diff options
Diffstat (limited to 'drivers/gpu/drm/amd/amdgpu')
295 files changed, 47405 insertions, 10140 deletions
diff --git a/drivers/gpu/drm/amd/amdgpu/Kconfig b/drivers/gpu/drm/amd/amdgpu/Kconfig index 74a8105fd2c0..7777d55275de 100644 --- a/drivers/gpu/drm/amd/amdgpu/Kconfig +++ b/drivers/gpu/drm/amd/amdgpu/Kconfig @@ -4,7 +4,7 @@ config DRM_AMDGPU_SI depends on DRM_AMDGPU help Choose this option if you want to enable experimental support - for SI asics. + for SI (Southern Islands) asics. SI is already supported in radeon. Experimental support for SI in amdgpu will be disabled by default and is still provided by @@ -16,7 +16,8 @@ config DRM_AMDGPU_CIK bool "Enable amdgpu support for CIK parts" depends on DRM_AMDGPU help - Choose this option if you want to enable support for CIK asics. + Choose this option if you want to enable support for CIK (Sea + Islands) asics. CIK is already supported in radeon. Support for CIK in amdgpu will be disabled by default and is still provided by radeon. diff --git a/drivers/gpu/drm/amd/amdgpu/Makefile b/drivers/gpu/drm/amd/amdgpu/Makefile index 7fedbb725e17..6ad39cf71bdd 100644 --- a/drivers/gpu/drm/amd/amdgpu/Makefile +++ b/drivers/gpu/drm/amd/amdgpu/Makefile @@ -46,19 +46,19 @@ amdgpu-y += amdgpu_device.o amdgpu_kms.o \ atom.o amdgpu_fence.o amdgpu_ttm.o amdgpu_object.o amdgpu_gart.o \ amdgpu_encoders.o amdgpu_display.o amdgpu_i2c.o \ amdgpu_gem.o amdgpu_ring.o \ - amdgpu_cs.o amdgpu_bios.o amdgpu_benchmark.o amdgpu_test.o \ + amdgpu_cs.o amdgpu_bios.o amdgpu_benchmark.o \ atombios_dp.o amdgpu_afmt.o amdgpu_trace_points.o \ atombios_encoders.o amdgpu_sa.o atombios_i2c.o \ - amdgpu_dma_buf.o amdgpu_vm.o amdgpu_ib.o amdgpu_pll.o \ + amdgpu_dma_buf.o amdgpu_vm.o amdgpu_vm_pt.o amdgpu_ib.o amdgpu_pll.o \ amdgpu_ucode.o amdgpu_bo_list.o amdgpu_ctx.o amdgpu_sync.o \ amdgpu_gtt_mgr.o amdgpu_preempt_mgr.o amdgpu_vram_mgr.o amdgpu_virt.o \ amdgpu_atomfirmware.o amdgpu_vf_error.o amdgpu_sched.o \ - amdgpu_debugfs.o amdgpu_ids.o amdgpu_gmc.o amdgpu_mmhub.o \ + amdgpu_debugfs.o amdgpu_ids.o amdgpu_gmc.o \ amdgpu_xgmi.o amdgpu_csa.o amdgpu_ras.o amdgpu_vm_cpu.o \ amdgpu_vm_sdma.o amdgpu_discovery.o amdgpu_ras_eeprom.o amdgpu_nbio.o \ amdgpu_umc.o smu_v11_0_i2c.o amdgpu_fru_eeprom.o amdgpu_rap.o \ - amdgpu_fw_attestation.o amdgpu_securedisplay.o amdgpu_hdp.o \ - amdgpu_eeprom.o amdgpu_mca.o + amdgpu_fw_attestation.o amdgpu_securedisplay.o \ + amdgpu_eeprom.o amdgpu_mca.o amdgpu_psp_ta.o amdgpu_lsdma.o amdgpu-$(CONFIG_PROC_FS) += amdgpu_fdinfo.o @@ -74,7 +74,8 @@ amdgpu-$(CONFIG_DRM_AMDGPU_SI)+= si.o gmc_v6_0.o gfx_v6_0.o si_ih.o si_dma.o dce amdgpu-y += \ vi.o mxgpu_vi.o nbio_v6_1.o soc15.o emu_soc.o mxgpu_ai.o nbio_v7_0.o vega10_reg_init.o \ vega20_reg_init.o nbio_v7_4.o nbio_v2_3.o nv.o arct_reg_init.o mxgpu_nv.o \ - nbio_v7_2.o hdp_v4_0.o hdp_v5_0.o aldebaran_reg_init.o aldebaran.o + nbio_v7_2.o hdp_v4_0.o hdp_v5_0.o aldebaran_reg_init.o aldebaran.o soc21.o \ + sienna_cichlid.o nbio_v4_3.o hdp_v6_0.o nbio_v7_7.o hdp_v5_2.o lsdma_v6_0.o # add DF block amdgpu-y += \ @@ -87,11 +88,12 @@ amdgpu-y += \ gmc_v8_0.o \ gfxhub_v1_0.o mmhub_v1_0.o gmc_v9_0.o gfxhub_v1_1.o mmhub_v9_4.o \ gfxhub_v2_0.o mmhub_v2_0.o gmc_v10_0.o gfxhub_v2_1.o mmhub_v2_3.o \ - mmhub_v1_7.o + mmhub_v1_7.o gfxhub_v3_0.o mmhub_v3_0.o mmhub_v3_0_2.o gmc_v11_0.o \ + mmhub_v3_0_1.o gfxhub_v3_0_3.o # add UMC block amdgpu-y += \ - umc_v6_0.o umc_v6_1.o umc_v6_7.o umc_v8_7.o + umc_v6_0.o umc_v6_1.o umc_v6_7.o umc_v8_7.o umc_v8_10.o # add IH block amdgpu-y += \ @@ -102,7 +104,8 @@ amdgpu-y += \ cz_ih.o \ vega10_ih.o \ vega20_ih.o \ - navi10_ih.o + navi10_ih.o \ + ih_v6_0.o # add PSP block amdgpu-y += \ @@ -112,7 +115,8 @@ amdgpu-y += \ psp_v11_0.o \ psp_v11_0_8.o \ psp_v12_0.o \ - psp_v13_0.o + psp_v13_0.o \ + psp_v13_0_4.o # add DCE block amdgpu-y += \ @@ -128,7 +132,10 @@ amdgpu-y += \ gfx_v9_0.o \ gfx_v9_4.o \ gfx_v9_4_2.o \ - gfx_v10_0.o + gfx_v10_0.o \ + imu_v11_0.o \ + gfx_v11_0.o \ + imu_v11_0_3.o # add async DMA block amdgpu-y += \ @@ -138,11 +145,14 @@ amdgpu-y += \ sdma_v4_0.o \ sdma_v4_4.o \ sdma_v5_0.o \ - sdma_v5_2.o + sdma_v5_2.o \ + sdma_v6_0.o # add MES block amdgpu-y += \ - mes_v10_1.o + amdgpu_mes.o \ + mes_v10_1.o \ + mes_v11_0.o # add UVD block amdgpu-y += \ @@ -160,28 +170,33 @@ amdgpu-y += \ # add VCN and JPEG block amdgpu-y += \ amdgpu_vcn.o \ + vcn_sw_ring.o \ vcn_v1_0.o \ vcn_v2_0.o \ vcn_v2_5.o \ vcn_v3_0.o \ + vcn_v4_0.o \ amdgpu_jpeg.o \ jpeg_v1_0.o \ jpeg_v2_0.o \ jpeg_v2_5.o \ - jpeg_v3_0.o + jpeg_v3_0.o \ + jpeg_v4_0.o # add ATHUB block amdgpu-y += \ athub_v1_0.o \ athub_v2_0.o \ - athub_v2_1.o + athub_v2_1.o \ + athub_v3_0.o # add SMUIO block amdgpu-y += \ smuio_v9_0.o \ smuio_v11_0.o \ smuio_v11_0_6.o \ - smuio_v13_0.o + smuio_v13_0.o \ + smuio_v13_0_6.o # add reset block amdgpu-y += \ @@ -207,7 +222,8 @@ amdgpu-y += \ amdgpu_amdkfd_arcturus.o \ amdgpu_amdkfd_aldebaran.o \ amdgpu_amdkfd_gfx_v10.o \ - amdgpu_amdkfd_gfx_v10_3.o + amdgpu_amdkfd_gfx_v10_3.o \ + amdgpu_amdkfd_gfx_v11.o ifneq ($(CONFIG_DRM_AMDGPU_CIK),) amdgpu-y += amdgpu_amdkfd_gfx_v7.o diff --git a/drivers/gpu/drm/amd/amdgpu/ObjectID.h b/drivers/gpu/drm/amd/amdgpu/ObjectID.h index 5b393622f592..a0f0a17e224f 100644 --- a/drivers/gpu/drm/amd/amdgpu/ObjectID.h +++ b/drivers/gpu/drm/amd/amdgpu/ObjectID.h @@ -119,6 +119,7 @@ #define CONNECTOR_OBJECT_ID_eDP 0x14 #define CONNECTOR_OBJECT_ID_MXM 0x15 #define CONNECTOR_OBJECT_ID_LVDS_eDP 0x16 +#define CONNECTOR_OBJECT_ID_USBC 0x17 /* deleted */ diff --git a/drivers/gpu/drm/amd/amdgpu/aldebaran.c b/drivers/gpu/drm/amd/amdgpu/aldebaran.c index bcfdb63b1d42..2b97b8a96fb4 100644 --- a/drivers/gpu/drm/amd/amdgpu/aldebaran.c +++ b/drivers/gpu/drm/amd/amdgpu/aldebaran.c @@ -31,6 +31,17 @@ #include "amdgpu_psp.h" #include "amdgpu_xgmi.h" +static bool aldebaran_is_mode2_default(struct amdgpu_reset_control *reset_ctl) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)reset_ctl->handle; + + if ((adev->ip_versions[MP1_HWIP][0] == IP_VERSION(13, 0, 2) && + adev->gmc.xgmi.connected_to_cpu)) + return true; + + return false; +} + static struct amdgpu_reset_handler * aldebaran_get_reset_handler(struct amdgpu_reset_control *reset_ctl, struct amdgpu_reset_context *reset_context) @@ -48,7 +59,7 @@ aldebaran_get_reset_handler(struct amdgpu_reset_control *reset_ctl, } } - if (adev->gmc.xgmi.connected_to_cpu) { + if (aldebaran_is_mode2_default(reset_ctl)) { list_for_each_entry(handler, &reset_ctl->reset_handlers, handler_list) { if (handler->reset_method == AMD_RESET_METHOD_MODE2) { @@ -136,18 +147,23 @@ static int aldebaran_mode2_perform_reset(struct amdgpu_reset_control *reset_ctl, struct amdgpu_reset_context *reset_context) { - struct amdgpu_device *tmp_adev = NULL; struct amdgpu_device *adev = (struct amdgpu_device *)reset_ctl->handle; + struct list_head *reset_device_list = reset_context->reset_device_list; + struct amdgpu_device *tmp_adev = NULL; int r = 0; dev_dbg(adev->dev, "aldebaran perform hw reset\n"); - if (reset_context->hive == NULL) { + + if (reset_device_list == NULL) + return -EINVAL; + + if (adev->ip_versions[MP1_HWIP][0] == IP_VERSION(13, 0, 2) && + reset_context->hive == NULL) { /* Wrong context, return error */ return -EINVAL; } - list_for_each_entry(tmp_adev, &reset_context->hive->device_list, - gmc.xgmi.head) { + list_for_each_entry(tmp_adev, reset_device_list, reset_list) { mutex_lock(&tmp_adev->reset_cntl->reset_lock); tmp_adev->reset_cntl->active_reset = AMD_RESET_METHOD_MODE2; } @@ -155,8 +171,7 @@ aldebaran_mode2_perform_reset(struct amdgpu_reset_control *reset_ctl, * Mode2 reset doesn't need any sync between nodes in XGMI hive, instead launch * them together so that they can be completed asynchronously on multiple nodes */ - list_for_each_entry(tmp_adev, &reset_context->hive->device_list, - gmc.xgmi.head) { + list_for_each_entry(tmp_adev, reset_device_list, reset_list) { /* For XGMI run all resets in parallel to speed up the process */ if (tmp_adev->gmc.xgmi.num_physical_nodes > 1) { if (!queue_work(system_unbound_wq, @@ -174,9 +189,7 @@ aldebaran_mode2_perform_reset(struct amdgpu_reset_control *reset_ctl, /* For XGMI wait for all resets to complete before proceed */ if (!r) { - list_for_each_entry(tmp_adev, - &reset_context->hive->device_list, - gmc.xgmi.head) { + list_for_each_entry(tmp_adev, reset_device_list, reset_list) { if (tmp_adev->gmc.xgmi.num_physical_nodes > 1) { flush_work(&tmp_adev->reset_cntl->reset_work); r = tmp_adev->asic_reset_res; @@ -186,8 +199,7 @@ aldebaran_mode2_perform_reset(struct amdgpu_reset_control *reset_ctl, } } - list_for_each_entry(tmp_adev, &reset_context->hive->device_list, - gmc.xgmi.head) { + list_for_each_entry(tmp_adev, reset_device_list, reset_list) { mutex_unlock(&tmp_adev->reset_cntl->reset_lock); tmp_adev->reset_cntl->active_reset = AMD_RESET_METHOD_NONE; } @@ -260,7 +272,7 @@ static int aldebaran_mode2_restore_ip(struct amdgpu_device *adev) adev->gfx.rlc.funcs->resume(adev); /* Wait for FW reset event complete */ - r = smu_wait_for_event(adev, SMU_EVENT_RESET_COMPLETE, 0); + r = amdgpu_dpm_wait_for_event(adev, SMU_EVENT_RESET_COMPLETE, 0); if (r) { dev_err(adev->dev, "Failed to get response from firmware after reset\n"); @@ -319,16 +331,21 @@ static int aldebaran_mode2_restore_hwcontext(struct amdgpu_reset_control *reset_ctl, struct amdgpu_reset_context *reset_context) { - int r; + struct list_head *reset_device_list = reset_context->reset_device_list; struct amdgpu_device *tmp_adev = NULL; + int r; + + if (reset_device_list == NULL) + return -EINVAL; - if (reset_context->hive == NULL) { + if (reset_context->reset_req_dev->ip_versions[MP1_HWIP][0] == + IP_VERSION(13, 0, 2) && + reset_context->hive == NULL) { /* Wrong context, return error */ return -EINVAL; } - list_for_each_entry(tmp_adev, &reset_context->hive->device_list, - gmc.xgmi.head) { + list_for_each_entry(tmp_adev, reset_device_list, reset_list) { dev_info(tmp_adev->dev, "GPU reset succeeded, trying to resume\n"); r = aldebaran_mode2_restore_ip(tmp_adev); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h b/drivers/gpu/drm/amd/amdgpu/amdgpu.h index d8b854fcbffa..8639a4f9c6e8 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu.h @@ -60,7 +60,6 @@ #include <drm/amdgpu_drm.h> #include <drm/drm_gem.h> #include <drm/drm_ioctl.h> -#include <drm/gpu_scheduler.h> #include <kgd_kfd_interface.h> #include "dm_pp_interface.h" @@ -87,11 +86,13 @@ #include "amdgpu_gmc.h" #include "amdgpu_gfx.h" #include "amdgpu_sdma.h" +#include "amdgpu_lsdma.h" #include "amdgpu_nbio.h" #include "amdgpu_hdp.h" #include "amdgpu_dm.h" #include "amdgpu_virt.h" #include "amdgpu_csa.h" +#include "amdgpu_mes_ctx.h" #include "amdgpu_gart.h" #include "amdgpu_debugfs.h" #include "amdgpu_job.h" @@ -99,7 +100,6 @@ #include "amdgpu_gem.h" #include "amdgpu_doorbell.h" #include "amdgpu_amdkfd.h" -#include "amdgpu_smu.h" #include "amdgpu_discovery.h" #include "amdgpu_mes.h" #include "amdgpu_umc.h" @@ -109,6 +109,7 @@ #include "amdgpu_smuio.h" #include "amdgpu_fdinfo.h" #include "amdgpu_mca.h" +#include "amdgpu_ras.h" #define MAX_GPU_INSTANCE 16 @@ -155,8 +156,6 @@ extern int amdgpu_vis_vram_limit; extern int amdgpu_gart_size; extern int amdgpu_gtt_size; extern int amdgpu_moverate; -extern int amdgpu_benchmarking; -extern int amdgpu_testing; extern int amdgpu_audio; extern int amdgpu_disp_priority; extern int amdgpu_hw_i2c; @@ -182,7 +181,7 @@ extern int amdgpu_sched_jobs; extern int amdgpu_sched_hw_submission; extern uint amdgpu_pcie_gen_cap; extern uint amdgpu_pcie_lane_cap; -extern uint amdgpu_cg_mask; +extern u64 amdgpu_cg_mask; extern uint amdgpu_pg_mask; extern uint amdgpu_sdma_phase_quantum; extern char *amdgpu_disable_cu; @@ -197,8 +196,8 @@ extern int amdgpu_emu_mode; extern uint amdgpu_smu_memory_pool_size; extern int amdgpu_smu_pptable_id; extern uint amdgpu_dc_feature_mask; -extern uint amdgpu_freesync_vid_mode; extern uint amdgpu_dc_debug_mask; +extern uint amdgpu_dc_visual_confirm; extern uint amdgpu_dm_abm_level; extern int amdgpu_backlight; extern struct amdgpu_mgpu_info mgpu_info; @@ -211,9 +210,11 @@ extern int amdgpu_async_gfx_ring; extern int amdgpu_mcbp; extern int amdgpu_discovery; extern int amdgpu_mes; +extern int amdgpu_mes_kiq; extern int amdgpu_noretry; extern int amdgpu_force_asic_type; extern int amdgpu_smartshift_bias; +extern int amdgpu_use_xgmi_p2p; #ifdef CONFIG_HSA_AMD extern int sched_policy; extern bool debug_evictions; @@ -223,6 +224,9 @@ static const int __maybe_unused sched_policy = KFD_SCHED_POLICY_HWS; static const bool __maybe_unused debug_evictions; /* = false */ static const bool __maybe_unused no_system_mem_limit; #endif +#ifdef CONFIG_HSA_AMD_P2P +extern bool pcie_p2p; +#endif extern int amdgpu_tmz; extern int amdgpu_reset_method; @@ -235,6 +239,9 @@ extern int amdgpu_cik_support; #endif extern int amdgpu_num_kcq; +#define AMDGPU_VCNFW_LOG_SIZE (32 * 1024) +extern int amdgpu_vcnfw_log; + #define AMDGPU_VM_MAX_NUM_CTX 4096 #define AMDGPU_SG_THRESHOLD (256*1024*1024) #define AMDGPU_DEFAULT_GTT_SIZE_MB 3072ULL /* 3GB by default */ @@ -271,14 +278,11 @@ extern int amdgpu_num_kcq; #define CIK_CURSOR_WIDTH 128 #define CIK_CURSOR_HEIGHT 128 -/* smasrt shift bias level limits */ +/* smart shift bias level limits */ #define AMDGPU_SMARTSHIFT_MAX_BIAS (100) #define AMDGPU_SMARTSHIFT_MIN_BIAS (-100) struct amdgpu_device; -struct amdgpu_ib; -struct amdgpu_cs_parser; -struct amdgpu_job; struct amdgpu_irq_src; struct amdgpu_fpriv; struct amdgpu_bo_va_mapping; @@ -313,7 +317,7 @@ enum amdgpu_kiq_irq { AMDGPU_CP_KIQ_IRQ_DRIVER0 = 0, AMDGPU_CP_KIQ_IRQ_LAST }; - +#define SRIOV_USEC_TIMEOUT 1200000 /* wait 12 * 100ms for SRIOV */ #define MAX_KIQ_REG_WAIT 5000 /* in usecs, 5ms */ #define MAX_KIQ_REG_BAILOUT_INTERVAL 5 /* in msecs, 5ms */ #define MAX_KIQ_REG_TRY 1000 @@ -325,7 +329,7 @@ int amdgpu_device_ip_set_powergating_state(void *dev, enum amd_ip_block_type block_type, enum amd_powergating_state state); void amdgpu_device_ip_get_clockgating_state(struct amdgpu_device *adev, - u32 *flags); + u64 *flags); int amdgpu_device_ip_wait_for_idle(struct amdgpu_device *adev, enum amd_ip_block_type block_type); bool amdgpu_device_ip_is_idle(struct amdgpu_device *adev, @@ -373,7 +377,8 @@ int amdgpu_device_ip_block_add(struct amdgpu_device *adev, */ bool amdgpu_get_bios(struct amdgpu_device *adev); bool amdgpu_read_bios(struct amdgpu_device *adev); - +bool amdgpu_soc15_read_bios_from_rom(struct amdgpu_device *adev, + u8 *bios, u32 length_bytes); /* * Clocks */ @@ -466,20 +471,6 @@ struct amdgpu_flip_work { /* - * CP & rings. - */ - -struct amdgpu_ib { - struct amdgpu_sa_bo *sa_bo; - uint32_t length_dw; - uint64_t gpu_addr; - uint32_t *ptr; - uint32_t flags; -}; - -extern const struct drm_sched_backend_ops amdgpu_sched_ops; - -/* * file private structure */ @@ -494,79 +485,6 @@ struct amdgpu_fpriv { int amdgpu_file_to_fpriv(struct file *filp, struct amdgpu_fpriv **fpriv); -int amdgpu_ib_get(struct amdgpu_device *adev, struct amdgpu_vm *vm, - unsigned size, - enum amdgpu_ib_pool_type pool, - struct amdgpu_ib *ib); -void amdgpu_ib_free(struct amdgpu_device *adev, struct amdgpu_ib *ib, - struct dma_fence *f); -int amdgpu_ib_schedule(struct amdgpu_ring *ring, unsigned num_ibs, - struct amdgpu_ib *ibs, struct amdgpu_job *job, - struct dma_fence **f); -int amdgpu_ib_pool_init(struct amdgpu_device *adev); -void amdgpu_ib_pool_fini(struct amdgpu_device *adev); -int amdgpu_ib_ring_tests(struct amdgpu_device *adev); - -/* - * CS. - */ -struct amdgpu_cs_chunk { - uint32_t chunk_id; - uint32_t length_dw; - void *kdata; -}; - -struct amdgpu_cs_post_dep { - struct drm_syncobj *syncobj; - struct dma_fence_chain *chain; - u64 point; -}; - -struct amdgpu_cs_parser { - struct amdgpu_device *adev; - struct drm_file *filp; - struct amdgpu_ctx *ctx; - - /* chunks */ - unsigned nchunks; - struct amdgpu_cs_chunk *chunks; - - /* scheduler job object */ - struct amdgpu_job *job; - struct drm_sched_entity *entity; - - /* buffer objects */ - struct ww_acquire_ctx ticket; - struct amdgpu_bo_list *bo_list; - struct amdgpu_mn *mn; - struct amdgpu_bo_list_entry vm_pd; - struct list_head validated; - struct dma_fence *fence; - uint64_t bytes_moved_threshold; - uint64_t bytes_moved_vis_threshold; - uint64_t bytes_moved; - uint64_t bytes_moved_vis; - - /* user fence */ - struct amdgpu_bo_list_entry uf_entry; - - unsigned num_post_deps; - struct amdgpu_cs_post_dep *post_deps; -}; - -static inline u32 amdgpu_get_ib_value(struct amdgpu_cs_parser *p, - uint32_t ib_idx, int idx) -{ - return p->job->ibs[ib_idx].ptr[idx]; -} - -static inline void amdgpu_set_ib_value(struct amdgpu_cs_parser *p, - uint32_t ib_idx, int idx, - uint32_t value) -{ - p->job->ibs[ib_idx].ptr[idx] = value; -} - /* * Writeback */ @@ -586,13 +504,7 @@ void amdgpu_device_wb_free(struct amdgpu_device *adev, u32 wb); /* * Benchmarking */ -void amdgpu_benchmark(struct amdgpu_device *adev, int test_number); - - -/* - * Testing - */ -void amdgpu_test_moves(struct amdgpu_device *adev); +int amdgpu_benchmark(struct amdgpu_device *adev, int test_number); /* * ASIC specific register table accessible by UMD @@ -736,6 +648,7 @@ enum amd_hw_ip_block_type { SDMA5_HWIP, SDMA6_HWIP, SDMA7_HWIP, + LSDMA_HWIP, MMHUB_HWIP, ATHUB_HWIP, NBIO_HWIP, @@ -758,19 +671,25 @@ enum amd_hw_ip_block_type { RSMU_HWIP, XGMI_HWIP, DCI_HWIP, + PCIE_HWIP, MAX_HWIP }; -#define HWIP_MAX_INSTANCE 10 +#define HWIP_MAX_INSTANCE 11 #define HW_ID_MAX 300 #define IP_VERSION(mj, mn, rv) (((mj) << 16) | ((mn) << 8) | (rv)) +#define IP_VERSION_MAJ(ver) ((ver) >> 16) +#define IP_VERSION_MIN(ver) (((ver) >> 8) & 0xFF) +#define IP_VERSION_REV(ver) ((ver) & 0xFF) struct amd_powerplay { void *pp_handle; const struct amd_pm_funcs *pp_funcs; }; +struct ip_discovery_top; + /* polaris10 kickers */ #define ASICID_IS_P20(did, rid) (((did == 0x67DF) && \ ((rid == 0xE3) || \ @@ -810,9 +729,31 @@ struct amd_powerplay { (rid == 0x01) || \ (rid == 0x10)))) +struct amdgpu_mqd_prop { + uint64_t mqd_gpu_addr; + uint64_t hqd_base_gpu_addr; + uint64_t rptr_gpu_addr; + uint64_t wptr_gpu_addr; + uint32_t queue_size; + bool use_doorbell; + uint32_t doorbell_index; + uint64_t eop_gpu_addr; + uint32_t hqd_pipe_priority; + uint32_t hqd_queue_priority; + bool hqd_active; +}; + +struct amdgpu_mqd { + unsigned mqd_size; + int (*init_mqd)(struct amdgpu_device *adev, void *mqd, + struct amdgpu_mqd_prop *p); +}; + #define AMDGPU_RESET_MAGIC_NUM 64 #define AMDGPU_MAX_DF_PERFMONS 4 #define AMDGPU_PRODUCT_NAME_LEN 64 +struct amdgpu_reset_domain; + struct amdgpu_device { struct device *dev; struct pci_dev *pdev; @@ -941,6 +882,7 @@ struct amdgpu_device { u64 fence_context; unsigned num_rings; struct amdgpu_ring *rings[AMDGPU_MAX_RINGS]; + struct dma_fence __rcu *gang_submit; bool ib_pool_ready; struct amdgpu_sa_manager ib_pools[AMDGPU_IB_POOL_MAX]; struct amdgpu_sched gpu_sched[AMDGPU_HW_IP_NUM][AMDGPU_RING_PRIO_MAX]; @@ -950,14 +892,8 @@ struct amdgpu_device { /* powerplay */ struct amd_powerplay powerplay; - bool pp_force_state_enabled; - - /* smu */ - struct smu_context smu; - - /* dpm */ struct amdgpu_pm pm; - u32 cg_flags; + u64 cg_flags; u32 pg_flags; /* nbio */ @@ -981,6 +917,9 @@ struct amdgpu_device { /* sdma */ struct amdgpu_sdma sdma; + /* lsdma */ + struct amdgpu_lsdma lsdma; + /* uvd */ struct amdgpu_uvd uvd; @@ -1013,7 +952,9 @@ struct amdgpu_device { /* mes */ bool enable_mes; + bool enable_mes_kiq; struct amdgpu_mes mes; + struct amdgpu_mqd mqds[AMDGPU_HW_IP_NUM]; /* df */ struct amdgpu_df df; @@ -1054,9 +995,7 @@ struct amdgpu_device { bool in_s4; bool in_s0ix; - atomic_t in_gpu_reset; enum pp_mp1_state mp1_state; - struct rw_semaphore reset_sem; struct amdgpu_doorbell_index doorbell_index; struct mutex notifier_lock; @@ -1074,16 +1013,15 @@ struct amdgpu_device { uint64_t df_perfmon_config_assign_mask[AMDGPU_MAX_DF_PERFMONS]; /* enable runtime pm on the device */ - bool runpm; bool in_runpm; bool has_pr3; - bool is_fw_fb; bool pm_sysfs_en; bool ucode_sysfs_en; + bool psp_sysfs_en; /* Chip product information */ - char product_number[16]; + char product_number[20]; char product_name[AMDGPU_PRODUCT_NAME_LEN]; char serial[20]; @@ -1100,6 +1038,31 @@ struct amdgpu_device { uint32_t ip_versions[MAX_HWIP][HWIP_MAX_INSTANCE]; bool ram_is_direct_mapped; + + struct list_head ras_list; + + struct ip_discovery_top *ip_top; + + struct amdgpu_reset_domain *reset_domain; + + struct mutex benchmark_mutex; + + /* reset dump register */ + uint32_t *reset_dump_reg_list; + uint32_t *reset_dump_reg_value; + int num_regs; +#ifdef CONFIG_DEV_COREDUMP + struct amdgpu_task_info reset_task_info; + bool reset_vram_lost; + struct timespec64 reset_time; +#endif + + bool scpm_enabled; + uint32_t scpm_status; + + struct work_struct reset_work; + + bool job_hang; }; static inline struct amdgpu_device *drm_to_adev(struct drm_device *ddev) @@ -1272,7 +1235,8 @@ int emu_soc_asic_init(struct amdgpu_device *adev); #define amdgpu_asic_flush_hdp(adev, r) \ ((adev)->asic_funcs->flush_hdp ? (adev)->asic_funcs->flush_hdp((adev), (r)) : (adev)->hdp.funcs->flush_hdp((adev), (r))) #define amdgpu_asic_invalidate_hdp(adev, r) \ - ((adev)->asic_funcs->invalidate_hdp ? (adev)->asic_funcs->invalidate_hdp((adev), (r)) : (adev)->hdp.funcs->invalidate_hdp((adev), (r))) + ((adev)->asic_funcs->invalidate_hdp ? (adev)->asic_funcs->invalidate_hdp((adev), (r)) : \ + ((adev)->hdp.funcs->invalidate_hdp ? (adev)->hdp.funcs->invalidate_hdp((adev), (r)) : 0)) #define amdgpu_asic_need_full_reset(adev) (adev)->asic_funcs->need_full_reset((adev)) #define amdgpu_asic_init_doorbell_index(adev) (adev)->asic_funcs->init_doorbell_index((adev)) #define amdgpu_asic_get_pcie_usage(adev, cnt0, cnt1) ((adev)->asic_funcs->get_pcie_usage((adev), (cnt0), (cnt1))) @@ -1292,10 +1256,12 @@ int emu_soc_asic_init(struct amdgpu_device *adev); bool amdgpu_device_has_job_running(struct amdgpu_device *adev); bool amdgpu_device_should_recover_gpu(struct amdgpu_device *adev); int amdgpu_device_gpu_recover(struct amdgpu_device *adev, - struct amdgpu_job* job); + struct amdgpu_job *job, + struct amdgpu_reset_context *reset_context); void amdgpu_device_pci_config_reset(struct amdgpu_device *adev); int amdgpu_device_pci_reset(struct amdgpu_device *adev); bool amdgpu_device_need_post(struct amdgpu_device *adev); +bool amdgpu_device_should_use_aspm(struct amdgpu_device *adev); void amdgpu_cs_report_moved_bytes(struct amdgpu_device *adev, u64 num_bytes, u64 num_vis_bytes); @@ -1321,6 +1287,12 @@ void amdgpu_device_invalidate_hdp(struct amdgpu_device *adev, struct amdgpu_ring *ring); void amdgpu_device_halt(struct amdgpu_device *adev); +u32 amdgpu_device_pcie_port_rreg(struct amdgpu_device *adev, + u32 reg); +void amdgpu_device_pcie_port_wreg(struct amdgpu_device *adev, + u32 reg, u32 v); +struct dma_fence *amdgpu_device_switch_gang(struct amdgpu_device *adev, + struct dma_fence *gang); /* atpx handler */ #if defined(CONFIG_VGA_SWITCHEROO) @@ -1408,12 +1380,10 @@ int amdgpu_acpi_smart_shift_update(struct drm_device *dev, enum amdgpu_ss ss_sta int amdgpu_acpi_pcie_notify_device_ready(struct amdgpu_device *adev); void amdgpu_acpi_get_backlight_caps(struct amdgpu_dm_backlight_caps *caps); -bool amdgpu_acpi_is_s0ix_active(struct amdgpu_device *adev); void amdgpu_acpi_detect(void); #else static inline int amdgpu_acpi_init(struct amdgpu_device *adev) { return 0; } static inline void amdgpu_acpi_fini(struct amdgpu_device *adev) { } -static inline bool amdgpu_acpi_is_s0ix_active(struct amdgpu_device *adev) { return false; } static inline void amdgpu_acpi_detect(void) { } static inline bool amdgpu_acpi_is_power_shift_control_supported(void) { return false; } static inline int amdgpu_acpi_power_shift_control(struct amdgpu_device *adev, @@ -1422,9 +1392,15 @@ static inline int amdgpu_acpi_smart_shift_update(struct drm_device *dev, enum amdgpu_ss ss_state) { return 0; } #endif -int amdgpu_cs_find_mapping(struct amdgpu_cs_parser *parser, - uint64_t addr, struct amdgpu_bo **bo, - struct amdgpu_bo_va_mapping **mapping); +#if defined(CONFIG_ACPI) && defined(CONFIG_SUSPEND) +bool amdgpu_acpi_is_s3_active(struct amdgpu_device *adev); +bool amdgpu_acpi_should_gpu_reset(struct amdgpu_device *adev); +bool amdgpu_acpi_is_s0ix_active(struct amdgpu_device *adev); +#else +static inline bool amdgpu_acpi_is_s0ix_active(struct amdgpu_device *adev) { return false; } +static inline bool amdgpu_acpi_should_gpu_reset(struct amdgpu_device *adev) { return false; } +static inline bool amdgpu_acpi_is_s3_active(struct amdgpu_device *adev) { return false; } +#endif #if defined(CONFIG_DRM_AMD_DC) int amdgpu_dm_display_resume(struct amdgpu_device *adev ); @@ -1452,6 +1428,15 @@ int amdgpu_device_set_cg_state(struct amdgpu_device *adev, int amdgpu_device_set_pg_state(struct amdgpu_device *adev, enum amd_powergating_state state); +static inline bool amdgpu_device_has_timeouts_enabled(struct amdgpu_device *adev) +{ + return amdgpu_gpu_recovery != 0 && + adev->gfx_timeout != MAX_SCHEDULE_TIMEOUT && + adev->compute_timeout != MAX_SCHEDULE_TIMEOUT && + adev->sdma_timeout != MAX_SCHEDULE_TIMEOUT && + adev->video_timeout != MAX_SCHEDULE_TIMEOUT; +} + #include "amdgpu_object.h" static inline bool amdgpu_is_tmz(struct amdgpu_device *adev) @@ -1459,8 +1444,6 @@ static inline bool amdgpu_is_tmz(struct amdgpu_device *adev) return adev->gmc.tmz_enabled; } -static inline int amdgpu_in_reset(struct amdgpu_device *adev) -{ - return atomic_read(&adev->in_gpu_reset); -} +int amdgpu_in_reset(struct amdgpu_device *adev); + #endif diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_acp.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_acp.c index cc9c9f8b23b2..6d72355ac492 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_acp.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_acp.c @@ -29,6 +29,8 @@ #include <linux/platform_device.h> #include <sound/designware_i2s.h> #include <sound/pcm.h> +#include <linux/acpi.h> +#include <linux/dmi.h> #include "amdgpu.h" #include "atom.h" @@ -36,6 +38,7 @@ #include "acp_gfx_if.h" +#define ST_JADEITE 1 #define ACP_TILE_ON_MASK 0x03 #define ACP_TILE_OFF_MASK 0x02 #define ACP_TILE_ON_RETAIN_REG_MASK 0x1f @@ -85,6 +88,8 @@ #define ACP_DEVS 4 #define ACP_SRC_ID 162 +static unsigned long acp_machine_id; + enum { ACP_TILE_P1 = 0, ACP_TILE_P2, @@ -128,16 +133,14 @@ static int acp_poweroff(struct generic_pm_domain *genpd) struct amdgpu_device *adev; apd = container_of(genpd, struct acp_pm_domain, gpd); - if (apd != NULL) { - adev = apd->adev; + adev = apd->adev; /* call smu to POWER GATE ACP block * smu will * 1. turn off the acp clock * 2. power off the acp tiles * 3. check and enter ulv state */ - amdgpu_dpm_set_powergating_by_smu(adev, AMD_IP_BLOCK_TYPE_ACP, true); - } + amdgpu_dpm_set_powergating_by_smu(adev, AMD_IP_BLOCK_TYPE_ACP, true); return 0; } @@ -147,16 +150,14 @@ static int acp_poweron(struct generic_pm_domain *genpd) struct amdgpu_device *adev; apd = container_of(genpd, struct acp_pm_domain, gpd); - if (apd != NULL) { - adev = apd->adev; + adev = apd->adev; /* call smu to UNGATE ACP block * smu will * 1. exit ulv * 2. turn on acp clock * 3. power on acp tiles */ - amdgpu_dpm_set_powergating_by_smu(adev, AMD_IP_BLOCK_TYPE_ACP, false); - } + amdgpu_dpm_set_powergating_by_smu(adev, AMD_IP_BLOCK_TYPE_ACP, false); return 0; } @@ -184,6 +185,37 @@ static int acp_genpd_remove_device(struct device *dev, void *data) return 0; } +static int acp_quirk_cb(const struct dmi_system_id *id) +{ + acp_machine_id = ST_JADEITE; + return 1; +} + +static const struct dmi_system_id acp_quirk_table[] = { + { + .callback = acp_quirk_cb, + .matches = { + DMI_EXACT_MATCH(DMI_BOARD_VENDOR, "AMD"), + DMI_EXACT_MATCH(DMI_PRODUCT_NAME, "Jadeite"), + } + }, + { + .callback = acp_quirk_cb, + .matches = { + DMI_EXACT_MATCH(DMI_BOARD_VENDOR, "IP3 Technology CO.,Ltd."), + DMI_EXACT_MATCH(DMI_PRODUCT_NAME, "ASN1D"), + }, + }, + { + .callback = acp_quirk_cb, + .matches = { + DMI_EXACT_MATCH(DMI_BOARD_VENDOR, "Standard"), + DMI_EXACT_MATCH(DMI_PRODUCT_NAME, "ASN10"), + }, + }, + {} +}; + /** * acp_hw_init - start and test ACP block * @@ -193,7 +225,7 @@ static int acp_genpd_remove_device(struct device *dev, void *data) static int acp_hw_init(void *handle) { int r; - uint64_t acp_base; + u64 acp_base; u32 val = 0; u32 count = 0; struct i2s_platform_data *i2s_pdata = NULL; @@ -220,141 +252,202 @@ static int acp_hw_init(void *handle) return -EINVAL; acp_base = adev->rmmio_base; - - adev->acp.acp_genpd = kzalloc(sizeof(struct acp_pm_domain), GFP_KERNEL); - if (adev->acp.acp_genpd == NULL) + if (!adev->acp.acp_genpd) return -ENOMEM; adev->acp.acp_genpd->gpd.name = "ACP_AUDIO"; adev->acp.acp_genpd->gpd.power_off = acp_poweroff; adev->acp.acp_genpd->gpd.power_on = acp_poweron; - - adev->acp.acp_genpd->adev = adev; pm_genpd_init(&adev->acp.acp_genpd->gpd, NULL, false); + dmi_check_system(acp_quirk_table); + switch (acp_machine_id) { + case ST_JADEITE: + { + adev->acp.acp_cell = kcalloc(2, sizeof(struct mfd_cell), + GFP_KERNEL); + if (!adev->acp.acp_cell) { + r = -ENOMEM; + goto failure; + } - adev->acp.acp_cell = kcalloc(ACP_DEVS, sizeof(struct mfd_cell), - GFP_KERNEL); - - if (adev->acp.acp_cell == NULL) { - r = -ENOMEM; - goto failure; - } - - adev->acp.acp_res = kcalloc(5, sizeof(struct resource), GFP_KERNEL); - if (adev->acp.acp_res == NULL) { - r = -ENOMEM; - goto failure; - } + adev->acp.acp_res = kcalloc(3, sizeof(struct resource), GFP_KERNEL); + if (!adev->acp.acp_res) { + r = -ENOMEM; + goto failure; + } - i2s_pdata = kcalloc(3, sizeof(struct i2s_platform_data), GFP_KERNEL); - if (i2s_pdata == NULL) { - r = -ENOMEM; - goto failure; - } + i2s_pdata = kcalloc(1, sizeof(struct i2s_platform_data), GFP_KERNEL); + if (!i2s_pdata) { + r = -ENOMEM; + goto failure; + } - switch (adev->asic_type) { - case CHIP_STONEY: i2s_pdata[0].quirks = DW_I2S_QUIRK_COMP_REG_OFFSET | - DW_I2S_QUIRK_16BIT_IDX_OVERRIDE; + DW_I2S_QUIRK_16BIT_IDX_OVERRIDE; + i2s_pdata[0].cap = DWC_I2S_PLAY | DWC_I2S_RECORD; + i2s_pdata[0].snd_rates = SNDRV_PCM_RATE_8000_96000; + i2s_pdata[0].i2s_reg_comp1 = ACP_I2S_COMP1_CAP_REG_OFFSET; + i2s_pdata[0].i2s_reg_comp2 = ACP_I2S_COMP2_CAP_REG_OFFSET; + + adev->acp.acp_res[0].name = "acp2x_dma"; + adev->acp.acp_res[0].flags = IORESOURCE_MEM; + adev->acp.acp_res[0].start = acp_base; + adev->acp.acp_res[0].end = acp_base + ACP_DMA_REGS_END; + + adev->acp.acp_res[1].name = "acp2x_dw_i2s_play_cap"; + adev->acp.acp_res[1].flags = IORESOURCE_MEM; + adev->acp.acp_res[1].start = acp_base + ACP_I2S_CAP_REGS_START; + adev->acp.acp_res[1].end = acp_base + ACP_I2S_CAP_REGS_END; + + adev->acp.acp_res[2].name = "acp2x_dma_irq"; + adev->acp.acp_res[2].flags = IORESOURCE_IRQ; + adev->acp.acp_res[2].start = amdgpu_irq_create_mapping(adev, 162); + adev->acp.acp_res[2].end = adev->acp.acp_res[2].start; + + adev->acp.acp_cell[0].name = "acp_audio_dma"; + adev->acp.acp_cell[0].num_resources = 3; + adev->acp.acp_cell[0].resources = &adev->acp.acp_res[0]; + adev->acp.acp_cell[0].platform_data = &adev->asic_type; + adev->acp.acp_cell[0].pdata_size = sizeof(adev->asic_type); + + adev->acp.acp_cell[1].name = "designware-i2s"; + adev->acp.acp_cell[1].num_resources = 1; + adev->acp.acp_cell[1].resources = &adev->acp.acp_res[1]; + adev->acp.acp_cell[1].platform_data = &i2s_pdata[0]; + adev->acp.acp_cell[1].pdata_size = sizeof(struct i2s_platform_data); + r = mfd_add_hotplug_devices(adev->acp.parent, adev->acp.acp_cell, 2); + if (r) + goto failure; + r = device_for_each_child(adev->acp.parent, &adev->acp.acp_genpd->gpd, + acp_genpd_add_device); + if (r) + goto failure; break; - default: - i2s_pdata[0].quirks = DW_I2S_QUIRK_COMP_REG_OFFSET; } - i2s_pdata[0].cap = DWC_I2S_PLAY; - i2s_pdata[0].snd_rates = SNDRV_PCM_RATE_8000_96000; - i2s_pdata[0].i2s_reg_comp1 = ACP_I2S_COMP1_PLAY_REG_OFFSET; - i2s_pdata[0].i2s_reg_comp2 = ACP_I2S_COMP2_PLAY_REG_OFFSET; - switch (adev->asic_type) { - case CHIP_STONEY: - i2s_pdata[1].quirks = DW_I2S_QUIRK_COMP_REG_OFFSET | - DW_I2S_QUIRK_COMP_PARAM1 | - DW_I2S_QUIRK_16BIT_IDX_OVERRIDE; - break; default: - i2s_pdata[1].quirks = DW_I2S_QUIRK_COMP_REG_OFFSET | - DW_I2S_QUIRK_COMP_PARAM1; - } + adev->acp.acp_cell = kcalloc(ACP_DEVS, sizeof(struct mfd_cell), + GFP_KERNEL); - i2s_pdata[1].cap = DWC_I2S_RECORD; - i2s_pdata[1].snd_rates = SNDRV_PCM_RATE_8000_96000; - i2s_pdata[1].i2s_reg_comp1 = ACP_I2S_COMP1_CAP_REG_OFFSET; - i2s_pdata[1].i2s_reg_comp2 = ACP_I2S_COMP2_CAP_REG_OFFSET; + if (!adev->acp.acp_cell) { + r = -ENOMEM; + goto failure; + } - i2s_pdata[2].quirks = DW_I2S_QUIRK_COMP_REG_OFFSET; - switch (adev->asic_type) { - case CHIP_STONEY: - i2s_pdata[2].quirks |= DW_I2S_QUIRK_16BIT_IDX_OVERRIDE; - break; - default: - break; - } + adev->acp.acp_res = kcalloc(5, sizeof(struct resource), GFP_KERNEL); + if (!adev->acp.acp_res) { + r = -ENOMEM; + goto failure; + } + + i2s_pdata = kcalloc(3, sizeof(struct i2s_platform_data), GFP_KERNEL); + if (!i2s_pdata) { + r = -ENOMEM; + goto failure; + } + + switch (adev->asic_type) { + case CHIP_STONEY: + i2s_pdata[0].quirks = DW_I2S_QUIRK_COMP_REG_OFFSET | + DW_I2S_QUIRK_16BIT_IDX_OVERRIDE; + break; + default: + i2s_pdata[0].quirks = DW_I2S_QUIRK_COMP_REG_OFFSET; + } + i2s_pdata[0].cap = DWC_I2S_PLAY; + i2s_pdata[0].snd_rates = SNDRV_PCM_RATE_8000_96000; + i2s_pdata[0].i2s_reg_comp1 = ACP_I2S_COMP1_PLAY_REG_OFFSET; + i2s_pdata[0].i2s_reg_comp2 = ACP_I2S_COMP2_PLAY_REG_OFFSET; + switch (adev->asic_type) { + case CHIP_STONEY: + i2s_pdata[1].quirks = DW_I2S_QUIRK_COMP_REG_OFFSET | + DW_I2S_QUIRK_COMP_PARAM1 | + DW_I2S_QUIRK_16BIT_IDX_OVERRIDE; + break; + default: + i2s_pdata[1].quirks = DW_I2S_QUIRK_COMP_REG_OFFSET | + DW_I2S_QUIRK_COMP_PARAM1; + } + + i2s_pdata[1].cap = DWC_I2S_RECORD; + i2s_pdata[1].snd_rates = SNDRV_PCM_RATE_8000_96000; + i2s_pdata[1].i2s_reg_comp1 = ACP_I2S_COMP1_CAP_REG_OFFSET; + i2s_pdata[1].i2s_reg_comp2 = ACP_I2S_COMP2_CAP_REG_OFFSET; + + i2s_pdata[2].quirks = DW_I2S_QUIRK_COMP_REG_OFFSET; + switch (adev->asic_type) { + case CHIP_STONEY: + i2s_pdata[2].quirks |= DW_I2S_QUIRK_16BIT_IDX_OVERRIDE; + break; + default: + break; + } - i2s_pdata[2].cap = DWC_I2S_PLAY | DWC_I2S_RECORD; - i2s_pdata[2].snd_rates = SNDRV_PCM_RATE_8000_96000; - i2s_pdata[2].i2s_reg_comp1 = ACP_BT_COMP1_REG_OFFSET; - i2s_pdata[2].i2s_reg_comp2 = ACP_BT_COMP2_REG_OFFSET; - - adev->acp.acp_res[0].name = "acp2x_dma"; - adev->acp.acp_res[0].flags = IORESOURCE_MEM; - adev->acp.acp_res[0].start = acp_base; - adev->acp.acp_res[0].end = acp_base + ACP_DMA_REGS_END; - - adev->acp.acp_res[1].name = "acp2x_dw_i2s_play"; - adev->acp.acp_res[1].flags = IORESOURCE_MEM; - adev->acp.acp_res[1].start = acp_base + ACP_I2S_PLAY_REGS_START; - adev->acp.acp_res[1].end = acp_base + ACP_I2S_PLAY_REGS_END; - - adev->acp.acp_res[2].name = "acp2x_dw_i2s_cap"; - adev->acp.acp_res[2].flags = IORESOURCE_MEM; - adev->acp.acp_res[2].start = acp_base + ACP_I2S_CAP_REGS_START; - adev->acp.acp_res[2].end = acp_base + ACP_I2S_CAP_REGS_END; - - adev->acp.acp_res[3].name = "acp2x_dw_bt_i2s_play_cap"; - adev->acp.acp_res[3].flags = IORESOURCE_MEM; - adev->acp.acp_res[3].start = acp_base + ACP_BT_PLAY_REGS_START; - adev->acp.acp_res[3].end = acp_base + ACP_BT_PLAY_REGS_END; - - adev->acp.acp_res[4].name = "acp2x_dma_irq"; - adev->acp.acp_res[4].flags = IORESOURCE_IRQ; - adev->acp.acp_res[4].start = amdgpu_irq_create_mapping(adev, 162); - adev->acp.acp_res[4].end = adev->acp.acp_res[4].start; - - adev->acp.acp_cell[0].name = "acp_audio_dma"; - adev->acp.acp_cell[0].num_resources = 5; - adev->acp.acp_cell[0].resources = &adev->acp.acp_res[0]; - adev->acp.acp_cell[0].platform_data = &adev->asic_type; - adev->acp.acp_cell[0].pdata_size = sizeof(adev->asic_type); - - adev->acp.acp_cell[1].name = "designware-i2s"; - adev->acp.acp_cell[1].num_resources = 1; - adev->acp.acp_cell[1].resources = &adev->acp.acp_res[1]; - adev->acp.acp_cell[1].platform_data = &i2s_pdata[0]; - adev->acp.acp_cell[1].pdata_size = sizeof(struct i2s_platform_data); - - adev->acp.acp_cell[2].name = "designware-i2s"; - adev->acp.acp_cell[2].num_resources = 1; - adev->acp.acp_cell[2].resources = &adev->acp.acp_res[2]; - adev->acp.acp_cell[2].platform_data = &i2s_pdata[1]; - adev->acp.acp_cell[2].pdata_size = sizeof(struct i2s_platform_data); - - adev->acp.acp_cell[3].name = "designware-i2s"; - adev->acp.acp_cell[3].num_resources = 1; - adev->acp.acp_cell[3].resources = &adev->acp.acp_res[3]; - adev->acp.acp_cell[3].platform_data = &i2s_pdata[2]; - adev->acp.acp_cell[3].pdata_size = sizeof(struct i2s_platform_data); - - r = mfd_add_hotplug_devices(adev->acp.parent, adev->acp.acp_cell, - ACP_DEVS); - if (r) - goto failure; - - r = device_for_each_child(adev->acp.parent, &adev->acp.acp_genpd->gpd, - acp_genpd_add_device); - if (r) - goto failure; + i2s_pdata[2].cap = DWC_I2S_PLAY | DWC_I2S_RECORD; + i2s_pdata[2].snd_rates = SNDRV_PCM_RATE_8000_96000; + i2s_pdata[2].i2s_reg_comp1 = ACP_BT_COMP1_REG_OFFSET; + i2s_pdata[2].i2s_reg_comp2 = ACP_BT_COMP2_REG_OFFSET; + + adev->acp.acp_res[0].name = "acp2x_dma"; + adev->acp.acp_res[0].flags = IORESOURCE_MEM; + adev->acp.acp_res[0].start = acp_base; + adev->acp.acp_res[0].end = acp_base + ACP_DMA_REGS_END; + + adev->acp.acp_res[1].name = "acp2x_dw_i2s_play"; + adev->acp.acp_res[1].flags = IORESOURCE_MEM; + adev->acp.acp_res[1].start = acp_base + ACP_I2S_PLAY_REGS_START; + adev->acp.acp_res[1].end = acp_base + ACP_I2S_PLAY_REGS_END; + + adev->acp.acp_res[2].name = "acp2x_dw_i2s_cap"; + adev->acp.acp_res[2].flags = IORESOURCE_MEM; + adev->acp.acp_res[2].start = acp_base + ACP_I2S_CAP_REGS_START; + adev->acp.acp_res[2].end = acp_base + ACP_I2S_CAP_REGS_END; + + adev->acp.acp_res[3].name = "acp2x_dw_bt_i2s_play_cap"; + adev->acp.acp_res[3].flags = IORESOURCE_MEM; + adev->acp.acp_res[3].start = acp_base + ACP_BT_PLAY_REGS_START; + adev->acp.acp_res[3].end = acp_base + ACP_BT_PLAY_REGS_END; + + adev->acp.acp_res[4].name = "acp2x_dma_irq"; + adev->acp.acp_res[4].flags = IORESOURCE_IRQ; + adev->acp.acp_res[4].start = amdgpu_irq_create_mapping(adev, 162); + adev->acp.acp_res[4].end = adev->acp.acp_res[4].start; + + adev->acp.acp_cell[0].name = "acp_audio_dma"; + adev->acp.acp_cell[0].num_resources = 5; + adev->acp.acp_cell[0].resources = &adev->acp.acp_res[0]; + adev->acp.acp_cell[0].platform_data = &adev->asic_type; + adev->acp.acp_cell[0].pdata_size = sizeof(adev->asic_type); + + adev->acp.acp_cell[1].name = "designware-i2s"; + adev->acp.acp_cell[1].num_resources = 1; + adev->acp.acp_cell[1].resources = &adev->acp.acp_res[1]; + adev->acp.acp_cell[1].platform_data = &i2s_pdata[0]; + adev->acp.acp_cell[1].pdata_size = sizeof(struct i2s_platform_data); + + adev->acp.acp_cell[2].name = "designware-i2s"; + adev->acp.acp_cell[2].num_resources = 1; + adev->acp.acp_cell[2].resources = &adev->acp.acp_res[2]; + adev->acp.acp_cell[2].platform_data = &i2s_pdata[1]; + adev->acp.acp_cell[2].pdata_size = sizeof(struct i2s_platform_data); + + adev->acp.acp_cell[3].name = "designware-i2s"; + adev->acp.acp_cell[3].num_resources = 1; + adev->acp.acp_cell[3].resources = &adev->acp.acp_res[3]; + adev->acp.acp_cell[3].platform_data = &i2s_pdata[2]; + adev->acp.acp_cell[3].pdata_size = sizeof(struct i2s_platform_data); + + r = mfd_add_hotplug_devices(adev->acp.parent, adev->acp.acp_cell, ACP_DEVS); + if (r) + goto failure; + + r = device_for_each_child(adev->acp.parent, &adev->acp.acp_genpd->gpd, + acp_genpd_add_device); + if (r) + goto failure; + } /* Assert Soft reset of ACP */ val = cgs_read_register(adev->acp.cgs_device, mmACP_SOFT_RESET); @@ -546,8 +639,7 @@ static const struct amd_ip_funcs acp_ip_funcs = { .set_powergating_state = acp_set_powergating_state, }; -const struct amdgpu_ip_block_version acp_ip_block = -{ +const struct amdgpu_ip_block_version acp_ip_block = { .type = AMD_IP_BLOCK_TYPE_ACP, .major = 2, .minor = 2, diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_acpi.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_acpi.c index 4811b0faafd9..b14800ac179e 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_acpi.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_acpi.c @@ -1,3 +1,4 @@ +// SPDX-License-Identifier: MIT /* * Copyright 2012 Advanced Micro Devices, Inc. * @@ -66,9 +67,7 @@ struct amdgpu_atif { struct amdgpu_atif_notifications notifications; struct amdgpu_atif_functions functions; struct amdgpu_atif_notification_cfg notification_cfg; -#if defined(CONFIG_BACKLIGHT_CLASS_DEVICE) || defined(CONFIG_BACKLIGHT_CLASS_DEVICE_MODULE) struct backlight_device *bd; -#endif struct amdgpu_dm_backlight_caps backlight_caps; }; @@ -436,7 +435,6 @@ static int amdgpu_atif_handler(struct amdgpu_device *adev, DRM_DEBUG_DRIVER("ATIF: %d pending SBIOS requests\n", count); if (req.pending & ATIF_PANEL_BRIGHTNESS_CHANGE_REQUEST) { -#if defined(CONFIG_BACKLIGHT_CLASS_DEVICE) || defined(CONFIG_BACKLIGHT_CLASS_DEVICE_MODULE) if (atif->bd) { DRM_DEBUG_DRIVER("Changing brightness to %d\n", req.backlight_level); @@ -447,7 +445,6 @@ static int amdgpu_atif_handler(struct amdgpu_device *adev, */ backlight_device_set_brightness(atif->bd, req.backlight_level); } -#endif } if (req.pending & ATIF_DGPU_DISPLAY_EVENT) { @@ -849,11 +846,11 @@ int amdgpu_acpi_init(struct amdgpu_device *adev) { struct amdgpu_atif *atif = &amdgpu_acpi_priv.atif; -#if defined(CONFIG_BACKLIGHT_CLASS_DEVICE) || defined(CONFIG_BACKLIGHT_CLASS_DEVICE_MODULE) if (atif->notifications.brightness_change) { if (amdgpu_device_has_dc_support(adev)) { #if defined(CONFIG_DRM_AMD_DC) struct amdgpu_display_manager *dm = &adev->dm; + if (dm->backlight_dev[0]) atif->bd = dm->backlight_dev[0]; #endif @@ -868,6 +865,7 @@ int amdgpu_acpi_init(struct amdgpu_device *adev) if ((enc->devices & (ATOM_DEVICE_LCD_SUPPORT)) && enc->enc_priv) { struct amdgpu_encoder_atom_dig *dig = enc->enc_priv; + if (dig->bl_dev) { atif->bd = dig->bl_dev; break; @@ -876,7 +874,6 @@ int amdgpu_acpi_init(struct amdgpu_device *adev) } } } -#endif adev->acpi_nb.notifier_call = amdgpu_acpi_event; register_acpi_notifier(&adev->acpi_nb); @@ -925,9 +922,9 @@ static bool amdgpu_atif_pci_probe_handle(struct pci_dev *pdev) return false; status = acpi_get_handle(dhandle, "ATIF", &atif_handle); - if (ACPI_FAILURE(status)) { + if (ACPI_FAILURE(status)) return false; - } + amdgpu_acpi_priv.atif.handle = atif_handle; acpi_get_name(amdgpu_acpi_priv.atif.handle, ACPI_FULL_PATHNAME, &buffer); DRM_DEBUG_DRIVER("Found ATIF handle %s\n", acpi_method_name); @@ -960,9 +957,9 @@ static bool amdgpu_atcs_pci_probe_handle(struct pci_dev *pdev) return false; status = acpi_get_handle(dhandle, "ATCS", &atcs_handle); - if (ACPI_FAILURE(status)) { + if (ACPI_FAILURE(status)) return false; - } + amdgpu_acpi_priv.atcs.handle = atcs_handle; acpi_get_name(amdgpu_acpi_priv.atcs.handle, ACPI_FULL_PATHNAME, &buffer); DRM_DEBUG_DRIVER("Found ATCS handle %s\n", acpi_method_name); @@ -1031,6 +1028,38 @@ void amdgpu_acpi_detect(void) } } +#if IS_ENABLED(CONFIG_SUSPEND) +/** + * amdgpu_acpi_is_s3_active + * + * @adev: amdgpu_device_pointer + * + * returns true if supported, false if not. + */ +bool amdgpu_acpi_is_s3_active(struct amdgpu_device *adev) +{ + return !(adev->flags & AMD_IS_APU) || + (pm_suspend_target_state == PM_SUSPEND_MEM); +} + +/** + * amdgpu_acpi_should_gpu_reset + * + * @adev: amdgpu_device_pointer + * + * returns true if should reset GPU, false if not + */ +bool amdgpu_acpi_should_gpu_reset(struct amdgpu_device *adev) +{ + if (adev->flags & AMD_IS_APU) + return false; + + if (amdgpu_sriov_vf(adev)) + return false; + + return pm_suspend_target_state != PM_SUSPEND_TO_IDLE; +} + /** * amdgpu_acpi_is_s0ix_active * @@ -1040,11 +1069,30 @@ void amdgpu_acpi_detect(void) */ bool amdgpu_acpi_is_s0ix_active(struct amdgpu_device *adev) { -#if IS_ENABLED(CONFIG_AMD_PMC) && IS_ENABLED(CONFIG_SUSPEND) - if (acpi_gbl_FADT.flags & ACPI_FADT_LOW_POWER_S0) { - if (adev->flags & AMD_IS_APU) - return pm_suspend_target_state == PM_SUSPEND_TO_IDLE; + if (!(adev->flags & AMD_IS_APU) || + (pm_suspend_target_state != PM_SUSPEND_TO_IDLE)) + return false; + + /* + * If ACPI_FADT_LOW_POWER_S0 is not set in the FADT, it is generally + * risky to do any special firmware-related preparations for entering + * S0ix even though the system is suspending to idle, so return false + * in that case. + */ + if (!(acpi_gbl_FADT.flags & ACPI_FADT_LOW_POWER_S0)) { + dev_warn_once(adev->dev, + "Power consumption will be higher as BIOS has not been configured for suspend-to-idle.\n" + "To use suspend-to-idle change the sleep mode in BIOS setup.\n"); + return false; } -#endif + +#if !IS_ENABLED(CONFIG_AMD_PMC) + dev_warn_once(adev->dev, + "Power consumption will be higher as the kernel has not been compiled with CONFIG_AMD_PMC.\n"); return false; +#else + return true; +#endif /* CONFIG_AMD_PMC */ } + +#endif /* CONFIG_SUSPEND */ diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c index 776a947b45df..5d9a34601a1a 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c @@ -1,3 +1,4 @@ +// SPDX-License-Identifier: MIT /* * Copyright 2014 Advanced Micro Devices, Inc. * @@ -33,6 +34,7 @@ #include <uapi/linux/kfd_ioctl.h> #include "amdgpu_ras.h" #include "amdgpu_umc.h" +#include "amdgpu_reset.h" /* Total memory size in system memory and all GPU VRAM. Used to * estimate worst case amount of memory to reserve for page tables @@ -73,9 +75,6 @@ void amdgpu_amdkfd_device_probe(struct amdgpu_device *adev) return; adev->kfd.dev = kgd2kfd_probe(adev, vf); - - if (adev->kfd.dev) - amdgpu_amdkfd_total_mem_size += adev->gmc.real_vram_size; } /** @@ -100,7 +99,18 @@ static void amdgpu_doorbell_get_kfd_info(struct amdgpu_device *adev, * The first num_doorbells are used by amdgpu. * amdkfd takes whatever's left in the aperture. */ - if (adev->doorbell.size > adev->doorbell.num_doorbells * sizeof(u32)) { + if (adev->enable_mes) { + /* + * With MES enabled, we only need to initialize + * the base address. The size and offset are + * not initialized as AMDGPU manages the whole + * doorbell space. + */ + *aperture_base = adev->doorbell.base; + *aperture_size = 0; + *start_offset = 0; + } else if (adev->doorbell.size > adev->doorbell.num_doorbells * + sizeof(u32)) { *aperture_base = adev->doorbell.base; *aperture_size = adev->doorbell.size; *start_offset = adev->doorbell.num_doorbells * sizeof(u32); @@ -111,6 +121,23 @@ static void amdgpu_doorbell_get_kfd_info(struct amdgpu_device *adev, } } + +static void amdgpu_amdkfd_reset_work(struct work_struct *work) +{ + struct amdgpu_device *adev = container_of(work, struct amdgpu_device, + kfd.reset_work); + + struct amdgpu_reset_context reset_context; + + memset(&reset_context, 0, sizeof(reset_context)); + + reset_context.method = AMD_RESET_METHOD_NONE; + reset_context.reset_req_dev = adev; + clear_bit(AMDGPU_NEED_FULL_RESET, &reset_context.flags); + + amdgpu_device_gpu_recover(adev, NULL, &reset_context); +} + void amdgpu_amdkfd_device_init(struct amdgpu_device *adev) { int i; @@ -128,7 +155,7 @@ void amdgpu_amdkfd_device_init(struct amdgpu_device *adev) AMDGPU_GMC_HOLE_START), .drm_render_minor = adev_to_drm(adev)->render->index, .sdma_doorbell_idx = adev->doorbell_index.sdma_engine, - + .enable_mes = adev->enable_mes, }; /* this is going to have a few of the MSBs set that we need to @@ -169,6 +196,10 @@ void amdgpu_amdkfd_device_init(struct amdgpu_device *adev) adev->kfd.init_complete = kgd2kfd_device_init(adev->kfd.dev, adev_to_drm(adev), &gpu_resources); + + amdgpu_amdkfd_total_mem_size += adev->gmc.real_vram_size; + + INIT_WORK(&adev->kfd.reset_work, amdgpu_amdkfd_reset_work); } } @@ -177,6 +208,7 @@ void amdgpu_amdkfd_device_fini_sw(struct amdgpu_device *adev) if (adev->kfd.dev) { kgd2kfd_device_exit(adev->kfd.dev); adev->kfd.dev = NULL; + amdgpu_amdkfd_total_mem_size -= adev->gmc.real_vram_size; } } @@ -236,7 +268,8 @@ int amdgpu_amdkfd_post_reset(struct amdgpu_device *adev) void amdgpu_amdkfd_gpu_reset(struct amdgpu_device *adev) { if (amdgpu_device_should_recover_gpu(adev)) - amdgpu_device_gpu_recover(adev, NULL); + amdgpu_reset_domain_schedule(adev->reset_domain, + &adev->kfd.reset_work); } int amdgpu_amdkfd_alloc_gtt_mem(struct amdgpu_device *adev, size_t size, @@ -514,13 +547,6 @@ out_put: return r; } -uint64_t amdgpu_amdkfd_get_vram_usage(struct amdgpu_device *adev) -{ - struct ttm_resource_manager *vram_man = ttm_manager_type(&adev->mman.bdev, TTM_PL_VRAM); - - return amdgpu_vram_mgr_usage(vram_man); -} - uint8_t amdgpu_amdkfd_get_xgmi_hops_count(struct amdgpu_device *dst, struct amdgpu_device *src) { @@ -659,6 +685,7 @@ int amdgpu_amdkfd_submit_ib(struct amdgpu_device *adev, ib->length_dw = ib_len; /* This works for NO_HWS. TODO: need to handle without knowing VMID */ job->vmid = vmid; + job->num_ibs = 1; ret = amdgpu_ib_schedule(ring, 1, ib, job, &f); @@ -667,6 +694,8 @@ int amdgpu_amdkfd_submit_ib(struct amdgpu_device *adev, goto err_ib_sched; } + /* Drop the initial kref_init count (see drm_sched_main as example) */ + dma_fence_put(f); ret = dma_fence_wait(f, false); err_ib_sched: @@ -677,6 +706,13 @@ err: void amdgpu_amdkfd_set_compute_idle(struct amdgpu_device *adev, bool idle) { + /* Temporary workaround to fix issues observed in some + * compute applications when GFXOFF is enabled on GFX11. + */ + if (IP_VERSION_MAJ(adev->ip_versions[GC_HWIP][0]) == 11) { + pr_debug("GFXOFF is %s\n", idle ? "enabled" : "disabled"); + amdgpu_gfx_off_ctrl(adev, idle); + } amdgpu_dpm_switch_power_profile(adev, PP_SMC_POWER_PROFILE_COMPUTE, !idle); @@ -710,7 +746,8 @@ int amdgpu_amdkfd_flush_gpu_tlb_pasid(struct amdgpu_device *adev, { bool all_hub = false; - if (adev->family == AMDGPU_FAMILY_AI) + if (adev->family == AMDGPU_FAMILY_AI || + adev->family == AMDGPU_FAMILY_RV) all_hub = true; return amdgpu_gmc_flush_gpu_tlb_pasid(adev, pasid, flush_type, all_hub); @@ -725,9 +762,13 @@ void amdgpu_amdkfd_ras_poison_consumption_handler(struct amdgpu_device *adev, bo { struct ras_err_data err_data = {0, 0, 0, NULL}; - /* CPU MCA will handle page retirement if connected_to_cpu is 1 */ - if (!adev->gmc.xgmi.connected_to_cpu) - amdgpu_umc_poison_handler(adev, &err_data, reset); - else if (reset) - amdgpu_amdkfd_gpu_reset(adev); + amdgpu_umc_poison_handler(adev, &err_data, reset); +} + +bool amdgpu_amdkfd_ras_query_utcl2_poison_status(struct amdgpu_device *adev) +{ + if (adev->gfx.ras && adev->gfx.ras->query_utcl2_poison_status) + return adev->gfx.ras->query_utcl2_poison_status(adev); + else + return false; } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h index 61f899e54fd5..647220a8762d 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h @@ -48,6 +48,7 @@ enum kfd_mem_attachment_type { KFD_MEM_ATT_SHARED, /* Share kgd_mem->bo or another attachment's */ KFD_MEM_ATT_USERPTR, /* SG bo to DMA map pages from a userptr bo */ KFD_MEM_ATT_DMABUF, /* DMAbuf to DMA map TTM BOs */ + KFD_MEM_ATT_SG /* Tag to DMA map SG BOs */ }; struct kfd_mem_attachment { @@ -95,7 +96,9 @@ struct amdgpu_amdkfd_fence { struct amdgpu_kfd_dev { struct kfd_dev *dev; uint64_t vram_used; + uint64_t vram_used_aligned; bool init_complete; + struct work_struct reset_work; }; enum kgd_engine_type { @@ -131,6 +134,7 @@ struct amdkfd_process_info { atomic_t evicted_bos; struct delayed_work restore_userptr_work; struct pid *pid; + bool block_mmu_notifications; }; int amdgpu_amdkfd_init(void); @@ -169,6 +173,9 @@ int amdgpu_queue_mask_bit_to_set_resource_bit(struct amdgpu_device *adev, struct amdgpu_amdkfd_fence *amdgpu_amdkfd_fence_create(u64 context, struct mm_struct *mm, struct svm_range_bo *svm_bo); +#if defined(CONFIG_DEBUG_FS) +int kfd_debugfs_kfd_mem_limits(struct seq_file *m, void *data); +#endif #if IS_ENABLED(CONFIG_HSA_AMD) bool amdkfd_fence_check_mm(struct dma_fence *f, struct mm_struct *mm); struct amdgpu_amdkfd_fence *to_amdgpu_amdkfd_fence(struct dma_fence *f); @@ -223,7 +230,6 @@ int amdgpu_amdkfd_get_dmabuf_info(struct amdgpu_device *adev, int dma_buf_fd, uint64_t *bo_size, void *metadata_buffer, size_t buffer_size, uint32_t *metadata_size, uint32_t *flags); -uint64_t amdgpu_amdkfd_get_vram_usage(struct amdgpu_device *adev); uint8_t amdgpu_amdkfd_get_xgmi_hops_count(struct amdgpu_device *dst, struct amdgpu_device *src); int amdgpu_amdkfd_get_xgmi_bandwidth_mbytes(struct amdgpu_device *dst, @@ -266,24 +272,25 @@ int amdgpu_amdkfd_gpuvm_acquire_process_vm(struct amdgpu_device *adev, void amdgpu_amdkfd_gpuvm_release_process_vm(struct amdgpu_device *adev, void *drm_priv); uint64_t amdgpu_amdkfd_gpuvm_get_process_page_dir(void *drm_priv); +size_t amdgpu_amdkfd_get_available_memory(struct amdgpu_device *adev); int amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu( struct amdgpu_device *adev, uint64_t va, uint64_t size, void *drm_priv, struct kgd_mem **mem, - uint64_t *offset, uint32_t flags); + uint64_t *offset, uint32_t flags, bool criu_resume); int amdgpu_amdkfd_gpuvm_free_memory_of_gpu( struct amdgpu_device *adev, struct kgd_mem *mem, void *drm_priv, uint64_t *size); -int amdgpu_amdkfd_gpuvm_map_memory_to_gpu( - struct amdgpu_device *adev, struct kgd_mem *mem, void *drm_priv, - bool *table_freed); +int amdgpu_amdkfd_gpuvm_map_memory_to_gpu(struct amdgpu_device *adev, + struct kgd_mem *mem, void *drm_priv); int amdgpu_amdkfd_gpuvm_unmap_memory_from_gpu( struct amdgpu_device *adev, struct kgd_mem *mem, void *drm_priv); int amdgpu_amdkfd_gpuvm_sync_memory( struct amdgpu_device *adev, struct kgd_mem *mem, bool intr); -int amdgpu_amdkfd_gpuvm_map_gtt_bo_to_kernel(struct amdgpu_device *adev, - struct kgd_mem *mem, void **kptr, uint64_t *size); -void amdgpu_amdkfd_gpuvm_unmap_gtt_bo_from_kernel(struct amdgpu_device *adev, - struct kgd_mem *mem); +int amdgpu_amdkfd_gpuvm_map_gtt_bo_to_kernel(struct kgd_mem *mem, + void **kptr, uint64_t *size); +void amdgpu_amdkfd_gpuvm_unmap_gtt_bo_from_kernel(struct kgd_mem *mem); + +int amdgpu_amdkfd_map_gtt_bo_to_gart(struct amdgpu_device *adev, struct amdgpu_bo *bo); int amdgpu_amdkfd_gpuvm_restore_process_bos(void *process_info, struct dma_fence **ef); @@ -298,6 +305,15 @@ int amdgpu_amdkfd_get_tile_config(struct amdgpu_device *adev, struct tile_config *config); void amdgpu_amdkfd_ras_poison_consumption_handler(struct amdgpu_device *adev, bool reset); +bool amdgpu_amdkfd_bo_mapped_to_dev(struct amdgpu_device *adev, struct kgd_mem *mem); +void amdgpu_amdkfd_block_mmu_notifications(void *p); +int amdgpu_amdkfd_criu_resume(void *p); +bool amdgpu_amdkfd_ras_query_utcl2_poison_status(struct amdgpu_device *adev); +int amdgpu_amdkfd_reserve_mem_limit(struct amdgpu_device *adev, + uint64_t size, u32 alloc_flag); +void amdgpu_amdkfd_unreserve_mem_limit(struct amdgpu_device *adev, + uint64_t size, u32 alloc_flag); + #if IS_ENABLED(CONFIG_HSA_AMD) void amdgpu_amdkfd_gpuvm_init_mem_limits(void); void amdgpu_amdkfd_gpuvm_destroy_cb(struct amdgpu_device *adev, @@ -328,7 +344,7 @@ void amdgpu_amdkfd_release_notify(struct amdgpu_bo *bo) } #endif /* KGD2KFD callbacks */ -int kgd2kfd_quiesce_mm(struct mm_struct *mm); +int kgd2kfd_quiesce_mm(struct mm_struct *mm, uint32_t trigger); int kgd2kfd_resume_mm(struct mm_struct *mm); int kgd2kfd_schedule_evict_and_restore_process(struct mm_struct *mm, struct dma_fence *fence); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_aldebaran.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_aldebaran.c index 46cd4ee6bafb..c8935d718207 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_aldebaran.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_aldebaran.c @@ -37,10 +37,7 @@ const struct kfd2kgd_calls aldebaran_kfd2kgd = { .hqd_sdma_is_occupied = kgd_arcturus_hqd_sdma_is_occupied, .hqd_destroy = kgd_gfx_v9_hqd_destroy, .hqd_sdma_destroy = kgd_arcturus_hqd_sdma_destroy, - .address_watch_disable = kgd_gfx_v9_address_watch_disable, - .address_watch_execute = kgd_gfx_v9_address_watch_execute, .wave_control_execute = kgd_gfx_v9_wave_control_execute, - .address_watch_get_offset = kgd_gfx_v9_address_watch_get_offset, .get_atc_vmid_pasid_mapping_info = kgd_gfx_v9_get_atc_vmid_pasid_mapping_info, .set_vm_context_page_table_base = kgd_gfx_v9_set_vm_context_page_table_base, diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_arcturus.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_arcturus.c index abe93b3ff765..4191af5a3f13 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_arcturus.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_arcturus.c @@ -289,10 +289,7 @@ const struct kfd2kgd_calls arcturus_kfd2kgd = { .hqd_sdma_is_occupied = kgd_arcturus_hqd_sdma_is_occupied, .hqd_destroy = kgd_gfx_v9_hqd_destroy, .hqd_sdma_destroy = kgd_arcturus_hqd_sdma_destroy, - .address_watch_disable = kgd_gfx_v9_address_watch_disable, - .address_watch_execute = kgd_gfx_v9_address_watch_execute, .wave_control_execute = kgd_gfx_v9_wave_control_execute, - .address_watch_get_offset = kgd_gfx_v9_address_watch_get_offset, .get_atc_vmid_pasid_mapping_info = kgd_gfx_v9_get_atc_vmid_pasid_mapping_info, .set_vm_context_page_table_base = diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_fence.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_fence.c index 1d0dbff87d3f..469785d33791 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_fence.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_fence.c @@ -159,11 +159,14 @@ static void amdkfd_fence_release(struct dma_fence *f) } /** - * amdkfd_fence_check_mm - Check if @mm is same as that of the fence @f - * if same return TRUE else return FALSE. + * amdkfd_fence_check_mm - Check whether to prevent eviction of @f by @mm * * @f: [IN] fence * @mm: [IN] mm that needs to be verified + * + * Check if @mm is same as that of the fence @f, if same return TRUE else + * return FALSE. + * For svm bo, which support vram overcommitment, always return FALSE. */ bool amdkfd_fence_check_mm(struct dma_fence *f, struct mm_struct *mm) { @@ -171,7 +174,7 @@ bool amdkfd_fence_check_mm(struct dma_fence *f, struct mm_struct *mm) if (!fence) return false; - else if (fence->mm == mm) + else if (fence->mm == mm && !fence->svm_bo) return true; return false; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10.c index 7b7f4b2764c1..9378fc79e9ea 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10.c @@ -671,20 +671,6 @@ static bool get_atc_vmid_pasid_mapping_info(struct amdgpu_device *adev, return !!(value & ATC_VMID0_PASID_MAPPING__VALID_MASK); } -static int kgd_address_watch_disable(struct amdgpu_device *adev) -{ - return 0; -} - -static int kgd_address_watch_execute(struct amdgpu_device *adev, - unsigned int watch_point_id, - uint32_t cntl_val, - uint32_t addr_hi, - uint32_t addr_lo) -{ - return 0; -} - static int kgd_wave_control_execute(struct amdgpu_device *adev, uint32_t gfx_index_val, uint32_t sq_cmd) @@ -709,13 +695,6 @@ static int kgd_wave_control_execute(struct amdgpu_device *adev, return 0; } -static uint32_t kgd_address_watch_get_offset(struct amdgpu_device *adev, - unsigned int watch_point_id, - unsigned int reg_offset) -{ - return 0; -} - static void set_vm_context_page_table_base(struct amdgpu_device *adev, uint32_t vmid, uint64_t page_table_base) { @@ -767,10 +746,7 @@ const struct kfd2kgd_calls gfx_v10_kfd2kgd = { .hqd_sdma_is_occupied = kgd_hqd_sdma_is_occupied, .hqd_destroy = kgd_hqd_destroy, .hqd_sdma_destroy = kgd_hqd_sdma_destroy, - .address_watch_disable = kgd_address_watch_disable, - .address_watch_execute = kgd_address_watch_execute, .wave_control_execute = kgd_wave_control_execute, - .address_watch_get_offset = kgd_address_watch_get_offset, .get_atc_vmid_pasid_mapping_info = get_atc_vmid_pasid_mapping_info, .set_vm_context_page_table_base = set_vm_context_page_table_base, diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10_3.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10_3.c index 1f37d3574001..ba21ec6b35e0 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10_3.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10_3.c @@ -26,6 +26,8 @@ #include "gc/gc_10_3_0_sh_mask.h" #include "oss/osssys_5_0_0_offset.h" #include "oss/osssys_5_0_0_sh_mask.h" +#include "athub/athub_2_1_0_offset.h" +#include "athub/athub_2_1_0_sh_mask.h" #include "soc15_common.h" #include "v10_structs.h" #include "nv.h" @@ -582,21 +584,6 @@ static int hqd_sdma_destroy_v10_3(struct amdgpu_device *adev, void *mqd, return 0; } - -static int address_watch_disable_v10_3(struct amdgpu_device *adev) -{ - return 0; -} - -static int address_watch_execute_v10_3(struct amdgpu_device *adev, - unsigned int watch_point_id, - uint32_t cntl_val, - uint32_t addr_hi, - uint32_t addr_lo) -{ - return 0; -} - static int wave_control_execute_v10_3(struct amdgpu_device *adev, uint32_t gfx_index_val, uint32_t sq_cmd) @@ -621,11 +608,16 @@ static int wave_control_execute_v10_3(struct amdgpu_device *adev, return 0; } -static uint32_t address_watch_get_offset_v10_3(struct amdgpu_device *adev, - unsigned int watch_point_id, - unsigned int reg_offset) +static bool get_atc_vmid_pasid_mapping_info_v10_3(struct amdgpu_device *adev, + uint8_t vmid, uint16_t *p_pasid) { - return 0; + uint32_t value; + + value = RREG32(SOC15_REG_OFFSET(ATHUB, 0, mmATC_VMID0_PASID_MAPPING) + + vmid); + *p_pasid = value & ATC_VMID0_PASID_MAPPING__PASID_MASK; + + return !!(value & ATC_VMID0_PASID_MAPPING__VALID_MASK); } static void set_vm_context_page_table_base_v10_3(struct amdgpu_device *adev, @@ -809,11 +801,8 @@ const struct kfd2kgd_calls gfx_v10_3_kfd2kgd = { .hqd_sdma_is_occupied = hqd_sdma_is_occupied_v10_3, .hqd_destroy = hqd_destroy_v10_3, .hqd_sdma_destroy = hqd_sdma_destroy_v10_3, - .address_watch_disable = address_watch_disable_v10_3, - .address_watch_execute = address_watch_execute_v10_3, .wave_control_execute = wave_control_execute_v10_3, - .address_watch_get_offset = address_watch_get_offset_v10_3, - .get_atc_vmid_pasid_mapping_info = NULL, + .get_atc_vmid_pasid_mapping_info = get_atc_vmid_pasid_mapping_info_v10_3, .set_vm_context_page_table_base = set_vm_context_page_table_base_v10_3, .program_trap_handler_settings = program_trap_handler_settings_v10_3, #if 0 diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v11.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v11.c new file mode 100644 index 000000000000..7e80caa05060 --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v11.c @@ -0,0 +1,625 @@ +/* + * Copyright 2021 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + */ +#include <linux/mmu_context.h> +#include "amdgpu.h" +#include "amdgpu_amdkfd.h" +#include "gc/gc_11_0_0_offset.h" +#include "gc/gc_11_0_0_sh_mask.h" +#include "oss/osssys_6_0_0_offset.h" +#include "oss/osssys_6_0_0_sh_mask.h" +#include "soc15_common.h" +#include "soc15d.h" +#include "v11_structs.h" +#include "soc21.h" + +enum hqd_dequeue_request_type { + NO_ACTION = 0, + DRAIN_PIPE, + RESET_WAVES, + SAVE_WAVES +}; + +static void lock_srbm(struct amdgpu_device *adev, uint32_t mec, uint32_t pipe, + uint32_t queue, uint32_t vmid) +{ + mutex_lock(&adev->srbm_mutex); + soc21_grbm_select(adev, mec, pipe, queue, vmid); +} + +static void unlock_srbm(struct amdgpu_device *adev) +{ + soc21_grbm_select(adev, 0, 0, 0, 0); + mutex_unlock(&adev->srbm_mutex); +} + +static void acquire_queue(struct amdgpu_device *adev, uint32_t pipe_id, + uint32_t queue_id) +{ + uint32_t mec = (pipe_id / adev->gfx.mec.num_pipe_per_mec) + 1; + uint32_t pipe = (pipe_id % adev->gfx.mec.num_pipe_per_mec); + + lock_srbm(adev, mec, pipe, queue_id, 0); +} + +static uint64_t get_queue_mask(struct amdgpu_device *adev, + uint32_t pipe_id, uint32_t queue_id) +{ + unsigned int bit = pipe_id * adev->gfx.mec.num_queue_per_pipe + + queue_id; + + return 1ull << bit; +} + +static void release_queue(struct amdgpu_device *adev) +{ + unlock_srbm(adev); +} + +static void program_sh_mem_settings_v11(struct amdgpu_device *adev, uint32_t vmid, + uint32_t sh_mem_config, + uint32_t sh_mem_ape1_base, + uint32_t sh_mem_ape1_limit, + uint32_t sh_mem_bases) +{ + lock_srbm(adev, 0, 0, 0, vmid); + + WREG32(SOC15_REG_OFFSET(GC, 0, regSH_MEM_CONFIG), sh_mem_config); + WREG32(SOC15_REG_OFFSET(GC, 0, regSH_MEM_BASES), sh_mem_bases); + + unlock_srbm(adev); +} + +static int set_pasid_vmid_mapping_v11(struct amdgpu_device *adev, unsigned int pasid, + unsigned int vmid) +{ + uint32_t value = pasid << IH_VMID_0_LUT__PASID__SHIFT; + + /* Mapping vmid to pasid also for IH block */ + pr_debug("mapping vmid %d -> pasid %d in IH block for GFX client\n", + vmid, pasid); + WREG32(SOC15_REG_OFFSET(OSSSYS, 0, regIH_VMID_0_LUT) + vmid, value); + + return 0; +} + +static int init_interrupts_v11(struct amdgpu_device *adev, uint32_t pipe_id) +{ + uint32_t mec; + uint32_t pipe; + + mec = (pipe_id / adev->gfx.mec.num_pipe_per_mec) + 1; + pipe = (pipe_id % adev->gfx.mec.num_pipe_per_mec); + + lock_srbm(adev, mec, pipe, 0, 0); + + WREG32_SOC15(GC, 0, regCPC_INT_CNTL, + CP_INT_CNTL_RING0__TIME_STAMP_INT_ENABLE_MASK | + CP_INT_CNTL_RING0__OPCODE_ERROR_INT_ENABLE_MASK); + + unlock_srbm(adev); + + return 0; +} + +static uint32_t get_sdma_rlc_reg_offset(struct amdgpu_device *adev, + unsigned int engine_id, + unsigned int queue_id) +{ + uint32_t sdma_engine_reg_base = 0; + uint32_t sdma_rlc_reg_offset; + + switch (engine_id) { + case 0: + sdma_engine_reg_base = SOC15_REG_OFFSET(SDMA0, 0, + regSDMA0_QUEUE0_RB_CNTL) - regSDMA0_QUEUE0_RB_CNTL; + break; + case 1: + sdma_engine_reg_base = SOC15_REG_OFFSET(SDMA1, 0, + regSDMA1_QUEUE0_RB_CNTL) - regSDMA0_QUEUE0_RB_CNTL; + break; + default: + BUG(); + } + + sdma_rlc_reg_offset = sdma_engine_reg_base + + queue_id * (regSDMA0_QUEUE1_RB_CNTL - regSDMA0_QUEUE0_RB_CNTL); + + pr_debug("RLC register offset for SDMA%d RLC%d: 0x%x\n", engine_id, + queue_id, sdma_rlc_reg_offset); + + return sdma_rlc_reg_offset; +} + +static inline struct v11_compute_mqd *get_mqd(void *mqd) +{ + return (struct v11_compute_mqd *)mqd; +} + +static inline struct v11_sdma_mqd *get_sdma_mqd(void *mqd) +{ + return (struct v11_sdma_mqd *)mqd; +} + +static int hqd_load_v11(struct amdgpu_device *adev, void *mqd, uint32_t pipe_id, + uint32_t queue_id, uint32_t __user *wptr, + uint32_t wptr_shift, uint32_t wptr_mask, + struct mm_struct *mm) +{ + struct v11_compute_mqd *m; + uint32_t *mqd_hqd; + uint32_t reg, hqd_base, data; + + m = get_mqd(mqd); + + pr_debug("Load hqd of pipe %d queue %d\n", pipe_id, queue_id); + acquire_queue(adev, pipe_id, queue_id); + + /* HIQ is set during driver init period with vmid set to 0*/ + if (m->cp_hqd_vmid == 0) { + uint32_t value, mec, pipe; + + mec = (pipe_id / adev->gfx.mec.num_pipe_per_mec) + 1; + pipe = (pipe_id % adev->gfx.mec.num_pipe_per_mec); + + pr_debug("kfd: set HIQ, mec:%d, pipe:%d, queue:%d.\n", + mec, pipe, queue_id); + value = RREG32(SOC15_REG_OFFSET(GC, 0, regRLC_CP_SCHEDULERS)); + value = REG_SET_FIELD(value, RLC_CP_SCHEDULERS, scheduler1, + ((mec << 5) | (pipe << 3) | queue_id | 0x80)); + WREG32(SOC15_REG_OFFSET(GC, 0, regRLC_CP_SCHEDULERS), value); + } + + /* HQD registers extend from CP_MQD_BASE_ADDR to CP_HQD_EOP_WPTR_MEM. */ + mqd_hqd = &m->cp_mqd_base_addr_lo; + hqd_base = SOC15_REG_OFFSET(GC, 0, regCP_MQD_BASE_ADDR); + + for (reg = hqd_base; + reg <= SOC15_REG_OFFSET(GC, 0, regCP_HQD_PQ_WPTR_HI); reg++) + WREG32(reg, mqd_hqd[reg - hqd_base]); + + + /* Activate doorbell logic before triggering WPTR poll. */ + data = REG_SET_FIELD(m->cp_hqd_pq_doorbell_control, + CP_HQD_PQ_DOORBELL_CONTROL, DOORBELL_EN, 1); + WREG32(SOC15_REG_OFFSET(GC, 0, regCP_HQD_PQ_DOORBELL_CONTROL), data); + + if (wptr) { + /* Don't read wptr with get_user because the user + * context may not be accessible (if this function + * runs in a work queue). Instead trigger a one-shot + * polling read from memory in the CP. This assumes + * that wptr is GPU-accessible in the queue's VMID via + * ATC or SVM. WPTR==RPTR before starting the poll so + * the CP starts fetching new commands from the right + * place. + * + * Guessing a 64-bit WPTR from a 32-bit RPTR is a bit + * tricky. Assume that the queue didn't overflow. The + * number of valid bits in the 32-bit RPTR depends on + * the queue size. The remaining bits are taken from + * the saved 64-bit WPTR. If the WPTR wrapped, add the + * queue size. + */ + uint32_t queue_size = + 2 << REG_GET_FIELD(m->cp_hqd_pq_control, + CP_HQD_PQ_CONTROL, QUEUE_SIZE); + uint64_t guessed_wptr = m->cp_hqd_pq_rptr & (queue_size - 1); + + if ((m->cp_hqd_pq_wptr_lo & (queue_size - 1)) < guessed_wptr) + guessed_wptr += queue_size; + guessed_wptr += m->cp_hqd_pq_wptr_lo & ~(queue_size - 1); + guessed_wptr += (uint64_t)m->cp_hqd_pq_wptr_hi << 32; + + WREG32(SOC15_REG_OFFSET(GC, 0, regCP_HQD_PQ_WPTR_LO), + lower_32_bits(guessed_wptr)); + WREG32(SOC15_REG_OFFSET(GC, 0, regCP_HQD_PQ_WPTR_HI), + upper_32_bits(guessed_wptr)); + WREG32(SOC15_REG_OFFSET(GC, 0, regCP_HQD_PQ_WPTR_POLL_ADDR), + lower_32_bits((uint64_t)wptr)); + WREG32(SOC15_REG_OFFSET(GC, 0, regCP_HQD_PQ_WPTR_POLL_ADDR_HI), + upper_32_bits((uint64_t)wptr)); + pr_debug("%s setting CP_PQ_WPTR_POLL_CNTL1 to %x\n", __func__, + (uint32_t)get_queue_mask(adev, pipe_id, queue_id)); + WREG32(SOC15_REG_OFFSET(GC, 0, regCP_PQ_WPTR_POLL_CNTL1), + (uint32_t)get_queue_mask(adev, pipe_id, queue_id)); + } + + /* Start the EOP fetcher */ + WREG32(SOC15_REG_OFFSET(GC, 0, regCP_HQD_EOP_RPTR), + REG_SET_FIELD(m->cp_hqd_eop_rptr, + CP_HQD_EOP_RPTR, INIT_FETCHER, 1)); + + data = REG_SET_FIELD(m->cp_hqd_active, CP_HQD_ACTIVE, ACTIVE, 1); + WREG32(SOC15_REG_OFFSET(GC, 0, regCP_HQD_ACTIVE), data); + + release_queue(adev); + + return 0; +} + +static int hiq_mqd_load_v11(struct amdgpu_device *adev, void *mqd, + uint32_t pipe_id, uint32_t queue_id, + uint32_t doorbell_off) +{ + struct amdgpu_ring *kiq_ring = &adev->gfx.kiq.ring; + struct v11_compute_mqd *m; + uint32_t mec, pipe; + int r; + + m = get_mqd(mqd); + + acquire_queue(adev, pipe_id, queue_id); + + mec = (pipe_id / adev->gfx.mec.num_pipe_per_mec) + 1; + pipe = (pipe_id % adev->gfx.mec.num_pipe_per_mec); + + pr_debug("kfd: set HIQ, mec:%d, pipe:%d, queue:%d.\n", + mec, pipe, queue_id); + + spin_lock(&adev->gfx.kiq.ring_lock); + r = amdgpu_ring_alloc(kiq_ring, 7); + if (r) { + pr_err("Failed to alloc KIQ (%d).\n", r); + goto out_unlock; + } + + amdgpu_ring_write(kiq_ring, PACKET3(PACKET3_MAP_QUEUES, 5)); + amdgpu_ring_write(kiq_ring, + PACKET3_MAP_QUEUES_QUEUE_SEL(0) | /* Queue_Sel */ + PACKET3_MAP_QUEUES_VMID(m->cp_hqd_vmid) | /* VMID */ + PACKET3_MAP_QUEUES_QUEUE(queue_id) | + PACKET3_MAP_QUEUES_PIPE(pipe) | + PACKET3_MAP_QUEUES_ME((mec - 1)) | + PACKET3_MAP_QUEUES_QUEUE_TYPE(0) | /*queue_type: normal compute queue */ + PACKET3_MAP_QUEUES_ALLOC_FORMAT(0) | /* alloc format: all_on_one_pipe */ + PACKET3_MAP_QUEUES_ENGINE_SEL(1) | /* engine_sel: hiq */ + PACKET3_MAP_QUEUES_NUM_QUEUES(1)); /* num_queues: must be 1 */ + amdgpu_ring_write(kiq_ring, + PACKET3_MAP_QUEUES_DOORBELL_OFFSET(doorbell_off)); + amdgpu_ring_write(kiq_ring, m->cp_mqd_base_addr_lo); + amdgpu_ring_write(kiq_ring, m->cp_mqd_base_addr_hi); + amdgpu_ring_write(kiq_ring, m->cp_hqd_pq_wptr_poll_addr_lo); + amdgpu_ring_write(kiq_ring, m->cp_hqd_pq_wptr_poll_addr_hi); + amdgpu_ring_commit(kiq_ring); + +out_unlock: + spin_unlock(&adev->gfx.kiq.ring_lock); + release_queue(adev); + + return r; +} + +static int hqd_dump_v11(struct amdgpu_device *adev, + uint32_t pipe_id, uint32_t queue_id, + uint32_t (**dump)[2], uint32_t *n_regs) +{ + uint32_t i = 0, reg; +#define HQD_N_REGS 56 +#define DUMP_REG(addr) do { \ + if (WARN_ON_ONCE(i >= HQD_N_REGS)) \ + break; \ + (*dump)[i][0] = (addr) << 2; \ + (*dump)[i++][1] = RREG32(addr); \ + } while (0) + + *dump = kmalloc(HQD_N_REGS*2*sizeof(uint32_t), GFP_KERNEL); + if (*dump == NULL) + return -ENOMEM; + + acquire_queue(adev, pipe_id, queue_id); + + for (reg = SOC15_REG_OFFSET(GC, 0, regCP_MQD_BASE_ADDR); + reg <= SOC15_REG_OFFSET(GC, 0, regCP_HQD_PQ_WPTR_HI); reg++) + DUMP_REG(reg); + + release_queue(adev); + + WARN_ON_ONCE(i != HQD_N_REGS); + *n_regs = i; + + return 0; +} + +static int hqd_sdma_load_v11(struct amdgpu_device *adev, void *mqd, + uint32_t __user *wptr, struct mm_struct *mm) +{ + struct v11_sdma_mqd *m; + uint32_t sdma_rlc_reg_offset; + unsigned long end_jiffies; + uint32_t data; + uint64_t data64; + uint64_t __user *wptr64 = (uint64_t __user *)wptr; + + m = get_sdma_mqd(mqd); + sdma_rlc_reg_offset = get_sdma_rlc_reg_offset(adev, m->sdma_engine_id, + m->sdma_queue_id); + + WREG32(sdma_rlc_reg_offset + regSDMA0_QUEUE0_RB_CNTL, + m->sdmax_rlcx_rb_cntl & (~SDMA0_QUEUE0_RB_CNTL__RB_ENABLE_MASK)); + + end_jiffies = msecs_to_jiffies(2000) + jiffies; + while (true) { + data = RREG32(sdma_rlc_reg_offset + regSDMA0_QUEUE0_CONTEXT_STATUS); + if (data & SDMA0_QUEUE0_CONTEXT_STATUS__IDLE_MASK) + break; + if (time_after(jiffies, end_jiffies)) { + pr_err("SDMA RLC not idle in %s\n", __func__); + return -ETIME; + } + usleep_range(500, 1000); + } + + WREG32(sdma_rlc_reg_offset + regSDMA0_QUEUE0_DOORBELL_OFFSET, + m->sdmax_rlcx_doorbell_offset); + + data = REG_SET_FIELD(m->sdmax_rlcx_doorbell, SDMA0_QUEUE0_DOORBELL, + ENABLE, 1); + WREG32(sdma_rlc_reg_offset + regSDMA0_QUEUE0_DOORBELL, data); + WREG32(sdma_rlc_reg_offset + regSDMA0_QUEUE0_RB_RPTR, + m->sdmax_rlcx_rb_rptr); + WREG32(sdma_rlc_reg_offset + regSDMA0_QUEUE0_RB_RPTR_HI, + m->sdmax_rlcx_rb_rptr_hi); + + WREG32(sdma_rlc_reg_offset + regSDMA0_QUEUE0_MINOR_PTR_UPDATE, 1); + if (read_user_wptr(mm, wptr64, data64)) { + WREG32(sdma_rlc_reg_offset + regSDMA0_QUEUE0_RB_WPTR, + lower_32_bits(data64)); + WREG32(sdma_rlc_reg_offset + regSDMA0_QUEUE0_RB_WPTR_HI, + upper_32_bits(data64)); + } else { + WREG32(sdma_rlc_reg_offset + regSDMA0_QUEUE0_RB_WPTR, + m->sdmax_rlcx_rb_rptr); + WREG32(sdma_rlc_reg_offset + regSDMA0_QUEUE0_RB_WPTR_HI, + m->sdmax_rlcx_rb_rptr_hi); + } + WREG32(sdma_rlc_reg_offset + regSDMA0_QUEUE0_MINOR_PTR_UPDATE, 0); + + WREG32(sdma_rlc_reg_offset + regSDMA0_QUEUE0_RB_BASE, m->sdmax_rlcx_rb_base); + WREG32(sdma_rlc_reg_offset + regSDMA0_QUEUE0_RB_BASE_HI, + m->sdmax_rlcx_rb_base_hi); + WREG32(sdma_rlc_reg_offset + regSDMA0_QUEUE0_RB_RPTR_ADDR_LO, + m->sdmax_rlcx_rb_rptr_addr_lo); + WREG32(sdma_rlc_reg_offset + regSDMA0_QUEUE0_RB_RPTR_ADDR_HI, + m->sdmax_rlcx_rb_rptr_addr_hi); + + data = REG_SET_FIELD(m->sdmax_rlcx_rb_cntl, SDMA0_QUEUE0_RB_CNTL, + RB_ENABLE, 1); + WREG32(sdma_rlc_reg_offset + regSDMA0_QUEUE0_RB_CNTL, data); + + return 0; +} + +static int hqd_sdma_dump_v11(struct amdgpu_device *adev, + uint32_t engine_id, uint32_t queue_id, + uint32_t (**dump)[2], uint32_t *n_regs) +{ + uint32_t sdma_rlc_reg_offset = get_sdma_rlc_reg_offset(adev, + engine_id, queue_id); + uint32_t i = 0, reg; +#undef HQD_N_REGS +#define HQD_N_REGS (7+11+1+12+12) + + *dump = kmalloc(HQD_N_REGS*2*sizeof(uint32_t), GFP_KERNEL); + if (*dump == NULL) + return -ENOMEM; + + for (reg = regSDMA0_QUEUE0_RB_CNTL; + reg <= regSDMA0_QUEUE0_RB_WPTR_HI; reg++) + DUMP_REG(sdma_rlc_reg_offset + reg); + for (reg = regSDMA0_QUEUE0_RB_RPTR_ADDR_HI; + reg <= regSDMA0_QUEUE0_DOORBELL; reg++) + DUMP_REG(sdma_rlc_reg_offset + reg); + for (reg = regSDMA0_QUEUE0_DOORBELL_LOG; + reg <= regSDMA0_QUEUE0_DOORBELL_LOG; reg++) + DUMP_REG(sdma_rlc_reg_offset + reg); + for (reg = regSDMA0_QUEUE0_DOORBELL_OFFSET; + reg <= regSDMA0_QUEUE0_RB_PREEMPT; reg++) + DUMP_REG(sdma_rlc_reg_offset + reg); + for (reg = regSDMA0_QUEUE0_MIDCMD_DATA0; + reg <= regSDMA0_QUEUE0_MIDCMD_CNTL; reg++) + DUMP_REG(sdma_rlc_reg_offset + reg); + + WARN_ON_ONCE(i != HQD_N_REGS); + *n_regs = i; + + return 0; +} + +static bool hqd_is_occupied_v11(struct amdgpu_device *adev, uint64_t queue_address, + uint32_t pipe_id, uint32_t queue_id) +{ + uint32_t act; + bool retval = false; + uint32_t low, high; + + acquire_queue(adev, pipe_id, queue_id); + act = RREG32(SOC15_REG_OFFSET(GC, 0, regCP_HQD_ACTIVE)); + if (act) { + low = lower_32_bits(queue_address >> 8); + high = upper_32_bits(queue_address >> 8); + + if (low == RREG32(SOC15_REG_OFFSET(GC, 0, regCP_HQD_PQ_BASE)) && + high == RREG32(SOC15_REG_OFFSET(GC, 0, regCP_HQD_PQ_BASE_HI))) + retval = true; + } + release_queue(adev); + return retval; +} + +static bool hqd_sdma_is_occupied_v11(struct amdgpu_device *adev, void *mqd) +{ + struct v11_sdma_mqd *m; + uint32_t sdma_rlc_reg_offset; + uint32_t sdma_rlc_rb_cntl; + + m = get_sdma_mqd(mqd); + sdma_rlc_reg_offset = get_sdma_rlc_reg_offset(adev, m->sdma_engine_id, + m->sdma_queue_id); + + sdma_rlc_rb_cntl = RREG32(sdma_rlc_reg_offset + regSDMA0_QUEUE0_RB_CNTL); + + if (sdma_rlc_rb_cntl & SDMA0_QUEUE0_RB_CNTL__RB_ENABLE_MASK) + return true; + + return false; +} + +static int hqd_destroy_v11(struct amdgpu_device *adev, void *mqd, + enum kfd_preempt_type reset_type, + unsigned int utimeout, uint32_t pipe_id, + uint32_t queue_id) +{ + enum hqd_dequeue_request_type type; + unsigned long end_jiffies; + uint32_t temp; + struct v11_compute_mqd *m = get_mqd(mqd); + + acquire_queue(adev, pipe_id, queue_id); + + if (m->cp_hqd_vmid == 0) + WREG32_FIELD15_PREREG(GC, 0, RLC_CP_SCHEDULERS, scheduler1, 0); + + switch (reset_type) { + case KFD_PREEMPT_TYPE_WAVEFRONT_DRAIN: + type = DRAIN_PIPE; + break; + case KFD_PREEMPT_TYPE_WAVEFRONT_RESET: + type = RESET_WAVES; + break; + default: + type = DRAIN_PIPE; + break; + } + + WREG32(SOC15_REG_OFFSET(GC, 0, regCP_HQD_DEQUEUE_REQUEST), type); + + end_jiffies = (utimeout * HZ / 1000) + jiffies; + while (true) { + temp = RREG32(SOC15_REG_OFFSET(GC, 0, regCP_HQD_ACTIVE)); + if (!(temp & CP_HQD_ACTIVE__ACTIVE_MASK)) + break; + if (time_after(jiffies, end_jiffies)) { + pr_err("cp queue pipe %d queue %d preemption failed\n", + pipe_id, queue_id); + release_queue(adev); + return -ETIME; + } + usleep_range(500, 1000); + } + + release_queue(adev); + return 0; +} + +static int hqd_sdma_destroy_v11(struct amdgpu_device *adev, void *mqd, + unsigned int utimeout) +{ + struct v11_sdma_mqd *m; + uint32_t sdma_rlc_reg_offset; + uint32_t temp; + unsigned long end_jiffies = (utimeout * HZ / 1000) + jiffies; + + m = get_sdma_mqd(mqd); + sdma_rlc_reg_offset = get_sdma_rlc_reg_offset(adev, m->sdma_engine_id, + m->sdma_queue_id); + + temp = RREG32(sdma_rlc_reg_offset + regSDMA0_QUEUE0_RB_CNTL); + temp = temp & ~SDMA0_QUEUE0_RB_CNTL__RB_ENABLE_MASK; + WREG32(sdma_rlc_reg_offset + regSDMA0_QUEUE0_RB_CNTL, temp); + + while (true) { + temp = RREG32(sdma_rlc_reg_offset + regSDMA0_QUEUE0_CONTEXT_STATUS); + if (temp & SDMA0_QUEUE0_CONTEXT_STATUS__IDLE_MASK) + break; + if (time_after(jiffies, end_jiffies)) { + pr_err("SDMA RLC not idle in %s\n", __func__); + return -ETIME; + } + usleep_range(500, 1000); + } + + WREG32(sdma_rlc_reg_offset + regSDMA0_QUEUE0_DOORBELL, 0); + WREG32(sdma_rlc_reg_offset + regSDMA0_QUEUE0_RB_CNTL, + RREG32(sdma_rlc_reg_offset + regSDMA0_QUEUE0_RB_CNTL) | + SDMA0_QUEUE0_RB_CNTL__RB_ENABLE_MASK); + + m->sdmax_rlcx_rb_rptr = RREG32(sdma_rlc_reg_offset + regSDMA0_QUEUE0_RB_RPTR); + m->sdmax_rlcx_rb_rptr_hi = + RREG32(sdma_rlc_reg_offset + regSDMA0_QUEUE0_RB_RPTR_HI); + + return 0; +} + +static int wave_control_execute_v11(struct amdgpu_device *adev, + uint32_t gfx_index_val, + uint32_t sq_cmd) +{ + uint32_t data = 0; + + mutex_lock(&adev->grbm_idx_mutex); + + WREG32(SOC15_REG_OFFSET(GC, 0, regGRBM_GFX_INDEX), gfx_index_val); + WREG32(SOC15_REG_OFFSET(GC, 0, regSQ_CMD), sq_cmd); + + data = REG_SET_FIELD(data, GRBM_GFX_INDEX, + INSTANCE_BROADCAST_WRITES, 1); + data = REG_SET_FIELD(data, GRBM_GFX_INDEX, + SA_BROADCAST_WRITES, 1); + data = REG_SET_FIELD(data, GRBM_GFX_INDEX, + SE_BROADCAST_WRITES, 1); + + WREG32(SOC15_REG_OFFSET(GC, 0, regGRBM_GFX_INDEX), data); + mutex_unlock(&adev->grbm_idx_mutex); + + return 0; +} + +static void set_vm_context_page_table_base_v11(struct amdgpu_device *adev, + uint32_t vmid, uint64_t page_table_base) +{ + if (!amdgpu_amdkfd_is_kfd_vmid(adev, vmid)) { + pr_err("trying to set page table base for wrong VMID %u\n", + vmid); + return; + } + + /* SDMA is on gfxhub as well for gfx11 adapters */ + adev->gfxhub.funcs->setup_vm_pt_regs(adev, vmid, page_table_base); +} + +const struct kfd2kgd_calls gfx_v11_kfd2kgd = { + .program_sh_mem_settings = program_sh_mem_settings_v11, + .set_pasid_vmid_mapping = set_pasid_vmid_mapping_v11, + .init_interrupts = init_interrupts_v11, + .hqd_load = hqd_load_v11, + .hiq_mqd_load = hiq_mqd_load_v11, + .hqd_sdma_load = hqd_sdma_load_v11, + .hqd_dump = hqd_dump_v11, + .hqd_sdma_dump = hqd_sdma_dump_v11, + .hqd_is_occupied = hqd_is_occupied_v11, + .hqd_sdma_is_occupied = hqd_sdma_is_occupied_v11, + .hqd_destroy = hqd_destroy_v11, + .hqd_sdma_destroy = hqd_sdma_destroy_v11, + .wave_control_execute = wave_control_execute_v11, + .get_atc_vmid_pasid_mapping_info = NULL, + .set_vm_context_page_table_base = set_vm_context_page_table_base_v11, +}; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c index 36528dad7684..e83cb1c09610 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c @@ -45,43 +45,6 @@ enum { MAX_WATCH_ADDRESSES = 4 }; -enum { - ADDRESS_WATCH_REG_ADDR_HI = 0, - ADDRESS_WATCH_REG_ADDR_LO, - ADDRESS_WATCH_REG_CNTL, - ADDRESS_WATCH_REG_MAX -}; - -/* not defined in the CI/KV reg file */ -enum { - ADDRESS_WATCH_REG_CNTL_ATC_BIT = 0x10000000UL, - ADDRESS_WATCH_REG_CNTL_DEFAULT_MASK = 0x00FFFFFF, - ADDRESS_WATCH_REG_ADDLOW_MASK_EXTENSION = 0x03000000, - /* extend the mask to 26 bits to match the low address field */ - ADDRESS_WATCH_REG_ADDLOW_SHIFT = 6, - ADDRESS_WATCH_REG_ADDHIGH_MASK = 0xFFFF -}; - -static const uint32_t watchRegs[MAX_WATCH_ADDRESSES * ADDRESS_WATCH_REG_MAX] = { - mmTCP_WATCH0_ADDR_H, mmTCP_WATCH0_ADDR_L, mmTCP_WATCH0_CNTL, - mmTCP_WATCH1_ADDR_H, mmTCP_WATCH1_ADDR_L, mmTCP_WATCH1_CNTL, - mmTCP_WATCH2_ADDR_H, mmTCP_WATCH2_ADDR_L, mmTCP_WATCH2_CNTL, - mmTCP_WATCH3_ADDR_H, mmTCP_WATCH3_ADDR_L, mmTCP_WATCH3_CNTL -}; - -union TCP_WATCH_CNTL_BITS { - struct { - uint32_t mask:24; - uint32_t vmid:4; - uint32_t atc:1; - uint32_t mode:2; - uint32_t valid:1; - } bitfields, bits; - uint32_t u32All; - signed int i32All; - float f32All; -}; - static void lock_srbm(struct amdgpu_device *adev, uint32_t mec, uint32_t pipe, uint32_t queue, uint32_t vmid) { @@ -221,7 +184,7 @@ static int kgd_hqd_load(struct amdgpu_device *adev, void *mqd, /* read_user_ptr may take the mm->mmap_lock. * release srbm_mutex to avoid circular dependency between - * srbm_mutex->mm_sem->reservation_ww_class_mutex->srbm_mutex. + * srbm_mutex->mmap_lock->reservation_ww_class_mutex->srbm_mutex. */ release_queue(adev); valid_wptr = read_user_wptr(mm, wptr, wptr_val); @@ -529,55 +492,6 @@ static int kgd_hqd_sdma_destroy(struct amdgpu_device *adev, void *mqd, return 0; } -static int kgd_address_watch_disable(struct amdgpu_device *adev) -{ - union TCP_WATCH_CNTL_BITS cntl; - unsigned int i; - - cntl.u32All = 0; - - cntl.bitfields.valid = 0; - cntl.bitfields.mask = ADDRESS_WATCH_REG_CNTL_DEFAULT_MASK; - cntl.bitfields.atc = 1; - - /* Turning off this address until we set all the registers */ - for (i = 0; i < MAX_WATCH_ADDRESSES; i++) - WREG32(watchRegs[i * ADDRESS_WATCH_REG_MAX + - ADDRESS_WATCH_REG_CNTL], cntl.u32All); - - return 0; -} - -static int kgd_address_watch_execute(struct amdgpu_device *adev, - unsigned int watch_point_id, - uint32_t cntl_val, - uint32_t addr_hi, - uint32_t addr_lo) -{ - union TCP_WATCH_CNTL_BITS cntl; - - cntl.u32All = cntl_val; - - /* Turning off this watch point until we set all the registers */ - cntl.bitfields.valid = 0; - WREG32(watchRegs[watch_point_id * ADDRESS_WATCH_REG_MAX + - ADDRESS_WATCH_REG_CNTL], cntl.u32All); - - WREG32(watchRegs[watch_point_id * ADDRESS_WATCH_REG_MAX + - ADDRESS_WATCH_REG_ADDR_HI], addr_hi); - - WREG32(watchRegs[watch_point_id * ADDRESS_WATCH_REG_MAX + - ADDRESS_WATCH_REG_ADDR_LO], addr_lo); - - /* Enable the watch point */ - cntl.bitfields.valid = 1; - - WREG32(watchRegs[watch_point_id * ADDRESS_WATCH_REG_MAX + - ADDRESS_WATCH_REG_CNTL], cntl.u32All); - - return 0; -} - static int kgd_wave_control_execute(struct amdgpu_device *adev, uint32_t gfx_index_val, uint32_t sq_cmd) @@ -602,13 +516,6 @@ static int kgd_wave_control_execute(struct amdgpu_device *adev, return 0; } -static uint32_t kgd_address_watch_get_offset(struct amdgpu_device *adev, - unsigned int watch_point_id, - unsigned int reg_offset) -{ - return watchRegs[watch_point_id * ADDRESS_WATCH_REG_MAX + reg_offset]; -} - static bool get_atc_vmid_pasid_mapping_info(struct amdgpu_device *adev, uint8_t vmid, uint16_t *p_pasid) { @@ -665,10 +572,7 @@ const struct kfd2kgd_calls gfx_v7_kfd2kgd = { .hqd_sdma_is_occupied = kgd_hqd_sdma_is_occupied, .hqd_destroy = kgd_hqd_destroy, .hqd_sdma_destroy = kgd_hqd_sdma_destroy, - .address_watch_disable = kgd_address_watch_disable, - .address_watch_execute = kgd_address_watch_execute, .wave_control_execute = kgd_wave_control_execute, - .address_watch_get_offset = kgd_address_watch_get_offset, .get_atc_vmid_pasid_mapping_info = get_atc_vmid_pasid_mapping_info, .set_scratch_backing_va = set_scratch_backing_va, .set_vm_context_page_table_base = set_vm_context_page_table_base, diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c index 52832cd69a93..870f352837fc 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c @@ -208,7 +208,7 @@ static int kgd_hqd_load(struct amdgpu_device *adev, void *mqd, /* read_user_ptr may take the mm->mmap_lock. * release srbm_mutex to avoid circular dependency between - * srbm_mutex->mm_sem->reservation_ww_class_mutex->srbm_mutex. + * srbm_mutex->mmap_lock->reservation_ww_class_mutex->srbm_mutex. */ release_queue(adev); valid_wptr = read_user_wptr(mm, wptr, wptr_val); @@ -538,20 +538,6 @@ static bool get_atc_vmid_pasid_mapping_info(struct amdgpu_device *adev, return !!(value & ATC_VMID0_PASID_MAPPING__VALID_MASK); } -static int kgd_address_watch_disable(struct amdgpu_device *adev) -{ - return 0; -} - -static int kgd_address_watch_execute(struct amdgpu_device *adev, - unsigned int watch_point_id, - uint32_t cntl_val, - uint32_t addr_hi, - uint32_t addr_lo) -{ - return 0; -} - static int kgd_wave_control_execute(struct amdgpu_device *adev, uint32_t gfx_index_val, uint32_t sq_cmd) @@ -576,13 +562,6 @@ static int kgd_wave_control_execute(struct amdgpu_device *adev, return 0; } -static uint32_t kgd_address_watch_get_offset(struct amdgpu_device *adev, - unsigned int watch_point_id, - unsigned int reg_offset) -{ - return 0; -} - static void set_scratch_backing_va(struct amdgpu_device *adev, uint64_t va, uint32_t vmid) { @@ -614,10 +593,7 @@ const struct kfd2kgd_calls gfx_v8_kfd2kgd = { .hqd_sdma_is_occupied = kgd_hqd_sdma_is_occupied, .hqd_destroy = kgd_hqd_destroy, .hqd_sdma_destroy = kgd_hqd_sdma_destroy, - .address_watch_disable = kgd_address_watch_disable, - .address_watch_execute = kgd_address_watch_execute, .wave_control_execute = kgd_wave_control_execute, - .address_watch_get_offset = kgd_address_watch_get_offset, .get_atc_vmid_pasid_mapping_info = get_atc_vmid_pasid_mapping_info, .set_scratch_backing_va = set_scratch_backing_va, diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c index 1abf662a0e91..81e3b528bbc9 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c @@ -622,20 +622,6 @@ bool kgd_gfx_v9_get_atc_vmid_pasid_mapping_info(struct amdgpu_device *adev, return !!(value & ATC_VMID0_PASID_MAPPING__VALID_MASK); } -int kgd_gfx_v9_address_watch_disable(struct amdgpu_device *adev) -{ - return 0; -} - -int kgd_gfx_v9_address_watch_execute(struct amdgpu_device *adev, - unsigned int watch_point_id, - uint32_t cntl_val, - uint32_t addr_hi, - uint32_t addr_lo) -{ - return 0; -} - int kgd_gfx_v9_wave_control_execute(struct amdgpu_device *adev, uint32_t gfx_index_val, uint32_t sq_cmd) @@ -660,13 +646,6 @@ int kgd_gfx_v9_wave_control_execute(struct amdgpu_device *adev, return 0; } -uint32_t kgd_gfx_v9_address_watch_get_offset(struct amdgpu_device *adev, - unsigned int watch_point_id, - unsigned int reg_offset) -{ - return 0; -} - void kgd_gfx_v9_set_vm_context_page_table_base(struct amdgpu_device *adev, uint32_t vmid, uint64_t page_table_base) { @@ -736,7 +715,7 @@ static void get_wave_count(struct amdgpu_device *adev, int queue_idx, * process whose pasid is provided as a parameter. The process could have ZERO * or more queues running and submitting waves to compute units. * - * @kgd: Handle of device from which to get number of waves in flight + * @adev: Handle of device from which to get number of waves in flight * @pasid: Identifies the process for which this query call is invoked * @pasid_wave_cnt: Output parameter updated with number of waves in flight that * belong to process with given pasid @@ -745,7 +724,7 @@ static void get_wave_count(struct amdgpu_device *adev, int queue_idx, * * Note: It's possible that the device has too many queues (oversubscription) * in which case a VMID could be remapped to a different PASID. This could lead - * to an iaccurate wave count. Following is a high-level sequence: + * to an inaccurate wave count. Following is a high-level sequence: * Time T1: vmid = getVmid(); vmid is associated with Pasid P1 * Time T2: passId = getPasId(vmid); vmid is associated with Pasid P2 * In the sequence above wave count obtained from time T1 will be incorrectly @@ -888,10 +867,7 @@ const struct kfd2kgd_calls gfx_v9_kfd2kgd = { .hqd_sdma_is_occupied = kgd_hqd_sdma_is_occupied, .hqd_destroy = kgd_gfx_v9_hqd_destroy, .hqd_sdma_destroy = kgd_hqd_sdma_destroy, - .address_watch_disable = kgd_gfx_v9_address_watch_disable, - .address_watch_execute = kgd_gfx_v9_address_watch_execute, .wave_control_execute = kgd_gfx_v9_wave_control_execute, - .address_watch_get_offset = kgd_gfx_v9_address_watch_get_offset, .get_atc_vmid_pasid_mapping_info = kgd_gfx_v9_get_atc_vmid_pasid_mapping_info, .set_vm_context_page_table_base = kgd_gfx_v9_set_vm_context_page_table_base, diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.h index 24be49df26fd..c7ed3bc9053c 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.h @@ -46,19 +46,9 @@ int kgd_gfx_v9_hqd_destroy(struct amdgpu_device *adev, void *mqd, enum kfd_preempt_type reset_type, unsigned int utimeout, uint32_t pipe_id, uint32_t queue_id); -int kgd_gfx_v9_address_watch_disable(struct amdgpu_device *adev); -int kgd_gfx_v9_address_watch_execute(struct amdgpu_device *adev, - unsigned int watch_point_id, - uint32_t cntl_val, - uint32_t addr_hi, - uint32_t addr_lo); int kgd_gfx_v9_wave_control_execute(struct amdgpu_device *adev, uint32_t gfx_index_val, uint32_t sq_cmd); -uint32_t kgd_gfx_v9_address_watch_get_offset(struct amdgpu_device *adev, - unsigned int watch_point_id, - unsigned int reg_offset); - bool kgd_gfx_v9_get_atc_vmid_pasid_mapping_info(struct amdgpu_device *adev, uint8_t vmid, uint16_t *p_pasid); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c index f9bab963a948..84f44f7e4111 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c @@ -1,3 +1,4 @@ +// SPDX-License-Identifier: MIT /* * Copyright 2014-2018 Advanced Micro Devices, Inc. * @@ -32,12 +33,19 @@ #include "amdgpu_dma_buf.h" #include <uapi/linux/kfd_ioctl.h> #include "amdgpu_xgmi.h" +#include "kfd_smi_events.h" /* Userptr restore delay, just long enough to allow consecutive VM * changes to accumulate */ #define AMDGPU_USERPTR_RESTORE_DELAY_MS 1 +/* + * Align VRAM availability to 2MB to avoid fragmentation caused by 4K allocations in the tail 2MB + * BO chunk + */ +#define VRAM_AVAILABLITY_ALIGN (1 << 21) + /* Impose limit on how much memory KFD can use */ static struct { uint64_t max_system_mem_limit; @@ -108,21 +116,12 @@ void amdgpu_amdkfd_reserve_system_mem(uint64_t size) * compromise that should work in most cases without reserving too * much memory for page tables unnecessarily (factor 16K, >> 14). */ -#define ESTIMATE_PT_SIZE(mem_size) ((mem_size) >> 14) -static size_t amdgpu_amdkfd_acc_size(uint64_t size) -{ - size >>= PAGE_SHIFT; - size *= sizeof(dma_addr_t) + sizeof(void *); - - return __roundup_pow_of_two(sizeof(struct amdgpu_bo)) + - __roundup_pow_of_two(sizeof(struct ttm_tt)) + - PAGE_ALIGN(size); -} +#define ESTIMATE_PT_SIZE(mem_size) max(((mem_size) >> 14), AMDGPU_VM_RESERVED_VRAM) /** - * @amdgpu_amdkfd_reserve_mem_limit() - Decrease available memory by size - * of buffer including any reserved for control structures + * amdgpu_amdkfd_reserve_mem_limit() - Decrease available memory by size + * of buffer. * * @adev: Device to which allocated BO belongs to * @size: Size of buffer, in bytes, encapsulated by B0. This should be @@ -131,33 +130,32 @@ static size_t amdgpu_amdkfd_acc_size(uint64_t size) * * Return: returns -ENOMEM in case of error, ZERO otherwise */ -static int amdgpu_amdkfd_reserve_mem_limit(struct amdgpu_device *adev, +int amdgpu_amdkfd_reserve_mem_limit(struct amdgpu_device *adev, uint64_t size, u32 alloc_flag) { uint64_t reserved_for_pt = ESTIMATE_PT_SIZE(amdgpu_amdkfd_total_mem_size); - size_t acc_size, system_mem_needed, ttm_mem_needed, vram_needed; + size_t system_mem_needed, ttm_mem_needed, vram_needed; int ret = 0; - acc_size = amdgpu_amdkfd_acc_size(size); - + system_mem_needed = 0; + ttm_mem_needed = 0; vram_needed = 0; if (alloc_flag & KFD_IOC_ALLOC_MEM_FLAGS_GTT) { - system_mem_needed = acc_size + size; - ttm_mem_needed = acc_size + size; + system_mem_needed = size; + ttm_mem_needed = size; } else if (alloc_flag & KFD_IOC_ALLOC_MEM_FLAGS_VRAM) { - system_mem_needed = acc_size; - ttm_mem_needed = acc_size; + /* + * Conservatively round up the allocation requirement to 2 MB + * to avoid fragmentation caused by 4K allocations in the tail + * 2M BO chunk. + */ vram_needed = size; } else if (alloc_flag & KFD_IOC_ALLOC_MEM_FLAGS_USERPTR) { - system_mem_needed = acc_size + size; - ttm_mem_needed = acc_size; - } else if (alloc_flag & - (KFD_IOC_ALLOC_MEM_FLAGS_DOORBELL | - KFD_IOC_ALLOC_MEM_FLAGS_MMIO_REMAP)) { - system_mem_needed = acc_size; - ttm_mem_needed = acc_size; - } else { + system_mem_needed = size; + } else if (!(alloc_flag & + (KFD_IOC_ALLOC_MEM_FLAGS_DOORBELL | + KFD_IOC_ALLOC_MEM_FLAGS_MMIO_REMAP))) { pr_err("%s: Invalid BO type %#x\n", __func__, alloc_flag); return -ENOMEM; } @@ -172,8 +170,10 @@ static int amdgpu_amdkfd_reserve_mem_limit(struct amdgpu_device *adev, kfd_mem_limit.max_system_mem_limit && !no_system_mem_limit) || (kfd_mem_limit.ttm_mem_used + ttm_mem_needed > kfd_mem_limit.max_ttm_mem_limit) || - (adev->kfd.vram_used + vram_needed > - adev->gmc.real_vram_size - reserved_for_pt)) { + (adev && adev->kfd.vram_used + vram_needed > + adev->gmc.real_vram_size - + atomic64_read(&adev->vram_pin_size) - + reserved_for_pt)) { ret = -ENOMEM; goto release; } @@ -181,7 +181,12 @@ static int amdgpu_amdkfd_reserve_mem_limit(struct amdgpu_device *adev, /* Update memory accounting by decreasing available system * memory, TTM memory and GPU memory as computed above */ - adev->kfd.vram_used += vram_needed; + WARN_ONCE(vram_needed && !adev, + "adev reference can't be null when vram is used"); + if (adev) { + adev->kfd.vram_used += vram_needed; + adev->kfd.vram_used_aligned += ALIGN(vram_needed, VRAM_AVAILABLITY_ALIGN); + } kfd_mem_limit.system_mem_used += system_mem_needed; kfd_mem_limit.ttm_mem_used += ttm_mem_needed; @@ -190,36 +195,30 @@ release: return ret; } -static void unreserve_mem_limit(struct amdgpu_device *adev, +void amdgpu_amdkfd_unreserve_mem_limit(struct amdgpu_device *adev, uint64_t size, u32 alloc_flag) { - size_t acc_size; - - acc_size = amdgpu_amdkfd_acc_size(size); - spin_lock(&kfd_mem_limit.mem_limit_lock); if (alloc_flag & KFD_IOC_ALLOC_MEM_FLAGS_GTT) { - kfd_mem_limit.system_mem_used -= (acc_size + size); - kfd_mem_limit.ttm_mem_used -= (acc_size + size); + kfd_mem_limit.system_mem_used -= size; + kfd_mem_limit.ttm_mem_used -= size; } else if (alloc_flag & KFD_IOC_ALLOC_MEM_FLAGS_VRAM) { - kfd_mem_limit.system_mem_used -= acc_size; - kfd_mem_limit.ttm_mem_used -= acc_size; - adev->kfd.vram_used -= size; + WARN_ONCE(!adev, + "adev reference can't be null when alloc mem flags vram is set"); + if (adev) { + adev->kfd.vram_used -= size; + adev->kfd.vram_used_aligned -= ALIGN(size, VRAM_AVAILABLITY_ALIGN); + } } else if (alloc_flag & KFD_IOC_ALLOC_MEM_FLAGS_USERPTR) { - kfd_mem_limit.system_mem_used -= (acc_size + size); - kfd_mem_limit.ttm_mem_used -= acc_size; - } else if (alloc_flag & - (KFD_IOC_ALLOC_MEM_FLAGS_DOORBELL | - KFD_IOC_ALLOC_MEM_FLAGS_MMIO_REMAP)) { - kfd_mem_limit.system_mem_used -= acc_size; - kfd_mem_limit.ttm_mem_used -= acc_size; - } else { + kfd_mem_limit.system_mem_used -= size; + } else if (!(alloc_flag & + (KFD_IOC_ALLOC_MEM_FLAGS_DOORBELL | + KFD_IOC_ALLOC_MEM_FLAGS_MMIO_REMAP))) { pr_err("%s: Invalid BO type %#x\n", __func__, alloc_flag); goto release; } - - WARN_ONCE(adev->kfd.vram_used < 0, + WARN_ONCE(adev && adev->kfd.vram_used < 0, "KFD VRAM memory accounting unbalanced"); WARN_ONCE(kfd_mem_limit.ttm_mem_used < 0, "KFD TTM memory accounting unbalanced"); @@ -236,11 +235,47 @@ void amdgpu_amdkfd_release_notify(struct amdgpu_bo *bo) u32 alloc_flags = bo->kfd_bo->alloc_flags; u64 size = amdgpu_bo_size(bo); - unreserve_mem_limit(adev, size, alloc_flags); + amdgpu_amdkfd_unreserve_mem_limit(adev, size, alloc_flags); kfree(bo->kfd_bo); } +/** + * @create_dmamap_sg_bo: Creates a amdgpu_bo object to reflect information + * about USERPTR or DOOREBELL or MMIO BO. + * @adev: Device for which dmamap BO is being created + * @mem: BO of peer device that is being DMA mapped. Provides parameters + * in building the dmamap BO + * @bo_out: Output parameter updated with handle of dmamap BO + */ +static int +create_dmamap_sg_bo(struct amdgpu_device *adev, + struct kgd_mem *mem, struct amdgpu_bo **bo_out) +{ + struct drm_gem_object *gem_obj; + int ret, align; + + ret = amdgpu_bo_reserve(mem->bo, false); + if (ret) + return ret; + + align = 1; + ret = amdgpu_gem_object_create(adev, mem->bo->tbo.base.size, align, + AMDGPU_GEM_DOMAIN_CPU, AMDGPU_GEM_CREATE_PREEMPTIBLE, + ttm_bo_type_sg, mem->bo->tbo.base.resv, &gem_obj); + + amdgpu_bo_unreserve(mem->bo); + + if (ret) { + pr_err("Error in creating DMA mappable SG BO on domain: %d\n", ret); + return -EINVAL; + } + + *bo_out = gem_to_amdgpu_bo(gem_obj); + (*bo_out)->parent = amdgpu_bo_ref(mem->bo); + return ret; +} + /* amdgpu_amdkfd_remove_eviction_fence - Removes eviction fence from BO's * reservation object. * @@ -253,53 +288,18 @@ void amdgpu_amdkfd_release_notify(struct amdgpu_bo *bo) static int amdgpu_amdkfd_remove_eviction_fence(struct amdgpu_bo *bo, struct amdgpu_amdkfd_fence *ef) { - struct dma_resv *resv = bo->tbo.base.resv; - struct dma_resv_list *old, *new; - unsigned int i, j, k; + struct dma_fence *replacement; if (!ef) return -EINVAL; - old = dma_resv_shared_list(resv); - if (!old) - return 0; - - new = kmalloc(struct_size(new, shared, old->shared_max), GFP_KERNEL); - if (!new) - return -ENOMEM; - - /* Go through all the shared fences in the resevation object and sort - * the interesting ones to the end of the list. + /* TODO: Instead of block before we should use the fence of the page + * table update and TLB flush here directly. */ - for (i = 0, j = old->shared_count, k = 0; i < old->shared_count; ++i) { - struct dma_fence *f; - - f = rcu_dereference_protected(old->shared[i], - dma_resv_held(resv)); - - if (f->context == ef->base.context) - RCU_INIT_POINTER(new->shared[--j], f); - else - RCU_INIT_POINTER(new->shared[k++], f); - } - new->shared_max = old->shared_max; - new->shared_count = k; - - /* Install the new fence list, seqcount provides the barriers */ - write_seqcount_begin(&resv->seq); - RCU_INIT_POINTER(resv->fence, new); - write_seqcount_end(&resv->seq); - - /* Drop the references to the removed fences or move them to ef_list */ - for (i = j; i < old->shared_count; ++i) { - struct dma_fence *f; - - f = rcu_dereference_protected(new->shared[i], - dma_resv_held(resv)); - dma_fence_put(f); - } - kfree_rcu(old, rcu); - + replacement = dma_fence_get_stub(); + dma_resv_replace_fences(bo->tbo.base.resv, ef->base.context, + replacement, DMA_RESV_USAGE_BOOKKEEP); + dma_fence_put(replacement); return 0; } @@ -385,22 +385,8 @@ static int vm_validate_pt_pd_bos(struct amdgpu_vm *vm) return ret; } - ret = amdgpu_amdkfd_validate_vm_bo(NULL, pd); - if (ret) { - pr_err("failed to validate PD\n"); - return ret; - } - vm->pd_phys_addr = amdgpu_gmc_pd_addr(vm->root.bo); - if (vm->use_cpu_for_update) { - ret = amdgpu_bo_kmap(pd, NULL); - if (ret) { - pr_err("failed to kmap PD, ret=%d\n", ret); - return ret; - } - } - return 0; } @@ -434,45 +420,42 @@ static uint64_t get_pte_flags(struct amdgpu_device *adev, struct kgd_mem *mem) switch (adev->asic_type) { case CHIP_ARCTURUS: - if (mem->alloc_flags & KFD_IOC_ALLOC_MEM_FLAGS_VRAM) { - if (bo_adev == adev) - mapping_flags |= coherent ? - AMDGPU_VM_MTYPE_CC : AMDGPU_VM_MTYPE_RW; - else - mapping_flags |= coherent ? - AMDGPU_VM_MTYPE_UC : AMDGPU_VM_MTYPE_NC; - } else { - mapping_flags |= coherent ? - AMDGPU_VM_MTYPE_UC : AMDGPU_VM_MTYPE_NC; - } - break; case CHIP_ALDEBARAN: - if (coherent && uncached) { - if (adev->gmc.xgmi.connected_to_cpu || - !(mem->alloc_flags & KFD_IOC_ALLOC_MEM_FLAGS_VRAM)) - snoop = true; - mapping_flags |= AMDGPU_VM_MTYPE_UC; - } else if (mem->alloc_flags & KFD_IOC_ALLOC_MEM_FLAGS_VRAM) { + if (mem->alloc_flags & KFD_IOC_ALLOC_MEM_FLAGS_VRAM) { if (bo_adev == adev) { - mapping_flags |= coherent ? - AMDGPU_VM_MTYPE_CC : AMDGPU_VM_MTYPE_RW; - if (adev->gmc.xgmi.connected_to_cpu) + if (uncached) + mapping_flags |= AMDGPU_VM_MTYPE_UC; + else if (coherent) + mapping_flags |= AMDGPU_VM_MTYPE_CC; + else + mapping_flags |= AMDGPU_VM_MTYPE_RW; + if (adev->asic_type == CHIP_ALDEBARAN && + adev->gmc.xgmi.connected_to_cpu) snoop = true; } else { - mapping_flags |= coherent ? - AMDGPU_VM_MTYPE_UC : AMDGPU_VM_MTYPE_NC; + if (uncached || coherent) + mapping_flags |= AMDGPU_VM_MTYPE_UC; + else + mapping_flags |= AMDGPU_VM_MTYPE_NC; if (amdgpu_xgmi_same_hive(adev, bo_adev)) snoop = true; } } else { + if (uncached || coherent) + mapping_flags |= AMDGPU_VM_MTYPE_UC; + else + mapping_flags |= AMDGPU_VM_MTYPE_NC; snoop = true; - mapping_flags |= coherent ? - AMDGPU_VM_MTYPE_UC : AMDGPU_VM_MTYPE_NC; } break; default: - mapping_flags |= coherent ? - AMDGPU_VM_MTYPE_UC : AMDGPU_VM_MTYPE_NC; + if (uncached || coherent) + mapping_flags |= AMDGPU_VM_MTYPE_UC; + else + mapping_flags |= AMDGPU_VM_MTYPE_NC; + + if (!(mem->alloc_flags & KFD_IOC_ALLOC_MEM_FLAGS_VRAM)) + snoop = true; } pte_flags = amdgpu_gem_va_map_flags(adev, mapping_flags); @@ -481,6 +464,38 @@ static uint64_t get_pte_flags(struct amdgpu_device *adev, struct kgd_mem *mem) return pte_flags; } +/** + * create_sg_table() - Create an sg_table for a contiguous DMA addr range + * @addr: The starting address to point to + * @size: Size of memory area in bytes being pointed to + * + * Allocates an instance of sg_table and initializes it to point to memory + * area specified by input parameters. The address used to build is assumed + * to be DMA mapped, if needed. + * + * DOORBELL or MMIO BOs use only one scatterlist node in their sg_table + * because they are physically contiguous. + * + * Return: Initialized instance of SG Table or NULL + */ +static struct sg_table *create_sg_table(uint64_t addr, uint32_t size) +{ + struct sg_table *sg = kmalloc(sizeof(*sg), GFP_KERNEL); + + if (!sg) + return NULL; + if (sg_alloc_table(sg, 1, GFP_KERNEL)) { + kfree(sg); + return NULL; + } + sg_dma_address(sg->sgl) = addr; + sg->sgl->length = size; +#ifdef CONFIG_NEED_SG_DMA_LENGTH + sg->sgl->dma_length = size; +#endif + return sg; +} + static int kfd_mem_dmamap_userptr(struct kgd_mem *mem, struct kfd_mem_attachment *attachment) @@ -495,13 +510,13 @@ kfd_mem_dmamap_userptr(struct kgd_mem *mem, struct ttm_tt *ttm = bo->tbo.ttm; int ret; + if (WARN_ON(ttm->num_pages != src_ttm->num_pages)) + return -EINVAL; + ttm->sg = kmalloc(sizeof(*ttm->sg), GFP_KERNEL); if (unlikely(!ttm->sg)) return -ENOMEM; - if (WARN_ON(ttm->num_pages != src_ttm->num_pages)) - return -EINVAL; - /* Same sequence as in amdgpu_ttm_tt_pin_userptr */ ret = sg_alloc_table_from_pages(ttm->sg, src_ttm->pages, ttm->num_pages, 0, @@ -545,6 +560,87 @@ kfd_mem_dmamap_dmabuf(struct kfd_mem_attachment *attachment) return ttm_bo_validate(&bo->tbo, &bo->placement, &ctx); } +/** + * kfd_mem_dmamap_sg_bo() - Create DMA mapped sg_table to access DOORBELL or MMIO BO + * @mem: SG BO of the DOORBELL or MMIO resource on the owning device + * @attachment: Virtual address attachment of the BO on accessing device + * + * An access request from the device that owns DOORBELL does not require DMA mapping. + * This is because the request doesn't go through PCIe root complex i.e. it instead + * loops back. The need to DMA map arises only when accessing peer device's DOORBELL + * + * In contrast, all access requests for MMIO need to be DMA mapped without regard to + * device ownership. This is because access requests for MMIO go through PCIe root + * complex. + * + * This is accomplished in two steps: + * - Obtain DMA mapped address of DOORBELL or MMIO memory that could be used + * in updating requesting device's page table + * - Signal TTM to mark memory pointed to by requesting device's BO as GPU + * accessible. This allows an update of requesting device's page table + * with entries associated with DOOREBELL or MMIO memory + * + * This method is invoked in the following contexts: + * - Mapping of DOORBELL or MMIO BO of same or peer device + * - Validating an evicted DOOREBELL or MMIO BO on device seeking access + * + * Return: ZERO if successful, NON-ZERO otherwise + */ +static int +kfd_mem_dmamap_sg_bo(struct kgd_mem *mem, + struct kfd_mem_attachment *attachment) +{ + struct ttm_operation_ctx ctx = {.interruptible = true}; + struct amdgpu_bo *bo = attachment->bo_va->base.bo; + struct amdgpu_device *adev = attachment->adev; + struct ttm_tt *ttm = bo->tbo.ttm; + enum dma_data_direction dir; + dma_addr_t dma_addr; + bool mmio; + int ret; + + /* Expect SG Table of dmapmap BO to be NULL */ + mmio = (mem->alloc_flags & KFD_IOC_ALLOC_MEM_FLAGS_MMIO_REMAP); + if (unlikely(ttm->sg)) { + pr_err("SG Table of %d BO for peer device is UNEXPECTEDLY NON-NULL", mmio); + return -EINVAL; + } + + dir = mem->alloc_flags & KFD_IOC_ALLOC_MEM_FLAGS_WRITABLE ? + DMA_BIDIRECTIONAL : DMA_TO_DEVICE; + dma_addr = mem->bo->tbo.sg->sgl->dma_address; + pr_debug("%d BO size: %d\n", mmio, mem->bo->tbo.sg->sgl->length); + pr_debug("%d BO address before DMA mapping: %llx\n", mmio, dma_addr); + dma_addr = dma_map_resource(adev->dev, dma_addr, + mem->bo->tbo.sg->sgl->length, dir, DMA_ATTR_SKIP_CPU_SYNC); + ret = dma_mapping_error(adev->dev, dma_addr); + if (unlikely(ret)) + return ret; + pr_debug("%d BO address after DMA mapping: %llx\n", mmio, dma_addr); + + ttm->sg = create_sg_table(dma_addr, mem->bo->tbo.sg->sgl->length); + if (unlikely(!ttm->sg)) { + ret = -ENOMEM; + goto unmap_sg; + } + + amdgpu_bo_placement_from_domain(bo, AMDGPU_GEM_DOMAIN_GTT); + ret = ttm_bo_validate(&bo->tbo, &bo->placement, &ctx); + if (unlikely(ret)) + goto free_sg; + + return ret; + +free_sg: + sg_free_table(ttm->sg); + kfree(ttm->sg); + ttm->sg = NULL; +unmap_sg: + dma_unmap_resource(adev->dev, dma_addr, mem->bo->tbo.sg->sgl->length, + dir, DMA_ATTR_SKIP_CPU_SYNC); + return ret; +} + static int kfd_mem_dmamap_attachment(struct kgd_mem *mem, struct kfd_mem_attachment *attachment) @@ -556,6 +652,8 @@ kfd_mem_dmamap_attachment(struct kgd_mem *mem, return kfd_mem_dmamap_userptr(mem, attachment); case KFD_MEM_ATT_DMABUF: return kfd_mem_dmamap_dmabuf(attachment); + case KFD_MEM_ATT_SG: + return kfd_mem_dmamap_sg_bo(mem, attachment); default: WARN_ON_ONCE(1); } @@ -596,6 +694,50 @@ kfd_mem_dmaunmap_dmabuf(struct kfd_mem_attachment *attachment) ttm_bo_validate(&bo->tbo, &bo->placement, &ctx); } +/** + * kfd_mem_dmaunmap_sg_bo() - Free DMA mapped sg_table of DOORBELL or MMIO BO + * @mem: SG BO of the DOORBELL or MMIO resource on the owning device + * @attachment: Virtual address attachment of the BO on accessing device + * + * The method performs following steps: + * - Signal TTM to mark memory pointed to by BO as GPU inaccessible + * - Free SG Table that is used to encapsulate DMA mapped memory of + * peer device's DOORBELL or MMIO memory + * + * This method is invoked in the following contexts: + * UNMapping of DOORBELL or MMIO BO on a device having access to its memory + * Eviction of DOOREBELL or MMIO BO on device having access to its memory + * + * Return: void + */ +static void +kfd_mem_dmaunmap_sg_bo(struct kgd_mem *mem, + struct kfd_mem_attachment *attachment) +{ + struct ttm_operation_ctx ctx = {.interruptible = true}; + struct amdgpu_bo *bo = attachment->bo_va->base.bo; + struct amdgpu_device *adev = attachment->adev; + struct ttm_tt *ttm = bo->tbo.ttm; + enum dma_data_direction dir; + + if (unlikely(!ttm->sg)) { + pr_err("SG Table of BO is UNEXPECTEDLY NULL"); + return; + } + + amdgpu_bo_placement_from_domain(bo, AMDGPU_GEM_DOMAIN_CPU); + ttm_bo_validate(&bo->tbo, &bo->placement, &ctx); + + dir = mem->alloc_flags & KFD_IOC_ALLOC_MEM_FLAGS_WRITABLE ? + DMA_BIDIRECTIONAL : DMA_TO_DEVICE; + dma_unmap_resource(adev->dev, ttm->sg->sgl->dma_address, + ttm->sg->sgl->length, dir, DMA_ATTR_SKIP_CPU_SYNC); + sg_free_table(ttm->sg); + kfree(ttm->sg); + ttm->sg = NULL; + bo->tbo.sg = NULL; +} + static void kfd_mem_dmaunmap_attachment(struct kgd_mem *mem, struct kfd_mem_attachment *attachment) @@ -609,39 +751,15 @@ kfd_mem_dmaunmap_attachment(struct kgd_mem *mem, case KFD_MEM_ATT_DMABUF: kfd_mem_dmaunmap_dmabuf(attachment); break; + case KFD_MEM_ATT_SG: + kfd_mem_dmaunmap_sg_bo(mem, attachment); + break; default: WARN_ON_ONCE(1); } } static int -kfd_mem_attach_userptr(struct amdgpu_device *adev, struct kgd_mem *mem, - struct amdgpu_bo **bo) -{ - unsigned long bo_size = mem->bo->tbo.base.size; - struct drm_gem_object *gobj; - int ret; - - ret = amdgpu_bo_reserve(mem->bo, false); - if (ret) - return ret; - - ret = amdgpu_gem_object_create(adev, bo_size, 1, - AMDGPU_GEM_DOMAIN_CPU, - AMDGPU_GEM_CREATE_PREEMPTIBLE, - ttm_bo_type_sg, mem->bo->tbo.base.resv, - &gobj); - amdgpu_bo_unreserve(mem->bo); - if (ret) - return ret; - - *bo = gem_to_amdgpu_bo(gobj); - (*bo)->parent = amdgpu_bo_ref(mem->bo); - - return 0; -} - -static int kfd_mem_attach_dmabuf(struct amdgpu_device *adev, struct kgd_mem *mem, struct amdgpu_bo **bo) { @@ -665,7 +783,6 @@ kfd_mem_attach_dmabuf(struct amdgpu_device *adev, struct kgd_mem *mem, *bo = gem_to_amdgpu_bo(gobj); (*bo)->flags |= AMDGPU_GEM_CREATE_PREEMPTIBLE; - (*bo)->parent = amdgpu_bo_ref(mem->bo); return 0; } @@ -691,6 +808,7 @@ static int kfd_mem_attach(struct amdgpu_device *adev, struct kgd_mem *mem, uint64_t va = mem->va; struct kfd_mem_attachment *attachment[2] = {NULL, NULL}; struct amdgpu_bo *bo[2] = {NULL, NULL}; + bool same_hive = false; int i, ret; if (!va) { @@ -698,6 +816,24 @@ static int kfd_mem_attach(struct amdgpu_device *adev, struct kgd_mem *mem, return -EINVAL; } + /* Determine access to VRAM, MMIO and DOORBELL BOs of peer devices + * + * The access path of MMIO and DOORBELL BOs of is always over PCIe. + * In contrast the access path of VRAM BOs depens upon the type of + * link that connects the peer device. Access over PCIe is allowed + * if peer device has large BAR. In contrast, access over xGMI is + * allowed for both small and large BAR configurations of peer device + */ + if ((adev != bo_adev) && + ((mem->domain == AMDGPU_GEM_DOMAIN_VRAM) || + (mem->alloc_flags & KFD_IOC_ALLOC_MEM_FLAGS_DOORBELL) || + (mem->alloc_flags & KFD_IOC_ALLOC_MEM_FLAGS_MMIO_REMAP))) { + if (mem->domain == AMDGPU_GEM_DOMAIN_VRAM) + same_hive = amdgpu_xgmi_same_hive(adev, bo_adev); + if (!same_hive && !amdgpu_device_is_peer_accessible(bo_adev, adev)) + return -EINVAL; + } + for (i = 0; i <= is_aql; i++) { attachment[i] = kzalloc(sizeof(*attachment[i]), GFP_KERNEL); if (unlikely(!attachment[i])) { @@ -708,9 +844,9 @@ static int kfd_mem_attach(struct amdgpu_device *adev, struct kgd_mem *mem, pr_debug("\t add VA 0x%llx - 0x%llx to vm %p\n", va, va + bo_size, vm); - if (adev == bo_adev || - (amdgpu_ttm_tt_get_usermm(mem->bo->tbo.ttm) && adev->ram_is_direct_mapped) || - (mem->domain == AMDGPU_GEM_DOMAIN_VRAM && amdgpu_xgmi_same_hive(adev, bo_adev))) { + if ((adev == bo_adev && !(mem->alloc_flags & KFD_IOC_ALLOC_MEM_FLAGS_MMIO_REMAP)) || + (amdgpu_ttm_tt_get_usermm(mem->bo->tbo.ttm) && adev->ram_is_direct_mapped) || + same_hive) { /* Mappings on the local GPU, or VRAM mappings in the * local hive, or userptr mapping IOMMU direct map mode * share the original BO @@ -726,26 +862,30 @@ static int kfd_mem_attach(struct amdgpu_device *adev, struct kgd_mem *mem, } else if (amdgpu_ttm_tt_get_usermm(mem->bo->tbo.ttm)) { /* Create an SG BO to DMA-map userptrs on other GPUs */ attachment[i]->type = KFD_MEM_ATT_USERPTR; - ret = kfd_mem_attach_userptr(adev, mem, &bo[i]); + ret = create_dmamap_sg_bo(adev, mem, &bo[i]); if (ret) goto unwind; - } else if (mem->domain == AMDGPU_GEM_DOMAIN_GTT && - mem->bo->tbo.type != ttm_bo_type_sg) { - /* GTT BOs use DMA-mapping ability of dynamic-attach - * DMA bufs. TODO: The same should work for VRAM on - * large-BAR GPUs. - */ + /* Handle DOORBELL BOs of peer devices and MMIO BOs of local and peer devices */ + } else if (mem->bo->tbo.type == ttm_bo_type_sg) { + WARN_ONCE(!(mem->alloc_flags & KFD_IOC_ALLOC_MEM_FLAGS_DOORBELL || + mem->alloc_flags & KFD_IOC_ALLOC_MEM_FLAGS_MMIO_REMAP), + "Handing invalid SG BO in ATTACH request"); + attachment[i]->type = KFD_MEM_ATT_SG; + ret = create_dmamap_sg_bo(adev, mem, &bo[i]); + if (ret) + goto unwind; + /* Enable acces to GTT and VRAM BOs of peer devices */ + } else if (mem->domain == AMDGPU_GEM_DOMAIN_GTT || + mem->domain == AMDGPU_GEM_DOMAIN_VRAM) { attachment[i]->type = KFD_MEM_ATT_DMABUF; ret = kfd_mem_attach_dmabuf(adev, mem, &bo[i]); if (ret) goto unwind; + pr_debug("Employ DMABUF mechanism to enable peer GPU access\n"); } else { - /* FIXME: Need to DMA-map other BO types: - * large-BAR VRAM, doorbells, MMIO remap - */ - attachment[i]->type = KFD_MEM_ATT_SHARED; - bo[i] = mem->bo; - drm_gem_object_get(&bo[i]->tbo.base); + WARN_ONCE(true, "Handling invalid ATTACH request"); + ret = -EINVAL; + goto unwind; } /* Add BO to VM internal data structures */ @@ -778,7 +918,7 @@ unwind: continue; if (attachment[i]->bo_va) { amdgpu_bo_reserve(bo[i], true); - amdgpu_vm_bo_rmv(adev, attachment[i]->bo_va); + amdgpu_vm_bo_del(adev, attachment[i]->bo_va); amdgpu_bo_unreserve(bo[i]); list_del(&attachment[i]->list); } @@ -795,7 +935,7 @@ static void kfd_mem_detach(struct kfd_mem_attachment *attachment) pr_debug("\t remove VA 0x%llx in entry %p\n", attachment->va, attachment); - amdgpu_vm_bo_rmv(attachment->adev, attachment->bo_va); + amdgpu_vm_bo_del(attachment->adev, attachment->bo_va); drm_gem_object_put(&bo->tbo.base); list_del(&attachment->list); kfree(attachment); @@ -842,7 +982,8 @@ static void remove_kgd_mem_from_kfd_bo_list(struct kgd_mem *mem, * * Returns 0 for success, negative errno for errors. */ -static int init_user_pages(struct kgd_mem *mem, uint64_t user_addr) +static int init_user_pages(struct kgd_mem *mem, uint64_t user_addr, + bool criu_resume) { struct amdkfd_process_info *process_info = mem->process_info; struct amdgpu_bo *bo = mem->bo; @@ -864,6 +1005,18 @@ static int init_user_pages(struct kgd_mem *mem, uint64_t user_addr) goto out; } + if (criu_resume) { + /* + * During a CRIU restore operation, the userptr buffer objects + * will be validated in the restore_userptr_work worker at a + * later stage when it is scheduled by another ioctl called by + * CRIU master process for the target pid for restore. + */ + atomic_inc(&mem->invalid); + mutex_unlock(&process_info->lock); + return 0; + } + ret = amdgpu_ttm_tt_get_user_pages(bo, bo->tbo.ttm->pages); if (ret) { pr_err("%s: Failed to get user pages: %d\n", __func__, ret); @@ -1080,8 +1233,7 @@ static void unmap_bo_from_gpuvm(struct kgd_mem *mem, static int update_gpuvm_pte(struct kgd_mem *mem, struct kfd_mem_attachment *entry, - struct amdgpu_sync *sync, - bool *table_freed) + struct amdgpu_sync *sync) { struct amdgpu_bo_va *bo_va = entry->bo_va; struct amdgpu_device *adev = entry->adev; @@ -1092,7 +1244,7 @@ static int update_gpuvm_pte(struct kgd_mem *mem, return ret; /* Update the page tables */ - ret = amdgpu_vm_bo_update(adev, bo_va, false, table_freed); + ret = amdgpu_vm_bo_update(adev, bo_va, false); if (ret) { pr_err("amdgpu_vm_bo_update failed\n"); return ret; @@ -1104,8 +1256,7 @@ static int update_gpuvm_pte(struct kgd_mem *mem, static int map_bo_to_gpuvm(struct kgd_mem *mem, struct kfd_mem_attachment *entry, struct amdgpu_sync *sync, - bool no_update_pte, - bool *table_freed) + bool no_update_pte) { int ret; @@ -1122,7 +1273,7 @@ static int map_bo_to_gpuvm(struct kgd_mem *mem, if (no_update_pte) return 0; - ret = update_gpuvm_pte(mem, entry, sync, table_freed); + ret = update_gpuvm_pte(mem, entry, sync); if (ret) { pr_err("update_gpuvm_pte() failed\n"); goto update_gpuvm_pte_failed; @@ -1135,24 +1286,6 @@ update_gpuvm_pte_failed: return ret; } -static struct sg_table *create_doorbell_sg(uint64_t addr, uint32_t size) -{ - struct sg_table *sg = kmalloc(sizeof(*sg), GFP_KERNEL); - - if (!sg) - return NULL; - if (sg_alloc_table(sg, 1, GFP_KERNEL)) { - kfree(sg); - return NULL; - } - sg->sgl->dma_address = addr; - sg->sgl->length = size; -#ifdef CONFIG_NEED_SG_DMA_LENGTH - sg->sgl->dma_length = size; -#endif - return sg; -} - static int process_validate_vms(struct amdkfd_process_info *process_info) { struct amdgpu_vm *peer_vm; @@ -1255,11 +1388,12 @@ static int init_kfd_vm(struct amdgpu_vm *vm, void **process_info, AMDGPU_FENCE_OWNER_KFD, false); if (ret) goto wait_pd_fail; - ret = dma_resv_reserve_shared(vm->root.bo->tbo.base.resv, 1); + ret = dma_resv_reserve_fences(vm->root.bo->tbo.base.resv, 1); if (ret) goto reserve_shared_fail; - amdgpu_bo_fence(vm->root.bo, - &vm->process_info->eviction_fence->base, true); + dma_resv_add_fence(vm->root.bo->tbo.base.resv, + &vm->process_info->eviction_fence->base, + DMA_RESV_USAGE_BOOKKEEP); amdgpu_bo_unreserve(vm->root.bo); /* Update process info */ @@ -1388,16 +1522,10 @@ void amdgpu_amdkfd_gpuvm_destroy_cb(struct amdgpu_device *adev, struct amdgpu_vm *vm) { struct amdkfd_process_info *process_info = vm->process_info; - struct amdgpu_bo *pd = vm->root.bo; if (!process_info) return; - /* Release eviction fence from PD */ - amdgpu_bo_reserve(pd, false); - amdgpu_bo_fence(pd, NULL, false); - amdgpu_bo_unreserve(pd); - /* Update process info */ mutex_lock(&process_info->lock); process_info->n_vms--; @@ -1452,10 +1580,55 @@ uint64_t amdgpu_amdkfd_gpuvm_get_process_page_dir(void *drm_priv) return avm->pd_phys_addr; } +void amdgpu_amdkfd_block_mmu_notifications(void *p) +{ + struct amdkfd_process_info *pinfo = (struct amdkfd_process_info *)p; + + mutex_lock(&pinfo->lock); + WRITE_ONCE(pinfo->block_mmu_notifications, true); + mutex_unlock(&pinfo->lock); +} + +int amdgpu_amdkfd_criu_resume(void *p) +{ + int ret = 0; + struct amdkfd_process_info *pinfo = (struct amdkfd_process_info *)p; + + mutex_lock(&pinfo->lock); + pr_debug("scheduling work\n"); + atomic_inc(&pinfo->evicted_bos); + if (!READ_ONCE(pinfo->block_mmu_notifications)) { + ret = -EINVAL; + goto out_unlock; + } + WRITE_ONCE(pinfo->block_mmu_notifications, false); + schedule_delayed_work(&pinfo->restore_userptr_work, 0); + +out_unlock: + mutex_unlock(&pinfo->lock); + return ret; +} + +size_t amdgpu_amdkfd_get_available_memory(struct amdgpu_device *adev) +{ + uint64_t reserved_for_pt = + ESTIMATE_PT_SIZE(amdgpu_amdkfd_total_mem_size); + size_t available; + + spin_lock(&kfd_mem_limit.mem_limit_lock); + available = adev->gmc.real_vram_size + - adev->kfd.vram_used_aligned + - atomic64_read(&adev->vram_pin_size) + - reserved_for_pt; + spin_unlock(&kfd_mem_limit.mem_limit_lock); + + return ALIGN_DOWN(available, VRAM_AVAILABLITY_ALIGN); +} + int amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu( struct amdgpu_device *adev, uint64_t va, uint64_t size, void *drm_priv, struct kgd_mem **mem, - uint64_t *offset, uint32_t flags) + uint64_t *offset, uint32_t flags, bool criu_resume) { struct amdgpu_vm *avm = drm_priv_to_vm(drm_priv); enum ttm_bo_type bo_type = ttm_bo_type_device; @@ -1478,26 +1651,26 @@ int amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu( } else if (flags & KFD_IOC_ALLOC_MEM_FLAGS_GTT) { domain = alloc_domain = AMDGPU_GEM_DOMAIN_GTT; alloc_flags = 0; - } else if (flags & KFD_IOC_ALLOC_MEM_FLAGS_USERPTR) { + } else { domain = AMDGPU_GEM_DOMAIN_GTT; alloc_domain = AMDGPU_GEM_DOMAIN_CPU; alloc_flags = AMDGPU_GEM_CREATE_PREEMPTIBLE; - if (!offset || !*offset) - return -EINVAL; - user_addr = untagged_addr(*offset); - } else if (flags & (KFD_IOC_ALLOC_MEM_FLAGS_DOORBELL | - KFD_IOC_ALLOC_MEM_FLAGS_MMIO_REMAP)) { - domain = AMDGPU_GEM_DOMAIN_GTT; - alloc_domain = AMDGPU_GEM_DOMAIN_CPU; - bo_type = ttm_bo_type_sg; - alloc_flags = 0; - if (size > UINT_MAX) + + if (flags & KFD_IOC_ALLOC_MEM_FLAGS_USERPTR) { + if (!offset || !*offset) + return -EINVAL; + user_addr = untagged_addr(*offset); + } else if (flags & (KFD_IOC_ALLOC_MEM_FLAGS_DOORBELL | + KFD_IOC_ALLOC_MEM_FLAGS_MMIO_REMAP)) { + bo_type = ttm_bo_type_sg; + if (size > UINT_MAX) + return -EINVAL; + sg = create_sg_table(*offset, size); + if (!sg) + return -ENOMEM; + } else { return -EINVAL; - sg = create_doorbell_sg(*offset, size); - if (!sg) - return -ENOMEM; - } else { - return -EINVAL; + } } *mem = kzalloc(sizeof(struct kgd_mem), GFP_KERNEL); @@ -1558,7 +1731,8 @@ int amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu( add_kgd_mem_to_kfd_bo_list(*mem, avm->process_info, user_addr); if (user_addr) { - ret = init_user_pages(*mem, user_addr); + pr_debug("creating userptr BO for user_addr = %llx\n", user_addr); + ret = init_user_pages(*mem, user_addr, criu_resume); if (ret) goto allocate_init_user_pages_failed; } else if (flags & (KFD_IOC_ALLOC_MEM_FLAGS_DOORBELL | @@ -1585,7 +1759,7 @@ err_node_allow: /* Don't unreserve system mem limit twice */ goto err_reserve_limit; err_bo_create: - unreserve_mem_limit(adev, size, flags); + amdgpu_amdkfd_unreserve_mem_limit(adev, size, flags); err_reserve_limit: mutex_destroy(&(*mem)->lock); if (gobj) @@ -1606,6 +1780,7 @@ int amdgpu_amdkfd_gpuvm_free_memory_of_gpu( { struct amdkfd_process_info *process_info = mem->process_info; unsigned long bo_size = mem->bo->tbo.base.size; + bool use_release_notifier = (mem->bo->kfd_bo == mem); struct kfd_mem_attachment *entry, *tmp; struct bo_vm_reservation_context ctx; struct ttm_validate_buffer *bo_list_entry; @@ -1615,7 +1790,7 @@ int amdgpu_amdkfd_gpuvm_free_memory_of_gpu( mutex_lock(&mem->lock); - /* Unpin MMIO/DOORBELL BO's that were pinnned during allocation */ + /* Unpin MMIO/DOORBELL BO's that were pinned during allocation */ if (mem->alloc_flags & (KFD_IOC_ALLOC_MEM_FLAGS_DOORBELL | KFD_IOC_ALLOC_MEM_FLAGS_MMIO_REMAP)) { @@ -1697,12 +1872,19 @@ int amdgpu_amdkfd_gpuvm_free_memory_of_gpu( */ drm_gem_object_put(&mem->bo->tbo.base); + /* + * For kgd_mem allocated in amdgpu_amdkfd_gpuvm_import_dmabuf(), + * explicitly free it here. + */ + if (!use_release_notifier) + kfree(mem); + return ret; } int amdgpu_amdkfd_gpuvm_map_memory_to_gpu( struct amdgpu_device *adev, struct kgd_mem *mem, - void *drm_priv, bool *table_freed) + void *drm_priv) { struct amdgpu_vm *avm = drm_priv_to_vm(drm_priv); int ret; @@ -1789,7 +1971,7 @@ int amdgpu_amdkfd_gpuvm_map_memory_to_gpu( entry->va, entry->va + bo_size, entry); ret = map_bo_to_gpuvm(mem, entry, ctx.sync, - is_invalid_userptr, table_freed); + is_invalid_userptr); if (ret) { pr_err("Failed to map bo to gpuvm\n"); goto out_unreserve; @@ -1808,17 +1990,11 @@ int amdgpu_amdkfd_gpuvm_map_memory_to_gpu( } if (!amdgpu_ttm_tt_get_usermm(bo->tbo.ttm) && !bo->tbo.pin_count) - amdgpu_bo_fence(bo, - &avm->process_info->eviction_fence->base, - true); + dma_resv_add_fence(bo->tbo.base.resv, + &avm->process_info->eviction_fence->base, + DMA_RESV_USAGE_BOOKKEEP); ret = unreserve_bo_and_vms(&ctx, false, false); - /* Only apply no TLB flush on Aldebaran to - * workaround regressions on other Asics. - */ - if (table_freed && (adev->asic_type != CHIP_ALDEBARAN)) - *table_freed = true; - goto out; out_unreserve: @@ -1907,8 +2083,69 @@ int amdgpu_amdkfd_gpuvm_sync_memory( return ret; } -int amdgpu_amdkfd_gpuvm_map_gtt_bo_to_kernel(struct amdgpu_device *adev, - struct kgd_mem *mem, void **kptr, uint64_t *size) +/** + * amdgpu_amdkfd_map_gtt_bo_to_gart - Map BO to GART and increment reference count + * @adev: Device to which allocated BO belongs + * @bo: Buffer object to be mapped + * + * Before return, bo reference count is incremented. To release the reference and unpin/ + * unmap the BO, call amdgpu_amdkfd_free_gtt_mem. + */ +int amdgpu_amdkfd_map_gtt_bo_to_gart(struct amdgpu_device *adev, struct amdgpu_bo *bo) +{ + int ret; + + ret = amdgpu_bo_reserve(bo, true); + if (ret) { + pr_err("Failed to reserve bo. ret %d\n", ret); + goto err_reserve_bo_failed; + } + + ret = amdgpu_bo_pin(bo, AMDGPU_GEM_DOMAIN_GTT); + if (ret) { + pr_err("Failed to pin bo. ret %d\n", ret); + goto err_pin_bo_failed; + } + + ret = amdgpu_ttm_alloc_gart(&bo->tbo); + if (ret) { + pr_err("Failed to bind bo to GART. ret %d\n", ret); + goto err_map_bo_gart_failed; + } + + amdgpu_amdkfd_remove_eviction_fence( + bo, bo->kfd_bo->process_info->eviction_fence); + + amdgpu_bo_unreserve(bo); + + bo = amdgpu_bo_ref(bo); + + return 0; + +err_map_bo_gart_failed: + amdgpu_bo_unpin(bo); +err_pin_bo_failed: + amdgpu_bo_unreserve(bo); +err_reserve_bo_failed: + + return ret; +} + +/** amdgpu_amdkfd_gpuvm_map_gtt_bo_to_kernel() - Map a GTT BO for kernel CPU access + * + * @mem: Buffer object to be mapped for CPU access + * @kptr[out]: pointer in kernel CPU address space + * @size[out]: size of the buffer + * + * Pins the BO and maps it for kernel CPU access. The eviction fence is removed + * from the BO, since pinned BOs cannot be evicted. The bo must remain on the + * validate_list, so the GPU mapping can be restored after a page table was + * evicted. + * + * Return: 0 on success, error code on failure + */ +int amdgpu_amdkfd_gpuvm_map_gtt_bo_to_kernel(struct kgd_mem *mem, + void **kptr, uint64_t *size) { int ret; struct amdgpu_bo *bo = mem->bo; @@ -1918,9 +2155,6 @@ int amdgpu_amdkfd_gpuvm_map_gtt_bo_to_kernel(struct amdgpu_device *adev, return -EINVAL; } - /* delete kgd_mem from kfd_bo_list to avoid re-validating - * this BO in BO's restoring after eviction. - */ mutex_lock(&mem->process_info->lock); ret = amdgpu_bo_reserve(bo, true); @@ -1943,7 +2177,6 @@ int amdgpu_amdkfd_gpuvm_map_gtt_bo_to_kernel(struct amdgpu_device *adev, amdgpu_amdkfd_remove_eviction_fence( bo, mem->process_info->eviction_fence); - list_del_init(&mem->validate_list.head); if (size) *size = amdgpu_bo_size(bo); @@ -1963,8 +2196,15 @@ bo_reserve_failed: return ret; } -void amdgpu_amdkfd_gpuvm_unmap_gtt_bo_from_kernel(struct amdgpu_device *adev, - struct kgd_mem *mem) +/** amdgpu_amdkfd_gpuvm_map_gtt_bo_to_kernel() - Unmap a GTT BO for kernel CPU access + * + * @mem: Buffer object to be unmapped for CPU access + * + * Removes the kernel CPU mapping and unpins the BO. It does not restore the + * eviction fence, so this function should only be used for cleanup before the + * BO is destroyed. + */ +void amdgpu_amdkfd_gpuvm_unmap_gtt_bo_from_kernel(struct kgd_mem *mem) { struct amdgpu_bo *bo = mem->bo; @@ -1979,7 +2219,7 @@ int amdgpu_amdkfd_gpuvm_get_vm_fault_info(struct amdgpu_device *adev, { if (atomic_read(&adev->gmc.vm_fault_info_updated) == 1) { *mem = *adev->gmc.vm_fault_info; - mb(); + mb(); /* make sure read happened */ atomic_set(&adev->gmc.vm_fault_info_updated, 0); } return 0; @@ -2068,11 +2308,15 @@ int amdgpu_amdkfd_evict_userptr(struct kgd_mem *mem, int evicted_bos; int r = 0; + /* Do not process MMU notifications until stage-4 IOCTL is received */ + if (READ_ONCE(process_info->block_mmu_notifications)) + return 0; + atomic_inc(&mem->invalid); evicted_bos = atomic_inc_return(&process_info->evicted_bos); if (evicted_bos == 1) { /* First eviction, stop the queues */ - r = kgd2kfd_quiesce_mm(mm); + r = kgd2kfd_quiesce_mm(mm, KFD_QUEUE_EVICTION_TRIGGER_USERPTR); if (r) pr_err("Failed to quiesce KFD\n"); schedule_delayed_work(&process_info->restore_userptr_work, @@ -2259,7 +2503,7 @@ static int validate_invalid_user_pages(struct amdkfd_process_info *process_info) continue; kfd_mem_dmaunmap_attachment(mem, attachment); - ret = update_gpuvm_pte(mem, attachment, &sync, NULL); + ret = update_gpuvm_pte(mem, attachment, &sync); if (ret) { pr_err("%s: update PTE failed\n", __func__); /* make sure this gets validated again */ @@ -2346,13 +2590,16 @@ static void amdgpu_amdkfd_restore_userptr_worker(struct work_struct *work) unlock_out: mutex_unlock(&process_info->lock); - mmput(mm); - put_task_struct(usertask); /* If validation failed, reschedule another attempt */ - if (evicted_bos) + if (evicted_bos) { schedule_delayed_work(&process_info->restore_userptr_work, msecs_to_jiffies(AMDGPU_USERPTR_RESTORE_DELAY_MS)); + + kfd_smi_event_queue_restore_rescheduled(mm); + } + mmput(mm); + put_task_struct(usertask); } /** amdgpu_amdkfd_gpuvm_restore_process_bos - Restore all BOs for the given @@ -2441,6 +2688,8 @@ int amdgpu_amdkfd_gpuvm_restore_process_bos(void *info, struct dma_fence **ef) struct amdgpu_bo *bo = mem->bo; uint32_t domain = mem->domain; struct kfd_mem_attachment *attachment; + struct dma_resv_iter cursor; + struct dma_fence *fence; total_size += amdgpu_bo_size(bo); @@ -2455,17 +2704,20 @@ int amdgpu_amdkfd_gpuvm_restore_process_bos(void *info, struct dma_fence **ef) goto validate_map_fail; } } - ret = amdgpu_sync_fence(&sync_obj, bo->tbo.moving); - if (ret) { - pr_debug("Memory eviction: Sync BO fence failed. Try again\n"); - goto validate_map_fail; + dma_resv_for_each_fence(&cursor, bo->tbo.base.resv, + DMA_RESV_USAGE_KERNEL, fence) { + ret = amdgpu_sync_fence(&sync_obj, fence); + if (ret) { + pr_debug("Memory eviction: Sync BO fence failed. Try again\n"); + goto validate_map_fail; + } } list_for_each_entry(attachment, &mem->attachments, list) { if (!attachment->is_mapped) continue; kfd_mem_dmaunmap_attachment(mem, attachment); - ret = update_gpuvm_pte(mem, attachment, &sync_obj, NULL); + ret = update_gpuvm_pte(mem, attachment, &sync_obj); if (ret) { pr_debug("Memory eviction: update PTE failed. Try again\n"); goto validate_map_fail; @@ -2503,18 +2755,24 @@ int amdgpu_amdkfd_gpuvm_restore_process_bos(void *info, struct dma_fence **ef) process_info->eviction_fence = new_fence; *ef = dma_fence_get(&new_fence->base); - /* Attach new eviction fence to all BOs */ + /* Attach new eviction fence to all BOs except pinned ones */ list_for_each_entry(mem, &process_info->kfd_bo_list, - validate_list.head) - amdgpu_bo_fence(mem->bo, - &process_info->eviction_fence->base, true); + validate_list.head) { + if (mem->bo->tbo.pin_count) + continue; + dma_resv_add_fence(mem->bo->tbo.base.resv, + &process_info->eviction_fence->base, + DMA_RESV_USAGE_BOOKKEEP); + } /* Attach eviction fence to PD / PT BOs */ list_for_each_entry(peer_vm, &process_info->vm_list_head, vm_list_node) { struct amdgpu_bo *bo = peer_vm->root.bo; - amdgpu_bo_fence(bo, &process_info->eviction_fence->base, true); + dma_resv_add_fence(bo->tbo.base.resv, + &process_info->eviction_fence->base, + DMA_RESV_USAGE_BOOKKEEP); } validate_map_fail: @@ -2565,10 +2823,12 @@ int amdgpu_amdkfd_add_gws_to_process(void *info, void *gws, struct kgd_mem **mem * Add process eviction fence to bo so they can * evict each other. */ - ret = dma_resv_reserve_shared(gws_bo->tbo.base.resv, 1); + ret = dma_resv_reserve_fences(gws_bo->tbo.base.resv, 1); if (ret) goto reserve_shared_fail; - amdgpu_bo_fence(gws_bo, &process_info->eviction_fence->base, true); + dma_resv_add_fence(gws_bo->tbo.base.resv, + &process_info->eviction_fence->base, + DMA_RESV_USAGE_BOOKKEEP); amdgpu_bo_unreserve(gws_bo); mutex_unlock(&(*mem)->process_info->lock); @@ -2635,3 +2895,33 @@ int amdgpu_amdkfd_get_tile_config(struct amdgpu_device *adev, return 0; } + +bool amdgpu_amdkfd_bo_mapped_to_dev(struct amdgpu_device *adev, struct kgd_mem *mem) +{ + struct kfd_mem_attachment *entry; + + list_for_each_entry(entry, &mem->attachments, list) { + if (entry->is_mapped && entry->adev == adev) + return true; + } + return false; +} + +#if defined(CONFIG_DEBUG_FS) + +int kfd_debugfs_kfd_mem_limits(struct seq_file *m, void *data) +{ + + spin_lock(&kfd_mem_limit.mem_limit_lock); + seq_printf(m, "System mem used %lldM out of %lluM\n", + (kfd_mem_limit.system_mem_used >> 20), + (kfd_mem_limit.max_system_mem_limit >> 20)); + seq_printf(m, "TTM mem used %lldM out of %lluM\n", + (kfd_mem_limit.ttm_mem_used >> 20), + (kfd_mem_limit.max_ttm_mem_limit >> 20)); + spin_unlock(&kfd_mem_limit.mem_limit_lock); + + return 0; +} + +#endif diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_atombios.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_atombios.c index 12a6b1c99c93..9ba4817a9148 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_atombios.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_atombios.c @@ -1083,6 +1083,7 @@ int amdgpu_atombios_get_clock_dividers(struct amdgpu_device *adev, return 0; } +#ifdef CONFIG_DRM_AMDGPU_SI int amdgpu_atombios_get_memory_pll_dividers(struct amdgpu_device *adev, u32 clock, bool strobe_mode, @@ -1503,6 +1504,7 @@ int amdgpu_atombios_init_mc_reg_table(struct amdgpu_device *adev, } return -EINVAL; } +#endif bool amdgpu_atombios_has_gpu_virtualization_table(struct amdgpu_device *adev) { diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_atombios.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_atombios.h index 27e74b1fc260..4153d520e2a3 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_atombios.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_atombios.h @@ -160,6 +160,7 @@ int amdgpu_atombios_get_clock_dividers(struct amdgpu_device *adev, bool strobe_mode, struct atom_clock_dividers *dividers); +#ifdef CONFIG_DRM_AMDGPU_SI int amdgpu_atombios_get_memory_pll_dividers(struct amdgpu_device *adev, u32 clock, bool strobe_mode, @@ -179,6 +180,17 @@ int amdgpu_atombios_get_voltage_table(struct amdgpu_device *adev, int amdgpu_atombios_init_mc_reg_table(struct amdgpu_device *adev, u8 module_index, struct atom_mc_reg_table *reg_table); +int amdgpu_atombios_get_max_vddc(struct amdgpu_device *adev, u8 voltage_type, + u16 voltage_id, u16 *voltage); +int amdgpu_atombios_get_leakage_vddc_based_on_leakage_idx(struct amdgpu_device *adev, + u16 *voltage, + u16 leakage_idx); +void amdgpu_atombios_get_default_voltages(struct amdgpu_device *adev, + u16 *vddc, u16 *vddci, u16 *mvdd); +int amdgpu_atombios_get_svi2_info(struct amdgpu_device *adev, + u8 voltage_type, + u8 *svd_gpio_id, u8 *svc_gpio_id); +#endif bool amdgpu_atombios_has_gpu_virtualization_table(struct amdgpu_device *adev); @@ -190,21 +202,11 @@ void amdgpu_atombios_scratch_regs_set_backlight_level(struct amdgpu_device *adev bool amdgpu_atombios_scratch_need_asic_init(struct amdgpu_device *adev); void amdgpu_atombios_copy_swap(u8 *dst, u8 *src, u8 num_bytes, bool to_le); -int amdgpu_atombios_get_max_vddc(struct amdgpu_device *adev, u8 voltage_type, - u16 voltage_id, u16 *voltage); -int amdgpu_atombios_get_leakage_vddc_based_on_leakage_idx(struct amdgpu_device *adev, - u16 *voltage, - u16 leakage_idx); -void amdgpu_atombios_get_default_voltages(struct amdgpu_device *adev, - u16 *vddc, u16 *vddci, u16 *mvdd); int amdgpu_atombios_get_clock_dividers(struct amdgpu_device *adev, u8 clock_type, u32 clock, bool strobe_mode, struct atom_clock_dividers *dividers); -int amdgpu_atombios_get_svi2_info(struct amdgpu_device *adev, - u8 voltage_type, - u8 *svd_gpio_id, u8 *svc_gpio_id); int amdgpu_atombios_get_data_table(struct amdgpu_device *adev, uint32_t table, diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_atomfirmware.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_atomfirmware.c index 4d4ddf026faf..b81b77a9efa6 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_atomfirmware.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_atomfirmware.c @@ -162,12 +162,14 @@ union vram_info { struct atom_vram_info_header_v2_4 v24; struct atom_vram_info_header_v2_5 v25; struct atom_vram_info_header_v2_6 v26; + struct atom_vram_info_header_v3_0 v30; }; union vram_module { struct atom_vram_module_v9 v9; struct atom_vram_module_v10 v10; struct atom_vram_module_v11 v11; + struct atom_vram_module_v3_0 v30; }; static int convert_atom_mem_type_to_vram_type(struct amdgpu_device *adev, @@ -186,13 +188,17 @@ static int convert_atom_mem_type_to_vram_type(struct amdgpu_device *adev, vram_type = AMDGPU_VRAM_TYPE_DDR3; break; case Ddr4MemType: - case LpDdr4MemType: vram_type = AMDGPU_VRAM_TYPE_DDR4; break; + case LpDdr4MemType: + vram_type = AMDGPU_VRAM_TYPE_LPDDR4; + break; case Ddr5MemType: - case LpDdr5MemType: vram_type = AMDGPU_VRAM_TYPE_DDR5; break; + case LpDdr5MemType: + vram_type = AMDGPU_VRAM_TYPE_LPDDR5; + break; default: vram_type = AMDGPU_VRAM_TYPE_UNKNOWN; break; @@ -294,88 +300,116 @@ amdgpu_atomfirmware_get_vram_info(struct amdgpu_device *adev, vram_info = (union vram_info *) (mode_info->atom_context->bios + data_offset); module_id = (RREG32(adev->bios_scratch_reg_offset + 4) & 0x00ff0000) >> 16; - switch (crev) { - case 3: - if (module_id > vram_info->v23.vram_module_num) - module_id = 0; - vram_module = (union vram_module *)vram_info->v23.vram_module; - while (i < module_id) { - vram_module = (union vram_module *) - ((u8 *)vram_module + vram_module->v9.vram_module_size); - i++; - } - mem_type = vram_module->v9.memory_type; - if (vram_type) - *vram_type = convert_atom_mem_type_to_vram_type(adev, mem_type); - mem_channel_number = vram_module->v9.channel_num; - mem_channel_width = vram_module->v9.channel_width; - if (vram_width) - *vram_width = mem_channel_number * (1 << mem_channel_width); - mem_vendor = (vram_module->v9.vender_rev_id) & 0xF; - if (vram_vendor) - *vram_vendor = mem_vendor; - break; - case 4: - if (module_id > vram_info->v24.vram_module_num) - module_id = 0; - vram_module = (union vram_module *)vram_info->v24.vram_module; - while (i < module_id) { - vram_module = (union vram_module *) - ((u8 *)vram_module + vram_module->v10.vram_module_size); - i++; - } - mem_type = vram_module->v10.memory_type; - if (vram_type) - *vram_type = convert_atom_mem_type_to_vram_type(adev, mem_type); - mem_channel_number = vram_module->v10.channel_num; - mem_channel_width = vram_module->v10.channel_width; - if (vram_width) - *vram_width = mem_channel_number * (1 << mem_channel_width); - mem_vendor = (vram_module->v10.vender_rev_id) & 0xF; - if (vram_vendor) - *vram_vendor = mem_vendor; - break; - case 5: - if (module_id > vram_info->v25.vram_module_num) - module_id = 0; - vram_module = (union vram_module *)vram_info->v25.vram_module; - while (i < module_id) { - vram_module = (union vram_module *) - ((u8 *)vram_module + vram_module->v11.vram_module_size); - i++; + if (frev == 3) { + switch (crev) { + /* v30 */ + case 0: + vram_module = (union vram_module *)vram_info->v30.vram_module; + mem_vendor = (vram_module->v30.dram_vendor_id) & 0xF; + if (vram_vendor) + *vram_vendor = mem_vendor; + mem_type = vram_info->v30.memory_type; + if (vram_type) + *vram_type = convert_atom_mem_type_to_vram_type(adev, mem_type); + mem_channel_number = vram_info->v30.channel_num; + mem_channel_width = vram_info->v30.channel_width; + if (vram_width) + *vram_width = mem_channel_number * (1 << mem_channel_width); + break; + default: + return -EINVAL; } - mem_type = vram_module->v11.memory_type; - if (vram_type) - *vram_type = convert_atom_mem_type_to_vram_type(adev, mem_type); - mem_channel_number = vram_module->v11.channel_num; - mem_channel_width = vram_module->v11.channel_width; - if (vram_width) - *vram_width = mem_channel_number * (1 << mem_channel_width); - mem_vendor = (vram_module->v11.vender_rev_id) & 0xF; - if (vram_vendor) - *vram_vendor = mem_vendor; - break; - case 6: - if (module_id > vram_info->v26.vram_module_num) - module_id = 0; - vram_module = (union vram_module *)vram_info->v26.vram_module; - while (i < module_id) { - vram_module = (union vram_module *) - ((u8 *)vram_module + vram_module->v9.vram_module_size); - i++; + } else if (frev == 2) { + switch (crev) { + /* v23 */ + case 3: + if (module_id > vram_info->v23.vram_module_num) + module_id = 0; + vram_module = (union vram_module *)vram_info->v23.vram_module; + while (i < module_id) { + vram_module = (union vram_module *) + ((u8 *)vram_module + vram_module->v9.vram_module_size); + i++; + } + mem_type = vram_module->v9.memory_type; + if (vram_type) + *vram_type = convert_atom_mem_type_to_vram_type(adev, mem_type); + mem_channel_number = vram_module->v9.channel_num; + mem_channel_width = vram_module->v9.channel_width; + if (vram_width) + *vram_width = mem_channel_number * (1 << mem_channel_width); + mem_vendor = (vram_module->v9.vender_rev_id) & 0xF; + if (vram_vendor) + *vram_vendor = mem_vendor; + break; + /* v24 */ + case 4: + if (module_id > vram_info->v24.vram_module_num) + module_id = 0; + vram_module = (union vram_module *)vram_info->v24.vram_module; + while (i < module_id) { + vram_module = (union vram_module *) + ((u8 *)vram_module + vram_module->v10.vram_module_size); + i++; + } + mem_type = vram_module->v10.memory_type; + if (vram_type) + *vram_type = convert_atom_mem_type_to_vram_type(adev, mem_type); + mem_channel_number = vram_module->v10.channel_num; + mem_channel_width = vram_module->v10.channel_width; + if (vram_width) + *vram_width = mem_channel_number * (1 << mem_channel_width); + mem_vendor = (vram_module->v10.vender_rev_id) & 0xF; + if (vram_vendor) + *vram_vendor = mem_vendor; + break; + /* v25 */ + case 5: + if (module_id > vram_info->v25.vram_module_num) + module_id = 0; + vram_module = (union vram_module *)vram_info->v25.vram_module; + while (i < module_id) { + vram_module = (union vram_module *) + ((u8 *)vram_module + vram_module->v11.vram_module_size); + i++; + } + mem_type = vram_module->v11.memory_type; + if (vram_type) + *vram_type = convert_atom_mem_type_to_vram_type(adev, mem_type); + mem_channel_number = vram_module->v11.channel_num; + mem_channel_width = vram_module->v11.channel_width; + if (vram_width) + *vram_width = mem_channel_number * (1 << mem_channel_width); + mem_vendor = (vram_module->v11.vender_rev_id) & 0xF; + if (vram_vendor) + *vram_vendor = mem_vendor; + break; + /* v26 */ + case 6: + if (module_id > vram_info->v26.vram_module_num) + module_id = 0; + vram_module = (union vram_module *)vram_info->v26.vram_module; + while (i < module_id) { + vram_module = (union vram_module *) + ((u8 *)vram_module + vram_module->v9.vram_module_size); + i++; + } + mem_type = vram_module->v9.memory_type; + if (vram_type) + *vram_type = convert_atom_mem_type_to_vram_type(adev, mem_type); + mem_channel_number = vram_module->v9.channel_num; + mem_channel_width = vram_module->v9.channel_width; + if (vram_width) + *vram_width = mem_channel_number * (1 << mem_channel_width); + mem_vendor = (vram_module->v9.vender_rev_id) & 0xF; + if (vram_vendor) + *vram_vendor = mem_vendor; + break; + default: + return -EINVAL; } - mem_type = vram_module->v9.memory_type; - if (vram_type) - *vram_type = convert_atom_mem_type_to_vram_type(adev, mem_type); - mem_channel_number = vram_module->v9.channel_num; - mem_channel_width = vram_module->v9.channel_width; - if (vram_width) - *vram_width = mem_channel_number * (1 << mem_channel_width); - mem_vendor = (vram_module->v9.vender_rev_id) & 0xF; - if (vram_vendor) - *vram_vendor = mem_vendor; - break; - default: + } else { + /* invalid frev */ return -EINVAL; } } @@ -526,6 +560,14 @@ bool amdgpu_atomfirmware_ras_rom_addr(struct amdgpu_device *adev, union smu_info { struct atom_smu_info_v3_1 v31; + struct atom_smu_info_v4_0 v40; +}; + +union gfx_info { + struct atom_gfx_info_v2_2 v22; + struct atom_gfx_info_v2_4 v24; + struct atom_gfx_info_v2_7 v27; + struct atom_gfx_info_v3_0 v30; }; int amdgpu_atomfirmware_get_clock_info(struct amdgpu_device *adev) @@ -565,7 +607,10 @@ int amdgpu_atomfirmware_get_clock_info(struct amdgpu_device *adev) data_offset); /* system clock */ - spll->reference_freq = le32_to_cpu(smu_info->v31.core_refclk_10khz); + if (frev == 3) + spll->reference_freq = le32_to_cpu(smu_info->v31.core_refclk_10khz); + else if (frev == 4) + spll->reference_freq = le32_to_cpu(smu_info->v40.core_refclk_10khz); spll->reference_div = 0; spll->min_post_div = 1; @@ -609,22 +654,26 @@ int amdgpu_atomfirmware_get_clock_info(struct amdgpu_device *adev) gfx_info); if (amdgpu_atom_parse_data_header(mode_info->atom_context, index, NULL, &frev, &crev, &data_offset)) { - struct atom_gfx_info_v2_2 *gfx_info = (struct atom_gfx_info_v2_2*) + union gfx_info *gfx_info = (union gfx_info *) (mode_info->atom_context->bios + data_offset); - if ((frev == 2) && (crev >= 2)) - spll->reference_freq = le32_to_cpu(gfx_info->rlc_gpu_timer_refclk); - ret = 0; + if ((frev == 3) || + (frev == 2 && crev == 6)) { + spll->reference_freq = le32_to_cpu(gfx_info->v30.golden_tsc_count_lower_refclk); + ret = 0; + } else if ((frev == 2) && + (crev >= 2) && + (crev != 6)) { + spll->reference_freq = le32_to_cpu(gfx_info->v22.rlc_gpu_timer_refclk); + ret = 0; + } else { + BUG(); + } } } return ret; } -union gfx_info { - struct atom_gfx_info_v2_4 v24; - struct atom_gfx_info_v2_7 v27; -}; - int amdgpu_atomfirmware_get_gfx_info(struct amdgpu_device *adev) { struct amdgpu_mode_info *mode_info = &adev->mode_info; @@ -638,42 +687,58 @@ int amdgpu_atomfirmware_get_gfx_info(struct amdgpu_device *adev) &frev, &crev, &data_offset)) { union gfx_info *gfx_info = (union gfx_info *) (mode_info->atom_context->bios + data_offset); - switch (crev) { - case 4: - adev->gfx.config.max_shader_engines = gfx_info->v24.max_shader_engines; - adev->gfx.config.max_cu_per_sh = gfx_info->v24.max_cu_per_sh; - adev->gfx.config.max_sh_per_se = gfx_info->v24.max_sh_per_se; - adev->gfx.config.max_backends_per_se = gfx_info->v24.max_backends_per_se; - adev->gfx.config.max_texture_channel_caches = gfx_info->v24.max_texture_channel_caches; - adev->gfx.config.max_gprs = le16_to_cpu(gfx_info->v24.gc_num_gprs); - adev->gfx.config.max_gs_threads = gfx_info->v24.gc_num_max_gs_thds; - adev->gfx.config.gs_vgt_table_depth = gfx_info->v24.gc_gs_table_depth; - adev->gfx.config.gs_prim_buffer_depth = - le16_to_cpu(gfx_info->v24.gc_gsprim_buff_depth); - adev->gfx.config.double_offchip_lds_buf = - gfx_info->v24.gc_double_offchip_lds_buffer; - adev->gfx.cu_info.wave_front_size = le16_to_cpu(gfx_info->v24.gc_wave_size); - adev->gfx.cu_info.max_waves_per_simd = le16_to_cpu(gfx_info->v24.gc_max_waves_per_simd); - adev->gfx.cu_info.max_scratch_slots_per_cu = gfx_info->v24.gc_max_scratch_slots_per_cu; - adev->gfx.cu_info.lds_size = le16_to_cpu(gfx_info->v24.gc_lds_size); - return 0; - case 7: - adev->gfx.config.max_shader_engines = gfx_info->v27.max_shader_engines; - adev->gfx.config.max_cu_per_sh = gfx_info->v27.max_cu_per_sh; - adev->gfx.config.max_sh_per_se = gfx_info->v27.max_sh_per_se; - adev->gfx.config.max_backends_per_se = gfx_info->v27.max_backends_per_se; - adev->gfx.config.max_texture_channel_caches = gfx_info->v27.max_texture_channel_caches; - adev->gfx.config.max_gprs = le16_to_cpu(gfx_info->v27.gc_num_gprs); - adev->gfx.config.max_gs_threads = gfx_info->v27.gc_num_max_gs_thds; - adev->gfx.config.gs_vgt_table_depth = gfx_info->v27.gc_gs_table_depth; - adev->gfx.config.gs_prim_buffer_depth = le16_to_cpu(gfx_info->v27.gc_gsprim_buff_depth); - adev->gfx.config.double_offchip_lds_buf = gfx_info->v27.gc_double_offchip_lds_buffer; - adev->gfx.cu_info.wave_front_size = le16_to_cpu(gfx_info->v27.gc_wave_size); - adev->gfx.cu_info.max_waves_per_simd = le16_to_cpu(gfx_info->v27.gc_max_waves_per_simd); - adev->gfx.cu_info.max_scratch_slots_per_cu = gfx_info->v27.gc_max_scratch_slots_per_cu; - adev->gfx.cu_info.lds_size = le16_to_cpu(gfx_info->v27.gc_lds_size); - return 0; - default: + if (frev == 2) { + switch (crev) { + case 4: + adev->gfx.config.max_shader_engines = gfx_info->v24.max_shader_engines; + adev->gfx.config.max_cu_per_sh = gfx_info->v24.max_cu_per_sh; + adev->gfx.config.max_sh_per_se = gfx_info->v24.max_sh_per_se; + adev->gfx.config.max_backends_per_se = gfx_info->v24.max_backends_per_se; + adev->gfx.config.max_texture_channel_caches = gfx_info->v24.max_texture_channel_caches; + adev->gfx.config.max_gprs = le16_to_cpu(gfx_info->v24.gc_num_gprs); + adev->gfx.config.max_gs_threads = gfx_info->v24.gc_num_max_gs_thds; + adev->gfx.config.gs_vgt_table_depth = gfx_info->v24.gc_gs_table_depth; + adev->gfx.config.gs_prim_buffer_depth = + le16_to_cpu(gfx_info->v24.gc_gsprim_buff_depth); + adev->gfx.config.double_offchip_lds_buf = + gfx_info->v24.gc_double_offchip_lds_buffer; + adev->gfx.cu_info.wave_front_size = le16_to_cpu(gfx_info->v24.gc_wave_size); + adev->gfx.cu_info.max_waves_per_simd = le16_to_cpu(gfx_info->v24.gc_max_waves_per_simd); + adev->gfx.cu_info.max_scratch_slots_per_cu = gfx_info->v24.gc_max_scratch_slots_per_cu; + adev->gfx.cu_info.lds_size = le16_to_cpu(gfx_info->v24.gc_lds_size); + return 0; + case 7: + adev->gfx.config.max_shader_engines = gfx_info->v27.max_shader_engines; + adev->gfx.config.max_cu_per_sh = gfx_info->v27.max_cu_per_sh; + adev->gfx.config.max_sh_per_se = gfx_info->v27.max_sh_per_se; + adev->gfx.config.max_backends_per_se = gfx_info->v27.max_backends_per_se; + adev->gfx.config.max_texture_channel_caches = gfx_info->v27.max_texture_channel_caches; + adev->gfx.config.max_gprs = le16_to_cpu(gfx_info->v27.gc_num_gprs); + adev->gfx.config.max_gs_threads = gfx_info->v27.gc_num_max_gs_thds; + adev->gfx.config.gs_vgt_table_depth = gfx_info->v27.gc_gs_table_depth; + adev->gfx.config.gs_prim_buffer_depth = le16_to_cpu(gfx_info->v27.gc_gsprim_buff_depth); + adev->gfx.config.double_offchip_lds_buf = gfx_info->v27.gc_double_offchip_lds_buffer; + adev->gfx.cu_info.wave_front_size = le16_to_cpu(gfx_info->v27.gc_wave_size); + adev->gfx.cu_info.max_waves_per_simd = le16_to_cpu(gfx_info->v27.gc_max_waves_per_simd); + adev->gfx.cu_info.max_scratch_slots_per_cu = gfx_info->v27.gc_max_scratch_slots_per_cu; + adev->gfx.cu_info.lds_size = le16_to_cpu(gfx_info->v27.gc_lds_size); + return 0; + default: + return -EINVAL; + } + } else if (frev == 3) { + switch (crev) { + case 0: + adev->gfx.config.max_shader_engines = gfx_info->v30.max_shader_engines; + adev->gfx.config.max_cu_per_sh = gfx_info->v30.max_cu_per_sh; + adev->gfx.config.max_sh_per_se = gfx_info->v30.max_sh_per_se; + adev->gfx.config.max_backends_per_se = gfx_info->v30.max_backends_per_se; + adev->gfx.config.max_texture_channel_caches = gfx_info->v30.max_texture_channel_caches; + return 0; + default: + return -EINVAL; + } + } else { return -EINVAL; } @@ -731,3 +796,67 @@ int amdgpu_atomfirmware_get_fw_reserved_fb_size(struct amdgpu_device *adev) return fw_reserved_fb_size; } + +/* + * Helper function to execute asic_init table + * + * @adev: amdgpu_device pointer + * @fb_reset: flag to indicate whether fb is reset or not + * + * Return 0 if succeed, otherwise failed + */ +int amdgpu_atomfirmware_asic_init(struct amdgpu_device *adev, bool fb_reset) +{ + struct amdgpu_mode_info *mode_info = &adev->mode_info; + struct atom_context *ctx; + uint8_t frev, crev; + uint16_t data_offset; + uint32_t bootup_sclk_in10khz, bootup_mclk_in10khz; + struct asic_init_ps_allocation_v2_1 asic_init_ps_v2_1; + int index; + + if (!mode_info) + return -EINVAL; + + ctx = mode_info->atom_context; + if (!ctx) + return -EINVAL; + + /* query bootup sclk/mclk from firmware_info table */ + index = get_index_into_master_table(atom_master_list_of_data_tables_v2_1, + firmwareinfo); + if (amdgpu_atom_parse_data_header(ctx, index, NULL, + &frev, &crev, &data_offset)) { + union firmware_info *firmware_info = + (union firmware_info *)(ctx->bios + + data_offset); + + bootup_sclk_in10khz = + le32_to_cpu(firmware_info->v31.bootup_sclk_in10khz); + bootup_mclk_in10khz = + le32_to_cpu(firmware_info->v31.bootup_mclk_in10khz); + } else { + return -EINVAL; + } + + index = get_index_into_master_table(atom_master_list_of_command_functions_v2_1, + asic_init); + if (amdgpu_atom_parse_cmd_header(mode_info->atom_context, index, &frev, &crev)) { + if (frev == 2 && crev >= 1) { + memset(&asic_init_ps_v2_1, 0, sizeof(asic_init_ps_v2_1)); + asic_init_ps_v2_1.param.engineparam.sclkfreqin10khz = bootup_sclk_in10khz; + asic_init_ps_v2_1.param.memparam.mclkfreqin10khz = bootup_mclk_in10khz; + asic_init_ps_v2_1.param.engineparam.engineflag = b3NORMAL_ENGINE_INIT; + if (!fb_reset) + asic_init_ps_v2_1.param.memparam.memflag = b3DRAM_SELF_REFRESH_EXIT; + else + asic_init_ps_v2_1.param.memparam.memflag = 0; + } else { + return -EINVAL; + } + } else { + return -EINVAL; + } + + return amdgpu_atom_execute_table(ctx, ATOM_CMD_INIT, (uint32_t *)&asic_init_ps_v2_1); +} diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_atomfirmware.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_atomfirmware.h index 751248b253de..c7eb2caec65a 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_atomfirmware.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_atomfirmware.h @@ -40,5 +40,6 @@ bool amdgpu_atomfirmware_ras_rom_addr(struct amdgpu_device *adev, uint8_t* i2c_a bool amdgpu_atomfirmware_mem_training_supported(struct amdgpu_device *adev); bool amdgpu_atomfirmware_dynamic_boot_config_supported(struct amdgpu_device *adev); int amdgpu_atomfirmware_get_fw_reserved_fb_size(struct amdgpu_device *adev); +int amdgpu_atomfirmware_asic_init(struct amdgpu_device *adev, bool fb_reset); #endif diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_benchmark.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_benchmark.c index 313517f7cf10..edc6377ec5ff 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_benchmark.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_benchmark.c @@ -29,14 +29,13 @@ #define AMDGPU_BENCHMARK_COMMON_MODES_N 17 static int amdgpu_benchmark_do_move(struct amdgpu_device *adev, unsigned size, - uint64_t saddr, uint64_t daddr, int n) + uint64_t saddr, uint64_t daddr, int n, s64 *time_ms) { - unsigned long start_jiffies; - unsigned long end_jiffies; + ktime_t stime, etime; struct dma_fence *fence; int i, r; - start_jiffies = jiffies; + stime = ktime_get(); for (i = 0; i < n; i++) { struct amdgpu_ring *ring = adev->mman.buffer_funcs_ring; r = amdgpu_copy_buffer(ring, saddr, daddr, size, NULL, &fence, @@ -48,120 +47,81 @@ static int amdgpu_benchmark_do_move(struct amdgpu_device *adev, unsigned size, if (r) goto exit_do_move; } - end_jiffies = jiffies; - r = jiffies_to_msecs(end_jiffies - start_jiffies); exit_do_move: + etime = ktime_get(); + *time_ms = ktime_ms_delta(etime, stime); + return r; } -static void amdgpu_benchmark_log_results(int n, unsigned size, - unsigned int time, +static void amdgpu_benchmark_log_results(struct amdgpu_device *adev, + int n, unsigned size, + s64 time_ms, unsigned sdomain, unsigned ddomain, char *kind) { - unsigned int throughput = (n * (size >> 10)) / time; - DRM_INFO("amdgpu: %s %u bo moves of %u kB from" - " %d to %d in %u ms, throughput: %u Mb/s or %u MB/s\n", - kind, n, size >> 10, sdomain, ddomain, time, + s64 throughput = (n * (size >> 10)); + + throughput = div64_s64(throughput, time_ms); + + dev_info(adev->dev, "amdgpu: %s %u bo moves of %u kB from" + " %d to %d in %lld ms, throughput: %lld Mb/s or %lld MB/s\n", + kind, n, size >> 10, sdomain, ddomain, time_ms, throughput * 8, throughput); } -static void amdgpu_benchmark_move(struct amdgpu_device *adev, unsigned size, - unsigned sdomain, unsigned ddomain) +static int amdgpu_benchmark_move(struct amdgpu_device *adev, unsigned size, + unsigned sdomain, unsigned ddomain) { struct amdgpu_bo *dobj = NULL; struct amdgpu_bo *sobj = NULL; - struct amdgpu_bo_param bp; uint64_t saddr, daddr; + s64 time_ms; int r, n; - int time; - - memset(&bp, 0, sizeof(bp)); - bp.size = size; - bp.byte_align = PAGE_SIZE; - bp.domain = sdomain; - bp.flags = 0; - bp.type = ttm_bo_type_kernel; - bp.resv = NULL; - bp.bo_ptr_size = sizeof(struct amdgpu_bo); n = AMDGPU_BENCHMARK_ITERATIONS; - r = amdgpu_bo_create(adev, &bp, &sobj); - if (r) { - goto out_cleanup; - } - r = amdgpu_bo_reserve(sobj, false); - if (unlikely(r != 0)) - goto out_cleanup; - r = amdgpu_bo_pin(sobj, sdomain); - if (r) { - amdgpu_bo_unreserve(sobj); - goto out_cleanup; - } - r = amdgpu_ttm_alloc_gart(&sobj->tbo); - amdgpu_bo_unreserve(sobj); - if (r) { - goto out_cleanup; - } - saddr = amdgpu_bo_gpu_offset(sobj); - bp.domain = ddomain; - r = amdgpu_bo_create(adev, &bp, &dobj); - if (r) { - goto out_cleanup; - } - r = amdgpu_bo_reserve(dobj, false); - if (unlikely(r != 0)) - goto out_cleanup; - r = amdgpu_bo_pin(dobj, ddomain); - if (r) { - amdgpu_bo_unreserve(sobj); + + r = amdgpu_bo_create_kernel(adev, size, + PAGE_SIZE, sdomain, + &sobj, + &saddr, + NULL); + if (r) goto out_cleanup; - } - r = amdgpu_ttm_alloc_gart(&dobj->tbo); - amdgpu_bo_unreserve(dobj); - if (r) { + r = amdgpu_bo_create_kernel(adev, size, + PAGE_SIZE, ddomain, + &dobj, + &daddr, + NULL); + if (r) goto out_cleanup; - } - daddr = amdgpu_bo_gpu_offset(dobj); if (adev->mman.buffer_funcs) { - time = amdgpu_benchmark_do_move(adev, size, saddr, daddr, n); - if (time < 0) + r = amdgpu_benchmark_do_move(adev, size, saddr, daddr, n, &time_ms); + if (r) goto out_cleanup; - if (time > 0) - amdgpu_benchmark_log_results(n, size, time, + else + amdgpu_benchmark_log_results(adev, n, size, time_ms, sdomain, ddomain, "dma"); } out_cleanup: /* Check error value now. The value can be overwritten when clean up.*/ - if (r) { - DRM_ERROR("Error while benchmarking BO move.\n"); - } + if (r < 0) + dev_info(adev->dev, "Error while benchmarking BO move.\n"); - if (sobj) { - r = amdgpu_bo_reserve(sobj, true); - if (likely(r == 0)) { - amdgpu_bo_unpin(sobj); - amdgpu_bo_unreserve(sobj); - } - amdgpu_bo_unref(&sobj); - } - if (dobj) { - r = amdgpu_bo_reserve(dobj, true); - if (likely(r == 0)) { - amdgpu_bo_unpin(dobj); - amdgpu_bo_unreserve(dobj); - } - amdgpu_bo_unref(&dobj); - } + if (sobj) + amdgpu_bo_free_kernel(&sobj, &saddr, NULL); + if (dobj) + amdgpu_bo_free_kernel(&dobj, &daddr, NULL); + return r; } -void amdgpu_benchmark(struct amdgpu_device *adev, int test_number) +int amdgpu_benchmark(struct amdgpu_device *adev, int test_number) { - int i; + int i, r; static const int common_modes[AMDGPU_BENCHMARK_COMMON_MODES_N] = { 640 * 480 * 4, 720 * 480 * 4, @@ -182,63 +142,119 @@ void amdgpu_benchmark(struct amdgpu_device *adev, int test_number) 1920 * 1200 * 4 }; + mutex_lock(&adev->benchmark_mutex); switch (test_number) { case 1: + dev_info(adev->dev, + "benchmark test: %d (simple test, VRAM to GTT and GTT to VRAM)\n", + test_number); /* simple test, VRAM to GTT and GTT to VRAM */ - amdgpu_benchmark_move(adev, 1024*1024, AMDGPU_GEM_DOMAIN_GTT, - AMDGPU_GEM_DOMAIN_VRAM); - amdgpu_benchmark_move(adev, 1024*1024, AMDGPU_GEM_DOMAIN_VRAM, - AMDGPU_GEM_DOMAIN_GTT); + r = amdgpu_benchmark_move(adev, 1024*1024, AMDGPU_GEM_DOMAIN_GTT, + AMDGPU_GEM_DOMAIN_VRAM); + if (r) + goto done; + r = amdgpu_benchmark_move(adev, 1024*1024, AMDGPU_GEM_DOMAIN_VRAM, + AMDGPU_GEM_DOMAIN_GTT); + if (r) + goto done; break; case 2: + dev_info(adev->dev, + "benchmark test: %d (simple test, VRAM to VRAM)\n", + test_number); /* simple test, VRAM to VRAM */ - amdgpu_benchmark_move(adev, 1024*1024, AMDGPU_GEM_DOMAIN_VRAM, - AMDGPU_GEM_DOMAIN_VRAM); + r = amdgpu_benchmark_move(adev, 1024*1024, AMDGPU_GEM_DOMAIN_VRAM, + AMDGPU_GEM_DOMAIN_VRAM); + if (r) + goto done; break; case 3: + dev_info(adev->dev, + "benchmark test: %d (GTT to VRAM, buffer size sweep, powers of 2)\n", + test_number); /* GTT to VRAM, buffer size sweep, powers of 2 */ - for (i = 1; i <= 16384; i <<= 1) - amdgpu_benchmark_move(adev, i * AMDGPU_GPU_PAGE_SIZE, - AMDGPU_GEM_DOMAIN_GTT, - AMDGPU_GEM_DOMAIN_VRAM); + for (i = 1; i <= 16384; i <<= 1) { + r = amdgpu_benchmark_move(adev, i * AMDGPU_GPU_PAGE_SIZE, + AMDGPU_GEM_DOMAIN_GTT, + AMDGPU_GEM_DOMAIN_VRAM); + if (r) + goto done; + } break; case 4: + dev_info(adev->dev, + "benchmark test: %d (VRAM to GTT, buffer size sweep, powers of 2)\n", + test_number); /* VRAM to GTT, buffer size sweep, powers of 2 */ - for (i = 1; i <= 16384; i <<= 1) - amdgpu_benchmark_move(adev, i * AMDGPU_GPU_PAGE_SIZE, - AMDGPU_GEM_DOMAIN_VRAM, - AMDGPU_GEM_DOMAIN_GTT); + for (i = 1; i <= 16384; i <<= 1) { + r = amdgpu_benchmark_move(adev, i * AMDGPU_GPU_PAGE_SIZE, + AMDGPU_GEM_DOMAIN_VRAM, + AMDGPU_GEM_DOMAIN_GTT); + if (r) + goto done; + } break; case 5: + dev_info(adev->dev, + "benchmark test: %d (VRAM to VRAM, buffer size sweep, powers of 2)\n", + test_number); /* VRAM to VRAM, buffer size sweep, powers of 2 */ - for (i = 1; i <= 16384; i <<= 1) - amdgpu_benchmark_move(adev, i * AMDGPU_GPU_PAGE_SIZE, - AMDGPU_GEM_DOMAIN_VRAM, - AMDGPU_GEM_DOMAIN_VRAM); + for (i = 1; i <= 16384; i <<= 1) { + r = amdgpu_benchmark_move(adev, i * AMDGPU_GPU_PAGE_SIZE, + AMDGPU_GEM_DOMAIN_VRAM, + AMDGPU_GEM_DOMAIN_VRAM); + if (r) + goto done; + } break; case 6: + dev_info(adev->dev, + "benchmark test: %d (GTT to VRAM, buffer size sweep, common modes)\n", + test_number); /* GTT to VRAM, buffer size sweep, common modes */ - for (i = 0; i < AMDGPU_BENCHMARK_COMMON_MODES_N; i++) - amdgpu_benchmark_move(adev, common_modes[i], - AMDGPU_GEM_DOMAIN_GTT, - AMDGPU_GEM_DOMAIN_VRAM); + for (i = 0; i < AMDGPU_BENCHMARK_COMMON_MODES_N; i++) { + r = amdgpu_benchmark_move(adev, common_modes[i], + AMDGPU_GEM_DOMAIN_GTT, + AMDGPU_GEM_DOMAIN_VRAM); + if (r) + goto done; + } break; case 7: + dev_info(adev->dev, + "benchmark test: %d (VRAM to GTT, buffer size sweep, common modes)\n", + test_number); /* VRAM to GTT, buffer size sweep, common modes */ - for (i = 0; i < AMDGPU_BENCHMARK_COMMON_MODES_N; i++) - amdgpu_benchmark_move(adev, common_modes[i], - AMDGPU_GEM_DOMAIN_VRAM, - AMDGPU_GEM_DOMAIN_GTT); + for (i = 0; i < AMDGPU_BENCHMARK_COMMON_MODES_N; i++) { + r = amdgpu_benchmark_move(adev, common_modes[i], + AMDGPU_GEM_DOMAIN_VRAM, + AMDGPU_GEM_DOMAIN_GTT); + if (r) + goto done; + } break; case 8: + dev_info(adev->dev, + "benchmark test: %d (VRAM to VRAM, buffer size sweep, common modes)\n", + test_number); /* VRAM to VRAM, buffer size sweep, common modes */ - for (i = 0; i < AMDGPU_BENCHMARK_COMMON_MODES_N; i++) - amdgpu_benchmark_move(adev, common_modes[i], + for (i = 0; i < AMDGPU_BENCHMARK_COMMON_MODES_N; i++) { + r = amdgpu_benchmark_move(adev, common_modes[i], AMDGPU_GEM_DOMAIN_VRAM, AMDGPU_GEM_DOMAIN_VRAM); + if (r) + goto done; + } break; default: - DRM_ERROR("Unknown benchmark\n"); + dev_info(adev->dev, "Unknown benchmark %d\n", test_number); + r = -EINVAL; + break; } + +done: + mutex_unlock(&adev->benchmark_mutex); + + return r; } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_bios.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_bios.c index 27b19503773b..e363f56c72af 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_bios.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_bios.c @@ -464,3 +464,50 @@ success: adev->is_atom_fw = (adev->asic_type >= CHIP_VEGA10) ? true : false; return true; } + +/* helper function for soc15 and onwards to read bios from rom */ +bool amdgpu_soc15_read_bios_from_rom(struct amdgpu_device *adev, + u8 *bios, u32 length_bytes) +{ + u32 *dw_ptr; + u32 i, length_dw; + u32 rom_offset; + u32 rom_index_offset; + u32 rom_data_offset; + + if (bios == NULL) + return false; + if (length_bytes == 0) + return false; + /* APU vbios image is part of sbios image */ + if (adev->flags & AMD_IS_APU) + return false; + if (!adev->smuio.funcs || + !adev->smuio.funcs->get_rom_index_offset || + !adev->smuio.funcs->get_rom_data_offset) + return false; + + dw_ptr = (u32 *)bios; + length_dw = ALIGN(length_bytes, 4) / 4; + + rom_index_offset = + adev->smuio.funcs->get_rom_index_offset(adev); + rom_data_offset = + adev->smuio.funcs->get_rom_data_offset(adev); + + if (adev->nbio.funcs && + adev->nbio.funcs->get_rom_offset) { + rom_offset = adev->nbio.funcs->get_rom_offset(adev); + rom_offset = rom_offset << 17; + } else { + rom_offset = 0; + } + + /* set rom index to rom_offset */ + WREG32(rom_index_offset, rom_offset); + /* read out the rom data */ + for (i = 0; i < length_dw; i++) + dw_ptr[i] = RREG32(rom_data_offset); + + return true; +} diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_bo_list.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_bo_list.c index 714178f1b6c6..2168163aad2d 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_bo_list.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_bo_list.c @@ -40,7 +40,7 @@ static void amdgpu_bo_list_free_rcu(struct rcu_head *rcu) { struct amdgpu_bo_list *list = container_of(rcu, struct amdgpu_bo_list, rhead); - + mutex_destroy(&list->bo_list_mutex); kvfree(list); } @@ -136,6 +136,7 @@ int amdgpu_bo_list_create(struct amdgpu_device *adev, struct drm_file *filp, trace_amdgpu_cs_bo_status(list->num_entries, total_size); + mutex_init(&list->bo_list_mutex); *result = list; return 0; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_bo_list.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_bo_list.h index 044b41f0bfd9..9caea1688fc3 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_bo_list.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_bo_list.h @@ -34,7 +34,6 @@ struct amdgpu_fpriv; struct amdgpu_bo_list_entry { struct ttm_validate_buffer tv; struct amdgpu_bo_va *bo_va; - struct dma_fence_chain *chain; uint32_t priority; struct page **user_pages; bool user_invalidated; @@ -48,6 +47,10 @@ struct amdgpu_bo_list { struct amdgpu_bo *oa_obj; unsigned first_userptr; unsigned num_entries; + + /* Protect access during command submission. + */ + struct mutex bo_list_mutex; }; int amdgpu_bo_list_get(struct amdgpu_fpriv *fpriv, int id, diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_connectors.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_connectors.c index c16a2704ced6..491d4846fc02 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_connectors.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_connectors.c @@ -24,9 +24,9 @@ * Alex Deucher */ +#include <drm/display/drm_dp_helper.h> #include <drm/drm_edid.h> #include <drm/drm_fb_helper.h> -#include <drm/drm_dp_helper.h> #include <drm/drm_probe_helper.h> #include <drm/amdgpu_drm.h> #include "amdgpu.h" @@ -175,7 +175,7 @@ int amdgpu_connector_get_monitor_bpc(struct drm_connector *connector) /* Check if bpc is within clock limit. Try to degrade gracefully otherwise */ if ((bpc == 12) && (mode_clock * 3/2 > max_tmds_clock)) { - if ((connector->display_info.edid_hdmi_dc_modes & DRM_EDID_HDMI_DC_30) && + if ((connector->display_info.edid_hdmi_rgb444_dc_modes & DRM_EDID_HDMI_DC_30) && (mode_clock * 5/4 <= max_tmds_clock)) bpc = 10; else @@ -626,7 +626,7 @@ amdgpu_connector_fixup_lcd_native_mode(struct drm_encoder *encoder, if (mode->type & DRM_MODE_TYPE_PREFERRED) { if (mode->hdisplay != native_mode->hdisplay || mode->vdisplay != native_mode->vdisplay) - memcpy(native_mode, mode, sizeof(*mode)); + drm_mode_copy(native_mode, mode); } } @@ -635,7 +635,7 @@ amdgpu_connector_fixup_lcd_native_mode(struct drm_encoder *encoder, list_for_each_entry_safe(mode, t, &connector->probed_modes, head) { if (mode->hdisplay == native_mode->hdisplay && mode->vdisplay == native_mode->vdisplay) { - *native_mode = *mode; + drm_mode_copy(native_mode, mode); drm_mode_set_crtcinfo(native_mode, CRTC_INTERLACE_HALVE_V); DRM_DEBUG_KMS("Determined LVDS native mode details from EDID\n"); break; @@ -1674,10 +1674,12 @@ amdgpu_connector_add(struct amdgpu_device *adev, adev->mode_info.dither_property, AMDGPU_FMT_DITHER_DISABLE); - if (amdgpu_audio != 0) + if (amdgpu_audio != 0) { drm_object_attach_property(&amdgpu_connector->base.base, adev->mode_info.audio_property, AMDGPU_AUDIO_AUTO); + amdgpu_connector->audio = AMDGPU_AUDIO_AUTO; + } subpixel_order = SubPixelHorizontalRGB; connector->interlace_allowed = true; @@ -1799,6 +1801,7 @@ amdgpu_connector_add(struct amdgpu_device *adev, drm_object_attach_property(&amdgpu_connector->base.base, adev->mode_info.audio_property, AMDGPU_AUDIO_AUTO); + amdgpu_connector->audio = AMDGPU_AUDIO_AUTO; } drm_object_attach_property(&amdgpu_connector->base.base, adev->mode_info.dither_property, @@ -1852,6 +1855,7 @@ amdgpu_connector_add(struct amdgpu_device *adev, drm_object_attach_property(&amdgpu_connector->base.base, adev->mode_info.audio_property, AMDGPU_AUDIO_AUTO); + amdgpu_connector->audio = AMDGPU_AUDIO_AUTO; } drm_object_attach_property(&amdgpu_connector->base.base, adev->mode_info.dither_property, @@ -1902,6 +1906,7 @@ amdgpu_connector_add(struct amdgpu_device *adev, drm_object_attach_property(&amdgpu_connector->base.base, adev->mode_info.audio_property, AMDGPU_AUDIO_AUTO); + amdgpu_connector->audio = AMDGPU_AUDIO_AUTO; } drm_object_attach_property(&amdgpu_connector->base.base, adev->mode_info.dither_property, diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c index 0311d799a010..d371000a5727 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c @@ -32,15 +32,89 @@ #include <drm/amdgpu_drm.h> #include <drm/drm_syncobj.h> +#include "amdgpu_cs.h" #include "amdgpu.h" #include "amdgpu_trace.h" #include "amdgpu_gmc.h" #include "amdgpu_gem.h" #include "amdgpu_ras.h" -static int amdgpu_cs_user_fence_chunk(struct amdgpu_cs_parser *p, - struct drm_amdgpu_cs_chunk_fence *data, - uint32_t *offset) +static int amdgpu_cs_parser_init(struct amdgpu_cs_parser *p, + struct amdgpu_device *adev, + struct drm_file *filp, + union drm_amdgpu_cs *cs) +{ + struct amdgpu_fpriv *fpriv = filp->driver_priv; + + if (cs->in.num_chunks == 0) + return -EINVAL; + + memset(p, 0, sizeof(*p)); + p->adev = adev; + p->filp = filp; + + p->ctx = amdgpu_ctx_get(fpriv, cs->in.ctx_id); + if (!p->ctx) + return -EINVAL; + + if (atomic_read(&p->ctx->guilty)) { + amdgpu_ctx_put(p->ctx); + return -ECANCELED; + } + return 0; +} + +static int amdgpu_cs_job_idx(struct amdgpu_cs_parser *p, + struct drm_amdgpu_cs_chunk_ib *chunk_ib) +{ + struct drm_sched_entity *entity; + unsigned int i; + int r; + + r = amdgpu_ctx_get_entity(p->ctx, chunk_ib->ip_type, + chunk_ib->ip_instance, + chunk_ib->ring, &entity); + if (r) + return r; + + /* + * Abort if there is no run queue associated with this entity. + * Possibly because of disabled HW IP. + */ + if (entity->rq == NULL) + return -EINVAL; + + /* Check if we can add this IB to some existing job */ + for (i = 0; i < p->gang_size; ++i) + if (p->entities[i] == entity) + return i; + + /* If not increase the gang size if possible */ + if (i == AMDGPU_CS_GANG_SIZE) + return -EINVAL; + + p->entities[i] = entity; + p->gang_size = i + 1; + return i; +} + +static int amdgpu_cs_p1_ib(struct amdgpu_cs_parser *p, + struct drm_amdgpu_cs_chunk_ib *chunk_ib, + unsigned int *num_ibs) +{ + int r; + + r = amdgpu_cs_job_idx(p, chunk_ib); + if (r < 0) + return r; + + ++(num_ibs[r]); + return 0; +} + +static int amdgpu_cs_p1_user_fence(struct amdgpu_cs_parser *p, + struct drm_amdgpu_cs_chunk_fence *data, + uint32_t *offset) { struct drm_gem_object *gobj; struct amdgpu_bo *bo; @@ -54,8 +128,8 @@ static int amdgpu_cs_user_fence_chunk(struct amdgpu_cs_parser *p, bo = amdgpu_bo_ref(gem_to_amdgpu_bo(gobj)); p->uf_entry.priority = 0; p->uf_entry.tv.bo = &bo->tbo; - /* One for TTM and one for the CS job */ - p->uf_entry.tv.num_shared = 2; + /* One for TTM and two for the CS job */ + p->uf_entry.tv.num_shared = 3; drm_gem_object_put(gobj); @@ -79,11 +153,11 @@ error_unref: return r; } -static int amdgpu_cs_bo_handles_chunk(struct amdgpu_cs_parser *p, - struct drm_amdgpu_bo_list_in *data) +static int amdgpu_cs_p1_bo_handles(struct amdgpu_cs_parser *p, + struct drm_amdgpu_bo_list_in *data) { + struct drm_amdgpu_bo_list_entry *info; int r; - struct drm_amdgpu_bo_list_entry *info = NULL; r = amdgpu_bo_create_list_entry_array(data, &info); if (r) @@ -103,38 +177,25 @@ error_free: return r; } -static int amdgpu_cs_parser_init(struct amdgpu_cs_parser *p, union drm_amdgpu_cs *cs) +/* Copy the data from userspace and go over it the first time */ +static int amdgpu_cs_pass1(struct amdgpu_cs_parser *p, + union drm_amdgpu_cs *cs) { struct amdgpu_fpriv *fpriv = p->filp->driver_priv; + unsigned int num_ibs[AMDGPU_CS_GANG_SIZE] = { }; struct amdgpu_vm *vm = &fpriv->vm; uint64_t *chunk_array_user; uint64_t *chunk_array; - unsigned size, num_ibs = 0; uint32_t uf_offset = 0; - int i; + unsigned int size; int ret; + int i; - if (cs->in.num_chunks == 0) - return 0; - - chunk_array = kvmalloc_array(cs->in.num_chunks, sizeof(uint64_t), GFP_KERNEL); + chunk_array = kvmalloc_array(cs->in.num_chunks, sizeof(uint64_t), + GFP_KERNEL); if (!chunk_array) return -ENOMEM; - p->ctx = amdgpu_ctx_get(fpriv, cs->in.ctx_id); - if (!p->ctx) { - ret = -EINVAL; - goto free_chunk; - } - - mutex_lock(&p->ctx->lock); - - /* skip guilty context job */ - if (atomic_read(&p->ctx->guilty) == 1) { - ret = -ECANCELED; - goto free_chunk; - } - /* get chunks */ chunk_array_user = u64_to_user_ptr(cs->in.chunks); if (copy_from_user(chunk_array, chunk_array_user, @@ -169,7 +230,8 @@ static int amdgpu_cs_parser_init(struct amdgpu_cs_parser *p, union drm_amdgpu_cs size = p->chunks[i].length_dw; cdata = u64_to_user_ptr(user_chunk.chunk_data); - p->chunks[i].kdata = kvmalloc_array(size, sizeof(uint32_t), GFP_KERNEL); + p->chunks[i].kdata = kvmalloc_array(size, sizeof(uint32_t), + GFP_KERNEL); if (p->chunks[i].kdata == NULL) { ret = -ENOMEM; i--; @@ -181,36 +243,35 @@ static int amdgpu_cs_parser_init(struct amdgpu_cs_parser *p, union drm_amdgpu_cs goto free_partial_kdata; } + /* Assume the worst on the following checks */ + ret = -EINVAL; switch (p->chunks[i].chunk_id) { case AMDGPU_CHUNK_ID_IB: - ++num_ibs; + if (size < sizeof(struct drm_amdgpu_cs_chunk_ib)) + goto free_partial_kdata; + + ret = amdgpu_cs_p1_ib(p, p->chunks[i].kdata, num_ibs); + if (ret) + goto free_partial_kdata; break; case AMDGPU_CHUNK_ID_FENCE: - size = sizeof(struct drm_amdgpu_cs_chunk_fence); - if (p->chunks[i].length_dw * sizeof(uint32_t) < size) { - ret = -EINVAL; + if (size < sizeof(struct drm_amdgpu_cs_chunk_fence)) goto free_partial_kdata; - } - ret = amdgpu_cs_user_fence_chunk(p, p->chunks[i].kdata, - &uf_offset); + ret = amdgpu_cs_p1_user_fence(p, p->chunks[i].kdata, + &uf_offset); if (ret) goto free_partial_kdata; - break; case AMDGPU_CHUNK_ID_BO_HANDLES: - size = sizeof(struct drm_amdgpu_bo_list_in); - if (p->chunks[i].length_dw * sizeof(uint32_t) < size) { - ret = -EINVAL; + if (size < sizeof(struct drm_amdgpu_bo_list_in)) goto free_partial_kdata; - } - ret = amdgpu_cs_bo_handles_chunk(p, p->chunks[i].kdata); + ret = amdgpu_cs_p1_bo_handles(p, p->chunks[i].kdata); if (ret) goto free_partial_kdata; - break; case AMDGPU_CHUNK_ID_DEPENDENCIES: @@ -222,22 +283,32 @@ static int amdgpu_cs_parser_init(struct amdgpu_cs_parser *p, union drm_amdgpu_cs break; default: - ret = -EINVAL; goto free_partial_kdata; } } - ret = amdgpu_job_alloc(p->adev, num_ibs, &p->job, vm); - if (ret) - goto free_all_kdata; + if (!p->gang_size) + return -EINVAL; + + for (i = 0; i < p->gang_size; ++i) { + ret = amdgpu_job_alloc(p->adev, num_ibs[i], &p->jobs[i], vm); + if (ret) + goto free_all_kdata; + + ret = drm_sched_job_init(&p->jobs[i]->base, p->entities[i], + &fpriv->vm); + if (ret) + goto free_all_kdata; + } + p->gang_leader = p->jobs[p->gang_size - 1]; - if (p->ctx->vram_lost_counter != p->job->vram_lost_counter) { + if (p->ctx->vram_lost_counter != p->gang_leader->vram_lost_counter) { ret = -ECANCELED; goto free_all_kdata; } if (p->uf_entry.tv.bo) - p->job->uf_addr = uf_offset; + p->gang_leader->uf_addr = uf_offset; kvfree(chunk_array); /* Use this opportunity to fill in task info for the vm */ @@ -259,6 +330,297 @@ free_chunk: return ret; } +static int amdgpu_cs_p2_ib(struct amdgpu_cs_parser *p, + struct amdgpu_cs_chunk *chunk, + unsigned int *ce_preempt, + unsigned int *de_preempt) +{ + struct drm_amdgpu_cs_chunk_ib *chunk_ib = chunk->kdata; + struct amdgpu_fpriv *fpriv = p->filp->driver_priv; + struct amdgpu_vm *vm = &fpriv->vm; + struct amdgpu_ring *ring; + struct amdgpu_job *job; + struct amdgpu_ib *ib; + int r; + + r = amdgpu_cs_job_idx(p, chunk_ib); + if (r < 0) + return r; + + job = p->jobs[r]; + ring = amdgpu_job_ring(job); + ib = &job->ibs[job->num_ibs++]; + + /* MM engine doesn't support user fences */ + if (p->uf_entry.tv.bo && ring->funcs->no_user_fence) + return -EINVAL; + + if (chunk_ib->ip_type == AMDGPU_HW_IP_GFX && + chunk_ib->flags & AMDGPU_IB_FLAG_PREEMPT) { + if (chunk_ib->flags & AMDGPU_IB_FLAG_CE) + (*ce_preempt)++; + else + (*de_preempt)++; + + /* Each GFX command submit allows only 1 IB max + * preemptible for CE & DE */ + if (*ce_preempt > 1 || *de_preempt > 1) + return -EINVAL; + } + + if (chunk_ib->flags & AMDGPU_IB_FLAG_PREAMBLE) + job->preamble_status |= AMDGPU_PREAMBLE_IB_PRESENT; + + r = amdgpu_ib_get(p->adev, vm, ring->funcs->parse_cs ? + chunk_ib->ib_bytes : 0, + AMDGPU_IB_POOL_DELAYED, ib); + if (r) { + DRM_ERROR("Failed to get ib !\n"); + return r; + } + + ib->gpu_addr = chunk_ib->va_start; + ib->length_dw = chunk_ib->ib_bytes / 4; + ib->flags = chunk_ib->flags; + return 0; +} + +static int amdgpu_cs_p2_dependencies(struct amdgpu_cs_parser *p, + struct amdgpu_cs_chunk *chunk) +{ + struct drm_amdgpu_cs_chunk_dep *deps = chunk->kdata; + struct amdgpu_fpriv *fpriv = p->filp->driver_priv; + unsigned num_deps; + int i, r; + + num_deps = chunk->length_dw * 4 / + sizeof(struct drm_amdgpu_cs_chunk_dep); + + for (i = 0; i < num_deps; ++i) { + struct amdgpu_ctx *ctx; + struct drm_sched_entity *entity; + struct dma_fence *fence; + + ctx = amdgpu_ctx_get(fpriv, deps[i].ctx_id); + if (ctx == NULL) + return -EINVAL; + + r = amdgpu_ctx_get_entity(ctx, deps[i].ip_type, + deps[i].ip_instance, + deps[i].ring, &entity); + if (r) { + amdgpu_ctx_put(ctx); + return r; + } + + fence = amdgpu_ctx_get_fence(ctx, entity, deps[i].handle); + amdgpu_ctx_put(ctx); + + if (IS_ERR(fence)) + return PTR_ERR(fence); + else if (!fence) + continue; + + if (chunk->chunk_id == AMDGPU_CHUNK_ID_SCHEDULED_DEPENDENCIES) { + struct drm_sched_fence *s_fence; + struct dma_fence *old = fence; + + s_fence = to_drm_sched_fence(fence); + fence = dma_fence_get(&s_fence->scheduled); + dma_fence_put(old); + } + + r = amdgpu_sync_fence(&p->gang_leader->sync, fence); + dma_fence_put(fence); + if (r) + return r; + } + return 0; +} + +static int amdgpu_syncobj_lookup_and_add(struct amdgpu_cs_parser *p, + uint32_t handle, u64 point, + u64 flags) +{ + struct dma_fence *fence; + int r; + + r = drm_syncobj_find_fence(p->filp, handle, point, flags, &fence); + if (r) { + DRM_ERROR("syncobj %u failed to find fence @ %llu (%d)!\n", + handle, point, r); + return r; + } + + r = amdgpu_sync_fence(&p->gang_leader->sync, fence); + dma_fence_put(fence); + + return r; +} + +static int amdgpu_cs_p2_syncobj_in(struct amdgpu_cs_parser *p, + struct amdgpu_cs_chunk *chunk) +{ + struct drm_amdgpu_cs_chunk_sem *deps = chunk->kdata; + unsigned num_deps; + int i, r; + + num_deps = chunk->length_dw * 4 / + sizeof(struct drm_amdgpu_cs_chunk_sem); + for (i = 0; i < num_deps; ++i) { + r = amdgpu_syncobj_lookup_and_add(p, deps[i].handle, 0, 0); + if (r) + return r; + } + + return 0; +} + +static int amdgpu_cs_p2_syncobj_timeline_wait(struct amdgpu_cs_parser *p, + struct amdgpu_cs_chunk *chunk) +{ + struct drm_amdgpu_cs_chunk_syncobj *syncobj_deps = chunk->kdata; + unsigned num_deps; + int i, r; + + num_deps = chunk->length_dw * 4 / + sizeof(struct drm_amdgpu_cs_chunk_syncobj); + for (i = 0; i < num_deps; ++i) { + r = amdgpu_syncobj_lookup_and_add(p, syncobj_deps[i].handle, + syncobj_deps[i].point, + syncobj_deps[i].flags); + if (r) + return r; + } + + return 0; +} + +static int amdgpu_cs_p2_syncobj_out(struct amdgpu_cs_parser *p, + struct amdgpu_cs_chunk *chunk) +{ + struct drm_amdgpu_cs_chunk_sem *deps = chunk->kdata; + unsigned num_deps; + int i; + + num_deps = chunk->length_dw * 4 / + sizeof(struct drm_amdgpu_cs_chunk_sem); + + if (p->post_deps) + return -EINVAL; + + p->post_deps = kmalloc_array(num_deps, sizeof(*p->post_deps), + GFP_KERNEL); + p->num_post_deps = 0; + + if (!p->post_deps) + return -ENOMEM; + + + for (i = 0; i < num_deps; ++i) { + p->post_deps[i].syncobj = + drm_syncobj_find(p->filp, deps[i].handle); + if (!p->post_deps[i].syncobj) + return -EINVAL; + p->post_deps[i].chain = NULL; + p->post_deps[i].point = 0; + p->num_post_deps++; + } + + return 0; +} + +static int amdgpu_cs_p2_syncobj_timeline_signal(struct amdgpu_cs_parser *p, + struct amdgpu_cs_chunk *chunk) +{ + struct drm_amdgpu_cs_chunk_syncobj *syncobj_deps = chunk->kdata; + unsigned num_deps; + int i; + + num_deps = chunk->length_dw * 4 / + sizeof(struct drm_amdgpu_cs_chunk_syncobj); + + if (p->post_deps) + return -EINVAL; + + p->post_deps = kmalloc_array(num_deps, sizeof(*p->post_deps), + GFP_KERNEL); + p->num_post_deps = 0; + + if (!p->post_deps) + return -ENOMEM; + + for (i = 0; i < num_deps; ++i) { + struct amdgpu_cs_post_dep *dep = &p->post_deps[i]; + + dep->chain = NULL; + if (syncobj_deps[i].point) { + dep->chain = dma_fence_chain_alloc(); + if (!dep->chain) + return -ENOMEM; + } + + dep->syncobj = drm_syncobj_find(p->filp, + syncobj_deps[i].handle); + if (!dep->syncobj) { + dma_fence_chain_free(dep->chain); + return -EINVAL; + } + dep->point = syncobj_deps[i].point; + p->num_post_deps++; + } + + return 0; +} + +static int amdgpu_cs_pass2(struct amdgpu_cs_parser *p) +{ + unsigned int ce_preempt = 0, de_preempt = 0; + int i, r; + + for (i = 0; i < p->nchunks; ++i) { + struct amdgpu_cs_chunk *chunk; + + chunk = &p->chunks[i]; + + switch (chunk->chunk_id) { + case AMDGPU_CHUNK_ID_IB: + r = amdgpu_cs_p2_ib(p, chunk, &ce_preempt, &de_preempt); + if (r) + return r; + break; + case AMDGPU_CHUNK_ID_DEPENDENCIES: + case AMDGPU_CHUNK_ID_SCHEDULED_DEPENDENCIES: + r = amdgpu_cs_p2_dependencies(p, chunk); + if (r) + return r; + break; + case AMDGPU_CHUNK_ID_SYNCOBJ_IN: + r = amdgpu_cs_p2_syncobj_in(p, chunk); + if (r) + return r; + break; + case AMDGPU_CHUNK_ID_SYNCOBJ_OUT: + r = amdgpu_cs_p2_syncobj_out(p, chunk); + if (r) + return r; + break; + case AMDGPU_CHUNK_ID_SYNCOBJ_TIMELINE_WAIT: + r = amdgpu_cs_p2_syncobj_timeline_wait(p, chunk); + if (r) + return r; + break; + case AMDGPU_CHUNK_ID_SYNCOBJ_TIMELINE_SIGNAL: + r = amdgpu_cs_p2_syncobj_timeline_signal(p, chunk); + if (r) + return r; + break; + } + } + + return 0; +} + /* Convert microseconds to bytes. */ static u64 us_to_bytes(struct amdgpu_device *adev, s64 us) { @@ -298,7 +660,6 @@ static void amdgpu_cs_get_threshold_for_moves(struct amdgpu_device *adev, { s64 time_us, increment_us; u64 free_vram, total_vram, used_vram; - struct ttm_resource_manager *vram_man = ttm_manager_type(&adev->mman.bdev, TTM_PL_VRAM); /* Allow a maximum of 200 accumulated ms. This is basically per-IB * throttling. * @@ -315,7 +676,7 @@ static void amdgpu_cs_get_threshold_for_moves(struct amdgpu_device *adev, } total_vram = adev->gmc.real_vram_size - atomic64_read(&adev->vram_pin_size); - used_vram = amdgpu_vram_mgr_usage(vram_man); + used_vram = ttm_resource_manager_usage(&adev->mman.vram_mgr.manager); free_vram = used_vram >= total_vram ? 0 : total_vram - used_vram; spin_lock(&adev->mm_stats.lock); @@ -342,7 +703,7 @@ static void amdgpu_cs_get_threshold_for_moves(struct amdgpu_device *adev, if (free_vram >= 128 * 1024 * 1024 || free_vram >= total_vram / 8) { s64 min_us; - /* Be more aggresive on dGPUs. Try to fill a portion of free + /* Be more aggressive on dGPUs. Try to fill a portion of free * VRAM now. */ if (!(adev->flags & AMD_IS_APU)) @@ -362,7 +723,7 @@ static void amdgpu_cs_get_threshold_for_moves(struct amdgpu_device *adev, if (!amdgpu_gmc_vram_full_visible(&adev->gmc)) { u64 total_vis_vram = adev->gmc.visible_vram_size; u64 used_vis_vram = - amdgpu_vram_mgr_vis_usage(vram_man); + amdgpu_vram_mgr_vis_usage(&adev->mman.vram_mgr); if (used_vis_vram < total_vis_vram) { u64 free_vis_vram = total_vis_vram - used_vis_vram; @@ -495,9 +856,7 @@ static int amdgpu_cs_parser_bos(struct amdgpu_cs_parser *p, struct amdgpu_vm *vm = &fpriv->vm; struct amdgpu_bo_list_entry *e; struct list_head duplicates; - struct amdgpu_bo *gds; - struct amdgpu_bo *gws; - struct amdgpu_bo *oa; + unsigned int i; int r; INIT_LIST_HEAD(&p->validated); @@ -519,6 +878,8 @@ static int amdgpu_cs_parser_bos(struct amdgpu_cs_parser *p, return r; } + mutex_lock(&p->bo_list->bo_list_mutex); + /* One for TTM and one for the CS job */ amdgpu_bo_list_for_each_entry(e, p->bo_list) e->tv.num_shared = 2; @@ -545,14 +906,15 @@ static int amdgpu_cs_parser_bos(struct amdgpu_cs_parser *p, GFP_KERNEL | __GFP_ZERO); if (!e->user_pages) { DRM_ERROR("kvmalloc_array failure\n"); - return -ENOMEM; + r = -ENOMEM; + goto out_free_user_pages; } r = amdgpu_ttm_tt_get_user_pages(bo, e->user_pages); if (r) { kvfree(e->user_pages); e->user_pages = NULL; - return r; + goto out_free_user_pages; } for (i = 0; i < bo->tbo.ttm->num_pages; i++) { @@ -569,21 +931,13 @@ static int amdgpu_cs_parser_bos(struct amdgpu_cs_parser *p, if (unlikely(r != 0)) { if (r != -ERESTARTSYS) DRM_ERROR("ttm_eu_reserve_buffers failed.\n"); - goto out; + goto out_free_user_pages; } amdgpu_bo_list_for_each_entry(e, p->bo_list) { struct amdgpu_bo *bo = ttm_to_amdgpu_bo(e->tv.bo); e->bo_va = amdgpu_vm_bo_find(vm, bo); - - if (bo->tbo.base.dma_buf && !amdgpu_bo_explicit_sync(bo)) { - e->chain = dma_fence_chain_alloc(); - if (!e->chain) { - r = -ENOMEM; - goto error_validate; - } - } } amdgpu_cs_get_threshold_for_moves(p->adev, &p->bytes_moved_threshold, @@ -606,215 +960,160 @@ static int amdgpu_cs_parser_bos(struct amdgpu_cs_parser *p, if (r) goto error_validate; - amdgpu_cs_report_moved_bytes(p->adev, p->bytes_moved, - p->bytes_moved_vis); + if (p->uf_entry.tv.bo) { + struct amdgpu_bo *uf = ttm_to_amdgpu_bo(p->uf_entry.tv.bo); - gds = p->bo_list->gds_obj; - gws = p->bo_list->gws_obj; - oa = p->bo_list->oa_obj; + r = amdgpu_ttm_alloc_gart(&uf->tbo); + if (r) + goto error_validate; - if (gds) { - p->job->gds_base = amdgpu_bo_gpu_offset(gds) >> PAGE_SHIFT; - p->job->gds_size = amdgpu_bo_size(gds) >> PAGE_SHIFT; - } - if (gws) { - p->job->gws_base = amdgpu_bo_gpu_offset(gws) >> PAGE_SHIFT; - p->job->gws_size = amdgpu_bo_size(gws) >> PAGE_SHIFT; - } - if (oa) { - p->job->oa_base = amdgpu_bo_gpu_offset(oa) >> PAGE_SHIFT; - p->job->oa_size = amdgpu_bo_size(oa) >> PAGE_SHIFT; + p->gang_leader->uf_addr += amdgpu_bo_gpu_offset(uf); } - if (!r && p->uf_entry.tv.bo) { - struct amdgpu_bo *uf = ttm_to_amdgpu_bo(p->uf_entry.tv.bo); + amdgpu_cs_report_moved_bytes(p->adev, p->bytes_moved, + p->bytes_moved_vis); - r = amdgpu_ttm_alloc_gart(&uf->tbo); - p->job->uf_addr += amdgpu_bo_gpu_offset(uf); - } + for (i = 0; i < p->gang_size; ++i) + amdgpu_job_set_resources(p->jobs[i], p->bo_list->gds_obj, + p->bo_list->gws_obj, + p->bo_list->oa_obj); + return 0; error_validate: - if (r) { - amdgpu_bo_list_for_each_entry(e, p->bo_list) { - dma_fence_chain_free(e->chain); - e->chain = NULL; - } - ttm_eu_backoff_reservation(&p->ticket, &p->validated); + ttm_eu_backoff_reservation(&p->ticket, &p->validated); + +out_free_user_pages: + amdgpu_bo_list_for_each_userptr_entry(e, p->bo_list) { + struct amdgpu_bo *bo = ttm_to_amdgpu_bo(e->tv.bo); + + if (!e->user_pages) + continue; + amdgpu_ttm_tt_get_user_pages_done(bo->tbo.ttm); + kvfree(e->user_pages); + e->user_pages = NULL; } -out: + mutex_unlock(&p->bo_list->bo_list_mutex); return r; } -static int amdgpu_cs_sync_rings(struct amdgpu_cs_parser *p) +static void trace_amdgpu_cs_ibs(struct amdgpu_cs_parser *p) { - struct amdgpu_fpriv *fpriv = p->filp->driver_priv; - struct amdgpu_bo_list_entry *e; - int r; + int i, j; - list_for_each_entry(e, &p->validated, tv.head) { - struct amdgpu_bo *bo = ttm_to_amdgpu_bo(e->tv.bo); - struct dma_resv *resv = bo->tbo.base.resv; - enum amdgpu_sync_mode sync_mode; + if (!trace_amdgpu_cs_enabled()) + return; - sync_mode = amdgpu_bo_explicit_sync(bo) ? - AMDGPU_SYNC_EXPLICIT : AMDGPU_SYNC_NE_OWNER; - r = amdgpu_sync_resv(p->adev, &p->job->sync, resv, sync_mode, - &fpriv->vm); - if (r) - return r; + for (i = 0; i < p->gang_size; ++i) { + struct amdgpu_job *job = p->jobs[i]; + + for (j = 0; j < job->num_ibs; ++j) + trace_amdgpu_cs(p, job, &job->ibs[j]); } - return 0; } -/** - * amdgpu_cs_parser_fini() - clean parser states - * @parser: parser structure holding parsing context. - * @error: error number - * @backoff: indicator to backoff the reservation - * - * If error is set then unvalidate buffer, otherwise just free memory - * used by parsing context. - **/ -static void amdgpu_cs_parser_fini(struct amdgpu_cs_parser *parser, int error, - bool backoff) +static int amdgpu_cs_patch_ibs(struct amdgpu_cs_parser *p, + struct amdgpu_job *job) { - unsigned i; + struct amdgpu_ring *ring = amdgpu_job_ring(job); + unsigned int i; + int r; + + /* Only for UVD/VCE VM emulation */ + if (!ring->funcs->parse_cs && !ring->funcs->patch_cs_in_place) + return 0; - if (error && backoff) { - struct amdgpu_bo_list_entry *e; + for (i = 0; i < job->num_ibs; ++i) { + struct amdgpu_ib *ib = &job->ibs[i]; + struct amdgpu_bo_va_mapping *m; + struct amdgpu_bo *aobj; + uint64_t va_start; + uint8_t *kptr; - amdgpu_bo_list_for_each_entry(e, parser->bo_list) { - dma_fence_chain_free(e->chain); - e->chain = NULL; + va_start = ib->gpu_addr & AMDGPU_GMC_HOLE_MASK; + r = amdgpu_cs_find_mapping(p, va_start, &aobj, &m); + if (r) { + DRM_ERROR("IB va_start is invalid\n"); + return r; } - ttm_eu_backoff_reservation(&parser->ticket, - &parser->validated); - } + if ((va_start + ib->length_dw * 4) > + (m->last + 1) * AMDGPU_GPU_PAGE_SIZE) { + DRM_ERROR("IB va_start+ib_bytes is invalid\n"); + return -EINVAL; + } - for (i = 0; i < parser->num_post_deps; i++) { - drm_syncobj_put(parser->post_deps[i].syncobj); - kfree(parser->post_deps[i].chain); - } - kfree(parser->post_deps); + /* the IB should be reserved at this point */ + r = amdgpu_bo_kmap(aobj, (void **)&kptr); + if (r) { + return r; + } - dma_fence_put(parser->fence); + kptr += va_start - (m->start * AMDGPU_GPU_PAGE_SIZE); - if (parser->ctx) { - mutex_unlock(&parser->ctx->lock); - amdgpu_ctx_put(parser->ctx); + if (ring->funcs->parse_cs) { + memcpy(ib->ptr, kptr, ib->length_dw * 4); + amdgpu_bo_kunmap(aobj); + + r = amdgpu_ring_parse_cs(ring, p, job, ib); + if (r) + return r; + } else { + ib->ptr = (uint32_t *)kptr; + r = amdgpu_ring_patch_cs_in_place(ring, p, job, ib); + amdgpu_bo_kunmap(aobj); + if (r) + return r; + } } - if (parser->bo_list) - amdgpu_bo_list_put(parser->bo_list); - for (i = 0; i < parser->nchunks; i++) - kvfree(parser->chunks[i].kdata); - kvfree(parser->chunks); - if (parser->job) - amdgpu_job_free(parser->job); - if (parser->uf_entry.tv.bo) { - struct amdgpu_bo *uf = ttm_to_amdgpu_bo(parser->uf_entry.tv.bo); + return 0; +} - amdgpu_bo_unref(&uf); +static int amdgpu_cs_patch_jobs(struct amdgpu_cs_parser *p) +{ + unsigned int i; + int r; + + for (i = 0; i < p->gang_size; ++i) { + r = amdgpu_cs_patch_ibs(p, p->jobs[i]); + if (r) + return r; } + return 0; } static int amdgpu_cs_vm_handling(struct amdgpu_cs_parser *p) { - struct amdgpu_ring *ring = to_amdgpu_ring(p->entity->rq->sched); struct amdgpu_fpriv *fpriv = p->filp->driver_priv; + struct amdgpu_job *job = p->gang_leader; struct amdgpu_device *adev = p->adev; struct amdgpu_vm *vm = &fpriv->vm; struct amdgpu_bo_list_entry *e; struct amdgpu_bo_va *bo_va; struct amdgpu_bo *bo; + unsigned int i; int r; - /* Only for UVD/VCE VM emulation */ - if (ring->funcs->parse_cs || ring->funcs->patch_cs_in_place) { - unsigned i, j; - - for (i = 0, j = 0; i < p->nchunks && j < p->job->num_ibs; i++) { - struct drm_amdgpu_cs_chunk_ib *chunk_ib; - struct amdgpu_bo_va_mapping *m; - struct amdgpu_bo *aobj = NULL; - struct amdgpu_cs_chunk *chunk; - uint64_t offset, va_start; - struct amdgpu_ib *ib; - uint8_t *kptr; - - chunk = &p->chunks[i]; - ib = &p->job->ibs[j]; - chunk_ib = chunk->kdata; - - if (chunk->chunk_id != AMDGPU_CHUNK_ID_IB) - continue; - - va_start = chunk_ib->va_start & AMDGPU_GMC_HOLE_MASK; - r = amdgpu_cs_find_mapping(p, va_start, &aobj, &m); - if (r) { - DRM_ERROR("IB va_start is invalid\n"); - return r; - } - - if ((va_start + chunk_ib->ib_bytes) > - (m->last + 1) * AMDGPU_GPU_PAGE_SIZE) { - DRM_ERROR("IB va_start+ib_bytes is invalid\n"); - return -EINVAL; - } - - /* the IB should be reserved at this point */ - r = amdgpu_bo_kmap(aobj, (void **)&kptr); - if (r) { - return r; - } - - offset = m->start * AMDGPU_GPU_PAGE_SIZE; - kptr += va_start - offset; - - if (ring->funcs->parse_cs) { - memcpy(ib->ptr, kptr, chunk_ib->ib_bytes); - amdgpu_bo_kunmap(aobj); - - r = amdgpu_ring_parse_cs(ring, p, j); - if (r) - return r; - } else { - ib->ptr = (uint32_t *)kptr; - r = amdgpu_ring_patch_cs_in_place(ring, p, j); - amdgpu_bo_kunmap(aobj); - if (r) - return r; - } - - j++; - } - } - - if (!p->job->vm) - return amdgpu_cs_sync_rings(p); - - r = amdgpu_vm_clear_freed(adev, vm, NULL); if (r) return r; - r = amdgpu_vm_bo_update(adev, fpriv->prt_va, false, NULL); + r = amdgpu_vm_bo_update(adev, fpriv->prt_va, false); if (r) return r; - r = amdgpu_sync_vm_fence(&p->job->sync, fpriv->prt_va->last_pt_update); + r = amdgpu_sync_fence(&job->sync, fpriv->prt_va->last_pt_update); if (r) return r; - if (amdgpu_mcbp || amdgpu_sriov_vf(adev)) { + if (fpriv->csa_va) { bo_va = fpriv->csa_va; BUG_ON(!bo_va); - r = amdgpu_vm_bo_update(adev, bo_va, false, NULL); + r = amdgpu_vm_bo_update(adev, bo_va, false); if (r) return r; - r = amdgpu_sync_vm_fence(&p->job->sync, bo_va->last_pt_update); + r = amdgpu_sync_fence(&job->sync, bo_va->last_pt_update); if (r) return r; } @@ -829,11 +1128,11 @@ static int amdgpu_cs_vm_handling(struct amdgpu_cs_parser *p) if (bo_va == NULL) continue; - r = amdgpu_vm_bo_update(adev, bo_va, false, NULL); + r = amdgpu_vm_bo_update(adev, bo_va, false); if (r) return r; - r = amdgpu_sync_vm_fence(&p->job->sync, bo_va->last_pt_update); + r = amdgpu_sync_fence(&job->sync, bo_va->last_pt_update); if (r) return r; } @@ -846,11 +1145,18 @@ static int amdgpu_cs_vm_handling(struct amdgpu_cs_parser *p) if (r) return r; - r = amdgpu_sync_vm_fence(&p->job->sync, vm->last_update); + r = amdgpu_sync_fence(&job->sync, vm->last_update); if (r) return r; - p->job->vm_pd_addr = amdgpu_gmc_pd_addr(vm->root.bo); + for (i = 0; i < p->gang_size; ++i) { + job = p->jobs[i]; + + if (!job->vm) + continue; + + job->vm_pd_addr = amdgpu_gmc_pd_addr(vm->root.bo); + } if (amdgpu_vm_debug) { /* Invalidate all BOs to test for userspace bugs */ @@ -865,327 +1171,41 @@ static int amdgpu_cs_vm_handling(struct amdgpu_cs_parser *p) } } - return amdgpu_cs_sync_rings(p); -} - -static int amdgpu_cs_ib_fill(struct amdgpu_device *adev, - struct amdgpu_cs_parser *parser) -{ - struct amdgpu_fpriv *fpriv = parser->filp->driver_priv; - struct amdgpu_vm *vm = &fpriv->vm; - int r, ce_preempt = 0, de_preempt = 0; - struct amdgpu_ring *ring; - int i, j; - - for (i = 0, j = 0; i < parser->nchunks && j < parser->job->num_ibs; i++) { - struct amdgpu_cs_chunk *chunk; - struct amdgpu_ib *ib; - struct drm_amdgpu_cs_chunk_ib *chunk_ib; - struct drm_sched_entity *entity; - - chunk = &parser->chunks[i]; - ib = &parser->job->ibs[j]; - chunk_ib = (struct drm_amdgpu_cs_chunk_ib *)chunk->kdata; - - if (chunk->chunk_id != AMDGPU_CHUNK_ID_IB) - continue; - - if (chunk_ib->ip_type == AMDGPU_HW_IP_GFX && - (amdgpu_mcbp || amdgpu_sriov_vf(adev))) { - if (chunk_ib->flags & AMDGPU_IB_FLAG_PREEMPT) { - if (chunk_ib->flags & AMDGPU_IB_FLAG_CE) - ce_preempt++; - else - de_preempt++; - } - - /* each GFX command submit allows 0 or 1 IB preemptible for CE & DE */ - if (ce_preempt > 1 || de_preempt > 1) - return -EINVAL; - } - - r = amdgpu_ctx_get_entity(parser->ctx, chunk_ib->ip_type, - chunk_ib->ip_instance, chunk_ib->ring, - &entity); - if (r) - return r; - - if (chunk_ib->flags & AMDGPU_IB_FLAG_PREAMBLE) - parser->job->preamble_status |= - AMDGPU_PREAMBLE_IB_PRESENT; - - if (parser->entity && parser->entity != entity) - return -EINVAL; - - /* Return if there is no run queue associated with this entity. - * Possibly because of disabled HW IP*/ - if (entity->rq == NULL) - return -EINVAL; - - parser->entity = entity; - - ring = to_amdgpu_ring(entity->rq->sched); - r = amdgpu_ib_get(adev, vm, ring->funcs->parse_cs ? - chunk_ib->ib_bytes : 0, - AMDGPU_IB_POOL_DELAYED, ib); - if (r) { - DRM_ERROR("Failed to get ib !\n"); - return r; - } - - ib->gpu_addr = chunk_ib->va_start; - ib->length_dw = chunk_ib->ib_bytes / 4; - ib->flags = chunk_ib->flags; - - j++; - } - - /* MM engine doesn't support user fences */ - ring = to_amdgpu_ring(parser->entity->rq->sched); - if (parser->job->uf_addr && ring->funcs->no_user_fence) - return -EINVAL; - - return amdgpu_ctx_wait_prev_fence(parser->ctx, parser->entity); -} - -static int amdgpu_cs_process_fence_dep(struct amdgpu_cs_parser *p, - struct amdgpu_cs_chunk *chunk) -{ - struct amdgpu_fpriv *fpriv = p->filp->driver_priv; - unsigned num_deps; - int i, r; - struct drm_amdgpu_cs_chunk_dep *deps; - - deps = (struct drm_amdgpu_cs_chunk_dep *)chunk->kdata; - num_deps = chunk->length_dw * 4 / - sizeof(struct drm_amdgpu_cs_chunk_dep); - - for (i = 0; i < num_deps; ++i) { - struct amdgpu_ctx *ctx; - struct drm_sched_entity *entity; - struct dma_fence *fence; - - ctx = amdgpu_ctx_get(fpriv, deps[i].ctx_id); - if (ctx == NULL) - return -EINVAL; - - r = amdgpu_ctx_get_entity(ctx, deps[i].ip_type, - deps[i].ip_instance, - deps[i].ring, &entity); - if (r) { - amdgpu_ctx_put(ctx); - return r; - } - - fence = amdgpu_ctx_get_fence(ctx, entity, deps[i].handle); - amdgpu_ctx_put(ctx); - - if (IS_ERR(fence)) - return PTR_ERR(fence); - else if (!fence) - continue; - - if (chunk->chunk_id == AMDGPU_CHUNK_ID_SCHEDULED_DEPENDENCIES) { - struct drm_sched_fence *s_fence; - struct dma_fence *old = fence; - - s_fence = to_drm_sched_fence(fence); - fence = dma_fence_get(&s_fence->scheduled); - dma_fence_put(old); - } - - r = amdgpu_sync_fence(&p->job->sync, fence); - dma_fence_put(fence); - if (r) - return r; - } return 0; } -static int amdgpu_syncobj_lookup_and_add_to_sync(struct amdgpu_cs_parser *p, - uint32_t handle, u64 point, - u64 flags) +static int amdgpu_cs_sync_rings(struct amdgpu_cs_parser *p) { - struct dma_fence *fence; + struct amdgpu_fpriv *fpriv = p->filp->driver_priv; + struct amdgpu_job *leader = p->gang_leader; + struct amdgpu_bo_list_entry *e; + unsigned int i; int r; - r = drm_syncobj_find_fence(p->filp, handle, point, flags, &fence); - if (r) { - DRM_ERROR("syncobj %u failed to find fence @ %llu (%d)!\n", - handle, point, r); - return r; - } - - r = amdgpu_sync_fence(&p->job->sync, fence); - dma_fence_put(fence); - - return r; -} - -static int amdgpu_cs_process_syncobj_in_dep(struct amdgpu_cs_parser *p, - struct amdgpu_cs_chunk *chunk) -{ - struct drm_amdgpu_cs_chunk_sem *deps; - unsigned num_deps; - int i, r; + list_for_each_entry(e, &p->validated, tv.head) { + struct amdgpu_bo *bo = ttm_to_amdgpu_bo(e->tv.bo); + struct dma_resv *resv = bo->tbo.base.resv; + enum amdgpu_sync_mode sync_mode; - deps = (struct drm_amdgpu_cs_chunk_sem *)chunk->kdata; - num_deps = chunk->length_dw * 4 / - sizeof(struct drm_amdgpu_cs_chunk_sem); - for (i = 0; i < num_deps; ++i) { - r = amdgpu_syncobj_lookup_and_add_to_sync(p, deps[i].handle, - 0, 0); + sync_mode = amdgpu_bo_explicit_sync(bo) ? + AMDGPU_SYNC_EXPLICIT : AMDGPU_SYNC_NE_OWNER; + r = amdgpu_sync_resv(p->adev, &leader->sync, resv, sync_mode, + &fpriv->vm); if (r) return r; } - return 0; -} - - -static int amdgpu_cs_process_syncobj_timeline_in_dep(struct amdgpu_cs_parser *p, - struct amdgpu_cs_chunk *chunk) -{ - struct drm_amdgpu_cs_chunk_syncobj *syncobj_deps; - unsigned num_deps; - int i, r; - - syncobj_deps = (struct drm_amdgpu_cs_chunk_syncobj *)chunk->kdata; - num_deps = chunk->length_dw * 4 / - sizeof(struct drm_amdgpu_cs_chunk_syncobj); - for (i = 0; i < num_deps; ++i) { - r = amdgpu_syncobj_lookup_and_add_to_sync(p, - syncobj_deps[i].handle, - syncobj_deps[i].point, - syncobj_deps[i].flags); + for (i = 0; i < p->gang_size - 1; ++i) { + r = amdgpu_sync_clone(&leader->sync, &p->jobs[i]->sync); if (r) return r; } - return 0; -} - -static int amdgpu_cs_process_syncobj_out_dep(struct amdgpu_cs_parser *p, - struct amdgpu_cs_chunk *chunk) -{ - struct drm_amdgpu_cs_chunk_sem *deps; - unsigned num_deps; - int i; - - deps = (struct drm_amdgpu_cs_chunk_sem *)chunk->kdata; - num_deps = chunk->length_dw * 4 / - sizeof(struct drm_amdgpu_cs_chunk_sem); - - if (p->post_deps) - return -EINVAL; - - p->post_deps = kmalloc_array(num_deps, sizeof(*p->post_deps), - GFP_KERNEL); - p->num_post_deps = 0; - - if (!p->post_deps) - return -ENOMEM; - - - for (i = 0; i < num_deps; ++i) { - p->post_deps[i].syncobj = - drm_syncobj_find(p->filp, deps[i].handle); - if (!p->post_deps[i].syncobj) - return -EINVAL; - p->post_deps[i].chain = NULL; - p->post_deps[i].point = 0; - p->num_post_deps++; - } - - return 0; -} - - -static int amdgpu_cs_process_syncobj_timeline_out_dep(struct amdgpu_cs_parser *p, - struct amdgpu_cs_chunk *chunk) -{ - struct drm_amdgpu_cs_chunk_syncobj *syncobj_deps; - unsigned num_deps; - int i; - - syncobj_deps = (struct drm_amdgpu_cs_chunk_syncobj *)chunk->kdata; - num_deps = chunk->length_dw * 4 / - sizeof(struct drm_amdgpu_cs_chunk_syncobj); - - if (p->post_deps) - return -EINVAL; - - p->post_deps = kmalloc_array(num_deps, sizeof(*p->post_deps), - GFP_KERNEL); - p->num_post_deps = 0; - - if (!p->post_deps) - return -ENOMEM; - - for (i = 0; i < num_deps; ++i) { - struct amdgpu_cs_post_dep *dep = &p->post_deps[i]; - - dep->chain = NULL; - if (syncobj_deps[i].point) { - dep->chain = dma_fence_chain_alloc(); - if (!dep->chain) - return -ENOMEM; - } - - dep->syncobj = drm_syncobj_find(p->filp, - syncobj_deps[i].handle); - if (!dep->syncobj) { - dma_fence_chain_free(dep->chain); - return -EINVAL; - } - dep->point = syncobj_deps[i].point; - p->num_post_deps++; - } - - return 0; -} - -static int amdgpu_cs_dependencies(struct amdgpu_device *adev, - struct amdgpu_cs_parser *p) -{ - int i, r; - - for (i = 0; i < p->nchunks; ++i) { - struct amdgpu_cs_chunk *chunk; - - chunk = &p->chunks[i]; - - switch (chunk->chunk_id) { - case AMDGPU_CHUNK_ID_DEPENDENCIES: - case AMDGPU_CHUNK_ID_SCHEDULED_DEPENDENCIES: - r = amdgpu_cs_process_fence_dep(p, chunk); - if (r) - return r; - break; - case AMDGPU_CHUNK_ID_SYNCOBJ_IN: - r = amdgpu_cs_process_syncobj_in_dep(p, chunk); - if (r) - return r; - break; - case AMDGPU_CHUNK_ID_SYNCOBJ_OUT: - r = amdgpu_cs_process_syncobj_out_dep(p, chunk); - if (r) - return r; - break; - case AMDGPU_CHUNK_ID_SYNCOBJ_TIMELINE_WAIT: - r = amdgpu_cs_process_syncobj_timeline_in_dep(p, chunk); - if (r) - return r; - break; - case AMDGPU_CHUNK_ID_SYNCOBJ_TIMELINE_SIGNAL: - r = amdgpu_cs_process_syncobj_timeline_out_dep(p, chunk); - if (r) - return r; - break; - } - } + r = amdgpu_ctx_wait_prev_fence(p->ctx, p->entities[p->gang_size - 1]); + if (r && r != -ERESTARTSYS) + DRM_ERROR("amdgpu_ctx_wait_prev_fence failed.\n"); - return 0; + return r; } static void amdgpu_cs_post_dependencies(struct amdgpu_cs_parser *p) @@ -1209,20 +1229,28 @@ static int amdgpu_cs_submit(struct amdgpu_cs_parser *p, union drm_amdgpu_cs *cs) { struct amdgpu_fpriv *fpriv = p->filp->driver_priv; - struct drm_sched_entity *entity = p->entity; + struct amdgpu_job *leader = p->gang_leader; struct amdgpu_bo_list_entry *e; - struct amdgpu_job *job; + unsigned int i; uint64_t seq; int r; - job = p->job; - p->job = NULL; + for (i = 0; i < p->gang_size; ++i) + drm_sched_job_arm(&p->jobs[i]->base); - r = drm_sched_job_init(&job->base, entity, &fpriv->vm); - if (r) - goto error_unlock; + for (i = 0; i < (p->gang_size - 1); ++i) { + struct dma_fence *fence; - drm_sched_job_arm(&job->base); + fence = &p->jobs[i]->base.s_fence->scheduled; + r = amdgpu_sync_fence(&leader->sync, fence); + if (r) + goto error_cleanup; + } + + if (p->gang_size > 1) { + for (i = 0; i < p->gang_size; ++i) + amdgpu_job_set_gang_leader(p->jobs[i], leader); + } /* No memory allocation is allowed while holding the notifier lock. * The lock is held until amdgpu_cs_submit is finished and fence is @@ -1233,6 +1261,7 @@ static int amdgpu_cs_submit(struct amdgpu_cs_parser *p, /* If userptr are invalidated after amdgpu_cs_parser_bos(), return * -EAGAIN, drmIoctl in libdrm will restart the amdgpu_cs_ioctl. */ + r = 0; amdgpu_bo_list_for_each_userptr_entry(e, p->bo_list) { struct amdgpu_bo *bo = ttm_to_amdgpu_bo(e->tv.bo); @@ -1240,84 +1269,96 @@ static int amdgpu_cs_submit(struct amdgpu_cs_parser *p, } if (r) { r = -EAGAIN; - goto error_abort; + goto error_unlock; } - p->fence = dma_fence_get(&job->base.s_fence->finished); + p->fence = dma_fence_get(&leader->base.s_fence->finished); + list_for_each_entry(e, &p->validated, tv.head) { + + /* Everybody except for the gang leader uses READ */ + for (i = 0; i < (p->gang_size - 1); ++i) { + dma_resv_add_fence(e->tv.bo->base.resv, + &p->jobs[i]->base.s_fence->finished, + DMA_RESV_USAGE_READ); + } + + /* The gang leader is remembered as writer */ + e->tv.num_shared = 0; + } - amdgpu_ctx_add_fence(p->ctx, entity, p->fence, &seq); + seq = amdgpu_ctx_add_fence(p->ctx, p->entities[p->gang_size - 1], + p->fence); amdgpu_cs_post_dependencies(p); - if ((job->preamble_status & AMDGPU_PREAMBLE_IB_PRESENT) && + if ((leader->preamble_status & AMDGPU_PREAMBLE_IB_PRESENT) && !p->ctx->preamble_presented) { - job->preamble_status |= AMDGPU_PREAMBLE_IB_PRESENT_FIRST; + leader->preamble_status |= AMDGPU_PREAMBLE_IB_PRESENT_FIRST; p->ctx->preamble_presented = true; } cs->out.handle = seq; - job->uf_sequence = seq; - - amdgpu_job_free_resources(job); + leader->uf_sequence = seq; - trace_amdgpu_cs_ioctl(job); amdgpu_vm_bo_trace_cs(&fpriv->vm, &p->ticket); - drm_sched_entity_push_job(&job->base); - - amdgpu_vm_move_to_lru_tail(p->adev, &fpriv->vm); - - amdgpu_bo_list_for_each_entry(e, p->bo_list) { - struct dma_resv *resv = e->tv.bo->base.resv; - struct dma_fence_chain *chain = e->chain; - - if (!chain) - continue; - - /* - * Work around dma_resv shortcommings by wrapping up the - * submission in a dma_fence_chain and add it as exclusive - * fence, but first add the submission as shared fence to make - * sure that shared fences never signal before the exclusive - * one. - */ - dma_fence_chain_init(chain, dma_resv_excl_fence(resv), - dma_fence_get(p->fence), 1); - - dma_resv_add_shared_fence(resv, p->fence); - rcu_assign_pointer(resv->fence_excl, &chain->base); - e->chain = NULL; + for (i = 0; i < p->gang_size; ++i) { + amdgpu_job_free_resources(p->jobs[i]); + trace_amdgpu_cs_ioctl(p->jobs[i]); + drm_sched_entity_push_job(&p->jobs[i]->base); + p->jobs[i] = NULL; } + amdgpu_vm_move_to_lru_tail(p->adev, &fpriv->vm); ttm_eu_fence_buffer_objects(&p->ticket, &p->validated, p->fence); - mutex_unlock(&p->adev->notifier_lock); + mutex_unlock(&p->adev->notifier_lock); + mutex_unlock(&p->bo_list->bo_list_mutex); return 0; -error_abort: - drm_sched_job_cleanup(&job->base); +error_unlock: mutex_unlock(&p->adev->notifier_lock); -error_unlock: - amdgpu_job_free(job); +error_cleanup: + for (i = 0; i < p->gang_size; ++i) + drm_sched_job_cleanup(&p->jobs[i]->base); return r; } -static void trace_amdgpu_cs_ibs(struct amdgpu_cs_parser *parser) +/* Cleanup the parser structure */ +static void amdgpu_cs_parser_fini(struct amdgpu_cs_parser *parser) { - int i; + unsigned i; - if (!trace_amdgpu_cs_enabled()) - return; + for (i = 0; i < parser->num_post_deps; i++) { + drm_syncobj_put(parser->post_deps[i].syncobj); + kfree(parser->post_deps[i].chain); + } + kfree(parser->post_deps); - for (i = 0; i < parser->job->num_ibs; i++) - trace_amdgpu_cs(parser, i); + dma_fence_put(parser->fence); + + if (parser->ctx) + amdgpu_ctx_put(parser->ctx); + if (parser->bo_list) + amdgpu_bo_list_put(parser->bo_list); + + for (i = 0; i < parser->nchunks; i++) + kvfree(parser->chunks[i].kdata); + kvfree(parser->chunks); + for (i = 0; i < parser->gang_size; ++i) { + if (parser->jobs[i]) + amdgpu_job_free(parser->jobs[i]); + } + if (parser->uf_entry.tv.bo) { + struct amdgpu_bo *uf = ttm_to_amdgpu_bo(parser->uf_entry.tv.bo); + + amdgpu_bo_unref(&uf); + } } int amdgpu_cs_ioctl(struct drm_device *dev, void *data, struct drm_file *filp) { struct amdgpu_device *adev = drm_to_adev(dev); - union drm_amdgpu_cs *cs = data; - struct amdgpu_cs_parser parser = {}; - bool reserved_buffers = false; + struct amdgpu_cs_parser parser; int r; if (amdgpu_ras_intr_triggered()) @@ -1326,25 +1367,20 @@ int amdgpu_cs_ioctl(struct drm_device *dev, void *data, struct drm_file *filp) if (!adev->accel_working) return -EBUSY; - parser.adev = adev; - parser.filp = filp; - - r = amdgpu_cs_parser_init(&parser, data); + r = amdgpu_cs_parser_init(&parser, adev, filp, data); if (r) { if (printk_ratelimit()) DRM_ERROR("Failed to initialize parser %d!\n", r); - goto out; + return r; } - r = amdgpu_cs_ib_fill(adev, &parser); + r = amdgpu_cs_pass1(&parser, data); if (r) - goto out; + goto error_fini; - r = amdgpu_cs_dependencies(adev, &parser); - if (r) { - DRM_ERROR("Failed in the dependencies handling %d!\n", r); - goto out; - } + r = amdgpu_cs_pass2(&parser); + if (r) + goto error_fini; r = amdgpu_cs_parser_bos(&parser, data); if (r) { @@ -1352,22 +1388,36 @@ int amdgpu_cs_ioctl(struct drm_device *dev, void *data, struct drm_file *filp) DRM_ERROR("Not enough memory for command submission!\n"); else if (r != -ERESTARTSYS && r != -EAGAIN) DRM_ERROR("Failed to process the buffer list %d!\n", r); - goto out; + goto error_fini; } - reserved_buffers = true; + r = amdgpu_cs_patch_jobs(&parser); + if (r) + goto error_backoff; + + r = amdgpu_cs_vm_handling(&parser); + if (r) + goto error_backoff; + + r = amdgpu_cs_sync_rings(&parser); + if (r) + goto error_backoff; trace_amdgpu_cs_ibs(&parser); - r = amdgpu_cs_vm_handling(&parser); + r = amdgpu_cs_submit(&parser, data); if (r) - goto out; + goto error_backoff; - r = amdgpu_cs_submit(&parser, cs); + amdgpu_cs_parser_fini(&parser); + return 0; -out: - amdgpu_cs_parser_fini(&parser, r, reserved_buffers); +error_backoff: + ttm_eu_backoff_reservation(&parser.ticket, &parser.validated); + mutex_unlock(&parser.bo_list->bo_list_mutex); +error_fini: + amdgpu_cs_parser_fini(&parser); return r; } @@ -1510,6 +1560,7 @@ int amdgpu_cs_fence_to_handle_ioctl(struct drm_device *dev, void *data, return 0; default: + dma_fence_put(fence); return -EINVAL; } } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.h new file mode 100644 index 000000000000..cbaa19b2b8a3 --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.h @@ -0,0 +1,84 @@ +/* + * Copyright 2022 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ +#ifndef __AMDGPU_CS_H__ +#define __AMDGPU_CS_H__ + +#include "amdgpu_job.h" +#include "amdgpu_bo_list.h" +#include "amdgpu_ring.h" + +#define AMDGPU_CS_GANG_SIZE 4 + +struct amdgpu_bo_va_mapping; + +struct amdgpu_cs_chunk { + uint32_t chunk_id; + uint32_t length_dw; + void *kdata; +}; + +struct amdgpu_cs_post_dep { + struct drm_syncobj *syncobj; + struct dma_fence_chain *chain; + u64 point; +}; + +struct amdgpu_cs_parser { + struct amdgpu_device *adev; + struct drm_file *filp; + struct amdgpu_ctx *ctx; + + /* chunks */ + unsigned nchunks; + struct amdgpu_cs_chunk *chunks; + + /* scheduler job objects */ + unsigned int gang_size; + struct drm_sched_entity *entities[AMDGPU_CS_GANG_SIZE]; + struct amdgpu_job *jobs[AMDGPU_CS_GANG_SIZE]; + struct amdgpu_job *gang_leader; + + /* buffer objects */ + struct ww_acquire_ctx ticket; + struct amdgpu_bo_list *bo_list; + struct amdgpu_mn *mn; + struct amdgpu_bo_list_entry vm_pd; + struct list_head validated; + struct dma_fence *fence; + uint64_t bytes_moved_threshold; + uint64_t bytes_moved_vis_threshold; + uint64_t bytes_moved; + uint64_t bytes_moved_vis; + + /* user fence */ + struct amdgpu_bo_list_entry uf_entry; + + unsigned num_post_deps; + struct amdgpu_cs_post_dep *post_deps; +}; + +int amdgpu_cs_find_mapping(struct amdgpu_cs_parser *parser, + uint64_t addr, struct amdgpu_bo **bo, + struct amdgpu_bo_va_mapping **mapping); + +#endif diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_csa.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_csa.c index da21e60bb827..c6d4d41c4393 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_csa.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_csa.c @@ -98,7 +98,7 @@ int amdgpu_map_static_csa(struct amdgpu_device *adev, struct amdgpu_vm *vm, if (r) { DRM_ERROR("failed to do bo_map on static CSA, err=%d\n", r); - amdgpu_vm_bo_rmv(adev, *bo_va); + amdgpu_vm_bo_del(adev, *bo_va); ttm_eu_backoff_reservation(&ticket, &list); return r; } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c index 468003583b2a..d2139ac12159 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c @@ -23,6 +23,7 @@ */ #include <drm/drm_auth.h> +#include <drm/drm_drv.h> #include "amdgpu.h" #include "amdgpu_sched.h" #include "amdgpu_ras.h" @@ -109,7 +110,7 @@ static int amdgpu_ctx_priority_permit(struct drm_file *filp, return -EACCES; } -static enum amdgpu_gfx_pipe_priority amdgpu_ctx_prio_to_compute_prio(int32_t prio) +static enum amdgpu_gfx_pipe_priority amdgpu_ctx_prio_to_gfx_pipe_prio(int32_t prio) { switch (prio) { case AMDGPU_CTX_PRIORITY_HIGH: @@ -134,16 +135,17 @@ static enum amdgpu_ring_priority_level amdgpu_ctx_sched_prio_to_ring_prio(int32_ static unsigned int amdgpu_ctx_get_hw_prio(struct amdgpu_ctx *ctx, u32 hw_ip) { - struct amdgpu_device *adev = ctx->adev; - int32_t ctx_prio; + struct amdgpu_device *adev = ctx->mgr->adev; unsigned int hw_prio; + int32_t ctx_prio; ctx_prio = (ctx->override_priority == AMDGPU_CTX_PRIORITY_UNSET) ? ctx->init_priority : ctx->override_priority; switch (hw_ip) { + case AMDGPU_HW_IP_GFX: case AMDGPU_HW_IP_COMPUTE: - hw_prio = amdgpu_ctx_prio_to_compute_prio(ctx_prio); + hw_prio = amdgpu_ctx_prio_to_gfx_pipe_prio(ctx_prio); break; case AMDGPU_HW_IP_VCE: case AMDGPU_HW_IP_VCN_ENC: @@ -161,17 +163,50 @@ static unsigned int amdgpu_ctx_get_hw_prio(struct amdgpu_ctx *ctx, u32 hw_ip) return hw_prio; } +/* Calculate the time spend on the hw */ +static ktime_t amdgpu_ctx_fence_time(struct dma_fence *fence) +{ + struct drm_sched_fence *s_fence; + + if (!fence) + return ns_to_ktime(0); + + /* When the fence is not even scheduled it can't have spend time */ + s_fence = to_drm_sched_fence(fence); + if (!test_bit(DMA_FENCE_FLAG_TIMESTAMP_BIT, &s_fence->scheduled.flags)) + return ns_to_ktime(0); + + /* When it is still running account how much already spend */ + if (!test_bit(DMA_FENCE_FLAG_TIMESTAMP_BIT, &s_fence->finished.flags)) + return ktime_sub(ktime_get(), s_fence->scheduled.timestamp); + + return ktime_sub(s_fence->finished.timestamp, + s_fence->scheduled.timestamp); +} + +static ktime_t amdgpu_ctx_entity_time(struct amdgpu_ctx *ctx, + struct amdgpu_ctx_entity *centity) +{ + ktime_t res = ns_to_ktime(0); + uint32_t i; + + spin_lock(&ctx->ring_lock); + for (i = 0; i < amdgpu_sched_jobs; i++) { + res = ktime_add(res, amdgpu_ctx_fence_time(centity->fences[i])); + } + spin_unlock(&ctx->ring_lock); + return res; +} static int amdgpu_ctx_init_entity(struct amdgpu_ctx *ctx, u32 hw_ip, const u32 ring) { - struct amdgpu_device *adev = ctx->adev; - struct amdgpu_ctx_entity *entity; struct drm_gpu_scheduler **scheds = NULL, *sched = NULL; - unsigned num_scheds = 0; - int32_t ctx_prio; - unsigned int hw_prio; + struct amdgpu_device *adev = ctx->mgr->adev; + struct amdgpu_ctx_entity *entity; enum drm_sched_priority drm_prio; + unsigned int hw_prio, num_scheds; + int32_t ctx_prio; int r; entity = kzalloc(struct_size(entity, fences, amdgpu_sched_jobs), @@ -181,6 +216,7 @@ static int amdgpu_ctx_init_entity(struct amdgpu_ctx *ctx, u32 hw_ip, ctx_prio = (ctx->override_priority == AMDGPU_CTX_PRIORITY_UNSET) ? ctx->init_priority : ctx->override_priority; + entity->hw_ip = hw_ip; entity->sequence = 1; hw_prio = amdgpu_ctx_get_hw_prio(ctx, hw_ip); drm_prio = amdgpu_ctx_to_drm_sched_prio(ctx_prio); @@ -204,20 +240,70 @@ static int amdgpu_ctx_init_entity(struct amdgpu_ctx *ctx, u32 hw_ip, if (r) goto error_free_entity; - ctx->entities[hw_ip][ring] = entity; + /* It's not an error if we fail to install the new entity */ + if (cmpxchg(&ctx->entities[hw_ip][ring], NULL, entity)) + goto cleanup_entity; + return 0; +cleanup_entity: + drm_sched_entity_fini(&entity->entity); + error_free_entity: kfree(entity); return r; } -static int amdgpu_ctx_init(struct amdgpu_device *adev, - int32_t priority, - struct drm_file *filp, - struct amdgpu_ctx *ctx) +static ktime_t amdgpu_ctx_fini_entity(struct amdgpu_ctx_entity *entity) +{ + ktime_t res = ns_to_ktime(0); + int i; + + if (!entity) + return res; + + for (i = 0; i < amdgpu_sched_jobs; ++i) { + res = ktime_add(res, amdgpu_ctx_fence_time(entity->fences[i])); + dma_fence_put(entity->fences[i]); + } + + kfree(entity); + return res; +} + +static int amdgpu_ctx_get_stable_pstate(struct amdgpu_ctx *ctx, + u32 *stable_pstate) { + struct amdgpu_device *adev = ctx->mgr->adev; + enum amd_dpm_forced_level current_level; + + current_level = amdgpu_dpm_get_performance_level(adev); + + switch (current_level) { + case AMD_DPM_FORCED_LEVEL_PROFILE_STANDARD: + *stable_pstate = AMDGPU_CTX_STABLE_PSTATE_STANDARD; + break; + case AMD_DPM_FORCED_LEVEL_PROFILE_MIN_SCLK: + *stable_pstate = AMDGPU_CTX_STABLE_PSTATE_MIN_SCLK; + break; + case AMD_DPM_FORCED_LEVEL_PROFILE_MIN_MCLK: + *stable_pstate = AMDGPU_CTX_STABLE_PSTATE_MIN_MCLK; + break; + case AMD_DPM_FORCED_LEVEL_PROFILE_PEAK: + *stable_pstate = AMDGPU_CTX_STABLE_PSTATE_PEAK; + break; + default: + *stable_pstate = AMDGPU_CTX_STABLE_PSTATE_NONE; + break; + } + return 0; +} + +static int amdgpu_ctx_init(struct amdgpu_ctx_mgr *mgr, int32_t priority, + struct drm_file *filp, struct amdgpu_ctx *ctx) +{ + u32 current_stable_pstate; int r; r = amdgpu_ctx_priority_permit(filp, priority); @@ -226,52 +312,103 @@ static int amdgpu_ctx_init(struct amdgpu_device *adev, memset(ctx, 0, sizeof(*ctx)); - ctx->adev = adev; - kref_init(&ctx->refcount); + ctx->mgr = mgr; spin_lock_init(&ctx->ring_lock); - mutex_init(&ctx->lock); - ctx->reset_counter = atomic_read(&adev->gpu_reset_counter); + ctx->reset_counter = atomic_read(&mgr->adev->gpu_reset_counter); ctx->reset_counter_query = ctx->reset_counter; - ctx->vram_lost_counter = atomic_read(&adev->vram_lost_counter); + ctx->vram_lost_counter = atomic_read(&mgr->adev->vram_lost_counter); ctx->init_priority = priority; ctx->override_priority = AMDGPU_CTX_PRIORITY_UNSET; + r = amdgpu_ctx_get_stable_pstate(ctx, ¤t_stable_pstate); + if (r) + return r; + + if (mgr->adev->pm.stable_pstate_ctx) + ctx->stable_pstate = mgr->adev->pm.stable_pstate_ctx->stable_pstate; + else + ctx->stable_pstate = current_stable_pstate; + return 0; } -static void amdgpu_ctx_fini_entity(struct amdgpu_ctx_entity *entity) +static int amdgpu_ctx_set_stable_pstate(struct amdgpu_ctx *ctx, + u32 stable_pstate) { + struct amdgpu_device *adev = ctx->mgr->adev; + enum amd_dpm_forced_level level; + u32 current_stable_pstate; + int r; - int i; + mutex_lock(&adev->pm.stable_pstate_ctx_lock); + if (adev->pm.stable_pstate_ctx && adev->pm.stable_pstate_ctx != ctx) { + r = -EBUSY; + goto done; + } - if (!entity) - return; + r = amdgpu_ctx_get_stable_pstate(ctx, ¤t_stable_pstate); + if (r || (stable_pstate == current_stable_pstate)) + goto done; - for (i = 0; i < amdgpu_sched_jobs; ++i) - dma_fence_put(entity->fences[i]); + switch (stable_pstate) { + case AMDGPU_CTX_STABLE_PSTATE_NONE: + level = AMD_DPM_FORCED_LEVEL_AUTO; + break; + case AMDGPU_CTX_STABLE_PSTATE_STANDARD: + level = AMD_DPM_FORCED_LEVEL_PROFILE_STANDARD; + break; + case AMDGPU_CTX_STABLE_PSTATE_MIN_SCLK: + level = AMD_DPM_FORCED_LEVEL_PROFILE_MIN_SCLK; + break; + case AMDGPU_CTX_STABLE_PSTATE_MIN_MCLK: + level = AMD_DPM_FORCED_LEVEL_PROFILE_MIN_MCLK; + break; + case AMDGPU_CTX_STABLE_PSTATE_PEAK: + level = AMD_DPM_FORCED_LEVEL_PROFILE_PEAK; + break; + default: + r = -EINVAL; + goto done; + } - kfree(entity); + r = amdgpu_dpm_force_performance_level(adev, level); + + if (level == AMD_DPM_FORCED_LEVEL_AUTO) + adev->pm.stable_pstate_ctx = NULL; + else + adev->pm.stable_pstate_ctx = ctx; +done: + mutex_unlock(&adev->pm.stable_pstate_ctx_lock); + + return r; } static void amdgpu_ctx_fini(struct kref *ref) { struct amdgpu_ctx *ctx = container_of(ref, struct amdgpu_ctx, refcount); - struct amdgpu_device *adev = ctx->adev; - unsigned i, j; + struct amdgpu_ctx_mgr *mgr = ctx->mgr; + struct amdgpu_device *adev = mgr->adev; + unsigned i, j, idx; if (!adev) return; for (i = 0; i < AMDGPU_HW_IP_NUM; ++i) { for (j = 0; j < AMDGPU_MAX_ENTITY_NUM; ++j) { - amdgpu_ctx_fini_entity(ctx->entities[i][j]); - ctx->entities[i][j] = NULL; + ktime_t spend; + + spend = amdgpu_ctx_fini_entity(ctx->entities[i][j]); + atomic64_add(ktime_to_ns(spend), &mgr->time_spend[i]); } } - mutex_destroy(&ctx->lock); + if (drm_dev_enter(adev_to_drm(adev), &idx)) { + amdgpu_ctx_set_stable_pstate(ctx, ctx->stable_pstate); + drm_dev_exit(idx); + } + kfree(ctx); } @@ -329,7 +466,7 @@ static int amdgpu_ctx_alloc(struct amdgpu_device *adev, } *id = (uint32_t)r; - r = amdgpu_ctx_init(adev, priority, filp, ctx); + r = amdgpu_ctx_init(mgr, priority, filp, ctx); if (r) { idr_remove(&mgr->ctx_handles, *id); *id = 0; @@ -467,11 +604,41 @@ static int amdgpu_ctx_query2(struct amdgpu_device *adev, return 0; } + + +static int amdgpu_ctx_stable_pstate(struct amdgpu_device *adev, + struct amdgpu_fpriv *fpriv, uint32_t id, + bool set, u32 *stable_pstate) +{ + struct amdgpu_ctx *ctx; + struct amdgpu_ctx_mgr *mgr; + int r; + + if (!fpriv) + return -EINVAL; + + mgr = &fpriv->ctx_mgr; + mutex_lock(&mgr->lock); + ctx = idr_find(&mgr->ctx_handles, id); + if (!ctx) { + mutex_unlock(&mgr->lock); + return -EINVAL; + } + + if (set) + r = amdgpu_ctx_set_stable_pstate(ctx, *stable_pstate); + else + r = amdgpu_ctx_get_stable_pstate(ctx, stable_pstate); + + mutex_unlock(&mgr->lock); + return r; +} + int amdgpu_ctx_ioctl(struct drm_device *dev, void *data, struct drm_file *filp) { int r; - uint32_t id; + uint32_t id, stable_pstate; int32_t priority; union drm_amdgpu_ctx *args = data; @@ -500,6 +667,21 @@ int amdgpu_ctx_ioctl(struct drm_device *dev, void *data, case AMDGPU_CTX_OP_QUERY_STATE2: r = amdgpu_ctx_query2(adev, fpriv, id, &args->out); break; + case AMDGPU_CTX_OP_GET_STABLE_PSTATE: + if (args->in.flags) + return -EINVAL; + r = amdgpu_ctx_stable_pstate(adev, fpriv, id, false, &stable_pstate); + if (!r) + args->out.pstate.flags = stable_pstate; + break; + case AMDGPU_CTX_OP_SET_STABLE_PSTATE: + if (args->in.flags & ~AMDGPU_CTX_STABLE_PSTATE_FLAGS_MASK) + return -EINVAL; + stable_pstate = args->in.flags & AMDGPU_CTX_STABLE_PSTATE_FLAGS_MASK; + if (stable_pstate > AMDGPU_CTX_STABLE_PSTATE_PEAK) + return -EINVAL; + r = amdgpu_ctx_stable_pstate(adev, fpriv, id, true, &stable_pstate); + break; default: return -EINVAL; } @@ -534,9 +716,9 @@ int amdgpu_ctx_put(struct amdgpu_ctx *ctx) return 0; } -void amdgpu_ctx_add_fence(struct amdgpu_ctx *ctx, - struct drm_sched_entity *entity, - struct dma_fence *fence, uint64_t *handle) +uint64_t amdgpu_ctx_add_fence(struct amdgpu_ctx *ctx, + struct drm_sched_entity *entity, + struct dma_fence *fence) { struct amdgpu_ctx_entity *centity = to_amdgpu_ctx_entity(entity); uint64_t seq = centity->sequence; @@ -545,8 +727,7 @@ void amdgpu_ctx_add_fence(struct amdgpu_ctx *ctx, idx = seq & (amdgpu_sched_jobs - 1); other = centity->fences[idx]; - if (other) - BUG_ON(!dma_fence_is_signaled(other)); + WARN_ON(other && !dma_fence_is_signaled(other)); dma_fence_get(fence); @@ -555,9 +736,11 @@ void amdgpu_ctx_add_fence(struct amdgpu_ctx *ctx, centity->sequence++; spin_unlock(&ctx->ring_lock); + atomic64_add(ktime_to_ns(amdgpu_ctx_fence_time(other)), + &ctx->mgr->time_spend[centity->hw_ip]); + dma_fence_put(other); - if (handle) - *handle = seq; + return seq; } struct dma_fence *amdgpu_ctx_get_fence(struct amdgpu_ctx *ctx, @@ -594,7 +777,7 @@ static void amdgpu_ctx_set_entity_priority(struct amdgpu_ctx *ctx, int hw_ip, int32_t priority) { - struct amdgpu_device *adev = ctx->adev; + struct amdgpu_device *adev = ctx->mgr->adev; unsigned int hw_prio; struct drm_gpu_scheduler **scheds = NULL; unsigned num_scheds; @@ -604,7 +787,7 @@ static void amdgpu_ctx_set_entity_priority(struct amdgpu_ctx *ctx, amdgpu_ctx_to_drm_sched_prio(priority)); /* set hw priority */ - if (hw_ip == AMDGPU_HW_IP_COMPUTE) { + if (hw_ip == AMDGPU_HW_IP_COMPUTE || hw_ip == AMDGPU_HW_IP_GFX) { hw_prio = amdgpu_ctx_get_hw_prio(ctx, hw_ip); hw_prio = array_index_nospec(hw_prio, AMDGPU_RING_PRIO_MAX); scheds = adev->gpu_sched[hw_ip][hw_prio].sched; @@ -659,10 +842,17 @@ int amdgpu_ctx_wait_prev_fence(struct amdgpu_ctx *ctx, return r; } -void amdgpu_ctx_mgr_init(struct amdgpu_ctx_mgr *mgr) +void amdgpu_ctx_mgr_init(struct amdgpu_ctx_mgr *mgr, + struct amdgpu_device *adev) { + unsigned int i; + + mgr->adev = adev; mutex_init(&mgr->lock); - idr_init(&mgr->ctx_handles); + idr_init_base(&mgr->ctx_handles, 1); + + for (i = 0; i < AMDGPU_HW_IP_NUM; ++i) + atomic64_set(&mgr->time_spend[i], 0); } long amdgpu_ctx_mgr_entity_flush(struct amdgpu_ctx_mgr *mgr, long timeout) @@ -738,80 +928,38 @@ void amdgpu_ctx_mgr_fini(struct amdgpu_ctx_mgr *mgr) mutex_destroy(&mgr->lock); } -static void amdgpu_ctx_fence_time(struct amdgpu_ctx *ctx, - struct amdgpu_ctx_entity *centity, ktime_t *total, ktime_t *max) -{ - ktime_t now, t1; - uint32_t i; - - *total = *max = 0; - - now = ktime_get(); - for (i = 0; i < amdgpu_sched_jobs; i++) { - struct dma_fence *fence; - struct drm_sched_fence *s_fence; - - spin_lock(&ctx->ring_lock); - fence = dma_fence_get(centity->fences[i]); - spin_unlock(&ctx->ring_lock); - if (!fence) - continue; - s_fence = to_drm_sched_fence(fence); - if (!dma_fence_is_signaled(&s_fence->scheduled)) { - dma_fence_put(fence); - continue; - } - t1 = s_fence->scheduled.timestamp; - if (!ktime_before(t1, now)) { - dma_fence_put(fence); - continue; - } - if (dma_fence_is_signaled(&s_fence->finished) && - s_fence->finished.timestamp < now) - *total += ktime_sub(s_fence->finished.timestamp, t1); - else - *total += ktime_sub(now, t1); - t1 = ktime_sub(now, t1); - dma_fence_put(fence); - *max = max(t1, *max); - } -} - -ktime_t amdgpu_ctx_mgr_fence_usage(struct amdgpu_ctx_mgr *mgr, uint32_t hwip, - uint32_t idx, uint64_t *elapsed) +void amdgpu_ctx_mgr_usage(struct amdgpu_ctx_mgr *mgr, + ktime_t usage[AMDGPU_HW_IP_NUM]) { - struct idr *idp; struct amdgpu_ctx *ctx; + unsigned int hw_ip, i; uint32_t id; - struct amdgpu_ctx_entity *centity; - ktime_t total = 0, max = 0; - if (idx >= AMDGPU_MAX_ENTITY_NUM) - return 0; - idp = &mgr->ctx_handles; + /* + * This is a little bit racy because it can be that a ctx or a fence are + * destroyed just in the moment we try to account them. But that is ok + * since exactly that case is explicitely allowed by the interface. + */ mutex_lock(&mgr->lock); - idr_for_each_entry(idp, ctx, id) { - ktime_t ttotal, tmax; - - if (!ctx->entities[hwip][idx]) - continue; + for (hw_ip = 0; hw_ip < AMDGPU_HW_IP_NUM; ++hw_ip) { + uint64_t ns = atomic64_read(&mgr->time_spend[hw_ip]); - centity = ctx->entities[hwip][idx]; - amdgpu_ctx_fence_time(ctx, centity, &ttotal, &tmax); + usage[hw_ip] = ns_to_ktime(ns); + } - /* Harmonic mean approximation diverges for very small - * values. If ratio < 0.01% ignore - */ - if (AMDGPU_CTX_FENCE_USAGE_MIN_RATIO(tmax, ttotal)) - continue; + idr_for_each_entry(&mgr->ctx_handles, ctx, id) { + for (hw_ip = 0; hw_ip < AMDGPU_HW_IP_NUM; ++hw_ip) { + for (i = 0; i < amdgpu_ctx_num_entities[hw_ip]; ++i) { + struct amdgpu_ctx_entity *centity; + ktime_t spend; - total = ktime_add(total, ttotal); - max = ktime_after(tmax, max) ? tmax : max; + centity = ctx->entities[hw_ip][i]; + if (!centity) + continue; + spend = amdgpu_ctx_entity_time(ctx, centity); + usage[hw_ip] = ktime_add(usage[hw_ip], spend); + } + } } - mutex_unlock(&mgr->lock); - if (elapsed) - *elapsed = max; - - return total; } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.h index a44b8b8ed39c..0fa0e56daf67 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.h @@ -23,16 +23,20 @@ #ifndef __AMDGPU_CTX_H__ #define __AMDGPU_CTX_H__ +#include <linux/ktime.h> +#include <linux/types.h> + #include "amdgpu_ring.h" struct drm_device; struct drm_file; struct amdgpu_fpriv; +struct amdgpu_ctx_mgr; #define AMDGPU_MAX_ENTITY_NUM 4 -#define AMDGPU_CTX_FENCE_USAGE_MIN_RATIO(max, total) ((max) > 16384ULL*(total)) struct amdgpu_ctx_entity { + uint32_t hw_ip; uint64_t sequence; struct drm_sched_entity entity; struct dma_fence *fences[]; @@ -40,7 +44,7 @@ struct amdgpu_ctx_entity { struct amdgpu_ctx { struct kref refcount; - struct amdgpu_device *adev; + struct amdgpu_ctx_mgr *mgr; unsigned reset_counter; unsigned reset_counter_query; uint32_t vram_lost_counter; @@ -49,10 +53,10 @@ struct amdgpu_ctx { bool preamble_presented; int32_t init_priority; int32_t override_priority; - struct mutex lock; atomic_t guilty; unsigned long ras_counter_ce; unsigned long ras_counter_ue; + uint32_t stable_pstate; }; struct amdgpu_ctx_mgr { @@ -60,6 +64,7 @@ struct amdgpu_ctx_mgr { struct mutex lock; /* protected by lock */ struct idr ctx_handles; + atomic64_t time_spend[AMDGPU_HW_IP_NUM]; }; extern const unsigned int amdgpu_ctx_num_entities[AMDGPU_HW_IP_NUM]; @@ -69,9 +74,9 @@ int amdgpu_ctx_put(struct amdgpu_ctx *ctx); int amdgpu_ctx_get_entity(struct amdgpu_ctx *ctx, u32 hw_ip, u32 instance, u32 ring, struct drm_sched_entity **entity); -void amdgpu_ctx_add_fence(struct amdgpu_ctx *ctx, - struct drm_sched_entity *entity, - struct dma_fence *fence, uint64_t *seq); +uint64_t amdgpu_ctx_add_fence(struct amdgpu_ctx *ctx, + struct drm_sched_entity *entity, + struct dma_fence *fence); struct dma_fence *amdgpu_ctx_get_fence(struct amdgpu_ctx *ctx, struct drm_sched_entity *entity, uint64_t seq); @@ -84,10 +89,12 @@ int amdgpu_ctx_ioctl(struct drm_device *dev, void *data, int amdgpu_ctx_wait_prev_fence(struct amdgpu_ctx *ctx, struct drm_sched_entity *entity); -void amdgpu_ctx_mgr_init(struct amdgpu_ctx_mgr *mgr); +void amdgpu_ctx_mgr_init(struct amdgpu_ctx_mgr *mgr, + struct amdgpu_device *adev); void amdgpu_ctx_mgr_entity_fini(struct amdgpu_ctx_mgr *mgr); long amdgpu_ctx_mgr_entity_flush(struct amdgpu_ctx_mgr *mgr, long timeout); void amdgpu_ctx_mgr_fini(struct amdgpu_ctx_mgr *mgr); -ktime_t amdgpu_ctx_mgr_fence_usage(struct amdgpu_ctx_mgr *mgr, uint32_t hwip, - uint32_t idx, uint64_t *elapsed); +void amdgpu_ctx_mgr_usage(struct amdgpu_ctx_mgr *mgr, + ktime_t usage[AMDGPU_HW_IP_NUM]); + #endif diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c index 25e2e5bf90eb..de61a85c4b02 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c @@ -37,6 +37,9 @@ #include "amdgpu_fw_attestation.h" #include "amdgpu_umr.h" +#include "amdgpu_reset.h" +#include "amdgpu_psp_ta.h" + #if defined(CONFIG_DEBUG_FS) /** @@ -380,12 +383,8 @@ static ssize_t amdgpu_debugfs_regs_pcie_read(struct file *f, char __user *buf, value = RREG32_PCIE(*pos); r = put_user(value, (uint32_t *)buf); - if (r) { - pm_runtime_mark_last_busy(adev_to_drm(adev)->dev); - pm_runtime_put_autosuspend(adev_to_drm(adev)->dev); - amdgpu_virt_disable_access_debugfs(adev); - return r; - } + if (r) + goto out; result += 4; buf += 4; @@ -393,11 +392,12 @@ static ssize_t amdgpu_debugfs_regs_pcie_read(struct file *f, char __user *buf, size -= 4; } + r = result; +out: pm_runtime_mark_last_busy(adev_to_drm(adev)->dev); pm_runtime_put_autosuspend(adev_to_drm(adev)->dev); - amdgpu_virt_disable_access_debugfs(adev); - return result; + return r; } /** @@ -438,12 +438,8 @@ static ssize_t amdgpu_debugfs_regs_pcie_write(struct file *f, const char __user uint32_t value; r = get_user(value, (uint32_t *)buf); - if (r) { - pm_runtime_mark_last_busy(adev_to_drm(adev)->dev); - pm_runtime_put_autosuspend(adev_to_drm(adev)->dev); - amdgpu_virt_disable_access_debugfs(adev); - return r; - } + if (r) + goto out; WREG32_PCIE(*pos, value); @@ -453,11 +449,12 @@ static ssize_t amdgpu_debugfs_regs_pcie_write(struct file *f, const char __user size -= 4; } + r = result; +out: pm_runtime_mark_last_busy(adev_to_drm(adev)->dev); pm_runtime_put_autosuspend(adev_to_drm(adev)->dev); - amdgpu_virt_disable_access_debugfs(adev); - return result; + return r; } /** @@ -499,12 +496,8 @@ static ssize_t amdgpu_debugfs_regs_didt_read(struct file *f, char __user *buf, value = RREG32_DIDT(*pos >> 2); r = put_user(value, (uint32_t *)buf); - if (r) { - pm_runtime_mark_last_busy(adev_to_drm(adev)->dev); - pm_runtime_put_autosuspend(adev_to_drm(adev)->dev); - amdgpu_virt_disable_access_debugfs(adev); - return r; - } + if (r) + goto out; result += 4; buf += 4; @@ -512,11 +505,12 @@ static ssize_t amdgpu_debugfs_regs_didt_read(struct file *f, char __user *buf, size -= 4; } + r = result; +out: pm_runtime_mark_last_busy(adev_to_drm(adev)->dev); pm_runtime_put_autosuspend(adev_to_drm(adev)->dev); - amdgpu_virt_disable_access_debugfs(adev); - return result; + return r; } /** @@ -557,12 +551,8 @@ static ssize_t amdgpu_debugfs_regs_didt_write(struct file *f, const char __user uint32_t value; r = get_user(value, (uint32_t *)buf); - if (r) { - pm_runtime_mark_last_busy(adev_to_drm(adev)->dev); - pm_runtime_put_autosuspend(adev_to_drm(adev)->dev); - amdgpu_virt_disable_access_debugfs(adev); - return r; - } + if (r) + goto out; WREG32_DIDT(*pos >> 2, value); @@ -572,11 +562,12 @@ static ssize_t amdgpu_debugfs_regs_didt_write(struct file *f, const char __user size -= 4; } + r = result; +out: pm_runtime_mark_last_busy(adev_to_drm(adev)->dev); pm_runtime_put_autosuspend(adev_to_drm(adev)->dev); - amdgpu_virt_disable_access_debugfs(adev); - return result; + return r; } /** @@ -618,12 +609,8 @@ static ssize_t amdgpu_debugfs_regs_smc_read(struct file *f, char __user *buf, value = RREG32_SMC(*pos); r = put_user(value, (uint32_t *)buf); - if (r) { - pm_runtime_mark_last_busy(adev_to_drm(adev)->dev); - pm_runtime_put_autosuspend(adev_to_drm(adev)->dev); - amdgpu_virt_disable_access_debugfs(adev); - return r; - } + if (r) + goto out; result += 4; buf += 4; @@ -631,11 +618,12 @@ static ssize_t amdgpu_debugfs_regs_smc_read(struct file *f, char __user *buf, size -= 4; } + r = result; +out: pm_runtime_mark_last_busy(adev_to_drm(adev)->dev); pm_runtime_put_autosuspend(adev_to_drm(adev)->dev); - amdgpu_virt_disable_access_debugfs(adev); - return result; + return r; } /** @@ -676,12 +664,8 @@ static ssize_t amdgpu_debugfs_regs_smc_write(struct file *f, const char __user * uint32_t value; r = get_user(value, (uint32_t *)buf); - if (r) { - pm_runtime_mark_last_busy(adev_to_drm(adev)->dev); - pm_runtime_put_autosuspend(adev_to_drm(adev)->dev); - amdgpu_virt_disable_access_debugfs(adev); - return r; - } + if (r) + goto out; WREG32_SMC(*pos, value); @@ -691,11 +675,12 @@ static ssize_t amdgpu_debugfs_regs_smc_write(struct file *f, const char __user * size -= 4; } + r = result; +out: pm_runtime_mark_last_busy(adev_to_drm(adev)->dev); pm_runtime_put_autosuspend(adev_to_drm(adev)->dev); - amdgpu_virt_disable_access_debugfs(adev); - return result; + return r; } /** @@ -728,7 +713,7 @@ static ssize_t amdgpu_debugfs_gca_config_read(struct file *f, char __user *buf, return -ENOMEM; /* version, increment each time something is added */ - config[no_regs++] = 3; + config[no_regs++] = 5; config[no_regs++] = adev->gfx.config.max_shader_engines; config[no_regs++] = adev->gfx.config.max_tile_pipes; config[no_regs++] = adev->gfx.config.max_cu_per_sh; @@ -755,8 +740,8 @@ static ssize_t amdgpu_debugfs_gca_config_read(struct file *f, char __user *buf, /* rev==1 */ config[no_regs++] = adev->rev_id; - config[no_regs++] = adev->pg_flags; - config[no_regs++] = adev->cg_flags; + config[no_regs++] = lower_32_bits(adev->pg_flags); + config[no_regs++] = lower_32_bits(adev->cg_flags); /* rev==2 */ config[no_regs++] = adev->family; @@ -768,6 +753,13 @@ static ssize_t amdgpu_debugfs_gca_config_read(struct file *f, char __user *buf, config[no_regs++] = adev->pdev->subsystem_device; config[no_regs++] = adev->pdev->subsystem_vendor; + /* rev==4 APU flag */ + config[no_regs++] = adev->flags & AMD_IS_APU ? 1 : 0; + + /* rev==5 PG/CG flag upper 32bit */ + config[no_regs++] = upper_32_bits(adev->pg_flags); + config[no_regs++] = upper_32_bits(adev->cg_flags); + while (size && (*pos < no_regs * 4)) { uint32_t value; @@ -1051,6 +1043,157 @@ err: } /** + * amdgpu_debugfs_gfxoff_residency_read - Read GFXOFF residency + * + * @f: open file handle + * @buf: User buffer to store read data in + * @size: Number of bytes to read + * @pos: Offset to seek to + * + * Read the last residency value logged. It doesn't auto update, one needs to + * stop logging before getting the current value. + */ +static ssize_t amdgpu_debugfs_gfxoff_residency_read(struct file *f, char __user *buf, + size_t size, loff_t *pos) +{ + struct amdgpu_device *adev = file_inode(f)->i_private; + ssize_t result = 0; + int r; + + if (size & 0x3 || *pos & 0x3) + return -EINVAL; + + r = pm_runtime_get_sync(adev_to_drm(adev)->dev); + if (r < 0) { + pm_runtime_put_autosuspend(adev_to_drm(adev)->dev); + return r; + } + + while (size) { + uint32_t value; + + r = amdgpu_get_gfx_off_residency(adev, &value); + if (r) + goto out; + + r = put_user(value, (uint32_t *)buf); + if (r) + goto out; + + result += 4; + buf += 4; + *pos += 4; + size -= 4; + } + + r = result; +out: + pm_runtime_mark_last_busy(adev_to_drm(adev)->dev); + pm_runtime_put_autosuspend(adev_to_drm(adev)->dev); + + return r; +} + +/** + * amdgpu_debugfs_gfxoff_residency_write - Log GFXOFF Residency + * + * @f: open file handle + * @buf: User buffer to write data from + * @size: Number of bytes to write + * @pos: Offset to seek to + * + * Write a 32-bit non-zero to start logging; write a 32-bit zero to stop + */ +static ssize_t amdgpu_debugfs_gfxoff_residency_write(struct file *f, const char __user *buf, + size_t size, loff_t *pos) +{ + struct amdgpu_device *adev = file_inode(f)->i_private; + ssize_t result = 0; + int r; + + if (size & 0x3 || *pos & 0x3) + return -EINVAL; + + r = pm_runtime_get_sync(adev_to_drm(adev)->dev); + if (r < 0) { + pm_runtime_put_autosuspend(adev_to_drm(adev)->dev); + return r; + } + + while (size) { + u32 value; + + r = get_user(value, (uint32_t *)buf); + if (r) + goto out; + + amdgpu_set_gfx_off_residency(adev, value ? true : false); + + result += 4; + buf += 4; + *pos += 4; + size -= 4; + } + + r = result; +out: + pm_runtime_mark_last_busy(adev_to_drm(adev)->dev); + pm_runtime_put_autosuspend(adev_to_drm(adev)->dev); + + return r; +} + + +/** + * amdgpu_debugfs_gfxoff_count_read - Read GFXOFF entry count + * + * @f: open file handle + * @buf: User buffer to store read data in + * @size: Number of bytes to read + * @pos: Offset to seek to + */ +static ssize_t amdgpu_debugfs_gfxoff_count_read(struct file *f, char __user *buf, + size_t size, loff_t *pos) +{ + struct amdgpu_device *adev = file_inode(f)->i_private; + ssize_t result = 0; + int r; + + if (size & 0x3 || *pos & 0x3) + return -EINVAL; + + r = pm_runtime_get_sync(adev_to_drm(adev)->dev); + if (r < 0) { + pm_runtime_put_autosuspend(adev_to_drm(adev)->dev); + return r; + } + + while (size) { + u64 value = 0; + + r = amdgpu_get_gfx_off_entrycount(adev, &value); + if (r) + goto out; + + r = put_user(value, (u64 *)buf); + if (r) + goto out; + + result += 4; + buf += 4; + *pos += 4; + size -= 4; + } + + r = result; +out: + pm_runtime_mark_last_busy(adev_to_drm(adev)->dev); + pm_runtime_put_autosuspend(adev_to_drm(adev)->dev); + + return r; +} + +/** * amdgpu_debugfs_gfxoff_write - Enable/disable GFXOFF * * @f: open file handle @@ -1080,11 +1223,8 @@ static ssize_t amdgpu_debugfs_gfxoff_write(struct file *f, const char __user *bu uint32_t value; r = get_user(value, (uint32_t *)buf); - if (r) { - pm_runtime_mark_last_busy(adev_to_drm(adev)->dev); - pm_runtime_put_autosuspend(adev_to_drm(adev)->dev); - return r; - } + if (r) + goto out; amdgpu_gfx_off_ctrl(adev, value ? true : false); @@ -1094,10 +1234,12 @@ static ssize_t amdgpu_debugfs_gfxoff_write(struct file *f, const char __user *bu size -= 4; } + r = result; +out: pm_runtime_mark_last_busy(adev_to_drm(adev)->dev); pm_runtime_put_autosuspend(adev_to_drm(adev)->dev); - return result; + return r; } @@ -1120,25 +1262,58 @@ static ssize_t amdgpu_debugfs_gfxoff_read(struct file *f, char __user *buf, return -EINVAL; r = pm_runtime_get_sync(adev_to_drm(adev)->dev); - if (r < 0) + if (r < 0) { + pm_runtime_put_autosuspend(adev_to_drm(adev)->dev); return r; + } while (size) { - uint32_t value; + u32 value = adev->gfx.gfx_off_state; + + r = put_user(value, (u32 *)buf); + if (r) + goto out; + + result += 4; + buf += 4; + *pos += 4; + size -= 4; + } + + r = result; +out: + pm_runtime_mark_last_busy(adev_to_drm(adev)->dev); + pm_runtime_put_autosuspend(adev_to_drm(adev)->dev); + + return r; +} + +static ssize_t amdgpu_debugfs_gfxoff_status_read(struct file *f, char __user *buf, + size_t size, loff_t *pos) +{ + struct amdgpu_device *adev = file_inode(f)->i_private; + ssize_t result = 0; + int r; + + if (size & 0x3 || *pos & 0x3) + return -EINVAL; + + r = pm_runtime_get_sync(adev_to_drm(adev)->dev); + if (r < 0) { + pm_runtime_put_autosuspend(adev_to_drm(adev)->dev); + return r; + } + + while (size) { + u32 value; r = amdgpu_get_gfx_off_status(adev, &value); - if (r) { - pm_runtime_mark_last_busy(adev_to_drm(adev)->dev); - pm_runtime_put_autosuspend(adev_to_drm(adev)->dev); - return r; - } + if (r) + goto out; - r = put_user(value, (uint32_t *)buf); - if (r) { - pm_runtime_mark_last_busy(adev_to_drm(adev)->dev); - pm_runtime_put_autosuspend(adev_to_drm(adev)->dev); - return r; - } + r = put_user(value, (u32 *)buf); + if (r) + goto out; result += 4; buf += 4; @@ -1146,10 +1321,12 @@ static ssize_t amdgpu_debugfs_gfxoff_read(struct file *f, char __user *buf, size -= 4; } + r = result; +out: pm_runtime_mark_last_busy(adev_to_drm(adev)->dev); pm_runtime_put_autosuspend(adev_to_drm(adev)->dev); - return result; + return r; } static const struct file_operations amdgpu_debugfs_regs2_fops = { @@ -1217,6 +1394,25 @@ static const struct file_operations amdgpu_debugfs_gfxoff_fops = { .llseek = default_llseek }; +static const struct file_operations amdgpu_debugfs_gfxoff_status_fops = { + .owner = THIS_MODULE, + .read = amdgpu_debugfs_gfxoff_status_read, + .llseek = default_llseek +}; + +static const struct file_operations amdgpu_debugfs_gfxoff_count_fops = { + .owner = THIS_MODULE, + .read = amdgpu_debugfs_gfxoff_count_read, + .llseek = default_llseek +}; + +static const struct file_operations amdgpu_debugfs_gfxoff_residency_fops = { + .owner = THIS_MODULE, + .read = amdgpu_debugfs_gfxoff_residency_read, + .write = amdgpu_debugfs_gfxoff_residency_write, + .llseek = default_llseek +}; + static const struct file_operations *debugfs_regs[] = { &amdgpu_debugfs_regs_fops, &amdgpu_debugfs_regs2_fops, @@ -1228,6 +1424,9 @@ static const struct file_operations *debugfs_regs[] = { &amdgpu_debugfs_wave_fops, &amdgpu_debugfs_gpr_fops, &amdgpu_debugfs_gfxoff_fops, + &amdgpu_debugfs_gfxoff_status_fops, + &amdgpu_debugfs_gfxoff_count_fops, + &amdgpu_debugfs_gfxoff_residency_fops, }; static const char *debugfs_regs_names[] = { @@ -1241,6 +1440,9 @@ static const char *debugfs_regs_names[] = { "amdgpu_wave", "amdgpu_gpr", "amdgpu_gfxoff", + "amdgpu_gfxoff_status", + "amdgpu_gfxoff_count", + "amdgpu_gfxoff_residency", }; /** @@ -1279,7 +1481,7 @@ static int amdgpu_debugfs_test_ib_show(struct seq_file *m, void *unused) } /* Avoid accidently unparking the sched thread during GPU reset */ - r = down_write_killable(&adev->reset_sem); + r = down_write_killable(&adev->reset_domain->sem); if (r) return r; @@ -1308,7 +1510,7 @@ static int amdgpu_debugfs_test_ib_show(struct seq_file *m, void *unused) kthread_unpark(ring->sched.thread); } - up_write(&adev->reset_sem); + up_write(&adev->reset_domain->sem); pm_runtime_mark_last_busy(dev->dev); pm_runtime_put_autosuspend(dev->dev); @@ -1357,6 +1559,25 @@ static int amdgpu_debugfs_evict_gtt(void *data, u64 *val) return 0; } +static int amdgpu_debugfs_benchmark(void *data, u64 val) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)data; + struct drm_device *dev = adev_to_drm(adev); + int r; + + r = pm_runtime_get_sync(dev->dev); + if (r < 0) { + pm_runtime_put_autosuspend(dev->dev); + return r; + } + + r = amdgpu_benchmark(adev, val); + + pm_runtime_mark_last_busy(dev->dev); + pm_runtime_put_autosuspend(dev->dev); + + return r; +} static int amdgpu_debugfs_vm_info_show(struct seq_file *m, void *unused) { @@ -1393,6 +1614,8 @@ DEFINE_DEBUGFS_ATTRIBUTE(amdgpu_evict_vram_fops, amdgpu_debugfs_evict_vram, NULL, "%lld\n"); DEFINE_DEBUGFS_ATTRIBUTE(amdgpu_evict_gtt_fops, amdgpu_debugfs_evict_gtt, NULL, "%lld\n"); +DEFINE_DEBUGFS_ATTRIBUTE(amdgpu_benchmark_fops, NULL, amdgpu_debugfs_benchmark, + "%lld\n"); static void amdgpu_ib_preempt_fences_swap(struct amdgpu_ring *ring, struct dma_fence **fences) @@ -1517,7 +1740,7 @@ static int amdgpu_debugfs_ib_preempt(void *data, u64 val) return -ENOMEM; /* Avoid accidently unparking the sched thread during GPU reset */ - r = down_read_killable(&adev->reset_sem); + r = down_read_killable(&adev->reset_domain->sem); if (r) goto pro_end; @@ -1560,7 +1783,7 @@ failure: /* restart the scheduler */ kthread_unpark(ring->sched.thread); - up_read(&adev->reset_sem); + up_read(&adev->reset_domain->sem); ttm_bo_unlock_delayed_workqueue(&adev->mman.bdev, resched); @@ -1585,22 +1808,25 @@ static int amdgpu_debugfs_sclk_set(void *data, u64 val) return ret; } - if (is_support_sw_smu(adev)) { - ret = smu_get_dpm_freq_range(&adev->smu, SMU_SCLK, &min_freq, &max_freq); - if (ret || val > max_freq || val < min_freq) - return -EINVAL; - ret = smu_set_soft_freq_range(&adev->smu, SMU_SCLK, (uint32_t)val, (uint32_t)val); - } else { - return 0; + ret = amdgpu_dpm_get_dpm_freq_range(adev, PP_SCLK, &min_freq, &max_freq); + if (ret == -EOPNOTSUPP) { + ret = 0; + goto out; + } + if (ret || val > max_freq || val < min_freq) { + ret = -EINVAL; + goto out; } + ret = amdgpu_dpm_set_soft_freq_range(adev, PP_SCLK, (uint32_t)val, (uint32_t)val); + if (ret) + ret = -EINVAL; + +out: pm_runtime_mark_last_busy(adev_to_drm(adev)->dev); pm_runtime_put_autosuspend(adev_to_drm(adev)->dev); - if (ret) - return -EINVAL; - - return 0; + return ret; } DEFINE_DEBUGFS_ATTRIBUTE(fops_ib_preempt, NULL, @@ -1609,6 +1835,99 @@ DEFINE_DEBUGFS_ATTRIBUTE(fops_ib_preempt, NULL, DEFINE_DEBUGFS_ATTRIBUTE(fops_sclk_set, NULL, amdgpu_debugfs_sclk_set, "%llu\n"); +static ssize_t amdgpu_reset_dump_register_list_read(struct file *f, + char __user *buf, size_t size, loff_t *pos) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)file_inode(f)->i_private; + char reg_offset[12]; + int i, ret, len = 0; + + if (*pos) + return 0; + + memset(reg_offset, 0, 12); + ret = down_read_killable(&adev->reset_domain->sem); + if (ret) + return ret; + + for (i = 0; i < adev->num_regs; i++) { + sprintf(reg_offset, "0x%x\n", adev->reset_dump_reg_list[i]); + up_read(&adev->reset_domain->sem); + if (copy_to_user(buf + len, reg_offset, strlen(reg_offset))) + return -EFAULT; + + len += strlen(reg_offset); + ret = down_read_killable(&adev->reset_domain->sem); + if (ret) + return ret; + } + + up_read(&adev->reset_domain->sem); + *pos += len; + + return len; +} + +static ssize_t amdgpu_reset_dump_register_list_write(struct file *f, + const char __user *buf, size_t size, loff_t *pos) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)file_inode(f)->i_private; + char reg_offset[11]; + uint32_t *new = NULL, *tmp = NULL; + int ret, i = 0, len = 0; + + do { + memset(reg_offset, 0, 11); + if (copy_from_user(reg_offset, buf + len, + min(10, ((int)size-len)))) { + ret = -EFAULT; + goto error_free; + } + + new = krealloc_array(tmp, i + 1, sizeof(uint32_t), GFP_KERNEL); + if (!new) { + ret = -ENOMEM; + goto error_free; + } + tmp = new; + if (sscanf(reg_offset, "%X %n", &tmp[i], &ret) != 1) { + ret = -EINVAL; + goto error_free; + } + + len += ret; + i++; + } while (len < size); + + new = kmalloc_array(i, sizeof(uint32_t), GFP_KERNEL); + if (!new) { + ret = -ENOMEM; + goto error_free; + } + ret = down_write_killable(&adev->reset_domain->sem); + if (ret) + goto error_free; + + swap(adev->reset_dump_reg_list, tmp); + swap(adev->reset_dump_reg_value, new); + adev->num_regs = i; + up_write(&adev->reset_domain->sem); + ret = size; + +error_free: + if (tmp != new) + kfree(tmp); + kfree(new); + return ret; +} + +static const struct file_operations amdgpu_reset_dump_register_list = { + .owner = THIS_MODULE, + .read = amdgpu_reset_dump_register_list_read, + .write = amdgpu_reset_dump_register_list_write, + .llseek = default_llseek +}; + int amdgpu_debugfs_init(struct amdgpu_device *adev) { struct dentry *root = adev_to_drm(adev)->primary->debugfs_root; @@ -1647,6 +1966,7 @@ int amdgpu_debugfs_init(struct amdgpu_device *adev) DRM_ERROR("registering register debugfs failed (%d).\n", r); amdgpu_debugfs_firmware_init(adev); + amdgpu_ta_if_debugfs_init(adev); #if defined(CONFIG_DRM_AMD_DC) if (amdgpu_device_has_dc_support(adev)) @@ -1662,6 +1982,16 @@ int amdgpu_debugfs_init(struct amdgpu_device *adev) amdgpu_debugfs_ring_init(adev, ring); } + for ( i = 0; i < adev->vcn.num_vcn_inst; i++) { + if (!amdgpu_vcnfw_log) + break; + + if (adev->vcn.harvest_config & (1 << i)) + continue; + + amdgpu_debugfs_vcn_fwlog_init(adev, i, &adev->vcn.inst[i]); + } + amdgpu_ras_debugfs_create_all(adev); amdgpu_rap_debugfs_init(adev); amdgpu_securedisplay_debugfs_init(adev); @@ -1675,6 +2005,10 @@ int amdgpu_debugfs_init(struct amdgpu_device *adev) &amdgpu_debugfs_test_ib_fops); debugfs_create_file("amdgpu_vm_info", 0444, root, adev, &amdgpu_debugfs_vm_info_fops); + debugfs_create_file("amdgpu_benchmark", 0200, root, adev, + &amdgpu_benchmark_fops); + debugfs_create_file("amdgpu_reset_dump_register_list", 0644, root, adev, + &amdgpu_reset_dump_register_list); adev->debugfs_vbios_blob.data = adev->bios; adev->debugfs_vbios_blob.size = adev->bios_size; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index a8b08a72b71b..64510898eedd 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -31,6 +31,10 @@ #include <linux/console.h> #include <linux/slab.h> #include <linux/iommu.h> +#include <linux/pci.h> +#include <linux/devcoredump.h> +#include <generated/utsrelease.h> +#include <linux/pci-p2pdma.h> #include <drm/drm_atomic_helper.h> #include <drm/drm_probe_helper.h> @@ -55,7 +59,6 @@ #include "soc15.h" #include "nv.h" #include "bif/bif_4_1_d.h" -#include <linux/pci.h> #include <linux/firmware.h> #include "amdgpu_vf_error.h" @@ -80,14 +83,11 @@ MODULE_FIRMWARE("amdgpu/raven_gpu_info.bin"); MODULE_FIRMWARE("amdgpu/picasso_gpu_info.bin"); MODULE_FIRMWARE("amdgpu/raven2_gpu_info.bin"); MODULE_FIRMWARE("amdgpu/arcturus_gpu_info.bin"); -MODULE_FIRMWARE("amdgpu/renoir_gpu_info.bin"); -MODULE_FIRMWARE("amdgpu/navi10_gpu_info.bin"); -MODULE_FIRMWARE("amdgpu/navi14_gpu_info.bin"); MODULE_FIRMWARE("amdgpu/navi12_gpu_info.bin"); -MODULE_FIRMWARE("amdgpu/vangogh_gpu_info.bin"); -MODULE_FIRMWARE("amdgpu/yellow_carp_gpu_info.bin"); #define AMDGPU_RESUME_MS 2000 +#define AMDGPU_MAX_RETRY_LIMIT 2 +#define AMDGPU_RETRY_SRIOV_RESET(r) ((r) == -EBUSY || (r) == -ETIMEDOUT || (r) == -EINVAL) const char *amdgpu_asic_name[] = { "TAHITI", @@ -424,10 +424,10 @@ bool amdgpu_device_skip_hw_access(struct amdgpu_device *adev) * the lock. */ if (in_task()) { - if (down_read_trylock(&adev->reset_sem)) - up_read(&adev->reset_sem); + if (down_read_trylock(&adev->reset_domain->sem)) + up_read(&adev->reset_domain->sem); else - lockdep_assert_held(&adev->reset_sem); + lockdep_assert_held(&adev->reset_domain->sem); } #endif return false; @@ -453,9 +453,9 @@ uint32_t amdgpu_device_rreg(struct amdgpu_device *adev, if ((reg * 4) < adev->rmmio_size) { if (!(acc_flags & AMDGPU_REGS_NO_KIQ) && amdgpu_sriov_runtime(adev) && - down_read_trylock(&adev->reset_sem)) { + down_read_trylock(&adev->reset_domain->sem)) { ret = amdgpu_kiq_rreg(adev, reg); - up_read(&adev->reset_sem); + up_read(&adev->reset_domain->sem); } else { ret = readl(((void __iomem *)adev->rmmio) + (reg * 4)); } @@ -538,9 +538,9 @@ void amdgpu_device_wreg(struct amdgpu_device *adev, if ((reg * 4) < adev->rmmio_size) { if (!(acc_flags & AMDGPU_REGS_NO_KIQ) && amdgpu_sriov_runtime(adev) && - down_read_trylock(&adev->reset_sem)) { + down_read_trylock(&adev->reset_domain->sem)) { amdgpu_kiq_wreg(adev, reg, v); - up_read(&adev->reset_sem); + up_read(&adev->reset_domain->sem); } else { writel(v, ((void __iomem *)adev->rmmio) + (reg * 4)); } @@ -552,9 +552,13 @@ void amdgpu_device_wreg(struct amdgpu_device *adev, } /** - * amdgpu_mm_wreg_mmio_rlc - write register either with mmio or with RLC path if in range + * amdgpu_mm_wreg_mmio_rlc - write register either with direct/indirect mmio or with RLC path if in range * - * this function is invoked only the debugfs register access + * @adev: amdgpu_device pointer + * @reg: mmio/rlc register + * @v: value to write + * + * this function is invoked only for the debugfs register access */ void amdgpu_mm_wreg_mmio_rlc(struct amdgpu_device *adev, uint32_t reg, uint32_t v) @@ -566,7 +570,9 @@ void amdgpu_mm_wreg_mmio_rlc(struct amdgpu_device *adev, adev->gfx.rlc.funcs && adev->gfx.rlc.funcs->is_rlcg_access_range) { if (adev->gfx.rlc.funcs->is_rlcg_access_range(adev, reg)) - return adev->gfx.rlc.funcs->sriov_wreg(adev, reg, v, 0, 0); + return amdgpu_sriov_wreg(adev, reg, v, 0, 0); + } else if ((reg * 4) >= adev->rmmio_size) { + adev->pcie_wreg(adev, reg * 4, v); } else { writel(v, ((void __iomem *)adev->rmmio) + (reg * 4)); } @@ -910,7 +916,10 @@ static int amdgpu_device_asic_init(struct amdgpu_device *adev) { amdgpu_asic_pre_asic_init(adev); - return amdgpu_atom_asic_init(adev->mode_info.atom_context); + if (adev->ip_versions[GC_HWIP][0] >= IP_VERSION(11, 0, 0)) + return amdgpu_atomfirmware_asic_init(adev, true); + else + return amdgpu_atom_asic_init(adev->mode_info.atom_context); } /** @@ -1038,19 +1047,25 @@ static int amdgpu_device_doorbell_init(struct amdgpu_device *adev) adev->doorbell.base = pci_resource_start(adev->pdev, 2); adev->doorbell.size = pci_resource_len(adev->pdev, 2); - adev->doorbell.num_doorbells = min_t(u32, adev->doorbell.size / sizeof(u32), - adev->doorbell_index.max_assignment+1); - if (adev->doorbell.num_doorbells == 0) - return -EINVAL; - - /* For Vega, reserve and map two pages on doorbell BAR since SDMA - * paging queue doorbell use the second page. The - * AMDGPU_DOORBELL64_MAX_ASSIGNMENT definition assumes all the - * doorbells are in the first page. So with paging queue enabled, - * the max num_doorbells should + 1 page (0x400 in dword) - */ - if (adev->asic_type >= CHIP_VEGA10) - adev->doorbell.num_doorbells += 0x400; + if (adev->enable_mes) { + adev->doorbell.num_doorbells = + adev->doorbell.size / sizeof(u32); + } else { + adev->doorbell.num_doorbells = + min_t(u32, adev->doorbell.size / sizeof(u32), + adev->doorbell_index.max_assignment+1); + if (adev->doorbell.num_doorbells == 0) + return -EINVAL; + + /* For Vega, reserve and map two pages on doorbell BAR since SDMA + * paging queue doorbell use the second page. The + * AMDGPU_DOORBELL64_MAX_ASSIGNMENT definition assumes all the + * doorbells are in the first page. So with paging queue enabled, + * the max num_doorbells should + 1 page (0x400 in dword) + */ + if (adev->asic_type >= CHIP_VEGA10) + adev->doorbell.num_doorbells += 0x400; + } adev->doorbell.ptr = ioremap(adev->doorbell.base, adev->doorbell.num_doorbells * @@ -1310,6 +1325,31 @@ bool amdgpu_device_need_post(struct amdgpu_device *adev) return true; } +/** + * amdgpu_device_should_use_aspm - check if the device should program ASPM + * + * @adev: amdgpu_device pointer + * + * Confirm whether the module parameter and pcie bridge agree that ASPM should + * be set for this device. + * + * Returns true if it should be used or false if not. + */ +bool amdgpu_device_should_use_aspm(struct amdgpu_device *adev) +{ + switch (amdgpu_aspm) { + case -1: + break; + case 0: + return false; + case 1: + return true; + default: + return false; + } + return pcie_aspm_enabled(adev->pdev); +} + /* if we get transitioned to only one device, take VGA back */ /** * amdgpu_device_vga_set_decode - enable/disable vga decode @@ -1444,11 +1484,12 @@ static int amdgpu_device_init_apu_flags(struct amdgpu_device *adev) case CHIP_YELLOW_CARP: break; case CHIP_CYAN_SKILLFISH: - if (adev->pdev->device == 0x13FE) + if ((adev->pdev->device == 0x13FE) || + (adev->pdev->device == 0x143F)) adev->apu_flags |= AMD_APU_IS_CYAN_SKILLFISH2; break; default: - return -EINVAL; + break; } return 0; @@ -1505,6 +1546,11 @@ static int amdgpu_device_check_arguments(struct amdgpu_device *adev) amdgpu_sched_hw_submission = roundup_pow_of_two(amdgpu_sched_hw_submission); } + if (amdgpu_reset_method < -1 || amdgpu_reset_method > 4) { + dev_warn(adev->dev, "invalid option for reset method, reverting to default\n"); + amdgpu_reset_method = -1; + } + amdgpu_device_check_smu_prv_buffer_size(adev); amdgpu_device_check_vm_size(adev); @@ -1513,10 +1559,6 @@ static int amdgpu_device_check_arguments(struct amdgpu_device *adev) adev->firmware.load_type = amdgpu_ucode_get_load_type(adev, amdgpu_fw_load_type); - amdgpu_gmc_tmz_set(adev); - - amdgpu_gmc_noretry_set(adev); - return 0; } @@ -1670,7 +1712,7 @@ int amdgpu_device_ip_set_powergating_state(void *dev, * clockgating is enabled. */ void amdgpu_device_ip_get_clockgating_state(struct amdgpu_device *adev, - u32 *flags) + u64 *flags) { int i; @@ -1893,11 +1935,9 @@ static int amdgpu_device_parse_gpu_info_fw(struct amdgpu_device *adev) adev->firmware.gpu_info_fw = NULL; if (adev->mman.discovery_bin) { - amdgpu_discovery_get_gfx_info(adev); - /* * FIXME: The bounding box is still needed by Navi12, so - * temporarily read it from gpu_info firmware. Should be droped + * temporarily read it from gpu_info firmware. Should be dropped * when DAL no longer needs it. */ if (adev->asic_type != CHIP_NAVI12) @@ -1905,35 +1945,6 @@ static int amdgpu_device_parse_gpu_info_fw(struct amdgpu_device *adev) } switch (adev->asic_type) { -#ifdef CONFIG_DRM_AMDGPU_SI - case CHIP_VERDE: - case CHIP_TAHITI: - case CHIP_PITCAIRN: - case CHIP_OLAND: - case CHIP_HAINAN: -#endif -#ifdef CONFIG_DRM_AMDGPU_CIK - case CHIP_BONAIRE: - case CHIP_HAWAII: - case CHIP_KAVERI: - case CHIP_KABINI: - case CHIP_MULLINS: -#endif - case CHIP_TOPAZ: - case CHIP_TONGA: - case CHIP_FIJI: - case CHIP_POLARIS10: - case CHIP_POLARIS11: - case CHIP_POLARIS12: - case CHIP_VEGAM: - case CHIP_CARRIZO: - case CHIP_STONEY: - case CHIP_VEGA20: - case CHIP_ALDEBARAN: - case CHIP_SIENNA_CICHLID: - case CHIP_NAVY_FLOUNDER: - case CHIP_DIMGREY_CAVEFISH: - case CHIP_BEIGE_GOBY: default: return 0; case CHIP_VEGA10: @@ -1953,27 +1964,9 @@ static int amdgpu_device_parse_gpu_info_fw(struct amdgpu_device *adev) case CHIP_ARCTURUS: chip_name = "arcturus"; break; - case CHIP_RENOIR: - if (adev->apu_flags & AMD_APU_IS_RENOIR) - chip_name = "renoir"; - else - chip_name = "green_sardine"; - break; - case CHIP_NAVI10: - chip_name = "navi10"; - break; - case CHIP_NAVI14: - chip_name = "navi14"; - break; case CHIP_NAVI12: chip_name = "navi12"; break; - case CHIP_VANGOGH: - chip_name = "vangogh"; - break; - case CHIP_YELLOW_CARP: - chip_name = "yellow_carp"; - break; } snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_gpu_info.bin", chip_name); @@ -2071,6 +2064,8 @@ out: */ static int amdgpu_device_ip_early_init(struct amdgpu_device *adev) { + struct drm_device *dev = adev_to_drm(adev); + struct pci_dev *parent; int i, r; amdgpu_device_enable_virtual_display(adev); @@ -2135,6 +2130,18 @@ static int amdgpu_device_ip_early_init(struct amdgpu_device *adev) break; } + if (amdgpu_has_atpx() && + (amdgpu_is_atpx_hybrid() || + amdgpu_has_atpx_dgpu_power_cntl()) && + ((adev->flags & AMD_IS_APU) == 0) && + !pci_is_thunderbolt_attached(to_pci_dev(dev->dev))) + adev->flags |= AMD_IS_PX; + + if (!(adev->flags & AMD_IS_APU)) { + parent = pci_upstream_bridge(adev->pdev); + adev->has_pr3 = parent ? pci_pr3_present(parent) : false; + } + amdgpu_amdkfd_device_probe(adev); adev->pm.pp_feature = amdgpu_pp_feature_mask; @@ -2285,6 +2292,49 @@ static int amdgpu_device_fw_loading(struct amdgpu_device *adev) return r; } +static int amdgpu_device_init_schedulers(struct amdgpu_device *adev) +{ + long timeout; + int r, i; + + for (i = 0; i < AMDGPU_MAX_RINGS; ++i) { + struct amdgpu_ring *ring = adev->rings[i]; + + /* No need to setup the GPU scheduler for rings that don't need it */ + if (!ring || ring->no_scheduler) + continue; + + switch (ring->funcs->type) { + case AMDGPU_RING_TYPE_GFX: + timeout = adev->gfx_timeout; + break; + case AMDGPU_RING_TYPE_COMPUTE: + timeout = adev->compute_timeout; + break; + case AMDGPU_RING_TYPE_SDMA: + timeout = adev->sdma_timeout; + break; + default: + timeout = adev->video_timeout; + break; + } + + r = drm_sched_init(&ring->sched, &amdgpu_sched_ops, + ring->num_hw_submission, amdgpu_job_hang_limit, + timeout, adev->reset_domain->wq, + ring->sched_score, ring->name, + adev->dev); + if (r) { + DRM_ERROR("Failed to create scheduler on ring %s.\n", + ring->name); + return r; + } + } + + return 0; +} + + /** * amdgpu_device_ip_init - run init for hardware IPs * @@ -2315,8 +2365,16 @@ static int amdgpu_device_ip_init(struct amdgpu_device *adev) } adev->ip_blocks[i].status.sw = true; - /* need to do gmc hw init early so we can allocate gpu mem */ - if (adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_GMC) { + if (adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_COMMON) { + /* need to do common hw init early so everything is set up for gmc */ + r = adev->ip_blocks[i].version->funcs->hw_init((void *)adev); + if (r) { + DRM_ERROR("hw_init %d failed %d\n", i, r); + goto init_failed; + } + adev->ip_blocks[i].status.hw = true; + } else if (adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_GMC) { + /* need to do gmc hw init early so we can allocate gpu mem */ /* Try to reserve bad pages early */ if (amdgpu_sriov_vf(adev)) amdgpu_virt_exchange_data(adev); @@ -2352,7 +2410,7 @@ static int amdgpu_device_ip_init(struct amdgpu_device *adev) } if (amdgpu_sriov_vf(adev)) - amdgpu_virt_exchange_data(adev); + amdgpu_virt_init_data_exchange(adev); r = amdgpu_ib_pool_init(adev); if (r) { @@ -2396,8 +2454,32 @@ static int amdgpu_device_ip_init(struct amdgpu_device *adev) if (r) goto init_failed; - if (adev->gmc.xgmi.num_physical_nodes > 1) - amdgpu_xgmi_add_device(adev); + /** + * In case of XGMI grab extra reference for reset domain for this device + */ + if (adev->gmc.xgmi.num_physical_nodes > 1) { + if (amdgpu_xgmi_add_device(adev) == 0) { + if (!amdgpu_sriov_vf(adev)) { + struct amdgpu_hive_info *hive = amdgpu_get_xgmi_hive(adev); + + if (!hive->reset_domain || + !amdgpu_reset_get_reset_domain(hive->reset_domain)) { + r = -ENOENT; + amdgpu_put_xgmi_hive(hive); + goto init_failed; + } + + /* Drop the early temporary reset domain we created for device */ + amdgpu_reset_put_reset_domain(adev->reset_domain); + adev->reset_domain = hive->reset_domain; + amdgpu_put_xgmi_hive(hive); + } + } + } + + r = amdgpu_device_init_schedulers(adev); + if (r) + goto init_failed; /* Don't init kfd if whole hive need to be reset during init */ if (!adev->gmc.xgmi.pending_reset) @@ -2608,6 +2690,12 @@ static int amdgpu_device_ip_late_init(struct amdgpu_device *adev) adev->ip_blocks[i].status.late_initialized = true; } + r = amdgpu_ras_late_init(adev); + if (r) { + DRM_ERROR("amdgpu_ras_late_init failed %d", r); + return r; + } + amdgpu_ras_set_error_query_ready(adev, true); amdgpu_device_set_cg_state(adev, AMD_CG_STATE_GATE); @@ -2622,7 +2710,7 @@ static int amdgpu_device_ip_late_init(struct amdgpu_device *adev) /* For passthrough configuration on arcturus and aldebaran, enable special handling SBR */ if (amdgpu_passthrough(adev) && ((adev->asic_type == CHIP_ARCTURUS && adev->gmc.xgmi.num_physical_nodes > 1)|| adev->asic_type == CHIP_ALDEBARAN )) - smu_handle_passthrough_sbr(&adev->smu, true); + amdgpu_dpm_handle_passthrough_sbr(adev, true); if (adev->gmc.xgmi.num_physical_nodes > 1) { mutex_lock(&mgpu_info.mutex); @@ -2706,11 +2794,11 @@ static int amdgpu_device_ip_fini_early(struct amdgpu_device *adev) } } - amdgpu_amdkfd_suspend(adev, false); - amdgpu_device_set_pg_state(adev, AMD_PG_STATE_UNGATE); amdgpu_device_set_cg_state(adev, AMD_CG_STATE_UNGATE); + amdgpu_amdkfd_suspend(adev, false); + /* Workaroud for ASICs need to disable SMC first */ amdgpu_device_smu_fini_early(adev); @@ -2840,6 +2928,14 @@ static int amdgpu_device_ip_suspend_phase1(struct amdgpu_device *adev) amdgpu_device_set_pg_state(adev, AMD_PG_STATE_UNGATE); amdgpu_device_set_cg_state(adev, AMD_CG_STATE_UNGATE); + /* + * Per PMFW team's suggestion, driver needs to handle gfxoff + * and df cstate features disablement for gpu reset(e.g. Mode1Reset) + * scenario. Add the missing df cstate disablement here. + */ + if (amdgpu_dpm_set_df_cstate(adev, DF_CSTATE_DISALLOW)) + dev_warn(adev->dev, "Failed to disallow df cstate"); + for (i = adev->num_ip_blocks - 1; i >= 0; i--) { if (!adev->ip_blocks[i].status.valid) continue; @@ -2879,7 +2975,7 @@ static int amdgpu_device_ip_suspend_phase2(struct amdgpu_device *adev) int i, r; if (adev->in_s0ix) - amdgpu_gfx_state_change_set(adev, sGpuChangeState_D3Entry); + amdgpu_dpm_gfx_state_change(adev, sGpuChangeState_D3Entry); for (i = adev->num_ip_blocks - 1; i >= 0; i--) { if (!adev->ip_blocks[i].status.valid) @@ -2974,8 +3070,8 @@ static int amdgpu_device_ip_reinit_early_sriov(struct amdgpu_device *adev) int i, r; static enum amd_ip_block_type ip_order[] = { - AMD_IP_BLOCK_TYPE_GMC, AMD_IP_BLOCK_TYPE_COMMON, + AMD_IP_BLOCK_TYPE_GMC, AMD_IP_BLOCK_TYPE_PSP, AMD_IP_BLOCK_TYPE_IH, }; @@ -3066,7 +3162,8 @@ static int amdgpu_device_ip_resume_phase1(struct amdgpu_device *adev) continue; if (adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_COMMON || adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_GMC || - adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_IH) { + adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_IH || + (adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_PSP && amdgpu_sriov_vf(adev))) { r = adev->ip_blocks[i].version->funcs->resume(adev); if (r) { @@ -3113,6 +3210,15 @@ static int amdgpu_device_ip_resume_phase2(struct amdgpu_device *adev) return r; } adev->ip_blocks[i].status.hw = true; + + if (adev->in_s0ix && adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_SMC) { + /* disable gfxoff for IP resume. The gfxoff will be re-enabled in + * amdgpu_device_resume() after IP resume. + */ + amdgpu_gfx_off_ctrl(adev, false); + DRM_DEBUG("will disable gfxoff for re-initializing other blocks\n"); + } + } return 0; @@ -3214,38 +3320,12 @@ bool amdgpu_device_asic_has_dc_support(enum amd_asic_type asic_type) case CHIP_MULLINS: /* * We have systems in the wild with these ASICs that require - * LVDS and VGA support which is not supported with DC. + * VGA support which is not supported with DC. * * Fallback to the non-DC driver here by default so as not to * cause regressions. */ return amdgpu_dc > 0; - case CHIP_HAWAII: - case CHIP_CARRIZO: - case CHIP_STONEY: - case CHIP_POLARIS10: - case CHIP_POLARIS11: - case CHIP_POLARIS12: - case CHIP_VEGAM: - case CHIP_TONGA: - case CHIP_FIJI: - case CHIP_VEGA10: - case CHIP_VEGA12: - case CHIP_VEGA20: -#if defined(CONFIG_DRM_AMD_DC_DCN) - case CHIP_RAVEN: - case CHIP_NAVI10: - case CHIP_NAVI14: - case CHIP_NAVI12: - case CHIP_RENOIR: - case CHIP_CYAN_SKILLFISH: - case CHIP_SIENNA_CICHLID: - case CHIP_NAVY_FLOUNDER: - case CHIP_DIMGREY_CAVEFISH: - case CHIP_BEIGE_GOBY: - case CHIP_VANGOGH: - case CHIP_YELLOW_CARP: -#endif default: return amdgpu_dc != 0; #else @@ -3267,7 +3347,7 @@ bool amdgpu_device_asic_has_dc_support(enum amd_asic_type asic_type) */ bool amdgpu_device_has_dc_support(struct amdgpu_device *adev) { - if (amdgpu_sriov_vf(adev) || + if (amdgpu_sriov_vf(adev) || adev->enable_virtual_display || (adev->harvest_ip_mask & AMD_HARVEST_IP_DMU_MASK)) return false; @@ -3305,9 +3385,9 @@ static void amdgpu_device_xgmi_reset_func(struct work_struct *__work) if (adev->asic_reset_res) goto fail; - if (adev->mmhub.ras_funcs && - adev->mmhub.ras_funcs->reset_ras_error_count) - adev->mmhub.ras_funcs->reset_ras_error_count(adev); + if (adev->mmhub.ras && adev->mmhub.ras->ras_block.hw_ops && + adev->mmhub.ras->ras_block.hw_ops->reset_ras_error_count) + adev->mmhub.ras->ras_block.hw_ops->reset_ras_error_count(adev); } else { task_barrier_full(&hive->tb); @@ -3449,6 +3529,7 @@ int amdgpu_device_init(struct amdgpu_device *adev, adev->gmc.gart_size = 512 * 1024 * 1024; adev->accel_working = false; adev->num_rings = 0; + RCU_INIT_POINTER(adev->gang_submit, dma_fence_get_stub()); adev->mman.buffer_funcs = NULL; adev->mman.buffer_funcs_ring = NULL; adev->vm_manager.vm_pte_funcs = NULL; @@ -3491,14 +3572,12 @@ int amdgpu_device_init(struct amdgpu_device *adev, mutex_init(&adev->mn_lock); mutex_init(&adev->virt.vf_errors.lock); hash_init(adev->mn_hash); - atomic_set(&adev->in_gpu_reset, 0); - init_rwsem(&adev->reset_sem); mutex_init(&adev->psp.mutex); mutex_init(&adev->notifier_lock); + mutex_init(&adev->pm.stable_pstate_ctx_lock); + mutex_init(&adev->benchmark_mutex); - r = amdgpu_device_init_apu_flags(adev); - if (r) - return r; + amdgpu_device_init_apu_flags(adev); r = amdgpu_device_check_arguments(adev); if (r) @@ -3519,6 +3598,8 @@ int amdgpu_device_init(struct amdgpu_device *adev, INIT_LIST_HEAD(&adev->reset_list); + INIT_LIST_HEAD(&adev->ras_list); + INIT_DELAYED_WORK(&adev->delayed_init_work, amdgpu_device_delayed_init_work_handler); INIT_DELAYED_WORK(&adev->gfx.gfx_off_delay_work, @@ -3527,6 +3608,8 @@ int amdgpu_device_init(struct amdgpu_device *adev, INIT_WORK(&adev->xgmi_reset_work, amdgpu_device_xgmi_reset_func); adev->gfx.gfx_off_req_count = 1; + adev->gfx.gfx_off_residency = 0; + adev->gfx.gfx_off_entrycount = 0; adev->pm.ac_power = power_supply_is_system_supplied() > 0; atomic_set(&adev->throttling_logging_enabled, 1); @@ -3565,8 +3648,14 @@ int amdgpu_device_init(struct amdgpu_device *adev, if (amdgpu_mcbp) DRM_INFO("MCBP is enabled\n"); - if (amdgpu_mes && adev->asic_type >= CHIP_NAVI10) - adev->enable_mes = true; + /* + * Reset domain needs to be present early, before XGMI hive discovered + * (if any) and intitialized to use reset sem and in_gpu reset flag + * early on during init and before calling to RREG32. + */ + adev->reset_domain = amdgpu_reset_create_reset_domain(SINGLE_DEVICE, "amdgpu-reset-dev"); + if (!adev->reset_domain) + return -ENOMEM; /* detect hw virtualization here */ amdgpu_detect_virtualization(adev); @@ -3582,6 +3671,10 @@ int amdgpu_device_init(struct amdgpu_device *adev, if (r) return r; + /* Enable TMZ based on IP_VERSION */ + amdgpu_gmc_tmz_set(adev); + + amdgpu_gmc_noretry_set(adev); /* Need to get xgmi info early to decide the reset behavior*/ if (adev->gmc.xgmi.supported) { r = adev->gfxhub.funcs->get_xgmi_info(adev); @@ -3592,7 +3685,7 @@ int amdgpu_device_init(struct amdgpu_device *adev, /* enable PCIE atomic ops */ if (amdgpu_sriov_vf(adev)) adev->have_atomics_support = ((struct amd_sriov_msg_pf2vf_info *) - adev->virt.fw_reserve.p_pf2vf)->pcie_atomic_ops_enabled_flags == + adev->virt.fw_reserve.p_pf2vf)->pcie_atomic_ops_support_flags == (PCI_EXP_DEVCAP2_ATOMIC_COMP32 | PCI_EXP_DEVCAP2_ATOMIC_COMP64); else adev->have_atomics_support = @@ -3749,18 +3842,13 @@ fence_driver_init: } else adev->ucode_sysfs_en = true; - if ((amdgpu_testing & 1)) { - if (adev->accel_working) - amdgpu_test_moves(adev); - else - DRM_INFO("amdgpu: acceleration disabled, skipping move tests\n"); - } - if (amdgpu_benchmarking) { - if (adev->accel_working) - amdgpu_benchmark(adev, amdgpu_benchmarking); - else - DRM_INFO("amdgpu: acceleration disabled, skipping benchmarks\n"); - } + r = amdgpu_psp_sysfs_init(adev); + if (r) { + adev->psp_sysfs_en = false; + if (!amdgpu_sriov_vf(adev)) + DRM_ERROR("Creating psp sysfs failed\n"); + } else + adev->psp_sysfs_en = true; /* * Register gpu instance before amdgpu_device_enable_mgpu_fan_boost. @@ -3833,6 +3921,7 @@ failed: static void amdgpu_device_unmap_mmio(struct amdgpu_device *adev) { + /* Clear all CPU mappings pointing to this device */ unmap_mapping_range(adev->ddev.anon_inode->i_mapping, 0, 0, 1); @@ -3864,10 +3953,6 @@ void amdgpu_device_fini_hw(struct amdgpu_device *adev) { dev_info(adev->dev, "amdgpu: finishing device.\n"); flush_delayed_work(&adev->delayed_init_work); - if (adev->mman.initialized) { - flush_delayed_work(&adev->mman.bdev.wq); - ttm_bo_lock_delayed_workqueue(&adev->mman.bdev); - } adev->shutdown = true; /* make sure IB test finished before entering exclusive mode @@ -3888,10 +3973,17 @@ void amdgpu_device_fini_hw(struct amdgpu_device *adev) } amdgpu_fence_driver_hw_fini(adev); + if (adev->mman.initialized) { + flush_delayed_work(&adev->mman.bdev.wq); + ttm_bo_lock_delayed_workqueue(&adev->mman.bdev); + } + if (adev->pm_sysfs_en) amdgpu_pm_sysfs_fini(adev); if (adev->ucode_sysfs_en) amdgpu_ucode_sysfs_fini(adev); + if (adev->psp_sysfs_en) + amdgpu_psp_sysfs_fini(adev); sysfs_remove_files(&adev->dev->kobj, amdgpu_dev_attributes); /* disable ras feature must before hw fini */ @@ -3906,18 +3998,20 @@ void amdgpu_device_fini_hw(struct amdgpu_device *adev) amdgpu_gart_dummy_page_fini(adev); - if (drm_dev_is_unplugged(adev_to_drm(adev))) - amdgpu_device_unmap_mmio(adev); + amdgpu_device_unmap_mmio(adev); } void amdgpu_device_fini_sw(struct amdgpu_device *adev) { + int idx; + amdgpu_fence_driver_sw_fini(adev); amdgpu_device_ip_fini(adev); release_firmware(adev->firmware.gpu_info_fw); adev->firmware.gpu_info_fw = NULL; adev->accel_working = false; + dma_fence_put(rcu_dereference_protected(adev->gang_submit, true)); amdgpu_reset_fini(adev); @@ -3937,11 +4031,22 @@ void amdgpu_device_fini_sw(struct amdgpu_device *adev) if ((adev->pdev->class >> 8) == PCI_CLASS_DISPLAY_VGA) vga_client_unregister(adev->pdev); + if (drm_dev_enter(adev_to_drm(adev), &idx)) { + + iounmap(adev->rmmio); + adev->rmmio = NULL; + amdgpu_device_doorbell_fini(adev); + drm_dev_exit(idx); + } + if (IS_ENABLED(CONFIG_PERF_EVENTS)) amdgpu_pmu_fini(adev); if (adev->mman.discovery_bin) amdgpu_discovery_fini(adev); + amdgpu_reset_put_reset_domain(adev->reset_domain); + adev->reset_domain = NULL; + kfree(adev->pci_state); } @@ -3955,15 +4060,18 @@ void amdgpu_device_fini_sw(struct amdgpu_device *adev) * at suspend time. * */ -static void amdgpu_device_evict_resources(struct amdgpu_device *adev) +static int amdgpu_device_evict_resources(struct amdgpu_device *adev) { - /* No need to evict vram on APUs for suspend to ram */ - if (adev->in_s3 && (adev->flags & AMD_IS_APU)) - return; + int ret; - if (amdgpu_ttm_evict_resources(adev, TTM_PL_VRAM)) - DRM_WARN("evicting device resources failed\n"); + /* No need to evict vram on APUs for suspend to ram or s2idle */ + if ((adev->in_s3 || adev->in_s0ix) && (adev->flags & AMD_IS_APU)) + return 0; + ret = amdgpu_ttm_evict_resources(adev, TTM_PL_VRAM); + if (ret) + DRM_WARN("evicting device resources failed\n"); + return ret; } /* @@ -3982,12 +4090,20 @@ static void amdgpu_device_evict_resources(struct amdgpu_device *adev) int amdgpu_device_suspend(struct drm_device *dev, bool fbcon) { struct amdgpu_device *adev = drm_to_adev(dev); + int r = 0; if (dev->switch_power_state == DRM_SWITCH_POWER_OFF) return 0; adev->in_suspend = true; + if (amdgpu_sriov_vf(adev)) { + amdgpu_virt_fini_data_exchange(adev); + r = amdgpu_virt_request_full_gpu(adev, false); + if (r) + return r; + } + if (amdgpu_acpi_smart_shift_update(dev, AMDGPU_SS_DEV_D3)) DRM_WARN("smart shift update failed\n"); @@ -4005,16 +4121,16 @@ int amdgpu_device_suspend(struct drm_device *dev, bool fbcon) if (!adev->in_s0ix) amdgpu_amdkfd_suspend(adev, adev->in_runpm); - /* First evict vram memory */ - amdgpu_device_evict_resources(adev); + r = amdgpu_device_evict_resources(adev); + if (r) + return r; amdgpu_fence_driver_hw_fini(adev); amdgpu_device_ip_suspend_phase2(adev); - /* This second call to evict device resources is to evict - * the gart page table using the CPU. - */ - amdgpu_device_evict_resources(adev); + + if (amdgpu_sriov_vf(adev)) + amdgpu_virt_release_full_gpu(adev, false); return 0; } @@ -4034,11 +4150,17 @@ int amdgpu_device_resume(struct drm_device *dev, bool fbcon) struct amdgpu_device *adev = drm_to_adev(dev); int r = 0; + if (amdgpu_sriov_vf(adev)) { + r = amdgpu_virt_request_full_gpu(adev, true); + if (r) + return r; + } + if (dev->switch_power_state == DRM_SWITCH_POWER_OFF) return 0; if (adev->in_s0ix) - amdgpu_gfx_state_change_set(adev, sGpuChangeState_D0Entry); + amdgpu_dpm_gfx_state_change(adev, sGpuChangeState_D0Entry); /* post card */ if (amdgpu_device_need_post(adev)) { @@ -4048,6 +4170,13 @@ int amdgpu_device_resume(struct drm_device *dev, bool fbcon) } r = amdgpu_device_ip_resume(adev); + + /* no matter what r is, always need to properly release full GPU */ + if (amdgpu_sriov_vf(adev)) { + amdgpu_virt_init_data_exchange(adev); + amdgpu_virt_release_full_gpu(adev, true); + } + if (r) { dev_err(adev->dev, "amdgpu_device_ip_resume failed (%d).\n", r); return r; @@ -4070,6 +4199,13 @@ int amdgpu_device_resume(struct drm_device *dev, bool fbcon) /* Make sure IB tests flushed */ flush_delayed_work(&adev->delayed_init_work); + if (adev->in_s0ix) { + /* re-enable gfxoff after IP resume. This re-enables gfxoff after + * it was disabled for IP resume in amdgpu_device_ip_resume_phase2(). + */ + amdgpu_gfx_off_ctrl(adev, true); + DRM_DEBUG("will enable gfxoff for the mission mode\n"); + } if (fbcon) drm_fb_helper_set_suspend_unlocked(adev_to_drm(adev)->fb_helper, false); @@ -4341,9 +4477,9 @@ static int amdgpu_device_reset_sriov(struct amdgpu_device *adev, { int r; struct amdgpu_hive_info *hive = NULL; + int retry_limit = 0; - amdgpu_amdkfd_pre_reset(adev); - +retry: amdgpu_amdkfd_pre_reset(adev); if (from_hypervisor) @@ -4359,8 +4495,6 @@ static int amdgpu_device_reset_sriov(struct amdgpu_device *adev, goto error; amdgpu_virt_init_data_exchange(adev); - /* we need recover gart prior to run SMC/CP/SDMA resume */ - amdgpu_gtt_mgr_recover(ttm_manager_type(&adev->mman.bdev, TTM_PL_TT)); r = amdgpu_device_fw_loading(adev); if (r) @@ -4382,6 +4516,7 @@ static int amdgpu_device_reset_sriov(struct amdgpu_device *adev, if (!r) { amdgpu_irq_gpu_reset_resume_helper(adev); r = amdgpu_ib_ring_tests(adev); + amdgpu_amdkfd_post_reset(adev); } @@ -4392,6 +4527,14 @@ error: } amdgpu_virt_release_full_gpu(adev, true); + if (AMDGPU_RETRY_SRIOV_RESET(r)) { + if (retry_limit < AMDGPU_MAX_RETRY_LIMIT) { + retry_limit++; + goto retry; + } else + DRM_ERROR("GPU reset retry is beyond the retry limit\n"); + } + return r; } @@ -4433,46 +4576,38 @@ bool amdgpu_device_has_job_running(struct amdgpu_device *adev) */ bool amdgpu_device_should_recover_gpu(struct amdgpu_device *adev) { - if (!amdgpu_device_ip_check_soft_reset(adev)) { - dev_info(adev->dev, "Timeout, but no hardware hang detected.\n"); - return false; - } if (amdgpu_gpu_recovery == 0) goto disabled; + if (!amdgpu_device_ip_check_soft_reset(adev)) { + dev_info(adev->dev,"Timeout, but no hardware hang detected.\n"); + return false; + } + if (amdgpu_sriov_vf(adev)) return true; if (amdgpu_gpu_recovery == -1) { switch (adev->asic_type) { - case CHIP_BONAIRE: - case CHIP_HAWAII: - case CHIP_TOPAZ: - case CHIP_TONGA: - case CHIP_FIJI: - case CHIP_POLARIS10: - case CHIP_POLARIS11: - case CHIP_POLARIS12: - case CHIP_VEGAM: - case CHIP_VEGA20: - case CHIP_VEGA10: - case CHIP_VEGA12: - case CHIP_RAVEN: - case CHIP_ARCTURUS: - case CHIP_RENOIR: - case CHIP_NAVI10: - case CHIP_NAVI14: - case CHIP_NAVI12: - case CHIP_SIENNA_CICHLID: - case CHIP_NAVY_FLOUNDER: - case CHIP_DIMGREY_CAVEFISH: - case CHIP_BEIGE_GOBY: - case CHIP_VANGOGH: - case CHIP_ALDEBARAN: - break; - default: +#ifdef CONFIG_DRM_AMDGPU_SI + case CHIP_VERDE: + case CHIP_TAHITI: + case CHIP_PITCAIRN: + case CHIP_OLAND: + case CHIP_HAINAN: +#endif +#ifdef CONFIG_DRM_AMDGPU_CIK + case CHIP_KAVERI: + case CHIP_KABINI: + case CHIP_MULLINS: +#endif + case CHIP_CARRIZO: + case CHIP_STONEY: + case CHIP_CYAN_SKILLFISH: goto disabled; + default: + break; } } @@ -4539,6 +4674,8 @@ int amdgpu_device_pre_asic_reset(struct amdgpu_device *adev, amdgpu_virt_fini_data_exchange(adev); } + amdgpu_fence_driver_isr_toggle(adev, true); + /* block all schedulers and reset given job's ring */ for (i = 0; i < AMDGPU_MAX_RINGS; ++i) { struct amdgpu_ring *ring = adev->rings[i]; @@ -4554,6 +4691,8 @@ int amdgpu_device_pre_asic_reset(struct amdgpu_device *adev, amdgpu_fence_driver_force_completion(ring); } + amdgpu_fence_driver_isr_toggle(adev, false); + if (job && job->vm) drm_sched_increase_karma(&job->base); @@ -4570,7 +4709,7 @@ int amdgpu_device_pre_asic_reset(struct amdgpu_device *adev, if (!need_full_reset) need_full_reset = amdgpu_device_ip_need_full_reset(adev); - if (!need_full_reset) { + if (!need_full_reset && amdgpu_gpu_recovery) { amdgpu_device_ip_pre_soft_reset(adev); r = amdgpu_device_ip_soft_reset(adev); amdgpu_device_ip_post_soft_reset(adev); @@ -4592,16 +4731,88 @@ int amdgpu_device_pre_asic_reset(struct amdgpu_device *adev, return r; } +static int amdgpu_reset_reg_dumps(struct amdgpu_device *adev) +{ + int i; + + lockdep_assert_held(&adev->reset_domain->sem); + + for (i = 0; i < adev->num_regs; i++) { + adev->reset_dump_reg_value[i] = RREG32(adev->reset_dump_reg_list[i]); + trace_amdgpu_reset_reg_dumps(adev->reset_dump_reg_list[i], + adev->reset_dump_reg_value[i]); + } + + return 0; +} + +#ifdef CONFIG_DEV_COREDUMP +static ssize_t amdgpu_devcoredump_read(char *buffer, loff_t offset, + size_t count, void *data, size_t datalen) +{ + struct drm_printer p; + struct amdgpu_device *adev = data; + struct drm_print_iterator iter; + int i; + + iter.data = buffer; + iter.offset = 0; + iter.start = offset; + iter.remain = count; + + p = drm_coredump_printer(&iter); + + drm_printf(&p, "**** AMDGPU Device Coredump ****\n"); + drm_printf(&p, "kernel: " UTS_RELEASE "\n"); + drm_printf(&p, "module: " KBUILD_MODNAME "\n"); + drm_printf(&p, "time: %lld.%09ld\n", adev->reset_time.tv_sec, adev->reset_time.tv_nsec); + if (adev->reset_task_info.pid) + drm_printf(&p, "process_name: %s PID: %d\n", + adev->reset_task_info.process_name, + adev->reset_task_info.pid); + + if (adev->reset_vram_lost) + drm_printf(&p, "VRAM is lost due to GPU reset!\n"); + if (adev->num_regs) { + drm_printf(&p, "AMDGPU register dumps:\nOffset: Value:\n"); + + for (i = 0; i < adev->num_regs; i++) + drm_printf(&p, "0x%08x: 0x%08x\n", + adev->reset_dump_reg_list[i], + adev->reset_dump_reg_value[i]); + } + + return count - iter.remain; +} + +static void amdgpu_devcoredump_free(void *data) +{ +} + +static void amdgpu_reset_capture_coredumpm(struct amdgpu_device *adev) +{ + struct drm_device *dev = adev_to_drm(adev); + + ktime_get_ts64(&adev->reset_time); + dev_coredumpm(dev->dev, THIS_MODULE, adev, 0, GFP_KERNEL, + amdgpu_devcoredump_read, amdgpu_devcoredump_free); +} +#endif + int amdgpu_do_asic_reset(struct list_head *device_list_handle, struct amdgpu_reset_context *reset_context) { struct amdgpu_device *tmp_adev = NULL; bool need_full_reset, skip_hw_reset, vram_lost = false; int r = 0; + bool gpu_reset_for_dev_remove = 0; /* Try reset handler method first */ tmp_adev = list_first_entry(device_list_handle, struct amdgpu_device, reset_list); + amdgpu_reset_reg_dumps(tmp_adev); + + reset_context->reset_device_list = device_list_handle; r = amdgpu_reset_perform_reset(tmp_adev, reset_context); /* If reset handler not implemented, continue; otherwise return */ if (r == -ENOSYS) @@ -4614,6 +4825,10 @@ int amdgpu_do_asic_reset(struct list_head *device_list_handle, test_bit(AMDGPU_NEED_FULL_RESET, &reset_context->flags); skip_hw_reset = test_bit(AMDGPU_SKIP_HW_RESET, &reset_context->flags); + gpu_reset_for_dev_remove = + test_bit(AMDGPU_RESET_FOR_DEVICE_REMOVE, &reset_context->flags) && + test_bit(AMDGPU_NEED_FULL_RESET, &reset_context->flags); + /* * ASIC reset has to be done on all XGMI hive nodes ASAP * to allow proper links negotiation in FW (within 1 sec) @@ -4650,14 +4865,26 @@ int amdgpu_do_asic_reset(struct list_head *device_list_handle, if (!r && amdgpu_ras_intr_triggered()) { list_for_each_entry(tmp_adev, device_list_handle, reset_list) { - if (tmp_adev->mmhub.ras_funcs && - tmp_adev->mmhub.ras_funcs->reset_ras_error_count) - tmp_adev->mmhub.ras_funcs->reset_ras_error_count(tmp_adev); + if (tmp_adev->mmhub.ras && tmp_adev->mmhub.ras->ras_block.hw_ops && + tmp_adev->mmhub.ras->ras_block.hw_ops->reset_ras_error_count) + tmp_adev->mmhub.ras->ras_block.hw_ops->reset_ras_error_count(tmp_adev); } amdgpu_ras_intr_cleared(); } + /* Since the mode1 reset affects base ip blocks, the + * phase1 ip blocks need to be resumed. Otherwise there + * will be a BIOS signature error and the psp bootloader + * can't load kdb on the next amdgpu install. + */ + if (gpu_reset_for_dev_remove) { + list_for_each_entry(tmp_adev, device_list_handle, reset_list) + amdgpu_device_ip_resume_phase1(tmp_adev); + + goto end; + } + list_for_each_entry(tmp_adev, device_list_handle, reset_list) { if (need_full_reset) { /* post card */ @@ -4675,15 +4902,20 @@ int amdgpu_do_asic_reset(struct list_head *device_list_handle, goto out; vram_lost = amdgpu_device_check_vram_lost(tmp_adev); +#ifdef CONFIG_DEV_COREDUMP + tmp_adev->reset_vram_lost = vram_lost; + memset(&tmp_adev->reset_task_info, 0, + sizeof(tmp_adev->reset_task_info)); + if (reset_context->job && reset_context->job->vm) + tmp_adev->reset_task_info = + reset_context->job->vm->task_info; + amdgpu_reset_capture_coredumpm(tmp_adev); +#endif if (vram_lost) { DRM_INFO("VRAM is lost due to GPU reset!\n"); amdgpu_inc_vram_lost(tmp_adev); } - r = amdgpu_gtt_mgr_recover(ttm_manager_type(&tmp_adev->mman.bdev, TTM_PL_TT)); - if (r) - goto out; - r = amdgpu_device_fw_loading(tmp_adev); if (r) return r; @@ -4763,17 +4995,8 @@ end: return r; } -static bool amdgpu_device_lock_adev(struct amdgpu_device *adev, - struct amdgpu_hive_info *hive) +static void amdgpu_device_set_mp1_state(struct amdgpu_device *adev) { - if (atomic_cmpxchg(&adev->in_gpu_reset, 0, 1) != 0) - return false; - - if (hive) { - down_write_nest_lock(&adev->reset_sem, &hive->hive_lock); - } else { - down_write(&adev->reset_sem); - } switch (amdgpu_asic_reset_method(adev)) { case AMD_RESET_METHOD_MODE1: @@ -4786,56 +5009,12 @@ static bool amdgpu_device_lock_adev(struct amdgpu_device *adev, adev->mp1_state = PP_MP1_STATE_NONE; break; } - - return true; } -static void amdgpu_device_unlock_adev(struct amdgpu_device *adev) +static void amdgpu_device_unset_mp1_state(struct amdgpu_device *adev) { amdgpu_vf_error_trans_all(adev); adev->mp1_state = PP_MP1_STATE_NONE; - atomic_set(&adev->in_gpu_reset, 0); - up_write(&adev->reset_sem); -} - -/* - * to lockup a list of amdgpu devices in a hive safely, if not a hive - * with multiple nodes, it will be similar as amdgpu_device_lock_adev. - * - * unlock won't require roll back. - */ -static int amdgpu_device_lock_hive_adev(struct amdgpu_device *adev, struct amdgpu_hive_info *hive) -{ - struct amdgpu_device *tmp_adev = NULL; - - if (!amdgpu_sriov_vf(adev) && (adev->gmc.xgmi.num_physical_nodes > 1)) { - if (!hive) { - dev_err(adev->dev, "Hive is NULL while device has multiple xgmi nodes"); - return -ENODEV; - } - list_for_each_entry(tmp_adev, &hive->device_list, gmc.xgmi.head) { - if (!amdgpu_device_lock_adev(tmp_adev, hive)) - goto roll_back; - } - } else if (!amdgpu_device_lock_adev(adev, hive)) - return -EAGAIN; - - return 0; -roll_back: - if (!list_is_first(&tmp_adev->gmc.xgmi.head, &hive->device_list)) { - /* - * if the lockup iteration break in the middle of a hive, - * it may means there may has a race issue, - * or a hive device locked up independently. - * we may be in trouble and may not, so will try to roll back - * the lock and give out a warnning. - */ - dev_warn(tmp_adev->dev, "Hive lock iteration broke in the middle. Rolling back to unlock"); - list_for_each_entry_continue_reverse(tmp_adev, &hive->device_list, gmc.xgmi.head) { - amdgpu_device_unlock_adev(tmp_adev); - } - } - return -EAGAIN; } static void amdgpu_device_resume_display_audio(struct amdgpu_device *adev) @@ -4917,18 +5096,35 @@ static void amdgpu_device_recheck_guilty_jobs( /* clear job's guilty and depend the folowing step to decide the real one */ drm_sched_reset_karma(s_job); - /* for the real bad job, it will be resubmitted twice, adding a dma_fence_get - * to make sure fence is balanced */ - dma_fence_get(s_job->s_fence->parent); drm_sched_resubmit_jobs_ext(&ring->sched, 1); + if (!s_job->s_fence->parent) { + DRM_WARN("Failed to get a HW fence for job!"); + continue; + } + ret = dma_fence_wait_timeout(s_job->s_fence->parent, false, ring->sched.timeout); if (ret == 0) { /* timeout */ DRM_ERROR("Found the real bad job! ring:%s, job_id:%llx\n", ring->sched.name, s_job->id); + + amdgpu_fence_driver_isr_toggle(adev, true); + + /* Clear this failed job from fence array */ + amdgpu_fence_driver_clear_job_fences(ring); + + amdgpu_fence_driver_isr_toggle(adev, false); + + /* Since the job won't signal and we go for + * another resubmit drop this parent pointer + */ + dma_fence_put(s_job->s_fence->parent); + s_job->s_fence->parent = NULL; + /* set guilty */ drm_sched_increase_karma(s_job); + amdgpu_reset_prepare_hwcontext(adev, reset_context); retry: /* do hw reset */ if (amdgpu_sriov_vf(adev)) { @@ -4955,7 +5151,6 @@ retry: /* got the hw fence, signal finished fence */ atomic_dec(ring->sched.score); - dma_fence_put(s_job->s_fence->parent); dma_fence_get(&s_job->s_fence->finished); dma_fence_signal(&s_job->s_fence->finished); dma_fence_put(&s_job->s_fence->finished); @@ -4968,6 +5163,27 @@ retry: } } +static inline void amdgpu_device_stop_pending_resets(struct amdgpu_device *adev) +{ + struct amdgpu_ras *con = amdgpu_ras_get_context(adev); + +#if defined(CONFIG_DEBUG_FS) + if (!amdgpu_sriov_vf(adev)) + cancel_work(&adev->reset_work); +#endif + + if (adev->kfd.dev) + cancel_work(&adev->kfd.reset_work); + + if (amdgpu_sriov_vf(adev)) + cancel_work(&adev->virt.flr_work); + + if (con && adev->ras_enabled) + cancel_work(&con->recovery_work); + +} + + /** * amdgpu_device_gpu_recover - reset the asic and recover scheduler * @@ -4980,7 +5196,8 @@ retry: */ int amdgpu_device_gpu_recover(struct amdgpu_device *adev, - struct amdgpu_job *job) + struct amdgpu_job *job, + struct amdgpu_reset_context *reset_context) { struct list_head device_list, *device_list_handle = NULL; bool job_signaled = false; @@ -4990,9 +5207,11 @@ int amdgpu_device_gpu_recover(struct amdgpu_device *adev, bool need_emergency_restart = false; bool audio_suspended = false; int tmp_vram_lost_counter; - struct amdgpu_reset_context reset_context; + bool gpu_reset_for_dev_remove = false; - memset(&reset_context, 0, sizeof(reset_context)); + gpu_reset_for_dev_remove = + test_bit(AMDGPU_RESET_FOR_DEVICE_REMOVE, &reset_context->flags) && + test_bit(AMDGPU_NEED_FULL_RESET, &reset_context->flags); /* * Special case: RAS triggered and full reset isn't supported @@ -5013,49 +5232,13 @@ int amdgpu_device_gpu_recover(struct amdgpu_device *adev, dev_info(adev->dev, "GPU %s begin!\n", need_emergency_restart ? "jobs stop":"reset"); - /* - * Here we trylock to avoid chain of resets executing from - * either trigger by jobs on different adevs in XGMI hive or jobs on - * different schedulers for same device while this TO handler is running. - * We always reset all schedulers for device and all devices for XGMI - * hive so that should take care of them too. - */ if (!amdgpu_sriov_vf(adev)) hive = amdgpu_get_xgmi_hive(adev); - if (hive) { - if (atomic_cmpxchg(&hive->in_reset, 0, 1) != 0) { - DRM_INFO("Bailing on TDR for s_job:%llx, hive: %llx as another already in progress", - job ? job->base.id : -1, hive->hive_id); - amdgpu_put_xgmi_hive(hive); - if (job && job->vm) - drm_sched_increase_karma(&job->base); - return 0; - } + if (hive) mutex_lock(&hive->hive_lock); - } - - reset_context.method = AMD_RESET_METHOD_NONE; - reset_context.reset_req_dev = adev; - reset_context.job = job; - reset_context.hive = hive; - clear_bit(AMDGPU_NEED_FULL_RESET, &reset_context.flags); - - /* - * lock the device before we try to operate the linked list - * if didn't get the device lock, don't touch the linked list since - * others may iterating it. - */ - r = amdgpu_device_lock_hive_adev(adev, hive); - if (r) { - dev_info(adev->dev, "Bailing on TDR for s_job:%llx, as another already in progress", - job ? job->base.id : -1); - - /* even we skipped this reset, still need to set the job to guilty */ - if (job && job->vm) - drm_sched_increase_karma(&job->base); - goto skip_recovery; - } + reset_context->job = job; + reset_context->hive = hive; /* * Build list of devices to reset. * In case we are in XGMI hive mode, resort the device list @@ -5063,8 +5246,11 @@ int amdgpu_device_gpu_recover(struct amdgpu_device *adev, */ INIT_LIST_HEAD(&device_list); if (!amdgpu_sriov_vf(adev) && (adev->gmc.xgmi.num_physical_nodes > 1)) { - list_for_each_entry(tmp_adev, &hive->device_list, gmc.xgmi.head) + list_for_each_entry(tmp_adev, &hive->device_list, gmc.xgmi.head) { list_add_tail(&tmp_adev->reset_list, &device_list); + if (gpu_reset_for_dev_remove && adev->shutdown) + tmp_adev->shutdown = true; + } if (!list_is_first(&adev->reset_list, &device_list)) list_rotate_to_front(&adev->reset_list, &device_list); device_list_handle = &device_list; @@ -5073,8 +5259,16 @@ int amdgpu_device_gpu_recover(struct amdgpu_device *adev, device_list_handle = &device_list; } + /* We need to lock reset domain only once both for XGMI and single device */ + tmp_adev = list_first_entry(device_list_handle, struct amdgpu_device, + reset_list); + amdgpu_device_lock_reset_domain(tmp_adev->reset_domain); + /* block all schedulers and reset given job's ring */ list_for_each_entry(tmp_adev, device_list_handle, reset_list) { + + amdgpu_device_set_mp1_state(tmp_adev); + /* * Try to put the audio codec into suspend state * before gpu reset started. @@ -5101,7 +5295,7 @@ int amdgpu_device_gpu_recover(struct amdgpu_device *adev, */ amdgpu_unregister_gpu_instance(tmp_adev); - drm_fb_helper_set_suspend_unlocked(adev_to_drm(adev)->fb_helper, true); + drm_fb_helper_set_suspend_unlocked(adev_to_drm(tmp_adev)->fb_helper, true); /* disable ras on ALL IPs */ if (!need_emergency_restart && @@ -5131,8 +5325,7 @@ int amdgpu_device_gpu_recover(struct amdgpu_device *adev, * * job->base holds a reference to parent fence */ - if (job && job->base.s_fence->parent && - dma_fence_is_signaled(job->base.s_fence->parent)) { + if (job && dma_fence_is_signaled(&job->hw_fence)) { job_signaled = true; dev_info(adev->dev, "Guilty job already signaled, skipping HW reset"); goto skip_hw_reset; @@ -5140,13 +5333,23 @@ int amdgpu_device_gpu_recover(struct amdgpu_device *adev, retry: /* Rest of adevs pre asic reset from XGMI hive. */ list_for_each_entry(tmp_adev, device_list_handle, reset_list) { - r = amdgpu_device_pre_asic_reset(tmp_adev, &reset_context); + if (gpu_reset_for_dev_remove) { + /* Workaroud for ASICs need to disable SMC first */ + amdgpu_device_smu_fini_early(tmp_adev); + } + r = amdgpu_device_pre_asic_reset(tmp_adev, reset_context); /*TODO Should we stop ?*/ if (r) { dev_err(tmp_adev->dev, "GPU pre asic reset failed with err, %d for drm dev, %s ", r, adev_to_drm(tmp_adev)->unique); tmp_adev->asic_reset_res = r; } + + /* + * Drop all pending non scheduler resets. Scheduler resets + * were already dropped during drm_sched_stop + */ + amdgpu_device_stop_pending_resets(tmp_adev); } tmp_vram_lost_counter = atomic_read(&((adev)->vram_lost_counter)); @@ -5156,10 +5359,17 @@ retry: /* Rest of adevs pre asic reset from XGMI hive. */ r = amdgpu_device_reset_sriov(adev, job ? false : true); if (r) adev->asic_reset_res = r; + + /* Aldebaran supports ras in SRIOV, so need resume ras during reset */ + if (adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 4, 2)) + amdgpu_ras_resume(adev); } else { - r = amdgpu_do_asic_reset(device_list_handle, &reset_context); + r = amdgpu_do_asic_reset(device_list_handle, reset_context); if (r && r == -EAGAIN) goto retry; + + if (!r && gpu_reset_for_dev_remove) + goto recover_end; } skip_hw_reset: @@ -5177,7 +5387,7 @@ skip_hw_reset: if (amdgpu_gpu_recovery == 2 && !(tmp_vram_lost_counter < atomic_read(&adev->vram_lost_counter))) amdgpu_device_recheck_guilty_jobs( - tmp_adev, device_list_handle, &reset_context); + tmp_adev, device_list_handle, reset_context); for (i = 0; i < AMDGPU_MAX_RINGS; ++i) { struct amdgpu_ring *ring = tmp_adev->rings[i]; @@ -5192,10 +5402,16 @@ skip_hw_reset: drm_sched_start(&ring->sched, !tmp_adev->asic_reset_res); } + if (adev->enable_mes && adev->ip_versions[GC_HWIP][0] != IP_VERSION(11, 0, 3)) + amdgpu_mes_self_test(tmp_adev); + if (!drm_drv_uses_atomic_modeset(adev_to_drm(tmp_adev)) && !job_signaled) { drm_helper_resume_force_mode(adev_to_drm(tmp_adev)); } + if (tmp_adev->asic_reset_res) + r = tmp_adev->asic_reset_res; + tmp_adev->asic_reset_res = 0; if (r) { @@ -5223,18 +5439,24 @@ skip_sched_resume: if (audio_suspended) amdgpu_device_resume_display_audio(tmp_adev); - amdgpu_device_unlock_adev(tmp_adev); + + amdgpu_device_unset_mp1_state(tmp_adev); } -skip_recovery: +recover_end: + tmp_adev = list_first_entry(device_list_handle, struct amdgpu_device, + reset_list); + amdgpu_device_unlock_reset_domain(tmp_adev->reset_domain); + if (hive) { - atomic_set(&hive->in_reset, 0); mutex_unlock(&hive->hive_lock); amdgpu_put_xgmi_hive(hive); } - if (r && r != -EAGAIN) + if (r) dev_info(adev->dev, "GPU reset end with ret = %d\n", r); + + atomic_set(&adev->reset_domain->reset_res, r); return r; } @@ -5386,6 +5608,37 @@ static void amdgpu_device_get_pcie_info(struct amdgpu_device *adev) } } +/** + * amdgpu_device_is_peer_accessible - Check peer access through PCIe BAR + * + * @adev: amdgpu_device pointer + * @peer_adev: amdgpu_device pointer for peer device trying to access @adev + * + * Return true if @peer_adev can access (DMA) @adev through the PCIe + * BAR, i.e. @adev is "large BAR" and the BAR matches the DMA mask of + * @peer_adev. + */ +bool amdgpu_device_is_peer_accessible(struct amdgpu_device *adev, + struct amdgpu_device *peer_adev) +{ +#ifdef CONFIG_HSA_AMD_P2P + uint64_t address_mask = peer_adev->dev->dma_mask ? + ~*peer_adev->dev->dma_mask : ~((1ULL << 32) - 1); + resource_size_t aper_limit = + adev->gmc.aper_base + adev->gmc.aper_size - 1; + bool p2p_access = + !adev->gmc.xgmi.connected_to_cpu && + !(pci_p2pdma_distance(adev->pdev, peer_adev->dev, false) < 0); + + return pcie_p2p && p2p_access && (adev->gmc.visible_vram_size && + adev->gmc.real_vram_size == adev->gmc.visible_vram_size && + !(adev->gmc.aper_base & address_mask || + aper_limit & address_mask)); +#else + return false; +#endif +} + int amdgpu_device_baco_enter(struct drm_device *dev) { struct amdgpu_device *adev = drm_to_adev(dev); @@ -5425,20 +5678,6 @@ int amdgpu_device_baco_exit(struct drm_device *dev) return 0; } -static void amdgpu_cancel_all_tdr(struct amdgpu_device *adev) -{ - int i; - - for (i = 0; i < AMDGPU_MAX_RINGS; ++i) { - struct amdgpu_ring *ring = adev->rings[i]; - - if (!ring || !ring->sched.thread) - continue; - - cancel_delayed_work_sync(&ring->sched.work_tdr); - } -} - /** * amdgpu_pci_error_detected - Called when a PCI error is detected. * @pdev: PCI device struct @@ -5469,14 +5708,11 @@ pci_ers_result_t amdgpu_pci_error_detected(struct pci_dev *pdev, pci_channel_sta /* Fatal error, prepare for slot reset */ case pci_channel_io_frozen: /* - * Cancel and wait for all TDRs in progress if failing to - * set adev->in_gpu_reset in amdgpu_device_lock_adev - * - * Locking adev->reset_sem will prevent any external access + * Locking adev->reset_domain->sem will prevent any external access * to GPU during PCI error recovery */ - while (!amdgpu_device_lock_adev(adev, NULL)) - amdgpu_cancel_all_tdr(adev); + amdgpu_device_lock_reset_domain(adev->reset_domain); + amdgpu_device_set_mp1_state(adev); /* * Block any work scheduling as we do for regular GPU reset @@ -5583,7 +5819,8 @@ out: DRM_INFO("PCIe error recovery succeeded\n"); } else { DRM_ERROR("PCIe error recovery failed, err:%d", r); - amdgpu_device_unlock_adev(adev); + amdgpu_device_unset_mp1_state(adev); + amdgpu_device_unlock_reset_domain(adev->reset_domain); } return r ? PCI_ERS_RESULT_DISCONNECT : PCI_ERS_RESULT_RECOVERED; @@ -5620,7 +5857,8 @@ void amdgpu_pci_resume(struct pci_dev *pdev) drm_sched_start(&ring->sched, true); } - amdgpu_device_unlock_adev(adev); + amdgpu_device_unset_mp1_state(adev); + amdgpu_device_unlock_reset_domain(adev->reset_domain); } bool amdgpu_device_cache_pci_state(struct pci_dev *pdev) @@ -5672,7 +5910,7 @@ void amdgpu_device_flush_hdp(struct amdgpu_device *adev, struct amdgpu_ring *ring) { #ifdef CONFIG_X86_64 - if (adev->flags & AMD_IS_APU) + if ((adev->flags & AMD_IS_APU) && !amdgpu_passthrough(adev)) return; #endif if (adev->gmc.xgmi.connected_to_cpu) @@ -5688,7 +5926,7 @@ void amdgpu_device_invalidate_hdp(struct amdgpu_device *adev, struct amdgpu_ring *ring) { #ifdef CONFIG_X86_64 - if (adev->flags & AMD_IS_APU) + if ((adev->flags & AMD_IS_APU) && !amdgpu_passthrough(adev)) return; #endif if (adev->gmc.xgmi.connected_to_cpu) @@ -5697,6 +5935,11 @@ void amdgpu_device_invalidate_hdp(struct amdgpu_device *adev, amdgpu_asic_invalidate_hdp(adev, ring); } +int amdgpu_in_reset(struct amdgpu_device *adev) +{ + return atomic_read(&adev->reset_domain->in_gpu_reset); + } + /** * amdgpu_device_halt() - bring hardware to some kind of halt state * @@ -5735,3 +5978,69 @@ void amdgpu_device_halt(struct amdgpu_device *adev) pci_disable_device(pdev); pci_wait_for_pending_transaction(pdev); } + +u32 amdgpu_device_pcie_port_rreg(struct amdgpu_device *adev, + u32 reg) +{ + unsigned long flags, address, data; + u32 r; + + address = adev->nbio.funcs->get_pcie_port_index_offset(adev); + data = adev->nbio.funcs->get_pcie_port_data_offset(adev); + + spin_lock_irqsave(&adev->pcie_idx_lock, flags); + WREG32(address, reg * 4); + (void)RREG32(address); + r = RREG32(data); + spin_unlock_irqrestore(&adev->pcie_idx_lock, flags); + return r; +} + +void amdgpu_device_pcie_port_wreg(struct amdgpu_device *adev, + u32 reg, u32 v) +{ + unsigned long flags, address, data; + + address = adev->nbio.funcs->get_pcie_port_index_offset(adev); + data = adev->nbio.funcs->get_pcie_port_data_offset(adev); + + spin_lock_irqsave(&adev->pcie_idx_lock, flags); + WREG32(address, reg * 4); + (void)RREG32(address); + WREG32(data, v); + (void)RREG32(data); + spin_unlock_irqrestore(&adev->pcie_idx_lock, flags); +} + +/** + * amdgpu_device_switch_gang - switch to a new gang + * @adev: amdgpu_device pointer + * @gang: the gang to switch to + * + * Try to switch to a new gang. + * Returns: NULL if we switched to the new gang or a reference to the current + * gang leader. + */ +struct dma_fence *amdgpu_device_switch_gang(struct amdgpu_device *adev, + struct dma_fence *gang) +{ + struct dma_fence *old = NULL; + + do { + dma_fence_put(old); + rcu_read_lock(); + old = dma_fence_get_rcu_safe(&adev->gang_submit); + rcu_read_unlock(); + + if (old == gang) + break; + + if (!dma_fence_is_signaled(old)) + return old; + + } while (cmpxchg((struct dma_fence __force **)&adev->gang_submit, + old, gang) != old); + + dma_fence_put(old); + return NULL; +} diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_df.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_df.h index 6b25837955c4..1538b2dbfff1 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_df.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_df.h @@ -40,7 +40,7 @@ struct amdgpu_df_funcs { void (*update_medium_grain_clock_gating)(struct amdgpu_device *adev, bool enable); void (*get_clockgating_state)(struct amdgpu_device *adev, - u32 *flags); + u64 *flags); void (*enable_ecc_force_par_wr_rmw)(struct amdgpu_device *adev, bool enable); int (*pmc_start)(struct amdgpu_device *adev, uint64_t config, diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c index 028190d42bb2..3993e6134914 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c @@ -47,25 +47,39 @@ #include "jpeg_v2_5.h" #include "smuio_v9_0.h" #include "gmc_v10_0.h" +#include "gmc_v11_0.h" #include "gfxhub_v2_0.h" #include "mmhub_v2_0.h" #include "nbio_v2_3.h" +#include "nbio_v4_3.h" #include "nbio_v7_2.h" +#include "nbio_v7_7.h" #include "hdp_v5_0.h" +#include "hdp_v5_2.h" +#include "hdp_v6_0.h" #include "nv.h" +#include "soc21.h" #include "navi10_ih.h" +#include "ih_v6_0.h" #include "gfx_v10_0.h" +#include "gfx_v11_0.h" #include "sdma_v5_0.h" #include "sdma_v5_2.h" +#include "sdma_v6_0.h" +#include "lsdma_v6_0.h" #include "vcn_v2_0.h" #include "jpeg_v2_0.h" #include "vcn_v3_0.h" #include "jpeg_v3_0.h" +#include "vcn_v4_0.h" +#include "jpeg_v4_0.h" #include "amdgpu_vkms.h" #include "mes_v10_1.h" +#include "mes_v11_0.h" #include "smuio_v11_0.h" #include "smuio_v11_0_6.h" #include "smuio_v13_0.h" +#include "smuio_v13_0_6.h" #define FIRMWARE_IP_DISCOVERY "amdgpu/ip_discovery.bin" MODULE_FIRMWARE(FIRMWARE_IP_DISCOVERY); @@ -111,6 +125,7 @@ static const char *hw_id_names[HW_ID_MAX] = { [SDMA1_HWID] = "SDMA1", [SDMA2_HWID] = "SDMA2", [SDMA3_HWID] = "SDMA3", + [LSDMA_HWID] = "LSDMA", [ISP_HWID] = "ISP", [DBGU_IO_HWID] = "DBGU_IO", [DF_HWID] = "DF", @@ -160,6 +175,7 @@ static int hw_id_map[MAX_HWIP] = { [SDMA1_HWIP] = SDMA1_HWID, [SDMA2_HWIP] = SDMA2_HWID, [SDMA3_HWIP] = SDMA3_HWID, + [LSDMA_HWIP] = LSDMA_HWID, [MMHUB_HWIP] = MMHUB_HWID, [ATHUB_HWIP] = ATHUB_HWID, [NBIO_HWIP] = NBIF_HWID, @@ -178,6 +194,7 @@ static int hw_id_map[MAX_HWIP] = { [UMC_HWIP] = UMC_HWID, [XGMI_HWIP] = XGMI_HWID, [DCI_HWIP] = DCI_HWID, + [PCIE_HWIP] = PCIE_HWID, }; static int amdgpu_discovery_read_binary_from_vram(struct amdgpu_device *adev, uint8_t *binary) @@ -212,7 +229,7 @@ static int amdgpu_discovery_read_binary_from_file(struct amdgpu_device *adev, ui return r; } - memcpy((u8 *)binary, (u8 *)fw->data, adev->mman.discovery_tmr_size); + memcpy((u8 *)binary, (u8 *)fw->data, fw->size); release_firmware(fw); return 0; @@ -243,12 +260,34 @@ static inline bool amdgpu_discovery_verify_binary_signature(uint8_t *binary) return (le32_to_cpu(bhdr->binary_signature) == BINARY_SIGNATURE); } +static void amdgpu_discovery_harvest_config_quirk(struct amdgpu_device *adev) +{ + /* + * So far, apply this quirk only on those Navy Flounder boards which + * have a bad harvest table of VCN config. + */ + if ((adev->ip_versions[UVD_HWIP][1] == IP_VERSION(3, 0, 1)) && + (adev->ip_versions[GC_HWIP][0] == IP_VERSION(10, 3, 2))) { + switch (adev->pdev->revision) { + case 0xC1: + case 0xC2: + case 0xC3: + case 0xC5: + case 0xC7: + case 0xCF: + case 0xDF: + adev->vcn.harvest_config |= AMDGPU_VCN_HARVEST_VCN1; + break; + default: + break; + } + } +} + static int amdgpu_discovery_init(struct amdgpu_device *adev) { struct table_info *info; struct binary_header *bhdr; - struct ip_discovery_header *ihdr; - struct gpu_info_header *ghdr; uint16_t offset; uint16_t size; uint16_t checksum; @@ -266,7 +305,7 @@ static int amdgpu_discovery_init(struct amdgpu_device *adev) goto out; } - if(!amdgpu_discovery_verify_binary_signature(adev->mman.discovery_bin)) { + if (!amdgpu_discovery_verify_binary_signature(adev->mman.discovery_bin)) { dev_warn(adev->dev, "get invalid ip discovery binary signature from vram\n"); /* retry read ip discovery binary from file */ r = amdgpu_discovery_read_binary_from_file(adev, adev->mman.discovery_bin); @@ -300,31 +339,110 @@ static int amdgpu_discovery_init(struct amdgpu_device *adev) info = &bhdr->table_list[IP_DISCOVERY]; offset = le16_to_cpu(info->offset); checksum = le16_to_cpu(info->checksum); - ihdr = (struct ip_discovery_header *)(adev->mman.discovery_bin + offset); - if (le32_to_cpu(ihdr->signature) != DISCOVERY_TABLE_SIGNATURE) { - dev_err(adev->dev, "invalid ip discovery data table signature\n"); - r = -EINVAL; - goto out; - } + if (offset) { + struct ip_discovery_header *ihdr = + (struct ip_discovery_header *)(adev->mman.discovery_bin + offset); + if (le32_to_cpu(ihdr->signature) != DISCOVERY_TABLE_SIGNATURE) { + dev_err(adev->dev, "invalid ip discovery data table signature\n"); + r = -EINVAL; + goto out; + } - if (!amdgpu_discovery_verify_checksum(adev->mman.discovery_bin + offset, - le16_to_cpu(ihdr->size), checksum)) { - dev_err(adev->dev, "invalid ip discovery data table checksum\n"); - r = -EINVAL; - goto out; + if (!amdgpu_discovery_verify_checksum(adev->mman.discovery_bin + offset, + le16_to_cpu(ihdr->size), checksum)) { + dev_err(adev->dev, "invalid ip discovery data table checksum\n"); + r = -EINVAL; + goto out; + } } info = &bhdr->table_list[GC]; offset = le16_to_cpu(info->offset); checksum = le16_to_cpu(info->checksum); - ghdr = (struct gpu_info_header *)(adev->mman.discovery_bin + offset); - if (!amdgpu_discovery_verify_checksum(adev->mman.discovery_bin + offset, - le32_to_cpu(ghdr->size), checksum)) { - dev_err(adev->dev, "invalid gc data table checksum\n"); - r = -EINVAL; - goto out; + if (offset) { + struct gpu_info_header *ghdr = + (struct gpu_info_header *)(adev->mman.discovery_bin + offset); + + if (le32_to_cpu(ghdr->table_id) != GC_TABLE_ID) { + dev_err(adev->dev, "invalid ip discovery gc table id\n"); + r = -EINVAL; + goto out; + } + + if (!amdgpu_discovery_verify_checksum(adev->mman.discovery_bin + offset, + le32_to_cpu(ghdr->size), checksum)) { + dev_err(adev->dev, "invalid gc data table checksum\n"); + r = -EINVAL; + goto out; + } + } + + info = &bhdr->table_list[HARVEST_INFO]; + offset = le16_to_cpu(info->offset); + checksum = le16_to_cpu(info->checksum); + + if (offset) { + struct harvest_info_header *hhdr = + (struct harvest_info_header *)(adev->mman.discovery_bin + offset); + + if (le32_to_cpu(hhdr->signature) != HARVEST_TABLE_SIGNATURE) { + dev_err(adev->dev, "invalid ip discovery harvest table signature\n"); + r = -EINVAL; + goto out; + } + + if (!amdgpu_discovery_verify_checksum(adev->mman.discovery_bin + offset, + sizeof(struct harvest_table), checksum)) { + dev_err(adev->dev, "invalid harvest data table checksum\n"); + r = -EINVAL; + goto out; + } + } + + info = &bhdr->table_list[VCN_INFO]; + offset = le16_to_cpu(info->offset); + checksum = le16_to_cpu(info->checksum); + + if (offset) { + struct vcn_info_header *vhdr = + (struct vcn_info_header *)(adev->mman.discovery_bin + offset); + + if (le32_to_cpu(vhdr->table_id) != VCN_INFO_TABLE_ID) { + dev_err(adev->dev, "invalid ip discovery vcn table id\n"); + r = -EINVAL; + goto out; + } + + if (!amdgpu_discovery_verify_checksum(adev->mman.discovery_bin + offset, + le32_to_cpu(vhdr->size_bytes), checksum)) { + dev_err(adev->dev, "invalid vcn data table checksum\n"); + r = -EINVAL; + goto out; + } + } + + info = &bhdr->table_list[MALL_INFO]; + offset = le16_to_cpu(info->offset); + checksum = le16_to_cpu(info->checksum); + + if (0 && offset) { + struct mall_info_header *mhdr = + (struct mall_info_header *)(adev->mman.discovery_bin + offset); + + if (le32_to_cpu(mhdr->table_id) != MALL_INFO_TABLE_ID) { + dev_err(adev->dev, "invalid ip discovery mall table id\n"); + r = -EINVAL; + goto out; + } + + if (!amdgpu_discovery_verify_checksum(adev->mman.discovery_bin + offset, + le32_to_cpu(mhdr->size_bytes), checksum)) { + dev_err(adev->dev, "invalid mall data table checksum\n"); + r = -EINVAL; + goto out; + } } return 0; @@ -336,8 +454,11 @@ out: return r; } +static void amdgpu_discovery_sysfs_fini(struct amdgpu_device *adev); + void amdgpu_discovery_fini(struct amdgpu_device *adev) { + amdgpu_discovery_sysfs_fini(adev); kfree(adev->mman.discovery_bin); adev->mman.discovery_bin = NULL; } @@ -358,7 +479,591 @@ static int amdgpu_discovery_validate_ip(const struct ip *ip) return 0; } -int amdgpu_discovery_reg_base_init(struct amdgpu_device *adev) +static void amdgpu_discovery_read_harvest_bit_per_ip(struct amdgpu_device *adev, + uint32_t *vcn_harvest_count) +{ + struct binary_header *bhdr; + struct ip_discovery_header *ihdr; + struct die_header *dhdr; + struct ip *ip; + uint16_t die_offset, ip_offset, num_dies, num_ips; + int i, j; + + bhdr = (struct binary_header *)adev->mman.discovery_bin; + ihdr = (struct ip_discovery_header *)(adev->mman.discovery_bin + + le16_to_cpu(bhdr->table_list[IP_DISCOVERY].offset)); + num_dies = le16_to_cpu(ihdr->num_dies); + + /* scan harvest bit of all IP data structures */ + for (i = 0; i < num_dies; i++) { + die_offset = le16_to_cpu(ihdr->die_info[i].die_offset); + dhdr = (struct die_header *)(adev->mman.discovery_bin + die_offset); + num_ips = le16_to_cpu(dhdr->num_ips); + ip_offset = die_offset + sizeof(*dhdr); + + for (j = 0; j < num_ips; j++) { + ip = (struct ip *)(adev->mman.discovery_bin + ip_offset); + + if (amdgpu_discovery_validate_ip(ip)) + goto next_ip; + + if (le16_to_cpu(ip->harvest) == 1) { + switch (le16_to_cpu(ip->hw_id)) { + case VCN_HWID: + (*vcn_harvest_count)++; + if (ip->number_instance == 0) + adev->vcn.harvest_config |= AMDGPU_VCN_HARVEST_VCN0; + else + adev->vcn.harvest_config |= AMDGPU_VCN_HARVEST_VCN1; + break; + case DMU_HWID: + adev->harvest_ip_mask |= AMD_HARVEST_IP_DMU_MASK; + break; + default: + break; + } + } +next_ip: + ip_offset += struct_size(ip, base_address, ip->num_base_address); + } + } +} + +static void amdgpu_discovery_read_from_harvest_table(struct amdgpu_device *adev, + uint32_t *vcn_harvest_count, + uint32_t *umc_harvest_count) +{ + struct binary_header *bhdr; + struct harvest_table *harvest_info; + u16 offset; + int i; + + bhdr = (struct binary_header *)adev->mman.discovery_bin; + offset = le16_to_cpu(bhdr->table_list[HARVEST_INFO].offset); + + if (!offset) { + dev_err(adev->dev, "invalid harvest table offset\n"); + return; + } + + harvest_info = (struct harvest_table *)(adev->mman.discovery_bin + offset); + + for (i = 0; i < 32; i++) { + if (le16_to_cpu(harvest_info->list[i].hw_id) == 0) + break; + + switch (le16_to_cpu(harvest_info->list[i].hw_id)) { + case VCN_HWID: + (*vcn_harvest_count)++; + if (harvest_info->list[i].number_instance == 0) + adev->vcn.harvest_config |= AMDGPU_VCN_HARVEST_VCN0; + else + adev->vcn.harvest_config |= AMDGPU_VCN_HARVEST_VCN1; + break; + case DMU_HWID: + adev->harvest_ip_mask |= AMD_HARVEST_IP_DMU_MASK; + break; + case UMC_HWID: + (*umc_harvest_count)++; + break; + default: + break; + } + } +} + +/* ================================================== */ + +struct ip_hw_instance { + struct kobject kobj; /* ip_discovery/die/#die/#hw_id/#instance/<attrs...> */ + + int hw_id; + u8 num_instance; + u8 major, minor, revision; + u8 harvest; + + int num_base_addresses; + u32 base_addr[]; +}; + +struct ip_hw_id { + struct kset hw_id_kset; /* ip_discovery/die/#die/#hw_id/, contains ip_hw_instance */ + int hw_id; +}; + +struct ip_die_entry { + struct kset ip_kset; /* ip_discovery/die/#die/, contains ip_hw_id */ + u16 num_ips; +}; + +/* -------------------------------------------------- */ + +struct ip_hw_instance_attr { + struct attribute attr; + ssize_t (*show)(struct ip_hw_instance *ip_hw_instance, char *buf); +}; + +static ssize_t hw_id_show(struct ip_hw_instance *ip_hw_instance, char *buf) +{ + return sysfs_emit(buf, "%d\n", ip_hw_instance->hw_id); +} + +static ssize_t num_instance_show(struct ip_hw_instance *ip_hw_instance, char *buf) +{ + return sysfs_emit(buf, "%d\n", ip_hw_instance->num_instance); +} + +static ssize_t major_show(struct ip_hw_instance *ip_hw_instance, char *buf) +{ + return sysfs_emit(buf, "%d\n", ip_hw_instance->major); +} + +static ssize_t minor_show(struct ip_hw_instance *ip_hw_instance, char *buf) +{ + return sysfs_emit(buf, "%d\n", ip_hw_instance->minor); +} + +static ssize_t revision_show(struct ip_hw_instance *ip_hw_instance, char *buf) +{ + return sysfs_emit(buf, "%d\n", ip_hw_instance->revision); +} + +static ssize_t harvest_show(struct ip_hw_instance *ip_hw_instance, char *buf) +{ + return sysfs_emit(buf, "0x%01X\n", ip_hw_instance->harvest); +} + +static ssize_t num_base_addresses_show(struct ip_hw_instance *ip_hw_instance, char *buf) +{ + return sysfs_emit(buf, "%d\n", ip_hw_instance->num_base_addresses); +} + +static ssize_t base_addr_show(struct ip_hw_instance *ip_hw_instance, char *buf) +{ + ssize_t res, at; + int ii; + + for (res = at = ii = 0; ii < ip_hw_instance->num_base_addresses; ii++) { + /* Here we satisfy the condition that, at + size <= PAGE_SIZE. + */ + if (at + 12 > PAGE_SIZE) + break; + res = sysfs_emit_at(buf, at, "0x%08X\n", + ip_hw_instance->base_addr[ii]); + if (res <= 0) + break; + at += res; + } + + return res < 0 ? res : at; +} + +static struct ip_hw_instance_attr ip_hw_attr[] = { + __ATTR_RO(hw_id), + __ATTR_RO(num_instance), + __ATTR_RO(major), + __ATTR_RO(minor), + __ATTR_RO(revision), + __ATTR_RO(harvest), + __ATTR_RO(num_base_addresses), + __ATTR_RO(base_addr), +}; + +static struct attribute *ip_hw_instance_attrs[ARRAY_SIZE(ip_hw_attr) + 1]; +ATTRIBUTE_GROUPS(ip_hw_instance); + +#define to_ip_hw_instance(x) container_of(x, struct ip_hw_instance, kobj) +#define to_ip_hw_instance_attr(x) container_of(x, struct ip_hw_instance_attr, attr) + +static ssize_t ip_hw_instance_attr_show(struct kobject *kobj, + struct attribute *attr, + char *buf) +{ + struct ip_hw_instance *ip_hw_instance = to_ip_hw_instance(kobj); + struct ip_hw_instance_attr *ip_hw_attr = to_ip_hw_instance_attr(attr); + + if (!ip_hw_attr->show) + return -EIO; + + return ip_hw_attr->show(ip_hw_instance, buf); +} + +static const struct sysfs_ops ip_hw_instance_sysfs_ops = { + .show = ip_hw_instance_attr_show, +}; + +static void ip_hw_instance_release(struct kobject *kobj) +{ + struct ip_hw_instance *ip_hw_instance = to_ip_hw_instance(kobj); + + kfree(ip_hw_instance); +} + +static struct kobj_type ip_hw_instance_ktype = { + .release = ip_hw_instance_release, + .sysfs_ops = &ip_hw_instance_sysfs_ops, + .default_groups = ip_hw_instance_groups, +}; + +/* -------------------------------------------------- */ + +#define to_ip_hw_id(x) container_of(to_kset(x), struct ip_hw_id, hw_id_kset) + +static void ip_hw_id_release(struct kobject *kobj) +{ + struct ip_hw_id *ip_hw_id = to_ip_hw_id(kobj); + + if (!list_empty(&ip_hw_id->hw_id_kset.list)) + DRM_ERROR("ip_hw_id->hw_id_kset is not empty"); + kfree(ip_hw_id); +} + +static struct kobj_type ip_hw_id_ktype = { + .release = ip_hw_id_release, + .sysfs_ops = &kobj_sysfs_ops, +}; + +/* -------------------------------------------------- */ + +static void die_kobj_release(struct kobject *kobj); +static void ip_disc_release(struct kobject *kobj); + +struct ip_die_entry_attribute { + struct attribute attr; + ssize_t (*show)(struct ip_die_entry *ip_die_entry, char *buf); +}; + +#define to_ip_die_entry_attr(x) container_of(x, struct ip_die_entry_attribute, attr) + +static ssize_t num_ips_show(struct ip_die_entry *ip_die_entry, char *buf) +{ + return sysfs_emit(buf, "%d\n", ip_die_entry->num_ips); +} + +/* If there are more ip_die_entry attrs, other than the number of IPs, + * we can make this intro an array of attrs, and then initialize + * ip_die_entry_attrs in a loop. + */ +static struct ip_die_entry_attribute num_ips_attr = + __ATTR_RO(num_ips); + +static struct attribute *ip_die_entry_attrs[] = { + &num_ips_attr.attr, + NULL, +}; +ATTRIBUTE_GROUPS(ip_die_entry); /* ip_die_entry_groups */ + +#define to_ip_die_entry(x) container_of(to_kset(x), struct ip_die_entry, ip_kset) + +static ssize_t ip_die_entry_attr_show(struct kobject *kobj, + struct attribute *attr, + char *buf) +{ + struct ip_die_entry_attribute *ip_die_entry_attr = to_ip_die_entry_attr(attr); + struct ip_die_entry *ip_die_entry = to_ip_die_entry(kobj); + + if (!ip_die_entry_attr->show) + return -EIO; + + return ip_die_entry_attr->show(ip_die_entry, buf); +} + +static void ip_die_entry_release(struct kobject *kobj) +{ + struct ip_die_entry *ip_die_entry = to_ip_die_entry(kobj); + + if (!list_empty(&ip_die_entry->ip_kset.list)) + DRM_ERROR("ip_die_entry->ip_kset is not empty"); + kfree(ip_die_entry); +} + +static const struct sysfs_ops ip_die_entry_sysfs_ops = { + .show = ip_die_entry_attr_show, +}; + +static struct kobj_type ip_die_entry_ktype = { + .release = ip_die_entry_release, + .sysfs_ops = &ip_die_entry_sysfs_ops, + .default_groups = ip_die_entry_groups, +}; + +static struct kobj_type die_kobj_ktype = { + .release = die_kobj_release, + .sysfs_ops = &kobj_sysfs_ops, +}; + +static struct kobj_type ip_discovery_ktype = { + .release = ip_disc_release, + .sysfs_ops = &kobj_sysfs_ops, +}; + +struct ip_discovery_top { + struct kobject kobj; /* ip_discovery/ */ + struct kset die_kset; /* ip_discovery/die/, contains ip_die_entry */ + struct amdgpu_device *adev; +}; + +static void die_kobj_release(struct kobject *kobj) +{ + struct ip_discovery_top *ip_top = container_of(to_kset(kobj), + struct ip_discovery_top, + die_kset); + if (!list_empty(&ip_top->die_kset.list)) + DRM_ERROR("ip_top->die_kset is not empty"); +} + +static void ip_disc_release(struct kobject *kobj) +{ + struct ip_discovery_top *ip_top = container_of(kobj, struct ip_discovery_top, + kobj); + struct amdgpu_device *adev = ip_top->adev; + + adev->ip_top = NULL; + kfree(ip_top); +} + +static int amdgpu_discovery_sysfs_ips(struct amdgpu_device *adev, + struct ip_die_entry *ip_die_entry, + const size_t _ip_offset, const int num_ips) +{ + int ii, jj, kk, res; + + DRM_DEBUG("num_ips:%d", num_ips); + + /* Find all IPs of a given HW ID, and add their instance to + * #die/#hw_id/#instance/<attributes> + */ + for (ii = 0; ii < HW_ID_MAX; ii++) { + struct ip_hw_id *ip_hw_id = NULL; + size_t ip_offset = _ip_offset; + + for (jj = 0; jj < num_ips; jj++) { + struct ip *ip; + struct ip_hw_instance *ip_hw_instance; + + ip = (struct ip *)(adev->mman.discovery_bin + ip_offset); + if (amdgpu_discovery_validate_ip(ip) || + le16_to_cpu(ip->hw_id) != ii) + goto next_ip; + + DRM_DEBUG("match:%d @ ip_offset:%zu", ii, ip_offset); + + /* We have a hw_id match; register the hw + * block if not yet registered. + */ + if (!ip_hw_id) { + ip_hw_id = kzalloc(sizeof(*ip_hw_id), GFP_KERNEL); + if (!ip_hw_id) + return -ENOMEM; + ip_hw_id->hw_id = ii; + + kobject_set_name(&ip_hw_id->hw_id_kset.kobj, "%d", ii); + ip_hw_id->hw_id_kset.kobj.kset = &ip_die_entry->ip_kset; + ip_hw_id->hw_id_kset.kobj.ktype = &ip_hw_id_ktype; + res = kset_register(&ip_hw_id->hw_id_kset); + if (res) { + DRM_ERROR("Couldn't register ip_hw_id kset"); + kfree(ip_hw_id); + return res; + } + if (hw_id_names[ii]) { + res = sysfs_create_link(&ip_die_entry->ip_kset.kobj, + &ip_hw_id->hw_id_kset.kobj, + hw_id_names[ii]); + if (res) { + DRM_ERROR("Couldn't create IP link %s in IP Die:%s\n", + hw_id_names[ii], + kobject_name(&ip_die_entry->ip_kset.kobj)); + } + } + } + + /* Now register its instance. + */ + ip_hw_instance = kzalloc(struct_size(ip_hw_instance, + base_addr, + ip->num_base_address), + GFP_KERNEL); + if (!ip_hw_instance) { + DRM_ERROR("no memory for ip_hw_instance"); + return -ENOMEM; + } + ip_hw_instance->hw_id = le16_to_cpu(ip->hw_id); /* == ii */ + ip_hw_instance->num_instance = ip->number_instance; + ip_hw_instance->major = ip->major; + ip_hw_instance->minor = ip->minor; + ip_hw_instance->revision = ip->revision; + ip_hw_instance->harvest = ip->harvest; + ip_hw_instance->num_base_addresses = ip->num_base_address; + + for (kk = 0; kk < ip_hw_instance->num_base_addresses; kk++) + ip_hw_instance->base_addr[kk] = ip->base_address[kk]; + + kobject_init(&ip_hw_instance->kobj, &ip_hw_instance_ktype); + ip_hw_instance->kobj.kset = &ip_hw_id->hw_id_kset; + res = kobject_add(&ip_hw_instance->kobj, NULL, + "%d", ip_hw_instance->num_instance); +next_ip: + ip_offset += struct_size(ip, base_address, ip->num_base_address); + } + } + + return 0; +} + +static int amdgpu_discovery_sysfs_recurse(struct amdgpu_device *adev) +{ + struct binary_header *bhdr; + struct ip_discovery_header *ihdr; + struct die_header *dhdr; + struct kset *die_kset = &adev->ip_top->die_kset; + u16 num_dies, die_offset, num_ips; + size_t ip_offset; + int ii, res; + + bhdr = (struct binary_header *)adev->mman.discovery_bin; + ihdr = (struct ip_discovery_header *)(adev->mman.discovery_bin + + le16_to_cpu(bhdr->table_list[IP_DISCOVERY].offset)); + num_dies = le16_to_cpu(ihdr->num_dies); + + DRM_DEBUG("number of dies: %d\n", num_dies); + + for (ii = 0; ii < num_dies; ii++) { + struct ip_die_entry *ip_die_entry; + + die_offset = le16_to_cpu(ihdr->die_info[ii].die_offset); + dhdr = (struct die_header *)(adev->mman.discovery_bin + die_offset); + num_ips = le16_to_cpu(dhdr->num_ips); + ip_offset = die_offset + sizeof(*dhdr); + + /* Add the die to the kset. + * + * dhdr->die_id == ii, which was checked in + * amdgpu_discovery_reg_base_init(). + */ + + ip_die_entry = kzalloc(sizeof(*ip_die_entry), GFP_KERNEL); + if (!ip_die_entry) + return -ENOMEM; + + ip_die_entry->num_ips = num_ips; + + kobject_set_name(&ip_die_entry->ip_kset.kobj, "%d", le16_to_cpu(dhdr->die_id)); + ip_die_entry->ip_kset.kobj.kset = die_kset; + ip_die_entry->ip_kset.kobj.ktype = &ip_die_entry_ktype; + res = kset_register(&ip_die_entry->ip_kset); + if (res) { + DRM_ERROR("Couldn't register ip_die_entry kset"); + kfree(ip_die_entry); + return res; + } + + amdgpu_discovery_sysfs_ips(adev, ip_die_entry, ip_offset, num_ips); + } + + return 0; +} + +static int amdgpu_discovery_sysfs_init(struct amdgpu_device *adev) +{ + struct kset *die_kset; + int res, ii; + + adev->ip_top = kzalloc(sizeof(*adev->ip_top), GFP_KERNEL); + if (!adev->ip_top) + return -ENOMEM; + + adev->ip_top->adev = adev; + + res = kobject_init_and_add(&adev->ip_top->kobj, &ip_discovery_ktype, + &adev->dev->kobj, "ip_discovery"); + if (res) { + DRM_ERROR("Couldn't init and add ip_discovery/"); + goto Err; + } + + die_kset = &adev->ip_top->die_kset; + kobject_set_name(&die_kset->kobj, "%s", "die"); + die_kset->kobj.parent = &adev->ip_top->kobj; + die_kset->kobj.ktype = &die_kobj_ktype; + res = kset_register(&adev->ip_top->die_kset); + if (res) { + DRM_ERROR("Couldn't register die_kset"); + goto Err; + } + + for (ii = 0; ii < ARRAY_SIZE(ip_hw_attr); ii++) + ip_hw_instance_attrs[ii] = &ip_hw_attr[ii].attr; + ip_hw_instance_attrs[ii] = NULL; + + res = amdgpu_discovery_sysfs_recurse(adev); + + return res; +Err: + kobject_put(&adev->ip_top->kobj); + return res; +} + +/* -------------------------------------------------- */ + +#define list_to_kobj(el) container_of(el, struct kobject, entry) + +static void amdgpu_discovery_sysfs_ip_hw_free(struct ip_hw_id *ip_hw_id) +{ + struct list_head *el, *tmp; + struct kset *hw_id_kset; + + hw_id_kset = &ip_hw_id->hw_id_kset; + spin_lock(&hw_id_kset->list_lock); + list_for_each_prev_safe(el, tmp, &hw_id_kset->list) { + list_del_init(el); + spin_unlock(&hw_id_kset->list_lock); + /* kobject is embedded in ip_hw_instance */ + kobject_put(list_to_kobj(el)); + spin_lock(&hw_id_kset->list_lock); + } + spin_unlock(&hw_id_kset->list_lock); + kobject_put(&ip_hw_id->hw_id_kset.kobj); +} + +static void amdgpu_discovery_sysfs_die_free(struct ip_die_entry *ip_die_entry) +{ + struct list_head *el, *tmp; + struct kset *ip_kset; + + ip_kset = &ip_die_entry->ip_kset; + spin_lock(&ip_kset->list_lock); + list_for_each_prev_safe(el, tmp, &ip_kset->list) { + list_del_init(el); + spin_unlock(&ip_kset->list_lock); + amdgpu_discovery_sysfs_ip_hw_free(to_ip_hw_id(list_to_kobj(el))); + spin_lock(&ip_kset->list_lock); + } + spin_unlock(&ip_kset->list_lock); + kobject_put(&ip_die_entry->ip_kset.kobj); +} + +static void amdgpu_discovery_sysfs_fini(struct amdgpu_device *adev) +{ + struct list_head *el, *tmp; + struct kset *die_kset; + + die_kset = &adev->ip_top->die_kset; + spin_lock(&die_kset->list_lock); + list_for_each_prev_safe(el, tmp, &die_kset->list) { + list_del_init(el); + spin_unlock(&die_kset->list_lock); + amdgpu_discovery_sysfs_die_free(to_ip_die_entry(list_to_kobj(el))); + spin_lock(&die_kset->list_lock); + } + spin_unlock(&die_kset->list_lock); + kobject_put(&adev->ip_top->die_kset.kobj); + kobject_put(&adev->ip_top->kobj); +} + +/* ================================================== */ + +static int amdgpu_discovery_reg_base_init(struct amdgpu_device *adev) { struct binary_header *bhdr; struct ip_discovery_header *ihdr; @@ -426,13 +1131,27 @@ int amdgpu_discovery_reg_base_init(struct amdgpu_device *adev) adev->vcn.vcn_config[adev->vcn.num_vcn_inst] = ip->revision & 0xc0; ip->revision &= ~0xc0; - adev->vcn.num_vcn_inst++; + if (adev->vcn.num_vcn_inst < AMDGPU_MAX_VCN_INSTANCES) + adev->vcn.num_vcn_inst++; + else + dev_err(adev->dev, "Too many VCN instances: %d vs %d\n", + adev->vcn.num_vcn_inst + 1, + AMDGPU_MAX_VCN_INSTANCES); } if (le16_to_cpu(ip->hw_id) == SDMA0_HWID || le16_to_cpu(ip->hw_id) == SDMA1_HWID || le16_to_cpu(ip->hw_id) == SDMA2_HWID || - le16_to_cpu(ip->hw_id) == SDMA3_HWID) - adev->sdma.num_instances++; + le16_to_cpu(ip->hw_id) == SDMA3_HWID) { + if (adev->sdma.num_instances < AMDGPU_MAX_SDMA_INSTANCES) + adev->sdma.num_instances++; + else + dev_err(adev->dev, "Too many SDMA instances: %d vs %d\n", + adev->sdma.num_instances + 1, + AMDGPU_MAX_SDMA_INSTANCES); + } + + if (le16_to_cpu(ip->hw_id) == UMC_HWID) + adev->gmc.num_umc++; for (k = 0; k < num_base_address; k++) { /* @@ -464,10 +1183,12 @@ int amdgpu_discovery_reg_base_init(struct amdgpu_device *adev) } next_ip: - ip_offset += sizeof(*ip) + 4 * (ip->num_base_address - 1); + ip_offset += struct_size(ip, base_address, ip->num_base_address); } } + amdgpu_discovery_sysfs_init(adev); + return 0; } @@ -512,68 +1233,63 @@ int amdgpu_discovery_get_ip_version(struct amdgpu_device *adev, int hw_id, int n *revision = ip->revision; return 0; } - ip_offset += sizeof(*ip) + 4 * (ip->num_base_address - 1); + ip_offset += struct_size(ip, base_address, ip->num_base_address); } } return -EINVAL; } -void amdgpu_discovery_harvest_ip(struct amdgpu_device *adev) +static void amdgpu_discovery_harvest_ip(struct amdgpu_device *adev) { - struct binary_header *bhdr; - struct harvest_table *harvest_info; - int i, vcn_harvest_count = 0; + int vcn_harvest_count = 0; + int umc_harvest_count = 0; + + /* + * Harvest table does not fit Navi1x and legacy GPUs, + * so read harvest bit per IP data structure to set + * harvest configuration. + */ + if (adev->ip_versions[GC_HWIP][0] < IP_VERSION(10, 2, 0)) { + if ((adev->pdev->device == 0x731E && + (adev->pdev->revision == 0xC6 || + adev->pdev->revision == 0xC7)) || + (adev->pdev->device == 0x7340 && + adev->pdev->revision == 0xC9) || + (adev->pdev->device == 0x7360 && + adev->pdev->revision == 0xC7)) + amdgpu_discovery_read_harvest_bit_per_ip(adev, + &vcn_harvest_count); + } else { + amdgpu_discovery_read_from_harvest_table(adev, + &vcn_harvest_count, + &umc_harvest_count); + } - bhdr = (struct binary_header *)adev->mman.discovery_bin; - harvest_info = (struct harvest_table *)(adev->mman.discovery_bin + - le16_to_cpu(bhdr->table_list[HARVEST_INFO].offset)); + amdgpu_discovery_harvest_config_quirk(adev); - for (i = 0; i < 32; i++) { - if (le16_to_cpu(harvest_info->list[i].hw_id) == 0) - break; - - switch (le16_to_cpu(harvest_info->list[i].hw_id)) { - case VCN_HWID: - vcn_harvest_count++; - if (harvest_info->list[i].number_instance == 0) - adev->vcn.harvest_config |= AMDGPU_VCN_HARVEST_VCN0; - else - adev->vcn.harvest_config |= AMDGPU_VCN_HARVEST_VCN1; - break; - case DMU_HWID: - adev->harvest_ip_mask |= AMD_HARVEST_IP_DMU_MASK; - break; - default: - break; - } - } - /* some IP discovery tables on Navy Flounder don't have this set correctly */ - if ((adev->ip_versions[UVD_HWIP][1] == IP_VERSION(3, 0, 1)) && - (adev->ip_versions[GC_HWIP][0] == IP_VERSION(10, 3, 2))) - adev->vcn.harvest_config |= AMDGPU_VCN_HARVEST_VCN1; if (vcn_harvest_count == adev->vcn.num_vcn_inst) { adev->harvest_ip_mask |= AMD_HARVEST_IP_VCN_MASK; adev->harvest_ip_mask |= AMD_HARVEST_IP_JPEG_MASK; } - if ((adev->pdev->device == 0x731E && - (adev->pdev->revision == 0xC6 || adev->pdev->revision == 0xC7)) || - (adev->pdev->device == 0x7340 && adev->pdev->revision == 0xC9) || - (adev->pdev->device == 0x7360 && adev->pdev->revision == 0xC7)) { - adev->harvest_ip_mask |= AMD_HARVEST_IP_VCN_MASK; - adev->harvest_ip_mask |= AMD_HARVEST_IP_JPEG_MASK; + + if (umc_harvest_count < adev->gmc.num_umc) { + adev->gmc.num_umc -= umc_harvest_count; } } union gc_info { struct gc_info_v1_0 v1; + struct gc_info_v1_1 v1_1; + struct gc_info_v1_2 v1_2; struct gc_info_v2_0 v2; }; -int amdgpu_discovery_get_gfx_info(struct amdgpu_device *adev) +static int amdgpu_discovery_get_gfx_info(struct amdgpu_device *adev) { struct binary_header *bhdr; union gc_info *gc_info; + u16 offset; if (!adev->mman.discovery_bin) { DRM_ERROR("ip discovery uninitialized\n"); @@ -581,9 +1297,14 @@ int amdgpu_discovery_get_gfx_info(struct amdgpu_device *adev) } bhdr = (struct binary_header *)adev->mman.discovery_bin; - gc_info = (union gc_info *)(adev->mman.discovery_bin + - le16_to_cpu(bhdr->table_list[GC].offset)); - switch (gc_info->v1.header.version_major) { + offset = le16_to_cpu(bhdr->table_list[GC].offset); + + if (!offset) + return 0; + + gc_info = (union gc_info *)(adev->mman.discovery_bin + offset); + + switch (le16_to_cpu(gc_info->v1.header.version_major)) { case 1: adev->gfx.config.max_shader_engines = le32_to_cpu(gc_info->v1.gc_num_se); adev->gfx.config.max_cu_per_sh = 2 * (le32_to_cpu(gc_info->v1.gc_num_wgp0_per_sa) + @@ -603,6 +1324,21 @@ int amdgpu_discovery_get_gfx_info(struct amdgpu_device *adev) adev->gfx.config.num_sc_per_sh = le32_to_cpu(gc_info->v1.gc_num_sc_per_se) / le32_to_cpu(gc_info->v1.gc_num_sa_per_se); adev->gfx.config.num_packer_per_sc = le32_to_cpu(gc_info->v1.gc_num_packer_per_sc); + if (gc_info->v1.header.version_minor >= 1) { + adev->gfx.config.gc_num_tcp_per_sa = le32_to_cpu(gc_info->v1_1.gc_num_tcp_per_sa); + adev->gfx.config.gc_num_sdp_interface = le32_to_cpu(gc_info->v1_1.gc_num_sdp_interface); + adev->gfx.config.gc_num_tcps = le32_to_cpu(gc_info->v1_1.gc_num_tcps); + } + if (gc_info->v1.header.version_minor >= 2) { + adev->gfx.config.gc_num_tcp_per_wpg = le32_to_cpu(gc_info->v1_2.gc_num_tcp_per_wpg); + adev->gfx.config.gc_tcp_l1_size = le32_to_cpu(gc_info->v1_2.gc_tcp_l1_size); + adev->gfx.config.gc_num_sqc_per_wgp = le32_to_cpu(gc_info->v1_2.gc_num_sqc_per_wgp); + adev->gfx.config.gc_l1_instruction_cache_size_per_sqc = le32_to_cpu(gc_info->v1_2.gc_l1_instruction_cache_size_per_sqc); + adev->gfx.config.gc_l1_data_cache_size_per_sqc = le32_to_cpu(gc_info->v1_2.gc_l1_data_cache_size_per_sqc); + adev->gfx.config.gc_gl1c_per_sa = le32_to_cpu(gc_info->v1_2.gc_gl1c_per_sa); + adev->gfx.config.gc_gl1c_size_per_instance = le32_to_cpu(gc_info->v1_2.gc_gl1c_size_per_instance); + adev->gfx.config.gc_gl2c_per_gpu = le32_to_cpu(gc_info->v1_2.gc_gl2c_per_gpu); + } break; case 2: adev->gfx.config.max_shader_engines = le32_to_cpu(gc_info->v2.gc_num_se); @@ -626,8 +1362,113 @@ int amdgpu_discovery_get_gfx_info(struct amdgpu_device *adev) default: dev_err(adev->dev, "Unhandled GC info table %d.%d\n", - gc_info->v1.header.version_major, - gc_info->v1.header.version_minor); + le16_to_cpu(gc_info->v1.header.version_major), + le16_to_cpu(gc_info->v1.header.version_minor)); + return -EINVAL; + } + return 0; +} + +union mall_info { + struct mall_info_v1_0 v1; +}; + +static int amdgpu_discovery_get_mall_info(struct amdgpu_device *adev) +{ + struct binary_header *bhdr; + union mall_info *mall_info; + u32 u, mall_size_per_umc, m_s_present, half_use; + u64 mall_size; + u16 offset; + + if (!adev->mman.discovery_bin) { + DRM_ERROR("ip discovery uninitialized\n"); + return -EINVAL; + } + + bhdr = (struct binary_header *)adev->mman.discovery_bin; + offset = le16_to_cpu(bhdr->table_list[MALL_INFO].offset); + + if (!offset) + return 0; + + mall_info = (union mall_info *)(adev->mman.discovery_bin + offset); + + switch (le16_to_cpu(mall_info->v1.header.version_major)) { + case 1: + mall_size = 0; + mall_size_per_umc = le32_to_cpu(mall_info->v1.mall_size_per_m); + m_s_present = le32_to_cpu(mall_info->v1.m_s_present); + half_use = le32_to_cpu(mall_info->v1.m_half_use); + for (u = 0; u < adev->gmc.num_umc; u++) { + if (m_s_present & (1 << u)) + mall_size += mall_size_per_umc * 2; + else if (half_use & (1 << u)) + mall_size += mall_size_per_umc / 2; + else + mall_size += mall_size_per_umc; + } + adev->gmc.mall_size = mall_size; + break; + default: + dev_err(adev->dev, + "Unhandled MALL info table %d.%d\n", + le16_to_cpu(mall_info->v1.header.version_major), + le16_to_cpu(mall_info->v1.header.version_minor)); + return -EINVAL; + } + return 0; +} + +union vcn_info { + struct vcn_info_v1_0 v1; +}; + +static int amdgpu_discovery_get_vcn_info(struct amdgpu_device *adev) +{ + struct binary_header *bhdr; + union vcn_info *vcn_info; + u16 offset; + int v; + + if (!adev->mman.discovery_bin) { + DRM_ERROR("ip discovery uninitialized\n"); + return -EINVAL; + } + + /* num_vcn_inst is currently limited to AMDGPU_MAX_VCN_INSTANCES + * which is smaller than VCN_INFO_TABLE_MAX_NUM_INSTANCES + * but that may change in the future with new GPUs so keep this + * check for defensive purposes. + */ + if (adev->vcn.num_vcn_inst > VCN_INFO_TABLE_MAX_NUM_INSTANCES) { + dev_err(adev->dev, "invalid vcn instances\n"); + return -EINVAL; + } + + bhdr = (struct binary_header *)adev->mman.discovery_bin; + offset = le16_to_cpu(bhdr->table_list[VCN_INFO].offset); + + if (!offset) + return 0; + + vcn_info = (union vcn_info *)(adev->mman.discovery_bin + offset); + + switch (le16_to_cpu(vcn_info->v1.header.version_major)) { + case 1: + /* num_vcn_inst is currently limited to AMDGPU_MAX_VCN_INSTANCES + * so this won't overflow. + */ + for (v = 0; v < adev->vcn.num_vcn_inst; v++) { + adev->vcn.vcn_codec_disable_mask[v] = + le32_to_cpu(vcn_info->v1.instance_info[v].fuse_data.all_bits); + } + break; + default: + dev_err(adev->dev, + "Unhandled VCN info table %d.%d\n", + le16_to_cpu(vcn_info->v1.header.version_major), + le16_to_cpu(vcn_info->v1.header.version_minor)); return -EINVAL; } return 0; @@ -651,14 +1492,23 @@ static int amdgpu_discovery_set_common_ip_blocks(struct amdgpu_device *adev) case IP_VERSION(10, 1, 1): case IP_VERSION(10, 1, 2): case IP_VERSION(10, 1, 3): + case IP_VERSION(10, 1, 4): case IP_VERSION(10, 3, 0): case IP_VERSION(10, 3, 1): case IP_VERSION(10, 3, 2): case IP_VERSION(10, 3, 3): case IP_VERSION(10, 3, 4): case IP_VERSION(10, 3, 5): + case IP_VERSION(10, 3, 6): + case IP_VERSION(10, 3, 7): amdgpu_device_ip_block_add(adev, &nv_common_ip_block); break; + case IP_VERSION(11, 0, 0): + case IP_VERSION(11, 0, 1): + case IP_VERSION(11, 0, 2): + case IP_VERSION(11, 0, 3): + amdgpu_device_ip_block_add(adev, &soc21_common_ip_block); + break; default: dev_err(adev->dev, "Failed to add common ip block(GC_HWIP:0x%x)\n", @@ -686,14 +1536,23 @@ static int amdgpu_discovery_set_gmc_ip_blocks(struct amdgpu_device *adev) case IP_VERSION(10, 1, 1): case IP_VERSION(10, 1, 2): case IP_VERSION(10, 1, 3): + case IP_VERSION(10, 1, 4): case IP_VERSION(10, 3, 0): case IP_VERSION(10, 3, 1): case IP_VERSION(10, 3, 2): case IP_VERSION(10, 3, 3): case IP_VERSION(10, 3, 4): case IP_VERSION(10, 3, 5): + case IP_VERSION(10, 3, 6): + case IP_VERSION(10, 3, 7): amdgpu_device_ip_block_add(adev, &gmc_v10_0_ip_block); break; + case IP_VERSION(11, 0, 0): + case IP_VERSION(11, 0, 1): + case IP_VERSION(11, 0, 2): + case IP_VERSION(11, 0, 3): + amdgpu_device_ip_block_add(adev, &gmc_v11_0_ip_block); + break; default: dev_err(adev->dev, "Failed to add gmc ip block(GC_HWIP:0x%x)\n", @@ -726,6 +1585,11 @@ static int amdgpu_discovery_set_ih_ip_blocks(struct amdgpu_device *adev) case IP_VERSION(5, 2, 1): amdgpu_device_ip_block_add(adev, &navi10_ih_ip_block); break; + case IP_VERSION(6, 0, 0): + case IP_VERSION(6, 0, 1): + case IP_VERSION(6, 0, 2): + amdgpu_device_ip_block_add(adev, &ih_v6_0_ip_block); + break; default: dev_err(adev->dev, "Failed to add ih ip block(OSSSYS_HWIP:0x%x)\n", @@ -764,11 +1628,19 @@ static int amdgpu_discovery_set_psp_ip_blocks(struct amdgpu_device *adev) case IP_VERSION(12, 0, 1): amdgpu_device_ip_block_add(adev, &psp_v12_0_ip_block); break; + case IP_VERSION(13, 0, 0): case IP_VERSION(13, 0, 1): case IP_VERSION(13, 0, 2): case IP_VERSION(13, 0, 3): + case IP_VERSION(13, 0, 5): + case IP_VERSION(13, 0, 7): + case IP_VERSION(13, 0, 8): + case IP_VERSION(13, 0, 10): amdgpu_device_ip_block_add(adev, &psp_v13_0_ip_block); break; + case IP_VERSION(13, 0, 4): + amdgpu_device_ip_block_add(adev, &psp_v13_0_4_ip_block); + break; default: dev_err(adev->dev, "Failed to add psp ip block(MP0_HWIP:0x%x)\n", @@ -805,9 +1677,15 @@ static int amdgpu_discovery_set_smu_ip_blocks(struct amdgpu_device *adev) case IP_VERSION(12, 0, 1): amdgpu_device_ip_block_add(adev, &smu_v12_0_ip_block); break; + case IP_VERSION(13, 0, 0): case IP_VERSION(13, 0, 1): case IP_VERSION(13, 0, 2): case IP_VERSION(13, 0, 3): + case IP_VERSION(13, 0, 4): + case IP_VERSION(13, 0, 5): + case IP_VERSION(13, 0, 7): + case IP_VERSION(13, 0, 8): + case IP_VERSION(13, 0, 10): amdgpu_device_ip_block_add(adev, &smu_v13_0_ip_block); break; default: @@ -823,8 +1701,14 @@ static int amdgpu_discovery_set_display_ip_blocks(struct amdgpu_device *adev) { if (adev->enable_virtual_display || amdgpu_sriov_vf(adev)) { amdgpu_device_ip_block_add(adev, &amdgpu_vkms_ip_block); + return 0; + } + + if (!amdgpu_device_has_dc_support(adev)) + return 0; + #if defined(CONFIG_DRM_AMD_DC) - } else if (adev->ip_versions[DCE_HWIP][0]) { + if (adev->ip_versions[DCE_HWIP][0]) { switch (adev->ip_versions[DCE_HWIP][0]) { case IP_VERSION(1, 0, 0): case IP_VERSION(1, 0, 1): @@ -838,6 +1722,11 @@ static int amdgpu_discovery_set_display_ip_blocks(struct amdgpu_device *adev) case IP_VERSION(3, 0, 1): case IP_VERSION(3, 1, 2): case IP_VERSION(3, 1, 3): + case IP_VERSION(3, 1, 4): + case IP_VERSION(3, 1, 5): + case IP_VERSION(3, 1, 6): + case IP_VERSION(3, 2, 0): + case IP_VERSION(3, 2, 1): amdgpu_device_ip_block_add(adev, &dm_ip_block); break; default: @@ -859,8 +1748,8 @@ static int amdgpu_discovery_set_display_ip_blocks(struct amdgpu_device *adev) adev->ip_versions[DCI_HWIP][0]); return -EINVAL; } -#endif } +#endif return 0; } @@ -881,14 +1770,23 @@ static int amdgpu_discovery_set_gc_ip_blocks(struct amdgpu_device *adev) case IP_VERSION(10, 1, 2): case IP_VERSION(10, 1, 1): case IP_VERSION(10, 1, 3): + case IP_VERSION(10, 1, 4): case IP_VERSION(10, 3, 0): case IP_VERSION(10, 3, 2): case IP_VERSION(10, 3, 1): case IP_VERSION(10, 3, 4): case IP_VERSION(10, 3, 5): + case IP_VERSION(10, 3, 6): case IP_VERSION(10, 3, 3): + case IP_VERSION(10, 3, 7): amdgpu_device_ip_block_add(adev, &gfx_v10_0_ip_block); break; + case IP_VERSION(11, 0, 0): + case IP_VERSION(11, 0, 1): + case IP_VERSION(11, 0, 2): + case IP_VERSION(11, 0, 3): + amdgpu_device_ip_block_add(adev, &gfx_v11_0_ip_block); + break; default: dev_err(adev->dev, "Failed to add gfx ip block(GC_HWIP:0x%x)\n", @@ -921,10 +1819,18 @@ static int amdgpu_discovery_set_sdma_ip_blocks(struct amdgpu_device *adev) case IP_VERSION(5, 2, 2): case IP_VERSION(5, 2, 4): case IP_VERSION(5, 2, 5): + case IP_VERSION(5, 2, 6): case IP_VERSION(5, 2, 3): case IP_VERSION(5, 2, 1): + case IP_VERSION(5, 2, 7): amdgpu_device_ip_block_add(adev, &sdma_v5_2_ip_block); break; + case IP_VERSION(6, 0, 0): + case IP_VERSION(6, 0, 1): + case IP_VERSION(6, 0, 2): + case IP_VERSION(6, 0, 3): + amdgpu_device_ip_block_add(adev, &sdma_v6_0_ip_block); + break; default: dev_err(adev->dev, "Failed to add sdma ip block(SDMA0_HWIP:0x%x)\n", @@ -989,6 +1895,7 @@ static int amdgpu_discovery_set_mm_ip_blocks(struct amdgpu_device *adev) case IP_VERSION(3, 0, 0): case IP_VERSION(3, 0, 16): case IP_VERSION(3, 1, 1): + case IP_VERSION(3, 1, 2): case IP_VERSION(3, 0, 2): case IP_VERSION(3, 0, 192): amdgpu_device_ip_block_add(adev, &vcn_v3_0_ip_block); @@ -998,6 +1905,13 @@ static int amdgpu_discovery_set_mm_ip_blocks(struct amdgpu_device *adev) case IP_VERSION(3, 0, 33): amdgpu_device_ip_block_add(adev, &vcn_v3_0_ip_block); break; + case IP_VERSION(4, 0, 0): + case IP_VERSION(4, 0, 2): + case IP_VERSION(4, 0, 4): + amdgpu_device_ip_block_add(adev, &vcn_v4_0_ip_block); + if (!amdgpu_sriov_vf(adev)) + amdgpu_device_ip_block_add(adev, &jpeg_v4_0_ip_block); + break; default: dev_err(adev->dev, "Failed to add vcn/jpeg ip block(UVD_HWIP:0x%x)\n", @@ -1015,13 +1929,28 @@ static int amdgpu_discovery_set_mes_ip_blocks(struct amdgpu_device *adev) case IP_VERSION(10, 1, 1): case IP_VERSION(10, 1, 2): case IP_VERSION(10, 1, 3): + case IP_VERSION(10, 1, 4): case IP_VERSION(10, 3, 0): case IP_VERSION(10, 3, 1): case IP_VERSION(10, 3, 2): case IP_VERSION(10, 3, 3): case IP_VERSION(10, 3, 4): case IP_VERSION(10, 3, 5): - amdgpu_device_ip_block_add(adev, &mes_v10_1_ip_block); + case IP_VERSION(10, 3, 6): + if (amdgpu_mes) { + amdgpu_device_ip_block_add(adev, &mes_v10_1_ip_block); + adev->enable_mes = true; + if (amdgpu_mes_kiq) + adev->enable_mes_kiq = true; + } + break; + case IP_VERSION(11, 0, 0): + case IP_VERSION(11, 0, 1): + case IP_VERSION(11, 0, 2): + case IP_VERSION(11, 0, 3): + amdgpu_device_ip_block_add(adev, &mes_v11_0_ip_block); + adev->enable_mes = true; + adev->enable_mes_kiq = true; break; default: break; @@ -1037,6 +1966,7 @@ int amdgpu_discovery_set_ip_blocks(struct amdgpu_device *adev) case CHIP_VEGA10: vega10_reg_base_init(adev); adev->sdma.num_instances = 2; + adev->gmc.num_umc = 4; adev->ip_versions[MMHUB_HWIP][0] = IP_VERSION(9, 0, 0); adev->ip_versions[ATHUB_HWIP][0] = IP_VERSION(9, 0, 0); adev->ip_versions[OSSSYS_HWIP][0] = IP_VERSION(4, 0, 0); @@ -1058,6 +1988,7 @@ int amdgpu_discovery_set_ip_blocks(struct amdgpu_device *adev) case CHIP_VEGA12: vega10_reg_base_init(adev); adev->sdma.num_instances = 2; + adev->gmc.num_umc = 4; adev->ip_versions[MMHUB_HWIP][0] = IP_VERSION(9, 3, 0); adev->ip_versions[ATHUB_HWIP][0] = IP_VERSION(9, 3, 0); adev->ip_versions[OSSSYS_HWIP][0] = IP_VERSION(4, 0, 1); @@ -1080,6 +2011,7 @@ int amdgpu_discovery_set_ip_blocks(struct amdgpu_device *adev) vega10_reg_base_init(adev); adev->sdma.num_instances = 1; adev->vcn.num_vcn_inst = 1; + adev->gmc.num_umc = 2; if (adev->apu_flags & AMD_APU_IS_RAVEN2) { adev->ip_versions[MMHUB_HWIP][0] = IP_VERSION(9, 2, 0); adev->ip_versions[ATHUB_HWIP][0] = IP_VERSION(9, 2, 0); @@ -1117,6 +2049,7 @@ int amdgpu_discovery_set_ip_blocks(struct amdgpu_device *adev) case CHIP_VEGA20: vega20_reg_base_init(adev); adev->sdma.num_instances = 2; + adev->gmc.num_umc = 8; adev->ip_versions[MMHUB_HWIP][0] = IP_VERSION(9, 4, 0); adev->ip_versions[ATHUB_HWIP][0] = IP_VERSION(9, 4, 0); adev->ip_versions[OSSSYS_HWIP][0] = IP_VERSION(4, 2, 0); @@ -1140,6 +2073,7 @@ int amdgpu_discovery_set_ip_blocks(struct amdgpu_device *adev) arct_reg_base_init(adev); adev->sdma.num_instances = 8; adev->vcn.num_vcn_inst = 2; + adev->gmc.num_umc = 8; adev->ip_versions[MMHUB_HWIP][0] = IP_VERSION(9, 4, 1); adev->ip_versions[ATHUB_HWIP][0] = IP_VERSION(9, 4, 1); adev->ip_versions[OSSSYS_HWIP][0] = IP_VERSION(4, 2, 1); @@ -1167,6 +2101,7 @@ int amdgpu_discovery_set_ip_blocks(struct amdgpu_device *adev) aldebaran_reg_base_init(adev); adev->sdma.num_instances = 5; adev->vcn.num_vcn_inst = 2; + adev->gmc.num_umc = 4; adev->ip_versions[MMHUB_HWIP][0] = IP_VERSION(9, 4, 2); adev->ip_versions[ATHUB_HWIP][0] = IP_VERSION(9, 4, 2); adev->ip_versions[OSSSYS_HWIP][0] = IP_VERSION(4, 4, 0); @@ -1194,11 +2129,9 @@ int amdgpu_discovery_set_ip_blocks(struct amdgpu_device *adev) return -EINVAL; amdgpu_discovery_harvest_ip(adev); - - if (!adev->mman.discovery_bin) { - DRM_ERROR("ip discovery uninitialized\n"); - return -EINVAL; - } + amdgpu_discovery_get_gfx_info(adev); + amdgpu_discovery_get_mall_info(adev); + amdgpu_discovery_get_vcn_info(adev); break; } @@ -1219,6 +2152,7 @@ int amdgpu_discovery_set_ip_blocks(struct amdgpu_device *adev) case IP_VERSION(10, 1, 1): case IP_VERSION(10, 1, 2): case IP_VERSION(10, 1, 3): + case IP_VERSION(10, 1, 4): case IP_VERSION(10, 3, 0): case IP_VERSION(10, 3, 2): case IP_VERSION(10, 3, 4): @@ -1231,10 +2165,41 @@ int amdgpu_discovery_set_ip_blocks(struct amdgpu_device *adev) case IP_VERSION(10, 3, 3): adev->family = AMDGPU_FAMILY_YC; break; + case IP_VERSION(10, 3, 6): + adev->family = AMDGPU_FAMILY_GC_10_3_6; + break; + case IP_VERSION(10, 3, 7): + adev->family = AMDGPU_FAMILY_GC_10_3_7; + break; + case IP_VERSION(11, 0, 0): + case IP_VERSION(11, 0, 2): + case IP_VERSION(11, 0, 3): + adev->family = AMDGPU_FAMILY_GC_11_0_0; + break; + case IP_VERSION(11, 0, 1): + adev->family = AMDGPU_FAMILY_GC_11_0_1; + break; default: return -EINVAL; } + switch (adev->ip_versions[GC_HWIP][0]) { + case IP_VERSION(9, 1, 0): + case IP_VERSION(9, 2, 2): + case IP_VERSION(9, 3, 0): + case IP_VERSION(10, 1, 3): + case IP_VERSION(10, 1, 4): + case IP_VERSION(10, 3, 1): + case IP_VERSION(10, 3, 3): + case IP_VERSION(10, 3, 6): + case IP_VERSION(10, 3, 7): + case IP_VERSION(11, 0, 1): + adev->flags |= AMD_IS_APU; + break; + default: + break; + } + if (adev->ip_versions[XGMI_HWIP][0] == IP_VERSION(4, 8, 0)) adev->gmc.xgmi.supported = true; @@ -1253,16 +2218,15 @@ int amdgpu_discovery_set_ip_blocks(struct amdgpu_device *adev) break; case IP_VERSION(7, 4, 0): case IP_VERSION(7, 4, 1): - adev->nbio.funcs = &nbio_v7_4_funcs; - adev->nbio.hdp_flush_reg = &nbio_v7_4_hdp_flush_reg; - break; case IP_VERSION(7, 4, 4): adev->nbio.funcs = &nbio_v7_4_funcs; - adev->nbio.hdp_flush_reg = &nbio_v7_4_hdp_flush_reg_ald; + adev->nbio.hdp_flush_reg = &nbio_v7_4_hdp_flush_reg; break; case IP_VERSION(7, 2, 0): case IP_VERSION(7, 2, 1): + case IP_VERSION(7, 3, 0): case IP_VERSION(7, 5, 0): + case IP_VERSION(7, 5, 1): adev->nbio.funcs = &nbio_v7_2_funcs; adev->nbio.hdp_flush_reg = &nbio_v7_2_hdp_flush_reg; break; @@ -1270,15 +2234,24 @@ int amdgpu_discovery_set_ip_blocks(struct amdgpu_device *adev) case IP_VERSION(2, 3, 0): case IP_VERSION(2, 3, 1): case IP_VERSION(2, 3, 2): - adev->nbio.funcs = &nbio_v2_3_funcs; - adev->nbio.hdp_flush_reg = &nbio_v2_3_hdp_flush_reg; - break; case IP_VERSION(3, 3, 0): case IP_VERSION(3, 3, 1): case IP_VERSION(3, 3, 2): case IP_VERSION(3, 3, 3): adev->nbio.funcs = &nbio_v2_3_funcs; - adev->nbio.hdp_flush_reg = &nbio_v2_3_hdp_flush_reg_sc; + adev->nbio.hdp_flush_reg = &nbio_v2_3_hdp_flush_reg; + break; + case IP_VERSION(4, 3, 0): + case IP_VERSION(4, 3, 1): + if (amdgpu_sriov_vf(adev)) + adev->nbio.funcs = &nbio_v4_3_sriov_funcs; + else + adev->nbio.funcs = &nbio_v4_3_funcs; + adev->nbio.hdp_flush_reg = &nbio_v4_3_hdp_flush_reg; + break; + case IP_VERSION(7, 7, 0): + adev->nbio.funcs = &nbio_v7_7_funcs; + adev->nbio.hdp_flush_reg = &nbio_v7_7_hdp_flush_reg; break; default: break; @@ -1303,6 +2276,13 @@ int amdgpu_discovery_set_ip_blocks(struct amdgpu_device *adev) case IP_VERSION(5, 2, 0): adev->hdp.funcs = &hdp_v5_0_funcs; break; + case IP_VERSION(5, 2, 1): + adev->hdp.funcs = &hdp_v5_2_funcs; + break; + case IP_VERSION(6, 0, 0): + case IP_VERSION(6, 0, 1): + adev->hdp.funcs = &hdp_v6_0_funcs; + break; default: break; } @@ -1345,11 +2325,28 @@ int amdgpu_discovery_set_ip_blocks(struct amdgpu_device *adev) case IP_VERSION(11, 0, 11): case IP_VERSION(11, 5, 0): case IP_VERSION(13, 0, 1): + case IP_VERSION(13, 0, 9): + case IP_VERSION(13, 0, 10): adev->smuio.funcs = &smuio_v11_0_6_funcs; break; case IP_VERSION(13, 0, 2): adev->smuio.funcs = &smuio_v13_0_funcs; break; + case IP_VERSION(13, 0, 6): + case IP_VERSION(13, 0, 8): + adev->smuio.funcs = &smuio_v13_0_6_funcs; + break; + default: + break; + } + + switch (adev->ip_versions[LSDMA_HWIP][0]) { + case IP_VERSION(6, 0, 0): + case IP_VERSION(6, 0, 1): + case IP_VERSION(6, 0, 2): + case IP_VERSION(6, 0, 3): + adev->lsdma.funcs = &lsdma_v6_0_funcs; + break; default: break; } @@ -1400,8 +2397,9 @@ int amdgpu_discovery_set_ip_blocks(struct amdgpu_device *adev) if (r) return r; - if (adev->firmware.load_type == AMDGPU_FW_LOAD_DIRECT && - !amdgpu_sriov_vf(adev)) { + if ((adev->firmware.load_type == AMDGPU_FW_LOAD_DIRECT && + !amdgpu_sriov_vf(adev)) || + (adev->firmware.load_type == AMDGPU_FW_LOAD_RLC_BACKDOOR_AUTO && amdgpu_dpm == 1)) { r = amdgpu_discovery_set_smu_ip_blocks(adev); if (r) return r; @@ -1411,11 +2409,9 @@ int amdgpu_discovery_set_ip_blocks(struct amdgpu_device *adev) if (r) return r; - if (adev->enable_mes) { - r = amdgpu_discovery_set_mes_ip_blocks(adev); - if (r) - return r; - } + r = amdgpu_discovery_set_mes_ip_blocks(adev); + if (r) + return r; return 0; } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.h index 14537cec19db..8563dd4a7dc2 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.h @@ -28,12 +28,8 @@ #define DISCOVERY_TMR_OFFSET (64 << 10) void amdgpu_discovery_fini(struct amdgpu_device *adev); -int amdgpu_discovery_reg_base_init(struct amdgpu_device *adev); -void amdgpu_discovery_harvest_ip(struct amdgpu_device *adev); int amdgpu_discovery_get_ip_version(struct amdgpu_device *adev, int hw_id, int number_instance, int *major, int *minor, int *revision); - -int amdgpu_discovery_get_gfx_info(struct amdgpu_device *adev); int amdgpu_discovery_set_ip_blocks(struct amdgpu_device *adev); #endif /* __AMDGPU_DISCOVERY__ */ diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_display.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_display.c index 82011e75ed85..1a06b8d724f3 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_display.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_display.c @@ -30,6 +30,9 @@ #include "atom.h" #include "amdgpu_connectors.h" #include "amdgpu_display.h" +#include "soc15_common.h" +#include "gc/gc_11_0_0_offset.h" +#include "gc/gc_11_0_0_sh_mask.h" #include <asm/div64.h> #include <linux/pci.h> @@ -41,6 +44,11 @@ #include <drm/drm_fourcc.h> #include <drm/drm_vblank.h> +static int amdgpu_display_framebuffer_init(struct drm_device *dev, + struct amdgpu_framebuffer *rfb, + const struct drm_mode_fb_cmd2 *mode_cmd, + struct drm_gem_object *obj); + static void amdgpu_display_flip_callback(struct dma_fence *f, struct dma_fence_cb *cb) { @@ -113,8 +121,9 @@ static void amdgpu_display_flip_work_func(struct work_struct *__work) spin_unlock_irqrestore(&crtc->dev->event_lock, flags); - DRM_DEBUG_DRIVER("crtc:%d[%p], pflip_stat:AMDGPU_FLIP_SUBMITTED, work: %p,\n", - amdgpu_crtc->crtc_id, amdgpu_crtc, work); + drm_dbg_vbl(adev_to_drm(adev), + "crtc:%d[%p], pflip_stat:AMDGPU_FLIP_SUBMITTED, work: %p,\n", + amdgpu_crtc->crtc_id, amdgpu_crtc, work); } @@ -200,8 +209,9 @@ int amdgpu_display_crtc_page_flip_target(struct drm_crtc *crtc, goto unpin; } - r = dma_resv_get_fences(new_abo->tbo.base.resv, NULL, - &work->shared_count, &work->shared); + r = dma_resv_get_fences(new_abo->tbo.base.resv, DMA_RESV_USAGE_WRITE, + &work->shared_count, + &work->shared); if (unlikely(r != 0)) { DRM_ERROR("failed to get fences for buffer\n"); goto unpin; @@ -504,28 +514,9 @@ uint32_t amdgpu_display_supported_domains(struct amdgpu_device *adev, */ if ((bo_flags & AMDGPU_GEM_CREATE_CPU_GTT_USWC) && amdgpu_bo_support_uswc(bo_flags) && - amdgpu_device_asic_has_dc_support(adev->asic_type)) { - switch (adev->asic_type) { - case CHIP_CARRIZO: - case CHIP_STONEY: - domain |= AMDGPU_GEM_DOMAIN_GTT; - break; - case CHIP_RAVEN: - /* enable S/G on PCO and RV2 */ - if ((adev->apu_flags & AMD_APU_IS_RAVEN2) || - (adev->apu_flags & AMD_APU_IS_PICASSO)) - domain |= AMDGPU_GEM_DOMAIN_GTT; - break; - case CHIP_RENOIR: - case CHIP_VANGOGH: - case CHIP_YELLOW_CARP: - domain |= AMDGPU_GEM_DOMAIN_GTT; - break; - - default: - break; - } - } + amdgpu_device_asic_has_dc_support(adev->asic_type) && + adev->mode_info.gpu_vm_support) + domain |= AMDGPU_GEM_DOMAIN_GTT; #endif return domain; @@ -675,6 +666,11 @@ static int convert_tiling_flags_to_modifier(struct amdgpu_framebuffer *afb) { struct amdgpu_device *adev = drm_to_adev(afb->base.dev); uint64_t modifier = 0; + int num_pipes = 0; + int num_pkrs = 0; + + num_pkrs = adev->gfx.config.gb_addr_config_fields.num_pkrs; + num_pipes = adev->gfx.config.gb_addr_config_fields.num_pipes; if (!afb->tiling_flags || !AMDGPU_TILING_GET(afb->tiling_flags, SWIZZLE_MODE)) { modifier = DRM_FORMAT_MOD_LINEAR; @@ -687,7 +683,7 @@ static int convert_tiling_flags_to_modifier(struct amdgpu_framebuffer *afb) int bank_xor_bits = 0; int packers = 0; int rb = 0; - int pipes = ilog2(adev->gfx.config.gb_addr_config_fields.num_pipes); + int pipes = ilog2(num_pipes); uint32_t dcc_offset = AMDGPU_TILING_GET(afb->tiling_flags, DCC_OFFSET_256B); switch (swizzle >> 2) { @@ -703,14 +699,19 @@ static int convert_tiling_flags_to_modifier(struct amdgpu_framebuffer *afb) case 6: /* 64 KiB _X */ block_size_bits = 16; break; + case 7: /* 256 KiB */ + block_size_bits = 18; + break; default: /* RESERVED or VAR */ return -EINVAL; } - if (adev->asic_type >= CHIP_SIENNA_CICHLID) + if (adev->ip_versions[GC_HWIP][0] >= IP_VERSION(11, 0, 0)) + version = AMD_FMT_MOD_TILE_VER_GFX11; + else if (adev->ip_versions[GC_HWIP][0] >= IP_VERSION(10, 3, 0)) version = AMD_FMT_MOD_TILE_VER_GFX10_RBPLUS; - else if (adev->family == AMDGPU_FAMILY_NV) + else if (adev->ip_versions[GC_HWIP][0] >= IP_VERSION(10, 0, 0)) version = AMD_FMT_MOD_TILE_VER_GFX10; else version = AMD_FMT_MOD_TILE_VER_GFX9; @@ -719,19 +720,32 @@ static int convert_tiling_flags_to_modifier(struct amdgpu_framebuffer *afb) case 0: /* Z microtiling */ return -EINVAL; case 1: /* S microtiling */ - if (!has_xor) - version = AMD_FMT_MOD_TILE_VER_GFX9; + if (adev->ip_versions[GC_HWIP][0] < IP_VERSION(11, 0, 0)) { + if (!has_xor) + version = AMD_FMT_MOD_TILE_VER_GFX9; + } break; case 2: - if (!has_xor && afb->base.format->cpp[0] != 4) - version = AMD_FMT_MOD_TILE_VER_GFX9; + if (adev->ip_versions[GC_HWIP][0] < IP_VERSION(11, 0, 0)) { + if (!has_xor && afb->base.format->cpp[0] != 4) + version = AMD_FMT_MOD_TILE_VER_GFX9; + } break; case 3: break; } if (has_xor) { + if (num_pipes == num_pkrs && num_pkrs == 0) { + DRM_ERROR("invalid number of pipes and packers\n"); + return -EINVAL; + } + switch (version) { + case AMD_FMT_MOD_TILE_VER_GFX11: + pipe_xor_bits = min(block_size_bits - 8, pipes); + packers = ilog2(adev->gfx.config.gb_addr_config_fields.num_pkrs); + break; case AMD_FMT_MOD_TILE_VER_GFX10_RBPLUS: pipe_xor_bits = min(block_size_bits - 8, pipes); packers = min(block_size_bits - 8 - pipe_xor_bits, @@ -765,9 +779,10 @@ static int convert_tiling_flags_to_modifier(struct amdgpu_framebuffer *afb) u64 render_dcc_offset; /* Enable constant encode on RAVEN2 and later. */ - bool dcc_constant_encode = adev->asic_type > CHIP_RAVEN || + bool dcc_constant_encode = (adev->asic_type > CHIP_RAVEN || (adev->asic_type == CHIP_RAVEN && - adev->external_rev_id >= 0x81); + adev->external_rev_id >= 0x81)) && + adev->ip_versions[GC_HWIP][0] < IP_VERSION(11, 0, 0); int max_cblock_size = dcc_i64b ? AMD_FMT_MOD_DCC_BLOCK_64B : dcc_i128b ? AMD_FMT_MOD_DCC_BLOCK_128B : @@ -804,7 +819,7 @@ static int convert_tiling_flags_to_modifier(struct amdgpu_framebuffer *afb) if (adev->family >= AMDGPU_FAMILY_NV) { int extra_pipe = 0; - if (adev->asic_type >= CHIP_SIENNA_CICHLID && + if ((adev->ip_versions[GC_HWIP][0] >= IP_VERSION(10, 3, 0)) && pipes == packers && pipes > 1) extra_pipe = 1; @@ -882,10 +897,11 @@ static unsigned int get_dcc_block_size(uint64_t modifier, bool rb_aligned, return max(10 + (rb_aligned ? (int)AMD_FMT_MOD_GET(RB, modifier) : 0), 12); } case AMD_FMT_MOD_TILE_VER_GFX10: - case AMD_FMT_MOD_TILE_VER_GFX10_RBPLUS: { + case AMD_FMT_MOD_TILE_VER_GFX10_RBPLUS: + case AMD_FMT_MOD_TILE_VER_GFX11: { int pipes_log2 = AMD_FMT_MOD_GET(PIPE_XOR_BITS, modifier); - if (ver == AMD_FMT_MOD_TILE_VER_GFX10_RBPLUS && pipes_log2 > 1 && + if (ver >= AMD_FMT_MOD_TILE_VER_GFX10_RBPLUS && pipes_log2 > 1 && AMD_FMT_MOD_GET(PACKERS, modifier) == pipes_log2) ++pipes_log2; @@ -954,7 +970,7 @@ static int amdgpu_display_verify_sizes(struct amdgpu_framebuffer *rfb) int ret; unsigned int i, block_width, block_height, block_size_log2; - if (!rfb->base.dev->mode_config.allow_fb_modifiers) + if (rfb->base.dev->mode_config.fb_modifiers_not_supported) return 0; for (i = 0; i < format_info->num_planes; ++i) { @@ -978,6 +994,9 @@ static int amdgpu_display_verify_sizes(struct amdgpu_framebuffer *rfb) case DC_SW_64KB_S_X: block_size_log2 = 16; break; + case DC_SW_VAR_S_X: + block_size_log2 = 18; + break; default: drm_dbg_kms(rfb->base.dev, "Swizzle mode with unknown block size: %d\n", swizzle); @@ -1056,35 +1075,11 @@ static int amdgpu_display_get_fb_info(const struct amdgpu_framebuffer *amdgpu_fb return r; } -int amdgpu_display_gem_fb_init(struct drm_device *dev, - struct amdgpu_framebuffer *rfb, - const struct drm_mode_fb_cmd2 *mode_cmd, - struct drm_gem_object *obj) -{ - int ret; - - rfb->base.obj[0] = obj; - drm_helper_mode_fill_fb_struct(dev, &rfb->base, mode_cmd); - - ret = amdgpu_display_framebuffer_init(dev, rfb, mode_cmd, obj); - if (ret) - goto err; - - ret = drm_framebuffer_init(dev, &rfb->base, &amdgpu_fb_funcs); - if (ret) - goto err; - - return 0; -err: - drm_dbg_kms(dev, "Failed to init gem fb: %d\n", ret); - rfb->base.obj[0] = NULL; - return ret; -} - -int amdgpu_display_gem_fb_verify_and_init( - struct drm_device *dev, struct amdgpu_framebuffer *rfb, - struct drm_file *file_priv, const struct drm_mode_fb_cmd2 *mode_cmd, - struct drm_gem_object *obj) +static int amdgpu_display_gem_fb_verify_and_init(struct drm_device *dev, + struct amdgpu_framebuffer *rfb, + struct drm_file *file_priv, + const struct drm_mode_fb_cmd2 *mode_cmd, + struct drm_gem_object *obj) { int ret; @@ -1106,6 +1101,7 @@ int amdgpu_display_gem_fb_verify_and_init( goto err; ret = drm_framebuffer_init(dev, &rfb->base, &amdgpu_fb_funcs); + if (ret) goto err; @@ -1116,10 +1112,10 @@ err: return ret; } -int amdgpu_display_framebuffer_init(struct drm_device *dev, - struct amdgpu_framebuffer *rfb, - const struct drm_mode_fb_cmd2 *mode_cmd, - struct drm_gem_object *obj) +static int amdgpu_display_framebuffer_init(struct drm_device *dev, + struct amdgpu_framebuffer *rfb, + const struct drm_mode_fb_cmd2 *mode_cmd, + struct drm_gem_object *obj) { struct amdgpu_device *adev = drm_to_adev(dev); int ret, i; @@ -1141,7 +1137,7 @@ int amdgpu_display_framebuffer_init(struct drm_device *dev, if (ret) return ret; - if (!dev->mode_config.allow_fb_modifiers) { + if (dev->mode_config.fb_modifiers_not_supported && !adev->enable_virtual_display) { drm_WARN_ONCE(dev, adev->family >= AMDGPU_FAMILY_AI, "GFX9+ requires FB check based on format modifier\n"); ret = check_tiling_flags_gfx6(rfb); @@ -1149,7 +1145,7 @@ int amdgpu_display_framebuffer_init(struct drm_device *dev, return ret; } - if (dev->mode_config.allow_fb_modifiers && + if (!dev->mode_config.fb_modifiers_not_supported && !(rfb->base.flags & DRM_MODE_FB_MODIFIERS)) { ret = convert_tiling_flags_to_modifier(rfb); if (ret) { @@ -1564,6 +1560,21 @@ bool amdgpu_crtc_get_scanout_position(struct drm_crtc *crtc, stime, etime, mode); } +static bool +amdgpu_display_robj_is_fb(struct amdgpu_device *adev, struct amdgpu_bo *robj) +{ + struct drm_device *dev = adev_to_drm(adev); + struct drm_fb_helper *fb_helper = dev->fb_helper; + + if (!fb_helper || !fb_helper->buffer) + return false; + + if (gem_to_amdgpu_bo(fb_helper->buffer->gem) != robj) + return false; + + return true; +} + int amdgpu_display_suspend_helper(struct amdgpu_device *adev) { struct drm_device *dev = adev_to_drm(adev); @@ -1599,10 +1610,12 @@ int amdgpu_display_suspend_helper(struct amdgpu_device *adev) continue; } robj = gem_to_amdgpu_bo(fb->obj[0]); - r = amdgpu_bo_reserve(robj, true); - if (r == 0) { - amdgpu_bo_unpin(robj); - amdgpu_bo_unreserve(robj); + if (!amdgpu_display_robj_is_fb(adev, robj)) { + r = amdgpu_bo_reserve(robj, true); + if (r == 0) { + amdgpu_bo_unpin(robj); + amdgpu_bo_unreserve(robj); + } } } return 0; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_display.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_display.h index 7b6d83e2b13c..560352f7c317 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_display.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_display.h @@ -35,8 +35,6 @@ #define amdgpu_display_add_encoder(adev, e, s, c) (adev)->mode_info.funcs->add_encoder((adev), (e), (s), (c)) #define amdgpu_display_add_connector(adev, ci, sd, ct, ib, coi, h, r) (adev)->mode_info.funcs->add_connector((adev), (ci), (sd), (ct), (ib), (coi), (h), (r)) -int amdgpu_display_freesync_ioctl(struct drm_device *dev, void *data, - struct drm_file *filp); void amdgpu_display_update_priority(struct amdgpu_device *adev); uint32_t amdgpu_display_supported_domains(struct amdgpu_device *adev, uint64_t bo_flags); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_dma_buf.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_dma_buf.c index 579adfafe4d0..7bd8e33b14be 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_dma_buf.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_dma_buf.c @@ -58,7 +58,7 @@ static int amdgpu_dma_buf_attach(struct dma_buf *dmabuf, struct amdgpu_device *adev = amdgpu_ttm_adev(bo->tbo.bdev); int r; - if (pci_p2pdma_distance_many(adev->pdev, &attach->dev, 1, true) < 0) + if (pci_p2pdma_distance(adev->pdev, attach->dev, false) < 0) attach->peer2peer = false; r = pm_runtime_get_sync(adev_to_drm(adev)->dev); @@ -102,21 +102,9 @@ static int amdgpu_dma_buf_pin(struct dma_buf_attachment *attach) { struct drm_gem_object *obj = attach->dmabuf->priv; struct amdgpu_bo *bo = gem_to_amdgpu_bo(obj); - int r; /* pin buffer into GTT */ - r = amdgpu_bo_pin(bo, AMDGPU_GEM_DOMAIN_GTT); - if (r) - return r; - - if (bo->tbo.moving) { - r = dma_fence_wait(bo->tbo.moving, true); - if (r) { - amdgpu_bo_unpin(bo); - return r; - } - } - return 0; + return amdgpu_bo_pin(bo, AMDGPU_GEM_DOMAIN_GTT); } /** diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_doorbell.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_doorbell.h index 89e6ad30396f..7199b6b0be81 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_doorbell.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_doorbell.h @@ -52,8 +52,11 @@ struct amdgpu_doorbell_index { uint32_t userqueue_end; uint32_t gfx_ring0; uint32_t gfx_ring1; + uint32_t gfx_userqueue_start; + uint32_t gfx_userqueue_end; uint32_t sdma_engine[8]; - uint32_t mes_ring; + uint32_t mes_ring0; + uint32_t mes_ring1; uint32_t ih; union { struct { @@ -174,11 +177,15 @@ typedef enum _AMDGPU_NAVI10_DOORBELL_ASSIGNMENT AMDGPU_NAVI10_DOORBELL_MEC_RING5 = 0x008, AMDGPU_NAVI10_DOORBELL_MEC_RING6 = 0x009, AMDGPU_NAVI10_DOORBELL_MEC_RING7 = 0x00A, - AMDGPU_NAVI10_DOORBELL_USERQUEUE_START = 0x00B, + AMDGPU_NAVI10_DOORBELL_MES_RING0 = 0x00B, + AMDGPU_NAVI10_DOORBELL_MES_RING1 = 0x00C, + AMDGPU_NAVI10_DOORBELL_USERQUEUE_START = 0x00D, AMDGPU_NAVI10_DOORBELL_USERQUEUE_END = 0x08A, AMDGPU_NAVI10_DOORBELL_GFX_RING0 = 0x08B, AMDGPU_NAVI10_DOORBELL_GFX_RING1 = 0x08C, - AMDGPU_NAVI10_DOORBELL_MES_RING = 0x090, + AMDGPU_NAVI10_DOORBELL_GFX_USERQUEUE_START = 0x08D, + AMDGPU_NAVI10_DOORBELL_GFX_USERQUEUE_END = 0x0FF, + /* SDMA:256~335*/ AMDGPU_NAVI10_DOORBELL_sDMA_ENGINE0 = 0x100, AMDGPU_NAVI10_DOORBELL_sDMA_ENGINE1 = 0x10A, diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c index b63ed1ddf713..bf2d50c8c92a 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c @@ -39,6 +39,7 @@ #include <linux/suspend.h> #include <linux/cc_platform.h> #include <linux/fb.h> +#include <linux/dynamic_debug.h> #include "amdgpu.h" #include "amdgpu_irq.h" @@ -81,7 +82,7 @@ * - 3.24.0 - Add high priority compute support for gfx9 * - 3.25.0 - Add support for sensor query info (stable pstate sclk/mclk). * - 3.26.0 - GFX9: Process AMDGPU_IB_FLAG_TC_WB_NOT_INVALIDATE. - * - 3.27.0 - Add new chunk to to AMDGPU_CS to enable BO_LIST creation. + * - 3.27.0 - Add new chunk to AMDGPU_CS to enable BO_LIST creation. * - 3.28.0 - Add AMDGPU_CHUNK_ID_SCHEDULED_DEPENDENCIES * - 3.29.0 - Add AMDGPU_IB_FLAG_RESET_GDS_MAX_WAVE_ID * - 3.30.0 - Add AMDGPU_SCHED_OP_CONTEXT_PRIORITY_OVERRIDE. @@ -99,9 +100,14 @@ * - 3.42.0 - Add 16bpc fixed point display support * - 3.43.0 - Add device hot plug/unplug support * - 3.44.0 - DCN3 supports DCC independent block settings: !64B && 128B, 64B && 128B + * - 3.45.0 - Add context ioctl stable pstate interface + * - 3.46.0 - To enable hot plug amdgpu tests in libdrm + * - 3.47.0 - Add AMDGPU_GEM_CREATE_DISCARDABLE and AMDGPU_VM_NOALLOC flags + * - 3.48.0 - Add IP discovery version info to HW INFO + * 3.49.0 - Add gang submit into CS IOCTL */ #define KMS_DRIVER_MAJOR 3 -#define KMS_DRIVER_MINOR 44 +#define KMS_DRIVER_MINOR 49 #define KMS_DRIVER_PATCHLEVEL 0 int amdgpu_vram_limit; @@ -109,8 +115,6 @@ int amdgpu_vis_vram_limit; int amdgpu_gart_size = -1; /* auto */ int amdgpu_gtt_size = -1; /* auto */ int amdgpu_moverate = -1; /* auto */ -int amdgpu_benchmarking; -int amdgpu_testing; int amdgpu_audio = -1; int amdgpu_disp_priority; int amdgpu_hw_i2c; @@ -136,7 +140,7 @@ int amdgpu_sched_jobs = 32; int amdgpu_sched_hw_submission = 2; uint amdgpu_pcie_gen_cap; uint amdgpu_pcie_lane_cap; -uint amdgpu_cg_mask = 0xffffffff; +u64 amdgpu_cg_mask = 0xffffffffffffffff; uint amdgpu_pg_mask = 0xffffffff; uint amdgpu_sdma_phase_quantum = 32; char *amdgpu_disable_cu = NULL; @@ -167,20 +171,35 @@ int amdgpu_smu_pptable_id = -1; */ uint amdgpu_dc_feature_mask = 2; uint amdgpu_dc_debug_mask; +uint amdgpu_dc_visual_confirm; int amdgpu_async_gfx_ring = 1; int amdgpu_mcbp; int amdgpu_discovery = -1; int amdgpu_mes; +int amdgpu_mes_kiq; int amdgpu_noretry = -1; int amdgpu_force_asic_type = -1; int amdgpu_tmz = -1; /* auto */ -uint amdgpu_freesync_vid_mode; int amdgpu_reset_method = -1; /* auto */ int amdgpu_num_kcq = -1; int amdgpu_smartshift_bias; +int amdgpu_use_xgmi_p2p = 1; +int amdgpu_vcnfw_log; static void amdgpu_drv_delayed_reset_work_handler(struct work_struct *work); +DECLARE_DYNDBG_CLASSMAP(drm_debug_classes, DD_CLASS_TYPE_DISJOINT_BITS, 0, + "DRM_UT_CORE", + "DRM_UT_DRIVER", + "DRM_UT_KMS", + "DRM_UT_PRIME", + "DRM_UT_ATOMIC", + "DRM_UT_VBL", + "DRM_UT_STATE", + "DRM_UT_LEASE", + "DRM_UT_DP", + "DRM_UT_DRMRES"); + struct amdgpu_mgpu_info mgpu_info = { .mutex = __MUTEX_INITIALIZER(mgpu_info.mutex), .delayed_reset_work = __DELAYED_WORK_INITIALIZER( @@ -232,20 +251,6 @@ MODULE_PARM_DESC(moverate, "Maximum buffer migration rate in MB/s. (32, 64, etc. module_param_named(moverate, amdgpu_moverate, int, 0600); /** - * DOC: benchmark (int) - * Run benchmarks. The default is 0 (Skip benchmarks). - */ -MODULE_PARM_DESC(benchmark, "Run benchmark"); -module_param_named(benchmark, amdgpu_benchmarking, int, 0444); - -/** - * DOC: test (int) - * Test BO GTT->VRAM and VRAM->GTT GPU copies. The default is 0 (Skip test, only set 1 to run test). - */ -MODULE_PARM_DESC(test, "Run tests"); -module_param_named(test, amdgpu_testing, int, 0444); - -/** * DOC: audio (int) * Set HDMI/DPAudio. Only affects non-DC display handling. The default is -1 (Enabled), set 0 to disabled it. */ @@ -319,7 +324,7 @@ module_param_named(dpm, amdgpu_dpm, int, 0444); * to -1 to select the default loading mode for the ASIC, as defined * by the driver. The default is -1 (auto). */ -MODULE_PARM_DESC(fw_load_type, "firmware loading type (0 = force direct if supported, -1 = auto)"); +MODULE_PARM_DESC(fw_load_type, "firmware loading type (3 = rlc backdoor autoload if supported, 2 = smu load if supported, 1 = psp load, 0 = force direct if supported, -1 = auto)"); module_param_named(fw_load_type, amdgpu_fw_load_type, int, 0444); /** @@ -467,12 +472,12 @@ MODULE_PARM_DESC(pcie_lane_cap, "PCIE Lane Caps (0: autodetect (default))"); module_param_named(pcie_lane_cap, amdgpu_pcie_lane_cap, uint, 0444); /** - * DOC: cg_mask (uint) + * DOC: cg_mask (ullong) * Override Clockgating features enabled on GPU (0 = disable clock gating). See the AMD_CG_SUPPORT flags in - * drivers/gpu/drm/amd/include/amd_shared.h. The default is 0xffffffff (all enabled). + * drivers/gpu/drm/amd/include/amd_shared.h. The default is 0xffffffffffffffff (all enabled). */ MODULE_PARM_DESC(cg_mask, "Clockgating flags mask (0 = disable clock gating)"); -module_param_named(cg_mask, amdgpu_cg_mask, uint, 0444); +module_param_named(cg_mask, amdgpu_cg_mask, ullong, 0444); /** * DOC: pg_mask (uint) @@ -650,6 +655,15 @@ MODULE_PARM_DESC(mes, module_param_named(mes, amdgpu_mes, int, 0444); /** + * DOC: mes_kiq (int) + * Enable Micro Engine Scheduler KIQ. This is a new engine pipe for kiq. + * (0 = disabled (default), 1 = enabled) + */ +MODULE_PARM_DESC(mes_kiq, + "Enable Micro Engine Scheduler KIQ (0 = disabled (default), 1 = enabled)"); +module_param_named(mes_kiq, amdgpu_mes_kiq, int, 0444); + +/** * DOC: noretry (int) * Disable XNACK retry in the SQ by default on GFXv9 hardware. On ASICs that * do not support per-process XNACK this also disables retry page faults. @@ -667,6 +681,13 @@ MODULE_PARM_DESC(force_asic_type, "A non negative value used to specify the asic type for all supported GPUs"); module_param_named(force_asic_type, amdgpu_force_asic_type, int, 0444); +/** + * DOC: use_xgmi_p2p (int) + * Enables/disables XGMI P2P interface (0 = disable, 1 = enable). + */ +MODULE_PARM_DESC(use_xgmi_p2p, + "Enable XGMI P2P interface (0 = disable; 1 = enable (default))"); +module_param_named(use_xgmi_p2p, amdgpu_use_xgmi_p2p, int, 0444); #ifdef CONFIG_HSA_AMD @@ -686,7 +707,7 @@ MODULE_PARM_DESC(sched_policy, * Maximum number of processes that HWS can schedule concurrently. The maximum is the * number of VMIDs assigned to the HWS, which is also the default. */ -int hws_max_conc_proc = 8; +int hws_max_conc_proc = -1; module_param(hws_max_conc_proc, int, 0444); MODULE_PARM_DESC(hws_max_conc_proc, "Max # processes HWS can execute concurrently when sched_policy=0 (0 = no concurrency, #VMIDs for KFD = Maximum(default))"); @@ -740,7 +761,7 @@ MODULE_PARM_DESC(debug_largebar, * systems with a broken CRAT table. * * Default is auto (according to asic type, iommu_v2, and crat table, to decide - * whehter use CRAT) + * whether use CRAT) */ int ignore_crat; module_param(ignore_crat, int, 0444); @@ -799,6 +820,16 @@ module_param_named(no_queue_eviction_on_vm_fault, amdgpu_no_queue_eviction_on_vm #endif /** + * DOC: pcie_p2p (bool) + * Enable PCIe P2P (requires large-BAR). Default value: true (on) + */ +#ifdef CONFIG_HSA_AMD_P2P +bool pcie_p2p = true; +module_param(pcie_p2p, bool, 0444); +MODULE_PARM_DESC(pcie_p2p, "Enable PCIe P2P (requires large-BAR). (N = off, Y = on(default))"); +#endif + +/** * DOC: dcfeaturemask (uint) * Override display features enabled. See enum DC_FEATURE_MASK in drivers/gpu/drm/amd/include/amd_shared.h. * The default is the current set of stable display features. @@ -813,6 +844,9 @@ module_param_named(dcfeaturemask, amdgpu_dc_feature_mask, uint, 0444); MODULE_PARM_DESC(dcdebugmask, "all debug options disabled (default))"); module_param_named(dcdebugmask, amdgpu_dc_debug_mask, uint, 0444); +MODULE_PARM_DESC(visualconfirm, "Visual confirm (0 = off (default), 1 = MPO, 5 = PSR)"); +module_param_named(visualconfirm, amdgpu_dc_visual_confirm, uint, 0444); + /** * DOC: abmlevel (uint) * Override the default ABM (Adaptive Backlight Management) level used for DC @@ -844,36 +878,10 @@ MODULE_PARM_DESC(tmz, "Enable TMZ feature (-1 = auto (default), 0 = off, 1 = on) module_param_named(tmz, amdgpu_tmz, int, 0444); /** - * DOC: freesync_video (uint) - * Enable the optimization to adjust front porch timing to achieve seamless - * mode change experience when setting a freesync supported mode for which full - * modeset is not needed. - * - * The Display Core will add a set of modes derived from the base FreeSync - * video mode into the corresponding connector's mode list based on commonly - * used refresh rates and VRR range of the connected display, when users enable - * this feature. From the userspace perspective, they can see a seamless mode - * change experience when the change between different refresh rates under the - * same resolution. Additionally, userspace applications such as Video playback - * can read this modeset list and change the refresh rate based on the video - * frame rate. Finally, the userspace can also derive an appropriate mode for a - * particular refresh rate based on the FreeSync Mode and add it to the - * connector's mode list. - * - * Note: This is an experimental feature. - * - * The default value: 0 (off). - */ -MODULE_PARM_DESC( - freesync_video, - "Enable freesync modesetting optimization feature (0 = off (default), 1 = on)"); -module_param_named(freesync_video, amdgpu_freesync_vid_mode, uint, 0444); - -/** * DOC: reset_method (int) - * GPU reset method (-1 = auto (default), 0 = legacy, 1 = mode0, 2 = mode1, 3 = mode2, 4 = baco, 5 = pci) + * GPU reset method (-1 = auto (default), 0 = legacy, 1 = mode0, 2 = mode1, 3 = mode2, 4 = baco) */ -MODULE_PARM_DESC(reset_method, "GPU reset method (-1 = auto (default), 0 = legacy, 1 = mode0, 2 = mode1, 3 = mode2, 4 = baco/bamaco, 5 = pci)"); +MODULE_PARM_DESC(reset_method, "GPU reset method (-1 = auto (default), 0 = legacy, 1 = mode0, 2 = mode1, 3 = mode2, 4 = baco/bamaco)"); module_param_named(reset_method, amdgpu_reset_method, int, 0444); /** @@ -889,6 +897,13 @@ MODULE_PARM_DESC(num_kcq, "number of kernel compute queue user want to setup (8 module_param_named(num_kcq, amdgpu_num_kcq, int, 0444); /** + * DOC: vcnfw_log (int) + * Enable vcnfw log output for debugging, the default is disabled. + */ +MODULE_PARM_DESC(vcnfw_log, "Enable vcnfw log(0 = disable (default value), 1 = enable)"); +module_param_named(vcnfw_log, amdgpu_vcnfw_log, int, 0444); + +/** * DOC: smu_pptable_id (int) * Used to override pptable id. id = 0 use VBIOS pptable. * id > 0 use the soft pptable with specicfied id. @@ -1525,6 +1540,87 @@ static const u16 amdgpu_unsupported_pciidlist[] = { 0x99A0, 0x99A2, 0x99A4, + /* radeon secondary ids */ + 0x3171, + 0x3e70, + 0x4164, + 0x4165, + 0x4166, + 0x4168, + 0x4170, + 0x4171, + 0x4172, + 0x4173, + 0x496e, + 0x4a69, + 0x4a6a, + 0x4a6b, + 0x4a70, + 0x4a74, + 0x4b69, + 0x4b6b, + 0x4b6c, + 0x4c6e, + 0x4e64, + 0x4e65, + 0x4e66, + 0x4e67, + 0x4e68, + 0x4e69, + 0x4e6a, + 0x4e71, + 0x4f73, + 0x5569, + 0x556b, + 0x556d, + 0x556f, + 0x5571, + 0x5854, + 0x5874, + 0x5940, + 0x5941, + 0x5b72, + 0x5b73, + 0x5b74, + 0x5b75, + 0x5d44, + 0x5d45, + 0x5d6d, + 0x5d6f, + 0x5d72, + 0x5d77, + 0x5e6b, + 0x5e6d, + 0x7120, + 0x7124, + 0x7129, + 0x712e, + 0x712f, + 0x7162, + 0x7163, + 0x7166, + 0x7167, + 0x7172, + 0x7173, + 0x71a0, + 0x71a1, + 0x71a3, + 0x71a7, + 0x71bb, + 0x71e0, + 0x71e1, + 0x71e2, + 0x71e6, + 0x71e7, + 0x71f2, + 0x7269, + 0x726b, + 0x726e, + 0x72a0, + 0x72a8, + 0x72b1, + 0x72b3, + 0x793f, }; static const struct pci_device_id pciidlist[] = { @@ -1861,19 +1957,21 @@ static const struct pci_device_id pciidlist[] = { {0x1002, 0x73FF, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_DIMGREY_CAVEFISH}, /* Aldebaran */ - {0x1002, 0x7408, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_ALDEBARAN|AMD_EXP_HW_SUPPORT}, - {0x1002, 0x740C, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_ALDEBARAN|AMD_EXP_HW_SUPPORT}, - {0x1002, 0x740F, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_ALDEBARAN|AMD_EXP_HW_SUPPORT}, - {0x1002, 0x7410, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_ALDEBARAN|AMD_EXP_HW_SUPPORT}, + {0x1002, 0x7408, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_ALDEBARAN}, + {0x1002, 0x740C, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_ALDEBARAN}, + {0x1002, 0x740F, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_ALDEBARAN}, + {0x1002, 0x7410, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_ALDEBARAN}, /* CYAN_SKILLFISH */ {0x1002, 0x13FE, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_CYAN_SKILLFISH|AMD_IS_APU}, + {0x1002, 0x143F, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_CYAN_SKILLFISH|AMD_IS_APU}, /* BEIGE_GOBY */ {0x1002, 0x7420, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_BEIGE_GOBY}, {0x1002, 0x7421, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_BEIGE_GOBY}, {0x1002, 0x7422, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_BEIGE_GOBY}, {0x1002, 0x7423, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_BEIGE_GOBY}, + {0x1002, 0x7424, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_BEIGE_GOBY}, {0x1002, 0x743F, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_BEIGE_GOBY}, { PCI_DEVICE(0x1002, PCI_ANY_ID), @@ -1893,24 +1991,26 @@ MODULE_DEVICE_TABLE(pci, pciidlist); static const struct drm_driver amdgpu_kms_driver; -static bool amdgpu_is_fw_framebuffer(resource_size_t base, - resource_size_t size) +static void amdgpu_get_secondary_funcs(struct amdgpu_device *adev) { - bool found = false; -#if IS_REACHABLE(CONFIG_FB) - struct apertures_struct *a; - - a = alloc_apertures(1); - if (!a) - return false; + struct pci_dev *p = NULL; + int i; - a->ranges[0].base = base; - a->ranges[0].size = size; - - found = is_firmware_framebuffer(a); - kfree(a); -#endif - return found; + /* 0 - GPU + * 1 - audio + * 2 - USB + * 3 - UCSI + */ + for (i = 1; i < 4; i++) { + p = pci_get_domain_bus_and_slot(pci_domain_nr(adev->pdev->bus), + adev->pdev->bus->number, i); + if (p) { + pm_runtime_get_sync(&p->dev); + pm_runtime_mark_last_busy(&p->dev); + pm_runtime_put_autosuspend(&p->dev); + pci_dev_put(p); + } + } } static int amdgpu_pci_probe(struct pci_dev *pdev, @@ -1921,8 +2021,6 @@ static int amdgpu_pci_probe(struct pci_dev *pdev, unsigned long flags = ent->driver_data; int ret, retry = 0, i; bool supports_atomic = false; - bool is_fw_fb; - resource_size_t base, size; /* skip devices which are owned by radeon */ for (i = 0; i < ARRAY_SIZE(amdgpu_unsupported_pciidlist); i++) { @@ -1930,10 +2028,8 @@ static int amdgpu_pci_probe(struct pci_dev *pdev, return -ENODEV; } - if (flags == 0) { - DRM_INFO("Unsupported asic. Remove me when IP discovery init is in place.\n"); - return -ENODEV; - } + if (amdgpu_aspm == -1 && !pcie_aspm_enabled(pdev)) + amdgpu_aspm = 0; if (amdgpu_virtual_display || amdgpu_device_asic_has_dc_support(flags & AMD_ASIC_MASK)) @@ -1991,10 +2087,6 @@ static int amdgpu_pci_probe(struct pci_dev *pdev, } #endif - base = pci_resource_start(pdev, 0); - size = pci_resource_len(pdev, 0); - is_fw_fb = amdgpu_is_fw_framebuffer(base, size); - /* Get rid of things like offb */ ret = drm_aperture_remove_conflicting_pci_framebuffers(pdev, &amdgpu_kms_driver); if (ret) @@ -2007,7 +2099,6 @@ static int amdgpu_pci_probe(struct pci_dev *pdev, adev->dev = &pdev->dev; adev->pdev = pdev; ddev = adev_to_drm(adev); - adev->is_fw_fb = is_fw_fb; if (!supports_atomic) ddev->driver_features &= ~DRIVER_ATOMIC; @@ -2050,6 +2141,48 @@ retry_init: if (ret) DRM_ERROR("Creating debugfs files failed (%d).\n", ret); + if (adev->pm.rpm_mode != AMDGPU_RUNPM_NONE) { + /* only need to skip on ATPX */ + if (amdgpu_device_supports_px(ddev)) + dev_pm_set_driver_flags(ddev->dev, DPM_FLAG_NO_DIRECT_COMPLETE); + /* we want direct complete for BOCO */ + if (amdgpu_device_supports_boco(ddev)) + dev_pm_set_driver_flags(ddev->dev, DPM_FLAG_SMART_PREPARE | + DPM_FLAG_SMART_SUSPEND | + DPM_FLAG_MAY_SKIP_RESUME); + pm_runtime_use_autosuspend(ddev->dev); + pm_runtime_set_autosuspend_delay(ddev->dev, 5000); + + pm_runtime_allow(ddev->dev); + + pm_runtime_mark_last_busy(ddev->dev); + pm_runtime_put_autosuspend(ddev->dev); + + /* + * For runpm implemented via BACO, PMFW will handle the + * timing for BACO in and out: + * - put ASIC into BACO state only when both video and + * audio functions are in D3 state. + * - pull ASIC out of BACO state when either video or + * audio function is in D0 state. + * Also, at startup, PMFW assumes both functions are in + * D0 state. + * + * So if snd driver was loaded prior to amdgpu driver + * and audio function was put into D3 state, there will + * be no PMFW-aware D-state transition(D0->D3) on runpm + * suspend. Thus the BACO will be not correctly kicked in. + * + * Via amdgpu_get_secondary_funcs(), the audio dev is put + * into D0 state. Then there will be a PMFW-aware D-state + * transition(D0->D3) on runpm suspend. + */ + if (amdgpu_device_supports_baco(ddev) && + !(adev->flags & AMD_IS_APU) && + (adev->asic_type >= CHIP_NAVI10)) + amdgpu_get_secondary_funcs(adev); + } + return 0; err_pci: @@ -2061,10 +2194,49 @@ static void amdgpu_pci_remove(struct pci_dev *pdev) { struct drm_device *dev = pci_get_drvdata(pdev); + struct amdgpu_device *adev = drm_to_adev(dev); + + if (adev->pm.rpm_mode != AMDGPU_RUNPM_NONE) { + pm_runtime_get_sync(dev->dev); + pm_runtime_forbid(dev->dev); + } + + if (adev->ip_versions[MP1_HWIP][0] == IP_VERSION(13, 0, 2) && + !amdgpu_sriov_vf(adev)) { + bool need_to_reset_gpu = false; + + if (adev->gmc.xgmi.num_physical_nodes > 1) { + struct amdgpu_hive_info *hive; + + hive = amdgpu_get_xgmi_hive(adev); + if (hive->device_remove_count == 0) + need_to_reset_gpu = true; + hive->device_remove_count++; + amdgpu_put_xgmi_hive(hive); + } else { + need_to_reset_gpu = true; + } + + /* Workaround for ASICs need to reset SMU. + * Called only when the first device is removed. + */ + if (need_to_reset_gpu) { + struct amdgpu_reset_context reset_context; + + adev->shutdown = true; + memset(&reset_context, 0, sizeof(reset_context)); + reset_context.method = AMD_RESET_METHOD_NONE; + reset_context.reset_req_dev = adev; + set_bit(AMDGPU_NEED_FULL_RESET, &reset_context.flags); + set_bit(AMDGPU_RESET_FOR_DEVICE_REMOVE, &reset_context.flags); + amdgpu_device_gpu_recover(adev, NULL, &reset_context); + } + } - drm_dev_unplug(dev); amdgpu_driver_unload_kms(dev); + drm_dev_unplug(dev); + /* * Flush any in flight DMA operations from device. * Clear the Bus Master Enable bit and then wait on the PCIe Device @@ -2170,13 +2342,20 @@ static void amdgpu_drv_delayed_reset_work_handler(struct work_struct *work) static int amdgpu_pmops_prepare(struct device *dev) { struct drm_device *drm_dev = dev_get_drvdata(dev); + struct amdgpu_device *adev = drm_to_adev(drm_dev); /* Return a positive number here so * DPM_FLAG_SMART_SUSPEND works properly */ if (amdgpu_device_supports_boco(drm_dev)) - return pm_runtime_suspended(dev) && - pm_suspend_via_firmware(); + return pm_runtime_suspended(dev); + + /* if we will not support s3 or s2i for the device + * then skip suspend + */ + if (!amdgpu_acpi_is_s0ix_active(adev) && + !amdgpu_acpi_is_s3_active(adev)) + return 1; return 0; } @@ -2190,18 +2369,23 @@ static int amdgpu_pmops_suspend(struct device *dev) { struct drm_device *drm_dev = dev_get_drvdata(dev); struct amdgpu_device *adev = drm_to_adev(drm_dev); - int r; if (amdgpu_acpi_is_s0ix_active(adev)) adev->in_s0ix = true; - adev->in_s3 = true; - r = amdgpu_device_suspend(drm_dev, true); - adev->in_s3 = false; - if (r) - return r; - if (!adev->in_s0ix) - r = amdgpu_asic_reset(adev); - return r; + else + adev->in_s3 = true; + return amdgpu_device_suspend(drm_dev, true); +} + +static int amdgpu_pmops_suspend_noirq(struct device *dev) +{ + struct drm_device *drm_dev = dev_get_drvdata(dev); + struct amdgpu_device *adev = drm_to_adev(drm_dev); + + if (amdgpu_acpi_should_gpu_reset(adev)) + return amdgpu_asic_reset(adev); + + return 0; } static int amdgpu_pmops_resume(struct device *dev) @@ -2217,6 +2401,8 @@ static int amdgpu_pmops_resume(struct device *dev) r = amdgpu_device_resume(drm_dev, true); if (amdgpu_acpi_is_s0ix_active(adev)) adev->in_s0ix = false; + else + adev->in_s3 = false; return r; } @@ -2255,6 +2441,71 @@ static int amdgpu_pmops_restore(struct device *dev) return amdgpu_device_resume(drm_dev, true); } +static int amdgpu_runtime_idle_check_display(struct device *dev) +{ + struct pci_dev *pdev = to_pci_dev(dev); + struct drm_device *drm_dev = pci_get_drvdata(pdev); + struct amdgpu_device *adev = drm_to_adev(drm_dev); + + if (adev->mode_info.num_crtc) { + struct drm_connector *list_connector; + struct drm_connector_list_iter iter; + int ret = 0; + + /* XXX: Return busy if any displays are connected to avoid + * possible display wakeups after runtime resume due to + * hotplug events in case any displays were connected while + * the GPU was in suspend. Remove this once that is fixed. + */ + mutex_lock(&drm_dev->mode_config.mutex); + drm_connector_list_iter_begin(drm_dev, &iter); + drm_for_each_connector_iter(list_connector, &iter) { + if (list_connector->status == connector_status_connected) { + ret = -EBUSY; + break; + } + } + drm_connector_list_iter_end(&iter); + mutex_unlock(&drm_dev->mode_config.mutex); + + if (ret) + return ret; + + if (amdgpu_device_has_dc_support(adev)) { + struct drm_crtc *crtc; + + drm_for_each_crtc(crtc, drm_dev) { + drm_modeset_lock(&crtc->mutex, NULL); + if (crtc->state->active) + ret = -EBUSY; + drm_modeset_unlock(&crtc->mutex); + if (ret < 0) + break; + } + } else { + mutex_lock(&drm_dev->mode_config.mutex); + drm_modeset_lock(&drm_dev->mode_config.connection_mutex, NULL); + + drm_connector_list_iter_begin(drm_dev, &iter); + drm_for_each_connector_iter(list_connector, &iter) { + if (list_connector->dpms == DRM_MODE_DPMS_ON) { + ret = -EBUSY; + break; + } + } + + drm_connector_list_iter_end(&iter); + + drm_modeset_unlock(&drm_dev->mode_config.connection_mutex); + mutex_unlock(&drm_dev->mode_config.mutex); + } + if (ret) + return ret; + } + + return 0; +} + static int amdgpu_pmops_runtime_suspend(struct device *dev) { struct pci_dev *pdev = to_pci_dev(dev); @@ -2262,11 +2513,15 @@ static int amdgpu_pmops_runtime_suspend(struct device *dev) struct amdgpu_device *adev = drm_to_adev(drm_dev); int ret, i; - if (!adev->runpm) { + if (adev->pm.rpm_mode == AMDGPU_RUNPM_NONE) { pm_runtime_forbid(dev); return -EBUSY; } + ret = amdgpu_runtime_idle_check_display(dev); + if (ret) + return ret; + /* wait for all rings to drain before suspending */ for (i = 0; i < AMDGPU_MAX_RINGS; i++) { struct amdgpu_ring *ring = adev->rings[i]; @@ -2327,7 +2582,7 @@ static int amdgpu_pmops_runtime_resume(struct device *dev) struct amdgpu_device *adev = drm_to_adev(drm_dev); int ret; - if (!adev->runpm) + if (adev->pm.rpm_mode == AMDGPU_RUNPM_NONE) return -EINVAL; /* Avoids registers access if device is physically gone */ @@ -2355,8 +2610,11 @@ static int amdgpu_pmops_runtime_resume(struct device *dev) amdgpu_device_baco_exit(drm_dev); } ret = amdgpu_device_resume(drm_dev, false); - if (ret) + if (ret) { + if (amdgpu_device_supports_px(drm_dev)) + pci_disable_device(pdev); return ret; + } if (amdgpu_device_supports_px(drm_dev)) drm_dev->switch_power_state = DRM_SWITCH_POWER_ON; @@ -2371,46 +2629,12 @@ static int amdgpu_pmops_runtime_idle(struct device *dev) /* we don't want the main rpm_idle to call suspend - we want to autosuspend */ int ret = 1; - if (!adev->runpm) { + if (adev->pm.rpm_mode == AMDGPU_RUNPM_NONE) { pm_runtime_forbid(dev); return -EBUSY; } - if (amdgpu_device_has_dc_support(adev)) { - struct drm_crtc *crtc; - - drm_for_each_crtc(crtc, drm_dev) { - drm_modeset_lock(&crtc->mutex, NULL); - if (crtc->state->active) - ret = -EBUSY; - drm_modeset_unlock(&crtc->mutex); - if (ret < 0) - break; - } - - } else { - struct drm_connector *list_connector; - struct drm_connector_list_iter iter; - - mutex_lock(&drm_dev->mode_config.mutex); - drm_modeset_lock(&drm_dev->mode_config.connection_mutex, NULL); - - drm_connector_list_iter_begin(drm_dev, &iter); - drm_for_each_connector_iter(list_connector, &iter) { - if (list_connector->dpms == DRM_MODE_DPMS_ON) { - ret = -EBUSY; - break; - } - } - - drm_connector_list_iter_end(&iter); - - drm_modeset_unlock(&drm_dev->mode_config.connection_mutex); - mutex_unlock(&drm_dev->mode_config.mutex); - } - - if (ret == -EBUSY) - DRM_DEBUG_DRIVER("failing to power off - crtc active\n"); + ret = amdgpu_runtime_idle_check_display(dev); pm_runtime_mark_last_busy(dev); pm_runtime_autosuspend(dev); @@ -2440,6 +2664,7 @@ static const struct dev_pm_ops amdgpu_pm_ops = { .prepare = amdgpu_pmops_prepare, .complete = amdgpu_pmops_complete, .suspend = amdgpu_pmops_suspend, + .suspend_noirq = amdgpu_pmops_suspend_noirq, .resume = amdgpu_pmops_resume, .freeze = amdgpu_pmops_freeze, .thaw = amdgpu_pmops_thaw, diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_fdinfo.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_fdinfo.c index 5a6857c44bb6..99a7855ab1bc 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_fdinfo.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_fdinfo.c @@ -32,6 +32,7 @@ #include <drm/amdgpu_drm.h> #include <drm/drm_debugfs.h> +#include <drm/drm_drv.h> #include "amdgpu.h" #include "amdgpu_vm.h" @@ -54,58 +55,49 @@ static const char *amdgpu_ip_name[AMDGPU_HW_IP_NUM] = { void amdgpu_show_fdinfo(struct seq_file *m, struct file *f) { - struct amdgpu_fpriv *fpriv; - uint32_t bus, dev, fn, i, domain; - uint64_t vram_mem = 0, gtt_mem = 0, cpu_mem = 0; struct drm_file *file = f->private_data; struct amdgpu_device *adev = drm_to_adev(file->minor->dev); - struct amdgpu_bo *root; + struct amdgpu_fpriv *fpriv = file->driver_priv; + struct amdgpu_vm *vm = &fpriv->vm; + + uint64_t vram_mem = 0, gtt_mem = 0, cpu_mem = 0; + ktime_t usage[AMDGPU_HW_IP_NUM]; + uint32_t bus, dev, fn, domain; + unsigned int hw_ip; int ret; - ret = amdgpu_file_to_fpriv(f, &fpriv); - if (ret) - return; bus = adev->pdev->bus->number; domain = pci_domain_nr(adev->pdev->bus); dev = PCI_SLOT(adev->pdev->devfn); fn = PCI_FUNC(adev->pdev->devfn); - root = amdgpu_bo_ref(fpriv->vm.root.bo); - if (!root) + ret = amdgpu_bo_reserve(vm->root.bo, false); + if (ret) return; - ret = amdgpu_bo_reserve(root, false); - if (ret) { - DRM_ERROR("Fail to reserve bo\n"); - return; - } - amdgpu_vm_get_memory(&fpriv->vm, &vram_mem, >t_mem, &cpu_mem); - amdgpu_bo_unreserve(root); - amdgpu_bo_unref(&root); + amdgpu_vm_get_memory(vm, &vram_mem, >t_mem, &cpu_mem); + amdgpu_bo_unreserve(vm->root.bo); - seq_printf(m, "pdev:\t%04x:%02x:%02x.%d\npasid:\t%u\n", domain, bus, - dev, fn, fpriv->vm.pasid); - seq_printf(m, "vram mem:\t%llu kB\n", vram_mem/1024UL); - seq_printf(m, "gtt mem:\t%llu kB\n", gtt_mem/1024UL); - seq_printf(m, "cpu mem:\t%llu kB\n", cpu_mem/1024UL); - for (i = 0; i < AMDGPU_HW_IP_NUM; i++) { - uint32_t count = amdgpu_ctx_num_entities[i]; - int idx = 0; - uint64_t total = 0, min = 0; - uint32_t perc, frac; + amdgpu_ctx_mgr_usage(&fpriv->ctx_mgr, usage); - for (idx = 0; idx < count; idx++) { - total = amdgpu_ctx_mgr_fence_usage(&fpriv->ctx_mgr, - i, idx, &min); - if ((total == 0) || (min == 0)) - continue; + /* + * ****************************************************************** + * For text output format description please see drm-usage-stats.rst! + * ****************************************************************** + */ - perc = div64_u64(10000 * total, min); - frac = perc % 100; + seq_printf(m, "pasid:\t%u\n", fpriv->vm.pasid); + seq_printf(m, "drm-driver:\t%s\n", file->minor->dev->driver->name); + seq_printf(m, "drm-pdev:\t%04x:%02x:%02x.%d\n", domain, bus, dev, fn); + seq_printf(m, "drm-client-id:\t%Lu\n", vm->immediate.fence_context); + seq_printf(m, "drm-memory-vram:\t%llu KiB\n", vram_mem/1024UL); + seq_printf(m, "drm-memory-gtt: \t%llu KiB\n", gtt_mem/1024UL); + seq_printf(m, "drm-memory-cpu: \t%llu KiB\n", cpu_mem/1024UL); + for (hw_ip = 0; hw_ip < AMDGPU_HW_IP_NUM; ++hw_ip) { + if (!usage[hw_ip]) + continue; - seq_printf(m, "%s%d:\t%d.%d%%\n", - amdgpu_ip_name[i], - idx, perc/100, frac); - } + seq_printf(m, "drm-engine-%s:\t%Ld ns\n", amdgpu_ip_name[hw_ip], + ktime_to_ns(usage[hw_ip])); } } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_fence.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_fence.c index 45977a72b5dd..d0d99ed607dd 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_fence.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_fence.c @@ -39,6 +39,7 @@ #include <drm/drm_drv.h> #include "amdgpu.h" #include "amdgpu_trace.h" +#include "amdgpu_reset.h" /* * Fences @@ -46,7 +47,7 @@ * for GPU/CPU synchronization. When the fence is written, * it is expected that all buffers associated with that fence * are no longer in use by the associated ring on the GPU and - * that the the relevant GPU caches have been flushed. + * that the relevant GPU caches have been flushed. */ struct amdgpu_fence { @@ -163,11 +164,16 @@ int amdgpu_fence_emit(struct amdgpu_ring *ring, struct dma_fence **f, struct amd if (job && job->job_run_counter) { /* reinit seq for resubmitted jobs */ fence->seqno = seq; + /* TO be inline with external fence creation and other drivers */ + dma_fence_get(fence); } else { - if (job) + if (job) { dma_fence_init(fence, &amdgpu_job_fence_ops, &ring->fence_drv.lock, adev->fence_context + ring->idx, seq); + /* Against remove in amdgpu_job_{free, free_cb} */ + dma_fence_get(fence); + } else dma_fence_init(fence, &amdgpu_fence_ops, &ring->fence_drv.lock, @@ -394,7 +400,6 @@ unsigned amdgpu_fence_count_emitted(struct amdgpu_ring *ring) /* We are not protected by ring lock when reading the last sequence * but it's ok to report slightly wrong fence count here. */ - amdgpu_fence_process(ring); emitted = 0x100000000ull; emitted -= atomic_read(&ring->fence_drv.last_seq); emitted += READ_ONCE(ring->fence_drv.sync_seq); @@ -422,8 +427,8 @@ int amdgpu_fence_driver_start_ring(struct amdgpu_ring *ring, uint64_t index; if (ring->funcs->type != AMDGPU_RING_TYPE_UVD) { - ring->fence_drv.cpu_addr = &adev->wb.wb[ring->fence_offs]; - ring->fence_drv.gpu_addr = adev->wb.gpu_addr + (ring->fence_offs * 4); + ring->fence_drv.cpu_addr = ring->fence_cpu_addr; + ring->fence_drv.gpu_addr = ring->fence_gpu_addr; } else { /* put fence directly behind firmware */ index = ALIGN(adev->uvd.fw->size, 8); @@ -446,24 +451,18 @@ int amdgpu_fence_driver_start_ring(struct amdgpu_ring *ring, * for the requested ring. * * @ring: ring to init the fence driver on - * @num_hw_submission: number of entries on the hardware queue - * @sched_score: optional score atomic shared with other schedulers * * Init the fence driver for the requested ring (all asics). * Helper function for amdgpu_fence_driver_init(). */ -int amdgpu_fence_driver_init_ring(struct amdgpu_ring *ring, - unsigned num_hw_submission, - atomic_t *sched_score) +int amdgpu_fence_driver_init_ring(struct amdgpu_ring *ring) { struct amdgpu_device *adev = ring->adev; - long timeout; - int r; if (!adev) return -EINVAL; - if (!is_power_of_2(num_hw_submission)) + if (!is_power_of_2(ring->num_hw_submission)) return -EINVAL; ring->fence_drv.cpu_addr = NULL; @@ -474,41 +473,14 @@ int amdgpu_fence_driver_init_ring(struct amdgpu_ring *ring, timer_setup(&ring->fence_drv.fallback_timer, amdgpu_fence_fallback, 0); - ring->fence_drv.num_fences_mask = num_hw_submission * 2 - 1; + ring->fence_drv.num_fences_mask = ring->num_hw_submission * 2 - 1; spin_lock_init(&ring->fence_drv.lock); - ring->fence_drv.fences = kcalloc(num_hw_submission * 2, sizeof(void *), + ring->fence_drv.fences = kcalloc(ring->num_hw_submission * 2, sizeof(void *), GFP_KERNEL); + if (!ring->fence_drv.fences) return -ENOMEM; - /* No need to setup the GPU scheduler for rings that don't need it */ - if (ring->no_scheduler) - return 0; - - switch (ring->funcs->type) { - case AMDGPU_RING_TYPE_GFX: - timeout = adev->gfx_timeout; - break; - case AMDGPU_RING_TYPE_COMPUTE: - timeout = adev->compute_timeout; - break; - case AMDGPU_RING_TYPE_SDMA: - timeout = adev->sdma_timeout; - break; - default: - timeout = adev->video_timeout; - break; - } - - r = drm_sched_init(&ring->sched, &amdgpu_sched_ops, - num_hw_submission, amdgpu_job_hang_limit, - timeout, NULL, sched_score, ring->name); - if (r) { - DRM_ERROR("Failed to create scheduler on ring %s.\n", - ring->name); - return r; - } - return 0; } @@ -564,6 +536,24 @@ void amdgpu_fence_driver_hw_fini(struct amdgpu_device *adev) } } +/* Will either stop and flush handlers for amdgpu interrupt or reanble it */ +void amdgpu_fence_driver_isr_toggle(struct amdgpu_device *adev, bool stop) +{ + int i; + + for (i = 0; i < AMDGPU_MAX_RINGS; i++) { + struct amdgpu_ring *ring = adev->rings[i]; + + if (!ring || !ring->fence_drv.initialized || !ring->fence_drv.irq_src) + continue; + + if (stop) + disable_irq(adev->irq.irq); + else + enable_irq(adev->irq.irq); + } +} + void amdgpu_fence_driver_sw_fini(struct amdgpu_device *adev) { unsigned int i, j; @@ -627,8 +617,10 @@ void amdgpu_fence_driver_clear_job_fences(struct amdgpu_ring *ring) for (i = 0; i <= ring->fence_drv.num_fences_mask; i++) { ptr = &ring->fence_drv.fences[i]; old = rcu_dereference_protected(*ptr, 1); - if (old && old->ops == &amdgpu_job_fence_ops) + if (old && old->ops == &amdgpu_job_fence_ops) { RCU_INIT_POINTER(*ptr, NULL); + dma_fence_put(old); + } } } @@ -831,7 +823,10 @@ static int gpu_recover_get(void *data, u64 *val) return 0; } - *val = amdgpu_device_gpu_recover(adev, NULL); + if (amdgpu_reset_domain_schedule(adev->reset_domain, &adev->reset_work)) + flush_work(&adev->reset_work); + + *val = atomic_read(&adev->reset_domain->reset_res); pm_runtime_mark_last_busy(dev->dev); pm_runtime_put_autosuspend(dev->dev); @@ -843,6 +838,21 @@ DEFINE_SHOW_ATTRIBUTE(amdgpu_debugfs_fence_info); DEFINE_DEBUGFS_ATTRIBUTE(amdgpu_debugfs_gpu_recover_fops, gpu_recover_get, NULL, "%lld\n"); +static void amdgpu_debugfs_reset_work(struct work_struct *work) +{ + struct amdgpu_device *adev = container_of(work, struct amdgpu_device, + reset_work); + + struct amdgpu_reset_context reset_context; + memset(&reset_context, 0, sizeof(reset_context)); + + reset_context.method = AMD_RESET_METHOD_NONE; + reset_context.reset_req_dev = adev; + set_bit(AMDGPU_NEED_FULL_RESET, &reset_context.flags); + + amdgpu_device_gpu_recover(adev, NULL, &reset_context); +} + #endif void amdgpu_debugfs_fence_init(struct amdgpu_device *adev) @@ -854,9 +864,12 @@ void amdgpu_debugfs_fence_init(struct amdgpu_device *adev) debugfs_create_file("amdgpu_fence_info", 0444, root, adev, &amdgpu_debugfs_fence_info_fops); - if (!amdgpu_sriov_vf(adev)) + if (!amdgpu_sriov_vf(adev)) { + + INIT_WORK(&adev->reset_work, amdgpu_debugfs_reset_work); debugfs_create_file("amdgpu_gpu_recover", 0444, root, adev, &amdgpu_debugfs_gpu_recover_fops); + } #endif } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_fru_eeprom.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_fru_eeprom.c index 2a786e788627..e325150879df 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_fru_eeprom.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_fru_eeprom.c @@ -30,7 +30,6 @@ #include "amdgpu_eeprom.h" #define FRU_EEPROM_MADDR 0x60000 -#define I2C_PRODUCT_INFO_OFFSET 0xC0 static bool is_fru_eeprom_supported(struct amdgpu_device *adev) { @@ -40,7 +39,13 @@ static bool is_fru_eeprom_supported(struct amdgpu_device *adev) */ struct atom_context *atom_ctx = adev->mode_info.atom_context; - /* VBIOS is of the format ###-DXXXYY-##. For SKU identification, + /* The i2c access is blocked on VF + * TODO: Need other way to get the info + */ + if (amdgpu_sriov_vf(adev)) + return false; + + /* VBIOS is of the format ###-DXXXYYYY-##. For SKU identification, * we can use just the "DXXX" portion. If there were more models, we * could convert the 3 characters to a hex integer and use a switch * for ease/speed/readability. For now, 2 string comparisons are @@ -59,17 +64,29 @@ static bool is_fru_eeprom_supported(struct amdgpu_device *adev) case CHIP_ALDEBARAN: /* All Aldebaran SKUs have the FRU */ return true; + case CHIP_SIENNA_CICHLID: + if (strnstr(atom_ctx->vbios_version, "D603", + sizeof(atom_ctx->vbios_version))) { + if (strnstr(atom_ctx->vbios_version, "D603GLXE", + sizeof(atom_ctx->vbios_version))) + return false; + else + return true; + } else { + return false; + } default: return false; } } static int amdgpu_fru_read_eeprom(struct amdgpu_device *adev, uint32_t addrptr, - unsigned char *buff) + unsigned char *buf, size_t buf_size) { - int ret, size; + int ret; + u8 size; - ret = amdgpu_eeprom_read(&adev->pm.smu_i2c, addrptr, buff, 1); + ret = amdgpu_eeprom_read(adev->pm.fru_eeprom_i2c_bus, addrptr, buf, 1); if (ret < 1) { DRM_WARN("FRU: Failed to get size field"); return ret; @@ -78,9 +95,11 @@ static int amdgpu_fru_read_eeprom(struct amdgpu_device *adev, uint32_t addrptr, /* The size returned by the i2c requires subtraction of 0xC0 since the * size apparently always reports as 0xC0+actual size. */ - size = buff[0] - I2C_PRODUCT_INFO_OFFSET; + size = buf[0] & 0x3F; + size = min_t(size_t, size, buf_size); - ret = amdgpu_eeprom_read(&adev->pm.smu_i2c, addrptr + 1, buff, size); + ret = amdgpu_eeprom_read(adev->pm.fru_eeprom_i2c_bus, addrptr + 1, + buf, size); if (ret < 1) { DRM_WARN("FRU: Failed to get data field"); return ret; @@ -91,19 +110,15 @@ static int amdgpu_fru_read_eeprom(struct amdgpu_device *adev, uint32_t addrptr, int amdgpu_fru_get_product_info(struct amdgpu_device *adev) { - unsigned char buff[AMDGPU_PRODUCT_NAME_LEN+2]; + unsigned char buf[AMDGPU_PRODUCT_NAME_LEN]; u32 addrptr; int size, len; - int offset = 2; if (!is_fru_eeprom_supported(adev)) return 0; - if (adev->asic_type == CHIP_ALDEBARAN) - offset = 0; - /* If algo exists, it means that the i2c_adapter's initialized */ - if (!adev->pm.smu_i2c.algo) { + if (!adev->pm.fru_eeprom_i2c_bus || !adev->pm.fru_eeprom_i2c_bus->algo) { DRM_WARN("Cannot access FRU, EEPROM accessor not initialized"); return -ENODEV; } @@ -121,7 +136,7 @@ int amdgpu_fru_get_product_info(struct amdgpu_device *adev) * and the language field, so just start from 0xb, manufacturer size */ addrptr = FRU_EEPROM_MADDR + 0xb; - size = amdgpu_fru_read_eeprom(adev, addrptr, buff); + size = amdgpu_fru_read_eeprom(adev, addrptr, buf, sizeof(buf)); if (size < 1) { DRM_ERROR("Failed to read FRU Manufacturer, ret:%d", size); return -EINVAL; @@ -131,7 +146,7 @@ int amdgpu_fru_get_product_info(struct amdgpu_device *adev) * size field being 1 byte. This pattern continues below. */ addrptr += size + 1; - size = amdgpu_fru_read_eeprom(adev, addrptr, buff); + size = amdgpu_fru_read_eeprom(adev, addrptr, buf, sizeof(buf)); if (size < 1) { DRM_ERROR("Failed to read FRU product name, ret:%d", size); return -EINVAL; @@ -143,12 +158,11 @@ int amdgpu_fru_get_product_info(struct amdgpu_device *adev) AMDGPU_PRODUCT_NAME_LEN); len = AMDGPU_PRODUCT_NAME_LEN - 1; } - /* Start at 2 due to buff using fields 0 and 1 for the address */ - memcpy(adev->product_name, &buff[offset], len); + memcpy(adev->product_name, buf, len); adev->product_name[len] = '\0'; addrptr += size + 1; - size = amdgpu_fru_read_eeprom(adev, addrptr, buff); + size = amdgpu_fru_read_eeprom(adev, addrptr, buf, sizeof(buf)); if (size < 1) { DRM_ERROR("Failed to read FRU product number, ret:%d", size); return -EINVAL; @@ -162,11 +176,11 @@ int amdgpu_fru_get_product_info(struct amdgpu_device *adev) DRM_WARN("FRU Product Number is larger than 16 characters. This is likely a mistake"); len = sizeof(adev->product_number) - 1; } - memcpy(adev->product_number, &buff[offset], len); + memcpy(adev->product_number, buf, len); adev->product_number[len] = '\0'; addrptr += size + 1; - size = amdgpu_fru_read_eeprom(adev, addrptr, buff); + size = amdgpu_fru_read_eeprom(adev, addrptr, buf, sizeof(buf)); if (size < 1) { DRM_ERROR("Failed to read FRU product version, ret:%d", size); @@ -174,7 +188,7 @@ int amdgpu_fru_get_product_info(struct amdgpu_device *adev) } addrptr += size + 1; - size = amdgpu_fru_read_eeprom(adev, addrptr, buff); + size = amdgpu_fru_read_eeprom(adev, addrptr, buf, sizeof(buf)); if (size < 1) { DRM_ERROR("Failed to read FRU serial number, ret:%d", size); @@ -189,7 +203,7 @@ int amdgpu_fru_get_product_info(struct amdgpu_device *adev) DRM_WARN("FRU Serial Number is larger than 16 characters. This is likely a mistake"); len = sizeof(adev->serial) - 1; } - memcpy(adev->serial, &buff[offset], len); + memcpy(adev->serial, buf, len); adev->serial[len] = '\0'; return 0; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gart.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gart.c index d3e4203f6217..01cb89ffbd56 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gart.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gart.c @@ -114,80 +114,12 @@ void amdgpu_gart_dummy_page_fini(struct amdgpu_device *adev) */ int amdgpu_gart_table_vram_alloc(struct amdgpu_device *adev) { - int r; - - if (adev->gart.bo == NULL) { - struct amdgpu_bo_param bp; - - memset(&bp, 0, sizeof(bp)); - bp.size = adev->gart.table_size; - bp.byte_align = PAGE_SIZE; - bp.domain = AMDGPU_GEM_DOMAIN_VRAM; - bp.flags = AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED | - AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS; - bp.type = ttm_bo_type_kernel; - bp.resv = NULL; - bp.bo_ptr_size = sizeof(struct amdgpu_bo); - - r = amdgpu_bo_create(adev, &bp, &adev->gart.bo); - if (r) { - return r; - } - } - return 0; -} - -/** - * amdgpu_gart_table_vram_pin - pin gart page table in vram - * - * @adev: amdgpu_device pointer - * - * Pin the GART page table in vram so it will not be moved - * by the memory manager (pcie r4xx, r5xx+). These asics require the - * gart table to be in video memory. - * Returns 0 for success, error for failure. - */ -int amdgpu_gart_table_vram_pin(struct amdgpu_device *adev) -{ - int r; - - r = amdgpu_bo_reserve(adev->gart.bo, false); - if (unlikely(r != 0)) - return r; - r = amdgpu_bo_pin(adev->gart.bo, AMDGPU_GEM_DOMAIN_VRAM); - if (r) { - amdgpu_bo_unreserve(adev->gart.bo); - return r; - } - r = amdgpu_bo_kmap(adev->gart.bo, &adev->gart.ptr); - if (r) - amdgpu_bo_unpin(adev->gart.bo); - amdgpu_bo_unreserve(adev->gart.bo); - return r; -} - -/** - * amdgpu_gart_table_vram_unpin - unpin gart page table in vram - * - * @adev: amdgpu_device pointer - * - * Unpin the GART page table in vram (pcie r4xx, r5xx+). - * These asics require the gart table to be in video memory. - */ -void amdgpu_gart_table_vram_unpin(struct amdgpu_device *adev) -{ - int r; + if (adev->gart.bo != NULL) + return 0; - if (adev->gart.bo == NULL) { - return; - } - r = amdgpu_bo_reserve(adev->gart.bo, true); - if (likely(r == 0)) { - amdgpu_bo_kunmap(adev->gart.bo); - amdgpu_bo_unpin(adev->gart.bo); - amdgpu_bo_unreserve(adev->gart.bo); - adev->gart.ptr = NULL; - } + return amdgpu_bo_create_kernel(adev, adev->gart.table_size, PAGE_SIZE, + AMDGPU_GEM_DOMAIN_VRAM, &adev->gart.bo, + NULL, (void *)&adev->gart.ptr); } /** @@ -201,11 +133,7 @@ void amdgpu_gart_table_vram_unpin(struct amdgpu_device *adev) */ void amdgpu_gart_table_vram_free(struct amdgpu_device *adev) { - if (adev->gart.bo == NULL) { - return; - } - amdgpu_bo_unref(&adev->gart.bo); - adev->gart.ptr = NULL; + amdgpu_bo_free_kernel(&adev->gart.bo, NULL, (void *)&adev->gart.ptr); } /* @@ -222,7 +150,7 @@ void amdgpu_gart_table_vram_free(struct amdgpu_device *adev) * replaces them with the dummy page (all asics). * Returns 0 for success, -EINVAL for failure. */ -int amdgpu_gart_unbind(struct amdgpu_device *adev, uint64_t offset, +void amdgpu_gart_unbind(struct amdgpu_device *adev, uint64_t offset, int pages) { unsigned t; @@ -233,13 +161,11 @@ int amdgpu_gart_unbind(struct amdgpu_device *adev, uint64_t offset, uint64_t flags = 0; int idx; - if (!adev->gart.ready) { - WARN(1, "trying to unbind memory from uninitialized GART !\n"); - return -EINVAL; - } + if (!adev->gart.ptr) + return; if (!drm_dev_enter(adev_to_drm(adev), &idx)) - return 0; + return; t = offset / AMDGPU_GPU_PAGE_SIZE; p = t / AMDGPU_GPU_PAGES_IN_CPU_PAGE; @@ -260,7 +186,6 @@ int amdgpu_gart_unbind(struct amdgpu_device *adev, uint64_t offset, amdgpu_gmc_flush_gpu_tlb(adev, 0, i, 0); drm_dev_exit(idx); - return 0; } /** @@ -276,7 +201,7 @@ int amdgpu_gart_unbind(struct amdgpu_device *adev, uint64_t offset, * Map the dma_addresses into GART entries (all asics). * Returns 0 for success, -EINVAL for failure. */ -int amdgpu_gart_map(struct amdgpu_device *adev, uint64_t offset, +void amdgpu_gart_map(struct amdgpu_device *adev, uint64_t offset, int pages, dma_addr_t *dma_addr, uint64_t flags, void *dst) { @@ -284,13 +209,8 @@ int amdgpu_gart_map(struct amdgpu_device *adev, uint64_t offset, unsigned i, j, t; int idx; - if (!adev->gart.ready) { - WARN(1, "trying to bind memory to uninitialized GART !\n"); - return -EINVAL; - } - if (!drm_dev_enter(adev_to_drm(adev), &idx)) - return 0; + return; t = offset / AMDGPU_GPU_PAGE_SIZE; @@ -302,7 +222,6 @@ int amdgpu_gart_map(struct amdgpu_device *adev, uint64_t offset, } } drm_dev_exit(idx); - return 0; } /** @@ -318,20 +237,14 @@ int amdgpu_gart_map(struct amdgpu_device *adev, uint64_t offset, * (all asics). * Returns 0 for success, -EINVAL for failure. */ -int amdgpu_gart_bind(struct amdgpu_device *adev, uint64_t offset, +void amdgpu_gart_bind(struct amdgpu_device *adev, uint64_t offset, int pages, dma_addr_t *dma_addr, uint64_t flags) { - if (!adev->gart.ready) { - WARN(1, "trying to bind memory to uninitialized GART !\n"); - return -EINVAL; - } - if (!adev->gart.ptr) - return 0; + return; - return amdgpu_gart_map(adev, offset, pages, dma_addr, flags, - adev->gart.ptr); + amdgpu_gart_map(adev, offset, pages, dma_addr, flags, adev->gart.ptr); } /** @@ -346,6 +259,9 @@ void amdgpu_gart_invalidate_tlb(struct amdgpu_device *adev) { int i; + if (!adev->gart.ptr) + return; + mb(); amdgpu_device_flush_hdp(adev, NULL); for (i = 0; i < adev->num_vmhubs; i++) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gart.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_gart.h index 78895413cf9f..8fea3e04e411 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gart.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gart.h @@ -46,7 +46,6 @@ struct amdgpu_gart { unsigned num_gpu_pages; unsigned num_cpu_pages; unsigned table_size; - bool ready; /* Asic default pte flags */ uint64_t gart_pte_flags; @@ -58,12 +57,12 @@ int amdgpu_gart_table_vram_pin(struct amdgpu_device *adev); void amdgpu_gart_table_vram_unpin(struct amdgpu_device *adev); int amdgpu_gart_init(struct amdgpu_device *adev); void amdgpu_gart_dummy_page_fini(struct amdgpu_device *adev); -int amdgpu_gart_unbind(struct amdgpu_device *adev, uint64_t offset, - int pages); -int amdgpu_gart_map(struct amdgpu_device *adev, uint64_t offset, - int pages, dma_addr_t *dma_addr, uint64_t flags, - void *dst); -int amdgpu_gart_bind(struct amdgpu_device *adev, uint64_t offset, - int pages, dma_addr_t *dma_addr, uint64_t flags); +void amdgpu_gart_unbind(struct amdgpu_device *adev, uint64_t offset, + int pages); +void amdgpu_gart_map(struct amdgpu_device *adev, uint64_t offset, + int pages, dma_addr_t *dma_addr, uint64_t flags, + void *dst); +void amdgpu_gart_bind(struct amdgpu_device *adev, uint64_t offset, + int pages, dma_addr_t *dma_addr, uint64_t flags); void amdgpu_gart_invalidate_tlb(struct amdgpu_device *adev); #endif diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c index 9a6507af1670..8ef31d687ef3 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c @@ -222,16 +222,10 @@ static void amdgpu_gem_object_close(struct drm_gem_object *obj, if (!bo_va || --bo_va->ref_count) goto out_unlock; - amdgpu_vm_bo_rmv(adev, bo_va); + amdgpu_vm_bo_del(adev, bo_va); if (!amdgpu_vm_ready(vm)) goto out_unlock; - fence = dma_resv_excl_fence(bo->tbo.base.resv); - if (fence) { - amdgpu_bo_fence(bo, fence, true); - fence = NULL; - } - r = amdgpu_vm_clear_freed(adev, vm, &fence); if (r || !fence) goto out_unlock; @@ -264,9 +258,6 @@ static int amdgpu_gem_object_mmap(struct drm_gem_object *obj, struct vm_area_str !(vma->vm_flags & (VM_READ | VM_WRITE | VM_EXEC))) vma->vm_flags &= ~VM_MAYWRITE; - if (bo->kfd_bo) - vma->vm_flags |= VM_DONTCOPY; - return drm_gem_ttm_mmap(obj, vma); } @@ -305,8 +296,8 @@ int amdgpu_gem_create_ioctl(struct drm_device *dev, void *data, AMDGPU_GEM_CREATE_VRAM_CLEARED | AMDGPU_GEM_CREATE_VM_ALWAYS_VALID | AMDGPU_GEM_CREATE_EXPLICIT_SYNC | - AMDGPU_GEM_CREATE_ENCRYPTED)) - + AMDGPU_GEM_CREATE_ENCRYPTED | + AMDGPU_GEM_CREATE_DISCARDABLE)) return -EINVAL; /* reject invalid gem domains */ @@ -535,7 +526,8 @@ int amdgpu_gem_wait_idle_ioctl(struct drm_device *dev, void *data, return -ENOENT; } robj = gem_to_amdgpu_bo(gobj); - ret = dma_resv_wait_timeout(robj->tbo.base.resv, true, true, timeout); + ret = dma_resv_wait_timeout(robj->tbo.base.resv, DMA_RESV_USAGE_READ, + true, timeout); /* ret == 0 means not signaled, * ret > 0 means signaled @@ -621,7 +613,7 @@ static void amdgpu_gem_va_update_vm(struct amdgpu_device *adev, if (operation == AMDGPU_VA_OP_MAP || operation == AMDGPU_VA_OP_REPLACE) { - r = amdgpu_vm_bo_update(adev, bo_va, false, NULL); + r = amdgpu_vm_bo_update(adev, bo_va, false); if (r) goto error; } @@ -653,6 +645,8 @@ uint64_t amdgpu_gem_va_map_flags(struct amdgpu_device *adev, uint32_t flags) pte_flag |= AMDGPU_PTE_WRITEABLE; if (flags & AMDGPU_VM_PAGE_PRT) pte_flag |= AMDGPU_PTE_PRT; + if (flags & AMDGPU_VM_PAGE_NOALLOC) + pte_flag |= AMDGPU_PTE_NOALLOC; if (adev->gmc.gmc_funcs->map_mtype) pte_flag |= amdgpu_gmc_map_mtype(adev, @@ -666,7 +660,8 @@ int amdgpu_gem_va_ioctl(struct drm_device *dev, void *data, { const uint32_t valid_flags = AMDGPU_VM_DELAY_UPDATE | AMDGPU_VM_PAGE_READABLE | AMDGPU_VM_PAGE_WRITEABLE | - AMDGPU_VM_PAGE_EXECUTABLE | AMDGPU_VM_MTYPE_MASK; + AMDGPU_VM_PAGE_EXECUTABLE | AMDGPU_VM_MTYPE_MASK | + AMDGPU_VM_PAGE_NOALLOC; const uint32_t prt_flags = AMDGPU_VM_DELAY_UPDATE | AMDGPU_VM_PAGE_PRT; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c index 1916ec84dd71..9546adc8a76f 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c @@ -23,6 +23,7 @@ * */ +#include <linux/firmware.h> #include "amdgpu.h" #include "amdgpu_gfx.h" #include "amdgpu_rlc.h" @@ -99,42 +100,6 @@ bool amdgpu_gfx_is_me_queue_enabled(struct amdgpu_device *adev, } /** - * amdgpu_gfx_scratch_get - Allocate a scratch register - * - * @adev: amdgpu_device pointer - * @reg: scratch register mmio offset - * - * Allocate a CP scratch register for use by the driver (all asics). - * Returns 0 on success or -EINVAL on failure. - */ -int amdgpu_gfx_scratch_get(struct amdgpu_device *adev, uint32_t *reg) -{ - int i; - - i = ffs(adev->gfx.scratch.free_mask); - if (i != 0 && i <= adev->gfx.scratch.num_reg) { - i--; - adev->gfx.scratch.free_mask &= ~(1u << i); - *reg = adev->gfx.scratch.reg_base + i; - return 0; - } - return -EINVAL; -} - -/** - * amdgpu_gfx_scratch_free - Free a scratch register - * - * @adev: amdgpu_device pointer - * @reg: scratch register mmio offset - * - * Free a CP scratch register allocated for use by the driver (all asics) - */ -void amdgpu_gfx_scratch_free(struct amdgpu_device *adev, uint32_t reg) -{ - adev->gfx.scratch.free_mask |= 1u << (reg - adev->gfx.scratch.reg_base); -} - -/** * amdgpu_gfx_parse_disable_cu - Parse the disable_cu module parameter * * @mask: array in which the per-shader array disable masks will be stored @@ -178,7 +143,12 @@ void amdgpu_gfx_parse_disable_cu(unsigned *mask, unsigned max_se, unsigned max_s } } -static bool amdgpu_gfx_is_multipipe_capable(struct amdgpu_device *adev) +static bool amdgpu_gfx_is_graphics_multipipe_capable(struct amdgpu_device *adev) +{ + return amdgpu_async_gfx_ring && adev->gfx.me.num_pipe_per_me > 1; +} + +static bool amdgpu_gfx_is_compute_multipipe_capable(struct amdgpu_device *adev) { if (amdgpu_compute_multipipe != -1) { DRM_INFO("amdgpu: forcing compute pipe policy %d\n", @@ -194,6 +164,28 @@ static bool amdgpu_gfx_is_multipipe_capable(struct amdgpu_device *adev) return adev->gfx.mec.num_mec > 1; } +bool amdgpu_gfx_is_high_priority_graphics_queue(struct amdgpu_device *adev, + struct amdgpu_ring *ring) +{ + int queue = ring->queue; + int pipe = ring->pipe; + + /* Policy: use pipe1 queue0 as high priority graphics queue if we + * have more than one gfx pipe. + */ + if (amdgpu_gfx_is_graphics_multipipe_capable(adev) && + adev->gfx.num_gfx_rings > 1 && pipe == 1 && queue == 0) { + int me = ring->me; + int bit; + + bit = amdgpu_gfx_me_queue_to_bit(adev, me, pipe, queue); + if (ring == &adev->gfx.gfx_ring[bit]) + return true; + } + + return false; +} + bool amdgpu_gfx_is_high_priority_compute_queue(struct amdgpu_device *adev, struct amdgpu_ring *ring) { @@ -210,7 +202,7 @@ bool amdgpu_gfx_is_high_priority_compute_queue(struct amdgpu_device *adev, void amdgpu_gfx_compute_queue_acquire(struct amdgpu_device *adev) { int i, queue, pipe; - bool multipipe_policy = amdgpu_gfx_is_multipipe_capable(adev); + bool multipipe_policy = amdgpu_gfx_is_compute_multipipe_capable(adev); int max_queues_per_mec = min(adev->gfx.mec.num_pipe_per_mec * adev->gfx.mec.num_queue_per_pipe, adev->gfx.num_compute_rings); @@ -236,18 +228,24 @@ void amdgpu_gfx_compute_queue_acquire(struct amdgpu_device *adev) void amdgpu_gfx_graphics_queue_acquire(struct amdgpu_device *adev) { - int i, queue, me; - - for (i = 0; i < AMDGPU_MAX_GFX_QUEUES; ++i) { - queue = i % adev->gfx.me.num_queue_per_pipe; - me = (i / adev->gfx.me.num_queue_per_pipe) - / adev->gfx.me.num_pipe_per_me; + int i, queue, pipe; + bool multipipe_policy = amdgpu_gfx_is_graphics_multipipe_capable(adev); + int max_queues_per_me = adev->gfx.me.num_pipe_per_me * + adev->gfx.me.num_queue_per_pipe; - if (me >= adev->gfx.me.num_me) - break; + if (multipipe_policy) { /* policy: amdgpu owns the first queue per pipe at this stage * will extend to mulitple queues per pipe later */ - if (me == 0 && queue < 1) + for (i = 0; i < max_queues_per_me; i++) { + pipe = i % adev->gfx.me.num_pipe_per_me; + queue = (i / adev->gfx.me.num_pipe_per_me) % + adev->gfx.me.num_queue_per_pipe; + + set_bit(pipe * adev->gfx.me.num_queue_per_pipe + queue, + adev->gfx.me.queue_bitmap); + } + } else { + for (i = 0; i < max_queues_per_me; ++i) set_bit(i, adev->gfx.me.queue_bitmap); } @@ -266,7 +264,7 @@ static int amdgpu_gfx_kiq_acquire(struct amdgpu_device *adev, * adev->gfx.mec.num_pipe_per_mec * adev->gfx.mec.num_queue_per_pipe; - while (queue_bit-- >= 0) { + while (--queue_bit >= 0) { if (test_bit(queue_bit, adev->gfx.mec.queue_bitmap)) continue; @@ -367,7 +365,7 @@ int amdgpu_gfx_mqd_sw_init(struct amdgpu_device *adev, /* create MQD for KIQ */ ring = &adev->gfx.kiq.ring; - if (!ring->mqd_obj) { + if (!adev->enable_mes_kiq && !ring->mqd_obj) { /* originaly the KIQ MQD is put in GTT domain, but for SRIOV VRAM domain is a must * otherwise hypervisor trigger SAVE_VF fail after driver unloaded which mean MQD * deallocated and gart_unbind, to strict diverage we decide to use VRAM domain for @@ -464,7 +462,7 @@ int amdgpu_gfx_disable_kcq(struct amdgpu_device *adev) { struct amdgpu_kiq *kiq = &adev->gfx.kiq; struct amdgpu_ring *kiq_ring = &kiq->ring; - int i, r; + int i, r = 0; if (!kiq->pmf || !kiq->pmf->kiq_unmap_queues) return -EINVAL; @@ -479,7 +477,9 @@ int amdgpu_gfx_disable_kcq(struct amdgpu_device *adev) for (i = 0; i < adev->gfx.num_compute_rings; i++) kiq->pmf->kiq_unmap_queues(kiq_ring, &adev->gfx.compute_ring[i], RESET_QUEUES, 0, 0); - r = amdgpu_ring_test_helper(kiq_ring); + + if (adev->gfx.kiq.ring.sched.ready && !adev->job_hang) + r = amdgpu_ring_test_helper(kiq_ring); spin_unlock(&adev->gfx.kiq.ring_lock); return r; @@ -535,6 +535,9 @@ int amdgpu_gfx_enable_kcq(struct amdgpu_device *adev) return r; } + if (adev->enable_mes) + queue_mask = ~0ULL; + kiq->pmf->kiq_set_resources(kiq_ring, queue_mask); for (i = 0; i < adev->gfx.num_compute_rings; i++) kiq->pmf->kiq_map_queues(kiq_ring, &adev->gfx.compute_ring[i]); @@ -608,6 +611,45 @@ unlock: mutex_unlock(&adev->gfx.gfx_off_mutex); } +int amdgpu_set_gfx_off_residency(struct amdgpu_device *adev, bool value) +{ + int r = 0; + + mutex_lock(&adev->gfx.gfx_off_mutex); + + r = amdgpu_dpm_set_residency_gfxoff(adev, value); + + mutex_unlock(&adev->gfx.gfx_off_mutex); + + return r; +} + +int amdgpu_get_gfx_off_residency(struct amdgpu_device *adev, u32 *value) +{ + int r = 0; + + mutex_lock(&adev->gfx.gfx_off_mutex); + + r = amdgpu_dpm_get_residency_gfxoff(adev, value); + + mutex_unlock(&adev->gfx.gfx_off_mutex); + + return r; +} + +int amdgpu_get_gfx_off_entrycount(struct amdgpu_device *adev, u64 *value) +{ + int r = 0; + + mutex_lock(&adev->gfx.gfx_off_mutex); + + r = amdgpu_dpm_get_entrycount_gfxoff(adev, value); + + mutex_unlock(&adev->gfx.gfx_off_mutex); + + return r; +} + int amdgpu_get_gfx_off_status(struct amdgpu_device *adev, uint32_t *value) { @@ -615,74 +657,38 @@ int amdgpu_get_gfx_off_status(struct amdgpu_device *adev, uint32_t *value) mutex_lock(&adev->gfx.gfx_off_mutex); - r = smu_get_status_gfxoff(adev, value); + r = amdgpu_dpm_get_status_gfxoff(adev, value); mutex_unlock(&adev->gfx.gfx_off_mutex); return r; } -int amdgpu_gfx_ras_late_init(struct amdgpu_device *adev) +int amdgpu_gfx_ras_late_init(struct amdgpu_device *adev, struct ras_common_if *ras_block) { int r; - struct ras_fs_if fs_info = { - .sysfs_name = "gfx_err_count", - }; - struct ras_ih_if ih_info = { - .cb = amdgpu_gfx_process_ras_data_cb, - }; - - if (!adev->gfx.ras_if) { - adev->gfx.ras_if = kmalloc(sizeof(struct ras_common_if), GFP_KERNEL); - if (!adev->gfx.ras_if) - return -ENOMEM; - adev->gfx.ras_if->block = AMDGPU_RAS_BLOCK__GFX; - adev->gfx.ras_if->type = AMDGPU_RAS_ERROR__MULTI_UNCORRECTABLE; - adev->gfx.ras_if->sub_block_index = 0; - } - fs_info.head = ih_info.head = *adev->gfx.ras_if; - r = amdgpu_ras_late_init(adev, adev->gfx.ras_if, - &fs_info, &ih_info); - if (r) - goto free; - if (amdgpu_ras_is_supported(adev, adev->gfx.ras_if->block)) { + if (amdgpu_ras_is_supported(adev, ras_block->block)) { if (!amdgpu_persistent_edc_harvesting_supported(adev)) amdgpu_ras_reset_error_status(adev, AMDGPU_RAS_BLOCK__GFX); + r = amdgpu_ras_block_late_init(adev, ras_block); + if (r) + return r; + r = amdgpu_irq_get(adev, &adev->gfx.cp_ecc_error_irq, 0); if (r) goto late_fini; } else { - /* free gfx ras_if if ras is not supported */ - r = 0; - goto free; + amdgpu_ras_feature_enable_on_boot(adev, ras_block, 0); } return 0; late_fini: - amdgpu_ras_late_fini(adev, adev->gfx.ras_if, &ih_info); -free: - kfree(adev->gfx.ras_if); - adev->gfx.ras_if = NULL; + amdgpu_ras_block_late_fini(adev, ras_block); return r; } -void amdgpu_gfx_ras_fini(struct amdgpu_device *adev) -{ - if (amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__GFX) && - adev->gfx.ras_if) { - struct ras_common_if *ras_if = adev->gfx.ras_if; - struct ras_ih_if ih_info = { - .head = *ras_if, - .cb = amdgpu_gfx_process_ras_data_cb, - }; - - amdgpu_ras_late_fini(adev, ras_if, &ih_info); - kfree(ras_if); - } -} - int amdgpu_gfx_process_ras_data_cb(struct amdgpu_device *adev, void *err_data, struct amdgpu_iv_entry *entry) @@ -695,9 +701,9 @@ int amdgpu_gfx_process_ras_data_cb(struct amdgpu_device *adev, */ if (!amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__GFX)) { kgd2kfd_set_sram_ecc_flag(adev->kfd.dev); - if (adev->gfx.ras_funcs && - adev->gfx.ras_funcs->query_ras_error_count) - adev->gfx.ras_funcs->query_ras_error_count(adev, err_data); + if (adev->gfx.ras && adev->gfx.ras->ras_block.hw_ops && + adev->gfx.ras->ras_block.hw_ops->query_ras_error_count) + adev->gfx.ras->ras_block.hw_ops->query_ras_error_count(adev, err_data); amdgpu_ras_reset_gpu(adev); } return AMDGPU_RAS_SUCCESS; @@ -733,6 +739,9 @@ uint32_t amdgpu_kiq_rreg(struct amdgpu_device *adev, uint32_t reg) if (amdgpu_device_skip_hw_access(adev)) return 0; + if (adev->mes.ring.sched.ready) + return amdgpu_mes_rreg(adev, reg); + BUG_ON(!ring->funcs->emit_rreg); spin_lock_irqsave(&kiq->ring_lock, flags); @@ -800,6 +809,11 @@ void amdgpu_kiq_wreg(struct amdgpu_device *adev, uint32_t reg, uint32_t v) if (amdgpu_device_skip_hw_access(adev)) return; + if (adev->mes.ring.sched.ready) { + amdgpu_mes_wreg(adev, reg, v); + return; + } + spin_lock_irqsave(&kiq->ring_lock, flags); amdgpu_ring_alloc(ring, 32); amdgpu_ring_emit_wreg(ring, reg, v); @@ -853,18 +867,141 @@ int amdgpu_gfx_get_num_kcq(struct amdgpu_device *adev) return amdgpu_num_kcq; } -/* amdgpu_gfx_state_change_set - Handle gfx power state change set - * @adev: amdgpu_device pointer - * @state: gfx power state(1 -sGpuChangeState_D0Entry and 2 -sGpuChangeState_D3Entry) - * - */ +void amdgpu_gfx_cp_init_microcode(struct amdgpu_device *adev, + uint32_t ucode_id) +{ + const struct gfx_firmware_header_v1_0 *cp_hdr; + const struct gfx_firmware_header_v2_0 *cp_hdr_v2_0; + struct amdgpu_firmware_info *info = NULL; + const struct firmware *ucode_fw; + unsigned int fw_size; + + switch (ucode_id) { + case AMDGPU_UCODE_ID_CP_PFP: + cp_hdr = (const struct gfx_firmware_header_v1_0 *) + adev->gfx.pfp_fw->data; + adev->gfx.pfp_fw_version = + le32_to_cpu(cp_hdr->header.ucode_version); + adev->gfx.pfp_feature_version = + le32_to_cpu(cp_hdr->ucode_feature_version); + ucode_fw = adev->gfx.pfp_fw; + fw_size = le32_to_cpu(cp_hdr->header.ucode_size_bytes); + break; + case AMDGPU_UCODE_ID_CP_RS64_PFP: + cp_hdr_v2_0 = (const struct gfx_firmware_header_v2_0 *) + adev->gfx.pfp_fw->data; + adev->gfx.pfp_fw_version = + le32_to_cpu(cp_hdr_v2_0->header.ucode_version); + adev->gfx.pfp_feature_version = + le32_to_cpu(cp_hdr_v2_0->ucode_feature_version); + ucode_fw = adev->gfx.pfp_fw; + fw_size = le32_to_cpu(cp_hdr_v2_0->ucode_size_bytes); + break; + case AMDGPU_UCODE_ID_CP_RS64_PFP_P0_STACK: + case AMDGPU_UCODE_ID_CP_RS64_PFP_P1_STACK: + cp_hdr_v2_0 = (const struct gfx_firmware_header_v2_0 *) + adev->gfx.pfp_fw->data; + ucode_fw = adev->gfx.pfp_fw; + fw_size = le32_to_cpu(cp_hdr_v2_0->data_size_bytes); + break; + case AMDGPU_UCODE_ID_CP_ME: + cp_hdr = (const struct gfx_firmware_header_v1_0 *) + adev->gfx.me_fw->data; + adev->gfx.me_fw_version = + le32_to_cpu(cp_hdr->header.ucode_version); + adev->gfx.me_feature_version = + le32_to_cpu(cp_hdr->ucode_feature_version); + ucode_fw = adev->gfx.me_fw; + fw_size = le32_to_cpu(cp_hdr->header.ucode_size_bytes); + break; + case AMDGPU_UCODE_ID_CP_RS64_ME: + cp_hdr_v2_0 = (const struct gfx_firmware_header_v2_0 *) + adev->gfx.me_fw->data; + adev->gfx.me_fw_version = + le32_to_cpu(cp_hdr_v2_0->header.ucode_version); + adev->gfx.me_feature_version = + le32_to_cpu(cp_hdr_v2_0->ucode_feature_version); + ucode_fw = adev->gfx.me_fw; + fw_size = le32_to_cpu(cp_hdr_v2_0->ucode_size_bytes); + break; + case AMDGPU_UCODE_ID_CP_RS64_ME_P0_STACK: + case AMDGPU_UCODE_ID_CP_RS64_ME_P1_STACK: + cp_hdr_v2_0 = (const struct gfx_firmware_header_v2_0 *) + adev->gfx.me_fw->data; + ucode_fw = adev->gfx.me_fw; + fw_size = le32_to_cpu(cp_hdr_v2_0->data_size_bytes); + break; + case AMDGPU_UCODE_ID_CP_CE: + cp_hdr = (const struct gfx_firmware_header_v1_0 *) + adev->gfx.ce_fw->data; + adev->gfx.ce_fw_version = + le32_to_cpu(cp_hdr->header.ucode_version); + adev->gfx.ce_feature_version = + le32_to_cpu(cp_hdr->ucode_feature_version); + ucode_fw = adev->gfx.ce_fw; + fw_size = le32_to_cpu(cp_hdr->header.ucode_size_bytes); + break; + case AMDGPU_UCODE_ID_CP_MEC1: + cp_hdr = (const struct gfx_firmware_header_v1_0 *) + adev->gfx.mec_fw->data; + adev->gfx.mec_fw_version = + le32_to_cpu(cp_hdr->header.ucode_version); + adev->gfx.mec_feature_version = + le32_to_cpu(cp_hdr->ucode_feature_version); + ucode_fw = adev->gfx.mec_fw; + fw_size = le32_to_cpu(cp_hdr->header.ucode_size_bytes) - + le32_to_cpu(cp_hdr->jt_size) * 4; + break; + case AMDGPU_UCODE_ID_CP_MEC1_JT: + cp_hdr = (const struct gfx_firmware_header_v1_0 *) + adev->gfx.mec_fw->data; + ucode_fw = adev->gfx.mec_fw; + fw_size = le32_to_cpu(cp_hdr->jt_size) * 4; + break; + case AMDGPU_UCODE_ID_CP_MEC2: + cp_hdr = (const struct gfx_firmware_header_v1_0 *) + adev->gfx.mec2_fw->data; + adev->gfx.mec2_fw_version = + le32_to_cpu(cp_hdr->header.ucode_version); + adev->gfx.mec2_feature_version = + le32_to_cpu(cp_hdr->ucode_feature_version); + ucode_fw = adev->gfx.mec2_fw; + fw_size = le32_to_cpu(cp_hdr->header.ucode_size_bytes) - + le32_to_cpu(cp_hdr->jt_size) * 4; + break; + case AMDGPU_UCODE_ID_CP_MEC2_JT: + cp_hdr = (const struct gfx_firmware_header_v1_0 *) + adev->gfx.mec2_fw->data; + ucode_fw = adev->gfx.mec2_fw; + fw_size = le32_to_cpu(cp_hdr->jt_size) * 4; + break; + case AMDGPU_UCODE_ID_CP_RS64_MEC: + cp_hdr_v2_0 = (const struct gfx_firmware_header_v2_0 *) + adev->gfx.mec_fw->data; + adev->gfx.mec_fw_version = + le32_to_cpu(cp_hdr_v2_0->header.ucode_version); + adev->gfx.mec_feature_version = + le32_to_cpu(cp_hdr_v2_0->ucode_feature_version); + ucode_fw = adev->gfx.mec_fw; + fw_size = le32_to_cpu(cp_hdr_v2_0->ucode_size_bytes); + break; + case AMDGPU_UCODE_ID_CP_RS64_MEC_P0_STACK: + case AMDGPU_UCODE_ID_CP_RS64_MEC_P1_STACK: + case AMDGPU_UCODE_ID_CP_RS64_MEC_P2_STACK: + case AMDGPU_UCODE_ID_CP_RS64_MEC_P3_STACK: + cp_hdr_v2_0 = (const struct gfx_firmware_header_v2_0 *) + adev->gfx.mec_fw->data; + ucode_fw = adev->gfx.mec_fw; + fw_size = le32_to_cpu(cp_hdr_v2_0->data_size_bytes); + break; + default: + break; + } -void amdgpu_gfx_state_change_set(struct amdgpu_device *adev, enum gfx_change_state state) -{ - mutex_lock(&adev->pm.mutex); - if (adev->powerplay.pp_funcs && - adev->powerplay.pp_funcs->gfx_state_change_set) - ((adev)->powerplay.pp_funcs->gfx_state_change_set( - (adev)->powerplay.pp_handle, state)); - mutex_unlock(&adev->pm.mutex); + if (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) { + info = &adev->firmware.ucode[ucode_id]; + info->ucode_id = ucode_id; + info->fw = ucode_fw; + adev->firmware.fw_size += ALIGN(fw_size, PAGE_SIZE); + } } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h index f851196c83a5..832b3807f1d6 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h @@ -30,7 +30,9 @@ #include "clearstate_defs.h" #include "amdgpu_ring.h" #include "amdgpu_rlc.h" +#include "amdgpu_imu.h" #include "soc15.h" +#include "amdgpu_ras.h" /* GFX current status */ #define AMDGPU_GFX_NORMAL_MODE 0x00000000L @@ -47,12 +49,6 @@ enum amdgpu_gfx_pipe_priority { AMDGPU_GFX_PIPE_PRIO_HIGH = AMDGPU_RING_PRIO_2 }; -/* Argument for PPSMC_MSG_GpuChangeState */ -enum gfx_change_state { - sGpuChangeState_D0Entry = 1, - sGpuChangeState_D3Entry, -}; - #define AMDGPU_GFX_QUEUE_PRIORITY_MINIMUM 0 #define AMDGPU_GFX_QUEUE_PRIORITY_MAXIMUM 15 @@ -61,6 +57,9 @@ struct amdgpu_mec { u64 hpd_eop_gpu_addr; struct amdgpu_bo *mec_fw_obj; u64 mec_fw_gpu_addr; + struct amdgpu_bo *mec_fw_data_obj; + u64 mec_fw_data_gpu_addr; + u32 num_mec; u32 num_pipe_per_mec; u32 num_queue_per_pipe; @@ -112,15 +111,6 @@ struct amdgpu_kiq { }; /* - * GPU scratch registers structures, functions & helpers - */ -struct amdgpu_scratch { - unsigned num_reg; - uint32_t reg_base; - uint32_t free_mask; -}; - -/* * GFX configurations */ #define AMDGPU_GFX_MAX_SE 4 @@ -188,6 +178,17 @@ struct amdgpu_gfx_config { uint32_t num_packer_per_sc; uint32_t pa_sc_tile_steering_override; uint64_t tcc_disabled_mask; + uint32_t gc_num_tcp_per_sa; + uint32_t gc_num_sdp_interface; + uint32_t gc_num_tcps; + uint32_t gc_num_tcp_per_wpg; + uint32_t gc_tcp_l1_size; + uint32_t gc_num_sqc_per_wgp; + uint32_t gc_l1_instruction_cache_size_per_sqc; + uint32_t gc_l1_data_cache_size_per_sqc; + uint32_t gc_gl1c_per_sa; + uint32_t gc_gl1c_size_per_instance; + uint32_t gc_gl2c_per_gpu; }; struct amdgpu_cu_info { @@ -204,17 +205,10 @@ struct amdgpu_cu_info { uint32_t bitmap[4][4]; }; -struct amdgpu_gfx_ras_funcs { - int (*ras_late_init)(struct amdgpu_device *adev); - void (*ras_fini)(struct amdgpu_device *adev); - int (*ras_error_inject)(struct amdgpu_device *adev, - void *inject_if); - int (*query_ras_error_count)(struct amdgpu_device *adev, - void *ras_error_status); - void (*reset_ras_error_count)(struct amdgpu_device *adev); - void (*query_ras_error_status)(struct amdgpu_device *adev); - void (*reset_ras_error_status)(struct amdgpu_device *adev); +struct amdgpu_gfx_ras { + struct amdgpu_ras_block_object ras_block; void (*enable_watchdog_timer)(struct amdgpu_device *adev); + bool (*query_utcl2_poison_status)(struct amdgpu_device *adev); }; struct amdgpu_gfx_funcs { @@ -245,6 +239,10 @@ struct amdgpu_pfp { struct amdgpu_bo *pfp_fw_obj; uint64_t pfp_fw_gpu_addr; uint32_t *pfp_fw_ptr; + + struct amdgpu_bo *pfp_fw_data_obj; + uint64_t pfp_fw_data_gpu_addr; + uint32_t *pfp_fw_data_ptr; }; struct amdgpu_ce { @@ -257,6 +255,11 @@ struct amdgpu_me { struct amdgpu_bo *me_fw_obj; uint64_t me_fw_gpu_addr; uint32_t *me_fw_ptr; + + struct amdgpu_bo *me_fw_data_obj; + uint64_t me_fw_data_gpu_addr; + uint32_t *me_fw_data_ptr; + uint32_t num_me; uint32_t num_pipe_per_me; uint32_t num_queue_per_pipe; @@ -275,7 +278,8 @@ struct amdgpu_gfx { struct amdgpu_me me; struct amdgpu_mec mec; struct amdgpu_kiq kiq; - struct amdgpu_scratch scratch; + struct amdgpu_imu imu; + bool rs64_enable; /* firmware format */ const struct firmware *me_fw; /* ME firmware */ uint32_t me_fw_version; const struct firmware *pfp_fw; /* PFP firmware */ @@ -288,6 +292,8 @@ struct amdgpu_gfx { uint32_t mec_fw_version; const struct firmware *mec2_fw; /* MEC2 firmware */ uint32_t mec2_fw_version; + const struct firmware *imu_fw; /* IMU firmware */ + uint32_t imu_fw_version; uint32_t me_feature_version; uint32_t ce_feature_version; uint32_t pfp_feature_version; @@ -298,6 +304,10 @@ struct amdgpu_gfx { uint32_t rlc_srlg_feature_version; uint32_t rlc_srls_fw_version; uint32_t rlc_srls_feature_version; + uint32_t rlcp_ucode_version; + uint32_t rlcp_ucode_feature_version; + uint32_t rlcv_ucode_version; + uint32_t rlcv_ucode_feature_version; uint32_t mec_feature_version; uint32_t mec2_feature_version; bool mec_fw_write_wait; @@ -326,18 +336,22 @@ struct amdgpu_gfx { uint32_t srbm_soft_reset; /* gfx off */ - bool gfx_off_state; /* true: enabled, false: disabled */ - struct mutex gfx_off_mutex; - uint32_t gfx_off_req_count; /* default 1, enable gfx off: dec 1, disable gfx off: add 1 */ - struct delayed_work gfx_off_delay_work; + bool gfx_off_state; /* true: enabled, false: disabled */ + struct mutex gfx_off_mutex; /* mutex to change gfxoff state */ + uint32_t gfx_off_req_count; /* default 1, enable gfx off: dec 1, disable gfx off: add 1 */ + struct delayed_work gfx_off_delay_work; /* async work to set gfx block off */ + uint32_t gfx_off_residency; /* last logged residency */ + uint64_t gfx_off_entrycount; /* count of times GPU has get into GFXOFF state */ /* pipe reservation */ struct mutex pipe_reserve_mutex; DECLARE_BITMAP (pipe_reserve_bitmap, AMDGPU_MAX_COMPUTE_QUEUES); /*ras */ - struct ras_common_if *ras_if; - const struct amdgpu_gfx_ras_funcs *ras_funcs; + struct ras_common_if *ras_if; + struct amdgpu_gfx_ras *ras; + + bool is_poweron; }; #define amdgpu_gfx_get_gpu_clock_counter(adev) (adev)->gfx.funcs->get_gpu_clock_counter((adev)) @@ -358,9 +372,6 @@ static inline u32 amdgpu_gfx_create_bitmask(u32 bit_width) return (u32)((1ULL << bit_width) - 1); } -int amdgpu_gfx_scratch_get(struct amdgpu_device *adev, uint32_t *reg); -void amdgpu_gfx_scratch_free(struct amdgpu_device *adev, uint32_t reg); - void amdgpu_gfx_parse_disable_cu(unsigned *mask, unsigned max_se, unsigned max_sh); @@ -391,6 +402,8 @@ bool amdgpu_gfx_is_mec_queue_enabled(struct amdgpu_device *adev, int mec, int pipe, int queue); bool amdgpu_gfx_is_high_priority_compute_queue(struct amdgpu_device *adev, struct amdgpu_ring *ring); +bool amdgpu_gfx_is_high_priority_graphics_queue(struct amdgpu_device *adev, + struct amdgpu_ring *ring); int amdgpu_gfx_me_queue_to_bit(struct amdgpu_device *adev, int me, int pipe, int queue); void amdgpu_gfx_bit_to_me_queue(struct amdgpu_device *adev, int bit, @@ -399,8 +412,11 @@ bool amdgpu_gfx_is_me_queue_enabled(struct amdgpu_device *adev, int me, int pipe, int queue); void amdgpu_gfx_off_ctrl(struct amdgpu_device *adev, bool enable); int amdgpu_get_gfx_off_status(struct amdgpu_device *adev, uint32_t *value); -int amdgpu_gfx_ras_late_init(struct amdgpu_device *adev); +int amdgpu_gfx_ras_late_init(struct amdgpu_device *adev, struct ras_common_if *ras_block); void amdgpu_gfx_ras_fini(struct amdgpu_device *adev); +int amdgpu_get_gfx_off_entrycount(struct amdgpu_device *adev, u64 *value); +int amdgpu_get_gfx_off_residency(struct amdgpu_device *adev, u32 *residency); +int amdgpu_set_gfx_off_residency(struct amdgpu_device *adev, bool value); int amdgpu_gfx_process_ras_data_cb(struct amdgpu_device *adev, void *err_data, struct amdgpu_iv_entry *entry); @@ -410,5 +426,6 @@ int amdgpu_gfx_cp_ecc_error_irq(struct amdgpu_device *adev, uint32_t amdgpu_kiq_rreg(struct amdgpu_device *adev, uint32_t reg); void amdgpu_kiq_wreg(struct amdgpu_device *adev, uint32_t reg, uint32_t v); int amdgpu_gfx_get_num_kcq(struct amdgpu_device *adev); -void amdgpu_gfx_state_change_set(struct amdgpu_device *adev, enum gfx_change_state state); +void amdgpu_gfx_cp_init_microcode(struct amdgpu_device *adev, uint32_t ucode_id); + #endif diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfxhub.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfxhub.h index beabab515836..c7b44aeb671b 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfxhub.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfxhub.h @@ -35,6 +35,9 @@ struct amdgpu_gfxhub_funcs { void (*init)(struct amdgpu_device *adev); int (*get_xgmi_info)(struct amdgpu_device *adev); void (*utcl2_harvest)(struct amdgpu_device *adev); + void (*mode2_save_regs)(struct amdgpu_device *adev); + void (*mode2_restore_regs)(struct amdgpu_device *adev); + void (*halt)(struct amdgpu_device *adev); }; struct amdgpu_gfxhub { diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c index 2430d6223c2d..34233a74248c 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c @@ -25,6 +25,9 @@ */ #include <linux/io-64-nonatomic-lo-hi.h> +#ifdef CONFIG_X86 +#include <asm/hypervisor.h> +#endif #include "amdgpu.h" #include "amdgpu_gmc.h" @@ -436,82 +439,25 @@ void amdgpu_gmc_filter_faults_remove(struct amdgpu_device *adev, uint64_t addr, } while (fault->timestamp < tmp); } -int amdgpu_gmc_ras_late_init(struct amdgpu_device *adev) +int amdgpu_gmc_ras_early_init(struct amdgpu_device *adev) { - int r; - - if (adev->umc.ras_funcs && - adev->umc.ras_funcs->ras_late_init) { - r = adev->umc.ras_funcs->ras_late_init(adev); - if (r) - return r; - } - - if (adev->mmhub.ras_funcs && - adev->mmhub.ras_funcs->ras_late_init) { - r = adev->mmhub.ras_funcs->ras_late_init(adev); - if (r) - return r; - } - - if (!adev->gmc.xgmi.connected_to_cpu) - adev->gmc.xgmi.ras_funcs = &xgmi_ras_funcs; - - if (adev->gmc.xgmi.ras_funcs && - adev->gmc.xgmi.ras_funcs->ras_late_init) { - r = adev->gmc.xgmi.ras_funcs->ras_late_init(adev); - if (r) - return r; - } - - if (adev->hdp.ras_funcs && - adev->hdp.ras_funcs->ras_late_init) { - r = adev->hdp.ras_funcs->ras_late_init(adev); - if (r) - return r; - } - - if (adev->mca.mp0.ras_funcs && - adev->mca.mp0.ras_funcs->ras_late_init) { - r = adev->mca.mp0.ras_funcs->ras_late_init(adev); - if (r) - return r; - } - - if (adev->mca.mp1.ras_funcs && - adev->mca.mp1.ras_funcs->ras_late_init) { - r = adev->mca.mp1.ras_funcs->ras_late_init(adev); - if (r) - return r; + if (!adev->gmc.xgmi.connected_to_cpu) { + adev->gmc.xgmi.ras = &xgmi_ras; + amdgpu_ras_register_ras_block(adev, &adev->gmc.xgmi.ras->ras_block); + adev->gmc.xgmi.ras_if = &adev->gmc.xgmi.ras->ras_block.ras_comm; } - if (adev->mca.mpio.ras_funcs && - adev->mca.mpio.ras_funcs->ras_late_init) { - r = adev->mca.mpio.ras_funcs->ras_late_init(adev); - if (r) - return r; - } + return 0; +} +int amdgpu_gmc_ras_late_init(struct amdgpu_device *adev) +{ return 0; } void amdgpu_gmc_ras_fini(struct amdgpu_device *adev) { - if (adev->umc.ras_funcs && - adev->umc.ras_funcs->ras_fini) - adev->umc.ras_funcs->ras_fini(adev); - - if (adev->mmhub.ras_funcs && - adev->mmhub.ras_funcs->ras_fini) - adev->mmhub.ras_funcs->ras_fini(adev); - - if (adev->gmc.xgmi.ras_funcs && - adev->gmc.xgmi.ras_funcs->ras_fini) - adev->gmc.xgmi.ras_funcs->ras_fini(adev); - if (adev->hdp.ras_funcs && - adev->hdp.ras_funcs->ras_fini) - adev->hdp.ras_funcs->ras_fini(adev); } /* @@ -566,9 +512,14 @@ int amdgpu_gmc_allocate_vm_inv_eng(struct amdgpu_device *adev) */ void amdgpu_gmc_tmz_set(struct amdgpu_device *adev) { - switch (adev->asic_type) { - case CHIP_RAVEN: - case CHIP_RENOIR: + switch (adev->ip_versions[GC_HWIP][0]) { + /* RAVEN */ + case IP_VERSION(9, 2, 2): + case IP_VERSION(9, 1, 0): + /* RENOIR looks like RAVEN */ + case IP_VERSION(9, 3, 0): + /* GC 10.3.7 */ + case IP_VERSION(10, 3, 7): if (amdgpu_tmz == 0) { adev->gmc.tmz_enabled = false; dev_info(adev->dev, @@ -579,11 +530,18 @@ void amdgpu_gmc_tmz_set(struct amdgpu_device *adev) "Trusted Memory Zone (TMZ) feature enabled\n"); } break; - case CHIP_NAVI10: - case CHIP_NAVI14: - case CHIP_NAVI12: - case CHIP_VANGOGH: - case CHIP_YELLOW_CARP: + case IP_VERSION(10, 1, 10): + case IP_VERSION(10, 1, 1): + case IP_VERSION(10, 1, 2): + case IP_VERSION(10, 1, 3): + case IP_VERSION(10, 3, 0): + case IP_VERSION(10, 3, 2): + case IP_VERSION(10, 3, 4): + case IP_VERSION(10, 3, 5): + /* VANGOGH */ + case IP_VERSION(10, 3, 1): + /* YELLOW_CARP*/ + case IP_VERSION(10, 3, 3): /* Don't enable it by default yet. */ if (amdgpu_tmz < 1) { @@ -614,40 +572,15 @@ void amdgpu_gmc_tmz_set(struct amdgpu_device *adev) void amdgpu_gmc_noretry_set(struct amdgpu_device *adev) { struct amdgpu_gmc *gmc = &adev->gmc; - - switch (adev->asic_type) { - case CHIP_VEGA10: - case CHIP_VEGA20: - case CHIP_ARCTURUS: - case CHIP_ALDEBARAN: - /* - * noretry = 0 will cause kfd page fault tests fail - * for some ASICs, so set default to 1 for these ASICs. - */ - if (amdgpu_noretry == -1) - gmc->noretry = 1; - else - gmc->noretry = amdgpu_noretry; - break; - case CHIP_RAVEN: - default: - /* Raven currently has issues with noretry - * regardless of what we decide for other - * asics, we should leave raven with - * noretry = 0 until we root cause the - * issues. - * - * default this to 0 for now, but we may want - * to change this in the future for certain - * GPUs as it can increase performance in - * certain cases. - */ - if (amdgpu_noretry == -1) - gmc->noretry = 0; - else - gmc->noretry = amdgpu_noretry; - break; - } + uint32_t gc_ver = adev->ip_versions[GC_HWIP][0]; + bool noretry_default = (gc_ver == IP_VERSION(9, 0, 1) || + gc_ver == IP_VERSION(9, 3, 0) || + gc_ver == IP_VERSION(9, 4, 0) || + gc_ver == IP_VERSION(9, 4, 1) || + gc_ver == IP_VERSION(9, 4, 2) || + gc_ver >= IP_VERSION(10, 3, 0)); + + gmc->noretry = (amdgpu_noretry == -1) ? noretry_default : amdgpu_noretry; } void amdgpu_gmc_set_vm_fault_masks(struct amdgpu_device *adev, int hub_type, @@ -680,6 +613,13 @@ void amdgpu_gmc_get_vbios_allocations(struct amdgpu_device *adev) unsigned size; /* + * Some ASICs need to reserve a region of video memory to avoid access + * from driver + */ + adev->mman.stolen_reserved_offset = 0; + adev->mman.stolen_reserved_size = 0; + + /* * TODO: * Currently there is a bug where some memory client outside * of the driver writes to first 8M of VRAM on S3 resume, @@ -689,10 +629,27 @@ void amdgpu_gmc_get_vbios_allocations(struct amdgpu_device *adev) */ switch (adev->asic_type) { case CHIP_VEGA10: + adev->mman.keep_stolen_vga_memory = true; + /* + * VEGA10 SRIOV VF with MS_HYPERV host needs some firmware reserved area. + */ +#ifdef CONFIG_X86 + if (amdgpu_sriov_vf(adev) && hypervisor_is_type(X86_HYPER_MS_HYPERV)) { + adev->mman.stolen_reserved_offset = 0x500000; + adev->mman.stolen_reserved_size = 0x200000; + } +#endif + break; case CHIP_RAVEN: case CHIP_RENOIR: adev->mman.keep_stolen_vga_memory = true; break; + case CHIP_YELLOW_CARP: + if (amdgpu_discovery == 0) { + adev->mman.stolen_reserved_offset = 0x1ffb0000; + adev->mman.stolen_reserved_size = 64 * PAGE_SIZE; + } + break; default: adev->mman.keep_stolen_vga_memory = false; break; @@ -813,21 +770,48 @@ uint64_t amdgpu_gmc_vram_cpu_pa(struct amdgpu_device *adev, struct amdgpu_bo *bo return amdgpu_bo_gpu_offset(bo) - adev->gmc.vram_start + adev->gmc.aper_base; } -void amdgpu_gmc_get_reserved_allocation(struct amdgpu_device *adev) +int amdgpu_gmc_vram_checking(struct amdgpu_device *adev) { - /* Some ASICs need to reserve a region of video memory to avoid access - * from driver */ - adev->mman.stolen_reserved_offset = 0; - adev->mman.stolen_reserved_size = 0; + struct amdgpu_bo *vram_bo = NULL; + uint64_t vram_gpu = 0; + void *vram_ptr = NULL; + + int ret, size = 0x100000; + uint8_t cptr[10]; + + ret = amdgpu_bo_create_kernel(adev, size, PAGE_SIZE, + AMDGPU_GEM_DOMAIN_VRAM, + &vram_bo, + &vram_gpu, + &vram_ptr); + if (ret) + return ret; + + memset(vram_ptr, 0x86, size); + memset(cptr, 0x86, 10); + + /** + * Check the start, the mid, and the end of the memory if the content of + * each byte is the pattern "0x86". If yes, we suppose the vram bo is + * workable. + * + * Note: If check the each byte of whole 1M bo, it will cost too many + * seconds, so here, we just pick up three parts for emulation. + */ + ret = memcmp(vram_ptr, cptr, 10); + if (ret) + return ret; - switch (adev->asic_type) { - case CHIP_YELLOW_CARP: - if (amdgpu_discovery == 0) { - adev->mman.stolen_reserved_offset = 0x1ffb0000; - adev->mman.stolen_reserved_size = 64 * PAGE_SIZE; - } - break; - default: - break; - } + ret = memcmp(vram_ptr + (size / 2), cptr, 10); + if (ret) + return ret; + + ret = memcmp(vram_ptr + size - 10, cptr, 10); + if (ret) + return ret; + + amdgpu_bo_free_kernel(&vram_bo, &vram_gpu, + &vram_ptr); + + return 0; } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.h index 8458cebc6d5b..0305b660cd17 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.h @@ -29,6 +29,7 @@ #include <linux/types.h> #include "amdgpu_irq.h" +#include "amdgpu_ras.h" /* VA hole for 48bit addresses on Vega10 */ #define AMDGPU_GMC_HOLE_START 0x0000800000000000ULL @@ -99,7 +100,9 @@ struct amdgpu_vmhub { uint32_t eng_distance; uint32_t eng_addr_distance; /* include LO32/HI32 */ + uint32_t vm_cntx_cntl; uint32_t vm_cntx_cntl_vm_fault; + uint32_t vm_l2_bank_select_reserved_cid2; const struct amdgpu_vmhub_funcs *vmhub_funcs; }; @@ -135,12 +138,8 @@ struct amdgpu_gmc_funcs { unsigned int (*get_vbios_fb_size)(struct amdgpu_device *adev); }; -struct amdgpu_xgmi_ras_funcs { - int (*ras_late_init)(struct amdgpu_device *adev); - void (*ras_fini)(struct amdgpu_device *adev); - int (*query_ras_error_count)(struct amdgpu_device *adev, - void *ras_error_status); - void (*reset_ras_error_count)(struct amdgpu_device *adev); +struct amdgpu_xgmi_ras { + struct amdgpu_ras_block_object ras_block; }; struct amdgpu_xgmi { @@ -159,7 +158,7 @@ struct amdgpu_xgmi { struct ras_common_if *ras_if; bool connected_to_cpu; bool pending_reset; - const struct amdgpu_xgmi_ras_funcs *ras_funcs; + struct amdgpu_xgmi_ras *ras; }; struct amdgpu_gmc { @@ -260,6 +259,37 @@ struct amdgpu_gmc { struct amdgpu_bo *pdb0_bo; /* CPU kmapped address of pdb0*/ void *ptr_pdb0; + + /* MALL size */ + u64 mall_size; + /* number of UMC instances */ + int num_umc; + /* mode2 save restore */ + u64 VM_L2_CNTL; + u64 VM_L2_CNTL2; + u64 VM_DUMMY_PAGE_FAULT_CNTL; + u64 VM_DUMMY_PAGE_FAULT_ADDR_LO32; + u64 VM_DUMMY_PAGE_FAULT_ADDR_HI32; + u64 VM_L2_PROTECTION_FAULT_CNTL; + u64 VM_L2_PROTECTION_FAULT_CNTL2; + u64 VM_L2_PROTECTION_FAULT_MM_CNTL3; + u64 VM_L2_PROTECTION_FAULT_MM_CNTL4; + u64 VM_L2_PROTECTION_FAULT_ADDR_LO32; + u64 VM_L2_PROTECTION_FAULT_ADDR_HI32; + u64 VM_DEBUG; + u64 VM_L2_MM_GROUP_RT_CLASSES; + u64 VM_L2_BANK_SELECT_RESERVED_CID; + u64 VM_L2_BANK_SELECT_RESERVED_CID2; + u64 VM_L2_CACHE_PARITY_CNTL; + u64 VM_L2_IH_LOG_CNTL; + u64 VM_CONTEXT_CNTL[16]; + u64 VM_CONTEXT_PAGE_TABLE_BASE_ADDR_LO32[16]; + u64 VM_CONTEXT_PAGE_TABLE_BASE_ADDR_HI32[16]; + u64 VM_CONTEXT_PAGE_TABLE_START_ADDR_LO32[16]; + u64 VM_CONTEXT_PAGE_TABLE_START_ADDR_HI32[16]; + u64 VM_CONTEXT_PAGE_TABLE_END_ADDR_LO32[16]; + u64 VM_CONTEXT_PAGE_TABLE_END_ADDR_HI32[16]; + u64 MC_VM_MX_L1_TLB_CNTL; }; #define amdgpu_gmc_flush_gpu_tlb(adev, vmid, vmhub, type) ((adev)->gmc.gmc_funcs->flush_gpu_tlb((adev), (vmid), (vmhub), (type))) @@ -321,6 +351,7 @@ bool amdgpu_gmc_filter_faults(struct amdgpu_device *adev, uint16_t pasid, uint64_t timestamp); void amdgpu_gmc_filter_faults_remove(struct amdgpu_device *adev, uint64_t addr, uint16_t pasid); +int amdgpu_gmc_ras_early_init(struct amdgpu_device *adev); int amdgpu_gmc_ras_late_init(struct amdgpu_device *adev); void amdgpu_gmc_ras_fini(struct amdgpu_device *adev); int amdgpu_gmc_allocate_vm_inv_eng(struct amdgpu_device *adev); @@ -333,10 +364,10 @@ amdgpu_gmc_set_vm_fault_masks(struct amdgpu_device *adev, int hub_type, bool enable); void amdgpu_gmc_get_vbios_allocations(struct amdgpu_device *adev); -void amdgpu_gmc_get_reserved_allocation(struct amdgpu_device *adev); void amdgpu_gmc_init_pdb0(struct amdgpu_device *adev); uint64_t amdgpu_gmc_vram_mc2pa(struct amdgpu_device *adev, uint64_t mc_addr); uint64_t amdgpu_gmc_vram_pa(struct amdgpu_device *adev, struct amdgpu_bo *bo); uint64_t amdgpu_gmc_vram_cpu_pa(struct amdgpu_device *adev, struct amdgpu_bo *bo); +int amdgpu_gmc_vram_checking(struct amdgpu_device *adev); #endif diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gtt_mgr.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gtt_mgr.c index 675a72ef305d..1f3302aebeff 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gtt_mgr.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gtt_mgr.c @@ -26,23 +26,12 @@ #include "amdgpu.h" -struct amdgpu_gtt_node { - struct ttm_buffer_object *tbo; - struct ttm_range_mgr_node base; -}; - static inline struct amdgpu_gtt_mgr * to_gtt_mgr(struct ttm_resource_manager *man) { return container_of(man, struct amdgpu_gtt_mgr, manager); } -static inline struct amdgpu_gtt_node * -to_amdgpu_gtt_node(struct ttm_resource *res) -{ - return container_of(res, struct amdgpu_gtt_node, base.base); -} - /** * DOC: mem_info_gtt_total * @@ -60,7 +49,7 @@ static ssize_t amdgpu_mem_info_gtt_total_show(struct device *dev, struct ttm_resource_manager *man; man = ttm_manager_type(&adev->mman.bdev, TTM_PL_TT); - return sysfs_emit(buf, "%llu\n", man->size * PAGE_SIZE); + return sysfs_emit(buf, "%llu\n", man->size); } /** @@ -77,10 +66,9 @@ static ssize_t amdgpu_mem_info_gtt_used_show(struct device *dev, { struct drm_device *ddev = dev_get_drvdata(dev); struct amdgpu_device *adev = drm_to_adev(ddev); - struct ttm_resource_manager *man; + struct ttm_resource_manager *man = &adev->mman.gtt_mgr.manager; - man = ttm_manager_type(&adev->mman.bdev, TTM_PL_TT); - return sysfs_emit(buf, "%llu\n", amdgpu_gtt_mgr_usage(man)); + return sysfs_emit(buf, "%llu\n", ttm_resource_manager_usage(man)); } static DEVICE_ATTR(mem_info_gtt_total, S_IRUGO, @@ -107,9 +95,9 @@ const struct attribute_group amdgpu_gtt_mgr_attr_group = { */ bool amdgpu_gtt_mgr_has_gart_addr(struct ttm_resource *res) { - struct amdgpu_gtt_node *node = to_amdgpu_gtt_node(res); + struct ttm_range_mgr_node *node = to_ttm_range_mgr_node(res); - return drm_mm_node_allocated(&node->base.mm_nodes[0]); + return drm_mm_node_allocated(&node->mm_nodes[0]); } /** @@ -129,28 +117,23 @@ static int amdgpu_gtt_mgr_new(struct ttm_resource_manager *man, { struct amdgpu_gtt_mgr *mgr = to_gtt_mgr(man); uint32_t num_pages = PFN_UP(tbo->base.size); - struct amdgpu_gtt_node *node; + struct ttm_range_mgr_node *node; int r; - if (!(place->flags & TTM_PL_FLAG_TEMPORARY) && - atomic64_add_return(num_pages, &mgr->used) > man->size) { - atomic64_sub(num_pages, &mgr->used); - return -ENOSPC; - } + node = kzalloc(struct_size(node, mm_nodes, 1), GFP_KERNEL); + if (!node) + return -ENOMEM; - node = kzalloc(struct_size(node, base.mm_nodes, 1), GFP_KERNEL); - if (!node) { - r = -ENOMEM; - goto err_out; + ttm_resource_init(tbo, place, &node->base); + if (!(place->flags & TTM_PL_FLAG_TEMPORARY) && + ttm_resource_manager_usage(man) > man->size) { + r = -ENOSPC; + goto err_free; } - node->tbo = tbo; - ttm_resource_init(tbo, place, &node->base.base); - if (place->lpfn) { spin_lock(&mgr->lock); - r = drm_mm_insert_node_in_range(&mgr->mm, - &node->base.mm_nodes[0], + r = drm_mm_insert_node_in_range(&mgr->mm, &node->mm_nodes[0], num_pages, tbo->page_alignment, 0, place->fpfn, place->lpfn, DRM_MM_INSERT_BEST); @@ -158,23 +141,19 @@ static int amdgpu_gtt_mgr_new(struct ttm_resource_manager *man, if (unlikely(r)) goto err_free; - node->base.base.start = node->base.mm_nodes[0].start; + node->base.start = node->mm_nodes[0].start; } else { - node->base.mm_nodes[0].start = 0; - node->base.mm_nodes[0].size = node->base.base.num_pages; - node->base.base.start = AMDGPU_BO_INVALID_OFFSET; + node->mm_nodes[0].start = 0; + node->mm_nodes[0].size = node->base.num_pages; + node->base.start = AMDGPU_BO_INVALID_OFFSET; } - *res = &node->base.base; + *res = &node->base; return 0; err_free: + ttm_resource_fini(man, &node->base); kfree(node); - -err_out: - if (!(place->flags & TTM_PL_FLAG_TEMPORARY)) - atomic64_sub(num_pages, &mgr->used); - return r; } @@ -189,62 +168,76 @@ err_out: static void amdgpu_gtt_mgr_del(struct ttm_resource_manager *man, struct ttm_resource *res) { - struct amdgpu_gtt_node *node = to_amdgpu_gtt_node(res); + struct ttm_range_mgr_node *node = to_ttm_range_mgr_node(res); struct amdgpu_gtt_mgr *mgr = to_gtt_mgr(man); spin_lock(&mgr->lock); - if (drm_mm_node_allocated(&node->base.mm_nodes[0])) - drm_mm_remove_node(&node->base.mm_nodes[0]); + if (drm_mm_node_allocated(&node->mm_nodes[0])) + drm_mm_remove_node(&node->mm_nodes[0]); spin_unlock(&mgr->lock); - if (!(res->placement & TTM_PL_FLAG_TEMPORARY)) - atomic64_sub(res->num_pages, &mgr->used); - + ttm_resource_fini(man, res); kfree(node); } /** - * amdgpu_gtt_mgr_usage - return usage of GTT domain - * - * @man: TTM memory type manager - * - * Return how many bytes are used in the GTT domain - */ -uint64_t amdgpu_gtt_mgr_usage(struct ttm_resource_manager *man) -{ - struct amdgpu_gtt_mgr *mgr = to_gtt_mgr(man); - - return atomic64_read(&mgr->used) * PAGE_SIZE; -} - -/** * amdgpu_gtt_mgr_recover - re-init gart * - * @man: TTM memory type manager + * @mgr: amdgpu_gtt_mgr pointer * * Re-init the gart for each known BO in the GTT. */ -int amdgpu_gtt_mgr_recover(struct ttm_resource_manager *man) +void amdgpu_gtt_mgr_recover(struct amdgpu_gtt_mgr *mgr) { - struct amdgpu_gtt_mgr *mgr = to_gtt_mgr(man); - struct amdgpu_device *adev; - struct amdgpu_gtt_node *node; + struct ttm_range_mgr_node *node; struct drm_mm_node *mm_node; - int r = 0; + struct amdgpu_device *adev; adev = container_of(mgr, typeof(*adev), mman.gtt_mgr); spin_lock(&mgr->lock); drm_mm_for_each_node(mm_node, &mgr->mm) { - node = container_of(mm_node, typeof(*node), base.mm_nodes[0]); - r = amdgpu_ttm_recover_gart(node->tbo); - if (r) - break; + node = container_of(mm_node, typeof(*node), mm_nodes[0]); + amdgpu_ttm_recover_gart(node->base.bo); } spin_unlock(&mgr->lock); amdgpu_gart_invalidate_tlb(adev); +} - return r; +/** + * amdgpu_gtt_mgr_intersects - test for intersection + * + * @man: Our manager object + * @res: The resource to test + * @place: The place for the new allocation + * @size: The size of the new allocation + * + * Simplified intersection test, only interesting if we need GART or not. + */ +static bool amdgpu_gtt_mgr_intersects(struct ttm_resource_manager *man, + struct ttm_resource *res, + const struct ttm_place *place, + size_t size) +{ + return !place->lpfn || amdgpu_gtt_mgr_has_gart_addr(res); +} + +/** + * amdgpu_gtt_mgr_compatible - test for compatibility + * + * @man: Our manager object + * @res: The resource to test + * @place: The place for the new allocation + * @size: The size of the new allocation + * + * Simplified compatibility test. + */ +static bool amdgpu_gtt_mgr_compatible(struct ttm_resource_manager *man, + struct ttm_resource *res, + const struct ttm_place *place, + size_t size) +{ + return !place->lpfn || amdgpu_gtt_mgr_has_gart_addr(res); } /** @@ -263,14 +256,13 @@ static void amdgpu_gtt_mgr_debug(struct ttm_resource_manager *man, spin_lock(&mgr->lock); drm_mm_print(&mgr->mm, printer); spin_unlock(&mgr->lock); - - drm_printf(printer, "man size:%llu pages, gtt used:%llu pages\n", - man->size, atomic64_read(&mgr->used)); } static const struct ttm_resource_manager_func amdgpu_gtt_mgr_func = { .alloc = amdgpu_gtt_mgr_new, .free = amdgpu_gtt_mgr_del, + .intersects = amdgpu_gtt_mgr_intersects, + .compatible = amdgpu_gtt_mgr_compatible, .debug = amdgpu_gtt_mgr_debug }; @@ -291,13 +283,12 @@ int amdgpu_gtt_mgr_init(struct amdgpu_device *adev, uint64_t gtt_size) man->use_tt = true; man->func = &amdgpu_gtt_mgr_func; - ttm_resource_manager_init(man, gtt_size >> PAGE_SHIFT); + ttm_resource_manager_init(man, &adev->mman.bdev, gtt_size); start = AMDGPU_GTT_MAX_TRANSFER_SIZE * AMDGPU_GTT_NUM_TRANSFER_WINDOWS; size = (adev->gmc.gart_size >> PAGE_SHIFT) - start; drm_mm_init(&mgr->mm, start, size); spin_lock_init(&mgr->lock); - atomic64_set(&mgr->used, 0); ttm_set_driver_manager(&adev->mman.bdev, TTM_PL_TT, &mgr->manager); ttm_resource_manager_set_used(man, true); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_hdp.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_hdp.h index 7ec99d591584..ac5c61d3de2b 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_hdp.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_hdp.h @@ -22,13 +22,10 @@ */ #ifndef __AMDGPU_HDP_H__ #define __AMDGPU_HDP_H__ +#include "amdgpu_ras.h" -struct amdgpu_hdp_ras_funcs { - int (*ras_late_init)(struct amdgpu_device *adev); - void (*ras_fini)(struct amdgpu_device *adev); - void (*query_ras_error_count)(struct amdgpu_device *adev, - void *ras_error_status); - void (*reset_ras_error_count)(struct amdgpu_device *adev); +struct amdgpu_hdp_ras { + struct amdgpu_ras_block_object ras_block; }; struct amdgpu_hdp_funcs { @@ -36,16 +33,15 @@ struct amdgpu_hdp_funcs { void (*invalidate_hdp)(struct amdgpu_device *adev, struct amdgpu_ring *ring); void (*update_clock_gating)(struct amdgpu_device *adev, bool enable); - void (*get_clock_gating_state)(struct amdgpu_device *adev, u32 *flags); + void (*get_clock_gating_state)(struct amdgpu_device *adev, u64 *flags); void (*init_registers)(struct amdgpu_device *adev); }; struct amdgpu_hdp { struct ras_common_if *ras_if; const struct amdgpu_hdp_funcs *funcs; - const struct amdgpu_hdp_ras_funcs *ras_funcs; + struct amdgpu_hdp_ras *ras; }; -int amdgpu_hdp_ras_late_init(struct amdgpu_device *adev); -void amdgpu_hdp_ras_fini(struct amdgpu_device *adev); +int amdgpu_hdp_ras_late_init(struct amdgpu_device *adev, struct ras_common_if *ras_block); #endif /* __AMDGPU_HDP_H__ */ diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c index bc1297dcdf97..258cffe3c06a 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c @@ -155,19 +155,19 @@ int amdgpu_ib_schedule(struct amdgpu_ring *ring, unsigned num_ibs, fence_ctx = 0; } - if (!ring->sched.ready) { + if (!ring->sched.ready && !ring->is_mes_queue) { dev_err(adev->dev, "couldn't schedule ib on ring <%s>\n", ring->name); return -EINVAL; } - if (vm && !job->vmid) { + if (vm && !job->vmid && !ring->is_mes_queue) { dev_err(adev->dev, "VM IB without ID\n"); return -EINVAL; } if ((ib->flags & AMDGPU_IB_FLAGS_SECURE) && - (ring->funcs->type == AMDGPU_RING_TYPE_COMPUTE)) { - dev_err(adev->dev, "secure submissions not supported on compute rings\n"); + (!ring->funcs->secure_submission_supported)) { + dev_err(adev->dev, "secure submissions not supported on ring <%s>\n", ring->name); return -EINVAL; } @@ -390,6 +390,10 @@ int amdgpu_ib_ring_tests(struct amdgpu_device *adev) if (!ring->sched.ready || !ring->funcs->test_ib) continue; + if (adev->enable_mes && + ring->funcs->type == AMDGPU_RING_TYPE_KIQ) + continue; + /* MM engine need more time */ if (ring->funcs->type == AMDGPU_RING_TYPE_UVD || ring->funcs->type == AMDGPU_RING_TYPE_VCE || diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ids.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ids.c index b7fb72bff2c1..03d115d2b5ed 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ids.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ids.c @@ -107,36 +107,19 @@ static void amdgpu_pasid_free_cb(struct dma_fence *fence, void amdgpu_pasid_free_delayed(struct dma_resv *resv, u32 pasid) { - struct dma_fence *fence, **fences; struct amdgpu_pasid_cb *cb; - unsigned count; + struct dma_fence *fence; int r; - r = dma_resv_get_fences(resv, NULL, &count, &fences); + r = dma_resv_get_singleton(resv, DMA_RESV_USAGE_BOOKKEEP, &fence); if (r) goto fallback; - if (count == 0) { + if (!fence) { amdgpu_pasid_free(pasid); return; } - if (count == 1) { - fence = fences[0]; - kfree(fences); - } else { - uint64_t context = dma_fence_context_alloc(1); - struct dma_fence_array *array; - - array = dma_fence_array_create(count, fences, context, - 1, false); - if (!array) { - kfree(fences); - goto fallback; - } - fence = &array->base; - } - cb = kmalloc(sizeof(*cb), GFP_KERNEL); if (!cb) { /* Last resort when we are OOM */ @@ -156,7 +139,8 @@ fallback: /* Not enough memory for the delayed delete, as last resort * block for all the fences to complete. */ - dma_resv_wait_timeout(resv, true, false, MAX_SCHEDULE_TIMEOUT); + dma_resv_wait_timeout(resv, DMA_RESV_USAGE_BOOKKEEP, + false, MAX_SCHEDULE_TIMEOUT); amdgpu_pasid_free(pasid); } @@ -204,7 +188,7 @@ static int amdgpu_vmid_grab_idle(struct amdgpu_vm *vm, unsigned i; int r; - if (ring->vmid_wait && !dma_fence_is_signaled(ring->vmid_wait)) + if (!dma_fence_is_signaled(ring->vmid_wait)) return amdgpu_sync_fence(sync, ring->vmid_wait); fences = kmalloc_array(id_mgr->num_ids, sizeof(void *), GFP_KERNEL); @@ -276,19 +260,15 @@ static int amdgpu_vmid_grab_reserved(struct amdgpu_vm *vm, struct amdgpu_device *adev = ring->adev; unsigned vmhub = ring->funcs->vmhub; uint64_t fence_context = adev->fence_context + ring->idx; - struct dma_fence *updates = sync->last_vm_update; bool needs_flush = vm->use_cpu_for_update; - int r = 0; + uint64_t updates = amdgpu_vm_tlb_seq(vm); + int r; *id = vm->reserved_vmid[vmhub]; - if (updates && (*id)->flushed_updates && - updates->context == (*id)->flushed_updates->context && - !dma_fence_is_later(updates, (*id)->flushed_updates)) - updates = NULL; - if ((*id)->owner != vm->immediate.fence_context || - job->vm_pd_addr != (*id)->pd_gpu_addr || - updates || !(*id)->last_flush || + (*id)->pd_gpu_addr != job->vm_pd_addr || + (*id)->flushed_updates < updates || + !(*id)->last_flush || ((*id)->last_flush->context != fence_context && !dma_fence_is_signaled((*id)->last_flush))) { struct dma_fence *tmp; @@ -302,8 +282,7 @@ static int amdgpu_vmid_grab_reserved(struct amdgpu_vm *vm, tmp = amdgpu_sync_peek_fence(&(*id)->active, ring); if (tmp) { *id = NULL; - r = amdgpu_sync_fence(sync, tmp); - return r; + return amdgpu_sync_fence(sync, tmp); } needs_flush = true; } @@ -315,10 +294,7 @@ static int amdgpu_vmid_grab_reserved(struct amdgpu_vm *vm, if (r) return r; - if (updates) { - dma_fence_put((*id)->flushed_updates); - (*id)->flushed_updates = dma_fence_get(updates); - } + (*id)->flushed_updates = updates; job->vm_needs_flush = needs_flush; return 0; } @@ -346,7 +322,7 @@ static int amdgpu_vmid_grab_used(struct amdgpu_vm *vm, unsigned vmhub = ring->funcs->vmhub; struct amdgpu_vmid_mgr *id_mgr = &adev->vm_manager.id_mgr[vmhub]; uint64_t fence_context = adev->fence_context + ring->idx; - struct dma_fence *updates = sync->last_vm_update; + uint64_t updates = amdgpu_vm_tlb_seq(vm); int r; job->vm_needs_flush = vm->use_cpu_for_update; @@ -354,7 +330,6 @@ static int amdgpu_vmid_grab_used(struct amdgpu_vm *vm, /* Check if we can use a VMID already assigned to this VM */ list_for_each_entry_reverse((*id), &id_mgr->ids_lru, list) { bool needs_flush = vm->use_cpu_for_update; - struct dma_fence *flushed; /* Check all the prerequisites to using this VMID */ if ((*id)->owner != vm->immediate.fence_context) @@ -368,8 +343,7 @@ static int amdgpu_vmid_grab_used(struct amdgpu_vm *vm, !dma_fence_is_signaled((*id)->last_flush))) needs_flush = true; - flushed = (*id)->flushed_updates; - if (updates && (!flushed || dma_fence_is_later(updates, flushed))) + if ((*id)->flushed_updates < updates) needs_flush = true; if (needs_flush && !adev->vm_manager.concurrent_flush) @@ -382,11 +356,7 @@ static int amdgpu_vmid_grab_used(struct amdgpu_vm *vm, if (r) return r; - if (updates && (!flushed || dma_fence_is_later(updates, flushed))) { - dma_fence_put((*id)->flushed_updates); - (*id)->flushed_updates = dma_fence_get(updates); - } - + (*id)->flushed_updates = updates; job->vm_needs_flush |= needs_flush; return 0; } @@ -432,8 +402,6 @@ int amdgpu_vmid_grab(struct amdgpu_vm *vm, struct amdgpu_ring *ring, goto error; if (!id) { - struct dma_fence *updates = sync->last_vm_update; - /* Still no ID to use? Then use the idle one found earlier */ id = idle; @@ -442,8 +410,7 @@ int amdgpu_vmid_grab(struct amdgpu_vm *vm, struct amdgpu_ring *ring, if (r) goto error; - dma_fence_put(id->flushed_updates); - id->flushed_updates = dma_fence_get(updates); + id->flushed_updates = amdgpu_vm_tlb_seq(vm); job->vm_needs_flush = true; } @@ -610,7 +577,6 @@ void amdgpu_vmid_mgr_fini(struct amdgpu_device *adev) struct amdgpu_vmid *id = &id_mgr->ids[j]; amdgpu_sync_free(&id->active); - dma_fence_put(id->flushed_updates); dma_fence_put(id->last_flush); dma_fence_put(id->pasid_mapping); } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ids.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ids.h index 0c3b4fa1f936..06c8a0034fa5 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ids.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ids.h @@ -47,7 +47,7 @@ struct amdgpu_vmid { uint64_t pd_gpu_addr; /* last flushed PD/PT update */ - struct dma_fence *flushed_updates; + uint64_t flushed_updates; uint32_t current_gpu_reset_count; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ih.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ih.c index 3df146579ad9..1d5af50331e4 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ih.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ih.c @@ -242,7 +242,7 @@ restart_ih: * @entry: IV entry * * Decodes the interrupt vector at the current rptr - * position and also advance the position for for Vega10 + * position and also advance the position for Vega10 * and later GPUs. */ void amdgpu_ih_decode_iv_helper(struct amdgpu_device *adev, diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_imu.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_imu.h new file mode 100644 index 000000000000..484e936812e4 --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_imu.h @@ -0,0 +1,58 @@ +/* + * Copyright 2021 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#ifndef __AMDGPU_IMU_H__ +#define __AMDGPU_IMU_H__ + +enum imu_work_mode { + DEBUG_MODE, + MISSION_MODE +}; + +struct amdgpu_imu_funcs { + int (*init_microcode)(struct amdgpu_device *adev); + int (*load_microcode)(struct amdgpu_device *adev); + void (*setup_imu)(struct amdgpu_device *adev); + int (*start_imu)(struct amdgpu_device *adev); + void (*program_rlc_ram)(struct amdgpu_device *adev); + int (*wait_for_reset_status)(struct amdgpu_device *adev); +}; + +struct imu_rlc_ram_golden { + u32 hwip; + u32 instance; + u32 segment; + u32 reg; + u32 data; + u32 addr_mask; +}; + +#define IMU_RLC_RAM_GOLDEN_VALUE(ip, inst, reg, data, addr_mask) \ + { ip##_HWIP, inst, reg##_BASE_IDX, reg, data, addr_mask } + +struct amdgpu_imu { + const struct amdgpu_imu_funcs *funcs; + enum imu_work_mode mode; +}; + +#endif diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c index f5cbc2747ac6..89011bae7588 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c @@ -193,20 +193,7 @@ static irqreturn_t amdgpu_irq_handler(int irq, void *arg) if (ret == IRQ_HANDLED) pm_runtime_mark_last_busy(dev->dev); - /* For the hardware that cannot enable bif ring for both ras_controller_irq - * and ras_err_evnet_athub_irq ih cookies, the driver has to poll status - * register to check whether the interrupt is triggered or not, and properly - * ack the interrupt if it is there - */ - if (amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__PCIE_BIF)) { - if (adev->nbio.ras_funcs && - adev->nbio.ras_funcs->handle_ras_controller_intr_no_bifring) - adev->nbio.ras_funcs->handle_ras_controller_intr_no_bifring(adev); - - if (adev->nbio.ras_funcs && - adev->nbio.ras_funcs->handle_ras_err_event_athub_intr_no_bifring) - adev->nbio.ras_funcs->handle_ras_err_event_athub_intr_no_bifring(adev); - } + amdgpu_ras_interrupt_fatal_error_handler(adev); return ret; } @@ -333,6 +320,7 @@ int amdgpu_irq_init(struct amdgpu_device *adev) if (!amdgpu_device_has_dc_support(adev)) { if (!adev->enable_virtual_display) /* Disable vblank IRQs aggressively for power-saving */ + /* XXX: can this be enabled for DC? */ adev_to_drm(adev)->vblank_disable_immediate = true; r = drm_vblank_init(adev_to_drm(adev), adev->mode_info.num_crtc); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c index bfc47bea23db..cd968e781077 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c @@ -29,6 +29,7 @@ #include "amdgpu.h" #include "amdgpu_trace.h" +#include "amdgpu_reset.h" static enum drm_gpu_sched_stat amdgpu_job_timedout(struct drm_sched_job *s_job) { @@ -37,6 +38,7 @@ static enum drm_gpu_sched_stat amdgpu_job_timedout(struct drm_sched_job *s_job) struct amdgpu_task_info ti; struct amdgpu_device *adev = ring->adev; int idx; + int r; if (!drm_dev_enter(adev_to_drm(adev), &idx)) { DRM_INFO("%s - device unplugged skipping recovery on scheduler:%s", @@ -47,6 +49,7 @@ static enum drm_gpu_sched_stat amdgpu_job_timedout(struct drm_sched_job *s_job) } memset(&ti, 0, sizeof(struct amdgpu_task_info)); + adev->job_hang = true; if (amdgpu_gpu_recovery && amdgpu_ring_soft_recovery(ring, job->vmid, s_job->s_fence->parent)) { @@ -63,7 +66,16 @@ static enum drm_gpu_sched_stat amdgpu_job_timedout(struct drm_sched_job *s_job) ti.process_name, ti.tgid, ti.task_name, ti.pid); if (amdgpu_device_should_recover_gpu(ring->adev)) { - amdgpu_device_gpu_recover(ring->adev, job); + struct amdgpu_reset_context reset_context; + memset(&reset_context, 0, sizeof(reset_context)); + + reset_context.method = AMD_RESET_METHOD_NONE; + reset_context.reset_req_dev = adev; + clear_bit(AMDGPU_NEED_FULL_RESET, &reset_context.flags); + + r = amdgpu_device_gpu_recover(ring->adev, job, &reset_context); + if (r) + DRM_ERROR("GPU Recovery Failed: %d\n", r); } else { drm_sched_suspend_timeout(&ring->sched); if (amdgpu_sriov_vf(adev)) @@ -71,6 +83,7 @@ static enum drm_gpu_sched_stat amdgpu_job_timedout(struct drm_sched_job *s_job) } exit: + adev->job_hang = false; drm_dev_exit(idx); return DRM_GPU_SCHED_STAT_NOMINAL; } @@ -78,14 +91,10 @@ exit: int amdgpu_job_alloc(struct amdgpu_device *adev, unsigned num_ibs, struct amdgpu_job **job, struct amdgpu_vm *vm) { - size_t size = sizeof(struct amdgpu_job); - if (num_ibs == 0) return -EINVAL; - size += sizeof(struct amdgpu_ib) * num_ibs; - - *job = kzalloc(size, GFP_KERNEL); + *job = kzalloc(struct_size(*job, ibs, num_ibs), GFP_KERNEL); if (!*job) return -ENOMEM; @@ -95,8 +104,6 @@ int amdgpu_job_alloc(struct amdgpu_device *adev, unsigned num_ibs, */ (*job)->base.sched = &adev->rings[0]->sched; (*job)->vm = vm; - (*job)->ibs = (void *)&(*job)[1]; - (*job)->num_ibs = num_ibs; amdgpu_sync_create(&(*job)->sync); amdgpu_sync_create(&(*job)->sched_sync); @@ -116,6 +123,7 @@ int amdgpu_job_alloc_with_ib(struct amdgpu_device *adev, unsigned size, if (r) return r; + (*job)->num_ibs = 1; r = amdgpu_ib_get(adev, NULL, size, pool_type, &(*job)->ibs[0]); if (r) kfree(*job); @@ -123,20 +131,31 @@ int amdgpu_job_alloc_with_ib(struct amdgpu_device *adev, unsigned size, return r; } +void amdgpu_job_set_resources(struct amdgpu_job *job, struct amdgpu_bo *gds, + struct amdgpu_bo *gws, struct amdgpu_bo *oa) +{ + if (gds) { + job->gds_base = amdgpu_bo_gpu_offset(gds) >> PAGE_SHIFT; + job->gds_size = amdgpu_bo_size(gds) >> PAGE_SHIFT; + } + if (gws) { + job->gws_base = amdgpu_bo_gpu_offset(gws) >> PAGE_SHIFT; + job->gws_size = amdgpu_bo_size(gws) >> PAGE_SHIFT; + } + if (oa) { + job->oa_base = amdgpu_bo_gpu_offset(oa) >> PAGE_SHIFT; + job->oa_size = amdgpu_bo_size(oa) >> PAGE_SHIFT; + } +} + void amdgpu_job_free_resources(struct amdgpu_job *job) { struct amdgpu_ring *ring = to_amdgpu_ring(job->base.sched); struct dma_fence *f; - struct dma_fence *hw_fence; unsigned i; - if (job->hw_fence.ops == NULL) - hw_fence = job->external_hw_fence; - else - hw_fence = &job->hw_fence; - /* use sched fence if available */ - f = job->base.s_fence ? &job->base.s_fence->finished : hw_fence; + f = job->base.s_fence ? &job->base.s_fence->finished : &job->hw_fence; for (i = 0; i < job->num_ibs; ++i) amdgpu_ib_free(ring->adev, &job->ibs[i], f); } @@ -150,11 +169,23 @@ static void amdgpu_job_free_cb(struct drm_sched_job *s_job) amdgpu_sync_free(&job->sync); amdgpu_sync_free(&job->sched_sync); - /* only put the hw fence if has embedded fence */ - if (job->hw_fence.ops != NULL) - dma_fence_put(&job->hw_fence); - else - kfree(job); + dma_fence_put(&job->hw_fence); +} + +void amdgpu_job_set_gang_leader(struct amdgpu_job *job, + struct amdgpu_job *leader) +{ + struct dma_fence *fence = &leader->base.s_fence->scheduled; + + WARN_ON(job->gang_submit); + + /* + * Don't add a reference when we are the gang leader to avoid circle + * dependency. + */ + if (job != leader) + dma_fence_get(fence); + job->gang_submit = fence; } void amdgpu_job_free(struct amdgpu_job *job) @@ -162,12 +193,13 @@ void amdgpu_job_free(struct amdgpu_job *job) amdgpu_job_free_resources(job); amdgpu_sync_free(&job->sync); amdgpu_sync_free(&job->sched_sync); + if (job->gang_submit != &job->base.s_fence->scheduled) + dma_fence_put(job->gang_submit); - /* only put the hw fence if has embedded fence */ - if (job->hw_fence.ops != NULL) - dma_fence_put(&job->hw_fence); - else + if (!job->hw_fence.ops) kfree(job); + else + dma_fence_put(&job->hw_fence); } int amdgpu_job_submit(struct amdgpu_job *job, struct drm_sched_entity *entity, @@ -197,15 +229,12 @@ int amdgpu_job_submit_direct(struct amdgpu_job *job, struct amdgpu_ring *ring, int r; job->base.sched = &ring->sched; - r = amdgpu_ib_schedule(ring, job->num_ibs, job->ibs, NULL, fence); - /* record external_hw_fence for direct submit */ - job->external_hw_fence = dma_fence_get(*fence); + r = amdgpu_ib_schedule(ring, job->num_ibs, job->ibs, job, fence); + if (r) return r; amdgpu_job_free(job); - dma_fence_put(*fence); - return 0; } @@ -235,12 +264,16 @@ static struct dma_fence *amdgpu_job_dependency(struct drm_sched_job *sched_job, fence = amdgpu_sync_get_fence(&job->sync); } + if (!fence && job->gang_submit) + fence = amdgpu_device_switch_gang(ring->adev, job->gang_submit); + return fence; } static struct dma_fence *amdgpu_job_run(struct drm_sched_job *sched_job) { struct amdgpu_ring *ring = to_amdgpu_ring(sched_job->sched); + struct amdgpu_device *adev = ring->adev; struct dma_fence *fence = NULL, *finished; struct amdgpu_job *job; int r = 0; @@ -252,8 +285,10 @@ static struct dma_fence *amdgpu_job_run(struct drm_sched_job *sched_job) trace_amdgpu_sched_run_job(job); - if (job->vram_lost_counter != atomic_read(&ring->adev->vram_lost_counter)) - dma_fence_set_error(finished, -ECANCELED);/* skip IB as well if VRAM lost */ + /* Skip job if VRAM is lost and never resubmit gangs */ + if (job->vram_lost_counter != atomic_read(&adev->vram_lost_counter) || + (job->job_run_counter && job->gang_submit)) + dma_fence_set_error(finished, -ECANCELED); if (finished->error < 0) { DRM_INFO("Skip scheduling IBs!\n"); @@ -264,10 +299,6 @@ static struct dma_fence *amdgpu_job_run(struct drm_sched_job *sched_job) DRM_ERROR("Error scheduling IBs (%d)\n", r); } - if (!job->job_run_counter) - dma_fence_get(fence); - else if (finished->error < 0) - dma_fence_put(&job->hw_fence); job->job_run_counter++; amdgpu_job_free_resources(job); @@ -287,10 +318,6 @@ void amdgpu_job_stop_all_jobs_on_sched(struct drm_gpu_scheduler *sched) /* Signal all jobs not yet scheduled */ for (i = DRM_SCHED_PRIORITY_COUNT - 1; i >= DRM_SCHED_PRIORITY_MIN; i--) { struct drm_sched_rq *rq = &sched->sched_rq[i]; - - if (!rq) - continue; - spin_lock(&rq->lock); list_for_each_entry(s_entity, &rq->entities, list) { while ((s_job = to_drm_sched_job(spsc_queue_pop(&s_entity->job_queue)))) { diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_job.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_job.h index 9e65730193b8..ab7b150e5d50 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_job.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_job.h @@ -23,6 +23,10 @@ #ifndef __AMDGPU_JOB_H__ #define __AMDGPU_JOB_H__ +#include <drm/gpu_scheduler.h> +#include "amdgpu_sync.h" +#include "amdgpu_ring.h" + /* bit set means command submit involves a preamble IB */ #define AMDGPU_PREAMBLE_IB_PRESENT (1 << 0) /* bit set means preamble IB is first presented in belonging context */ @@ -45,12 +49,10 @@ struct amdgpu_job { struct amdgpu_vm *vm; struct amdgpu_sync sync; struct amdgpu_sync sched_sync; - struct amdgpu_ib *ibs; struct dma_fence hw_fence; - struct dma_fence *external_hw_fence; + struct dma_fence *gang_submit; uint32_t preamble_status; uint32_t preemption_status; - uint32_t num_ibs; bool vm_needs_flush; uint64_t vm_pd_addr; unsigned vmid; @@ -66,13 +68,25 @@ struct amdgpu_job { /* job_run_counter >= 1 means a resubmit job */ uint32_t job_run_counter; + + uint32_t num_ibs; + struct amdgpu_ib ibs[]; }; +static inline struct amdgpu_ring *amdgpu_job_ring(struct amdgpu_job *job) +{ + return to_amdgpu_ring(job->base.entity->rq->sched); +} + int amdgpu_job_alloc(struct amdgpu_device *adev, unsigned num_ibs, struct amdgpu_job **job, struct amdgpu_vm *vm); int amdgpu_job_alloc_with_ib(struct amdgpu_device *adev, unsigned size, enum amdgpu_ib_pool_type pool, struct amdgpu_job **job); +void amdgpu_job_set_resources(struct amdgpu_job *job, struct amdgpu_bo *gds, + struct amdgpu_bo *gws, struct amdgpu_bo *oa); void amdgpu_job_free_resources(struct amdgpu_job *job); +void amdgpu_job_set_gang_leader(struct amdgpu_job *job, + struct amdgpu_job *leader); void amdgpu_job_free(struct amdgpu_job *job); int amdgpu_job_submit(struct amdgpu_job *job, struct drm_sched_entity *entity, void *owner, struct dma_fence **f); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_jpeg.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_jpeg.c index 9342aa23ebd2..518eb0e40d32 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_jpeg.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_jpeg.c @@ -216,3 +216,21 @@ int amdgpu_jpeg_dec_ring_test_ib(struct amdgpu_ring *ring, long timeout) error: return r; } + +int amdgpu_jpeg_process_poison_irq(struct amdgpu_device *adev, + struct amdgpu_irq_src *source, + struct amdgpu_iv_entry *entry) +{ + struct ras_common_if *ras_if = adev->jpeg.ras_if; + struct ras_dispatch_if ih_data = { + .entry = entry, + }; + + if (!ras_if) + return 0; + + ih_data.head = *ras_if; + amdgpu_ras_interrupt_dispatch(adev, &ih_data); + + return 0; +} diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_jpeg.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_jpeg.h index 55fbff2be761..635dca59a70a 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_jpeg.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_jpeg.h @@ -24,6 +24,8 @@ #ifndef __AMDGPU_JPEG_H__ #define __AMDGPU_JPEG_H__ +#include "amdgpu_ras.h" + #define AMDGPU_MAX_JPEG_INSTANCES 2 #define AMDGPU_JPEG_HARVEST_JPEG0 (1 << 0) @@ -39,6 +41,10 @@ struct amdgpu_jpeg_inst { struct amdgpu_jpeg_reg external; }; +struct amdgpu_jpeg_ras { + struct amdgpu_ras_block_object ras_block; +}; + struct amdgpu_jpeg { uint8_t num_jpeg_inst; struct amdgpu_jpeg_inst inst[AMDGPU_MAX_JPEG_INSTANCES]; @@ -48,6 +54,8 @@ struct amdgpu_jpeg { enum amd_powergating_state cur_state; struct mutex jpeg_pg_lock; atomic_t total_submission_cnt; + struct ras_common_if *ras_if; + struct amdgpu_jpeg_ras *ras; }; int amdgpu_jpeg_sw_init(struct amdgpu_device *adev); @@ -61,4 +69,8 @@ void amdgpu_jpeg_ring_end_use(struct amdgpu_ring *ring); int amdgpu_jpeg_dec_ring_test_ring(struct amdgpu_ring *ring); int amdgpu_jpeg_dec_ring_test_ib(struct amdgpu_ring *ring, long timeout); +int amdgpu_jpeg_process_poison_irq(struct amdgpu_device *adev, + struct amdgpu_irq_src *source, + struct amdgpu_iv_entry *entry); + #endif /*__AMDGPU_JPEG_H__*/ diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c index 09ad17944eb2..4e42dcb1950f 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c @@ -87,11 +87,6 @@ void amdgpu_driver_unload_kms(struct drm_device *dev) if (adev->rmmio == NULL) return; - if (adev->runpm) { - pm_runtime_get_sync(dev->dev); - pm_runtime_forbid(dev->dev); - } - if (amdgpu_acpi_smart_shift_update(dev, AMDGPU_SS_DRV_UNLOAD)) DRM_WARN("smart shift update failed\n"); @@ -124,22 +119,6 @@ void amdgpu_register_gpu_instance(struct amdgpu_device *adev) mutex_unlock(&mgpu_info.mutex); } -static void amdgpu_get_audio_func(struct amdgpu_device *adev) -{ - struct pci_dev *p = NULL; - - p = pci_get_domain_bus_and_slot(pci_domain_nr(adev->pdev->bus), - adev->pdev->bus->number, 1); - if (p) { - pm_runtime_get_sync(&p->dev); - - pm_runtime_mark_last_busy(&p->dev); - pm_runtime_put_autosuspend(&p->dev); - - pci_dev_put(p); - } -} - /** * amdgpu_driver_load_kms - Main load function for KMS. * @@ -152,21 +131,10 @@ static void amdgpu_get_audio_func(struct amdgpu_device *adev) int amdgpu_driver_load_kms(struct amdgpu_device *adev, unsigned long flags) { struct drm_device *dev; - struct pci_dev *parent; int r, acpi_status; dev = adev_to_drm(adev); - if (amdgpu_has_atpx() && - (amdgpu_is_atpx_hybrid() || - amdgpu_has_atpx_dgpu_power_cntl()) && - ((flags & AMD_IS_APU) == 0) && - !pci_is_thunderbolt_attached(to_pci_dev(dev->dev))) - flags |= AMD_IS_PX; - - parent = pci_upstream_bridge(adev->pdev); - adev->has_pr3 = parent ? pci_pr3_present(parent) : false; - /* amdgpu_device_init should report only fatal error * like memory allocation failure or iomapping failure, * or memory manager initialization failure, it must @@ -179,40 +147,36 @@ int amdgpu_driver_load_kms(struct amdgpu_device *adev, unsigned long flags) goto out; } + adev->pm.rpm_mode = AMDGPU_RUNPM_NONE; if (amdgpu_device_supports_px(dev) && - (amdgpu_runtime_pm != 0)) { /* enable runpm by default for atpx */ - adev->runpm = true; + (amdgpu_runtime_pm != 0)) { /* enable PX as runtime mode */ + adev->pm.rpm_mode = AMDGPU_RUNPM_PX; dev_info(adev->dev, "Using ATPX for runtime pm\n"); } else if (amdgpu_device_supports_boco(dev) && - (amdgpu_runtime_pm != 0)) { /* enable runpm by default for boco */ - adev->runpm = true; + (amdgpu_runtime_pm != 0)) { /* enable boco as runtime mode */ + adev->pm.rpm_mode = AMDGPU_RUNPM_BOCO; dev_info(adev->dev, "Using BOCO for runtime pm\n"); } else if (amdgpu_device_supports_baco(dev) && (amdgpu_runtime_pm != 0)) { switch (adev->asic_type) { case CHIP_VEGA20: case CHIP_ARCTURUS: - /* enable runpm if runpm=1 */ + /* enable BACO as runpm mode if runpm=1 */ if (amdgpu_runtime_pm > 0) - adev->runpm = true; + adev->pm.rpm_mode = AMDGPU_RUNPM_BACO; break; case CHIP_VEGA10: - /* turn runpm on if noretry=0 */ + /* enable BACO as runpm mode if noretry=0 */ if (!adev->gmc.noretry) - adev->runpm = true; + adev->pm.rpm_mode = AMDGPU_RUNPM_BACO; break; default: - /* enable runpm on CI+ */ - adev->runpm = true; + /* enable BACO as runpm mode on CI+ */ + adev->pm.rpm_mode = AMDGPU_RUNPM_BACO; break; } - /* XXX: disable runtime pm if we are the primary adapter - * to avoid displays being re-enabled after DPMS. - * This needs to be sorted out and fixed properly. - */ - if (adev->is_fw_fb) - adev->runpm = false; - if (adev->runpm) + + if (adev->pm.rpm_mode == AMDGPU_RUNPM_BACO) dev_info(adev->dev, "Using BACO for runtime pm\n"); } @@ -224,58 +188,12 @@ int amdgpu_driver_load_kms(struct amdgpu_device *adev, unsigned long flags) if (acpi_status) dev_dbg(dev->dev, "Error during ACPI methods call\n"); - if (adev->runpm) { - /* only need to skip on ATPX */ - if (amdgpu_device_supports_px(dev)) - dev_pm_set_driver_flags(dev->dev, DPM_FLAG_NO_DIRECT_COMPLETE); - /* we want direct complete for BOCO */ - if (amdgpu_device_supports_boco(dev)) - dev_pm_set_driver_flags(dev->dev, DPM_FLAG_SMART_PREPARE | - DPM_FLAG_SMART_SUSPEND | - DPM_FLAG_MAY_SKIP_RESUME); - pm_runtime_use_autosuspend(dev->dev); - pm_runtime_set_autosuspend_delay(dev->dev, 5000); - - pm_runtime_allow(dev->dev); - - pm_runtime_mark_last_busy(dev->dev); - pm_runtime_put_autosuspend(dev->dev); - - /* - * For runpm implemented via BACO, PMFW will handle the - * timing for BACO in and out: - * - put ASIC into BACO state only when both video and - * audio functions are in D3 state. - * - pull ASIC out of BACO state when either video or - * audio function is in D0 state. - * Also, at startup, PMFW assumes both functions are in - * D0 state. - * - * So if snd driver was loaded prior to amdgpu driver - * and audio function was put into D3 state, there will - * be no PMFW-aware D-state transition(D0->D3) on runpm - * suspend. Thus the BACO will be not correctly kicked in. - * - * Via amdgpu_get_audio_func(), the audio dev is put - * into D0 state. Then there will be a PMFW-aware D-state - * transition(D0->D3) on runpm suspend. - */ - if (amdgpu_device_supports_baco(dev) && - !(adev->flags & AMD_IS_APU) && - (adev->asic_type >= CHIP_NAVI10)) - amdgpu_get_audio_func(adev); - } - if (amdgpu_acpi_smart_shift_update(dev, AMDGPU_SS_DRV_LOAD)) DRM_WARN("smart shift update failed\n"); out: - if (r) { - /* balance pm_runtime_get_sync in amdgpu_driver_unload_kms */ - if (adev->rmmio && adev->runpm) - pm_runtime_put_noidle(dev->dev); + if (r) amdgpu_driver_unload_kms(dev); - } return r; } @@ -329,6 +247,14 @@ static int amdgpu_firmware_info(struct drm_amdgpu_info_firmware *fw_info, fw_info->ver = adev->gfx.rlc_srls_fw_version; fw_info->feature = adev->gfx.rlc_srls_feature_version; break; + case AMDGPU_INFO_FW_GFX_RLCP: + fw_info->ver = adev->gfx.rlcp_ucode_version; + fw_info->feature = adev->gfx.rlcp_ucode_feature_version; + break; + case AMDGPU_INFO_FW_GFX_RLCV: + fw_info->ver = adev->gfx.rlcv_ucode_version; + fw_info->feature = adev->gfx.rlcv_ucode_feature_version; + break; case AMDGPU_INFO_FW_GFX_MEC: if (query_fw->index == 0) { fw_info->ver = adev->gfx.mec_fw_version; @@ -406,6 +332,24 @@ static int amdgpu_firmware_info(struct drm_amdgpu_info_firmware *fw_info, fw_info->ver = adev->psp.toc.fw_version; fw_info->feature = adev->psp.toc.feature_version; break; + case AMDGPU_INFO_FW_CAP: + fw_info->ver = adev->psp.cap_fw_version; + fw_info->feature = adev->psp.cap_feature_version; + break; + case AMDGPU_INFO_FW_MES_KIQ: + fw_info->ver = adev->mes.kiq_version & AMDGPU_MES_VERSION_MASK; + fw_info->feature = (adev->mes.kiq_version & AMDGPU_MES_FEAT_VERSION_MASK) + >> AMDGPU_MES_FEAT_VERSION_SHIFT; + break; + case AMDGPU_INFO_FW_MES: + fw_info->ver = adev->mes.sched_version & AMDGPU_MES_VERSION_MASK; + fw_info->feature = (adev->mes.sched_version & AMDGPU_MES_FEAT_VERSION_MASK) + >> AMDGPU_MES_FEAT_VERSION_SHIFT; + break; + case AMDGPU_INFO_FW_IMU: + fw_info->ver = adev->gfx.imu_fw_version; + fw_info->feature = 0; + break; default: return -EINVAL; } @@ -539,6 +483,30 @@ static int amdgpu_hw_ip_info(struct amdgpu_device *adev, result->hw_ip_version_major = adev->ip_blocks[i].version->major; result->hw_ip_version_minor = adev->ip_blocks[i].version->minor; + + if (adev->asic_type >= CHIP_VEGA10) { + switch (type) { + case AMD_IP_BLOCK_TYPE_GFX: + result->ip_discovery_version = adev->ip_versions[GC_HWIP][0]; + break; + case AMD_IP_BLOCK_TYPE_SDMA: + result->ip_discovery_version = adev->ip_versions[SDMA0_HWIP][0]; + break; + case AMD_IP_BLOCK_TYPE_UVD: + case AMD_IP_BLOCK_TYPE_VCN: + case AMD_IP_BLOCK_TYPE_JPEG: + result->ip_discovery_version = adev->ip_versions[UVD_HWIP][0]; + break; + case AMD_IP_BLOCK_TYPE_VCE: + result->ip_discovery_version = adev->ip_versions[VCE_HWIP][0]; + break; + default: + result->ip_discovery_version = 0; + break; + } + } else { + result->ip_discovery_version = 0; + } result->capabilities_flags = 0; result->available_rings = (1 << num_rings) - 1; result->ib_start_alignment = ib_start_alignment; @@ -678,13 +646,13 @@ int amdgpu_info_ioctl(struct drm_device *dev, void *data, struct drm_file *filp) ui64 = atomic64_read(&adev->num_vram_cpu_page_faults); return copy_to_user(out, &ui64, min(size, 8u)) ? -EFAULT : 0; case AMDGPU_INFO_VRAM_USAGE: - ui64 = amdgpu_vram_mgr_usage(ttm_manager_type(&adev->mman.bdev, TTM_PL_VRAM)); + ui64 = ttm_resource_manager_usage(&adev->mman.vram_mgr.manager); return copy_to_user(out, &ui64, min(size, 8u)) ? -EFAULT : 0; case AMDGPU_INFO_VIS_VRAM_USAGE: - ui64 = amdgpu_vram_mgr_vis_usage(ttm_manager_type(&adev->mman.bdev, TTM_PL_VRAM)); + ui64 = amdgpu_vram_mgr_vis_usage(&adev->mman.vram_mgr); return copy_to_user(out, &ui64, min(size, 8u)) ? -EFAULT : 0; case AMDGPU_INFO_GTT_USAGE: - ui64 = amdgpu_gtt_mgr_usage(ttm_manager_type(&adev->mman.bdev, TTM_PL_TT)); + ui64 = ttm_resource_manager_usage(&adev->mman.gtt_mgr.manager); return copy_to_user(out, &ui64, min(size, 8u)) ? -EFAULT : 0; case AMDGPU_INFO_GDS_CONFIG: { struct drm_amdgpu_info_gds gds_info; @@ -708,24 +676,24 @@ int amdgpu_info_ioctl(struct drm_device *dev, void *data, struct drm_file *filp) atomic64_read(&adev->visible_pin_size), vram_gtt.vram_size); vram_gtt.gtt_size = ttm_manager_type(&adev->mman.bdev, TTM_PL_TT)->size; - vram_gtt.gtt_size *= PAGE_SIZE; vram_gtt.gtt_size -= atomic64_read(&adev->gart_pin_size); return copy_to_user(out, &vram_gtt, min((size_t)size, sizeof(vram_gtt))) ? -EFAULT : 0; } case AMDGPU_INFO_MEMORY: { struct drm_amdgpu_memory_info mem; - struct ttm_resource_manager *vram_man = - ttm_manager_type(&adev->mman.bdev, TTM_PL_VRAM); struct ttm_resource_manager *gtt_man = - ttm_manager_type(&adev->mman.bdev, TTM_PL_TT); + &adev->mman.gtt_mgr.manager; + struct ttm_resource_manager *vram_man = + &adev->mman.vram_mgr.manager; + memset(&mem, 0, sizeof(mem)); mem.vram.total_heap_size = adev->gmc.real_vram_size; mem.vram.usable_heap_size = adev->gmc.real_vram_size - atomic64_read(&adev->vram_pin_size) - AMDGPU_VM_RESERVED_VRAM; mem.vram.heap_usage = - amdgpu_vram_mgr_usage(vram_man); + ttm_resource_manager_usage(vram_man); mem.vram.max_allocation = mem.vram.usable_heap_size * 3 / 4; mem.cpu_accessible_vram.total_heap_size = @@ -735,16 +703,14 @@ int amdgpu_info_ioctl(struct drm_device *dev, void *data, struct drm_file *filp) atomic64_read(&adev->visible_pin_size), mem.vram.usable_heap_size); mem.cpu_accessible_vram.heap_usage = - amdgpu_vram_mgr_vis_usage(vram_man); + amdgpu_vram_mgr_vis_usage(&adev->mman.vram_mgr); mem.cpu_accessible_vram.max_allocation = mem.cpu_accessible_vram.usable_heap_size * 3 / 4; mem.gtt.total_heap_size = gtt_man->size; - mem.gtt.total_heap_size *= PAGE_SIZE; mem.gtt.usable_heap_size = mem.gtt.total_heap_size - atomic64_read(&adev->gart_pin_size); - mem.gtt.heap_usage = - amdgpu_gtt_mgr_usage(gtt_man); + mem.gtt.heap_usage = ttm_resource_manager_usage(gtt_man); mem.gtt.max_allocation = mem.gtt.usable_heap_size * 3 / 4; return copy_to_user(out, &mem, @@ -1216,9 +1182,9 @@ int amdgpu_driver_open_kms(struct drm_device *dev, struct drm_file *file_priv) } mutex_init(&fpriv->bo_list_lock); - idr_init(&fpriv->bo_list_handles); + idr_init_base(&fpriv->bo_list_handles, 1); - amdgpu_ctx_mgr_init(&fpriv->ctx_mgr); + amdgpu_ctx_mgr_init(&fpriv->ctx_mgr, adev); file_priv->driver_priv = fpriv; goto out_suspend; @@ -1270,18 +1236,20 @@ void amdgpu_driver_postclose_kms(struct drm_device *dev, if (amdgpu_device_ip_get_ip_block(adev, AMD_IP_BLOCK_TYPE_VCE) != NULL) amdgpu_vce_free_handles(adev, file_priv); - amdgpu_vm_bo_rmv(adev, fpriv->prt_va); - if (amdgpu_mcbp || amdgpu_sriov_vf(adev)) { /* TODO: how to handle reserve failure */ BUG_ON(amdgpu_bo_reserve(adev->virt.csa_obj, true)); - amdgpu_vm_bo_rmv(adev, fpriv->csa_va); + amdgpu_vm_bo_del(adev, fpriv->csa_va); fpriv->csa_va = NULL; amdgpu_bo_unreserve(adev->virt.csa_obj); } pasid = fpriv->vm.pasid; pd = amdgpu_bo_ref(fpriv->vm.root.bo); + if (!WARN_ON(amdgpu_bo_reserve(pd, true))) { + amdgpu_vm_bo_del(adev, fpriv->prt_va); + amdgpu_bo_unreserve(pd); + } amdgpu_ctx_mgr_fini(&fpriv->ctx_mgr); amdgpu_vm_fini(adev, &fpriv->vm); @@ -1429,8 +1397,7 @@ static int amdgpu_debugfs_firmware_info_show(struct seq_file *m, void *unused) struct drm_amdgpu_info_firmware fw_info; struct drm_amdgpu_query_fw query_fw; struct atom_context *ctx = adev->mode_info.atom_context; - uint8_t smu_minor, smu_debug; - uint16_t smu_major; + uint8_t smu_program, smu_major, smu_minor, smu_debug; int ret, i; static const char *ta_fw_name[TA_FW_TYPE_MAX_INDEX] = { @@ -1524,6 +1491,22 @@ static int amdgpu_debugfs_firmware_info_show(struct seq_file *m, void *unused) seq_printf(m, "RLC SRLS feature version: %u, firmware version: 0x%08x\n", fw_info.feature, fw_info.ver); + /* RLCP */ + query_fw.fw_type = AMDGPU_INFO_FW_GFX_RLCP; + ret = amdgpu_firmware_info(&fw_info, &query_fw, adev); + if (ret) + return ret; + seq_printf(m, "RLCP feature version: %u, firmware version: 0x%08x\n", + fw_info.feature, fw_info.ver); + + /* RLCV */ + query_fw.fw_type = AMDGPU_INFO_FW_GFX_RLCV; + ret = amdgpu_firmware_info(&fw_info, &query_fw, adev); + if (ret) + return ret; + seq_printf(m, "RLCV feature version: %u, firmware version: 0x%08x\n", + fw_info.feature, fw_info.ver); + /* MEC */ query_fw.fw_type = AMDGPU_INFO_FW_GFX_MEC; query_fw.index = 0; @@ -1543,6 +1526,15 @@ static int amdgpu_debugfs_firmware_info_show(struct seq_file *m, void *unused) fw_info.feature, fw_info.ver); } + /* IMU */ + query_fw.fw_type = AMDGPU_INFO_FW_IMU; + query_fw.index = 0; + ret = amdgpu_firmware_info(&fw_info, &query_fw, adev); + if (ret) + return ret; + seq_printf(m, "IMU feature version: %u, firmware version: 0x%08x\n", + fw_info.feature, fw_info.ver); + /* PSP SOS */ query_fw.fw_type = AMDGPU_INFO_FW_SOS; ret = amdgpu_firmware_info(&fw_info, &query_fw, adev); @@ -1576,11 +1568,12 @@ static int amdgpu_debugfs_firmware_info_show(struct seq_file *m, void *unused) ret = amdgpu_firmware_info(&fw_info, &query_fw, adev); if (ret) return ret; - smu_major = (fw_info.ver >> 16) & 0xffff; + smu_program = (fw_info.ver >> 24) & 0xff; + smu_major = (fw_info.ver >> 16) & 0xff; smu_minor = (fw_info.ver >> 8) & 0xff; smu_debug = (fw_info.ver >> 0) & 0xff; - seq_printf(m, "SMC feature version: %u, firmware version: 0x%08x (%d.%d.%d)\n", - fw_info.feature, fw_info.ver, smu_major, smu_minor, smu_debug); + seq_printf(m, "SMC feature version: %u, program: %d, firmware version: 0x%08x (%d.%d.%d)\n", + fw_info.feature, smu_program, fw_info.ver, smu_major, smu_minor, smu_debug); /* SDMA */ query_fw.fw_type = AMDGPU_INFO_FW_SDMA; @@ -1625,6 +1618,32 @@ static int amdgpu_debugfs_firmware_info_show(struct seq_file *m, void *unused) seq_printf(m, "TOC feature version: %u, firmware version: 0x%08x\n", fw_info.feature, fw_info.ver); + /* CAP */ + if (adev->psp.cap_fw) { + query_fw.fw_type = AMDGPU_INFO_FW_CAP; + ret = amdgpu_firmware_info(&fw_info, &query_fw, adev); + if (ret) + return ret; + seq_printf(m, "CAP feature version: %u, firmware version: 0x%08x\n", + fw_info.feature, fw_info.ver); + } + + /* MES_KIQ */ + query_fw.fw_type = AMDGPU_INFO_FW_MES_KIQ; + ret = amdgpu_firmware_info(&fw_info, &query_fw, adev); + if (ret) + return ret; + seq_printf(m, "MES_KIQ feature version: %u, firmware version: 0x%08x\n", + fw_info.feature, fw_info.ver); + + /* MES */ + query_fw.fw_type = AMDGPU_INFO_FW_MES; + ret = amdgpu_firmware_info(&fw_info, &query_fw, adev); + if (ret) + return ret; + seq_printf(m, "MES feature version: %u, firmware version: 0x%08x\n", + fw_info.feature, fw_info.ver); + seq_printf(m, "VBIOS version: %s\n", ctx->vbios_version); return 0; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_lsdma.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_lsdma.c new file mode 100644 index 000000000000..4d1d4994ea3f --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_lsdma.c @@ -0,0 +1,91 @@ +/* + * Copyright 2022 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#include "amdgpu.h" +#include "amdgpu_lsdma.h" + +#define AMDGPU_LSDMA_MAX_SIZE 0x2000000ULL + +int amdgpu_lsdma_wait_for(struct amdgpu_device *adev, + uint32_t reg_index, uint32_t reg_val, + uint32_t mask) +{ + uint32_t val; + int i; + + for (i = 0; i < adev->usec_timeout; i++) { + val = RREG32(reg_index); + if ((val & mask) == reg_val) + return 0; + udelay(1); + } + + return -ETIME; +} + +int amdgpu_lsdma_copy_mem(struct amdgpu_device *adev, + uint64_t src_addr, + uint64_t dst_addr, + uint64_t mem_size) +{ + int ret; + + if (mem_size == 0) + return -EINVAL; + + while (mem_size > 0) { + uint64_t current_copy_size = min(mem_size, AMDGPU_LSDMA_MAX_SIZE); + + ret = adev->lsdma.funcs->copy_mem(adev, src_addr, dst_addr, current_copy_size); + if (ret) + return ret; + src_addr += current_copy_size; + dst_addr += current_copy_size; + mem_size -= current_copy_size; + } + + return 0; +} + +int amdgpu_lsdma_fill_mem(struct amdgpu_device *adev, + uint64_t dst_addr, + uint32_t data, + uint64_t mem_size) +{ + int ret; + + if (mem_size == 0) + return -EINVAL; + + while (mem_size > 0) { + uint64_t current_fill_size = min(mem_size, AMDGPU_LSDMA_MAX_SIZE); + + ret = adev->lsdma.funcs->fill_mem(adev, dst_addr, data, current_fill_size); + if (ret) + return ret; + dst_addr += current_fill_size; + mem_size -= current_fill_size; + } + + return 0; +} diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_lsdma.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_lsdma.h new file mode 100644 index 000000000000..c61ba58c5ee0 --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_lsdma.h @@ -0,0 +1,46 @@ +/* + * Copyright 2022 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#ifndef __AMDGPU_LSDMA_H__ +#define __AMDGPU_LSDMA_H__ + +struct amdgpu_lsdma { + const struct amdgpu_lsdma_funcs *funcs; +}; + +struct amdgpu_lsdma_funcs { + int (*copy_mem)(struct amdgpu_device *adev, uint64_t src_addr, + uint64_t dst_addr, uint64_t size); + int (*fill_mem)(struct amdgpu_device *adev, uint64_t dst_addr, + uint32_t data, uint64_t size); + void (*update_memory_power_gating)(struct amdgpu_device *adev, bool enable); +}; + +int amdgpu_lsdma_copy_mem(struct amdgpu_device *adev, uint64_t src_addr, + uint64_t dst_addr, uint64_t mem_size); +int amdgpu_lsdma_fill_mem(struct amdgpu_device *adev, uint64_t dst_addr, + uint32_t data, uint64_t mem_size); +int amdgpu_lsdma_wait_for(struct amdgpu_device *adev, uint32_t reg_index, + uint32_t reg_val, uint32_t mask); + +#endif diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_mca.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_mca.c index ce538f4819f9..51c2a82e2fa4 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_mca.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_mca.c @@ -70,48 +70,3 @@ void amdgpu_mca_query_ras_error_count(struct amdgpu_device *adev, amdgpu_mca_reset_error_count(adev, mc_status_addr); } - -int amdgpu_mca_ras_late_init(struct amdgpu_device *adev, - struct amdgpu_mca_ras *mca_dev) -{ - int r; - struct ras_ih_if ih_info = { - .cb = NULL, - }; - struct ras_fs_if fs_info = { - .sysfs_name = mca_dev->ras_funcs->sysfs_name, - }; - - if (!mca_dev->ras_if) { - mca_dev->ras_if = kmalloc(sizeof(struct ras_common_if), GFP_KERNEL); - if (!mca_dev->ras_if) - return -ENOMEM; - mca_dev->ras_if->block = mca_dev->ras_funcs->ras_block; - mca_dev->ras_if->sub_block_index = mca_dev->ras_funcs->ras_sub_block; - mca_dev->ras_if->type = AMDGPU_RAS_ERROR__MULTI_UNCORRECTABLE; - } - ih_info.head = fs_info.head = *mca_dev->ras_if; - r = amdgpu_ras_late_init(adev, mca_dev->ras_if, - &fs_info, &ih_info); - if (r || !amdgpu_ras_is_supported(adev, mca_dev->ras_if->block)) { - kfree(mca_dev->ras_if); - mca_dev->ras_if = NULL; - } - - return r; -} - -void amdgpu_mca_ras_fini(struct amdgpu_device *adev, - struct amdgpu_mca_ras *mca_dev) -{ - struct ras_ih_if ih_info = { - .cb = NULL, - }; - - if (!mca_dev->ras_if) - return; - - amdgpu_ras_late_fini(adev, mca_dev->ras_if, &ih_info); - kfree(mca_dev->ras_if); - mca_dev->ras_if = NULL; -}
\ No newline at end of file diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_mca.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_mca.h index c74bc7177066..7ce16d16e34b 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_mca.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_mca.h @@ -21,21 +21,13 @@ #ifndef __AMDGPU_MCA_H__ #define __AMDGPU_MCA_H__ -struct amdgpu_mca_ras_funcs { - int (*ras_late_init)(struct amdgpu_device *adev); - void (*ras_fini)(struct amdgpu_device *adev); - void (*query_ras_error_count)(struct amdgpu_device *adev, - void *ras_error_status); - void (*query_ras_error_address)(struct amdgpu_device *adev, - void *ras_error_status); - uint32_t ras_block; - uint32_t ras_sub_block; - const char* sysfs_name; +struct amdgpu_mca_ras_block { + struct amdgpu_ras_block_object ras_block; }; struct amdgpu_mca_ras { struct ras_common_if *ras_if; - const struct amdgpu_mca_ras_funcs *ras_funcs; + struct amdgpu_mca_ras_block *ras; }; struct amdgpu_mca_funcs { @@ -64,10 +56,4 @@ void amdgpu_mca_query_ras_error_count(struct amdgpu_device *adev, uint64_t mc_status_addr, void *ras_error_status); -int amdgpu_mca_ras_late_init(struct amdgpu_device *adev, - struct amdgpu_mca_ras *mca_dev); - -void amdgpu_mca_ras_fini(struct amdgpu_device *adev, - struct amdgpu_mca_ras *mca_dev); - #endif diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.c new file mode 100644 index 000000000000..0c546245793b --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.c @@ -0,0 +1,1425 @@ +/* + * Copyright 2019 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#include "amdgpu_mes.h" +#include "amdgpu.h" +#include "soc15_common.h" +#include "amdgpu_mes_ctx.h" + +#define AMDGPU_MES_MAX_NUM_OF_QUEUES_PER_PROCESS 1024 +#define AMDGPU_ONE_DOORBELL_SIZE 8 + +int amdgpu_mes_doorbell_process_slice(struct amdgpu_device *adev) +{ + return roundup(AMDGPU_ONE_DOORBELL_SIZE * + AMDGPU_MES_MAX_NUM_OF_QUEUES_PER_PROCESS, + PAGE_SIZE); +} + +int amdgpu_mes_alloc_process_doorbells(struct amdgpu_device *adev, + unsigned int *doorbell_index) +{ + int r = ida_simple_get(&adev->mes.doorbell_ida, 2, + adev->mes.max_doorbell_slices, + GFP_KERNEL); + if (r > 0) + *doorbell_index = r; + + return r; +} + +void amdgpu_mes_free_process_doorbells(struct amdgpu_device *adev, + unsigned int doorbell_index) +{ + if (doorbell_index) + ida_simple_remove(&adev->mes.doorbell_ida, doorbell_index); +} + +unsigned int amdgpu_mes_get_doorbell_dw_offset_in_bar( + struct amdgpu_device *adev, + uint32_t doorbell_index, + unsigned int doorbell_id) +{ + return ((doorbell_index * + amdgpu_mes_doorbell_process_slice(adev)) / sizeof(u32) + + doorbell_id * 2); +} + +static int amdgpu_mes_queue_doorbell_get(struct amdgpu_device *adev, + struct amdgpu_mes_process *process, + int ip_type, uint64_t *doorbell_index) +{ + unsigned int offset, found; + + if (ip_type == AMDGPU_RING_TYPE_SDMA) { + offset = adev->doorbell_index.sdma_engine[0]; + found = find_next_zero_bit(process->doorbell_bitmap, + AMDGPU_MES_MAX_NUM_OF_QUEUES_PER_PROCESS, + offset); + } else { + found = find_first_zero_bit(process->doorbell_bitmap, + AMDGPU_MES_MAX_NUM_OF_QUEUES_PER_PROCESS); + } + + if (found >= AMDGPU_MES_MAX_NUM_OF_QUEUES_PER_PROCESS) { + DRM_WARN("No doorbell available\n"); + return -ENOSPC; + } + + set_bit(found, process->doorbell_bitmap); + + *doorbell_index = amdgpu_mes_get_doorbell_dw_offset_in_bar(adev, + process->doorbell_index, found); + + return 0; +} + +static void amdgpu_mes_queue_doorbell_free(struct amdgpu_device *adev, + struct amdgpu_mes_process *process, + uint32_t doorbell_index) +{ + unsigned int old, doorbell_id; + + doorbell_id = doorbell_index - + (process->doorbell_index * + amdgpu_mes_doorbell_process_slice(adev)) / sizeof(u32); + doorbell_id /= 2; + + old = test_and_clear_bit(doorbell_id, process->doorbell_bitmap); + WARN_ON(!old); +} + +static int amdgpu_mes_doorbell_init(struct amdgpu_device *adev) +{ + size_t doorbell_start_offset; + size_t doorbell_aperture_size; + size_t doorbell_process_limit; + size_t aggregated_doorbell_start; + int i; + + aggregated_doorbell_start = (adev->doorbell_index.max_assignment + 1) * sizeof(u32); + aggregated_doorbell_start = + roundup(aggregated_doorbell_start, PAGE_SIZE); + + doorbell_start_offset = aggregated_doorbell_start + PAGE_SIZE; + doorbell_start_offset = + roundup(doorbell_start_offset, + amdgpu_mes_doorbell_process_slice(adev)); + + doorbell_aperture_size = adev->doorbell.size; + doorbell_aperture_size = + rounddown(doorbell_aperture_size, + amdgpu_mes_doorbell_process_slice(adev)); + + if (doorbell_aperture_size > doorbell_start_offset) + doorbell_process_limit = + (doorbell_aperture_size - doorbell_start_offset) / + amdgpu_mes_doorbell_process_slice(adev); + else + return -ENOSPC; + + adev->mes.doorbell_id_offset = doorbell_start_offset / sizeof(u32); + adev->mes.max_doorbell_slices = doorbell_process_limit; + + /* allocate Qword range for aggregated doorbell */ + for (i = 0; i < AMDGPU_MES_PRIORITY_NUM_LEVELS; i++) + adev->mes.aggregated_doorbells[i] = + aggregated_doorbell_start / sizeof(u32) + i * 2; + + DRM_INFO("max_doorbell_slices=%zu\n", doorbell_process_limit); + return 0; +} + +int amdgpu_mes_init(struct amdgpu_device *adev) +{ + int i, r; + + adev->mes.adev = adev; + + idr_init(&adev->mes.pasid_idr); + idr_init(&adev->mes.gang_id_idr); + idr_init(&adev->mes.queue_id_idr); + ida_init(&adev->mes.doorbell_ida); + spin_lock_init(&adev->mes.queue_id_lock); + spin_lock_init(&adev->mes.ring_lock); + mutex_init(&adev->mes.mutex_hidden); + + adev->mes.total_max_queue = AMDGPU_FENCE_MES_QUEUE_ID_MASK; + adev->mes.vmid_mask_mmhub = 0xffffff00; + adev->mes.vmid_mask_gfxhub = 0xffffff00; + + for (i = 0; i < AMDGPU_MES_MAX_COMPUTE_PIPES; i++) { + /* use only 1st MEC pipes */ + if (i >= 4) + continue; + adev->mes.compute_hqd_mask[i] = 0xc; + } + + for (i = 0; i < AMDGPU_MES_MAX_GFX_PIPES; i++) + adev->mes.gfx_hqd_mask[i] = i ? 0 : 0xfffffffe; + + for (i = 0; i < AMDGPU_MES_MAX_SDMA_PIPES; i++) { + if (adev->ip_versions[SDMA0_HWIP][0] < IP_VERSION(6, 0, 0)) + adev->mes.sdma_hqd_mask[i] = i ? 0 : 0x3fc; + /* zero sdma_hqd_mask for non-existent engine */ + else if (adev->sdma.num_instances == 1) + adev->mes.sdma_hqd_mask[i] = i ? 0 : 0xfc; + else + adev->mes.sdma_hqd_mask[i] = 0xfc; + } + + r = amdgpu_device_wb_get(adev, &adev->mes.sch_ctx_offs); + if (r) { + dev_err(adev->dev, + "(%d) ring trail_fence_offs wb alloc failed\n", r); + goto error_ids; + } + adev->mes.sch_ctx_gpu_addr = + adev->wb.gpu_addr + (adev->mes.sch_ctx_offs * 4); + adev->mes.sch_ctx_ptr = + (uint64_t *)&adev->wb.wb[adev->mes.sch_ctx_offs]; + + r = amdgpu_device_wb_get(adev, &adev->mes.query_status_fence_offs); + if (r) { + amdgpu_device_wb_free(adev, adev->mes.sch_ctx_offs); + dev_err(adev->dev, + "(%d) query_status_fence_offs wb alloc failed\n", r); + goto error_ids; + } + adev->mes.query_status_fence_gpu_addr = + adev->wb.gpu_addr + (adev->mes.query_status_fence_offs * 4); + adev->mes.query_status_fence_ptr = + (uint64_t *)&adev->wb.wb[adev->mes.query_status_fence_offs]; + + r = amdgpu_device_wb_get(adev, &adev->mes.read_val_offs); + if (r) { + amdgpu_device_wb_free(adev, adev->mes.sch_ctx_offs); + amdgpu_device_wb_free(adev, adev->mes.query_status_fence_offs); + dev_err(adev->dev, + "(%d) read_val_offs alloc failed\n", r); + goto error_ids; + } + adev->mes.read_val_gpu_addr = + adev->wb.gpu_addr + (adev->mes.read_val_offs * 4); + adev->mes.read_val_ptr = + (uint32_t *)&adev->wb.wb[adev->mes.read_val_offs]; + + r = amdgpu_mes_doorbell_init(adev); + if (r) + goto error; + + return 0; + +error: + amdgpu_device_wb_free(adev, adev->mes.sch_ctx_offs); + amdgpu_device_wb_free(adev, adev->mes.query_status_fence_offs); + amdgpu_device_wb_free(adev, adev->mes.read_val_offs); +error_ids: + idr_destroy(&adev->mes.pasid_idr); + idr_destroy(&adev->mes.gang_id_idr); + idr_destroy(&adev->mes.queue_id_idr); + ida_destroy(&adev->mes.doorbell_ida); + mutex_destroy(&adev->mes.mutex_hidden); + return r; +} + +void amdgpu_mes_fini(struct amdgpu_device *adev) +{ + amdgpu_device_wb_free(adev, adev->mes.sch_ctx_offs); + amdgpu_device_wb_free(adev, adev->mes.query_status_fence_offs); + amdgpu_device_wb_free(adev, adev->mes.read_val_offs); + + idr_destroy(&adev->mes.pasid_idr); + idr_destroy(&adev->mes.gang_id_idr); + idr_destroy(&adev->mes.queue_id_idr); + ida_destroy(&adev->mes.doorbell_ida); + mutex_destroy(&adev->mes.mutex_hidden); +} + +static void amdgpu_mes_queue_free_mqd(struct amdgpu_mes_queue *q) +{ + amdgpu_bo_free_kernel(&q->mqd_obj, + &q->mqd_gpu_addr, + &q->mqd_cpu_ptr); +} + +int amdgpu_mes_create_process(struct amdgpu_device *adev, int pasid, + struct amdgpu_vm *vm) +{ + struct amdgpu_mes_process *process; + int r; + + /* allocate the mes process buffer */ + process = kzalloc(sizeof(struct amdgpu_mes_process), GFP_KERNEL); + if (!process) { + DRM_ERROR("no more memory to create mes process\n"); + return -ENOMEM; + } + + process->doorbell_bitmap = + kzalloc(DIV_ROUND_UP(AMDGPU_MES_MAX_NUM_OF_QUEUES_PER_PROCESS, + BITS_PER_BYTE), GFP_KERNEL); + if (!process->doorbell_bitmap) { + DRM_ERROR("failed to allocate doorbell bitmap\n"); + kfree(process); + return -ENOMEM; + } + + /* allocate the process context bo and map it */ + r = amdgpu_bo_create_kernel(adev, AMDGPU_MES_PROC_CTX_SIZE, PAGE_SIZE, + AMDGPU_GEM_DOMAIN_GTT, + &process->proc_ctx_bo, + &process->proc_ctx_gpu_addr, + &process->proc_ctx_cpu_ptr); + if (r) { + DRM_ERROR("failed to allocate process context bo\n"); + goto clean_up_memory; + } + memset(process->proc_ctx_cpu_ptr, 0, AMDGPU_MES_PROC_CTX_SIZE); + + /* + * Avoid taking any other locks under MES lock to avoid circular + * lock dependencies. + */ + amdgpu_mes_lock(&adev->mes); + + /* add the mes process to idr list */ + r = idr_alloc(&adev->mes.pasid_idr, process, pasid, pasid + 1, + GFP_KERNEL); + if (r < 0) { + DRM_ERROR("failed to lock pasid=%d\n", pasid); + goto clean_up_ctx; + } + + /* allocate the starting doorbell index of the process */ + r = amdgpu_mes_alloc_process_doorbells(adev, &process->doorbell_index); + if (r < 0) { + DRM_ERROR("failed to allocate doorbell for process\n"); + goto clean_up_pasid; + } + + DRM_DEBUG("process doorbell index = %d\n", process->doorbell_index); + + INIT_LIST_HEAD(&process->gang_list); + process->vm = vm; + process->pasid = pasid; + process->process_quantum = adev->mes.default_process_quantum; + process->pd_gpu_addr = amdgpu_bo_gpu_offset(vm->root.bo); + + amdgpu_mes_unlock(&adev->mes); + return 0; + +clean_up_pasid: + idr_remove(&adev->mes.pasid_idr, pasid); + amdgpu_mes_unlock(&adev->mes); +clean_up_ctx: + amdgpu_bo_free_kernel(&process->proc_ctx_bo, + &process->proc_ctx_gpu_addr, + &process->proc_ctx_cpu_ptr); +clean_up_memory: + kfree(process->doorbell_bitmap); + kfree(process); + return r; +} + +void amdgpu_mes_destroy_process(struct amdgpu_device *adev, int pasid) +{ + struct amdgpu_mes_process *process; + struct amdgpu_mes_gang *gang, *tmp1; + struct amdgpu_mes_queue *queue, *tmp2; + struct mes_remove_queue_input queue_input; + unsigned long flags; + int r; + + /* + * Avoid taking any other locks under MES lock to avoid circular + * lock dependencies. + */ + amdgpu_mes_lock(&adev->mes); + + process = idr_find(&adev->mes.pasid_idr, pasid); + if (!process) { + DRM_WARN("pasid %d doesn't exist\n", pasid); + amdgpu_mes_unlock(&adev->mes); + return; + } + + /* Remove all queues from hardware */ + list_for_each_entry_safe(gang, tmp1, &process->gang_list, list) { + list_for_each_entry_safe(queue, tmp2, &gang->queue_list, list) { + spin_lock_irqsave(&adev->mes.queue_id_lock, flags); + idr_remove(&adev->mes.queue_id_idr, queue->queue_id); + spin_unlock_irqrestore(&adev->mes.queue_id_lock, flags); + + queue_input.doorbell_offset = queue->doorbell_off; + queue_input.gang_context_addr = gang->gang_ctx_gpu_addr; + + r = adev->mes.funcs->remove_hw_queue(&adev->mes, + &queue_input); + if (r) + DRM_WARN("failed to remove hardware queue\n"); + } + + idr_remove(&adev->mes.gang_id_idr, gang->gang_id); + } + + amdgpu_mes_free_process_doorbells(adev, process->doorbell_index); + idr_remove(&adev->mes.pasid_idr, pasid); + amdgpu_mes_unlock(&adev->mes); + + /* free all memory allocated by the process */ + list_for_each_entry_safe(gang, tmp1, &process->gang_list, list) { + /* free all queues in the gang */ + list_for_each_entry_safe(queue, tmp2, &gang->queue_list, list) { + amdgpu_mes_queue_free_mqd(queue); + list_del(&queue->list); + kfree(queue); + } + amdgpu_bo_free_kernel(&gang->gang_ctx_bo, + &gang->gang_ctx_gpu_addr, + &gang->gang_ctx_cpu_ptr); + list_del(&gang->list); + kfree(gang); + + } + amdgpu_bo_free_kernel(&process->proc_ctx_bo, + &process->proc_ctx_gpu_addr, + &process->proc_ctx_cpu_ptr); + kfree(process->doorbell_bitmap); + kfree(process); +} + +int amdgpu_mes_add_gang(struct amdgpu_device *adev, int pasid, + struct amdgpu_mes_gang_properties *gprops, + int *gang_id) +{ + struct amdgpu_mes_process *process; + struct amdgpu_mes_gang *gang; + int r; + + /* allocate the mes gang buffer */ + gang = kzalloc(sizeof(struct amdgpu_mes_gang), GFP_KERNEL); + if (!gang) { + return -ENOMEM; + } + + /* allocate the gang context bo and map it to cpu space */ + r = amdgpu_bo_create_kernel(adev, AMDGPU_MES_GANG_CTX_SIZE, PAGE_SIZE, + AMDGPU_GEM_DOMAIN_GTT, + &gang->gang_ctx_bo, + &gang->gang_ctx_gpu_addr, + &gang->gang_ctx_cpu_ptr); + if (r) { + DRM_ERROR("failed to allocate process context bo\n"); + goto clean_up_mem; + } + memset(gang->gang_ctx_cpu_ptr, 0, AMDGPU_MES_GANG_CTX_SIZE); + + /* + * Avoid taking any other locks under MES lock to avoid circular + * lock dependencies. + */ + amdgpu_mes_lock(&adev->mes); + + process = idr_find(&adev->mes.pasid_idr, pasid); + if (!process) { + DRM_ERROR("pasid %d doesn't exist\n", pasid); + r = -EINVAL; + goto clean_up_ctx; + } + + /* add the mes gang to idr list */ + r = idr_alloc(&adev->mes.gang_id_idr, gang, 1, 0, + GFP_KERNEL); + if (r < 0) { + DRM_ERROR("failed to allocate idr for gang\n"); + goto clean_up_ctx; + } + + gang->gang_id = r; + *gang_id = r; + + INIT_LIST_HEAD(&gang->queue_list); + gang->process = process; + gang->priority = gprops->priority; + gang->gang_quantum = gprops->gang_quantum ? + gprops->gang_quantum : adev->mes.default_gang_quantum; + gang->global_priority_level = gprops->global_priority_level; + gang->inprocess_gang_priority = gprops->inprocess_gang_priority; + list_add_tail(&gang->list, &process->gang_list); + + amdgpu_mes_unlock(&adev->mes); + return 0; + +clean_up_ctx: + amdgpu_mes_unlock(&adev->mes); + amdgpu_bo_free_kernel(&gang->gang_ctx_bo, + &gang->gang_ctx_gpu_addr, + &gang->gang_ctx_cpu_ptr); +clean_up_mem: + kfree(gang); + return r; +} + +int amdgpu_mes_remove_gang(struct amdgpu_device *adev, int gang_id) +{ + struct amdgpu_mes_gang *gang; + + /* + * Avoid taking any other locks under MES lock to avoid circular + * lock dependencies. + */ + amdgpu_mes_lock(&adev->mes); + + gang = idr_find(&adev->mes.gang_id_idr, gang_id); + if (!gang) { + DRM_ERROR("gang id %d doesn't exist\n", gang_id); + amdgpu_mes_unlock(&adev->mes); + return -EINVAL; + } + + if (!list_empty(&gang->queue_list)) { + DRM_ERROR("queue list is not empty\n"); + amdgpu_mes_unlock(&adev->mes); + return -EBUSY; + } + + idr_remove(&adev->mes.gang_id_idr, gang->gang_id); + list_del(&gang->list); + amdgpu_mes_unlock(&adev->mes); + + amdgpu_bo_free_kernel(&gang->gang_ctx_bo, + &gang->gang_ctx_gpu_addr, + &gang->gang_ctx_cpu_ptr); + + kfree(gang); + + return 0; +} + +int amdgpu_mes_suspend(struct amdgpu_device *adev) +{ + struct idr *idp; + struct amdgpu_mes_process *process; + struct amdgpu_mes_gang *gang; + struct mes_suspend_gang_input input; + int r, pasid; + + /* + * Avoid taking any other locks under MES lock to avoid circular + * lock dependencies. + */ + amdgpu_mes_lock(&adev->mes); + + idp = &adev->mes.pasid_idr; + + idr_for_each_entry(idp, process, pasid) { + list_for_each_entry(gang, &process->gang_list, list) { + r = adev->mes.funcs->suspend_gang(&adev->mes, &input); + if (r) + DRM_ERROR("failed to suspend pasid %d gangid %d", + pasid, gang->gang_id); + } + } + + amdgpu_mes_unlock(&adev->mes); + return 0; +} + +int amdgpu_mes_resume(struct amdgpu_device *adev) +{ + struct idr *idp; + struct amdgpu_mes_process *process; + struct amdgpu_mes_gang *gang; + struct mes_resume_gang_input input; + int r, pasid; + + /* + * Avoid taking any other locks under MES lock to avoid circular + * lock dependencies. + */ + amdgpu_mes_lock(&adev->mes); + + idp = &adev->mes.pasid_idr; + + idr_for_each_entry(idp, process, pasid) { + list_for_each_entry(gang, &process->gang_list, list) { + r = adev->mes.funcs->resume_gang(&adev->mes, &input); + if (r) + DRM_ERROR("failed to resume pasid %d gangid %d", + pasid, gang->gang_id); + } + } + + amdgpu_mes_unlock(&adev->mes); + return 0; +} + +static int amdgpu_mes_queue_alloc_mqd(struct amdgpu_device *adev, + struct amdgpu_mes_queue *q, + struct amdgpu_mes_queue_properties *p) +{ + struct amdgpu_mqd *mqd_mgr = &adev->mqds[p->queue_type]; + u32 mqd_size = mqd_mgr->mqd_size; + int r; + + r = amdgpu_bo_create_kernel(adev, mqd_size, PAGE_SIZE, + AMDGPU_GEM_DOMAIN_GTT, + &q->mqd_obj, + &q->mqd_gpu_addr, &q->mqd_cpu_ptr); + if (r) { + dev_warn(adev->dev, "failed to create queue mqd bo (%d)", r); + return r; + } + memset(q->mqd_cpu_ptr, 0, mqd_size); + + r = amdgpu_bo_reserve(q->mqd_obj, false); + if (unlikely(r != 0)) + goto clean_up; + + return 0; + +clean_up: + amdgpu_bo_free_kernel(&q->mqd_obj, + &q->mqd_gpu_addr, + &q->mqd_cpu_ptr); + return r; +} + +static void amdgpu_mes_queue_init_mqd(struct amdgpu_device *adev, + struct amdgpu_mes_queue *q, + struct amdgpu_mes_queue_properties *p) +{ + struct amdgpu_mqd *mqd_mgr = &adev->mqds[p->queue_type]; + struct amdgpu_mqd_prop mqd_prop = {0}; + + mqd_prop.mqd_gpu_addr = q->mqd_gpu_addr; + mqd_prop.hqd_base_gpu_addr = p->hqd_base_gpu_addr; + mqd_prop.rptr_gpu_addr = p->rptr_gpu_addr; + mqd_prop.wptr_gpu_addr = p->wptr_gpu_addr; + mqd_prop.queue_size = p->queue_size; + mqd_prop.use_doorbell = true; + mqd_prop.doorbell_index = p->doorbell_off; + mqd_prop.eop_gpu_addr = p->eop_gpu_addr; + mqd_prop.hqd_pipe_priority = p->hqd_pipe_priority; + mqd_prop.hqd_queue_priority = p->hqd_queue_priority; + mqd_prop.hqd_active = false; + + mqd_mgr->init_mqd(adev, q->mqd_cpu_ptr, &mqd_prop); + + amdgpu_bo_unreserve(q->mqd_obj); +} + +int amdgpu_mes_add_hw_queue(struct amdgpu_device *adev, int gang_id, + struct amdgpu_mes_queue_properties *qprops, + int *queue_id) +{ + struct amdgpu_mes_queue *queue; + struct amdgpu_mes_gang *gang; + struct mes_add_queue_input queue_input; + unsigned long flags; + int r; + + /* allocate the mes queue buffer */ + queue = kzalloc(sizeof(struct amdgpu_mes_queue), GFP_KERNEL); + if (!queue) { + DRM_ERROR("Failed to allocate memory for queue\n"); + return -ENOMEM; + } + + /* Allocate the queue mqd */ + r = amdgpu_mes_queue_alloc_mqd(adev, queue, qprops); + if (r) + goto clean_up_memory; + + /* + * Avoid taking any other locks under MES lock to avoid circular + * lock dependencies. + */ + amdgpu_mes_lock(&adev->mes); + + gang = idr_find(&adev->mes.gang_id_idr, gang_id); + if (!gang) { + DRM_ERROR("gang id %d doesn't exist\n", gang_id); + r = -EINVAL; + goto clean_up_mqd; + } + + /* add the mes gang to idr list */ + spin_lock_irqsave(&adev->mes.queue_id_lock, flags); + r = idr_alloc(&adev->mes.queue_id_idr, queue, 1, 0, + GFP_ATOMIC); + if (r < 0) { + spin_unlock_irqrestore(&adev->mes.queue_id_lock, flags); + goto clean_up_mqd; + } + spin_unlock_irqrestore(&adev->mes.queue_id_lock, flags); + *queue_id = queue->queue_id = r; + + /* allocate a doorbell index for the queue */ + r = amdgpu_mes_queue_doorbell_get(adev, gang->process, + qprops->queue_type, + &qprops->doorbell_off); + if (r) + goto clean_up_queue_id; + + /* initialize the queue mqd */ + amdgpu_mes_queue_init_mqd(adev, queue, qprops); + + /* add hw queue to mes */ + queue_input.process_id = gang->process->pasid; + + queue_input.page_table_base_addr = + adev->vm_manager.vram_base_offset + gang->process->pd_gpu_addr - + adev->gmc.vram_start; + + queue_input.process_va_start = 0; + queue_input.process_va_end = + (adev->vm_manager.max_pfn - 1) << AMDGPU_GPU_PAGE_SHIFT; + queue_input.process_quantum = gang->process->process_quantum; + queue_input.process_context_addr = gang->process->proc_ctx_gpu_addr; + queue_input.gang_quantum = gang->gang_quantum; + queue_input.gang_context_addr = gang->gang_ctx_gpu_addr; + queue_input.inprocess_gang_priority = gang->inprocess_gang_priority; + queue_input.gang_global_priority_level = gang->global_priority_level; + queue_input.doorbell_offset = qprops->doorbell_off; + queue_input.mqd_addr = queue->mqd_gpu_addr; + queue_input.wptr_addr = qprops->wptr_gpu_addr; + queue_input.wptr_mc_addr = qprops->wptr_mc_addr; + queue_input.queue_type = qprops->queue_type; + queue_input.paging = qprops->paging; + queue_input.is_kfd_process = 0; + + r = adev->mes.funcs->add_hw_queue(&adev->mes, &queue_input); + if (r) { + DRM_ERROR("failed to add hardware queue to MES, doorbell=0x%llx\n", + qprops->doorbell_off); + goto clean_up_doorbell; + } + + DRM_DEBUG("MES hw queue was added, pasid=%d, gang id=%d, " + "queue type=%d, doorbell=0x%llx\n", + gang->process->pasid, gang_id, qprops->queue_type, + qprops->doorbell_off); + + queue->ring = qprops->ring; + queue->doorbell_off = qprops->doorbell_off; + queue->wptr_gpu_addr = qprops->wptr_gpu_addr; + queue->queue_type = qprops->queue_type; + queue->paging = qprops->paging; + queue->gang = gang; + queue->ring->mqd_ptr = queue->mqd_cpu_ptr; + list_add_tail(&queue->list, &gang->queue_list); + + amdgpu_mes_unlock(&adev->mes); + return 0; + +clean_up_doorbell: + amdgpu_mes_queue_doorbell_free(adev, gang->process, + qprops->doorbell_off); +clean_up_queue_id: + spin_lock_irqsave(&adev->mes.queue_id_lock, flags); + idr_remove(&adev->mes.queue_id_idr, queue->queue_id); + spin_unlock_irqrestore(&adev->mes.queue_id_lock, flags); +clean_up_mqd: + amdgpu_mes_unlock(&adev->mes); + amdgpu_mes_queue_free_mqd(queue); +clean_up_memory: + kfree(queue); + return r; +} + +int amdgpu_mes_remove_hw_queue(struct amdgpu_device *adev, int queue_id) +{ + unsigned long flags; + struct amdgpu_mes_queue *queue; + struct amdgpu_mes_gang *gang; + struct mes_remove_queue_input queue_input; + int r; + + /* + * Avoid taking any other locks under MES lock to avoid circular + * lock dependencies. + */ + amdgpu_mes_lock(&adev->mes); + + /* remove the mes gang from idr list */ + spin_lock_irqsave(&adev->mes.queue_id_lock, flags); + + queue = idr_find(&adev->mes.queue_id_idr, queue_id); + if (!queue) { + spin_unlock_irqrestore(&adev->mes.queue_id_lock, flags); + amdgpu_mes_unlock(&adev->mes); + DRM_ERROR("queue id %d doesn't exist\n", queue_id); + return -EINVAL; + } + + idr_remove(&adev->mes.queue_id_idr, queue_id); + spin_unlock_irqrestore(&adev->mes.queue_id_lock, flags); + + DRM_DEBUG("try to remove queue, doorbell off = 0x%llx\n", + queue->doorbell_off); + + gang = queue->gang; + queue_input.doorbell_offset = queue->doorbell_off; + queue_input.gang_context_addr = gang->gang_ctx_gpu_addr; + + r = adev->mes.funcs->remove_hw_queue(&adev->mes, &queue_input); + if (r) + DRM_ERROR("failed to remove hardware queue, queue id = %d\n", + queue_id); + + list_del(&queue->list); + amdgpu_mes_queue_doorbell_free(adev, gang->process, + queue->doorbell_off); + amdgpu_mes_unlock(&adev->mes); + + amdgpu_mes_queue_free_mqd(queue); + kfree(queue); + return 0; +} + +int amdgpu_mes_unmap_legacy_queue(struct amdgpu_device *adev, + struct amdgpu_ring *ring, + enum amdgpu_unmap_queues_action action, + u64 gpu_addr, u64 seq) +{ + struct mes_unmap_legacy_queue_input queue_input; + int r; + + queue_input.action = action; + queue_input.queue_type = ring->funcs->type; + queue_input.doorbell_offset = ring->doorbell_index; + queue_input.pipe_id = ring->pipe; + queue_input.queue_id = ring->queue; + queue_input.trail_fence_addr = gpu_addr; + queue_input.trail_fence_data = seq; + + r = adev->mes.funcs->unmap_legacy_queue(&adev->mes, &queue_input); + if (r) + DRM_ERROR("failed to unmap legacy queue\n"); + + return r; +} + +uint32_t amdgpu_mes_rreg(struct amdgpu_device *adev, uint32_t reg) +{ + struct mes_misc_op_input op_input; + int r, val = 0; + + op_input.op = MES_MISC_OP_READ_REG; + op_input.read_reg.reg_offset = reg; + op_input.read_reg.buffer_addr = adev->mes.read_val_gpu_addr; + + if (!adev->mes.funcs->misc_op) { + DRM_ERROR("mes rreg is not supported!\n"); + goto error; + } + + r = adev->mes.funcs->misc_op(&adev->mes, &op_input); + if (r) + DRM_ERROR("failed to read reg (0x%x)\n", reg); + else + val = *(adev->mes.read_val_ptr); + +error: + return val; +} + +int amdgpu_mes_wreg(struct amdgpu_device *adev, + uint32_t reg, uint32_t val) +{ + struct mes_misc_op_input op_input; + int r; + + op_input.op = MES_MISC_OP_WRITE_REG; + op_input.write_reg.reg_offset = reg; + op_input.write_reg.reg_value = val; + + if (!adev->mes.funcs->misc_op) { + DRM_ERROR("mes wreg is not supported!\n"); + r = -EINVAL; + goto error; + } + + r = adev->mes.funcs->misc_op(&adev->mes, &op_input); + if (r) + DRM_ERROR("failed to write reg (0x%x)\n", reg); + +error: + return r; +} + +int amdgpu_mes_reg_write_reg_wait(struct amdgpu_device *adev, + uint32_t reg0, uint32_t reg1, + uint32_t ref, uint32_t mask) +{ + struct mes_misc_op_input op_input; + int r; + + op_input.op = MES_MISC_OP_WRM_REG_WR_WAIT; + op_input.wrm_reg.reg0 = reg0; + op_input.wrm_reg.reg1 = reg1; + op_input.wrm_reg.ref = ref; + op_input.wrm_reg.mask = mask; + + if (!adev->mes.funcs->misc_op) { + DRM_ERROR("mes reg_write_reg_wait is not supported!\n"); + r = -EINVAL; + goto error; + } + + r = adev->mes.funcs->misc_op(&adev->mes, &op_input); + if (r) + DRM_ERROR("failed to reg_write_reg_wait\n"); + +error: + return r; +} + +int amdgpu_mes_reg_wait(struct amdgpu_device *adev, uint32_t reg, + uint32_t val, uint32_t mask) +{ + struct mes_misc_op_input op_input; + int r; + + op_input.op = MES_MISC_OP_WRM_REG_WAIT; + op_input.wrm_reg.reg0 = reg; + op_input.wrm_reg.ref = val; + op_input.wrm_reg.mask = mask; + + if (!adev->mes.funcs->misc_op) { + DRM_ERROR("mes reg wait is not supported!\n"); + r = -EINVAL; + goto error; + } + + r = adev->mes.funcs->misc_op(&adev->mes, &op_input); + if (r) + DRM_ERROR("failed to reg_write_reg_wait\n"); + +error: + return r; +} + +static void +amdgpu_mes_ring_to_queue_props(struct amdgpu_device *adev, + struct amdgpu_ring *ring, + struct amdgpu_mes_queue_properties *props) +{ + props->queue_type = ring->funcs->type; + props->hqd_base_gpu_addr = ring->gpu_addr; + props->rptr_gpu_addr = ring->rptr_gpu_addr; + props->wptr_gpu_addr = ring->wptr_gpu_addr; + props->wptr_mc_addr = + ring->mes_ctx->meta_data_mc_addr + ring->wptr_offs; + props->queue_size = ring->ring_size; + props->eop_gpu_addr = ring->eop_gpu_addr; + props->hqd_pipe_priority = AMDGPU_GFX_PIPE_PRIO_NORMAL; + props->hqd_queue_priority = AMDGPU_GFX_QUEUE_PRIORITY_MINIMUM; + props->paging = false; + props->ring = ring; +} + +#define DEFINE_AMDGPU_MES_CTX_GET_OFFS_ENG(_eng) \ +do { \ + if (id_offs < AMDGPU_MES_CTX_MAX_OFFS) \ + return offsetof(struct amdgpu_mes_ctx_meta_data, \ + _eng[ring->idx].slots[id_offs]); \ + else if (id_offs == AMDGPU_MES_CTX_RING_OFFS) \ + return offsetof(struct amdgpu_mes_ctx_meta_data, \ + _eng[ring->idx].ring); \ + else if (id_offs == AMDGPU_MES_CTX_IB_OFFS) \ + return offsetof(struct amdgpu_mes_ctx_meta_data, \ + _eng[ring->idx].ib); \ + else if (id_offs == AMDGPU_MES_CTX_PADDING_OFFS) \ + return offsetof(struct amdgpu_mes_ctx_meta_data, \ + _eng[ring->idx].padding); \ +} while(0) + +int amdgpu_mes_ctx_get_offs(struct amdgpu_ring *ring, unsigned int id_offs) +{ + switch (ring->funcs->type) { + case AMDGPU_RING_TYPE_GFX: + DEFINE_AMDGPU_MES_CTX_GET_OFFS_ENG(gfx); + break; + case AMDGPU_RING_TYPE_COMPUTE: + DEFINE_AMDGPU_MES_CTX_GET_OFFS_ENG(compute); + break; + case AMDGPU_RING_TYPE_SDMA: + DEFINE_AMDGPU_MES_CTX_GET_OFFS_ENG(sdma); + break; + default: + break; + } + + WARN_ON(1); + return -EINVAL; +} + +int amdgpu_mes_add_ring(struct amdgpu_device *adev, int gang_id, + int queue_type, int idx, + struct amdgpu_mes_ctx_data *ctx_data, + struct amdgpu_ring **out) +{ + struct amdgpu_ring *ring; + struct amdgpu_mes_gang *gang; + struct amdgpu_mes_queue_properties qprops = {0}; + int r, queue_id, pasid; + + /* + * Avoid taking any other locks under MES lock to avoid circular + * lock dependencies. + */ + amdgpu_mes_lock(&adev->mes); + gang = idr_find(&adev->mes.gang_id_idr, gang_id); + if (!gang) { + DRM_ERROR("gang id %d doesn't exist\n", gang_id); + amdgpu_mes_unlock(&adev->mes); + return -EINVAL; + } + pasid = gang->process->pasid; + + ring = kzalloc(sizeof(struct amdgpu_ring), GFP_KERNEL); + if (!ring) { + amdgpu_mes_unlock(&adev->mes); + return -ENOMEM; + } + + ring->ring_obj = NULL; + ring->use_doorbell = true; + ring->is_mes_queue = true; + ring->mes_ctx = ctx_data; + ring->idx = idx; + ring->no_scheduler = true; + + if (queue_type == AMDGPU_RING_TYPE_COMPUTE) { + int offset = offsetof(struct amdgpu_mes_ctx_meta_data, + compute[ring->idx].mec_hpd); + ring->eop_gpu_addr = + amdgpu_mes_ctx_get_offs_gpu_addr(ring, offset); + } + + switch (queue_type) { + case AMDGPU_RING_TYPE_GFX: + ring->funcs = adev->gfx.gfx_ring[0].funcs; + break; + case AMDGPU_RING_TYPE_COMPUTE: + ring->funcs = adev->gfx.compute_ring[0].funcs; + break; + case AMDGPU_RING_TYPE_SDMA: + ring->funcs = adev->sdma.instance[0].ring.funcs; + break; + default: + BUG(); + } + + r = amdgpu_ring_init(adev, ring, 1024, NULL, 0, + AMDGPU_RING_PRIO_DEFAULT, NULL); + if (r) + goto clean_up_memory; + + amdgpu_mes_ring_to_queue_props(adev, ring, &qprops); + + dma_fence_wait(gang->process->vm->last_update, false); + dma_fence_wait(ctx_data->meta_data_va->last_pt_update, false); + amdgpu_mes_unlock(&adev->mes); + + r = amdgpu_mes_add_hw_queue(adev, gang_id, &qprops, &queue_id); + if (r) + goto clean_up_ring; + + ring->hw_queue_id = queue_id; + ring->doorbell_index = qprops.doorbell_off; + + if (queue_type == AMDGPU_RING_TYPE_GFX) + sprintf(ring->name, "gfx_%d.%d.%d", pasid, gang_id, queue_id); + else if (queue_type == AMDGPU_RING_TYPE_COMPUTE) + sprintf(ring->name, "compute_%d.%d.%d", pasid, gang_id, + queue_id); + else if (queue_type == AMDGPU_RING_TYPE_SDMA) + sprintf(ring->name, "sdma_%d.%d.%d", pasid, gang_id, + queue_id); + else + BUG(); + + *out = ring; + return 0; + +clean_up_ring: + amdgpu_ring_fini(ring); +clean_up_memory: + kfree(ring); + amdgpu_mes_unlock(&adev->mes); + return r; +} + +void amdgpu_mes_remove_ring(struct amdgpu_device *adev, + struct amdgpu_ring *ring) +{ + if (!ring) + return; + + amdgpu_mes_remove_hw_queue(adev, ring->hw_queue_id); + amdgpu_ring_fini(ring); + kfree(ring); +} + +uint32_t amdgpu_mes_get_aggregated_doorbell_index(struct amdgpu_device *adev, + enum amdgpu_mes_priority_level prio) +{ + return adev->mes.aggregated_doorbells[prio]; +} + +int amdgpu_mes_ctx_alloc_meta_data(struct amdgpu_device *adev, + struct amdgpu_mes_ctx_data *ctx_data) +{ + int r; + + r = amdgpu_bo_create_kernel(adev, + sizeof(struct amdgpu_mes_ctx_meta_data), + PAGE_SIZE, AMDGPU_GEM_DOMAIN_GTT, + &ctx_data->meta_data_obj, + &ctx_data->meta_data_mc_addr, + &ctx_data->meta_data_ptr); + if (!ctx_data->meta_data_obj) + return -ENOMEM; + + memset(ctx_data->meta_data_ptr, 0, + sizeof(struct amdgpu_mes_ctx_meta_data)); + + return 0; +} + +void amdgpu_mes_ctx_free_meta_data(struct amdgpu_mes_ctx_data *ctx_data) +{ + if (ctx_data->meta_data_obj) + amdgpu_bo_free_kernel(&ctx_data->meta_data_obj, + &ctx_data->meta_data_mc_addr, + &ctx_data->meta_data_ptr); +} + +int amdgpu_mes_ctx_map_meta_data(struct amdgpu_device *adev, + struct amdgpu_vm *vm, + struct amdgpu_mes_ctx_data *ctx_data) +{ + struct amdgpu_bo_va *bo_va; + struct ww_acquire_ctx ticket; + struct list_head list; + struct amdgpu_bo_list_entry pd; + struct ttm_validate_buffer csa_tv; + struct amdgpu_sync sync; + int r; + + amdgpu_sync_create(&sync); + INIT_LIST_HEAD(&list); + INIT_LIST_HEAD(&csa_tv.head); + + csa_tv.bo = &ctx_data->meta_data_obj->tbo; + csa_tv.num_shared = 1; + + list_add(&csa_tv.head, &list); + amdgpu_vm_get_pd_bo(vm, &list, &pd); + + r = ttm_eu_reserve_buffers(&ticket, &list, true, NULL); + if (r) { + DRM_ERROR("failed to reserve meta data BO: err=%d\n", r); + return r; + } + + bo_va = amdgpu_vm_bo_add(adev, vm, ctx_data->meta_data_obj); + if (!bo_va) { + ttm_eu_backoff_reservation(&ticket, &list); + DRM_ERROR("failed to create bo_va for meta data BO\n"); + return -ENOMEM; + } + + r = amdgpu_vm_bo_map(adev, bo_va, ctx_data->meta_data_gpu_addr, 0, + sizeof(struct amdgpu_mes_ctx_meta_data), + AMDGPU_PTE_READABLE | AMDGPU_PTE_WRITEABLE | + AMDGPU_PTE_EXECUTABLE); + + if (r) { + DRM_ERROR("failed to do bo_map on meta data, err=%d\n", r); + goto error; + } + + r = amdgpu_vm_bo_update(adev, bo_va, false); + if (r) { + DRM_ERROR("failed to do vm_bo_update on meta data\n"); + goto error; + } + amdgpu_sync_fence(&sync, bo_va->last_pt_update); + + r = amdgpu_vm_update_pdes(adev, vm, false); + if (r) { + DRM_ERROR("failed to update pdes on meta data\n"); + goto error; + } + amdgpu_sync_fence(&sync, vm->last_update); + + amdgpu_sync_wait(&sync, false); + ttm_eu_backoff_reservation(&ticket, &list); + + amdgpu_sync_free(&sync); + ctx_data->meta_data_va = bo_va; + return 0; + +error: + amdgpu_vm_bo_del(adev, bo_va); + ttm_eu_backoff_reservation(&ticket, &list); + amdgpu_sync_free(&sync); + return r; +} + +int amdgpu_mes_ctx_unmap_meta_data(struct amdgpu_device *adev, + struct amdgpu_mes_ctx_data *ctx_data) +{ + struct amdgpu_bo_va *bo_va = ctx_data->meta_data_va; + struct amdgpu_bo *bo = ctx_data->meta_data_obj; + struct amdgpu_vm *vm = bo_va->base.vm; + struct amdgpu_bo_list_entry vm_pd; + struct list_head list, duplicates; + struct dma_fence *fence = NULL; + struct ttm_validate_buffer tv; + struct ww_acquire_ctx ticket; + long r = 0; + + INIT_LIST_HEAD(&list); + INIT_LIST_HEAD(&duplicates); + + tv.bo = &bo->tbo; + tv.num_shared = 2; + list_add(&tv.head, &list); + + amdgpu_vm_get_pd_bo(vm, &list, &vm_pd); + + r = ttm_eu_reserve_buffers(&ticket, &list, false, &duplicates); + if (r) { + dev_err(adev->dev, "leaking bo va because " + "we fail to reserve bo (%ld)\n", r); + return r; + } + + amdgpu_vm_bo_del(adev, bo_va); + if (!amdgpu_vm_ready(vm)) + goto out_unlock; + + r = dma_resv_get_singleton(bo->tbo.base.resv, DMA_RESV_USAGE_BOOKKEEP, &fence); + if (r) + goto out_unlock; + if (fence) { + amdgpu_bo_fence(bo, fence, true); + fence = NULL; + } + + r = amdgpu_vm_clear_freed(adev, vm, &fence); + if (r || !fence) + goto out_unlock; + + dma_fence_wait(fence, false); + amdgpu_bo_fence(bo, fence, true); + dma_fence_put(fence); + +out_unlock: + if (unlikely(r < 0)) + dev_err(adev->dev, "failed to clear page tables (%ld)\n", r); + ttm_eu_backoff_reservation(&ticket, &list); + + return r; +} + +static int amdgpu_mes_test_create_gang_and_queues(struct amdgpu_device *adev, + int pasid, int *gang_id, + int queue_type, int num_queue, + struct amdgpu_ring **added_rings, + struct amdgpu_mes_ctx_data *ctx_data) +{ + struct amdgpu_ring *ring; + struct amdgpu_mes_gang_properties gprops = {0}; + int r, j; + + /* create a gang for the process */ + gprops.priority = AMDGPU_MES_PRIORITY_LEVEL_NORMAL; + gprops.gang_quantum = adev->mes.default_gang_quantum; + gprops.inprocess_gang_priority = AMDGPU_MES_PRIORITY_LEVEL_NORMAL; + gprops.priority_level = AMDGPU_MES_PRIORITY_LEVEL_NORMAL; + gprops.global_priority_level = AMDGPU_MES_PRIORITY_LEVEL_NORMAL; + + r = amdgpu_mes_add_gang(adev, pasid, &gprops, gang_id); + if (r) { + DRM_ERROR("failed to add gang\n"); + return r; + } + + /* create queues for the gang */ + for (j = 0; j < num_queue; j++) { + r = amdgpu_mes_add_ring(adev, *gang_id, queue_type, j, + ctx_data, &ring); + if (r) { + DRM_ERROR("failed to add ring\n"); + break; + } + + DRM_INFO("ring %s was added\n", ring->name); + added_rings[j] = ring; + } + + return 0; +} + +static int amdgpu_mes_test_queues(struct amdgpu_ring **added_rings) +{ + struct amdgpu_ring *ring; + int i, r; + + for (i = 0; i < AMDGPU_MES_CTX_MAX_RINGS; i++) { + ring = added_rings[i]; + if (!ring) + continue; + + r = amdgpu_ring_test_ring(ring); + if (r) { + DRM_DEV_ERROR(ring->adev->dev, + "ring %s test failed (%d)\n", + ring->name, r); + return r; + } else + DRM_INFO("ring %s test pass\n", ring->name); + + r = amdgpu_ring_test_ib(ring, 1000 * 10); + if (r) { + DRM_DEV_ERROR(ring->adev->dev, + "ring %s ib test failed (%d)\n", + ring->name, r); + return r; + } else + DRM_INFO("ring %s ib test pass\n", ring->name); + } + + return 0; +} + +int amdgpu_mes_self_test(struct amdgpu_device *adev) +{ + struct amdgpu_vm *vm = NULL; + struct amdgpu_mes_ctx_data ctx_data = {0}; + struct amdgpu_ring *added_rings[AMDGPU_MES_CTX_MAX_RINGS] = { NULL }; + int gang_ids[3] = {0}; + int queue_types[][2] = { { AMDGPU_RING_TYPE_GFX, + AMDGPU_MES_CTX_MAX_GFX_RINGS}, + { AMDGPU_RING_TYPE_COMPUTE, + AMDGPU_MES_CTX_MAX_COMPUTE_RINGS}, + { AMDGPU_RING_TYPE_SDMA, + AMDGPU_MES_CTX_MAX_SDMA_RINGS } }; + int i, r, pasid, k = 0; + + pasid = amdgpu_pasid_alloc(16); + if (pasid < 0) { + dev_warn(adev->dev, "No more PASIDs available!"); + pasid = 0; + } + + vm = kzalloc(sizeof(*vm), GFP_KERNEL); + if (!vm) { + r = -ENOMEM; + goto error_pasid; + } + + r = amdgpu_vm_init(adev, vm); + if (r) { + DRM_ERROR("failed to initialize vm\n"); + goto error_pasid; + } + + r = amdgpu_mes_ctx_alloc_meta_data(adev, &ctx_data); + if (r) { + DRM_ERROR("failed to alloc ctx meta data\n"); + goto error_fini; + } + + ctx_data.meta_data_gpu_addr = AMDGPU_VA_RESERVED_SIZE; + r = amdgpu_mes_ctx_map_meta_data(adev, vm, &ctx_data); + if (r) { + DRM_ERROR("failed to map ctx meta data\n"); + goto error_vm; + } + + r = amdgpu_mes_create_process(adev, pasid, vm); + if (r) { + DRM_ERROR("failed to create MES process\n"); + goto error_vm; + } + + for (i = 0; i < ARRAY_SIZE(queue_types); i++) { + /* On GFX v10.3, fw hasn't supported to map sdma queue. */ + if (adev->ip_versions[GC_HWIP][0] >= IP_VERSION(10, 3, 0) && + adev->ip_versions[GC_HWIP][0] < IP_VERSION(11, 0, 0) && + queue_types[i][0] == AMDGPU_RING_TYPE_SDMA) + continue; + + r = amdgpu_mes_test_create_gang_and_queues(adev, pasid, + &gang_ids[i], + queue_types[i][0], + queue_types[i][1], + &added_rings[k], + &ctx_data); + if (r) + goto error_queues; + + k += queue_types[i][1]; + } + + /* start ring test and ib test for MES queues */ + amdgpu_mes_test_queues(added_rings); + +error_queues: + /* remove all queues */ + for (i = 0; i < ARRAY_SIZE(added_rings); i++) { + if (!added_rings[i]) + continue; + amdgpu_mes_remove_ring(adev, added_rings[i]); + } + + for (i = 0; i < ARRAY_SIZE(gang_ids); i++) { + if (!gang_ids[i]) + continue; + amdgpu_mes_remove_gang(adev, gang_ids[i]); + } + + amdgpu_mes_destroy_process(adev, pasid); + +error_vm: + amdgpu_mes_ctx_unmap_meta_data(adev, &ctx_data); + +error_fini: + amdgpu_vm_fini(adev, vm); + +error_pasid: + if (pasid) + amdgpu_pasid_free(pasid); + + amdgpu_mes_ctx_free_meta_data(&ctx_data); + kfree(vm); + return 0; +} diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.h index 7334982ea702..ad980f4b66e1 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.h @@ -24,10 +24,22 @@ #ifndef __AMDGPU_MES_H__ #define __AMDGPU_MES_H__ +#include "amdgpu_irq.h" +#include "kgd_kfd_interface.h" +#include "amdgpu_gfx.h" +#include <linux/sched/mm.h> + #define AMDGPU_MES_MAX_COMPUTE_PIPES 8 #define AMDGPU_MES_MAX_GFX_PIPES 2 #define AMDGPU_MES_MAX_SDMA_PIPES 2 +#define AMDGPU_MES_API_VERSION_SHIFT 12 +#define AMDGPU_MES_FEAT_VERSION_SHIFT 24 + +#define AMDGPU_MES_VERSION_MASK 0x00000fff +#define AMDGPU_MES_API_VERSION_MASK 0x00fff000 +#define AMDGPU_MES_FEAT_VERSION_MASK 0xff000000 + enum amdgpu_mes_priority_level { AMDGPU_MES_PRIORITY_LEVEL_LOW = 0, AMDGPU_MES_PRIORITY_LEVEL_NORMAL = 1, @@ -37,11 +49,32 @@ enum amdgpu_mes_priority_level { AMDGPU_MES_PRIORITY_NUM_LEVELS }; +#define AMDGPU_MES_PROC_CTX_SIZE 0x1000 /* one page area */ +#define AMDGPU_MES_GANG_CTX_SIZE 0x1000 /* one page area */ + struct amdgpu_mes_funcs; +enum admgpu_mes_pipe { + AMDGPU_MES_SCHED_PIPE = 0, + AMDGPU_MES_KIQ_PIPE, + AMDGPU_MAX_MES_PIPES = 2, +}; + struct amdgpu_mes { struct amdgpu_device *adev; + struct mutex mutex_hidden; + + struct idr pasid_idr; + struct idr gang_id_idr; + struct idr queue_id_idr; + struct ida doorbell_ida; + + spinlock_t queue_id_lock; + + uint32_t sched_version; + uint32_t kiq_version; + uint32_t total_max_queue; uint32_t doorbell_id_offset; uint32_t max_doorbell_slices; @@ -50,46 +83,123 @@ struct amdgpu_mes { uint64_t default_gang_quantum; struct amdgpu_ring ring; + spinlock_t ring_lock; - const struct firmware *fw; + const struct firmware *fw[AMDGPU_MAX_MES_PIPES]; /* mes ucode */ - struct amdgpu_bo *ucode_fw_obj; - uint64_t ucode_fw_gpu_addr; - uint32_t *ucode_fw_ptr; - uint32_t ucode_fw_version; - uint64_t uc_start_addr; + struct amdgpu_bo *ucode_fw_obj[AMDGPU_MAX_MES_PIPES]; + uint64_t ucode_fw_gpu_addr[AMDGPU_MAX_MES_PIPES]; + uint32_t *ucode_fw_ptr[AMDGPU_MAX_MES_PIPES]; + uint32_t ucode_fw_version[AMDGPU_MAX_MES_PIPES]; + uint64_t uc_start_addr[AMDGPU_MAX_MES_PIPES]; /* mes ucode data */ - struct amdgpu_bo *data_fw_obj; - uint64_t data_fw_gpu_addr; - uint32_t *data_fw_ptr; - uint32_t data_fw_version; - uint64_t data_start_addr; + struct amdgpu_bo *data_fw_obj[AMDGPU_MAX_MES_PIPES]; + uint64_t data_fw_gpu_addr[AMDGPU_MAX_MES_PIPES]; + uint32_t *data_fw_ptr[AMDGPU_MAX_MES_PIPES]; + uint32_t data_fw_version[AMDGPU_MAX_MES_PIPES]; + uint64_t data_start_addr[AMDGPU_MAX_MES_PIPES]; /* eop gpu obj */ - struct amdgpu_bo *eop_gpu_obj; - uint64_t eop_gpu_addr; + struct amdgpu_bo *eop_gpu_obj[AMDGPU_MAX_MES_PIPES]; + uint64_t eop_gpu_addr[AMDGPU_MAX_MES_PIPES]; - void *mqd_backup; + void *mqd_backup[AMDGPU_MAX_MES_PIPES]; + struct amdgpu_irq_src irq[AMDGPU_MAX_MES_PIPES]; uint32_t vmid_mask_gfxhub; uint32_t vmid_mask_mmhub; uint32_t compute_hqd_mask[AMDGPU_MES_MAX_COMPUTE_PIPES]; uint32_t gfx_hqd_mask[AMDGPU_MES_MAX_GFX_PIPES]; uint32_t sdma_hqd_mask[AMDGPU_MES_MAX_SDMA_PIPES]; - uint32_t agreegated_doorbells[AMDGPU_MES_PRIORITY_NUM_LEVELS]; + uint32_t aggregated_doorbells[AMDGPU_MES_PRIORITY_NUM_LEVELS]; uint32_t sch_ctx_offs; uint64_t sch_ctx_gpu_addr; uint64_t *sch_ctx_ptr; uint32_t query_status_fence_offs; uint64_t query_status_fence_gpu_addr; uint64_t *query_status_fence_ptr; + uint32_t read_val_offs; + uint64_t read_val_gpu_addr; + uint32_t *read_val_ptr; + + uint32_t saved_flags; + + /* initialize kiq pipe */ + int (*kiq_hw_init)(struct amdgpu_device *adev); + int (*kiq_hw_fini)(struct amdgpu_device *adev); /* ip specific functions */ const struct amdgpu_mes_funcs *funcs; }; +struct amdgpu_mes_process { + int pasid; + struct amdgpu_vm *vm; + uint64_t pd_gpu_addr; + struct amdgpu_bo *proc_ctx_bo; + uint64_t proc_ctx_gpu_addr; + void *proc_ctx_cpu_ptr; + uint64_t process_quantum; + struct list_head gang_list; + uint32_t doorbell_index; + unsigned long *doorbell_bitmap; + struct mutex doorbell_lock; +}; + +struct amdgpu_mes_gang { + int gang_id; + int priority; + int inprocess_gang_priority; + int global_priority_level; + struct list_head list; + struct amdgpu_mes_process *process; + struct amdgpu_bo *gang_ctx_bo; + uint64_t gang_ctx_gpu_addr; + void *gang_ctx_cpu_ptr; + uint64_t gang_quantum; + struct list_head queue_list; +}; + +struct amdgpu_mes_queue { + struct list_head list; + struct amdgpu_mes_gang *gang; + int queue_id; + uint64_t doorbell_off; + struct amdgpu_bo *mqd_obj; + void *mqd_cpu_ptr; + uint64_t mqd_gpu_addr; + uint64_t wptr_gpu_addr; + int queue_type; + int paging; + struct amdgpu_ring *ring; +}; + +struct amdgpu_mes_queue_properties { + int queue_type; + uint64_t hqd_base_gpu_addr; + uint64_t rptr_gpu_addr; + uint64_t wptr_gpu_addr; + uint64_t wptr_mc_addr; + uint32_t queue_size; + uint64_t eop_gpu_addr; + uint32_t hqd_pipe_priority; + uint32_t hqd_queue_priority; + bool paging; + struct amdgpu_ring *ring; + /* out */ + uint64_t doorbell_off; +}; + +struct amdgpu_mes_gang_properties { + uint32_t priority; + uint32_t gang_quantum; + uint32_t inprocess_gang_priority; + uint32_t priority_level; + int global_priority_level; +}; + struct mes_add_queue_input { uint32_t process_id; uint64_t page_table_base_addr; @@ -104,8 +214,16 @@ struct mes_add_queue_input { uint32_t doorbell_offset; uint64_t mqd_addr; uint64_t wptr_addr; + uint64_t wptr_mc_addr; uint32_t queue_type; uint32_t paging; + uint32_t gws_base; + uint32_t gws_size; + uint64_t tba_addr; + uint64_t tma_addr; + uint32_t is_kfd_process; + uint32_t is_aql_queue; + uint32_t queue_size; }; struct mes_remove_queue_input { @@ -113,6 +231,16 @@ struct mes_remove_queue_input { uint64_t gang_context_addr; }; +struct mes_unmap_legacy_queue_input { + enum amdgpu_unmap_queues_action action; + uint32_t queue_type; + uint32_t doorbell_offset; + uint32_t pipe_id; + uint32_t queue_id; + uint64_t trail_fence_addr; + uint64_t trail_fence_data; +}; + struct mes_suspend_gang_input { bool suspend_all_gangs; uint64_t gang_context_addr; @@ -125,6 +253,36 @@ struct mes_resume_gang_input { uint64_t gang_context_addr; }; +enum mes_misc_opcode { + MES_MISC_OP_WRITE_REG, + MES_MISC_OP_READ_REG, + MES_MISC_OP_WRM_REG_WAIT, + MES_MISC_OP_WRM_REG_WR_WAIT, +}; + +struct mes_misc_op_input { + enum mes_misc_opcode op; + + union { + struct { + uint32_t reg_offset; + uint64_t buffer_addr; + } read_reg; + + struct { + uint32_t reg_offset; + uint32_t reg_value; + } write_reg; + + struct { + uint32_t ref; + uint32_t mask; + uint32_t reg0; + uint32_t reg1; + } wrm_reg; + }; +}; + struct amdgpu_mes_funcs { int (*add_hw_queue)(struct amdgpu_mes *mes, struct mes_add_queue_input *input); @@ -132,11 +290,144 @@ struct amdgpu_mes_funcs { int (*remove_hw_queue)(struct amdgpu_mes *mes, struct mes_remove_queue_input *input); + int (*unmap_legacy_queue)(struct amdgpu_mes *mes, + struct mes_unmap_legacy_queue_input *input); + int (*suspend_gang)(struct amdgpu_mes *mes, struct mes_suspend_gang_input *input); int (*resume_gang)(struct amdgpu_mes *mes, struct mes_resume_gang_input *input); + + int (*misc_op)(struct amdgpu_mes *mes, + struct mes_misc_op_input *input); }; +#define amdgpu_mes_kiq_hw_init(adev) (adev)->mes.kiq_hw_init((adev)) +#define amdgpu_mes_kiq_hw_fini(adev) (adev)->mes.kiq_hw_fini((adev)) + +int amdgpu_mes_ctx_get_offs(struct amdgpu_ring *ring, unsigned int id_offs); + +int amdgpu_mes_init(struct amdgpu_device *adev); +void amdgpu_mes_fini(struct amdgpu_device *adev); + +int amdgpu_mes_create_process(struct amdgpu_device *adev, int pasid, + struct amdgpu_vm *vm); +void amdgpu_mes_destroy_process(struct amdgpu_device *adev, int pasid); + +int amdgpu_mes_add_gang(struct amdgpu_device *adev, int pasid, + struct amdgpu_mes_gang_properties *gprops, + int *gang_id); +int amdgpu_mes_remove_gang(struct amdgpu_device *adev, int gang_id); + +int amdgpu_mes_suspend(struct amdgpu_device *adev); +int amdgpu_mes_resume(struct amdgpu_device *adev); + +int amdgpu_mes_add_hw_queue(struct amdgpu_device *adev, int gang_id, + struct amdgpu_mes_queue_properties *qprops, + int *queue_id); +int amdgpu_mes_remove_hw_queue(struct amdgpu_device *adev, int queue_id); + +int amdgpu_mes_unmap_legacy_queue(struct amdgpu_device *adev, + struct amdgpu_ring *ring, + enum amdgpu_unmap_queues_action action, + u64 gpu_addr, u64 seq); + +uint32_t amdgpu_mes_rreg(struct amdgpu_device *adev, uint32_t reg); +int amdgpu_mes_wreg(struct amdgpu_device *adev, + uint32_t reg, uint32_t val); +int amdgpu_mes_reg_wait(struct amdgpu_device *adev, uint32_t reg, + uint32_t val, uint32_t mask); +int amdgpu_mes_reg_write_reg_wait(struct amdgpu_device *adev, + uint32_t reg0, uint32_t reg1, + uint32_t ref, uint32_t mask); + +int amdgpu_mes_add_ring(struct amdgpu_device *adev, int gang_id, + int queue_type, int idx, + struct amdgpu_mes_ctx_data *ctx_data, + struct amdgpu_ring **out); +void amdgpu_mes_remove_ring(struct amdgpu_device *adev, + struct amdgpu_ring *ring); + +uint32_t amdgpu_mes_get_aggregated_doorbell_index(struct amdgpu_device *adev, + enum amdgpu_mes_priority_level prio); + +int amdgpu_mes_ctx_alloc_meta_data(struct amdgpu_device *adev, + struct amdgpu_mes_ctx_data *ctx_data); +void amdgpu_mes_ctx_free_meta_data(struct amdgpu_mes_ctx_data *ctx_data); +int amdgpu_mes_ctx_map_meta_data(struct amdgpu_device *adev, + struct amdgpu_vm *vm, + struct amdgpu_mes_ctx_data *ctx_data); +int amdgpu_mes_ctx_unmap_meta_data(struct amdgpu_device *adev, + struct amdgpu_mes_ctx_data *ctx_data); + +int amdgpu_mes_self_test(struct amdgpu_device *adev); + +int amdgpu_mes_alloc_process_doorbells(struct amdgpu_device *adev, + unsigned int *doorbell_index); +void amdgpu_mes_free_process_doorbells(struct amdgpu_device *adev, + unsigned int doorbell_index); +unsigned int amdgpu_mes_get_doorbell_dw_offset_in_bar( + struct amdgpu_device *adev, + uint32_t doorbell_index, + unsigned int doorbell_id); +int amdgpu_mes_doorbell_process_slice(struct amdgpu_device *adev); + +/* + * MES lock can be taken in MMU notifiers. + * + * A bit more detail about why to set no-FS reclaim with MES lock: + * + * The purpose of the MMU notifier is to stop GPU access to memory so + * that the Linux VM subsystem can move pages around safely. This is + * done by preempting user mode queues for the affected process. When + * MES is used, MES lock needs to be taken to preempt the queues. + * + * The MMU notifier callback entry point in the driver is + * amdgpu_mn_invalidate_range_start_hsa. The relevant call chain from + * there is: + * amdgpu_amdkfd_evict_userptr -> kgd2kfd_quiesce_mm -> + * kfd_process_evict_queues -> pdd->dev->dqm->ops.evict_process_queues + * + * The last part of the chain is a function pointer where we take the + * MES lock. + * + * The problem with taking locks in the MMU notifier is, that MMU + * notifiers can be called in reclaim-FS context. That's where the + * kernel frees up pages to make room for new page allocations under + * memory pressure. While we are running in reclaim-FS context, we must + * not trigger another memory reclaim operation because that would + * recursively reenter the reclaim code and cause a deadlock. The + * memalloc_nofs_save/restore calls guarantee that. + * + * In addition we also need to avoid lock dependencies on other locks taken + * under the MES lock, for example reservation locks. Here is a possible + * scenario of a deadlock: + * Thread A: takes and holds reservation lock | triggers reclaim-FS | + * MMU notifier | blocks trying to take MES lock + * Thread B: takes and holds MES lock | blocks trying to take reservation lock + * + * In this scenario Thread B gets involved in a deadlock even without + * triggering a reclaim-FS operation itself. + * To fix this and break the lock dependency chain you'd need to either: + * 1. protect reservation locks with memalloc_nofs_save/restore, or + * 2. avoid taking reservation locks under the MES lock. + * + * Reservation locks are taken all over the kernel in different subsystems, we + * have no control over them and their lock dependencies.So the only workable + * solution is to avoid taking other locks under the MES lock. + * As a result, make sure no reclaim-FS happens while holding this lock anywhere + * to prevent deadlocks when an MMU notifier runs in reclaim-FS context. + */ +static inline void amdgpu_mes_lock(struct amdgpu_mes *mes) +{ + mutex_lock(&mes->mutex_hidden); + mes->saved_flags = memalloc_noreclaim_save(); +} + +static inline void amdgpu_mes_unlock(struct amdgpu_mes *mes) +{ + memalloc_noreclaim_restore(mes->saved_flags); + mutex_unlock(&mes->mutex_hidden); +} #endif /* __AMDGPU_MES_H__ */ diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_mes_ctx.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_mes_ctx.h new file mode 100644 index 000000000000..912a5be2ece6 --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_mes_ctx.h @@ -0,0 +1,122 @@ +/* + * Copyright 2019 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#ifndef __AMDGPU_MES_CTX_H__ +#define __AMDGPU_MES_CTX_H__ + +#include "v10_structs.h" + +enum { + AMDGPU_MES_CTX_RPTR_OFFS = 0, + AMDGPU_MES_CTX_WPTR_OFFS, + AMDGPU_MES_CTX_FENCE_OFFS, + AMDGPU_MES_CTX_COND_EXE_OFFS, + AMDGPU_MES_CTX_TRAIL_FENCE_OFFS, + AMDGPU_MES_CTX_MAX_OFFS, +}; + +enum { + AMDGPU_MES_CTX_RING_OFFS = AMDGPU_MES_CTX_MAX_OFFS, + AMDGPU_MES_CTX_IB_OFFS, + AMDGPU_MES_CTX_PADDING_OFFS, +}; + +#define AMDGPU_MES_CTX_MAX_GFX_RINGS 1 +#define AMDGPU_MES_CTX_MAX_COMPUTE_RINGS 4 +#define AMDGPU_MES_CTX_MAX_SDMA_RINGS 2 +#define AMDGPU_MES_CTX_MAX_RINGS \ + (AMDGPU_MES_CTX_MAX_GFX_RINGS + \ + AMDGPU_MES_CTX_MAX_COMPUTE_RINGS + \ + AMDGPU_MES_CTX_MAX_SDMA_RINGS) + +#define AMDGPU_CSA_SDMA_SIZE 64 +#define GFX10_MEC_HPD_SIZE 2048 + +struct amdgpu_wb_slot { + uint32_t data[8]; +}; + +struct amdgpu_mes_ctx_meta_data { + struct { + uint8_t ring[PAGE_SIZE * 4]; + + /* gfx csa */ + struct v10_gfx_meta_data gfx_meta_data; + + uint8_t gds_backup[64 * 1024]; + + struct amdgpu_wb_slot slots[AMDGPU_MES_CTX_MAX_OFFS]; + + /* only for ib test */ + uint32_t ib[256] __aligned(256); + + uint32_t padding[64]; + + } __aligned(PAGE_SIZE) gfx[AMDGPU_MES_CTX_MAX_GFX_RINGS]; + + struct { + uint8_t ring[PAGE_SIZE * 4]; + + uint8_t mec_hpd[GFX10_MEC_HPD_SIZE]; + + struct amdgpu_wb_slot slots[AMDGPU_MES_CTX_MAX_OFFS]; + + /* only for ib test */ + uint32_t ib[256] __aligned(256); + + uint32_t padding[64]; + + } __aligned(PAGE_SIZE) compute[AMDGPU_MES_CTX_MAX_COMPUTE_RINGS]; + + struct { + uint8_t ring[PAGE_SIZE * 4]; + + /* sdma csa for mcbp */ + uint8_t sdma_meta_data[AMDGPU_CSA_SDMA_SIZE]; + + struct amdgpu_wb_slot slots[AMDGPU_MES_CTX_MAX_OFFS]; + + /* only for ib test */ + uint32_t ib[256] __aligned(256); + + uint32_t padding[64]; + + } __aligned(PAGE_SIZE) sdma[AMDGPU_MES_CTX_MAX_SDMA_RINGS]; +}; + +struct amdgpu_mes_ctx_data { + struct amdgpu_bo *meta_data_obj; + uint64_t meta_data_gpu_addr; + uint64_t meta_data_mc_addr; + struct amdgpu_bo_va *meta_data_va; + void *meta_data_ptr; + uint32_t gang_ids[AMDGPU_HW_IP_DMA+1]; +}; + +#define AMDGPU_FENCE_MES_QUEUE_FLAG 0x1000000u +#define AMDGPU_FENCE_MES_QUEUE_ID_MASK (AMDGPU_FENCE_MES_QUEUE_FLAG - 1) + +#define AMDGPU_FENCE_MES_QUEUE_FLAG 0x1000000u +#define AMDGPU_FENCE_MES_QUEUE_ID_MASK (AMDGPU_FENCE_MES_QUEUE_FLAG - 1) + +#endif diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_mmhub.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_mmhub.c deleted file mode 100644 index 24297dc51434..000000000000 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_mmhub.c +++ /dev/null @@ -1,68 +0,0 @@ -/* - * Copyright 2019 Advanced Micro Devices, Inc. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * the rights to use, copy, modify, merge, publish, distribute, sublicense, - * and/or sell copies of the Software, and to permit persons to whom the - * Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice shall be included in - * all copies or substantial portions of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL - * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR - * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, - * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR - * OTHER DEALINGS IN THE SOFTWARE. - * - */ - -#include "amdgpu.h" -#include "amdgpu_ras.h" - -int amdgpu_mmhub_ras_late_init(struct amdgpu_device *adev) -{ - int r; - struct ras_ih_if ih_info = { - .cb = NULL, - }; - struct ras_fs_if fs_info = { - .sysfs_name = "mmhub_err_count", - }; - - if (!adev->mmhub.ras_if) { - adev->mmhub.ras_if = kmalloc(sizeof(struct ras_common_if), GFP_KERNEL); - if (!adev->mmhub.ras_if) - return -ENOMEM; - adev->mmhub.ras_if->block = AMDGPU_RAS_BLOCK__MMHUB; - adev->mmhub.ras_if->type = AMDGPU_RAS_ERROR__MULTI_UNCORRECTABLE; - adev->mmhub.ras_if->sub_block_index = 0; - } - ih_info.head = fs_info.head = *adev->mmhub.ras_if; - r = amdgpu_ras_late_init(adev, adev->mmhub.ras_if, - &fs_info, &ih_info); - if (r || !amdgpu_ras_is_supported(adev, adev->mmhub.ras_if->block)) { - kfree(adev->mmhub.ras_if); - adev->mmhub.ras_if = NULL; - } - - return r; -} - -void amdgpu_mmhub_ras_fini(struct amdgpu_device *adev) -{ - if (amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__MMHUB) && - adev->mmhub.ras_if) { - struct ras_common_if *ras_if = adev->mmhub.ras_if; - struct ras_ih_if ih_info = { - .cb = NULL, - }; - - amdgpu_ras_late_fini(adev, ras_if, &ih_info); - kfree(ras_if); - } -} diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_mmhub.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_mmhub.h index b27fcbccce2b..93430d3823c9 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_mmhub.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_mmhub.h @@ -21,18 +21,13 @@ #ifndef __AMDGPU_MMHUB_H__ #define __AMDGPU_MMHUB_H__ -struct amdgpu_mmhub_ras_funcs { - int (*ras_late_init)(struct amdgpu_device *adev); - void (*ras_fini)(struct amdgpu_device *adev); - void (*query_ras_error_count)(struct amdgpu_device *adev, - void *ras_error_status); - void (*query_ras_error_status)(struct amdgpu_device *adev); - void (*reset_ras_error_count)(struct amdgpu_device *adev); - void (*reset_ras_error_status)(struct amdgpu_device *adev); +struct amdgpu_mmhub_ras { + struct amdgpu_ras_block_object ras_block; }; struct amdgpu_mmhub_funcs { u64 (*get_fb_location)(struct amdgpu_device *adev); + u64 (*get_mc_fb_offset)(struct amdgpu_device *adev); void (*init)(struct amdgpu_device *adev); int (*gart_enable)(struct amdgpu_device *adev); void (*set_fault_enable_default)(struct amdgpu_device *adev, @@ -40,7 +35,7 @@ struct amdgpu_mmhub_funcs { void (*gart_disable)(struct amdgpu_device *adev); int (*set_clockgating)(struct amdgpu_device *adev, enum amd_clockgating_state state); - void (*get_clockgating)(struct amdgpu_device *adev, u32 *flags); + void (*get_clockgating)(struct amdgpu_device *adev, u64 *flags); void (*setup_vm_pt_regs)(struct amdgpu_device *adev, uint32_t vmid, uint64_t page_table_base); void (*update_power_gating)(struct amdgpu_device *adev, @@ -50,10 +45,8 @@ struct amdgpu_mmhub_funcs { struct amdgpu_mmhub { struct ras_common_if *ras_if; const struct amdgpu_mmhub_funcs *funcs; - const struct amdgpu_mmhub_ras_funcs *ras_funcs; + struct amdgpu_mmhub_ras *ras; }; -int amdgpu_mmhub_ras_late_init(struct amdgpu_device *adev); -void amdgpu_mmhub_ras_fini(struct amdgpu_device *adev); #endif diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_mn.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_mn.c index 4b153daf283d..b86c0b8252a5 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_mn.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_mn.c @@ -75,8 +75,8 @@ static bool amdgpu_mn_invalidate_gfx(struct mmu_interval_notifier *mni, mmu_interval_set_seq(mni, cur_seq); - r = dma_resv_wait_timeout(bo->tbo.base.resv, true, false, - MAX_SCHEDULE_TIMEOUT); + r = dma_resv_wait_timeout(bo->tbo.base.resv, DMA_RESV_USAGE_BOOKKEEP, + false, MAX_SCHEDULE_TIMEOUT); mutex_unlock(&adev->notifier_lock); if (r <= 0) DRM_ERROR("(%ld) failed to wait for user bo\n", r); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_mode.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_mode.h index 6043bf6fd414..37322550d750 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_mode.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_mode.h @@ -30,21 +30,21 @@ #ifndef AMDGPU_MODE_H #define AMDGPU_MODE_H +#include <drm/display/drm_dp_helper.h> #include <drm/drm_crtc.h> #include <drm/drm_edid.h> #include <drm/drm_encoder.h> -#include <drm/drm_dp_helper.h> #include <drm/drm_fixed.h> #include <drm/drm_crtc_helper.h> #include <drm/drm_fb_helper.h> -#include <drm/drm_plane_helper.h> +#include <drm/drm_framebuffer.h> #include <drm/drm_probe_helper.h> #include <linux/i2c.h> #include <linux/i2c-algo-bit.h> #include <linux/hrtimer.h> #include "amdgpu_irq.h" -#include <drm/drm_dp_mst_helper.h> +#include <drm/display/drm_dp_mst_helper.h> #include "modules/inc/mod_freesync.h" #include "amdgpu_dm_irq_params.h" @@ -341,6 +341,7 @@ struct amdgpu_mode_info { int num_crtc; /* number of crtcs */ int num_hpd; /* number of hpd pins */ int num_dig; /* number of dig blocks */ + bool gpu_vm_support; /* supports display from GTT */ int disp_priority; const struct amdgpu_display_funcs *funcs; const enum drm_plane_type *plane_type; @@ -348,15 +349,11 @@ struct amdgpu_mode_info { #define AMDGPU_MAX_BL_LEVEL 0xFF -#if defined(CONFIG_BACKLIGHT_CLASS_DEVICE) || defined(CONFIG_BACKLIGHT_CLASS_DEVICE_MODULE) - struct amdgpu_backlight_privdata { struct amdgpu_encoder *encoder; uint8_t negative; }; -#endif - struct amdgpu_atom_ss { uint16_t percentage; uint16_t percentage_divider; @@ -591,19 +588,6 @@ int amdgpu_display_get_crtc_scanoutpos(struct drm_device *dev, int *hpos, ktime_t *stime, ktime_t *etime, const struct drm_display_mode *mode); -int amdgpu_display_gem_fb_init(struct drm_device *dev, - struct amdgpu_framebuffer *rfb, - const struct drm_mode_fb_cmd2 *mode_cmd, - struct drm_gem_object *obj); -int amdgpu_display_gem_fb_verify_and_init( - struct drm_device *dev, struct amdgpu_framebuffer *rfb, - struct drm_file *file_priv, const struct drm_mode_fb_cmd2 *mode_cmd, - struct drm_gem_object *obj); -int amdgpu_display_framebuffer_init(struct drm_device *dev, - struct amdgpu_framebuffer *rfb, - const struct drm_mode_fb_cmd2 *mode_cmd, - struct drm_gem_object *obj); - int amdgpufb_remove(struct drm_device *dev, struct drm_framebuffer *fb); void amdgpu_enc_destroy(struct drm_encoder *encoder); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_nbio.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_nbio.c index 6afb02fef8cf..37d779b8e4a6 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_nbio.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_nbio.c @@ -22,61 +22,24 @@ #include "amdgpu.h" #include "amdgpu_ras.h" -int amdgpu_nbio_ras_late_init(struct amdgpu_device *adev) +int amdgpu_nbio_ras_late_init(struct amdgpu_device *adev, struct ras_common_if *ras_block) { int r; - struct ras_ih_if ih_info = { - .cb = NULL, - }; - struct ras_fs_if fs_info = { - .sysfs_name = "pcie_bif_err_count", - }; - - if (!adev->nbio.ras_if) { - adev->nbio.ras_if = kmalloc(sizeof(struct ras_common_if), GFP_KERNEL); - if (!adev->nbio.ras_if) - return -ENOMEM; - adev->nbio.ras_if->block = AMDGPU_RAS_BLOCK__PCIE_BIF; - adev->nbio.ras_if->type = AMDGPU_RAS_ERROR__MULTI_UNCORRECTABLE; - adev->nbio.ras_if->sub_block_index = 0; - } - ih_info.head = fs_info.head = *adev->nbio.ras_if; - r = amdgpu_ras_late_init(adev, adev->nbio.ras_if, - &fs_info, &ih_info); + r = amdgpu_ras_block_late_init(adev, ras_block); if (r) - goto free; + return r; - if (amdgpu_ras_is_supported(adev, adev->nbio.ras_if->block)) { + if (amdgpu_ras_is_supported(adev, ras_block->block)) { r = amdgpu_irq_get(adev, &adev->nbio.ras_controller_irq, 0); if (r) goto late_fini; r = amdgpu_irq_get(adev, &adev->nbio.ras_err_event_athub_irq, 0); if (r) goto late_fini; - } else { - r = 0; - goto free; } return 0; late_fini: - amdgpu_ras_late_fini(adev, adev->nbio.ras_if, &ih_info); -free: - kfree(adev->nbio.ras_if); - adev->nbio.ras_if = NULL; + amdgpu_ras_block_late_fini(adev, ras_block); return r; } - -void amdgpu_nbio_ras_fini(struct amdgpu_device *adev) -{ - if (amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__PCIE_BIF) && - adev->nbio.ras_if) { - struct ras_common_if *ras_if = adev->nbio.ras_if; - struct ras_ih_if ih_info = { - .cb = NULL, - }; - - amdgpu_ras_late_fini(adev, ras_if, &ih_info); - kfree(ras_if); - } -} diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_nbio.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_nbio.h index 843052205bd5..a240336bbc6b 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_nbio.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_nbio.h @@ -47,15 +47,12 @@ struct nbio_hdp_flush_reg { u32 ref_and_mask_sdma7; }; -struct amdgpu_nbio_ras_funcs { +struct amdgpu_nbio_ras { + struct amdgpu_ras_block_object ras_block; void (*handle_ras_controller_intr_no_bifring)(struct amdgpu_device *adev); void (*handle_ras_err_event_athub_intr_no_bifring)(struct amdgpu_device *adev); int (*init_ras_controller_interrupt)(struct amdgpu_device *adev); int (*init_ras_err_event_athub_interrupt)(struct amdgpu_device *adev); - void (*query_ras_error_count)(struct amdgpu_device *adev, - void *ras_error_status); - int (*ras_late_init)(struct amdgpu_device *adev); - void (*ras_fini)(struct amdgpu_device *adev); }; struct amdgpu_nbio_funcs { @@ -73,6 +70,7 @@ struct amdgpu_nbio_funcs { bool use_doorbell, int doorbell_index, int doorbell_size); void (*vcn_doorbell_range)(struct amdgpu_device *adev, bool use_doorbell, int doorbell_index, int instance); + void (*gc_doorbell_init)(struct amdgpu_device *adev); void (*enable_doorbell_aperture)(struct amdgpu_device *adev, bool enable); void (*enable_doorbell_selfring_aperture)(struct amdgpu_device *adev, @@ -86,7 +84,7 @@ struct amdgpu_nbio_funcs { void (*update_medium_grain_light_sleep)(struct amdgpu_device *adev, bool enable); void (*get_clockgating_state)(struct amdgpu_device *adev, - u32 *flags); + u64 *flags); void (*ih_control)(struct amdgpu_device *adev); void (*init_registers)(struct amdgpu_device *adev); void (*remap_hdp_registers)(struct amdgpu_device *adev); @@ -96,6 +94,7 @@ struct amdgpu_nbio_funcs { void (*apply_lc_spc_mode_wa)(struct amdgpu_device *adev); void (*apply_l1_link_width_reconfig_wa)(struct amdgpu_device *adev); void (*clear_doorbell_interrupt)(struct amdgpu_device *adev); + u32 (*get_rom_offset)(struct amdgpu_device *adev); }; struct amdgpu_nbio { @@ -104,9 +103,8 @@ struct amdgpu_nbio { struct amdgpu_irq_src ras_err_event_athub_irq; struct ras_common_if *ras_if; const struct amdgpu_nbio_funcs *funcs; - const struct amdgpu_nbio_ras_funcs *ras_funcs; + struct amdgpu_nbio_ras *ras; }; -int amdgpu_nbio_ras_late_init(struct amdgpu_device *adev); -void amdgpu_nbio_ras_fini(struct amdgpu_device *adev); +int amdgpu_nbio_ras_late_init(struct amdgpu_device *adev, struct ras_common_if *ras_block); #endif diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c index 3a7b56e57cec..2e8f6cd7a729 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c @@ -33,6 +33,7 @@ #include <linux/slab.h> #include <linux/dma-buf.h> +#include <drm/drm_drv.h> #include <drm/amdgpu_drm.h> #include <drm/drm_cache.h> #include "amdgpu.h" @@ -450,7 +451,7 @@ static bool amdgpu_bo_validate_size(struct amdgpu_device *adev, if (domain & AMDGPU_GEM_DOMAIN_GTT) { man = ttm_manager_type(&adev->mman.bdev, TTM_PL_TT); - if (size < (man->size << PAGE_SHIFT)) + if (size < man->size) return true; else goto fail; @@ -459,7 +460,7 @@ static bool amdgpu_bo_validate_size(struct amdgpu_device *adev, if (domain & AMDGPU_GEM_DOMAIN_VRAM) { man = ttm_manager_type(&adev->mman.bdev, TTM_PL_VRAM); - if (size < (man->size << PAGE_SHIFT)) + if (size < man->size) return true; else goto fail; @@ -471,7 +472,7 @@ static bool amdgpu_bo_validate_size(struct amdgpu_device *adev, fail: DRM_DEBUG("BO size %lu > total memory in domain: %llu\n", size, - man->size << PAGE_SHIFT); + man->size); return false; } @@ -566,6 +567,7 @@ int amdgpu_bo_create(struct amdgpu_device *adev, bp->domain; bo->allowed_domains = bo->preferred_domains; if (bp->type != ttm_bo_type_kernel && + !(bp->flags & AMDGPU_GEM_CREATE_DISCARDABLE) && bo->allowed_domains == AMDGPU_GEM_DOMAIN_VRAM) bo->allowed_domains |= AMDGPU_GEM_DOMAIN_GTT; @@ -574,6 +576,9 @@ int amdgpu_bo_create(struct amdgpu_device *adev, if (!amdgpu_bo_support_uswc(bo->flags)) bo->flags &= ~AMDGPU_GEM_CREATE_CPU_GTT_USWC; + if (adev->ras_enabled) + bo->flags |= AMDGPU_GEM_CREATE_VRAM_WIPE_ON_RELEASE; + bo->tbo.bdev = &adev->mman.bdev; if (bp->domain & (AMDGPU_GEM_DOMAIN_GWS | AMDGPU_GEM_DOMAIN_OA | AMDGPU_GEM_DOMAIN_GDS)) @@ -586,7 +591,7 @@ int amdgpu_bo_create(struct amdgpu_device *adev, if (!bp->destroy) bp->destroy = &amdgpu_bo_destroy; - r = ttm_bo_init_reserved(&adev->mman.bdev, &bo->tbo, size, bp->type, + r = ttm_bo_init_reserved(&adev->mman.bdev, &bo->tbo, bp->type, &bo->placement, page_align, &ctx, NULL, bp->resv, bp->destroy); if (unlikely(r != 0)) @@ -608,9 +613,8 @@ int amdgpu_bo_create(struct amdgpu_device *adev, if (unlikely(r)) goto fail_unreserve; - amdgpu_bo_fence(bo, fence, false); - dma_fence_put(bo->tbo.moving); - bo->tbo.moving = dma_fence_get(fence); + dma_resv_add_fence(bo->tbo.base.resv, fence, + DMA_RESV_USAGE_KERNEL); dma_fence_put(fence); } if (!bp->resv) @@ -684,13 +688,16 @@ int amdgpu_bo_create_vm(struct amdgpu_device *adev, * num of amdgpu_vm_pt entries. */ BUG_ON(bp->bo_ptr_size < sizeof(struct amdgpu_bo_vm)); - bp->destroy = &amdgpu_bo_vm_destroy; r = amdgpu_bo_create(adev, bp, &bo_ptr); if (r) return r; *vmbo_ptr = to_amdgpu_bo_vm(bo_ptr); INIT_LIST_HEAD(&(*vmbo_ptr)->shadow_list); + /* Set destroy callback to amdgpu_bo_vm_destroy after vmbo->shadow_list + * is initialized. + */ + bo_ptr->tbo.destroy = &amdgpu_bo_vm_destroy; return r; } @@ -757,6 +764,11 @@ int amdgpu_bo_kmap(struct amdgpu_bo *bo, void **ptr) if (bo->flags & AMDGPU_GEM_CREATE_NO_CPU_ACCESS) return -EPERM; + r = dma_resv_wait_timeout(bo->tbo.base.resv, DMA_RESV_USAGE_KERNEL, + false, MAX_SCHEDULE_TIMEOUT); + if (r < 0) + return r; + kptr = amdgpu_bo_kptr(bo); if (kptr) { if (ptr) @@ -764,11 +776,6 @@ int amdgpu_bo_kmap(struct amdgpu_bo *bo, void **ptr) return 0; } - r = dma_resv_wait_timeout(bo->tbo.base.resv, false, false, - MAX_SCHEDULE_TIMEOUT); - if (r < 0) - return r; - r = ttm_bo_kmap(&bo->tbo, 0, bo->tbo.resource->num_pages, &bo->kmap); if (r) return r; @@ -878,6 +885,10 @@ int amdgpu_bo_pin_restricted(struct amdgpu_bo *bo, u32 domain, if (WARN_ON_ONCE(min_offset > max_offset)) return -EINVAL; + /* Check domain to be pinned to against preferred domains */ + if (bo->preferred_domains & domain) + domain = bo->preferred_domains & domain; + /* A shared bo cannot be migrated to VRAM */ if (bo->tbo.base.import_attach) { if (domain & AMDGPU_GEM_DOMAIN_GTT) @@ -1015,7 +1026,9 @@ static const char *amdgpu_vram_names[] = { "DDR3", "DDR4", "GDDR6", - "DDR5" + "DDR5", + "LPDDR4", + "LPDDR5" }; /** @@ -1061,7 +1074,18 @@ int amdgpu_bo_init(struct amdgpu_device *adev) */ void amdgpu_bo_fini(struct amdgpu_device *adev) { + int idx; + amdgpu_ttm_fini(adev); + + if (drm_dev_enter(adev_to_drm(adev), &idx)) { + + if (!adev->gmc.xgmi.connected_to_cpu) { + arch_phys_wc_del(adev->gmc.vram_mtrr); + arch_io_free_memtype_wc(adev->gmc.aper_base, adev->gmc.aper_size); + } + drm_dev_exit(idx); + } } /** @@ -1269,6 +1293,7 @@ void amdgpu_bo_get_memory(struct amdgpu_bo *bo, uint64_t *vram_mem, */ void amdgpu_bo_release_notify(struct ttm_buffer_object *bo) { + struct amdgpu_device *adev = amdgpu_ttm_adev(bo->bdev); struct dma_fence *fence = NULL; struct amdgpu_bo *abo; int r; @@ -1287,11 +1312,13 @@ void amdgpu_bo_release_notify(struct ttm_buffer_object *bo) if (bo->base.resv == &bo->base._resv) amdgpu_amdkfd_remove_fence_on_pt_pd_bos(abo); - if (bo->resource->mem_type != TTM_PL_VRAM || - !(abo->flags & AMDGPU_GEM_CREATE_VRAM_WIPE_ON_RELEASE)) + if (!bo->resource || bo->resource->mem_type != TTM_PL_VRAM || + !(abo->flags & AMDGPU_GEM_CREATE_VRAM_WIPE_ON_RELEASE) || + adev->in_suspend || adev->shutdown) return; - dma_resv_lock(bo->base.resv, NULL); + if (WARN_ON_ONCE(!dma_resv_trylock(bo->base.resv))) + return; r = amdgpu_fill_buffer(abo, AMDGPU_POISON, bo->base.resv, &fence); if (!WARN_ON(r)) { @@ -1372,11 +1399,17 @@ void amdgpu_bo_fence(struct amdgpu_bo *bo, struct dma_fence *fence, bool shared) { struct dma_resv *resv = bo->tbo.base.resv; + int r; - if (shared) - dma_resv_add_shared_fence(resv, fence); - else - dma_resv_add_excl_fence(resv, fence); + r = dma_resv_reserve_fences(resv, 1); + if (r) { + /* As last resort on OOM we block for the fence */ + dma_fence_wait(fence, false); + return; + } + + dma_resv_add_fence(resv, fence, shared ? DMA_RESV_USAGE_READ : + DMA_RESV_USAGE_WRITE); } /** diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h index 4c9cbdc66995..147b79c10cbb 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h @@ -41,7 +41,6 @@ /* BO flag to indicate a KFD userptr BO */ #define AMDGPU_AMDKFD_CREATE_USERPTR_BO (1ULL << 63) -#define AMDGPU_AMDKFD_CREATE_SVM_BO (1ULL << 62) #define to_amdgpu_bo_user(abo) container_of((abo), struct amdgpu_bo_user, bo) #define to_amdgpu_bo_vm(abo) container_of((abo), struct amdgpu_bo_vm, bo) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_preempt_mgr.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_preempt_mgr.c index 786afe4f58f9..e8adfd0a570a 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_preempt_mgr.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_preempt_mgr.c @@ -25,12 +25,6 @@ #include "amdgpu.h" -static inline struct amdgpu_preempt_mgr * -to_preempt_mgr(struct ttm_resource_manager *man) -{ - return container_of(man, struct amdgpu_preempt_mgr, manager); -} - /** * DOC: mem_info_preempt_used * @@ -45,10 +39,9 @@ static ssize_t mem_info_preempt_used_show(struct device *dev, { struct drm_device *ddev = dev_get_drvdata(dev); struct amdgpu_device *adev = drm_to_adev(ddev); - struct ttm_resource_manager *man; + struct ttm_resource_manager *man = &adev->mman.preempt_mgr; - man = ttm_manager_type(&adev->mman.bdev, AMDGPU_PL_PREEMPT); - return sysfs_emit(buf, "%llu\n", amdgpu_preempt_mgr_usage(man)); + return sysfs_emit(buf, "%llu\n", ttm_resource_manager_usage(man)); } static DEVICE_ATTR_RO(mem_info_preempt_used); @@ -68,16 +61,12 @@ static int amdgpu_preempt_mgr_new(struct ttm_resource_manager *man, const struct ttm_place *place, struct ttm_resource **res) { - struct amdgpu_preempt_mgr *mgr = to_preempt_mgr(man); - *res = kzalloc(sizeof(**res), GFP_KERNEL); if (!*res) return -ENOMEM; ttm_resource_init(tbo, place, *res); (*res)->start = AMDGPU_BO_INVALID_OFFSET; - - atomic64_add((*res)->num_pages, &mgr->used); return 0; } @@ -92,48 +81,13 @@ static int amdgpu_preempt_mgr_new(struct ttm_resource_manager *man, static void amdgpu_preempt_mgr_del(struct ttm_resource_manager *man, struct ttm_resource *res) { - struct amdgpu_preempt_mgr *mgr = to_preempt_mgr(man); - - atomic64_sub(res->num_pages, &mgr->used); + ttm_resource_fini(man, res); kfree(res); } -/** - * amdgpu_preempt_mgr_usage - return usage of PREEMPT domain - * - * @man: TTM memory type manager - * - * Return how many bytes are used in the GTT domain - */ -uint64_t amdgpu_preempt_mgr_usage(struct ttm_resource_manager *man) -{ - struct amdgpu_preempt_mgr *mgr = to_preempt_mgr(man); - s64 result = atomic64_read(&mgr->used); - - return (result > 0 ? result : 0) * PAGE_SIZE; -} - -/** - * amdgpu_preempt_mgr_debug - dump VRAM table - * - * @man: TTM memory type manager - * @printer: DRM printer to use - * - * Dump the table content using printk. - */ -static void amdgpu_preempt_mgr_debug(struct ttm_resource_manager *man, - struct drm_printer *printer) -{ - struct amdgpu_preempt_mgr *mgr = to_preempt_mgr(man); - - drm_printf(printer, "man size:%llu pages, preempt used:%lld pages\n", - man->size, (u64)atomic64_read(&mgr->used)); -} - static const struct ttm_resource_manager_func amdgpu_preempt_mgr_func = { .alloc = amdgpu_preempt_mgr_new, .free = amdgpu_preempt_mgr_del, - .debug = amdgpu_preempt_mgr_debug }; /** @@ -145,16 +99,13 @@ static const struct ttm_resource_manager_func amdgpu_preempt_mgr_func = { */ int amdgpu_preempt_mgr_init(struct amdgpu_device *adev) { - struct amdgpu_preempt_mgr *mgr = &adev->mman.preempt_mgr; - struct ttm_resource_manager *man = &mgr->manager; + struct ttm_resource_manager *man = &adev->mman.preempt_mgr; int ret; man->use_tt = true; man->func = &amdgpu_preempt_mgr_func; - ttm_resource_manager_init(man, (1 << 30)); - - atomic64_set(&mgr->used, 0); + ttm_resource_manager_init(man, &adev->mman.bdev, (1 << 30)); ret = device_create_file(adev->dev, &dev_attr_mem_info_preempt_used); if (ret) { @@ -162,8 +113,7 @@ int amdgpu_preempt_mgr_init(struct amdgpu_device *adev) return ret; } - ttm_set_driver_manager(&adev->mman.bdev, AMDGPU_PL_PREEMPT, - &mgr->manager); + ttm_set_driver_manager(&adev->mman.bdev, AMDGPU_PL_PREEMPT, man); ttm_resource_manager_set_used(man, true); return 0; } @@ -178,8 +128,7 @@ int amdgpu_preempt_mgr_init(struct amdgpu_device *adev) */ void amdgpu_preempt_mgr_fini(struct amdgpu_device *adev) { - struct amdgpu_preempt_mgr *mgr = &adev->mman.preempt_mgr; - struct ttm_resource_manager *man = &mgr->manager; + struct ttm_resource_manager *man = &adev->mman.preempt_mgr; int ret; ttm_resource_manager_set_used(man, false); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c index dee17a0e1187..effa7df3ddbf 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c @@ -37,17 +37,18 @@ #include "psp_v11_0_8.h" #include "psp_v12_0.h" #include "psp_v13_0.h" +#include "psp_v13_0_4.h" #include "amdgpu_ras.h" #include "amdgpu_securedisplay.h" #include "amdgpu_atomfirmware.h" +#define AMD_VBIOS_FILE_MAX_SIZE_B (1024*1024*3) + static int psp_sysfs_init(struct amdgpu_device *adev); static void psp_sysfs_fini(struct amdgpu_device *adev); static int psp_load_smu_fw(struct psp_context *psp); -static int psp_ta_unload(struct psp_context *psp, struct ta_context *context); -static int psp_ta_load(struct psp_context *psp, struct ta_context *context); static int psp_rap_terminate(struct psp_context *psp); static int psp_securedisplay_terminate(struct psp_context *psp); @@ -84,7 +85,9 @@ static void psp_check_pmfw_centralized_cstate_management(struct psp_context *psp case IP_VERSION(11, 0, 11): case IP_VERSION(11, 0, 12): case IP_VERSION(11, 0, 13): + case IP_VERSION(13, 0, 0): case IP_VERSION(13, 0, 2): + case IP_VERSION(13, 0, 7): psp->pmfw_centralized_cstate_management = true; break; default: @@ -133,6 +136,9 @@ static int psp_early_init(void *handle) break; case IP_VERSION(13, 0, 1): case IP_VERSION(13, 0, 3): + case IP_VERSION(13, 0, 5): + case IP_VERSION(13, 0, 8): + case IP_VERSION(13, 0, 10): psp_v13_0_set_psp_funcs(psp); psp->autoload_supported = true; break; @@ -142,6 +148,15 @@ static int psp_early_init(void *handle) psp->autoload_supported = false; } break; + case IP_VERSION(13, 0, 0): + case IP_VERSION(13, 0, 7): + psp_v13_0_set_psp_funcs(psp); + psp->autoload_supported = true; + break; + case IP_VERSION(13, 0, 4): + psp_v13_0_4_set_psp_funcs(psp); + psp->autoload_supported = true; + break; default: return -EINVAL; } @@ -153,6 +168,42 @@ static int psp_early_init(void *handle) return 0; } +void psp_ta_free_shared_buf(struct ta_mem_context *mem_ctx) +{ + amdgpu_bo_free_kernel(&mem_ctx->shared_bo, &mem_ctx->shared_mc_addr, + &mem_ctx->shared_buf); +} + +static void psp_free_shared_bufs(struct psp_context *psp) +{ + void *tmr_buf; + void **pptr; + + /* free TMR memory buffer */ + pptr = amdgpu_sriov_vf(psp->adev) ? &tmr_buf : NULL; + amdgpu_bo_free_kernel(&psp->tmr_bo, &psp->tmr_mc_addr, pptr); + + /* free xgmi shared memory */ + psp_ta_free_shared_buf(&psp->xgmi_context.context.mem_context); + + /* free ras shared memory */ + psp_ta_free_shared_buf(&psp->ras_context.context.mem_context); + + /* free hdcp shared memory */ + psp_ta_free_shared_buf(&psp->hdcp_context.context.mem_context); + + /* free dtm shared memory */ + psp_ta_free_shared_buf(&psp->dtm_context.context.mem_context); + + /* free rap shared memory */ + psp_ta_free_shared_buf(&psp->rap_context.context.mem_context); + + /* free securedisplay shared memory */ + psp_ta_free_shared_buf(&psp->securedisplay_context.context.mem_context); + + +} + static void psp_memory_training_fini(struct psp_context *psp) { struct psp_memory_training_context *ctx = &psp->mem_train_ctx; @@ -241,7 +292,7 @@ static bool psp_get_runtime_db_entry(struct amdgpu_device *adev, case PSP_RUNTIME_ENTRY_TYPE_BOOT_CONFIG: if (db_dir.entry_list[i].size < sizeof(struct psp_runtime_boot_cfg_entry)) { /* invalid db entry size */ - dev_warn(adev->dev, "Invalid PSP runtime database entry size\n"); + dev_warn(adev->dev, "Invalid PSP runtime database boot cfg entry size\n"); return false; } /* read runtime database entry */ @@ -249,6 +300,17 @@ static bool psp_get_runtime_db_entry(struct amdgpu_device *adev, (uint32_t *)db_entry, sizeof(struct psp_runtime_boot_cfg_entry), false); ret = true; break; + case PSP_RUNTIME_ENTRY_TYPE_PPTABLE_ERR_STATUS: + if (db_dir.entry_list[i].size < sizeof(struct psp_runtime_scpm_entry)) { + /* invalid db entry size */ + dev_warn(adev->dev, "Invalid PSP runtime database scpm entry size\n"); + return false; + } + /* read runtime database entry */ + amdgpu_device_vram_access(adev, db_header_pos + db_dir.entry_list[i].offset, + (uint32_t *)db_entry, sizeof(struct psp_runtime_scpm_entry), false); + ret = true; + break; default: ret = false; break; @@ -259,6 +321,42 @@ static bool psp_get_runtime_db_entry(struct amdgpu_device *adev, return ret; } +static int psp_init_sriov_microcode(struct psp_context *psp) +{ + struct amdgpu_device *adev = psp->adev; + int ret = 0; + + switch (adev->ip_versions[MP0_HWIP][0]) { + case IP_VERSION(9, 0, 0): + adev->virt.autoload_ucode_id = AMDGPU_UCODE_ID_CP_MEC2; + ret = psp_init_cap_microcode(psp, "vega10"); + break; + case IP_VERSION(11, 0, 9): + adev->virt.autoload_ucode_id = AMDGPU_UCODE_ID_CP_MEC2; + ret = psp_init_cap_microcode(psp, "navi12"); + break; + case IP_VERSION(11, 0, 7): + adev->virt.autoload_ucode_id = AMDGPU_UCODE_ID_CP_MEC2; + ret = psp_init_cap_microcode(psp, "sienna_cichlid"); + break; + case IP_VERSION(13, 0, 2): + adev->virt.autoload_ucode_id = AMDGPU_UCODE_ID_CP_MEC2; + ret = psp_init_cap_microcode(psp, "aldebaran"); + ret &= psp_init_ta_microcode(psp, "aldebaran"); + break; + case IP_VERSION(13, 0, 0): + adev->virt.autoload_ucode_id = 0; + break; + case IP_VERSION(13, 0, 10): + adev->virt.autoload_ucode_id = AMDGPU_UCODE_ID_CP_MES1_DATA; + break; + default: + BUG(); + break; + } + return ret; +} + static int psp_sw_init(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; @@ -266,6 +364,7 @@ static int psp_sw_init(void *handle) int ret; struct psp_runtime_boot_cfg_entry boot_cfg_entry; struct psp_memory_training_context *mem_training_ctx = &psp->mem_train_ctx; + struct psp_runtime_scpm_entry scpm_entry; psp->cmd = kzalloc(sizeof(struct psp_gfx_cmd_resp), GFP_KERNEL); if (!psp->cmd) { @@ -273,21 +372,33 @@ static int psp_sw_init(void *handle) ret = -ENOMEM; } - if (!amdgpu_sriov_vf(adev)) { + if (amdgpu_sriov_vf(adev)) + ret = psp_init_sriov_microcode(psp); + else ret = psp_init_microcode(psp); - if (ret) { - DRM_ERROR("Failed to load psp firmware!\n"); - return ret; - } - } else if (amdgpu_sriov_vf(adev) && - adev->ip_versions[MP0_HWIP][0] == IP_VERSION(13, 0, 2)) { - ret = psp_init_ta_microcode(psp, "aldebaran"); - if (ret) { - DRM_ERROR("Failed to initialize ta microcode!\n"); - return ret; - } + if (ret) { + DRM_ERROR("Failed to load psp firmware!\n"); + return ret; + } + + adev->psp.xgmi_context.supports_extended_data = + !adev->gmc.xgmi.connected_to_cpu && + adev->ip_versions[MP0_HWIP][0] == IP_VERSION(13, 0, 2); + + memset(&scpm_entry, 0, sizeof(scpm_entry)); + if ((psp_get_runtime_db_entry(adev, + PSP_RUNTIME_ENTRY_TYPE_PPTABLE_ERR_STATUS, + &scpm_entry)) && + (SCPM_DISABLE != scpm_entry.scpm_status)) { + adev->scpm_enabled = true; + adev->scpm_status = scpm_entry.scpm_status; + } else { + adev->scpm_enabled = false; + adev->scpm_status = SCPM_DISABLE; } + /* TODO: stop gpu driver services and print alarm if scpm is enabled with error status */ + memset(&boot_cfg_entry, 0, sizeof(boot_cfg_entry)); if (psp_get_runtime_db_entry(adev, PSP_RUNTIME_ENTRY_TYPE_BOOT_CONFIG, @@ -331,7 +442,39 @@ static int psp_sw_init(void *handle) } } + ret = amdgpu_bo_create_kernel(adev, PSP_1_MEG, PSP_1_MEG, + amdgpu_sriov_vf(adev) ? + AMDGPU_GEM_DOMAIN_VRAM : AMDGPU_GEM_DOMAIN_GTT, + &psp->fw_pri_bo, + &psp->fw_pri_mc_addr, + &psp->fw_pri_buf); + if (ret) + return ret; + + ret = amdgpu_bo_create_kernel(adev, PSP_FENCE_BUFFER_SIZE, PAGE_SIZE, + AMDGPU_GEM_DOMAIN_VRAM, + &psp->fence_buf_bo, + &psp->fence_buf_mc_addr, + &psp->fence_buf); + if (ret) + goto failed1; + + ret = amdgpu_bo_create_kernel(adev, PSP_CMD_BUFFER_SIZE, PAGE_SIZE, + AMDGPU_GEM_DOMAIN_VRAM, + &psp->cmd_buf_bo, &psp->cmd_buf_mc_addr, + (void **)&psp->cmd_buf_mem); + if (ret) + goto failed2; + return 0; + +failed2: + amdgpu_bo_free_kernel(&psp->fw_pri_bo, + &psp->fw_pri_mc_addr, &psp->fw_pri_buf); +failed1: + amdgpu_bo_free_kernel(&psp->fence_buf_bo, + &psp->fence_buf_mc_addr, &psp->fence_buf); + return ret; } static int psp_sw_fini(void *handle) @@ -353,7 +496,14 @@ static int psp_sw_fini(void *handle) release_firmware(psp->ta_fw); psp->ta_fw = NULL; } - + if (psp->cap_fw) { + release_firmware(psp->cap_fw); + psp->cap_fw = NULL; + } + if (psp->toc_fw) { + release_firmware(psp->toc_fw); + psp->toc_fw = NULL; + } if (adev->ip_versions[MP0_HWIP][0] == IP_VERSION(11, 0, 0) || adev->ip_versions[MP0_HWIP][0] == IP_VERSION(11, 0, 7)) psp_sysfs_fini(adev); @@ -361,6 +511,18 @@ static int psp_sw_fini(void *handle) kfree(cmd); cmd = NULL; + if (psp->km_ring.ring_mem) + amdgpu_bo_free_kernel(&adev->firmware.rbuf, + &psp->km_ring.ring_mem_mc_addr, + (void **)&psp->km_ring.ring_mem); + + amdgpu_bo_free_kernel(&psp->fw_pri_bo, + &psp->fw_pri_mc_addr, &psp->fw_pri_buf); + amdgpu_bo_free_kernel(&psp->fence_buf_bo, + &psp->fence_buf_mc_addr, &psp->fence_buf); + amdgpu_bo_free_kernel(&psp->cmd_buf_bo, &psp->cmd_buf_mc_addr, + (void **)&psp->cmd_buf_mem); + return 0; } @@ -491,7 +653,11 @@ psp_cmd_submit_buf(struct psp_context *psp, DRM_WARN("psp gfx command %s(0x%X) failed and response status is (0x%X)\n", psp_gfx_cmd_name(psp->cmd_buf_mem->cmd_id), psp->cmd_buf_mem->cmd_id, psp->cmd_buf_mem->resp.status); - if (!timeout) { + /* If any firmware (including CAP) load fails under SRIOV, it should + * return failure to stop the VF from initializing. + * Also return failure in case of timeout + */ + if ((ucode && amdgpu_sriov_vf(psp->adev)) || !timeout) { ret = -EINVAL; goto exit; } @@ -605,7 +771,7 @@ static int psp_tmr_init(struct psp_context *psp) } pptr = amdgpu_sriov_vf(psp->adev) ? &tmr_buf : NULL; - ret = amdgpu_bo_create_kernel(psp->adev, tmr_size, PSP_TMR_SIZE(psp->adev), + ret = amdgpu_bo_create_kernel(psp->adev, tmr_size, PSP_TMR_ALIGNMENT, AMDGPU_GEM_DOMAIN_VRAM, &psp->tmr_bo, &psp->tmr_mc_addr, pptr); @@ -618,6 +784,7 @@ static bool psp_skip_tmr(struct psp_context *psp) case IP_VERSION(11, 0, 9): case IP_VERSION(11, 0, 7): case IP_VERSION(13, 0, 2): + case IP_VERSION(13, 0, 10): return true; default: return false; @@ -664,7 +831,7 @@ static int psp_tmr_unload(struct psp_context *psp) struct psp_gfx_cmd_resp *cmd = acquire_psp_cmd_buf(psp); psp_prep_tmr_unload_cmd_buf(psp, cmd); - DRM_INFO("free PSP TMR buffer\n"); + dev_info(psp->adev->dev, "free PSP TMR buffer\n"); ret = psp_cmd_submit_buf(psp, NULL, cmd, psp->fence_buf_mc_addr); @@ -676,19 +843,7 @@ static int psp_tmr_unload(struct psp_context *psp) static int psp_tmr_terminate(struct psp_context *psp) { - int ret; - void *tmr_buf; - void **pptr; - - ret = psp_tmr_unload(psp); - if (ret) - return ret; - - /* free TMR memory buffer */ - pptr = amdgpu_sriov_vf(psp->adev) ? &tmr_buf : NULL; - amdgpu_bo_free_kernel(&psp->tmr_bo, &psp->tmr_mc_addr, pptr); - - return 0; + return psp_tmr_unload(psp); } int psp_get_fw_attestation_records_addr(struct psp_context *psp, @@ -795,11 +950,6 @@ static int psp_rl_load(struct amdgpu_device *adev) return ret; } -static int psp_asd_load(struct psp_context *psp) -{ - return psp_ta_load(psp, &psp->asd_context); -} - static int psp_asd_initialize(struct psp_context *psp) { int ret; @@ -815,7 +965,7 @@ static int psp_asd_initialize(struct psp_context *psp) psp->asd_context.mem_context.shared_mem_size = PSP_ASD_SHARED_MEM_SIZE; psp->asd_context.ta_load_type = GFX_CMD_ID_LOAD_ASD; - ret = psp_asd_load(psp); + ret = psp_ta_load(psp, &psp->asd_context); if (!ret) psp->asd_context.initialized = true; @@ -829,7 +979,7 @@ static void psp_prep_ta_unload_cmd_buf(struct psp_gfx_cmd_resp *cmd, cmd->cmd.cmd_unload_ta.session_id = session_id; } -static int psp_ta_unload(struct psp_context *psp, struct ta_context *context) +int psp_ta_unload(struct psp_context *psp, struct ta_context *context) { int ret; struct psp_gfx_cmd_resp *cmd = acquire_psp_cmd_buf(psp); @@ -843,11 +993,6 @@ static int psp_ta_unload(struct psp_context *psp, struct ta_context *context) return ret; } -static int psp_asd_unload(struct psp_context *psp) -{ - return psp_ta_unload(psp, &psp->asd_context); -} - static int psp_asd_terminate(struct psp_context *psp) { int ret; @@ -858,8 +1003,7 @@ static int psp_asd_terminate(struct psp_context *psp) if (!psp->asd_context.initialized) return 0; - ret = psp_asd_unload(psp); - + ret = psp_ta_unload(psp, &psp->asd_context); if (!ret) psp->asd_context.initialized = false; @@ -911,33 +1055,54 @@ static void psp_prep_ta_load_cmd_buf(struct psp_gfx_cmd_resp *cmd, cmd->cmd.cmd_load_ta.cmd_buf_len = context->mem_context.shared_mem_size; } -static int psp_ta_init_shared_buf(struct psp_context *psp, +int psp_ta_init_shared_buf(struct psp_context *psp, struct ta_mem_context *mem_ctx) { - int ret; - /* * Allocate 16k memory aligned to 4k from Frame Buffer (local * physical) for ta to host memory */ - ret = amdgpu_bo_create_kernel(psp->adev, mem_ctx->shared_mem_size, + return amdgpu_bo_create_kernel(psp->adev, mem_ctx->shared_mem_size, PAGE_SIZE, AMDGPU_GEM_DOMAIN_VRAM, &mem_ctx->shared_bo, &mem_ctx->shared_mc_addr, &mem_ctx->shared_buf); - - return ret; } -static void psp_ta_free_shared_buf(struct ta_mem_context *mem_ctx) +static void psp_prep_ta_invoke_indirect_cmd_buf(struct psp_gfx_cmd_resp *cmd, + uint32_t ta_cmd_id, + struct ta_context *context) { - amdgpu_bo_free_kernel(&mem_ctx->shared_bo, &mem_ctx->shared_mc_addr, - &mem_ctx->shared_buf); + cmd->cmd_id = GFX_CMD_ID_INVOKE_CMD; + cmd->cmd.cmd_invoke_cmd.session_id = context->session_id; + cmd->cmd.cmd_invoke_cmd.ta_cmd_id = ta_cmd_id; + + cmd->cmd.cmd_invoke_cmd.buf.num_desc = 1; + cmd->cmd.cmd_invoke_cmd.buf.total_size = context->mem_context.shared_mem_size; + cmd->cmd.cmd_invoke_cmd.buf.buf_desc[0].buf_size = context->mem_context.shared_mem_size; + cmd->cmd.cmd_invoke_cmd.buf.buf_desc[0].buf_phy_addr_lo = + lower_32_bits(context->mem_context.shared_mc_addr); + cmd->cmd.cmd_invoke_cmd.buf.buf_desc[0].buf_phy_addr_hi = + upper_32_bits(context->mem_context.shared_mc_addr); } -static int psp_xgmi_init_shared_buf(struct psp_context *psp) +int psp_ta_invoke_indirect(struct psp_context *psp, + uint32_t ta_cmd_id, + struct ta_context *context) { - return psp_ta_init_shared_buf(psp, &psp->xgmi_context.context.mem_context); + int ret; + struct psp_gfx_cmd_resp *cmd = acquire_psp_cmd_buf(psp); + + psp_prep_ta_invoke_indirect_cmd_buf(cmd, ta_cmd_id, context); + + ret = psp_cmd_submit_buf(psp, NULL, cmd, + psp->fence_buf_mc_addr); + + context->resp_status = cmd->resp.status; + + release_psp_cmd_buf(psp); + + return ret; } static void psp_prep_ta_invoke_cmd_buf(struct psp_gfx_cmd_resp *cmd, @@ -949,7 +1114,7 @@ static void psp_prep_ta_invoke_cmd_buf(struct psp_gfx_cmd_resp *cmd, cmd->cmd.cmd_invoke_cmd.ta_cmd_id = ta_cmd_id; } -static int psp_ta_invoke(struct psp_context *psp, +int psp_ta_invoke(struct psp_context *psp, uint32_t ta_cmd_id, struct ta_context *context) { @@ -961,12 +1126,14 @@ static int psp_ta_invoke(struct psp_context *psp, ret = psp_cmd_submit_buf(psp, NULL, cmd, psp->fence_buf_mc_addr); + context->resp_status = cmd->resp.status; + release_psp_cmd_buf(psp); return ret; } -static int psp_ta_load(struct psp_context *psp, struct ta_context *context) +int psp_ta_load(struct psp_context *psp, struct ta_context *context) { int ret; struct psp_gfx_cmd_resp *cmd; @@ -981,6 +1148,8 @@ static int psp_ta_load(struct psp_context *psp, struct ta_context *context) ret = psp_cmd_submit_buf(psp, NULL, cmd, psp->fence_buf_mc_addr); + context->resp_status = cmd->resp.status; + if (!ret) { context->session_id = cmd->resp.session_id; } @@ -990,16 +1159,6 @@ static int psp_ta_load(struct psp_context *psp, struct ta_context *context) return ret; } -static int psp_xgmi_load(struct psp_context *psp) -{ - return psp_ta_load(psp, &psp->xgmi_context.context); -} - -static int psp_xgmi_unload(struct psp_context *psp) -{ - return psp_ta_unload(psp, &psp->xgmi_context.context); -} - int psp_xgmi_invoke(struct psp_context *psp, uint32_t ta_cmd_id) { return psp_ta_invoke(psp, ta_cmd_id, &psp->xgmi_context.context); @@ -1019,16 +1178,11 @@ int psp_xgmi_terminate(struct psp_context *psp) if (!psp->xgmi_context.context.initialized) return 0; - ret = psp_xgmi_unload(psp); - if (ret) - return ret; + ret = psp_ta_unload(psp, &psp->xgmi_context.context); psp->xgmi_context.context.initialized = false; - /* free xgmi shared memory */ - psp_ta_free_shared_buf(&psp->xgmi_context.context.mem_context); - - return 0; + return ret; } int psp_xgmi_initialize(struct psp_context *psp, bool set_extended_data, bool load_ta) @@ -1047,14 +1201,14 @@ int psp_xgmi_initialize(struct psp_context *psp, bool set_extended_data, bool lo psp->xgmi_context.context.mem_context.shared_mem_size = PSP_XGMI_SHARED_MEM_SIZE; psp->xgmi_context.context.ta_load_type = GFX_CMD_ID_LOAD_TA; - if (!psp->xgmi_context.context.initialized) { - ret = psp_xgmi_init_shared_buf(psp); + if (!psp->xgmi_context.context.mem_context.shared_buf) { + ret = psp_ta_init_shared_buf(psp, &psp->xgmi_context.context.mem_context); if (ret) return ret; } /* Load XGMI TA */ - ret = psp_xgmi_load(psp); + ret = psp_ta_load(psp, &psp->xgmi_context.context); if (!ret) psp->xgmi_context.context.initialized = true; else @@ -1162,6 +1316,8 @@ static void psp_xgmi_reflect_topology_info(struct psp_context *psp, break; } + + amdgpu_put_xgmi_hive(hive); } int psp_xgmi_get_topology_info(struct psp_context *psp, @@ -1277,21 +1433,6 @@ int psp_xgmi_set_topology_info(struct psp_context *psp, } // ras begin -static int psp_ras_init_shared_buf(struct psp_context *psp) -{ - return psp_ta_init_shared_buf(psp, &psp->ras_context.context.mem_context); -} - -static int psp_ras_load(struct psp_context *psp) -{ - return psp_ta_load(psp, &psp->ras_context.context); -} - -static int psp_ras_unload(struct psp_context *psp) -{ - return psp_ta_unload(psp, &psp->ras_context.context); -} - static void psp_ras_ta_check_status(struct psp_context *psp) { struct ta_ras_shared_memory *ras_cmd = @@ -1308,6 +1449,11 @@ static void psp_ras_ta_check_status(struct psp_context *psp) break; case TA_RAS_STATUS__SUCCESS: break; + case TA_RAS_STATUS__TEE_ERROR_ACCESS_DENIED: + if (ras_cmd->cmd_id == TA_RAS_COMMAND__TRIGGER_ERROR) + dev_warn(psp->adev->dev, + "RAS WARNING: Inject error to critical region is not allowed\n"); + break; default: dev_warn(psp->adev->dev, "RAS WARNING: ras status = 0x%X\n", ras_cmd->ras_status); @@ -1381,7 +1527,7 @@ int psp_ras_enable_features(struct psp_context *psp, return 0; } -static int psp_ras_terminate(struct psp_context *psp) +int psp_ras_terminate(struct psp_context *psp) { int ret; @@ -1394,16 +1540,11 @@ static int psp_ras_terminate(struct psp_context *psp) if (!psp->ras_context.context.initialized) return 0; - ret = psp_ras_unload(psp); - if (ret) - return ret; + ret = psp_ta_unload(psp, &psp->ras_context.context); psp->ras_context.context.initialized = false; - /* free ras shared memory */ - psp_ta_free_shared_buf(&psp->ras_context.context.mem_context); - - return 0; + return ret; } static int psp_ras_initialize(struct psp_context *psp) @@ -1470,7 +1611,7 @@ static int psp_ras_initialize(struct psp_context *psp) psp->ras_context.context.ta_load_type = GFX_CMD_ID_LOAD_TA; if (!psp->ras_context.context.initialized) { - ret = psp_ras_init_shared_buf(psp); + ret = psp_ta_init_shared_buf(psp, &psp->ras_context.context.mem_context); if (ret) return ret; } @@ -1483,7 +1624,7 @@ static int psp_ras_initialize(struct psp_context *psp) if (!adev->gmc.xgmi.connected_to_cpu) ras_cmd->ras_in_message.init_flags.dgpu_mode = 1; - ret = psp_ras_load(psp); + ret = psp_ta_load(psp, &psp->ras_context.context); if (!ret && !ras_cmd->ras_status) psp->ras_context.context.initialized = true; @@ -1520,7 +1661,9 @@ int psp_ras_trigger_error(struct psp_context *psp, if (amdgpu_ras_intr_triggered()) return 0; - if (ras_cmd->ras_status) + if (ras_cmd->ras_status == TA_RAS_STATUS__TEE_ERROR_ACCESS_DENIED) + return -EACCES; + else if (ras_cmd->ras_status) return -EINVAL; return 0; @@ -1528,16 +1671,6 @@ int psp_ras_trigger_error(struct psp_context *psp, // ras end // HDCP start -static int psp_hdcp_init_shared_buf(struct psp_context *psp) -{ - return psp_ta_init_shared_buf(psp, &psp->hdcp_context.context.mem_context); -} - -static int psp_hdcp_load(struct psp_context *psp) -{ - return psp_ta_load(psp, &psp->hdcp_context.context); -} - static int psp_hdcp_initialize(struct psp_context *psp) { int ret; @@ -1558,12 +1691,12 @@ static int psp_hdcp_initialize(struct psp_context *psp) psp->hdcp_context.context.ta_load_type = GFX_CMD_ID_LOAD_TA; if (!psp->hdcp_context.context.initialized) { - ret = psp_hdcp_init_shared_buf(psp); + ret = psp_ta_init_shared_buf(psp, &psp->hdcp_context.context.mem_context); if (ret) return ret; } - ret = psp_hdcp_load(psp); + ret = psp_ta_load(psp, &psp->hdcp_context.context); if (!ret) { psp->hdcp_context.context.initialized = true; mutex_init(&psp->hdcp_context.mutex); @@ -1572,11 +1705,6 @@ static int psp_hdcp_initialize(struct psp_context *psp) return ret; } -static int psp_hdcp_unload(struct psp_context *psp) -{ - return psp_ta_unload(psp, &psp->hdcp_context.context); -} - int psp_hdcp_invoke(struct psp_context *psp, uint32_t ta_cmd_id) { /* @@ -1598,38 +1726,18 @@ static int psp_hdcp_terminate(struct psp_context *psp) if (amdgpu_sriov_vf(psp->adev)) return 0; - if (!psp->hdcp_context.context.initialized) { - if (psp->hdcp_context.context.mem_context.shared_buf) - goto out; - else - return 0; - } + if (!psp->hdcp_context.context.initialized) + return 0; - ret = psp_hdcp_unload(psp); - if (ret) - return ret; + ret = psp_ta_unload(psp, &psp->hdcp_context.context); psp->hdcp_context.context.initialized = false; -out: - /* free hdcp shared memory */ - psp_ta_free_shared_buf(&psp->hdcp_context.context.mem_context); - - return 0; + return ret; } // HDCP end // DTM start -static int psp_dtm_init_shared_buf(struct psp_context *psp) -{ - return psp_ta_init_shared_buf(psp, &psp->dtm_context.context.mem_context); -} - -static int psp_dtm_load(struct psp_context *psp) -{ - return psp_ta_load(psp, &psp->dtm_context.context); -} - static int psp_dtm_initialize(struct psp_context *psp) { int ret; @@ -1650,12 +1758,12 @@ static int psp_dtm_initialize(struct psp_context *psp) psp->dtm_context.context.ta_load_type = GFX_CMD_ID_LOAD_TA; if (!psp->dtm_context.context.initialized) { - ret = psp_dtm_init_shared_buf(psp); + ret = psp_ta_init_shared_buf(psp, &psp->dtm_context.context.mem_context); if (ret) return ret; } - ret = psp_dtm_load(psp); + ret = psp_ta_load(psp, &psp->dtm_context.context); if (!ret) { psp->dtm_context.context.initialized = true; mutex_init(&psp->dtm_context.mutex); @@ -1664,11 +1772,6 @@ static int psp_dtm_initialize(struct psp_context *psp) return ret; } -static int psp_dtm_unload(struct psp_context *psp) -{ - return psp_ta_unload(psp, &psp->dtm_context.context); -} - int psp_dtm_invoke(struct psp_context *psp, uint32_t ta_cmd_id) { /* @@ -1690,43 +1793,18 @@ static int psp_dtm_terminate(struct psp_context *psp) if (amdgpu_sriov_vf(psp->adev)) return 0; - if (!psp->dtm_context.context.initialized) { - if (psp->dtm_context.context.mem_context.shared_buf) - goto out; - else - return 0; - } + if (!psp->dtm_context.context.initialized) + return 0; - ret = psp_dtm_unload(psp); - if (ret) - return ret; + ret = psp_ta_unload(psp, &psp->dtm_context.context); psp->dtm_context.context.initialized = false; -out: - /* free dtm shared memory */ - psp_ta_free_shared_buf(&psp->dtm_context.context.mem_context); - - return 0; + return ret; } // DTM end // RAP start -static int psp_rap_init_shared_buf(struct psp_context *psp) -{ - return psp_ta_init_shared_buf(psp, &psp->rap_context.context.mem_context); -} - -static int psp_rap_load(struct psp_context *psp) -{ - return psp_ta_load(psp, &psp->rap_context.context); -} - -static int psp_rap_unload(struct psp_context *psp) -{ - return psp_ta_unload(psp, &psp->rap_context.context); -} - static int psp_rap_initialize(struct psp_context *psp) { int ret; @@ -1748,12 +1826,12 @@ static int psp_rap_initialize(struct psp_context *psp) psp->rap_context.context.ta_load_type = GFX_CMD_ID_LOAD_TA; if (!psp->rap_context.context.initialized) { - ret = psp_rap_init_shared_buf(psp); + ret = psp_ta_init_shared_buf(psp, &psp->rap_context.context.mem_context); if (ret) return ret; } - ret = psp_rap_load(psp); + ret = psp_ta_load(psp, &psp->rap_context.context); if (!ret) { psp->rap_context.context.initialized = true; mutex_init(&psp->rap_context.mutex); @@ -1763,6 +1841,8 @@ static int psp_rap_initialize(struct psp_context *psp) ret = psp_rap_invoke(psp, TA_CMD_RAP__INITIALIZE, &status); if (ret || status != TA_RAP_STATUS__SUCCESS) { psp_rap_terminate(psp); + /* free rap shared memory */ + psp_ta_free_shared_buf(&psp->rap_context.context.mem_context); dev_warn(psp->adev->dev, "RAP TA initialize fail (%d) status %d.\n", ret, status); @@ -1780,13 +1860,10 @@ static int psp_rap_terminate(struct psp_context *psp) if (!psp->rap_context.context.initialized) return 0; - ret = psp_rap_unload(psp); + ret = psp_ta_unload(psp, &psp->rap_context.context); psp->rap_context.context.initialized = false; - /* free rap shared memory */ - psp_ta_free_shared_buf(&psp->rap_context.context.mem_context); - return ret; } @@ -1826,22 +1903,6 @@ out_unlock: // RAP end /* securedisplay start */ -static int psp_securedisplay_init_shared_buf(struct psp_context *psp) -{ - return psp_ta_init_shared_buf( - psp, &psp->securedisplay_context.context.mem_context); -} - -static int psp_securedisplay_load(struct psp_context *psp) -{ - return psp_ta_load(psp, &psp->securedisplay_context.context); -} - -static int psp_securedisplay_unload(struct psp_context *psp) -{ - return psp_ta_unload(psp, &psp->securedisplay_context.context); -} - static int psp_securedisplay_initialize(struct psp_context *psp) { int ret; @@ -1864,12 +1925,13 @@ static int psp_securedisplay_initialize(struct psp_context *psp) psp->securedisplay_context.context.ta_load_type = GFX_CMD_ID_LOAD_TA; if (!psp->securedisplay_context.context.initialized) { - ret = psp_securedisplay_init_shared_buf(psp); + ret = psp_ta_init_shared_buf(psp, + &psp->securedisplay_context.context.mem_context); if (ret) return ret; } - ret = psp_securedisplay_load(psp); + ret = psp_ta_load(psp, &psp->securedisplay_context.context); if (!ret) { psp->securedisplay_context.context.initialized = true; mutex_init(&psp->securedisplay_context.mutex); @@ -1882,6 +1944,8 @@ static int psp_securedisplay_initialize(struct psp_context *psp) ret = psp_securedisplay_invoke(psp, TA_SECUREDISPLAY_COMMAND__QUERY_TA); if (ret) { psp_securedisplay_terminate(psp); + /* free securedisplay shared memory */ + psp_ta_free_shared_buf(&psp->securedisplay_context.context.mem_context); dev_err(psp->adev->dev, "SECUREDISPLAY TA initialize fail.\n"); return -EINVAL; } @@ -1908,15 +1972,10 @@ static int psp_securedisplay_terminate(struct psp_context *psp) if (!psp->securedisplay_context.context.initialized) return 0; - ret = psp_securedisplay_unload(psp); - if (ret) - return ret; + ret = psp_ta_unload(psp, &psp->securedisplay_context.context); psp->securedisplay_context.context.initialized = false; - /* free securedisplay shared memory */ - psp_ta_free_shared_buf(&psp->securedisplay_context.context.mem_context); - return ret; } @@ -2001,6 +2060,15 @@ static int psp_hw_start(struct psp_context *psp) } } + if ((is_psp_fw_valid(psp->ras_drv)) && + (psp->funcs->bootloader_load_ras_drv != NULL)) { + ret = psp_bootloader_load_ras_drv(psp); + if (ret) { + DRM_ERROR("PSP load ras_drv failed!\n"); + return ret; + } + } + if ((is_psp_fw_valid(psp->sos)) && (psp->funcs->bootloader_load_sos != NULL)) { ret = psp_bootloader_load_sos(psp); @@ -2051,6 +2119,9 @@ static int psp_get_fw_type(struct amdgpu_firmware_info *ucode, enum psp_gfx_fw_type *type) { switch (ucode->ucode_id) { + case AMDGPU_UCODE_ID_CAP: + *type = GFX_FW_TYPE_CAP; + break; case AMDGPU_UCODE_ID_SDMA0: *type = GFX_FW_TYPE_SDMA0; break; @@ -2081,6 +2152,12 @@ static int psp_get_fw_type(struct amdgpu_firmware_info *ucode, case AMDGPU_UCODE_ID_CP_MES_DATA: *type = GFX_FW_TYPE_MES_STACK; break; + case AMDGPU_UCODE_ID_CP_MES1: + *type = GFX_FW_TYPE_CP_MES_KIQ; + break; + case AMDGPU_UCODE_ID_CP_MES1_DATA: + *type = GFX_FW_TYPE_MES_KIQ_STACK; + break; case AMDGPU_UCODE_ID_CP_CE: *type = GFX_FW_TYPE_CP_CE; break; @@ -2102,6 +2179,12 @@ static int psp_get_fw_type(struct amdgpu_firmware_info *ucode, case AMDGPU_UCODE_ID_CP_MEC2_JT: *type = GFX_FW_TYPE_CP_MEC_ME2; break; + case AMDGPU_UCODE_ID_RLC_P: + *type = GFX_FW_TYPE_RLC_P; + break; + case AMDGPU_UCODE_ID_RLC_V: + *type = GFX_FW_TYPE_RLC_V; + break; case AMDGPU_UCODE_ID_RLC_G: *type = GFX_FW_TYPE_RLC_G; break; @@ -2120,9 +2203,27 @@ static int psp_get_fw_type(struct amdgpu_firmware_info *ucode, case AMDGPU_UCODE_ID_RLC_DRAM: *type = GFX_FW_TYPE_RLC_DRAM_BOOT; break; + case AMDGPU_UCODE_ID_GLOBAL_TAP_DELAYS: + *type = GFX_FW_TYPE_GLOBAL_TAP_DELAYS; + break; + case AMDGPU_UCODE_ID_SE0_TAP_DELAYS: + *type = GFX_FW_TYPE_SE0_TAP_DELAYS; + break; + case AMDGPU_UCODE_ID_SE1_TAP_DELAYS: + *type = GFX_FW_TYPE_SE1_TAP_DELAYS; + break; + case AMDGPU_UCODE_ID_SE2_TAP_DELAYS: + *type = GFX_FW_TYPE_SE2_TAP_DELAYS; + break; + case AMDGPU_UCODE_ID_SE3_TAP_DELAYS: + *type = GFX_FW_TYPE_SE3_TAP_DELAYS; + break; case AMDGPU_UCODE_ID_SMC: *type = GFX_FW_TYPE_SMU; break; + case AMDGPU_UCODE_ID_PPTABLE: + *type = GFX_FW_TYPE_PPTABLE; + break; case AMDGPU_UCODE_ID_UVD: *type = GFX_FW_TYPE_UVD; break; @@ -2153,6 +2254,51 @@ static int psp_get_fw_type(struct amdgpu_firmware_info *ucode, case AMDGPU_UCODE_ID_DMCUB: *type = GFX_FW_TYPE_DMUB; break; + case AMDGPU_UCODE_ID_SDMA_UCODE_TH0: + *type = GFX_FW_TYPE_SDMA_UCODE_TH0; + break; + case AMDGPU_UCODE_ID_SDMA_UCODE_TH1: + *type = GFX_FW_TYPE_SDMA_UCODE_TH1; + break; + case AMDGPU_UCODE_ID_IMU_I: + *type = GFX_FW_TYPE_IMU_I; + break; + case AMDGPU_UCODE_ID_IMU_D: + *type = GFX_FW_TYPE_IMU_D; + break; + case AMDGPU_UCODE_ID_CP_RS64_PFP: + *type = GFX_FW_TYPE_RS64_PFP; + break; + case AMDGPU_UCODE_ID_CP_RS64_ME: + *type = GFX_FW_TYPE_RS64_ME; + break; + case AMDGPU_UCODE_ID_CP_RS64_MEC: + *type = GFX_FW_TYPE_RS64_MEC; + break; + case AMDGPU_UCODE_ID_CP_RS64_PFP_P0_STACK: + *type = GFX_FW_TYPE_RS64_PFP_P0_STACK; + break; + case AMDGPU_UCODE_ID_CP_RS64_PFP_P1_STACK: + *type = GFX_FW_TYPE_RS64_PFP_P1_STACK; + break; + case AMDGPU_UCODE_ID_CP_RS64_ME_P0_STACK: + *type = GFX_FW_TYPE_RS64_ME_P0_STACK; + break; + case AMDGPU_UCODE_ID_CP_RS64_ME_P1_STACK: + *type = GFX_FW_TYPE_RS64_ME_P1_STACK; + break; + case AMDGPU_UCODE_ID_CP_RS64_MEC_P0_STACK: + *type = GFX_FW_TYPE_RS64_MEC_P0_STACK; + break; + case AMDGPU_UCODE_ID_CP_RS64_MEC_P1_STACK: + *type = GFX_FW_TYPE_RS64_MEC_P1_STACK; + break; + case AMDGPU_UCODE_ID_CP_RS64_MEC_P2_STACK: + *type = GFX_FW_TYPE_RS64_MEC_P2_STACK; + break; + case AMDGPU_UCODE_ID_CP_RS64_MEC_P3_STACK: + *type = GFX_FW_TYPE_RS64_MEC_P3_STACK; + break; case AMDGPU_UCODE_ID_MAXIMUM: default: return -EINVAL; @@ -2252,6 +2398,13 @@ static int psp_load_smu_fw(struct psp_context *psp) &adev->firmware.ucode[AMDGPU_UCODE_ID_SMC]; struct amdgpu_ras *ras = psp->ras_context.ras; + /* + * Skip SMU FW reloading in case of using BACO for runpm only, + * as SMU is always alive. + */ + if (adev->in_runpm && (adev->pm.rpm_mode == AMDGPU_RUNPM_BACO)) + return 0; + if (!ucode->fw || amdgpu_sriov_vf(psp->adev)) return 0; @@ -2276,7 +2429,7 @@ static int psp_load_smu_fw(struct psp_context *psp) static bool fw_load_skip_check(struct psp_context *psp, struct amdgpu_firmware_info *ucode) { - if (!ucode->fw) + if (!ucode->fw || !ucode->ucode_size) return true; if (ucode->ucode_id == AMDGPU_UCODE_ID_SMC && @@ -2286,20 +2439,7 @@ static bool fw_load_skip_check(struct psp_context *psp, return true; if (amdgpu_sriov_vf(psp->adev) && - (ucode->ucode_id == AMDGPU_UCODE_ID_SDMA0 - || ucode->ucode_id == AMDGPU_UCODE_ID_SDMA1 - || ucode->ucode_id == AMDGPU_UCODE_ID_SDMA2 - || ucode->ucode_id == AMDGPU_UCODE_ID_SDMA3 - || ucode->ucode_id == AMDGPU_UCODE_ID_SDMA4 - || ucode->ucode_id == AMDGPU_UCODE_ID_SDMA5 - || ucode->ucode_id == AMDGPU_UCODE_ID_SDMA6 - || ucode->ucode_id == AMDGPU_UCODE_ID_SDMA7 - || ucode->ucode_id == AMDGPU_UCODE_ID_RLC_G - || ucode->ucode_id == AMDGPU_UCODE_ID_RLC_RESTORE_LIST_CNTL - || ucode->ucode_id == AMDGPU_UCODE_ID_RLC_RESTORE_LIST_GPM_MEM - || ucode->ucode_id == AMDGPU_UCODE_ID_RLC_RESTORE_LIST_SRM_MEM - || ucode->ucode_id == AMDGPU_UCODE_ID_SMC)) - /*skip ucode loading in SRIOV VF */ + amdgpu_virt_fw_load_skip_check(psp->adev, ucode->ucode_id)) return true; if (psp->autoload_supported && @@ -2373,7 +2513,7 @@ static int psp_load_non_psp_fw(struct psp_context *psp) /* Start rlc autoload after psp recieved all the gfx firmware */ if (psp->autoload_supported && ucode->ucode_id == (amdgpu_sriov_vf(adev) ? - AMDGPU_UCODE_ID_CP_MEC2 : AMDGPU_UCODE_ID_RLC_G)) { + adev->virt.autoload_ucode_id : AMDGPU_UCODE_ID_RLC_G)) { ret = psp_rlc_autoload_start(psp); if (ret) { DRM_ERROR("Failed to start rlc autoload\n"); @@ -2391,69 +2531,36 @@ static int psp_load_fw(struct amdgpu_device *adev) struct psp_context *psp = &adev->psp; if (amdgpu_sriov_vf(adev) && amdgpu_in_reset(adev)) { - psp_ring_stop(psp, PSP_RING_TYPE__KM); /* should not destroy ring, only stop */ - goto skip_memalloc; - } - - if (amdgpu_sriov_vf(adev)) { - ret = amdgpu_bo_create_kernel(adev, PSP_1_MEG, PSP_1_MEG, - AMDGPU_GEM_DOMAIN_VRAM, - &psp->fw_pri_bo, - &psp->fw_pri_mc_addr, - &psp->fw_pri_buf); + /* should not destroy ring, only stop */ + psp_ring_stop(psp, PSP_RING_TYPE__KM); } else { - ret = amdgpu_bo_create_kernel(adev, PSP_1_MEG, PSP_1_MEG, - AMDGPU_GEM_DOMAIN_GTT, - &psp->fw_pri_bo, - &psp->fw_pri_mc_addr, - &psp->fw_pri_buf); - } - - if (ret) - goto failed; - - ret = amdgpu_bo_create_kernel(adev, PSP_FENCE_BUFFER_SIZE, PAGE_SIZE, - AMDGPU_GEM_DOMAIN_VRAM, - &psp->fence_buf_bo, - &psp->fence_buf_mc_addr, - &psp->fence_buf); - if (ret) - goto failed; + memset(psp->fence_buf, 0, PSP_FENCE_BUFFER_SIZE); - ret = amdgpu_bo_create_kernel(adev, PSP_CMD_BUFFER_SIZE, PAGE_SIZE, - AMDGPU_GEM_DOMAIN_VRAM, - &psp->cmd_buf_bo, &psp->cmd_buf_mc_addr, - (void **)&psp->cmd_buf_mem); - if (ret) - goto failed; - - memset(psp->fence_buf, 0, PSP_FENCE_BUFFER_SIZE); - - ret = psp_ring_init(psp, PSP_RING_TYPE__KM); - if (ret) { - DRM_ERROR("PSP ring init failed!\n"); - goto failed; + ret = psp_ring_init(psp, PSP_RING_TYPE__KM); + if (ret) { + DRM_ERROR("PSP ring init failed!\n"); + goto failed; + } } -skip_memalloc: ret = psp_hw_start(psp); if (ret) goto failed; ret = psp_load_non_psp_fw(psp); if (ret) - goto failed; + goto failed1; ret = psp_asd_initialize(psp); if (ret) { DRM_ERROR("PSP load asd failed!\n"); - return ret; + goto failed1; } ret = psp_rl_load(adev); if (ret) { DRM_ERROR("PSP load RL failed!\n"); - return ret; + goto failed1; } if (amdgpu_sriov_vf(adev) && amdgpu_in_reset(adev)) { @@ -2497,12 +2604,15 @@ skip_memalloc: return 0; +failed1: + psp_free_shared_bufs(psp); failed: /* * all cleanup jobs (xgmi terminate, ras terminate, * ring destroy, cmd/fence/fw buffers destory, * psp->cmd destory) are delayed to psp_hw_fini */ + psp_ring_destroy(psp, PSP_RING_TYPE__KM); return ret; } @@ -2546,26 +2656,24 @@ static int psp_hw_fini(void *handle) psp_rap_terminate(psp); psp_dtm_terminate(psp); psp_hdcp_terminate(psp); + + if (adev->gmc.xgmi.num_physical_nodes > 1) + psp_xgmi_terminate(psp); } psp_asd_terminate(psp); - psp_tmr_terminate(psp); + psp_ring_destroy(psp, PSP_RING_TYPE__KM); - amdgpu_bo_free_kernel(&psp->fw_pri_bo, - &psp->fw_pri_mc_addr, &psp->fw_pri_buf); - amdgpu_bo_free_kernel(&psp->fence_buf_bo, - &psp->fence_buf_mc_addr, &psp->fence_buf); - amdgpu_bo_free_kernel(&psp->cmd_buf_bo, &psp->cmd_buf_mc_addr, - (void **)&psp->cmd_buf_mem); + psp_free_shared_bufs(psp); return 0; } static int psp_suspend(void *handle) { - int ret; + int ret = 0; struct amdgpu_device *adev = (struct amdgpu_device *)handle; struct psp_context *psp = &adev->psp; @@ -2574,7 +2682,7 @@ static int psp_suspend(void *handle) ret = psp_xgmi_terminate(psp); if (ret) { DRM_ERROR("Failed to terminate xgmi ta\n"); - return ret; + goto out; } } @@ -2582,49 +2690,51 @@ static int psp_suspend(void *handle) ret = psp_ras_terminate(psp); if (ret) { DRM_ERROR("Failed to terminate ras ta\n"); - return ret; + goto out; } ret = psp_hdcp_terminate(psp); if (ret) { DRM_ERROR("Failed to terminate hdcp ta\n"); - return ret; + goto out; } ret = psp_dtm_terminate(psp); if (ret) { DRM_ERROR("Failed to terminate dtm ta\n"); - return ret; + goto out; } ret = psp_rap_terminate(psp); if (ret) { DRM_ERROR("Failed to terminate rap ta\n"); - return ret; + goto out; } ret = psp_securedisplay_terminate(psp); if (ret) { DRM_ERROR("Failed to terminate securedisplay ta\n"); - return ret; + goto out; } } ret = psp_asd_terminate(psp); if (ret) { DRM_ERROR("Failed to terminate asd\n"); - return ret; + goto out; } ret = psp_tmr_terminate(psp); if (ret) { DRM_ERROR("Failed to terminate tmr\n"); - return ret; + goto out; } ret = psp_ring_stop(psp, PSP_RING_TYPE__KM); if (ret) { DRM_ERROR("PSP ring stop failed\n"); - return ret; } - return 0; +out: + psp_free_shared_bufs(psp); + + return ret; } static int psp_resume(void *handle) @@ -2944,6 +3054,12 @@ static int parse_sos_bin_descriptor(struct psp_context *psp, psp->dbg_drv.size_bytes = le32_to_cpu(desc->size_bytes); psp->dbg_drv.start_addr = ucode_start_addr; break; + case PSP_FW_TYPE_PSP_RAS_DRV: + psp->ras_drv.fw_version = le32_to_cpu(desc->fw_version); + psp->ras_drv.feature_version = le32_to_cpu(desc->fw_version); + psp->ras_drv.size_bytes = le32_to_cpu(desc->size_bytes); + psp->ras_drv.start_addr = ucode_start_addr; + break; default: dev_warn(psp->adev->dev, "Unsupported PSP FW type: %d\n", desc->fw_type); break; @@ -2973,7 +3089,6 @@ static int psp_init_sos_base_fw(struct amdgpu_device *adev) adev->psp.sos.size_bytes = le32_to_cpu(sos_hdr->sos.size_bytes); adev->psp.sos.start_addr = ucode_array_start_addr + le32_to_cpu(sos_hdr->sos.offset_bytes); - adev->psp.xgmi_context.supports_extended_data = false; } else { /* Load alternate PSP SOS FW */ sos_hdr_v1_3 = (const struct psp_firmware_header_v1_3 *)adev->psp.sos_fw->data; @@ -2988,7 +3103,6 @@ static int psp_init_sos_base_fw(struct amdgpu_device *adev) adev->psp.sos.size_bytes = le32_to_cpu(sos_hdr_v1_3->sos_aux.size_bytes); adev->psp.sos.start_addr = ucode_array_start_addr + le32_to_cpu(sos_hdr_v1_3->sos_aux.offset_bytes); - adev->psp.xgmi_context.supports_extended_data = true; } if ((adev->psp.sys.size_bytes == 0) || (adev->psp.sos.size_bytes == 0)) { @@ -3217,6 +3331,58 @@ out: return err; } +int psp_init_cap_microcode(struct psp_context *psp, + const char *chip_name) +{ + struct amdgpu_device *adev = psp->adev; + char fw_name[PSP_FW_NAME_LEN]; + const struct psp_firmware_header_v1_0 *cap_hdr_v1_0; + struct amdgpu_firmware_info *info = NULL; + int err = 0; + + if (!chip_name) { + dev_err(adev->dev, "invalid chip name for cap microcode\n"); + return -EINVAL; + } + + if (!amdgpu_sriov_vf(adev)) { + dev_err(adev->dev, "cap microcode should only be loaded under SRIOV\n"); + return -EINVAL; + } + + snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_cap.bin", chip_name); + err = request_firmware(&adev->psp.cap_fw, fw_name, adev->dev); + if (err) { + dev_warn(adev->dev, "cap microcode does not exist, skip\n"); + err = 0; + goto out; + } + + err = amdgpu_ucode_validate(adev->psp.cap_fw); + if (err) { + dev_err(adev->dev, "fail to initialize cap microcode\n"); + goto out; + } + + info = &adev->firmware.ucode[AMDGPU_UCODE_ID_CAP]; + info->ucode_id = AMDGPU_UCODE_ID_CAP; + info->fw = adev->psp.cap_fw; + cap_hdr_v1_0 = (const struct psp_firmware_header_v1_0 *) + adev->psp.cap_fw->data; + adev->firmware.fw_size += ALIGN( + le32_to_cpu(cap_hdr_v1_0->header.ucode_size_bytes), PAGE_SIZE); + adev->psp.cap_fw_version = le32_to_cpu(cap_hdr_v1_0->header.ucode_version); + adev->psp.cap_feature_version = le32_to_cpu(cap_hdr_v1_0->sos.fw_version); + adev->psp.cap_ucode_size = le32_to_cpu(cap_hdr_v1_0->header.ucode_size_bytes); + + return 0; + +out: + release_firmware(adev->psp.cap_fw); + adev->psp.cap_fw = NULL; + return err; +} + static int psp_set_clockgating_state(void *handle, enum amd_clockgating_state state) { @@ -3333,6 +3499,140 @@ int is_psp_fw_valid(struct psp_bin_desc bin) return bin.size_bytes; } +static ssize_t amdgpu_psp_vbflash_write(struct file *filp, struct kobject *kobj, + struct bin_attribute *bin_attr, + char *buffer, loff_t pos, size_t count) +{ + struct device *dev = kobj_to_dev(kobj); + struct drm_device *ddev = dev_get_drvdata(dev); + struct amdgpu_device *adev = drm_to_adev(ddev); + + adev->psp.vbflash_done = false; + + /* Safeguard against memory drain */ + if (adev->psp.vbflash_image_size > AMD_VBIOS_FILE_MAX_SIZE_B) { + dev_err(adev->dev, "File size cannot exceed %u", AMD_VBIOS_FILE_MAX_SIZE_B); + kvfree(adev->psp.vbflash_tmp_buf); + adev->psp.vbflash_tmp_buf = NULL; + adev->psp.vbflash_image_size = 0; + return -ENOMEM; + } + + /* TODO Just allocate max for now and optimize to realloc later if needed */ + if (!adev->psp.vbflash_tmp_buf) { + adev->psp.vbflash_tmp_buf = kvmalloc(AMD_VBIOS_FILE_MAX_SIZE_B, GFP_KERNEL); + if (!adev->psp.vbflash_tmp_buf) + return -ENOMEM; + } + + mutex_lock(&adev->psp.mutex); + memcpy(adev->psp.vbflash_tmp_buf + pos, buffer, count); + adev->psp.vbflash_image_size += count; + mutex_unlock(&adev->psp.mutex); + + dev_info(adev->dev, "VBIOS flash write PSP done"); + + return count; +} + +static ssize_t amdgpu_psp_vbflash_read(struct file *filp, struct kobject *kobj, + struct bin_attribute *bin_attr, char *buffer, + loff_t pos, size_t count) +{ + struct device *dev = kobj_to_dev(kobj); + struct drm_device *ddev = dev_get_drvdata(dev); + struct amdgpu_device *adev = drm_to_adev(ddev); + struct amdgpu_bo *fw_buf_bo = NULL; + uint64_t fw_pri_mc_addr; + void *fw_pri_cpu_addr; + int ret; + + dev_info(adev->dev, "VBIOS flash to PSP started"); + + ret = amdgpu_bo_create_kernel(adev, adev->psp.vbflash_image_size, + AMDGPU_GPU_PAGE_SIZE, + AMDGPU_GEM_DOMAIN_VRAM, + &fw_buf_bo, + &fw_pri_mc_addr, + &fw_pri_cpu_addr); + if (ret) + goto rel_buf; + + memcpy_toio(fw_pri_cpu_addr, adev->psp.vbflash_tmp_buf, adev->psp.vbflash_image_size); + + mutex_lock(&adev->psp.mutex); + ret = psp_update_spirom(&adev->psp, fw_pri_mc_addr); + mutex_unlock(&adev->psp.mutex); + + amdgpu_bo_free_kernel(&fw_buf_bo, &fw_pri_mc_addr, &fw_pri_cpu_addr); + +rel_buf: + kvfree(adev->psp.vbflash_tmp_buf); + adev->psp.vbflash_tmp_buf = NULL; + adev->psp.vbflash_image_size = 0; + + if (ret) { + dev_err(adev->dev, "Failed to load VBIOS FW, err = %d", ret); + return ret; + } + + dev_info(adev->dev, "VBIOS flash to PSP done"); + return 0; +} + +static ssize_t amdgpu_psp_vbflash_status(struct device *dev, + struct device_attribute *attr, + char *buf) +{ + struct drm_device *ddev = dev_get_drvdata(dev); + struct amdgpu_device *adev = drm_to_adev(ddev); + uint32_t vbflash_status; + + vbflash_status = psp_vbflash_status(&adev->psp); + if (!adev->psp.vbflash_done) + vbflash_status = 0; + else if (adev->psp.vbflash_done && !(vbflash_status & 0x80000000)) + vbflash_status = 1; + + return sysfs_emit(buf, "0x%x\n", vbflash_status); +} + +static const struct bin_attribute psp_vbflash_bin_attr = { + .attr = {.name = "psp_vbflash", .mode = 0664}, + .size = 0, + .write = amdgpu_psp_vbflash_write, + .read = amdgpu_psp_vbflash_read, +}; + +static DEVICE_ATTR(psp_vbflash_status, 0444, amdgpu_psp_vbflash_status, NULL); + +int amdgpu_psp_sysfs_init(struct amdgpu_device *adev) +{ + int ret = 0; + struct psp_context *psp = &adev->psp; + + if (amdgpu_sriov_vf(adev)) + return -EINVAL; + + switch (adev->ip_versions[MP0_HWIP][0]) { + case IP_VERSION(13, 0, 0): + case IP_VERSION(13, 0, 7): + if (!psp->adev) { + psp->adev = adev; + psp_v13_0_set_psp_funcs(psp); + } + ret = sysfs_create_bin_file(&adev->dev->kobj, &psp_vbflash_bin_attr); + if (ret) + dev_err(adev->dev, "Failed to create device file psp_vbflash"); + ret = device_create_file(adev->dev, &dev_attr_psp_vbflash_status); + if (ret) + dev_err(adev->dev, "Failed to create device file psp_vbflash_status"); + return ret; + default: + return 0; + } +} + const struct amd_ip_funcs psp_ip_funcs = { .name = "psp", .early_init = psp_early_init, @@ -3361,6 +3661,12 @@ static int psp_sysfs_init(struct amdgpu_device *adev) return ret; } +void amdgpu_psp_sysfs_fini(struct amdgpu_device *adev) +{ + sysfs_remove_bin_file(&adev->dev->kobj, &psp_vbflash_bin_attr); + device_remove_file(adev->dev, &dev_attr_psp_vbflash_status); +} + static void psp_sysfs_fini(struct amdgpu_device *adev) { device_remove_file(adev->dev, &dev_attr_usbc_pd_fw); @@ -3417,3 +3723,11 @@ const struct amdgpu_ip_block_version psp_v13_0_ip_block = { .rev = 0, .funcs = &psp_ip_funcs, }; + +const struct amdgpu_ip_block_version psp_v13_0_4_ip_block = { + .type = AMD_IP_BLOCK_TYPE_PSP, + .major = 13, + .minor = 0, + .rev = 4, + .funcs = &psp_ip_funcs, +}; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.h index f29afabbff1f..58ce3ebb446c 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.h @@ -36,6 +36,7 @@ #define PSP_CMD_BUFFER_SIZE 0x1000 #define PSP_1_MEG 0x100000 #define PSP_TMR_SIZE(adev) ((adev)->asic_type == CHIP_ALDEBARAN ? 0x800000 : 0x400000) +#define PSP_TMR_ALIGNMENT 0x100000 #define PSP_FW_NAME_LEN 0x24 enum psp_shared_mem_size { @@ -48,6 +49,17 @@ enum psp_shared_mem_size { PSP_SECUREDISPLAY_SHARED_MEM_SIZE = 0x4000, }; +enum ta_type_id { + TA_TYPE_XGMI = 1, + TA_TYPE_RAS, + TA_TYPE_HDCP, + TA_TYPE_DTM, + TA_TYPE_RAP, + TA_TYPE_SECUREDISPLAY, + + TA_TYPE_MAX_INDEX, +}; + struct psp_context; struct psp_xgmi_node_info; struct psp_xgmi_topology_info; @@ -58,8 +70,9 @@ enum psp_bootloader_cmd { PSP_BL__LOAD_SOSDRV = 0x20000, PSP_BL__LOAD_KEY_DATABASE = 0x80000, PSP_BL__LOAD_SOCDRV = 0xB0000, - PSP_BL__LOAD_INTFDRV = 0xC0000, - PSP_BL__LOAD_DBGDRV = 0xD0000, + PSP_BL__LOAD_DBGDRV = 0xC0000, + PSP_BL__LOAD_INTFDRV = 0xD0000, + PSP_BL__LOAD_RASDRV = 0xE0000, PSP_BL__DRAM_LONG_TRAIN = 0x100000, PSP_BL__DRAM_SHORT_TRAIN = 0x200000, PSP_BL__LOAD_TOS_SPL_TABLE = 0x10000000, @@ -103,6 +116,7 @@ struct psp_funcs int (*bootloader_load_soc_drv)(struct psp_context *psp); int (*bootloader_load_intf_drv)(struct psp_context *psp); int (*bootloader_load_dbg_drv)(struct psp_context *psp); + int (*bootloader_load_ras_drv)(struct psp_context *psp); int (*bootloader_load_sos)(struct psp_context *psp); int (*ring_init)(struct psp_context *psp, enum psp_ring_type ring_type); int (*ring_create)(struct psp_context *psp, @@ -118,6 +132,8 @@ struct psp_funcs void (*ring_set_wptr)(struct psp_context *psp, uint32_t value); int (*load_usbc_pd_fw)(struct psp_context *psp, uint64_t fw_pri_mc_addr); int (*read_usbc_pd_fw)(struct psp_context *psp, uint32_t *fw_ver); + int (*update_spirom)(struct psp_context *psp, uint64_t fw_pri_mc_addr); + int (*vbflash_stat)(struct psp_context *psp); }; #define AMDGPU_XGMI_MAX_CONNECTED_NODES 64 @@ -151,9 +167,11 @@ struct ta_mem_context { struct ta_context { bool initialized; uint32_t session_id; + uint32_t resp_status; struct ta_mem_context mem_context; struct psp_bin_desc bin_desc; enum psp_gfx_cmd_id ta_load_type; + enum ta_type_id ta_type; }; struct ta_cp_context { @@ -231,6 +249,7 @@ enum psp_runtime_entry_type { PSP_RUNTIME_ENTRY_TYPE_MGPU_WAFL = 0x3, /* WAFL runtime data */ PSP_RUNTIME_ENTRY_TYPE_MGPU_XGMI = 0x4, /* XGMI runtime data */ PSP_RUNTIME_ENTRY_TYPE_BOOT_CONFIG = 0x5, /* Boot Config runtime data */ + PSP_RUNTIME_ENTRY_TYPE_PPTABLE_ERR_STATUS = 0x6, /* SCPM validation data */ }; /* PSP runtime DB header */ @@ -265,12 +284,24 @@ enum psp_runtime_boot_cfg_feature { BOOT_CFG_FEATURE_TWO_STAGE_DRAM_TRAINING = 0x2, }; +/* PSP run time DB SCPM authentication defines */ +enum psp_runtime_scpm_authentication { + SCPM_DISABLE = 0x0, + SCPM_ENABLE = 0x1, + SCPM_ENABLE_WITH_SCPM_ERR = 0x2, +}; + /* PSP runtime DB boot config entry */ struct psp_runtime_boot_cfg_entry { uint32_t boot_cfg_bitmask; uint32_t reserved; }; +/* PSP runtime DB SCPM entry */ +struct psp_runtime_scpm_entry { + enum psp_runtime_scpm_authentication scpm_status; +}; + struct psp_context { struct amdgpu_device *adev; @@ -295,6 +326,7 @@ struct psp_context struct psp_bin_desc soc_drv; struct psp_bin_desc intf_drv; struct psp_bin_desc dbg_drv; + struct psp_bin_desc ras_drv; /* tmr buffer */ struct amdgpu_bo *tmr_bo; @@ -306,6 +338,9 @@ struct psp_context /* toc firmware */ const struct firmware *toc_fw; + /* cap firmware */ + const struct firmware *cap_fw; + /* fence buffer */ struct amdgpu_bo *fence_buf_bo; uint64_t fence_buf_mc_addr; @@ -327,6 +362,10 @@ struct psp_context const struct firmware *ta_fw; uint32_t ta_fw_version; + uint32_t cap_fw_version; + uint32_t cap_feature_version; + uint32_t cap_ucode_size; + struct ta_context asd_context; struct psp_xgmi_context xgmi_context; struct psp_ras_context ras_context; @@ -338,6 +377,10 @@ struct psp_context struct psp_memory_training_context mem_train_ctx; uint32_t boot_cfg_bitmask; + + char *vbflash_tmp_buf; + size_t vbflash_image_size; + bool vbflash_done; }; struct amdgpu_psp_funcs { @@ -364,6 +407,9 @@ struct amdgpu_psp_funcs { ((psp)->funcs->bootloader_load_intf_drv ? (psp)->funcs->bootloader_load_intf_drv((psp)) : 0) #define psp_bootloader_load_dbg_drv(psp) \ ((psp)->funcs->bootloader_load_dbg_drv ? (psp)->funcs->bootloader_load_dbg_drv((psp)) : 0) +#define psp_bootloader_load_ras_drv(psp) \ + ((psp)->funcs->bootloader_load_ras_drv ? \ + (psp)->funcs->bootloader_load_ras_drv((psp)) : 0) #define psp_bootloader_load_sos(psp) \ ((psp)->funcs->bootloader_load_sos ? (psp)->funcs->bootloader_load_sos((psp)) : 0) #define psp_smu_reload_quirk(psp) \ @@ -384,6 +430,14 @@ struct amdgpu_psp_funcs { ((psp)->funcs->read_usbc_pd_fw ? \ (psp)->funcs->read_usbc_pd_fw((psp), fw_ver) : -EINVAL) +#define psp_update_spirom(psp, fw_pri_mc_addr) \ + ((psp)->funcs->update_spirom ? \ + (psp)->funcs->update_spirom((psp), fw_pri_mc_addr) : -EINVAL) + +#define psp_vbflash_status(psp) \ + ((psp)->funcs->vbflash_stat ? \ + (psp)->funcs->vbflash_stat((psp)) : -EINVAL) + extern const struct amd_ip_funcs psp_ip_funcs; extern const struct amdgpu_ip_block_version psp_v3_1_ip_block; @@ -392,6 +446,7 @@ extern const struct amdgpu_ip_block_version psp_v11_0_ip_block; extern const struct amdgpu_ip_block_version psp_v11_0_8_ip_block; extern const struct amdgpu_ip_block_version psp_v12_0_ip_block; extern const struct amdgpu_ip_block_version psp_v13_0_ip_block; +extern const struct amdgpu_ip_block_version psp_v13_0_4_ip_block; extern int psp_wait_for(struct psp_context *psp, uint32_t reg_index, uint32_t field_val, uint32_t mask, bool check_changed); @@ -400,6 +455,18 @@ int psp_gpu_reset(struct amdgpu_device *adev); int psp_update_vcn_sram(struct amdgpu_device *adev, int inst_idx, uint64_t cmd_gpu_addr, int cmd_size); +int psp_ta_init_shared_buf(struct psp_context *psp, + struct ta_mem_context *mem_ctx); +void psp_ta_free_shared_buf(struct ta_mem_context *mem_ctx); +int psp_ta_unload(struct psp_context *psp, struct ta_context *context); +int psp_ta_load(struct psp_context *psp, struct ta_context *context); +int psp_ta_invoke(struct psp_context *psp, + uint32_t ta_cmd_id, + struct ta_context *context); +int psp_ta_invoke_indirect(struct psp_context *psp, + uint32_t ta_cmd_id, + struct ta_context *context); + int psp_xgmi_initialize(struct psp_context *psp, bool set_extended_data, bool load_ta); int psp_xgmi_terminate(struct psp_context *psp); int psp_xgmi_invoke(struct psp_context *psp, uint32_t ta_cmd_id); @@ -418,6 +485,7 @@ int psp_ras_enable_features(struct psp_context *psp, union ta_ras_cmd_input *info, bool enable); int psp_ras_trigger_error(struct psp_context *psp, struct ta_ras_trigger_error_input *info); +int psp_ras_terminate(struct psp_context *psp); int psp_hdcp_invoke(struct psp_context *psp, uint32_t ta_cmd_id); int psp_dtm_invoke(struct psp_context *psp, uint32_t ta_cmd_id); @@ -440,6 +508,8 @@ int psp_init_sos_microcode(struct psp_context *psp, const char *chip_name); int psp_init_ta_microcode(struct psp_context *psp, const char *chip_name); +int psp_init_cap_microcode(struct psp_context *psp, + const char *chip_name); int psp_get_fw_attestation_records_addr(struct psp_context *psp, uint64_t *output_ptr); @@ -448,4 +518,7 @@ int psp_load_fw_list(struct psp_context *psp, void psp_copy_fw(struct psp_context *psp, uint8_t *start_addr, uint32_t bin_size); int is_psp_fw_valid(struct psp_bin_desc bin); + +int amdgpu_psp_sysfs_init(struct amdgpu_device *adev); +void amdgpu_psp_sysfs_fini(struct amdgpu_device *adev); #endif diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp_ta.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp_ta.c new file mode 100644 index 000000000000..0988e00612e5 --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp_ta.c @@ -0,0 +1,304 @@ +/* + * Copyright 2022 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#include "amdgpu.h" +#include "amdgpu_psp_ta.h" + +#if defined(CONFIG_DEBUG_FS) + +static ssize_t ta_if_load_debugfs_write(struct file *fp, const char *buf, + size_t len, loff_t *off); +static ssize_t ta_if_unload_debugfs_write(struct file *fp, const char *buf, + size_t len, loff_t *off); +static ssize_t ta_if_invoke_debugfs_write(struct file *fp, const char *buf, + size_t len, loff_t *off); + +static uint32_t get_bin_version(const uint8_t *bin) +{ + const struct common_firmware_header *hdr = + (const struct common_firmware_header *)bin; + + return hdr->ucode_version; +} + +static void prep_ta_mem_context(struct psp_context *psp, + struct ta_context *context, + uint8_t *shared_buf, + uint32_t shared_buf_len) +{ + context->mem_context.shared_mem_size = PAGE_ALIGN(shared_buf_len); + psp_ta_init_shared_buf(psp, &context->mem_context); + + memcpy((void *)context->mem_context.shared_buf, shared_buf, shared_buf_len); +} + +static bool is_ta_type_valid(enum ta_type_id ta_type) +{ + bool ret = false; + + switch (ta_type) { + case TA_TYPE_RAS: + ret = true; + break; + default: + break; + } + + return ret; +} + +static const struct file_operations ta_load_debugfs_fops = { + .write = ta_if_load_debugfs_write, + .llseek = default_llseek, + .owner = THIS_MODULE +}; + +static const struct file_operations ta_unload_debugfs_fops = { + .write = ta_if_unload_debugfs_write, + .llseek = default_llseek, + .owner = THIS_MODULE +}; + +static const struct file_operations ta_invoke_debugfs_fops = { + .write = ta_if_invoke_debugfs_write, + .llseek = default_llseek, + .owner = THIS_MODULE +}; + + +/** + * DOC: AMDGPU TA debugfs interfaces + * + * Three debugfs interfaces can be opened by a program to + * load/invoke/unload TA, + * + * - /sys/kernel/debug/dri/<N>/ta_if/ta_load + * - /sys/kernel/debug/dri/<N>/ta_if/ta_invoke + * - /sys/kernel/debug/dri/<N>/ta_if/ta_unload + * + * How to use the interfaces in a program? + * + * A program needs to provide transmit buffer to the interfaces + * and will receive buffer from the interfaces below, + * + * - For TA load debugfs interface: + * Transmit buffer: + * - TA type (4bytes) + * - TA bin length (4bytes) + * - TA bin + * Receive buffer: + * - TA ID (4bytes) + * + * - For TA invoke debugfs interface: + * Transmit buffer: + * - TA ID (4bytes) + * - TA CMD ID (4bytes) + * - TA shard buf length (4bytes) + * - TA shared buf + * Receive buffer: + * - TA shared buf + * + * - For TA unload debugfs interface: + * Transmit buffer: + * - TA ID (4bytes) + */ + +static ssize_t ta_if_load_debugfs_write(struct file *fp, const char *buf, size_t len, loff_t *off) +{ + uint32_t ta_type = 0; + uint32_t ta_bin_len = 0; + uint8_t *ta_bin = NULL; + uint32_t copy_pos = 0; + int ret = 0; + + struct amdgpu_device *adev = (struct amdgpu_device *)file_inode(fp)->i_private; + struct psp_context *psp = &adev->psp; + struct ta_context context = {0}; + + if (!buf) + return -EINVAL; + + ret = copy_from_user((void *)&ta_type, &buf[copy_pos], sizeof(uint32_t)); + if (ret || (!is_ta_type_valid(ta_type))) + return -EINVAL; + + copy_pos += sizeof(uint32_t); + + ret = copy_from_user((void *)&ta_bin_len, &buf[copy_pos], sizeof(uint32_t)); + if (ret) + return -EINVAL; + + copy_pos += sizeof(uint32_t); + + ta_bin = kzalloc(ta_bin_len, GFP_KERNEL); + if (!ta_bin) + ret = -ENOMEM; + if (copy_from_user((void *)ta_bin, &buf[copy_pos], ta_bin_len)) { + ret = -EFAULT; + goto err_free_bin; + } + + ret = psp_ras_terminate(psp); + if (ret) { + dev_err(adev->dev, "Failed to unload embedded RAS TA\n"); + goto err_free_bin; + } + + context.ta_type = ta_type; + context.ta_load_type = GFX_CMD_ID_LOAD_TA; + context.bin_desc.fw_version = get_bin_version(ta_bin); + context.bin_desc.size_bytes = ta_bin_len; + context.bin_desc.start_addr = ta_bin; + + ret = psp_ta_load(psp, &context); + + if (ret || context.resp_status) { + dev_err(adev->dev, "TA load via debugfs failed (%d) status %d\n", + ret, context.resp_status); + if (!ret) + ret = -EINVAL; + goto err_free_bin; + } + + context.initialized = true; + if (copy_to_user((char *)buf, (void *)&context.session_id, sizeof(uint32_t))) + ret = -EFAULT; + +err_free_bin: + kfree(ta_bin); + + return ret; +} + +static ssize_t ta_if_unload_debugfs_write(struct file *fp, const char *buf, size_t len, loff_t *off) +{ + uint32_t ta_id = 0; + int ret = 0; + + struct amdgpu_device *adev = (struct amdgpu_device *)file_inode(fp)->i_private; + struct psp_context *psp = &adev->psp; + struct ta_context context = {0}; + + if (!buf) + return -EINVAL; + + ret = copy_from_user((void *)&ta_id, buf, sizeof(uint32_t)); + if (ret) + return -EINVAL; + + context.session_id = ta_id; + + ret = psp_ta_unload(psp, &context); + if (!ret) + context.initialized = false; + + return ret; +} + +static ssize_t ta_if_invoke_debugfs_write(struct file *fp, const char *buf, size_t len, loff_t *off) +{ + uint32_t ta_id = 0; + uint32_t cmd_id = 0; + uint32_t shared_buf_len = 0; + uint8_t *shared_buf = NULL; + uint32_t copy_pos = 0; + int ret = 0; + + struct amdgpu_device *adev = (struct amdgpu_device *)file_inode(fp)->i_private; + struct psp_context *psp = &adev->psp; + struct ta_context context = {0}; + + if (!buf) + return -EINVAL; + + ret = copy_from_user((void *)&ta_id, &buf[copy_pos], sizeof(uint32_t)); + if (ret) + return -EINVAL; + copy_pos += sizeof(uint32_t); + + ret = copy_from_user((void *)&cmd_id, &buf[copy_pos], sizeof(uint32_t)); + if (ret) + return -EINVAL; + copy_pos += sizeof(uint32_t); + + ret = copy_from_user((void *)&shared_buf_len, &buf[copy_pos], sizeof(uint32_t)); + if (ret) + return -EINVAL; + copy_pos += sizeof(uint32_t); + + shared_buf = kzalloc(shared_buf_len, GFP_KERNEL); + if (!shared_buf) + return -ENOMEM; + if (copy_from_user((void *)shared_buf, &buf[copy_pos], shared_buf_len)) { + ret = -EFAULT; + goto err_free_shared_buf; + } + + context.session_id = ta_id; + + prep_ta_mem_context(psp, &context, shared_buf, shared_buf_len); + + ret = psp_ta_invoke_indirect(psp, cmd_id, &context); + + if (ret || context.resp_status) { + dev_err(adev->dev, "TA invoke via debugfs failed (%d) status %d\n", + ret, context.resp_status); + if (!ret) + ret = -EINVAL; + goto err_free_ta_shared_buf; + } + + if (copy_to_user((char *)buf, context.mem_context.shared_buf, shared_buf_len)) + ret = -EFAULT; + +err_free_ta_shared_buf: + psp_ta_free_shared_buf(&context.mem_context); + +err_free_shared_buf: + kfree(shared_buf); + + return ret; +} + +void amdgpu_ta_if_debugfs_init(struct amdgpu_device *adev) +{ + struct drm_minor *minor = adev_to_drm(adev)->primary; + + struct dentry *dir = debugfs_create_dir("ta_if", minor->debugfs_root); + + debugfs_create_file("ta_load", 0200, dir, adev, + &ta_load_debugfs_fops); + + debugfs_create_file("ta_unload", 0200, dir, + adev, &ta_unload_debugfs_fops); + + debugfs_create_file("ta_invoke", 0200, dir, + adev, &ta_invoke_debugfs_fops); +} + +#else +void amdgpu_ta_if_debugfs_init(struct amdgpu_device *adev) +{ + +} +#endif diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp_ta.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp_ta.h new file mode 100644 index 000000000000..cfc1542f63ef --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp_ta.h @@ -0,0 +1,29 @@ +/* + * Copyright 2022 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#ifndef __AMDGPU_PSP_TA_H__ +#define __AMDGPU_PSP_TA_H__ + +void amdgpu_ta_if_debugfs_init(struct amdgpu_device *adev); + +#endif diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c index 91e6e87562ac..a4b47e1bd111 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c @@ -35,6 +35,8 @@ #include "amdgpu_xgmi.h" #include "ivsrcid/nbio/irqsrcs_nbif_7_4.h" #include "atom.h" +#include "amdgpu_reset.h" + #ifdef CONFIG_X86_MCE_AMD #include <asm/mce.h> @@ -66,6 +68,8 @@ const char *ras_block_string[] = { "mp1", "fuse", "mca", + "vcn", + "jpeg", }; const char *ras_mca_block_string[] = { @@ -75,6 +79,13 @@ const char *ras_mca_block_string[] = { "mca_iohc", }; +struct amdgpu_ras_block_list { + /* ras block link */ + struct list_head node; + + struct amdgpu_ras_block_object *ras_obj; +}; + const char *get_ras_block_str(struct ras_common_if *ras_block) { if (!ras_block) @@ -89,6 +100,9 @@ const char *get_ras_block_str(struct ras_common_if *ras_block) return ras_block_string[ras_block->block]; } +#define ras_block_str(_BLOCK_) \ + (((_BLOCK_) < ARRAY_SIZE(ras_block_string)) ? ras_block_string[_BLOCK_] : "Out Of Range") + #define ras_err_str(i) (ras_error_string[ffs(i)]) #define RAS_DEFAULT_FLAGS (AMDGPU_RAS_FLAG_INIT_BY_VBIOS) @@ -155,14 +169,9 @@ static int amdgpu_reserve_page_direct(struct amdgpu_device *adev, uint64_t addre } memset(&err_rec, 0x0, sizeof(struct eeprom_table_record)); - - err_rec.address = address; - err_rec.retired_page = address >> AMDGPU_GPU_PAGE_SHIFT; - err_rec.ts = (uint64_t)ktime_get_real_seconds(); - err_rec.err_type = AMDGPU_RAS_EEPROM_ERR_NON_RECOVERABLE; - err_data.err_addr = &err_rec; - err_data.err_addr_cnt = 1; + amdgpu_umc_fill_error_record(&err_data, address, + (address >> AMDGPU_GPU_PAGE_SHIFT), 0, 0); if (amdgpu_bad_page_threshold != 0) { amdgpu_ras_add_bad_pages(adev, err_data.err_addr, @@ -190,6 +199,13 @@ static ssize_t amdgpu_ras_debugfs_read(struct file *f, char __user *buf, if (amdgpu_ras_query_error_status(obj->adev, &info)) return -EINVAL; + /* Hardware counter will be reset automatically after the query on Vega20 and Arcturus */ + if (obj->adev->ip_versions[MP0_HWIP][0] != IP_VERSION(11, 0, 2) && + obj->adev->ip_versions[MP0_HWIP][0] != IP_VERSION(11, 0, 4)) { + if (amdgpu_ras_reset_error_status(obj->adev, info.head.block)) + dev_warn(obj->adev->dev, "Failed to reset error counter and error status"); + } + s = snprintf(val, sizeof(val), "%s: %lu\n%s: %lu\n", "ue", info.ue_count, "ce", info.ce_count); @@ -452,7 +468,7 @@ static ssize_t amdgpu_ras_debugfs_ctrl_write(struct file *f, } if (ret) - return -EINVAL; + return ret; return size; } @@ -543,9 +559,10 @@ static ssize_t amdgpu_ras_sysfs_read(struct device *dev, if (amdgpu_ras_query_error_status(obj->adev, &info)) return -EINVAL; - if (obj->adev->asic_type == CHIP_ALDEBARAN) { + if (obj->adev->ip_versions[MP0_HWIP][0] != IP_VERSION(11, 0, 2) && + obj->adev->ip_versions[MP0_HWIP][0] != IP_VERSION(11, 0, 4)) { if (amdgpu_ras_reset_error_status(obj->adev, info.head.block)) - DRM_WARN("Failed to reset error counter and error status"); + dev_warn(obj->adev->dev, "Failed to reset error counter and error status"); } return sysfs_emit(buf, "%s: %lu\n%s: %lu\n", "ue", info.ue_count, @@ -700,26 +717,31 @@ int amdgpu_ras_feature_enable(struct amdgpu_device *adev, if (!con) return -EINVAL; - info = kzalloc(sizeof(union ta_ras_cmd_input), GFP_KERNEL); - if (!info) - return -ENOMEM; + if (head->block == AMDGPU_RAS_BLOCK__GFX) { + info = kzalloc(sizeof(union ta_ras_cmd_input), GFP_KERNEL); + if (!info) + return -ENOMEM; - if (!enable) { - info->disable_features = (struct ta_ras_disable_features_input) { - .block_id = amdgpu_ras_block_to_ta(head->block), - .error_type = amdgpu_ras_error_to_ta(head->type), - }; - } else { - info->enable_features = (struct ta_ras_enable_features_input) { - .block_id = amdgpu_ras_block_to_ta(head->block), - .error_type = amdgpu_ras_error_to_ta(head->type), - }; + if (!enable) { + info->disable_features = (struct ta_ras_disable_features_input) { + .block_id = amdgpu_ras_block_to_ta(head->block), + .error_type = amdgpu_ras_error_to_ta(head->type), + }; + } else { + info->enable_features = (struct ta_ras_enable_features_input) { + .block_id = amdgpu_ras_block_to_ta(head->block), + .error_type = amdgpu_ras_error_to_ta(head->type), + }; + } } /* Do not enable if it is not allowed. */ WARN_ON(enable && !amdgpu_ras_is_feature_allowed(adev, head)); - if (!amdgpu_ras_intr_triggered()) { + /* Only enable ras feature operation handle on host side */ + if (head->block == AMDGPU_RAS_BLOCK__GFX && + !amdgpu_sriov_vf(adev) && + !amdgpu_ras_intr_triggered()) { ret = psp_ras_enable_features(&adev->psp, info, enable); if (ret) { dev_err(adev->dev, "ras %s %s failed poison:%d ret:%d\n", @@ -734,7 +756,8 @@ int amdgpu_ras_feature_enable(struct amdgpu_device *adev, __amdgpu_ras_feature_enable(adev, head, enable); ret = 0; out: - kfree(info); + if (head->block == AMDGPU_RAS_BLOCK__GFX) + kfree(info); return ret; } @@ -866,30 +889,47 @@ static int amdgpu_ras_enable_all_features(struct amdgpu_device *adev, } /* feature ctl end */ +static int amdgpu_ras_block_match_default(struct amdgpu_ras_block_object *block_obj, + enum amdgpu_ras_block block) +{ + if (!block_obj) + return -EINVAL; + + if (block_obj->ras_comm.block == block) + return 0; + + return -EINVAL; +} -static void amdgpu_ras_mca_query_error_status(struct amdgpu_device *adev, - struct ras_common_if *ras_block, - struct ras_err_data *err_data) +static struct amdgpu_ras_block_object *amdgpu_ras_get_ras_block(struct amdgpu_device *adev, + enum amdgpu_ras_block block, uint32_t sub_block_index) { - switch (ras_block->sub_block_index) { - case AMDGPU_RAS_MCA_BLOCK__MP0: - if (adev->mca.mp0.ras_funcs && - adev->mca.mp0.ras_funcs->query_ras_error_count) - adev->mca.mp0.ras_funcs->query_ras_error_count(adev, &err_data); - break; - case AMDGPU_RAS_MCA_BLOCK__MP1: - if (adev->mca.mp1.ras_funcs && - adev->mca.mp1.ras_funcs->query_ras_error_count) - adev->mca.mp1.ras_funcs->query_ras_error_count(adev, &err_data); - break; - case AMDGPU_RAS_MCA_BLOCK__MPIO: - if (adev->mca.mpio.ras_funcs && - adev->mca.mpio.ras_funcs->query_ras_error_count) - adev->mca.mpio.ras_funcs->query_ras_error_count(adev, &err_data); - break; - default: - break; + struct amdgpu_ras_block_list *node, *tmp; + struct amdgpu_ras_block_object *obj; + + if (block >= AMDGPU_RAS_BLOCK__LAST) + return NULL; + + if (!amdgpu_ras_is_supported(adev, block)) + return NULL; + + list_for_each_entry_safe(node, tmp, &adev->ras_list, node) { + if (!node->ras_obj) { + dev_warn(adev->dev, "Warning: abnormal ras list node.\n"); + continue; + } + + obj = node->ras_obj; + if (obj->ras_block_match) { + if (obj->ras_block_match(obj, block, sub_block_index) == 0) + return obj; + } else { + if (amdgpu_ras_block_match_default(obj, block) == 0) + return obj; + } } + + return NULL; } static void amdgpu_ras_get_ecc_info(struct amdgpu_device *adev, struct ras_err_data *err_data) @@ -901,26 +941,26 @@ static void amdgpu_ras_get_ecc_info(struct amdgpu_device *adev, struct ras_err_d * choosing right query method according to * whether smu support query error information */ - ret = smu_get_ecc_info(&adev->smu, (void *)&(ras->umc_ecc)); + ret = amdgpu_dpm_get_ecc_info(adev, (void *)&(ras->umc_ecc)); if (ret == -EOPNOTSUPP) { - if (adev->umc.ras_funcs && - adev->umc.ras_funcs->query_ras_error_count) - adev->umc.ras_funcs->query_ras_error_count(adev, err_data); + if (adev->umc.ras && adev->umc.ras->ras_block.hw_ops && + adev->umc.ras->ras_block.hw_ops->query_ras_error_count) + adev->umc.ras->ras_block.hw_ops->query_ras_error_count(adev, err_data); /* umc query_ras_error_address is also responsible for clearing * error status */ - if (adev->umc.ras_funcs && - adev->umc.ras_funcs->query_ras_error_address) - adev->umc.ras_funcs->query_ras_error_address(adev, err_data); + if (adev->umc.ras && adev->umc.ras->ras_block.hw_ops && + adev->umc.ras->ras_block.hw_ops->query_ras_error_address) + adev->umc.ras->ras_block.hw_ops->query_ras_error_address(adev, err_data); } else if (!ret) { - if (adev->umc.ras_funcs && - adev->umc.ras_funcs->ecc_info_query_ras_error_count) - adev->umc.ras_funcs->ecc_info_query_ras_error_count(adev, err_data); + if (adev->umc.ras && + adev->umc.ras->ecc_info_query_ras_error_count) + adev->umc.ras->ecc_info_query_ras_error_count(adev, err_data); - if (adev->umc.ras_funcs && - adev->umc.ras_funcs->ecc_info_query_ras_error_address) - adev->umc.ras_funcs->ecc_info_query_ras_error_address(adev, err_data); + if (adev->umc.ras && + adev->umc.ras->ecc_info_query_ras_error_address) + adev->umc.ras->ecc_info_query_ras_error_address(adev, err_data); } } @@ -928,62 +968,32 @@ static void amdgpu_ras_get_ecc_info(struct amdgpu_device *adev, struct ras_err_d int amdgpu_ras_query_error_status(struct amdgpu_device *adev, struct ras_query_if *info) { + struct amdgpu_ras_block_object *block_obj = NULL; struct ras_manager *obj = amdgpu_ras_find_obj(adev, &info->head); struct ras_err_data err_data = {0, 0, 0, NULL}; - int i; if (!obj) return -EINVAL; - switch (info->head.block) { - case AMDGPU_RAS_BLOCK__UMC: + if (info->head.block == AMDGPU_RAS_BLOCK__UMC) { amdgpu_ras_get_ecc_info(adev, &err_data); - break; - case AMDGPU_RAS_BLOCK__SDMA: - if (adev->sdma.funcs->query_ras_error_count) { - for (i = 0; i < adev->sdma.num_instances; i++) - adev->sdma.funcs->query_ras_error_count(adev, i, - &err_data); + } else { + block_obj = amdgpu_ras_get_ras_block(adev, info->head.block, 0); + if (!block_obj || !block_obj->hw_ops) { + dev_dbg_once(adev->dev, "%s doesn't config RAS function\n", + get_ras_block_str(&info->head)); + return -EINVAL; } - break; - case AMDGPU_RAS_BLOCK__GFX: - if (adev->gfx.ras_funcs && - adev->gfx.ras_funcs->query_ras_error_count) - adev->gfx.ras_funcs->query_ras_error_count(adev, &err_data); - - if (adev->gfx.ras_funcs && - adev->gfx.ras_funcs->query_ras_error_status) - adev->gfx.ras_funcs->query_ras_error_status(adev); - break; - case AMDGPU_RAS_BLOCK__MMHUB: - if (adev->mmhub.ras_funcs && - adev->mmhub.ras_funcs->query_ras_error_count) - adev->mmhub.ras_funcs->query_ras_error_count(adev, &err_data); - - if (adev->mmhub.ras_funcs && - adev->mmhub.ras_funcs->query_ras_error_status) - adev->mmhub.ras_funcs->query_ras_error_status(adev); - break; - case AMDGPU_RAS_BLOCK__PCIE_BIF: - if (adev->nbio.ras_funcs && - adev->nbio.ras_funcs->query_ras_error_count) - adev->nbio.ras_funcs->query_ras_error_count(adev, &err_data); - break; - case AMDGPU_RAS_BLOCK__XGMI_WAFL: - if (adev->gmc.xgmi.ras_funcs && - adev->gmc.xgmi.ras_funcs->query_ras_error_count) - adev->gmc.xgmi.ras_funcs->query_ras_error_count(adev, &err_data); - break; - case AMDGPU_RAS_BLOCK__HDP: - if (adev->hdp.ras_funcs && - adev->hdp.ras_funcs->query_ras_error_count) - adev->hdp.ras_funcs->query_ras_error_count(adev, &err_data); - break; - case AMDGPU_RAS_BLOCK__MCA: - amdgpu_ras_mca_query_error_status(adev, &info->head, &err_data); - break; - default: - break; + + if (block_obj->hw_ops->query_ras_error_count) + block_obj->hw_ops->query_ras_error_count(adev, &err_data); + + if ((info->head.block == AMDGPU_RAS_BLOCK__SDMA) || + (info->head.block == AMDGPU_RAS_BLOCK__GFX) || + (info->head.block == AMDGPU_RAS_BLOCK__MMHUB)) { + if (block_obj->hw_ops->query_ras_error_status) + block_obj->hw_ops->query_ras_error_status(adev); + } } obj->err_data.ue_count += err_data.ue_count; @@ -1031,77 +1041,33 @@ int amdgpu_ras_query_error_status(struct amdgpu_device *adev, } } - if (!amdgpu_persistent_edc_harvesting_supported(adev)) - amdgpu_ras_reset_error_status(adev, info->head.block); - return 0; } int amdgpu_ras_reset_error_status(struct amdgpu_device *adev, enum amdgpu_ras_block block) { + struct amdgpu_ras_block_object *block_obj = amdgpu_ras_get_ras_block(adev, block, 0); + if (!amdgpu_ras_is_supported(adev, block)) return -EINVAL; - switch (block) { - case AMDGPU_RAS_BLOCK__GFX: - if (adev->gfx.ras_funcs && - adev->gfx.ras_funcs->reset_ras_error_count) - adev->gfx.ras_funcs->reset_ras_error_count(adev); - - if (adev->gfx.ras_funcs && - adev->gfx.ras_funcs->reset_ras_error_status) - adev->gfx.ras_funcs->reset_ras_error_status(adev); - break; - case AMDGPU_RAS_BLOCK__MMHUB: - if (adev->mmhub.ras_funcs && - adev->mmhub.ras_funcs->reset_ras_error_count) - adev->mmhub.ras_funcs->reset_ras_error_count(adev); - - if (adev->mmhub.ras_funcs && - adev->mmhub.ras_funcs->reset_ras_error_status) - adev->mmhub.ras_funcs->reset_ras_error_status(adev); - break; - case AMDGPU_RAS_BLOCK__SDMA: - if (adev->sdma.funcs->reset_ras_error_count) - adev->sdma.funcs->reset_ras_error_count(adev); - break; - case AMDGPU_RAS_BLOCK__HDP: - if (adev->hdp.ras_funcs && - adev->hdp.ras_funcs->reset_ras_error_count) - adev->hdp.ras_funcs->reset_ras_error_count(adev); - break; - default: - break; + if (!block_obj || !block_obj->hw_ops) { + dev_dbg_once(adev->dev, "%s doesn't config RAS function\n", + ras_block_str(block)); + return -EINVAL; } - return 0; -} - -/* Trigger XGMI/WAFL error */ -static int amdgpu_ras_error_inject_xgmi(struct amdgpu_device *adev, - struct ta_ras_trigger_error_input *block_info) -{ - int ret; + if (block_obj->hw_ops->reset_ras_error_count) + block_obj->hw_ops->reset_ras_error_count(adev); - if (amdgpu_dpm_set_df_cstate(adev, DF_CSTATE_DISALLOW)) - dev_warn(adev->dev, "Failed to disallow df cstate"); - - if (amdgpu_dpm_allow_xgmi_power_down(adev, false)) - dev_warn(adev->dev, "Failed to disallow XGMI power down"); - - ret = psp_ras_trigger_error(&adev->psp, block_info); - - if (amdgpu_ras_intr_triggered()) - return ret; - - if (amdgpu_dpm_allow_xgmi_power_down(adev, true)) - dev_warn(adev->dev, "Failed to allow XGMI power down"); - - if (amdgpu_dpm_set_df_cstate(adev, DF_CSTATE_ALLOW)) - dev_warn(adev->dev, "Failed to allow df cstate"); + if ((block == AMDGPU_RAS_BLOCK__GFX) || + (block == AMDGPU_RAS_BLOCK__MMHUB)) { + if (block_obj->hw_ops->reset_ras_error_status) + block_obj->hw_ops->reset_ras_error_status(adev); + } - return ret; + return 0; } /* wrapper of psp_ras_trigger_error */ @@ -1116,11 +1082,20 @@ int amdgpu_ras_error_inject(struct amdgpu_device *adev, .address = info->address, .value = info->value, }; - int ret = 0; + int ret = -EINVAL; + struct amdgpu_ras_block_object *block_obj = amdgpu_ras_get_ras_block(adev, + info->head.block, + info->head.sub_block_index); if (!obj) return -EINVAL; + if (!block_obj || !block_obj->hw_ops) { + dev_dbg_once(adev->dev, "%s doesn't config RAS function\n", + get_ras_block_str(&info->head)); + return -EINVAL; + } + /* Calculate XGMI relative offset */ if (adev->gmc.xgmi.num_physical_nodes > 1) { block_info.address = @@ -1128,28 +1103,15 @@ int amdgpu_ras_error_inject(struct amdgpu_device *adev, block_info.address); } - switch (info->head.block) { - case AMDGPU_RAS_BLOCK__GFX: - if (adev->gfx.ras_funcs && - adev->gfx.ras_funcs->ras_error_inject) - ret = adev->gfx.ras_funcs->ras_error_inject(adev, info); - else - ret = -EINVAL; - break; - case AMDGPU_RAS_BLOCK__UMC: - case AMDGPU_RAS_BLOCK__SDMA: - case AMDGPU_RAS_BLOCK__MMHUB: - case AMDGPU_RAS_BLOCK__PCIE_BIF: - case AMDGPU_RAS_BLOCK__MCA: - ret = psp_ras_trigger_error(&adev->psp, &block_info); - break; - case AMDGPU_RAS_BLOCK__XGMI_WAFL: - ret = amdgpu_ras_error_inject_xgmi(adev, &block_info); - break; - default: - dev_info(adev->dev, "%s error injection is not supported yet\n", - get_ras_block_str(&info->head)); - ret = -EINVAL; + if (info->head.block == AMDGPU_RAS_BLOCK__GFX) { + if (block_obj->hw_ops->ras_error_inject) + ret = block_obj->hw_ops->ras_error_inject(adev, info); + } else { + /* If defined special ras_error_inject(e.g: xgmi), implement special ras_error_inject */ + if (block_obj->hw_ops->ras_error_inject) + ret = block_obj->hw_ops->ras_error_inject(adev, &block_info); + else /*If not defined .ras_error_inject, use default ras_error_inject*/ + ret = psp_ras_trigger_error(&adev->psp, &block_info); } if (ret) @@ -1198,6 +1160,12 @@ int amdgpu_ras_query_error_count(struct amdgpu_device *adev, if (res) return res; + if (adev->ip_versions[MP0_HWIP][0] != IP_VERSION(11, 0, 2) && + adev->ip_versions[MP0_HWIP][0] != IP_VERSION(11, 0, 4)) { + if (amdgpu_ras_reset_error_status(adev, info.head.block)) + dev_warn(adev->dev, "Failed to reset error counter and error status"); + } + ce += info.ce_count; ue += info.ue_count; } @@ -1329,18 +1297,17 @@ static int amdgpu_ras_sysfs_remove_feature_node(struct amdgpu_device *adev) } int amdgpu_ras_sysfs_create(struct amdgpu_device *adev, - struct ras_fs_if *head) + struct ras_common_if *head) { - struct ras_manager *obj = amdgpu_ras_find_obj(adev, &head->head); + struct ras_manager *obj = amdgpu_ras_find_obj(adev, head); if (!obj || obj->attr_inuse) return -EINVAL; get_obj(obj); - memcpy(obj->fs_data.sysfs_name, - head->sysfs_name, - sizeof(obj->fs_data.sysfs_name)); + snprintf(obj->fs_data.sysfs_name, sizeof(obj->fs_data.sysfs_name), + "%s_err_count", head->name); obj->sysfs_attr = (struct device_attribute){ .attr = { @@ -1567,12 +1534,108 @@ static int amdgpu_ras_fs_fini(struct amdgpu_device *adev) /* ras fs end */ /* ih begin */ + +/* For the hardware that cannot enable bif ring for both ras_controller_irq + * and ras_err_evnet_athub_irq ih cookies, the driver has to poll status + * register to check whether the interrupt is triggered or not, and properly + * ack the interrupt if it is there + */ +void amdgpu_ras_interrupt_fatal_error_handler(struct amdgpu_device *adev) +{ + /* Fatal error events are handled on host side */ + if (amdgpu_sriov_vf(adev) || + !amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__PCIE_BIF)) + return; + + if (adev->nbio.ras && + adev->nbio.ras->handle_ras_controller_intr_no_bifring) + adev->nbio.ras->handle_ras_controller_intr_no_bifring(adev); + + if (adev->nbio.ras && + adev->nbio.ras->handle_ras_err_event_athub_intr_no_bifring) + adev->nbio.ras->handle_ras_err_event_athub_intr_no_bifring(adev); +} + +static void amdgpu_ras_interrupt_poison_consumption_handler(struct ras_manager *obj, + struct amdgpu_iv_entry *entry) +{ + bool poison_stat = false; + struct amdgpu_device *adev = obj->adev; + struct ras_err_data err_data = {0, 0, 0, NULL}; + struct amdgpu_ras_block_object *block_obj = + amdgpu_ras_get_ras_block(adev, obj->head.block, 0); + + if (!block_obj || !block_obj->hw_ops) + return; + + /* both query_poison_status and handle_poison_consumption are optional, + * but at least one of them should be implemented if we need poison + * consumption handler + */ + if (block_obj->hw_ops->query_poison_status) { + poison_stat = block_obj->hw_ops->query_poison_status(adev); + if (!poison_stat) { + /* Not poison consumption interrupt, no need to handle it */ + dev_info(adev->dev, "No RAS poison status in %s poison IH.\n", + block_obj->ras_comm.name); + + return; + } + } + + if (!adev->gmc.xgmi.connected_to_cpu) + amdgpu_umc_poison_handler(adev, &err_data, false); + + if (block_obj->hw_ops->handle_poison_consumption) + poison_stat = block_obj->hw_ops->handle_poison_consumption(adev); + + /* gpu reset is fallback for failed and default cases */ + if (poison_stat) { + dev_info(adev->dev, "GPU reset for %s RAS poison consumption is issued!\n", + block_obj->ras_comm.name); + amdgpu_ras_reset_gpu(adev); + } +} + +static void amdgpu_ras_interrupt_poison_creation_handler(struct ras_manager *obj, + struct amdgpu_iv_entry *entry) +{ + dev_info(obj->adev->dev, + "Poison is created, no user action is needed.\n"); +} + +static void amdgpu_ras_interrupt_umc_handler(struct ras_manager *obj, + struct amdgpu_iv_entry *entry) +{ + struct ras_ih_data *data = &obj->ih_data; + struct ras_err_data err_data = {0, 0, 0, NULL}; + int ret; + + if (!data->cb) + return; + + /* Let IP handle its data, maybe we need get the output + * from the callback to update the error type/count, etc + */ + ret = data->cb(obj->adev, &err_data, entry); + /* ue will trigger an interrupt, and in that case + * we need do a reset to recovery the whole system. + * But leave IP do that recovery, here we just dispatch + * the error. + */ + if (ret == AMDGPU_RAS_SUCCESS) { + /* these counts could be left as 0 if + * some blocks do not count error number + */ + obj->err_data.ue_count += err_data.ue_count; + obj->err_data.ce_count += err_data.ce_count; + } +} + static void amdgpu_ras_interrupt_handler(struct ras_manager *obj) { struct ras_ih_data *data = &obj->ih_data; struct amdgpu_iv_entry entry; - int ret; - struct ras_err_data err_data = {0, 0, 0, NULL}; while (data->rptr != data->wptr) { rmb(); @@ -1583,29 +1646,17 @@ static void amdgpu_ras_interrupt_handler(struct ras_manager *obj) data->rptr = (data->aligned_element_size + data->rptr) % data->ring_size; - if (data->cb) { - if (amdgpu_ras_is_poison_mode_supported(obj->adev) && - obj->head.block == AMDGPU_RAS_BLOCK__UMC) - dev_info(obj->adev->dev, - "Poison is created, no user action is needed.\n"); - else { - /* Let IP handle its data, maybe we need get the output - * from the callback to udpate the error type/count, etc - */ - ret = data->cb(obj->adev, &err_data, &entry); - /* ue will trigger an interrupt, and in that case - * we need do a reset to recovery the whole system. - * But leave IP do that recovery, here we just dispatch - * the error. - */ - if (ret == AMDGPU_RAS_SUCCESS) { - /* these counts could be left as 0 if - * some blocks do not count error number - */ - obj->err_data.ue_count += err_data.ue_count; - obj->err_data.ce_count += err_data.ce_count; - } - } + if (amdgpu_ras_is_poison_mode_supported(obj->adev)) { + if (obj->head.block == AMDGPU_RAS_BLOCK__UMC) + amdgpu_ras_interrupt_poison_creation_handler(obj, &entry); + else + amdgpu_ras_interrupt_poison_consumption_handler(obj, &entry); + } else { + if (obj->head.block == AMDGPU_RAS_BLOCK__UMC) + amdgpu_ras_interrupt_umc_handler(obj, &entry); + else + dev_warn(obj->adev->dev, + "No RAS interrupt handler for non-UMC block with poison disabled.\n"); } } } @@ -1646,9 +1697,9 @@ int amdgpu_ras_interrupt_dispatch(struct amdgpu_device *adev, } int amdgpu_ras_interrupt_remove_handler(struct amdgpu_device *adev, - struct ras_ih_if *info) + struct ras_common_if *head) { - struct ras_manager *obj = amdgpu_ras_find_obj(adev, &info->head); + struct ras_manager *obj = amdgpu_ras_find_obj(adev, head); struct ras_ih_data *data; if (!obj) @@ -1668,24 +1719,27 @@ int amdgpu_ras_interrupt_remove_handler(struct amdgpu_device *adev, } int amdgpu_ras_interrupt_add_handler(struct amdgpu_device *adev, - struct ras_ih_if *info) + struct ras_common_if *head) { - struct ras_manager *obj = amdgpu_ras_find_obj(adev, &info->head); + struct ras_manager *obj = amdgpu_ras_find_obj(adev, head); struct ras_ih_data *data; + struct amdgpu_ras_block_object *ras_obj; if (!obj) { /* in case we registe the IH before enable ras feature */ - obj = amdgpu_ras_create_obj(adev, &info->head); + obj = amdgpu_ras_create_obj(adev, head); if (!obj) return -EINVAL; } else get_obj(obj); + ras_obj = container_of(head, struct amdgpu_ras_block_object, ras_comm); + data = &obj->ih_data; /* add the callback.etc */ *data = (struct ras_ih_data) { .inuse = 0, - .cb = info->cb, + .cb = ras_obj->ras_cb, .element_size = sizeof(struct amdgpu_iv_entry), .rptr = 0, .wptr = 0, @@ -1714,10 +1768,7 @@ static int amdgpu_ras_interrupt_remove_all(struct amdgpu_device *adev) struct ras_manager *obj, *tmp; list_for_each_entry_safe(obj, tmp, &con->head, node) { - struct ras_ih_if info = { - .head = obj->head, - }; - amdgpu_ras_interrupt_remove_handler(adev, &info); + amdgpu_ras_interrupt_remove_handler(adev, &obj->head); } return 0; @@ -1758,6 +1809,13 @@ static void amdgpu_ras_log_on_err_counter(struct amdgpu_device *adev) continue; amdgpu_ras_query_error_status(adev, &info); + + if (adev->ip_versions[MP0_HWIP][0] != IP_VERSION(11, 0, 2) && + adev->ip_versions[MP0_HWIP][0] != IP_VERSION(11, 0, 4) && + adev->ip_versions[MP0_HWIP][0] != IP_VERSION(13, 0, 0)) { + if (amdgpu_ras_reset_error_status(adev, info.head.block)) + dev_warn(adev->dev, "Failed to reset error counter and error status"); + } } } @@ -1765,24 +1823,28 @@ static void amdgpu_ras_log_on_err_counter(struct amdgpu_device *adev) static void amdgpu_ras_error_status_query(struct amdgpu_device *adev, struct ras_query_if *info) { + struct amdgpu_ras_block_object *block_obj; /* * Only two block need to query read/write * RspStatus at current state */ - switch (info->head.block) { - case AMDGPU_RAS_BLOCK__GFX: - if (adev->gfx.ras_funcs && - adev->gfx.ras_funcs->query_ras_error_status) - adev->gfx.ras_funcs->query_ras_error_status(adev); - break; - case AMDGPU_RAS_BLOCK__MMHUB: - if (adev->mmhub.ras_funcs && - adev->mmhub.ras_funcs->query_ras_error_status) - adev->mmhub.ras_funcs->query_ras_error_status(adev); - break; - default: - break; + if ((info->head.block != AMDGPU_RAS_BLOCK__GFX) && + (info->head.block != AMDGPU_RAS_BLOCK__MMHUB)) + return; + + block_obj = amdgpu_ras_get_ras_block(adev, + info->head.block, + info->head.sub_block_index); + + if (!block_obj || !block_obj->hw_ops) { + dev_dbg_once(adev->dev, "%s doesn't config RAS function\n", + get_ras_block_str(&info->head)); + return; } + + if (block_obj->hw_ops->query_ras_error_status) + block_obj->hw_ops->query_ras_error_status(adev); + } static void amdgpu_ras_query_err_status(struct amdgpu_device *adev) @@ -1838,8 +1900,7 @@ static int amdgpu_ras_badpages_read(struct amdgpu_device *adev, .size = AMDGPU_GPU_PAGE_SIZE, .flags = AMDGPU_RAS_RETIRE_PAGE_RESERVED, }; - status = amdgpu_vram_mgr_query_page_status( - ttm_manager_type(&adev->mman.bdev, TTM_PL_VRAM), + status = amdgpu_vram_mgr_query_page_status(&adev->mman.vram_mgr, data->bps[i].retired_page); if (status == -EBUSY) (*bps)[i].flags = AMDGPU_RAS_RETIRE_PAGE_PENDING; @@ -1882,8 +1943,16 @@ static void amdgpu_ras_do_recovery(struct work_struct *work) amdgpu_put_xgmi_hive(hive); } - if (amdgpu_device_should_recover_gpu(ras->adev)) - amdgpu_device_gpu_recover(ras->adev, NULL); + if (amdgpu_device_should_recover_gpu(ras->adev)) { + struct amdgpu_reset_context reset_context; + memset(&reset_context, 0, sizeof(reset_context)); + + reset_context.method = AMD_RESET_METHOD_NONE; + reset_context.reset_req_dev = adev; + clear_bit(AMDGPU_NEED_FULL_RESET, &reset_context.flags); + + amdgpu_device_gpu_recover(ras->adev, NULL, &reset_context); + } atomic_set(&ras->in_recovery, 0); } @@ -1897,7 +1966,6 @@ static int amdgpu_ras_realloc_eh_data_space(struct amdgpu_device *adev, void *bps = kmalloc(align_space * sizeof(*data->bps), GFP_KERNEL); if (!bps) { - kfree(bps); return -ENOMEM; } @@ -1940,8 +2008,7 @@ int amdgpu_ras_add_bad_pages(struct amdgpu_device *adev, goto out; } - amdgpu_vram_mgr_reserve_range( - ttm_manager_type(&adev->mman.bdev, TTM_PL_VRAM), + amdgpu_vram_mgr_reserve_range(&adev->mman.vram_mgr, bps[i].retired_page << AMDGPU_GPU_PAGE_SHIFT, AMDGPU_GPU_PAGE_SIZE); @@ -2096,7 +2163,7 @@ int amdgpu_ras_recovery_init(struct amdgpu_device *adev) bool exc_err_limit = false; int ret; - if (!con) + if (!con || amdgpu_sriov_vf(adev)) return 0; /* Allow access to RAS EEPROM via debugfs, when the ASIC @@ -2119,6 +2186,7 @@ int amdgpu_ras_recovery_init(struct amdgpu_device *adev) mutex_init(&con->recovery_lock); INIT_WORK(&con->recovery_work, amdgpu_ras_do_recovery); atomic_set(&con->in_recovery, 0); + con->eeprom_control.bad_channel_bitmap = 0; max_eeprom_records_count = amdgpu_ras_eeprom_max_record_count(); amdgpu_ras_validate_threshold(adev, max_eeprom_records_count); @@ -2142,8 +2210,12 @@ int amdgpu_ras_recovery_init(struct amdgpu_device *adev) if (ret) goto free; - if (adev->smu.ppt_funcs && adev->smu.ppt_funcs->send_hbm_bad_pages_num) - adev->smu.ppt_funcs->send_hbm_bad_pages_num(&adev->smu, con->eeprom_control.ras_num_recs); + amdgpu_dpm_send_hbm_bad_pages_num(adev, con->eeprom_control.ras_num_recs); + + if (con->update_channel_flag == true) { + amdgpu_dpm_send_hbm_bad_channel_flag(adev, con->eeprom_control.bad_channel_bitmap); + con->update_channel_flag = false; + } } #ifdef CONFIG_X86_MCE_AMD @@ -2195,6 +2267,25 @@ static int amdgpu_ras_recovery_fini(struct amdgpu_device *adev) static bool amdgpu_ras_asic_supported(struct amdgpu_device *adev) { + if (amdgpu_sriov_vf(adev)) { + switch (adev->ip_versions[MP0_HWIP][0]) { + case IP_VERSION(13, 0, 2): + return true; + default: + return false; + } + } + + if (adev->asic_type == CHIP_IP_DISCOVERY) { + switch (adev->ip_versions[MP0_HWIP][0]) { + case IP_VERSION(13, 0, 0): + case IP_VERSION(13, 0, 10): + return true; + default: + return false; + } + } + return adev->asic_type == CHIP_VEGA10 || adev->asic_type == CHIP_VEGA20 || adev->asic_type == CHIP_ARCTURUS || @@ -2234,7 +2325,7 @@ static void amdgpu_ras_check_supported(struct amdgpu_device *adev) { adev->ras_hw_enabled = adev->ras_enabled = 0; - if (amdgpu_sriov_vf(adev) || !adev->is_atom_fw || + if (!adev->is_atom_fw || !amdgpu_ras_asic_supported(adev)) return; @@ -2249,8 +2340,21 @@ static void amdgpu_ras_check_supported(struct amdgpu_device *adev) if (amdgpu_atomfirmware_sram_ecc_supported(adev)) { dev_info(adev->dev, "SRAM ECC is active.\n"); - adev->ras_hw_enabled |= ~(1 << AMDGPU_RAS_BLOCK__UMC | - 1 << AMDGPU_RAS_BLOCK__DF); + if (!amdgpu_sriov_vf(adev)) { + adev->ras_hw_enabled |= ~(1 << AMDGPU_RAS_BLOCK__UMC | + 1 << AMDGPU_RAS_BLOCK__DF); + + if (adev->ip_versions[VCN_HWIP][0] == IP_VERSION(2, 6, 0)) + adev->ras_hw_enabled |= (1 << AMDGPU_RAS_BLOCK__VCN | + 1 << AMDGPU_RAS_BLOCK__JPEG); + else + adev->ras_hw_enabled &= ~(1 << AMDGPU_RAS_BLOCK__VCN | + 1 << AMDGPU_RAS_BLOCK__JPEG); + } else { + adev->ras_hw_enabled |= (1 << AMDGPU_RAS_BLOCK__PCIE_BIF | + 1 << AMDGPU_RAS_BLOCK__SDMA | + 1 << AMDGPU_RAS_BLOCK__GFX); + } } else { dev_info(adev->dev, "SRAM ECC is not presented.\n"); } @@ -2337,6 +2441,7 @@ int amdgpu_ras_init(struct amdgpu_device *adev) goto release_con; } + con->update_channel_flag = false; con->features = 0; INIT_LIST_HEAD(&con->head); /* Might need get this flag from vbios. */ @@ -2349,24 +2454,27 @@ int amdgpu_ras_init(struct amdgpu_device *adev) case CHIP_VEGA20: case CHIP_ARCTURUS: case CHIP_ALDEBARAN: - if (!adev->gmc.xgmi.connected_to_cpu) - adev->nbio.ras_funcs = &nbio_v7_4_ras_funcs; + if (!adev->gmc.xgmi.connected_to_cpu) { + adev->nbio.ras = &nbio_v7_4_ras; + amdgpu_ras_register_ras_block(adev, &adev->nbio.ras->ras_block); + adev->nbio.ras_if = &adev->nbio.ras->ras_block.ras_comm; + } break; default: /* nbio ras is not available */ break; } - if (adev->nbio.ras_funcs && - adev->nbio.ras_funcs->init_ras_controller_interrupt) { - r = adev->nbio.ras_funcs->init_ras_controller_interrupt(adev); + if (adev->nbio.ras && + adev->nbio.ras->init_ras_controller_interrupt) { + r = adev->nbio.ras->init_ras_controller_interrupt(adev); if (r) goto release_con; } - if (adev->nbio.ras_funcs && - adev->nbio.ras_funcs->init_ras_err_event_athub_interrupt) { - r = adev->nbio.ras_funcs->init_ras_err_event_athub_interrupt(adev); + if (adev->nbio.ras && + adev->nbio.ras->init_ras_err_event_athub_interrupt) { + r = adev->nbio.ras->init_ras_err_event_athub_interrupt(adev); if (r) goto release_con; } @@ -2378,12 +2486,12 @@ int amdgpu_ras_init(struct amdgpu_device *adev) } else if (adev->df.funcs && adev->df.funcs->query_ras_poison_mode && - adev->umc.ras_funcs && - adev->umc.ras_funcs->query_ras_poison_mode) { + adev->umc.ras && + adev->umc.ras->query_ras_poison_mode) { df_poison = adev->df.funcs->query_ras_poison_mode(adev); umc_poison = - adev->umc.ras_funcs->query_ras_poison_mode(adev); + adev->umc.ras->query_ras_poison_mode(adev); /* Only poison is set in both DF and UMC, we can support it */ if (df_poison && umc_poison) con->poison_supported = true; @@ -2446,11 +2554,10 @@ bool amdgpu_ras_is_poison_mode_supported(struct amdgpu_device *adev) } /* helper function to handle common stuff in ip late init phase */ -int amdgpu_ras_late_init(struct amdgpu_device *adev, - struct ras_common_if *ras_block, - struct ras_fs_if *fs_info, - struct ras_ih_if *ih_info) +int amdgpu_ras_block_late_init(struct amdgpu_device *adev, + struct ras_common_if *ras_block) { + struct amdgpu_ras_block_object *ras_obj = NULL; struct amdgpu_ras *con = amdgpu_ras_get_context(adev); unsigned long ue_count, ce_count; int r; @@ -2478,15 +2585,18 @@ int amdgpu_ras_late_init(struct amdgpu_device *adev, if (adev->in_suspend || amdgpu_in_reset(adev)) return 0; - if (ih_info->cb) { - r = amdgpu_ras_interrupt_add_handler(adev, ih_info); + ras_obj = container_of(ras_block, struct amdgpu_ras_block_object, ras_comm); + if (ras_obj->ras_cb || (ras_obj->hw_ops && + (ras_obj->hw_ops->query_poison_status || + ras_obj->hw_ops->handle_poison_consumption))) { + r = amdgpu_ras_interrupt_add_handler(adev, ras_block); if (r) - goto interrupt; + goto cleanup; } - r = amdgpu_ras_sysfs_create(adev, fs_info); + r = amdgpu_ras_sysfs_create(adev, ras_block); if (r) - goto sysfs; + goto interrupt; /* Those are the cached values at init. */ @@ -2496,27 +2606,40 @@ int amdgpu_ras_late_init(struct amdgpu_device *adev, } return 0; -cleanup: - amdgpu_ras_sysfs_remove(adev, ras_block); -sysfs: - if (ih_info->cb) - amdgpu_ras_interrupt_remove_handler(adev, ih_info); + interrupt: + if (ras_obj->ras_cb) + amdgpu_ras_interrupt_remove_handler(adev, ras_block); +cleanup: amdgpu_ras_feature_enable(adev, ras_block, 0); return r; } +static int amdgpu_ras_block_late_init_default(struct amdgpu_device *adev, + struct ras_common_if *ras_block) +{ + return amdgpu_ras_block_late_init(adev, ras_block); +} + /* helper function to remove ras fs node and interrupt handler */ -void amdgpu_ras_late_fini(struct amdgpu_device *adev, - struct ras_common_if *ras_block, - struct ras_ih_if *ih_info) +void amdgpu_ras_block_late_fini(struct amdgpu_device *adev, + struct ras_common_if *ras_block) { - if (!ras_block || !ih_info) + struct amdgpu_ras_block_object *ras_obj; + if (!ras_block) return; amdgpu_ras_sysfs_remove(adev, ras_block); - if (ih_info->cb) - amdgpu_ras_interrupt_remove_handler(adev, ih_info); + + ras_obj = container_of(ras_block, struct amdgpu_ras_block_object, ras_comm); + if (ras_obj->ras_cb) + amdgpu_ras_interrupt_remove_handler(adev, ras_block); +} + +static void amdgpu_ras_block_late_fini_default(struct amdgpu_device *adev, + struct ras_common_if *ras_block) +{ + return amdgpu_ras_block_late_fini(adev, ras_block); } /* do some init work after IP late init as dependence. @@ -2569,6 +2692,37 @@ void amdgpu_ras_suspend(struct amdgpu_device *adev) amdgpu_ras_disable_all_features(adev, 1); } +int amdgpu_ras_late_init(struct amdgpu_device *adev) +{ + struct amdgpu_ras_block_list *node, *tmp; + struct amdgpu_ras_block_object *obj; + int r; + + /* Guest side doesn't need init ras feature */ + if (amdgpu_sriov_vf(adev)) + return 0; + + list_for_each_entry_safe(node, tmp, &adev->ras_list, node) { + if (!node->ras_obj) { + dev_warn(adev->dev, "Warning: abnormal ras list node.\n"); + continue; + } + + obj = node->ras_obj; + if (obj->ras_late_init) { + r = obj->ras_late_init(adev, &obj->ras_comm); + if (r) { + dev_err(adev->dev, "%s failed to execute ras_late_init! ret:%d\n", + obj->ras_comm.name, r); + return r; + } + } else + amdgpu_ras_block_late_init_default(adev, &obj->ras_comm); + } + + return 0; +} + /* do some fini work before IP fini as dependence */ int amdgpu_ras_pre_fini(struct amdgpu_device *adev) { @@ -2579,18 +2733,36 @@ int amdgpu_ras_pre_fini(struct amdgpu_device *adev) /* Need disable ras on all IPs here before ip [hw/sw]fini */ - amdgpu_ras_disable_all_features(adev, 0); + if (con->features) + amdgpu_ras_disable_all_features(adev, 0); amdgpu_ras_recovery_fini(adev); return 0; } int amdgpu_ras_fini(struct amdgpu_device *adev) { + struct amdgpu_ras_block_list *ras_node, *tmp; + struct amdgpu_ras_block_object *obj = NULL; struct amdgpu_ras *con = amdgpu_ras_get_context(adev); if (!adev->ras_enabled || !con) return 0; + list_for_each_entry_safe(ras_node, tmp, &adev->ras_list, node) { + if (ras_node->ras_obj) { + obj = ras_node->ras_obj; + if (amdgpu_ras_is_supported(adev, obj->ras_comm.block) && + obj->ras_fini) + obj->ras_fini(adev, &obj->ras_comm); + else + amdgpu_ras_block_late_fini_default(adev, &obj->ras_comm); + } + + /* Clear ras blocks from ras_list and free ras block list node */ + list_del(&ras_node->node); + kfree(ras_node); + } + amdgpu_ras_fs_fini(adev); amdgpu_ras_interrupt_remove_all(adev); @@ -2675,11 +2847,8 @@ static int amdgpu_bad_page_notifier(struct notifier_block *nb, struct mce *m = (struct mce *)data; struct amdgpu_device *adev = NULL; uint32_t gpu_id = 0; - uint32_t umc_inst = 0; - uint32_t ch_inst, channel_index = 0; + uint32_t umc_inst = 0, ch_inst = 0; struct ras_err_data err_data = {0, 0, 0, NULL}; - struct eeprom_table_record err_rec; - uint64_t retired_page; /* * If the error was generated in UMC_V2, which belongs to GPU UMCs, @@ -2718,29 +2887,22 @@ static int amdgpu_bad_page_notifier(struct notifier_block *nb, dev_info(adev->dev, "Uncorrectable error detected in UMC inst: %d, chan_idx: %d", umc_inst, ch_inst); - memset(&err_rec, 0x0, sizeof(struct eeprom_table_record)); + err_data.err_addr = + kcalloc(adev->umc.max_ras_err_cnt_per_query, + sizeof(struct eeprom_table_record), GFP_KERNEL); + if (!err_data.err_addr) { + dev_warn(adev->dev, + "Failed to alloc memory for umc error record in mca notifier!\n"); + return NOTIFY_DONE; + } /* * Translate UMC channel address to Physical address */ - channel_index = - adev->umc.channel_idx_tbl[umc_inst * adev->umc.channel_inst_num - + ch_inst]; - - retired_page = ADDR_OF_8KB_BLOCK(m->addr) | - ADDR_OF_256B_BLOCK(channel_index) | - OFFSET_IN_256B_BLOCK(m->addr); - - err_rec.address = m->addr; - err_rec.retired_page = retired_page >> AMDGPU_GPU_PAGE_SHIFT; - err_rec.ts = (uint64_t)ktime_get_real_seconds(); - err_rec.err_type = AMDGPU_RAS_EEPROM_ERR_NON_RECOVERABLE; - err_rec.cu = 0; - err_rec.mem_channel = channel_index; - err_rec.mcumc_id = umc_inst; - - err_data.err_addr = &err_rec; - err_data.err_addr_cnt = 1; + if (adev->umc.ras && + adev->umc.ras->convert_ras_error_address) + adev->umc.ras->convert_ras_error_address(adev, + &err_data, m->addr, ch_inst, umc_inst); if (amdgpu_bad_page_threshold != 0) { amdgpu_ras_add_bad_pages(adev, err_data.err_addr, @@ -2748,6 +2910,7 @@ static int amdgpu_bad_page_notifier(struct notifier_block *nb, amdgpu_ras_save_bad_pages(adev); } + kfree(err_data.err_addr); return NOTIFY_OK; } @@ -2778,3 +2941,63 @@ static void amdgpu_register_bad_pages_mca_notifier(struct amdgpu_device *adev) } } #endif + +struct amdgpu_ras *amdgpu_ras_get_context(struct amdgpu_device *adev) +{ + if (!adev) + return NULL; + + return adev->psp.ras_context.ras; +} + +int amdgpu_ras_set_context(struct amdgpu_device *adev, struct amdgpu_ras *ras_con) +{ + if (!adev) + return -EINVAL; + + adev->psp.ras_context.ras = ras_con; + return 0; +} + +/* check if ras is supported on block, say, sdma, gfx */ +int amdgpu_ras_is_supported(struct amdgpu_device *adev, + unsigned int block) +{ + struct amdgpu_ras *ras = amdgpu_ras_get_context(adev); + + if (block >= AMDGPU_RAS_BLOCK_COUNT) + return 0; + return ras && (adev->ras_enabled & (1 << block)); +} + +int amdgpu_ras_reset_gpu(struct amdgpu_device *adev) +{ + struct amdgpu_ras *ras = amdgpu_ras_get_context(adev); + + if (atomic_cmpxchg(&ras->in_recovery, 0, 1) == 0) + amdgpu_reset_domain_schedule(ras->adev->reset_domain, &ras->recovery_work); + return 0; +} + + +/* Register each ip ras block into amdgpu ras */ +int amdgpu_ras_register_ras_block(struct amdgpu_device *adev, + struct amdgpu_ras_block_object *ras_block_obj) +{ + struct amdgpu_ras_block_list *ras_node; + if (!adev || !ras_block_obj) + return -EINVAL; + + if (!amdgpu_ras_asic_supported(adev)) + return 0; + + ras_node = kzalloc(sizeof(*ras_node), GFP_KERNEL); + if (!ras_node) + return -ENOMEM; + + INIT_LIST_HEAD(&ras_node->node); + ras_node->ras_obj = ras_block_obj; + list_add_tail(&ras_node->node, &adev->ras_list); + + return 0; +} diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h index 1c708122d492..bf5a95104ec1 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h @@ -26,11 +26,11 @@ #include <linux/debugfs.h> #include <linux/list.h> -#include "amdgpu.h" -#include "amdgpu_psp.h" #include "ta_ras_if.h" #include "amdgpu_ras_eeprom.h" +struct amdgpu_iv_entry; + #define AMDGPU_RAS_FLAG_INIT_BY_VBIOS (0x1 << 0) enum amdgpu_ras_block { @@ -49,6 +49,8 @@ enum amdgpu_ras_block { AMDGPU_RAS_BLOCK__MP1, AMDGPU_RAS_BLOCK__FUSE, AMDGPU_RAS_BLOCK__MCA, + AMDGPU_RAS_BLOCK__VCN, + AMDGPU_RAS_BLOCK__JPEG, AMDGPU_RAS_BLOCK__LAST }; @@ -326,10 +328,16 @@ struct ecc_info_per_ch { uint16_t ce_count_hi_chip; uint64_t mca_umc_status; uint64_t mca_umc_addr; + uint64_t mca_ceumc_addr; }; struct umc_ecc_info { struct ecc_info_per_ch ecc[MAX_UMC_CHANNEL_NUM]; + + /* Determine smu ecctable whether support + * record correctable error address + */ + int record_ce_addr_supported; }; struct amdgpu_ras { @@ -374,6 +382,9 @@ struct amdgpu_ras { /* record umc error info queried from smu */ struct umc_ecc_info umc_ecc; + + /* Indicates smu whether need update bad channel info */ + bool update_channel_flag; }; struct ras_fs_data { @@ -484,6 +495,29 @@ struct ras_debug_if { }; int op; }; + +struct amdgpu_ras_block_object { + struct ras_common_if ras_comm; + + int (*ras_block_match)(struct amdgpu_ras_block_object *block_obj, + enum amdgpu_ras_block block, uint32_t sub_block_index); + int (*ras_late_init)(struct amdgpu_device *adev, struct ras_common_if *ras_block); + void (*ras_fini)(struct amdgpu_device *adev, struct ras_common_if *ras_block); + ras_ih_cb ras_cb; + const struct amdgpu_ras_block_hw_ops *hw_ops; +}; + +struct amdgpu_ras_block_hw_ops { + int (*ras_error_inject)(struct amdgpu_device *adev, void *inject_if); + void (*query_ras_error_count)(struct amdgpu_device *adev, void *ras_error_status); + void (*query_ras_error_status)(struct amdgpu_device *adev); + void (*query_ras_error_address)(struct amdgpu_device *adev, void *ras_error_status); + void (*reset_ras_error_count)(struct amdgpu_device *adev); + void (*reset_ras_error_status)(struct amdgpu_device *adev); + bool (*query_poison_status)(struct amdgpu_device *adev); + bool (*handle_poison_consumption)(struct amdgpu_device *adev); +}; + /* work flow * vbios * 1: ras feature enable (enabled by default) @@ -498,19 +532,6 @@ struct ras_debug_if { * 8: feature disable */ -#define amdgpu_ras_get_context(adev) ((adev)->psp.ras_context.ras) -#define amdgpu_ras_set_context(adev, ras_con) ((adev)->psp.ras_context.ras = (ras_con)) - -/* check if ras is supported on block, say, sdma, gfx */ -static inline int amdgpu_ras_is_supported(struct amdgpu_device *adev, - unsigned int block) -{ - struct amdgpu_ras *ras = amdgpu_ras_get_context(adev); - - if (block >= AMDGPU_RAS_BLOCK_COUNT) - return 0; - return ras && (adev->ras_enabled & (1 << block)); -} int amdgpu_ras_recovery_init(struct amdgpu_device *adev); @@ -527,15 +548,6 @@ int amdgpu_ras_add_bad_pages(struct amdgpu_device *adev, int amdgpu_ras_save_bad_pages(struct amdgpu_device *adev); -static inline int amdgpu_ras_reset_gpu(struct amdgpu_device *adev) -{ - struct amdgpu_ras *ras = amdgpu_ras_get_context(adev); - - if (atomic_cmpxchg(&ras->in_recovery, 0, 1) == 0) - schedule_work(&ras->recovery_work); - return 0; -} - static inline enum ta_ras_block amdgpu_ras_block_to_ta(enum amdgpu_ras_block block) { switch (block) { @@ -596,15 +608,15 @@ amdgpu_ras_error_to_ta(enum amdgpu_ras_error_type error) { /* called in ip_init and ip_fini */ int amdgpu_ras_init(struct amdgpu_device *adev); +int amdgpu_ras_late_init(struct amdgpu_device *adev); int amdgpu_ras_fini(struct amdgpu_device *adev); int amdgpu_ras_pre_fini(struct amdgpu_device *adev); -int amdgpu_ras_late_init(struct amdgpu_device *adev, - struct ras_common_if *ras_block, - struct ras_fs_if *fs_info, - struct ras_ih_if *ih_info); -void amdgpu_ras_late_fini(struct amdgpu_device *adev, - struct ras_common_if *ras_block, - struct ras_ih_if *ih_info); + +int amdgpu_ras_block_late_init(struct amdgpu_device *adev, + struct ras_common_if *ras_block); + +void amdgpu_ras_block_late_fini(struct amdgpu_device *adev, + struct ras_common_if *ras_block); int amdgpu_ras_feature_enable(struct amdgpu_device *adev, struct ras_common_if *head, bool enable); @@ -613,7 +625,7 @@ int amdgpu_ras_feature_enable_on_boot(struct amdgpu_device *adev, struct ras_common_if *head, bool enable); int amdgpu_ras_sysfs_create(struct amdgpu_device *adev, - struct ras_fs_if *head); + struct ras_common_if *head); int amdgpu_ras_sysfs_remove(struct amdgpu_device *adev, struct ras_common_if *head); @@ -630,10 +642,10 @@ int amdgpu_ras_error_inject(struct amdgpu_device *adev, struct ras_inject_if *info); int amdgpu_ras_interrupt_add_handler(struct amdgpu_device *adev, - struct ras_ih_if *info); + struct ras_common_if *head); int amdgpu_ras_interrupt_remove_handler(struct amdgpu_device *adev, - struct ras_ih_if *info); + struct ras_common_if *head); int amdgpu_ras_interrupt_dispatch(struct amdgpu_device *adev, struct ras_dispatch_if *info); @@ -667,4 +679,15 @@ const char *get_ras_block_str(struct ras_common_if *ras_block); bool amdgpu_ras_is_poison_mode_supported(struct amdgpu_device *adev); +int amdgpu_ras_is_supported(struct amdgpu_device *adev, unsigned int block); + +int amdgpu_ras_reset_gpu(struct amdgpu_device *adev); + +struct amdgpu_ras* amdgpu_ras_get_context(struct amdgpu_device *adev); + +int amdgpu_ras_set_context(struct amdgpu_device *adev, struct amdgpu_ras *ras_con); + +int amdgpu_ras_register_ras_block(struct amdgpu_device *adev, + struct amdgpu_ras_block_object *ras_block_obj); +void amdgpu_ras_interrupt_fatal_error_handler(struct amdgpu_device *adev); #endif diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras_eeprom.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras_eeprom.c index 05117eda105b..84c241b9a2a1 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras_eeprom.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras_eeprom.c @@ -31,11 +31,14 @@ #include <linux/debugfs.h> #include <linux/uaccess.h> +#include "amdgpu_reset.h" + #define EEPROM_I2C_MADDR_VEGA20 0x0 #define EEPROM_I2C_MADDR_ARCTURUS 0x40000 #define EEPROM_I2C_MADDR_ARCTURUS_D342 0x0 #define EEPROM_I2C_MADDR_SIENNA_CICHLID 0x0 #define EEPROM_I2C_MADDR_ALDEBARAN 0x0 +#define EEPROM_I2C_MADDR_SMU_13_0_0 (0x54UL << 16) /* * The 2 macros bellow represent the actual size in bytes that @@ -154,6 +157,15 @@ static bool __get_eeprom_i2c_addr(struct amdgpu_device *adev, return false; } + switch (adev->ip_versions[MP1_HWIP][0]) { + case IP_VERSION(13, 0, 0): + control->i2c_address = EEPROM_I2C_MADDR_SMU_13_0_0; + break; + + default: + break; + } + return true; } @@ -193,12 +205,12 @@ static int __write_table_header(struct amdgpu_ras_eeprom_control *control) __encode_table_header_to_buf(&control->tbl_hdr, buf); /* i2c may be unstable in gpu reset */ - down_read(&adev->reset_sem); - res = amdgpu_eeprom_write(&adev->pm.smu_i2c, + down_read(&adev->reset_domain->sem); + res = amdgpu_eeprom_write(adev->pm.ras_eeprom_i2c_bus, control->i2c_address + control->ras_header_offset, buf, RAS_TABLE_HEADER_SIZE); - up_read(&adev->reset_sem); + up_read(&adev->reset_domain->sem); if (res < 0) { DRM_ERROR("Failed to write EEPROM table header:%d", res); @@ -263,7 +275,9 @@ static int amdgpu_ras_eeprom_correct_header_tag( */ int amdgpu_ras_eeprom_reset_table(struct amdgpu_ras_eeprom_control *control) { + struct amdgpu_device *adev = to_amdgpu_device(control); struct amdgpu_ras_eeprom_table_header *hdr = &control->tbl_hdr; + struct amdgpu_ras *con = amdgpu_ras_get_context(adev); u8 csum; int res; @@ -282,6 +296,12 @@ int amdgpu_ras_eeprom_reset_table(struct amdgpu_ras_eeprom_control *control) control->ras_num_recs = 0; control->ras_fri = 0; + amdgpu_dpm_send_hbm_bad_pages_num(adev, control->ras_num_recs); + + control->bad_channel_bitmap = 0; + amdgpu_dpm_send_hbm_bad_channel_flag(adev, control->bad_channel_bitmap); + con->update_channel_flag = false; + amdgpu_ras_debugfs_set_ret_size(control); mutex_unlock(&control->ras_tbl_mutex); @@ -387,13 +407,13 @@ static int __amdgpu_ras_eeprom_write(struct amdgpu_ras_eeprom_control *control, int res; /* i2c may be unstable in gpu reset */ - down_read(&adev->reset_sem); + down_read(&adev->reset_domain->sem); buf_size = num * RAS_TABLE_RECORD_SIZE; - res = amdgpu_eeprom_write(&adev->pm.smu_i2c, + res = amdgpu_eeprom_write(adev->pm.ras_eeprom_i2c_bus, control->i2c_address + RAS_INDEX_TO_OFFSET(control, fri), buf, buf_size); - up_read(&adev->reset_sem); + up_read(&adev->reset_domain->sem); if (res < 0) { DRM_ERROR("Writing %d EEPROM table records error:%d", num, res); @@ -415,6 +435,7 @@ amdgpu_ras_eeprom_append_table(struct amdgpu_ras_eeprom_control *control, struct eeprom_table_record *record, const u32 num) { + struct amdgpu_ras *con = amdgpu_ras_get_context(to_amdgpu_device(control)); u32 a, b, i; u8 *buf, *pp; int res; @@ -426,9 +447,16 @@ amdgpu_ras_eeprom_append_table(struct amdgpu_ras_eeprom_control *control, /* Encode all of them in one go. */ pp = buf; - for (i = 0; i < num; i++, pp += RAS_TABLE_RECORD_SIZE) + for (i = 0; i < num; i++, pp += RAS_TABLE_RECORD_SIZE) { __encode_table_record_to_buf(control, &record[i], pp); + /* update bad channel bitmap */ + if (!(control->bad_channel_bitmap & (1 << record[i].mem_channel))) { + control->bad_channel_bitmap |= 1 << record[i].mem_channel; + con->update_channel_flag = true; + } + } + /* a, first record index to write into. * b, last record index to write into. * a = first index to read (fri) + number of records in the table, @@ -547,12 +575,12 @@ amdgpu_ras_eeprom_update_header(struct amdgpu_ras_eeprom_control *control) goto Out; } - down_read(&adev->reset_sem); - res = amdgpu_eeprom_read(&adev->pm.smu_i2c, + down_read(&adev->reset_domain->sem); + res = amdgpu_eeprom_read(adev->pm.ras_eeprom_i2c_bus, control->i2c_address + control->ras_record_offset, buf, buf_size); - up_read(&adev->reset_sem); + up_read(&adev->reset_domain->sem); if (res < 0) { DRM_ERROR("EEPROM failed reading records:%d\n", res); @@ -642,13 +670,13 @@ static int __amdgpu_ras_eeprom_read(struct amdgpu_ras_eeprom_control *control, int res; /* i2c may be unstable in gpu reset */ - down_read(&adev->reset_sem); + down_read(&adev->reset_domain->sem); buf_size = num * RAS_TABLE_RECORD_SIZE; - res = amdgpu_eeprom_read(&adev->pm.smu_i2c, + res = amdgpu_eeprom_read(adev->pm.ras_eeprom_i2c_bus, control->i2c_address + RAS_INDEX_TO_OFFSET(control, fri), buf, buf_size); - up_read(&adev->reset_sem); + up_read(&adev->reset_domain->sem); if (res < 0) { DRM_ERROR("Reading %d EEPROM table records error:%d", num, res); @@ -681,6 +709,7 @@ int amdgpu_ras_eeprom_read(struct amdgpu_ras_eeprom_control *control, const u32 num) { struct amdgpu_device *adev = to_amdgpu_device(control); + struct amdgpu_ras *con = amdgpu_ras_get_context(adev); int i, res; u8 *buf, *pp; u32 g0, g1; @@ -748,8 +777,15 @@ int amdgpu_ras_eeprom_read(struct amdgpu_ras_eeprom_control *control, /* Read up everything? Then transform. */ pp = buf; - for (i = 0; i < num; i++, pp += RAS_TABLE_RECORD_SIZE) + for (i = 0; i < num; i++, pp += RAS_TABLE_RECORD_SIZE) { __decode_table_record_from_buf(control, &record[i], pp); + + /* update bad channel bitmap */ + if (!(control->bad_channel_bitmap & (1 << record[i].mem_channel))) { + control->bad_channel_bitmap |= 1 << record[i].mem_channel; + con->update_channel_flag = true; + } + } Out: kfree(buf); mutex_unlock(&control->ras_tbl_mutex); @@ -1009,7 +1045,7 @@ static int __verify_ras_table_checksum(struct amdgpu_ras_eeprom_control *control return -ENOMEM; } - res = amdgpu_eeprom_read(&adev->pm.smu_i2c, + res = amdgpu_eeprom_read(adev->pm.ras_eeprom_i2c_bus, control->i2c_address + control->ras_header_offset, buf, buf_size); @@ -1045,7 +1081,7 @@ int amdgpu_ras_eeprom_init(struct amdgpu_ras_eeprom_control *control, return 0; /* Verify i2c adapter is initialized */ - if (!adev->pm.smu_i2c.algo) + if (!adev->pm.ras_eeprom_i2c_bus || !adev->pm.ras_eeprom_i2c_bus->algo) return -ENOENT; if (!__get_eeprom_i2c_addr(adev, control)) @@ -1057,7 +1093,7 @@ int amdgpu_ras_eeprom_init(struct amdgpu_ras_eeprom_control *control, mutex_init(&control->ras_tbl_mutex); /* Read the table header from EEPROM address */ - res = amdgpu_eeprom_read(&adev->pm.smu_i2c, + res = amdgpu_eeprom_read(adev->pm.ras_eeprom_i2c_bus, control->i2c_address + control->ras_header_offset, buf, RAS_TABLE_HEADER_SIZE); if (res < RAS_TABLE_HEADER_SIZE) { diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras_eeprom.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras_eeprom.h index 6bb00578bfbb..54d9bfe0881d 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras_eeprom.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras_eeprom.h @@ -80,6 +80,10 @@ struct amdgpu_ras_eeprom_control { /* Protect table access via this mutex. */ struct mutex ras_tbl_mutex; + + /* Record channel info which occurred bad pages + */ + u32 bad_channel_bitmap; }; /* diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_res_cursor.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_res_cursor.h index acfa207cf970..6546552e596c 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_res_cursor.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_res_cursor.h @@ -30,12 +30,15 @@ #include <drm/ttm/ttm_resource.h> #include <drm/ttm/ttm_range_manager.h> +#include "amdgpu_vram_mgr.h" + /* state back for walking over vram_mgr and gtt_mgr allocations */ struct amdgpu_res_cursor { uint64_t start; uint64_t size; uint64_t remaining; - struct drm_mm_node *node; + void *node; + uint32_t mem_type; }; /** @@ -52,27 +55,63 @@ static inline void amdgpu_res_first(struct ttm_resource *res, uint64_t start, uint64_t size, struct amdgpu_res_cursor *cur) { + struct drm_buddy_block *block; + struct list_head *head, *next; struct drm_mm_node *node; - if (!res || res->mem_type == TTM_PL_SYSTEM) { - cur->start = start; - cur->size = size; - cur->remaining = size; - cur->node = NULL; - WARN_ON(res && start + size > res->num_pages << PAGE_SHIFT); - return; - } + if (!res) + goto fallback; BUG_ON(start + size > res->num_pages << PAGE_SHIFT); - node = to_ttm_range_mgr_node(res)->mm_nodes; - while (start >= node->size << PAGE_SHIFT) - start -= node++->size << PAGE_SHIFT; + cur->mem_type = res->mem_type; + + switch (cur->mem_type) { + case TTM_PL_VRAM: + head = &to_amdgpu_vram_mgr_resource(res)->blocks; + + block = list_first_entry_or_null(head, + struct drm_buddy_block, + link); + if (!block) + goto fallback; + + while (start >= amdgpu_vram_mgr_block_size(block)) { + start -= amdgpu_vram_mgr_block_size(block); + + next = block->link.next; + if (next != head) + block = list_entry(next, struct drm_buddy_block, link); + } + + cur->start = amdgpu_vram_mgr_block_start(block) + start; + cur->size = min(amdgpu_vram_mgr_block_size(block) - start, size); + cur->remaining = size; + cur->node = block; + break; + case TTM_PL_TT: + node = to_ttm_range_mgr_node(res)->mm_nodes; + while (start >= node->size << PAGE_SHIFT) + start -= node++->size << PAGE_SHIFT; + + cur->start = (node->start << PAGE_SHIFT) + start; + cur->size = min((node->size << PAGE_SHIFT) - start, size); + cur->remaining = size; + cur->node = node; + break; + default: + goto fallback; + } - cur->start = (node->start << PAGE_SHIFT) + start; - cur->size = min((node->size << PAGE_SHIFT) - start, size); + return; + +fallback: + cur->start = start; + cur->size = size; cur->remaining = size; - cur->node = node; + cur->node = NULL; + WARN_ON(res && start + size > res->num_pages << PAGE_SHIFT); + return; } /** @@ -85,7 +124,9 @@ static inline void amdgpu_res_first(struct ttm_resource *res, */ static inline void amdgpu_res_next(struct amdgpu_res_cursor *cur, uint64_t size) { - struct drm_mm_node *node = cur->node; + struct drm_buddy_block *block; + struct drm_mm_node *node; + struct list_head *next; BUG_ON(size > cur->remaining); @@ -99,9 +140,27 @@ static inline void amdgpu_res_next(struct amdgpu_res_cursor *cur, uint64_t size) return; } - cur->node = ++node; - cur->start = node->start << PAGE_SHIFT; - cur->size = min(node->size << PAGE_SHIFT, cur->remaining); + switch (cur->mem_type) { + case TTM_PL_VRAM: + block = cur->node; + + next = block->link.next; + block = list_entry(next, struct drm_buddy_block, link); + + cur->node = block; + cur->start = amdgpu_vram_mgr_block_start(block); + cur->size = min(amdgpu_vram_mgr_block_size(block), cur->remaining); + break; + case TTM_PL_TT: + node = cur->node; + + cur->node = ++node; + cur->start = node->start << PAGE_SHIFT; + cur->size = min(node->size << PAGE_SHIFT, cur->remaining); + break; + default: + return; + } } #endif diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_reset.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_reset.c index 02afd4115675..f778466bb9db 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_reset.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_reset.c @@ -23,6 +23,7 @@ #include "amdgpu_reset.h" #include "aldebaran.h" +#include "sienna_cichlid.h" int amdgpu_reset_add_handler(struct amdgpu_reset_control *reset_ctl, struct amdgpu_reset_handler *handler) @@ -36,10 +37,13 @@ int amdgpu_reset_init(struct amdgpu_device *adev) { int ret = 0; - switch (adev->asic_type) { - case CHIP_ALDEBARAN: + switch (adev->ip_versions[MP1_HWIP][0]) { + case IP_VERSION(13, 0, 2): ret = aldebaran_reset_init(adev); break; + case IP_VERSION(11, 0, 7): + ret = sienna_cichlid_reset_init(adev); + break; default: break; } @@ -51,10 +55,13 @@ int amdgpu_reset_fini(struct amdgpu_device *adev) { int ret = 0; - switch (adev->asic_type) { - case CHIP_ALDEBARAN: + switch (adev->ip_versions[MP1_HWIP][0]) { + case IP_VERSION(13, 0, 2): ret = aldebaran_reset_fini(adev); break; + case IP_VERSION(11, 0, 7): + ret = sienna_cichlid_reset_fini(adev); + break; default: break; } @@ -96,3 +103,60 @@ int amdgpu_reset_perform_reset(struct amdgpu_device *adev, return reset_handler->restore_hwcontext(adev->reset_cntl, reset_context); } + + +void amdgpu_reset_destroy_reset_domain(struct kref *ref) +{ + struct amdgpu_reset_domain *reset_domain = container_of(ref, + struct amdgpu_reset_domain, + refcount); + if (reset_domain->wq) + destroy_workqueue(reset_domain->wq); + + kvfree(reset_domain); +} + +struct amdgpu_reset_domain *amdgpu_reset_create_reset_domain(enum amdgpu_reset_domain_type type, + char *wq_name) +{ + struct amdgpu_reset_domain *reset_domain; + + reset_domain = kvzalloc(sizeof(struct amdgpu_reset_domain), GFP_KERNEL); + if (!reset_domain) { + DRM_ERROR("Failed to allocate amdgpu_reset_domain!"); + return NULL; + } + + reset_domain->type = type; + kref_init(&reset_domain->refcount); + + reset_domain->wq = create_singlethread_workqueue(wq_name); + if (!reset_domain->wq) { + DRM_ERROR("Failed to allocate wq for amdgpu_reset_domain!"); + amdgpu_reset_put_reset_domain(reset_domain); + return NULL; + + } + + atomic_set(&reset_domain->in_gpu_reset, 0); + atomic_set(&reset_domain->reset_res, 0); + init_rwsem(&reset_domain->sem); + + return reset_domain; +} + +void amdgpu_device_lock_reset_domain(struct amdgpu_reset_domain *reset_domain) +{ + atomic_set(&reset_domain->in_gpu_reset, 1); + down_write(&reset_domain->sem); +} + + +void amdgpu_device_unlock_reset_domain(struct amdgpu_reset_domain *reset_domain) +{ + atomic_set(&reset_domain->in_gpu_reset, 0); + up_write(&reset_domain->sem); +} + + + diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_reset.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_reset.h index e00d38d9160a..f4a501ff87d9 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_reset.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_reset.h @@ -30,6 +30,7 @@ enum AMDGPU_RESET_FLAGS { AMDGPU_NEED_FULL_RESET = 0, AMDGPU_SKIP_HW_RESET = 1, + AMDGPU_RESET_FOR_DEVICE_REMOVE = 2, }; struct amdgpu_reset_context { @@ -37,6 +38,7 @@ struct amdgpu_reset_context { struct amdgpu_device *reset_req_dev; struct amdgpu_job *job; struct amdgpu_hive_info *hive; + struct list_head *reset_device_list; unsigned long flags; }; @@ -70,6 +72,22 @@ struct amdgpu_reset_control { void (*async_reset)(struct work_struct *work); }; + +enum amdgpu_reset_domain_type { + SINGLE_DEVICE, + XGMI_HIVE +}; + +struct amdgpu_reset_domain { + struct kref refcount; + struct workqueue_struct *wq; + enum amdgpu_reset_domain_type type; + struct rw_semaphore sem; + atomic_t in_gpu_reset; + atomic_t reset_res; +}; + + int amdgpu_reset_init(struct amdgpu_device *adev); int amdgpu_reset_fini(struct amdgpu_device *adev); @@ -82,4 +100,30 @@ int amdgpu_reset_perform_reset(struct amdgpu_device *adev, int amdgpu_reset_add_handler(struct amdgpu_reset_control *reset_ctl, struct amdgpu_reset_handler *handler); +struct amdgpu_reset_domain *amdgpu_reset_create_reset_domain(enum amdgpu_reset_domain_type type, + char *wq_name); + +void amdgpu_reset_destroy_reset_domain(struct kref *ref); + +static inline bool amdgpu_reset_get_reset_domain(struct amdgpu_reset_domain *domain) +{ + return kref_get_unless_zero(&domain->refcount) != 0; +} + +static inline void amdgpu_reset_put_reset_domain(struct amdgpu_reset_domain *domain) +{ + if (domain) + kref_put(&domain->refcount, amdgpu_reset_destroy_reset_domain); +} + +static inline bool amdgpu_reset_domain_schedule(struct amdgpu_reset_domain *domain, + struct work_struct *work) +{ + return queue_work(domain->wq, work); +} + +void amdgpu_device_lock_reset_domain(struct amdgpu_reset_domain *reset_domain); + +void amdgpu_device_unlock_reset_domain(struct amdgpu_reset_domain *reset_domain); + #endif diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c index ab2351ba9574..d3558c34d406 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c @@ -149,6 +149,16 @@ void amdgpu_ring_undo(struct amdgpu_ring *ring) ring->funcs->end_use(ring); } +#define amdgpu_ring_get_gpu_addr(ring, offset) \ + (ring->is_mes_queue ? \ + (ring->mes_ctx->meta_data_gpu_addr + offset) : \ + (ring->adev->wb.gpu_addr + offset * 4)) + +#define amdgpu_ring_get_cpu_addr(ring, offset) \ + (ring->is_mes_queue ? \ + (void *)((uint8_t *)(ring->mes_ctx->meta_data_ptr) + offset) : \ + (&ring->adev->wb.wb[offset])) + /** * amdgpu_ring_init - init driver ring struct. * @@ -189,49 +199,88 @@ int amdgpu_ring_init(struct amdgpu_device *adev, struct amdgpu_ring *ring, return -EINVAL; ring->adev = adev; - ring->idx = adev->num_rings++; - adev->rings[ring->idx] = ring; - r = amdgpu_fence_driver_init_ring(ring, sched_hw_submission, - sched_score); + ring->num_hw_submission = sched_hw_submission; + ring->sched_score = sched_score; + ring->vmid_wait = dma_fence_get_stub(); + + if (!ring->is_mes_queue) { + ring->idx = adev->num_rings++; + adev->rings[ring->idx] = ring; + } + + r = amdgpu_fence_driver_init_ring(ring); if (r) return r; } - r = amdgpu_device_wb_get(adev, &ring->rptr_offs); - if (r) { - dev_err(adev->dev, "(%d) ring rptr_offs wb alloc failed\n", r); - return r; - } + if (ring->is_mes_queue) { + ring->rptr_offs = amdgpu_mes_ctx_get_offs(ring, + AMDGPU_MES_CTX_RPTR_OFFS); + ring->wptr_offs = amdgpu_mes_ctx_get_offs(ring, + AMDGPU_MES_CTX_WPTR_OFFS); + ring->fence_offs = amdgpu_mes_ctx_get_offs(ring, + AMDGPU_MES_CTX_FENCE_OFFS); + ring->trail_fence_offs = amdgpu_mes_ctx_get_offs(ring, + AMDGPU_MES_CTX_TRAIL_FENCE_OFFS); + ring->cond_exe_offs = amdgpu_mes_ctx_get_offs(ring, + AMDGPU_MES_CTX_COND_EXE_OFFS); + } else { + r = amdgpu_device_wb_get(adev, &ring->rptr_offs); + if (r) { + dev_err(adev->dev, "(%d) ring rptr_offs wb alloc failed\n", r); + return r; + } - r = amdgpu_device_wb_get(adev, &ring->wptr_offs); - if (r) { - dev_err(adev->dev, "(%d) ring wptr_offs wb alloc failed\n", r); - return r; - } + r = amdgpu_device_wb_get(adev, &ring->wptr_offs); + if (r) { + dev_err(adev->dev, "(%d) ring wptr_offs wb alloc failed\n", r); + return r; + } - r = amdgpu_device_wb_get(adev, &ring->fence_offs); - if (r) { - dev_err(adev->dev, "(%d) ring fence_offs wb alloc failed\n", r); - return r; - } + r = amdgpu_device_wb_get(adev, &ring->fence_offs); + if (r) { + dev_err(adev->dev, "(%d) ring fence_offs wb alloc failed\n", r); + return r; + } - r = amdgpu_device_wb_get(adev, &ring->trail_fence_offs); - if (r) { - dev_err(adev->dev, - "(%d) ring trail_fence_offs wb alloc failed\n", r); - return r; + r = amdgpu_device_wb_get(adev, &ring->trail_fence_offs); + if (r) { + dev_err(adev->dev, "(%d) ring trail_fence_offs wb alloc failed\n", r); + return r; + } + + r = amdgpu_device_wb_get(adev, &ring->cond_exe_offs); + if (r) { + dev_err(adev->dev, "(%d) ring cond_exec_polling wb alloc failed\n", r); + return r; + } } + + ring->fence_gpu_addr = + amdgpu_ring_get_gpu_addr(ring, ring->fence_offs); + ring->fence_cpu_addr = + amdgpu_ring_get_cpu_addr(ring, ring->fence_offs); + + ring->rptr_gpu_addr = + amdgpu_ring_get_gpu_addr(ring, ring->rptr_offs); + ring->rptr_cpu_addr = + amdgpu_ring_get_cpu_addr(ring, ring->rptr_offs); + + ring->wptr_gpu_addr = + amdgpu_ring_get_gpu_addr(ring, ring->wptr_offs); + ring->wptr_cpu_addr = + amdgpu_ring_get_cpu_addr(ring, ring->wptr_offs); + ring->trail_fence_gpu_addr = - adev->wb.gpu_addr + (ring->trail_fence_offs * 4); - ring->trail_fence_cpu_addr = &adev->wb.wb[ring->trail_fence_offs]; + amdgpu_ring_get_gpu_addr(ring, ring->trail_fence_offs); + ring->trail_fence_cpu_addr = + amdgpu_ring_get_cpu_addr(ring, ring->trail_fence_offs); + + ring->cond_exe_gpu_addr = + amdgpu_ring_get_gpu_addr(ring, ring->cond_exe_offs); + ring->cond_exe_cpu_addr = + amdgpu_ring_get_cpu_addr(ring, ring->cond_exe_offs); - r = amdgpu_device_wb_get(adev, &ring->cond_exe_offs); - if (r) { - dev_err(adev->dev, "(%d) ring cond_exec_polling wb alloc failed\n", r); - return r; - } - ring->cond_exe_gpu_addr = adev->wb.gpu_addr + (ring->cond_exe_offs * 4); - ring->cond_exe_cpu_addr = &adev->wb.wb[ring->cond_exe_offs]; /* always set cond_exec_polling to CONTINUE */ *ring->cond_exe_cpu_addr = 1; @@ -246,8 +295,20 @@ int amdgpu_ring_init(struct amdgpu_device *adev, struct amdgpu_ring *ring, ring->buf_mask = (ring->ring_size / 4) - 1; ring->ptr_mask = ring->funcs->support_64bit_ptrs ? 0xffffffffffffffff : ring->buf_mask; + /* Allocate ring buffer */ - if (ring->ring_obj == NULL) { + if (ring->is_mes_queue) { + int offset = 0; + + BUG_ON(ring->ring_size > PAGE_SIZE*4); + + offset = amdgpu_mes_ctx_get_offs(ring, + AMDGPU_MES_CTX_RING_OFFS); + ring->gpu_addr = amdgpu_mes_ctx_get_offs_gpu_addr(ring, offset); + ring->ring = amdgpu_mes_ctx_get_offs_cpu_addr(ring, offset); + amdgpu_ring_clear_ring(ring); + + } else if (ring->ring_obj == NULL) { r = amdgpu_bo_create_kernel(adev, ring->ring_size + ring->funcs->extra_dw, PAGE_SIZE, AMDGPU_GEM_DOMAIN_GTT, &ring->ring_obj, @@ -284,26 +345,30 @@ void amdgpu_ring_fini(struct amdgpu_ring *ring) { /* Not to finish a ring which is not initialized */ - if (!(ring->adev) || !(ring->adev->rings[ring->idx])) + if (!(ring->adev) || + (!ring->is_mes_queue && !(ring->adev->rings[ring->idx]))) return; ring->sched.ready = false; - amdgpu_device_wb_free(ring->adev, ring->rptr_offs); - amdgpu_device_wb_free(ring->adev, ring->wptr_offs); + if (!ring->is_mes_queue) { + amdgpu_device_wb_free(ring->adev, ring->rptr_offs); + amdgpu_device_wb_free(ring->adev, ring->wptr_offs); - amdgpu_device_wb_free(ring->adev, ring->cond_exe_offs); - amdgpu_device_wb_free(ring->adev, ring->fence_offs); + amdgpu_device_wb_free(ring->adev, ring->cond_exe_offs); + amdgpu_device_wb_free(ring->adev, ring->fence_offs); - amdgpu_bo_free_kernel(&ring->ring_obj, - &ring->gpu_addr, - (void **)&ring->ring); + amdgpu_bo_free_kernel(&ring->ring_obj, + &ring->gpu_addr, + (void **)&ring->ring); + } dma_fence_put(ring->vmid_wait); ring->vmid_wait = NULL; ring->me = 0; - ring->adev->rings[ring->idx] = NULL; + if (!ring->is_mes_queue) + ring->adev->rings[ring->idx] = NULL; } /** @@ -456,3 +521,51 @@ int amdgpu_ring_test_helper(struct amdgpu_ring *ring) ring->sched.ready = !r; return r; } + +static void amdgpu_ring_to_mqd_prop(struct amdgpu_ring *ring, + struct amdgpu_mqd_prop *prop) +{ + struct amdgpu_device *adev = ring->adev; + + memset(prop, 0, sizeof(*prop)); + + prop->mqd_gpu_addr = ring->mqd_gpu_addr; + prop->hqd_base_gpu_addr = ring->gpu_addr; + prop->rptr_gpu_addr = ring->rptr_gpu_addr; + prop->wptr_gpu_addr = ring->wptr_gpu_addr; + prop->queue_size = ring->ring_size; + prop->eop_gpu_addr = ring->eop_gpu_addr; + prop->use_doorbell = ring->use_doorbell; + prop->doorbell_index = ring->doorbell_index; + + /* map_queues packet doesn't need activate the queue, + * so only kiq need set this field. + */ + prop->hqd_active = ring->funcs->type == AMDGPU_RING_TYPE_KIQ; + + if ((ring->funcs->type == AMDGPU_RING_TYPE_COMPUTE && + amdgpu_gfx_is_high_priority_compute_queue(adev, ring)) || + (ring->funcs->type == AMDGPU_RING_TYPE_GFX && + amdgpu_gfx_is_high_priority_graphics_queue(adev, ring))) { + prop->hqd_pipe_priority = AMDGPU_GFX_PIPE_PRIO_HIGH; + prop->hqd_queue_priority = AMDGPU_GFX_QUEUE_PRIORITY_MAXIMUM; + } +} + +int amdgpu_ring_init_mqd(struct amdgpu_ring *ring) +{ + struct amdgpu_device *adev = ring->adev; + struct amdgpu_mqd *mqd_mgr; + struct amdgpu_mqd_prop prop; + + amdgpu_ring_to_mqd_prop(ring, &prop); + + ring->wptr = 0; + + if (ring->funcs->type == AMDGPU_RING_TYPE_KIQ) + mqd_mgr = &adev->mqds[AMDGPU_HW_IP_COMPUTE]; + else + mqd_mgr = &adev->mqds[ring->funcs->type]; + + return mqd_mgr->init_mqd(adev, ring->mqd_ptr, &prop); +} diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h index fae7d185ad0d..82c178a9033a 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h @@ -28,6 +28,13 @@ #include <drm/gpu_scheduler.h> #include <drm/drm_print.h> +struct amdgpu_device; +struct amdgpu_ring; +struct amdgpu_ib; +struct amdgpu_cs_parser; +struct amdgpu_job; +struct amdgpu_vm; + /* max number of rings */ #define AMDGPU_MAX_RINGS 28 #define AMDGPU_MAX_HWIP_RINGS 8 @@ -82,11 +89,13 @@ enum amdgpu_ib_pool_type { AMDGPU_IB_POOL_MAX }; -struct amdgpu_device; -struct amdgpu_ring; -struct amdgpu_ib; -struct amdgpu_cs_parser; -struct amdgpu_job; +struct amdgpu_ib { + struct amdgpu_sa_bo *sa_bo; + uint32_t length_dw; + uint64_t gpu_addr; + uint32_t *ptr; + uint32_t flags; +}; struct amdgpu_sched { u32 num_scheds; @@ -111,12 +120,12 @@ struct amdgpu_fence_driver { struct dma_fence **fences; }; +extern const struct drm_sched_backend_ops amdgpu_sched_ops; + void amdgpu_fence_driver_clear_job_fences(struct amdgpu_ring *ring); void amdgpu_fence_driver_force_completion(struct amdgpu_ring *ring); -int amdgpu_fence_driver_init_ring(struct amdgpu_ring *ring, - unsigned num_hw_submission, - atomic_t *sched_score); +int amdgpu_fence_driver_init_ring(struct amdgpu_ring *ring); int amdgpu_fence_driver_start_ring(struct amdgpu_ring *ring, struct amdgpu_irq_src *irq_src, unsigned irq_type); @@ -134,6 +143,7 @@ signed long amdgpu_fence_wait_polling(struct amdgpu_ring *ring, uint32_t wait_seq, signed long timeout); unsigned amdgpu_fence_count_emitted(struct amdgpu_ring *ring); +void amdgpu_fence_driver_isr_toggle(struct amdgpu_device *adev, bool stop); /* * Rings. @@ -146,6 +156,7 @@ struct amdgpu_ring_funcs { u32 nop; bool support_64bit_ptrs; bool no_user_fence; + bool secure_submission_supported; unsigned vmhub; unsigned extra_dw; @@ -154,8 +165,12 @@ struct amdgpu_ring_funcs { u64 (*get_wptr)(struct amdgpu_ring *ring); void (*set_wptr)(struct amdgpu_ring *ring); /* validating and patching of IBs */ - int (*parse_cs)(struct amdgpu_cs_parser *p, uint32_t ib_idx); - int (*patch_cs_in_place)(struct amdgpu_cs_parser *p, uint32_t ib_idx); + int (*parse_cs)(struct amdgpu_cs_parser *p, + struct amdgpu_job *job, + struct amdgpu_ib *ib); + int (*patch_cs_in_place)(struct amdgpu_cs_parser *p, + struct amdgpu_job *job, + struct amdgpu_ib *ib); /* constants to calculate how many DW are needed for an emit */ unsigned emit_frame_size; unsigned emit_ib_size; @@ -216,6 +231,8 @@ struct amdgpu_ring { struct amdgpu_bo *ring_obj; volatile uint32_t *ring; unsigned rptr_offs; + u64 rptr_gpu_addr; + volatile u32 *rptr_cpu_addr; u64 wptr; u64 wptr_old; unsigned ring_size; @@ -236,7 +253,11 @@ struct amdgpu_ring { bool use_doorbell; bool use_pollmem; unsigned wptr_offs; + u64 wptr_gpu_addr; + volatile u32 *wptr_cpu_addr; unsigned fence_offs; + u64 fence_gpu_addr; + volatile u32 *fence_cpu_addr; uint64_t current_ctx; char name[16]; u32 trail_seq; @@ -251,10 +272,17 @@ struct amdgpu_ring { bool has_compute_vm_bug; bool no_scheduler; int hw_prio; + unsigned num_hw_submission; + atomic_t *sched_score; + + /* used for mes */ + bool is_mes_queue; + uint32_t hw_queue_id; + struct amdgpu_mes_ctx_data *mes_ctx; }; -#define amdgpu_ring_parse_cs(r, p, ib) ((r)->funcs->parse_cs((p), (ib))) -#define amdgpu_ring_patch_cs_in_place(r, p, ib) ((r)->funcs->patch_cs_in_place((p), (ib))) +#define amdgpu_ring_parse_cs(r, p, job, ib) ((r)->funcs->parse_cs((p), (job), (ib))) +#define amdgpu_ring_patch_cs_in_place(r, p, job, ib) ((r)->funcs->patch_cs_in_place((p), (job), (ib))) #define amdgpu_ring_test_ring(r) (r)->funcs->test_ring((r)) #define amdgpu_ring_test_ib(r, t) (r)->funcs->test_ib((r), (t)) #define amdgpu_ring_get_rptr(r) (r)->funcs->get_rptr((r)) @@ -284,8 +312,8 @@ void amdgpu_ring_generic_pad_ib(struct amdgpu_ring *ring, struct amdgpu_ib *ib); void amdgpu_ring_commit(struct amdgpu_ring *ring); void amdgpu_ring_undo(struct amdgpu_ring *ring); int amdgpu_ring_init(struct amdgpu_device *adev, struct amdgpu_ring *ring, - unsigned int ring_size, struct amdgpu_irq_src *irq_src, - unsigned int irq_type, unsigned int prio, + unsigned int max_dw, struct amdgpu_irq_src *irq_src, + unsigned int irq_type, unsigned int hw_prio, atomic_t *sched_score); void amdgpu_ring_fini(struct amdgpu_ring *ring); void amdgpu_ring_emit_reg_write_reg_wait_helper(struct amdgpu_ring *ring, @@ -348,8 +376,44 @@ static inline void amdgpu_ring_write_multiple(struct amdgpu_ring *ring, ring->count_dw -= count_dw; } +#define amdgpu_mes_ctx_get_offs_gpu_addr(ring, offset) \ + (ring->is_mes_queue && ring->mes_ctx ? \ + (ring->mes_ctx->meta_data_gpu_addr + offset) : 0) + +#define amdgpu_mes_ctx_get_offs_cpu_addr(ring, offset) \ + (ring->is_mes_queue && ring->mes_ctx ? \ + (void *)((uint8_t *)(ring->mes_ctx->meta_data_ptr) + offset) : \ + NULL) + int amdgpu_ring_test_helper(struct amdgpu_ring *ring); void amdgpu_debugfs_ring_init(struct amdgpu_device *adev, struct amdgpu_ring *ring); + +int amdgpu_ring_init_mqd(struct amdgpu_ring *ring); + +static inline u32 amdgpu_ib_get_value(struct amdgpu_ib *ib, int idx) +{ + return ib->ptr[idx]; +} + +static inline void amdgpu_ib_set_value(struct amdgpu_ib *ib, int idx, + uint32_t value) +{ + ib->ptr[idx] = value; +} + +int amdgpu_ib_get(struct amdgpu_device *adev, struct amdgpu_vm *vm, + unsigned size, + enum amdgpu_ib_pool_type pool, + struct amdgpu_ib *ib); +void amdgpu_ib_free(struct amdgpu_device *adev, struct amdgpu_ib *ib, + struct dma_fence *f); +int amdgpu_ib_schedule(struct amdgpu_ring *ring, unsigned num_ibs, + struct amdgpu_ib *ibs, struct amdgpu_job *job, + struct dma_fence **f); +int amdgpu_ib_pool_init(struct amdgpu_device *adev); +void amdgpu_ib_pool_fini(struct amdgpu_device *adev); +int amdgpu_ib_ring_tests(struct amdgpu_device *adev); + #endif diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_rlc.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_rlc.c index 6373bfb47d55..012b72d00e04 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_rlc.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_rlc.c @@ -272,3 +272,275 @@ void amdgpu_gfx_rlc_fini(struct amdgpu_device *adev) &adev->gfx.rlc.cp_table_gpu_addr, (void **)&adev->gfx.rlc.cp_table_ptr); } + +static int amdgpu_gfx_rlc_init_microcode_v2_0(struct amdgpu_device *adev) +{ + const struct common_firmware_header *common_hdr; + const struct rlc_firmware_header_v2_0 *rlc_hdr; + struct amdgpu_firmware_info *info; + unsigned int *tmp; + unsigned int i; + + rlc_hdr = (const struct rlc_firmware_header_v2_0 *)adev->gfx.rlc_fw->data; + + adev->gfx.rlc_fw_version = le32_to_cpu(rlc_hdr->header.ucode_version); + adev->gfx.rlc_feature_version = le32_to_cpu(rlc_hdr->ucode_feature_version); + adev->gfx.rlc.save_and_restore_offset = + le32_to_cpu(rlc_hdr->save_and_restore_offset); + adev->gfx.rlc.clear_state_descriptor_offset = + le32_to_cpu(rlc_hdr->clear_state_descriptor_offset); + adev->gfx.rlc.avail_scratch_ram_locations = + le32_to_cpu(rlc_hdr->avail_scratch_ram_locations); + adev->gfx.rlc.reg_restore_list_size = + le32_to_cpu(rlc_hdr->reg_restore_list_size); + adev->gfx.rlc.reg_list_format_start = + le32_to_cpu(rlc_hdr->reg_list_format_start); + adev->gfx.rlc.reg_list_format_separate_start = + le32_to_cpu(rlc_hdr->reg_list_format_separate_start); + adev->gfx.rlc.starting_offsets_start = + le32_to_cpu(rlc_hdr->starting_offsets_start); + adev->gfx.rlc.reg_list_format_size_bytes = + le32_to_cpu(rlc_hdr->reg_list_format_size_bytes); + adev->gfx.rlc.reg_list_size_bytes = + le32_to_cpu(rlc_hdr->reg_list_size_bytes); + adev->gfx.rlc.register_list_format = + kmalloc(adev->gfx.rlc.reg_list_format_size_bytes + + adev->gfx.rlc.reg_list_size_bytes, GFP_KERNEL); + if (!adev->gfx.rlc.register_list_format) { + dev_err(adev->dev, "failed to allocate memory for rlc register_list_format\n"); + return -ENOMEM; + } + + tmp = (unsigned int *)((uintptr_t)rlc_hdr + + le32_to_cpu(rlc_hdr->reg_list_format_array_offset_bytes)); + for (i = 0 ; i < (rlc_hdr->reg_list_format_size_bytes >> 2); i++) + adev->gfx.rlc.register_list_format[i] = le32_to_cpu(tmp[i]); + + adev->gfx.rlc.register_restore = adev->gfx.rlc.register_list_format + i; + + tmp = (unsigned int *)((uintptr_t)rlc_hdr + + le32_to_cpu(rlc_hdr->reg_list_array_offset_bytes)); + for (i = 0 ; i < (rlc_hdr->reg_list_size_bytes >> 2); i++) + adev->gfx.rlc.register_restore[i] = le32_to_cpu(tmp[i]); + + if (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) { + info = &adev->firmware.ucode[AMDGPU_UCODE_ID_RLC_G]; + info->ucode_id = AMDGPU_UCODE_ID_RLC_G; + info->fw = adev->gfx.rlc_fw; + if (info->fw) { + common_hdr = (const struct common_firmware_header *)info->fw->data; + adev->firmware.fw_size += + ALIGN(le32_to_cpu(common_hdr->ucode_size_bytes), PAGE_SIZE); + } + } + + return 0; +} + +static void amdgpu_gfx_rlc_init_microcode_v2_1(struct amdgpu_device *adev) +{ + const struct rlc_firmware_header_v2_1 *rlc_hdr; + struct amdgpu_firmware_info *info; + + rlc_hdr = (const struct rlc_firmware_header_v2_1 *)adev->gfx.rlc_fw->data; + adev->gfx.rlc_srlc_fw_version = le32_to_cpu(rlc_hdr->save_restore_list_cntl_ucode_ver); + adev->gfx.rlc_srlc_feature_version = le32_to_cpu(rlc_hdr->save_restore_list_cntl_feature_ver); + adev->gfx.rlc.save_restore_list_cntl_size_bytes = le32_to_cpu(rlc_hdr->save_restore_list_cntl_size_bytes); + adev->gfx.rlc.save_restore_list_cntl = (u8 *)rlc_hdr + le32_to_cpu(rlc_hdr->save_restore_list_cntl_offset_bytes); + adev->gfx.rlc_srlg_fw_version = le32_to_cpu(rlc_hdr->save_restore_list_gpm_ucode_ver); + adev->gfx.rlc_srlg_feature_version = le32_to_cpu(rlc_hdr->save_restore_list_gpm_feature_ver); + adev->gfx.rlc.save_restore_list_gpm_size_bytes = le32_to_cpu(rlc_hdr->save_restore_list_gpm_size_bytes); + adev->gfx.rlc.save_restore_list_gpm = (u8 *)rlc_hdr + le32_to_cpu(rlc_hdr->save_restore_list_gpm_offset_bytes); + adev->gfx.rlc_srls_fw_version = le32_to_cpu(rlc_hdr->save_restore_list_srm_ucode_ver); + adev->gfx.rlc_srls_feature_version = le32_to_cpu(rlc_hdr->save_restore_list_srm_feature_ver); + adev->gfx.rlc.save_restore_list_srm_size_bytes = le32_to_cpu(rlc_hdr->save_restore_list_srm_size_bytes); + adev->gfx.rlc.save_restore_list_srm = (u8 *)rlc_hdr + le32_to_cpu(rlc_hdr->save_restore_list_srm_offset_bytes); + adev->gfx.rlc.reg_list_format_direct_reg_list_length = + le32_to_cpu(rlc_hdr->reg_list_format_direct_reg_list_length); + + if (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) { + if (adev->gfx.rlc.save_restore_list_cntl_size_bytes) { + info = &adev->firmware.ucode[AMDGPU_UCODE_ID_RLC_RESTORE_LIST_CNTL]; + info->ucode_id = AMDGPU_UCODE_ID_RLC_RESTORE_LIST_CNTL; + info->fw = adev->gfx.rlc_fw; + adev->firmware.fw_size += + ALIGN(adev->gfx.rlc.save_restore_list_cntl_size_bytes, PAGE_SIZE); + } + + if (adev->gfx.rlc.save_restore_list_gpm_size_bytes) { + info = &adev->firmware.ucode[AMDGPU_UCODE_ID_RLC_RESTORE_LIST_GPM_MEM]; + info->ucode_id = AMDGPU_UCODE_ID_RLC_RESTORE_LIST_GPM_MEM; + info->fw = adev->gfx.rlc_fw; + adev->firmware.fw_size += + ALIGN(adev->gfx.rlc.save_restore_list_gpm_size_bytes, PAGE_SIZE); + } + + if (adev->gfx.rlc.save_restore_list_srm_size_bytes) { + info = &adev->firmware.ucode[AMDGPU_UCODE_ID_RLC_RESTORE_LIST_SRM_MEM]; + info->ucode_id = AMDGPU_UCODE_ID_RLC_RESTORE_LIST_SRM_MEM; + info->fw = adev->gfx.rlc_fw; + adev->firmware.fw_size += + ALIGN(adev->gfx.rlc.save_restore_list_srm_size_bytes, PAGE_SIZE); + } + } +} + +static void amdgpu_gfx_rlc_init_microcode_v2_2(struct amdgpu_device *adev) +{ + const struct rlc_firmware_header_v2_2 *rlc_hdr; + struct amdgpu_firmware_info *info; + + rlc_hdr = (const struct rlc_firmware_header_v2_2 *)adev->gfx.rlc_fw->data; + adev->gfx.rlc.rlc_iram_ucode_size_bytes = le32_to_cpu(rlc_hdr->rlc_iram_ucode_size_bytes); + adev->gfx.rlc.rlc_iram_ucode = (u8 *)rlc_hdr + le32_to_cpu(rlc_hdr->rlc_iram_ucode_offset_bytes); + adev->gfx.rlc.rlc_dram_ucode_size_bytes = le32_to_cpu(rlc_hdr->rlc_dram_ucode_size_bytes); + adev->gfx.rlc.rlc_dram_ucode = (u8 *)rlc_hdr + le32_to_cpu(rlc_hdr->rlc_dram_ucode_offset_bytes); + + if (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) { + if (adev->gfx.rlc.rlc_iram_ucode_size_bytes) { + info = &adev->firmware.ucode[AMDGPU_UCODE_ID_RLC_IRAM]; + info->ucode_id = AMDGPU_UCODE_ID_RLC_IRAM; + info->fw = adev->gfx.rlc_fw; + adev->firmware.fw_size += + ALIGN(adev->gfx.rlc.rlc_iram_ucode_size_bytes, PAGE_SIZE); + } + + if (adev->gfx.rlc.rlc_dram_ucode_size_bytes) { + info = &adev->firmware.ucode[AMDGPU_UCODE_ID_RLC_DRAM]; + info->ucode_id = AMDGPU_UCODE_ID_RLC_DRAM; + info->fw = adev->gfx.rlc_fw; + adev->firmware.fw_size += + ALIGN(adev->gfx.rlc.rlc_dram_ucode_size_bytes, PAGE_SIZE); + } + } +} + +static void amdgpu_gfx_rlc_init_microcode_v2_3(struct amdgpu_device *adev) +{ + const struct rlc_firmware_header_v2_3 *rlc_hdr; + struct amdgpu_firmware_info *info; + + rlc_hdr = (const struct rlc_firmware_header_v2_3 *)adev->gfx.rlc_fw->data; + adev->gfx.rlcp_ucode_version = le32_to_cpu(rlc_hdr->rlcp_ucode_version); + adev->gfx.rlcp_ucode_feature_version = le32_to_cpu(rlc_hdr->rlcp_ucode_feature_version); + adev->gfx.rlc.rlcp_ucode_size_bytes = le32_to_cpu(rlc_hdr->rlcp_ucode_size_bytes); + adev->gfx.rlc.rlcp_ucode = (u8 *)rlc_hdr + le32_to_cpu(rlc_hdr->rlcp_ucode_offset_bytes); + + adev->gfx.rlcv_ucode_version = le32_to_cpu(rlc_hdr->rlcv_ucode_version); + adev->gfx.rlcv_ucode_feature_version = le32_to_cpu(rlc_hdr->rlcv_ucode_feature_version); + adev->gfx.rlc.rlcv_ucode_size_bytes = le32_to_cpu(rlc_hdr->rlcv_ucode_size_bytes); + adev->gfx.rlc.rlcv_ucode = (u8 *)rlc_hdr + le32_to_cpu(rlc_hdr->rlcv_ucode_offset_bytes); + + if (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) { + if (adev->gfx.rlc.rlcp_ucode_size_bytes) { + info = &adev->firmware.ucode[AMDGPU_UCODE_ID_RLC_P]; + info->ucode_id = AMDGPU_UCODE_ID_RLC_P; + info->fw = adev->gfx.rlc_fw; + adev->firmware.fw_size += + ALIGN(adev->gfx.rlc.rlcp_ucode_size_bytes, PAGE_SIZE); + } + + if (adev->gfx.rlc.rlcv_ucode_size_bytes) { + info = &adev->firmware.ucode[AMDGPU_UCODE_ID_RLC_V]; + info->ucode_id = AMDGPU_UCODE_ID_RLC_V; + info->fw = adev->gfx.rlc_fw; + adev->firmware.fw_size += + ALIGN(adev->gfx.rlc.rlcv_ucode_size_bytes, PAGE_SIZE); + } + } +} + +static void amdgpu_gfx_rlc_init_microcode_v2_4(struct amdgpu_device *adev) +{ + const struct rlc_firmware_header_v2_4 *rlc_hdr; + struct amdgpu_firmware_info *info; + + rlc_hdr = (const struct rlc_firmware_header_v2_4 *)adev->gfx.rlc_fw->data; + adev->gfx.rlc.global_tap_delays_ucode_size_bytes = le32_to_cpu(rlc_hdr->global_tap_delays_ucode_size_bytes); + adev->gfx.rlc.global_tap_delays_ucode = (u8 *)rlc_hdr + le32_to_cpu(rlc_hdr->global_tap_delays_ucode_offset_bytes); + adev->gfx.rlc.se0_tap_delays_ucode_size_bytes = le32_to_cpu(rlc_hdr->se0_tap_delays_ucode_size_bytes); + adev->gfx.rlc.se0_tap_delays_ucode = (u8 *)rlc_hdr + le32_to_cpu(rlc_hdr->se0_tap_delays_ucode_offset_bytes); + adev->gfx.rlc.se1_tap_delays_ucode_size_bytes = le32_to_cpu(rlc_hdr->se1_tap_delays_ucode_size_bytes); + adev->gfx.rlc.se1_tap_delays_ucode = (u8 *)rlc_hdr + le32_to_cpu(rlc_hdr->se1_tap_delays_ucode_offset_bytes); + adev->gfx.rlc.se2_tap_delays_ucode_size_bytes = le32_to_cpu(rlc_hdr->se2_tap_delays_ucode_size_bytes); + adev->gfx.rlc.se2_tap_delays_ucode = (u8 *)rlc_hdr + le32_to_cpu(rlc_hdr->se2_tap_delays_ucode_offset_bytes); + adev->gfx.rlc.se3_tap_delays_ucode_size_bytes = le32_to_cpu(rlc_hdr->se3_tap_delays_ucode_size_bytes); + adev->gfx.rlc.se3_tap_delays_ucode = (u8 *)rlc_hdr + le32_to_cpu(rlc_hdr->se3_tap_delays_ucode_offset_bytes); + + if (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) { + if (adev->gfx.rlc.global_tap_delays_ucode_size_bytes) { + info = &adev->firmware.ucode[AMDGPU_UCODE_ID_GLOBAL_TAP_DELAYS]; + info->ucode_id = AMDGPU_UCODE_ID_GLOBAL_TAP_DELAYS; + info->fw = adev->gfx.rlc_fw; + adev->firmware.fw_size += + ALIGN(adev->gfx.rlc.global_tap_delays_ucode_size_bytes, PAGE_SIZE); + } + + if (adev->gfx.rlc.se0_tap_delays_ucode_size_bytes) { + info = &adev->firmware.ucode[AMDGPU_UCODE_ID_SE0_TAP_DELAYS]; + info->ucode_id = AMDGPU_UCODE_ID_SE0_TAP_DELAYS; + info->fw = adev->gfx.rlc_fw; + adev->firmware.fw_size += + ALIGN(adev->gfx.rlc.se0_tap_delays_ucode_size_bytes, PAGE_SIZE); + } + + if (adev->gfx.rlc.se1_tap_delays_ucode_size_bytes) { + info = &adev->firmware.ucode[AMDGPU_UCODE_ID_SE1_TAP_DELAYS]; + info->ucode_id = AMDGPU_UCODE_ID_SE1_TAP_DELAYS; + info->fw = adev->gfx.rlc_fw; + adev->firmware.fw_size += + ALIGN(adev->gfx.rlc.se1_tap_delays_ucode_size_bytes, PAGE_SIZE); + } + + if (adev->gfx.rlc.se2_tap_delays_ucode_size_bytes) { + info = &adev->firmware.ucode[AMDGPU_UCODE_ID_SE2_TAP_DELAYS]; + info->ucode_id = AMDGPU_UCODE_ID_SE2_TAP_DELAYS; + info->fw = adev->gfx.rlc_fw; + adev->firmware.fw_size += + ALIGN(adev->gfx.rlc.se2_tap_delays_ucode_size_bytes, PAGE_SIZE); + } + + if (adev->gfx.rlc.se3_tap_delays_ucode_size_bytes) { + info = &adev->firmware.ucode[AMDGPU_UCODE_ID_SE3_TAP_DELAYS]; + info->ucode_id = AMDGPU_UCODE_ID_SE3_TAP_DELAYS; + info->fw = adev->gfx.rlc_fw; + adev->firmware.fw_size += + ALIGN(adev->gfx.rlc.se3_tap_delays_ucode_size_bytes, PAGE_SIZE); + } + } +} + +int amdgpu_gfx_rlc_init_microcode(struct amdgpu_device *adev, + uint16_t version_major, + uint16_t version_minor) +{ + int err; + + if (version_major < 2) { + /* only support rlc_hdr v2.x and onwards */ + dev_err(adev->dev, "unsupported rlc fw hdr\n"); + return -EINVAL; + } + + /* is_rlc_v2_1 is still used in APU code path */ + if (version_major == 2 && version_minor == 1) + adev->gfx.rlc.is_rlc_v2_1 = true; + + if (version_minor >= 0) { + err = amdgpu_gfx_rlc_init_microcode_v2_0(adev); + if (err) { + dev_err(adev->dev, "fail to init rlc v2_0 microcode\n"); + return err; + } + } + if (version_minor >= 1) + amdgpu_gfx_rlc_init_microcode_v2_1(adev); + if (version_minor >= 2) + amdgpu_gfx_rlc_init_microcode_v2_2(adev); + if (version_minor == 3) + amdgpu_gfx_rlc_init_microcode_v2_3(adev); + if (version_minor == 4) + amdgpu_gfx_rlc_init_microcode_v2_4(adev); + + return 0; +} diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_rlc.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_rlc.h index 00afd0dcae86..23f060db9255 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_rlc.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_rlc.h @@ -69,6 +69,47 @@ typedef enum _FIRMWARE_ID_ { FIRMWARE_ID_MAX = 38, } FIRMWARE_ID; +typedef enum _SOC21_FIRMWARE_ID_ { + SOC21_FIRMWARE_ID_INVALID = 0, + SOC21_FIRMWARE_ID_RLC_G_UCODE = 1, + SOC21_FIRMWARE_ID_RLC_TOC = 2, + SOC21_FIRMWARE_ID_RLCG_SCRATCH = 3, + SOC21_FIRMWARE_ID_RLC_SRM_ARAM = 4, + SOC21_FIRMWARE_ID_RLC_P_UCODE = 5, + SOC21_FIRMWARE_ID_RLC_V_UCODE = 6, + SOC21_FIRMWARE_ID_RLX6_UCODE = 7, + SOC21_FIRMWARE_ID_RLX6_UCODE_CORE1 = 8, + SOC21_FIRMWARE_ID_RLX6_DRAM_BOOT = 9, + SOC21_FIRMWARE_ID_RLX6_DRAM_BOOT_CORE1 = 10, + SOC21_FIRMWARE_ID_SDMA_UCODE_TH0 = 11, + SOC21_FIRMWARE_ID_SDMA_UCODE_TH1 = 12, + SOC21_FIRMWARE_ID_CP_PFP = 13, + SOC21_FIRMWARE_ID_CP_ME = 14, + SOC21_FIRMWARE_ID_CP_MEC = 15, + SOC21_FIRMWARE_ID_RS64_MES_P0 = 16, + SOC21_FIRMWARE_ID_RS64_MES_P1 = 17, + SOC21_FIRMWARE_ID_RS64_PFP = 18, + SOC21_FIRMWARE_ID_RS64_ME = 19, + SOC21_FIRMWARE_ID_RS64_MEC = 20, + SOC21_FIRMWARE_ID_RS64_MES_P0_STACK = 21, + SOC21_FIRMWARE_ID_RS64_MES_P1_STACK = 22, + SOC21_FIRMWARE_ID_RS64_PFP_P0_STACK = 23, + SOC21_FIRMWARE_ID_RS64_PFP_P1_STACK = 24, + SOC21_FIRMWARE_ID_RS64_ME_P0_STACK = 25, + SOC21_FIRMWARE_ID_RS64_ME_P1_STACK = 26, + SOC21_FIRMWARE_ID_RS64_MEC_P0_STACK = 27, + SOC21_FIRMWARE_ID_RS64_MEC_P1_STACK = 28, + SOC21_FIRMWARE_ID_RS64_MEC_P2_STACK = 29, + SOC21_FIRMWARE_ID_RS64_MEC_P3_STACK = 30, + SOC21_FIRMWARE_ID_RLC_SRM_DRAM_SR = 31, + SOC21_FIRMWARE_ID_RLCG_SCRATCH_SR = 32, + SOC21_FIRMWARE_ID_RLCP_SCRATCH_SR = 33, + SOC21_FIRMWARE_ID_RLCV_SCRATCH_SR = 34, + SOC21_FIRMWARE_ID_RLX6_DRAM_SR = 35, + SOC21_FIRMWARE_ID_RLX6_DRAM_SR_CORE1 = 36, + SOC21_FIRMWARE_ID_MAX = 37 +} SOC21_FIRMWARE_ID; + typedef struct _RLC_TABLE_OF_CONTENT { union { unsigned int DW0; @@ -127,11 +168,19 @@ struct amdgpu_rlc_funcs { void (*reset)(struct amdgpu_device *adev); void (*start)(struct amdgpu_device *adev); void (*update_spm_vmid)(struct amdgpu_device *adev, unsigned vmid); - void (*sriov_wreg)(struct amdgpu_device *adev, u32 offset, u32 v, u32 acc_flags, u32 hwip); - u32 (*sriov_rreg)(struct amdgpu_device *adev, u32 offset, u32 acc_flags, u32 hwip); bool (*is_rlcg_access_range)(struct amdgpu_device *adev, uint32_t reg); }; +struct amdgpu_rlcg_reg_access_ctrl { + uint32_t scratch_reg0; + uint32_t scratch_reg1; + uint32_t scratch_reg2; + uint32_t scratch_reg3; + uint32_t grbm_cntl; + uint32_t grbm_idx; + uint32_t spare_int; +}; + struct amdgpu_rlc { /* for power gating */ struct amdgpu_bo *save_restore_obj; @@ -171,6 +220,13 @@ struct amdgpu_rlc { u32 save_restore_list_srm_size_bytes; u32 rlc_iram_ucode_size_bytes; u32 rlc_dram_ucode_size_bytes; + u32 rlcp_ucode_size_bytes; + u32 rlcv_ucode_size_bytes; + u32 global_tap_delays_ucode_size_bytes; + u32 se0_tap_delays_ucode_size_bytes; + u32 se1_tap_delays_ucode_size_bytes; + u32 se2_tap_delays_ucode_size_bytes; + u32 se3_tap_delays_ucode_size_bytes; u32 *register_list_format; u32 *register_restore; @@ -179,6 +235,13 @@ struct amdgpu_rlc { u8 *save_restore_list_srm; u8 *rlc_iram_ucode; u8 *rlc_dram_ucode; + u8 *rlcp_ucode; + u8 *rlcv_ucode; + u8 *global_tap_delays_ucode; + u8 *se0_tap_delays_ucode; + u8 *se1_tap_delays_ucode; + u8 *se2_tap_delays_ucode; + u8 *se3_tap_delays_ucode; bool is_rlc_v2_1; @@ -191,6 +254,10 @@ struct amdgpu_rlc { struct amdgpu_bo *rlc_toc_bo; uint64_t rlc_toc_gpu_addr; void *rlc_toc_buf; + + bool rlcg_reg_access_supported; + /* registers for rlcg indirect reg access */ + struct amdgpu_rlcg_reg_access_ctrl reg_access_ctrl; }; void amdgpu_gfx_rlc_enter_safe_mode(struct amdgpu_device *adev); @@ -200,5 +267,7 @@ int amdgpu_gfx_rlc_init_csb(struct amdgpu_device *adev); int amdgpu_gfx_rlc_init_cpt(struct amdgpu_device *adev); void amdgpu_gfx_rlc_setup_cp_table(struct amdgpu_device *adev); void amdgpu_gfx_rlc_fini(struct amdgpu_device *adev); - +int amdgpu_gfx_rlc_init_microcode(struct amdgpu_device *adev, + uint16_t version_major, + uint16_t version_minor); #endif diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_sdma.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_sdma.c index 65debb65a5df..ea5278f094c0 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_sdma.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_sdma.c @@ -21,6 +21,7 @@ * */ +#include <linux/firmware.h> #include "amdgpu.h" #include "amdgpu_sdma.h" #include "amdgpu_ras.h" @@ -74,91 +75,60 @@ uint64_t amdgpu_sdma_get_csa_mc_addr(struct amdgpu_ring *ring, if (amdgpu_sriov_vf(adev) || vmid == 0 || !amdgpu_mcbp) return 0; - r = amdgpu_sdma_get_index_from_ring(ring, &index); + if (ring->is_mes_queue) { + uint32_t offset = 0; - if (r || index > 31) - csa_mc_addr = 0; - else - csa_mc_addr = amdgpu_csa_vaddr(adev) + - AMDGPU_CSA_SDMA_OFFSET + - index * AMDGPU_CSA_SDMA_SIZE; + offset = offsetof(struct amdgpu_mes_ctx_meta_data, + sdma[ring->idx].sdma_meta_data); + csa_mc_addr = amdgpu_mes_ctx_get_offs_gpu_addr(ring, offset); + } else { + r = amdgpu_sdma_get_index_from_ring(ring, &index); + + if (r || index > 31) + csa_mc_addr = 0; + else + csa_mc_addr = amdgpu_csa_vaddr(adev) + + AMDGPU_CSA_SDMA_OFFSET + + index * AMDGPU_CSA_SDMA_SIZE; + } return csa_mc_addr; } int amdgpu_sdma_ras_late_init(struct amdgpu_device *adev, - void *ras_ih_info) + struct ras_common_if *ras_block) { int r, i; - struct ras_ih_if *ih_info = (struct ras_ih_if *)ras_ih_info; - struct ras_fs_if fs_info = { - .sysfs_name = "sdma_err_count", - }; - - if (!ih_info) - return -EINVAL; - - if (!adev->sdma.ras_if) { - adev->sdma.ras_if = kmalloc(sizeof(struct ras_common_if), GFP_KERNEL); - if (!adev->sdma.ras_if) - return -ENOMEM; - adev->sdma.ras_if->block = AMDGPU_RAS_BLOCK__SDMA; - adev->sdma.ras_if->type = AMDGPU_RAS_ERROR__MULTI_UNCORRECTABLE; - adev->sdma.ras_if->sub_block_index = 0; - } - fs_info.head = ih_info->head = *adev->sdma.ras_if; - r = amdgpu_ras_late_init(adev, adev->sdma.ras_if, - &fs_info, ih_info); + r = amdgpu_ras_block_late_init(adev, ras_block); if (r) - goto free; + return r; - if (amdgpu_ras_is_supported(adev, adev->sdma.ras_if->block)) { + if (amdgpu_ras_is_supported(adev, ras_block->block)) { for (i = 0; i < adev->sdma.num_instances; i++) { r = amdgpu_irq_get(adev, &adev->sdma.ecc_irq, AMDGPU_SDMA_IRQ_INSTANCE0 + i); if (r) goto late_fini; } - } else { - r = 0; - goto free; } return 0; late_fini: - amdgpu_ras_late_fini(adev, adev->sdma.ras_if, ih_info); -free: - kfree(adev->sdma.ras_if); - adev->sdma.ras_if = NULL; + amdgpu_ras_block_late_fini(adev, ras_block); return r; } -void amdgpu_sdma_ras_fini(struct amdgpu_device *adev) -{ - if (amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__SDMA) && - adev->sdma.ras_if) { - struct ras_common_if *ras_if = adev->sdma.ras_if; - struct ras_ih_if ih_info = { - .head = *ras_if, - /* the cb member will not be used by - * amdgpu_ras_interrupt_remove_handler, init it only - * to cheat the check in ras_late_fini - */ - .cb = amdgpu_sdma_process_ras_data_cb, - }; - - amdgpu_ras_late_fini(adev, ras_if, &ih_info); - kfree(ras_if); - } -} - int amdgpu_sdma_process_ras_data_cb(struct amdgpu_device *adev, void *err_data, struct amdgpu_iv_entry *entry) { kgd2kfd_set_sram_ecc_flag(adev->kfd.dev); + + if (amdgpu_sriov_vf(adev)) + return AMDGPU_RAS_SUCCESS; + amdgpu_ras_reset_gpu(adev); return AMDGPU_RAS_SUCCESS; @@ -181,3 +151,158 @@ int amdgpu_sdma_process_ecc_irq(struct amdgpu_device *adev, amdgpu_ras_interrupt_dispatch(adev, &ih_data); return 0; } + +static int amdgpu_sdma_init_inst_ctx(struct amdgpu_sdma_instance *sdma_inst) +{ + int err = 0; + uint16_t version_major; + const struct common_firmware_header *header = NULL; + const struct sdma_firmware_header_v1_0 *hdr; + const struct sdma_firmware_header_v2_0 *hdr_v2; + + err = amdgpu_ucode_validate(sdma_inst->fw); + if (err) + return err; + + header = (const struct common_firmware_header *) + sdma_inst->fw->data; + version_major = le16_to_cpu(header->header_version_major); + + switch (version_major) { + case 1: + hdr = (const struct sdma_firmware_header_v1_0 *)sdma_inst->fw->data; + sdma_inst->fw_version = le32_to_cpu(hdr->header.ucode_version); + sdma_inst->feature_version = le32_to_cpu(hdr->ucode_feature_version); + break; + case 2: + hdr_v2 = (const struct sdma_firmware_header_v2_0 *)sdma_inst->fw->data; + sdma_inst->fw_version = le32_to_cpu(hdr_v2->header.ucode_version); + sdma_inst->feature_version = le32_to_cpu(hdr_v2->ucode_feature_version); + break; + default: + return -EINVAL; + } + + if (sdma_inst->feature_version >= 20) + sdma_inst->burst_nop = true; + + return 0; +} + +void amdgpu_sdma_destroy_inst_ctx(struct amdgpu_device *adev, + bool duplicate) +{ + int i; + + for (i = 0; i < adev->sdma.num_instances; i++) { + release_firmware(adev->sdma.instance[i].fw); + if (duplicate) + break; + } + + memset((void *)adev->sdma.instance, 0, + sizeof(struct amdgpu_sdma_instance) * AMDGPU_MAX_SDMA_INSTANCES); +} + +int amdgpu_sdma_init_microcode(struct amdgpu_device *adev, + char *fw_name, u32 instance, + bool duplicate) +{ + struct amdgpu_firmware_info *info = NULL; + const struct common_firmware_header *header = NULL; + int err = 0, i; + const struct sdma_firmware_header_v2_0 *sdma_hdr; + uint16_t version_major; + + err = request_firmware(&adev->sdma.instance[instance].fw, fw_name, adev->dev); + if (err) + goto out; + + header = (const struct common_firmware_header *) + adev->sdma.instance[instance].fw->data; + version_major = le16_to_cpu(header->header_version_major); + + if ((duplicate && instance) || (!duplicate && version_major > 1)) { + err = -EINVAL; + goto out; + } + + err = amdgpu_sdma_init_inst_ctx(&adev->sdma.instance[instance]); + if (err) + goto out; + + if (duplicate) { + for (i = 1; i < adev->sdma.num_instances; i++) + memcpy((void *)&adev->sdma.instance[i], + (void *)&adev->sdma.instance[0], + sizeof(struct amdgpu_sdma_instance)); + } + + if (amdgpu_sriov_vf(adev)) + return 0; + + DRM_DEBUG("psp_load == '%s'\n", + adev->firmware.load_type == AMDGPU_FW_LOAD_PSP ? "true" : "false"); + + if (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) { + switch (version_major) { + case 1: + for (i = 0; i < adev->sdma.num_instances; i++) { + if (!duplicate && (instance != i)) + continue; + else { + info = &adev->firmware.ucode[AMDGPU_UCODE_ID_SDMA0 + i]; + info->ucode_id = AMDGPU_UCODE_ID_SDMA0 + i; + info->fw = adev->sdma.instance[i].fw; + adev->firmware.fw_size += + ALIGN(le32_to_cpu(header->ucode_size_bytes), PAGE_SIZE); + } + } + break; + case 2: + sdma_hdr = (const struct sdma_firmware_header_v2_0 *) + adev->sdma.instance[0].fw->data; + info = &adev->firmware.ucode[AMDGPU_UCODE_ID_SDMA_UCODE_TH0]; + info->ucode_id = AMDGPU_UCODE_ID_SDMA_UCODE_TH0; + info->fw = adev->sdma.instance[0].fw; + adev->firmware.fw_size += + ALIGN(le32_to_cpu(sdma_hdr->ctx_ucode_size_bytes), PAGE_SIZE); + info = &adev->firmware.ucode[AMDGPU_UCODE_ID_SDMA_UCODE_TH1]; + info->ucode_id = AMDGPU_UCODE_ID_SDMA_UCODE_TH1; + info->fw = adev->sdma.instance[0].fw; + adev->firmware.fw_size += + ALIGN(le32_to_cpu(sdma_hdr->ctl_ucode_size_bytes), PAGE_SIZE); + break; + default: + err = -EINVAL; + } + } + +out: + if (err) { + DRM_ERROR("SDMA: Failed to init firmware \"%s\"\n", fw_name); + amdgpu_sdma_destroy_inst_ctx(adev, duplicate); + } + return err; +} + +void amdgpu_sdma_unset_buffer_funcs_helper(struct amdgpu_device *adev) +{ + struct amdgpu_ring *sdma; + int i; + + for (i = 0; i < adev->sdma.num_instances; i++) { + if (adev->sdma.has_page_queue) { + sdma = &adev->sdma.instance[i].page; + if (adev->mman.buffer_funcs_ring == sdma) { + amdgpu_ttm_set_buffer_funcs_status(adev, false); + break; + } + } + sdma = &adev->sdma.instance[i].ring; + if (adev->mman.buffer_funcs_ring == sdma) { + amdgpu_ttm_set_buffer_funcs_status(adev, false); + break; + } + } +} diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_sdma.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_sdma.h index f8fb755e3aa6..7d99205c2e01 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_sdma.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_sdma.h @@ -23,6 +23,7 @@ #ifndef __AMDGPU_SDMA_H__ #define __AMDGPU_SDMA_H__ +#include "amdgpu_ras.h" /* max number of IP instances */ #define AMDGPU_MAX_SDMA_INSTANCES 8 @@ -50,13 +51,8 @@ struct amdgpu_sdma_instance { bool burst_nop; }; -struct amdgpu_sdma_ras_funcs { - int (*ras_late_init)(struct amdgpu_device *adev, - void *ras_ih_info); - void (*ras_fini)(struct amdgpu_device *adev); - int (*query_ras_error_count)(struct amdgpu_device *adev, - uint32_t instance, void *ras_error_status); - void (*reset_ras_error_count)(struct amdgpu_device *adev); +struct amdgpu_sdma_ras { + struct amdgpu_ras_block_object ras_block; }; struct amdgpu_sdma { @@ -73,7 +69,7 @@ struct amdgpu_sdma { uint32_t srbm_soft_reset; bool has_page_queue; struct ras_common_if *ras_if; - const struct amdgpu_sdma_ras_funcs *funcs; + struct amdgpu_sdma_ras *ras; }; /* @@ -121,12 +117,17 @@ amdgpu_sdma_get_instance_from_ring(struct amdgpu_ring *ring); int amdgpu_sdma_get_index_from_ring(struct amdgpu_ring *ring, uint32_t *index); uint64_t amdgpu_sdma_get_csa_mc_addr(struct amdgpu_ring *ring, unsigned vmid); int amdgpu_sdma_ras_late_init(struct amdgpu_device *adev, - void *ras_ih_info); -void amdgpu_sdma_ras_fini(struct amdgpu_device *adev); + struct ras_common_if *ras_block); int amdgpu_sdma_process_ras_data_cb(struct amdgpu_device *adev, void *err_data, struct amdgpu_iv_entry *entry); int amdgpu_sdma_process_ecc_irq(struct amdgpu_device *adev, struct amdgpu_irq_src *source, struct amdgpu_iv_entry *entry); +int amdgpu_sdma_init_microcode(struct amdgpu_device *adev, + char *fw_name, u32 instance, bool duplicate); +void amdgpu_sdma_destroy_inst_ctx(struct amdgpu_device *adev, + bool duplicate); +void amdgpu_sdma_unset_buffer_funcs_helper(struct amdgpu_device *adev); + #endif diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_smuio.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_smuio.h index 484bb3dcec47..c7a823f3f2c5 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_smuio.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_smuio.h @@ -27,7 +27,7 @@ struct amdgpu_smuio_funcs { u32 (*get_rom_index_offset)(struct amdgpu_device *adev); u32 (*get_rom_data_offset)(struct amdgpu_device *adev); void (*update_rom_clock_gating)(struct amdgpu_device *adev, bool enable); - void (*get_clock_gating_state)(struct amdgpu_device *adev, u32 *flags); + void (*get_clock_gating_state)(struct amdgpu_device *adev, u64 *flags); u32 (*get_die_id)(struct amdgpu_device *adev); u32 (*get_socket_id)(struct amdgpu_device *adev); bool (*is_host_gpu_xgmi_supported)(struct amdgpu_device *adev); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_sync.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_sync.c index f7d8487799b2..090e66a1b284 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_sync.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_sync.c @@ -1,3 +1,4 @@ +// SPDX-License-Identifier: MIT /* * Copyright 2014 Advanced Micro Devices, Inc. * All Rights Reserved. @@ -51,7 +52,6 @@ static struct kmem_cache *amdgpu_sync_slab; void amdgpu_sync_create(struct amdgpu_sync *sync) { hash_init(sync->fences); - sync->last_vm_update = NULL; } /** @@ -171,23 +171,6 @@ int amdgpu_sync_fence(struct amdgpu_sync *sync, struct dma_fence *f) return 0; } -/** - * amdgpu_sync_vm_fence - remember to sync to this VM fence - * - * @sync: sync object to add fence to - * @fence: the VM fence to add - * - * Add the fence to the sync object and remember it as VM update. - */ -int amdgpu_sync_vm_fence(struct amdgpu_sync *sync, struct dma_fence *fence) -{ - if (!fence) - return 0; - - amdgpu_sync_keep_later(&sync->last_vm_update, fence); - return amdgpu_sync_fence(sync, fence); -} - /* Determine based on the owner and mode if we should sync to a fence or not */ static bool amdgpu_sync_test_fence(struct amdgpu_device *adev, enum amdgpu_sync_mode mode, @@ -259,12 +242,12 @@ int amdgpu_sync_resv(struct amdgpu_device *adev, struct amdgpu_sync *sync, if (resv == NULL) return -EINVAL; - dma_resv_for_each_fence(&cursor, resv, true, f) { + /* TODO: Use DMA_RESV_USAGE_READ here */ + dma_resv_for_each_fence(&cursor, resv, DMA_RESV_USAGE_BOOKKEEP, f) { dma_fence_chain_for_each(f, f) { - struct dma_fence_chain *chain = to_dma_fence_chain(f); + struct dma_fence *tmp = dma_fence_chain_contained(f); - if (amdgpu_sync_test_fence(adev, mode, owner, chain ? - chain->fence : f)) { + if (amdgpu_sync_test_fence(adev, mode, owner, tmp)) { r = amdgpu_sync_fence(sync, f); dma_fence_put(f); if (r) @@ -333,6 +316,7 @@ struct dma_fence *amdgpu_sync_get_fence(struct amdgpu_sync *sync) struct hlist_node *tmp; struct dma_fence *f; int i; + hash_for_each_safe(sync->fences, i, tmp, e, node) { f = e->fence; @@ -377,9 +361,6 @@ int amdgpu_sync_clone(struct amdgpu_sync *source, struct amdgpu_sync *clone) } } - dma_fence_put(clone->last_vm_update); - clone->last_vm_update = dma_fence_get(source->last_vm_update); - return 0; } @@ -413,15 +394,13 @@ void amdgpu_sync_free(struct amdgpu_sync *sync) { struct amdgpu_sync_entry *e; struct hlist_node *tmp; - unsigned i; + unsigned int i; hash_for_each_safe(sync->fences, i, tmp, e, node) { hash_del(&e->node); dma_fence_put(e->fence); kmem_cache_free(amdgpu_sync_slab, e); } - - dma_fence_put(sync->last_vm_update); } /** diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_sync.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_sync.h index 7c0fe20c470d..2d5c613cda10 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_sync.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_sync.h @@ -43,12 +43,10 @@ enum amdgpu_sync_mode { */ struct amdgpu_sync { DECLARE_HASHTABLE(fences, 4); - struct dma_fence *last_vm_update; }; void amdgpu_sync_create(struct amdgpu_sync *sync); int amdgpu_sync_fence(struct amdgpu_sync *sync, struct dma_fence *f); -int amdgpu_sync_vm_fence(struct amdgpu_sync *sync, struct dma_fence *fence); int amdgpu_sync_resv(struct amdgpu_device *adev, struct amdgpu_sync *sync, struct dma_resv *resv, enum amdgpu_sync_mode mode, void *owner); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_test.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_test.c deleted file mode 100644 index 909d830b513e..000000000000 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_test.c +++ /dev/null @@ -1,250 +0,0 @@ -// SPDX-License-Identifier: GPL-2.0 OR MIT -/* - * Copyright 2009 VMware, Inc. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * the rights to use, copy, modify, merge, publish, distribute, sublicense, - * and/or sell copies of the Software, and to permit persons to whom the - * Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice shall be included in - * all copies or substantial portions of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL - * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR - * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, - * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR - * OTHER DEALINGS IN THE SOFTWARE. - * - * Authors: Michel Dänzer - */ - -#include <drm/amdgpu_drm.h> -#include "amdgpu.h" -#include "amdgpu_uvd.h" -#include "amdgpu_vce.h" - -/* Test BO GTT->VRAM and VRAM->GTT GPU copies across the whole GTT aperture */ -static void amdgpu_do_test_moves(struct amdgpu_device *adev) -{ - struct amdgpu_ring *ring = adev->mman.buffer_funcs_ring; - struct amdgpu_bo *vram_obj = NULL; - struct amdgpu_bo **gtt_obj = NULL; - struct amdgpu_bo_param bp; - uint64_t gart_addr, vram_addr; - unsigned n, size; - int i, r; - - size = 1024 * 1024; - - /* Number of tests = - * (Total GTT - gart_pin_size - (2 transfer windows for buffer moves)) / test size - */ - n = adev->gmc.gart_size - atomic64_read(&adev->gart_pin_size); - n -= AMDGPU_GTT_MAX_TRANSFER_SIZE * AMDGPU_GTT_NUM_TRANSFER_WINDOWS * - AMDGPU_GPU_PAGE_SIZE; - n /= size; - - gtt_obj = kcalloc(n, sizeof(*gtt_obj), GFP_KERNEL); - if (!gtt_obj) { - DRM_ERROR("Failed to allocate %d pointers\n", n); - r = 1; - goto out_cleanup; - } - memset(&bp, 0, sizeof(bp)); - bp.size = size; - bp.byte_align = PAGE_SIZE; - bp.domain = AMDGPU_GEM_DOMAIN_VRAM; - bp.flags = 0; - bp.type = ttm_bo_type_kernel; - bp.resv = NULL; - bp.bo_ptr_size = sizeof(struct amdgpu_bo); - - r = amdgpu_bo_create(adev, &bp, &vram_obj); - if (r) { - DRM_ERROR("Failed to create VRAM object\n"); - goto out_cleanup; - } - r = amdgpu_bo_reserve(vram_obj, false); - if (unlikely(r != 0)) - goto out_unref; - r = amdgpu_bo_pin(vram_obj, AMDGPU_GEM_DOMAIN_VRAM); - if (r) { - DRM_ERROR("Failed to pin VRAM object\n"); - goto out_unres; - } - vram_addr = amdgpu_bo_gpu_offset(vram_obj); - for (i = 0; i < n; i++) { - void *gtt_map, *vram_map; - void **gart_start, **gart_end; - void **vram_start, **vram_end; - struct dma_fence *fence = NULL; - - bp.domain = AMDGPU_GEM_DOMAIN_GTT; - r = amdgpu_bo_create(adev, &bp, gtt_obj + i); - if (r) { - DRM_ERROR("Failed to create GTT object %d\n", i); - goto out_lclean; - } - - r = amdgpu_bo_reserve(gtt_obj[i], false); - if (unlikely(r != 0)) - goto out_lclean_unref; - r = amdgpu_bo_pin(gtt_obj[i], AMDGPU_GEM_DOMAIN_GTT); - if (r) { - DRM_ERROR("Failed to pin GTT object %d\n", i); - goto out_lclean_unres; - } - r = amdgpu_ttm_alloc_gart(>t_obj[i]->tbo); - if (r) { - DRM_ERROR("%p bind failed\n", gtt_obj[i]); - goto out_lclean_unpin; - } - gart_addr = amdgpu_bo_gpu_offset(gtt_obj[i]); - - r = amdgpu_bo_kmap(gtt_obj[i], >t_map); - if (r) { - DRM_ERROR("Failed to map GTT object %d\n", i); - goto out_lclean_unpin; - } - - for (gart_start = gtt_map, gart_end = gtt_map + size; - gart_start < gart_end; - gart_start++) - *gart_start = gart_start; - - amdgpu_bo_kunmap(gtt_obj[i]); - - r = amdgpu_copy_buffer(ring, gart_addr, vram_addr, - size, NULL, &fence, false, false, false); - - if (r) { - DRM_ERROR("Failed GTT->VRAM copy %d\n", i); - goto out_lclean_unpin; - } - - r = dma_fence_wait(fence, false); - if (r) { - DRM_ERROR("Failed to wait for GTT->VRAM fence %d\n", i); - goto out_lclean_unpin; - } - - dma_fence_put(fence); - fence = NULL; - - r = amdgpu_bo_kmap(vram_obj, &vram_map); - if (r) { - DRM_ERROR("Failed to map VRAM object after copy %d\n", i); - goto out_lclean_unpin; - } - - for (gart_start = gtt_map, gart_end = gtt_map + size, - vram_start = vram_map, vram_end = vram_map + size; - vram_start < vram_end; - gart_start++, vram_start++) { - if (*vram_start != gart_start) { - DRM_ERROR("Incorrect GTT->VRAM copy %d: Got 0x%p, " - "expected 0x%p (GTT/VRAM offset " - "0x%16llx/0x%16llx)\n", - i, *vram_start, gart_start, - (unsigned long long) - (gart_addr - adev->gmc.gart_start + - (void *)gart_start - gtt_map), - (unsigned long long) - (vram_addr - adev->gmc.vram_start + - (void *)gart_start - gtt_map)); - amdgpu_bo_kunmap(vram_obj); - goto out_lclean_unpin; - } - *vram_start = vram_start; - } - - amdgpu_bo_kunmap(vram_obj); - - r = amdgpu_copy_buffer(ring, vram_addr, gart_addr, - size, NULL, &fence, false, false, false); - - if (r) { - DRM_ERROR("Failed VRAM->GTT copy %d\n", i); - goto out_lclean_unpin; - } - - r = dma_fence_wait(fence, false); - if (r) { - DRM_ERROR("Failed to wait for VRAM->GTT fence %d\n", i); - goto out_lclean_unpin; - } - - dma_fence_put(fence); - fence = NULL; - - r = amdgpu_bo_kmap(gtt_obj[i], >t_map); - if (r) { - DRM_ERROR("Failed to map GTT object after copy %d\n", i); - goto out_lclean_unpin; - } - - for (gart_start = gtt_map, gart_end = gtt_map + size, - vram_start = vram_map, vram_end = vram_map + size; - gart_start < gart_end; - gart_start++, vram_start++) { - if (*gart_start != vram_start) { - DRM_ERROR("Incorrect VRAM->GTT copy %d: Got 0x%p, " - "expected 0x%p (VRAM/GTT offset " - "0x%16llx/0x%16llx)\n", - i, *gart_start, vram_start, - (unsigned long long) - (vram_addr - adev->gmc.vram_start + - (void *)vram_start - vram_map), - (unsigned long long) - (gart_addr - adev->gmc.gart_start + - (void *)vram_start - vram_map)); - amdgpu_bo_kunmap(gtt_obj[i]); - goto out_lclean_unpin; - } - } - - amdgpu_bo_kunmap(gtt_obj[i]); - - DRM_INFO("Tested GTT->VRAM and VRAM->GTT copy for GTT offset 0x%llx\n", - gart_addr - adev->gmc.gart_start); - continue; - -out_lclean_unpin: - amdgpu_bo_unpin(gtt_obj[i]); -out_lclean_unres: - amdgpu_bo_unreserve(gtt_obj[i]); -out_lclean_unref: - amdgpu_bo_unref(>t_obj[i]); -out_lclean: - for (--i; i >= 0; --i) { - amdgpu_bo_unpin(gtt_obj[i]); - amdgpu_bo_unreserve(gtt_obj[i]); - amdgpu_bo_unref(>t_obj[i]); - } - if (fence) - dma_fence_put(fence); - break; - } - - amdgpu_bo_unpin(vram_obj); -out_unres: - amdgpu_bo_unreserve(vram_obj); -out_unref: - amdgpu_bo_unref(&vram_obj); -out_cleanup: - kfree(gtt_obj); - if (r) { - pr_warn("Error while testing BO move\n"); - } -} - -void amdgpu_test_moves(struct amdgpu_device *adev) -{ - if (adev->mman.buffer_funcs) - amdgpu_do_test_moves(adev); -} diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_trace.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_trace.h index d855cb53c7e0..5e6ddc7e101c 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_trace.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_trace.h @@ -140,8 +140,10 @@ TRACE_EVENT(amdgpu_bo_create, ); TRACE_EVENT(amdgpu_cs, - TP_PROTO(struct amdgpu_cs_parser *p, int i), - TP_ARGS(p, i), + TP_PROTO(struct amdgpu_cs_parser *p, + struct amdgpu_job *job, + struct amdgpu_ib *ib), + TP_ARGS(p, job, ib), TP_STRUCT__entry( __field(struct amdgpu_bo_list *, bo_list) __field(u32, ring) @@ -151,10 +153,10 @@ TRACE_EVENT(amdgpu_cs, TP_fast_assign( __entry->bo_list = p->bo_list; - __entry->ring = to_amdgpu_ring(p->entity->rq->sched)->idx; - __entry->dw = p->job->ibs[i].length_dw; + __entry->ring = to_amdgpu_ring(job->base.sched)->idx; + __entry->dw = ib->length_dw; __entry->fences = amdgpu_fence_count_emitted( - to_amdgpu_ring(p->entity->rq->sched)); + to_amdgpu_ring(job->base.sched)); ), TP_printk("bo_list=%p, ring=%u, dw=%u, fences=%u", __entry->bo_list, __entry->ring, __entry->dw, @@ -358,11 +360,10 @@ TRACE_EVENT(amdgpu_vm_update_ptes, } ), TP_printk("pid:%u vm_ctx:0x%llx start:0x%010llx end:0x%010llx," - " flags:0x%llx, incr:%llu, dst:\n%s%s", __entry->pid, + " flags:0x%llx, incr:%llu, dst:\n%s", __entry->pid, __entry->vm_ctx, __entry->start, __entry->end, __entry->flags, __entry->incr, __print_array( - __get_dynamic_array(dst), min(__entry->nptes, 32u), 8), - __entry->nptes > 32 ? "..." : "") + __get_dynamic_array(dst), __entry->nptes, 8)) ); TRACE_EVENT(amdgpu_vm_set_ptes, @@ -537,6 +538,22 @@ TRACE_EVENT(amdgpu_ib_pipe_sync, __entry->seqno) ); +TRACE_EVENT(amdgpu_reset_reg_dumps, + TP_PROTO(uint32_t address, uint32_t value), + TP_ARGS(address, value), + TP_STRUCT__entry( + __field(uint32_t, address) + __field(uint32_t, value) + ), + TP_fast_assign( + __entry->address = address; + __entry->value = value; + ), + TP_printk("amdgpu register dump 0x%x: 0x%x", + __entry->address, + __entry->value) +); + #undef AMDGPU_JOB_GET_TIMELINE_NAME #endif diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_trace_points.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_trace_points.c index 57c6c39ba064..b96d885f6e33 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_trace_points.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_trace_points.c @@ -23,6 +23,7 @@ */ #include <drm/amdgpu_drm.h> +#include "amdgpu_cs.h" #include "amdgpu.h" #define CREATE_TRACE_POINTS diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c index fb0d8bffdce2..57277b1cf183 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c @@ -43,12 +43,14 @@ #include <linux/sizes.h> #include <linux/module.h> +#include <drm/drm_drv.h> #include <drm/ttm/ttm_bo_api.h> #include <drm/ttm/ttm_bo_driver.h> #include <drm/ttm/ttm_placement.h> #include <drm/ttm/ttm_range_manager.h> #include <drm/amdgpu_drm.h> +#include <drm/drm_drv.h> #include "amdgpu.h" #include "amdgpu_object.h" @@ -115,7 +117,7 @@ static void amdgpu_evict_flags(struct ttm_buffer_object *bo, } abo = ttm_to_amdgpu_bo(bo); - if (abo->flags & AMDGPU_AMDKFD_CREATE_SVM_BO) { + if (abo->flags & AMDGPU_GEM_CREATE_DISCARDABLE) { placement->num_placement = 0; placement->num_busy_placement = 0; return; @@ -169,10 +171,10 @@ static void amdgpu_evict_flags(struct ttm_buffer_object *bo, * @bo: buffer object to map * @mem: memory object to map * @mm_cur: range to map - * @num_pages: number of pages to map * @window: which GART window to use * @ring: DMA ring to use for the copy * @tmz: if we should setup a TMZ enabled mapping + * @size: in number of bytes to map, out number of bytes mapped * @addr: resulting address inside the MC address space * * Setup one of the GART windows to access a specific piece of memory or return @@ -181,15 +183,14 @@ static void amdgpu_evict_flags(struct ttm_buffer_object *bo, static int amdgpu_ttm_map_buffer(struct ttm_buffer_object *bo, struct ttm_resource *mem, struct amdgpu_res_cursor *mm_cur, - unsigned num_pages, unsigned window, - struct amdgpu_ring *ring, bool tmz, - uint64_t *addr) + unsigned window, struct amdgpu_ring *ring, + bool tmz, uint64_t *size, uint64_t *addr) { struct amdgpu_device *adev = ring->adev; - struct amdgpu_job *job; - unsigned num_dw, num_bytes; - struct dma_fence *fence; + unsigned offset, num_pages, num_dw, num_bytes; uint64_t src_addr, dst_addr; + struct dma_fence *fence; + struct amdgpu_job *job; void *cpu_addr; uint64_t flags; unsigned int i; @@ -197,7 +198,9 @@ static int amdgpu_ttm_map_buffer(struct ttm_buffer_object *bo, BUG_ON(adev->mman.buffer_funcs->copy_max_bytes < AMDGPU_GTT_MAX_TRANSFER_SIZE * 8); - BUG_ON(mem->mem_type == AMDGPU_PL_PREEMPT); + + if (WARN_ON(mem->mem_type == AMDGPU_PL_PREEMPT)) + return -EINVAL; /* Map only what can't be accessed directly */ if (!tmz && mem->start != AMDGPU_BO_INVALID_OFFSET) { @@ -206,10 +209,22 @@ static int amdgpu_ttm_map_buffer(struct ttm_buffer_object *bo, return 0; } + + /* + * If start begins at an offset inside the page, then adjust the size + * and addr accordingly + */ + offset = mm_cur->start & ~PAGE_MASK; + + num_pages = PFN_UP(*size + offset); + num_pages = min_t(uint32_t, num_pages, AMDGPU_GTT_MAX_TRANSFER_SIZE); + + *size = min(*size, (uint64_t)num_pages * PAGE_SIZE - offset); + *addr = adev->gmc.gart_start; *addr += (u64)window * AMDGPU_GTT_MAX_TRANSFER_SIZE * AMDGPU_GPU_PAGE_SIZE; - *addr += mm_cur->start & ~PAGE_MASK; + *addr += offset; num_dw = ALIGN(adev->mman.buffer_funcs->copy_num_dw, 8); num_bytes = num_pages * 8 * AMDGPU_GPU_PAGES_IN_CPU_PAGE; @@ -240,10 +255,7 @@ static int amdgpu_ttm_map_buffer(struct ttm_buffer_object *bo, dma_addr_t *dma_addr; dma_addr = &bo->ttm->dma_address[mm_cur->start >> PAGE_SHIFT]; - r = amdgpu_gart_map(adev, 0, num_pages, dma_addr, flags, - cpu_addr); - if (r) - goto error_free; + amdgpu_gart_map(adev, 0, num_pages, dma_addr, flags, cpu_addr); } else { dma_addr_t dma_address; @@ -251,11 +263,8 @@ static int amdgpu_ttm_map_buffer(struct ttm_buffer_object *bo, dma_address += adev->vm_manager.vram_base_offset; for (i = 0; i < num_pages; ++i) { - r = amdgpu_gart_map(adev, i << PAGE_SHIFT, 1, - &dma_address, flags, cpu_addr); - if (r) - goto error_free; - + amdgpu_gart_map(adev, i << PAGE_SHIFT, 1, &dma_address, + flags, cpu_addr); dma_address += PAGE_SIZE; } } @@ -296,9 +305,6 @@ int amdgpu_ttm_copy_mem_to_mem(struct amdgpu_device *adev, struct dma_resv *resv, struct dma_fence **f) { - const uint32_t GTT_MAX_BYTES = (AMDGPU_GTT_MAX_TRANSFER_SIZE * - AMDGPU_GPU_PAGE_SIZE); - struct amdgpu_ring *ring = adev->mman.buffer_funcs_ring; struct amdgpu_res_cursor src_mm, dst_mm; struct dma_fence *fence = NULL; @@ -314,29 +320,20 @@ int amdgpu_ttm_copy_mem_to_mem(struct amdgpu_device *adev, mutex_lock(&adev->mman.gtt_window_lock); while (src_mm.remaining) { - uint32_t src_page_offset = src_mm.start & ~PAGE_MASK; - uint32_t dst_page_offset = dst_mm.start & ~PAGE_MASK; + uint64_t from, to, cur_size; struct dma_fence *next; - uint32_t cur_size; - uint64_t from, to; - /* Copy size cannot exceed GTT_MAX_BYTES. So if src or dst - * begins at an offset, then adjust the size accordingly - */ - cur_size = max(src_page_offset, dst_page_offset); - cur_size = min(min3(src_mm.size, dst_mm.size, size), - (uint64_t)(GTT_MAX_BYTES - cur_size)); + /* Never copy more than 256MiB at once to avoid a timeout */ + cur_size = min3(src_mm.size, dst_mm.size, 256ULL << 20); /* Map src to window 0 and dst to window 1. */ r = amdgpu_ttm_map_buffer(src->bo, src->mem, &src_mm, - PFN_UP(cur_size + src_page_offset), - 0, ring, tmz, &from); + 0, ring, tmz, &cur_size, &from); if (r) goto error; r = amdgpu_ttm_map_buffer(dst->bo, dst->mem, &dst_mm, - PFN_UP(cur_size + dst_page_offset), - 1, ring, tmz, &to); + 1, ring, tmz, &cur_size, &to); if (r) goto error; @@ -395,8 +392,7 @@ static int amdgpu_move_blit(struct ttm_buffer_object *bo, (abo->flags & AMDGPU_GEM_CREATE_VRAM_WIPE_ON_RELEASE)) { struct dma_fence *wipe_fence = NULL; - r = amdgpu_fill_buffer(ttm_to_amdgpu_bo(bo), AMDGPU_POISON, - NULL, &wipe_fence); + r = amdgpu_fill_buffer(abo, AMDGPU_POISON, NULL, &wipe_fence); if (r) { goto error; } else if (wipe_fence) { @@ -428,8 +424,9 @@ error: static bool amdgpu_mem_visible(struct amdgpu_device *adev, struct ttm_resource *mem) { - uint64_t mem_size = (u64)mem->num_pages << PAGE_SHIFT; + u64 mem_size = (u64)mem->num_pages << PAGE_SHIFT; struct amdgpu_res_cursor cursor; + u64 end; if (mem->mem_type == TTM_PL_SYSTEM || mem->mem_type == TTM_PL_TT) @@ -438,12 +435,21 @@ static bool amdgpu_mem_visible(struct amdgpu_device *adev, return false; amdgpu_res_first(mem, 0, mem_size, &cursor); + end = cursor.start + cursor.size; + while (cursor.remaining) { + amdgpu_res_next(&cursor, cursor.size); - /* ttm_resource_ioremap only supports contiguous memory */ - if (cursor.size != mem_size) - return false; + if (!cursor.remaining) + break; + + /* ttm_resource_ioremap only supports contiguous memory */ + if (end != cursor.start) + return false; + + end = cursor.start + cursor.size; + } - return cursor.start + cursor.size <= adev->gmc.visible_vram_size; + return end <= adev->gmc.visible_vram_size; } /* @@ -475,7 +481,8 @@ static int amdgpu_bo_move(struct ttm_buffer_object *bo, bool evict, adev = amdgpu_ttm_adev(bo->bdev); - if (old_mem->mem_type == TTM_PL_SYSTEM && bo->ttm == NULL) { + if (!old_mem || (old_mem->mem_type == TTM_PL_SYSTEM && + bo->ttm == NULL)) { ttm_bo_move_null(bo, new_mem); goto out; } @@ -641,6 +648,8 @@ struct amdgpu_ttm_tt { #endif }; +#define ttm_to_amdgpu_ttm_tt(ptr) container_of(ptr, struct amdgpu_ttm_tt, ttm) + #ifdef CONFIG_DRM_AMDGPU_USERPTR /* * amdgpu_ttm_tt_get_user_pages - get device accessible pages that back user @@ -652,7 +661,7 @@ struct amdgpu_ttm_tt { int amdgpu_ttm_tt_get_user_pages(struct amdgpu_bo *bo, struct page **pages) { struct ttm_tt *ttm = bo->tbo.ttm; - struct amdgpu_ttm_tt *gtt = (void *)ttm; + struct amdgpu_ttm_tt *gtt = ttm_to_amdgpu_ttm_tt(ttm); unsigned long start = gtt->userptr; struct vm_area_struct *vma; struct mm_struct *mm; @@ -706,7 +715,7 @@ out_unlock: */ bool amdgpu_ttm_tt_get_user_pages_done(struct ttm_tt *ttm) { - struct amdgpu_ttm_tt *gtt = (void *)ttm; + struct amdgpu_ttm_tt *gtt = ttm_to_amdgpu_ttm_tt(ttm); bool r = false; if (!gtt || !gtt->userptr) @@ -755,7 +764,7 @@ static int amdgpu_ttm_tt_pin_userptr(struct ttm_device *bdev, struct ttm_tt *ttm) { struct amdgpu_device *adev = amdgpu_ttm_adev(bdev); - struct amdgpu_ttm_tt *gtt = (void *)ttm; + struct amdgpu_ttm_tt *gtt = ttm_to_amdgpu_ttm_tt(ttm); int write = !(gtt->userflags & AMDGPU_GEM_USERPTR_READONLY); enum dma_data_direction direction = write ? DMA_BIDIRECTIONAL : DMA_TO_DEVICE; @@ -792,7 +801,7 @@ static void amdgpu_ttm_tt_unpin_userptr(struct ttm_device *bdev, struct ttm_tt *ttm) { struct amdgpu_device *adev = amdgpu_ttm_adev(bdev); - struct amdgpu_ttm_tt *gtt = (void *)ttm; + struct amdgpu_ttm_tt *gtt = ttm_to_amdgpu_ttm_tt(ttm); int write = !(gtt->userflags & AMDGPU_GEM_USERPTR_READONLY); enum dma_data_direction direction = write ? DMA_BIDIRECTIONAL : DMA_TO_DEVICE; @@ -820,14 +829,13 @@ static void amdgpu_ttm_tt_unpin_userptr(struct ttm_device *bdev, #endif } -static int amdgpu_ttm_gart_bind(struct amdgpu_device *adev, - struct ttm_buffer_object *tbo, - uint64_t flags) +static void amdgpu_ttm_gart_bind(struct amdgpu_device *adev, + struct ttm_buffer_object *tbo, + uint64_t flags) { struct amdgpu_bo *abo = ttm_to_amdgpu_bo(tbo); struct ttm_tt *ttm = tbo->ttm; - struct amdgpu_ttm_tt *gtt = (void *)ttm; - int r; + struct amdgpu_ttm_tt *gtt = ttm_to_amdgpu_ttm_tt(ttm); if (amdgpu_bo_encrypted(abo)) flags |= AMDGPU_PTE_TMZ; @@ -835,10 +843,8 @@ static int amdgpu_ttm_gart_bind(struct amdgpu_device *adev, if (abo->flags & AMDGPU_GEM_CREATE_CP_MQD_GFX9) { uint64_t page_idx = 1; - r = amdgpu_gart_bind(adev, gtt->offset, page_idx, - gtt->ttm.dma_address, flags); - if (r) - goto gart_bind_fail; + amdgpu_gart_bind(adev, gtt->offset, page_idx, + gtt->ttm.dma_address, flags); /* The memory type of the first page defaults to UC. Now * modify the memory type to NC from the second page of @@ -847,21 +853,13 @@ static int amdgpu_ttm_gart_bind(struct amdgpu_device *adev, flags &= ~AMDGPU_PTE_MTYPE_VG10_MASK; flags |= AMDGPU_PTE_MTYPE_VG10(AMDGPU_MTYPE_NC); - r = amdgpu_gart_bind(adev, - gtt->offset + (page_idx << PAGE_SHIFT), - ttm->num_pages - page_idx, - &(gtt->ttm.dma_address[page_idx]), flags); + amdgpu_gart_bind(adev, gtt->offset + (page_idx << PAGE_SHIFT), + ttm->num_pages - page_idx, + &(gtt->ttm.dma_address[page_idx]), flags); } else { - r = amdgpu_gart_bind(adev, gtt->offset, ttm->num_pages, - gtt->ttm.dma_address, flags); + amdgpu_gart_bind(adev, gtt->offset, ttm->num_pages, + gtt->ttm.dma_address, flags); } - -gart_bind_fail: - if (r) - DRM_ERROR("failed to bind %u pages at 0x%08llX\n", - ttm->num_pages, gtt->offset); - - return r; } /* @@ -875,9 +873,9 @@ static int amdgpu_ttm_backend_bind(struct ttm_device *bdev, struct ttm_resource *bo_mem) { struct amdgpu_device *adev = amdgpu_ttm_adev(bdev); - struct amdgpu_ttm_tt *gtt = (void*)ttm; + struct amdgpu_ttm_tt *gtt = ttm_to_amdgpu_ttm_tt(ttm); uint64_t flags; - int r = 0; + int r; if (!bo_mem) return -EINVAL; @@ -924,14 +922,10 @@ static int amdgpu_ttm_backend_bind(struct ttm_device *bdev, /* bind pages into GART page tables */ gtt->offset = (u64)bo_mem->start << PAGE_SHIFT; - r = amdgpu_gart_bind(adev, gtt->offset, ttm->num_pages, - gtt->ttm.dma_address, flags); - - if (r) - DRM_ERROR("failed to bind %u pages at 0x%08llX\n", - ttm->num_pages, gtt->offset); + amdgpu_gart_bind(adev, gtt->offset, ttm->num_pages, + gtt->ttm.dma_address, flags); gtt->bound = true; - return r; + return 0; } /* @@ -946,7 +940,7 @@ int amdgpu_ttm_alloc_gart(struct ttm_buffer_object *bo) { struct amdgpu_device *adev = amdgpu_ttm_adev(bo->bdev); struct ttm_operation_ctx ctx = { false, false }; - struct amdgpu_ttm_tt *gtt = (void *)bo->ttm; + struct amdgpu_ttm_tt *gtt = ttm_to_amdgpu_ttm_tt(bo->ttm); struct ttm_placement placement; struct ttm_place placements; struct ttm_resource *tmp; @@ -981,12 +975,7 @@ int amdgpu_ttm_alloc_gart(struct ttm_buffer_object *bo) /* Bind pages */ gtt->offset = (u64)tmp->start << PAGE_SHIFT; - r = amdgpu_ttm_gart_bind(adev, bo, flags); - if (unlikely(r)) { - ttm_resource_free(bo, &tmp); - return r; - } - + amdgpu_ttm_gart_bind(adev, bo, flags); amdgpu_gart_invalidate_tlb(adev); ttm_resource_free(bo, &bo->resource); ttm_bo_assign_mem(bo, tmp); @@ -1000,19 +989,16 @@ int amdgpu_ttm_alloc_gart(struct ttm_buffer_object *bo) * Called by amdgpu_gtt_mgr_recover() from amdgpu_device_reset() to * rebind GTT pages during a GPU reset. */ -int amdgpu_ttm_recover_gart(struct ttm_buffer_object *tbo) +void amdgpu_ttm_recover_gart(struct ttm_buffer_object *tbo) { struct amdgpu_device *adev = amdgpu_ttm_adev(tbo->bdev); uint64_t flags; - int r; if (!tbo->ttm) - return 0; + return; flags = amdgpu_ttm_tt_pte_flags(adev, tbo->ttm, tbo->resource); - r = amdgpu_ttm_gart_bind(adev, tbo, flags); - - return r; + amdgpu_ttm_gart_bind(adev, tbo, flags); } /* @@ -1025,8 +1011,7 @@ static void amdgpu_ttm_backend_unbind(struct ttm_device *bdev, struct ttm_tt *ttm) { struct amdgpu_device *adev = amdgpu_ttm_adev(bdev); - struct amdgpu_ttm_tt *gtt = (void *)ttm; - int r; + struct amdgpu_ttm_tt *gtt = ttm_to_amdgpu_ttm_tt(ttm); /* if the pages have userptr pinning then clear that first */ if (gtt->userptr) { @@ -1046,17 +1031,14 @@ static void amdgpu_ttm_backend_unbind(struct ttm_device *bdev, return; /* unbind shouldn't be done for GDS/GWS/OA in ttm_bo_clean_mm */ - r = amdgpu_gart_unbind(adev, gtt->offset, ttm->num_pages); - if (r) - DRM_ERROR("failed to unbind %u pages at 0x%08llX\n", - gtt->ttm.num_pages, gtt->offset); + amdgpu_gart_unbind(adev, gtt->offset, ttm->num_pages); gtt->bound = false; } static void amdgpu_ttm_backend_destroy(struct ttm_device *bdev, struct ttm_tt *ttm) { - struct amdgpu_ttm_tt *gtt = (void *)ttm; + struct amdgpu_ttm_tt *gtt = ttm_to_amdgpu_ttm_tt(ttm); if (gtt->usertask) put_task_struct(gtt->usertask); @@ -1110,7 +1092,7 @@ static int amdgpu_ttm_tt_populate(struct ttm_device *bdev, struct ttm_operation_ctx *ctx) { struct amdgpu_device *adev = amdgpu_ttm_adev(bdev); - struct amdgpu_ttm_tt *gtt = (void *)ttm; + struct amdgpu_ttm_tt *gtt = ttm_to_amdgpu_ttm_tt(ttm); pgoff_t i; int ret; @@ -1144,7 +1126,7 @@ static int amdgpu_ttm_tt_populate(struct ttm_device *bdev, static void amdgpu_ttm_tt_unpopulate(struct ttm_device *bdev, struct ttm_tt *ttm) { - struct amdgpu_ttm_tt *gtt = (void *)ttm; + struct amdgpu_ttm_tt *gtt = ttm_to_amdgpu_ttm_tt(ttm); struct amdgpu_device *adev; pgoff_t i; @@ -1168,6 +1150,26 @@ static void amdgpu_ttm_tt_unpopulate(struct ttm_device *bdev, } /** + * amdgpu_ttm_tt_get_userptr - Return the userptr GTT ttm_tt for the current + * task + * + * @tbo: The ttm_buffer_object that contains the userptr + * @user_addr: The returned value + */ +int amdgpu_ttm_tt_get_userptr(const struct ttm_buffer_object *tbo, + uint64_t *user_addr) +{ + struct amdgpu_ttm_tt *gtt; + + if (!tbo->ttm) + return -EINVAL; + + gtt = (void *)tbo->ttm; + *user_addr = gtt->userptr; + return 0; +} + +/** * amdgpu_ttm_tt_set_userptr - Initialize userptr GTT ttm_tt for the current * task * @@ -1193,7 +1195,7 @@ int amdgpu_ttm_tt_set_userptr(struct ttm_buffer_object *bo, /* Set TTM_TT_FLAG_EXTERNAL before populate but after create. */ bo->ttm->page_flags |= TTM_TT_FLAG_EXTERNAL; - gtt = (void *)bo->ttm; + gtt = ttm_to_amdgpu_ttm_tt(bo->ttm); gtt->userptr = addr; gtt->userflags = flags; @@ -1210,7 +1212,7 @@ int amdgpu_ttm_tt_set_userptr(struct ttm_buffer_object *bo, */ struct mm_struct *amdgpu_ttm_tt_get_usermm(struct ttm_tt *ttm) { - struct amdgpu_ttm_tt *gtt = (void *)ttm; + struct amdgpu_ttm_tt *gtt = ttm_to_amdgpu_ttm_tt(ttm); if (gtt == NULL) return NULL; @@ -1229,7 +1231,7 @@ struct mm_struct *amdgpu_ttm_tt_get_usermm(struct ttm_tt *ttm) bool amdgpu_ttm_tt_affect_userptr(struct ttm_tt *ttm, unsigned long start, unsigned long end, unsigned long *userptr) { - struct amdgpu_ttm_tt *gtt = (void *)ttm; + struct amdgpu_ttm_tt *gtt = ttm_to_amdgpu_ttm_tt(ttm); unsigned long size; if (gtt == NULL || !gtt->userptr) @@ -1252,7 +1254,7 @@ bool amdgpu_ttm_tt_affect_userptr(struct ttm_tt *ttm, unsigned long start, */ bool amdgpu_ttm_tt_is_userptr(struct ttm_tt *ttm) { - struct amdgpu_ttm_tt *gtt = (void *)ttm; + struct amdgpu_ttm_tt *gtt = ttm_to_amdgpu_ttm_tt(ttm); if (gtt == NULL || !gtt->userptr) return false; @@ -1265,7 +1267,7 @@ bool amdgpu_ttm_tt_is_userptr(struct ttm_tt *ttm) */ bool amdgpu_ttm_tt_is_readonly(struct ttm_tt *ttm) { - struct amdgpu_ttm_tt *gtt = (void *)ttm; + struct amdgpu_ttm_tt *gtt = ttm_to_amdgpu_ttm_tt(ttm); if (gtt == NULL) return false; @@ -1338,11 +1340,12 @@ uint64_t amdgpu_ttm_tt_pte_flags(struct amdgpu_device *adev, struct ttm_tt *ttm, static bool amdgpu_ttm_bo_eviction_valuable(struct ttm_buffer_object *bo, const struct ttm_place *place) { - unsigned long num_pages = bo->resource->num_pages; struct dma_resv_iter resv_cursor; - struct amdgpu_res_cursor cursor; struct dma_fence *f; + if (!amdgpu_bo_is_amdgpu_bo(bo)) + return ttm_bo_eviction_valuable(bo, place); + /* Swapout? */ if (bo->resource->mem_type == TTM_PL_SYSTEM) return true; @@ -1355,45 +1358,26 @@ static bool amdgpu_ttm_bo_eviction_valuable(struct ttm_buffer_object *bo, * If true, then return false as any KFD process needs all its BOs to * be resident to run successfully */ - dma_resv_for_each_fence(&resv_cursor, bo->base.resv, true, f) { + dma_resv_for_each_fence(&resv_cursor, bo->base.resv, + DMA_RESV_USAGE_BOOKKEEP, f) { if (amdkfd_fence_check_mm(f, current->mm)) return false; } - switch (bo->resource->mem_type) { - case AMDGPU_PL_PREEMPT: - /* Preemptible BOs don't own system resources managed by the - * driver (pages, VRAM, GART space). They point to resources - * owned by someone else (e.g. pageable memory in user mode - * or a DMABuf). They are used in a preemptible context so we - * can guarantee no deadlocks and good QoS in case of MMU - * notifiers or DMABuf move notifiers from the resource owner. - */ + /* Preemptible BOs don't own system resources managed by the + * driver (pages, VRAM, GART space). They point to resources + * owned by someone else (e.g. pageable memory in user mode + * or a DMABuf). They are used in a preemptible context so we + * can guarantee no deadlocks and good QoS in case of MMU + * notifiers or DMABuf move notifiers from the resource owner. + */ + if (bo->resource->mem_type == AMDGPU_PL_PREEMPT) return false; - case TTM_PL_TT: - if (amdgpu_bo_is_amdgpu_bo(bo) && - amdgpu_bo_encrypted(ttm_to_amdgpu_bo(bo))) - return false; - return true; - case TTM_PL_VRAM: - /* Check each drm MM node individually */ - amdgpu_res_first(bo->resource, 0, (u64)num_pages << PAGE_SHIFT, - &cursor); - while (cursor.remaining) { - if (place->fpfn < PFN_DOWN(cursor.start + cursor.size) - && !(place->lpfn && - place->lpfn <= PFN_DOWN(cursor.start))) - return true; - - amdgpu_res_next(&cursor, cursor.size); - } + if (bo->resource->mem_type == TTM_PL_TT && + amdgpu_bo_encrypted(ttm_to_amdgpu_bo(bo))) return false; - default: - break; - } - return ttm_bo_eviction_valuable(bo, place); } @@ -1432,6 +1416,63 @@ static void amdgpu_ttm_vram_mm_access(struct amdgpu_device *adev, loff_t pos, } } +static int amdgpu_ttm_access_memory_sdma(struct ttm_buffer_object *bo, + unsigned long offset, void *buf, int len, int write) +{ + struct amdgpu_bo *abo = ttm_to_amdgpu_bo(bo); + struct amdgpu_device *adev = amdgpu_ttm_adev(abo->tbo.bdev); + struct amdgpu_res_cursor src_mm; + struct amdgpu_job *job; + struct dma_fence *fence; + uint64_t src_addr, dst_addr; + unsigned int num_dw; + int r, idx; + + if (len != PAGE_SIZE) + return -EINVAL; + + if (!adev->mman.sdma_access_ptr) + return -EACCES; + + if (!drm_dev_enter(adev_to_drm(adev), &idx)) + return -ENODEV; + + if (write) + memcpy(adev->mman.sdma_access_ptr, buf, len); + + num_dw = ALIGN(adev->mman.buffer_funcs->copy_num_dw, 8); + r = amdgpu_job_alloc_with_ib(adev, num_dw * 4, AMDGPU_IB_POOL_DELAYED, &job); + if (r) + goto out; + + amdgpu_res_first(abo->tbo.resource, offset, len, &src_mm); + src_addr = amdgpu_ttm_domain_start(adev, bo->resource->mem_type) + src_mm.start; + dst_addr = amdgpu_bo_gpu_offset(adev->mman.sdma_access_bo); + if (write) + swap(src_addr, dst_addr); + + amdgpu_emit_copy_buffer(adev, &job->ibs[0], src_addr, dst_addr, PAGE_SIZE, false); + + amdgpu_ring_pad_ib(adev->mman.buffer_funcs_ring, &job->ibs[0]); + WARN_ON(job->ibs[0].length_dw > num_dw); + + r = amdgpu_job_submit(job, &adev->mman.entity, AMDGPU_FENCE_OWNER_UNDEFINED, &fence); + if (r) { + amdgpu_job_free(job); + goto out; + } + + if (!dma_fence_wait_timeout(fence, false, adev->sdma_timeout)) + r = -ETIMEDOUT; + dma_fence_put(fence); + + if (!(r || write)) + memcpy(buf, adev->mman.sdma_access_ptr, len); +out: + drm_dev_exit(idx); + return r; +} + /** * amdgpu_ttm_access_memory - Read or Write memory that backs a buffer object. * @@ -1456,6 +1497,10 @@ static int amdgpu_ttm_access_memory(struct ttm_buffer_object *bo, if (bo->resource->mem_type != TTM_PL_VRAM) return -EIO; + if (amdgpu_device_has_timeouts_enabled(adev) && + !amdgpu_ttm_access_memory_sdma(bo, offset, buf, len, write)) + return len; + amdgpu_res_first(bo->resource, offset, len, &cursor); while (cursor.remaining) { size_t count, size = cursor.size; @@ -1497,7 +1542,6 @@ static struct ttm_device_funcs amdgpu_bo_driver = { .io_mem_reserve = &amdgpu_ttm_io_mem_reserve, .io_mem_pfn = amdgpu_ttm_io_mem_pfn, .access_memory = &amdgpu_ttm_access_memory, - .del_from_lru_notify = &amdgpu_vm_del_from_lru_notify }; /* @@ -1748,18 +1792,26 @@ int amdgpu_ttm_init(struct amdgpu_device *adev) DRM_INFO("amdgpu: %uM of VRAM memory ready\n", (unsigned) (adev->gmc.real_vram_size / (1024 * 1024))); - /* Compute GTT size, either bsaed on 3/4th the size of RAM size + /* Compute GTT size, either based on 1/2 the size of RAM size * or whatever the user passed on module init */ if (amdgpu_gtt_size == -1) { struct sysinfo si; si_meminfo(&si); - gtt_size = min(max((AMDGPU_DEFAULT_GTT_SIZE_MB << 20), - adev->gmc.mc_vram_size), - ((uint64_t)si.totalram * si.mem_unit * 3/4)); - } - else + /* Certain GL unit tests for large textures can cause problems + * with the OOM killer since there is no way to link this memory + * to a process. This was originally mitigated (but not necessarily + * eliminated) by limiting the GTT size. The problem is this limit + * is often too low for many modern games so just make the limit 1/2 + * of system memory which aligns with TTM. The OOM accounting needs + * to be addressed, but we shouldn't prevent common 3D applications + * from being usable just to potentially mitigate that corner case. + */ + gtt_size = max((AMDGPU_DEFAULT_GTT_SIZE_MB << 20), + (u64)si.totalram * si.mem_unit / 2); + } else { gtt_size = (uint64_t)amdgpu_gtt_size << 20; + } /* Initialize GTT memory pool */ r = amdgpu_gtt_mgr_init(adev, gtt_size); @@ -1796,6 +1848,12 @@ int amdgpu_ttm_init(struct amdgpu_device *adev) return r; } + if (amdgpu_bo_create_kernel(adev, PAGE_SIZE, PAGE_SIZE, + AMDGPU_GEM_DOMAIN_GTT, + &adev->mman.sdma_access_bo, NULL, + &adev->mman.sdma_access_ptr)) + DRM_WARN("Debug VRAM access will use slowpath MM access\n"); + return 0; } @@ -1804,6 +1862,7 @@ int amdgpu_ttm_init(struct amdgpu_device *adev) */ void amdgpu_ttm_fini(struct amdgpu_device *adev) { + int idx; if (!adev->mman.initialized) return; @@ -1816,8 +1875,19 @@ void amdgpu_ttm_fini(struct amdgpu_device *adev) if (adev->mman.stolen_reserved_size) amdgpu_bo_free_kernel(&adev->mman.stolen_reserved_memory, NULL, NULL); + amdgpu_bo_free_kernel(&adev->mman.sdma_access_bo, NULL, + &adev->mman.sdma_access_ptr); amdgpu_ttm_fw_reserve_vram_fini(adev); + if (drm_dev_enter(adev_to_drm(adev), &idx)) { + + if (adev->mman.aper_base_kaddr) + iounmap(adev->mman.aper_base_kaddr); + adev->mman.aper_base_kaddr = NULL; + + drm_dev_exit(idx); + } + amdgpu_vram_mgr_fini(adev); amdgpu_gtt_mgr_fini(adev); amdgpu_preempt_mgr_fini(adev); @@ -1873,27 +1943,59 @@ void amdgpu_ttm_set_buffer_funcs_status(struct amdgpu_device *adev, bool enable) size = adev->gmc.real_vram_size; else size = adev->gmc.visible_vram_size; - man->size = size >> PAGE_SHIFT; + man->size = size; adev->mman.buffer_funcs_enabled = enable; } +static int amdgpu_ttm_prepare_job(struct amdgpu_device *adev, + bool direct_submit, + unsigned int num_dw, + struct dma_resv *resv, + bool vm_needs_flush, + struct amdgpu_job **job) +{ + enum amdgpu_ib_pool_type pool = direct_submit ? + AMDGPU_IB_POOL_DIRECT : + AMDGPU_IB_POOL_DELAYED; + int r; + + r = amdgpu_job_alloc_with_ib(adev, num_dw * 4, pool, job); + if (r) + return r; + + if (vm_needs_flush) { + (*job)->vm_pd_addr = amdgpu_gmc_pd_addr(adev->gmc.pdb0_bo ? + adev->gmc.pdb0_bo : + adev->gart.bo); + (*job)->vm_needs_flush = true; + } + if (resv) { + r = amdgpu_sync_resv(adev, &(*job)->sync, resv, + AMDGPU_SYNC_ALWAYS, + AMDGPU_FENCE_OWNER_UNDEFINED); + if (r) { + DRM_ERROR("sync failed (%d).\n", r); + amdgpu_job_free(*job); + return r; + } + } + return 0; +} + int amdgpu_copy_buffer(struct amdgpu_ring *ring, uint64_t src_offset, uint64_t dst_offset, uint32_t byte_count, struct dma_resv *resv, struct dma_fence **fence, bool direct_submit, bool vm_needs_flush, bool tmz) { - enum amdgpu_ib_pool_type pool = direct_submit ? AMDGPU_IB_POOL_DIRECT : - AMDGPU_IB_POOL_DELAYED; struct amdgpu_device *adev = ring->adev; + unsigned num_loops, num_dw; struct amdgpu_job *job; - uint32_t max_bytes; - unsigned num_loops, num_dw; unsigned i; int r; - if (direct_submit && !ring->sched.ready) { + if (!direct_submit && !ring->sched.ready) { DRM_ERROR("Trying to move memory with ring turned off.\n"); return -EINVAL; } @@ -1901,26 +2003,11 @@ int amdgpu_copy_buffer(struct amdgpu_ring *ring, uint64_t src_offset, max_bytes = adev->mman.buffer_funcs->copy_max_bytes; num_loops = DIV_ROUND_UP(byte_count, max_bytes); num_dw = ALIGN(num_loops * adev->mman.buffer_funcs->copy_num_dw, 8); - - r = amdgpu_job_alloc_with_ib(adev, num_dw * 4, pool, &job); + r = amdgpu_ttm_prepare_job(adev, direct_submit, num_dw, + resv, vm_needs_flush, &job); if (r) return r; - if (vm_needs_flush) { - job->vm_pd_addr = amdgpu_gmc_pd_addr(adev->gmc.pdb0_bo ? - adev->gmc.pdb0_bo : adev->gart.bo); - job->vm_needs_flush = true; - } - if (resv) { - r = amdgpu_sync_resv(adev, &job->sync, resv, - AMDGPU_SYNC_ALWAYS, - AMDGPU_FENCE_OWNER_UNDEFINED); - if (r) { - DRM_ERROR("sync failed (%d).\n", r); - goto error_free; - } - } - for (i = 0; i < num_loops; i++) { uint32_t cur_size_in_bytes = min(byte_count, max_bytes); @@ -1950,77 +2037,35 @@ error_free: return r; } -int amdgpu_fill_buffer(struct amdgpu_bo *bo, - uint32_t src_data, - struct dma_resv *resv, - struct dma_fence **fence) +static int amdgpu_ttm_fill_mem(struct amdgpu_ring *ring, uint32_t src_data, + uint64_t dst_addr, uint32_t byte_count, + struct dma_resv *resv, + struct dma_fence **fence, + bool vm_needs_flush) { - struct amdgpu_device *adev = amdgpu_ttm_adev(bo->tbo.bdev); - uint32_t max_bytes = adev->mman.buffer_funcs->fill_max_bytes; - struct amdgpu_ring *ring = adev->mman.buffer_funcs_ring; - - struct amdgpu_res_cursor cursor; + struct amdgpu_device *adev = ring->adev; unsigned int num_loops, num_dw; - uint64_t num_bytes; - struct amdgpu_job *job; + uint32_t max_bytes; + unsigned int i; int r; - if (!adev->mman.buffer_funcs_enabled) { - DRM_ERROR("Trying to clear memory with ring turned off.\n"); - return -EINVAL; - } - - if (bo->tbo.resource->mem_type == AMDGPU_PL_PREEMPT) { - DRM_ERROR("Trying to clear preemptible memory.\n"); - return -EINVAL; - } - - if (bo->tbo.resource->mem_type == TTM_PL_TT) { - r = amdgpu_ttm_alloc_gart(&bo->tbo); - if (r) - return r; - } - - num_bytes = bo->tbo.resource->num_pages << PAGE_SHIFT; - num_loops = 0; - - amdgpu_res_first(bo->tbo.resource, 0, num_bytes, &cursor); - while (cursor.remaining) { - num_loops += DIV_ROUND_UP_ULL(cursor.size, max_bytes); - amdgpu_res_next(&cursor, cursor.size); - } - num_dw = num_loops * adev->mman.buffer_funcs->fill_num_dw; - - /* for IB padding */ - num_dw += 64; - - r = amdgpu_job_alloc_with_ib(adev, num_dw * 4, AMDGPU_IB_POOL_DELAYED, - &job); + max_bytes = adev->mman.buffer_funcs->fill_max_bytes; + num_loops = DIV_ROUND_UP_ULL(byte_count, max_bytes); + num_dw = ALIGN(num_loops * adev->mman.buffer_funcs->fill_num_dw, 8); + r = amdgpu_ttm_prepare_job(adev, false, num_dw, resv, vm_needs_flush, + &job); if (r) return r; - if (resv) { - r = amdgpu_sync_resv(adev, &job->sync, resv, - AMDGPU_SYNC_ALWAYS, - AMDGPU_FENCE_OWNER_UNDEFINED); - if (r) { - DRM_ERROR("sync failed (%d).\n", r); - goto error_free; - } - } - - amdgpu_res_first(bo->tbo.resource, 0, num_bytes, &cursor); - while (cursor.remaining) { - uint32_t cur_size = min_t(uint64_t, cursor.size, max_bytes); - uint64_t dst_addr = cursor.start; + for (i = 0; i < num_loops; i++) { + uint32_t cur_size = min(byte_count, max_bytes); - dst_addr += amdgpu_ttm_domain_start(adev, - bo->tbo.resource->mem_type); amdgpu_emit_fill_buffer(adev, &job->ibs[0], src_data, dst_addr, cur_size); - amdgpu_res_next(&cursor, cur_size); + dst_addr += cur_size; + byte_count -= cur_size; } amdgpu_ring_pad_ib(ring, &job->ibs[0]); @@ -2037,6 +2082,55 @@ error_free: return r; } +int amdgpu_fill_buffer(struct amdgpu_bo *bo, + uint32_t src_data, + struct dma_resv *resv, + struct dma_fence **f) +{ + struct amdgpu_device *adev = amdgpu_ttm_adev(bo->tbo.bdev); + struct amdgpu_ring *ring = adev->mman.buffer_funcs_ring; + struct dma_fence *fence = NULL; + struct amdgpu_res_cursor dst; + int r; + + if (!adev->mman.buffer_funcs_enabled) { + DRM_ERROR("Trying to clear memory with ring turned off.\n"); + return -EINVAL; + } + + amdgpu_res_first(bo->tbo.resource, 0, amdgpu_bo_size(bo), &dst); + + mutex_lock(&adev->mman.gtt_window_lock); + while (dst.remaining) { + struct dma_fence *next; + uint64_t cur_size, to; + + /* Never fill more than 256MiB at once to avoid timeouts */ + cur_size = min(dst.size, 256ULL << 20); + + r = amdgpu_ttm_map_buffer(&bo->tbo, bo->tbo.resource, &dst, + 1, ring, false, &cur_size, &to); + if (r) + goto error; + + r = amdgpu_ttm_fill_mem(ring, src_data, to, cur_size, resv, + &next, true); + if (r) + goto error; + + dma_fence_put(fence); + fence = next; + + amdgpu_res_next(&dst, cur_size); + } +error: + mutex_unlock(&adev->mman.gtt_window_lock); + if (f) + *f = dma_fence_get(fence); + dma_fence_put(fence); + return r; +} + /** * amdgpu_ttm_evict_resources - evict memory buffers * @adev: amdgpu device object @@ -2069,17 +2163,6 @@ int amdgpu_ttm_evict_resources(struct amdgpu_device *adev, int mem_type) #if defined(CONFIG_DEBUG_FS) -static int amdgpu_mm_vram_table_show(struct seq_file *m, void *unused) -{ - struct amdgpu_device *adev = (struct amdgpu_device *)m->private; - struct ttm_resource_manager *man = ttm_manager_type(&adev->mman.bdev, - TTM_PL_VRAM); - struct drm_printer p = drm_seq_file_printer(m); - - man->func->debug(man, &p); - return 0; -} - static int amdgpu_ttm_page_pool_show(struct seq_file *m, void *unused) { struct amdgpu_device *adev = (struct amdgpu_device *)m->private; @@ -2087,55 +2170,6 @@ static int amdgpu_ttm_page_pool_show(struct seq_file *m, void *unused) return ttm_pool_debugfs(&adev->mman.bdev.pool, m); } -static int amdgpu_mm_tt_table_show(struct seq_file *m, void *unused) -{ - struct amdgpu_device *adev = (struct amdgpu_device *)m->private; - struct ttm_resource_manager *man = ttm_manager_type(&adev->mman.bdev, - TTM_PL_TT); - struct drm_printer p = drm_seq_file_printer(m); - - man->func->debug(man, &p); - return 0; -} - -static int amdgpu_mm_gds_table_show(struct seq_file *m, void *unused) -{ - struct amdgpu_device *adev = (struct amdgpu_device *)m->private; - struct ttm_resource_manager *man = ttm_manager_type(&adev->mman.bdev, - AMDGPU_PL_GDS); - struct drm_printer p = drm_seq_file_printer(m); - - man->func->debug(man, &p); - return 0; -} - -static int amdgpu_mm_gws_table_show(struct seq_file *m, void *unused) -{ - struct amdgpu_device *adev = (struct amdgpu_device *)m->private; - struct ttm_resource_manager *man = ttm_manager_type(&adev->mman.bdev, - AMDGPU_PL_GWS); - struct drm_printer p = drm_seq_file_printer(m); - - man->func->debug(man, &p); - return 0; -} - -static int amdgpu_mm_oa_table_show(struct seq_file *m, void *unused) -{ - struct amdgpu_device *adev = (struct amdgpu_device *)m->private; - struct ttm_resource_manager *man = ttm_manager_type(&adev->mman.bdev, - AMDGPU_PL_OA); - struct drm_printer p = drm_seq_file_printer(m); - - man->func->debug(man, &p); - return 0; -} - -DEFINE_SHOW_ATTRIBUTE(amdgpu_mm_vram_table); -DEFINE_SHOW_ATTRIBUTE(amdgpu_mm_tt_table); -DEFINE_SHOW_ATTRIBUTE(amdgpu_mm_gds_table); -DEFINE_SHOW_ATTRIBUTE(amdgpu_mm_gws_table); -DEFINE_SHOW_ATTRIBUTE(amdgpu_mm_oa_table); DEFINE_SHOW_ATTRIBUTE(amdgpu_ttm_page_pool); /* @@ -2345,17 +2379,23 @@ void amdgpu_ttm_debugfs_init(struct amdgpu_device *adev) &amdgpu_ttm_vram_fops, adev->gmc.mc_vram_size); debugfs_create_file("amdgpu_iomem", 0444, root, adev, &amdgpu_ttm_iomem_fops); - debugfs_create_file("amdgpu_vram_mm", 0444, root, adev, - &amdgpu_mm_vram_table_fops); - debugfs_create_file("amdgpu_gtt_mm", 0444, root, adev, - &amdgpu_mm_tt_table_fops); - debugfs_create_file("amdgpu_gds_mm", 0444, root, adev, - &amdgpu_mm_gds_table_fops); - debugfs_create_file("amdgpu_gws_mm", 0444, root, adev, - &amdgpu_mm_gws_table_fops); - debugfs_create_file("amdgpu_oa_mm", 0444, root, adev, - &amdgpu_mm_oa_table_fops); debugfs_create_file("ttm_page_pool", 0444, root, adev, &amdgpu_ttm_page_pool_fops); + ttm_resource_manager_create_debugfs(ttm_manager_type(&adev->mman.bdev, + TTM_PL_VRAM), + root, "amdgpu_vram_mm"); + ttm_resource_manager_create_debugfs(ttm_manager_type(&adev->mman.bdev, + TTM_PL_TT), + root, "amdgpu_gtt_mm"); + ttm_resource_manager_create_debugfs(ttm_manager_type(&adev->mman.bdev, + AMDGPU_PL_GDS), + root, "amdgpu_gds_mm"); + ttm_resource_manager_create_debugfs(ttm_manager_type(&adev->mman.bdev, + AMDGPU_PL_GWS), + root, "amdgpu_gws_mm"); + ttm_resource_manager_create_debugfs(ttm_manager_type(&adev->mman.bdev, + AMDGPU_PL_OA), + root, "amdgpu_oa_mm"); + #endif } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.h index 7346ecff4438..6a70818039dd 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.h @@ -26,6 +26,7 @@ #include <linux/dma-direction.h> #include <drm/gpu_scheduler.h> +#include "amdgpu_vram_mgr.h" #include "amdgpu.h" #define AMDGPU_PL_GDS (TTM_PL_PRIV + 0) @@ -38,26 +39,10 @@ #define AMDGPU_POISON 0xd0bed0be -struct amdgpu_vram_mgr { - struct ttm_resource_manager manager; - struct drm_mm mm; - spinlock_t lock; - struct list_head reservations_pending; - struct list_head reserved_pages; - atomic64_t usage; - atomic64_t vis_usage; -}; - struct amdgpu_gtt_mgr { struct ttm_resource_manager manager; struct drm_mm mm; spinlock_t lock; - atomic64_t used; -}; - -struct amdgpu_preempt_mgr { - struct ttm_resource_manager manager; - atomic64_t used; }; struct amdgpu_mman { @@ -76,7 +61,7 @@ struct amdgpu_mman { struct amdgpu_vram_mgr vram_mgr; struct amdgpu_gtt_mgr gtt_mgr; - struct amdgpu_preempt_mgr preempt_mgr; + struct ttm_resource_manager preempt_mgr; uint64_t stolen_vga_size; struct amdgpu_bo *stolen_vga_memory; @@ -98,6 +83,10 @@ struct amdgpu_mman { u64 fw_vram_usage_size; struct amdgpu_bo *fw_vram_usage_reserved_bo; void *fw_vram_usage_va; + + /* PAGE_SIZE'd BO for process memory r/w over SDMA. */ + struct amdgpu_bo *sdma_access_bo; + void *sdma_access_ptr; }; struct amdgpu_copy_mem { @@ -114,8 +103,7 @@ int amdgpu_vram_mgr_init(struct amdgpu_device *adev); void amdgpu_vram_mgr_fini(struct amdgpu_device *adev); bool amdgpu_gtt_mgr_has_gart_addr(struct ttm_resource *mem); -uint64_t amdgpu_gtt_mgr_usage(struct ttm_resource_manager *man); -int amdgpu_gtt_mgr_recover(struct ttm_resource_manager *man); +void amdgpu_gtt_mgr_recover(struct amdgpu_gtt_mgr *mgr); uint64_t amdgpu_preempt_mgr_usage(struct ttm_resource_manager *man); @@ -129,11 +117,10 @@ int amdgpu_vram_mgr_alloc_sgt(struct amdgpu_device *adev, void amdgpu_vram_mgr_free_sgt(struct device *dev, enum dma_data_direction dir, struct sg_table *sgt); -uint64_t amdgpu_vram_mgr_usage(struct ttm_resource_manager *man); -uint64_t amdgpu_vram_mgr_vis_usage(struct ttm_resource_manager *man); -int amdgpu_vram_mgr_reserve_range(struct ttm_resource_manager *man, +uint64_t amdgpu_vram_mgr_vis_usage(struct amdgpu_vram_mgr *mgr); +int amdgpu_vram_mgr_reserve_range(struct amdgpu_vram_mgr *mgr, uint64_t start, uint64_t size); -int amdgpu_vram_mgr_query_page_status(struct ttm_resource_manager *man, +int amdgpu_vram_mgr_query_page_status(struct amdgpu_vram_mgr *mgr, uint64_t start); int amdgpu_ttm_init(struct amdgpu_device *adev); @@ -158,7 +145,7 @@ int amdgpu_fill_buffer(struct amdgpu_bo *bo, struct dma_fence **fence); int amdgpu_ttm_alloc_gart(struct ttm_buffer_object *bo); -int amdgpu_ttm_recover_gart(struct ttm_buffer_object *tbo); +void amdgpu_ttm_recover_gart(struct ttm_buffer_object *tbo); uint64_t amdgpu_ttm_domain_start(struct amdgpu_device *adev, uint32_t type); #if IS_ENABLED(CONFIG_DRM_AMDGPU_USERPTR) @@ -177,6 +164,8 @@ static inline bool amdgpu_ttm_tt_get_user_pages_done(struct ttm_tt *ttm) #endif void amdgpu_ttm_tt_set_user_pages(struct ttm_tt *ttm, struct page **pages); +int amdgpu_ttm_tt_get_userptr(const struct ttm_buffer_object *tbo, + uint64_t *user_addr); int amdgpu_ttm_tt_set_userptr(struct ttm_buffer_object *bo, uint64_t addr, uint32_t flags); bool amdgpu_ttm_tt_has_userptr(struct ttm_tt *ttm); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.c index ca3350502618..5cb62e6249c2 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.c @@ -115,11 +115,30 @@ void amdgpu_ucode_print_gfx_hdr(const struct common_firmware_header *hdr) le32_to_cpu(gfx_hdr->ucode_feature_version)); DRM_DEBUG("jt_offset: %u\n", le32_to_cpu(gfx_hdr->jt_offset)); DRM_DEBUG("jt_size: %u\n", le32_to_cpu(gfx_hdr->jt_size)); + } else if (version_major == 2) { + const struct gfx_firmware_header_v2_0 *gfx_hdr = + container_of(hdr, struct gfx_firmware_header_v2_0, header); + + DRM_DEBUG("ucode_feature_version: %u\n", + le32_to_cpu(gfx_hdr->ucode_feature_version)); } else { DRM_ERROR("Unknown GFX ucode version: %u.%u\n", version_major, version_minor); } } +void amdgpu_ucode_print_imu_hdr(const struct common_firmware_header *hdr) +{ + uint16_t version_major = le16_to_cpu(hdr->header_version_major); + uint16_t version_minor = le16_to_cpu(hdr->header_version_minor); + + DRM_DEBUG("IMU\n"); + amdgpu_ucode_print_common_hdr(hdr); + + if (version_major != 1) { + DRM_ERROR("Unknown GFX ucode version: %u.%u\n", version_major, version_minor); + } +} + void amdgpu_ucode_print_rlc_hdr(const struct common_firmware_header *hdr) { uint16_t version_major = le16_to_cpu(hdr->header_version_major); @@ -145,70 +164,138 @@ void amdgpu_ucode_print_rlc_hdr(const struct common_firmware_header *hdr) } else if (version_major == 2) { const struct rlc_firmware_header_v2_0 *rlc_hdr = container_of(hdr, struct rlc_firmware_header_v2_0, header); + const struct rlc_firmware_header_v2_1 *rlc_hdr_v2_1 = + container_of(rlc_hdr, struct rlc_firmware_header_v2_1, v2_0); + const struct rlc_firmware_header_v2_2 *rlc_hdr_v2_2 = + container_of(rlc_hdr_v2_1, struct rlc_firmware_header_v2_2, v2_1); + const struct rlc_firmware_header_v2_3 *rlc_hdr_v2_3 = + container_of(rlc_hdr_v2_2, struct rlc_firmware_header_v2_3, v2_2); + const struct rlc_firmware_header_v2_4 *rlc_hdr_v2_4 = + container_of(rlc_hdr_v2_3, struct rlc_firmware_header_v2_4, v2_3); - DRM_DEBUG("ucode_feature_version: %u\n", - le32_to_cpu(rlc_hdr->ucode_feature_version)); - DRM_DEBUG("jt_offset: %u\n", le32_to_cpu(rlc_hdr->jt_offset)); - DRM_DEBUG("jt_size: %u\n", le32_to_cpu(rlc_hdr->jt_size)); - DRM_DEBUG("save_and_restore_offset: %u\n", - le32_to_cpu(rlc_hdr->save_and_restore_offset)); - DRM_DEBUG("clear_state_descriptor_offset: %u\n", - le32_to_cpu(rlc_hdr->clear_state_descriptor_offset)); - DRM_DEBUG("avail_scratch_ram_locations: %u\n", - le32_to_cpu(rlc_hdr->avail_scratch_ram_locations)); - DRM_DEBUG("reg_restore_list_size: %u\n", - le32_to_cpu(rlc_hdr->reg_restore_list_size)); - DRM_DEBUG("reg_list_format_start: %u\n", - le32_to_cpu(rlc_hdr->reg_list_format_start)); - DRM_DEBUG("reg_list_format_separate_start: %u\n", - le32_to_cpu(rlc_hdr->reg_list_format_separate_start)); - DRM_DEBUG("starting_offsets_start: %u\n", - le32_to_cpu(rlc_hdr->starting_offsets_start)); - DRM_DEBUG("reg_list_format_size_bytes: %u\n", - le32_to_cpu(rlc_hdr->reg_list_format_size_bytes)); - DRM_DEBUG("reg_list_format_array_offset_bytes: %u\n", - le32_to_cpu(rlc_hdr->reg_list_format_array_offset_bytes)); - DRM_DEBUG("reg_list_size_bytes: %u\n", - le32_to_cpu(rlc_hdr->reg_list_size_bytes)); - DRM_DEBUG("reg_list_array_offset_bytes: %u\n", - le32_to_cpu(rlc_hdr->reg_list_array_offset_bytes)); - DRM_DEBUG("reg_list_format_separate_size_bytes: %u\n", - le32_to_cpu(rlc_hdr->reg_list_format_separate_size_bytes)); - DRM_DEBUG("reg_list_format_separate_array_offset_bytes: %u\n", - le32_to_cpu(rlc_hdr->reg_list_format_separate_array_offset_bytes)); - DRM_DEBUG("reg_list_separate_size_bytes: %u\n", - le32_to_cpu(rlc_hdr->reg_list_separate_size_bytes)); - DRM_DEBUG("reg_list_separate_array_offset_bytes: %u\n", - le32_to_cpu(rlc_hdr->reg_list_separate_array_offset_bytes)); - if (version_minor == 1) { - const struct rlc_firmware_header_v2_1 *v2_1 = - container_of(rlc_hdr, struct rlc_firmware_header_v2_1, v2_0); + switch (version_minor) { + case 0: + /* rlc_hdr v2_0 */ + DRM_DEBUG("ucode_feature_version: %u\n", + le32_to_cpu(rlc_hdr->ucode_feature_version)); + DRM_DEBUG("jt_offset: %u\n", le32_to_cpu(rlc_hdr->jt_offset)); + DRM_DEBUG("jt_size: %u\n", le32_to_cpu(rlc_hdr->jt_size)); + DRM_DEBUG("save_and_restore_offset: %u\n", + le32_to_cpu(rlc_hdr->save_and_restore_offset)); + DRM_DEBUG("clear_state_descriptor_offset: %u\n", + le32_to_cpu(rlc_hdr->clear_state_descriptor_offset)); + DRM_DEBUG("avail_scratch_ram_locations: %u\n", + le32_to_cpu(rlc_hdr->avail_scratch_ram_locations)); + DRM_DEBUG("reg_restore_list_size: %u\n", + le32_to_cpu(rlc_hdr->reg_restore_list_size)); + DRM_DEBUG("reg_list_format_start: %u\n", + le32_to_cpu(rlc_hdr->reg_list_format_start)); + DRM_DEBUG("reg_list_format_separate_start: %u\n", + le32_to_cpu(rlc_hdr->reg_list_format_separate_start)); + DRM_DEBUG("starting_offsets_start: %u\n", + le32_to_cpu(rlc_hdr->starting_offsets_start)); + DRM_DEBUG("reg_list_format_size_bytes: %u\n", + le32_to_cpu(rlc_hdr->reg_list_format_size_bytes)); + DRM_DEBUG("reg_list_format_array_offset_bytes: %u\n", + le32_to_cpu(rlc_hdr->reg_list_format_array_offset_bytes)); + DRM_DEBUG("reg_list_size_bytes: %u\n", + le32_to_cpu(rlc_hdr->reg_list_size_bytes)); + DRM_DEBUG("reg_list_array_offset_bytes: %u\n", + le32_to_cpu(rlc_hdr->reg_list_array_offset_bytes)); + DRM_DEBUG("reg_list_format_separate_size_bytes: %u\n", + le32_to_cpu(rlc_hdr->reg_list_format_separate_size_bytes)); + DRM_DEBUG("reg_list_format_separate_array_offset_bytes: %u\n", + le32_to_cpu(rlc_hdr->reg_list_format_separate_array_offset_bytes)); + DRM_DEBUG("reg_list_separate_size_bytes: %u\n", + le32_to_cpu(rlc_hdr->reg_list_separate_size_bytes)); + DRM_DEBUG("reg_list_separate_array_offset_bytes: %u\n", + le32_to_cpu(rlc_hdr->reg_list_separate_array_offset_bytes)); + break; + case 1: + /* rlc_hdr v2_1 */ DRM_DEBUG("reg_list_format_direct_reg_list_length: %u\n", - le32_to_cpu(v2_1->reg_list_format_direct_reg_list_length)); + le32_to_cpu(rlc_hdr_v2_1->reg_list_format_direct_reg_list_length)); DRM_DEBUG("save_restore_list_cntl_ucode_ver: %u\n", - le32_to_cpu(v2_1->save_restore_list_cntl_ucode_ver)); + le32_to_cpu(rlc_hdr_v2_1->save_restore_list_cntl_ucode_ver)); DRM_DEBUG("save_restore_list_cntl_feature_ver: %u\n", - le32_to_cpu(v2_1->save_restore_list_cntl_feature_ver)); + le32_to_cpu(rlc_hdr_v2_1->save_restore_list_cntl_feature_ver)); DRM_DEBUG("save_restore_list_cntl_size_bytes %u\n", - le32_to_cpu(v2_1->save_restore_list_cntl_size_bytes)); + le32_to_cpu(rlc_hdr_v2_1->save_restore_list_cntl_size_bytes)); DRM_DEBUG("save_restore_list_cntl_offset_bytes: %u\n", - le32_to_cpu(v2_1->save_restore_list_cntl_offset_bytes)); + le32_to_cpu(rlc_hdr_v2_1->save_restore_list_cntl_offset_bytes)); DRM_DEBUG("save_restore_list_gpm_ucode_ver: %u\n", - le32_to_cpu(v2_1->save_restore_list_gpm_ucode_ver)); + le32_to_cpu(rlc_hdr_v2_1->save_restore_list_gpm_ucode_ver)); DRM_DEBUG("save_restore_list_gpm_feature_ver: %u\n", - le32_to_cpu(v2_1->save_restore_list_gpm_feature_ver)); + le32_to_cpu(rlc_hdr_v2_1->save_restore_list_gpm_feature_ver)); DRM_DEBUG("save_restore_list_gpm_size_bytes %u\n", - le32_to_cpu(v2_1->save_restore_list_gpm_size_bytes)); + le32_to_cpu(rlc_hdr_v2_1->save_restore_list_gpm_size_bytes)); DRM_DEBUG("save_restore_list_gpm_offset_bytes: %u\n", - le32_to_cpu(v2_1->save_restore_list_gpm_offset_bytes)); + le32_to_cpu(rlc_hdr_v2_1->save_restore_list_gpm_offset_bytes)); DRM_DEBUG("save_restore_list_srm_ucode_ver: %u\n", - le32_to_cpu(v2_1->save_restore_list_srm_ucode_ver)); + le32_to_cpu(rlc_hdr_v2_1->save_restore_list_srm_ucode_ver)); DRM_DEBUG("save_restore_list_srm_feature_ver: %u\n", - le32_to_cpu(v2_1->save_restore_list_srm_feature_ver)); + le32_to_cpu(rlc_hdr_v2_1->save_restore_list_srm_feature_ver)); DRM_DEBUG("save_restore_list_srm_size_bytes %u\n", - le32_to_cpu(v2_1->save_restore_list_srm_size_bytes)); + le32_to_cpu(rlc_hdr_v2_1->save_restore_list_srm_size_bytes)); DRM_DEBUG("save_restore_list_srm_offset_bytes: %u\n", - le32_to_cpu(v2_1->save_restore_list_srm_offset_bytes)); + le32_to_cpu(rlc_hdr_v2_1->save_restore_list_srm_offset_bytes)); + break; + case 2: + /* rlc_hdr v2_2 */ + DRM_DEBUG("rlc_iram_ucode_size_bytes: %u\n", + le32_to_cpu(rlc_hdr_v2_2->rlc_iram_ucode_size_bytes)); + DRM_DEBUG("rlc_iram_ucode_offset_bytes: %u\n", + le32_to_cpu(rlc_hdr_v2_2->rlc_iram_ucode_offset_bytes)); + DRM_DEBUG("rlc_dram_ucode_size_bytes: %u\n", + le32_to_cpu(rlc_hdr_v2_2->rlc_dram_ucode_size_bytes)); + DRM_DEBUG("rlc_dram_ucode_offset_bytes: %u\n", + le32_to_cpu(rlc_hdr_v2_2->rlc_dram_ucode_offset_bytes)); + break; + case 3: + /* rlc_hdr v2_3 */ + DRM_DEBUG("rlcp_ucode_version: %u\n", + le32_to_cpu(rlc_hdr_v2_3->rlcp_ucode_version)); + DRM_DEBUG("rlcp_ucode_feature_version: %u\n", + le32_to_cpu(rlc_hdr_v2_3->rlcp_ucode_feature_version)); + DRM_DEBUG("rlcp_ucode_size_bytes: %u\n", + le32_to_cpu(rlc_hdr_v2_3->rlcp_ucode_size_bytes)); + DRM_DEBUG("rlcp_ucode_offset_bytes: %u\n", + le32_to_cpu(rlc_hdr_v2_3->rlcp_ucode_offset_bytes)); + DRM_DEBUG("rlcv_ucode_version: %u\n", + le32_to_cpu(rlc_hdr_v2_3->rlcv_ucode_version)); + DRM_DEBUG("rlcv_ucode_feature_version: %u\n", + le32_to_cpu(rlc_hdr_v2_3->rlcv_ucode_feature_version)); + DRM_DEBUG("rlcv_ucode_size_bytes: %u\n", + le32_to_cpu(rlc_hdr_v2_3->rlcv_ucode_size_bytes)); + DRM_DEBUG("rlcv_ucode_offset_bytes: %u\n", + le32_to_cpu(rlc_hdr_v2_3->rlcv_ucode_offset_bytes)); + break; + case 4: + /* rlc_hdr v2_4 */ + DRM_DEBUG("global_tap_delays_ucode_size_bytes :%u\n", + le32_to_cpu(rlc_hdr_v2_4->global_tap_delays_ucode_size_bytes)); + DRM_DEBUG("global_tap_delays_ucode_offset_bytes: %u\n", + le32_to_cpu(rlc_hdr_v2_4->global_tap_delays_ucode_offset_bytes)); + DRM_DEBUG("se0_tap_delays_ucode_size_bytes :%u\n", + le32_to_cpu(rlc_hdr_v2_4->se0_tap_delays_ucode_size_bytes)); + DRM_DEBUG("se0_tap_delays_ucode_offset_bytes: %u\n", + le32_to_cpu(rlc_hdr_v2_4->se0_tap_delays_ucode_offset_bytes)); + DRM_DEBUG("se1_tap_delays_ucode_size_bytes :%u\n", + le32_to_cpu(rlc_hdr_v2_4->se1_tap_delays_ucode_size_bytes)); + DRM_DEBUG("se1_tap_delays_ucode_offset_bytes: %u\n", + le32_to_cpu(rlc_hdr_v2_4->se1_tap_delays_ucode_offset_bytes)); + DRM_DEBUG("se2_tap_delays_ucode_size_bytes :%u\n", + le32_to_cpu(rlc_hdr_v2_4->se2_tap_delays_ucode_size_bytes)); + DRM_DEBUG("se2_tap_delays_ucode_offset_bytes: %u\n", + le32_to_cpu(rlc_hdr_v2_4->se2_tap_delays_ucode_offset_bytes)); + DRM_DEBUG("se3_tap_delays_ucode_size_bytes :%u\n", + le32_to_cpu(rlc_hdr_v2_4->se3_tap_delays_ucode_size_bytes)); + DRM_DEBUG("se3_tap_delays_ucode_offset_bytes: %u\n", + le32_to_cpu(rlc_hdr_v2_4->se3_tap_delays_ucode_offset_bytes)); + break; + default: + DRM_ERROR("Unknown RLC v2 ucode: v2.%u\n", version_minor); + break; } } else { DRM_ERROR("Unknown RLC ucode version: %u.%u\n", version_major, version_minor); @@ -238,6 +325,17 @@ void amdgpu_ucode_print_sdma_hdr(const struct common_firmware_header *hdr) container_of(sdma_hdr, struct sdma_firmware_header_v1_1, v1_0); DRM_DEBUG("digest_size: %u\n", le32_to_cpu(sdma_v1_1_hdr->digest_size)); } + } else if (version_major == 2) { + const struct sdma_firmware_header_v2_0 *sdma_hdr = + container_of(hdr, struct sdma_firmware_header_v2_0, header); + + DRM_DEBUG("ucode_feature_version: %u\n", + le32_to_cpu(sdma_hdr->ucode_feature_version)); + DRM_DEBUG("ctx_jt_offset: %u\n", le32_to_cpu(sdma_hdr->ctx_jt_offset)); + DRM_DEBUG("ctx_jt_size: %u\n", le32_to_cpu(sdma_hdr->ctx_jt_size)); + DRM_DEBUG("ctl_ucode_offset: %u\n", le32_to_cpu(sdma_hdr->ctl_ucode_offset)); + DRM_DEBUG("ctl_jt_offset: %u\n", le32_to_cpu(sdma_hdr->ctl_jt_offset)); + DRM_DEBUG("ctl_jt_size: %u\n", le32_to_cpu(sdma_hdr->ctl_jt_size)); } else { DRM_ERROR("Unknown SDMA ucode version: %u.%u\n", version_major, version_minor); @@ -248,6 +346,8 @@ void amdgpu_ucode_print_psp_hdr(const struct common_firmware_header *hdr) { uint16_t version_major = le16_to_cpu(hdr->header_version_major); uint16_t version_minor = le16_to_cpu(hdr->header_version_minor); + uint32_t fw_index; + const struct psp_fw_bin_desc *desc; DRM_DEBUG("PSP\n"); amdgpu_ucode_print_common_hdr(hdr); @@ -312,6 +412,71 @@ void amdgpu_ucode_print_psp_hdr(const struct common_firmware_header *hdr) DRM_DEBUG("spl_size_bytes: %u\n", le32_to_cpu(psp_hdr_v1_3->spl.size_bytes)); } + } else if (version_major == 2) { + const struct psp_firmware_header_v2_0 *psp_hdr_v2_0 = + container_of(hdr, struct psp_firmware_header_v2_0, header); + for (fw_index = 0; fw_index < le32_to_cpu(psp_hdr_v2_0->psp_fw_bin_count); fw_index++) { + desc = &(psp_hdr_v2_0->psp_fw_bin[fw_index]); + switch (desc->fw_type) { + case PSP_FW_TYPE_PSP_SOS: + DRM_DEBUG("psp_sos_version: %u\n", + le32_to_cpu(desc->fw_version)); + DRM_DEBUG("psp_sos_size_bytes: %u\n", + le32_to_cpu(desc->size_bytes)); + break; + case PSP_FW_TYPE_PSP_SYS_DRV: + DRM_DEBUG("psp_sys_drv_version: %u\n", + le32_to_cpu(desc->fw_version)); + DRM_DEBUG("psp_sys_drv_size_bytes: %u\n", + le32_to_cpu(desc->size_bytes)); + break; + case PSP_FW_TYPE_PSP_KDB: + DRM_DEBUG("psp_kdb_version: %u\n", + le32_to_cpu(desc->fw_version)); + DRM_DEBUG("psp_kdb_size_bytes: %u\n", + le32_to_cpu(desc->size_bytes)); + break; + case PSP_FW_TYPE_PSP_TOC: + DRM_DEBUG("psp_toc_version: %u\n", + le32_to_cpu(desc->fw_version)); + DRM_DEBUG("psp_toc_size_bytes: %u\n", + le32_to_cpu(desc->size_bytes)); + break; + case PSP_FW_TYPE_PSP_SPL: + DRM_DEBUG("psp_spl_version: %u\n", + le32_to_cpu(desc->fw_version)); + DRM_DEBUG("psp_spl_size_bytes: %u\n", + le32_to_cpu(desc->size_bytes)); + break; + case PSP_FW_TYPE_PSP_RL: + DRM_DEBUG("psp_rl_version: %u\n", + le32_to_cpu(desc->fw_version)); + DRM_DEBUG("psp_rl_size_bytes: %u\n", + le32_to_cpu(desc->size_bytes)); + break; + case PSP_FW_TYPE_PSP_SOC_DRV: + DRM_DEBUG("psp_soc_drv_version: %u\n", + le32_to_cpu(desc->fw_version)); + DRM_DEBUG("psp_soc_drv_size_bytes: %u\n", + le32_to_cpu(desc->size_bytes)); + break; + case PSP_FW_TYPE_PSP_INTF_DRV: + DRM_DEBUG("psp_intf_drv_version: %u\n", + le32_to_cpu(desc->fw_version)); + DRM_DEBUG("psp_intf_drv_size_bytes: %u\n", + le32_to_cpu(desc->size_bytes)); + break; + case PSP_FW_TYPE_PSP_DBG_DRV: + DRM_DEBUG("psp_dbg_drv_version: %u\n", + le32_to_cpu(desc->fw_version)); + DRM_DEBUG("psp_dbg_drv_size_bytes: %u\n", + le32_to_cpu(desc->size_bytes)); + break; + default: + DRM_DEBUG("Unsupported PSP fw type: %d\n", desc->fw_type); + break; + } + } } else { DRM_ERROR("Unknown PSP ucode version: %u.%u\n", version_major, version_minor); @@ -355,8 +520,8 @@ bool amdgpu_ucode_hdr_version(union amdgpu_firmware_header *hdr, { if ((hdr->common.header_version_major == hdr_major) && (hdr->common.header_version_minor == hdr_minor)) - return false; - return true; + return true; + return false; } enum amdgpu_firmware_load_type @@ -389,26 +554,6 @@ amdgpu_ucode_get_load_type(struct amdgpu_device *adev, int load_type) case CHIP_POLARIS12: case CHIP_VEGAM: return AMDGPU_FW_LOAD_SMU; - case CHIP_VEGA10: - case CHIP_RAVEN: - case CHIP_VEGA12: - case CHIP_VEGA20: - case CHIP_ARCTURUS: - case CHIP_RENOIR: - case CHIP_NAVI10: - case CHIP_NAVI14: - case CHIP_NAVI12: - case CHIP_SIENNA_CICHLID: - case CHIP_NAVY_FLOUNDER: - case CHIP_VANGOGH: - case CHIP_DIMGREY_CAVEFISH: - case CHIP_ALDEBARAN: - case CHIP_BEIGE_GOBY: - case CHIP_YELLOW_CARP: - if (!load_type) - return AMDGPU_FW_LOAD_DIRECT; - else - return AMDGPU_FW_LOAD_PSP; case CHIP_CYAN_SKILLFISH: if (!(load_type && adev->apu_flags & AMD_APU_IS_CYAN_SKILLFISH2)) @@ -442,6 +587,10 @@ const char *amdgpu_ucode_name(enum AMDGPU_UCODE_ID ucode_id) return "SDMA6"; case AMDGPU_UCODE_ID_SDMA7: return "SDMA7"; + case AMDGPU_UCODE_ID_SDMA_UCODE_TH0: + return "SDMA_CTX"; + case AMDGPU_UCODE_ID_SDMA_UCODE_TH1: + return "SDMA_CTL"; case AMDGPU_UCODE_ID_CP_CE: return "CP_CE"; case AMDGPU_UCODE_ID_CP_PFP: @@ -460,6 +609,10 @@ const char *amdgpu_ucode_name(enum AMDGPU_UCODE_ID ucode_id) return "CP_MES"; case AMDGPU_UCODE_ID_CP_MES_DATA: return "CP_MES_DATA"; + case AMDGPU_UCODE_ID_CP_MES1: + return "CP_MES_KIQ"; + case AMDGPU_UCODE_ID_CP_MES1_DATA: + return "CP_MES_KIQ_DATA"; case AMDGPU_UCODE_ID_RLC_RESTORE_LIST_CNTL: return "RLC_RESTORE_LIST_CNTL"; case AMDGPU_UCODE_ID_RLC_RESTORE_LIST_GPM_MEM: @@ -472,10 +625,30 @@ const char *amdgpu_ucode_name(enum AMDGPU_UCODE_ID ucode_id) return "RLC_DRAM"; case AMDGPU_UCODE_ID_RLC_G: return "RLC_G"; + case AMDGPU_UCODE_ID_RLC_P: + return "RLC_P"; + case AMDGPU_UCODE_ID_RLC_V: + return "RLC_V"; + case AMDGPU_UCODE_ID_GLOBAL_TAP_DELAYS: + return "GLOBAL_TAP_DELAYS"; + case AMDGPU_UCODE_ID_SE0_TAP_DELAYS: + return "SE0_TAP_DELAYS"; + case AMDGPU_UCODE_ID_SE1_TAP_DELAYS: + return "SE1_TAP_DELAYS"; + case AMDGPU_UCODE_ID_SE2_TAP_DELAYS: + return "SE2_TAP_DELAYS"; + case AMDGPU_UCODE_ID_SE3_TAP_DELAYS: + return "SE3_TAP_DELAYS"; + case AMDGPU_UCODE_ID_IMU_I: + return "IMU_I"; + case AMDGPU_UCODE_ID_IMU_D: + return "IMU_D"; case AMDGPU_UCODE_ID_STORAGE: return "STORAGE"; case AMDGPU_UCODE_ID_SMC: return "SMC"; + case AMDGPU_UCODE_ID_PPTABLE: + return "PPTABLE"; case AMDGPU_UCODE_ID_UVD: return "UVD"; case AMDGPU_UCODE_ID_UVD1: @@ -525,6 +698,7 @@ FW_VERSION_ATTR(rlc_srlg_fw_version, 0444, gfx.rlc_srlg_fw_version); FW_VERSION_ATTR(rlc_srls_fw_version, 0444, gfx.rlc_srls_fw_version); FW_VERSION_ATTR(mec_fw_version, 0444, gfx.mec_fw_version); FW_VERSION_ATTR(mec2_fw_version, 0444, gfx.mec2_fw_version); +FW_VERSION_ATTR(imu_fw_version, 0444, gfx.imu_fw_version); FW_VERSION_ATTR(sos_fw_version, 0444, psp.sos.fw_version); FW_VERSION_ATTR(asd_fw_version, 0444, psp.asd_context.bin_desc.fw_version); FW_VERSION_ATTR(ta_ras_fw_version, 0444, psp.ras_context.context.bin_desc.fw_version); @@ -546,7 +720,8 @@ static struct attribute *fw_attrs[] = { &dev_attr_ta_ras_fw_version.attr, &dev_attr_ta_xgmi_fw_version.attr, &dev_attr_smc_fw_version.attr, &dev_attr_sdma_fw_version.attr, &dev_attr_sdma2_fw_version.attr, &dev_attr_vcn_fw_version.attr, - &dev_attr_dmcu_fw_version.attr, NULL + &dev_attr_dmcu_fw_version.attr, &dev_attr_imu_fw_version.attr, + NULL }; static const struct attribute_group fw_attr_group = { @@ -570,9 +745,12 @@ static int amdgpu_ucode_init_single_fw(struct amdgpu_device *adev, { const struct common_firmware_header *header = NULL; const struct gfx_firmware_header_v1_0 *cp_hdr = NULL; + const struct gfx_firmware_header_v2_0 *cpv2_hdr = NULL; const struct dmcu_firmware_header_v1_0 *dmcu_hdr = NULL; const struct dmcub_firmware_header_v1_0 *dmcub_hdr = NULL; const struct mes_firmware_header_v1_0 *mes_hdr = NULL; + const struct sdma_firmware_header_v2_0 *sdma_hdr = NULL; + const struct imu_firmware_header_v1_0 *imu_hdr = NULL; u8 *ucode_addr; if (NULL == ucode->fw) @@ -586,12 +764,25 @@ static int amdgpu_ucode_init_single_fw(struct amdgpu_device *adev, header = (const struct common_firmware_header *)ucode->fw->data; cp_hdr = (const struct gfx_firmware_header_v1_0 *)ucode->fw->data; + cpv2_hdr = (const struct gfx_firmware_header_v2_0 *)ucode->fw->data; dmcu_hdr = (const struct dmcu_firmware_header_v1_0 *)ucode->fw->data; dmcub_hdr = (const struct dmcub_firmware_header_v1_0 *)ucode->fw->data; mes_hdr = (const struct mes_firmware_header_v1_0 *)ucode->fw->data; + sdma_hdr = (const struct sdma_firmware_header_v2_0 *)ucode->fw->data; + imu_hdr = (const struct imu_firmware_header_v1_0 *)ucode->fw->data; if (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) { switch (ucode->ucode_id) { + case AMDGPU_UCODE_ID_SDMA_UCODE_TH0: + ucode->ucode_size = le32_to_cpu(sdma_hdr->ctx_ucode_size_bytes); + ucode_addr = (u8 *)ucode->fw->data + + le32_to_cpu(sdma_hdr->header.ucode_array_offset_bytes); + break; + case AMDGPU_UCODE_ID_SDMA_UCODE_TH1: + ucode->ucode_size = le32_to_cpu(sdma_hdr->ctl_ucode_size_bytes); + ucode_addr = (u8 *)ucode->fw->data + + le32_to_cpu(sdma_hdr->ctl_ucode_offset); + break; case AMDGPU_UCODE_ID_CP_MEC1: case AMDGPU_UCODE_ID_CP_MEC2: ucode->ucode_size = le32_to_cpu(header->ucode_size_bytes) - @@ -626,6 +817,34 @@ static int amdgpu_ucode_init_single_fw(struct amdgpu_device *adev, ucode->ucode_size = adev->gfx.rlc.rlc_dram_ucode_size_bytes; ucode_addr = adev->gfx.rlc.rlc_dram_ucode; break; + case AMDGPU_UCODE_ID_RLC_P: + ucode->ucode_size = adev->gfx.rlc.rlcp_ucode_size_bytes; + ucode_addr = adev->gfx.rlc.rlcp_ucode; + break; + case AMDGPU_UCODE_ID_RLC_V: + ucode->ucode_size = adev->gfx.rlc.rlcv_ucode_size_bytes; + ucode_addr = adev->gfx.rlc.rlcv_ucode; + break; + case AMDGPU_UCODE_ID_GLOBAL_TAP_DELAYS: + ucode->ucode_size = adev->gfx.rlc.global_tap_delays_ucode_size_bytes; + ucode_addr = adev->gfx.rlc.global_tap_delays_ucode; + break; + case AMDGPU_UCODE_ID_SE0_TAP_DELAYS: + ucode->ucode_size = adev->gfx.rlc.se0_tap_delays_ucode_size_bytes; + ucode_addr = adev->gfx.rlc.se0_tap_delays_ucode; + break; + case AMDGPU_UCODE_ID_SE1_TAP_DELAYS: + ucode->ucode_size = adev->gfx.rlc.se1_tap_delays_ucode_size_bytes; + ucode_addr = adev->gfx.rlc.se1_tap_delays_ucode; + break; + case AMDGPU_UCODE_ID_SE2_TAP_DELAYS: + ucode->ucode_size = adev->gfx.rlc.se2_tap_delays_ucode_size_bytes; + ucode_addr = adev->gfx.rlc.se2_tap_delays_ucode; + break; + case AMDGPU_UCODE_ID_SE3_TAP_DELAYS: + ucode->ucode_size = adev->gfx.rlc.se3_tap_delays_ucode_size_bytes; + ucode_addr = adev->gfx.rlc.se3_tap_delays_ucode; + break; case AMDGPU_UCODE_ID_CP_MES: ucode->ucode_size = le32_to_cpu(mes_hdr->mes_ucode_size_bytes); ucode_addr = (u8 *)ucode->fw->data + @@ -636,6 +855,16 @@ static int amdgpu_ucode_init_single_fw(struct amdgpu_device *adev, ucode_addr = (u8 *)ucode->fw->data + le32_to_cpu(mes_hdr->mes_ucode_data_offset_bytes); break; + case AMDGPU_UCODE_ID_CP_MES1: + ucode->ucode_size = le32_to_cpu(mes_hdr->mes_ucode_size_bytes); + ucode_addr = (u8 *)ucode->fw->data + + le32_to_cpu(mes_hdr->mes_ucode_offset_bytes); + break; + case AMDGPU_UCODE_ID_CP_MES1_DATA: + ucode->ucode_size = le32_to_cpu(mes_hdr->mes_ucode_data_size_bytes); + ucode_addr = (u8 *)ucode->fw->data + + le32_to_cpu(mes_hdr->mes_ucode_data_offset_bytes); + break; case AMDGPU_UCODE_ID_DMCU_ERAM: ucode->ucode_size = le32_to_cpu(header->ucode_size_bytes) - le32_to_cpu(dmcu_hdr->intv_size_bytes); @@ -653,6 +882,76 @@ static int amdgpu_ucode_init_single_fw(struct amdgpu_device *adev, ucode_addr = (u8 *)ucode->fw->data + le32_to_cpu(header->ucode_array_offset_bytes); break; + case AMDGPU_UCODE_ID_PPTABLE: + ucode->ucode_size = ucode->fw->size; + ucode_addr = (u8 *)ucode->fw->data; + break; + case AMDGPU_UCODE_ID_IMU_I: + ucode->ucode_size = le32_to_cpu(imu_hdr->imu_iram_ucode_size_bytes); + ucode_addr = (u8 *)ucode->fw->data + + le32_to_cpu(imu_hdr->header.ucode_array_offset_bytes); + break; + case AMDGPU_UCODE_ID_IMU_D: + ucode->ucode_size = le32_to_cpu(imu_hdr->imu_dram_ucode_size_bytes); + ucode_addr = (u8 *)ucode->fw->data + + le32_to_cpu(imu_hdr->header.ucode_array_offset_bytes) + + le32_to_cpu(imu_hdr->imu_iram_ucode_size_bytes); + break; + case AMDGPU_UCODE_ID_CP_RS64_PFP: + ucode->ucode_size = le32_to_cpu(cpv2_hdr->ucode_size_bytes); + ucode_addr = (u8 *)ucode->fw->data + + le32_to_cpu(header->ucode_array_offset_bytes); + break; + case AMDGPU_UCODE_ID_CP_RS64_PFP_P0_STACK: + ucode->ucode_size = le32_to_cpu(cpv2_hdr->data_size_bytes); + ucode_addr = (u8 *)ucode->fw->data + + le32_to_cpu(cpv2_hdr->data_offset_bytes); + break; + case AMDGPU_UCODE_ID_CP_RS64_PFP_P1_STACK: + ucode->ucode_size = le32_to_cpu(cpv2_hdr->data_size_bytes); + ucode_addr = (u8 *)ucode->fw->data + + le32_to_cpu(cpv2_hdr->data_offset_bytes); + break; + case AMDGPU_UCODE_ID_CP_RS64_ME: + ucode->ucode_size = le32_to_cpu(cpv2_hdr->ucode_size_bytes); + ucode_addr = (u8 *)ucode->fw->data + + le32_to_cpu(header->ucode_array_offset_bytes); + break; + case AMDGPU_UCODE_ID_CP_RS64_ME_P0_STACK: + ucode->ucode_size = le32_to_cpu(cpv2_hdr->data_size_bytes); + ucode_addr = (u8 *)ucode->fw->data + + le32_to_cpu(cpv2_hdr->data_offset_bytes); + break; + case AMDGPU_UCODE_ID_CP_RS64_ME_P1_STACK: + ucode->ucode_size = le32_to_cpu(cpv2_hdr->data_size_bytes); + ucode_addr = (u8 *)ucode->fw->data + + le32_to_cpu(cpv2_hdr->data_offset_bytes); + break; + case AMDGPU_UCODE_ID_CP_RS64_MEC: + ucode->ucode_size = le32_to_cpu(cpv2_hdr->ucode_size_bytes); + ucode_addr = (u8 *)ucode->fw->data + + le32_to_cpu(header->ucode_array_offset_bytes); + break; + case AMDGPU_UCODE_ID_CP_RS64_MEC_P0_STACK: + ucode->ucode_size = le32_to_cpu(cpv2_hdr->data_size_bytes); + ucode_addr = (u8 *)ucode->fw->data + + le32_to_cpu(cpv2_hdr->data_offset_bytes); + break; + case AMDGPU_UCODE_ID_CP_RS64_MEC_P1_STACK: + ucode->ucode_size = le32_to_cpu(cpv2_hdr->data_size_bytes); + ucode_addr = (u8 *)ucode->fw->data + + le32_to_cpu(cpv2_hdr->data_offset_bytes); + break; + case AMDGPU_UCODE_ID_CP_RS64_MEC_P2_STACK: + ucode->ucode_size = le32_to_cpu(cpv2_hdr->data_size_bytes); + ucode_addr = (u8 *)ucode->fw->data + + le32_to_cpu(cpv2_hdr->data_offset_bytes); + break; + case AMDGPU_UCODE_ID_CP_RS64_MEC_P3_STACK: + ucode->ucode_size = le32_to_cpu(cpv2_hdr->data_size_bytes); + ucode_addr = (u8 *)ucode->fw->data + + le32_to_cpu(cpv2_hdr->data_offset_bytes); + break; default: ucode->ucode_size = le32_to_cpu(header->ucode_size_bytes); ucode_addr = (u8 *)ucode->fw->data + @@ -714,8 +1013,7 @@ int amdgpu_ucode_create_bo(struct amdgpu_device *adev) void amdgpu_ucode_free_bo(struct amdgpu_device *adev) { - if (adev->firmware.load_type != AMDGPU_FW_LOAD_DIRECT) - amdgpu_bo_free_kernel(&adev->firmware.fw_buf, + amdgpu_bo_free_kernel(&adev->firmware.fw_buf, &adev->firmware.fw_buf_mc, &adev->firmware.fw_buf_ptr); } @@ -760,3 +1058,36 @@ int amdgpu_ucode_init_bo(struct amdgpu_device *adev) } return 0; } + +void amdgpu_ucode_ip_version_decode(struct amdgpu_device *adev, int block_type, char *ucode_prefix, int len) +{ + int maj, min, rev; + char *ip_name; + uint32_t version = adev->ip_versions[block_type][0]; + + switch (block_type) { + case GC_HWIP: + ip_name = "gc"; + break; + case SDMA0_HWIP: + ip_name = "sdma"; + break; + case MP0_HWIP: + ip_name = "psp"; + break; + case MP1_HWIP: + ip_name = "smu"; + break; + case UVD_HWIP: + ip_name = "vcn"; + break; + default: + BUG(); + } + + maj = IP_VERSION_MAJ(version); + min = IP_VERSION_MIN(version); + rev = IP_VERSION_REV(version); + + snprintf(ucode_prefix, len, "%s_%d_%d_%d", ip_name, maj, min, rev); +} diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.h index 7c2538db3cd5..1c36235b4539 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.h @@ -124,6 +124,7 @@ enum psp_fw_type { PSP_FW_TYPE_PSP_SOC_DRV, PSP_FW_TYPE_PSP_INTF_DRV, PSP_FW_TYPE_PSP_DBG_DRV, + PSP_FW_TYPE_PSP_RAS_DRV, }; /* version_major=2, version_minor=0 */ @@ -170,6 +171,18 @@ struct gfx_firmware_header_v1_0 { uint32_t jt_size; /* size of jt */ }; +/* version_major=2, version_minor=0 */ +struct gfx_firmware_header_v2_0 { + struct common_firmware_header header; + uint32_t ucode_feature_version; + uint32_t ucode_size_bytes; + uint32_t ucode_offset_bytes; + uint32_t data_size_bytes; + uint32_t data_offset_bytes; + uint32_t ucode_start_addr_lo; + uint32_t ucode_start_addr_hi; +}; + /* version_major=1, version_minor=0 */ struct mes_firmware_header_v1_0 { struct common_firmware_header header; @@ -236,7 +249,7 @@ struct rlc_firmware_header_v2_1 { uint32_t save_restore_list_srm_offset_bytes; }; -/* version_major=2, version_minor=1 */ +/* version_major=2, version_minor=2 */ struct rlc_firmware_header_v2_2 { struct rlc_firmware_header_v2_1 v2_1; uint32_t rlc_iram_ucode_size_bytes; @@ -245,6 +258,34 @@ struct rlc_firmware_header_v2_2 { uint32_t rlc_dram_ucode_offset_bytes; }; +/* version_major=2, version_minor=3 */ +struct rlc_firmware_header_v2_3 { + struct rlc_firmware_header_v2_2 v2_2; + uint32_t rlcp_ucode_version; + uint32_t rlcp_ucode_feature_version; + uint32_t rlcp_ucode_size_bytes; + uint32_t rlcp_ucode_offset_bytes; + uint32_t rlcv_ucode_version; + uint32_t rlcv_ucode_feature_version; + uint32_t rlcv_ucode_size_bytes; + uint32_t rlcv_ucode_offset_bytes; +}; + +/* version_major=2, version_minor=4 */ +struct rlc_firmware_header_v2_4 { + struct rlc_firmware_header_v2_3 v2_3; + uint32_t global_tap_delays_ucode_size_bytes; + uint32_t global_tap_delays_ucode_offset_bytes; + uint32_t se0_tap_delays_ucode_size_bytes; + uint32_t se0_tap_delays_ucode_offset_bytes; + uint32_t se1_tap_delays_ucode_size_bytes; + uint32_t se1_tap_delays_ucode_offset_bytes; + uint32_t se2_tap_delays_ucode_size_bytes; + uint32_t se2_tap_delays_ucode_offset_bytes; + uint32_t se3_tap_delays_ucode_size_bytes; + uint32_t se3_tap_delays_ucode_offset_bytes; +}; + /* version_major=1, version_minor=0 */ struct sdma_firmware_header_v1_0 { struct common_firmware_header header; @@ -260,6 +301,19 @@ struct sdma_firmware_header_v1_1 { uint32_t digest_size; }; +/* version_major=2, version_minor=0 */ +struct sdma_firmware_header_v2_0 { + struct common_firmware_header header; + uint32_t ucode_feature_version; + uint32_t ctx_ucode_size_bytes; /* context thread ucode size */ + uint32_t ctx_jt_offset; /* context thread jt location */ + uint32_t ctx_jt_size; /* context thread size of jt */ + uint32_t ctl_ucode_offset; + uint32_t ctl_ucode_size_bytes; /* control thread ucode size */ + uint32_t ctl_jt_offset; /* control thread jt location */ + uint32_t ctl_jt_size; /* control thread size of jt */ +}; + /* gpu info payload */ struct gpu_info_firmware_v1_0 { uint32_t gc_num_se; @@ -313,6 +367,15 @@ struct dmcub_firmware_header_v1_0 { uint32_t bss_data_bytes; /* size of bss/data region, in bytes */ }; +/* version_major=1, version_minor=0 */ +struct imu_firmware_header_v1_0 { + struct common_firmware_header header; + uint32_t imu_iram_ucode_size_bytes; + uint32_t imu_iram_ucode_offset_bytes; + uint32_t imu_dram_ucode_size_bytes; + uint32_t imu_dram_ucode_offset_bytes; +}; + /* header is fixed size */ union amdgpu_firmware_header { struct common_firmware_header common; @@ -326,14 +389,20 @@ union amdgpu_firmware_header { struct ta_firmware_header_v1_0 ta; struct ta_firmware_header_v2_0 ta_v2_0; struct gfx_firmware_header_v1_0 gfx; + struct gfx_firmware_header_v2_0 gfx_v2_0; struct rlc_firmware_header_v1_0 rlc; struct rlc_firmware_header_v2_0 rlc_v2_0; struct rlc_firmware_header_v2_1 rlc_v2_1; + struct rlc_firmware_header_v2_2 rlc_v2_2; + struct rlc_firmware_header_v2_3 rlc_v2_3; + struct rlc_firmware_header_v2_4 rlc_v2_4; struct sdma_firmware_header_v1_0 sdma; struct sdma_firmware_header_v1_1 sdma_v1_1; + struct sdma_firmware_header_v2_0 sdma_v2_0; struct gpu_info_firmware_header_v1_0 gpu_info; struct dmcu_firmware_header_v1_0 dmcu; struct dmcub_firmware_header_v1_0 dmcub; + struct imu_firmware_header_v1_0 imu; uint8_t raw[0x100]; }; @@ -343,7 +412,8 @@ union amdgpu_firmware_header { * fw loading support */ enum AMDGPU_UCODE_ID { - AMDGPU_UCODE_ID_SDMA0 = 0, + AMDGPU_UCODE_ID_CAP = 0, + AMDGPU_UCODE_ID_SDMA0, AMDGPU_UCODE_ID_SDMA1, AMDGPU_UCODE_ID_SDMA2, AMDGPU_UCODE_ID_SDMA3, @@ -351,23 +421,48 @@ enum AMDGPU_UCODE_ID { AMDGPU_UCODE_ID_SDMA5, AMDGPU_UCODE_ID_SDMA6, AMDGPU_UCODE_ID_SDMA7, + AMDGPU_UCODE_ID_SDMA_UCODE_TH0, + AMDGPU_UCODE_ID_SDMA_UCODE_TH1, AMDGPU_UCODE_ID_CP_CE, AMDGPU_UCODE_ID_CP_PFP, AMDGPU_UCODE_ID_CP_ME, + AMDGPU_UCODE_ID_CP_RS64_PFP, + AMDGPU_UCODE_ID_CP_RS64_ME, + AMDGPU_UCODE_ID_CP_RS64_MEC, + AMDGPU_UCODE_ID_CP_RS64_PFP_P0_STACK, + AMDGPU_UCODE_ID_CP_RS64_PFP_P1_STACK, + AMDGPU_UCODE_ID_CP_RS64_ME_P0_STACK, + AMDGPU_UCODE_ID_CP_RS64_ME_P1_STACK, + AMDGPU_UCODE_ID_CP_RS64_MEC_P0_STACK, + AMDGPU_UCODE_ID_CP_RS64_MEC_P1_STACK, + AMDGPU_UCODE_ID_CP_RS64_MEC_P2_STACK, + AMDGPU_UCODE_ID_CP_RS64_MEC_P3_STACK, AMDGPU_UCODE_ID_CP_MEC1, AMDGPU_UCODE_ID_CP_MEC1_JT, AMDGPU_UCODE_ID_CP_MEC2, AMDGPU_UCODE_ID_CP_MEC2_JT, AMDGPU_UCODE_ID_CP_MES, AMDGPU_UCODE_ID_CP_MES_DATA, + AMDGPU_UCODE_ID_CP_MES1, + AMDGPU_UCODE_ID_CP_MES1_DATA, + AMDGPU_UCODE_ID_IMU_I, + AMDGPU_UCODE_ID_IMU_D, + AMDGPU_UCODE_ID_GLOBAL_TAP_DELAYS, + AMDGPU_UCODE_ID_SE0_TAP_DELAYS, + AMDGPU_UCODE_ID_SE1_TAP_DELAYS, + AMDGPU_UCODE_ID_SE2_TAP_DELAYS, + AMDGPU_UCODE_ID_SE3_TAP_DELAYS, AMDGPU_UCODE_ID_RLC_RESTORE_LIST_CNTL, AMDGPU_UCODE_ID_RLC_RESTORE_LIST_GPM_MEM, AMDGPU_UCODE_ID_RLC_RESTORE_LIST_SRM_MEM, AMDGPU_UCODE_ID_RLC_IRAM, AMDGPU_UCODE_ID_RLC_DRAM, + AMDGPU_UCODE_ID_RLC_P, + AMDGPU_UCODE_ID_RLC_V, AMDGPU_UCODE_ID_RLC_G, AMDGPU_UCODE_ID_STORAGE, AMDGPU_UCODE_ID_SMC, + AMDGPU_UCODE_ID_PPTABLE, AMDGPU_UCODE_ID_UVD, AMDGPU_UCODE_ID_UVD1, AMDGPU_UCODE_ID_VCE, @@ -390,8 +485,8 @@ enum AMDGPU_UCODE_STATUS { enum amdgpu_firmware_load_type { AMDGPU_FW_LOAD_DIRECT = 0, - AMDGPU_FW_LOAD_SMU, AMDGPU_FW_LOAD_PSP, + AMDGPU_FW_LOAD_SMU, AMDGPU_FW_LOAD_RLC_BACKDOOR_AUTO, }; @@ -462,4 +557,6 @@ amdgpu_ucode_get_load_type(struct amdgpu_device *adev, int load_type); const char *amdgpu_ucode_name(enum AMDGPU_UCODE_ID ucode_id); +void amdgpu_ucode_ip_version_decode(struct amdgpu_device *adev, int block_type, char *ucode_prefix, int len); + #endif diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_umc.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_umc.c index 46264a4002f7..aad3c8b4c810 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_umc.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_umc.c @@ -21,7 +21,7 @@ * */ -#include "amdgpu_ras.h" +#include "amdgpu.h" static int amdgpu_umc_do_page_retirement(struct amdgpu_device *adev, void *ras_error_status, @@ -33,14 +33,14 @@ static int amdgpu_umc_do_page_retirement(struct amdgpu_device *adev, int ret = 0; kgd2kfd_set_sram_ecc_flag(adev->kfd.dev); - ret = smu_get_ecc_info(&adev->smu, (void *)&(con->umc_ecc)); + ret = amdgpu_dpm_get_ecc_info(adev, (void *)&(con->umc_ecc)); if (ret == -EOPNOTSUPP) { - if (adev->umc.ras_funcs && - adev->umc.ras_funcs->query_ras_error_count) - adev->umc.ras_funcs->query_ras_error_count(adev, ras_error_status); + if (adev->umc.ras && adev->umc.ras->ras_block.hw_ops && + adev->umc.ras->ras_block.hw_ops->query_ras_error_count) + adev->umc.ras->ras_block.hw_ops->query_ras_error_count(adev, ras_error_status); - if (adev->umc.ras_funcs && - adev->umc.ras_funcs->query_ras_error_address && + if (adev->umc.ras && adev->umc.ras->ras_block.hw_ops && + adev->umc.ras->ras_block.hw_ops->query_ras_error_address && adev->umc.max_ras_err_cnt_per_query) { err_data->err_addr = kcalloc(adev->umc.max_ras_err_cnt_per_query, @@ -56,15 +56,15 @@ static int amdgpu_umc_do_page_retirement(struct amdgpu_device *adev, /* umc query_ras_error_address is also responsible for clearing * error status */ - adev->umc.ras_funcs->query_ras_error_address(adev, ras_error_status); + adev->umc.ras->ras_block.hw_ops->query_ras_error_address(adev, ras_error_status); } } else if (!ret) { - if (adev->umc.ras_funcs && - adev->umc.ras_funcs->ecc_info_query_ras_error_count) - adev->umc.ras_funcs->ecc_info_query_ras_error_count(adev, ras_error_status); + if (adev->umc.ras && + adev->umc.ras->ecc_info_query_ras_error_count) + adev->umc.ras->ecc_info_query_ras_error_count(adev, ras_error_status); - if (adev->umc.ras_funcs && - adev->umc.ras_funcs->ecc_info_query_ras_error_address && + if (adev->umc.ras && + adev->umc.ras->ecc_info_query_ras_error_address && adev->umc.max_ras_err_cnt_per_query) { err_data->err_addr = kcalloc(adev->umc.max_ras_err_cnt_per_query, @@ -80,7 +80,7 @@ static int amdgpu_umc_do_page_retirement(struct amdgpu_device *adev, /* umc query_ras_error_address is also responsible for clearing * error status */ - adev->umc.ras_funcs->ecc_info_query_ras_error_address(adev, ras_error_status); + adev->umc.ras->ecc_info_query_ras_error_address(adev, ras_error_status); } } @@ -96,8 +96,12 @@ static int amdgpu_umc_do_page_retirement(struct amdgpu_device *adev, err_data->err_addr_cnt); amdgpu_ras_save_bad_pages(adev); - if (adev->smu.ppt_funcs && adev->smu.ppt_funcs->send_hbm_bad_pages_num) - adev->smu.ppt_funcs->send_hbm_bad_pages_num(&adev->smu, con->eeprom_control.ras_num_recs); + amdgpu_dpm_send_hbm_bad_pages_num(adev, con->eeprom_control.ras_num_recs); + + if (con->update_channel_flag == true) { + amdgpu_dpm_send_hbm_bad_channel_flag(adev, con->eeprom_control.bad_channel_bitmap); + con->update_channel_flag = false; + } } if (reset) @@ -130,78 +134,39 @@ int amdgpu_umc_poison_handler(struct amdgpu_device *adev, return ret; } -static int amdgpu_umc_process_ras_data_cb(struct amdgpu_device *adev, +int amdgpu_umc_process_ras_data_cb(struct amdgpu_device *adev, void *ras_error_status, struct amdgpu_iv_entry *entry) { return amdgpu_umc_do_page_retirement(adev, ras_error_status, entry, true); } -int amdgpu_umc_ras_late_init(struct amdgpu_device *adev) +int amdgpu_umc_ras_late_init(struct amdgpu_device *adev, struct ras_common_if *ras_block) { int r; - struct ras_fs_if fs_info = { - .sysfs_name = "umc_err_count", - }; - struct ras_ih_if ih_info = { - .cb = amdgpu_umc_process_ras_data_cb, - }; - if (!adev->umc.ras_if) { - adev->umc.ras_if = - kmalloc(sizeof(struct ras_common_if), GFP_KERNEL); - if (!adev->umc.ras_if) - return -ENOMEM; - adev->umc.ras_if->block = AMDGPU_RAS_BLOCK__UMC; - adev->umc.ras_if->type = AMDGPU_RAS_ERROR__MULTI_UNCORRECTABLE; - adev->umc.ras_if->sub_block_index = 0; - } - ih_info.head = fs_info.head = *adev->umc.ras_if; - - r = amdgpu_ras_late_init(adev, adev->umc.ras_if, - &fs_info, &ih_info); + r = amdgpu_ras_block_late_init(adev, ras_block); if (r) - goto free; + return r; - if (amdgpu_ras_is_supported(adev, adev->umc.ras_if->block)) { + if (amdgpu_ras_is_supported(adev, ras_block->block)) { r = amdgpu_irq_get(adev, &adev->gmc.ecc_irq, 0); if (r) goto late_fini; - } else { - r = 0; - goto free; } /* ras init of specific umc version */ - if (adev->umc.ras_funcs && - adev->umc.ras_funcs->err_cnt_init) - adev->umc.ras_funcs->err_cnt_init(adev); + if (adev->umc.ras && + adev->umc.ras->err_cnt_init) + adev->umc.ras->err_cnt_init(adev); return 0; late_fini: - amdgpu_ras_late_fini(adev, adev->umc.ras_if, &ih_info); -free: - kfree(adev->umc.ras_if); - adev->umc.ras_if = NULL; + amdgpu_ras_block_late_fini(adev, ras_block); return r; } -void amdgpu_umc_ras_fini(struct amdgpu_device *adev) -{ - if (amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__UMC) && - adev->umc.ras_if) { - struct ras_common_if *ras_if = adev->umc.ras_if; - struct ras_ih_if ih_info = { - .head = *ras_if, - .cb = amdgpu_umc_process_ras_data_cb, - }; - - amdgpu_ras_late_fini(adev, ras_if, &ih_info); - kfree(ras_if); - } -} - int amdgpu_umc_process_ecc_irq(struct amdgpu_device *adev, struct amdgpu_irq_src *source, struct amdgpu_iv_entry *entry) @@ -219,3 +184,24 @@ int amdgpu_umc_process_ecc_irq(struct amdgpu_device *adev, amdgpu_ras_interrupt_dispatch(adev, &ih_data); return 0; } + +void amdgpu_umc_fill_error_record(struct ras_err_data *err_data, + uint64_t err_addr, + uint64_t retired_page, + uint32_t channel_index, + uint32_t umc_inst) +{ + struct eeprom_table_record *err_rec = + &err_data->err_addr[err_data->err_addr_cnt]; + + err_rec->address = err_addr; + /* page frame address is saved */ + err_rec->retired_page = retired_page >> AMDGPU_GPU_PAGE_SHIFT; + err_rec->ts = (uint64_t)ktime_get_real_seconds(); + err_rec->err_type = AMDGPU_RAS_EEPROM_ERR_NON_RECOVERABLE; + err_rec->cu = 0; + err_rec->mem_channel = channel_index; + err_rec->mcumc_id = umc_inst; + + err_data->err_addr_cnt++; +} diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_umc.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_umc.h index b72194e8bfe5..e46439274f3a 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_umc.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_umc.h @@ -20,6 +20,7 @@ */ #ifndef __AMDGPU_UMC_H__ #define __AMDGPU_UMC_H__ +#include "amdgpu_ras.h" /* * (addr / 256) * 4096, the higher 26 bits in ErrorAddr @@ -40,15 +41,19 @@ #define LOOP_UMC_CH_INST(ch_inst) for ((ch_inst) = 0; (ch_inst) < adev->umc.channel_inst_num; (ch_inst)++) #define LOOP_UMC_INST_AND_CH(umc_inst, ch_inst) LOOP_UMC_INST((umc_inst)) LOOP_UMC_CH_INST((ch_inst)) -struct amdgpu_umc_ras_funcs { +#define LOOP_UMC_NODE_INST(node_inst) \ + for ((node_inst) = 0; (node_inst) < adev->umc.node_inst_num; (node_inst)++) + +#define LOOP_UMC_EACH_NODE_INST_AND_CH(node_inst, umc_inst, ch_inst) \ + LOOP_UMC_NODE_INST((node_inst)) LOOP_UMC_INST_AND_CH((umc_inst), (ch_inst)) + +struct amdgpu_umc_ras { + struct amdgpu_ras_block_object ras_block; void (*err_cnt_init)(struct amdgpu_device *adev); - int (*ras_late_init)(struct amdgpu_device *adev); - void (*ras_fini)(struct amdgpu_device *adev); - void (*query_ras_error_count)(struct amdgpu_device *adev, - void *ras_error_status); - void (*query_ras_error_address)(struct amdgpu_device *adev, - void *ras_error_status); bool (*query_ras_poison_mode)(struct amdgpu_device *adev); + void (*convert_ras_error_address)(struct amdgpu_device *adev, + struct ras_err_data *err_data, uint64_t err_addr, + uint32_t ch_inst, uint32_t umc_inst); void (*ecc_info_query_ras_error_count)(struct amdgpu_device *adev, void *ras_error_status); void (*ecc_info_query_ras_error_address)(struct amdgpu_device *adev, @@ -66,6 +71,10 @@ struct amdgpu_umc { uint32_t channel_inst_num; /* number of umc instance with memory map register access */ uint32_t umc_inst_num; + + /*number of umc node instance with memory map register access*/ + uint32_t node_inst_num; + /* UMC regiser per channel offset */ uint32_t channel_offs; /* channel index table of interleaved memory */ @@ -73,15 +82,23 @@ struct amdgpu_umc { struct ras_common_if *ras_if; const struct amdgpu_umc_funcs *funcs; - const struct amdgpu_umc_ras_funcs *ras_funcs; + struct amdgpu_umc_ras *ras; }; -int amdgpu_umc_ras_late_init(struct amdgpu_device *adev); -void amdgpu_umc_ras_fini(struct amdgpu_device *adev); +int amdgpu_umc_ras_late_init(struct amdgpu_device *adev, struct ras_common_if *ras_block); int amdgpu_umc_poison_handler(struct amdgpu_device *adev, void *ras_error_status, bool reset); int amdgpu_umc_process_ecc_irq(struct amdgpu_device *adev, struct amdgpu_irq_src *source, struct amdgpu_iv_entry *entry); +void amdgpu_umc_fill_error_record(struct ras_err_data *err_data, + uint64_t err_addr, + uint64_t retired_page, + uint32_t channel_index, + uint32_t umc_inst); + +int amdgpu_umc_process_ras_data_cb(struct amdgpu_device *adev, + void *ras_error_status, + struct amdgpu_iv_entry *entry); #endif diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c index 6f8de11a17f1..6eac649499d3 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c @@ -37,6 +37,7 @@ #include "amdgpu.h" #include "amdgpu_pm.h" #include "amdgpu_uvd.h" +#include "amdgpu_cs.h" #include "cikd.h" #include "uvd/uvd_4_2_d.h" @@ -98,7 +99,7 @@ struct amdgpu_uvd_cs_ctx { unsigned reg, count; unsigned data0, data1; unsigned idx; - unsigned ib_idx; + struct amdgpu_ib *ib; /* does the IB has a msg command */ bool has_msg_cmd; @@ -557,8 +558,8 @@ static u64 amdgpu_uvd_get_addr_from_ctx(struct amdgpu_uvd_cs_ctx *ctx) uint32_t lo, hi; uint64_t addr; - lo = amdgpu_get_ib_value(ctx->parser, ctx->ib_idx, ctx->data0); - hi = amdgpu_get_ib_value(ctx->parser, ctx->ib_idx, ctx->data1); + lo = amdgpu_ib_get_value(ctx->ib, ctx->data0); + hi = amdgpu_ib_get_value(ctx->ib, ctx->data1); addr = ((uint64_t)lo) | (((uint64_t)hi) << 32); return addr; @@ -589,7 +590,7 @@ static int amdgpu_uvd_cs_pass1(struct amdgpu_uvd_cs_ctx *ctx) if (!ctx->parser->adev->uvd.address_64_bit) { /* check if it's a message or feedback command */ - cmd = amdgpu_get_ib_value(ctx->parser, ctx->ib_idx, ctx->idx) >> 1; + cmd = amdgpu_ib_get_value(ctx->ib, ctx->idx) >> 1; if (cmd == 0x0 || cmd == 0x3) { /* yes, force it into VRAM */ uint32_t domain = AMDGPU_GEM_DOMAIN_VRAM; @@ -834,6 +835,7 @@ static int amdgpu_uvd_cs_msg(struct amdgpu_uvd_cs_ctx *ctx, handle = msg[2]; if (handle == 0) { + amdgpu_bo_kunmap(bo); DRM_ERROR("Invalid UVD handle!\n"); return -EINVAL; } @@ -892,6 +894,7 @@ static int amdgpu_uvd_cs_msg(struct amdgpu_uvd_cs_ctx *ctx, DRM_ERROR("Illegal UVD message type (%d)!\n", msg_type); } + amdgpu_bo_kunmap(bo); return -EINVAL; } @@ -925,12 +928,10 @@ static int amdgpu_uvd_cs_pass2(struct amdgpu_uvd_cs_ctx *ctx) addr -= mapping->start * AMDGPU_GPU_PAGE_SIZE; start += addr; - amdgpu_set_ib_value(ctx->parser, ctx->ib_idx, ctx->data0, - lower_32_bits(start)); - amdgpu_set_ib_value(ctx->parser, ctx->ib_idx, ctx->data1, - upper_32_bits(start)); + amdgpu_ib_set_value(ctx->ib, ctx->data0, lower_32_bits(start)); + amdgpu_ib_set_value(ctx->ib, ctx->data1, upper_32_bits(start)); - cmd = amdgpu_get_ib_value(ctx->parser, ctx->ib_idx, ctx->idx) >> 1; + cmd = amdgpu_ib_get_value(ctx->ib, ctx->idx) >> 1; if (cmd < 0x4) { if ((end - start) < ctx->buf_sizes[cmd]) { DRM_ERROR("buffer (%d) to small (%d / %d)!\n", cmd, @@ -990,14 +991,13 @@ static int amdgpu_uvd_cs_pass2(struct amdgpu_uvd_cs_ctx *ctx) static int amdgpu_uvd_cs_reg(struct amdgpu_uvd_cs_ctx *ctx, int (*cb)(struct amdgpu_uvd_cs_ctx *ctx)) { - struct amdgpu_ib *ib = &ctx->parser->job->ibs[ctx->ib_idx]; int i, r; ctx->idx++; for (i = 0; i <= ctx->count; ++i) { unsigned reg = ctx->reg + i; - if (ctx->idx >= ib->length_dw) { + if (ctx->idx >= ctx->ib->length_dw) { DRM_ERROR("Register command after end of CS!\n"); return -EINVAL; } @@ -1037,11 +1037,10 @@ static int amdgpu_uvd_cs_reg(struct amdgpu_uvd_cs_ctx *ctx, static int amdgpu_uvd_cs_packets(struct amdgpu_uvd_cs_ctx *ctx, int (*cb)(struct amdgpu_uvd_cs_ctx *ctx)) { - struct amdgpu_ib *ib = &ctx->parser->job->ibs[ctx->ib_idx]; int r; - for (ctx->idx = 0 ; ctx->idx < ib->length_dw; ) { - uint32_t cmd = amdgpu_get_ib_value(ctx->parser, ctx->ib_idx, ctx->idx); + for (ctx->idx = 0 ; ctx->idx < ctx->ib->length_dw; ) { + uint32_t cmd = amdgpu_ib_get_value(ctx->ib, ctx->idx); unsigned type = CP_PACKET_GET_TYPE(cmd); switch (type) { case PACKET_TYPE0: @@ -1066,11 +1065,14 @@ static int amdgpu_uvd_cs_packets(struct amdgpu_uvd_cs_ctx *ctx, * amdgpu_uvd_ring_parse_cs - UVD command submission parser * * @parser: Command submission parser context - * @ib_idx: Which indirect buffer to use + * @job: the job to parse + * @ib: the IB to patch * * Parse the command stream, patch in addresses as necessary. */ -int amdgpu_uvd_ring_parse_cs(struct amdgpu_cs_parser *parser, uint32_t ib_idx) +int amdgpu_uvd_ring_parse_cs(struct amdgpu_cs_parser *parser, + struct amdgpu_job *job, + struct amdgpu_ib *ib) { struct amdgpu_uvd_cs_ctx ctx = {}; unsigned buf_sizes[] = { @@ -1080,10 +1082,9 @@ int amdgpu_uvd_ring_parse_cs(struct amdgpu_cs_parser *parser, uint32_t ib_idx) [0x00000003] = 2048, [0x00000004] = 0xFFFFFFFF, }; - struct amdgpu_ib *ib = &parser->job->ibs[ib_idx]; int r; - parser->job->vm = NULL; + job->vm = NULL; ib->gpu_addr = amdgpu_sa_bo_gpu_addr(ib->sa_bo); if (ib->length_dw % 16) { @@ -1094,7 +1095,7 @@ int amdgpu_uvd_ring_parse_cs(struct amdgpu_cs_parser *parser, uint32_t ib_idx) ctx.parser = parser; ctx.buf_sizes = buf_sizes; - ctx.ib_idx = ib_idx; + ctx.ib = ib; /* first round only required on chips without UVD 64 bit address support */ if (!parser->adev->uvd.address_64_bit) { @@ -1162,7 +1163,8 @@ static int amdgpu_uvd_send_msg(struct amdgpu_ring *ring, struct amdgpu_bo *bo, ib->length_dw = 16; if (direct) { - r = dma_resv_wait_timeout(bo->tbo.base.resv, true, false, + r = dma_resv_wait_timeout(bo->tbo.base.resv, + DMA_RESV_USAGE_KERNEL, false, msecs_to_jiffies(10)); if (r == 0) r = -ETIMEDOUT; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.h index 76ac9699885d..9f89bb7cd60b 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.h @@ -82,7 +82,9 @@ int amdgpu_uvd_get_destroy_msg(struct amdgpu_ring *ring, uint32_t handle, bool direct, struct dma_fence **fence); void amdgpu_uvd_free_handles(struct amdgpu_device *adev, struct drm_file *filp); -int amdgpu_uvd_ring_parse_cs(struct amdgpu_cs_parser *parser, uint32_t ib_idx); +int amdgpu_uvd_ring_parse_cs(struct amdgpu_cs_parser *parser, + struct amdgpu_job *job, + struct amdgpu_ib *ib); void amdgpu_uvd_ring_begin_use(struct amdgpu_ring *ring); void amdgpu_uvd_ring_end_use(struct amdgpu_ring *ring); int amdgpu_uvd_ring_test_ib(struct amdgpu_ring *ring, long timeout); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c index 344f711ad144..02cb3a12dd76 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c @@ -34,6 +34,7 @@ #include "amdgpu.h" #include "amdgpu_pm.h" #include "amdgpu_vce.h" +#include "amdgpu_cs.h" #include "cikd.h" /* 1 second timeout */ @@ -587,8 +588,7 @@ err: /** * amdgpu_vce_validate_bo - make sure not to cross 4GB boundary * - * @p: parser context - * @ib_idx: indirect buffer to use + * @ib: indirect buffer to use * @lo: address of lower dword * @hi: address of higher dword * @size: minimum size @@ -596,8 +596,9 @@ err: * * Make sure that no BO cross a 4GB boundary. */ -static int amdgpu_vce_validate_bo(struct amdgpu_cs_parser *p, uint32_t ib_idx, - int lo, int hi, unsigned size, int32_t index) +static int amdgpu_vce_validate_bo(struct amdgpu_cs_parser *p, + struct amdgpu_ib *ib, int lo, int hi, + unsigned size, int32_t index) { int64_t offset = ((uint64_t)size) * ((int64_t)index); struct ttm_operation_ctx ctx = { false, false }; @@ -607,8 +608,8 @@ static int amdgpu_vce_validate_bo(struct amdgpu_cs_parser *p, uint32_t ib_idx, uint64_t addr; int r; - addr = ((uint64_t)amdgpu_get_ib_value(p, ib_idx, lo)) | - ((uint64_t)amdgpu_get_ib_value(p, ib_idx, hi)) << 32; + addr = ((uint64_t)amdgpu_ib_get_value(ib, lo)) | + ((uint64_t)amdgpu_ib_get_value(ib, hi)) << 32; if (index >= 0) { addr += offset; fpfn = PAGE_ALIGN(offset) >> PAGE_SHIFT; @@ -638,7 +639,7 @@ static int amdgpu_vce_validate_bo(struct amdgpu_cs_parser *p, uint32_t ib_idx, * amdgpu_vce_cs_reloc - command submission relocation * * @p: parser context - * @ib_idx: indirect buffer to use + * @ib: indirect buffer to use * @lo: address of lower dword * @hi: address of higher dword * @size: minimum size @@ -646,7 +647,7 @@ static int amdgpu_vce_validate_bo(struct amdgpu_cs_parser *p, uint32_t ib_idx, * * Patch relocation inside command stream with real buffer address */ -static int amdgpu_vce_cs_reloc(struct amdgpu_cs_parser *p, uint32_t ib_idx, +static int amdgpu_vce_cs_reloc(struct amdgpu_cs_parser *p, struct amdgpu_ib *ib, int lo, int hi, unsigned size, uint32_t index) { struct amdgpu_bo_va_mapping *mapping; @@ -657,8 +658,8 @@ static int amdgpu_vce_cs_reloc(struct amdgpu_cs_parser *p, uint32_t ib_idx, if (index == 0xffffffff) index = 0; - addr = ((uint64_t)amdgpu_get_ib_value(p, ib_idx, lo)) | - ((uint64_t)amdgpu_get_ib_value(p, ib_idx, hi)) << 32; + addr = ((uint64_t)amdgpu_ib_get_value(ib, lo)) | + ((uint64_t)amdgpu_ib_get_value(ib, hi)) << 32; addr += ((uint64_t)size) * ((uint64_t)index); r = amdgpu_cs_find_mapping(p, addr, &bo, &mapping); @@ -679,8 +680,8 @@ static int amdgpu_vce_cs_reloc(struct amdgpu_cs_parser *p, uint32_t ib_idx, addr += amdgpu_bo_gpu_offset(bo); addr -= ((uint64_t)size) * ((uint64_t)index); - amdgpu_set_ib_value(p, ib_idx, lo, lower_32_bits(addr)); - amdgpu_set_ib_value(p, ib_idx, hi, upper_32_bits(addr)); + amdgpu_ib_set_value(ib, lo, lower_32_bits(addr)); + amdgpu_ib_set_value(ib, hi, upper_32_bits(addr)); return 0; } @@ -729,11 +730,13 @@ static int amdgpu_vce_validate_handle(struct amdgpu_cs_parser *p, * amdgpu_vce_ring_parse_cs - parse and validate the command stream * * @p: parser context - * @ib_idx: indirect buffer to use + * @job: the job to parse + * @ib: the IB to patch */ -int amdgpu_vce_ring_parse_cs(struct amdgpu_cs_parser *p, uint32_t ib_idx) +int amdgpu_vce_ring_parse_cs(struct amdgpu_cs_parser *p, + struct amdgpu_job *job, + struct amdgpu_ib *ib) { - struct amdgpu_ib *ib = &p->job->ibs[ib_idx]; unsigned fb_idx = 0, bs_idx = 0; int session_idx = -1; uint32_t destroyed = 0; @@ -744,12 +747,12 @@ int amdgpu_vce_ring_parse_cs(struct amdgpu_cs_parser *p, uint32_t ib_idx) unsigned idx; int i, r = 0; - p->job->vm = NULL; + job->vm = NULL; ib->gpu_addr = amdgpu_sa_bo_gpu_addr(ib->sa_bo); for (idx = 0; idx < ib->length_dw;) { - uint32_t len = amdgpu_get_ib_value(p, ib_idx, idx); - uint32_t cmd = amdgpu_get_ib_value(p, ib_idx, idx + 1); + uint32_t len = amdgpu_ib_get_value(ib, idx); + uint32_t cmd = amdgpu_ib_get_value(ib, idx + 1); if ((len < 8) || (len & 3)) { DRM_ERROR("invalid VCE command length (%d)!\n", len); @@ -759,52 +762,52 @@ int amdgpu_vce_ring_parse_cs(struct amdgpu_cs_parser *p, uint32_t ib_idx) switch (cmd) { case 0x00000002: /* task info */ - fb_idx = amdgpu_get_ib_value(p, ib_idx, idx + 6); - bs_idx = amdgpu_get_ib_value(p, ib_idx, idx + 7); + fb_idx = amdgpu_ib_get_value(ib, idx + 6); + bs_idx = amdgpu_ib_get_value(ib, idx + 7); break; case 0x03000001: /* encode */ - r = amdgpu_vce_validate_bo(p, ib_idx, idx + 10, - idx + 9, 0, 0); + r = amdgpu_vce_validate_bo(p, ib, idx + 10, idx + 9, + 0, 0); if (r) goto out; - r = amdgpu_vce_validate_bo(p, ib_idx, idx + 12, - idx + 11, 0, 0); + r = amdgpu_vce_validate_bo(p, ib, idx + 12, idx + 11, + 0, 0); if (r) goto out; break; case 0x05000001: /* context buffer */ - r = amdgpu_vce_validate_bo(p, ib_idx, idx + 3, - idx + 2, 0, 0); + r = amdgpu_vce_validate_bo(p, ib, idx + 3, idx + 2, + 0, 0); if (r) goto out; break; case 0x05000004: /* video bitstream buffer */ - tmp = amdgpu_get_ib_value(p, ib_idx, idx + 4); - r = amdgpu_vce_validate_bo(p, ib_idx, idx + 3, idx + 2, + tmp = amdgpu_ib_get_value(ib, idx + 4); + r = amdgpu_vce_validate_bo(p, ib, idx + 3, idx + 2, tmp, bs_idx); if (r) goto out; break; case 0x05000005: /* feedback buffer */ - r = amdgpu_vce_validate_bo(p, ib_idx, idx + 3, idx + 2, + r = amdgpu_vce_validate_bo(p, ib, idx + 3, idx + 2, 4096, fb_idx); if (r) goto out; break; case 0x0500000d: /* MV buffer */ - r = amdgpu_vce_validate_bo(p, ib_idx, idx + 3, - idx + 2, 0, 0); + r = amdgpu_vce_validate_bo(p, ib, idx + 3, idx + 2, + 0, 0); if (r) goto out; - r = amdgpu_vce_validate_bo(p, ib_idx, idx + 8, - idx + 7, 0, 0); + r = amdgpu_vce_validate_bo(p, ib, idx + 8, idx + 7, + 0, 0); if (r) goto out; break; @@ -814,12 +817,12 @@ int amdgpu_vce_ring_parse_cs(struct amdgpu_cs_parser *p, uint32_t ib_idx) } for (idx = 0; idx < ib->length_dw;) { - uint32_t len = amdgpu_get_ib_value(p, ib_idx, idx); - uint32_t cmd = amdgpu_get_ib_value(p, ib_idx, idx + 1); + uint32_t len = amdgpu_ib_get_value(ib, idx); + uint32_t cmd = amdgpu_ib_get_value(ib, idx + 1); switch (cmd) { case 0x00000001: /* session */ - handle = amdgpu_get_ib_value(p, ib_idx, idx + 2); + handle = amdgpu_ib_get_value(ib, idx + 2); session_idx = amdgpu_vce_validate_handle(p, handle, &allocated); if (session_idx < 0) { @@ -830,8 +833,8 @@ int amdgpu_vce_ring_parse_cs(struct amdgpu_cs_parser *p, uint32_t ib_idx) break; case 0x00000002: /* task info */ - fb_idx = amdgpu_get_ib_value(p, ib_idx, idx + 6); - bs_idx = amdgpu_get_ib_value(p, ib_idx, idx + 7); + fb_idx = amdgpu_ib_get_value(ib, idx + 6); + bs_idx = amdgpu_ib_get_value(ib, idx + 7); break; case 0x01000001: /* create */ @@ -846,8 +849,8 @@ int amdgpu_vce_ring_parse_cs(struct amdgpu_cs_parser *p, uint32_t ib_idx) goto out; } - *size = amdgpu_get_ib_value(p, ib_idx, idx + 8) * - amdgpu_get_ib_value(p, ib_idx, idx + 10) * + *size = amdgpu_ib_get_value(ib, idx + 8) * + amdgpu_ib_get_value(ib, idx + 10) * 8 * 3 / 2; break; @@ -876,12 +879,12 @@ int amdgpu_vce_ring_parse_cs(struct amdgpu_cs_parser *p, uint32_t ib_idx) break; case 0x03000001: /* encode */ - r = amdgpu_vce_cs_reloc(p, ib_idx, idx + 10, idx + 9, + r = amdgpu_vce_cs_reloc(p, ib, idx + 10, idx + 9, *size, 0); if (r) goto out; - r = amdgpu_vce_cs_reloc(p, ib_idx, idx + 12, idx + 11, + r = amdgpu_vce_cs_reloc(p, ib, idx + 12, idx + 11, *size / 3, 0); if (r) goto out; @@ -892,35 +895,35 @@ int amdgpu_vce_ring_parse_cs(struct amdgpu_cs_parser *p, uint32_t ib_idx) break; case 0x05000001: /* context buffer */ - r = amdgpu_vce_cs_reloc(p, ib_idx, idx + 3, idx + 2, + r = amdgpu_vce_cs_reloc(p, ib, idx + 3, idx + 2, *size * 2, 0); if (r) goto out; break; case 0x05000004: /* video bitstream buffer */ - tmp = amdgpu_get_ib_value(p, ib_idx, idx + 4); - r = amdgpu_vce_cs_reloc(p, ib_idx, idx + 3, idx + 2, + tmp = amdgpu_ib_get_value(ib, idx + 4); + r = amdgpu_vce_cs_reloc(p, ib, idx + 3, idx + 2, tmp, bs_idx); if (r) goto out; break; case 0x05000005: /* feedback buffer */ - r = amdgpu_vce_cs_reloc(p, ib_idx, idx + 3, idx + 2, + r = amdgpu_vce_cs_reloc(p, ib, idx + 3, idx + 2, 4096, fb_idx); if (r) goto out; break; case 0x0500000d: /* MV buffer */ - r = amdgpu_vce_cs_reloc(p, ib_idx, idx + 3, - idx + 2, *size, 0); + r = amdgpu_vce_cs_reloc(p, ib, idx + 3, + idx + 2, *size, 0); if (r) goto out; - r = amdgpu_vce_cs_reloc(p, ib_idx, idx + 8, - idx + 7, *size / 12, 0); + r = amdgpu_vce_cs_reloc(p, ib, idx + 8, + idx + 7, *size / 12, 0); if (r) goto out; break; @@ -965,11 +968,13 @@ out: * amdgpu_vce_ring_parse_cs_vm - parse the command stream in VM mode * * @p: parser context - * @ib_idx: indirect buffer to use + * @job: the job to parse + * @ib: the IB to patch */ -int amdgpu_vce_ring_parse_cs_vm(struct amdgpu_cs_parser *p, uint32_t ib_idx) +int amdgpu_vce_ring_parse_cs_vm(struct amdgpu_cs_parser *p, + struct amdgpu_job *job, + struct amdgpu_ib *ib) { - struct amdgpu_ib *ib = &p->job->ibs[ib_idx]; int session_idx = -1; uint32_t destroyed = 0; uint32_t created = 0; @@ -978,8 +983,8 @@ int amdgpu_vce_ring_parse_cs_vm(struct amdgpu_cs_parser *p, uint32_t ib_idx) int i, r = 0, idx = 0; while (idx < ib->length_dw) { - uint32_t len = amdgpu_get_ib_value(p, ib_idx, idx); - uint32_t cmd = amdgpu_get_ib_value(p, ib_idx, idx + 1); + uint32_t len = amdgpu_ib_get_value(ib, idx); + uint32_t cmd = amdgpu_ib_get_value(ib, idx + 1); if ((len < 8) || (len & 3)) { DRM_ERROR("invalid VCE command length (%d)!\n", len); @@ -989,7 +994,7 @@ int amdgpu_vce_ring_parse_cs_vm(struct amdgpu_cs_parser *p, uint32_t ib_idx) switch (cmd) { case 0x00000001: /* session */ - handle = amdgpu_get_ib_value(p, ib_idx, idx + 2); + handle = amdgpu_ib_get_value(ib, idx + 2); session_idx = amdgpu_vce_validate_handle(p, handle, &allocated); if (session_idx < 0) { diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.h index be4a6e773c5b..ea680fc9a6c3 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.h @@ -59,8 +59,11 @@ int amdgpu_vce_entity_init(struct amdgpu_device *adev); int amdgpu_vce_suspend(struct amdgpu_device *adev); int amdgpu_vce_resume(struct amdgpu_device *adev); void amdgpu_vce_free_handles(struct amdgpu_device *adev, struct drm_file *filp); -int amdgpu_vce_ring_parse_cs(struct amdgpu_cs_parser *p, uint32_t ib_idx); -int amdgpu_vce_ring_parse_cs_vm(struct amdgpu_cs_parser *p, uint32_t ib_idx); +int amdgpu_vce_ring_parse_cs(struct amdgpu_cs_parser *p, struct amdgpu_job *job, + struct amdgpu_ib *ib); +int amdgpu_vce_ring_parse_cs_vm(struct amdgpu_cs_parser *p, + struct amdgpu_job *job, + struct amdgpu_ib *ib); void amdgpu_vce_ring_emit_ib(struct amdgpu_ring *ring, struct amdgpu_job *job, struct amdgpu_ib *ib, uint32_t flags); void amdgpu_vce_ring_emit_fence(struct amdgpu_ring *ring, u64 addr, u64 seq, diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c index 9a19a6a57b23..0b52af415b28 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c @@ -27,6 +27,7 @@ #include <linux/firmware.h> #include <linux/module.h> #include <linux/pci.h> +#include <linux/debugfs.h> #include <drm/drm_drv.h> #include "amdgpu.h" @@ -51,6 +52,10 @@ #define FIRMWARE_ALDEBARAN "amdgpu/aldebaran_vcn.bin" #define FIRMWARE_BEIGE_GOBY "amdgpu/beige_goby_vcn.bin" #define FIRMWARE_YELLOW_CARP "amdgpu/yellow_carp_vcn.bin" +#define FIRMWARE_VCN_3_1_2 "amdgpu/vcn_3_1_2.bin" +#define FIRMWARE_VCN4_0_0 "amdgpu/vcn_4_0_0.bin" +#define FIRMWARE_VCN4_0_2 "amdgpu/vcn_4_0_2.bin" +#define FIRMWARE_VCN4_0_4 "amdgpu/vcn_4_0_4.bin" MODULE_FIRMWARE(FIRMWARE_RAVEN); MODULE_FIRMWARE(FIRMWARE_PICASSO); @@ -68,6 +73,10 @@ MODULE_FIRMWARE(FIRMWARE_VANGOGH); MODULE_FIRMWARE(FIRMWARE_DIMGREY_CAVEFISH); MODULE_FIRMWARE(FIRMWARE_BEIGE_GOBY); MODULE_FIRMWARE(FIRMWARE_YELLOW_CARP); +MODULE_FIRMWARE(FIRMWARE_VCN_3_1_2); +MODULE_FIRMWARE(FIRMWARE_VCN4_0_0); +MODULE_FIRMWARE(FIRMWARE_VCN4_0_2); +MODULE_FIRMWARE(FIRMWARE_VCN4_0_4); static void amdgpu_vcn_idle_work_handler(struct work_struct *work); @@ -77,6 +86,7 @@ int amdgpu_vcn_sw_init(struct amdgpu_device *adev) const char *fw_name; const struct common_firmware_header *hdr; unsigned char fw_check; + unsigned int fw_shared_size, log_offset; int i, r; INIT_DELAYED_WORK(&adev->vcn.idle_work, amdgpu_vcn_idle_work_handler); @@ -165,6 +175,30 @@ int amdgpu_vcn_sw_init(struct amdgpu_device *adev) (adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG)) adev->vcn.indirect_sram = true; break; + case IP_VERSION(3, 1, 2): + fw_name = FIRMWARE_VCN_3_1_2; + if ((adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) && + (adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG)) + adev->vcn.indirect_sram = true; + break; + case IP_VERSION(4, 0, 0): + fw_name = FIRMWARE_VCN4_0_0; + if ((adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) && + (adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG)) + adev->vcn.indirect_sram = true; + break; + case IP_VERSION(4, 0, 2): + fw_name = FIRMWARE_VCN4_0_2; + if ((adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) && + (adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG)) + adev->vcn.indirect_sram = true; + break; + case IP_VERSION(4, 0, 4): + fw_name = FIRMWARE_VCN4_0_4; + if ((adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) && + (adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG)) + adev->vcn.indirect_sram = true; + break; default: return -EINVAL; } @@ -218,7 +252,19 @@ int amdgpu_vcn_sw_init(struct amdgpu_device *adev) bo_size = AMDGPU_VCN_STACK_SIZE + AMDGPU_VCN_CONTEXT_SIZE; if (adev->firmware.load_type != AMDGPU_FW_LOAD_PSP) bo_size += AMDGPU_GPU_PAGE_ALIGN(le32_to_cpu(hdr->ucode_size_bytes) + 8); - bo_size += AMDGPU_GPU_PAGE_ALIGN(sizeof(struct amdgpu_fw_shared)); + + if (adev->ip_versions[UVD_HWIP][0] >= IP_VERSION(4, 0, 0)){ + fw_shared_size = AMDGPU_GPU_PAGE_ALIGN(sizeof(struct amdgpu_vcn4_fw_shared)); + log_offset = offsetof(struct amdgpu_vcn4_fw_shared, fw_log); + } else { + fw_shared_size = AMDGPU_GPU_PAGE_ALIGN(sizeof(struct amdgpu_fw_shared)); + log_offset = offsetof(struct amdgpu_fw_shared, fw_log); + } + + bo_size += fw_shared_size; + + if (amdgpu_vcnfw_log) + bo_size += AMDGPU_VCNFW_LOG_SIZE; for (i = 0; i < adev->vcn.num_vcn_inst; i++) { if (adev->vcn.harvest_config & (1 << i)) @@ -232,10 +278,18 @@ int amdgpu_vcn_sw_init(struct amdgpu_device *adev) return r; } - adev->vcn.inst[i].fw_shared_cpu_addr = adev->vcn.inst[i].cpu_addr + - bo_size - AMDGPU_GPU_PAGE_ALIGN(sizeof(struct amdgpu_fw_shared)); - adev->vcn.inst[i].fw_shared_gpu_addr = adev->vcn.inst[i].gpu_addr + - bo_size - AMDGPU_GPU_PAGE_ALIGN(sizeof(struct amdgpu_fw_shared)); + adev->vcn.inst[i].fw_shared.cpu_addr = adev->vcn.inst[i].cpu_addr + + bo_size - fw_shared_size; + adev->vcn.inst[i].fw_shared.gpu_addr = adev->vcn.inst[i].gpu_addr + + bo_size - fw_shared_size; + + adev->vcn.inst[i].fw_shared.mem_size = fw_shared_size; + + if (amdgpu_vcnfw_log) { + adev->vcn.inst[i].fw_shared.cpu_addr -= AMDGPU_VCNFW_LOG_SIZE; + adev->vcn.inst[i].fw_shared.gpu_addr -= AMDGPU_VCNFW_LOG_SIZE; + adev->vcn.inst[i].fw_shared.log_offset = log_offset; + } if (adev->vcn.indirect_sram) { r = amdgpu_bo_create_kernel(adev, 64 * 2 * 4, PAGE_SIZE, @@ -283,6 +337,18 @@ int amdgpu_vcn_sw_fini(struct amdgpu_device *adev) return 0; } +/* from vcn4 and above, only unified queue is used */ +static bool amdgpu_vcn_using_unified_queue(struct amdgpu_ring *ring) +{ + struct amdgpu_device *adev = ring->adev; + bool ret = false; + + if (adev->ip_versions[UVD_HWIP][0] >= IP_VERSION(4, 0, 0)) + ret = true; + + return ret; +} + bool amdgpu_vcn_is_disabled_vcn(struct amdgpu_device *adev, enum vcn_ring_type type, uint32_t vcn_instance) { bool ret = false; @@ -672,19 +738,55 @@ error: return r; } +static uint32_t *amdgpu_vcn_unified_ring_ib_header(struct amdgpu_ib *ib, + uint32_t ib_pack_in_dw, bool enc) +{ + uint32_t *ib_checksum; + + ib->ptr[ib->length_dw++] = 0x00000010; /* single queue checksum */ + ib->ptr[ib->length_dw++] = 0x30000002; + ib_checksum = &ib->ptr[ib->length_dw++]; + ib->ptr[ib->length_dw++] = ib_pack_in_dw; + + ib->ptr[ib->length_dw++] = 0x00000010; /* engine info */ + ib->ptr[ib->length_dw++] = 0x30000001; + ib->ptr[ib->length_dw++] = enc ? 0x2 : 0x3; + ib->ptr[ib->length_dw++] = ib_pack_in_dw * sizeof(uint32_t); + + return ib_checksum; +} + +static void amdgpu_vcn_unified_ring_ib_checksum(uint32_t **ib_checksum, + uint32_t ib_pack_in_dw) +{ + uint32_t i; + uint32_t checksum = 0; + + for (i = 0; i < ib_pack_in_dw; i++) + checksum += *(*ib_checksum + 2 + i); + + **ib_checksum = checksum; +} + static int amdgpu_vcn_dec_sw_send_msg(struct amdgpu_ring *ring, struct amdgpu_ib *ib_msg, struct dma_fence **fence) { struct amdgpu_vcn_decode_buffer *decode_buffer = NULL; - const unsigned int ib_size_dw = 64; + unsigned int ib_size_dw = 64; struct amdgpu_device *adev = ring->adev; struct dma_fence *f = NULL; struct amdgpu_job *job; struct amdgpu_ib *ib; uint64_t addr = AMDGPU_GPU_PAGE_ALIGN(ib_msg->gpu_addr); + bool sq = amdgpu_vcn_using_unified_queue(ring); + uint32_t *ib_checksum; + uint32_t ib_pack_in_dw; int i, r; + if (sq) + ib_size_dw += 8; + r = amdgpu_job_alloc_with_ib(adev, ib_size_dw * 4, AMDGPU_IB_POOL_DIRECT, &job); if (r) @@ -693,6 +795,13 @@ static int amdgpu_vcn_dec_sw_send_msg(struct amdgpu_ring *ring, ib = &job->ibs[0]; ib->length_dw = 0; + /* single queue headers */ + if (sq) { + ib_pack_in_dw = sizeof(struct amdgpu_vcn_decode_buffer) / sizeof(uint32_t) + + 4 + 2; /* engine info + decoding ib in dw */ + ib_checksum = amdgpu_vcn_unified_ring_ib_header(ib, ib_pack_in_dw, false); + } + ib->ptr[ib->length_dw++] = sizeof(struct amdgpu_vcn_decode_buffer) + 8; ib->ptr[ib->length_dw++] = cpu_to_le32(AMDGPU_VCN_IB_FLAG_DECODE_BUFFER); decode_buffer = (struct amdgpu_vcn_decode_buffer *)&(ib->ptr[ib->length_dw]); @@ -706,6 +815,9 @@ static int amdgpu_vcn_dec_sw_send_msg(struct amdgpu_ring *ring, for (i = ib->length_dw; i < ib_size_dw; ++i) ib->ptr[i] = 0x0; + if (sq) + amdgpu_vcn_unified_ring_ib_checksum(&ib_checksum, ib_pack_in_dw); + r = amdgpu_job_submit_direct(job, ring, &f); if (r) goto err_free; @@ -792,13 +904,18 @@ static int amdgpu_vcn_enc_get_create_msg(struct amdgpu_ring *ring, uint32_t hand struct amdgpu_ib *ib_msg, struct dma_fence **fence) { - const unsigned ib_size_dw = 16; + unsigned int ib_size_dw = 16; struct amdgpu_job *job; struct amdgpu_ib *ib; struct dma_fence *f = NULL; + uint32_t *ib_checksum = NULL; uint64_t addr; + bool sq = amdgpu_vcn_using_unified_queue(ring); int i, r; + if (sq) + ib_size_dw += 8; + r = amdgpu_job_alloc_with_ib(ring->adev, ib_size_dw * 4, AMDGPU_IB_POOL_DIRECT, &job); if (r) @@ -808,6 +925,10 @@ static int amdgpu_vcn_enc_get_create_msg(struct amdgpu_ring *ring, uint32_t hand addr = AMDGPU_GPU_PAGE_ALIGN(ib_msg->gpu_addr); ib->length_dw = 0; + + if (sq) + ib_checksum = amdgpu_vcn_unified_ring_ib_header(ib, 0x11, true); + ib->ptr[ib->length_dw++] = 0x00000018; ib->ptr[ib->length_dw++] = 0x00000001; /* session info */ ib->ptr[ib->length_dw++] = handle; @@ -827,6 +948,9 @@ static int amdgpu_vcn_enc_get_create_msg(struct amdgpu_ring *ring, uint32_t hand for (i = ib->length_dw; i < ib_size_dw; ++i) ib->ptr[i] = 0x0; + if (sq) + amdgpu_vcn_unified_ring_ib_checksum(&ib_checksum, 0x11); + r = amdgpu_job_submit_direct(job, ring, &f); if (r) goto err; @@ -846,13 +970,18 @@ static int amdgpu_vcn_enc_get_destroy_msg(struct amdgpu_ring *ring, uint32_t han struct amdgpu_ib *ib_msg, struct dma_fence **fence) { - const unsigned ib_size_dw = 16; + unsigned int ib_size_dw = 16; struct amdgpu_job *job; struct amdgpu_ib *ib; struct dma_fence *f = NULL; + uint32_t *ib_checksum = NULL; uint64_t addr; + bool sq = amdgpu_vcn_using_unified_queue(ring); int i, r; + if (sq) + ib_size_dw += 8; + r = amdgpu_job_alloc_with_ib(ring->adev, ib_size_dw * 4, AMDGPU_IB_POOL_DIRECT, &job); if (r) @@ -862,6 +991,10 @@ static int amdgpu_vcn_enc_get_destroy_msg(struct amdgpu_ring *ring, uint32_t han addr = AMDGPU_GPU_PAGE_ALIGN(ib_msg->gpu_addr); ib->length_dw = 0; + + if (sq) + ib_checksum = amdgpu_vcn_unified_ring_ib_header(ib, 0x11, true); + ib->ptr[ib->length_dw++] = 0x00000018; ib->ptr[ib->length_dw++] = 0x00000001; ib->ptr[ib->length_dw++] = handle; @@ -881,6 +1014,9 @@ static int amdgpu_vcn_enc_get_destroy_msg(struct amdgpu_ring *ring, uint32_t han for (i = ib->length_dw; i < ib_size_dw; ++i) ib->ptr[i] = 0x0; + if (sq) + amdgpu_vcn_unified_ring_ib_checksum(&ib_checksum, 0x11); + r = amdgpu_job_submit_direct(job, ring, &f); if (r) goto err; @@ -931,6 +1067,20 @@ error: return r; } +int amdgpu_vcn_unified_ring_test_ib(struct amdgpu_ring *ring, long timeout) +{ + long r; + + r = amdgpu_vcn_enc_ring_test_ib(ring, timeout); + if (r) + goto error; + + r = amdgpu_vcn_dec_sw_ring_test_ib(ring, timeout); + +error: + return r; +} + enum amdgpu_ring_priority_level amdgpu_vcn_get_enc_ring_prio(int ring) { switch(ring) { @@ -971,3 +1121,130 @@ void amdgpu_vcn_setup_ucode(struct amdgpu_device *adev) dev_info(adev->dev, "Will use PSP to load VCN firmware\n"); } } + +/* + * debugfs for mapping vcn firmware log buffer. + */ +#if defined(CONFIG_DEBUG_FS) +static ssize_t amdgpu_debugfs_vcn_fwlog_read(struct file *f, char __user *buf, + size_t size, loff_t *pos) +{ + struct amdgpu_vcn_inst *vcn; + void *log_buf; + volatile struct amdgpu_vcn_fwlog *plog; + unsigned int read_pos, write_pos, available, i, read_bytes = 0; + unsigned int read_num[2] = {0}; + + vcn = file_inode(f)->i_private; + if (!vcn) + return -ENODEV; + + if (!vcn->fw_shared.cpu_addr || !amdgpu_vcnfw_log) + return -EFAULT; + + log_buf = vcn->fw_shared.cpu_addr + vcn->fw_shared.mem_size; + + plog = (volatile struct amdgpu_vcn_fwlog *)log_buf; + read_pos = plog->rptr; + write_pos = plog->wptr; + + if (read_pos > AMDGPU_VCNFW_LOG_SIZE || write_pos > AMDGPU_VCNFW_LOG_SIZE) + return -EFAULT; + + if (!size || (read_pos == write_pos)) + return 0; + + if (write_pos > read_pos) { + available = write_pos - read_pos; + read_num[0] = min(size, (size_t)available); + } else { + read_num[0] = AMDGPU_VCNFW_LOG_SIZE - read_pos; + available = read_num[0] + write_pos - plog->header_size; + if (size > available) + read_num[1] = write_pos - plog->header_size; + else if (size > read_num[0]) + read_num[1] = size - read_num[0]; + else + read_num[0] = size; + } + + for (i = 0; i < 2; i++) { + if (read_num[i]) { + if (read_pos == AMDGPU_VCNFW_LOG_SIZE) + read_pos = plog->header_size; + if (read_num[i] == copy_to_user((buf + read_bytes), + (log_buf + read_pos), read_num[i])) + return -EFAULT; + + read_bytes += read_num[i]; + read_pos += read_num[i]; + } + } + + plog->rptr = read_pos; + *pos += read_bytes; + return read_bytes; +} + +static const struct file_operations amdgpu_debugfs_vcnfwlog_fops = { + .owner = THIS_MODULE, + .read = amdgpu_debugfs_vcn_fwlog_read, + .llseek = default_llseek +}; +#endif + +void amdgpu_debugfs_vcn_fwlog_init(struct amdgpu_device *adev, uint8_t i, + struct amdgpu_vcn_inst *vcn) +{ +#if defined(CONFIG_DEBUG_FS) + struct drm_minor *minor = adev_to_drm(adev)->primary; + struct dentry *root = minor->debugfs_root; + char name[32]; + + sprintf(name, "amdgpu_vcn_%d_fwlog", i); + debugfs_create_file_size(name, S_IFREG | S_IRUGO, root, vcn, + &amdgpu_debugfs_vcnfwlog_fops, + AMDGPU_VCNFW_LOG_SIZE); +#endif +} + +void amdgpu_vcn_fwlog_init(struct amdgpu_vcn_inst *vcn) +{ +#if defined(CONFIG_DEBUG_FS) + volatile uint32_t *flag = vcn->fw_shared.cpu_addr; + void *fw_log_cpu_addr = vcn->fw_shared.cpu_addr + vcn->fw_shared.mem_size; + uint64_t fw_log_gpu_addr = vcn->fw_shared.gpu_addr + vcn->fw_shared.mem_size; + volatile struct amdgpu_vcn_fwlog *log_buf = fw_log_cpu_addr; + volatile struct amdgpu_fw_shared_fw_logging *fw_log = vcn->fw_shared.cpu_addr + + vcn->fw_shared.log_offset; + *flag |= cpu_to_le32(AMDGPU_VCN_FW_LOGGING_FLAG); + fw_log->is_enabled = 1; + fw_log->addr_lo = cpu_to_le32(fw_log_gpu_addr & 0xFFFFFFFF); + fw_log->addr_hi = cpu_to_le32(fw_log_gpu_addr >> 32); + fw_log->size = cpu_to_le32(AMDGPU_VCNFW_LOG_SIZE); + + log_buf->header_size = sizeof(struct amdgpu_vcn_fwlog); + log_buf->buffer_size = AMDGPU_VCNFW_LOG_SIZE; + log_buf->rptr = log_buf->header_size; + log_buf->wptr = log_buf->header_size; + log_buf->wrapped = 0; +#endif +} + +int amdgpu_vcn_process_poison_irq(struct amdgpu_device *adev, + struct amdgpu_irq_src *source, + struct amdgpu_iv_entry *entry) +{ + struct ras_common_if *ras_if = adev->vcn.ras_if; + struct ras_dispatch_if ih_data = { + .entry = entry, + }; + + if (!ras_if) + return 0; + + ih_data.head = *ras_if; + amdgpu_ras_interrupt_dispatch(adev, &ih_data); + + return 0; +} diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.h index 5d3728b027d3..253ea6b159df 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.h @@ -24,6 +24,8 @@ #ifndef __AMDGPU_VCN_H__ #define __AMDGPU_VCN_H__ +#include "amdgpu_ras.h" + #define AMDGPU_VCN_STACK_SIZE (128*1024) #define AMDGPU_VCN_CONTEXT_SIZE (512*1024) @@ -63,8 +65,6 @@ #define VCN_ENC_CMD_REG_WRITE 0x0000000b #define VCN_ENC_CMD_REG_WAIT 0x0000000c -#define VCN_VID_SOC_ADDRESS_2_0 0x1fa00 -#define VCN1_VID_SOC_ADDRESS_3_0 0x48200 #define VCN_AON_SOC_ADDRESS_2_0 0x1f800 #define VCN1_AON_SOC_ADDRESS_3_0 0x48000 #define VCN_VID_IP_ADDRESS_2_0 0x0 @@ -155,13 +155,26 @@ } \ } while (0) +#define AMDGPU_FW_SHARED_FLAG_0_UNIFIED_QUEUE (1 << 2) #define AMDGPU_VCN_FW_SHARED_FLAG_0_RB (1 << 6) #define AMDGPU_VCN_MULTI_QUEUE_FLAG (1 << 8) #define AMDGPU_VCN_SW_RING_FLAG (1 << 9) +#define AMDGPU_VCN_FW_LOGGING_FLAG (1 << 10) +#define AMDGPU_VCN_SMU_VERSION_INFO_FLAG (1 << 11) +#define AMDGPU_VCN_SMU_DPM_INTERFACE_FLAG (1 << 11) +#define AMDGPU_VCN_VF_RB_SETUP_FLAG (1 << 14) #define AMDGPU_VCN_IB_FLAG_DECODE_BUFFER 0x00000001 #define AMDGPU_VCN_CMD_FLAG_MSG_BUFFER 0x00000001 +#define VCN_CODEC_DISABLE_MASK_AV1 (1 << 0) +#define VCN_CODEC_DISABLE_MASK_VP9 (1 << 1) +#define VCN_CODEC_DISABLE_MASK_HEVC (1 << 2) +#define VCN_CODEC_DISABLE_MASK_H264 (1 << 3) + +#define AMDGPU_VCN_SMU_DPM_INTERFACE_DGPU (0) +#define AMDGPU_VCN_SMU_DPM_INTERFACE_APU (1) + enum fw_queue_mode { FW_QUEUE_RING_RESET = 1, FW_QUEUE_DPG_HOLD_OFF = 2, @@ -205,6 +218,13 @@ struct amdgpu_vcn_reg{ unsigned scratch9; }; +struct amdgpu_vcn_fw_shared { + void *cpu_addr; + uint64_t gpu_addr; + uint32_t mem_size; + uint32_t log_offset; +}; + struct amdgpu_vcn_inst { struct amdgpu_bo *vcpu_bo; void *cpu_addr; @@ -221,8 +241,11 @@ struct amdgpu_vcn_inst { uint64_t dpg_sram_gpu_addr; uint32_t *dpg_sram_curr_addr; atomic_t dpg_enc_submission_cnt; - void *fw_shared_cpu_addr; - uint64_t fw_shared_gpu_addr; + struct amdgpu_vcn_fw_shared fw_shared; +}; + +struct amdgpu_vcn_ras { + struct amdgpu_ras_block_object ras_block; }; struct amdgpu_vcn { @@ -236,6 +259,7 @@ struct amdgpu_vcn { uint8_t num_vcn_inst; struct amdgpu_vcn_inst inst[AMDGPU_MAX_VCN_INSTANCES]; uint8_t vcn_config[AMDGPU_MAX_VCN_INSTANCES]; + uint32_t vcn_codec_disable_mask[AMDGPU_MAX_VCN_INSTANCES]; struct amdgpu_vcn_reg internal; struct mutex vcn_pg_lock; struct mutex vcn1_jpeg1_workaround; @@ -244,6 +268,9 @@ struct amdgpu_vcn { unsigned harvest_config; int (*pause_dpg_mode)(struct amdgpu_device *adev, int inst_idx, struct dpg_pause_state *new_state); + + struct ras_common_if *ras_if; + struct amdgpu_vcn_ras *ras; }; struct amdgpu_fw_shared_rb_ptrs_struct { @@ -265,6 +292,25 @@ struct amdgpu_fw_shared_sw_ring { uint8_t padding[3]; }; +struct amdgpu_fw_shared_unified_queue_struct { + uint8_t is_enabled; + uint8_t queue_mode; + uint8_t queue_status; + uint8_t padding[5]; +}; + +struct amdgpu_fw_shared_fw_logging { + uint8_t is_enabled; + uint32_t addr_lo; + uint32_t addr_hi; + uint32_t size; +}; + +struct amdgpu_fw_shared_smu_interface_info { + uint8_t smu_interface_type; + uint8_t padding[3]; +}; + struct amdgpu_fw_shared { uint32_t present_flag_0; uint8_t pad[44]; @@ -272,6 +318,38 @@ struct amdgpu_fw_shared { uint8_t pad1[1]; struct amdgpu_fw_shared_multi_queue multi_queue; struct amdgpu_fw_shared_sw_ring sw_ring; + struct amdgpu_fw_shared_fw_logging fw_log; + struct amdgpu_fw_shared_smu_interface_info smu_interface_info; +}; + +struct amdgpu_fw_shared_rb_setup { + uint32_t is_rb_enabled_flags; + uint32_t rb_addr_lo; + uint32_t rb_addr_hi; + uint32_t rb_size; + uint32_t rb4_addr_lo; + uint32_t rb4_addr_hi; + uint32_t rb4_size; + uint32_t reserved[6]; +}; + +struct amdgpu_vcn4_fw_shared { + uint32_t present_flag_0; + uint8_t pad[12]; + struct amdgpu_fw_shared_unified_queue_struct sq; + uint8_t pad1[8]; + struct amdgpu_fw_shared_fw_logging fw_log; + uint8_t pad2[20]; + struct amdgpu_fw_shared_rb_setup rb_setup; + struct amdgpu_fw_shared_smu_interface_info smu_dpm_interface; +}; + +struct amdgpu_vcn_fwlog { + uint32_t rptr; + uint32_t wptr; + uint32_t buffer_size; + uint32_t header_size; + uint8_t wrapped; }; struct amdgpu_vcn_decode_buffer { @@ -305,6 +383,7 @@ int amdgpu_vcn_dec_ring_test_ring(struct amdgpu_ring *ring); int amdgpu_vcn_dec_ring_test_ib(struct amdgpu_ring *ring, long timeout); int amdgpu_vcn_dec_sw_ring_test_ring(struct amdgpu_ring *ring); int amdgpu_vcn_dec_sw_ring_test_ib(struct amdgpu_ring *ring, long timeout); +int amdgpu_vcn_unified_ring_test_ib(struct amdgpu_ring *ring, long timeout); int amdgpu_vcn_enc_ring_test_ring(struct amdgpu_ring *ring); int amdgpu_vcn_enc_ring_test_ib(struct amdgpu_ring *ring, long timeout); @@ -313,4 +392,12 @@ enum amdgpu_ring_priority_level amdgpu_vcn_get_enc_ring_prio(int ring); void amdgpu_vcn_setup_ucode(struct amdgpu_device *adev); +void amdgpu_vcn_fwlog_init(struct amdgpu_vcn_inst *vcn); +void amdgpu_debugfs_vcn_fwlog_init(struct amdgpu_device *adev, + uint8_t i, struct amdgpu_vcn_inst *vcn); + +int amdgpu_vcn_process_poison_irq(struct amdgpu_device *adev, + struct amdgpu_irq_src *source, + struct amdgpu_iv_entry *entry); + #endif diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.c index f8e574cc0e22..c73abe54d974 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.c @@ -23,7 +23,12 @@ #include <linux/module.h> +#ifdef CONFIG_X86 +#include <asm/hypervisor.h> +#endif + #include <drm/drm_drv.h> +#include <xen/xen.h> #include "amdgpu.h" #include "amdgpu_ras.h" @@ -71,6 +76,12 @@ void amdgpu_virt_kiq_reg_write_reg_wait(struct amdgpu_device *adev, unsigned long flags; uint32_t seq; + if (adev->mes.ring.sched.ready) { + amdgpu_mes_reg_write_reg_wait(adev, reg0, reg1, + ref, mask); + return; + } + spin_lock_irqsave(&kiq->ring_lock, flags); amdgpu_ring_alloc(ring, 32); amdgpu_ring_emit_reg_write_reg_wait(ring, reg0, reg1, @@ -536,6 +547,7 @@ static void amdgpu_virt_populate_vf2pf_ucode_info(struct amdgpu_device *adev) POPULATE_UCODE_INFO(vf2pf_info, AMD_SRIOV_UCODE_ID_RLC_SRLS, adev->gfx.rlc_srls_fw_version); POPULATE_UCODE_INFO(vf2pf_info, AMD_SRIOV_UCODE_ID_MEC, adev->gfx.mec_fw_version); POPULATE_UCODE_INFO(vf2pf_info, AMD_SRIOV_UCODE_ID_MEC2, adev->gfx.mec2_fw_version); + POPULATE_UCODE_INFO(vf2pf_info, AMD_SRIOV_UCODE_ID_IMU, adev->gfx.imu_fw_version); POPULATE_UCODE_INFO(vf2pf_info, AMD_SRIOV_UCODE_ID_SOS, adev->psp.sos.fw_version); POPULATE_UCODE_INFO(vf2pf_info, AMD_SRIOV_UCODE_ID_ASD, adev->psp.asd_context.bin_desc.fw_version); @@ -553,7 +565,6 @@ static void amdgpu_virt_populate_vf2pf_ucode_info(struct amdgpu_device *adev) static int amdgpu_virt_write_vf2pf_data(struct amdgpu_device *adev) { struct amd_sriov_msg_vf2pf_info *vf2pf_info; - struct ttm_resource_manager *vram_man = ttm_manager_type(&adev->mman.bdev, TTM_PL_VRAM); vf2pf_info = (struct amd_sriov_msg_vf2pf_info *) adev->virt.fw_reserve.p_vf2pf; @@ -576,8 +587,10 @@ static int amdgpu_virt_write_vf2pf_data(struct amdgpu_device *adev) vf2pf_info->driver_cert = 0; vf2pf_info->os_info.all = 0; - vf2pf_info->fb_usage = amdgpu_vram_mgr_usage(vram_man) >> 20; - vf2pf_info->fb_vis_usage = amdgpu_vram_mgr_vis_usage(vram_man) >> 20; + vf2pf_info->fb_usage = + ttm_resource_manager_usage(&adev->mman.vram_mgr.manager) >> 20; + vf2pf_info->fb_vis_usage = + amdgpu_vram_mgr_vis_usage(&adev->mman.vram_mgr) >> 20; vf2pf_info->fb_size = adev->gmc.real_vram_size >> 20; vf2pf_info->fb_vis_size = adev->gmc.visible_vram_size >> 20; @@ -626,20 +639,20 @@ void amdgpu_virt_init_data_exchange(struct amdgpu_device *adev) adev->virt.fw_reserve.p_vf2pf = NULL; adev->virt.vf2pf_update_interval_ms = 0; - if (adev->bios != NULL) { - adev->virt.vf2pf_update_interval_ms = 2000; + if (adev->mman.fw_vram_usage_va != NULL) { + /* go through this logic in ip_init and reset to init workqueue*/ + amdgpu_virt_exchange_data(adev); + INIT_DELAYED_WORK(&adev->virt.vf2pf_work, amdgpu_virt_update_vf2pf_work_item); + schedule_delayed_work(&(adev->virt.vf2pf_work), msecs_to_jiffies(adev->virt.vf2pf_update_interval_ms)); + } else if (adev->bios != NULL) { + /* got through this logic in early init stage to get necessary flags, e.g. rlcg_acc related*/ adev->virt.fw_reserve.p_pf2vf = (struct amd_sriov_msg_pf2vf_info_header *) (adev->bios + (AMD_SRIOV_MSG_PF2VF_OFFSET_KB << 10)); amdgpu_virt_read_pf2vf_data(adev); } - - if (adev->virt.vf2pf_update_interval_ms != 0) { - INIT_DELAYED_WORK(&adev->virt.vf2pf_work, amdgpu_virt_update_vf2pf_work_item); - schedule_delayed_work(&(adev->virt.vf2pf_work), msecs_to_jiffies(adev->virt.vf2pf_update_interval_ms)); - } } @@ -675,16 +688,9 @@ void amdgpu_virt_exchange_data(struct amdgpu_device *adev) if (adev->virt.ras_init_done) amdgpu_virt_add_bad_page(adev, bp_block_offset, bp_block_size); } - } else if (adev->bios != NULL) { - adev->virt.fw_reserve.p_pf2vf = - (struct amd_sriov_msg_pf2vf_info_header *) - (adev->bios + (AMD_SRIOV_MSG_PF2VF_OFFSET_KB << 10)); - - amdgpu_virt_read_pf2vf_data(adev); } } - void amdgpu_detect_virtualization(struct amdgpu_device *adev) { uint32_t reg; @@ -701,6 +707,7 @@ void amdgpu_detect_virtualization(struct amdgpu_device *adev) case CHIP_SIENNA_CICHLID: case CHIP_ARCTURUS: case CHIP_ALDEBARAN: + case CHIP_IP_DISCOVERY: reg = RREG32(mmRCC_IOV_FUNC_IDENTIFIER); break; default: /* other chip doesn't support SRIOV */ @@ -715,10 +722,17 @@ void amdgpu_detect_virtualization(struct amdgpu_device *adev) adev->virt.caps |= AMDGPU_SRIOV_CAPS_ENABLE_IOV; if (!reg) { - if (is_virtual_machine()) /* passthrough mode exclus sriov mod */ + /* passthrough mode exclus sriov mod */ + if (is_virtual_machine() && !xen_initial_domain()) adev->virt.caps |= AMDGPU_PASSTHROUGH_MODE; } + if (amdgpu_sriov_vf(adev) && adev->asic_type == CHIP_SIENNA_CICHLID) + /* VF MMIO access (except mailbox range) from CPU + * will be blocked during sriov runtime + */ + adev->virt.caps |= AMDGPU_VF_MMIO_ACCESS_PROTECT; + /* we have the ability to check now */ if (amdgpu_sriov_vf(adev)) { switch (adev->asic_type) { @@ -727,6 +741,14 @@ void amdgpu_detect_virtualization(struct amdgpu_device *adev) vi_set_virt_ops(adev); break; case CHIP_VEGA10: + soc15_set_virt_ops(adev); +#ifdef CONFIG_X86 + /* not send GPU_INIT_DATA with MS_HYPERV*/ + if (!hypervisor_is_type(X86_HYPER_MS_HYPERV)) +#endif + /* send a dummy GPU_INIT_DATA request to host on vega10 */ + amdgpu_virt_request_init_data(adev); + break; case CHIP_VEGA20: case CHIP_ARCTURUS: case CHIP_ALDEBARAN: @@ -735,6 +757,7 @@ void amdgpu_detect_virtualization(struct amdgpu_device *adev) case CHIP_NAVI10: case CHIP_NAVI12: case CHIP_SIENNA_CICHLID: + case CHIP_IP_DISCOVERY: nv_set_virt_ops(adev); /* try send GPU_INIT_DATA request to host */ amdgpu_virt_request_init_data(adev); @@ -792,6 +815,60 @@ enum amdgpu_sriov_vf_mode amdgpu_virt_get_sriov_vf_mode(struct amdgpu_device *ad return mode; } +bool amdgpu_virt_fw_load_skip_check(struct amdgpu_device *adev, uint32_t ucode_id) +{ + switch (adev->ip_versions[MP0_HWIP][0]) { + case IP_VERSION(13, 0, 0): + /* no vf autoload, white list */ + if (ucode_id == AMDGPU_UCODE_ID_VCN1 || + ucode_id == AMDGPU_UCODE_ID_VCN) + return false; + else + return true; + case IP_VERSION(13, 0, 10): + /* white list */ + if (ucode_id == AMDGPU_UCODE_ID_CAP + || ucode_id == AMDGPU_UCODE_ID_CP_RS64_PFP + || ucode_id == AMDGPU_UCODE_ID_CP_RS64_ME + || ucode_id == AMDGPU_UCODE_ID_CP_RS64_MEC + || ucode_id == AMDGPU_UCODE_ID_CP_RS64_PFP_P0_STACK + || ucode_id == AMDGPU_UCODE_ID_CP_RS64_PFP_P1_STACK + || ucode_id == AMDGPU_UCODE_ID_CP_RS64_ME_P0_STACK + || ucode_id == AMDGPU_UCODE_ID_CP_RS64_ME_P1_STACK + || ucode_id == AMDGPU_UCODE_ID_CP_RS64_MEC_P0_STACK + || ucode_id == AMDGPU_UCODE_ID_CP_RS64_MEC_P1_STACK + || ucode_id == AMDGPU_UCODE_ID_CP_RS64_MEC_P2_STACK + || ucode_id == AMDGPU_UCODE_ID_CP_RS64_MEC_P3_STACK + || ucode_id == AMDGPU_UCODE_ID_CP_MES + || ucode_id == AMDGPU_UCODE_ID_CP_MES_DATA + || ucode_id == AMDGPU_UCODE_ID_CP_MES1 + || ucode_id == AMDGPU_UCODE_ID_CP_MES1_DATA + || ucode_id == AMDGPU_UCODE_ID_VCN1 + || ucode_id == AMDGPU_UCODE_ID_VCN) + return false; + else + return true; + default: + /* lagacy black list */ + if (ucode_id == AMDGPU_UCODE_ID_SDMA0 + || ucode_id == AMDGPU_UCODE_ID_SDMA1 + || ucode_id == AMDGPU_UCODE_ID_SDMA2 + || ucode_id == AMDGPU_UCODE_ID_SDMA3 + || ucode_id == AMDGPU_UCODE_ID_SDMA4 + || ucode_id == AMDGPU_UCODE_ID_SDMA5 + || ucode_id == AMDGPU_UCODE_ID_SDMA6 + || ucode_id == AMDGPU_UCODE_ID_SDMA7 + || ucode_id == AMDGPU_UCODE_ID_RLC_G + || ucode_id == AMDGPU_UCODE_ID_RLC_RESTORE_LIST_CNTL + || ucode_id == AMDGPU_UCODE_ID_RLC_RESTORE_LIST_GPM_MEM + || ucode_id == AMDGPU_UCODE_ID_RLC_RESTORE_LIST_SRM_MEM + || ucode_id == AMDGPU_UCODE_ID_SMC) + return true; + else + return false; + } +} + void amdgpu_virt_update_sriov_video_codec(struct amdgpu_device *adev, struct amdgpu_video_codec_info *encode, uint32_t encode_array_size, struct amdgpu_video_codec_info *decode, uint32_t decode_array_size) @@ -823,3 +900,148 @@ void amdgpu_virt_update_sriov_video_codec(struct amdgpu_device *adev, } } } + +static bool amdgpu_virt_get_rlcg_reg_access_flag(struct amdgpu_device *adev, + u32 acc_flags, u32 hwip, + bool write, u32 *rlcg_flag) +{ + bool ret = false; + + switch (hwip) { + case GC_HWIP: + if (amdgpu_sriov_reg_indirect_gc(adev)) { + *rlcg_flag = + write ? AMDGPU_RLCG_GC_WRITE : AMDGPU_RLCG_GC_READ; + ret = true; + /* only in new version, AMDGPU_REGS_NO_KIQ and + * AMDGPU_REGS_RLC are enabled simultaneously */ + } else if ((acc_flags & AMDGPU_REGS_RLC) && + !(acc_flags & AMDGPU_REGS_NO_KIQ) && write) { + *rlcg_flag = AMDGPU_RLCG_GC_WRITE_LEGACY; + ret = true; + } + break; + case MMHUB_HWIP: + if (amdgpu_sriov_reg_indirect_mmhub(adev) && + (acc_flags & AMDGPU_REGS_RLC) && write) { + *rlcg_flag = AMDGPU_RLCG_MMHUB_WRITE; + ret = true; + } + break; + default: + break; + } + return ret; +} + +static u32 amdgpu_virt_rlcg_reg_rw(struct amdgpu_device *adev, u32 offset, u32 v, u32 flag) +{ + struct amdgpu_rlcg_reg_access_ctrl *reg_access_ctrl; + uint32_t timeout = 50000; + uint32_t i, tmp; + uint32_t ret = 0; + void *scratch_reg0; + void *scratch_reg1; + void *scratch_reg2; + void *scratch_reg3; + void *spare_int; + + if (!adev->gfx.rlc.rlcg_reg_access_supported) { + dev_err(adev->dev, + "indirect registers access through rlcg is not available\n"); + return 0; + } + + reg_access_ctrl = &adev->gfx.rlc.reg_access_ctrl; + scratch_reg0 = (void __iomem *)adev->rmmio + 4 * reg_access_ctrl->scratch_reg0; + scratch_reg1 = (void __iomem *)adev->rmmio + 4 * reg_access_ctrl->scratch_reg1; + scratch_reg2 = (void __iomem *)adev->rmmio + 4 * reg_access_ctrl->scratch_reg2; + scratch_reg3 = (void __iomem *)adev->rmmio + 4 * reg_access_ctrl->scratch_reg3; + if (reg_access_ctrl->spare_int) + spare_int = (void __iomem *)adev->rmmio + 4 * reg_access_ctrl->spare_int; + + if (offset == reg_access_ctrl->grbm_cntl) { + /* if the target reg offset is grbm_cntl, write to scratch_reg2 */ + writel(v, scratch_reg2); + writel(v, ((void __iomem *)adev->rmmio) + (offset * 4)); + } else if (offset == reg_access_ctrl->grbm_idx) { + /* if the target reg offset is grbm_idx, write to scratch_reg3 */ + writel(v, scratch_reg3); + writel(v, ((void __iomem *)adev->rmmio) + (offset * 4)); + } else { + /* + * SCRATCH_REG0 = read/write value + * SCRATCH_REG1[30:28] = command + * SCRATCH_REG1[19:0] = address in dword + * SCRATCH_REG1[26:24] = Error reporting + */ + writel(v, scratch_reg0); + writel((offset | flag), scratch_reg1); + if (reg_access_ctrl->spare_int) + writel(1, spare_int); + + for (i = 0; i < timeout; i++) { + tmp = readl(scratch_reg1); + if (!(tmp & AMDGPU_RLCG_SCRATCH1_ADDRESS_MASK)) + break; + udelay(10); + } + + if (i >= timeout) { + if (amdgpu_sriov_rlcg_error_report_enabled(adev)) { + if (tmp & AMDGPU_RLCG_VFGATE_DISABLED) { + dev_err(adev->dev, + "vfgate is disabled, rlcg failed to program reg: 0x%05x\n", offset); + } else if (tmp & AMDGPU_RLCG_WRONG_OPERATION_TYPE) { + dev_err(adev->dev, + "wrong operation type, rlcg failed to program reg: 0x%05x\n", offset); + } else if (tmp & AMDGPU_RLCG_REG_NOT_IN_RANGE) { + dev_err(adev->dev, + "register is not in range, rlcg failed to program reg: 0x%05x\n", offset); + } else { + dev_err(adev->dev, + "unknown error type, rlcg failed to program reg: 0x%05x\n", offset); + } + } else { + dev_err(adev->dev, + "timeout: rlcg faled to program reg: 0x%05x\n", offset); + } + } + } + + ret = readl(scratch_reg0); + return ret; +} + +void amdgpu_sriov_wreg(struct amdgpu_device *adev, + u32 offset, u32 value, + u32 acc_flags, u32 hwip) +{ + u32 rlcg_flag; + + if (!amdgpu_sriov_runtime(adev) && + amdgpu_virt_get_rlcg_reg_access_flag(adev, acc_flags, hwip, true, &rlcg_flag)) { + amdgpu_virt_rlcg_reg_rw(adev, offset, value, rlcg_flag); + return; + } + + if (acc_flags & AMDGPU_REGS_NO_KIQ) + WREG32_NO_KIQ(offset, value); + else + WREG32(offset, value); +} + +u32 amdgpu_sriov_rreg(struct amdgpu_device *adev, + u32 offset, u32 acc_flags, u32 hwip) +{ + u32 rlcg_flag; + + if (!amdgpu_sriov_runtime(adev) && + amdgpu_virt_get_rlcg_reg_access_flag(adev, acc_flags, hwip, false, &rlcg_flag)) + return amdgpu_virt_rlcg_reg_rw(adev, offset, 0, rlcg_flag); + + if (acc_flags & AMDGPU_REGS_NO_KIQ) + return RREG32_NO_KIQ(offset); + else + return RREG32(offset); +} diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.h index 9adfb8d63280..49c4347d154c 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.h @@ -31,6 +31,20 @@ #define AMDGPU_SRIOV_CAPS_IS_VF (1 << 2) /* this GPU is a virtual function */ #define AMDGPU_PASSTHROUGH_MODE (1 << 3) /* thw whole GPU is pass through for VM */ #define AMDGPU_SRIOV_CAPS_RUNTIME (1 << 4) /* is out of full access mode */ +#define AMDGPU_VF_MMIO_ACCESS_PROTECT (1 << 5) /* MMIO write access is not allowed in sriov runtime */ + +/* flags for indirect register access path supported by rlcg for sriov */ +#define AMDGPU_RLCG_GC_WRITE_LEGACY (0x8 << 28) +#define AMDGPU_RLCG_GC_WRITE (0x0 << 28) +#define AMDGPU_RLCG_GC_READ (0x1 << 28) +#define AMDGPU_RLCG_MMHUB_WRITE (0x2 << 28) + +/* error code for indirect register access path supported by rlcg for sriov */ +#define AMDGPU_RLCG_VFGATE_DISABLED 0x4000000 +#define AMDGPU_RLCG_WRONG_OPERATION_TYPE 0x2000000 +#define AMDGPU_RLCG_REG_NOT_IN_RANGE 0x1000000 + +#define AMDGPU_RLCG_SCRATCH1_ADDRESS_MASK 0xFFFFF /* all asic after AI use this offset */ #define mmRCC_IOV_FUNC_IDENTIFIER 0xDE5 @@ -240,6 +254,9 @@ struct amdgpu_virt { uint32_t decode_max_frame_pixels; uint32_t encode_max_dimension_pixels; uint32_t encode_max_frame_pixels; + + /* the ucode id to signal the autoload */ + uint32_t autoload_ucode_id; }; struct amdgpu_video_codec_info; @@ -275,13 +292,21 @@ struct amdgpu_video_codec_info; (amdgpu_sriov_vf((adev)) && \ ((adev)->virt.reg_access & (AMDGIM_FEATURE_GC_REG_RLC_EN))) +#define amdgpu_sriov_rlcg_error_report_enabled(adev) \ + (amdgpu_sriov_reg_indirect_mmhub(adev) || amdgpu_sriov_reg_indirect_gc(adev)) + #define amdgpu_passthrough(adev) \ ((adev)->virt.caps & AMDGPU_PASSTHROUGH_MODE) +#define amdgpu_sriov_vf_mmio_access_protection(adev) \ +((adev)->virt.caps & AMDGPU_VF_MMIO_ACCESS_PROTECT) + static inline bool is_virtual_machine(void) { -#ifdef CONFIG_X86 +#if defined(CONFIG_X86) return boot_cpu_has(X86_FEATURE_HYPERVISOR); +#elif defined(CONFIG_ARM64) + return !is_kernel_in_hyp_mode(); #else return false; #endif @@ -293,7 +318,6 @@ static inline bool is_virtual_machine(void) ((!amdgpu_in_reset(adev)) && adev->virt.tdr_debug) #define amdgpu_sriov_is_normal(adev) \ ((!amdgpu_in_reset(adev)) && (!adev->virt.tdr_debug)) - bool amdgpu_virt_mmio_blocked(struct amdgpu_device *adev); void amdgpu_virt_init_setting(struct amdgpu_device *adev); void amdgpu_virt_kiq_reg_write_reg_wait(struct amdgpu_device *adev, @@ -321,4 +345,11 @@ enum amdgpu_sriov_vf_mode amdgpu_virt_get_sriov_vf_mode(struct amdgpu_device *ad void amdgpu_virt_update_sriov_video_codec(struct amdgpu_device *adev, struct amdgpu_video_codec_info *encode, uint32_t encode_array_size, struct amdgpu_video_codec_info *decode, uint32_t decode_array_size); +void amdgpu_sriov_wreg(struct amdgpu_device *adev, + u32 offset, u32 value, + u32 acc_flags, u32 hwip); +u32 amdgpu_sriov_rreg(struct amdgpu_device *adev, + u32 offset, u32 acc_flags, u32 hwip); +bool amdgpu_virt_fw_load_skip_check(struct amdgpu_device *adev, + uint32_t ucode_id); #endif diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vkms.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vkms.c index 2dcc68e04e84..500a1dc4fe02 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vkms.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vkms.c @@ -144,15 +144,16 @@ static void amdgpu_vkms_crtc_atomic_disable(struct drm_crtc *crtc, static void amdgpu_vkms_crtc_atomic_flush(struct drm_crtc *crtc, struct drm_atomic_state *state) { + unsigned long flags; if (crtc->state->event) { - spin_lock(&crtc->dev->event_lock); + spin_lock_irqsave(&crtc->dev->event_lock, flags); if (drm_crtc_vblank_get(crtc) != 0) drm_crtc_send_vblank_event(crtc, crtc->state->event); else drm_crtc_arm_vblank_event(crtc, crtc->state->event); - spin_unlock(&crtc->dev->event_lock); + spin_unlock_irqrestore(&crtc->dev->event_lock, flags); crtc->state->event = NULL; } @@ -281,8 +282,8 @@ static int amdgpu_vkms_plane_atomic_check(struct drm_plane *plane, return PTR_ERR(crtc_state); ret = drm_atomic_helper_check_plane_state(new_plane_state, crtc_state, - DRM_PLANE_HELPER_NO_SCALING, - DRM_PLANE_HELPER_NO_SCALING, + DRM_PLANE_NO_SCALING, + DRM_PLANE_NO_SCALING, false, true); if (ret != 0) return ret; @@ -301,9 +302,6 @@ static int amdgpu_vkms_prepare_fb(struct drm_plane *plane, struct drm_gem_object *obj; struct amdgpu_device *adev; struct amdgpu_bo *rbo; - struct list_head list; - struct ttm_validate_buffer tv; - struct ww_acquire_ctx ticket; uint32_t domain; int r; @@ -315,18 +313,19 @@ static int amdgpu_vkms_prepare_fb(struct drm_plane *plane, obj = new_state->fb->obj[0]; rbo = gem_to_amdgpu_bo(obj); adev = amdgpu_ttm_adev(rbo->tbo.bdev); - INIT_LIST_HEAD(&list); - tv.bo = &rbo->tbo; - tv.num_shared = 1; - list_add(&tv.head, &list); - - r = ttm_eu_reserve_buffers(&ticket, &list, false, NULL); + r = amdgpu_bo_reserve(rbo, true); if (r) { dev_err(adev->dev, "fail to reserve bo (%d)\n", r); return r; } + r = dma_resv_reserve_fences(rbo->tbo.base.resv, 1); + if (r) { + dev_err(adev->dev, "allocating fence slot failed (%d)\n", r); + goto error_unlock; + } + if (plane->type != DRM_PLANE_TYPE_CURSOR) domain = amdgpu_display_supported_domains(adev, rbo->flags); else @@ -336,25 +335,29 @@ static int amdgpu_vkms_prepare_fb(struct drm_plane *plane, if (unlikely(r != 0)) { if (r != -ERESTARTSYS) DRM_ERROR("Failed to pin framebuffer with error %d\n", r); - ttm_eu_backoff_reservation(&ticket, &list); - return r; + goto error_unlock; } r = amdgpu_ttm_alloc_gart(&rbo->tbo); if (unlikely(r != 0)) { - amdgpu_bo_unpin(rbo); - ttm_eu_backoff_reservation(&ticket, &list); DRM_ERROR("%p bind failed\n", rbo); - return r; + goto error_unpin; } - ttm_eu_backoff_reservation(&ticket, &list); + amdgpu_bo_unreserve(rbo); afb->address = amdgpu_bo_gpu_offset(rbo); amdgpu_bo_ref(rbo); return 0; + +error_unpin: + amdgpu_bo_unpin(rbo); + +error_unlock: + amdgpu_bo_unreserve(rbo); + return r; } static void amdgpu_vkms_cleanup_fb(struct drm_plane *plane, @@ -390,7 +393,6 @@ static struct drm_plane *amdgpu_vkms_plane_init(struct drm_device *dev, int index) { struct drm_plane *plane; - uint64_t modifiers[] = {DRM_FORMAT_MOD_LINEAR, DRM_FORMAT_MOD_INVALID}; int ret; plane = kzalloc(sizeof(*plane), GFP_KERNEL); @@ -401,7 +403,7 @@ static struct drm_plane *amdgpu_vkms_plane_init(struct drm_device *dev, &amdgpu_vkms_plane_funcs, amdgpu_vkms_formats, ARRAY_SIZE(amdgpu_vkms_formats), - modifiers, type, NULL); + NULL, type, NULL); if (ret) { kfree(plane); return ERR_PTR(ret); @@ -498,6 +500,8 @@ static int amdgpu_vkms_sw_init(void *handle) adev_to_drm(adev)->mode_config.fb_base = adev->gmc.aper_base; + adev_to_drm(adev)->mode_config.fb_modifiers_not_supported = true; + r = amdgpu_display_modeset_create_props(adev); if (r) return r; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c index b37fc7d7d2c7..003aa9e47085 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c @@ -54,7 +54,7 @@ * (uncached system pages). * Each VM has an ID associated with it and there is a page table * associated with each VMID. When executing a command buffer, - * the kernel tells the the ring what VMID to use for that command + * the kernel tells the ring what VMID to use for that command * buffer. VMIDs are allocated dynamically as commands are submitted. * The userspace drivers maintain their own address space and the kernel * sets up their pages tables accordingly when they submit their @@ -89,6 +89,21 @@ struct amdgpu_prt_cb { }; /** + * struct amdgpu_vm_tlb_seq_cb - Helper to increment the TLB flush sequence + */ +struct amdgpu_vm_tlb_seq_cb { + /** + * @vm: pointer to the amdgpu_vm structure to set the fence sequence on + */ + struct amdgpu_vm *vm; + + /** + * @cb: callback + */ + struct dma_fence_cb cb; +}; + +/** * amdgpu_vm_set_pasid - manage pasid and vm ptr mapping * * @adev: amdgpu_device pointer @@ -128,134 +143,6 @@ int amdgpu_vm_set_pasid(struct amdgpu_device *adev, struct amdgpu_vm *vm, return 0; } -/* - * vm eviction_lock can be taken in MMU notifiers. Make sure no reclaim-FS - * happens while holding this lock anywhere to prevent deadlocks when - * an MMU notifier runs in reclaim-FS context. - */ -static inline void amdgpu_vm_eviction_lock(struct amdgpu_vm *vm) -{ - mutex_lock(&vm->eviction_lock); - vm->saved_flags = memalloc_noreclaim_save(); -} - -static inline int amdgpu_vm_eviction_trylock(struct amdgpu_vm *vm) -{ - if (mutex_trylock(&vm->eviction_lock)) { - vm->saved_flags = memalloc_noreclaim_save(); - return 1; - } - return 0; -} - -static inline void amdgpu_vm_eviction_unlock(struct amdgpu_vm *vm) -{ - memalloc_noreclaim_restore(vm->saved_flags); - mutex_unlock(&vm->eviction_lock); -} - -/** - * amdgpu_vm_level_shift - return the addr shift for each level - * - * @adev: amdgpu_device pointer - * @level: VMPT level - * - * Returns: - * The number of bits the pfn needs to be right shifted for a level. - */ -static unsigned amdgpu_vm_level_shift(struct amdgpu_device *adev, - unsigned level) -{ - switch (level) { - case AMDGPU_VM_PDB2: - case AMDGPU_VM_PDB1: - case AMDGPU_VM_PDB0: - return 9 * (AMDGPU_VM_PDB0 - level) + - adev->vm_manager.block_size; - case AMDGPU_VM_PTB: - return 0; - default: - return ~0; - } -} - -/** - * amdgpu_vm_num_entries - return the number of entries in a PD/PT - * - * @adev: amdgpu_device pointer - * @level: VMPT level - * - * Returns: - * The number of entries in a page directory or page table. - */ -static unsigned amdgpu_vm_num_entries(struct amdgpu_device *adev, - unsigned level) -{ - unsigned shift = amdgpu_vm_level_shift(adev, - adev->vm_manager.root_level); - - if (level == adev->vm_manager.root_level) - /* For the root directory */ - return round_up(adev->vm_manager.max_pfn, 1ULL << shift) - >> shift; - else if (level != AMDGPU_VM_PTB) - /* Everything in between */ - return 512; - else - /* For the page tables on the leaves */ - return AMDGPU_VM_PTE_COUNT(adev); -} - -/** - * amdgpu_vm_num_ats_entries - return the number of ATS entries in the root PD - * - * @adev: amdgpu_device pointer - * - * Returns: - * The number of entries in the root page directory which needs the ATS setting. - */ -static unsigned amdgpu_vm_num_ats_entries(struct amdgpu_device *adev) -{ - unsigned shift; - - shift = amdgpu_vm_level_shift(adev, adev->vm_manager.root_level); - return AMDGPU_GMC_HOLE_START >> (shift + AMDGPU_GPU_PAGE_SHIFT); -} - -/** - * amdgpu_vm_entries_mask - the mask to get the entry number of a PD/PT - * - * @adev: amdgpu_device pointer - * @level: VMPT level - * - * Returns: - * The mask to extract the entry number of a PD/PT from an address. - */ -static uint32_t amdgpu_vm_entries_mask(struct amdgpu_device *adev, - unsigned int level) -{ - if (level <= adev->vm_manager.root_level) - return 0xffffffff; - else if (level != AMDGPU_VM_PTB) - return 0x1ff; - else - return AMDGPU_VM_PTE_COUNT(adev) - 1; -} - -/** - * amdgpu_vm_bo_size - returns the size of the BOs in bytes - * - * @adev: amdgpu_device pointer - * @level: VMPT level - * - * Returns: - * The size of the BO for a page directory or page table in bytes. - */ -static unsigned amdgpu_vm_bo_size(struct amdgpu_device *adev, unsigned level) -{ - return AMDGPU_GPU_PAGE_ALIGN(amdgpu_vm_num_entries(adev, level) * 8); -} - /** * amdgpu_vm_bo_evicted - vm_bo is evicted * @@ -270,10 +157,12 @@ static void amdgpu_vm_bo_evicted(struct amdgpu_vm_bo_base *vm_bo) struct amdgpu_bo *bo = vm_bo->bo; vm_bo->moved = true; + spin_lock(&vm_bo->vm->status_lock); if (bo->tbo.type == ttm_bo_type_kernel) list_move(&vm_bo->vm_status, &vm->evicted); else list_move_tail(&vm_bo->vm_status, &vm->evicted); + spin_unlock(&vm_bo->vm->status_lock); } /** * amdgpu_vm_bo_moved - vm_bo is moved @@ -285,7 +174,9 @@ static void amdgpu_vm_bo_evicted(struct amdgpu_vm_bo_base *vm_bo) */ static void amdgpu_vm_bo_moved(struct amdgpu_vm_bo_base *vm_bo) { + spin_lock(&vm_bo->vm->status_lock); list_move(&vm_bo->vm_status, &vm_bo->vm->moved); + spin_unlock(&vm_bo->vm->status_lock); } /** @@ -298,7 +189,9 @@ static void amdgpu_vm_bo_moved(struct amdgpu_vm_bo_base *vm_bo) */ static void amdgpu_vm_bo_idle(struct amdgpu_vm_bo_base *vm_bo) { + spin_lock(&vm_bo->vm->status_lock); list_move(&vm_bo->vm_status, &vm_bo->vm->idle); + spin_unlock(&vm_bo->vm->status_lock); vm_bo->moved = false; } @@ -312,9 +205,9 @@ static void amdgpu_vm_bo_idle(struct amdgpu_vm_bo_base *vm_bo) */ static void amdgpu_vm_bo_invalidated(struct amdgpu_vm_bo_base *vm_bo) { - spin_lock(&vm_bo->vm->invalidated_lock); + spin_lock(&vm_bo->vm->status_lock); list_move(&vm_bo->vm_status, &vm_bo->vm->invalidated); - spin_unlock(&vm_bo->vm->invalidated_lock); + spin_unlock(&vm_bo->vm->status_lock); } /** @@ -327,10 +220,13 @@ static void amdgpu_vm_bo_invalidated(struct amdgpu_vm_bo_base *vm_bo) */ static void amdgpu_vm_bo_relocated(struct amdgpu_vm_bo_base *vm_bo) { - if (vm_bo->bo->parent) + if (vm_bo->bo->parent) { + spin_lock(&vm_bo->vm->status_lock); list_move(&vm_bo->vm_status, &vm_bo->vm->relocated); - else + spin_unlock(&vm_bo->vm->status_lock); + } else { amdgpu_vm_bo_idle(vm_bo); + } } /** @@ -343,9 +239,9 @@ static void amdgpu_vm_bo_relocated(struct amdgpu_vm_bo_base *vm_bo) */ static void amdgpu_vm_bo_done(struct amdgpu_vm_bo_base *vm_bo) { - spin_lock(&vm_bo->vm->invalidated_lock); + spin_lock(&vm_bo->vm->status_lock); list_move(&vm_bo->vm_status, &vm_bo->vm->done); - spin_unlock(&vm_bo->vm->invalidated_lock); + spin_unlock(&vm_bo->vm->status_lock); } /** @@ -358,9 +254,8 @@ static void amdgpu_vm_bo_done(struct amdgpu_vm_bo_base *vm_bo) * Initialize a bo_va_base structure and add it to the appropriate lists * */ -static void amdgpu_vm_bo_base_init(struct amdgpu_vm_bo_base *base, - struct amdgpu_vm *vm, - struct amdgpu_bo *bo) +void amdgpu_vm_bo_base_init(struct amdgpu_vm_bo_base *base, + struct amdgpu_vm *vm, struct amdgpu_bo *bo) { base->vm = vm; base->bo = bo; @@ -375,7 +270,9 @@ static void amdgpu_vm_bo_base_init(struct amdgpu_vm_bo_base *base, if (bo->tbo.base.resv != vm->root.bo->tbo.base.resv) return; - vm->bulk_moveable = false; + dma_resv_assert_held(vm->root.bo->tbo.base.resv); + + ttm_bo_set_bulk_move(&bo->tbo, &vm->lru_bulk_move); if (bo->tbo.type == ttm_bo_type_kernel && bo->parent) amdgpu_vm_bo_relocated(base); else @@ -394,228 +291,6 @@ static void amdgpu_vm_bo_base_init(struct amdgpu_vm_bo_base *base, } /** - * amdgpu_vm_pt_parent - get the parent page directory - * - * @pt: child page table - * - * Helper to get the parent entry for the child page table. NULL if we are at - * the root page directory. - */ -static struct amdgpu_vm_bo_base *amdgpu_vm_pt_parent(struct amdgpu_vm_bo_base *pt) -{ - struct amdgpu_bo *parent = pt->bo->parent; - - if (!parent) - return NULL; - - return parent->vm_bo; -} - -/* - * amdgpu_vm_pt_cursor - state for for_each_amdgpu_vm_pt - */ -struct amdgpu_vm_pt_cursor { - uint64_t pfn; - struct amdgpu_vm_bo_base *parent; - struct amdgpu_vm_bo_base *entry; - unsigned level; -}; - -/** - * amdgpu_vm_pt_start - start PD/PT walk - * - * @adev: amdgpu_device pointer - * @vm: amdgpu_vm structure - * @start: start address of the walk - * @cursor: state to initialize - * - * Initialize a amdgpu_vm_pt_cursor to start a walk. - */ -static void amdgpu_vm_pt_start(struct amdgpu_device *adev, - struct amdgpu_vm *vm, uint64_t start, - struct amdgpu_vm_pt_cursor *cursor) -{ - cursor->pfn = start; - cursor->parent = NULL; - cursor->entry = &vm->root; - cursor->level = adev->vm_manager.root_level; -} - -/** - * amdgpu_vm_pt_descendant - go to child node - * - * @adev: amdgpu_device pointer - * @cursor: current state - * - * Walk to the child node of the current node. - * Returns: - * True if the walk was possible, false otherwise. - */ -static bool amdgpu_vm_pt_descendant(struct amdgpu_device *adev, - struct amdgpu_vm_pt_cursor *cursor) -{ - unsigned mask, shift, idx; - - if ((cursor->level == AMDGPU_VM_PTB) || !cursor->entry || - !cursor->entry->bo) - return false; - - mask = amdgpu_vm_entries_mask(adev, cursor->level); - shift = amdgpu_vm_level_shift(adev, cursor->level); - - ++cursor->level; - idx = (cursor->pfn >> shift) & mask; - cursor->parent = cursor->entry; - cursor->entry = &to_amdgpu_bo_vm(cursor->entry->bo)->entries[idx]; - return true; -} - -/** - * amdgpu_vm_pt_sibling - go to sibling node - * - * @adev: amdgpu_device pointer - * @cursor: current state - * - * Walk to the sibling node of the current node. - * Returns: - * True if the walk was possible, false otherwise. - */ -static bool amdgpu_vm_pt_sibling(struct amdgpu_device *adev, - struct amdgpu_vm_pt_cursor *cursor) -{ - unsigned shift, num_entries; - - /* Root doesn't have a sibling */ - if (!cursor->parent) - return false; - - /* Go to our parents and see if we got a sibling */ - shift = amdgpu_vm_level_shift(adev, cursor->level - 1); - num_entries = amdgpu_vm_num_entries(adev, cursor->level - 1); - - if (cursor->entry == &to_amdgpu_bo_vm(cursor->parent->bo)->entries[num_entries - 1]) - return false; - - cursor->pfn += 1ULL << shift; - cursor->pfn &= ~((1ULL << shift) - 1); - ++cursor->entry; - return true; -} - -/** - * amdgpu_vm_pt_ancestor - go to parent node - * - * @cursor: current state - * - * Walk to the parent node of the current node. - * Returns: - * True if the walk was possible, false otherwise. - */ -static bool amdgpu_vm_pt_ancestor(struct amdgpu_vm_pt_cursor *cursor) -{ - if (!cursor->parent) - return false; - - --cursor->level; - cursor->entry = cursor->parent; - cursor->parent = amdgpu_vm_pt_parent(cursor->parent); - return true; -} - -/** - * amdgpu_vm_pt_next - get next PD/PT in hieratchy - * - * @adev: amdgpu_device pointer - * @cursor: current state - * - * Walk the PD/PT tree to the next node. - */ -static void amdgpu_vm_pt_next(struct amdgpu_device *adev, - struct amdgpu_vm_pt_cursor *cursor) -{ - /* First try a newborn child */ - if (amdgpu_vm_pt_descendant(adev, cursor)) - return; - - /* If that didn't worked try to find a sibling */ - while (!amdgpu_vm_pt_sibling(adev, cursor)) { - /* No sibling, go to our parents and grandparents */ - if (!amdgpu_vm_pt_ancestor(cursor)) { - cursor->pfn = ~0ll; - return; - } - } -} - -/** - * amdgpu_vm_pt_first_dfs - start a deep first search - * - * @adev: amdgpu_device structure - * @vm: amdgpu_vm structure - * @start: optional cursor to start with - * @cursor: state to initialize - * - * Starts a deep first traversal of the PD/PT tree. - */ -static void amdgpu_vm_pt_first_dfs(struct amdgpu_device *adev, - struct amdgpu_vm *vm, - struct amdgpu_vm_pt_cursor *start, - struct amdgpu_vm_pt_cursor *cursor) -{ - if (start) - *cursor = *start; - else - amdgpu_vm_pt_start(adev, vm, 0, cursor); - while (amdgpu_vm_pt_descendant(adev, cursor)); -} - -/** - * amdgpu_vm_pt_continue_dfs - check if the deep first search should continue - * - * @start: starting point for the search - * @entry: current entry - * - * Returns: - * True when the search should continue, false otherwise. - */ -static bool amdgpu_vm_pt_continue_dfs(struct amdgpu_vm_pt_cursor *start, - struct amdgpu_vm_bo_base *entry) -{ - return entry && (!start || entry != start->entry); -} - -/** - * amdgpu_vm_pt_next_dfs - get the next node for a deep first search - * - * @adev: amdgpu_device structure - * @cursor: current state - * - * Move the cursor to the next node in a deep first search. - */ -static void amdgpu_vm_pt_next_dfs(struct amdgpu_device *adev, - struct amdgpu_vm_pt_cursor *cursor) -{ - if (!cursor->entry) - return; - - if (!cursor->parent) - cursor->entry = NULL; - else if (amdgpu_vm_pt_sibling(adev, cursor)) - while (amdgpu_vm_pt_descendant(adev, cursor)); - else - amdgpu_vm_pt_ancestor(cursor); -} - -/* - * for_each_amdgpu_vm_pt_dfs_safe - safe deep first search of all PDs/PTs - */ -#define for_each_amdgpu_vm_pt_dfs_safe(adev, vm, start, cursor, entry) \ - for (amdgpu_vm_pt_first_dfs((adev), (vm), (start), &(cursor)), \ - (entry) = (cursor).entry, amdgpu_vm_pt_next_dfs((adev), &(cursor));\ - amdgpu_vm_pt_continue_dfs((start), (entry)); \ - (entry) = (cursor).entry, amdgpu_vm_pt_next_dfs((adev), &(cursor))) - -/** * amdgpu_vm_get_pd_bo - add the VM PD to a validation list * * @vm: vm providing the BOs @@ -638,36 +313,6 @@ void amdgpu_vm_get_pd_bo(struct amdgpu_vm *vm, } /** - * amdgpu_vm_del_from_lru_notify - update bulk_moveable flag - * - * @bo: BO which was removed from the LRU - * - * Make sure the bulk_moveable flag is updated when a BO is removed from the - * LRU. - */ -void amdgpu_vm_del_from_lru_notify(struct ttm_buffer_object *bo) -{ - struct amdgpu_bo *abo; - struct amdgpu_vm_bo_base *bo_base; - - if (!amdgpu_bo_is_amdgpu_bo(bo)) - return; - - if (bo->pin_count) - return; - - abo = ttm_to_amdgpu_bo(bo); - if (!abo->parent) - return; - for (bo_base = abo->vm_bo; bo_base; bo_base = bo_base->next) { - struct amdgpu_vm *vm = bo_base->vm; - - if (abo->tbo.base.resv == vm->root.bo->tbo.base.resv) - vm->bulk_moveable = false; - } - -} -/** * amdgpu_vm_move_to_lru_tail - move all BOs to the end of LRU * * @adev: amdgpu device pointer @@ -679,35 +324,9 @@ void amdgpu_vm_del_from_lru_notify(struct ttm_buffer_object *bo) void amdgpu_vm_move_to_lru_tail(struct amdgpu_device *adev, struct amdgpu_vm *vm) { - struct amdgpu_vm_bo_base *bo_base; - - if (vm->bulk_moveable) { - spin_lock(&adev->mman.bdev.lru_lock); - ttm_bo_bulk_move_lru_tail(&vm->lru_bulk_move); - spin_unlock(&adev->mman.bdev.lru_lock); - return; - } - - memset(&vm->lru_bulk_move, 0, sizeof(vm->lru_bulk_move)); - spin_lock(&adev->mman.bdev.lru_lock); - list_for_each_entry(bo_base, &vm->idle, vm_status) { - struct amdgpu_bo *bo = bo_base->bo; - struct amdgpu_bo *shadow = amdgpu_bo_shadowed(bo); - - if (!bo->parent) - continue; - - ttm_bo_move_to_lru_tail(&bo->tbo, bo->tbo.resource, - &vm->lru_bulk_move); - if (shadow) - ttm_bo_move_to_lru_tail(&shadow->tbo, - shadow->tbo.resource, - &vm->lru_bulk_move); - } + ttm_lru_bulk_move_tail(&vm->lru_bulk_move); spin_unlock(&adev->mman.bdev.lru_lock); - - vm->bulk_moveable = true; } /** @@ -727,14 +346,20 @@ int amdgpu_vm_validate_pt_bos(struct amdgpu_device *adev, struct amdgpu_vm *vm, int (*validate)(void *p, struct amdgpu_bo *bo), void *param) { - struct amdgpu_vm_bo_base *bo_base, *tmp; + struct amdgpu_vm_bo_base *bo_base; + struct amdgpu_bo *shadow; + struct amdgpu_bo *bo; int r; - vm->bulk_moveable &= list_empty(&vm->evicted); + spin_lock(&vm->status_lock); + while (!list_empty(&vm->evicted)) { + bo_base = list_first_entry(&vm->evicted, + struct amdgpu_vm_bo_base, + vm_status); + spin_unlock(&vm->status_lock); - list_for_each_entry_safe(bo_base, tmp, &vm->evicted, vm_status) { - struct amdgpu_bo *bo = bo_base->bo; - struct amdgpu_bo *shadow = amdgpu_bo_shadowed(bo); + bo = bo_base->bo; + shadow = amdgpu_bo_shadowed(bo); r = validate(param, bo); if (r) @@ -751,7 +376,9 @@ int amdgpu_vm_validate_pt_bos(struct amdgpu_device *adev, struct amdgpu_vm *vm, vm->update_funcs->map_table(to_amdgpu_bo_vm(bo)); amdgpu_vm_bo_relocated(bo_base); } + spin_lock(&vm->status_lock); } + spin_unlock(&vm->status_lock); amdgpu_vm_eviction_lock(vm); vm->evicting = false; @@ -768,317 +395,22 @@ int amdgpu_vm_validate_pt_bos(struct amdgpu_device *adev, struct amdgpu_vm *vm, * Check if all VM PDs/PTs are ready for updates * * Returns: - * True if eviction list is empty. + * True if VM is not evicting. */ bool amdgpu_vm_ready(struct amdgpu_vm *vm) { - return list_empty(&vm->evicted); -} - -/** - * amdgpu_vm_clear_bo - initially clear the PDs/PTs - * - * @adev: amdgpu_device pointer - * @vm: VM to clear BO from - * @vmbo: BO to clear - * @immediate: use an immediate update - * - * Root PD needs to be reserved when calling this. - * - * Returns: - * 0 on success, errno otherwise. - */ -static int amdgpu_vm_clear_bo(struct amdgpu_device *adev, - struct amdgpu_vm *vm, - struct amdgpu_bo_vm *vmbo, - bool immediate) -{ - struct ttm_operation_ctx ctx = { true, false }; - unsigned level = adev->vm_manager.root_level; - struct amdgpu_vm_update_params params; - struct amdgpu_bo *ancestor = &vmbo->bo; - struct amdgpu_bo *bo = &vmbo->bo; - unsigned entries, ats_entries; - uint64_t addr; - int r, idx; - - /* Figure out our place in the hierarchy */ - if (ancestor->parent) { - ++level; - while (ancestor->parent->parent) { - ++level; - ancestor = ancestor->parent; - } - } - - entries = amdgpu_bo_size(bo) / 8; - if (!vm->pte_support_ats) { - ats_entries = 0; - - } else if (!bo->parent) { - ats_entries = amdgpu_vm_num_ats_entries(adev); - ats_entries = min(ats_entries, entries); - entries -= ats_entries; - - } else { - struct amdgpu_vm_bo_base *pt; - - pt = ancestor->vm_bo; - ats_entries = amdgpu_vm_num_ats_entries(adev); - if ((pt - to_amdgpu_bo_vm(vm->root.bo)->entries) >= ats_entries) { - ats_entries = 0; - } else { - ats_entries = entries; - entries = 0; - } - } - - r = ttm_bo_validate(&bo->tbo, &bo->placement, &ctx); - if (r) - return r; - - if (vmbo->shadow) { - struct amdgpu_bo *shadow = vmbo->shadow; - - r = ttm_bo_validate(&shadow->tbo, &shadow->placement, &ctx); - if (r) - return r; - } - - if (!drm_dev_enter(adev_to_drm(adev), &idx)) - return -ENODEV; - - r = vm->update_funcs->map_table(vmbo); - if (r) - goto exit; - - memset(¶ms, 0, sizeof(params)); - params.adev = adev; - params.vm = vm; - params.immediate = immediate; - - r = vm->update_funcs->prepare(¶ms, NULL, AMDGPU_SYNC_EXPLICIT); - if (r) - goto exit; - - addr = 0; - if (ats_entries) { - uint64_t value = 0, flags; - - flags = AMDGPU_PTE_DEFAULT_ATC; - if (level != AMDGPU_VM_PTB) { - /* Handle leaf PDEs as PTEs */ - flags |= AMDGPU_PDE_PTE; - amdgpu_gmc_get_vm_pde(adev, level, &value, &flags); - } - - r = vm->update_funcs->update(¶ms, vmbo, addr, 0, ats_entries, - value, flags); - if (r) - goto exit; - - addr += ats_entries * 8; - } - - if (entries) { - uint64_t value = 0, flags = 0; - - if (adev->asic_type >= CHIP_VEGA10) { - if (level != AMDGPU_VM_PTB) { - /* Handle leaf PDEs as PTEs */ - flags |= AMDGPU_PDE_PTE; - amdgpu_gmc_get_vm_pde(adev, level, - &value, &flags); - } else { - /* Workaround for fault priority problem on GMC9 */ - flags = AMDGPU_PTE_EXECUTABLE; - } - } - - r = vm->update_funcs->update(¶ms, vmbo, addr, 0, entries, - value, flags); - if (r) - goto exit; - } - - r = vm->update_funcs->commit(¶ms, NULL); -exit: - drm_dev_exit(idx); - return r; -} - -/** - * amdgpu_vm_pt_create - create bo for PD/PT - * - * @adev: amdgpu_device pointer - * @vm: requesting vm - * @level: the page table level - * @immediate: use a immediate update - * @vmbo: pointer to the buffer object pointer - */ -static int amdgpu_vm_pt_create(struct amdgpu_device *adev, - struct amdgpu_vm *vm, - int level, bool immediate, - struct amdgpu_bo_vm **vmbo) -{ - struct amdgpu_bo_param bp; - struct amdgpu_bo *bo; - struct dma_resv *resv; - unsigned int num_entries; - int r; - - memset(&bp, 0, sizeof(bp)); - - bp.size = amdgpu_vm_bo_size(adev, level); - bp.byte_align = AMDGPU_GPU_PAGE_SIZE; - bp.domain = AMDGPU_GEM_DOMAIN_VRAM; - bp.domain = amdgpu_bo_get_preferred_domain(adev, bp.domain); - bp.flags = AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS | - AMDGPU_GEM_CREATE_CPU_GTT_USWC; - - if (level < AMDGPU_VM_PTB) - num_entries = amdgpu_vm_num_entries(adev, level); - else - num_entries = 0; - - bp.bo_ptr_size = struct_size((*vmbo), entries, num_entries); - - if (vm->use_cpu_for_update) - bp.flags |= AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED; - - bp.type = ttm_bo_type_kernel; - bp.no_wait_gpu = immediate; - if (vm->root.bo) - bp.resv = vm->root.bo->tbo.base.resv; - - r = amdgpu_bo_create_vm(adev, &bp, vmbo); - if (r) - return r; - - bo = &(*vmbo)->bo; - if (vm->is_compute_context || (adev->flags & AMD_IS_APU)) { - (*vmbo)->shadow = NULL; - return 0; - } - - if (!bp.resv) - WARN_ON(dma_resv_lock(bo->tbo.base.resv, - NULL)); - resv = bp.resv; - memset(&bp, 0, sizeof(bp)); - bp.size = amdgpu_vm_bo_size(adev, level); - bp.domain = AMDGPU_GEM_DOMAIN_GTT; - bp.flags = AMDGPU_GEM_CREATE_CPU_GTT_USWC; - bp.type = ttm_bo_type_kernel; - bp.resv = bo->tbo.base.resv; - bp.bo_ptr_size = sizeof(struct amdgpu_bo); - - r = amdgpu_bo_create(adev, &bp, &(*vmbo)->shadow); + bool empty; + bool ret; - if (!resv) - dma_resv_unlock(bo->tbo.base.resv); - - if (r) { - amdgpu_bo_unref(&bo); - return r; - } - - (*vmbo)->shadow->parent = amdgpu_bo_ref(bo); - amdgpu_bo_add_to_shadow_list(*vmbo); - - return 0; -} - -/** - * amdgpu_vm_alloc_pts - Allocate a specific page table - * - * @adev: amdgpu_device pointer - * @vm: VM to allocate page tables for - * @cursor: Which page table to allocate - * @immediate: use an immediate update - * - * Make sure a specific page table or directory is allocated. - * - * Returns: - * 1 if page table needed to be allocated, 0 if page table was already - * allocated, negative errno if an error occurred. - */ -static int amdgpu_vm_alloc_pts(struct amdgpu_device *adev, - struct amdgpu_vm *vm, - struct amdgpu_vm_pt_cursor *cursor, - bool immediate) -{ - struct amdgpu_vm_bo_base *entry = cursor->entry; - struct amdgpu_bo *pt_bo; - struct amdgpu_bo_vm *pt; - int r; - - if (entry->bo) - return 0; - - r = amdgpu_vm_pt_create(adev, vm, cursor->level, immediate, &pt); - if (r) - return r; - - /* Keep a reference to the root directory to avoid - * freeing them up in the wrong order. - */ - pt_bo = &pt->bo; - pt_bo->parent = amdgpu_bo_ref(cursor->parent->bo); - amdgpu_vm_bo_base_init(entry, vm, pt_bo); - r = amdgpu_vm_clear_bo(adev, vm, pt, immediate); - if (r) - goto error_free_pt; - - return 0; - -error_free_pt: - amdgpu_bo_unref(&pt->shadow); - amdgpu_bo_unref(&pt_bo); - return r; -} - -/** - * amdgpu_vm_free_table - fre one PD/PT - * - * @entry: PDE to free - */ -static void amdgpu_vm_free_table(struct amdgpu_vm_bo_base *entry) -{ - struct amdgpu_bo *shadow; - - if (!entry->bo) - return; - shadow = amdgpu_bo_shadowed(entry->bo); - entry->bo->vm_bo = NULL; - list_del(&entry->vm_status); - amdgpu_bo_unref(&shadow); - amdgpu_bo_unref(&entry->bo); -} - -/** - * amdgpu_vm_free_pts - free PD/PT levels - * - * @adev: amdgpu device structure - * @vm: amdgpu vm structure - * @start: optional cursor where to start freeing PDs/PTs - * - * Free the page directory or page table level and all sub levels. - */ -static void amdgpu_vm_free_pts(struct amdgpu_device *adev, - struct amdgpu_vm *vm, - struct amdgpu_vm_pt_cursor *start) -{ - struct amdgpu_vm_pt_cursor cursor; - struct amdgpu_vm_bo_base *entry; - - vm->bulk_moveable = false; + amdgpu_vm_eviction_lock(vm); + ret = !vm->evicting; + amdgpu_vm_eviction_unlock(vm); - for_each_amdgpu_vm_pt_dfs_safe(adev, vm, start, cursor, entry) - amdgpu_vm_free_table(entry); + spin_lock(&vm->status_lock); + empty = list_empty(&vm->evicted); + spin_unlock(&vm->status_lock); - if (start) - amdgpu_vm_free_table(start->entry); + return ret && empty; } /** @@ -1247,7 +579,8 @@ int amdgpu_vm_flush(struct amdgpu_ring *ring, struct amdgpu_job *job, } dma_fence_put(fence); - if (ring->funcs->emit_gds_switch && gds_switch_needed) { + if (!ring->is_mes_queue && ring->funcs->emit_gds_switch && + gds_switch_needed) { id->gds_base = job->gds_base; id->gds_size = job->gds_size; id->gws_base = job->gws_base; @@ -1328,53 +661,6 @@ uint64_t amdgpu_vm_map_gart(const dma_addr_t *pages_addr, uint64_t addr) } /** - * amdgpu_vm_update_pde - update a single level in the hierarchy - * - * @params: parameters for the update - * @vm: requested vm - * @entry: entry to update - * - * Makes sure the requested entry in parent is up to date. - */ -static int amdgpu_vm_update_pde(struct amdgpu_vm_update_params *params, - struct amdgpu_vm *vm, - struct amdgpu_vm_bo_base *entry) -{ - struct amdgpu_vm_bo_base *parent = amdgpu_vm_pt_parent(entry); - struct amdgpu_bo *bo = parent->bo, *pbo; - uint64_t pde, pt, flags; - unsigned level; - - for (level = 0, pbo = bo->parent; pbo; ++level) - pbo = pbo->parent; - - level += params->adev->vm_manager.root_level; - amdgpu_gmc_get_pde_for_bo(entry->bo, level, &pt, &flags); - pde = (entry - to_amdgpu_bo_vm(parent->bo)->entries) * 8; - return vm->update_funcs->update(params, to_amdgpu_bo_vm(bo), pde, pt, - 1, 0, flags); -} - -/** - * amdgpu_vm_invalidate_pds - mark all PDs as invalid - * - * @adev: amdgpu_device pointer - * @vm: related vm - * - * Mark all PD level as invalid after an error. - */ -static void amdgpu_vm_invalidate_pds(struct amdgpu_device *adev, - struct amdgpu_vm *vm) -{ - struct amdgpu_vm_pt_cursor cursor; - struct amdgpu_vm_bo_base *entry; - - for_each_amdgpu_vm_pt_dfs_safe(adev, vm, NULL, cursor, entry) - if (entry->bo && !entry->moved) - amdgpu_vm_bo_relocated(entry); -} - -/** * amdgpu_vm_update_pdes - make sure that all directories are valid * * @adev: amdgpu_device pointer @@ -1390,9 +676,16 @@ int amdgpu_vm_update_pdes(struct amdgpu_device *adev, struct amdgpu_vm *vm, bool immediate) { struct amdgpu_vm_update_params params; + struct amdgpu_vm_bo_base *entry; + bool flush_tlb_needed = false; + LIST_HEAD(relocated); int r, idx; - if (list_empty(&vm->relocated)) + spin_lock(&vm->status_lock); + list_splice_init(&vm->relocated, &relocated); + spin_unlock(&vm->status_lock); + + if (list_empty(&relocated)) return 0; if (!drm_dev_enter(adev_to_drm(adev), &idx)) @@ -1405,17 +698,13 @@ int amdgpu_vm_update_pdes(struct amdgpu_device *adev, r = vm->update_funcs->prepare(¶ms, NULL, AMDGPU_SYNC_EXPLICIT); if (r) - goto exit; - - while (!list_empty(&vm->relocated)) { - struct amdgpu_vm_bo_base *entry; + goto error; - entry = list_first_entry(&vm->relocated, - struct amdgpu_vm_bo_base, - vm_status); - amdgpu_vm_bo_idle(entry); + list_for_each_entry(entry, &relocated, vm_status) { + /* vm_flush_needed after updating moved PDEs */ + flush_tlb_needed |= entry->moved; - r = amdgpu_vm_update_pde(¶ms, vm, entry); + r = amdgpu_vm_pde_update(¶ms, entry); if (r) goto error; } @@ -1423,297 +712,70 @@ int amdgpu_vm_update_pdes(struct amdgpu_device *adev, r = vm->update_funcs->commit(¶ms, &vm->last_update); if (r) goto error; - drm_dev_exit(idx); - return 0; -error: - amdgpu_vm_invalidate_pds(adev, vm); -exit: - drm_dev_exit(idx); - return r; -} - -/* - * amdgpu_vm_update_flags - figure out flags for PTE updates - * - * Make sure to set the right flags for the PTEs at the desired level. - */ -static void amdgpu_vm_update_flags(struct amdgpu_vm_update_params *params, - struct amdgpu_bo_vm *pt, unsigned int level, - uint64_t pe, uint64_t addr, - unsigned int count, uint32_t incr, - uint64_t flags) - -{ - if (level != AMDGPU_VM_PTB) { - flags |= AMDGPU_PDE_PTE; - amdgpu_gmc_get_vm_pde(params->adev, level, &addr, &flags); - - } else if (params->adev->asic_type >= CHIP_VEGA10 && - !(flags & AMDGPU_PTE_VALID) && - !(flags & AMDGPU_PTE_PRT)) { - - /* Workaround for fault priority problem on GMC9 */ - flags |= AMDGPU_PTE_EXECUTABLE; - } - - params->vm->update_funcs->update(params, pt, pe, addr, count, incr, - flags); -} + if (flush_tlb_needed) + atomic64_inc(&vm->tlb_seq); -/** - * amdgpu_vm_fragment - get fragment for PTEs - * - * @params: see amdgpu_vm_update_params definition - * @start: first PTE to handle - * @end: last PTE to handle - * @flags: hw mapping flags - * @frag: resulting fragment size - * @frag_end: end of this fragment - * - * Returns the first possible fragment for the start and end address. - */ -static void amdgpu_vm_fragment(struct amdgpu_vm_update_params *params, - uint64_t start, uint64_t end, uint64_t flags, - unsigned int *frag, uint64_t *frag_end) -{ - /** - * The MC L1 TLB supports variable sized pages, based on a fragment - * field in the PTE. When this field is set to a non-zero value, page - * granularity is increased from 4KB to (1 << (12 + frag)). The PTE - * flags are considered valid for all PTEs within the fragment range - * and corresponding mappings are assumed to be physically contiguous. - * - * The L1 TLB can store a single PTE for the whole fragment, - * significantly increasing the space available for translation - * caching. This leads to large improvements in throughput when the - * TLB is under pressure. - * - * The L2 TLB distributes small and large fragments into two - * asymmetric partitions. The large fragment cache is significantly - * larger. Thus, we try to use large fragments wherever possible. - * Userspace can support this by aligning virtual base address and - * allocation size to the fragment size. - * - * Starting with Vega10 the fragment size only controls the L1. The L2 - * is now directly feed with small/huge/giant pages from the walker. - */ - unsigned max_frag; - - if (params->adev->asic_type < CHIP_VEGA10) - max_frag = params->adev->vm_manager.fragment_size; - else - max_frag = 31; - - /* system pages are non continuously */ - if (params->pages_addr) { - *frag = 0; - *frag_end = end; - return; + while (!list_empty(&relocated)) { + entry = list_first_entry(&relocated, struct amdgpu_vm_bo_base, + vm_status); + amdgpu_vm_bo_idle(entry); } - /* This intentionally wraps around if no bit is set */ - *frag = min((unsigned)ffs(start) - 1, (unsigned)fls64(end - start) - 1); - if (*frag >= max_frag) { - *frag = max_frag; - *frag_end = end & ~((1ULL << max_frag) - 1); - } else { - *frag_end = start + (1 << *frag); - } +error: + drm_dev_exit(idx); + return r; } /** - * amdgpu_vm_update_ptes - make sure that page tables are valid - * - * @params: see amdgpu_vm_update_params definition - * @start: start of GPU address range - * @end: end of GPU address range - * @dst: destination address to map to, the next dst inside the function - * @flags: mapping flags + * amdgpu_vm_tlb_seq_cb - make sure to increment tlb sequence + * @fence: unused + * @cb: the callback structure * - * Update the page tables in the range @start - @end. - * - * Returns: - * 0 for success, -EINVAL for failure. + * Increments the tlb sequence to make sure that future CS execute a VM flush. */ -static int amdgpu_vm_update_ptes(struct amdgpu_vm_update_params *params, - uint64_t start, uint64_t end, - uint64_t dst, uint64_t flags) +static void amdgpu_vm_tlb_seq_cb(struct dma_fence *fence, + struct dma_fence_cb *cb) { - struct amdgpu_device *adev = params->adev; - struct amdgpu_vm_pt_cursor cursor; - uint64_t frag_start = start, frag_end; - unsigned int frag; - int r; - - /* figure out the initial fragment */ - amdgpu_vm_fragment(params, frag_start, end, flags, &frag, &frag_end); - - /* walk over the address space and update the PTs */ - amdgpu_vm_pt_start(adev, params->vm, start, &cursor); - while (cursor.pfn < end) { - unsigned shift, parent_shift, mask; - uint64_t incr, entry_end, pe_start; - struct amdgpu_bo *pt; - - if (!params->unlocked) { - /* make sure that the page tables covering the - * address range are actually allocated - */ - r = amdgpu_vm_alloc_pts(params->adev, params->vm, - &cursor, params->immediate); - if (r) - return r; - } - - shift = amdgpu_vm_level_shift(adev, cursor.level); - parent_shift = amdgpu_vm_level_shift(adev, cursor.level - 1); - if (params->unlocked) { - /* Unlocked updates are only allowed on the leaves */ - if (amdgpu_vm_pt_descendant(adev, &cursor)) - continue; - } else if (adev->asic_type < CHIP_VEGA10 && - (flags & AMDGPU_PTE_VALID)) { - /* No huge page support before GMC v9 */ - if (cursor.level != AMDGPU_VM_PTB) { - if (!amdgpu_vm_pt_descendant(adev, &cursor)) - return -ENOENT; - continue; - } - } else if (frag < shift) { - /* We can't use this level when the fragment size is - * smaller than the address shift. Go to the next - * child entry and try again. - */ - if (amdgpu_vm_pt_descendant(adev, &cursor)) - continue; - } else if (frag >= parent_shift) { - /* If the fragment size is even larger than the parent - * shift we should go up one level and check it again. - */ - if (!amdgpu_vm_pt_ancestor(&cursor)) - return -EINVAL; - continue; - } - - pt = cursor.entry->bo; - if (!pt) { - /* We need all PDs and PTs for mapping something, */ - if (flags & AMDGPU_PTE_VALID) - return -ENOENT; - - /* but unmapping something can happen at a higher - * level. - */ - if (!amdgpu_vm_pt_ancestor(&cursor)) - return -EINVAL; - - pt = cursor.entry->bo; - shift = parent_shift; - frag_end = max(frag_end, ALIGN(frag_start + 1, - 1ULL << shift)); - } - - /* Looks good so far, calculate parameters for the update */ - incr = (uint64_t)AMDGPU_GPU_PAGE_SIZE << shift; - mask = amdgpu_vm_entries_mask(adev, cursor.level); - pe_start = ((cursor.pfn >> shift) & mask) * 8; - entry_end = ((uint64_t)mask + 1) << shift; - entry_end += cursor.pfn & ~(entry_end - 1); - entry_end = min(entry_end, end); - - do { - struct amdgpu_vm *vm = params->vm; - uint64_t upd_end = min(entry_end, frag_end); - unsigned nptes = (upd_end - frag_start) >> shift; - uint64_t upd_flags = flags | AMDGPU_PTE_FRAG(frag); - - /* This can happen when we set higher level PDs to - * silent to stop fault floods. - */ - nptes = max(nptes, 1u); - - trace_amdgpu_vm_update_ptes(params, frag_start, upd_end, - nptes, dst, incr, upd_flags, - vm->task_info.pid, - vm->immediate.fence_context); - amdgpu_vm_update_flags(params, to_amdgpu_bo_vm(pt), - cursor.level, pe_start, dst, - nptes, incr, upd_flags); - - pe_start += nptes * 8; - dst += nptes * incr; - - frag_start = upd_end; - if (frag_start >= frag_end) { - /* figure out the next fragment */ - amdgpu_vm_fragment(params, frag_start, end, - flags, &frag, &frag_end); - if (frag < shift) - break; - } - } while (frag_start < entry_end); - - if (amdgpu_vm_pt_descendant(adev, &cursor)) { - /* Free all child entries. - * Update the tables with the flags and addresses and free up subsequent - * tables in the case of huge pages or freed up areas. - * This is the maximum you can free, because all other page tables are not - * completely covered by the range and so potentially still in use. - */ - while (cursor.pfn < frag_start) { - /* Make sure previous mapping is freed */ - if (cursor.entry->bo) { - params->table_freed = true; - amdgpu_vm_free_pts(adev, params->vm, &cursor); - } - amdgpu_vm_pt_next(adev, &cursor); - } - - } else if (frag >= shift) { - /* or just move on to the next on the same level. */ - amdgpu_vm_pt_next(adev, &cursor); - } - } + struct amdgpu_vm_tlb_seq_cb *tlb_cb; - return 0; + tlb_cb = container_of(cb, typeof(*tlb_cb), cb); + atomic64_inc(&tlb_cb->vm->tlb_seq); + kfree(tlb_cb); } /** - * amdgpu_vm_bo_update_mapping - update a mapping in the vm page table + * amdgpu_vm_update_range - update a range in the vm page table * - * @adev: amdgpu_device pointer of the VM - * @bo_adev: amdgpu_device pointer of the mapped BO - * @vm: requested vm + * @adev: amdgpu_device pointer to use for commands + * @vm: the VM to update the range * @immediate: immediate submission in a page fault * @unlocked: unlocked invalidation during MM callback + * @flush_tlb: trigger tlb invalidation after update completed * @resv: fences we need to sync to * @start: start of mapped range * @last: last mapped entry * @flags: flags for the entries * @offset: offset into nodes and pages_addr + * @vram_base: base for vram mappings * @res: ttm_resource to map * @pages_addr: DMA addresses to use for mapping * @fence: optional resulting fence - * @table_freed: return true if page table is freed * * Fill in the page table entries between @start and @last. * * Returns: - * 0 for success, -EINVAL for failure. + * 0 for success, negative erro code for failure. */ -int amdgpu_vm_bo_update_mapping(struct amdgpu_device *adev, - struct amdgpu_device *bo_adev, - struct amdgpu_vm *vm, bool immediate, - bool unlocked, struct dma_resv *resv, - uint64_t start, uint64_t last, - uint64_t flags, uint64_t offset, - struct ttm_resource *res, - dma_addr_t *pages_addr, - struct dma_fence **fence, - bool *table_freed) +int amdgpu_vm_update_range(struct amdgpu_device *adev, struct amdgpu_vm *vm, + bool immediate, bool unlocked, bool flush_tlb, + struct dma_resv *resv, uint64_t start, uint64_t last, + uint64_t flags, uint64_t offset, uint64_t vram_base, + struct ttm_resource *res, dma_addr_t *pages_addr, + struct dma_fence **fence) { struct amdgpu_vm_update_params params; + struct amdgpu_vm_tlb_seq_cb *tlb_cb; struct amdgpu_res_cursor cursor; enum amdgpu_sync_mode sync_mode; int r, idx; @@ -1721,6 +783,23 @@ int amdgpu_vm_bo_update_mapping(struct amdgpu_device *adev, if (!drm_dev_enter(adev_to_drm(adev), &idx)) return -ENODEV; + tlb_cb = kmalloc(sizeof(*tlb_cb), GFP_KERNEL); + if (!tlb_cb) { + r = -ENOMEM; + goto error_unlock; + } + + /* Vega20+XGMI where PTEs get inadvertently cached in L2 texture cache, + * heavy-weight flush TLB unconditionally. + */ + flush_tlb |= adev->gmc.xgmi.num_physical_nodes && + adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 4, 0); + + /* + * On GFX8 and older any 8 PTE block with a valid bit set enters the TLB + */ + flush_tlb |= adev->ip_versions[GC_HWIP][0] < IP_VERSION(9, 0, 0); + memset(¶ms, 0, sizeof(params)); params.adev = adev; params.vm = vm; @@ -1739,7 +818,7 @@ int amdgpu_vm_bo_update_mapping(struct amdgpu_device *adev, amdgpu_vm_eviction_lock(vm); if (vm->evicting) { r = -EBUSY; - goto error_unlock; + goto error_free; } if (!unlocked && !dma_fence_is_signaled(vm->last_unlocked)) { @@ -1752,7 +831,7 @@ int amdgpu_vm_bo_update_mapping(struct amdgpu_device *adev, r = vm->update_funcs->prepare(¶ms, resv, sync_mode); if (r) - goto error_unlock; + goto error_free; amdgpu_res_first(pages_addr ? NULL : res, offset, (last - start + 1) * AMDGPU_GPU_PAGE_SIZE, &cursor); @@ -1792,16 +871,15 @@ int amdgpu_vm_bo_update_mapping(struct amdgpu_device *adev, } } else if (flags & (AMDGPU_PTE_VALID | AMDGPU_PTE_PRT)) { - addr = bo_adev->vm_manager.vram_base_offset + - cursor.start; + addr = vram_base + cursor.start; } else { addr = 0; } tmp = start + num_entries; - r = amdgpu_vm_update_ptes(¶ms, start, tmp, addr, flags); + r = amdgpu_vm_ptes_update(¶ms, start, tmp, addr, flags); if (r) - goto error_unlock; + goto error_free; amdgpu_res_next(&cursor, num_entries * AMDGPU_GPU_PAGE_SIZE); start = tmp; @@ -1809,8 +887,21 @@ int amdgpu_vm_bo_update_mapping(struct amdgpu_device *adev, r = vm->update_funcs->commit(¶ms, fence); - if (table_freed) - *table_freed = *table_freed || params.table_freed; + if (flush_tlb || params.table_freed) { + tlb_cb->vm = vm; + if (fence && *fence && + !dma_fence_add_callback(*fence, &tlb_cb->cb, + amdgpu_vm_tlb_seq_cb)) { + dma_fence_put(vm->last_tlb_flush); + vm->last_tlb_flush = dma_fence_get(*fence); + } else { + amdgpu_vm_tlb_seq_cb(NULL, &tlb_cb->cb); + } + tlb_cb = NULL; + } + +error_free: + kfree(tlb_cb); error_unlock: amdgpu_vm_eviction_unlock(vm); @@ -1823,6 +914,7 @@ void amdgpu_vm_get_memory(struct amdgpu_vm *vm, uint64_t *vram_mem, { struct amdgpu_bo_va *bo_va, *tmp; + spin_lock(&vm->status_lock); list_for_each_entry_safe(bo_va, tmp, &vm->idle, base.vm_status) { if (!bo_va->base.bo) continue; @@ -1847,7 +939,6 @@ void amdgpu_vm_get_memory(struct amdgpu_vm *vm, uint64_t *vram_mem, amdgpu_bo_get_memory(bo_va->base.bo, vram_mem, gtt_mem, cpu_mem); } - spin_lock(&vm->invalidated_lock); list_for_each_entry_safe(bo_va, tmp, &vm->invalidated, base.vm_status) { if (!bo_va->base.bo) continue; @@ -1860,7 +951,7 @@ void amdgpu_vm_get_memory(struct amdgpu_vm *vm, uint64_t *vram_mem, amdgpu_bo_get_memory(bo_va->base.bo, vram_mem, gtt_mem, cpu_mem); } - spin_unlock(&vm->invalidated_lock); + spin_unlock(&vm->status_lock); } /** * amdgpu_vm_bo_update - update all BO mappings in the vm page table @@ -1868,7 +959,6 @@ void amdgpu_vm_get_memory(struct amdgpu_vm *vm, uint64_t *vram_mem, * @adev: amdgpu_device pointer * @bo_va: requested BO and VM object * @clear: if true clear the entries - * @table_freed: return true if page table is freed * * Fill in the page table entries for @bo_va. * @@ -1876,7 +966,7 @@ void amdgpu_vm_get_memory(struct amdgpu_vm *vm, uint64_t *vram_mem, * 0 for success, -EINVAL for failure. */ int amdgpu_vm_bo_update(struct amdgpu_device *adev, struct amdgpu_bo_va *bo_va, - bool clear, bool *table_freed) + bool clear) { struct amdgpu_bo *bo = bo_va->base.bo; struct amdgpu_vm *vm = bo_va->base.vm; @@ -1884,9 +974,10 @@ int amdgpu_vm_bo_update(struct amdgpu_device *adev, struct amdgpu_bo_va *bo_va, dma_addr_t *pages_addr = NULL; struct ttm_resource *mem; struct dma_fence **last_update; + bool flush_tlb = clear; struct dma_resv *resv; + uint64_t vram_base; uint64_t flags; - struct amdgpu_device *bo_adev = adev; int r; if (clear || !bo) { @@ -1911,14 +1002,18 @@ int amdgpu_vm_bo_update(struct amdgpu_device *adev, struct amdgpu_bo_va *bo_va, } if (bo) { + struct amdgpu_device *bo_adev; + flags = amdgpu_ttm_tt_pte_flags(adev, bo->tbo.ttm, mem); if (amdgpu_bo_encrypted(bo)) flags |= AMDGPU_PTE_TMZ; bo_adev = amdgpu_ttm_adev(bo->tbo.bdev); + vram_base = bo_adev->vm_manager.vram_base_offset; } else { flags = 0x0; + vram_base = 0; } if (clear || (bo && bo->tbo.base.resv == @@ -1928,7 +1023,7 @@ int amdgpu_vm_bo_update(struct amdgpu_device *adev, struct amdgpu_bo_va *bo_va, last_update = &bo_va->last_pt_update; if (!clear && bo_va->base.moved) { - bo_va->base.moved = false; + flush_tlb = true; list_splice_init(&bo_va->valids, &bo_va->invalids); } else if (bo_va->cleared != clear) { @@ -1951,11 +1046,11 @@ int amdgpu_vm_bo_update(struct amdgpu_device *adev, struct amdgpu_bo_va *bo_va, trace_amdgpu_vm_bo_update(mapping); - r = amdgpu_vm_bo_update_mapping(adev, bo_adev, vm, false, false, - resv, mapping->start, - mapping->last, update_flags, - mapping->offset, mem, - pages_addr, last_update, table_freed); + r = amdgpu_vm_update_range(adev, vm, false, false, flush_tlb, + resv, mapping->start, mapping->last, + update_flags, mapping->offset, + vram_base, mem, pages_addr, + last_update); if (r) return r; } @@ -1978,6 +1073,7 @@ int amdgpu_vm_bo_update(struct amdgpu_device *adev, struct amdgpu_bo_va *bo_va, list_splice_init(&bo_va->invalids, &bo_va->valids); bo_va->cleared = clear; + bo_va->base.moved = false; if (trace_amdgpu_vm_bo_mapping_enabled()) { list_for_each_entry(mapping, &bo_va->valids, list) @@ -2105,7 +1201,7 @@ static void amdgpu_vm_prt_fini(struct amdgpu_device *adev, struct amdgpu_vm *vm) struct dma_resv_iter cursor; struct dma_fence *fence; - dma_resv_for_each_fence(&cursor, resv, true, fence) { + dma_resv_for_each_fence(&cursor, resv, DMA_RESV_USAGE_BOOKKEEP, fence) { /* Add a callback for each fence in the reservation object */ amdgpu_vm_prt_get(adev); amdgpu_vm_add_prt_cb(adev, fence); @@ -2146,10 +1242,10 @@ int amdgpu_vm_clear_freed(struct amdgpu_device *adev, mapping->start < AMDGPU_GMC_HOLE_START) init_pte_value = AMDGPU_PTE_DEFAULT_ATC; - r = amdgpu_vm_bo_update_mapping(adev, adev, vm, false, false, - resv, mapping->start, - mapping->last, init_pte_value, - 0, NULL, NULL, &f, NULL); + r = amdgpu_vm_update_range(adev, vm, false, false, true, resv, + mapping->start, mapping->last, + init_pte_value, 0, 0, NULL, NULL, + &f); amdgpu_vm_free_mapping(adev, vm, mapping, f); if (r) { dma_fence_put(f); @@ -2184,24 +1280,29 @@ int amdgpu_vm_clear_freed(struct amdgpu_device *adev, int amdgpu_vm_handle_moved(struct amdgpu_device *adev, struct amdgpu_vm *vm) { - struct amdgpu_bo_va *bo_va, *tmp; + struct amdgpu_bo_va *bo_va; struct dma_resv *resv; bool clear; int r; - list_for_each_entry_safe(bo_va, tmp, &vm->moved, base.vm_status) { + spin_lock(&vm->status_lock); + while (!list_empty(&vm->moved)) { + bo_va = list_first_entry(&vm->moved, struct amdgpu_bo_va, + base.vm_status); + spin_unlock(&vm->status_lock); + /* Per VM BOs never need to bo cleared in the page tables */ - r = amdgpu_vm_bo_update(adev, bo_va, false, NULL); + r = amdgpu_vm_bo_update(adev, bo_va, false); if (r) return r; + spin_lock(&vm->status_lock); } - spin_lock(&vm->invalidated_lock); while (!list_empty(&vm->invalidated)) { bo_va = list_first_entry(&vm->invalidated, struct amdgpu_bo_va, base.vm_status); resv = bo_va->base.bo->tbo.base.resv; - spin_unlock(&vm->invalidated_lock); + spin_unlock(&vm->status_lock); /* Try to reserve the BO to avoid clearing its ptes */ if (!amdgpu_vm_debug && dma_resv_trylock(resv)) @@ -2210,15 +1311,15 @@ int amdgpu_vm_handle_moved(struct amdgpu_device *adev, else clear = true; - r = amdgpu_vm_bo_update(adev, bo_va, clear, NULL); + r = amdgpu_vm_bo_update(adev, bo_va, clear); if (r) return r; if (!clear) dma_resv_unlock(resv); - spin_lock(&vm->invalidated_lock); + spin_lock(&vm->status_lock); } - spin_unlock(&vm->invalidated_lock); + spin_unlock(&vm->status_lock); return 0; } @@ -2257,6 +1358,7 @@ struct amdgpu_bo_va *amdgpu_vm_bo_add(struct amdgpu_device *adev, if (!bo) return bo_va; + dma_resv_assert_held(bo->tbo.base.resv); if (amdgpu_dmabuf_is_xgmi_accessible(adev, bo)) { bo_va->is_xgmi = true; /* Power up XGMI if it can be potentially used */ @@ -2292,7 +1394,7 @@ static void amdgpu_vm_bo_insert_map(struct amdgpu_device *adev, if (bo && bo->tbo.base.resv == vm->root.bo->tbo.base.resv && !bo_va->base.moved) { - list_move(&bo_va->base.vm_status, &vm->moved); + amdgpu_vm_bo_moved(&bo_va->base); } trace_amdgpu_vm_bo_map(bo_va, mapping); } @@ -2634,7 +1736,7 @@ void amdgpu_vm_bo_trace_cs(struct amdgpu_vm *vm, struct ww_acquire_ctx *ticket) } /** - * amdgpu_vm_bo_rmv - remove a bo to a specific vm + * amdgpu_vm_bo_del - remove a bo from a specific vm * * @adev: amdgpu_device pointer * @bo_va: requested bo_va @@ -2643,7 +1745,7 @@ void amdgpu_vm_bo_trace_cs(struct amdgpu_vm *vm, struct ww_acquire_ctx *ticket) * * Object have to be reserved! */ -void amdgpu_vm_bo_rmv(struct amdgpu_device *adev, +void amdgpu_vm_bo_del(struct amdgpu_device *adev, struct amdgpu_bo_va *bo_va) { struct amdgpu_bo_va_mapping *mapping, *next; @@ -2651,9 +1753,12 @@ void amdgpu_vm_bo_rmv(struct amdgpu_device *adev, struct amdgpu_vm *vm = bo_va->base.vm; struct amdgpu_vm_bo_base **base; + dma_resv_assert_held(vm->root.bo->tbo.base.resv); + if (bo) { + dma_resv_assert_held(bo->tbo.base.resv); if (bo->tbo.base.resv == vm->root.bo->tbo.base.resv) - vm->bulk_moveable = false; + ttm_bo_set_bulk_move(&bo->tbo, NULL); for (base = &bo_va->base.bo->vm_bo; *base; base = &(*base)->next) { @@ -2665,9 +1770,9 @@ void amdgpu_vm_bo_rmv(struct amdgpu_device *adev, } } - spin_lock(&vm->invalidated_lock); + spin_lock(&vm->status_lock); list_del(&bo_va->base.vm_status); - spin_unlock(&vm->invalidated_lock); + spin_unlock(&vm->status_lock); list_for_each_entry_safe(mapping, next, &bo_va->valids, list) { list_del(&mapping->list); @@ -2707,7 +1812,7 @@ bool amdgpu_vm_evictable(struct amdgpu_bo *bo) return true; /* Don't evict VM page tables while they are busy */ - if (!dma_resv_test_signaled(bo->tbo.base.resv, true)) + if (!dma_resv_test_signaled(bo->tbo.base.resv, DMA_RESV_USAGE_BOOKKEEP)) return false; /* Try to block ongoing updates */ @@ -2887,7 +1992,8 @@ void amdgpu_vm_adjust_size(struct amdgpu_device *adev, uint32_t min_vm_size, */ long amdgpu_vm_wait_idle(struct amdgpu_vm *vm, long timeout) { - timeout = dma_resv_wait_timeout(vm->root.bo->tbo.base.resv, true, + timeout = dma_resv_wait_timeout(vm->root.bo->tbo.base.resv, + DMA_RESV_USAGE_BOOKKEEP, true, timeout); if (timeout <= 0) return timeout; @@ -2920,9 +2026,11 @@ int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm) INIT_LIST_HEAD(&vm->moved); INIT_LIST_HEAD(&vm->idle); INIT_LIST_HEAD(&vm->invalidated); - spin_lock_init(&vm->invalidated_lock); + spin_lock_init(&vm->status_lock); INIT_LIST_HEAD(&vm->freed); INIT_LIST_HEAD(&vm->done); + INIT_LIST_HEAD(&vm->pt_freed); + INIT_WORK(&vm->pt_free_work, amdgpu_vm_pt_free_work); /* create scheduler entities for page table updates */ r = drm_sched_entity_init(&vm->immediate, DRM_SCHED_PRIORITY_NORMAL, @@ -2955,6 +2063,7 @@ int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm) vm->update_funcs = &amdgpu_vm_sdma_funcs; vm->last_update = NULL; vm->last_unlocked = dma_fence_get_stub(); + vm->last_tlb_flush = dma_fence_get_stub(); mutex_init(&vm->eviction_lock); vm->evicting = false; @@ -2968,13 +2077,13 @@ int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm) if (r) goto error_free_root; - r = dma_resv_reserve_shared(root_bo->tbo.base.resv, 1); + r = dma_resv_reserve_fences(root_bo->tbo.base.resv, 1); if (r) goto error_unreserve; amdgpu_vm_bo_base_init(&vm->root, vm, root_bo); - r = amdgpu_vm_clear_bo(adev, vm, root, false); + r = amdgpu_vm_pt_clear(adev, vm, root, false); if (r) goto error_unreserve; @@ -2993,6 +2102,7 @@ error_free_root: vm->root.bo = NULL; error_free_delayed: + dma_fence_put(vm->last_tlb_flush); dma_fence_put(vm->last_unlocked); drm_sched_entity_destroy(&vm->delayed); @@ -3003,34 +2113,6 @@ error_free_immediate: } /** - * amdgpu_vm_check_clean_reserved - check if a VM is clean - * - * @adev: amdgpu_device pointer - * @vm: the VM to check - * - * check all entries of the root PD, if any subsequent PDs are allocated, - * it means there are page table creating and filling, and is no a clean - * VM - * - * Returns: - * 0 if this VM is clean - */ -static int amdgpu_vm_check_clean_reserved(struct amdgpu_device *adev, - struct amdgpu_vm *vm) -{ - enum amdgpu_vm_level root = adev->vm_manager.root_level; - unsigned int entries = amdgpu_vm_num_entries(adev, root); - unsigned int i = 0; - - for (i = 0; i < entries; i++) { - if (to_amdgpu_bo_vm(vm->root.bo)->entries[i].bo) - return -EINVAL; - } - - return 0; -} - -/** * amdgpu_vm_make_compute - Turn a GFX VM into a compute VM * * @adev: amdgpu_device pointer @@ -3059,17 +2141,17 @@ int amdgpu_vm_make_compute(struct amdgpu_device *adev, struct amdgpu_vm *vm) return r; /* Sanity checks */ - r = amdgpu_vm_check_clean_reserved(adev, vm); - if (r) + if (!amdgpu_vm_pt_is_root_clean(adev, vm)) { + r = -EINVAL; goto unreserve_bo; + } /* Check if PD needs to be reinitialized and do it before * changing any other state, in case it fails. */ if (pte_support_ats != vm->pte_support_ats) { vm->pte_support_ats = pte_support_ats; - r = amdgpu_vm_clear_bo(adev, vm, - to_amdgpu_bo_vm(vm->root.bo), + r = amdgpu_vm_pt_clear(adev, vm, to_amdgpu_bo_vm(vm->root.bo), false); if (r) goto unreserve_bo; @@ -3095,6 +2177,14 @@ int amdgpu_vm_make_compute(struct amdgpu_device *adev, struct amdgpu_vm *vm) } else { vm->update_funcs = &amdgpu_vm_sdma_funcs; } + /* + * Make sure root PD gets mapped. As vm_update_mode could be changed + * when turning a GFX VM into a compute VM. + */ + r = vm->update_funcs->map_table(to_amdgpu_bo_vm(vm->root.bo)); + if (r) + goto unreserve_bo; + dma_fence_put(vm->last_update); vm->last_update = NULL; vm->is_compute_context = true; @@ -3137,15 +2227,23 @@ void amdgpu_vm_fini(struct amdgpu_device *adev, struct amdgpu_vm *vm) struct amdgpu_bo_va_mapping *mapping, *tmp; bool prt_fini_needed = !!adev->gmc.gmc_funcs->set_prt; struct amdgpu_bo *root; + unsigned long flags; int i; amdgpu_amdkfd_gpuvm_destroy_cb(adev, vm); + flush_work(&vm->pt_free_work); + root = amdgpu_bo_ref(vm->root.bo); amdgpu_bo_reserve(root, true); amdgpu_vm_set_pasid(adev, vm, 0); dma_fence_wait(vm->last_unlocked, false); dma_fence_put(vm->last_unlocked); + dma_fence_wait(vm->last_tlb_flush, false); + /* Make sure that all fence callbacks have completed */ + spin_lock_irqsave(vm->last_tlb_flush->lock, flags); + spin_unlock_irqrestore(vm->last_tlb_flush->lock, flags); + dma_fence_put(vm->last_tlb_flush); list_for_each_entry_safe(mapping, tmp, &vm->freed, list) { if (mapping->flags & AMDGPU_PTE_PRT && prt_fini_needed) { @@ -3157,7 +2255,7 @@ void amdgpu_vm_fini(struct amdgpu_device *adev, struct amdgpu_vm *vm) amdgpu_vm_free_mapping(adev, vm, mapping, NULL); } - amdgpu_vm_free_pts(adev, vm, NULL); + amdgpu_vm_pt_free_root(adev, vm); amdgpu_bo_unreserve(root); amdgpu_bo_unref(&root); WARN_ON(vm->root.bo); @@ -3214,7 +2312,11 @@ void amdgpu_vm_manager_init(struct amdgpu_device *adev) */ #ifdef CONFIG_X86_64 if (amdgpu_vm_update_mode == -1) { - if (amdgpu_gmc_vram_full_visible(&adev->gmc)) + /* For asic with VF MMIO access protection + * avoid using CPU for VM table updates + */ + if (amdgpu_gmc_vram_full_visible(&adev->gmc) && + !amdgpu_sriov_vf_mmio_access_protection(adev)) adev->vm_manager.vm_update_mode = AMDGPU_VM_USE_CPU_FOR_COMPUTE; else @@ -3397,8 +2499,7 @@ bool amdgpu_vm_handle_fault(struct amdgpu_device *adev, u32 pasid, /* Intentionally setting invalid PTE flag * combination to force a no-retry-fault */ - flags = AMDGPU_PTE_EXECUTABLE | AMDGPU_PDE_PTE | - AMDGPU_PTE_TF; + flags = AMDGPU_PTE_SNOOPED | AMDGPU_PTE_PRT; value = 0; } else if (amdgpu_vm_fault_stop == AMDGPU_VM_FAULT_STOP_NEVER) { /* Redirect the access to the dummy page */ @@ -3411,15 +2512,14 @@ bool amdgpu_vm_handle_fault(struct amdgpu_device *adev, u32 pasid, value = 0; } - r = dma_resv_reserve_shared(root->tbo.base.resv, 1); + r = dma_resv_reserve_fences(root->tbo.base.resv, 1); if (r) { pr_debug("failed %d to reserve fence slot\n", r); goto error_unlock; } - r = amdgpu_vm_bo_update_mapping(adev, adev, vm, true, false, NULL, addr, - addr, flags, value, NULL, NULL, NULL, - NULL); + r = amdgpu_vm_update_range(adev, vm, true, false, false, NULL, addr, + addr, flags, value, 0, NULL, NULL, NULL); if (r) goto error_unlock; @@ -3462,6 +2562,7 @@ void amdgpu_debugfs_vm_bo_info(struct amdgpu_vm *vm, struct seq_file *m) unsigned int total_done_objs = 0; unsigned int id = 0; + spin_lock(&vm->status_lock); seq_puts(m, "\tIdle BOs:\n"); list_for_each_entry_safe(bo_va, tmp, &vm->idle, base.vm_status) { if (!bo_va->base.bo) @@ -3499,7 +2600,6 @@ void amdgpu_debugfs_vm_bo_info(struct amdgpu_vm *vm, struct seq_file *m) id = 0; seq_puts(m, "\tInvalidated BOs:\n"); - spin_lock(&vm->invalidated_lock); list_for_each_entry_safe(bo_va, tmp, &vm->invalidated, base.vm_status) { if (!bo_va->base.bo) continue; @@ -3514,7 +2614,7 @@ void amdgpu_debugfs_vm_bo_info(struct amdgpu_vm *vm, struct seq_file *m) continue; total_done += amdgpu_bo_print_info(id++, bo_va->base.bo, m); } - spin_unlock(&vm->invalidated_lock); + spin_unlock(&vm->status_lock); total_done_objs = id; seq_printf(m, "\tTotal idle size: %12lld\tobjs:\t%d\n", total_idle, diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h index 85fcfb8c5efd..6546e786bf00 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h @@ -254,6 +254,9 @@ struct amdgpu_vm { bool evicting; unsigned int saved_flags; + /* Lock to protect vm_bo add/del/move on all lists of vm */ + spinlock_t status_lock; + /* BOs who needs a validation */ struct list_head evicted; @@ -268,7 +271,6 @@ struct amdgpu_vm { /* regular invalidated BOs, but not yet updated in the PT */ struct list_head invalidated; - spinlock_t invalidated_lock; /* BO mappings freed, but not yet updated in the PT */ struct list_head freed; @@ -276,6 +278,10 @@ struct amdgpu_vm { /* BOs which are invalidated, has been updated in the PTs */ struct list_head done; + /* PT BOs scheduled to free and fill with zero if vm_resv is not hold */ + struct list_head pt_freed; + struct work_struct pt_free_work; + /* contains the page directory */ struct amdgpu_vm_bo_base root; struct dma_fence *last_update; @@ -284,6 +290,10 @@ struct amdgpu_vm { struct drm_sched_entity immediate; struct drm_sched_entity delayed; + /* Last finished delayed update */ + atomic64_t tlb_seq; + struct dma_fence *last_tlb_flush; + /* Last unlocked submission to the scheduler entities */ struct dma_fence *last_unlocked; @@ -317,8 +327,6 @@ struct amdgpu_vm { /* Store positions of group of BOs */ struct ttm_lru_bulk_move lru_bulk_move; - /* mark whether can do the bulk move */ - bool bulk_moveable; /* Flag to indicate if VM is used for compute */ bool is_compute_context; }; @@ -397,18 +405,17 @@ int amdgpu_vm_clear_freed(struct amdgpu_device *adev, struct dma_fence **fence); int amdgpu_vm_handle_moved(struct amdgpu_device *adev, struct amdgpu_vm *vm); -int amdgpu_vm_bo_update_mapping(struct amdgpu_device *adev, - struct amdgpu_device *bo_adev, - struct amdgpu_vm *vm, bool immediate, - bool unlocked, struct dma_resv *resv, - uint64_t start, uint64_t last, - uint64_t flags, uint64_t offset, - struct ttm_resource *res, - dma_addr_t *pages_addr, - struct dma_fence **fence, bool *free_table); +void amdgpu_vm_bo_base_init(struct amdgpu_vm_bo_base *base, + struct amdgpu_vm *vm, struct amdgpu_bo *bo); +int amdgpu_vm_update_range(struct amdgpu_device *adev, struct amdgpu_vm *vm, + bool immediate, bool unlocked, bool flush_tlb, + struct dma_resv *resv, uint64_t start, uint64_t last, + uint64_t flags, uint64_t offset, uint64_t vram_base, + struct ttm_resource *res, dma_addr_t *pages_addr, + struct dma_fence **fence); int amdgpu_vm_bo_update(struct amdgpu_device *adev, struct amdgpu_bo_va *bo_va, - bool clear, bool *table_freed); + bool clear); bool amdgpu_vm_evictable(struct amdgpu_bo *bo); void amdgpu_vm_bo_invalidate(struct amdgpu_device *adev, struct amdgpu_bo *bo, bool evicted); @@ -435,7 +442,7 @@ int amdgpu_vm_bo_clear_mappings(struct amdgpu_device *adev, struct amdgpu_bo_va_mapping *amdgpu_vm_bo_lookup_mapping(struct amdgpu_vm *vm, uint64_t addr); void amdgpu_vm_bo_trace_cs(struct amdgpu_vm *vm, struct ww_acquire_ctx *ticket); -void amdgpu_vm_bo_rmv(struct amdgpu_device *adev, +void amdgpu_vm_bo_del(struct amdgpu_device *adev, struct amdgpu_bo_va *bo_va); void amdgpu_vm_adjust_size(struct amdgpu_device *adev, uint32_t min_vm_size, uint32_t fragment_size_default, unsigned max_level, @@ -454,12 +461,79 @@ void amdgpu_vm_set_task_info(struct amdgpu_vm *vm); void amdgpu_vm_move_to_lru_tail(struct amdgpu_device *adev, struct amdgpu_vm *vm); -void amdgpu_vm_del_from_lru_notify(struct ttm_buffer_object *bo); void amdgpu_vm_get_memory(struct amdgpu_vm *vm, uint64_t *vram_mem, uint64_t *gtt_mem, uint64_t *cpu_mem); +int amdgpu_vm_pt_clear(struct amdgpu_device *adev, struct amdgpu_vm *vm, + struct amdgpu_bo_vm *vmbo, bool immediate); +int amdgpu_vm_pt_create(struct amdgpu_device *adev, struct amdgpu_vm *vm, + int level, bool immediate, struct amdgpu_bo_vm **vmbo); +void amdgpu_vm_pt_free_root(struct amdgpu_device *adev, struct amdgpu_vm *vm); +bool amdgpu_vm_pt_is_root_clean(struct amdgpu_device *adev, + struct amdgpu_vm *vm); + +int amdgpu_vm_pde_update(struct amdgpu_vm_update_params *params, + struct amdgpu_vm_bo_base *entry); +int amdgpu_vm_ptes_update(struct amdgpu_vm_update_params *params, + uint64_t start, uint64_t end, + uint64_t dst, uint64_t flags); +void amdgpu_vm_pt_free_work(struct work_struct *work); + #if defined(CONFIG_DEBUG_FS) void amdgpu_debugfs_vm_bo_info(struct amdgpu_vm *vm, struct seq_file *m); #endif +/** + * amdgpu_vm_tlb_seq - return tlb flush sequence number + * @vm: the amdgpu_vm structure to query + * + * Returns the tlb flush sequence number which indicates that the VM TLBs needs + * to be invalidated whenever the sequence number change. + */ +static inline uint64_t amdgpu_vm_tlb_seq(struct amdgpu_vm *vm) +{ + unsigned long flags; + spinlock_t *lock; + + /* + * Workaround to stop racing between the fence signaling and handling + * the cb. The lock is static after initially setting it up, just make + * sure that the dma_fence structure isn't freed up. + */ + rcu_read_lock(); + lock = vm->last_tlb_flush->lock; + rcu_read_unlock(); + + spin_lock_irqsave(lock, flags); + spin_unlock_irqrestore(lock, flags); + + return atomic64_read(&vm->tlb_seq); +} + +/* + * vm eviction_lock can be taken in MMU notifiers. Make sure no reclaim-FS + * happens while holding this lock anywhere to prevent deadlocks when + * an MMU notifier runs in reclaim-FS context. + */ +static inline void amdgpu_vm_eviction_lock(struct amdgpu_vm *vm) +{ + mutex_lock(&vm->eviction_lock); + vm->saved_flags = memalloc_noreclaim_save(); +} + +static inline bool amdgpu_vm_eviction_trylock(struct amdgpu_vm *vm) +{ + if (mutex_trylock(&vm->eviction_lock)) { + vm->saved_flags = memalloc_noreclaim_save(); + return true; + } + return false; +} + +static inline void amdgpu_vm_eviction_unlock(struct amdgpu_vm *vm) +{ + memalloc_noreclaim_restore(vm->saved_flags); + mutex_unlock(&vm->eviction_lock); +} + #endif diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm_cpu.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm_cpu.c index e3fbf0f10add..31913ae86de6 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm_cpu.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm_cpu.c @@ -74,13 +74,12 @@ static int amdgpu_vm_cpu_update(struct amdgpu_vm_update_params *p, { unsigned int i; uint64_t value; - int r; + long r; - if (vmbo->bo.tbo.moving) { - r = dma_fence_wait(vmbo->bo.tbo.moving, true); - if (r) - return r; - } + r = dma_resv_wait_timeout(vmbo->bo.tbo.base.resv, DMA_RESV_USAGE_KERNEL, + true, MAX_SCHEDULE_TIMEOUT); + if (r < 0) + return r; pe += (unsigned long)amdgpu_bo_kptr(&vmbo->bo); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm_pt.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm_pt.c new file mode 100644 index 000000000000..b5f3bba851db --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm_pt.c @@ -0,0 +1,1021 @@ +// SPDX-License-Identifier: GPL-2.0 OR MIT +/* + * Copyright 2022 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + */ + +#include <drm/drm_drv.h> + +#include "amdgpu.h" +#include "amdgpu_trace.h" +#include "amdgpu_vm.h" + +/* + * amdgpu_vm_pt_cursor - state for for_each_amdgpu_vm_pt + */ +struct amdgpu_vm_pt_cursor { + uint64_t pfn; + struct amdgpu_vm_bo_base *parent; + struct amdgpu_vm_bo_base *entry; + unsigned int level; +}; + +/** + * amdgpu_vm_pt_level_shift - return the addr shift for each level + * + * @adev: amdgpu_device pointer + * @level: VMPT level + * + * Returns: + * The number of bits the pfn needs to be right shifted for a level. + */ +static unsigned int amdgpu_vm_pt_level_shift(struct amdgpu_device *adev, + unsigned int level) +{ + switch (level) { + case AMDGPU_VM_PDB2: + case AMDGPU_VM_PDB1: + case AMDGPU_VM_PDB0: + return 9 * (AMDGPU_VM_PDB0 - level) + + adev->vm_manager.block_size; + case AMDGPU_VM_PTB: + return 0; + default: + return ~0; + } +} + +/** + * amdgpu_vm_pt_num_entries - return the number of entries in a PD/PT + * + * @adev: amdgpu_device pointer + * @level: VMPT level + * + * Returns: + * The number of entries in a page directory or page table. + */ +static unsigned int amdgpu_vm_pt_num_entries(struct amdgpu_device *adev, + unsigned int level) +{ + unsigned int shift; + + shift = amdgpu_vm_pt_level_shift(adev, adev->vm_manager.root_level); + if (level == adev->vm_manager.root_level) + /* For the root directory */ + return round_up(adev->vm_manager.max_pfn, 1ULL << shift) + >> shift; + else if (level != AMDGPU_VM_PTB) + /* Everything in between */ + return 512; + + /* For the page tables on the leaves */ + return AMDGPU_VM_PTE_COUNT(adev); +} + +/** + * amdgpu_vm_pt_num_ats_entries - return the number of ATS entries in the root PD + * + * @adev: amdgpu_device pointer + * + * Returns: + * The number of entries in the root page directory which needs the ATS setting. + */ +static unsigned int amdgpu_vm_pt_num_ats_entries(struct amdgpu_device *adev) +{ + unsigned int shift; + + shift = amdgpu_vm_pt_level_shift(adev, adev->vm_manager.root_level); + return AMDGPU_GMC_HOLE_START >> (shift + AMDGPU_GPU_PAGE_SHIFT); +} + +/** + * amdgpu_vm_pt_entries_mask - the mask to get the entry number of a PD/PT + * + * @adev: amdgpu_device pointer + * @level: VMPT level + * + * Returns: + * The mask to extract the entry number of a PD/PT from an address. + */ +static uint32_t amdgpu_vm_pt_entries_mask(struct amdgpu_device *adev, + unsigned int level) +{ + if (level <= adev->vm_manager.root_level) + return 0xffffffff; + else if (level != AMDGPU_VM_PTB) + return 0x1ff; + else + return AMDGPU_VM_PTE_COUNT(adev) - 1; +} + +/** + * amdgpu_vm_pt_size - returns the size of the page table in bytes + * + * @adev: amdgpu_device pointer + * @level: VMPT level + * + * Returns: + * The size of the BO for a page directory or page table in bytes. + */ +static unsigned int amdgpu_vm_pt_size(struct amdgpu_device *adev, + unsigned int level) +{ + return AMDGPU_GPU_PAGE_ALIGN(amdgpu_vm_pt_num_entries(adev, level) * 8); +} + +/** + * amdgpu_vm_pt_parent - get the parent page directory + * + * @pt: child page table + * + * Helper to get the parent entry for the child page table. NULL if we are at + * the root page directory. + */ +static struct amdgpu_vm_bo_base * +amdgpu_vm_pt_parent(struct amdgpu_vm_bo_base *pt) +{ + struct amdgpu_bo *parent = pt->bo->parent; + + if (!parent) + return NULL; + + return parent->vm_bo; +} + +/** + * amdgpu_vm_pt_start - start PD/PT walk + * + * @adev: amdgpu_device pointer + * @vm: amdgpu_vm structure + * @start: start address of the walk + * @cursor: state to initialize + * + * Initialize a amdgpu_vm_pt_cursor to start a walk. + */ +static void amdgpu_vm_pt_start(struct amdgpu_device *adev, + struct amdgpu_vm *vm, uint64_t start, + struct amdgpu_vm_pt_cursor *cursor) +{ + cursor->pfn = start; + cursor->parent = NULL; + cursor->entry = &vm->root; + cursor->level = adev->vm_manager.root_level; +} + +/** + * amdgpu_vm_pt_descendant - go to child node + * + * @adev: amdgpu_device pointer + * @cursor: current state + * + * Walk to the child node of the current node. + * Returns: + * True if the walk was possible, false otherwise. + */ +static bool amdgpu_vm_pt_descendant(struct amdgpu_device *adev, + struct amdgpu_vm_pt_cursor *cursor) +{ + unsigned int mask, shift, idx; + + if ((cursor->level == AMDGPU_VM_PTB) || !cursor->entry || + !cursor->entry->bo) + return false; + + mask = amdgpu_vm_pt_entries_mask(adev, cursor->level); + shift = amdgpu_vm_pt_level_shift(adev, cursor->level); + + ++cursor->level; + idx = (cursor->pfn >> shift) & mask; + cursor->parent = cursor->entry; + cursor->entry = &to_amdgpu_bo_vm(cursor->entry->bo)->entries[idx]; + return true; +} + +/** + * amdgpu_vm_pt_sibling - go to sibling node + * + * @adev: amdgpu_device pointer + * @cursor: current state + * + * Walk to the sibling node of the current node. + * Returns: + * True if the walk was possible, false otherwise. + */ +static bool amdgpu_vm_pt_sibling(struct amdgpu_device *adev, + struct amdgpu_vm_pt_cursor *cursor) +{ + + unsigned int shift, num_entries; + struct amdgpu_bo_vm *parent; + + /* Root doesn't have a sibling */ + if (!cursor->parent) + return false; + + /* Go to our parents and see if we got a sibling */ + shift = amdgpu_vm_pt_level_shift(adev, cursor->level - 1); + num_entries = amdgpu_vm_pt_num_entries(adev, cursor->level - 1); + parent = to_amdgpu_bo_vm(cursor->parent->bo); + + if (cursor->entry == &parent->entries[num_entries - 1]) + return false; + + cursor->pfn += 1ULL << shift; + cursor->pfn &= ~((1ULL << shift) - 1); + ++cursor->entry; + return true; +} + +/** + * amdgpu_vm_pt_ancestor - go to parent node + * + * @cursor: current state + * + * Walk to the parent node of the current node. + * Returns: + * True if the walk was possible, false otherwise. + */ +static bool amdgpu_vm_pt_ancestor(struct amdgpu_vm_pt_cursor *cursor) +{ + if (!cursor->parent) + return false; + + --cursor->level; + cursor->entry = cursor->parent; + cursor->parent = amdgpu_vm_pt_parent(cursor->parent); + return true; +} + +/** + * amdgpu_vm_pt_next - get next PD/PT in hieratchy + * + * @adev: amdgpu_device pointer + * @cursor: current state + * + * Walk the PD/PT tree to the next node. + */ +static void amdgpu_vm_pt_next(struct amdgpu_device *adev, + struct amdgpu_vm_pt_cursor *cursor) +{ + /* First try a newborn child */ + if (amdgpu_vm_pt_descendant(adev, cursor)) + return; + + /* If that didn't worked try to find a sibling */ + while (!amdgpu_vm_pt_sibling(adev, cursor)) { + /* No sibling, go to our parents and grandparents */ + if (!amdgpu_vm_pt_ancestor(cursor)) { + cursor->pfn = ~0ll; + return; + } + } +} + +/** + * amdgpu_vm_pt_first_dfs - start a deep first search + * + * @adev: amdgpu_device structure + * @vm: amdgpu_vm structure + * @start: optional cursor to start with + * @cursor: state to initialize + * + * Starts a deep first traversal of the PD/PT tree. + */ +static void amdgpu_vm_pt_first_dfs(struct amdgpu_device *adev, + struct amdgpu_vm *vm, + struct amdgpu_vm_pt_cursor *start, + struct amdgpu_vm_pt_cursor *cursor) +{ + if (start) + *cursor = *start; + else + amdgpu_vm_pt_start(adev, vm, 0, cursor); + + while (amdgpu_vm_pt_descendant(adev, cursor)) + ; +} + +/** + * amdgpu_vm_pt_continue_dfs - check if the deep first search should continue + * + * @start: starting point for the search + * @entry: current entry + * + * Returns: + * True when the search should continue, false otherwise. + */ +static bool amdgpu_vm_pt_continue_dfs(struct amdgpu_vm_pt_cursor *start, + struct amdgpu_vm_bo_base *entry) +{ + return entry && (!start || entry != start->entry); +} + +/** + * amdgpu_vm_pt_next_dfs - get the next node for a deep first search + * + * @adev: amdgpu_device structure + * @cursor: current state + * + * Move the cursor to the next node in a deep first search. + */ +static void amdgpu_vm_pt_next_dfs(struct amdgpu_device *adev, + struct amdgpu_vm_pt_cursor *cursor) +{ + if (!cursor->entry) + return; + + if (!cursor->parent) + cursor->entry = NULL; + else if (amdgpu_vm_pt_sibling(adev, cursor)) + while (amdgpu_vm_pt_descendant(adev, cursor)) + ; + else + amdgpu_vm_pt_ancestor(cursor); +} + +/* + * for_each_amdgpu_vm_pt_dfs_safe - safe deep first search of all PDs/PTs + */ +#define for_each_amdgpu_vm_pt_dfs_safe(adev, vm, start, cursor, entry) \ + for (amdgpu_vm_pt_first_dfs((adev), (vm), (start), &(cursor)), \ + (entry) = (cursor).entry, amdgpu_vm_pt_next_dfs((adev), &(cursor));\ + amdgpu_vm_pt_continue_dfs((start), (entry)); \ + (entry) = (cursor).entry, amdgpu_vm_pt_next_dfs((adev), &(cursor))) + +/** + * amdgpu_vm_pt_clear - initially clear the PDs/PTs + * + * @adev: amdgpu_device pointer + * @vm: VM to clear BO from + * @vmbo: BO to clear + * @immediate: use an immediate update + * + * Root PD needs to be reserved when calling this. + * + * Returns: + * 0 on success, errno otherwise. + */ +int amdgpu_vm_pt_clear(struct amdgpu_device *adev, struct amdgpu_vm *vm, + struct amdgpu_bo_vm *vmbo, bool immediate) +{ + unsigned int level = adev->vm_manager.root_level; + struct ttm_operation_ctx ctx = { true, false }; + struct amdgpu_vm_update_params params; + struct amdgpu_bo *ancestor = &vmbo->bo; + unsigned int entries, ats_entries; + struct amdgpu_bo *bo = &vmbo->bo; + uint64_t addr; + int r, idx; + + /* Figure out our place in the hierarchy */ + if (ancestor->parent) { + ++level; + while (ancestor->parent->parent) { + ++level; + ancestor = ancestor->parent; + } + } + + entries = amdgpu_bo_size(bo) / 8; + if (!vm->pte_support_ats) { + ats_entries = 0; + + } else if (!bo->parent) { + ats_entries = amdgpu_vm_pt_num_ats_entries(adev); + ats_entries = min(ats_entries, entries); + entries -= ats_entries; + + } else { + struct amdgpu_vm_bo_base *pt; + + pt = ancestor->vm_bo; + ats_entries = amdgpu_vm_pt_num_ats_entries(adev); + if ((pt - to_amdgpu_bo_vm(vm->root.bo)->entries) >= + ats_entries) { + ats_entries = 0; + } else { + ats_entries = entries; + entries = 0; + } + } + + r = ttm_bo_validate(&bo->tbo, &bo->placement, &ctx); + if (r) + return r; + + if (vmbo->shadow) { + struct amdgpu_bo *shadow = vmbo->shadow; + + r = ttm_bo_validate(&shadow->tbo, &shadow->placement, &ctx); + if (r) + return r; + } + + if (!drm_dev_enter(adev_to_drm(adev), &idx)) + return -ENODEV; + + r = vm->update_funcs->map_table(vmbo); + if (r) + goto exit; + + memset(¶ms, 0, sizeof(params)); + params.adev = adev; + params.vm = vm; + params.immediate = immediate; + + r = vm->update_funcs->prepare(¶ms, NULL, AMDGPU_SYNC_EXPLICIT); + if (r) + goto exit; + + addr = 0; + if (ats_entries) { + uint64_t value = 0, flags; + + flags = AMDGPU_PTE_DEFAULT_ATC; + if (level != AMDGPU_VM_PTB) { + /* Handle leaf PDEs as PTEs */ + flags |= AMDGPU_PDE_PTE; + amdgpu_gmc_get_vm_pde(adev, level, &value, &flags); + } + + r = vm->update_funcs->update(¶ms, vmbo, addr, 0, + ats_entries, value, flags); + if (r) + goto exit; + + addr += ats_entries * 8; + } + + if (entries) { + uint64_t value = 0, flags = 0; + + if (adev->asic_type >= CHIP_VEGA10) { + if (level != AMDGPU_VM_PTB) { + /* Handle leaf PDEs as PTEs */ + flags |= AMDGPU_PDE_PTE; + amdgpu_gmc_get_vm_pde(adev, level, + &value, &flags); + } else { + /* Workaround for fault priority problem on GMC9 */ + flags = AMDGPU_PTE_EXECUTABLE; + } + } + + r = vm->update_funcs->update(¶ms, vmbo, addr, 0, entries, + value, flags); + if (r) + goto exit; + } + + r = vm->update_funcs->commit(¶ms, NULL); +exit: + drm_dev_exit(idx); + return r; +} + +/** + * amdgpu_vm_pt_create - create bo for PD/PT + * + * @adev: amdgpu_device pointer + * @vm: requesting vm + * @level: the page table level + * @immediate: use a immediate update + * @vmbo: pointer to the buffer object pointer + */ +int amdgpu_vm_pt_create(struct amdgpu_device *adev, struct amdgpu_vm *vm, + int level, bool immediate, struct amdgpu_bo_vm **vmbo) +{ + struct amdgpu_bo_param bp; + struct amdgpu_bo *bo; + struct dma_resv *resv; + unsigned int num_entries; + int r; + + memset(&bp, 0, sizeof(bp)); + + bp.size = amdgpu_vm_pt_size(adev, level); + bp.byte_align = AMDGPU_GPU_PAGE_SIZE; + bp.domain = AMDGPU_GEM_DOMAIN_VRAM; + bp.domain = amdgpu_bo_get_preferred_domain(adev, bp.domain); + bp.flags = AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS | + AMDGPU_GEM_CREATE_CPU_GTT_USWC; + + if (level < AMDGPU_VM_PTB) + num_entries = amdgpu_vm_pt_num_entries(adev, level); + else + num_entries = 0; + + bp.bo_ptr_size = struct_size((*vmbo), entries, num_entries); + + if (vm->use_cpu_for_update) + bp.flags |= AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED; + + bp.type = ttm_bo_type_kernel; + bp.no_wait_gpu = immediate; + if (vm->root.bo) + bp.resv = vm->root.bo->tbo.base.resv; + + r = amdgpu_bo_create_vm(adev, &bp, vmbo); + if (r) + return r; + + bo = &(*vmbo)->bo; + if (vm->is_compute_context || (adev->flags & AMD_IS_APU)) { + (*vmbo)->shadow = NULL; + return 0; + } + + if (!bp.resv) + WARN_ON(dma_resv_lock(bo->tbo.base.resv, + NULL)); + resv = bp.resv; + memset(&bp, 0, sizeof(bp)); + bp.size = amdgpu_vm_pt_size(adev, level); + bp.domain = AMDGPU_GEM_DOMAIN_GTT; + bp.flags = AMDGPU_GEM_CREATE_CPU_GTT_USWC; + bp.type = ttm_bo_type_kernel; + bp.resv = bo->tbo.base.resv; + bp.bo_ptr_size = sizeof(struct amdgpu_bo); + + r = amdgpu_bo_create(adev, &bp, &(*vmbo)->shadow); + + if (!resv) + dma_resv_unlock(bo->tbo.base.resv); + + if (r) { + amdgpu_bo_unref(&bo); + return r; + } + + (*vmbo)->shadow->parent = amdgpu_bo_ref(bo); + amdgpu_bo_add_to_shadow_list(*vmbo); + + return 0; +} + +/** + * amdgpu_vm_pt_alloc - Allocate a specific page table + * + * @adev: amdgpu_device pointer + * @vm: VM to allocate page tables for + * @cursor: Which page table to allocate + * @immediate: use an immediate update + * + * Make sure a specific page table or directory is allocated. + * + * Returns: + * 1 if page table needed to be allocated, 0 if page table was already + * allocated, negative errno if an error occurred. + */ +static int amdgpu_vm_pt_alloc(struct amdgpu_device *adev, + struct amdgpu_vm *vm, + struct amdgpu_vm_pt_cursor *cursor, + bool immediate) +{ + struct amdgpu_vm_bo_base *entry = cursor->entry; + struct amdgpu_bo *pt_bo; + struct amdgpu_bo_vm *pt; + int r; + + if (entry->bo) + return 0; + + amdgpu_vm_eviction_unlock(vm); + r = amdgpu_vm_pt_create(adev, vm, cursor->level, immediate, &pt); + amdgpu_vm_eviction_lock(vm); + if (r) + return r; + + /* Keep a reference to the root directory to avoid + * freeing them up in the wrong order. + */ + pt_bo = &pt->bo; + pt_bo->parent = amdgpu_bo_ref(cursor->parent->bo); + amdgpu_vm_bo_base_init(entry, vm, pt_bo); + r = amdgpu_vm_pt_clear(adev, vm, pt, immediate); + if (r) + goto error_free_pt; + + return 0; + +error_free_pt: + amdgpu_bo_unref(&pt->shadow); + amdgpu_bo_unref(&pt_bo); + return r; +} + +/** + * amdgpu_vm_pt_free - free one PD/PT + * + * @entry: PDE to free + */ +static void amdgpu_vm_pt_free(struct amdgpu_vm_bo_base *entry) +{ + struct amdgpu_bo *shadow; + + if (!entry->bo) + return; + shadow = amdgpu_bo_shadowed(entry->bo); + if (shadow) { + ttm_bo_set_bulk_move(&shadow->tbo, NULL); + amdgpu_bo_unref(&shadow); + } + ttm_bo_set_bulk_move(&entry->bo->tbo, NULL); + entry->bo->vm_bo = NULL; + + spin_lock(&entry->vm->status_lock); + list_del(&entry->vm_status); + spin_unlock(&entry->vm->status_lock); + amdgpu_bo_unref(&entry->bo); +} + +void amdgpu_vm_pt_free_work(struct work_struct *work) +{ + struct amdgpu_vm_bo_base *entry, *next; + struct amdgpu_vm *vm; + LIST_HEAD(pt_freed); + + vm = container_of(work, struct amdgpu_vm, pt_free_work); + + spin_lock(&vm->status_lock); + list_splice_init(&vm->pt_freed, &pt_freed); + spin_unlock(&vm->status_lock); + + /* flush_work in amdgpu_vm_fini ensure vm->root.bo is valid. */ + amdgpu_bo_reserve(vm->root.bo, true); + + list_for_each_entry_safe(entry, next, &pt_freed, vm_status) + amdgpu_vm_pt_free(entry); + + amdgpu_bo_unreserve(vm->root.bo); +} + +/** + * amdgpu_vm_pt_free_dfs - free PD/PT levels + * + * @adev: amdgpu device structure + * @vm: amdgpu vm structure + * @start: optional cursor where to start freeing PDs/PTs + * + * Free the page directory or page table level and all sub levels. + */ +static void amdgpu_vm_pt_free_dfs(struct amdgpu_device *adev, + struct amdgpu_vm *vm, + struct amdgpu_vm_pt_cursor *start, + bool unlocked) +{ + struct amdgpu_vm_pt_cursor cursor; + struct amdgpu_vm_bo_base *entry; + + if (unlocked) { + spin_lock(&vm->status_lock); + for_each_amdgpu_vm_pt_dfs_safe(adev, vm, start, cursor, entry) + list_move(&entry->vm_status, &vm->pt_freed); + + if (start) + list_move(&start->entry->vm_status, &vm->pt_freed); + spin_unlock(&vm->status_lock); + schedule_work(&vm->pt_free_work); + return; + } + + for_each_amdgpu_vm_pt_dfs_safe(adev, vm, start, cursor, entry) + amdgpu_vm_pt_free(entry); + + if (start) + amdgpu_vm_pt_free(start->entry); +} + +/** + * amdgpu_vm_pt_free_root - free root PD + * @adev: amdgpu device structure + * @vm: amdgpu vm structure + * + * Free the root page directory and everything below it. + */ +void amdgpu_vm_pt_free_root(struct amdgpu_device *adev, struct amdgpu_vm *vm) +{ + amdgpu_vm_pt_free_dfs(adev, vm, NULL, false); +} + +/** + * amdgpu_vm_pt_is_root_clean - check if a root PD is clean + * + * @adev: amdgpu_device pointer + * @vm: the VM to check + * + * Check all entries of the root PD, if any subsequent PDs are allocated, + * it means there are page table creating and filling, and is no a clean + * VM + * + * Returns: + * 0 if this VM is clean + */ +bool amdgpu_vm_pt_is_root_clean(struct amdgpu_device *adev, + struct amdgpu_vm *vm) +{ + enum amdgpu_vm_level root = adev->vm_manager.root_level; + unsigned int entries = amdgpu_vm_pt_num_entries(adev, root); + unsigned int i = 0; + + for (i = 0; i < entries; i++) { + if (to_amdgpu_bo_vm(vm->root.bo)->entries[i].bo) + return false; + } + return true; +} + +/** + * amdgpu_vm_pde_update - update a single level in the hierarchy + * + * @params: parameters for the update + * @entry: entry to update + * + * Makes sure the requested entry in parent is up to date. + */ +int amdgpu_vm_pde_update(struct amdgpu_vm_update_params *params, + struct amdgpu_vm_bo_base *entry) +{ + struct amdgpu_vm_bo_base *parent = amdgpu_vm_pt_parent(entry); + struct amdgpu_bo *bo = parent->bo, *pbo; + struct amdgpu_vm *vm = params->vm; + uint64_t pde, pt, flags; + unsigned int level; + + for (level = 0, pbo = bo->parent; pbo; ++level) + pbo = pbo->parent; + + level += params->adev->vm_manager.root_level; + amdgpu_gmc_get_pde_for_bo(entry->bo, level, &pt, &flags); + pde = (entry - to_amdgpu_bo_vm(parent->bo)->entries) * 8; + return vm->update_funcs->update(params, to_amdgpu_bo_vm(bo), pde, pt, + 1, 0, flags); +} + +/* + * amdgpu_vm_pte_update_flags - figure out flags for PTE updates + * + * Make sure to set the right flags for the PTEs at the desired level. + */ +static void amdgpu_vm_pte_update_flags(struct amdgpu_vm_update_params *params, + struct amdgpu_bo_vm *pt, + unsigned int level, + uint64_t pe, uint64_t addr, + unsigned int count, uint32_t incr, + uint64_t flags) + +{ + if (level != AMDGPU_VM_PTB) { + flags |= AMDGPU_PDE_PTE; + amdgpu_gmc_get_vm_pde(params->adev, level, &addr, &flags); + + } else if (params->adev->asic_type >= CHIP_VEGA10 && + !(flags & AMDGPU_PTE_VALID) && + !(flags & AMDGPU_PTE_PRT)) { + + /* Workaround for fault priority problem on GMC9 */ + flags |= AMDGPU_PTE_EXECUTABLE; + } + + params->vm->update_funcs->update(params, pt, pe, addr, count, incr, + flags); +} + +/** + * amdgpu_vm_pte_fragment - get fragment for PTEs + * + * @params: see amdgpu_vm_update_params definition + * @start: first PTE to handle + * @end: last PTE to handle + * @flags: hw mapping flags + * @frag: resulting fragment size + * @frag_end: end of this fragment + * + * Returns the first possible fragment for the start and end address. + */ +static void amdgpu_vm_pte_fragment(struct amdgpu_vm_update_params *params, + uint64_t start, uint64_t end, uint64_t flags, + unsigned int *frag, uint64_t *frag_end) +{ + /** + * The MC L1 TLB supports variable sized pages, based on a fragment + * field in the PTE. When this field is set to a non-zero value, page + * granularity is increased from 4KB to (1 << (12 + frag)). The PTE + * flags are considered valid for all PTEs within the fragment range + * and corresponding mappings are assumed to be physically contiguous. + * + * The L1 TLB can store a single PTE for the whole fragment, + * significantly increasing the space available for translation + * caching. This leads to large improvements in throughput when the + * TLB is under pressure. + * + * The L2 TLB distributes small and large fragments into two + * asymmetric partitions. The large fragment cache is significantly + * larger. Thus, we try to use large fragments wherever possible. + * Userspace can support this by aligning virtual base address and + * allocation size to the fragment size. + * + * Starting with Vega10 the fragment size only controls the L1. The L2 + * is now directly feed with small/huge/giant pages from the walker. + */ + unsigned int max_frag; + + if (params->adev->asic_type < CHIP_VEGA10) + max_frag = params->adev->vm_manager.fragment_size; + else + max_frag = 31; + + /* system pages are non continuously */ + if (params->pages_addr) { + *frag = 0; + *frag_end = end; + return; + } + + /* This intentionally wraps around if no bit is set */ + *frag = min_t(unsigned int, ffs(start) - 1, fls64(end - start) - 1); + if (*frag >= max_frag) { + *frag = max_frag; + *frag_end = end & ~((1ULL << max_frag) - 1); + } else { + *frag_end = start + (1 << *frag); + } +} + +/** + * amdgpu_vm_ptes_update - make sure that page tables are valid + * + * @params: see amdgpu_vm_update_params definition + * @start: start of GPU address range + * @end: end of GPU address range + * @dst: destination address to map to, the next dst inside the function + * @flags: mapping flags + * + * Update the page tables in the range @start - @end. + * + * Returns: + * 0 for success, -EINVAL for failure. + */ +int amdgpu_vm_ptes_update(struct amdgpu_vm_update_params *params, + uint64_t start, uint64_t end, + uint64_t dst, uint64_t flags) +{ + struct amdgpu_device *adev = params->adev; + struct amdgpu_vm_pt_cursor cursor; + uint64_t frag_start = start, frag_end; + unsigned int frag; + int r; + + /* figure out the initial fragment */ + amdgpu_vm_pte_fragment(params, frag_start, end, flags, &frag, + &frag_end); + + /* walk over the address space and update the PTs */ + amdgpu_vm_pt_start(adev, params->vm, start, &cursor); + while (cursor.pfn < end) { + unsigned int shift, parent_shift, mask; + uint64_t incr, entry_end, pe_start; + struct amdgpu_bo *pt; + + if (!params->unlocked) { + /* make sure that the page tables covering the + * address range are actually allocated + */ + r = amdgpu_vm_pt_alloc(params->adev, params->vm, + &cursor, params->immediate); + if (r) + return r; + } + + shift = amdgpu_vm_pt_level_shift(adev, cursor.level); + parent_shift = amdgpu_vm_pt_level_shift(adev, cursor.level - 1); + if (params->unlocked) { + /* Unlocked updates are only allowed on the leaves */ + if (amdgpu_vm_pt_descendant(adev, &cursor)) + continue; + } else if (adev->asic_type < CHIP_VEGA10 && + (flags & AMDGPU_PTE_VALID)) { + /* No huge page support before GMC v9 */ + if (cursor.level != AMDGPU_VM_PTB) { + if (!amdgpu_vm_pt_descendant(adev, &cursor)) + return -ENOENT; + continue; + } + } else if (frag < shift) { + /* We can't use this level when the fragment size is + * smaller than the address shift. Go to the next + * child entry and try again. + */ + if (amdgpu_vm_pt_descendant(adev, &cursor)) + continue; + } else if (frag >= parent_shift) { + /* If the fragment size is even larger than the parent + * shift we should go up one level and check it again. + */ + if (!amdgpu_vm_pt_ancestor(&cursor)) + return -EINVAL; + continue; + } + + pt = cursor.entry->bo; + if (!pt) { + /* We need all PDs and PTs for mapping something, */ + if (flags & AMDGPU_PTE_VALID) + return -ENOENT; + + /* but unmapping something can happen at a higher + * level. + */ + if (!amdgpu_vm_pt_ancestor(&cursor)) + return -EINVAL; + + pt = cursor.entry->bo; + shift = parent_shift; + frag_end = max(frag_end, ALIGN(frag_start + 1, + 1ULL << shift)); + } + + /* Looks good so far, calculate parameters for the update */ + incr = (uint64_t)AMDGPU_GPU_PAGE_SIZE << shift; + mask = amdgpu_vm_pt_entries_mask(adev, cursor.level); + pe_start = ((cursor.pfn >> shift) & mask) * 8; + entry_end = ((uint64_t)mask + 1) << shift; + entry_end += cursor.pfn & ~(entry_end - 1); + entry_end = min(entry_end, end); + + do { + struct amdgpu_vm *vm = params->vm; + uint64_t upd_end = min(entry_end, frag_end); + unsigned int nptes = (upd_end - frag_start) >> shift; + uint64_t upd_flags = flags | AMDGPU_PTE_FRAG(frag); + + /* This can happen when we set higher level PDs to + * silent to stop fault floods. + */ + nptes = max(nptes, 1u); + + trace_amdgpu_vm_update_ptes(params, frag_start, upd_end, + min(nptes, 32u), dst, incr, + upd_flags, + vm->task_info.pid, + vm->immediate.fence_context); + amdgpu_vm_pte_update_flags(params, to_amdgpu_bo_vm(pt), + cursor.level, pe_start, dst, + nptes, incr, upd_flags); + + pe_start += nptes * 8; + dst += nptes * incr; + + frag_start = upd_end; + if (frag_start >= frag_end) { + /* figure out the next fragment */ + amdgpu_vm_pte_fragment(params, frag_start, end, + flags, &frag, &frag_end); + if (frag < shift) + break; + } + } while (frag_start < entry_end); + + if (amdgpu_vm_pt_descendant(adev, &cursor)) { + /* Free all child entries. + * Update the tables with the flags and addresses and free up subsequent + * tables in the case of huge pages or freed up areas. + * This is the maximum you can free, because all other page tables are not + * completely covered by the range and so potentially still in use. + */ + while (cursor.pfn < frag_start) { + /* Make sure previous mapping is freed */ + if (cursor.entry->bo) { + params->table_freed = true; + amdgpu_vm_pt_free_dfs(adev, params->vm, + &cursor, + params->unlocked); + } + amdgpu_vm_pt_next(adev, &cursor); + } + + } else if (frag >= shift) { + /* or just move on to the next on the same level. */ + amdgpu_vm_pt_next(adev, &cursor); + } + } + + return 0; +} diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm_sdma.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm_sdma.c index dbb551762805..69e105fa41f6 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm_sdma.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm_sdma.c @@ -109,14 +109,22 @@ static int amdgpu_vm_sdma_commit(struct amdgpu_vm_update_params *p, if (p->unlocked) { struct dma_fence *tmp = dma_fence_get(f); - swap(p->vm->last_unlocked, f); + swap(p->vm->last_unlocked, tmp); dma_fence_put(tmp); } else { - amdgpu_bo_fence(p->vm->root.bo, f, true); + dma_resv_add_fence(p->vm->root.bo->tbo.base.resv, f, + DMA_RESV_USAGE_BOOKKEEP); } - if (fence && !p->immediate) + if (fence && !p->immediate) { + /* + * Most hw generations now have a separate queue for page table + * updates, but when the queue is shared with userspace we need + * the extra CPU round trip to correctly flush the TLB. + */ + set_bit(DRM_SCHED_FENCE_DONT_PIPELINE, &f->flags); swap(*fence, f); + } dma_fence_put(f); return 0; @@ -204,14 +212,22 @@ static int amdgpu_vm_sdma_update(struct amdgpu_vm_update_params *p, struct amdgpu_bo *bo = &vmbo->bo; enum amdgpu_ib_pool_type pool = p->immediate ? AMDGPU_IB_POOL_IMMEDIATE : AMDGPU_IB_POOL_DELAYED; + struct dma_resv_iter cursor; unsigned int i, ndw, nptes; + struct dma_fence *fence; uint64_t *pte; int r; /* Wait for PD/PT moves to be completed */ - r = amdgpu_sync_fence(&p->job->sync, bo->tbo.moving); - if (r) - return r; + dma_resv_iter_begin(&cursor, bo->tbo.base.resv, DMA_RESV_USAGE_KERNEL); + dma_resv_for_each_fence_unlocked(&cursor, fence) { + r = amdgpu_sync_fence(&p->job->sync, fence); + if (r) { + dma_resv_iter_end(&cursor); + return r; + } + } + dma_resv_iter_end(&cursor); do { ndw = p->num_dw_left; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vram_mgr.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vram_mgr.c index 7b2b0980ec41..80dd1343594c 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vram_mgr.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vram_mgr.c @@ -32,8 +32,10 @@ #include "atom.h" struct amdgpu_vram_reservation { - struct list_head node; - struct drm_mm_node mm_node; + u64 start; + u64 size; + struct list_head allocated; + struct list_head blocks; }; static inline struct amdgpu_vram_mgr * @@ -48,6 +50,35 @@ to_amdgpu_device(struct amdgpu_vram_mgr *mgr) return container_of(mgr, struct amdgpu_device, mman.vram_mgr); } +static inline struct drm_buddy_block * +amdgpu_vram_mgr_first_block(struct list_head *list) +{ + return list_first_entry_or_null(list, struct drm_buddy_block, link); +} + +static inline bool amdgpu_is_vram_mgr_blocks_contiguous(struct list_head *head) +{ + struct drm_buddy_block *block; + u64 start, size; + + block = amdgpu_vram_mgr_first_block(head); + if (!block) + return false; + + while (head != block->link.next) { + start = amdgpu_vram_mgr_block_start(block); + size = amdgpu_vram_mgr_block_size(block); + + block = list_entry(block->link.next, struct drm_buddy_block, link); + if (start + size != amdgpu_vram_mgr_block_start(block)) + return false; + } + + return true; +} + + + /** * DOC: mem_info_vram_total * @@ -96,10 +127,9 @@ static ssize_t amdgpu_mem_info_vram_used_show(struct device *dev, { struct drm_device *ddev = dev_get_drvdata(dev); struct amdgpu_device *adev = drm_to_adev(ddev); - struct ttm_resource_manager *man; + struct ttm_resource_manager *man = &adev->mman.vram_mgr.manager; - man = ttm_manager_type(&adev->mman.bdev, TTM_PL_VRAM); - return sysfs_emit(buf, "%llu\n", amdgpu_vram_mgr_usage(man)); + return sysfs_emit(buf, "%llu\n", ttm_resource_manager_usage(man)); } /** @@ -116,10 +146,9 @@ static ssize_t amdgpu_mem_info_vis_vram_used_show(struct device *dev, { struct drm_device *ddev = dev_get_drvdata(dev); struct amdgpu_device *adev = drm_to_adev(ddev); - struct ttm_resource_manager *man; - man = ttm_manager_type(&adev->mman.bdev, TTM_PL_VRAM); - return sysfs_emit(buf, "%llu\n", amdgpu_vram_mgr_vis_usage(man)); + return sysfs_emit(buf, "%llu\n", + amdgpu_vram_mgr_vis_usage(&adev->mman.vram_mgr)); } /** @@ -188,18 +217,18 @@ const struct attribute_group amdgpu_vram_mgr_attr_group = { }; /** - * amdgpu_vram_mgr_vis_size - Calculate visible node size + * amdgpu_vram_mgr_vis_size - Calculate visible block size * * @adev: amdgpu_device pointer - * @node: MM node structure + * @block: DRM BUDDY block structure * - * Calculate how many bytes of the MM node are inside visible VRAM + * Calculate how many bytes of the DRM BUDDY block are inside visible VRAM */ static u64 amdgpu_vram_mgr_vis_size(struct amdgpu_device *adev, - struct drm_mm_node *node) + struct drm_buddy_block *block) { - uint64_t start = node->start << PAGE_SHIFT; - uint64_t end = (node->size + node->start) << PAGE_SHIFT; + u64 start = amdgpu_vram_mgr_block_start(block); + u64 end = start + amdgpu_vram_mgr_block_size(block); if (start >= adev->gmc.visible_vram_size) return 0; @@ -220,9 +249,9 @@ u64 amdgpu_vram_mgr_bo_visible_size(struct amdgpu_bo *bo) { struct amdgpu_device *adev = amdgpu_ttm_adev(bo->tbo.bdev); struct ttm_resource *res = bo->tbo.resource; - unsigned pages = res->num_pages; - struct drm_mm_node *mm; - u64 usage; + struct amdgpu_vram_mgr_resource *vres = to_amdgpu_vram_mgr_resource(res); + struct drm_buddy_block *block; + u64 usage = 0; if (amdgpu_gmc_vram_full_visible(&adev->gmc)) return amdgpu_bo_size(bo); @@ -230,9 +259,8 @@ u64 amdgpu_vram_mgr_bo_visible_size(struct amdgpu_bo *bo) if (res->start >= adev->gmc.visible_vram_size >> PAGE_SHIFT) return 0; - mm = &container_of(res, struct ttm_range_mgr_node, base)->mm_nodes[0]; - for (usage = 0; pages; pages -= mm->size, mm++) - usage += amdgpu_vram_mgr_vis_size(adev, mm); + list_for_each_entry(block, &vres->blocks, link) + usage += amdgpu_vram_mgr_vis_size(adev, block); return usage; } @@ -242,51 +270,61 @@ static void amdgpu_vram_mgr_do_reserve(struct ttm_resource_manager *man) { struct amdgpu_vram_mgr *mgr = to_vram_mgr(man); struct amdgpu_device *adev = to_amdgpu_device(mgr); - struct drm_mm *mm = &mgr->mm; + struct drm_buddy *mm = &mgr->mm; struct amdgpu_vram_reservation *rsv, *temp; + struct drm_buddy_block *block; uint64_t vis_usage; - list_for_each_entry_safe(rsv, temp, &mgr->reservations_pending, node) { - if (drm_mm_reserve_node(mm, &rsv->mm_node)) + list_for_each_entry_safe(rsv, temp, &mgr->reservations_pending, blocks) { + if (drm_buddy_alloc_blocks(mm, rsv->start, rsv->start + rsv->size, + rsv->size, mm->chunk_size, &rsv->allocated, + DRM_BUDDY_RANGE_ALLOCATION)) + continue; + + block = amdgpu_vram_mgr_first_block(&rsv->allocated); + if (!block) continue; dev_dbg(adev->dev, "Reservation 0x%llx - %lld, Succeeded\n", - rsv->mm_node.start, rsv->mm_node.size); + rsv->start, rsv->size); - vis_usage = amdgpu_vram_mgr_vis_size(adev, &rsv->mm_node); + vis_usage = amdgpu_vram_mgr_vis_size(adev, block); atomic64_add(vis_usage, &mgr->vis_usage); - atomic64_add(rsv->mm_node.size << PAGE_SHIFT, &mgr->usage); - list_move(&rsv->node, &mgr->reserved_pages); + spin_lock(&man->bdev->lru_lock); + man->usage += rsv->size; + spin_unlock(&man->bdev->lru_lock); + list_move(&rsv->blocks, &mgr->reserved_pages); } } /** * amdgpu_vram_mgr_reserve_range - Reserve a range from VRAM * - * @man: TTM memory type manager + * @mgr: amdgpu_vram_mgr pointer * @start: start address of the range in VRAM * @size: size of the range * - * Reserve memory from start addess with the specified size in VRAM + * Reserve memory from start address with the specified size in VRAM */ -int amdgpu_vram_mgr_reserve_range(struct ttm_resource_manager *man, +int amdgpu_vram_mgr_reserve_range(struct amdgpu_vram_mgr *mgr, uint64_t start, uint64_t size) { - struct amdgpu_vram_mgr *mgr = to_vram_mgr(man); struct amdgpu_vram_reservation *rsv; rsv = kzalloc(sizeof(*rsv), GFP_KERNEL); if (!rsv) return -ENOMEM; - INIT_LIST_HEAD(&rsv->node); - rsv->mm_node.start = start >> PAGE_SHIFT; - rsv->mm_node.size = size >> PAGE_SHIFT; + INIT_LIST_HEAD(&rsv->allocated); + INIT_LIST_HEAD(&rsv->blocks); - spin_lock(&mgr->lock); - list_add_tail(&mgr->reservations_pending, &rsv->node); - amdgpu_vram_mgr_do_reserve(man); - spin_unlock(&mgr->lock); + rsv->start = start; + rsv->size = size; + + mutex_lock(&mgr->lock); + list_add_tail(&rsv->blocks, &mgr->reservations_pending); + amdgpu_vram_mgr_do_reserve(&mgr->manager); + mutex_unlock(&mgr->lock); return 0; } @@ -294,7 +332,7 @@ int amdgpu_vram_mgr_reserve_range(struct ttm_resource_manager *man, /** * amdgpu_vram_mgr_query_page_status - query the reservation status * - * @man: TTM memory type manager + * @mgr: amdgpu_vram_mgr pointer * @start: start address of a page in VRAM * * Returns: @@ -302,26 +340,25 @@ int amdgpu_vram_mgr_reserve_range(struct ttm_resource_manager *man, * 0: the page has been reserved * -ENOENT: the input page is not a reservation */ -int amdgpu_vram_mgr_query_page_status(struct ttm_resource_manager *man, +int amdgpu_vram_mgr_query_page_status(struct amdgpu_vram_mgr *mgr, uint64_t start) { - struct amdgpu_vram_mgr *mgr = to_vram_mgr(man); struct amdgpu_vram_reservation *rsv; int ret; - spin_lock(&mgr->lock); + mutex_lock(&mgr->lock); - list_for_each_entry(rsv, &mgr->reservations_pending, node) { - if ((rsv->mm_node.start <= start) && - (start < (rsv->mm_node.start + rsv->mm_node.size))) { + list_for_each_entry(rsv, &mgr->reservations_pending, blocks) { + if (rsv->start <= start && + (start < (rsv->start + rsv->size))) { ret = -EBUSY; goto out; } } - list_for_each_entry(rsv, &mgr->reserved_pages, node) { - if ((rsv->mm_node.start <= start) && - (start < (rsv->mm_node.start + rsv->mm_node.size))) { + list_for_each_entry(rsv, &mgr->reserved_pages, blocks) { + if (rsv->start <= start && + (start < (rsv->start + rsv->size))) { ret = 0; goto out; } @@ -329,33 +366,11 @@ int amdgpu_vram_mgr_query_page_status(struct ttm_resource_manager *man, ret = -ENOENT; out: - spin_unlock(&mgr->lock); + mutex_unlock(&mgr->lock); return ret; } /** - * amdgpu_vram_mgr_virt_start - update virtual start address - * - * @mem: ttm_resource to update - * @node: just allocated node - * - * Calculate a virtual BO start address to easily check if everything is CPU - * accessible. - */ -static void amdgpu_vram_mgr_virt_start(struct ttm_resource *mem, - struct drm_mm_node *node) -{ - unsigned long start; - - start = node->start + node->size; - if (start > mem->num_pages) - start -= mem->num_pages; - else - start = 0; - mem->start = max(mem->start, start); -} - -/** * amdgpu_vram_mgr_new - allocate new ranges * * @man: TTM memory type manager @@ -370,116 +385,182 @@ static int amdgpu_vram_mgr_new(struct ttm_resource_manager *man, const struct ttm_place *place, struct ttm_resource **res) { - unsigned long lpfn, num_nodes, pages_per_node, pages_left, pages; + u64 vis_usage = 0, max_bytes, cur_size, min_block_size; struct amdgpu_vram_mgr *mgr = to_vram_mgr(man); struct amdgpu_device *adev = to_amdgpu_device(mgr); - uint64_t vis_usage = 0, mem_bytes, max_bytes; - struct ttm_range_mgr_node *node; - struct drm_mm *mm = &mgr->mm; - enum drm_mm_insert_mode mode; - unsigned i; + struct amdgpu_vram_mgr_resource *vres; + u64 size, remaining_size, lpfn, fpfn; + struct drm_buddy *mm = &mgr->mm; + struct drm_buddy_block *block; + unsigned long pages_per_block; int r; - lpfn = place->lpfn; + lpfn = (u64)place->lpfn << PAGE_SHIFT; if (!lpfn) lpfn = man->size; + fpfn = (u64)place->fpfn << PAGE_SHIFT; + max_bytes = adev->gmc.mc_vram_size; if (tbo->type != ttm_bo_type_kernel) max_bytes -= AMDGPU_VM_RESERVED_VRAM; - /* bail out quickly if there's likely not enough VRAM for this BO */ - mem_bytes = tbo->base.size; - if (atomic64_add_return(mem_bytes, &mgr->usage) > max_bytes) { - r = -ENOSPC; - goto error_sub; - } - if (place->flags & TTM_PL_FLAG_CONTIGUOUS) { - pages_per_node = ~0ul; - num_nodes = 1; + pages_per_block = ~0ul; } else { #ifdef CONFIG_TRANSPARENT_HUGEPAGE - pages_per_node = HPAGE_PMD_NR; + pages_per_block = HPAGE_PMD_NR; #else /* default to 2MB */ - pages_per_node = 2UL << (20UL - PAGE_SHIFT); + pages_per_block = 2UL << (20UL - PAGE_SHIFT); #endif - pages_per_node = max_t(uint32_t, pages_per_node, - tbo->page_alignment); - num_nodes = DIV_ROUND_UP_ULL(PFN_UP(mem_bytes), pages_per_node); + pages_per_block = max_t(uint32_t, pages_per_block, + tbo->page_alignment); } - node = kvmalloc(struct_size(node, mm_nodes, num_nodes), - GFP_KERNEL | __GFP_ZERO); - if (!node) { - r = -ENOMEM; - goto error_sub; + vres = kzalloc(sizeof(*vres), GFP_KERNEL); + if (!vres) + return -ENOMEM; + + ttm_resource_init(tbo, place, &vres->base); + + /* bail out quickly if there's likely not enough VRAM for this BO */ + if (ttm_resource_manager_usage(man) > max_bytes) { + r = -ENOSPC; + goto error_fini; } - ttm_resource_init(tbo, place, &node->base); + INIT_LIST_HEAD(&vres->blocks); - mode = DRM_MM_INSERT_BEST; if (place->flags & TTM_PL_FLAG_TOPDOWN) - mode = DRM_MM_INSERT_HIGH; - - pages_left = node->base.num_pages; - - /* Limit maximum size to 2GB due to SG table limitations */ - pages = min(pages_left, 2UL << (30 - PAGE_SHIFT)); - - i = 0; - spin_lock(&mgr->lock); - while (pages_left) { - uint32_t alignment = tbo->page_alignment; - - if (pages >= pages_per_node) - alignment = pages_per_node; - - r = drm_mm_insert_node_in_range(mm, &node->mm_nodes[i], pages, - alignment, 0, place->fpfn, - lpfn, mode); - if (unlikely(r)) { - if (pages > pages_per_node) { - if (is_power_of_2(pages)) - pages = pages / 2; - else - pages = rounddown_pow_of_two(pages); - continue; + vres->flags |= DRM_BUDDY_TOPDOWN_ALLOCATION; + + if (fpfn || lpfn != mgr->mm.size) + /* Allocate blocks in desired range */ + vres->flags |= DRM_BUDDY_RANGE_ALLOCATION; + + remaining_size = (u64)vres->base.num_pages << PAGE_SHIFT; + + mutex_lock(&mgr->lock); + while (remaining_size) { + if (tbo->page_alignment) + min_block_size = (u64)tbo->page_alignment << PAGE_SHIFT; + else + min_block_size = mgr->default_page_size; + + BUG_ON(min_block_size < mm->chunk_size); + + /* Limit maximum size to 2GiB due to SG table limitations */ + size = min(remaining_size, 2ULL << 30); + + if (size >= (u64)pages_per_block << PAGE_SHIFT) + min_block_size = (u64)pages_per_block << PAGE_SHIFT; + + cur_size = size; + + if (fpfn + size != (u64)place->lpfn << PAGE_SHIFT) { + /* + * Except for actual range allocation, modify the size and + * min_block_size conforming to continuous flag enablement + */ + if (place->flags & TTM_PL_FLAG_CONTIGUOUS) { + size = roundup_pow_of_two(size); + min_block_size = size; + /* + * Modify the size value if size is not + * aligned with min_block_size + */ + } else if (!IS_ALIGNED(size, min_block_size)) { + size = round_up(size, min_block_size); } - goto error_free; } - vis_usage += amdgpu_vram_mgr_vis_size(adev, &node->mm_nodes[i]); - amdgpu_vram_mgr_virt_start(&node->base, &node->mm_nodes[i]); - pages_left -= pages; - ++i; + r = drm_buddy_alloc_blocks(mm, fpfn, + lpfn, + size, + min_block_size, + &vres->blocks, + vres->flags); + if (unlikely(r)) + goto error_free_blocks; + + if (size > remaining_size) + remaining_size = 0; + else + remaining_size -= size; + } + mutex_unlock(&mgr->lock); + + if (cur_size != size) { + struct drm_buddy_block *block; + struct list_head *trim_list; + u64 original_size; + LIST_HEAD(temp); + + trim_list = &vres->blocks; + original_size = (u64)vres->base.num_pages << PAGE_SHIFT; + + /* + * If size value is rounded up to min_block_size, trim the last + * block to the required size + */ + if (!list_is_singular(&vres->blocks)) { + block = list_last_entry(&vres->blocks, typeof(*block), link); + list_move_tail(&block->link, &temp); + trim_list = &temp; + /* + * Compute the original_size value by subtracting the + * last block size with (aligned size - original size) + */ + original_size = amdgpu_vram_mgr_block_size(block) - (size - cur_size); + } + + mutex_lock(&mgr->lock); + drm_buddy_block_trim(mm, + original_size, + trim_list); + mutex_unlock(&mgr->lock); + + if (!list_empty(&temp)) + list_splice_tail(trim_list, &vres->blocks); + } + + vres->base.start = 0; + list_for_each_entry(block, &vres->blocks, link) { + unsigned long start; + + start = amdgpu_vram_mgr_block_start(block) + + amdgpu_vram_mgr_block_size(block); + start >>= PAGE_SHIFT; + + if (start > vres->base.num_pages) + start -= vres->base.num_pages; + else + start = 0; + vres->base.start = max(vres->base.start, start); - if (pages > pages_left) - pages = pages_left; + vis_usage += amdgpu_vram_mgr_vis_size(adev, block); } - spin_unlock(&mgr->lock); - if (i == 1) - node->base.placement |= TTM_PL_FLAG_CONTIGUOUS; + if (amdgpu_is_vram_mgr_blocks_contiguous(&vres->blocks)) + vres->base.placement |= TTM_PL_FLAG_CONTIGUOUS; if (adev->gmc.xgmi.connected_to_cpu) - node->base.bus.caching = ttm_cached; + vres->base.bus.caching = ttm_cached; else - node->base.bus.caching = ttm_write_combined; + vres->base.bus.caching = ttm_write_combined; atomic64_add(vis_usage, &mgr->vis_usage); - *res = &node->base; + *res = &vres->base; return 0; -error_free: - while (i--) - drm_mm_remove_node(&node->mm_nodes[i]); - spin_unlock(&mgr->lock); - kvfree(node); +error_free_blocks: + drm_buddy_free_list(mm, &vres->blocks); + mutex_unlock(&mgr->lock); +error_fini: + ttm_resource_fini(man, &vres->base); + kfree(vres); -error_sub: - atomic64_sub(mem_bytes, &mgr->usage); return r; } @@ -494,28 +575,26 @@ error_sub: static void amdgpu_vram_mgr_del(struct ttm_resource_manager *man, struct ttm_resource *res) { - struct ttm_range_mgr_node *node = to_ttm_range_mgr_node(res); + struct amdgpu_vram_mgr_resource *vres = to_amdgpu_vram_mgr_resource(res); struct amdgpu_vram_mgr *mgr = to_vram_mgr(man); struct amdgpu_device *adev = to_amdgpu_device(mgr); - uint64_t usage = 0, vis_usage = 0; - unsigned i, pages; + struct drm_buddy *mm = &mgr->mm; + struct drm_buddy_block *block; + uint64_t vis_usage = 0; - spin_lock(&mgr->lock); - for (i = 0, pages = res->num_pages; pages; - pages -= node->mm_nodes[i].size, ++i) { - struct drm_mm_node *mm = &node->mm_nodes[i]; + mutex_lock(&mgr->lock); + list_for_each_entry(block, &vres->blocks, link) + vis_usage += amdgpu_vram_mgr_vis_size(adev, block); - drm_mm_remove_node(mm); - usage += mm->size << PAGE_SHIFT; - vis_usage += amdgpu_vram_mgr_vis_size(adev, mm); - } amdgpu_vram_mgr_do_reserve(man); - spin_unlock(&mgr->lock); - atomic64_sub(usage, &mgr->usage); + drm_buddy_free_list(mm, &vres->blocks); + mutex_unlock(&mgr->lock); + atomic64_sub(vis_usage, &mgr->vis_usage); - kvfree(node); + ttm_resource_fini(man, res); + kfree(vres); } /** @@ -547,7 +626,7 @@ int amdgpu_vram_mgr_alloc_sgt(struct amdgpu_device *adev, if (!*sgt) return -ENOMEM; - /* Determine the number of DRM_MM nodes to export */ + /* Determine the number of DRM_BUDDY blocks to export */ amdgpu_res_first(res, offset, length, &cursor); while (cursor.remaining) { num_entries++; @@ -563,10 +642,10 @@ int amdgpu_vram_mgr_alloc_sgt(struct amdgpu_device *adev, sg->length = 0; /* - * Walk down DRM_MM nodes to populate scatterlist nodes - * @note: Use iterator api to get first the DRM_MM node + * Walk down DRM_BUDDY blocks to populate scatterlist nodes + * @note: Use iterator api to get first the DRM_BUDDY block * and the number of bytes from it. Access the following - * DRM_MM node(s) if more buffer needs to exported + * DRM_BUDDY block(s) if more buffer needs to exported */ amdgpu_res_first(res, offset, length, &cursor); for_each_sgtable_sg((*sgt), sg, i) { @@ -630,31 +709,81 @@ void amdgpu_vram_mgr_free_sgt(struct device *dev, } /** - * amdgpu_vram_mgr_usage - how many bytes are used in this domain + * amdgpu_vram_mgr_vis_usage - how many bytes are used in the visible part + * + * @mgr: amdgpu_vram_mgr pointer + * + * Returns how many bytes are used in the visible part of VRAM + */ +uint64_t amdgpu_vram_mgr_vis_usage(struct amdgpu_vram_mgr *mgr) +{ + return atomic64_read(&mgr->vis_usage); +} + +/** + * amdgpu_vram_mgr_intersects - test each drm buddy block for intersection * * @man: TTM memory type manager + * @res: The resource to test + * @place: The place to test against + * @size: Size of the new allocation * - * Returns how many bytes are used in this domain. + * Test each drm buddy block for intersection for eviction decision. */ -uint64_t amdgpu_vram_mgr_usage(struct ttm_resource_manager *man) +static bool amdgpu_vram_mgr_intersects(struct ttm_resource_manager *man, + struct ttm_resource *res, + const struct ttm_place *place, + size_t size) { - struct amdgpu_vram_mgr *mgr = to_vram_mgr(man); + struct amdgpu_vram_mgr_resource *mgr = to_amdgpu_vram_mgr_resource(res); + struct drm_buddy_block *block; + + /* Check each drm buddy block individually */ + list_for_each_entry(block, &mgr->blocks, link) { + unsigned long fpfn = + amdgpu_vram_mgr_block_start(block) >> PAGE_SHIFT; + unsigned long lpfn = fpfn + + (amdgpu_vram_mgr_block_size(block) >> PAGE_SHIFT); + + if (place->fpfn < lpfn && + (!place->lpfn || place->lpfn > fpfn)) + return true; + } - return atomic64_read(&mgr->usage); + return false; } /** - * amdgpu_vram_mgr_vis_usage - how many bytes are used in the visible part + * amdgpu_vram_mgr_compatible - test each drm buddy block for compatibility * * @man: TTM memory type manager + * @res: The resource to test + * @place: The place to test against + * @size: Size of the new allocation * - * Returns how many bytes are used in the visible part of VRAM + * Test each drm buddy block for placement compatibility. */ -uint64_t amdgpu_vram_mgr_vis_usage(struct ttm_resource_manager *man) +static bool amdgpu_vram_mgr_compatible(struct ttm_resource_manager *man, + struct ttm_resource *res, + const struct ttm_place *place, + size_t size) { - struct amdgpu_vram_mgr *mgr = to_vram_mgr(man); + struct amdgpu_vram_mgr_resource *mgr = to_amdgpu_vram_mgr_resource(res); + struct drm_buddy_block *block; + + /* Check each drm buddy block individually */ + list_for_each_entry(block, &mgr->blocks, link) { + unsigned long fpfn = + amdgpu_vram_mgr_block_start(block) >> PAGE_SHIFT; + unsigned long lpfn = fpfn + + (amdgpu_vram_mgr_block_size(block) >> PAGE_SHIFT); + + if (fpfn < place->fpfn || + (place->lpfn && lpfn > place->lpfn)) + return false; + } - return atomic64_read(&mgr->vis_usage); + return true; } /** @@ -669,19 +798,29 @@ static void amdgpu_vram_mgr_debug(struct ttm_resource_manager *man, struct drm_printer *printer) { struct amdgpu_vram_mgr *mgr = to_vram_mgr(man); + struct drm_buddy *mm = &mgr->mm; + struct drm_buddy_block *block; + + drm_printf(printer, " vis usage:%llu\n", + amdgpu_vram_mgr_vis_usage(mgr)); - spin_lock(&mgr->lock); - drm_mm_print(&mgr->mm, printer); - spin_unlock(&mgr->lock); + mutex_lock(&mgr->lock); + drm_printf(printer, "default_page_size: %lluKiB\n", + mgr->default_page_size >> 10); - drm_printf(printer, "man size:%llu pages, ram usage:%lluMB, vis usage:%lluMB\n", - man->size, amdgpu_vram_mgr_usage(man) >> 20, - amdgpu_vram_mgr_vis_usage(man) >> 20); + drm_buddy_print(mm, printer); + + drm_printf(printer, "reserved:\n"); + list_for_each_entry(block, &mgr->reserved_pages, link) + drm_buddy_block_print(mm, block, printer); + mutex_unlock(&mgr->lock); } static const struct ttm_resource_manager_func amdgpu_vram_mgr_func = { .alloc = amdgpu_vram_mgr_new, .free = amdgpu_vram_mgr_del, + .intersects = amdgpu_vram_mgr_intersects, + .compatible = amdgpu_vram_mgr_compatible, .debug = amdgpu_vram_mgr_debug }; @@ -696,15 +835,21 @@ int amdgpu_vram_mgr_init(struct amdgpu_device *adev) { struct amdgpu_vram_mgr *mgr = &adev->mman.vram_mgr; struct ttm_resource_manager *man = &mgr->manager; + int err; - ttm_resource_manager_init(man, adev->gmc.real_vram_size >> PAGE_SHIFT); + ttm_resource_manager_init(man, &adev->mman.bdev, + adev->gmc.real_vram_size); man->func = &amdgpu_vram_mgr_func; - drm_mm_init(&mgr->mm, 0, man->size); - spin_lock_init(&mgr->lock); + err = drm_buddy_init(&mgr->mm, man->size, PAGE_SIZE); + if (err) + return err; + + mutex_init(&mgr->lock); INIT_LIST_HEAD(&mgr->reservations_pending); INIT_LIST_HEAD(&mgr->reserved_pages); + mgr->default_page_size = PAGE_SIZE; ttm_set_driver_manager(&adev->mman.bdev, TTM_PL_VRAM, &mgr->manager); ttm_resource_manager_set_used(man, true); @@ -732,16 +877,16 @@ void amdgpu_vram_mgr_fini(struct amdgpu_device *adev) if (ret) return; - spin_lock(&mgr->lock); - list_for_each_entry_safe(rsv, temp, &mgr->reservations_pending, node) + mutex_lock(&mgr->lock); + list_for_each_entry_safe(rsv, temp, &mgr->reservations_pending, blocks) kfree(rsv); - list_for_each_entry_safe(rsv, temp, &mgr->reserved_pages, node) { - drm_mm_remove_node(&rsv->mm_node); + list_for_each_entry_safe(rsv, temp, &mgr->reserved_pages, blocks) { + drm_buddy_free_list(&mgr->mm, &rsv->blocks); kfree(rsv); } - drm_mm_takedown(&mgr->mm); - spin_unlock(&mgr->lock); + drm_buddy_fini(&mgr->mm); + mutex_unlock(&mgr->lock); ttm_resource_manager_cleanup(man); ttm_set_driver_manager(&adev->mman.bdev, TTM_PL_VRAM, NULL); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vram_mgr.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_vram_mgr.h new file mode 100644 index 000000000000..0e04e42cf809 --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vram_mgr.h @@ -0,0 +1,62 @@ +/* SPDX-License-Identifier: MIT + * Copyright 2021 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#ifndef __AMDGPU_VRAM_MGR_H__ +#define __AMDGPU_VRAM_MGR_H__ + +#include <drm/drm_buddy.h> + +struct amdgpu_vram_mgr { + struct ttm_resource_manager manager; + struct drm_buddy mm; + /* protects access to buffer objects */ + struct mutex lock; + struct list_head reservations_pending; + struct list_head reserved_pages; + atomic64_t vis_usage; + u64 default_page_size; +}; + +struct amdgpu_vram_mgr_resource { + struct ttm_resource base; + struct list_head blocks; + unsigned long flags; +}; + +static inline u64 amdgpu_vram_mgr_block_start(struct drm_buddy_block *block) +{ + return drm_buddy_block_offset(block); +} + +static inline u64 amdgpu_vram_mgr_block_size(struct drm_buddy_block *block) +{ + return (u64)PAGE_SIZE << drm_buddy_block_order(block); +} + +static inline struct amdgpu_vram_mgr_resource * +to_amdgpu_vram_mgr_resource(struct ttm_resource *res) +{ + return container_of(res, struct amdgpu_vram_mgr_resource, base); +} + +#endif diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c index a38c6a747fa4..47159e9a0884 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c @@ -32,7 +32,8 @@ #include "wafl/wafl2_4_0_0_smn.h" #include "wafl/wafl2_4_0_0_sh_mask.h" -#define smnPCS_XGMI23_PCS_ERROR_STATUS 0x11a01210 +#include "amdgpu_reset.h" + #define smnPCS_XGMI3X16_PCS_ERROR_STATUS 0x11a0020c #define smnPCS_GOPX1_PCS_ERROR_STATUS 0x12200210 @@ -67,17 +68,6 @@ static const int wafl_pcs_err_status_reg_arct[] = { smnPCS_GOPX1_0_PCS_GOPX1_PCS_ERROR_STATUS + 0x100000, }; -static const int xgmi23_pcs_err_status_reg_aldebaran[] = { - smnPCS_XGMI23_PCS_ERROR_STATUS, - smnPCS_XGMI23_PCS_ERROR_STATUS + 0x100000, - smnPCS_XGMI23_PCS_ERROR_STATUS + 0x200000, - smnPCS_XGMI23_PCS_ERROR_STATUS + 0x300000, - smnPCS_XGMI23_PCS_ERROR_STATUS + 0x400000, - smnPCS_XGMI23_PCS_ERROR_STATUS + 0x500000, - smnPCS_XGMI23_PCS_ERROR_STATUS + 0x600000, - smnPCS_XGMI23_PCS_ERROR_STATUS + 0x700000 -}; - static const int xgmi3x16_pcs_err_status_reg_aldebaran[] = { smnPCS_XGMI3X16_PCS_ERROR_STATUS, smnPCS_XGMI3X16_PCS_ERROR_STATUS + 0x100000, @@ -208,6 +198,7 @@ static struct attribute *amdgpu_xgmi_hive_attrs[] = { &amdgpu_xgmi_hive_id, NULL }; +ATTRIBUTE_GROUPS(amdgpu_xgmi_hive); static ssize_t amdgpu_xgmi_show_attrs(struct kobject *kobj, struct attribute *attr, char *buf) @@ -226,6 +217,9 @@ static void amdgpu_xgmi_hive_release(struct kobject *kobj) struct amdgpu_hive_info *hive = container_of( kobj, struct amdgpu_hive_info, kobj); + amdgpu_reset_put_reset_domain(hive->reset_domain); + hive->reset_domain = NULL; + mutex_destroy(&hive->hive_lock); kfree(hive); } @@ -237,7 +231,7 @@ static const struct sysfs_ops amdgpu_xgmi_hive_ops = { struct kobj_type amdgpu_xgmi_hive_type = { .release = amdgpu_xgmi_hive_release, .sysfs_ops = &amdgpu_xgmi_hive_ops, - .default_attrs = amdgpu_xgmi_hive_attrs, + .default_groups = amdgpu_xgmi_hive_groups, }; static ssize_t amdgpu_xgmi_show_device_id(struct device *dev, @@ -397,15 +391,44 @@ struct amdgpu_hive_info *amdgpu_get_xgmi_hive(struct amdgpu_device *adev) goto pro_end; } + /** + * Only init hive->reset_domain for none SRIOV configuration. For SRIOV, + * Host driver decide how to reset the GPU either through FLR or chain reset. + * Guest side will get individual notifications from the host for the FLR + * if necessary. + */ + if (!amdgpu_sriov_vf(adev)) { + /** + * Avoid recreating reset domain when hive is reconstructed for the case + * of reset the devices in the XGMI hive during probe for passthrough GPU + * See https://www.spinics.net/lists/amd-gfx/msg58836.html + */ + if (adev->reset_domain->type != XGMI_HIVE) { + hive->reset_domain = + amdgpu_reset_create_reset_domain(XGMI_HIVE, "amdgpu-reset-hive"); + if (!hive->reset_domain) { + dev_err(adev->dev, "XGMI: failed initializing reset domain for xgmi hive\n"); + ret = -ENOMEM; + kobject_put(&hive->kobj); + kfree(hive); + hive = NULL; + goto pro_end; + } + } else { + amdgpu_reset_get_reset_domain(adev->reset_domain); + hive->reset_domain = adev->reset_domain; + } + } + hive->hive_id = adev->gmc.xgmi.hive_id; INIT_LIST_HEAD(&hive->device_list); INIT_LIST_HEAD(&hive->node); mutex_init(&hive->hive_lock); - atomic_set(&hive->in_reset, 0); atomic_set(&hive->number_devices, 0); task_barrier_init(&hive->tb); hive->pstate = AMDGPU_XGMI_PSTATE_UNKNOWN; hive->hi_req_gpu = NULL; + /* * hive pstate on boot is high in vega20 so we have to go to low * pstate on after boot. @@ -490,6 +513,9 @@ int amdgpu_xgmi_update_topology(struct amdgpu_hive_info *hive, struct amdgpu_dev { int ret; + if (amdgpu_sriov_vf(adev)) + return 0; + /* Each psp need to set the latest topology */ ret = psp_xgmi_set_topology_info(&adev->psp, atomic_read(&hive->number_devices), @@ -728,56 +754,18 @@ int amdgpu_xgmi_remove_device(struct amdgpu_device *adev) amdgpu_put_xgmi_hive(hive); } - return psp_xgmi_terminate(&adev->psp); + return 0; } -static int amdgpu_xgmi_ras_late_init(struct amdgpu_device *adev) +static int amdgpu_xgmi_ras_late_init(struct amdgpu_device *adev, struct ras_common_if *ras_block) { - int r; - struct ras_ih_if ih_info = { - .cb = NULL, - }; - struct ras_fs_if fs_info = { - .sysfs_name = "xgmi_wafl_err_count", - }; - if (!adev->gmc.xgmi.supported || adev->gmc.xgmi.num_physical_nodes == 0) return 0; - adev->gmc.xgmi.ras_funcs->reset_ras_error_count(adev); + adev->gmc.xgmi.ras->ras_block.hw_ops->reset_ras_error_count(adev); - if (!adev->gmc.xgmi.ras_if) { - adev->gmc.xgmi.ras_if = kmalloc(sizeof(struct ras_common_if), GFP_KERNEL); - if (!adev->gmc.xgmi.ras_if) - return -ENOMEM; - adev->gmc.xgmi.ras_if->block = AMDGPU_RAS_BLOCK__XGMI_WAFL; - adev->gmc.xgmi.ras_if->type = AMDGPU_RAS_ERROR__MULTI_UNCORRECTABLE; - adev->gmc.xgmi.ras_if->sub_block_index = 0; - } - ih_info.head = fs_info.head = *adev->gmc.xgmi.ras_if; - r = amdgpu_ras_late_init(adev, adev->gmc.xgmi.ras_if, - &fs_info, &ih_info); - if (r || !amdgpu_ras_is_supported(adev, adev->gmc.xgmi.ras_if->block)) { - kfree(adev->gmc.xgmi.ras_if); - adev->gmc.xgmi.ras_if = NULL; - } - - return r; -} - -static void amdgpu_xgmi_ras_fini(struct amdgpu_device *adev) -{ - if (amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__XGMI_WAFL) && - adev->gmc.xgmi.ras_if) { - struct ras_common_if *ras_if = adev->gmc.xgmi.ras_if; - struct ras_ih_if ih_info = { - .cb = NULL, - }; - - amdgpu_ras_late_fini(adev, ras_if, &ih_info); - kfree(ras_if); - } + return amdgpu_ras_block_late_init(adev, ras_block); } uint64_t amdgpu_xgmi_get_relative_phy_addr(struct amdgpu_device *adev, @@ -809,9 +797,6 @@ static void amdgpu_xgmi_reset_ras_error_count(struct amdgpu_device *adev) xgmi_pcs_err_status_reg_vg20[i]); break; case CHIP_ALDEBARAN: - for (i = 0; i < ARRAY_SIZE(xgmi23_pcs_err_status_reg_aldebaran); i++) - pcs_clear_status(adev, - xgmi23_pcs_err_status_reg_aldebaran[i]); for (i = 0; i < ARRAY_SIZE(xgmi3x16_pcs_err_status_reg_aldebaran); i++) pcs_clear_status(adev, xgmi3x16_pcs_err_status_reg_aldebaran[i]); @@ -864,7 +849,7 @@ static int amdgpu_xgmi_query_pcs_error_status(struct amdgpu_device *adev, return 0; } -static int amdgpu_xgmi_query_ras_error_count(struct amdgpu_device *adev, +static void amdgpu_xgmi_query_ras_error_count(struct amdgpu_device *adev, void *ras_error_status) { struct ras_err_data *err_data = (struct ras_err_data *)ras_error_status; @@ -873,7 +858,7 @@ static int amdgpu_xgmi_query_ras_error_count(struct amdgpu_device *adev, uint32_t ue_cnt = 0, ce_cnt = 0; if (!amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__XGMI_WAFL)) - return -EINVAL; + return ; err_data->ue_count = 0; err_data->ce_count = 0; @@ -912,13 +897,6 @@ static int amdgpu_xgmi_query_ras_error_count(struct amdgpu_device *adev, } break; case CHIP_ALDEBARAN: - /* check xgmi23 pcs error */ - for (i = 0; i < ARRAY_SIZE(xgmi23_pcs_err_status_reg_aldebaran); i++) { - data = RREG32_PCIE(xgmi23_pcs_err_status_reg_aldebaran[i]); - if (data) - amdgpu_xgmi_query_pcs_error_status(adev, - data, &ue_cnt, &ce_cnt, true); - } /* check xgmi3x16 pcs error */ for (i = 0; i < ARRAY_SIZE(xgmi3x16_pcs_err_status_reg_aldebaran); i++) { data = RREG32_PCIE(xgmi3x16_pcs_err_status_reg_aldebaran[i]); @@ -939,17 +917,53 @@ static int amdgpu_xgmi_query_ras_error_count(struct amdgpu_device *adev, break; } - adev->gmc.xgmi.ras_funcs->reset_ras_error_count(adev); + adev->gmc.xgmi.ras->ras_block.hw_ops->reset_ras_error_count(adev); err_data->ue_count += ue_cnt; err_data->ce_count += ce_cnt; +} - return 0; +/* Trigger XGMI/WAFL error */ +static int amdgpu_ras_error_inject_xgmi(struct amdgpu_device *adev, void *inject_if) +{ + int ret = 0; + struct ta_ras_trigger_error_input *block_info = + (struct ta_ras_trigger_error_input *)inject_if; + + if (amdgpu_dpm_set_df_cstate(adev, DF_CSTATE_DISALLOW)) + dev_warn(adev->dev, "Failed to disallow df cstate"); + + if (amdgpu_dpm_allow_xgmi_power_down(adev, false)) + dev_warn(adev->dev, "Failed to disallow XGMI power down"); + + ret = psp_ras_trigger_error(&adev->psp, block_info); + + if (amdgpu_ras_intr_triggered()) + return ret; + + if (amdgpu_dpm_allow_xgmi_power_down(adev, true)) + dev_warn(adev->dev, "Failed to allow XGMI power down"); + + if (amdgpu_dpm_set_df_cstate(adev, DF_CSTATE_ALLOW)) + dev_warn(adev->dev, "Failed to allow df cstate"); + + return ret; } -const struct amdgpu_xgmi_ras_funcs xgmi_ras_funcs = { - .ras_late_init = amdgpu_xgmi_ras_late_init, - .ras_fini = amdgpu_xgmi_ras_fini, +struct amdgpu_ras_block_hw_ops xgmi_ras_hw_ops = { .query_ras_error_count = amdgpu_xgmi_query_ras_error_count, .reset_ras_error_count = amdgpu_xgmi_reset_ras_error_count, + .ras_error_inject = amdgpu_ras_error_inject_xgmi, +}; + +struct amdgpu_xgmi_ras xgmi_ras = { + .ras_block = { + .ras_comm = { + .name = "xgmi_wafl", + .block = AMDGPU_RAS_BLOCK__XGMI_WAFL, + .type = AMDGPU_RAS_ERROR__MULTI_UNCORRECTABLE, + }, + .hw_ops = &xgmi_ras_hw_ops, + .ras_late_init = amdgpu_xgmi_ras_late_init, + }, }; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.h index d2189bf7d428..30dcc1681b4e 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.h @@ -24,7 +24,7 @@ #include <drm/task_barrier.h> #include "amdgpu_psp.h" - +#include "amdgpu_ras.h" struct amdgpu_hive_info { struct kobject kobj; @@ -33,7 +33,6 @@ struct amdgpu_hive_info { struct list_head node; atomic_t number_devices; struct mutex hive_lock; - atomic_t in_reset; int hi_req_count; struct amdgpu_device *hi_req_gpu; struct task_barrier tb; @@ -42,6 +41,9 @@ struct amdgpu_hive_info { AMDGPU_XGMI_PSTATE_MAX_VEGA20, AMDGPU_XGMI_PSTATE_UNKNOWN } pstate; + + struct amdgpu_reset_domain *reset_domain; + uint32_t device_remove_count; }; struct amdgpu_pcs_ras_field { @@ -50,7 +52,7 @@ struct amdgpu_pcs_ras_field { uint32_t pcs_err_shift; }; -extern const struct amdgpu_xgmi_ras_funcs xgmi_ras_funcs; +extern struct amdgpu_xgmi_ras xgmi_ras; struct amdgpu_hive_info *amdgpu_get_xgmi_hive(struct amdgpu_device *adev); void amdgpu_put_xgmi_hive(struct amdgpu_hive_info *hive); int amdgpu_xgmi_update_topology(struct amdgpu_hive_info *hive, struct amdgpu_device *adev); @@ -66,7 +68,8 @@ uint64_t amdgpu_xgmi_get_relative_phy_addr(struct amdgpu_device *adev, static inline bool amdgpu_xgmi_same_hive(struct amdgpu_device *adev, struct amdgpu_device *bo_adev) { - return (adev != bo_adev && + return (amdgpu_use_xgmi_p2p && + adev != bo_adev && adev->gmc.xgmi.hive_id && adev->gmc.xgmi.hive_id == bo_adev->gmc.xgmi.hive_id); } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgv_sriovmsg.h b/drivers/gpu/drm/amd/amdgpu/amdgv_sriovmsg.h index 7326b6c1b71c..6c97148ca0ed 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgv_sriovmsg.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgv_sriovmsg.h @@ -1,34 +1,33 @@ /* - * Copyright 2018-2019 Advanced Micro Devices, Inc. + * Copyright (c) 2018-2021 Advanced Micro Devices, Inc. All rights reserved. * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * the rights to use, copy, modify, merge, publish, distribute, sublicense, - * and/or sell copies of the Software, and to permit persons to whom the - * Software is furnished to do so, subject to the following conditions: + * Permission is hereby granted, free of charge, to any person obtaining a copy + * of this software and associated documentation files (the "Software"), to deal + * in the Software without restriction, including without limitation the rights + * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell + * copies of the Software, and to permit persons to whom the Software is + * furnished to do so, subject to the following conditions: * * The above copyright notice and this permission notice shall be included in * all copies or substantial portions of the Software. * * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL - * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR - * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, - * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR - * OTHER DEALINGS IN THE SOFTWARE. - * + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE + * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, + * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN + * THE SOFTWARE. */ #ifndef AMDGV_SRIOV_MSG__H_ #define AMDGV_SRIOV_MSG__H_ /* unit in kilobytes */ -#define AMD_SRIOV_MSG_VBIOS_OFFSET 0 -#define AMD_SRIOV_MSG_VBIOS_SIZE_KB 64 -#define AMD_SRIOV_MSG_DATAEXCHANGE_OFFSET_KB AMD_SRIOV_MSG_VBIOS_SIZE_KB -#define AMD_SRIOV_MSG_DATAEXCHANGE_SIZE_KB 4 +#define AMD_SRIOV_MSG_VBIOS_OFFSET 0 +#define AMD_SRIOV_MSG_VBIOS_SIZE_KB 64 +#define AMD_SRIOV_MSG_DATAEXCHANGE_OFFSET_KB AMD_SRIOV_MSG_VBIOS_SIZE_KB +#define AMD_SRIOV_MSG_DATAEXCHANGE_SIZE_KB 4 /* * layout @@ -51,10 +50,10 @@ * v2 defined in amdgim * v3 current */ -#define AMD_SRIOV_MSG_FW_VRAM_PF2VF_VER 2 -#define AMD_SRIOV_MSG_FW_VRAM_VF2PF_VER 3 +#define AMD_SRIOV_MSG_FW_VRAM_PF2VF_VER 2 +#define AMD_SRIOV_MSG_FW_VRAM_VF2PF_VER 3 -#define AMD_SRIOV_MSG_RESERVE_UCODE 24 +#define AMD_SRIOV_MSG_RESERVE_UCODE 24 #define AMD_SRIOV_MSG_RESERVE_VCN_INST 4 @@ -71,6 +70,7 @@ enum amd_sriov_ucode_engine_id { AMD_SRIOV_UCODE_ID_RLC_SRLS, AMD_SRIOV_UCODE_ID_MEC, AMD_SRIOV_UCODE_ID_MEC2, + AMD_SRIOV_UCODE_ID_IMU, AMD_SRIOV_UCODE_ID_SOS, AMD_SRIOV_UCODE_ID_ASD, AMD_SRIOV_UCODE_ID_TA_RAS, @@ -83,19 +83,19 @@ enum amd_sriov_ucode_engine_id { AMD_SRIOV_UCODE_ID__MAX }; -#pragma pack(push, 1) // PF2VF / VF2PF data areas are byte packed +#pragma pack(push, 1) // PF2VF / VF2PF data areas are byte packed union amd_sriov_msg_feature_flags { struct { - uint32_t error_log_collect : 1; - uint32_t host_load_ucodes : 1; - uint32_t host_flr_vramlost : 1; - uint32_t mm_bw_management : 1; - uint32_t pp_one_vf_mode : 1; - uint32_t reg_indirect_acc : 1; - uint32_t reserved : 26; + uint32_t error_log_collect : 1; + uint32_t host_load_ucodes : 1; + uint32_t host_flr_vramlost : 1; + uint32_t mm_bw_management : 1; + uint32_t pp_one_vf_mode : 1; + uint32_t reg_indirect_acc : 1; + uint32_t reserved : 26; } flags; - uint32_t all; + uint32_t all; }; union amd_sriov_reg_access_flags { @@ -110,10 +110,10 @@ union amd_sriov_reg_access_flags { union amd_sriov_msg_os_info { struct { - uint32_t windows : 1; - uint32_t reserved : 31; + uint32_t windows : 1; + uint32_t reserved : 31; } info; - uint32_t all; + uint32_t all; }; struct amd_sriov_msg_uuid_info { @@ -156,6 +156,7 @@ struct amd_sriov_msg_pf2vf_info_header { uint32_t reserved[2]; }; +#define AMD_SRIOV_MSG_PF2VF_INFO_FILLED_SIZE (48) struct amd_sriov_msg_pf2vf_info { /* header contains size and version */ struct amd_sriov_msg_pf2vf_info_header header; @@ -204,10 +205,10 @@ struct amd_sriov_msg_pf2vf_info { } mm_bw_management[AMD_SRIOV_MSG_RESERVE_VCN_INST]; /* UUID info */ struct amd_sriov_msg_uuid_info uuid_info; - /* pcie atomic Ops info */ - uint32_t pcie_atomic_ops_enabled_flags; + /* PCIE atomic ops support flag */ + uint32_t pcie_atomic_ops_support_flags; /* reserved */ - uint32_t reserved[256 - 48]; + uint32_t reserved[256 - AMD_SRIOV_MSG_PF2VF_INFO_FILLED_SIZE]; }; struct amd_sriov_msg_vf2pf_info_header { @@ -219,12 +220,13 @@ struct amd_sriov_msg_vf2pf_info_header { uint32_t reserved[2]; }; +#define AMD_SRIOV_MSG_VF2PF_INFO_FILLED_SIZE (70) struct amd_sriov_msg_vf2pf_info { /* header contains size and version */ struct amd_sriov_msg_vf2pf_info_header header; uint32_t checksum; /* driver version */ - uint8_t driver_version[64]; + uint8_t driver_version[64]; /* driver certification, 1=WHQL, 0=None */ uint32_t driver_cert; /* guest OS type and version */ @@ -258,13 +260,13 @@ struct amd_sriov_msg_vf2pf_info { uint32_t fb_size; /* guest ucode data, each one is 1.25 Dword */ struct { - uint8_t id; + uint8_t id; uint32_t version; } ucode_info[AMD_SRIOV_MSG_RESERVE_UCODE]; uint64_t dummy_page_addr; /* reserved */ - uint32_t reserved[256-70]; + uint32_t reserved[256 - AMD_SRIOV_MSG_VF2PF_INFO_FILLED_SIZE]; }; /* mailbox message send from guest to host */ @@ -276,7 +278,7 @@ enum amd_sriov_mailbox_request_message { MB_REQ_MSG_REQ_GPU_RESET_ACCESS, MB_REQ_MSG_REQ_GPU_INIT_DATA, - MB_REQ_MSG_LOG_VF_ERROR = 200, + MB_REQ_MSG_LOG_VF_ERROR = 200, }; /* mailbox message send from host to guest */ @@ -298,17 +300,15 @@ enum amd_sriov_gpu_init_data_version { GPU_INIT_DATA_READY_V1 = 1, }; -#pragma pack(pop) // Restore previous packing option +#pragma pack(pop) // Restore previous packing option /* checksum function between host and guest */ -unsigned int amd_sriov_msg_checksum(void *obj, - unsigned long obj_size, - unsigned int key, - unsigned int checksum); +unsigned int amd_sriov_msg_checksum(void *obj, unsigned long obj_size, unsigned int key, + unsigned int checksum); /* assertion at compile time */ #ifdef __linux__ -#define stringification(s) _stringification(s) +#define stringification(s) _stringification(s) #define _stringification(s) #s _Static_assert( @@ -319,13 +319,11 @@ _Static_assert( sizeof(struct amd_sriov_msg_pf2vf_info) == AMD_SRIOV_MSG_SIZE_KB << 10, "amd_sriov_msg_pf2vf_info must be " stringification(AMD_SRIOV_MSG_SIZE_KB) " KB"); -_Static_assert( - AMD_SRIOV_MSG_RESERVE_UCODE % 4 == 0, - "AMD_SRIOV_MSG_RESERVE_UCODE must be multiple of 4"); +_Static_assert(AMD_SRIOV_MSG_RESERVE_UCODE % 4 == 0, + "AMD_SRIOV_MSG_RESERVE_UCODE must be multiple of 4"); -_Static_assert( - AMD_SRIOV_MSG_RESERVE_UCODE > AMD_SRIOV_UCODE_ID__MAX, - "AMD_SRIOV_MSG_RESERVE_UCODE must be bigger than AMD_SRIOV_UCODE_ID__MAX"); +_Static_assert(AMD_SRIOV_MSG_RESERVE_UCODE > AMD_SRIOV_UCODE_ID__MAX, + "AMD_SRIOV_MSG_RESERVE_UCODE must be bigger than AMD_SRIOV_UCODE_ID__MAX"); #undef _stringification #undef stringification diff --git a/drivers/gpu/drm/amd/amdgpu/athub_v1_0.c b/drivers/gpu/drm/amd/amdgpu/athub_v1_0.c index 3ea557864320..a13c443ea10f 100644 --- a/drivers/gpu/drm/amd/amdgpu/athub_v1_0.c +++ b/drivers/gpu/drm/amd/amdgpu/athub_v1_0.c @@ -68,12 +68,13 @@ int athub_v1_0_set_clockgating(struct amdgpu_device *adev, if (amdgpu_sriov_vf(adev)) return 0; - switch (adev->asic_type) { - case CHIP_VEGA10: - case CHIP_VEGA12: - case CHIP_VEGA20: - case CHIP_RAVEN: - case CHIP_RENOIR: + switch (adev->ip_versions[ATHUB_HWIP][0]) { + case IP_VERSION(9, 0, 0): + case IP_VERSION(9, 1, 0): + case IP_VERSION(9, 2, 0): + case IP_VERSION(9, 3, 0): + case IP_VERSION(9, 4, 0): + case IP_VERSION(1, 5, 0): athub_update_medium_grain_clock_gating(adev, state == AMD_CG_STATE_GATE); athub_update_medium_grain_light_sleep(adev, @@ -86,7 +87,7 @@ int athub_v1_0_set_clockgating(struct amdgpu_device *adev, return 0; } -void athub_v1_0_get_clockgating(struct amdgpu_device *adev, u32 *flags) +void athub_v1_0_get_clockgating(struct amdgpu_device *adev, u64 *flags) { int data; diff --git a/drivers/gpu/drm/amd/amdgpu/athub_v1_0.h b/drivers/gpu/drm/amd/amdgpu/athub_v1_0.h index b279af59e34f..6be0a6704ea7 100644 --- a/drivers/gpu/drm/amd/amdgpu/athub_v1_0.h +++ b/drivers/gpu/drm/amd/amdgpu/athub_v1_0.h @@ -25,6 +25,6 @@ int athub_v1_0_set_clockgating(struct amdgpu_device *adev, enum amd_clockgating_state state); -void athub_v1_0_get_clockgating(struct amdgpu_device *adev, u32 *flags); +void athub_v1_0_get_clockgating(struct amdgpu_device *adev, u64 *flags); #endif diff --git a/drivers/gpu/drm/amd/amdgpu/athub_v2_0.c b/drivers/gpu/drm/amd/amdgpu/athub_v2_0.c index ab6a07e5e8c4..a9521c98e7f7 100644 --- a/drivers/gpu/drm/amd/amdgpu/athub_v2_0.c +++ b/drivers/gpu/drm/amd/amdgpu/athub_v2_0.c @@ -78,6 +78,7 @@ int athub_v2_0_set_clockgating(struct amdgpu_device *adev, return 0; switch (adev->ip_versions[ATHUB_HWIP][0]) { + case IP_VERSION(1, 3, 1): case IP_VERSION(2, 0, 0): case IP_VERSION(2, 0, 2): athub_v2_0_update_medium_grain_clock_gating(adev, @@ -92,7 +93,7 @@ int athub_v2_0_set_clockgating(struct amdgpu_device *adev, return 0; } -void athub_v2_0_get_clockgating(struct amdgpu_device *adev, u32 *flags) +void athub_v2_0_get_clockgating(struct amdgpu_device *adev, u64 *flags) { int data; diff --git a/drivers/gpu/drm/amd/amdgpu/athub_v2_0.h b/drivers/gpu/drm/amd/amdgpu/athub_v2_0.h index 02932c1c8bab..8b763f6dfd81 100644 --- a/drivers/gpu/drm/amd/amdgpu/athub_v2_0.h +++ b/drivers/gpu/drm/amd/amdgpu/athub_v2_0.h @@ -25,6 +25,6 @@ int athub_v2_0_set_clockgating(struct amdgpu_device *adev, enum amd_clockgating_state state); -void athub_v2_0_get_clockgating(struct amdgpu_device *adev, u32 *flags); +void athub_v2_0_get_clockgating(struct amdgpu_device *adev, u64 *flags); #endif diff --git a/drivers/gpu/drm/amd/amdgpu/athub_v2_1.c b/drivers/gpu/drm/amd/amdgpu/athub_v2_1.c index 2edefd10e56c..78508ae6a670 100644 --- a/drivers/gpu/drm/amd/amdgpu/athub_v2_1.c +++ b/drivers/gpu/drm/amd/amdgpu/athub_v2_1.c @@ -74,6 +74,7 @@ int athub_v2_1_set_clockgating(struct amdgpu_device *adev, case IP_VERSION(2, 1, 0): case IP_VERSION(2, 1, 1): case IP_VERSION(2, 1, 2): + case IP_VERSION(2, 4, 0): athub_v2_1_update_medium_grain_clock_gating(adev, state == AMD_CG_STATE_GATE); athub_v2_1_update_medium_grain_light_sleep(adev, state == AMD_CG_STATE_GATE); break; @@ -84,7 +85,7 @@ int athub_v2_1_set_clockgating(struct amdgpu_device *adev, return 0; } -void athub_v2_1_get_clockgating(struct amdgpu_device *adev, u32 *flags) +void athub_v2_1_get_clockgating(struct amdgpu_device *adev, u64 *flags) { int data; diff --git a/drivers/gpu/drm/amd/amdgpu/athub_v2_1.h b/drivers/gpu/drm/amd/amdgpu/athub_v2_1.h index 5e6824c0f591..b799f14bce03 100644 --- a/drivers/gpu/drm/amd/amdgpu/athub_v2_1.h +++ b/drivers/gpu/drm/amd/amdgpu/athub_v2_1.h @@ -25,6 +25,6 @@ int athub_v2_1_set_clockgating(struct amdgpu_device *adev, enum amd_clockgating_state state); -void athub_v2_1_get_clockgating(struct amdgpu_device *adev, u32 *flags); +void athub_v2_1_get_clockgating(struct amdgpu_device *adev, u64 *flags); #endif diff --git a/drivers/gpu/drm/amd/amdgpu/athub_v3_0.c b/drivers/gpu/drm/amd/amdgpu/athub_v3_0.c new file mode 100644 index 000000000000..f0e235f98afb --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/athub_v3_0.c @@ -0,0 +1,130 @@ +/* + * Copyright 2021 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#include "amdgpu.h" +#include "athub_v3_0.h" +#include "athub/athub_3_0_0_offset.h" +#include "athub/athub_3_0_0_sh_mask.h" +#include "navi10_enum.h" +#include "soc15_common.h" + +#define regATHUB_MISC_CNTL_V3_0_1 0x00d7 +#define regATHUB_MISC_CNTL_V3_0_1_BASE_IDX 0 + + +static uint32_t athub_v3_0_get_cg_cntl(struct amdgpu_device *adev) +{ + uint32_t data; + + switch (adev->ip_versions[ATHUB_HWIP][0]) { + case IP_VERSION(3, 0, 1): + data = RREG32_SOC15(ATHUB, 0, regATHUB_MISC_CNTL_V3_0_1); + break; + default: + data = RREG32_SOC15(ATHUB, 0, regATHUB_MISC_CNTL); + break; + } + return data; +} + +static void athub_v3_0_set_cg_cntl(struct amdgpu_device *adev, uint32_t data) +{ + switch (adev->ip_versions[ATHUB_HWIP][0]) { + case IP_VERSION(3, 0, 1): + WREG32_SOC15(ATHUB, 0, regATHUB_MISC_CNTL_V3_0_1, data); + break; + default: + WREG32_SOC15(ATHUB, 0, regATHUB_MISC_CNTL, data); + break; + } +} + +static void +athub_v3_0_update_medium_grain_clock_gating(struct amdgpu_device *adev, + bool enable) +{ + uint32_t def, data; + + def = data = athub_v3_0_get_cg_cntl(adev); + + if (enable && (adev->cg_flags & AMD_CG_SUPPORT_ATHUB_MGCG)) + data |= ATHUB_MISC_CNTL__CG_ENABLE_MASK; + else + data &= ~ATHUB_MISC_CNTL__CG_ENABLE_MASK; + + if (def != data) + athub_v3_0_set_cg_cntl(adev, data); +} + +static void +athub_v3_0_update_medium_grain_light_sleep(struct amdgpu_device *adev, + bool enable) +{ + uint32_t def, data; + + def = data = athub_v3_0_get_cg_cntl(adev); + + if (enable && (adev->cg_flags & AMD_CG_SUPPORT_ATHUB_LS)) + data |= ATHUB_MISC_CNTL__CG_MEM_LS_ENABLE_MASK; + else + data &= ~ATHUB_MISC_CNTL__CG_MEM_LS_ENABLE_MASK; + + if (def != data) + athub_v3_0_set_cg_cntl(adev, data); +} + +int athub_v3_0_set_clockgating(struct amdgpu_device *adev, + enum amd_clockgating_state state) +{ + if (amdgpu_sriov_vf(adev)) + return 0; + + switch (adev->ip_versions[ATHUB_HWIP][0]) { + case IP_VERSION(3, 0, 0): + case IP_VERSION(3, 0, 1): + case IP_VERSION(3, 0, 2): + athub_v3_0_update_medium_grain_clock_gating(adev, + state == AMD_CG_STATE_GATE); + athub_v3_0_update_medium_grain_light_sleep(adev, + state == AMD_CG_STATE_GATE); + break; + default: + break; + } + + return 0; +} + +void athub_v3_0_get_clockgating(struct amdgpu_device *adev, u64 *flags) +{ + int data; + + /* AMD_CG_SUPPORT_ATHUB_MGCG */ + data = athub_v3_0_get_cg_cntl(adev); + if (data & ATHUB_MISC_CNTL__CG_ENABLE_MASK) + *flags |= AMD_CG_SUPPORT_ATHUB_MGCG; + + /* AMD_CG_SUPPORT_ATHUB_LS */ + if (data & ATHUB_MISC_CNTL__CG_MEM_LS_ENABLE_MASK) + *flags |= AMD_CG_SUPPORT_ATHUB_LS; +} diff --git a/drivers/gpu/drm/amd/amdgpu/athub_v3_0.h b/drivers/gpu/drm/amd/amdgpu/athub_v3_0.h new file mode 100644 index 000000000000..e08a7d564365 --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/athub_v3_0.h @@ -0,0 +1,30 @@ +/* + * Copyright 2021 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ +#ifndef __ATHUB_V3_0_H__ +#define __ATHUB_V3_0_H__ + +int athub_v3_0_set_clockgating(struct amdgpu_device *adev, + enum amd_clockgating_state state); +void athub_v3_0_get_clockgating(struct amdgpu_device *adev, u64 *flags); + +#endif diff --git a/drivers/gpu/drm/amd/amdgpu/atom.c b/drivers/gpu/drm/amd/amdgpu/atom.c index 6fa2229b7229..1c5d9388ad0b 100644 --- a/drivers/gpu/drm/amd/amdgpu/atom.c +++ b/drivers/gpu/drm/amd/amdgpu/atom.c @@ -25,6 +25,8 @@ #include <linux/module.h> #include <linux/sched.h> #include <linux/slab.h> +#include <linux/string_helpers.h> + #include <asm/unaligned.h> #include <drm/drm_util.h> @@ -740,7 +742,7 @@ static void atom_op_jump(atom_exec_context *ctx, int *ptr, int arg) break; } if (arg != ATOM_COND_ALWAYS) - SDEBUG(" taken: %s\n", execute ? "yes" : "no"); + SDEBUG(" taken: %s\n", str_yes_no(execute)); SDEBUG(" target: 0x%04X\n", target); if (execute) { if (ctx->last_jump == (ctx->start + target)) { diff --git a/drivers/gpu/drm/amd/amdgpu/atombios_dp.c b/drivers/gpu/drm/amd/amdgpu/atombios_dp.c index f327becb022f..87c41e0e9b7c 100644 --- a/drivers/gpu/drm/amd/amdgpu/atombios_dp.c +++ b/drivers/gpu/drm/amd/amdgpu/atombios_dp.c @@ -26,6 +26,8 @@ */ #include <drm/amdgpu_drm.h> +#include <drm/display/drm_dp_helper.h> + #include "amdgpu.h" #include "atom.h" @@ -34,7 +36,6 @@ #include "atombios_dp.h" #include "amdgpu_connectors.h" #include "amdgpu_atombios.h" -#include <drm/drm_dp_helper.h> /* move these to drm_dp_helper.c/h */ #define DP_LINK_CONFIGURATION_SIZE 9 diff --git a/drivers/gpu/drm/amd/amdgpu/atombios_encoders.c b/drivers/gpu/drm/amd/amdgpu/atombios_encoders.c index a92d86e12718..6be9ac2b9c5b 100644 --- a/drivers/gpu/drm/amd/amdgpu/atombios_encoders.c +++ b/drivers/gpu/drm/amd/amdgpu/atombios_encoders.c @@ -26,6 +26,8 @@ #include <linux/pci.h> +#include <acpi/video.h> + #include <drm/drm_crtc_helper.h> #include <drm/amdgpu_drm.h> #include "amdgpu.h" @@ -118,8 +120,6 @@ amdgpu_atombios_encoder_set_backlight_level(struct amdgpu_encoder *amdgpu_encode } } -#if defined(CONFIG_BACKLIGHT_CLASS_DEVICE) || defined(CONFIG_BACKLIGHT_CLASS_DEVICE_MODULE) - static u8 amdgpu_atombios_encoder_backlight_level(struct backlight_device *bd) { u8 level; @@ -184,7 +184,12 @@ void amdgpu_atombios_encoder_init_backlight(struct amdgpu_encoder *amdgpu_encode return; if (!(adev->mode_info.firmware_flags & ATOM_BIOS_INFO_BL_CONTROLLED_BY_GPU)) - return; + goto register_acpi_backlight; + + if (!acpi_video_backlight_use_native()) { + drm_info(dev, "Skipping amdgpu atom DIG backlight registration\n"); + goto register_acpi_backlight; + } pdata = kmalloc(sizeof(struct amdgpu_backlight_privdata), GFP_KERNEL); if (!pdata) { @@ -220,6 +225,11 @@ void amdgpu_atombios_encoder_init_backlight(struct amdgpu_encoder *amdgpu_encode error: kfree(pdata); return; + +register_acpi_backlight: + /* Try registering an ACPI video backlight device instead. */ + acpi_video_register_backlight(); + return; } void @@ -251,18 +261,6 @@ amdgpu_atombios_encoder_fini_backlight(struct amdgpu_encoder *amdgpu_encoder) } } -#else /* !CONFIG_BACKLIGHT_CLASS_DEVICE */ - -void amdgpu_atombios_encoder_init_backlight(struct amdgpu_encoder *encoder) -{ -} - -void amdgpu_atombios_encoder_fini_backlight(struct amdgpu_encoder *encoder) -{ -} - -#endif - bool amdgpu_atombios_encoder_is_digital(struct drm_encoder *encoder) { struct amdgpu_encoder *amdgpu_encoder = to_amdgpu_encoder(encoder); @@ -765,7 +763,6 @@ amdgpu_atombios_encoder_setup_dig_transmitter(struct drm_encoder *encoder, int a int dp_clock = 0; int dp_lane_count = 0; int connector_object_id = 0; - int igp_lane_info = 0; int dig_encoder = dig->dig_encoder; int hpd_id = AMDGPU_HPD_NONE; @@ -848,26 +845,6 @@ amdgpu_atombios_encoder_setup_dig_transmitter(struct drm_encoder *encoder, int a else args.v1.ucConfig |= ATOM_TRANSMITTER_CONFIG_DIG1_ENCODER; - if ((adev->flags & AMD_IS_APU) && - (amdgpu_encoder->encoder_id == ENCODER_OBJECT_ID_INTERNAL_UNIPHY)) { - if (is_dp || - !amdgpu_dig_monitor_is_duallink(encoder, amdgpu_encoder->pixel_clock)) { - if (igp_lane_info & 0x1) - args.v1.ucConfig |= ATOM_TRANSMITTER_CONFIG_LANE_0_3; - else if (igp_lane_info & 0x2) - args.v1.ucConfig |= ATOM_TRANSMITTER_CONFIG_LANE_4_7; - else if (igp_lane_info & 0x4) - args.v1.ucConfig |= ATOM_TRANSMITTER_CONFIG_LANE_8_11; - else if (igp_lane_info & 0x8) - args.v1.ucConfig |= ATOM_TRANSMITTER_CONFIG_LANE_12_15; - } else { - if (igp_lane_info & 0x3) - args.v1.ucConfig |= ATOM_TRANSMITTER_CONFIG_LANE_0_7; - else if (igp_lane_info & 0xc) - args.v1.ucConfig |= ATOM_TRANSMITTER_CONFIG_LANE_8_15; - } - } - if (dig->linkb) args.v1.ucConfig |= ATOM_TRANSMITTER_CONFIG_LINKB; else diff --git a/drivers/gpu/drm/amd/amdgpu/cik.c b/drivers/gpu/drm/amd/amdgpu/cik.c index 54f28c075f21..de6d10390ab2 100644 --- a/drivers/gpu/drm/amd/amdgpu/cik.c +++ b/drivers/gpu/drm/amd/amdgpu/cik.c @@ -1428,6 +1428,10 @@ static int cik_asic_reset(struct amdgpu_device *adev) { int r; + /* APUs don't have full asic reset */ + if (adev->flags & AMD_IS_APU) + return 0; + if (cik_asic_reset_method(adev) == AMD_RESET_METHOD_BACO) { dev_info(adev->dev, "BACO reset\n"); r = amdgpu_dpm_baco_reset(adev); @@ -1715,7 +1719,7 @@ static void cik_program_aspm(struct amdgpu_device *adev) bool disable_l0s = false, disable_l1 = false, disable_plloff_in_l1 = false; bool disable_clkreq = false; - if (amdgpu_aspm == 0) + if (!amdgpu_device_should_use_aspm(adev)) return; if (pci_is_root_bus(adev->pdev->bus)) diff --git a/drivers/gpu/drm/amd/amdgpu/cik_sdma.c b/drivers/gpu/drm/amd/amdgpu/cik_sdma.c index c8ebd108548d..cbca9866645c 100644 --- a/drivers/gpu/drm/amd/amdgpu/cik_sdma.c +++ b/drivers/gpu/drm/amd/amdgpu/cik_sdma.c @@ -164,7 +164,7 @@ static uint64_t cik_sdma_ring_get_rptr(struct amdgpu_ring *ring) { u32 rptr; - rptr = ring->adev->wb.wb[ring->rptr_offs]; + rptr = *ring->rptr_cpu_addr; return (rptr & 0x3fffc) >> 2; } @@ -195,7 +195,7 @@ static void cik_sdma_ring_set_wptr(struct amdgpu_ring *ring) struct amdgpu_device *adev = ring->adev; WREG32(mmSDMA0_GFX_RB_WPTR + sdma_offsets[ring->me], - (lower_32_bits(ring->wptr) << 2) & 0x3fffc); + (ring->wptr << 2) & 0x3fffc); } static void cik_sdma_ring_insert_nop(struct amdgpu_ring *ring, uint32_t count) @@ -309,14 +309,10 @@ static void cik_sdma_ring_emit_fence(struct amdgpu_ring *ring, u64 addr, u64 seq */ static void cik_sdma_gfx_stop(struct amdgpu_device *adev) { - struct amdgpu_ring *sdma0 = &adev->sdma.instance[0].ring; - struct amdgpu_ring *sdma1 = &adev->sdma.instance[1].ring; u32 rb_cntl; int i; - if ((adev->mman.buffer_funcs_ring == sdma0) || - (adev->mman.buffer_funcs_ring == sdma1)) - amdgpu_ttm_set_buffer_funcs_status(adev, false); + amdgpu_sdma_unset_buffer_funcs_helper(adev); for (i = 0; i < adev->sdma.num_instances; i++) { rb_cntl = RREG32(mmSDMA0_GFX_RB_CNTL + sdma_offsets[i]); @@ -436,12 +432,10 @@ static int cik_sdma_gfx_resume(struct amdgpu_device *adev) struct amdgpu_ring *ring; u32 rb_cntl, ib_cntl; u32 rb_bufsz; - u32 wb_offset; int i, j, r; for (i = 0; i < adev->sdma.num_instances; i++) { ring = &adev->sdma.instance[i].ring; - wb_offset = (ring->rptr_offs * 4); mutex_lock(&adev->srbm_mutex); for (j = 0; j < 16; j++) { @@ -477,9 +471,9 @@ static int cik_sdma_gfx_resume(struct amdgpu_device *adev) /* set the wb address whether it's enabled or not */ WREG32(mmSDMA0_GFX_RB_RPTR_ADDR_HI + sdma_offsets[i], - upper_32_bits(adev->wb.gpu_addr + wb_offset) & 0xFFFFFFFF); + upper_32_bits(ring->rptr_gpu_addr) & 0xFFFFFFFF); WREG32(mmSDMA0_GFX_RB_RPTR_ADDR_LO + sdma_offsets[i], - ((adev->wb.gpu_addr + wb_offset) & 0xFFFFFFFC)); + ((ring->rptr_gpu_addr) & 0xFFFFFFFC)); rb_cntl |= SDMA0_GFX_RB_CNTL__RPTR_WRITEBACK_ENABLE_MASK; @@ -487,7 +481,7 @@ static int cik_sdma_gfx_resume(struct amdgpu_device *adev) WREG32(mmSDMA0_GFX_RB_BASE_HI + sdma_offsets[i], ring->gpu_addr >> 40); ring->wptr = 0; - WREG32(mmSDMA0_GFX_RB_WPTR + sdma_offsets[i], lower_32_bits(ring->wptr) << 2); + WREG32(mmSDMA0_GFX_RB_WPTR + sdma_offsets[i], ring->wptr << 2); /* enable DMA RB */ WREG32(mmSDMA0_GFX_RB_CNTL + sdma_offsets[i], diff --git a/drivers/gpu/drm/amd/amdgpu/clearstate_gfx11.h b/drivers/gpu/drm/amd/amdgpu/clearstate_gfx11.h new file mode 100644 index 000000000000..a8b29d33c464 --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/clearstate_gfx11.h @@ -0,0 +1,997 @@ +/* + * Copyright 2021 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ +#ifndef __CLEARSTATE_GFX11_H_ +#define __CLEARSTATE_GFX11_H_ + +static const unsigned int gfx11_SECT_CONTEXT_def_1[] = +{ + 0x00000000, // DB_RENDER_CONTROL + 0x00000000, // DB_COUNT_CONTROL + 0x00000000, // DB_DEPTH_VIEW + 0x00000000, // DB_RENDER_OVERRIDE + 0x00000000, // DB_RENDER_OVERRIDE2 + 0x00000000, // DB_HTILE_DATA_BASE + 0, // HOLE + 0x00000000, // DB_DEPTH_SIZE_XY + 0x00000000, // DB_DEPTH_BOUNDS_MIN + 0x00000000, // DB_DEPTH_BOUNDS_MAX + 0x00000000, // DB_STENCIL_CLEAR + 0x00000000, // DB_DEPTH_CLEAR + 0x00000000, // PA_SC_SCREEN_SCISSOR_TL + 0x40004000, // PA_SC_SCREEN_SCISSOR_BR + 0, // HOLE + 0x00000000, // DB_RESERVED_REG_2 + 0x00000000, // DB_Z_INFO + 0x00000000, // DB_STENCIL_INFO + 0x00000000, // DB_Z_READ_BASE + 0x00000000, // DB_STENCIL_READ_BASE + 0x00000000, // DB_Z_WRITE_BASE + 0x00000000, // DB_STENCIL_WRITE_BASE + 0x00000000, // DB_RESERVED_REG_1 + 0x00000000, // DB_RESERVED_REG_3 + 0x00000000, // DB_SPI_VRS_CENTER_LOCATION + 0, // HOLE + 0x00000000, // DB_Z_READ_BASE_HI + 0x00000000, // DB_STENCIL_READ_BASE_HI + 0x00000000, // DB_Z_WRITE_BASE_HI + 0x00000000, // DB_STENCIL_WRITE_BASE_HI + 0x00000000, // DB_HTILE_DATA_BASE_HI + 0x00150055, // DB_RMI_L2_CACHE_CONTROL + 0x00000000, // TA_BC_BASE_ADDR + 0x00000000, // TA_BC_BASE_ADDR_HI + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0x00000000, // COHER_DEST_BASE_HI_0 + 0x00000000, // COHER_DEST_BASE_HI_1 + 0x00000000, // COHER_DEST_BASE_HI_2 + 0x00000000, // COHER_DEST_BASE_HI_3 + 0x00000000, // COHER_DEST_BASE_2 + 0x00000000, // COHER_DEST_BASE_3 + 0x00000000, // PA_SC_WINDOW_OFFSET + 0x80000000, // PA_SC_WINDOW_SCISSOR_TL + 0x40004000, // PA_SC_WINDOW_SCISSOR_BR + 0x0000ffff, // PA_SC_CLIPRECT_RULE + 0x00000000, // PA_SC_CLIPRECT_0_TL + 0x40004000, // PA_SC_CLIPRECT_0_BR + 0x00000000, // PA_SC_CLIPRECT_1_TL + 0x40004000, // PA_SC_CLIPRECT_1_BR + 0x00000000, // PA_SC_CLIPRECT_2_TL + 0x40004000, // PA_SC_CLIPRECT_2_BR + 0x00000000, // PA_SC_CLIPRECT_3_TL + 0x40004000, // PA_SC_CLIPRECT_3_BR + 0xaa99aaaa, // PA_SC_EDGERULE + 0x00000000, // PA_SU_HARDWARE_SCREEN_OFFSET + 0xffffffff, // CB_TARGET_MASK + 0xffffffff, // CB_SHADER_MASK + 0x80000000, // PA_SC_GENERIC_SCISSOR_TL + 0x40004000, // PA_SC_GENERIC_SCISSOR_BR + 0x00000000, // COHER_DEST_BASE_0 + 0x00000000, // COHER_DEST_BASE_1 + 0x80000000, // PA_SC_VPORT_SCISSOR_0_TL + 0x40004000, // PA_SC_VPORT_SCISSOR_0_BR + 0x80000000, // PA_SC_VPORT_SCISSOR_1_TL + 0x40004000, // PA_SC_VPORT_SCISSOR_1_BR + 0x80000000, // PA_SC_VPORT_SCISSOR_2_TL + 0x40004000, // PA_SC_VPORT_SCISSOR_2_BR + 0x80000000, // PA_SC_VPORT_SCISSOR_3_TL + 0x40004000, // PA_SC_VPORT_SCISSOR_3_BR + 0x80000000, // PA_SC_VPORT_SCISSOR_4_TL + 0x40004000, // PA_SC_VPORT_SCISSOR_4_BR + 0x80000000, // PA_SC_VPORT_SCISSOR_5_TL + 0x40004000, // PA_SC_VPORT_SCISSOR_5_BR + 0x80000000, // PA_SC_VPORT_SCISSOR_6_TL + 0x40004000, // PA_SC_VPORT_SCISSOR_6_BR + 0x80000000, // PA_SC_VPORT_SCISSOR_7_TL + 0x40004000, // PA_SC_VPORT_SCISSOR_7_BR + 0x80000000, // PA_SC_VPORT_SCISSOR_8_TL + 0x40004000, // PA_SC_VPORT_SCISSOR_8_BR + 0x80000000, // PA_SC_VPORT_SCISSOR_9_TL + 0x40004000, // PA_SC_VPORT_SCISSOR_9_BR + 0x80000000, // PA_SC_VPORT_SCISSOR_10_TL + 0x40004000, // PA_SC_VPORT_SCISSOR_10_BR + 0x80000000, // PA_SC_VPORT_SCISSOR_11_TL + 0x40004000, // PA_SC_VPORT_SCISSOR_11_BR + 0x80000000, // PA_SC_VPORT_SCISSOR_12_TL + 0x40004000, // PA_SC_VPORT_SCISSOR_12_BR + 0x80000000, // PA_SC_VPORT_SCISSOR_13_TL + 0x40004000, // PA_SC_VPORT_SCISSOR_13_BR + 0x80000000, // PA_SC_VPORT_SCISSOR_14_TL + 0x40004000, // PA_SC_VPORT_SCISSOR_14_BR + 0x80000000, // PA_SC_VPORT_SCISSOR_15_TL + 0x40004000, // PA_SC_VPORT_SCISSOR_15_BR + 0x00000000, // PA_SC_VPORT_ZMIN_0 + 0x3f800000, // PA_SC_VPORT_ZMAX_0 + 0x00000000, // PA_SC_VPORT_ZMIN_1 + 0x3f800000, // PA_SC_VPORT_ZMAX_1 + 0x00000000, // PA_SC_VPORT_ZMIN_2 + 0x3f800000, // PA_SC_VPORT_ZMAX_2 + 0x00000000, // PA_SC_VPORT_ZMIN_3 + 0x3f800000, // PA_SC_VPORT_ZMAX_3 + 0x00000000, // PA_SC_VPORT_ZMIN_4 + 0x3f800000, // PA_SC_VPORT_ZMAX_4 + 0x00000000, // PA_SC_VPORT_ZMIN_5 + 0x3f800000, // PA_SC_VPORT_ZMAX_5 + 0x00000000, // PA_SC_VPORT_ZMIN_6 + 0x3f800000, // PA_SC_VPORT_ZMAX_6 + 0x00000000, // PA_SC_VPORT_ZMIN_7 + 0x3f800000, // PA_SC_VPORT_ZMAX_7 + 0x00000000, // PA_SC_VPORT_ZMIN_8 + 0x3f800000, // PA_SC_VPORT_ZMAX_8 + 0x00000000, // PA_SC_VPORT_ZMIN_9 + 0x3f800000, // PA_SC_VPORT_ZMAX_9 + 0x00000000, // PA_SC_VPORT_ZMIN_10 + 0x3f800000, // PA_SC_VPORT_ZMAX_10 + 0x00000000, // PA_SC_VPORT_ZMIN_11 + 0x3f800000, // PA_SC_VPORT_ZMAX_11 + 0x00000000, // PA_SC_VPORT_ZMIN_12 + 0x3f800000, // PA_SC_VPORT_ZMAX_12 + 0x00000000, // PA_SC_VPORT_ZMIN_13 + 0x3f800000, // PA_SC_VPORT_ZMAX_13 + 0x00000000, // PA_SC_VPORT_ZMIN_14 + 0x3f800000, // PA_SC_VPORT_ZMAX_14 + 0x00000000, // PA_SC_VPORT_ZMIN_15 + 0x3f800000, // PA_SC_VPORT_ZMAX_15 + 0x00000000, // PA_SC_RASTER_CONFIG + 0x00000000, // PA_SC_RASTER_CONFIG_1 + 0x00000000, // PA_SC_SCREEN_EXTENT_CONTROL +}; +static const unsigned int gfx11_SECT_CONTEXT_def_2[] = +{ + 0x00000000, // CP_PERFMON_CNTX_CNTL + 0x00000000, // CP_PIPEID + 0x00000000, // CP_VMID + 0x00000000, // CONTEXT_RESERVED_REG0 + 0x00000000, // CONTEXT_RESERVED_REG1 + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0x00000000, // PA_SC_FSR_EN + 0x00000000, // PA_SC_FSR_FBW_RECURSIONS_X + 0x00000000, // PA_SC_FSR_FBW_RECURSIONS_Y + 0, // HOLE + 0x00000000, // PA_SC_VRS_OVERRIDE_CNTL + 0x00000000, // PA_SC_VRS_RATE_FEEDBACK_BASE + 0x00000000, // PA_SC_VRS_RATE_FEEDBACK_BASE_EXT + 0x00000000, // PA_SC_VRS_RATE_FEEDBACK_SIZE_XY + 0x00000000, // PA_SC_BINNER_OUTPUT_TIMEOUT_CNTL + 0x00000000, // PA_SC_VRS_RATE_CACHE_CNTL + 0, // HOLE + 0, // HOLE + 0x00000000, // PA_SC_VRS_RATE_BASE + 0x00000000, // PA_SC_VRS_RATE_BASE_EXT + 0x00000000, // PA_SC_VRS_RATE_SIZE_XY + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0x00000000, // VGT_MULTI_PRIM_IB_RESET_INDX + 0x00550055, // CB_RMI_GL2_CACHE_CONTROL + 0x00000000, // CB_BLEND_RED + 0x00000000, // CB_BLEND_GREEN + 0x00000000, // CB_BLEND_BLUE + 0x00000000, // CB_BLEND_ALPHA + 0x00000000, // CB_FDCC_CONTROL + 0x00000000, // CB_COVERAGE_OUT_CONTROL + 0x00000000, // DB_STENCIL_CONTROL + 0x01000000, // DB_STENCILREFMASK + 0x01000000, // DB_STENCILREFMASK_BF + 0, // HOLE + 0x00000000, // PA_CL_VPORT_XSCALE + 0x00000000, // PA_CL_VPORT_XOFFSET + 0x00000000, // PA_CL_VPORT_YSCALE + 0x00000000, // PA_CL_VPORT_YOFFSET + 0x00000000, // PA_CL_VPORT_ZSCALE + 0x00000000, // PA_CL_VPORT_ZOFFSET + 0x00000000, // PA_CL_VPORT_XSCALE_1 + 0x00000000, // PA_CL_VPORT_XOFFSET_1 + 0x00000000, // PA_CL_VPORT_YSCALE_1 + 0x00000000, // PA_CL_VPORT_YOFFSET_1 + 0x00000000, // PA_CL_VPORT_ZSCALE_1 + 0x00000000, // PA_CL_VPORT_ZOFFSET_1 + 0x00000000, // PA_CL_VPORT_XSCALE_2 + 0x00000000, // PA_CL_VPORT_XOFFSET_2 + 0x00000000, // PA_CL_VPORT_YSCALE_2 + 0x00000000, // PA_CL_VPORT_YOFFSET_2 + 0x00000000, // PA_CL_VPORT_ZSCALE_2 + 0x00000000, // PA_CL_VPORT_ZOFFSET_2 + 0x00000000, // PA_CL_VPORT_XSCALE_3 + 0x00000000, // PA_CL_VPORT_XOFFSET_3 + 0x00000000, // PA_CL_VPORT_YSCALE_3 + 0x00000000, // PA_CL_VPORT_YOFFSET_3 + 0x00000000, // PA_CL_VPORT_ZSCALE_3 + 0x00000000, // PA_CL_VPORT_ZOFFSET_3 + 0x00000000, // PA_CL_VPORT_XSCALE_4 + 0x00000000, // PA_CL_VPORT_XOFFSET_4 + 0x00000000, // PA_CL_VPORT_YSCALE_4 + 0x00000000, // PA_CL_VPORT_YOFFSET_4 + 0x00000000, // PA_CL_VPORT_ZSCALE_4 + 0x00000000, // PA_CL_VPORT_ZOFFSET_4 + 0x00000000, // PA_CL_VPORT_XSCALE_5 + 0x00000000, // PA_CL_VPORT_XOFFSET_5 + 0x00000000, // PA_CL_VPORT_YSCALE_5 + 0x00000000, // PA_CL_VPORT_YOFFSET_5 + 0x00000000, // PA_CL_VPORT_ZSCALE_5 + 0x00000000, // PA_CL_VPORT_ZOFFSET_5 + 0x00000000, // PA_CL_VPORT_XSCALE_6 + 0x00000000, // PA_CL_VPORT_XOFFSET_6 + 0x00000000, // PA_CL_VPORT_YSCALE_6 + 0x00000000, // PA_CL_VPORT_YOFFSET_6 + 0x00000000, // PA_CL_VPORT_ZSCALE_6 + 0x00000000, // PA_CL_VPORT_ZOFFSET_6 + 0x00000000, // PA_CL_VPORT_XSCALE_7 + 0x00000000, // PA_CL_VPORT_XOFFSET_7 + 0x00000000, // PA_CL_VPORT_YSCALE_7 + 0x00000000, // PA_CL_VPORT_YOFFSET_7 + 0x00000000, // PA_CL_VPORT_ZSCALE_7 + 0x00000000, // PA_CL_VPORT_ZOFFSET_7 + 0x00000000, // PA_CL_VPORT_XSCALE_8 + 0x00000000, // PA_CL_VPORT_XOFFSET_8 + 0x00000000, // PA_CL_VPORT_YSCALE_8 + 0x00000000, // PA_CL_VPORT_YOFFSET_8 + 0x00000000, // PA_CL_VPORT_ZSCALE_8 + 0x00000000, // PA_CL_VPORT_ZOFFSET_8 + 0x00000000, // PA_CL_VPORT_XSCALE_9 + 0x00000000, // PA_CL_VPORT_XOFFSET_9 + 0x00000000, // PA_CL_VPORT_YSCALE_9 + 0x00000000, // PA_CL_VPORT_YOFFSET_9 + 0x00000000, // PA_CL_VPORT_ZSCALE_9 + 0x00000000, // PA_CL_VPORT_ZOFFSET_9 + 0x00000000, // PA_CL_VPORT_XSCALE_10 + 0x00000000, // PA_CL_VPORT_XOFFSET_10 + 0x00000000, // PA_CL_VPORT_YSCALE_10 + 0x00000000, // PA_CL_VPORT_YOFFSET_10 + 0x00000000, // PA_CL_VPORT_ZSCALE_10 + 0x00000000, // PA_CL_VPORT_ZOFFSET_10 + 0x00000000, // PA_CL_VPORT_XSCALE_11 + 0x00000000, // PA_CL_VPORT_XOFFSET_11 + 0x00000000, // PA_CL_VPORT_YSCALE_11 + 0x00000000, // PA_CL_VPORT_YOFFSET_11 + 0x00000000, // PA_CL_VPORT_ZSCALE_11 + 0x00000000, // PA_CL_VPORT_ZOFFSET_11 + 0x00000000, // PA_CL_VPORT_XSCALE_12 + 0x00000000, // PA_CL_VPORT_XOFFSET_12 + 0x00000000, // PA_CL_VPORT_YSCALE_12 + 0x00000000, // PA_CL_VPORT_YOFFSET_12 + 0x00000000, // PA_CL_VPORT_ZSCALE_12 + 0x00000000, // PA_CL_VPORT_ZOFFSET_12 + 0x00000000, // PA_CL_VPORT_XSCALE_13 + 0x00000000, // PA_CL_VPORT_XOFFSET_13 + 0x00000000, // PA_CL_VPORT_YSCALE_13 + 0x00000000, // PA_CL_VPORT_YOFFSET_13 + 0x00000000, // PA_CL_VPORT_ZSCALE_13 + 0x00000000, // PA_CL_VPORT_ZOFFSET_13 + 0x00000000, // PA_CL_VPORT_XSCALE_14 + 0x00000000, // PA_CL_VPORT_XOFFSET_14 + 0x00000000, // PA_CL_VPORT_YSCALE_14 + 0x00000000, // PA_CL_VPORT_YOFFSET_14 + 0x00000000, // PA_CL_VPORT_ZSCALE_14 + 0x00000000, // PA_CL_VPORT_ZOFFSET_14 + 0x00000000, // PA_CL_VPORT_XSCALE_15 + 0x00000000, // PA_CL_VPORT_XOFFSET_15 + 0x00000000, // PA_CL_VPORT_YSCALE_15 + 0x00000000, // PA_CL_VPORT_YOFFSET_15 + 0x00000000, // PA_CL_VPORT_ZSCALE_15 + 0x00000000, // PA_CL_VPORT_ZOFFSET_15 + 0x00000000, // PA_CL_UCP_0_X + 0x00000000, // PA_CL_UCP_0_Y + 0x00000000, // PA_CL_UCP_0_Z + 0x00000000, // PA_CL_UCP_0_W + 0x00000000, // PA_CL_UCP_1_X + 0x00000000, // PA_CL_UCP_1_Y + 0x00000000, // PA_CL_UCP_1_Z + 0x00000000, // PA_CL_UCP_1_W + 0x00000000, // PA_CL_UCP_2_X + 0x00000000, // PA_CL_UCP_2_Y + 0x00000000, // PA_CL_UCP_2_Z + 0x00000000, // PA_CL_UCP_2_W + 0x00000000, // PA_CL_UCP_3_X + 0x00000000, // PA_CL_UCP_3_Y + 0x00000000, // PA_CL_UCP_3_Z + 0x00000000, // PA_CL_UCP_3_W + 0x00000000, // PA_CL_UCP_4_X + 0x00000000, // PA_CL_UCP_4_Y + 0x00000000, // PA_CL_UCP_4_Z + 0x00000000, // PA_CL_UCP_4_W + 0x00000000, // PA_CL_UCP_5_X + 0x00000000, // PA_CL_UCP_5_Y + 0x00000000, // PA_CL_UCP_5_Z + 0x00000000, // PA_CL_UCP_5_W + 0x00000000, // PA_CL_PROG_NEAR_CLIP_Z + 0x00000000, // PA_RATE_CNTL + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0x00000000, // SPI_PS_INPUT_CNTL_0 + 0x00000000, // SPI_PS_INPUT_CNTL_1 + 0x00000000, // SPI_PS_INPUT_CNTL_2 + 0x00000000, // SPI_PS_INPUT_CNTL_3 + 0x00000000, // SPI_PS_INPUT_CNTL_4 + 0x00000000, // SPI_PS_INPUT_CNTL_5 + 0x00000000, // SPI_PS_INPUT_CNTL_6 + 0x00000000, // SPI_PS_INPUT_CNTL_7 + 0x00000000, // SPI_PS_INPUT_CNTL_8 + 0x00000000, // SPI_PS_INPUT_CNTL_9 + 0x00000000, // SPI_PS_INPUT_CNTL_10 + 0x00000000, // SPI_PS_INPUT_CNTL_11 + 0x00000000, // SPI_PS_INPUT_CNTL_12 + 0x00000000, // SPI_PS_INPUT_CNTL_13 + 0x00000000, // SPI_PS_INPUT_CNTL_14 + 0x00000000, // SPI_PS_INPUT_CNTL_15 + 0x00000000, // SPI_PS_INPUT_CNTL_16 + 0x00000000, // SPI_PS_INPUT_CNTL_17 + 0x00000000, // SPI_PS_INPUT_CNTL_18 + 0x00000000, // SPI_PS_INPUT_CNTL_19 + 0x00000000, // SPI_PS_INPUT_CNTL_20 + 0x00000000, // SPI_PS_INPUT_CNTL_21 + 0x00000000, // SPI_PS_INPUT_CNTL_22 + 0x00000000, // SPI_PS_INPUT_CNTL_23 + 0x00000000, // SPI_PS_INPUT_CNTL_24 + 0x00000000, // SPI_PS_INPUT_CNTL_25 + 0x00000000, // SPI_PS_INPUT_CNTL_26 + 0x00000000, // SPI_PS_INPUT_CNTL_27 + 0x00000000, // SPI_PS_INPUT_CNTL_28 + 0x00000000, // SPI_PS_INPUT_CNTL_29 + 0x00000000, // SPI_PS_INPUT_CNTL_30 + 0x00000000, // SPI_PS_INPUT_CNTL_31 + 0x00000000, // SPI_VS_OUT_CONFIG + 0, // HOLE + 0x00000000, // SPI_PS_INPUT_ENA + 0x00000000, // SPI_PS_INPUT_ADDR + 0x00000000, // SPI_INTERP_CONTROL_0 + 0x00000002, // SPI_PS_IN_CONTROL + 0x00000000, // SPI_BARYC_SSAA_CNTL + 0x00000000, // SPI_BARYC_CNTL + 0, // HOLE + 0x00000000, // SPI_TMPRING_SIZE + 0x00000000, // SPI_GFX_SCRATCH_BASE_LO + 0x00000000, // SPI_GFX_SCRATCH_BASE_HI + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0x00000000, // SPI_SHADER_IDX_FORMAT + 0x00000000, // SPI_SHADER_POS_FORMAT + 0x00000000, // SPI_SHADER_Z_FORMAT + 0x00000000, // SPI_SHADER_COL_FORMAT + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0x00000000, // SX_PS_DOWNCONVERT_CONTROL + 0x00000000, // SX_PS_DOWNCONVERT + 0x00000000, // SX_BLEND_OPT_EPSILON + 0x00000000, // SX_BLEND_OPT_CONTROL + 0x00000000, // SX_MRT0_BLEND_OPT + 0x00000000, // SX_MRT1_BLEND_OPT + 0x00000000, // SX_MRT2_BLEND_OPT + 0x00000000, // SX_MRT3_BLEND_OPT + 0x00000000, // SX_MRT4_BLEND_OPT + 0x00000000, // SX_MRT5_BLEND_OPT + 0x00000000, // SX_MRT6_BLEND_OPT + 0x00000000, // SX_MRT7_BLEND_OPT + 0x00000000, // CB_BLEND0_CONTROL + 0x00000000, // CB_BLEND1_CONTROL + 0x00000000, // CB_BLEND2_CONTROL + 0x00000000, // CB_BLEND3_CONTROL + 0x00000000, // CB_BLEND4_CONTROL + 0x00000000, // CB_BLEND5_CONTROL + 0x00000000, // CB_BLEND6_CONTROL + 0x00000000, // CB_BLEND7_CONTROL +}; +static const unsigned int gfx11_SECT_CONTEXT_def_3[] = +{ + 0x00000000, // PA_CL_POINT_X_RAD + 0x00000000, // PA_CL_POINT_Y_RAD + 0x00000000, // PA_CL_POINT_SIZE + 0x00000000, // PA_CL_POINT_CULL_RAD +}; +static const unsigned int gfx11_SECT_CONTEXT_def_4[] = +{ + 0x00000000, // GE_MAX_OUTPUT_PER_SUBGROUP + 0x00000000, // DB_DEPTH_CONTROL + 0x00000000, // DB_EQAA + 0x00000000, // CB_COLOR_CONTROL + 0x00000000, // DB_SHADER_CONTROL + 0x00090000, // PA_CL_CLIP_CNTL + 0x00000004, // PA_SU_SC_MODE_CNTL + 0x00000000, // PA_CL_VTE_CNTL + 0x00000000, // PA_CL_VS_OUT_CNTL + 0x00000000, // PA_CL_NANINF_CNTL + 0x00000000, // PA_SU_LINE_STIPPLE_CNTL + 0x00000000, // PA_SU_LINE_STIPPLE_SCALE + 0x00000000, // PA_SU_PRIM_FILTER_CNTL + 0x00000000, // PA_SU_SMALL_PRIM_FILTER_CNTL + 0, // HOLE + 0x00000000, // PA_CL_NGG_CNTL + 0x00000000, // PA_SU_OVER_RASTERIZATION_CNTL + 0x00000000, // PA_STEREO_CNTL + 0x00000000, // PA_STATE_STEREO_X + 0x00000000, // PA_CL_VRS_CNTL + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0x00000000, // PA_SU_POINT_SIZE + 0x00000000, // PA_SU_POINT_MINMAX + 0x00000000, // PA_SU_LINE_CNTL + 0x00000000, // PA_SC_LINE_STIPPLE + 0, // HOLE + 0, // HOLE + 0x00000000, // VGT_HOS_MAX_TESS_LEVEL + 0x00000000, // VGT_HOS_MIN_TESS_LEVEL + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0x00000000, // VGT_GS_ONCHIP_CNTL + 0x00000000, // PA_SC_MODE_CNTL_0 + 0x00000000, // PA_SC_MODE_CNTL_1 + 0x00000000, // VGT_ENHANCE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0x00000000, // IA_ENHANCE +}; +static const unsigned int gfx11_SECT_CONTEXT_def_5[] = +{ + 0x00000000, // WD_ENHANCE + 0x00000000, // VGT_PRIMITIVEID_EN +}; +static const unsigned int gfx11_SECT_CONTEXT_def_6[] = +{ + 0x00000000, // VGT_PRIMITIVEID_RESET +}; +static const unsigned int gfx11_SECT_CONTEXT_def_7[] = +{ + 0x00000000, // VGT_DRAW_PAYLOAD_CNTL + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0x00000000, // VGT_ESGS_RING_ITEMSIZE + 0, // HOLE + 0x00000000, // VGT_REUSE_OFF + 0, // HOLE + 0x00000000, // DB_HTILE_SURFACE + 0x00000000, // DB_SRESULTS_COMPARE_STATE0 + 0x00000000, // DB_SRESULTS_COMPARE_STATE1 + 0x00000000, // DB_PRELOAD_CONTROL + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0x00000000, // VGT_STRMOUT_DRAW_OPAQUE_OFFSET + 0x00000000, // VGT_STRMOUT_DRAW_OPAQUE_BUFFER_FILLED_SIZE + 0x00000000, // VGT_STRMOUT_DRAW_OPAQUE_VERTEX_STRIDE + 0, // HOLE + 0x00000000, // VGT_GS_MAX_VERT_OUT + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0x00000000, // GE_NGG_SUBGRP_CNTL + 0x00000000, // VGT_TESS_DISTRIBUTION + 0x00000000, // VGT_SHADER_STAGES_EN + 0x00000000, // VGT_LS_HS_CONFIG + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0x00000000, // VGT_TF_PARAM + 0x00000000, // DB_ALPHA_TO_MASK + 0, // HOLE + 0x00000000, // PA_SU_POLY_OFFSET_DB_FMT_CNTL + 0x00000000, // PA_SU_POLY_OFFSET_CLAMP + 0x00000000, // PA_SU_POLY_OFFSET_FRONT_SCALE + 0x00000000, // PA_SU_POLY_OFFSET_FRONT_OFFSET + 0x00000000, // PA_SU_POLY_OFFSET_BACK_SCALE + 0x00000000, // PA_SU_POLY_OFFSET_BACK_OFFSET + 0x00000000, // VGT_GS_INSTANCE_CNT + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0x00000000, // PA_SC_CENTROID_PRIORITY_0 + 0x00000000, // PA_SC_CENTROID_PRIORITY_1 + 0x00001000, // PA_SC_LINE_CNTL + 0x00000000, // PA_SC_AA_CONFIG + 0x00000005, // PA_SU_VTX_CNTL + 0x3f800000, // PA_CL_GB_VERT_CLIP_ADJ + 0x3f800000, // PA_CL_GB_VERT_DISC_ADJ + 0x3f800000, // PA_CL_GB_HORZ_CLIP_ADJ + 0x3f800000, // PA_CL_GB_HORZ_DISC_ADJ + 0x00000000, // PA_SC_AA_SAMPLE_LOCS_PIXEL_X0Y0_0 + 0x00000000, // PA_SC_AA_SAMPLE_LOCS_PIXEL_X0Y0_1 + 0x00000000, // PA_SC_AA_SAMPLE_LOCS_PIXEL_X0Y0_2 + 0x00000000, // PA_SC_AA_SAMPLE_LOCS_PIXEL_X0Y0_3 + 0x00000000, // PA_SC_AA_SAMPLE_LOCS_PIXEL_X1Y0_0 + 0x00000000, // PA_SC_AA_SAMPLE_LOCS_PIXEL_X1Y0_1 + 0x00000000, // PA_SC_AA_SAMPLE_LOCS_PIXEL_X1Y0_2 + 0x00000000, // PA_SC_AA_SAMPLE_LOCS_PIXEL_X1Y0_3 + 0x00000000, // PA_SC_AA_SAMPLE_LOCS_PIXEL_X0Y1_0 + 0x00000000, // PA_SC_AA_SAMPLE_LOCS_PIXEL_X0Y1_1 + 0x00000000, // PA_SC_AA_SAMPLE_LOCS_PIXEL_X0Y1_2 + 0x00000000, // PA_SC_AA_SAMPLE_LOCS_PIXEL_X0Y1_3 + 0x00000000, // PA_SC_AA_SAMPLE_LOCS_PIXEL_X1Y1_0 + 0x00000000, // PA_SC_AA_SAMPLE_LOCS_PIXEL_X1Y1_1 + 0x00000000, // PA_SC_AA_SAMPLE_LOCS_PIXEL_X1Y1_2 + 0x00000000, // PA_SC_AA_SAMPLE_LOCS_PIXEL_X1Y1_3 + 0xffffffff, // PA_SC_AA_MASK_X0Y0_X1Y0 + 0xffffffff, // PA_SC_AA_MASK_X0Y1_X1Y1 + 0x00000000, // PA_SC_SHADER_CONTROL + 0x00000003, // PA_SC_BINNER_CNTL_0 + 0x00000000, // PA_SC_BINNER_CNTL_1 + 0x00100000, // PA_SC_CONSERVATIVE_RASTERIZATION_CNTL + 0x00000000, // PA_SC_NGG_MODE_CNTL + 0x00000000, // PA_SC_BINNER_CNTL_2 + 0, // HOLE + 0, // HOLE + 0x00000000, // CB_COLOR0_BASE + 0, // HOLE + 0, // HOLE + 0x00000000, // CB_COLOR0_VIEW + 0x00000000, // CB_COLOR0_INFO + 0x00000000, // CB_COLOR0_ATTRIB + 0x00000000, // CB_COLOR0_FDCC_CONTROL + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0x00000000, // CB_COLOR0_DCC_BASE + 0, // HOLE + 0x00000000, // CB_COLOR1_BASE + 0, // HOLE + 0, // HOLE + 0x00000000, // CB_COLOR1_VIEW + 0x00000000, // CB_COLOR1_INFO + 0x00000000, // CB_COLOR1_ATTRIB + 0x00000000, // CB_COLOR1_FDCC_CONTROL + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0x00000000, // CB_COLOR1_DCC_BASE + 0, // HOLE + 0x00000000, // CB_COLOR2_BASE + 0, // HOLE + 0, // HOLE + 0x00000000, // CB_COLOR2_VIEW + 0x00000000, // CB_COLOR2_INFO + 0x00000000, // CB_COLOR2_ATTRIB + 0x00000000, // CB_COLOR2_FDCC_CONTROL + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0x00000000, // CB_COLOR2_DCC_BASE + 0, // HOLE + 0x00000000, // CB_COLOR3_BASE + 0, // HOLE + 0, // HOLE + 0x00000000, // CB_COLOR3_VIEW + 0x00000000, // CB_COLOR3_INFO + 0x00000000, // CB_COLOR3_ATTRIB + 0x00000000, // CB_COLOR3_FDCC_CONTROL + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0x00000000, // CB_COLOR3_DCC_BASE + 0, // HOLE + 0x00000000, // CB_COLOR4_BASE + 0, // HOLE + 0, // HOLE + 0x00000000, // CB_COLOR4_VIEW + 0x00000000, // CB_COLOR4_INFO + 0x00000000, // CB_COLOR4_ATTRIB + 0x00000000, // CB_COLOR4_FDCC_CONTROL + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0x00000000, // CB_COLOR4_DCC_BASE + 0, // HOLE + 0x00000000, // CB_COLOR5_BASE + 0, // HOLE + 0, // HOLE + 0x00000000, // CB_COLOR5_VIEW + 0x00000000, // CB_COLOR5_INFO + 0x00000000, // CB_COLOR5_ATTRIB + 0x00000000, // CB_COLOR5_FDCC_CONTROL + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0x00000000, // CB_COLOR5_DCC_BASE + 0, // HOLE + 0x00000000, // CB_COLOR6_BASE + 0, // HOLE + 0, // HOLE + 0x00000000, // CB_COLOR6_VIEW + 0x00000000, // CB_COLOR6_INFO + 0x00000000, // CB_COLOR6_ATTRIB + 0x00000000, // CB_COLOR6_FDCC_CONTROL + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0x00000000, // CB_COLOR6_DCC_BASE + 0, // HOLE + 0x00000000, // CB_COLOR7_BASE + 0, // HOLE + 0, // HOLE + 0x00000000, // CB_COLOR7_VIEW + 0x00000000, // CB_COLOR7_INFO + 0x00000000, // CB_COLOR7_ATTRIB + 0x00000000, // CB_COLOR7_FDCC_CONTROL + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0x00000000, // CB_COLOR7_DCC_BASE + 0, // HOLE + 0x00000000, // CB_COLOR0_BASE_EXT + 0x00000000, // CB_COLOR1_BASE_EXT + 0x00000000, // CB_COLOR2_BASE_EXT + 0x00000000, // CB_COLOR3_BASE_EXT + 0x00000000, // CB_COLOR4_BASE_EXT + 0x00000000, // CB_COLOR5_BASE_EXT + 0x00000000, // CB_COLOR6_BASE_EXT + 0x00000000, // CB_COLOR7_BASE_EXT + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0x00000000, // CB_COLOR0_DCC_BASE_EXT + 0x00000000, // CB_COLOR1_DCC_BASE_EXT + 0x00000000, // CB_COLOR2_DCC_BASE_EXT + 0x00000000, // CB_COLOR3_DCC_BASE_EXT + 0x00000000, // CB_COLOR4_DCC_BASE_EXT + 0x00000000, // CB_COLOR5_DCC_BASE_EXT + 0x00000000, // CB_COLOR6_DCC_BASE_EXT + 0x00000000, // CB_COLOR7_DCC_BASE_EXT + 0x00000000, // CB_COLOR0_ATTRIB2 + 0x00000000, // CB_COLOR1_ATTRIB2 + 0x00000000, // CB_COLOR2_ATTRIB2 + 0x00000000, // CB_COLOR3_ATTRIB2 + 0x00000000, // CB_COLOR4_ATTRIB2 + 0x00000000, // CB_COLOR5_ATTRIB2 + 0x00000000, // CB_COLOR6_ATTRIB2 + 0x00000000, // CB_COLOR7_ATTRIB2 + 0x00000000, // CB_COLOR0_ATTRIB3 + 0x00000000, // CB_COLOR1_ATTRIB3 + 0x00000000, // CB_COLOR2_ATTRIB3 + 0x00000000, // CB_COLOR3_ATTRIB3 + 0x00000000, // CB_COLOR4_ATTRIB3 + 0x00000000, // CB_COLOR5_ATTRIB3 + 0x00000000, // CB_COLOR6_ATTRIB3 + 0x00000000, // CB_COLOR7_ATTRIB3 +}; +static const struct cs_extent_def gfx11_SECT_CONTEXT_defs[] = +{ + {gfx11_SECT_CONTEXT_def_1, 0x0000a000, 215 }, + {gfx11_SECT_CONTEXT_def_2, 0x0000a0d8, 272 }, + {gfx11_SECT_CONTEXT_def_3, 0x0000a1f5, 4 }, + {gfx11_SECT_CONTEXT_def_4, 0x0000a1ff, 158 }, + {gfx11_SECT_CONTEXT_def_5, 0x0000a2a0, 2 }, + {gfx11_SECT_CONTEXT_def_6, 0x0000a2a3, 1 }, + {gfx11_SECT_CONTEXT_def_7, 0x0000a2a6, 282 }, + { 0, 0, 0 } +}; +static const struct cs_section_def gfx11_cs_data[] = { + { gfx11_SECT_CONTEXT_defs, SECT_CONTEXT }, + { 0, SECT_NONE } +}; + +#endif /* __CLEARSTATE_GFX11_H_ */ diff --git a/drivers/gpu/drm/amd/amdgpu/dce_v10_0.c b/drivers/gpu/drm/amd/amdgpu/dce_v10_0.c index d1570a462a51..288fce7dc0ed 100644 --- a/drivers/gpu/drm/amd/amdgpu/dce_v10_0.c +++ b/drivers/gpu/drm/amd/amdgpu/dce_v10_0.c @@ -2532,7 +2532,7 @@ static void dce_v10_0_crtc_dpms(struct drm_crtc *crtc, int mode) break; } /* adjust pm to dpms */ - amdgpu_pm_compute_clocks(adev); + amdgpu_dpm_compute_clocks(adev); } static void dce_v10_0_crtc_prepare(struct drm_crtc *crtc) @@ -2798,6 +2798,8 @@ static int dce_v10_0_sw_init(void *handle) adev_to_drm(adev)->mode_config.preferred_depth = 24; adev_to_drm(adev)->mode_config.prefer_shadow = 1; + adev_to_drm(adev)->mode_config.fb_modifiers_not_supported = true; + adev_to_drm(adev)->mode_config.fb_base = adev->gmc.aper_base; r = amdgpu_display_modeset_create_props(adev); diff --git a/drivers/gpu/drm/amd/amdgpu/dce_v11_0.c b/drivers/gpu/drm/amd/amdgpu/dce_v11_0.c index 18a7b3bd633b..cbe5250b31cb 100644 --- a/drivers/gpu/drm/amd/amdgpu/dce_v11_0.c +++ b/drivers/gpu/drm/amd/amdgpu/dce_v11_0.c @@ -2608,7 +2608,7 @@ static void dce_v11_0_crtc_dpms(struct drm_crtc *crtc, int mode) break; } /* adjust pm to dpms */ - amdgpu_pm_compute_clocks(adev); + amdgpu_dpm_compute_clocks(adev); } static void dce_v11_0_crtc_prepare(struct drm_crtc *crtc) @@ -2916,6 +2916,8 @@ static int dce_v11_0_sw_init(void *handle) adev_to_drm(adev)->mode_config.preferred_depth = 24; adev_to_drm(adev)->mode_config.prefer_shadow = 1; + adev_to_drm(adev)->mode_config.fb_modifiers_not_supported = true; + adev_to_drm(adev)->mode_config.fb_base = adev->gmc.aper_base; r = amdgpu_display_modeset_create_props(adev); diff --git a/drivers/gpu/drm/amd/amdgpu/dce_v6_0.c b/drivers/gpu/drm/amd/amdgpu/dce_v6_0.c index c7803dc2b2d5..b1c44fab074f 100644 --- a/drivers/gpu/drm/amd/amdgpu/dce_v6_0.c +++ b/drivers/gpu/drm/amd/amdgpu/dce_v6_0.c @@ -339,7 +339,7 @@ static void dce_v6_0_hpd_fini(struct amdgpu_device *adev) tmp = RREG32(mmDC_HPD1_CONTROL + hpd_offsets[amdgpu_connector->hpd.hpd]); tmp &= ~DC_HPD1_CONTROL__DC_HPD1_EN_MASK; - WREG32(mmDC_HPD1_CONTROL + hpd_offsets[amdgpu_connector->hpd.hpd], 0); + WREG32(mmDC_HPD1_CONTROL + hpd_offsets[amdgpu_connector->hpd.hpd], tmp); amdgpu_irq_put(adev, &adev->hpd_irq, amdgpu_connector->hpd.hpd); } @@ -2424,7 +2424,7 @@ static void dce_v6_0_crtc_dpms(struct drm_crtc *crtc, int mode) break; } /* adjust pm to dpms */ - amdgpu_pm_compute_clocks(adev); + amdgpu_dpm_compute_clocks(adev); } static void dce_v6_0_crtc_prepare(struct drm_crtc *crtc) @@ -2674,6 +2674,7 @@ static int dce_v6_0_sw_init(void *handle) adev_to_drm(adev)->mode_config.max_height = 16384; adev_to_drm(adev)->mode_config.preferred_depth = 24; adev_to_drm(adev)->mode_config.prefer_shadow = 1; + adev_to_drm(adev)->mode_config.fb_modifiers_not_supported = true; adev_to_drm(adev)->mode_config.fb_base = adev->gmc.aper_base; r = amdgpu_display_modeset_create_props(adev); diff --git a/drivers/gpu/drm/amd/amdgpu/dce_v8_0.c b/drivers/gpu/drm/amd/amdgpu/dce_v8_0.c index 8318ee8339f1..a22b45c92792 100644 --- a/drivers/gpu/drm/amd/amdgpu/dce_v8_0.c +++ b/drivers/gpu/drm/amd/amdgpu/dce_v8_0.c @@ -333,7 +333,7 @@ static void dce_v8_0_hpd_fini(struct amdgpu_device *adev) tmp = RREG32(mmDC_HPD1_CONTROL + hpd_offsets[amdgpu_connector->hpd.hpd]); tmp &= ~DC_HPD1_CONTROL__DC_HPD1_EN_MASK; - WREG32(mmDC_HPD1_CONTROL + hpd_offsets[amdgpu_connector->hpd.hpd], 0); + WREG32(mmDC_HPD1_CONTROL + hpd_offsets[amdgpu_connector->hpd.hpd], tmp); amdgpu_irq_put(adev, &adev->hpd_irq, amdgpu_connector->hpd.hpd); } @@ -2433,7 +2433,7 @@ static void dce_v8_0_crtc_dpms(struct drm_crtc *crtc, int mode) break; } /* adjust pm to dpms */ - amdgpu_pm_compute_clocks(adev); + amdgpu_dpm_compute_clocks(adev); } static void dce_v8_0_crtc_prepare(struct drm_crtc *crtc) @@ -2693,7 +2693,13 @@ static int dce_v8_0_sw_init(void *handle) adev_to_drm(adev)->mode_config.max_height = 16384; adev_to_drm(adev)->mode_config.preferred_depth = 24; - adev_to_drm(adev)->mode_config.prefer_shadow = 1; + if (adev->asic_type == CHIP_HAWAII) + /* disable prefer shadow for now due to hibernation issues */ + adev_to_drm(adev)->mode_config.prefer_shadow = 0; + else + adev_to_drm(adev)->mode_config.prefer_shadow = 1; + + adev_to_drm(adev)->mode_config.fb_modifiers_not_supported = true; adev_to_drm(adev)->mode_config.fb_base = adev->gmc.aper_base; diff --git a/drivers/gpu/drm/amd/amdgpu/df_v1_7.c b/drivers/gpu/drm/amd/amdgpu/df_v1_7.c index 2d01ac0d4c11..b991609f46c1 100644 --- a/drivers/gpu/drm/amd/amdgpu/df_v1_7.c +++ b/drivers/gpu/drm/amd/amdgpu/df_v1_7.c @@ -99,7 +99,7 @@ static void df_v1_7_update_medium_grain_clock_gating(struct amdgpu_device *adev, } static void df_v1_7_get_clockgating_state(struct amdgpu_device *adev, - u32 *flags) + u64 *flags) { u32 tmp; diff --git a/drivers/gpu/drm/amd/amdgpu/df_v3_6.c b/drivers/gpu/drm/amd/amdgpu/df_v3_6.c index 43c5e3ec9a39..483a441b46aa 100644 --- a/drivers/gpu/drm/amd/amdgpu/df_v3_6.c +++ b/drivers/gpu/drm/amd/amdgpu/df_v3_6.c @@ -332,7 +332,7 @@ static void df_v3_6_update_medium_grain_clock_gating(struct amdgpu_device *adev, } static void df_v3_6_get_clockgating_state(struct amdgpu_device *adev, - u32 *flags) + u64 *flags) { u32 tmp; @@ -458,7 +458,7 @@ static int df_v3_6_pmc_add_cntr(struct amdgpu_device *adev, #define DEFERRED_ARM_MASK (1 << 31) static int df_v3_6_pmc_set_deferred(struct amdgpu_device *adev, - int counter_idx, uint64_t config, + uint64_t config, int counter_idx, bool is_deferred) { @@ -476,8 +476,8 @@ static int df_v3_6_pmc_set_deferred(struct amdgpu_device *adev, } static bool df_v3_6_pmc_is_deferred(struct amdgpu_device *adev, - int counter_idx, - uint64_t config) + uint64_t config, + int counter_idx) { return (df_v3_6_pmc_has_counter(adev, config, counter_idx) && (adev->df_perfmon_config_assign_mask[counter_idx] diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c index dbe7442fb25c..af94ac580d3e 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c @@ -53,13 +53,9 @@ * 2. Async ring */ #define GFX10_NUM_GFX_RINGS_NV1X 1 -#define GFX10_NUM_GFX_RINGS_Sienna_Cichlid 1 +#define GFX10_NUM_GFX_RINGS_Sienna_Cichlid 2 #define GFX10_MEC_HPD_SIZE 2048 -#define RLCG_VFGATE_DISABLED 0x4000000 -#define RLCG_WRONG_OPERATION_TYPE 0x2000000 -#define RLCG_NOT_IN_RANGE 0x1000000 - #define F32_CE_PROGRAM_RAM_SIZE 65536 #define RLCG_UCODE_LOADING_START_ADDRESS 0x00002000L @@ -110,6 +106,12 @@ #define mmGOLDEN_TSC_COUNT_UPPER_Vangogh_BASE_IDX 1 #define mmGOLDEN_TSC_COUNT_LOWER_Vangogh 0x0026 #define mmGOLDEN_TSC_COUNT_LOWER_Vangogh_BASE_IDX 1 + +#define mmGOLDEN_TSC_COUNT_UPPER_GC_10_3_6 0x002d +#define mmGOLDEN_TSC_COUNT_UPPER_GC_10_3_6_BASE_IDX 1 +#define mmGOLDEN_TSC_COUNT_LOWER_GC_10_3_6 0x002e +#define mmGOLDEN_TSC_COUNT_LOWER_GC_10_3_6_BASE_IDX 1 + #define mmSPI_CONFIG_CNTL_1_Vangogh 0x2441 #define mmSPI_CONFIG_CNTL_1_Vangogh_BASE_IDX 1 #define mmVGT_TF_MEMORY_BASE_HI_Vangogh 0x2261 @@ -180,14 +182,6 @@ #define mmRLC_SPARE_INT_0_Sienna_Cichlid 0x4ca5 #define mmRLC_SPARE_INT_0_Sienna_Cichlid_BASE_IDX 1 -#define GFX_RLCG_GC_WRITE_OLD (0x8 << 28) -#define GFX_RLCG_GC_WRITE (0x0 << 28) -#define GFX_RLCG_GC_READ (0x1 << 28) -#define GFX_RLCG_MMHUB_WRITE (0x2 << 28) - -#define RLCG_ERROR_REPORT_ENABLED(adev) \ - (amdgpu_sriov_reg_indirect_mmhub(adev) || amdgpu_sriov_reg_indirect_gc(adev)) - MODULE_FIRMWARE("amdgpu/navi10_ce.bin"); MODULE_FIRMWARE("amdgpu/navi10_pfp.bin"); MODULE_FIRMWARE("amdgpu/navi10_me.bin"); @@ -256,13 +250,6 @@ MODULE_FIRMWARE("amdgpu/yellow_carp_mec.bin"); MODULE_FIRMWARE("amdgpu/yellow_carp_mec2.bin"); MODULE_FIRMWARE("amdgpu/yellow_carp_rlc.bin"); -MODULE_FIRMWARE("amdgpu/cyan_skillfish_ce.bin"); -MODULE_FIRMWARE("amdgpu/cyan_skillfish_pfp.bin"); -MODULE_FIRMWARE("amdgpu/cyan_skillfish_me.bin"); -MODULE_FIRMWARE("amdgpu/cyan_skillfish_mec.bin"); -MODULE_FIRMWARE("amdgpu/cyan_skillfish_mec2.bin"); -MODULE_FIRMWARE("amdgpu/cyan_skillfish_rlc.bin"); - MODULE_FIRMWARE("amdgpu/cyan_skillfish2_ce.bin"); MODULE_FIRMWARE("amdgpu/cyan_skillfish2_pfp.bin"); MODULE_FIRMWARE("amdgpu/cyan_skillfish2_me.bin"); @@ -270,6 +257,20 @@ MODULE_FIRMWARE("amdgpu/cyan_skillfish2_mec.bin"); MODULE_FIRMWARE("amdgpu/cyan_skillfish2_mec2.bin"); MODULE_FIRMWARE("amdgpu/cyan_skillfish2_rlc.bin"); +MODULE_FIRMWARE("amdgpu/gc_10_3_6_ce.bin"); +MODULE_FIRMWARE("amdgpu/gc_10_3_6_pfp.bin"); +MODULE_FIRMWARE("amdgpu/gc_10_3_6_me.bin"); +MODULE_FIRMWARE("amdgpu/gc_10_3_6_mec.bin"); +MODULE_FIRMWARE("amdgpu/gc_10_3_6_mec2.bin"); +MODULE_FIRMWARE("amdgpu/gc_10_3_6_rlc.bin"); + +MODULE_FIRMWARE("amdgpu/gc_10_3_7_ce.bin"); +MODULE_FIRMWARE("amdgpu/gc_10_3_7_pfp.bin"); +MODULE_FIRMWARE("amdgpu/gc_10_3_7_me.bin"); +MODULE_FIRMWARE("amdgpu/gc_10_3_7_mec.bin"); +MODULE_FIRMWARE("amdgpu/gc_10_3_7_mec2.bin"); +MODULE_FIRMWARE("amdgpu/gc_10_3_7_rlc.bin"); + static const struct soc15_reg_golden golden_settings_gc_10_1[] = { SOC15_REG_GOLDEN_VALUE(GC, 0, mmCB_HW_CONTROL_4, 0xffffffff, 0x00400014), @@ -1463,143 +1464,6 @@ static const struct soc15_reg_golden golden_settings_gc_10_1_2[] = SOC15_REG_GOLDEN_VALUE(GC, 0, mmUTCL1_CTRL, 0xffffffff, 0x00c00000) }; -static bool gfx_v10_get_rlcg_flag(struct amdgpu_device *adev, u32 acc_flags, u32 hwip, - int write, u32 *rlcg_flag) -{ - switch (hwip) { - case GC_HWIP: - if (amdgpu_sriov_reg_indirect_gc(adev)) { - *rlcg_flag = write ? GFX_RLCG_GC_WRITE : GFX_RLCG_GC_READ; - - return true; - /* only in new version, AMDGPU_REGS_NO_KIQ and AMDGPU_REGS_RLC enabled simultaneously */ - } else if ((acc_flags & AMDGPU_REGS_RLC) && !(acc_flags & AMDGPU_REGS_NO_KIQ)) { - *rlcg_flag = GFX_RLCG_GC_WRITE_OLD; - - return true; - } - - break; - case MMHUB_HWIP: - if (amdgpu_sriov_reg_indirect_mmhub(adev) && - (acc_flags & AMDGPU_REGS_RLC) && write) { - *rlcg_flag = GFX_RLCG_MMHUB_WRITE; - return true; - } - - break; - default: - DRM_DEBUG("Not program register by RLCG\n"); - } - - return false; -} - -static u32 gfx_v10_rlcg_rw(struct amdgpu_device *adev, u32 offset, u32 v, uint32_t flag) -{ - static void *scratch_reg0; - static void *scratch_reg1; - static void *scratch_reg2; - static void *scratch_reg3; - static void *spare_int; - static uint32_t grbm_cntl; - static uint32_t grbm_idx; - uint32_t i = 0; - uint32_t retries = 50000; - u32 ret = 0; - u32 tmp; - - scratch_reg0 = adev->rmmio + - (adev->reg_offset[GC_HWIP][0][mmSCRATCH_REG0_BASE_IDX] + mmSCRATCH_REG0) * 4; - scratch_reg1 = adev->rmmio + - (adev->reg_offset[GC_HWIP][0][mmSCRATCH_REG1_BASE_IDX] + mmSCRATCH_REG1) * 4; - scratch_reg2 = adev->rmmio + - (adev->reg_offset[GC_HWIP][0][mmSCRATCH_REG0_BASE_IDX] + mmSCRATCH_REG2) * 4; - scratch_reg3 = adev->rmmio + - (adev->reg_offset[GC_HWIP][0][mmSCRATCH_REG1_BASE_IDX] + mmSCRATCH_REG3) * 4; - - if (adev->ip_versions[GC_HWIP][0] >= IP_VERSION(10, 3, 0)) { - spare_int = adev->rmmio + - (adev->reg_offset[GC_HWIP][0][mmRLC_SPARE_INT_0_Sienna_Cichlid_BASE_IDX] - + mmRLC_SPARE_INT_0_Sienna_Cichlid) * 4; - } else { - spare_int = adev->rmmio + - (adev->reg_offset[GC_HWIP][0][mmRLC_SPARE_INT_BASE_IDX] + mmRLC_SPARE_INT) * 4; - } - - grbm_cntl = adev->reg_offset[GC_HWIP][0][mmGRBM_GFX_CNTL_BASE_IDX] + mmGRBM_GFX_CNTL; - grbm_idx = adev->reg_offset[GC_HWIP][0][mmGRBM_GFX_INDEX_BASE_IDX] + mmGRBM_GFX_INDEX; - - if (offset == grbm_cntl || offset == grbm_idx) { - if (offset == grbm_cntl) - writel(v, scratch_reg2); - else if (offset == grbm_idx) - writel(v, scratch_reg3); - - writel(v, ((void __iomem *)adev->rmmio) + (offset * 4)); - } else { - writel(v, scratch_reg0); - writel(offset | flag, scratch_reg1); - writel(1, spare_int); - - for (i = 0; i < retries; i++) { - tmp = readl(scratch_reg1); - if (!(tmp & flag)) - break; - - udelay(10); - } - - if (i >= retries) { - if (RLCG_ERROR_REPORT_ENABLED(adev)) { - if (tmp & RLCG_VFGATE_DISABLED) - pr_err("The vfgate is disabled, program reg:0x%05x failed!\n", offset); - else if (tmp & RLCG_WRONG_OPERATION_TYPE) - pr_err("Wrong operation type, program reg:0x%05x failed!\n", offset); - else if (tmp & RLCG_NOT_IN_RANGE) - pr_err("The register is not in range, program reg:0x%05x failed!\n", offset); - else - pr_err("Unknown error type, program reg:0x%05x failed!\n", offset); - } else - pr_err("timeout: rlcg program reg:0x%05x failed!\n", offset); - } - } - - ret = readl(scratch_reg0); - - return ret; -} - -static void gfx_v10_sriov_wreg(struct amdgpu_device *adev, u32 offset, u32 value, u32 acc_flags, u32 hwip) -{ - u32 rlcg_flag; - - if (!amdgpu_sriov_runtime(adev) && - gfx_v10_get_rlcg_flag(adev, acc_flags, hwip, 1, &rlcg_flag)) { - gfx_v10_rlcg_rw(adev, offset, value, rlcg_flag); - return; - } - - if (acc_flags & AMDGPU_REGS_NO_KIQ) - WREG32_NO_KIQ(offset, value); - else - WREG32(offset, value); -} - -static u32 gfx_v10_sriov_rreg(struct amdgpu_device *adev, u32 offset, u32 acc_flags, u32 hwip) -{ - u32 rlcg_flag; - - if (!amdgpu_sriov_runtime(adev) && - gfx_v10_get_rlcg_flag(adev, acc_flags, hwip, 0, &rlcg_flag)) - return gfx_v10_rlcg_rw(adev, offset, 0, rlcg_flag); - - if (acc_flags & AMDGPU_REGS_NO_KIQ) - return RREG32_NO_KIQ(offset); - else - return RREG32(offset); -} - static const struct soc15_reg_golden golden_settings_gc_10_1_nv14[] = { /* Pending on emulation bring up */ @@ -3429,7 +3293,7 @@ static const struct soc15_reg_golden golden_settings_gc_10_3_3[] = SOC15_REG_GOLDEN_VALUE(GC, 0, mmDB_DEBUG3, 0xffffffff, 0x00000280), SOC15_REG_GOLDEN_VALUE(GC, 0, mmDB_DEBUG4, 0xffffffff, 0x00800000), SOC15_REG_GOLDEN_VALUE(GC, 0, mmGB_ADDR_CONFIG, 0x0c1807ff, 0x00000242), - SOC15_REG_GOLDEN_VALUE(GC, 0, mmGCR_GENERAL_CNTL, 0x1ff1ffff, 0x00000500), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmGCR_GENERAL_CNTL_Vangogh, 0x1ff1ffff, 0x00000500), SOC15_REG_GOLDEN_VALUE(GC, 0, mmGL1_PIPE_STEER, 0x000000ff, 0x000000e4), SOC15_REG_GOLDEN_VALUE(GC, 0, mmGL2_PIPE_STEER_0, 0x77777777, 0x32103210), SOC15_REG_GOLDEN_VALUE(GC, 0, mmGL2_PIPE_STEER_1, 0x77777777, 0x32103210), @@ -3557,6 +3421,57 @@ static const struct soc15_reg_golden golden_settings_gc_10_0_cyan_skillfish[] = SOC15_REG_GOLDEN_VALUE(GC, 0, mmUTCL1_CTRL, 0x00800000, 0x00800000) }; +static const struct soc15_reg_golden golden_settings_gc_10_3_6[] = +{ + SOC15_REG_GOLDEN_VALUE(GC, 0, mmCGTT_SPI_CS_CLK_CTRL, 0xff7f0fff, 0x78000100), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmCH_PIPE_STEER, 0x000000ff, 0x00000044), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmCPF_GCR_CNTL, 0x0007ffff, 0x0000c200), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmDB_DEBUG3, 0xffffffff, 0x00000280), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmDB_DEBUG4, 0xffffffff, 0x00800000), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmGB_ADDR_CONFIG, 0x0c1807ff, 0x00000042), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmGCR_GENERAL_CNTL_Vangogh, 0x1ff1ffff, 0x00000500), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmGL1_PIPE_STEER, 0x000000ff, 0x00000044), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmGL2_PIPE_STEER_0, 0x77777777, 0x32103210), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmGL2_PIPE_STEER_1, 0x77777777, 0x32103210), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmGL2A_ADDR_MATCH_MASK, 0xffffffff, 0xfffffff3), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmGL2C_ADDR_MATCH_MASK, 0xffffffff, 0xfffffff3), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmGL2C_CM_CTRL1, 0xff8fff0f, 0x580f1008), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmGL2C_CTRL3, 0xf7ffffff, 0x00f80988), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmLDS_CONFIG, 0x000001ff, 0x00000020), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmPA_CL_ENHANCE, 0xf17fffff, 0x01200007), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmPA_SC_BINNER_TIMEOUT_COUNTER, 0xffffffff, 0x00000800), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmPA_SC_ENHANCE_2, 0xffffffbf, 0x00000820), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmSQG_CONFIG, 0x000017ff, 0x00001000), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmSX_DEBUG_1, 0xffffff7f, 0x00010020), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmTA_CNTL_AUX, 0xfff7ffff, 0x01030000), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmUTCL1_CTRL, 0xffffffff, 0x00100000) +}; + +static const struct soc15_reg_golden golden_settings_gc_10_3_7[] = { + SOC15_REG_GOLDEN_VALUE(GC, 0, mmCGTT_SPI_CS_CLK_CTRL, 0xff7f0fff, 0x78000100), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmCH_PIPE_STEER, 0x000000ff, 0x000000e4), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmCPF_GCR_CNTL, 0x0007ffff, 0x0000c200), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmDB_DEBUG3, 0xffffffff, 0x00000280), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmDB_DEBUG4, 0xffffffff, 0x00800000), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmGB_ADDR_CONFIG, 0x0c1807ff, 0x00000041), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmGCR_GENERAL_CNTL_Vangogh, 0x1ff1ffff, 0x00000500), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmGL1_PIPE_STEER, 0x000000ff, 0x000000e4), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmGL2_PIPE_STEER_0, 0x77777777, 0x32103210), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmGL2_PIPE_STEER_1, 0x77777777, 0x32103210), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmGL2A_ADDR_MATCH_MASK, 0xffffffff, 0xffffffff), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmGL2C_ADDR_MATCH_MASK, 0xffffffff, 0xffffffff), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmGL2C_CM_CTRL1, 0xff8fff0f, 0x580f1008), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmGL2C_CTRL3, 0xf7ffffff, 0x00f80988), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmLDS_CONFIG, 0x000001ff, 0x00000020), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmPA_CL_ENHANCE, 0xf000003f, 0x01200007), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmPA_SC_BINNER_TIMEOUT_COUNTER, 0xffffffff, 0x00000800), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmPA_SC_ENHANCE_2, 0xffffffbf, 0x00000820), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmSQG_CONFIG, 0x000017ff, 0x00001000), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmSX_DEBUG_1, 0xffffff7f, 0x00010020), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmTA_CNTL_AUX, 0xfff7ffff, 0x01030000), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmUTCL1_CTRL, 0xffffffff, 0x00100000) +}; + #define DEFAULT_SH_MEM_CONFIG \ ((SH_MEM_ADDRESS_MODE_64 << SH_MEM_CONFIG__ADDRESS_MODE__SHIFT) | \ (SH_MEM_ALIGNMENT_MODE_UNALIGNED << SH_MEM_CONFIG__ALIGNMENT_MODE__SHIFT) | \ @@ -3570,6 +3485,7 @@ static void gfx_v10_0_set_ring_funcs(struct amdgpu_device *adev); static void gfx_v10_0_set_irq_funcs(struct amdgpu_device *adev); static void gfx_v10_0_set_gds_init(struct amdgpu_device *adev); static void gfx_v10_0_set_rlc_funcs(struct amdgpu_device *adev); +static void gfx_v10_0_set_mqd_funcs(struct amdgpu_device *adev); static int gfx_v10_0_get_cu_info(struct amdgpu_device *adev, struct amdgpu_cu_info *cu_info); static uint64_t gfx_v10_0_get_gpu_clock_counter(struct amdgpu_device *adev); @@ -3587,6 +3503,9 @@ static void gfx_v10_0_ring_emit_frame_cntl(struct amdgpu_ring *ring, bool start, static u32 gfx_v10_3_get_disabled_sa(struct amdgpu_device *adev); static void gfx_v10_3_program_pbb_mode(struct amdgpu_device *adev); static void gfx_v10_3_set_power_brake_sequence(struct amdgpu_device *adev); +static void gfx_v10_0_ring_invalidate_tlbs(struct amdgpu_ring *ring, + uint16_t pasid, uint32_t flush_type, + bool all_hub, uint8_t dst_sel); static void gfx10_kiq_set_resources(struct amdgpu_ring *kiq_ring, uint64_t queue_mask) { @@ -3604,10 +3523,23 @@ static void gfx10_kiq_set_resources(struct amdgpu_ring *kiq_ring, uint64_t queue static void gfx10_kiq_map_queues(struct amdgpu_ring *kiq_ring, struct amdgpu_ring *ring) { - struct amdgpu_device *adev = kiq_ring->adev; uint64_t mqd_addr = amdgpu_bo_gpu_offset(ring->mqd_obj); - uint64_t wptr_addr = adev->wb.gpu_addr + (ring->wptr_offs * 4); - uint32_t eng_sel = ring->funcs->type == AMDGPU_RING_TYPE_GFX ? 4 : 0; + uint64_t wptr_addr = ring->wptr_gpu_addr; + uint32_t eng_sel = 0; + + switch (ring->funcs->type) { + case AMDGPU_RING_TYPE_COMPUTE: + eng_sel = 0; + break; + case AMDGPU_RING_TYPE_GFX: + eng_sel = 4; + break; + case AMDGPU_RING_TYPE_MES: + eng_sel = 5; + break; + default: + WARN_ON(1); + } amdgpu_ring_write(kiq_ring, PACKET3(PACKET3_MAP_QUEUES, 5)); /* Q_sel:0, vmid:0, vidmem: 1, engine:0, num_Q:1*/ @@ -3633,8 +3565,14 @@ static void gfx10_kiq_unmap_queues(struct amdgpu_ring *kiq_ring, enum amdgpu_unmap_queues_action action, u64 gpu_addr, u64 seq) { + struct amdgpu_device *adev = kiq_ring->adev; uint32_t eng_sel = ring->funcs->type == AMDGPU_RING_TYPE_GFX ? 4 : 0; + if (adev->enable_mes && !adev->gfx.kiq.ring.sched.ready) { + amdgpu_mes_unmap_legacy_queue(adev, ring, action, gpu_addr, seq); + return; + } + amdgpu_ring_write(kiq_ring, PACKET3(PACKET3_UNMAP_QUEUES, 4)); amdgpu_ring_write(kiq_ring, /* Q_sel: 0, vmid: 0, engine: 0, num_Q: 1 */ PACKET3_UNMAP_QUEUES_ACTION(action) | @@ -3680,12 +3618,7 @@ static void gfx10_kiq_invalidate_tlbs(struct amdgpu_ring *kiq_ring, uint16_t pasid, uint32_t flush_type, bool all_hub) { - amdgpu_ring_write(kiq_ring, PACKET3(PACKET3_INVALIDATE_TLBS, 0)); - amdgpu_ring_write(kiq_ring, - PACKET3_INVALIDATE_TLBS_DST_SEL(1) | - PACKET3_INVALIDATE_TLBS_ALL_HUB(all_hub) | - PACKET3_INVALIDATE_TLBS_PASID(pasid) | - PACKET3_INVALIDATE_TLBS_FLUSH_TYPE(flush_type)); + gfx_v10_0_ring_invalidate_tlbs(kiq_ring, pasid, flush_type, all_hub, 1); } static const struct kiq_pm4_funcs gfx_v10_0_kiq_pm4_funcs = { @@ -3790,23 +3723,27 @@ static void gfx_v10_0_init_golden_registers(struct amdgpu_device *adev) (const u32)ARRAY_SIZE(golden_settings_gc_10_3_5)); break; case IP_VERSION(10, 1, 3): + case IP_VERSION(10, 1, 4): soc15_program_register_sequence(adev, golden_settings_gc_10_0_cyan_skillfish, (const u32)ARRAY_SIZE(golden_settings_gc_10_0_cyan_skillfish)); break; + case IP_VERSION(10, 3, 6): + soc15_program_register_sequence(adev, + golden_settings_gc_10_3_6, + (const u32)ARRAY_SIZE(golden_settings_gc_10_3_6)); + break; + case IP_VERSION(10, 3, 7): + soc15_program_register_sequence(adev, + golden_settings_gc_10_3_7, + (const u32)ARRAY_SIZE(golden_settings_gc_10_3_7)); + break; default: break; } gfx_v10_0_init_spm_golden_registers(adev); } -static void gfx_v10_0_scratch_init(struct amdgpu_device *adev) -{ - adev->gfx.scratch.num_reg = 8; - adev->gfx.scratch.reg_base = SOC15_REG_OFFSET(GC, 0, mmSCRATCH_REG0); - adev->gfx.scratch.free_mask = (1u << adev->gfx.scratch.num_reg) - 1; -} - static void gfx_v10_0_write_data_to_reg(struct amdgpu_ring *ring, int eng_sel, bool wc, uint32_t reg, uint32_t val) { @@ -3843,29 +3780,22 @@ static void gfx_v10_0_wait_reg_mem(struct amdgpu_ring *ring, int eng_sel, static int gfx_v10_0_ring_test_ring(struct amdgpu_ring *ring) { struct amdgpu_device *adev = ring->adev; - uint32_t scratch; + uint32_t scratch = SOC15_REG_OFFSET(GC, 0, mmSCRATCH_REG0); uint32_t tmp = 0; unsigned i; int r; - r = amdgpu_gfx_scratch_get(adev, &scratch); - if (r) { - DRM_ERROR("amdgpu: cp failed to get scratch reg (%d).\n", r); - return r; - } - WREG32(scratch, 0xCAFEDEAD); - r = amdgpu_ring_alloc(ring, 3); if (r) { DRM_ERROR("amdgpu: cp failed to lock ring %d (%d).\n", ring->idx, r); - amdgpu_gfx_scratch_free(adev, scratch); return r; } amdgpu_ring_write(ring, PACKET3(PACKET3_SET_UCONFIG_REG, 1)); - amdgpu_ring_write(ring, (scratch - PACKET3_SET_UCONFIG_REG_START)); + amdgpu_ring_write(ring, scratch - + PACKET3_SET_UCONFIG_REG_START); amdgpu_ring_write(ring, 0xDEADBEEF); amdgpu_ring_commit(ring); @@ -3882,8 +3812,6 @@ static int gfx_v10_0_ring_test_ring(struct amdgpu_ring *ring) if (i >= adev->usec_timeout) r = -ETIMEDOUT; - amdgpu_gfx_scratch_free(adev, scratch); - return r; } @@ -3894,20 +3822,39 @@ static int gfx_v10_0_ring_test_ib(struct amdgpu_ring *ring, long timeout) struct dma_fence *f = NULL; unsigned index; uint64_t gpu_addr; - uint32_t tmp; + volatile uint32_t *cpu_ptr; long r; - r = amdgpu_device_wb_get(adev, &index); - if (r) - return r; - - gpu_addr = adev->wb.gpu_addr + (index * 4); - adev->wb.wb[index] = cpu_to_le32(0xCAFEDEAD); memset(&ib, 0, sizeof(ib)); - r = amdgpu_ib_get(adev, NULL, 16, - AMDGPU_IB_POOL_DIRECT, &ib); - if (r) - goto err1; + + if (ring->is_mes_queue) { + uint32_t padding, offset; + + offset = amdgpu_mes_ctx_get_offs(ring, AMDGPU_MES_CTX_IB_OFFS); + padding = amdgpu_mes_ctx_get_offs(ring, + AMDGPU_MES_CTX_PADDING_OFFS); + + ib.gpu_addr = amdgpu_mes_ctx_get_offs_gpu_addr(ring, offset); + ib.ptr = amdgpu_mes_ctx_get_offs_cpu_addr(ring, offset); + + gpu_addr = amdgpu_mes_ctx_get_offs_gpu_addr(ring, padding); + cpu_ptr = amdgpu_mes_ctx_get_offs_cpu_addr(ring, padding); + *cpu_ptr = cpu_to_le32(0xCAFEDEAD); + } else { + r = amdgpu_device_wb_get(adev, &index); + if (r) + return r; + + gpu_addr = adev->wb.gpu_addr + (index * 4); + adev->wb.wb[index] = cpu_to_le32(0xCAFEDEAD); + cpu_ptr = &adev->wb.wb[index]; + + r = amdgpu_ib_get(adev, NULL, 20, AMDGPU_IB_POOL_DIRECT, &ib); + if (r) { + DRM_ERROR("amdgpu: failed to get ib (%ld).\n", r); + goto err1; + } + } ib.ptr[0] = PACKET3(PACKET3_WRITE_DATA, 3); ib.ptr[1] = WRITE_DATA_DST_SEL(5) | WR_CONFIRM; @@ -3928,16 +3875,17 @@ static int gfx_v10_0_ring_test_ib(struct amdgpu_ring *ring, long timeout) goto err2; } - tmp = adev->wb.wb[index]; - if (tmp == 0xDEADBEEF) + if (le32_to_cpu(*cpu_ptr) == 0xDEADBEEF) r = 0; else r = -EINVAL; err2: - amdgpu_ib_free(adev, &ib, NULL); + if (!ring->is_mes_queue) + amdgpu_ib_free(adev, &ib, NULL); dma_fence_put(f); err1: - amdgpu_device_wb_free(adev, index); + if (!ring->is_mes_queue) + amdgpu_device_wb_free(adev, index); return r; } @@ -3968,6 +3916,7 @@ static void gfx_v10_0_check_fw_write_wait(struct amdgpu_device *adev) case IP_VERSION(10, 1, 2): case IP_VERSION(10, 1, 1): case IP_VERSION(10, 1, 3): + case IP_VERSION(10, 1, 4): if ((adev->gfx.me_fw_version >= 0x00000046) && (adev->gfx.me_feature_version >= 27) && (adev->gfx.pfp_fw_version >= 0x00000068) && @@ -3981,7 +3930,9 @@ static void gfx_v10_0_check_fw_write_wait(struct amdgpu_device *adev) case IP_VERSION(10, 3, 1): case IP_VERSION(10, 3, 4): case IP_VERSION(10, 3, 5): + case IP_VERSION(10, 3, 6): case IP_VERSION(10, 3, 3): + case IP_VERSION(10, 3, 7): adev->gfx.cp_fw_write_wait = true; break; default: @@ -3992,39 +3943,6 @@ static void gfx_v10_0_check_fw_write_wait(struct amdgpu_device *adev) DRM_WARN_ONCE("CP firmware version too old, please update!"); } - -static void gfx_v10_0_init_rlc_ext_microcode(struct amdgpu_device *adev) -{ - const struct rlc_firmware_header_v2_1 *rlc_hdr; - - rlc_hdr = (const struct rlc_firmware_header_v2_1 *)adev->gfx.rlc_fw->data; - adev->gfx.rlc_srlc_fw_version = le32_to_cpu(rlc_hdr->save_restore_list_cntl_ucode_ver); - adev->gfx.rlc_srlc_feature_version = le32_to_cpu(rlc_hdr->save_restore_list_cntl_feature_ver); - adev->gfx.rlc.save_restore_list_cntl_size_bytes = le32_to_cpu(rlc_hdr->save_restore_list_cntl_size_bytes); - adev->gfx.rlc.save_restore_list_cntl = (u8 *)rlc_hdr + le32_to_cpu(rlc_hdr->save_restore_list_cntl_offset_bytes); - adev->gfx.rlc_srlg_fw_version = le32_to_cpu(rlc_hdr->save_restore_list_gpm_ucode_ver); - adev->gfx.rlc_srlg_feature_version = le32_to_cpu(rlc_hdr->save_restore_list_gpm_feature_ver); - adev->gfx.rlc.save_restore_list_gpm_size_bytes = le32_to_cpu(rlc_hdr->save_restore_list_gpm_size_bytes); - adev->gfx.rlc.save_restore_list_gpm = (u8 *)rlc_hdr + le32_to_cpu(rlc_hdr->save_restore_list_gpm_offset_bytes); - adev->gfx.rlc_srls_fw_version = le32_to_cpu(rlc_hdr->save_restore_list_srm_ucode_ver); - adev->gfx.rlc_srls_feature_version = le32_to_cpu(rlc_hdr->save_restore_list_srm_feature_ver); - adev->gfx.rlc.save_restore_list_srm_size_bytes = le32_to_cpu(rlc_hdr->save_restore_list_srm_size_bytes); - adev->gfx.rlc.save_restore_list_srm = (u8 *)rlc_hdr + le32_to_cpu(rlc_hdr->save_restore_list_srm_offset_bytes); - adev->gfx.rlc.reg_list_format_direct_reg_list_length = - le32_to_cpu(rlc_hdr->reg_list_format_direct_reg_list_length); -} - -static void gfx_v10_0_init_rlc_iram_dram_microcode(struct amdgpu_device *adev) -{ - const struct rlc_firmware_header_v2_2 *rlc_hdr; - - rlc_hdr = (const struct rlc_firmware_header_v2_2 *)adev->gfx.rlc_fw->data; - adev->gfx.rlc.rlc_iram_ucode_size_bytes = le32_to_cpu(rlc_hdr->rlc_iram_ucode_size_bytes); - adev->gfx.rlc.rlc_iram_ucode = (u8 *)rlc_hdr + le32_to_cpu(rlc_hdr->rlc_iram_ucode_offset_bytes); - adev->gfx.rlc.rlc_dram_ucode_size_bytes = le32_to_cpu(rlc_hdr->rlc_dram_ucode_size_bytes); - adev->gfx.rlc.rlc_dram_ucode = (u8 *)rlc_hdr + le32_to_cpu(rlc_hdr->rlc_dram_ucode_offset_bytes); -} - static bool gfx_v10_0_navi10_gfxoff_should_enable(struct amdgpu_device *adev) { bool ret = false; @@ -4060,12 +3978,7 @@ static int gfx_v10_0_init_microcode(struct amdgpu_device *adev) char fw_name[40]; char *wks = ""; int err; - struct amdgpu_firmware_info *info = NULL; - const struct common_firmware_header *header = NULL; - const struct gfx_firmware_header_v1_0 *cp_hdr; const struct rlc_firmware_header_v2_0 *rlc_hdr; - unsigned int *tmp = NULL; - unsigned int i = 0; uint16_t version_major; uint16_t version_minor; @@ -4102,11 +4015,15 @@ static int gfx_v10_0_init_microcode(struct amdgpu_device *adev) case IP_VERSION(10, 3, 3): chip_name = "yellow_carp"; break; + case IP_VERSION(10, 3, 6): + chip_name = "gc_10_3_6"; + break; case IP_VERSION(10, 1, 3): - if (adev->apu_flags & AMD_APU_IS_CYAN_SKILLFISH2) - chip_name = "cyan_skillfish2"; - else - chip_name = "cyan_skillfish"; + case IP_VERSION(10, 1, 4): + chip_name = "cyan_skillfish2"; + break; + case IP_VERSION(10, 3, 7): + chip_name = "gc_10_3_7"; break; default: BUG(); @@ -4119,9 +4036,7 @@ static int gfx_v10_0_init_microcode(struct amdgpu_device *adev) err = amdgpu_ucode_validate(adev->gfx.pfp_fw); if (err) goto out; - cp_hdr = (const struct gfx_firmware_header_v1_0 *)adev->gfx.pfp_fw->data; - adev->gfx.pfp_fw_version = le32_to_cpu(cp_hdr->header.ucode_version); - adev->gfx.pfp_feature_version = le32_to_cpu(cp_hdr->ucode_feature_version); + amdgpu_gfx_cp_init_microcode(adev, AMDGPU_UCODE_ID_CP_PFP); snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_me%s.bin", chip_name, wks); err = request_firmware(&adev->gfx.me_fw, fw_name, adev->dev); @@ -4130,9 +4045,7 @@ static int gfx_v10_0_init_microcode(struct amdgpu_device *adev) err = amdgpu_ucode_validate(adev->gfx.me_fw); if (err) goto out; - cp_hdr = (const struct gfx_firmware_header_v1_0 *)adev->gfx.me_fw->data; - adev->gfx.me_fw_version = le32_to_cpu(cp_hdr->header.ucode_version); - adev->gfx.me_feature_version = le32_to_cpu(cp_hdr->ucode_feature_version); + amdgpu_gfx_cp_init_microcode(adev, AMDGPU_UCODE_ID_CP_ME); snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_ce%s.bin", chip_name, wks); err = request_firmware(&adev->gfx.ce_fw, fw_name, adev->dev); @@ -4141,66 +4054,27 @@ static int gfx_v10_0_init_microcode(struct amdgpu_device *adev) err = amdgpu_ucode_validate(adev->gfx.ce_fw); if (err) goto out; - cp_hdr = (const struct gfx_firmware_header_v1_0 *)adev->gfx.ce_fw->data; - adev->gfx.ce_fw_version = le32_to_cpu(cp_hdr->header.ucode_version); - adev->gfx.ce_feature_version = le32_to_cpu(cp_hdr->ucode_feature_version); + amdgpu_gfx_cp_init_microcode(adev, AMDGPU_UCODE_ID_CP_CE); if (!amdgpu_sriov_vf(adev)) { snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_rlc.bin", chip_name); err = request_firmware(&adev->gfx.rlc_fw, fw_name, adev->dev); if (err) goto out; + /* don't check this. There are apparently firmwares in the wild with + * incorrect size in the header + */ err = amdgpu_ucode_validate(adev->gfx.rlc_fw); + if (err) + dev_dbg(adev->dev, + "gfx10: amdgpu_ucode_validate() failed \"%s\"\n", + fw_name); rlc_hdr = (const struct rlc_firmware_header_v2_0 *)adev->gfx.rlc_fw->data; version_major = le16_to_cpu(rlc_hdr->header.header_version_major); version_minor = le16_to_cpu(rlc_hdr->header.header_version_minor); - - adev->gfx.rlc_fw_version = le32_to_cpu(rlc_hdr->header.ucode_version); - adev->gfx.rlc_feature_version = le32_to_cpu(rlc_hdr->ucode_feature_version); - adev->gfx.rlc.save_and_restore_offset = - le32_to_cpu(rlc_hdr->save_and_restore_offset); - adev->gfx.rlc.clear_state_descriptor_offset = - le32_to_cpu(rlc_hdr->clear_state_descriptor_offset); - adev->gfx.rlc.avail_scratch_ram_locations = - le32_to_cpu(rlc_hdr->avail_scratch_ram_locations); - adev->gfx.rlc.reg_restore_list_size = - le32_to_cpu(rlc_hdr->reg_restore_list_size); - adev->gfx.rlc.reg_list_format_start = - le32_to_cpu(rlc_hdr->reg_list_format_start); - adev->gfx.rlc.reg_list_format_separate_start = - le32_to_cpu(rlc_hdr->reg_list_format_separate_start); - adev->gfx.rlc.starting_offsets_start = - le32_to_cpu(rlc_hdr->starting_offsets_start); - adev->gfx.rlc.reg_list_format_size_bytes = - le32_to_cpu(rlc_hdr->reg_list_format_size_bytes); - adev->gfx.rlc.reg_list_size_bytes = - le32_to_cpu(rlc_hdr->reg_list_size_bytes); - adev->gfx.rlc.register_list_format = - kmalloc(adev->gfx.rlc.reg_list_format_size_bytes + - adev->gfx.rlc.reg_list_size_bytes, GFP_KERNEL); - if (!adev->gfx.rlc.register_list_format) { - err = -ENOMEM; + err = amdgpu_gfx_rlc_init_microcode(adev, version_major, version_minor); + if (err) goto out; - } - - tmp = (unsigned int *)((uintptr_t)rlc_hdr + - le32_to_cpu(rlc_hdr->reg_list_format_array_offset_bytes)); - for (i = 0 ; i < (rlc_hdr->reg_list_format_size_bytes >> 2); i++) - adev->gfx.rlc.register_list_format[i] = le32_to_cpu(tmp[i]); - - adev->gfx.rlc.register_restore = adev->gfx.rlc.register_list_format + i; - - tmp = (unsigned int *)((uintptr_t)rlc_hdr + - le32_to_cpu(rlc_hdr->reg_list_array_offset_bytes)); - for (i = 0 ; i < (rlc_hdr->reg_list_size_bytes >> 2); i++) - adev->gfx.rlc.register_restore[i] = le32_to_cpu(tmp[i]); - - if (version_major == 2) { - if (version_minor >= 1) - gfx_v10_0_init_rlc_ext_microcode(adev); - if (version_minor == 2) - gfx_v10_0_init_rlc_iram_dram_microcode(adev); - } } snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_mec%s.bin", chip_name, wks); @@ -4210,9 +4084,8 @@ static int gfx_v10_0_init_microcode(struct amdgpu_device *adev) err = amdgpu_ucode_validate(adev->gfx.mec_fw); if (err) goto out; - cp_hdr = (const struct gfx_firmware_header_v1_0 *)adev->gfx.mec_fw->data; - adev->gfx.mec_fw_version = le32_to_cpu(cp_hdr->header.ucode_version); - adev->gfx.mec_feature_version = le32_to_cpu(cp_hdr->ucode_feature_version); + amdgpu_gfx_cp_init_microcode(adev, AMDGPU_UCODE_ID_CP_MEC1); + amdgpu_gfx_cp_init_microcode(adev, AMDGPU_UCODE_ID_CP_MEC1_JT); snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_mec2%s.bin", chip_name, wks); err = request_firmware(&adev->gfx.mec2_fw, fw_name, adev->dev); @@ -4220,123 +4093,18 @@ static int gfx_v10_0_init_microcode(struct amdgpu_device *adev) err = amdgpu_ucode_validate(adev->gfx.mec2_fw); if (err) goto out; - cp_hdr = (const struct gfx_firmware_header_v1_0 *) - adev->gfx.mec2_fw->data; - adev->gfx.mec2_fw_version = - le32_to_cpu(cp_hdr->header.ucode_version); - adev->gfx.mec2_feature_version = - le32_to_cpu(cp_hdr->ucode_feature_version); + amdgpu_gfx_cp_init_microcode(adev, AMDGPU_UCODE_ID_CP_MEC2); + amdgpu_gfx_cp_init_microcode(adev, AMDGPU_UCODE_ID_CP_MEC2_JT); } else { err = 0; adev->gfx.mec2_fw = NULL; } - if (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) { - info = &adev->firmware.ucode[AMDGPU_UCODE_ID_CP_PFP]; - info->ucode_id = AMDGPU_UCODE_ID_CP_PFP; - info->fw = adev->gfx.pfp_fw; - header = (const struct common_firmware_header *)info->fw->data; - adev->firmware.fw_size += - ALIGN(le32_to_cpu(header->ucode_size_bytes), PAGE_SIZE); - - info = &adev->firmware.ucode[AMDGPU_UCODE_ID_CP_ME]; - info->ucode_id = AMDGPU_UCODE_ID_CP_ME; - info->fw = adev->gfx.me_fw; - header = (const struct common_firmware_header *)info->fw->data; - adev->firmware.fw_size += - ALIGN(le32_to_cpu(header->ucode_size_bytes), PAGE_SIZE); - - info = &adev->firmware.ucode[AMDGPU_UCODE_ID_CP_CE]; - info->ucode_id = AMDGPU_UCODE_ID_CP_CE; - info->fw = adev->gfx.ce_fw; - header = (const struct common_firmware_header *)info->fw->data; - adev->firmware.fw_size += - ALIGN(le32_to_cpu(header->ucode_size_bytes), PAGE_SIZE); - - info = &adev->firmware.ucode[AMDGPU_UCODE_ID_RLC_G]; - info->ucode_id = AMDGPU_UCODE_ID_RLC_G; - info->fw = adev->gfx.rlc_fw; - if (info->fw) { - header = (const struct common_firmware_header *)info->fw->data; - adev->firmware.fw_size += - ALIGN(le32_to_cpu(header->ucode_size_bytes), PAGE_SIZE); - } - if (adev->gfx.rlc.save_restore_list_cntl_size_bytes && - adev->gfx.rlc.save_restore_list_gpm_size_bytes && - adev->gfx.rlc.save_restore_list_srm_size_bytes) { - info = &adev->firmware.ucode[AMDGPU_UCODE_ID_RLC_RESTORE_LIST_CNTL]; - info->ucode_id = AMDGPU_UCODE_ID_RLC_RESTORE_LIST_CNTL; - info->fw = adev->gfx.rlc_fw; - adev->firmware.fw_size += - ALIGN(adev->gfx.rlc.save_restore_list_cntl_size_bytes, PAGE_SIZE); - - info = &adev->firmware.ucode[AMDGPU_UCODE_ID_RLC_RESTORE_LIST_GPM_MEM]; - info->ucode_id = AMDGPU_UCODE_ID_RLC_RESTORE_LIST_GPM_MEM; - info->fw = adev->gfx.rlc_fw; - adev->firmware.fw_size += - ALIGN(adev->gfx.rlc.save_restore_list_gpm_size_bytes, PAGE_SIZE); - - info = &adev->firmware.ucode[AMDGPU_UCODE_ID_RLC_RESTORE_LIST_SRM_MEM]; - info->ucode_id = AMDGPU_UCODE_ID_RLC_RESTORE_LIST_SRM_MEM; - info->fw = adev->gfx.rlc_fw; - adev->firmware.fw_size += - ALIGN(adev->gfx.rlc.save_restore_list_srm_size_bytes, PAGE_SIZE); - - if (adev->gfx.rlc.rlc_iram_ucode_size_bytes && - adev->gfx.rlc.rlc_dram_ucode_size_bytes) { - info = &adev->firmware.ucode[AMDGPU_UCODE_ID_RLC_IRAM]; - info->ucode_id = AMDGPU_UCODE_ID_RLC_IRAM; - info->fw = adev->gfx.rlc_fw; - adev->firmware.fw_size += - ALIGN(adev->gfx.rlc.rlc_iram_ucode_size_bytes, PAGE_SIZE); - - info = &adev->firmware.ucode[AMDGPU_UCODE_ID_RLC_DRAM]; - info->ucode_id = AMDGPU_UCODE_ID_RLC_DRAM; - info->fw = adev->gfx.rlc_fw; - adev->firmware.fw_size += - ALIGN(adev->gfx.rlc.rlc_dram_ucode_size_bytes, PAGE_SIZE); - } - } - - info = &adev->firmware.ucode[AMDGPU_UCODE_ID_CP_MEC1]; - info->ucode_id = AMDGPU_UCODE_ID_CP_MEC1; - info->fw = adev->gfx.mec_fw; - header = (const struct common_firmware_header *)info->fw->data; - cp_hdr = (const struct gfx_firmware_header_v1_0 *)info->fw->data; - adev->firmware.fw_size += - ALIGN(le32_to_cpu(header->ucode_size_bytes) - - le32_to_cpu(cp_hdr->jt_size) * 4, PAGE_SIZE); - - info = &adev->firmware.ucode[AMDGPU_UCODE_ID_CP_MEC1_JT]; - info->ucode_id = AMDGPU_UCODE_ID_CP_MEC1_JT; - info->fw = adev->gfx.mec_fw; - adev->firmware.fw_size += - ALIGN(le32_to_cpu(cp_hdr->jt_size) * 4, PAGE_SIZE); - - if (adev->gfx.mec2_fw) { - info = &adev->firmware.ucode[AMDGPU_UCODE_ID_CP_MEC2]; - info->ucode_id = AMDGPU_UCODE_ID_CP_MEC2; - info->fw = adev->gfx.mec2_fw; - header = (const struct common_firmware_header *)info->fw->data; - cp_hdr = (const struct gfx_firmware_header_v1_0 *)info->fw->data; - adev->firmware.fw_size += - ALIGN(le32_to_cpu(header->ucode_size_bytes) - - le32_to_cpu(cp_hdr->jt_size) * 4, - PAGE_SIZE); - info = &adev->firmware.ucode[AMDGPU_UCODE_ID_CP_MEC2_JT]; - info->ucode_id = AMDGPU_UCODE_ID_CP_MEC2_JT; - info->fw = adev->gfx.mec2_fw; - adev->firmware.fw_size += - ALIGN(le32_to_cpu(cp_hdr->jt_size) * 4, - PAGE_SIZE); - } - } - gfx_v10_0_check_fw_write_wait(adev); out: if (err) { dev_err(adev->dev, - "gfx10: Failed to load firmware \"%s\"\n", + "gfx10: Failed to init firmware \"%s\"\n", fw_name); release_firmware(adev->gfx.pfp_fw); adev->gfx.pfp_fw = NULL; @@ -4448,6 +4216,30 @@ static void gfx_v10_0_rlc_fini(struct amdgpu_device *adev) (void **)&adev->gfx.rlc.cp_table_ptr); } +static void gfx_v10_0_init_rlcg_reg_access_ctrl(struct amdgpu_device *adev) +{ + struct amdgpu_rlcg_reg_access_ctrl *reg_access_ctrl; + + reg_access_ctrl = &adev->gfx.rlc.reg_access_ctrl; + reg_access_ctrl->scratch_reg0 = SOC15_REG_OFFSET(GC, 0, mmSCRATCH_REG0); + reg_access_ctrl->scratch_reg1 = SOC15_REG_OFFSET(GC, 0, mmSCRATCH_REG1); + reg_access_ctrl->scratch_reg2 = SOC15_REG_OFFSET(GC, 0, mmSCRATCH_REG2); + reg_access_ctrl->scratch_reg3 = SOC15_REG_OFFSET(GC, 0, mmSCRATCH_REG3); + reg_access_ctrl->grbm_cntl = SOC15_REG_OFFSET(GC, 0, mmGRBM_GFX_CNTL); + reg_access_ctrl->grbm_idx = SOC15_REG_OFFSET(GC, 0, mmGRBM_GFX_INDEX); + switch (adev->ip_versions[GC_HWIP][0]) { + case IP_VERSION(10, 3, 0): + reg_access_ctrl->spare_int = + SOC15_REG_OFFSET(GC, 0, mmRLC_SPARE_INT_0_Sienna_Cichlid); + break; + default: + reg_access_ctrl->spare_int = + SOC15_REG_OFFSET(GC, 0, mmRLC_SPARE_INT); + break; + } + adev->gfx.rlc.rlcg_reg_access_supported = true; +} + static int gfx_v10_0_rlc_init(struct amdgpu_device *adev) { const struct cs_section_def *cs_data; @@ -4468,6 +4260,7 @@ static int gfx_v10_0_rlc_init(struct amdgpu_device *adev) if (adev->gfx.rlc.funcs->update_spm_vmid) adev->gfx.rlc.funcs->update_spm_vmid(adev, 0xf); + return 0; } @@ -4678,7 +4471,9 @@ static void gfx_v10_0_gpu_early_init(struct amdgpu_device *adev) case IP_VERSION(10, 3, 1): case IP_VERSION(10, 3, 4): case IP_VERSION(10, 3, 5): + case IP_VERSION(10, 3, 6): case IP_VERSION(10, 3, 3): + case IP_VERSION(10, 3, 7): adev->gfx.config.max_hw_contexts = 8; adev->gfx.config.sc_prim_fifo_size_frontend = 0x20; adev->gfx.config.sc_prim_fifo_size_backend = 0x100; @@ -4689,6 +4484,7 @@ static void gfx_v10_0_gpu_early_init(struct amdgpu_device *adev) 1 << REG_GET_FIELD(gb_addr_config, GB_ADDR_CONFIG, NUM_PKRS); break; case IP_VERSION(10, 1, 3): + case IP_VERSION(10, 1, 4): adev->gfx.config.max_hw_contexts = 8; adev->gfx.config.sc_prim_fifo_size_frontend = 0x20; adev->gfx.config.sc_prim_fifo_size_backend = 0x100; @@ -4727,9 +4523,9 @@ static void gfx_v10_0_gpu_early_init(struct amdgpu_device *adev) static int gfx_v10_0_gfx_ring_init(struct amdgpu_device *adev, int ring_id, int me, int pipe, int queue) { - int r; struct amdgpu_ring *ring; unsigned int irq_type; + unsigned int hw_prio; ring = &adev->gfx.gfx_ring[ring_id]; @@ -4747,17 +4543,15 @@ static int gfx_v10_0_gfx_ring_init(struct amdgpu_device *adev, int ring_id, sprintf(ring->name, "gfx_%d.%d.%d", ring->me, ring->pipe, ring->queue); irq_type = AMDGPU_CP_IRQ_GFX_ME0_PIPE0_EOP + ring->pipe; - r = amdgpu_ring_init(adev, ring, 1024, &adev->gfx.eop_irq, irq_type, - AMDGPU_RING_PRIO_DEFAULT, NULL); - if (r) - return r; - return 0; + hw_prio = amdgpu_gfx_is_high_priority_graphics_queue(adev, ring) ? + AMDGPU_GFX_PIPE_PRIO_HIGH : AMDGPU_GFX_PIPE_PRIO_NORMAL; + return amdgpu_ring_init(adev, ring, 1024, &adev->gfx.eop_irq, irq_type, + hw_prio, NULL); } static int gfx_v10_0_compute_ring_init(struct amdgpu_device *adev, int ring_id, int mec, int pipe, int queue) { - int r; unsigned irq_type; struct amdgpu_ring *ring; unsigned int hw_prio; @@ -4780,14 +4574,10 @@ static int gfx_v10_0_compute_ring_init(struct amdgpu_device *adev, int ring_id, + ((ring->me - 1) * adev->gfx.mec.num_pipe_per_mec) + ring->pipe; hw_prio = amdgpu_gfx_is_high_priority_compute_queue(adev, ring) ? - AMDGPU_GFX_PIPE_PRIO_HIGH : AMDGPU_GFX_PIPE_PRIO_NORMAL; + AMDGPU_RING_PRIO_2 : AMDGPU_RING_PRIO_DEFAULT; /* type-2 packets are deprecated on MEC, use type-3 instead */ - r = amdgpu_ring_init(adev, ring, 1024, &adev->gfx.eop_irq, irq_type, + return amdgpu_ring_init(adev, ring, 1024, &adev->gfx.eop_irq, irq_type, hw_prio, NULL); - if (r) - return r; - - return 0; } static int gfx_v10_0_sw_init(void *handle) @@ -4801,6 +4591,7 @@ static int gfx_v10_0_sw_init(void *handle) case IP_VERSION(10, 1, 1): case IP_VERSION(10, 1, 2): case IP_VERSION(10, 1, 3): + case IP_VERSION(10, 1, 4): adev->gfx.me.num_me = 1; adev->gfx.me.num_pipe_per_me = 1; adev->gfx.me.num_queue_per_pipe = 1; @@ -4813,7 +4604,9 @@ static int gfx_v10_0_sw_init(void *handle) case IP_VERSION(10, 3, 1): case IP_VERSION(10, 3, 4): case IP_VERSION(10, 3, 5): + case IP_VERSION(10, 3, 6): case IP_VERSION(10, 3, 3): + case IP_VERSION(10, 3, 7): adev->gfx.me.num_me = 1; adev->gfx.me.num_pipe_per_me = 1; adev->gfx.me.num_queue_per_pipe = 1; @@ -4859,16 +4652,18 @@ static int gfx_v10_0_sw_init(void *handle) adev->gfx.gfx_current_status = AMDGPU_GFX_NORMAL_MODE; - gfx_v10_0_scratch_init(adev); - r = gfx_v10_0_me_init(adev); if (r) return r; - r = gfx_v10_0_rlc_init(adev); - if (r) { - DRM_ERROR("Failed to init rlc BOs!\n"); - return r; + if (adev->gfx.rlc.funcs) { + if (adev->gfx.rlc.funcs->init) { + r = adev->gfx.rlc.funcs->init(adev); + if (r) { + dev_err(adev->dev, "Failed to init rlc BOs!\n"); + return r; + } + } } r = gfx_v10_0_mec_init(adev); @@ -4912,16 +4707,18 @@ static int gfx_v10_0_sw_init(void *handle) } } - r = amdgpu_gfx_kiq_init(adev, GFX10_MEC_HPD_SIZE); - if (r) { - DRM_ERROR("Failed to init KIQ BOs!\n"); - return r; - } + if (!adev->enable_mes_kiq) { + r = amdgpu_gfx_kiq_init(adev, GFX10_MEC_HPD_SIZE); + if (r) { + DRM_ERROR("Failed to init KIQ BOs!\n"); + return r; + } - kiq = &adev->gfx.kiq; - r = amdgpu_gfx_kiq_init_ring(adev, &kiq->ring, &kiq->irq); - if (r) - return r; + kiq = &adev->gfx.kiq; + r = amdgpu_gfx_kiq_init_ring(adev, &kiq->ring, &kiq->irq); + if (r) + return r; + } r = amdgpu_gfx_mqd_sw_init(adev, sizeof(struct v10_compute_mqd)); if (r) @@ -4973,8 +4770,11 @@ static int gfx_v10_0_sw_fini(void *handle) amdgpu_ring_fini(&adev->gfx.compute_ring[i]); amdgpu_gfx_mqd_sw_fini(adev); - amdgpu_gfx_kiq_free_ring(&adev->gfx.kiq.ring); - amdgpu_gfx_kiq_fini(adev); + + if (!adev->enable_mes_kiq) { + amdgpu_gfx_kiq_free_ring(&adev->gfx.kiq.ring); + amdgpu_gfx_kiq_fini(adev); + } gfx_v10_0_pfp_fini(adev); gfx_v10_0_ce_fini(adev); @@ -5047,7 +4847,8 @@ static void gfx_v10_0_setup_rb(struct amdgpu_device *adev) for (j = 0; j < adev->gfx.config.max_sh_per_se; j++) { bitmap = i * adev->gfx.config.max_sh_per_se + j; if (((adev->ip_versions[GC_HWIP][0] == IP_VERSION(10, 3, 0)) || - (adev->ip_versions[GC_HWIP][0] == IP_VERSION(10, 3, 3))) && + (adev->ip_versions[GC_HWIP][0] == IP_VERSION(10, 3, 3)) || + (adev->ip_versions[GC_HWIP][0] == IP_VERSION(10, 3, 6))) && ((gfx_v10_3_get_disabled_sa(adev) >> bitmap) & 1)) continue; gfx_v10_0_select_se_sh(adev, i, j, 0xffffffff); @@ -5127,7 +4928,7 @@ static void gfx_v10_0_init_compute_vmid(struct amdgpu_device *adev) mutex_unlock(&adev->srbm_mutex); /* Initialize all compute VMIDs to have no GDS, GWS, or OA - acccess. These should be enabled by FW for target VMIDs. */ + access. These should be enabled by FW for target VMIDs. */ for (i = adev->vm_manager.first_kfd_vmid; i < AMDGPU_NUM_VMID; i++) { WREG32_SOC15_OFFSET(GC, 0, mmGDS_VMID0_BASE, 2 * i, 0); WREG32_SOC15_OFFSET(GC, 0, mmGDS_VMID0_SIZE, 2 * i, 0); @@ -5932,6 +5733,9 @@ static int gfx_v10_0_cp_gfx_enable(struct amdgpu_device *adev, bool enable) WREG32_SOC15(GC, 0, mmCP_ME_CNTL, tmp); } + if (adev->job_hang && !enable) + return 0; + for (i = 0; i < adev->usec_timeout; i++) { if (RREG32_SOC15(GC, 0, mmCP_STAT) == 0) break; @@ -6321,7 +6125,9 @@ static void gfx_v10_0_cp_gfx_set_doorbell(struct amdgpu_device *adev, case IP_VERSION(10, 3, 1): case IP_VERSION(10, 3, 4): case IP_VERSION(10, 3, 5): + case IP_VERSION(10, 3, 6): case IP_VERSION(10, 3, 3): + case IP_VERSION(10, 3, 7): tmp = REG_SET_FIELD(0, CP_RB_DOORBELL_RANGE_LOWER, DOORBELL_RANGE_LOWER_Sienna_Cichlid, ring->doorbell_index); WREG32_SOC15(GC, 0, mmCP_RB_DOORBELL_RANGE_LOWER, tmp); @@ -6374,12 +6180,12 @@ static int gfx_v10_0_cp_gfx_resume(struct amdgpu_device *adev) WREG32_SOC15(GC, 0, mmCP_RB0_WPTR_HI, upper_32_bits(ring->wptr)); /* set the wb address wether it's enabled or not */ - rptr_addr = adev->wb.gpu_addr + (ring->rptr_offs * 4); + rptr_addr = ring->rptr_gpu_addr; WREG32_SOC15(GC, 0, mmCP_RB0_RPTR_ADDR, lower_32_bits(rptr_addr)); WREG32_SOC15(GC, 0, mmCP_RB0_RPTR_ADDR_HI, upper_32_bits(rptr_addr) & CP_RB_RPTR_ADDR_HI__RB_RPTR_ADDR_HI_MASK); - wptr_gpu_addr = adev->wb.gpu_addr + (ring->wptr_offs * 4); + wptr_gpu_addr = ring->wptr_gpu_addr; WREG32_SOC15(GC, 0, mmCP_RB_WPTR_POLL_ADDR_LO, lower_32_bits(wptr_gpu_addr)); WREG32_SOC15(GC, 0, mmCP_RB_WPTR_POLL_ADDR_HI, @@ -6412,11 +6218,11 @@ static int gfx_v10_0_cp_gfx_resume(struct amdgpu_device *adev) WREG32_SOC15(GC, 0, mmCP_RB1_WPTR, lower_32_bits(ring->wptr)); WREG32_SOC15(GC, 0, mmCP_RB1_WPTR_HI, upper_32_bits(ring->wptr)); /* Set the wb address wether it's enabled or not */ - rptr_addr = adev->wb.gpu_addr + (ring->rptr_offs * 4); + rptr_addr = ring->rptr_gpu_addr; WREG32_SOC15(GC, 0, mmCP_RB1_RPTR_ADDR, lower_32_bits(rptr_addr)); WREG32_SOC15(GC, 0, mmCP_RB1_RPTR_ADDR_HI, upper_32_bits(rptr_addr) & CP_RB1_RPTR_ADDR_HI__RB_RPTR_ADDR_HI_MASK); - wptr_gpu_addr = adev->wb.gpu_addr + (ring->wptr_offs * 4); + wptr_gpu_addr = ring->wptr_gpu_addr; WREG32_SOC15(GC, 0, mmCP_RB_WPTR_POLL_ADDR_LO, lower_32_bits(wptr_gpu_addr)); WREG32_SOC15(GC, 0, mmCP_RB_WPTR_POLL_ADDR_HI, @@ -6458,7 +6264,9 @@ static void gfx_v10_0_cp_compute_enable(struct amdgpu_device *adev, bool enable) case IP_VERSION(10, 3, 1): case IP_VERSION(10, 3, 4): case IP_VERSION(10, 3, 5): + case IP_VERSION(10, 3, 6): case IP_VERSION(10, 3, 3): + case IP_VERSION(10, 3, 7): WREG32_SOC15(GC, 0, mmCP_MEC_CNTL_Sienna_Cichlid, 0); break; default: @@ -6472,7 +6280,9 @@ static void gfx_v10_0_cp_compute_enable(struct amdgpu_device *adev, bool enable) case IP_VERSION(10, 3, 1): case IP_VERSION(10, 3, 4): case IP_VERSION(10, 3, 5): + case IP_VERSION(10, 3, 6): case IP_VERSION(10, 3, 3): + case IP_VERSION(10, 3, 7): WREG32_SOC15(GC, 0, mmCP_MEC_CNTL_Sienna_Cichlid, (CP_MEC_CNTL__MEC_ME1_HALT_MASK | CP_MEC_CNTL__MEC_ME2_HALT_MASK)); @@ -6570,7 +6380,9 @@ static void gfx_v10_0_kiq_setting(struct amdgpu_ring *ring) case IP_VERSION(10, 3, 1): case IP_VERSION(10, 3, 4): case IP_VERSION(10, 3, 5): + case IP_VERSION(10, 3, 6): case IP_VERSION(10, 3, 3): + case IP_VERSION(10, 3, 7): tmp = RREG32_SOC15(GC, 0, mmRLC_CP_SCHEDULERS_Sienna_Cichlid); tmp &= 0xffffff00; tmp |= (ring->me << 5) | (ring->pipe << 3) | (ring->queue); @@ -6589,10 +6401,28 @@ static void gfx_v10_0_kiq_setting(struct amdgpu_ring *ring) } } -static int gfx_v10_0_gfx_mqd_init(struct amdgpu_ring *ring) +static void gfx_v10_0_gfx_mqd_set_priority(struct amdgpu_device *adev, + struct v10_gfx_mqd *mqd, + struct amdgpu_mqd_prop *prop) { - struct amdgpu_device *adev = ring->adev; - struct v10_gfx_mqd *mqd = ring->mqd_ptr; + bool priority = 0; + u32 tmp; + + /* set up default queue priority level + * 0x0 = low priority, 0x1 = high priority + */ + if (prop->hqd_pipe_priority == AMDGPU_GFX_PIPE_PRIO_HIGH) + priority = 1; + + tmp = RREG32_SOC15(GC, 0, mmCP_GFX_HQD_QUEUE_PRIORITY); + tmp = REG_SET_FIELD(tmp, CP_GFX_HQD_QUEUE_PRIORITY, PRIORITY_LEVEL, priority); + mqd->cp_gfx_hqd_queue_priority = tmp; +} + +static int gfx_v10_0_gfx_mqd_init(struct amdgpu_device *adev, void *m, + struct amdgpu_mqd_prop *prop) +{ + struct v10_gfx_mqd *mqd = m; uint64_t hqd_gpu_addr, wb_gpu_addr; uint32_t tmp; uint32_t rb_bufsz; @@ -6602,8 +6432,8 @@ static int gfx_v10_0_gfx_mqd_init(struct amdgpu_ring *ring) mqd->cp_gfx_hqd_wptr_hi = 0; /* set the pointer to the MQD */ - mqd->cp_mqd_base_addr = ring->mqd_gpu_addr & 0xfffffffc; - mqd->cp_mqd_base_addr_hi = upper_32_bits(ring->mqd_gpu_addr); + mqd->cp_mqd_base_addr = prop->mqd_gpu_addr & 0xfffffffc; + mqd->cp_mqd_base_addr_hi = upper_32_bits(prop->mqd_gpu_addr); /* set up mqd control */ tmp = RREG32_SOC15(GC, 0, mmCP_GFX_MQD_CONTROL); @@ -6617,11 +6447,8 @@ static int gfx_v10_0_gfx_mqd_init(struct amdgpu_ring *ring) tmp = REG_SET_FIELD(tmp, CP_GFX_HQD_VMID, VMID, 0); mqd->cp_gfx_hqd_vmid = 0; - /* set up default queue priority level - * 0x0 = low priority, 0x1 = high priority */ - tmp = RREG32_SOC15(GC, 0, mmCP_GFX_HQD_QUEUE_PRIORITY); - tmp = REG_SET_FIELD(tmp, CP_GFX_HQD_QUEUE_PRIORITY, PRIORITY_LEVEL, 0); - mqd->cp_gfx_hqd_queue_priority = tmp; + /* set up gfx queue priority */ + gfx_v10_0_gfx_mqd_set_priority(adev, mqd, prop); /* set up time quantum */ tmp = RREG32_SOC15(GC, 0, mmCP_GFX_HQD_QUANTUM); @@ -6629,23 +6456,23 @@ static int gfx_v10_0_gfx_mqd_init(struct amdgpu_ring *ring) mqd->cp_gfx_hqd_quantum = tmp; /* set up gfx hqd base. this is similar as CP_RB_BASE */ - hqd_gpu_addr = ring->gpu_addr >> 8; + hqd_gpu_addr = prop->hqd_base_gpu_addr >> 8; mqd->cp_gfx_hqd_base = hqd_gpu_addr; mqd->cp_gfx_hqd_base_hi = upper_32_bits(hqd_gpu_addr); /* set up hqd_rptr_addr/_hi, similar as CP_RB_RPTR */ - wb_gpu_addr = adev->wb.gpu_addr + (ring->rptr_offs * 4); + wb_gpu_addr = prop->rptr_gpu_addr; mqd->cp_gfx_hqd_rptr_addr = wb_gpu_addr & 0xfffffffc; mqd->cp_gfx_hqd_rptr_addr_hi = upper_32_bits(wb_gpu_addr) & 0xffff; /* set up rb_wptr_poll addr */ - wb_gpu_addr = adev->wb.gpu_addr + (ring->wptr_offs * 4); + wb_gpu_addr = prop->wptr_gpu_addr; mqd->cp_rb_wptr_poll_addr_lo = wb_gpu_addr & 0xfffffffc; mqd->cp_rb_wptr_poll_addr_hi = upper_32_bits(wb_gpu_addr) & 0xffff; /* set up the gfx_hqd_control, similar as CP_RB0_CNTL */ - rb_bufsz = order_base_2(ring->ring_size / 4) - 1; + rb_bufsz = order_base_2(prop->queue_size / 4) - 1; tmp = RREG32_SOC15(GC, 0, mmCP_GFX_HQD_CNTL); tmp = REG_SET_FIELD(tmp, CP_GFX_HQD_CNTL, RB_BUFSZ, rb_bufsz); tmp = REG_SET_FIELD(tmp, CP_GFX_HQD_CNTL, RB_BLKSZ, rb_bufsz - 2); @@ -6656,9 +6483,9 @@ static int gfx_v10_0_gfx_mqd_init(struct amdgpu_ring *ring) /* set up cp_doorbell_control */ tmp = RREG32_SOC15(GC, 0, mmCP_RB_DOORBELL_CONTROL); - if (ring->use_doorbell) { + if (prop->use_doorbell) { tmp = REG_SET_FIELD(tmp, CP_RB_DOORBELL_CONTROL, - DOORBELL_OFFSET, ring->doorbell_index); + DOORBELL_OFFSET, prop->doorbell_index); tmp = REG_SET_FIELD(tmp, CP_RB_DOORBELL_CONTROL, DOORBELL_EN, 1); } else @@ -6666,13 +6493,7 @@ static int gfx_v10_0_gfx_mqd_init(struct amdgpu_ring *ring) DOORBELL_EN, 0); mqd->cp_rb_doorbell_control = tmp; - /*if there are 2 gfx rings, set the lower doorbell range of the first ring, - *otherwise the range of the second ring will override the first ring */ - if (ring->doorbell_index == adev->doorbell_index.gfx_ring0 << 1) - gfx_v10_0_cp_gfx_set_doorbell(adev, ring); - /* reset read and write pointers, similar to CP_RB0_WPTR/_RPTR */ - ring->wptr = 0; mqd->cp_gfx_hqd_rptr = RREG32_SOC15(GC, 0, mmCP_GFX_HQD_RPTR); /* active the queue */ @@ -6740,7 +6561,16 @@ static int gfx_v10_0_gfx_init_queue(struct amdgpu_ring *ring) memset((void *)mqd, 0, sizeof(*mqd)); mutex_lock(&adev->srbm_mutex); nv_grbm_select(adev, ring->me, ring->pipe, ring->queue, 0); - gfx_v10_0_gfx_mqd_init(ring); + amdgpu_ring_init_mqd(ring); + + /* + * if there are 2 gfx rings, set the lower doorbell + * range of the first ring, otherwise the range of + * the second ring will override the first ring + */ + if (ring->doorbell_index == adev->doorbell_index.gfx_ring0 << 1) + gfx_v10_0_cp_gfx_set_doorbell(adev, ring); + #ifdef BRING_UP_DEBUG gfx_v10_0_gfx_queue_init_register(ring); #endif @@ -6754,7 +6584,7 @@ static int gfx_v10_0_gfx_init_queue(struct amdgpu_ring *ring) memcpy(mqd, adev->gfx.me.mqd_backup[mqd_idx], sizeof(*mqd)); /* reset the ring */ ring->wptr = 0; - adev->wb.wb[ring->wptr_offs] = 0; + *ring->wptr_cpu_addr = 0; amdgpu_ring_clear_ring(ring); #ifdef BRING_UP_DEBUG mutex_lock(&adev->srbm_mutex); @@ -6833,23 +6663,10 @@ done: return r; } -static void gfx_v10_0_compute_mqd_set_priority(struct amdgpu_ring *ring, struct v10_compute_mqd *mqd) +static int gfx_v10_0_compute_mqd_init(struct amdgpu_device *adev, void *m, + struct amdgpu_mqd_prop *prop) { - struct amdgpu_device *adev = ring->adev; - - if (ring->funcs->type == AMDGPU_RING_TYPE_COMPUTE) { - if (amdgpu_gfx_is_high_priority_compute_queue(adev, ring)) { - mqd->cp_hqd_pipe_priority = AMDGPU_GFX_PIPE_PRIO_HIGH; - mqd->cp_hqd_queue_priority = - AMDGPU_GFX_QUEUE_PRIORITY_MAXIMUM; - } - } -} - -static int gfx_v10_0_compute_mqd_init(struct amdgpu_ring *ring) -{ - struct amdgpu_device *adev = ring->adev; - struct v10_compute_mqd *mqd = ring->mqd_ptr; + struct v10_compute_mqd *mqd = m; uint64_t hqd_gpu_addr, wb_gpu_addr, eop_base_addr; uint32_t tmp; @@ -6861,7 +6678,7 @@ static int gfx_v10_0_compute_mqd_init(struct amdgpu_ring *ring) mqd->compute_static_thread_mgmt_se3 = 0xffffffff; mqd->compute_misc_reserved = 0x00000003; - eop_base_addr = ring->eop_gpu_addr >> 8; + eop_base_addr = prop->eop_gpu_addr >> 8; mqd->cp_hqd_eop_base_addr_lo = eop_base_addr; mqd->cp_hqd_eop_base_addr_hi = upper_32_bits(eop_base_addr); @@ -6875,9 +6692,9 @@ static int gfx_v10_0_compute_mqd_init(struct amdgpu_ring *ring) /* enable doorbell? */ tmp = RREG32_SOC15(GC, 0, mmCP_HQD_PQ_DOORBELL_CONTROL); - if (ring->use_doorbell) { + if (prop->use_doorbell) { tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_DOORBELL_CONTROL, - DOORBELL_OFFSET, ring->doorbell_index); + DOORBELL_OFFSET, prop->doorbell_index); tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_DOORBELL_CONTROL, DOORBELL_EN, 1); tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_DOORBELL_CONTROL, @@ -6892,15 +6709,14 @@ static int gfx_v10_0_compute_mqd_init(struct amdgpu_ring *ring) mqd->cp_hqd_pq_doorbell_control = tmp; /* disable the queue if it's active */ - ring->wptr = 0; mqd->cp_hqd_dequeue_request = 0; mqd->cp_hqd_pq_rptr = 0; mqd->cp_hqd_pq_wptr_lo = 0; mqd->cp_hqd_pq_wptr_hi = 0; /* set the pointer to the MQD */ - mqd->cp_mqd_base_addr_lo = ring->mqd_gpu_addr & 0xfffffffc; - mqd->cp_mqd_base_addr_hi = upper_32_bits(ring->mqd_gpu_addr); + mqd->cp_mqd_base_addr_lo = prop->mqd_gpu_addr & 0xfffffffc; + mqd->cp_mqd_base_addr_hi = upper_32_bits(prop->mqd_gpu_addr); /* set MQD vmid to 0 */ tmp = RREG32_SOC15(GC, 0, mmCP_MQD_CONTROL); @@ -6908,16 +6724,16 @@ static int gfx_v10_0_compute_mqd_init(struct amdgpu_ring *ring) mqd->cp_mqd_control = tmp; /* set the pointer to the HQD, this is similar CP_RB0_BASE/_HI */ - hqd_gpu_addr = ring->gpu_addr >> 8; + hqd_gpu_addr = prop->hqd_base_gpu_addr >> 8; mqd->cp_hqd_pq_base_lo = hqd_gpu_addr; mqd->cp_hqd_pq_base_hi = upper_32_bits(hqd_gpu_addr); /* set up the HQD, this is similar to CP_RB0_CNTL */ tmp = RREG32_SOC15(GC, 0, mmCP_HQD_PQ_CONTROL); tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, QUEUE_SIZE, - (order_base_2(ring->ring_size / 4) - 1)); + (order_base_2(prop->queue_size / 4) - 1)); tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, RPTR_BLOCK_SIZE, - ((order_base_2(AMDGPU_GPU_PAGE_SIZE / 4) - 1) << 8)); + (order_base_2(AMDGPU_GPU_PAGE_SIZE / 4) - 1)); #ifdef __BIG_ENDIAN tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, ENDIAN_SWAP, 1); #endif @@ -6928,35 +6744,17 @@ static int gfx_v10_0_compute_mqd_init(struct amdgpu_ring *ring) mqd->cp_hqd_pq_control = tmp; /* set the wb address whether it's enabled or not */ - wb_gpu_addr = adev->wb.gpu_addr + (ring->rptr_offs * 4); + wb_gpu_addr = prop->rptr_gpu_addr; mqd->cp_hqd_pq_rptr_report_addr_lo = wb_gpu_addr & 0xfffffffc; mqd->cp_hqd_pq_rptr_report_addr_hi = upper_32_bits(wb_gpu_addr) & 0xffff; /* only used if CP_PQ_WPTR_POLL_CNTL.CP_PQ_WPTR_POLL_CNTL__EN_MASK=1 */ - wb_gpu_addr = adev->wb.gpu_addr + (ring->wptr_offs * 4); + wb_gpu_addr = prop->wptr_gpu_addr; mqd->cp_hqd_pq_wptr_poll_addr_lo = wb_gpu_addr & 0xfffffffc; mqd->cp_hqd_pq_wptr_poll_addr_hi = upper_32_bits(wb_gpu_addr) & 0xffff; - tmp = 0; - /* enable the doorbell if requested */ - if (ring->use_doorbell) { - tmp = RREG32_SOC15(GC, 0, mmCP_HQD_PQ_DOORBELL_CONTROL); - tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_DOORBELL_CONTROL, - DOORBELL_OFFSET, ring->doorbell_index); - - tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_DOORBELL_CONTROL, - DOORBELL_EN, 1); - tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_DOORBELL_CONTROL, - DOORBELL_SOURCE, 0); - tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_DOORBELL_CONTROL, - DOORBELL_HIT, 0); - } - - mqd->cp_hqd_pq_doorbell_control = tmp; - /* reset read and write pointers, similar to CP_RB0_WPTR/_RPTR */ - ring->wptr = 0; mqd->cp_hqd_pq_rptr = RREG32_SOC15(GC, 0, mmCP_HQD_PQ_RPTR); /* set the vmid for the queue */ @@ -6972,13 +6770,10 @@ static int gfx_v10_0_compute_mqd_init(struct amdgpu_ring *ring) mqd->cp_hqd_ib_control = tmp; /* set static priority for a compute queue/ring */ - gfx_v10_0_compute_mqd_set_priority(ring, mqd); + mqd->cp_hqd_pipe_priority = prop->hqd_pipe_priority; + mqd->cp_hqd_queue_priority = prop->hqd_queue_priority; - /* map_queues packet doesn't need activate the queue, - * so only kiq need set this field. - */ - if (ring->funcs->type == AMDGPU_RING_TYPE_KIQ) - mqd->cp_hqd_active = 1; + mqd->cp_hqd_active = prop->hqd_active; return 0; } @@ -6996,20 +6791,6 @@ static int gfx_v10_0_kiq_init_register(struct amdgpu_ring *ring) /* disable wptr polling */ WREG32_FIELD15(GC, 0, CP_PQ_WPTR_POLL_CNTL, EN, 0); - /* write the EOP addr */ - WREG32_SOC15(GC, 0, mmCP_HQD_EOP_BASE_ADDR, - mqd->cp_hqd_eop_base_addr_lo); - WREG32_SOC15(GC, 0, mmCP_HQD_EOP_BASE_ADDR_HI, - mqd->cp_hqd_eop_base_addr_hi); - - /* set the EOP size, register value is 2^(EOP_SIZE+1) dwords */ - WREG32_SOC15(GC, 0, mmCP_HQD_EOP_CONTROL, - mqd->cp_hqd_eop_control); - - /* enable doorbell? */ - WREG32_SOC15(GC, 0, mmCP_HQD_PQ_DOORBELL_CONTROL, - mqd->cp_hqd_pq_doorbell_control); - /* disable the queue if it's active */ if (RREG32_SOC15(GC, 0, mmCP_HQD_ACTIVE) & 1) { WREG32_SOC15(GC, 0, mmCP_HQD_DEQUEUE_REQUEST, 1); @@ -7028,6 +6809,19 @@ static int gfx_v10_0_kiq_init_register(struct amdgpu_ring *ring) mqd->cp_hqd_pq_wptr_hi); } + /* disable doorbells */ + WREG32_SOC15(GC, 0, mmCP_HQD_PQ_DOORBELL_CONTROL, 0); + + /* write the EOP addr */ + WREG32_SOC15(GC, 0, mmCP_HQD_EOP_BASE_ADDR, + mqd->cp_hqd_eop_base_addr_lo); + WREG32_SOC15(GC, 0, mmCP_HQD_EOP_BASE_ADDR_HI, + mqd->cp_hqd_eop_base_addr_hi); + + /* set the EOP size, register value is 2^(EOP_SIZE+1) dwords */ + WREG32_SOC15(GC, 0, mmCP_HQD_EOP_CONTROL, + mqd->cp_hqd_eop_control); + /* set the pointer to the MQD */ WREG32_SOC15(GC, 0, mmCP_MQD_BASE_ADDR, mqd->cp_mqd_base_addr_lo); @@ -7119,7 +6913,7 @@ static int gfx_v10_0_kiq_init_queue(struct amdgpu_ring *ring) memset((void *)mqd, 0, sizeof(*mqd)); mutex_lock(&adev->srbm_mutex); nv_grbm_select(adev, ring->me, ring->pipe, ring->queue, 0); - gfx_v10_0_compute_mqd_init(ring); + amdgpu_ring_init_mqd(ring); gfx_v10_0_kiq_init_register(ring); nv_grbm_select(adev, 0, 0, 0, 0); mutex_unlock(&adev->srbm_mutex); @@ -7141,7 +6935,7 @@ static int gfx_v10_0_kcq_init_queue(struct amdgpu_ring *ring) memset((void *)mqd, 0, sizeof(*mqd)); mutex_lock(&adev->srbm_mutex); nv_grbm_select(adev, ring->me, ring->pipe, ring->queue, 0); - gfx_v10_0_compute_mqd_init(ring); + amdgpu_ring_init_mqd(ring); nv_grbm_select(adev, 0, 0, 0, 0); mutex_unlock(&adev->srbm_mutex); @@ -7154,7 +6948,7 @@ static int gfx_v10_0_kcq_init_queue(struct amdgpu_ring *ring) /* reset ring buffer */ ring->wptr = 0; - atomic64_set((atomic64_t *)&adev->wb.wb[ring->wptr_offs], 0); + atomic64_set((atomic64_t *)ring->wptr_cpu_addr, 0); amdgpu_ring_clear_ring(ring); } else { amdgpu_ring_clear_ring(ring); @@ -7234,7 +7028,10 @@ static int gfx_v10_0_cp_resume(struct amdgpu_device *adev) return r; } - r = gfx_v10_0_kiq_resume(adev); + if (adev->enable_mes_kiq && adev->mes.kiq_hw_init) + r = amdgpu_mes_kiq_hw_init(adev); + else + r = gfx_v10_0_kiq_resume(adev); if (r) return r; @@ -7300,6 +7097,8 @@ static bool gfx_v10_0_check_grbm_cam_remapping(struct amdgpu_device *adev) break; case IP_VERSION(10, 3, 1): case IP_VERSION(10, 3, 3): + case IP_VERSION(10, 3, 6): + case IP_VERSION(10, 3, 7): return true; default: data = RREG32_SOC15(GC, 0, mmVGT_ESGS_RING_SIZE); @@ -7334,7 +7133,9 @@ static void gfx_v10_0_setup_grbm_cam_remapping(struct amdgpu_device *adev) case IP_VERSION(10, 3, 1): case IP_VERSION(10, 3, 4): case IP_VERSION(10, 3, 5): + case IP_VERSION(10, 3, 6): case IP_VERSION(10, 3, 3): + case IP_VERSION(10, 3, 7): /* mmVGT_TF_RING_SIZE_UMD -> mmVGT_TF_RING_SIZE */ data = (SOC15_REG_OFFSET(GC, 0, mmVGT_TF_RING_SIZE_UMD) << GRBM_CAM_DATA__CAM_ADDR__SHIFT) | @@ -7533,8 +7334,10 @@ static int gfx_v10_0_kiq_disable_kgq(struct amdgpu_device *adev) for (i = 0; i < adev->gfx.num_gfx_rings; i++) kiq->pmf->kiq_unmap_queues(kiq_ring, &adev->gfx.gfx_ring[i], PREEMPT_QUEUES, 0, 0); - - return amdgpu_ring_test_helper(kiq_ring); + if (!adev->job_hang) + return amdgpu_ring_test_helper(kiq_ring); + else + return 0; } #endif @@ -7654,6 +7457,7 @@ static int gfx_v10_0_soft_reset(void *handle) case IP_VERSION(10, 3, 1): case IP_VERSION(10, 3, 4): case IP_VERSION(10, 3, 5): + case IP_VERSION(10, 3, 6): case IP_VERSION(10, 3, 3): if (REG_GET_FIELD(tmp, GRBM_STATUS2, RLC_BUSY_Sienna_Cichlid)) grbm_soft_reset = REG_SET_FIELD(grbm_soft_reset, @@ -7707,6 +7511,7 @@ static uint64_t gfx_v10_0_get_gpu_clock_counter(struct amdgpu_device *adev) switch (adev->ip_versions[GC_HWIP][0]) { case IP_VERSION(10, 3, 1): case IP_VERSION(10, 3, 3): + case IP_VERSION(10, 3, 7): preempt_disable(); clock_hi = RREG32_SOC15_NO_KIQ(SMUIO, 0, mmGOLDEN_TSC_COUNT_UPPER_Vangogh); clock_lo = RREG32_SOC15_NO_KIQ(SMUIO, 0, mmGOLDEN_TSC_COUNT_LOWER_Vangogh); @@ -7721,6 +7526,21 @@ static uint64_t gfx_v10_0_get_gpu_clock_counter(struct amdgpu_device *adev) preempt_enable(); clock = clock_lo | (clock_hi << 32ULL); break; + case IP_VERSION(10, 3, 6): + preempt_disable(); + clock_hi = RREG32_SOC15_NO_KIQ(SMUIO, 0, mmGOLDEN_TSC_COUNT_UPPER_GC_10_3_6); + clock_lo = RREG32_SOC15_NO_KIQ(SMUIO, 0, mmGOLDEN_TSC_COUNT_LOWER_GC_10_3_6); + hi_check = RREG32_SOC15_NO_KIQ(SMUIO, 0, mmGOLDEN_TSC_COUNT_UPPER_GC_10_3_6); + /* The SMUIO TSC clock frequency is 100MHz, which sets 32-bit carry over + * roughly every 42 seconds. + */ + if (hi_check != clock_hi) { + clock_lo = RREG32_SOC15_NO_KIQ(SMUIO, 0, mmGOLDEN_TSC_COUNT_LOWER_GC_10_3_6); + clock_hi = hi_check; + } + preempt_enable(); + clock = clock_lo | (clock_hi << 32ULL); + break; default: preempt_disable(); clock_hi = RREG32_SOC15_NO_KIQ(SMUIO, 0, mmGOLDEN_TSC_COUNT_UPPER); @@ -7778,6 +7598,7 @@ static int gfx_v10_0_early_init(void *handle) case IP_VERSION(10, 1, 1): case IP_VERSION(10, 1, 2): case IP_VERSION(10, 1, 3): + case IP_VERSION(10, 1, 4): adev->gfx.num_gfx_rings = GFX10_NUM_GFX_RINGS_NV1X; break; case IP_VERSION(10, 3, 0): @@ -7785,7 +7606,9 @@ static int gfx_v10_0_early_init(void *handle) case IP_VERSION(10, 3, 1): case IP_VERSION(10, 3, 4): case IP_VERSION(10, 3, 5): + case IP_VERSION(10, 3, 6): case IP_VERSION(10, 3, 3): + case IP_VERSION(10, 3, 7): adev->gfx.num_gfx_rings = GFX10_NUM_GFX_RINGS_Sienna_Cichlid; break; default: @@ -7800,6 +7623,10 @@ static int gfx_v10_0_early_init(void *handle) gfx_v10_0_set_irq_funcs(adev); gfx_v10_0_set_gds_init(adev); gfx_v10_0_set_rlc_funcs(adev); + gfx_v10_0_set_mqd_funcs(adev); + + /* init rlcg reg access ctrl */ + gfx_v10_0_init_rlcg_reg_access_ctrl(adev); return 0; } @@ -7843,7 +7670,9 @@ static void gfx_v10_0_set_safe_mode(struct amdgpu_device *adev) case IP_VERSION(10, 3, 1): case IP_VERSION(10, 3, 4): case IP_VERSION(10, 3, 5): + case IP_VERSION(10, 3, 6): case IP_VERSION(10, 3, 3): + case IP_VERSION(10, 3, 7): WREG32_SOC15(GC, 0, mmRLC_SAFE_MODE_Sienna_Cichlid, data); /* wait for RLC_SAFE_MODE */ @@ -7879,7 +7708,9 @@ static void gfx_v10_0_unset_safe_mode(struct amdgpu_device *adev) case IP_VERSION(10, 3, 1): case IP_VERSION(10, 3, 4): case IP_VERSION(10, 3, 5): + case IP_VERSION(10, 3, 6): case IP_VERSION(10, 3, 3): + case IP_VERSION(10, 3, 7): WREG32_SOC15(GC, 0, mmRLC_SAFE_MODE_Sienna_Cichlid, data); break; default: @@ -8333,6 +8164,8 @@ static void gfx_v10_cntl_power_gating(struct amdgpu_device *adev, bool enable) switch (adev->ip_versions[GC_HWIP][0]) { case IP_VERSION(10, 3, 1): case IP_VERSION(10, 3, 3): + case IP_VERSION(10, 3, 6): + case IP_VERSION(10, 3, 7): data = 0x4E20 & RLC_PG_DELAY_3__CGCG_ACTIVE_BEFORE_CGPG_MASK_Vangogh; WREG32_SOC15(GC, 0, mmRLC_PG_DELAY_3, data); break; @@ -8377,8 +8210,6 @@ static const struct amdgpu_rlc_funcs gfx_v10_0_rlc_funcs_sriov = { .reset = gfx_v10_0_rlc_reset, .start = gfx_v10_0_rlc_start, .update_spm_vmid = gfx_v10_0_update_spm_vmid, - .sriov_wreg = gfx_v10_sriov_wreg, - .sriov_rreg = gfx_v10_sriov_rreg, .is_rlcg_access_range = gfx_v10_0_is_rlcg_access_range, }; @@ -8403,6 +8234,8 @@ static int gfx_v10_0_set_powergating_state(void *handle, break; case IP_VERSION(10, 3, 1): case IP_VERSION(10, 3, 3): + case IP_VERSION(10, 3, 6): + case IP_VERSION(10, 3, 7): gfx_v10_cntl_pg(adev, enable); amdgpu_gfx_off_ctrl(adev, enable); break; @@ -8429,7 +8262,9 @@ static int gfx_v10_0_set_clockgating_state(void *handle, case IP_VERSION(10, 3, 1): case IP_VERSION(10, 3, 4): case IP_VERSION(10, 3, 5): + case IP_VERSION(10, 3, 6): case IP_VERSION(10, 3, 3): + case IP_VERSION(10, 3, 7): gfx_v10_0_update_gfx_clock_gating(adev, state == AMD_CG_STATE_GATE); break; @@ -8439,7 +8274,7 @@ static int gfx_v10_0_set_clockgating_state(void *handle, return 0; } -static void gfx_v10_0_get_clockgating_state(void *handle, u32 *flags) +static void gfx_v10_0_get_clockgating_state(void *handle, u64 *flags) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; int data; @@ -8485,7 +8320,8 @@ static void gfx_v10_0_get_clockgating_state(void *handle, u32 *flags) static u64 gfx_v10_0_ring_get_rptr_gfx(struct amdgpu_ring *ring) { - return ring->adev->wb.wb[ring->rptr_offs]; /* gfx10 is 32bit rptr*/ + /* gfx10 is 32bit rptr*/ + return *(uint32_t *)ring->rptr_cpu_addr; } static u64 gfx_v10_0_ring_get_wptr_gfx(struct amdgpu_ring *ring) @@ -8495,7 +8331,7 @@ static u64 gfx_v10_0_ring_get_wptr_gfx(struct amdgpu_ring *ring) /* XXX check if swapping is necessary on BE */ if (ring->use_doorbell) { - wptr = atomic64_read((atomic64_t *)&adev->wb.wb[ring->wptr_offs]); + wptr = atomic64_read((atomic64_t *)ring->wptr_cpu_addr); } else { wptr = RREG32_SOC15(GC, 0, mmCP_RB0_WPTR); wptr += (u64)RREG32_SOC15(GC, 0, mmCP_RB0_WPTR_HI) << 32; @@ -8507,20 +8343,52 @@ static u64 gfx_v10_0_ring_get_wptr_gfx(struct amdgpu_ring *ring) static void gfx_v10_0_ring_set_wptr_gfx(struct amdgpu_ring *ring) { struct amdgpu_device *adev = ring->adev; + uint32_t *wptr_saved; + uint32_t *is_queue_unmap; + uint64_t aggregated_db_index; + uint32_t mqd_size = adev->mqds[AMDGPU_HW_IP_GFX].mqd_size; + uint64_t wptr_tmp; + + if (ring->is_mes_queue) { + wptr_saved = (uint32_t *)(ring->mqd_ptr + mqd_size); + is_queue_unmap = (uint32_t *)(ring->mqd_ptr + mqd_size + + sizeof(uint32_t)); + aggregated_db_index = + amdgpu_mes_get_aggregated_doorbell_index(adev, + AMDGPU_MES_PRIORITY_LEVEL_NORMAL); + + wptr_tmp = ring->wptr & ring->buf_mask; + atomic64_set((atomic64_t *)ring->wptr_cpu_addr, wptr_tmp); + *wptr_saved = wptr_tmp; + /* assume doorbell always being used by mes mapped queue */ + if (*is_queue_unmap) { + WDOORBELL64(aggregated_db_index, wptr_tmp); + WDOORBELL64(ring->doorbell_index, wptr_tmp); + } else { + WDOORBELL64(ring->doorbell_index, wptr_tmp); - if (ring->use_doorbell) { - /* XXX check if swapping is necessary on BE */ - atomic64_set((atomic64_t *)&adev->wb.wb[ring->wptr_offs], ring->wptr); - WDOORBELL64(ring->doorbell_index, ring->wptr); + if (*is_queue_unmap) + WDOORBELL64(aggregated_db_index, wptr_tmp); + } } else { - WREG32_SOC15(GC, 0, mmCP_RB0_WPTR, lower_32_bits(ring->wptr)); - WREG32_SOC15(GC, 0, mmCP_RB0_WPTR_HI, upper_32_bits(ring->wptr)); + if (ring->use_doorbell) { + /* XXX check if swapping is necessary on BE */ + atomic64_set((atomic64_t *)ring->wptr_cpu_addr, + ring->wptr); + WDOORBELL64(ring->doorbell_index, ring->wptr); + } else { + WREG32_SOC15(GC, 0, mmCP_RB0_WPTR, + lower_32_bits(ring->wptr)); + WREG32_SOC15(GC, 0, mmCP_RB0_WPTR_HI, + upper_32_bits(ring->wptr)); + } } } static u64 gfx_v10_0_ring_get_rptr_compute(struct amdgpu_ring *ring) { - return ring->adev->wb.wb[ring->rptr_offs]; /* gfx10 hardware is 32bit rptr */ + /* gfx10 hardware is 32bit rptr */ + return *(uint32_t *)ring->rptr_cpu_addr; } static u64 gfx_v10_0_ring_get_wptr_compute(struct amdgpu_ring *ring) @@ -8529,7 +8397,7 @@ static u64 gfx_v10_0_ring_get_wptr_compute(struct amdgpu_ring *ring) /* XXX check if swapping is necessary on BE */ if (ring->use_doorbell) - wptr = atomic64_read((atomic64_t *)&ring->adev->wb.wb[ring->wptr_offs]); + wptr = atomic64_read((atomic64_t *)ring->wptr_cpu_addr); else BUG(); return wptr; @@ -8538,13 +8406,42 @@ static u64 gfx_v10_0_ring_get_wptr_compute(struct amdgpu_ring *ring) static void gfx_v10_0_ring_set_wptr_compute(struct amdgpu_ring *ring) { struct amdgpu_device *adev = ring->adev; + uint32_t *wptr_saved; + uint32_t *is_queue_unmap; + uint64_t aggregated_db_index; + uint32_t mqd_size = adev->mqds[AMDGPU_HW_IP_COMPUTE].mqd_size; + uint64_t wptr_tmp; + + if (ring->is_mes_queue) { + wptr_saved = (uint32_t *)(ring->mqd_ptr + mqd_size); + is_queue_unmap = (uint32_t *)(ring->mqd_ptr + mqd_size + + sizeof(uint32_t)); + aggregated_db_index = + amdgpu_mes_get_aggregated_doorbell_index(adev, + AMDGPU_MES_PRIORITY_LEVEL_NORMAL); + + wptr_tmp = ring->wptr & ring->buf_mask; + atomic64_set((atomic64_t *)ring->wptr_cpu_addr, wptr_tmp); + *wptr_saved = wptr_tmp; + /* assume doorbell always used by mes mapped queue */ + if (*is_queue_unmap) { + WDOORBELL64(aggregated_db_index, wptr_tmp); + WDOORBELL64(ring->doorbell_index, wptr_tmp); + } else { + WDOORBELL64(ring->doorbell_index, wptr_tmp); - /* XXX check if swapping is necessary on BE */ - if (ring->use_doorbell) { - atomic64_set((atomic64_t *)&adev->wb.wb[ring->wptr_offs], ring->wptr); - WDOORBELL64(ring->doorbell_index, ring->wptr); + if (*is_queue_unmap) + WDOORBELL64(aggregated_db_index, wptr_tmp); + } } else { - BUG(); /* only DOORBELL method supported on gfx10 now */ + /* XXX check if swapping is necessary on BE */ + if (ring->use_doorbell) { + atomic64_set((atomic64_t *)ring->wptr_cpu_addr, + ring->wptr); + WDOORBELL64(ring->doorbell_index, ring->wptr); + } else { + BUG(); /* only DOORBELL method supported on gfx10 now */ + } } } @@ -8603,6 +8500,10 @@ static void gfx_v10_0_ring_emit_ib_gfx(struct amdgpu_ring *ring, (!amdgpu_sriov_vf(ring->adev) && flags & AMDGPU_IB_PREEMPTED) ? true : false); } + if (ring->is_mes_queue) + /* inherit vmid from mqd */ + control |= 0x400000; + amdgpu_ring_write(ring, header); BUG_ON(ib->gpu_addr & 0x3); /* Dword align */ amdgpu_ring_write(ring, @@ -8622,6 +8523,10 @@ static void gfx_v10_0_ring_emit_ib_compute(struct amdgpu_ring *ring, unsigned vmid = AMDGPU_JOB_GET_VMID(job); u32 control = INDIRECT_BUFFER_VALID | ib->length_dw | (vmid << 24); + if (ring->is_mes_queue) + /* inherit vmid from mqd */ + control |= 0x40000000; + /* Currently, there is a high possibility to get wave ID mismatch * between ME and GDS, leading to a hw deadlock, because ME generates * different wave IDs than the GDS expects. This situation happens @@ -8679,7 +8584,8 @@ static void gfx_v10_0_ring_emit_fence(struct amdgpu_ring *ring, u64 addr, amdgpu_ring_write(ring, upper_32_bits(addr)); amdgpu_ring_write(ring, lower_32_bits(seq)); amdgpu_ring_write(ring, upper_32_bits(seq)); - amdgpu_ring_write(ring, 0); + amdgpu_ring_write(ring, ring->is_mes_queue ? + (ring->hw_queue_id | AMDGPU_FENCE_MES_QUEUE_FLAG) : 0); } static void gfx_v10_0_ring_emit_pipeline_sync(struct amdgpu_ring *ring) @@ -8692,10 +8598,25 @@ static void gfx_v10_0_ring_emit_pipeline_sync(struct amdgpu_ring *ring) upper_32_bits(addr), seq, 0xffffffff, 4); } +static void gfx_v10_0_ring_invalidate_tlbs(struct amdgpu_ring *ring, + uint16_t pasid, uint32_t flush_type, + bool all_hub, uint8_t dst_sel) +{ + amdgpu_ring_write(ring, PACKET3(PACKET3_INVALIDATE_TLBS, 0)); + amdgpu_ring_write(ring, + PACKET3_INVALIDATE_TLBS_DST_SEL(dst_sel) | + PACKET3_INVALIDATE_TLBS_ALL_HUB(all_hub) | + PACKET3_INVALIDATE_TLBS_PASID(pasid) | + PACKET3_INVALIDATE_TLBS_FLUSH_TYPE(flush_type)); +} + static void gfx_v10_0_ring_emit_vm_flush(struct amdgpu_ring *ring, unsigned vmid, uint64_t pd_addr) { - amdgpu_gmc_emit_flush_gpu_tlb(ring, vmid, pd_addr); + if (ring->is_mes_queue) + gfx_v10_0_ring_invalidate_tlbs(ring, 0, 0, false, 0); + else + amdgpu_gmc_emit_flush_gpu_tlb(ring, vmid, pd_addr); /* compute doesn't have PFP */ if (ring->funcs->type == AMDGPU_RING_TYPE_GFX) { @@ -8850,26 +8771,36 @@ static void gfx_v10_0_ring_emit_ce_meta(struct amdgpu_ring *ring, bool resume) { struct amdgpu_device *adev = ring->adev; struct v10_ce_ib_state ce_payload = {0}; - uint64_t csa_addr; + uint64_t offset, ce_payload_gpu_addr; + void *ce_payload_cpu_addr; int cnt; cnt = (sizeof(ce_payload) >> 2) + 4 - 2; - csa_addr = amdgpu_csa_vaddr(ring->adev); + + if (ring->is_mes_queue) { + offset = offsetof(struct amdgpu_mes_ctx_meta_data, + gfx[0].gfx_meta_data) + + offsetof(struct v10_gfx_meta_data, ce_payload); + ce_payload_gpu_addr = + amdgpu_mes_ctx_get_offs_gpu_addr(ring, offset); + ce_payload_cpu_addr = + amdgpu_mes_ctx_get_offs_cpu_addr(ring, offset); + } else { + offset = offsetof(struct v10_gfx_meta_data, ce_payload); + ce_payload_gpu_addr = amdgpu_csa_vaddr(ring->adev) + offset; + ce_payload_cpu_addr = adev->virt.csa_cpu_addr + offset; + } amdgpu_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, cnt)); amdgpu_ring_write(ring, (WRITE_DATA_ENGINE_SEL(2) | WRITE_DATA_DST_SEL(8) | WR_CONFIRM) | WRITE_DATA_CACHE_POLICY(0)); - amdgpu_ring_write(ring, lower_32_bits(csa_addr + - offsetof(struct v10_gfx_meta_data, ce_payload))); - amdgpu_ring_write(ring, upper_32_bits(csa_addr + - offsetof(struct v10_gfx_meta_data, ce_payload))); + amdgpu_ring_write(ring, lower_32_bits(ce_payload_gpu_addr)); + amdgpu_ring_write(ring, upper_32_bits(ce_payload_gpu_addr)); if (resume) - amdgpu_ring_write_multiple(ring, adev->virt.csa_cpu_addr + - offsetof(struct v10_gfx_meta_data, - ce_payload), + amdgpu_ring_write_multiple(ring, ce_payload_cpu_addr, sizeof(ce_payload) >> 2); else amdgpu_ring_write_multiple(ring, (void *)&ce_payload, @@ -8880,12 +8811,33 @@ static void gfx_v10_0_ring_emit_de_meta(struct amdgpu_ring *ring, bool resume) { struct amdgpu_device *adev = ring->adev; struct v10_de_ib_state de_payload = {0}; - uint64_t csa_addr, gds_addr; + uint64_t offset, gds_addr, de_payload_gpu_addr; + void *de_payload_cpu_addr; int cnt; - csa_addr = amdgpu_csa_vaddr(ring->adev); - gds_addr = ALIGN(csa_addr + AMDGPU_CSA_SIZE - adev->gds.gds_size, - PAGE_SIZE); + if (ring->is_mes_queue) { + offset = offsetof(struct amdgpu_mes_ctx_meta_data, + gfx[0].gfx_meta_data) + + offsetof(struct v10_gfx_meta_data, de_payload); + de_payload_gpu_addr = + amdgpu_mes_ctx_get_offs_gpu_addr(ring, offset); + de_payload_cpu_addr = + amdgpu_mes_ctx_get_offs_cpu_addr(ring, offset); + + offset = offsetof(struct amdgpu_mes_ctx_meta_data, + gfx[0].gds_backup) + + offsetof(struct v10_gfx_meta_data, de_payload); + gds_addr = amdgpu_mes_ctx_get_offs_gpu_addr(ring, offset); + } else { + offset = offsetof(struct v10_gfx_meta_data, de_payload); + de_payload_gpu_addr = amdgpu_csa_vaddr(ring->adev) + offset; + de_payload_cpu_addr = adev->virt.csa_cpu_addr + offset; + + gds_addr = ALIGN(amdgpu_csa_vaddr(ring->adev) + + AMDGPU_CSA_SIZE - adev->gds.gds_size, + PAGE_SIZE); + } + de_payload.gds_backup_addrlo = lower_32_bits(gds_addr); de_payload.gds_backup_addrhi = upper_32_bits(gds_addr); @@ -8895,15 +8847,11 @@ static void gfx_v10_0_ring_emit_de_meta(struct amdgpu_ring *ring, bool resume) WRITE_DATA_DST_SEL(8) | WR_CONFIRM) | WRITE_DATA_CACHE_POLICY(0)); - amdgpu_ring_write(ring, lower_32_bits(csa_addr + - offsetof(struct v10_gfx_meta_data, de_payload))); - amdgpu_ring_write(ring, upper_32_bits(csa_addr + - offsetof(struct v10_gfx_meta_data, de_payload))); + amdgpu_ring_write(ring, lower_32_bits(de_payload_gpu_addr)); + amdgpu_ring_write(ring, upper_32_bits(de_payload_gpu_addr)); if (resume) - amdgpu_ring_write_multiple(ring, adev->virt.csa_cpu_addr + - offsetof(struct v10_gfx_meta_data, - de_payload), + amdgpu_ring_write_multiple(ring, de_payload_cpu_addr, sizeof(de_payload) >> 2); else amdgpu_ring_write_multiple(ring, (void *)&de_payload, @@ -9140,31 +9088,51 @@ static int gfx_v10_0_eop_irq(struct amdgpu_device *adev, int i; u8 me_id, pipe_id, queue_id; struct amdgpu_ring *ring; + uint32_t mes_queue_id = entry->src_data[0]; DRM_DEBUG("IH: CP EOP\n"); - me_id = (entry->ring_id & 0x0c) >> 2; - pipe_id = (entry->ring_id & 0x03) >> 0; - queue_id = (entry->ring_id & 0x70) >> 4; - switch (me_id) { - case 0: - if (pipe_id == 0) - amdgpu_fence_process(&adev->gfx.gfx_ring[0]); - else - amdgpu_fence_process(&adev->gfx.gfx_ring[1]); - break; - case 1: - case 2: - for (i = 0; i < adev->gfx.num_compute_rings; i++) { - ring = &adev->gfx.compute_ring[i]; - /* Per-queue interrupt is supported for MEC starting from VI. - * The interrupt can only be enabled/disabled per pipe instead of per queue. - */ - if ((ring->me == me_id) && (ring->pipe == pipe_id) && (ring->queue == queue_id)) - amdgpu_fence_process(ring); + if (adev->enable_mes && (mes_queue_id & AMDGPU_FENCE_MES_QUEUE_FLAG)) { + struct amdgpu_mes_queue *queue; + + mes_queue_id &= AMDGPU_FENCE_MES_QUEUE_ID_MASK; + + spin_lock(&adev->mes.queue_id_lock); + queue = idr_find(&adev->mes.queue_id_idr, mes_queue_id); + if (queue) { + DRM_DEBUG("process mes queue id = %d\n", mes_queue_id); + amdgpu_fence_process(queue->ring); + } + spin_unlock(&adev->mes.queue_id_lock); + } else { + me_id = (entry->ring_id & 0x0c) >> 2; + pipe_id = (entry->ring_id & 0x03) >> 0; + queue_id = (entry->ring_id & 0x70) >> 4; + + switch (me_id) { + case 0: + if (pipe_id == 0) + amdgpu_fence_process(&adev->gfx.gfx_ring[0]); + else + amdgpu_fence_process(&adev->gfx.gfx_ring[1]); + break; + case 1: + case 2: + for (i = 0; i < adev->gfx.num_compute_rings; i++) { + ring = &adev->gfx.compute_ring[i]; + /* Per-queue interrupt is supported for MEC starting from VI. + * The interrupt can only be enabled/disabled per pipe instead + * of per queue. + */ + if ((ring->me == me_id) && + (ring->pipe == pipe_id) && + (ring->queue == queue_id)) + amdgpu_fence_process(ring); + } + break; } - break; } + return 0; } @@ -9366,6 +9334,7 @@ static const struct amdgpu_ring_funcs gfx_v10_0_ring_funcs_gfx = { .align_mask = 0xff, .nop = PACKET3(PACKET3_NOP, 0x3FFF), .support_64bit_ptrs = true, + .secure_submission_supported = true, .vmhub = AMDGPU_GFXHUB_0, .get_rptr = gfx_v10_0_ring_get_rptr_gfx, .get_wptr = gfx_v10_0_ring_get_wptr_gfx, @@ -9537,11 +9506,14 @@ static void gfx_v10_0_set_rlc_funcs(struct amdgpu_device *adev) case IP_VERSION(10, 1, 10): case IP_VERSION(10, 1, 1): case IP_VERSION(10, 1, 3): + case IP_VERSION(10, 1, 4): case IP_VERSION(10, 3, 2): case IP_VERSION(10, 3, 1): case IP_VERSION(10, 3, 4): case IP_VERSION(10, 3, 5): + case IP_VERSION(10, 3, 6): case IP_VERSION(10, 3, 3): + case IP_VERSION(10, 3, 7): adev->gfx.rlc.funcs = &gfx_v10_0_rlc_funcs; break; case IP_VERSION(10, 1, 2): @@ -9565,6 +9537,20 @@ static void gfx_v10_0_set_gds_init(struct amdgpu_device *adev) adev->gds.oa_size = 16; } +static void gfx_v10_0_set_mqd_funcs(struct amdgpu_device *adev) +{ + /* set gfx eng mqd */ + adev->mqds[AMDGPU_HW_IP_GFX].mqd_size = + sizeof(struct v10_gfx_mqd); + adev->mqds[AMDGPU_HW_IP_GFX].init_mqd = + gfx_v10_0_gfx_mqd_init; + /* set compute eng mqd */ + adev->mqds[AMDGPU_HW_IP_COMPUTE].mqd_size = + sizeof(struct v10_compute_mqd); + adev->mqds[AMDGPU_HW_IP_COMPUTE].init_mqd = + gfx_v10_0_compute_mqd_init; +} + static void gfx_v10_0_set_user_wgp_inactive_bitmap_per_sh(struct amdgpu_device *adev, u32 bitmap) { @@ -9634,7 +9620,9 @@ static int gfx_v10_0_get_cu_info(struct amdgpu_device *adev, for (j = 0; j < adev->gfx.config.max_sh_per_se; j++) { bitmap = i * adev->gfx.config.max_sh_per_se + j; if (((adev->ip_versions[GC_HWIP][0] == IP_VERSION(10, 3, 0)) || - (adev->ip_versions[GC_HWIP][0] == IP_VERSION(10, 3, 3))) && + (adev->ip_versions[GC_HWIP][0] == IP_VERSION(10, 3, 3)) || + (adev->ip_versions[GC_HWIP][0] == IP_VERSION(10, 3, 6)) || + (adev->ip_versions[GC_HWIP][0] == IP_VERSION(10, 3, 7))) && ((gfx_v10_3_get_disabled_sa(adev) >> bitmap) & 1)) continue; mask = 1; diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c new file mode 100644 index 000000000000..0fecc5bf45bc --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c @@ -0,0 +1,6366 @@ +/* + * Copyright 2021 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ +#include <linux/delay.h> +#include <linux/kernel.h> +#include <linux/firmware.h> +#include <linux/module.h> +#include <linux/pci.h> +#include "amdgpu.h" +#include "amdgpu_gfx.h" +#include "amdgpu_psp.h" +#include "amdgpu_smu.h" +#include "amdgpu_atomfirmware.h" +#include "imu_v11_0.h" +#include "soc21.h" +#include "nvd.h" + +#include "gc/gc_11_0_0_offset.h" +#include "gc/gc_11_0_0_sh_mask.h" +#include "smuio/smuio_13_0_6_offset.h" +#include "smuio/smuio_13_0_6_sh_mask.h" +#include "navi10_enum.h" +#include "ivsrcid/gfx/irqsrcs_gfx_11_0_0.h" + +#include "soc15.h" +#include "soc15d.h" +#include "clearstate_gfx11.h" +#include "v11_structs.h" +#include "gfx_v11_0.h" +#include "nbio_v4_3.h" +#include "mes_v11_0.h" + +#define GFX11_NUM_GFX_RINGS 1 +#define GFX11_MEC_HPD_SIZE 2048 + +#define RLCG_UCODE_LOADING_START_ADDRESS 0x00002000L +#define RLC_PG_DELAY_3_DEFAULT_GC_11_0_1 0x1388 + +#define regCGTT_WD_CLK_CTRL 0x5086 +#define regCGTT_WD_CLK_CTRL_BASE_IDX 1 +#define regRLC_RLCS_BOOTLOAD_STATUS_gc_11_0_1 0x4e7e +#define regRLC_RLCS_BOOTLOAD_STATUS_gc_11_0_1_BASE_IDX 1 + +MODULE_FIRMWARE("amdgpu/gc_11_0_0_pfp.bin"); +MODULE_FIRMWARE("amdgpu/gc_11_0_0_me.bin"); +MODULE_FIRMWARE("amdgpu/gc_11_0_0_mec.bin"); +MODULE_FIRMWARE("amdgpu/gc_11_0_0_rlc.bin"); +MODULE_FIRMWARE("amdgpu/gc_11_0_0_toc.bin"); +MODULE_FIRMWARE("amdgpu/gc_11_0_1_pfp.bin"); +MODULE_FIRMWARE("amdgpu/gc_11_0_1_me.bin"); +MODULE_FIRMWARE("amdgpu/gc_11_0_1_mec.bin"); +MODULE_FIRMWARE("amdgpu/gc_11_0_1_rlc.bin"); +MODULE_FIRMWARE("amdgpu/gc_11_0_2_pfp.bin"); +MODULE_FIRMWARE("amdgpu/gc_11_0_2_me.bin"); +MODULE_FIRMWARE("amdgpu/gc_11_0_2_mec.bin"); +MODULE_FIRMWARE("amdgpu/gc_11_0_2_rlc.bin"); +MODULE_FIRMWARE("amdgpu/gc_11_0_3_pfp.bin"); +MODULE_FIRMWARE("amdgpu/gc_11_0_3_me.bin"); +MODULE_FIRMWARE("amdgpu/gc_11_0_3_mec.bin"); +MODULE_FIRMWARE("amdgpu/gc_11_0_3_rlc.bin"); + +static const struct soc15_reg_golden golden_settings_gc_11_0_1[] = +{ + SOC15_REG_GOLDEN_VALUE(GC, 0, regCGTT_GS_NGG_CLK_CTRL, 0x9fff8fff, 0x00000010), + SOC15_REG_GOLDEN_VALUE(GC, 0, regCGTT_WD_CLK_CTRL, 0xffff8fff, 0x00000010), + SOC15_REG_GOLDEN_VALUE(GC, 0, regCPF_GCR_CNTL, 0x0007ffff, 0x0000c200), + SOC15_REG_GOLDEN_VALUE(GC, 0, regGL2C_CTRL3, 0xffff001b, 0x00f01988), + SOC15_REG_GOLDEN_VALUE(GC, 0, regPA_CL_ENHANCE, 0xf0ffffff, 0x00880007), + SOC15_REG_GOLDEN_VALUE(GC, 0, regPA_SC_ENHANCE_3, 0xfffffffd, 0x00000008), + SOC15_REG_GOLDEN_VALUE(GC, 0, regPA_SC_VRS_SURFACE_CNTL_1, 0xfff891ff, 0x55480100), + SOC15_REG_GOLDEN_VALUE(GC, 0, regTA_CNTL_AUX, 0xf7f7ffff, 0x01030000), + SOC15_REG_GOLDEN_VALUE(GC, 0, regTCP_CNTL2, 0xfcffffff, 0x0000000a) +}; + +#define DEFAULT_SH_MEM_CONFIG \ + ((SH_MEM_ADDRESS_MODE_64 << SH_MEM_CONFIG__ADDRESS_MODE__SHIFT) | \ + (SH_MEM_ALIGNMENT_MODE_UNALIGNED << SH_MEM_CONFIG__ALIGNMENT_MODE__SHIFT) | \ + (3 << SH_MEM_CONFIG__INITIAL_INST_PREFETCH__SHIFT)) + +static void gfx_v11_0_disable_gpa_mode(struct amdgpu_device *adev); +static void gfx_v11_0_set_ring_funcs(struct amdgpu_device *adev); +static void gfx_v11_0_set_irq_funcs(struct amdgpu_device *adev); +static void gfx_v11_0_set_gds_init(struct amdgpu_device *adev); +static void gfx_v11_0_set_rlc_funcs(struct amdgpu_device *adev); +static void gfx_v11_0_set_mqd_funcs(struct amdgpu_device *adev); +static void gfx_v11_0_set_imu_funcs(struct amdgpu_device *adev); +static int gfx_v11_0_get_cu_info(struct amdgpu_device *adev, + struct amdgpu_cu_info *cu_info); +static uint64_t gfx_v11_0_get_gpu_clock_counter(struct amdgpu_device *adev); +static void gfx_v11_0_select_se_sh(struct amdgpu_device *adev, u32 se_num, + u32 sh_num, u32 instance); +static u32 gfx_v11_0_get_wgp_active_bitmap_per_sh(struct amdgpu_device *adev); + +static void gfx_v11_0_ring_emit_de_meta(struct amdgpu_ring *ring, bool resume); +static void gfx_v11_0_ring_emit_frame_cntl(struct amdgpu_ring *ring, bool start, bool secure); +static void gfx_v11_0_ring_emit_wreg(struct amdgpu_ring *ring, uint32_t reg, + uint32_t val); +static int gfx_v11_0_wait_for_rlc_autoload_complete(struct amdgpu_device *adev); +static void gfx_v11_0_ring_invalidate_tlbs(struct amdgpu_ring *ring, + uint16_t pasid, uint32_t flush_type, + bool all_hub, uint8_t dst_sel); +static void gfx_v11_0_set_safe_mode(struct amdgpu_device *adev); +static void gfx_v11_0_unset_safe_mode(struct amdgpu_device *adev); +static void gfx_v11_0_update_perf_clk(struct amdgpu_device *adev, + bool enable); + +static void gfx11_kiq_set_resources(struct amdgpu_ring *kiq_ring, uint64_t queue_mask) +{ + amdgpu_ring_write(kiq_ring, PACKET3(PACKET3_SET_RESOURCES, 6)); + amdgpu_ring_write(kiq_ring, PACKET3_SET_RESOURCES_VMID_MASK(0) | + PACKET3_SET_RESOURCES_QUEUE_TYPE(0)); /* vmid_mask:0 queue_type:0 (KIQ) */ + amdgpu_ring_write(kiq_ring, lower_32_bits(queue_mask)); /* queue mask lo */ + amdgpu_ring_write(kiq_ring, upper_32_bits(queue_mask)); /* queue mask hi */ + amdgpu_ring_write(kiq_ring, 0); /* gws mask lo */ + amdgpu_ring_write(kiq_ring, 0); /* gws mask hi */ + amdgpu_ring_write(kiq_ring, 0); /* oac mask */ + amdgpu_ring_write(kiq_ring, 0); /* gds heap base:0, gds heap size:0 */ +} + +static void gfx11_kiq_map_queues(struct amdgpu_ring *kiq_ring, + struct amdgpu_ring *ring) +{ + uint64_t mqd_addr = amdgpu_bo_gpu_offset(ring->mqd_obj); + uint64_t wptr_addr = ring->wptr_gpu_addr; + uint32_t me = 0, eng_sel = 0; + + switch (ring->funcs->type) { + case AMDGPU_RING_TYPE_COMPUTE: + me = 1; + eng_sel = 0; + break; + case AMDGPU_RING_TYPE_GFX: + me = 0; + eng_sel = 4; + break; + case AMDGPU_RING_TYPE_MES: + me = 2; + eng_sel = 5; + break; + default: + WARN_ON(1); + } + + amdgpu_ring_write(kiq_ring, PACKET3(PACKET3_MAP_QUEUES, 5)); + /* Q_sel:0, vmid:0, vidmem: 1, engine:0, num_Q:1*/ + amdgpu_ring_write(kiq_ring, /* Q_sel: 0, vmid: 0, engine: 0, num_Q: 1 */ + PACKET3_MAP_QUEUES_QUEUE_SEL(0) | /* Queue_Sel */ + PACKET3_MAP_QUEUES_VMID(0) | /* VMID */ + PACKET3_MAP_QUEUES_QUEUE(ring->queue) | + PACKET3_MAP_QUEUES_PIPE(ring->pipe) | + PACKET3_MAP_QUEUES_ME((me)) | + PACKET3_MAP_QUEUES_QUEUE_TYPE(0) | /*queue_type: normal compute queue */ + PACKET3_MAP_QUEUES_ALLOC_FORMAT(0) | /* alloc format: all_on_one_pipe */ + PACKET3_MAP_QUEUES_ENGINE_SEL(eng_sel) | + PACKET3_MAP_QUEUES_NUM_QUEUES(1)); /* num_queues: must be 1 */ + amdgpu_ring_write(kiq_ring, PACKET3_MAP_QUEUES_DOORBELL_OFFSET(ring->doorbell_index)); + amdgpu_ring_write(kiq_ring, lower_32_bits(mqd_addr)); + amdgpu_ring_write(kiq_ring, upper_32_bits(mqd_addr)); + amdgpu_ring_write(kiq_ring, lower_32_bits(wptr_addr)); + amdgpu_ring_write(kiq_ring, upper_32_bits(wptr_addr)); +} + +static void gfx11_kiq_unmap_queues(struct amdgpu_ring *kiq_ring, + struct amdgpu_ring *ring, + enum amdgpu_unmap_queues_action action, + u64 gpu_addr, u64 seq) +{ + struct amdgpu_device *adev = kiq_ring->adev; + uint32_t eng_sel = ring->funcs->type == AMDGPU_RING_TYPE_GFX ? 4 : 0; + + if (adev->enable_mes && !adev->gfx.kiq.ring.sched.ready) { + amdgpu_mes_unmap_legacy_queue(adev, ring, action, gpu_addr, seq); + return; + } + + amdgpu_ring_write(kiq_ring, PACKET3(PACKET3_UNMAP_QUEUES, 4)); + amdgpu_ring_write(kiq_ring, /* Q_sel: 0, vmid: 0, engine: 0, num_Q: 1 */ + PACKET3_UNMAP_QUEUES_ACTION(action) | + PACKET3_UNMAP_QUEUES_QUEUE_SEL(0) | + PACKET3_UNMAP_QUEUES_ENGINE_SEL(eng_sel) | + PACKET3_UNMAP_QUEUES_NUM_QUEUES(1)); + amdgpu_ring_write(kiq_ring, + PACKET3_UNMAP_QUEUES_DOORBELL_OFFSET0(ring->doorbell_index)); + + if (action == PREEMPT_QUEUES_NO_UNMAP) { + amdgpu_ring_write(kiq_ring, lower_32_bits(gpu_addr)); + amdgpu_ring_write(kiq_ring, upper_32_bits(gpu_addr)); + amdgpu_ring_write(kiq_ring, seq); + } else { + amdgpu_ring_write(kiq_ring, 0); + amdgpu_ring_write(kiq_ring, 0); + amdgpu_ring_write(kiq_ring, 0); + } +} + +static void gfx11_kiq_query_status(struct amdgpu_ring *kiq_ring, + struct amdgpu_ring *ring, + u64 addr, + u64 seq) +{ + uint32_t eng_sel = ring->funcs->type == AMDGPU_RING_TYPE_GFX ? 4 : 0; + + amdgpu_ring_write(kiq_ring, PACKET3(PACKET3_QUERY_STATUS, 5)); + amdgpu_ring_write(kiq_ring, + PACKET3_QUERY_STATUS_CONTEXT_ID(0) | + PACKET3_QUERY_STATUS_INTERRUPT_SEL(0) | + PACKET3_QUERY_STATUS_COMMAND(2)); + amdgpu_ring_write(kiq_ring, /* Q_sel: 0, vmid: 0, engine: 0, num_Q: 1 */ + PACKET3_QUERY_STATUS_DOORBELL_OFFSET(ring->doorbell_index) | + PACKET3_QUERY_STATUS_ENG_SEL(eng_sel)); + amdgpu_ring_write(kiq_ring, lower_32_bits(addr)); + amdgpu_ring_write(kiq_ring, upper_32_bits(addr)); + amdgpu_ring_write(kiq_ring, lower_32_bits(seq)); + amdgpu_ring_write(kiq_ring, upper_32_bits(seq)); +} + +static void gfx11_kiq_invalidate_tlbs(struct amdgpu_ring *kiq_ring, + uint16_t pasid, uint32_t flush_type, + bool all_hub) +{ + gfx_v11_0_ring_invalidate_tlbs(kiq_ring, pasid, flush_type, all_hub, 1); +} + +static const struct kiq_pm4_funcs gfx_v11_0_kiq_pm4_funcs = { + .kiq_set_resources = gfx11_kiq_set_resources, + .kiq_map_queues = gfx11_kiq_map_queues, + .kiq_unmap_queues = gfx11_kiq_unmap_queues, + .kiq_query_status = gfx11_kiq_query_status, + .kiq_invalidate_tlbs = gfx11_kiq_invalidate_tlbs, + .set_resources_size = 8, + .map_queues_size = 7, + .unmap_queues_size = 6, + .query_status_size = 7, + .invalidate_tlbs_size = 2, +}; + +static void gfx_v11_0_set_kiq_pm4_funcs(struct amdgpu_device *adev) +{ + adev->gfx.kiq.pmf = &gfx_v11_0_kiq_pm4_funcs; +} + +static void gfx_v11_0_init_golden_registers(struct amdgpu_device *adev) +{ + switch (adev->ip_versions[GC_HWIP][0]) { + case IP_VERSION(11, 0, 1): + soc15_program_register_sequence(adev, + golden_settings_gc_11_0_1, + (const u32)ARRAY_SIZE(golden_settings_gc_11_0_1)); + break; + default: + break; + } +} + +static void gfx_v11_0_write_data_to_reg(struct amdgpu_ring *ring, int eng_sel, + bool wc, uint32_t reg, uint32_t val) +{ + amdgpu_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, 3)); + amdgpu_ring_write(ring, WRITE_DATA_ENGINE_SEL(eng_sel) | + WRITE_DATA_DST_SEL(0) | (wc ? WR_CONFIRM : 0)); + amdgpu_ring_write(ring, reg); + amdgpu_ring_write(ring, 0); + amdgpu_ring_write(ring, val); +} + +static void gfx_v11_0_wait_reg_mem(struct amdgpu_ring *ring, int eng_sel, + int mem_space, int opt, uint32_t addr0, + uint32_t addr1, uint32_t ref, uint32_t mask, + uint32_t inv) +{ + amdgpu_ring_write(ring, PACKET3(PACKET3_WAIT_REG_MEM, 5)); + amdgpu_ring_write(ring, + /* memory (1) or register (0) */ + (WAIT_REG_MEM_MEM_SPACE(mem_space) | + WAIT_REG_MEM_OPERATION(opt) | /* wait */ + WAIT_REG_MEM_FUNCTION(3) | /* equal */ + WAIT_REG_MEM_ENGINE(eng_sel))); + + if (mem_space) + BUG_ON(addr0 & 0x3); /* Dword align */ + amdgpu_ring_write(ring, addr0); + amdgpu_ring_write(ring, addr1); + amdgpu_ring_write(ring, ref); + amdgpu_ring_write(ring, mask); + amdgpu_ring_write(ring, inv); /* poll interval */ +} + +static int gfx_v11_0_ring_test_ring(struct amdgpu_ring *ring) +{ + struct amdgpu_device *adev = ring->adev; + uint32_t scratch = SOC15_REG_OFFSET(GC, 0, regSCRATCH_REG0); + uint32_t tmp = 0; + unsigned i; + int r; + + WREG32(scratch, 0xCAFEDEAD); + r = amdgpu_ring_alloc(ring, 5); + if (r) { + DRM_ERROR("amdgpu: cp failed to lock ring %d (%d).\n", + ring->idx, r); + return r; + } + + if (ring->funcs->type == AMDGPU_RING_TYPE_KIQ) { + gfx_v11_0_ring_emit_wreg(ring, scratch, 0xDEADBEEF); + } else { + amdgpu_ring_write(ring, PACKET3(PACKET3_SET_UCONFIG_REG, 1)); + amdgpu_ring_write(ring, scratch - + PACKET3_SET_UCONFIG_REG_START); + amdgpu_ring_write(ring, 0xDEADBEEF); + } + amdgpu_ring_commit(ring); + + for (i = 0; i < adev->usec_timeout; i++) { + tmp = RREG32(scratch); + if (tmp == 0xDEADBEEF) + break; + if (amdgpu_emu_mode == 1) + msleep(1); + else + udelay(1); + } + + if (i >= adev->usec_timeout) + r = -ETIMEDOUT; + return r; +} + +static int gfx_v11_0_ring_test_ib(struct amdgpu_ring *ring, long timeout) +{ + struct amdgpu_device *adev = ring->adev; + struct amdgpu_ib ib; + struct dma_fence *f = NULL; + unsigned index; + uint64_t gpu_addr; + volatile uint32_t *cpu_ptr; + long r; + + /* MES KIQ fw hasn't indirect buffer support for now */ + if (adev->enable_mes_kiq && + ring->funcs->type == AMDGPU_RING_TYPE_KIQ) + return 0; + + memset(&ib, 0, sizeof(ib)); + + if (ring->is_mes_queue) { + uint32_t padding, offset; + + offset = amdgpu_mes_ctx_get_offs(ring, AMDGPU_MES_CTX_IB_OFFS); + padding = amdgpu_mes_ctx_get_offs(ring, + AMDGPU_MES_CTX_PADDING_OFFS); + + ib.gpu_addr = amdgpu_mes_ctx_get_offs_gpu_addr(ring, offset); + ib.ptr = amdgpu_mes_ctx_get_offs_cpu_addr(ring, offset); + + gpu_addr = amdgpu_mes_ctx_get_offs_gpu_addr(ring, padding); + cpu_ptr = amdgpu_mes_ctx_get_offs_cpu_addr(ring, padding); + *cpu_ptr = cpu_to_le32(0xCAFEDEAD); + } else { + r = amdgpu_device_wb_get(adev, &index); + if (r) + return r; + + gpu_addr = adev->wb.gpu_addr + (index * 4); + adev->wb.wb[index] = cpu_to_le32(0xCAFEDEAD); + cpu_ptr = &adev->wb.wb[index]; + + r = amdgpu_ib_get(adev, NULL, 16, AMDGPU_IB_POOL_DIRECT, &ib); + if (r) { + DRM_ERROR("amdgpu: failed to get ib (%ld).\n", r); + goto err1; + } + } + + ib.ptr[0] = PACKET3(PACKET3_WRITE_DATA, 3); + ib.ptr[1] = WRITE_DATA_DST_SEL(5) | WR_CONFIRM; + ib.ptr[2] = lower_32_bits(gpu_addr); + ib.ptr[3] = upper_32_bits(gpu_addr); + ib.ptr[4] = 0xDEADBEEF; + ib.length_dw = 5; + + r = amdgpu_ib_schedule(ring, 1, &ib, NULL, &f); + if (r) + goto err2; + + r = dma_fence_wait_timeout(f, false, timeout); + if (r == 0) { + r = -ETIMEDOUT; + goto err2; + } else if (r < 0) { + goto err2; + } + + if (le32_to_cpu(*cpu_ptr) == 0xDEADBEEF) + r = 0; + else + r = -EINVAL; +err2: + if (!ring->is_mes_queue) + amdgpu_ib_free(adev, &ib, NULL); + dma_fence_put(f); +err1: + if (!ring->is_mes_queue) + amdgpu_device_wb_free(adev, index); + return r; +} + +static void gfx_v11_0_free_microcode(struct amdgpu_device *adev) +{ + release_firmware(adev->gfx.pfp_fw); + adev->gfx.pfp_fw = NULL; + release_firmware(adev->gfx.me_fw); + adev->gfx.me_fw = NULL; + release_firmware(adev->gfx.rlc_fw); + adev->gfx.rlc_fw = NULL; + release_firmware(adev->gfx.mec_fw); + adev->gfx.mec_fw = NULL; + + kfree(adev->gfx.rlc.register_list_format); +} + +static int gfx_v11_0_init_microcode(struct amdgpu_device *adev) +{ + char fw_name[40]; + char ucode_prefix[30]; + int err; + const struct rlc_firmware_header_v2_0 *rlc_hdr; + uint16_t version_major; + uint16_t version_minor; + + DRM_DEBUG("\n"); + + amdgpu_ucode_ip_version_decode(adev, GC_HWIP, ucode_prefix, sizeof(ucode_prefix)); + + snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_pfp.bin", ucode_prefix); + err = request_firmware(&adev->gfx.pfp_fw, fw_name, adev->dev); + if (err) + goto out; + err = amdgpu_ucode_validate(adev->gfx.pfp_fw); + if (err) + goto out; + /* check pfp fw hdr version to decide if enable rs64 for gfx11.*/ + adev->gfx.rs64_enable = amdgpu_ucode_hdr_version( + (union amdgpu_firmware_header *) + adev->gfx.pfp_fw->data, 2, 0); + if (adev->gfx.rs64_enable) { + dev_info(adev->dev, "CP RS64 enable\n"); + amdgpu_gfx_cp_init_microcode(adev, AMDGPU_UCODE_ID_CP_RS64_PFP); + amdgpu_gfx_cp_init_microcode(adev, AMDGPU_UCODE_ID_CP_RS64_PFP_P0_STACK); + amdgpu_gfx_cp_init_microcode(adev, AMDGPU_UCODE_ID_CP_RS64_PFP_P1_STACK); + } else { + amdgpu_gfx_cp_init_microcode(adev, AMDGPU_UCODE_ID_CP_PFP); + } + + snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_me.bin", ucode_prefix); + err = request_firmware(&adev->gfx.me_fw, fw_name, adev->dev); + if (err) + goto out; + err = amdgpu_ucode_validate(adev->gfx.me_fw); + if (err) + goto out; + if (adev->gfx.rs64_enable) { + amdgpu_gfx_cp_init_microcode(adev, AMDGPU_UCODE_ID_CP_RS64_ME); + amdgpu_gfx_cp_init_microcode(adev, AMDGPU_UCODE_ID_CP_RS64_ME_P0_STACK); + amdgpu_gfx_cp_init_microcode(adev, AMDGPU_UCODE_ID_CP_RS64_ME_P1_STACK); + } else { + amdgpu_gfx_cp_init_microcode(adev, AMDGPU_UCODE_ID_CP_ME); + } + + if (!amdgpu_sriov_vf(adev)) { + snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_rlc.bin", ucode_prefix); + err = request_firmware(&adev->gfx.rlc_fw, fw_name, adev->dev); + if (err) + goto out; + err = amdgpu_ucode_validate(adev->gfx.rlc_fw); + if (err) + goto out; + rlc_hdr = (const struct rlc_firmware_header_v2_0 *)adev->gfx.rlc_fw->data; + version_major = le16_to_cpu(rlc_hdr->header.header_version_major); + version_minor = le16_to_cpu(rlc_hdr->header.header_version_minor); + err = amdgpu_gfx_rlc_init_microcode(adev, version_major, version_minor); + if (err) + goto out; + } + + snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_mec.bin", ucode_prefix); + err = request_firmware(&adev->gfx.mec_fw, fw_name, adev->dev); + if (err) + goto out; + err = amdgpu_ucode_validate(adev->gfx.mec_fw); + if (err) + goto out; + if (adev->gfx.rs64_enable) { + amdgpu_gfx_cp_init_microcode(adev, AMDGPU_UCODE_ID_CP_RS64_MEC); + amdgpu_gfx_cp_init_microcode(adev, AMDGPU_UCODE_ID_CP_RS64_MEC_P0_STACK); + amdgpu_gfx_cp_init_microcode(adev, AMDGPU_UCODE_ID_CP_RS64_MEC_P1_STACK); + amdgpu_gfx_cp_init_microcode(adev, AMDGPU_UCODE_ID_CP_RS64_MEC_P2_STACK); + amdgpu_gfx_cp_init_microcode(adev, AMDGPU_UCODE_ID_CP_RS64_MEC_P3_STACK); + } else { + amdgpu_gfx_cp_init_microcode(adev, AMDGPU_UCODE_ID_CP_MEC1); + amdgpu_gfx_cp_init_microcode(adev, AMDGPU_UCODE_ID_CP_MEC1_JT); + } + + /* only one MEC for gfx 11.0.0. */ + adev->gfx.mec2_fw = NULL; + +out: + if (err) { + dev_err(adev->dev, + "gfx11: Failed to init firmware \"%s\"\n", + fw_name); + release_firmware(adev->gfx.pfp_fw); + adev->gfx.pfp_fw = NULL; + release_firmware(adev->gfx.me_fw); + adev->gfx.me_fw = NULL; + release_firmware(adev->gfx.rlc_fw); + adev->gfx.rlc_fw = NULL; + release_firmware(adev->gfx.mec_fw); + adev->gfx.mec_fw = NULL; + } + + return err; +} + +static int gfx_v11_0_init_toc_microcode(struct amdgpu_device *adev) +{ + const struct psp_firmware_header_v1_0 *toc_hdr; + int err = 0; + char fw_name[40]; + char ucode_prefix[30]; + + amdgpu_ucode_ip_version_decode(adev, GC_HWIP, ucode_prefix, sizeof(ucode_prefix)); + + snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_toc.bin", ucode_prefix); + err = request_firmware(&adev->psp.toc_fw, fw_name, adev->dev); + if (err) + goto out; + + err = amdgpu_ucode_validate(adev->psp.toc_fw); + if (err) + goto out; + + toc_hdr = (const struct psp_firmware_header_v1_0 *)adev->psp.toc_fw->data; + adev->psp.toc.fw_version = le32_to_cpu(toc_hdr->header.ucode_version); + adev->psp.toc.feature_version = le32_to_cpu(toc_hdr->sos.fw_version); + adev->psp.toc.size_bytes = le32_to_cpu(toc_hdr->header.ucode_size_bytes); + adev->psp.toc.start_addr = (uint8_t *)toc_hdr + + le32_to_cpu(toc_hdr->header.ucode_array_offset_bytes); + return 0; +out: + dev_err(adev->dev, "Failed to load TOC microcode\n"); + release_firmware(adev->psp.toc_fw); + adev->psp.toc_fw = NULL; + return err; +} + +static u32 gfx_v11_0_get_csb_size(struct amdgpu_device *adev) +{ + u32 count = 0; + const struct cs_section_def *sect = NULL; + const struct cs_extent_def *ext = NULL; + + /* begin clear state */ + count += 2; + /* context control state */ + count += 3; + + for (sect = gfx11_cs_data; sect->section != NULL; ++sect) { + for (ext = sect->section; ext->extent != NULL; ++ext) { + if (sect->id == SECT_CONTEXT) + count += 2 + ext->reg_count; + else + return 0; + } + } + + /* set PA_SC_TILE_STEERING_OVERRIDE */ + count += 3; + /* end clear state */ + count += 2; + /* clear state */ + count += 2; + + return count; +} + +static void gfx_v11_0_get_csb_buffer(struct amdgpu_device *adev, + volatile u32 *buffer) +{ + u32 count = 0, i; + const struct cs_section_def *sect = NULL; + const struct cs_extent_def *ext = NULL; + int ctx_reg_offset; + + if (adev->gfx.rlc.cs_data == NULL) + return; + if (buffer == NULL) + return; + + buffer[count++] = cpu_to_le32(PACKET3(PACKET3_PREAMBLE_CNTL, 0)); + buffer[count++] = cpu_to_le32(PACKET3_PREAMBLE_BEGIN_CLEAR_STATE); + + buffer[count++] = cpu_to_le32(PACKET3(PACKET3_CONTEXT_CONTROL, 1)); + buffer[count++] = cpu_to_le32(0x80000000); + buffer[count++] = cpu_to_le32(0x80000000); + + for (sect = adev->gfx.rlc.cs_data; sect->section != NULL; ++sect) { + for (ext = sect->section; ext->extent != NULL; ++ext) { + if (sect->id == SECT_CONTEXT) { + buffer[count++] = + cpu_to_le32(PACKET3(PACKET3_SET_CONTEXT_REG, ext->reg_count)); + buffer[count++] = cpu_to_le32(ext->reg_index - + PACKET3_SET_CONTEXT_REG_START); + for (i = 0; i < ext->reg_count; i++) + buffer[count++] = cpu_to_le32(ext->extent[i]); + } else { + return; + } + } + } + + ctx_reg_offset = + SOC15_REG_OFFSET(GC, 0, regPA_SC_TILE_STEERING_OVERRIDE) - PACKET3_SET_CONTEXT_REG_START; + buffer[count++] = cpu_to_le32(PACKET3(PACKET3_SET_CONTEXT_REG, 1)); + buffer[count++] = cpu_to_le32(ctx_reg_offset); + buffer[count++] = cpu_to_le32(adev->gfx.config.pa_sc_tile_steering_override); + + buffer[count++] = cpu_to_le32(PACKET3(PACKET3_PREAMBLE_CNTL, 0)); + buffer[count++] = cpu_to_le32(PACKET3_PREAMBLE_END_CLEAR_STATE); + + buffer[count++] = cpu_to_le32(PACKET3(PACKET3_CLEAR_STATE, 0)); + buffer[count++] = cpu_to_le32(0); +} + +static void gfx_v11_0_rlc_fini(struct amdgpu_device *adev) +{ + /* clear state block */ + amdgpu_bo_free_kernel(&adev->gfx.rlc.clear_state_obj, + &adev->gfx.rlc.clear_state_gpu_addr, + (void **)&adev->gfx.rlc.cs_ptr); + + /* jump table block */ + amdgpu_bo_free_kernel(&adev->gfx.rlc.cp_table_obj, + &adev->gfx.rlc.cp_table_gpu_addr, + (void **)&adev->gfx.rlc.cp_table_ptr); +} + +static void gfx_v11_0_init_rlcg_reg_access_ctrl(struct amdgpu_device *adev) +{ + struct amdgpu_rlcg_reg_access_ctrl *reg_access_ctrl; + + reg_access_ctrl = &adev->gfx.rlc.reg_access_ctrl; + reg_access_ctrl->scratch_reg0 = SOC15_REG_OFFSET(GC, 0, regSCRATCH_REG0); + reg_access_ctrl->scratch_reg1 = SOC15_REG_OFFSET(GC, 0, regSCRATCH_REG1); + reg_access_ctrl->scratch_reg2 = SOC15_REG_OFFSET(GC, 0, regSCRATCH_REG2); + reg_access_ctrl->scratch_reg3 = SOC15_REG_OFFSET(GC, 0, regSCRATCH_REG3); + reg_access_ctrl->grbm_cntl = SOC15_REG_OFFSET(GC, 0, regGRBM_GFX_CNTL); + reg_access_ctrl->grbm_idx = SOC15_REG_OFFSET(GC, 0, regGRBM_GFX_INDEX); + reg_access_ctrl->spare_int = SOC15_REG_OFFSET(GC, 0, regRLC_SPARE_INT_0); + adev->gfx.rlc.rlcg_reg_access_supported = true; +} + +static int gfx_v11_0_rlc_init(struct amdgpu_device *adev) +{ + const struct cs_section_def *cs_data; + int r; + + adev->gfx.rlc.cs_data = gfx11_cs_data; + + cs_data = adev->gfx.rlc.cs_data; + + if (cs_data) { + /* init clear state block */ + r = amdgpu_gfx_rlc_init_csb(adev); + if (r) + return r; + } + + /* init spm vmid with 0xf */ + if (adev->gfx.rlc.funcs->update_spm_vmid) + adev->gfx.rlc.funcs->update_spm_vmid(adev, 0xf); + + return 0; +} + +static void gfx_v11_0_mec_fini(struct amdgpu_device *adev) +{ + amdgpu_bo_free_kernel(&adev->gfx.mec.hpd_eop_obj, NULL, NULL); + amdgpu_bo_free_kernel(&adev->gfx.mec.mec_fw_obj, NULL, NULL); + amdgpu_bo_free_kernel(&adev->gfx.mec.mec_fw_data_obj, NULL, NULL); +} + +static int gfx_v11_0_me_init(struct amdgpu_device *adev) +{ + int r; + + bitmap_zero(adev->gfx.me.queue_bitmap, AMDGPU_MAX_GFX_QUEUES); + + amdgpu_gfx_graphics_queue_acquire(adev); + + r = gfx_v11_0_init_microcode(adev); + if (r) + DRM_ERROR("Failed to load gfx firmware!\n"); + + return r; +} + +static int gfx_v11_0_mec_init(struct amdgpu_device *adev) +{ + int r; + u32 *hpd; + size_t mec_hpd_size; + + bitmap_zero(adev->gfx.mec.queue_bitmap, AMDGPU_MAX_COMPUTE_QUEUES); + + /* take ownership of the relevant compute queues */ + amdgpu_gfx_compute_queue_acquire(adev); + mec_hpd_size = adev->gfx.num_compute_rings * GFX11_MEC_HPD_SIZE; + + if (mec_hpd_size) { + r = amdgpu_bo_create_reserved(adev, mec_hpd_size, PAGE_SIZE, + AMDGPU_GEM_DOMAIN_GTT, + &adev->gfx.mec.hpd_eop_obj, + &adev->gfx.mec.hpd_eop_gpu_addr, + (void **)&hpd); + if (r) { + dev_warn(adev->dev, "(%d) create HDP EOP bo failed\n", r); + gfx_v11_0_mec_fini(adev); + return r; + } + + memset(hpd, 0, mec_hpd_size); + + amdgpu_bo_kunmap(adev->gfx.mec.hpd_eop_obj); + amdgpu_bo_unreserve(adev->gfx.mec.hpd_eop_obj); + } + + return 0; +} + +static uint32_t wave_read_ind(struct amdgpu_device *adev, uint32_t wave, uint32_t address) +{ + WREG32_SOC15(GC, 0, regSQ_IND_INDEX, + (wave << SQ_IND_INDEX__WAVE_ID__SHIFT) | + (address << SQ_IND_INDEX__INDEX__SHIFT)); + return RREG32_SOC15(GC, 0, regSQ_IND_DATA); +} + +static void wave_read_regs(struct amdgpu_device *adev, uint32_t wave, + uint32_t thread, uint32_t regno, + uint32_t num, uint32_t *out) +{ + WREG32_SOC15(GC, 0, regSQ_IND_INDEX, + (wave << SQ_IND_INDEX__WAVE_ID__SHIFT) | + (regno << SQ_IND_INDEX__INDEX__SHIFT) | + (thread << SQ_IND_INDEX__WORKITEM_ID__SHIFT) | + (SQ_IND_INDEX__AUTO_INCR_MASK)); + while (num--) + *(out++) = RREG32_SOC15(GC, 0, regSQ_IND_DATA); +} + +static void gfx_v11_0_read_wave_data(struct amdgpu_device *adev, uint32_t simd, uint32_t wave, uint32_t *dst, int *no_fields) +{ + /* in gfx11 the SIMD_ID is specified as part of the INSTANCE + * field when performing a select_se_sh so it should be + * zero here */ + WARN_ON(simd != 0); + + /* type 2 wave data */ + dst[(*no_fields)++] = 2; + dst[(*no_fields)++] = wave_read_ind(adev, wave, ixSQ_WAVE_STATUS); + dst[(*no_fields)++] = wave_read_ind(adev, wave, ixSQ_WAVE_PC_LO); + dst[(*no_fields)++] = wave_read_ind(adev, wave, ixSQ_WAVE_PC_HI); + dst[(*no_fields)++] = wave_read_ind(adev, wave, ixSQ_WAVE_EXEC_LO); + dst[(*no_fields)++] = wave_read_ind(adev, wave, ixSQ_WAVE_EXEC_HI); + dst[(*no_fields)++] = wave_read_ind(adev, wave, ixSQ_WAVE_HW_ID1); + dst[(*no_fields)++] = wave_read_ind(adev, wave, ixSQ_WAVE_HW_ID2); + dst[(*no_fields)++] = wave_read_ind(adev, wave, ixSQ_WAVE_GPR_ALLOC); + dst[(*no_fields)++] = wave_read_ind(adev, wave, ixSQ_WAVE_LDS_ALLOC); + dst[(*no_fields)++] = wave_read_ind(adev, wave, ixSQ_WAVE_TRAPSTS); + dst[(*no_fields)++] = wave_read_ind(adev, wave, ixSQ_WAVE_IB_STS); + dst[(*no_fields)++] = wave_read_ind(adev, wave, ixSQ_WAVE_IB_STS2); + dst[(*no_fields)++] = wave_read_ind(adev, wave, ixSQ_WAVE_IB_DBG1); + dst[(*no_fields)++] = wave_read_ind(adev, wave, ixSQ_WAVE_M0); + dst[(*no_fields)++] = wave_read_ind(adev, wave, ixSQ_WAVE_MODE); +} + +static void gfx_v11_0_read_wave_sgprs(struct amdgpu_device *adev, uint32_t simd, + uint32_t wave, uint32_t start, + uint32_t size, uint32_t *dst) +{ + WARN_ON(simd != 0); + + wave_read_regs( + adev, wave, 0, start + SQIND_WAVE_SGPRS_OFFSET, size, + dst); +} + +static void gfx_v11_0_read_wave_vgprs(struct amdgpu_device *adev, uint32_t simd, + uint32_t wave, uint32_t thread, + uint32_t start, uint32_t size, + uint32_t *dst) +{ + wave_read_regs( + adev, wave, thread, + start + SQIND_WAVE_VGPRS_OFFSET, size, dst); +} + +static void gfx_v11_0_select_me_pipe_q(struct amdgpu_device *adev, + u32 me, u32 pipe, u32 q, u32 vm) +{ + soc21_grbm_select(adev, me, pipe, q, vm); +} + +static const struct amdgpu_gfx_funcs gfx_v11_0_gfx_funcs = { + .get_gpu_clock_counter = &gfx_v11_0_get_gpu_clock_counter, + .select_se_sh = &gfx_v11_0_select_se_sh, + .read_wave_data = &gfx_v11_0_read_wave_data, + .read_wave_sgprs = &gfx_v11_0_read_wave_sgprs, + .read_wave_vgprs = &gfx_v11_0_read_wave_vgprs, + .select_me_pipe_q = &gfx_v11_0_select_me_pipe_q, + .update_perfmon_mgcg = &gfx_v11_0_update_perf_clk, +}; + +static int gfx_v11_0_gpu_early_init(struct amdgpu_device *adev) +{ + adev->gfx.funcs = &gfx_v11_0_gfx_funcs; + + switch (adev->ip_versions[GC_HWIP][0]) { + case IP_VERSION(11, 0, 0): + case IP_VERSION(11, 0, 2): + case IP_VERSION(11, 0, 3): + adev->gfx.config.max_hw_contexts = 8; + adev->gfx.config.sc_prim_fifo_size_frontend = 0x20; + adev->gfx.config.sc_prim_fifo_size_backend = 0x100; + adev->gfx.config.sc_hiz_tile_fifo_size = 0; + adev->gfx.config.sc_earlyz_tile_fifo_size = 0x4C0; + break; + case IP_VERSION(11, 0, 1): + adev->gfx.config.max_hw_contexts = 8; + adev->gfx.config.sc_prim_fifo_size_frontend = 0x20; + adev->gfx.config.sc_prim_fifo_size_backend = 0x100; + adev->gfx.config.sc_hiz_tile_fifo_size = 0x80; + adev->gfx.config.sc_earlyz_tile_fifo_size = 0x300; + break; + default: + BUG(); + break; + } + + return 0; +} + +static int gfx_v11_0_gfx_ring_init(struct amdgpu_device *adev, int ring_id, + int me, int pipe, int queue) +{ + int r; + struct amdgpu_ring *ring; + unsigned int irq_type; + + ring = &adev->gfx.gfx_ring[ring_id]; + + ring->me = me; + ring->pipe = pipe; + ring->queue = queue; + + ring->ring_obj = NULL; + ring->use_doorbell = true; + + if (!ring_id) + ring->doorbell_index = adev->doorbell_index.gfx_ring0 << 1; + else + ring->doorbell_index = adev->doorbell_index.gfx_ring1 << 1; + sprintf(ring->name, "gfx_%d.%d.%d", ring->me, ring->pipe, ring->queue); + + irq_type = AMDGPU_CP_IRQ_GFX_ME0_PIPE0_EOP + ring->pipe; + r = amdgpu_ring_init(adev, ring, 1024, &adev->gfx.eop_irq, irq_type, + AMDGPU_RING_PRIO_DEFAULT, NULL); + if (r) + return r; + return 0; +} + +static int gfx_v11_0_compute_ring_init(struct amdgpu_device *adev, int ring_id, + int mec, int pipe, int queue) +{ + int r; + unsigned irq_type; + struct amdgpu_ring *ring; + unsigned int hw_prio; + + ring = &adev->gfx.compute_ring[ring_id]; + + /* mec0 is me1 */ + ring->me = mec + 1; + ring->pipe = pipe; + ring->queue = queue; + + ring->ring_obj = NULL; + ring->use_doorbell = true; + ring->doorbell_index = (adev->doorbell_index.mec_ring0 + ring_id) << 1; + ring->eop_gpu_addr = adev->gfx.mec.hpd_eop_gpu_addr + + (ring_id * GFX11_MEC_HPD_SIZE); + sprintf(ring->name, "comp_%d.%d.%d", ring->me, ring->pipe, ring->queue); + + irq_type = AMDGPU_CP_IRQ_COMPUTE_MEC1_PIPE0_EOP + + ((ring->me - 1) * adev->gfx.mec.num_pipe_per_mec) + + ring->pipe; + hw_prio = amdgpu_gfx_is_high_priority_compute_queue(adev, ring) ? + AMDGPU_GFX_PIPE_PRIO_HIGH : AMDGPU_GFX_PIPE_PRIO_NORMAL; + /* type-2 packets are deprecated on MEC, use type-3 instead */ + r = amdgpu_ring_init(adev, ring, 1024, &adev->gfx.eop_irq, irq_type, + hw_prio, NULL); + if (r) + return r; + + return 0; +} + +static struct { + SOC21_FIRMWARE_ID id; + unsigned int offset; + unsigned int size; +} rlc_autoload_info[SOC21_FIRMWARE_ID_MAX]; + +static void gfx_v11_0_parse_rlc_toc(struct amdgpu_device *adev, void *rlc_toc) +{ + RLC_TABLE_OF_CONTENT *ucode = rlc_toc; + + while (ucode && (ucode->id > SOC21_FIRMWARE_ID_INVALID) && + (ucode->id < SOC21_FIRMWARE_ID_MAX)) { + rlc_autoload_info[ucode->id].id = ucode->id; + rlc_autoload_info[ucode->id].offset = ucode->offset * 4; + rlc_autoload_info[ucode->id].size = ucode->size * 4; + + ucode++; + } +} + +static uint32_t gfx_v11_0_calc_toc_total_size(struct amdgpu_device *adev) +{ + uint32_t total_size = 0; + SOC21_FIRMWARE_ID id; + + gfx_v11_0_parse_rlc_toc(adev, adev->psp.toc.start_addr); + + for (id = SOC21_FIRMWARE_ID_RLC_G_UCODE; id < SOC21_FIRMWARE_ID_MAX; id++) + total_size += rlc_autoload_info[id].size; + + /* In case the offset in rlc toc ucode is aligned */ + if (total_size < rlc_autoload_info[SOC21_FIRMWARE_ID_MAX-1].offset) + total_size = rlc_autoload_info[SOC21_FIRMWARE_ID_MAX-1].offset + + rlc_autoload_info[SOC21_FIRMWARE_ID_MAX-1].size; + + return total_size; +} + +static int gfx_v11_0_rlc_autoload_buffer_init(struct amdgpu_device *adev) +{ + int r; + uint32_t total_size; + + total_size = gfx_v11_0_calc_toc_total_size(adev); + + r = amdgpu_bo_create_reserved(adev, total_size, 64 * 1024, + AMDGPU_GEM_DOMAIN_VRAM, + &adev->gfx.rlc.rlc_autoload_bo, + &adev->gfx.rlc.rlc_autoload_gpu_addr, + (void **)&adev->gfx.rlc.rlc_autoload_ptr); + + if (r) { + dev_err(adev->dev, "(%d) failed to create fw autoload bo\n", r); + return r; + } + + return 0; +} + +static void gfx_v11_0_rlc_backdoor_autoload_copy_ucode(struct amdgpu_device *adev, + SOC21_FIRMWARE_ID id, + const void *fw_data, + uint32_t fw_size, + uint32_t *fw_autoload_mask) +{ + uint32_t toc_offset; + uint32_t toc_fw_size; + char *ptr = adev->gfx.rlc.rlc_autoload_ptr; + + if (id <= SOC21_FIRMWARE_ID_INVALID || id >= SOC21_FIRMWARE_ID_MAX) + return; + + toc_offset = rlc_autoload_info[id].offset; + toc_fw_size = rlc_autoload_info[id].size; + + if (fw_size == 0) + fw_size = toc_fw_size; + + if (fw_size > toc_fw_size) + fw_size = toc_fw_size; + + memcpy(ptr + toc_offset, fw_data, fw_size); + + if (fw_size < toc_fw_size) + memset(ptr + toc_offset + fw_size, 0, toc_fw_size - fw_size); + + if ((id != SOC21_FIRMWARE_ID_RS64_PFP) && (id != SOC21_FIRMWARE_ID_RS64_ME)) + *(uint64_t *)fw_autoload_mask |= 1ULL << id; +} + +static void gfx_v11_0_rlc_backdoor_autoload_copy_toc_ucode(struct amdgpu_device *adev, + uint32_t *fw_autoload_mask) +{ + void *data; + uint32_t size; + uint64_t *toc_ptr; + + *(uint64_t *)fw_autoload_mask |= 0x1; + + DRM_DEBUG("rlc autoload enabled fw: 0x%llx\n", *(uint64_t *)fw_autoload_mask); + + data = adev->psp.toc.start_addr; + size = rlc_autoload_info[SOC21_FIRMWARE_ID_RLC_TOC].size; + + toc_ptr = (uint64_t *)data + size / 8 - 1; + *toc_ptr = *(uint64_t *)fw_autoload_mask; + + gfx_v11_0_rlc_backdoor_autoload_copy_ucode(adev, SOC21_FIRMWARE_ID_RLC_TOC, + data, size, fw_autoload_mask); +} + +static void gfx_v11_0_rlc_backdoor_autoload_copy_gfx_ucode(struct amdgpu_device *adev, + uint32_t *fw_autoload_mask) +{ + const __le32 *fw_data; + uint32_t fw_size; + const struct gfx_firmware_header_v1_0 *cp_hdr; + const struct gfx_firmware_header_v2_0 *cpv2_hdr; + const struct rlc_firmware_header_v2_0 *rlc_hdr; + const struct rlc_firmware_header_v2_2 *rlcv22_hdr; + uint16_t version_major, version_minor; + + if (adev->gfx.rs64_enable) { + /* pfp ucode */ + cpv2_hdr = (const struct gfx_firmware_header_v2_0 *) + adev->gfx.pfp_fw->data; + /* instruction */ + fw_data = (const __le32 *)(adev->gfx.pfp_fw->data + + le32_to_cpu(cpv2_hdr->ucode_offset_bytes)); + fw_size = le32_to_cpu(cpv2_hdr->ucode_size_bytes); + gfx_v11_0_rlc_backdoor_autoload_copy_ucode(adev, SOC21_FIRMWARE_ID_RS64_PFP, + fw_data, fw_size, fw_autoload_mask); + /* data */ + fw_data = (const __le32 *)(adev->gfx.pfp_fw->data + + le32_to_cpu(cpv2_hdr->data_offset_bytes)); + fw_size = le32_to_cpu(cpv2_hdr->data_size_bytes); + gfx_v11_0_rlc_backdoor_autoload_copy_ucode(adev, SOC21_FIRMWARE_ID_RS64_PFP_P0_STACK, + fw_data, fw_size, fw_autoload_mask); + gfx_v11_0_rlc_backdoor_autoload_copy_ucode(adev, SOC21_FIRMWARE_ID_RS64_PFP_P1_STACK, + fw_data, fw_size, fw_autoload_mask); + /* me ucode */ + cpv2_hdr = (const struct gfx_firmware_header_v2_0 *) + adev->gfx.me_fw->data; + /* instruction */ + fw_data = (const __le32 *)(adev->gfx.me_fw->data + + le32_to_cpu(cpv2_hdr->ucode_offset_bytes)); + fw_size = le32_to_cpu(cpv2_hdr->ucode_size_bytes); + gfx_v11_0_rlc_backdoor_autoload_copy_ucode(adev, SOC21_FIRMWARE_ID_RS64_ME, + fw_data, fw_size, fw_autoload_mask); + /* data */ + fw_data = (const __le32 *)(adev->gfx.me_fw->data + + le32_to_cpu(cpv2_hdr->data_offset_bytes)); + fw_size = le32_to_cpu(cpv2_hdr->data_size_bytes); + gfx_v11_0_rlc_backdoor_autoload_copy_ucode(adev, SOC21_FIRMWARE_ID_RS64_ME_P0_STACK, + fw_data, fw_size, fw_autoload_mask); + gfx_v11_0_rlc_backdoor_autoload_copy_ucode(adev, SOC21_FIRMWARE_ID_RS64_ME_P1_STACK, + fw_data, fw_size, fw_autoload_mask); + /* mec ucode */ + cpv2_hdr = (const struct gfx_firmware_header_v2_0 *) + adev->gfx.mec_fw->data; + /* instruction */ + fw_data = (const __le32 *) (adev->gfx.mec_fw->data + + le32_to_cpu(cpv2_hdr->ucode_offset_bytes)); + fw_size = le32_to_cpu(cpv2_hdr->ucode_size_bytes); + gfx_v11_0_rlc_backdoor_autoload_copy_ucode(adev, SOC21_FIRMWARE_ID_RS64_MEC, + fw_data, fw_size, fw_autoload_mask); + /* data */ + fw_data = (const __le32 *) (adev->gfx.mec_fw->data + + le32_to_cpu(cpv2_hdr->data_offset_bytes)); + fw_size = le32_to_cpu(cpv2_hdr->data_size_bytes); + gfx_v11_0_rlc_backdoor_autoload_copy_ucode(adev, SOC21_FIRMWARE_ID_RS64_MEC_P0_STACK, + fw_data, fw_size, fw_autoload_mask); + gfx_v11_0_rlc_backdoor_autoload_copy_ucode(adev, SOC21_FIRMWARE_ID_RS64_MEC_P1_STACK, + fw_data, fw_size, fw_autoload_mask); + gfx_v11_0_rlc_backdoor_autoload_copy_ucode(adev, SOC21_FIRMWARE_ID_RS64_MEC_P2_STACK, + fw_data, fw_size, fw_autoload_mask); + gfx_v11_0_rlc_backdoor_autoload_copy_ucode(adev, SOC21_FIRMWARE_ID_RS64_MEC_P3_STACK, + fw_data, fw_size, fw_autoload_mask); + } else { + /* pfp ucode */ + cp_hdr = (const struct gfx_firmware_header_v1_0 *) + adev->gfx.pfp_fw->data; + fw_data = (const __le32 *)(adev->gfx.pfp_fw->data + + le32_to_cpu(cp_hdr->header.ucode_array_offset_bytes)); + fw_size = le32_to_cpu(cp_hdr->header.ucode_size_bytes); + gfx_v11_0_rlc_backdoor_autoload_copy_ucode(adev, SOC21_FIRMWARE_ID_CP_PFP, + fw_data, fw_size, fw_autoload_mask); + + /* me ucode */ + cp_hdr = (const struct gfx_firmware_header_v1_0 *) + adev->gfx.me_fw->data; + fw_data = (const __le32 *)(adev->gfx.me_fw->data + + le32_to_cpu(cp_hdr->header.ucode_array_offset_bytes)); + fw_size = le32_to_cpu(cp_hdr->header.ucode_size_bytes); + gfx_v11_0_rlc_backdoor_autoload_copy_ucode(adev, SOC21_FIRMWARE_ID_CP_ME, + fw_data, fw_size, fw_autoload_mask); + + /* mec ucode */ + cp_hdr = (const struct gfx_firmware_header_v1_0 *) + adev->gfx.mec_fw->data; + fw_data = (const __le32 *) (adev->gfx.mec_fw->data + + le32_to_cpu(cp_hdr->header.ucode_array_offset_bytes)); + fw_size = le32_to_cpu(cp_hdr->header.ucode_size_bytes) - + cp_hdr->jt_size * 4; + gfx_v11_0_rlc_backdoor_autoload_copy_ucode(adev, SOC21_FIRMWARE_ID_CP_MEC, + fw_data, fw_size, fw_autoload_mask); + } + + /* rlc ucode */ + rlc_hdr = (const struct rlc_firmware_header_v2_0 *) + adev->gfx.rlc_fw->data; + fw_data = (const __le32 *)(adev->gfx.rlc_fw->data + + le32_to_cpu(rlc_hdr->header.ucode_array_offset_bytes)); + fw_size = le32_to_cpu(rlc_hdr->header.ucode_size_bytes); + gfx_v11_0_rlc_backdoor_autoload_copy_ucode(adev, SOC21_FIRMWARE_ID_RLC_G_UCODE, + fw_data, fw_size, fw_autoload_mask); + + version_major = le16_to_cpu(rlc_hdr->header.header_version_major); + version_minor = le16_to_cpu(rlc_hdr->header.header_version_minor); + if (version_major == 2) { + if (version_minor >= 2) { + rlcv22_hdr = (const struct rlc_firmware_header_v2_2 *)adev->gfx.rlc_fw->data; + + fw_data = (const __le32 *)(adev->gfx.rlc_fw->data + + le32_to_cpu(rlcv22_hdr->rlc_iram_ucode_offset_bytes)); + fw_size = le32_to_cpu(rlcv22_hdr->rlc_iram_ucode_size_bytes); + gfx_v11_0_rlc_backdoor_autoload_copy_ucode(adev, SOC21_FIRMWARE_ID_RLX6_UCODE, + fw_data, fw_size, fw_autoload_mask); + + fw_data = (const __le32 *)(adev->gfx.rlc_fw->data + + le32_to_cpu(rlcv22_hdr->rlc_dram_ucode_offset_bytes)); + fw_size = le32_to_cpu(rlcv22_hdr->rlc_dram_ucode_size_bytes); + gfx_v11_0_rlc_backdoor_autoload_copy_ucode(adev, SOC21_FIRMWARE_ID_RLX6_DRAM_BOOT, + fw_data, fw_size, fw_autoload_mask); + } + } +} + +static void gfx_v11_0_rlc_backdoor_autoload_copy_sdma_ucode(struct amdgpu_device *adev, + uint32_t *fw_autoload_mask) +{ + const __le32 *fw_data; + uint32_t fw_size; + const struct sdma_firmware_header_v2_0 *sdma_hdr; + + sdma_hdr = (const struct sdma_firmware_header_v2_0 *) + adev->sdma.instance[0].fw->data; + fw_data = (const __le32 *) (adev->sdma.instance[0].fw->data + + le32_to_cpu(sdma_hdr->header.ucode_array_offset_bytes)); + fw_size = le32_to_cpu(sdma_hdr->ctx_ucode_size_bytes); + + gfx_v11_0_rlc_backdoor_autoload_copy_ucode(adev, + SOC21_FIRMWARE_ID_SDMA_UCODE_TH0, fw_data, fw_size, fw_autoload_mask); + + fw_data = (const __le32 *) (adev->sdma.instance[0].fw->data + + le32_to_cpu(sdma_hdr->ctl_ucode_offset)); + fw_size = le32_to_cpu(sdma_hdr->ctl_ucode_size_bytes); + + gfx_v11_0_rlc_backdoor_autoload_copy_ucode(adev, + SOC21_FIRMWARE_ID_SDMA_UCODE_TH1, fw_data, fw_size, fw_autoload_mask); +} + +static void gfx_v11_0_rlc_backdoor_autoload_copy_mes_ucode(struct amdgpu_device *adev, + uint32_t *fw_autoload_mask) +{ + const __le32 *fw_data; + unsigned fw_size; + const struct mes_firmware_header_v1_0 *mes_hdr; + int pipe, ucode_id, data_id; + + for (pipe = 0; pipe < 2; pipe++) { + if (pipe==0) { + ucode_id = SOC21_FIRMWARE_ID_RS64_MES_P0; + data_id = SOC21_FIRMWARE_ID_RS64_MES_P0_STACK; + } else { + ucode_id = SOC21_FIRMWARE_ID_RS64_MES_P1; + data_id = SOC21_FIRMWARE_ID_RS64_MES_P1_STACK; + } + + mes_hdr = (const struct mes_firmware_header_v1_0 *) + adev->mes.fw[pipe]->data; + + fw_data = (const __le32 *)(adev->mes.fw[pipe]->data + + le32_to_cpu(mes_hdr->mes_ucode_offset_bytes)); + fw_size = le32_to_cpu(mes_hdr->mes_ucode_size_bytes); + + gfx_v11_0_rlc_backdoor_autoload_copy_ucode(adev, + ucode_id, fw_data, fw_size, fw_autoload_mask); + + fw_data = (const __le32 *)(adev->mes.fw[pipe]->data + + le32_to_cpu(mes_hdr->mes_ucode_data_offset_bytes)); + fw_size = le32_to_cpu(mes_hdr->mes_ucode_data_size_bytes); + + gfx_v11_0_rlc_backdoor_autoload_copy_ucode(adev, + data_id, fw_data, fw_size, fw_autoload_mask); + } +} + +static int gfx_v11_0_rlc_backdoor_autoload_enable(struct amdgpu_device *adev) +{ + uint32_t rlc_g_offset, rlc_g_size; + uint64_t gpu_addr; + uint32_t autoload_fw_id[2]; + + memset(autoload_fw_id, 0, sizeof(uint32_t) * 2); + + /* RLC autoload sequence 2: copy ucode */ + gfx_v11_0_rlc_backdoor_autoload_copy_sdma_ucode(adev, autoload_fw_id); + gfx_v11_0_rlc_backdoor_autoload_copy_gfx_ucode(adev, autoload_fw_id); + gfx_v11_0_rlc_backdoor_autoload_copy_mes_ucode(adev, autoload_fw_id); + gfx_v11_0_rlc_backdoor_autoload_copy_toc_ucode(adev, autoload_fw_id); + + rlc_g_offset = rlc_autoload_info[SOC21_FIRMWARE_ID_RLC_G_UCODE].offset; + rlc_g_size = rlc_autoload_info[SOC21_FIRMWARE_ID_RLC_G_UCODE].size; + gpu_addr = adev->gfx.rlc.rlc_autoload_gpu_addr + rlc_g_offset; + + WREG32_SOC15(GC, 0, regGFX_IMU_RLC_BOOTLOADER_ADDR_HI, upper_32_bits(gpu_addr)); + WREG32_SOC15(GC, 0, regGFX_IMU_RLC_BOOTLOADER_ADDR_LO, lower_32_bits(gpu_addr)); + + WREG32_SOC15(GC, 0, regGFX_IMU_RLC_BOOTLOADER_SIZE, rlc_g_size); + + /* RLC autoload sequence 3: load IMU fw */ + if (adev->gfx.imu.funcs->load_microcode) + adev->gfx.imu.funcs->load_microcode(adev); + /* RLC autoload sequence 4 init IMU fw */ + if (adev->gfx.imu.funcs->setup_imu) + adev->gfx.imu.funcs->setup_imu(adev); + if (adev->gfx.imu.funcs->start_imu) + adev->gfx.imu.funcs->start_imu(adev); + + /* RLC autoload sequence 5 disable gpa mode */ + gfx_v11_0_disable_gpa_mode(adev); + + return 0; +} + +static int gfx_v11_0_sw_init(void *handle) +{ + int i, j, k, r, ring_id = 0; + struct amdgpu_kiq *kiq; + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + + adev->gfxhub.funcs->init(adev); + + switch (adev->ip_versions[GC_HWIP][0]) { + case IP_VERSION(11, 0, 0): + case IP_VERSION(11, 0, 1): + case IP_VERSION(11, 0, 2): + case IP_VERSION(11, 0, 3): + adev->gfx.me.num_me = 1; + adev->gfx.me.num_pipe_per_me = 1; + adev->gfx.me.num_queue_per_pipe = 1; + adev->gfx.mec.num_mec = 2; + adev->gfx.mec.num_pipe_per_mec = 4; + adev->gfx.mec.num_queue_per_pipe = 4; + break; + default: + adev->gfx.me.num_me = 1; + adev->gfx.me.num_pipe_per_me = 1; + adev->gfx.me.num_queue_per_pipe = 1; + adev->gfx.mec.num_mec = 1; + adev->gfx.mec.num_pipe_per_mec = 4; + adev->gfx.mec.num_queue_per_pipe = 8; + break; + } + + /* EOP Event */ + r = amdgpu_irq_add_id(adev, SOC21_IH_CLIENTID_GRBM_CP, + GFX_11_0_0__SRCID__CP_EOP_INTERRUPT, + &adev->gfx.eop_irq); + if (r) + return r; + + /* Privileged reg */ + r = amdgpu_irq_add_id(adev, SOC21_IH_CLIENTID_GRBM_CP, + GFX_11_0_0__SRCID__CP_PRIV_REG_FAULT, + &adev->gfx.priv_reg_irq); + if (r) + return r; + + /* Privileged inst */ + r = amdgpu_irq_add_id(adev, SOC21_IH_CLIENTID_GRBM_CP, + GFX_11_0_0__SRCID__CP_PRIV_INSTR_FAULT, + &adev->gfx.priv_inst_irq); + if (r) + return r; + + adev->gfx.gfx_current_status = AMDGPU_GFX_NORMAL_MODE; + + if (adev->gfx.imu.funcs) { + if (adev->gfx.imu.funcs->init_microcode) { + r = adev->gfx.imu.funcs->init_microcode(adev); + if (r) + DRM_ERROR("Failed to load imu firmware!\n"); + } + } + + r = gfx_v11_0_me_init(adev); + if (r) + return r; + + r = gfx_v11_0_rlc_init(adev); + if (r) { + DRM_ERROR("Failed to init rlc BOs!\n"); + return r; + } + + r = gfx_v11_0_mec_init(adev); + if (r) { + DRM_ERROR("Failed to init MEC BOs!\n"); + return r; + } + + /* set up the gfx ring */ + for (i = 0; i < adev->gfx.me.num_me; i++) { + for (j = 0; j < adev->gfx.me.num_queue_per_pipe; j++) { + for (k = 0; k < adev->gfx.me.num_pipe_per_me; k++) { + if (!amdgpu_gfx_is_me_queue_enabled(adev, i, k, j)) + continue; + + r = gfx_v11_0_gfx_ring_init(adev, ring_id, + i, k, j); + if (r) + return r; + ring_id++; + } + } + } + + ring_id = 0; + /* set up the compute queues - allocate horizontally across pipes */ + for (i = 0; i < adev->gfx.mec.num_mec; ++i) { + for (j = 0; j < adev->gfx.mec.num_queue_per_pipe; j++) { + for (k = 0; k < adev->gfx.mec.num_pipe_per_mec; k++) { + if (!amdgpu_gfx_is_mec_queue_enabled(adev, i, k, + j)) + continue; + + r = gfx_v11_0_compute_ring_init(adev, ring_id, + i, k, j); + if (r) + return r; + + ring_id++; + } + } + } + + if (!adev->enable_mes_kiq) { + r = amdgpu_gfx_kiq_init(adev, GFX11_MEC_HPD_SIZE); + if (r) { + DRM_ERROR("Failed to init KIQ BOs!\n"); + return r; + } + + kiq = &adev->gfx.kiq; + r = amdgpu_gfx_kiq_init_ring(adev, &kiq->ring, &kiq->irq); + if (r) + return r; + } + + r = amdgpu_gfx_mqd_sw_init(adev, sizeof(struct v11_compute_mqd)); + if (r) + return r; + + /* allocate visible FB for rlc auto-loading fw */ + if (adev->firmware.load_type == AMDGPU_FW_LOAD_RLC_BACKDOOR_AUTO) { + r = gfx_v11_0_init_toc_microcode(adev); + if (r) + dev_err(adev->dev, "Failed to load toc firmware!\n"); + r = gfx_v11_0_rlc_autoload_buffer_init(adev); + if (r) + return r; + } + + r = gfx_v11_0_gpu_early_init(adev); + if (r) + return r; + + return 0; +} + +static void gfx_v11_0_pfp_fini(struct amdgpu_device *adev) +{ + amdgpu_bo_free_kernel(&adev->gfx.pfp.pfp_fw_obj, + &adev->gfx.pfp.pfp_fw_gpu_addr, + (void **)&adev->gfx.pfp.pfp_fw_ptr); + + amdgpu_bo_free_kernel(&adev->gfx.pfp.pfp_fw_data_obj, + &adev->gfx.pfp.pfp_fw_data_gpu_addr, + (void **)&adev->gfx.pfp.pfp_fw_data_ptr); +} + +static void gfx_v11_0_me_fini(struct amdgpu_device *adev) +{ + amdgpu_bo_free_kernel(&adev->gfx.me.me_fw_obj, + &adev->gfx.me.me_fw_gpu_addr, + (void **)&adev->gfx.me.me_fw_ptr); + + amdgpu_bo_free_kernel(&adev->gfx.me.me_fw_data_obj, + &adev->gfx.me.me_fw_data_gpu_addr, + (void **)&adev->gfx.me.me_fw_data_ptr); +} + +static void gfx_v11_0_rlc_autoload_buffer_fini(struct amdgpu_device *adev) +{ + amdgpu_bo_free_kernel(&adev->gfx.rlc.rlc_autoload_bo, + &adev->gfx.rlc.rlc_autoload_gpu_addr, + (void **)&adev->gfx.rlc.rlc_autoload_ptr); +} + +static int gfx_v11_0_sw_fini(void *handle) +{ + int i; + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + + for (i = 0; i < adev->gfx.num_gfx_rings; i++) + amdgpu_ring_fini(&adev->gfx.gfx_ring[i]); + for (i = 0; i < adev->gfx.num_compute_rings; i++) + amdgpu_ring_fini(&adev->gfx.compute_ring[i]); + + amdgpu_gfx_mqd_sw_fini(adev); + + if (!adev->enable_mes_kiq) { + amdgpu_gfx_kiq_free_ring(&adev->gfx.kiq.ring); + amdgpu_gfx_kiq_fini(adev); + } + + gfx_v11_0_pfp_fini(adev); + gfx_v11_0_me_fini(adev); + gfx_v11_0_rlc_fini(adev); + gfx_v11_0_mec_fini(adev); + + if (adev->firmware.load_type == AMDGPU_FW_LOAD_RLC_BACKDOOR_AUTO) + gfx_v11_0_rlc_autoload_buffer_fini(adev); + + gfx_v11_0_free_microcode(adev); + + return 0; +} + +static void gfx_v11_0_select_se_sh(struct amdgpu_device *adev, u32 se_num, + u32 sh_num, u32 instance) +{ + u32 data; + + if (instance == 0xffffffff) + data = REG_SET_FIELD(0, GRBM_GFX_INDEX, + INSTANCE_BROADCAST_WRITES, 1); + else + data = REG_SET_FIELD(0, GRBM_GFX_INDEX, INSTANCE_INDEX, + instance); + + if (se_num == 0xffffffff) + data = REG_SET_FIELD(data, GRBM_GFX_INDEX, SE_BROADCAST_WRITES, + 1); + else + data = REG_SET_FIELD(data, GRBM_GFX_INDEX, SE_INDEX, se_num); + + if (sh_num == 0xffffffff) + data = REG_SET_FIELD(data, GRBM_GFX_INDEX, SA_BROADCAST_WRITES, + 1); + else + data = REG_SET_FIELD(data, GRBM_GFX_INDEX, SA_INDEX, sh_num); + + WREG32_SOC15(GC, 0, regGRBM_GFX_INDEX, data); +} + +static u32 gfx_v11_0_get_rb_active_bitmap(struct amdgpu_device *adev) +{ + u32 data, mask; + + data = RREG32_SOC15(GC, 0, regCC_RB_BACKEND_DISABLE); + data |= RREG32_SOC15(GC, 0, regGC_USER_RB_BACKEND_DISABLE); + + data &= CC_RB_BACKEND_DISABLE__BACKEND_DISABLE_MASK; + data >>= GC_USER_RB_BACKEND_DISABLE__BACKEND_DISABLE__SHIFT; + + mask = amdgpu_gfx_create_bitmask(adev->gfx.config.max_backends_per_se / + adev->gfx.config.max_sh_per_se); + + return (~data) & mask; +} + +static void gfx_v11_0_setup_rb(struct amdgpu_device *adev) +{ + int i, j; + u32 data; + u32 active_rbs = 0; + u32 rb_bitmap_width_per_sh = adev->gfx.config.max_backends_per_se / + adev->gfx.config.max_sh_per_se; + + mutex_lock(&adev->grbm_idx_mutex); + for (i = 0; i < adev->gfx.config.max_shader_engines; i++) { + for (j = 0; j < adev->gfx.config.max_sh_per_se; j++) { + gfx_v11_0_select_se_sh(adev, i, j, 0xffffffff); + data = gfx_v11_0_get_rb_active_bitmap(adev); + active_rbs |= data << ((i * adev->gfx.config.max_sh_per_se + j) * + rb_bitmap_width_per_sh); + } + } + gfx_v11_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff); + mutex_unlock(&adev->grbm_idx_mutex); + + adev->gfx.config.backend_enable_mask = active_rbs; + adev->gfx.config.num_rbs = hweight32(active_rbs); +} + +#define DEFAULT_SH_MEM_BASES (0x6000) +#define LDS_APP_BASE 0x1 +#define SCRATCH_APP_BASE 0x2 + +static void gfx_v11_0_init_compute_vmid(struct amdgpu_device *adev) +{ + int i; + uint32_t sh_mem_bases; + uint32_t data; + + /* + * Configure apertures: + * LDS: 0x60000000'00000000 - 0x60000001'00000000 (4GB) + * Scratch: 0x60000001'00000000 - 0x60000002'00000000 (4GB) + * GPUVM: 0x60010000'00000000 - 0x60020000'00000000 (1TB) + */ + sh_mem_bases = (LDS_APP_BASE << SH_MEM_BASES__SHARED_BASE__SHIFT) | + SCRATCH_APP_BASE; + + mutex_lock(&adev->srbm_mutex); + for (i = adev->vm_manager.first_kfd_vmid; i < AMDGPU_NUM_VMID; i++) { + soc21_grbm_select(adev, 0, 0, 0, i); + /* CP and shaders */ + WREG32_SOC15(GC, 0, regSH_MEM_CONFIG, DEFAULT_SH_MEM_CONFIG); + WREG32_SOC15(GC, 0, regSH_MEM_BASES, sh_mem_bases); + + /* Enable trap for each kfd vmid. */ + data = RREG32_SOC15(GC, 0, regSPI_GDBG_PER_VMID_CNTL); + data = REG_SET_FIELD(data, SPI_GDBG_PER_VMID_CNTL, TRAP_EN, 1); + } + soc21_grbm_select(adev, 0, 0, 0, 0); + mutex_unlock(&adev->srbm_mutex); + + /* Initialize all compute VMIDs to have no GDS, GWS, or OA + acccess. These should be enabled by FW for target VMIDs. */ + for (i = adev->vm_manager.first_kfd_vmid; i < AMDGPU_NUM_VMID; i++) { + WREG32_SOC15_OFFSET(GC, 0, regGDS_VMID0_BASE, 2 * i, 0); + WREG32_SOC15_OFFSET(GC, 0, regGDS_VMID0_SIZE, 2 * i, 0); + WREG32_SOC15_OFFSET(GC, 0, regGDS_GWS_VMID0, i, 0); + WREG32_SOC15_OFFSET(GC, 0, regGDS_OA_VMID0, i, 0); + } +} + +static void gfx_v11_0_init_gds_vmid(struct amdgpu_device *adev) +{ + int vmid; + + /* + * Initialize all compute and user-gfx VMIDs to have no GDS, GWS, or OA + * access. Compute VMIDs should be enabled by FW for target VMIDs, + * the driver can enable them for graphics. VMID0 should maintain + * access so that HWS firmware can save/restore entries. + */ + for (vmid = 1; vmid < 16; vmid++) { + WREG32_SOC15_OFFSET(GC, 0, regGDS_VMID0_BASE, 2 * vmid, 0); + WREG32_SOC15_OFFSET(GC, 0, regGDS_VMID0_SIZE, 2 * vmid, 0); + WREG32_SOC15_OFFSET(GC, 0, regGDS_GWS_VMID0, vmid, 0); + WREG32_SOC15_OFFSET(GC, 0, regGDS_OA_VMID0, vmid, 0); + } +} + +static void gfx_v11_0_tcp_harvest(struct amdgpu_device *adev) +{ + /* TODO: harvest feature to be added later. */ +} + +static void gfx_v11_0_get_tcc_info(struct amdgpu_device *adev) +{ + /* TCCs are global (not instanced). */ + uint32_t tcc_disable = RREG32_SOC15(GC, 0, regCGTS_TCC_DISABLE) | + RREG32_SOC15(GC, 0, regCGTS_USER_TCC_DISABLE); + + adev->gfx.config.tcc_disabled_mask = + REG_GET_FIELD(tcc_disable, CGTS_TCC_DISABLE, TCC_DISABLE) | + (REG_GET_FIELD(tcc_disable, CGTS_TCC_DISABLE, HI_TCC_DISABLE) << 16); +} + +static void gfx_v11_0_constants_init(struct amdgpu_device *adev) +{ + u32 tmp; + int i; + + WREG32_FIELD15_PREREG(GC, 0, GRBM_CNTL, READ_TIMEOUT, 0xff); + + gfx_v11_0_setup_rb(adev); + gfx_v11_0_get_cu_info(adev, &adev->gfx.cu_info); + gfx_v11_0_get_tcc_info(adev); + adev->gfx.config.pa_sc_tile_steering_override = 0; + + /* XXX SH_MEM regs */ + /* where to put LDS, scratch, GPUVM in FSA64 space */ + mutex_lock(&adev->srbm_mutex); + for (i = 0; i < adev->vm_manager.id_mgr[AMDGPU_GFXHUB_0].num_ids; i++) { + soc21_grbm_select(adev, 0, 0, 0, i); + /* CP and shaders */ + WREG32_SOC15(GC, 0, regSH_MEM_CONFIG, DEFAULT_SH_MEM_CONFIG); + if (i != 0) { + tmp = REG_SET_FIELD(0, SH_MEM_BASES, PRIVATE_BASE, + (adev->gmc.private_aperture_start >> 48)); + tmp = REG_SET_FIELD(tmp, SH_MEM_BASES, SHARED_BASE, + (adev->gmc.shared_aperture_start >> 48)); + WREG32_SOC15(GC, 0, regSH_MEM_BASES, tmp); + } + } + soc21_grbm_select(adev, 0, 0, 0, 0); + + mutex_unlock(&adev->srbm_mutex); + + gfx_v11_0_init_compute_vmid(adev); + gfx_v11_0_init_gds_vmid(adev); +} + +static void gfx_v11_0_enable_gui_idle_interrupt(struct amdgpu_device *adev, + bool enable) +{ + u32 tmp; + + if (amdgpu_sriov_vf(adev)) + return; + + tmp = RREG32_SOC15(GC, 0, regCP_INT_CNTL_RING0); + + tmp = REG_SET_FIELD(tmp, CP_INT_CNTL_RING0, CNTX_BUSY_INT_ENABLE, + enable ? 1 : 0); + tmp = REG_SET_FIELD(tmp, CP_INT_CNTL_RING0, CNTX_EMPTY_INT_ENABLE, + enable ? 1 : 0); + tmp = REG_SET_FIELD(tmp, CP_INT_CNTL_RING0, CMP_BUSY_INT_ENABLE, + enable ? 1 : 0); + tmp = REG_SET_FIELD(tmp, CP_INT_CNTL_RING0, GFX_IDLE_INT_ENABLE, + enable ? 1 : 0); + + WREG32_SOC15(GC, 0, regCP_INT_CNTL_RING0, tmp); +} + +static int gfx_v11_0_init_csb(struct amdgpu_device *adev) +{ + adev->gfx.rlc.funcs->get_csb_buffer(adev, adev->gfx.rlc.cs_ptr); + + WREG32_SOC15(GC, 0, regRLC_CSIB_ADDR_HI, + adev->gfx.rlc.clear_state_gpu_addr >> 32); + WREG32_SOC15(GC, 0, regRLC_CSIB_ADDR_LO, + adev->gfx.rlc.clear_state_gpu_addr & 0xfffffffc); + WREG32_SOC15(GC, 0, regRLC_CSIB_LENGTH, adev->gfx.rlc.clear_state_size); + + return 0; +} + +static void gfx_v11_0_rlc_stop(struct amdgpu_device *adev) +{ + u32 tmp = RREG32_SOC15(GC, 0, regRLC_CNTL); + + tmp = REG_SET_FIELD(tmp, RLC_CNTL, RLC_ENABLE_F32, 0); + WREG32_SOC15(GC, 0, regRLC_CNTL, tmp); +} + +static void gfx_v11_0_rlc_reset(struct amdgpu_device *adev) +{ + WREG32_FIELD15_PREREG(GC, 0, GRBM_SOFT_RESET, SOFT_RESET_RLC, 1); + udelay(50); + WREG32_FIELD15_PREREG(GC, 0, GRBM_SOFT_RESET, SOFT_RESET_RLC, 0); + udelay(50); +} + +static void gfx_v11_0_rlc_smu_handshake_cntl(struct amdgpu_device *adev, + bool enable) +{ + uint32_t rlc_pg_cntl; + + rlc_pg_cntl = RREG32_SOC15(GC, 0, regRLC_PG_CNTL); + + if (!enable) { + /* RLC_PG_CNTL[23] = 0 (default) + * RLC will wait for handshake acks with SMU + * GFXOFF will be enabled + * RLC_PG_CNTL[23] = 1 + * RLC will not issue any message to SMU + * hence no handshake between SMU & RLC + * GFXOFF will be disabled + */ + rlc_pg_cntl |= RLC_PG_CNTL__SMU_HANDSHAKE_DISABLE_MASK; + } else + rlc_pg_cntl &= ~RLC_PG_CNTL__SMU_HANDSHAKE_DISABLE_MASK; + WREG32_SOC15(GC, 0, regRLC_PG_CNTL, rlc_pg_cntl); +} + +static void gfx_v11_0_rlc_start(struct amdgpu_device *adev) +{ + /* TODO: enable rlc & smu handshake until smu + * and gfxoff feature works as expected */ + if (!(amdgpu_pp_feature_mask & PP_GFXOFF_MASK)) + gfx_v11_0_rlc_smu_handshake_cntl(adev, false); + + WREG32_FIELD15_PREREG(GC, 0, RLC_CNTL, RLC_ENABLE_F32, 1); + udelay(50); +} + +static void gfx_v11_0_rlc_enable_srm(struct amdgpu_device *adev) +{ + uint32_t tmp; + + /* enable Save Restore Machine */ + tmp = RREG32(SOC15_REG_OFFSET(GC, 0, regRLC_SRM_CNTL)); + tmp |= RLC_SRM_CNTL__AUTO_INCR_ADDR_MASK; + tmp |= RLC_SRM_CNTL__SRM_ENABLE_MASK; + WREG32(SOC15_REG_OFFSET(GC, 0, regRLC_SRM_CNTL), tmp); +} + +static void gfx_v11_0_load_rlcg_microcode(struct amdgpu_device *adev) +{ + const struct rlc_firmware_header_v2_0 *hdr; + const __le32 *fw_data; + unsigned i, fw_size; + + hdr = (const struct rlc_firmware_header_v2_0 *)adev->gfx.rlc_fw->data; + fw_data = (const __le32 *)(adev->gfx.rlc_fw->data + + le32_to_cpu(hdr->header.ucode_array_offset_bytes)); + fw_size = le32_to_cpu(hdr->header.ucode_size_bytes) / 4; + + WREG32_SOC15(GC, 0, regRLC_GPM_UCODE_ADDR, + RLCG_UCODE_LOADING_START_ADDRESS); + + for (i = 0; i < fw_size; i++) + WREG32_SOC15(GC, 0, regRLC_GPM_UCODE_DATA, + le32_to_cpup(fw_data++)); + + WREG32_SOC15(GC, 0, regRLC_GPM_UCODE_ADDR, adev->gfx.rlc_fw_version); +} + +static void gfx_v11_0_load_rlc_iram_dram_microcode(struct amdgpu_device *adev) +{ + const struct rlc_firmware_header_v2_2 *hdr; + const __le32 *fw_data; + unsigned i, fw_size; + u32 tmp; + + hdr = (const struct rlc_firmware_header_v2_2 *)adev->gfx.rlc_fw->data; + + fw_data = (const __le32 *)(adev->gfx.rlc_fw->data + + le32_to_cpu(hdr->rlc_iram_ucode_offset_bytes)); + fw_size = le32_to_cpu(hdr->rlc_iram_ucode_size_bytes) / 4; + + WREG32_SOC15(GC, 0, regRLC_LX6_IRAM_ADDR, 0); + + for (i = 0; i < fw_size; i++) { + if ((amdgpu_emu_mode == 1) && (i % 100 == 99)) + msleep(1); + WREG32_SOC15(GC, 0, regRLC_LX6_IRAM_DATA, + le32_to_cpup(fw_data++)); + } + + WREG32_SOC15(GC, 0, regRLC_LX6_IRAM_ADDR, adev->gfx.rlc_fw_version); + + fw_data = (const __le32 *)(adev->gfx.rlc_fw->data + + le32_to_cpu(hdr->rlc_dram_ucode_offset_bytes)); + fw_size = le32_to_cpu(hdr->rlc_dram_ucode_size_bytes) / 4; + + WREG32_SOC15(GC, 0, regRLC_LX6_DRAM_ADDR, 0); + for (i = 0; i < fw_size; i++) { + if ((amdgpu_emu_mode == 1) && (i % 100 == 99)) + msleep(1); + WREG32_SOC15(GC, 0, regRLC_LX6_DRAM_DATA, + le32_to_cpup(fw_data++)); + } + + WREG32_SOC15(GC, 0, regRLC_LX6_IRAM_ADDR, adev->gfx.rlc_fw_version); + + tmp = RREG32_SOC15(GC, 0, regRLC_LX6_CNTL); + tmp = REG_SET_FIELD(tmp, RLC_LX6_CNTL, PDEBUG_ENABLE, 1); + tmp = REG_SET_FIELD(tmp, RLC_LX6_CNTL, BRESET, 0); + WREG32_SOC15(GC, 0, regRLC_LX6_CNTL, tmp); +} + +static void gfx_v11_0_load_rlcp_rlcv_microcode(struct amdgpu_device *adev) +{ + const struct rlc_firmware_header_v2_3 *hdr; + const __le32 *fw_data; + unsigned i, fw_size; + u32 tmp; + + hdr = (const struct rlc_firmware_header_v2_3 *)adev->gfx.rlc_fw->data; + + fw_data = (const __le32 *)(adev->gfx.rlc_fw->data + + le32_to_cpu(hdr->rlcp_ucode_offset_bytes)); + fw_size = le32_to_cpu(hdr->rlcp_ucode_size_bytes) / 4; + + WREG32_SOC15(GC, 0, regRLC_PACE_UCODE_ADDR, 0); + + for (i = 0; i < fw_size; i++) { + if ((amdgpu_emu_mode == 1) && (i % 100 == 99)) + msleep(1); + WREG32_SOC15(GC, 0, regRLC_PACE_UCODE_DATA, + le32_to_cpup(fw_data++)); + } + + WREG32_SOC15(GC, 0, regRLC_PACE_UCODE_ADDR, adev->gfx.rlc_fw_version); + + tmp = RREG32_SOC15(GC, 0, regRLC_GPM_THREAD_ENABLE); + tmp = REG_SET_FIELD(tmp, RLC_GPM_THREAD_ENABLE, THREAD1_ENABLE, 1); + WREG32_SOC15(GC, 0, regRLC_GPM_THREAD_ENABLE, tmp); + + fw_data = (const __le32 *)(adev->gfx.rlc_fw->data + + le32_to_cpu(hdr->rlcv_ucode_offset_bytes)); + fw_size = le32_to_cpu(hdr->rlcv_ucode_size_bytes) / 4; + + WREG32_SOC15(GC, 0, regRLC_GPU_IOV_UCODE_ADDR, 0); + + for (i = 0; i < fw_size; i++) { + if ((amdgpu_emu_mode == 1) && (i % 100 == 99)) + msleep(1); + WREG32_SOC15(GC, 0, regRLC_GPU_IOV_UCODE_DATA, + le32_to_cpup(fw_data++)); + } + + WREG32_SOC15(GC, 0, regRLC_GPU_IOV_UCODE_ADDR, adev->gfx.rlc_fw_version); + + tmp = RREG32_SOC15(GC, 0, regRLC_GPU_IOV_F32_CNTL); + tmp = REG_SET_FIELD(tmp, RLC_GPU_IOV_F32_CNTL, ENABLE, 1); + WREG32_SOC15(GC, 0, regRLC_GPU_IOV_F32_CNTL, tmp); +} + +static int gfx_v11_0_rlc_load_microcode(struct amdgpu_device *adev) +{ + const struct rlc_firmware_header_v2_0 *hdr; + uint16_t version_major; + uint16_t version_minor; + + if (!adev->gfx.rlc_fw) + return -EINVAL; + + hdr = (const struct rlc_firmware_header_v2_0 *)adev->gfx.rlc_fw->data; + amdgpu_ucode_print_rlc_hdr(&hdr->header); + + version_major = le16_to_cpu(hdr->header.header_version_major); + version_minor = le16_to_cpu(hdr->header.header_version_minor); + + if (version_major == 2) { + gfx_v11_0_load_rlcg_microcode(adev); + if (amdgpu_dpm == 1) { + if (version_minor >= 2) + gfx_v11_0_load_rlc_iram_dram_microcode(adev); + if (version_minor == 3) + gfx_v11_0_load_rlcp_rlcv_microcode(adev); + } + + return 0; + } + + return -EINVAL; +} + +static int gfx_v11_0_rlc_resume(struct amdgpu_device *adev) +{ + int r; + + if (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) { + gfx_v11_0_init_csb(adev); + + if (!amdgpu_sriov_vf(adev)) /* enable RLC SRM */ + gfx_v11_0_rlc_enable_srm(adev); + } else { + if (amdgpu_sriov_vf(adev)) { + gfx_v11_0_init_csb(adev); + return 0; + } + + adev->gfx.rlc.funcs->stop(adev); + + /* disable CG */ + WREG32_SOC15(GC, 0, regRLC_CGCG_CGLS_CTRL, 0); + + /* disable PG */ + WREG32_SOC15(GC, 0, regRLC_PG_CNTL, 0); + + if (adev->firmware.load_type == AMDGPU_FW_LOAD_DIRECT) { + /* legacy rlc firmware loading */ + r = gfx_v11_0_rlc_load_microcode(adev); + if (r) + return r; + } + + gfx_v11_0_init_csb(adev); + + adev->gfx.rlc.funcs->start(adev); + } + return 0; +} + +static int gfx_v11_0_config_me_cache(struct amdgpu_device *adev, uint64_t addr) +{ + uint32_t usec_timeout = 50000; /* wait for 50ms */ + uint32_t tmp; + int i; + + /* Trigger an invalidation of the L1 instruction caches */ + tmp = RREG32_SOC15(GC, 0, regCP_ME_IC_OP_CNTL); + tmp = REG_SET_FIELD(tmp, CP_ME_IC_OP_CNTL, INVALIDATE_CACHE, 1); + WREG32_SOC15(GC, 0, regCP_ME_IC_OP_CNTL, tmp); + + /* Wait for invalidation complete */ + for (i = 0; i < usec_timeout; i++) { + tmp = RREG32_SOC15(GC, 0, regCP_ME_IC_OP_CNTL); + if (1 == REG_GET_FIELD(tmp, CP_ME_IC_OP_CNTL, + INVALIDATE_CACHE_COMPLETE)) + break; + udelay(1); + } + + if (i >= usec_timeout) { + dev_err(adev->dev, "failed to invalidate instruction cache\n"); + return -EINVAL; + } + + if (amdgpu_emu_mode == 1) + adev->hdp.funcs->flush_hdp(adev, NULL); + + tmp = RREG32_SOC15(GC, 0, regCP_ME_IC_BASE_CNTL); + tmp = REG_SET_FIELD(tmp, CP_ME_IC_BASE_CNTL, VMID, 0); + tmp = REG_SET_FIELD(tmp, CP_ME_IC_BASE_CNTL, CACHE_POLICY, 0); + tmp = REG_SET_FIELD(tmp, CP_ME_IC_BASE_CNTL, EXE_DISABLE, 0); + tmp = REG_SET_FIELD(tmp, CP_ME_IC_BASE_CNTL, ADDRESS_CLAMP, 1); + WREG32_SOC15(GC, 0, regCP_ME_IC_BASE_CNTL, tmp); + + /* Program me ucode address into intruction cache address register */ + WREG32_SOC15(GC, 0, regCP_ME_IC_BASE_LO, + lower_32_bits(addr) & 0xFFFFF000); + WREG32_SOC15(GC, 0, regCP_ME_IC_BASE_HI, + upper_32_bits(addr)); + + return 0; +} + +static int gfx_v11_0_config_pfp_cache(struct amdgpu_device *adev, uint64_t addr) +{ + uint32_t usec_timeout = 50000; /* wait for 50ms */ + uint32_t tmp; + int i; + + /* Trigger an invalidation of the L1 instruction caches */ + tmp = RREG32_SOC15(GC, 0, regCP_PFP_IC_OP_CNTL); + tmp = REG_SET_FIELD(tmp, CP_PFP_IC_OP_CNTL, INVALIDATE_CACHE, 1); + WREG32_SOC15(GC, 0, regCP_PFP_IC_OP_CNTL, tmp); + + /* Wait for invalidation complete */ + for (i = 0; i < usec_timeout; i++) { + tmp = RREG32_SOC15(GC, 0, regCP_PFP_IC_OP_CNTL); + if (1 == REG_GET_FIELD(tmp, CP_PFP_IC_OP_CNTL, + INVALIDATE_CACHE_COMPLETE)) + break; + udelay(1); + } + + if (i >= usec_timeout) { + dev_err(adev->dev, "failed to invalidate instruction cache\n"); + return -EINVAL; + } + + if (amdgpu_emu_mode == 1) + adev->hdp.funcs->flush_hdp(adev, NULL); + + tmp = RREG32_SOC15(GC, 0, regCP_PFP_IC_BASE_CNTL); + tmp = REG_SET_FIELD(tmp, CP_PFP_IC_BASE_CNTL, VMID, 0); + tmp = REG_SET_FIELD(tmp, CP_PFP_IC_BASE_CNTL, CACHE_POLICY, 0); + tmp = REG_SET_FIELD(tmp, CP_PFP_IC_BASE_CNTL, EXE_DISABLE, 0); + tmp = REG_SET_FIELD(tmp, CP_PFP_IC_BASE_CNTL, ADDRESS_CLAMP, 1); + WREG32_SOC15(GC, 0, regCP_PFP_IC_BASE_CNTL, tmp); + + /* Program pfp ucode address into intruction cache address register */ + WREG32_SOC15(GC, 0, regCP_PFP_IC_BASE_LO, + lower_32_bits(addr) & 0xFFFFF000); + WREG32_SOC15(GC, 0, regCP_PFP_IC_BASE_HI, + upper_32_bits(addr)); + + return 0; +} + +static int gfx_v11_0_config_mec_cache(struct amdgpu_device *adev, uint64_t addr) +{ + uint32_t usec_timeout = 50000; /* wait for 50ms */ + uint32_t tmp; + int i; + + /* Trigger an invalidation of the L1 instruction caches */ + tmp = RREG32_SOC15(GC, 0, regCP_CPC_IC_OP_CNTL); + tmp = REG_SET_FIELD(tmp, CP_CPC_IC_OP_CNTL, INVALIDATE_CACHE, 1); + + WREG32_SOC15(GC, 0, regCP_CPC_IC_OP_CNTL, tmp); + + /* Wait for invalidation complete */ + for (i = 0; i < usec_timeout; i++) { + tmp = RREG32_SOC15(GC, 0, regCP_CPC_IC_OP_CNTL); + if (1 == REG_GET_FIELD(tmp, CP_CPC_IC_OP_CNTL, + INVALIDATE_CACHE_COMPLETE)) + break; + udelay(1); + } + + if (i >= usec_timeout) { + dev_err(adev->dev, "failed to invalidate instruction cache\n"); + return -EINVAL; + } + + if (amdgpu_emu_mode == 1) + adev->hdp.funcs->flush_hdp(adev, NULL); + + tmp = RREG32_SOC15(GC, 0, regCP_CPC_IC_BASE_CNTL); + tmp = REG_SET_FIELD(tmp, CP_CPC_IC_BASE_CNTL, CACHE_POLICY, 0); + tmp = REG_SET_FIELD(tmp, CP_CPC_IC_BASE_CNTL, EXE_DISABLE, 0); + tmp = REG_SET_FIELD(tmp, CP_CPC_IC_BASE_CNTL, ADDRESS_CLAMP, 1); + WREG32_SOC15(GC, 0, regCP_CPC_IC_BASE_CNTL, tmp); + + /* Program mec1 ucode address into intruction cache address register */ + WREG32_SOC15(GC, 0, regCP_CPC_IC_BASE_LO, + lower_32_bits(addr) & 0xFFFFF000); + WREG32_SOC15(GC, 0, regCP_CPC_IC_BASE_HI, + upper_32_bits(addr)); + + return 0; +} + +static int gfx_v11_0_config_pfp_cache_rs64(struct amdgpu_device *adev, uint64_t addr, uint64_t addr2) +{ + uint32_t usec_timeout = 50000; /* wait for 50ms */ + uint32_t tmp; + unsigned i, pipe_id; + const struct gfx_firmware_header_v2_0 *pfp_hdr; + + pfp_hdr = (const struct gfx_firmware_header_v2_0 *) + adev->gfx.pfp_fw->data; + + WREG32_SOC15(GC, 0, regCP_PFP_IC_BASE_LO, + lower_32_bits(addr)); + WREG32_SOC15(GC, 0, regCP_PFP_IC_BASE_HI, + upper_32_bits(addr)); + + tmp = RREG32_SOC15(GC, 0, regCP_PFP_IC_BASE_CNTL); + tmp = REG_SET_FIELD(tmp, CP_PFP_IC_BASE_CNTL, VMID, 0); + tmp = REG_SET_FIELD(tmp, CP_PFP_IC_BASE_CNTL, CACHE_POLICY, 0); + tmp = REG_SET_FIELD(tmp, CP_PFP_IC_BASE_CNTL, EXE_DISABLE, 0); + WREG32_SOC15(GC, 0, regCP_PFP_IC_BASE_CNTL, tmp); + + /* + * Programming any of the CP_PFP_IC_BASE registers + * forces invalidation of the ME L1 I$. Wait for the + * invalidation complete + */ + for (i = 0; i < usec_timeout; i++) { + tmp = RREG32_SOC15(GC, 0, regCP_PFP_IC_OP_CNTL); + if (1 == REG_GET_FIELD(tmp, CP_PFP_IC_OP_CNTL, + INVALIDATE_CACHE_COMPLETE)) + break; + udelay(1); + } + + if (i >= usec_timeout) { + dev_err(adev->dev, "failed to invalidate instruction cache\n"); + return -EINVAL; + } + + /* Prime the L1 instruction caches */ + tmp = RREG32_SOC15(GC, 0, regCP_PFP_IC_OP_CNTL); + tmp = REG_SET_FIELD(tmp, CP_PFP_IC_OP_CNTL, PRIME_ICACHE, 1); + WREG32_SOC15(GC, 0, regCP_PFP_IC_OP_CNTL, tmp); + /* Waiting for cache primed*/ + for (i = 0; i < usec_timeout; i++) { + tmp = RREG32_SOC15(GC, 0, regCP_PFP_IC_OP_CNTL); + if (1 == REG_GET_FIELD(tmp, CP_PFP_IC_OP_CNTL, + ICACHE_PRIMED)) + break; + udelay(1); + } + + if (i >= usec_timeout) { + dev_err(adev->dev, "failed to prime instruction cache\n"); + return -EINVAL; + } + + mutex_lock(&adev->srbm_mutex); + for (pipe_id = 0; pipe_id < adev->gfx.me.num_pipe_per_me; pipe_id++) { + soc21_grbm_select(adev, 0, pipe_id, 0, 0); + WREG32_SOC15(GC, 0, regCP_PFP_PRGRM_CNTR_START, + (pfp_hdr->ucode_start_addr_hi << 30) | + (pfp_hdr->ucode_start_addr_lo >> 2)); + WREG32_SOC15(GC, 0, regCP_PFP_PRGRM_CNTR_START_HI, + pfp_hdr->ucode_start_addr_hi >> 2); + + /* + * Program CP_ME_CNTL to reset given PIPE to take + * effect of CP_PFP_PRGRM_CNTR_START. + */ + tmp = RREG32_SOC15(GC, 0, regCP_ME_CNTL); + if (pipe_id == 0) + tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, + PFP_PIPE0_RESET, 1); + else + tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, + PFP_PIPE1_RESET, 1); + WREG32_SOC15(GC, 0, regCP_ME_CNTL, tmp); + + /* Clear pfp pipe0 reset bit. */ + if (pipe_id == 0) + tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, + PFP_PIPE0_RESET, 0); + else + tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, + PFP_PIPE1_RESET, 0); + WREG32_SOC15(GC, 0, regCP_ME_CNTL, tmp); + + WREG32_SOC15(GC, 0, regCP_GFX_RS64_DC_BASE0_LO, + lower_32_bits(addr2)); + WREG32_SOC15(GC, 0, regCP_GFX_RS64_DC_BASE0_HI, + upper_32_bits(addr2)); + } + soc21_grbm_select(adev, 0, 0, 0, 0); + mutex_unlock(&adev->srbm_mutex); + + tmp = RREG32_SOC15(GC, 0, regCP_GFX_RS64_DC_BASE_CNTL); + tmp = REG_SET_FIELD(tmp, CP_GFX_RS64_DC_BASE_CNTL, VMID, 0); + tmp = REG_SET_FIELD(tmp, CP_GFX_RS64_DC_BASE_CNTL, CACHE_POLICY, 0); + WREG32_SOC15(GC, 0, regCP_GFX_RS64_DC_BASE_CNTL, tmp); + + /* Invalidate the data caches */ + tmp = RREG32_SOC15(GC, 0, regCP_GFX_RS64_DC_OP_CNTL); + tmp = REG_SET_FIELD(tmp, CP_GFX_RS64_DC_OP_CNTL, INVALIDATE_DCACHE, 1); + WREG32_SOC15(GC, 0, regCP_GFX_RS64_DC_OP_CNTL, tmp); + + for (i = 0; i < usec_timeout; i++) { + tmp = RREG32_SOC15(GC, 0, regCP_GFX_RS64_DC_OP_CNTL); + if (1 == REG_GET_FIELD(tmp, CP_GFX_RS64_DC_OP_CNTL, + INVALIDATE_DCACHE_COMPLETE)) + break; + udelay(1); + } + + if (i >= usec_timeout) { + dev_err(adev->dev, "failed to invalidate RS64 data cache\n"); + return -EINVAL; + } + + return 0; +} + +static int gfx_v11_0_config_me_cache_rs64(struct amdgpu_device *adev, uint64_t addr, uint64_t addr2) +{ + uint32_t usec_timeout = 50000; /* wait for 50ms */ + uint32_t tmp; + unsigned i, pipe_id; + const struct gfx_firmware_header_v2_0 *me_hdr; + + me_hdr = (const struct gfx_firmware_header_v2_0 *) + adev->gfx.me_fw->data; + + WREG32_SOC15(GC, 0, regCP_ME_IC_BASE_LO, + lower_32_bits(addr)); + WREG32_SOC15(GC, 0, regCP_ME_IC_BASE_HI, + upper_32_bits(addr)); + + tmp = RREG32_SOC15(GC, 0, regCP_ME_IC_BASE_CNTL); + tmp = REG_SET_FIELD(tmp, CP_ME_IC_BASE_CNTL, VMID, 0); + tmp = REG_SET_FIELD(tmp, CP_ME_IC_BASE_CNTL, CACHE_POLICY, 0); + tmp = REG_SET_FIELD(tmp, CP_ME_IC_BASE_CNTL, EXE_DISABLE, 0); + WREG32_SOC15(GC, 0, regCP_ME_IC_BASE_CNTL, tmp); + + /* + * Programming any of the CP_ME_IC_BASE registers + * forces invalidation of the ME L1 I$. Wait for the + * invalidation complete + */ + for (i = 0; i < usec_timeout; i++) { + tmp = RREG32_SOC15(GC, 0, regCP_ME_IC_OP_CNTL); + if (1 == REG_GET_FIELD(tmp, CP_ME_IC_OP_CNTL, + INVALIDATE_CACHE_COMPLETE)) + break; + udelay(1); + } + + if (i >= usec_timeout) { + dev_err(adev->dev, "failed to invalidate instruction cache\n"); + return -EINVAL; + } + + /* Prime the instruction caches */ + tmp = RREG32_SOC15(GC, 0, regCP_ME_IC_OP_CNTL); + tmp = REG_SET_FIELD(tmp, CP_ME_IC_OP_CNTL, PRIME_ICACHE, 1); + WREG32_SOC15(GC, 0, regCP_ME_IC_OP_CNTL, tmp); + + /* Waiting for instruction cache primed*/ + for (i = 0; i < usec_timeout; i++) { + tmp = RREG32_SOC15(GC, 0, regCP_ME_IC_OP_CNTL); + if (1 == REG_GET_FIELD(tmp, CP_ME_IC_OP_CNTL, + ICACHE_PRIMED)) + break; + udelay(1); + } + + if (i >= usec_timeout) { + dev_err(adev->dev, "failed to prime instruction cache\n"); + return -EINVAL; + } + + mutex_lock(&adev->srbm_mutex); + for (pipe_id = 0; pipe_id < adev->gfx.me.num_pipe_per_me; pipe_id++) { + soc21_grbm_select(adev, 0, pipe_id, 0, 0); + WREG32_SOC15(GC, 0, regCP_ME_PRGRM_CNTR_START, + (me_hdr->ucode_start_addr_hi << 30) | + (me_hdr->ucode_start_addr_lo >> 2) ); + WREG32_SOC15(GC, 0, regCP_ME_PRGRM_CNTR_START_HI, + me_hdr->ucode_start_addr_hi>>2); + + /* + * Program CP_ME_CNTL to reset given PIPE to take + * effect of CP_PFP_PRGRM_CNTR_START. + */ + tmp = RREG32_SOC15(GC, 0, regCP_ME_CNTL); + if (pipe_id == 0) + tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, + ME_PIPE0_RESET, 1); + else + tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, + ME_PIPE1_RESET, 1); + WREG32_SOC15(GC, 0, regCP_ME_CNTL, tmp); + + /* Clear pfp pipe0 reset bit. */ + if (pipe_id == 0) + tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, + ME_PIPE0_RESET, 0); + else + tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, + ME_PIPE1_RESET, 0); + WREG32_SOC15(GC, 0, regCP_ME_CNTL, tmp); + + WREG32_SOC15(GC, 0, regCP_GFX_RS64_DC_BASE1_LO, + lower_32_bits(addr2)); + WREG32_SOC15(GC, 0, regCP_GFX_RS64_DC_BASE1_HI, + upper_32_bits(addr2)); + } + soc21_grbm_select(adev, 0, 0, 0, 0); + mutex_unlock(&adev->srbm_mutex); + + tmp = RREG32_SOC15(GC, 0, regCP_GFX_RS64_DC_BASE_CNTL); + tmp = REG_SET_FIELD(tmp, CP_GFX_RS64_DC_BASE_CNTL, VMID, 0); + tmp = REG_SET_FIELD(tmp, CP_GFX_RS64_DC_BASE_CNTL, CACHE_POLICY, 0); + WREG32_SOC15(GC, 0, regCP_GFX_RS64_DC_BASE_CNTL, tmp); + + /* Invalidate the data caches */ + tmp = RREG32_SOC15(GC, 0, regCP_GFX_RS64_DC_OP_CNTL); + tmp = REG_SET_FIELD(tmp, CP_GFX_RS64_DC_OP_CNTL, INVALIDATE_DCACHE, 1); + WREG32_SOC15(GC, 0, regCP_GFX_RS64_DC_OP_CNTL, tmp); + + for (i = 0; i < usec_timeout; i++) { + tmp = RREG32_SOC15(GC, 0, regCP_GFX_RS64_DC_OP_CNTL); + if (1 == REG_GET_FIELD(tmp, CP_GFX_RS64_DC_OP_CNTL, + INVALIDATE_DCACHE_COMPLETE)) + break; + udelay(1); + } + + if (i >= usec_timeout) { + dev_err(adev->dev, "failed to invalidate RS64 data cache\n"); + return -EINVAL; + } + + return 0; +} + +static int gfx_v11_0_config_mec_cache_rs64(struct amdgpu_device *adev, uint64_t addr, uint64_t addr2) +{ + uint32_t usec_timeout = 50000; /* wait for 50ms */ + uint32_t tmp; + unsigned i; + const struct gfx_firmware_header_v2_0 *mec_hdr; + + mec_hdr = (const struct gfx_firmware_header_v2_0 *) + adev->gfx.mec_fw->data; + + tmp = RREG32_SOC15(GC, 0, regCP_CPC_IC_BASE_CNTL); + tmp = REG_SET_FIELD(tmp, CP_CPC_IC_BASE_CNTL, VMID, 0); + tmp = REG_SET_FIELD(tmp, CP_CPC_IC_BASE_CNTL, EXE_DISABLE, 0); + tmp = REG_SET_FIELD(tmp, CP_CPC_IC_BASE_CNTL, CACHE_POLICY, 0); + WREG32_SOC15(GC, 0, regCP_CPC_IC_BASE_CNTL, tmp); + + tmp = RREG32_SOC15(GC, 0, regCP_MEC_DC_BASE_CNTL); + tmp = REG_SET_FIELD(tmp, CP_MEC_DC_BASE_CNTL, VMID, 0); + tmp = REG_SET_FIELD(tmp, CP_MEC_DC_BASE_CNTL, CACHE_POLICY, 0); + WREG32_SOC15(GC, 0, regCP_MEC_DC_BASE_CNTL, tmp); + + mutex_lock(&adev->srbm_mutex); + for (i = 0; i < adev->gfx.mec.num_pipe_per_mec; i++) { + soc21_grbm_select(adev, 1, i, 0, 0); + + WREG32_SOC15(GC, 0, regCP_MEC_MDBASE_LO, addr2); + WREG32_SOC15(GC, 0, regCP_MEC_MDBASE_HI, + upper_32_bits(addr2)); + + WREG32_SOC15(GC, 0, regCP_MEC_RS64_PRGRM_CNTR_START, + mec_hdr->ucode_start_addr_lo >> 2 | + mec_hdr->ucode_start_addr_hi << 30); + WREG32_SOC15(GC, 0, regCP_MEC_RS64_PRGRM_CNTR_START_HI, + mec_hdr->ucode_start_addr_hi >> 2); + + WREG32_SOC15(GC, 0, regCP_CPC_IC_BASE_LO, addr); + WREG32_SOC15(GC, 0, regCP_CPC_IC_BASE_HI, + upper_32_bits(addr)); + } + mutex_unlock(&adev->srbm_mutex); + soc21_grbm_select(adev, 0, 0, 0, 0); + + /* Trigger an invalidation of the L1 instruction caches */ + tmp = RREG32_SOC15(GC, 0, regCP_MEC_DC_OP_CNTL); + tmp = REG_SET_FIELD(tmp, CP_MEC_DC_OP_CNTL, INVALIDATE_DCACHE, 1); + WREG32_SOC15(GC, 0, regCP_MEC_DC_OP_CNTL, tmp); + + /* Wait for invalidation complete */ + for (i = 0; i < usec_timeout; i++) { + tmp = RREG32_SOC15(GC, 0, regCP_MEC_DC_OP_CNTL); + if (1 == REG_GET_FIELD(tmp, CP_MEC_DC_OP_CNTL, + INVALIDATE_DCACHE_COMPLETE)) + break; + udelay(1); + } + + if (i >= usec_timeout) { + dev_err(adev->dev, "failed to invalidate instruction cache\n"); + return -EINVAL; + } + + /* Trigger an invalidation of the L1 instruction caches */ + tmp = RREG32_SOC15(GC, 0, regCP_CPC_IC_OP_CNTL); + tmp = REG_SET_FIELD(tmp, CP_CPC_IC_OP_CNTL, INVALIDATE_CACHE, 1); + WREG32_SOC15(GC, 0, regCP_CPC_IC_OP_CNTL, tmp); + + /* Wait for invalidation complete */ + for (i = 0; i < usec_timeout; i++) { + tmp = RREG32_SOC15(GC, 0, regCP_CPC_IC_OP_CNTL); + if (1 == REG_GET_FIELD(tmp, CP_CPC_IC_OP_CNTL, + INVALIDATE_CACHE_COMPLETE)) + break; + udelay(1); + } + + if (i >= usec_timeout) { + dev_err(adev->dev, "failed to invalidate instruction cache\n"); + return -EINVAL; + } + + return 0; +} + +static void gfx_v11_0_config_gfx_rs64(struct amdgpu_device *adev) +{ + const struct gfx_firmware_header_v2_0 *pfp_hdr; + const struct gfx_firmware_header_v2_0 *me_hdr; + const struct gfx_firmware_header_v2_0 *mec_hdr; + uint32_t pipe_id, tmp; + + mec_hdr = (const struct gfx_firmware_header_v2_0 *) + adev->gfx.mec_fw->data; + me_hdr = (const struct gfx_firmware_header_v2_0 *) + adev->gfx.me_fw->data; + pfp_hdr = (const struct gfx_firmware_header_v2_0 *) + adev->gfx.pfp_fw->data; + + /* config pfp program start addr */ + for (pipe_id = 0; pipe_id < 2; pipe_id++) { + soc21_grbm_select(adev, 0, pipe_id, 0, 0); + WREG32_SOC15(GC, 0, regCP_PFP_PRGRM_CNTR_START, + (pfp_hdr->ucode_start_addr_hi << 30) | + (pfp_hdr->ucode_start_addr_lo >> 2)); + WREG32_SOC15(GC, 0, regCP_PFP_PRGRM_CNTR_START_HI, + pfp_hdr->ucode_start_addr_hi >> 2); + } + soc21_grbm_select(adev, 0, 0, 0, 0); + + /* reset pfp pipe */ + tmp = RREG32_SOC15(GC, 0, regCP_ME_CNTL); + tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, PFP_PIPE0_RESET, 1); + tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, PFP_PIPE1_RESET, 1); + WREG32_SOC15(GC, 0, regCP_ME_CNTL, tmp); + + /* clear pfp pipe reset */ + tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, PFP_PIPE0_RESET, 0); + tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, PFP_PIPE1_RESET, 0); + WREG32_SOC15(GC, 0, regCP_ME_CNTL, tmp); + + /* config me program start addr */ + for (pipe_id = 0; pipe_id < 2; pipe_id++) { + soc21_grbm_select(adev, 0, pipe_id, 0, 0); + WREG32_SOC15(GC, 0, regCP_ME_PRGRM_CNTR_START, + (me_hdr->ucode_start_addr_hi << 30) | + (me_hdr->ucode_start_addr_lo >> 2) ); + WREG32_SOC15(GC, 0, regCP_ME_PRGRM_CNTR_START_HI, + me_hdr->ucode_start_addr_hi>>2); + } + soc21_grbm_select(adev, 0, 0, 0, 0); + + /* reset me pipe */ + tmp = RREG32_SOC15(GC, 0, regCP_ME_CNTL); + tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, ME_PIPE0_RESET, 1); + tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, ME_PIPE1_RESET, 1); + WREG32_SOC15(GC, 0, regCP_ME_CNTL, tmp); + + /* clear me pipe reset */ + tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, ME_PIPE0_RESET, 0); + tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, ME_PIPE1_RESET, 0); + WREG32_SOC15(GC, 0, regCP_ME_CNTL, tmp); + + /* config mec program start addr */ + for (pipe_id = 0; pipe_id < 4; pipe_id++) { + soc21_grbm_select(adev, 1, pipe_id, 0, 0); + WREG32_SOC15(GC, 0, regCP_MEC_RS64_PRGRM_CNTR_START, + mec_hdr->ucode_start_addr_lo >> 2 | + mec_hdr->ucode_start_addr_hi << 30); + WREG32_SOC15(GC, 0, regCP_MEC_RS64_PRGRM_CNTR_START_HI, + mec_hdr->ucode_start_addr_hi >> 2); + } + soc21_grbm_select(adev, 0, 0, 0, 0); + + /* reset mec pipe */ + tmp = RREG32_SOC15(GC, 0, regCP_MEC_RS64_CNTL); + tmp = REG_SET_FIELD(tmp, CP_MEC_RS64_CNTL, MEC_PIPE0_RESET, 1); + tmp = REG_SET_FIELD(tmp, CP_MEC_RS64_CNTL, MEC_PIPE1_RESET, 1); + tmp = REG_SET_FIELD(tmp, CP_MEC_RS64_CNTL, MEC_PIPE2_RESET, 1); + tmp = REG_SET_FIELD(tmp, CP_MEC_RS64_CNTL, MEC_PIPE3_RESET, 1); + WREG32_SOC15(GC, 0, regCP_MEC_RS64_CNTL, tmp); + + /* clear mec pipe reset */ + tmp = REG_SET_FIELD(tmp, CP_MEC_RS64_CNTL, MEC_PIPE0_RESET, 0); + tmp = REG_SET_FIELD(tmp, CP_MEC_RS64_CNTL, MEC_PIPE1_RESET, 0); + tmp = REG_SET_FIELD(tmp, CP_MEC_RS64_CNTL, MEC_PIPE2_RESET, 0); + tmp = REG_SET_FIELD(tmp, CP_MEC_RS64_CNTL, MEC_PIPE3_RESET, 0); + WREG32_SOC15(GC, 0, regCP_MEC_RS64_CNTL, tmp); +} + +static int gfx_v11_0_wait_for_rlc_autoload_complete(struct amdgpu_device *adev) +{ + uint32_t cp_status; + uint32_t bootload_status; + int i, r; + uint64_t addr, addr2; + + for (i = 0; i < adev->usec_timeout; i++) { + cp_status = RREG32_SOC15(GC, 0, regCP_STAT); + + if (adev->ip_versions[GC_HWIP][0] == IP_VERSION(11, 0, 1)) + bootload_status = RREG32_SOC15(GC, 0, + regRLC_RLCS_BOOTLOAD_STATUS_gc_11_0_1); + else + bootload_status = RREG32_SOC15(GC, 0, regRLC_RLCS_BOOTLOAD_STATUS); + + if ((cp_status == 0) && + (REG_GET_FIELD(bootload_status, + RLC_RLCS_BOOTLOAD_STATUS, BOOTLOAD_COMPLETE) == 1)) { + break; + } + udelay(1); + } + + if (i >= adev->usec_timeout) { + dev_err(adev->dev, "rlc autoload: gc ucode autoload timeout\n"); + return -ETIMEDOUT; + } + + if (adev->firmware.load_type == AMDGPU_FW_LOAD_RLC_BACKDOOR_AUTO) { + if (adev->gfx.rs64_enable) { + addr = adev->gfx.rlc.rlc_autoload_gpu_addr + + rlc_autoload_info[SOC21_FIRMWARE_ID_RS64_ME].offset; + addr2 = adev->gfx.rlc.rlc_autoload_gpu_addr + + rlc_autoload_info[SOC21_FIRMWARE_ID_RS64_ME_P0_STACK].offset; + r = gfx_v11_0_config_me_cache_rs64(adev, addr, addr2); + if (r) + return r; + addr = adev->gfx.rlc.rlc_autoload_gpu_addr + + rlc_autoload_info[SOC21_FIRMWARE_ID_RS64_PFP].offset; + addr2 = adev->gfx.rlc.rlc_autoload_gpu_addr + + rlc_autoload_info[SOC21_FIRMWARE_ID_RS64_PFP_P0_STACK].offset; + r = gfx_v11_0_config_pfp_cache_rs64(adev, addr, addr2); + if (r) + return r; + addr = adev->gfx.rlc.rlc_autoload_gpu_addr + + rlc_autoload_info[SOC21_FIRMWARE_ID_RS64_MEC].offset; + addr2 = adev->gfx.rlc.rlc_autoload_gpu_addr + + rlc_autoload_info[SOC21_FIRMWARE_ID_RS64_MEC_P0_STACK].offset; + r = gfx_v11_0_config_mec_cache_rs64(adev, addr, addr2); + if (r) + return r; + } else { + addr = adev->gfx.rlc.rlc_autoload_gpu_addr + + rlc_autoload_info[SOC21_FIRMWARE_ID_CP_ME].offset; + r = gfx_v11_0_config_me_cache(adev, addr); + if (r) + return r; + addr = adev->gfx.rlc.rlc_autoload_gpu_addr + + rlc_autoload_info[SOC21_FIRMWARE_ID_CP_PFP].offset; + r = gfx_v11_0_config_pfp_cache(adev, addr); + if (r) + return r; + addr = adev->gfx.rlc.rlc_autoload_gpu_addr + + rlc_autoload_info[SOC21_FIRMWARE_ID_CP_MEC].offset; + r = gfx_v11_0_config_mec_cache(adev, addr); + if (r) + return r; + } + } + + return 0; +} + +static int gfx_v11_0_cp_gfx_enable(struct amdgpu_device *adev, bool enable) +{ + int i; + u32 tmp = RREG32_SOC15(GC, 0, regCP_ME_CNTL); + + tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, ME_HALT, enable ? 0 : 1); + tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, PFP_HALT, enable ? 0 : 1); + WREG32_SOC15(GC, 0, regCP_ME_CNTL, tmp); + + for (i = 0; i < adev->usec_timeout; i++) { + if (RREG32_SOC15(GC, 0, regCP_STAT) == 0) + break; + udelay(1); + } + + if (i >= adev->usec_timeout) + DRM_ERROR("failed to %s cp gfx\n", enable ? "unhalt" : "halt"); + + return 0; +} + +static int gfx_v11_0_cp_gfx_load_pfp_microcode(struct amdgpu_device *adev) +{ + int r; + const struct gfx_firmware_header_v1_0 *pfp_hdr; + const __le32 *fw_data; + unsigned i, fw_size; + + pfp_hdr = (const struct gfx_firmware_header_v1_0 *) + adev->gfx.pfp_fw->data; + + amdgpu_ucode_print_gfx_hdr(&pfp_hdr->header); + + fw_data = (const __le32 *)(adev->gfx.pfp_fw->data + + le32_to_cpu(pfp_hdr->header.ucode_array_offset_bytes)); + fw_size = le32_to_cpu(pfp_hdr->header.ucode_size_bytes); + + r = amdgpu_bo_create_reserved(adev, pfp_hdr->header.ucode_size_bytes, + PAGE_SIZE, AMDGPU_GEM_DOMAIN_GTT, + &adev->gfx.pfp.pfp_fw_obj, + &adev->gfx.pfp.pfp_fw_gpu_addr, + (void **)&adev->gfx.pfp.pfp_fw_ptr); + if (r) { + dev_err(adev->dev, "(%d) failed to create pfp fw bo\n", r); + gfx_v11_0_pfp_fini(adev); + return r; + } + + memcpy(adev->gfx.pfp.pfp_fw_ptr, fw_data, fw_size); + + amdgpu_bo_kunmap(adev->gfx.pfp.pfp_fw_obj); + amdgpu_bo_unreserve(adev->gfx.pfp.pfp_fw_obj); + + gfx_v11_0_config_pfp_cache(adev, adev->gfx.pfp.pfp_fw_gpu_addr); + + WREG32_SOC15(GC, 0, regCP_HYP_PFP_UCODE_ADDR, 0); + + for (i = 0; i < pfp_hdr->jt_size; i++) + WREG32_SOC15(GC, 0, regCP_HYP_PFP_UCODE_DATA, + le32_to_cpup(fw_data + pfp_hdr->jt_offset + i)); + + WREG32_SOC15(GC, 0, regCP_HYP_PFP_UCODE_ADDR, adev->gfx.pfp_fw_version); + + return 0; +} + +static int gfx_v11_0_cp_gfx_load_pfp_microcode_rs64(struct amdgpu_device *adev) +{ + int r; + const struct gfx_firmware_header_v2_0 *pfp_hdr; + const __le32 *fw_ucode, *fw_data; + unsigned i, pipe_id, fw_ucode_size, fw_data_size; + uint32_t tmp; + uint32_t usec_timeout = 50000; /* wait for 50ms */ + + pfp_hdr = (const struct gfx_firmware_header_v2_0 *) + adev->gfx.pfp_fw->data; + + amdgpu_ucode_print_gfx_hdr(&pfp_hdr->header); + + /* instruction */ + fw_ucode = (const __le32 *)(adev->gfx.pfp_fw->data + + le32_to_cpu(pfp_hdr->ucode_offset_bytes)); + fw_ucode_size = le32_to_cpu(pfp_hdr->ucode_size_bytes); + /* data */ + fw_data = (const __le32 *)(adev->gfx.pfp_fw->data + + le32_to_cpu(pfp_hdr->data_offset_bytes)); + fw_data_size = le32_to_cpu(pfp_hdr->data_size_bytes); + + /* 64kb align */ + r = amdgpu_bo_create_reserved(adev, fw_ucode_size, + 64 * 1024, AMDGPU_GEM_DOMAIN_VRAM, + &adev->gfx.pfp.pfp_fw_obj, + &adev->gfx.pfp.pfp_fw_gpu_addr, + (void **)&adev->gfx.pfp.pfp_fw_ptr); + if (r) { + dev_err(adev->dev, "(%d) failed to create pfp ucode fw bo\n", r); + gfx_v11_0_pfp_fini(adev); + return r; + } + + r = amdgpu_bo_create_reserved(adev, fw_data_size, + 64 * 1024, AMDGPU_GEM_DOMAIN_VRAM, + &adev->gfx.pfp.pfp_fw_data_obj, + &adev->gfx.pfp.pfp_fw_data_gpu_addr, + (void **)&adev->gfx.pfp.pfp_fw_data_ptr); + if (r) { + dev_err(adev->dev, "(%d) failed to create pfp data fw bo\n", r); + gfx_v11_0_pfp_fini(adev); + return r; + } + + memcpy(adev->gfx.pfp.pfp_fw_ptr, fw_ucode, fw_ucode_size); + memcpy(adev->gfx.pfp.pfp_fw_data_ptr, fw_data, fw_data_size); + + amdgpu_bo_kunmap(adev->gfx.pfp.pfp_fw_obj); + amdgpu_bo_kunmap(adev->gfx.pfp.pfp_fw_data_obj); + amdgpu_bo_unreserve(adev->gfx.pfp.pfp_fw_obj); + amdgpu_bo_unreserve(adev->gfx.pfp.pfp_fw_data_obj); + + if (amdgpu_emu_mode == 1) + adev->hdp.funcs->flush_hdp(adev, NULL); + + WREG32_SOC15(GC, 0, regCP_PFP_IC_BASE_LO, + lower_32_bits(adev->gfx.pfp.pfp_fw_gpu_addr)); + WREG32_SOC15(GC, 0, regCP_PFP_IC_BASE_HI, + upper_32_bits(adev->gfx.pfp.pfp_fw_gpu_addr)); + + tmp = RREG32_SOC15(GC, 0, regCP_PFP_IC_BASE_CNTL); + tmp = REG_SET_FIELD(tmp, CP_PFP_IC_BASE_CNTL, VMID, 0); + tmp = REG_SET_FIELD(tmp, CP_PFP_IC_BASE_CNTL, CACHE_POLICY, 0); + tmp = REG_SET_FIELD(tmp, CP_PFP_IC_BASE_CNTL, EXE_DISABLE, 0); + WREG32_SOC15(GC, 0, regCP_PFP_IC_BASE_CNTL, tmp); + + /* + * Programming any of the CP_PFP_IC_BASE registers + * forces invalidation of the ME L1 I$. Wait for the + * invalidation complete + */ + for (i = 0; i < usec_timeout; i++) { + tmp = RREG32_SOC15(GC, 0, regCP_PFP_IC_OP_CNTL); + if (1 == REG_GET_FIELD(tmp, CP_PFP_IC_OP_CNTL, + INVALIDATE_CACHE_COMPLETE)) + break; + udelay(1); + } + + if (i >= usec_timeout) { + dev_err(adev->dev, "failed to invalidate instruction cache\n"); + return -EINVAL; + } + + /* Prime the L1 instruction caches */ + tmp = RREG32_SOC15(GC, 0, regCP_PFP_IC_OP_CNTL); + tmp = REG_SET_FIELD(tmp, CP_PFP_IC_OP_CNTL, PRIME_ICACHE, 1); + WREG32_SOC15(GC, 0, regCP_PFP_IC_OP_CNTL, tmp); + /* Waiting for cache primed*/ + for (i = 0; i < usec_timeout; i++) { + tmp = RREG32_SOC15(GC, 0, regCP_PFP_IC_OP_CNTL); + if (1 == REG_GET_FIELD(tmp, CP_PFP_IC_OP_CNTL, + ICACHE_PRIMED)) + break; + udelay(1); + } + + if (i >= usec_timeout) { + dev_err(adev->dev, "failed to prime instruction cache\n"); + return -EINVAL; + } + + mutex_lock(&adev->srbm_mutex); + for (pipe_id = 0; pipe_id < adev->gfx.me.num_pipe_per_me; pipe_id++) { + soc21_grbm_select(adev, 0, pipe_id, 0, 0); + WREG32_SOC15(GC, 0, regCP_PFP_PRGRM_CNTR_START, + (pfp_hdr->ucode_start_addr_hi << 30) | + (pfp_hdr->ucode_start_addr_lo >> 2) ); + WREG32_SOC15(GC, 0, regCP_PFP_PRGRM_CNTR_START_HI, + pfp_hdr->ucode_start_addr_hi>>2); + + /* + * Program CP_ME_CNTL to reset given PIPE to take + * effect of CP_PFP_PRGRM_CNTR_START. + */ + tmp = RREG32_SOC15(GC, 0, regCP_ME_CNTL); + if (pipe_id == 0) + tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, + PFP_PIPE0_RESET, 1); + else + tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, + PFP_PIPE1_RESET, 1); + WREG32_SOC15(GC, 0, regCP_ME_CNTL, tmp); + + /* Clear pfp pipe0 reset bit. */ + if (pipe_id == 0) + tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, + PFP_PIPE0_RESET, 0); + else + tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, + PFP_PIPE1_RESET, 0); + WREG32_SOC15(GC, 0, regCP_ME_CNTL, tmp); + + WREG32_SOC15(GC, 0, regCP_GFX_RS64_DC_BASE0_LO, + lower_32_bits(adev->gfx.pfp.pfp_fw_data_gpu_addr)); + WREG32_SOC15(GC, 0, regCP_GFX_RS64_DC_BASE0_HI, + upper_32_bits(adev->gfx.pfp.pfp_fw_data_gpu_addr)); + } + soc21_grbm_select(adev, 0, 0, 0, 0); + mutex_unlock(&adev->srbm_mutex); + + tmp = RREG32_SOC15(GC, 0, regCP_GFX_RS64_DC_BASE_CNTL); + tmp = REG_SET_FIELD(tmp, CP_GFX_RS64_DC_BASE_CNTL, VMID, 0); + tmp = REG_SET_FIELD(tmp, CP_GFX_RS64_DC_BASE_CNTL, CACHE_POLICY, 0); + WREG32_SOC15(GC, 0, regCP_GFX_RS64_DC_BASE_CNTL, tmp); + + /* Invalidate the data caches */ + tmp = RREG32_SOC15(GC, 0, regCP_GFX_RS64_DC_OP_CNTL); + tmp = REG_SET_FIELD(tmp, CP_GFX_RS64_DC_OP_CNTL, INVALIDATE_DCACHE, 1); + WREG32_SOC15(GC, 0, regCP_GFX_RS64_DC_OP_CNTL, tmp); + + for (i = 0; i < usec_timeout; i++) { + tmp = RREG32_SOC15(GC, 0, regCP_GFX_RS64_DC_OP_CNTL); + if (1 == REG_GET_FIELD(tmp, CP_GFX_RS64_DC_OP_CNTL, + INVALIDATE_DCACHE_COMPLETE)) + break; + udelay(1); + } + + if (i >= usec_timeout) { + dev_err(adev->dev, "failed to invalidate RS64 data cache\n"); + return -EINVAL; + } + + return 0; +} + +static int gfx_v11_0_cp_gfx_load_me_microcode(struct amdgpu_device *adev) +{ + int r; + const struct gfx_firmware_header_v1_0 *me_hdr; + const __le32 *fw_data; + unsigned i, fw_size; + + me_hdr = (const struct gfx_firmware_header_v1_0 *) + adev->gfx.me_fw->data; + + amdgpu_ucode_print_gfx_hdr(&me_hdr->header); + + fw_data = (const __le32 *)(adev->gfx.me_fw->data + + le32_to_cpu(me_hdr->header.ucode_array_offset_bytes)); + fw_size = le32_to_cpu(me_hdr->header.ucode_size_bytes); + + r = amdgpu_bo_create_reserved(adev, me_hdr->header.ucode_size_bytes, + PAGE_SIZE, AMDGPU_GEM_DOMAIN_GTT, + &adev->gfx.me.me_fw_obj, + &adev->gfx.me.me_fw_gpu_addr, + (void **)&adev->gfx.me.me_fw_ptr); + if (r) { + dev_err(adev->dev, "(%d) failed to create me fw bo\n", r); + gfx_v11_0_me_fini(adev); + return r; + } + + memcpy(adev->gfx.me.me_fw_ptr, fw_data, fw_size); + + amdgpu_bo_kunmap(adev->gfx.me.me_fw_obj); + amdgpu_bo_unreserve(adev->gfx.me.me_fw_obj); + + gfx_v11_0_config_me_cache(adev, adev->gfx.me.me_fw_gpu_addr); + + WREG32_SOC15(GC, 0, regCP_HYP_ME_UCODE_ADDR, 0); + + for (i = 0; i < me_hdr->jt_size; i++) + WREG32_SOC15(GC, 0, regCP_HYP_ME_UCODE_DATA, + le32_to_cpup(fw_data + me_hdr->jt_offset + i)); + + WREG32_SOC15(GC, 0, regCP_HYP_ME_UCODE_ADDR, adev->gfx.me_fw_version); + + return 0; +} + +static int gfx_v11_0_cp_gfx_load_me_microcode_rs64(struct amdgpu_device *adev) +{ + int r; + const struct gfx_firmware_header_v2_0 *me_hdr; + const __le32 *fw_ucode, *fw_data; + unsigned i, pipe_id, fw_ucode_size, fw_data_size; + uint32_t tmp; + uint32_t usec_timeout = 50000; /* wait for 50ms */ + + me_hdr = (const struct gfx_firmware_header_v2_0 *) + adev->gfx.me_fw->data; + + amdgpu_ucode_print_gfx_hdr(&me_hdr->header); + + /* instruction */ + fw_ucode = (const __le32 *)(adev->gfx.me_fw->data + + le32_to_cpu(me_hdr->ucode_offset_bytes)); + fw_ucode_size = le32_to_cpu(me_hdr->ucode_size_bytes); + /* data */ + fw_data = (const __le32 *)(adev->gfx.me_fw->data + + le32_to_cpu(me_hdr->data_offset_bytes)); + fw_data_size = le32_to_cpu(me_hdr->data_size_bytes); + + /* 64kb align*/ + r = amdgpu_bo_create_reserved(adev, fw_ucode_size, + 64 * 1024, AMDGPU_GEM_DOMAIN_VRAM, + &adev->gfx.me.me_fw_obj, + &adev->gfx.me.me_fw_gpu_addr, + (void **)&adev->gfx.me.me_fw_ptr); + if (r) { + dev_err(adev->dev, "(%d) failed to create me ucode bo\n", r); + gfx_v11_0_me_fini(adev); + return r; + } + + r = amdgpu_bo_create_reserved(adev, fw_data_size, + 64 * 1024, AMDGPU_GEM_DOMAIN_VRAM, + &adev->gfx.me.me_fw_data_obj, + &adev->gfx.me.me_fw_data_gpu_addr, + (void **)&adev->gfx.me.me_fw_data_ptr); + if (r) { + dev_err(adev->dev, "(%d) failed to create me data bo\n", r); + gfx_v11_0_pfp_fini(adev); + return r; + } + + memcpy(adev->gfx.me.me_fw_ptr, fw_ucode, fw_ucode_size); + memcpy(adev->gfx.me.me_fw_data_ptr, fw_data, fw_data_size); + + amdgpu_bo_kunmap(adev->gfx.me.me_fw_obj); + amdgpu_bo_kunmap(adev->gfx.me.me_fw_data_obj); + amdgpu_bo_unreserve(adev->gfx.me.me_fw_obj); + amdgpu_bo_unreserve(adev->gfx.me.me_fw_data_obj); + + if (amdgpu_emu_mode == 1) + adev->hdp.funcs->flush_hdp(adev, NULL); + + WREG32_SOC15(GC, 0, regCP_ME_IC_BASE_LO, + lower_32_bits(adev->gfx.me.me_fw_gpu_addr)); + WREG32_SOC15(GC, 0, regCP_ME_IC_BASE_HI, + upper_32_bits(adev->gfx.me.me_fw_gpu_addr)); + + tmp = RREG32_SOC15(GC, 0, regCP_ME_IC_BASE_CNTL); + tmp = REG_SET_FIELD(tmp, CP_ME_IC_BASE_CNTL, VMID, 0); + tmp = REG_SET_FIELD(tmp, CP_ME_IC_BASE_CNTL, CACHE_POLICY, 0); + tmp = REG_SET_FIELD(tmp, CP_ME_IC_BASE_CNTL, EXE_DISABLE, 0); + WREG32_SOC15(GC, 0, regCP_ME_IC_BASE_CNTL, tmp); + + /* + * Programming any of the CP_ME_IC_BASE registers + * forces invalidation of the ME L1 I$. Wait for the + * invalidation complete + */ + for (i = 0; i < usec_timeout; i++) { + tmp = RREG32_SOC15(GC, 0, regCP_ME_IC_OP_CNTL); + if (1 == REG_GET_FIELD(tmp, CP_ME_IC_OP_CNTL, + INVALIDATE_CACHE_COMPLETE)) + break; + udelay(1); + } + + if (i >= usec_timeout) { + dev_err(adev->dev, "failed to invalidate instruction cache\n"); + return -EINVAL; + } + + /* Prime the instruction caches */ + tmp = RREG32_SOC15(GC, 0, regCP_ME_IC_OP_CNTL); + tmp = REG_SET_FIELD(tmp, CP_ME_IC_OP_CNTL, PRIME_ICACHE, 1); + WREG32_SOC15(GC, 0, regCP_ME_IC_OP_CNTL, tmp); + + /* Waiting for instruction cache primed*/ + for (i = 0; i < usec_timeout; i++) { + tmp = RREG32_SOC15(GC, 0, regCP_ME_IC_OP_CNTL); + if (1 == REG_GET_FIELD(tmp, CP_ME_IC_OP_CNTL, + ICACHE_PRIMED)) + break; + udelay(1); + } + + if (i >= usec_timeout) { + dev_err(adev->dev, "failed to prime instruction cache\n"); + return -EINVAL; + } + + mutex_lock(&adev->srbm_mutex); + for (pipe_id = 0; pipe_id < adev->gfx.me.num_pipe_per_me; pipe_id++) { + soc21_grbm_select(adev, 0, pipe_id, 0, 0); + WREG32_SOC15(GC, 0, regCP_ME_PRGRM_CNTR_START, + (me_hdr->ucode_start_addr_hi << 30) | + (me_hdr->ucode_start_addr_lo >> 2) ); + WREG32_SOC15(GC, 0, regCP_ME_PRGRM_CNTR_START_HI, + me_hdr->ucode_start_addr_hi>>2); + + /* + * Program CP_ME_CNTL to reset given PIPE to take + * effect of CP_PFP_PRGRM_CNTR_START. + */ + tmp = RREG32_SOC15(GC, 0, regCP_ME_CNTL); + if (pipe_id == 0) + tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, + ME_PIPE0_RESET, 1); + else + tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, + ME_PIPE1_RESET, 1); + WREG32_SOC15(GC, 0, regCP_ME_CNTL, tmp); + + /* Clear pfp pipe0 reset bit. */ + if (pipe_id == 0) + tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, + ME_PIPE0_RESET, 0); + else + tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, + ME_PIPE1_RESET, 0); + WREG32_SOC15(GC, 0, regCP_ME_CNTL, tmp); + + WREG32_SOC15(GC, 0, regCP_GFX_RS64_DC_BASE1_LO, + lower_32_bits(adev->gfx.me.me_fw_data_gpu_addr)); + WREG32_SOC15(GC, 0, regCP_GFX_RS64_DC_BASE1_HI, + upper_32_bits(adev->gfx.me.me_fw_data_gpu_addr)); + } + soc21_grbm_select(adev, 0, 0, 0, 0); + mutex_unlock(&adev->srbm_mutex); + + tmp = RREG32_SOC15(GC, 0, regCP_GFX_RS64_DC_BASE_CNTL); + tmp = REG_SET_FIELD(tmp, CP_GFX_RS64_DC_BASE_CNTL, VMID, 0); + tmp = REG_SET_FIELD(tmp, CP_GFX_RS64_DC_BASE_CNTL, CACHE_POLICY, 0); + WREG32_SOC15(GC, 0, regCP_GFX_RS64_DC_BASE_CNTL, tmp); + + /* Invalidate the data caches */ + tmp = RREG32_SOC15(GC, 0, regCP_GFX_RS64_DC_OP_CNTL); + tmp = REG_SET_FIELD(tmp, CP_GFX_RS64_DC_OP_CNTL, INVALIDATE_DCACHE, 1); + WREG32_SOC15(GC, 0, regCP_GFX_RS64_DC_OP_CNTL, tmp); + + for (i = 0; i < usec_timeout; i++) { + tmp = RREG32_SOC15(GC, 0, regCP_GFX_RS64_DC_OP_CNTL); + if (1 == REG_GET_FIELD(tmp, CP_GFX_RS64_DC_OP_CNTL, + INVALIDATE_DCACHE_COMPLETE)) + break; + udelay(1); + } + + if (i >= usec_timeout) { + dev_err(adev->dev, "failed to invalidate RS64 data cache\n"); + return -EINVAL; + } + + return 0; +} + +static int gfx_v11_0_cp_gfx_load_microcode(struct amdgpu_device *adev) +{ + int r; + + if (!adev->gfx.me_fw || !adev->gfx.pfp_fw) + return -EINVAL; + + gfx_v11_0_cp_gfx_enable(adev, false); + + if (adev->gfx.rs64_enable) + r = gfx_v11_0_cp_gfx_load_pfp_microcode_rs64(adev); + else + r = gfx_v11_0_cp_gfx_load_pfp_microcode(adev); + if (r) { + dev_err(adev->dev, "(%d) failed to load pfp fw\n", r); + return r; + } + + if (adev->gfx.rs64_enable) + r = gfx_v11_0_cp_gfx_load_me_microcode_rs64(adev); + else + r = gfx_v11_0_cp_gfx_load_me_microcode(adev); + if (r) { + dev_err(adev->dev, "(%d) failed to load me fw\n", r); + return r; + } + + return 0; +} + +static int gfx_v11_0_cp_gfx_start(struct amdgpu_device *adev) +{ + struct amdgpu_ring *ring; + const struct cs_section_def *sect = NULL; + const struct cs_extent_def *ext = NULL; + int r, i; + int ctx_reg_offset; + + /* init the CP */ + WREG32_SOC15(GC, 0, regCP_MAX_CONTEXT, + adev->gfx.config.max_hw_contexts - 1); + WREG32_SOC15(GC, 0, regCP_DEVICE_ID, 1); + + if (!amdgpu_async_gfx_ring) + gfx_v11_0_cp_gfx_enable(adev, true); + + ring = &adev->gfx.gfx_ring[0]; + r = amdgpu_ring_alloc(ring, gfx_v11_0_get_csb_size(adev)); + if (r) { + DRM_ERROR("amdgpu: cp failed to lock ring (%d).\n", r); + return r; + } + + amdgpu_ring_write(ring, PACKET3(PACKET3_PREAMBLE_CNTL, 0)); + amdgpu_ring_write(ring, PACKET3_PREAMBLE_BEGIN_CLEAR_STATE); + + amdgpu_ring_write(ring, PACKET3(PACKET3_CONTEXT_CONTROL, 1)); + amdgpu_ring_write(ring, 0x80000000); + amdgpu_ring_write(ring, 0x80000000); + + for (sect = gfx11_cs_data; sect->section != NULL; ++sect) { + for (ext = sect->section; ext->extent != NULL; ++ext) { + if (sect->id == SECT_CONTEXT) { + amdgpu_ring_write(ring, + PACKET3(PACKET3_SET_CONTEXT_REG, + ext->reg_count)); + amdgpu_ring_write(ring, ext->reg_index - + PACKET3_SET_CONTEXT_REG_START); + for (i = 0; i < ext->reg_count; i++) + amdgpu_ring_write(ring, ext->extent[i]); + } + } + } + + ctx_reg_offset = + SOC15_REG_OFFSET(GC, 0, regPA_SC_TILE_STEERING_OVERRIDE) - PACKET3_SET_CONTEXT_REG_START; + amdgpu_ring_write(ring, PACKET3(PACKET3_SET_CONTEXT_REG, 1)); + amdgpu_ring_write(ring, ctx_reg_offset); + amdgpu_ring_write(ring, adev->gfx.config.pa_sc_tile_steering_override); + + amdgpu_ring_write(ring, PACKET3(PACKET3_PREAMBLE_CNTL, 0)); + amdgpu_ring_write(ring, PACKET3_PREAMBLE_END_CLEAR_STATE); + + amdgpu_ring_write(ring, PACKET3(PACKET3_CLEAR_STATE, 0)); + amdgpu_ring_write(ring, 0); + + amdgpu_ring_commit(ring); + + /* submit cs packet to copy state 0 to next available state */ + if (adev->gfx.num_gfx_rings > 1) { + /* maximum supported gfx ring is 2 */ + ring = &adev->gfx.gfx_ring[1]; + r = amdgpu_ring_alloc(ring, 2); + if (r) { + DRM_ERROR("amdgpu: cp failed to lock ring (%d).\n", r); + return r; + } + + amdgpu_ring_write(ring, PACKET3(PACKET3_CLEAR_STATE, 0)); + amdgpu_ring_write(ring, 0); + + amdgpu_ring_commit(ring); + } + return 0; +} + +static void gfx_v11_0_cp_gfx_switch_pipe(struct amdgpu_device *adev, + CP_PIPE_ID pipe) +{ + u32 tmp; + + tmp = RREG32_SOC15(GC, 0, regGRBM_GFX_CNTL); + tmp = REG_SET_FIELD(tmp, GRBM_GFX_CNTL, PIPEID, pipe); + + WREG32_SOC15(GC, 0, regGRBM_GFX_CNTL, tmp); +} + +static void gfx_v11_0_cp_gfx_set_doorbell(struct amdgpu_device *adev, + struct amdgpu_ring *ring) +{ + u32 tmp; + + tmp = RREG32_SOC15(GC, 0, regCP_RB_DOORBELL_CONTROL); + if (ring->use_doorbell) { + tmp = REG_SET_FIELD(tmp, CP_RB_DOORBELL_CONTROL, + DOORBELL_OFFSET, ring->doorbell_index); + tmp = REG_SET_FIELD(tmp, CP_RB_DOORBELL_CONTROL, + DOORBELL_EN, 1); + } else { + tmp = REG_SET_FIELD(tmp, CP_RB_DOORBELL_CONTROL, + DOORBELL_EN, 0); + } + WREG32_SOC15(GC, 0, regCP_RB_DOORBELL_CONTROL, tmp); + + tmp = REG_SET_FIELD(0, CP_RB_DOORBELL_RANGE_LOWER, + DOORBELL_RANGE_LOWER, ring->doorbell_index); + WREG32_SOC15(GC, 0, regCP_RB_DOORBELL_RANGE_LOWER, tmp); + + WREG32_SOC15(GC, 0, regCP_RB_DOORBELL_RANGE_UPPER, + CP_RB_DOORBELL_RANGE_UPPER__DOORBELL_RANGE_UPPER_MASK); +} + +static int gfx_v11_0_cp_gfx_resume(struct amdgpu_device *adev) +{ + struct amdgpu_ring *ring; + u32 tmp; + u32 rb_bufsz; + u64 rb_addr, rptr_addr, wptr_gpu_addr; + u32 i; + + /* Set the write pointer delay */ + WREG32_SOC15(GC, 0, regCP_RB_WPTR_DELAY, 0); + + /* set the RB to use vmid 0 */ + WREG32_SOC15(GC, 0, regCP_RB_VMID, 0); + + /* Init gfx ring 0 for pipe 0 */ + mutex_lock(&adev->srbm_mutex); + gfx_v11_0_cp_gfx_switch_pipe(adev, PIPE_ID0); + + /* Set ring buffer size */ + ring = &adev->gfx.gfx_ring[0]; + rb_bufsz = order_base_2(ring->ring_size / 8); + tmp = REG_SET_FIELD(0, CP_RB0_CNTL, RB_BUFSZ, rb_bufsz); + tmp = REG_SET_FIELD(tmp, CP_RB0_CNTL, RB_BLKSZ, rb_bufsz - 2); + WREG32_SOC15(GC, 0, regCP_RB0_CNTL, tmp); + + /* Initialize the ring buffer's write pointers */ + ring->wptr = 0; + WREG32_SOC15(GC, 0, regCP_RB0_WPTR, lower_32_bits(ring->wptr)); + WREG32_SOC15(GC, 0, regCP_RB0_WPTR_HI, upper_32_bits(ring->wptr)); + + /* set the wb address wether it's enabled or not */ + rptr_addr = ring->rptr_gpu_addr; + WREG32_SOC15(GC, 0, regCP_RB0_RPTR_ADDR, lower_32_bits(rptr_addr)); + WREG32_SOC15(GC, 0, regCP_RB0_RPTR_ADDR_HI, upper_32_bits(rptr_addr) & + CP_RB_RPTR_ADDR_HI__RB_RPTR_ADDR_HI_MASK); + + wptr_gpu_addr = ring->wptr_gpu_addr; + WREG32_SOC15(GC, 0, regCP_RB_WPTR_POLL_ADDR_LO, + lower_32_bits(wptr_gpu_addr)); + WREG32_SOC15(GC, 0, regCP_RB_WPTR_POLL_ADDR_HI, + upper_32_bits(wptr_gpu_addr)); + + mdelay(1); + WREG32_SOC15(GC, 0, regCP_RB0_CNTL, tmp); + + rb_addr = ring->gpu_addr >> 8; + WREG32_SOC15(GC, 0, regCP_RB0_BASE, rb_addr); + WREG32_SOC15(GC, 0, regCP_RB0_BASE_HI, upper_32_bits(rb_addr)); + + WREG32_SOC15(GC, 0, regCP_RB_ACTIVE, 1); + + gfx_v11_0_cp_gfx_set_doorbell(adev, ring); + mutex_unlock(&adev->srbm_mutex); + + /* Init gfx ring 1 for pipe 1 */ + if (adev->gfx.num_gfx_rings > 1) { + mutex_lock(&adev->srbm_mutex); + gfx_v11_0_cp_gfx_switch_pipe(adev, PIPE_ID1); + /* maximum supported gfx ring is 2 */ + ring = &adev->gfx.gfx_ring[1]; + rb_bufsz = order_base_2(ring->ring_size / 8); + tmp = REG_SET_FIELD(0, CP_RB1_CNTL, RB_BUFSZ, rb_bufsz); + tmp = REG_SET_FIELD(tmp, CP_RB1_CNTL, RB_BLKSZ, rb_bufsz - 2); + WREG32_SOC15(GC, 0, regCP_RB1_CNTL, tmp); + /* Initialize the ring buffer's write pointers */ + ring->wptr = 0; + WREG32_SOC15(GC, 0, regCP_RB1_WPTR, lower_32_bits(ring->wptr)); + WREG32_SOC15(GC, 0, regCP_RB1_WPTR_HI, upper_32_bits(ring->wptr)); + /* Set the wb address wether it's enabled or not */ + rptr_addr = ring->rptr_gpu_addr; + WREG32_SOC15(GC, 0, regCP_RB1_RPTR_ADDR, lower_32_bits(rptr_addr)); + WREG32_SOC15(GC, 0, regCP_RB1_RPTR_ADDR_HI, upper_32_bits(rptr_addr) & + CP_RB1_RPTR_ADDR_HI__RB_RPTR_ADDR_HI_MASK); + wptr_gpu_addr = ring->wptr_gpu_addr; + WREG32_SOC15(GC, 0, regCP_RB_WPTR_POLL_ADDR_LO, + lower_32_bits(wptr_gpu_addr)); + WREG32_SOC15(GC, 0, regCP_RB_WPTR_POLL_ADDR_HI, + upper_32_bits(wptr_gpu_addr)); + + mdelay(1); + WREG32_SOC15(GC, 0, regCP_RB1_CNTL, tmp); + + rb_addr = ring->gpu_addr >> 8; + WREG32_SOC15(GC, 0, regCP_RB1_BASE, rb_addr); + WREG32_SOC15(GC, 0, regCP_RB1_BASE_HI, upper_32_bits(rb_addr)); + WREG32_SOC15(GC, 0, regCP_RB1_ACTIVE, 1); + + gfx_v11_0_cp_gfx_set_doorbell(adev, ring); + mutex_unlock(&adev->srbm_mutex); + } + /* Switch to pipe 0 */ + mutex_lock(&adev->srbm_mutex); + gfx_v11_0_cp_gfx_switch_pipe(adev, PIPE_ID0); + mutex_unlock(&adev->srbm_mutex); + + /* start the ring */ + gfx_v11_0_cp_gfx_start(adev); + + for (i = 0; i < adev->gfx.num_gfx_rings; i++) { + ring = &adev->gfx.gfx_ring[i]; + ring->sched.ready = true; + } + + return 0; +} + +static void gfx_v11_0_cp_compute_enable(struct amdgpu_device *adev, bool enable) +{ + u32 data; + + if (adev->gfx.rs64_enable) { + data = RREG32_SOC15(GC, 0, regCP_MEC_RS64_CNTL); + data = REG_SET_FIELD(data, CP_MEC_RS64_CNTL, MEC_INVALIDATE_ICACHE, + enable ? 0 : 1); + data = REG_SET_FIELD(data, CP_MEC_RS64_CNTL, MEC_PIPE0_RESET, + enable ? 0 : 1); + data = REG_SET_FIELD(data, CP_MEC_RS64_CNTL, MEC_PIPE1_RESET, + enable ? 0 : 1); + data = REG_SET_FIELD(data, CP_MEC_RS64_CNTL, MEC_PIPE2_RESET, + enable ? 0 : 1); + data = REG_SET_FIELD(data, CP_MEC_RS64_CNTL, MEC_PIPE3_RESET, + enable ? 0 : 1); + data = REG_SET_FIELD(data, CP_MEC_RS64_CNTL, MEC_PIPE0_ACTIVE, + enable ? 1 : 0); + data = REG_SET_FIELD(data, CP_MEC_RS64_CNTL, MEC_PIPE1_ACTIVE, + enable ? 1 : 0); + data = REG_SET_FIELD(data, CP_MEC_RS64_CNTL, MEC_PIPE2_ACTIVE, + enable ? 1 : 0); + data = REG_SET_FIELD(data, CP_MEC_RS64_CNTL, MEC_PIPE3_ACTIVE, + enable ? 1 : 0); + data = REG_SET_FIELD(data, CP_MEC_RS64_CNTL, MEC_HALT, + enable ? 0 : 1); + WREG32_SOC15(GC, 0, regCP_MEC_RS64_CNTL, data); + } else { + data = RREG32_SOC15(GC, 0, regCP_MEC_CNTL); + + if (enable) { + data = REG_SET_FIELD(data, CP_MEC_CNTL, MEC_ME1_HALT, 0); + if (!adev->enable_mes_kiq) + data = REG_SET_FIELD(data, CP_MEC_CNTL, + MEC_ME2_HALT, 0); + } else { + data = REG_SET_FIELD(data, CP_MEC_CNTL, MEC_ME1_HALT, 1); + data = REG_SET_FIELD(data, CP_MEC_CNTL, MEC_ME2_HALT, 1); + } + WREG32_SOC15(GC, 0, regCP_MEC_CNTL, data); + } + + adev->gfx.kiq.ring.sched.ready = enable; + + udelay(50); +} + +static int gfx_v11_0_cp_compute_load_microcode(struct amdgpu_device *adev) +{ + const struct gfx_firmware_header_v1_0 *mec_hdr; + const __le32 *fw_data; + unsigned i, fw_size; + u32 *fw = NULL; + int r; + + if (!adev->gfx.mec_fw) + return -EINVAL; + + gfx_v11_0_cp_compute_enable(adev, false); + + mec_hdr = (const struct gfx_firmware_header_v1_0 *)adev->gfx.mec_fw->data; + amdgpu_ucode_print_gfx_hdr(&mec_hdr->header); + + fw_data = (const __le32 *) + (adev->gfx.mec_fw->data + + le32_to_cpu(mec_hdr->header.ucode_array_offset_bytes)); + fw_size = le32_to_cpu(mec_hdr->header.ucode_size_bytes); + + r = amdgpu_bo_create_reserved(adev, mec_hdr->header.ucode_size_bytes, + PAGE_SIZE, AMDGPU_GEM_DOMAIN_GTT, + &adev->gfx.mec.mec_fw_obj, + &adev->gfx.mec.mec_fw_gpu_addr, + (void **)&fw); + if (r) { + dev_err(adev->dev, "(%d) failed to create mec fw bo\n", r); + gfx_v11_0_mec_fini(adev); + return r; + } + + memcpy(fw, fw_data, fw_size); + + amdgpu_bo_kunmap(adev->gfx.mec.mec_fw_obj); + amdgpu_bo_unreserve(adev->gfx.mec.mec_fw_obj); + + gfx_v11_0_config_mec_cache(adev, adev->gfx.mec.mec_fw_gpu_addr); + + /* MEC1 */ + WREG32_SOC15(GC, 0, regCP_MEC_ME1_UCODE_ADDR, 0); + + for (i = 0; i < mec_hdr->jt_size; i++) + WREG32_SOC15(GC, 0, regCP_MEC_ME1_UCODE_DATA, + le32_to_cpup(fw_data + mec_hdr->jt_offset + i)); + + WREG32_SOC15(GC, 0, regCP_MEC_ME1_UCODE_ADDR, adev->gfx.mec_fw_version); + + return 0; +} + +static int gfx_v11_0_cp_compute_load_microcode_rs64(struct amdgpu_device *adev) +{ + const struct gfx_firmware_header_v2_0 *mec_hdr; + const __le32 *fw_ucode, *fw_data; + u32 tmp, fw_ucode_size, fw_data_size; + u32 i, usec_timeout = 50000; /* Wait for 50 ms */ + u32 *fw_ucode_ptr, *fw_data_ptr; + int r; + + if (!adev->gfx.mec_fw) + return -EINVAL; + + gfx_v11_0_cp_compute_enable(adev, false); + + mec_hdr = (const struct gfx_firmware_header_v2_0 *)adev->gfx.mec_fw->data; + amdgpu_ucode_print_gfx_hdr(&mec_hdr->header); + + fw_ucode = (const __le32 *) (adev->gfx.mec_fw->data + + le32_to_cpu(mec_hdr->ucode_offset_bytes)); + fw_ucode_size = le32_to_cpu(mec_hdr->ucode_size_bytes); + + fw_data = (const __le32 *) (adev->gfx.mec_fw->data + + le32_to_cpu(mec_hdr->data_offset_bytes)); + fw_data_size = le32_to_cpu(mec_hdr->data_size_bytes); + + r = amdgpu_bo_create_reserved(adev, fw_ucode_size, + 64 * 1024, AMDGPU_GEM_DOMAIN_VRAM, + &adev->gfx.mec.mec_fw_obj, + &adev->gfx.mec.mec_fw_gpu_addr, + (void **)&fw_ucode_ptr); + if (r) { + dev_err(adev->dev, "(%d) failed to create mec fw ucode bo\n", r); + gfx_v11_0_mec_fini(adev); + return r; + } + + r = amdgpu_bo_create_reserved(adev, fw_data_size, + 64 * 1024, AMDGPU_GEM_DOMAIN_VRAM, + &adev->gfx.mec.mec_fw_data_obj, + &adev->gfx.mec.mec_fw_data_gpu_addr, + (void **)&fw_data_ptr); + if (r) { + dev_err(adev->dev, "(%d) failed to create mec fw ucode bo\n", r); + gfx_v11_0_mec_fini(adev); + return r; + } + + memcpy(fw_ucode_ptr, fw_ucode, fw_ucode_size); + memcpy(fw_data_ptr, fw_data, fw_data_size); + + amdgpu_bo_kunmap(adev->gfx.mec.mec_fw_obj); + amdgpu_bo_kunmap(adev->gfx.mec.mec_fw_data_obj); + amdgpu_bo_unreserve(adev->gfx.mec.mec_fw_obj); + amdgpu_bo_unreserve(adev->gfx.mec.mec_fw_data_obj); + + tmp = RREG32_SOC15(GC, 0, regCP_CPC_IC_BASE_CNTL); + tmp = REG_SET_FIELD(tmp, CP_CPC_IC_BASE_CNTL, VMID, 0); + tmp = REG_SET_FIELD(tmp, CP_CPC_IC_BASE_CNTL, EXE_DISABLE, 0); + tmp = REG_SET_FIELD(tmp, CP_CPC_IC_BASE_CNTL, CACHE_POLICY, 0); + WREG32_SOC15(GC, 0, regCP_CPC_IC_BASE_CNTL, tmp); + + tmp = RREG32_SOC15(GC, 0, regCP_MEC_DC_BASE_CNTL); + tmp = REG_SET_FIELD(tmp, CP_MEC_DC_BASE_CNTL, VMID, 0); + tmp = REG_SET_FIELD(tmp, CP_MEC_DC_BASE_CNTL, CACHE_POLICY, 0); + WREG32_SOC15(GC, 0, regCP_MEC_DC_BASE_CNTL, tmp); + + mutex_lock(&adev->srbm_mutex); + for (i = 0; i < adev->gfx.mec.num_pipe_per_mec; i++) { + soc21_grbm_select(adev, 1, i, 0, 0); + + WREG32_SOC15(GC, 0, regCP_MEC_MDBASE_LO, adev->gfx.mec.mec_fw_data_gpu_addr); + WREG32_SOC15(GC, 0, regCP_MEC_MDBASE_HI, + upper_32_bits(adev->gfx.mec.mec_fw_data_gpu_addr)); + + WREG32_SOC15(GC, 0, regCP_MEC_RS64_PRGRM_CNTR_START, + mec_hdr->ucode_start_addr_lo >> 2 | + mec_hdr->ucode_start_addr_hi << 30); + WREG32_SOC15(GC, 0, regCP_MEC_RS64_PRGRM_CNTR_START_HI, + mec_hdr->ucode_start_addr_hi >> 2); + + WREG32_SOC15(GC, 0, regCP_CPC_IC_BASE_LO, adev->gfx.mec.mec_fw_gpu_addr); + WREG32_SOC15(GC, 0, regCP_CPC_IC_BASE_HI, + upper_32_bits(adev->gfx.mec.mec_fw_gpu_addr)); + } + mutex_unlock(&adev->srbm_mutex); + soc21_grbm_select(adev, 0, 0, 0, 0); + + /* Trigger an invalidation of the L1 instruction caches */ + tmp = RREG32_SOC15(GC, 0, regCP_MEC_DC_OP_CNTL); + tmp = REG_SET_FIELD(tmp, CP_MEC_DC_OP_CNTL, INVALIDATE_DCACHE, 1); + WREG32_SOC15(GC, 0, regCP_MEC_DC_OP_CNTL, tmp); + + /* Wait for invalidation complete */ + for (i = 0; i < usec_timeout; i++) { + tmp = RREG32_SOC15(GC, 0, regCP_MEC_DC_OP_CNTL); + if (1 == REG_GET_FIELD(tmp, CP_MEC_DC_OP_CNTL, + INVALIDATE_DCACHE_COMPLETE)) + break; + udelay(1); + } + + if (i >= usec_timeout) { + dev_err(adev->dev, "failed to invalidate instruction cache\n"); + return -EINVAL; + } + + /* Trigger an invalidation of the L1 instruction caches */ + tmp = RREG32_SOC15(GC, 0, regCP_CPC_IC_OP_CNTL); + tmp = REG_SET_FIELD(tmp, CP_CPC_IC_OP_CNTL, INVALIDATE_CACHE, 1); + WREG32_SOC15(GC, 0, regCP_CPC_IC_OP_CNTL, tmp); + + /* Wait for invalidation complete */ + for (i = 0; i < usec_timeout; i++) { + tmp = RREG32_SOC15(GC, 0, regCP_CPC_IC_OP_CNTL); + if (1 == REG_GET_FIELD(tmp, CP_CPC_IC_OP_CNTL, + INVALIDATE_CACHE_COMPLETE)) + break; + udelay(1); + } + + if (i >= usec_timeout) { + dev_err(adev->dev, "failed to invalidate instruction cache\n"); + return -EINVAL; + } + + return 0; +} + +static void gfx_v11_0_kiq_setting(struct amdgpu_ring *ring) +{ + uint32_t tmp; + struct amdgpu_device *adev = ring->adev; + + /* tell RLC which is KIQ queue */ + tmp = RREG32_SOC15(GC, 0, regRLC_CP_SCHEDULERS); + tmp &= 0xffffff00; + tmp |= (ring->me << 5) | (ring->pipe << 3) | (ring->queue); + WREG32_SOC15(GC, 0, regRLC_CP_SCHEDULERS, tmp); + tmp |= 0x80; + WREG32_SOC15(GC, 0, regRLC_CP_SCHEDULERS, tmp); +} + +static void gfx_v11_0_cp_set_doorbell_range(struct amdgpu_device *adev) +{ + /* set graphics engine doorbell range */ + WREG32_SOC15(GC, 0, regCP_RB_DOORBELL_RANGE_LOWER, + (adev->doorbell_index.gfx_ring0 * 2) << 2); + WREG32_SOC15(GC, 0, regCP_RB_DOORBELL_RANGE_UPPER, + (adev->doorbell_index.gfx_userqueue_end * 2) << 2); + + /* set compute engine doorbell range */ + WREG32_SOC15(GC, 0, regCP_MEC_DOORBELL_RANGE_LOWER, + (adev->doorbell_index.kiq * 2) << 2); + WREG32_SOC15(GC, 0, regCP_MEC_DOORBELL_RANGE_UPPER, + (adev->doorbell_index.userqueue_end * 2) << 2); +} + +static int gfx_v11_0_gfx_mqd_init(struct amdgpu_device *adev, void *m, + struct amdgpu_mqd_prop *prop) +{ + struct v11_gfx_mqd *mqd = m; + uint64_t hqd_gpu_addr, wb_gpu_addr; + uint32_t tmp; + uint32_t rb_bufsz; + + /* set up gfx hqd wptr */ + mqd->cp_gfx_hqd_wptr = 0; + mqd->cp_gfx_hqd_wptr_hi = 0; + + /* set the pointer to the MQD */ + mqd->cp_mqd_base_addr = prop->mqd_gpu_addr & 0xfffffffc; + mqd->cp_mqd_base_addr_hi = upper_32_bits(prop->mqd_gpu_addr); + + /* set up mqd control */ + tmp = RREG32_SOC15(GC, 0, regCP_GFX_MQD_CONTROL); + tmp = REG_SET_FIELD(tmp, CP_GFX_MQD_CONTROL, VMID, 0); + tmp = REG_SET_FIELD(tmp, CP_GFX_MQD_CONTROL, PRIV_STATE, 1); + tmp = REG_SET_FIELD(tmp, CP_GFX_MQD_CONTROL, CACHE_POLICY, 0); + mqd->cp_gfx_mqd_control = tmp; + + /* set up gfx_hqd_vimd with 0x0 to indicate the ring buffer's vmid */ + tmp = RREG32_SOC15(GC, 0, regCP_GFX_HQD_VMID); + tmp = REG_SET_FIELD(tmp, CP_GFX_HQD_VMID, VMID, 0); + mqd->cp_gfx_hqd_vmid = 0; + + /* set up default queue priority level + * 0x0 = low priority, 0x1 = high priority */ + tmp = RREG32_SOC15(GC, 0, regCP_GFX_HQD_QUEUE_PRIORITY); + tmp = REG_SET_FIELD(tmp, CP_GFX_HQD_QUEUE_PRIORITY, PRIORITY_LEVEL, 0); + mqd->cp_gfx_hqd_queue_priority = tmp; + + /* set up time quantum */ + tmp = RREG32_SOC15(GC, 0, regCP_GFX_HQD_QUANTUM); + tmp = REG_SET_FIELD(tmp, CP_GFX_HQD_QUANTUM, QUANTUM_EN, 1); + mqd->cp_gfx_hqd_quantum = tmp; + + /* set up gfx hqd base. this is similar as CP_RB_BASE */ + hqd_gpu_addr = prop->hqd_base_gpu_addr >> 8; + mqd->cp_gfx_hqd_base = hqd_gpu_addr; + mqd->cp_gfx_hqd_base_hi = upper_32_bits(hqd_gpu_addr); + + /* set up hqd_rptr_addr/_hi, similar as CP_RB_RPTR */ + wb_gpu_addr = prop->rptr_gpu_addr; + mqd->cp_gfx_hqd_rptr_addr = wb_gpu_addr & 0xfffffffc; + mqd->cp_gfx_hqd_rptr_addr_hi = + upper_32_bits(wb_gpu_addr) & 0xffff; + + /* set up rb_wptr_poll addr */ + wb_gpu_addr = prop->wptr_gpu_addr; + mqd->cp_rb_wptr_poll_addr_lo = wb_gpu_addr & 0xfffffffc; + mqd->cp_rb_wptr_poll_addr_hi = upper_32_bits(wb_gpu_addr) & 0xffff; + + /* set up the gfx_hqd_control, similar as CP_RB0_CNTL */ + rb_bufsz = order_base_2(prop->queue_size / 4) - 1; + tmp = RREG32_SOC15(GC, 0, regCP_GFX_HQD_CNTL); + tmp = REG_SET_FIELD(tmp, CP_GFX_HQD_CNTL, RB_BUFSZ, rb_bufsz); + tmp = REG_SET_FIELD(tmp, CP_GFX_HQD_CNTL, RB_BLKSZ, rb_bufsz - 2); +#ifdef __BIG_ENDIAN + tmp = REG_SET_FIELD(tmp, CP_GFX_HQD_CNTL, BUF_SWAP, 1); +#endif + mqd->cp_gfx_hqd_cntl = tmp; + + /* set up cp_doorbell_control */ + tmp = RREG32_SOC15(GC, 0, regCP_RB_DOORBELL_CONTROL); + if (prop->use_doorbell) { + tmp = REG_SET_FIELD(tmp, CP_RB_DOORBELL_CONTROL, + DOORBELL_OFFSET, prop->doorbell_index); + tmp = REG_SET_FIELD(tmp, CP_RB_DOORBELL_CONTROL, + DOORBELL_EN, 1); + } else + tmp = REG_SET_FIELD(tmp, CP_RB_DOORBELL_CONTROL, + DOORBELL_EN, 0); + mqd->cp_rb_doorbell_control = tmp; + + /* reset read and write pointers, similar to CP_RB0_WPTR/_RPTR */ + mqd->cp_gfx_hqd_rptr = RREG32_SOC15(GC, 0, regCP_GFX_HQD_RPTR); + + /* active the queue */ + mqd->cp_gfx_hqd_active = 1; + + return 0; +} + +#ifdef BRING_UP_DEBUG +static int gfx_v11_0_gfx_queue_init_register(struct amdgpu_ring *ring) +{ + struct amdgpu_device *adev = ring->adev; + struct v11_gfx_mqd *mqd = ring->mqd_ptr; + + /* set mmCP_GFX_HQD_WPTR/_HI to 0 */ + WREG32_SOC15(GC, 0, regCP_GFX_HQD_WPTR, mqd->cp_gfx_hqd_wptr); + WREG32_SOC15(GC, 0, regCP_GFX_HQD_WPTR_HI, mqd->cp_gfx_hqd_wptr_hi); + + /* set GFX_MQD_BASE */ + WREG32_SOC15(GC, 0, regCP_MQD_BASE_ADDR, mqd->cp_mqd_base_addr); + WREG32_SOC15(GC, 0, regCP_MQD_BASE_ADDR_HI, mqd->cp_mqd_base_addr_hi); + + /* set GFX_MQD_CONTROL */ + WREG32_SOC15(GC, 0, regCP_GFX_MQD_CONTROL, mqd->cp_gfx_mqd_control); + + /* set GFX_HQD_VMID to 0 */ + WREG32_SOC15(GC, 0, regCP_GFX_HQD_VMID, mqd->cp_gfx_hqd_vmid); + + WREG32_SOC15(GC, 0, regCP_GFX_HQD_QUEUE_PRIORITY, + mqd->cp_gfx_hqd_queue_priority); + WREG32_SOC15(GC, 0, regCP_GFX_HQD_QUANTUM, mqd->cp_gfx_hqd_quantum); + + /* set GFX_HQD_BASE, similar as CP_RB_BASE */ + WREG32_SOC15(GC, 0, regCP_GFX_HQD_BASE, mqd->cp_gfx_hqd_base); + WREG32_SOC15(GC, 0, regCP_GFX_HQD_BASE_HI, mqd->cp_gfx_hqd_base_hi); + + /* set GFX_HQD_RPTR_ADDR, similar as CP_RB_RPTR */ + WREG32_SOC15(GC, 0, regCP_GFX_HQD_RPTR_ADDR, mqd->cp_gfx_hqd_rptr_addr); + WREG32_SOC15(GC, 0, regCP_GFX_HQD_RPTR_ADDR_HI, mqd->cp_gfx_hqd_rptr_addr_hi); + + /* set GFX_HQD_CNTL, similar as CP_RB_CNTL */ + WREG32_SOC15(GC, 0, regCP_GFX_HQD_CNTL, mqd->cp_gfx_hqd_cntl); + + /* set RB_WPTR_POLL_ADDR */ + WREG32_SOC15(GC, 0, regCP_RB_WPTR_POLL_ADDR_LO, mqd->cp_rb_wptr_poll_addr_lo); + WREG32_SOC15(GC, 0, regCP_RB_WPTR_POLL_ADDR_HI, mqd->cp_rb_wptr_poll_addr_hi); + + /* set RB_DOORBELL_CONTROL */ + WREG32_SOC15(GC, 0, regCP_RB_DOORBELL_CONTROL, mqd->cp_rb_doorbell_control); + + /* active the queue */ + WREG32_SOC15(GC, 0, regCP_GFX_HQD_ACTIVE, mqd->cp_gfx_hqd_active); + + return 0; +} +#endif + +static int gfx_v11_0_gfx_init_queue(struct amdgpu_ring *ring) +{ + struct amdgpu_device *adev = ring->adev; + struct v11_gfx_mqd *mqd = ring->mqd_ptr; + int mqd_idx = ring - &adev->gfx.gfx_ring[0]; + + if (!amdgpu_in_reset(adev) && !adev->in_suspend) { + memset((void *)mqd, 0, sizeof(*mqd)); + mutex_lock(&adev->srbm_mutex); + soc21_grbm_select(adev, ring->me, ring->pipe, ring->queue, 0); + amdgpu_ring_init_mqd(ring); +#ifdef BRING_UP_DEBUG + gfx_v11_0_gfx_queue_init_register(ring); +#endif + soc21_grbm_select(adev, 0, 0, 0, 0); + mutex_unlock(&adev->srbm_mutex); + if (adev->gfx.me.mqd_backup[mqd_idx]) + memcpy(adev->gfx.me.mqd_backup[mqd_idx], mqd, sizeof(*mqd)); + } else if (amdgpu_in_reset(adev)) { + /* reset mqd with the backup copy */ + if (adev->gfx.me.mqd_backup[mqd_idx]) + memcpy(mqd, adev->gfx.me.mqd_backup[mqd_idx], sizeof(*mqd)); + /* reset the ring */ + ring->wptr = 0; + *ring->wptr_cpu_addr = 0; + amdgpu_ring_clear_ring(ring); +#ifdef BRING_UP_DEBUG + mutex_lock(&adev->srbm_mutex); + soc21_grbm_select(adev, ring->me, ring->pipe, ring->queue, 0); + gfx_v11_0_gfx_queue_init_register(ring); + soc21_grbm_select(adev, 0, 0, 0, 0); + mutex_unlock(&adev->srbm_mutex); +#endif + } else { + amdgpu_ring_clear_ring(ring); + } + + return 0; +} + +#ifndef BRING_UP_DEBUG +static int gfx_v11_0_kiq_enable_kgq(struct amdgpu_device *adev) +{ + struct amdgpu_kiq *kiq = &adev->gfx.kiq; + struct amdgpu_ring *kiq_ring = &adev->gfx.kiq.ring; + int r, i; + + if (!kiq->pmf || !kiq->pmf->kiq_map_queues) + return -EINVAL; + + r = amdgpu_ring_alloc(kiq_ring, kiq->pmf->map_queues_size * + adev->gfx.num_gfx_rings); + if (r) { + DRM_ERROR("Failed to lock KIQ (%d).\n", r); + return r; + } + + for (i = 0; i < adev->gfx.num_gfx_rings; i++) + kiq->pmf->kiq_map_queues(kiq_ring, &adev->gfx.gfx_ring[i]); + + return amdgpu_ring_test_helper(kiq_ring); +} +#endif + +static int gfx_v11_0_cp_async_gfx_ring_resume(struct amdgpu_device *adev) +{ + int r, i; + struct amdgpu_ring *ring; + + for (i = 0; i < adev->gfx.num_gfx_rings; i++) { + ring = &adev->gfx.gfx_ring[i]; + + r = amdgpu_bo_reserve(ring->mqd_obj, false); + if (unlikely(r != 0)) + goto done; + + r = amdgpu_bo_kmap(ring->mqd_obj, (void **)&ring->mqd_ptr); + if (!r) { + r = gfx_v11_0_gfx_init_queue(ring); + amdgpu_bo_kunmap(ring->mqd_obj); + ring->mqd_ptr = NULL; + } + amdgpu_bo_unreserve(ring->mqd_obj); + if (r) + goto done; + } +#ifndef BRING_UP_DEBUG + r = gfx_v11_0_kiq_enable_kgq(adev); + if (r) + goto done; +#endif + r = gfx_v11_0_cp_gfx_start(adev); + if (r) + goto done; + + for (i = 0; i < adev->gfx.num_gfx_rings; i++) { + ring = &adev->gfx.gfx_ring[i]; + ring->sched.ready = true; + } +done: + return r; +} + +static int gfx_v11_0_compute_mqd_init(struct amdgpu_device *adev, void *m, + struct amdgpu_mqd_prop *prop) +{ + struct v11_compute_mqd *mqd = m; + uint64_t hqd_gpu_addr, wb_gpu_addr, eop_base_addr; + uint32_t tmp; + + mqd->header = 0xC0310800; + mqd->compute_pipelinestat_enable = 0x00000001; + mqd->compute_static_thread_mgmt_se0 = 0xffffffff; + mqd->compute_static_thread_mgmt_se1 = 0xffffffff; + mqd->compute_static_thread_mgmt_se2 = 0xffffffff; + mqd->compute_static_thread_mgmt_se3 = 0xffffffff; + mqd->compute_misc_reserved = 0x00000007; + + eop_base_addr = prop->eop_gpu_addr >> 8; + mqd->cp_hqd_eop_base_addr_lo = eop_base_addr; + mqd->cp_hqd_eop_base_addr_hi = upper_32_bits(eop_base_addr); + + /* set the EOP size, register value is 2^(EOP_SIZE+1) dwords */ + tmp = RREG32_SOC15(GC, 0, regCP_HQD_EOP_CONTROL); + tmp = REG_SET_FIELD(tmp, CP_HQD_EOP_CONTROL, EOP_SIZE, + (order_base_2(GFX11_MEC_HPD_SIZE / 4) - 1)); + + mqd->cp_hqd_eop_control = tmp; + + /* enable doorbell? */ + tmp = RREG32_SOC15(GC, 0, regCP_HQD_PQ_DOORBELL_CONTROL); + + if (prop->use_doorbell) { + tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_DOORBELL_CONTROL, + DOORBELL_OFFSET, prop->doorbell_index); + tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_DOORBELL_CONTROL, + DOORBELL_EN, 1); + tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_DOORBELL_CONTROL, + DOORBELL_SOURCE, 0); + tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_DOORBELL_CONTROL, + DOORBELL_HIT, 0); + } else { + tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_DOORBELL_CONTROL, + DOORBELL_EN, 0); + } + + mqd->cp_hqd_pq_doorbell_control = tmp; + + /* disable the queue if it's active */ + mqd->cp_hqd_dequeue_request = 0; + mqd->cp_hqd_pq_rptr = 0; + mqd->cp_hqd_pq_wptr_lo = 0; + mqd->cp_hqd_pq_wptr_hi = 0; + + /* set the pointer to the MQD */ + mqd->cp_mqd_base_addr_lo = prop->mqd_gpu_addr & 0xfffffffc; + mqd->cp_mqd_base_addr_hi = upper_32_bits(prop->mqd_gpu_addr); + + /* set MQD vmid to 0 */ + tmp = RREG32_SOC15(GC, 0, regCP_MQD_CONTROL); + tmp = REG_SET_FIELD(tmp, CP_MQD_CONTROL, VMID, 0); + mqd->cp_mqd_control = tmp; + + /* set the pointer to the HQD, this is similar CP_RB0_BASE/_HI */ + hqd_gpu_addr = prop->hqd_base_gpu_addr >> 8; + mqd->cp_hqd_pq_base_lo = hqd_gpu_addr; + mqd->cp_hqd_pq_base_hi = upper_32_bits(hqd_gpu_addr); + + /* set up the HQD, this is similar to CP_RB0_CNTL */ + tmp = RREG32_SOC15(GC, 0, regCP_HQD_PQ_CONTROL); + tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, QUEUE_SIZE, + (order_base_2(prop->queue_size / 4) - 1)); + tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, RPTR_BLOCK_SIZE, + (order_base_2(AMDGPU_GPU_PAGE_SIZE / 4) - 1)); + tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, UNORD_DISPATCH, 0); + tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, TUNNEL_DISPATCH, 0); + tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, PRIV_STATE, 1); + tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, KMD_QUEUE, 1); + mqd->cp_hqd_pq_control = tmp; + + /* set the wb address whether it's enabled or not */ + wb_gpu_addr = prop->rptr_gpu_addr; + mqd->cp_hqd_pq_rptr_report_addr_lo = wb_gpu_addr & 0xfffffffc; + mqd->cp_hqd_pq_rptr_report_addr_hi = + upper_32_bits(wb_gpu_addr) & 0xffff; + + /* only used if CP_PQ_WPTR_POLL_CNTL.CP_PQ_WPTR_POLL_CNTL__EN_MASK=1 */ + wb_gpu_addr = prop->wptr_gpu_addr; + mqd->cp_hqd_pq_wptr_poll_addr_lo = wb_gpu_addr & 0xfffffffc; + mqd->cp_hqd_pq_wptr_poll_addr_hi = upper_32_bits(wb_gpu_addr) & 0xffff; + + tmp = 0; + /* enable the doorbell if requested */ + if (prop->use_doorbell) { + tmp = RREG32_SOC15(GC, 0, regCP_HQD_PQ_DOORBELL_CONTROL); + tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_DOORBELL_CONTROL, + DOORBELL_OFFSET, prop->doorbell_index); + + tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_DOORBELL_CONTROL, + DOORBELL_EN, 1); + tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_DOORBELL_CONTROL, + DOORBELL_SOURCE, 0); + tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_DOORBELL_CONTROL, + DOORBELL_HIT, 0); + } + + mqd->cp_hqd_pq_doorbell_control = tmp; + + /* reset read and write pointers, similar to CP_RB0_WPTR/_RPTR */ + mqd->cp_hqd_pq_rptr = RREG32_SOC15(GC, 0, regCP_HQD_PQ_RPTR); + + /* set the vmid for the queue */ + mqd->cp_hqd_vmid = 0; + + tmp = RREG32_SOC15(GC, 0, regCP_HQD_PERSISTENT_STATE); + tmp = REG_SET_FIELD(tmp, CP_HQD_PERSISTENT_STATE, PRELOAD_SIZE, 0x55); + mqd->cp_hqd_persistent_state = tmp; + + /* set MIN_IB_AVAIL_SIZE */ + tmp = RREG32_SOC15(GC, 0, regCP_HQD_IB_CONTROL); + tmp = REG_SET_FIELD(tmp, CP_HQD_IB_CONTROL, MIN_IB_AVAIL_SIZE, 3); + mqd->cp_hqd_ib_control = tmp; + + /* set static priority for a compute queue/ring */ + mqd->cp_hqd_pipe_priority = prop->hqd_pipe_priority; + mqd->cp_hqd_queue_priority = prop->hqd_queue_priority; + + mqd->cp_hqd_active = prop->hqd_active; + + return 0; +} + +static int gfx_v11_0_kiq_init_register(struct amdgpu_ring *ring) +{ + struct amdgpu_device *adev = ring->adev; + struct v11_compute_mqd *mqd = ring->mqd_ptr; + int j; + + /* inactivate the queue */ + if (amdgpu_sriov_vf(adev)) + WREG32_SOC15(GC, 0, regCP_HQD_ACTIVE, 0); + + /* disable wptr polling */ + WREG32_FIELD15_PREREG(GC, 0, CP_PQ_WPTR_POLL_CNTL, EN, 0); + + /* write the EOP addr */ + WREG32_SOC15(GC, 0, regCP_HQD_EOP_BASE_ADDR, + mqd->cp_hqd_eop_base_addr_lo); + WREG32_SOC15(GC, 0, regCP_HQD_EOP_BASE_ADDR_HI, + mqd->cp_hqd_eop_base_addr_hi); + + /* set the EOP size, register value is 2^(EOP_SIZE+1) dwords */ + WREG32_SOC15(GC, 0, regCP_HQD_EOP_CONTROL, + mqd->cp_hqd_eop_control); + + /* enable doorbell? */ + WREG32_SOC15(GC, 0, regCP_HQD_PQ_DOORBELL_CONTROL, + mqd->cp_hqd_pq_doorbell_control); + + /* disable the queue if it's active */ + if (RREG32_SOC15(GC, 0, regCP_HQD_ACTIVE) & 1) { + WREG32_SOC15(GC, 0, regCP_HQD_DEQUEUE_REQUEST, 1); + for (j = 0; j < adev->usec_timeout; j++) { + if (!(RREG32_SOC15(GC, 0, regCP_HQD_ACTIVE) & 1)) + break; + udelay(1); + } + WREG32_SOC15(GC, 0, regCP_HQD_DEQUEUE_REQUEST, + mqd->cp_hqd_dequeue_request); + WREG32_SOC15(GC, 0, regCP_HQD_PQ_RPTR, + mqd->cp_hqd_pq_rptr); + WREG32_SOC15(GC, 0, regCP_HQD_PQ_WPTR_LO, + mqd->cp_hqd_pq_wptr_lo); + WREG32_SOC15(GC, 0, regCP_HQD_PQ_WPTR_HI, + mqd->cp_hqd_pq_wptr_hi); + } + + /* set the pointer to the MQD */ + WREG32_SOC15(GC, 0, regCP_MQD_BASE_ADDR, + mqd->cp_mqd_base_addr_lo); + WREG32_SOC15(GC, 0, regCP_MQD_BASE_ADDR_HI, + mqd->cp_mqd_base_addr_hi); + + /* set MQD vmid to 0 */ + WREG32_SOC15(GC, 0, regCP_MQD_CONTROL, + mqd->cp_mqd_control); + + /* set the pointer to the HQD, this is similar CP_RB0_BASE/_HI */ + WREG32_SOC15(GC, 0, regCP_HQD_PQ_BASE, + mqd->cp_hqd_pq_base_lo); + WREG32_SOC15(GC, 0, regCP_HQD_PQ_BASE_HI, + mqd->cp_hqd_pq_base_hi); + + /* set up the HQD, this is similar to CP_RB0_CNTL */ + WREG32_SOC15(GC, 0, regCP_HQD_PQ_CONTROL, + mqd->cp_hqd_pq_control); + + /* set the wb address whether it's enabled or not */ + WREG32_SOC15(GC, 0, regCP_HQD_PQ_RPTR_REPORT_ADDR, + mqd->cp_hqd_pq_rptr_report_addr_lo); + WREG32_SOC15(GC, 0, regCP_HQD_PQ_RPTR_REPORT_ADDR_HI, + mqd->cp_hqd_pq_rptr_report_addr_hi); + + /* only used if CP_PQ_WPTR_POLL_CNTL.CP_PQ_WPTR_POLL_CNTL__EN_MASK=1 */ + WREG32_SOC15(GC, 0, regCP_HQD_PQ_WPTR_POLL_ADDR, + mqd->cp_hqd_pq_wptr_poll_addr_lo); + WREG32_SOC15(GC, 0, regCP_HQD_PQ_WPTR_POLL_ADDR_HI, + mqd->cp_hqd_pq_wptr_poll_addr_hi); + + /* enable the doorbell if requested */ + if (ring->use_doorbell) { + WREG32_SOC15(GC, 0, regCP_MEC_DOORBELL_RANGE_LOWER, + (adev->doorbell_index.kiq * 2) << 2); + WREG32_SOC15(GC, 0, regCP_MEC_DOORBELL_RANGE_UPPER, + (adev->doorbell_index.userqueue_end * 2) << 2); + } + + WREG32_SOC15(GC, 0, regCP_HQD_PQ_DOORBELL_CONTROL, + mqd->cp_hqd_pq_doorbell_control); + + /* reset read and write pointers, similar to CP_RB0_WPTR/_RPTR */ + WREG32_SOC15(GC, 0, regCP_HQD_PQ_WPTR_LO, + mqd->cp_hqd_pq_wptr_lo); + WREG32_SOC15(GC, 0, regCP_HQD_PQ_WPTR_HI, + mqd->cp_hqd_pq_wptr_hi); + + /* set the vmid for the queue */ + WREG32_SOC15(GC, 0, regCP_HQD_VMID, mqd->cp_hqd_vmid); + + WREG32_SOC15(GC, 0, regCP_HQD_PERSISTENT_STATE, + mqd->cp_hqd_persistent_state); + + /* activate the queue */ + WREG32_SOC15(GC, 0, regCP_HQD_ACTIVE, + mqd->cp_hqd_active); + + if (ring->use_doorbell) + WREG32_FIELD15_PREREG(GC, 0, CP_PQ_STATUS, DOORBELL_ENABLE, 1); + + return 0; +} + +static int gfx_v11_0_kiq_init_queue(struct amdgpu_ring *ring) +{ + struct amdgpu_device *adev = ring->adev; + struct v11_compute_mqd *mqd = ring->mqd_ptr; + int mqd_idx = AMDGPU_MAX_COMPUTE_RINGS; + + gfx_v11_0_kiq_setting(ring); + + if (amdgpu_in_reset(adev)) { /* for GPU_RESET case */ + /* reset MQD to a clean status */ + if (adev->gfx.mec.mqd_backup[mqd_idx]) + memcpy(mqd, adev->gfx.mec.mqd_backup[mqd_idx], sizeof(*mqd)); + + /* reset ring buffer */ + ring->wptr = 0; + amdgpu_ring_clear_ring(ring); + + mutex_lock(&adev->srbm_mutex); + soc21_grbm_select(adev, ring->me, ring->pipe, ring->queue, 0); + gfx_v11_0_kiq_init_register(ring); + soc21_grbm_select(adev, 0, 0, 0, 0); + mutex_unlock(&adev->srbm_mutex); + } else { + memset((void *)mqd, 0, sizeof(*mqd)); + mutex_lock(&adev->srbm_mutex); + soc21_grbm_select(adev, ring->me, ring->pipe, ring->queue, 0); + amdgpu_ring_init_mqd(ring); + gfx_v11_0_kiq_init_register(ring); + soc21_grbm_select(adev, 0, 0, 0, 0); + mutex_unlock(&adev->srbm_mutex); + + if (adev->gfx.mec.mqd_backup[mqd_idx]) + memcpy(adev->gfx.mec.mqd_backup[mqd_idx], mqd, sizeof(*mqd)); + } + + return 0; +} + +static int gfx_v11_0_kcq_init_queue(struct amdgpu_ring *ring) +{ + struct amdgpu_device *adev = ring->adev; + struct v11_compute_mqd *mqd = ring->mqd_ptr; + int mqd_idx = ring - &adev->gfx.compute_ring[0]; + + if (!amdgpu_in_reset(adev) && !adev->in_suspend) { + memset((void *)mqd, 0, sizeof(*mqd)); + mutex_lock(&adev->srbm_mutex); + soc21_grbm_select(adev, ring->me, ring->pipe, ring->queue, 0); + amdgpu_ring_init_mqd(ring); + soc21_grbm_select(adev, 0, 0, 0, 0); + mutex_unlock(&adev->srbm_mutex); + + if (adev->gfx.mec.mqd_backup[mqd_idx]) + memcpy(adev->gfx.mec.mqd_backup[mqd_idx], mqd, sizeof(*mqd)); + } else if (amdgpu_in_reset(adev)) { /* for GPU_RESET case */ + /* reset MQD to a clean status */ + if (adev->gfx.mec.mqd_backup[mqd_idx]) + memcpy(mqd, adev->gfx.mec.mqd_backup[mqd_idx], sizeof(*mqd)); + + /* reset ring buffer */ + ring->wptr = 0; + atomic64_set((atomic64_t *)ring->wptr_cpu_addr, 0); + amdgpu_ring_clear_ring(ring); + } else { + amdgpu_ring_clear_ring(ring); + } + + return 0; +} + +static int gfx_v11_0_kiq_resume(struct amdgpu_device *adev) +{ + struct amdgpu_ring *ring; + int r; + + ring = &adev->gfx.kiq.ring; + + r = amdgpu_bo_reserve(ring->mqd_obj, false); + if (unlikely(r != 0)) + return r; + + r = amdgpu_bo_kmap(ring->mqd_obj, (void **)&ring->mqd_ptr); + if (unlikely(r != 0)) { + amdgpu_bo_unreserve(ring->mqd_obj); + return r; + } + + gfx_v11_0_kiq_init_queue(ring); + amdgpu_bo_kunmap(ring->mqd_obj); + ring->mqd_ptr = NULL; + amdgpu_bo_unreserve(ring->mqd_obj); + ring->sched.ready = true; + return 0; +} + +static int gfx_v11_0_kcq_resume(struct amdgpu_device *adev) +{ + struct amdgpu_ring *ring = NULL; + int r = 0, i; + + if (!amdgpu_async_gfx_ring) + gfx_v11_0_cp_compute_enable(adev, true); + + for (i = 0; i < adev->gfx.num_compute_rings; i++) { + ring = &adev->gfx.compute_ring[i]; + + r = amdgpu_bo_reserve(ring->mqd_obj, false); + if (unlikely(r != 0)) + goto done; + r = amdgpu_bo_kmap(ring->mqd_obj, (void **)&ring->mqd_ptr); + if (!r) { + r = gfx_v11_0_kcq_init_queue(ring); + amdgpu_bo_kunmap(ring->mqd_obj); + ring->mqd_ptr = NULL; + } + amdgpu_bo_unreserve(ring->mqd_obj); + if (r) + goto done; + } + + r = amdgpu_gfx_enable_kcq(adev); +done: + return r; +} + +static int gfx_v11_0_cp_resume(struct amdgpu_device *adev) +{ + int r, i; + struct amdgpu_ring *ring; + + if (!(adev->flags & AMD_IS_APU)) + gfx_v11_0_enable_gui_idle_interrupt(adev, false); + + if (adev->firmware.load_type == AMDGPU_FW_LOAD_DIRECT) { + /* legacy firmware loading */ + r = gfx_v11_0_cp_gfx_load_microcode(adev); + if (r) + return r; + + if (adev->gfx.rs64_enable) + r = gfx_v11_0_cp_compute_load_microcode_rs64(adev); + else + r = gfx_v11_0_cp_compute_load_microcode(adev); + if (r) + return r; + } + + gfx_v11_0_cp_set_doorbell_range(adev); + + if (amdgpu_async_gfx_ring) { + gfx_v11_0_cp_compute_enable(adev, true); + gfx_v11_0_cp_gfx_enable(adev, true); + } + + if (adev->enable_mes_kiq && adev->mes.kiq_hw_init) + r = amdgpu_mes_kiq_hw_init(adev); + else + r = gfx_v11_0_kiq_resume(adev); + if (r) + return r; + + r = gfx_v11_0_kcq_resume(adev); + if (r) + return r; + + if (!amdgpu_async_gfx_ring) { + r = gfx_v11_0_cp_gfx_resume(adev); + if (r) + return r; + } else { + r = gfx_v11_0_cp_async_gfx_ring_resume(adev); + if (r) + return r; + } + + for (i = 0; i < adev->gfx.num_gfx_rings; i++) { + ring = &adev->gfx.gfx_ring[i]; + r = amdgpu_ring_test_helper(ring); + if (r) + return r; + } + + for (i = 0; i < adev->gfx.num_compute_rings; i++) { + ring = &adev->gfx.compute_ring[i]; + r = amdgpu_ring_test_helper(ring); + if (r) + return r; + } + + return 0; +} + +static void gfx_v11_0_cp_enable(struct amdgpu_device *adev, bool enable) +{ + gfx_v11_0_cp_gfx_enable(adev, enable); + gfx_v11_0_cp_compute_enable(adev, enable); +} + +static int gfx_v11_0_gfxhub_enable(struct amdgpu_device *adev) +{ + int r; + bool value; + + r = adev->gfxhub.funcs->gart_enable(adev); + if (r) + return r; + + adev->hdp.funcs->flush_hdp(adev, NULL); + + value = (amdgpu_vm_fault_stop == AMDGPU_VM_FAULT_STOP_ALWAYS) ? + false : true; + + adev->gfxhub.funcs->set_fault_enable_default(adev, value); + amdgpu_gmc_flush_gpu_tlb(adev, 0, AMDGPU_GFXHUB_0, 0); + + return 0; +} + +static void gfx_v11_0_select_cp_fw_arch(struct amdgpu_device *adev) +{ + u32 tmp; + + /* select RS64 */ + if (adev->gfx.rs64_enable) { + tmp = RREG32_SOC15(GC, 0, regCP_GFX_CNTL); + tmp = REG_SET_FIELD(tmp, CP_GFX_CNTL, ENGINE_SEL, 1); + WREG32_SOC15(GC, 0, regCP_GFX_CNTL, tmp); + + tmp = RREG32_SOC15(GC, 0, regCP_MEC_ISA_CNTL); + tmp = REG_SET_FIELD(tmp, CP_MEC_ISA_CNTL, ISA_MODE, 1); + WREG32_SOC15(GC, 0, regCP_MEC_ISA_CNTL, tmp); + } + + if (amdgpu_emu_mode == 1) + msleep(100); +} + +static int get_gb_addr_config(struct amdgpu_device * adev) +{ + u32 gb_addr_config; + + gb_addr_config = RREG32_SOC15(GC, 0, regGB_ADDR_CONFIG); + if (gb_addr_config == 0) + return -EINVAL; + + adev->gfx.config.gb_addr_config_fields.num_pkrs = + 1 << REG_GET_FIELD(gb_addr_config, GB_ADDR_CONFIG, NUM_PKRS); + + adev->gfx.config.gb_addr_config = gb_addr_config; + + adev->gfx.config.gb_addr_config_fields.num_pipes = 1 << + REG_GET_FIELD(adev->gfx.config.gb_addr_config, + GB_ADDR_CONFIG, NUM_PIPES); + + adev->gfx.config.max_tile_pipes = + adev->gfx.config.gb_addr_config_fields.num_pipes; + + adev->gfx.config.gb_addr_config_fields.max_compress_frags = 1 << + REG_GET_FIELD(adev->gfx.config.gb_addr_config, + GB_ADDR_CONFIG, MAX_COMPRESSED_FRAGS); + adev->gfx.config.gb_addr_config_fields.num_rb_per_se = 1 << + REG_GET_FIELD(adev->gfx.config.gb_addr_config, + GB_ADDR_CONFIG, NUM_RB_PER_SE); + adev->gfx.config.gb_addr_config_fields.num_se = 1 << + REG_GET_FIELD(adev->gfx.config.gb_addr_config, + GB_ADDR_CONFIG, NUM_SHADER_ENGINES); + adev->gfx.config.gb_addr_config_fields.pipe_interleave_size = 1 << (8 + + REG_GET_FIELD(adev->gfx.config.gb_addr_config, + GB_ADDR_CONFIG, PIPE_INTERLEAVE_SIZE)); + + return 0; +} + +static void gfx_v11_0_disable_gpa_mode(struct amdgpu_device *adev) +{ + uint32_t data; + + data = RREG32_SOC15(GC, 0, regCPC_PSP_DEBUG); + data |= CPC_PSP_DEBUG__GPA_OVERRIDE_MASK; + WREG32_SOC15(GC, 0, regCPC_PSP_DEBUG, data); + + data = RREG32_SOC15(GC, 0, regCPG_PSP_DEBUG); + data |= CPG_PSP_DEBUG__GPA_OVERRIDE_MASK; + WREG32_SOC15(GC, 0, regCPG_PSP_DEBUG, data); +} + +static int gfx_v11_0_hw_init(void *handle) +{ + int r; + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + + if (adev->firmware.load_type == AMDGPU_FW_LOAD_RLC_BACKDOOR_AUTO) { + if (adev->gfx.imu.funcs) { + /* RLC autoload sequence 1: Program rlc ram */ + if (adev->gfx.imu.funcs->program_rlc_ram) + adev->gfx.imu.funcs->program_rlc_ram(adev); + } + /* rlc autoload firmware */ + r = gfx_v11_0_rlc_backdoor_autoload_enable(adev); + if (r) + return r; + } else { + if (adev->firmware.load_type == AMDGPU_FW_LOAD_DIRECT) { + if (adev->gfx.imu.funcs && (amdgpu_dpm > 0)) { + if (adev->gfx.imu.funcs->load_microcode) + adev->gfx.imu.funcs->load_microcode(adev); + if (adev->gfx.imu.funcs->setup_imu) + adev->gfx.imu.funcs->setup_imu(adev); + if (adev->gfx.imu.funcs->start_imu) + adev->gfx.imu.funcs->start_imu(adev); + } + + /* disable gpa mode in backdoor loading */ + gfx_v11_0_disable_gpa_mode(adev); + } + } + + if ((adev->firmware.load_type == AMDGPU_FW_LOAD_RLC_BACKDOOR_AUTO) || + (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP)) { + r = gfx_v11_0_wait_for_rlc_autoload_complete(adev); + if (r) { + dev_err(adev->dev, "(%d) failed to wait rlc autoload complete\n", r); + return r; + } + } + + adev->gfx.is_poweron = true; + + if(get_gb_addr_config(adev)) + DRM_WARN("Invalid gb_addr_config !\n"); + + if (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP && + adev->gfx.rs64_enable) + gfx_v11_0_config_gfx_rs64(adev); + + r = gfx_v11_0_gfxhub_enable(adev); + if (r) + return r; + + if (!amdgpu_emu_mode) + gfx_v11_0_init_golden_registers(adev); + + if ((adev->firmware.load_type == AMDGPU_FW_LOAD_DIRECT) || + (adev->firmware.load_type == AMDGPU_FW_LOAD_RLC_BACKDOOR_AUTO && amdgpu_dpm == 1)) { + /** + * For gfx 11, rlc firmware loading relies on smu firmware is + * loaded firstly, so in direct type, it has to load smc ucode + * here before rlc. + */ + if (!(adev->flags & AMD_IS_APU)) { + r = amdgpu_pm_load_smu_firmware(adev, NULL); + if (r) + return r; + } + } + + gfx_v11_0_constants_init(adev); + + if (adev->firmware.load_type != AMDGPU_FW_LOAD_PSP) + gfx_v11_0_select_cp_fw_arch(adev); + + if (adev->nbio.funcs->gc_doorbell_init) + adev->nbio.funcs->gc_doorbell_init(adev); + + r = gfx_v11_0_rlc_resume(adev); + if (r) + return r; + + /* + * init golden registers and rlc resume may override some registers, + * reconfig them here + */ + gfx_v11_0_tcp_harvest(adev); + + r = gfx_v11_0_cp_resume(adev); + if (r) + return r; + + return r; +} + +#ifndef BRING_UP_DEBUG +static int gfx_v11_0_kiq_disable_kgq(struct amdgpu_device *adev) +{ + struct amdgpu_kiq *kiq = &adev->gfx.kiq; + struct amdgpu_ring *kiq_ring = &kiq->ring; + int i, r = 0; + + if (!kiq->pmf || !kiq->pmf->kiq_unmap_queues) + return -EINVAL; + + if (amdgpu_ring_alloc(kiq_ring, kiq->pmf->unmap_queues_size * + adev->gfx.num_gfx_rings)) + return -ENOMEM; + + for (i = 0; i < adev->gfx.num_gfx_rings; i++) + kiq->pmf->kiq_unmap_queues(kiq_ring, &adev->gfx.gfx_ring[i], + PREEMPT_QUEUES, 0, 0); + + if (adev->gfx.kiq.ring.sched.ready) + r = amdgpu_ring_test_helper(kiq_ring); + + return r; +} +#endif + +static int gfx_v11_0_hw_fini(void *handle) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + int r; + uint32_t tmp; + + amdgpu_irq_put(adev, &adev->gfx.priv_reg_irq, 0); + amdgpu_irq_put(adev, &adev->gfx.priv_inst_irq, 0); + + if (!adev->no_hw_access) { +#ifndef BRING_UP_DEBUG + if (amdgpu_async_gfx_ring) { + r = gfx_v11_0_kiq_disable_kgq(adev); + if (r) + DRM_ERROR("KGQ disable failed\n"); + } +#endif + if (amdgpu_gfx_disable_kcq(adev)) + DRM_ERROR("KCQ disable failed\n"); + + amdgpu_mes_kiq_hw_fini(adev); + } + + if (amdgpu_sriov_vf(adev)) { + gfx_v11_0_cp_gfx_enable(adev, false); + /* Program KIQ position of RLC_CP_SCHEDULERS during destroy */ + tmp = RREG32_SOC15(GC, 0, regRLC_CP_SCHEDULERS); + tmp &= 0xffffff00; + WREG32_SOC15(GC, 0, regRLC_CP_SCHEDULERS, tmp); + + return 0; + } + gfx_v11_0_cp_enable(adev, false); + gfx_v11_0_enable_gui_idle_interrupt(adev, false); + + adev->gfxhub.funcs->gart_disable(adev); + + adev->gfx.is_poweron = false; + + return 0; +} + +static int gfx_v11_0_suspend(void *handle) +{ + return gfx_v11_0_hw_fini(handle); +} + +static int gfx_v11_0_resume(void *handle) +{ + return gfx_v11_0_hw_init(handle); +} + +static bool gfx_v11_0_is_idle(void *handle) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + + if (REG_GET_FIELD(RREG32_SOC15(GC, 0, regGRBM_STATUS), + GRBM_STATUS, GUI_ACTIVE)) + return false; + else + return true; +} + +static int gfx_v11_0_wait_for_idle(void *handle) +{ + unsigned i; + u32 tmp; + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + + for (i = 0; i < adev->usec_timeout; i++) { + /* read MC_STATUS */ + tmp = RREG32_SOC15(GC, 0, regGRBM_STATUS) & + GRBM_STATUS__GUI_ACTIVE_MASK; + + if (!REG_GET_FIELD(tmp, GRBM_STATUS, GUI_ACTIVE)) + return 0; + udelay(1); + } + return -ETIMEDOUT; +} + +static int gfx_v11_0_soft_reset(void *handle) +{ + u32 grbm_soft_reset = 0; + u32 tmp; + int i, j, k; + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + + tmp = RREG32_SOC15(GC, 0, regCP_INT_CNTL); + tmp = REG_SET_FIELD(tmp, CP_INT_CNTL, CMP_BUSY_INT_ENABLE, 0); + tmp = REG_SET_FIELD(tmp, CP_INT_CNTL, CNTX_BUSY_INT_ENABLE, 0); + tmp = REG_SET_FIELD(tmp, CP_INT_CNTL, CNTX_EMPTY_INT_ENABLE, 0); + tmp = REG_SET_FIELD(tmp, CP_INT_CNTL, GFX_IDLE_INT_ENABLE, 0); + WREG32_SOC15(GC, 0, regCP_INT_CNTL, tmp); + + gfx_v11_0_set_safe_mode(adev); + + for (i = 0; i < adev->gfx.mec.num_mec; ++i) { + for (j = 0; j < adev->gfx.mec.num_queue_per_pipe; j++) { + for (k = 0; k < adev->gfx.mec.num_pipe_per_mec; k++) { + tmp = RREG32_SOC15(GC, 0, regGRBM_GFX_CNTL); + tmp = REG_SET_FIELD(tmp, GRBM_GFX_CNTL, MEID, i); + tmp = REG_SET_FIELD(tmp, GRBM_GFX_CNTL, QUEUEID, j); + tmp = REG_SET_FIELD(tmp, GRBM_GFX_CNTL, PIPEID, k); + WREG32_SOC15(GC, 0, regGRBM_GFX_CNTL, tmp); + + WREG32_SOC15(GC, 0, regCP_HQD_DEQUEUE_REQUEST, 0x2); + WREG32_SOC15(GC, 0, regSPI_COMPUTE_QUEUE_RESET, 0x1); + } + } + } + for (i = 0; i < adev->gfx.me.num_me; ++i) { + for (j = 0; j < adev->gfx.me.num_queue_per_pipe; j++) { + for (k = 0; k < adev->gfx.me.num_pipe_per_me; k++) { + tmp = RREG32_SOC15(GC, 0, regGRBM_GFX_CNTL); + tmp = REG_SET_FIELD(tmp, GRBM_GFX_CNTL, MEID, i); + tmp = REG_SET_FIELD(tmp, GRBM_GFX_CNTL, QUEUEID, j); + tmp = REG_SET_FIELD(tmp, GRBM_GFX_CNTL, PIPEID, k); + WREG32_SOC15(GC, 0, regGRBM_GFX_CNTL, tmp); + + WREG32_SOC15(GC, 0, regCP_GFX_HQD_DEQUEUE_REQUEST, 0x1); + } + } + } + + WREG32_SOC15(GC, 0, regCP_VMID_RESET, 0xfffffffe); + + // Read CP_VMID_RESET register three times. + // to get sufficient time for GFX_HQD_ACTIVE reach 0 + RREG32_SOC15(GC, 0, regCP_VMID_RESET); + RREG32_SOC15(GC, 0, regCP_VMID_RESET); + RREG32_SOC15(GC, 0, regCP_VMID_RESET); + + for (i = 0; i < adev->usec_timeout; i++) { + if (!RREG32_SOC15(GC, 0, regCP_HQD_ACTIVE) && + !RREG32_SOC15(GC, 0, regCP_GFX_HQD_ACTIVE)) + break; + udelay(1); + } + if (i >= adev->usec_timeout) { + printk("Failed to wait all pipes clean\n"); + return -EINVAL; + } + + /********** trigger soft reset ***********/ + grbm_soft_reset = RREG32_SOC15(GC, 0, regGRBM_SOFT_RESET); + grbm_soft_reset = REG_SET_FIELD(grbm_soft_reset, GRBM_SOFT_RESET, + SOFT_RESET_CP, 1); + grbm_soft_reset = REG_SET_FIELD(grbm_soft_reset, GRBM_SOFT_RESET, + SOFT_RESET_GFX, 1); + grbm_soft_reset = REG_SET_FIELD(grbm_soft_reset, GRBM_SOFT_RESET, + SOFT_RESET_CPF, 1); + grbm_soft_reset = REG_SET_FIELD(grbm_soft_reset, GRBM_SOFT_RESET, + SOFT_RESET_CPC, 1); + grbm_soft_reset = REG_SET_FIELD(grbm_soft_reset, GRBM_SOFT_RESET, + SOFT_RESET_CPG, 1); + WREG32_SOC15(GC, 0, regGRBM_SOFT_RESET, grbm_soft_reset); + /********** exit soft reset ***********/ + grbm_soft_reset = RREG32_SOC15(GC, 0, regGRBM_SOFT_RESET); + grbm_soft_reset = REG_SET_FIELD(grbm_soft_reset, GRBM_SOFT_RESET, + SOFT_RESET_CP, 0); + grbm_soft_reset = REG_SET_FIELD(grbm_soft_reset, GRBM_SOFT_RESET, + SOFT_RESET_GFX, 0); + grbm_soft_reset = REG_SET_FIELD(grbm_soft_reset, GRBM_SOFT_RESET, + SOFT_RESET_CPF, 0); + grbm_soft_reset = REG_SET_FIELD(grbm_soft_reset, GRBM_SOFT_RESET, + SOFT_RESET_CPC, 0); + grbm_soft_reset = REG_SET_FIELD(grbm_soft_reset, GRBM_SOFT_RESET, + SOFT_RESET_CPG, 0); + WREG32_SOC15(GC, 0, regGRBM_SOFT_RESET, grbm_soft_reset); + + tmp = RREG32_SOC15(GC, 0, regCP_SOFT_RESET_CNTL); + tmp = REG_SET_FIELD(tmp, CP_SOFT_RESET_CNTL, CMP_HQD_REG_RESET, 0x1); + WREG32_SOC15(GC, 0, regCP_SOFT_RESET_CNTL, tmp); + + WREG32_SOC15(GC, 0, regCP_ME_CNTL, 0x0); + WREG32_SOC15(GC, 0, regCP_MEC_RS64_CNTL, 0x0); + + for (i = 0; i < adev->usec_timeout; i++) { + if (!RREG32_SOC15(GC, 0, regCP_VMID_RESET)) + break; + udelay(1); + } + if (i >= adev->usec_timeout) { + printk("Failed to wait CP_VMID_RESET to 0\n"); + return -EINVAL; + } + + tmp = RREG32_SOC15(GC, 0, regCP_INT_CNTL); + tmp = REG_SET_FIELD(tmp, CP_INT_CNTL, CMP_BUSY_INT_ENABLE, 1); + tmp = REG_SET_FIELD(tmp, CP_INT_CNTL, CNTX_BUSY_INT_ENABLE, 1); + tmp = REG_SET_FIELD(tmp, CP_INT_CNTL, CNTX_EMPTY_INT_ENABLE, 1); + tmp = REG_SET_FIELD(tmp, CP_INT_CNTL, GFX_IDLE_INT_ENABLE, 1); + WREG32_SOC15(GC, 0, regCP_INT_CNTL, tmp); + + gfx_v11_0_unset_safe_mode(adev); + + return gfx_v11_0_cp_resume(adev); +} + +static bool gfx_v11_0_check_soft_reset(void *handle) +{ + int i, r; + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + struct amdgpu_ring *ring; + long tmo = msecs_to_jiffies(1000); + + for (i = 0; i < adev->gfx.num_gfx_rings; i++) { + ring = &adev->gfx.gfx_ring[i]; + r = amdgpu_ring_test_ib(ring, tmo); + if (r) + return true; + } + + for (i = 0; i < adev->gfx.num_compute_rings; i++) { + ring = &adev->gfx.compute_ring[i]; + r = amdgpu_ring_test_ib(ring, tmo); + if (r) + return true; + } + + return false; +} + +static uint64_t gfx_v11_0_get_gpu_clock_counter(struct amdgpu_device *adev) +{ + uint64_t clock; + + amdgpu_gfx_off_ctrl(adev, false); + mutex_lock(&adev->gfx.gpu_clock_mutex); + clock = (uint64_t)RREG32_SOC15(SMUIO, 0, regGOLDEN_TSC_COUNT_LOWER) | + ((uint64_t)RREG32_SOC15(SMUIO, 0, regGOLDEN_TSC_COUNT_UPPER) << 32ULL); + mutex_unlock(&adev->gfx.gpu_clock_mutex); + amdgpu_gfx_off_ctrl(adev, true); + return clock; +} + +static void gfx_v11_0_ring_emit_gds_switch(struct amdgpu_ring *ring, + uint32_t vmid, + uint32_t gds_base, uint32_t gds_size, + uint32_t gws_base, uint32_t gws_size, + uint32_t oa_base, uint32_t oa_size) +{ + struct amdgpu_device *adev = ring->adev; + + /* GDS Base */ + gfx_v11_0_write_data_to_reg(ring, 0, false, + SOC15_REG_OFFSET(GC, 0, regGDS_VMID0_BASE) + 2 * vmid, + gds_base); + + /* GDS Size */ + gfx_v11_0_write_data_to_reg(ring, 0, false, + SOC15_REG_OFFSET(GC, 0, regGDS_VMID0_SIZE) + 2 * vmid, + gds_size); + + /* GWS */ + gfx_v11_0_write_data_to_reg(ring, 0, false, + SOC15_REG_OFFSET(GC, 0, regGDS_GWS_VMID0) + vmid, + gws_size << GDS_GWS_VMID0__SIZE__SHIFT | gws_base); + + /* OA */ + gfx_v11_0_write_data_to_reg(ring, 0, false, + SOC15_REG_OFFSET(GC, 0, regGDS_OA_VMID0) + vmid, + (1 << (oa_size + oa_base)) - (1 << oa_base)); +} + +static int gfx_v11_0_early_init(void *handle) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + + adev->gfx.num_gfx_rings = GFX11_NUM_GFX_RINGS; + adev->gfx.num_compute_rings = min(amdgpu_gfx_get_num_kcq(adev), + AMDGPU_MAX_COMPUTE_RINGS); + + gfx_v11_0_set_kiq_pm4_funcs(adev); + gfx_v11_0_set_ring_funcs(adev); + gfx_v11_0_set_irq_funcs(adev); + gfx_v11_0_set_gds_init(adev); + gfx_v11_0_set_rlc_funcs(adev); + gfx_v11_0_set_mqd_funcs(adev); + gfx_v11_0_set_imu_funcs(adev); + + gfx_v11_0_init_rlcg_reg_access_ctrl(adev); + + return 0; +} + +static int gfx_v11_0_late_init(void *handle) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + int r; + + r = amdgpu_irq_get(adev, &adev->gfx.priv_reg_irq, 0); + if (r) + return r; + + r = amdgpu_irq_get(adev, &adev->gfx.priv_inst_irq, 0); + if (r) + return r; + + return 0; +} + +static bool gfx_v11_0_is_rlc_enabled(struct amdgpu_device *adev) +{ + uint32_t rlc_cntl; + + /* if RLC is not enabled, do nothing */ + rlc_cntl = RREG32_SOC15(GC, 0, regRLC_CNTL); + return (REG_GET_FIELD(rlc_cntl, RLC_CNTL, RLC_ENABLE_F32)) ? true : false; +} + +static void gfx_v11_0_set_safe_mode(struct amdgpu_device *adev) +{ + uint32_t data; + unsigned i; + + data = RLC_SAFE_MODE__CMD_MASK; + data |= (1 << RLC_SAFE_MODE__MESSAGE__SHIFT); + + WREG32_SOC15(GC, 0, regRLC_SAFE_MODE, data); + + /* wait for RLC_SAFE_MODE */ + for (i = 0; i < adev->usec_timeout; i++) { + if (!REG_GET_FIELD(RREG32_SOC15(GC, 0, regRLC_SAFE_MODE), + RLC_SAFE_MODE, CMD)) + break; + udelay(1); + } +} + +static void gfx_v11_0_unset_safe_mode(struct amdgpu_device *adev) +{ + WREG32_SOC15(GC, 0, regRLC_SAFE_MODE, RLC_SAFE_MODE__CMD_MASK); +} + +static void gfx_v11_0_update_perf_clk(struct amdgpu_device *adev, + bool enable) +{ + uint32_t def, data; + + if (!(adev->cg_flags & AMD_CG_SUPPORT_GFX_PERF_CLK)) + return; + + def = data = RREG32_SOC15(GC, 0, regRLC_CGTT_MGCG_OVERRIDE); + + if (enable) + data &= ~RLC_CGTT_MGCG_OVERRIDE__PERFMON_CLOCK_STATE_MASK; + else + data |= RLC_CGTT_MGCG_OVERRIDE__PERFMON_CLOCK_STATE_MASK; + + if (def != data) + WREG32_SOC15(GC, 0, regRLC_CGTT_MGCG_OVERRIDE, data); +} + +static void gfx_v11_0_update_sram_fgcg(struct amdgpu_device *adev, + bool enable) +{ + uint32_t def, data; + + if (!(adev->cg_flags & AMD_CG_SUPPORT_GFX_FGCG)) + return; + + def = data = RREG32_SOC15(GC, 0, regRLC_CGTT_MGCG_OVERRIDE); + + if (enable) + data &= ~RLC_CGTT_MGCG_OVERRIDE__GFXIP_FGCG_OVERRIDE_MASK; + else + data |= RLC_CGTT_MGCG_OVERRIDE__GFXIP_FGCG_OVERRIDE_MASK; + + if (def != data) + WREG32_SOC15(GC, 0, regRLC_CGTT_MGCG_OVERRIDE, data); +} + +static void gfx_v11_0_update_repeater_fgcg(struct amdgpu_device *adev, + bool enable) +{ + uint32_t def, data; + + if (!(adev->cg_flags & AMD_CG_SUPPORT_REPEATER_FGCG)) + return; + + def = data = RREG32_SOC15(GC, 0, regRLC_CGTT_MGCG_OVERRIDE); + + if (enable) + data &= ~RLC_CGTT_MGCG_OVERRIDE__GFXIP_REPEATER_FGCG_OVERRIDE_MASK; + else + data |= RLC_CGTT_MGCG_OVERRIDE__GFXIP_REPEATER_FGCG_OVERRIDE_MASK; + + if (def != data) + WREG32_SOC15(GC, 0, regRLC_CGTT_MGCG_OVERRIDE, data); +} + +static void gfx_v11_0_update_medium_grain_clock_gating(struct amdgpu_device *adev, + bool enable) +{ + uint32_t data, def; + + if (!(adev->cg_flags & (AMD_CG_SUPPORT_GFX_MGCG | AMD_CG_SUPPORT_GFX_MGLS))) + return; + + /* It is disabled by HW by default */ + if (enable) { + if (adev->cg_flags & AMD_CG_SUPPORT_GFX_MGCG) { + /* 1 - RLC_CGTT_MGCG_OVERRIDE */ + def = data = RREG32_SOC15(GC, 0, regRLC_CGTT_MGCG_OVERRIDE); + + data &= ~(RLC_CGTT_MGCG_OVERRIDE__GRBM_CGTT_SCLK_OVERRIDE_MASK | + RLC_CGTT_MGCG_OVERRIDE__RLC_CGTT_SCLK_OVERRIDE_MASK | + RLC_CGTT_MGCG_OVERRIDE__GFXIP_MGCG_OVERRIDE_MASK); + + if (def != data) + WREG32_SOC15(GC, 0, regRLC_CGTT_MGCG_OVERRIDE, data); + } + } else { + if (adev->cg_flags & AMD_CG_SUPPORT_GFX_MGCG) { + def = data = RREG32_SOC15(GC, 0, regRLC_CGTT_MGCG_OVERRIDE); + + data |= (RLC_CGTT_MGCG_OVERRIDE__RLC_CGTT_SCLK_OVERRIDE_MASK | + RLC_CGTT_MGCG_OVERRIDE__GRBM_CGTT_SCLK_OVERRIDE_MASK | + RLC_CGTT_MGCG_OVERRIDE__GFXIP_MGCG_OVERRIDE_MASK); + + if (def != data) + WREG32_SOC15(GC, 0, regRLC_CGTT_MGCG_OVERRIDE, data); + } + } +} + +static void gfx_v11_0_update_coarse_grain_clock_gating(struct amdgpu_device *adev, + bool enable) +{ + uint32_t def, data; + + if (!(adev->cg_flags & + (AMD_CG_SUPPORT_GFX_CGCG | + AMD_CG_SUPPORT_GFX_CGLS | + AMD_CG_SUPPORT_GFX_3D_CGCG | + AMD_CG_SUPPORT_GFX_3D_CGLS))) + return; + + if (enable) { + def = data = RREG32_SOC15(GC, 0, regRLC_CGTT_MGCG_OVERRIDE); + + /* unset CGCG override */ + if (adev->cg_flags & AMD_CG_SUPPORT_GFX_CGCG) + data &= ~RLC_CGTT_MGCG_OVERRIDE__GFXIP_CGCG_OVERRIDE_MASK; + if (adev->cg_flags & AMD_CG_SUPPORT_GFX_CGLS) + data &= ~RLC_CGTT_MGCG_OVERRIDE__GFXIP_CGLS_OVERRIDE_MASK; + if (adev->cg_flags & AMD_CG_SUPPORT_GFX_3D_CGCG || + adev->cg_flags & AMD_CG_SUPPORT_GFX_3D_CGLS) + data &= ~RLC_CGTT_MGCG_OVERRIDE__GFXIP_GFX3D_CG_OVERRIDE_MASK; + + /* update CGCG override bits */ + if (def != data) + WREG32_SOC15(GC, 0, regRLC_CGTT_MGCG_OVERRIDE, data); + + /* enable cgcg FSM(0x0000363F) */ + def = data = RREG32_SOC15(GC, 0, regRLC_CGCG_CGLS_CTRL); + + if (adev->cg_flags & AMD_CG_SUPPORT_GFX_CGCG) { + data &= ~RLC_CGCG_CGLS_CTRL__CGCG_GFX_IDLE_THRESHOLD_MASK; + data |= (0x36 << RLC_CGCG_CGLS_CTRL__CGCG_GFX_IDLE_THRESHOLD__SHIFT) | + RLC_CGCG_CGLS_CTRL__CGCG_EN_MASK; + } + + if (adev->cg_flags & AMD_CG_SUPPORT_GFX_CGLS) { + data &= ~RLC_CGCG_CGLS_CTRL__CGLS_REP_COMPANSAT_DELAY_MASK; + data |= (0x000F << RLC_CGCG_CGLS_CTRL__CGLS_REP_COMPANSAT_DELAY__SHIFT) | + RLC_CGCG_CGLS_CTRL__CGLS_EN_MASK; + } + + if (def != data) + WREG32_SOC15(GC, 0, regRLC_CGCG_CGLS_CTRL, data); + + /* Program RLC_CGCG_CGLS_CTRL_3D */ + def = data = RREG32_SOC15(GC, 0, regRLC_CGCG_CGLS_CTRL_3D); + + if (adev->cg_flags & AMD_CG_SUPPORT_GFX_3D_CGCG) { + data &= ~RLC_CGCG_CGLS_CTRL_3D__CGCG_GFX_IDLE_THRESHOLD_MASK; + data |= (0x36 << RLC_CGCG_CGLS_CTRL_3D__CGCG_GFX_IDLE_THRESHOLD__SHIFT) | + RLC_CGCG_CGLS_CTRL_3D__CGCG_EN_MASK; + } + + if (adev->cg_flags & AMD_CG_SUPPORT_GFX_3D_CGLS) { + data &= ~RLC_CGCG_CGLS_CTRL_3D__CGLS_REP_COMPANSAT_DELAY_MASK; + data |= (0xf << RLC_CGCG_CGLS_CTRL_3D__CGLS_REP_COMPANSAT_DELAY__SHIFT) | + RLC_CGCG_CGLS_CTRL_3D__CGLS_EN_MASK; + } + + if (def != data) + WREG32_SOC15(GC, 0, regRLC_CGCG_CGLS_CTRL_3D, data); + + /* set IDLE_POLL_COUNT(0x00900100) */ + def = data = RREG32_SOC15(GC, 0, regCP_RB_WPTR_POLL_CNTL); + + data &= ~(CP_RB_WPTR_POLL_CNTL__POLL_FREQUENCY_MASK | CP_RB_WPTR_POLL_CNTL__IDLE_POLL_COUNT_MASK); + data |= (0x0100 << CP_RB_WPTR_POLL_CNTL__POLL_FREQUENCY__SHIFT) | + (0x0090 << CP_RB_WPTR_POLL_CNTL__IDLE_POLL_COUNT__SHIFT); + + if (def != data) + WREG32_SOC15(GC, 0, regCP_RB_WPTR_POLL_CNTL, data); + + data = RREG32_SOC15(GC, 0, regCP_INT_CNTL); + data = REG_SET_FIELD(data, CP_INT_CNTL, CNTX_BUSY_INT_ENABLE, 1); + data = REG_SET_FIELD(data, CP_INT_CNTL, CNTX_EMPTY_INT_ENABLE, 1); + data = REG_SET_FIELD(data, CP_INT_CNTL, CMP_BUSY_INT_ENABLE, 1); + data = REG_SET_FIELD(data, CP_INT_CNTL, GFX_IDLE_INT_ENABLE, 1); + WREG32_SOC15(GC, 0, regCP_INT_CNTL, data); + + data = RREG32_SOC15(GC, 0, regSDMA0_RLC_CGCG_CTRL); + data = REG_SET_FIELD(data, SDMA0_RLC_CGCG_CTRL, CGCG_INT_ENABLE, 1); + WREG32_SOC15(GC, 0, regSDMA0_RLC_CGCG_CTRL, data); + + /* Some ASICs only have one SDMA instance, not need to configure SDMA1 */ + if (adev->sdma.num_instances > 1) { + data = RREG32_SOC15(GC, 0, regSDMA1_RLC_CGCG_CTRL); + data = REG_SET_FIELD(data, SDMA1_RLC_CGCG_CTRL, CGCG_INT_ENABLE, 1); + WREG32_SOC15(GC, 0, regSDMA1_RLC_CGCG_CTRL, data); + } + } else { + /* Program RLC_CGCG_CGLS_CTRL */ + def = data = RREG32_SOC15(GC, 0, regRLC_CGCG_CGLS_CTRL); + + if (adev->cg_flags & AMD_CG_SUPPORT_GFX_CGCG) + data &= ~RLC_CGCG_CGLS_CTRL__CGCG_EN_MASK; + + if (adev->cg_flags & AMD_CG_SUPPORT_GFX_CGLS) + data &= ~RLC_CGCG_CGLS_CTRL__CGLS_EN_MASK; + + if (def != data) + WREG32_SOC15(GC, 0, regRLC_CGCG_CGLS_CTRL, data); + + /* Program RLC_CGCG_CGLS_CTRL_3D */ + def = data = RREG32_SOC15(GC, 0, regRLC_CGCG_CGLS_CTRL_3D); + + if (adev->cg_flags & AMD_CG_SUPPORT_GFX_3D_CGCG) + data &= ~RLC_CGCG_CGLS_CTRL_3D__CGCG_EN_MASK; + if (adev->cg_flags & AMD_CG_SUPPORT_GFX_3D_CGLS) + data &= ~RLC_CGCG_CGLS_CTRL_3D__CGLS_EN_MASK; + + if (def != data) + WREG32_SOC15(GC, 0, regRLC_CGCG_CGLS_CTRL_3D, data); + + data = RREG32_SOC15(GC, 0, regSDMA0_RLC_CGCG_CTRL); + data &= ~SDMA0_RLC_CGCG_CTRL__CGCG_INT_ENABLE_MASK; + WREG32_SOC15(GC, 0, regSDMA0_RLC_CGCG_CTRL, data); + + /* Some ASICs only have one SDMA instance, not need to configure SDMA1 */ + if (adev->sdma.num_instances > 1) { + data = RREG32_SOC15(GC, 0, regSDMA1_RLC_CGCG_CTRL); + data &= ~SDMA1_RLC_CGCG_CTRL__CGCG_INT_ENABLE_MASK; + WREG32_SOC15(GC, 0, regSDMA1_RLC_CGCG_CTRL, data); + } + } +} + +static int gfx_v11_0_update_gfx_clock_gating(struct amdgpu_device *adev, + bool enable) +{ + amdgpu_gfx_rlc_enter_safe_mode(adev); + + gfx_v11_0_update_coarse_grain_clock_gating(adev, enable); + + gfx_v11_0_update_medium_grain_clock_gating(adev, enable); + + gfx_v11_0_update_repeater_fgcg(adev, enable); + + gfx_v11_0_update_sram_fgcg(adev, enable); + + gfx_v11_0_update_perf_clk(adev, enable); + + if (adev->cg_flags & + (AMD_CG_SUPPORT_GFX_MGCG | + AMD_CG_SUPPORT_GFX_CGLS | + AMD_CG_SUPPORT_GFX_CGCG | + AMD_CG_SUPPORT_GFX_3D_CGCG | + AMD_CG_SUPPORT_GFX_3D_CGLS)) + gfx_v11_0_enable_gui_idle_interrupt(adev, enable); + + amdgpu_gfx_rlc_exit_safe_mode(adev); + + return 0; +} + +static void gfx_v11_0_update_spm_vmid(struct amdgpu_device *adev, unsigned vmid) +{ + u32 reg, data; + + amdgpu_gfx_off_ctrl(adev, false); + + reg = SOC15_REG_OFFSET(GC, 0, regRLC_SPM_MC_CNTL); + if (amdgpu_sriov_is_pp_one_vf(adev)) + data = RREG32_NO_KIQ(reg); + else + data = RREG32(reg); + + data &= ~RLC_SPM_MC_CNTL__RLC_SPM_VMID_MASK; + data |= (vmid & RLC_SPM_MC_CNTL__RLC_SPM_VMID_MASK) << RLC_SPM_MC_CNTL__RLC_SPM_VMID__SHIFT; + + if (amdgpu_sriov_is_pp_one_vf(adev)) + WREG32_SOC15_NO_KIQ(GC, 0, regRLC_SPM_MC_CNTL, data); + else + WREG32_SOC15(GC, 0, regRLC_SPM_MC_CNTL, data); + + amdgpu_gfx_off_ctrl(adev, true); +} + +static const struct amdgpu_rlc_funcs gfx_v11_0_rlc_funcs = { + .is_rlc_enabled = gfx_v11_0_is_rlc_enabled, + .set_safe_mode = gfx_v11_0_set_safe_mode, + .unset_safe_mode = gfx_v11_0_unset_safe_mode, + .init = gfx_v11_0_rlc_init, + .get_csb_size = gfx_v11_0_get_csb_size, + .get_csb_buffer = gfx_v11_0_get_csb_buffer, + .resume = gfx_v11_0_rlc_resume, + .stop = gfx_v11_0_rlc_stop, + .reset = gfx_v11_0_rlc_reset, + .start = gfx_v11_0_rlc_start, + .update_spm_vmid = gfx_v11_0_update_spm_vmid, +}; + +static void gfx_v11_cntl_power_gating(struct amdgpu_device *adev, bool enable) +{ + u32 data = RREG32_SOC15(GC, 0, regRLC_PG_CNTL); + + if (enable && (adev->pg_flags & AMD_PG_SUPPORT_GFX_PG)) + data |= RLC_PG_CNTL__GFX_POWER_GATING_ENABLE_MASK; + else + data &= ~RLC_PG_CNTL__GFX_POWER_GATING_ENABLE_MASK; + + WREG32_SOC15(GC, 0, regRLC_PG_CNTL, data); + + // Program RLC_PG_DELAY3 for CGPG hysteresis + if (enable && (adev->pg_flags & AMD_PG_SUPPORT_GFX_PG)) { + switch (adev->ip_versions[GC_HWIP][0]) { + case IP_VERSION(11, 0, 1): + WREG32_SOC15(GC, 0, regRLC_PG_DELAY_3, RLC_PG_DELAY_3_DEFAULT_GC_11_0_1); + break; + default: + break; + } + } +} + +static void gfx_v11_cntl_pg(struct amdgpu_device *adev, bool enable) +{ + amdgpu_gfx_rlc_enter_safe_mode(adev); + + gfx_v11_cntl_power_gating(adev, enable); + + amdgpu_gfx_rlc_exit_safe_mode(adev); +} + +static int gfx_v11_0_set_powergating_state(void *handle, + enum amd_powergating_state state) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + bool enable = (state == AMD_PG_STATE_GATE); + + if (amdgpu_sriov_vf(adev)) + return 0; + + switch (adev->ip_versions[GC_HWIP][0]) { + case IP_VERSION(11, 0, 0): + case IP_VERSION(11, 0, 2): + case IP_VERSION(11, 0, 3): + amdgpu_gfx_off_ctrl(adev, enable); + break; + case IP_VERSION(11, 0, 1): + gfx_v11_cntl_pg(adev, enable); + amdgpu_gfx_off_ctrl(adev, enable); + break; + default: + break; + } + + return 0; +} + +static int gfx_v11_0_set_clockgating_state(void *handle, + enum amd_clockgating_state state) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + + if (amdgpu_sriov_vf(adev)) + return 0; + + switch (adev->ip_versions[GC_HWIP][0]) { + case IP_VERSION(11, 0, 0): + case IP_VERSION(11, 0, 1): + case IP_VERSION(11, 0, 2): + case IP_VERSION(11, 0, 3): + gfx_v11_0_update_gfx_clock_gating(adev, + state == AMD_CG_STATE_GATE); + break; + default: + break; + } + + return 0; +} + +static void gfx_v11_0_get_clockgating_state(void *handle, u64 *flags) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + int data; + + /* AMD_CG_SUPPORT_GFX_MGCG */ + data = RREG32_SOC15(GC, 0, regRLC_CGTT_MGCG_OVERRIDE); + if (!(data & RLC_CGTT_MGCG_OVERRIDE__GFXIP_MGCG_OVERRIDE_MASK)) + *flags |= AMD_CG_SUPPORT_GFX_MGCG; + + /* AMD_CG_SUPPORT_REPEATER_FGCG */ + if (!(data & RLC_CGTT_MGCG_OVERRIDE__GFXIP_REPEATER_FGCG_OVERRIDE_MASK)) + *flags |= AMD_CG_SUPPORT_REPEATER_FGCG; + + /* AMD_CG_SUPPORT_GFX_FGCG */ + if (!(data & RLC_CGTT_MGCG_OVERRIDE__GFXIP_FGCG_OVERRIDE_MASK)) + *flags |= AMD_CG_SUPPORT_GFX_FGCG; + + /* AMD_CG_SUPPORT_GFX_PERF_CLK */ + if (!(data & RLC_CGTT_MGCG_OVERRIDE__PERFMON_CLOCK_STATE_MASK)) + *flags |= AMD_CG_SUPPORT_GFX_PERF_CLK; + + /* AMD_CG_SUPPORT_GFX_CGCG */ + data = RREG32_SOC15(GC, 0, regRLC_CGCG_CGLS_CTRL); + if (data & RLC_CGCG_CGLS_CTRL__CGCG_EN_MASK) + *flags |= AMD_CG_SUPPORT_GFX_CGCG; + + /* AMD_CG_SUPPORT_GFX_CGLS */ + if (data & RLC_CGCG_CGLS_CTRL__CGLS_EN_MASK) + *flags |= AMD_CG_SUPPORT_GFX_CGLS; + + /* AMD_CG_SUPPORT_GFX_3D_CGCG */ + data = RREG32_SOC15(GC, 0, regRLC_CGCG_CGLS_CTRL_3D); + if (data & RLC_CGCG_CGLS_CTRL_3D__CGCG_EN_MASK) + *flags |= AMD_CG_SUPPORT_GFX_3D_CGCG; + + /* AMD_CG_SUPPORT_GFX_3D_CGLS */ + if (data & RLC_CGCG_CGLS_CTRL_3D__CGLS_EN_MASK) + *flags |= AMD_CG_SUPPORT_GFX_3D_CGLS; +} + +static u64 gfx_v11_0_ring_get_rptr_gfx(struct amdgpu_ring *ring) +{ + /* gfx11 is 32bit rptr*/ + return *(uint32_t *)ring->rptr_cpu_addr; +} + +static u64 gfx_v11_0_ring_get_wptr_gfx(struct amdgpu_ring *ring) +{ + struct amdgpu_device *adev = ring->adev; + u64 wptr; + + /* XXX check if swapping is necessary on BE */ + if (ring->use_doorbell) { + wptr = atomic64_read((atomic64_t *)ring->wptr_cpu_addr); + } else { + wptr = RREG32_SOC15(GC, 0, regCP_RB0_WPTR); + wptr += (u64)RREG32_SOC15(GC, 0, regCP_RB0_WPTR_HI) << 32; + } + + return wptr; +} + +static void gfx_v11_0_ring_set_wptr_gfx(struct amdgpu_ring *ring) +{ + struct amdgpu_device *adev = ring->adev; + uint32_t *wptr_saved; + uint32_t *is_queue_unmap; + uint64_t aggregated_db_index; + uint32_t mqd_size = adev->mqds[AMDGPU_HW_IP_GFX].mqd_size; + uint64_t wptr_tmp; + + if (ring->is_mes_queue) { + wptr_saved = (uint32_t *)(ring->mqd_ptr + mqd_size); + is_queue_unmap = (uint32_t *)(ring->mqd_ptr + mqd_size + + sizeof(uint32_t)); + aggregated_db_index = + amdgpu_mes_get_aggregated_doorbell_index(adev, + ring->hw_prio); + + wptr_tmp = ring->wptr & ring->buf_mask; + atomic64_set((atomic64_t *)ring->wptr_cpu_addr, wptr_tmp); + *wptr_saved = wptr_tmp; + /* assume doorbell always being used by mes mapped queue */ + if (*is_queue_unmap) { + WDOORBELL64(aggregated_db_index, wptr_tmp); + WDOORBELL64(ring->doorbell_index, wptr_tmp); + } else { + WDOORBELL64(ring->doorbell_index, wptr_tmp); + + if (*is_queue_unmap) + WDOORBELL64(aggregated_db_index, wptr_tmp); + } + } else { + if (ring->use_doorbell) { + /* XXX check if swapping is necessary on BE */ + atomic64_set((atomic64_t *)ring->wptr_cpu_addr, + ring->wptr); + WDOORBELL64(ring->doorbell_index, ring->wptr); + } else { + WREG32_SOC15(GC, 0, regCP_RB0_WPTR, + lower_32_bits(ring->wptr)); + WREG32_SOC15(GC, 0, regCP_RB0_WPTR_HI, + upper_32_bits(ring->wptr)); + } + } +} + +static u64 gfx_v11_0_ring_get_rptr_compute(struct amdgpu_ring *ring) +{ + /* gfx11 hardware is 32bit rptr */ + return *(uint32_t *)ring->rptr_cpu_addr; +} + +static u64 gfx_v11_0_ring_get_wptr_compute(struct amdgpu_ring *ring) +{ + u64 wptr; + + /* XXX check if swapping is necessary on BE */ + if (ring->use_doorbell) + wptr = atomic64_read((atomic64_t *)ring->wptr_cpu_addr); + else + BUG(); + return wptr; +} + +static void gfx_v11_0_ring_set_wptr_compute(struct amdgpu_ring *ring) +{ + struct amdgpu_device *adev = ring->adev; + uint32_t *wptr_saved; + uint32_t *is_queue_unmap; + uint64_t aggregated_db_index; + uint32_t mqd_size = adev->mqds[AMDGPU_HW_IP_COMPUTE].mqd_size; + uint64_t wptr_tmp; + + if (ring->is_mes_queue) { + wptr_saved = (uint32_t *)(ring->mqd_ptr + mqd_size); + is_queue_unmap = (uint32_t *)(ring->mqd_ptr + mqd_size + + sizeof(uint32_t)); + aggregated_db_index = + amdgpu_mes_get_aggregated_doorbell_index(adev, + ring->hw_prio); + + wptr_tmp = ring->wptr & ring->buf_mask; + atomic64_set((atomic64_t *)ring->wptr_cpu_addr, wptr_tmp); + *wptr_saved = wptr_tmp; + /* assume doorbell always used by mes mapped queue */ + if (*is_queue_unmap) { + WDOORBELL64(aggregated_db_index, wptr_tmp); + WDOORBELL64(ring->doorbell_index, wptr_tmp); + } else { + WDOORBELL64(ring->doorbell_index, wptr_tmp); + + if (*is_queue_unmap) + WDOORBELL64(aggregated_db_index, wptr_tmp); + } + } else { + /* XXX check if swapping is necessary on BE */ + if (ring->use_doorbell) { + atomic64_set((atomic64_t *)ring->wptr_cpu_addr, + ring->wptr); + WDOORBELL64(ring->doorbell_index, ring->wptr); + } else { + BUG(); /* only DOORBELL method supported on gfx11 now */ + } + } +} + +static void gfx_v11_0_ring_emit_hdp_flush(struct amdgpu_ring *ring) +{ + struct amdgpu_device *adev = ring->adev; + u32 ref_and_mask, reg_mem_engine; + const struct nbio_hdp_flush_reg *nbio_hf_reg = adev->nbio.hdp_flush_reg; + + if (ring->funcs->type == AMDGPU_RING_TYPE_COMPUTE) { + switch (ring->me) { + case 1: + ref_and_mask = nbio_hf_reg->ref_and_mask_cp2 << ring->pipe; + break; + case 2: + ref_and_mask = nbio_hf_reg->ref_and_mask_cp6 << ring->pipe; + break; + default: + return; + } + reg_mem_engine = 0; + } else { + ref_and_mask = nbio_hf_reg->ref_and_mask_cp0; + reg_mem_engine = 1; /* pfp */ + } + + gfx_v11_0_wait_reg_mem(ring, reg_mem_engine, 0, 1, + adev->nbio.funcs->get_hdp_flush_req_offset(adev), + adev->nbio.funcs->get_hdp_flush_done_offset(adev), + ref_and_mask, ref_and_mask, 0x20); +} + +static void gfx_v11_0_ring_emit_ib_gfx(struct amdgpu_ring *ring, + struct amdgpu_job *job, + struct amdgpu_ib *ib, + uint32_t flags) +{ + unsigned vmid = AMDGPU_JOB_GET_VMID(job); + u32 header, control = 0; + + BUG_ON(ib->flags & AMDGPU_IB_FLAG_CE); + + header = PACKET3(PACKET3_INDIRECT_BUFFER, 2); + + control |= ib->length_dw | (vmid << 24); + + if ((amdgpu_sriov_vf(ring->adev) || amdgpu_mcbp) && (ib->flags & AMDGPU_IB_FLAG_PREEMPT)) { + control |= INDIRECT_BUFFER_PRE_ENB(1); + + if (flags & AMDGPU_IB_PREEMPTED) + control |= INDIRECT_BUFFER_PRE_RESUME(1); + + if (vmid) + gfx_v11_0_ring_emit_de_meta(ring, + (!amdgpu_sriov_vf(ring->adev) && flags & AMDGPU_IB_PREEMPTED) ? true : false); + } + + if (ring->is_mes_queue) + /* inherit vmid from mqd */ + control |= 0x400000; + + amdgpu_ring_write(ring, header); + BUG_ON(ib->gpu_addr & 0x3); /* Dword align */ + amdgpu_ring_write(ring, +#ifdef __BIG_ENDIAN + (2 << 0) | +#endif + lower_32_bits(ib->gpu_addr)); + amdgpu_ring_write(ring, upper_32_bits(ib->gpu_addr)); + amdgpu_ring_write(ring, control); +} + +static void gfx_v11_0_ring_emit_ib_compute(struct amdgpu_ring *ring, + struct amdgpu_job *job, + struct amdgpu_ib *ib, + uint32_t flags) +{ + unsigned vmid = AMDGPU_JOB_GET_VMID(job); + u32 control = INDIRECT_BUFFER_VALID | ib->length_dw | (vmid << 24); + + if (ring->is_mes_queue) + /* inherit vmid from mqd */ + control |= 0x40000000; + + /* Currently, there is a high possibility to get wave ID mismatch + * between ME and GDS, leading to a hw deadlock, because ME generates + * different wave IDs than the GDS expects. This situation happens + * randomly when at least 5 compute pipes use GDS ordered append. + * The wave IDs generated by ME are also wrong after suspend/resume. + * Those are probably bugs somewhere else in the kernel driver. + * + * Writing GDS_COMPUTE_MAX_WAVE_ID resets wave ID counters in ME and + * GDS to 0 for this ring (me/pipe). + */ + if (ib->flags & AMDGPU_IB_FLAG_RESET_GDS_MAX_WAVE_ID) { + amdgpu_ring_write(ring, PACKET3(PACKET3_SET_CONFIG_REG, 1)); + amdgpu_ring_write(ring, regGDS_COMPUTE_MAX_WAVE_ID); + amdgpu_ring_write(ring, ring->adev->gds.gds_compute_max_wave_id); + } + + amdgpu_ring_write(ring, PACKET3(PACKET3_INDIRECT_BUFFER, 2)); + BUG_ON(ib->gpu_addr & 0x3); /* Dword align */ + amdgpu_ring_write(ring, +#ifdef __BIG_ENDIAN + (2 << 0) | +#endif + lower_32_bits(ib->gpu_addr)); + amdgpu_ring_write(ring, upper_32_bits(ib->gpu_addr)); + amdgpu_ring_write(ring, control); +} + +static void gfx_v11_0_ring_emit_fence(struct amdgpu_ring *ring, u64 addr, + u64 seq, unsigned flags) +{ + bool write64bit = flags & AMDGPU_FENCE_FLAG_64BIT; + bool int_sel = flags & AMDGPU_FENCE_FLAG_INT; + + /* RELEASE_MEM - flush caches, send int */ + amdgpu_ring_write(ring, PACKET3(PACKET3_RELEASE_MEM, 6)); + amdgpu_ring_write(ring, (PACKET3_RELEASE_MEM_GCR_SEQ | + PACKET3_RELEASE_MEM_GCR_GL2_WB | + PACKET3_RELEASE_MEM_GCR_GL2_INV | + PACKET3_RELEASE_MEM_GCR_GL2_US | + PACKET3_RELEASE_MEM_GCR_GL1_INV | + PACKET3_RELEASE_MEM_GCR_GLV_INV | + PACKET3_RELEASE_MEM_GCR_GLM_INV | + PACKET3_RELEASE_MEM_GCR_GLM_WB | + PACKET3_RELEASE_MEM_CACHE_POLICY(3) | + PACKET3_RELEASE_MEM_EVENT_TYPE(CACHE_FLUSH_AND_INV_TS_EVENT) | + PACKET3_RELEASE_MEM_EVENT_INDEX(5))); + amdgpu_ring_write(ring, (PACKET3_RELEASE_MEM_DATA_SEL(write64bit ? 2 : 1) | + PACKET3_RELEASE_MEM_INT_SEL(int_sel ? 2 : 0))); + + /* + * the address should be Qword aligned if 64bit write, Dword + * aligned if only send 32bit data low (discard data high) + */ + if (write64bit) + BUG_ON(addr & 0x7); + else + BUG_ON(addr & 0x3); + amdgpu_ring_write(ring, lower_32_bits(addr)); + amdgpu_ring_write(ring, upper_32_bits(addr)); + amdgpu_ring_write(ring, lower_32_bits(seq)); + amdgpu_ring_write(ring, upper_32_bits(seq)); + amdgpu_ring_write(ring, ring->is_mes_queue ? + (ring->hw_queue_id | AMDGPU_FENCE_MES_QUEUE_FLAG) : 0); +} + +static void gfx_v11_0_ring_emit_pipeline_sync(struct amdgpu_ring *ring) +{ + int usepfp = (ring->funcs->type == AMDGPU_RING_TYPE_GFX); + uint32_t seq = ring->fence_drv.sync_seq; + uint64_t addr = ring->fence_drv.gpu_addr; + + gfx_v11_0_wait_reg_mem(ring, usepfp, 1, 0, lower_32_bits(addr), + upper_32_bits(addr), seq, 0xffffffff, 4); +} + +static void gfx_v11_0_ring_invalidate_tlbs(struct amdgpu_ring *ring, + uint16_t pasid, uint32_t flush_type, + bool all_hub, uint8_t dst_sel) +{ + amdgpu_ring_write(ring, PACKET3(PACKET3_INVALIDATE_TLBS, 0)); + amdgpu_ring_write(ring, + PACKET3_INVALIDATE_TLBS_DST_SEL(dst_sel) | + PACKET3_INVALIDATE_TLBS_ALL_HUB(all_hub) | + PACKET3_INVALIDATE_TLBS_PASID(pasid) | + PACKET3_INVALIDATE_TLBS_FLUSH_TYPE(flush_type)); +} + +static void gfx_v11_0_ring_emit_vm_flush(struct amdgpu_ring *ring, + unsigned vmid, uint64_t pd_addr) +{ + if (ring->is_mes_queue) + gfx_v11_0_ring_invalidate_tlbs(ring, 0, 0, false, 0); + else + amdgpu_gmc_emit_flush_gpu_tlb(ring, vmid, pd_addr); + + /* compute doesn't have PFP */ + if (ring->funcs->type == AMDGPU_RING_TYPE_GFX) { + /* sync PFP to ME, otherwise we might get invalid PFP reads */ + amdgpu_ring_write(ring, PACKET3(PACKET3_PFP_SYNC_ME, 0)); + amdgpu_ring_write(ring, 0x0); + } +} + +static void gfx_v11_0_ring_emit_fence_kiq(struct amdgpu_ring *ring, u64 addr, + u64 seq, unsigned int flags) +{ + struct amdgpu_device *adev = ring->adev; + + /* we only allocate 32bit for each seq wb address */ + BUG_ON(flags & AMDGPU_FENCE_FLAG_64BIT); + + /* write fence seq to the "addr" */ + amdgpu_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, 3)); + amdgpu_ring_write(ring, (WRITE_DATA_ENGINE_SEL(0) | + WRITE_DATA_DST_SEL(5) | WR_CONFIRM)); + amdgpu_ring_write(ring, lower_32_bits(addr)); + amdgpu_ring_write(ring, upper_32_bits(addr)); + amdgpu_ring_write(ring, lower_32_bits(seq)); + + if (flags & AMDGPU_FENCE_FLAG_INT) { + /* set register to trigger INT */ + amdgpu_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, 3)); + amdgpu_ring_write(ring, (WRITE_DATA_ENGINE_SEL(0) | + WRITE_DATA_DST_SEL(0) | WR_CONFIRM)); + amdgpu_ring_write(ring, SOC15_REG_OFFSET(GC, 0, regCPC_INT_STATUS)); + amdgpu_ring_write(ring, 0); + amdgpu_ring_write(ring, 0x20000000); /* src_id is 178 */ + } +} + +static void gfx_v11_0_ring_emit_cntxcntl(struct amdgpu_ring *ring, + uint32_t flags) +{ + uint32_t dw2 = 0; + + dw2 |= 0x80000000; /* set load_enable otherwise this package is just NOPs */ + if (flags & AMDGPU_HAVE_CTX_SWITCH) { + /* set load_global_config & load_global_uconfig */ + dw2 |= 0x8001; + /* set load_cs_sh_regs */ + dw2 |= 0x01000000; + /* set load_per_context_state & load_gfx_sh_regs for GFX */ + dw2 |= 0x10002; + } + + amdgpu_ring_write(ring, PACKET3(PACKET3_CONTEXT_CONTROL, 1)); + amdgpu_ring_write(ring, dw2); + amdgpu_ring_write(ring, 0); +} + +static unsigned gfx_v11_0_ring_emit_init_cond_exec(struct amdgpu_ring *ring) +{ + unsigned ret; + + amdgpu_ring_write(ring, PACKET3(PACKET3_COND_EXEC, 3)); + amdgpu_ring_write(ring, lower_32_bits(ring->cond_exe_gpu_addr)); + amdgpu_ring_write(ring, upper_32_bits(ring->cond_exe_gpu_addr)); + amdgpu_ring_write(ring, 0); /* discard following DWs if *cond_exec_gpu_addr==0 */ + ret = ring->wptr & ring->buf_mask; + amdgpu_ring_write(ring, 0x55aa55aa); /* patch dummy value later */ + + return ret; +} + +static void gfx_v11_0_ring_emit_patch_cond_exec(struct amdgpu_ring *ring, unsigned offset) +{ + unsigned cur; + BUG_ON(offset > ring->buf_mask); + BUG_ON(ring->ring[offset] != 0x55aa55aa); + + cur = (ring->wptr - 1) & ring->buf_mask; + if (likely(cur > offset)) + ring->ring[offset] = cur - offset; + else + ring->ring[offset] = (ring->buf_mask + 1) - offset + cur; +} + +static int gfx_v11_0_ring_preempt_ib(struct amdgpu_ring *ring) +{ + int i, r = 0; + struct amdgpu_device *adev = ring->adev; + struct amdgpu_kiq *kiq = &adev->gfx.kiq; + struct amdgpu_ring *kiq_ring = &kiq->ring; + unsigned long flags; + + if (!kiq->pmf || !kiq->pmf->kiq_unmap_queues) + return -EINVAL; + + spin_lock_irqsave(&kiq->ring_lock, flags); + + if (amdgpu_ring_alloc(kiq_ring, kiq->pmf->unmap_queues_size)) { + spin_unlock_irqrestore(&kiq->ring_lock, flags); + return -ENOMEM; + } + + /* assert preemption condition */ + amdgpu_ring_set_preempt_cond_exec(ring, false); + + /* assert IB preemption, emit the trailing fence */ + kiq->pmf->kiq_unmap_queues(kiq_ring, ring, PREEMPT_QUEUES_NO_UNMAP, + ring->trail_fence_gpu_addr, + ++ring->trail_seq); + amdgpu_ring_commit(kiq_ring); + + spin_unlock_irqrestore(&kiq->ring_lock, flags); + + /* poll the trailing fence */ + for (i = 0; i < adev->usec_timeout; i++) { + if (ring->trail_seq == + le32_to_cpu(*(ring->trail_fence_cpu_addr))) + break; + udelay(1); + } + + if (i >= adev->usec_timeout) { + r = -EINVAL; + DRM_ERROR("ring %d failed to preempt ib\n", ring->idx); + } + + /* deassert preemption condition */ + amdgpu_ring_set_preempt_cond_exec(ring, true); + return r; +} + +static void gfx_v11_0_ring_emit_de_meta(struct amdgpu_ring *ring, bool resume) +{ + struct amdgpu_device *adev = ring->adev; + struct v10_de_ib_state de_payload = {0}; + uint64_t offset, gds_addr, de_payload_gpu_addr; + void *de_payload_cpu_addr; + int cnt; + + if (ring->is_mes_queue) { + offset = offsetof(struct amdgpu_mes_ctx_meta_data, + gfx[0].gfx_meta_data) + + offsetof(struct v10_gfx_meta_data, de_payload); + de_payload_gpu_addr = + amdgpu_mes_ctx_get_offs_gpu_addr(ring, offset); + de_payload_cpu_addr = + amdgpu_mes_ctx_get_offs_cpu_addr(ring, offset); + + offset = offsetof(struct amdgpu_mes_ctx_meta_data, + gfx[0].gds_backup) + + offsetof(struct v10_gfx_meta_data, de_payload); + gds_addr = amdgpu_mes_ctx_get_offs_gpu_addr(ring, offset); + } else { + offset = offsetof(struct v10_gfx_meta_data, de_payload); + de_payload_gpu_addr = amdgpu_csa_vaddr(ring->adev) + offset; + de_payload_cpu_addr = adev->virt.csa_cpu_addr + offset; + + gds_addr = ALIGN(amdgpu_csa_vaddr(ring->adev) + + AMDGPU_CSA_SIZE - adev->gds.gds_size, + PAGE_SIZE); + } + + de_payload.gds_backup_addrlo = lower_32_bits(gds_addr); + de_payload.gds_backup_addrhi = upper_32_bits(gds_addr); + + cnt = (sizeof(de_payload) >> 2) + 4 - 2; + amdgpu_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, cnt)); + amdgpu_ring_write(ring, (WRITE_DATA_ENGINE_SEL(1) | + WRITE_DATA_DST_SEL(8) | + WR_CONFIRM) | + WRITE_DATA_CACHE_POLICY(0)); + amdgpu_ring_write(ring, lower_32_bits(de_payload_gpu_addr)); + amdgpu_ring_write(ring, upper_32_bits(de_payload_gpu_addr)); + + if (resume) + amdgpu_ring_write_multiple(ring, de_payload_cpu_addr, + sizeof(de_payload) >> 2); + else + amdgpu_ring_write_multiple(ring, (void *)&de_payload, + sizeof(de_payload) >> 2); +} + +static void gfx_v11_0_ring_emit_frame_cntl(struct amdgpu_ring *ring, bool start, + bool secure) +{ + uint32_t v = secure ? FRAME_TMZ : 0; + + amdgpu_ring_write(ring, PACKET3(PACKET3_FRAME_CONTROL, 0)); + amdgpu_ring_write(ring, v | FRAME_CMD(start ? 0 : 1)); +} + +static void gfx_v11_0_ring_emit_rreg(struct amdgpu_ring *ring, uint32_t reg, + uint32_t reg_val_offs) +{ + struct amdgpu_device *adev = ring->adev; + + amdgpu_ring_write(ring, PACKET3(PACKET3_COPY_DATA, 4)); + amdgpu_ring_write(ring, 0 | /* src: register*/ + (5 << 8) | /* dst: memory */ + (1 << 20)); /* write confirm */ + amdgpu_ring_write(ring, reg); + amdgpu_ring_write(ring, 0); + amdgpu_ring_write(ring, lower_32_bits(adev->wb.gpu_addr + + reg_val_offs * 4)); + amdgpu_ring_write(ring, upper_32_bits(adev->wb.gpu_addr + + reg_val_offs * 4)); +} + +static void gfx_v11_0_ring_emit_wreg(struct amdgpu_ring *ring, uint32_t reg, + uint32_t val) +{ + uint32_t cmd = 0; + + switch (ring->funcs->type) { + case AMDGPU_RING_TYPE_GFX: + cmd = WRITE_DATA_ENGINE_SEL(1) | WR_CONFIRM; + break; + case AMDGPU_RING_TYPE_KIQ: + cmd = (1 << 16); /* no inc addr */ + break; + default: + cmd = WR_CONFIRM; + break; + } + amdgpu_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, 3)); + amdgpu_ring_write(ring, cmd); + amdgpu_ring_write(ring, reg); + amdgpu_ring_write(ring, 0); + amdgpu_ring_write(ring, val); +} + +static void gfx_v11_0_ring_emit_reg_wait(struct amdgpu_ring *ring, uint32_t reg, + uint32_t val, uint32_t mask) +{ + gfx_v11_0_wait_reg_mem(ring, 0, 0, 0, reg, 0, val, mask, 0x20); +} + +static void gfx_v11_0_ring_emit_reg_write_reg_wait(struct amdgpu_ring *ring, + uint32_t reg0, uint32_t reg1, + uint32_t ref, uint32_t mask) +{ + int usepfp = (ring->funcs->type == AMDGPU_RING_TYPE_GFX); + + gfx_v11_0_wait_reg_mem(ring, usepfp, 0, 1, reg0, reg1, + ref, mask, 0x20); +} + +static void gfx_v11_0_ring_soft_recovery(struct amdgpu_ring *ring, + unsigned vmid) +{ + struct amdgpu_device *adev = ring->adev; + uint32_t value = 0; + + value = REG_SET_FIELD(value, SQ_CMD, CMD, 0x03); + value = REG_SET_FIELD(value, SQ_CMD, MODE, 0x01); + value = REG_SET_FIELD(value, SQ_CMD, CHECK_VMID, 1); + value = REG_SET_FIELD(value, SQ_CMD, VM_ID, vmid); + WREG32_SOC15(GC, 0, regSQ_CMD, value); +} + +static void +gfx_v11_0_set_gfx_eop_interrupt_state(struct amdgpu_device *adev, + uint32_t me, uint32_t pipe, + enum amdgpu_interrupt_state state) +{ + uint32_t cp_int_cntl, cp_int_cntl_reg; + + if (!me) { + switch (pipe) { + case 0: + cp_int_cntl_reg = SOC15_REG_OFFSET(GC, 0, regCP_INT_CNTL_RING0); + break; + case 1: + cp_int_cntl_reg = SOC15_REG_OFFSET(GC, 0, regCP_INT_CNTL_RING1); + break; + default: + DRM_DEBUG("invalid pipe %d\n", pipe); + return; + } + } else { + DRM_DEBUG("invalid me %d\n", me); + return; + } + + switch (state) { + case AMDGPU_IRQ_STATE_DISABLE: + cp_int_cntl = RREG32_SOC15_IP(GC, cp_int_cntl_reg); + cp_int_cntl = REG_SET_FIELD(cp_int_cntl, CP_INT_CNTL_RING0, + TIME_STAMP_INT_ENABLE, 0); + cp_int_cntl = REG_SET_FIELD(cp_int_cntl, CP_INT_CNTL_RING0, + GENERIC0_INT_ENABLE, 0); + WREG32_SOC15_IP(GC, cp_int_cntl_reg, cp_int_cntl); + break; + case AMDGPU_IRQ_STATE_ENABLE: + cp_int_cntl = RREG32_SOC15_IP(GC, cp_int_cntl_reg); + cp_int_cntl = REG_SET_FIELD(cp_int_cntl, CP_INT_CNTL_RING0, + TIME_STAMP_INT_ENABLE, 1); + cp_int_cntl = REG_SET_FIELD(cp_int_cntl, CP_INT_CNTL_RING0, + GENERIC0_INT_ENABLE, 1); + WREG32_SOC15_IP(GC, cp_int_cntl_reg, cp_int_cntl); + break; + default: + break; + } +} + +static void gfx_v11_0_set_compute_eop_interrupt_state(struct amdgpu_device *adev, + int me, int pipe, + enum amdgpu_interrupt_state state) +{ + u32 mec_int_cntl, mec_int_cntl_reg; + + /* + * amdgpu controls only the first MEC. That's why this function only + * handles the setting of interrupts for this specific MEC. All other + * pipes' interrupts are set by amdkfd. + */ + + if (me == 1) { + switch (pipe) { + case 0: + mec_int_cntl_reg = SOC15_REG_OFFSET(GC, 0, regCP_ME1_PIPE0_INT_CNTL); + break; + case 1: + mec_int_cntl_reg = SOC15_REG_OFFSET(GC, 0, regCP_ME1_PIPE1_INT_CNTL); + break; + case 2: + mec_int_cntl_reg = SOC15_REG_OFFSET(GC, 0, regCP_ME1_PIPE2_INT_CNTL); + break; + case 3: + mec_int_cntl_reg = SOC15_REG_OFFSET(GC, 0, regCP_ME1_PIPE3_INT_CNTL); + break; + default: + DRM_DEBUG("invalid pipe %d\n", pipe); + return; + } + } else { + DRM_DEBUG("invalid me %d\n", me); + return; + } + + switch (state) { + case AMDGPU_IRQ_STATE_DISABLE: + mec_int_cntl = RREG32_SOC15_IP(GC, mec_int_cntl_reg); + mec_int_cntl = REG_SET_FIELD(mec_int_cntl, CP_ME1_PIPE0_INT_CNTL, + TIME_STAMP_INT_ENABLE, 0); + mec_int_cntl = REG_SET_FIELD(mec_int_cntl, CP_ME1_PIPE0_INT_CNTL, + GENERIC0_INT_ENABLE, 0); + WREG32_SOC15_IP(GC, mec_int_cntl_reg, mec_int_cntl); + break; + case AMDGPU_IRQ_STATE_ENABLE: + mec_int_cntl = RREG32_SOC15_IP(GC, mec_int_cntl_reg); + mec_int_cntl = REG_SET_FIELD(mec_int_cntl, CP_ME1_PIPE0_INT_CNTL, + TIME_STAMP_INT_ENABLE, 1); + mec_int_cntl = REG_SET_FIELD(mec_int_cntl, CP_ME1_PIPE0_INT_CNTL, + GENERIC0_INT_ENABLE, 1); + WREG32_SOC15_IP(GC, mec_int_cntl_reg, mec_int_cntl); + break; + default: + break; + } +} + +static int gfx_v11_0_set_eop_interrupt_state(struct amdgpu_device *adev, + struct amdgpu_irq_src *src, + unsigned type, + enum amdgpu_interrupt_state state) +{ + switch (type) { + case AMDGPU_CP_IRQ_GFX_ME0_PIPE0_EOP: + gfx_v11_0_set_gfx_eop_interrupt_state(adev, 0, 0, state); + break; + case AMDGPU_CP_IRQ_GFX_ME0_PIPE1_EOP: + gfx_v11_0_set_gfx_eop_interrupt_state(adev, 0, 1, state); + break; + case AMDGPU_CP_IRQ_COMPUTE_MEC1_PIPE0_EOP: + gfx_v11_0_set_compute_eop_interrupt_state(adev, 1, 0, state); + break; + case AMDGPU_CP_IRQ_COMPUTE_MEC1_PIPE1_EOP: + gfx_v11_0_set_compute_eop_interrupt_state(adev, 1, 1, state); + break; + case AMDGPU_CP_IRQ_COMPUTE_MEC1_PIPE2_EOP: + gfx_v11_0_set_compute_eop_interrupt_state(adev, 1, 2, state); + break; + case AMDGPU_CP_IRQ_COMPUTE_MEC1_PIPE3_EOP: + gfx_v11_0_set_compute_eop_interrupt_state(adev, 1, 3, state); + break; + default: + break; + } + return 0; +} + +static int gfx_v11_0_eop_irq(struct amdgpu_device *adev, + struct amdgpu_irq_src *source, + struct amdgpu_iv_entry *entry) +{ + int i; + u8 me_id, pipe_id, queue_id; + struct amdgpu_ring *ring; + uint32_t mes_queue_id = entry->src_data[0]; + + DRM_DEBUG("IH: CP EOP\n"); + + if (adev->enable_mes && (mes_queue_id & AMDGPU_FENCE_MES_QUEUE_FLAG)) { + struct amdgpu_mes_queue *queue; + + mes_queue_id &= AMDGPU_FENCE_MES_QUEUE_ID_MASK; + + spin_lock(&adev->mes.queue_id_lock); + queue = idr_find(&adev->mes.queue_id_idr, mes_queue_id); + if (queue) { + DRM_DEBUG("process mes queue id = %d\n", mes_queue_id); + amdgpu_fence_process(queue->ring); + } + spin_unlock(&adev->mes.queue_id_lock); + } else { + me_id = (entry->ring_id & 0x0c) >> 2; + pipe_id = (entry->ring_id & 0x03) >> 0; + queue_id = (entry->ring_id & 0x70) >> 4; + + switch (me_id) { + case 0: + if (pipe_id == 0) + amdgpu_fence_process(&adev->gfx.gfx_ring[0]); + else + amdgpu_fence_process(&adev->gfx.gfx_ring[1]); + break; + case 1: + case 2: + for (i = 0; i < adev->gfx.num_compute_rings; i++) { + ring = &adev->gfx.compute_ring[i]; + /* Per-queue interrupt is supported for MEC starting from VI. + * The interrupt can only be enabled/disabled per pipe instead + * of per queue. + */ + if ((ring->me == me_id) && + (ring->pipe == pipe_id) && + (ring->queue == queue_id)) + amdgpu_fence_process(ring); + } + break; + } + } + + return 0; +} + +static int gfx_v11_0_set_priv_reg_fault_state(struct amdgpu_device *adev, + struct amdgpu_irq_src *source, + unsigned type, + enum amdgpu_interrupt_state state) +{ + switch (state) { + case AMDGPU_IRQ_STATE_DISABLE: + case AMDGPU_IRQ_STATE_ENABLE: + WREG32_FIELD15_PREREG(GC, 0, CP_INT_CNTL_RING0, + PRIV_REG_INT_ENABLE, + state == AMDGPU_IRQ_STATE_ENABLE ? 1 : 0); + break; + default: + break; + } + + return 0; +} + +static int gfx_v11_0_set_priv_inst_fault_state(struct amdgpu_device *adev, + struct amdgpu_irq_src *source, + unsigned type, + enum amdgpu_interrupt_state state) +{ + switch (state) { + case AMDGPU_IRQ_STATE_DISABLE: + case AMDGPU_IRQ_STATE_ENABLE: + WREG32_FIELD15_PREREG(GC, 0, CP_INT_CNTL_RING0, + PRIV_INSTR_INT_ENABLE, + state == AMDGPU_IRQ_STATE_ENABLE ? 1 : 0); + break; + default: + break; + } + + return 0; +} + +static void gfx_v11_0_handle_priv_fault(struct amdgpu_device *adev, + struct amdgpu_iv_entry *entry) +{ + u8 me_id, pipe_id, queue_id; + struct amdgpu_ring *ring; + int i; + + me_id = (entry->ring_id & 0x0c) >> 2; + pipe_id = (entry->ring_id & 0x03) >> 0; + queue_id = (entry->ring_id & 0x70) >> 4; + + switch (me_id) { + case 0: + for (i = 0; i < adev->gfx.num_gfx_rings; i++) { + ring = &adev->gfx.gfx_ring[i]; + /* we only enabled 1 gfx queue per pipe for now */ + if (ring->me == me_id && ring->pipe == pipe_id) + drm_sched_fault(&ring->sched); + } + break; + case 1: + case 2: + for (i = 0; i < adev->gfx.num_compute_rings; i++) { + ring = &adev->gfx.compute_ring[i]; + if (ring->me == me_id && ring->pipe == pipe_id && + ring->queue == queue_id) + drm_sched_fault(&ring->sched); + } + break; + default: + BUG(); + break; + } +} + +static int gfx_v11_0_priv_reg_irq(struct amdgpu_device *adev, + struct amdgpu_irq_src *source, + struct amdgpu_iv_entry *entry) +{ + DRM_ERROR("Illegal register access in command stream\n"); + gfx_v11_0_handle_priv_fault(adev, entry); + return 0; +} + +static int gfx_v11_0_priv_inst_irq(struct amdgpu_device *adev, + struct amdgpu_irq_src *source, + struct amdgpu_iv_entry *entry) +{ + DRM_ERROR("Illegal instruction in command stream\n"); + gfx_v11_0_handle_priv_fault(adev, entry); + return 0; +} + +#if 0 +static int gfx_v11_0_kiq_set_interrupt_state(struct amdgpu_device *adev, + struct amdgpu_irq_src *src, + unsigned int type, + enum amdgpu_interrupt_state state) +{ + uint32_t tmp, target; + struct amdgpu_ring *ring = &(adev->gfx.kiq.ring); + + target = SOC15_REG_OFFSET(GC, 0, regCP_ME1_PIPE0_INT_CNTL); + target += ring->pipe; + + switch (type) { + case AMDGPU_CP_KIQ_IRQ_DRIVER0: + if (state == AMDGPU_IRQ_STATE_DISABLE) { + tmp = RREG32_SOC15(GC, 0, regCPC_INT_CNTL); + tmp = REG_SET_FIELD(tmp, CPC_INT_CNTL, + GENERIC2_INT_ENABLE, 0); + WREG32_SOC15(GC, 0, regCPC_INT_CNTL, tmp); + + tmp = RREG32_SOC15_IP(GC, target); + tmp = REG_SET_FIELD(tmp, CP_ME1_PIPE0_INT_CNTL, + GENERIC2_INT_ENABLE, 0); + WREG32_SOC15_IP(GC, target, tmp); + } else { + tmp = RREG32_SOC15(GC, 0, regCPC_INT_CNTL); + tmp = REG_SET_FIELD(tmp, CPC_INT_CNTL, + GENERIC2_INT_ENABLE, 1); + WREG32_SOC15(GC, 0, regCPC_INT_CNTL, tmp); + + tmp = RREG32_SOC15_IP(GC, target); + tmp = REG_SET_FIELD(tmp, CP_ME1_PIPE0_INT_CNTL, + GENERIC2_INT_ENABLE, 1); + WREG32_SOC15_IP(GC, target, tmp); + } + break; + default: + BUG(); /* kiq only support GENERIC2_INT now */ + break; + } + return 0; +} +#endif + +static void gfx_v11_0_emit_mem_sync(struct amdgpu_ring *ring) +{ + const unsigned int gcr_cntl = + PACKET3_ACQUIRE_MEM_GCR_CNTL_GL2_INV(1) | + PACKET3_ACQUIRE_MEM_GCR_CNTL_GL2_WB(1) | + PACKET3_ACQUIRE_MEM_GCR_CNTL_GLM_INV(1) | + PACKET3_ACQUIRE_MEM_GCR_CNTL_GLM_WB(1) | + PACKET3_ACQUIRE_MEM_GCR_CNTL_GL1_INV(1) | + PACKET3_ACQUIRE_MEM_GCR_CNTL_GLV_INV(1) | + PACKET3_ACQUIRE_MEM_GCR_CNTL_GLK_INV(1) | + PACKET3_ACQUIRE_MEM_GCR_CNTL_GLI_INV(1); + + /* ACQUIRE_MEM - make one or more surfaces valid for use by the subsequent operations */ + amdgpu_ring_write(ring, PACKET3(PACKET3_ACQUIRE_MEM, 6)); + amdgpu_ring_write(ring, 0); /* CP_COHER_CNTL */ + amdgpu_ring_write(ring, 0xffffffff); /* CP_COHER_SIZE */ + amdgpu_ring_write(ring, 0xffffff); /* CP_COHER_SIZE_HI */ + amdgpu_ring_write(ring, 0); /* CP_COHER_BASE */ + amdgpu_ring_write(ring, 0); /* CP_COHER_BASE_HI */ + amdgpu_ring_write(ring, 0x0000000A); /* POLL_INTERVAL */ + amdgpu_ring_write(ring, gcr_cntl); /* GCR_CNTL */ +} + +static const struct amd_ip_funcs gfx_v11_0_ip_funcs = { + .name = "gfx_v11_0", + .early_init = gfx_v11_0_early_init, + .late_init = gfx_v11_0_late_init, + .sw_init = gfx_v11_0_sw_init, + .sw_fini = gfx_v11_0_sw_fini, + .hw_init = gfx_v11_0_hw_init, + .hw_fini = gfx_v11_0_hw_fini, + .suspend = gfx_v11_0_suspend, + .resume = gfx_v11_0_resume, + .is_idle = gfx_v11_0_is_idle, + .wait_for_idle = gfx_v11_0_wait_for_idle, + .soft_reset = gfx_v11_0_soft_reset, + .check_soft_reset = gfx_v11_0_check_soft_reset, + .set_clockgating_state = gfx_v11_0_set_clockgating_state, + .set_powergating_state = gfx_v11_0_set_powergating_state, + .get_clockgating_state = gfx_v11_0_get_clockgating_state, +}; + +static const struct amdgpu_ring_funcs gfx_v11_0_ring_funcs_gfx = { + .type = AMDGPU_RING_TYPE_GFX, + .align_mask = 0xff, + .nop = PACKET3(PACKET3_NOP, 0x3FFF), + .support_64bit_ptrs = true, + .vmhub = AMDGPU_GFXHUB_0, + .get_rptr = gfx_v11_0_ring_get_rptr_gfx, + .get_wptr = gfx_v11_0_ring_get_wptr_gfx, + .set_wptr = gfx_v11_0_ring_set_wptr_gfx, + .emit_frame_size = /* totally 242 maximum if 16 IBs */ + 5 + /* COND_EXEC */ + 7 + /* PIPELINE_SYNC */ + SOC15_FLUSH_GPU_TLB_NUM_WREG * 5 + + SOC15_FLUSH_GPU_TLB_NUM_REG_WAIT * 7 + + 2 + /* VM_FLUSH */ + 8 + /* FENCE for VM_FLUSH */ + 20 + /* GDS switch */ + 5 + /* COND_EXEC */ + 7 + /* HDP_flush */ + 4 + /* VGT_flush */ + 31 + /* DE_META */ + 3 + /* CNTX_CTRL */ + 5 + /* HDP_INVL */ + 8 + 8 + /* FENCE x2 */ + 8, /* gfx_v11_0_emit_mem_sync */ + .emit_ib_size = 4, /* gfx_v11_0_ring_emit_ib_gfx */ + .emit_ib = gfx_v11_0_ring_emit_ib_gfx, + .emit_fence = gfx_v11_0_ring_emit_fence, + .emit_pipeline_sync = gfx_v11_0_ring_emit_pipeline_sync, + .emit_vm_flush = gfx_v11_0_ring_emit_vm_flush, + .emit_gds_switch = gfx_v11_0_ring_emit_gds_switch, + .emit_hdp_flush = gfx_v11_0_ring_emit_hdp_flush, + .test_ring = gfx_v11_0_ring_test_ring, + .test_ib = gfx_v11_0_ring_test_ib, + .insert_nop = amdgpu_ring_insert_nop, + .pad_ib = amdgpu_ring_generic_pad_ib, + .emit_cntxcntl = gfx_v11_0_ring_emit_cntxcntl, + .init_cond_exec = gfx_v11_0_ring_emit_init_cond_exec, + .patch_cond_exec = gfx_v11_0_ring_emit_patch_cond_exec, + .preempt_ib = gfx_v11_0_ring_preempt_ib, + .emit_frame_cntl = gfx_v11_0_ring_emit_frame_cntl, + .emit_wreg = gfx_v11_0_ring_emit_wreg, + .emit_reg_wait = gfx_v11_0_ring_emit_reg_wait, + .emit_reg_write_reg_wait = gfx_v11_0_ring_emit_reg_write_reg_wait, + .soft_recovery = gfx_v11_0_ring_soft_recovery, + .emit_mem_sync = gfx_v11_0_emit_mem_sync, +}; + +static const struct amdgpu_ring_funcs gfx_v11_0_ring_funcs_compute = { + .type = AMDGPU_RING_TYPE_COMPUTE, + .align_mask = 0xff, + .nop = PACKET3(PACKET3_NOP, 0x3FFF), + .support_64bit_ptrs = true, + .vmhub = AMDGPU_GFXHUB_0, + .get_rptr = gfx_v11_0_ring_get_rptr_compute, + .get_wptr = gfx_v11_0_ring_get_wptr_compute, + .set_wptr = gfx_v11_0_ring_set_wptr_compute, + .emit_frame_size = + 20 + /* gfx_v11_0_ring_emit_gds_switch */ + 7 + /* gfx_v11_0_ring_emit_hdp_flush */ + 5 + /* hdp invalidate */ + 7 + /* gfx_v11_0_ring_emit_pipeline_sync */ + SOC15_FLUSH_GPU_TLB_NUM_WREG * 5 + + SOC15_FLUSH_GPU_TLB_NUM_REG_WAIT * 7 + + 2 + /* gfx_v11_0_ring_emit_vm_flush */ + 8 + 8 + 8 + /* gfx_v11_0_ring_emit_fence x3 for user fence, vm fence */ + 8, /* gfx_v11_0_emit_mem_sync */ + .emit_ib_size = 7, /* gfx_v11_0_ring_emit_ib_compute */ + .emit_ib = gfx_v11_0_ring_emit_ib_compute, + .emit_fence = gfx_v11_0_ring_emit_fence, + .emit_pipeline_sync = gfx_v11_0_ring_emit_pipeline_sync, + .emit_vm_flush = gfx_v11_0_ring_emit_vm_flush, + .emit_gds_switch = gfx_v11_0_ring_emit_gds_switch, + .emit_hdp_flush = gfx_v11_0_ring_emit_hdp_flush, + .test_ring = gfx_v11_0_ring_test_ring, + .test_ib = gfx_v11_0_ring_test_ib, + .insert_nop = amdgpu_ring_insert_nop, + .pad_ib = amdgpu_ring_generic_pad_ib, + .emit_wreg = gfx_v11_0_ring_emit_wreg, + .emit_reg_wait = gfx_v11_0_ring_emit_reg_wait, + .emit_reg_write_reg_wait = gfx_v11_0_ring_emit_reg_write_reg_wait, + .emit_mem_sync = gfx_v11_0_emit_mem_sync, +}; + +static const struct amdgpu_ring_funcs gfx_v11_0_ring_funcs_kiq = { + .type = AMDGPU_RING_TYPE_KIQ, + .align_mask = 0xff, + .nop = PACKET3(PACKET3_NOP, 0x3FFF), + .support_64bit_ptrs = true, + .vmhub = AMDGPU_GFXHUB_0, + .get_rptr = gfx_v11_0_ring_get_rptr_compute, + .get_wptr = gfx_v11_0_ring_get_wptr_compute, + .set_wptr = gfx_v11_0_ring_set_wptr_compute, + .emit_frame_size = + 20 + /* gfx_v11_0_ring_emit_gds_switch */ + 7 + /* gfx_v11_0_ring_emit_hdp_flush */ + 5 + /*hdp invalidate */ + 7 + /* gfx_v11_0_ring_emit_pipeline_sync */ + SOC15_FLUSH_GPU_TLB_NUM_WREG * 5 + + SOC15_FLUSH_GPU_TLB_NUM_REG_WAIT * 7 + + 2 + /* gfx_v11_0_ring_emit_vm_flush */ + 8 + 8 + 8, /* gfx_v11_0_ring_emit_fence_kiq x3 for user fence, vm fence */ + .emit_ib_size = 7, /* gfx_v11_0_ring_emit_ib_compute */ + .emit_ib = gfx_v11_0_ring_emit_ib_compute, + .emit_fence = gfx_v11_0_ring_emit_fence_kiq, + .test_ring = gfx_v11_0_ring_test_ring, + .test_ib = gfx_v11_0_ring_test_ib, + .insert_nop = amdgpu_ring_insert_nop, + .pad_ib = amdgpu_ring_generic_pad_ib, + .emit_rreg = gfx_v11_0_ring_emit_rreg, + .emit_wreg = gfx_v11_0_ring_emit_wreg, + .emit_reg_wait = gfx_v11_0_ring_emit_reg_wait, + .emit_reg_write_reg_wait = gfx_v11_0_ring_emit_reg_write_reg_wait, +}; + +static void gfx_v11_0_set_ring_funcs(struct amdgpu_device *adev) +{ + int i; + + adev->gfx.kiq.ring.funcs = &gfx_v11_0_ring_funcs_kiq; + + for (i = 0; i < adev->gfx.num_gfx_rings; i++) + adev->gfx.gfx_ring[i].funcs = &gfx_v11_0_ring_funcs_gfx; + + for (i = 0; i < adev->gfx.num_compute_rings; i++) + adev->gfx.compute_ring[i].funcs = &gfx_v11_0_ring_funcs_compute; +} + +static const struct amdgpu_irq_src_funcs gfx_v11_0_eop_irq_funcs = { + .set = gfx_v11_0_set_eop_interrupt_state, + .process = gfx_v11_0_eop_irq, +}; + +static const struct amdgpu_irq_src_funcs gfx_v11_0_priv_reg_irq_funcs = { + .set = gfx_v11_0_set_priv_reg_fault_state, + .process = gfx_v11_0_priv_reg_irq, +}; + +static const struct amdgpu_irq_src_funcs gfx_v11_0_priv_inst_irq_funcs = { + .set = gfx_v11_0_set_priv_inst_fault_state, + .process = gfx_v11_0_priv_inst_irq, +}; + +static void gfx_v11_0_set_irq_funcs(struct amdgpu_device *adev) +{ + adev->gfx.eop_irq.num_types = AMDGPU_CP_IRQ_LAST; + adev->gfx.eop_irq.funcs = &gfx_v11_0_eop_irq_funcs; + + adev->gfx.priv_reg_irq.num_types = 1; + adev->gfx.priv_reg_irq.funcs = &gfx_v11_0_priv_reg_irq_funcs; + + adev->gfx.priv_inst_irq.num_types = 1; + adev->gfx.priv_inst_irq.funcs = &gfx_v11_0_priv_inst_irq_funcs; +} + +static void gfx_v11_0_set_imu_funcs(struct amdgpu_device *adev) +{ + if (adev->flags & AMD_IS_APU) + adev->gfx.imu.mode = MISSION_MODE; + else + adev->gfx.imu.mode = DEBUG_MODE; + + adev->gfx.imu.funcs = &gfx_v11_0_imu_funcs; +} + +static void gfx_v11_0_set_rlc_funcs(struct amdgpu_device *adev) +{ + adev->gfx.rlc.funcs = &gfx_v11_0_rlc_funcs; +} + +static void gfx_v11_0_set_gds_init(struct amdgpu_device *adev) +{ + unsigned total_cu = adev->gfx.config.max_cu_per_sh * + adev->gfx.config.max_sh_per_se * + adev->gfx.config.max_shader_engines; + + adev->gds.gds_size = 0x1000; + adev->gds.gds_compute_max_wave_id = total_cu * 32 - 1; + adev->gds.gws_size = 64; + adev->gds.oa_size = 16; +} + +static void gfx_v11_0_set_mqd_funcs(struct amdgpu_device *adev) +{ + /* set gfx eng mqd */ + adev->mqds[AMDGPU_HW_IP_GFX].mqd_size = + sizeof(struct v11_gfx_mqd); + adev->mqds[AMDGPU_HW_IP_GFX].init_mqd = + gfx_v11_0_gfx_mqd_init; + /* set compute eng mqd */ + adev->mqds[AMDGPU_HW_IP_COMPUTE].mqd_size = + sizeof(struct v11_compute_mqd); + adev->mqds[AMDGPU_HW_IP_COMPUTE].init_mqd = + gfx_v11_0_compute_mqd_init; +} + +static void gfx_v11_0_set_user_wgp_inactive_bitmap_per_sh(struct amdgpu_device *adev, + u32 bitmap) +{ + u32 data; + + if (!bitmap) + return; + + data = bitmap << GC_USER_SHADER_ARRAY_CONFIG__INACTIVE_WGPS__SHIFT; + data &= GC_USER_SHADER_ARRAY_CONFIG__INACTIVE_WGPS_MASK; + + WREG32_SOC15(GC, 0, regGC_USER_SHADER_ARRAY_CONFIG, data); +} + +static u32 gfx_v11_0_get_wgp_active_bitmap_per_sh(struct amdgpu_device *adev) +{ + u32 data, wgp_bitmask; + data = RREG32_SOC15(GC, 0, regCC_GC_SHADER_ARRAY_CONFIG); + data |= RREG32_SOC15(GC, 0, regGC_USER_SHADER_ARRAY_CONFIG); + + data &= CC_GC_SHADER_ARRAY_CONFIG__INACTIVE_WGPS_MASK; + data >>= CC_GC_SHADER_ARRAY_CONFIG__INACTIVE_WGPS__SHIFT; + + wgp_bitmask = + amdgpu_gfx_create_bitmask(adev->gfx.config.max_cu_per_sh >> 1); + + return (~data) & wgp_bitmask; +} + +static u32 gfx_v11_0_get_cu_active_bitmap_per_sh(struct amdgpu_device *adev) +{ + u32 wgp_idx, wgp_active_bitmap; + u32 cu_bitmap_per_wgp, cu_active_bitmap; + + wgp_active_bitmap = gfx_v11_0_get_wgp_active_bitmap_per_sh(adev); + cu_active_bitmap = 0; + + for (wgp_idx = 0; wgp_idx < 16; wgp_idx++) { + /* if there is one WGP enabled, it means 2 CUs will be enabled */ + cu_bitmap_per_wgp = 3 << (2 * wgp_idx); + if (wgp_active_bitmap & (1 << wgp_idx)) + cu_active_bitmap |= cu_bitmap_per_wgp; + } + + return cu_active_bitmap; +} + +static int gfx_v11_0_get_cu_info(struct amdgpu_device *adev, + struct amdgpu_cu_info *cu_info) +{ + int i, j, k, counter, active_cu_number = 0; + u32 mask, bitmap; + unsigned disable_masks[8 * 2]; + + if (!adev || !cu_info) + return -EINVAL; + + amdgpu_gfx_parse_disable_cu(disable_masks, 8, 2); + + mutex_lock(&adev->grbm_idx_mutex); + for (i = 0; i < adev->gfx.config.max_shader_engines; i++) { + for (j = 0; j < adev->gfx.config.max_sh_per_se; j++) { + mask = 1; + counter = 0; + gfx_v11_0_select_se_sh(adev, i, j, 0xffffffff); + if (i < 8 && j < 2) + gfx_v11_0_set_user_wgp_inactive_bitmap_per_sh( + adev, disable_masks[i * 2 + j]); + bitmap = gfx_v11_0_get_cu_active_bitmap_per_sh(adev); + + /** + * GFX11 could support more than 4 SEs, while the bitmap + * in cu_info struct is 4x4 and ioctl interface struct + * drm_amdgpu_info_device should keep stable. + * So we use last two columns of bitmap to store cu mask for + * SEs 4 to 7, the layout of the bitmap is as below: + * SE0: {SH0,SH1} --> {bitmap[0][0], bitmap[0][1]} + * SE1: {SH0,SH1} --> {bitmap[1][0], bitmap[1][1]} + * SE2: {SH0,SH1} --> {bitmap[2][0], bitmap[2][1]} + * SE3: {SH0,SH1} --> {bitmap[3][0], bitmap[3][1]} + * SE4: {SH0,SH1} --> {bitmap[0][2], bitmap[0][3]} + * SE5: {SH0,SH1} --> {bitmap[1][2], bitmap[1][3]} + * SE6: {SH0,SH1} --> {bitmap[2][2], bitmap[2][3]} + * SE7: {SH0,SH1} --> {bitmap[3][2], bitmap[3][3]} + */ + cu_info->bitmap[i % 4][j + (i / 4) * 2] = bitmap; + + for (k = 0; k < adev->gfx.config.max_cu_per_sh; k++) { + if (bitmap & mask) + counter++; + + mask <<= 1; + } + active_cu_number += counter; + } + } + gfx_v11_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff); + mutex_unlock(&adev->grbm_idx_mutex); + + cu_info->number = active_cu_number; + cu_info->simd_per_cu = NUM_SIMD_PER_CU; + + return 0; +} + +const struct amdgpu_ip_block_version gfx_v11_0_ip_block = +{ + .type = AMD_IP_BLOCK_TYPE_GFX, + .major = 11, + .minor = 0, + .rev = 0, + .funcs = &gfx_v11_0_ip_funcs, +}; diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v11_0.h b/drivers/gpu/drm/amd/amdgpu/gfx_v11_0.h new file mode 100644 index 000000000000..10cfc29c27c9 --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v11_0.h @@ -0,0 +1,29 @@ +/* + * Copyright 2021 dvanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#ifndef __GFX_V11_0_H__ +#define __GFX_V11_0_H__ + +extern const struct amdgpu_ip_block_version gfx_v11_0_ip_block; + +#endif diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v6_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v6_0.c index 6a8dadea40f9..204b246f0e3f 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v6_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v6_0.c @@ -1778,39 +1778,26 @@ static void gfx_v6_0_constants_init(struct amdgpu_device *adev) udelay(50); } - -static void gfx_v6_0_scratch_init(struct amdgpu_device *adev) -{ - adev->gfx.scratch.num_reg = 8; - adev->gfx.scratch.reg_base = mmSCRATCH_REG0; - adev->gfx.scratch.free_mask = (1u << adev->gfx.scratch.num_reg) - 1; -} - static int gfx_v6_0_ring_test_ring(struct amdgpu_ring *ring) { struct amdgpu_device *adev = ring->adev; - uint32_t scratch; uint32_t tmp = 0; unsigned i; int r; - r = amdgpu_gfx_scratch_get(adev, &scratch); - if (r) - return r; - - WREG32(scratch, 0xCAFEDEAD); + WREG32(mmSCRATCH_REG0, 0xCAFEDEAD); r = amdgpu_ring_alloc(ring, 3); if (r) - goto error_free_scratch; + return r; amdgpu_ring_write(ring, PACKET3(PACKET3_SET_CONFIG_REG, 1)); - amdgpu_ring_write(ring, (scratch - PACKET3_SET_CONFIG_REG_START)); + amdgpu_ring_write(ring, mmSCRATCH_REG0 - PACKET3_SET_CONFIG_REG_START); amdgpu_ring_write(ring, 0xDEADBEEF); amdgpu_ring_commit(ring); for (i = 0; i < adev->usec_timeout; i++) { - tmp = RREG32(scratch); + tmp = RREG32(mmSCRATCH_REG0); if (tmp == 0xDEADBEEF) break; udelay(1); @@ -1818,9 +1805,6 @@ static int gfx_v6_0_ring_test_ring(struct amdgpu_ring *ring) if (i >= adev->usec_timeout) r = -ETIMEDOUT; - -error_free_scratch: - amdgpu_gfx_scratch_free(adev, scratch); return r; } @@ -1903,50 +1887,42 @@ static void gfx_v6_0_ring_emit_ib(struct amdgpu_ring *ring, static int gfx_v6_0_ring_test_ib(struct amdgpu_ring *ring, long timeout) { struct amdgpu_device *adev = ring->adev; - struct amdgpu_ib ib; struct dma_fence *f = NULL; - uint32_t scratch; + struct amdgpu_ib ib; uint32_t tmp = 0; long r; - r = amdgpu_gfx_scratch_get(adev, &scratch); - if (r) - return r; - - WREG32(scratch, 0xCAFEDEAD); + WREG32(mmSCRATCH_REG0, 0xCAFEDEAD); memset(&ib, 0, sizeof(ib)); - r = amdgpu_ib_get(adev, NULL, 256, - AMDGPU_IB_POOL_DIRECT, &ib); + r = amdgpu_ib_get(adev, NULL, 256, AMDGPU_IB_POOL_DIRECT, &ib); if (r) - goto err1; + return r; ib.ptr[0] = PACKET3(PACKET3_SET_CONFIG_REG, 1); - ib.ptr[1] = ((scratch - PACKET3_SET_CONFIG_REG_START)); + ib.ptr[1] = mmSCRATCH_REG0 - PACKET3_SET_CONFIG_REG_START; ib.ptr[2] = 0xDEADBEEF; ib.length_dw = 3; r = amdgpu_ib_schedule(ring, 1, &ib, NULL, &f); if (r) - goto err2; + goto error; r = dma_fence_wait_timeout(f, false, timeout); if (r == 0) { r = -ETIMEDOUT; - goto err2; + goto error; } else if (r < 0) { - goto err2; + goto error; } - tmp = RREG32(scratch); + tmp = RREG32(mmSCRATCH_REG0); if (tmp == 0xDEADBEEF) r = 0; else r = -EINVAL; -err2: +error: amdgpu_ib_free(adev, &ib, NULL); dma_fence_put(f); -err1: - amdgpu_gfx_scratch_free(adev, scratch); return r; } @@ -2117,7 +2093,7 @@ static int gfx_v6_0_cp_gfx_resume(struct amdgpu_device *adev) WREG32(mmCP_RB0_WPTR, ring->wptr); /* set the wb address whether it's enabled or not */ - rptr_addr = adev->wb.gpu_addr + (ring->rptr_offs * 4); + rptr_addr = ring->rptr_gpu_addr; WREG32(mmCP_RB0_RPTR_ADDR, lower_32_bits(rptr_addr)); WREG32(mmCP_RB0_RPTR_ADDR_HI, upper_32_bits(rptr_addr) & 0xFF); @@ -2139,7 +2115,7 @@ static int gfx_v6_0_cp_gfx_resume(struct amdgpu_device *adev) static u64 gfx_v6_0_ring_get_rptr(struct amdgpu_ring *ring) { - return ring->adev->wb.wb[ring->rptr_offs]; + return *ring->rptr_cpu_addr; } static u64 gfx_v6_0_ring_get_wptr(struct amdgpu_ring *ring) @@ -2203,7 +2179,7 @@ static int gfx_v6_0_cp_compute_resume(struct amdgpu_device *adev) ring->wptr = 0; WREG32(mmCP_RB1_WPTR, ring->wptr); - rptr_addr = adev->wb.gpu_addr + (ring->rptr_offs * 4); + rptr_addr = ring->rptr_gpu_addr; WREG32(mmCP_RB1_RPTR_ADDR, lower_32_bits(rptr_addr)); WREG32(mmCP_RB1_RPTR_ADDR_HI, upper_32_bits(rptr_addr) & 0xFF); @@ -2222,7 +2198,7 @@ static int gfx_v6_0_cp_compute_resume(struct amdgpu_device *adev) WREG32(mmCP_RB2_CNTL, tmp | CP_RB2_CNTL__RB_RPTR_WR_ENA_MASK); ring->wptr = 0; WREG32(mmCP_RB2_WPTR, ring->wptr); - rptr_addr = adev->wb.gpu_addr + (ring->rptr_offs * 4); + rptr_addr = ring->rptr_gpu_addr; WREG32(mmCP_RB2_RPTR_ADDR, lower_32_bits(rptr_addr)); WREG32(mmCP_RB2_RPTR_ADDR_HI, upper_32_bits(rptr_addr) & 0xFF); @@ -3094,8 +3070,6 @@ static int gfx_v6_0_sw_init(void *handle) if (r) return r; - gfx_v6_0_scratch_init(adev); - r = gfx_v6_0_init_microcode(adev); if (r) { DRM_ERROR("Failed to load gfx firmware!\n"); diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c index d17a6f399347..0f2976507e48 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c @@ -2049,26 +2049,6 @@ static void gfx_v7_0_constants_init(struct amdgpu_device *adev) udelay(50); } -/* - * GPU scratch registers helpers function. - */ -/** - * gfx_v7_0_scratch_init - setup driver info for CP scratch regs - * - * @adev: amdgpu_device pointer - * - * Set up the number and offset of the CP scratch registers. - * NOTE: use of CP scratch registers is a legacy interface and - * is not used by default on newer asics (r6xx+). On newer asics, - * memory buffers are used for fences rather than scratch regs. - */ -static void gfx_v7_0_scratch_init(struct amdgpu_device *adev) -{ - adev->gfx.scratch.num_reg = 8; - adev->gfx.scratch.reg_base = mmSCRATCH_REG0; - adev->gfx.scratch.free_mask = (1u << adev->gfx.scratch.num_reg) - 1; -} - /** * gfx_v7_0_ring_test_ring - basic gfx ring test * @@ -2082,36 +2062,28 @@ static void gfx_v7_0_scratch_init(struct amdgpu_device *adev) static int gfx_v7_0_ring_test_ring(struct amdgpu_ring *ring) { struct amdgpu_device *adev = ring->adev; - uint32_t scratch; uint32_t tmp = 0; unsigned i; int r; - r = amdgpu_gfx_scratch_get(adev, &scratch); - if (r) - return r; - - WREG32(scratch, 0xCAFEDEAD); + WREG32(mmSCRATCH_REG0, 0xCAFEDEAD); r = amdgpu_ring_alloc(ring, 3); if (r) - goto error_free_scratch; + return r; amdgpu_ring_write(ring, PACKET3(PACKET3_SET_UCONFIG_REG, 1)); - amdgpu_ring_write(ring, (scratch - PACKET3_SET_UCONFIG_REG_START)); + amdgpu_ring_write(ring, mmSCRATCH_REG0 - PACKET3_SET_UCONFIG_REG_START); amdgpu_ring_write(ring, 0xDEADBEEF); amdgpu_ring_commit(ring); for (i = 0; i < adev->usec_timeout; i++) { - tmp = RREG32(scratch); + tmp = RREG32(mmSCRATCH_REG0); if (tmp == 0xDEADBEEF) break; udelay(1); } if (i >= adev->usec_timeout) r = -ETIMEDOUT; - -error_free_scratch: - amdgpu_gfx_scratch_free(adev, scratch); return r; } @@ -2355,48 +2327,40 @@ static int gfx_v7_0_ring_test_ib(struct amdgpu_ring *ring, long timeout) struct amdgpu_device *adev = ring->adev; struct amdgpu_ib ib; struct dma_fence *f = NULL; - uint32_t scratch; uint32_t tmp = 0; long r; - r = amdgpu_gfx_scratch_get(adev, &scratch); - if (r) - return r; - - WREG32(scratch, 0xCAFEDEAD); + WREG32(mmSCRATCH_REG0, 0xCAFEDEAD); memset(&ib, 0, sizeof(ib)); - r = amdgpu_ib_get(adev, NULL, 256, - AMDGPU_IB_POOL_DIRECT, &ib); + r = amdgpu_ib_get(adev, NULL, 256, AMDGPU_IB_POOL_DIRECT, &ib); if (r) - goto err1; + return r; ib.ptr[0] = PACKET3(PACKET3_SET_UCONFIG_REG, 1); - ib.ptr[1] = ((scratch - PACKET3_SET_UCONFIG_REG_START)); + ib.ptr[1] = mmSCRATCH_REG0 - PACKET3_SET_UCONFIG_REG_START; ib.ptr[2] = 0xDEADBEEF; ib.length_dw = 3; r = amdgpu_ib_schedule(ring, 1, &ib, NULL, &f); if (r) - goto err2; + goto error; r = dma_fence_wait_timeout(f, false, timeout); if (r == 0) { r = -ETIMEDOUT; - goto err2; + goto error; } else if (r < 0) { - goto err2; + goto error; } - tmp = RREG32(scratch); + tmp = RREG32(mmSCRATCH_REG0); if (tmp == 0xDEADBEEF) r = 0; else r = -EINVAL; -err2: +error: amdgpu_ib_free(adev, &ib, NULL); dma_fence_put(f); -err1: - amdgpu_gfx_scratch_free(adev, scratch); return r; } @@ -2630,8 +2594,8 @@ static int gfx_v7_0_cp_gfx_resume(struct amdgpu_device *adev) ring->wptr = 0; WREG32(mmCP_RB0_WPTR, lower_32_bits(ring->wptr)); - /* set the wb address whether it's enabled or not */ - rptr_addr = adev->wb.gpu_addr + (ring->rptr_offs * 4); + /* set the wb address wether it's enabled or not */ + rptr_addr = ring->rptr_gpu_addr; WREG32(mmCP_RB0_RPTR_ADDR, lower_32_bits(rptr_addr)); WREG32(mmCP_RB0_RPTR_ADDR_HI, upper_32_bits(rptr_addr) & 0xFF); @@ -2656,7 +2620,7 @@ static int gfx_v7_0_cp_gfx_resume(struct amdgpu_device *adev) static u64 gfx_v7_0_ring_get_rptr(struct amdgpu_ring *ring) { - return ring->adev->wb.wb[ring->rptr_offs]; + return *ring->rptr_cpu_addr; } static u64 gfx_v7_0_ring_get_wptr_gfx(struct amdgpu_ring *ring) @@ -2677,7 +2641,7 @@ static void gfx_v7_0_ring_set_wptr_gfx(struct amdgpu_ring *ring) static u64 gfx_v7_0_ring_get_wptr_compute(struct amdgpu_ring *ring) { /* XXX check if swapping is necessary on BE */ - return ring->adev->wb.wb[ring->wptr_offs]; + return *ring->wptr_cpu_addr; } static void gfx_v7_0_ring_set_wptr_compute(struct amdgpu_ring *ring) @@ -2685,7 +2649,7 @@ static void gfx_v7_0_ring_set_wptr_compute(struct amdgpu_ring *ring) struct amdgpu_device *adev = ring->adev; /* XXX check if swapping is necessary on BE */ - adev->wb.wb[ring->wptr_offs] = lower_32_bits(ring->wptr); + *ring->wptr_cpu_addr = lower_32_bits(ring->wptr); WDOORBELL32(ring->doorbell_index, lower_32_bits(ring->wptr)); } @@ -2981,12 +2945,12 @@ static void gfx_v7_0_mqd_init(struct amdgpu_device *adev, CP_HQD_PQ_CONTROL__KMD_QUEUE_MASK; /* assuming kernel queue control */ /* only used if CP_PQ_WPTR_POLL_CNTL.CP_PQ_WPTR_POLL_CNTL__EN_MASK=1 */ - wb_gpu_addr = adev->wb.gpu_addr + (ring->wptr_offs * 4); + wb_gpu_addr = ring->wptr_gpu_addr; mqd->cp_hqd_pq_wptr_poll_addr_lo = wb_gpu_addr & 0xfffffffc; mqd->cp_hqd_pq_wptr_poll_addr_hi = upper_32_bits(wb_gpu_addr) & 0xffff; - /* set the wb address whether it's enabled or not */ - wb_gpu_addr = adev->wb.gpu_addr + (ring->rptr_offs * 4); + /* set the wb address wether it's enabled or not */ + wb_gpu_addr = ring->rptr_gpu_addr; mqd->cp_hqd_pq_rptr_report_addr_lo = wb_gpu_addr & 0xfffffffc; mqd->cp_hqd_pq_rptr_report_addr_hi = upper_32_bits(wb_gpu_addr) & 0xffff; @@ -4489,8 +4453,6 @@ static int gfx_v7_0_sw_init(void *handle) if (r) return r; - gfx_v7_0_scratch_init(adev); - r = gfx_v7_0_init_microcode(adev); if (r) { DRM_ERROR("Failed to load gfx firmware!\n"); diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c index 5f112efda634..7f0b18b0d4c4 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c @@ -835,37 +835,25 @@ static void gfx_v8_0_init_golden_registers(struct amdgpu_device *adev) } } -static void gfx_v8_0_scratch_init(struct amdgpu_device *adev) -{ - adev->gfx.scratch.num_reg = 8; - adev->gfx.scratch.reg_base = mmSCRATCH_REG0; - adev->gfx.scratch.free_mask = (1u << adev->gfx.scratch.num_reg) - 1; -} - static int gfx_v8_0_ring_test_ring(struct amdgpu_ring *ring) { struct amdgpu_device *adev = ring->adev; - uint32_t scratch; uint32_t tmp = 0; unsigned i; int r; - r = amdgpu_gfx_scratch_get(adev, &scratch); - if (r) - return r; - - WREG32(scratch, 0xCAFEDEAD); + WREG32(mmSCRATCH_REG0, 0xCAFEDEAD); r = amdgpu_ring_alloc(ring, 3); if (r) - goto error_free_scratch; + return r; amdgpu_ring_write(ring, PACKET3(PACKET3_SET_UCONFIG_REG, 1)); - amdgpu_ring_write(ring, (scratch - PACKET3_SET_UCONFIG_REG_START)); + amdgpu_ring_write(ring, mmSCRATCH_REG0 - PACKET3_SET_UCONFIG_REG_START); amdgpu_ring_write(ring, 0xDEADBEEF); amdgpu_ring_commit(ring); for (i = 0; i < adev->usec_timeout; i++) { - tmp = RREG32(scratch); + tmp = RREG32(mmSCRATCH_REG0); if (tmp == 0xDEADBEEF) break; udelay(1); @@ -874,8 +862,6 @@ static int gfx_v8_0_ring_test_ring(struct amdgpu_ring *ring) if (i >= adev->usec_timeout) r = -ETIMEDOUT; -error_free_scratch: - amdgpu_gfx_scratch_free(adev, scratch); return r; } @@ -1925,7 +1911,7 @@ static int gfx_v8_0_compute_ring_init(struct amdgpu_device *adev, int ring_id, + ring->pipe; hw_prio = amdgpu_gfx_is_high_priority_compute_queue(adev, ring) ? - AMDGPU_GFX_PIPE_PRIO_HIGH : AMDGPU_RING_PRIO_DEFAULT; + AMDGPU_RING_PRIO_2 : AMDGPU_RING_PRIO_DEFAULT; /* type-2 packets are deprecated on MEC, use type-3 instead */ r = amdgpu_ring_init(adev, ring, 1024, &adev->gfx.eop_irq, irq_type, hw_prio, NULL); @@ -2000,8 +1986,6 @@ static int gfx_v8_0_sw_init(void *handle) adev->gfx.gfx_current_status = AMDGPU_GFX_NORMAL_MODE; - gfx_v8_0_scratch_init(adev); - r = gfx_v8_0_init_microcode(adev); if (r) { DRM_ERROR("Failed to load gfx firmware!\n"); @@ -3730,7 +3714,7 @@ static void gfx_v8_0_init_compute_vmid(struct amdgpu_device *adev) mutex_unlock(&adev->srbm_mutex); /* Initialize all compute VMIDs to have no GDS, GWS, or OA - acccess. These should be enabled by FW for target VMIDs. */ + access. These should be enabled by FW for target VMIDs. */ for (i = adev->vm_manager.first_kfd_vmid; i < AMDGPU_NUM_VMID; i++) { WREG32(amdgpu_gds_reg_offset[i].mem_base, 0); WREG32(amdgpu_gds_reg_offset[i].mem_size, 0); @@ -4306,11 +4290,11 @@ static int gfx_v8_0_cp_gfx_resume(struct amdgpu_device *adev) WREG32(mmCP_RB0_WPTR, lower_32_bits(ring->wptr)); /* set the wb address wether it's enabled or not */ - rptr_addr = adev->wb.gpu_addr + (ring->rptr_offs * 4); + rptr_addr = ring->rptr_gpu_addr; WREG32(mmCP_RB0_RPTR_ADDR, lower_32_bits(rptr_addr)); WREG32(mmCP_RB0_RPTR_ADDR_HI, upper_32_bits(rptr_addr) & 0xFF); - wptr_gpu_addr = adev->wb.gpu_addr + (ring->wptr_offs * 4); + wptr_gpu_addr = ring->wptr_gpu_addr; WREG32(mmCP_RB_WPTR_POLL_ADDR_LO, lower_32_bits(wptr_gpu_addr)); WREG32(mmCP_RB_WPTR_POLL_ADDR_HI, upper_32_bits(wptr_gpu_addr)); mdelay(1); @@ -4393,7 +4377,7 @@ static int gfx_v8_0_kiq_kcq_enable(struct amdgpu_device *adev) for (i = 0; i < adev->gfx.num_compute_rings; i++) { struct amdgpu_ring *ring = &adev->gfx.compute_ring[i]; uint64_t mqd_addr = amdgpu_bo_gpu_offset(ring->mqd_obj); - uint64_t wptr_addr = adev->wb.gpu_addr + (ring->wptr_offs * 4); + uint64_t wptr_addr = ring->wptr_gpu_addr; /* map queues */ amdgpu_ring_write(kiq_ring, PACKET3(PACKET3_MAP_QUEUES, 5)); @@ -4506,7 +4490,7 @@ static int gfx_v8_0_mqd_init(struct amdgpu_ring *ring) tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, QUEUE_SIZE, (order_base_2(ring->ring_size / 4) - 1)); tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, RPTR_BLOCK_SIZE, - ((order_base_2(AMDGPU_GPU_PAGE_SIZE / 4) - 1) << 8)); + (order_base_2(AMDGPU_GPU_PAGE_SIZE / 4) - 1)); #ifdef __BIG_ENDIAN tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, ENDIAN_SWAP, 1); #endif @@ -4517,13 +4501,13 @@ static int gfx_v8_0_mqd_init(struct amdgpu_ring *ring) mqd->cp_hqd_pq_control = tmp; /* set the wb address whether it's enabled or not */ - wb_gpu_addr = adev->wb.gpu_addr + (ring->rptr_offs * 4); + wb_gpu_addr = ring->rptr_gpu_addr; mqd->cp_hqd_pq_rptr_report_addr_lo = wb_gpu_addr & 0xfffffffc; mqd->cp_hqd_pq_rptr_report_addr_hi = upper_32_bits(wb_gpu_addr) & 0xffff; /* only used if CP_PQ_WPTR_POLL_CNTL.CP_PQ_WPTR_POLL_CNTL__EN_MASK=1 */ - wb_gpu_addr = adev->wb.gpu_addr + (ring->wptr_offs * 4); + wb_gpu_addr = ring->wptr_gpu_addr; mqd->cp_hqd_pq_wptr_poll_addr_lo = wb_gpu_addr & 0xfffffffc; mqd->cp_hqd_pq_wptr_poll_addr_hi = upper_32_bits(wb_gpu_addr) & 0xffff; @@ -5475,7 +5459,7 @@ static int gfx_v8_0_set_powergating_state(void *handle, return 0; } -static void gfx_v8_0_get_clockgating_state(void *handle, u32 *flags) +static void gfx_v8_0_get_clockgating_state(void *handle, u64 *flags) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; int data; @@ -5831,7 +5815,7 @@ static void gfx_v8_0_update_coarse_grain_clock_gating(struct amdgpu_device *adev /* wait for RLC_SERDES_CU_MASTER & RLC_SERDES_NONCU_MASTER idle */ gfx_v8_0_wait_for_rlc_serdes(adev); - /* write cmd to Set CGCG Overrride */ + /* write cmd to Set CGCG Override */ gfx_v8_0_send_serdes_cmd(adev, BPM_REG_CGCG_OVERRIDE, SET_BPM_SERDES_CMD); /* wait for RLC_SERDES_CU_MASTER & RLC_SERDES_NONCU_MASTER idle */ @@ -6051,7 +6035,7 @@ static int gfx_v8_0_set_clockgating_state(void *handle, static u64 gfx_v8_0_ring_get_rptr(struct amdgpu_ring *ring) { - return ring->adev->wb.wb[ring->rptr_offs]; + return *ring->rptr_cpu_addr; } static u64 gfx_v8_0_ring_get_wptr_gfx(struct amdgpu_ring *ring) @@ -6060,7 +6044,7 @@ static u64 gfx_v8_0_ring_get_wptr_gfx(struct amdgpu_ring *ring) if (ring->use_doorbell) /* XXX check if swapping is necessary on BE */ - return ring->adev->wb.wb[ring->wptr_offs]; + return *ring->wptr_cpu_addr; else return RREG32(mmCP_RB0_WPTR); } @@ -6071,7 +6055,7 @@ static void gfx_v8_0_ring_set_wptr_gfx(struct amdgpu_ring *ring) if (ring->use_doorbell) { /* XXX check if swapping is necessary on BE */ - adev->wb.wb[ring->wptr_offs] = lower_32_bits(ring->wptr); + *ring->wptr_cpu_addr = lower_32_bits(ring->wptr); WDOORBELL32(ring->doorbell_index, lower_32_bits(ring->wptr)); } else { WREG32(mmCP_RB0_WPTR, lower_32_bits(ring->wptr)); @@ -6271,7 +6255,7 @@ static void gfx_v8_0_ring_emit_vm_flush(struct amdgpu_ring *ring, static u64 gfx_v8_0_ring_get_wptr_compute(struct amdgpu_ring *ring) { - return ring->adev->wb.wb[ring->wptr_offs]; + return *ring->wptr_cpu_addr; } static void gfx_v8_0_ring_set_wptr_compute(struct amdgpu_ring *ring) @@ -6279,7 +6263,7 @@ static void gfx_v8_0_ring_set_wptr_compute(struct amdgpu_ring *ring) struct amdgpu_device *adev = ring->adev; /* XXX check if swapping is necessary on BE */ - adev->wb.wb[ring->wptr_offs] = lower_32_bits(ring->wptr); + *ring->wptr_cpu_addr = lower_32_bits(ring->wptr); WDOORBELL32(ring->doorbell_index, lower_32_bits(ring->wptr)); } diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c index 9189fb85a4dd..0320be4a5fc6 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c @@ -63,13 +63,6 @@ #define mmGCEA_PROBE_MAP 0x070c #define mmGCEA_PROBE_MAP_BASE_IDX 0 -#define GFX9_RLCG_GC_WRITE_OLD (0x8 << 28) -#define GFX9_RLCG_GC_WRITE (0x0 << 28) -#define GFX9_RLCG_GC_READ (0x1 << 28) -#define GFX9_RLCG_VFGATE_DISABLED 0x4000000 -#define GFX9_RLCG_WRONG_OPERATION_TYPE 0x2000000 -#define GFX9_RLCG_NOT_IN_RANGE 0x1000000 - MODULE_FIRMWARE("amdgpu/vega10_ce.bin"); MODULE_FIRMWARE("amdgpu/vega10_pfp.bin"); MODULE_FIRMWARE("amdgpu/vega10_me.bin"); @@ -133,6 +126,8 @@ MODULE_FIRMWARE("amdgpu/green_sardine_rlc.bin"); MODULE_FIRMWARE("amdgpu/aldebaran_mec.bin"); MODULE_FIRMWARE("amdgpu/aldebaran_mec2.bin"); MODULE_FIRMWARE("amdgpu/aldebaran_rlc.bin"); +MODULE_FIRMWARE("amdgpu/aldebaran_sjt_mec.bin"); +MODULE_FIRMWARE("amdgpu/aldebaran_sjt_mec2.bin"); #define mmTCP_CHAN_STEER_0_ARCT 0x0b03 #define mmTCP_CHAN_STEER_0_ARCT_BASE_IDX 0 @@ -746,128 +741,6 @@ static const u32 GFX_RLC_SRM_INDEX_CNTL_DATA_OFFSETS[] = mmRLC_SRM_INDEX_CNTL_DATA_7 - mmRLC_SRM_INDEX_CNTL_DATA_0, }; -static u32 gfx_v9_0_rlcg_rw(struct amdgpu_device *adev, u32 offset, u32 v, uint32_t flag) -{ - static void *scratch_reg0; - static void *scratch_reg1; - static void *scratch_reg2; - static void *scratch_reg3; - static void *spare_int; - static uint32_t grbm_cntl; - static uint32_t grbm_idx; - uint32_t i = 0; - uint32_t retries = 50000; - u32 ret = 0; - u32 tmp; - - scratch_reg0 = adev->rmmio + (adev->reg_offset[GC_HWIP][0][mmSCRATCH_REG0_BASE_IDX] + mmSCRATCH_REG0)*4; - scratch_reg1 = adev->rmmio + (adev->reg_offset[GC_HWIP][0][mmSCRATCH_REG1_BASE_IDX] + mmSCRATCH_REG1)*4; - scratch_reg2 = adev->rmmio + (adev->reg_offset[GC_HWIP][0][mmSCRATCH_REG2_BASE_IDX] + mmSCRATCH_REG2)*4; - scratch_reg3 = adev->rmmio + (adev->reg_offset[GC_HWIP][0][mmSCRATCH_REG3_BASE_IDX] + mmSCRATCH_REG3)*4; - spare_int = adev->rmmio + (adev->reg_offset[GC_HWIP][0][mmRLC_SPARE_INT_BASE_IDX] + mmRLC_SPARE_INT)*4; - - grbm_cntl = adev->reg_offset[GC_HWIP][0][mmGRBM_GFX_CNTL_BASE_IDX] + mmGRBM_GFX_CNTL; - grbm_idx = adev->reg_offset[GC_HWIP][0][mmGRBM_GFX_INDEX_BASE_IDX] + mmGRBM_GFX_INDEX; - - if (offset == grbm_cntl || offset == grbm_idx) { - if (offset == grbm_cntl) - writel(v, scratch_reg2); - else if (offset == grbm_idx) - writel(v, scratch_reg3); - - writel(v, ((void __iomem *)adev->rmmio) + (offset * 4)); - } else { - /* - * SCRATCH_REG0 = read/write value - * SCRATCH_REG1[30:28] = command - * SCRATCH_REG1[19:0] = address in dword - * SCRATCH_REG1[26:24] = Error reporting - */ - writel(v, scratch_reg0); - writel(offset | flag, scratch_reg1); - writel(1, spare_int); - - for (i = 0; i < retries; i++) { - tmp = readl(scratch_reg1); - if (!(tmp & flag)) - break; - - udelay(10); - } - - if (i >= retries) { - if (amdgpu_sriov_reg_indirect_gc(adev)) { - if (tmp & GFX9_RLCG_VFGATE_DISABLED) - pr_err("The vfgate is disabled, program reg:0x%05x failed!\n", offset); - else if (tmp & GFX9_RLCG_WRONG_OPERATION_TYPE) - pr_err("Wrong operation type, program reg:0x%05x failed!\n", offset); - else if (tmp & GFX9_RLCG_NOT_IN_RANGE) - pr_err("The register is not in range, program reg:0x%05x failed!\n", offset); - else - pr_err("Unknown error type, program reg:0x%05x failed!\n", offset); - } else - pr_err("timeout: rlcg program reg:0x%05x failed!\n", offset); - } - } - - ret = readl(scratch_reg0); - - return ret; -} - -static bool gfx_v9_0_get_rlcg_flag(struct amdgpu_device *adev, u32 acc_flags, u32 hwip, - int write, u32 *rlcg_flag) -{ - - switch (hwip) { - case GC_HWIP: - if (amdgpu_sriov_reg_indirect_gc(adev)) { - *rlcg_flag = write ? GFX9_RLCG_GC_WRITE : GFX9_RLCG_GC_READ; - - return true; - /* only in new version, AMDGPU_REGS_NO_KIQ and AMDGPU_REGS_RLC enabled simultaneously */ - } else if ((acc_flags & AMDGPU_REGS_RLC) && !(acc_flags & AMDGPU_REGS_NO_KIQ) && write) { - *rlcg_flag = GFX9_RLCG_GC_WRITE_OLD; - return true; - } - - break; - default: - return false; - } - - return false; -} - -static u32 gfx_v9_0_sriov_rreg(struct amdgpu_device *adev, u32 offset, u32 acc_flags, u32 hwip) -{ - u32 rlcg_flag; - - if (!amdgpu_sriov_runtime(adev) && gfx_v9_0_get_rlcg_flag(adev, acc_flags, hwip, 0, &rlcg_flag)) - return gfx_v9_0_rlcg_rw(adev, offset, 0, rlcg_flag); - - if (acc_flags & AMDGPU_REGS_NO_KIQ) - return RREG32_NO_KIQ(offset); - else - return RREG32(offset); -} - -static void gfx_v9_0_sriov_wreg(struct amdgpu_device *adev, u32 offset, - u32 value, u32 acc_flags, u32 hwip) -{ - u32 rlcg_flag; - - if (!amdgpu_sriov_runtime(adev) && gfx_v9_0_get_rlcg_flag(adev, acc_flags, hwip, 1, &rlcg_flag)) { - gfx_v9_0_rlcg_rw(adev, offset, value, rlcg_flag); - return; - } - - if (acc_flags & AMDGPU_REGS_NO_KIQ) - WREG32_NO_KIQ(offset, value); - else - WREG32(offset, value); -} - #define VEGA10_GB_ADDR_CONFIG_GOLDEN 0x2a114042 #define VEGA12_GB_ADDR_CONFIG_GOLDEN 0x24104041 #define RAVEN_GB_ADDR_CONFIG_GOLDEN 0x24000042 @@ -882,7 +755,7 @@ static int gfx_v9_0_get_cu_info(struct amdgpu_device *adev, static uint64_t gfx_v9_0_get_gpu_clock_counter(struct amdgpu_device *adev); static void gfx_v9_0_ring_emit_de_meta(struct amdgpu_ring *ring); static u64 gfx_v9_0_ring_get_rptr_compute(struct amdgpu_ring *ring); -static int gfx_v9_0_query_ras_error_count(struct amdgpu_device *adev, +static void gfx_v9_0_query_ras_error_count(struct amdgpu_device *adev, void *ras_error_status); static int gfx_v9_0_ras_error_inject(struct amdgpu_device *adev, void *inject_if); @@ -909,9 +782,8 @@ static void gfx_v9_0_kiq_set_resources(struct amdgpu_ring *kiq_ring, static void gfx_v9_0_kiq_map_queues(struct amdgpu_ring *kiq_ring, struct amdgpu_ring *ring) { - struct amdgpu_device *adev = kiq_ring->adev; uint64_t mqd_addr = amdgpu_bo_gpu_offset(ring->mqd_obj); - uint64_t wptr_addr = adev->wb.gpu_addr + (ring->wptr_offs * 4); + uint64_t wptr_addr = ring->wptr_gpu_addr; uint32_t eng_sel = ring->funcs->type == AMDGPU_RING_TYPE_GFX ? 4 : 0; amdgpu_ring_write(kiq_ring, PACKET3(PACKET3_MAP_QUEUES, 5)); @@ -1080,13 +952,6 @@ static void gfx_v9_0_init_golden_registers(struct amdgpu_device *adev) (const u32)ARRAY_SIZE(golden_settings_gc_9_x_common)); } -static void gfx_v9_0_scratch_init(struct amdgpu_device *adev) -{ - adev->gfx.scratch.num_reg = 8; - adev->gfx.scratch.reg_base = SOC15_REG_OFFSET(GC, 0, mmSCRATCH_REG0); - adev->gfx.scratch.free_mask = (1u << adev->gfx.scratch.num_reg) - 1; -} - static void gfx_v9_0_write_data_to_reg(struct amdgpu_ring *ring, int eng_sel, bool wc, uint32_t reg, uint32_t val) { @@ -1124,22 +989,18 @@ static void gfx_v9_0_wait_reg_mem(struct amdgpu_ring *ring, int eng_sel, static int gfx_v9_0_ring_test_ring(struct amdgpu_ring *ring) { struct amdgpu_device *adev = ring->adev; - uint32_t scratch; + uint32_t scratch = SOC15_REG_OFFSET(GC, 0, mmSCRATCH_REG0); uint32_t tmp = 0; unsigned i; int r; - r = amdgpu_gfx_scratch_get(adev, &scratch); - if (r) - return r; - WREG32(scratch, 0xCAFEDEAD); r = amdgpu_ring_alloc(ring, 3); if (r) - goto error_free_scratch; + return r; amdgpu_ring_write(ring, PACKET3(PACKET3_SET_UCONFIG_REG, 1)); - amdgpu_ring_write(ring, (scratch - PACKET3_SET_UCONFIG_REG_START)); + amdgpu_ring_write(ring, scratch - PACKET3_SET_UCONFIG_REG_START); amdgpu_ring_write(ring, 0xDEADBEEF); amdgpu_ring_commit(ring); @@ -1152,9 +1013,6 @@ static int gfx_v9_0_ring_test_ring(struct amdgpu_ring *ring) if (i >= adev->usec_timeout) r = -ETIMEDOUT; - -error_free_scratch: - amdgpu_gfx_scratch_free(adev, scratch); return r; } @@ -1233,27 +1091,6 @@ static void gfx_v9_0_free_microcode(struct amdgpu_device *adev) kfree(adev->gfx.rlc.register_list_format); } -static void gfx_v9_0_init_rlc_ext_microcode(struct amdgpu_device *adev) -{ - const struct rlc_firmware_header_v2_1 *rlc_hdr; - - rlc_hdr = (const struct rlc_firmware_header_v2_1 *)adev->gfx.rlc_fw->data; - adev->gfx.rlc_srlc_fw_version = le32_to_cpu(rlc_hdr->save_restore_list_cntl_ucode_ver); - adev->gfx.rlc_srlc_feature_version = le32_to_cpu(rlc_hdr->save_restore_list_cntl_feature_ver); - adev->gfx.rlc.save_restore_list_cntl_size_bytes = le32_to_cpu(rlc_hdr->save_restore_list_cntl_size_bytes); - adev->gfx.rlc.save_restore_list_cntl = (u8 *)rlc_hdr + le32_to_cpu(rlc_hdr->save_restore_list_cntl_offset_bytes); - adev->gfx.rlc_srlg_fw_version = le32_to_cpu(rlc_hdr->save_restore_list_gpm_ucode_ver); - adev->gfx.rlc_srlg_feature_version = le32_to_cpu(rlc_hdr->save_restore_list_gpm_feature_ver); - adev->gfx.rlc.save_restore_list_gpm_size_bytes = le32_to_cpu(rlc_hdr->save_restore_list_gpm_size_bytes); - adev->gfx.rlc.save_restore_list_gpm = (u8 *)rlc_hdr + le32_to_cpu(rlc_hdr->save_restore_list_gpm_offset_bytes); - adev->gfx.rlc_srls_fw_version = le32_to_cpu(rlc_hdr->save_restore_list_srm_ucode_ver); - adev->gfx.rlc_srls_feature_version = le32_to_cpu(rlc_hdr->save_restore_list_srm_feature_ver); - adev->gfx.rlc.save_restore_list_srm_size_bytes = le32_to_cpu(rlc_hdr->save_restore_list_srm_size_bytes); - adev->gfx.rlc.save_restore_list_srm = (u8 *)rlc_hdr + le32_to_cpu(rlc_hdr->save_restore_list_srm_offset_bytes); - adev->gfx.rlc.reg_list_format_direct_reg_list_length = - le32_to_cpu(rlc_hdr->reg_list_format_direct_reg_list_length); -} - static void gfx_v9_0_check_fw_write_wait(struct amdgpu_device *adev) { adev->gfx.me_fw_write_wait = false; @@ -1334,6 +1171,8 @@ static const struct amdgpu_gfxoff_quirk amdgpu_gfxoff_quirk_list[] = { { 0x1002, 0x15dd, 0x103c, 0x83e7, 0xd3 }, /* GFXOFF is unstable on C6 parts with a VBIOS 113-RAVEN-114 */ { 0x1002, 0x15dd, 0x1002, 0x15dd, 0xc6 }, + /* Apple MacBook Pro (15-inch, 2019) Radeon Pro Vega 20 4 GB */ + { 0x1002, 0x69af, 0x106b, 0x019a, 0xc0 }, { 0, 0, 0, 0, 0 }, }; @@ -1413,9 +1252,6 @@ static int gfx_v9_0_init_cp_gfx_microcode(struct amdgpu_device *adev, { char fw_name[30]; int err; - struct amdgpu_firmware_info *info = NULL; - const struct common_firmware_header *header = NULL; - const struct gfx_firmware_header_v1_0 *cp_hdr; snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_pfp.bin", chip_name); err = request_firmware(&adev->gfx.pfp_fw, fw_name, adev->dev); @@ -1424,9 +1260,7 @@ static int gfx_v9_0_init_cp_gfx_microcode(struct amdgpu_device *adev, err = amdgpu_ucode_validate(adev->gfx.pfp_fw); if (err) goto out; - cp_hdr = (const struct gfx_firmware_header_v1_0 *)adev->gfx.pfp_fw->data; - adev->gfx.pfp_fw_version = le32_to_cpu(cp_hdr->header.ucode_version); - adev->gfx.pfp_feature_version = le32_to_cpu(cp_hdr->ucode_feature_version); + amdgpu_gfx_cp_init_microcode(adev, AMDGPU_UCODE_ID_CP_PFP); snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_me.bin", chip_name); err = request_firmware(&adev->gfx.me_fw, fw_name, adev->dev); @@ -1435,9 +1269,7 @@ static int gfx_v9_0_init_cp_gfx_microcode(struct amdgpu_device *adev, err = amdgpu_ucode_validate(adev->gfx.me_fw); if (err) goto out; - cp_hdr = (const struct gfx_firmware_header_v1_0 *)adev->gfx.me_fw->data; - adev->gfx.me_fw_version = le32_to_cpu(cp_hdr->header.ucode_version); - adev->gfx.me_feature_version = le32_to_cpu(cp_hdr->ucode_feature_version); + amdgpu_gfx_cp_init_microcode(adev, AMDGPU_UCODE_ID_CP_ME); snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_ce.bin", chip_name); err = request_firmware(&adev->gfx.ce_fw, fw_name, adev->dev); @@ -1446,37 +1278,12 @@ static int gfx_v9_0_init_cp_gfx_microcode(struct amdgpu_device *adev, err = amdgpu_ucode_validate(adev->gfx.ce_fw); if (err) goto out; - cp_hdr = (const struct gfx_firmware_header_v1_0 *)adev->gfx.ce_fw->data; - adev->gfx.ce_fw_version = le32_to_cpu(cp_hdr->header.ucode_version); - adev->gfx.ce_feature_version = le32_to_cpu(cp_hdr->ucode_feature_version); - - if (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) { - info = &adev->firmware.ucode[AMDGPU_UCODE_ID_CP_PFP]; - info->ucode_id = AMDGPU_UCODE_ID_CP_PFP; - info->fw = adev->gfx.pfp_fw; - header = (const struct common_firmware_header *)info->fw->data; - adev->firmware.fw_size += - ALIGN(le32_to_cpu(header->ucode_size_bytes), PAGE_SIZE); - - info = &adev->firmware.ucode[AMDGPU_UCODE_ID_CP_ME]; - info->ucode_id = AMDGPU_UCODE_ID_CP_ME; - info->fw = adev->gfx.me_fw; - header = (const struct common_firmware_header *)info->fw->data; - adev->firmware.fw_size += - ALIGN(le32_to_cpu(header->ucode_size_bytes), PAGE_SIZE); - - info = &adev->firmware.ucode[AMDGPU_UCODE_ID_CP_CE]; - info->ucode_id = AMDGPU_UCODE_ID_CP_CE; - info->fw = adev->gfx.ce_fw; - header = (const struct common_firmware_header *)info->fw->data; - adev->firmware.fw_size += - ALIGN(le32_to_cpu(header->ucode_size_bytes), PAGE_SIZE); - } + amdgpu_gfx_cp_init_microcode(adev, AMDGPU_UCODE_ID_CP_CE); out: if (err) { dev_err(adev->dev, - "gfx9: Failed to load firmware \"%s\"\n", + "gfx9: Failed to init firmware \"%s\"\n", fw_name); release_firmware(adev->gfx.pfp_fw); adev->gfx.pfp_fw = NULL; @@ -1493,11 +1300,7 @@ static int gfx_v9_0_init_rlc_microcode(struct amdgpu_device *adev, { char fw_name[30]; int err; - struct amdgpu_firmware_info *info = NULL; - const struct common_firmware_header *header = NULL; const struct rlc_firmware_header_v2_0 *rlc_hdr; - unsigned int *tmp = NULL; - unsigned int i = 0; uint16_t version_major; uint16_t version_minor; uint32_t smu_version; @@ -1526,92 +1329,17 @@ static int gfx_v9_0_init_rlc_microcode(struct amdgpu_device *adev, if (err) goto out; err = amdgpu_ucode_validate(adev->gfx.rlc_fw); + if (err) + goto out; rlc_hdr = (const struct rlc_firmware_header_v2_0 *)adev->gfx.rlc_fw->data; version_major = le16_to_cpu(rlc_hdr->header.header_version_major); version_minor = le16_to_cpu(rlc_hdr->header.header_version_minor); - if (version_major == 2 && version_minor == 1) - adev->gfx.rlc.is_rlc_v2_1 = true; - - adev->gfx.rlc_fw_version = le32_to_cpu(rlc_hdr->header.ucode_version); - adev->gfx.rlc_feature_version = le32_to_cpu(rlc_hdr->ucode_feature_version); - adev->gfx.rlc.save_and_restore_offset = - le32_to_cpu(rlc_hdr->save_and_restore_offset); - adev->gfx.rlc.clear_state_descriptor_offset = - le32_to_cpu(rlc_hdr->clear_state_descriptor_offset); - adev->gfx.rlc.avail_scratch_ram_locations = - le32_to_cpu(rlc_hdr->avail_scratch_ram_locations); - adev->gfx.rlc.reg_restore_list_size = - le32_to_cpu(rlc_hdr->reg_restore_list_size); - adev->gfx.rlc.reg_list_format_start = - le32_to_cpu(rlc_hdr->reg_list_format_start); - adev->gfx.rlc.reg_list_format_separate_start = - le32_to_cpu(rlc_hdr->reg_list_format_separate_start); - adev->gfx.rlc.starting_offsets_start = - le32_to_cpu(rlc_hdr->starting_offsets_start); - adev->gfx.rlc.reg_list_format_size_bytes = - le32_to_cpu(rlc_hdr->reg_list_format_size_bytes); - adev->gfx.rlc.reg_list_size_bytes = - le32_to_cpu(rlc_hdr->reg_list_size_bytes); - adev->gfx.rlc.register_list_format = - kmalloc(adev->gfx.rlc.reg_list_format_size_bytes + - adev->gfx.rlc.reg_list_size_bytes, GFP_KERNEL); - if (!adev->gfx.rlc.register_list_format) { - err = -ENOMEM; - goto out; - } - - tmp = (unsigned int *)((uintptr_t)rlc_hdr + - le32_to_cpu(rlc_hdr->reg_list_format_array_offset_bytes)); - for (i = 0 ; i < (adev->gfx.rlc.reg_list_format_size_bytes >> 2); i++) - adev->gfx.rlc.register_list_format[i] = le32_to_cpu(tmp[i]); - - adev->gfx.rlc.register_restore = adev->gfx.rlc.register_list_format + i; - - tmp = (unsigned int *)((uintptr_t)rlc_hdr + - le32_to_cpu(rlc_hdr->reg_list_array_offset_bytes)); - for (i = 0 ; i < (adev->gfx.rlc.reg_list_size_bytes >> 2); i++) - adev->gfx.rlc.register_restore[i] = le32_to_cpu(tmp[i]); - - if (adev->gfx.rlc.is_rlc_v2_1) - gfx_v9_0_init_rlc_ext_microcode(adev); - - if (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) { - info = &adev->firmware.ucode[AMDGPU_UCODE_ID_RLC_G]; - info->ucode_id = AMDGPU_UCODE_ID_RLC_G; - info->fw = adev->gfx.rlc_fw; - header = (const struct common_firmware_header *)info->fw->data; - adev->firmware.fw_size += - ALIGN(le32_to_cpu(header->ucode_size_bytes), PAGE_SIZE); - - if (adev->gfx.rlc.is_rlc_v2_1 && - adev->gfx.rlc.save_restore_list_cntl_size_bytes && - adev->gfx.rlc.save_restore_list_gpm_size_bytes && - adev->gfx.rlc.save_restore_list_srm_size_bytes) { - info = &adev->firmware.ucode[AMDGPU_UCODE_ID_RLC_RESTORE_LIST_CNTL]; - info->ucode_id = AMDGPU_UCODE_ID_RLC_RESTORE_LIST_CNTL; - info->fw = adev->gfx.rlc_fw; - adev->firmware.fw_size += - ALIGN(adev->gfx.rlc.save_restore_list_cntl_size_bytes, PAGE_SIZE); - - info = &adev->firmware.ucode[AMDGPU_UCODE_ID_RLC_RESTORE_LIST_GPM_MEM]; - info->ucode_id = AMDGPU_UCODE_ID_RLC_RESTORE_LIST_GPM_MEM; - info->fw = adev->gfx.rlc_fw; - adev->firmware.fw_size += - ALIGN(adev->gfx.rlc.save_restore_list_gpm_size_bytes, PAGE_SIZE); - - info = &adev->firmware.ucode[AMDGPU_UCODE_ID_RLC_RESTORE_LIST_SRM_MEM]; - info->ucode_id = AMDGPU_UCODE_ID_RLC_RESTORE_LIST_SRM_MEM; - info->fw = adev->gfx.rlc_fw; - adev->firmware.fw_size += - ALIGN(adev->gfx.rlc.save_restore_list_srm_size_bytes, PAGE_SIZE); - } - } - + err = amdgpu_gfx_rlc_init_microcode(adev, version_major, version_minor); out: if (err) { dev_err(adev->dev, - "gfx9: Failed to load firmware \"%s\"\n", + "gfx9: Failed to init firmware \"%s\"\n", fw_name); release_firmware(adev->gfx.rlc_fw); adev->gfx.rlc_fw = NULL; @@ -1634,35 +1362,34 @@ static int gfx_v9_0_init_cp_compute_microcode(struct amdgpu_device *adev, { char fw_name[30]; int err; - struct amdgpu_firmware_info *info = NULL; - const struct common_firmware_header *header = NULL; - const struct gfx_firmware_header_v1_0 *cp_hdr; - snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_mec.bin", chip_name); + if (amdgpu_sriov_vf(adev) && (adev->asic_type == CHIP_ALDEBARAN)) + snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_sjt_mec.bin", chip_name); + else + snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_mec.bin", chip_name); + err = request_firmware(&adev->gfx.mec_fw, fw_name, adev->dev); if (err) goto out; err = amdgpu_ucode_validate(adev->gfx.mec_fw); if (err) goto out; - cp_hdr = (const struct gfx_firmware_header_v1_0 *)adev->gfx.mec_fw->data; - adev->gfx.mec_fw_version = le32_to_cpu(cp_hdr->header.ucode_version); - adev->gfx.mec_feature_version = le32_to_cpu(cp_hdr->ucode_feature_version); - + amdgpu_gfx_cp_init_microcode(adev, AMDGPU_UCODE_ID_CP_MEC1); + amdgpu_gfx_cp_init_microcode(adev, AMDGPU_UCODE_ID_CP_MEC1_JT); if (gfx_v9_0_load_mec2_fw_bin_support(adev)) { - snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_mec2.bin", chip_name); + if (amdgpu_sriov_vf(adev) && (adev->asic_type == CHIP_ALDEBARAN)) + snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_sjt_mec2.bin", chip_name); + else + snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_mec2.bin", chip_name); + err = request_firmware(&adev->gfx.mec2_fw, fw_name, adev->dev); if (!err) { err = amdgpu_ucode_validate(adev->gfx.mec2_fw); if (err) goto out; - cp_hdr = (const struct gfx_firmware_header_v1_0 *) - adev->gfx.mec2_fw->data; - adev->gfx.mec2_fw_version = - le32_to_cpu(cp_hdr->header.ucode_version); - adev->gfx.mec2_feature_version = - le32_to_cpu(cp_hdr->ucode_feature_version); + amdgpu_gfx_cp_init_microcode(adev, AMDGPU_UCODE_ID_CP_MEC2); + amdgpu_gfx_cp_init_microcode(adev, AMDGPU_UCODE_ID_CP_MEC2_JT); } else { err = 0; adev->gfx.mec2_fw = NULL; @@ -1672,49 +1399,12 @@ static int gfx_v9_0_init_cp_compute_microcode(struct amdgpu_device *adev, adev->gfx.mec2_feature_version = adev->gfx.mec_feature_version; } - if (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) { - info = &adev->firmware.ucode[AMDGPU_UCODE_ID_CP_MEC1]; - info->ucode_id = AMDGPU_UCODE_ID_CP_MEC1; - info->fw = adev->gfx.mec_fw; - header = (const struct common_firmware_header *)info->fw->data; - cp_hdr = (const struct gfx_firmware_header_v1_0 *)info->fw->data; - adev->firmware.fw_size += - ALIGN(le32_to_cpu(header->ucode_size_bytes) - le32_to_cpu(cp_hdr->jt_size) * 4, PAGE_SIZE); - - info = &adev->firmware.ucode[AMDGPU_UCODE_ID_CP_MEC1_JT]; - info->ucode_id = AMDGPU_UCODE_ID_CP_MEC1_JT; - info->fw = adev->gfx.mec_fw; - adev->firmware.fw_size += - ALIGN(le32_to_cpu(cp_hdr->jt_size) * 4, PAGE_SIZE); - - if (adev->gfx.mec2_fw) { - info = &adev->firmware.ucode[AMDGPU_UCODE_ID_CP_MEC2]; - info->ucode_id = AMDGPU_UCODE_ID_CP_MEC2; - info->fw = adev->gfx.mec2_fw; - header = (const struct common_firmware_header *)info->fw->data; - cp_hdr = (const struct gfx_firmware_header_v1_0 *)info->fw->data; - adev->firmware.fw_size += - ALIGN(le32_to_cpu(header->ucode_size_bytes) - le32_to_cpu(cp_hdr->jt_size) * 4, PAGE_SIZE); - - /* TODO: Determine if MEC2 JT FW loading can be removed - for all GFX V9 asic and above */ - if (gfx_v9_0_load_mec2_fw_bin_support(adev)) { - info = &adev->firmware.ucode[AMDGPU_UCODE_ID_CP_MEC2_JT]; - info->ucode_id = AMDGPU_UCODE_ID_CP_MEC2_JT; - info->fw = adev->gfx.mec2_fw; - adev->firmware.fw_size += - ALIGN(le32_to_cpu(cp_hdr->jt_size) * 4, - PAGE_SIZE); - } - } - } - out: gfx_v9_0_check_if_need_gfxoff(adev); gfx_v9_0_check_fw_write_wait(adev); if (err) { dev_err(adev->dev, - "gfx9: Failed to load firmware \"%s\"\n", + "gfx9: Failed to init firmware \"%s\"\n", fw_name); release_firmware(adev->gfx.mec_fw); adev->gfx.mec_fw = NULL; @@ -2008,6 +1698,21 @@ static int gfx_v9_0_cp_jump_table_num(struct amdgpu_device *adev) return 4; } +static void gfx_v9_0_init_rlcg_reg_access_ctrl(struct amdgpu_device *adev) +{ + struct amdgpu_rlcg_reg_access_ctrl *reg_access_ctrl; + + reg_access_ctrl = &adev->gfx.rlc.reg_access_ctrl; + reg_access_ctrl->scratch_reg0 = SOC15_REG_OFFSET(GC, 0, mmSCRATCH_REG0); + reg_access_ctrl->scratch_reg1 = SOC15_REG_OFFSET(GC, 0, mmSCRATCH_REG1); + reg_access_ctrl->scratch_reg2 = SOC15_REG_OFFSET(GC, 0, mmSCRATCH_REG2); + reg_access_ctrl->scratch_reg3 = SOC15_REG_OFFSET(GC, 0, mmSCRATCH_REG3); + reg_access_ctrl->grbm_cntl = SOC15_REG_OFFSET(GC, 0, mmGRBM_GFX_CNTL); + reg_access_ctrl->grbm_idx = SOC15_REG_OFFSET(GC, 0, mmGRBM_GFX_INDEX); + reg_access_ctrl->spare_int = SOC15_REG_OFFSET(GC, 0, mmRLC_SPARE_INT); + adev->gfx.rlc.rlcg_reg_access_supported = true; +} + static int gfx_v9_0_rlc_init(struct amdgpu_device *adev) { const struct cs_section_def *cs_data; @@ -2197,12 +1902,16 @@ static const struct amdgpu_gfx_funcs gfx_v9_0_gfx_funcs = { .select_me_pipe_q = &gfx_v9_0_select_me_pipe_q, }; -static const struct amdgpu_gfx_ras_funcs gfx_v9_0_ras_funcs = { - .ras_late_init = amdgpu_gfx_ras_late_init, - .ras_fini = amdgpu_gfx_ras_fini, - .ras_error_inject = &gfx_v9_0_ras_error_inject, - .query_ras_error_count = &gfx_v9_0_query_ras_error_count, - .reset_ras_error_count = &gfx_v9_0_reset_ras_error_count, +const struct amdgpu_ras_block_hw_ops gfx_v9_0_ras_ops = { + .ras_error_inject = &gfx_v9_0_ras_error_inject, + .query_ras_error_count = &gfx_v9_0_query_ras_error_count, + .reset_ras_error_count = &gfx_v9_0_reset_ras_error_count, +}; + +static struct amdgpu_gfx_ras gfx_v9_0_ras = { + .ras_block = { + .hw_ops = &gfx_v9_0_ras_ops, + }, }; static int gfx_v9_0_gpu_early_init(struct amdgpu_device *adev) @@ -2231,7 +1940,7 @@ static int gfx_v9_0_gpu_early_init(struct amdgpu_device *adev) DRM_INFO("fix gfx.config for vega12\n"); break; case IP_VERSION(9, 4, 0): - adev->gfx.ras_funcs = &gfx_v9_0_ras_funcs; + adev->gfx.ras = &gfx_v9_0_ras; adev->gfx.config.max_hw_contexts = 8; adev->gfx.config.sc_prim_fifo_size_frontend = 0x20; adev->gfx.config.sc_prim_fifo_size_backend = 0x100; @@ -2258,7 +1967,7 @@ static int gfx_v9_0_gpu_early_init(struct amdgpu_device *adev) gb_addr_config = RAVEN_GB_ADDR_CONFIG_GOLDEN; break; case IP_VERSION(9, 4, 1): - adev->gfx.ras_funcs = &gfx_v9_4_ras_funcs; + adev->gfx.ras = &gfx_v9_4_ras; adev->gfx.config.max_hw_contexts = 8; adev->gfx.config.sc_prim_fifo_size_frontend = 0x20; adev->gfx.config.sc_prim_fifo_size_backend = 0x100; @@ -2279,7 +1988,7 @@ static int gfx_v9_0_gpu_early_init(struct amdgpu_device *adev) gb_addr_config |= 0x22010042; break; case IP_VERSION(9, 4, 2): - adev->gfx.ras_funcs = &gfx_v9_4_2_ras_funcs; + adev->gfx.ras = &gfx_v9_4_2_ras; adev->gfx.config.max_hw_contexts = 8; adev->gfx.config.sc_prim_fifo_size_frontend = 0x20; adev->gfx.config.sc_prim_fifo_size_backend = 0x100; @@ -2298,6 +2007,27 @@ static int gfx_v9_0_gpu_early_init(struct amdgpu_device *adev) break; } + if (adev->gfx.ras) { + err = amdgpu_ras_register_ras_block(adev, &adev->gfx.ras->ras_block); + if (err) { + DRM_ERROR("Failed to register gfx ras block!\n"); + return err; + } + + strcpy(adev->gfx.ras->ras_block.ras_comm.name, "gfx"); + adev->gfx.ras->ras_block.ras_comm.block = AMDGPU_RAS_BLOCK__GFX; + adev->gfx.ras->ras_block.ras_comm.type = AMDGPU_RAS_ERROR__MULTI_UNCORRECTABLE; + adev->gfx.ras_if = &adev->gfx.ras->ras_block.ras_comm; + + /* If not define special ras_late_init function, use gfx default ras_late_init */ + if (!adev->gfx.ras->ras_block.ras_late_init) + adev->gfx.ras->ras_block.ras_late_init = amdgpu_gfx_ras_late_init; + + /* If not defined special ras_cb function, use default ras_cb */ + if (!adev->gfx.ras->ras_block.ras_cb) + adev->gfx.ras->ras_block.ras_cb = amdgpu_gfx_process_ras_data_cb; + } + adev->gfx.config.gb_addr_config = gb_addr_config; adev->gfx.config.gb_addr_config_fields.num_pipes = 1 << @@ -2363,7 +2093,7 @@ static int gfx_v9_0_compute_ring_init(struct amdgpu_device *adev, int ring_id, + ((ring->me - 1) * adev->gfx.mec.num_pipe_per_mec) + ring->pipe; hw_prio = amdgpu_gfx_is_high_priority_compute_queue(adev, ring) ? - AMDGPU_GFX_PIPE_PRIO_HIGH : AMDGPU_GFX_PIPE_PRIO_NORMAL; + AMDGPU_RING_PRIO_2 : AMDGPU_RING_PRIO_DEFAULT; /* type-2 packets are deprecated on MEC, use type-3 instead */ return amdgpu_ring_init(adev, ring, 1024, &adev->gfx.eop_irq, irq_type, hw_prio, NULL); @@ -2426,18 +2156,20 @@ static int gfx_v9_0_sw_init(void *handle) adev->gfx.gfx_current_status = AMDGPU_GFX_NORMAL_MODE; - gfx_v9_0_scratch_init(adev); - r = gfx_v9_0_init_microcode(adev); if (r) { DRM_ERROR("Failed to load gfx firmware!\n"); return r; } - r = adev->gfx.rlc.funcs->init(adev); - if (r) { - DRM_ERROR("Failed to init rlc BOs!\n"); - return r; + if (adev->gfx.rlc.funcs) { + if (adev->gfx.rlc.funcs->init) { + r = adev->gfx.rlc.funcs->init(adev); + if (r) { + dev_err(adev->dev, "Failed to init rlc BOs!\n"); + return r; + } + } } r = gfx_v9_0_mec_init(adev); @@ -2513,10 +2245,6 @@ static int gfx_v9_0_sw_fini(void *handle) int i; struct amdgpu_device *adev = (struct amdgpu_device *)handle; - if (adev->gfx.ras_funcs && - adev->gfx.ras_funcs->ras_fini) - adev->gfx.ras_funcs->ras_fini(adev); - for (i = 0; i < adev->gfx.num_gfx_rings; i++) amdgpu_ring_fini(&adev->gfx.gfx_ring[i]); for (i = 0; i < adev->gfx.num_compute_rings; i++) @@ -2639,7 +2367,7 @@ static void gfx_v9_0_init_compute_vmid(struct amdgpu_device *adev) mutex_unlock(&adev->srbm_mutex); /* Initialize all compute VMIDs to have no GDS, GWS, or OA - acccess. These should be enabled by FW for target VMIDs. */ + access. These should be enabled by FW for target VMIDs. */ for (i = adev->vm_manager.first_kfd_vmid; i < AMDGPU_NUM_VMID; i++) { WREG32_SOC15_OFFSET(GC, 0, mmGDS_VMID0_BASE, 2 * i, 0); WREG32_SOC15_OFFSET(GC, 0, mmGDS_VMID0_SIZE, 2 * i, 0); @@ -2691,7 +2419,8 @@ static void gfx_v9_0_constants_init(struct amdgpu_device *adev) gfx_v9_0_tiling_mode_table_init(adev); - gfx_v9_0_setup_rb(adev); + if (adev->gfx.num_gfx_rings) + gfx_v9_0_setup_rb(adev); gfx_v9_0_get_cu_info(adev, &adev->gfx.cu_info); adev->gfx.config.db_debug2 = RREG32_SOC15(GC, 0, mmDB_DEBUG2); @@ -3413,11 +3142,11 @@ static int gfx_v9_0_cp_gfx_resume(struct amdgpu_device *adev) WREG32_SOC15(GC, 0, mmCP_RB0_WPTR_HI, upper_32_bits(ring->wptr)); /* set the wb address wether it's enabled or not */ - rptr_addr = adev->wb.gpu_addr + (ring->rptr_offs * 4); + rptr_addr = ring->rptr_gpu_addr; WREG32_SOC15(GC, 0, mmCP_RB0_RPTR_ADDR, lower_32_bits(rptr_addr)); WREG32_SOC15(GC, 0, mmCP_RB0_RPTR_ADDR_HI, upper_32_bits(rptr_addr) & CP_RB_RPTR_ADDR_HI__RB_RPTR_ADDR_HI_MASK); - wptr_gpu_addr = adev->wb.gpu_addr + (ring->wptr_offs * 4); + wptr_gpu_addr = ring->wptr_gpu_addr; WREG32_SOC15(GC, 0, mmCP_RB_WPTR_POLL_ADDR_LO, lower_32_bits(wptr_gpu_addr)); WREG32_SOC15(GC, 0, mmCP_RB_WPTR_POLL_ADDR_HI, upper_32_bits(wptr_gpu_addr)); @@ -3618,7 +3347,7 @@ static int gfx_v9_0_mqd_init(struct amdgpu_ring *ring) tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, QUEUE_SIZE, (order_base_2(ring->ring_size / 4) - 1)); tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, RPTR_BLOCK_SIZE, - ((order_base_2(AMDGPU_GPU_PAGE_SIZE / 4) - 1) << 8)); + (order_base_2(AMDGPU_GPU_PAGE_SIZE / 4) - 1)); #ifdef __BIG_ENDIAN tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, ENDIAN_SWAP, 1); #endif @@ -3629,33 +3358,16 @@ static int gfx_v9_0_mqd_init(struct amdgpu_ring *ring) mqd->cp_hqd_pq_control = tmp; /* set the wb address whether it's enabled or not */ - wb_gpu_addr = adev->wb.gpu_addr + (ring->rptr_offs * 4); + wb_gpu_addr = ring->rptr_gpu_addr; mqd->cp_hqd_pq_rptr_report_addr_lo = wb_gpu_addr & 0xfffffffc; mqd->cp_hqd_pq_rptr_report_addr_hi = upper_32_bits(wb_gpu_addr) & 0xffff; /* only used if CP_PQ_WPTR_POLL_CNTL.CP_PQ_WPTR_POLL_CNTL__EN_MASK=1 */ - wb_gpu_addr = adev->wb.gpu_addr + (ring->wptr_offs * 4); + wb_gpu_addr = ring->wptr_gpu_addr; mqd->cp_hqd_pq_wptr_poll_addr_lo = wb_gpu_addr & 0xfffffffc; mqd->cp_hqd_pq_wptr_poll_addr_hi = upper_32_bits(wb_gpu_addr) & 0xffff; - tmp = 0; - /* enable the doorbell if requested */ - if (ring->use_doorbell) { - tmp = RREG32_SOC15(GC, 0, mmCP_HQD_PQ_DOORBELL_CONTROL); - tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_DOORBELL_CONTROL, - DOORBELL_OFFSET, ring->doorbell_index); - - tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_DOORBELL_CONTROL, - DOORBELL_EN, 1); - tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_DOORBELL_CONTROL, - DOORBELL_SOURCE, 0); - tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_DOORBELL_CONTROL, - DOORBELL_HIT, 0); - } - - mqd->cp_hqd_pq_doorbell_control = tmp; - /* reset read and write pointers, similar to CP_RB0_WPTR/_RPTR */ ring->wptr = 0; mqd->cp_hqd_pq_rptr = RREG32_SOC15(GC, 0, mmCP_HQD_PQ_RPTR); @@ -3917,7 +3629,7 @@ static int gfx_v9_0_kcq_init_queue(struct amdgpu_ring *ring) /* reset ring buffer */ ring->wptr = 0; - atomic64_set((atomic64_t *)&adev->wb.wb[ring->wptr_offs], 0); + atomic64_set((atomic64_t *)ring->wptr_cpu_addr, 0); amdgpu_ring_clear_ring(ring); } else { amdgpu_ring_clear_ring(ring); @@ -4840,6 +4552,9 @@ static int gfx_v9_0_early_init(void *handle) gfx_v9_0_set_gds_init(adev); gfx_v9_0_set_rlc_funcs(adev); + /* init rlcg reg access ctrl */ + gfx_v9_0_init_rlcg_reg_access_ctrl(adev); + return 0; } @@ -4870,16 +4585,9 @@ static int gfx_v9_0_ecc_late_init(void *handle) if (r) return r; - if (adev->gfx.ras_funcs && - adev->gfx.ras_funcs->ras_late_init) { - r = adev->gfx.ras_funcs->ras_late_init(adev); - if (r) - return r; - } - - if (adev->gfx.ras_funcs && - adev->gfx.ras_funcs->enable_watchdog_timer) - adev->gfx.ras_funcs->enable_watchdog_timer(adev); + if (adev->gfx.ras && + adev->gfx.ras->enable_watchdog_timer) + adev->gfx.ras->enable_watchdog_timer(adev); return 0; } @@ -5250,8 +4958,6 @@ static const struct amdgpu_rlc_funcs gfx_v9_0_rlc_funcs = { .reset = gfx_v9_0_rlc_reset, .start = gfx_v9_0_rlc_start, .update_spm_vmid = gfx_v9_0_update_spm_vmid, - .sriov_wreg = gfx_v9_0_sriov_wreg, - .sriov_rreg = gfx_v9_0_sriov_rreg, .is_rlcg_access_range = gfx_v9_0_is_rlcg_access_range, }; @@ -5326,7 +5032,7 @@ static int gfx_v9_0_set_clockgating_state(void *handle, return 0; } -static void gfx_v9_0_get_clockgating_state(void *handle, u32 *flags) +static void gfx_v9_0_get_clockgating_state(void *handle, u64 *flags) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; int data; @@ -5372,7 +5078,7 @@ static void gfx_v9_0_get_clockgating_state(void *handle, u32 *flags) static u64 gfx_v9_0_ring_get_rptr_gfx(struct amdgpu_ring *ring) { - return ring->adev->wb.wb[ring->rptr_offs]; /* gfx9 is 32bit rptr*/ + return *ring->rptr_cpu_addr; /* gfx9 is 32bit rptr*/ } static u64 gfx_v9_0_ring_get_wptr_gfx(struct amdgpu_ring *ring) @@ -5382,7 +5088,7 @@ static u64 gfx_v9_0_ring_get_wptr_gfx(struct amdgpu_ring *ring) /* XXX check if swapping is necessary on BE */ if (ring->use_doorbell) { - wptr = atomic64_read((atomic64_t *)&adev->wb.wb[ring->wptr_offs]); + wptr = atomic64_read((atomic64_t *)ring->wptr_cpu_addr); } else { wptr = RREG32_SOC15(GC, 0, mmCP_RB0_WPTR); wptr += (u64)RREG32_SOC15(GC, 0, mmCP_RB0_WPTR_HI) << 32; @@ -5397,7 +5103,7 @@ static void gfx_v9_0_ring_set_wptr_gfx(struct amdgpu_ring *ring) if (ring->use_doorbell) { /* XXX check if swapping is necessary on BE */ - atomic64_set((atomic64_t *)&adev->wb.wb[ring->wptr_offs], ring->wptr); + atomic64_set((atomic64_t *)ring->wptr_cpu_addr, ring->wptr); WDOORBELL64(ring->doorbell_index, ring->wptr); } else { WREG32_SOC15(GC, 0, mmCP_RB0_WPTR, lower_32_bits(ring->wptr)); @@ -5562,7 +5268,7 @@ static void gfx_v9_0_ring_emit_vm_flush(struct amdgpu_ring *ring, static u64 gfx_v9_0_ring_get_rptr_compute(struct amdgpu_ring *ring) { - return ring->adev->wb.wb[ring->rptr_offs]; /* gfx9 hardware is 32bit rptr */ + return *ring->rptr_cpu_addr; /* gfx9 hardware is 32bit rptr */ } static u64 gfx_v9_0_ring_get_wptr_compute(struct amdgpu_ring *ring) @@ -5571,7 +5277,7 @@ static u64 gfx_v9_0_ring_get_wptr_compute(struct amdgpu_ring *ring) /* XXX check if swapping is necessary on BE */ if (ring->use_doorbell) - wptr = atomic64_read((atomic64_t *)&ring->adev->wb.wb[ring->wptr_offs]); + wptr = atomic64_read((atomic64_t *)ring->wptr_cpu_addr); else BUG(); return wptr; @@ -5583,7 +5289,7 @@ static void gfx_v9_0_ring_set_wptr_compute(struct amdgpu_ring *ring) /* XXX check if swapping is necessary on BE */ if (ring->use_doorbell) { - atomic64_set((atomic64_t *)&adev->wb.wb[ring->wptr_offs], ring->wptr); + atomic64_set((atomic64_t *)ring->wptr_cpu_addr, ring->wptr); WDOORBELL64(ring->doorbell_index, ring->wptr); } else{ BUG(); /* only DOORBELL method supported on gfx9 now */ @@ -5723,7 +5429,7 @@ static void gfx_v9_0_ring_emit_patch_cond_exec(struct amdgpu_ring *ring, unsigne BUG_ON(offset > ring->buf_mask); BUG_ON(ring->ring[offset] != 0x55aa55aa); - cur = (ring->wptr & ring->buf_mask) - 1; + cur = (ring->wptr - 1) & ring->buf_mask; if (likely(cur > offset)) ring->ring[offset] = cur - offset; else @@ -6819,7 +6525,7 @@ static void gfx_v9_0_reset_ras_error_count(struct amdgpu_device *adev) WREG32_SOC15(GC, 0, mmATC_L2_CACHE_4K_EDC_INDEX, 255); } -static int gfx_v9_0_query_ras_error_count(struct amdgpu_device *adev, +static void gfx_v9_0_query_ras_error_count(struct amdgpu_device *adev, void *ras_error_status) { struct ras_err_data *err_data = (struct ras_err_data *)ras_error_status; @@ -6828,7 +6534,7 @@ static int gfx_v9_0_query_ras_error_count(struct amdgpu_device *adev, uint32_t reg_value; if (!amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__GFX)) - return -EINVAL; + return; err_data->ue_count = 0; err_data->ce_count = 0; @@ -6857,8 +6563,6 @@ static int gfx_v9_0_query_ras_error_count(struct amdgpu_device *adev, mutex_unlock(&adev->grbm_idx_mutex); gfx_v9_0_query_utc_edc_status(adev, err_data); - - return 0; } static void gfx_v9_0_emit_mem_sync(struct amdgpu_ring *ring) @@ -6962,6 +6666,7 @@ static const struct amdgpu_ring_funcs gfx_v9_0_ring_funcs_gfx = { .align_mask = 0xff, .nop = PACKET3(PACKET3_NOP, 0x3FFF), .support_64bit_ptrs = true, + .secure_submission_supported = true, .vmhub = AMDGPU_GFXHUB_0, .get_rptr = gfx_v9_0_ring_get_rptr_gfx, .get_wptr = gfx_v9_0_ring_get_wptr_gfx, diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_4.c b/drivers/gpu/drm/amd/amdgpu/gfx_v9_4.c index b4789dfc2bb9..c67e387a97f5 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_4.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_4.c @@ -863,7 +863,7 @@ static int gfx_v9_4_ras_error_count(struct amdgpu_device *adev, return 0; } -static int gfx_v9_4_query_ras_error_count(struct amdgpu_device *adev, +static void gfx_v9_4_query_ras_error_count(struct amdgpu_device *adev, void *ras_error_status) { struct ras_err_data *err_data = (struct ras_err_data *)ras_error_status; @@ -872,7 +872,7 @@ static int gfx_v9_4_query_ras_error_count(struct amdgpu_device *adev, uint32_t reg_value; if (!amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__GFX)) - return -EINVAL; + return; err_data->ue_count = 0; err_data->ce_count = 0; @@ -903,7 +903,6 @@ static int gfx_v9_4_query_ras_error_count(struct amdgpu_device *adev, gfx_v9_4_query_utc_edc_status(adev, err_data); - return 0; } static void gfx_v9_4_reset_ras_error_count(struct amdgpu_device *adev) @@ -1029,11 +1028,16 @@ static void gfx_v9_4_query_ras_error_status(struct amdgpu_device *adev) mutex_unlock(&adev->grbm_idx_mutex); } -const struct amdgpu_gfx_ras_funcs gfx_v9_4_ras_funcs = { - .ras_late_init = amdgpu_gfx_ras_late_init, - .ras_fini = amdgpu_gfx_ras_fini, - .ras_error_inject = &gfx_v9_4_ras_error_inject, - .query_ras_error_count = &gfx_v9_4_query_ras_error_count, - .reset_ras_error_count = &gfx_v9_4_reset_ras_error_count, - .query_ras_error_status = &gfx_v9_4_query_ras_error_status, + +const struct amdgpu_ras_block_hw_ops gfx_v9_4_ras_ops = { + .ras_error_inject = &gfx_v9_4_ras_error_inject, + .query_ras_error_count = &gfx_v9_4_query_ras_error_count, + .reset_ras_error_count = &gfx_v9_4_reset_ras_error_count, + .query_ras_error_status = &gfx_v9_4_query_ras_error_status, +}; + +struct amdgpu_gfx_ras gfx_v9_4_ras = { + .ras_block = { + .hw_ops = &gfx_v9_4_ras_ops, + }, }; diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_4.h b/drivers/gpu/drm/amd/amdgpu/gfx_v9_4.h index bdd16b568021..ca520a767267 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_4.h +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_4.h @@ -24,6 +24,6 @@ #ifndef __GFX_V9_4_H__ #define __GFX_V9_4_H__ -extern const struct amdgpu_gfx_ras_funcs gfx_v9_4_ras_funcs; +extern struct amdgpu_gfx_ras gfx_v9_4_ras; #endif /* __GFX_V9_4_H__ */ diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_2.c b/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_2.c index c4f37a161875..3a797424579c 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_2.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_2.c @@ -1641,14 +1641,14 @@ static int gfx_v9_4_2_query_utc_edc_count(struct amdgpu_device *adev, return 0; } -static int gfx_v9_4_2_query_ras_error_count(struct amdgpu_device *adev, +static void gfx_v9_4_2_query_ras_error_count(struct amdgpu_device *adev, void *ras_error_status) { struct ras_err_data *err_data = (struct ras_err_data *)ras_error_status; uint32_t sec_count = 0, ded_count = 0; if (!amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__GFX)) - return -EINVAL; + return; err_data->ue_count = 0; err_data->ce_count = 0; @@ -1661,7 +1661,6 @@ static int gfx_v9_4_2_query_ras_error_count(struct amdgpu_device *adev, err_data->ce_count += sec_count; err_data->ue_count += ded_count; - return 0; } static void gfx_v9_4_2_reset_utc_err_status(struct amdgpu_device *adev) @@ -1931,13 +1930,31 @@ static void gfx_v9_4_2_reset_sq_timeout_status(struct amdgpu_device *adev) mutex_unlock(&adev->grbm_idx_mutex); } -const struct amdgpu_gfx_ras_funcs gfx_v9_4_2_ras_funcs = { - .ras_late_init = amdgpu_gfx_ras_late_init, - .ras_fini = amdgpu_gfx_ras_fini, - .ras_error_inject = &gfx_v9_4_2_ras_error_inject, - .query_ras_error_count = &gfx_v9_4_2_query_ras_error_count, - .reset_ras_error_count = &gfx_v9_4_2_reset_ras_error_count, - .query_ras_error_status = &gfx_v9_4_2_query_ras_error_status, - .reset_ras_error_status = &gfx_v9_4_2_reset_ras_error_status, +static bool gfx_v9_4_2_query_uctl2_poison_status(struct amdgpu_device *adev) +{ + u32 status = 0; + struct amdgpu_vmhub *hub; + + hub = &adev->vmhub[AMDGPU_GFXHUB_0]; + status = RREG32(hub->vm_l2_pro_fault_status); + /* reset page fault status */ + WREG32_P(hub->vm_l2_pro_fault_cntl, 1, ~1); + + return REG_GET_FIELD(status, VM_L2_PROTECTION_FAULT_STATUS, FED); +} + +struct amdgpu_ras_block_hw_ops gfx_v9_4_2_ras_ops = { + .ras_error_inject = &gfx_v9_4_2_ras_error_inject, + .query_ras_error_count = &gfx_v9_4_2_query_ras_error_count, + .reset_ras_error_count = &gfx_v9_4_2_reset_ras_error_count, + .query_ras_error_status = &gfx_v9_4_2_query_ras_error_status, + .reset_ras_error_status = &gfx_v9_4_2_reset_ras_error_status, +}; + +struct amdgpu_gfx_ras gfx_v9_4_2_ras = { + .ras_block = { + .hw_ops = &gfx_v9_4_2_ras_ops, + }, .enable_watchdog_timer = &gfx_v9_4_2_enable_watchdog_timer, + .query_utcl2_poison_status = gfx_v9_4_2_query_uctl2_poison_status, }; diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_2.h b/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_2.h index 6db1f88509af..7584624b641c 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_2.h +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_2.h @@ -31,6 +31,6 @@ void gfx_v9_4_2_init_golden_registers(struct amdgpu_device *adev, void gfx_v9_4_2_set_power_brake_sequence(struct amdgpu_device *adev); int gfx_v9_4_2_do_edc_gpr_workarounds(struct amdgpu_device *adev); -extern const struct amdgpu_gfx_ras_funcs gfx_v9_4_2_ras_funcs; +extern struct amdgpu_gfx_ras gfx_v9_4_2_ras; #endif /* __GFX_V9_4_2_H__ */ diff --git a/drivers/gpu/drm/amd/amdgpu/gfxhub_v2_0.c b/drivers/gpu/drm/amd/amdgpu/gfxhub_v2_0.c index 6e0ace2fbfab..34513e8e1519 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfxhub_v2_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfxhub_v2_0.c @@ -325,6 +325,8 @@ static void gfxhub_v2_0_setup_vmid_config(struct amdgpu_device *adev) i * hub->ctx_addr_distance, upper_32_bits(adev->vm_manager.max_pfn - 1)); } + + hub->vm_cntx_cntl = tmp; } static void gfxhub_v2_0_program_invalidation(struct amdgpu_device *adev) diff --git a/drivers/gpu/drm/amd/amdgpu/gfxhub_v2_1.c b/drivers/gpu/drm/amd/amdgpu/gfxhub_v2_1.c index b4eddf6e98a6..8cf53e039c11 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfxhub_v2_1.c +++ b/drivers/gpu/drm/amd/amdgpu/gfxhub_v2_1.c @@ -334,6 +334,8 @@ static void gfxhub_v2_1_setup_vmid_config(struct amdgpu_device *adev) i * hub->ctx_addr_distance, upper_32_bits(adev->vm_manager.max_pfn - 1)); } + + hub->vm_cntx_cntl = tmp; } static void gfxhub_v2_1_program_invalidation(struct amdgpu_device *adev) @@ -543,7 +545,9 @@ static void gfxhub_v2_1_utcl2_harvest(struct amdgpu_device *adev) adev->gfx.config.max_sh_per_se * adev->gfx.config.max_shader_engines); - if (adev->ip_versions[GC_HWIP][0] == IP_VERSION(10, 3, 3)) { + switch (adev->ip_versions[GC_HWIP][0]) { + case IP_VERSION(10, 3, 1): + case IP_VERSION(10, 3, 3): /* Get SA disabled bitmap from eFuse setting */ efuse_setting = RREG32_SOC15(GC, 0, mmCC_GC_SA_UNIT_DISABLE); efuse_setting &= CC_GC_SA_UNIT_DISABLE__SA_DISABLE_MASK; @@ -566,6 +570,114 @@ static void gfxhub_v2_1_utcl2_harvest(struct amdgpu_device *adev) disabled_sa = tmp; WREG32_SOC15(GC, 0, mmGCUTCL2_HARVEST_BYPASS_GROUPS_YELLOW_CARP, disabled_sa); + break; + default: + break; + } +} + +static void gfxhub_v2_1_save_regs(struct amdgpu_device *adev) +{ + int i; + adev->gmc.VM_L2_CNTL = RREG32_SOC15(GC, 0, mmGCVM_L2_CNTL); + adev->gmc.VM_L2_CNTL2 = RREG32_SOC15(GC, 0, mmGCVM_L2_CNTL2); + adev->gmc.VM_DUMMY_PAGE_FAULT_CNTL = RREG32_SOC15(GC, 0, mmGCVM_DUMMY_PAGE_FAULT_CNTL); + adev->gmc.VM_DUMMY_PAGE_FAULT_ADDR_LO32 = RREG32_SOC15(GC, 0, mmGCVM_DUMMY_PAGE_FAULT_ADDR_LO32); + adev->gmc.VM_DUMMY_PAGE_FAULT_ADDR_HI32 = RREG32_SOC15(GC, 0, mmGCVM_DUMMY_PAGE_FAULT_ADDR_HI32); + adev->gmc.VM_L2_PROTECTION_FAULT_CNTL = RREG32_SOC15(GC, 0, mmGCVM_L2_PROTECTION_FAULT_CNTL); + adev->gmc.VM_L2_PROTECTION_FAULT_CNTL2 = RREG32_SOC15(GC, 0, mmGCVM_L2_PROTECTION_FAULT_CNTL2); + adev->gmc.VM_L2_PROTECTION_FAULT_MM_CNTL3 = RREG32_SOC15(GC, 0, mmGCVM_L2_PROTECTION_FAULT_MM_CNTL3); + adev->gmc.VM_L2_PROTECTION_FAULT_MM_CNTL4 = RREG32_SOC15(GC, 0, mmGCVM_L2_PROTECTION_FAULT_MM_CNTL4); + adev->gmc.VM_L2_PROTECTION_FAULT_ADDR_LO32 = RREG32_SOC15(GC, 0, mmGCVM_L2_PROTECTION_FAULT_ADDR_LO32); + adev->gmc.VM_L2_PROTECTION_FAULT_ADDR_HI32 = RREG32_SOC15(GC, 0, mmGCVM_L2_PROTECTION_FAULT_ADDR_HI32); + adev->gmc.VM_DEBUG = RREG32_SOC15(GC, 0, mmGCVM_DEBUG); + adev->gmc.VM_L2_MM_GROUP_RT_CLASSES = RREG32_SOC15(GC, 0, mmGCVM_L2_MM_GROUP_RT_CLASSES); + adev->gmc.VM_L2_BANK_SELECT_RESERVED_CID = RREG32_SOC15(GC, 0, mmGCVM_L2_BANK_SELECT_RESERVED_CID); + adev->gmc.VM_L2_BANK_SELECT_RESERVED_CID2 = RREG32_SOC15(GC, 0, mmGCVM_L2_BANK_SELECT_RESERVED_CID2); + adev->gmc.VM_L2_CACHE_PARITY_CNTL = RREG32_SOC15(GC, 0, mmGCVM_L2_CACHE_PARITY_CNTL); + adev->gmc.VM_L2_IH_LOG_CNTL = RREG32_SOC15(GC, 0, mmGCVM_L2_IH_LOG_CNTL); + + for (i = 0; i <= 15; i++) { + adev->gmc.VM_CONTEXT_CNTL[i] = RREG32_SOC15_OFFSET(GC, 0, mmGCVM_CONTEXT0_CNTL, i); + adev->gmc.VM_CONTEXT_PAGE_TABLE_BASE_ADDR_LO32[i] = RREG32_SOC15_OFFSET(GC, 0, mmGCVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32, i * 2); + adev->gmc.VM_CONTEXT_PAGE_TABLE_BASE_ADDR_HI32[i] = RREG32_SOC15_OFFSET(GC, 0, mmGCVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_HI32, i * 2); + adev->gmc.VM_CONTEXT_PAGE_TABLE_START_ADDR_LO32[i] = RREG32_SOC15_OFFSET(GC, 0, mmGCVM_CONTEXT0_PAGE_TABLE_START_ADDR_LO32, i * 2); + adev->gmc.VM_CONTEXT_PAGE_TABLE_START_ADDR_HI32[i] = RREG32_SOC15_OFFSET(GC, 0, mmGCVM_CONTEXT0_PAGE_TABLE_START_ADDR_HI32, i * 2); + adev->gmc.VM_CONTEXT_PAGE_TABLE_END_ADDR_LO32[i] = RREG32_SOC15_OFFSET(GC, 0, mmGCVM_CONTEXT0_PAGE_TABLE_END_ADDR_LO32, i * 2); + adev->gmc.VM_CONTEXT_PAGE_TABLE_END_ADDR_HI32[i] = RREG32_SOC15_OFFSET(GC, 0, mmGCVM_CONTEXT0_PAGE_TABLE_END_ADDR_HI32, i * 2); + } + + adev->gmc.MC_VM_MX_L1_TLB_CNTL = RREG32_SOC15(GC, 0, mmGCMC_VM_MX_L1_TLB_CNTL); +} + +static void gfxhub_v2_1_restore_regs(struct amdgpu_device *adev) +{ + int i; + WREG32_SOC15(GC, 0, mmGCVM_L2_CNTL, adev->gmc.VM_L2_CNTL); + WREG32_SOC15(GC, 0, mmGCVM_L2_CNTL2, adev->gmc.VM_L2_CNTL2); + WREG32_SOC15(GC, 0, mmGCVM_DUMMY_PAGE_FAULT_CNTL, adev->gmc.VM_DUMMY_PAGE_FAULT_CNTL); + WREG32_SOC15(GC, 0, mmGCVM_DUMMY_PAGE_FAULT_ADDR_LO32, adev->gmc.VM_DUMMY_PAGE_FAULT_ADDR_LO32); + WREG32_SOC15(GC, 0, mmGCVM_DUMMY_PAGE_FAULT_ADDR_HI32, adev->gmc.VM_DUMMY_PAGE_FAULT_ADDR_HI32); + WREG32_SOC15(GC, 0, mmGCVM_L2_PROTECTION_FAULT_CNTL, adev->gmc.VM_L2_PROTECTION_FAULT_CNTL); + WREG32_SOC15(GC, 0, mmGCVM_L2_PROTECTION_FAULT_CNTL2, adev->gmc.VM_L2_PROTECTION_FAULT_CNTL2); + WREG32_SOC15(GC, 0, mmGCVM_L2_PROTECTION_FAULT_MM_CNTL3, adev->gmc.VM_L2_PROTECTION_FAULT_MM_CNTL3); + WREG32_SOC15(GC, 0, mmGCVM_L2_PROTECTION_FAULT_MM_CNTL4, adev->gmc.VM_L2_PROTECTION_FAULT_MM_CNTL4); + WREG32_SOC15(GC, 0, mmGCVM_L2_PROTECTION_FAULT_ADDR_LO32, adev->gmc.VM_L2_PROTECTION_FAULT_ADDR_LO32); + WREG32_SOC15(GC, 0, mmGCVM_L2_PROTECTION_FAULT_ADDR_HI32, adev->gmc.VM_L2_PROTECTION_FAULT_ADDR_HI32); + WREG32_SOC15(GC, 0, mmGCVM_DEBUG, adev->gmc.VM_DEBUG); + WREG32_SOC15(GC, 0, mmGCVM_L2_MM_GROUP_RT_CLASSES, adev->gmc.VM_L2_MM_GROUP_RT_CLASSES); + WREG32_SOC15(GC, 0, mmGCVM_L2_BANK_SELECT_RESERVED_CID, adev->gmc.VM_L2_BANK_SELECT_RESERVED_CID); + WREG32_SOC15(GC, 0, mmGCVM_L2_BANK_SELECT_RESERVED_CID2, adev->gmc.VM_L2_BANK_SELECT_RESERVED_CID2); + WREG32_SOC15(GC, 0, mmGCVM_L2_CACHE_PARITY_CNTL, adev->gmc.VM_L2_CACHE_PARITY_CNTL); + WREG32_SOC15(GC, 0, mmGCVM_L2_IH_LOG_CNTL, adev->gmc.VM_L2_IH_LOG_CNTL); + + for (i = 0; i <= 15; i++) { + WREG32_SOC15_OFFSET(GC, 0, mmGCVM_CONTEXT0_CNTL, i, adev->gmc.VM_CONTEXT_CNTL[i]); + WREG32_SOC15_OFFSET(GC, 0, mmGCVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32, i * 2, adev->gmc.VM_CONTEXT_PAGE_TABLE_BASE_ADDR_LO32[i]); + WREG32_SOC15_OFFSET(GC, 0, mmGCVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_HI32, i * 2, adev->gmc.VM_CONTEXT_PAGE_TABLE_BASE_ADDR_HI32[i]); + WREG32_SOC15_OFFSET(GC, 0, mmGCVM_CONTEXT0_PAGE_TABLE_START_ADDR_LO32, i * 2, adev->gmc.VM_CONTEXT_PAGE_TABLE_START_ADDR_LO32[i]); + WREG32_SOC15_OFFSET(GC, 0, mmGCVM_CONTEXT0_PAGE_TABLE_START_ADDR_HI32, i * 2, adev->gmc.VM_CONTEXT_PAGE_TABLE_START_ADDR_HI32[i]); + WREG32_SOC15_OFFSET(GC, 0, mmGCVM_CONTEXT0_PAGE_TABLE_END_ADDR_LO32, i * 2, adev->gmc.VM_CONTEXT_PAGE_TABLE_END_ADDR_LO32[i]); + WREG32_SOC15_OFFSET(GC, 0, mmGCVM_CONTEXT0_PAGE_TABLE_END_ADDR_HI32, i * 2, adev->gmc.VM_CONTEXT_PAGE_TABLE_END_ADDR_HI32[i]); + } + + WREG32_SOC15(GC, 0, mmGCMC_VM_FB_LOCATION_BASE, adev->gmc.vram_start >> 24); + WREG32_SOC15(GC, 0, mmGCMC_VM_FB_LOCATION_TOP, adev->gmc.vram_end >> 24); + WREG32_SOC15(GC, 0, mmGCMC_VM_MX_L1_TLB_CNTL, adev->gmc.MC_VM_MX_L1_TLB_CNTL); +} + +static void gfxhub_v2_1_halt(struct amdgpu_device *adev) +{ + struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0]; + int i; + uint32_t tmp; + int time = 1000; + + gfxhub_v2_1_set_fault_enable_default(adev, false); + + for (i = 0; i <= 14; i++) { + WREG32_SOC15_OFFSET(GC, 0, mmGCVM_CONTEXT1_PAGE_TABLE_START_ADDR_LO32, + i * hub->ctx_addr_distance, ~0); + WREG32_SOC15_OFFSET(GC, 0, mmGCVM_CONTEXT1_PAGE_TABLE_START_ADDR_HI32, + i * hub->ctx_addr_distance, ~0); + WREG32_SOC15_OFFSET(GC, 0, mmGCVM_CONTEXT1_PAGE_TABLE_END_ADDR_LO32, + i * hub->ctx_addr_distance, + 0); + WREG32_SOC15_OFFSET(GC, 0, mmGCVM_CONTEXT1_PAGE_TABLE_END_ADDR_HI32, + i * hub->ctx_addr_distance, + 0); + } + tmp = RREG32_SOC15(GC, 0, mmGRBM_STATUS2); + while ((tmp & (GRBM_STATUS2__EA_BUSY_MASK | + GRBM_STATUS2__EA_LINK_BUSY_MASK)) != 0 && + time) { + udelay(100); + time--; + tmp = RREG32_SOC15(GC, 0, mmGRBM_STATUS2); + } + + if (!time) { + DRM_WARN("failed to wait for GRBM(EA) idle\n"); } } @@ -579,4 +691,7 @@ const struct amdgpu_gfxhub_funcs gfxhub_v2_1_funcs = { .init = gfxhub_v2_1_init, .get_xgmi_info = gfxhub_v2_1_get_xgmi_info, .utcl2_harvest = gfxhub_v2_1_utcl2_harvest, + .mode2_save_regs = gfxhub_v2_1_save_regs, + .mode2_restore_regs = gfxhub_v2_1_restore_regs, + .halt = gfxhub_v2_1_halt, }; diff --git a/drivers/gpu/drm/amd/amdgpu/gfxhub_v3_0.c b/drivers/gpu/drm/amd/amdgpu/gfxhub_v3_0.c new file mode 100644 index 000000000000..0e13370c2057 --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/gfxhub_v3_0.c @@ -0,0 +1,535 @@ +/* + * Copyright 2021 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#include "amdgpu.h" +#include "gfxhub_v3_0.h" + +#include "gc/gc_11_0_0_offset.h" +#include "gc/gc_11_0_0_sh_mask.h" +#include "gc/gc_11_0_0_default.h" +#include "navi10_enum.h" +#include "soc15_common.h" + +static const char *gfxhub_client_ids[] = { + "CB/DB", + "Reserved", + "GE1", + "GE2", + "CPF", + "CPC", + "CPG", + "RLC", + "TCP", + "SQC (inst)", + "SQC (data)", + "SQG", + "Reserved", + "SDMA0", + "SDMA1", + "GCR", + "SDMA2", + "SDMA3", +}; + +static uint32_t gfxhub_v3_0_get_invalidate_req(unsigned int vmid, + uint32_t flush_type) +{ + u32 req = 0; + + /* invalidate using legacy mode on vmid*/ + req = REG_SET_FIELD(req, GCVM_INVALIDATE_ENG0_REQ, + PER_VMID_INVALIDATE_REQ, 1 << vmid); + req = REG_SET_FIELD(req, GCVM_INVALIDATE_ENG0_REQ, FLUSH_TYPE, flush_type); + req = REG_SET_FIELD(req, GCVM_INVALIDATE_ENG0_REQ, INVALIDATE_L2_PTES, 1); + req = REG_SET_FIELD(req, GCVM_INVALIDATE_ENG0_REQ, INVALIDATE_L2_PDE0, 1); + req = REG_SET_FIELD(req, GCVM_INVALIDATE_ENG0_REQ, INVALIDATE_L2_PDE1, 1); + req = REG_SET_FIELD(req, GCVM_INVALIDATE_ENG0_REQ, INVALIDATE_L2_PDE2, 1); + req = REG_SET_FIELD(req, GCVM_INVALIDATE_ENG0_REQ, INVALIDATE_L1_PTES, 1); + req = REG_SET_FIELD(req, GCVM_INVALIDATE_ENG0_REQ, + CLEAR_PROTECTION_FAULT_STATUS_ADDR, 0); + + return req; +} + +static void +gfxhub_v3_0_print_l2_protection_fault_status(struct amdgpu_device *adev, + uint32_t status) +{ + u32 cid = REG_GET_FIELD(status, + GCVM_L2_PROTECTION_FAULT_STATUS, CID); + + dev_err(adev->dev, + "GCVM_L2_PROTECTION_FAULT_STATUS:0x%08X\n", + status); + dev_err(adev->dev, "\t Faulty UTCL2 client ID: %s (0x%x)\n", + cid >= ARRAY_SIZE(gfxhub_client_ids) ? "unknown" : gfxhub_client_ids[cid], + cid); + dev_err(adev->dev, "\t MORE_FAULTS: 0x%lx\n", + REG_GET_FIELD(status, + GCVM_L2_PROTECTION_FAULT_STATUS, MORE_FAULTS)); + dev_err(adev->dev, "\t WALKER_ERROR: 0x%lx\n", + REG_GET_FIELD(status, + GCVM_L2_PROTECTION_FAULT_STATUS, WALKER_ERROR)); + dev_err(adev->dev, "\t PERMISSION_FAULTS: 0x%lx\n", + REG_GET_FIELD(status, + GCVM_L2_PROTECTION_FAULT_STATUS, PERMISSION_FAULTS)); + dev_err(adev->dev, "\t MAPPING_ERROR: 0x%lx\n", + REG_GET_FIELD(status, + GCVM_L2_PROTECTION_FAULT_STATUS, MAPPING_ERROR)); + dev_err(adev->dev, "\t RW: 0x%lx\n", + REG_GET_FIELD(status, + GCVM_L2_PROTECTION_FAULT_STATUS, RW)); +} + +static u64 gfxhub_v3_0_get_fb_location(struct amdgpu_device *adev) +{ + u64 base = RREG32_SOC15(GC, 0, regGCMC_VM_FB_LOCATION_BASE); + + base &= GCMC_VM_FB_LOCATION_BASE__FB_BASE_MASK; + base <<= 24; + + return base; +} + +static u64 gfxhub_v3_0_get_mc_fb_offset(struct amdgpu_device *adev) +{ + return (u64)RREG32_SOC15(GC, 0, regGCMC_VM_FB_OFFSET) << 24; +} + +static void gfxhub_v3_0_setup_vm_pt_regs(struct amdgpu_device *adev, uint32_t vmid, + uint64_t page_table_base) +{ + struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0]; + + WREG32_SOC15_OFFSET(GC, 0, regGCVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32, + hub->ctx_addr_distance * vmid, + lower_32_bits(page_table_base)); + + WREG32_SOC15_OFFSET(GC, 0, regGCVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_HI32, + hub->ctx_addr_distance * vmid, + upper_32_bits(page_table_base)); +} + +static void gfxhub_v3_0_init_gart_aperture_regs(struct amdgpu_device *adev) +{ + uint64_t pt_base = amdgpu_gmc_pd_addr(adev->gart.bo); + + gfxhub_v3_0_setup_vm_pt_regs(adev, 0, pt_base); + + WREG32_SOC15(GC, 0, regGCVM_CONTEXT0_PAGE_TABLE_START_ADDR_LO32, + (u32)(adev->gmc.gart_start >> 12)); + WREG32_SOC15(GC, 0, regGCVM_CONTEXT0_PAGE_TABLE_START_ADDR_HI32, + (u32)(adev->gmc.gart_start >> 44)); + + WREG32_SOC15(GC, 0, regGCVM_CONTEXT0_PAGE_TABLE_END_ADDR_LO32, + (u32)(adev->gmc.gart_end >> 12)); + WREG32_SOC15(GC, 0, regGCVM_CONTEXT0_PAGE_TABLE_END_ADDR_HI32, + (u32)(adev->gmc.gart_end >> 44)); +} + +static void gfxhub_v3_0_init_system_aperture_regs(struct amdgpu_device *adev) +{ + uint64_t value; + + /* Disable AGP. */ + WREG32_SOC15(GC, 0, regGCMC_VM_AGP_BASE, 0); + WREG32_SOC15(GC, 0, regGCMC_VM_AGP_TOP, 0); + WREG32_SOC15(GC, 0, regGCMC_VM_AGP_BOT, 0x00FFFFFF); + + /* Program the system aperture low logical page number. */ + WREG32_SOC15(GC, 0, regGCMC_VM_SYSTEM_APERTURE_LOW_ADDR, + adev->gmc.vram_start >> 18); + WREG32_SOC15(GC, 0, regGCMC_VM_SYSTEM_APERTURE_HIGH_ADDR, + adev->gmc.vram_end >> 18); + + /* Set default page address. */ + value = adev->vram_scratch.gpu_addr - adev->gmc.vram_start + + adev->vm_manager.vram_base_offset; + WREG32_SOC15(GC, 0, regGCMC_VM_SYSTEM_APERTURE_DEFAULT_ADDR_LSB, + (u32)(value >> 12)); + WREG32_SOC15(GC, 0, regGCMC_VM_SYSTEM_APERTURE_DEFAULT_ADDR_MSB, + (u32)(value >> 44)); + + /* Program "protection fault". */ + WREG32_SOC15(GC, 0, regGCVM_L2_PROTECTION_FAULT_DEFAULT_ADDR_LO32, + (u32)(adev->dummy_page_addr >> 12)); + WREG32_SOC15(GC, 0, regGCVM_L2_PROTECTION_FAULT_DEFAULT_ADDR_HI32, + (u32)((u64)adev->dummy_page_addr >> 44)); + + WREG32_FIELD15_PREREG(GC, 0, GCVM_L2_PROTECTION_FAULT_CNTL2, + ACTIVE_PAGE_MIGRATION_PTE_READ_RETRY, 1); +} + + +static void gfxhub_v3_0_init_tlb_regs(struct amdgpu_device *adev) +{ + uint32_t tmp; + + /* Setup TLB control */ + tmp = RREG32_SOC15(GC, 0, regGCMC_VM_MX_L1_TLB_CNTL); + + tmp = REG_SET_FIELD(tmp, GCMC_VM_MX_L1_TLB_CNTL, ENABLE_L1_TLB, 1); + tmp = REG_SET_FIELD(tmp, GCMC_VM_MX_L1_TLB_CNTL, SYSTEM_ACCESS_MODE, 3); + tmp = REG_SET_FIELD(tmp, GCMC_VM_MX_L1_TLB_CNTL, + ENABLE_ADVANCED_DRIVER_MODEL, 1); + tmp = REG_SET_FIELD(tmp, GCMC_VM_MX_L1_TLB_CNTL, + SYSTEM_APERTURE_UNMAPPED_ACCESS, 0); + tmp = REG_SET_FIELD(tmp, GCMC_VM_MX_L1_TLB_CNTL, ECO_BITS, 0); + tmp = REG_SET_FIELD(tmp, GCMC_VM_MX_L1_TLB_CNTL, + MTYPE, MTYPE_UC); /* UC, uncached */ + + WREG32_SOC15(GC, 0, regGCMC_VM_MX_L1_TLB_CNTL, tmp); +} + +static void gfxhub_v3_0_init_cache_regs(struct amdgpu_device *adev) +{ + uint32_t tmp; + + /* These registers are not accessible to VF-SRIOV. + * The PF will program them instead. + */ + if (amdgpu_sriov_vf(adev)) + return; + + /* Setup L2 cache */ + tmp = RREG32_SOC15(GC, 0, regGCVM_L2_CNTL); + tmp = REG_SET_FIELD(tmp, GCVM_L2_CNTL, ENABLE_L2_CACHE, 1); + tmp = REG_SET_FIELD(tmp, GCVM_L2_CNTL, ENABLE_L2_FRAGMENT_PROCESSING, 0); + tmp = REG_SET_FIELD(tmp, GCVM_L2_CNTL, + ENABLE_DEFAULT_PAGE_OUT_TO_SYSTEM_MEMORY, 1); + /* XXX for emulation, Refer to closed source code.*/ + tmp = REG_SET_FIELD(tmp, GCVM_L2_CNTL, + L2_PDE0_CACHE_TAG_GENERATION_MODE, 0); + tmp = REG_SET_FIELD(tmp, GCVM_L2_CNTL, PDE_FAULT_CLASSIFICATION, 0); + tmp = REG_SET_FIELD(tmp, GCVM_L2_CNTL, CONTEXT1_IDENTITY_ACCESS_MODE, 1); + tmp = REG_SET_FIELD(tmp, GCVM_L2_CNTL, IDENTITY_MODE_FRAGMENT_SIZE, 0); + WREG32_SOC15(GC, 0, regGCVM_L2_CNTL, tmp); + + tmp = RREG32_SOC15(GC, 0, regGCVM_L2_CNTL2); + tmp = REG_SET_FIELD(tmp, GCVM_L2_CNTL2, INVALIDATE_ALL_L1_TLBS, 1); + tmp = REG_SET_FIELD(tmp, GCVM_L2_CNTL2, INVALIDATE_L2_CACHE, 1); + WREG32_SOC15(GC, 0, regGCVM_L2_CNTL2, tmp); + + tmp = regGCVM_L2_CNTL3_DEFAULT; + if (adev->gmc.translate_further) { + tmp = REG_SET_FIELD(tmp, GCVM_L2_CNTL3, BANK_SELECT, 12); + tmp = REG_SET_FIELD(tmp, GCVM_L2_CNTL3, + L2_CACHE_BIGK_FRAGMENT_SIZE, 9); + } else { + tmp = REG_SET_FIELD(tmp, GCVM_L2_CNTL3, BANK_SELECT, 9); + tmp = REG_SET_FIELD(tmp, GCVM_L2_CNTL3, + L2_CACHE_BIGK_FRAGMENT_SIZE, 6); + } + WREG32_SOC15(GC, 0, regGCVM_L2_CNTL3, tmp); + + tmp = regGCVM_L2_CNTL4_DEFAULT; + tmp = REG_SET_FIELD(tmp, GCVM_L2_CNTL4, VMC_TAP_PDE_REQUEST_PHYSICAL, 0); + tmp = REG_SET_FIELD(tmp, GCVM_L2_CNTL4, VMC_TAP_PTE_REQUEST_PHYSICAL, 0); + WREG32_SOC15(GC, 0, regGCVM_L2_CNTL4, tmp); + + tmp = regGCVM_L2_CNTL5_DEFAULT; + tmp = REG_SET_FIELD(tmp, GCVM_L2_CNTL5, L2_CACHE_SMALLK_FRAGMENT_SIZE, 0); + WREG32_SOC15(GC, 0, regGCVM_L2_CNTL5, tmp); +} + +static void gfxhub_v3_0_enable_system_domain(struct amdgpu_device *adev) +{ + uint32_t tmp; + + tmp = RREG32_SOC15(GC, 0, regGCVM_CONTEXT0_CNTL); + tmp = REG_SET_FIELD(tmp, GCVM_CONTEXT0_CNTL, ENABLE_CONTEXT, 1); + tmp = REG_SET_FIELD(tmp, GCVM_CONTEXT0_CNTL, PAGE_TABLE_DEPTH, 0); + tmp = REG_SET_FIELD(tmp, GCVM_CONTEXT0_CNTL, + RETRY_PERMISSION_OR_INVALID_PAGE_FAULT, 0); + WREG32_SOC15(GC, 0, regGCVM_CONTEXT0_CNTL, tmp); +} + +static void gfxhub_v3_0_disable_identity_aperture(struct amdgpu_device *adev) +{ + /* These registers are not accessible to VF-SRIOV. + * The PF will program them instead. + */ + if (amdgpu_sriov_vf(adev)) + return; + + WREG32_SOC15(GC, 0, regGCVM_L2_CONTEXT1_IDENTITY_APERTURE_LOW_ADDR_LO32, + 0xFFFFFFFF); + WREG32_SOC15(GC, 0, regGCVM_L2_CONTEXT1_IDENTITY_APERTURE_LOW_ADDR_HI32, + 0x0000000F); + + WREG32_SOC15(GC, 0, regGCVM_L2_CONTEXT1_IDENTITY_APERTURE_HIGH_ADDR_LO32, + 0); + WREG32_SOC15(GC, 0, regGCVM_L2_CONTEXT1_IDENTITY_APERTURE_HIGH_ADDR_HI32, + 0); + + WREG32_SOC15(GC, 0, regGCVM_L2_CONTEXT_IDENTITY_PHYSICAL_OFFSET_LO32, 0); + WREG32_SOC15(GC, 0, regGCVM_L2_CONTEXT_IDENTITY_PHYSICAL_OFFSET_HI32, 0); + +} + +static void gfxhub_v3_0_setup_vmid_config(struct amdgpu_device *adev) +{ + struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0]; + int i; + uint32_t tmp; + + for (i = 0; i <= 14; i++) { + tmp = RREG32_SOC15_OFFSET(GC, 0, regGCVM_CONTEXT1_CNTL, i); + tmp = REG_SET_FIELD(tmp, GCVM_CONTEXT1_CNTL, ENABLE_CONTEXT, 1); + tmp = REG_SET_FIELD(tmp, GCVM_CONTEXT1_CNTL, PAGE_TABLE_DEPTH, + adev->vm_manager.num_level); + tmp = REG_SET_FIELD(tmp, GCVM_CONTEXT1_CNTL, + RANGE_PROTECTION_FAULT_ENABLE_DEFAULT, 1); + tmp = REG_SET_FIELD(tmp, GCVM_CONTEXT1_CNTL, + DUMMY_PAGE_PROTECTION_FAULT_ENABLE_DEFAULT, 1); + tmp = REG_SET_FIELD(tmp, GCVM_CONTEXT1_CNTL, + PDE0_PROTECTION_FAULT_ENABLE_DEFAULT, 1); + tmp = REG_SET_FIELD(tmp, GCVM_CONTEXT1_CNTL, + VALID_PROTECTION_FAULT_ENABLE_DEFAULT, 1); + tmp = REG_SET_FIELD(tmp, GCVM_CONTEXT1_CNTL, + READ_PROTECTION_FAULT_ENABLE_DEFAULT, 1); + tmp = REG_SET_FIELD(tmp, GCVM_CONTEXT1_CNTL, + WRITE_PROTECTION_FAULT_ENABLE_DEFAULT, 1); + tmp = REG_SET_FIELD(tmp, GCVM_CONTEXT1_CNTL, + EXECUTE_PROTECTION_FAULT_ENABLE_DEFAULT, 1); + tmp = REG_SET_FIELD(tmp, GCVM_CONTEXT1_CNTL, + PAGE_TABLE_BLOCK_SIZE, + adev->vm_manager.block_size - 9); + /* Send no-retry XNACK on fault to suppress VM fault storm. */ + tmp = REG_SET_FIELD(tmp, GCVM_CONTEXT1_CNTL, + RETRY_PERMISSION_OR_INVALID_PAGE_FAULT, + !amdgpu_noretry); + WREG32_SOC15_OFFSET(GC, 0, regGCVM_CONTEXT1_CNTL, + i * hub->ctx_distance, tmp); + WREG32_SOC15_OFFSET(GC, 0, regGCVM_CONTEXT1_PAGE_TABLE_START_ADDR_LO32, + i * hub->ctx_addr_distance, 0); + WREG32_SOC15_OFFSET(GC, 0, regGCVM_CONTEXT1_PAGE_TABLE_START_ADDR_HI32, + i * hub->ctx_addr_distance, 0); + WREG32_SOC15_OFFSET(GC, 0, regGCVM_CONTEXT1_PAGE_TABLE_END_ADDR_LO32, + i * hub->ctx_addr_distance, + lower_32_bits(adev->vm_manager.max_pfn - 1)); + WREG32_SOC15_OFFSET(GC, 0, regGCVM_CONTEXT1_PAGE_TABLE_END_ADDR_HI32, + i * hub->ctx_addr_distance, + upper_32_bits(adev->vm_manager.max_pfn - 1)); + } + + hub->vm_cntx_cntl = tmp; +} + +static void gfxhub_v3_0_program_invalidation(struct amdgpu_device *adev) +{ + struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0]; + unsigned i; + + for (i = 0 ; i < 18; ++i) { + WREG32_SOC15_OFFSET(GC, 0, regGCVM_INVALIDATE_ENG0_ADDR_RANGE_LO32, + i * hub->eng_addr_distance, 0xffffffff); + WREG32_SOC15_OFFSET(GC, 0, regGCVM_INVALIDATE_ENG0_ADDR_RANGE_HI32, + i * hub->eng_addr_distance, 0x1f); + } +} + +static int gfxhub_v3_0_gart_enable(struct amdgpu_device *adev) +{ + if (amdgpu_sriov_vf(adev)) { + /* + * GCMC_VM_FB_LOCATION_BASE/TOP is NULL for VF, becuase they are + * VF copy registers so vbios post doesn't program them, for + * SRIOV driver need to program them + */ + WREG32_SOC15(GC, 0, regGCMC_VM_FB_LOCATION_BASE, + adev->gmc.vram_start >> 24); + WREG32_SOC15(GC, 0, regGCMC_VM_FB_LOCATION_TOP, + adev->gmc.vram_end >> 24); + } + + /* GART Enable. */ + gfxhub_v3_0_init_gart_aperture_regs(adev); + gfxhub_v3_0_init_system_aperture_regs(adev); + gfxhub_v3_0_init_tlb_regs(adev); + gfxhub_v3_0_init_cache_regs(adev); + + gfxhub_v3_0_enable_system_domain(adev); + gfxhub_v3_0_disable_identity_aperture(adev); + gfxhub_v3_0_setup_vmid_config(adev); + gfxhub_v3_0_program_invalidation(adev); + + return 0; +} + +static void gfxhub_v3_0_gart_disable(struct amdgpu_device *adev) +{ + struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0]; + u32 tmp; + u32 i; + + /* Disable all tables */ + for (i = 0; i < 16; i++) + WREG32_SOC15_OFFSET(GC, 0, regGCVM_CONTEXT0_CNTL, + i * hub->ctx_distance, 0); + + /* Setup TLB control */ + tmp = RREG32_SOC15(GC, 0, regGCMC_VM_MX_L1_TLB_CNTL); + tmp = REG_SET_FIELD(tmp, GCMC_VM_MX_L1_TLB_CNTL, ENABLE_L1_TLB, 0); + tmp = REG_SET_FIELD(tmp, GCMC_VM_MX_L1_TLB_CNTL, + ENABLE_ADVANCED_DRIVER_MODEL, 0); + WREG32_SOC15(GC, 0, regGCMC_VM_MX_L1_TLB_CNTL, tmp); + + /* Setup L2 cache */ + WREG32_FIELD15_PREREG(GC, 0, GCVM_L2_CNTL, ENABLE_L2_CACHE, 0); + WREG32_SOC15(GC, 0, regGCVM_L2_CNTL3, 0); +} + +/** + * gfxhub_v3_0_set_fault_enable_default - update GART/VM fault handling + * + * @adev: amdgpu_device pointer + * @value: true redirects VM faults to the default page + */ +static void gfxhub_v3_0_set_fault_enable_default(struct amdgpu_device *adev, + bool value) +{ + u32 tmp; + + /* NO halt CP when page fault */ + tmp = RREG32_SOC15(GC, 0, regCP_DEBUG); + tmp = REG_SET_FIELD(tmp, CP_DEBUG, CPG_UTCL1_ERROR_HALT_DISABLE, 1); + WREG32_SOC15(GC, 0, regCP_DEBUG, tmp); + + /** + * Set GRBM_GFX_INDEX in broad cast mode + * before programming GL1C_UTCL0_CNTL1 and SQG_CONFIG + */ + WREG32_SOC15(GC, 0, regGRBM_GFX_INDEX, regGRBM_GFX_INDEX_DEFAULT); + + /** + * Retry respond mode: RETRY + * Error (no retry) respond mode: SUCCESS + */ + tmp = RREG32_SOC15(GC, 0, regGL1C_UTCL0_CNTL1); + tmp = REG_SET_FIELD(tmp, GL1C_UTCL0_CNTL1, RESP_MODE, 0); + tmp = REG_SET_FIELD(tmp, GL1C_UTCL0_CNTL1, RESP_FAULT_MODE, 0x2); + WREG32_SOC15(GC, 0, regGL1C_UTCL0_CNTL1, tmp); + + /* These registers are not accessible to VF-SRIOV. + * The PF will program them instead. + */ + if (amdgpu_sriov_vf(adev)) + return; + + /* Disable SQ XNACK interrupt for all VMIDs */ + tmp = RREG32_SOC15(GC, 0, regSQG_CONFIG); + tmp = REG_SET_FIELD(tmp, SQG_CONFIG, XNACK_INTR_MASK, + SQG_CONFIG__XNACK_INTR_MASK_MASK >> + SQG_CONFIG__XNACK_INTR_MASK__SHIFT); + WREG32_SOC15(GC, 0, regSQG_CONFIG, tmp); + + tmp = RREG32_SOC15(GC, 0, regGCVM_L2_PROTECTION_FAULT_CNTL); + tmp = REG_SET_FIELD(tmp, GCVM_L2_PROTECTION_FAULT_CNTL, + RANGE_PROTECTION_FAULT_ENABLE_DEFAULT, value); + tmp = REG_SET_FIELD(tmp, GCVM_L2_PROTECTION_FAULT_CNTL, + PDE0_PROTECTION_FAULT_ENABLE_DEFAULT, value); + tmp = REG_SET_FIELD(tmp, GCVM_L2_PROTECTION_FAULT_CNTL, + PDE1_PROTECTION_FAULT_ENABLE_DEFAULT, value); + tmp = REG_SET_FIELD(tmp, GCVM_L2_PROTECTION_FAULT_CNTL, + PDE2_PROTECTION_FAULT_ENABLE_DEFAULT, value); + tmp = REG_SET_FIELD(tmp, GCVM_L2_PROTECTION_FAULT_CNTL, + TRANSLATE_FURTHER_PROTECTION_FAULT_ENABLE_DEFAULT, + value); + tmp = REG_SET_FIELD(tmp, GCVM_L2_PROTECTION_FAULT_CNTL, + NACK_PROTECTION_FAULT_ENABLE_DEFAULT, value); + tmp = REG_SET_FIELD(tmp, GCVM_L2_PROTECTION_FAULT_CNTL, + DUMMY_PAGE_PROTECTION_FAULT_ENABLE_DEFAULT, value); + tmp = REG_SET_FIELD(tmp, GCVM_L2_PROTECTION_FAULT_CNTL, + VALID_PROTECTION_FAULT_ENABLE_DEFAULT, value); + tmp = REG_SET_FIELD(tmp, GCVM_L2_PROTECTION_FAULT_CNTL, + READ_PROTECTION_FAULT_ENABLE_DEFAULT, value); + tmp = REG_SET_FIELD(tmp, GCVM_L2_PROTECTION_FAULT_CNTL, + WRITE_PROTECTION_FAULT_ENABLE_DEFAULT, value); + tmp = REG_SET_FIELD(tmp, GCVM_L2_PROTECTION_FAULT_CNTL, + EXECUTE_PROTECTION_FAULT_ENABLE_DEFAULT, value); + if (!value) { + tmp = REG_SET_FIELD(tmp, GCVM_L2_PROTECTION_FAULT_CNTL, + CRASH_ON_NO_RETRY_FAULT, 1); + tmp = REG_SET_FIELD(tmp, GCVM_L2_PROTECTION_FAULT_CNTL, + CRASH_ON_RETRY_FAULT, 1); + } + WREG32_SOC15(GC, 0, regGCVM_L2_PROTECTION_FAULT_CNTL, tmp); +} + +static const struct amdgpu_vmhub_funcs gfxhub_v3_0_vmhub_funcs = { + .print_l2_protection_fault_status = gfxhub_v3_0_print_l2_protection_fault_status, + .get_invalidate_req = gfxhub_v3_0_get_invalidate_req, +}; + +static void gfxhub_v3_0_init(struct amdgpu_device *adev) +{ + struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0]; + + hub->ctx0_ptb_addr_lo32 = + SOC15_REG_OFFSET(GC, 0, + regGCVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32); + hub->ctx0_ptb_addr_hi32 = + SOC15_REG_OFFSET(GC, 0, + regGCVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_HI32); + hub->vm_inv_eng0_sem = + SOC15_REG_OFFSET(GC, 0, regGCVM_INVALIDATE_ENG0_SEM); + hub->vm_inv_eng0_req = + SOC15_REG_OFFSET(GC, 0, regGCVM_INVALIDATE_ENG0_REQ); + hub->vm_inv_eng0_ack = + SOC15_REG_OFFSET(GC, 0, regGCVM_INVALIDATE_ENG0_ACK); + hub->vm_context0_cntl = + SOC15_REG_OFFSET(GC, 0, regGCVM_CONTEXT0_CNTL); + hub->vm_l2_pro_fault_status = + SOC15_REG_OFFSET(GC, 0, regGCVM_L2_PROTECTION_FAULT_STATUS); + hub->vm_l2_pro_fault_cntl = + SOC15_REG_OFFSET(GC, 0, regGCVM_L2_PROTECTION_FAULT_CNTL); + + hub->ctx_distance = regGCVM_CONTEXT1_CNTL - regGCVM_CONTEXT0_CNTL; + hub->ctx_addr_distance = regGCVM_CONTEXT1_PAGE_TABLE_BASE_ADDR_LO32 - + regGCVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32; + hub->eng_distance = regGCVM_INVALIDATE_ENG1_REQ - + regGCVM_INVALIDATE_ENG0_REQ; + hub->eng_addr_distance = regGCVM_INVALIDATE_ENG1_ADDR_RANGE_LO32 - + regGCVM_INVALIDATE_ENG0_ADDR_RANGE_LO32; + + hub->vm_cntx_cntl_vm_fault = GCVM_CONTEXT1_CNTL__RANGE_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK | + GCVM_CONTEXT1_CNTL__DUMMY_PAGE_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK | + GCVM_CONTEXT1_CNTL__PDE0_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK | + GCVM_CONTEXT1_CNTL__VALID_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK | + GCVM_CONTEXT1_CNTL__READ_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK | + GCVM_CONTEXT1_CNTL__WRITE_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK | + GCVM_CONTEXT1_CNTL__EXECUTE_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK; + + hub->vmhub_funcs = &gfxhub_v3_0_vmhub_funcs; +} + +const struct amdgpu_gfxhub_funcs gfxhub_v3_0_funcs = { + .get_fb_location = gfxhub_v3_0_get_fb_location, + .get_mc_fb_offset = gfxhub_v3_0_get_mc_fb_offset, + .setup_vm_pt_regs = gfxhub_v3_0_setup_vm_pt_regs, + .gart_enable = gfxhub_v3_0_gart_enable, + .gart_disable = gfxhub_v3_0_gart_disable, + .set_fault_enable_default = gfxhub_v3_0_set_fault_enable_default, + .init = gfxhub_v3_0_init, +}; diff --git a/drivers/gpu/drm/amd/amdgpu/gfxhub_v3_0.h b/drivers/gpu/drm/amd/amdgpu/gfxhub_v3_0.h new file mode 100644 index 000000000000..ea345e4e072a --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/gfxhub_v3_0.h @@ -0,0 +1,29 @@ +/* + * Copyright 2021 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#ifndef __GFXHUB_V3_0_H__ +#define __GFXHUB_V3_0_H__ + +extern const struct amdgpu_gfxhub_funcs gfxhub_v3_0_funcs; + +#endif diff --git a/drivers/gpu/drm/amd/amdgpu/gfxhub_v3_0_3.c b/drivers/gpu/drm/amd/amdgpu/gfxhub_v3_0_3.c new file mode 100644 index 000000000000..5d3fffd4929f --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/gfxhub_v3_0_3.c @@ -0,0 +1,511 @@ +/* + * Copyright 2022 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#include "amdgpu.h" +#include "gfxhub_v3_0_3.h" + +#include "gc/gc_11_0_3_offset.h" +#include "gc/gc_11_0_3_sh_mask.h" +#include "navi10_enum.h" +#include "soc15_common.h" + +#define regGCVM_L2_CNTL3_DEFAULT 0x80100007 +#define regGCVM_L2_CNTL4_DEFAULT 0x000000c1 +#define regGCVM_L2_CNTL5_DEFAULT 0x00003fe0 + +static const char *gfxhub_client_ids[] = { + "CB/DB", + "Reserved", + "GE1", + "GE2", + "CPF", + "CPC", + "CPG", + "RLC", + "TCP", + "SQC (inst)", + "SQC (data)", + "SQG", + "Reserved", + "SDMA0", + "SDMA1", + "GCR", + "SDMA2", + "SDMA3", +}; + +static uint32_t gfxhub_v3_0_3_get_invalidate_req(unsigned int vmid, + uint32_t flush_type) +{ + u32 req = 0; + + /* invalidate using legacy mode on vmid*/ + req = REG_SET_FIELD(req, GCVM_INVALIDATE_ENG0_REQ, + PER_VMID_INVALIDATE_REQ, 1 << vmid); + req = REG_SET_FIELD(req, GCVM_INVALIDATE_ENG0_REQ, FLUSH_TYPE, flush_type); + req = REG_SET_FIELD(req, GCVM_INVALIDATE_ENG0_REQ, INVALIDATE_L2_PTES, 1); + req = REG_SET_FIELD(req, GCVM_INVALIDATE_ENG0_REQ, INVALIDATE_L2_PDE0, 1); + req = REG_SET_FIELD(req, GCVM_INVALIDATE_ENG0_REQ, INVALIDATE_L2_PDE1, 1); + req = REG_SET_FIELD(req, GCVM_INVALIDATE_ENG0_REQ, INVALIDATE_L2_PDE2, 1); + req = REG_SET_FIELD(req, GCVM_INVALIDATE_ENG0_REQ, INVALIDATE_L1_PTES, 1); + req = REG_SET_FIELD(req, GCVM_INVALIDATE_ENG0_REQ, + CLEAR_PROTECTION_FAULT_STATUS_ADDR, 0); + + return req; +} + +static void +gfxhub_v3_0_3_print_l2_protection_fault_status(struct amdgpu_device *adev, + uint32_t status) +{ + u32 cid = REG_GET_FIELD(status, + GCVM_L2_PROTECTION_FAULT_STATUS, CID); + + dev_err(adev->dev, + "GCVM_L2_PROTECTION_FAULT_STATUS:0x%08X\n", + status); + dev_err(adev->dev, "\t Faulty UTCL2 client ID: %s (0x%x)\n", + cid >= ARRAY_SIZE(gfxhub_client_ids) ? "unknown" : gfxhub_client_ids[cid], + cid); + dev_err(adev->dev, "\t MORE_FAULTS: 0x%lx\n", + REG_GET_FIELD(status, + GCVM_L2_PROTECTION_FAULT_STATUS, MORE_FAULTS)); + dev_err(adev->dev, "\t WALKER_ERROR: 0x%lx\n", + REG_GET_FIELD(status, + GCVM_L2_PROTECTION_FAULT_STATUS, WALKER_ERROR)); + dev_err(adev->dev, "\t PERMISSION_FAULTS: 0x%lx\n", + REG_GET_FIELD(status, + GCVM_L2_PROTECTION_FAULT_STATUS, PERMISSION_FAULTS)); + dev_err(adev->dev, "\t MAPPING_ERROR: 0x%lx\n", + REG_GET_FIELD(status, + GCVM_L2_PROTECTION_FAULT_STATUS, MAPPING_ERROR)); + dev_err(adev->dev, "\t RW: 0x%lx\n", + REG_GET_FIELD(status, + GCVM_L2_PROTECTION_FAULT_STATUS, RW)); +} + +static u64 gfxhub_v3_0_3_get_fb_location(struct amdgpu_device *adev) +{ + u64 base = RREG32_SOC15(GC, 0, regGCMC_VM_FB_LOCATION_BASE); + + base &= GCMC_VM_FB_LOCATION_BASE__FB_BASE_MASK; + base <<= 24; + + return base; +} + +static u64 gfxhub_v3_0_3_get_mc_fb_offset(struct amdgpu_device *adev) +{ + return (u64)RREG32_SOC15(GC, 0, regGCMC_VM_FB_OFFSET) << 24; +} + +static void gfxhub_v3_0_3_setup_vm_pt_regs(struct amdgpu_device *adev, uint32_t vmid, + uint64_t page_table_base) +{ + struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0]; + + WREG32_SOC15_OFFSET(GC, 0, regGCVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32, + hub->ctx_addr_distance * vmid, + lower_32_bits(page_table_base)); + + WREG32_SOC15_OFFSET(GC, 0, regGCVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_HI32, + hub->ctx_addr_distance * vmid, + upper_32_bits(page_table_base)); +} + +static void gfxhub_v3_0_3_init_gart_aperture_regs(struct amdgpu_device *adev) +{ + uint64_t pt_base = amdgpu_gmc_pd_addr(adev->gart.bo); + + gfxhub_v3_0_3_setup_vm_pt_regs(adev, 0, pt_base); + + WREG32_SOC15(GC, 0, regGCVM_CONTEXT0_PAGE_TABLE_START_ADDR_LO32, + (u32)(adev->gmc.gart_start >> 12)); + WREG32_SOC15(GC, 0, regGCVM_CONTEXT0_PAGE_TABLE_START_ADDR_HI32, + (u32)(adev->gmc.gart_start >> 44)); + + WREG32_SOC15(GC, 0, regGCVM_CONTEXT0_PAGE_TABLE_END_ADDR_LO32, + (u32)(adev->gmc.gart_end >> 12)); + WREG32_SOC15(GC, 0, regGCVM_CONTEXT0_PAGE_TABLE_END_ADDR_HI32, + (u32)(adev->gmc.gart_end >> 44)); +} + +static void gfxhub_v3_0_3_init_system_aperture_regs(struct amdgpu_device *adev) +{ + uint64_t value; + + /* Disable AGP. */ + WREG32_SOC15(GC, 0, regGCMC_VM_AGP_BASE, 0); + WREG32_SOC15(GC, 0, regGCMC_VM_AGP_TOP, 0); + WREG32_SOC15(GC, 0, regGCMC_VM_AGP_BOT, 0x00FFFFFF); + + /* Program the system aperture low logical page number. */ + WREG32_SOC15(GC, 0, regGCMC_VM_SYSTEM_APERTURE_LOW_ADDR, + adev->gmc.vram_start >> 18); + WREG32_SOC15(GC, 0, regGCMC_VM_SYSTEM_APERTURE_HIGH_ADDR, + adev->gmc.vram_end >> 18); + + /* Set default page address. */ + value = adev->vram_scratch.gpu_addr - adev->gmc.vram_start + + adev->vm_manager.vram_base_offset; + WREG32_SOC15(GC, 0, regGCMC_VM_SYSTEM_APERTURE_DEFAULT_ADDR_LSB, + (u32)(value >> 12)); + WREG32_SOC15(GC, 0, regGCMC_VM_SYSTEM_APERTURE_DEFAULT_ADDR_MSB, + (u32)(value >> 44)); + + /* Program "protection fault". */ + WREG32_SOC15(GC, 0, regGCVM_L2_PROTECTION_FAULT_DEFAULT_ADDR_LO32, + (u32)(adev->dummy_page_addr >> 12)); + WREG32_SOC15(GC, 0, regGCVM_L2_PROTECTION_FAULT_DEFAULT_ADDR_HI32, + (u32)((u64)adev->dummy_page_addr >> 44)); + + WREG32_FIELD15_PREREG(GC, 0, GCVM_L2_PROTECTION_FAULT_CNTL2, + ACTIVE_PAGE_MIGRATION_PTE_READ_RETRY, 1); +} + + +static void gfxhub_v3_0_3_init_tlb_regs(struct amdgpu_device *adev) +{ + uint32_t tmp; + + /* Setup TLB control */ + tmp = RREG32_SOC15(GC, 0, regGCMC_VM_MX_L1_TLB_CNTL); + + tmp = REG_SET_FIELD(tmp, GCMC_VM_MX_L1_TLB_CNTL, ENABLE_L1_TLB, 1); + tmp = REG_SET_FIELD(tmp, GCMC_VM_MX_L1_TLB_CNTL, SYSTEM_ACCESS_MODE, 3); + tmp = REG_SET_FIELD(tmp, GCMC_VM_MX_L1_TLB_CNTL, + ENABLE_ADVANCED_DRIVER_MODEL, 1); + tmp = REG_SET_FIELD(tmp, GCMC_VM_MX_L1_TLB_CNTL, + SYSTEM_APERTURE_UNMAPPED_ACCESS, 0); + tmp = REG_SET_FIELD(tmp, GCMC_VM_MX_L1_TLB_CNTL, ECO_BITS, 0); + tmp = REG_SET_FIELD(tmp, GCMC_VM_MX_L1_TLB_CNTL, + MTYPE, MTYPE_UC); /* UC, uncached */ + + WREG32_SOC15(GC, 0, regGCMC_VM_MX_L1_TLB_CNTL, tmp); +} + +static void gfxhub_v3_0_3_init_cache_regs(struct amdgpu_device *adev) +{ + uint32_t tmp; + + /* These registers are not accessible to VF-SRIOV. + * The PF will program them instead. + */ + if (amdgpu_sriov_vf(adev)) + return; + + /* Setup L2 cache */ + tmp = RREG32_SOC15(GC, 0, regGCVM_L2_CNTL); + tmp = REG_SET_FIELD(tmp, GCVM_L2_CNTL, ENABLE_L2_CACHE, 1); + tmp = REG_SET_FIELD(tmp, GCVM_L2_CNTL, ENABLE_L2_FRAGMENT_PROCESSING, 0); + tmp = REG_SET_FIELD(tmp, GCVM_L2_CNTL, + ENABLE_DEFAULT_PAGE_OUT_TO_SYSTEM_MEMORY, 1); + /* XXX for emulation, Refer to closed source code.*/ + tmp = REG_SET_FIELD(tmp, GCVM_L2_CNTL, + L2_PDE0_CACHE_TAG_GENERATION_MODE, 0); + tmp = REG_SET_FIELD(tmp, GCVM_L2_CNTL, PDE_FAULT_CLASSIFICATION, 0); + tmp = REG_SET_FIELD(tmp, GCVM_L2_CNTL, CONTEXT1_IDENTITY_ACCESS_MODE, 1); + tmp = REG_SET_FIELD(tmp, GCVM_L2_CNTL, IDENTITY_MODE_FRAGMENT_SIZE, 0); + WREG32_SOC15(GC, 0, regGCVM_L2_CNTL, tmp); + + tmp = RREG32_SOC15(GC, 0, regGCVM_L2_CNTL2); + tmp = REG_SET_FIELD(tmp, GCVM_L2_CNTL2, INVALIDATE_ALL_L1_TLBS, 1); + tmp = REG_SET_FIELD(tmp, GCVM_L2_CNTL2, INVALIDATE_L2_CACHE, 1); + WREG32_SOC15(GC, 0, regGCVM_L2_CNTL2, tmp); + + tmp = regGCVM_L2_CNTL3_DEFAULT; + if (adev->gmc.translate_further) { + tmp = REG_SET_FIELD(tmp, GCVM_L2_CNTL3, BANK_SELECT, 12); + tmp = REG_SET_FIELD(tmp, GCVM_L2_CNTL3, + L2_CACHE_BIGK_FRAGMENT_SIZE, 9); + } else { + tmp = REG_SET_FIELD(tmp, GCVM_L2_CNTL3, BANK_SELECT, 9); + tmp = REG_SET_FIELD(tmp, GCVM_L2_CNTL3, + L2_CACHE_BIGK_FRAGMENT_SIZE, 6); + } + WREG32_SOC15(GC, 0, regGCVM_L2_CNTL3, tmp); + + tmp = regGCVM_L2_CNTL4_DEFAULT; + tmp = REG_SET_FIELD(tmp, GCVM_L2_CNTL4, VMC_TAP_PDE_REQUEST_PHYSICAL, 0); + tmp = REG_SET_FIELD(tmp, GCVM_L2_CNTL4, VMC_TAP_PTE_REQUEST_PHYSICAL, 0); + WREG32_SOC15(GC, 0, regGCVM_L2_CNTL4, tmp); + + tmp = regGCVM_L2_CNTL5_DEFAULT; + tmp = REG_SET_FIELD(tmp, GCVM_L2_CNTL5, L2_CACHE_SMALLK_FRAGMENT_SIZE, 0); + WREG32_SOC15(GC, 0, regGCVM_L2_CNTL5, tmp); +} + +static void gfxhub_v3_0_3_enable_system_domain(struct amdgpu_device *adev) +{ + uint32_t tmp; + + tmp = RREG32_SOC15(GC, 0, regGCVM_CONTEXT0_CNTL); + tmp = REG_SET_FIELD(tmp, GCVM_CONTEXT0_CNTL, ENABLE_CONTEXT, 1); + tmp = REG_SET_FIELD(tmp, GCVM_CONTEXT0_CNTL, PAGE_TABLE_DEPTH, 0); + tmp = REG_SET_FIELD(tmp, GCVM_CONTEXT0_CNTL, + RETRY_PERMISSION_OR_INVALID_PAGE_FAULT, 0); + WREG32_SOC15(GC, 0, regGCVM_CONTEXT0_CNTL, tmp); +} + +static void gfxhub_v3_0_3_disable_identity_aperture(struct amdgpu_device *adev) +{ + /* These registers are not accessible to VF-SRIOV. + * The PF will program them instead. + */ + if (amdgpu_sriov_vf(adev)) + return; + + WREG32_SOC15(GC, 0, regGCVM_L2_CONTEXT1_IDENTITY_APERTURE_LOW_ADDR_LO32, + 0xFFFFFFFF); + WREG32_SOC15(GC, 0, regGCVM_L2_CONTEXT1_IDENTITY_APERTURE_LOW_ADDR_HI32, + 0x0000000F); + + WREG32_SOC15(GC, 0, regGCVM_L2_CONTEXT1_IDENTITY_APERTURE_HIGH_ADDR_LO32, + 0); + WREG32_SOC15(GC, 0, regGCVM_L2_CONTEXT1_IDENTITY_APERTURE_HIGH_ADDR_HI32, + 0); + + WREG32_SOC15(GC, 0, regGCVM_L2_CONTEXT_IDENTITY_PHYSICAL_OFFSET_LO32, 0); + WREG32_SOC15(GC, 0, regGCVM_L2_CONTEXT_IDENTITY_PHYSICAL_OFFSET_HI32, 0); + +} + +static void gfxhub_v3_0_3_setup_vmid_config(struct amdgpu_device *adev) +{ + struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0]; + int i; + uint32_t tmp; + + for (i = 0; i <= 14; i++) { + tmp = RREG32_SOC15_OFFSET(GC, 0, regGCVM_CONTEXT1_CNTL, i); + tmp = REG_SET_FIELD(tmp, GCVM_CONTEXT1_CNTL, ENABLE_CONTEXT, 1); + tmp = REG_SET_FIELD(tmp, GCVM_CONTEXT1_CNTL, PAGE_TABLE_DEPTH, + adev->vm_manager.num_level); + tmp = REG_SET_FIELD(tmp, GCVM_CONTEXT1_CNTL, + RANGE_PROTECTION_FAULT_ENABLE_DEFAULT, 1); + tmp = REG_SET_FIELD(tmp, GCVM_CONTEXT1_CNTL, + DUMMY_PAGE_PROTECTION_FAULT_ENABLE_DEFAULT, 1); + tmp = REG_SET_FIELD(tmp, GCVM_CONTEXT1_CNTL, + PDE0_PROTECTION_FAULT_ENABLE_DEFAULT, 1); + tmp = REG_SET_FIELD(tmp, GCVM_CONTEXT1_CNTL, + VALID_PROTECTION_FAULT_ENABLE_DEFAULT, 1); + tmp = REG_SET_FIELD(tmp, GCVM_CONTEXT1_CNTL, + READ_PROTECTION_FAULT_ENABLE_DEFAULT, 1); + tmp = REG_SET_FIELD(tmp, GCVM_CONTEXT1_CNTL, + WRITE_PROTECTION_FAULT_ENABLE_DEFAULT, 1); + tmp = REG_SET_FIELD(tmp, GCVM_CONTEXT1_CNTL, + EXECUTE_PROTECTION_FAULT_ENABLE_DEFAULT, 1); + tmp = REG_SET_FIELD(tmp, GCVM_CONTEXT1_CNTL, + PAGE_TABLE_BLOCK_SIZE, + adev->vm_manager.block_size - 9); + /* Send no-retry XNACK on fault to suppress VM fault storm. */ + tmp = REG_SET_FIELD(tmp, GCVM_CONTEXT1_CNTL, + RETRY_PERMISSION_OR_INVALID_PAGE_FAULT, + !amdgpu_noretry); + WREG32_SOC15_OFFSET(GC, 0, regGCVM_CONTEXT1_CNTL, + i * hub->ctx_distance, tmp); + WREG32_SOC15_OFFSET(GC, 0, regGCVM_CONTEXT1_PAGE_TABLE_START_ADDR_LO32, + i * hub->ctx_addr_distance, 0); + WREG32_SOC15_OFFSET(GC, 0, regGCVM_CONTEXT1_PAGE_TABLE_START_ADDR_HI32, + i * hub->ctx_addr_distance, 0); + WREG32_SOC15_OFFSET(GC, 0, regGCVM_CONTEXT1_PAGE_TABLE_END_ADDR_LO32, + i * hub->ctx_addr_distance, + lower_32_bits(adev->vm_manager.max_pfn - 1)); + WREG32_SOC15_OFFSET(GC, 0, regGCVM_CONTEXT1_PAGE_TABLE_END_ADDR_HI32, + i * hub->ctx_addr_distance, + upper_32_bits(adev->vm_manager.max_pfn - 1)); + } + + hub->vm_cntx_cntl = tmp; +} + +static void gfxhub_v3_0_3_program_invalidation(struct amdgpu_device *adev) +{ + struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0]; + unsigned i; + + for (i = 0 ; i < 18; ++i) { + WREG32_SOC15_OFFSET(GC, 0, regGCVM_INVALIDATE_ENG0_ADDR_RANGE_LO32, + i * hub->eng_addr_distance, 0xffffffff); + WREG32_SOC15_OFFSET(GC, 0, regGCVM_INVALIDATE_ENG0_ADDR_RANGE_HI32, + i * hub->eng_addr_distance, 0x1f); + } +} + +static int gfxhub_v3_0_3_gart_enable(struct amdgpu_device *adev) +{ + if (amdgpu_sriov_vf(adev)) { + /* + * GCMC_VM_FB_LOCATION_BASE/TOP is NULL for VF, becuase they are + * VF copy registers so vbios post doesn't program them, for + * SRIOV driver need to program them + */ + WREG32_SOC15(GC, 0, regGCMC_VM_FB_LOCATION_BASE, + adev->gmc.vram_start >> 24); + WREG32_SOC15(GC, 0, regGCMC_VM_FB_LOCATION_TOP, + adev->gmc.vram_end >> 24); + } + + /* GART Enable. */ + gfxhub_v3_0_3_init_gart_aperture_regs(adev); + gfxhub_v3_0_3_init_system_aperture_regs(adev); + gfxhub_v3_0_3_init_tlb_regs(adev); + gfxhub_v3_0_3_init_cache_regs(adev); + + gfxhub_v3_0_3_enable_system_domain(adev); + gfxhub_v3_0_3_disable_identity_aperture(adev); + gfxhub_v3_0_3_setup_vmid_config(adev); + gfxhub_v3_0_3_program_invalidation(adev); + + return 0; +} + +static void gfxhub_v3_0_3_gart_disable(struct amdgpu_device *adev) +{ + struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0]; + u32 tmp; + u32 i; + + /* Disable all tables */ + for (i = 0; i < 16; i++) + WREG32_SOC15_OFFSET(GC, 0, regGCVM_CONTEXT0_CNTL, + i * hub->ctx_distance, 0); + + /* Setup TLB control */ + tmp = RREG32_SOC15(GC, 0, regGCMC_VM_MX_L1_TLB_CNTL); + tmp = REG_SET_FIELD(tmp, GCMC_VM_MX_L1_TLB_CNTL, ENABLE_L1_TLB, 0); + tmp = REG_SET_FIELD(tmp, GCMC_VM_MX_L1_TLB_CNTL, + ENABLE_ADVANCED_DRIVER_MODEL, 0); + WREG32_SOC15(GC, 0, regGCMC_VM_MX_L1_TLB_CNTL, tmp); + + /* Setup L2 cache */ + WREG32_FIELD15_PREREG(GC, 0, GCVM_L2_CNTL, ENABLE_L2_CACHE, 0); + WREG32_SOC15(GC, 0, regGCVM_L2_CNTL3, 0); +} + +/** + * gfxhub_v3_0_3_set_fault_enable_default - update GART/VM fault handling + * + * @adev: amdgpu_device pointer + * @value: true redirects VM faults to the default page + */ +static void gfxhub_v3_0_3_set_fault_enable_default(struct amdgpu_device *adev, + bool value) +{ + u32 tmp; + + /* These registers are not accessible to VF-SRIOV. + * The PF will program them instead. + */ + if (amdgpu_sriov_vf(adev)) + return; + + tmp = RREG32_SOC15(GC, 0, regGCVM_L2_PROTECTION_FAULT_CNTL); + tmp = REG_SET_FIELD(tmp, GCVM_L2_PROTECTION_FAULT_CNTL, + RANGE_PROTECTION_FAULT_ENABLE_DEFAULT, value); + tmp = REG_SET_FIELD(tmp, GCVM_L2_PROTECTION_FAULT_CNTL, + PDE0_PROTECTION_FAULT_ENABLE_DEFAULT, value); + tmp = REG_SET_FIELD(tmp, GCVM_L2_PROTECTION_FAULT_CNTL, + PDE1_PROTECTION_FAULT_ENABLE_DEFAULT, value); + tmp = REG_SET_FIELD(tmp, GCVM_L2_PROTECTION_FAULT_CNTL, + PDE2_PROTECTION_FAULT_ENABLE_DEFAULT, value); + tmp = REG_SET_FIELD(tmp, GCVM_L2_PROTECTION_FAULT_CNTL, + TRANSLATE_FURTHER_PROTECTION_FAULT_ENABLE_DEFAULT, + value); + tmp = REG_SET_FIELD(tmp, GCVM_L2_PROTECTION_FAULT_CNTL, + NACK_PROTECTION_FAULT_ENABLE_DEFAULT, value); + tmp = REG_SET_FIELD(tmp, GCVM_L2_PROTECTION_FAULT_CNTL, + DUMMY_PAGE_PROTECTION_FAULT_ENABLE_DEFAULT, value); + tmp = REG_SET_FIELD(tmp, GCVM_L2_PROTECTION_FAULT_CNTL, + VALID_PROTECTION_FAULT_ENABLE_DEFAULT, value); + tmp = REG_SET_FIELD(tmp, GCVM_L2_PROTECTION_FAULT_CNTL, + READ_PROTECTION_FAULT_ENABLE_DEFAULT, value); + tmp = REG_SET_FIELD(tmp, GCVM_L2_PROTECTION_FAULT_CNTL, + WRITE_PROTECTION_FAULT_ENABLE_DEFAULT, value); + tmp = REG_SET_FIELD(tmp, GCVM_L2_PROTECTION_FAULT_CNTL, + EXECUTE_PROTECTION_FAULT_ENABLE_DEFAULT, value); + if (!value) { + tmp = REG_SET_FIELD(tmp, GCVM_L2_PROTECTION_FAULT_CNTL, + CRASH_ON_NO_RETRY_FAULT, 1); + tmp = REG_SET_FIELD(tmp, GCVM_L2_PROTECTION_FAULT_CNTL, + CRASH_ON_RETRY_FAULT, 1); + } + WREG32_SOC15(GC, 0, regGCVM_L2_PROTECTION_FAULT_CNTL, tmp); +} + +static const struct amdgpu_vmhub_funcs gfxhub_v3_0_3_vmhub_funcs = { + .print_l2_protection_fault_status = gfxhub_v3_0_3_print_l2_protection_fault_status, + .get_invalidate_req = gfxhub_v3_0_3_get_invalidate_req, +}; + +static void gfxhub_v3_0_3_init(struct amdgpu_device *adev) +{ + struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0]; + + hub->ctx0_ptb_addr_lo32 = + SOC15_REG_OFFSET(GC, 0, + regGCVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32); + hub->ctx0_ptb_addr_hi32 = + SOC15_REG_OFFSET(GC, 0, + regGCVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_HI32); + hub->vm_inv_eng0_sem = + SOC15_REG_OFFSET(GC, 0, regGCVM_INVALIDATE_ENG0_SEM); + hub->vm_inv_eng0_req = + SOC15_REG_OFFSET(GC, 0, regGCVM_INVALIDATE_ENG0_REQ); + hub->vm_inv_eng0_ack = + SOC15_REG_OFFSET(GC, 0, regGCVM_INVALIDATE_ENG0_ACK); + hub->vm_context0_cntl = + SOC15_REG_OFFSET(GC, 0, regGCVM_CONTEXT0_CNTL); + hub->vm_l2_pro_fault_status = + SOC15_REG_OFFSET(GC, 0, regGCVM_L2_PROTECTION_FAULT_STATUS); + hub->vm_l2_pro_fault_cntl = + SOC15_REG_OFFSET(GC, 0, regGCVM_L2_PROTECTION_FAULT_CNTL); + + hub->ctx_distance = regGCVM_CONTEXT1_CNTL - regGCVM_CONTEXT0_CNTL; + hub->ctx_addr_distance = regGCVM_CONTEXT1_PAGE_TABLE_BASE_ADDR_LO32 - + regGCVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32; + hub->eng_distance = regGCVM_INVALIDATE_ENG1_REQ - + regGCVM_INVALIDATE_ENG0_REQ; + hub->eng_addr_distance = regGCVM_INVALIDATE_ENG1_ADDR_RANGE_LO32 - + regGCVM_INVALIDATE_ENG0_ADDR_RANGE_LO32; + + hub->vm_cntx_cntl_vm_fault = GCVM_CONTEXT1_CNTL__RANGE_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK | + GCVM_CONTEXT1_CNTL__DUMMY_PAGE_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK | + GCVM_CONTEXT1_CNTL__PDE0_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK | + GCVM_CONTEXT1_CNTL__VALID_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK | + GCVM_CONTEXT1_CNTL__READ_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK | + GCVM_CONTEXT1_CNTL__WRITE_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK | + GCVM_CONTEXT1_CNTL__EXECUTE_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK; + + hub->vmhub_funcs = &gfxhub_v3_0_3_vmhub_funcs; +} + +const struct amdgpu_gfxhub_funcs gfxhub_v3_0_3_funcs = { + .get_fb_location = gfxhub_v3_0_3_get_fb_location, + .get_mc_fb_offset = gfxhub_v3_0_3_get_mc_fb_offset, + .setup_vm_pt_regs = gfxhub_v3_0_3_setup_vm_pt_regs, + .gart_enable = gfxhub_v3_0_3_gart_enable, + .gart_disable = gfxhub_v3_0_3_gart_disable, + .set_fault_enable_default = gfxhub_v3_0_3_set_fault_enable_default, + .init = gfxhub_v3_0_3_init, +}; diff --git a/drivers/gpu/drm/amd/amdgpu/gfxhub_v3_0_3.h b/drivers/gpu/drm/amd/amdgpu/gfxhub_v3_0_3.h new file mode 100644 index 000000000000..6153bd5e3083 --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/gfxhub_v3_0_3.h @@ -0,0 +1,29 @@ +/* + * Copyright 2022 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#ifndef __GFXHUB_V3_0_3_H__ +#define __GFXHUB_V3_0_3_H__ + +extern const struct amdgpu_gfxhub_funcs gfxhub_v3_0_3_funcs; + +#endif diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c index 3d5d47a799e3..f513e2c2e964 100644 --- a/drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c @@ -22,6 +22,9 @@ */ #include <linux/firmware.h> #include <linux/pci.h> + +#include <drm/drm_cache.h> + #include "amdgpu.h" #include "amdgpu_atomfirmware.h" #include "gmc_v10_0.h" @@ -48,6 +51,8 @@ #include "athub_v2_0.h" #include "athub_v2_1.h" +#include "amdgpu_reset.h" + #if 0 static const struct soc15_reg_golden golden_settings_navi10_hdp[] = { @@ -326,9 +331,9 @@ static void gmc_v10_0_flush_gpu_tlb(struct amdgpu_device *adev, uint32_t vmid, /* For SRIOV run time, driver shouldn't access the register through MMIO * Directly use kiq to do the vm invalidation instead */ - if (adev->gfx.kiq.ring.sched.ready && + if (adev->gfx.kiq.ring.sched.ready && !adev->enable_mes && (amdgpu_sriov_runtime(adev) || !amdgpu_sriov_vf(adev)) && - down_read_trylock(&adev->reset_sem)) { + down_read_trylock(&adev->reset_domain->sem)) { struct amdgpu_vmhub *hub = &adev->vmhub[vmhub]; const unsigned eng = 17; u32 inv_req = hub->vmhub_funcs->get_invalidate_req(vmid, flush_type); @@ -338,7 +343,7 @@ static void gmc_v10_0_flush_gpu_tlb(struct amdgpu_device *adev, uint32_t vmid, amdgpu_virt_kiq_reg_write_reg_wait(adev, req, ack, inv_req, 1 << vmid); - up_read(&adev->reset_sem); + up_read(&adev->reset_domain->sem); return; } @@ -414,6 +419,7 @@ static int gmc_v10_0_flush_gpu_tlb_pasid(struct amdgpu_device *adev, uint32_t seq; uint16_t queried_pasid; bool ret; + u32 usec_timeout = amdgpu_sriov_vf(adev) ? SRIOV_USEC_TIMEOUT : adev->usec_timeout; struct amdgpu_ring *ring = &adev->gfx.kiq.ring; struct amdgpu_kiq *kiq = &adev->gfx.kiq; @@ -432,7 +438,7 @@ static int gmc_v10_0_flush_gpu_tlb_pasid(struct amdgpu_device *adev, amdgpu_ring_commit(ring); spin_unlock(&adev->gfx.kiq.ring_lock); - r = amdgpu_fence_wait_polling(ring, seq, adev->usec_timeout); + r = amdgpu_fence_wait_polling(ring, seq, usec_timeout); if (r < 1) { dev_err(adev->dev, "wait for kiq fence error: %ld.\n", r); return -ETIME; @@ -454,7 +460,8 @@ static int gmc_v10_0_flush_gpu_tlb_pasid(struct amdgpu_device *adev, gmc_v10_0_flush_gpu_tlb(adev, vmid, AMDGPU_GFXHUB_0, flush_type); } - break; + if (!adev->enable_mes) + break; } } @@ -515,6 +522,10 @@ static void gmc_v10_0_emit_pasid_mapping(struct amdgpu_ring *ring, unsigned vmid struct amdgpu_device *adev = ring->adev; uint32_t reg; + /* MES fw manages IH_VMID_x_LUT updating */ + if (ring->is_mes_queue) + return; + if (ring->funcs->vmhub == AMDGPU_GFXHUB_0) reg = SOC15_REG_OFFSET(OSSSYS, 0, mmIH_VMID_0_LUT) + vmid; else @@ -607,6 +618,9 @@ static void gmc_v10_0_get_vm_pte(struct amdgpu_device *adev, *flags &= ~AMDGPU_PTE_MTYPE_NV10_MASK; *flags |= (mapping->flags & AMDGPU_PTE_MTYPE_NV10_MASK); + *flags &= ~AMDGPU_PTE_NOALLOC; + *flags |= (mapping->flags & AMDGPU_PTE_NOALLOC); + if (mapping->flags & AMDGPU_PTE_PRT) { *flags |= AMDGPU_PTE_PRT; *flags |= AMDGPU_PTE_SNOOPED; @@ -664,11 +678,27 @@ static void gmc_v10_0_set_umc_funcs(struct amdgpu_device *adev) adev->umc.umc_inst_num = UMC_V8_7_UMC_INSTANCE_NUM; adev->umc.channel_offs = UMC_V8_7_PER_CHANNEL_OFFSET_SIENNA; adev->umc.channel_idx_tbl = &umc_v8_7_channel_idx_tbl[0][0]; - adev->umc.ras_funcs = &umc_v8_7_ras_funcs; + adev->umc.ras = &umc_v8_7_ras; break; default: break; } + if (adev->umc.ras) { + amdgpu_ras_register_ras_block(adev, &adev->umc.ras->ras_block); + + strcpy(adev->umc.ras->ras_block.ras_comm.name, "umc"); + adev->umc.ras->ras_block.ras_comm.block = AMDGPU_RAS_BLOCK__UMC; + adev->umc.ras->ras_block.ras_comm.type = AMDGPU_RAS_ERROR__MULTI_UNCORRECTABLE; + adev->umc.ras_if = &adev->umc.ras->ras_block.ras_comm; + + /* If don't define special ras_late_init function, use default ras_late_init */ + if (!adev->umc.ras->ras_block.ras_late_init) + adev->umc.ras->ras_block.ras_late_init = amdgpu_umc_ras_late_init; + + /* If not defined special ras_cb function, use default ras_cb */ + if (!adev->umc.ras->ras_block.ras_cb) + adev->umc.ras->ras_block.ras_cb = amdgpu_umc_process_ras_data_cb; + } } @@ -677,6 +707,7 @@ static void gmc_v10_0_set_mmhub_funcs(struct amdgpu_device *adev) switch (adev->ip_versions[MMHUB_HWIP][0]) { case IP_VERSION(2, 3, 0): case IP_VERSION(2, 4, 0): + case IP_VERSION(2, 4, 1): adev->mmhub.funcs = &mmhub_v2_3_funcs; break; default: @@ -693,7 +724,9 @@ static void gmc_v10_0_set_gfxhub_funcs(struct amdgpu_device *adev) case IP_VERSION(10, 3, 1): case IP_VERSION(10, 3, 4): case IP_VERSION(10, 3, 5): + case IP_VERSION(10, 3, 6): case IP_VERSION(10, 3, 3): + case IP_VERSION(10, 3, 7): adev->gfxhub.funcs = &gfxhub_v2_1_funcs; break; default: @@ -705,6 +738,7 @@ static void gmc_v10_0_set_gfxhub_funcs(struct amdgpu_device *adev) static int gmc_v10_0_early_init(void *handle) { + int r; struct amdgpu_device *adev = (struct amdgpu_device *)handle; gmc_v10_0_set_mmhub_funcs(adev); @@ -720,6 +754,10 @@ static int gmc_v10_0_early_init(void *handle) adev->gmc.private_aperture_end = adev->gmc.private_aperture_start + (4ULL << 30) - 1; + r = amdgpu_gmc_ras_early_init(adev); + if (r) + return r; + return 0; } @@ -788,7 +826,7 @@ static int gmc_v10_0_mc_init(struct amdgpu_device *adev) adev->gmc.aper_size = pci_resource_len(adev->pdev, 0); #ifdef CONFIG_X86_64 - if (adev->flags & AMD_IS_APU) { + if ((adev->flags & AMD_IS_APU) && !amdgpu_passthrough(adev)) { adev->gmc.aper_base = adev->gfxhub.funcs->get_mc_fb_offset(adev); adev->gmc.aper_size = adev->gmc.real_vram_size; } @@ -800,10 +838,21 @@ static int gmc_v10_0_mc_init(struct amdgpu_device *adev) adev->gmc.visible_vram_size = adev->gmc.real_vram_size; /* set the gart size */ - if (amdgpu_gart_size == -1) - adev->gmc.gart_size = 512ULL << 20; - else + if (amdgpu_gart_size == -1) { + switch (adev->ip_versions[GC_HWIP][0]) { + default: + adev->gmc.gart_size = 512ULL << 20; + break; + case IP_VERSION(10, 3, 1): /* DCE SG support */ + case IP_VERSION(10, 3, 3): /* DCE SG support */ + case IP_VERSION(10, 3, 6): /* DCE SG support */ + case IP_VERSION(10, 3, 7): /* DCE SG support */ + adev->gmc.gart_size = 1024ULL << 20; + break; + } + } else { adev->gmc.gart_size = (u64)amdgpu_gart_size << 20; + } gmc_v10_0_vram_gtt_location(adev, &adev->gmc); @@ -858,16 +907,37 @@ static int gmc_v10_0_sw_init(void *handle) } switch (adev->ip_versions[GC_HWIP][0]) { + case IP_VERSION(10, 3, 0): + adev->gmc.mall_size = 128 * 1024 * 1024; + break; + case IP_VERSION(10, 3, 2): + adev->gmc.mall_size = 96 * 1024 * 1024; + break; + case IP_VERSION(10, 3, 4): + adev->gmc.mall_size = 32 * 1024 * 1024; + break; + case IP_VERSION(10, 3, 5): + adev->gmc.mall_size = 16 * 1024 * 1024; + break; + default: + adev->gmc.mall_size = 0; + break; + } + + switch (adev->ip_versions[GC_HWIP][0]) { case IP_VERSION(10, 1, 10): case IP_VERSION(10, 1, 1): case IP_VERSION(10, 1, 2): case IP_VERSION(10, 1, 3): + case IP_VERSION(10, 1, 4): case IP_VERSION(10, 3, 0): case IP_VERSION(10, 3, 2): case IP_VERSION(10, 3, 1): case IP_VERSION(10, 3, 4): case IP_VERSION(10, 3, 5): + case IP_VERSION(10, 3, 6): case IP_VERSION(10, 3, 3): + case IP_VERSION(10, 3, 7): adev->num_vmhubs = 2; /* * To fulfill 4-level page support, @@ -914,12 +984,13 @@ static int gmc_v10_0_sw_init(void *handle) return r; } + adev->need_swiotlb = drm_need_swiotlb(44); + r = gmc_v10_0_mc_init(adev); if (r) return r; amdgpu_gmc_get_vbios_allocations(adev); - amdgpu_gmc_get_reserved_allocation(adev); /* Memory manager */ r = amdgpu_bo_init(adev); @@ -986,14 +1057,7 @@ static int gmc_v10_0_gart_enable(struct amdgpu_device *adev) return -EINVAL; } - if (amdgpu_sriov_vf(adev) && amdgpu_in_reset(adev)) - goto skip_pin_bo; - - r = amdgpu_gart_table_vram_pin(adev); - if (r) - return r; - -skip_pin_bo: + amdgpu_gtt_mgr_recover(&adev->mman.gtt_mgr); r = adev->gfxhub.funcs->gart_enable(adev); if (r) return r; @@ -1019,8 +1083,6 @@ skip_pin_bo: (unsigned)(adev->gmc.gart_size >> 20), (unsigned long long)amdgpu_bo_gpu_offset(adev->gart.bo)); - adev->gart.ready = true; - return 0; } @@ -1043,6 +1105,12 @@ static int gmc_v10_0_hw_init(void *handle) if (r) return r; + if (amdgpu_emu_mode == 1) { + r = amdgpu_gmc_vram_checking(adev); + if (r) + return r; + } + if (adev->umc.funcs && adev->umc.funcs->init_registers) adev->umc.funcs->init_registers(adev); @@ -1060,7 +1128,6 @@ static void gmc_v10_0_gart_disable(struct amdgpu_device *adev) { adev->gfxhub.funcs->gart_disable(adev); adev->mmhub.funcs->gart_disable(adev); - amdgpu_gart_table_vram_unpin(adev); } static int gmc_v10_0_hw_fini(void *handle) @@ -1127,6 +1194,16 @@ static int gmc_v10_0_set_clockgating_state(void *handle, int r; struct amdgpu_device *adev = (struct amdgpu_device *)handle; + /* + * The issue mmhub can't disconnect from DF with MMHUB clock gating being disabled + * is a new problem observed at DF 3.0.3, however with the same suspend sequence not + * seen any issue on the DF 3.0.2 series platform. + */ + if (adev->in_s0ix && adev->ip_versions[DF_HWIP][0] > IP_VERSION(3, 0, 2)) { + dev_dbg(adev->dev, "keep mmhub clock gating being enabled for s0ix\n"); + return 0; + } + r = adev->mmhub.funcs->set_clockgating(adev, state); if (r) return r; @@ -1137,10 +1214,14 @@ static int gmc_v10_0_set_clockgating_state(void *handle, return athub_v2_0_set_clockgating(adev, state); } -static void gmc_v10_0_get_clockgating_state(void *handle, u32 *flags) +static void gmc_v10_0_get_clockgating_state(void *handle, u64 *flags) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; + if (adev->ip_versions[GC_HWIP][0] == IP_VERSION(10, 1, 3) || + adev->ip_versions[GC_HWIP][0] == IP_VERSION(10, 1, 4)) + return; + adev->mmhub.funcs->get_clockgating(adev, flags); if (adev->ip_versions[ATHUB_HWIP][0] >= IP_VERSION(2, 1, 0)) diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v11_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v11_0.c new file mode 100644 index 000000000000..66dfb574cc7d --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/gmc_v11_0.c @@ -0,0 +1,1032 @@ +/* + * Copyright 2021 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ +#include <linux/firmware.h> +#include <linux/pci.h> + +#include <drm/drm_cache.h> + +#include "amdgpu.h" +#include "amdgpu_atomfirmware.h" +#include "gmc_v11_0.h" +#include "umc_v8_10.h" +#include "athub/athub_3_0_0_sh_mask.h" +#include "athub/athub_3_0_0_offset.h" +#include "oss/osssys_6_0_0_offset.h" +#include "ivsrcid/vmc/irqsrcs_vmc_1_0.h" +#include "navi10_enum.h" +#include "soc15.h" +#include "soc15d.h" +#include "soc15_common.h" +#include "nbio_v4_3.h" +#include "gfxhub_v3_0.h" +#include "gfxhub_v3_0_3.h" +#include "mmhub_v3_0.h" +#include "mmhub_v3_0_1.h" +#include "mmhub_v3_0_2.h" +#include "athub_v3_0.h" + + +static int gmc_v11_0_ecc_interrupt_state(struct amdgpu_device *adev, + struct amdgpu_irq_src *src, + unsigned type, + enum amdgpu_interrupt_state state) +{ + return 0; +} + +static int +gmc_v11_0_vm_fault_interrupt_state(struct amdgpu_device *adev, + struct amdgpu_irq_src *src, unsigned type, + enum amdgpu_interrupt_state state) +{ + switch (state) { + case AMDGPU_IRQ_STATE_DISABLE: + /* MM HUB */ + amdgpu_gmc_set_vm_fault_masks(adev, AMDGPU_MMHUB_0, false); + /* GFX HUB */ + amdgpu_gmc_set_vm_fault_masks(adev, AMDGPU_GFXHUB_0, false); + break; + case AMDGPU_IRQ_STATE_ENABLE: + /* MM HUB */ + amdgpu_gmc_set_vm_fault_masks(adev, AMDGPU_MMHUB_0, true); + /* GFX HUB */ + amdgpu_gmc_set_vm_fault_masks(adev, AMDGPU_GFXHUB_0, true); + break; + default: + break; + } + + return 0; +} + +static int gmc_v11_0_process_interrupt(struct amdgpu_device *adev, + struct amdgpu_irq_src *source, + struct amdgpu_iv_entry *entry) +{ + struct amdgpu_vmhub *hub = &adev->vmhub[entry->vmid_src]; + uint32_t status = 0; + u64 addr; + + addr = (u64)entry->src_data[0] << 12; + addr |= ((u64)entry->src_data[1] & 0xf) << 44; + + if (!amdgpu_sriov_vf(adev)) { + /* + * Issue a dummy read to wait for the status register to + * be updated to avoid reading an incorrect value due to + * the new fast GRBM interface. + */ + if (entry->vmid_src == AMDGPU_GFXHUB_0) + RREG32(hub->vm_l2_pro_fault_status); + + status = RREG32(hub->vm_l2_pro_fault_status); + WREG32_P(hub->vm_l2_pro_fault_cntl, 1, ~1); + } + + if (printk_ratelimit()) { + struct amdgpu_task_info task_info; + + memset(&task_info, 0, sizeof(struct amdgpu_task_info)); + amdgpu_vm_get_task_info(adev, entry->pasid, &task_info); + + dev_err(adev->dev, + "[%s] page fault (src_id:%u ring:%u vmid:%u pasid:%u, " + "for process %s pid %d thread %s pid %d)\n", + entry->vmid_src ? "mmhub" : "gfxhub", + entry->src_id, entry->ring_id, entry->vmid, + entry->pasid, task_info.process_name, task_info.tgid, + task_info.task_name, task_info.pid); + dev_err(adev->dev, " in page starting at address 0x%016llx from client %d\n", + addr, entry->client_id); + if (!amdgpu_sriov_vf(adev)) + hub->vmhub_funcs->print_l2_protection_fault_status(adev, status); + } + + return 0; +} + +static const struct amdgpu_irq_src_funcs gmc_v11_0_irq_funcs = { + .set = gmc_v11_0_vm_fault_interrupt_state, + .process = gmc_v11_0_process_interrupt, +}; + +static const struct amdgpu_irq_src_funcs gmc_v11_0_ecc_funcs = { + .set = gmc_v11_0_ecc_interrupt_state, + .process = amdgpu_umc_process_ecc_irq, +}; + +static void gmc_v11_0_set_irq_funcs(struct amdgpu_device *adev) +{ + adev->gmc.vm_fault.num_types = 1; + adev->gmc.vm_fault.funcs = &gmc_v11_0_irq_funcs; + + if (!amdgpu_sriov_vf(adev)) { + adev->gmc.ecc_irq.num_types = 1; + adev->gmc.ecc_irq.funcs = &gmc_v11_0_ecc_funcs; + } +} + +/** + * gmc_v11_0_use_invalidate_semaphore - judge whether to use semaphore + * + * @adev: amdgpu_device pointer + * @vmhub: vmhub type + * + */ +static bool gmc_v11_0_use_invalidate_semaphore(struct amdgpu_device *adev, + uint32_t vmhub) +{ + return ((vmhub == AMDGPU_MMHUB_0) && + (!amdgpu_sriov_vf(adev))); +} + +static bool gmc_v11_0_get_vmid_pasid_mapping_info( + struct amdgpu_device *adev, + uint8_t vmid, uint16_t *p_pasid) +{ + *p_pasid = RREG32(SOC15_REG_OFFSET(OSSSYS, 0, regIH_VMID_0_LUT) + vmid) & 0xffff; + + return !!(*p_pasid); +} + +/* + * GART + * VMID 0 is the physical GPU addresses as used by the kernel. + * VMIDs 1-15 are used for userspace clients and are handled + * by the amdgpu vm/hsa code. + */ + +static void gmc_v11_0_flush_vm_hub(struct amdgpu_device *adev, uint32_t vmid, + unsigned int vmhub, uint32_t flush_type) +{ + bool use_semaphore = gmc_v11_0_use_invalidate_semaphore(adev, vmhub); + struct amdgpu_vmhub *hub = &adev->vmhub[vmhub]; + u32 inv_req = hub->vmhub_funcs->get_invalidate_req(vmid, flush_type); + u32 tmp; + /* Use register 17 for GART */ + const unsigned eng = 17; + unsigned int i; + unsigned char hub_ip = 0; + + hub_ip = (vmhub == AMDGPU_GFXHUB_0) ? + GC_HWIP : MMHUB_HWIP; + + spin_lock(&adev->gmc.invalidate_lock); + /* + * It may lose gpuvm invalidate acknowldege state across power-gating + * off cycle, add semaphore acquire before invalidation and semaphore + * release after invalidation to avoid entering power gated state + * to WA the Issue + */ + + /* TODO: It needs to continue working on debugging with semaphore for GFXHUB as well. */ + if (use_semaphore) { + for (i = 0; i < adev->usec_timeout; i++) { + /* a read return value of 1 means semaphore acuqire */ + tmp = RREG32_RLC_NO_KIQ(hub->vm_inv_eng0_sem + + hub->eng_distance * eng, hub_ip); + if (tmp & 0x1) + break; + udelay(1); + } + + if (i >= adev->usec_timeout) + DRM_ERROR("Timeout waiting for sem acquire in VM flush!\n"); + } + + WREG32_RLC_NO_KIQ(hub->vm_inv_eng0_req + hub->eng_distance * eng, inv_req, hub_ip); + + /* Wait for ACK with a delay.*/ + for (i = 0; i < adev->usec_timeout; i++) { + tmp = RREG32_RLC_NO_KIQ(hub->vm_inv_eng0_ack + + hub->eng_distance * eng, hub_ip); + tmp &= 1 << vmid; + if (tmp) + break; + + udelay(1); + } + + /* TODO: It needs to continue working on debugging with semaphore for GFXHUB as well. */ + if (use_semaphore) + /* + * add semaphore release after invalidation, + * write with 0 means semaphore release + */ + WREG32_RLC_NO_KIQ(hub->vm_inv_eng0_sem + + hub->eng_distance * eng, 0, hub_ip); + + /* Issue additional private vm invalidation to MMHUB */ + if ((vmhub != AMDGPU_GFXHUB_0) && + (hub->vm_l2_bank_select_reserved_cid2) && + !amdgpu_sriov_vf(adev)) { + inv_req = RREG32_NO_KIQ(hub->vm_l2_bank_select_reserved_cid2); + /* bit 25: RSERVED_CACHE_PRIVATE_INVALIDATION */ + inv_req |= (1 << 25); + /* Issue private invalidation */ + WREG32_NO_KIQ(hub->vm_l2_bank_select_reserved_cid2, inv_req); + /* Read back to ensure invalidation is done*/ + RREG32_NO_KIQ(hub->vm_l2_bank_select_reserved_cid2); + } + + spin_unlock(&adev->gmc.invalidate_lock); + + if (i < adev->usec_timeout) + return; + + DRM_ERROR("Timeout waiting for VM flush ACK!\n"); +} + +/** + * gmc_v11_0_flush_gpu_tlb - gart tlb flush callback + * + * @adev: amdgpu_device pointer + * @vmid: vm instance to flush + * + * Flush the TLB for the requested page table. + */ +static void gmc_v11_0_flush_gpu_tlb(struct amdgpu_device *adev, uint32_t vmid, + uint32_t vmhub, uint32_t flush_type) +{ + if ((vmhub == AMDGPU_GFXHUB_0) && !adev->gfx.is_poweron) + return; + + /* flush hdp cache */ + adev->hdp.funcs->flush_hdp(adev, NULL); + + /* For SRIOV run time, driver shouldn't access the register through MMIO + * Directly use kiq to do the vm invalidation instead + */ + if ((adev->gfx.kiq.ring.sched.ready || adev->mes.ring.sched.ready) && + (amdgpu_sriov_runtime(adev) || !amdgpu_sriov_vf(adev))) { + struct amdgpu_vmhub *hub = &adev->vmhub[vmhub]; + const unsigned eng = 17; + u32 inv_req = hub->vmhub_funcs->get_invalidate_req(vmid, flush_type); + u32 req = hub->vm_inv_eng0_req + hub->eng_distance * eng; + u32 ack = hub->vm_inv_eng0_ack + hub->eng_distance * eng; + + amdgpu_virt_kiq_reg_write_reg_wait(adev, req, ack, inv_req, + 1 << vmid); + return; + } + + mutex_lock(&adev->mman.gtt_window_lock); + gmc_v11_0_flush_vm_hub(adev, vmid, vmhub, 0); + mutex_unlock(&adev->mman.gtt_window_lock); + return; +} + +/** + * gmc_v11_0_flush_gpu_tlb_pasid - tlb flush via pasid + * + * @adev: amdgpu_device pointer + * @pasid: pasid to be flush + * + * Flush the TLB for the requested pasid. + */ +static int gmc_v11_0_flush_gpu_tlb_pasid(struct amdgpu_device *adev, + uint16_t pasid, uint32_t flush_type, + bool all_hub) +{ + int vmid, i; + signed long r; + uint32_t seq; + uint16_t queried_pasid; + bool ret; + struct amdgpu_ring *ring = &adev->gfx.kiq.ring; + struct amdgpu_kiq *kiq = &adev->gfx.kiq; + + if (amdgpu_emu_mode == 0 && ring->sched.ready) { + spin_lock(&adev->gfx.kiq.ring_lock); + /* 2 dwords flush + 8 dwords fence */ + amdgpu_ring_alloc(ring, kiq->pmf->invalidate_tlbs_size + 8); + kiq->pmf->kiq_invalidate_tlbs(ring, + pasid, flush_type, all_hub); + r = amdgpu_fence_emit_polling(ring, &seq, MAX_KIQ_REG_WAIT); + if (r) { + amdgpu_ring_undo(ring); + spin_unlock(&adev->gfx.kiq.ring_lock); + return -ETIME; + } + + amdgpu_ring_commit(ring); + spin_unlock(&adev->gfx.kiq.ring_lock); + r = amdgpu_fence_wait_polling(ring, seq, adev->usec_timeout); + if (r < 1) { + dev_err(adev->dev, "wait for kiq fence error: %ld.\n", r); + return -ETIME; + } + + return 0; + } + + for (vmid = 1; vmid < 16; vmid++) { + + ret = gmc_v11_0_get_vmid_pasid_mapping_info(adev, vmid, + &queried_pasid); + if (ret && queried_pasid == pasid) { + if (all_hub) { + for (i = 0; i < adev->num_vmhubs; i++) + gmc_v11_0_flush_gpu_tlb(adev, vmid, + i, flush_type); + } else { + gmc_v11_0_flush_gpu_tlb(adev, vmid, + AMDGPU_GFXHUB_0, flush_type); + } + } + } + + return 0; +} + +static uint64_t gmc_v11_0_emit_flush_gpu_tlb(struct amdgpu_ring *ring, + unsigned vmid, uint64_t pd_addr) +{ + bool use_semaphore = gmc_v11_0_use_invalidate_semaphore(ring->adev, ring->funcs->vmhub); + struct amdgpu_vmhub *hub = &ring->adev->vmhub[ring->funcs->vmhub]; + uint32_t req = hub->vmhub_funcs->get_invalidate_req(vmid, 0); + unsigned eng = ring->vm_inv_eng; + + /* + * It may lose gpuvm invalidate acknowldege state across power-gating + * off cycle, add semaphore acquire before invalidation and semaphore + * release after invalidation to avoid entering power gated state + * to WA the Issue + */ + + /* TODO: It needs to continue working on debugging with semaphore for GFXHUB as well. */ + if (use_semaphore) + /* a read return value of 1 means semaphore acuqire */ + amdgpu_ring_emit_reg_wait(ring, + hub->vm_inv_eng0_sem + + hub->eng_distance * eng, 0x1, 0x1); + + amdgpu_ring_emit_wreg(ring, hub->ctx0_ptb_addr_lo32 + + (hub->ctx_addr_distance * vmid), + lower_32_bits(pd_addr)); + + amdgpu_ring_emit_wreg(ring, hub->ctx0_ptb_addr_hi32 + + (hub->ctx_addr_distance * vmid), + upper_32_bits(pd_addr)); + + amdgpu_ring_emit_reg_write_reg_wait(ring, hub->vm_inv_eng0_req + + hub->eng_distance * eng, + hub->vm_inv_eng0_ack + + hub->eng_distance * eng, + req, 1 << vmid); + + /* TODO: It needs to continue working on debugging with semaphore for GFXHUB as well. */ + if (use_semaphore) + /* + * add semaphore release after invalidation, + * write with 0 means semaphore release + */ + amdgpu_ring_emit_wreg(ring, hub->vm_inv_eng0_sem + + hub->eng_distance * eng, 0); + + return pd_addr; +} + +static void gmc_v11_0_emit_pasid_mapping(struct amdgpu_ring *ring, unsigned vmid, + unsigned pasid) +{ + struct amdgpu_device *adev = ring->adev; + uint32_t reg; + + /* MES fw manages IH_VMID_x_LUT updating */ + if (ring->is_mes_queue) + return; + + if (ring->funcs->vmhub == AMDGPU_GFXHUB_0) + reg = SOC15_REG_OFFSET(OSSSYS, 0, regIH_VMID_0_LUT) + vmid; + else + reg = SOC15_REG_OFFSET(OSSSYS, 0, regIH_VMID_0_LUT_MM) + vmid; + + amdgpu_ring_emit_wreg(ring, reg, pasid); +} + +/* + * PTE format: + * 63:59 reserved + * 58:57 reserved + * 56 F + * 55 L + * 54 reserved + * 53:52 SW + * 51 T + * 50:48 mtype + * 47:12 4k physical page base address + * 11:7 fragment + * 6 write + * 5 read + * 4 exe + * 3 Z + * 2 snooped + * 1 system + * 0 valid + * + * PDE format: + * 63:59 block fragment size + * 58:55 reserved + * 54 P + * 53:48 reserved + * 47:6 physical base address of PD or PTE + * 5:3 reserved + * 2 C + * 1 system + * 0 valid + */ + +static uint64_t gmc_v11_0_map_mtype(struct amdgpu_device *adev, uint32_t flags) +{ + switch (flags) { + case AMDGPU_VM_MTYPE_DEFAULT: + return AMDGPU_PTE_MTYPE_NV10(MTYPE_NC); + case AMDGPU_VM_MTYPE_NC: + return AMDGPU_PTE_MTYPE_NV10(MTYPE_NC); + case AMDGPU_VM_MTYPE_WC: + return AMDGPU_PTE_MTYPE_NV10(MTYPE_WC); + case AMDGPU_VM_MTYPE_CC: + return AMDGPU_PTE_MTYPE_NV10(MTYPE_CC); + case AMDGPU_VM_MTYPE_UC: + return AMDGPU_PTE_MTYPE_NV10(MTYPE_UC); + default: + return AMDGPU_PTE_MTYPE_NV10(MTYPE_NC); + } +} + +static void gmc_v11_0_get_vm_pde(struct amdgpu_device *adev, int level, + uint64_t *addr, uint64_t *flags) +{ + if (!(*flags & AMDGPU_PDE_PTE) && !(*flags & AMDGPU_PTE_SYSTEM)) + *addr = adev->vm_manager.vram_base_offset + *addr - + adev->gmc.vram_start; + BUG_ON(*addr & 0xFFFF00000000003FULL); + + if (!adev->gmc.translate_further) + return; + + if (level == AMDGPU_VM_PDB1) { + /* Set the block fragment size */ + if (!(*flags & AMDGPU_PDE_PTE)) + *flags |= AMDGPU_PDE_BFS(0x9); + + } else if (level == AMDGPU_VM_PDB0) { + if (*flags & AMDGPU_PDE_PTE) + *flags &= ~AMDGPU_PDE_PTE; + else + *flags |= AMDGPU_PTE_TF; + } +} + +static void gmc_v11_0_get_vm_pte(struct amdgpu_device *adev, + struct amdgpu_bo_va_mapping *mapping, + uint64_t *flags) +{ + *flags &= ~AMDGPU_PTE_EXECUTABLE; + *flags |= mapping->flags & AMDGPU_PTE_EXECUTABLE; + + *flags &= ~AMDGPU_PTE_MTYPE_NV10_MASK; + *flags |= (mapping->flags & AMDGPU_PTE_MTYPE_NV10_MASK); + + *flags &= ~AMDGPU_PTE_NOALLOC; + *flags |= (mapping->flags & AMDGPU_PTE_NOALLOC); + + if (mapping->flags & AMDGPU_PTE_PRT) { + *flags |= AMDGPU_PTE_PRT; + *flags |= AMDGPU_PTE_SNOOPED; + *flags |= AMDGPU_PTE_LOG; + *flags |= AMDGPU_PTE_SYSTEM; + *flags &= ~AMDGPU_PTE_VALID; + } +} + +static unsigned gmc_v11_0_get_vbios_fb_size(struct amdgpu_device *adev) +{ + return 0; +} + +static const struct amdgpu_gmc_funcs gmc_v11_0_gmc_funcs = { + .flush_gpu_tlb = gmc_v11_0_flush_gpu_tlb, + .flush_gpu_tlb_pasid = gmc_v11_0_flush_gpu_tlb_pasid, + .emit_flush_gpu_tlb = gmc_v11_0_emit_flush_gpu_tlb, + .emit_pasid_mapping = gmc_v11_0_emit_pasid_mapping, + .map_mtype = gmc_v11_0_map_mtype, + .get_vm_pde = gmc_v11_0_get_vm_pde, + .get_vm_pte = gmc_v11_0_get_vm_pte, + .get_vbios_fb_size = gmc_v11_0_get_vbios_fb_size, +}; + +static void gmc_v11_0_set_gmc_funcs(struct amdgpu_device *adev) +{ + adev->gmc.gmc_funcs = &gmc_v11_0_gmc_funcs; +} + +static void gmc_v11_0_set_umc_funcs(struct amdgpu_device *adev) +{ + switch (adev->ip_versions[UMC_HWIP][0]) { + case IP_VERSION(8, 10, 0): + adev->umc.channel_inst_num = UMC_V8_10_CHANNEL_INSTANCE_NUM; + adev->umc.umc_inst_num = UMC_V8_10_UMC_INSTANCE_NUM; + adev->umc.node_inst_num = adev->gmc.num_umc; + adev->umc.max_ras_err_cnt_per_query = UMC_V8_10_TOTAL_CHANNEL_NUM(adev); + adev->umc.channel_offs = UMC_V8_10_PER_CHANNEL_OFFSET; + adev->umc.channel_idx_tbl = &umc_v8_10_channel_idx_tbl[0][0][0]; + adev->umc.ras = &umc_v8_10_ras; + break; + case IP_VERSION(8, 11, 0): + break; + default: + break; + } + + if (adev->umc.ras) { + amdgpu_ras_register_ras_block(adev, &adev->umc.ras->ras_block); + + strcpy(adev->umc.ras->ras_block.ras_comm.name, "umc"); + adev->umc.ras->ras_block.ras_comm.block = AMDGPU_RAS_BLOCK__UMC; + adev->umc.ras->ras_block.ras_comm.type = AMDGPU_RAS_ERROR__MULTI_UNCORRECTABLE; + adev->umc.ras_if = &adev->umc.ras->ras_block.ras_comm; + + /* If don't define special ras_late_init function, use default ras_late_init */ + if (!adev->umc.ras->ras_block.ras_late_init) + adev->umc.ras->ras_block.ras_late_init = amdgpu_umc_ras_late_init; + + /* If not define special ras_cb function, use default ras_cb */ + if (!adev->umc.ras->ras_block.ras_cb) + adev->umc.ras->ras_block.ras_cb = amdgpu_umc_process_ras_data_cb; + } +} + + +static void gmc_v11_0_set_mmhub_funcs(struct amdgpu_device *adev) +{ + switch (adev->ip_versions[MMHUB_HWIP][0]) { + case IP_VERSION(3, 0, 1): + adev->mmhub.funcs = &mmhub_v3_0_1_funcs; + break; + case IP_VERSION(3, 0, 2): + adev->mmhub.funcs = &mmhub_v3_0_2_funcs; + break; + default: + adev->mmhub.funcs = &mmhub_v3_0_funcs; + break; + } +} + +static void gmc_v11_0_set_gfxhub_funcs(struct amdgpu_device *adev) +{ + switch (adev->ip_versions[GC_HWIP][0]) { + case IP_VERSION(11, 0, 3): + adev->gfxhub.funcs = &gfxhub_v3_0_3_funcs; + break; + default: + adev->gfxhub.funcs = &gfxhub_v3_0_funcs; + break; + } +} + +static int gmc_v11_0_early_init(void *handle) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + + gmc_v11_0_set_gfxhub_funcs(adev); + gmc_v11_0_set_mmhub_funcs(adev); + gmc_v11_0_set_gmc_funcs(adev); + gmc_v11_0_set_irq_funcs(adev); + gmc_v11_0_set_umc_funcs(adev); + + adev->gmc.shared_aperture_start = 0x2000000000000000ULL; + adev->gmc.shared_aperture_end = + adev->gmc.shared_aperture_start + (4ULL << 30) - 1; + adev->gmc.private_aperture_start = 0x1000000000000000ULL; + adev->gmc.private_aperture_end = + adev->gmc.private_aperture_start + (4ULL << 30) - 1; + + return 0; +} + +static int gmc_v11_0_late_init(void *handle) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + int r; + + r = amdgpu_gmc_allocate_vm_inv_eng(adev); + if (r) + return r; + + r = amdgpu_gmc_ras_late_init(adev); + if (r) + return r; + + return amdgpu_irq_get(adev, &adev->gmc.vm_fault, 0); +} + +static void gmc_v11_0_vram_gtt_location(struct amdgpu_device *adev, + struct amdgpu_gmc *mc) +{ + u64 base = 0; + + base = adev->mmhub.funcs->get_fb_location(adev); + + amdgpu_gmc_vram_location(adev, &adev->gmc, base); + amdgpu_gmc_gart_location(adev, mc); + + /* base offset of vram pages */ + if (amdgpu_sriov_vf(adev)) + adev->vm_manager.vram_base_offset = 0; + else + adev->vm_manager.vram_base_offset = adev->mmhub.funcs->get_mc_fb_offset(adev); +} + +/** + * gmc_v11_0_mc_init - initialize the memory controller driver params + * + * @adev: amdgpu_device pointer + * + * Look up the amount of vram, vram width, and decide how to place + * vram and gart within the GPU's physical address space. + * Returns 0 for success. + */ +static int gmc_v11_0_mc_init(struct amdgpu_device *adev) +{ + int r; + + /* size in MB on si */ + adev->gmc.mc_vram_size = + adev->nbio.funcs->get_memsize(adev) * 1024ULL * 1024ULL; + adev->gmc.real_vram_size = adev->gmc.mc_vram_size; + + if (!(adev->flags & AMD_IS_APU)) { + r = amdgpu_device_resize_fb_bar(adev); + if (r) + return r; + } + adev->gmc.aper_base = pci_resource_start(adev->pdev, 0); + adev->gmc.aper_size = pci_resource_len(adev->pdev, 0); + +#ifdef CONFIG_X86_64 + if ((adev->flags & AMD_IS_APU) && !amdgpu_passthrough(adev)) { + adev->gmc.aper_base = adev->mmhub.funcs->get_mc_fb_offset(adev); + adev->gmc.aper_size = adev->gmc.real_vram_size; + } +#endif + /* In case the PCI BAR is larger than the actual amount of vram */ + adev->gmc.visible_vram_size = adev->gmc.aper_size; + if (adev->gmc.visible_vram_size > adev->gmc.real_vram_size) + adev->gmc.visible_vram_size = adev->gmc.real_vram_size; + + /* set the gart size */ + if (amdgpu_gart_size == -1) { + adev->gmc.gart_size = 512ULL << 20; + } else + adev->gmc.gart_size = (u64)amdgpu_gart_size << 20; + + gmc_v11_0_vram_gtt_location(adev, &adev->gmc); + + return 0; +} + +static int gmc_v11_0_gart_init(struct amdgpu_device *adev) +{ + int r; + + if (adev->gart.bo) { + WARN(1, "PCIE GART already initialized\n"); + return 0; + } + + /* Initialize common gart structure */ + r = amdgpu_gart_init(adev); + if (r) + return r; + + adev->gart.table_size = adev->gart.num_gpu_pages * 8; + adev->gart.gart_pte_flags = AMDGPU_PTE_MTYPE_NV10(MTYPE_UC) | + AMDGPU_PTE_EXECUTABLE; + + return amdgpu_gart_table_vram_alloc(adev); +} + +static int gmc_v11_0_sw_init(void *handle) +{ + int r, vram_width = 0, vram_type = 0, vram_vendor = 0; + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + + adev->mmhub.funcs->init(adev); + + spin_lock_init(&adev->gmc.invalidate_lock); + + r = amdgpu_atomfirmware_get_vram_info(adev, + &vram_width, &vram_type, &vram_vendor); + adev->gmc.vram_width = vram_width; + + adev->gmc.vram_type = vram_type; + adev->gmc.vram_vendor = vram_vendor; + + switch (adev->ip_versions[GC_HWIP][0]) { + case IP_VERSION(11, 0, 0): + case IP_VERSION(11, 0, 1): + case IP_VERSION(11, 0, 2): + case IP_VERSION(11, 0, 3): + adev->num_vmhubs = 2; + /* + * To fulfill 4-level page support, + * vm size is 256TB (48bit), maximum size, + * block size 512 (9bit) + */ + amdgpu_vm_adjust_size(adev, 256 * 1024, 9, 3, 48); + break; + default: + break; + } + + /* This interrupt is VMC page fault.*/ + r = amdgpu_irq_add_id(adev, SOC21_IH_CLIENTID_VMC, + VMC_1_0__SRCID__VM_FAULT, + &adev->gmc.vm_fault); + + if (r) + return r; + + r = amdgpu_irq_add_id(adev, SOC21_IH_CLIENTID_GFX, + UTCL2_1_0__SRCID__FAULT, + &adev->gmc.vm_fault); + if (r) + return r; + + if (!amdgpu_sriov_vf(adev)) { + /* interrupt sent to DF. */ + r = amdgpu_irq_add_id(adev, SOC21_IH_CLIENTID_DF, 0, + &adev->gmc.ecc_irq); + if (r) + return r; + } + + /* + * Set the internal MC address mask This is the max address of the GPU's + * internal address space. + */ + adev->gmc.mc_mask = 0xffffffffffffULL; /* 48 bit MC */ + + r = dma_set_mask_and_coherent(adev->dev, DMA_BIT_MASK(44)); + if (r) { + printk(KERN_WARNING "amdgpu: No suitable DMA available.\n"); + return r; + } + + adev->need_swiotlb = drm_need_swiotlb(44); + + r = gmc_v11_0_mc_init(adev); + if (r) + return r; + + amdgpu_gmc_get_vbios_allocations(adev); + + /* Memory manager */ + r = amdgpu_bo_init(adev); + if (r) + return r; + + r = gmc_v11_0_gart_init(adev); + if (r) + return r; + + /* + * number of VMs + * VMID 0 is reserved for System + * amdgpu graphics/compute will use VMIDs 1-7 + * amdkfd will use VMIDs 8-15 + */ + adev->vm_manager.first_kfd_vmid = 8; + + amdgpu_vm_manager_init(adev); + + return 0; +} + +/** + * gmc_v11_0_gart_fini - vm fini callback + * + * @adev: amdgpu_device pointer + * + * Tears down the driver GART/VM setup (CIK). + */ +static void gmc_v11_0_gart_fini(struct amdgpu_device *adev) +{ + amdgpu_gart_table_vram_free(adev); +} + +static int gmc_v11_0_sw_fini(void *handle) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + + amdgpu_vm_manager_fini(adev); + gmc_v11_0_gart_fini(adev); + amdgpu_gem_force_release(adev); + amdgpu_bo_fini(adev); + + return 0; +} + +static void gmc_v11_0_init_golden_registers(struct amdgpu_device *adev) +{ +} + +/** + * gmc_v11_0_gart_enable - gart enable + * + * @adev: amdgpu_device pointer + */ +static int gmc_v11_0_gart_enable(struct amdgpu_device *adev) +{ + int r; + bool value; + + if (adev->gart.bo == NULL) { + dev_err(adev->dev, "No VRAM object for PCIE GART.\n"); + return -EINVAL; + } + + amdgpu_gtt_mgr_recover(&adev->mman.gtt_mgr); + + r = adev->mmhub.funcs->gart_enable(adev); + if (r) + return r; + + /* Flush HDP after it is initialized */ + adev->hdp.funcs->flush_hdp(adev, NULL); + + value = (amdgpu_vm_fault_stop == AMDGPU_VM_FAULT_STOP_ALWAYS) ? + false : true; + + adev->mmhub.funcs->set_fault_enable_default(adev, value); + gmc_v11_0_flush_gpu_tlb(adev, 0, AMDGPU_MMHUB_0, 0); + + DRM_INFO("PCIE GART of %uM enabled (table at 0x%016llX).\n", + (unsigned)(adev->gmc.gart_size >> 20), + (unsigned long long)amdgpu_bo_gpu_offset(adev->gart.bo)); + + return 0; +} + +static int gmc_v11_0_hw_init(void *handle) +{ + int r; + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + + /* The sequence of these two function calls matters.*/ + gmc_v11_0_init_golden_registers(adev); + + r = gmc_v11_0_gart_enable(adev); + if (r) + return r; + + if (adev->umc.funcs && adev->umc.funcs->init_registers) + adev->umc.funcs->init_registers(adev); + + return 0; +} + +/** + * gmc_v11_0_gart_disable - gart disable + * + * @adev: amdgpu_device pointer + * + * This disables all VM page table. + */ +static void gmc_v11_0_gart_disable(struct amdgpu_device *adev) +{ + adev->mmhub.funcs->gart_disable(adev); +} + +static int gmc_v11_0_hw_fini(void *handle) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + + if (amdgpu_sriov_vf(adev)) { + /* full access mode, so don't touch any GMC register */ + DRM_DEBUG("For SRIOV client, shouldn't do anything.\n"); + return 0; + } + + amdgpu_irq_put(adev, &adev->gmc.ecc_irq, 0); + amdgpu_irq_put(adev, &adev->gmc.vm_fault, 0); + gmc_v11_0_gart_disable(adev); + + return 0; +} + +static int gmc_v11_0_suspend(void *handle) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + + gmc_v11_0_hw_fini(adev); + + return 0; +} + +static int gmc_v11_0_resume(void *handle) +{ + int r; + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + + r = gmc_v11_0_hw_init(adev); + if (r) + return r; + + amdgpu_vmid_reset_all(adev); + + return 0; +} + +static bool gmc_v11_0_is_idle(void *handle) +{ + /* MC is always ready in GMC v11.*/ + return true; +} + +static int gmc_v11_0_wait_for_idle(void *handle) +{ + /* There is no need to wait for MC idle in GMC v11.*/ + return 0; +} + +static int gmc_v11_0_soft_reset(void *handle) +{ + return 0; +} + +static int gmc_v11_0_set_clockgating_state(void *handle, + enum amd_clockgating_state state) +{ + int r; + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + + r = adev->mmhub.funcs->set_clockgating(adev, state); + if (r) + return r; + + return athub_v3_0_set_clockgating(adev, state); +} + +static void gmc_v11_0_get_clockgating_state(void *handle, u64 *flags) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + + adev->mmhub.funcs->get_clockgating(adev, flags); + + athub_v3_0_get_clockgating(adev, flags); +} + +static int gmc_v11_0_set_powergating_state(void *handle, + enum amd_powergating_state state) +{ + return 0; +} + +const struct amd_ip_funcs gmc_v11_0_ip_funcs = { + .name = "gmc_v11_0", + .early_init = gmc_v11_0_early_init, + .sw_init = gmc_v11_0_sw_init, + .hw_init = gmc_v11_0_hw_init, + .late_init = gmc_v11_0_late_init, + .sw_fini = gmc_v11_0_sw_fini, + .hw_fini = gmc_v11_0_hw_fini, + .suspend = gmc_v11_0_suspend, + .resume = gmc_v11_0_resume, + .is_idle = gmc_v11_0_is_idle, + .wait_for_idle = gmc_v11_0_wait_for_idle, + .soft_reset = gmc_v11_0_soft_reset, + .set_clockgating_state = gmc_v11_0_set_clockgating_state, + .set_powergating_state = gmc_v11_0_set_powergating_state, + .get_clockgating_state = gmc_v11_0_get_clockgating_state, +}; + +const struct amdgpu_ip_block_version gmc_v11_0_ip_block = { + .type = AMD_IP_BLOCK_TYPE_GMC, + .major = 11, + .minor = 0, + .rev = 0, + .funcs = &gmc_v11_0_ip_funcs, +}; diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v11_0.h b/drivers/gpu/drm/amd/amdgpu/gmc_v11_0.h new file mode 100644 index 000000000000..def4d5516f82 --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/gmc_v11_0.h @@ -0,0 +1,30 @@ +/* + * Copyright 2021 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#ifndef __GMC_V11_0_H__ +#define __GMC_V11_0_H__ + +extern const struct amd_ip_funcs gmc_v11_0_ip_funcs; +extern const struct amdgpu_ip_block_version gmc_v11_0_ip_block; + +#endif diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v6_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v6_0.c index 0fe714f54cca..ec291d28edff 100644 --- a/drivers/gpu/drm/amd/amdgpu/gmc_v6_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gmc_v6_0.c @@ -469,16 +469,14 @@ static void gmc_v6_0_set_prt(struct amdgpu_device *adev, bool enable) static int gmc_v6_0_gart_enable(struct amdgpu_device *adev) { uint64_t table_addr; - int r, i; u32 field; + int i; if (adev->gart.bo == NULL) { dev_err(adev->dev, "No VRAM object for PCIE GART.\n"); return -EINVAL; } - r = amdgpu_gart_table_vram_pin(adev); - if (r) - return r; + amdgpu_gtt_mgr_recover(&adev->mman.gtt_mgr); table_addr = amdgpu_bo_gpu_offset(adev->gart.bo); @@ -558,7 +556,6 @@ static int gmc_v6_0_gart_enable(struct amdgpu_device *adev) dev_info(adev->dev, "PCIE GART of %uM enabled (table at 0x%016llX).\n", (unsigned)(adev->gmc.gart_size >> 20), (unsigned long long)table_addr); - adev->gart.ready = true; return 0; } @@ -608,7 +605,6 @@ static void gmc_v6_0_gart_disable(struct amdgpu_device *adev) WREG32(mmVM_L2_CNTL3, VM_L2_CNTL3__L2_CACHE_BIGK_ASSOCIATIVITY_MASK | (0UL << VM_L2_CNTL3__L2_CACHE_BIGK_FRAGMENT_SIZE__SHIFT)); - amdgpu_gart_table_vram_unpin(adev); } static void gmc_v6_0_vm_decode_fault(struct amdgpu_device *adev, @@ -923,7 +919,10 @@ static int gmc_v6_0_hw_init(void *handle) if (r) return r; - return r; + if (amdgpu_emu_mode == 1) + return amdgpu_gmc_vram_checking(adev); + else + return r; } static int gmc_v6_0_hw_fini(void *handle) diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v7_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v7_0.c index 0a50fdaced7e..979da6f510e8 100644 --- a/drivers/gpu/drm/amd/amdgpu/gmc_v7_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gmc_v7_0.c @@ -381,8 +381,9 @@ static int gmc_v7_0_mc_init(struct amdgpu_device *adev) adev->gmc.aper_size = pci_resource_len(adev->pdev, 0); #ifdef CONFIG_X86_64 - if (adev->flags & AMD_IS_APU && - adev->gmc.real_vram_size > adev->gmc.aper_size) { + if ((adev->flags & AMD_IS_APU) && + adev->gmc.real_vram_size > adev->gmc.aper_size && + !amdgpu_passthrough(adev)) { adev->gmc.aper_base = ((u64)RREG32(mmMC_VM_FB_OFFSET)) << 22; adev->gmc.aper_size = adev->gmc.real_vram_size; } @@ -613,17 +614,14 @@ static void gmc_v7_0_set_prt(struct amdgpu_device *adev, bool enable) static int gmc_v7_0_gart_enable(struct amdgpu_device *adev) { uint64_t table_addr; - int r, i; u32 tmp, field; + int i; if (adev->gart.bo == NULL) { dev_err(adev->dev, "No VRAM object for PCIE GART.\n"); return -EINVAL; } - r = amdgpu_gart_table_vram_pin(adev); - if (r) - return r; - + amdgpu_gtt_mgr_recover(&adev->mman.gtt_mgr); table_addr = amdgpu_bo_gpu_offset(adev->gart.bo); /* Setup TLB control */ @@ -712,7 +710,6 @@ static int gmc_v7_0_gart_enable(struct amdgpu_device *adev) DRM_INFO("PCIE GART of %uM enabled (table at 0x%016llX).\n", (unsigned)(adev->gmc.gart_size >> 20), (unsigned long long)table_addr); - adev->gart.ready = true; return 0; } @@ -758,7 +755,6 @@ static void gmc_v7_0_gart_disable(struct amdgpu_device *adev) tmp = REG_SET_FIELD(tmp, VM_L2_CNTL, ENABLE_L2_CACHE, 0); WREG32(mmVM_L2_CNTL, tmp); WREG32(mmVM_L2_CNTL2, 0); - amdgpu_gart_table_vram_unpin(adev); } /** @@ -1112,7 +1108,10 @@ static int gmc_v7_0_hw_init(void *handle) if (r) return r; - return r; + if (amdgpu_emu_mode == 1) + return amdgpu_gmc_vram_checking(adev); + else + return r; } static int gmc_v7_0_hw_fini(void *handle) diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c index 63b890f1e8af..382dde1ce74c 100644 --- a/drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c @@ -581,7 +581,7 @@ static int gmc_v8_0_mc_init(struct amdgpu_device *adev) adev->gmc.aper_size = pci_resource_len(adev->pdev, 0); #ifdef CONFIG_X86_64 - if (adev->flags & AMD_IS_APU) { + if ((adev->flags & AMD_IS_APU) && !amdgpu_passthrough(adev)) { adev->gmc.aper_base = ((u64)RREG32(mmMC_VM_FB_OFFSET)) << 22; adev->gmc.aper_size = adev->gmc.real_vram_size; } @@ -837,17 +837,14 @@ static void gmc_v8_0_set_prt(struct amdgpu_device *adev, bool enable) static int gmc_v8_0_gart_enable(struct amdgpu_device *adev) { uint64_t table_addr; - int r, i; u32 tmp, field; + int i; if (adev->gart.bo == NULL) { dev_err(adev->dev, "No VRAM object for PCIE GART.\n"); return -EINVAL; } - r = amdgpu_gart_table_vram_pin(adev); - if (r) - return r; - + amdgpu_gtt_mgr_recover(&adev->mman.gtt_mgr); table_addr = amdgpu_bo_gpu_offset(adev->gart.bo); /* Setup TLB control */ @@ -953,7 +950,6 @@ static int gmc_v8_0_gart_enable(struct amdgpu_device *adev) DRM_INFO("PCIE GART of %uM enabled (table at 0x%016llX).\n", (unsigned)(adev->gmc.gart_size >> 20), (unsigned long long)table_addr); - adev->gart.ready = true; return 0; } @@ -999,7 +995,6 @@ static void gmc_v8_0_gart_disable(struct amdgpu_device *adev) tmp = REG_SET_FIELD(tmp, VM_L2_CNTL, ENABLE_L2_CACHE, 0); WREG32(mmVM_L2_CNTL, tmp); WREG32(mmVM_L2_CNTL2, 0); - amdgpu_gart_table_vram_unpin(adev); } /** @@ -1243,7 +1238,10 @@ static int gmc_v8_0_hw_init(void *handle) if (r) return r; - return r; + if (amdgpu_emu_mode == 1) + return amdgpu_gmc_vram_checking(adev); + else + return r; } static int gmc_v8_0_hw_fini(void *handle) @@ -1692,7 +1690,7 @@ static int gmc_v8_0_set_powergating_state(void *handle, return 0; } -static void gmc_v8_0_get_clockgating_state(void *handle, u32 *flags) +static void gmc_v8_0_get_clockgating_state(void *handle, u64 *flags) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; int data; diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c index 57f2729a7bd0..67ca16a8027c 100644 --- a/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c @@ -62,6 +62,8 @@ #include "amdgpu_ras.h" #include "amdgpu_xgmi.h" +#include "amdgpu_reset.h" + /* add these here since we already include dce12 headers and these are for DCN */ #define mmHUBP0_DCSURF_PRI_VIEWPORT_DIMENSION 0x055d #define mmHUBP0_DCSURF_PRI_VIEWPORT_DIMENSION_BASE_IDX 2 @@ -72,6 +74,9 @@ #define mmDCHUBBUB_SDPIF_MMIO_CNTRL_0 0x049d #define mmDCHUBBUB_SDPIF_MMIO_CNTRL_0_BASE_IDX 2 +#define mmHUBP0_DCSURF_PRI_VIEWPORT_DIMENSION_DCN2 0x05ea +#define mmHUBP0_DCSURF_PRI_VIEWPORT_DIMENSION_DCN2_BASE_IDX 2 + static const char *gfxhub_client_ids[] = { "CB", @@ -784,13 +789,13 @@ static void gmc_v9_0_flush_gpu_tlb(struct amdgpu_device *adev, uint32_t vmid, */ if (adev->gfx.kiq.ring.sched.ready && (amdgpu_sriov_runtime(adev) || !amdgpu_sriov_vf(adev)) && - down_read_trylock(&adev->reset_sem)) { + down_read_trylock(&adev->reset_domain->sem)) { uint32_t req = hub->vm_inv_eng0_req + hub->eng_distance * eng; uint32_t ack = hub->vm_inv_eng0_ack + hub->eng_distance * eng; amdgpu_virt_kiq_reg_write_reg_wait(adev, req, ack, inv_req, 1 << vmid); - up_read(&adev->reset_sem); + up_read(&adev->reset_domain->sem); return; } @@ -891,13 +896,14 @@ static int gmc_v9_0_flush_gpu_tlb_pasid(struct amdgpu_device *adev, uint32_t seq; uint16_t queried_pasid; bool ret; + u32 usec_timeout = amdgpu_sriov_vf(adev) ? SRIOV_USEC_TIMEOUT : adev->usec_timeout; struct amdgpu_ring *ring = &adev->gfx.kiq.ring; struct amdgpu_kiq *kiq = &adev->gfx.kiq; if (amdgpu_in_reset(adev)) return -EIO; - if (ring->sched.ready && down_read_trylock(&adev->reset_sem)) { + if (ring->sched.ready && down_read_trylock(&adev->reset_domain->sem)) { /* Vega20+XGMI caches PTEs in TC and TLB. Add a * heavy-weight TLB flush (type 2), which flushes * both. Due to a race condition with concurrent @@ -924,19 +930,19 @@ static int gmc_v9_0_flush_gpu_tlb_pasid(struct amdgpu_device *adev, if (r) { amdgpu_ring_undo(ring); spin_unlock(&adev->gfx.kiq.ring_lock); - up_read(&adev->reset_sem); + up_read(&adev->reset_domain->sem); return -ETIME; } amdgpu_ring_commit(ring); spin_unlock(&adev->gfx.kiq.ring_lock); - r = amdgpu_fence_wait_polling(ring, seq, adev->usec_timeout); + r = amdgpu_fence_wait_polling(ring, seq, usec_timeout); if (r < 1) { dev_err(adev->dev, "wait for kiq fence error: %ld.\n", r); - up_read(&adev->reset_sem); + up_read(&adev->reset_domain->sem); return -ETIME; } - up_read(&adev->reset_sem); + up_read(&adev->reset_domain->sem); return 0; } @@ -1097,10 +1103,13 @@ static void gmc_v9_0_get_vm_pde(struct amdgpu_device *adev, int level, *flags |= AMDGPU_PDE_BFS(0x9); } else if (level == AMDGPU_VM_PDB0) { - if (*flags & AMDGPU_PDE_PTE) + if (*flags & AMDGPU_PDE_PTE) { *flags &= ~AMDGPU_PDE_PTE; - else + if (!(*flags & AMDGPU_PTE_VALID)) + *addr |= 1 << PAGE_SHIFT; + } else { *flags |= AMDGPU_PTE_TF; + } } } @@ -1134,6 +1143,8 @@ static unsigned gmc_v9_0_get_vbios_fb_size(struct amdgpu_device *adev) u32 d1vga_control = RREG32_SOC15(DCE, 0, mmD1VGA_CONTROL); unsigned size; + /* TODO move to DC so GMC doesn't need to hard-code DCN registers */ + if (REG_GET_FIELD(d1vga_control, D1VGA_CONTROL, D1VGA_MODE_ENABLE)) { size = AMDGPU_VBIOS_VGA_ALLOCATION; } else { @@ -1142,7 +1153,6 @@ static unsigned gmc_v9_0_get_vbios_fb_size(struct amdgpu_device *adev) switch (adev->ip_versions[DCE_HWIP][0]) { case IP_VERSION(1, 0, 0): case IP_VERSION(1, 0, 1): - case IP_VERSION(2, 1, 0): viewport = RREG32_SOC15(DCE, 0, mmHUBP0_DCSURF_PRI_VIEWPORT_DIMENSION); size = (REG_GET_FIELD(viewport, HUBP0_DCSURF_PRI_VIEWPORT_DIMENSION, PRI_VIEWPORT_HEIGHT) * @@ -1150,6 +1160,14 @@ static unsigned gmc_v9_0_get_vbios_fb_size(struct amdgpu_device *adev) HUBP0_DCSURF_PRI_VIEWPORT_DIMENSION, PRI_VIEWPORT_WIDTH) * 4); break; + case IP_VERSION(2, 1, 0): + viewport = RREG32_SOC15(DCE, 0, mmHUBP0_DCSURF_PRI_VIEWPORT_DIMENSION_DCN2); + size = (REG_GET_FIELD(viewport, + HUBP0_DCSURF_PRI_VIEWPORT_DIMENSION, PRI_VIEWPORT_HEIGHT) * + REG_GET_FIELD(viewport, + HUBP0_DCSURF_PRI_VIEWPORT_DIMENSION, PRI_VIEWPORT_WIDTH) * + 4); + break; default: viewport = RREG32_SOC15(DCE, 0, mmSCL0_VIEWPORT_SIZE); size = (REG_GET_FIELD(viewport, SCL0_VIEWPORT_SIZE, VIEWPORT_HEIGHT) * @@ -1190,7 +1208,7 @@ static void gmc_v9_0_set_umc_funcs(struct amdgpu_device *adev) adev->umc.umc_inst_num = UMC_V6_1_UMC_INSTANCE_NUM; adev->umc.channel_offs = UMC_V6_1_PER_CHANNEL_OFFSET_VG20; adev->umc.channel_idx_tbl = &umc_v6_1_channel_idx_tbl[0][0]; - adev->umc.ras_funcs = &umc_v6_1_ras_funcs; + adev->umc.ras = &umc_v6_1_ras; break; case IP_VERSION(6, 1, 2): adev->umc.max_ras_err_cnt_per_query = UMC_V6_1_TOTAL_CHANNEL_NUM; @@ -1198,15 +1216,16 @@ static void gmc_v9_0_set_umc_funcs(struct amdgpu_device *adev) adev->umc.umc_inst_num = UMC_V6_1_UMC_INSTANCE_NUM; adev->umc.channel_offs = UMC_V6_1_PER_CHANNEL_OFFSET_ARCT; adev->umc.channel_idx_tbl = &umc_v6_1_channel_idx_tbl[0][0]; - adev->umc.ras_funcs = &umc_v6_1_ras_funcs; + adev->umc.ras = &umc_v6_1_ras; break; case IP_VERSION(6, 7, 0): - adev->umc.max_ras_err_cnt_per_query = UMC_V6_7_TOTAL_CHANNEL_NUM; + adev->umc.max_ras_err_cnt_per_query = + UMC_V6_7_TOTAL_CHANNEL_NUM * UMC_V6_7_BAD_PAGE_NUM_PER_CHANNEL; adev->umc.channel_inst_num = UMC_V6_7_CHANNEL_INSTANCE_NUM; adev->umc.umc_inst_num = UMC_V6_7_UMC_INSTANCE_NUM; adev->umc.channel_offs = UMC_V6_7_PER_CHANNEL_OFFSET; if (!adev->gmc.xgmi.connected_to_cpu) - adev->umc.ras_funcs = &umc_v6_7_ras_funcs; + adev->umc.ras = &umc_v6_7_ras; if (1 & adev->smuio.funcs->get_die_id(adev)) adev->umc.channel_idx_tbl = &umc_v6_7_channel_idx_tbl_first[0][0]; else @@ -1215,6 +1234,23 @@ static void gmc_v9_0_set_umc_funcs(struct amdgpu_device *adev) default: break; } + + if (adev->umc.ras) { + amdgpu_ras_register_ras_block(adev, &adev->umc.ras->ras_block); + + strcpy(adev->umc.ras->ras_block.ras_comm.name, "umc"); + adev->umc.ras->ras_block.ras_comm.block = AMDGPU_RAS_BLOCK__UMC; + adev->umc.ras->ras_block.ras_comm.type = AMDGPU_RAS_ERROR__MULTI_UNCORRECTABLE; + adev->umc.ras_if = &adev->umc.ras->ras_block.ras_comm; + + /* If don't define special ras_late_init function, use default ras_late_init */ + if (!adev->umc.ras->ras_block.ras_late_init) + adev->umc.ras->ras_block.ras_late_init = amdgpu_umc_ras_late_init; + + /* If not defined special ras_cb function, use default ras_cb */ + if (!adev->umc.ras->ras_block.ras_cb) + adev->umc.ras->ras_block.ras_cb = amdgpu_umc_process_ras_data_cb; + } } static void gmc_v9_0_set_mmhub_funcs(struct amdgpu_device *adev) @@ -1236,18 +1272,27 @@ static void gmc_v9_0_set_mmhub_ras_funcs(struct amdgpu_device *adev) { switch (adev->ip_versions[MMHUB_HWIP][0]) { case IP_VERSION(9, 4, 0): - adev->mmhub.ras_funcs = &mmhub_v1_0_ras_funcs; + adev->mmhub.ras = &mmhub_v1_0_ras; break; case IP_VERSION(9, 4, 1): - adev->mmhub.ras_funcs = &mmhub_v9_4_ras_funcs; + adev->mmhub.ras = &mmhub_v9_4_ras; break; case IP_VERSION(9, 4, 2): - adev->mmhub.ras_funcs = &mmhub_v1_7_ras_funcs; + adev->mmhub.ras = &mmhub_v1_7_ras; break; default: /* mmhub ras is not available */ break; } + + if (adev->mmhub.ras) { + amdgpu_ras_register_ras_block(adev, &adev->mmhub.ras->ras_block); + + strcpy(adev->mmhub.ras->ras_block.ras_comm.name, "mmhub"); + adev->mmhub.ras->ras_block.ras_comm.block = AMDGPU_RAS_BLOCK__MMHUB; + adev->mmhub.ras->ras_block.ras_comm.type = AMDGPU_RAS_ERROR__MULTI_UNCORRECTABLE; + adev->mmhub.ras_if = &adev->mmhub.ras->ras_block.ras_comm; + } } static void gmc_v9_0_set_gfxhub_funcs(struct amdgpu_device *adev) @@ -1257,7 +1302,9 @@ static void gmc_v9_0_set_gfxhub_funcs(struct amdgpu_device *adev) static void gmc_v9_0_set_hdp_ras_funcs(struct amdgpu_device *adev) { - adev->hdp.ras_funcs = &hdp_v4_0_ras_funcs; + adev->hdp.ras = &hdp_v4_0_ras; + amdgpu_ras_register_ras_block(adev, &adev->hdp.ras->ras_block); + adev->hdp.ras_if = &adev->hdp.ras->ras_block.ras_comm; } static void gmc_v9_0_set_mca_funcs(struct amdgpu_device *adev) @@ -1275,6 +1322,7 @@ static void gmc_v9_0_set_mca_funcs(struct amdgpu_device *adev) static int gmc_v9_0_early_init(void *handle) { + int r; struct amdgpu_device *adev = (struct amdgpu_device *)handle; /* ARCT and VEGA20 don't have XGMI defined in their IP discovery tables */ @@ -1304,6 +1352,10 @@ static int gmc_v9_0_early_init(void *handle) adev->gmc.private_aperture_end = adev->gmc.private_aperture_start + (4ULL << 30) - 1; + r = amdgpu_gmc_ras_early_init(adev); + if (r) + return r; + return 0; } @@ -1330,13 +1382,13 @@ static int gmc_v9_0_late_init(void *handle) } if (!amdgpu_persistent_edc_harvesting_supported(adev)) { - if (adev->mmhub.ras_funcs && - adev->mmhub.ras_funcs->reset_ras_error_count) - adev->mmhub.ras_funcs->reset_ras_error_count(adev); + if (adev->mmhub.ras && adev->mmhub.ras->ras_block.hw_ops && + adev->mmhub.ras->ras_block.hw_ops->reset_ras_error_count) + adev->mmhub.ras->ras_block.hw_ops->reset_ras_error_count(adev); - if (adev->hdp.ras_funcs && - adev->hdp.ras_funcs->reset_ras_error_count) - adev->hdp.ras_funcs->reset_ras_error_count(adev); + if (adev->hdp.ras && adev->hdp.ras->ras_block.hw_ops && + adev->hdp.ras->ras_block.hw_ops->reset_ras_error_count) + adev->hdp.ras->ras_block.hw_ops->reset_ras_error_count(adev); } r = amdgpu_gmc_ras_late_init(adev); @@ -1408,7 +1460,7 @@ static int gmc_v9_0_mc_init(struct amdgpu_device *adev) */ /* check whether both host-gpu and gpu-gpu xgmi links exist */ - if ((adev->flags & AMD_IS_APU) || + if (((adev->flags & AMD_IS_APU) && !amdgpu_passthrough(adev)) || (adev->gmc.xgmi.supported && adev->gmc.xgmi.connected_to_cpu)) { adev->gmc.aper_base = @@ -1505,7 +1557,7 @@ static void gmc_v9_0_save_registers(struct amdgpu_device *adev) static int gmc_v9_0_sw_init(void *handle) { - int r, vram_width = 0, vram_type = 0, vram_vendor = 0; + int r, vram_width = 0, vram_type = 0, vram_vendor = 0, dma_addr_bits; struct amdgpu_device *adev = (struct amdgpu_device *)handle; adev->gfxhub.funcs->init(adev); @@ -1576,12 +1628,15 @@ static int gmc_v9_0_sw_init(void *handle) amdgpu_vm_adjust_size(adev, 256 * 1024, 9, 3, 47); else amdgpu_vm_adjust_size(adev, 256 * 1024, 9, 3, 48); + if (adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 4, 2)) + adev->gmc.translate_further = adev->vm_manager.num_level > 1; break; case IP_VERSION(9, 4, 1): adev->num_vmhubs = 3; /* Keep the vm size same with Vega20 */ amdgpu_vm_adjust_size(adev, 256 * 1024, 9, 3, 48); + adev->gmc.translate_further = adev->vm_manager.num_level > 1; break; default: break; @@ -1621,12 +1676,13 @@ static int gmc_v9_0_sw_init(void *handle) */ adev->gmc.mc_mask = 0xffffffffffffULL; /* 48 bit MC */ - r = dma_set_mask_and_coherent(adev->dev, DMA_BIT_MASK(44)); + dma_addr_bits = adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 4, 2) ? 48:44; + r = dma_set_mask_and_coherent(adev->dev, DMA_BIT_MASK(dma_addr_bits)); if (r) { printk(KERN_WARNING "amdgpu: No suitable DMA available.\n"); return r; } - adev->need_swiotlb = drm_need_swiotlb(44); + adev->need_swiotlb = drm_need_swiotlb(dma_addr_bits); r = gmc_v9_0_mc_init(adev); if (r) @@ -1672,7 +1728,7 @@ static int gmc_v9_0_sw_fini(void *handle) amdgpu_gem_force_release(adev); amdgpu_vm_manager_fini(adev); amdgpu_gart_table_vram_free(adev); - amdgpu_bo_unref(&adev->gmc.pdb0_bo); + amdgpu_bo_free_kernel(&adev->gmc.pdb0_bo, NULL, &adev->gmc.ptr_pdb0); amdgpu_bo_fini(adev); return 0; @@ -1740,14 +1796,7 @@ static int gmc_v9_0_gart_enable(struct amdgpu_device *adev) return -EINVAL; } - if (amdgpu_sriov_vf(adev) && amdgpu_in_reset(adev)) - goto skip_pin_bo; - - r = amdgpu_gart_table_vram_pin(adev); - if (r) - return r; - -skip_pin_bo: + amdgpu_gtt_mgr_recover(&adev->mman.gtt_mgr); r = adev->gfxhub.funcs->gart_enable(adev); if (r) return r; @@ -1764,7 +1813,6 @@ skip_pin_bo: DRM_INFO("PTB located at 0x%016llX\n", (unsigned long long)amdgpu_bo_gpu_offset(adev->gart.bo)); - adev->gart.ready = true; return 0; } @@ -1772,7 +1820,7 @@ static int gmc_v9_0_hw_init(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; bool value; - int i; + int i, r; /* The sequence of these two function calls matters.*/ gmc_v9_0_init_golden_registers(adev); @@ -1807,7 +1855,14 @@ static int gmc_v9_0_hw_init(void *handle) if (adev->umc.funcs && adev->umc.funcs->init_registers) adev->umc.funcs->init_registers(adev); - return gmc_v9_0_gart_enable(adev); + r = gmc_v9_0_gart_enable(adev); + if (r) + return r; + + if (amdgpu_emu_mode == 1) + return amdgpu_gmc_vram_checking(adev); + else + return r; } /** @@ -1821,7 +1876,6 @@ static void gmc_v9_0_gart_disable(struct amdgpu_device *adev) { adev->gfxhub.funcs->gart_disable(adev); adev->mmhub.funcs->gart_disable(adev); - amdgpu_gart_table_vram_unpin(adev); } static int gmc_v9_0_hw_fini(void *handle) @@ -1901,7 +1955,7 @@ static int gmc_v9_0_set_clockgating_state(void *handle, return 0; } -static void gmc_v9_0_get_clockgating_state(void *handle, u32 *flags) +static void gmc_v9_0_get_clockgating_state(void *handle, u64 *flags) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; diff --git a/drivers/gpu/drm/amd/amdgpu/hdp_v4_0.c b/drivers/gpu/drm/amd/amdgpu/hdp_v4_0.c index eecfb1545c1e..adf89680f53e 100644 --- a/drivers/gpu/drm/amd/amdgpu/hdp_v4_0.c +++ b/drivers/gpu/drm/amd/amdgpu/hdp_v4_0.c @@ -124,7 +124,7 @@ static void hdp_v4_0_update_clock_gating(struct amdgpu_device *adev, } static void hdp_v4_0_get_clockgating_state(struct amdgpu_device *adev, - u32 *flags) + u64 *flags) { int data; @@ -146,17 +146,29 @@ static void hdp_v4_0_init_registers(struct amdgpu_device *adev) WREG32_FIELD15(HDP, 0, HDP_MISC_CNTL, FLUSH_INVALIDATE_CACHE, 1); + if (adev->ip_versions[HDP_HWIP][0] == IP_VERSION(4, 4, 0)) + WREG32_FIELD15(HDP, 0, HDP_MISC_CNTL, READ_BUFFER_WATERMARK, 2); + WREG32_SOC15(HDP, 0, mmHDP_NONSURFACE_BASE, (adev->gmc.vram_start >> 8)); WREG32_SOC15(HDP, 0, mmHDP_NONSURFACE_BASE_HI, (adev->gmc.vram_start >> 40)); } -const struct amdgpu_hdp_ras_funcs hdp_v4_0_ras_funcs = { - .ras_late_init = amdgpu_hdp_ras_late_init, - .ras_fini = amdgpu_hdp_ras_fini, +struct amdgpu_ras_block_hw_ops hdp_v4_0_ras_hw_ops = { .query_ras_error_count = hdp_v4_0_query_ras_error_count, .reset_ras_error_count = hdp_v4_0_reset_ras_error_count, }; +struct amdgpu_hdp_ras hdp_v4_0_ras = { + .ras_block = { + .ras_comm = { + .name = "hdp", + .block = AMDGPU_RAS_BLOCK__HDP, + .type = AMDGPU_RAS_ERROR__MULTI_UNCORRECTABLE, + }, + .hw_ops = &hdp_v4_0_ras_hw_ops, + }, +}; + const struct amdgpu_hdp_funcs hdp_v4_0_funcs = { .flush_hdp = hdp_v4_0_flush_hdp, .invalidate_hdp = hdp_v4_0_invalidate_hdp, diff --git a/drivers/gpu/drm/amd/amdgpu/hdp_v4_0.h b/drivers/gpu/drm/amd/amdgpu/hdp_v4_0.h index dc3a1b81dd62..c44eee9282ab 100644 --- a/drivers/gpu/drm/amd/amdgpu/hdp_v4_0.h +++ b/drivers/gpu/drm/amd/amdgpu/hdp_v4_0.h @@ -27,6 +27,6 @@ #include "soc15_common.h" extern const struct amdgpu_hdp_funcs hdp_v4_0_funcs; -extern const struct amdgpu_hdp_ras_funcs hdp_v4_0_ras_funcs; +extern struct amdgpu_hdp_ras hdp_v4_0_ras; #endif diff --git a/drivers/gpu/drm/amd/amdgpu/hdp_v5_0.c b/drivers/gpu/drm/amd/amdgpu/hdp_v5_0.c index 5793977953cc..a9ea23fa0def 100644 --- a/drivers/gpu/drm/amd/amdgpu/hdp_v5_0.c +++ b/drivers/gpu/drm/amd/amdgpu/hdp_v5_0.c @@ -181,7 +181,7 @@ static void hdp_v5_0_update_clock_gating(struct amdgpu_device *adev, } static void hdp_v5_0_get_clockgating_state(struct amdgpu_device *adev, - u32 *flags) + u64 *flags) { uint32_t tmp; diff --git a/drivers/gpu/drm/amd/amdgpu/hdp_v5_2.c b/drivers/gpu/drm/amd/amdgpu/hdp_v5_2.c new file mode 100644 index 000000000000..29c3484ae1f1 --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/hdp_v5_2.c @@ -0,0 +1,195 @@ +/* + * Copyright 2021 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ +#include "amdgpu.h" +#include "amdgpu_atombios.h" +#include "hdp_v5_2.h" + +#include "hdp/hdp_5_2_1_offset.h" +#include "hdp/hdp_5_2_1_sh_mask.h" +#include <uapi/linux/kfd_ioctl.h> + +static void hdp_v5_2_flush_hdp(struct amdgpu_device *adev, + struct amdgpu_ring *ring) +{ + if (!ring || !ring->funcs->emit_wreg) + WREG32_NO_KIQ((adev->rmmio_remap.reg_offset + KFD_MMIO_REMAP_HDP_MEM_FLUSH_CNTL) >> 2, + 0); + else + amdgpu_ring_emit_wreg(ring, + (adev->rmmio_remap.reg_offset + KFD_MMIO_REMAP_HDP_MEM_FLUSH_CNTL) >> 2, + 0); +} + +static void hdp_v5_2_update_mem_power_gating(struct amdgpu_device *adev, + bool enable) +{ + uint32_t hdp_clk_cntl; + uint32_t hdp_mem_pwr_cntl; + + if (!(adev->cg_flags & (AMD_CG_SUPPORT_HDP_LS | + AMD_CG_SUPPORT_HDP_DS | + AMD_CG_SUPPORT_HDP_SD))) + return; + + hdp_clk_cntl = RREG32_SOC15(HDP, 0, regHDP_CLK_CNTL); + hdp_mem_pwr_cntl = RREG32_SOC15(HDP, 0, regHDP_MEM_POWER_CTRL); + + /* Before doing clock/power mode switch, forced on MEM clock */ + hdp_clk_cntl = REG_SET_FIELD(hdp_clk_cntl, HDP_CLK_CNTL, + ATOMIC_MEM_CLK_SOFT_OVERRIDE, 1); + hdp_clk_cntl = REG_SET_FIELD(hdp_clk_cntl, HDP_CLK_CNTL, + RC_MEM_CLK_SOFT_OVERRIDE, 1); + WREG32_SOC15(HDP, 0, regHDP_CLK_CNTL, hdp_clk_cntl); + + /* disable clock and power gating before any changing */ + hdp_mem_pwr_cntl = REG_SET_FIELD(hdp_mem_pwr_cntl, HDP_MEM_POWER_CTRL, + ATOMIC_MEM_POWER_CTRL_EN, 0); + hdp_mem_pwr_cntl = REG_SET_FIELD(hdp_mem_pwr_cntl, HDP_MEM_POWER_CTRL, + ATOMIC_MEM_POWER_LS_EN, 0); + hdp_mem_pwr_cntl = REG_SET_FIELD(hdp_mem_pwr_cntl, HDP_MEM_POWER_CTRL, + ATOMIC_MEM_POWER_DS_EN, 0); + hdp_mem_pwr_cntl = REG_SET_FIELD(hdp_mem_pwr_cntl, HDP_MEM_POWER_CTRL, + ATOMIC_MEM_POWER_SD_EN, 0); + hdp_mem_pwr_cntl = REG_SET_FIELD(hdp_mem_pwr_cntl, HDP_MEM_POWER_CTRL, + RC_MEM_POWER_CTRL_EN, 0); + hdp_mem_pwr_cntl = REG_SET_FIELD(hdp_mem_pwr_cntl, HDP_MEM_POWER_CTRL, + RC_MEM_POWER_LS_EN, 0); + hdp_mem_pwr_cntl = REG_SET_FIELD(hdp_mem_pwr_cntl, HDP_MEM_POWER_CTRL, + RC_MEM_POWER_DS_EN, 0); + hdp_mem_pwr_cntl = REG_SET_FIELD(hdp_mem_pwr_cntl, HDP_MEM_POWER_CTRL, + RC_MEM_POWER_SD_EN, 0); + WREG32_SOC15(HDP, 0, regHDP_MEM_POWER_CTRL, hdp_mem_pwr_cntl); + + /* Already disabled above. The actions below are for "enabled" only */ + if (enable) { + /* only one clock gating mode (LS/DS/SD) can be enabled */ + if (adev->cg_flags & AMD_CG_SUPPORT_HDP_SD) { + hdp_mem_pwr_cntl = REG_SET_FIELD(hdp_mem_pwr_cntl, + HDP_MEM_POWER_CTRL, + ATOMIC_MEM_POWER_SD_EN, 1); + hdp_mem_pwr_cntl = REG_SET_FIELD(hdp_mem_pwr_cntl, + HDP_MEM_POWER_CTRL, + RC_MEM_POWER_SD_EN, 1); + } else if (adev->cg_flags & AMD_CG_SUPPORT_HDP_LS) { + hdp_mem_pwr_cntl = REG_SET_FIELD(hdp_mem_pwr_cntl, + HDP_MEM_POWER_CTRL, + ATOMIC_MEM_POWER_LS_EN, 1); + hdp_mem_pwr_cntl = REG_SET_FIELD(hdp_mem_pwr_cntl, + HDP_MEM_POWER_CTRL, + RC_MEM_POWER_LS_EN, 1); + } else if (adev->cg_flags & AMD_CG_SUPPORT_HDP_DS) { + hdp_mem_pwr_cntl = REG_SET_FIELD(hdp_mem_pwr_cntl, + HDP_MEM_POWER_CTRL, + ATOMIC_MEM_POWER_DS_EN, 1); + hdp_mem_pwr_cntl = REG_SET_FIELD(hdp_mem_pwr_cntl, + HDP_MEM_POWER_CTRL, + RC_MEM_POWER_DS_EN, 1); + } + + /* confirmed that ATOMIC/RC_MEM_POWER_CTRL_EN have to be set for SRAM LS/DS/SD */ + if (adev->cg_flags & (AMD_CG_SUPPORT_HDP_LS | AMD_CG_SUPPORT_HDP_DS | + AMD_CG_SUPPORT_HDP_SD)) { + hdp_mem_pwr_cntl = REG_SET_FIELD(hdp_mem_pwr_cntl, HDP_MEM_POWER_CTRL, + ATOMIC_MEM_POWER_CTRL_EN, 1); + hdp_mem_pwr_cntl = REG_SET_FIELD(hdp_mem_pwr_cntl, HDP_MEM_POWER_CTRL, + RC_MEM_POWER_CTRL_EN, 1); + WREG32_SOC15(HDP, 0, regHDP_MEM_POWER_CTRL, hdp_mem_pwr_cntl); + } + } + + /* disable MEM clock override after clock/power mode changing */ + hdp_clk_cntl = REG_SET_FIELD(hdp_clk_cntl, HDP_CLK_CNTL, + ATOMIC_MEM_CLK_SOFT_OVERRIDE, 0); + hdp_clk_cntl = REG_SET_FIELD(hdp_clk_cntl, HDP_CLK_CNTL, + RC_MEM_CLK_SOFT_OVERRIDE, 0); + WREG32_SOC15(HDP, 0, regHDP_CLK_CNTL, hdp_clk_cntl); +} + +static void hdp_v5_2_update_medium_grain_clock_gating(struct amdgpu_device *adev, + bool enable) +{ + uint32_t hdp_clk_cntl; + + if (!(adev->cg_flags & AMD_CG_SUPPORT_HDP_MGCG)) + return; + + hdp_clk_cntl = RREG32_SOC15(HDP, 0, regHDP_CLK_CNTL); + + if (enable) { + hdp_clk_cntl &= + ~(uint32_t) + (HDP_CLK_CNTL__ATOMIC_MEM_CLK_SOFT_OVERRIDE_MASK | + HDP_CLK_CNTL__RC_MEM_CLK_SOFT_OVERRIDE_MASK | + HDP_CLK_CNTL__DBUS_CLK_SOFT_OVERRIDE_MASK | + HDP_CLK_CNTL__DYN_CLK_SOFT_OVERRIDE_MASK | + HDP_CLK_CNTL__XDP_REG_CLK_SOFT_OVERRIDE_MASK | + HDP_CLK_CNTL__HDP_REG_CLK_SOFT_OVERRIDE_MASK); + } else { + hdp_clk_cntl |= HDP_CLK_CNTL__ATOMIC_MEM_CLK_SOFT_OVERRIDE_MASK | + HDP_CLK_CNTL__RC_MEM_CLK_SOFT_OVERRIDE_MASK | + HDP_CLK_CNTL__DBUS_CLK_SOFT_OVERRIDE_MASK | + HDP_CLK_CNTL__DYN_CLK_SOFT_OVERRIDE_MASK | + HDP_CLK_CNTL__XDP_REG_CLK_SOFT_OVERRIDE_MASK | + HDP_CLK_CNTL__HDP_REG_CLK_SOFT_OVERRIDE_MASK; + } + + WREG32_SOC15(HDP, 0, regHDP_CLK_CNTL, hdp_clk_cntl); +} + +static void hdp_v5_2_get_clockgating_state(struct amdgpu_device *adev, + u64 *flags) +{ + uint32_t tmp; + + /* AMD_CG_SUPPORT_HDP_MGCG */ + tmp = RREG32_SOC15(HDP, 0, regHDP_CLK_CNTL); + if (!(tmp & (HDP_CLK_CNTL__ATOMIC_MEM_CLK_SOFT_OVERRIDE_MASK | + HDP_CLK_CNTL__RC_MEM_CLK_SOFT_OVERRIDE_MASK | + HDP_CLK_CNTL__DBUS_CLK_SOFT_OVERRIDE_MASK | + HDP_CLK_CNTL__DYN_CLK_SOFT_OVERRIDE_MASK | + HDP_CLK_CNTL__XDP_REG_CLK_SOFT_OVERRIDE_MASK | + HDP_CLK_CNTL__HDP_REG_CLK_SOFT_OVERRIDE_MASK))) + *flags |= AMD_CG_SUPPORT_HDP_MGCG; + + /* AMD_CG_SUPPORT_HDP_LS/DS/SD */ + tmp = RREG32_SOC15(HDP, 0, regHDP_MEM_POWER_CTRL); + if (tmp & HDP_MEM_POWER_CTRL__ATOMIC_MEM_POWER_LS_EN_MASK) + *flags |= AMD_CG_SUPPORT_HDP_LS; + else if (tmp & HDP_MEM_POWER_CTRL__ATOMIC_MEM_POWER_DS_EN_MASK) + *flags |= AMD_CG_SUPPORT_HDP_DS; + else if (tmp & HDP_MEM_POWER_CTRL__ATOMIC_MEM_POWER_SD_EN_MASK) + *flags |= AMD_CG_SUPPORT_HDP_SD; +} + +static void hdp_v5_2_update_clock_gating(struct amdgpu_device *adev, + bool enable) +{ + hdp_v5_2_update_mem_power_gating(adev, enable); + hdp_v5_2_update_medium_grain_clock_gating(adev, enable); +} + +const struct amdgpu_hdp_funcs hdp_v5_2_funcs = { + .flush_hdp = hdp_v5_2_flush_hdp, + .update_clock_gating = hdp_v5_2_update_clock_gating, + .get_clock_gating_state = hdp_v5_2_get_clockgating_state, +}; diff --git a/drivers/gpu/drm/amd/amdgpu/hdp_v5_2.h b/drivers/gpu/drm/amd/amdgpu/hdp_v5_2.h new file mode 100644 index 000000000000..cb2abc0c80ee --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/hdp_v5_2.h @@ -0,0 +1,31 @@ +/* + * Copyright 2021 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#ifndef __HDP_V5_2_H__ +#define __HDP_V5_2_H__ + +#include "soc15_common.h" + +extern const struct amdgpu_hdp_funcs hdp_v5_2_funcs; + +#endif diff --git a/drivers/gpu/drm/amd/amdgpu/hdp_v6_0.c b/drivers/gpu/drm/amd/amdgpu/hdp_v6_0.c new file mode 100644 index 000000000000..063eba619f2f --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/hdp_v6_0.c @@ -0,0 +1,142 @@ +/* + * Copyright 2020 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ +#include "amdgpu.h" +#include "amdgpu_atombios.h" +#include "hdp_v6_0.h" + +#include "hdp/hdp_6_0_0_offset.h" +#include "hdp/hdp_6_0_0_sh_mask.h" +#include <uapi/linux/kfd_ioctl.h> + +static void hdp_v6_0_flush_hdp(struct amdgpu_device *adev, + struct amdgpu_ring *ring) +{ + if (!ring || !ring->funcs->emit_wreg) + WREG32_NO_KIQ((adev->rmmio_remap.reg_offset + KFD_MMIO_REMAP_HDP_MEM_FLUSH_CNTL) >> 2, 0); + else + amdgpu_ring_emit_wreg(ring, (adev->rmmio_remap.reg_offset + KFD_MMIO_REMAP_HDP_MEM_FLUSH_CNTL) >> 2, 0); +} + +static void hdp_v6_0_update_clock_gating(struct amdgpu_device *adev, + bool enable) +{ + uint32_t hdp_clk_cntl, hdp_clk_cntl1; + uint32_t hdp_mem_pwr_cntl; + + if (!(adev->cg_flags & (AMD_CG_SUPPORT_HDP_LS | + AMD_CG_SUPPORT_HDP_DS | + AMD_CG_SUPPORT_HDP_SD))) + return; + + hdp_clk_cntl = hdp_clk_cntl1 = RREG32_SOC15(HDP, 0,regHDP_CLK_CNTL); + hdp_mem_pwr_cntl = RREG32_SOC15(HDP, 0, regHDP_MEM_POWER_CTRL); + + /* Before doing clock/power mode switch, + * forced on IPH & RC clock */ + hdp_clk_cntl = REG_SET_FIELD(hdp_clk_cntl, HDP_CLK_CNTL, + RC_MEM_CLK_SOFT_OVERRIDE, 1); + WREG32_SOC15(HDP, 0, regHDP_CLK_CNTL, hdp_clk_cntl); + + /* disable clock and power gating before any changing */ + hdp_mem_pwr_cntl = REG_SET_FIELD(hdp_mem_pwr_cntl, HDP_MEM_POWER_CTRL, + ATOMIC_MEM_POWER_CTRL_EN, 0); + hdp_mem_pwr_cntl = REG_SET_FIELD(hdp_mem_pwr_cntl, HDP_MEM_POWER_CTRL, + ATOMIC_MEM_POWER_LS_EN, 0); + hdp_mem_pwr_cntl = REG_SET_FIELD(hdp_mem_pwr_cntl, HDP_MEM_POWER_CTRL, + ATOMIC_MEM_POWER_DS_EN, 0); + hdp_mem_pwr_cntl = REG_SET_FIELD(hdp_mem_pwr_cntl, HDP_MEM_POWER_CTRL, + ATOMIC_MEM_POWER_SD_EN, 0); + hdp_mem_pwr_cntl = REG_SET_FIELD(hdp_mem_pwr_cntl, HDP_MEM_POWER_CTRL, + RC_MEM_POWER_CTRL_EN, 0); + hdp_mem_pwr_cntl = REG_SET_FIELD(hdp_mem_pwr_cntl, HDP_MEM_POWER_CTRL, + RC_MEM_POWER_LS_EN, 0); + hdp_mem_pwr_cntl = REG_SET_FIELD(hdp_mem_pwr_cntl, HDP_MEM_POWER_CTRL, + RC_MEM_POWER_DS_EN, 0); + hdp_mem_pwr_cntl = REG_SET_FIELD(hdp_mem_pwr_cntl, HDP_MEM_POWER_CTRL, + RC_MEM_POWER_SD_EN, 0); + WREG32_SOC15(HDP, 0, regHDP_MEM_POWER_CTRL, hdp_mem_pwr_cntl); + + /* Already disabled above. The actions below are for "enabled" only */ + if (enable) { + /* only one clock gating mode (LS/DS/SD) can be enabled */ + if (adev->cg_flags & AMD_CG_SUPPORT_HDP_SD) { + hdp_mem_pwr_cntl = REG_SET_FIELD(hdp_mem_pwr_cntl, + HDP_MEM_POWER_CTRL, + ATOMIC_MEM_POWER_SD_EN, 1); + hdp_mem_pwr_cntl = REG_SET_FIELD(hdp_mem_pwr_cntl, + HDP_MEM_POWER_CTRL, + RC_MEM_POWER_SD_EN, 1); + } else if (adev->cg_flags & AMD_CG_SUPPORT_HDP_LS) { + hdp_mem_pwr_cntl = REG_SET_FIELD(hdp_mem_pwr_cntl, + HDP_MEM_POWER_CTRL, + ATOMIC_MEM_POWER_LS_EN, 1); + hdp_mem_pwr_cntl = REG_SET_FIELD(hdp_mem_pwr_cntl, + HDP_MEM_POWER_CTRL, + RC_MEM_POWER_LS_EN, 1); + } else if (adev->cg_flags & AMD_CG_SUPPORT_HDP_DS) { + hdp_mem_pwr_cntl = REG_SET_FIELD(hdp_mem_pwr_cntl, + HDP_MEM_POWER_CTRL, + ATOMIC_MEM_POWER_DS_EN, 1); + hdp_mem_pwr_cntl = REG_SET_FIELD(hdp_mem_pwr_cntl, + HDP_MEM_POWER_CTRL, + RC_MEM_POWER_DS_EN, 1); + } + + /* confirmed that IPH_MEM_POWER_CTRL_EN and RC_MEM_POWER_CTRL_EN have to + * be set for SRAM LS/DS/SD */ + if (adev->cg_flags & (AMD_CG_SUPPORT_HDP_LS | AMD_CG_SUPPORT_HDP_DS | + AMD_CG_SUPPORT_HDP_SD)) { + hdp_mem_pwr_cntl = REG_SET_FIELD(hdp_mem_pwr_cntl, HDP_MEM_POWER_CTRL, + ATOMIC_MEM_POWER_CTRL_EN, 1); + hdp_mem_pwr_cntl = REG_SET_FIELD(hdp_mem_pwr_cntl, HDP_MEM_POWER_CTRL, + RC_MEM_POWER_CTRL_EN, 1); + WREG32_SOC15(HDP, 0, regHDP_MEM_POWER_CTRL, hdp_mem_pwr_cntl); + } + } + + /* disable IPH & RC clock override after clock/power mode changing */ + hdp_clk_cntl = REG_SET_FIELD(hdp_clk_cntl, HDP_CLK_CNTL, + RC_MEM_CLK_SOFT_OVERRIDE, 0); + WREG32_SOC15(HDP, 0, regHDP_CLK_CNTL, hdp_clk_cntl); +} + +static void hdp_v6_0_get_clockgating_state(struct amdgpu_device *adev, + u64 *flags) +{ + uint32_t tmp; + + /* AMD_CG_SUPPORT_HDP_LS/DS/SD */ + tmp = RREG32_SOC15(HDP, 0, regHDP_MEM_POWER_CTRL); + if (tmp & HDP_MEM_POWER_CTRL__ATOMIC_MEM_POWER_LS_EN_MASK) + *flags |= AMD_CG_SUPPORT_HDP_LS; + else if (tmp & HDP_MEM_POWER_CTRL__ATOMIC_MEM_POWER_DS_EN_MASK) + *flags |= AMD_CG_SUPPORT_HDP_DS; + else if (tmp & HDP_MEM_POWER_CTRL__ATOMIC_MEM_POWER_SD_EN_MASK) + *flags |= AMD_CG_SUPPORT_HDP_SD; +} + +const struct amdgpu_hdp_funcs hdp_v6_0_funcs = { + .flush_hdp = hdp_v6_0_flush_hdp, + .update_clock_gating = hdp_v6_0_update_clock_gating, + .get_clock_gating_state = hdp_v6_0_get_clockgating_state, +}; diff --git a/drivers/gpu/drm/amd/amdgpu/hdp_v6_0.h b/drivers/gpu/drm/amd/amdgpu/hdp_v6_0.h new file mode 100644 index 000000000000..533ecd8c0800 --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/hdp_v6_0.h @@ -0,0 +1,31 @@ +/* + * Copyright 2020 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#ifndef __HDP_V6_0_H__ +#define __HDP_V6_0_H__ + +#include "soc15_common.h" + +extern const struct amdgpu_hdp_funcs hdp_v6_0_funcs; + +#endif diff --git a/drivers/gpu/drm/amd/amdgpu/iceland_ih.c b/drivers/gpu/drm/amd/amdgpu/iceland_ih.c index ddfe4eaeea05..aecad530b10a 100644 --- a/drivers/gpu/drm/amd/amdgpu/iceland_ih.c +++ b/drivers/gpu/drm/amd/amdgpu/iceland_ih.c @@ -308,14 +308,9 @@ static int iceland_ih_sw_fini(void *handle) static int iceland_ih_hw_init(void *handle) { - int r; struct amdgpu_device *adev = (struct amdgpu_device *)handle; - r = iceland_ih_irq_init(adev); - if (r) - return r; - - return 0; + return iceland_ih_irq_init(adev); } static int iceland_ih_hw_fini(void *handle) diff --git a/drivers/gpu/drm/amd/amdgpu/ih_v6_0.c b/drivers/gpu/drm/amd/amdgpu/ih_v6_0.c new file mode 100644 index 000000000000..7cd79a3844b2 --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/ih_v6_0.c @@ -0,0 +1,766 @@ +/* + * Copyright 2021 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#include <linux/pci.h> + +#include "amdgpu.h" +#include "amdgpu_ih.h" + +#include "oss/osssys_6_0_0_offset.h" +#include "oss/osssys_6_0_0_sh_mask.h" + +#include "soc15_common.h" +#include "ih_v6_0.h" + +#define MAX_REARM_RETRY 10 + +static void ih_v6_0_set_interrupt_funcs(struct amdgpu_device *adev); + +/** + * ih_v6_0_init_register_offset - Initialize register offset for ih rings + * + * @adev: amdgpu_device pointer + * + * Initialize register offset ih rings (IH_V6_0). + */ +static void ih_v6_0_init_register_offset(struct amdgpu_device *adev) +{ + struct amdgpu_ih_regs *ih_regs; + + /* ih ring 2 is removed + * ih ring and ih ring 1 are available */ + if (adev->irq.ih.ring_size) { + ih_regs = &adev->irq.ih.ih_regs; + ih_regs->ih_rb_base = SOC15_REG_OFFSET(OSSSYS, 0, regIH_RB_BASE); + ih_regs->ih_rb_base_hi = SOC15_REG_OFFSET(OSSSYS, 0, regIH_RB_BASE_HI); + ih_regs->ih_rb_cntl = SOC15_REG_OFFSET(OSSSYS, 0, regIH_RB_CNTL); + ih_regs->ih_rb_wptr = SOC15_REG_OFFSET(OSSSYS, 0, regIH_RB_WPTR); + ih_regs->ih_rb_rptr = SOC15_REG_OFFSET(OSSSYS, 0, regIH_RB_RPTR); + ih_regs->ih_doorbell_rptr = SOC15_REG_OFFSET(OSSSYS, 0, regIH_DOORBELL_RPTR); + ih_regs->ih_rb_wptr_addr_lo = SOC15_REG_OFFSET(OSSSYS, 0, regIH_RB_WPTR_ADDR_LO); + ih_regs->ih_rb_wptr_addr_hi = SOC15_REG_OFFSET(OSSSYS, 0, regIH_RB_WPTR_ADDR_HI); + ih_regs->psp_reg_id = PSP_REG_IH_RB_CNTL; + } + + if (adev->irq.ih1.ring_size) { + ih_regs = &adev->irq.ih1.ih_regs; + ih_regs->ih_rb_base = SOC15_REG_OFFSET(OSSSYS, 0, regIH_RB_BASE_RING1); + ih_regs->ih_rb_base_hi = SOC15_REG_OFFSET(OSSSYS, 0, regIH_RB_BASE_HI_RING1); + ih_regs->ih_rb_cntl = SOC15_REG_OFFSET(OSSSYS, 0, regIH_RB_CNTL_RING1); + ih_regs->ih_rb_wptr = SOC15_REG_OFFSET(OSSSYS, 0, regIH_RB_WPTR_RING1); + ih_regs->ih_rb_rptr = SOC15_REG_OFFSET(OSSSYS, 0, regIH_RB_RPTR_RING1); + ih_regs->ih_doorbell_rptr = SOC15_REG_OFFSET(OSSSYS, 0, regIH_DOORBELL_RPTR_RING1); + ih_regs->psp_reg_id = PSP_REG_IH_RB_CNTL_RING1; + } +} + +/** + * force_update_wptr_for_self_int - Force update the wptr for self interrupt + * + * @adev: amdgpu_device pointer + * @threshold: threshold to trigger the wptr reporting + * @timeout: timeout to trigger the wptr reporting + * @enabled: Enable/disable timeout flush mechanism + * + * threshold input range: 0 ~ 15, default 0, + * real_threshold = 2^threshold + * timeout input range: 0 ~ 20, default 8, + * real_timeout = (2^timeout) * 1024 / (socclk_freq) + * + * Force update wptr for self interrupt ( >= SIENNA_CICHLID). + */ +static void +force_update_wptr_for_self_int(struct amdgpu_device *adev, + u32 threshold, u32 timeout, bool enabled) +{ + u32 ih_cntl, ih_rb_cntl; + + ih_cntl = RREG32_SOC15(OSSSYS, 0, regIH_CNTL2); + ih_rb_cntl = RREG32_SOC15(OSSSYS, 0, regIH_RB_CNTL_RING1); + + ih_cntl = REG_SET_FIELD(ih_cntl, IH_CNTL2, + SELF_IV_FORCE_WPTR_UPDATE_TIMEOUT, timeout); + ih_cntl = REG_SET_FIELD(ih_cntl, IH_CNTL2, + SELF_IV_FORCE_WPTR_UPDATE_ENABLE, enabled); + ih_rb_cntl = REG_SET_FIELD(ih_rb_cntl, IH_RB_CNTL_RING1, + RB_USED_INT_THRESHOLD, threshold); + + if (amdgpu_sriov_vf(adev) && amdgpu_sriov_reg_indirect_ih(adev)) { + if (psp_reg_program(&adev->psp, PSP_REG_IH_RB_CNTL_RING1, ih_rb_cntl)) + return; + } else { + WREG32_SOC15(OSSSYS, 0, regIH_RB_CNTL_RING1, ih_rb_cntl); + } + + WREG32_SOC15(OSSSYS, 0, regIH_CNTL2, ih_cntl); +} + +/** + * ih_v6_0_toggle_ring_interrupts - toggle the interrupt ring buffer + * + * @adev: amdgpu_device pointer + * @ih: amdgpu_ih_ring pointet + * @enable: true - enable the interrupts, false - disable the interrupts + * + * Toggle the interrupt ring buffer (IH_V6_0) + */ +static int ih_v6_0_toggle_ring_interrupts(struct amdgpu_device *adev, + struct amdgpu_ih_ring *ih, + bool enable) +{ + struct amdgpu_ih_regs *ih_regs; + uint32_t tmp; + + ih_regs = &ih->ih_regs; + + tmp = RREG32(ih_regs->ih_rb_cntl); + tmp = REG_SET_FIELD(tmp, IH_RB_CNTL, RB_ENABLE, (enable ? 1 : 0)); + /* enable_intr field is only valid in ring0 */ + if (ih == &adev->irq.ih) + tmp = REG_SET_FIELD(tmp, IH_RB_CNTL, ENABLE_INTR, (enable ? 1 : 0)); + + if (amdgpu_sriov_vf(adev) && amdgpu_sriov_reg_indirect_ih(adev)) { + if (psp_reg_program(&adev->psp, ih_regs->psp_reg_id, tmp)) + return -ETIMEDOUT; + } else { + WREG32(ih_regs->ih_rb_cntl, tmp); + } + + if (enable) { + ih->enabled = true; + } else { + /* set rptr, wptr to 0 */ + WREG32(ih_regs->ih_rb_rptr, 0); + WREG32(ih_regs->ih_rb_wptr, 0); + ih->enabled = false; + ih->rptr = 0; + } + + return 0; +} + +/** + * ih_v6_0_toggle_interrupts - Toggle all the available interrupt ring buffers + * + * @adev: amdgpu_device pointer + * @enable: enable or disable interrupt ring buffers + * + * Toggle all the available interrupt ring buffers (IH_V6_0). + */ +static int ih_v6_0_toggle_interrupts(struct amdgpu_device *adev, bool enable) +{ + struct amdgpu_ih_ring *ih[] = {&adev->irq.ih, &adev->irq.ih1}; + int i; + int r; + + for (i = 0; i < ARRAY_SIZE(ih); i++) { + if (ih[i]->ring_size) { + r = ih_v6_0_toggle_ring_interrupts(adev, ih[i], enable); + if (r) + return r; + } + } + + return 0; +} + +static uint32_t ih_v6_0_rb_cntl(struct amdgpu_ih_ring *ih, uint32_t ih_rb_cntl) +{ + int rb_bufsz = order_base_2(ih->ring_size / 4); + + ih_rb_cntl = REG_SET_FIELD(ih_rb_cntl, IH_RB_CNTL, + MC_SPACE, ih->use_bus_addr ? 2 : 4); + ih_rb_cntl = REG_SET_FIELD(ih_rb_cntl, IH_RB_CNTL, + WPTR_OVERFLOW_CLEAR, 1); + ih_rb_cntl = REG_SET_FIELD(ih_rb_cntl, IH_RB_CNTL, + WPTR_OVERFLOW_ENABLE, 1); + ih_rb_cntl = REG_SET_FIELD(ih_rb_cntl, IH_RB_CNTL, RB_SIZE, rb_bufsz); + /* Ring Buffer write pointer writeback. If enabled, IH_RB_WPTR register + * value is written to memory + */ + ih_rb_cntl = REG_SET_FIELD(ih_rb_cntl, IH_RB_CNTL, + WPTR_WRITEBACK_ENABLE, 1); + ih_rb_cntl = REG_SET_FIELD(ih_rb_cntl, IH_RB_CNTL, MC_SNOOP, 1); + ih_rb_cntl = REG_SET_FIELD(ih_rb_cntl, IH_RB_CNTL, MC_RO, 0); + ih_rb_cntl = REG_SET_FIELD(ih_rb_cntl, IH_RB_CNTL, MC_VMID, 0); + + return ih_rb_cntl; +} + +static uint32_t ih_v6_0_doorbell_rptr(struct amdgpu_ih_ring *ih) +{ + u32 ih_doorbell_rtpr = 0; + + if (ih->use_doorbell) { + ih_doorbell_rtpr = REG_SET_FIELD(ih_doorbell_rtpr, + IH_DOORBELL_RPTR, OFFSET, + ih->doorbell_index); + ih_doorbell_rtpr = REG_SET_FIELD(ih_doorbell_rtpr, + IH_DOORBELL_RPTR, + ENABLE, 1); + } else { + ih_doorbell_rtpr = REG_SET_FIELD(ih_doorbell_rtpr, + IH_DOORBELL_RPTR, + ENABLE, 0); + } + return ih_doorbell_rtpr; +} + +/** + * ih_v6_0_enable_ring - enable an ih ring buffer + * + * @adev: amdgpu_device pointer + * @ih: amdgpu_ih_ring pointer + * + * Enable an ih ring buffer (IH_V6_0) + */ +static int ih_v6_0_enable_ring(struct amdgpu_device *adev, + struct amdgpu_ih_ring *ih) +{ + struct amdgpu_ih_regs *ih_regs; + uint32_t tmp; + + ih_regs = &ih->ih_regs; + + /* Ring Buffer base. [39:8] of 40-bit address of the beginning of the ring buffer*/ + WREG32(ih_regs->ih_rb_base, ih->gpu_addr >> 8); + WREG32(ih_regs->ih_rb_base_hi, (ih->gpu_addr >> 40) & 0xff); + + tmp = RREG32(ih_regs->ih_rb_cntl); + tmp = ih_v6_0_rb_cntl(ih, tmp); + if (ih == &adev->irq.ih) + tmp = REG_SET_FIELD(tmp, IH_RB_CNTL, RPTR_REARM, !!adev->irq.msi_enabled); + if (ih == &adev->irq.ih1) { + tmp = REG_SET_FIELD(tmp, IH_RB_CNTL, WPTR_OVERFLOW_ENABLE, 0); + tmp = REG_SET_FIELD(tmp, IH_RB_CNTL, RB_FULL_DRAIN_ENABLE, 1); + } + + if (amdgpu_sriov_vf(adev) && amdgpu_sriov_reg_indirect_ih(adev)) { + if (psp_reg_program(&adev->psp, ih_regs->psp_reg_id, tmp)) { + DRM_ERROR("PSP program IH_RB_CNTL failed!\n"); + return -ETIMEDOUT; + } + } else { + WREG32(ih_regs->ih_rb_cntl, tmp); + } + + if (ih == &adev->irq.ih) { + /* set the ih ring 0 writeback address whether it's enabled or not */ + WREG32(ih_regs->ih_rb_wptr_addr_lo, lower_32_bits(ih->wptr_addr)); + WREG32(ih_regs->ih_rb_wptr_addr_hi, upper_32_bits(ih->wptr_addr) & 0xFFFF); + } + + /* set rptr, wptr to 0 */ + WREG32(ih_regs->ih_rb_wptr, 0); + WREG32(ih_regs->ih_rb_rptr, 0); + + WREG32(ih_regs->ih_doorbell_rptr, ih_v6_0_doorbell_rptr(ih)); + + return 0; +} + +/** + * ih_v6_0_irq_init - init and enable the interrupt ring + * + * @adev: amdgpu_device pointer + * + * Allocate a ring buffer for the interrupt controller, + * enable the RLC, disable interrupts, enable the IH + * ring buffer and enable it. + * Called at device load and reume. + * Returns 0 for success, errors for failure. + */ +static int ih_v6_0_irq_init(struct amdgpu_device *adev) +{ + struct amdgpu_ih_ring *ih[] = {&adev->irq.ih, &adev->irq.ih1}; + u32 ih_chicken; + u32 tmp; + int ret; + int i; + + /* disable irqs */ + ret = ih_v6_0_toggle_interrupts(adev, false); + if (ret) + return ret; + + adev->nbio.funcs->ih_control(adev); + + if (unlikely((adev->firmware.load_type == AMDGPU_FW_LOAD_DIRECT) || + (adev->firmware.load_type == AMDGPU_FW_LOAD_RLC_BACKDOOR_AUTO))) { + if (ih[0]->use_bus_addr) { + ih_chicken = RREG32_SOC15(OSSSYS, 0, regIH_CHICKEN); + ih_chicken = REG_SET_FIELD(ih_chicken, + IH_CHICKEN, MC_SPACE_GPA_ENABLE, 1); + WREG32_SOC15(OSSSYS, 0, regIH_CHICKEN, ih_chicken); + } + } + + for (i = 0; i < ARRAY_SIZE(ih); i++) { + if (ih[i]->ring_size) { + ret = ih_v6_0_enable_ring(adev, ih[i]); + if (ret) + return ret; + } + } + + /* update doorbell range for ih ring 0 */ + adev->nbio.funcs->ih_doorbell_range(adev, ih[0]->use_doorbell, + ih[0]->doorbell_index); + + tmp = RREG32_SOC15(OSSSYS, 0, regIH_STORM_CLIENT_LIST_CNTL); + tmp = REG_SET_FIELD(tmp, IH_STORM_CLIENT_LIST_CNTL, + CLIENT18_IS_STORM_CLIENT, 1); + WREG32_SOC15(OSSSYS, 0, regIH_STORM_CLIENT_LIST_CNTL, tmp); + + tmp = RREG32_SOC15(OSSSYS, 0, regIH_INT_FLOOD_CNTL); + tmp = REG_SET_FIELD(tmp, IH_INT_FLOOD_CNTL, FLOOD_CNTL_ENABLE, 1); + WREG32_SOC15(OSSSYS, 0, regIH_INT_FLOOD_CNTL, tmp); + + /* GC/MMHUB UTCL2 page fault interrupts are configured as + * MSI storm capable interrupts by deafult. The delay is + * used to avoid ISR being called too frequently + * when page fault happens on several continuous page + * and thus avoid MSI storm */ + tmp = RREG32_SOC15(OSSSYS, 0, regIH_MSI_STORM_CTRL); + tmp = REG_SET_FIELD(tmp, IH_MSI_STORM_CTRL, + DELAY, 3); + WREG32_SOC15(OSSSYS, 0, regIH_MSI_STORM_CTRL, tmp); + + pci_set_master(adev->pdev); + + /* enable interrupts */ + ret = ih_v6_0_toggle_interrupts(adev, true); + if (ret) + return ret; + /* enable wptr force update for self int */ + force_update_wptr_for_self_int(adev, 0, 8, true); + + if (adev->irq.ih_soft.ring_size) + adev->irq.ih_soft.enabled = true; + + return 0; +} + +/** + * ih_v6_0_irq_disable - disable interrupts + * + * @adev: amdgpu_device pointer + * + * Disable interrupts on the hw. + */ +static void ih_v6_0_irq_disable(struct amdgpu_device *adev) +{ + force_update_wptr_for_self_int(adev, 0, 8, false); + ih_v6_0_toggle_interrupts(adev, false); + + /* Wait and acknowledge irq */ + mdelay(1); +} + +/** + * ih_v6_0_get_wptr - get the IH ring buffer wptr + * + * @adev: amdgpu_device pointer + * + * Get the IH ring buffer wptr from either the register + * or the writeback memory buffer. Also check for + * ring buffer overflow and deal with it. + * Returns the value of the wptr. + */ +static u32 ih_v6_0_get_wptr(struct amdgpu_device *adev, + struct amdgpu_ih_ring *ih) +{ + u32 wptr, tmp; + struct amdgpu_ih_regs *ih_regs; + + wptr = le32_to_cpu(*ih->wptr_cpu); + ih_regs = &ih->ih_regs; + + if (!REG_GET_FIELD(wptr, IH_RB_WPTR, RB_OVERFLOW)) + goto out; + + wptr = RREG32_NO_KIQ(ih_regs->ih_rb_wptr); + if (!REG_GET_FIELD(wptr, IH_RB_WPTR, RB_OVERFLOW)) + goto out; + wptr = REG_SET_FIELD(wptr, IH_RB_WPTR, RB_OVERFLOW, 0); + + /* When a ring buffer overflow happen start parsing interrupt + * from the last not overwritten vector (wptr + 32). Hopefully + * this should allow us to catch up. + */ + tmp = (wptr + 32) & ih->ptr_mask; + dev_warn(adev->dev, "IH ring buffer overflow " + "(0x%08X, 0x%08X, 0x%08X)\n", + wptr, ih->rptr, tmp); + ih->rptr = tmp; + + tmp = RREG32_NO_KIQ(ih_regs->ih_rb_cntl); + tmp = REG_SET_FIELD(tmp, IH_RB_CNTL, WPTR_OVERFLOW_CLEAR, 1); + WREG32_NO_KIQ(ih_regs->ih_rb_cntl, tmp); +out: + return (wptr & ih->ptr_mask); +} + +/** + * ih_v6_0_irq_rearm - rearm IRQ if lost + * + * @adev: amdgpu_device pointer + * + */ +static void ih_v6_0_irq_rearm(struct amdgpu_device *adev, + struct amdgpu_ih_ring *ih) +{ + uint32_t v = 0; + uint32_t i = 0; + struct amdgpu_ih_regs *ih_regs; + + ih_regs = &ih->ih_regs; + + /* Rearm IRQ / re-write doorbell if doorbell write is lost */ + for (i = 0; i < MAX_REARM_RETRY; i++) { + v = RREG32_NO_KIQ(ih_regs->ih_rb_rptr); + if ((v < ih->ring_size) && (v != ih->rptr)) + WDOORBELL32(ih->doorbell_index, ih->rptr); + else + break; + } +} + +/** + * ih_v6_0_set_rptr - set the IH ring buffer rptr + * + * @adev: amdgpu_device pointer + * + * Set the IH ring buffer rptr. + */ +static void ih_v6_0_set_rptr(struct amdgpu_device *adev, + struct amdgpu_ih_ring *ih) +{ + struct amdgpu_ih_regs *ih_regs; + + if (ih->use_doorbell) { + /* XXX check if swapping is necessary on BE */ + *ih->rptr_cpu = ih->rptr; + WDOORBELL32(ih->doorbell_index, ih->rptr); + + if (amdgpu_sriov_vf(adev)) + ih_v6_0_irq_rearm(adev, ih); + } else { + ih_regs = &ih->ih_regs; + WREG32(ih_regs->ih_rb_rptr, ih->rptr); + } +} + +/** + * ih_v6_0_self_irq - dispatch work for ring 1 + * + * @adev: amdgpu_device pointer + * @source: irq source + * @entry: IV with WPTR update + * + * Update the WPTR from the IV and schedule work to handle the entries. + */ +static int ih_v6_0_self_irq(struct amdgpu_device *adev, + struct amdgpu_irq_src *source, + struct amdgpu_iv_entry *entry) +{ + uint32_t wptr = cpu_to_le32(entry->src_data[0]); + + switch (entry->ring_id) { + case 1: + *adev->irq.ih1.wptr_cpu = wptr; + schedule_work(&adev->irq.ih1_work); + break; + default: break; + } + return 0; +} + +static const struct amdgpu_irq_src_funcs ih_v6_0_self_irq_funcs = { + .process = ih_v6_0_self_irq, +}; + +static void ih_v6_0_set_self_irq_funcs(struct amdgpu_device *adev) +{ + adev->irq.self_irq.num_types = 0; + adev->irq.self_irq.funcs = &ih_v6_0_self_irq_funcs; +} + +static int ih_v6_0_early_init(void *handle) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + + ih_v6_0_set_interrupt_funcs(adev); + ih_v6_0_set_self_irq_funcs(adev); + return 0; +} + +static int ih_v6_0_sw_init(void *handle) +{ + int r; + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + bool use_bus_addr; + + r = amdgpu_irq_add_id(adev, SOC21_IH_CLIENTID_IH, 0, + &adev->irq.self_irq); + + if (r) + return r; + + /* use gpu virtual address for ih ring + * until ih_checken is programmed to allow + * use bus address for ih ring by psp bl */ + use_bus_addr = + (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) ? false : true; + r = amdgpu_ih_ring_init(adev, &adev->irq.ih, 256 * 1024, use_bus_addr); + if (r) + return r; + + adev->irq.ih.use_doorbell = true; + adev->irq.ih.doorbell_index = adev->doorbell_index.ih << 1; + + adev->irq.ih1.ring_size = 0; + adev->irq.ih2.ring_size = 0; + + /* initialize ih control register offset */ + ih_v6_0_init_register_offset(adev); + + r = amdgpu_ih_ring_init(adev, &adev->irq.ih_soft, PAGE_SIZE, true); + if (r) + return r; + + r = amdgpu_irq_init(adev); + + return r; +} + +static int ih_v6_0_sw_fini(void *handle) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + + amdgpu_irq_fini_sw(adev); + + return 0; +} + +static int ih_v6_0_hw_init(void *handle) +{ + int r; + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + + r = ih_v6_0_irq_init(adev); + if (r) + return r; + + return 0; +} + +static int ih_v6_0_hw_fini(void *handle) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + + ih_v6_0_irq_disable(adev); + + return 0; +} + +static int ih_v6_0_suspend(void *handle) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + + return ih_v6_0_hw_fini(adev); +} + +static int ih_v6_0_resume(void *handle) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + + return ih_v6_0_hw_init(adev); +} + +static bool ih_v6_0_is_idle(void *handle) +{ + /* todo */ + return true; +} + +static int ih_v6_0_wait_for_idle(void *handle) +{ + /* todo */ + return -ETIMEDOUT; +} + +static int ih_v6_0_soft_reset(void *handle) +{ + /* todo */ + return 0; +} + +static void ih_v6_0_update_clockgating_state(struct amdgpu_device *adev, + bool enable) +{ + uint32_t data, def, field_val; + + if (adev->cg_flags & AMD_CG_SUPPORT_IH_CG) { + def = data = RREG32_SOC15(OSSSYS, 0, regIH_CLK_CTRL); + field_val = enable ? 0 : 1; + data = REG_SET_FIELD(data, IH_CLK_CTRL, + DBUS_MUX_CLK_SOFT_OVERRIDE, field_val); + data = REG_SET_FIELD(data, IH_CLK_CTRL, + OSSSYS_SHARE_CLK_SOFT_OVERRIDE, field_val); + data = REG_SET_FIELD(data, IH_CLK_CTRL, + LIMIT_SMN_CLK_SOFT_OVERRIDE, field_val); + data = REG_SET_FIELD(data, IH_CLK_CTRL, + DYN_CLK_SOFT_OVERRIDE, field_val); + data = REG_SET_FIELD(data, IH_CLK_CTRL, + REG_CLK_SOFT_OVERRIDE, field_val); + if (def != data) + WREG32_SOC15(OSSSYS, 0, regIH_CLK_CTRL, data); + } + + return; +} + +static int ih_v6_0_set_clockgating_state(void *handle, + enum amd_clockgating_state state) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + + ih_v6_0_update_clockgating_state(adev, + state == AMD_CG_STATE_GATE); + return 0; +} + +static void ih_v6_0_update_ih_mem_power_gating(struct amdgpu_device *adev, + bool enable) +{ + uint32_t ih_mem_pwr_cntl; + + /* Disable ih sram power cntl before switch powergating mode */ + ih_mem_pwr_cntl = RREG32_SOC15(OSSSYS, 0, regIH_MEM_POWER_CTRL); + ih_mem_pwr_cntl = REG_SET_FIELD(ih_mem_pwr_cntl, IH_MEM_POWER_CTRL, + IH_BUFFER_MEM_POWER_CTRL_EN, 0); + WREG32_SOC15(OSSSYS, 0, regIH_MEM_POWER_CTRL, ih_mem_pwr_cntl); + + /* It is recommended to set mem powergating mode to DS mode */ + if (enable) { + /* mem power mode */ + ih_mem_pwr_cntl = REG_SET_FIELD(ih_mem_pwr_cntl, IH_MEM_POWER_CTRL, + IH_BUFFER_MEM_POWER_LS_EN, 0); + ih_mem_pwr_cntl = REG_SET_FIELD(ih_mem_pwr_cntl, IH_MEM_POWER_CTRL, + IH_BUFFER_MEM_POWER_DS_EN, 1); + ih_mem_pwr_cntl = REG_SET_FIELD(ih_mem_pwr_cntl, IH_MEM_POWER_CTRL, + IH_BUFFER_MEM_POWER_SD_EN, 0); + /* cam mem power mode */ + ih_mem_pwr_cntl = REG_SET_FIELD(ih_mem_pwr_cntl, IH_MEM_POWER_CTRL, + IH_RETRY_INT_CAM_MEM_POWER_LS_EN, 0); + ih_mem_pwr_cntl = REG_SET_FIELD(ih_mem_pwr_cntl, IH_MEM_POWER_CTRL, + IH_RETRY_INT_CAM_MEM_POWER_DS_EN, 1); + ih_mem_pwr_cntl = REG_SET_FIELD(ih_mem_pwr_cntl, IH_MEM_POWER_CTRL, + IH_RETRY_INT_CAM_MEM_POWER_SD_EN, 0); + /* re-enable power cntl */ + ih_mem_pwr_cntl = REG_SET_FIELD(ih_mem_pwr_cntl, IH_MEM_POWER_CTRL, + IH_BUFFER_MEM_POWER_CTRL_EN, 1); + } else { + /* mem power mode */ + ih_mem_pwr_cntl = REG_SET_FIELD(ih_mem_pwr_cntl, IH_MEM_POWER_CTRL, + IH_BUFFER_MEM_POWER_LS_EN, 0); + ih_mem_pwr_cntl = REG_SET_FIELD(ih_mem_pwr_cntl, IH_MEM_POWER_CTRL, + IH_BUFFER_MEM_POWER_DS_EN, 0); + ih_mem_pwr_cntl = REG_SET_FIELD(ih_mem_pwr_cntl, IH_MEM_POWER_CTRL, + IH_BUFFER_MEM_POWER_SD_EN, 0); + /* cam mem power mode */ + ih_mem_pwr_cntl = REG_SET_FIELD(ih_mem_pwr_cntl, IH_MEM_POWER_CTRL, + IH_RETRY_INT_CAM_MEM_POWER_LS_EN, 0); + ih_mem_pwr_cntl = REG_SET_FIELD(ih_mem_pwr_cntl, IH_MEM_POWER_CTRL, + IH_RETRY_INT_CAM_MEM_POWER_DS_EN, 0); + ih_mem_pwr_cntl = REG_SET_FIELD(ih_mem_pwr_cntl, IH_MEM_POWER_CTRL, + IH_RETRY_INT_CAM_MEM_POWER_SD_EN, 0); + /* re-enable power cntl*/ + ih_mem_pwr_cntl = REG_SET_FIELD(ih_mem_pwr_cntl, IH_MEM_POWER_CTRL, + IH_BUFFER_MEM_POWER_CTRL_EN, 1); + } + + WREG32_SOC15(OSSSYS, 0, regIH_MEM_POWER_CTRL, ih_mem_pwr_cntl); +} + +static int ih_v6_0_set_powergating_state(void *handle, + enum amd_powergating_state state) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + bool enable = (state == AMD_PG_STATE_GATE); + + if (adev->pg_flags & AMD_PG_SUPPORT_IH_SRAM_PG) + ih_v6_0_update_ih_mem_power_gating(adev, enable); + + return 0; +} + +static void ih_v6_0_get_clockgating_state(void *handle, u64 *flags) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + + if (!RREG32_SOC15(OSSSYS, 0, regIH_CLK_CTRL)) + *flags |= AMD_CG_SUPPORT_IH_CG; + + return; +} + +static const struct amd_ip_funcs ih_v6_0_ip_funcs = { + .name = "ih_v6_0", + .early_init = ih_v6_0_early_init, + .late_init = NULL, + .sw_init = ih_v6_0_sw_init, + .sw_fini = ih_v6_0_sw_fini, + .hw_init = ih_v6_0_hw_init, + .hw_fini = ih_v6_0_hw_fini, + .suspend = ih_v6_0_suspend, + .resume = ih_v6_0_resume, + .is_idle = ih_v6_0_is_idle, + .wait_for_idle = ih_v6_0_wait_for_idle, + .soft_reset = ih_v6_0_soft_reset, + .set_clockgating_state = ih_v6_0_set_clockgating_state, + .set_powergating_state = ih_v6_0_set_powergating_state, + .get_clockgating_state = ih_v6_0_get_clockgating_state, +}; + +static const struct amdgpu_ih_funcs ih_v6_0_funcs = { + .get_wptr = ih_v6_0_get_wptr, + .decode_iv = amdgpu_ih_decode_iv_helper, + .decode_iv_ts = amdgpu_ih_decode_iv_ts_helper, + .set_rptr = ih_v6_0_set_rptr +}; + +static void ih_v6_0_set_interrupt_funcs(struct amdgpu_device *adev) +{ + adev->irq.ih_funcs = &ih_v6_0_funcs; +} + +const struct amdgpu_ip_block_version ih_v6_0_ip_block = +{ + .type = AMD_IP_BLOCK_TYPE_IH, + .major = 6, + .minor = 0, + .rev = 0, + .funcs = &ih_v6_0_ip_funcs, +}; diff --git a/drivers/gpu/drm/amd/amdgpu/ih_v6_0.h b/drivers/gpu/drm/amd/amdgpu/ih_v6_0.h new file mode 100644 index 000000000000..f27b55580716 --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/ih_v6_0.h @@ -0,0 +1,28 @@ +/* + * Copyright 2021 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ +#ifndef __IH_V6_0_IH_H__ +#define __IH_V6_0_IH_H__ + +extern const struct amdgpu_ip_block_version ih_v6_0_ip_block; + +#endif diff --git a/drivers/gpu/drm/amd/amdgpu/imu_v11_0.c b/drivers/gpu/drm/amd/amdgpu/imu_v11_0.c new file mode 100644 index 000000000000..95548c512f4f --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/imu_v11_0.c @@ -0,0 +1,387 @@ +/* + * Copyright 2021 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#include <linux/firmware.h> +#include "amdgpu.h" +#include "amdgpu_imu.h" +#include "amdgpu_dpm.h" + +#include "imu_v11_0_3.h" + +#include "gc/gc_11_0_0_offset.h" +#include "gc/gc_11_0_0_sh_mask.h" + +MODULE_FIRMWARE("amdgpu/gc_11_0_0_imu.bin"); +MODULE_FIRMWARE("amdgpu/gc_11_0_1_imu.bin"); +MODULE_FIRMWARE("amdgpu/gc_11_0_2_imu.bin"); +MODULE_FIRMWARE("amdgpu/gc_11_0_3_imu.bin"); + +static int imu_v11_0_init_microcode(struct amdgpu_device *adev) +{ + char fw_name[40]; + char ucode_prefix[30]; + int err; + const struct imu_firmware_header_v1_0 *imu_hdr; + struct amdgpu_firmware_info *info = NULL; + + DRM_DEBUG("\n"); + + amdgpu_ucode_ip_version_decode(adev, GC_HWIP, ucode_prefix, sizeof(ucode_prefix)); + + snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_imu.bin", ucode_prefix); + err = request_firmware(&adev->gfx.imu_fw, fw_name, adev->dev); + if (err) + goto out; + err = amdgpu_ucode_validate(adev->gfx.imu_fw); + if (err) + goto out; + imu_hdr = (const struct imu_firmware_header_v1_0 *)adev->gfx.imu_fw->data; + adev->gfx.imu_fw_version = le32_to_cpu(imu_hdr->header.ucode_version); + //adev->gfx.imu_feature_version = le32_to_cpu(imu_hdr->ucode_feature_version); + + if (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) { + info = &adev->firmware.ucode[AMDGPU_UCODE_ID_IMU_I]; + info->ucode_id = AMDGPU_UCODE_ID_IMU_I; + info->fw = adev->gfx.imu_fw; + adev->firmware.fw_size += + ALIGN(le32_to_cpu(imu_hdr->imu_iram_ucode_size_bytes), PAGE_SIZE); + info = &adev->firmware.ucode[AMDGPU_UCODE_ID_IMU_D]; + info->ucode_id = AMDGPU_UCODE_ID_IMU_D; + info->fw = adev->gfx.imu_fw; + adev->firmware.fw_size += + ALIGN(le32_to_cpu(imu_hdr->imu_dram_ucode_size_bytes), PAGE_SIZE); + } + +out: + if (err) { + dev_err(adev->dev, + "gfx11: Failed to load firmware \"%s\"\n", + fw_name); + release_firmware(adev->gfx.imu_fw); + } + + return err; +} + +static int imu_v11_0_load_microcode(struct amdgpu_device *adev) +{ + const struct imu_firmware_header_v1_0 *hdr; + const __le32 *fw_data; + unsigned i, fw_size; + + if (!adev->gfx.imu_fw) + return -EINVAL; + + hdr = (const struct imu_firmware_header_v1_0 *)adev->gfx.imu_fw->data; + //amdgpu_ucode_print_rlc_hdr(&hdr->header); + + fw_data = (const __le32 *)(adev->gfx.imu_fw->data + + le32_to_cpu(hdr->header.ucode_array_offset_bytes)); + fw_size = le32_to_cpu(hdr->imu_iram_ucode_size_bytes) / 4; + + WREG32_SOC15(GC, 0, regGFX_IMU_I_RAM_ADDR, 0); + + for (i = 0; i < fw_size; i++) + WREG32_SOC15(GC, 0, regGFX_IMU_I_RAM_DATA, le32_to_cpup(fw_data++)); + + WREG32_SOC15(GC, 0, regGFX_IMU_I_RAM_ADDR, adev->gfx.imu_fw_version); + + fw_data = (const __le32 *)(adev->gfx.imu_fw->data + + le32_to_cpu(hdr->header.ucode_array_offset_bytes) + + le32_to_cpu(hdr->imu_iram_ucode_size_bytes)); + fw_size = le32_to_cpu(hdr->imu_dram_ucode_size_bytes) / 4; + + WREG32_SOC15(GC, 0, regGFX_IMU_D_RAM_ADDR, 0); + + for (i = 0; i < fw_size; i++) + WREG32_SOC15(GC, 0, regGFX_IMU_D_RAM_DATA, le32_to_cpup(fw_data++)); + + WREG32_SOC15(GC, 0, regGFX_IMU_D_RAM_ADDR, adev->gfx.imu_fw_version); + + return 0; +} + +static int imu_v11_0_wait_for_reset_status(struct amdgpu_device *adev) +{ + int i, imu_reg_val = 0; + + for (i = 0; i < adev->usec_timeout; i++) { + imu_reg_val = RREG32_SOC15(GC, 0, regGFX_IMU_GFX_RESET_CTRL); + if ((imu_reg_val & 0x1f) == 0x1f) + break; + udelay(1); + } + + if (i >= adev->usec_timeout) { + dev_err(adev->dev, "init imu: IMU start timeout\n"); + return -ETIMEDOUT; + } + + return 0; +} + +static void imu_v11_0_setup(struct amdgpu_device *adev) +{ + int imu_reg_val; + + //enable IMU debug mode + WREG32_SOC15(GC, 0, regGFX_IMU_C2PMSG_ACCESS_CTRL0, 0xffffff); + WREG32_SOC15(GC, 0, regGFX_IMU_C2PMSG_ACCESS_CTRL1, 0xffff); + + if (adev->gfx.imu.mode == DEBUG_MODE) { + imu_reg_val = RREG32_SOC15(GC, 0, regGFX_IMU_C2PMSG_16); + imu_reg_val |= 0x1; + WREG32_SOC15(GC, 0, regGFX_IMU_C2PMSG_16, imu_reg_val); + } + + //disble imu Rtavfs, SmsRepair, DfllBTC, and ClkB + imu_reg_val = RREG32_SOC15(GC, 0, regGFX_IMU_SCRATCH_10); + imu_reg_val |= 0x10007; + WREG32_SOC15(GC, 0, regGFX_IMU_SCRATCH_10, imu_reg_val); +} + +static int imu_v11_0_start(struct amdgpu_device *adev) +{ + int imu_reg_val; + + //Start IMU by set GFX_IMU_CORE_CTRL.CRESET = 0 + imu_reg_val = RREG32_SOC15(GC, 0, regGFX_IMU_CORE_CTRL); + imu_reg_val &= 0xfffffffe; + WREG32_SOC15(GC, 0, regGFX_IMU_CORE_CTRL, imu_reg_val); + + if (adev->flags & AMD_IS_APU) + amdgpu_dpm_set_gfx_power_up_by_imu(adev); + + return imu_v11_0_wait_for_reset_status(adev); +} + +static const struct imu_rlc_ram_golden imu_rlc_ram_golden_11[] = +{ + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGUS_IO_RD_COMBINE_FLUSH, 0x00055555, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGUS_IO_WR_COMBINE_FLUSH, 0x00055555, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGUS_DRAM_COMBINE_FLUSH, 0x00555555, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGUS_MISC2, 0x00001ffe, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGUS_SDP_CREDITS , 0x003f3fff, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGUS_SDP_TAG_RESERVE1, 0x00000000, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGUS_SDP_VCC_RESERVE0, 0x00041000, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGUS_SDP_VCC_RESERVE1, 0x00000000, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGUS_SDP_VCD_RESERVE0, 0x00040000, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGUS_SDP_VCD_RESERVE1, 0x00000000, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGUS_MISC, 0x00000017, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGUS_SDP_ENABLE, 0x00000001, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCEA_SDP_CREDITS , 0x003f3fbf, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCEA_SDP_TAG_RESERVE0, 0x10201000, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCEA_SDP_TAG_RESERVE1, 0x00000080, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCEA_SDP_VCC_RESERVE0, 0x1d041040, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCEA_SDP_VCC_RESERVE1, 0x80000000, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCEA_SDP_IO_PRIORITY, 0x88888888, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCEA_MAM_CTRL, 0x0000d800, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCEA_SDP_ARB_FINAL, 0x000003f7, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCEA_SDP_ENABLE, 0x00000001, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCVM_L2_PROTECTION_FAULT_CNTL2, 0x00020000, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCMC_VM_APT_CNTL, 0x0000000c, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCMC_VM_CACHEABLE_DRAM_ADDRESS_END, 0x000fffff, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCEA_MISC, 0x0c48bff0, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regCC_GC_SA_UNIT_DISABLE, 0x00fffc01, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regCC_GC_PRIM_CONFIG, 0x000fffe1, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regCC_RB_BACKEND_DISABLE, 0x0fffff01, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regCC_GC_SHADER_ARRAY_CONFIG, 0xfffe0001, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCMC_VM_MX_L1_TLB_CNTL, 0x00000500, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCMC_VM_SYSTEM_APERTURE_LOW_ADDR, 0x00000001, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCMC_VM_SYSTEM_APERTURE_HIGH_ADDR, 0x00000000, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCMC_VM_LOCAL_FB_ADDRESS_START, 0x00000000, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCMC_VM_LOCAL_FB_ADDRESS_END, 0x000fffff, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCVM_CONTEXT0_CNTL, 0x00000000, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCVM_CONTEXT1_CNTL, 0x00000000, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCMC_VM_NB_TOP_OF_DRAM_SLOT1, 0xff800000, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCMC_VM_NB_LOWER_TOP_OF_DRAM2, 0x00000001, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCMC_VM_NB_UPPER_TOP_OF_DRAM2, 0x00000fff, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCVM_L2_PROTECTION_FAULT_CNTL, 0x00001ffc, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCMC_VM_MX_L1_TLB_CNTL, 0x00000501, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCVM_L2_CNTL, 0x00080603, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCVM_L2_CNTL2, 0x00000003, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCVM_L2_CNTL3, 0x00100003, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCVM_L2_CNTL5, 0x00003fe0, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCVM_CONTEXT0_CNTL, 0x00000001, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCVM_L2_CONTEXT0_PER_PFVF_PTE_CACHE_FRAGMENT_SIZES, 0x00000c00, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCVM_CONTEXT1_CNTL, 0x00000001, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCVM_L2_CONTEXT1_PER_PFVF_PTE_CACHE_FRAGMENT_SIZES, 0x00000c00, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGB_ADDR_CONFIG, 0x00000545, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGL2_PIPE_STEER_0, 0x13455431, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGL2_PIPE_STEER_1, 0x13455431, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGL2_PIPE_STEER_2, 0x76027602, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGL2_PIPE_STEER_3, 0x76207620, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGB_ADDR_CONFIG, 0x00000345, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCUTCL2_HARVEST_BYPASS_GROUPS, 0x0000003e, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCMC_VM_FB_LOCATION_BASE, 0x00006000, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCMC_VM_FB_LOCATION_TOP, 0x000061ff, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCMC_VM_APT_CNTL, 0x0000000c, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCMC_VM_AGP_BASE, 0x00000000, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCMC_VM_AGP_BOT, 0x00000002, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCMC_VM_AGP_TOP, 0x00000000, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCVM_L2_PROTECTION_FAULT_CNTL2, 0x00020000, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regSDMA0_UCODE_SELFLOAD_CONTROL, 0x00000210, 0), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regSDMA1_UCODE_SELFLOAD_CONTROL, 0x00000210, 0), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regCPC_PSP_DEBUG, CPC_PSP_DEBUG__GPA_OVERRIDE_MASK, 0), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regCPG_PSP_DEBUG, CPG_PSP_DEBUG__GPA_OVERRIDE_MASK, 0) +}; + +static const struct imu_rlc_ram_golden imu_rlc_ram_golden_11_0_2[] = +{ + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCEA_MISC, 0x0c48bff0, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCEA_SDP_CREDITS, 0x003f3fbf, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCEA_SDP_TAG_RESERVE0, 0x10200800, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCEA_SDP_TAG_RESERVE1, 0x00000088, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCEA_SDP_VCC_RESERVE0, 0x1d041040, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCEA_SDP_VCC_RESERVE1, 0x80000000, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCEA_SDP_IO_PRIORITY, 0x88888888, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCEA_MAM_CTRL, 0x0000d800, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCEA_SDP_ARB_FINAL, 0x000007ef, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCEA_DRAM_PAGE_BURST, 0x20080200, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCEA_SDP_ENABLE, 0x00000001, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCMC_VM_APT_CNTL, 0x0000000c, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCMC_VM_CACHEABLE_DRAM_ADDRESS_END, 0x000fffff, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGUS_IO_RD_COMBINE_FLUSH, 0x00055555, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGUS_IO_WR_COMBINE_FLUSH, 0x00055555, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGUS_DRAM_COMBINE_FLUSH, 0x00555555, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGUS_MISC2, 0x00001ffe, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGUS_SDP_CREDITS, 0x003f3fff, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGUS_SDP_TAG_RESERVE1, 0x00000000, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGUS_SDP_VCC_RESERVE0, 0x00041000, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGUS_SDP_VCC_RESERVE1, 0x00000000, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGUS_SDP_VCD_RESERVE0, 0x00040000, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGUS_SDP_VCD_RESERVE1, 0x00000000, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGUS_MISC, 0x00000017, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGUS_SDP_ENABLE, 0x00000001, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regCC_GC_SA_UNIT_DISABLE, 0x00fffc01, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regCC_GC_PRIM_CONFIG, 0x000fffe1, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regCC_RB_BACKEND_DISABLE, 0x00000f01, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regCC_GC_SHADER_ARRAY_CONFIG, 0xfffe0001, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGL1_PIPE_STEER, 0x000000e4, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regCH_PIPE_STEER, 0x000000e4, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGL2_PIPE_STEER_0, 0x01231023, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGB_ADDR_CONFIG, 0x00000243, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCUTCL2_HARVEST_BYPASS_GROUPS, 0x00000002, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCMC_VM_MX_L1_TLB_CNTL, 0x00000500, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCMC_VM_SYSTEM_APERTURE_LOW_ADDR, 0x00000001, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCMC_VM_SYSTEM_APERTURE_HIGH_ADDR, 0x00000000, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCMC_VM_LOCAL_FB_ADDRESS_START, 0x00000000, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCMC_VM_LOCAL_FB_ADDRESS_END, 0x000001ff, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCMC_VM_FB_LOCATION_BASE, 0x00006000, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCMC_VM_FB_LOCATION_TOP, 0x000061ff, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCVM_CONTEXT0_CNTL, 0x00000000, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCVM_CONTEXT1_CNTL, 0x00000000, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCMC_VM_APT_CNTL, 0x0000000c, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCMC_VM_NB_TOP_OF_DRAM_SLOT1, 0xff800000, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCMC_VM_NB_LOWER_TOP_OF_DRAM2, 0x00000001, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCMC_VM_NB_UPPER_TOP_OF_DRAM2, 0x00000fff, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCMC_VM_AGP_BASE, 0x00000000, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCMC_VM_AGP_BOT, 0x00000002, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCMC_VM_AGP_TOP, 0x00000000, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCVM_L2_PROTECTION_FAULT_CNTL, 0x00001ffc, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCVM_L2_PROTECTION_FAULT_CNTL2, 0x00002825, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCMC_VM_MX_L1_TLB_CNTL, 0x00000501, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCVM_L2_CNTL, 0x00080603, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCVM_L2_CNTL2, 0x00000003, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCVM_L2_CNTL3, 0x00100003, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCVM_L2_CNTL5, 0x00003fe0, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCVM_CONTEXT0_CNTL, 0x00000001, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCVM_L2_CONTEXT0_PER_PFVF_PTE_CACHE_FRAGMENT_SIZES, 0x00000c00, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCVM_CONTEXT1_CNTL, 0x00000001, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCVM_L2_CONTEXT1_PER_PFVF_PTE_CACHE_FRAGMENT_SIZES, 0x00000c00, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regSDMA0_UCODE_SELFLOAD_CONTROL, 0x00000210, 0), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regSDMA1_UCODE_SELFLOAD_CONTROL, 0x00000210, 0), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regCPC_PSP_DEBUG, CPC_PSP_DEBUG__GPA_OVERRIDE_MASK, 0), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regCPG_PSP_DEBUG, CPG_PSP_DEBUG__GPA_OVERRIDE_MASK, 0) +}; + +static void program_imu_rlc_ram(struct amdgpu_device *adev, + const struct imu_rlc_ram_golden *regs, + const u32 array_size) +{ + const struct imu_rlc_ram_golden *entry; + u32 reg, data; + int i; + + for (i = 0; i < array_size; ++i) { + entry = ®s[i]; + reg = adev->reg_offset[entry->hwip][entry->instance][entry->segment] + entry->reg; + reg |= entry->addr_mask; + + data = entry->data; + if (entry->reg == regGCMC_VM_AGP_BASE) + data = 0x00ffffff; + else if (entry->reg == regGCMC_VM_AGP_TOP) + data = 0x0; + else if (entry->reg == regGCMC_VM_FB_LOCATION_BASE) + data = adev->gmc.vram_start >> 24; + else if (entry->reg == regGCMC_VM_FB_LOCATION_TOP) + data = adev->gmc.vram_end >> 24; + + WREG32_SOC15(GC, 0, regGFX_IMU_RLC_RAM_ADDR_HIGH, 0); + WREG32_SOC15(GC, 0, regGFX_IMU_RLC_RAM_ADDR_LOW, reg); + WREG32_SOC15(GC, 0, regGFX_IMU_RLC_RAM_DATA, data); + } + //Indicate the latest entry + WREG32_SOC15(GC, 0, regGFX_IMU_RLC_RAM_ADDR_HIGH, 0); + WREG32_SOC15(GC, 0, regGFX_IMU_RLC_RAM_ADDR_LOW, 0); + WREG32_SOC15(GC, 0, regGFX_IMU_RLC_RAM_DATA, 0); +} + +static void imu_v11_0_program_rlc_ram(struct amdgpu_device *adev) +{ + u32 reg_data; + + WREG32_SOC15(GC, 0, regGFX_IMU_RLC_RAM_INDEX, 0x2); + + switch (adev->ip_versions[GC_HWIP][0]) { + case IP_VERSION(11, 0, 0): + program_imu_rlc_ram(adev, imu_rlc_ram_golden_11, + (const u32)ARRAY_SIZE(imu_rlc_ram_golden_11)); + break; + case IP_VERSION(11, 0, 2): + program_imu_rlc_ram(adev, imu_rlc_ram_golden_11_0_2, + (const u32)ARRAY_SIZE(imu_rlc_ram_golden_11_0_2)); + break; + case IP_VERSION(11, 0, 3): + imu_v11_0_3_program_rlc_ram(adev); + break; + default: + BUG(); + break; + } + + //Indicate the contents of the RAM are valid + reg_data = RREG32_SOC15(GC, 0, regGFX_IMU_RLC_RAM_INDEX); + reg_data |= GFX_IMU_RLC_RAM_INDEX__RAM_VALID_MASK; + WREG32_SOC15(GC, 0, regGFX_IMU_RLC_RAM_INDEX, reg_data); +} + +const struct amdgpu_imu_funcs gfx_v11_0_imu_funcs = { + .init_microcode = imu_v11_0_init_microcode, + .load_microcode = imu_v11_0_load_microcode, + .setup_imu = imu_v11_0_setup, + .start_imu = imu_v11_0_start, + .program_rlc_ram = imu_v11_0_program_rlc_ram, + .wait_for_reset_status = imu_v11_0_wait_for_reset_status, +}; diff --git a/drivers/gpu/drm/amd/amdgpu/imu_v11_0.h b/drivers/gpu/drm/amd/amdgpu/imu_v11_0.h new file mode 100644 index 000000000000..e71f96fc2f06 --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/imu_v11_0.h @@ -0,0 +1,30 @@ +/* + * Copyright 2021 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#ifndef __IMU_V11_0_H__ +#define __IMU_V11_0_H__ + +extern const struct amdgpu_imu_funcs gfx_v11_0_imu_funcs; + +#endif + diff --git a/drivers/gpu/drm/amd/amdgpu/imu_v11_0_3.c b/drivers/gpu/drm/amd/amdgpu/imu_v11_0_3.c new file mode 100644 index 000000000000..fc69c1a29e23 --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/imu_v11_0_3.c @@ -0,0 +1,145 @@ +/* + * Copyright 2022 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ +#include "amdgpu.h" +#include "amdgpu_imu.h" +#include "imu_v11_0_3.h" + +#include "gc/gc_11_0_3_offset.h" +#include "gc/gc_11_0_3_sh_mask.h" + +static const struct imu_rlc_ram_golden imu_rlc_ram_golden_11_0_3[] = { + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGUS_IO_RD_COMBINE_FLUSH, 0x00055555, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGUS_IO_WR_COMBINE_FLUSH, 0x00055555, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGUS_DRAM_COMBINE_FLUSH, 0x00555555, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGUS_MISC2, 0x00001ffe, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGUS_SDP_CREDITS, 0x003f3fff, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGUS_SDP_TAG_RESERVE1, 0x00000000, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGUS_SDP_VCC_RESERVE0, 0x00041000, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGUS_SDP_VCC_RESERVE1, 0x00000000, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGUS_SDP_VCD_RESERVE0, 0x00040000, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGUS_SDP_VCD_RESERVE1, 0x00000000, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGUS_MISC, 0x00000017, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGUS_SDP_ENABLE, 0x00000001, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCEA_SDP_CREDITS, 0x003f3fbf, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCEA_SDP_TAG_RESERVE0, 0x10200800, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCEA_SDP_TAG_RESERVE1, 0x00000088, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCEA_SDP_VCC_RESERVE0, 0x1d041040, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCEA_SDP_VCC_RESERVE1, 0x80000000, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCEA_SDP_IO_PRIORITY, 0x88888888, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCEA_MAM_CTRL, 0x0000d800, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCEA_SDP_ARB_FINAL, 0x000007ff, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCEA_DRAM_PAGE_BURST, 0x20080200, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCEA_SDP_ENABLE, 0x00000001, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCVM_L2_PROTECTION_FAULT_CNTL2, 0x00020000, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCMC_VM_APT_CNTL, 0x0000000c, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCMC_VM_CACHEABLE_DRAM_ADDRESS_END, 0x000fffff, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCEA_MISC, 0x0c48bff0, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regCC_GC_SA_UNIT_DISABLE, 0x00fffc01, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regCC_GC_PRIM_CONFIG, 0x000fffe1, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regCC_RB_BACKEND_DISABLE, 0xffffff01, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regCC_GC_SHADER_ARRAY_CONFIG, 0xfffe0001, 0x40000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regCC_GC_SHADER_ARRAY_CONFIG, 0xfffe0001, 0x42000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regCC_GC_SHADER_ARRAY_CONFIG, 0xffff0001, 0x44000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regCC_GC_SHADER_ARRAY_CONFIG, 0xffff0001, 0x46000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regCC_GC_SHADER_ARRAY_CONFIG, 0xffff0001, 0x48000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regCC_GC_SHADER_ARRAY_CONFIG, 0xffff0001, 0x4A000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regCGTS_TCC_DISABLE, 0x00000001, 0x00000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regCC_GC_SHADER_RATE_CONFIG, 0x00000001, 0x00000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regCC_GC_EDC_CONFIG, 0x00000001, 0x00000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCMC_VM_MX_L1_TLB_CNTL, 0x00000500, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCMC_VM_SYSTEM_APERTURE_LOW_ADDR, 0x00000001, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCMC_VM_SYSTEM_APERTURE_HIGH_ADDR, 0x00000000, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCMC_VM_LOCAL_FB_ADDRESS_START, 0x00000000, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCMC_VM_LOCAL_FB_ADDRESS_END, 0x000005ff, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCMC_VM_FB_LOCATION_BASE, 0x00006000, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCMC_VM_FB_LOCATION_TOP, 0x000065ff, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCVM_CONTEXT0_CNTL, 0x00000000, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCVM_CONTEXT1_CNTL, 0x00000000, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCMC_VM_NB_TOP_OF_DRAM_SLOT1, 0xff800000, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCMC_VM_NB_LOWER_TOP_OF_DRAM2, 0x00000001, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCMC_VM_NB_UPPER_TOP_OF_DRAM2, 0x00000fff, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCVM_L2_PROTECTION_FAULT_CNTL, 0x00001ffc, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCMC_VM_MX_L1_TLB_CNTL, 0x00000551, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCVM_L2_CNTL, 0x00080603, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCVM_L2_CNTL2, 0x00000003, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCVM_L2_CNTL3, 0x00100003, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCVM_L2_CNTL5, 0x00003fe0, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCVM_CONTEXT0_CNTL, 0x00000001, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCVM_L2_CONTEXT0_PER_PFVF_PTE_CACHE_FRAGMENT_SIZES, 0x00000c00, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCVM_CONTEXT1_CNTL, 0x00000001, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCVM_L2_CONTEXT1_PER_PFVF_PTE_CACHE_FRAGMENT_SIZES, 0x00000c00, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGB_ADDR_CONFIG, 0x00000444, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGL2_PIPE_STEER_0, 0x54105410, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGL2_PIPE_STEER_2, 0x76323276, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGB_ADDR_CONFIG, 0x00000244, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCUTCL2_HARVEST_BYPASS_GROUPS, 0x00000006, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCMC_VM_APT_CNTL, 0x0000000c, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCMC_VM_AGP_BASE, 0x00000000, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCMC_VM_AGP_BOT, 0x00000002, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCMC_VM_AGP_TOP, 0x00000000, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCVM_L2_PROTECTION_FAULT_CNTL2, 0x00020000, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regSDMA0_UCODE_SELFLOAD_CONTROL, 0x00000210, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regSDMA1_UCODE_SELFLOAD_CONTROL, 0x00000210, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regCPC_PSP_DEBUG, CPC_PSP_DEBUG__GPA_OVERRIDE_MASK, 0xe0000000), + IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regCPG_PSP_DEBUG, CPG_PSP_DEBUG__GPA_OVERRIDE_MASK, 0xe0000000), +}; + +static void program_rlc_ram_register_setting(struct amdgpu_device *adev, + const struct imu_rlc_ram_golden *regs, + const u32 array_size) +{ + const struct imu_rlc_ram_golden *entry; + u32 reg, data; + int i; + + for (i = 0; i < array_size; ++i) { + entry = ®s[i]; + reg = adev->reg_offset[entry->hwip][entry->instance][entry->segment] + entry->reg; + reg |= entry->addr_mask; + + data = entry->data; + if (entry->reg == regGCMC_VM_AGP_BASE) + data = 0x00ffffff; + else if (entry->reg == regGCMC_VM_AGP_TOP) + data = 0x0; + else if (entry->reg == regGCMC_VM_FB_LOCATION_BASE) + data = adev->gmc.vram_start >> 24; + else if (entry->reg == regGCMC_VM_FB_LOCATION_TOP) + data = adev->gmc.vram_end >> 24; + + WREG32_SOC15(GC, 0, regGFX_IMU_RLC_RAM_ADDR_HIGH, 0); + WREG32_SOC15(GC, 0, regGFX_IMU_RLC_RAM_ADDR_LOW, reg); + WREG32_SOC15(GC, 0, regGFX_IMU_RLC_RAM_DATA, data); + } + //Indicate the latest entry + WREG32_SOC15(GC, 0, regGFX_IMU_RLC_RAM_ADDR_HIGH, 0); + WREG32_SOC15(GC, 0, regGFX_IMU_RLC_RAM_ADDR_LOW, 0); + WREG32_SOC15(GC, 0, regGFX_IMU_RLC_RAM_DATA, 0); +} + +void imu_v11_0_3_program_rlc_ram(struct amdgpu_device *adev) +{ + program_rlc_ram_register_setting(adev, + imu_rlc_ram_golden_11_0_3, + (const u32)ARRAY_SIZE(imu_rlc_ram_golden_11_0_3)); +} diff --git a/drivers/gpu/drm/amd/amdgpu/imu_v11_0_3.h b/drivers/gpu/drm/amd/amdgpu/imu_v11_0_3.h new file mode 100644 index 000000000000..702be568f26b --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/imu_v11_0_3.h @@ -0,0 +1,29 @@ +/* + * Copyright 2022 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#ifndef __IMU_V11_0_3_H__ +#define __IMU_V11_0_3_H__ + +void imu_v11_0_3_program_rlc_ram(struct amdgpu_device *adev); + +#endif diff --git a/drivers/gpu/drm/amd/amdgpu/jpeg_v2_0.c b/drivers/gpu/drm/amd/amdgpu/jpeg_v2_0.c index 299de1d131d8..f3c1af5130ab 100644 --- a/drivers/gpu/drm/amd/amdgpu/jpeg_v2_0.c +++ b/drivers/gpu/drm/amd/amdgpu/jpeg_v2_0.c @@ -407,7 +407,7 @@ static uint64_t jpeg_v2_0_dec_ring_get_wptr(struct amdgpu_ring *ring) struct amdgpu_device *adev = ring->adev; if (ring->use_doorbell) - return adev->wb.wb[ring->wptr_offs]; + return *ring->wptr_cpu_addr; else return RREG32_SOC15(JPEG, 0, mmUVD_JRBC_RB_WPTR); } @@ -424,7 +424,7 @@ static void jpeg_v2_0_dec_ring_set_wptr(struct amdgpu_ring *ring) struct amdgpu_device *adev = ring->adev; if (ring->use_doorbell) { - adev->wb.wb[ring->wptr_offs] = lower_32_bits(ring->wptr); + *ring->wptr_cpu_addr = lower_32_bits(ring->wptr); WDOORBELL32(ring->doorbell_index, lower_32_bits(ring->wptr)); } else { WREG32_SOC15(JPEG, 0, mmUVD_JRBC_RB_WPTR, lower_32_bits(ring->wptr)); @@ -535,6 +535,10 @@ void jpeg_v2_0_dec_ring_emit_ib(struct amdgpu_ring *ring, { unsigned vmid = AMDGPU_JOB_GET_VMID(job); + amdgpu_ring_write(ring, PACKETJ(mmUVD_JPEG_IH_CTRL_INTERNAL_OFFSET, + 0, 0, PACKETJ_TYPE0)); + amdgpu_ring_write(ring, (vmid << JPEG_IH_CTRL__IH_VMID__SHIFT)); + amdgpu_ring_write(ring, PACKETJ(mmUVD_LMI_JRBC_IB_VMID_INTERNAL_OFFSET, 0, 0, PACKETJ_TYPE0)); amdgpu_ring_write(ring, (vmid | (vmid << 4))); @@ -768,7 +772,7 @@ static const struct amdgpu_ring_funcs jpeg_v2_0_dec_ring_vm_funcs = { 8 + /* jpeg_v2_0_dec_ring_emit_vm_flush */ 18 + 18 + /* jpeg_v2_0_dec_ring_emit_fence x2 vm fence */ 8 + 16, - .emit_ib_size = 22, /* jpeg_v2_0_dec_ring_emit_ib */ + .emit_ib_size = 24, /* jpeg_v2_0_dec_ring_emit_ib */ .emit_ib = jpeg_v2_0_dec_ring_emit_ib, .emit_fence = jpeg_v2_0_dec_ring_emit_fence, .emit_vm_flush = jpeg_v2_0_dec_ring_emit_vm_flush, diff --git a/drivers/gpu/drm/amd/amdgpu/jpeg_v2_0.h b/drivers/gpu/drm/amd/amdgpu/jpeg_v2_0.h index 1a03baa59755..654e43e83e2c 100644 --- a/drivers/gpu/drm/amd/amdgpu/jpeg_v2_0.h +++ b/drivers/gpu/drm/amd/amdgpu/jpeg_v2_0.h @@ -41,6 +41,7 @@ #define mmUVD_JRBC_RB_REF_DATA_INTERNAL_OFFSET 0x4084 #define mmUVD_JRBC_STATUS_INTERNAL_OFFSET 0x4089 #define mmUVD_JPEG_PITCH_INTERNAL_OFFSET 0x401f +#define mmUVD_JPEG_IH_CTRL_INTERNAL_OFFSET 0x4149 #define JRBC_DEC_EXTERNAL_REG_WRITE_ADDR 0x18000 diff --git a/drivers/gpu/drm/amd/amdgpu/jpeg_v2_5.c b/drivers/gpu/drm/amd/amdgpu/jpeg_v2_5.c index a29c86617fb5..f87d0f6ffc93 100644 --- a/drivers/gpu/drm/amd/amdgpu/jpeg_v2_5.c +++ b/drivers/gpu/drm/amd/amdgpu/jpeg_v2_5.c @@ -26,6 +26,7 @@ #include "soc15.h" #include "soc15d.h" #include "jpeg_v2_0.h" +#include "jpeg_v2_5.h" #include "vcn/vcn_2_5_offset.h" #include "vcn/vcn_2_5_sh_mask.h" @@ -39,6 +40,7 @@ static void jpeg_v2_5_set_dec_ring_funcs(struct amdgpu_device *adev); static void jpeg_v2_5_set_irq_funcs(struct amdgpu_device *adev); static int jpeg_v2_5_set_powergating_state(void *handle, enum amd_powergating_state state); +static void jpeg_v2_5_set_ras_funcs(struct amdgpu_device *adev); static int amdgpu_ih_clientid_jpeg[] = { SOC15_IH_CLIENTID_VCN, @@ -70,6 +72,7 @@ static int jpeg_v2_5_early_init(void *handle) jpeg_v2_5_set_dec_ring_funcs(adev); jpeg_v2_5_set_irq_funcs(adev); + jpeg_v2_5_set_ras_funcs(adev); return 0; } @@ -96,6 +99,18 @@ static int jpeg_v2_5_sw_init(void *handle) VCN_2_0__SRCID__JPEG_DECODE, &adev->jpeg.inst[i].irq); if (r) return r; + + /* JPEG DJPEG POISON EVENT */ + r = amdgpu_irq_add_id(adev, amdgpu_ih_clientid_jpeg[i], + VCN_2_6__SRCID_DJPEG0_POISON, &adev->jpeg.inst[i].irq); + if (r) + return r; + + /* JPEG EJPEG POISON EVENT */ + r = amdgpu_irq_add_id(adev, amdgpu_ih_clientid_jpeg[i], + VCN_2_6__SRCID_EJPEG0_POISON, &adev->jpeg.inst[i].irq); + if (r) + return r; } r = amdgpu_jpeg_sw_init(adev); @@ -399,7 +414,7 @@ static uint64_t jpeg_v2_5_dec_ring_get_wptr(struct amdgpu_ring *ring) struct amdgpu_device *adev = ring->adev; if (ring->use_doorbell) - return adev->wb.wb[ring->wptr_offs]; + return *ring->wptr_cpu_addr; else return RREG32_SOC15(JPEG, ring->me, mmUVD_JRBC_RB_WPTR); } @@ -416,7 +431,7 @@ static void jpeg_v2_5_dec_ring_set_wptr(struct amdgpu_ring *ring) struct amdgpu_device *adev = ring->adev; if (ring->use_doorbell) { - adev->wb.wb[ring->wptr_offs] = lower_32_bits(ring->wptr); + *ring->wptr_cpu_addr = lower_32_bits(ring->wptr); WDOORBELL32(ring->doorbell_index, lower_32_bits(ring->wptr)); } else { WREG32_SOC15(JPEG, ring->me, mmUVD_JRBC_RB_WPTR, lower_32_bits(ring->wptr)); @@ -570,6 +585,10 @@ static int jpeg_v2_5_process_interrupt(struct amdgpu_device *adev, case VCN_2_0__SRCID__JPEG_DECODE: amdgpu_fence_process(&adev->jpeg.inst[ip_instance].ring_dec); break; + case VCN_2_6__SRCID_DJPEG0_POISON: + case VCN_2_6__SRCID_EJPEG0_POISON: + amdgpu_jpeg_process_poison_irq(adev, source, entry); + break; default: DRM_ERROR("Unhandled interrupt: %d %d\n", entry->src_id, entry->src_data[0]); @@ -730,3 +749,74 @@ const struct amdgpu_ip_block_version jpeg_v2_6_ip_block = .rev = 0, .funcs = &jpeg_v2_6_ip_funcs, }; + +static uint32_t jpeg_v2_6_query_poison_by_instance(struct amdgpu_device *adev, + uint32_t instance, uint32_t sub_block) +{ + uint32_t poison_stat = 0, reg_value = 0; + + switch (sub_block) { + case AMDGPU_JPEG_V2_6_JPEG0: + reg_value = RREG32_SOC15(JPEG, instance, mmUVD_RAS_JPEG0_STATUS); + poison_stat = REG_GET_FIELD(reg_value, UVD_RAS_JPEG0_STATUS, POISONED_PF); + break; + case AMDGPU_JPEG_V2_6_JPEG1: + reg_value = RREG32_SOC15(JPEG, instance, mmUVD_RAS_JPEG1_STATUS); + poison_stat = REG_GET_FIELD(reg_value, UVD_RAS_JPEG1_STATUS, POISONED_PF); + break; + default: + break; + } + + if (poison_stat) + dev_info(adev->dev, "Poison detected in JPEG%d sub_block%d\n", + instance, sub_block); + + return poison_stat; +} + +static bool jpeg_v2_6_query_ras_poison_status(struct amdgpu_device *adev) +{ + uint32_t inst = 0, sub = 0, poison_stat = 0; + + for (inst = 0; inst < adev->jpeg.num_jpeg_inst; inst++) + for (sub = 0; sub < AMDGPU_JPEG_V2_6_MAX_SUB_BLOCK; sub++) + poison_stat += + jpeg_v2_6_query_poison_by_instance(adev, inst, sub); + + return !!poison_stat; +} + +const struct amdgpu_ras_block_hw_ops jpeg_v2_6_ras_hw_ops = { + .query_poison_status = jpeg_v2_6_query_ras_poison_status, +}; + +static struct amdgpu_jpeg_ras jpeg_v2_6_ras = { + .ras_block = { + .hw_ops = &jpeg_v2_6_ras_hw_ops, + }, +}; + +static void jpeg_v2_5_set_ras_funcs(struct amdgpu_device *adev) +{ + switch (adev->ip_versions[JPEG_HWIP][0]) { + case IP_VERSION(2, 6, 0): + adev->jpeg.ras = &jpeg_v2_6_ras; + break; + default: + break; + } + + if (adev->jpeg.ras) { + amdgpu_ras_register_ras_block(adev, &adev->jpeg.ras->ras_block); + + strcpy(adev->jpeg.ras->ras_block.ras_comm.name, "jpeg"); + adev->jpeg.ras->ras_block.ras_comm.block = AMDGPU_RAS_BLOCK__JPEG; + adev->jpeg.ras->ras_block.ras_comm.type = AMDGPU_RAS_ERROR__POISON; + adev->jpeg.ras_if = &adev->jpeg.ras->ras_block.ras_comm; + + /* If don't define special ras_late_init function, use default ras_late_init */ + if (!adev->jpeg.ras->ras_block.ras_late_init) + adev->jpeg.ras->ras_block.ras_late_init = amdgpu_ras_block_late_init; + } +} diff --git a/drivers/gpu/drm/amd/amdgpu/jpeg_v2_5.h b/drivers/gpu/drm/amd/amdgpu/jpeg_v2_5.h index 3b0aa29b9879..1e858c6cdf13 100644 --- a/drivers/gpu/drm/amd/amdgpu/jpeg_v2_5.h +++ b/drivers/gpu/drm/amd/amdgpu/jpeg_v2_5.h @@ -24,6 +24,13 @@ #ifndef __JPEG_V2_5_H__ #define __JPEG_V2_5_H__ +enum amdgpu_jpeg_v2_6_sub_block { + AMDGPU_JPEG_V2_6_JPEG0 = 0, + AMDGPU_JPEG_V2_6_JPEG1, + + AMDGPU_JPEG_V2_6_MAX_SUB_BLOCK, +}; + extern const struct amdgpu_ip_block_version jpeg_v2_5_ip_block; extern const struct amdgpu_ip_block_version jpeg_v2_6_ip_block; diff --git a/drivers/gpu/drm/amd/amdgpu/jpeg_v3_0.c b/drivers/gpu/drm/amd/amdgpu/jpeg_v3_0.c index 01c242c5abc3..a1b751d9ac06 100644 --- a/drivers/gpu/drm/amd/amdgpu/jpeg_v3_0.c +++ b/drivers/gpu/drm/amd/amdgpu/jpeg_v3_0.c @@ -50,11 +50,16 @@ static int jpeg_v3_0_early_init(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; - if (adev->asic_type != CHIP_YELLOW_CARP) { - u32 harvest = RREG32_SOC15(JPEG, 0, mmCC_UVD_HARVESTING); + u32 harvest; + switch (adev->ip_versions[UVD_HWIP][0]) { + case IP_VERSION(3, 1, 1): + break; + default: + harvest = RREG32_SOC15(JPEG, 0, mmCC_UVD_HARVESTING); if (harvest & CC_UVD_HARVESTING__UVD_DISABLE_MASK) return -ENOENT; + break; } adev->jpeg.num_jpeg_inst = 1; @@ -422,7 +427,7 @@ static uint64_t jpeg_v3_0_dec_ring_get_wptr(struct amdgpu_ring *ring) struct amdgpu_device *adev = ring->adev; if (ring->use_doorbell) - return adev->wb.wb[ring->wptr_offs]; + return *ring->wptr_cpu_addr; else return RREG32_SOC15(JPEG, 0, mmUVD_JRBC_RB_WPTR); } @@ -439,7 +444,7 @@ static void jpeg_v3_0_dec_ring_set_wptr(struct amdgpu_ring *ring) struct amdgpu_device *adev = ring->adev; if (ring->use_doorbell) { - adev->wb.wb[ring->wptr_offs] = lower_32_bits(ring->wptr); + *ring->wptr_cpu_addr = lower_32_bits(ring->wptr); WDOORBELL32(ring->doorbell_index, lower_32_bits(ring->wptr)); } else { WREG32_SOC15(JPEG, 0, mmUVD_JRBC_RB_WPTR, lower_32_bits(ring->wptr)); diff --git a/drivers/gpu/drm/amd/amdgpu/jpeg_v4_0.c b/drivers/gpu/drm/amd/amdgpu/jpeg_v4_0.c new file mode 100644 index 000000000000..63b0d0b810ec --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/jpeg_v4_0.c @@ -0,0 +1,609 @@ +/* + * Copyright 2021 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#include "amdgpu.h" +#include "amdgpu_jpeg.h" +#include "amdgpu_pm.h" +#include "soc15.h" +#include "soc15d.h" +#include "jpeg_v2_0.h" + +#include "vcn/vcn_4_0_0_offset.h" +#include "vcn/vcn_4_0_0_sh_mask.h" +#include "ivsrcid/vcn/irqsrcs_vcn_4_0.h" + +#define regUVD_JPEG_PITCH_INTERNAL_OFFSET 0x401f + +static void jpeg_v4_0_set_dec_ring_funcs(struct amdgpu_device *adev); +static void jpeg_v4_0_set_irq_funcs(struct amdgpu_device *adev); +static int jpeg_v4_0_set_powergating_state(void *handle, + enum amd_powergating_state state); + +/** + * jpeg_v4_0_early_init - set function pointers + * + * @handle: amdgpu_device pointer + * + * Set ring and irq function pointers + */ +static int jpeg_v4_0_early_init(void *handle) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + + + adev->jpeg.num_jpeg_inst = 1; + + jpeg_v4_0_set_dec_ring_funcs(adev); + jpeg_v4_0_set_irq_funcs(adev); + + return 0; +} + +/** + * jpeg_v4_0_sw_init - sw init for JPEG block + * + * @handle: amdgpu_device pointer + * + * Load firmware and sw initialization + */ +static int jpeg_v4_0_sw_init(void *handle) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + struct amdgpu_ring *ring; + int r; + + /* JPEG TRAP */ + r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_VCN, + VCN_4_0__SRCID__JPEG_DECODE, &adev->jpeg.inst->irq); + if (r) + return r; + + r = amdgpu_jpeg_sw_init(adev); + if (r) + return r; + + r = amdgpu_jpeg_resume(adev); + if (r) + return r; + + ring = &adev->jpeg.inst->ring_dec; + ring->use_doorbell = true; + ring->doorbell_index = (adev->doorbell_index.vcn.vcn_ring0_1 << 1) + 1; + sprintf(ring->name, "jpeg_dec"); + r = amdgpu_ring_init(adev, ring, 512, &adev->jpeg.inst->irq, 0, + AMDGPU_RING_PRIO_DEFAULT, NULL); + if (r) + return r; + + adev->jpeg.internal.jpeg_pitch = regUVD_JPEG_PITCH_INTERNAL_OFFSET; + adev->jpeg.inst->external.jpeg_pitch = SOC15_REG_OFFSET(JPEG, 0, regUVD_JPEG_PITCH); + + return 0; +} + +/** + * jpeg_v4_0_sw_fini - sw fini for JPEG block + * + * @handle: amdgpu_device pointer + * + * JPEG suspend and free up sw allocation + */ +static int jpeg_v4_0_sw_fini(void *handle) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + int r; + + r = amdgpu_jpeg_suspend(adev); + if (r) + return r; + + r = amdgpu_jpeg_sw_fini(adev); + + return r; +} + +/** + * jpeg_v4_0_hw_init - start and test JPEG block + * + * @handle: amdgpu_device pointer + * + */ +static int jpeg_v4_0_hw_init(void *handle) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + struct amdgpu_ring *ring = &adev->jpeg.inst->ring_dec; + int r; + + adev->nbio.funcs->vcn_doorbell_range(adev, ring->use_doorbell, + (adev->doorbell_index.vcn.vcn_ring0_1 << 1), 0); + + WREG32_SOC15(VCN, 0, regVCN_JPEG_DB_CTRL, + ring->doorbell_index << VCN_JPEG_DB_CTRL__OFFSET__SHIFT | + VCN_JPEG_DB_CTRL__EN_MASK); + + r = amdgpu_ring_test_helper(ring); + if (r) + return r; + + DRM_DEV_INFO(adev->dev, "JPEG decode initialized successfully.\n"); + + return 0; +} + +/** + * jpeg_v4_0_hw_fini - stop the hardware block + * + * @handle: amdgpu_device pointer + * + * Stop the JPEG block, mark ring as not ready any more + */ +static int jpeg_v4_0_hw_fini(void *handle) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + + cancel_delayed_work_sync(&adev->vcn.idle_work); + + if (adev->jpeg.cur_state != AMD_PG_STATE_GATE && + RREG32_SOC15(JPEG, 0, regUVD_JRBC_STATUS)) + jpeg_v4_0_set_powergating_state(adev, AMD_PG_STATE_GATE); + + return 0; +} + +/** + * jpeg_v4_0_suspend - suspend JPEG block + * + * @handle: amdgpu_device pointer + * + * HW fini and suspend JPEG block + */ +static int jpeg_v4_0_suspend(void *handle) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + int r; + + r = jpeg_v4_0_hw_fini(adev); + if (r) + return r; + + r = amdgpu_jpeg_suspend(adev); + + return r; +} + +/** + * jpeg_v4_0_resume - resume JPEG block + * + * @handle: amdgpu_device pointer + * + * Resume firmware and hw init JPEG block + */ +static int jpeg_v4_0_resume(void *handle) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + int r; + + r = amdgpu_jpeg_resume(adev); + if (r) + return r; + + r = jpeg_v4_0_hw_init(adev); + + return r; +} + +static void jpeg_v4_0_disable_clock_gating(struct amdgpu_device *adev) +{ + uint32_t data = 0; + + data = RREG32_SOC15(JPEG, 0, regJPEG_CGC_CTRL); + if (adev->cg_flags & AMD_CG_SUPPORT_JPEG_MGCG) { + data |= 1 << JPEG_CGC_CTRL__DYN_CLOCK_MODE__SHIFT; + data &= (~JPEG_CGC_CTRL__JPEG_DEC_MODE_MASK); + } else { + data &= ~JPEG_CGC_CTRL__DYN_CLOCK_MODE__SHIFT; + } + + data |= 1 << JPEG_CGC_CTRL__CLK_GATE_DLY_TIMER__SHIFT; + data |= 4 << JPEG_CGC_CTRL__CLK_OFF_DELAY__SHIFT; + WREG32_SOC15(JPEG, 0, regJPEG_CGC_CTRL, data); + + data = RREG32_SOC15(JPEG, 0, regJPEG_CGC_GATE); + data &= ~(JPEG_CGC_GATE__JPEG_DEC_MASK + | JPEG_CGC_GATE__JPEG2_DEC_MASK + | JPEG_CGC_GATE__JMCIF_MASK + | JPEG_CGC_GATE__JRBBM_MASK); + WREG32_SOC15(JPEG, 0, regJPEG_CGC_GATE, data); +} + +static void jpeg_v4_0_enable_clock_gating(struct amdgpu_device *adev) +{ + uint32_t data = 0; + + data = RREG32_SOC15(JPEG, 0, regJPEG_CGC_CTRL); + if (adev->cg_flags & AMD_CG_SUPPORT_JPEG_MGCG) { + data |= 1 << JPEG_CGC_CTRL__DYN_CLOCK_MODE__SHIFT; + data |= JPEG_CGC_CTRL__JPEG_DEC_MODE_MASK; + } else { + data &= ~JPEG_CGC_CTRL__DYN_CLOCK_MODE__SHIFT; + } + + data |= 1 << JPEG_CGC_CTRL__CLK_GATE_DLY_TIMER__SHIFT; + data |= 4 << JPEG_CGC_CTRL__CLK_OFF_DELAY__SHIFT; + WREG32_SOC15(JPEG, 0, regJPEG_CGC_CTRL, data); + + data = RREG32_SOC15(JPEG, 0, regJPEG_CGC_GATE); + data |= (JPEG_CGC_GATE__JPEG_DEC_MASK + |JPEG_CGC_GATE__JPEG2_DEC_MASK + |JPEG_CGC_GATE__JMCIF_MASK + |JPEG_CGC_GATE__JRBBM_MASK); + WREG32_SOC15(JPEG, 0, regJPEG_CGC_GATE, data); +} + +static int jpeg_v4_0_disable_static_power_gating(struct amdgpu_device *adev) +{ + if (adev->pg_flags & AMD_PG_SUPPORT_JPEG) { + uint32_t data = 0; + int r = 0; + + data = 1 << UVD_PGFSM_CONFIG__UVDJ_PWR_CONFIG__SHIFT; + WREG32(SOC15_REG_OFFSET(JPEG, 0, regUVD_PGFSM_CONFIG), data); + + r = SOC15_WAIT_ON_RREG(JPEG, 0, + regUVD_PGFSM_STATUS, UVD_PGFSM_STATUS_UVDJ_PWR_ON, + UVD_PGFSM_STATUS__UVDJ_PWR_STATUS_MASK); + + if (r) { + DRM_DEV_ERROR(adev->dev, "amdgpu: JPEG disable power gating failed\n"); + return r; + } + } + + /* disable anti hang mechanism */ + WREG32_P(SOC15_REG_OFFSET(JPEG, 0, regUVD_JPEG_POWER_STATUS), 0, + ~UVD_JPEG_POWER_STATUS__JPEG_POWER_STATUS_MASK); + + /* keep the JPEG in static PG mode */ + WREG32_P(SOC15_REG_OFFSET(JPEG, 0, regUVD_JPEG_POWER_STATUS), 0, + ~UVD_JPEG_POWER_STATUS__JPEG_PG_MODE_MASK); + + return 0; +} + +static int jpeg_v4_0_enable_static_power_gating(struct amdgpu_device *adev) +{ + /* enable anti hang mechanism */ + WREG32_P(SOC15_REG_OFFSET(JPEG, 0, regUVD_JPEG_POWER_STATUS), + UVD_JPEG_POWER_STATUS__JPEG_POWER_STATUS_MASK, + ~UVD_JPEG_POWER_STATUS__JPEG_POWER_STATUS_MASK); + + if (adev->pg_flags & AMD_PG_SUPPORT_JPEG) { + uint32_t data = 0; + int r = 0; + + data = 2 << UVD_PGFSM_CONFIG__UVDJ_PWR_CONFIG__SHIFT; + WREG32(SOC15_REG_OFFSET(JPEG, 0, regUVD_PGFSM_CONFIG), data); + + r = SOC15_WAIT_ON_RREG(JPEG, 0, regUVD_PGFSM_STATUS, + (2 << UVD_PGFSM_STATUS__UVDJ_PWR_STATUS__SHIFT), + UVD_PGFSM_STATUS__UVDJ_PWR_STATUS_MASK); + + if (r) { + DRM_DEV_ERROR(adev->dev, "amdgpu: JPEG enable power gating failed\n"); + return r; + } + } + + return 0; +} + +/** + * jpeg_v4_0_start - start JPEG block + * + * @adev: amdgpu_device pointer + * + * Setup and start the JPEG block + */ +static int jpeg_v4_0_start(struct amdgpu_device *adev) +{ + struct amdgpu_ring *ring = &adev->jpeg.inst->ring_dec; + int r; + + if (adev->pm.dpm_enabled) + amdgpu_dpm_enable_jpeg(adev, true); + + /* disable power gating */ + r = jpeg_v4_0_disable_static_power_gating(adev); + if (r) + return r; + + /* JPEG disable CGC */ + jpeg_v4_0_disable_clock_gating(adev); + + /* MJPEG global tiling registers */ + WREG32_SOC15(JPEG, 0, regJPEG_DEC_GFX10_ADDR_CONFIG, + adev->gfx.config.gb_addr_config); + + + /* enable JMI channel */ + WREG32_P(SOC15_REG_OFFSET(JPEG, 0, regUVD_JMI_CNTL), 0, + ~UVD_JMI_CNTL__SOFT_RESET_MASK); + + /* enable System Interrupt for JRBC */ + WREG32_P(SOC15_REG_OFFSET(JPEG, 0, regJPEG_SYS_INT_EN), + JPEG_SYS_INT_EN__DJRBC_MASK, + ~JPEG_SYS_INT_EN__DJRBC_MASK); + + WREG32_SOC15(JPEG, 0, regUVD_LMI_JRBC_RB_VMID, 0); + WREG32_SOC15(JPEG, 0, regUVD_JRBC_RB_CNTL, (0x00000001L | 0x00000002L)); + WREG32_SOC15(JPEG, 0, regUVD_LMI_JRBC_RB_64BIT_BAR_LOW, + lower_32_bits(ring->gpu_addr)); + WREG32_SOC15(JPEG, 0, regUVD_LMI_JRBC_RB_64BIT_BAR_HIGH, + upper_32_bits(ring->gpu_addr)); + WREG32_SOC15(JPEG, 0, regUVD_JRBC_RB_RPTR, 0); + WREG32_SOC15(JPEG, 0, regUVD_JRBC_RB_WPTR, 0); + WREG32_SOC15(JPEG, 0, regUVD_JRBC_RB_CNTL, 0x00000002L); + WREG32_SOC15(JPEG, 0, regUVD_JRBC_RB_SIZE, ring->ring_size / 4); + ring->wptr = RREG32_SOC15(JPEG, 0, regUVD_JRBC_RB_WPTR); + + return 0; +} + +/** + * jpeg_v4_0_stop - stop JPEG block + * + * @adev: amdgpu_device pointer + * + * stop the JPEG block + */ +static int jpeg_v4_0_stop(struct amdgpu_device *adev) +{ + int r; + + /* reset JMI */ + WREG32_P(SOC15_REG_OFFSET(JPEG, 0, regUVD_JMI_CNTL), + UVD_JMI_CNTL__SOFT_RESET_MASK, + ~UVD_JMI_CNTL__SOFT_RESET_MASK); + + jpeg_v4_0_enable_clock_gating(adev); + + /* enable power gating */ + r = jpeg_v4_0_enable_static_power_gating(adev); + if (r) + return r; + + if (adev->pm.dpm_enabled) + amdgpu_dpm_enable_jpeg(adev, false); + + return 0; +} + +/** + * jpeg_v4_0_dec_ring_get_rptr - get read pointer + * + * @ring: amdgpu_ring pointer + * + * Returns the current hardware read pointer + */ +static uint64_t jpeg_v4_0_dec_ring_get_rptr(struct amdgpu_ring *ring) +{ + struct amdgpu_device *adev = ring->adev; + + return RREG32_SOC15(JPEG, 0, regUVD_JRBC_RB_RPTR); +} + +/** + * jpeg_v4_0_dec_ring_get_wptr - get write pointer + * + * @ring: amdgpu_ring pointer + * + * Returns the current hardware write pointer + */ +static uint64_t jpeg_v4_0_dec_ring_get_wptr(struct amdgpu_ring *ring) +{ + struct amdgpu_device *adev = ring->adev; + + if (ring->use_doorbell) + return *ring->wptr_cpu_addr; + else + return RREG32_SOC15(JPEG, 0, regUVD_JRBC_RB_WPTR); +} + +/** + * jpeg_v4_0_dec_ring_set_wptr - set write pointer + * + * @ring: amdgpu_ring pointer + * + * Commits the write pointer to the hardware + */ +static void jpeg_v4_0_dec_ring_set_wptr(struct amdgpu_ring *ring) +{ + struct amdgpu_device *adev = ring->adev; + + if (ring->use_doorbell) { + *ring->wptr_cpu_addr = lower_32_bits(ring->wptr); + WDOORBELL32(ring->doorbell_index, lower_32_bits(ring->wptr)); + } else { + WREG32_SOC15(JPEG, 0, regUVD_JRBC_RB_WPTR, lower_32_bits(ring->wptr)); + } +} + +static bool jpeg_v4_0_is_idle(void *handle) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + int ret = 1; + + ret &= (((RREG32_SOC15(JPEG, 0, regUVD_JRBC_STATUS) & + UVD_JRBC_STATUS__RB_JOB_DONE_MASK) == + UVD_JRBC_STATUS__RB_JOB_DONE_MASK)); + + return ret; +} + +static int jpeg_v4_0_wait_for_idle(void *handle) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + + return SOC15_WAIT_ON_RREG(JPEG, 0, regUVD_JRBC_STATUS, + UVD_JRBC_STATUS__RB_JOB_DONE_MASK, + UVD_JRBC_STATUS__RB_JOB_DONE_MASK); +} + +static int jpeg_v4_0_set_clockgating_state(void *handle, + enum amd_clockgating_state state) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + bool enable = (state == AMD_CG_STATE_GATE) ? true : false; + + if (enable) { + if (!jpeg_v4_0_is_idle(handle)) + return -EBUSY; + jpeg_v4_0_enable_clock_gating(adev); + } else { + jpeg_v4_0_disable_clock_gating(adev); + } + + return 0; +} + +static int jpeg_v4_0_set_powergating_state(void *handle, + enum amd_powergating_state state) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + int ret; + + if (state == adev->jpeg.cur_state) + return 0; + + if (state == AMD_PG_STATE_GATE) + ret = jpeg_v4_0_stop(adev); + else + ret = jpeg_v4_0_start(adev); + + if (!ret) + adev->jpeg.cur_state = state; + + return ret; +} + +static int jpeg_v4_0_set_interrupt_state(struct amdgpu_device *adev, + struct amdgpu_irq_src *source, + unsigned type, + enum amdgpu_interrupt_state state) +{ + return 0; +} + +static int jpeg_v4_0_process_interrupt(struct amdgpu_device *adev, + struct amdgpu_irq_src *source, + struct amdgpu_iv_entry *entry) +{ + DRM_DEBUG("IH: JPEG TRAP\n"); + + switch (entry->src_id) { + case VCN_4_0__SRCID__JPEG_DECODE: + amdgpu_fence_process(&adev->jpeg.inst->ring_dec); + break; + default: + DRM_DEV_ERROR(adev->dev, "Unhandled interrupt: %d %d\n", + entry->src_id, entry->src_data[0]); + break; + } + + return 0; +} + +static const struct amd_ip_funcs jpeg_v4_0_ip_funcs = { + .name = "jpeg_v4_0", + .early_init = jpeg_v4_0_early_init, + .late_init = NULL, + .sw_init = jpeg_v4_0_sw_init, + .sw_fini = jpeg_v4_0_sw_fini, + .hw_init = jpeg_v4_0_hw_init, + .hw_fini = jpeg_v4_0_hw_fini, + .suspend = jpeg_v4_0_suspend, + .resume = jpeg_v4_0_resume, + .is_idle = jpeg_v4_0_is_idle, + .wait_for_idle = jpeg_v4_0_wait_for_idle, + .check_soft_reset = NULL, + .pre_soft_reset = NULL, + .soft_reset = NULL, + .post_soft_reset = NULL, + .set_clockgating_state = jpeg_v4_0_set_clockgating_state, + .set_powergating_state = jpeg_v4_0_set_powergating_state, +}; + +static const struct amdgpu_ring_funcs jpeg_v4_0_dec_ring_vm_funcs = { + .type = AMDGPU_RING_TYPE_VCN_JPEG, + .align_mask = 0xf, + .vmhub = AMDGPU_MMHUB_0, + .get_rptr = jpeg_v4_0_dec_ring_get_rptr, + .get_wptr = jpeg_v4_0_dec_ring_get_wptr, + .set_wptr = jpeg_v4_0_dec_ring_set_wptr, + .emit_frame_size = + SOC15_FLUSH_GPU_TLB_NUM_WREG * 6 + + SOC15_FLUSH_GPU_TLB_NUM_REG_WAIT * 8 + + 8 + /* jpeg_v4_0_dec_ring_emit_vm_flush */ + 18 + 18 + /* jpeg_v4_0_dec_ring_emit_fence x2 vm fence */ + 8 + 16, + .emit_ib_size = 22, /* jpeg_v4_0_dec_ring_emit_ib */ + .emit_ib = jpeg_v2_0_dec_ring_emit_ib, + .emit_fence = jpeg_v2_0_dec_ring_emit_fence, + .emit_vm_flush = jpeg_v2_0_dec_ring_emit_vm_flush, + .test_ring = amdgpu_jpeg_dec_ring_test_ring, + .test_ib = amdgpu_jpeg_dec_ring_test_ib, + .insert_nop = jpeg_v2_0_dec_ring_nop, + .insert_start = jpeg_v2_0_dec_ring_insert_start, + .insert_end = jpeg_v2_0_dec_ring_insert_end, + .pad_ib = amdgpu_ring_generic_pad_ib, + .begin_use = amdgpu_jpeg_ring_begin_use, + .end_use = amdgpu_jpeg_ring_end_use, + .emit_wreg = jpeg_v2_0_dec_ring_emit_wreg, + .emit_reg_wait = jpeg_v2_0_dec_ring_emit_reg_wait, + .emit_reg_write_reg_wait = amdgpu_ring_emit_reg_write_reg_wait_helper, +}; + +static void jpeg_v4_0_set_dec_ring_funcs(struct amdgpu_device *adev) +{ + adev->jpeg.inst->ring_dec.funcs = &jpeg_v4_0_dec_ring_vm_funcs; + DRM_DEV_INFO(adev->dev, "JPEG decode is enabled in VM mode\n"); +} + +static const struct amdgpu_irq_src_funcs jpeg_v4_0_irq_funcs = { + .set = jpeg_v4_0_set_interrupt_state, + .process = jpeg_v4_0_process_interrupt, +}; + +static void jpeg_v4_0_set_irq_funcs(struct amdgpu_device *adev) +{ + adev->jpeg.inst->irq.num_types = 1; + adev->jpeg.inst->irq.funcs = &jpeg_v4_0_irq_funcs; +} + +const struct amdgpu_ip_block_version jpeg_v4_0_ip_block = { + .type = AMD_IP_BLOCK_TYPE_JPEG, + .major = 4, + .minor = 0, + .rev = 0, + .funcs = &jpeg_v4_0_ip_funcs, +}; diff --git a/drivers/gpu/drm/amd/amdgpu/jpeg_v4_0.h b/drivers/gpu/drm/amd/amdgpu/jpeg_v4_0.h new file mode 100644 index 000000000000..f1ed6ccfedca --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/jpeg_v4_0.h @@ -0,0 +1,29 @@ +/* + * Copyright 2021 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#ifndef __JPEG_V4_0_H__ +#define __JPEG_V4_0_H__ + +extern const struct amdgpu_ip_block_version jpeg_v4_0_ip_block; + +#endif /* __JPEG_V4_0_H__ */ diff --git a/drivers/gpu/drm/amd/amdgpu/lsdma_v6_0.c b/drivers/gpu/drm/amd/amdgpu/lsdma_v6_0.c new file mode 100644 index 000000000000..1a285b531881 --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/lsdma_v6_0.c @@ -0,0 +1,121 @@ +/* + * Copyright 2022 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#include <linux/delay.h> +#include "amdgpu.h" +#include "lsdma_v6_0.h" +#include "amdgpu_lsdma.h" + +#include "lsdma/lsdma_6_0_0_offset.h" +#include "lsdma/lsdma_6_0_0_sh_mask.h" + +static int lsdma_v6_0_wait_pio_status(struct amdgpu_device *adev) +{ + return amdgpu_lsdma_wait_for(adev, SOC15_REG_OFFSET(LSDMA, 0, regLSDMA_PIO_STATUS), + LSDMA_PIO_STATUS__PIO_IDLE_MASK | LSDMA_PIO_STATUS__PIO_FIFO_EMPTY_MASK, + LSDMA_PIO_STATUS__PIO_IDLE_MASK | LSDMA_PIO_STATUS__PIO_FIFO_EMPTY_MASK); +} + +static int lsdma_v6_0_copy_mem(struct amdgpu_device *adev, + uint64_t src_addr, + uint64_t dst_addr, + uint64_t size) +{ + int ret; + uint32_t tmp; + + WREG32_SOC15(LSDMA, 0, regLSDMA_PIO_SRC_ADDR_LO, lower_32_bits(src_addr)); + WREG32_SOC15(LSDMA, 0, regLSDMA_PIO_SRC_ADDR_HI, upper_32_bits(src_addr)); + + WREG32_SOC15(LSDMA, 0, regLSDMA_PIO_DST_ADDR_LO, lower_32_bits(dst_addr)); + WREG32_SOC15(LSDMA, 0, regLSDMA_PIO_DST_ADDR_HI, upper_32_bits(dst_addr)); + + WREG32_SOC15(LSDMA, 0, regLSDMA_PIO_CONTROL, 0x0); + + tmp = RREG32_SOC15(LSDMA, 0, regLSDMA_PIO_COMMAND); + tmp = REG_SET_FIELD(tmp, LSDMA_PIO_COMMAND, BYTE_COUNT, size); + tmp = REG_SET_FIELD(tmp, LSDMA_PIO_COMMAND, SRC_LOCATION, 0); + tmp = REG_SET_FIELD(tmp, LSDMA_PIO_COMMAND, DST_LOCATION, 0); + tmp = REG_SET_FIELD(tmp, LSDMA_PIO_COMMAND, SRC_ADDR_INC, 0); + tmp = REG_SET_FIELD(tmp, LSDMA_PIO_COMMAND, DST_ADDR_INC, 0); + tmp = REG_SET_FIELD(tmp, LSDMA_PIO_COMMAND, OVERLAP_DISABLE, 0); + tmp = REG_SET_FIELD(tmp, LSDMA_PIO_COMMAND, CONSTANT_FILL, 0); + WREG32_SOC15(LSDMA, 0, regLSDMA_PIO_COMMAND, tmp); + + ret = lsdma_v6_0_wait_pio_status(adev); + if (ret) + dev_err(adev->dev, "LSDMA PIO failed to copy memory!\n"); + + return ret; +} + +static int lsdma_v6_0_fill_mem(struct amdgpu_device *adev, + uint64_t dst_addr, + uint32_t data, + uint64_t size) +{ + int ret; + uint32_t tmp; + + WREG32_SOC15(LSDMA, 0, regLSDMA_PIO_CONSTFILL_DATA, data); + + WREG32_SOC15(LSDMA, 0, regLSDMA_PIO_DST_ADDR_LO, lower_32_bits(dst_addr)); + WREG32_SOC15(LSDMA, 0, regLSDMA_PIO_DST_ADDR_HI, upper_32_bits(dst_addr)); + + WREG32_SOC15(LSDMA, 0, regLSDMA_PIO_CONTROL, 0x0); + + tmp = RREG32_SOC15(LSDMA, 0, regLSDMA_PIO_COMMAND); + tmp = REG_SET_FIELD(tmp, LSDMA_PIO_COMMAND, BYTE_COUNT, size); + tmp = REG_SET_FIELD(tmp, LSDMA_PIO_COMMAND, SRC_LOCATION, 0); + tmp = REG_SET_FIELD(tmp, LSDMA_PIO_COMMAND, DST_LOCATION, 0); + tmp = REG_SET_FIELD(tmp, LSDMA_PIO_COMMAND, SRC_ADDR_INC, 0); + tmp = REG_SET_FIELD(tmp, LSDMA_PIO_COMMAND, DST_ADDR_INC, 0); + tmp = REG_SET_FIELD(tmp, LSDMA_PIO_COMMAND, OVERLAP_DISABLE, 0); + tmp = REG_SET_FIELD(tmp, LSDMA_PIO_COMMAND, CONSTANT_FILL, 1); + WREG32_SOC15(LSDMA, 0, regLSDMA_PIO_COMMAND, tmp); + + ret = lsdma_v6_0_wait_pio_status(adev); + if (ret) + dev_err(adev->dev, "LSDMA PIO failed to fill memory!\n"); + + return ret; +} + +static void lsdma_v6_0_update_memory_power_gating(struct amdgpu_device *adev, + bool enable) +{ + uint32_t tmp; + + tmp = RREG32_SOC15(LSDMA, 0, regLSDMA_MEM_POWER_CTRL); + tmp = REG_SET_FIELD(tmp, LSDMA_MEM_POWER_CTRL, MEM_POWER_CTRL_EN, 0); + WREG32_SOC15(LSDMA, 0, regLSDMA_MEM_POWER_CTRL, tmp); + + tmp = REG_SET_FIELD(tmp, LSDMA_MEM_POWER_CTRL, MEM_POWER_CTRL_EN, enable); + WREG32_SOC15(LSDMA, 0, regLSDMA_MEM_POWER_CTRL, tmp); +} + +const struct amdgpu_lsdma_funcs lsdma_v6_0_funcs = { + .copy_mem = lsdma_v6_0_copy_mem, + .fill_mem = lsdma_v6_0_fill_mem, + .update_memory_power_gating = lsdma_v6_0_update_memory_power_gating +}; diff --git a/drivers/gpu/drm/amd/amdgpu/lsdma_v6_0.h b/drivers/gpu/drm/amd/amdgpu/lsdma_v6_0.h new file mode 100644 index 000000000000..3ef79be1a9bf --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/lsdma_v6_0.h @@ -0,0 +1,31 @@ +/* + * Copyright 2022 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#ifndef __LSDMA_V6_0_H__ +#define __LSDMA_V6_0_H__ + +#include "soc15_common.h" + +extern const struct amdgpu_lsdma_funcs lsdma_v6_0_funcs; + +#endif /* __LSDMA_V6_0_H__ */ diff --git a/drivers/gpu/drm/amd/amdgpu/mca_v3_0.c b/drivers/gpu/drm/amd/amdgpu/mca_v3_0.c index 8f7107d392af..d4bd7d1d2649 100644 --- a/drivers/gpu/drm/amd/amdgpu/mca_v3_0.c +++ b/drivers/gpu/drm/amd/amdgpu/mca_v3_0.c @@ -37,24 +37,36 @@ static void mca_v3_0_mp0_query_ras_error_count(struct amdgpu_device *adev, ras_error_status); } -static int mca_v3_0_mp0_ras_late_init(struct amdgpu_device *adev) +static int mca_v3_0_ras_block_match(struct amdgpu_ras_block_object *block_obj, + enum amdgpu_ras_block block, uint32_t sub_block_index) { - return amdgpu_mca_ras_late_init(adev, &adev->mca.mp0); -} + if (!block_obj) + return -EINVAL; -static void mca_v3_0_mp0_ras_fini(struct amdgpu_device *adev) -{ - amdgpu_mca_ras_fini(adev, &adev->mca.mp0); + if ((block_obj->ras_comm.block == block) && + (block_obj->ras_comm.sub_block_index == sub_block_index)) { + return 0; + } + + return -EINVAL; } -const struct amdgpu_mca_ras_funcs mca_v3_0_mp0_ras_funcs = { - .ras_late_init = mca_v3_0_mp0_ras_late_init, - .ras_fini = mca_v3_0_mp0_ras_fini, +const struct amdgpu_ras_block_hw_ops mca_v3_0_mp0_hw_ops = { .query_ras_error_count = mca_v3_0_mp0_query_ras_error_count, .query_ras_error_address = NULL, - .ras_block = AMDGPU_RAS_BLOCK__MCA, - .ras_sub_block = AMDGPU_RAS_MCA_BLOCK__MP0, - .sysfs_name = "mp0_err_count", +}; + +struct amdgpu_mca_ras_block mca_v3_0_mp0_ras = { + .ras_block = { + .ras_comm = { + .block = AMDGPU_RAS_BLOCK__MCA, + .sub_block_index = AMDGPU_RAS_MCA_BLOCK__MP0, + .type = AMDGPU_RAS_ERROR__MULTI_UNCORRECTABLE, + .name = "mp0", + }, + .hw_ops = &mca_v3_0_mp0_hw_ops, + .ras_block_match = mca_v3_0_ras_block_match, + }, }; static void mca_v3_0_mp1_query_ras_error_count(struct amdgpu_device *adev, @@ -65,24 +77,22 @@ static void mca_v3_0_mp1_query_ras_error_count(struct amdgpu_device *adev, ras_error_status); } -static int mca_v3_0_mp1_ras_late_init(struct amdgpu_device *adev) -{ - return amdgpu_mca_ras_late_init(adev, &adev->mca.mp1); -} - -static void mca_v3_0_mp1_ras_fini(struct amdgpu_device *adev) -{ - amdgpu_mca_ras_fini(adev, &adev->mca.mp1); -} - -const struct amdgpu_mca_ras_funcs mca_v3_0_mp1_ras_funcs = { - .ras_late_init = mca_v3_0_mp1_ras_late_init, - .ras_fini = mca_v3_0_mp1_ras_fini, +const struct amdgpu_ras_block_hw_ops mca_v3_0_mp1_hw_ops = { .query_ras_error_count = mca_v3_0_mp1_query_ras_error_count, .query_ras_error_address = NULL, - .ras_block = AMDGPU_RAS_BLOCK__MCA, - .ras_sub_block = AMDGPU_RAS_MCA_BLOCK__MP1, - .sysfs_name = "mp1_err_count", +}; + +struct amdgpu_mca_ras_block mca_v3_0_mp1_ras = { + .ras_block = { + .ras_comm = { + .block = AMDGPU_RAS_BLOCK__MCA, + .sub_block_index = AMDGPU_RAS_MCA_BLOCK__MP1, + .type = AMDGPU_RAS_ERROR__MULTI_UNCORRECTABLE, + .name = "mp1", + }, + .hw_ops = &mca_v3_0_mp1_hw_ops, + .ras_block_match = mca_v3_0_ras_block_match, + }, }; static void mca_v3_0_mpio_query_ras_error_count(struct amdgpu_device *adev, @@ -93,24 +103,22 @@ static void mca_v3_0_mpio_query_ras_error_count(struct amdgpu_device *adev, ras_error_status); } -static int mca_v3_0_mpio_ras_late_init(struct amdgpu_device *adev) -{ - return amdgpu_mca_ras_late_init(adev, &adev->mca.mpio); -} - -static void mca_v3_0_mpio_ras_fini(struct amdgpu_device *adev) -{ - amdgpu_mca_ras_fini(adev, &adev->mca.mpio); -} - -const struct amdgpu_mca_ras_funcs mca_v3_0_mpio_ras_funcs = { - .ras_late_init = mca_v3_0_mpio_ras_late_init, - .ras_fini = mca_v3_0_mpio_ras_fini, +const struct amdgpu_ras_block_hw_ops mca_v3_0_mpio_hw_ops = { .query_ras_error_count = mca_v3_0_mpio_query_ras_error_count, .query_ras_error_address = NULL, - .ras_block = AMDGPU_RAS_BLOCK__MCA, - .ras_sub_block = AMDGPU_RAS_MCA_BLOCK__MPIO, - .sysfs_name = "mpio_err_count", +}; + +struct amdgpu_mca_ras_block mca_v3_0_mpio_ras = { + .ras_block = { + .ras_comm = { + .block = AMDGPU_RAS_BLOCK__MCA, + .sub_block_index = AMDGPU_RAS_MCA_BLOCK__MPIO, + .type = AMDGPU_RAS_ERROR__MULTI_UNCORRECTABLE, + .name = "mpio", + }, + .hw_ops = &mca_v3_0_mpio_hw_ops, + .ras_block_match = mca_v3_0_ras_block_match, + }, }; @@ -118,9 +126,15 @@ static void mca_v3_0_init(struct amdgpu_device *adev) { struct amdgpu_mca *mca = &adev->mca; - mca->mp0.ras_funcs = &mca_v3_0_mp0_ras_funcs; - mca->mp1.ras_funcs = &mca_v3_0_mp1_ras_funcs; - mca->mpio.ras_funcs = &mca_v3_0_mpio_ras_funcs; + mca->mp0.ras = &mca_v3_0_mp0_ras; + mca->mp1.ras = &mca_v3_0_mp1_ras; + mca->mpio.ras = &mca_v3_0_mpio_ras; + amdgpu_ras_register_ras_block(adev, &mca->mp0.ras->ras_block); + amdgpu_ras_register_ras_block(adev, &mca->mp1.ras->ras_block); + amdgpu_ras_register_ras_block(adev, &mca->mpio.ras->ras_block); + mca->mp0.ras_if = &mca->mp0.ras->ras_block.ras_comm; + mca->mp1.ras_if = &mca->mp1.ras->ras_block.ras_comm; + mca->mpio.ras_if = &mca->mpio.ras->ras_block.ras_comm; } const struct amdgpu_mca_funcs mca_v3_0_funcs = { diff --git a/drivers/gpu/drm/amd/amdgpu/mes_api_def.h b/drivers/gpu/drm/amd/amdgpu/mes_api_def.h deleted file mode 100644 index 3f4fca5fd1da..000000000000 --- a/drivers/gpu/drm/amd/amdgpu/mes_api_def.h +++ /dev/null @@ -1,443 +0,0 @@ -/* - * Copyright 2019 Advanced Micro Devices, Inc. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * the rights to use, copy, modify, merge, publish, distribute, sublicense, - * and/or sell copies of the Software, and to permit persons to whom the - * Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice shall be included in - * all copies or substantial portions of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL - * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR - * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, - * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR - * OTHER DEALINGS IN THE SOFTWARE. - * - */ - -#ifndef __MES_API_DEF_H__ -#define __MES_API_DEF_H__ - -#pragma pack(push, 4) - -#define MES_API_VERSION 1 - -/* Driver submits one API(cmd) as a single Frame and this command size is same - * for all API to ease the debugging and parsing of ring buffer. - */ -enum { API_FRAME_SIZE_IN_DWORDS = 64 }; - -/* To avoid command in scheduler context to be overwritten whenenver mutilple - * interrupts come in, this creates another queue. - */ -enum { API_NUMBER_OF_COMMAND_MAX = 32 }; - -enum MES_API_TYPE { - MES_API_TYPE_SCHEDULER = 1, - MES_API_TYPE_MAX -}; - -enum MES_SCH_API_OPCODE { - MES_SCH_API_SET_HW_RSRC = 0, - MES_SCH_API_SET_SCHEDULING_CONFIG = 1, /* agreegated db, quantums, etc */ - MES_SCH_API_ADD_QUEUE = 2, - MES_SCH_API_REMOVE_QUEUE = 3, - MES_SCH_API_PERFORM_YIELD = 4, - MES_SCH_API_SET_GANG_PRIORITY_LEVEL = 5, - MES_SCH_API_SUSPEND = 6, - MES_SCH_API_RESUME = 7, - MES_SCH_API_RESET = 8, - MES_SCH_API_SET_LOG_BUFFER = 9, - MES_SCH_API_CHANGE_GANG_PRORITY = 10, - MES_SCH_API_QUERY_SCHEDULER_STATUS = 11, - MES_SCH_API_PROGRAM_GDS = 12, - MES_SCH_API_SET_DEBUG_VMID = 13, - MES_SCH_API_MISC = 14, - MES_SCH_API_MAX = 0xFF -}; - -union MES_API_HEADER { - struct { - uint32_t type : 4; /* 0 - Invalid; 1 - Scheduling; 2 - TBD */ - uint32_t opcode : 8; - uint32_t dwsize : 8; /* including header */ - uint32_t reserved : 12; - }; - - uint32_t u32All; -}; - -enum MES_AMD_PRIORITY_LEVEL { - AMD_PRIORITY_LEVEL_LOW = 0, - AMD_PRIORITY_LEVEL_NORMAL = 1, - AMD_PRIORITY_LEVEL_MEDIUM = 2, - AMD_PRIORITY_LEVEL_HIGH = 3, - AMD_PRIORITY_LEVEL_REALTIME = 4, - AMD_PRIORITY_NUM_LEVELS -}; - -enum MES_QUEUE_TYPE { - MES_QUEUE_TYPE_GFX, - MES_QUEUE_TYPE_COMPUTE, - MES_QUEUE_TYPE_SDMA, - MES_QUEUE_TYPE_MAX, -}; - -struct MES_API_STATUS { - uint64_t api_completion_fence_addr; - uint64_t api_completion_fence_value; -}; - -enum { MAX_COMPUTE_PIPES = 8 }; -enum { MAX_GFX_PIPES = 2 }; -enum { MAX_SDMA_PIPES = 2 }; - -enum { MAX_COMPUTE_HQD_PER_PIPE = 8 }; -enum { MAX_GFX_HQD_PER_PIPE = 8 }; -enum { MAX_SDMA_HQD_PER_PIPE = 10 }; - -enum { MAX_QUEUES_IN_A_GANG = 8 }; - -enum VM_HUB_TYPE { - VM_HUB_TYPE_GC = 0, - VM_HUB_TYPE_MM = 1, - VM_HUB_TYPE_MAX, -}; - -enum { VMID_INVALID = 0xffff }; - -enum { MAX_VMID_GCHUB = 16 }; -enum { MAX_VMID_MMHUB = 16 }; - -enum MES_LOG_OPERATION { - MES_LOG_OPERATION_CONTEXT_STATE_CHANGE = 0 -}; - -enum MES_LOG_CONTEXT_STATE { - MES_LOG_CONTEXT_STATE_IDLE = 0, - MES_LOG_CONTEXT_STATE_RUNNING = 1, - MES_LOG_CONTEXT_STATE_READY = 2, - MES_LOG_CONTEXT_STATE_READY_STANDBY = 3, -}; - -struct MES_LOG_CONTEXT_STATE_CHANGE { - void *h_context; - enum MES_LOG_CONTEXT_STATE new_context_state; -}; - -struct MES_LOG_ENTRY_HEADER { - uint32_t first_free_entry_index; - uint32_t wraparound_count; - uint64_t number_of_entries; - uint64_t reserved[2]; -}; - -struct MES_LOG_ENTRY_DATA { - uint64_t gpu_time_stamp; - uint32_t operation_type; /* operation_type is of MES_LOG_OPERATION type */ - uint32_t reserved_operation_type_bits; - union { - struct MES_LOG_CONTEXT_STATE_CHANGE context_state_change; - uint64_t reserved_operation_data[2]; - }; -}; - -struct MES_LOG_BUFFER { - struct MES_LOG_ENTRY_HEADER header; - struct MES_LOG_ENTRY_DATA entries[1]; -}; - -union MESAPI_SET_HW_RESOURCES { - struct { - union MES_API_HEADER header; - uint32_t vmid_mask_mmhub; - uint32_t vmid_mask_gfxhub; - uint32_t gds_size; - uint32_t paging_vmid; - uint32_t compute_hqd_mask[MAX_COMPUTE_PIPES]; - uint32_t gfx_hqd_mask[MAX_GFX_PIPES]; - uint32_t sdma_hqd_mask[MAX_SDMA_PIPES]; - uint32_t agreegated_doorbells[AMD_PRIORITY_NUM_LEVELS]; - uint64_t g_sch_ctx_gpu_mc_ptr; - uint64_t query_status_fence_gpu_mc_ptr; - struct MES_API_STATUS api_status; - union { - struct { - uint32_t disable_reset : 1; - uint32_t reserved : 31; - }; - uint32_t uint32_t_all; - }; - }; - - uint32_t max_dwords_in_api[API_FRAME_SIZE_IN_DWORDS]; -}; - -union MESAPI__ADD_QUEUE { - struct { - union MES_API_HEADER header; - uint32_t process_id; - uint64_t page_table_base_addr; - uint64_t process_va_start; - uint64_t process_va_end; - uint64_t process_quantum; - uint64_t process_context_addr; - uint64_t gang_quantum; - uint64_t gang_context_addr; - uint32_t inprocess_gang_priority; - enum MES_AMD_PRIORITY_LEVEL gang_global_priority_level; - uint32_t doorbell_offset; - uint64_t mqd_addr; - uint64_t wptr_addr; - enum MES_QUEUE_TYPE queue_type; - uint32_t gds_base; - uint32_t gds_size; - uint32_t gws_base; - uint32_t gws_size; - uint32_t oa_mask; - - struct { - uint32_t paging : 1; - uint32_t debug_vmid : 4; - uint32_t program_gds : 1; - uint32_t is_gang_suspended : 1; - uint32_t is_tmz_queue : 1; - uint32_t reserved : 24; - }; - struct MES_API_STATUS api_status; - }; - - uint32_t max_dwords_in_api[API_FRAME_SIZE_IN_DWORDS]; -}; - -union MESAPI__REMOVE_QUEUE { - struct { - union MES_API_HEADER header; - uint32_t doorbell_offset; - uint64_t gang_context_addr; - - struct { - uint32_t unmap_legacy_gfx_queue : 1; - uint32_t reserved : 31; - }; - struct MES_API_STATUS api_status; - }; - - uint32_t max_dwords_in_api[API_FRAME_SIZE_IN_DWORDS]; -}; - -union MESAPI__SET_SCHEDULING_CONFIG { - struct { - union MES_API_HEADER header; - /* Grace period when preempting another priority band for this - * priority band. The value for idle priority band is ignored, - * as it never preempts other bands. - */ - uint64_t grace_period_other_levels[AMD_PRIORITY_NUM_LEVELS]; - /* Default quantum for scheduling across processes within - * a priority band. - */ - uint64_t process_quantum_for_level[AMD_PRIORITY_NUM_LEVELS]; - /* Default grace period for processes that preempt each other - * within a priority band. - */ - uint64_t process_grace_period_same_level[AMD_PRIORITY_NUM_LEVELS]; - /* For normal level this field specifies the target GPU - * percentage in situations when it's starved by the high level. - * Valid values are between 0 and 50, with the default being 10. - */ - uint32_t normal_yield_percent; - struct MES_API_STATUS api_status; - }; - - uint32_t max_dwords_in_api[API_FRAME_SIZE_IN_DWORDS]; -}; - -union MESAPI__PERFORM_YIELD { - struct { - union MES_API_HEADER header; - uint32_t dummy; - struct MES_API_STATUS api_status; - }; - - uint32_t max_dwords_in_api[API_FRAME_SIZE_IN_DWORDS]; -}; - -union MESAPI__CHANGE_GANG_PRIORITY_LEVEL { - struct { - union MES_API_HEADER header; - uint32_t inprocess_gang_priority; - enum MES_AMD_PRIORITY_LEVEL gang_global_priority_level; - uint64_t gang_quantum; - uint64_t gang_context_addr; - struct MES_API_STATUS api_status; - }; - - uint32_t max_dwords_in_api[API_FRAME_SIZE_IN_DWORDS]; -}; - -union MESAPI__SUSPEND { - struct { - union MES_API_HEADER header; - /* false - suspend all gangs; true - specific gang */ - struct { - uint32_t suspend_all_gangs : 1; - uint32_t reserved : 31; - }; - /* gang_context_addr is valid only if suspend_all = false */ - uint64_t gang_context_addr; - - uint64_t suspend_fence_addr; - uint32_t suspend_fence_value; - - struct MES_API_STATUS api_status; - }; - - uint32_t max_dwords_in_api[API_FRAME_SIZE_IN_DWORDS]; -}; - -union MESAPI__RESUME { - struct { - union MES_API_HEADER header; - /* false - resume all gangs; true - specified gang */ - struct { - uint32_t resume_all_gangs : 1; - uint32_t reserved : 31; - }; - /* valid only if resume_all_gangs = false */ - uint64_t gang_context_addr; - - struct MES_API_STATUS api_status; - }; - - uint32_t max_dwords_in_api[API_FRAME_SIZE_IN_DWORDS]; -}; - -union MESAPI__RESET { - struct { - union MES_API_HEADER header; - - struct { - uint32_t reset_queue : 1; - uint32_t reserved : 31; - }; - - uint64_t gang_context_addr; - uint32_t doorbell_offset; /* valid only if reset_queue = true */ - struct MES_API_STATUS api_status; - }; - - uint32_t max_dwords_in_api[API_FRAME_SIZE_IN_DWORDS]; -}; - -union MESAPI__SET_LOGGING_BUFFER { - struct { - union MES_API_HEADER header; - /* There are separate log buffers for each queue type */ - enum MES_QUEUE_TYPE log_type; - /* Log buffer GPU Address */ - uint64_t logging_buffer_addr; - /* number of entries in the log buffer */ - uint32_t number_of_entries; - /* Entry index at which CPU interrupt needs to be signalled */ - uint32_t interrupt_entry; - - struct MES_API_STATUS api_status; - }; - - uint32_t max_dwords_in_api[API_FRAME_SIZE_IN_DWORDS]; -}; - -union MESAPI__QUERY_MES_STATUS { - struct { - union MES_API_HEADER header; - bool mes_healthy; /* 0 - not healthy, 1 - healthy */ - struct MES_API_STATUS api_status; - }; - - uint32_t max_dwords_in_api[API_FRAME_SIZE_IN_DWORDS]; -}; - -union MESAPI__PROGRAM_GDS { - struct { - union MES_API_HEADER header; - uint64_t process_context_addr; - uint32_t gds_base; - uint32_t gds_size; - uint32_t gws_base; - uint32_t gws_size; - uint32_t oa_mask; - struct MES_API_STATUS api_status; - }; - - uint32_t max_dwords_in_api[API_FRAME_SIZE_IN_DWORDS]; -}; - -union MESAPI__SET_DEBUG_VMID { - struct { - union MES_API_HEADER header; - struct MES_API_STATUS api_status; - union { - struct { - uint32_t use_gds : 1; - uint32_t reserved : 31; - } flags; - uint32_t u32All; - }; - uint32_t reserved; - uint32_t debug_vmid; - uint64_t process_context_addr; - uint64_t page_table_base_addr; - uint64_t process_va_start; - uint64_t process_va_end; - uint32_t gds_base; - uint32_t gds_size; - uint32_t gws_base; - uint32_t gws_size; - uint32_t oa_mask; - }; - - uint32_t max_dwords_in_api[API_FRAME_SIZE_IN_DWORDS]; -}; - -enum MESAPI_MISC_OPCODE { - MESAPI_MISC__MODIFY_REG, - MESAPI_MISC__MAX, -}; - -enum MODIFY_REG_SUBCODE { - MODIFY_REG__OVERWRITE, - MODIFY_REG__RMW_OR, - MODIFY_REG__RMW_AND, - MODIFY_REG__MAX, -}; - -enum { MISC_DATA_MAX_SIZE_IN_DWORDS = 20 }; - -union MESAPI__MISC { - struct { - union MES_API_HEADER header; - enum MESAPI_MISC_OPCODE opcode; - struct MES_API_STATUS api_status; - - union { - struct { - enum MODIFY_REG_SUBCODE subcode; - uint32_t reg_offset; - uint32_t reg_value; - } modify_reg; - uint32_t data[MISC_DATA_MAX_SIZE_IN_DWORDS]; - }; - }; - - uint32_t max_dwords_in_api[API_FRAME_SIZE_IN_DWORDS]; -}; - -#pragma pack(pop) -#endif diff --git a/drivers/gpu/drm/amd/amdgpu/mes_v10_1.c b/drivers/gpu/drm/amd/amdgpu/mes_v10_1.c index a7ec4ac89da5..067d10073a56 100644 --- a/drivers/gpu/drm/amd/amdgpu/mes_v10_1.c +++ b/drivers/gpu/drm/amd/amdgpu/mes_v10_1.c @@ -28,16 +28,21 @@ #include "nv.h" #include "gc/gc_10_1_0_offset.h" #include "gc/gc_10_1_0_sh_mask.h" +#include "gc/gc_10_1_0_default.h" #include "v10_structs.h" #include "mes_api_def.h" #define mmCP_MES_IC_OP_CNTL_Sienna_Cichlid 0x2820 #define mmCP_MES_IC_OP_CNTL_Sienna_Cichlid_BASE_IDX 1 +#define mmRLC_CP_SCHEDULERS_Sienna_Cichlid 0x4ca1 +#define mmRLC_CP_SCHEDULERS_Sienna_Cichlid_BASE_IDX 1 MODULE_FIRMWARE("amdgpu/navi10_mes.bin"); MODULE_FIRMWARE("amdgpu/sienna_cichlid_mes.bin"); +MODULE_FIRMWARE("amdgpu/sienna_cichlid_mes1.bin"); static int mes_v10_1_hw_fini(void *handle); +static int mes_v10_1_kiq_hw_init(struct amdgpu_device *adev); #define MES_EOP_SIZE 2048 @@ -46,7 +51,7 @@ static void mes_v10_1_ring_set_wptr(struct amdgpu_ring *ring) struct amdgpu_device *adev = ring->adev; if (ring->use_doorbell) { - atomic64_set((atomic64_t *)&adev->wb.wb[ring->wptr_offs], + atomic64_set((atomic64_t *)ring->wptr_cpu_addr, ring->wptr); WDOORBELL64(ring->doorbell_index, ring->wptr); } else { @@ -56,7 +61,7 @@ static void mes_v10_1_ring_set_wptr(struct amdgpu_ring *ring) static u64 mes_v10_1_ring_get_rptr(struct amdgpu_ring *ring) { - return ring->adev->wb.wb[ring->rptr_offs]; + return *ring->rptr_cpu_addr; } static u64 mes_v10_1_ring_get_wptr(struct amdgpu_ring *ring) @@ -64,8 +69,7 @@ static u64 mes_v10_1_ring_get_wptr(struct amdgpu_ring *ring) u64 wptr; if (ring->use_doorbell) - wptr = atomic64_read((atomic64_t *) - &ring->adev->wb.wb[ring->wptr_offs]); + wptr = atomic64_read((atomic64_t *)ring->wptr_cpu_addr); else BUG(); return wptr; @@ -83,21 +87,32 @@ static const struct amdgpu_ring_funcs mes_v10_1_ring_funcs = { }; static int mes_v10_1_submit_pkt_and_poll_completion(struct amdgpu_mes *mes, - void *pkt, int size) + void *pkt, int size, + int api_status_off) { int ndw = size / 4; signed long r; union MESAPI__ADD_QUEUE *x_pkt = pkt; + struct MES_API_STATUS *api_status; struct amdgpu_device *adev = mes->adev; struct amdgpu_ring *ring = &mes->ring; + unsigned long flags; BUG_ON(size % 4 != 0); - if (amdgpu_ring_alloc(ring, ndw)) + spin_lock_irqsave(&mes->ring_lock, flags); + if (amdgpu_ring_alloc(ring, ndw)) { + spin_unlock_irqrestore(&mes->ring_lock, flags); return -ENOMEM; + } + + api_status = (struct MES_API_STATUS *)((char *)pkt + api_status_off); + api_status->api_completion_fence_addr = mes->ring.fence_drv.gpu_addr; + api_status->api_completion_fence_value = ++mes->ring.fence_drv.sync_seq; amdgpu_ring_write_multiple(ring, pkt, ndw); amdgpu_ring_commit(ring); + spin_unlock_irqrestore(&mes->ring_lock, flags); DRM_DEBUG("MES msg=%d was emitted\n", x_pkt->header.opcode); @@ -130,6 +145,8 @@ static int mes_v10_1_add_hw_queue(struct amdgpu_mes *mes, { struct amdgpu_device *adev = mes->adev; union MESAPI__ADD_QUEUE mes_add_queue_pkt; + struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0]; + uint32_t vm_cntx_cntl = hub->vm_cntx_cntl; memset(&mes_add_queue_pkt, 0, sizeof(mes_add_queue_pkt)); @@ -138,8 +155,7 @@ static int mes_v10_1_add_hw_queue(struct amdgpu_mes *mes, mes_add_queue_pkt.header.dwsize = API_FRAME_SIZE_IN_DWORDS; mes_add_queue_pkt.process_id = input->process_id; - mes_add_queue_pkt.page_table_base_addr = - input->page_table_base_addr - adev->gmc.vram_start; + mes_add_queue_pkt.page_table_base_addr = input->page_table_base_addr; mes_add_queue_pkt.process_va_start = input->process_va_start; mes_add_queue_pkt.process_va_end = input->process_va_end; mes_add_queue_pkt.process_quantum = input->process_quantum; @@ -156,14 +172,14 @@ static int mes_v10_1_add_hw_queue(struct amdgpu_mes *mes, mes_add_queue_pkt.queue_type = convert_to_mes_queue_type(input->queue_type); mes_add_queue_pkt.paging = input->paging; - - mes_add_queue_pkt.api_status.api_completion_fence_addr = - mes->ring.fence_drv.gpu_addr; - mes_add_queue_pkt.api_status.api_completion_fence_value = - ++mes->ring.fence_drv.sync_seq; + mes_add_queue_pkt.vm_context_cntl = vm_cntx_cntl; + mes_add_queue_pkt.gws_base = input->gws_base; + mes_add_queue_pkt.gws_size = input->gws_size; + mes_add_queue_pkt.trap_handler_addr = input->tba_addr; return mes_v10_1_submit_pkt_and_poll_completion(mes, - &mes_add_queue_pkt, sizeof(mes_add_queue_pkt)); + &mes_add_queue_pkt, sizeof(mes_add_queue_pkt), + offsetof(union MESAPI__ADD_QUEUE, api_status)); } static int mes_v10_1_remove_hw_queue(struct amdgpu_mes *mes, @@ -180,13 +196,43 @@ static int mes_v10_1_remove_hw_queue(struct amdgpu_mes *mes, mes_remove_queue_pkt.doorbell_offset = input->doorbell_offset; mes_remove_queue_pkt.gang_context_addr = input->gang_context_addr; - mes_remove_queue_pkt.api_status.api_completion_fence_addr = - mes->ring.fence_drv.gpu_addr; - mes_remove_queue_pkt.api_status.api_completion_fence_value = - ++mes->ring.fence_drv.sync_seq; + return mes_v10_1_submit_pkt_and_poll_completion(mes, + &mes_remove_queue_pkt, sizeof(mes_remove_queue_pkt), + offsetof(union MESAPI__REMOVE_QUEUE, api_status)); +} + +static int mes_v10_1_unmap_legacy_queue(struct amdgpu_mes *mes, + struct mes_unmap_legacy_queue_input *input) +{ + union MESAPI__REMOVE_QUEUE mes_remove_queue_pkt; + + memset(&mes_remove_queue_pkt, 0, sizeof(mes_remove_queue_pkt)); + + mes_remove_queue_pkt.header.type = MES_API_TYPE_SCHEDULER; + mes_remove_queue_pkt.header.opcode = MES_SCH_API_REMOVE_QUEUE; + mes_remove_queue_pkt.header.dwsize = API_FRAME_SIZE_IN_DWORDS; + + mes_remove_queue_pkt.doorbell_offset = input->doorbell_offset; + mes_remove_queue_pkt.gang_context_addr = 0; + + mes_remove_queue_pkt.pipe_id = input->pipe_id; + mes_remove_queue_pkt.queue_id = input->queue_id; + + if (input->action == PREEMPT_QUEUES_NO_UNMAP) { + mes_remove_queue_pkt.preempt_legacy_gfx_queue = 1; + mes_remove_queue_pkt.tf_addr = input->trail_fence_addr; + mes_remove_queue_pkt.tf_data = + lower_32_bits(input->trail_fence_data); + } else { + if (input->queue_type == AMDGPU_RING_TYPE_GFX) + mes_remove_queue_pkt.unmap_legacy_gfx_queue = 1; + else + mes_remove_queue_pkt.unmap_kiq_utility_queue = 1; + } return mes_v10_1_submit_pkt_and_poll_completion(mes, - &mes_remove_queue_pkt, sizeof(mes_remove_queue_pkt)); + &mes_remove_queue_pkt, sizeof(mes_remove_queue_pkt), + offsetof(union MESAPI__REMOVE_QUEUE, api_status)); } static int mes_v10_1_suspend_gang(struct amdgpu_mes *mes, @@ -211,13 +257,9 @@ static int mes_v10_1_query_sched_status(struct amdgpu_mes *mes) mes_status_pkt.header.opcode = MES_SCH_API_QUERY_SCHEDULER_STATUS; mes_status_pkt.header.dwsize = API_FRAME_SIZE_IN_DWORDS; - mes_status_pkt.api_status.api_completion_fence_addr = - mes->ring.fence_drv.gpu_addr; - mes_status_pkt.api_status.api_completion_fence_value = - ++mes->ring.fence_drv.sync_seq; - return mes_v10_1_submit_pkt_and_poll_completion(mes, - &mes_status_pkt, sizeof(mes_status_pkt)); + &mes_status_pkt, sizeof(mes_status_pkt), + offsetof(union MESAPI__QUERY_MES_STATUS, api_status)); } static int mes_v10_1_set_hw_resources(struct amdgpu_mes *mes) @@ -251,26 +293,90 @@ static int mes_v10_1_set_hw_resources(struct amdgpu_mes *mes) mes_set_hw_res_pkt.sdma_hqd_mask[i] = mes->sdma_hqd_mask[i]; for (i = 0; i < AMD_PRIORITY_NUM_LEVELS; i++) - mes_set_hw_res_pkt.agreegated_doorbells[i] = - mes->agreegated_doorbells[i]; + mes_set_hw_res_pkt.aggregated_doorbells[i] = + mes->aggregated_doorbells[i]; + + for (i = 0; i < 5; i++) { + mes_set_hw_res_pkt.gc_base[i] = adev->reg_offset[GC_HWIP][0][i]; + mes_set_hw_res_pkt.mmhub_base[i] = + adev->reg_offset[MMHUB_HWIP][0][i]; + mes_set_hw_res_pkt.osssys_base[i] = + adev->reg_offset[OSSSYS_HWIP][0][i]; + } - mes_set_hw_res_pkt.api_status.api_completion_fence_addr = - mes->ring.fence_drv.gpu_addr; - mes_set_hw_res_pkt.api_status.api_completion_fence_value = - ++mes->ring.fence_drv.sync_seq; + mes_set_hw_res_pkt.disable_reset = 1; + mes_set_hw_res_pkt.disable_mes_log = 1; + mes_set_hw_res_pkt.use_different_vmid_compute = 1; return mes_v10_1_submit_pkt_and_poll_completion(mes, - &mes_set_hw_res_pkt, sizeof(mes_set_hw_res_pkt)); + &mes_set_hw_res_pkt, sizeof(mes_set_hw_res_pkt), + offsetof(union MESAPI_SET_HW_RESOURCES, api_status)); +} + +static void mes_v10_1_init_aggregated_doorbell(struct amdgpu_mes *mes) +{ + struct amdgpu_device *adev = mes->adev; + uint32_t data; + + data = RREG32_SOC15(GC, 0, mmCP_MES_DOORBELL_CONTROL1); + data &= ~(CP_MES_DOORBELL_CONTROL1__DOORBELL_OFFSET_MASK | + CP_MES_DOORBELL_CONTROL1__DOORBELL_EN_MASK | + CP_MES_DOORBELL_CONTROL1__DOORBELL_HIT_MASK); + data |= mes->aggregated_doorbells[AMDGPU_MES_PRIORITY_LEVEL_LOW] << + CP_MES_DOORBELL_CONTROL1__DOORBELL_OFFSET__SHIFT; + data |= 1 << CP_MES_DOORBELL_CONTROL1__DOORBELL_EN__SHIFT; + WREG32_SOC15(GC, 0, mmCP_MES_DOORBELL_CONTROL1, data); + + data = RREG32_SOC15(GC, 0, mmCP_MES_DOORBELL_CONTROL2); + data &= ~(CP_MES_DOORBELL_CONTROL2__DOORBELL_OFFSET_MASK | + CP_MES_DOORBELL_CONTROL2__DOORBELL_EN_MASK | + CP_MES_DOORBELL_CONTROL2__DOORBELL_HIT_MASK); + data |= mes->aggregated_doorbells[AMDGPU_MES_PRIORITY_LEVEL_NORMAL] << + CP_MES_DOORBELL_CONTROL2__DOORBELL_OFFSET__SHIFT; + data |= 1 << CP_MES_DOORBELL_CONTROL2__DOORBELL_EN__SHIFT; + WREG32_SOC15(GC, 0, mmCP_MES_DOORBELL_CONTROL2, data); + + data = RREG32_SOC15(GC, 0, mmCP_MES_DOORBELL_CONTROL3); + data &= ~(CP_MES_DOORBELL_CONTROL3__DOORBELL_OFFSET_MASK | + CP_MES_DOORBELL_CONTROL3__DOORBELL_EN_MASK | + CP_MES_DOORBELL_CONTROL3__DOORBELL_HIT_MASK); + data |= mes->aggregated_doorbells[AMDGPU_MES_PRIORITY_LEVEL_MEDIUM] << + CP_MES_DOORBELL_CONTROL3__DOORBELL_OFFSET__SHIFT; + data |= 1 << CP_MES_DOORBELL_CONTROL3__DOORBELL_EN__SHIFT; + WREG32_SOC15(GC, 0, mmCP_MES_DOORBELL_CONTROL3, data); + + data = RREG32_SOC15(GC, 0, mmCP_MES_DOORBELL_CONTROL4); + data &= ~(CP_MES_DOORBELL_CONTROL4__DOORBELL_OFFSET_MASK | + CP_MES_DOORBELL_CONTROL4__DOORBELL_EN_MASK | + CP_MES_DOORBELL_CONTROL4__DOORBELL_HIT_MASK); + data |= mes->aggregated_doorbells[AMDGPU_MES_PRIORITY_LEVEL_HIGH] << + CP_MES_DOORBELL_CONTROL4__DOORBELL_OFFSET__SHIFT; + data |= 1 << CP_MES_DOORBELL_CONTROL4__DOORBELL_EN__SHIFT; + WREG32_SOC15(GC, 0, mmCP_MES_DOORBELL_CONTROL4, data); + + data = RREG32_SOC15(GC, 0, mmCP_MES_DOORBELL_CONTROL5); + data &= ~(CP_MES_DOORBELL_CONTROL5__DOORBELL_OFFSET_MASK | + CP_MES_DOORBELL_CONTROL5__DOORBELL_EN_MASK | + CP_MES_DOORBELL_CONTROL5__DOORBELL_HIT_MASK); + data |= mes->aggregated_doorbells[AMDGPU_MES_PRIORITY_LEVEL_REALTIME] << + CP_MES_DOORBELL_CONTROL5__DOORBELL_OFFSET__SHIFT; + data |= 1 << CP_MES_DOORBELL_CONTROL5__DOORBELL_EN__SHIFT; + WREG32_SOC15(GC, 0, mmCP_MES_DOORBELL_CONTROL5, data); + + data = 1 << CP_HQD_GFX_CONTROL__DB_UPDATED_MSG_EN__SHIFT; + WREG32_SOC15(GC, 0, mmCP_HQD_GFX_CONTROL, data); } static const struct amdgpu_mes_funcs mes_v10_1_funcs = { .add_hw_queue = mes_v10_1_add_hw_queue, .remove_hw_queue = mes_v10_1_remove_hw_queue, + .unmap_legacy_queue = mes_v10_1_unmap_legacy_queue, .suspend_gang = mes_v10_1_suspend_gang, .resume_gang = mes_v10_1_resume_gang, }; -static int mes_v10_1_init_microcode(struct amdgpu_device *adev) +static int mes_v10_1_init_microcode(struct amdgpu_device *adev, + enum admgpu_mes_pipe pipe) { const char *chip_name; char fw_name[30]; @@ -278,51 +384,69 @@ static int mes_v10_1_init_microcode(struct amdgpu_device *adev) const struct mes_firmware_header_v1_0 *mes_hdr; struct amdgpu_firmware_info *info; - switch (adev->asic_type) { - case CHIP_NAVI10: + switch (adev->ip_versions[GC_HWIP][0]) { + case IP_VERSION(10, 1, 10): chip_name = "navi10"; break; - case CHIP_SIENNA_CICHLID: + case IP_VERSION(10, 3, 0): chip_name = "sienna_cichlid"; break; default: BUG(); } - snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_mes.bin", chip_name); - err = request_firmware(&adev->mes.fw, fw_name, adev->dev); + if (pipe == AMDGPU_MES_SCHED_PIPE) + snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_mes.bin", + chip_name); + else + snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_mes1.bin", + chip_name); + + err = request_firmware(&adev->mes.fw[pipe], fw_name, adev->dev); if (err) return err; - err = amdgpu_ucode_validate(adev->mes.fw); + err = amdgpu_ucode_validate(adev->mes.fw[pipe]); if (err) { - release_firmware(adev->mes.fw); - adev->mes.fw = NULL; + release_firmware(adev->mes.fw[pipe]); + adev->mes.fw[pipe] = NULL; return err; } - mes_hdr = (const struct mes_firmware_header_v1_0 *)adev->mes.fw->data; - adev->mes.ucode_fw_version = le32_to_cpu(mes_hdr->mes_ucode_version); - adev->mes.ucode_fw_version = + mes_hdr = (const struct mes_firmware_header_v1_0 *) + adev->mes.fw[pipe]->data; + adev->mes.ucode_fw_version[pipe] = + le32_to_cpu(mes_hdr->mes_ucode_version); + adev->mes.ucode_fw_version[pipe] = le32_to_cpu(mes_hdr->mes_ucode_data_version); - adev->mes.uc_start_addr = + adev->mes.uc_start_addr[pipe] = le32_to_cpu(mes_hdr->mes_uc_start_addr_lo) | ((uint64_t)(le32_to_cpu(mes_hdr->mes_uc_start_addr_hi)) << 32); - adev->mes.data_start_addr = + adev->mes.data_start_addr[pipe] = le32_to_cpu(mes_hdr->mes_data_start_addr_lo) | ((uint64_t)(le32_to_cpu(mes_hdr->mes_data_start_addr_hi)) << 32); if (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) { - info = &adev->firmware.ucode[AMDGPU_UCODE_ID_CP_MES]; - info->ucode_id = AMDGPU_UCODE_ID_CP_MES; - info->fw = adev->mes.fw; + int ucode, ucode_data; + + if (pipe == AMDGPU_MES_SCHED_PIPE) { + ucode = AMDGPU_UCODE_ID_CP_MES; + ucode_data = AMDGPU_UCODE_ID_CP_MES_DATA; + } else { + ucode = AMDGPU_UCODE_ID_CP_MES1; + ucode_data = AMDGPU_UCODE_ID_CP_MES1_DATA; + } + + info = &adev->firmware.ucode[ucode]; + info->ucode_id = ucode; + info->fw = adev->mes.fw[pipe]; adev->firmware.fw_size += ALIGN(le32_to_cpu(mes_hdr->mes_ucode_size_bytes), PAGE_SIZE); - info = &adev->firmware.ucode[AMDGPU_UCODE_ID_CP_MES_DATA]; - info->ucode_id = AMDGPU_UCODE_ID_CP_MES_DATA; - info->fw = adev->mes.fw; + info = &adev->firmware.ucode[ucode_data]; + info->ucode_id = ucode_data; + info->fw = adev->mes.fw[pipe]; adev->firmware.fw_size += ALIGN(le32_to_cpu(mes_hdr->mes_ucode_data_size_bytes), PAGE_SIZE); @@ -331,13 +455,15 @@ static int mes_v10_1_init_microcode(struct amdgpu_device *adev) return 0; } -static void mes_v10_1_free_microcode(struct amdgpu_device *adev) +static void mes_v10_1_free_microcode(struct amdgpu_device *adev, + enum admgpu_mes_pipe pipe) { - release_firmware(adev->mes.fw); - adev->mes.fw = NULL; + release_firmware(adev->mes.fw[pipe]); + adev->mes.fw[pipe] = NULL; } -static int mes_v10_1_allocate_ucode_buffer(struct amdgpu_device *adev) +static int mes_v10_1_allocate_ucode_buffer(struct amdgpu_device *adev, + enum admgpu_mes_pipe pipe) { int r; const struct mes_firmware_header_v1_0 *mes_hdr; @@ -345,31 +471,32 @@ static int mes_v10_1_allocate_ucode_buffer(struct amdgpu_device *adev) unsigned fw_size; mes_hdr = (const struct mes_firmware_header_v1_0 *) - adev->mes.fw->data; + adev->mes.fw[pipe]->data; - fw_data = (const __le32 *)(adev->mes.fw->data + + fw_data = (const __le32 *)(adev->mes.fw[pipe]->data + le32_to_cpu(mes_hdr->mes_ucode_offset_bytes)); fw_size = le32_to_cpu(mes_hdr->mes_ucode_size_bytes); r = amdgpu_bo_create_reserved(adev, fw_size, PAGE_SIZE, AMDGPU_GEM_DOMAIN_GTT, - &adev->mes.ucode_fw_obj, - &adev->mes.ucode_fw_gpu_addr, - (void **)&adev->mes.ucode_fw_ptr); + &adev->mes.ucode_fw_obj[pipe], + &adev->mes.ucode_fw_gpu_addr[pipe], + (void **)&adev->mes.ucode_fw_ptr[pipe]); if (r) { dev_err(adev->dev, "(%d) failed to create mes fw bo\n", r); return r; } - memcpy(adev->mes.ucode_fw_ptr, fw_data, fw_size); + memcpy(adev->mes.ucode_fw_ptr[pipe], fw_data, fw_size); - amdgpu_bo_kunmap(adev->mes.ucode_fw_obj); - amdgpu_bo_unreserve(adev->mes.ucode_fw_obj); + amdgpu_bo_kunmap(adev->mes.ucode_fw_obj[pipe]); + amdgpu_bo_unreserve(adev->mes.ucode_fw_obj[pipe]); return 0; } -static int mes_v10_1_allocate_ucode_data_buffer(struct amdgpu_device *adev) +static int mes_v10_1_allocate_ucode_data_buffer(struct amdgpu_device *adev, + enum admgpu_mes_pipe pipe) { int r; const struct mes_firmware_header_v1_0 *mes_hdr; @@ -377,53 +504,65 @@ static int mes_v10_1_allocate_ucode_data_buffer(struct amdgpu_device *adev) unsigned fw_size; mes_hdr = (const struct mes_firmware_header_v1_0 *) - adev->mes.fw->data; + adev->mes.fw[pipe]->data; - fw_data = (const __le32 *)(adev->mes.fw->data + + fw_data = (const __le32 *)(adev->mes.fw[pipe]->data + le32_to_cpu(mes_hdr->mes_ucode_data_offset_bytes)); fw_size = le32_to_cpu(mes_hdr->mes_ucode_data_size_bytes); r = amdgpu_bo_create_reserved(adev, fw_size, 64 * 1024, AMDGPU_GEM_DOMAIN_GTT, - &adev->mes.data_fw_obj, - &adev->mes.data_fw_gpu_addr, - (void **)&adev->mes.data_fw_ptr); + &adev->mes.data_fw_obj[pipe], + &adev->mes.data_fw_gpu_addr[pipe], + (void **)&adev->mes.data_fw_ptr[pipe]); if (r) { dev_err(adev->dev, "(%d) failed to create mes data fw bo\n", r); return r; } - memcpy(adev->mes.data_fw_ptr, fw_data, fw_size); + memcpy(adev->mes.data_fw_ptr[pipe], fw_data, fw_size); - amdgpu_bo_kunmap(adev->mes.data_fw_obj); - amdgpu_bo_unreserve(adev->mes.data_fw_obj); + amdgpu_bo_kunmap(adev->mes.data_fw_obj[pipe]); + amdgpu_bo_unreserve(adev->mes.data_fw_obj[pipe]); return 0; } -static void mes_v10_1_free_ucode_buffers(struct amdgpu_device *adev) +static void mes_v10_1_free_ucode_buffers(struct amdgpu_device *adev, + enum admgpu_mes_pipe pipe) { - amdgpu_bo_free_kernel(&adev->mes.data_fw_obj, - &adev->mes.data_fw_gpu_addr, - (void **)&adev->mes.data_fw_ptr); + amdgpu_bo_free_kernel(&adev->mes.data_fw_obj[pipe], + &adev->mes.data_fw_gpu_addr[pipe], + (void **)&adev->mes.data_fw_ptr[pipe]); - amdgpu_bo_free_kernel(&adev->mes.ucode_fw_obj, - &adev->mes.ucode_fw_gpu_addr, - (void **)&adev->mes.ucode_fw_ptr); + amdgpu_bo_free_kernel(&adev->mes.ucode_fw_obj[pipe], + &adev->mes.ucode_fw_gpu_addr[pipe], + (void **)&adev->mes.ucode_fw_ptr[pipe]); } static void mes_v10_1_enable(struct amdgpu_device *adev, bool enable) { - uint32_t data = 0; + uint32_t pipe, data = 0; if (enable) { data = RREG32_SOC15(GC, 0, mmCP_MES_CNTL); data = REG_SET_FIELD(data, CP_MES_CNTL, MES_PIPE0_RESET, 1); + data = REG_SET_FIELD(data, CP_MES_CNTL, + MES_PIPE1_RESET, adev->enable_mes_kiq ? 1 : 0); WREG32_SOC15(GC, 0, mmCP_MES_CNTL, data); - /* set ucode start address */ - WREG32_SOC15(GC, 0, mmCP_MES_PRGRM_CNTR_START, - (uint32_t)(adev->mes.uc_start_addr) >> 2); + mutex_lock(&adev->srbm_mutex); + for (pipe = 0; pipe < AMDGPU_MAX_MES_PIPES; pipe++) { + if (!adev->enable_mes_kiq && + pipe == AMDGPU_MES_KIQ_PIPE) + continue; + + nv_grbm_select(adev, 3, pipe, 0, 0); + WREG32_SOC15(GC, 0, mmCP_MES_PRGRM_CNTR_START, + (uint32_t)(adev->mes.uc_start_addr[pipe]) >> 2); + } + nv_grbm_select(adev, 0, 0, 0, 0); + mutex_unlock(&adev->srbm_mutex); /* clear BYPASS_UNCACHED to avoid hangs after interrupt. */ data = RREG32_SOC15(GC, 0, mmCP_MES_DC_OP_CNTL); @@ -433,70 +572,77 @@ static void mes_v10_1_enable(struct amdgpu_device *adev, bool enable) /* unhalt MES and activate pipe0 */ data = REG_SET_FIELD(0, CP_MES_CNTL, MES_PIPE0_ACTIVE, 1); + data = REG_SET_FIELD(data, CP_MES_CNTL, MES_PIPE1_ACTIVE, + adev->enable_mes_kiq ? 1 : 0); WREG32_SOC15(GC, 0, mmCP_MES_CNTL, data); + udelay(100); } else { data = RREG32_SOC15(GC, 0, mmCP_MES_CNTL); data = REG_SET_FIELD(data, CP_MES_CNTL, MES_PIPE0_ACTIVE, 0); + data = REG_SET_FIELD(data, CP_MES_CNTL, MES_PIPE1_ACTIVE, 0); data = REG_SET_FIELD(data, CP_MES_CNTL, MES_INVALIDATE_ICACHE, 1); data = REG_SET_FIELD(data, CP_MES_CNTL, MES_PIPE0_RESET, 1); + data = REG_SET_FIELD(data, CP_MES_CNTL, MES_PIPE1_RESET, + adev->enable_mes_kiq ? 1 : 0); data = REG_SET_FIELD(data, CP_MES_CNTL, MES_HALT, 1); WREG32_SOC15(GC, 0, mmCP_MES_CNTL, data); } } /* This function is for backdoor MES firmware */ -static int mes_v10_1_load_microcode(struct amdgpu_device *adev) +static int mes_v10_1_load_microcode(struct amdgpu_device *adev, + enum admgpu_mes_pipe pipe) { int r; uint32_t data; - if (!adev->mes.fw) + mes_v10_1_enable(adev, false); + + if (!adev->mes.fw[pipe]) return -EINVAL; - r = mes_v10_1_allocate_ucode_buffer(adev); + r = mes_v10_1_allocate_ucode_buffer(adev, pipe); if (r) return r; - r = mes_v10_1_allocate_ucode_data_buffer(adev); + r = mes_v10_1_allocate_ucode_data_buffer(adev, pipe); if (r) { - mes_v10_1_free_ucode_buffers(adev); + mes_v10_1_free_ucode_buffers(adev, pipe); return r; } - mes_v10_1_enable(adev, false); - WREG32_SOC15(GC, 0, mmCP_MES_IC_BASE_CNTL, 0); mutex_lock(&adev->srbm_mutex); /* me=3, pipe=0, queue=0 */ - nv_grbm_select(adev, 3, 0, 0, 0); + nv_grbm_select(adev, 3, pipe, 0, 0); /* set ucode start address */ WREG32_SOC15(GC, 0, mmCP_MES_PRGRM_CNTR_START, - (uint32_t)(adev->mes.uc_start_addr) >> 2); + (uint32_t)(adev->mes.uc_start_addr[pipe]) >> 2); /* set ucode fimrware address */ WREG32_SOC15(GC, 0, mmCP_MES_IC_BASE_LO, - lower_32_bits(adev->mes.ucode_fw_gpu_addr)); + lower_32_bits(adev->mes.ucode_fw_gpu_addr[pipe])); WREG32_SOC15(GC, 0, mmCP_MES_IC_BASE_HI, - upper_32_bits(adev->mes.ucode_fw_gpu_addr)); + upper_32_bits(adev->mes.ucode_fw_gpu_addr[pipe])); /* set ucode instruction cache boundary to 2M-1 */ WREG32_SOC15(GC, 0, mmCP_MES_MIBOUND_LO, 0x1FFFFF); /* set ucode data firmware address */ WREG32_SOC15(GC, 0, mmCP_MES_MDBASE_LO, - lower_32_bits(adev->mes.data_fw_gpu_addr)); + lower_32_bits(adev->mes.data_fw_gpu_addr[pipe])); WREG32_SOC15(GC, 0, mmCP_MES_MDBASE_HI, - upper_32_bits(adev->mes.data_fw_gpu_addr)); + upper_32_bits(adev->mes.data_fw_gpu_addr[pipe])); /* Set 0x3FFFF (256K-1) to CP_MES_MDBOUND_LO */ WREG32_SOC15(GC, 0, mmCP_MES_MDBOUND_LO, 0x3FFFF); /* invalidate ICACHE */ - switch (adev->asic_type) { - case CHIP_SIENNA_CICHLID: + switch (adev->ip_versions[GC_HWIP][0]) { + case IP_VERSION(10, 3, 0): data = RREG32_SOC15(GC, 0, mmCP_MES_IC_OP_CNTL_Sienna_Cichlid); break; default: @@ -505,8 +651,8 @@ static int mes_v10_1_load_microcode(struct amdgpu_device *adev) } data = REG_SET_FIELD(data, CP_MES_IC_OP_CNTL, PRIME_ICACHE, 0); data = REG_SET_FIELD(data, CP_MES_IC_OP_CNTL, INVALIDATE_CACHE, 1); - switch (adev->asic_type) { - case CHIP_SIENNA_CICHLID: + switch (adev->ip_versions[GC_HWIP][0]) { + case IP_VERSION(10, 3, 0): WREG32_SOC15(GC, 0, mmCP_MES_IC_OP_CNTL_Sienna_Cichlid, data); break; default: @@ -515,8 +661,8 @@ static int mes_v10_1_load_microcode(struct amdgpu_device *adev) } /* prime the ICACHE. */ - switch (adev->asic_type) { - case CHIP_SIENNA_CICHLID: + switch (adev->ip_versions[GC_HWIP][0]) { + case IP_VERSION(10, 3, 0): data = RREG32_SOC15(GC, 0, mmCP_MES_IC_OP_CNTL_Sienna_Cichlid); break; default: @@ -524,8 +670,8 @@ static int mes_v10_1_load_microcode(struct amdgpu_device *adev) break; } data = REG_SET_FIELD(data, CP_MES_IC_OP_CNTL, PRIME_ICACHE, 1); - switch (adev->asic_type) { - case CHIP_SIENNA_CICHLID: + switch (adev->ip_versions[GC_HWIP][0]) { + case IP_VERSION(10, 3, 0): WREG32_SOC15(GC, 0, mmCP_MES_IC_OP_CNTL_Sienna_Cichlid, data); break; default: @@ -539,61 +685,32 @@ static int mes_v10_1_load_microcode(struct amdgpu_device *adev) return 0; } -static int mes_v10_1_allocate_eop_buf(struct amdgpu_device *adev) +static int mes_v10_1_allocate_eop_buf(struct amdgpu_device *adev, + enum admgpu_mes_pipe pipe) { int r; u32 *eop; r = amdgpu_bo_create_reserved(adev, MES_EOP_SIZE, PAGE_SIZE, - AMDGPU_GEM_DOMAIN_GTT, - &adev->mes.eop_gpu_obj, - &adev->mes.eop_gpu_addr, - (void **)&eop); + AMDGPU_GEM_DOMAIN_GTT, + &adev->mes.eop_gpu_obj[pipe], + &adev->mes.eop_gpu_addr[pipe], + (void **)&eop); if (r) { dev_warn(adev->dev, "(%d) create EOP bo failed\n", r); return r; } - memset(eop, 0, adev->mes.eop_gpu_obj->tbo.base.size); - - amdgpu_bo_kunmap(adev->mes.eop_gpu_obj); - amdgpu_bo_unreserve(adev->mes.eop_gpu_obj); + memset(eop, 0, adev->mes.eop_gpu_obj[pipe]->tbo.base.size); - return 0; -} - -static int mes_v10_1_allocate_mem_slots(struct amdgpu_device *adev) -{ - int r; - - r = amdgpu_device_wb_get(adev, &adev->mes.sch_ctx_offs); - if (r) { - dev_err(adev->dev, - "(%d) mes sch_ctx_offs wb alloc failed\n", r); - return r; - } - adev->mes.sch_ctx_gpu_addr = - adev->wb.gpu_addr + (adev->mes.sch_ctx_offs * 4); - adev->mes.sch_ctx_ptr = - (uint64_t *)&adev->wb.wb[adev->mes.sch_ctx_offs]; - - r = amdgpu_device_wb_get(adev, &adev->mes.query_status_fence_offs); - if (r) { - dev_err(adev->dev, - "(%d) query_status_fence_offs wb alloc failed\n", r); - return r; - } - adev->mes.query_status_fence_gpu_addr = - adev->wb.gpu_addr + (adev->mes.query_status_fence_offs * 4); - adev->mes.query_status_fence_ptr = - (uint64_t *)&adev->wb.wb[adev->mes.query_status_fence_offs]; + amdgpu_bo_kunmap(adev->mes.eop_gpu_obj[pipe]); + amdgpu_bo_unreserve(adev->mes.eop_gpu_obj[pipe]); return 0; } static int mes_v10_1_mqd_init(struct amdgpu_ring *ring) { - struct amdgpu_device *adev = ring->adev; struct v10_compute_mqd *mqd = ring->mqd_ptr; uint64_t hqd_gpu_addr, wb_gpu_addr, eop_base_addr; uint32_t tmp; @@ -607,38 +724,18 @@ static int mes_v10_1_mqd_init(struct amdgpu_ring *ring) mqd->compute_misc_reserved = 0x00000003; eop_base_addr = ring->eop_gpu_addr >> 8; - mqd->cp_hqd_eop_base_addr_lo = eop_base_addr; - mqd->cp_hqd_eop_base_addr_hi = upper_32_bits(eop_base_addr); /* set the EOP size, register value is 2^(EOP_SIZE+1) dwords */ - tmp = RREG32_SOC15(GC, 0, mmCP_HQD_EOP_CONTROL); + tmp = mmCP_HQD_EOP_CONTROL_DEFAULT; tmp = REG_SET_FIELD(tmp, CP_HQD_EOP_CONTROL, EOP_SIZE, (order_base_2(MES_EOP_SIZE / 4) - 1)); + mqd->cp_hqd_eop_base_addr_lo = lower_32_bits(eop_base_addr); + mqd->cp_hqd_eop_base_addr_hi = upper_32_bits(eop_base_addr); mqd->cp_hqd_eop_control = tmp; - /* enable doorbell? */ - tmp = RREG32_SOC15(GC, 0, mmCP_HQD_PQ_DOORBELL_CONTROL); - - if (ring->use_doorbell) { - tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_DOORBELL_CONTROL, - DOORBELL_OFFSET, ring->doorbell_index); - tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_DOORBELL_CONTROL, - DOORBELL_EN, 1); - tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_DOORBELL_CONTROL, - DOORBELL_SOURCE, 0); - tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_DOORBELL_CONTROL, - DOORBELL_HIT, 0); - } - else - tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_DOORBELL_CONTROL, - DOORBELL_EN, 0); - - mqd->cp_hqd_pq_doorbell_control = tmp; - /* disable the queue if it's active */ ring->wptr = 0; - mqd->cp_hqd_dequeue_request = 0; mqd->cp_hqd_pq_rptr = 0; mqd->cp_hqd_pq_wptr_lo = 0; mqd->cp_hqd_pq_wptr_hi = 0; @@ -648,17 +745,28 @@ static int mes_v10_1_mqd_init(struct amdgpu_ring *ring) mqd->cp_mqd_base_addr_hi = upper_32_bits(ring->mqd_gpu_addr); /* set MQD vmid to 0 */ - tmp = RREG32_SOC15(GC, 0, mmCP_MQD_CONTROL); + tmp = mmCP_MQD_CONTROL_DEFAULT; tmp = REG_SET_FIELD(tmp, CP_MQD_CONTROL, VMID, 0); mqd->cp_mqd_control = tmp; /* set the pointer to the HQD, this is similar CP_RB0_BASE/_HI */ hqd_gpu_addr = ring->gpu_addr >> 8; - mqd->cp_hqd_pq_base_lo = hqd_gpu_addr; + mqd->cp_hqd_pq_base_lo = lower_32_bits(hqd_gpu_addr); mqd->cp_hqd_pq_base_hi = upper_32_bits(hqd_gpu_addr); + /* set the wb address whether it's enabled or not */ + wb_gpu_addr = ring->rptr_gpu_addr; + mqd->cp_hqd_pq_rptr_report_addr_lo = wb_gpu_addr & 0xfffffffc; + mqd->cp_hqd_pq_rptr_report_addr_hi = + upper_32_bits(wb_gpu_addr) & 0xffff; + + /* only used if CP_PQ_WPTR_POLL_CNTL.CP_PQ_WPTR_POLL_CNTL__EN_MASK=1 */ + wb_gpu_addr = ring->wptr_gpu_addr; + mqd->cp_hqd_pq_wptr_poll_addr_lo = wb_gpu_addr & 0xfffffff8; + mqd->cp_hqd_pq_wptr_poll_addr_hi = upper_32_bits(wb_gpu_addr) & 0xffff; + /* set up the HQD, this is similar to CP_RB0_CNTL */ - tmp = RREG32_SOC15(GC, 0, mmCP_HQD_PQ_CONTROL); + tmp = mmCP_HQD_PQ_CONTROL_DEFAULT; tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, QUEUE_SIZE, (order_base_2(ring->ring_size / 4) - 1)); tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, RPTR_BLOCK_SIZE, @@ -666,30 +774,18 @@ static int mes_v10_1_mqd_init(struct amdgpu_ring *ring) #ifdef __BIG_ENDIAN tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, ENDIAN_SWAP, 1); #endif - tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, UNORD_DISPATCH, 0); + tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, UNORD_DISPATCH, 1); tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, TUNNEL_DISPATCH, 0); tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, PRIV_STATE, 1); tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, KMD_QUEUE, 1); + tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, NO_UPDATE_RPTR, 1); mqd->cp_hqd_pq_control = tmp; - /* set the wb address whether it's enabled or not */ - wb_gpu_addr = adev->wb.gpu_addr + (ring->rptr_offs * 4); - mqd->cp_hqd_pq_rptr_report_addr_lo = wb_gpu_addr & 0xfffffffc; - mqd->cp_hqd_pq_rptr_report_addr_hi = - upper_32_bits(wb_gpu_addr) & 0xffff; - - /* only used if CP_PQ_WPTR_POLL_CNTL.CP_PQ_WPTR_POLL_CNTL__EN_MASK=1 */ - wb_gpu_addr = adev->wb.gpu_addr + (ring->wptr_offs * 4); - mqd->cp_hqd_pq_wptr_poll_addr_lo = wb_gpu_addr & 0xfffffff8; - mqd->cp_hqd_pq_wptr_poll_addr_hi = upper_32_bits(wb_gpu_addr) & 0xffff; - + /* enable doorbell? */ tmp = 0; - /* enable the doorbell if requested */ if (ring->use_doorbell) { - tmp = RREG32_SOC15(GC, 0, mmCP_HQD_PQ_DOORBELL_CONTROL); tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_DOORBELL_CONTROL, - DOORBELL_OFFSET, ring->doorbell_index); - + DOORBELL_OFFSET, ring->doorbell_index); tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_DOORBELL_CONTROL, DOORBELL_EN, 1); tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_DOORBELL_CONTROL, @@ -697,30 +793,28 @@ static int mes_v10_1_mqd_init(struct amdgpu_ring *ring) tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_DOORBELL_CONTROL, DOORBELL_HIT, 0); } - + else + tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_DOORBELL_CONTROL, + DOORBELL_EN, 0); mqd->cp_hqd_pq_doorbell_control = tmp; - /* reset read and write pointers, similar to CP_RB0_WPTR/_RPTR */ - ring->wptr = 0; - mqd->cp_hqd_pq_rptr = RREG32_SOC15(GC, 0, mmCP_HQD_PQ_RPTR); - - /* set the vmid for the queue */ mqd->cp_hqd_vmid = 0; - - tmp = RREG32_SOC15(GC, 0, mmCP_HQD_PERSISTENT_STATE); - tmp = REG_SET_FIELD(tmp, CP_HQD_PERSISTENT_STATE, PRELOAD_SIZE, 0x53); - mqd->cp_hqd_persistent_state = tmp; - - /* set MIN_IB_AVAIL_SIZE */ - tmp = RREG32_SOC15(GC, 0, mmCP_HQD_IB_CONTROL); - tmp = REG_SET_FIELD(tmp, CP_HQD_IB_CONTROL, MIN_IB_AVAIL_SIZE, 3); - mqd->cp_hqd_ib_control = tmp; - /* activate the queue */ mqd->cp_hqd_active = 1; + mqd->cp_hqd_persistent_state = mmCP_HQD_PERSISTENT_STATE_DEFAULT; + mqd->cp_hqd_ib_control = mmCP_HQD_IB_CONTROL_DEFAULT; + mqd->cp_hqd_iq_timer = mmCP_HQD_IQ_TIMER_DEFAULT; + mqd->cp_hqd_quantum = mmCP_HQD_QUANTUM_DEFAULT; + + tmp = mmCP_HQD_GFX_CONTROL_DEFAULT; + tmp = REG_SET_FIELD(tmp, CP_HQD_GFX_CONTROL, DB_UPDATED_MSG_EN, 1); + /* offset: 184 - this is used for CP_HQD_GFX_CONTROL */ + mqd->cp_hqd_suspend_cntl_stack_offset = tmp; + return 0; } +#if 0 static void mes_v10_1_queue_init_register(struct amdgpu_ring *ring) { struct v10_compute_mqd *mqd = ring->mqd_ptr; @@ -728,7 +822,7 @@ static void mes_v10_1_queue_init_register(struct amdgpu_ring *ring) uint32_t data = 0; mutex_lock(&adev->srbm_mutex); - nv_grbm_select(adev, 3, 0, 0, 0); + nv_grbm_select(adev, 3, ring->pipe, 0, 0); /* set CP_HQD_VMID.VMID = 0. */ data = RREG32_SOC15(GC, 0, mmCP_HQD_VMID); @@ -782,8 +876,8 @@ static void mes_v10_1_queue_init_register(struct amdgpu_ring *ring) nv_grbm_select(adev, 0, 0, 0, 0); mutex_unlock(&adev->srbm_mutex); } +#endif -#if 0 static int mes_v10_1_kiq_enable_queue(struct amdgpu_device *adev) { struct amdgpu_kiq *kiq = &adev->gfx.kiq; @@ -806,9 +900,9 @@ static int mes_v10_1_kiq_enable_queue(struct amdgpu_device *adev) DRM_ERROR("kfq enable failed\n"); kiq_ring->sched.ready = false; } + return r; } -#endif static int mes_v10_1_queue_init(struct amdgpu_device *adev) { @@ -818,13 +912,9 @@ static int mes_v10_1_queue_init(struct amdgpu_device *adev) if (r) return r; -#if 0 r = mes_v10_1_kiq_enable_queue(adev); if (r) return r; -#else - mes_v10_1_queue_init_register(&adev->mes.ring); -#endif return 0; } @@ -843,8 +933,8 @@ static int mes_v10_1_ring_init(struct amdgpu_device *adev) ring->ring_obj = NULL; ring->use_doorbell = true; - ring->doorbell_index = adev->doorbell_index.mes_ring << 1; - ring->eop_gpu_addr = adev->mes.eop_gpu_addr; + ring->doorbell_index = adev->doorbell_index.mes_ring0 << 1; + ring->eop_gpu_addr = adev->mes.eop_gpu_addr[AMDGPU_MES_SCHED_PIPE]; ring->no_scheduler = true; sprintf(ring->name, "mes_%d.%d.%d", ring->me, ring->pipe, ring->queue); @@ -852,10 +942,43 @@ static int mes_v10_1_ring_init(struct amdgpu_device *adev) AMDGPU_RING_PRIO_DEFAULT, NULL); } -static int mes_v10_1_mqd_sw_init(struct amdgpu_device *adev) +static int mes_v10_1_kiq_ring_init(struct amdgpu_device *adev) +{ + struct amdgpu_ring *ring; + + spin_lock_init(&adev->gfx.kiq.ring_lock); + + ring = &adev->gfx.kiq.ring; + + ring->me = 3; + ring->pipe = 1; + ring->queue = 0; + + ring->adev = NULL; + ring->ring_obj = NULL; + ring->use_doorbell = true; + ring->doorbell_index = adev->doorbell_index.mes_ring1 << 1; + ring->eop_gpu_addr = adev->mes.eop_gpu_addr[AMDGPU_MES_KIQ_PIPE]; + ring->no_scheduler = true; + sprintf(ring->name, "mes_kiq_%d.%d.%d", + ring->me, ring->pipe, ring->queue); + + return amdgpu_ring_init(adev, ring, 1024, NULL, 0, + AMDGPU_RING_PRIO_DEFAULT, NULL); +} + +static int mes_v10_1_mqd_sw_init(struct amdgpu_device *adev, + enum admgpu_mes_pipe pipe) { int r, mqd_size = sizeof(struct v10_compute_mqd); - struct amdgpu_ring *ring = &adev->mes.ring; + struct amdgpu_ring *ring; + + if (pipe == AMDGPU_MES_KIQ_PIPE) + ring = &adev->gfx.kiq.ring; + else if (pipe == AMDGPU_MES_SCHED_PIPE) + ring = &adev->mes.ring; + else + BUG(); if (ring->mqd_obj) return 0; @@ -867,10 +990,11 @@ static int mes_v10_1_mqd_sw_init(struct amdgpu_device *adev) dev_warn(adev->dev, "failed to create ring mqd bo (%d)", r); return r; } + memset(ring->mqd_ptr, 0, mqd_size); /* prepare MQD backup */ - adev->mes.mqd_backup = kmalloc(mqd_size, GFP_KERNEL); - if (!adev->mes.mqd_backup) + adev->mes.mqd_backup[pipe] = kmalloc(mqd_size, GFP_KERNEL); + if (!adev->mes.mqd_backup[pipe]) dev_warn(adev->dev, "no memory to create MQD backup for ring %s\n", ring->name); @@ -880,29 +1004,41 @@ static int mes_v10_1_mqd_sw_init(struct amdgpu_device *adev) static int mes_v10_1_sw_init(void *handle) { - int r; struct amdgpu_device *adev = (struct amdgpu_device *)handle; + int pipe, r; adev->mes.adev = adev; adev->mes.funcs = &mes_v10_1_funcs; + adev->mes.kiq_hw_init = &mes_v10_1_kiq_hw_init; - r = mes_v10_1_init_microcode(adev); + r = amdgpu_mes_init(adev); if (r) return r; - r = mes_v10_1_allocate_eop_buf(adev); - if (r) - return r; + for (pipe = 0; pipe < AMDGPU_MAX_MES_PIPES; pipe++) { + if (!adev->enable_mes_kiq && pipe == AMDGPU_MES_KIQ_PIPE) + continue; - r = mes_v10_1_mqd_sw_init(adev); - if (r) - return r; + r = mes_v10_1_init_microcode(adev, pipe); + if (r) + return r; - r = mes_v10_1_ring_init(adev); - if (r) - return r; + r = mes_v10_1_allocate_eop_buf(adev, pipe); + if (r) + return r; - r = mes_v10_1_allocate_mem_slots(adev); + r = mes_v10_1_mqd_sw_init(adev, pipe); + if (r) + return r; + } + + if (adev->enable_mes_kiq) { + r = mes_v10_1_kiq_ring_init(adev); + if (r) + return r; + } + + r = mes_v10_1_ring_init(adev); if (r) return r; @@ -912,40 +1048,116 @@ static int mes_v10_1_sw_init(void *handle) static int mes_v10_1_sw_fini(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; + int pipe; amdgpu_device_wb_free(adev, adev->mes.sch_ctx_offs); amdgpu_device_wb_free(adev, adev->mes.query_status_fence_offs); - kfree(adev->mes.mqd_backup); + for (pipe = 0; pipe < AMDGPU_MAX_MES_PIPES; pipe++) { + kfree(adev->mes.mqd_backup[pipe]); + + amdgpu_bo_free_kernel(&adev->mes.eop_gpu_obj[pipe], + &adev->mes.eop_gpu_addr[pipe], + NULL); + + mes_v10_1_free_microcode(adev, pipe); + } + + amdgpu_bo_free_kernel(&adev->gfx.kiq.ring.mqd_obj, + &adev->gfx.kiq.ring.mqd_gpu_addr, + &adev->gfx.kiq.ring.mqd_ptr); amdgpu_bo_free_kernel(&adev->mes.ring.mqd_obj, &adev->mes.ring.mqd_gpu_addr, &adev->mes.ring.mqd_ptr); - amdgpu_bo_free_kernel(&adev->mes.eop_gpu_obj, - &adev->mes.eop_gpu_addr, - NULL); - - mes_v10_1_free_microcode(adev); + amdgpu_ring_fini(&adev->gfx.kiq.ring); + amdgpu_ring_fini(&adev->mes.ring); + amdgpu_mes_fini(adev); return 0; } -static int mes_v10_1_hw_init(void *handle) +static void mes_v10_1_kiq_setting(struct amdgpu_ring *ring) { - int r; - struct amdgpu_device *adev = (struct amdgpu_device *)handle; + uint32_t tmp; + struct amdgpu_device *adev = ring->adev; + + /* tell RLC which is KIQ queue */ + switch (adev->ip_versions[GC_HWIP][0]) { + case IP_VERSION(10, 3, 0): + case IP_VERSION(10, 3, 2): + case IP_VERSION(10, 3, 1): + case IP_VERSION(10, 3, 4): + tmp = RREG32_SOC15(GC, 0, mmRLC_CP_SCHEDULERS_Sienna_Cichlid); + tmp &= 0xffffff00; + tmp |= (ring->me << 5) | (ring->pipe << 3) | (ring->queue); + WREG32_SOC15(GC, 0, mmRLC_CP_SCHEDULERS_Sienna_Cichlid, tmp); + tmp |= 0x80; + WREG32_SOC15(GC, 0, mmRLC_CP_SCHEDULERS_Sienna_Cichlid, tmp); + break; + default: + tmp = RREG32_SOC15(GC, 0, mmRLC_CP_SCHEDULERS); + tmp &= 0xffffff00; + tmp |= (ring->me << 5) | (ring->pipe << 3) | (ring->queue); + WREG32_SOC15(GC, 0, mmRLC_CP_SCHEDULERS, tmp); + tmp |= 0x80; + WREG32_SOC15(GC, 0, mmRLC_CP_SCHEDULERS, tmp); + break; + } +} + +static int mes_v10_1_kiq_hw_init(struct amdgpu_device *adev) +{ + int r = 0; if (adev->firmware.load_type == AMDGPU_FW_LOAD_DIRECT) { - r = mes_v10_1_load_microcode(adev); + r = mes_v10_1_load_microcode(adev, AMDGPU_MES_KIQ_PIPE); + if (r) { + DRM_ERROR("failed to load MES kiq fw, r=%d\n", r); + return r; + } + + r = mes_v10_1_load_microcode(adev, AMDGPU_MES_SCHED_PIPE); if (r) { - DRM_ERROR("failed to MES fw, r=%d\n", r); + DRM_ERROR("failed to load MES fw, r=%d\n", r); return r; } } mes_v10_1_enable(adev, true); + mes_v10_1_kiq_setting(&adev->gfx.kiq.ring); + + r = mes_v10_1_queue_init(adev); + if (r) + goto failure; + + return r; + +failure: + mes_v10_1_hw_fini(adev); + return r; +} + +static int mes_v10_1_hw_init(void *handle) +{ + int r; + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + + if (!adev->enable_mes_kiq) { + if (adev->firmware.load_type == AMDGPU_FW_LOAD_DIRECT) { + r = mes_v10_1_load_microcode(adev, + AMDGPU_MES_SCHED_PIPE); + if (r) { + DRM_ERROR("failed to MES fw, r=%d\n", r); + return r; + } + } + + mes_v10_1_enable(adev, true); + } + r = mes_v10_1_queue_init(adev); if (r) goto failure; @@ -954,12 +1166,22 @@ static int mes_v10_1_hw_init(void *handle) if (r) goto failure; + mes_v10_1_init_aggregated_doorbell(&adev->mes); + r = mes_v10_1_query_sched_status(&adev->mes); if (r) { DRM_ERROR("MES is busy\n"); goto failure; } + /* + * Disable KIQ ring usage from the driver once MES is enabled. + * MES uses KIQ ring exclusively so driver cannot access KIQ ring + * with MES enabled. + */ + adev->gfx.kiq.ring.sched.ready = false; + adev->mes.ring.sched.ready = true; + return 0; failure: @@ -971,26 +1193,55 @@ static int mes_v10_1_hw_fini(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; + adev->mes.ring.sched.ready = false; + mes_v10_1_enable(adev, false); - if (adev->firmware.load_type == AMDGPU_FW_LOAD_DIRECT) - mes_v10_1_free_ucode_buffers(adev); + if (adev->firmware.load_type == AMDGPU_FW_LOAD_DIRECT) { + mes_v10_1_free_ucode_buffers(adev, AMDGPU_MES_KIQ_PIPE); + mes_v10_1_free_ucode_buffers(adev, AMDGPU_MES_SCHED_PIPE); + } return 0; } static int mes_v10_1_suspend(void *handle) { - return 0; + int r; + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + + r = amdgpu_mes_suspend(adev); + if (r) + return r; + + return mes_v10_1_hw_fini(adev); } static int mes_v10_1_resume(void *handle) { + int r; + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + + r = mes_v10_1_hw_init(adev); + if (r) + return r; + + return amdgpu_mes_resume(adev); +} + +static int mes_v10_0_late_init(void *handle) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + + if (!amdgpu_in_reset(adev)) + amdgpu_mes_self_test(adev); + return 0; } static const struct amd_ip_funcs mes_v10_1_ip_funcs = { .name = "mes_v10_1", + .late_init = mes_v10_0_late_init, .sw_init = mes_v10_1_sw_init, .sw_fini = mes_v10_1_sw_fini, .hw_init = mes_v10_1_hw_init, diff --git a/drivers/gpu/drm/amd/amdgpu/mes_v11_0.c b/drivers/gpu/drm/amd/amdgpu/mes_v11_0.c new file mode 100644 index 000000000000..f141fadd2d86 --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/mes_v11_0.c @@ -0,0 +1,1366 @@ +/* + * Copyright 2019 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#include <linux/firmware.h> +#include <linux/module.h> +#include "amdgpu.h" +#include "soc15_common.h" +#include "soc21.h" +#include "gc/gc_11_0_0_offset.h" +#include "gc/gc_11_0_0_sh_mask.h" +#include "gc/gc_11_0_0_default.h" +#include "v11_structs.h" +#include "mes_v11_api_def.h" + +MODULE_FIRMWARE("amdgpu/gc_11_0_0_mes.bin"); +MODULE_FIRMWARE("amdgpu/gc_11_0_0_mes1.bin"); +MODULE_FIRMWARE("amdgpu/gc_11_0_1_mes.bin"); +MODULE_FIRMWARE("amdgpu/gc_11_0_1_mes1.bin"); +MODULE_FIRMWARE("amdgpu/gc_11_0_2_mes.bin"); +MODULE_FIRMWARE("amdgpu/gc_11_0_2_mes1.bin"); +MODULE_FIRMWARE("amdgpu/gc_11_0_3_mes.bin"); +MODULE_FIRMWARE("amdgpu/gc_11_0_3_mes1.bin"); + +static int mes_v11_0_hw_fini(void *handle); +static int mes_v11_0_kiq_hw_init(struct amdgpu_device *adev); +static int mes_v11_0_kiq_hw_fini(struct amdgpu_device *adev); + +#define MES_EOP_SIZE 2048 + +static void mes_v11_0_ring_set_wptr(struct amdgpu_ring *ring) +{ + struct amdgpu_device *adev = ring->adev; + + if (ring->use_doorbell) { + atomic64_set((atomic64_t *)ring->wptr_cpu_addr, + ring->wptr); + WDOORBELL64(ring->doorbell_index, ring->wptr); + } else { + BUG(); + } +} + +static u64 mes_v11_0_ring_get_rptr(struct amdgpu_ring *ring) +{ + return *ring->rptr_cpu_addr; +} + +static u64 mes_v11_0_ring_get_wptr(struct amdgpu_ring *ring) +{ + u64 wptr; + + if (ring->use_doorbell) + wptr = atomic64_read((atomic64_t *)ring->wptr_cpu_addr); + else + BUG(); + return wptr; +} + +static const struct amdgpu_ring_funcs mes_v11_0_ring_funcs = { + .type = AMDGPU_RING_TYPE_MES, + .align_mask = 1, + .nop = 0, + .support_64bit_ptrs = true, + .get_rptr = mes_v11_0_ring_get_rptr, + .get_wptr = mes_v11_0_ring_get_wptr, + .set_wptr = mes_v11_0_ring_set_wptr, + .insert_nop = amdgpu_ring_insert_nop, +}; + +static int mes_v11_0_submit_pkt_and_poll_completion(struct amdgpu_mes *mes, + void *pkt, int size, + int api_status_off) +{ + int ndw = size / 4; + signed long r; + union MESAPI__ADD_QUEUE *x_pkt = pkt; + struct MES_API_STATUS *api_status; + struct amdgpu_device *adev = mes->adev; + struct amdgpu_ring *ring = &mes->ring; + unsigned long flags; + signed long timeout = adev->usec_timeout; + + if (amdgpu_emu_mode) { + timeout *= 100; + } else if (amdgpu_sriov_vf(adev)) { + /* Worst case in sriov where all other 15 VF timeout, each VF needs about 600ms */ + timeout = 15 * 600 * 1000; + } + BUG_ON(size % 4 != 0); + + spin_lock_irqsave(&mes->ring_lock, flags); + if (amdgpu_ring_alloc(ring, ndw)) { + spin_unlock_irqrestore(&mes->ring_lock, flags); + return -ENOMEM; + } + + api_status = (struct MES_API_STATUS *)((char *)pkt + api_status_off); + api_status->api_completion_fence_addr = mes->ring.fence_drv.gpu_addr; + api_status->api_completion_fence_value = ++mes->ring.fence_drv.sync_seq; + + amdgpu_ring_write_multiple(ring, pkt, ndw); + amdgpu_ring_commit(ring); + spin_unlock_irqrestore(&mes->ring_lock, flags); + + DRM_DEBUG("MES msg=%d was emitted\n", x_pkt->header.opcode); + + r = amdgpu_fence_wait_polling(ring, ring->fence_drv.sync_seq, + timeout); + if (r < 1) { + DRM_ERROR("MES failed to response msg=%d\n", + x_pkt->header.opcode); + return -ETIMEDOUT; + } + + return 0; +} + +static int convert_to_mes_queue_type(int queue_type) +{ + if (queue_type == AMDGPU_RING_TYPE_GFX) + return MES_QUEUE_TYPE_GFX; + else if (queue_type == AMDGPU_RING_TYPE_COMPUTE) + return MES_QUEUE_TYPE_COMPUTE; + else if (queue_type == AMDGPU_RING_TYPE_SDMA) + return MES_QUEUE_TYPE_SDMA; + else + BUG(); + return -1; +} + +static int mes_v11_0_add_hw_queue(struct amdgpu_mes *mes, + struct mes_add_queue_input *input) +{ + struct amdgpu_device *adev = mes->adev; + union MESAPI__ADD_QUEUE mes_add_queue_pkt; + struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0]; + uint32_t vm_cntx_cntl = hub->vm_cntx_cntl; + + memset(&mes_add_queue_pkt, 0, sizeof(mes_add_queue_pkt)); + + mes_add_queue_pkt.header.type = MES_API_TYPE_SCHEDULER; + mes_add_queue_pkt.header.opcode = MES_SCH_API_ADD_QUEUE; + mes_add_queue_pkt.header.dwsize = API_FRAME_SIZE_IN_DWORDS; + + mes_add_queue_pkt.process_id = input->process_id; + mes_add_queue_pkt.page_table_base_addr = input->page_table_base_addr; + mes_add_queue_pkt.process_va_start = input->process_va_start; + mes_add_queue_pkt.process_va_end = input->process_va_end; + mes_add_queue_pkt.process_quantum = input->process_quantum; + mes_add_queue_pkt.process_context_addr = input->process_context_addr; + mes_add_queue_pkt.gang_quantum = input->gang_quantum; + mes_add_queue_pkt.gang_context_addr = input->gang_context_addr; + mes_add_queue_pkt.inprocess_gang_priority = + input->inprocess_gang_priority; + mes_add_queue_pkt.gang_global_priority_level = + input->gang_global_priority_level; + mes_add_queue_pkt.doorbell_offset = input->doorbell_offset; + mes_add_queue_pkt.mqd_addr = input->mqd_addr; + + if (((adev->mes.sched_version & AMDGPU_MES_API_VERSION_MASK) >> + AMDGPU_MES_API_VERSION_SHIFT) >= 2) + mes_add_queue_pkt.wptr_addr = input->wptr_mc_addr; + else + mes_add_queue_pkt.wptr_addr = input->wptr_addr; + + mes_add_queue_pkt.queue_type = + convert_to_mes_queue_type(input->queue_type); + mes_add_queue_pkt.paging = input->paging; + mes_add_queue_pkt.vm_context_cntl = vm_cntx_cntl; + mes_add_queue_pkt.gws_base = input->gws_base; + mes_add_queue_pkt.gws_size = input->gws_size; + mes_add_queue_pkt.trap_handler_addr = input->tba_addr; + mes_add_queue_pkt.tma_addr = input->tma_addr; + mes_add_queue_pkt.is_kfd_process = input->is_kfd_process; + mes_add_queue_pkt.trap_en = 1; + + /* For KFD, gds_size is re-used for queue size (needed in MES for AQL queues) */ + mes_add_queue_pkt.is_aql_queue = input->is_aql_queue; + mes_add_queue_pkt.gds_size = input->queue_size; + + if (!(((adev->mes.sched_version & AMDGPU_MES_VERSION_MASK) >= 4) && + (adev->ip_versions[GC_HWIP][0] >= IP_VERSION(11, 0, 0)) && + (adev->ip_versions[GC_HWIP][0] <= IP_VERSION(11, 0, 3)))) + mes_add_queue_pkt.trap_en = 1; + + /* For KFD, gds_size is re-used for queue size (needed in MES for AQL queues) */ + mes_add_queue_pkt.is_aql_queue = input->is_aql_queue; + mes_add_queue_pkt.gds_size = input->queue_size; + + return mes_v11_0_submit_pkt_and_poll_completion(mes, + &mes_add_queue_pkt, sizeof(mes_add_queue_pkt), + offsetof(union MESAPI__ADD_QUEUE, api_status)); +} + +static int mes_v11_0_remove_hw_queue(struct amdgpu_mes *mes, + struct mes_remove_queue_input *input) +{ + union MESAPI__REMOVE_QUEUE mes_remove_queue_pkt; + + memset(&mes_remove_queue_pkt, 0, sizeof(mes_remove_queue_pkt)); + + mes_remove_queue_pkt.header.type = MES_API_TYPE_SCHEDULER; + mes_remove_queue_pkt.header.opcode = MES_SCH_API_REMOVE_QUEUE; + mes_remove_queue_pkt.header.dwsize = API_FRAME_SIZE_IN_DWORDS; + + mes_remove_queue_pkt.doorbell_offset = input->doorbell_offset; + mes_remove_queue_pkt.gang_context_addr = input->gang_context_addr; + + return mes_v11_0_submit_pkt_and_poll_completion(mes, + &mes_remove_queue_pkt, sizeof(mes_remove_queue_pkt), + offsetof(union MESAPI__REMOVE_QUEUE, api_status)); +} + +static int mes_v11_0_unmap_legacy_queue(struct amdgpu_mes *mes, + struct mes_unmap_legacy_queue_input *input) +{ + union MESAPI__REMOVE_QUEUE mes_remove_queue_pkt; + + memset(&mes_remove_queue_pkt, 0, sizeof(mes_remove_queue_pkt)); + + mes_remove_queue_pkt.header.type = MES_API_TYPE_SCHEDULER; + mes_remove_queue_pkt.header.opcode = MES_SCH_API_REMOVE_QUEUE; + mes_remove_queue_pkt.header.dwsize = API_FRAME_SIZE_IN_DWORDS; + + mes_remove_queue_pkt.doorbell_offset = input->doorbell_offset; + mes_remove_queue_pkt.gang_context_addr = 0; + + mes_remove_queue_pkt.pipe_id = input->pipe_id; + mes_remove_queue_pkt.queue_id = input->queue_id; + + if (input->action == PREEMPT_QUEUES_NO_UNMAP) { + mes_remove_queue_pkt.preempt_legacy_gfx_queue = 1; + mes_remove_queue_pkt.tf_addr = input->trail_fence_addr; + mes_remove_queue_pkt.tf_data = + lower_32_bits(input->trail_fence_data); + } else { + mes_remove_queue_pkt.unmap_legacy_queue = 1; + mes_remove_queue_pkt.queue_type = + convert_to_mes_queue_type(input->queue_type); + } + + return mes_v11_0_submit_pkt_and_poll_completion(mes, + &mes_remove_queue_pkt, sizeof(mes_remove_queue_pkt), + offsetof(union MESAPI__REMOVE_QUEUE, api_status)); +} + +static int mes_v11_0_suspend_gang(struct amdgpu_mes *mes, + struct mes_suspend_gang_input *input) +{ + return 0; +} + +static int mes_v11_0_resume_gang(struct amdgpu_mes *mes, + struct mes_resume_gang_input *input) +{ + return 0; +} + +static int mes_v11_0_query_sched_status(struct amdgpu_mes *mes) +{ + union MESAPI__QUERY_MES_STATUS mes_status_pkt; + + memset(&mes_status_pkt, 0, sizeof(mes_status_pkt)); + + mes_status_pkt.header.type = MES_API_TYPE_SCHEDULER; + mes_status_pkt.header.opcode = MES_SCH_API_QUERY_SCHEDULER_STATUS; + mes_status_pkt.header.dwsize = API_FRAME_SIZE_IN_DWORDS; + + return mes_v11_0_submit_pkt_and_poll_completion(mes, + &mes_status_pkt, sizeof(mes_status_pkt), + offsetof(union MESAPI__QUERY_MES_STATUS, api_status)); +} + +static int mes_v11_0_misc_op(struct amdgpu_mes *mes, + struct mes_misc_op_input *input) +{ + union MESAPI__MISC misc_pkt; + + memset(&misc_pkt, 0, sizeof(misc_pkt)); + + misc_pkt.header.type = MES_API_TYPE_SCHEDULER; + misc_pkt.header.opcode = MES_SCH_API_MISC; + misc_pkt.header.dwsize = API_FRAME_SIZE_IN_DWORDS; + + switch (input->op) { + case MES_MISC_OP_READ_REG: + misc_pkt.opcode = MESAPI_MISC__READ_REG; + misc_pkt.read_reg.reg_offset = input->read_reg.reg_offset; + misc_pkt.read_reg.buffer_addr = input->read_reg.buffer_addr; + break; + case MES_MISC_OP_WRITE_REG: + misc_pkt.opcode = MESAPI_MISC__WRITE_REG; + misc_pkt.write_reg.reg_offset = input->write_reg.reg_offset; + misc_pkt.write_reg.reg_value = input->write_reg.reg_value; + break; + case MES_MISC_OP_WRM_REG_WAIT: + misc_pkt.opcode = MESAPI_MISC__WAIT_REG_MEM; + misc_pkt.wait_reg_mem.op = WRM_OPERATION__WAIT_REG_MEM; + misc_pkt.wait_reg_mem.reference = input->wrm_reg.ref; + misc_pkt.wait_reg_mem.mask = input->wrm_reg.mask; + misc_pkt.wait_reg_mem.reg_offset1 = input->wrm_reg.reg0; + misc_pkt.wait_reg_mem.reg_offset2 = 0; + break; + case MES_MISC_OP_WRM_REG_WR_WAIT: + misc_pkt.opcode = MESAPI_MISC__WAIT_REG_MEM; + misc_pkt.wait_reg_mem.op = WRM_OPERATION__WR_WAIT_WR_REG; + misc_pkt.wait_reg_mem.reference = input->wrm_reg.ref; + misc_pkt.wait_reg_mem.mask = input->wrm_reg.mask; + misc_pkt.wait_reg_mem.reg_offset1 = input->wrm_reg.reg0; + misc_pkt.wait_reg_mem.reg_offset2 = input->wrm_reg.reg1; + break; + default: + DRM_ERROR("unsupported misc op (%d) \n", input->op); + return -EINVAL; + } + + return mes_v11_0_submit_pkt_and_poll_completion(mes, + &misc_pkt, sizeof(misc_pkt), + offsetof(union MESAPI__MISC, api_status)); +} + +static int mes_v11_0_set_hw_resources(struct amdgpu_mes *mes) +{ + int i; + struct amdgpu_device *adev = mes->adev; + union MESAPI_SET_HW_RESOURCES mes_set_hw_res_pkt; + + memset(&mes_set_hw_res_pkt, 0, sizeof(mes_set_hw_res_pkt)); + + mes_set_hw_res_pkt.header.type = MES_API_TYPE_SCHEDULER; + mes_set_hw_res_pkt.header.opcode = MES_SCH_API_SET_HW_RSRC; + mes_set_hw_res_pkt.header.dwsize = API_FRAME_SIZE_IN_DWORDS; + + mes_set_hw_res_pkt.vmid_mask_mmhub = mes->vmid_mask_mmhub; + mes_set_hw_res_pkt.vmid_mask_gfxhub = mes->vmid_mask_gfxhub; + mes_set_hw_res_pkt.gds_size = adev->gds.gds_size; + mes_set_hw_res_pkt.paging_vmid = 0; + mes_set_hw_res_pkt.g_sch_ctx_gpu_mc_ptr = mes->sch_ctx_gpu_addr; + mes_set_hw_res_pkt.query_status_fence_gpu_mc_ptr = + mes->query_status_fence_gpu_addr; + + for (i = 0; i < MAX_COMPUTE_PIPES; i++) + mes_set_hw_res_pkt.compute_hqd_mask[i] = + mes->compute_hqd_mask[i]; + + for (i = 0; i < MAX_GFX_PIPES; i++) + mes_set_hw_res_pkt.gfx_hqd_mask[i] = mes->gfx_hqd_mask[i]; + + for (i = 0; i < MAX_SDMA_PIPES; i++) + mes_set_hw_res_pkt.sdma_hqd_mask[i] = mes->sdma_hqd_mask[i]; + + for (i = 0; i < AMD_PRIORITY_NUM_LEVELS; i++) + mes_set_hw_res_pkt.aggregated_doorbells[i] = + mes->aggregated_doorbells[i]; + + for (i = 0; i < 5; i++) { + mes_set_hw_res_pkt.gc_base[i] = adev->reg_offset[GC_HWIP][0][i]; + mes_set_hw_res_pkt.mmhub_base[i] = + adev->reg_offset[MMHUB_HWIP][0][i]; + mes_set_hw_res_pkt.osssys_base[i] = + adev->reg_offset[OSSSYS_HWIP][0][i]; + } + + mes_set_hw_res_pkt.disable_reset = 1; + mes_set_hw_res_pkt.disable_mes_log = 1; + mes_set_hw_res_pkt.use_different_vmid_compute = 1; + mes_set_hw_res_pkt.oversubscription_timer = 50; + + return mes_v11_0_submit_pkt_and_poll_completion(mes, + &mes_set_hw_res_pkt, sizeof(mes_set_hw_res_pkt), + offsetof(union MESAPI_SET_HW_RESOURCES, api_status)); +} + +static void mes_v11_0_init_aggregated_doorbell(struct amdgpu_mes *mes) +{ + struct amdgpu_device *adev = mes->adev; + uint32_t data; + + data = RREG32_SOC15(GC, 0, regCP_MES_DOORBELL_CONTROL1); + data &= ~(CP_MES_DOORBELL_CONTROL1__DOORBELL_OFFSET_MASK | + CP_MES_DOORBELL_CONTROL1__DOORBELL_EN_MASK | + CP_MES_DOORBELL_CONTROL1__DOORBELL_HIT_MASK); + data |= mes->aggregated_doorbells[AMDGPU_MES_PRIORITY_LEVEL_LOW] << + CP_MES_DOORBELL_CONTROL1__DOORBELL_OFFSET__SHIFT; + data |= 1 << CP_MES_DOORBELL_CONTROL1__DOORBELL_EN__SHIFT; + WREG32_SOC15(GC, 0, regCP_MES_DOORBELL_CONTROL1, data); + + data = RREG32_SOC15(GC, 0, regCP_MES_DOORBELL_CONTROL2); + data &= ~(CP_MES_DOORBELL_CONTROL2__DOORBELL_OFFSET_MASK | + CP_MES_DOORBELL_CONTROL2__DOORBELL_EN_MASK | + CP_MES_DOORBELL_CONTROL2__DOORBELL_HIT_MASK); + data |= mes->aggregated_doorbells[AMDGPU_MES_PRIORITY_LEVEL_NORMAL] << + CP_MES_DOORBELL_CONTROL2__DOORBELL_OFFSET__SHIFT; + data |= 1 << CP_MES_DOORBELL_CONTROL2__DOORBELL_EN__SHIFT; + WREG32_SOC15(GC, 0, regCP_MES_DOORBELL_CONTROL2, data); + + data = RREG32_SOC15(GC, 0, regCP_MES_DOORBELL_CONTROL3); + data &= ~(CP_MES_DOORBELL_CONTROL3__DOORBELL_OFFSET_MASK | + CP_MES_DOORBELL_CONTROL3__DOORBELL_EN_MASK | + CP_MES_DOORBELL_CONTROL3__DOORBELL_HIT_MASK); + data |= mes->aggregated_doorbells[AMDGPU_MES_PRIORITY_LEVEL_MEDIUM] << + CP_MES_DOORBELL_CONTROL3__DOORBELL_OFFSET__SHIFT; + data |= 1 << CP_MES_DOORBELL_CONTROL3__DOORBELL_EN__SHIFT; + WREG32_SOC15(GC, 0, regCP_MES_DOORBELL_CONTROL3, data); + + data = RREG32_SOC15(GC, 0, regCP_MES_DOORBELL_CONTROL4); + data &= ~(CP_MES_DOORBELL_CONTROL4__DOORBELL_OFFSET_MASK | + CP_MES_DOORBELL_CONTROL4__DOORBELL_EN_MASK | + CP_MES_DOORBELL_CONTROL4__DOORBELL_HIT_MASK); + data |= mes->aggregated_doorbells[AMDGPU_MES_PRIORITY_LEVEL_HIGH] << + CP_MES_DOORBELL_CONTROL4__DOORBELL_OFFSET__SHIFT; + data |= 1 << CP_MES_DOORBELL_CONTROL4__DOORBELL_EN__SHIFT; + WREG32_SOC15(GC, 0, regCP_MES_DOORBELL_CONTROL4, data); + + data = RREG32_SOC15(GC, 0, regCP_MES_DOORBELL_CONTROL5); + data &= ~(CP_MES_DOORBELL_CONTROL5__DOORBELL_OFFSET_MASK | + CP_MES_DOORBELL_CONTROL5__DOORBELL_EN_MASK | + CP_MES_DOORBELL_CONTROL5__DOORBELL_HIT_MASK); + data |= mes->aggregated_doorbells[AMDGPU_MES_PRIORITY_LEVEL_REALTIME] << + CP_MES_DOORBELL_CONTROL5__DOORBELL_OFFSET__SHIFT; + data |= 1 << CP_MES_DOORBELL_CONTROL5__DOORBELL_EN__SHIFT; + WREG32_SOC15(GC, 0, regCP_MES_DOORBELL_CONTROL5, data); + + data = 1 << CP_HQD_GFX_CONTROL__DB_UPDATED_MSG_EN__SHIFT; + WREG32_SOC15(GC, 0, regCP_HQD_GFX_CONTROL, data); +} + +static const struct amdgpu_mes_funcs mes_v11_0_funcs = { + .add_hw_queue = mes_v11_0_add_hw_queue, + .remove_hw_queue = mes_v11_0_remove_hw_queue, + .unmap_legacy_queue = mes_v11_0_unmap_legacy_queue, + .suspend_gang = mes_v11_0_suspend_gang, + .resume_gang = mes_v11_0_resume_gang, + .misc_op = mes_v11_0_misc_op, +}; + +static int mes_v11_0_init_microcode(struct amdgpu_device *adev, + enum admgpu_mes_pipe pipe) +{ + char fw_name[30]; + char ucode_prefix[30]; + int err; + const struct mes_firmware_header_v1_0 *mes_hdr; + struct amdgpu_firmware_info *info; + + amdgpu_ucode_ip_version_decode(adev, GC_HWIP, ucode_prefix, sizeof(ucode_prefix)); + + if (pipe == AMDGPU_MES_SCHED_PIPE) + snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_mes.bin", + ucode_prefix); + else + snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_mes1.bin", + ucode_prefix); + + err = request_firmware(&adev->mes.fw[pipe], fw_name, adev->dev); + if (err) + return err; + + err = amdgpu_ucode_validate(adev->mes.fw[pipe]); + if (err) { + release_firmware(adev->mes.fw[pipe]); + adev->mes.fw[pipe] = NULL; + return err; + } + + mes_hdr = (const struct mes_firmware_header_v1_0 *) + adev->mes.fw[pipe]->data; + adev->mes.ucode_fw_version[pipe] = + le32_to_cpu(mes_hdr->mes_ucode_version); + adev->mes.ucode_fw_version[pipe] = + le32_to_cpu(mes_hdr->mes_ucode_data_version); + adev->mes.uc_start_addr[pipe] = + le32_to_cpu(mes_hdr->mes_uc_start_addr_lo) | + ((uint64_t)(le32_to_cpu(mes_hdr->mes_uc_start_addr_hi)) << 32); + adev->mes.data_start_addr[pipe] = + le32_to_cpu(mes_hdr->mes_data_start_addr_lo) | + ((uint64_t)(le32_to_cpu(mes_hdr->mes_data_start_addr_hi)) << 32); + + if (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) { + int ucode, ucode_data; + + if (pipe == AMDGPU_MES_SCHED_PIPE) { + ucode = AMDGPU_UCODE_ID_CP_MES; + ucode_data = AMDGPU_UCODE_ID_CP_MES_DATA; + } else { + ucode = AMDGPU_UCODE_ID_CP_MES1; + ucode_data = AMDGPU_UCODE_ID_CP_MES1_DATA; + } + + info = &adev->firmware.ucode[ucode]; + info->ucode_id = ucode; + info->fw = adev->mes.fw[pipe]; + adev->firmware.fw_size += + ALIGN(le32_to_cpu(mes_hdr->mes_ucode_size_bytes), + PAGE_SIZE); + + info = &adev->firmware.ucode[ucode_data]; + info->ucode_id = ucode_data; + info->fw = adev->mes.fw[pipe]; + adev->firmware.fw_size += + ALIGN(le32_to_cpu(mes_hdr->mes_ucode_data_size_bytes), + PAGE_SIZE); + } + + return 0; +} + +static void mes_v11_0_free_microcode(struct amdgpu_device *adev, + enum admgpu_mes_pipe pipe) +{ + release_firmware(adev->mes.fw[pipe]); + adev->mes.fw[pipe] = NULL; +} + +static int mes_v11_0_allocate_ucode_buffer(struct amdgpu_device *adev, + enum admgpu_mes_pipe pipe) +{ + int r; + const struct mes_firmware_header_v1_0 *mes_hdr; + const __le32 *fw_data; + unsigned fw_size; + + mes_hdr = (const struct mes_firmware_header_v1_0 *) + adev->mes.fw[pipe]->data; + + fw_data = (const __le32 *)(adev->mes.fw[pipe]->data + + le32_to_cpu(mes_hdr->mes_ucode_offset_bytes)); + fw_size = le32_to_cpu(mes_hdr->mes_ucode_size_bytes); + + r = amdgpu_bo_create_reserved(adev, fw_size, + PAGE_SIZE, AMDGPU_GEM_DOMAIN_VRAM, + &adev->mes.ucode_fw_obj[pipe], + &adev->mes.ucode_fw_gpu_addr[pipe], + (void **)&adev->mes.ucode_fw_ptr[pipe]); + if (r) { + dev_err(adev->dev, "(%d) failed to create mes fw bo\n", r); + return r; + } + + memcpy(adev->mes.ucode_fw_ptr[pipe], fw_data, fw_size); + + amdgpu_bo_kunmap(adev->mes.ucode_fw_obj[pipe]); + amdgpu_bo_unreserve(adev->mes.ucode_fw_obj[pipe]); + + return 0; +} + +static int mes_v11_0_allocate_ucode_data_buffer(struct amdgpu_device *adev, + enum admgpu_mes_pipe pipe) +{ + int r; + const struct mes_firmware_header_v1_0 *mes_hdr; + const __le32 *fw_data; + unsigned fw_size; + + mes_hdr = (const struct mes_firmware_header_v1_0 *) + adev->mes.fw[pipe]->data; + + fw_data = (const __le32 *)(adev->mes.fw[pipe]->data + + le32_to_cpu(mes_hdr->mes_ucode_data_offset_bytes)); + fw_size = le32_to_cpu(mes_hdr->mes_ucode_data_size_bytes); + + r = amdgpu_bo_create_reserved(adev, fw_size, + 64 * 1024, AMDGPU_GEM_DOMAIN_VRAM, + &adev->mes.data_fw_obj[pipe], + &adev->mes.data_fw_gpu_addr[pipe], + (void **)&adev->mes.data_fw_ptr[pipe]); + if (r) { + dev_err(adev->dev, "(%d) failed to create mes data fw bo\n", r); + return r; + } + + memcpy(adev->mes.data_fw_ptr[pipe], fw_data, fw_size); + + amdgpu_bo_kunmap(adev->mes.data_fw_obj[pipe]); + amdgpu_bo_unreserve(adev->mes.data_fw_obj[pipe]); + + return 0; +} + +static void mes_v11_0_free_ucode_buffers(struct amdgpu_device *adev, + enum admgpu_mes_pipe pipe) +{ + amdgpu_bo_free_kernel(&adev->mes.data_fw_obj[pipe], + &adev->mes.data_fw_gpu_addr[pipe], + (void **)&adev->mes.data_fw_ptr[pipe]); + + amdgpu_bo_free_kernel(&adev->mes.ucode_fw_obj[pipe], + &adev->mes.ucode_fw_gpu_addr[pipe], + (void **)&adev->mes.ucode_fw_ptr[pipe]); +} + +static void mes_v11_0_enable(struct amdgpu_device *adev, bool enable) +{ + uint64_t ucode_addr; + uint32_t pipe, data = 0; + + if (enable) { + data = RREG32_SOC15(GC, 0, regCP_MES_CNTL); + data = REG_SET_FIELD(data, CP_MES_CNTL, MES_PIPE0_RESET, 1); + data = REG_SET_FIELD(data, CP_MES_CNTL, + MES_PIPE1_RESET, adev->enable_mes_kiq ? 1 : 0); + WREG32_SOC15(GC, 0, regCP_MES_CNTL, data); + + mutex_lock(&adev->srbm_mutex); + for (pipe = 0; pipe < AMDGPU_MAX_MES_PIPES; pipe++) { + if (!adev->enable_mes_kiq && + pipe == AMDGPU_MES_KIQ_PIPE) + continue; + + soc21_grbm_select(adev, 3, pipe, 0, 0); + + ucode_addr = adev->mes.uc_start_addr[pipe] >> 2; + WREG32_SOC15(GC, 0, regCP_MES_PRGRM_CNTR_START, + lower_32_bits(ucode_addr)); + WREG32_SOC15(GC, 0, regCP_MES_PRGRM_CNTR_START_HI, + upper_32_bits(ucode_addr)); + } + soc21_grbm_select(adev, 0, 0, 0, 0); + mutex_unlock(&adev->srbm_mutex); + + /* unhalt MES and activate pipe0 */ + data = REG_SET_FIELD(0, CP_MES_CNTL, MES_PIPE0_ACTIVE, 1); + data = REG_SET_FIELD(data, CP_MES_CNTL, MES_PIPE1_ACTIVE, + adev->enable_mes_kiq ? 1 : 0); + WREG32_SOC15(GC, 0, regCP_MES_CNTL, data); + + if (amdgpu_emu_mode) + msleep(100); + else + udelay(50); + } else { + data = RREG32_SOC15(GC, 0, regCP_MES_CNTL); + data = REG_SET_FIELD(data, CP_MES_CNTL, MES_PIPE0_ACTIVE, 0); + data = REG_SET_FIELD(data, CP_MES_CNTL, MES_PIPE1_ACTIVE, 0); + data = REG_SET_FIELD(data, CP_MES_CNTL, + MES_INVALIDATE_ICACHE, 1); + data = REG_SET_FIELD(data, CP_MES_CNTL, MES_PIPE0_RESET, 1); + data = REG_SET_FIELD(data, CP_MES_CNTL, MES_PIPE1_RESET, + adev->enable_mes_kiq ? 1 : 0); + data = REG_SET_FIELD(data, CP_MES_CNTL, MES_HALT, 1); + WREG32_SOC15(GC, 0, regCP_MES_CNTL, data); + } +} + +/* This function is for backdoor MES firmware */ +static int mes_v11_0_load_microcode(struct amdgpu_device *adev, + enum admgpu_mes_pipe pipe, bool prime_icache) +{ + int r; + uint32_t data; + uint64_t ucode_addr; + + mes_v11_0_enable(adev, false); + + if (!adev->mes.fw[pipe]) + return -EINVAL; + + r = mes_v11_0_allocate_ucode_buffer(adev, pipe); + if (r) + return r; + + r = mes_v11_0_allocate_ucode_data_buffer(adev, pipe); + if (r) { + mes_v11_0_free_ucode_buffers(adev, pipe); + return r; + } + + mutex_lock(&adev->srbm_mutex); + /* me=3, pipe=0, queue=0 */ + soc21_grbm_select(adev, 3, pipe, 0, 0); + + WREG32_SOC15(GC, 0, regCP_MES_IC_BASE_CNTL, 0); + + /* set ucode start address */ + ucode_addr = adev->mes.uc_start_addr[pipe] >> 2; + WREG32_SOC15(GC, 0, regCP_MES_PRGRM_CNTR_START, + lower_32_bits(ucode_addr)); + WREG32_SOC15(GC, 0, regCP_MES_PRGRM_CNTR_START_HI, + upper_32_bits(ucode_addr)); + + /* set ucode fimrware address */ + WREG32_SOC15(GC, 0, regCP_MES_IC_BASE_LO, + lower_32_bits(adev->mes.ucode_fw_gpu_addr[pipe])); + WREG32_SOC15(GC, 0, regCP_MES_IC_BASE_HI, + upper_32_bits(adev->mes.ucode_fw_gpu_addr[pipe])); + + /* set ucode instruction cache boundary to 2M-1 */ + WREG32_SOC15(GC, 0, regCP_MES_MIBOUND_LO, 0x1FFFFF); + + /* set ucode data firmware address */ + WREG32_SOC15(GC, 0, regCP_MES_MDBASE_LO, + lower_32_bits(adev->mes.data_fw_gpu_addr[pipe])); + WREG32_SOC15(GC, 0, regCP_MES_MDBASE_HI, + upper_32_bits(adev->mes.data_fw_gpu_addr[pipe])); + + /* Set 0x3FFFF (256K-1) to CP_MES_MDBOUND_LO */ + WREG32_SOC15(GC, 0, regCP_MES_MDBOUND_LO, 0x3FFFF); + + if (prime_icache) { + /* invalidate ICACHE */ + data = RREG32_SOC15(GC, 0, regCP_MES_IC_OP_CNTL); + data = REG_SET_FIELD(data, CP_MES_IC_OP_CNTL, PRIME_ICACHE, 0); + data = REG_SET_FIELD(data, CP_MES_IC_OP_CNTL, INVALIDATE_CACHE, 1); + WREG32_SOC15(GC, 0, regCP_MES_IC_OP_CNTL, data); + + /* prime the ICACHE. */ + data = RREG32_SOC15(GC, 0, regCP_MES_IC_OP_CNTL); + data = REG_SET_FIELD(data, CP_MES_IC_OP_CNTL, PRIME_ICACHE, 1); + WREG32_SOC15(GC, 0, regCP_MES_IC_OP_CNTL, data); + } + + soc21_grbm_select(adev, 0, 0, 0, 0); + mutex_unlock(&adev->srbm_mutex); + + return 0; +} + +static int mes_v11_0_allocate_eop_buf(struct amdgpu_device *adev, + enum admgpu_mes_pipe pipe) +{ + int r; + u32 *eop; + + r = amdgpu_bo_create_reserved(adev, MES_EOP_SIZE, PAGE_SIZE, + AMDGPU_GEM_DOMAIN_GTT, + &adev->mes.eop_gpu_obj[pipe], + &adev->mes.eop_gpu_addr[pipe], + (void **)&eop); + if (r) { + dev_warn(adev->dev, "(%d) create EOP bo failed\n", r); + return r; + } + + memset(eop, 0, + adev->mes.eop_gpu_obj[pipe]->tbo.base.size); + + amdgpu_bo_kunmap(adev->mes.eop_gpu_obj[pipe]); + amdgpu_bo_unreserve(adev->mes.eop_gpu_obj[pipe]); + + return 0; +} + +static int mes_v11_0_mqd_init(struct amdgpu_ring *ring) +{ + struct v11_compute_mqd *mqd = ring->mqd_ptr; + uint64_t hqd_gpu_addr, wb_gpu_addr, eop_base_addr; + uint32_t tmp; + + mqd->header = 0xC0310800; + mqd->compute_pipelinestat_enable = 0x00000001; + mqd->compute_static_thread_mgmt_se0 = 0xffffffff; + mqd->compute_static_thread_mgmt_se1 = 0xffffffff; + mqd->compute_static_thread_mgmt_se2 = 0xffffffff; + mqd->compute_static_thread_mgmt_se3 = 0xffffffff; + mqd->compute_misc_reserved = 0x00000007; + + eop_base_addr = ring->eop_gpu_addr >> 8; + + /* set the EOP size, register value is 2^(EOP_SIZE+1) dwords */ + tmp = regCP_HQD_EOP_CONTROL_DEFAULT; + tmp = REG_SET_FIELD(tmp, CP_HQD_EOP_CONTROL, EOP_SIZE, + (order_base_2(MES_EOP_SIZE / 4) - 1)); + + mqd->cp_hqd_eop_base_addr_lo = lower_32_bits(eop_base_addr); + mqd->cp_hqd_eop_base_addr_hi = upper_32_bits(eop_base_addr); + mqd->cp_hqd_eop_control = tmp; + + /* disable the queue if it's active */ + ring->wptr = 0; + mqd->cp_hqd_pq_rptr = 0; + mqd->cp_hqd_pq_wptr_lo = 0; + mqd->cp_hqd_pq_wptr_hi = 0; + + /* set the pointer to the MQD */ + mqd->cp_mqd_base_addr_lo = ring->mqd_gpu_addr & 0xfffffffc; + mqd->cp_mqd_base_addr_hi = upper_32_bits(ring->mqd_gpu_addr); + + /* set MQD vmid to 0 */ + tmp = regCP_MQD_CONTROL_DEFAULT; + tmp = REG_SET_FIELD(tmp, CP_MQD_CONTROL, VMID, 0); + mqd->cp_mqd_control = tmp; + + /* set the pointer to the HQD, this is similar CP_RB0_BASE/_HI */ + hqd_gpu_addr = ring->gpu_addr >> 8; + mqd->cp_hqd_pq_base_lo = lower_32_bits(hqd_gpu_addr); + mqd->cp_hqd_pq_base_hi = upper_32_bits(hqd_gpu_addr); + + /* set the wb address whether it's enabled or not */ + wb_gpu_addr = ring->rptr_gpu_addr; + mqd->cp_hqd_pq_rptr_report_addr_lo = wb_gpu_addr & 0xfffffffc; + mqd->cp_hqd_pq_rptr_report_addr_hi = + upper_32_bits(wb_gpu_addr) & 0xffff; + + /* only used if CP_PQ_WPTR_POLL_CNTL.CP_PQ_WPTR_POLL_CNTL__EN_MASK=1 */ + wb_gpu_addr = ring->wptr_gpu_addr; + mqd->cp_hqd_pq_wptr_poll_addr_lo = wb_gpu_addr & 0xfffffff8; + mqd->cp_hqd_pq_wptr_poll_addr_hi = upper_32_bits(wb_gpu_addr) & 0xffff; + + /* set up the HQD, this is similar to CP_RB0_CNTL */ + tmp = regCP_HQD_PQ_CONTROL_DEFAULT; + tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, QUEUE_SIZE, + (order_base_2(ring->ring_size / 4) - 1)); + tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, RPTR_BLOCK_SIZE, + ((order_base_2(AMDGPU_GPU_PAGE_SIZE / 4) - 1) << 8)); + tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, UNORD_DISPATCH, 1); + tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, TUNNEL_DISPATCH, 0); + tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, PRIV_STATE, 1); + tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, KMD_QUEUE, 1); + tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, NO_UPDATE_RPTR, 1); + mqd->cp_hqd_pq_control = tmp; + + /* enable doorbell */ + tmp = 0; + if (ring->use_doorbell) { + tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_DOORBELL_CONTROL, + DOORBELL_OFFSET, ring->doorbell_index); + tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_DOORBELL_CONTROL, + DOORBELL_EN, 1); + tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_DOORBELL_CONTROL, + DOORBELL_SOURCE, 0); + tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_DOORBELL_CONTROL, + DOORBELL_HIT, 0); + } + else + tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_DOORBELL_CONTROL, + DOORBELL_EN, 0); + mqd->cp_hqd_pq_doorbell_control = tmp; + + mqd->cp_hqd_vmid = 0; + /* activate the queue */ + mqd->cp_hqd_active = 1; + + tmp = regCP_HQD_PERSISTENT_STATE_DEFAULT; + tmp = REG_SET_FIELD(tmp, CP_HQD_PERSISTENT_STATE, + PRELOAD_SIZE, 0x55); + mqd->cp_hqd_persistent_state = tmp; + + mqd->cp_hqd_ib_control = regCP_HQD_IB_CONTROL_DEFAULT; + mqd->cp_hqd_iq_timer = regCP_HQD_IQ_TIMER_DEFAULT; + mqd->cp_hqd_quantum = regCP_HQD_QUANTUM_DEFAULT; + + return 0; +} + +static void mes_v11_0_queue_init_register(struct amdgpu_ring *ring) +{ + struct v11_compute_mqd *mqd = ring->mqd_ptr; + struct amdgpu_device *adev = ring->adev; + uint32_t data = 0; + + mutex_lock(&adev->srbm_mutex); + soc21_grbm_select(adev, 3, ring->pipe, 0, 0); + + /* set CP_HQD_VMID.VMID = 0. */ + data = RREG32_SOC15(GC, 0, regCP_HQD_VMID); + data = REG_SET_FIELD(data, CP_HQD_VMID, VMID, 0); + WREG32_SOC15(GC, 0, regCP_HQD_VMID, data); + + /* set CP_HQD_PQ_DOORBELL_CONTROL.DOORBELL_EN=0 */ + data = RREG32_SOC15(GC, 0, regCP_HQD_PQ_DOORBELL_CONTROL); + data = REG_SET_FIELD(data, CP_HQD_PQ_DOORBELL_CONTROL, + DOORBELL_EN, 0); + WREG32_SOC15(GC, 0, regCP_HQD_PQ_DOORBELL_CONTROL, data); + + /* set CP_MQD_BASE_ADDR/HI with the MQD base address */ + WREG32_SOC15(GC, 0, regCP_MQD_BASE_ADDR, mqd->cp_mqd_base_addr_lo); + WREG32_SOC15(GC, 0, regCP_MQD_BASE_ADDR_HI, mqd->cp_mqd_base_addr_hi); + + /* set CP_MQD_CONTROL.VMID=0 */ + data = RREG32_SOC15(GC, 0, regCP_MQD_CONTROL); + data = REG_SET_FIELD(data, CP_MQD_CONTROL, VMID, 0); + WREG32_SOC15(GC, 0, regCP_MQD_CONTROL, 0); + + /* set CP_HQD_PQ_BASE/HI with the ring buffer base address */ + WREG32_SOC15(GC, 0, regCP_HQD_PQ_BASE, mqd->cp_hqd_pq_base_lo); + WREG32_SOC15(GC, 0, regCP_HQD_PQ_BASE_HI, mqd->cp_hqd_pq_base_hi); + + /* set CP_HQD_PQ_RPTR_REPORT_ADDR/HI */ + WREG32_SOC15(GC, 0, regCP_HQD_PQ_RPTR_REPORT_ADDR, + mqd->cp_hqd_pq_rptr_report_addr_lo); + WREG32_SOC15(GC, 0, regCP_HQD_PQ_RPTR_REPORT_ADDR_HI, + mqd->cp_hqd_pq_rptr_report_addr_hi); + + /* set CP_HQD_PQ_CONTROL */ + WREG32_SOC15(GC, 0, regCP_HQD_PQ_CONTROL, mqd->cp_hqd_pq_control); + + /* set CP_HQD_PQ_WPTR_POLL_ADDR/HI */ + WREG32_SOC15(GC, 0, regCP_HQD_PQ_WPTR_POLL_ADDR, + mqd->cp_hqd_pq_wptr_poll_addr_lo); + WREG32_SOC15(GC, 0, regCP_HQD_PQ_WPTR_POLL_ADDR_HI, + mqd->cp_hqd_pq_wptr_poll_addr_hi); + + /* set CP_HQD_PQ_DOORBELL_CONTROL */ + WREG32_SOC15(GC, 0, regCP_HQD_PQ_DOORBELL_CONTROL, + mqd->cp_hqd_pq_doorbell_control); + + /* set CP_HQD_PERSISTENT_STATE.PRELOAD_SIZE=0x53 */ + WREG32_SOC15(GC, 0, regCP_HQD_PERSISTENT_STATE, mqd->cp_hqd_persistent_state); + + /* set CP_HQD_ACTIVE.ACTIVE=1 */ + WREG32_SOC15(GC, 0, regCP_HQD_ACTIVE, mqd->cp_hqd_active); + + soc21_grbm_select(adev, 0, 0, 0, 0); + mutex_unlock(&adev->srbm_mutex); +} + +static int mes_v11_0_kiq_enable_queue(struct amdgpu_device *adev) +{ + struct amdgpu_kiq *kiq = &adev->gfx.kiq; + struct amdgpu_ring *kiq_ring = &adev->gfx.kiq.ring; + int r; + + if (!kiq->pmf || !kiq->pmf->kiq_map_queues) + return -EINVAL; + + r = amdgpu_ring_alloc(kiq_ring, kiq->pmf->map_queues_size); + if (r) { + DRM_ERROR("Failed to lock KIQ (%d).\n", r); + return r; + } + + kiq->pmf->kiq_map_queues(kiq_ring, &adev->mes.ring); + + r = amdgpu_ring_test_ring(kiq_ring); + if (r) { + DRM_ERROR("kfq enable failed\n"); + kiq_ring->sched.ready = false; + } + return r; +} + +static int mes_v11_0_queue_init(struct amdgpu_device *adev, + enum admgpu_mes_pipe pipe) +{ + struct amdgpu_ring *ring; + int r; + + if (pipe == AMDGPU_MES_KIQ_PIPE) + ring = &adev->gfx.kiq.ring; + else if (pipe == AMDGPU_MES_SCHED_PIPE) + ring = &adev->mes.ring; + else + BUG(); + + if ((pipe == AMDGPU_MES_SCHED_PIPE) && + (amdgpu_in_reset(adev) || adev->in_suspend)) { + *(ring->wptr_cpu_addr) = 0; + *(ring->rptr_cpu_addr) = 0; + amdgpu_ring_clear_ring(ring); + } + + r = mes_v11_0_mqd_init(ring); + if (r) + return r; + + if (pipe == AMDGPU_MES_SCHED_PIPE) { + r = mes_v11_0_kiq_enable_queue(adev); + if (r) + return r; + } else { + mes_v11_0_queue_init_register(ring); + } + + /* get MES scheduler/KIQ versions */ + mutex_lock(&adev->srbm_mutex); + soc21_grbm_select(adev, 3, pipe, 0, 0); + + if (pipe == AMDGPU_MES_SCHED_PIPE) + adev->mes.sched_version = RREG32_SOC15(GC, 0, regCP_MES_GP3_LO); + else if (pipe == AMDGPU_MES_KIQ_PIPE && adev->enable_mes_kiq) + adev->mes.kiq_version = RREG32_SOC15(GC, 0, regCP_MES_GP3_LO); + + soc21_grbm_select(adev, 0, 0, 0, 0); + mutex_unlock(&adev->srbm_mutex); + + return 0; +} + +static int mes_v11_0_ring_init(struct amdgpu_device *adev) +{ + struct amdgpu_ring *ring; + + ring = &adev->mes.ring; + + ring->funcs = &mes_v11_0_ring_funcs; + + ring->me = 3; + ring->pipe = 0; + ring->queue = 0; + + ring->ring_obj = NULL; + ring->use_doorbell = true; + ring->doorbell_index = adev->doorbell_index.mes_ring0 << 1; + ring->eop_gpu_addr = adev->mes.eop_gpu_addr[AMDGPU_MES_SCHED_PIPE]; + ring->no_scheduler = true; + sprintf(ring->name, "mes_%d.%d.%d", ring->me, ring->pipe, ring->queue); + + return amdgpu_ring_init(adev, ring, 1024, NULL, 0, + AMDGPU_RING_PRIO_DEFAULT, NULL); +} + +static int mes_v11_0_kiq_ring_init(struct amdgpu_device *adev) +{ + struct amdgpu_ring *ring; + + spin_lock_init(&adev->gfx.kiq.ring_lock); + + ring = &adev->gfx.kiq.ring; + + ring->me = 3; + ring->pipe = 1; + ring->queue = 0; + + ring->adev = NULL; + ring->ring_obj = NULL; + ring->use_doorbell = true; + ring->doorbell_index = adev->doorbell_index.mes_ring1 << 1; + ring->eop_gpu_addr = adev->mes.eop_gpu_addr[AMDGPU_MES_KIQ_PIPE]; + ring->no_scheduler = true; + sprintf(ring->name, "mes_kiq_%d.%d.%d", + ring->me, ring->pipe, ring->queue); + + return amdgpu_ring_init(adev, ring, 1024, NULL, 0, + AMDGPU_RING_PRIO_DEFAULT, NULL); +} + +static int mes_v11_0_mqd_sw_init(struct amdgpu_device *adev, + enum admgpu_mes_pipe pipe) +{ + int r, mqd_size = sizeof(struct v11_compute_mqd); + struct amdgpu_ring *ring; + + if (pipe == AMDGPU_MES_KIQ_PIPE) + ring = &adev->gfx.kiq.ring; + else if (pipe == AMDGPU_MES_SCHED_PIPE) + ring = &adev->mes.ring; + else + BUG(); + + if (ring->mqd_obj) + return 0; + + r = amdgpu_bo_create_kernel(adev, mqd_size, PAGE_SIZE, + AMDGPU_GEM_DOMAIN_GTT, &ring->mqd_obj, + &ring->mqd_gpu_addr, &ring->mqd_ptr); + if (r) { + dev_warn(adev->dev, "failed to create ring mqd bo (%d)", r); + return r; + } + + memset(ring->mqd_ptr, 0, mqd_size); + + /* prepare MQD backup */ + adev->mes.mqd_backup[pipe] = kmalloc(mqd_size, GFP_KERNEL); + if (!adev->mes.mqd_backup[pipe]) + dev_warn(adev->dev, + "no memory to create MQD backup for ring %s\n", + ring->name); + + return 0; +} + +static int mes_v11_0_sw_init(void *handle) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + int pipe, r; + + adev->mes.adev = adev; + adev->mes.funcs = &mes_v11_0_funcs; + adev->mes.kiq_hw_init = &mes_v11_0_kiq_hw_init; + adev->mes.kiq_hw_fini = &mes_v11_0_kiq_hw_fini; + + r = amdgpu_mes_init(adev); + if (r) + return r; + + for (pipe = 0; pipe < AMDGPU_MAX_MES_PIPES; pipe++) { + if (!adev->enable_mes_kiq && pipe == AMDGPU_MES_KIQ_PIPE) + continue; + + r = mes_v11_0_init_microcode(adev, pipe); + if (r) + return r; + + r = mes_v11_0_allocate_eop_buf(adev, pipe); + if (r) + return r; + + r = mes_v11_0_mqd_sw_init(adev, pipe); + if (r) + return r; + } + + if (adev->enable_mes_kiq) { + r = mes_v11_0_kiq_ring_init(adev); + if (r) + return r; + } + + r = mes_v11_0_ring_init(adev); + if (r) + return r; + + return 0; +} + +static int mes_v11_0_sw_fini(void *handle) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + int pipe; + + amdgpu_device_wb_free(adev, adev->mes.sch_ctx_offs); + amdgpu_device_wb_free(adev, adev->mes.query_status_fence_offs); + + for (pipe = 0; pipe < AMDGPU_MAX_MES_PIPES; pipe++) { + kfree(adev->mes.mqd_backup[pipe]); + + amdgpu_bo_free_kernel(&adev->mes.eop_gpu_obj[pipe], + &adev->mes.eop_gpu_addr[pipe], + NULL); + + mes_v11_0_free_microcode(adev, pipe); + } + + amdgpu_bo_free_kernel(&adev->gfx.kiq.ring.mqd_obj, + &adev->gfx.kiq.ring.mqd_gpu_addr, + &adev->gfx.kiq.ring.mqd_ptr); + + amdgpu_bo_free_kernel(&adev->mes.ring.mqd_obj, + &adev->mes.ring.mqd_gpu_addr, + &adev->mes.ring.mqd_ptr); + + amdgpu_ring_fini(&adev->gfx.kiq.ring); + amdgpu_ring_fini(&adev->mes.ring); + + if (adev->firmware.load_type == AMDGPU_FW_LOAD_DIRECT) { + mes_v11_0_free_ucode_buffers(adev, AMDGPU_MES_KIQ_PIPE); + mes_v11_0_free_ucode_buffers(adev, AMDGPU_MES_SCHED_PIPE); + } + + amdgpu_mes_fini(adev); + return 0; +} + +static void mes_v11_0_kiq_dequeue_sched(struct amdgpu_device *adev) +{ + uint32_t data; + int i; + + mutex_lock(&adev->srbm_mutex); + soc21_grbm_select(adev, 3, AMDGPU_MES_SCHED_PIPE, 0, 0); + + /* disable the queue if it's active */ + if (RREG32_SOC15(GC, 0, regCP_HQD_ACTIVE) & 1) { + WREG32_SOC15(GC, 0, regCP_HQD_DEQUEUE_REQUEST, 1); + for (i = 0; i < adev->usec_timeout; i++) { + if (!(RREG32_SOC15(GC, 0, regCP_HQD_ACTIVE) & 1)) + break; + udelay(1); + } + } + data = RREG32_SOC15(GC, 0, regCP_HQD_PQ_DOORBELL_CONTROL); + data = REG_SET_FIELD(data, CP_HQD_PQ_DOORBELL_CONTROL, + DOORBELL_EN, 0); + data = REG_SET_FIELD(data, CP_HQD_PQ_DOORBELL_CONTROL, + DOORBELL_HIT, 1); + WREG32_SOC15(GC, 0, regCP_HQD_PQ_DOORBELL_CONTROL, data); + + WREG32_SOC15(GC, 0, regCP_HQD_PQ_DOORBELL_CONTROL, 0); + + WREG32_SOC15(GC, 0, regCP_HQD_PQ_WPTR_LO, 0); + WREG32_SOC15(GC, 0, regCP_HQD_PQ_WPTR_HI, 0); + WREG32_SOC15(GC, 0, regCP_HQD_PQ_RPTR, 0); + + soc21_grbm_select(adev, 0, 0, 0, 0); + mutex_unlock(&adev->srbm_mutex); + + adev->mes.ring.sched.ready = false; +} + +static void mes_v11_0_kiq_setting(struct amdgpu_ring *ring) +{ + uint32_t tmp; + struct amdgpu_device *adev = ring->adev; + + /* tell RLC which is KIQ queue */ + tmp = RREG32_SOC15(GC, 0, regRLC_CP_SCHEDULERS); + tmp &= 0xffffff00; + tmp |= (ring->me << 5) | (ring->pipe << 3) | (ring->queue); + WREG32_SOC15(GC, 0, regRLC_CP_SCHEDULERS, tmp); + tmp |= 0x80; + WREG32_SOC15(GC, 0, regRLC_CP_SCHEDULERS, tmp); +} + +static int mes_v11_0_kiq_hw_init(struct amdgpu_device *adev) +{ + int r = 0; + + if (adev->firmware.load_type == AMDGPU_FW_LOAD_DIRECT) { + + r = mes_v11_0_load_microcode(adev, AMDGPU_MES_SCHED_PIPE, false); + if (r) { + DRM_ERROR("failed to load MES fw, r=%d\n", r); + return r; + } + + r = mes_v11_0_load_microcode(adev, AMDGPU_MES_KIQ_PIPE, true); + if (r) { + DRM_ERROR("failed to load MES kiq fw, r=%d\n", r); + return r; + } + + } + + mes_v11_0_enable(adev, true); + + mes_v11_0_kiq_setting(&adev->gfx.kiq.ring); + + r = mes_v11_0_queue_init(adev, AMDGPU_MES_KIQ_PIPE); + if (r) + goto failure; + + return r; + +failure: + mes_v11_0_hw_fini(adev); + return r; +} + +static int mes_v11_0_kiq_hw_fini(struct amdgpu_device *adev) +{ + if (adev->mes.ring.sched.ready) + mes_v11_0_kiq_dequeue_sched(adev); + + mes_v11_0_enable(adev, false); + return 0; +} + +static int mes_v11_0_hw_init(void *handle) +{ + int r; + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + + if (!adev->enable_mes_kiq) { + if (adev->firmware.load_type == AMDGPU_FW_LOAD_DIRECT) { + r = mes_v11_0_load_microcode(adev, + AMDGPU_MES_SCHED_PIPE, true); + if (r) { + DRM_ERROR("failed to MES fw, r=%d\n", r); + return r; + } + } + + mes_v11_0_enable(adev, true); + } + + r = mes_v11_0_queue_init(adev, AMDGPU_MES_SCHED_PIPE); + if (r) + goto failure; + + r = mes_v11_0_set_hw_resources(&adev->mes); + if (r) + goto failure; + + mes_v11_0_init_aggregated_doorbell(&adev->mes); + + r = mes_v11_0_query_sched_status(&adev->mes); + if (r) { + DRM_ERROR("MES is busy\n"); + goto failure; + } + + /* + * Disable KIQ ring usage from the driver once MES is enabled. + * MES uses KIQ ring exclusively so driver cannot access KIQ ring + * with MES enabled. + */ + adev->gfx.kiq.ring.sched.ready = false; + adev->mes.ring.sched.ready = true; + + return 0; + +failure: + mes_v11_0_hw_fini(adev); + return r; +} + +static int mes_v11_0_hw_fini(void *handle) +{ + return 0; +} + +static int mes_v11_0_suspend(void *handle) +{ + int r; + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + + r = amdgpu_mes_suspend(adev); + if (r) + return r; + + return mes_v11_0_hw_fini(adev); +} + +static int mes_v11_0_resume(void *handle) +{ + int r; + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + + r = mes_v11_0_hw_init(adev); + if (r) + return r; + + return amdgpu_mes_resume(adev); +} + +static int mes_v11_0_late_init(void *handle) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + + if (!amdgpu_in_reset(adev) && + (adev->ip_versions[GC_HWIP][0] != IP_VERSION(11, 0, 3))) + amdgpu_mes_self_test(adev); + + return 0; +} + +static const struct amd_ip_funcs mes_v11_0_ip_funcs = { + .name = "mes_v11_0", + .late_init = mes_v11_0_late_init, + .sw_init = mes_v11_0_sw_init, + .sw_fini = mes_v11_0_sw_fini, + .hw_init = mes_v11_0_hw_init, + .hw_fini = mes_v11_0_hw_fini, + .suspend = mes_v11_0_suspend, + .resume = mes_v11_0_resume, +}; + +const struct amdgpu_ip_block_version mes_v11_0_ip_block = { + .type = AMD_IP_BLOCK_TYPE_MES, + .major = 11, + .minor = 0, + .rev = 0, + .funcs = &mes_v11_0_ip_funcs, +}; diff --git a/drivers/gpu/drm/amd/amdgpu/mes_v11_0.h b/drivers/gpu/drm/amd/amdgpu/mes_v11_0.h new file mode 100644 index 000000000000..b3519e1df2b2 --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/mes_v11_0.h @@ -0,0 +1,29 @@ +/* + * Copyright 2021 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#ifndef __MES_V11_0_H__ +#define __MES_V11_0_H__ + +extern const struct amdgpu_ip_block_version mes_v11_0_ip_block; + +#endif diff --git a/drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.c b/drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.c index 1da2ec692057..3e51e773f92b 100644 --- a/drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.c +++ b/drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.c @@ -176,6 +176,7 @@ static void mmhub_v1_0_init_cache_regs(struct amdgpu_device *adev) tmp = REG_SET_FIELD(tmp, VM_L2_CNTL2, INVALIDATE_L2_CACHE, 1); WREG32_SOC15(MMHUB, 0, mmVM_L2_CNTL2, tmp); + tmp = mmVM_L2_CNTL3_DEFAULT; if (adev->gmc.translate_further) { tmp = REG_SET_FIELD(tmp, VM_L2_CNTL3, BANK_SELECT, 12); tmp = REG_SET_FIELD(tmp, VM_L2_CNTL3, @@ -546,7 +547,7 @@ static int mmhub_v1_0_set_clockgating(struct amdgpu_device *adev, return 0; } -static void mmhub_v1_0_get_clockgating(struct amdgpu_device *adev, u32 *flags) +static void mmhub_v1_0_get_clockgating(struct amdgpu_device *adev, u64 *flags) { int data, data1; @@ -774,13 +775,17 @@ static void mmhub_v1_0_reset_ras_error_count(struct amdgpu_device *adev) } } -const struct amdgpu_mmhub_ras_funcs mmhub_v1_0_ras_funcs = { - .ras_late_init = amdgpu_mmhub_ras_late_init, - .ras_fini = amdgpu_mmhub_ras_fini, +struct amdgpu_ras_block_hw_ops mmhub_v1_0_ras_hw_ops = { .query_ras_error_count = mmhub_v1_0_query_ras_error_count, .reset_ras_error_count = mmhub_v1_0_reset_ras_error_count, }; +struct amdgpu_mmhub_ras mmhub_v1_0_ras = { + .ras_block = { + .hw_ops = &mmhub_v1_0_ras_hw_ops, + }, +}; + const struct amdgpu_mmhub_funcs mmhub_v1_0_funcs = { .get_fb_location = mmhub_v1_0_get_fb_location, .init = mmhub_v1_0_init, diff --git a/drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.h b/drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.h index 4661b094e007..dae7ca48bd8b 100644 --- a/drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.h +++ b/drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.h @@ -24,6 +24,6 @@ #define __MMHUB_V1_0_H__ extern const struct amdgpu_mmhub_funcs mmhub_v1_0_funcs; -extern const struct amdgpu_mmhub_ras_funcs mmhub_v1_0_ras_funcs; +extern struct amdgpu_mmhub_ras mmhub_v1_0_ras; #endif diff --git a/drivers/gpu/drm/amd/amdgpu/mmhub_v1_7.c b/drivers/gpu/drm/amd/amdgpu/mmhub_v1_7.c index f5f7181f9af5..6fa7090bc6cb 100644 --- a/drivers/gpu/drm/amd/amdgpu/mmhub_v1_7.c +++ b/drivers/gpu/drm/amd/amdgpu/mmhub_v1_7.c @@ -542,7 +542,7 @@ static int mmhub_v1_7_set_clockgating(struct amdgpu_device *adev, return 0; } -static void mmhub_v1_7_get_clockgating(struct amdgpu_device *adev, u32 *flags) +static void mmhub_v1_7_get_clockgating(struct amdgpu_device *adev, u64 *flags) { int data, data1; @@ -1321,15 +1321,19 @@ static void mmhub_v1_7_reset_ras_error_status(struct amdgpu_device *adev) } } -const struct amdgpu_mmhub_ras_funcs mmhub_v1_7_ras_funcs = { - .ras_late_init = amdgpu_mmhub_ras_late_init, - .ras_fini = amdgpu_mmhub_ras_fini, +struct amdgpu_ras_block_hw_ops mmhub_v1_7_ras_hw_ops = { .query_ras_error_count = mmhub_v1_7_query_ras_error_count, .reset_ras_error_count = mmhub_v1_7_reset_ras_error_count, .query_ras_error_status = mmhub_v1_7_query_ras_error_status, .reset_ras_error_status = mmhub_v1_7_reset_ras_error_status, }; +struct amdgpu_mmhub_ras mmhub_v1_7_ras = { + .ras_block = { + .hw_ops = &mmhub_v1_7_ras_hw_ops, + }, +}; + const struct amdgpu_mmhub_funcs mmhub_v1_7_funcs = { .get_fb_location = mmhub_v1_7_get_fb_location, .init = mmhub_v1_7_init, diff --git a/drivers/gpu/drm/amd/amdgpu/mmhub_v1_7.h b/drivers/gpu/drm/amd/amdgpu/mmhub_v1_7.h index a7f9dfc24697..629f49052137 100644 --- a/drivers/gpu/drm/amd/amdgpu/mmhub_v1_7.h +++ b/drivers/gpu/drm/amd/amdgpu/mmhub_v1_7.h @@ -24,6 +24,6 @@ #define __MMHUB_V1_7_H__ extern const struct amdgpu_mmhub_funcs mmhub_v1_7_funcs; -extern const struct amdgpu_mmhub_ras_funcs mmhub_v1_7_ras_funcs; +extern struct amdgpu_mmhub_ras mmhub_v1_7_ras; #endif diff --git a/drivers/gpu/drm/amd/amdgpu/mmhub_v2_0.c b/drivers/gpu/drm/amd/amdgpu/mmhub_v2_0.c index 3718ff610ab2..998b5d17b271 100644 --- a/drivers/gpu/drm/amd/amdgpu/mmhub_v2_0.c +++ b/drivers/gpu/drm/amd/amdgpu/mmhub_v2_0.c @@ -32,8 +32,6 @@ #include "gc/gc_10_1_0_offset.h" #include "soc15_common.h" -#define mmMM_ATC_L2_MISC_CG_Sienna_Cichlid 0x064d -#define mmMM_ATC_L2_MISC_CG_Sienna_Cichlid_BASE_IDX 0 #define mmDAGB0_CNTL_MISC2_Sienna_Cichlid 0x0070 #define mmDAGB0_CNTL_MISC2_Sienna_Cichlid_BASE_IDX 0 @@ -408,6 +406,8 @@ static void mmhub_v2_0_setup_vmid_config(struct amdgpu_device *adev) i * hub->ctx_addr_distance, upper_32_bits(adev->vm_manager.max_pfn - 1)); } + + hub->vm_cntx_cntl = tmp; } static void mmhub_v2_0_program_invalidation(struct amdgpu_device *adev) @@ -572,7 +572,6 @@ static void mmhub_v2_0_update_medium_grain_clock_gating(struct amdgpu_device *ad case IP_VERSION(2, 1, 0): case IP_VERSION(2, 1, 1): case IP_VERSION(2, 1, 2): - def = data = RREG32_SOC15(MMHUB, 0, mmMM_ATC_L2_MISC_CG_Sienna_Cichlid); def1 = data1 = RREG32_SOC15(MMHUB, 0, mmDAGB0_CNTL_MISC2_Sienna_Cichlid); break; default: @@ -606,8 +605,6 @@ static void mmhub_v2_0_update_medium_grain_clock_gating(struct amdgpu_device *ad case IP_VERSION(2, 1, 0): case IP_VERSION(2, 1, 1): case IP_VERSION(2, 1, 2): - if (def != data) - WREG32_SOC15(MMHUB, 0, mmMM_ATC_L2_MISC_CG_Sienna_Cichlid, data); if (def1 != data1) WREG32_SOC15(MMHUB, 0, mmDAGB0_CNTL_MISC2_Sienna_Cichlid, data1); break; @@ -632,8 +629,8 @@ static void mmhub_v2_0_update_medium_grain_light_sleep(struct amdgpu_device *ade case IP_VERSION(2, 1, 0): case IP_VERSION(2, 1, 1): case IP_VERSION(2, 1, 2): - def = data = RREG32_SOC15(MMHUB, 0, mmMM_ATC_L2_MISC_CG_Sienna_Cichlid); - break; + /* There is no ATCL2 in MMHUB for 2.1.x */ + return; default: def = data = RREG32_SOC15(MMHUB, 0, mmMM_ATC_L2_MISC_CG); break; @@ -644,18 +641,8 @@ static void mmhub_v2_0_update_medium_grain_light_sleep(struct amdgpu_device *ade else data &= ~MM_ATC_L2_MISC_CG__MEM_LS_ENABLE_MASK; - if (def != data) { - switch (adev->ip_versions[MMHUB_HWIP][0]) { - case IP_VERSION(2, 1, 0): - case IP_VERSION(2, 1, 1): - case IP_VERSION(2, 1, 2): - WREG32_SOC15(MMHUB, 0, mmMM_ATC_L2_MISC_CG_Sienna_Cichlid, data); - break; - default: - WREG32_SOC15(MMHUB, 0, mmMM_ATC_L2_MISC_CG, data); - break; - } - } + if (def != data) + WREG32_SOC15(MMHUB, 0, mmMM_ATC_L2_MISC_CG, data); } static int mmhub_v2_0_set_clockgating(struct amdgpu_device *adev, @@ -682,7 +669,7 @@ static int mmhub_v2_0_set_clockgating(struct amdgpu_device *adev, return 0; } -static void mmhub_v2_0_get_clockgating(struct amdgpu_device *adev, u32 *flags) +static void mmhub_v2_0_get_clockgating(struct amdgpu_device *adev, u64 *flags) { int data, data1; @@ -693,7 +680,10 @@ static void mmhub_v2_0_get_clockgating(struct amdgpu_device *adev, u32 *flags) case IP_VERSION(2, 1, 0): case IP_VERSION(2, 1, 1): case IP_VERSION(2, 1, 2): - data = RREG32_SOC15(MMHUB, 0, mmMM_ATC_L2_MISC_CG_Sienna_Cichlid); + /* There is no ATCL2 in MMHUB for 2.1.x. Keep the status + * based on DAGB + */ + data = MM_ATC_L2_MISC_CG__ENABLE_MASK; data1 = RREG32_SOC15(MMHUB, 0, mmDAGB0_CNTL_MISC2_Sienna_Cichlid); break; default: diff --git a/drivers/gpu/drm/amd/amdgpu/mmhub_v2_3.c b/drivers/gpu/drm/amd/amdgpu/mmhub_v2_3.c index 9e16da28505a..1b027d069ab4 100644 --- a/drivers/gpu/drm/amd/amdgpu/mmhub_v2_3.c +++ b/drivers/gpu/drm/amd/amdgpu/mmhub_v2_3.c @@ -93,6 +93,7 @@ mmhub_v2_3_print_l2_protection_fault_status(struct amdgpu_device *adev, switch (adev->ip_versions[MMHUB_HWIP][0]) { case IP_VERSION(2, 3, 0): case IP_VERSION(2, 4, 0): + case IP_VERSION(2, 4, 1): mmhub_cid = mmhub_client_ids_vangogh[cid][rw]; break; default: @@ -323,6 +324,8 @@ static void mmhub_v2_3_setup_vmid_config(struct amdgpu_device *adev) i * hub->ctx_addr_distance, upper_32_bits(adev->vm_manager.max_pfn - 1)); } + + hub->vm_cntx_cntl = tmp; } static void mmhub_v2_3_program_invalidation(struct amdgpu_device *adev) @@ -576,7 +579,7 @@ static int mmhub_v2_3_set_clockgating(struct amdgpu_device *adev, return 0; } -static void mmhub_v2_3_get_clockgating(struct amdgpu_device *adev, u32 *flags) +static void mmhub_v2_3_get_clockgating(struct amdgpu_device *adev, u64 *flags) { int data, data1, data2, data3; diff --git a/drivers/gpu/drm/amd/amdgpu/mmhub_v3_0.c b/drivers/gpu/drm/amd/amdgpu/mmhub_v3_0.c new file mode 100644 index 000000000000..a1d26c4d80b8 --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/mmhub_v3_0.c @@ -0,0 +1,661 @@ +/* + * Copyright 2021 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#include "amdgpu.h" +#include "mmhub_v3_0.h" + +#include "mmhub/mmhub_3_0_0_offset.h" +#include "mmhub/mmhub_3_0_0_sh_mask.h" +#include "navi10_enum.h" + +#include "soc15_common.h" + +#define regMMVM_L2_CNTL3_DEFAULT 0x80100007 +#define regMMVM_L2_CNTL4_DEFAULT 0x000000c1 +#define regMMVM_L2_CNTL5_DEFAULT 0x00003fe0 + +static const char *mmhub_client_ids_v3_0_0[][2] = { + [0][0] = "VMC", + [4][0] = "DCEDMC", + [5][0] = "DCEVGA", + [6][0] = "MP0", + [7][0] = "MP1", + [8][0] = "MPIO", + [16][0] = "HDP", + [17][0] = "LSDMA", + [18][0] = "JPEG", + [19][0] = "VCNU0", + [21][0] = "VSCH", + [22][0] = "VCNU1", + [23][0] = "VCN1", + [32+20][0] = "VCN0", + [2][1] = "DBGUNBIO", + [3][1] = "DCEDWB", + [4][1] = "DCEDMC", + [5][1] = "DCEVGA", + [6][1] = "MP0", + [7][1] = "MP1", + [8][1] = "MPIO", + [10][1] = "DBGU0", + [11][1] = "DBGU1", + [12][1] = "DBGU2", + [13][1] = "DBGU3", + [14][1] = "XDP", + [15][1] = "OSSSYS", + [16][1] = "HDP", + [17][1] = "LSDMA", + [18][1] = "JPEG", + [19][1] = "VCNU0", + [20][1] = "VCN0", + [21][1] = "VSCH", + [22][1] = "VCNU1", + [23][1] = "VCN1", +}; + +static uint32_t mmhub_v3_0_get_invalidate_req(unsigned int vmid, + uint32_t flush_type) +{ + u32 req = 0; + + /* invalidate using legacy mode on vmid*/ + req = REG_SET_FIELD(req, MMVM_INVALIDATE_ENG0_REQ, + PER_VMID_INVALIDATE_REQ, 1 << vmid); + req = REG_SET_FIELD(req, MMVM_INVALIDATE_ENG0_REQ, FLUSH_TYPE, flush_type); + req = REG_SET_FIELD(req, MMVM_INVALIDATE_ENG0_REQ, INVALIDATE_L2_PTES, 1); + req = REG_SET_FIELD(req, MMVM_INVALIDATE_ENG0_REQ, INVALIDATE_L2_PDE0, 1); + req = REG_SET_FIELD(req, MMVM_INVALIDATE_ENG0_REQ, INVALIDATE_L2_PDE1, 1); + req = REG_SET_FIELD(req, MMVM_INVALIDATE_ENG0_REQ, INVALIDATE_L2_PDE2, 1); + req = REG_SET_FIELD(req, MMVM_INVALIDATE_ENG0_REQ, INVALIDATE_L1_PTES, 1); + req = REG_SET_FIELD(req, MMVM_INVALIDATE_ENG0_REQ, + CLEAR_PROTECTION_FAULT_STATUS_ADDR, 0); + + return req; +} + +static void +mmhub_v3_0_print_l2_protection_fault_status(struct amdgpu_device *adev, + uint32_t status) +{ + uint32_t cid, rw; + const char *mmhub_cid = NULL; + + cid = REG_GET_FIELD(status, + MMVM_L2_PROTECTION_FAULT_STATUS, CID); + rw = REG_GET_FIELD(status, + MMVM_L2_PROTECTION_FAULT_STATUS, RW); + + dev_err(adev->dev, + "MMVM_L2_PROTECTION_FAULT_STATUS:0x%08X\n", + status); + switch (adev->ip_versions[MMHUB_HWIP][0]) { + case IP_VERSION(3, 0, 0): + case IP_VERSION(3, 0, 1): + mmhub_cid = mmhub_client_ids_v3_0_0[cid][rw]; + break; + default: + mmhub_cid = NULL; + break; + } + dev_err(adev->dev, "\t Faulty UTCL2 client ID: %s (0x%x)\n", + mmhub_cid ? mmhub_cid : "unknown", cid); + dev_err(adev->dev, "\t MORE_FAULTS: 0x%lx\n", + REG_GET_FIELD(status, + MMVM_L2_PROTECTION_FAULT_STATUS, MORE_FAULTS)); + dev_err(adev->dev, "\t WALKER_ERROR: 0x%lx\n", + REG_GET_FIELD(status, + MMVM_L2_PROTECTION_FAULT_STATUS, WALKER_ERROR)); + dev_err(adev->dev, "\t PERMISSION_FAULTS: 0x%lx\n", + REG_GET_FIELD(status, + MMVM_L2_PROTECTION_FAULT_STATUS, PERMISSION_FAULTS)); + dev_err(adev->dev, "\t MAPPING_ERROR: 0x%lx\n", + REG_GET_FIELD(status, + MMVM_L2_PROTECTION_FAULT_STATUS, MAPPING_ERROR)); + dev_err(adev->dev, "\t RW: 0x%x\n", rw); +} + +static void mmhub_v3_0_setup_vm_pt_regs(struct amdgpu_device *adev, uint32_t vmid, + uint64_t page_table_base) +{ + struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0]; + + WREG32_SOC15_OFFSET(MMHUB, 0, regMMVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32, + hub->ctx_addr_distance * vmid, + lower_32_bits(page_table_base)); + + WREG32_SOC15_OFFSET(MMHUB, 0, regMMVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_HI32, + hub->ctx_addr_distance * vmid, + upper_32_bits(page_table_base)); +} + +static void mmhub_v3_0_init_gart_aperture_regs(struct amdgpu_device *adev) +{ + uint64_t pt_base = amdgpu_gmc_pd_addr(adev->gart.bo); + + mmhub_v3_0_setup_vm_pt_regs(adev, 0, pt_base); + + WREG32_SOC15(MMHUB, 0, regMMVM_CONTEXT0_PAGE_TABLE_START_ADDR_LO32, + (u32)(adev->gmc.gart_start >> 12)); + WREG32_SOC15(MMHUB, 0, regMMVM_CONTEXT0_PAGE_TABLE_START_ADDR_HI32, + (u32)(adev->gmc.gart_start >> 44)); + + WREG32_SOC15(MMHUB, 0, regMMVM_CONTEXT0_PAGE_TABLE_END_ADDR_LO32, + (u32)(adev->gmc.gart_end >> 12)); + WREG32_SOC15(MMHUB, 0, regMMVM_CONTEXT0_PAGE_TABLE_END_ADDR_HI32, + (u32)(adev->gmc.gart_end >> 44)); +} + +static void mmhub_v3_0_init_system_aperture_regs(struct amdgpu_device *adev) +{ + uint64_t value; + uint32_t tmp; + + if (!amdgpu_sriov_vf(adev)) { + /* + * the new L1 policy will block SRIOV guest from writing + * these regs, and they will be programed at host. + * so skip programing these regs. + */ + /* Disable AGP. */ + WREG32_SOC15(MMHUB, 0, regMMMC_VM_AGP_BASE, 0); + WREG32_SOC15(MMHUB, 0, regMMMC_VM_AGP_TOP, 0); + WREG32_SOC15(MMHUB, 0, regMMMC_VM_AGP_BOT, 0x00FFFFFF); + + /* Program the system aperture low logical page number. */ + WREG32_SOC15(MMHUB, 0, regMMMC_VM_SYSTEM_APERTURE_LOW_ADDR, + adev->gmc.vram_start >> 18); + WREG32_SOC15(MMHUB, 0, regMMMC_VM_SYSTEM_APERTURE_HIGH_ADDR, + adev->gmc.vram_end >> 18); + } + + /* Set default page address. */ + value = adev->vram_scratch.gpu_addr - adev->gmc.vram_start + + adev->vm_manager.vram_base_offset; + WREG32_SOC15(MMHUB, 0, regMMMC_VM_SYSTEM_APERTURE_DEFAULT_ADDR_LSB, + (u32)(value >> 12)); + WREG32_SOC15(MMHUB, 0, regMMMC_VM_SYSTEM_APERTURE_DEFAULT_ADDR_MSB, + (u32)(value >> 44)); + + /* Program "protection fault". */ + WREG32_SOC15(MMHUB, 0, regMMVM_L2_PROTECTION_FAULT_DEFAULT_ADDR_LO32, + (u32)(adev->dummy_page_addr >> 12)); + WREG32_SOC15(MMHUB, 0, regMMVM_L2_PROTECTION_FAULT_DEFAULT_ADDR_HI32, + (u32)((u64)adev->dummy_page_addr >> 44)); + + tmp = RREG32_SOC15(MMHUB, 0, regMMVM_L2_PROTECTION_FAULT_CNTL2); + tmp = REG_SET_FIELD(tmp, MMVM_L2_PROTECTION_FAULT_CNTL2, + ACTIVE_PAGE_MIGRATION_PTE_READ_RETRY, 1); + WREG32_SOC15(MMHUB, 0, regMMVM_L2_PROTECTION_FAULT_CNTL2, tmp); +} + +static void mmhub_v3_0_init_tlb_regs(struct amdgpu_device *adev) +{ + uint32_t tmp; + + /* Setup TLB control */ + tmp = RREG32_SOC15(MMHUB, 0, regMMMC_VM_MX_L1_TLB_CNTL); + + tmp = REG_SET_FIELD(tmp, MMMC_VM_MX_L1_TLB_CNTL, ENABLE_L1_TLB, 1); + tmp = REG_SET_FIELD(tmp, MMMC_VM_MX_L1_TLB_CNTL, SYSTEM_ACCESS_MODE, 3); + tmp = REG_SET_FIELD(tmp, MMMC_VM_MX_L1_TLB_CNTL, + ENABLE_ADVANCED_DRIVER_MODEL, 1); + tmp = REG_SET_FIELD(tmp, MMMC_VM_MX_L1_TLB_CNTL, + SYSTEM_APERTURE_UNMAPPED_ACCESS, 0); + tmp = REG_SET_FIELD(tmp, MMMC_VM_MX_L1_TLB_CNTL, ECO_BITS, 0); + tmp = REG_SET_FIELD(tmp, MMMC_VM_MX_L1_TLB_CNTL, + MTYPE, MTYPE_UC); /* UC, uncached */ + + WREG32_SOC15(MMHUB, 0, regMMMC_VM_MX_L1_TLB_CNTL, tmp); +} + +static void mmhub_v3_0_init_cache_regs(struct amdgpu_device *adev) +{ + uint32_t tmp; + + /* These registers are not accessible to VF-SRIOV. + * The PF will program them instead. + */ + if (amdgpu_sriov_vf(adev)) + return; + + /* Setup L2 cache */ + tmp = RREG32_SOC15(MMHUB, 0, regMMVM_L2_CNTL); + tmp = REG_SET_FIELD(tmp, MMVM_L2_CNTL, ENABLE_L2_CACHE, 1); + tmp = REG_SET_FIELD(tmp, MMVM_L2_CNTL, ENABLE_L2_FRAGMENT_PROCESSING, 0); + tmp = REG_SET_FIELD(tmp, MMVM_L2_CNTL, + ENABLE_DEFAULT_PAGE_OUT_TO_SYSTEM_MEMORY, 1); + /* XXX for emulation, Refer to closed source code.*/ + tmp = REG_SET_FIELD(tmp, MMVM_L2_CNTL, L2_PDE0_CACHE_TAG_GENERATION_MODE, + 0); + tmp = REG_SET_FIELD(tmp, MMVM_L2_CNTL, PDE_FAULT_CLASSIFICATION, 0); + tmp = REG_SET_FIELD(tmp, MMVM_L2_CNTL, CONTEXT1_IDENTITY_ACCESS_MODE, 1); + tmp = REG_SET_FIELD(tmp, MMVM_L2_CNTL, IDENTITY_MODE_FRAGMENT_SIZE, 0); + WREG32_SOC15(MMHUB, 0, regMMVM_L2_CNTL, tmp); + + tmp = RREG32_SOC15(MMHUB, 0, regMMVM_L2_CNTL2); + tmp = REG_SET_FIELD(tmp, MMVM_L2_CNTL2, INVALIDATE_ALL_L1_TLBS, 1); + tmp = REG_SET_FIELD(tmp, MMVM_L2_CNTL2, INVALIDATE_L2_CACHE, 1); + WREG32_SOC15(MMHUB, 0, regMMVM_L2_CNTL2, tmp); + + tmp = regMMVM_L2_CNTL3_DEFAULT; + if (adev->gmc.translate_further) { + tmp = REG_SET_FIELD(tmp, MMVM_L2_CNTL3, BANK_SELECT, 12); + tmp = REG_SET_FIELD(tmp, MMVM_L2_CNTL3, + L2_CACHE_BIGK_FRAGMENT_SIZE, 9); + } else { + tmp = REG_SET_FIELD(tmp, MMVM_L2_CNTL3, BANK_SELECT, 9); + tmp = REG_SET_FIELD(tmp, MMVM_L2_CNTL3, + L2_CACHE_BIGK_FRAGMENT_SIZE, 6); + } + WREG32_SOC15(MMHUB, 0, regMMVM_L2_CNTL3, tmp); + + tmp = regMMVM_L2_CNTL4_DEFAULT; + tmp = REG_SET_FIELD(tmp, MMVM_L2_CNTL4, VMC_TAP_PDE_REQUEST_PHYSICAL, 0); + tmp = REG_SET_FIELD(tmp, MMVM_L2_CNTL4, VMC_TAP_PTE_REQUEST_PHYSICAL, 0); + WREG32_SOC15(MMHUB, 0, regMMVM_L2_CNTL4, tmp); + + tmp = regMMVM_L2_CNTL5_DEFAULT; + tmp = REG_SET_FIELD(tmp, MMVM_L2_CNTL5, L2_CACHE_SMALLK_FRAGMENT_SIZE, 0); + WREG32_SOC15(GC, 0, regMMVM_L2_CNTL5, tmp); +} + +static void mmhub_v3_0_enable_system_domain(struct amdgpu_device *adev) +{ + uint32_t tmp; + + tmp = RREG32_SOC15(MMHUB, 0, regMMVM_CONTEXT0_CNTL); + tmp = REG_SET_FIELD(tmp, MMVM_CONTEXT0_CNTL, ENABLE_CONTEXT, 1); + tmp = REG_SET_FIELD(tmp, MMVM_CONTEXT0_CNTL, PAGE_TABLE_DEPTH, 0); + tmp = REG_SET_FIELD(tmp, MMVM_CONTEXT0_CNTL, + RETRY_PERMISSION_OR_INVALID_PAGE_FAULT, 0); + WREG32_SOC15(MMHUB, 0, regMMVM_CONTEXT0_CNTL, tmp); +} + +static void mmhub_v3_0_disable_identity_aperture(struct amdgpu_device *adev) +{ + /* These registers are not accessible to VF-SRIOV. + * The PF will program them instead. + */ + if (amdgpu_sriov_vf(adev)) + return; + + WREG32_SOC15(MMHUB, 0, + regMMVM_L2_CONTEXT1_IDENTITY_APERTURE_LOW_ADDR_LO32, + 0xFFFFFFFF); + WREG32_SOC15(MMHUB, 0, + regMMVM_L2_CONTEXT1_IDENTITY_APERTURE_LOW_ADDR_HI32, + 0x0000000F); + + WREG32_SOC15(MMHUB, 0, + regMMVM_L2_CONTEXT1_IDENTITY_APERTURE_HIGH_ADDR_LO32, 0); + WREG32_SOC15(MMHUB, 0, + regMMVM_L2_CONTEXT1_IDENTITY_APERTURE_HIGH_ADDR_HI32, 0); + + WREG32_SOC15(MMHUB, 0, regMMVM_L2_CONTEXT_IDENTITY_PHYSICAL_OFFSET_LO32, + 0); + WREG32_SOC15(MMHUB, 0, regMMVM_L2_CONTEXT_IDENTITY_PHYSICAL_OFFSET_HI32, + 0); +} + +static void mmhub_v3_0_setup_vmid_config(struct amdgpu_device *adev) +{ + struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0]; + int i; + uint32_t tmp; + + for (i = 0; i <= 14; i++) { + tmp = RREG32_SOC15_OFFSET(MMHUB, 0, regMMVM_CONTEXT1_CNTL, i); + tmp = REG_SET_FIELD(tmp, MMVM_CONTEXT1_CNTL, ENABLE_CONTEXT, 1); + tmp = REG_SET_FIELD(tmp, MMVM_CONTEXT1_CNTL, PAGE_TABLE_DEPTH, + adev->vm_manager.num_level); + tmp = REG_SET_FIELD(tmp, MMVM_CONTEXT1_CNTL, + RANGE_PROTECTION_FAULT_ENABLE_DEFAULT, 1); + tmp = REG_SET_FIELD(tmp, MMVM_CONTEXT1_CNTL, + DUMMY_PAGE_PROTECTION_FAULT_ENABLE_DEFAULT, + 1); + tmp = REG_SET_FIELD(tmp, MMVM_CONTEXT1_CNTL, + PDE0_PROTECTION_FAULT_ENABLE_DEFAULT, 1); + tmp = REG_SET_FIELD(tmp, MMVM_CONTEXT1_CNTL, + VALID_PROTECTION_FAULT_ENABLE_DEFAULT, 1); + tmp = REG_SET_FIELD(tmp, MMVM_CONTEXT1_CNTL, + READ_PROTECTION_FAULT_ENABLE_DEFAULT, 1); + tmp = REG_SET_FIELD(tmp, MMVM_CONTEXT1_CNTL, + WRITE_PROTECTION_FAULT_ENABLE_DEFAULT, 1); + tmp = REG_SET_FIELD(tmp, MMVM_CONTEXT1_CNTL, + EXECUTE_PROTECTION_FAULT_ENABLE_DEFAULT, 1); + tmp = REG_SET_FIELD(tmp, MMVM_CONTEXT1_CNTL, + PAGE_TABLE_BLOCK_SIZE, + adev->vm_manager.block_size - 9); + /* Send no-retry XNACK on fault to suppress VM fault storm. */ + tmp = REG_SET_FIELD(tmp, MMVM_CONTEXT1_CNTL, + RETRY_PERMISSION_OR_INVALID_PAGE_FAULT, + !amdgpu_noretry); + WREG32_SOC15_OFFSET(MMHUB, 0, regMMVM_CONTEXT1_CNTL, + i * hub->ctx_distance, tmp); + WREG32_SOC15_OFFSET(MMHUB, 0, regMMVM_CONTEXT1_PAGE_TABLE_START_ADDR_LO32, + i * hub->ctx_addr_distance, 0); + WREG32_SOC15_OFFSET(MMHUB, 0, regMMVM_CONTEXT1_PAGE_TABLE_START_ADDR_HI32, + i * hub->ctx_addr_distance, 0); + WREG32_SOC15_OFFSET(MMHUB, 0, regMMVM_CONTEXT1_PAGE_TABLE_END_ADDR_LO32, + i * hub->ctx_addr_distance, + lower_32_bits(adev->vm_manager.max_pfn - 1)); + WREG32_SOC15_OFFSET(MMHUB, 0, regMMVM_CONTEXT1_PAGE_TABLE_END_ADDR_HI32, + i * hub->ctx_addr_distance, + upper_32_bits(adev->vm_manager.max_pfn - 1)); + } + + hub->vm_cntx_cntl = tmp; +} + +static void mmhub_v3_0_program_invalidation(struct amdgpu_device *adev) +{ + struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0]; + unsigned i; + + for (i = 0; i < 18; ++i) { + WREG32_SOC15_OFFSET(MMHUB, 0, regMMVM_INVALIDATE_ENG0_ADDR_RANGE_LO32, + i * hub->eng_addr_distance, 0xffffffff); + WREG32_SOC15_OFFSET(MMHUB, 0, regMMVM_INVALIDATE_ENG0_ADDR_RANGE_HI32, + i * hub->eng_addr_distance, 0x1f); + } +} + +static int mmhub_v3_0_gart_enable(struct amdgpu_device *adev) +{ + /* GART Enable. */ + mmhub_v3_0_init_gart_aperture_regs(adev); + mmhub_v3_0_init_system_aperture_regs(adev); + mmhub_v3_0_init_tlb_regs(adev); + mmhub_v3_0_init_cache_regs(adev); + + mmhub_v3_0_enable_system_domain(adev); + mmhub_v3_0_disable_identity_aperture(adev); + mmhub_v3_0_setup_vmid_config(adev); + mmhub_v3_0_program_invalidation(adev); + + return 0; +} + +static void mmhub_v3_0_gart_disable(struct amdgpu_device *adev) +{ + struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0]; + u32 tmp; + u32 i; + + /* Disable all tables */ + for (i = 0; i < 16; i++) + WREG32_SOC15_OFFSET(MMHUB, 0, regMMVM_CONTEXT0_CNTL, + i * hub->ctx_distance, 0); + + /* Setup TLB control */ + tmp = RREG32_SOC15(MMHUB, 0, regMMMC_VM_MX_L1_TLB_CNTL); + tmp = REG_SET_FIELD(tmp, MMMC_VM_MX_L1_TLB_CNTL, ENABLE_L1_TLB, 0); + tmp = REG_SET_FIELD(tmp, MMMC_VM_MX_L1_TLB_CNTL, + ENABLE_ADVANCED_DRIVER_MODEL, 0); + WREG32_SOC15(MMHUB, 0, regMMMC_VM_MX_L1_TLB_CNTL, tmp); + + /* Setup L2 cache */ + tmp = RREG32_SOC15(MMHUB, 0, regMMVM_L2_CNTL); + tmp = REG_SET_FIELD(tmp, MMVM_L2_CNTL, ENABLE_L2_CACHE, 0); + WREG32_SOC15(MMHUB, 0, regMMVM_L2_CNTL, tmp); + WREG32_SOC15(MMHUB, 0, regMMVM_L2_CNTL3, 0); +} + +/** + * mmhub_v3_0_set_fault_enable_default - update GART/VM fault handling + * + * @adev: amdgpu_device pointer + * @value: true redirects VM faults to the default page + */ +static void mmhub_v3_0_set_fault_enable_default(struct amdgpu_device *adev, bool value) +{ + u32 tmp; + + /* These registers are not accessible to VF-SRIOV. + * The PF will program them instead. + */ + if (amdgpu_sriov_vf(adev)) + return; + + tmp = RREG32_SOC15(MMHUB, 0, regMMVM_L2_PROTECTION_FAULT_CNTL); + tmp = REG_SET_FIELD(tmp, MMVM_L2_PROTECTION_FAULT_CNTL, + RANGE_PROTECTION_FAULT_ENABLE_DEFAULT, value); + tmp = REG_SET_FIELD(tmp, MMVM_L2_PROTECTION_FAULT_CNTL, + PDE0_PROTECTION_FAULT_ENABLE_DEFAULT, value); + tmp = REG_SET_FIELD(tmp, MMVM_L2_PROTECTION_FAULT_CNTL, + PDE1_PROTECTION_FAULT_ENABLE_DEFAULT, value); + tmp = REG_SET_FIELD(tmp, MMVM_L2_PROTECTION_FAULT_CNTL, + PDE2_PROTECTION_FAULT_ENABLE_DEFAULT, value); + tmp = REG_SET_FIELD(tmp, MMVM_L2_PROTECTION_FAULT_CNTL, + TRANSLATE_FURTHER_PROTECTION_FAULT_ENABLE_DEFAULT, + value); + tmp = REG_SET_FIELD(tmp, MMVM_L2_PROTECTION_FAULT_CNTL, + NACK_PROTECTION_FAULT_ENABLE_DEFAULT, value); + tmp = REG_SET_FIELD(tmp, MMVM_L2_PROTECTION_FAULT_CNTL, + DUMMY_PAGE_PROTECTION_FAULT_ENABLE_DEFAULT, value); + tmp = REG_SET_FIELD(tmp, MMVM_L2_PROTECTION_FAULT_CNTL, + VALID_PROTECTION_FAULT_ENABLE_DEFAULT, value); + tmp = REG_SET_FIELD(tmp, MMVM_L2_PROTECTION_FAULT_CNTL, + READ_PROTECTION_FAULT_ENABLE_DEFAULT, value); + tmp = REG_SET_FIELD(tmp, MMVM_L2_PROTECTION_FAULT_CNTL, + WRITE_PROTECTION_FAULT_ENABLE_DEFAULT, value); + tmp = REG_SET_FIELD(tmp, MMVM_L2_PROTECTION_FAULT_CNTL, + EXECUTE_PROTECTION_FAULT_ENABLE_DEFAULT, value); + if (!value) { + tmp = REG_SET_FIELD(tmp, MMVM_L2_PROTECTION_FAULT_CNTL, + CRASH_ON_NO_RETRY_FAULT, 1); + tmp = REG_SET_FIELD(tmp, MMVM_L2_PROTECTION_FAULT_CNTL, + CRASH_ON_RETRY_FAULT, 1); + } + WREG32_SOC15(MMHUB, 0, regMMVM_L2_PROTECTION_FAULT_CNTL, tmp); +} + +static const struct amdgpu_vmhub_funcs mmhub_v3_0_vmhub_funcs = { + .print_l2_protection_fault_status = mmhub_v3_0_print_l2_protection_fault_status, + .get_invalidate_req = mmhub_v3_0_get_invalidate_req, +}; + +static void mmhub_v3_0_init(struct amdgpu_device *adev) +{ + struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0]; + + hub->ctx0_ptb_addr_lo32 = + SOC15_REG_OFFSET(MMHUB, 0, + regMMVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32); + hub->ctx0_ptb_addr_hi32 = + SOC15_REG_OFFSET(MMHUB, 0, + regMMVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_HI32); + hub->vm_inv_eng0_sem = + SOC15_REG_OFFSET(MMHUB, 0, regMMVM_INVALIDATE_ENG0_SEM); + hub->vm_inv_eng0_req = + SOC15_REG_OFFSET(MMHUB, 0, regMMVM_INVALIDATE_ENG0_REQ); + hub->vm_inv_eng0_ack = + SOC15_REG_OFFSET(MMHUB, 0, regMMVM_INVALIDATE_ENG0_ACK); + hub->vm_context0_cntl = + SOC15_REG_OFFSET(MMHUB, 0, regMMVM_CONTEXT0_CNTL); + hub->vm_l2_pro_fault_status = + SOC15_REG_OFFSET(MMHUB, 0, regMMVM_L2_PROTECTION_FAULT_STATUS); + hub->vm_l2_pro_fault_cntl = + SOC15_REG_OFFSET(MMHUB, 0, regMMVM_L2_PROTECTION_FAULT_CNTL); + + hub->ctx_distance = regMMVM_CONTEXT1_CNTL - regMMVM_CONTEXT0_CNTL; + hub->ctx_addr_distance = regMMVM_CONTEXT1_PAGE_TABLE_BASE_ADDR_LO32 - + regMMVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32; + hub->eng_distance = regMMVM_INVALIDATE_ENG1_REQ - + regMMVM_INVALIDATE_ENG0_REQ; + hub->eng_addr_distance = regMMVM_INVALIDATE_ENG1_ADDR_RANGE_LO32 - + regMMVM_INVALIDATE_ENG0_ADDR_RANGE_LO32; + + hub->vm_cntx_cntl_vm_fault = MMVM_CONTEXT1_CNTL__RANGE_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK | + MMVM_CONTEXT1_CNTL__DUMMY_PAGE_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK | + MMVM_CONTEXT1_CNTL__PDE0_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK | + MMVM_CONTEXT1_CNTL__VALID_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK | + MMVM_CONTEXT1_CNTL__READ_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK | + MMVM_CONTEXT1_CNTL__WRITE_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK | + MMVM_CONTEXT1_CNTL__EXECUTE_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK; + + hub->vm_l2_bank_select_reserved_cid2 = + SOC15_REG_OFFSET(MMHUB, 0, regMMVM_L2_BANK_SELECT_RESERVED_CID2); + + hub->vmhub_funcs = &mmhub_v3_0_vmhub_funcs; +} + +static u64 mmhub_v3_0_get_fb_location(struct amdgpu_device *adev) +{ + u64 base; + + base = RREG32_SOC15(MMHUB, 0, regMMMC_VM_FB_LOCATION_BASE); + + base &= MMMC_VM_FB_LOCATION_BASE__FB_BASE_MASK; + base <<= 24; + + return base; +} + +static u64 mmhub_v3_0_get_mc_fb_offset(struct amdgpu_device *adev) +{ + return (u64)RREG32_SOC15(MMHUB, 0, regMMMC_VM_FB_OFFSET) << 24; +} + +static void mmhub_v3_0_update_medium_grain_clock_gating(struct amdgpu_device *adev, + bool enable) +{ + uint32_t def, data; +#if 0 + uint32_t def1, data1, def2 = 0, data2 = 0; +#endif + + def = data = RREG32_SOC15(MMHUB, 0, regMM_ATC_L2_MISC_CG); +#if 0 + def1 = data1 = RREG32_SOC15(MMHUB, 0, regDAGB0_CNTL_MISC2); + def2 = data2 = RREG32_SOC15(MMHUB, 0, regDAGB1_CNTL_MISC2); +#endif + + if (enable) { + data |= MM_ATC_L2_MISC_CG__ENABLE_MASK; +#if 0 + data1 &= ~(DAGB0_CNTL_MISC2__DISABLE_WRREQ_CG_MASK | + DAGB0_CNTL_MISC2__DISABLE_WRRET_CG_MASK | + DAGB0_CNTL_MISC2__DISABLE_RDREQ_CG_MASK | + DAGB0_CNTL_MISC2__DISABLE_RDRET_CG_MASK | + DAGB0_CNTL_MISC2__DISABLE_TLBWR_CG_MASK | + DAGB0_CNTL_MISC2__DISABLE_TLBRD_CG_MASK); + + data2 &= ~(DAGB1_CNTL_MISC2__DISABLE_WRREQ_CG_MASK | + DAGB1_CNTL_MISC2__DISABLE_WRRET_CG_MASK | + DAGB1_CNTL_MISC2__DISABLE_RDREQ_CG_MASK | + DAGB1_CNTL_MISC2__DISABLE_RDRET_CG_MASK | + DAGB1_CNTL_MISC2__DISABLE_TLBWR_CG_MASK | + DAGB1_CNTL_MISC2__DISABLE_TLBRD_CG_MASK); +#endif + } else { + data &= ~MM_ATC_L2_MISC_CG__ENABLE_MASK; +#if 0 + data1 |= (DAGB0_CNTL_MISC2__DISABLE_WRREQ_CG_MASK | + DAGB0_CNTL_MISC2__DISABLE_WRRET_CG_MASK | + DAGB0_CNTL_MISC2__DISABLE_RDREQ_CG_MASK | + DAGB0_CNTL_MISC2__DISABLE_RDRET_CG_MASK | + DAGB0_CNTL_MISC2__DISABLE_TLBWR_CG_MASK | + DAGB0_CNTL_MISC2__DISABLE_TLBRD_CG_MASK); + + data2 |= (DAGB1_CNTL_MISC2__DISABLE_WRREQ_CG_MASK | + DAGB1_CNTL_MISC2__DISABLE_WRRET_CG_MASK | + DAGB1_CNTL_MISC2__DISABLE_RDREQ_CG_MASK | + DAGB1_CNTL_MISC2__DISABLE_RDRET_CG_MASK | + DAGB1_CNTL_MISC2__DISABLE_TLBWR_CG_MASK | + DAGB1_CNTL_MISC2__DISABLE_TLBRD_CG_MASK); +#endif + } + + if (def != data) + WREG32_SOC15(MMHUB, 0, regMM_ATC_L2_MISC_CG, data); +#if 0 + if (def1 != data1) + WREG32_SOC15(MMHUB, 0, regDAGB0_CNTL_MISC2, data1); + + if (def2 != data2) + WREG32_SOC15(MMHUB, 0, regDAGB1_CNTL_MISC2, data2); +#endif +} + +static void mmhub_v3_0_update_medium_grain_light_sleep(struct amdgpu_device *adev, + bool enable) +{ + uint32_t def, data; + + def = data = RREG32_SOC15(MMHUB, 0, regMM_ATC_L2_MISC_CG); + + if (enable) + data |= MM_ATC_L2_MISC_CG__MEM_LS_ENABLE_MASK; + else + data &= ~MM_ATC_L2_MISC_CG__MEM_LS_ENABLE_MASK; + + if (def != data) + WREG32_SOC15(MMHUB, 0, regMM_ATC_L2_MISC_CG, data); +} + +static int mmhub_v3_0_set_clockgating(struct amdgpu_device *adev, + enum amd_clockgating_state state) +{ + if (amdgpu_sriov_vf(adev)) + return 0; + + if (adev->cg_flags & AMD_CG_SUPPORT_MC_MGCG) + mmhub_v3_0_update_medium_grain_clock_gating(adev, + state == AMD_CG_STATE_GATE); + + if (adev->cg_flags & AMD_CG_SUPPORT_MC_LS) + mmhub_v3_0_update_medium_grain_light_sleep(adev, + state == AMD_CG_STATE_GATE); + + return 0; +} + +static void mmhub_v3_0_get_clockgating(struct amdgpu_device *adev, u64 *flags) +{ + int data; + + if (amdgpu_sriov_vf(adev)) + *flags = 0; + + data = RREG32_SOC15(MMHUB, 0, regMM_ATC_L2_MISC_CG); + + /* AMD_CG_SUPPORT_MC_MGCG */ + if (data & MM_ATC_L2_MISC_CG__ENABLE_MASK) + *flags |= AMD_CG_SUPPORT_MC_MGCG; + + /* AMD_CG_SUPPORT_MC_LS */ + if (data & MM_ATC_L2_MISC_CG__MEM_LS_ENABLE_MASK) + *flags |= AMD_CG_SUPPORT_MC_LS; +} + +const struct amdgpu_mmhub_funcs mmhub_v3_0_funcs = { + .init = mmhub_v3_0_init, + .get_fb_location = mmhub_v3_0_get_fb_location, + .get_mc_fb_offset = mmhub_v3_0_get_mc_fb_offset, + .gart_enable = mmhub_v3_0_gart_enable, + .set_fault_enable_default = mmhub_v3_0_set_fault_enable_default, + .gart_disable = mmhub_v3_0_gart_disable, + .set_clockgating = mmhub_v3_0_set_clockgating, + .get_clockgating = mmhub_v3_0_get_clockgating, + .setup_vm_pt_regs = mmhub_v3_0_setup_vm_pt_regs, +}; diff --git a/drivers/gpu/drm/amd/amdgpu/mmhub_v3_0.h b/drivers/gpu/drm/amd/amdgpu/mmhub_v3_0.h new file mode 100644 index 000000000000..3ced20f350bb --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/mmhub_v3_0.h @@ -0,0 +1,28 @@ +/* + * Copyright 2021 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ +#ifndef __MMHUB_V3_0_H__ +#define __MMHUB_V3_0_H__ + +extern const struct amdgpu_mmhub_funcs mmhub_v3_0_funcs; + +#endif diff --git a/drivers/gpu/drm/amd/amdgpu/mmhub_v3_0_1.c b/drivers/gpu/drm/amd/amdgpu/mmhub_v3_0_1.c new file mode 100644 index 000000000000..e8058edc1d10 --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/mmhub_v3_0_1.c @@ -0,0 +1,591 @@ +/* + * Copyright 2022 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#include "amdgpu.h" +#include "mmhub_v3_0_1.h" + +#include "mmhub/mmhub_3_0_1_offset.h" +#include "mmhub/mmhub_3_0_1_sh_mask.h" +#include "navi10_enum.h" + +#include "soc15_common.h" + +#define regMMVM_L2_CNTL3_DEFAULT 0x80100007 +#define regMMVM_L2_CNTL4_DEFAULT 0x000000c1 +#define regMMVM_L2_CNTL5_DEFAULT 0x00003fe0 + +static const char *mmhub_client_ids_v3_0_1[][2] = { + [0][0] = "VMC", + [4][0] = "DCEDMC", + [5][0] = "DCEVGA", + [6][0] = "MP0", + [7][0] = "MP1", + [8][0] = "MPIO", + [16][0] = "HDP", + [17][0] = "LSDMA", + [18][0] = "JPEG", + [19][0] = "VCNU0", + [21][0] = "VSCH", + [22][0] = "VCNU1", + [23][0] = "VCN1", + [32+20][0] = "VCN0", + [2][1] = "DBGUNBIO", + [3][1] = "DCEDWB", + [4][1] = "DCEDMC", + [5][1] = "DCEVGA", + [6][1] = "MP0", + [7][1] = "MP1", + [8][1] = "MPIO", + [10][1] = "DBGU0", + [11][1] = "DBGU1", + [12][1] = "DBGU2", + [13][1] = "DBGU3", + [14][1] = "XDP", + [15][1] = "OSSSYS", + [16][1] = "HDP", + [17][1] = "LSDMA", + [18][1] = "JPEG", + [19][1] = "VCNU0", + [20][1] = "VCN0", + [21][1] = "VSCH", + [22][1] = "VCNU1", + [23][1] = "VCN1", +}; + +static uint32_t mmhub_v3_0_1_get_invalidate_req(unsigned int vmid, + uint32_t flush_type) +{ + u32 req = 0; + + /* invalidate using legacy mode on vmid*/ + req = REG_SET_FIELD(req, MMVM_INVALIDATE_ENG0_REQ, + PER_VMID_INVALIDATE_REQ, 1 << vmid); + req = REG_SET_FIELD(req, MMVM_INVALIDATE_ENG0_REQ, FLUSH_TYPE, flush_type); + req = REG_SET_FIELD(req, MMVM_INVALIDATE_ENG0_REQ, INVALIDATE_L2_PTES, 1); + req = REG_SET_FIELD(req, MMVM_INVALIDATE_ENG0_REQ, INVALIDATE_L2_PDE0, 1); + req = REG_SET_FIELD(req, MMVM_INVALIDATE_ENG0_REQ, INVALIDATE_L2_PDE1, 1); + req = REG_SET_FIELD(req, MMVM_INVALIDATE_ENG0_REQ, INVALIDATE_L2_PDE2, 1); + req = REG_SET_FIELD(req, MMVM_INVALIDATE_ENG0_REQ, INVALIDATE_L1_PTES, 1); + req = REG_SET_FIELD(req, MMVM_INVALIDATE_ENG0_REQ, + CLEAR_PROTECTION_FAULT_STATUS_ADDR, 0); + + return req; +} + +static void +mmhub_v3_0_1_print_l2_protection_fault_status(struct amdgpu_device *adev, + uint32_t status) +{ + uint32_t cid, rw; + const char *mmhub_cid = NULL; + + cid = REG_GET_FIELD(status, + MMVM_L2_PROTECTION_FAULT_STATUS, CID); + rw = REG_GET_FIELD(status, + MMVM_L2_PROTECTION_FAULT_STATUS, RW); + + dev_err(adev->dev, + "MMVM_L2_PROTECTION_FAULT_STATUS:0x%08X\n", + status); + + switch (adev->ip_versions[MMHUB_HWIP][0]) { + case IP_VERSION(3, 0, 1): + mmhub_cid = mmhub_client_ids_v3_0_1[cid][rw]; + break; + default: + mmhub_cid = NULL; + break; + } + + dev_err(adev->dev, "\t Faulty UTCL2 client ID: %s (0x%x)\n", + mmhub_cid ? mmhub_cid : "unknown", cid); + dev_err(adev->dev, "\t MORE_FAULTS: 0x%lx\n", + REG_GET_FIELD(status, + MMVM_L2_PROTECTION_FAULT_STATUS, MORE_FAULTS)); + dev_err(adev->dev, "\t WALKER_ERROR: 0x%lx\n", + REG_GET_FIELD(status, + MMVM_L2_PROTECTION_FAULT_STATUS, WALKER_ERROR)); + dev_err(adev->dev, "\t PERMISSION_FAULTS: 0x%lx\n", + REG_GET_FIELD(status, + MMVM_L2_PROTECTION_FAULT_STATUS, PERMISSION_FAULTS)); + dev_err(adev->dev, "\t MAPPING_ERROR: 0x%lx\n", + REG_GET_FIELD(status, + MMVM_L2_PROTECTION_FAULT_STATUS, MAPPING_ERROR)); + dev_err(adev->dev, "\t RW: 0x%x\n", rw); +} + +static void mmhub_v3_0_1_setup_vm_pt_regs(struct amdgpu_device *adev, + uint32_t vmid, + uint64_t page_table_base) +{ + struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0]; + + WREG32_SOC15_OFFSET(MMHUB, 0, regMMVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32, + hub->ctx_addr_distance * vmid, + lower_32_bits(page_table_base)); + + WREG32_SOC15_OFFSET(MMHUB, 0, regMMVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_HI32, + hub->ctx_addr_distance * vmid, + upper_32_bits(page_table_base)); +} + +static void mmhub_v3_0_1_init_gart_aperture_regs(struct amdgpu_device *adev) +{ + uint64_t pt_base = amdgpu_gmc_pd_addr(adev->gart.bo); + + mmhub_v3_0_1_setup_vm_pt_regs(adev, 0, pt_base); + + WREG32_SOC15(MMHUB, 0, regMMVM_CONTEXT0_PAGE_TABLE_START_ADDR_LO32, + (u32)(adev->gmc.gart_start >> 12)); + WREG32_SOC15(MMHUB, 0, regMMVM_CONTEXT0_PAGE_TABLE_START_ADDR_HI32, + (u32)(adev->gmc.gart_start >> 44)); + + WREG32_SOC15(MMHUB, 0, regMMVM_CONTEXT0_PAGE_TABLE_END_ADDR_LO32, + (u32)(adev->gmc.gart_end >> 12)); + WREG32_SOC15(MMHUB, 0, regMMVM_CONTEXT0_PAGE_TABLE_END_ADDR_HI32, + (u32)(adev->gmc.gart_end >> 44)); +} + +static void mmhub_v3_0_1_init_system_aperture_regs(struct amdgpu_device *adev) +{ + uint64_t value; + uint32_t tmp; + + /* Program the AGP BAR */ + WREG32_SOC15(MMHUB, 0, regMMMC_VM_AGP_BASE, 0); + WREG32_SOC15(MMHUB, 0, regMMMC_VM_AGP_BOT, adev->gmc.agp_start >> 24); + WREG32_SOC15(MMHUB, 0, regMMMC_VM_AGP_TOP, adev->gmc.agp_end >> 24); + + /* + * the new L1 policy will block SRIOV guest from writing + * these regs, and they will be programed at host. + * so skip programing these regs. + */ + /* Program the system aperture low logical page number. */ + WREG32_SOC15(MMHUB, 0, regMMMC_VM_SYSTEM_APERTURE_LOW_ADDR, + adev->gmc.vram_start >> 18); + WREG32_SOC15(MMHUB, 0, regMMMC_VM_SYSTEM_APERTURE_HIGH_ADDR, + adev->gmc.vram_end >> 18); + + /* Set default page address. */ + value = adev->vram_scratch.gpu_addr - adev->gmc.vram_start + + adev->vm_manager.vram_base_offset; + WREG32_SOC15(MMHUB, 0, regMMMC_VM_SYSTEM_APERTURE_DEFAULT_ADDR_LSB, + (u32)(value >> 12)); + WREG32_SOC15(MMHUB, 0, regMMMC_VM_SYSTEM_APERTURE_DEFAULT_ADDR_MSB, + (u32)(value >> 44)); + + /* Program "protection fault". */ + WREG32_SOC15(MMHUB, 0, regMMVM_L2_PROTECTION_FAULT_DEFAULT_ADDR_LO32, + (u32)(adev->dummy_page_addr >> 12)); + WREG32_SOC15(MMHUB, 0, regMMVM_L2_PROTECTION_FAULT_DEFAULT_ADDR_HI32, + (u32)((u64)adev->dummy_page_addr >> 44)); + + tmp = RREG32_SOC15(MMHUB, 0, regMMVM_L2_PROTECTION_FAULT_CNTL2); + tmp = REG_SET_FIELD(tmp, MMVM_L2_PROTECTION_FAULT_CNTL2, + ACTIVE_PAGE_MIGRATION_PTE_READ_RETRY, 1); + WREG32_SOC15(MMHUB, 0, regMMVM_L2_PROTECTION_FAULT_CNTL2, tmp); +} + +static void mmhub_v3_0_1_init_tlb_regs(struct amdgpu_device *adev) +{ + uint32_t tmp; + + /* Setup TLB control */ + tmp = RREG32_SOC15(MMHUB, 0, regMMMC_VM_MX_L1_TLB_CNTL); + + tmp = REG_SET_FIELD(tmp, MMMC_VM_MX_L1_TLB_CNTL, ENABLE_L1_TLB, 1); + tmp = REG_SET_FIELD(tmp, MMMC_VM_MX_L1_TLB_CNTL, SYSTEM_ACCESS_MODE, 3); + tmp = REG_SET_FIELD(tmp, MMMC_VM_MX_L1_TLB_CNTL, + ENABLE_ADVANCED_DRIVER_MODEL, 1); + tmp = REG_SET_FIELD(tmp, MMMC_VM_MX_L1_TLB_CNTL, + SYSTEM_APERTURE_UNMAPPED_ACCESS, 0); + tmp = REG_SET_FIELD(tmp, MMMC_VM_MX_L1_TLB_CNTL, ECO_BITS, 0); + tmp = REG_SET_FIELD(tmp, MMMC_VM_MX_L1_TLB_CNTL, + MTYPE, MTYPE_UC); /* UC, uncached */ + + WREG32_SOC15(MMHUB, 0, regMMMC_VM_MX_L1_TLB_CNTL, tmp); +} + +static void mmhub_v3_0_1_init_cache_regs(struct amdgpu_device *adev) +{ + uint32_t tmp; + + /* Setup L2 cache */ + tmp = RREG32_SOC15(MMHUB, 0, regMMVM_L2_CNTL); + tmp = REG_SET_FIELD(tmp, MMVM_L2_CNTL, ENABLE_L2_CACHE, 1); + tmp = REG_SET_FIELD(tmp, MMVM_L2_CNTL, ENABLE_L2_FRAGMENT_PROCESSING, 0); + tmp = REG_SET_FIELD(tmp, MMVM_L2_CNTL, + ENABLE_DEFAULT_PAGE_OUT_TO_SYSTEM_MEMORY, 1); + /* XXX for emulation, Refer to closed source code.*/ + tmp = REG_SET_FIELD(tmp, MMVM_L2_CNTL, L2_PDE0_CACHE_TAG_GENERATION_MODE, + 0); + tmp = REG_SET_FIELD(tmp, MMVM_L2_CNTL, PDE_FAULT_CLASSIFICATION, 0); + tmp = REG_SET_FIELD(tmp, MMVM_L2_CNTL, CONTEXT1_IDENTITY_ACCESS_MODE, 1); + tmp = REG_SET_FIELD(tmp, MMVM_L2_CNTL, IDENTITY_MODE_FRAGMENT_SIZE, 0); + WREG32_SOC15(MMHUB, 0, regMMVM_L2_CNTL, tmp); + + tmp = RREG32_SOC15(MMHUB, 0, regMMVM_L2_CNTL2); + tmp = REG_SET_FIELD(tmp, MMVM_L2_CNTL2, INVALIDATE_ALL_L1_TLBS, 1); + tmp = REG_SET_FIELD(tmp, MMVM_L2_CNTL2, INVALIDATE_L2_CACHE, 1); + WREG32_SOC15(MMHUB, 0, regMMVM_L2_CNTL2, tmp); + + tmp = regMMVM_L2_CNTL3_DEFAULT; + if (adev->gmc.translate_further) { + tmp = REG_SET_FIELD(tmp, MMVM_L2_CNTL3, BANK_SELECT, 12); + tmp = REG_SET_FIELD(tmp, MMVM_L2_CNTL3, + L2_CACHE_BIGK_FRAGMENT_SIZE, 9); + } else { + tmp = REG_SET_FIELD(tmp, MMVM_L2_CNTL3, BANK_SELECT, 9); + tmp = REG_SET_FIELD(tmp, MMVM_L2_CNTL3, + L2_CACHE_BIGK_FRAGMENT_SIZE, 6); + } + WREG32_SOC15(MMHUB, 0, regMMVM_L2_CNTL3, tmp); + + tmp = regMMVM_L2_CNTL4_DEFAULT; + tmp = REG_SET_FIELD(tmp, MMVM_L2_CNTL4, VMC_TAP_PDE_REQUEST_PHYSICAL, 0); + tmp = REG_SET_FIELD(tmp, MMVM_L2_CNTL4, VMC_TAP_PTE_REQUEST_PHYSICAL, 0); + WREG32_SOC15(MMHUB, 0, regMMVM_L2_CNTL4, tmp); + + tmp = regMMVM_L2_CNTL5_DEFAULT; + tmp = REG_SET_FIELD(tmp, MMVM_L2_CNTL5, L2_CACHE_SMALLK_FRAGMENT_SIZE, 0); + WREG32_SOC15(GC, 0, regMMVM_L2_CNTL5, tmp); +} + +static void mmhub_v3_0_1_enable_system_domain(struct amdgpu_device *adev) +{ + uint32_t tmp; + + tmp = RREG32_SOC15(MMHUB, 0, regMMVM_CONTEXT0_CNTL); + tmp = REG_SET_FIELD(tmp, MMVM_CONTEXT0_CNTL, ENABLE_CONTEXT, 1); + tmp = REG_SET_FIELD(tmp, MMVM_CONTEXT0_CNTL, PAGE_TABLE_DEPTH, 0); + tmp = REG_SET_FIELD(tmp, MMVM_CONTEXT0_CNTL, + RETRY_PERMISSION_OR_INVALID_PAGE_FAULT, 0); + WREG32_SOC15(MMHUB, 0, regMMVM_CONTEXT0_CNTL, tmp); +} + +static void mmhub_v3_0_1_disable_identity_aperture(struct amdgpu_device *adev) +{ + WREG32_SOC15(MMHUB, 0, + regMMVM_L2_CONTEXT1_IDENTITY_APERTURE_LOW_ADDR_LO32, + 0xFFFFFFFF); + WREG32_SOC15(MMHUB, 0, + regMMVM_L2_CONTEXT1_IDENTITY_APERTURE_LOW_ADDR_HI32, + 0x0000000F); + + WREG32_SOC15(MMHUB, 0, + regMMVM_L2_CONTEXT1_IDENTITY_APERTURE_HIGH_ADDR_LO32, 0); + WREG32_SOC15(MMHUB, 0, + regMMVM_L2_CONTEXT1_IDENTITY_APERTURE_HIGH_ADDR_HI32, 0); + + WREG32_SOC15(MMHUB, 0, regMMVM_L2_CONTEXT_IDENTITY_PHYSICAL_OFFSET_LO32, + 0); + WREG32_SOC15(MMHUB, 0, regMMVM_L2_CONTEXT_IDENTITY_PHYSICAL_OFFSET_HI32, + 0); +} + +static void mmhub_v3_0_1_setup_vmid_config(struct amdgpu_device *adev) +{ + struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0]; + int i; + uint32_t tmp; + + for (i = 0; i <= 14; i++) { + tmp = RREG32_SOC15_OFFSET(MMHUB, 0, regMMVM_CONTEXT1_CNTL, i); + tmp = REG_SET_FIELD(tmp, MMVM_CONTEXT1_CNTL, ENABLE_CONTEXT, 1); + tmp = REG_SET_FIELD(tmp, MMVM_CONTEXT1_CNTL, PAGE_TABLE_DEPTH, + adev->vm_manager.num_level); + tmp = REG_SET_FIELD(tmp, MMVM_CONTEXT1_CNTL, + RANGE_PROTECTION_FAULT_ENABLE_DEFAULT, 1); + tmp = REG_SET_FIELD(tmp, MMVM_CONTEXT1_CNTL, + DUMMY_PAGE_PROTECTION_FAULT_ENABLE_DEFAULT, + 1); + tmp = REG_SET_FIELD(tmp, MMVM_CONTEXT1_CNTL, + PDE0_PROTECTION_FAULT_ENABLE_DEFAULT, 1); + tmp = REG_SET_FIELD(tmp, MMVM_CONTEXT1_CNTL, + VALID_PROTECTION_FAULT_ENABLE_DEFAULT, 1); + tmp = REG_SET_FIELD(tmp, MMVM_CONTEXT1_CNTL, + READ_PROTECTION_FAULT_ENABLE_DEFAULT, 1); + tmp = REG_SET_FIELD(tmp, MMVM_CONTEXT1_CNTL, + WRITE_PROTECTION_FAULT_ENABLE_DEFAULT, 1); + tmp = REG_SET_FIELD(tmp, MMVM_CONTEXT1_CNTL, + EXECUTE_PROTECTION_FAULT_ENABLE_DEFAULT, 1); + tmp = REG_SET_FIELD(tmp, MMVM_CONTEXT1_CNTL, + PAGE_TABLE_BLOCK_SIZE, + adev->vm_manager.block_size - 9); + /* Send no-retry XNACK on fault to suppress VM fault storm. */ + tmp = REG_SET_FIELD(tmp, MMVM_CONTEXT1_CNTL, + RETRY_PERMISSION_OR_INVALID_PAGE_FAULT, + !amdgpu_noretry); + WREG32_SOC15_OFFSET(MMHUB, 0, regMMVM_CONTEXT1_CNTL, + i * hub->ctx_distance, tmp); + WREG32_SOC15_OFFSET(MMHUB, 0, regMMVM_CONTEXT1_PAGE_TABLE_START_ADDR_LO32, + i * hub->ctx_addr_distance, 0); + WREG32_SOC15_OFFSET(MMHUB, 0, regMMVM_CONTEXT1_PAGE_TABLE_START_ADDR_HI32, + i * hub->ctx_addr_distance, 0); + WREG32_SOC15_OFFSET(MMHUB, 0, regMMVM_CONTEXT1_PAGE_TABLE_END_ADDR_LO32, + i * hub->ctx_addr_distance, + lower_32_bits(adev->vm_manager.max_pfn - 1)); + WREG32_SOC15_OFFSET(MMHUB, 0, regMMVM_CONTEXT1_PAGE_TABLE_END_ADDR_HI32, + i * hub->ctx_addr_distance, + upper_32_bits(adev->vm_manager.max_pfn - 1)); + } + + hub->vm_cntx_cntl = tmp; +} + +static void mmhub_v3_0_1_program_invalidation(struct amdgpu_device *adev) +{ + struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0]; + unsigned i; + + for (i = 0; i < 18; ++i) { + WREG32_SOC15_OFFSET(MMHUB, 0, regMMVM_INVALIDATE_ENG0_ADDR_RANGE_LO32, + i * hub->eng_addr_distance, 0xffffffff); + WREG32_SOC15_OFFSET(MMHUB, 0, regMMVM_INVALIDATE_ENG0_ADDR_RANGE_HI32, + i * hub->eng_addr_distance, 0x1f); + } +} + +static int mmhub_v3_0_1_gart_enable(struct amdgpu_device *adev) +{ + /* GART Enable. */ + mmhub_v3_0_1_init_gart_aperture_regs(adev); + mmhub_v3_0_1_init_system_aperture_regs(adev); + mmhub_v3_0_1_init_tlb_regs(adev); + mmhub_v3_0_1_init_cache_regs(adev); + + mmhub_v3_0_1_enable_system_domain(adev); + mmhub_v3_0_1_disable_identity_aperture(adev); + mmhub_v3_0_1_setup_vmid_config(adev); + mmhub_v3_0_1_program_invalidation(adev); + + return 0; +} + +static void mmhub_v3_0_1_gart_disable(struct amdgpu_device *adev) +{ + struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0]; + u32 tmp; + u32 i; + + /* Disable all tables */ + for (i = 0; i < 16; i++) + WREG32_SOC15_OFFSET(MMHUB, 0, regMMVM_CONTEXT0_CNTL, + i * hub->ctx_distance, 0); + + /* Setup TLB control */ + tmp = RREG32_SOC15(MMHUB, 0, regMMMC_VM_MX_L1_TLB_CNTL); + tmp = REG_SET_FIELD(tmp, MMMC_VM_MX_L1_TLB_CNTL, ENABLE_L1_TLB, 0); + tmp = REG_SET_FIELD(tmp, MMMC_VM_MX_L1_TLB_CNTL, + ENABLE_ADVANCED_DRIVER_MODEL, 0); + WREG32_SOC15(MMHUB, 0, regMMMC_VM_MX_L1_TLB_CNTL, tmp); + + /* Setup L2 cache */ + tmp = RREG32_SOC15(MMHUB, 0, regMMVM_L2_CNTL); + tmp = REG_SET_FIELD(tmp, MMVM_L2_CNTL, ENABLE_L2_CACHE, 0); + WREG32_SOC15(MMHUB, 0, regMMVM_L2_CNTL, tmp); + WREG32_SOC15(MMHUB, 0, regMMVM_L2_CNTL3, 0); +} + +/** + * mmhub_v3_0_1_set_fault_enable_default - update GART/VM fault handling + * + * @adev: amdgpu_device pointer + * @value: true redirects VM faults to the default page + */ +static void mmhub_v3_0_1_set_fault_enable_default(struct amdgpu_device *adev, + bool value) +{ + u32 tmp; + + tmp = RREG32_SOC15(MMHUB, 0, regMMVM_L2_PROTECTION_FAULT_CNTL); + tmp = REG_SET_FIELD(tmp, MMVM_L2_PROTECTION_FAULT_CNTL, + RANGE_PROTECTION_FAULT_ENABLE_DEFAULT, value); + tmp = REG_SET_FIELD(tmp, MMVM_L2_PROTECTION_FAULT_CNTL, + PDE0_PROTECTION_FAULT_ENABLE_DEFAULT, value); + tmp = REG_SET_FIELD(tmp, MMVM_L2_PROTECTION_FAULT_CNTL, + PDE1_PROTECTION_FAULT_ENABLE_DEFAULT, value); + tmp = REG_SET_FIELD(tmp, MMVM_L2_PROTECTION_FAULT_CNTL, + PDE2_PROTECTION_FAULT_ENABLE_DEFAULT, value); + tmp = REG_SET_FIELD(tmp, MMVM_L2_PROTECTION_FAULT_CNTL, + TRANSLATE_FURTHER_PROTECTION_FAULT_ENABLE_DEFAULT, + value); + tmp = REG_SET_FIELD(tmp, MMVM_L2_PROTECTION_FAULT_CNTL, + NACK_PROTECTION_FAULT_ENABLE_DEFAULT, value); + tmp = REG_SET_FIELD(tmp, MMVM_L2_PROTECTION_FAULT_CNTL, + DUMMY_PAGE_PROTECTION_FAULT_ENABLE_DEFAULT, value); + tmp = REG_SET_FIELD(tmp, MMVM_L2_PROTECTION_FAULT_CNTL, + VALID_PROTECTION_FAULT_ENABLE_DEFAULT, value); + tmp = REG_SET_FIELD(tmp, MMVM_L2_PROTECTION_FAULT_CNTL, + READ_PROTECTION_FAULT_ENABLE_DEFAULT, value); + tmp = REG_SET_FIELD(tmp, MMVM_L2_PROTECTION_FAULT_CNTL, + WRITE_PROTECTION_FAULT_ENABLE_DEFAULT, value); + tmp = REG_SET_FIELD(tmp, MMVM_L2_PROTECTION_FAULT_CNTL, + EXECUTE_PROTECTION_FAULT_ENABLE_DEFAULT, value); + if (!value) { + tmp = REG_SET_FIELD(tmp, MMVM_L2_PROTECTION_FAULT_CNTL, + CRASH_ON_NO_RETRY_FAULT, 1); + tmp = REG_SET_FIELD(tmp, MMVM_L2_PROTECTION_FAULT_CNTL, + CRASH_ON_RETRY_FAULT, 1); + } + WREG32_SOC15(MMHUB, 0, regMMVM_L2_PROTECTION_FAULT_CNTL, tmp); +} + +static const struct amdgpu_vmhub_funcs mmhub_v3_0_1_vmhub_funcs = { + .print_l2_protection_fault_status = mmhub_v3_0_1_print_l2_protection_fault_status, + .get_invalidate_req = mmhub_v3_0_1_get_invalidate_req, +}; + +static void mmhub_v3_0_1_init(struct amdgpu_device *adev) +{ + struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0]; + + hub->ctx0_ptb_addr_lo32 = + SOC15_REG_OFFSET(MMHUB, 0, + regMMVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32); + hub->ctx0_ptb_addr_hi32 = + SOC15_REG_OFFSET(MMHUB, 0, + regMMVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_HI32); + hub->vm_inv_eng0_sem = + SOC15_REG_OFFSET(MMHUB, 0, regMMVM_INVALIDATE_ENG0_SEM); + hub->vm_inv_eng0_req = + SOC15_REG_OFFSET(MMHUB, 0, regMMVM_INVALIDATE_ENG0_REQ); + hub->vm_inv_eng0_ack = + SOC15_REG_OFFSET(MMHUB, 0, regMMVM_INVALIDATE_ENG0_ACK); + hub->vm_context0_cntl = + SOC15_REG_OFFSET(MMHUB, 0, regMMVM_CONTEXT0_CNTL); + hub->vm_l2_pro_fault_status = + SOC15_REG_OFFSET(MMHUB, 0, regMMVM_L2_PROTECTION_FAULT_STATUS); + hub->vm_l2_pro_fault_cntl = + SOC15_REG_OFFSET(MMHUB, 0, regMMVM_L2_PROTECTION_FAULT_CNTL); + + hub->ctx_distance = regMMVM_CONTEXT1_CNTL - regMMVM_CONTEXT0_CNTL; + hub->ctx_addr_distance = regMMVM_CONTEXT1_PAGE_TABLE_BASE_ADDR_LO32 - + regMMVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32; + hub->eng_distance = regMMVM_INVALIDATE_ENG1_REQ - + regMMVM_INVALIDATE_ENG0_REQ; + hub->eng_addr_distance = regMMVM_INVALIDATE_ENG1_ADDR_RANGE_LO32 - + regMMVM_INVALIDATE_ENG0_ADDR_RANGE_LO32; + + hub->vm_cntx_cntl_vm_fault = MMVM_CONTEXT1_CNTL__RANGE_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK | + MMVM_CONTEXT1_CNTL__DUMMY_PAGE_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK | + MMVM_CONTEXT1_CNTL__PDE0_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK | + MMVM_CONTEXT1_CNTL__VALID_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK | + MMVM_CONTEXT1_CNTL__READ_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK | + MMVM_CONTEXT1_CNTL__WRITE_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK | + MMVM_CONTEXT1_CNTL__EXECUTE_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK; + + hub->vmhub_funcs = &mmhub_v3_0_1_vmhub_funcs; +} + +static u64 mmhub_v3_0_1_get_fb_location(struct amdgpu_device *adev) +{ + u64 base; + + base = RREG32_SOC15(MMHUB, 0, regMMMC_VM_FB_LOCATION_BASE); + base &= MMMC_VM_FB_LOCATION_BASE__FB_BASE_MASK; + base <<= 24; + + return base; +} + +static u64 mmhub_v3_0_1_get_mc_fb_offset(struct amdgpu_device *adev) +{ + return (u64)RREG32_SOC15(MMHUB, 0, regMMMC_VM_FB_OFFSET) << 24; +} + +static void mmhub_v3_0_1_update_medium_grain_clock_gating(struct amdgpu_device *adev, + bool enable) +{ + uint32_t def, data; + + def = data = RREG32_SOC15(MMHUB, 0, regMM_ATC_L2_MISC_CG); + + if (enable) + data |= MM_ATC_L2_MISC_CG__ENABLE_MASK; + else + data &= ~MM_ATC_L2_MISC_CG__ENABLE_MASK; + + if (def != data) + WREG32_SOC15(MMHUB, 0, regMM_ATC_L2_MISC_CG, data); +} + +static void mmhub_v3_0_1_update_medium_grain_light_sleep(struct amdgpu_device *adev, + bool enable) +{ + uint32_t def, data; + + def = data = RREG32_SOC15(MMHUB, 0, regMM_ATC_L2_MISC_CG); + + if (enable) + data |= MM_ATC_L2_MISC_CG__MEM_LS_ENABLE_MASK; + else + data &= ~MM_ATC_L2_MISC_CG__MEM_LS_ENABLE_MASK; + + if (def != data) + WREG32_SOC15(MMHUB, 0, regMM_ATC_L2_MISC_CG, data); +} + +static int mmhub_v3_0_1_set_clockgating(struct amdgpu_device *adev, + enum amd_clockgating_state state) +{ + if (amdgpu_sriov_vf(adev)) + return 0; + + mmhub_v3_0_1_update_medium_grain_clock_gating(adev, + state == AMD_CG_STATE_GATE); + mmhub_v3_0_1_update_medium_grain_light_sleep(adev, + state == AMD_CG_STATE_GATE); + return 0; +} + +static void mmhub_v3_0_1_get_clockgating(struct amdgpu_device *adev, u64 *flags) +{ + int data; + + if (amdgpu_sriov_vf(adev)) + *flags = 0; + + data = RREG32_SOC15(MMHUB, 0, regMM_ATC_L2_MISC_CG); + + /* AMD_CG_SUPPORT_MC_MGCG */ + if (data & MM_ATC_L2_MISC_CG__ENABLE_MASK) + *flags |= AMD_CG_SUPPORT_MC_MGCG; + + /* AMD_CG_SUPPORT_MC_LS */ + if (data & MM_ATC_L2_MISC_CG__MEM_LS_ENABLE_MASK) + *flags |= AMD_CG_SUPPORT_MC_LS; +} + +const struct amdgpu_mmhub_funcs mmhub_v3_0_1_funcs = { + .init = mmhub_v3_0_1_init, + .get_fb_location = mmhub_v3_0_1_get_fb_location, + .get_mc_fb_offset = mmhub_v3_0_1_get_mc_fb_offset, + .gart_enable = mmhub_v3_0_1_gart_enable, + .set_fault_enable_default = mmhub_v3_0_1_set_fault_enable_default, + .gart_disable = mmhub_v3_0_1_gart_disable, + .set_clockgating = mmhub_v3_0_1_set_clockgating, + .get_clockgating = mmhub_v3_0_1_get_clockgating, + .setup_vm_pt_regs = mmhub_v3_0_1_setup_vm_pt_regs, +}; diff --git a/drivers/gpu/drm/amd/amdgpu/mmhub_v3_0_1.h b/drivers/gpu/drm/amd/amdgpu/mmhub_v3_0_1.h new file mode 100644 index 000000000000..4c1246735e7d --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/mmhub_v3_0_1.h @@ -0,0 +1,28 @@ +/* + * Copyright 2022 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ +#ifndef __MMHUB_V3_0_1_H__ +#define __MMHUB_V3_0_1_H__ + +extern const struct amdgpu_mmhub_funcs mmhub_v3_0_1_funcs; + +#endif diff --git a/drivers/gpu/drm/amd/amdgpu/mmhub_v3_0_2.c b/drivers/gpu/drm/amd/amdgpu/mmhub_v3_0_2.c new file mode 100644 index 000000000000..770be0a8f7ce --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/mmhub_v3_0_2.c @@ -0,0 +1,571 @@ +/* + * Copyright 2022 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#include "amdgpu.h" +#include "mmhub_v3_0_2.h" + +#include "mmhub/mmhub_3_0_2_offset.h" +#include "mmhub/mmhub_3_0_2_sh_mask.h" +#include "navi10_enum.h" + +#include "soc15_common.h" + +#define regMMVM_L2_CNTL3_DEFAULT 0x80100007 +#define regMMVM_L2_CNTL4_DEFAULT 0x000000c1 +#define regMMVM_L2_CNTL5_DEFAULT 0x00003fe0 + +static const char *mmhub_client_ids_v3_0_2[][2] = { + [0][0] = "VMC", + [4][0] = "DCEDMC", + [5][0] = "DCEVGA", + [6][0] = "MP0", + [7][0] = "MP1", + [8][0] = "MPIO", + [16][0] = "HDP", + [17][0] = "LSDMA", + [18][0] = "JPEG", + [19][0] = "VCNU0", + [21][0] = "VSCH", + [22][0] = "VCNU1", + [23][0] = "VCN1", + [32+20][0] = "VCN0", + [2][1] = "DBGUNBIO", + [3][1] = "DCEDWB", + [4][1] = "DCEDMC", + [5][1] = "DCEVGA", + [6][1] = "MP0", + [7][1] = "MP1", + [8][1] = "MPIO", + [10][1] = "DBGU0", + [11][1] = "DBGU1", + [12][1] = "DBGU2", + [13][1] = "DBGU3", + [14][1] = "XDP", + [15][1] = "OSSSYS", + [16][1] = "HDP", + [17][1] = "LSDMA", + [18][1] = "JPEG", + [19][1] = "VCNU0", + [20][1] = "VCN0", + [21][1] = "VSCH", + [22][1] = "VCNU1", + [23][1] = "VCN1", +}; + +static uint32_t mmhub_v3_0_2_get_invalidate_req(unsigned int vmid, + uint32_t flush_type) +{ + u32 req = 0; + + /* invalidate using legacy mode on vmid*/ + req = REG_SET_FIELD(req, MMVM_INVALIDATE_ENG0_REQ, + PER_VMID_INVALIDATE_REQ, 1 << vmid); + req = REG_SET_FIELD(req, MMVM_INVALIDATE_ENG0_REQ, FLUSH_TYPE, flush_type); + req = REG_SET_FIELD(req, MMVM_INVALIDATE_ENG0_REQ, INVALIDATE_L2_PTES, 1); + req = REG_SET_FIELD(req, MMVM_INVALIDATE_ENG0_REQ, INVALIDATE_L2_PDE0, 1); + req = REG_SET_FIELD(req, MMVM_INVALIDATE_ENG0_REQ, INVALIDATE_L2_PDE1, 1); + req = REG_SET_FIELD(req, MMVM_INVALIDATE_ENG0_REQ, INVALIDATE_L2_PDE2, 1); + req = REG_SET_FIELD(req, MMVM_INVALIDATE_ENG0_REQ, INVALIDATE_L1_PTES, 1); + req = REG_SET_FIELD(req, MMVM_INVALIDATE_ENG0_REQ, + CLEAR_PROTECTION_FAULT_STATUS_ADDR, 0); + + return req; +} + +static void +mmhub_v3_0_2_print_l2_protection_fault_status(struct amdgpu_device *adev, + uint32_t status) +{ + uint32_t cid, rw; + const char *mmhub_cid = NULL; + + cid = REG_GET_FIELD(status, + MMVM_L2_PROTECTION_FAULT_STATUS, CID); + rw = REG_GET_FIELD(status, + MMVM_L2_PROTECTION_FAULT_STATUS, RW); + + dev_err(adev->dev, + "MMVM_L2_PROTECTION_FAULT_STATUS:0x%08X\n", + status); + + mmhub_cid = mmhub_client_ids_v3_0_2[cid][rw]; + dev_err(adev->dev, "\t Faulty UTCL2 client ID: %s (0x%x)\n", + mmhub_cid ? mmhub_cid : "unknown", cid); + dev_err(adev->dev, "\t MORE_FAULTS: 0x%lx\n", + REG_GET_FIELD(status, + MMVM_L2_PROTECTION_FAULT_STATUS, MORE_FAULTS)); + dev_err(adev->dev, "\t WALKER_ERROR: 0x%lx\n", + REG_GET_FIELD(status, + MMVM_L2_PROTECTION_FAULT_STATUS, WALKER_ERROR)); + dev_err(adev->dev, "\t PERMISSION_FAULTS: 0x%lx\n", + REG_GET_FIELD(status, + MMVM_L2_PROTECTION_FAULT_STATUS, PERMISSION_FAULTS)); + dev_err(adev->dev, "\t MAPPING_ERROR: 0x%lx\n", + REG_GET_FIELD(status, + MMVM_L2_PROTECTION_FAULT_STATUS, MAPPING_ERROR)); + dev_err(adev->dev, "\t RW: 0x%x\n", rw); +} + +static void mmhub_v3_0_2_setup_vm_pt_regs(struct amdgpu_device *adev, uint32_t vmid, + uint64_t page_table_base) +{ + struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0]; + + WREG32_SOC15_OFFSET(MMHUB, 0, regMMVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32, + hub->ctx_addr_distance * vmid, + lower_32_bits(page_table_base)); + + WREG32_SOC15_OFFSET(MMHUB, 0, regMMVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_HI32, + hub->ctx_addr_distance * vmid, + upper_32_bits(page_table_base)); +} + +static void mmhub_v3_0_2_init_gart_aperture_regs(struct amdgpu_device *adev) +{ + uint64_t pt_base = amdgpu_gmc_pd_addr(adev->gart.bo); + + mmhub_v3_0_2_setup_vm_pt_regs(adev, 0, pt_base); + + WREG32_SOC15(MMHUB, 0, regMMVM_CONTEXT0_PAGE_TABLE_START_ADDR_LO32, + (u32)(adev->gmc.gart_start >> 12)); + WREG32_SOC15(MMHUB, 0, regMMVM_CONTEXT0_PAGE_TABLE_START_ADDR_HI32, + (u32)(adev->gmc.gart_start >> 44)); + + WREG32_SOC15(MMHUB, 0, regMMVM_CONTEXT0_PAGE_TABLE_END_ADDR_LO32, + (u32)(adev->gmc.gart_end >> 12)); + WREG32_SOC15(MMHUB, 0, regMMVM_CONTEXT0_PAGE_TABLE_END_ADDR_HI32, + (u32)(adev->gmc.gart_end >> 44)); +} + +static void mmhub_v3_0_2_init_system_aperture_regs(struct amdgpu_device *adev) +{ + uint64_t value; + uint32_t tmp; + + /* Disable AGP. */ + WREG32_SOC15(MMHUB, 0, regMMMC_VM_AGP_BASE, 0); + WREG32_SOC15(MMHUB, 0, regMMMC_VM_AGP_TOP, 0); + WREG32_SOC15(MMHUB, 0, regMMMC_VM_AGP_BOT, 0x00FFFFFF); + + if (!amdgpu_sriov_vf(adev)) { + /* + * the new L1 policy will block SRIOV guest from writing + * these regs, and they will be programed at host. + * so skip programing these regs. + */ + /* Program the system aperture low logical page number. */ + WREG32_SOC15(MMHUB, 0, regMMMC_VM_SYSTEM_APERTURE_LOW_ADDR, + adev->gmc.vram_start >> 18); + WREG32_SOC15(MMHUB, 0, regMMMC_VM_SYSTEM_APERTURE_HIGH_ADDR, + adev->gmc.vram_end >> 18); + } + + /* Set default page address. */ + value = adev->vram_scratch.gpu_addr - adev->gmc.vram_start + + adev->vm_manager.vram_base_offset; + WREG32_SOC15(MMHUB, 0, regMMMC_VM_SYSTEM_APERTURE_DEFAULT_ADDR_LSB, + (u32)(value >> 12)); + WREG32_SOC15(MMHUB, 0, regMMMC_VM_SYSTEM_APERTURE_DEFAULT_ADDR_MSB, + (u32)(value >> 44)); + + /* Program "protection fault". */ + WREG32_SOC15(MMHUB, 0, regMMVM_L2_PROTECTION_FAULT_DEFAULT_ADDR_LO32, + (u32)(adev->dummy_page_addr >> 12)); + WREG32_SOC15(MMHUB, 0, regMMVM_L2_PROTECTION_FAULT_DEFAULT_ADDR_HI32, + (u32)((u64)adev->dummy_page_addr >> 44)); + + tmp = RREG32_SOC15(MMHUB, 0, regMMVM_L2_PROTECTION_FAULT_CNTL2); + tmp = REG_SET_FIELD(tmp, MMVM_L2_PROTECTION_FAULT_CNTL2, + ACTIVE_PAGE_MIGRATION_PTE_READ_RETRY, 1); + WREG32_SOC15(MMHUB, 0, regMMVM_L2_PROTECTION_FAULT_CNTL2, tmp); +} + +static void mmhub_v3_0_2_init_tlb_regs(struct amdgpu_device *adev) +{ + uint32_t tmp; + + /* Setup TLB control */ + tmp = RREG32_SOC15(MMHUB, 0, regMMMC_VM_MX_L1_TLB_CNTL); + + tmp = REG_SET_FIELD(tmp, MMMC_VM_MX_L1_TLB_CNTL, ENABLE_L1_TLB, 1); + tmp = REG_SET_FIELD(tmp, MMMC_VM_MX_L1_TLB_CNTL, SYSTEM_ACCESS_MODE, 3); + tmp = REG_SET_FIELD(tmp, MMMC_VM_MX_L1_TLB_CNTL, + ENABLE_ADVANCED_DRIVER_MODEL, 1); + tmp = REG_SET_FIELD(tmp, MMMC_VM_MX_L1_TLB_CNTL, + SYSTEM_APERTURE_UNMAPPED_ACCESS, 0); + tmp = REG_SET_FIELD(tmp, MMMC_VM_MX_L1_TLB_CNTL, ECO_BITS, 0); + tmp = REG_SET_FIELD(tmp, MMMC_VM_MX_L1_TLB_CNTL, + MTYPE, MTYPE_UC); /* UC, uncached */ + + WREG32_SOC15(MMHUB, 0, regMMMC_VM_MX_L1_TLB_CNTL, tmp); +} + +static void mmhub_v3_0_2_init_cache_regs(struct amdgpu_device *adev) +{ + uint32_t tmp; + + /* These registers are not accessible to VF-SRIOV. + * The PF will program them instead. + */ + if (amdgpu_sriov_vf(adev)) + return; + + /* Setup L2 cache */ + tmp = RREG32_SOC15(MMHUB, 0, regMMVM_L2_CNTL); + tmp = REG_SET_FIELD(tmp, MMVM_L2_CNTL, ENABLE_L2_CACHE, 1); + tmp = REG_SET_FIELD(tmp, MMVM_L2_CNTL, ENABLE_L2_FRAGMENT_PROCESSING, 0); + tmp = REG_SET_FIELD(tmp, MMVM_L2_CNTL, + ENABLE_DEFAULT_PAGE_OUT_TO_SYSTEM_MEMORY, 1); + /* XXX for emulation, Refer to closed source code.*/ + tmp = REG_SET_FIELD(tmp, MMVM_L2_CNTL, L2_PDE0_CACHE_TAG_GENERATION_MODE, + 0); + tmp = REG_SET_FIELD(tmp, MMVM_L2_CNTL, PDE_FAULT_CLASSIFICATION, 0); + tmp = REG_SET_FIELD(tmp, MMVM_L2_CNTL, CONTEXT1_IDENTITY_ACCESS_MODE, 1); + tmp = REG_SET_FIELD(tmp, MMVM_L2_CNTL, IDENTITY_MODE_FRAGMENT_SIZE, 0); + WREG32_SOC15(MMHUB, 0, regMMVM_L2_CNTL, tmp); + + tmp = RREG32_SOC15(MMHUB, 0, regMMVM_L2_CNTL2); + tmp = REG_SET_FIELD(tmp, MMVM_L2_CNTL2, INVALIDATE_ALL_L1_TLBS, 1); + tmp = REG_SET_FIELD(tmp, MMVM_L2_CNTL2, INVALIDATE_L2_CACHE, 1); + WREG32_SOC15(MMHUB, 0, regMMVM_L2_CNTL2, tmp); + + tmp = regMMVM_L2_CNTL3_DEFAULT; + if (adev->gmc.translate_further) { + tmp = REG_SET_FIELD(tmp, MMVM_L2_CNTL3, BANK_SELECT, 12); + tmp = REG_SET_FIELD(tmp, MMVM_L2_CNTL3, + L2_CACHE_BIGK_FRAGMENT_SIZE, 9); + } else { + tmp = REG_SET_FIELD(tmp, MMVM_L2_CNTL3, BANK_SELECT, 9); + tmp = REG_SET_FIELD(tmp, MMVM_L2_CNTL3, + L2_CACHE_BIGK_FRAGMENT_SIZE, 6); + } + WREG32_SOC15(MMHUB, 0, regMMVM_L2_CNTL3, tmp); + + tmp = regMMVM_L2_CNTL4_DEFAULT; + tmp = REG_SET_FIELD(tmp, MMVM_L2_CNTL4, VMC_TAP_PDE_REQUEST_PHYSICAL, 0); + tmp = REG_SET_FIELD(tmp, MMVM_L2_CNTL4, VMC_TAP_PTE_REQUEST_PHYSICAL, 0); + WREG32_SOC15(MMHUB, 0, regMMVM_L2_CNTL4, tmp); + + tmp = regMMVM_L2_CNTL5_DEFAULT; + tmp = REG_SET_FIELD(tmp, MMVM_L2_CNTL5, L2_CACHE_SMALLK_FRAGMENT_SIZE, 0); + WREG32_SOC15(GC, 0, regMMVM_L2_CNTL5, tmp); +} + +static void mmhub_v3_0_2_enable_system_domain(struct amdgpu_device *adev) +{ + uint32_t tmp; + + tmp = RREG32_SOC15(MMHUB, 0, regMMVM_CONTEXT0_CNTL); + tmp = REG_SET_FIELD(tmp, MMVM_CONTEXT0_CNTL, ENABLE_CONTEXT, 1); + tmp = REG_SET_FIELD(tmp, MMVM_CONTEXT0_CNTL, PAGE_TABLE_DEPTH, 0); + tmp = REG_SET_FIELD(tmp, MMVM_CONTEXT0_CNTL, + RETRY_PERMISSION_OR_INVALID_PAGE_FAULT, 0); + WREG32_SOC15(MMHUB, 0, regMMVM_CONTEXT0_CNTL, tmp); +} + +static void mmhub_v3_0_2_disable_identity_aperture(struct amdgpu_device *adev) +{ + /* These registers are not accessible to VF-SRIOV. + * The PF will program them instead. + */ + if (amdgpu_sriov_vf(adev)) + return; + + WREG32_SOC15(MMHUB, 0, + regMMVM_L2_CONTEXT1_IDENTITY_APERTURE_LOW_ADDR_LO32, + 0xFFFFFFFF); + WREG32_SOC15(MMHUB, 0, + regMMVM_L2_CONTEXT1_IDENTITY_APERTURE_LOW_ADDR_HI32, + 0x0000000F); + + WREG32_SOC15(MMHUB, 0, + regMMVM_L2_CONTEXT1_IDENTITY_APERTURE_HIGH_ADDR_LO32, 0); + WREG32_SOC15(MMHUB, 0, + regMMVM_L2_CONTEXT1_IDENTITY_APERTURE_HIGH_ADDR_HI32, 0); + + WREG32_SOC15(MMHUB, 0, regMMVM_L2_CONTEXT_IDENTITY_PHYSICAL_OFFSET_LO32, + 0); + WREG32_SOC15(MMHUB, 0, regMMVM_L2_CONTEXT_IDENTITY_PHYSICAL_OFFSET_HI32, + 0); +} + +static void mmhub_v3_0_2_setup_vmid_config(struct amdgpu_device *adev) +{ + struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0]; + int i; + uint32_t tmp; + + for (i = 0; i <= 14; i++) { + tmp = RREG32_SOC15_OFFSET(MMHUB, 0, regMMVM_CONTEXT1_CNTL, i); + tmp = REG_SET_FIELD(tmp, MMVM_CONTEXT1_CNTL, ENABLE_CONTEXT, 1); + tmp = REG_SET_FIELD(tmp, MMVM_CONTEXT1_CNTL, PAGE_TABLE_DEPTH, + adev->vm_manager.num_level); + tmp = REG_SET_FIELD(tmp, MMVM_CONTEXT1_CNTL, + RANGE_PROTECTION_FAULT_ENABLE_DEFAULT, 1); + tmp = REG_SET_FIELD(tmp, MMVM_CONTEXT1_CNTL, + DUMMY_PAGE_PROTECTION_FAULT_ENABLE_DEFAULT, + 1); + tmp = REG_SET_FIELD(tmp, MMVM_CONTEXT1_CNTL, + PDE0_PROTECTION_FAULT_ENABLE_DEFAULT, 1); + tmp = REG_SET_FIELD(tmp, MMVM_CONTEXT1_CNTL, + VALID_PROTECTION_FAULT_ENABLE_DEFAULT, 1); + tmp = REG_SET_FIELD(tmp, MMVM_CONTEXT1_CNTL, + READ_PROTECTION_FAULT_ENABLE_DEFAULT, 1); + tmp = REG_SET_FIELD(tmp, MMVM_CONTEXT1_CNTL, + WRITE_PROTECTION_FAULT_ENABLE_DEFAULT, 1); + tmp = REG_SET_FIELD(tmp, MMVM_CONTEXT1_CNTL, + EXECUTE_PROTECTION_FAULT_ENABLE_DEFAULT, 1); + tmp = REG_SET_FIELD(tmp, MMVM_CONTEXT1_CNTL, + PAGE_TABLE_BLOCK_SIZE, + adev->vm_manager.block_size - 9); + /* Send no-retry XNACK on fault to suppress VM fault storm. */ + tmp = REG_SET_FIELD(tmp, MMVM_CONTEXT1_CNTL, + RETRY_PERMISSION_OR_INVALID_PAGE_FAULT, + !amdgpu_noretry); + WREG32_SOC15_OFFSET(MMHUB, 0, regMMVM_CONTEXT1_CNTL, + i * hub->ctx_distance, tmp); + WREG32_SOC15_OFFSET(MMHUB, 0, regMMVM_CONTEXT1_PAGE_TABLE_START_ADDR_LO32, + i * hub->ctx_addr_distance, 0); + WREG32_SOC15_OFFSET(MMHUB, 0, regMMVM_CONTEXT1_PAGE_TABLE_START_ADDR_HI32, + i * hub->ctx_addr_distance, 0); + WREG32_SOC15_OFFSET(MMHUB, 0, regMMVM_CONTEXT1_PAGE_TABLE_END_ADDR_LO32, + i * hub->ctx_addr_distance, + lower_32_bits(adev->vm_manager.max_pfn - 1)); + WREG32_SOC15_OFFSET(MMHUB, 0, regMMVM_CONTEXT1_PAGE_TABLE_END_ADDR_HI32, + i * hub->ctx_addr_distance, + upper_32_bits(adev->vm_manager.max_pfn - 1)); + } + + hub->vm_cntx_cntl = tmp; +} + +static void mmhub_v3_0_2_program_invalidation(struct amdgpu_device *adev) +{ + struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0]; + unsigned i; + + for (i = 0; i < 18; ++i) { + WREG32_SOC15_OFFSET(MMHUB, 0, regMMVM_INVALIDATE_ENG0_ADDR_RANGE_LO32, + i * hub->eng_addr_distance, 0xffffffff); + WREG32_SOC15_OFFSET(MMHUB, 0, regMMVM_INVALIDATE_ENG0_ADDR_RANGE_HI32, + i * hub->eng_addr_distance, 0x1f); + } +} + +static int mmhub_v3_0_2_gart_enable(struct amdgpu_device *adev) +{ + /* GART Enable. */ + mmhub_v3_0_2_init_gart_aperture_regs(adev); + mmhub_v3_0_2_init_system_aperture_regs(adev); + mmhub_v3_0_2_init_tlb_regs(adev); + mmhub_v3_0_2_init_cache_regs(adev); + + mmhub_v3_0_2_enable_system_domain(adev); + mmhub_v3_0_2_disable_identity_aperture(adev); + mmhub_v3_0_2_setup_vmid_config(adev); + mmhub_v3_0_2_program_invalidation(adev); + + return 0; +} + +static void mmhub_v3_0_2_gart_disable(struct amdgpu_device *adev) +{ + struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0]; + u32 tmp; + u32 i; + + /* Disable all tables */ + for (i = 0; i < 16; i++) + WREG32_SOC15_OFFSET(MMHUB, 0, regMMVM_CONTEXT0_CNTL, + i * hub->ctx_distance, 0); + + /* Setup TLB control */ + tmp = RREG32_SOC15(MMHUB, 0, regMMMC_VM_MX_L1_TLB_CNTL); + tmp = REG_SET_FIELD(tmp, MMMC_VM_MX_L1_TLB_CNTL, ENABLE_L1_TLB, 0); + tmp = REG_SET_FIELD(tmp, MMMC_VM_MX_L1_TLB_CNTL, + ENABLE_ADVANCED_DRIVER_MODEL, 0); + WREG32_SOC15(MMHUB, 0, regMMMC_VM_MX_L1_TLB_CNTL, tmp); + + /* Setup L2 cache */ + tmp = RREG32_SOC15(MMHUB, 0, regMMVM_L2_CNTL); + tmp = REG_SET_FIELD(tmp, MMVM_L2_CNTL, ENABLE_L2_CACHE, 0); + WREG32_SOC15(MMHUB, 0, regMMVM_L2_CNTL, tmp); + WREG32_SOC15(MMHUB, 0, regMMVM_L2_CNTL3, 0); +} + +/** + * mmhub_v3_0_2_set_fault_enable_default - update GART/VM fault handling + * + * @adev: amdgpu_device pointer + * @value: true redirects VM faults to the default page + */ +static void mmhub_v3_0_2_set_fault_enable_default(struct amdgpu_device *adev, bool value) +{ + u32 tmp; + + /* These registers are not accessible to VF-SRIOV. + * The PF will program them instead. + */ + if (amdgpu_sriov_vf(adev)) + return; + + tmp = RREG32_SOC15(MMHUB, 0, regMMVM_L2_PROTECTION_FAULT_CNTL); + tmp = REG_SET_FIELD(tmp, MMVM_L2_PROTECTION_FAULT_CNTL, + RANGE_PROTECTION_FAULT_ENABLE_DEFAULT, value); + tmp = REG_SET_FIELD(tmp, MMVM_L2_PROTECTION_FAULT_CNTL, + PDE0_PROTECTION_FAULT_ENABLE_DEFAULT, value); + tmp = REG_SET_FIELD(tmp, MMVM_L2_PROTECTION_FAULT_CNTL, + PDE1_PROTECTION_FAULT_ENABLE_DEFAULT, value); + tmp = REG_SET_FIELD(tmp, MMVM_L2_PROTECTION_FAULT_CNTL, + PDE2_PROTECTION_FAULT_ENABLE_DEFAULT, value); + tmp = REG_SET_FIELD(tmp, MMVM_L2_PROTECTION_FAULT_CNTL, + TRANSLATE_FURTHER_PROTECTION_FAULT_ENABLE_DEFAULT, + value); + tmp = REG_SET_FIELD(tmp, MMVM_L2_PROTECTION_FAULT_CNTL, + NACK_PROTECTION_FAULT_ENABLE_DEFAULT, value); + tmp = REG_SET_FIELD(tmp, MMVM_L2_PROTECTION_FAULT_CNTL, + DUMMY_PAGE_PROTECTION_FAULT_ENABLE_DEFAULT, value); + tmp = REG_SET_FIELD(tmp, MMVM_L2_PROTECTION_FAULT_CNTL, + VALID_PROTECTION_FAULT_ENABLE_DEFAULT, value); + tmp = REG_SET_FIELD(tmp, MMVM_L2_PROTECTION_FAULT_CNTL, + READ_PROTECTION_FAULT_ENABLE_DEFAULT, value); + tmp = REG_SET_FIELD(tmp, MMVM_L2_PROTECTION_FAULT_CNTL, + WRITE_PROTECTION_FAULT_ENABLE_DEFAULT, value); + tmp = REG_SET_FIELD(tmp, MMVM_L2_PROTECTION_FAULT_CNTL, + EXECUTE_PROTECTION_FAULT_ENABLE_DEFAULT, value); + if (!value) { + tmp = REG_SET_FIELD(tmp, MMVM_L2_PROTECTION_FAULT_CNTL, + CRASH_ON_NO_RETRY_FAULT, 1); + tmp = REG_SET_FIELD(tmp, MMVM_L2_PROTECTION_FAULT_CNTL, + CRASH_ON_RETRY_FAULT, 1); + } + WREG32_SOC15(MMHUB, 0, regMMVM_L2_PROTECTION_FAULT_CNTL, tmp); +} + +static const struct amdgpu_vmhub_funcs mmhub_v3_0_2_vmhub_funcs = { + .print_l2_protection_fault_status = mmhub_v3_0_2_print_l2_protection_fault_status, + .get_invalidate_req = mmhub_v3_0_2_get_invalidate_req, +}; + +static void mmhub_v3_0_2_init(struct amdgpu_device *adev) +{ + struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0]; + + hub->ctx0_ptb_addr_lo32 = + SOC15_REG_OFFSET(MMHUB, 0, + regMMVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32); + hub->ctx0_ptb_addr_hi32 = + SOC15_REG_OFFSET(MMHUB, 0, + regMMVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_HI32); + hub->vm_inv_eng0_sem = + SOC15_REG_OFFSET(MMHUB, 0, regMMVM_INVALIDATE_ENG0_SEM); + hub->vm_inv_eng0_req = + SOC15_REG_OFFSET(MMHUB, 0, regMMVM_INVALIDATE_ENG0_REQ); + hub->vm_inv_eng0_ack = + SOC15_REG_OFFSET(MMHUB, 0, regMMVM_INVALIDATE_ENG0_ACK); + hub->vm_context0_cntl = + SOC15_REG_OFFSET(MMHUB, 0, regMMVM_CONTEXT0_CNTL); + hub->vm_l2_pro_fault_status = + SOC15_REG_OFFSET(MMHUB, 0, regMMVM_L2_PROTECTION_FAULT_STATUS); + hub->vm_l2_pro_fault_cntl = + SOC15_REG_OFFSET(MMHUB, 0, regMMVM_L2_PROTECTION_FAULT_CNTL); + + hub->ctx_distance = regMMVM_CONTEXT1_CNTL - regMMVM_CONTEXT0_CNTL; + hub->ctx_addr_distance = regMMVM_CONTEXT1_PAGE_TABLE_BASE_ADDR_LO32 - + regMMVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32; + hub->eng_distance = regMMVM_INVALIDATE_ENG1_REQ - + regMMVM_INVALIDATE_ENG0_REQ; + hub->eng_addr_distance = regMMVM_INVALIDATE_ENG1_ADDR_RANGE_LO32 - + regMMVM_INVALIDATE_ENG0_ADDR_RANGE_LO32; + + hub->vm_cntx_cntl_vm_fault = MMVM_CONTEXT1_CNTL__RANGE_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK | + MMVM_CONTEXT1_CNTL__DUMMY_PAGE_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK | + MMVM_CONTEXT1_CNTL__PDE0_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK | + MMVM_CONTEXT1_CNTL__VALID_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK | + MMVM_CONTEXT1_CNTL__READ_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK | + MMVM_CONTEXT1_CNTL__WRITE_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK | + MMVM_CONTEXT1_CNTL__EXECUTE_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK; + + hub->vm_l2_bank_select_reserved_cid2 = + SOC15_REG_OFFSET(MMHUB, 0, regMMVM_L2_BANK_SELECT_RESERVED_CID2); + + hub->vmhub_funcs = &mmhub_v3_0_2_vmhub_funcs; +} + +static u64 mmhub_v3_0_2_get_fb_location(struct amdgpu_device *adev) +{ + u64 base; + + base = RREG32_SOC15(MMHUB, 0, regMMMC_VM_FB_LOCATION_BASE); + base &= MMMC_VM_FB_LOCATION_BASE__FB_BASE_MASK; + base <<= 24; + + return base; +} + +static u64 mmhub_v3_0_2_get_mc_fb_offset(struct amdgpu_device *adev) +{ + return (u64)RREG32_SOC15(MMHUB, 0, regMMMC_VM_FB_OFFSET) << 24; +} + +static void mmhub_v3_0_2_update_medium_grain_clock_gating(struct amdgpu_device *adev, + bool enable) +{ + //TODO +} + +static void mmhub_v3_0_2_update_medium_grain_light_sleep(struct amdgpu_device *adev, + bool enable) +{ + //TODO +} + +static int mmhub_v3_0_2_set_clockgating(struct amdgpu_device *adev, + enum amd_clockgating_state state) +{ + if (amdgpu_sriov_vf(adev)) + return 0; + + mmhub_v3_0_2_update_medium_grain_clock_gating(adev, + state == AMD_CG_STATE_GATE); + mmhub_v3_0_2_update_medium_grain_light_sleep(adev, + state == AMD_CG_STATE_GATE); + return 0; +} + +static void mmhub_v3_0_2_get_clockgating(struct amdgpu_device *adev, u64 *flags) +{ + //TODO +} + +const struct amdgpu_mmhub_funcs mmhub_v3_0_2_funcs = { + .init = mmhub_v3_0_2_init, + .get_fb_location = mmhub_v3_0_2_get_fb_location, + .get_mc_fb_offset = mmhub_v3_0_2_get_mc_fb_offset, + .gart_enable = mmhub_v3_0_2_gart_enable, + .set_fault_enable_default = mmhub_v3_0_2_set_fault_enable_default, + .gart_disable = mmhub_v3_0_2_gart_disable, + .set_clockgating = mmhub_v3_0_2_set_clockgating, + .get_clockgating = mmhub_v3_0_2_get_clockgating, + .setup_vm_pt_regs = mmhub_v3_0_2_setup_vm_pt_regs, +}; diff --git a/drivers/gpu/drm/amd/amdgpu/mmhub_v3_0_2.h b/drivers/gpu/drm/amd/amdgpu/mmhub_v3_0_2.h new file mode 100644 index 000000000000..23ad7b156cdb --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/mmhub_v3_0_2.h @@ -0,0 +1,28 @@ +/* + * Copyright 2022 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ +#ifndef __MMHUB_V3_0_2_H__ +#define __MMHUB_V3_0_2_H__ + +extern const struct amdgpu_mmhub_funcs mmhub_v3_0_2_funcs; + +#endif diff --git a/drivers/gpu/drm/amd/amdgpu/mmhub_v9_4.c b/drivers/gpu/drm/amd/amdgpu/mmhub_v9_4.c index ff49eeaf7882..445cb06b9d26 100644 --- a/drivers/gpu/drm/amd/amdgpu/mmhub_v9_4.c +++ b/drivers/gpu/drm/amd/amdgpu/mmhub_v9_4.c @@ -295,9 +295,17 @@ static void mmhub_v9_4_disable_identity_aperture(struct amdgpu_device *adev, static void mmhub_v9_4_setup_vmid_config(struct amdgpu_device *adev, int hubid) { struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0]; + unsigned int num_level, block_size; uint32_t tmp; int i; + num_level = adev->vm_manager.num_level; + block_size = adev->vm_manager.block_size; + if (adev->gmc.translate_further) + num_level -= 1; + else + block_size -= 9; + for (i = 0; i <= 14; i++) { tmp = RREG32_SOC15_OFFSET(MMHUB, 0, mmVML2VC0_VM_CONTEXT1_CNTL, hubid * MMHUB_INSTANCE_REGISTER_OFFSET + i); @@ -305,7 +313,7 @@ static void mmhub_v9_4_setup_vmid_config(struct amdgpu_device *adev, int hubid) ENABLE_CONTEXT, 1); tmp = REG_SET_FIELD(tmp, VML2VC0_VM_CONTEXT1_CNTL, PAGE_TABLE_DEPTH, - adev->vm_manager.num_level); + num_level); tmp = REG_SET_FIELD(tmp, VML2VC0_VM_CONTEXT1_CNTL, RANGE_PROTECTION_FAULT_ENABLE_DEFAULT, 1); tmp = REG_SET_FIELD(tmp, VML2VC0_VM_CONTEXT1_CNTL, @@ -323,7 +331,7 @@ static void mmhub_v9_4_setup_vmid_config(struct amdgpu_device *adev, int hubid) EXECUTE_PROTECTION_FAULT_ENABLE_DEFAULT, 1); tmp = REG_SET_FIELD(tmp, VML2VC0_VM_CONTEXT1_CNTL, PAGE_TABLE_BLOCK_SIZE, - adev->vm_manager.block_size - 9); + block_size); /* Send no-retry XNACK on fault to suppress VM fault storm. */ tmp = REG_SET_FIELD(tmp, VML2VC0_VM_CONTEXT1_CNTL, RETRY_PERMISSION_OR_INVALID_PAGE_FAULT, @@ -647,7 +655,7 @@ static int mmhub_v9_4_set_clockgating(struct amdgpu_device *adev, return 0; } -static void mmhub_v9_4_get_clockgating(struct amdgpu_device *adev, u32 *flags) +static void mmhub_v9_4_get_clockgating(struct amdgpu_device *adev, u64 *flags) { int data, data1; @@ -1655,14 +1663,18 @@ static void mmhub_v9_4_query_ras_error_status(struct amdgpu_device *adev) } } -const struct amdgpu_mmhub_ras_funcs mmhub_v9_4_ras_funcs = { - .ras_late_init = amdgpu_mmhub_ras_late_init, - .ras_fini = amdgpu_mmhub_ras_fini, +const struct amdgpu_ras_block_hw_ops mmhub_v9_4_ras_hw_ops = { .query_ras_error_count = mmhub_v9_4_query_ras_error_count, .reset_ras_error_count = mmhub_v9_4_reset_ras_error_count, .query_ras_error_status = mmhub_v9_4_query_ras_error_status, }; +struct amdgpu_mmhub_ras mmhub_v9_4_ras = { + .ras_block = { + .hw_ops = &mmhub_v9_4_ras_hw_ops, + }, +}; + const struct amdgpu_mmhub_funcs mmhub_v9_4_funcs = { .get_fb_location = mmhub_v9_4_get_fb_location, .init = mmhub_v9_4_init, diff --git a/drivers/gpu/drm/amd/amdgpu/mmhub_v9_4.h b/drivers/gpu/drm/amd/amdgpu/mmhub_v9_4.h index 90436efa92ef..a48329d95f71 100644 --- a/drivers/gpu/drm/amd/amdgpu/mmhub_v9_4.h +++ b/drivers/gpu/drm/amd/amdgpu/mmhub_v9_4.h @@ -24,6 +24,6 @@ #define __MMHUB_V9_4_H__ extern const struct amdgpu_mmhub_funcs mmhub_v9_4_funcs; -extern const struct amdgpu_mmhub_ras_funcs mmhub_v9_4_ras_funcs; +extern struct amdgpu_mmhub_ras mmhub_v9_4_ras; #endif diff --git a/drivers/gpu/drm/amd/amdgpu/mmsch_v4_0.h b/drivers/gpu/drm/amd/amdgpu/mmsch_v4_0.h new file mode 100644 index 000000000000..f772bb499f3e --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/mmsch_v4_0.h @@ -0,0 +1,140 @@ +/* + * Copyright 2022 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#ifndef __MMSCH_V4_0_H__ +#define __MMSCH_V4_0_H__ + +#include "amdgpu_vcn.h" + +#define MMSCH_VERSION_MAJOR 4 +#define MMSCH_VERSION_MINOR 0 +#define MMSCH_VERSION (MMSCH_VERSION_MAJOR << 16 | MMSCH_VERSION_MINOR) + +#define RB_ENABLED (1 << 0) +#define RB4_ENABLED (1 << 1) +#define MMSCH_DOORBELL_OFFSET 0x8 + +#define MMSCH_VF_ENGINE_STATUS__PASS 0x1 + +#define MMSCH_VF_MAILBOX_RESP__OK 0x1 +#define MMSCH_VF_MAILBOX_RESP__INCOMPLETE 0x2 + +enum mmsch_v4_0_command_type { + MMSCH_COMMAND__DIRECT_REG_WRITE = 0, + MMSCH_COMMAND__DIRECT_REG_POLLING = 2, + MMSCH_COMMAND__DIRECT_REG_READ_MODIFY_WRITE = 3, + MMSCH_COMMAND__INDIRECT_REG_WRITE = 8, + MMSCH_COMMAND__END = 0xf +}; + +struct mmsch_v4_0_table_info { + uint32_t init_status; + uint32_t table_offset; + uint32_t table_size; +}; + +struct mmsch_v4_0_init_header { + uint32_t version; + uint32_t total_size; + struct mmsch_v4_0_table_info inst[AMDGPU_MAX_VCN_INSTANCES]; + struct mmsch_v4_0_table_info jpegdec; +}; + +struct mmsch_v4_0_cmd_direct_reg_header { + uint32_t reg_offset : 28; + uint32_t command_type : 4; +}; + +struct mmsch_v4_0_cmd_indirect_reg_header { + uint32_t reg_offset : 20; + uint32_t reg_idx_space : 8; + uint32_t command_type : 4; +}; + +struct mmsch_v4_0_cmd_direct_write { + struct mmsch_v4_0_cmd_direct_reg_header cmd_header; + uint32_t reg_value; +}; + +struct mmsch_v4_0_cmd_direct_read_modify_write { + struct mmsch_v4_0_cmd_direct_reg_header cmd_header; + uint32_t write_data; + uint32_t mask_value; +}; + +struct mmsch_v4_0_cmd_direct_polling { + struct mmsch_v4_0_cmd_direct_reg_header cmd_header; + uint32_t mask_value; + uint32_t wait_value; +}; + +struct mmsch_v4_0_cmd_end { + struct mmsch_v4_0_cmd_direct_reg_header cmd_header; +}; + +struct mmsch_v4_0_cmd_indirect_write { + struct mmsch_v4_0_cmd_indirect_reg_header cmd_header; + uint32_t reg_value; +}; + +#define MMSCH_V4_0_INSERT_DIRECT_RD_MOD_WT(reg, mask, data) { \ + size = sizeof(struct mmsch_v4_0_cmd_direct_read_modify_write); \ + size_dw = size / 4; \ + direct_rd_mod_wt.cmd_header.reg_offset = reg; \ + direct_rd_mod_wt.mask_value = mask; \ + direct_rd_mod_wt.write_data = data; \ + memcpy((void *)table_loc, &direct_rd_mod_wt, size); \ + table_loc += size_dw; \ + table_size += size_dw; \ +} + +#define MMSCH_V4_0_INSERT_DIRECT_WT(reg, value) { \ + size = sizeof(struct mmsch_v4_0_cmd_direct_write); \ + size_dw = size / 4; \ + direct_wt.cmd_header.reg_offset = reg; \ + direct_wt.reg_value = value; \ + memcpy((void *)table_loc, &direct_wt, size); \ + table_loc += size_dw; \ + table_size += size_dw; \ +} + +#define MMSCH_V4_0_INSERT_DIRECT_POLL(reg, mask, wait) { \ + size = sizeof(struct mmsch_v4_0_cmd_direct_polling); \ + size_dw = size / 4; \ + direct_poll.cmd_header.reg_offset = reg; \ + direct_poll.mask_value = mask; \ + direct_poll.wait_value = wait; \ + memcpy((void *)table_loc, &direct_poll, size); \ + table_loc += size_dw; \ + table_size += size_dw; \ +} + +#define MMSCH_V4_0_INSERT_END() { \ + size = sizeof(struct mmsch_v4_0_cmd_end); \ + size_dw = size / 4; \ + memcpy((void *)table_loc, &end, size); \ + table_loc += size_dw; \ + table_size += size_dw; \ +} + +#endif diff --git a/drivers/gpu/drm/amd/amdgpu/mxgpu_ai.c b/drivers/gpu/drm/amd/amdgpu/mxgpu_ai.c index 0077e738db31..12906ba74462 100644 --- a/drivers/gpu/drm/amd/amdgpu/mxgpu_ai.c +++ b/drivers/gpu/drm/amd/amdgpu/mxgpu_ai.c @@ -32,6 +32,8 @@ #include "soc15_common.h" #include "mxgpu_ai.h" +#include "amdgpu_reset.h" + static void xgpu_ai_mailbox_send_ack(struct amdgpu_device *adev) { WREG8(AI_MAIBOX_CONTROL_RCV_OFFSET_BYTE, 2); @@ -180,6 +182,11 @@ static int xgpu_ai_send_access_requests(struct amdgpu_device *adev, RREG32_NO_KIQ(SOC15_REG_OFFSET(NBIO, 0, mmBIF_BX_PF0_MAILBOX_MSGBUF_RCV_DW2)); } + } else if (req == IDH_REQ_GPU_INIT_DATA){ + /* Dummy REQ_GPU_INIT_DATA handling */ + r = xgpu_ai_poll_msg(adev, IDH_REQ_GPU_INIT_DATA_READY); + /* version set to 0 since dummy */ + adev->virt.req_init_data_ver = 0; } return 0; @@ -252,10 +259,10 @@ static void xgpu_ai_mailbox_flr_work(struct work_struct *work) * otherwise the mailbox msg will be ruined/reseted by * the VF FLR. */ - if (atomic_cmpxchg(&adev->in_gpu_reset, 0, 1) != 0) + if (atomic_cmpxchg(&adev->reset_domain->in_gpu_reset, 0, 1) != 0) return; - down_write(&adev->reset_sem); + down_write(&adev->reset_domain->sem); amdgpu_virt_fini_data_exchange(adev); @@ -270,14 +277,22 @@ static void xgpu_ai_mailbox_flr_work(struct work_struct *work) } while (timeout > 1); flr_done: - atomic_set(&adev->in_gpu_reset, 0); - up_write(&adev->reset_sem); + atomic_set(&adev->reset_domain->in_gpu_reset, 0); + up_write(&adev->reset_domain->sem); /* Trigger recovery for world switch failure if no TDR */ if (amdgpu_device_should_recover_gpu(adev) && (!amdgpu_device_has_job_running(adev) || - adev->sdma_timeout == MAX_SCHEDULE_TIMEOUT)) - amdgpu_device_gpu_recover(adev, NULL); + adev->sdma_timeout == MAX_SCHEDULE_TIMEOUT)) { + struct amdgpu_reset_context reset_context; + memset(&reset_context, 0, sizeof(reset_context)); + + reset_context.method = AMD_RESET_METHOD_NONE; + reset_context.reset_req_dev = adev; + clear_bit(AMDGPU_NEED_FULL_RESET, &reset_context.flags); + + amdgpu_device_gpu_recover(adev, NULL, &reset_context); + } } static int xgpu_ai_set_mailbox_rcv_irq(struct amdgpu_device *adev, @@ -302,8 +317,11 @@ static int xgpu_ai_mailbox_rcv_irq(struct amdgpu_device *adev, switch (event) { case IDH_FLR_NOTIFICATION: - if (amdgpu_sriov_runtime(adev)) - schedule_work(&adev->virt.flr_work); + if (amdgpu_sriov_runtime(adev) && !amdgpu_in_reset(adev)) + WARN_ONCE(!amdgpu_reset_domain_schedule(adev->reset_domain, + &adev->virt.flr_work), + "Failed to queue work! at %s", + __func__); break; case IDH_QUERY_ALIVE: xgpu_ai_mailbox_send_ack(adev); @@ -381,10 +399,16 @@ void xgpu_ai_mailbox_put_irq(struct amdgpu_device *adev) amdgpu_irq_put(adev, &adev->virt.rcv_irq, 0); } +static int xgpu_ai_request_init_data(struct amdgpu_device *adev) +{ + return xgpu_ai_send_access_requests(adev, IDH_REQ_GPU_INIT_DATA); +} + const struct amdgpu_virt_ops xgpu_ai_virt_ops = { .req_full_gpu = xgpu_ai_request_full_gpu_access, .rel_full_gpu = xgpu_ai_release_full_gpu_access, .reset_gpu = xgpu_ai_request_reset, .wait_reset = NULL, .trans_msg = xgpu_ai_mailbox_trans_msg, + .req_init_data = xgpu_ai_request_init_data, }; diff --git a/drivers/gpu/drm/amd/amdgpu/mxgpu_ai.h b/drivers/gpu/drm/amd/amdgpu/mxgpu_ai.h index f9aa4d0bb638..fa7e13e0459e 100644 --- a/drivers/gpu/drm/amd/amdgpu/mxgpu_ai.h +++ b/drivers/gpu/drm/amd/amdgpu/mxgpu_ai.h @@ -35,6 +35,7 @@ enum idh_request { IDH_REQ_GPU_FINI_ACCESS, IDH_REL_GPU_FINI_ACCESS, IDH_REQ_GPU_RESET_ACCESS, + IDH_REQ_GPU_INIT_DATA, IDH_LOG_VF_ERROR = 200, IDH_READY_TO_RESET = 201, @@ -48,6 +49,7 @@ enum idh_event { IDH_SUCCESS, IDH_FAIL, IDH_QUERY_ALIVE, + IDH_REQ_GPU_INIT_DATA_READY, IDH_TEXT_MESSAGE = 255, }; diff --git a/drivers/gpu/drm/amd/amdgpu/mxgpu_nv.c b/drivers/gpu/drm/amd/amdgpu/mxgpu_nv.c index 477d0dde19c5..e07757eea7ad 100644 --- a/drivers/gpu/drm/amd/amdgpu/mxgpu_nv.c +++ b/drivers/gpu/drm/amd/amdgpu/mxgpu_nv.c @@ -31,6 +31,8 @@ #include "soc15_common.h" #include "mxgpu_nv.h" +#include "amdgpu_reset.h" + static void xgpu_nv_mailbox_send_ack(struct amdgpu_device *adev) { WREG8(NV_MAIBOX_CONTROL_RCV_OFFSET_BYTE, 2); @@ -281,10 +283,10 @@ static void xgpu_nv_mailbox_flr_work(struct work_struct *work) * otherwise the mailbox msg will be ruined/reseted by * the VF FLR. */ - if (atomic_cmpxchg(&adev->in_gpu_reset, 0, 1) != 0) + if (atomic_cmpxchg(&adev->reset_domain->in_gpu_reset, 0, 1) != 0) return; - down_write(&adev->reset_sem); + down_write(&adev->reset_domain->sem); amdgpu_virt_fini_data_exchange(adev); @@ -299,8 +301,8 @@ static void xgpu_nv_mailbox_flr_work(struct work_struct *work) } while (timeout > 1); flr_done: - atomic_set(&adev->in_gpu_reset, 0); - up_write(&adev->reset_sem); + atomic_set(&adev->reset_domain->in_gpu_reset, 0); + up_write(&adev->reset_domain->sem); /* Trigger recovery for world switch failure if no TDR */ if (amdgpu_device_should_recover_gpu(adev) @@ -308,8 +310,16 @@ flr_done: adev->sdma_timeout == MAX_SCHEDULE_TIMEOUT || adev->gfx_timeout == MAX_SCHEDULE_TIMEOUT || adev->compute_timeout == MAX_SCHEDULE_TIMEOUT || - adev->video_timeout == MAX_SCHEDULE_TIMEOUT)) - amdgpu_device_gpu_recover(adev, NULL); + adev->video_timeout == MAX_SCHEDULE_TIMEOUT)) { + struct amdgpu_reset_context reset_context; + memset(&reset_context, 0, sizeof(reset_context)); + + reset_context.method = AMD_RESET_METHOD_NONE; + reset_context.reset_req_dev = adev; + clear_bit(AMDGPU_NEED_FULL_RESET, &reset_context.flags); + + amdgpu_device_gpu_recover(adev, NULL, &reset_context); + } } static int xgpu_nv_set_mailbox_rcv_irq(struct amdgpu_device *adev, @@ -337,8 +347,11 @@ static int xgpu_nv_mailbox_rcv_irq(struct amdgpu_device *adev, switch (event) { case IDH_FLR_NOTIFICATION: - if (amdgpu_sriov_runtime(adev)) - schedule_work(&adev->virt.flr_work); + if (amdgpu_sriov_runtime(adev) && !amdgpu_in_reset(adev)) + WARN_ONCE(!amdgpu_reset_domain_schedule(adev->reset_domain, + &adev->virt.flr_work), + "Failed to queue work! at %s", + __func__); break; /* READY_TO_ACCESS_GPU is fetched by kernel polling, IRQ can ignore * it byfar since that polling thread will handle it, diff --git a/drivers/gpu/drm/amd/amdgpu/mxgpu_vi.c b/drivers/gpu/drm/amd/amdgpu/mxgpu_vi.c index aef9d059ae52..288c414babdf 100644 --- a/drivers/gpu/drm/amd/amdgpu/mxgpu_vi.c +++ b/drivers/gpu/drm/amd/amdgpu/mxgpu_vi.c @@ -42,6 +42,8 @@ #include "smu/smu_7_1_3_d.h" #include "mxgpu_vi.h" +#include "amdgpu_reset.h" + /* VI golden setting */ static const u32 xgpu_fiji_mgcg_cgcg_init[] = { mmRLC_CGTT_MGCG_OVERRIDE, 0xffffffff, 0xffffffff, @@ -520,8 +522,16 @@ static void xgpu_vi_mailbox_flr_work(struct work_struct *work) } /* Trigger recovery due to world switch failure */ - if (amdgpu_device_should_recover_gpu(adev)) - amdgpu_device_gpu_recover(adev, NULL); + if (amdgpu_device_should_recover_gpu(adev)) { + struct amdgpu_reset_context reset_context; + memset(&reset_context, 0, sizeof(reset_context)); + + reset_context.method = AMD_RESET_METHOD_NONE; + reset_context.reset_req_dev = adev; + clear_bit(AMDGPU_NEED_FULL_RESET, &reset_context.flags); + + amdgpu_device_gpu_recover(adev, NULL, &reset_context); + } } static int xgpu_vi_set_mailbox_rcv_irq(struct amdgpu_device *adev, @@ -544,14 +554,17 @@ static int xgpu_vi_mailbox_rcv_irq(struct amdgpu_device *adev, { int r; - /* trigger gpu-reset by hypervisor only if TDR disbaled */ + /* trigger gpu-reset by hypervisor only if TDR disabled */ if (!amdgpu_gpu_recovery) { /* see what event we get */ r = xgpu_vi_mailbox_rcv_msg(adev, IDH_FLR_NOTIFICATION); /* only handle FLR_NOTIFY now */ - if (!r) - schedule_work(&adev->virt.flr_work); + if (!r && !amdgpu_in_reset(adev)) + WARN_ONCE(!amdgpu_reset_domain_schedule(adev->reset_domain, + &adev->virt.flr_work), + "Failed to queue work! at %s", + __func__); } return 0; diff --git a/drivers/gpu/drm/amd/amdgpu/navi10_ih.c b/drivers/gpu/drm/amd/amdgpu/navi10_ih.c index 8ce5b8ca1fd7..eec13cb5bf75 100644 --- a/drivers/gpu/drm/amd/amdgpu/navi10_ih.c +++ b/drivers/gpu/drm/amd/amdgpu/navi10_ih.c @@ -409,9 +409,11 @@ static u32 navi10_ih_get_wptr(struct amdgpu_device *adev, u32 wptr, tmp; struct amdgpu_ih_regs *ih_regs; - if (ih == &adev->irq.ih) { + if (ih == &adev->irq.ih || ih == &adev->irq.ih_soft) { /* Only ring0 supports writeback. On other rings fall back * to register-based code with overflow checking below. + * ih_soft ring doesn't have any backing hardware registers, + * update wptr and return. */ wptr = le32_to_cpu(*ih->wptr_cpu); @@ -483,6 +485,9 @@ static void navi10_ih_set_rptr(struct amdgpu_device *adev, { struct amdgpu_ih_regs *ih_regs; + if (ih == &adev->irq.ih_soft) + return; + if (ih->use_doorbell) { /* XXX check if swapping is necessary on BE */ *ih->rptr_cpu = ih->rptr; @@ -593,14 +598,9 @@ static int navi10_ih_sw_fini(void *handle) static int navi10_ih_hw_init(void *handle) { - int r; struct amdgpu_device *adev = (struct amdgpu_device *)handle; - r = navi10_ih_irq_init(adev); - if (r) - return r; - - return 0; + return navi10_ih_irq_init(adev); } static int navi10_ih_hw_fini(void *handle) @@ -685,7 +685,7 @@ static int navi10_ih_set_powergating_state(void *handle, return 0; } -static void navi10_ih_get_clockgating_state(void *handle, u32 *flags) +static void navi10_ih_get_clockgating_state(void *handle, u64 *flags) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; diff --git a/drivers/gpu/drm/amd/amdgpu/nbio_v2_3.c b/drivers/gpu/drm/amd/amdgpu/nbio_v2_3.c index ee7cab37dfd5..aa761ff3a5fa 100644 --- a/drivers/gpu/drm/amd/amdgpu/nbio_v2_3.c +++ b/drivers/gpu/drm/amd/amdgpu/nbio_v2_3.c @@ -278,7 +278,7 @@ static void nbio_v2_3_update_medium_grain_light_sleep(struct amdgpu_device *adev } static void nbio_v2_3_get_clockgating_state(struct amdgpu_device *adev, - u32 *flags) + u64 *flags) { int data; @@ -328,27 +328,6 @@ const struct nbio_hdp_flush_reg nbio_v2_3_hdp_flush_reg = { .ref_and_mask_sdma1 = BIF_BX_PF_GPU_HDP_FLUSH_DONE__SDMA1_MASK, }; -const struct nbio_hdp_flush_reg nbio_v2_3_hdp_flush_reg_sc = { - .ref_and_mask_cp0 = BIF_BX_PF_GPU_HDP_FLUSH_DONE__CP0_MASK, - .ref_and_mask_cp1 = BIF_BX_PF_GPU_HDP_FLUSH_DONE__CP1_MASK, - .ref_and_mask_cp2 = BIF_BX_PF_GPU_HDP_FLUSH_DONE__CP2_MASK, - .ref_and_mask_cp3 = BIF_BX_PF_GPU_HDP_FLUSH_DONE__CP3_MASK, - .ref_and_mask_cp4 = BIF_BX_PF_GPU_HDP_FLUSH_DONE__CP4_MASK, - .ref_and_mask_cp5 = BIF_BX_PF_GPU_HDP_FLUSH_DONE__CP5_MASK, - .ref_and_mask_cp6 = BIF_BX_PF_GPU_HDP_FLUSH_DONE__CP6_MASK, - .ref_and_mask_cp7 = BIF_BX_PF_GPU_HDP_FLUSH_DONE__CP7_MASK, - .ref_and_mask_cp8 = BIF_BX_PF_GPU_HDP_FLUSH_DONE__CP8_MASK, - .ref_and_mask_cp9 = BIF_BX_PF_GPU_HDP_FLUSH_DONE__CP9_MASK, - .ref_and_mask_sdma0 = GPU_HDP_FLUSH_DONE__RSVD_ENG1_MASK, - .ref_and_mask_sdma1 = GPU_HDP_FLUSH_DONE__RSVD_ENG2_MASK, - .ref_and_mask_sdma2 = GPU_HDP_FLUSH_DONE__RSVD_ENG3_MASK, - .ref_and_mask_sdma3 = GPU_HDP_FLUSH_DONE__RSVD_ENG4_MASK, - .ref_and_mask_sdma4 = GPU_HDP_FLUSH_DONE__RSVD_ENG5_MASK, - .ref_and_mask_sdma5 = GPU_HDP_FLUSH_DONE__RSVD_ENG6_MASK, - .ref_and_mask_sdma6 = GPU_HDP_FLUSH_DONE__RSVD_ENG7_MASK, - .ref_and_mask_sdma7 = GPU_HDP_FLUSH_DONE__RSVD_ENG8_MASK, -}; - static void nbio_v2_3_init_registers(struct amdgpu_device *adev) { uint32_t def, data; @@ -401,6 +380,7 @@ static void nbio_v2_3_enable_aspm(struct amdgpu_device *adev, WREG32_PCIE(smnPCIE_LC_CNTL, data); } +#ifdef CONFIG_PCIEASPM static void nbio_v2_3_program_ltr(struct amdgpu_device *adev) { uint32_t def, data; @@ -422,9 +402,11 @@ static void nbio_v2_3_program_ltr(struct amdgpu_device *adev) if (def != data) WREG32_PCIE(smnBIF_CFG_DEV0_EPF0_DEVICE_CNTL2, data); } +#endif static void nbio_v2_3_program_aspm(struct amdgpu_device *adev) { +#ifdef CONFIG_PCIEASPM uint32_t def, data; def = data = RREG32_PCIE(smnPCIE_LC_CNTL); @@ -480,7 +462,10 @@ static void nbio_v2_3_program_aspm(struct amdgpu_device *adev) if (def != data) WREG32_PCIE(smnPCIE_LC_CNTL6, data); - nbio_v2_3_program_ltr(adev); + /* Don't bother about LTR if LTR is not enabled + * in the path */ + if (adev->pdev->ltr_path) + nbio_v2_3_program_ltr(adev); def = data = RREG32_SOC15(NBIO, 0, mmRCC_BIF_STRAP3); data |= 0x5DE0 << RCC_BIF_STRAP3__STRAP_VLINK_ASPM_IDLE_TIMER__SHIFT; @@ -504,6 +489,7 @@ static void nbio_v2_3_program_aspm(struct amdgpu_device *adev) data &= ~PCIE_LC_CNTL3__LC_DSC_DONT_ENTER_L23_AFTER_PME_ACK_MASK; if (def != data) WREG32_PCIE(smnPCIE_LC_CNTL3, data); +#endif } static void nbio_v2_3_apply_lc_spc_mode_wa(struct amdgpu_device *adev) @@ -547,7 +533,7 @@ static void nbio_v2_3_clear_doorbell_interrupt(struct amdgpu_device *adev) { uint32_t reg, reg_data; - if (adev->asic_type != CHIP_SIENNA_CICHLID) + if (adev->ip_versions[NBIO_HWIP][0] != IP_VERSION(3, 3, 0)) return; reg = RREG32_SOC15(NBIO, 0, mmBIF_RB_CNTL); diff --git a/drivers/gpu/drm/amd/amdgpu/nbio_v2_3.h b/drivers/gpu/drm/amd/amdgpu/nbio_v2_3.h index 6074dd3a1ed8..a43b60acf7f6 100644 --- a/drivers/gpu/drm/amd/amdgpu/nbio_v2_3.h +++ b/drivers/gpu/drm/amd/amdgpu/nbio_v2_3.h @@ -27,7 +27,6 @@ #include "soc15_common.h" extern const struct nbio_hdp_flush_reg nbio_v2_3_hdp_flush_reg; -extern const struct nbio_hdp_flush_reg nbio_v2_3_hdp_flush_reg_sc; extern const struct amdgpu_nbio_funcs nbio_v2_3_funcs; #endif diff --git a/drivers/gpu/drm/amd/amdgpu/nbio_v4_3.c b/drivers/gpu/drm/amd/amdgpu/nbio_v4_3.c new file mode 100644 index 000000000000..15eb3658d70e --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/nbio_v4_3.c @@ -0,0 +1,534 @@ +/* + * Copyright 2021 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ +#include "amdgpu.h" +#include "amdgpu_atombios.h" +#include "nbio_v4_3.h" + +#include "nbio/nbio_4_3_0_offset.h" +#include "nbio/nbio_4_3_0_sh_mask.h" +#include <uapi/linux/kfd_ioctl.h> + +static void nbio_v4_3_remap_hdp_registers(struct amdgpu_device *adev) +{ + WREG32_SOC15(NBIO, 0, regBIF_BX0_REMAP_HDP_MEM_FLUSH_CNTL, + adev->rmmio_remap.reg_offset + KFD_MMIO_REMAP_HDP_MEM_FLUSH_CNTL); + WREG32_SOC15(NBIO, 0, regBIF_BX0_REMAP_HDP_REG_FLUSH_CNTL, + adev->rmmio_remap.reg_offset + KFD_MMIO_REMAP_HDP_REG_FLUSH_CNTL); +} + +static u32 nbio_v4_3_get_rev_id(struct amdgpu_device *adev) +{ + u32 tmp = RREG32_SOC15(NBIO, 0, regRCC_STRAP0_RCC_DEV0_EPF0_STRAP0); + + tmp &= RCC_STRAP0_RCC_DEV0_EPF0_STRAP0__STRAP_ATI_REV_ID_DEV0_F0_MASK; + tmp >>= RCC_STRAP0_RCC_DEV0_EPF0_STRAP0__STRAP_ATI_REV_ID_DEV0_F0__SHIFT; + + return tmp; +} + +static void nbio_v4_3_mc_access_enable(struct amdgpu_device *adev, bool enable) +{ + if (enable) + WREG32_SOC15(NBIO, 0, regBIF_BX0_BIF_FB_EN, + BIF_BX0_BIF_FB_EN__FB_READ_EN_MASK | + BIF_BX0_BIF_FB_EN__FB_WRITE_EN_MASK); + else + WREG32_SOC15(NBIO, 0, regBIF_BX0_BIF_FB_EN, 0); +} + +static u32 nbio_v4_3_get_memsize(struct amdgpu_device *adev) +{ + return RREG32_SOC15(NBIO, 0, regRCC_DEV0_EPF0_RCC_CONFIG_MEMSIZE); +} + +static void nbio_v4_3_sdma_doorbell_range(struct amdgpu_device *adev, int instance, + bool use_doorbell, int doorbell_index, + int doorbell_size) +{ + if (instance == 0) { + u32 doorbell_range = RREG32_SOC15(NBIO, 0, regS2A_DOORBELL_ENTRY_2_CTRL); + + if (use_doorbell) { + doorbell_range = REG_SET_FIELD(doorbell_range, + S2A_DOORBELL_ENTRY_2_CTRL, + S2A_DOORBELL_PORT2_ENABLE, + 0x1); + doorbell_range = REG_SET_FIELD(doorbell_range, + S2A_DOORBELL_ENTRY_2_CTRL, + S2A_DOORBELL_PORT2_AWID, + 0xe); + doorbell_range = REG_SET_FIELD(doorbell_range, + S2A_DOORBELL_ENTRY_2_CTRL, + S2A_DOORBELL_PORT2_RANGE_OFFSET, + doorbell_index); + doorbell_range = REG_SET_FIELD(doorbell_range, + S2A_DOORBELL_ENTRY_2_CTRL, + S2A_DOORBELL_PORT2_RANGE_SIZE, + doorbell_size); + doorbell_range = REG_SET_FIELD(doorbell_range, + S2A_DOORBELL_ENTRY_2_CTRL, + S2A_DOORBELL_PORT2_AWADDR_31_28_VALUE, + 0x3); + } else + doorbell_range = REG_SET_FIELD(doorbell_range, + S2A_DOORBELL_ENTRY_2_CTRL, + S2A_DOORBELL_PORT2_RANGE_SIZE, + 0); + + WREG32_SOC15(NBIO, 0, regS2A_DOORBELL_ENTRY_2_CTRL, doorbell_range); + } +} + +static void nbio_v4_3_vcn_doorbell_range(struct amdgpu_device *adev, bool use_doorbell, + int doorbell_index, int instance) +{ + u32 doorbell_range; + + if (instance) + doorbell_range = RREG32_SOC15(NBIO, 0, regS2A_DOORBELL_ENTRY_5_CTRL); + else + doorbell_range = RREG32_SOC15(NBIO, 0, regS2A_DOORBELL_ENTRY_4_CTRL); + + if (use_doorbell) { + doorbell_range = REG_SET_FIELD(doorbell_range, + S2A_DOORBELL_ENTRY_4_CTRL, + S2A_DOORBELL_PORT4_ENABLE, + 0x1); + doorbell_range = REG_SET_FIELD(doorbell_range, + S2A_DOORBELL_ENTRY_4_CTRL, + S2A_DOORBELL_PORT4_AWID, + instance ? 0x7 : 0x4); + doorbell_range = REG_SET_FIELD(doorbell_range, + S2A_DOORBELL_ENTRY_4_CTRL, + S2A_DOORBELL_PORT4_RANGE_OFFSET, + doorbell_index); + doorbell_range = REG_SET_FIELD(doorbell_range, + S2A_DOORBELL_ENTRY_4_CTRL, + S2A_DOORBELL_PORT4_RANGE_SIZE, + 8); + doorbell_range = REG_SET_FIELD(doorbell_range, + S2A_DOORBELL_ENTRY_4_CTRL, + S2A_DOORBELL_PORT4_AWADDR_31_28_VALUE, + instance ? 0x7 : 0x4); + } else + doorbell_range = REG_SET_FIELD(doorbell_range, + S2A_DOORBELL_ENTRY_4_CTRL, + S2A_DOORBELL_PORT4_RANGE_SIZE, + 0); + + if (instance) + WREG32_SOC15(NBIO, 0, regS2A_DOORBELL_ENTRY_5_CTRL, doorbell_range); + else + WREG32_SOC15(NBIO, 0, regS2A_DOORBELL_ENTRY_4_CTRL, doorbell_range); +} + +static void nbio_v4_3_gc_doorbell_init(struct amdgpu_device *adev) +{ + WREG32_SOC15(NBIO, 0, regS2A_DOORBELL_ENTRY_0_CTRL, 0x30000007); + WREG32_SOC15(NBIO, 0, regS2A_DOORBELL_ENTRY_3_CTRL, 0x3000000d); +} + +static void nbio_v4_3_enable_doorbell_aperture(struct amdgpu_device *adev, + bool enable) +{ + WREG32_FIELD15_PREREG(NBIO, 0, RCC_DEV0_EPF0_RCC_DOORBELL_APER_EN, + BIF_DOORBELL_APER_EN, enable ? 1 : 0); +} + +static void nbio_v4_3_enable_doorbell_selfring_aperture(struct amdgpu_device *adev, + bool enable) +{ + u32 tmp = 0; + + if (enable) { + tmp = REG_SET_FIELD(tmp, BIF_BX_PF0_DOORBELL_SELFRING_GPA_APER_CNTL, + DOORBELL_SELFRING_GPA_APER_EN, 1) | + REG_SET_FIELD(tmp, BIF_BX_PF0_DOORBELL_SELFRING_GPA_APER_CNTL, + DOORBELL_SELFRING_GPA_APER_MODE, 1) | + REG_SET_FIELD(tmp, BIF_BX_PF0_DOORBELL_SELFRING_GPA_APER_CNTL, + DOORBELL_SELFRING_GPA_APER_SIZE, 0); + + WREG32_SOC15(NBIO, 0, regBIF_BX_PF0_DOORBELL_SELFRING_GPA_APER_BASE_LOW, + lower_32_bits(adev->doorbell.base)); + WREG32_SOC15(NBIO, 0, regBIF_BX_PF0_DOORBELL_SELFRING_GPA_APER_BASE_HIGH, + upper_32_bits(adev->doorbell.base)); + } + + WREG32_SOC15(NBIO, 0, regBIF_BX_PF0_DOORBELL_SELFRING_GPA_APER_CNTL, + tmp); +} + +static void nbio_v4_3_ih_doorbell_range(struct amdgpu_device *adev, + bool use_doorbell, int doorbell_index) +{ + u32 ih_doorbell_range = RREG32_SOC15(NBIO, 0, regS2A_DOORBELL_ENTRY_1_CTRL); + + if (use_doorbell) { + ih_doorbell_range = REG_SET_FIELD(ih_doorbell_range, + S2A_DOORBELL_ENTRY_1_CTRL, + S2A_DOORBELL_PORT1_ENABLE, + 0x1); + ih_doorbell_range = REG_SET_FIELD(ih_doorbell_range, + S2A_DOORBELL_ENTRY_1_CTRL, + S2A_DOORBELL_PORT1_AWID, + 0x0); + ih_doorbell_range = REG_SET_FIELD(ih_doorbell_range, + S2A_DOORBELL_ENTRY_1_CTRL, + S2A_DOORBELL_PORT1_RANGE_OFFSET, + doorbell_index); + ih_doorbell_range = REG_SET_FIELD(ih_doorbell_range, + S2A_DOORBELL_ENTRY_1_CTRL, + S2A_DOORBELL_PORT1_RANGE_SIZE, + 2); + ih_doorbell_range = REG_SET_FIELD(ih_doorbell_range, + S2A_DOORBELL_ENTRY_1_CTRL, + S2A_DOORBELL_PORT1_AWADDR_31_28_VALUE, + 0x0); + } else + ih_doorbell_range = REG_SET_FIELD(ih_doorbell_range, + S2A_DOORBELL_ENTRY_1_CTRL, + S2A_DOORBELL_PORT1_RANGE_SIZE, + 0); + + WREG32_SOC15(NBIO, 0, regS2A_DOORBELL_ENTRY_1_CTRL, ih_doorbell_range); +} + +static void nbio_v4_3_ih_control(struct amdgpu_device *adev) +{ + u32 interrupt_cntl; + + /* setup interrupt control */ + WREG32_SOC15(NBIO, 0, regBIF_BX0_INTERRUPT_CNTL2, adev->dummy_page_addr >> 8); + + interrupt_cntl = RREG32_SOC15(NBIO, 0, regBIF_BX0_INTERRUPT_CNTL); + /* + * BIF_BX0_INTERRUPT_CNTL__IH_DUMMY_RD_OVERRIDE_MASK=0 - dummy read disabled with msi, enabled without msi + * BIF_BX0_INTERRUPT_CNTL__IH_DUMMY_RD_OVERRIDE_MASK=1 - dummy read controlled by IH_DUMMY_RD_EN + */ + interrupt_cntl = REG_SET_FIELD(interrupt_cntl, BIF_BX0_INTERRUPT_CNTL, + IH_DUMMY_RD_OVERRIDE, 0); + + /* BIF_BX0_INTERRUPT_CNTL__IH_REQ_NONSNOOP_EN_MASK=1 if ring is in non-cacheable memory, e.g., vram */ + interrupt_cntl = REG_SET_FIELD(interrupt_cntl, BIF_BX0_INTERRUPT_CNTL, + IH_REQ_NONSNOOP_EN, 0); + + WREG32_SOC15(NBIO, 0, regBIF_BX0_INTERRUPT_CNTL, interrupt_cntl); +} + +static void nbio_v4_3_update_medium_grain_clock_gating(struct amdgpu_device *adev, + bool enable) +{ + uint32_t def, data; + + if (enable && !(adev->cg_flags & AMD_CG_SUPPORT_BIF_MGCG)) + return; + + def = data = RREG32_SOC15(NBIO, 0, regCPM_CONTROL); + if (enable) { + data |= (CPM_CONTROL__LCLK_DYN_GATE_ENABLE_MASK | + CPM_CONTROL__TXCLK_DYN_GATE_ENABLE_MASK | + CPM_CONTROL__TXCLK_LCNT_GATE_ENABLE_MASK | + CPM_CONTROL__TXCLK_REGS_GATE_ENABLE_MASK | + CPM_CONTROL__TXCLK_PRBS_GATE_ENABLE_MASK | + CPM_CONTROL__REFCLK_REGS_GATE_ENABLE_MASK); + } else { + data &= ~(CPM_CONTROL__LCLK_DYN_GATE_ENABLE_MASK | + CPM_CONTROL__TXCLK_DYN_GATE_ENABLE_MASK | + CPM_CONTROL__TXCLK_LCNT_GATE_ENABLE_MASK | + CPM_CONTROL__TXCLK_REGS_GATE_ENABLE_MASK | + CPM_CONTROL__TXCLK_PRBS_GATE_ENABLE_MASK | + CPM_CONTROL__REFCLK_REGS_GATE_ENABLE_MASK); + } + + if (def != data) + WREG32_SOC15(NBIO, 0, regCPM_CONTROL, data); +} + +static void nbio_v4_3_update_medium_grain_light_sleep(struct amdgpu_device *adev, + bool enable) +{ + uint32_t def, data; + + if (enable && !(adev->cg_flags & AMD_CG_SUPPORT_BIF_LS)) + return; + + /* TODO: need update in future */ + def = data = RREG32_SOC15(NBIO, 0, regPCIE_CNTL2); + if (enable) { + data |= PCIE_CNTL2__SLV_MEM_LS_EN_MASK; + } else { + data &= ~PCIE_CNTL2__SLV_MEM_LS_EN_MASK; + } + + if (def != data) + WREG32_SOC15(NBIO, 0, regPCIE_CNTL2, data); +} + +static void nbio_v4_3_get_clockgating_state(struct amdgpu_device *adev, + u64 *flags) +{ + int data; + + /* AMD_CG_SUPPORT_BIF_MGCG */ + data = RREG32_SOC15(NBIO, 0, regCPM_CONTROL); + if (data & CPM_CONTROL__LCLK_DYN_GATE_ENABLE_MASK) + *flags |= AMD_CG_SUPPORT_BIF_MGCG; + + /* AMD_CG_SUPPORT_BIF_LS */ + data = RREG32_SOC15(NBIO, 0, regPCIE_CNTL2); + if (data & PCIE_CNTL2__SLV_MEM_LS_EN_MASK) + *flags |= AMD_CG_SUPPORT_BIF_LS; +} + +static u32 nbio_v4_3_get_hdp_flush_req_offset(struct amdgpu_device *adev) +{ + return SOC15_REG_OFFSET(NBIO, 0, regBIF_BX_PF0_GPU_HDP_FLUSH_REQ); +} + +static u32 nbio_v4_3_get_hdp_flush_done_offset(struct amdgpu_device *adev) +{ + return SOC15_REG_OFFSET(NBIO, 0, regBIF_BX_PF0_GPU_HDP_FLUSH_DONE); +} + +static u32 nbio_v4_3_get_pcie_index_offset(struct amdgpu_device *adev) +{ + return SOC15_REG_OFFSET(NBIO, 0, regBIF_BX_PF0_RSMU_INDEX); +} + +static u32 nbio_v4_3_get_pcie_data_offset(struct amdgpu_device *adev) +{ + return SOC15_REG_OFFSET(NBIO, 0, regBIF_BX_PF0_RSMU_DATA); +} + +const struct nbio_hdp_flush_reg nbio_v4_3_hdp_flush_reg = { + .ref_and_mask_cp0 = BIF_BX_PF_GPU_HDP_FLUSH_DONE__CP0_MASK, + .ref_and_mask_cp1 = BIF_BX_PF_GPU_HDP_FLUSH_DONE__CP1_MASK, + .ref_and_mask_cp2 = BIF_BX_PF_GPU_HDP_FLUSH_DONE__CP2_MASK, + .ref_and_mask_cp3 = BIF_BX_PF_GPU_HDP_FLUSH_DONE__CP3_MASK, + .ref_and_mask_cp4 = BIF_BX_PF_GPU_HDP_FLUSH_DONE__CP4_MASK, + .ref_and_mask_cp5 = BIF_BX_PF_GPU_HDP_FLUSH_DONE__CP5_MASK, + .ref_and_mask_cp6 = BIF_BX_PF_GPU_HDP_FLUSH_DONE__CP6_MASK, + .ref_and_mask_cp7 = BIF_BX_PF_GPU_HDP_FLUSH_DONE__CP7_MASK, + .ref_and_mask_cp8 = BIF_BX_PF_GPU_HDP_FLUSH_DONE__CP8_MASK, + .ref_and_mask_cp9 = BIF_BX_PF_GPU_HDP_FLUSH_DONE__CP9_MASK, + .ref_and_mask_sdma0 = BIF_BX_PF_GPU_HDP_FLUSH_DONE__SDMA0_MASK, + .ref_and_mask_sdma1 = BIF_BX_PF_GPU_HDP_FLUSH_DONE__SDMA1_MASK, +}; + +static void nbio_v4_3_init_registers(struct amdgpu_device *adev) +{ + return; +} + +static u32 nbio_v4_3_get_rom_offset(struct amdgpu_device *adev) +{ + u32 data, rom_offset; + + data = RREG32_SOC15(NBIO, 0, regREGS_ROM_OFFSET_CTRL); + rom_offset = REG_GET_FIELD(data, REGS_ROM_OFFSET_CTRL, ROM_OFFSET); + + return rom_offset; +} + +#ifdef CONFIG_PCIEASPM +static void nbio_v4_3_program_ltr(struct amdgpu_device *adev) +{ + uint32_t def, data; + + def = RREG32_SOC15(NBIO, 0, regRCC_EP_DEV0_0_EP_PCIE_TX_LTR_CNTL); + data = 0x35EB; + data &= ~EP_PCIE_TX_LTR_CNTL__LTR_PRIV_MSG_DIS_IN_PM_NON_D0_MASK; + data &= ~EP_PCIE_TX_LTR_CNTL__LTR_PRIV_RST_LTR_IN_DL_DOWN_MASK; + if (def != data) + WREG32_SOC15(NBIO, 0, regRCC_EP_DEV0_0_EP_PCIE_TX_LTR_CNTL, data); + + def = data = RREG32_SOC15(NBIO, 0, regRCC_STRAP0_RCC_BIF_STRAP2); + data &= ~RCC_BIF_STRAP2__STRAP_LTR_IN_ASPML1_DIS_MASK; + if (def != data) + WREG32_SOC15(NBIO, 0, regRCC_STRAP0_RCC_BIF_STRAP2, data); + + def = data = RREG32_SOC15(NBIO, 0, regBIF_CFG_DEV0_EPF0_DEVICE_CNTL2); + if (adev->pdev->ltr_path) + data |= BIF_CFG_DEV0_EPF0_DEVICE_CNTL2__LTR_EN_MASK; + else + data &= ~BIF_CFG_DEV0_EPF0_DEVICE_CNTL2__LTR_EN_MASK; + if (def != data) + WREG32_SOC15(NBIO, 0, regBIF_CFG_DEV0_EPF0_DEVICE_CNTL2, data); +} +#endif + +static void nbio_v4_3_program_aspm(struct amdgpu_device *adev) +{ +#ifdef CONFIG_PCIEASPM + uint32_t def, data; + + if (!(adev->ip_versions[PCIE_HWIP][0] == IP_VERSION(7, 4, 0)) && + !(adev->ip_versions[PCIE_HWIP][0] == IP_VERSION(7, 6, 0))) + return; + + def = data = RREG32_SOC15(NBIO, 0, regPCIE_LC_CNTL); + data &= ~PCIE_LC_CNTL__LC_L1_INACTIVITY_MASK; + data &= ~PCIE_LC_CNTL__LC_L0S_INACTIVITY_MASK; + data |= PCIE_LC_CNTL__LC_PMI_TO_L1_DIS_MASK; + if (def != data) + WREG32_SOC15(NBIO, 0, regPCIE_LC_CNTL, data); + + def = data = RREG32_SOC15(NBIO, 0, regPCIE_LC_CNTL7); + data |= PCIE_LC_CNTL7__LC_NBIF_ASPM_INPUT_EN_MASK; + if (def != data) + WREG32_SOC15(NBIO, 0, regPCIE_LC_CNTL7, data); + + def = data = RREG32_SOC15(NBIO, 0, regPCIE_LC_CNTL3); + data |= PCIE_LC_CNTL3__LC_DSC_DONT_ENTER_L23_AFTER_PME_ACK_MASK; + if (def != data) + WREG32_SOC15(NBIO, 0, regPCIE_LC_CNTL3, data); + + def = data = RREG32_SOC15(NBIO, 0, regRCC_STRAP0_RCC_BIF_STRAP3); + data &= ~RCC_BIF_STRAP3__STRAP_VLINK_ASPM_IDLE_TIMER_MASK; + data &= ~RCC_BIF_STRAP3__STRAP_VLINK_PM_L1_ENTRY_TIMER_MASK; + if (def != data) + WREG32_SOC15(NBIO, 0, regRCC_STRAP0_RCC_BIF_STRAP3, data); + + def = data = RREG32_SOC15(NBIO, 0, regRCC_STRAP0_RCC_BIF_STRAP5); + data &= ~RCC_BIF_STRAP5__STRAP_VLINK_LDN_ENTRY_TIMER_MASK; + if (def != data) + WREG32_SOC15(NBIO, 0, regRCC_STRAP0_RCC_BIF_STRAP5, data); + + def = data = RREG32_SOC15(NBIO, 0, regBIF_CFG_DEV0_EPF0_DEVICE_CNTL2); + data &= ~BIF_CFG_DEV0_EPF0_DEVICE_CNTL2__LTR_EN_MASK; + if (def != data) + WREG32_SOC15(NBIO, 0, regBIF_CFG_DEV0_EPF0_DEVICE_CNTL2, data); + + WREG32_SOC15(NBIO, 0, regBIF_CFG_DEV0_EPF0_PCIE_LTR_CAP, 0x10011001); + + def = data = RREG32_SOC15(NBIO, 0, regPSWUSP0_PCIE_LC_CNTL2); + data |= PSWUSP0_PCIE_LC_CNTL2__LC_ALLOW_PDWN_IN_L1_MASK | + PSWUSP0_PCIE_LC_CNTL2__LC_ALLOW_PDWN_IN_L23_MASK; + data &= ~PSWUSP0_PCIE_LC_CNTL2__LC_RCV_L0_TO_RCV_L0S_DIS_MASK; + if (def != data) + WREG32_SOC15(NBIO, 0, regPSWUSP0_PCIE_LC_CNTL2, data); + + def = data = RREG32_SOC15(NBIO, 0, regPCIE_LC_CNTL4); + data |= PCIE_LC_CNTL4__LC_L1_POWERDOWN_MASK; + if (def != data) + WREG32_SOC15(NBIO, 0, regPCIE_LC_CNTL4, data); + + def = data = RREG32_SOC15(NBIO, 0, regPCIE_LC_RXRECOVER_RXSTANDBY_CNTL); + data |= PCIE_LC_RXRECOVER_RXSTANDBY_CNTL__LC_RX_L0S_STANDBY_EN_MASK; + if (def != data) + WREG32_SOC15(NBIO, 0, regPCIE_LC_RXRECOVER_RXSTANDBY_CNTL, data); + + nbio_v4_3_program_ltr(adev); + + def = data = RREG32_SOC15(NBIO, 0, regRCC_STRAP0_RCC_BIF_STRAP3); + data |= 0x5DE0 << RCC_BIF_STRAP3__STRAP_VLINK_ASPM_IDLE_TIMER__SHIFT; + data |= 0x0010 << RCC_BIF_STRAP3__STRAP_VLINK_PM_L1_ENTRY_TIMER__SHIFT; + if (def != data) + WREG32_SOC15(NBIO, 0, regRCC_STRAP0_RCC_BIF_STRAP3, data); + + def = data = RREG32_SOC15(NBIO, 0, regRCC_STRAP0_RCC_BIF_STRAP5); + data |= 0x0010 << RCC_BIF_STRAP5__STRAP_VLINK_LDN_ENTRY_TIMER__SHIFT; + if (def != data) + WREG32_SOC15(NBIO, 0, regRCC_STRAP0_RCC_BIF_STRAP5, data); + + def = data = RREG32_SOC15(NBIO, 0, regPCIE_LC_CNTL); + data |= 0x0 << PCIE_LC_CNTL__LC_L0S_INACTIVITY__SHIFT; + data |= 0x9 << PCIE_LC_CNTL__LC_L1_INACTIVITY__SHIFT; + data &= ~PCIE_LC_CNTL__LC_PMI_TO_L1_DIS_MASK; + if (def != data) + WREG32_SOC15(NBIO, 0, regPCIE_LC_CNTL, data); + + def = data = RREG32_SOC15(NBIO, 0, regPCIE_LC_CNTL3); + data &= ~PCIE_LC_CNTL3__LC_DSC_DONT_ENTER_L23_AFTER_PME_ACK_MASK; + if (def != data) + WREG32_SOC15(NBIO, 0, regPCIE_LC_CNTL3, data); +#endif +} + +const struct amdgpu_nbio_funcs nbio_v4_3_funcs = { + .get_hdp_flush_req_offset = nbio_v4_3_get_hdp_flush_req_offset, + .get_hdp_flush_done_offset = nbio_v4_3_get_hdp_flush_done_offset, + .get_pcie_index_offset = nbio_v4_3_get_pcie_index_offset, + .get_pcie_data_offset = nbio_v4_3_get_pcie_data_offset, + .get_rev_id = nbio_v4_3_get_rev_id, + .mc_access_enable = nbio_v4_3_mc_access_enable, + .get_memsize = nbio_v4_3_get_memsize, + .sdma_doorbell_range = nbio_v4_3_sdma_doorbell_range, + .vcn_doorbell_range = nbio_v4_3_vcn_doorbell_range, + .gc_doorbell_init = nbio_v4_3_gc_doorbell_init, + .enable_doorbell_aperture = nbio_v4_3_enable_doorbell_aperture, + .enable_doorbell_selfring_aperture = nbio_v4_3_enable_doorbell_selfring_aperture, + .ih_doorbell_range = nbio_v4_3_ih_doorbell_range, + .update_medium_grain_clock_gating = nbio_v4_3_update_medium_grain_clock_gating, + .update_medium_grain_light_sleep = nbio_v4_3_update_medium_grain_light_sleep, + .get_clockgating_state = nbio_v4_3_get_clockgating_state, + .ih_control = nbio_v4_3_ih_control, + .init_registers = nbio_v4_3_init_registers, + .remap_hdp_registers = nbio_v4_3_remap_hdp_registers, + .get_rom_offset = nbio_v4_3_get_rom_offset, + .program_aspm = nbio_v4_3_program_aspm, +}; + + +static void nbio_v4_3_sriov_ih_doorbell_range(struct amdgpu_device *adev, + bool use_doorbell, int doorbell_index) +{ +} + +static void nbio_v4_3_sriov_sdma_doorbell_range(struct amdgpu_device *adev, int instance, + bool use_doorbell, int doorbell_index, + int doorbell_size) +{ +} + +static void nbio_v4_3_sriov_vcn_doorbell_range(struct amdgpu_device *adev, bool use_doorbell, + int doorbell_index, int instance) +{ +} + +static void nbio_v4_3_sriov_gc_doorbell_init(struct amdgpu_device *adev) +{ +} + +const struct amdgpu_nbio_funcs nbio_v4_3_sriov_funcs = { + .get_hdp_flush_req_offset = nbio_v4_3_get_hdp_flush_req_offset, + .get_hdp_flush_done_offset = nbio_v4_3_get_hdp_flush_done_offset, + .get_pcie_index_offset = nbio_v4_3_get_pcie_index_offset, + .get_pcie_data_offset = nbio_v4_3_get_pcie_data_offset, + .get_rev_id = nbio_v4_3_get_rev_id, + .mc_access_enable = nbio_v4_3_mc_access_enable, + .get_memsize = nbio_v4_3_get_memsize, + .sdma_doorbell_range = nbio_v4_3_sriov_sdma_doorbell_range, + .vcn_doorbell_range = nbio_v4_3_sriov_vcn_doorbell_range, + .gc_doorbell_init = nbio_v4_3_sriov_gc_doorbell_init, + .enable_doorbell_aperture = nbio_v4_3_enable_doorbell_aperture, + .enable_doorbell_selfring_aperture = nbio_v4_3_enable_doorbell_selfring_aperture, + .ih_doorbell_range = nbio_v4_3_sriov_ih_doorbell_range, + .update_medium_grain_clock_gating = nbio_v4_3_update_medium_grain_clock_gating, + .update_medium_grain_light_sleep = nbio_v4_3_update_medium_grain_light_sleep, + .get_clockgating_state = nbio_v4_3_get_clockgating_state, + .ih_control = nbio_v4_3_ih_control, + .init_registers = nbio_v4_3_init_registers, + .remap_hdp_registers = nbio_v4_3_remap_hdp_registers, + .get_rom_offset = nbio_v4_3_get_rom_offset, +}; diff --git a/drivers/gpu/drm/amd/amdgpu/nbio_v4_3.h b/drivers/gpu/drm/amd/amdgpu/nbio_v4_3.h new file mode 100644 index 000000000000..711999ceedf4 --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/nbio_v4_3.h @@ -0,0 +1,33 @@ +/* + * Copyright 2021 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#ifndef __NBIO_V4_3_H__ +#define __NBIO_V4_3_H__ + +#include "soc15_common.h" + +extern const struct nbio_hdp_flush_reg nbio_v4_3_hdp_flush_reg; +extern const struct amdgpu_nbio_funcs nbio_v4_3_funcs; +extern const struct amdgpu_nbio_funcs nbio_v4_3_sriov_funcs; + +#endif diff --git a/drivers/gpu/drm/amd/amdgpu/nbio_v6_1.c b/drivers/gpu/drm/amd/amdgpu/nbio_v6_1.c index 4bbacf1be25a..37615a77287b 100644 --- a/drivers/gpu/drm/amd/amdgpu/nbio_v6_1.c +++ b/drivers/gpu/drm/amd/amdgpu/nbio_v6_1.c @@ -210,7 +210,7 @@ static void nbio_v6_1_update_medium_grain_light_sleep(struct amdgpu_device *adev } static void nbio_v6_1_get_clockgating_state(struct amdgpu_device *adev, - u32 *flags) + u64 *flags) { int data; @@ -282,6 +282,7 @@ static void nbio_v6_1_init_registers(struct amdgpu_device *adev) mmBIF_BX_DEV0_EPF0_VF0_HDP_MEM_COHERENCY_FLUSH_CNTL) << 2; } +#ifdef CONFIG_PCIEASPM static void nbio_v6_1_program_ltr(struct amdgpu_device *adev) { uint32_t def, data; @@ -303,9 +304,11 @@ static void nbio_v6_1_program_ltr(struct amdgpu_device *adev) if (def != data) WREG32_PCIE(smnBIF_CFG_DEV0_EPF0_DEVICE_CNTL2, data); } +#endif static void nbio_v6_1_program_aspm(struct amdgpu_device *adev) { +#ifdef CONFIG_PCIEASPM uint32_t def, data; def = data = RREG32_PCIE(smnPCIE_LC_CNTL); @@ -361,7 +364,10 @@ static void nbio_v6_1_program_aspm(struct amdgpu_device *adev) if (def != data) WREG32_PCIE(smnPCIE_LC_CNTL6, data); - nbio_v6_1_program_ltr(adev); + /* Don't bother about LTR if LTR is not enabled + * in the path */ + if (adev->pdev->ltr_path) + nbio_v6_1_program_ltr(adev); def = data = RREG32_PCIE(smnRCC_BIF_STRAP3); data |= 0x5DE0 << RCC_BIF_STRAP3__STRAP_VLINK_ASPM_IDLE_TIMER__SHIFT; @@ -385,6 +391,7 @@ static void nbio_v6_1_program_aspm(struct amdgpu_device *adev) data &= ~PCIE_LC_CNTL3__LC_DSC_DONT_ENTER_L23_AFTER_PME_ACK_MASK; if (def != data) WREG32_PCIE(smnPCIE_LC_CNTL3, data); +#endif } const struct amdgpu_nbio_funcs nbio_v6_1_funcs = { diff --git a/drivers/gpu/drm/amd/amdgpu/nbio_v7_0.c b/drivers/gpu/drm/amd/amdgpu/nbio_v7_0.c index 37a4039fdfc5..aa0326d00c72 100644 --- a/drivers/gpu/drm/amd/amdgpu/nbio_v7_0.c +++ b/drivers/gpu/drm/amd/amdgpu/nbio_v7_0.c @@ -205,7 +205,7 @@ static void nbio_v7_0_update_medium_grain_light_sleep(struct amdgpu_device *adev } static void nbio_v7_0_get_clockgating_state(struct amdgpu_device *adev, - u32 *flags) + u64 *flags) { int data; diff --git a/drivers/gpu/drm/amd/amdgpu/nbio_v7_2.c b/drivers/gpu/drm/amd/amdgpu/nbio_v7_2.c index 3444332ea110..31776b12e4c4 100644 --- a/drivers/gpu/drm/amd/amdgpu/nbio_v7_2.c +++ b/drivers/gpu/drm/amd/amdgpu/nbio_v7_2.c @@ -59,10 +59,16 @@ static u32 nbio_v7_2_get_rev_id(struct amdgpu_device *adev) { u32 tmp; - if (adev->asic_type == CHIP_YELLOW_CARP) + switch (adev->ip_versions[NBIO_HWIP][0]) { + case IP_VERSION(7, 2, 1): + case IP_VERSION(7, 3, 0): + case IP_VERSION(7, 5, 0): tmp = RREG32_SOC15(NBIO, 0, regRCC_STRAP0_RCC_DEV0_EPF0_STRAP0_YC); - else + break; + default: tmp = RREG32_SOC15(NBIO, 0, regRCC_STRAP0_RCC_DEV0_EPF0_STRAP0); + break; + } tmp &= RCC_STRAP0_RCC_DEV0_EPF0_STRAP0__STRAP_ATI_REV_ID_DEV0_F0_MASK; tmp >>= RCC_STRAP0_RCC_DEV0_EPF0_STRAP0__STRAP_ATI_REV_ID_DEV0_F0__SHIFT; @@ -72,20 +78,26 @@ static u32 nbio_v7_2_get_rev_id(struct amdgpu_device *adev) static void nbio_v7_2_mc_access_enable(struct amdgpu_device *adev, bool enable) { - if (enable) - if (adev->asic_type == CHIP_YELLOW_CARP) + switch (adev->ip_versions[NBIO_HWIP][0]) { + case IP_VERSION(7, 2, 1): + case IP_VERSION(7, 3, 0): + case IP_VERSION(7, 5, 0): + if (enable) WREG32_SOC15(NBIO, 0, regBIF_BX0_BIF_FB_EN_YC, BIF_BX0_BIF_FB_EN__FB_READ_EN_MASK | BIF_BX0_BIF_FB_EN__FB_WRITE_EN_MASK); else + WREG32_SOC15(NBIO, 0, regBIF_BX0_BIF_FB_EN_YC, 0); + break; + default: + if (enable) WREG32_SOC15(NBIO, 0, regBIF_BX0_BIF_FB_EN, BIF_BX0_BIF_FB_EN__FB_READ_EN_MASK | BIF_BX0_BIF_FB_EN__FB_WRITE_EN_MASK); - else - if (adev->asic_type == CHIP_YELLOW_CARP) - WREG32_SOC15(NBIO, 0, regBIF_BX0_BIF_FB_EN_YC, 0); else WREG32_SOC15(NBIO, 0, regBIF_BX0_BIF_FB_EN, 0); + break; + } } static u32 nbio_v7_2_get_memsize(struct amdgpu_device *adev) @@ -250,7 +262,10 @@ static void nbio_v7_2_update_medium_grain_light_sleep(struct amdgpu_device *adev { uint32_t def, data; - if (adev->asic_type == CHIP_YELLOW_CARP) { + switch (adev->ip_versions[NBIO_HWIP][0]) { + case IP_VERSION(7, 2, 1): + case IP_VERSION(7, 3, 0): + case IP_VERSION(7, 5, 0): def = data = RREG32_PCIE_PORT(SOC15_REG_OFFSET(NBIO, 0, regPCIE_CNTL2)); if (enable && (adev->cg_flags & AMD_CG_SUPPORT_BIF_LS)) data |= PCIE_CNTL2__SLV_MEM_LS_EN_MASK; @@ -260,8 +275,8 @@ static void nbio_v7_2_update_medium_grain_light_sleep(struct amdgpu_device *adev if (def != data) WREG32_PCIE_PORT(SOC15_REG_OFFSET(NBIO, 0, regPCIE_CNTL2), data); - data = RREG32_PCIE_PORT(SOC15_REG_OFFSET(NBIO, 0, regBIF1_PCIE_TX_POWER_CTRL_1)); - def = data; + def = data = RREG32_PCIE_PORT(SOC15_REG_OFFSET(NBIO, 0, + regBIF1_PCIE_TX_POWER_CTRL_1)); if (enable && (adev->cg_flags & AMD_CG_SUPPORT_BIF_LS)) data |= (BIF1_PCIE_TX_POWER_CTRL_1__MST_MEM_LS_EN_MASK | BIF1_PCIE_TX_POWER_CTRL_1__REPLAY_MEM_LS_EN_MASK); @@ -272,7 +287,8 @@ static void nbio_v7_2_update_medium_grain_light_sleep(struct amdgpu_device *adev if (def != data) WREG32_PCIE_PORT(SOC15_REG_OFFSET(NBIO, 0, regBIF1_PCIE_TX_POWER_CTRL_1), data); - } else { + break; + default: def = data = RREG32_PCIE_PORT(SOC15_REG_OFFSET(NBIO, 0, regPCIE_CNTL2)); if (enable && (adev->cg_flags & AMD_CG_SUPPORT_BIF_LS)) data |= (PCIE_CNTL2__SLV_MEM_LS_EN_MASK | @@ -285,11 +301,12 @@ static void nbio_v7_2_update_medium_grain_light_sleep(struct amdgpu_device *adev if (def != data) WREG32_PCIE_PORT(SOC15_REG_OFFSET(NBIO, 0, regPCIE_CNTL2), data); + break; } } static void nbio_v7_2_get_clockgating_state(struct amdgpu_device *adev, - u32 *flags) + u64 *flags) { int data; @@ -352,7 +369,10 @@ const struct nbio_hdp_flush_reg nbio_v7_2_hdp_flush_reg = { static void nbio_v7_2_init_registers(struct amdgpu_device *adev) { uint32_t def, data; - if (adev->asic_type == CHIP_YELLOW_CARP) { + switch (adev->ip_versions[NBIO_HWIP][0]) { + case IP_VERSION(7, 2, 1): + case IP_VERSION(7, 3, 0): + case IP_VERSION(7, 5, 0): def = data = RREG32_PCIE_PORT(SOC15_REG_OFFSET(NBIO, 0, regBIF1_PCIE_MST_CTRL_3)); data = REG_SET_FIELD(data, BIF1_PCIE_MST_CTRL_3, CI_SWUS_MAX_READ_REQUEST_SIZE_MODE, 1); @@ -361,7 +381,8 @@ static void nbio_v7_2_init_registers(struct amdgpu_device *adev) if (def != data) WREG32_PCIE_PORT(SOC15_REG_OFFSET(NBIO, 0, regBIF1_PCIE_MST_CTRL_3), data); - } else { + break; + default: def = data = RREG32_PCIE_PORT(SOC15_REG_OFFSET(NBIO, 0, regPCIE_CONFIG_CNTL)); data = REG_SET_FIELD(data, PCIE_CONFIG_CNTL, CI_SWUS_MAX_READ_REQUEST_SIZE_MODE, 1); @@ -370,6 +391,7 @@ static void nbio_v7_2_init_registers(struct amdgpu_device *adev) if (def != data) WREG32_PCIE_PORT(SOC15_REG_OFFSET(NBIO, 0, regPCIE_CONFIG_CNTL), data); + break; } if (amdgpu_sriov_vf(adev)) diff --git a/drivers/gpu/drm/amd/amdgpu/nbio_v7_4.c b/drivers/gpu/drm/amd/amdgpu/nbio_v7_4.c index dc5e93756fea..19455a725939 100644 --- a/drivers/gpu/drm/amd/amdgpu/nbio_v7_4.c +++ b/drivers/gpu/drm/amd/amdgpu/nbio_v7_4.c @@ -273,7 +273,7 @@ static void nbio_v7_4_update_medium_grain_light_sleep(struct amdgpu_device *adev } static void nbio_v7_4_get_clockgating_state(struct amdgpu_device *adev, - u32 *flags) + u64 *flags) { int data; @@ -339,32 +339,26 @@ const struct nbio_hdp_flush_reg nbio_v7_4_hdp_flush_reg = { .ref_and_mask_sdma1 = GPU_HDP_FLUSH_DONE__SDMA1_MASK, }; -const struct nbio_hdp_flush_reg nbio_v7_4_hdp_flush_reg_ald = { - .ref_and_mask_cp0 = GPU_HDP_FLUSH_DONE__CP0_MASK, - .ref_and_mask_cp1 = GPU_HDP_FLUSH_DONE__CP1_MASK, - .ref_and_mask_cp2 = GPU_HDP_FLUSH_DONE__CP2_MASK, - .ref_and_mask_cp3 = GPU_HDP_FLUSH_DONE__CP3_MASK, - .ref_and_mask_cp4 = GPU_HDP_FLUSH_DONE__CP4_MASK, - .ref_and_mask_cp5 = GPU_HDP_FLUSH_DONE__CP5_MASK, - .ref_and_mask_cp6 = GPU_HDP_FLUSH_DONE__CP6_MASK, - .ref_and_mask_cp7 = GPU_HDP_FLUSH_DONE__CP7_MASK, - .ref_and_mask_cp8 = GPU_HDP_FLUSH_DONE__CP8_MASK, - .ref_and_mask_cp9 = GPU_HDP_FLUSH_DONE__CP9_MASK, - .ref_and_mask_sdma0 = GPU_HDP_FLUSH_DONE__RSVD_ENG1_MASK, - .ref_and_mask_sdma1 = GPU_HDP_FLUSH_DONE__RSVD_ENG2_MASK, - .ref_and_mask_sdma2 = GPU_HDP_FLUSH_DONE__RSVD_ENG3_MASK, - .ref_and_mask_sdma3 = GPU_HDP_FLUSH_DONE__RSVD_ENG4_MASK, - .ref_and_mask_sdma4 = GPU_HDP_FLUSH_DONE__RSVD_ENG5_MASK, - .ref_and_mask_sdma5 = GPU_HDP_FLUSH_DONE__RSVD_ENG6_MASK, - .ref_and_mask_sdma6 = GPU_HDP_FLUSH_DONE__RSVD_ENG7_MASK, - .ref_and_mask_sdma7 = GPU_HDP_FLUSH_DONE__RSVD_ENG8_MASK, -}; - static void nbio_v7_4_init_registers(struct amdgpu_device *adev) { + uint32_t baco_cntl; + if (amdgpu_sriov_vf(adev)) adev->rmmio_remap.reg_offset = SOC15_REG_OFFSET(NBIO, 0, mmBIF_BX_DEV0_EPF0_VF0_HDP_MEM_COHERENCY_FLUSH_CNTL) << 2; + + if (adev->ip_versions[NBIO_HWIP][0] == IP_VERSION(7, 4, 4) && + !amdgpu_sriov_vf(adev)) { + baco_cntl = RREG32_SOC15(NBIO, 0, mmBACO_CNTL); + if (baco_cntl & + (BACO_CNTL__BACO_DUMMY_EN_MASK | BACO_CNTL__BACO_EN_MASK)) { + baco_cntl &= ~(BACO_CNTL__BACO_DUMMY_EN_MASK | + BACO_CNTL__BACO_EN_MASK); + dev_dbg(adev->dev, "Unsetting baco dummy mode %x", + baco_cntl); + WREG32_SOC15(NBIO, 0, mmBACO_CNTL, baco_cntl); + } + } } static void nbio_v7_4_handle_ras_controller_intr_no_bifring(struct amdgpu_device *adev) @@ -658,16 +652,28 @@ static void nbio_v7_4_enable_doorbell_interrupt(struct amdgpu_device *adev, DOORBELL_INTERRUPT_DISABLE, enable ? 0 : 1); } -const struct amdgpu_nbio_ras_funcs nbio_v7_4_ras_funcs = { +const struct amdgpu_ras_block_hw_ops nbio_v7_4_ras_hw_ops = { + .query_ras_error_count = nbio_v7_4_query_ras_error_count, +}; + +struct amdgpu_nbio_ras nbio_v7_4_ras = { + .ras_block = { + .ras_comm = { + .name = "pcie_bif", + .block = AMDGPU_RAS_BLOCK__PCIE_BIF, + .type = AMDGPU_RAS_ERROR__MULTI_UNCORRECTABLE, + }, + .hw_ops = &nbio_v7_4_ras_hw_ops, + .ras_late_init = amdgpu_nbio_ras_late_init, + }, .handle_ras_controller_intr_no_bifring = nbio_v7_4_handle_ras_controller_intr_no_bifring, .handle_ras_err_event_athub_intr_no_bifring = nbio_v7_4_handle_ras_err_event_athub_intr_no_bifring, .init_ras_controller_interrupt = nbio_v7_4_init_ras_controller_interrupt, .init_ras_err_event_athub_interrupt = nbio_v7_4_init_ras_err_event_athub_interrupt, - .query_ras_error_count = nbio_v7_4_query_ras_error_count, - .ras_late_init = amdgpu_nbio_ras_late_init, - .ras_fini = amdgpu_nbio_ras_fini, }; + +#ifdef CONFIG_PCIEASPM static void nbio_v7_4_program_ltr(struct amdgpu_device *adev) { uint32_t def, data; @@ -689,9 +695,11 @@ static void nbio_v7_4_program_ltr(struct amdgpu_device *adev) if (def != data) WREG32_PCIE(smnBIF_CFG_DEV0_EPF0_DEVICE_CNTL2, data); } +#endif static void nbio_v7_4_program_aspm(struct amdgpu_device *adev) { +#ifdef CONFIG_PCIEASPM uint32_t def, data; if (adev->ip_versions[NBIO_HWIP][0] == IP_VERSION(7, 4, 4)) @@ -750,7 +758,10 @@ static void nbio_v7_4_program_aspm(struct amdgpu_device *adev) if (def != data) WREG32_PCIE(smnPCIE_LC_CNTL6, data); - nbio_v7_4_program_ltr(adev); + /* Don't bother about LTR if LTR is not enabled + * in the path */ + if (adev->pdev->ltr_path) + nbio_v7_4_program_ltr(adev); def = data = RREG32_PCIE(smnRCC_BIF_STRAP3); data |= 0x5DE0 << RCC_BIF_STRAP3__STRAP_VLINK_ASPM_IDLE_TIMER__SHIFT; @@ -774,6 +785,7 @@ static void nbio_v7_4_program_aspm(struct amdgpu_device *adev) data &= ~PCIE_LC_CNTL3__LC_DSC_DONT_ENTER_L23_AFTER_PME_ACK_MASK; if (def != data) WREG32_PCIE(smnPCIE_LC_CNTL3, data); +#endif } const struct amdgpu_nbio_funcs nbio_v7_4_funcs = { diff --git a/drivers/gpu/drm/amd/amdgpu/nbio_v7_4.h b/drivers/gpu/drm/amd/amdgpu/nbio_v7_4.h index cc5692db6f98..f27c41728822 100644 --- a/drivers/gpu/drm/amd/amdgpu/nbio_v7_4.h +++ b/drivers/gpu/drm/amd/amdgpu/nbio_v7_4.h @@ -27,8 +27,7 @@ #include "soc15_common.h" extern const struct nbio_hdp_flush_reg nbio_v7_4_hdp_flush_reg; -extern const struct nbio_hdp_flush_reg nbio_v7_4_hdp_flush_reg_ald; extern const struct amdgpu_nbio_funcs nbio_v7_4_funcs; -extern const struct amdgpu_nbio_ras_funcs nbio_v7_4_ras_funcs; +extern struct amdgpu_nbio_ras nbio_v7_4_ras; #endif diff --git a/drivers/gpu/drm/amd/amdgpu/nbio_v7_7.c b/drivers/gpu/drm/amd/amdgpu/nbio_v7_7.c new file mode 100644 index 000000000000..def89379b51a --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/nbio_v7_7.c @@ -0,0 +1,348 @@ +/* + * Copyright 2021 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ +#include "amdgpu.h" +#include "amdgpu_atombios.h" +#include "nbio_v7_7.h" + +#include "nbio/nbio_7_7_0_offset.h" +#include "nbio/nbio_7_7_0_sh_mask.h" +#include <uapi/linux/kfd_ioctl.h> + +static void nbio_v7_7_remap_hdp_registers(struct amdgpu_device *adev) +{ + WREG32_SOC15(NBIO, 0, regBIF_BX0_REMAP_HDP_MEM_FLUSH_CNTL, + adev->rmmio_remap.reg_offset + KFD_MMIO_REMAP_HDP_MEM_FLUSH_CNTL); + WREG32_SOC15(NBIO, 0, regBIF_BX0_REMAP_HDP_REG_FLUSH_CNTL, + adev->rmmio_remap.reg_offset + KFD_MMIO_REMAP_HDP_REG_FLUSH_CNTL); +} + +static u32 nbio_v7_7_get_rev_id(struct amdgpu_device *adev) +{ + u32 tmp; + + tmp = RREG32_SOC15(NBIO, 0, regRCC_STRAP0_RCC_DEV0_EPF0_STRAP0); + tmp &= RCC_STRAP0_RCC_DEV0_EPF0_STRAP0__STRAP_ATI_REV_ID_DEV0_F0_MASK; + tmp >>= RCC_STRAP0_RCC_DEV0_EPF0_STRAP0__STRAP_ATI_REV_ID_DEV0_F0__SHIFT; + + return tmp; +} + +static void nbio_v7_7_mc_access_enable(struct amdgpu_device *adev, bool enable) +{ + if (enable) + WREG32_SOC15(NBIO, 0, regBIF_BX1_BIF_FB_EN, + BIF_BX1_BIF_FB_EN__FB_READ_EN_MASK | + BIF_BX1_BIF_FB_EN__FB_WRITE_EN_MASK); + else + WREG32_SOC15(NBIO, 0, regBIF_BX1_BIF_FB_EN, 0); +} + +static u32 nbio_v7_7_get_memsize(struct amdgpu_device *adev) +{ + return RREG32_SOC15(NBIO, 0, regRCC_DEV0_EPF0_0_RCC_CONFIG_MEMSIZE); +} + +static void nbio_v7_7_sdma_doorbell_range(struct amdgpu_device *adev, int instance, + bool use_doorbell, int doorbell_index, + int doorbell_size) +{ + u32 reg = SOC15_REG_OFFSET(NBIO, 0, regGDC0_BIF_CSDMA_DOORBELL_RANGE); + u32 doorbell_range = RREG32_PCIE_PORT(reg); + + if (use_doorbell) { + doorbell_range = REG_SET_FIELD(doorbell_range, + GDC0_BIF_CSDMA_DOORBELL_RANGE, + OFFSET, doorbell_index); + doorbell_range = REG_SET_FIELD(doorbell_range, + GDC0_BIF_CSDMA_DOORBELL_RANGE, + SIZE, doorbell_size); + } else { + doorbell_range = REG_SET_FIELD(doorbell_range, + GDC0_BIF_SDMA0_DOORBELL_RANGE, + SIZE, 0); + } + + WREG32_PCIE_PORT(reg, doorbell_range); +} + +static void nbio_v7_7_vcn_doorbell_range(struct amdgpu_device *adev, bool use_doorbell, + int doorbell_index, int instance) +{ + u32 reg = SOC15_REG_OFFSET(NBIO, 0, regGDC0_BIF_VCN0_DOORBELL_RANGE); + u32 doorbell_range = RREG32_PCIE_PORT(reg); + + if (use_doorbell) { + doorbell_range = REG_SET_FIELD(doorbell_range, + GDC0_BIF_VCN0_DOORBELL_RANGE, OFFSET, + doorbell_index); + doorbell_range = REG_SET_FIELD(doorbell_range, + GDC0_BIF_VCN0_DOORBELL_RANGE, SIZE, 8); + } else { + doorbell_range = REG_SET_FIELD(doorbell_range, + GDC0_BIF_VCN0_DOORBELL_RANGE, SIZE, 0); + } + + WREG32_PCIE_PORT(reg, doorbell_range); +} + +static void nbio_v7_7_enable_doorbell_aperture(struct amdgpu_device *adev, + bool enable) +{ + u32 reg; + + reg = RREG32_SOC15(NBIO, 0, regRCC_DEV0_EPF0_0_RCC_DOORBELL_APER_EN); + reg = REG_SET_FIELD(reg, RCC_DEV0_EPF0_0_RCC_DOORBELL_APER_EN, + BIF_DOORBELL_APER_EN, enable ? 1 : 0); + + WREG32_SOC15(NBIO, 0, regRCC_DEV0_EPF0_0_RCC_DOORBELL_APER_EN, reg); +} + +static void nbio_v7_7_enable_doorbell_selfring_aperture(struct amdgpu_device *adev, + bool enable) +{ + u32 tmp = 0; + + if (enable) { + tmp = REG_SET_FIELD(tmp, BIF_BX_PF0_DOORBELL_SELFRING_GPA_APER_CNTL, + DOORBELL_SELFRING_GPA_APER_EN, 1) | + REG_SET_FIELD(tmp, BIF_BX_PF0_DOORBELL_SELFRING_GPA_APER_CNTL, + DOORBELL_SELFRING_GPA_APER_MODE, 1) | + REG_SET_FIELD(tmp, BIF_BX_PF0_DOORBELL_SELFRING_GPA_APER_CNTL, + DOORBELL_SELFRING_GPA_APER_SIZE, 0); + + WREG32_SOC15(NBIO, 0, + regBIF_BX_PF0_DOORBELL_SELFRING_GPA_APER_BASE_LOW, + lower_32_bits(adev->doorbell.base)); + WREG32_SOC15(NBIO, 0, + regBIF_BX_PF0_DOORBELL_SELFRING_GPA_APER_BASE_HIGH, + upper_32_bits(adev->doorbell.base)); + } + + WREG32_SOC15(NBIO, 0, regBIF_BX_PF0_DOORBELL_SELFRING_GPA_APER_CNTL, + tmp); +} + + +static void nbio_v7_7_ih_doorbell_range(struct amdgpu_device *adev, + bool use_doorbell, int doorbell_index) +{ + u32 ih_doorbell_range = RREG32_SOC15(NBIO, 0, + regGDC0_BIF_IH_DOORBELL_RANGE); + + if (use_doorbell) { + ih_doorbell_range = REG_SET_FIELD(ih_doorbell_range, + GDC0_BIF_IH_DOORBELL_RANGE, OFFSET, + doorbell_index); + ih_doorbell_range = REG_SET_FIELD(ih_doorbell_range, + GDC0_BIF_IH_DOORBELL_RANGE, SIZE, + 2); + } else { + ih_doorbell_range = REG_SET_FIELD(ih_doorbell_range, + GDC0_BIF_IH_DOORBELL_RANGE, SIZE, + 0); + } + + WREG32_SOC15(NBIO, 0, regGDC0_BIF_IH_DOORBELL_RANGE, + ih_doorbell_range); +} + +static void nbio_v7_7_ih_control(struct amdgpu_device *adev) +{ + u32 interrupt_cntl; + + /* setup interrupt control */ + WREG32_SOC15(NBIO, 0, regBIF_BX1_INTERRUPT_CNTL2, + adev->dummy_page_addr >> 8); + + interrupt_cntl = RREG32_SOC15(NBIO, 0, regBIF_BX1_INTERRUPT_CNTL); + /* + * INTERRUPT_CNTL__IH_DUMMY_RD_OVERRIDE_MASK=0 - dummy read disabled with msi, enabled without msi + * INTERRUPT_CNTL__IH_DUMMY_RD_OVERRIDE_MASK=1 - dummy read controlled by IH_DUMMY_RD_EN + */ + interrupt_cntl = REG_SET_FIELD(interrupt_cntl, BIF_BX1_INTERRUPT_CNTL, + IH_DUMMY_RD_OVERRIDE, 0); + + /* INTERRUPT_CNTL__IH_REQ_NONSNOOP_EN_MASK=1 if ring is in non-cacheable memory, e.g., vram */ + interrupt_cntl = REG_SET_FIELD(interrupt_cntl, BIF_BX1_INTERRUPT_CNTL, + IH_REQ_NONSNOOP_EN, 0); + + WREG32_SOC15(NBIO, 0, regBIF_BX1_INTERRUPT_CNTL, interrupt_cntl); +} + +static u32 nbio_v7_7_get_hdp_flush_req_offset(struct amdgpu_device *adev) +{ + return SOC15_REG_OFFSET(NBIO, 0, regBIF_BX_PF0_GPU_HDP_FLUSH_REQ); +} + +static u32 nbio_v7_7_get_hdp_flush_done_offset(struct amdgpu_device *adev) +{ + return SOC15_REG_OFFSET(NBIO, 0, regBIF_BX_PF0_GPU_HDP_FLUSH_DONE); +} + +static u32 nbio_v7_7_get_pcie_index_offset(struct amdgpu_device *adev) +{ + return SOC15_REG_OFFSET(NBIO, 0, regBIF_BX0_PCIE_INDEX2); +} + +static u32 nbio_v7_7_get_pcie_data_offset(struct amdgpu_device *adev) +{ + return SOC15_REG_OFFSET(NBIO, 0, regBIF_BX0_PCIE_DATA2); +} + +static u32 nbio_v7_7_get_pcie_port_index_offset(struct amdgpu_device *adev) +{ + return SOC15_REG_OFFSET(NBIO, 0, regBIF_BX_PF0_RSMU_INDEX); +} + +static u32 nbio_v7_7_get_pcie_port_data_offset(struct amdgpu_device *adev) +{ + return SOC15_REG_OFFSET(NBIO, 0, regBIF_BX_PF0_RSMU_DATA); +} + +const struct nbio_hdp_flush_reg nbio_v7_7_hdp_flush_reg = { + .ref_and_mask_cp0 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__CP0_MASK, + .ref_and_mask_cp1 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__CP1_MASK, + .ref_and_mask_cp2 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__CP2_MASK, + .ref_and_mask_cp3 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__CP3_MASK, + .ref_and_mask_cp4 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__CP4_MASK, + .ref_and_mask_cp5 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__CP5_MASK, + .ref_and_mask_cp6 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__CP6_MASK, + .ref_and_mask_cp7 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__CP7_MASK, + .ref_and_mask_cp8 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__CP8_MASK, + .ref_and_mask_cp9 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__CP9_MASK, + .ref_and_mask_sdma0 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__SDMA0_MASK, + .ref_and_mask_sdma1 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__SDMA1_MASK, +}; + +static void nbio_v7_7_init_registers(struct amdgpu_device *adev) +{ + uint32_t def, data; + + def = data = RREG32_SOC15(NBIO, 0, regBIF0_PCIE_MST_CTRL_3); + data = REG_SET_FIELD(data, BIF0_PCIE_MST_CTRL_3, + CI_SWUS_MAX_READ_REQUEST_SIZE_MODE, 1); + data = REG_SET_FIELD(data, BIF0_PCIE_MST_CTRL_3, + CI_SWUS_MAX_READ_REQUEST_SIZE_PRIV, 1); + + if (def != data) + WREG32_SOC15(NBIO, 0, regBIF0_PCIE_MST_CTRL_3, data); + +} + +static void nbio_v7_7_update_medium_grain_clock_gating(struct amdgpu_device *adev, + bool enable) +{ + uint32_t def, data; + + if (enable && !(adev->cg_flags & AMD_CG_SUPPORT_BIF_MGCG)) + return; + + def = data = RREG32_SOC15(NBIO, 0, regBIF0_CPM_CONTROL); + if (enable) { + data |= (BIF0_CPM_CONTROL__LCLK_DYN_GATE_ENABLE_MASK | + BIF0_CPM_CONTROL__TXCLK_DYN_GATE_ENABLE_MASK | + BIF0_CPM_CONTROL__TXCLK_LCNT_GATE_ENABLE_MASK | + BIF0_CPM_CONTROL__TXCLK_REGS_GATE_ENABLE_MASK | + BIF0_CPM_CONTROL__TXCLK_PRBS_GATE_ENABLE_MASK | + BIF0_CPM_CONTROL__REFCLK_REGS_GATE_ENABLE_MASK); + } else { + data &= ~(BIF0_CPM_CONTROL__LCLK_DYN_GATE_ENABLE_MASK | + BIF0_CPM_CONTROL__TXCLK_DYN_GATE_ENABLE_MASK | + BIF0_CPM_CONTROL__TXCLK_LCNT_GATE_ENABLE_MASK | + BIF0_CPM_CONTROL__TXCLK_REGS_GATE_ENABLE_MASK | + BIF0_CPM_CONTROL__TXCLK_PRBS_GATE_ENABLE_MASK | + BIF0_CPM_CONTROL__REFCLK_REGS_GATE_ENABLE_MASK); + } + + if (def != data) + WREG32_SOC15(NBIO, 0, regBIF0_CPM_CONTROL, data); +} + +static void nbio_v7_7_update_medium_grain_light_sleep(struct amdgpu_device *adev, + bool enable) +{ + uint32_t def, data; + + if (enable && !(adev->cg_flags & AMD_CG_SUPPORT_BIF_LS)) + return; + + def = data = RREG32_SOC15(NBIO, 0, regBIF0_PCIE_CNTL2); + if (enable) + data |= BIF0_PCIE_CNTL2__SLV_MEM_LS_EN_MASK; + else + data &= ~BIF0_PCIE_CNTL2__SLV_MEM_LS_EN_MASK; + + if (def != data) + WREG32_SOC15(NBIO, 0, regBIF0_PCIE_CNTL2, data); + + def = data = RREG32_SOC15(NBIO, 0, regBIF0_PCIE_TX_POWER_CTRL_1); + if (enable) { + data |= (BIF0_PCIE_TX_POWER_CTRL_1__MST_MEM_LS_EN_MASK | + BIF0_PCIE_TX_POWER_CTRL_1__REPLAY_MEM_LS_EN_MASK); + } else { + data &= ~(BIF0_PCIE_TX_POWER_CTRL_1__MST_MEM_LS_EN_MASK | + BIF0_PCIE_TX_POWER_CTRL_1__REPLAY_MEM_LS_EN_MASK); + } + + if (def != data) + WREG32_SOC15(NBIO, 0, regBIF0_PCIE_TX_POWER_CTRL_1, data); +} + +static void nbio_v7_7_get_clockgating_state(struct amdgpu_device *adev, + u64 *flags) +{ + uint32_t data; + + /* AMD_CG_SUPPORT_BIF_MGCG */ + data = RREG32_SOC15(NBIO, 0, regBIF0_CPM_CONTROL); + if (data & BIF0_CPM_CONTROL__LCLK_DYN_GATE_ENABLE_MASK) + *flags |= AMD_CG_SUPPORT_BIF_MGCG; + + /* AMD_CG_SUPPORT_BIF_LS */ + data = RREG32_SOC15(NBIO, 0, regBIF0_PCIE_CNTL2); + if (data & BIF0_PCIE_CNTL2__SLV_MEM_LS_EN_MASK) + *flags |= AMD_CG_SUPPORT_BIF_LS; +} + +const struct amdgpu_nbio_funcs nbio_v7_7_funcs = { + .get_hdp_flush_req_offset = nbio_v7_7_get_hdp_flush_req_offset, + .get_hdp_flush_done_offset = nbio_v7_7_get_hdp_flush_done_offset, + .get_pcie_index_offset = nbio_v7_7_get_pcie_index_offset, + .get_pcie_data_offset = nbio_v7_7_get_pcie_data_offset, + .get_pcie_port_index_offset = nbio_v7_7_get_pcie_port_index_offset, + .get_pcie_port_data_offset = nbio_v7_7_get_pcie_port_data_offset, + .get_rev_id = nbio_v7_7_get_rev_id, + .mc_access_enable = nbio_v7_7_mc_access_enable, + .get_memsize = nbio_v7_7_get_memsize, + .sdma_doorbell_range = nbio_v7_7_sdma_doorbell_range, + .vcn_doorbell_range = nbio_v7_7_vcn_doorbell_range, + .enable_doorbell_aperture = nbio_v7_7_enable_doorbell_aperture, + .enable_doorbell_selfring_aperture = nbio_v7_7_enable_doorbell_selfring_aperture, + .ih_doorbell_range = nbio_v7_7_ih_doorbell_range, + .update_medium_grain_clock_gating = nbio_v7_7_update_medium_grain_clock_gating, + .update_medium_grain_light_sleep = nbio_v7_7_update_medium_grain_light_sleep, + .get_clockgating_state = nbio_v7_7_get_clockgating_state, + .ih_control = nbio_v7_7_ih_control, + .init_registers = nbio_v7_7_init_registers, + .remap_hdp_registers = nbio_v7_7_remap_hdp_registers, +}; diff --git a/drivers/gpu/drm/amd/amdgpu/nbio_v7_7.h b/drivers/gpu/drm/amd/amdgpu/nbio_v7_7.h new file mode 100644 index 000000000000..2a33b256ba81 --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/nbio_v7_7.h @@ -0,0 +1,33 @@ +/* + * Copyright 2021 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#ifndef __NBIO_V7_7_H__ +#define __NBIO_V7_7_H__ + +#include "soc15_common.h" + +extern const struct nbio_hdp_flush_reg nbio_v7_7_hdp_flush_reg; +extern const struct amdgpu_nbio_funcs nbio_v7_7_funcs; +extern const struct amdgpu_nbio_ras_funcs nbio_v7_7_ras_funcs; + +#endif diff --git a/drivers/gpu/drm/amd/amdgpu/nv.c b/drivers/gpu/drm/amd/amdgpu/nv.c index 2ec1ffb36b1f..b3fba8dea63c 100644 --- a/drivers/gpu/drm/amd/amdgpu/nv.c +++ b/drivers/gpu/drm/amd/amdgpu/nv.c @@ -170,6 +170,7 @@ static const struct amdgpu_video_codec_info yc_video_codecs_decode_array[] = { {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_HEVC, 8192, 4352, 186)}, {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_VP9, 8192, 4352, 0)}, {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_JPEG, 4096, 4096, 0)}, + {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_AV1, 8192, 4352, 0)}, }; static const struct amdgpu_video_codecs yc_video_codecs_decode = { @@ -204,6 +205,7 @@ static int nv_query_video_codecs(struct amdgpu_device *adev, bool encode, *codecs = &sc_video_codecs_decode; return 0; case IP_VERSION(3, 1, 1): + case IP_VERSION(3, 1, 2): if (encode) *codecs = &nv_video_codecs_encode; else @@ -258,21 +260,6 @@ static u64 nv_pcie_rreg64(struct amdgpu_device *adev, u32 reg) return amdgpu_device_indirect_rreg64(adev, address, data, reg); } -static u32 nv_pcie_port_rreg(struct amdgpu_device *adev, u32 reg) -{ - unsigned long flags, address, data; - u32 r; - address = adev->nbio.funcs->get_pcie_port_index_offset(adev); - data = adev->nbio.funcs->get_pcie_port_data_offset(adev); - - spin_lock_irqsave(&adev->pcie_idx_lock, flags); - WREG32(address, reg * 4); - (void)RREG32(address); - r = RREG32(data); - spin_unlock_irqrestore(&adev->pcie_idx_lock, flags); - return r; -} - static void nv_pcie_wreg64(struct amdgpu_device *adev, u32 reg, u64 v) { unsigned long address, data; @@ -283,21 +270,6 @@ static void nv_pcie_wreg64(struct amdgpu_device *adev, u32 reg, u64 v) amdgpu_device_indirect_wreg64(adev, address, data, reg, v); } -static void nv_pcie_port_wreg(struct amdgpu_device *adev, u32 reg, u32 v) -{ - unsigned long flags, address, data; - - address = adev->nbio.funcs->get_pcie_port_index_offset(adev); - data = adev->nbio.funcs->get_pcie_port_data_offset(adev); - - spin_lock_irqsave(&adev->pcie_idx_lock, flags); - WREG32(address, reg * 4); - (void)RREG32(address); - WREG32(data, v); - (void)RREG32(data); - spin_unlock_irqrestore(&adev->pcie_idx_lock, flags); -} - static u32 nv_didt_rreg(struct amdgpu_device *adev, u32 reg) { unsigned long flags, address, data; @@ -360,38 +332,6 @@ static bool nv_read_disabled_bios(struct amdgpu_device *adev) return false; } -static bool nv_read_bios_from_rom(struct amdgpu_device *adev, - u8 *bios, u32 length_bytes) -{ - u32 *dw_ptr; - u32 i, length_dw; - u32 rom_index_offset, rom_data_offset; - - if (bios == NULL) - return false; - if (length_bytes == 0) - return false; - /* APU vbios image is part of sbios image */ - if (adev->flags & AMD_IS_APU) - return false; - - dw_ptr = (u32 *)bios; - length_dw = ALIGN(length_bytes, 4) / 4; - - rom_index_offset = - adev->smuio.funcs->get_rom_index_offset(adev); - rom_data_offset = - adev->smuio.funcs->get_rom_data_offset(adev); - - /* set rom index to 0 */ - WREG32(rom_index_offset, 0); - /* read out the rom data */ - for (i = 0; i < length_dw; i++) - dw_ptr[i] = RREG32(rom_data_offset); - - return true; -} - static struct soc15_allowed_register_entry nv_allowed_read_registers[] = { { SOC15_REG_ENTRY(GC, 0, mmGRBM_STATUS)}, { SOC15_REG_ENTRY(GC, 0, mmGRBM_STATUS2)}, @@ -453,9 +393,9 @@ static int nv_read_register(struct amdgpu_device *adev, u32 se_num, *value = 0; for (i = 0; i < ARRAY_SIZE(nv_allowed_read_registers); i++) { en = &nv_allowed_read_registers[i]; - if ((i == 7 && (adev->sdma.num_instances == 1)) || /* some asics don't have SDMA1 */ - reg_offset != - (adev->reg_offset[en->hwip][en->inst][en->seg] + en->reg_offset)) + if (adev->reg_offset[en->hwip][en->inst] && + reg_offset != (adev->reg_offset[en->hwip][en->inst][en->seg] + + en->reg_offset)) continue; *value = nv_get_register_value(adev, @@ -515,6 +455,8 @@ nv_asic_reset_method(struct amdgpu_device *adev) case IP_VERSION(11, 5, 0): case IP_VERSION(13, 0, 1): case IP_VERSION(13, 0, 3): + case IP_VERSION(13, 0, 5): + case IP_VERSION(13, 0, 8): return AMD_RESET_METHOD_MODE2; case IP_VERSION(11, 0, 7): case IP_VERSION(11, 0, 11): @@ -584,7 +526,7 @@ static void nv_pcie_gen3_enable(struct amdgpu_device *adev) static void nv_program_aspm(struct amdgpu_device *adev) { - if (!amdgpu_aspm) + if (!amdgpu_device_should_use_aspm(adev)) return; if (!(adev->flags & AMD_IS_APU) && @@ -666,7 +608,12 @@ static void nv_init_doorbell_index(struct amdgpu_device *adev) adev->doorbell_index.userqueue_end = AMDGPU_NAVI10_DOORBELL_USERQUEUE_END; adev->doorbell_index.gfx_ring0 = AMDGPU_NAVI10_DOORBELL_GFX_RING0; adev->doorbell_index.gfx_ring1 = AMDGPU_NAVI10_DOORBELL_GFX_RING1; - adev->doorbell_index.mes_ring = AMDGPU_NAVI10_DOORBELL_MES_RING; + adev->doorbell_index.gfx_userqueue_start = + AMDGPU_NAVI10_DOORBELL_GFX_USERQUEUE_START; + adev->doorbell_index.gfx_userqueue_end = + AMDGPU_NAVI10_DOORBELL_GFX_USERQUEUE_END; + adev->doorbell_index.mes_ring0 = AMDGPU_NAVI10_DOORBELL_MES_RING0; + adev->doorbell_index.mes_ring1 = AMDGPU_NAVI10_DOORBELL_MES_RING1; adev->doorbell_index.sdma_engine[0] = AMDGPU_NAVI10_DOORBELL_sDMA_ENGINE0; adev->doorbell_index.sdma_engine[1] = AMDGPU_NAVI10_DOORBELL_sDMA_ENGINE1; adev->doorbell_index.sdma_engine[2] = AMDGPU_NAVI10_DOORBELL_sDMA_ENGINE2; @@ -699,7 +646,8 @@ static int nv_update_umd_stable_pstate(struct amdgpu_device *adev, adev->gfx.funcs->update_perfmon_mgcg(adev, !enter); if (!(adev->flags & AMD_IS_APU) && - (adev->nbio.funcs->enable_aspm)) + (adev->nbio.funcs->enable_aspm) && + amdgpu_device_should_use_aspm(adev)) adev->nbio.funcs->enable_aspm(adev, !enter); return 0; @@ -708,7 +656,7 @@ static int nv_update_umd_stable_pstate(struct amdgpu_device *adev, static const struct amdgpu_asic_funcs nv_asic_funcs = { .read_disabled_bios = &nv_read_disabled_bios, - .read_bios_from_rom = &nv_read_bios_from_rom, + .read_bios_from_rom = &amdgpu_soc15_read_bios_from_rom, .read_register = &nv_read_register, .reset = &nv_asic_reset, .reset_method = &nv_asic_reset_method, @@ -742,8 +690,8 @@ static int nv_common_early_init(void *handle) adev->pcie_wreg = &nv_pcie_wreg; adev->pcie_rreg64 = &nv_pcie_rreg64; adev->pcie_wreg64 = &nv_pcie_wreg64; - adev->pciep_rreg = &nv_pcie_port_rreg; - adev->pciep_wreg = &nv_pcie_port_wreg; + adev->pciep_rreg = amdgpu_device_pcie_port_rreg; + adev->pciep_wreg = amdgpu_device_pcie_port_wreg; /* TODO: will add them during VCN v2 implementation */ adev->uvd_ctx_rreg = NULL; @@ -964,10 +912,63 @@ static int nv_common_early_init(void *handle) adev->external_rev_id = adev->rev_id + 0x01; break; case IP_VERSION(10, 1, 3): + case IP_VERSION(10, 1, 4): adev->cg_flags = 0; adev->pg_flags = 0; adev->external_rev_id = adev->rev_id + 0x82; break; + case IP_VERSION(10, 3, 6): + adev->cg_flags = AMD_CG_SUPPORT_GFX_MGCG | + AMD_CG_SUPPORT_GFX_MGLS | + AMD_CG_SUPPORT_GFX_CGCG | + AMD_CG_SUPPORT_GFX_CGLS | + AMD_CG_SUPPORT_GFX_3D_CGCG | + AMD_CG_SUPPORT_GFX_3D_CGLS | + AMD_CG_SUPPORT_GFX_RLC_LS | + AMD_CG_SUPPORT_GFX_CP_LS | + AMD_CG_SUPPORT_GFX_FGCG | + AMD_CG_SUPPORT_MC_MGCG | + AMD_CG_SUPPORT_MC_LS | + AMD_CG_SUPPORT_SDMA_LS | + AMD_CG_SUPPORT_HDP_MGCG | + AMD_CG_SUPPORT_HDP_LS | + AMD_CG_SUPPORT_ATHUB_MGCG | + AMD_CG_SUPPORT_ATHUB_LS | + AMD_CG_SUPPORT_IH_CG | + AMD_CG_SUPPORT_VCN_MGCG | + AMD_CG_SUPPORT_JPEG_MGCG; + adev->pg_flags = AMD_PG_SUPPORT_GFX_PG | + AMD_PG_SUPPORT_VCN | + AMD_PG_SUPPORT_VCN_DPG | + AMD_PG_SUPPORT_JPEG; + adev->external_rev_id = adev->rev_id + 0x01; + break; + case IP_VERSION(10, 3, 7): + adev->cg_flags = AMD_CG_SUPPORT_GFX_MGCG | + AMD_CG_SUPPORT_GFX_MGLS | + AMD_CG_SUPPORT_GFX_CGCG | + AMD_CG_SUPPORT_GFX_CGLS | + AMD_CG_SUPPORT_GFX_3D_CGCG | + AMD_CG_SUPPORT_GFX_3D_CGLS | + AMD_CG_SUPPORT_GFX_RLC_LS | + AMD_CG_SUPPORT_GFX_CP_LS | + AMD_CG_SUPPORT_GFX_FGCG | + AMD_CG_SUPPORT_MC_MGCG | + AMD_CG_SUPPORT_MC_LS | + AMD_CG_SUPPORT_SDMA_LS | + AMD_CG_SUPPORT_HDP_MGCG | + AMD_CG_SUPPORT_HDP_LS | + AMD_CG_SUPPORT_ATHUB_MGCG | + AMD_CG_SUPPORT_ATHUB_LS | + AMD_CG_SUPPORT_IH_CG | + AMD_CG_SUPPORT_VCN_MGCG | + AMD_CG_SUPPORT_JPEG_MGCG; + adev->pg_flags = AMD_PG_SUPPORT_VCN | + AMD_PG_SUPPORT_VCN_DPG | + AMD_PG_SUPPORT_JPEG | + AMD_PG_SUPPORT_GFX_PG; + adev->external_rev_id = adev->rev_id + 0x01; + break; default: /* FIXME: not supported yet */ return -EINVAL; @@ -1120,7 +1121,7 @@ static int nv_common_set_powergating_state(void *handle, return 0; } -static void nv_common_get_clockgating_state(void *handle, u32 *flags) +static void nv_common_get_clockgating_state(void *handle, u64 *flags) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; diff --git a/drivers/gpu/drm/amd/amdgpu/psp_gfx_if.h b/drivers/gpu/drm/amd/amdgpu/psp_gfx_if.h index dd0dce254901..22c775f39119 100644 --- a/drivers/gpu/drm/amd/amdgpu/psp_gfx_if.h +++ b/drivers/gpu/drm/amd/amdgpu/psp_gfx_if.h @@ -258,7 +258,40 @@ enum psp_gfx_fw_type { GFX_FW_TYPE_SDMA6 = 56, /* SDMA6 MI */ GFX_FW_TYPE_SDMA7 = 57, /* SDMA7 MI */ GFX_FW_TYPE_VCN1 = 58, /* VCN1 MI */ + GFX_FW_TYPE_CAP = 62, /* CAP_FW */ + GFX_FW_TYPE_SE2_TAP_DELAYS = 65, /* SE2 TAP DELAYS NV */ + GFX_FW_TYPE_SE3_TAP_DELAYS = 66, /* SE3 TAP DELAYS NV */ GFX_FW_TYPE_REG_LIST = 67, /* REG_LIST MI */ + GFX_FW_TYPE_IMU_I = 68, /* IMU Instruction FW SOC21 */ + GFX_FW_TYPE_IMU_D = 69, /* IMU Data FW SOC21 */ + GFX_FW_TYPE_LSDMA = 70, /* LSDMA FW SOC21 */ + GFX_FW_TYPE_SDMA_UCODE_TH0 = 71, /* SDMA Thread 0/CTX SOC21 */ + GFX_FW_TYPE_SDMA_UCODE_TH1 = 72, /* SDMA Thread 1/CTL SOC21 */ + GFX_FW_TYPE_PPTABLE = 73, /* PPTABLE SOC21 */ + GFX_FW_TYPE_DISCRETE_USB4 = 74, /* dUSB4 FW SOC21 */ + GFX_FW_TYPE_TA = 75, /* SRIOV TA FW UUID SOC21 */ + GFX_FW_TYPE_RS64_MES = 76, /* RS64 MES ucode SOC21 */ + GFX_FW_TYPE_RS64_MES_STACK = 77, /* RS64 MES stack ucode SOC21 */ + GFX_FW_TYPE_RS64_KIQ = 78, /* RS64 KIQ ucode SOC21 */ + GFX_FW_TYPE_RS64_KIQ_STACK = 79, /* RS64 KIQ Heap stack SOC21 */ + GFX_FW_TYPE_ISP_DATA = 80, /* ISP DATA SOC21 */ + GFX_FW_TYPE_CP_MES_KIQ = 81, /* MES KIQ ucode SOC21 */ + GFX_FW_TYPE_MES_KIQ_STACK = 82, /* MES KIQ stack SOC21 */ + GFX_FW_TYPE_UMSCH_DATA = 83, /* User Mode Scheduler Data SOC21 */ + GFX_FW_TYPE_UMSCH_UCODE = 84, /* User Mode Scheduler Ucode SOC21 */ + GFX_FW_TYPE_UMSCH_CMD_BUFFER = 85, /* User Mode Scheduler Command Buffer SOC21 */ + GFX_FW_TYPE_USB_DP_COMBO_PHY = 86, /* USB-Display port Combo SOC21 */ + GFX_FW_TYPE_RS64_PFP = 87, /* RS64 PFP SOC21 */ + GFX_FW_TYPE_RS64_ME = 88, /* RS64 ME SOC21 */ + GFX_FW_TYPE_RS64_MEC = 89, /* RS64 MEC SOC21 */ + GFX_FW_TYPE_RS64_PFP_P0_STACK = 90, /* RS64 PFP stack P0 SOC21 */ + GFX_FW_TYPE_RS64_PFP_P1_STACK = 91, /* RS64 PFP stack P1 SOC21 */ + GFX_FW_TYPE_RS64_ME_P0_STACK = 92, /* RS64 ME stack P0 SOC21 */ + GFX_FW_TYPE_RS64_ME_P1_STACK = 93, /* RS64 ME stack P1 SOC21 */ + GFX_FW_TYPE_RS64_MEC_P0_STACK = 94, /* RS64 MEC stack P0 SOC21 */ + GFX_FW_TYPE_RS64_MEC_P1_STACK = 95, /* RS64 MEC stack P1 SOC21 */ + GFX_FW_TYPE_RS64_MEC_P2_STACK = 96, /* RS64 MEC stack P2 SOC21 */ + GFX_FW_TYPE_RS64_MEC_P3_STACK = 97, /* RS64 MEC stack P3 SOC21 */ GFX_FW_TYPE_MAX }; diff --git a/drivers/gpu/drm/amd/amdgpu/psp_v11_0.c b/drivers/gpu/drm/amd/amdgpu/psp_v11_0.c index d0e76b36d4ab..9518b4394a6e 100644 --- a/drivers/gpu/drm/amd/amdgpu/psp_v11_0.c +++ b/drivers/gpu/drm/amd/amdgpu/psp_v11_0.c @@ -53,11 +53,13 @@ MODULE_FIRMWARE("amdgpu/navi14_ta.bin"); MODULE_FIRMWARE("amdgpu/navi12_sos.bin"); MODULE_FIRMWARE("amdgpu/navi12_asd.bin"); MODULE_FIRMWARE("amdgpu/navi12_ta.bin"); +MODULE_FIRMWARE("amdgpu/navi12_cap.bin"); MODULE_FIRMWARE("amdgpu/arcturus_sos.bin"); MODULE_FIRMWARE("amdgpu/arcturus_asd.bin"); MODULE_FIRMWARE("amdgpu/arcturus_ta.bin"); MODULE_FIRMWARE("amdgpu/sienna_cichlid_sos.bin"); MODULE_FIRMWARE("amdgpu/sienna_cichlid_ta.bin"); +MODULE_FIRMWARE("amdgpu/sienna_cichlid_cap.bin"); MODULE_FIRMWARE("amdgpu/navy_flounder_sos.bin"); MODULE_FIRMWARE("amdgpu/navy_flounder_ta.bin"); MODULE_FIRMWARE("amdgpu/vangogh_asd.bin"); @@ -177,8 +179,6 @@ static int psp_v11_0_init_microcode(struct psp_context *psp) err = psp_init_asd_microcode(psp, chip_name); if (err) return err; - if (amdgpu_sriov_vf(adev)) - break; snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_ta.bin", chip_name); err = request_firmware(&adev->psp.ta_fw, fw_name, adev->dev); if (err) { diff --git a/drivers/gpu/drm/amd/amdgpu/psp_v12_0.c b/drivers/gpu/drm/amd/amdgpu/psp_v12_0.c index a2588200ea58..0b2ac418e4ac 100644 --- a/drivers/gpu/drm/amd/amdgpu/psp_v12_0.c +++ b/drivers/gpu/drm/amd/amdgpu/psp_v12_0.c @@ -101,6 +101,16 @@ static int psp_v12_0_init_microcode(struct psp_context *psp) adev->psp.dtm_context.context.bin_desc.start_addr = (uint8_t *)adev->psp.hdcp_context.context.bin_desc.start_addr + le32_to_cpu(ta_hdr->dtm.offset_bytes); + + if (adev->apu_flags & AMD_APU_IS_RENOIR) { + adev->psp.securedisplay_context.context.bin_desc.fw_version = + le32_to_cpu(ta_hdr->securedisplay.fw_version); + adev->psp.securedisplay_context.context.bin_desc.size_bytes = + le32_to_cpu(ta_hdr->securedisplay.size_bytes); + adev->psp.securedisplay_context.context.bin_desc.start_addr = + (uint8_t *)adev->psp.hdcp_context.context.bin_desc.start_addr + + le32_to_cpu(ta_hdr->securedisplay.offset_bytes); + } } return 0; diff --git a/drivers/gpu/drm/amd/amdgpu/psp_v13_0.c b/drivers/gpu/drm/amd/amdgpu/psp_v13_0.c index 17655bc6d2f1..21d822b1d589 100644 --- a/drivers/gpu/drm/amd/amdgpu/psp_v13_0.c +++ b/drivers/gpu/drm/amd/amdgpu/psp_v13_0.c @@ -20,6 +20,8 @@ * OTHER DEALINGS IN THE SOFTWARE. * */ +#include <drm/drm_drv.h> +#include <linux/vmalloc.h> #include "amdgpu.h" #include "amdgpu_psp.h" #include "amdgpu_ucode.h" @@ -31,9 +33,18 @@ MODULE_FIRMWARE("amdgpu/aldebaran_sos.bin"); MODULE_FIRMWARE("amdgpu/aldebaran_ta.bin"); -MODULE_FIRMWARE("amdgpu/yellow_carp_asd.bin"); +MODULE_FIRMWARE("amdgpu/aldebaran_cap.bin"); MODULE_FIRMWARE("amdgpu/yellow_carp_toc.bin"); MODULE_FIRMWARE("amdgpu/yellow_carp_ta.bin"); +MODULE_FIRMWARE("amdgpu/psp_13_0_5_toc.bin"); +MODULE_FIRMWARE("amdgpu/psp_13_0_5_ta.bin"); +MODULE_FIRMWARE("amdgpu/psp_13_0_8_toc.bin"); +MODULE_FIRMWARE("amdgpu/psp_13_0_8_ta.bin"); +MODULE_FIRMWARE("amdgpu/psp_13_0_0_sos.bin"); +MODULE_FIRMWARE("amdgpu/psp_13_0_0_ta.bin"); +MODULE_FIRMWARE("amdgpu/psp_13_0_7_sos.bin"); +MODULE_FIRMWARE("amdgpu/psp_13_0_7_ta.bin"); +MODULE_FIRMWARE("amdgpu/psp_13_0_10_sos.bin"); /* For large FW files the time to complete can be very long */ #define USBC_PD_POLLING_LIMIT_S 240 @@ -41,10 +52,23 @@ MODULE_FIRMWARE("amdgpu/yellow_carp_ta.bin"); /* Read USB-PD from LFB */ #define GFX_CMD_USB_PD_USE_LFB 0x480 +/* VBIOS gfl defines */ +#define MBOX_READY_MASK 0x80000000 +#define MBOX_STATUS_MASK 0x0000FFFF +#define MBOX_COMMAND_MASK 0x00FF0000 +#define MBOX_READY_FLAG 0x80000000 +#define C2PMSG_CMD_SPI_UPDATE_ROM_IMAGE_ADDR_LO 0x2 +#define C2PMSG_CMD_SPI_UPDATE_ROM_IMAGE_ADDR_HI 0x3 +#define C2PMSG_CMD_SPI_UPDATE_FLASH_IMAGE 0x4 + +/* memory training timeout define */ +#define MEM_TRAIN_SEND_MSG_TIMEOUT_US 3000000 + static int psp_v13_0_init_microcode(struct psp_context *psp) { struct amdgpu_device *adev = psp->adev; const char *chip_name; + char ucode_prefix[30]; int err = 0; switch (adev->ip_versions[MP0_HWIP][0]) { @@ -56,25 +80,41 @@ static int psp_v13_0_init_microcode(struct psp_context *psp) chip_name = "yellow_carp"; break; default: - BUG(); + amdgpu_ucode_ip_version_decode(adev, MP0_HWIP, ucode_prefix, sizeof(ucode_prefix)); + chip_name = ucode_prefix; + break; } + switch (adev->ip_versions[MP0_HWIP][0]) { case IP_VERSION(13, 0, 2): err = psp_init_sos_microcode(psp, chip_name); if (err) return err; - err = psp_init_ta_microcode(&adev->psp, chip_name); - if (err) - return err; + /* It's not necessary to load ras ta on Guest side */ + if (!amdgpu_sriov_vf(adev)) { + err = psp_init_ta_microcode(&adev->psp, chip_name); + if (err) + return err; + } break; case IP_VERSION(13, 0, 1): case IP_VERSION(13, 0, 3): - err = psp_init_asd_microcode(psp, chip_name); + case IP_VERSION(13, 0, 5): + case IP_VERSION(13, 0, 8): + err = psp_init_toc_microcode(psp, chip_name); if (err) return err; - err = psp_init_toc_microcode(psp, chip_name); + err = psp_init_ta_microcode(psp, chip_name); + if (err) + return err; + break; + case IP_VERSION(13, 0, 0): + case IP_VERSION(13, 0, 7): + case IP_VERSION(13, 0, 10): + err = psp_init_sos_microcode(psp, chip_name); if (err) return err; + /* It's not necessary to load ras ta on Guest side */ err = psp_init_ta_microcode(psp, chip_name); if (err) return err; @@ -159,6 +199,11 @@ static int psp_v13_0_bootloader_load_kdb(struct psp_context *psp) return psp_v13_0_bootloader_load_component(psp, &psp->kdb, PSP_BL__LOAD_KEY_DATABASE); } +static int psp_v13_0_bootloader_load_spl(struct psp_context *psp) +{ + return psp_v13_0_bootloader_load_component(psp, &psp->kdb, PSP_BL__LOAD_TOS_SPL_TABLE); +} + static int psp_v13_0_bootloader_load_sysdrv(struct psp_context *psp) { return psp_v13_0_bootloader_load_component(psp, &psp->sys, PSP_BL__LOAD_SYSDRV); @@ -179,6 +224,12 @@ static int psp_v13_0_bootloader_load_dbg_drv(struct psp_context *psp) return psp_v13_0_bootloader_load_component(psp, &psp->dbg_drv, PSP_BL__LOAD_DBGDRV); } +static int psp_v13_0_bootloader_load_ras_drv(struct psp_context *psp) +{ + return psp_v13_0_bootloader_load_component(psp, &psp->ras_drv, PSP_BL__LOAD_RASDRV); +} + + static int psp_v13_0_bootloader_load_sos(struct psp_context *psp) { int ret; @@ -381,6 +432,159 @@ static void psp_v13_0_ring_set_wptr(struct psp_context *psp, uint32_t value) WREG32_SOC15(MP0, 0, regMP0_SMN_C2PMSG_67, value); } +static int psp_v13_0_memory_training_send_msg(struct psp_context *psp, int msg) +{ + int ret; + int i; + uint32_t data_32; + int max_wait; + struct amdgpu_device *adev = psp->adev; + + data_32 = (psp->mem_train_ctx.c2p_train_data_offset >> 20); + WREG32_SOC15(MP0, 0, regMP0_SMN_C2PMSG_36, data_32); + WREG32_SOC15(MP0, 0, regMP0_SMN_C2PMSG_35, msg); + + max_wait = MEM_TRAIN_SEND_MSG_TIMEOUT_US / adev->usec_timeout; + for (i = 0; i < max_wait; i++) { + ret = psp_wait_for(psp, SOC15_REG_OFFSET(MP0, 0, regMP0_SMN_C2PMSG_35), + 0x80000000, 0x80000000, false); + if (ret == 0) + break; + } + if (i < max_wait) + ret = 0; + else + ret = -ETIME; + + dev_dbg(adev->dev, "training %s %s, cost %d @ %d ms\n", + (msg == PSP_BL__DRAM_SHORT_TRAIN) ? "short" : "long", + (ret == 0) ? "succeed" : "failed", + i, adev->usec_timeout/1000); + return ret; +} + + +static int psp_v13_0_memory_training(struct psp_context *psp, uint32_t ops) +{ + struct psp_memory_training_context *ctx = &psp->mem_train_ctx; + uint32_t *pcache = (uint32_t *)ctx->sys_cache; + struct amdgpu_device *adev = psp->adev; + uint32_t p2c_header[4]; + uint32_t sz; + void *buf; + int ret, idx; + + if (ctx->init == PSP_MEM_TRAIN_NOT_SUPPORT) { + dev_dbg(adev->dev, "Memory training is not supported.\n"); + return 0; + } else if (ctx->init != PSP_MEM_TRAIN_INIT_SUCCESS) { + dev_err(adev->dev, "Memory training initialization failure.\n"); + return -EINVAL; + } + + if (psp_v13_0_is_sos_alive(psp)) { + dev_dbg(adev->dev, "SOS is alive, skip memory training.\n"); + return 0; + } + + amdgpu_device_vram_access(adev, ctx->p2c_train_data_offset, p2c_header, sizeof(p2c_header), false); + dev_dbg(adev->dev, "sys_cache[%08x,%08x,%08x,%08x] p2c_header[%08x,%08x,%08x,%08x]\n", + pcache[0], pcache[1], pcache[2], pcache[3], + p2c_header[0], p2c_header[1], p2c_header[2], p2c_header[3]); + + if (ops & PSP_MEM_TRAIN_SEND_SHORT_MSG) { + dev_dbg(adev->dev, "Short training depends on restore.\n"); + ops |= PSP_MEM_TRAIN_RESTORE; + } + + if ((ops & PSP_MEM_TRAIN_RESTORE) && + pcache[0] != MEM_TRAIN_SYSTEM_SIGNATURE) { + dev_dbg(adev->dev, "sys_cache[0] is invalid, restore depends on save.\n"); + ops |= PSP_MEM_TRAIN_SAVE; + } + + if (p2c_header[0] == MEM_TRAIN_SYSTEM_SIGNATURE && + !(pcache[0] == MEM_TRAIN_SYSTEM_SIGNATURE && + pcache[3] == p2c_header[3])) { + dev_dbg(adev->dev, "sys_cache is invalid or out-of-date, need save training data to sys_cache.\n"); + ops |= PSP_MEM_TRAIN_SAVE; + } + + if ((ops & PSP_MEM_TRAIN_SAVE) && + p2c_header[0] != MEM_TRAIN_SYSTEM_SIGNATURE) { + dev_dbg(adev->dev, "p2c_header[0] is invalid, save depends on long training.\n"); + ops |= PSP_MEM_TRAIN_SEND_LONG_MSG; + } + + if (ops & PSP_MEM_TRAIN_SEND_LONG_MSG) { + ops &= ~PSP_MEM_TRAIN_SEND_SHORT_MSG; + ops |= PSP_MEM_TRAIN_SAVE; + } + + dev_dbg(adev->dev, "Memory training ops:%x.\n", ops); + + if (ops & PSP_MEM_TRAIN_SEND_LONG_MSG) { + /* + * Long training will encroach a certain amount on the bottom of VRAM; + * save the content from the bottom of VRAM to system memory + * before training, and restore it after training to avoid + * VRAM corruption. + */ + sz = GDDR6_MEM_TRAINING_ENCROACHED_SIZE; + + if (adev->gmc.visible_vram_size < sz || !adev->mman.aper_base_kaddr) { + dev_err(adev->dev, "visible_vram_size %llx or aper_base_kaddr %p is not initialized.\n", + adev->gmc.visible_vram_size, + adev->mman.aper_base_kaddr); + return -EINVAL; + } + + buf = vmalloc(sz); + if (!buf) { + dev_err(adev->dev, "failed to allocate system memory.\n"); + return -ENOMEM; + } + + if (drm_dev_enter(adev_to_drm(adev), &idx)) { + memcpy_fromio(buf, adev->mman.aper_base_kaddr, sz); + ret = psp_v13_0_memory_training_send_msg(psp, PSP_BL__DRAM_LONG_TRAIN); + if (ret) { + DRM_ERROR("Send long training msg failed.\n"); + vfree(buf); + drm_dev_exit(idx); + return ret; + } + + memcpy_toio(adev->mman.aper_base_kaddr, buf, sz); + adev->hdp.funcs->flush_hdp(adev, NULL); + vfree(buf); + drm_dev_exit(idx); + } else { + vfree(buf); + return -ENODEV; + } + } + + if (ops & PSP_MEM_TRAIN_SAVE) { + amdgpu_device_vram_access(psp->adev, ctx->p2c_train_data_offset, ctx->sys_cache, ctx->train_data_size, false); + } + + if (ops & PSP_MEM_TRAIN_RESTORE) { + amdgpu_device_vram_access(psp->adev, ctx->c2p_train_data_offset, ctx->sys_cache, ctx->train_data_size, true); + } + + if (ops & PSP_MEM_TRAIN_SEND_SHORT_MSG) { + ret = psp_v13_0_memory_training_send_msg(psp, (amdgpu_force_long_training > 0) ? + PSP_BL__DRAM_LONG_TRAIN : PSP_BL__DRAM_SHORT_TRAIN); + if (ret) { + dev_err(adev->dev, "send training msg failed.\n"); + return ret; + } + } + ctx->training_cnt++; + return 0; +} + static int psp_v13_0_load_usbc_pd_fw(struct psp_context *psp, uint64_t fw_pri_mc_addr) { struct amdgpu_device *adev = psp->adev; @@ -439,13 +643,90 @@ static int psp_v13_0_read_usbc_pd_fw(struct psp_context *psp, uint32_t *fw_ver) return ret; } +static int psp_v13_0_exec_spi_cmd(struct psp_context *psp, int cmd) +{ + uint32_t reg_status = 0, reg_val = 0; + struct amdgpu_device *adev = psp->adev; + int ret; + + /* clear MBX ready (MBOX_READY_MASK bit is 0) and set update command */ + reg_val |= (cmd << 16); + WREG32_SOC15(MP0, 0, regMP0_SMN_C2PMSG_115, reg_val); + + /* Ring the doorbell */ + WREG32_SOC15(MP0, 0, regMP0_SMN_C2PMSG_73, 1); + + if (cmd == C2PMSG_CMD_SPI_UPDATE_FLASH_IMAGE) + return 0; + + ret = psp_wait_for(psp, SOC15_REG_OFFSET(MP0, 0, regMP0_SMN_C2PMSG_115), + MBOX_READY_FLAG, MBOX_READY_MASK, false); + if (ret) { + dev_err(adev->dev, "SPI cmd %x timed out, ret = %d", cmd, ret); + return ret; + } + + reg_status = RREG32_SOC15(MP0, 0, regMP0_SMN_C2PMSG_115); + if ((reg_status & 0xFFFF) != 0) { + dev_err(adev->dev, "SPI cmd %x failed, fail status = %04x\n", + cmd, reg_status & 0xFFFF); + return -EIO; + } + + return 0; +} + +static int psp_v13_0_update_spirom(struct psp_context *psp, + uint64_t fw_pri_mc_addr) +{ + struct amdgpu_device *adev = psp->adev; + int ret; + + /* Confirm PSP is ready to start */ + ret = psp_wait_for(psp, SOC15_REG_OFFSET(MP0, 0, regMP0_SMN_C2PMSG_115), + MBOX_READY_FLAG, MBOX_READY_MASK, false); + if (ret) { + dev_err(adev->dev, "PSP Not ready to start processing, ret = %d", ret); + return ret; + } + + WREG32_SOC15(MP0, 0, regMP0_SMN_C2PMSG_116, lower_32_bits(fw_pri_mc_addr)); + + ret = psp_v13_0_exec_spi_cmd(psp, C2PMSG_CMD_SPI_UPDATE_ROM_IMAGE_ADDR_LO); + if (ret) + return ret; + + WREG32_SOC15(MP0, 0, regMP0_SMN_C2PMSG_116, upper_32_bits(fw_pri_mc_addr)); + + ret = psp_v13_0_exec_spi_cmd(psp, C2PMSG_CMD_SPI_UPDATE_ROM_IMAGE_ADDR_HI); + if (ret) + return ret; + + psp->vbflash_done = true; + + ret = psp_v13_0_exec_spi_cmd(psp, C2PMSG_CMD_SPI_UPDATE_FLASH_IMAGE); + if (ret) + return ret; + + return 0; +} + +static int psp_v13_0_vbflash_status(struct psp_context *psp) +{ + struct amdgpu_device *adev = psp->adev; + + return RREG32_SOC15(MP0, 0, regMP0_SMN_C2PMSG_115); +} + static const struct psp_funcs psp_v13_0_funcs = { .init_microcode = psp_v13_0_init_microcode, .bootloader_load_kdb = psp_v13_0_bootloader_load_kdb, + .bootloader_load_spl = psp_v13_0_bootloader_load_spl, .bootloader_load_sysdrv = psp_v13_0_bootloader_load_sysdrv, .bootloader_load_soc_drv = psp_v13_0_bootloader_load_soc_drv, .bootloader_load_intf_drv = psp_v13_0_bootloader_load_intf_drv, .bootloader_load_dbg_drv = psp_v13_0_bootloader_load_dbg_drv, + .bootloader_load_ras_drv = psp_v13_0_bootloader_load_ras_drv, .bootloader_load_sos = psp_v13_0_bootloader_load_sos, .ring_init = psp_v13_0_ring_init, .ring_create = psp_v13_0_ring_create, @@ -453,8 +734,11 @@ static const struct psp_funcs psp_v13_0_funcs = { .ring_destroy = psp_v13_0_ring_destroy, .ring_get_wptr = psp_v13_0_ring_get_wptr, .ring_set_wptr = psp_v13_0_ring_set_wptr, + .mem_training = psp_v13_0_memory_training, .load_usbc_pd_fw = psp_v13_0_load_usbc_pd_fw, - .read_usbc_pd_fw = psp_v13_0_read_usbc_pd_fw + .read_usbc_pd_fw = psp_v13_0_read_usbc_pd_fw, + .update_spirom = psp_v13_0_update_spirom, + .vbflash_stat = psp_v13_0_vbflash_status }; void psp_v13_0_set_psp_funcs(struct psp_context *psp) diff --git a/drivers/gpu/drm/amd/amdgpu/psp_v13_0_4.c b/drivers/gpu/drm/amd/amdgpu/psp_v13_0_4.c new file mode 100644 index 000000000000..321089dfa7db --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/psp_v13_0_4.c @@ -0,0 +1,387 @@ +/* + * Copyright 2020 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ +#include "amdgpu.h" +#include "amdgpu_psp.h" +#include "amdgpu_ucode.h" +#include "soc15_common.h" +#include "psp_v13_0_4.h" + +#include "mp/mp_13_0_4_offset.h" +#include "mp/mp_13_0_4_sh_mask.h" + +MODULE_FIRMWARE("amdgpu/psp_13_0_4_toc.bin"); +MODULE_FIRMWARE("amdgpu/psp_13_0_4_ta.bin"); + +static int psp_v13_0_4_init_microcode(struct psp_context *psp) +{ + struct amdgpu_device *adev = psp->adev; + const char *chip_name; + char ucode_prefix[30]; + int err = 0; + + switch (adev->ip_versions[MP0_HWIP][0]) { + case IP_VERSION(13, 0, 4): + amdgpu_ucode_ip_version_decode(adev, MP0_HWIP, ucode_prefix, sizeof(ucode_prefix)); + chip_name = ucode_prefix; + break; + default: + BUG(); + } + + switch (adev->ip_versions[MP0_HWIP][0]) { + case IP_VERSION(13, 0, 4): + err = psp_init_toc_microcode(psp, chip_name); + if (err) + return err; + err = psp_init_ta_microcode(psp, chip_name); + if (err) + return err; + break; + default: + BUG(); + } + + return 0; +} + +static bool psp_v13_0_4_is_sos_alive(struct psp_context *psp) +{ + struct amdgpu_device *adev = psp->adev; + uint32_t sol_reg; + + sol_reg = RREG32_SOC15(MP0, 0, regMP0_SMN_C2PMSG_81); + + return sol_reg != 0x0; +} + +static int psp_v13_0_4_wait_for_bootloader(struct psp_context *psp) +{ + struct amdgpu_device *adev = psp->adev; + + int ret; + int retry_loop; + + for (retry_loop = 0; retry_loop < 10; retry_loop++) { + /* Wait for bootloader to signify that is + ready having bit 31 of C2PMSG_35 set to 1 */ + ret = psp_wait_for(psp, + SOC15_REG_OFFSET(MP0, 0, regMP0_SMN_C2PMSG_35), + 0x80000000, + 0x80000000, + false); + + if (ret == 0) + return 0; + } + + return ret; +} + +static int psp_v13_0_4_bootloader_load_component(struct psp_context *psp, + struct psp_bin_desc *bin_desc, + enum psp_bootloader_cmd bl_cmd) +{ + int ret; + uint32_t psp_gfxdrv_command_reg = 0; + struct amdgpu_device *adev = psp->adev; + + /* Check tOS sign of life register to confirm sys driver and sOS + * are already been loaded. + */ + if (psp_v13_0_4_is_sos_alive(psp)) + return 0; + + ret = psp_v13_0_4_wait_for_bootloader(psp); + if (ret) + return ret; + + memset(psp->fw_pri_buf, 0, PSP_1_MEG); + + /* Copy PSP KDB binary to memory */ + memcpy(psp->fw_pri_buf, bin_desc->start_addr, bin_desc->size_bytes); + + /* Provide the PSP KDB to bootloader */ + WREG32_SOC15(MP0, 0, regMP0_SMN_C2PMSG_36, + (uint32_t)(psp->fw_pri_mc_addr >> 20)); + psp_gfxdrv_command_reg = bl_cmd; + WREG32_SOC15(MP0, 0, regMP0_SMN_C2PMSG_35, + psp_gfxdrv_command_reg); + + ret = psp_v13_0_4_wait_for_bootloader(psp); + + return ret; +} + +static int psp_v13_0_4_bootloader_load_kdb(struct psp_context *psp) +{ + return psp_v13_0_4_bootloader_load_component(psp, &psp->kdb, PSP_BL__LOAD_KEY_DATABASE); +} + +static int psp_v13_0_4_bootloader_load_spl(struct psp_context *psp) +{ + return psp_v13_0_4_bootloader_load_component(psp, &psp->kdb, PSP_BL__LOAD_TOS_SPL_TABLE); +} + +static int psp_v13_0_4_bootloader_load_sysdrv(struct psp_context *psp) +{ + return psp_v13_0_4_bootloader_load_component(psp, &psp->sys, PSP_BL__LOAD_SYSDRV); +} + +static int psp_v13_0_4_bootloader_load_soc_drv(struct psp_context *psp) +{ + return psp_v13_0_4_bootloader_load_component(psp, &psp->soc_drv, PSP_BL__LOAD_SOCDRV); +} + +static int psp_v13_0_4_bootloader_load_intf_drv(struct psp_context *psp) +{ + return psp_v13_0_4_bootloader_load_component(psp, &psp->intf_drv, PSP_BL__LOAD_INTFDRV); +} + +static int psp_v13_0_4_bootloader_load_dbg_drv(struct psp_context *psp) +{ + return psp_v13_0_4_bootloader_load_component(psp, &psp->dbg_drv, PSP_BL__LOAD_DBGDRV); +} + +static int psp_v13_0_4_bootloader_load_sos(struct psp_context *psp) +{ + int ret; + unsigned int psp_gfxdrv_command_reg = 0; + struct amdgpu_device *adev = psp->adev; + + /* Check sOS sign of life register to confirm sys driver and sOS + * are already been loaded. + */ + if (psp_v13_0_4_is_sos_alive(psp)) + return 0; + + ret = psp_v13_0_4_wait_for_bootloader(psp); + if (ret) + return ret; + + memset(psp->fw_pri_buf, 0, PSP_1_MEG); + + /* Copy Secure OS binary to PSP memory */ + memcpy(psp->fw_pri_buf, psp->sos.start_addr, psp->sos.size_bytes); + + /* Provide the PSP secure OS to bootloader */ + WREG32_SOC15(MP0, 0, regMP0_SMN_C2PMSG_36, + (uint32_t)(psp->fw_pri_mc_addr >> 20)); + psp_gfxdrv_command_reg = PSP_BL__LOAD_SOSDRV; + WREG32_SOC15(MP0, 0, regMP0_SMN_C2PMSG_35, + psp_gfxdrv_command_reg); + + /* there might be handshake issue with hardware which needs delay */ + mdelay(20); + ret = psp_wait_for(psp, SOC15_REG_OFFSET(MP0, 0, regMP0_SMN_C2PMSG_81), + RREG32_SOC15(MP0, 0, regMP0_SMN_C2PMSG_81), + 0, true); + + return ret; +} + +static int psp_v13_0_4_ring_init(struct psp_context *psp, + enum psp_ring_type ring_type) +{ + int ret = 0; + struct psp_ring *ring; + struct amdgpu_device *adev = psp->adev; + + ring = &psp->km_ring; + + ring->ring_type = ring_type; + + /* allocate 4k Page of Local Frame Buffer memory for ring */ + ring->ring_size = 0x1000; + ret = amdgpu_bo_create_kernel(adev, ring->ring_size, PAGE_SIZE, + AMDGPU_GEM_DOMAIN_VRAM, + &adev->firmware.rbuf, + &ring->ring_mem_mc_addr, + (void **)&ring->ring_mem); + if (ret) { + ring->ring_size = 0; + return ret; + } + + return 0; +} + +static int psp_v13_0_4_ring_stop(struct psp_context *psp, + enum psp_ring_type ring_type) +{ + int ret = 0; + struct amdgpu_device *adev = psp->adev; + + if (amdgpu_sriov_vf(adev)) { + /* Write the ring destroy command*/ + WREG32_SOC15(MP0, 0, regMP0_SMN_C2PMSG_101, + GFX_CTRL_CMD_ID_DESTROY_GPCOM_RING); + /* there might be handshake issue with hardware which needs delay */ + mdelay(20); + /* Wait for response flag (bit 31) */ + ret = psp_wait_for(psp, SOC15_REG_OFFSET(MP0, 0, regMP0_SMN_C2PMSG_101), + 0x80000000, 0x80000000, false); + } else { + /* Write the ring destroy command*/ + WREG32_SOC15(MP0, 0, regMP0_SMN_C2PMSG_64, + GFX_CTRL_CMD_ID_DESTROY_RINGS); + /* there might be handshake issue with hardware which needs delay */ + mdelay(20); + /* Wait for response flag (bit 31) */ + ret = psp_wait_for(psp, SOC15_REG_OFFSET(MP0, 0, regMP0_SMN_C2PMSG_64), + 0x80000000, 0x80000000, false); + } + + return ret; +} + +static int psp_v13_0_4_ring_create(struct psp_context *psp, + enum psp_ring_type ring_type) +{ + int ret = 0; + unsigned int psp_ring_reg = 0; + struct psp_ring *ring = &psp->km_ring; + struct amdgpu_device *adev = psp->adev; + + if (amdgpu_sriov_vf(adev)) { + ret = psp_v13_0_4_ring_stop(psp, ring_type); + if (ret) { + DRM_ERROR("psp_v13_0_ring_stop_sriov failed!\n"); + return ret; + } + + /* Write low address of the ring to C2PMSG_102 */ + psp_ring_reg = lower_32_bits(ring->ring_mem_mc_addr); + WREG32_SOC15(MP0, 0, regMP0_SMN_C2PMSG_102, psp_ring_reg); + /* Write high address of the ring to C2PMSG_103 */ + psp_ring_reg = upper_32_bits(ring->ring_mem_mc_addr); + WREG32_SOC15(MP0, 0, regMP0_SMN_C2PMSG_103, psp_ring_reg); + + /* Write the ring initialization command to C2PMSG_101 */ + WREG32_SOC15(MP0, 0, regMP0_SMN_C2PMSG_101, + GFX_CTRL_CMD_ID_INIT_GPCOM_RING); + + /* there might be handshake issue with hardware which needs delay */ + mdelay(20); + + /* Wait for response flag (bit 31) in C2PMSG_101 */ + ret = psp_wait_for(psp, SOC15_REG_OFFSET(MP0, 0, regMP0_SMN_C2PMSG_101), + 0x80000000, 0x8000FFFF, false); + + } else { + /* Wait for sOS ready for ring creation */ + ret = psp_wait_for(psp, SOC15_REG_OFFSET(MP0, 0, regMP0_SMN_C2PMSG_64), + 0x80000000, 0x80000000, false); + if (ret) { + DRM_ERROR("Failed to wait for trust OS ready for ring creation\n"); + return ret; + } + + /* Write low address of the ring to C2PMSG_69 */ + psp_ring_reg = lower_32_bits(ring->ring_mem_mc_addr); + WREG32_SOC15(MP0, 0, regMP0_SMN_C2PMSG_69, psp_ring_reg); + /* Write high address of the ring to C2PMSG_70 */ + psp_ring_reg = upper_32_bits(ring->ring_mem_mc_addr); + WREG32_SOC15(MP0, 0, regMP0_SMN_C2PMSG_70, psp_ring_reg); + /* Write size of ring to C2PMSG_71 */ + psp_ring_reg = ring->ring_size; + WREG32_SOC15(MP0, 0, regMP0_SMN_C2PMSG_71, psp_ring_reg); + /* Write the ring initialization command to C2PMSG_64 */ + psp_ring_reg = ring_type; + psp_ring_reg = psp_ring_reg << 16; + WREG32_SOC15(MP0, 0, regMP0_SMN_C2PMSG_64, psp_ring_reg); + + /* there might be handshake issue with hardware which needs delay */ + mdelay(20); + + /* Wait for response flag (bit 31) in C2PMSG_64 */ + ret = psp_wait_for(psp, SOC15_REG_OFFSET(MP0, 0, regMP0_SMN_C2PMSG_64), + 0x80000000, 0x8000FFFF, false); + } + + return ret; +} + +static int psp_v13_0_4_ring_destroy(struct psp_context *psp, + enum psp_ring_type ring_type) +{ + int ret = 0; + struct psp_ring *ring = &psp->km_ring; + struct amdgpu_device *adev = psp->adev; + + ret = psp_v13_0_4_ring_stop(psp, ring_type); + if (ret) + DRM_ERROR("Fail to stop psp ring\n"); + + amdgpu_bo_free_kernel(&adev->firmware.rbuf, + &ring->ring_mem_mc_addr, + (void **)&ring->ring_mem); + + return ret; +} + +static uint32_t psp_v13_0_4_ring_get_wptr(struct psp_context *psp) +{ + uint32_t data; + struct amdgpu_device *adev = psp->adev; + + if (amdgpu_sriov_vf(adev)) + data = RREG32_SOC15(MP0, 0, regMP0_SMN_C2PMSG_102); + else + data = RREG32_SOC15(MP0, 0, regMP0_SMN_C2PMSG_67); + + return data; +} + +static void psp_v13_0_4_ring_set_wptr(struct psp_context *psp, uint32_t value) +{ + struct amdgpu_device *adev = psp->adev; + + if (amdgpu_sriov_vf(adev)) { + WREG32_SOC15(MP0, 0, regMP0_SMN_C2PMSG_102, value); + WREG32_SOC15(MP0, 0, regMP0_SMN_C2PMSG_101, + GFX_CTRL_CMD_ID_CONSUME_CMD); + } else + WREG32_SOC15(MP0, 0, regMP0_SMN_C2PMSG_67, value); +} + +static const struct psp_funcs psp_v13_0_4_funcs = { + .init_microcode = psp_v13_0_4_init_microcode, + .bootloader_load_kdb = psp_v13_0_4_bootloader_load_kdb, + .bootloader_load_spl = psp_v13_0_4_bootloader_load_spl, + .bootloader_load_sysdrv = psp_v13_0_4_bootloader_load_sysdrv, + .bootloader_load_soc_drv = psp_v13_0_4_bootloader_load_soc_drv, + .bootloader_load_intf_drv = psp_v13_0_4_bootloader_load_intf_drv, + .bootloader_load_dbg_drv = psp_v13_0_4_bootloader_load_dbg_drv, + .bootloader_load_sos = psp_v13_0_4_bootloader_load_sos, + .ring_init = psp_v13_0_4_ring_init, + .ring_create = psp_v13_0_4_ring_create, + .ring_stop = psp_v13_0_4_ring_stop, + .ring_destroy = psp_v13_0_4_ring_destroy, + .ring_get_wptr = psp_v13_0_4_ring_get_wptr, + .ring_set_wptr = psp_v13_0_4_ring_set_wptr, +}; + +void psp_v13_0_4_set_psp_funcs(struct psp_context *psp) +{ + psp->funcs = &psp_v13_0_4_funcs; +} diff --git a/drivers/gpu/drm/amd/amdgpu/psp_v13_0_4.h b/drivers/gpu/drm/amd/amdgpu/psp_v13_0_4.h new file mode 100644 index 000000000000..8547b8d514d5 --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/psp_v13_0_4.h @@ -0,0 +1,30 @@ +/* + * Copyright 2020 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ +#ifndef __PSP_V13_0_4_H__ +#define __PSP_V13_0_4_H__ + +#include "amdgpu_psp.h" + +void psp_v13_0_4_set_psp_funcs(struct psp_context *psp); + +#endif diff --git a/drivers/gpu/drm/amd/amdgpu/psp_v3_1.c b/drivers/gpu/drm/amd/amdgpu/psp_v3_1.c index 1ed357cb0f49..01f3bcc62a6c 100644 --- a/drivers/gpu/drm/amd/amdgpu/psp_v3_1.c +++ b/drivers/gpu/drm/amd/amdgpu/psp_v3_1.c @@ -44,6 +44,7 @@ MODULE_FIRMWARE("amdgpu/vega10_sos.bin"); MODULE_FIRMWARE("amdgpu/vega10_asd.bin"); +MODULE_FIRMWARE("amdgpu/vega10_cap.bin"); MODULE_FIRMWARE("amdgpu/vega12_sos.bin"); MODULE_FIRMWARE("amdgpu/vega12_asd.bin"); diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c b/drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c index 4509bd4cce2d..c52d246a1d96 100644 --- a/drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c +++ b/drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c @@ -194,7 +194,7 @@ out: static uint64_t sdma_v2_4_ring_get_rptr(struct amdgpu_ring *ring) { /* XXX check if swapping is necessary on BE */ - return ring->adev->wb.wb[ring->rptr_offs] >> 2; + return *ring->rptr_cpu_addr >> 2; } /** @@ -223,7 +223,7 @@ static void sdma_v2_4_ring_set_wptr(struct amdgpu_ring *ring) { struct amdgpu_device *adev = ring->adev; - WREG32(mmSDMA0_GFX_RB_WPTR + sdma_offsets[ring->me], lower_32_bits(ring->wptr) << 2); + WREG32(mmSDMA0_GFX_RB_WPTR + sdma_offsets[ring->me], ring->wptr << 2); } static void sdma_v2_4_ring_insert_nop(struct amdgpu_ring *ring, uint32_t count) @@ -342,14 +342,10 @@ static void sdma_v2_4_ring_emit_fence(struct amdgpu_ring *ring, u64 addr, u64 se */ static void sdma_v2_4_gfx_stop(struct amdgpu_device *adev) { - struct amdgpu_ring *sdma0 = &adev->sdma.instance[0].ring; - struct amdgpu_ring *sdma1 = &adev->sdma.instance[1].ring; u32 rb_cntl, ib_cntl; int i; - if ((adev->mman.buffer_funcs_ring == sdma0) || - (adev->mman.buffer_funcs_ring == sdma1)) - amdgpu_ttm_set_buffer_funcs_status(adev, false); + amdgpu_sdma_unset_buffer_funcs_helper(adev); for (i = 0; i < adev->sdma.num_instances; i++) { rb_cntl = RREG32(mmSDMA0_GFX_RB_CNTL + sdma_offsets[i]); @@ -414,12 +410,10 @@ static int sdma_v2_4_gfx_resume(struct amdgpu_device *adev) struct amdgpu_ring *ring; u32 rb_cntl, ib_cntl; u32 rb_bufsz; - u32 wb_offset; int i, j, r; for (i = 0; i < adev->sdma.num_instances; i++) { ring = &adev->sdma.instance[i].ring; - wb_offset = (ring->rptr_offs * 4); mutex_lock(&adev->srbm_mutex); for (j = 0; j < 16; j++) { @@ -455,9 +449,9 @@ static int sdma_v2_4_gfx_resume(struct amdgpu_device *adev) /* set the wb address whether it's enabled or not */ WREG32(mmSDMA0_GFX_RB_RPTR_ADDR_HI + sdma_offsets[i], - upper_32_bits(adev->wb.gpu_addr + wb_offset) & 0xFFFFFFFF); + upper_32_bits(ring->rptr_gpu_addr) & 0xFFFFFFFF); WREG32(mmSDMA0_GFX_RB_RPTR_ADDR_LO + sdma_offsets[i], - lower_32_bits(adev->wb.gpu_addr + wb_offset) & 0xFFFFFFFC); + lower_32_bits(ring->rptr_gpu_addr) & 0xFFFFFFFC); rb_cntl = REG_SET_FIELD(rb_cntl, SDMA0_GFX_RB_CNTL, RPTR_WRITEBACK_ENABLE, 1); @@ -465,7 +459,7 @@ static int sdma_v2_4_gfx_resume(struct amdgpu_device *adev) WREG32(mmSDMA0_GFX_RB_BASE_HI + sdma_offsets[i], ring->gpu_addr >> 40); ring->wptr = 0; - WREG32(mmSDMA0_GFX_RB_WPTR + sdma_offsets[i], lower_32_bits(ring->wptr) << 2); + WREG32(mmSDMA0_GFX_RB_WPTR + sdma_offsets[i], ring->wptr << 2); /* enable DMA RB */ rb_cntl = REG_SET_FIELD(rb_cntl, SDMA0_GFX_RB_CNTL, RB_ENABLE, 1); @@ -1142,6 +1136,7 @@ static const struct amdgpu_ring_funcs sdma_v2_4_ring_funcs = { .align_mask = 0xf, .nop = SDMA_PKT_NOP_HEADER_OP(SDMA_OP_NOP), .support_64bit_ptrs = false, + .secure_submission_supported = true, .get_rptr = sdma_v2_4_ring_get_rptr, .get_wptr = sdma_v2_4_ring_get_wptr, .set_wptr = sdma_v2_4_ring_set_wptr, diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c b/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c index 135727b59c41..486d9b5c1b9e 100644 --- a/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c +++ b/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c @@ -350,7 +350,7 @@ out: static uint64_t sdma_v3_0_ring_get_rptr(struct amdgpu_ring *ring) { /* XXX check if swapping is necessary on BE */ - return ring->adev->wb.wb[ring->rptr_offs] >> 2; + return *ring->rptr_cpu_addr >> 2; } /** @@ -367,7 +367,7 @@ static uint64_t sdma_v3_0_ring_get_wptr(struct amdgpu_ring *ring) if (ring->use_doorbell || ring->use_pollmem) { /* XXX check if swapping is necessary on BE */ - wptr = ring->adev->wb.wb[ring->wptr_offs] >> 2; + wptr = *ring->wptr_cpu_addr >> 2; } else { wptr = RREG32(mmSDMA0_GFX_RB_WPTR + sdma_offsets[ring->me]) >> 2; } @@ -387,16 +387,16 @@ static void sdma_v3_0_ring_set_wptr(struct amdgpu_ring *ring) struct amdgpu_device *adev = ring->adev; if (ring->use_doorbell) { - u32 *wb = (u32 *)&adev->wb.wb[ring->wptr_offs]; + u32 *wb = (u32 *)ring->wptr_cpu_addr; /* XXX check if swapping is necessary on BE */ - WRITE_ONCE(*wb, (lower_32_bits(ring->wptr) << 2)); - WDOORBELL32(ring->doorbell_index, lower_32_bits(ring->wptr) << 2); + WRITE_ONCE(*wb, ring->wptr << 2); + WDOORBELL32(ring->doorbell_index, ring->wptr << 2); } else if (ring->use_pollmem) { - u32 *wb = (u32 *)&adev->wb.wb[ring->wptr_offs]; + u32 *wb = (u32 *)ring->wptr_cpu_addr; - WRITE_ONCE(*wb, (lower_32_bits(ring->wptr) << 2)); + WRITE_ONCE(*wb, ring->wptr << 2); } else { - WREG32(mmSDMA0_GFX_RB_WPTR + sdma_offsets[ring->me], lower_32_bits(ring->wptr) << 2); + WREG32(mmSDMA0_GFX_RB_WPTR + sdma_offsets[ring->me], ring->wptr << 2); } } @@ -516,14 +516,10 @@ static void sdma_v3_0_ring_emit_fence(struct amdgpu_ring *ring, u64 addr, u64 se */ static void sdma_v3_0_gfx_stop(struct amdgpu_device *adev) { - struct amdgpu_ring *sdma0 = &adev->sdma.instance[0].ring; - struct amdgpu_ring *sdma1 = &adev->sdma.instance[1].ring; u32 rb_cntl, ib_cntl; int i; - if ((adev->mman.buffer_funcs_ring == sdma0) || - (adev->mman.buffer_funcs_ring == sdma1)) - amdgpu_ttm_set_buffer_funcs_status(adev, false); + amdgpu_sdma_unset_buffer_funcs_helper(adev); for (i = 0; i < adev->sdma.num_instances; i++) { rb_cntl = RREG32(mmSDMA0_GFX_RB_CNTL + sdma_offsets[i]); @@ -649,7 +645,6 @@ static int sdma_v3_0_gfx_resume(struct amdgpu_device *adev) struct amdgpu_ring *ring; u32 rb_cntl, ib_cntl, wptr_poll_cntl; u32 rb_bufsz; - u32 wb_offset; u32 doorbell; u64 wptr_gpu_addr; int i, j, r; @@ -657,7 +652,6 @@ static int sdma_v3_0_gfx_resume(struct amdgpu_device *adev) for (i = 0; i < adev->sdma.num_instances; i++) { ring = &adev->sdma.instance[i].ring; amdgpu_ring_clear_ring(ring); - wb_offset = (ring->rptr_offs * 4); mutex_lock(&adev->srbm_mutex); for (j = 0; j < 16; j++) { @@ -694,9 +688,9 @@ static int sdma_v3_0_gfx_resume(struct amdgpu_device *adev) /* set the wb address whether it's enabled or not */ WREG32(mmSDMA0_GFX_RB_RPTR_ADDR_HI + sdma_offsets[i], - upper_32_bits(adev->wb.gpu_addr + wb_offset) & 0xFFFFFFFF); + upper_32_bits(ring->rptr_gpu_addr) & 0xFFFFFFFF); WREG32(mmSDMA0_GFX_RB_RPTR_ADDR_LO + sdma_offsets[i], - lower_32_bits(adev->wb.gpu_addr + wb_offset) & 0xFFFFFFFC); + lower_32_bits(ring->rptr_gpu_addr) & 0xFFFFFFFC); rb_cntl = REG_SET_FIELD(rb_cntl, SDMA0_GFX_RB_CNTL, RPTR_WRITEBACK_ENABLE, 1); @@ -715,7 +709,7 @@ static int sdma_v3_0_gfx_resume(struct amdgpu_device *adev) WREG32(mmSDMA0_GFX_DOORBELL + sdma_offsets[i], doorbell); /* setup the wptr shadow polling */ - wptr_gpu_addr = adev->wb.gpu_addr + (ring->wptr_offs * 4); + wptr_gpu_addr = ring->wptr_gpu_addr; WREG32(mmSDMA0_GFX_RB_WPTR_POLL_ADDR_LO + sdma_offsets[i], lower_32_bits(wptr_gpu_addr)); @@ -1535,7 +1529,7 @@ static int sdma_v3_0_set_powergating_state(void *handle, return 0; } -static void sdma_v3_0_get_clockgating_state(void *handle, u32 *flags) +static void sdma_v3_0_get_clockgating_state(void *handle, u64 *flags) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; int data; @@ -1580,6 +1574,7 @@ static const struct amdgpu_ring_funcs sdma_v3_0_ring_funcs = { .align_mask = 0xf, .nop = SDMA_PKT_NOP_HEADER_OP(SDMA_OP_NOP), .support_64bit_ptrs = false, + .secure_submission_supported = true, .get_rptr = sdma_v3_0_ring_get_rptr, .get_wptr = sdma_v3_0_ring_get_wptr, .set_wptr = sdma_v3_0_ring_set_wptr, diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c b/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c index e8e4749e9c79..1122bd4eae98 100644 --- a/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c +++ b/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c @@ -561,44 +561,6 @@ static void sdma_v4_0_setup_ulv(struct amdgpu_device *adev) } } -static int sdma_v4_0_init_inst_ctx(struct amdgpu_sdma_instance *sdma_inst) -{ - int err = 0; - const struct sdma_firmware_header_v1_0 *hdr; - - err = amdgpu_ucode_validate(sdma_inst->fw); - if (err) - return err; - - hdr = (const struct sdma_firmware_header_v1_0 *)sdma_inst->fw->data; - sdma_inst->fw_version = le32_to_cpu(hdr->header.ucode_version); - sdma_inst->feature_version = le32_to_cpu(hdr->ucode_feature_version); - - if (sdma_inst->feature_version >= 20) - sdma_inst->burst_nop = true; - - return 0; -} - -static void sdma_v4_0_destroy_inst_ctx(struct amdgpu_device *adev) -{ - int i; - - for (i = 0; i < adev->sdma.num_instances; i++) { - release_firmware(adev->sdma.instance[i].fw); - adev->sdma.instance[i].fw = NULL; - - /* arcturus shares the same FW memory across - all SDMA isntances */ - if (adev->ip_versions[SDMA0_HWIP][0] == IP_VERSION(4, 2, 2) || - adev->ip_versions[SDMA0_HWIP][0] == IP_VERSION(4, 4, 0)) - break; - } - - memset((void *)adev->sdma.instance, 0, - sizeof(struct amdgpu_sdma_instance) * AMDGPU_MAX_SDMA_INSTANCES); -} - /** * sdma_v4_0_init_microcode - load ucode images from disk * @@ -615,9 +577,7 @@ static int sdma_v4_0_init_microcode(struct amdgpu_device *adev) { const char *chip_name; char fw_name[30]; - int err = 0, i; - struct amdgpu_firmware_info *info = NULL; - const struct common_firmware_header *header = NULL; + int ret, i; DRM_DEBUG("\n"); @@ -656,58 +616,25 @@ static int sdma_v4_0_init_microcode(struct amdgpu_device *adev) BUG(); } - snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_sdma.bin", chip_name); - - err = request_firmware(&adev->sdma.instance[0].fw, fw_name, adev->dev); - if (err) - goto out; - - err = sdma_v4_0_init_inst_ctx(&adev->sdma.instance[0]); - if (err) - goto out; - - for (i = 1; i < adev->sdma.num_instances; i++) { + for (i = 0; i < adev->sdma.num_instances; i++) { + if (i == 0) + snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_sdma.bin", chip_name); + else + snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_sdma%d.bin", chip_name, i); if (adev->ip_versions[SDMA0_HWIP][0] == IP_VERSION(4, 2, 2) || adev->ip_versions[SDMA0_HWIP][0] == IP_VERSION(4, 4, 0)) { /* Acturus & Aldebaran will leverage the same FW memory for every SDMA instance */ - memcpy((void *)&adev->sdma.instance[i], - (void *)&adev->sdma.instance[0], - sizeof(struct amdgpu_sdma_instance)); - } - else { - snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_sdma%d.bin", chip_name, i); - - err = request_firmware(&adev->sdma.instance[i].fw, fw_name, adev->dev); - if (err) - goto out; - - err = sdma_v4_0_init_inst_ctx(&adev->sdma.instance[i]); - if (err) - goto out; - } - } - - DRM_DEBUG("psp_load == '%s'\n", - adev->firmware.load_type == AMDGPU_FW_LOAD_PSP ? "true" : "false"); - - if (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) { - for (i = 0; i < adev->sdma.num_instances; i++) { - info = &adev->firmware.ucode[AMDGPU_UCODE_ID_SDMA0 + i]; - info->ucode_id = AMDGPU_UCODE_ID_SDMA0 + i; - info->fw = adev->sdma.instance[i].fw; - header = (const struct common_firmware_header *)info->fw->data; - adev->firmware.fw_size += - ALIGN(le32_to_cpu(header->ucode_size_bytes), PAGE_SIZE); + ret = amdgpu_sdma_init_microcode(adev, fw_name, 0, true); + break; + } else { + ret = amdgpu_sdma_init_microcode(adev, fw_name, i, false); + if (ret) + return ret; } } -out: - if (err) { - DRM_ERROR("sdma_v4_0: Failed to load firmware \"%s\"\n", fw_name); - sdma_v4_0_destroy_inst_ctx(adev); - } - return err; + return ret; } /** @@ -722,7 +649,7 @@ static uint64_t sdma_v4_0_ring_get_rptr(struct amdgpu_ring *ring) u64 *rptr; /* XXX check if swapping is necessary on BE */ - rptr = ((u64 *)&ring->adev->wb.wb[ring->rptr_offs]); + rptr = ((u64 *)ring->rptr_cpu_addr); DRM_DEBUG("rptr before shift == 0x%016llx\n", *rptr); return ((*rptr) >> 2); @@ -742,7 +669,7 @@ static uint64_t sdma_v4_0_ring_get_wptr(struct amdgpu_ring *ring) if (ring->use_doorbell) { /* XXX check if swapping is necessary on BE */ - wptr = READ_ONCE(*((u64 *)&adev->wb.wb[ring->wptr_offs])); + wptr = READ_ONCE(*((u64 *)ring->wptr_cpu_addr)); DRM_DEBUG("wptr/doorbell before shift == 0x%016llx\n", wptr); } else { wptr = RREG32_SDMA(ring->me, mmSDMA0_GFX_RB_WPTR_HI); @@ -768,12 +695,12 @@ static void sdma_v4_0_ring_set_wptr(struct amdgpu_ring *ring) DRM_DEBUG("Setting write pointer\n"); if (ring->use_doorbell) { - u64 *wb = (u64 *)&adev->wb.wb[ring->wptr_offs]; + u64 *wb = (u64 *)ring->wptr_cpu_addr; DRM_DEBUG("Using doorbell -- " "wptr_offs == 0x%08x " - "lower_32_bits(ring->wptr) << 2 == 0x%08x " - "upper_32_bits(ring->wptr) << 2 == 0x%08x\n", + "lower_32_bits(ring->wptr << 2) == 0x%08x " + "upper_32_bits(ring->wptr << 2) == 0x%08x\n", ring->wptr_offs, lower_32_bits(ring->wptr << 2), upper_32_bits(ring->wptr << 2)); @@ -811,7 +738,7 @@ static uint64_t sdma_v4_0_page_ring_get_wptr(struct amdgpu_ring *ring) if (ring->use_doorbell) { /* XXX check if swapping is necessary on BE */ - wptr = READ_ONCE(*((u64 *)&adev->wb.wb[ring->wptr_offs])); + wptr = READ_ONCE(*((u64 *)ring->wptr_cpu_addr)); } else { wptr = RREG32_SDMA(ring->me, mmSDMA0_PAGE_RB_WPTR_HI); wptr = wptr << 32; @@ -833,7 +760,7 @@ static void sdma_v4_0_page_ring_set_wptr(struct amdgpu_ring *ring) struct amdgpu_device *adev = ring->adev; if (ring->use_doorbell) { - u64 *wb = (u64 *)&adev->wb.wb[ring->wptr_offs]; + u64 *wb = (u64 *)ring->wptr_cpu_addr; /* XXX check if swapping is necessary on BE */ WRITE_ONCE(*wb, (ring->wptr << 2)); @@ -988,18 +915,12 @@ static void sdma_v4_0_ring_emit_fence(struct amdgpu_ring *ring, u64 addr, u64 se */ static void sdma_v4_0_gfx_stop(struct amdgpu_device *adev) { - struct amdgpu_ring *sdma[AMDGPU_MAX_SDMA_INSTANCES]; u32 rb_cntl, ib_cntl; - int i, unset = 0; - - for (i = 0; i < adev->sdma.num_instances; i++) { - sdma[i] = &adev->sdma.instance[i].ring; + int i; - if ((adev->mman.buffer_funcs_ring == sdma[i]) && unset != 1) { - amdgpu_ttm_set_buffer_funcs_status(adev, false); - unset = 1; - } + amdgpu_sdma_unset_buffer_funcs_helper(adev); + for (i = 0; i < adev->sdma.num_instances; i++) { rb_cntl = RREG32_SDMA(i, mmSDMA0_GFX_RB_CNTL); rb_cntl = REG_SET_FIELD(rb_cntl, SDMA0_GFX_RB_CNTL, RB_ENABLE, 0); WREG32_SDMA(i, mmSDMA0_GFX_RB_CNTL, rb_cntl); @@ -1030,20 +951,12 @@ static void sdma_v4_0_rlc_stop(struct amdgpu_device *adev) */ static void sdma_v4_0_page_stop(struct amdgpu_device *adev) { - struct amdgpu_ring *sdma[AMDGPU_MAX_SDMA_INSTANCES]; u32 rb_cntl, ib_cntl; int i; - bool unset = false; - - for (i = 0; i < adev->sdma.num_instances; i++) { - sdma[i] = &adev->sdma.instance[i].page; - if ((adev->mman.buffer_funcs_ring == sdma[i]) && - (!unset)) { - amdgpu_ttm_set_buffer_funcs_status(adev, false); - unset = true; - } + amdgpu_sdma_unset_buffer_funcs_helper(adev); + for (i = 0; i < adev->sdma.num_instances; i++) { rb_cntl = RREG32_SDMA(i, mmSDMA0_PAGE_RB_CNTL); rb_cntl = REG_SET_FIELD(rb_cntl, SDMA0_PAGE_RB_CNTL, RB_ENABLE, 0); @@ -1174,13 +1087,10 @@ static void sdma_v4_0_gfx_resume(struct amdgpu_device *adev, unsigned int i) { struct amdgpu_ring *ring = &adev->sdma.instance[i].ring; u32 rb_cntl, ib_cntl, wptr_poll_cntl; - u32 wb_offset; u32 doorbell; u32 doorbell_offset; u64 wptr_gpu_addr; - wb_offset = (ring->rptr_offs * 4); - rb_cntl = RREG32_SDMA(i, mmSDMA0_GFX_RB_CNTL); rb_cntl = sdma_v4_0_rb_cntl(ring, rb_cntl); WREG32_SDMA(i, mmSDMA0_GFX_RB_CNTL, rb_cntl); @@ -1193,9 +1103,9 @@ static void sdma_v4_0_gfx_resume(struct amdgpu_device *adev, unsigned int i) /* set the wb address whether it's enabled or not */ WREG32_SDMA(i, mmSDMA0_GFX_RB_RPTR_ADDR_HI, - upper_32_bits(adev->wb.gpu_addr + wb_offset) & 0xFFFFFFFF); + upper_32_bits(ring->rptr_gpu_addr) & 0xFFFFFFFF); WREG32_SDMA(i, mmSDMA0_GFX_RB_RPTR_ADDR_LO, - lower_32_bits(adev->wb.gpu_addr + wb_offset) & 0xFFFFFFFC); + lower_32_bits(ring->rptr_gpu_addr) & 0xFFFFFFFC); rb_cntl = REG_SET_FIELD(rb_cntl, SDMA0_GFX_RB_CNTL, RPTR_WRITEBACK_ENABLE, 1); @@ -1225,7 +1135,7 @@ static void sdma_v4_0_gfx_resume(struct amdgpu_device *adev, unsigned int i) WREG32_SDMA(i, mmSDMA0_GFX_MINOR_PTR_UPDATE, 0); /* setup the wptr shadow polling */ - wptr_gpu_addr = adev->wb.gpu_addr + (ring->wptr_offs * 4); + wptr_gpu_addr = ring->wptr_gpu_addr; WREG32_SDMA(i, mmSDMA0_GFX_RB_WPTR_POLL_ADDR_LO, lower_32_bits(wptr_gpu_addr)); WREG32_SDMA(i, mmSDMA0_GFX_RB_WPTR_POLL_ADDR_HI, @@ -1264,13 +1174,10 @@ static void sdma_v4_0_page_resume(struct amdgpu_device *adev, unsigned int i) { struct amdgpu_ring *ring = &adev->sdma.instance[i].page; u32 rb_cntl, ib_cntl, wptr_poll_cntl; - u32 wb_offset; u32 doorbell; u32 doorbell_offset; u64 wptr_gpu_addr; - wb_offset = (ring->rptr_offs * 4); - rb_cntl = RREG32_SDMA(i, mmSDMA0_PAGE_RB_CNTL); rb_cntl = sdma_v4_0_rb_cntl(ring, rb_cntl); WREG32_SDMA(i, mmSDMA0_PAGE_RB_CNTL, rb_cntl); @@ -1283,9 +1190,9 @@ static void sdma_v4_0_page_resume(struct amdgpu_device *adev, unsigned int i) /* set the wb address whether it's enabled or not */ WREG32_SDMA(i, mmSDMA0_PAGE_RB_RPTR_ADDR_HI, - upper_32_bits(adev->wb.gpu_addr + wb_offset) & 0xFFFFFFFF); + upper_32_bits(ring->rptr_gpu_addr) & 0xFFFFFFFF); WREG32_SDMA(i, mmSDMA0_PAGE_RB_RPTR_ADDR_LO, - lower_32_bits(adev->wb.gpu_addr + wb_offset) & 0xFFFFFFFC); + lower_32_bits(ring->rptr_gpu_addr) & 0xFFFFFFFC); rb_cntl = REG_SET_FIELD(rb_cntl, SDMA0_PAGE_RB_CNTL, RPTR_WRITEBACK_ENABLE, 1); @@ -1316,7 +1223,7 @@ static void sdma_v4_0_page_resume(struct amdgpu_device *adev, unsigned int i) WREG32_SDMA(i, mmSDMA0_PAGE_MINOR_PTR_UPDATE, 0); /* setup the wptr shadow polling */ - wptr_gpu_addr = adev->wb.gpu_addr + (ring->wptr_offs * 4); + wptr_gpu_addr = ring->wptr_gpu_addr; WREG32_SDMA(i, mmSDMA0_PAGE_RB_WPTR_POLL_ADDR_LO, lower_32_bits(wptr_gpu_addr)); WREG32_SDMA(i, mmSDMA0_PAGE_RB_WPTR_POLL_ADDR_HI, @@ -1885,22 +1792,16 @@ static int sdma_v4_0_process_ras_data_cb(struct amdgpu_device *adev, static int sdma_v4_0_late_init(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; - struct ras_ih_if ih_info = { - .cb = sdma_v4_0_process_ras_data_cb, - }; sdma_v4_0_setup_ulv(adev); if (!amdgpu_persistent_edc_harvesting_supported(adev)) { - if (adev->sdma.funcs && - adev->sdma.funcs->reset_ras_error_count) - adev->sdma.funcs->reset_ras_error_count(adev); + if (adev->sdma.ras && adev->sdma.ras->ras_block.hw_ops && + adev->sdma.ras->ras_block.hw_ops->reset_ras_error_count) + adev->sdma.ras->ras_block.hw_ops->reset_ras_error_count(adev); } - if (adev->sdma.funcs && adev->sdma.funcs->ras_late_init) - return adev->sdma.funcs->ras_late_init(adev, &ih_info); - else - return 0; + return 0; } static int sdma_v4_0_sw_init(void *handle) @@ -2001,23 +1902,23 @@ static int sdma_v4_0_sw_fini(void *handle) struct amdgpu_device *adev = (struct amdgpu_device *)handle; int i; - if (adev->sdma.funcs && adev->sdma.funcs->ras_fini) - adev->sdma.funcs->ras_fini(adev); - for (i = 0; i < adev->sdma.num_instances; i++) { amdgpu_ring_fini(&adev->sdma.instance[i].ring); if (adev->sdma.has_page_queue) amdgpu_ring_fini(&adev->sdma.instance[i].page); } - sdma_v4_0_destroy_inst_ctx(adev); + if (adev->ip_versions[SDMA0_HWIP][0] == IP_VERSION(4, 2, 0) || + adev->ip_versions[SDMA0_HWIP][0] == IP_VERSION(4, 4, 0)) + amdgpu_sdma_destroy_inst_ctx(adev, true); + else + amdgpu_sdma_destroy_inst_ctx(adev, false); return 0; } static int sdma_v4_0_hw_init(void *handle) { - int r; struct amdgpu_device *adev = (struct amdgpu_device *)handle; if (adev->flags & AMD_IS_APU) @@ -2026,9 +1927,7 @@ static int sdma_v4_0_hw_init(void *handle) if (!amdgpu_sriov_vf(adev)) sdma_v4_0_init_golden_registers(adev); - r = sdma_v4_0_start(adev); - - return r; + return sdma_v4_0_start(adev); } static int sdma_v4_0_hw_fini(void *handle) @@ -2036,8 +1935,11 @@ static int sdma_v4_0_hw_fini(void *handle) struct amdgpu_device *adev = (struct amdgpu_device *)handle; int i; - if (amdgpu_sriov_vf(adev)) + if (amdgpu_sriov_vf(adev)) { + /* disable the scheduler for SDMA */ + amdgpu_sdma_unset_buffer_funcs_helper(adev); return 0; + } for (i = 0; i < adev->sdma.num_instances; i++) { amdgpu_irq_put(adev, &adev->sdma.ecc_irq, @@ -2057,6 +1959,10 @@ static int sdma_v4_0_suspend(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; + /* SMU saves SDMA state for us */ + if (adev->in_s0ix) + return 0; + return sdma_v4_0_hw_fini(adev); } @@ -2064,6 +1970,10 @@ static int sdma_v4_0_resume(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; + /* SMU restores SDMA state for us */ + if (adev->in_s0ix) + return 0; + return sdma_v4_0_hw_init(adev); } @@ -2373,7 +2283,7 @@ static int sdma_v4_0_set_powergating_state(void *handle, return 0; } -static void sdma_v4_0_get_clockgating_state(void *handle, u32 *flags) +static void sdma_v4_0_get_clockgating_state(void *handle, u64 *flags) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; int data; @@ -2415,6 +2325,7 @@ static const struct amdgpu_ring_funcs sdma_v4_0_ring_funcs = { .align_mask = 0xf, .nop = SDMA_PKT_NOP_HEADER_OP(SDMA_OP_NOP), .support_64bit_ptrs = true, + .secure_submission_supported = true, .vmhub = AMDGPU_MMHUB_0, .get_rptr = sdma_v4_0_ring_get_rptr, .get_wptr = sdma_v4_0_ring_get_wptr, @@ -2451,6 +2362,7 @@ static const struct amdgpu_ring_funcs sdma_v4_0_ring_funcs_2nd_mmhub = { .align_mask = 0xf, .nop = SDMA_PKT_NOP_HEADER_OP(SDMA_OP_NOP), .support_64bit_ptrs = true, + .secure_submission_supported = true, .vmhub = AMDGPU_MMHUB_1, .get_rptr = sdma_v4_0_ring_get_rptr, .get_wptr = sdma_v4_0_ring_get_wptr, @@ -2483,6 +2395,7 @@ static const struct amdgpu_ring_funcs sdma_v4_0_page_ring_funcs = { .align_mask = 0xf, .nop = SDMA_PKT_NOP_HEADER_OP(SDMA_OP_NOP), .support_64bit_ptrs = true, + .secure_submission_supported = true, .vmhub = AMDGPU_MMHUB_0, .get_rptr = sdma_v4_0_ring_get_rptr, .get_wptr = sdma_v4_0_page_ring_get_wptr, @@ -2515,6 +2428,7 @@ static const struct amdgpu_ring_funcs sdma_v4_0_page_ring_funcs_2nd_mmhub = { .align_mask = 0xf, .nop = SDMA_PKT_NOP_HEADER_OP(SDMA_OP_NOP), .support_64bit_ptrs = true, + .secure_submission_supported = true, .vmhub = AMDGPU_MMHUB_1, .get_rptr = sdma_v4_0_ring_get_rptr, .get_wptr = sdma_v4_0_page_ring_get_wptr, @@ -2740,7 +2654,7 @@ static void sdma_v4_0_get_ras_error_count(uint32_t value, } } -static int sdma_v4_0_query_ras_error_count(struct amdgpu_device *adev, +static int sdma_v4_0_query_ras_error_count_by_instance(struct amdgpu_device *adev, uint32_t instance, void *ras_error_status) { struct ras_err_data *err_data = (struct ras_err_data *)ras_error_status; @@ -2762,6 +2676,18 @@ static int sdma_v4_0_query_ras_error_count(struct amdgpu_device *adev, return 0; }; +static void sdma_v4_0_query_ras_error_count(struct amdgpu_device *adev, void *ras_error_status) +{ + int i = 0; + + for (i = 0; i < adev->sdma.num_instances; i++) { + if (sdma_v4_0_query_ras_error_count_by_instance(adev, i, ras_error_status)) { + dev_err(adev->dev, "Query ras error count failed in SDMA%d\n", i); + return; + } + } +} + static void sdma_v4_0_reset_ras_error_count(struct amdgpu_device *adev) { int i; @@ -2773,26 +2699,48 @@ static void sdma_v4_0_reset_ras_error_count(struct amdgpu_device *adev) } } -static const struct amdgpu_sdma_ras_funcs sdma_v4_0_ras_funcs = { - .ras_late_init = amdgpu_sdma_ras_late_init, - .ras_fini = amdgpu_sdma_ras_fini, +const struct amdgpu_ras_block_hw_ops sdma_v4_0_ras_hw_ops = { .query_ras_error_count = sdma_v4_0_query_ras_error_count, .reset_ras_error_count = sdma_v4_0_reset_ras_error_count, }; +static struct amdgpu_sdma_ras sdma_v4_0_ras = { + .ras_block = { + .hw_ops = &sdma_v4_0_ras_hw_ops, + .ras_cb = sdma_v4_0_process_ras_data_cb, + }, +}; + static void sdma_v4_0_set_ras_funcs(struct amdgpu_device *adev) { switch (adev->ip_versions[SDMA0_HWIP][0]) { case IP_VERSION(4, 2, 0): case IP_VERSION(4, 2, 2): - adev->sdma.funcs = &sdma_v4_0_ras_funcs; + adev->sdma.ras = &sdma_v4_0_ras; break; case IP_VERSION(4, 4, 0): - adev->sdma.funcs = &sdma_v4_4_ras_funcs; + adev->sdma.ras = &sdma_v4_4_ras; break; default: break; } + + if (adev->sdma.ras) { + amdgpu_ras_register_ras_block(adev, &adev->sdma.ras->ras_block); + + strcpy(adev->sdma.ras->ras_block.ras_comm.name, "sdma"); + adev->sdma.ras->ras_block.ras_comm.block = AMDGPU_RAS_BLOCK__SDMA; + adev->sdma.ras->ras_block.ras_comm.type = AMDGPU_RAS_ERROR__MULTI_UNCORRECTABLE; + adev->sdma.ras_if = &adev->sdma.ras->ras_block.ras_comm; + + /* If don't define special ras_late_init function, use default ras_late_init */ + if (!adev->sdma.ras->ras_block.ras_late_init) + adev->sdma.ras->ras_block.ras_late_init = amdgpu_sdma_ras_late_init; + + /* If not defined special ras_cb function, use default ras_cb */ + if (!adev->sdma.ras->ras_block.ras_cb) + adev->sdma.ras->ras_block.ras_cb = amdgpu_sdma_process_ras_data_cb; + } } const struct amdgpu_ip_block_version sdma_v4_0_ip_block = { diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v4_4.c b/drivers/gpu/drm/amd/amdgpu/sdma_v4_4.c index bf95007f0843..6f9895cdddb1 100644 --- a/drivers/gpu/drm/amd/amdgpu/sdma_v4_4.c +++ b/drivers/gpu/drm/amd/amdgpu/sdma_v4_4.c @@ -188,7 +188,7 @@ static void sdma_v4_4_get_ras_error_count(struct amdgpu_device *adev, } } -static int sdma_v4_4_query_ras_error_count(struct amdgpu_device *adev, +static int sdma_v4_4_query_ras_error_count_by_instance(struct amdgpu_device *adev, uint32_t instance, void *ras_error_status) { @@ -245,9 +245,26 @@ static void sdma_v4_4_reset_ras_error_count(struct amdgpu_device *adev) } } -const struct amdgpu_sdma_ras_funcs sdma_v4_4_ras_funcs = { - .ras_late_init = amdgpu_sdma_ras_late_init, - .ras_fini = amdgpu_sdma_ras_fini, +static void sdma_v4_4_query_ras_error_count(struct amdgpu_device *adev, void *ras_error_status) +{ + int i = 0; + + for (i = 0; i < adev->sdma.num_instances; i++) { + if (sdma_v4_4_query_ras_error_count_by_instance(adev, i, ras_error_status)) { + dev_err(adev->dev, "Query ras error count failed in SDMA%d\n", i); + return; + } + } + +} + +const struct amdgpu_ras_block_hw_ops sdma_v4_4_ras_hw_ops = { .query_ras_error_count = sdma_v4_4_query_ras_error_count, .reset_ras_error_count = sdma_v4_4_reset_ras_error_count, }; + +struct amdgpu_sdma_ras sdma_v4_4_ras = { + .ras_block = { + .hw_ops = &sdma_v4_4_ras_hw_ops, + }, +}; diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v4_4.h b/drivers/gpu/drm/amd/amdgpu/sdma_v4_4.h index 74a6e5b5e949..a9f0c68359e0 100644 --- a/drivers/gpu/drm/amd/amdgpu/sdma_v4_4.h +++ b/drivers/gpu/drm/amd/amdgpu/sdma_v4_4.h @@ -23,6 +23,6 @@ #ifndef __SDMA_V4_4_H__ #define __SDMA_V4_4_H__ -extern const struct amdgpu_sdma_ras_funcs sdma_v4_4_ras_funcs; +extern struct amdgpu_sdma_ras sdma_v4_4_ras; #endif diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v5_0.c b/drivers/gpu/drm/amd/amdgpu/sdma_v5_0.c index 81e033549dda..d4d9f196db83 100644 --- a/drivers/gpu/drm/amd/amdgpu/sdma_v5_0.c +++ b/drivers/gpu/drm/amd/amdgpu/sdma_v5_0.c @@ -51,9 +51,6 @@ MODULE_FIRMWARE("amdgpu/navi14_sdma1.bin"); MODULE_FIRMWARE("amdgpu/navi12_sdma.bin"); MODULE_FIRMWARE("amdgpu/navi12_sdma1.bin"); -MODULE_FIRMWARE("amdgpu/cyan_skillfish_sdma.bin"); -MODULE_FIRMWARE("amdgpu/cyan_skillfish_sdma1.bin"); - MODULE_FIRMWARE("amdgpu/cyan_skillfish2_sdma.bin"); MODULE_FIRMWARE("amdgpu/cyan_skillfish2_sdma1.bin"); @@ -243,10 +240,7 @@ static int sdma_v5_0_init_microcode(struct amdgpu_device *adev) { const char *chip_name; char fw_name[40]; - int err = 0, i; - struct amdgpu_firmware_info *info = NULL; - const struct common_firmware_header *header = NULL; - const struct sdma_firmware_header_v1_0 *hdr; + int ret, i; if (amdgpu_sriov_vf(adev) && (adev->ip_versions[SDMA0_HWIP][0] == IP_VERSION(5, 0, 5))) return 0; @@ -264,10 +258,7 @@ static int sdma_v5_0_init_microcode(struct amdgpu_device *adev) chip_name = "navi12"; break; case IP_VERSION(5, 0, 1): - if (adev->apu_flags & AMD_APU_IS_CYAN_SKILLFISH2) - chip_name = "cyan_skillfish2"; - else - chip_name = "cyan_skillfish"; + chip_name = "cyan_skillfish2"; break; default: BUG(); @@ -278,38 +269,12 @@ static int sdma_v5_0_init_microcode(struct amdgpu_device *adev) snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_sdma.bin", chip_name); else snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_sdma1.bin", chip_name); - err = request_firmware(&adev->sdma.instance[i].fw, fw_name, adev->dev); - if (err) - goto out; - err = amdgpu_ucode_validate(adev->sdma.instance[i].fw); - if (err) - goto out; - hdr = (const struct sdma_firmware_header_v1_0 *)adev->sdma.instance[i].fw->data; - adev->sdma.instance[i].fw_version = le32_to_cpu(hdr->header.ucode_version); - adev->sdma.instance[i].feature_version = le32_to_cpu(hdr->ucode_feature_version); - if (adev->sdma.instance[i].feature_version >= 20) - adev->sdma.instance[i].burst_nop = true; - DRM_DEBUG("psp_load == '%s'\n", - adev->firmware.load_type == AMDGPU_FW_LOAD_PSP ? "true" : "false"); - - if (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) { - info = &adev->firmware.ucode[AMDGPU_UCODE_ID_SDMA0 + i]; - info->ucode_id = AMDGPU_UCODE_ID_SDMA0 + i; - info->fw = adev->sdma.instance[i].fw; - header = (const struct common_firmware_header *)info->fw->data; - adev->firmware.fw_size += - ALIGN(le32_to_cpu(header->ucode_size_bytes), PAGE_SIZE); - } + ret = amdgpu_sdma_init_microcode(adev, fw_name, i, false); + if (ret) + return ret; } -out: - if (err) { - DRM_ERROR("sdma_v5_0: Failed to load firmware \"%s\"\n", fw_name); - for (i = 0; i < adev->sdma.num_instances; i++) { - release_firmware(adev->sdma.instance[i].fw); - adev->sdma.instance[i].fw = NULL; - } - } - return err; + + return ret; } static unsigned sdma_v5_0_ring_init_cond_exec(struct amdgpu_ring *ring) @@ -353,7 +318,7 @@ static uint64_t sdma_v5_0_ring_get_rptr(struct amdgpu_ring *ring) u64 *rptr; /* XXX check if swapping is necessary on BE */ - rptr = ((u64 *)&ring->adev->wb.wb[ring->rptr_offs]); + rptr = (u64 *)ring->rptr_cpu_addr; DRM_DEBUG("rptr before shift == 0x%016llx\n", *rptr); return ((*rptr) >> 2); @@ -373,7 +338,7 @@ static uint64_t sdma_v5_0_ring_get_wptr(struct amdgpu_ring *ring) if (ring->use_doorbell) { /* XXX check if swapping is necessary on BE */ - wptr = READ_ONCE(*((u64 *)&adev->wb.wb[ring->wptr_offs])); + wptr = READ_ONCE(*((u64 *)ring->wptr_cpu_addr)); DRM_DEBUG("wptr/doorbell before shift == 0x%016llx\n", wptr); } else { wptr = RREG32_SOC15_IP(GC, sdma_v5_0_get_reg_offset(adev, ring->me, mmSDMA0_GFX_RB_WPTR_HI)); @@ -395,34 +360,67 @@ static uint64_t sdma_v5_0_ring_get_wptr(struct amdgpu_ring *ring) static void sdma_v5_0_ring_set_wptr(struct amdgpu_ring *ring) { struct amdgpu_device *adev = ring->adev; + uint32_t *wptr_saved; + uint32_t *is_queue_unmap; + uint64_t aggregated_db_index; + uint32_t mqd_size = adev->mqds[AMDGPU_HW_IP_DMA].mqd_size; DRM_DEBUG("Setting write pointer\n"); - if (ring->use_doorbell) { - DRM_DEBUG("Using doorbell -- " - "wptr_offs == 0x%08x " - "lower_32_bits(ring->wptr) << 2 == 0x%08x " - "upper_32_bits(ring->wptr) << 2 == 0x%08x\n", - ring->wptr_offs, - lower_32_bits(ring->wptr << 2), - upper_32_bits(ring->wptr << 2)); - /* XXX check if swapping is necessary on BE */ - adev->wb.wb[ring->wptr_offs] = lower_32_bits(ring->wptr << 2); - adev->wb.wb[ring->wptr_offs + 1] = upper_32_bits(ring->wptr << 2); - DRM_DEBUG("calling WDOORBELL64(0x%08x, 0x%016llx)\n", - ring->doorbell_index, ring->wptr << 2); - WDOORBELL64(ring->doorbell_index, ring->wptr << 2); + if (ring->is_mes_queue) { + wptr_saved = (uint32_t *)(ring->mqd_ptr + mqd_size); + is_queue_unmap = (uint32_t *)(ring->mqd_ptr + mqd_size + + sizeof(uint32_t)); + aggregated_db_index = + amdgpu_mes_get_aggregated_doorbell_index(adev, + AMDGPU_MES_PRIORITY_LEVEL_NORMAL); + + atomic64_set((atomic64_t *)ring->wptr_cpu_addr, + ring->wptr << 2); + *wptr_saved = ring->wptr << 2; + if (*is_queue_unmap) { + WDOORBELL64(aggregated_db_index, ring->wptr << 2); + DRM_DEBUG("calling WDOORBELL64(0x%08x, 0x%016llx)\n", + ring->doorbell_index, ring->wptr << 2); + WDOORBELL64(ring->doorbell_index, ring->wptr << 2); + } else { + DRM_DEBUG("calling WDOORBELL64(0x%08x, 0x%016llx)\n", + ring->doorbell_index, ring->wptr << 2); + WDOORBELL64(ring->doorbell_index, ring->wptr << 2); + + if (*is_queue_unmap) + WDOORBELL64(aggregated_db_index, + ring->wptr << 2); + } } else { - DRM_DEBUG("Not using doorbell -- " - "mmSDMA%i_GFX_RB_WPTR == 0x%08x " - "mmSDMA%i_GFX_RB_WPTR_HI == 0x%08x\n", - ring->me, - lower_32_bits(ring->wptr << 2), - ring->me, - upper_32_bits(ring->wptr << 2)); - WREG32_SOC15_IP(GC, sdma_v5_0_get_reg_offset(adev, ring->me, mmSDMA0_GFX_RB_WPTR), - lower_32_bits(ring->wptr << 2)); - WREG32_SOC15_IP(GC, sdma_v5_0_get_reg_offset(adev, ring->me, mmSDMA0_GFX_RB_WPTR_HI), - upper_32_bits(ring->wptr << 2)); + if (ring->use_doorbell) { + DRM_DEBUG("Using doorbell -- " + "wptr_offs == 0x%08x " + "lower_32_bits(ring->wptr) << 2 == 0x%08x " + "upper_32_bits(ring->wptr) << 2 == 0x%08x\n", + ring->wptr_offs, + lower_32_bits(ring->wptr << 2), + upper_32_bits(ring->wptr << 2)); + /* XXX check if swapping is necessary on BE */ + atomic64_set((atomic64_t *)ring->wptr_cpu_addr, + ring->wptr << 2); + DRM_DEBUG("calling WDOORBELL64(0x%08x, 0x%016llx)\n", + ring->doorbell_index, ring->wptr << 2); + WDOORBELL64(ring->doorbell_index, ring->wptr << 2); + } else { + DRM_DEBUG("Not using doorbell -- " + "mmSDMA%i_GFX_RB_WPTR == 0x%08x " + "mmSDMA%i_GFX_RB_WPTR_HI == 0x%08x\n", + ring->me, + lower_32_bits(ring->wptr << 2), + ring->me, + upper_32_bits(ring->wptr << 2)); + WREG32_SOC15_IP(GC, sdma_v5_0_get_reg_offset(adev, + ring->me, mmSDMA0_GFX_RB_WPTR), + lower_32_bits(ring->wptr << 2)); + WREG32_SOC15_IP(GC, sdma_v5_0_get_reg_offset(adev, + ring->me, mmSDMA0_GFX_RB_WPTR_HI), + upper_32_bits(ring->wptr << 2)); + } } } @@ -568,9 +566,11 @@ static void sdma_v5_0_ring_emit_fence(struct amdgpu_ring *ring, u64 addr, u64 se } if (flags & AMDGPU_FENCE_FLAG_INT) { + uint32_t ctx = ring->is_mes_queue ? + (ring->hw_queue_id | AMDGPU_FENCE_MES_QUEUE_FLAG) : 0; /* generate an interrupt */ amdgpu_ring_write(ring, SDMA_PKT_HEADER_OP(SDMA_OP_TRAP)); - amdgpu_ring_write(ring, SDMA_PKT_TRAP_INT_CONTEXT_INT_CONTEXT(0)); + amdgpu_ring_write(ring, SDMA_PKT_TRAP_INT_CONTEXT_INT_CONTEXT(ctx)); } } @@ -584,14 +584,10 @@ static void sdma_v5_0_ring_emit_fence(struct amdgpu_ring *ring, u64 addr, u64 se */ static void sdma_v5_0_gfx_stop(struct amdgpu_device *adev) { - struct amdgpu_ring *sdma0 = &adev->sdma.instance[0].ring; - struct amdgpu_ring *sdma1 = &adev->sdma.instance[1].ring; u32 rb_cntl, ib_cntl; int i; - if ((adev->mman.buffer_funcs_ring == sdma0) || - (adev->mman.buffer_funcs_ring == sdma1)) - amdgpu_ttm_set_buffer_funcs_status(adev, false); + amdgpu_sdma_unset_buffer_funcs_helper(adev); for (i = 0; i < adev->sdma.num_instances; i++) { rb_cntl = RREG32_SOC15_IP(GC, sdma_v5_0_get_reg_offset(adev, i, mmSDMA0_GFX_RB_CNTL)); @@ -714,7 +710,6 @@ static int sdma_v5_0_gfx_resume(struct amdgpu_device *adev) struct amdgpu_ring *ring; u32 rb_cntl, ib_cntl; u32 rb_bufsz; - u32 wb_offset; u32 doorbell; u32 doorbell_offset; u32 temp; @@ -724,7 +719,6 @@ static int sdma_v5_0_gfx_resume(struct amdgpu_device *adev) for (i = 0; i < adev->sdma.num_instances; i++) { ring = &adev->sdma.instance[i].ring; - wb_offset = (ring->rptr_offs * 4); if (!amdgpu_sriov_vf(adev)) WREG32(sdma_v5_0_get_reg_offset(adev, i, mmSDMA0_SEM_WAIT_FAIL_TIMER_CNTL), 0); @@ -747,7 +741,7 @@ static int sdma_v5_0_gfx_resume(struct amdgpu_device *adev) WREG32_SOC15_IP(GC, sdma_v5_0_get_reg_offset(adev, i, mmSDMA0_GFX_RB_WPTR_HI), 0); /* setup the wptr shadow polling */ - wptr_gpu_addr = adev->wb.gpu_addr + (ring->wptr_offs * 4); + wptr_gpu_addr = ring->wptr_gpu_addr; WREG32_SOC15_IP(GC, sdma_v5_0_get_reg_offset(adev, i, mmSDMA0_GFX_RB_WPTR_POLL_ADDR_LO), lower_32_bits(wptr_gpu_addr)); WREG32_SOC15_IP(GC, sdma_v5_0_get_reg_offset(adev, i, mmSDMA0_GFX_RB_WPTR_POLL_ADDR_HI), @@ -762,9 +756,9 @@ static int sdma_v5_0_gfx_resume(struct amdgpu_device *adev) /* set the wb address whether it's enabled or not */ WREG32_SOC15_IP(GC, sdma_v5_0_get_reg_offset(adev, i, mmSDMA0_GFX_RB_RPTR_ADDR_HI), - upper_32_bits(adev->wb.gpu_addr + wb_offset) & 0xFFFFFFFF); + upper_32_bits(ring->rptr_gpu_addr) & 0xFFFFFFFF); WREG32_SOC15_IP(GC, sdma_v5_0_get_reg_offset(adev, i, mmSDMA0_GFX_RB_RPTR_ADDR_LO), - lower_32_bits(adev->wb.gpu_addr + wb_offset) & 0xFFFFFFFC); + lower_32_bits(ring->rptr_gpu_addr) & 0xFFFFFFFC); rb_cntl = REG_SET_FIELD(rb_cntl, SDMA0_GFX_RB_CNTL, RPTR_WRITEBACK_ENABLE, 1); @@ -780,9 +774,9 @@ static int sdma_v5_0_gfx_resume(struct amdgpu_device *adev) if (!amdgpu_sriov_vf(adev)) { /* only bare-metal use register write for wptr */ WREG32(sdma_v5_0_get_reg_offset(adev, i, mmSDMA0_GFX_RB_WPTR), - lower_32_bits(ring->wptr) << 2); + lower_32_bits(ring->wptr << 2)); WREG32(sdma_v5_0_get_reg_offset(adev, i, mmSDMA0_GFX_RB_WPTR_HI), - upper_32_bits(ring->wptr) << 2); + upper_32_bits(ring->wptr << 2)); } doorbell = RREG32_SOC15_IP(GC, sdma_v5_0_get_reg_offset(adev, i, mmSDMA0_GFX_DOORBELL)); @@ -967,6 +961,49 @@ static int sdma_v5_0_start(struct amdgpu_device *adev) return r; } +static int sdma_v5_0_mqd_init(struct amdgpu_device *adev, void *mqd, + struct amdgpu_mqd_prop *prop) +{ + struct v10_sdma_mqd *m = mqd; + uint64_t wb_gpu_addr; + + m->sdmax_rlcx_rb_cntl = + order_base_2(prop->queue_size / 4) << SDMA0_RLC0_RB_CNTL__RB_SIZE__SHIFT | + 1 << SDMA0_RLC0_RB_CNTL__RPTR_WRITEBACK_ENABLE__SHIFT | + 6 << SDMA0_RLC0_RB_CNTL__RPTR_WRITEBACK_TIMER__SHIFT | + 1 << SDMA0_RLC0_RB_CNTL__RB_PRIV__SHIFT; + + m->sdmax_rlcx_rb_base = lower_32_bits(prop->hqd_base_gpu_addr >> 8); + m->sdmax_rlcx_rb_base_hi = upper_32_bits(prop->hqd_base_gpu_addr >> 8); + + m->sdmax_rlcx_rb_wptr_poll_cntl = RREG32(sdma_v5_0_get_reg_offset(adev, 0, + mmSDMA0_GFX_RB_WPTR_POLL_CNTL)); + + wb_gpu_addr = prop->wptr_gpu_addr; + m->sdmax_rlcx_rb_wptr_poll_addr_lo = lower_32_bits(wb_gpu_addr); + m->sdmax_rlcx_rb_wptr_poll_addr_hi = upper_32_bits(wb_gpu_addr); + + wb_gpu_addr = prop->rptr_gpu_addr; + m->sdmax_rlcx_rb_rptr_addr_lo = lower_32_bits(wb_gpu_addr); + m->sdmax_rlcx_rb_rptr_addr_hi = upper_32_bits(wb_gpu_addr); + + m->sdmax_rlcx_ib_cntl = RREG32(sdma_v5_0_get_reg_offset(adev, 0, + mmSDMA0_GFX_IB_CNTL)); + + m->sdmax_rlcx_doorbell_offset = + prop->doorbell_index << SDMA0_RLC0_DOORBELL_OFFSET__OFFSET__SHIFT; + + m->sdmax_rlcx_doorbell = REG_SET_FIELD(0, SDMA0_RLC0_DOORBELL, ENABLE, 1); + + return 0; +} + +static void sdma_v5_0_set_mqd_funcs(struct amdgpu_device *adev) +{ + adev->mqds[AMDGPU_HW_IP_DMA].mqd_size = sizeof(struct v10_sdma_mqd); + adev->mqds[AMDGPU_HW_IP_DMA].init_mqd = sdma_v5_0_mqd_init; +} + /** * sdma_v5_0_ring_test_ring - simple async dma engine test * @@ -984,18 +1021,29 @@ static int sdma_v5_0_ring_test_ring(struct amdgpu_ring *ring) int r; u32 tmp; u64 gpu_addr; + volatile uint32_t *cpu_ptr = NULL; - r = amdgpu_device_wb_get(adev, &index); - if (r) { - dev_err(adev->dev, "(%d) failed to allocate wb slot\n", r); - return r; - } - - gpu_addr = adev->wb.gpu_addr + (index * 4); tmp = 0xCAFEDEAD; - adev->wb.wb[index] = cpu_to_le32(tmp); - r = amdgpu_ring_alloc(ring, 5); + if (ring->is_mes_queue) { + uint32_t offset = 0; + offset = amdgpu_mes_ctx_get_offs(ring, + AMDGPU_MES_CTX_PADDING_OFFS); + gpu_addr = amdgpu_mes_ctx_get_offs_gpu_addr(ring, offset); + cpu_ptr = amdgpu_mes_ctx_get_offs_cpu_addr(ring, offset); + *cpu_ptr = tmp; + } else { + r = amdgpu_device_wb_get(adev, &index); + if (r) { + dev_err(adev->dev, "(%d) failed to allocate wb slot\n", r); + return r; + } + + gpu_addr = adev->wb.gpu_addr + (index * 4); + adev->wb.wb[index] = cpu_to_le32(tmp); + } + + r = amdgpu_ring_alloc(ring, 20); if (r) { DRM_ERROR("amdgpu: dma failed to lock ring %d (%d).\n", ring->idx, r); amdgpu_device_wb_free(adev, index); @@ -1011,7 +1059,10 @@ static int sdma_v5_0_ring_test_ring(struct amdgpu_ring *ring) amdgpu_ring_commit(ring); for (i = 0; i < adev->usec_timeout; i++) { - tmp = le32_to_cpu(adev->wb.wb[index]); + if (ring->is_mes_queue) + tmp = le32_to_cpu(*cpu_ptr); + else + tmp = le32_to_cpu(adev->wb.wb[index]); if (tmp == 0xDEADBEEF) break; if (amdgpu_emu_mode == 1) @@ -1023,7 +1074,8 @@ static int sdma_v5_0_ring_test_ring(struct amdgpu_ring *ring) if (i >= adev->usec_timeout) r = -ETIMEDOUT; - amdgpu_device_wb_free(adev, index); + if (!ring->is_mes_queue) + amdgpu_device_wb_free(adev, index); return r; } @@ -1046,22 +1098,38 @@ static int sdma_v5_0_ring_test_ib(struct amdgpu_ring *ring, long timeout) long r; u32 tmp = 0; u64 gpu_addr; + volatile uint32_t *cpu_ptr = NULL; - r = amdgpu_device_wb_get(adev, &index); - if (r) { - dev_err(adev->dev, "(%ld) failed to allocate wb slot\n", r); - return r; - } - - gpu_addr = adev->wb.gpu_addr + (index * 4); tmp = 0xCAFEDEAD; - adev->wb.wb[index] = cpu_to_le32(tmp); memset(&ib, 0, sizeof(ib)); - r = amdgpu_ib_get(adev, NULL, 256, + + if (ring->is_mes_queue) { + uint32_t offset = 0; + offset = amdgpu_mes_ctx_get_offs(ring, AMDGPU_MES_CTX_IB_OFFS); + ib.gpu_addr = amdgpu_mes_ctx_get_offs_gpu_addr(ring, offset); + ib.ptr = (void *)amdgpu_mes_ctx_get_offs_cpu_addr(ring, offset); + + offset = amdgpu_mes_ctx_get_offs(ring, + AMDGPU_MES_CTX_PADDING_OFFS); + gpu_addr = amdgpu_mes_ctx_get_offs_gpu_addr(ring, offset); + cpu_ptr = amdgpu_mes_ctx_get_offs_cpu_addr(ring, offset); + *cpu_ptr = tmp; + } else { + r = amdgpu_device_wb_get(adev, &index); + if (r) { + dev_err(adev->dev, "(%ld) failed to allocate wb slot\n", r); + return r; + } + + gpu_addr = adev->wb.gpu_addr + (index * 4); + adev->wb.wb[index] = cpu_to_le32(tmp); + + r = amdgpu_ib_get(adev, NULL, 256, AMDGPU_IB_POOL_DIRECT, &ib); - if (r) { - DRM_ERROR("amdgpu: failed to get ib (%ld).\n", r); - goto err0; + if (r) { + DRM_ERROR("amdgpu: failed to get ib (%ld).\n", r); + goto err0; + } } ib.ptr[0] = SDMA_PKT_HEADER_OP(SDMA_OP_WRITE) | @@ -1088,7 +1156,12 @@ static int sdma_v5_0_ring_test_ib(struct amdgpu_ring *ring, long timeout) DRM_ERROR("amdgpu: fence wait failed (%ld).\n", r); goto err1; } - tmp = le32_to_cpu(adev->wb.wb[index]); + + if (ring->is_mes_queue) + tmp = le32_to_cpu(*cpu_ptr); + else + tmp = le32_to_cpu(adev->wb.wb[index]); + if (tmp == 0xDEADBEEF) r = 0; else @@ -1098,7 +1171,8 @@ err1: amdgpu_ib_free(adev, &ib, NULL); dma_fence_put(f); err0: - amdgpu_device_wb_free(adev, index); + if (!ring->is_mes_queue) + amdgpu_device_wb_free(adev, index); return r; } @@ -1297,6 +1371,7 @@ static int sdma_v5_0_early_init(void *handle) sdma_v5_0_set_buffer_funcs(adev); sdma_v5_0_set_vm_pte_funcs(adev); sdma_v5_0_set_irq_funcs(adev); + sdma_v5_0_set_mqd_funcs(adev); return 0; } @@ -1357,12 +1432,10 @@ static int sdma_v5_0_sw_fini(void *handle) struct amdgpu_device *adev = (struct amdgpu_device *)handle; int i; - for (i = 0; i < adev->sdma.num_instances; i++) { - release_firmware(adev->sdma.instance[i].fw); - adev->sdma.instance[i].fw = NULL; - + for (i = 0; i < adev->sdma.num_instances; i++) amdgpu_ring_fini(&adev->sdma.instance[i].ring); - } + + amdgpu_sdma_destroy_inst_ctx(adev, false); return 0; } @@ -1383,8 +1456,11 @@ static int sdma_v5_0_hw_fini(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; - if (amdgpu_sriov_vf(adev)) + if (amdgpu_sriov_vf(adev)) { + /* disable the scheduler for SDMA */ + amdgpu_sdma_unset_buffer_funcs_helper(adev); return 0; + } sdma_v5_0_ctx_switch_enable(adev, false); sdma_v5_0_enable(adev, false); @@ -1517,7 +1593,25 @@ static int sdma_v5_0_process_trap_irq(struct amdgpu_device *adev, struct amdgpu_irq_src *source, struct amdgpu_iv_entry *entry) { + uint32_t mes_queue_id = entry->src_data[0]; + DRM_DEBUG("IH: SDMA trap\n"); + + if (adev->enable_mes && (mes_queue_id & AMDGPU_FENCE_MES_QUEUE_FLAG)) { + struct amdgpu_mes_queue *queue; + + mes_queue_id &= AMDGPU_FENCE_MES_QUEUE_ID_MASK; + + spin_lock(&adev->mes.queue_id_lock); + queue = idr_find(&adev->mes.queue_id_idr, mes_queue_id); + if (queue) { + DRM_DEBUG("process smda queue id = %d\n", mes_queue_id); + amdgpu_fence_process(queue->ring); + } + spin_unlock(&adev->mes.queue_id_lock); + return 0; + } + switch (entry->client_id) { case SOC15_IH_CLIENTID_SDMA0: switch (entry->ring_id) { @@ -1654,7 +1748,7 @@ static int sdma_v5_0_set_powergating_state(void *handle, return 0; } -static void sdma_v5_0_get_clockgating_state(void *handle, u32 *flags) +static void sdma_v5_0_get_clockgating_state(void *handle, u64 *flags) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; int data; @@ -1696,6 +1790,7 @@ static const struct amdgpu_ring_funcs sdma_v5_0_ring_funcs = { .align_mask = 0xf, .nop = SDMA_PKT_NOP_HEADER_OP(SDMA_OP_NOP), .support_64bit_ptrs = true, + .secure_submission_supported = true, .vmhub = AMDGPU_GFXHUB_0, .get_rptr = sdma_v5_0_ring_get_rptr, .get_wptr = sdma_v5_0_ring_get_wptr, diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v5_2.c b/drivers/gpu/drm/amd/amdgpu/sdma_v5_2.c index d3d6d5b045b8..809eca54fc61 100644 --- a/drivers/gpu/drm/amd/amdgpu/sdma_v5_2.c +++ b/drivers/gpu/drm/amd/amdgpu/sdma_v5_2.c @@ -51,6 +51,8 @@ MODULE_FIRMWARE("amdgpu/beige_goby_sdma.bin"); MODULE_FIRMWARE("amdgpu/vangogh_sdma.bin"); MODULE_FIRMWARE("amdgpu/yellow_carp_sdma.bin"); +MODULE_FIRMWARE("amdgpu/sdma_5_2_6.bin"); +MODULE_FIRMWARE("amdgpu/sdma_5_2_7.bin"); #define SDMA1_REG_OFFSET 0x600 #define SDMA3_REG_OFFSET 0x400 @@ -87,33 +89,6 @@ static u32 sdma_v5_2_get_reg_offset(struct amdgpu_device *adev, u32 instance, u3 return base + internal_offset; } -static int sdma_v5_2_init_inst_ctx(struct amdgpu_sdma_instance *sdma_inst) -{ - int err = 0; - const struct sdma_firmware_header_v1_0 *hdr; - - err = amdgpu_ucode_validate(sdma_inst->fw); - if (err) - return err; - - hdr = (const struct sdma_firmware_header_v1_0 *)sdma_inst->fw->data; - sdma_inst->fw_version = le32_to_cpu(hdr->header.ucode_version); - sdma_inst->feature_version = le32_to_cpu(hdr->ucode_feature_version); - - if (sdma_inst->feature_version >= 20) - sdma_inst->burst_nop = true; - - return 0; -} - -static void sdma_v5_2_destroy_inst_ctx(struct amdgpu_device *adev) -{ - release_firmware(adev->sdma.instance[0].fw); - - memset((void *)adev->sdma.instance, 0, - sizeof(struct amdgpu_sdma_instance) * AMDGPU_MAX_SDMA_INSTANCES); -} - /** * sdma_v5_2_init_microcode - load ucode images from disk * @@ -130,73 +105,41 @@ static int sdma_v5_2_init_microcode(struct amdgpu_device *adev) { const char *chip_name; char fw_name[40]; - int err = 0, i; - struct amdgpu_firmware_info *info = NULL; - const struct common_firmware_header *header = NULL; DRM_DEBUG("\n"); switch (adev->ip_versions[SDMA0_HWIP][0]) { case IP_VERSION(5, 2, 0): - chip_name = "sienna_cichlid"; + chip_name = "sienna_cichlid_sdma"; break; case IP_VERSION(5, 2, 2): - chip_name = "navy_flounder"; + chip_name = "navy_flounder_sdma"; break; case IP_VERSION(5, 2, 1): - chip_name = "vangogh"; + chip_name = "vangogh_sdma"; break; case IP_VERSION(5, 2, 4): - chip_name = "dimgrey_cavefish"; + chip_name = "dimgrey_cavefish_sdma"; break; case IP_VERSION(5, 2, 5): - chip_name = "beige_goby"; + chip_name = "beige_goby_sdma"; break; case IP_VERSION(5, 2, 3): - chip_name = "yellow_carp"; + chip_name = "yellow_carp_sdma"; + break; + case IP_VERSION(5, 2, 6): + chip_name = "sdma_5_2_6"; + break; + case IP_VERSION(5, 2, 7): + chip_name = "sdma_5_2_7"; break; default: BUG(); } - snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_sdma.bin", chip_name); - - err = request_firmware(&adev->sdma.instance[0].fw, fw_name, adev->dev); - if (err) - goto out; - - err = sdma_v5_2_init_inst_ctx(&adev->sdma.instance[0]); - if (err) - goto out; - - for (i = 1; i < adev->sdma.num_instances; i++) - memcpy((void *)&adev->sdma.instance[i], - (void *)&adev->sdma.instance[0], - sizeof(struct amdgpu_sdma_instance)); + snprintf(fw_name, sizeof(fw_name), "amdgpu/%s.bin", chip_name); - if (amdgpu_sriov_vf(adev) && (adev->ip_versions[SDMA0_HWIP][0] == IP_VERSION(5, 2, 0))) - return 0; - - DRM_DEBUG("psp_load == '%s'\n", - adev->firmware.load_type == AMDGPU_FW_LOAD_PSP ? "true" : "false"); - - if (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) { - for (i = 0; i < adev->sdma.num_instances; i++) { - info = &adev->firmware.ucode[AMDGPU_UCODE_ID_SDMA0 + i]; - info->ucode_id = AMDGPU_UCODE_ID_SDMA0 + i; - info->fw = adev->sdma.instance[i].fw; - header = (const struct common_firmware_header *)info->fw->data; - adev->firmware.fw_size += - ALIGN(le32_to_cpu(header->ucode_size_bytes), PAGE_SIZE); - } - } - -out: - if (err) { - DRM_ERROR("sdma_v5_2: Failed to load firmware \"%s\"\n", fw_name); - sdma_v5_2_destroy_inst_ctx(adev); - } - return err; + return amdgpu_sdma_init_microcode(adev, fw_name, 0, true); } static unsigned sdma_v5_2_ring_init_cond_exec(struct amdgpu_ring *ring) @@ -240,7 +183,7 @@ static uint64_t sdma_v5_2_ring_get_rptr(struct amdgpu_ring *ring) u64 *rptr; /* XXX check if swapping is necessary on BE */ - rptr = ((u64 *)&ring->adev->wb.wb[ring->rptr_offs]); + rptr = (u64 *)ring->rptr_cpu_addr; DRM_DEBUG("rptr before shift == 0x%016llx\n", *rptr); return ((*rptr) >> 2); @@ -260,7 +203,7 @@ static uint64_t sdma_v5_2_ring_get_wptr(struct amdgpu_ring *ring) if (ring->use_doorbell) { /* XXX check if swapping is necessary on BE */ - wptr = READ_ONCE(*((u64 *)&adev->wb.wb[ring->wptr_offs])); + wptr = READ_ONCE(*((u64 *)ring->wptr_cpu_addr)); DRM_DEBUG("wptr/doorbell before shift == 0x%016llx\n", wptr); } else { wptr = RREG32(sdma_v5_2_get_reg_offset(adev, ring->me, mmSDMA0_GFX_RB_WPTR_HI)); @@ -287,14 +230,14 @@ static void sdma_v5_2_ring_set_wptr(struct amdgpu_ring *ring) if (ring->use_doorbell) { DRM_DEBUG("Using doorbell -- " "wptr_offs == 0x%08x " - "lower_32_bits(ring->wptr) << 2 == 0x%08x " - "upper_32_bits(ring->wptr) << 2 == 0x%08x\n", + "lower_32_bits(ring->wptr << 2) == 0x%08x " + "upper_32_bits(ring->wptr << 2) == 0x%08x\n", ring->wptr_offs, lower_32_bits(ring->wptr << 2), upper_32_bits(ring->wptr << 2)); /* XXX check if swapping is necessary on BE */ - adev->wb.wb[ring->wptr_offs] = lower_32_bits(ring->wptr << 2); - adev->wb.wb[ring->wptr_offs + 1] = upper_32_bits(ring->wptr << 2); + atomic64_set((atomic64_t *)ring->wptr_cpu_addr, + ring->wptr << 2); DRM_DEBUG("calling WDOORBELL64(0x%08x, 0x%016llx)\n", ring->doorbell_index, ring->wptr << 2); WDOORBELL64(ring->doorbell_index, ring->wptr << 2); @@ -452,10 +395,12 @@ static void sdma_v5_2_ring_emit_fence(struct amdgpu_ring *ring, u64 addr, u64 se amdgpu_ring_write(ring, upper_32_bits(seq)); } - if (flags & AMDGPU_FENCE_FLAG_INT) { + if ((flags & AMDGPU_FENCE_FLAG_INT)) { + uint32_t ctx = ring->is_mes_queue ? + (ring->hw_queue_id | AMDGPU_FENCE_MES_QUEUE_FLAG) : 0; /* generate an interrupt */ amdgpu_ring_write(ring, SDMA_PKT_HEADER_OP(SDMA_OP_TRAP)); - amdgpu_ring_write(ring, SDMA_PKT_TRAP_INT_CONTEXT_INT_CONTEXT(0)); + amdgpu_ring_write(ring, SDMA_PKT_TRAP_INT_CONTEXT_INT_CONTEXT(ctx)); } } @@ -469,18 +414,10 @@ static void sdma_v5_2_ring_emit_fence(struct amdgpu_ring *ring, u64 addr, u64 se */ static void sdma_v5_2_gfx_stop(struct amdgpu_device *adev) { - struct amdgpu_ring *sdma0 = &adev->sdma.instance[0].ring; - struct amdgpu_ring *sdma1 = &adev->sdma.instance[1].ring; - struct amdgpu_ring *sdma2 = &adev->sdma.instance[2].ring; - struct amdgpu_ring *sdma3 = &adev->sdma.instance[3].ring; u32 rb_cntl, ib_cntl; int i; - if ((adev->mman.buffer_funcs_ring == sdma0) || - (adev->mman.buffer_funcs_ring == sdma1) || - (adev->mman.buffer_funcs_ring == sdma2) || - (adev->mman.buffer_funcs_ring == sdma3)) - amdgpu_ttm_set_buffer_funcs_status(adev, false); + amdgpu_sdma_unset_buffer_funcs_helper(adev); for (i = 0; i < adev->sdma.num_instances; i++) { rb_cntl = RREG32_SOC15_IP(GC, sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_GFX_RB_CNTL)); @@ -601,7 +538,6 @@ static int sdma_v5_2_gfx_resume(struct amdgpu_device *adev) struct amdgpu_ring *ring; u32 rb_cntl, ib_cntl; u32 rb_bufsz; - u32 wb_offset; u32 doorbell; u32 doorbell_offset; u32 temp; @@ -611,7 +547,6 @@ static int sdma_v5_2_gfx_resume(struct amdgpu_device *adev) for (i = 0; i < adev->sdma.num_instances; i++) { ring = &adev->sdma.instance[i].ring; - wb_offset = (ring->rptr_offs * 4); if (!amdgpu_sriov_vf(adev)) WREG32_SOC15_IP(GC, sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_SEM_WAIT_FAIL_TIMER_CNTL), 0); @@ -634,7 +569,7 @@ static int sdma_v5_2_gfx_resume(struct amdgpu_device *adev) WREG32_SOC15_IP(GC, sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_GFX_RB_WPTR_HI), 0); /* setup the wptr shadow polling */ - wptr_gpu_addr = adev->wb.gpu_addr + (ring->wptr_offs * 4); + wptr_gpu_addr = ring->wptr_gpu_addr; WREG32_SOC15_IP(GC, sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_GFX_RB_WPTR_POLL_ADDR_LO), lower_32_bits(wptr_gpu_addr)); WREG32_SOC15_IP(GC, sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_GFX_RB_WPTR_POLL_ADDR_HI), @@ -649,9 +584,9 @@ static int sdma_v5_2_gfx_resume(struct amdgpu_device *adev) /* set the wb address whether it's enabled or not */ WREG32_SOC15_IP(GC, sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_GFX_RB_RPTR_ADDR_HI), - upper_32_bits(adev->wb.gpu_addr + wb_offset) & 0xFFFFFFFF); + upper_32_bits(ring->rptr_gpu_addr) & 0xFFFFFFFF); WREG32_SOC15_IP(GC, sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_GFX_RB_RPTR_ADDR_LO), - lower_32_bits(adev->wb.gpu_addr + wb_offset) & 0xFFFFFFFC); + lower_32_bits(ring->rptr_gpu_addr) & 0xFFFFFFFC); rb_cntl = REG_SET_FIELD(rb_cntl, SDMA0_GFX_RB_CNTL, RPTR_WRITEBACK_ENABLE, 1); @@ -664,8 +599,8 @@ static int sdma_v5_2_gfx_resume(struct amdgpu_device *adev) WREG32_SOC15_IP(GC, sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_GFX_MINOR_PTR_UPDATE), 1); if (!amdgpu_sriov_vf(adev)) { /* only bare-metal use register write for wptr */ - WREG32(sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_GFX_RB_WPTR), lower_32_bits(ring->wptr) << 2); - WREG32(sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_GFX_RB_WPTR_HI), upper_32_bits(ring->wptr) << 2); + WREG32(sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_GFX_RB_WPTR), lower_32_bits(ring->wptr << 2)); + WREG32(sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_GFX_RB_WPTR_HI), upper_32_bits(ring->wptr << 2)); } doorbell = RREG32_SOC15_IP(GC, sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_GFX_DOORBELL)); @@ -897,6 +832,49 @@ static int sdma_v5_2_start(struct amdgpu_device *adev) return r; } +static int sdma_v5_2_mqd_init(struct amdgpu_device *adev, void *mqd, + struct amdgpu_mqd_prop *prop) +{ + struct v10_sdma_mqd *m = mqd; + uint64_t wb_gpu_addr; + + m->sdmax_rlcx_rb_cntl = + order_base_2(prop->queue_size / 4) << SDMA0_RLC0_RB_CNTL__RB_SIZE__SHIFT | + 1 << SDMA0_RLC0_RB_CNTL__RPTR_WRITEBACK_ENABLE__SHIFT | + 6 << SDMA0_RLC0_RB_CNTL__RPTR_WRITEBACK_TIMER__SHIFT | + 1 << SDMA0_RLC0_RB_CNTL__RB_PRIV__SHIFT; + + m->sdmax_rlcx_rb_base = lower_32_bits(prop->hqd_base_gpu_addr >> 8); + m->sdmax_rlcx_rb_base_hi = upper_32_bits(prop->hqd_base_gpu_addr >> 8); + + m->sdmax_rlcx_rb_wptr_poll_cntl = RREG32(sdma_v5_2_get_reg_offset(adev, 0, + mmSDMA0_GFX_RB_WPTR_POLL_CNTL)); + + wb_gpu_addr = prop->wptr_gpu_addr; + m->sdmax_rlcx_rb_wptr_poll_addr_lo = lower_32_bits(wb_gpu_addr); + m->sdmax_rlcx_rb_wptr_poll_addr_hi = upper_32_bits(wb_gpu_addr); + + wb_gpu_addr = prop->rptr_gpu_addr; + m->sdmax_rlcx_rb_rptr_addr_lo = lower_32_bits(wb_gpu_addr); + m->sdmax_rlcx_rb_rptr_addr_hi = upper_32_bits(wb_gpu_addr); + + m->sdmax_rlcx_ib_cntl = RREG32(sdma_v5_2_get_reg_offset(adev, 0, + mmSDMA0_GFX_IB_CNTL)); + + m->sdmax_rlcx_doorbell_offset = + prop->doorbell_index << SDMA0_RLC0_DOORBELL_OFFSET__OFFSET__SHIFT; + + m->sdmax_rlcx_doorbell = REG_SET_FIELD(0, SDMA0_RLC0_DOORBELL, ENABLE, 1); + + return 0; +} + +static void sdma_v5_2_set_mqd_funcs(struct amdgpu_device *adev) +{ + adev->mqds[AMDGPU_HW_IP_DMA].mqd_size = sizeof(struct v10_sdma_mqd); + adev->mqds[AMDGPU_HW_IP_DMA].init_mqd = sdma_v5_2_mqd_init; +} + /** * sdma_v5_2_ring_test_ring - simple async dma engine test * @@ -914,18 +892,29 @@ static int sdma_v5_2_ring_test_ring(struct amdgpu_ring *ring) int r; u32 tmp; u64 gpu_addr; + volatile uint32_t *cpu_ptr = NULL; - r = amdgpu_device_wb_get(adev, &index); - if (r) { - dev_err(adev->dev, "(%d) failed to allocate wb slot\n", r); - return r; - } - - gpu_addr = adev->wb.gpu_addr + (index * 4); tmp = 0xCAFEDEAD; - adev->wb.wb[index] = cpu_to_le32(tmp); - r = amdgpu_ring_alloc(ring, 5); + if (ring->is_mes_queue) { + uint32_t offset = 0; + offset = amdgpu_mes_ctx_get_offs(ring, + AMDGPU_MES_CTX_PADDING_OFFS); + gpu_addr = amdgpu_mes_ctx_get_offs_gpu_addr(ring, offset); + cpu_ptr = amdgpu_mes_ctx_get_offs_cpu_addr(ring, offset); + *cpu_ptr = tmp; + } else { + r = amdgpu_device_wb_get(adev, &index); + if (r) { + dev_err(adev->dev, "(%d) failed to allocate wb slot\n", r); + return r; + } + + gpu_addr = adev->wb.gpu_addr + (index * 4); + adev->wb.wb[index] = cpu_to_le32(tmp); + } + + r = amdgpu_ring_alloc(ring, 20); if (r) { DRM_ERROR("amdgpu: dma failed to lock ring %d (%d).\n", ring->idx, r); amdgpu_device_wb_free(adev, index); @@ -941,7 +930,10 @@ static int sdma_v5_2_ring_test_ring(struct amdgpu_ring *ring) amdgpu_ring_commit(ring); for (i = 0; i < adev->usec_timeout; i++) { - tmp = le32_to_cpu(adev->wb.wb[index]); + if (ring->is_mes_queue) + tmp = le32_to_cpu(*cpu_ptr); + else + tmp = le32_to_cpu(adev->wb.wb[index]); if (tmp == 0xDEADBEEF) break; if (amdgpu_emu_mode == 1) @@ -953,7 +945,8 @@ static int sdma_v5_2_ring_test_ring(struct amdgpu_ring *ring) if (i >= adev->usec_timeout) r = -ETIMEDOUT; - amdgpu_device_wb_free(adev, index); + if (!ring->is_mes_queue) + amdgpu_device_wb_free(adev, index); return r; } @@ -976,21 +969,37 @@ static int sdma_v5_2_ring_test_ib(struct amdgpu_ring *ring, long timeout) long r; u32 tmp = 0; u64 gpu_addr; + volatile uint32_t *cpu_ptr = NULL; - r = amdgpu_device_wb_get(adev, &index); - if (r) { - dev_err(adev->dev, "(%ld) failed to allocate wb slot\n", r); - return r; - } - - gpu_addr = adev->wb.gpu_addr + (index * 4); tmp = 0xCAFEDEAD; - adev->wb.wb[index] = cpu_to_le32(tmp); memset(&ib, 0, sizeof(ib)); - r = amdgpu_ib_get(adev, NULL, 256, AMDGPU_IB_POOL_DIRECT, &ib); - if (r) { - DRM_ERROR("amdgpu: failed to get ib (%ld).\n", r); - goto err0; + + if (ring->is_mes_queue) { + uint32_t offset = 0; + offset = amdgpu_mes_ctx_get_offs(ring, AMDGPU_MES_CTX_IB_OFFS); + ib.gpu_addr = amdgpu_mes_ctx_get_offs_gpu_addr(ring, offset); + ib.ptr = (void *)amdgpu_mes_ctx_get_offs_cpu_addr(ring, offset); + + offset = amdgpu_mes_ctx_get_offs(ring, + AMDGPU_MES_CTX_PADDING_OFFS); + gpu_addr = amdgpu_mes_ctx_get_offs_gpu_addr(ring, offset); + cpu_ptr = amdgpu_mes_ctx_get_offs_cpu_addr(ring, offset); + *cpu_ptr = tmp; + } else { + r = amdgpu_device_wb_get(adev, &index); + if (r) { + dev_err(adev->dev, "(%ld) failed to allocate wb slot\n", r); + return r; + } + + gpu_addr = adev->wb.gpu_addr + (index * 4); + adev->wb.wb[index] = cpu_to_le32(tmp); + + r = amdgpu_ib_get(adev, NULL, 256, AMDGPU_IB_POOL_DIRECT, &ib); + if (r) { + DRM_ERROR("amdgpu: failed to get ib (%ld).\n", r); + goto err0; + } } ib.ptr[0] = SDMA_PKT_HEADER_OP(SDMA_OP_WRITE) | @@ -1017,7 +1026,12 @@ static int sdma_v5_2_ring_test_ib(struct amdgpu_ring *ring, long timeout) DRM_ERROR("amdgpu: fence wait failed (%ld).\n", r); goto err1; } - tmp = le32_to_cpu(adev->wb.wb[index]); + + if (ring->is_mes_queue) + tmp = le32_to_cpu(*cpu_ptr); + else + tmp = le32_to_cpu(adev->wb.wb[index]); + if (tmp == 0xDEADBEEF) r = 0; else @@ -1027,7 +1041,8 @@ err1: amdgpu_ib_free(adev, &ib, NULL); dma_fence_put(f); err0: - amdgpu_device_wb_free(adev, index); + if (!ring->is_mes_queue) + amdgpu_device_wb_free(adev, index); return r; } @@ -1227,6 +1242,7 @@ static int sdma_v5_2_early_init(void *handle) sdma_v5_2_set_buffer_funcs(adev); sdma_v5_2_set_vm_pte_funcs(adev); sdma_v5_2_set_irq_funcs(adev); + sdma_v5_2_set_mqd_funcs(adev); return 0; } @@ -1317,27 +1333,27 @@ static int sdma_v5_2_sw_fini(void *handle) for (i = 0; i < adev->sdma.num_instances; i++) amdgpu_ring_fini(&adev->sdma.instance[i].ring); - sdma_v5_2_destroy_inst_ctx(adev); + amdgpu_sdma_destroy_inst_ctx(adev, true); return 0; } static int sdma_v5_2_hw_init(void *handle) { - int r; struct amdgpu_device *adev = (struct amdgpu_device *)handle; - r = sdma_v5_2_start(adev); - - return r; + return sdma_v5_2_start(adev); } static int sdma_v5_2_hw_fini(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; - if (amdgpu_sriov_vf(adev)) + if (amdgpu_sriov_vf(adev)) { + /* disable the scheduler for SDMA */ + amdgpu_sdma_unset_buffer_funcs_helper(adev); return 0; + } sdma_v5_2_ctx_switch_enable(adev, false); sdma_v5_2_enable(adev, false); @@ -1460,7 +1476,25 @@ static int sdma_v5_2_process_trap_irq(struct amdgpu_device *adev, struct amdgpu_irq_src *source, struct amdgpu_iv_entry *entry) { + uint32_t mes_queue_id = entry->src_data[0]; + DRM_DEBUG("IH: SDMA trap\n"); + + if (adev->enable_mes && (mes_queue_id & AMDGPU_FENCE_MES_QUEUE_FLAG)) { + struct amdgpu_mes_queue *queue; + + mes_queue_id &= AMDGPU_FENCE_MES_QUEUE_ID_MASK; + + spin_lock(&adev->mes.queue_id_lock); + queue = idr_find(&adev->mes.queue_id_idr, mes_queue_id); + if (queue) { + DRM_DEBUG("process smda queue id = %d\n", mes_queue_id); + amdgpu_fence_process(queue->ring); + } + spin_unlock(&adev->mes.queue_id_lock); + return 0; + } + switch (entry->client_id) { case SOC15_IH_CLIENTID_SDMA0: switch (entry->ring_id) { @@ -1617,6 +1651,7 @@ static int sdma_v5_2_set_clockgating_state(void *handle, case IP_VERSION(5, 2, 1): case IP_VERSION(5, 2, 4): case IP_VERSION(5, 2, 5): + case IP_VERSION(5, 2, 6): case IP_VERSION(5, 2, 3): sdma_v5_2_update_medium_grain_clock_gating(adev, state == AMD_CG_STATE_GATE); @@ -1636,7 +1671,7 @@ static int sdma_v5_2_set_powergating_state(void *handle, return 0; } -static void sdma_v5_2_get_clockgating_state(void *handle, u32 *flags) +static void sdma_v5_2_get_clockgating_state(void *handle, u64 *flags) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; int data; @@ -1644,6 +1679,11 @@ static void sdma_v5_2_get_clockgating_state(void *handle, u32 *flags) if (amdgpu_sriov_vf(adev)) *flags = 0; + /* AMD_CG_SUPPORT_SDMA_MGCG */ + data = RREG32(sdma_v5_2_get_reg_offset(adev, 0, mmSDMA0_CLK_CTRL)); + if (!(data & SDMA0_CLK_CTRL__CGCG_EN_OVERRIDE_MASK)) + *flags |= AMD_CG_SUPPORT_SDMA_MGCG; + /* AMD_CG_SUPPORT_SDMA_LS */ data = RREG32_KIQ(sdma_v5_2_get_reg_offset(adev, 0, mmSDMA0_POWER_CNTL)); if (data & SDMA0_POWER_CNTL__MEM_POWER_OVERRIDE_MASK) @@ -1673,6 +1713,7 @@ static const struct amdgpu_ring_funcs sdma_v5_2_ring_funcs = { .align_mask = 0xf, .nop = SDMA_PKT_NOP_HEADER_OP(SDMA_OP_NOP), .support_64bit_ptrs = true, + .secure_submission_supported = true, .vmhub = AMDGPU_GFXHUB_0, .get_rptr = sdma_v5_2_ring_get_rptr, .get_wptr = sdma_v5_2_ring_get_wptr, diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v6_0.c b/drivers/gpu/drm/amd/amdgpu/sdma_v6_0.c new file mode 100644 index 000000000000..da3beb0bf2fa --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/sdma_v6_0.c @@ -0,0 +1,1676 @@ +/* + * Copyright 2020 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#include <linux/delay.h> +#include <linux/firmware.h> +#include <linux/module.h> +#include <linux/pci.h> + +#include "amdgpu.h" +#include "amdgpu_ucode.h" +#include "amdgpu_trace.h" + +#include "gc/gc_11_0_0_offset.h" +#include "gc/gc_11_0_0_sh_mask.h" +#include "gc/gc_11_0_0_default.h" +#include "hdp/hdp_6_0_0_offset.h" +#include "ivsrcid/gfx/irqsrcs_gfx_11_0_0.h" + +#include "soc15_common.h" +#include "soc15.h" +#include "sdma_v6_0_0_pkt_open.h" +#include "nbio_v4_3.h" +#include "sdma_common.h" +#include "sdma_v6_0.h" +#include "v11_structs.h" + +MODULE_FIRMWARE("amdgpu/sdma_6_0_0.bin"); +MODULE_FIRMWARE("amdgpu/sdma_6_0_1.bin"); +MODULE_FIRMWARE("amdgpu/sdma_6_0_2.bin"); +MODULE_FIRMWARE("amdgpu/sdma_6_0_3.bin"); + +#define SDMA1_REG_OFFSET 0x600 +#define SDMA0_HYP_DEC_REG_START 0x5880 +#define SDMA0_HYP_DEC_REG_END 0x589a +#define SDMA1_HYP_DEC_REG_OFFSET 0x20 + +static void sdma_v6_0_set_ring_funcs(struct amdgpu_device *adev); +static void sdma_v6_0_set_buffer_funcs(struct amdgpu_device *adev); +static void sdma_v6_0_set_vm_pte_funcs(struct amdgpu_device *adev); +static void sdma_v6_0_set_irq_funcs(struct amdgpu_device *adev); +static int sdma_v6_0_start(struct amdgpu_device *adev); + +static u32 sdma_v6_0_get_reg_offset(struct amdgpu_device *adev, u32 instance, u32 internal_offset) +{ + u32 base; + + if (internal_offset >= SDMA0_HYP_DEC_REG_START && + internal_offset <= SDMA0_HYP_DEC_REG_END) { + base = adev->reg_offset[GC_HWIP][0][1]; + if (instance != 0) + internal_offset += SDMA1_HYP_DEC_REG_OFFSET * instance; + } else { + base = adev->reg_offset[GC_HWIP][0][0]; + if (instance == 1) + internal_offset += SDMA1_REG_OFFSET; + } + + return base + internal_offset; +} + +/** + * sdma_v6_0_init_microcode - load ucode images from disk + * + * @adev: amdgpu_device pointer + * + * Use the firmware interface to load the ucode images into + * the driver (not loaded into hw). + * Returns 0 on success, error on failure. + */ +static int sdma_v6_0_init_microcode(struct amdgpu_device *adev) +{ + char fw_name[30]; + char ucode_prefix[30]; + + DRM_DEBUG("\n"); + + amdgpu_ucode_ip_version_decode(adev, SDMA0_HWIP, ucode_prefix, sizeof(ucode_prefix)); + + snprintf(fw_name, sizeof(fw_name), "amdgpu/%s.bin", ucode_prefix); + + return amdgpu_sdma_init_microcode(adev, fw_name, 0, true); +} + +static unsigned sdma_v6_0_ring_init_cond_exec(struct amdgpu_ring *ring) +{ + unsigned ret; + + amdgpu_ring_write(ring, SDMA_PKT_COPY_LINEAR_HEADER_OP(SDMA_OP_COND_EXE)); + amdgpu_ring_write(ring, lower_32_bits(ring->cond_exe_gpu_addr)); + amdgpu_ring_write(ring, upper_32_bits(ring->cond_exe_gpu_addr)); + amdgpu_ring_write(ring, 1); + ret = ring->wptr & ring->buf_mask;/* this is the offset we need patch later */ + amdgpu_ring_write(ring, 0x55aa55aa);/* insert dummy here and patch it later */ + + return ret; +} + +static void sdma_v6_0_ring_patch_cond_exec(struct amdgpu_ring *ring, + unsigned offset) +{ + unsigned cur; + + BUG_ON(offset > ring->buf_mask); + BUG_ON(ring->ring[offset] != 0x55aa55aa); + + cur = (ring->wptr - 1) & ring->buf_mask; + if (cur > offset) + ring->ring[offset] = cur - offset; + else + ring->ring[offset] = (ring->buf_mask + 1) - offset + cur; +} + +/** + * sdma_v6_0_ring_get_rptr - get the current read pointer + * + * @ring: amdgpu ring pointer + * + * Get the current rptr from the hardware. + */ +static uint64_t sdma_v6_0_ring_get_rptr(struct amdgpu_ring *ring) +{ + u64 *rptr; + + /* XXX check if swapping is necessary on BE */ + rptr = (u64 *)ring->rptr_cpu_addr; + + DRM_DEBUG("rptr before shift == 0x%016llx\n", *rptr); + return ((*rptr) >> 2); +} + +/** + * sdma_v6_0_ring_get_wptr - get the current write pointer + * + * @ring: amdgpu ring pointer + * + * Get the current wptr from the hardware. + */ +static uint64_t sdma_v6_0_ring_get_wptr(struct amdgpu_ring *ring) +{ + u64 wptr = 0; + + if (ring->use_doorbell) { + /* XXX check if swapping is necessary on BE */ + wptr = READ_ONCE(*((u64 *)ring->wptr_cpu_addr)); + DRM_DEBUG("wptr/doorbell before shift == 0x%016llx\n", wptr); + } + + return wptr >> 2; +} + +/** + * sdma_v6_0_ring_set_wptr - commit the write pointer + * + * @ring: amdgpu ring pointer + * + * Write the wptr back to the hardware. + */ +static void sdma_v6_0_ring_set_wptr(struct amdgpu_ring *ring) +{ + struct amdgpu_device *adev = ring->adev; + uint32_t *wptr_saved; + uint32_t *is_queue_unmap; + uint64_t aggregated_db_index; + uint32_t mqd_size = adev->mqds[AMDGPU_HW_IP_DMA].mqd_size; + + DRM_DEBUG("Setting write pointer\n"); + + if (ring->is_mes_queue) { + wptr_saved = (uint32_t *)(ring->mqd_ptr + mqd_size); + is_queue_unmap = (uint32_t *)(ring->mqd_ptr + mqd_size + + sizeof(uint32_t)); + aggregated_db_index = + amdgpu_mes_get_aggregated_doorbell_index(adev, + ring->hw_prio); + + atomic64_set((atomic64_t *)ring->wptr_cpu_addr, + ring->wptr << 2); + *wptr_saved = ring->wptr << 2; + if (*is_queue_unmap) { + WDOORBELL64(aggregated_db_index, ring->wptr << 2); + DRM_DEBUG("calling WDOORBELL64(0x%08x, 0x%016llx)\n", + ring->doorbell_index, ring->wptr << 2); + WDOORBELL64(ring->doorbell_index, ring->wptr << 2); + } else { + DRM_DEBUG("calling WDOORBELL64(0x%08x, 0x%016llx)\n", + ring->doorbell_index, ring->wptr << 2); + WDOORBELL64(ring->doorbell_index, ring->wptr << 2); + + if (*is_queue_unmap) + WDOORBELL64(aggregated_db_index, + ring->wptr << 2); + } + } else { + if (ring->use_doorbell) { + DRM_DEBUG("Using doorbell -- " + "wptr_offs == 0x%08x " + "lower_32_bits(ring->wptr) << 2 == 0x%08x " + "upper_32_bits(ring->wptr) << 2 == 0x%08x\n", + ring->wptr_offs, + lower_32_bits(ring->wptr << 2), + upper_32_bits(ring->wptr << 2)); + /* XXX check if swapping is necessary on BE */ + atomic64_set((atomic64_t *)ring->wptr_cpu_addr, + ring->wptr << 2); + DRM_DEBUG("calling WDOORBELL64(0x%08x, 0x%016llx)\n", + ring->doorbell_index, ring->wptr << 2); + WDOORBELL64(ring->doorbell_index, ring->wptr << 2); + } else { + DRM_DEBUG("Not using doorbell -- " + "regSDMA%i_GFX_RB_WPTR == 0x%08x " + "regSDMA%i_GFX_RB_WPTR_HI == 0x%08x\n", + ring->me, + lower_32_bits(ring->wptr << 2), + ring->me, + upper_32_bits(ring->wptr << 2)); + WREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev, + ring->me, regSDMA0_QUEUE0_RB_WPTR), + lower_32_bits(ring->wptr << 2)); + WREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev, + ring->me, regSDMA0_QUEUE0_RB_WPTR_HI), + upper_32_bits(ring->wptr << 2)); + } + } +} + +static void sdma_v6_0_ring_insert_nop(struct amdgpu_ring *ring, uint32_t count) +{ + struct amdgpu_sdma_instance *sdma = amdgpu_sdma_get_instance_from_ring(ring); + int i; + + for (i = 0; i < count; i++) + if (sdma && sdma->burst_nop && (i == 0)) + amdgpu_ring_write(ring, ring->funcs->nop | + SDMA_PKT_NOP_HEADER_COUNT(count - 1)); + else + amdgpu_ring_write(ring, ring->funcs->nop); +} + +/** + * sdma_v6_0_ring_emit_ib - Schedule an IB on the DMA engine + * + * @ring: amdgpu ring pointer + * @ib: IB object to schedule + * + * Schedule an IB in the DMA ring. + */ +static void sdma_v6_0_ring_emit_ib(struct amdgpu_ring *ring, + struct amdgpu_job *job, + struct amdgpu_ib *ib, + uint32_t flags) +{ + unsigned vmid = AMDGPU_JOB_GET_VMID(job); + uint64_t csa_mc_addr = amdgpu_sdma_get_csa_mc_addr(ring, vmid); + + /* An IB packet must end on a 8 DW boundary--the next dword + * must be on a 8-dword boundary. Our IB packet below is 6 + * dwords long, thus add x number of NOPs, such that, in + * modular arithmetic, + * wptr + 6 + x = 8k, k >= 0, which in C is, + * (wptr + 6 + x) % 8 = 0. + * The expression below, is a solution of x. + */ + sdma_v6_0_ring_insert_nop(ring, (2 - lower_32_bits(ring->wptr)) & 7); + + amdgpu_ring_write(ring, SDMA_PKT_COPY_LINEAR_HEADER_OP(SDMA_OP_INDIRECT) | + SDMA_PKT_INDIRECT_HEADER_VMID(vmid & 0xf)); + /* base must be 32 byte aligned */ + amdgpu_ring_write(ring, lower_32_bits(ib->gpu_addr) & 0xffffffe0); + amdgpu_ring_write(ring, upper_32_bits(ib->gpu_addr)); + amdgpu_ring_write(ring, ib->length_dw); + amdgpu_ring_write(ring, lower_32_bits(csa_mc_addr)); + amdgpu_ring_write(ring, upper_32_bits(csa_mc_addr)); +} + +/** + * sdma_v6_0_ring_emit_mem_sync - flush the IB by graphics cache rinse + * + * @ring: amdgpu ring pointer + * @job: job to retrieve vmid from + * @ib: IB object to schedule + * + * flush the IB by graphics cache rinse. + */ +static void sdma_v6_0_ring_emit_mem_sync(struct amdgpu_ring *ring) +{ + uint32_t gcr_cntl = SDMA_GCR_GL2_INV | SDMA_GCR_GL2_WB | SDMA_GCR_GLM_INV | + SDMA_GCR_GL1_INV | SDMA_GCR_GLV_INV | SDMA_GCR_GLK_INV | + SDMA_GCR_GLI_INV(1); + + /* flush entire cache L0/L1/L2, this can be optimized by performance requirement */ + amdgpu_ring_write(ring, SDMA_PKT_COPY_LINEAR_HEADER_OP(SDMA_OP_GCR_REQ)); + amdgpu_ring_write(ring, SDMA_PKT_GCR_REQ_PAYLOAD1_BASE_VA_31_7(0)); + amdgpu_ring_write(ring, SDMA_PKT_GCR_REQ_PAYLOAD2_GCR_CONTROL_15_0(gcr_cntl) | + SDMA_PKT_GCR_REQ_PAYLOAD2_BASE_VA_47_32(0)); + amdgpu_ring_write(ring, SDMA_PKT_GCR_REQ_PAYLOAD3_LIMIT_VA_31_7(0) | + SDMA_PKT_GCR_REQ_PAYLOAD3_GCR_CONTROL_18_16(gcr_cntl >> 16)); + amdgpu_ring_write(ring, SDMA_PKT_GCR_REQ_PAYLOAD4_LIMIT_VA_47_32(0) | + SDMA_PKT_GCR_REQ_PAYLOAD4_VMID(0)); +} + + +/** + * sdma_v6_0_ring_emit_hdp_flush - emit an hdp flush on the DMA ring + * + * @ring: amdgpu ring pointer + * + * Emit an hdp flush packet on the requested DMA ring. + */ +static void sdma_v6_0_ring_emit_hdp_flush(struct amdgpu_ring *ring) +{ + struct amdgpu_device *adev = ring->adev; + u32 ref_and_mask = 0; + const struct nbio_hdp_flush_reg *nbio_hf_reg = adev->nbio.hdp_flush_reg; + + ref_and_mask = nbio_hf_reg->ref_and_mask_sdma0 << ring->me; + + amdgpu_ring_write(ring, SDMA_PKT_COPY_LINEAR_HEADER_OP(SDMA_OP_POLL_REGMEM) | + SDMA_PKT_POLL_REGMEM_HEADER_HDP_FLUSH(1) | + SDMA_PKT_POLL_REGMEM_HEADER_FUNC(3)); /* == */ + amdgpu_ring_write(ring, (adev->nbio.funcs->get_hdp_flush_done_offset(adev)) << 2); + amdgpu_ring_write(ring, (adev->nbio.funcs->get_hdp_flush_req_offset(adev)) << 2); + amdgpu_ring_write(ring, ref_and_mask); /* reference */ + amdgpu_ring_write(ring, ref_and_mask); /* mask */ + amdgpu_ring_write(ring, SDMA_PKT_POLL_REGMEM_DW5_RETRY_COUNT(0xfff) | + SDMA_PKT_POLL_REGMEM_DW5_INTERVAL(10)); /* retry count, poll interval */ +} + +/** + * sdma_v6_0_ring_emit_fence - emit a fence on the DMA ring + * + * @ring: amdgpu ring pointer + * @fence: amdgpu fence object + * + * Add a DMA fence packet to the ring to write + * the fence seq number and DMA trap packet to generate + * an interrupt if needed. + */ +static void sdma_v6_0_ring_emit_fence(struct amdgpu_ring *ring, u64 addr, u64 seq, + unsigned flags) +{ + bool write64bit = flags & AMDGPU_FENCE_FLAG_64BIT; + /* write the fence */ + amdgpu_ring_write(ring, SDMA_PKT_COPY_LINEAR_HEADER_OP(SDMA_OP_FENCE) | + SDMA_PKT_FENCE_HEADER_MTYPE(0x3)); /* Ucached(UC) */ + /* zero in first two bits */ + BUG_ON(addr & 0x3); + amdgpu_ring_write(ring, lower_32_bits(addr)); + amdgpu_ring_write(ring, upper_32_bits(addr)); + amdgpu_ring_write(ring, lower_32_bits(seq)); + + /* optionally write high bits as well */ + if (write64bit) { + addr += 4; + amdgpu_ring_write(ring, SDMA_PKT_COPY_LINEAR_HEADER_OP(SDMA_OP_FENCE) | + SDMA_PKT_FENCE_HEADER_MTYPE(0x3)); + /* zero in first two bits */ + BUG_ON(addr & 0x3); + amdgpu_ring_write(ring, lower_32_bits(addr)); + amdgpu_ring_write(ring, upper_32_bits(addr)); + amdgpu_ring_write(ring, upper_32_bits(seq)); + } + + if (flags & AMDGPU_FENCE_FLAG_INT) { + uint32_t ctx = ring->is_mes_queue ? + (ring->hw_queue_id | AMDGPU_FENCE_MES_QUEUE_FLAG) : 0; + /* generate an interrupt */ + amdgpu_ring_write(ring, SDMA_PKT_COPY_LINEAR_HEADER_OP(SDMA_OP_TRAP)); + amdgpu_ring_write(ring, SDMA_PKT_TRAP_INT_CONTEXT_INT_CONTEXT(ctx)); + } +} + +/** + * sdma_v6_0_gfx_stop - stop the gfx async dma engines + * + * @adev: amdgpu_device pointer + * + * Stop the gfx async dma ring buffers. + */ +static void sdma_v6_0_gfx_stop(struct amdgpu_device *adev) +{ + u32 rb_cntl, ib_cntl; + int i; + + amdgpu_sdma_unset_buffer_funcs_helper(adev); + + for (i = 0; i < adev->sdma.num_instances; i++) { + rb_cntl = RREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev, i, regSDMA0_QUEUE0_RB_CNTL)); + rb_cntl = REG_SET_FIELD(rb_cntl, SDMA0_QUEUE0_RB_CNTL, RB_ENABLE, 0); + WREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev, i, regSDMA0_QUEUE0_RB_CNTL), rb_cntl); + ib_cntl = RREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev, i, regSDMA0_QUEUE0_IB_CNTL)); + ib_cntl = REG_SET_FIELD(ib_cntl, SDMA0_QUEUE0_IB_CNTL, IB_ENABLE, 0); + WREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev, i, regSDMA0_QUEUE0_IB_CNTL), ib_cntl); + } +} + +/** + * sdma_v6_0_rlc_stop - stop the compute async dma engines + * + * @adev: amdgpu_device pointer + * + * Stop the compute async dma queues. + */ +static void sdma_v6_0_rlc_stop(struct amdgpu_device *adev) +{ + /* XXX todo */ +} + +/** + * sdma_v6_0_ctx_switch_enable - stop the async dma engines context switch + * + * @adev: amdgpu_device pointer + * @enable: enable/disable the DMA MEs context switch. + * + * Halt or unhalt the async dma engines context switch. + */ +static void sdma_v6_0_ctx_switch_enable(struct amdgpu_device *adev, bool enable) +{ +} + +/** + * sdma_v6_0_enable - stop the async dma engines + * + * @adev: amdgpu_device pointer + * @enable: enable/disable the DMA MEs. + * + * Halt or unhalt the async dma engines. + */ +static void sdma_v6_0_enable(struct amdgpu_device *adev, bool enable) +{ + u32 f32_cntl; + int i; + + if (!enable) { + sdma_v6_0_gfx_stop(adev); + sdma_v6_0_rlc_stop(adev); + } + + for (i = 0; i < adev->sdma.num_instances; i++) { + f32_cntl = RREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev, i, regSDMA0_F32_CNTL)); + f32_cntl = REG_SET_FIELD(f32_cntl, SDMA0_F32_CNTL, HALT, enable ? 0 : 1); + WREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev, i, regSDMA0_F32_CNTL), f32_cntl); + } +} + +/** + * sdma_v6_0_gfx_resume - setup and start the async dma engines + * + * @adev: amdgpu_device pointer + * + * Set up the gfx DMA ring buffers and enable them. + * Returns 0 for success, error for failure. + */ +static int sdma_v6_0_gfx_resume(struct amdgpu_device *adev) +{ + struct amdgpu_ring *ring; + u32 rb_cntl, ib_cntl; + u32 rb_bufsz; + u32 doorbell; + u32 doorbell_offset; + u32 temp; + u64 wptr_gpu_addr; + int i, r; + + for (i = 0; i < adev->sdma.num_instances; i++) { + ring = &adev->sdma.instance[i].ring; + + if (!amdgpu_sriov_vf(adev)) + WREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev, i, regSDMA0_SEM_WAIT_FAIL_TIMER_CNTL), 0); + + /* Set ring buffer size in dwords */ + rb_bufsz = order_base_2(ring->ring_size / 4); + rb_cntl = RREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev, i, regSDMA0_QUEUE0_RB_CNTL)); + rb_cntl = REG_SET_FIELD(rb_cntl, SDMA0_QUEUE0_RB_CNTL, RB_SIZE, rb_bufsz); +#ifdef __BIG_ENDIAN + rb_cntl = REG_SET_FIELD(rb_cntl, SDMA0_QUEUE0_RB_CNTL, RB_SWAP_ENABLE, 1); + rb_cntl = REG_SET_FIELD(rb_cntl, SDMA0_QUEUE0_RB_CNTL, + RPTR_WRITEBACK_SWAP_ENABLE, 1); +#endif + rb_cntl = REG_SET_FIELD(rb_cntl, SDMA0_QUEUE0_RB_CNTL, RB_PRIV, 1); + WREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev, i, regSDMA0_QUEUE0_RB_CNTL), rb_cntl); + + /* Initialize the ring buffer's read and write pointers */ + WREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev, i, regSDMA0_QUEUE0_RB_RPTR), 0); + WREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev, i, regSDMA0_QUEUE0_RB_RPTR_HI), 0); + WREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev, i, regSDMA0_QUEUE0_RB_WPTR), 0); + WREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev, i, regSDMA0_QUEUE0_RB_WPTR_HI), 0); + + /* setup the wptr shadow polling */ + wptr_gpu_addr = ring->wptr_gpu_addr; + WREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev, i, regSDMA0_QUEUE0_RB_WPTR_POLL_ADDR_LO), + lower_32_bits(wptr_gpu_addr)); + WREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev, i, regSDMA0_QUEUE0_RB_WPTR_POLL_ADDR_HI), + upper_32_bits(wptr_gpu_addr)); + + /* set the wb address whether it's enabled or not */ + WREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev, i, regSDMA0_QUEUE0_RB_RPTR_ADDR_HI), + upper_32_bits(ring->rptr_gpu_addr) & 0xFFFFFFFF); + WREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev, i, regSDMA0_QUEUE0_RB_RPTR_ADDR_LO), + lower_32_bits(ring->rptr_gpu_addr) & 0xFFFFFFFC); + + rb_cntl = REG_SET_FIELD(rb_cntl, SDMA0_QUEUE0_RB_CNTL, RPTR_WRITEBACK_ENABLE, 1); + if (amdgpu_sriov_vf(adev)) + rb_cntl = REG_SET_FIELD(rb_cntl, SDMA0_QUEUE0_RB_CNTL, WPTR_POLL_ENABLE, 1); + else + rb_cntl = REG_SET_FIELD(rb_cntl, SDMA0_QUEUE0_RB_CNTL, WPTR_POLL_ENABLE, 0); + rb_cntl = REG_SET_FIELD(rb_cntl, SDMA0_QUEUE0_RB_CNTL, F32_WPTR_POLL_ENABLE, 1); + + WREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev, i, regSDMA0_QUEUE0_RB_BASE), ring->gpu_addr >> 8); + WREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev, i, regSDMA0_QUEUE0_RB_BASE_HI), ring->gpu_addr >> 40); + + ring->wptr = 0; + + /* before programing wptr to a less value, need set minor_ptr_update first */ + WREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev, i, regSDMA0_QUEUE0_MINOR_PTR_UPDATE), 1); + + if (!amdgpu_sriov_vf(adev)) { /* only bare-metal use register write for wptr */ + WREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev, i, regSDMA0_QUEUE0_RB_WPTR), lower_32_bits(ring->wptr) << 2); + WREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev, i, regSDMA0_QUEUE0_RB_WPTR_HI), upper_32_bits(ring->wptr) << 2); + } + + doorbell = RREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev, i, regSDMA0_QUEUE0_DOORBELL)); + doorbell_offset = RREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev, i, regSDMA0_QUEUE0_DOORBELL_OFFSET)); + + if (ring->use_doorbell) { + doorbell = REG_SET_FIELD(doorbell, SDMA0_QUEUE0_DOORBELL, ENABLE, 1); + doorbell_offset = REG_SET_FIELD(doorbell_offset, SDMA0_QUEUE0_DOORBELL_OFFSET, + OFFSET, ring->doorbell_index); + } else { + doorbell = REG_SET_FIELD(doorbell, SDMA0_QUEUE0_DOORBELL, ENABLE, 0); + } + WREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev, i, regSDMA0_QUEUE0_DOORBELL), doorbell); + WREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev, i, regSDMA0_QUEUE0_DOORBELL_OFFSET), doorbell_offset); + + if (i == 0) + adev->nbio.funcs->sdma_doorbell_range(adev, i, ring->use_doorbell, + ring->doorbell_index, + adev->doorbell_index.sdma_doorbell_range * adev->sdma.num_instances); + + if (amdgpu_sriov_vf(adev)) + sdma_v6_0_ring_set_wptr(ring); + + /* set minor_ptr_update to 0 after wptr programed */ + WREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev, i, regSDMA0_QUEUE0_MINOR_PTR_UPDATE), 0); + + /* Set up RESP_MODE to non-copy addresses */ + temp = RREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev, i, regSDMA0_UTCL1_CNTL)); + temp = REG_SET_FIELD(temp, SDMA0_UTCL1_CNTL, RESP_MODE, 3); + temp = REG_SET_FIELD(temp, SDMA0_UTCL1_CNTL, REDO_DELAY, 9); + WREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev, i, regSDMA0_UTCL1_CNTL), temp); + + /* program default cache read and write policy */ + temp = RREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev, i, regSDMA0_UTCL1_PAGE)); + /* clean read policy and write policy bits */ + temp &= 0xFF0FFF; + temp |= ((CACHE_READ_POLICY_L2__DEFAULT << 12) | + (CACHE_WRITE_POLICY_L2__DEFAULT << 14) | + SDMA0_UTCL1_PAGE__LLC_NOALLOC_MASK); + WREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev, i, regSDMA0_UTCL1_PAGE), temp); + + if (!amdgpu_sriov_vf(adev)) { + /* unhalt engine */ + temp = RREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev, i, regSDMA0_F32_CNTL)); + temp = REG_SET_FIELD(temp, SDMA0_F32_CNTL, HALT, 0); + temp = REG_SET_FIELD(temp, SDMA0_F32_CNTL, TH1_RESET, 0); + WREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev, i, regSDMA0_F32_CNTL), temp); + } + + /* enable DMA RB */ + rb_cntl = REG_SET_FIELD(rb_cntl, SDMA0_QUEUE0_RB_CNTL, RB_ENABLE, 1); + WREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev, i, regSDMA0_QUEUE0_RB_CNTL), rb_cntl); + + ib_cntl = RREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev, i, regSDMA0_QUEUE0_IB_CNTL)); + ib_cntl = REG_SET_FIELD(ib_cntl, SDMA0_QUEUE0_IB_CNTL, IB_ENABLE, 1); +#ifdef __BIG_ENDIAN + ib_cntl = REG_SET_FIELD(ib_cntl, SDMA0_QUEUE0_IB_CNTL, IB_SWAP_ENABLE, 1); +#endif + /* enable DMA IBs */ + WREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev, i, regSDMA0_QUEUE0_IB_CNTL), ib_cntl); + + ring->sched.ready = true; + + if (amdgpu_sriov_vf(adev)) { /* bare-metal sequence doesn't need below to lines */ + sdma_v6_0_ctx_switch_enable(adev, true); + sdma_v6_0_enable(adev, true); + } + + r = amdgpu_ring_test_helper(ring); + if (r) { + ring->sched.ready = false; + return r; + } + + if (adev->mman.buffer_funcs_ring == ring) + amdgpu_ttm_set_buffer_funcs_status(adev, true); + } + + return 0; +} + +/** + * sdma_v6_0_rlc_resume - setup and start the async dma engines + * + * @adev: amdgpu_device pointer + * + * Set up the compute DMA queues and enable them. + * Returns 0 for success, error for failure. + */ +static int sdma_v6_0_rlc_resume(struct amdgpu_device *adev) +{ + return 0; +} + +/** + * sdma_v6_0_load_microcode - load the sDMA ME ucode + * + * @adev: amdgpu_device pointer + * + * Loads the sDMA0/1 ucode. + * Returns 0 for success, -EINVAL if the ucode is not available. + */ +static int sdma_v6_0_load_microcode(struct amdgpu_device *adev) +{ + const struct sdma_firmware_header_v2_0 *hdr; + const __le32 *fw_data; + u32 fw_size; + int i, j; + bool use_broadcast; + + /* halt the MEs */ + sdma_v6_0_enable(adev, false); + + if (!adev->sdma.instance[0].fw) + return -EINVAL; + + /* use broadcast mode to load SDMA microcode by default */ + use_broadcast = true; + + if (use_broadcast) { + dev_info(adev->dev, "Use broadcast method to load SDMA firmware\n"); + /* load Control Thread microcode */ + hdr = (const struct sdma_firmware_header_v2_0 *)adev->sdma.instance[0].fw->data; + amdgpu_ucode_print_sdma_hdr(&hdr->header); + fw_size = le32_to_cpu(hdr->ctx_jt_offset + hdr->ctx_jt_size) / 4; + + fw_data = (const __le32 *) + (adev->sdma.instance[0].fw->data + + le32_to_cpu(hdr->header.ucode_array_offset_bytes)); + + WREG32(sdma_v6_0_get_reg_offset(adev, 0, regSDMA0_BROADCAST_UCODE_ADDR), 0); + + for (j = 0; j < fw_size; j++) { + if (amdgpu_emu_mode == 1 && j % 500 == 0) + msleep(1); + WREG32(sdma_v6_0_get_reg_offset(adev, 0, regSDMA0_BROADCAST_UCODE_DATA), le32_to_cpup(fw_data++)); + } + + /* load Context Switch microcode */ + fw_size = le32_to_cpu(hdr->ctl_jt_offset + hdr->ctl_jt_size) / 4; + + fw_data = (const __le32 *) + (adev->sdma.instance[0].fw->data + + le32_to_cpu(hdr->ctl_ucode_offset)); + + WREG32(sdma_v6_0_get_reg_offset(adev, 0, regSDMA0_BROADCAST_UCODE_ADDR), 0x8000); + + for (j = 0; j < fw_size; j++) { + if (amdgpu_emu_mode == 1 && j % 500 == 0) + msleep(1); + WREG32(sdma_v6_0_get_reg_offset(adev, 0, regSDMA0_BROADCAST_UCODE_DATA), le32_to_cpup(fw_data++)); + } + } else { + dev_info(adev->dev, "Use legacy method to load SDMA firmware\n"); + for (i = 0; i < adev->sdma.num_instances; i++) { + /* load Control Thread microcode */ + hdr = (const struct sdma_firmware_header_v2_0 *)adev->sdma.instance[0].fw->data; + amdgpu_ucode_print_sdma_hdr(&hdr->header); + fw_size = le32_to_cpu(hdr->ctx_jt_offset + hdr->ctx_jt_size) / 4; + + fw_data = (const __le32 *) + (adev->sdma.instance[0].fw->data + + le32_to_cpu(hdr->header.ucode_array_offset_bytes)); + + WREG32(sdma_v6_0_get_reg_offset(adev, i, regSDMA0_UCODE_ADDR), 0); + + for (j = 0; j < fw_size; j++) { + if (amdgpu_emu_mode == 1 && j % 500 == 0) + msleep(1); + WREG32(sdma_v6_0_get_reg_offset(adev, i, regSDMA0_UCODE_DATA), le32_to_cpup(fw_data++)); + } + + WREG32(sdma_v6_0_get_reg_offset(adev, i, regSDMA0_UCODE_ADDR), adev->sdma.instance[0].fw_version); + + /* load Context Switch microcode */ + fw_size = le32_to_cpu(hdr->ctl_jt_offset + hdr->ctl_jt_size) / 4; + + fw_data = (const __le32 *) + (adev->sdma.instance[0].fw->data + + le32_to_cpu(hdr->ctl_ucode_offset)); + + WREG32(sdma_v6_0_get_reg_offset(adev, i, regSDMA0_UCODE_ADDR), 0x8000); + + for (j = 0; j < fw_size; j++) { + if (amdgpu_emu_mode == 1 && j % 500 == 0) + msleep(1); + WREG32(sdma_v6_0_get_reg_offset(adev, i, regSDMA0_UCODE_DATA), le32_to_cpup(fw_data++)); + } + + WREG32(sdma_v6_0_get_reg_offset(adev, i, regSDMA0_UCODE_ADDR), adev->sdma.instance[0].fw_version); + } + } + + return 0; +} + +static int sdma_v6_0_soft_reset(void *handle) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + u32 tmp; + int i; + + sdma_v6_0_gfx_stop(adev); + + for (i = 0; i < adev->sdma.num_instances; i++) { + tmp = RREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev, i, regSDMA0_FREEZE)); + tmp |= SDMA0_FREEZE__FREEZE_MASK; + WREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev, i, regSDMA0_FREEZE), tmp); + tmp = RREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev, i, regSDMA0_F32_CNTL)); + tmp |= SDMA0_F32_CNTL__HALT_MASK; + tmp |= SDMA0_F32_CNTL__TH1_RESET_MASK; + WREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev, i, regSDMA0_F32_CNTL), tmp); + + WREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev, i, regSDMA0_QUEUE0_PREEMPT), 0); + + udelay(100); + + tmp = GRBM_SOFT_RESET__SOFT_RESET_SDMA0_MASK << i; + WREG32_SOC15(GC, 0, regGRBM_SOFT_RESET, tmp); + tmp = RREG32_SOC15(GC, 0, regGRBM_SOFT_RESET); + + udelay(100); + + WREG32_SOC15(GC, 0, regGRBM_SOFT_RESET, 0); + tmp = RREG32_SOC15(GC, 0, regGRBM_SOFT_RESET); + + udelay(100); + } + + return sdma_v6_0_start(adev); +} + +static bool sdma_v6_0_check_soft_reset(void *handle) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + struct amdgpu_ring *ring; + int i, r; + long tmo = msecs_to_jiffies(1000); + + for (i = 0; i < adev->sdma.num_instances; i++) { + ring = &adev->sdma.instance[i].ring; + r = amdgpu_ring_test_ib(ring, tmo); + if (r) + return true; + } + + return false; +} + +/** + * sdma_v6_0_start - setup and start the async dma engines + * + * @adev: amdgpu_device pointer + * + * Set up the DMA engines and enable them. + * Returns 0 for success, error for failure. + */ +static int sdma_v6_0_start(struct amdgpu_device *adev) +{ + int r = 0; + + if (amdgpu_sriov_vf(adev)) { + sdma_v6_0_ctx_switch_enable(adev, false); + sdma_v6_0_enable(adev, false); + + /* set RB registers */ + r = sdma_v6_0_gfx_resume(adev); + return r; + } + + if (adev->firmware.load_type == AMDGPU_FW_LOAD_DIRECT) { + r = sdma_v6_0_load_microcode(adev); + if (r) + return r; + + /* The value of regSDMA_F32_CNTL is invalid the moment after loading fw */ + if (amdgpu_emu_mode == 1) + msleep(1000); + } + + /* unhalt the MEs */ + sdma_v6_0_enable(adev, true); + /* enable sdma ring preemption */ + sdma_v6_0_ctx_switch_enable(adev, true); + + /* start the gfx rings and rlc compute queues */ + r = sdma_v6_0_gfx_resume(adev); + if (r) + return r; + r = sdma_v6_0_rlc_resume(adev); + + return r; +} + +static int sdma_v6_0_mqd_init(struct amdgpu_device *adev, void *mqd, + struct amdgpu_mqd_prop *prop) +{ + struct v11_sdma_mqd *m = mqd; + uint64_t wb_gpu_addr; + + m->sdmax_rlcx_rb_cntl = + order_base_2(prop->queue_size / 4) << SDMA0_QUEUE0_RB_CNTL__RB_SIZE__SHIFT | + 1 << SDMA0_QUEUE0_RB_CNTL__RPTR_WRITEBACK_ENABLE__SHIFT | + 4 << SDMA0_QUEUE0_RB_CNTL__RPTR_WRITEBACK_TIMER__SHIFT | + 1 << SDMA0_QUEUE0_RB_CNTL__F32_WPTR_POLL_ENABLE__SHIFT; + + m->sdmax_rlcx_rb_base = lower_32_bits(prop->hqd_base_gpu_addr >> 8); + m->sdmax_rlcx_rb_base_hi = upper_32_bits(prop->hqd_base_gpu_addr >> 8); + + wb_gpu_addr = prop->wptr_gpu_addr; + m->sdmax_rlcx_rb_wptr_poll_addr_lo = lower_32_bits(wb_gpu_addr); + m->sdmax_rlcx_rb_wptr_poll_addr_hi = upper_32_bits(wb_gpu_addr); + + wb_gpu_addr = prop->rptr_gpu_addr; + m->sdmax_rlcx_rb_rptr_addr_lo = lower_32_bits(wb_gpu_addr); + m->sdmax_rlcx_rb_rptr_addr_hi = upper_32_bits(wb_gpu_addr); + + m->sdmax_rlcx_ib_cntl = RREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev, 0, + regSDMA0_QUEUE0_IB_CNTL)); + + m->sdmax_rlcx_doorbell_offset = + prop->doorbell_index << SDMA0_QUEUE0_DOORBELL_OFFSET__OFFSET__SHIFT; + + m->sdmax_rlcx_doorbell = REG_SET_FIELD(0, SDMA0_QUEUE0_DOORBELL, ENABLE, 1); + + m->sdmax_rlcx_skip_cntl = 0; + m->sdmax_rlcx_context_status = 0; + m->sdmax_rlcx_doorbell_log = 0; + + m->sdmax_rlcx_rb_aql_cntl = regSDMA0_QUEUE0_RB_AQL_CNTL_DEFAULT; + m->sdmax_rlcx_dummy_reg = regSDMA0_QUEUE0_DUMMY_REG_DEFAULT; + + return 0; +} + +static void sdma_v6_0_set_mqd_funcs(struct amdgpu_device *adev) +{ + adev->mqds[AMDGPU_HW_IP_DMA].mqd_size = sizeof(struct v11_sdma_mqd); + adev->mqds[AMDGPU_HW_IP_DMA].init_mqd = sdma_v6_0_mqd_init; +} + +/** + * sdma_v6_0_ring_test_ring - simple async dma engine test + * + * @ring: amdgpu_ring structure holding ring information + * + * Test the DMA engine by writing using it to write an + * value to memory. + * Returns 0 for success, error for failure. + */ +static int sdma_v6_0_ring_test_ring(struct amdgpu_ring *ring) +{ + struct amdgpu_device *adev = ring->adev; + unsigned i; + unsigned index; + int r; + u32 tmp; + u64 gpu_addr; + volatile uint32_t *cpu_ptr = NULL; + + tmp = 0xCAFEDEAD; + + if (ring->is_mes_queue) { + uint32_t offset = 0; + offset = amdgpu_mes_ctx_get_offs(ring, + AMDGPU_MES_CTX_PADDING_OFFS); + gpu_addr = amdgpu_mes_ctx_get_offs_gpu_addr(ring, offset); + cpu_ptr = amdgpu_mes_ctx_get_offs_cpu_addr(ring, offset); + *cpu_ptr = tmp; + } else { + r = amdgpu_device_wb_get(adev, &index); + if (r) { + dev_err(adev->dev, "(%d) failed to allocate wb slot\n", r); + return r; + } + + gpu_addr = adev->wb.gpu_addr + (index * 4); + adev->wb.wb[index] = cpu_to_le32(tmp); + } + + r = amdgpu_ring_alloc(ring, 5); + if (r) { + DRM_ERROR("amdgpu: dma failed to lock ring %d (%d).\n", ring->idx, r); + amdgpu_device_wb_free(adev, index); + return r; + } + + amdgpu_ring_write(ring, SDMA_PKT_COPY_LINEAR_HEADER_OP(SDMA_OP_WRITE) | + SDMA_PKT_COPY_LINEAR_HEADER_SUB_OP(SDMA_SUBOP_WRITE_LINEAR)); + amdgpu_ring_write(ring, lower_32_bits(gpu_addr)); + amdgpu_ring_write(ring, upper_32_bits(gpu_addr)); + amdgpu_ring_write(ring, SDMA_PKT_WRITE_UNTILED_DW_3_COUNT(0)); + amdgpu_ring_write(ring, 0xDEADBEEF); + amdgpu_ring_commit(ring); + + for (i = 0; i < adev->usec_timeout; i++) { + if (ring->is_mes_queue) + tmp = le32_to_cpu(*cpu_ptr); + else + tmp = le32_to_cpu(adev->wb.wb[index]); + if (tmp == 0xDEADBEEF) + break; + if (amdgpu_emu_mode == 1) + msleep(1); + else + udelay(1); + } + + if (i >= adev->usec_timeout) + r = -ETIMEDOUT; + + if (!ring->is_mes_queue) + amdgpu_device_wb_free(adev, index); + + return r; +} + +/** + * sdma_v6_0_ring_test_ib - test an IB on the DMA engine + * + * @ring: amdgpu_ring structure holding ring information + * + * Test a simple IB in the DMA ring. + * Returns 0 on success, error on failure. + */ +static int sdma_v6_0_ring_test_ib(struct amdgpu_ring *ring, long timeout) +{ + struct amdgpu_device *adev = ring->adev; + struct amdgpu_ib ib; + struct dma_fence *f = NULL; + unsigned index; + long r; + u32 tmp = 0; + u64 gpu_addr; + volatile uint32_t *cpu_ptr = NULL; + + tmp = 0xCAFEDEAD; + memset(&ib, 0, sizeof(ib)); + + if (ring->is_mes_queue) { + uint32_t offset = 0; + offset = amdgpu_mes_ctx_get_offs(ring, AMDGPU_MES_CTX_IB_OFFS); + ib.gpu_addr = amdgpu_mes_ctx_get_offs_gpu_addr(ring, offset); + ib.ptr = (void *)amdgpu_mes_ctx_get_offs_cpu_addr(ring, offset); + + offset = amdgpu_mes_ctx_get_offs(ring, + AMDGPU_MES_CTX_PADDING_OFFS); + gpu_addr = amdgpu_mes_ctx_get_offs_gpu_addr(ring, offset); + cpu_ptr = amdgpu_mes_ctx_get_offs_cpu_addr(ring, offset); + *cpu_ptr = tmp; + } else { + r = amdgpu_device_wb_get(adev, &index); + if (r) { + dev_err(adev->dev, "(%ld) failed to allocate wb slot\n", r); + return r; + } + + gpu_addr = adev->wb.gpu_addr + (index * 4); + adev->wb.wb[index] = cpu_to_le32(tmp); + + r = amdgpu_ib_get(adev, NULL, 256, AMDGPU_IB_POOL_DIRECT, &ib); + if (r) { + DRM_ERROR("amdgpu: failed to get ib (%ld).\n", r); + goto err0; + } + } + + ib.ptr[0] = SDMA_PKT_COPY_LINEAR_HEADER_OP(SDMA_OP_WRITE) | + SDMA_PKT_COPY_LINEAR_HEADER_SUB_OP(SDMA_SUBOP_WRITE_LINEAR); + ib.ptr[1] = lower_32_bits(gpu_addr); + ib.ptr[2] = upper_32_bits(gpu_addr); + ib.ptr[3] = SDMA_PKT_WRITE_UNTILED_DW_3_COUNT(0); + ib.ptr[4] = 0xDEADBEEF; + ib.ptr[5] = SDMA_PKT_NOP_HEADER_OP(SDMA_OP_NOP); + ib.ptr[6] = SDMA_PKT_NOP_HEADER_OP(SDMA_OP_NOP); + ib.ptr[7] = SDMA_PKT_NOP_HEADER_OP(SDMA_OP_NOP); + ib.length_dw = 8; + + r = amdgpu_ib_schedule(ring, 1, &ib, NULL, &f); + if (r) + goto err1; + + r = dma_fence_wait_timeout(f, false, timeout); + if (r == 0) { + DRM_ERROR("amdgpu: IB test timed out\n"); + r = -ETIMEDOUT; + goto err1; + } else if (r < 0) { + DRM_ERROR("amdgpu: fence wait failed (%ld).\n", r); + goto err1; + } + + if (ring->is_mes_queue) + tmp = le32_to_cpu(*cpu_ptr); + else + tmp = le32_to_cpu(adev->wb.wb[index]); + + if (tmp == 0xDEADBEEF) + r = 0; + else + r = -EINVAL; + +err1: + amdgpu_ib_free(adev, &ib, NULL); + dma_fence_put(f); +err0: + if (!ring->is_mes_queue) + amdgpu_device_wb_free(adev, index); + return r; +} + + +/** + * sdma_v6_0_vm_copy_pte - update PTEs by copying them from the GART + * + * @ib: indirect buffer to fill with commands + * @pe: addr of the page entry + * @src: src addr to copy from + * @count: number of page entries to update + * + * Update PTEs by copying them from the GART using sDMA. + */ +static void sdma_v6_0_vm_copy_pte(struct amdgpu_ib *ib, + uint64_t pe, uint64_t src, + unsigned count) +{ + unsigned bytes = count * 8; + + ib->ptr[ib->length_dw++] = SDMA_PKT_COPY_LINEAR_HEADER_OP(SDMA_OP_COPY) | + SDMA_PKT_COPY_LINEAR_HEADER_SUB_OP(SDMA_SUBOP_COPY_LINEAR); + ib->ptr[ib->length_dw++] = bytes - 1; + ib->ptr[ib->length_dw++] = 0; /* src/dst endian swap */ + ib->ptr[ib->length_dw++] = lower_32_bits(src); + ib->ptr[ib->length_dw++] = upper_32_bits(src); + ib->ptr[ib->length_dw++] = lower_32_bits(pe); + ib->ptr[ib->length_dw++] = upper_32_bits(pe); + +} + +/** + * sdma_v6_0_vm_write_pte - update PTEs by writing them manually + * + * @ib: indirect buffer to fill with commands + * @pe: addr of the page entry + * @addr: dst addr to write into pe + * @count: number of page entries to update + * @incr: increase next addr by incr bytes + * @flags: access flags + * + * Update PTEs by writing them manually using sDMA. + */ +static void sdma_v6_0_vm_write_pte(struct amdgpu_ib *ib, uint64_t pe, + uint64_t value, unsigned count, + uint32_t incr) +{ + unsigned ndw = count * 2; + + ib->ptr[ib->length_dw++] = SDMA_PKT_COPY_LINEAR_HEADER_OP(SDMA_OP_WRITE) | + SDMA_PKT_COPY_LINEAR_HEADER_SUB_OP(SDMA_SUBOP_WRITE_LINEAR); + ib->ptr[ib->length_dw++] = lower_32_bits(pe); + ib->ptr[ib->length_dw++] = upper_32_bits(pe); + ib->ptr[ib->length_dw++] = ndw - 1; + for (; ndw > 0; ndw -= 2) { + ib->ptr[ib->length_dw++] = lower_32_bits(value); + ib->ptr[ib->length_dw++] = upper_32_bits(value); + value += incr; + } +} + +/** + * sdma_v6_0_vm_set_pte_pde - update the page tables using sDMA + * + * @ib: indirect buffer to fill with commands + * @pe: addr of the page entry + * @addr: dst addr to write into pe + * @count: number of page entries to update + * @incr: increase next addr by incr bytes + * @flags: access flags + * + * Update the page tables using sDMA. + */ +static void sdma_v6_0_vm_set_pte_pde(struct amdgpu_ib *ib, + uint64_t pe, + uint64_t addr, unsigned count, + uint32_t incr, uint64_t flags) +{ + /* for physically contiguous pages (vram) */ + ib->ptr[ib->length_dw++] = SDMA_PKT_COPY_LINEAR_HEADER_OP(SDMA_OP_PTEPDE); + ib->ptr[ib->length_dw++] = lower_32_bits(pe); /* dst addr */ + ib->ptr[ib->length_dw++] = upper_32_bits(pe); + ib->ptr[ib->length_dw++] = lower_32_bits(flags); /* mask */ + ib->ptr[ib->length_dw++] = upper_32_bits(flags); + ib->ptr[ib->length_dw++] = lower_32_bits(addr); /* value */ + ib->ptr[ib->length_dw++] = upper_32_bits(addr); + ib->ptr[ib->length_dw++] = incr; /* increment size */ + ib->ptr[ib->length_dw++] = 0; + ib->ptr[ib->length_dw++] = count - 1; /* number of entries */ +} + +/** + * sdma_v6_0_ring_pad_ib - pad the IB + * @ib: indirect buffer to fill with padding + * + * Pad the IB with NOPs to a boundary multiple of 8. + */ +static void sdma_v6_0_ring_pad_ib(struct amdgpu_ring *ring, struct amdgpu_ib *ib) +{ + struct amdgpu_sdma_instance *sdma = amdgpu_sdma_get_instance_from_ring(ring); + u32 pad_count; + int i; + + pad_count = (-ib->length_dw) & 0x7; + for (i = 0; i < pad_count; i++) + if (sdma && sdma->burst_nop && (i == 0)) + ib->ptr[ib->length_dw++] = + SDMA_PKT_COPY_LINEAR_HEADER_OP(SDMA_OP_NOP) | + SDMA_PKT_NOP_HEADER_COUNT(pad_count - 1); + else + ib->ptr[ib->length_dw++] = + SDMA_PKT_COPY_LINEAR_HEADER_OP(SDMA_OP_NOP); +} + +/** + * sdma_v6_0_ring_emit_pipeline_sync - sync the pipeline + * + * @ring: amdgpu_ring pointer + * + * Make sure all previous operations are completed (CIK). + */ +static void sdma_v6_0_ring_emit_pipeline_sync(struct amdgpu_ring *ring) +{ + uint32_t seq = ring->fence_drv.sync_seq; + uint64_t addr = ring->fence_drv.gpu_addr; + + /* wait for idle */ + amdgpu_ring_write(ring, SDMA_PKT_COPY_LINEAR_HEADER_OP(SDMA_OP_POLL_REGMEM) | + SDMA_PKT_POLL_REGMEM_HEADER_HDP_FLUSH(0) | + SDMA_PKT_POLL_REGMEM_HEADER_FUNC(3) | /* equal */ + SDMA_PKT_POLL_REGMEM_HEADER_MEM_POLL(1)); + amdgpu_ring_write(ring, addr & 0xfffffffc); + amdgpu_ring_write(ring, upper_32_bits(addr) & 0xffffffff); + amdgpu_ring_write(ring, seq); /* reference */ + amdgpu_ring_write(ring, 0xffffffff); /* mask */ + amdgpu_ring_write(ring, SDMA_PKT_POLL_REGMEM_DW5_RETRY_COUNT(0xfff) | + SDMA_PKT_POLL_REGMEM_DW5_INTERVAL(4)); /* retry count, poll interval */ +} + +/** + * sdma_v6_0_ring_emit_vm_flush - vm flush using sDMA + * + * @ring: amdgpu_ring pointer + * @vm: amdgpu_vm pointer + * + * Update the page table base and flush the VM TLB + * using sDMA. + */ +static void sdma_v6_0_ring_emit_vm_flush(struct amdgpu_ring *ring, + unsigned vmid, uint64_t pd_addr) +{ + amdgpu_gmc_emit_flush_gpu_tlb(ring, vmid, pd_addr); +} + +static void sdma_v6_0_ring_emit_wreg(struct amdgpu_ring *ring, + uint32_t reg, uint32_t val) +{ + amdgpu_ring_write(ring, SDMA_PKT_COPY_LINEAR_HEADER_OP(SDMA_OP_SRBM_WRITE) | + SDMA_PKT_SRBM_WRITE_HEADER_BYTE_EN(0xf)); + amdgpu_ring_write(ring, reg); + amdgpu_ring_write(ring, val); +} + +static void sdma_v6_0_ring_emit_reg_wait(struct amdgpu_ring *ring, uint32_t reg, + uint32_t val, uint32_t mask) +{ + amdgpu_ring_write(ring, SDMA_PKT_COPY_LINEAR_HEADER_OP(SDMA_OP_POLL_REGMEM) | + SDMA_PKT_POLL_REGMEM_HEADER_HDP_FLUSH(0) | + SDMA_PKT_POLL_REGMEM_HEADER_FUNC(3)); /* equal */ + amdgpu_ring_write(ring, reg << 2); + amdgpu_ring_write(ring, 0); + amdgpu_ring_write(ring, val); /* reference */ + amdgpu_ring_write(ring, mask); /* mask */ + amdgpu_ring_write(ring, SDMA_PKT_POLL_REGMEM_DW5_RETRY_COUNT(0xfff) | + SDMA_PKT_POLL_REGMEM_DW5_INTERVAL(10)); +} + +static void sdma_v6_0_ring_emit_reg_write_reg_wait(struct amdgpu_ring *ring, + uint32_t reg0, uint32_t reg1, + uint32_t ref, uint32_t mask) +{ + amdgpu_ring_emit_wreg(ring, reg0, ref); + /* wait for a cycle to reset vm_inv_eng*_ack */ + amdgpu_ring_emit_reg_wait(ring, reg0, 0, 0); + amdgpu_ring_emit_reg_wait(ring, reg1, mask, mask); +} + +static int sdma_v6_0_early_init(void *handle) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + + sdma_v6_0_set_ring_funcs(adev); + sdma_v6_0_set_buffer_funcs(adev); + sdma_v6_0_set_vm_pte_funcs(adev); + sdma_v6_0_set_irq_funcs(adev); + sdma_v6_0_set_mqd_funcs(adev); + + return 0; +} + +static int sdma_v6_0_sw_init(void *handle) +{ + struct amdgpu_ring *ring; + int r, i; + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + + /* SDMA trap event */ + r = amdgpu_irq_add_id(adev, SOC21_IH_CLIENTID_GFX, + GFX_11_0_0__SRCID__SDMA_TRAP, + &adev->sdma.trap_irq); + if (r) + return r; + + r = sdma_v6_0_init_microcode(adev); + if (r) { + DRM_ERROR("Failed to load sdma firmware!\n"); + return r; + } + + for (i = 0; i < adev->sdma.num_instances; i++) { + ring = &adev->sdma.instance[i].ring; + ring->ring_obj = NULL; + ring->use_doorbell = true; + ring->me = i; + + DRM_DEBUG("SDMA %d use_doorbell being set to: [%s]\n", i, + ring->use_doorbell?"true":"false"); + + ring->doorbell_index = + (adev->doorbell_index.sdma_engine[i] << 1); // get DWORD offset + + sprintf(ring->name, "sdma%d", i); + r = amdgpu_ring_init(adev, ring, 1024, + &adev->sdma.trap_irq, + AMDGPU_SDMA_IRQ_INSTANCE0 + i, + AMDGPU_RING_PRIO_DEFAULT, NULL); + if (r) + return r; + } + + return r; +} + +static int sdma_v6_0_sw_fini(void *handle) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + int i; + + for (i = 0; i < adev->sdma.num_instances; i++) + amdgpu_ring_fini(&adev->sdma.instance[i].ring); + + amdgpu_sdma_destroy_inst_ctx(adev, true); + + return 0; +} + +static int sdma_v6_0_hw_init(void *handle) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + + return sdma_v6_0_start(adev); +} + +static int sdma_v6_0_hw_fini(void *handle) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + + if (amdgpu_sriov_vf(adev)) { + /* disable the scheduler for SDMA */ + amdgpu_sdma_unset_buffer_funcs_helper(adev); + return 0; + } + + sdma_v6_0_ctx_switch_enable(adev, false); + sdma_v6_0_enable(adev, false); + + return 0; +} + +static int sdma_v6_0_suspend(void *handle) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + + return sdma_v6_0_hw_fini(adev); +} + +static int sdma_v6_0_resume(void *handle) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + + return sdma_v6_0_hw_init(adev); +} + +static bool sdma_v6_0_is_idle(void *handle) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + u32 i; + + for (i = 0; i < adev->sdma.num_instances; i++) { + u32 tmp = RREG32(sdma_v6_0_get_reg_offset(adev, i, regSDMA0_STATUS_REG)); + + if (!(tmp & SDMA0_STATUS_REG__IDLE_MASK)) + return false; + } + + return true; +} + +static int sdma_v6_0_wait_for_idle(void *handle) +{ + unsigned i; + u32 sdma0, sdma1; + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + + for (i = 0; i < adev->usec_timeout; i++) { + sdma0 = RREG32(sdma_v6_0_get_reg_offset(adev, 0, regSDMA0_STATUS_REG)); + sdma1 = RREG32(sdma_v6_0_get_reg_offset(adev, 1, regSDMA0_STATUS_REG)); + + if (sdma0 & sdma1 & SDMA0_STATUS_REG__IDLE_MASK) + return 0; + udelay(1); + } + return -ETIMEDOUT; +} + +static int sdma_v6_0_ring_preempt_ib(struct amdgpu_ring *ring) +{ + int i, r = 0; + struct amdgpu_device *adev = ring->adev; + u32 index = 0; + u64 sdma_gfx_preempt; + + amdgpu_sdma_get_index_from_ring(ring, &index); + sdma_gfx_preempt = + sdma_v6_0_get_reg_offset(adev, index, regSDMA0_QUEUE0_PREEMPT); + + /* assert preemption condition */ + amdgpu_ring_set_preempt_cond_exec(ring, false); + + /* emit the trailing fence */ + ring->trail_seq += 1; + amdgpu_ring_alloc(ring, 10); + sdma_v6_0_ring_emit_fence(ring, ring->trail_fence_gpu_addr, + ring->trail_seq, 0); + amdgpu_ring_commit(ring); + + /* assert IB preemption */ + WREG32(sdma_gfx_preempt, 1); + + /* poll the trailing fence */ + for (i = 0; i < adev->usec_timeout; i++) { + if (ring->trail_seq == + le32_to_cpu(*(ring->trail_fence_cpu_addr))) + break; + udelay(1); + } + + if (i >= adev->usec_timeout) { + r = -EINVAL; + DRM_ERROR("ring %d failed to be preempted\n", ring->idx); + } + + /* deassert IB preemption */ + WREG32(sdma_gfx_preempt, 0); + + /* deassert the preemption condition */ + amdgpu_ring_set_preempt_cond_exec(ring, true); + return r; +} + +static int sdma_v6_0_set_trap_irq_state(struct amdgpu_device *adev, + struct amdgpu_irq_src *source, + unsigned type, + enum amdgpu_interrupt_state state) +{ + u32 sdma_cntl; + + u32 reg_offset = sdma_v6_0_get_reg_offset(adev, type, regSDMA0_CNTL); + + sdma_cntl = RREG32(reg_offset); + sdma_cntl = REG_SET_FIELD(sdma_cntl, SDMA0_CNTL, TRAP_ENABLE, + state == AMDGPU_IRQ_STATE_ENABLE ? 1 : 0); + WREG32(reg_offset, sdma_cntl); + + return 0; +} + +static int sdma_v6_0_process_trap_irq(struct amdgpu_device *adev, + struct amdgpu_irq_src *source, + struct amdgpu_iv_entry *entry) +{ + int instances, queue; + uint32_t mes_queue_id = entry->src_data[0]; + + DRM_DEBUG("IH: SDMA trap\n"); + + if (adev->enable_mes && (mes_queue_id & AMDGPU_FENCE_MES_QUEUE_FLAG)) { + struct amdgpu_mes_queue *queue; + + mes_queue_id &= AMDGPU_FENCE_MES_QUEUE_ID_MASK; + + spin_lock(&adev->mes.queue_id_lock); + queue = idr_find(&adev->mes.queue_id_idr, mes_queue_id); + if (queue) { + DRM_DEBUG("process smda queue id = %d\n", mes_queue_id); + amdgpu_fence_process(queue->ring); + } + spin_unlock(&adev->mes.queue_id_lock); + return 0; + } + + queue = entry->ring_id & 0xf; + instances = (entry->ring_id & 0xf0) >> 4; + if (instances > 1) { + DRM_ERROR("IH: wrong ring_ID detected, as wrong sdma instance\n"); + return -EINVAL; + } + + switch (entry->client_id) { + case SOC21_IH_CLIENTID_GFX: + switch (queue) { + case 0: + amdgpu_fence_process(&adev->sdma.instance[instances].ring); + break; + default: + break; + } + break; + } + return 0; +} + +static int sdma_v6_0_process_illegal_inst_irq(struct amdgpu_device *adev, + struct amdgpu_irq_src *source, + struct amdgpu_iv_entry *entry) +{ + return 0; +} + +static int sdma_v6_0_set_clockgating_state(void *handle, + enum amd_clockgating_state state) +{ + return 0; +} + +static int sdma_v6_0_set_powergating_state(void *handle, + enum amd_powergating_state state) +{ + return 0; +} + +static void sdma_v6_0_get_clockgating_state(void *handle, u64 *flags) +{ +} + +const struct amd_ip_funcs sdma_v6_0_ip_funcs = { + .name = "sdma_v6_0", + .early_init = sdma_v6_0_early_init, + .late_init = NULL, + .sw_init = sdma_v6_0_sw_init, + .sw_fini = sdma_v6_0_sw_fini, + .hw_init = sdma_v6_0_hw_init, + .hw_fini = sdma_v6_0_hw_fini, + .suspend = sdma_v6_0_suspend, + .resume = sdma_v6_0_resume, + .is_idle = sdma_v6_0_is_idle, + .wait_for_idle = sdma_v6_0_wait_for_idle, + .soft_reset = sdma_v6_0_soft_reset, + .check_soft_reset = sdma_v6_0_check_soft_reset, + .set_clockgating_state = sdma_v6_0_set_clockgating_state, + .set_powergating_state = sdma_v6_0_set_powergating_state, + .get_clockgating_state = sdma_v6_0_get_clockgating_state, +}; + +static const struct amdgpu_ring_funcs sdma_v6_0_ring_funcs = { + .type = AMDGPU_RING_TYPE_SDMA, + .align_mask = 0xf, + .nop = SDMA_PKT_NOP_HEADER_OP(SDMA_OP_NOP), + .support_64bit_ptrs = true, + .vmhub = AMDGPU_GFXHUB_0, + .get_rptr = sdma_v6_0_ring_get_rptr, + .get_wptr = sdma_v6_0_ring_get_wptr, + .set_wptr = sdma_v6_0_ring_set_wptr, + .emit_frame_size = + 5 + /* sdma_v6_0_ring_init_cond_exec */ + 6 + /* sdma_v6_0_ring_emit_hdp_flush */ + 6 + /* sdma_v6_0_ring_emit_pipeline_sync */ + /* sdma_v6_0_ring_emit_vm_flush */ + SOC15_FLUSH_GPU_TLB_NUM_WREG * 3 + + SOC15_FLUSH_GPU_TLB_NUM_REG_WAIT * 6 + + 10 + 10 + 10, /* sdma_v6_0_ring_emit_fence x3 for user fence, vm fence */ + .emit_ib_size = 5 + 7 + 6, /* sdma_v6_0_ring_emit_ib */ + .emit_ib = sdma_v6_0_ring_emit_ib, + .emit_mem_sync = sdma_v6_0_ring_emit_mem_sync, + .emit_fence = sdma_v6_0_ring_emit_fence, + .emit_pipeline_sync = sdma_v6_0_ring_emit_pipeline_sync, + .emit_vm_flush = sdma_v6_0_ring_emit_vm_flush, + .emit_hdp_flush = sdma_v6_0_ring_emit_hdp_flush, + .test_ring = sdma_v6_0_ring_test_ring, + .test_ib = sdma_v6_0_ring_test_ib, + .insert_nop = sdma_v6_0_ring_insert_nop, + .pad_ib = sdma_v6_0_ring_pad_ib, + .emit_wreg = sdma_v6_0_ring_emit_wreg, + .emit_reg_wait = sdma_v6_0_ring_emit_reg_wait, + .emit_reg_write_reg_wait = sdma_v6_0_ring_emit_reg_write_reg_wait, + .init_cond_exec = sdma_v6_0_ring_init_cond_exec, + .patch_cond_exec = sdma_v6_0_ring_patch_cond_exec, + .preempt_ib = sdma_v6_0_ring_preempt_ib, +}; + +static void sdma_v6_0_set_ring_funcs(struct amdgpu_device *adev) +{ + int i; + + for (i = 0; i < adev->sdma.num_instances; i++) { + adev->sdma.instance[i].ring.funcs = &sdma_v6_0_ring_funcs; + adev->sdma.instance[i].ring.me = i; + } +} + +static const struct amdgpu_irq_src_funcs sdma_v6_0_trap_irq_funcs = { + .set = sdma_v6_0_set_trap_irq_state, + .process = sdma_v6_0_process_trap_irq, +}; + +static const struct amdgpu_irq_src_funcs sdma_v6_0_illegal_inst_irq_funcs = { + .process = sdma_v6_0_process_illegal_inst_irq, +}; + +static void sdma_v6_0_set_irq_funcs(struct amdgpu_device *adev) +{ + adev->sdma.trap_irq.num_types = AMDGPU_SDMA_IRQ_INSTANCE0 + + adev->sdma.num_instances; + adev->sdma.trap_irq.funcs = &sdma_v6_0_trap_irq_funcs; + adev->sdma.illegal_inst_irq.funcs = &sdma_v6_0_illegal_inst_irq_funcs; +} + +/** + * sdma_v6_0_emit_copy_buffer - copy buffer using the sDMA engine + * + * @ring: amdgpu_ring structure holding ring information + * @src_offset: src GPU address + * @dst_offset: dst GPU address + * @byte_count: number of bytes to xfer + * + * Copy GPU buffers using the DMA engine. + * Used by the amdgpu ttm implementation to move pages if + * registered as the asic copy callback. + */ +static void sdma_v6_0_emit_copy_buffer(struct amdgpu_ib *ib, + uint64_t src_offset, + uint64_t dst_offset, + uint32_t byte_count, + bool tmz) +{ + ib->ptr[ib->length_dw++] = SDMA_PKT_COPY_LINEAR_HEADER_OP(SDMA_OP_COPY) | + SDMA_PKT_COPY_LINEAR_HEADER_SUB_OP(SDMA_SUBOP_COPY_LINEAR) | + SDMA_PKT_COPY_LINEAR_HEADER_TMZ(tmz ? 1 : 0); + ib->ptr[ib->length_dw++] = byte_count - 1; + ib->ptr[ib->length_dw++] = 0; /* src/dst endian swap */ + ib->ptr[ib->length_dw++] = lower_32_bits(src_offset); + ib->ptr[ib->length_dw++] = upper_32_bits(src_offset); + ib->ptr[ib->length_dw++] = lower_32_bits(dst_offset); + ib->ptr[ib->length_dw++] = upper_32_bits(dst_offset); +} + +/** + * sdma_v6_0_emit_fill_buffer - fill buffer using the sDMA engine + * + * @ring: amdgpu_ring structure holding ring information + * @src_data: value to write to buffer + * @dst_offset: dst GPU address + * @byte_count: number of bytes to xfer + * + * Fill GPU buffers using the DMA engine. + */ +static void sdma_v6_0_emit_fill_buffer(struct amdgpu_ib *ib, + uint32_t src_data, + uint64_t dst_offset, + uint32_t byte_count) +{ + ib->ptr[ib->length_dw++] = SDMA_PKT_COPY_LINEAR_HEADER_OP(SDMA_OP_CONST_FILL); + ib->ptr[ib->length_dw++] = lower_32_bits(dst_offset); + ib->ptr[ib->length_dw++] = upper_32_bits(dst_offset); + ib->ptr[ib->length_dw++] = src_data; + ib->ptr[ib->length_dw++] = byte_count - 1; +} + +static const struct amdgpu_buffer_funcs sdma_v6_0_buffer_funcs = { + .copy_max_bytes = 0x400000, + .copy_num_dw = 7, + .emit_copy_buffer = sdma_v6_0_emit_copy_buffer, + + .fill_max_bytes = 0x400000, + .fill_num_dw = 5, + .emit_fill_buffer = sdma_v6_0_emit_fill_buffer, +}; + +static void sdma_v6_0_set_buffer_funcs(struct amdgpu_device *adev) +{ + adev->mman.buffer_funcs = &sdma_v6_0_buffer_funcs; + adev->mman.buffer_funcs_ring = &adev->sdma.instance[0].ring; +} + +static const struct amdgpu_vm_pte_funcs sdma_v6_0_vm_pte_funcs = { + .copy_pte_num_dw = 7, + .copy_pte = sdma_v6_0_vm_copy_pte, + .write_pte = sdma_v6_0_vm_write_pte, + .set_pte_pde = sdma_v6_0_vm_set_pte_pde, +}; + +static void sdma_v6_0_set_vm_pte_funcs(struct amdgpu_device *adev) +{ + unsigned i; + + adev->vm_manager.vm_pte_funcs = &sdma_v6_0_vm_pte_funcs; + for (i = 0; i < adev->sdma.num_instances; i++) { + adev->vm_manager.vm_pte_scheds[i] = + &adev->sdma.instance[i].ring.sched; + } + adev->vm_manager.vm_pte_num_scheds = adev->sdma.num_instances; +} + +const struct amdgpu_ip_block_version sdma_v6_0_ip_block = { + .type = AMD_IP_BLOCK_TYPE_SDMA, + .major = 6, + .minor = 0, + .rev = 0, + .funcs = &sdma_v6_0_ip_funcs, +}; diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v6_0.h b/drivers/gpu/drm/amd/amdgpu/sdma_v6_0.h new file mode 100644 index 000000000000..e473ec7dfc8f --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/sdma_v6_0.h @@ -0,0 +1,30 @@ +/* + * Copyright 2020 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#ifndef __SDMA_V6_0_H__ +#define __SDMA_V6_0_H__ + +extern const struct amd_ip_funcs sdma_v6_0_ip_funcs; +extern const struct amdgpu_ip_block_version sdma_v6_0_ip_block; + +#endif /* __SDMA_V6_0_H__ */ diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v6_0_0_pkt_open.h b/drivers/gpu/drm/amd/amdgpu/sdma_v6_0_0_pkt_open.h new file mode 100644 index 000000000000..6af23e7888ca --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/sdma_v6_0_0_pkt_open.h @@ -0,0 +1,5664 @@ +/* + * Copyright 2021 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ +#ifndef __SDMA_V6_0_0_PKT_OPEN_H_ +#define __SDMA_V6_0_0_PKT_OPEN_H_ + +#define SDMA_OP_NOP 0 +#define SDMA_OP_COPY 1 +#define SDMA_OP_WRITE 2 +#define SDMA_OP_INDIRECT 4 +#define SDMA_OP_FENCE 5 +#define SDMA_OP_TRAP 6 +#define SDMA_OP_SEM 7 +#define SDMA_OP_POLL_REGMEM 8 +#define SDMA_OP_COND_EXE 9 +#define SDMA_OP_ATOMIC 10 +#define SDMA_OP_CONST_FILL 11 +#define SDMA_OP_PTEPDE 12 +#define SDMA_OP_TIMESTAMP 13 +#define SDMA_OP_SRBM_WRITE 14 +#define SDMA_OP_PRE_EXE 15 +#define SDMA_OP_GPUVM_INV 16 +#define SDMA_OP_GCR_REQ 17 +#define SDMA_OP_DUMMY_TRAP 32 +#define SDMA_SUBOP_TIMESTAMP_SET 0 +#define SDMA_SUBOP_TIMESTAMP_GET 1 +#define SDMA_SUBOP_TIMESTAMP_GET_GLOBAL 2 +#define SDMA_SUBOP_COPY_LINEAR 0 +#define SDMA_SUBOP_COPY_LINEAR_SUB_WIND 4 +#define SDMA_SUBOP_COPY_TILED 1 +#define SDMA_SUBOP_COPY_TILED_SUB_WIND 5 +#define SDMA_SUBOP_COPY_T2T_SUB_WIND 6 +#define SDMA_SUBOP_COPY_SOA 3 +#define SDMA_SUBOP_COPY_DIRTY_PAGE 7 +#define SDMA_SUBOP_COPY_LINEAR_PHY 8 +#define SDMA_SUBOP_COPY_LINEAR_SUB_WIND_LARGE 36 +#define SDMA_SUBOP_COPY_LINEAR_BC 16 +#define SDMA_SUBOP_COPY_TILED_BC 17 +#define SDMA_SUBOP_COPY_LINEAR_SUB_WIND_BC 20 +#define SDMA_SUBOP_COPY_TILED_SUB_WIND_BC 21 +#define SDMA_SUBOP_COPY_T2T_SUB_WIND_BC 22 +#define SDMA_SUBOP_WRITE_LINEAR 0 +#define SDMA_SUBOP_WRITE_TILED 1 +#define SDMA_SUBOP_WRITE_TILED_BC 17 +#define SDMA_SUBOP_PTEPDE_GEN 0 +#define SDMA_SUBOP_PTEPDE_COPY 1 +#define SDMA_SUBOP_PTEPDE_RMW 2 +#define SDMA_SUBOP_PTEPDE_COPY_BACKWARDS 3 +#define SDMA_SUBOP_MEM_INCR 1 +#define SDMA_SUBOP_DATA_FILL_MULTI 1 +#define SDMA_SUBOP_POLL_REG_WRITE_MEM 1 +#define SDMA_SUBOP_POLL_DBIT_WRITE_MEM 2 +#define SDMA_SUBOP_POLL_MEM_VERIFY 3 +#define SDMA_SUBOP_VM_INVALIDATION 4 +#define HEADER_AGENT_DISPATCH 4 +#define HEADER_BARRIER 5 +#define SDMA_OP_AQL_COPY 0 +#define SDMA_OP_AQL_BARRIER_OR 0 + +#define SDMA_GCR_RANGE_IS_PA (1 << 18) +#define SDMA_GCR_SEQ(x) (((x) & 0x3) << 16) +#define SDMA_GCR_GL2_WB (1 << 15) +#define SDMA_GCR_GL2_INV (1 << 14) +#define SDMA_GCR_GL2_DISCARD (1 << 13) +#define SDMA_GCR_GL2_RANGE(x) (((x) & 0x3) << 11) +#define SDMA_GCR_GL2_US (1 << 10) +#define SDMA_GCR_GL1_INV (1 << 9) +#define SDMA_GCR_GLV_INV (1 << 8) +#define SDMA_GCR_GLK_INV (1 << 7) +#define SDMA_GCR_GLK_WB (1 << 6) +#define SDMA_GCR_GLM_INV (1 << 5) +#define SDMA_GCR_GLM_WB (1 << 4) +#define SDMA_GCR_GL1_RANGE(x) (((x) & 0x3) << 2) +#define SDMA_GCR_GLI_INV(x) (((x) & 0x3) << 0) +/* +** Definitions for SDMA_PKT_COPY_LINEAR packet +*/ + +/*define for HEADER word*/ +/*define for op field*/ +#define SDMA_PKT_COPY_LINEAR_HEADER_op_offset 0 +#define SDMA_PKT_COPY_LINEAR_HEADER_op_mask 0x000000FF +#define SDMA_PKT_COPY_LINEAR_HEADER_op_shift 0 +#define SDMA_PKT_COPY_LINEAR_HEADER_OP(x) (((x) & SDMA_PKT_COPY_LINEAR_HEADER_op_mask) << SDMA_PKT_COPY_LINEAR_HEADER_op_shift) + +/*define for sub_op field*/ +#define SDMA_PKT_COPY_LINEAR_HEADER_sub_op_offset 0 +#define SDMA_PKT_COPY_LINEAR_HEADER_sub_op_mask 0x000000FF +#define SDMA_PKT_COPY_LINEAR_HEADER_sub_op_shift 8 +#define SDMA_PKT_COPY_LINEAR_HEADER_SUB_OP(x) (((x) & SDMA_PKT_COPY_LINEAR_HEADER_sub_op_mask) << SDMA_PKT_COPY_LINEAR_HEADER_sub_op_shift) + +/*define for encrypt field*/ +#define SDMA_PKT_COPY_LINEAR_HEADER_encrypt_offset 0 +#define SDMA_PKT_COPY_LINEAR_HEADER_encrypt_mask 0x00000001 +#define SDMA_PKT_COPY_LINEAR_HEADER_encrypt_shift 16 +#define SDMA_PKT_COPY_LINEAR_HEADER_ENCRYPT(x) (((x) & SDMA_PKT_COPY_LINEAR_HEADER_encrypt_mask) << SDMA_PKT_COPY_LINEAR_HEADER_encrypt_shift) + +/*define for tmz field*/ +#define SDMA_PKT_COPY_LINEAR_HEADER_tmz_offset 0 +#define SDMA_PKT_COPY_LINEAR_HEADER_tmz_mask 0x00000001 +#define SDMA_PKT_COPY_LINEAR_HEADER_tmz_shift 18 +#define SDMA_PKT_COPY_LINEAR_HEADER_TMZ(x) (((x) & SDMA_PKT_COPY_LINEAR_HEADER_tmz_mask) << SDMA_PKT_COPY_LINEAR_HEADER_tmz_shift) + +/*define for cpv field*/ +#define SDMA_PKT_COPY_LINEAR_HEADER_cpv_offset 0 +#define SDMA_PKT_COPY_LINEAR_HEADER_cpv_mask 0x00000001 +#define SDMA_PKT_COPY_LINEAR_HEADER_cpv_shift 19 +#define SDMA_PKT_COPY_LINEAR_HEADER_CPV(x) (((x) & SDMA_PKT_COPY_LINEAR_HEADER_cpv_mask) << SDMA_PKT_COPY_LINEAR_HEADER_cpv_shift) + +/*define for backwards field*/ +#define SDMA_PKT_COPY_LINEAR_HEADER_backwards_offset 0 +#define SDMA_PKT_COPY_LINEAR_HEADER_backwards_mask 0x00000001 +#define SDMA_PKT_COPY_LINEAR_HEADER_backwards_shift 25 +#define SDMA_PKT_COPY_LINEAR_HEADER_BACKWARDS(x) (((x) & SDMA_PKT_COPY_LINEAR_HEADER_backwards_mask) << SDMA_PKT_COPY_LINEAR_HEADER_backwards_shift) + +/*define for broadcast field*/ +#define SDMA_PKT_COPY_LINEAR_HEADER_broadcast_offset 0 +#define SDMA_PKT_COPY_LINEAR_HEADER_broadcast_mask 0x00000001 +#define SDMA_PKT_COPY_LINEAR_HEADER_broadcast_shift 27 +#define SDMA_PKT_COPY_LINEAR_HEADER_BROADCAST(x) (((x) & SDMA_PKT_COPY_LINEAR_HEADER_broadcast_mask) << SDMA_PKT_COPY_LINEAR_HEADER_broadcast_shift) + +/*define for COUNT word*/ +/*define for count field*/ +#define SDMA_PKT_COPY_LINEAR_COUNT_count_offset 1 +#define SDMA_PKT_COPY_LINEAR_COUNT_count_mask 0x3FFFFFFF +#define SDMA_PKT_COPY_LINEAR_COUNT_count_shift 0 +#define SDMA_PKT_COPY_LINEAR_COUNT_COUNT(x) (((x) & SDMA_PKT_COPY_LINEAR_COUNT_count_mask) << SDMA_PKT_COPY_LINEAR_COUNT_count_shift) + +/*define for PARAMETER word*/ +/*define for dst_sw field*/ +#define SDMA_PKT_COPY_LINEAR_PARAMETER_dst_sw_offset 2 +#define SDMA_PKT_COPY_LINEAR_PARAMETER_dst_sw_mask 0x00000003 +#define SDMA_PKT_COPY_LINEAR_PARAMETER_dst_sw_shift 16 +#define SDMA_PKT_COPY_LINEAR_PARAMETER_DST_SW(x) (((x) & SDMA_PKT_COPY_LINEAR_PARAMETER_dst_sw_mask) << SDMA_PKT_COPY_LINEAR_PARAMETER_dst_sw_shift) + +/*define for dst_cache_policy field*/ +#define SDMA_PKT_COPY_LINEAR_PARAMETER_dst_cache_policy_offset 2 +#define SDMA_PKT_COPY_LINEAR_PARAMETER_dst_cache_policy_mask 0x00000007 +#define SDMA_PKT_COPY_LINEAR_PARAMETER_dst_cache_policy_shift 18 +#define SDMA_PKT_COPY_LINEAR_PARAMETER_DST_CACHE_POLICY(x) (((x) & SDMA_PKT_COPY_LINEAR_PARAMETER_dst_cache_policy_mask) << SDMA_PKT_COPY_LINEAR_PARAMETER_dst_cache_policy_shift) + +/*define for src_sw field*/ +#define SDMA_PKT_COPY_LINEAR_PARAMETER_src_sw_offset 2 +#define SDMA_PKT_COPY_LINEAR_PARAMETER_src_sw_mask 0x00000003 +#define SDMA_PKT_COPY_LINEAR_PARAMETER_src_sw_shift 24 +#define SDMA_PKT_COPY_LINEAR_PARAMETER_SRC_SW(x) (((x) & SDMA_PKT_COPY_LINEAR_PARAMETER_src_sw_mask) << SDMA_PKT_COPY_LINEAR_PARAMETER_src_sw_shift) + +/*define for src_cache_policy field*/ +#define SDMA_PKT_COPY_LINEAR_PARAMETER_src_cache_policy_offset 2 +#define SDMA_PKT_COPY_LINEAR_PARAMETER_src_cache_policy_mask 0x00000007 +#define SDMA_PKT_COPY_LINEAR_PARAMETER_src_cache_policy_shift 26 +#define SDMA_PKT_COPY_LINEAR_PARAMETER_SRC_CACHE_POLICY(x) (((x) & SDMA_PKT_COPY_LINEAR_PARAMETER_src_cache_policy_mask) << SDMA_PKT_COPY_LINEAR_PARAMETER_src_cache_policy_shift) + +/*define for SRC_ADDR_LO word*/ +/*define for src_addr_31_0 field*/ +#define SDMA_PKT_COPY_LINEAR_SRC_ADDR_LO_src_addr_31_0_offset 3 +#define SDMA_PKT_COPY_LINEAR_SRC_ADDR_LO_src_addr_31_0_mask 0xFFFFFFFF +#define SDMA_PKT_COPY_LINEAR_SRC_ADDR_LO_src_addr_31_0_shift 0 +#define SDMA_PKT_COPY_LINEAR_SRC_ADDR_LO_SRC_ADDR_31_0(x) (((x) & SDMA_PKT_COPY_LINEAR_SRC_ADDR_LO_src_addr_31_0_mask) << SDMA_PKT_COPY_LINEAR_SRC_ADDR_LO_src_addr_31_0_shift) + +/*define for SRC_ADDR_HI word*/ +/*define for src_addr_63_32 field*/ +#define SDMA_PKT_COPY_LINEAR_SRC_ADDR_HI_src_addr_63_32_offset 4 +#define SDMA_PKT_COPY_LINEAR_SRC_ADDR_HI_src_addr_63_32_mask 0xFFFFFFFF +#define SDMA_PKT_COPY_LINEAR_SRC_ADDR_HI_src_addr_63_32_shift 0 +#define SDMA_PKT_COPY_LINEAR_SRC_ADDR_HI_SRC_ADDR_63_32(x) (((x) & SDMA_PKT_COPY_LINEAR_SRC_ADDR_HI_src_addr_63_32_mask) << SDMA_PKT_COPY_LINEAR_SRC_ADDR_HI_src_addr_63_32_shift) + +/*define for DST_ADDR_LO word*/ +/*define for dst_addr_31_0 field*/ +#define SDMA_PKT_COPY_LINEAR_DST_ADDR_LO_dst_addr_31_0_offset 5 +#define SDMA_PKT_COPY_LINEAR_DST_ADDR_LO_dst_addr_31_0_mask 0xFFFFFFFF +#define SDMA_PKT_COPY_LINEAR_DST_ADDR_LO_dst_addr_31_0_shift 0 +#define SDMA_PKT_COPY_LINEAR_DST_ADDR_LO_DST_ADDR_31_0(x) (((x) & SDMA_PKT_COPY_LINEAR_DST_ADDR_LO_dst_addr_31_0_mask) << SDMA_PKT_COPY_LINEAR_DST_ADDR_LO_dst_addr_31_0_shift) + +/*define for DST_ADDR_HI word*/ +/*define for dst_addr_63_32 field*/ +#define SDMA_PKT_COPY_LINEAR_DST_ADDR_HI_dst_addr_63_32_offset 6 +#define SDMA_PKT_COPY_LINEAR_DST_ADDR_HI_dst_addr_63_32_mask 0xFFFFFFFF +#define SDMA_PKT_COPY_LINEAR_DST_ADDR_HI_dst_addr_63_32_shift 0 +#define SDMA_PKT_COPY_LINEAR_DST_ADDR_HI_DST_ADDR_63_32(x) (((x) & SDMA_PKT_COPY_LINEAR_DST_ADDR_HI_dst_addr_63_32_mask) << SDMA_PKT_COPY_LINEAR_DST_ADDR_HI_dst_addr_63_32_shift) + + +/* +** Definitions for SDMA_PKT_COPY_LINEAR_BC packet +*/ + +/*define for HEADER word*/ +/*define for op field*/ +#define SDMA_PKT_COPY_LINEAR_BC_HEADER_op_offset 0 +#define SDMA_PKT_COPY_LINEAR_BC_HEADER_op_mask 0x000000FF +#define SDMA_PKT_COPY_LINEAR_BC_HEADER_op_shift 0 +#define SDMA_PKT_COPY_LINEAR_BC_HEADER_OP(x) (((x) & SDMA_PKT_COPY_LINEAR_BC_HEADER_op_mask) << SDMA_PKT_COPY_LINEAR_BC_HEADER_op_shift) + +/*define for sub_op field*/ +#define SDMA_PKT_COPY_LINEAR_BC_HEADER_sub_op_offset 0 +#define SDMA_PKT_COPY_LINEAR_BC_HEADER_sub_op_mask 0x000000FF +#define SDMA_PKT_COPY_LINEAR_BC_HEADER_sub_op_shift 8 +#define SDMA_PKT_COPY_LINEAR_BC_HEADER_SUB_OP(x) (((x) & SDMA_PKT_COPY_LINEAR_BC_HEADER_sub_op_mask) << SDMA_PKT_COPY_LINEAR_BC_HEADER_sub_op_shift) + +/*define for COUNT word*/ +/*define for count field*/ +#define SDMA_PKT_COPY_LINEAR_BC_COUNT_count_offset 1 +#define SDMA_PKT_COPY_LINEAR_BC_COUNT_count_mask 0x003FFFFF +#define SDMA_PKT_COPY_LINEAR_BC_COUNT_count_shift 0 +#define SDMA_PKT_COPY_LINEAR_BC_COUNT_COUNT(x) (((x) & SDMA_PKT_COPY_LINEAR_BC_COUNT_count_mask) << SDMA_PKT_COPY_LINEAR_BC_COUNT_count_shift) + +/*define for PARAMETER word*/ +/*define for dst_sw field*/ +#define SDMA_PKT_COPY_LINEAR_BC_PARAMETER_dst_sw_offset 2 +#define SDMA_PKT_COPY_LINEAR_BC_PARAMETER_dst_sw_mask 0x00000003 +#define SDMA_PKT_COPY_LINEAR_BC_PARAMETER_dst_sw_shift 16 +#define SDMA_PKT_COPY_LINEAR_BC_PARAMETER_DST_SW(x) (((x) & SDMA_PKT_COPY_LINEAR_BC_PARAMETER_dst_sw_mask) << SDMA_PKT_COPY_LINEAR_BC_PARAMETER_dst_sw_shift) + +/*define for dst_ha field*/ +#define SDMA_PKT_COPY_LINEAR_BC_PARAMETER_dst_ha_offset 2 +#define SDMA_PKT_COPY_LINEAR_BC_PARAMETER_dst_ha_mask 0x00000001 +#define SDMA_PKT_COPY_LINEAR_BC_PARAMETER_dst_ha_shift 19 +#define SDMA_PKT_COPY_LINEAR_BC_PARAMETER_DST_HA(x) (((x) & SDMA_PKT_COPY_LINEAR_BC_PARAMETER_dst_ha_mask) << SDMA_PKT_COPY_LINEAR_BC_PARAMETER_dst_ha_shift) + +/*define for src_sw field*/ +#define SDMA_PKT_COPY_LINEAR_BC_PARAMETER_src_sw_offset 2 +#define SDMA_PKT_COPY_LINEAR_BC_PARAMETER_src_sw_mask 0x00000003 +#define SDMA_PKT_COPY_LINEAR_BC_PARAMETER_src_sw_shift 24 +#define SDMA_PKT_COPY_LINEAR_BC_PARAMETER_SRC_SW(x) (((x) & SDMA_PKT_COPY_LINEAR_BC_PARAMETER_src_sw_mask) << SDMA_PKT_COPY_LINEAR_BC_PARAMETER_src_sw_shift) + +/*define for src_ha field*/ +#define SDMA_PKT_COPY_LINEAR_BC_PARAMETER_src_ha_offset 2 +#define SDMA_PKT_COPY_LINEAR_BC_PARAMETER_src_ha_mask 0x00000001 +#define SDMA_PKT_COPY_LINEAR_BC_PARAMETER_src_ha_shift 27 +#define SDMA_PKT_COPY_LINEAR_BC_PARAMETER_SRC_HA(x) (((x) & SDMA_PKT_COPY_LINEAR_BC_PARAMETER_src_ha_mask) << SDMA_PKT_COPY_LINEAR_BC_PARAMETER_src_ha_shift) + +/*define for SRC_ADDR_LO word*/ +/*define for src_addr_31_0 field*/ +#define SDMA_PKT_COPY_LINEAR_BC_SRC_ADDR_LO_src_addr_31_0_offset 3 +#define SDMA_PKT_COPY_LINEAR_BC_SRC_ADDR_LO_src_addr_31_0_mask 0xFFFFFFFF +#define SDMA_PKT_COPY_LINEAR_BC_SRC_ADDR_LO_src_addr_31_0_shift 0 +#define SDMA_PKT_COPY_LINEAR_BC_SRC_ADDR_LO_SRC_ADDR_31_0(x) (((x) & SDMA_PKT_COPY_LINEAR_BC_SRC_ADDR_LO_src_addr_31_0_mask) << SDMA_PKT_COPY_LINEAR_BC_SRC_ADDR_LO_src_addr_31_0_shift) + +/*define for SRC_ADDR_HI word*/ +/*define for src_addr_63_32 field*/ +#define SDMA_PKT_COPY_LINEAR_BC_SRC_ADDR_HI_src_addr_63_32_offset 4 +#define SDMA_PKT_COPY_LINEAR_BC_SRC_ADDR_HI_src_addr_63_32_mask 0xFFFFFFFF +#define SDMA_PKT_COPY_LINEAR_BC_SRC_ADDR_HI_src_addr_63_32_shift 0 +#define SDMA_PKT_COPY_LINEAR_BC_SRC_ADDR_HI_SRC_ADDR_63_32(x) (((x) & SDMA_PKT_COPY_LINEAR_BC_SRC_ADDR_HI_src_addr_63_32_mask) << SDMA_PKT_COPY_LINEAR_BC_SRC_ADDR_HI_src_addr_63_32_shift) + +/*define for DST_ADDR_LO word*/ +/*define for dst_addr_31_0 field*/ +#define SDMA_PKT_COPY_LINEAR_BC_DST_ADDR_LO_dst_addr_31_0_offset 5 +#define SDMA_PKT_COPY_LINEAR_BC_DST_ADDR_LO_dst_addr_31_0_mask 0xFFFFFFFF +#define SDMA_PKT_COPY_LINEAR_BC_DST_ADDR_LO_dst_addr_31_0_shift 0 +#define SDMA_PKT_COPY_LINEAR_BC_DST_ADDR_LO_DST_ADDR_31_0(x) (((x) & SDMA_PKT_COPY_LINEAR_BC_DST_ADDR_LO_dst_addr_31_0_mask) << SDMA_PKT_COPY_LINEAR_BC_DST_ADDR_LO_dst_addr_31_0_shift) + +/*define for DST_ADDR_HI word*/ +/*define for dst_addr_63_32 field*/ +#define SDMA_PKT_COPY_LINEAR_BC_DST_ADDR_HI_dst_addr_63_32_offset 6 +#define SDMA_PKT_COPY_LINEAR_BC_DST_ADDR_HI_dst_addr_63_32_mask 0xFFFFFFFF +#define SDMA_PKT_COPY_LINEAR_BC_DST_ADDR_HI_dst_addr_63_32_shift 0 +#define SDMA_PKT_COPY_LINEAR_BC_DST_ADDR_HI_DST_ADDR_63_32(x) (((x) & SDMA_PKT_COPY_LINEAR_BC_DST_ADDR_HI_dst_addr_63_32_mask) << SDMA_PKT_COPY_LINEAR_BC_DST_ADDR_HI_dst_addr_63_32_shift) + + +/* +** Definitions for SDMA_PKT_COPY_DIRTY_PAGE packet +*/ + +/*define for HEADER word*/ +/*define for op field*/ +#define SDMA_PKT_COPY_DIRTY_PAGE_HEADER_op_offset 0 +#define SDMA_PKT_COPY_DIRTY_PAGE_HEADER_op_mask 0x000000FF +#define SDMA_PKT_COPY_DIRTY_PAGE_HEADER_op_shift 0 +#define SDMA_PKT_COPY_DIRTY_PAGE_HEADER_OP(x) (((x) & SDMA_PKT_COPY_DIRTY_PAGE_HEADER_op_mask) << SDMA_PKT_COPY_DIRTY_PAGE_HEADER_op_shift) + +/*define for sub_op field*/ +#define SDMA_PKT_COPY_DIRTY_PAGE_HEADER_sub_op_offset 0 +#define SDMA_PKT_COPY_DIRTY_PAGE_HEADER_sub_op_mask 0x000000FF +#define SDMA_PKT_COPY_DIRTY_PAGE_HEADER_sub_op_shift 8 +#define SDMA_PKT_COPY_DIRTY_PAGE_HEADER_SUB_OP(x) (((x) & SDMA_PKT_COPY_DIRTY_PAGE_HEADER_sub_op_mask) << SDMA_PKT_COPY_DIRTY_PAGE_HEADER_sub_op_shift) + +/*define for tmz field*/ +#define SDMA_PKT_COPY_DIRTY_PAGE_HEADER_tmz_offset 0 +#define SDMA_PKT_COPY_DIRTY_PAGE_HEADER_tmz_mask 0x00000001 +#define SDMA_PKT_COPY_DIRTY_PAGE_HEADER_tmz_shift 18 +#define SDMA_PKT_COPY_DIRTY_PAGE_HEADER_TMZ(x) (((x) & SDMA_PKT_COPY_DIRTY_PAGE_HEADER_tmz_mask) << SDMA_PKT_COPY_DIRTY_PAGE_HEADER_tmz_shift) + +/*define for cpv field*/ +#define SDMA_PKT_COPY_DIRTY_PAGE_HEADER_cpv_offset 0 +#define SDMA_PKT_COPY_DIRTY_PAGE_HEADER_cpv_mask 0x00000001 +#define SDMA_PKT_COPY_DIRTY_PAGE_HEADER_cpv_shift 19 +#define SDMA_PKT_COPY_DIRTY_PAGE_HEADER_CPV(x) (((x) & SDMA_PKT_COPY_DIRTY_PAGE_HEADER_cpv_mask) << SDMA_PKT_COPY_DIRTY_PAGE_HEADER_cpv_shift) + +/*define for all field*/ +#define SDMA_PKT_COPY_DIRTY_PAGE_HEADER_all_offset 0 +#define SDMA_PKT_COPY_DIRTY_PAGE_HEADER_all_mask 0x00000001 +#define SDMA_PKT_COPY_DIRTY_PAGE_HEADER_all_shift 31 +#define SDMA_PKT_COPY_DIRTY_PAGE_HEADER_ALL(x) (((x) & SDMA_PKT_COPY_DIRTY_PAGE_HEADER_all_mask) << SDMA_PKT_COPY_DIRTY_PAGE_HEADER_all_shift) + +/*define for COUNT word*/ +/*define for count field*/ +#define SDMA_PKT_COPY_DIRTY_PAGE_COUNT_count_offset 1 +#define SDMA_PKT_COPY_DIRTY_PAGE_COUNT_count_mask 0x003FFFFF +#define SDMA_PKT_COPY_DIRTY_PAGE_COUNT_count_shift 0 +#define SDMA_PKT_COPY_DIRTY_PAGE_COUNT_COUNT(x) (((x) & SDMA_PKT_COPY_DIRTY_PAGE_COUNT_count_mask) << SDMA_PKT_COPY_DIRTY_PAGE_COUNT_count_shift) + +/*define for PARAMETER word*/ +/*define for dst_mtype field*/ +#define SDMA_PKT_COPY_DIRTY_PAGE_PARAMETER_dst_mtype_offset 2 +#define SDMA_PKT_COPY_DIRTY_PAGE_PARAMETER_dst_mtype_mask 0x00000007 +#define SDMA_PKT_COPY_DIRTY_PAGE_PARAMETER_dst_mtype_shift 3 +#define SDMA_PKT_COPY_DIRTY_PAGE_PARAMETER_DST_MTYPE(x) (((x) & SDMA_PKT_COPY_DIRTY_PAGE_PARAMETER_dst_mtype_mask) << SDMA_PKT_COPY_DIRTY_PAGE_PARAMETER_dst_mtype_shift) + +/*define for dst_l2_policy field*/ +#define SDMA_PKT_COPY_DIRTY_PAGE_PARAMETER_dst_l2_policy_offset 2 +#define SDMA_PKT_COPY_DIRTY_PAGE_PARAMETER_dst_l2_policy_mask 0x00000003 +#define SDMA_PKT_COPY_DIRTY_PAGE_PARAMETER_dst_l2_policy_shift 6 +#define SDMA_PKT_COPY_DIRTY_PAGE_PARAMETER_DST_L2_POLICY(x) (((x) & SDMA_PKT_COPY_DIRTY_PAGE_PARAMETER_dst_l2_policy_mask) << SDMA_PKT_COPY_DIRTY_PAGE_PARAMETER_dst_l2_policy_shift) + +/*define for dst_llc field*/ +#define SDMA_PKT_COPY_DIRTY_PAGE_PARAMETER_dst_llc_offset 2 +#define SDMA_PKT_COPY_DIRTY_PAGE_PARAMETER_dst_llc_mask 0x00000001 +#define SDMA_PKT_COPY_DIRTY_PAGE_PARAMETER_dst_llc_shift 8 +#define SDMA_PKT_COPY_DIRTY_PAGE_PARAMETER_DST_LLC(x) (((x) & SDMA_PKT_COPY_DIRTY_PAGE_PARAMETER_dst_llc_mask) << SDMA_PKT_COPY_DIRTY_PAGE_PARAMETER_dst_llc_shift) + +/*define for src_mtype field*/ +#define SDMA_PKT_COPY_DIRTY_PAGE_PARAMETER_src_mtype_offset 2 +#define SDMA_PKT_COPY_DIRTY_PAGE_PARAMETER_src_mtype_mask 0x00000007 +#define SDMA_PKT_COPY_DIRTY_PAGE_PARAMETER_src_mtype_shift 11 +#define SDMA_PKT_COPY_DIRTY_PAGE_PARAMETER_SRC_MTYPE(x) (((x) & SDMA_PKT_COPY_DIRTY_PAGE_PARAMETER_src_mtype_mask) << SDMA_PKT_COPY_DIRTY_PAGE_PARAMETER_src_mtype_shift) + +/*define for src_l2_policy field*/ +#define SDMA_PKT_COPY_DIRTY_PAGE_PARAMETER_src_l2_policy_offset 2 +#define SDMA_PKT_COPY_DIRTY_PAGE_PARAMETER_src_l2_policy_mask 0x00000003 +#define SDMA_PKT_COPY_DIRTY_PAGE_PARAMETER_src_l2_policy_shift 14 +#define SDMA_PKT_COPY_DIRTY_PAGE_PARAMETER_SRC_L2_POLICY(x) (((x) & SDMA_PKT_COPY_DIRTY_PAGE_PARAMETER_src_l2_policy_mask) << SDMA_PKT_COPY_DIRTY_PAGE_PARAMETER_src_l2_policy_shift) + +/*define for src_llc field*/ +#define SDMA_PKT_COPY_DIRTY_PAGE_PARAMETER_src_llc_offset 2 +#define SDMA_PKT_COPY_DIRTY_PAGE_PARAMETER_src_llc_mask 0x00000001 +#define SDMA_PKT_COPY_DIRTY_PAGE_PARAMETER_src_llc_shift 16 +#define SDMA_PKT_COPY_DIRTY_PAGE_PARAMETER_SRC_LLC(x) (((x) & SDMA_PKT_COPY_DIRTY_PAGE_PARAMETER_src_llc_mask) << SDMA_PKT_COPY_DIRTY_PAGE_PARAMETER_src_llc_shift) + +/*define for dst_sw field*/ +#define SDMA_PKT_COPY_DIRTY_PAGE_PARAMETER_dst_sw_offset 2 +#define SDMA_PKT_COPY_DIRTY_PAGE_PARAMETER_dst_sw_mask 0x00000003 +#define SDMA_PKT_COPY_DIRTY_PAGE_PARAMETER_dst_sw_shift 17 +#define SDMA_PKT_COPY_DIRTY_PAGE_PARAMETER_DST_SW(x) (((x) & SDMA_PKT_COPY_DIRTY_PAGE_PARAMETER_dst_sw_mask) << SDMA_PKT_COPY_DIRTY_PAGE_PARAMETER_dst_sw_shift) + +/*define for dst_gcc field*/ +#define SDMA_PKT_COPY_DIRTY_PAGE_PARAMETER_dst_gcc_offset 2 +#define SDMA_PKT_COPY_DIRTY_PAGE_PARAMETER_dst_gcc_mask 0x00000001 +#define SDMA_PKT_COPY_DIRTY_PAGE_PARAMETER_dst_gcc_shift 19 +#define SDMA_PKT_COPY_DIRTY_PAGE_PARAMETER_DST_GCC(x) (((x) & SDMA_PKT_COPY_DIRTY_PAGE_PARAMETER_dst_gcc_mask) << SDMA_PKT_COPY_DIRTY_PAGE_PARAMETER_dst_gcc_shift) + +/*define for dst_sys field*/ +#define SDMA_PKT_COPY_DIRTY_PAGE_PARAMETER_dst_sys_offset 2 +#define SDMA_PKT_COPY_DIRTY_PAGE_PARAMETER_dst_sys_mask 0x00000001 +#define SDMA_PKT_COPY_DIRTY_PAGE_PARAMETER_dst_sys_shift 20 +#define SDMA_PKT_COPY_DIRTY_PAGE_PARAMETER_DST_SYS(x) (((x) & SDMA_PKT_COPY_DIRTY_PAGE_PARAMETER_dst_sys_mask) << SDMA_PKT_COPY_DIRTY_PAGE_PARAMETER_dst_sys_shift) + +/*define for dst_snoop field*/ +#define SDMA_PKT_COPY_DIRTY_PAGE_PARAMETER_dst_snoop_offset 2 +#define SDMA_PKT_COPY_DIRTY_PAGE_PARAMETER_dst_snoop_mask 0x00000001 +#define SDMA_PKT_COPY_DIRTY_PAGE_PARAMETER_dst_snoop_shift 22 +#define SDMA_PKT_COPY_DIRTY_PAGE_PARAMETER_DST_SNOOP(x) (((x) & SDMA_PKT_COPY_DIRTY_PAGE_PARAMETER_dst_snoop_mask) << SDMA_PKT_COPY_DIRTY_PAGE_PARAMETER_dst_snoop_shift) + +/*define for dst_gpa field*/ +#define SDMA_PKT_COPY_DIRTY_PAGE_PARAMETER_dst_gpa_offset 2 +#define SDMA_PKT_COPY_DIRTY_PAGE_PARAMETER_dst_gpa_mask 0x00000001 +#define SDMA_PKT_COPY_DIRTY_PAGE_PARAMETER_dst_gpa_shift 23 +#define SDMA_PKT_COPY_DIRTY_PAGE_PARAMETER_DST_GPA(x) (((x) & SDMA_PKT_COPY_DIRTY_PAGE_PARAMETER_dst_gpa_mask) << SDMA_PKT_COPY_DIRTY_PAGE_PARAMETER_dst_gpa_shift) + +/*define for src_sw field*/ +#define SDMA_PKT_COPY_DIRTY_PAGE_PARAMETER_src_sw_offset 2 +#define SDMA_PKT_COPY_DIRTY_PAGE_PARAMETER_src_sw_mask 0x00000003 +#define SDMA_PKT_COPY_DIRTY_PAGE_PARAMETER_src_sw_shift 24 +#define SDMA_PKT_COPY_DIRTY_PAGE_PARAMETER_SRC_SW(x) (((x) & SDMA_PKT_COPY_DIRTY_PAGE_PARAMETER_src_sw_mask) << SDMA_PKT_COPY_DIRTY_PAGE_PARAMETER_src_sw_shift) + +/*define for src_sys field*/ +#define SDMA_PKT_COPY_DIRTY_PAGE_PARAMETER_src_sys_offset 2 +#define SDMA_PKT_COPY_DIRTY_PAGE_PARAMETER_src_sys_mask 0x00000001 +#define SDMA_PKT_COPY_DIRTY_PAGE_PARAMETER_src_sys_shift 28 +#define SDMA_PKT_COPY_DIRTY_PAGE_PARAMETER_SRC_SYS(x) (((x) & SDMA_PKT_COPY_DIRTY_PAGE_PARAMETER_src_sys_mask) << SDMA_PKT_COPY_DIRTY_PAGE_PARAMETER_src_sys_shift) + +/*define for src_snoop field*/ +#define SDMA_PKT_COPY_DIRTY_PAGE_PARAMETER_src_snoop_offset 2 +#define SDMA_PKT_COPY_DIRTY_PAGE_PARAMETER_src_snoop_mask 0x00000001 +#define SDMA_PKT_COPY_DIRTY_PAGE_PARAMETER_src_snoop_shift 30 +#define SDMA_PKT_COPY_DIRTY_PAGE_PARAMETER_SRC_SNOOP(x) (((x) & SDMA_PKT_COPY_DIRTY_PAGE_PARAMETER_src_snoop_mask) << SDMA_PKT_COPY_DIRTY_PAGE_PARAMETER_src_snoop_shift) + +/*define for src_gpa field*/ +#define SDMA_PKT_COPY_DIRTY_PAGE_PARAMETER_src_gpa_offset 2 +#define SDMA_PKT_COPY_DIRTY_PAGE_PARAMETER_src_gpa_mask 0x00000001 +#define SDMA_PKT_COPY_DIRTY_PAGE_PARAMETER_src_gpa_shift 31 +#define SDMA_PKT_COPY_DIRTY_PAGE_PARAMETER_SRC_GPA(x) (((x) & SDMA_PKT_COPY_DIRTY_PAGE_PARAMETER_src_gpa_mask) << SDMA_PKT_COPY_DIRTY_PAGE_PARAMETER_src_gpa_shift) + +/*define for SRC_ADDR_LO word*/ +/*define for src_addr_31_0 field*/ +#define SDMA_PKT_COPY_DIRTY_PAGE_SRC_ADDR_LO_src_addr_31_0_offset 3 +#define SDMA_PKT_COPY_DIRTY_PAGE_SRC_ADDR_LO_src_addr_31_0_mask 0xFFFFFFFF +#define SDMA_PKT_COPY_DIRTY_PAGE_SRC_ADDR_LO_src_addr_31_0_shift 0 +#define SDMA_PKT_COPY_DIRTY_PAGE_SRC_ADDR_LO_SRC_ADDR_31_0(x) (((x) & SDMA_PKT_COPY_DIRTY_PAGE_SRC_ADDR_LO_src_addr_31_0_mask) << SDMA_PKT_COPY_DIRTY_PAGE_SRC_ADDR_LO_src_addr_31_0_shift) + +/*define for SRC_ADDR_HI word*/ +/*define for src_addr_63_32 field*/ +#define SDMA_PKT_COPY_DIRTY_PAGE_SRC_ADDR_HI_src_addr_63_32_offset 4 +#define SDMA_PKT_COPY_DIRTY_PAGE_SRC_ADDR_HI_src_addr_63_32_mask 0xFFFFFFFF +#define SDMA_PKT_COPY_DIRTY_PAGE_SRC_ADDR_HI_src_addr_63_32_shift 0 +#define SDMA_PKT_COPY_DIRTY_PAGE_SRC_ADDR_HI_SRC_ADDR_63_32(x) (((x) & SDMA_PKT_COPY_DIRTY_PAGE_SRC_ADDR_HI_src_addr_63_32_mask) << SDMA_PKT_COPY_DIRTY_PAGE_SRC_ADDR_HI_src_addr_63_32_shift) + +/*define for DST_ADDR_LO word*/ +/*define for dst_addr_31_0 field*/ +#define SDMA_PKT_COPY_DIRTY_PAGE_DST_ADDR_LO_dst_addr_31_0_offset 5 +#define SDMA_PKT_COPY_DIRTY_PAGE_DST_ADDR_LO_dst_addr_31_0_mask 0xFFFFFFFF +#define SDMA_PKT_COPY_DIRTY_PAGE_DST_ADDR_LO_dst_addr_31_0_shift 0 +#define SDMA_PKT_COPY_DIRTY_PAGE_DST_ADDR_LO_DST_ADDR_31_0(x) (((x) & SDMA_PKT_COPY_DIRTY_PAGE_DST_ADDR_LO_dst_addr_31_0_mask) << SDMA_PKT_COPY_DIRTY_PAGE_DST_ADDR_LO_dst_addr_31_0_shift) + +/*define for DST_ADDR_HI word*/ +/*define for dst_addr_63_32 field*/ +#define SDMA_PKT_COPY_DIRTY_PAGE_DST_ADDR_HI_dst_addr_63_32_offset 6 +#define SDMA_PKT_COPY_DIRTY_PAGE_DST_ADDR_HI_dst_addr_63_32_mask 0xFFFFFFFF +#define SDMA_PKT_COPY_DIRTY_PAGE_DST_ADDR_HI_dst_addr_63_32_shift 0 +#define SDMA_PKT_COPY_DIRTY_PAGE_DST_ADDR_HI_DST_ADDR_63_32(x) (((x) & SDMA_PKT_COPY_DIRTY_PAGE_DST_ADDR_HI_dst_addr_63_32_mask) << SDMA_PKT_COPY_DIRTY_PAGE_DST_ADDR_HI_dst_addr_63_32_shift) + + +/* +** Definitions for SDMA_PKT_COPY_PHYSICAL_LINEAR packet +*/ + +/*define for HEADER word*/ +/*define for op field*/ +#define SDMA_PKT_COPY_PHYSICAL_LINEAR_HEADER_op_offset 0 +#define SDMA_PKT_COPY_PHYSICAL_LINEAR_HEADER_op_mask 0x000000FF +#define SDMA_PKT_COPY_PHYSICAL_LINEAR_HEADER_op_shift 0 +#define SDMA_PKT_COPY_PHYSICAL_LINEAR_HEADER_OP(x) (((x) & SDMA_PKT_COPY_PHYSICAL_LINEAR_HEADER_op_mask) << SDMA_PKT_COPY_PHYSICAL_LINEAR_HEADER_op_shift) + +/*define for sub_op field*/ +#define SDMA_PKT_COPY_PHYSICAL_LINEAR_HEADER_sub_op_offset 0 +#define SDMA_PKT_COPY_PHYSICAL_LINEAR_HEADER_sub_op_mask 0x000000FF +#define SDMA_PKT_COPY_PHYSICAL_LINEAR_HEADER_sub_op_shift 8 +#define SDMA_PKT_COPY_PHYSICAL_LINEAR_HEADER_SUB_OP(x) (((x) & SDMA_PKT_COPY_PHYSICAL_LINEAR_HEADER_sub_op_mask) << SDMA_PKT_COPY_PHYSICAL_LINEAR_HEADER_sub_op_shift) + +/*define for tmz field*/ +#define SDMA_PKT_COPY_PHYSICAL_LINEAR_HEADER_tmz_offset 0 +#define SDMA_PKT_COPY_PHYSICAL_LINEAR_HEADER_tmz_mask 0x00000001 +#define SDMA_PKT_COPY_PHYSICAL_LINEAR_HEADER_tmz_shift 18 +#define SDMA_PKT_COPY_PHYSICAL_LINEAR_HEADER_TMZ(x) (((x) & SDMA_PKT_COPY_PHYSICAL_LINEAR_HEADER_tmz_mask) << SDMA_PKT_COPY_PHYSICAL_LINEAR_HEADER_tmz_shift) + +/*define for cpv field*/ +#define SDMA_PKT_COPY_PHYSICAL_LINEAR_HEADER_cpv_offset 0 +#define SDMA_PKT_COPY_PHYSICAL_LINEAR_HEADER_cpv_mask 0x00000001 +#define SDMA_PKT_COPY_PHYSICAL_LINEAR_HEADER_cpv_shift 19 +#define SDMA_PKT_COPY_PHYSICAL_LINEAR_HEADER_CPV(x) (((x) & SDMA_PKT_COPY_PHYSICAL_LINEAR_HEADER_cpv_mask) << SDMA_PKT_COPY_PHYSICAL_LINEAR_HEADER_cpv_shift) + +/*define for COUNT word*/ +/*define for count field*/ +#define SDMA_PKT_COPY_PHYSICAL_LINEAR_COUNT_count_offset 1 +#define SDMA_PKT_COPY_PHYSICAL_LINEAR_COUNT_count_mask 0x003FFFFF +#define SDMA_PKT_COPY_PHYSICAL_LINEAR_COUNT_count_shift 0 +#define SDMA_PKT_COPY_PHYSICAL_LINEAR_COUNT_COUNT(x) (((x) & SDMA_PKT_COPY_PHYSICAL_LINEAR_COUNT_count_mask) << SDMA_PKT_COPY_PHYSICAL_LINEAR_COUNT_count_shift) + +/*define for addr_pair_num field*/ +#define SDMA_PKT_COPY_PHYSICAL_LINEAR_COUNT_addr_pair_num_offset 1 +#define SDMA_PKT_COPY_PHYSICAL_LINEAR_COUNT_addr_pair_num_mask 0x000000FF +#define SDMA_PKT_COPY_PHYSICAL_LINEAR_COUNT_addr_pair_num_shift 24 +#define SDMA_PKT_COPY_PHYSICAL_LINEAR_COUNT_ADDR_PAIR_NUM(x) (((x) & SDMA_PKT_COPY_PHYSICAL_LINEAR_COUNT_addr_pair_num_mask) << SDMA_PKT_COPY_PHYSICAL_LINEAR_COUNT_addr_pair_num_shift) + +/*define for PARAMETER word*/ +/*define for dst_mtype field*/ +#define SDMA_PKT_COPY_PHYSICAL_LINEAR_PARAMETER_dst_mtype_offset 2 +#define SDMA_PKT_COPY_PHYSICAL_LINEAR_PARAMETER_dst_mtype_mask 0x00000007 +#define SDMA_PKT_COPY_PHYSICAL_LINEAR_PARAMETER_dst_mtype_shift 3 +#define SDMA_PKT_COPY_PHYSICAL_LINEAR_PARAMETER_DST_MTYPE(x) (((x) & SDMA_PKT_COPY_PHYSICAL_LINEAR_PARAMETER_dst_mtype_mask) << SDMA_PKT_COPY_PHYSICAL_LINEAR_PARAMETER_dst_mtype_shift) + +/*define for dst_l2_policy field*/ +#define SDMA_PKT_COPY_PHYSICAL_LINEAR_PARAMETER_dst_l2_policy_offset 2 +#define SDMA_PKT_COPY_PHYSICAL_LINEAR_PARAMETER_dst_l2_policy_mask 0x00000003 +#define SDMA_PKT_COPY_PHYSICAL_LINEAR_PARAMETER_dst_l2_policy_shift 6 +#define SDMA_PKT_COPY_PHYSICAL_LINEAR_PARAMETER_DST_L2_POLICY(x) (((x) & SDMA_PKT_COPY_PHYSICAL_LINEAR_PARAMETER_dst_l2_policy_mask) << SDMA_PKT_COPY_PHYSICAL_LINEAR_PARAMETER_dst_l2_policy_shift) + +/*define for dst_llc field*/ +#define SDMA_PKT_COPY_PHYSICAL_LINEAR_PARAMETER_dst_llc_offset 2 +#define SDMA_PKT_COPY_PHYSICAL_LINEAR_PARAMETER_dst_llc_mask 0x00000001 +#define SDMA_PKT_COPY_PHYSICAL_LINEAR_PARAMETER_dst_llc_shift 8 +#define SDMA_PKT_COPY_PHYSICAL_LINEAR_PARAMETER_DST_LLC(x) (((x) & SDMA_PKT_COPY_PHYSICAL_LINEAR_PARAMETER_dst_llc_mask) << SDMA_PKT_COPY_PHYSICAL_LINEAR_PARAMETER_dst_llc_shift) + +/*define for src_mtype field*/ +#define SDMA_PKT_COPY_PHYSICAL_LINEAR_PARAMETER_src_mtype_offset 2 +#define SDMA_PKT_COPY_PHYSICAL_LINEAR_PARAMETER_src_mtype_mask 0x00000007 +#define SDMA_PKT_COPY_PHYSICAL_LINEAR_PARAMETER_src_mtype_shift 11 +#define SDMA_PKT_COPY_PHYSICAL_LINEAR_PARAMETER_SRC_MTYPE(x) (((x) & SDMA_PKT_COPY_PHYSICAL_LINEAR_PARAMETER_src_mtype_mask) << SDMA_PKT_COPY_PHYSICAL_LINEAR_PARAMETER_src_mtype_shift) + +/*define for src_l2_policy field*/ +#define SDMA_PKT_COPY_PHYSICAL_LINEAR_PARAMETER_src_l2_policy_offset 2 +#define SDMA_PKT_COPY_PHYSICAL_LINEAR_PARAMETER_src_l2_policy_mask 0x00000003 +#define SDMA_PKT_COPY_PHYSICAL_LINEAR_PARAMETER_src_l2_policy_shift 14 +#define SDMA_PKT_COPY_PHYSICAL_LINEAR_PARAMETER_SRC_L2_POLICY(x) (((x) & SDMA_PKT_COPY_PHYSICAL_LINEAR_PARAMETER_src_l2_policy_mask) << SDMA_PKT_COPY_PHYSICAL_LINEAR_PARAMETER_src_l2_policy_shift) + +/*define for src_llc field*/ +#define SDMA_PKT_COPY_PHYSICAL_LINEAR_PARAMETER_src_llc_offset 2 +#define SDMA_PKT_COPY_PHYSICAL_LINEAR_PARAMETER_src_llc_mask 0x00000001 +#define SDMA_PKT_COPY_PHYSICAL_LINEAR_PARAMETER_src_llc_shift 16 +#define SDMA_PKT_COPY_PHYSICAL_LINEAR_PARAMETER_SRC_LLC(x) (((x) & SDMA_PKT_COPY_PHYSICAL_LINEAR_PARAMETER_src_llc_mask) << SDMA_PKT_COPY_PHYSICAL_LINEAR_PARAMETER_src_llc_shift) + +/*define for dst_sw field*/ +#define SDMA_PKT_COPY_PHYSICAL_LINEAR_PARAMETER_dst_sw_offset 2 +#define SDMA_PKT_COPY_PHYSICAL_LINEAR_PARAMETER_dst_sw_mask 0x00000003 +#define SDMA_PKT_COPY_PHYSICAL_LINEAR_PARAMETER_dst_sw_shift 17 +#define SDMA_PKT_COPY_PHYSICAL_LINEAR_PARAMETER_DST_SW(x) (((x) & SDMA_PKT_COPY_PHYSICAL_LINEAR_PARAMETER_dst_sw_mask) << SDMA_PKT_COPY_PHYSICAL_LINEAR_PARAMETER_dst_sw_shift) + +/*define for dst_gcc field*/ +#define SDMA_PKT_COPY_PHYSICAL_LINEAR_PARAMETER_dst_gcc_offset 2 +#define SDMA_PKT_COPY_PHYSICAL_LINEAR_PARAMETER_dst_gcc_mask 0x00000001 +#define SDMA_PKT_COPY_PHYSICAL_LINEAR_PARAMETER_dst_gcc_shift 19 +#define SDMA_PKT_COPY_PHYSICAL_LINEAR_PARAMETER_DST_GCC(x) (((x) & SDMA_PKT_COPY_PHYSICAL_LINEAR_PARAMETER_dst_gcc_mask) << SDMA_PKT_COPY_PHYSICAL_LINEAR_PARAMETER_dst_gcc_shift) + +/*define for dst_sys field*/ +#define SDMA_PKT_COPY_PHYSICAL_LINEAR_PARAMETER_dst_sys_offset 2 +#define SDMA_PKT_COPY_PHYSICAL_LINEAR_PARAMETER_dst_sys_mask 0x00000001 +#define SDMA_PKT_COPY_PHYSICAL_LINEAR_PARAMETER_dst_sys_shift 20 +#define SDMA_PKT_COPY_PHYSICAL_LINEAR_PARAMETER_DST_SYS(x) (((x) & SDMA_PKT_COPY_PHYSICAL_LINEAR_PARAMETER_dst_sys_mask) << SDMA_PKT_COPY_PHYSICAL_LINEAR_PARAMETER_dst_sys_shift) + +/*define for dst_log field*/ +#define SDMA_PKT_COPY_PHYSICAL_LINEAR_PARAMETER_dst_log_offset 2 +#define SDMA_PKT_COPY_PHYSICAL_LINEAR_PARAMETER_dst_log_mask 0x00000001 +#define SDMA_PKT_COPY_PHYSICAL_LINEAR_PARAMETER_dst_log_shift 21 +#define SDMA_PKT_COPY_PHYSICAL_LINEAR_PARAMETER_DST_LOG(x) (((x) & SDMA_PKT_COPY_PHYSICAL_LINEAR_PARAMETER_dst_log_mask) << SDMA_PKT_COPY_PHYSICAL_LINEAR_PARAMETER_dst_log_shift) + +/*define for dst_snoop field*/ +#define SDMA_PKT_COPY_PHYSICAL_LINEAR_PARAMETER_dst_snoop_offset 2 +#define SDMA_PKT_COPY_PHYSICAL_LINEAR_PARAMETER_dst_snoop_mask 0x00000001 +#define SDMA_PKT_COPY_PHYSICAL_LINEAR_PARAMETER_dst_snoop_shift 22 +#define SDMA_PKT_COPY_PHYSICAL_LINEAR_PARAMETER_DST_SNOOP(x) (((x) & SDMA_PKT_COPY_PHYSICAL_LINEAR_PARAMETER_dst_snoop_mask) << SDMA_PKT_COPY_PHYSICAL_LINEAR_PARAMETER_dst_snoop_shift) + +/*define for dst_gpa field*/ +#define SDMA_PKT_COPY_PHYSICAL_LINEAR_PARAMETER_dst_gpa_offset 2 +#define SDMA_PKT_COPY_PHYSICAL_LINEAR_PARAMETER_dst_gpa_mask 0x00000001 +#define SDMA_PKT_COPY_PHYSICAL_LINEAR_PARAMETER_dst_gpa_shift 23 +#define SDMA_PKT_COPY_PHYSICAL_LINEAR_PARAMETER_DST_GPA(x) (((x) & SDMA_PKT_COPY_PHYSICAL_LINEAR_PARAMETER_dst_gpa_mask) << SDMA_PKT_COPY_PHYSICAL_LINEAR_PARAMETER_dst_gpa_shift) + +/*define for src_sw field*/ +#define SDMA_PKT_COPY_PHYSICAL_LINEAR_PARAMETER_src_sw_offset 2 +#define SDMA_PKT_COPY_PHYSICAL_LINEAR_PARAMETER_src_sw_mask 0x00000003 +#define SDMA_PKT_COPY_PHYSICAL_LINEAR_PARAMETER_src_sw_shift 24 +#define SDMA_PKT_COPY_PHYSICAL_LINEAR_PARAMETER_SRC_SW(x) (((x) & SDMA_PKT_COPY_PHYSICAL_LINEAR_PARAMETER_src_sw_mask) << SDMA_PKT_COPY_PHYSICAL_LINEAR_PARAMETER_src_sw_shift) + +/*define for src_gcc field*/ +#define SDMA_PKT_COPY_PHYSICAL_LINEAR_PARAMETER_src_gcc_offset 2 +#define SDMA_PKT_COPY_PHYSICAL_LINEAR_PARAMETER_src_gcc_mask 0x00000001 +#define SDMA_PKT_COPY_PHYSICAL_LINEAR_PARAMETER_src_gcc_shift 27 +#define SDMA_PKT_COPY_PHYSICAL_LINEAR_PARAMETER_SRC_GCC(x) (((x) & SDMA_PKT_COPY_PHYSICAL_LINEAR_PARAMETER_src_gcc_mask) << SDMA_PKT_COPY_PHYSICAL_LINEAR_PARAMETER_src_gcc_shift) + +/*define for src_sys field*/ +#define SDMA_PKT_COPY_PHYSICAL_LINEAR_PARAMETER_src_sys_offset 2 +#define SDMA_PKT_COPY_PHYSICAL_LINEAR_PARAMETER_src_sys_mask 0x00000001 +#define SDMA_PKT_COPY_PHYSICAL_LINEAR_PARAMETER_src_sys_shift 28 +#define SDMA_PKT_COPY_PHYSICAL_LINEAR_PARAMETER_SRC_SYS(x) (((x) & SDMA_PKT_COPY_PHYSICAL_LINEAR_PARAMETER_src_sys_mask) << SDMA_PKT_COPY_PHYSICAL_LINEAR_PARAMETER_src_sys_shift) + +/*define for src_snoop field*/ +#define SDMA_PKT_COPY_PHYSICAL_LINEAR_PARAMETER_src_snoop_offset 2 +#define SDMA_PKT_COPY_PHYSICAL_LINEAR_PARAMETER_src_snoop_mask 0x00000001 +#define SDMA_PKT_COPY_PHYSICAL_LINEAR_PARAMETER_src_snoop_shift 30 +#define SDMA_PKT_COPY_PHYSICAL_LINEAR_PARAMETER_SRC_SNOOP(x) (((x) & SDMA_PKT_COPY_PHYSICAL_LINEAR_PARAMETER_src_snoop_mask) << SDMA_PKT_COPY_PHYSICAL_LINEAR_PARAMETER_src_snoop_shift) + +/*define for src_gpa field*/ +#define SDMA_PKT_COPY_PHYSICAL_LINEAR_PARAMETER_src_gpa_offset 2 +#define SDMA_PKT_COPY_PHYSICAL_LINEAR_PARAMETER_src_gpa_mask 0x00000001 +#define SDMA_PKT_COPY_PHYSICAL_LINEAR_PARAMETER_src_gpa_shift 31 +#define SDMA_PKT_COPY_PHYSICAL_LINEAR_PARAMETER_SRC_GPA(x) (((x) & SDMA_PKT_COPY_PHYSICAL_LINEAR_PARAMETER_src_gpa_mask) << SDMA_PKT_COPY_PHYSICAL_LINEAR_PARAMETER_src_gpa_shift) + +/*define for SRC_ADDR_LO word*/ +/*define for src_addr_31_0 field*/ +#define SDMA_PKT_COPY_PHYSICAL_LINEAR_SRC_ADDR_LO_src_addr_31_0_offset 3 +#define SDMA_PKT_COPY_PHYSICAL_LINEAR_SRC_ADDR_LO_src_addr_31_0_mask 0xFFFFFFFF +#define SDMA_PKT_COPY_PHYSICAL_LINEAR_SRC_ADDR_LO_src_addr_31_0_shift 0 +#define SDMA_PKT_COPY_PHYSICAL_LINEAR_SRC_ADDR_LO_SRC_ADDR_31_0(x) (((x) & SDMA_PKT_COPY_PHYSICAL_LINEAR_SRC_ADDR_LO_src_addr_31_0_mask) << SDMA_PKT_COPY_PHYSICAL_LINEAR_SRC_ADDR_LO_src_addr_31_0_shift) + +/*define for SRC_ADDR_HI word*/ +/*define for src_addr_63_32 field*/ +#define SDMA_PKT_COPY_PHYSICAL_LINEAR_SRC_ADDR_HI_src_addr_63_32_offset 4 +#define SDMA_PKT_COPY_PHYSICAL_LINEAR_SRC_ADDR_HI_src_addr_63_32_mask 0xFFFFFFFF +#define SDMA_PKT_COPY_PHYSICAL_LINEAR_SRC_ADDR_HI_src_addr_63_32_shift 0 +#define SDMA_PKT_COPY_PHYSICAL_LINEAR_SRC_ADDR_HI_SRC_ADDR_63_32(x) (((x) & SDMA_PKT_COPY_PHYSICAL_LINEAR_SRC_ADDR_HI_src_addr_63_32_mask) << SDMA_PKT_COPY_PHYSICAL_LINEAR_SRC_ADDR_HI_src_addr_63_32_shift) + +/*define for DST_ADDR_LO word*/ +/*define for dst_addr_31_0 field*/ +#define SDMA_PKT_COPY_PHYSICAL_LINEAR_DST_ADDR_LO_dst_addr_31_0_offset 5 +#define SDMA_PKT_COPY_PHYSICAL_LINEAR_DST_ADDR_LO_dst_addr_31_0_mask 0xFFFFFFFF +#define SDMA_PKT_COPY_PHYSICAL_LINEAR_DST_ADDR_LO_dst_addr_31_0_shift 0 +#define SDMA_PKT_COPY_PHYSICAL_LINEAR_DST_ADDR_LO_DST_ADDR_31_0(x) (((x) & SDMA_PKT_COPY_PHYSICAL_LINEAR_DST_ADDR_LO_dst_addr_31_0_mask) << SDMA_PKT_COPY_PHYSICAL_LINEAR_DST_ADDR_LO_dst_addr_31_0_shift) + +/*define for DST_ADDR_HI word*/ +/*define for dst_addr_63_32 field*/ +#define SDMA_PKT_COPY_PHYSICAL_LINEAR_DST_ADDR_HI_dst_addr_63_32_offset 6 +#define SDMA_PKT_COPY_PHYSICAL_LINEAR_DST_ADDR_HI_dst_addr_63_32_mask 0xFFFFFFFF +#define SDMA_PKT_COPY_PHYSICAL_LINEAR_DST_ADDR_HI_dst_addr_63_32_shift 0 +#define SDMA_PKT_COPY_PHYSICAL_LINEAR_DST_ADDR_HI_DST_ADDR_63_32(x) (((x) & SDMA_PKT_COPY_PHYSICAL_LINEAR_DST_ADDR_HI_dst_addr_63_32_mask) << SDMA_PKT_COPY_PHYSICAL_LINEAR_DST_ADDR_HI_dst_addr_63_32_shift) + + +/* +** Definitions for SDMA_PKT_COPY_BROADCAST_LINEAR packet +*/ + +/*define for HEADER word*/ +/*define for op field*/ +#define SDMA_PKT_COPY_BROADCAST_LINEAR_HEADER_op_offset 0 +#define SDMA_PKT_COPY_BROADCAST_LINEAR_HEADER_op_mask 0x000000FF +#define SDMA_PKT_COPY_BROADCAST_LINEAR_HEADER_op_shift 0 +#define SDMA_PKT_COPY_BROADCAST_LINEAR_HEADER_OP(x) (((x) & SDMA_PKT_COPY_BROADCAST_LINEAR_HEADER_op_mask) << SDMA_PKT_COPY_BROADCAST_LINEAR_HEADER_op_shift) + +/*define for sub_op field*/ +#define SDMA_PKT_COPY_BROADCAST_LINEAR_HEADER_sub_op_offset 0 +#define SDMA_PKT_COPY_BROADCAST_LINEAR_HEADER_sub_op_mask 0x000000FF +#define SDMA_PKT_COPY_BROADCAST_LINEAR_HEADER_sub_op_shift 8 +#define SDMA_PKT_COPY_BROADCAST_LINEAR_HEADER_SUB_OP(x) (((x) & SDMA_PKT_COPY_BROADCAST_LINEAR_HEADER_sub_op_mask) << SDMA_PKT_COPY_BROADCAST_LINEAR_HEADER_sub_op_shift) + +/*define for encrypt field*/ +#define SDMA_PKT_COPY_BROADCAST_LINEAR_HEADER_encrypt_offset 0 +#define SDMA_PKT_COPY_BROADCAST_LINEAR_HEADER_encrypt_mask 0x00000001 +#define SDMA_PKT_COPY_BROADCAST_LINEAR_HEADER_encrypt_shift 16 +#define SDMA_PKT_COPY_BROADCAST_LINEAR_HEADER_ENCRYPT(x) (((x) & SDMA_PKT_COPY_BROADCAST_LINEAR_HEADER_encrypt_mask) << SDMA_PKT_COPY_BROADCAST_LINEAR_HEADER_encrypt_shift) + +/*define for tmz field*/ +#define SDMA_PKT_COPY_BROADCAST_LINEAR_HEADER_tmz_offset 0 +#define SDMA_PKT_COPY_BROADCAST_LINEAR_HEADER_tmz_mask 0x00000001 +#define SDMA_PKT_COPY_BROADCAST_LINEAR_HEADER_tmz_shift 18 +#define SDMA_PKT_COPY_BROADCAST_LINEAR_HEADER_TMZ(x) (((x) & SDMA_PKT_COPY_BROADCAST_LINEAR_HEADER_tmz_mask) << SDMA_PKT_COPY_BROADCAST_LINEAR_HEADER_tmz_shift) + +/*define for cpv field*/ +#define SDMA_PKT_COPY_BROADCAST_LINEAR_HEADER_cpv_offset 0 +#define SDMA_PKT_COPY_BROADCAST_LINEAR_HEADER_cpv_mask 0x00000001 +#define SDMA_PKT_COPY_BROADCAST_LINEAR_HEADER_cpv_shift 19 +#define SDMA_PKT_COPY_BROADCAST_LINEAR_HEADER_CPV(x) (((x) & SDMA_PKT_COPY_BROADCAST_LINEAR_HEADER_cpv_mask) << SDMA_PKT_COPY_BROADCAST_LINEAR_HEADER_cpv_shift) + +/*define for broadcast field*/ +#define SDMA_PKT_COPY_BROADCAST_LINEAR_HEADER_broadcast_offset 0 +#define SDMA_PKT_COPY_BROADCAST_LINEAR_HEADER_broadcast_mask 0x00000001 +#define SDMA_PKT_COPY_BROADCAST_LINEAR_HEADER_broadcast_shift 27 +#define SDMA_PKT_COPY_BROADCAST_LINEAR_HEADER_BROADCAST(x) (((x) & SDMA_PKT_COPY_BROADCAST_LINEAR_HEADER_broadcast_mask) << SDMA_PKT_COPY_BROADCAST_LINEAR_HEADER_broadcast_shift) + +/*define for COUNT word*/ +/*define for count field*/ +#define SDMA_PKT_COPY_BROADCAST_LINEAR_COUNT_count_offset 1 +#define SDMA_PKT_COPY_BROADCAST_LINEAR_COUNT_count_mask 0x3FFFFFFF +#define SDMA_PKT_COPY_BROADCAST_LINEAR_COUNT_count_shift 0 +#define SDMA_PKT_COPY_BROADCAST_LINEAR_COUNT_COUNT(x) (((x) & SDMA_PKT_COPY_BROADCAST_LINEAR_COUNT_count_mask) << SDMA_PKT_COPY_BROADCAST_LINEAR_COUNT_count_shift) + +/*define for PARAMETER word*/ +/*define for dst2_sw field*/ +#define SDMA_PKT_COPY_BROADCAST_LINEAR_PARAMETER_dst2_sw_offset 2 +#define SDMA_PKT_COPY_BROADCAST_LINEAR_PARAMETER_dst2_sw_mask 0x00000003 +#define SDMA_PKT_COPY_BROADCAST_LINEAR_PARAMETER_dst2_sw_shift 8 +#define SDMA_PKT_COPY_BROADCAST_LINEAR_PARAMETER_DST2_SW(x) (((x) & SDMA_PKT_COPY_BROADCAST_LINEAR_PARAMETER_dst2_sw_mask) << SDMA_PKT_COPY_BROADCAST_LINEAR_PARAMETER_dst2_sw_shift) + +/*define for dst2_cache_policy field*/ +#define SDMA_PKT_COPY_BROADCAST_LINEAR_PARAMETER_dst2_cache_policy_offset 2 +#define SDMA_PKT_COPY_BROADCAST_LINEAR_PARAMETER_dst2_cache_policy_mask 0x00000007 +#define SDMA_PKT_COPY_BROADCAST_LINEAR_PARAMETER_dst2_cache_policy_shift 10 +#define SDMA_PKT_COPY_BROADCAST_LINEAR_PARAMETER_DST2_CACHE_POLICY(x) (((x) & SDMA_PKT_COPY_BROADCAST_LINEAR_PARAMETER_dst2_cache_policy_mask) << SDMA_PKT_COPY_BROADCAST_LINEAR_PARAMETER_dst2_cache_policy_shift) + +/*define for dst1_sw field*/ +#define SDMA_PKT_COPY_BROADCAST_LINEAR_PARAMETER_dst1_sw_offset 2 +#define SDMA_PKT_COPY_BROADCAST_LINEAR_PARAMETER_dst1_sw_mask 0x00000003 +#define SDMA_PKT_COPY_BROADCAST_LINEAR_PARAMETER_dst1_sw_shift 16 +#define SDMA_PKT_COPY_BROADCAST_LINEAR_PARAMETER_DST1_SW(x) (((x) & SDMA_PKT_COPY_BROADCAST_LINEAR_PARAMETER_dst1_sw_mask) << SDMA_PKT_COPY_BROADCAST_LINEAR_PARAMETER_dst1_sw_shift) + +/*define for dst1_cache_policy field*/ +#define SDMA_PKT_COPY_BROADCAST_LINEAR_PARAMETER_dst1_cache_policy_offset 2 +#define SDMA_PKT_COPY_BROADCAST_LINEAR_PARAMETER_dst1_cache_policy_mask 0x00000007 +#define SDMA_PKT_COPY_BROADCAST_LINEAR_PARAMETER_dst1_cache_policy_shift 18 +#define SDMA_PKT_COPY_BROADCAST_LINEAR_PARAMETER_DST1_CACHE_POLICY(x) (((x) & SDMA_PKT_COPY_BROADCAST_LINEAR_PARAMETER_dst1_cache_policy_mask) << SDMA_PKT_COPY_BROADCAST_LINEAR_PARAMETER_dst1_cache_policy_shift) + +/*define for src_sw field*/ +#define SDMA_PKT_COPY_BROADCAST_LINEAR_PARAMETER_src_sw_offset 2 +#define SDMA_PKT_COPY_BROADCAST_LINEAR_PARAMETER_src_sw_mask 0x00000003 +#define SDMA_PKT_COPY_BROADCAST_LINEAR_PARAMETER_src_sw_shift 24 +#define SDMA_PKT_COPY_BROADCAST_LINEAR_PARAMETER_SRC_SW(x) (((x) & SDMA_PKT_COPY_BROADCAST_LINEAR_PARAMETER_src_sw_mask) << SDMA_PKT_COPY_BROADCAST_LINEAR_PARAMETER_src_sw_shift) + +/*define for src_cache_policy field*/ +#define SDMA_PKT_COPY_BROADCAST_LINEAR_PARAMETER_src_cache_policy_offset 2 +#define SDMA_PKT_COPY_BROADCAST_LINEAR_PARAMETER_src_cache_policy_mask 0x00000007 +#define SDMA_PKT_COPY_BROADCAST_LINEAR_PARAMETER_src_cache_policy_shift 26 +#define SDMA_PKT_COPY_BROADCAST_LINEAR_PARAMETER_SRC_CACHE_POLICY(x) (((x) & SDMA_PKT_COPY_BROADCAST_LINEAR_PARAMETER_src_cache_policy_mask) << SDMA_PKT_COPY_BROADCAST_LINEAR_PARAMETER_src_cache_policy_shift) + +/*define for SRC_ADDR_LO word*/ +/*define for src_addr_31_0 field*/ +#define SDMA_PKT_COPY_BROADCAST_LINEAR_SRC_ADDR_LO_src_addr_31_0_offset 3 +#define SDMA_PKT_COPY_BROADCAST_LINEAR_SRC_ADDR_LO_src_addr_31_0_mask 0xFFFFFFFF +#define SDMA_PKT_COPY_BROADCAST_LINEAR_SRC_ADDR_LO_src_addr_31_0_shift 0 +#define SDMA_PKT_COPY_BROADCAST_LINEAR_SRC_ADDR_LO_SRC_ADDR_31_0(x) (((x) & SDMA_PKT_COPY_BROADCAST_LINEAR_SRC_ADDR_LO_src_addr_31_0_mask) << SDMA_PKT_COPY_BROADCAST_LINEAR_SRC_ADDR_LO_src_addr_31_0_shift) + +/*define for SRC_ADDR_HI word*/ +/*define for src_addr_63_32 field*/ +#define SDMA_PKT_COPY_BROADCAST_LINEAR_SRC_ADDR_HI_src_addr_63_32_offset 4 +#define SDMA_PKT_COPY_BROADCAST_LINEAR_SRC_ADDR_HI_src_addr_63_32_mask 0xFFFFFFFF +#define SDMA_PKT_COPY_BROADCAST_LINEAR_SRC_ADDR_HI_src_addr_63_32_shift 0 +#define SDMA_PKT_COPY_BROADCAST_LINEAR_SRC_ADDR_HI_SRC_ADDR_63_32(x) (((x) & SDMA_PKT_COPY_BROADCAST_LINEAR_SRC_ADDR_HI_src_addr_63_32_mask) << SDMA_PKT_COPY_BROADCAST_LINEAR_SRC_ADDR_HI_src_addr_63_32_shift) + +/*define for DST1_ADDR_LO word*/ +/*define for dst1_addr_31_0 field*/ +#define SDMA_PKT_COPY_BROADCAST_LINEAR_DST1_ADDR_LO_dst1_addr_31_0_offset 5 +#define SDMA_PKT_COPY_BROADCAST_LINEAR_DST1_ADDR_LO_dst1_addr_31_0_mask 0xFFFFFFFF +#define SDMA_PKT_COPY_BROADCAST_LINEAR_DST1_ADDR_LO_dst1_addr_31_0_shift 0 +#define SDMA_PKT_COPY_BROADCAST_LINEAR_DST1_ADDR_LO_DST1_ADDR_31_0(x) (((x) & SDMA_PKT_COPY_BROADCAST_LINEAR_DST1_ADDR_LO_dst1_addr_31_0_mask) << SDMA_PKT_COPY_BROADCAST_LINEAR_DST1_ADDR_LO_dst1_addr_31_0_shift) + +/*define for DST1_ADDR_HI word*/ +/*define for dst1_addr_63_32 field*/ +#define SDMA_PKT_COPY_BROADCAST_LINEAR_DST1_ADDR_HI_dst1_addr_63_32_offset 6 +#define SDMA_PKT_COPY_BROADCAST_LINEAR_DST1_ADDR_HI_dst1_addr_63_32_mask 0xFFFFFFFF +#define SDMA_PKT_COPY_BROADCAST_LINEAR_DST1_ADDR_HI_dst1_addr_63_32_shift 0 +#define SDMA_PKT_COPY_BROADCAST_LINEAR_DST1_ADDR_HI_DST1_ADDR_63_32(x) (((x) & SDMA_PKT_COPY_BROADCAST_LINEAR_DST1_ADDR_HI_dst1_addr_63_32_mask) << SDMA_PKT_COPY_BROADCAST_LINEAR_DST1_ADDR_HI_dst1_addr_63_32_shift) + +/*define for DST2_ADDR_LO word*/ +/*define for dst2_addr_31_0 field*/ +#define SDMA_PKT_COPY_BROADCAST_LINEAR_DST2_ADDR_LO_dst2_addr_31_0_offset 7 +#define SDMA_PKT_COPY_BROADCAST_LINEAR_DST2_ADDR_LO_dst2_addr_31_0_mask 0xFFFFFFFF +#define SDMA_PKT_COPY_BROADCAST_LINEAR_DST2_ADDR_LO_dst2_addr_31_0_shift 0 +#define SDMA_PKT_COPY_BROADCAST_LINEAR_DST2_ADDR_LO_DST2_ADDR_31_0(x) (((x) & SDMA_PKT_COPY_BROADCAST_LINEAR_DST2_ADDR_LO_dst2_addr_31_0_mask) << SDMA_PKT_COPY_BROADCAST_LINEAR_DST2_ADDR_LO_dst2_addr_31_0_shift) + +/*define for DST2_ADDR_HI word*/ +/*define for dst2_addr_63_32 field*/ +#define SDMA_PKT_COPY_BROADCAST_LINEAR_DST2_ADDR_HI_dst2_addr_63_32_offset 8 +#define SDMA_PKT_COPY_BROADCAST_LINEAR_DST2_ADDR_HI_dst2_addr_63_32_mask 0xFFFFFFFF +#define SDMA_PKT_COPY_BROADCAST_LINEAR_DST2_ADDR_HI_dst2_addr_63_32_shift 0 +#define SDMA_PKT_COPY_BROADCAST_LINEAR_DST2_ADDR_HI_DST2_ADDR_63_32(x) (((x) & SDMA_PKT_COPY_BROADCAST_LINEAR_DST2_ADDR_HI_dst2_addr_63_32_mask) << SDMA_PKT_COPY_BROADCAST_LINEAR_DST2_ADDR_HI_dst2_addr_63_32_shift) + + +/* +** Definitions for SDMA_PKT_COPY_LINEAR_SUBWIN packet +*/ + +/*define for HEADER word*/ +/*define for op field*/ +#define SDMA_PKT_COPY_LINEAR_SUBWIN_HEADER_op_offset 0 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_HEADER_op_mask 0x000000FF +#define SDMA_PKT_COPY_LINEAR_SUBWIN_HEADER_op_shift 0 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_HEADER_OP(x) (((x) & SDMA_PKT_COPY_LINEAR_SUBWIN_HEADER_op_mask) << SDMA_PKT_COPY_LINEAR_SUBWIN_HEADER_op_shift) + +/*define for sub_op field*/ +#define SDMA_PKT_COPY_LINEAR_SUBWIN_HEADER_sub_op_offset 0 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_HEADER_sub_op_mask 0x000000FF +#define SDMA_PKT_COPY_LINEAR_SUBWIN_HEADER_sub_op_shift 8 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_HEADER_SUB_OP(x) (((x) & SDMA_PKT_COPY_LINEAR_SUBWIN_HEADER_sub_op_mask) << SDMA_PKT_COPY_LINEAR_SUBWIN_HEADER_sub_op_shift) + +/*define for tmz field*/ +#define SDMA_PKT_COPY_LINEAR_SUBWIN_HEADER_tmz_offset 0 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_HEADER_tmz_mask 0x00000001 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_HEADER_tmz_shift 18 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_HEADER_TMZ(x) (((x) & SDMA_PKT_COPY_LINEAR_SUBWIN_HEADER_tmz_mask) << SDMA_PKT_COPY_LINEAR_SUBWIN_HEADER_tmz_shift) + +/*define for cpv field*/ +#define SDMA_PKT_COPY_LINEAR_SUBWIN_HEADER_cpv_offset 0 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_HEADER_cpv_mask 0x00000001 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_HEADER_cpv_shift 19 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_HEADER_CPV(x) (((x) & SDMA_PKT_COPY_LINEAR_SUBWIN_HEADER_cpv_mask) << SDMA_PKT_COPY_LINEAR_SUBWIN_HEADER_cpv_shift) + +/*define for elementsize field*/ +#define SDMA_PKT_COPY_LINEAR_SUBWIN_HEADER_elementsize_offset 0 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_HEADER_elementsize_mask 0x00000007 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_HEADER_elementsize_shift 29 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_HEADER_ELEMENTSIZE(x) (((x) & SDMA_PKT_COPY_LINEAR_SUBWIN_HEADER_elementsize_mask) << SDMA_PKT_COPY_LINEAR_SUBWIN_HEADER_elementsize_shift) + +/*define for SRC_ADDR_LO word*/ +/*define for src_addr_31_0 field*/ +#define SDMA_PKT_COPY_LINEAR_SUBWIN_SRC_ADDR_LO_src_addr_31_0_offset 1 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_SRC_ADDR_LO_src_addr_31_0_mask 0xFFFFFFFF +#define SDMA_PKT_COPY_LINEAR_SUBWIN_SRC_ADDR_LO_src_addr_31_0_shift 0 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_SRC_ADDR_LO_SRC_ADDR_31_0(x) (((x) & SDMA_PKT_COPY_LINEAR_SUBWIN_SRC_ADDR_LO_src_addr_31_0_mask) << SDMA_PKT_COPY_LINEAR_SUBWIN_SRC_ADDR_LO_src_addr_31_0_shift) + +/*define for SRC_ADDR_HI word*/ +/*define for src_addr_63_32 field*/ +#define SDMA_PKT_COPY_LINEAR_SUBWIN_SRC_ADDR_HI_src_addr_63_32_offset 2 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_SRC_ADDR_HI_src_addr_63_32_mask 0xFFFFFFFF +#define SDMA_PKT_COPY_LINEAR_SUBWIN_SRC_ADDR_HI_src_addr_63_32_shift 0 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_SRC_ADDR_HI_SRC_ADDR_63_32(x) (((x) & SDMA_PKT_COPY_LINEAR_SUBWIN_SRC_ADDR_HI_src_addr_63_32_mask) << SDMA_PKT_COPY_LINEAR_SUBWIN_SRC_ADDR_HI_src_addr_63_32_shift) + +/*define for DW_3 word*/ +/*define for src_x field*/ +#define SDMA_PKT_COPY_LINEAR_SUBWIN_DW_3_src_x_offset 3 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_DW_3_src_x_mask 0x00003FFF +#define SDMA_PKT_COPY_LINEAR_SUBWIN_DW_3_src_x_shift 0 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_DW_3_SRC_X(x) (((x) & SDMA_PKT_COPY_LINEAR_SUBWIN_DW_3_src_x_mask) << SDMA_PKT_COPY_LINEAR_SUBWIN_DW_3_src_x_shift) + +/*define for src_y field*/ +#define SDMA_PKT_COPY_LINEAR_SUBWIN_DW_3_src_y_offset 3 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_DW_3_src_y_mask 0x00003FFF +#define SDMA_PKT_COPY_LINEAR_SUBWIN_DW_3_src_y_shift 16 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_DW_3_SRC_Y(x) (((x) & SDMA_PKT_COPY_LINEAR_SUBWIN_DW_3_src_y_mask) << SDMA_PKT_COPY_LINEAR_SUBWIN_DW_3_src_y_shift) + +/*define for DW_4 word*/ +/*define for src_z field*/ +#define SDMA_PKT_COPY_LINEAR_SUBWIN_DW_4_src_z_offset 4 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_DW_4_src_z_mask 0x00001FFF +#define SDMA_PKT_COPY_LINEAR_SUBWIN_DW_4_src_z_shift 0 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_DW_4_SRC_Z(x) (((x) & SDMA_PKT_COPY_LINEAR_SUBWIN_DW_4_src_z_mask) << SDMA_PKT_COPY_LINEAR_SUBWIN_DW_4_src_z_shift) + +/*define for src_pitch field*/ +#define SDMA_PKT_COPY_LINEAR_SUBWIN_DW_4_src_pitch_offset 4 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_DW_4_src_pitch_mask 0x0007FFFF +#define SDMA_PKT_COPY_LINEAR_SUBWIN_DW_4_src_pitch_shift 13 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_DW_4_SRC_PITCH(x) (((x) & SDMA_PKT_COPY_LINEAR_SUBWIN_DW_4_src_pitch_mask) << SDMA_PKT_COPY_LINEAR_SUBWIN_DW_4_src_pitch_shift) + +/*define for DW_5 word*/ +/*define for src_slice_pitch field*/ +#define SDMA_PKT_COPY_LINEAR_SUBWIN_DW_5_src_slice_pitch_offset 5 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_DW_5_src_slice_pitch_mask 0x0FFFFFFF +#define SDMA_PKT_COPY_LINEAR_SUBWIN_DW_5_src_slice_pitch_shift 0 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_DW_5_SRC_SLICE_PITCH(x) (((x) & SDMA_PKT_COPY_LINEAR_SUBWIN_DW_5_src_slice_pitch_mask) << SDMA_PKT_COPY_LINEAR_SUBWIN_DW_5_src_slice_pitch_shift) + +/*define for DST_ADDR_LO word*/ +/*define for dst_addr_31_0 field*/ +#define SDMA_PKT_COPY_LINEAR_SUBWIN_DST_ADDR_LO_dst_addr_31_0_offset 6 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_DST_ADDR_LO_dst_addr_31_0_mask 0xFFFFFFFF +#define SDMA_PKT_COPY_LINEAR_SUBWIN_DST_ADDR_LO_dst_addr_31_0_shift 0 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_DST_ADDR_LO_DST_ADDR_31_0(x) (((x) & SDMA_PKT_COPY_LINEAR_SUBWIN_DST_ADDR_LO_dst_addr_31_0_mask) << SDMA_PKT_COPY_LINEAR_SUBWIN_DST_ADDR_LO_dst_addr_31_0_shift) + +/*define for DST_ADDR_HI word*/ +/*define for dst_addr_63_32 field*/ +#define SDMA_PKT_COPY_LINEAR_SUBWIN_DST_ADDR_HI_dst_addr_63_32_offset 7 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_DST_ADDR_HI_dst_addr_63_32_mask 0xFFFFFFFF +#define SDMA_PKT_COPY_LINEAR_SUBWIN_DST_ADDR_HI_dst_addr_63_32_shift 0 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_DST_ADDR_HI_DST_ADDR_63_32(x) (((x) & SDMA_PKT_COPY_LINEAR_SUBWIN_DST_ADDR_HI_dst_addr_63_32_mask) << SDMA_PKT_COPY_LINEAR_SUBWIN_DST_ADDR_HI_dst_addr_63_32_shift) + +/*define for DW_8 word*/ +/*define for dst_x field*/ +#define SDMA_PKT_COPY_LINEAR_SUBWIN_DW_8_dst_x_offset 8 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_DW_8_dst_x_mask 0x00003FFF +#define SDMA_PKT_COPY_LINEAR_SUBWIN_DW_8_dst_x_shift 0 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_DW_8_DST_X(x) (((x) & SDMA_PKT_COPY_LINEAR_SUBWIN_DW_8_dst_x_mask) << SDMA_PKT_COPY_LINEAR_SUBWIN_DW_8_dst_x_shift) + +/*define for dst_y field*/ +#define SDMA_PKT_COPY_LINEAR_SUBWIN_DW_8_dst_y_offset 8 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_DW_8_dst_y_mask 0x00003FFF +#define SDMA_PKT_COPY_LINEAR_SUBWIN_DW_8_dst_y_shift 16 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_DW_8_DST_Y(x) (((x) & SDMA_PKT_COPY_LINEAR_SUBWIN_DW_8_dst_y_mask) << SDMA_PKT_COPY_LINEAR_SUBWIN_DW_8_dst_y_shift) + +/*define for DW_9 word*/ +/*define for dst_z field*/ +#define SDMA_PKT_COPY_LINEAR_SUBWIN_DW_9_dst_z_offset 9 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_DW_9_dst_z_mask 0x00001FFF +#define SDMA_PKT_COPY_LINEAR_SUBWIN_DW_9_dst_z_shift 0 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_DW_9_DST_Z(x) (((x) & SDMA_PKT_COPY_LINEAR_SUBWIN_DW_9_dst_z_mask) << SDMA_PKT_COPY_LINEAR_SUBWIN_DW_9_dst_z_shift) + +/*define for dst_pitch field*/ +#define SDMA_PKT_COPY_LINEAR_SUBWIN_DW_9_dst_pitch_offset 9 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_DW_9_dst_pitch_mask 0x0007FFFF +#define SDMA_PKT_COPY_LINEAR_SUBWIN_DW_9_dst_pitch_shift 13 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_DW_9_DST_PITCH(x) (((x) & SDMA_PKT_COPY_LINEAR_SUBWIN_DW_9_dst_pitch_mask) << SDMA_PKT_COPY_LINEAR_SUBWIN_DW_9_dst_pitch_shift) + +/*define for DW_10 word*/ +/*define for dst_slice_pitch field*/ +#define SDMA_PKT_COPY_LINEAR_SUBWIN_DW_10_dst_slice_pitch_offset 10 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_DW_10_dst_slice_pitch_mask 0x0FFFFFFF +#define SDMA_PKT_COPY_LINEAR_SUBWIN_DW_10_dst_slice_pitch_shift 0 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_DW_10_DST_SLICE_PITCH(x) (((x) & SDMA_PKT_COPY_LINEAR_SUBWIN_DW_10_dst_slice_pitch_mask) << SDMA_PKT_COPY_LINEAR_SUBWIN_DW_10_dst_slice_pitch_shift) + +/*define for DW_11 word*/ +/*define for rect_x field*/ +#define SDMA_PKT_COPY_LINEAR_SUBWIN_DW_11_rect_x_offset 11 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_DW_11_rect_x_mask 0x00003FFF +#define SDMA_PKT_COPY_LINEAR_SUBWIN_DW_11_rect_x_shift 0 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_DW_11_RECT_X(x) (((x) & SDMA_PKT_COPY_LINEAR_SUBWIN_DW_11_rect_x_mask) << SDMA_PKT_COPY_LINEAR_SUBWIN_DW_11_rect_x_shift) + +/*define for rect_y field*/ +#define SDMA_PKT_COPY_LINEAR_SUBWIN_DW_11_rect_y_offset 11 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_DW_11_rect_y_mask 0x00003FFF +#define SDMA_PKT_COPY_LINEAR_SUBWIN_DW_11_rect_y_shift 16 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_DW_11_RECT_Y(x) (((x) & SDMA_PKT_COPY_LINEAR_SUBWIN_DW_11_rect_y_mask) << SDMA_PKT_COPY_LINEAR_SUBWIN_DW_11_rect_y_shift) + +/*define for DW_12 word*/ +/*define for rect_z field*/ +#define SDMA_PKT_COPY_LINEAR_SUBWIN_DW_12_rect_z_offset 12 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_DW_12_rect_z_mask 0x00001FFF +#define SDMA_PKT_COPY_LINEAR_SUBWIN_DW_12_rect_z_shift 0 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_DW_12_RECT_Z(x) (((x) & SDMA_PKT_COPY_LINEAR_SUBWIN_DW_12_rect_z_mask) << SDMA_PKT_COPY_LINEAR_SUBWIN_DW_12_rect_z_shift) + +/*define for dst_sw field*/ +#define SDMA_PKT_COPY_LINEAR_SUBWIN_DW_12_dst_sw_offset 12 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_DW_12_dst_sw_mask 0x00000003 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_DW_12_dst_sw_shift 16 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_DW_12_DST_SW(x) (((x) & SDMA_PKT_COPY_LINEAR_SUBWIN_DW_12_dst_sw_mask) << SDMA_PKT_COPY_LINEAR_SUBWIN_DW_12_dst_sw_shift) + +/*define for dst_cache_policy field*/ +#define SDMA_PKT_COPY_LINEAR_SUBWIN_DW_12_dst_cache_policy_offset 12 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_DW_12_dst_cache_policy_mask 0x00000007 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_DW_12_dst_cache_policy_shift 18 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_DW_12_DST_CACHE_POLICY(x) (((x) & SDMA_PKT_COPY_LINEAR_SUBWIN_DW_12_dst_cache_policy_mask) << SDMA_PKT_COPY_LINEAR_SUBWIN_DW_12_dst_cache_policy_shift) + +/*define for src_sw field*/ +#define SDMA_PKT_COPY_LINEAR_SUBWIN_DW_12_src_sw_offset 12 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_DW_12_src_sw_mask 0x00000003 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_DW_12_src_sw_shift 24 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_DW_12_SRC_SW(x) (((x) & SDMA_PKT_COPY_LINEAR_SUBWIN_DW_12_src_sw_mask) << SDMA_PKT_COPY_LINEAR_SUBWIN_DW_12_src_sw_shift) + +/*define for src_cache_policy field*/ +#define SDMA_PKT_COPY_LINEAR_SUBWIN_DW_12_src_cache_policy_offset 12 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_DW_12_src_cache_policy_mask 0x00000007 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_DW_12_src_cache_policy_shift 26 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_DW_12_SRC_CACHE_POLICY(x) (((x) & SDMA_PKT_COPY_LINEAR_SUBWIN_DW_12_src_cache_policy_mask) << SDMA_PKT_COPY_LINEAR_SUBWIN_DW_12_src_cache_policy_shift) + + +/* +** Definitions for SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE packet +*/ + +/*define for HEADER word*/ +/*define for op field*/ +#define SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_HEADER_op_offset 0 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_HEADER_op_mask 0x000000FF +#define SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_HEADER_op_shift 0 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_HEADER_OP(x) (((x) & SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_HEADER_op_mask) << SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_HEADER_op_shift) + +/*define for sub_op field*/ +#define SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_HEADER_sub_op_offset 0 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_HEADER_sub_op_mask 0x000000FF +#define SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_HEADER_sub_op_shift 8 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_HEADER_SUB_OP(x) (((x) & SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_HEADER_sub_op_mask) << SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_HEADER_sub_op_shift) + +/*define for tmz field*/ +#define SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_HEADER_tmz_offset 0 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_HEADER_tmz_mask 0x00000001 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_HEADER_tmz_shift 18 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_HEADER_TMZ(x) (((x) & SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_HEADER_tmz_mask) << SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_HEADER_tmz_shift) + +/*define for cpv field*/ +#define SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_HEADER_cpv_offset 0 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_HEADER_cpv_mask 0x00000001 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_HEADER_cpv_shift 19 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_HEADER_CPV(x) (((x) & SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_HEADER_cpv_mask) << SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_HEADER_cpv_shift) + +/*define for SRC_ADDR_LO word*/ +/*define for src_addr_31_0 field*/ +#define SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_SRC_ADDR_LO_src_addr_31_0_offset 1 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_SRC_ADDR_LO_src_addr_31_0_mask 0xFFFFFFFF +#define SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_SRC_ADDR_LO_src_addr_31_0_shift 0 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_SRC_ADDR_LO_SRC_ADDR_31_0(x) (((x) & SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_SRC_ADDR_LO_src_addr_31_0_mask) << SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_SRC_ADDR_LO_src_addr_31_0_shift) + +/*define for SRC_ADDR_HI word*/ +/*define for src_addr_63_32 field*/ +#define SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_SRC_ADDR_HI_src_addr_63_32_offset 2 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_SRC_ADDR_HI_src_addr_63_32_mask 0xFFFFFFFF +#define SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_SRC_ADDR_HI_src_addr_63_32_shift 0 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_SRC_ADDR_HI_SRC_ADDR_63_32(x) (((x) & SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_SRC_ADDR_HI_src_addr_63_32_mask) << SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_SRC_ADDR_HI_src_addr_63_32_shift) + +/*define for DW_3 word*/ +/*define for src_x field*/ +#define SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_3_src_x_offset 3 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_3_src_x_mask 0xFFFFFFFF +#define SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_3_src_x_shift 0 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_3_SRC_X(x) (((x) & SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_3_src_x_mask) << SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_3_src_x_shift) + +/*define for DW_4 word*/ +/*define for src_y field*/ +#define SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_4_src_y_offset 4 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_4_src_y_mask 0xFFFFFFFF +#define SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_4_src_y_shift 0 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_4_SRC_Y(x) (((x) & SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_4_src_y_mask) << SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_4_src_y_shift) + +/*define for DW_5 word*/ +/*define for src_z field*/ +#define SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_5_src_z_offset 5 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_5_src_z_mask 0xFFFFFFFF +#define SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_5_src_z_shift 0 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_5_SRC_Z(x) (((x) & SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_5_src_z_mask) << SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_5_src_z_shift) + +/*define for DW_6 word*/ +/*define for src_pitch field*/ +#define SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_6_src_pitch_offset 6 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_6_src_pitch_mask 0xFFFFFFFF +#define SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_6_src_pitch_shift 0 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_6_SRC_PITCH(x) (((x) & SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_6_src_pitch_mask) << SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_6_src_pitch_shift) + +/*define for DW_7 word*/ +/*define for src_slice_pitch_31_0 field*/ +#define SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_7_src_slice_pitch_31_0_offset 7 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_7_src_slice_pitch_31_0_mask 0xFFFFFFFF +#define SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_7_src_slice_pitch_31_0_shift 0 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_7_SRC_SLICE_PITCH_31_0(x) (((x) & SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_7_src_slice_pitch_31_0_mask) << SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_7_src_slice_pitch_31_0_shift) + +/*define for DW_8 word*/ +/*define for src_slice_pitch_47_32 field*/ +#define SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_8_src_slice_pitch_47_32_offset 8 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_8_src_slice_pitch_47_32_mask 0x0000FFFF +#define SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_8_src_slice_pitch_47_32_shift 0 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_8_SRC_SLICE_PITCH_47_32(x) (((x) & SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_8_src_slice_pitch_47_32_mask) << SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_8_src_slice_pitch_47_32_shift) + +/*define for DST_ADDR_LO word*/ +/*define for dst_addr_31_0 field*/ +#define SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DST_ADDR_LO_dst_addr_31_0_offset 9 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DST_ADDR_LO_dst_addr_31_0_mask 0xFFFFFFFF +#define SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DST_ADDR_LO_dst_addr_31_0_shift 0 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DST_ADDR_LO_DST_ADDR_31_0(x) (((x) & SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DST_ADDR_LO_dst_addr_31_0_mask) << SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DST_ADDR_LO_dst_addr_31_0_shift) + +/*define for DST_ADDR_HI word*/ +/*define for dst_addr_63_32 field*/ +#define SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DST_ADDR_HI_dst_addr_63_32_offset 10 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DST_ADDR_HI_dst_addr_63_32_mask 0xFFFFFFFF +#define SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DST_ADDR_HI_dst_addr_63_32_shift 0 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DST_ADDR_HI_DST_ADDR_63_32(x) (((x) & SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DST_ADDR_HI_dst_addr_63_32_mask) << SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DST_ADDR_HI_dst_addr_63_32_shift) + +/*define for DW_11 word*/ +/*define for dst_x field*/ +#define SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_11_dst_x_offset 11 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_11_dst_x_mask 0xFFFFFFFF +#define SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_11_dst_x_shift 0 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_11_DST_X(x) (((x) & SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_11_dst_x_mask) << SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_11_dst_x_shift) + +/*define for DW_12 word*/ +/*define for dst_y field*/ +#define SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_12_dst_y_offset 12 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_12_dst_y_mask 0xFFFFFFFF +#define SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_12_dst_y_shift 0 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_12_DST_Y(x) (((x) & SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_12_dst_y_mask) << SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_12_dst_y_shift) + +/*define for DW_13 word*/ +/*define for dst_z field*/ +#define SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_13_dst_z_offset 13 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_13_dst_z_mask 0xFFFFFFFF +#define SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_13_dst_z_shift 0 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_13_DST_Z(x) (((x) & SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_13_dst_z_mask) << SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_13_dst_z_shift) + +/*define for DW_14 word*/ +/*define for dst_pitch field*/ +#define SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_14_dst_pitch_offset 14 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_14_dst_pitch_mask 0xFFFFFFFF +#define SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_14_dst_pitch_shift 0 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_14_DST_PITCH(x) (((x) & SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_14_dst_pitch_mask) << SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_14_dst_pitch_shift) + +/*define for DW_15 word*/ +/*define for dst_slice_pitch_31_0 field*/ +#define SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_15_dst_slice_pitch_31_0_offset 15 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_15_dst_slice_pitch_31_0_mask 0xFFFFFFFF +#define SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_15_dst_slice_pitch_31_0_shift 0 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_15_DST_SLICE_PITCH_31_0(x) (((x) & SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_15_dst_slice_pitch_31_0_mask) << SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_15_dst_slice_pitch_31_0_shift) + +/*define for DW_16 word*/ +/*define for dst_slice_pitch_47_32 field*/ +#define SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_16_dst_slice_pitch_47_32_offset 16 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_16_dst_slice_pitch_47_32_mask 0x0000FFFF +#define SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_16_dst_slice_pitch_47_32_shift 0 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_16_DST_SLICE_PITCH_47_32(x) (((x) & SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_16_dst_slice_pitch_47_32_mask) << SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_16_dst_slice_pitch_47_32_shift) + +/*define for dst_sw field*/ +#define SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_16_dst_sw_offset 16 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_16_dst_sw_mask 0x00000003 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_16_dst_sw_shift 16 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_16_DST_SW(x) (((x) & SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_16_dst_sw_mask) << SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_16_dst_sw_shift) + +/*define for dst_policy field*/ +#define SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_16_dst_policy_offset 16 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_16_dst_policy_mask 0x00000007 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_16_dst_policy_shift 18 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_16_DST_POLICY(x) (((x) & SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_16_dst_policy_mask) << SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_16_dst_policy_shift) + +/*define for src_sw field*/ +#define SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_16_src_sw_offset 16 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_16_src_sw_mask 0x00000003 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_16_src_sw_shift 24 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_16_SRC_SW(x) (((x) & SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_16_src_sw_mask) << SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_16_src_sw_shift) + +/*define for src_policy field*/ +#define SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_16_src_policy_offset 16 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_16_src_policy_mask 0x00000007 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_16_src_policy_shift 26 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_16_SRC_POLICY(x) (((x) & SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_16_src_policy_mask) << SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_16_src_policy_shift) + +/*define for DW_17 word*/ +/*define for rect_x field*/ +#define SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_17_rect_x_offset 17 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_17_rect_x_mask 0xFFFFFFFF +#define SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_17_rect_x_shift 0 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_17_RECT_X(x) (((x) & SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_17_rect_x_mask) << SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_17_rect_x_shift) + +/*define for DW_18 word*/ +/*define for rect_y field*/ +#define SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_18_rect_y_offset 18 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_18_rect_y_mask 0xFFFFFFFF +#define SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_18_rect_y_shift 0 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_18_RECT_Y(x) (((x) & SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_18_rect_y_mask) << SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_18_rect_y_shift) + +/*define for DW_19 word*/ +/*define for rect_z field*/ +#define SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_19_rect_z_offset 19 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_19_rect_z_mask 0xFFFFFFFF +#define SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_19_rect_z_shift 0 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_19_RECT_Z(x) (((x) & SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_19_rect_z_mask) << SDMA_PKT_COPY_LINEAR_SUBWIN_LARGE_DW_19_rect_z_shift) + + +/* +** Definitions for SDMA_PKT_COPY_LINEAR_SUBWIN_BC packet +*/ + +/*define for HEADER word*/ +/*define for op field*/ +#define SDMA_PKT_COPY_LINEAR_SUBWIN_BC_HEADER_op_offset 0 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_BC_HEADER_op_mask 0x000000FF +#define SDMA_PKT_COPY_LINEAR_SUBWIN_BC_HEADER_op_shift 0 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_BC_HEADER_OP(x) (((x) & SDMA_PKT_COPY_LINEAR_SUBWIN_BC_HEADER_op_mask) << SDMA_PKT_COPY_LINEAR_SUBWIN_BC_HEADER_op_shift) + +/*define for sub_op field*/ +#define SDMA_PKT_COPY_LINEAR_SUBWIN_BC_HEADER_sub_op_offset 0 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_BC_HEADER_sub_op_mask 0x000000FF +#define SDMA_PKT_COPY_LINEAR_SUBWIN_BC_HEADER_sub_op_shift 8 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_BC_HEADER_SUB_OP(x) (((x) & SDMA_PKT_COPY_LINEAR_SUBWIN_BC_HEADER_sub_op_mask) << SDMA_PKT_COPY_LINEAR_SUBWIN_BC_HEADER_sub_op_shift) + +/*define for elementsize field*/ +#define SDMA_PKT_COPY_LINEAR_SUBWIN_BC_HEADER_elementsize_offset 0 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_BC_HEADER_elementsize_mask 0x00000007 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_BC_HEADER_elementsize_shift 29 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_BC_HEADER_ELEMENTSIZE(x) (((x) & SDMA_PKT_COPY_LINEAR_SUBWIN_BC_HEADER_elementsize_mask) << SDMA_PKT_COPY_LINEAR_SUBWIN_BC_HEADER_elementsize_shift) + +/*define for SRC_ADDR_LO word*/ +/*define for src_addr_31_0 field*/ +#define SDMA_PKT_COPY_LINEAR_SUBWIN_BC_SRC_ADDR_LO_src_addr_31_0_offset 1 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_BC_SRC_ADDR_LO_src_addr_31_0_mask 0xFFFFFFFF +#define SDMA_PKT_COPY_LINEAR_SUBWIN_BC_SRC_ADDR_LO_src_addr_31_0_shift 0 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_BC_SRC_ADDR_LO_SRC_ADDR_31_0(x) (((x) & SDMA_PKT_COPY_LINEAR_SUBWIN_BC_SRC_ADDR_LO_src_addr_31_0_mask) << SDMA_PKT_COPY_LINEAR_SUBWIN_BC_SRC_ADDR_LO_src_addr_31_0_shift) + +/*define for SRC_ADDR_HI word*/ +/*define for src_addr_63_32 field*/ +#define SDMA_PKT_COPY_LINEAR_SUBWIN_BC_SRC_ADDR_HI_src_addr_63_32_offset 2 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_BC_SRC_ADDR_HI_src_addr_63_32_mask 0xFFFFFFFF +#define SDMA_PKT_COPY_LINEAR_SUBWIN_BC_SRC_ADDR_HI_src_addr_63_32_shift 0 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_BC_SRC_ADDR_HI_SRC_ADDR_63_32(x) (((x) & SDMA_PKT_COPY_LINEAR_SUBWIN_BC_SRC_ADDR_HI_src_addr_63_32_mask) << SDMA_PKT_COPY_LINEAR_SUBWIN_BC_SRC_ADDR_HI_src_addr_63_32_shift) + +/*define for DW_3 word*/ +/*define for src_x field*/ +#define SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DW_3_src_x_offset 3 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DW_3_src_x_mask 0x00003FFF +#define SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DW_3_src_x_shift 0 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DW_3_SRC_X(x) (((x) & SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DW_3_src_x_mask) << SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DW_3_src_x_shift) + +/*define for src_y field*/ +#define SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DW_3_src_y_offset 3 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DW_3_src_y_mask 0x00003FFF +#define SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DW_3_src_y_shift 16 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DW_3_SRC_Y(x) (((x) & SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DW_3_src_y_mask) << SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DW_3_src_y_shift) + +/*define for DW_4 word*/ +/*define for src_z field*/ +#define SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DW_4_src_z_offset 4 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DW_4_src_z_mask 0x000007FF +#define SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DW_4_src_z_shift 0 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DW_4_SRC_Z(x) (((x) & SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DW_4_src_z_mask) << SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DW_4_src_z_shift) + +/*define for src_pitch field*/ +#define SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DW_4_src_pitch_offset 4 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DW_4_src_pitch_mask 0x00003FFF +#define SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DW_4_src_pitch_shift 13 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DW_4_SRC_PITCH(x) (((x) & SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DW_4_src_pitch_mask) << SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DW_4_src_pitch_shift) + +/*define for DW_5 word*/ +/*define for src_slice_pitch field*/ +#define SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DW_5_src_slice_pitch_offset 5 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DW_5_src_slice_pitch_mask 0x0FFFFFFF +#define SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DW_5_src_slice_pitch_shift 0 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DW_5_SRC_SLICE_PITCH(x) (((x) & SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DW_5_src_slice_pitch_mask) << SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DW_5_src_slice_pitch_shift) + +/*define for DST_ADDR_LO word*/ +/*define for dst_addr_31_0 field*/ +#define SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DST_ADDR_LO_dst_addr_31_0_offset 6 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DST_ADDR_LO_dst_addr_31_0_mask 0xFFFFFFFF +#define SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DST_ADDR_LO_dst_addr_31_0_shift 0 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DST_ADDR_LO_DST_ADDR_31_0(x) (((x) & SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DST_ADDR_LO_dst_addr_31_0_mask) << SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DST_ADDR_LO_dst_addr_31_0_shift) + +/*define for DST_ADDR_HI word*/ +/*define for dst_addr_63_32 field*/ +#define SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DST_ADDR_HI_dst_addr_63_32_offset 7 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DST_ADDR_HI_dst_addr_63_32_mask 0xFFFFFFFF +#define SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DST_ADDR_HI_dst_addr_63_32_shift 0 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DST_ADDR_HI_DST_ADDR_63_32(x) (((x) & SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DST_ADDR_HI_dst_addr_63_32_mask) << SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DST_ADDR_HI_dst_addr_63_32_shift) + +/*define for DW_8 word*/ +/*define for dst_x field*/ +#define SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DW_8_dst_x_offset 8 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DW_8_dst_x_mask 0x00003FFF +#define SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DW_8_dst_x_shift 0 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DW_8_DST_X(x) (((x) & SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DW_8_dst_x_mask) << SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DW_8_dst_x_shift) + +/*define for dst_y field*/ +#define SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DW_8_dst_y_offset 8 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DW_8_dst_y_mask 0x00003FFF +#define SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DW_8_dst_y_shift 16 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DW_8_DST_Y(x) (((x) & SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DW_8_dst_y_mask) << SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DW_8_dst_y_shift) + +/*define for DW_9 word*/ +/*define for dst_z field*/ +#define SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DW_9_dst_z_offset 9 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DW_9_dst_z_mask 0x000007FF +#define SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DW_9_dst_z_shift 0 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DW_9_DST_Z(x) (((x) & SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DW_9_dst_z_mask) << SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DW_9_dst_z_shift) + +/*define for dst_pitch field*/ +#define SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DW_9_dst_pitch_offset 9 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DW_9_dst_pitch_mask 0x00003FFF +#define SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DW_9_dst_pitch_shift 13 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DW_9_DST_PITCH(x) (((x) & SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DW_9_dst_pitch_mask) << SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DW_9_dst_pitch_shift) + +/*define for DW_10 word*/ +/*define for dst_slice_pitch field*/ +#define SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DW_10_dst_slice_pitch_offset 10 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DW_10_dst_slice_pitch_mask 0x0FFFFFFF +#define SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DW_10_dst_slice_pitch_shift 0 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DW_10_DST_SLICE_PITCH(x) (((x) & SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DW_10_dst_slice_pitch_mask) << SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DW_10_dst_slice_pitch_shift) + +/*define for DW_11 word*/ +/*define for rect_x field*/ +#define SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DW_11_rect_x_offset 11 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DW_11_rect_x_mask 0x00003FFF +#define SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DW_11_rect_x_shift 0 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DW_11_RECT_X(x) (((x) & SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DW_11_rect_x_mask) << SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DW_11_rect_x_shift) + +/*define for rect_y field*/ +#define SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DW_11_rect_y_offset 11 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DW_11_rect_y_mask 0x00003FFF +#define SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DW_11_rect_y_shift 16 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DW_11_RECT_Y(x) (((x) & SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DW_11_rect_y_mask) << SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DW_11_rect_y_shift) + +/*define for DW_12 word*/ +/*define for rect_z field*/ +#define SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DW_12_rect_z_offset 12 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DW_12_rect_z_mask 0x000007FF +#define SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DW_12_rect_z_shift 0 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DW_12_RECT_Z(x) (((x) & SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DW_12_rect_z_mask) << SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DW_12_rect_z_shift) + +/*define for dst_sw field*/ +#define SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DW_12_dst_sw_offset 12 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DW_12_dst_sw_mask 0x00000003 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DW_12_dst_sw_shift 16 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DW_12_DST_SW(x) (((x) & SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DW_12_dst_sw_mask) << SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DW_12_dst_sw_shift) + +/*define for dst_ha field*/ +#define SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DW_12_dst_ha_offset 12 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DW_12_dst_ha_mask 0x00000001 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DW_12_dst_ha_shift 19 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DW_12_DST_HA(x) (((x) & SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DW_12_dst_ha_mask) << SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DW_12_dst_ha_shift) + +/*define for src_sw field*/ +#define SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DW_12_src_sw_offset 12 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DW_12_src_sw_mask 0x00000003 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DW_12_src_sw_shift 24 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DW_12_SRC_SW(x) (((x) & SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DW_12_src_sw_mask) << SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DW_12_src_sw_shift) + +/*define for src_ha field*/ +#define SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DW_12_src_ha_offset 12 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DW_12_src_ha_mask 0x00000001 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DW_12_src_ha_shift 27 +#define SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DW_12_SRC_HA(x) (((x) & SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DW_12_src_ha_mask) << SDMA_PKT_COPY_LINEAR_SUBWIN_BC_DW_12_src_ha_shift) + + +/* +** Definitions for SDMA_PKT_COPY_TILED packet +*/ + +/*define for HEADER word*/ +/*define for op field*/ +#define SDMA_PKT_COPY_TILED_HEADER_op_offset 0 +#define SDMA_PKT_COPY_TILED_HEADER_op_mask 0x000000FF +#define SDMA_PKT_COPY_TILED_HEADER_op_shift 0 +#define SDMA_PKT_COPY_TILED_HEADER_OP(x) (((x) & SDMA_PKT_COPY_TILED_HEADER_op_mask) << SDMA_PKT_COPY_TILED_HEADER_op_shift) + +/*define for sub_op field*/ +#define SDMA_PKT_COPY_TILED_HEADER_sub_op_offset 0 +#define SDMA_PKT_COPY_TILED_HEADER_sub_op_mask 0x000000FF +#define SDMA_PKT_COPY_TILED_HEADER_sub_op_shift 8 +#define SDMA_PKT_COPY_TILED_HEADER_SUB_OP(x) (((x) & SDMA_PKT_COPY_TILED_HEADER_sub_op_mask) << SDMA_PKT_COPY_TILED_HEADER_sub_op_shift) + +/*define for encrypt field*/ +#define SDMA_PKT_COPY_TILED_HEADER_encrypt_offset 0 +#define SDMA_PKT_COPY_TILED_HEADER_encrypt_mask 0x00000001 +#define SDMA_PKT_COPY_TILED_HEADER_encrypt_shift 16 +#define SDMA_PKT_COPY_TILED_HEADER_ENCRYPT(x) (((x) & SDMA_PKT_COPY_TILED_HEADER_encrypt_mask) << SDMA_PKT_COPY_TILED_HEADER_encrypt_shift) + +/*define for tmz field*/ +#define SDMA_PKT_COPY_TILED_HEADER_tmz_offset 0 +#define SDMA_PKT_COPY_TILED_HEADER_tmz_mask 0x00000001 +#define SDMA_PKT_COPY_TILED_HEADER_tmz_shift 18 +#define SDMA_PKT_COPY_TILED_HEADER_TMZ(x) (((x) & SDMA_PKT_COPY_TILED_HEADER_tmz_mask) << SDMA_PKT_COPY_TILED_HEADER_tmz_shift) + +/*define for cpv field*/ +#define SDMA_PKT_COPY_TILED_HEADER_cpv_offset 0 +#define SDMA_PKT_COPY_TILED_HEADER_cpv_mask 0x00000001 +#define SDMA_PKT_COPY_TILED_HEADER_cpv_shift 19 +#define SDMA_PKT_COPY_TILED_HEADER_CPV(x) (((x) & SDMA_PKT_COPY_TILED_HEADER_cpv_mask) << SDMA_PKT_COPY_TILED_HEADER_cpv_shift) + +/*define for detile field*/ +#define SDMA_PKT_COPY_TILED_HEADER_detile_offset 0 +#define SDMA_PKT_COPY_TILED_HEADER_detile_mask 0x00000001 +#define SDMA_PKT_COPY_TILED_HEADER_detile_shift 31 +#define SDMA_PKT_COPY_TILED_HEADER_DETILE(x) (((x) & SDMA_PKT_COPY_TILED_HEADER_detile_mask) << SDMA_PKT_COPY_TILED_HEADER_detile_shift) + +/*define for TILED_ADDR_LO word*/ +/*define for tiled_addr_31_0 field*/ +#define SDMA_PKT_COPY_TILED_TILED_ADDR_LO_tiled_addr_31_0_offset 1 +#define SDMA_PKT_COPY_TILED_TILED_ADDR_LO_tiled_addr_31_0_mask 0xFFFFFFFF +#define SDMA_PKT_COPY_TILED_TILED_ADDR_LO_tiled_addr_31_0_shift 0 +#define SDMA_PKT_COPY_TILED_TILED_ADDR_LO_TILED_ADDR_31_0(x) (((x) & SDMA_PKT_COPY_TILED_TILED_ADDR_LO_tiled_addr_31_0_mask) << SDMA_PKT_COPY_TILED_TILED_ADDR_LO_tiled_addr_31_0_shift) + +/*define for TILED_ADDR_HI word*/ +/*define for tiled_addr_63_32 field*/ +#define SDMA_PKT_COPY_TILED_TILED_ADDR_HI_tiled_addr_63_32_offset 2 +#define SDMA_PKT_COPY_TILED_TILED_ADDR_HI_tiled_addr_63_32_mask 0xFFFFFFFF +#define SDMA_PKT_COPY_TILED_TILED_ADDR_HI_tiled_addr_63_32_shift 0 +#define SDMA_PKT_COPY_TILED_TILED_ADDR_HI_TILED_ADDR_63_32(x) (((x) & SDMA_PKT_COPY_TILED_TILED_ADDR_HI_tiled_addr_63_32_mask) << SDMA_PKT_COPY_TILED_TILED_ADDR_HI_tiled_addr_63_32_shift) + +/*define for DW_3 word*/ +/*define for width field*/ +#define SDMA_PKT_COPY_TILED_DW_3_width_offset 3 +#define SDMA_PKT_COPY_TILED_DW_3_width_mask 0x00003FFF +#define SDMA_PKT_COPY_TILED_DW_3_width_shift 0 +#define SDMA_PKT_COPY_TILED_DW_3_WIDTH(x) (((x) & SDMA_PKT_COPY_TILED_DW_3_width_mask) << SDMA_PKT_COPY_TILED_DW_3_width_shift) + +/*define for DW_4 word*/ +/*define for height field*/ +#define SDMA_PKT_COPY_TILED_DW_4_height_offset 4 +#define SDMA_PKT_COPY_TILED_DW_4_height_mask 0x00003FFF +#define SDMA_PKT_COPY_TILED_DW_4_height_shift 0 +#define SDMA_PKT_COPY_TILED_DW_4_HEIGHT(x) (((x) & SDMA_PKT_COPY_TILED_DW_4_height_mask) << SDMA_PKT_COPY_TILED_DW_4_height_shift) + +/*define for depth field*/ +#define SDMA_PKT_COPY_TILED_DW_4_depth_offset 4 +#define SDMA_PKT_COPY_TILED_DW_4_depth_mask 0x00001FFF +#define SDMA_PKT_COPY_TILED_DW_4_depth_shift 16 +#define SDMA_PKT_COPY_TILED_DW_4_DEPTH(x) (((x) & SDMA_PKT_COPY_TILED_DW_4_depth_mask) << SDMA_PKT_COPY_TILED_DW_4_depth_shift) + +/*define for DW_5 word*/ +/*define for element_size field*/ +#define SDMA_PKT_COPY_TILED_DW_5_element_size_offset 5 +#define SDMA_PKT_COPY_TILED_DW_5_element_size_mask 0x00000007 +#define SDMA_PKT_COPY_TILED_DW_5_element_size_shift 0 +#define SDMA_PKT_COPY_TILED_DW_5_ELEMENT_SIZE(x) (((x) & SDMA_PKT_COPY_TILED_DW_5_element_size_mask) << SDMA_PKT_COPY_TILED_DW_5_element_size_shift) + +/*define for swizzle_mode field*/ +#define SDMA_PKT_COPY_TILED_DW_5_swizzle_mode_offset 5 +#define SDMA_PKT_COPY_TILED_DW_5_swizzle_mode_mask 0x0000001F +#define SDMA_PKT_COPY_TILED_DW_5_swizzle_mode_shift 3 +#define SDMA_PKT_COPY_TILED_DW_5_SWIZZLE_MODE(x) (((x) & SDMA_PKT_COPY_TILED_DW_5_swizzle_mode_mask) << SDMA_PKT_COPY_TILED_DW_5_swizzle_mode_shift) + +/*define for dimension field*/ +#define SDMA_PKT_COPY_TILED_DW_5_dimension_offset 5 +#define SDMA_PKT_COPY_TILED_DW_5_dimension_mask 0x00000003 +#define SDMA_PKT_COPY_TILED_DW_5_dimension_shift 9 +#define SDMA_PKT_COPY_TILED_DW_5_DIMENSION(x) (((x) & SDMA_PKT_COPY_TILED_DW_5_dimension_mask) << SDMA_PKT_COPY_TILED_DW_5_dimension_shift) + +/*define for mip_max field*/ +#define SDMA_PKT_COPY_TILED_DW_5_mip_max_offset 5 +#define SDMA_PKT_COPY_TILED_DW_5_mip_max_mask 0x0000000F +#define SDMA_PKT_COPY_TILED_DW_5_mip_max_shift 16 +#define SDMA_PKT_COPY_TILED_DW_5_MIP_MAX(x) (((x) & SDMA_PKT_COPY_TILED_DW_5_mip_max_mask) << SDMA_PKT_COPY_TILED_DW_5_mip_max_shift) + +/*define for DW_6 word*/ +/*define for x field*/ +#define SDMA_PKT_COPY_TILED_DW_6_x_offset 6 +#define SDMA_PKT_COPY_TILED_DW_6_x_mask 0x00003FFF +#define SDMA_PKT_COPY_TILED_DW_6_x_shift 0 +#define SDMA_PKT_COPY_TILED_DW_6_X(x) (((x) & SDMA_PKT_COPY_TILED_DW_6_x_mask) << SDMA_PKT_COPY_TILED_DW_6_x_shift) + +/*define for y field*/ +#define SDMA_PKT_COPY_TILED_DW_6_y_offset 6 +#define SDMA_PKT_COPY_TILED_DW_6_y_mask 0x00003FFF +#define SDMA_PKT_COPY_TILED_DW_6_y_shift 16 +#define SDMA_PKT_COPY_TILED_DW_6_Y(x) (((x) & SDMA_PKT_COPY_TILED_DW_6_y_mask) << SDMA_PKT_COPY_TILED_DW_6_y_shift) + +/*define for DW_7 word*/ +/*define for z field*/ +#define SDMA_PKT_COPY_TILED_DW_7_z_offset 7 +#define SDMA_PKT_COPY_TILED_DW_7_z_mask 0x00001FFF +#define SDMA_PKT_COPY_TILED_DW_7_z_shift 0 +#define SDMA_PKT_COPY_TILED_DW_7_Z(x) (((x) & SDMA_PKT_COPY_TILED_DW_7_z_mask) << SDMA_PKT_COPY_TILED_DW_7_z_shift) + +/*define for linear_sw field*/ +#define SDMA_PKT_COPY_TILED_DW_7_linear_sw_offset 7 +#define SDMA_PKT_COPY_TILED_DW_7_linear_sw_mask 0x00000003 +#define SDMA_PKT_COPY_TILED_DW_7_linear_sw_shift 16 +#define SDMA_PKT_COPY_TILED_DW_7_LINEAR_SW(x) (((x) & SDMA_PKT_COPY_TILED_DW_7_linear_sw_mask) << SDMA_PKT_COPY_TILED_DW_7_linear_sw_shift) + +/*define for linear_cache_policy field*/ +#define SDMA_PKT_COPY_TILED_DW_7_linear_cache_policy_offset 7 +#define SDMA_PKT_COPY_TILED_DW_7_linear_cache_policy_mask 0x00000007 +#define SDMA_PKT_COPY_TILED_DW_7_linear_cache_policy_shift 18 +#define SDMA_PKT_COPY_TILED_DW_7_LINEAR_CACHE_POLICY(x) (((x) & SDMA_PKT_COPY_TILED_DW_7_linear_cache_policy_mask) << SDMA_PKT_COPY_TILED_DW_7_linear_cache_policy_shift) + +/*define for tile_sw field*/ +#define SDMA_PKT_COPY_TILED_DW_7_tile_sw_offset 7 +#define SDMA_PKT_COPY_TILED_DW_7_tile_sw_mask 0x00000003 +#define SDMA_PKT_COPY_TILED_DW_7_tile_sw_shift 24 +#define SDMA_PKT_COPY_TILED_DW_7_TILE_SW(x) (((x) & SDMA_PKT_COPY_TILED_DW_7_tile_sw_mask) << SDMA_PKT_COPY_TILED_DW_7_tile_sw_shift) + +/*define for tile_cache_policy field*/ +#define SDMA_PKT_COPY_TILED_DW_7_tile_cache_policy_offset 7 +#define SDMA_PKT_COPY_TILED_DW_7_tile_cache_policy_mask 0x00000007 +#define SDMA_PKT_COPY_TILED_DW_7_tile_cache_policy_shift 26 +#define SDMA_PKT_COPY_TILED_DW_7_TILE_CACHE_POLICY(x) (((x) & SDMA_PKT_COPY_TILED_DW_7_tile_cache_policy_mask) << SDMA_PKT_COPY_TILED_DW_7_tile_cache_policy_shift) + +/*define for LINEAR_ADDR_LO word*/ +/*define for linear_addr_31_0 field*/ +#define SDMA_PKT_COPY_TILED_LINEAR_ADDR_LO_linear_addr_31_0_offset 8 +#define SDMA_PKT_COPY_TILED_LINEAR_ADDR_LO_linear_addr_31_0_mask 0xFFFFFFFF +#define SDMA_PKT_COPY_TILED_LINEAR_ADDR_LO_linear_addr_31_0_shift 0 +#define SDMA_PKT_COPY_TILED_LINEAR_ADDR_LO_LINEAR_ADDR_31_0(x) (((x) & SDMA_PKT_COPY_TILED_LINEAR_ADDR_LO_linear_addr_31_0_mask) << SDMA_PKT_COPY_TILED_LINEAR_ADDR_LO_linear_addr_31_0_shift) + +/*define for LINEAR_ADDR_HI word*/ +/*define for linear_addr_63_32 field*/ +#define SDMA_PKT_COPY_TILED_LINEAR_ADDR_HI_linear_addr_63_32_offset 9 +#define SDMA_PKT_COPY_TILED_LINEAR_ADDR_HI_linear_addr_63_32_mask 0xFFFFFFFF +#define SDMA_PKT_COPY_TILED_LINEAR_ADDR_HI_linear_addr_63_32_shift 0 +#define SDMA_PKT_COPY_TILED_LINEAR_ADDR_HI_LINEAR_ADDR_63_32(x) (((x) & SDMA_PKT_COPY_TILED_LINEAR_ADDR_HI_linear_addr_63_32_mask) << SDMA_PKT_COPY_TILED_LINEAR_ADDR_HI_linear_addr_63_32_shift) + +/*define for LINEAR_PITCH word*/ +/*define for linear_pitch field*/ +#define SDMA_PKT_COPY_TILED_LINEAR_PITCH_linear_pitch_offset 10 +#define SDMA_PKT_COPY_TILED_LINEAR_PITCH_linear_pitch_mask 0x0007FFFF +#define SDMA_PKT_COPY_TILED_LINEAR_PITCH_linear_pitch_shift 0 +#define SDMA_PKT_COPY_TILED_LINEAR_PITCH_LINEAR_PITCH(x) (((x) & SDMA_PKT_COPY_TILED_LINEAR_PITCH_linear_pitch_mask) << SDMA_PKT_COPY_TILED_LINEAR_PITCH_linear_pitch_shift) + +/*define for LINEAR_SLICE_PITCH word*/ +/*define for linear_slice_pitch field*/ +#define SDMA_PKT_COPY_TILED_LINEAR_SLICE_PITCH_linear_slice_pitch_offset 11 +#define SDMA_PKT_COPY_TILED_LINEAR_SLICE_PITCH_linear_slice_pitch_mask 0xFFFFFFFF +#define SDMA_PKT_COPY_TILED_LINEAR_SLICE_PITCH_linear_slice_pitch_shift 0 +#define SDMA_PKT_COPY_TILED_LINEAR_SLICE_PITCH_LINEAR_SLICE_PITCH(x) (((x) & SDMA_PKT_COPY_TILED_LINEAR_SLICE_PITCH_linear_slice_pitch_mask) << SDMA_PKT_COPY_TILED_LINEAR_SLICE_PITCH_linear_slice_pitch_shift) + +/*define for COUNT word*/ +/*define for count field*/ +#define SDMA_PKT_COPY_TILED_COUNT_count_offset 12 +#define SDMA_PKT_COPY_TILED_COUNT_count_mask 0x3FFFFFFF +#define SDMA_PKT_COPY_TILED_COUNT_count_shift 0 +#define SDMA_PKT_COPY_TILED_COUNT_COUNT(x) (((x) & SDMA_PKT_COPY_TILED_COUNT_count_mask) << SDMA_PKT_COPY_TILED_COUNT_count_shift) + + +/* +** Definitions for SDMA_PKT_COPY_TILED_BC packet +*/ + +/*define for HEADER word*/ +/*define for op field*/ +#define SDMA_PKT_COPY_TILED_BC_HEADER_op_offset 0 +#define SDMA_PKT_COPY_TILED_BC_HEADER_op_mask 0x000000FF +#define SDMA_PKT_COPY_TILED_BC_HEADER_op_shift 0 +#define SDMA_PKT_COPY_TILED_BC_HEADER_OP(x) (((x) & SDMA_PKT_COPY_TILED_BC_HEADER_op_mask) << SDMA_PKT_COPY_TILED_BC_HEADER_op_shift) + +/*define for sub_op field*/ +#define SDMA_PKT_COPY_TILED_BC_HEADER_sub_op_offset 0 +#define SDMA_PKT_COPY_TILED_BC_HEADER_sub_op_mask 0x000000FF +#define SDMA_PKT_COPY_TILED_BC_HEADER_sub_op_shift 8 +#define SDMA_PKT_COPY_TILED_BC_HEADER_SUB_OP(x) (((x) & SDMA_PKT_COPY_TILED_BC_HEADER_sub_op_mask) << SDMA_PKT_COPY_TILED_BC_HEADER_sub_op_shift) + +/*define for detile field*/ +#define SDMA_PKT_COPY_TILED_BC_HEADER_detile_offset 0 +#define SDMA_PKT_COPY_TILED_BC_HEADER_detile_mask 0x00000001 +#define SDMA_PKT_COPY_TILED_BC_HEADER_detile_shift 31 +#define SDMA_PKT_COPY_TILED_BC_HEADER_DETILE(x) (((x) & SDMA_PKT_COPY_TILED_BC_HEADER_detile_mask) << SDMA_PKT_COPY_TILED_BC_HEADER_detile_shift) + +/*define for TILED_ADDR_LO word*/ +/*define for tiled_addr_31_0 field*/ +#define SDMA_PKT_COPY_TILED_BC_TILED_ADDR_LO_tiled_addr_31_0_offset 1 +#define SDMA_PKT_COPY_TILED_BC_TILED_ADDR_LO_tiled_addr_31_0_mask 0xFFFFFFFF +#define SDMA_PKT_COPY_TILED_BC_TILED_ADDR_LO_tiled_addr_31_0_shift 0 +#define SDMA_PKT_COPY_TILED_BC_TILED_ADDR_LO_TILED_ADDR_31_0(x) (((x) & SDMA_PKT_COPY_TILED_BC_TILED_ADDR_LO_tiled_addr_31_0_mask) << SDMA_PKT_COPY_TILED_BC_TILED_ADDR_LO_tiled_addr_31_0_shift) + +/*define for TILED_ADDR_HI word*/ +/*define for tiled_addr_63_32 field*/ +#define SDMA_PKT_COPY_TILED_BC_TILED_ADDR_HI_tiled_addr_63_32_offset 2 +#define SDMA_PKT_COPY_TILED_BC_TILED_ADDR_HI_tiled_addr_63_32_mask 0xFFFFFFFF +#define SDMA_PKT_COPY_TILED_BC_TILED_ADDR_HI_tiled_addr_63_32_shift 0 +#define SDMA_PKT_COPY_TILED_BC_TILED_ADDR_HI_TILED_ADDR_63_32(x) (((x) & SDMA_PKT_COPY_TILED_BC_TILED_ADDR_HI_tiled_addr_63_32_mask) << SDMA_PKT_COPY_TILED_BC_TILED_ADDR_HI_tiled_addr_63_32_shift) + +/*define for DW_3 word*/ +/*define for width field*/ +#define SDMA_PKT_COPY_TILED_BC_DW_3_width_offset 3 +#define SDMA_PKT_COPY_TILED_BC_DW_3_width_mask 0x00003FFF +#define SDMA_PKT_COPY_TILED_BC_DW_3_width_shift 0 +#define SDMA_PKT_COPY_TILED_BC_DW_3_WIDTH(x) (((x) & SDMA_PKT_COPY_TILED_BC_DW_3_width_mask) << SDMA_PKT_COPY_TILED_BC_DW_3_width_shift) + +/*define for DW_4 word*/ +/*define for height field*/ +#define SDMA_PKT_COPY_TILED_BC_DW_4_height_offset 4 +#define SDMA_PKT_COPY_TILED_BC_DW_4_height_mask 0x00003FFF +#define SDMA_PKT_COPY_TILED_BC_DW_4_height_shift 0 +#define SDMA_PKT_COPY_TILED_BC_DW_4_HEIGHT(x) (((x) & SDMA_PKT_COPY_TILED_BC_DW_4_height_mask) << SDMA_PKT_COPY_TILED_BC_DW_4_height_shift) + +/*define for depth field*/ +#define SDMA_PKT_COPY_TILED_BC_DW_4_depth_offset 4 +#define SDMA_PKT_COPY_TILED_BC_DW_4_depth_mask 0x000007FF +#define SDMA_PKT_COPY_TILED_BC_DW_4_depth_shift 16 +#define SDMA_PKT_COPY_TILED_BC_DW_4_DEPTH(x) (((x) & SDMA_PKT_COPY_TILED_BC_DW_4_depth_mask) << SDMA_PKT_COPY_TILED_BC_DW_4_depth_shift) + +/*define for DW_5 word*/ +/*define for element_size field*/ +#define SDMA_PKT_COPY_TILED_BC_DW_5_element_size_offset 5 +#define SDMA_PKT_COPY_TILED_BC_DW_5_element_size_mask 0x00000007 +#define SDMA_PKT_COPY_TILED_BC_DW_5_element_size_shift 0 +#define SDMA_PKT_COPY_TILED_BC_DW_5_ELEMENT_SIZE(x) (((x) & SDMA_PKT_COPY_TILED_BC_DW_5_element_size_mask) << SDMA_PKT_COPY_TILED_BC_DW_5_element_size_shift) + +/*define for array_mode field*/ +#define SDMA_PKT_COPY_TILED_BC_DW_5_array_mode_offset 5 +#define SDMA_PKT_COPY_TILED_BC_DW_5_array_mode_mask 0x0000000F +#define SDMA_PKT_COPY_TILED_BC_DW_5_array_mode_shift 3 +#define SDMA_PKT_COPY_TILED_BC_DW_5_ARRAY_MODE(x) (((x) & SDMA_PKT_COPY_TILED_BC_DW_5_array_mode_mask) << SDMA_PKT_COPY_TILED_BC_DW_5_array_mode_shift) + +/*define for mit_mode field*/ +#define SDMA_PKT_COPY_TILED_BC_DW_5_mit_mode_offset 5 +#define SDMA_PKT_COPY_TILED_BC_DW_5_mit_mode_mask 0x00000007 +#define SDMA_PKT_COPY_TILED_BC_DW_5_mit_mode_shift 8 +#define SDMA_PKT_COPY_TILED_BC_DW_5_MIT_MODE(x) (((x) & SDMA_PKT_COPY_TILED_BC_DW_5_mit_mode_mask) << SDMA_PKT_COPY_TILED_BC_DW_5_mit_mode_shift) + +/*define for tilesplit_size field*/ +#define SDMA_PKT_COPY_TILED_BC_DW_5_tilesplit_size_offset 5 +#define SDMA_PKT_COPY_TILED_BC_DW_5_tilesplit_size_mask 0x00000007 +#define SDMA_PKT_COPY_TILED_BC_DW_5_tilesplit_size_shift 11 +#define SDMA_PKT_COPY_TILED_BC_DW_5_TILESPLIT_SIZE(x) (((x) & SDMA_PKT_COPY_TILED_BC_DW_5_tilesplit_size_mask) << SDMA_PKT_COPY_TILED_BC_DW_5_tilesplit_size_shift) + +/*define for bank_w field*/ +#define SDMA_PKT_COPY_TILED_BC_DW_5_bank_w_offset 5 +#define SDMA_PKT_COPY_TILED_BC_DW_5_bank_w_mask 0x00000003 +#define SDMA_PKT_COPY_TILED_BC_DW_5_bank_w_shift 15 +#define SDMA_PKT_COPY_TILED_BC_DW_5_BANK_W(x) (((x) & SDMA_PKT_COPY_TILED_BC_DW_5_bank_w_mask) << SDMA_PKT_COPY_TILED_BC_DW_5_bank_w_shift) + +/*define for bank_h field*/ +#define SDMA_PKT_COPY_TILED_BC_DW_5_bank_h_offset 5 +#define SDMA_PKT_COPY_TILED_BC_DW_5_bank_h_mask 0x00000003 +#define SDMA_PKT_COPY_TILED_BC_DW_5_bank_h_shift 18 +#define SDMA_PKT_COPY_TILED_BC_DW_5_BANK_H(x) (((x) & SDMA_PKT_COPY_TILED_BC_DW_5_bank_h_mask) << SDMA_PKT_COPY_TILED_BC_DW_5_bank_h_shift) + +/*define for num_bank field*/ +#define SDMA_PKT_COPY_TILED_BC_DW_5_num_bank_offset 5 +#define SDMA_PKT_COPY_TILED_BC_DW_5_num_bank_mask 0x00000003 +#define SDMA_PKT_COPY_TILED_BC_DW_5_num_bank_shift 21 +#define SDMA_PKT_COPY_TILED_BC_DW_5_NUM_BANK(x) (((x) & SDMA_PKT_COPY_TILED_BC_DW_5_num_bank_mask) << SDMA_PKT_COPY_TILED_BC_DW_5_num_bank_shift) + +/*define for mat_aspt field*/ +#define SDMA_PKT_COPY_TILED_BC_DW_5_mat_aspt_offset 5 +#define SDMA_PKT_COPY_TILED_BC_DW_5_mat_aspt_mask 0x00000003 +#define SDMA_PKT_COPY_TILED_BC_DW_5_mat_aspt_shift 24 +#define SDMA_PKT_COPY_TILED_BC_DW_5_MAT_ASPT(x) (((x) & SDMA_PKT_COPY_TILED_BC_DW_5_mat_aspt_mask) << SDMA_PKT_COPY_TILED_BC_DW_5_mat_aspt_shift) + +/*define for pipe_config field*/ +#define SDMA_PKT_COPY_TILED_BC_DW_5_pipe_config_offset 5 +#define SDMA_PKT_COPY_TILED_BC_DW_5_pipe_config_mask 0x0000001F +#define SDMA_PKT_COPY_TILED_BC_DW_5_pipe_config_shift 26 +#define SDMA_PKT_COPY_TILED_BC_DW_5_PIPE_CONFIG(x) (((x) & SDMA_PKT_COPY_TILED_BC_DW_5_pipe_config_mask) << SDMA_PKT_COPY_TILED_BC_DW_5_pipe_config_shift) + +/*define for DW_6 word*/ +/*define for x field*/ +#define SDMA_PKT_COPY_TILED_BC_DW_6_x_offset 6 +#define SDMA_PKT_COPY_TILED_BC_DW_6_x_mask 0x00003FFF +#define SDMA_PKT_COPY_TILED_BC_DW_6_x_shift 0 +#define SDMA_PKT_COPY_TILED_BC_DW_6_X(x) (((x) & SDMA_PKT_COPY_TILED_BC_DW_6_x_mask) << SDMA_PKT_COPY_TILED_BC_DW_6_x_shift) + +/*define for y field*/ +#define SDMA_PKT_COPY_TILED_BC_DW_6_y_offset 6 +#define SDMA_PKT_COPY_TILED_BC_DW_6_y_mask 0x00003FFF +#define SDMA_PKT_COPY_TILED_BC_DW_6_y_shift 16 +#define SDMA_PKT_COPY_TILED_BC_DW_6_Y(x) (((x) & SDMA_PKT_COPY_TILED_BC_DW_6_y_mask) << SDMA_PKT_COPY_TILED_BC_DW_6_y_shift) + +/*define for DW_7 word*/ +/*define for z field*/ +#define SDMA_PKT_COPY_TILED_BC_DW_7_z_offset 7 +#define SDMA_PKT_COPY_TILED_BC_DW_7_z_mask 0x000007FF +#define SDMA_PKT_COPY_TILED_BC_DW_7_z_shift 0 +#define SDMA_PKT_COPY_TILED_BC_DW_7_Z(x) (((x) & SDMA_PKT_COPY_TILED_BC_DW_7_z_mask) << SDMA_PKT_COPY_TILED_BC_DW_7_z_shift) + +/*define for linear_sw field*/ +#define SDMA_PKT_COPY_TILED_BC_DW_7_linear_sw_offset 7 +#define SDMA_PKT_COPY_TILED_BC_DW_7_linear_sw_mask 0x00000003 +#define SDMA_PKT_COPY_TILED_BC_DW_7_linear_sw_shift 16 +#define SDMA_PKT_COPY_TILED_BC_DW_7_LINEAR_SW(x) (((x) & SDMA_PKT_COPY_TILED_BC_DW_7_linear_sw_mask) << SDMA_PKT_COPY_TILED_BC_DW_7_linear_sw_shift) + +/*define for tile_sw field*/ +#define SDMA_PKT_COPY_TILED_BC_DW_7_tile_sw_offset 7 +#define SDMA_PKT_COPY_TILED_BC_DW_7_tile_sw_mask 0x00000003 +#define SDMA_PKT_COPY_TILED_BC_DW_7_tile_sw_shift 24 +#define SDMA_PKT_COPY_TILED_BC_DW_7_TILE_SW(x) (((x) & SDMA_PKT_COPY_TILED_BC_DW_7_tile_sw_mask) << SDMA_PKT_COPY_TILED_BC_DW_7_tile_sw_shift) + +/*define for LINEAR_ADDR_LO word*/ +/*define for linear_addr_31_0 field*/ +#define SDMA_PKT_COPY_TILED_BC_LINEAR_ADDR_LO_linear_addr_31_0_offset 8 +#define SDMA_PKT_COPY_TILED_BC_LINEAR_ADDR_LO_linear_addr_31_0_mask 0xFFFFFFFF +#define SDMA_PKT_COPY_TILED_BC_LINEAR_ADDR_LO_linear_addr_31_0_shift 0 +#define SDMA_PKT_COPY_TILED_BC_LINEAR_ADDR_LO_LINEAR_ADDR_31_0(x) (((x) & SDMA_PKT_COPY_TILED_BC_LINEAR_ADDR_LO_linear_addr_31_0_mask) << SDMA_PKT_COPY_TILED_BC_LINEAR_ADDR_LO_linear_addr_31_0_shift) + +/*define for LINEAR_ADDR_HI word*/ +/*define for linear_addr_63_32 field*/ +#define SDMA_PKT_COPY_TILED_BC_LINEAR_ADDR_HI_linear_addr_63_32_offset 9 +#define SDMA_PKT_COPY_TILED_BC_LINEAR_ADDR_HI_linear_addr_63_32_mask 0xFFFFFFFF +#define SDMA_PKT_COPY_TILED_BC_LINEAR_ADDR_HI_linear_addr_63_32_shift 0 +#define SDMA_PKT_COPY_TILED_BC_LINEAR_ADDR_HI_LINEAR_ADDR_63_32(x) (((x) & SDMA_PKT_COPY_TILED_BC_LINEAR_ADDR_HI_linear_addr_63_32_mask) << SDMA_PKT_COPY_TILED_BC_LINEAR_ADDR_HI_linear_addr_63_32_shift) + +/*define for LINEAR_PITCH word*/ +/*define for linear_pitch field*/ +#define SDMA_PKT_COPY_TILED_BC_LINEAR_PITCH_linear_pitch_offset 10 +#define SDMA_PKT_COPY_TILED_BC_LINEAR_PITCH_linear_pitch_mask 0x0007FFFF +#define SDMA_PKT_COPY_TILED_BC_LINEAR_PITCH_linear_pitch_shift 0 +#define SDMA_PKT_COPY_TILED_BC_LINEAR_PITCH_LINEAR_PITCH(x) (((x) & SDMA_PKT_COPY_TILED_BC_LINEAR_PITCH_linear_pitch_mask) << SDMA_PKT_COPY_TILED_BC_LINEAR_PITCH_linear_pitch_shift) + +/*define for LINEAR_SLICE_PITCH word*/ +/*define for linear_slice_pitch field*/ +#define SDMA_PKT_COPY_TILED_BC_LINEAR_SLICE_PITCH_linear_slice_pitch_offset 11 +#define SDMA_PKT_COPY_TILED_BC_LINEAR_SLICE_PITCH_linear_slice_pitch_mask 0xFFFFFFFF +#define SDMA_PKT_COPY_TILED_BC_LINEAR_SLICE_PITCH_linear_slice_pitch_shift 0 +#define SDMA_PKT_COPY_TILED_BC_LINEAR_SLICE_PITCH_LINEAR_SLICE_PITCH(x) (((x) & SDMA_PKT_COPY_TILED_BC_LINEAR_SLICE_PITCH_linear_slice_pitch_mask) << SDMA_PKT_COPY_TILED_BC_LINEAR_SLICE_PITCH_linear_slice_pitch_shift) + +/*define for COUNT word*/ +/*define for count field*/ +#define SDMA_PKT_COPY_TILED_BC_COUNT_count_offset 12 +#define SDMA_PKT_COPY_TILED_BC_COUNT_count_mask 0x000FFFFF +#define SDMA_PKT_COPY_TILED_BC_COUNT_count_shift 2 +#define SDMA_PKT_COPY_TILED_BC_COUNT_COUNT(x) (((x) & SDMA_PKT_COPY_TILED_BC_COUNT_count_mask) << SDMA_PKT_COPY_TILED_BC_COUNT_count_shift) + + +/* +** Definitions for SDMA_PKT_COPY_L2T_BROADCAST packet +*/ + +/*define for HEADER word*/ +/*define for op field*/ +#define SDMA_PKT_COPY_L2T_BROADCAST_HEADER_op_offset 0 +#define SDMA_PKT_COPY_L2T_BROADCAST_HEADER_op_mask 0x000000FF +#define SDMA_PKT_COPY_L2T_BROADCAST_HEADER_op_shift 0 +#define SDMA_PKT_COPY_L2T_BROADCAST_HEADER_OP(x) (((x) & SDMA_PKT_COPY_L2T_BROADCAST_HEADER_op_mask) << SDMA_PKT_COPY_L2T_BROADCAST_HEADER_op_shift) + +/*define for sub_op field*/ +#define SDMA_PKT_COPY_L2T_BROADCAST_HEADER_sub_op_offset 0 +#define SDMA_PKT_COPY_L2T_BROADCAST_HEADER_sub_op_mask 0x000000FF +#define SDMA_PKT_COPY_L2T_BROADCAST_HEADER_sub_op_shift 8 +#define SDMA_PKT_COPY_L2T_BROADCAST_HEADER_SUB_OP(x) (((x) & SDMA_PKT_COPY_L2T_BROADCAST_HEADER_sub_op_mask) << SDMA_PKT_COPY_L2T_BROADCAST_HEADER_sub_op_shift) + +/*define for encrypt field*/ +#define SDMA_PKT_COPY_L2T_BROADCAST_HEADER_encrypt_offset 0 +#define SDMA_PKT_COPY_L2T_BROADCAST_HEADER_encrypt_mask 0x00000001 +#define SDMA_PKT_COPY_L2T_BROADCAST_HEADER_encrypt_shift 16 +#define SDMA_PKT_COPY_L2T_BROADCAST_HEADER_ENCRYPT(x) (((x) & SDMA_PKT_COPY_L2T_BROADCAST_HEADER_encrypt_mask) << SDMA_PKT_COPY_L2T_BROADCAST_HEADER_encrypt_shift) + +/*define for tmz field*/ +#define SDMA_PKT_COPY_L2T_BROADCAST_HEADER_tmz_offset 0 +#define SDMA_PKT_COPY_L2T_BROADCAST_HEADER_tmz_mask 0x00000001 +#define SDMA_PKT_COPY_L2T_BROADCAST_HEADER_tmz_shift 18 +#define SDMA_PKT_COPY_L2T_BROADCAST_HEADER_TMZ(x) (((x) & SDMA_PKT_COPY_L2T_BROADCAST_HEADER_tmz_mask) << SDMA_PKT_COPY_L2T_BROADCAST_HEADER_tmz_shift) + +/*define for cpv field*/ +#define SDMA_PKT_COPY_L2T_BROADCAST_HEADER_cpv_offset 0 +#define SDMA_PKT_COPY_L2T_BROADCAST_HEADER_cpv_mask 0x00000001 +#define SDMA_PKT_COPY_L2T_BROADCAST_HEADER_cpv_shift 19 +#define SDMA_PKT_COPY_L2T_BROADCAST_HEADER_CPV(x) (((x) & SDMA_PKT_COPY_L2T_BROADCAST_HEADER_cpv_mask) << SDMA_PKT_COPY_L2T_BROADCAST_HEADER_cpv_shift) + +/*define for videocopy field*/ +#define SDMA_PKT_COPY_L2T_BROADCAST_HEADER_videocopy_offset 0 +#define SDMA_PKT_COPY_L2T_BROADCAST_HEADER_videocopy_mask 0x00000001 +#define SDMA_PKT_COPY_L2T_BROADCAST_HEADER_videocopy_shift 26 +#define SDMA_PKT_COPY_L2T_BROADCAST_HEADER_VIDEOCOPY(x) (((x) & SDMA_PKT_COPY_L2T_BROADCAST_HEADER_videocopy_mask) << SDMA_PKT_COPY_L2T_BROADCAST_HEADER_videocopy_shift) + +/*define for broadcast field*/ +#define SDMA_PKT_COPY_L2T_BROADCAST_HEADER_broadcast_offset 0 +#define SDMA_PKT_COPY_L2T_BROADCAST_HEADER_broadcast_mask 0x00000001 +#define SDMA_PKT_COPY_L2T_BROADCAST_HEADER_broadcast_shift 27 +#define SDMA_PKT_COPY_L2T_BROADCAST_HEADER_BROADCAST(x) (((x) & SDMA_PKT_COPY_L2T_BROADCAST_HEADER_broadcast_mask) << SDMA_PKT_COPY_L2T_BROADCAST_HEADER_broadcast_shift) + +/*define for TILED_ADDR_LO_0 word*/ +/*define for tiled_addr0_31_0 field*/ +#define SDMA_PKT_COPY_L2T_BROADCAST_TILED_ADDR_LO_0_tiled_addr0_31_0_offset 1 +#define SDMA_PKT_COPY_L2T_BROADCAST_TILED_ADDR_LO_0_tiled_addr0_31_0_mask 0xFFFFFFFF +#define SDMA_PKT_COPY_L2T_BROADCAST_TILED_ADDR_LO_0_tiled_addr0_31_0_shift 0 +#define SDMA_PKT_COPY_L2T_BROADCAST_TILED_ADDR_LO_0_TILED_ADDR0_31_0(x) (((x) & SDMA_PKT_COPY_L2T_BROADCAST_TILED_ADDR_LO_0_tiled_addr0_31_0_mask) << SDMA_PKT_COPY_L2T_BROADCAST_TILED_ADDR_LO_0_tiled_addr0_31_0_shift) + +/*define for TILED_ADDR_HI_0 word*/ +/*define for tiled_addr0_63_32 field*/ +#define SDMA_PKT_COPY_L2T_BROADCAST_TILED_ADDR_HI_0_tiled_addr0_63_32_offset 2 +#define SDMA_PKT_COPY_L2T_BROADCAST_TILED_ADDR_HI_0_tiled_addr0_63_32_mask 0xFFFFFFFF +#define SDMA_PKT_COPY_L2T_BROADCAST_TILED_ADDR_HI_0_tiled_addr0_63_32_shift 0 +#define SDMA_PKT_COPY_L2T_BROADCAST_TILED_ADDR_HI_0_TILED_ADDR0_63_32(x) (((x) & SDMA_PKT_COPY_L2T_BROADCAST_TILED_ADDR_HI_0_tiled_addr0_63_32_mask) << SDMA_PKT_COPY_L2T_BROADCAST_TILED_ADDR_HI_0_tiled_addr0_63_32_shift) + +/*define for TILED_ADDR_LO_1 word*/ +/*define for tiled_addr1_31_0 field*/ +#define SDMA_PKT_COPY_L2T_BROADCAST_TILED_ADDR_LO_1_tiled_addr1_31_0_offset 3 +#define SDMA_PKT_COPY_L2T_BROADCAST_TILED_ADDR_LO_1_tiled_addr1_31_0_mask 0xFFFFFFFF +#define SDMA_PKT_COPY_L2T_BROADCAST_TILED_ADDR_LO_1_tiled_addr1_31_0_shift 0 +#define SDMA_PKT_COPY_L2T_BROADCAST_TILED_ADDR_LO_1_TILED_ADDR1_31_0(x) (((x) & SDMA_PKT_COPY_L2T_BROADCAST_TILED_ADDR_LO_1_tiled_addr1_31_0_mask) << SDMA_PKT_COPY_L2T_BROADCAST_TILED_ADDR_LO_1_tiled_addr1_31_0_shift) + +/*define for TILED_ADDR_HI_1 word*/ +/*define for tiled_addr1_63_32 field*/ +#define SDMA_PKT_COPY_L2T_BROADCAST_TILED_ADDR_HI_1_tiled_addr1_63_32_offset 4 +#define SDMA_PKT_COPY_L2T_BROADCAST_TILED_ADDR_HI_1_tiled_addr1_63_32_mask 0xFFFFFFFF +#define SDMA_PKT_COPY_L2T_BROADCAST_TILED_ADDR_HI_1_tiled_addr1_63_32_shift 0 +#define SDMA_PKT_COPY_L2T_BROADCAST_TILED_ADDR_HI_1_TILED_ADDR1_63_32(x) (((x) & SDMA_PKT_COPY_L2T_BROADCAST_TILED_ADDR_HI_1_tiled_addr1_63_32_mask) << SDMA_PKT_COPY_L2T_BROADCAST_TILED_ADDR_HI_1_tiled_addr1_63_32_shift) + +/*define for DW_5 word*/ +/*define for width field*/ +#define SDMA_PKT_COPY_L2T_BROADCAST_DW_5_width_offset 5 +#define SDMA_PKT_COPY_L2T_BROADCAST_DW_5_width_mask 0x00003FFF +#define SDMA_PKT_COPY_L2T_BROADCAST_DW_5_width_shift 0 +#define SDMA_PKT_COPY_L2T_BROADCAST_DW_5_WIDTH(x) (((x) & SDMA_PKT_COPY_L2T_BROADCAST_DW_5_width_mask) << SDMA_PKT_COPY_L2T_BROADCAST_DW_5_width_shift) + +/*define for DW_6 word*/ +/*define for height field*/ +#define SDMA_PKT_COPY_L2T_BROADCAST_DW_6_height_offset 6 +#define SDMA_PKT_COPY_L2T_BROADCAST_DW_6_height_mask 0x00003FFF +#define SDMA_PKT_COPY_L2T_BROADCAST_DW_6_height_shift 0 +#define SDMA_PKT_COPY_L2T_BROADCAST_DW_6_HEIGHT(x) (((x) & SDMA_PKT_COPY_L2T_BROADCAST_DW_6_height_mask) << SDMA_PKT_COPY_L2T_BROADCAST_DW_6_height_shift) + +/*define for depth field*/ +#define SDMA_PKT_COPY_L2T_BROADCAST_DW_6_depth_offset 6 +#define SDMA_PKT_COPY_L2T_BROADCAST_DW_6_depth_mask 0x00001FFF +#define SDMA_PKT_COPY_L2T_BROADCAST_DW_6_depth_shift 16 +#define SDMA_PKT_COPY_L2T_BROADCAST_DW_6_DEPTH(x) (((x) & SDMA_PKT_COPY_L2T_BROADCAST_DW_6_depth_mask) << SDMA_PKT_COPY_L2T_BROADCAST_DW_6_depth_shift) + +/*define for DW_7 word*/ +/*define for element_size field*/ +#define SDMA_PKT_COPY_L2T_BROADCAST_DW_7_element_size_offset 7 +#define SDMA_PKT_COPY_L2T_BROADCAST_DW_7_element_size_mask 0x00000007 +#define SDMA_PKT_COPY_L2T_BROADCAST_DW_7_element_size_shift 0 +#define SDMA_PKT_COPY_L2T_BROADCAST_DW_7_ELEMENT_SIZE(x) (((x) & SDMA_PKT_COPY_L2T_BROADCAST_DW_7_element_size_mask) << SDMA_PKT_COPY_L2T_BROADCAST_DW_7_element_size_shift) + +/*define for swizzle_mode field*/ +#define SDMA_PKT_COPY_L2T_BROADCAST_DW_7_swizzle_mode_offset 7 +#define SDMA_PKT_COPY_L2T_BROADCAST_DW_7_swizzle_mode_mask 0x0000001F +#define SDMA_PKT_COPY_L2T_BROADCAST_DW_7_swizzle_mode_shift 3 +#define SDMA_PKT_COPY_L2T_BROADCAST_DW_7_SWIZZLE_MODE(x) (((x) & SDMA_PKT_COPY_L2T_BROADCAST_DW_7_swizzle_mode_mask) << SDMA_PKT_COPY_L2T_BROADCAST_DW_7_swizzle_mode_shift) + +/*define for dimension field*/ +#define SDMA_PKT_COPY_L2T_BROADCAST_DW_7_dimension_offset 7 +#define SDMA_PKT_COPY_L2T_BROADCAST_DW_7_dimension_mask 0x00000003 +#define SDMA_PKT_COPY_L2T_BROADCAST_DW_7_dimension_shift 9 +#define SDMA_PKT_COPY_L2T_BROADCAST_DW_7_DIMENSION(x) (((x) & SDMA_PKT_COPY_L2T_BROADCAST_DW_7_dimension_mask) << SDMA_PKT_COPY_L2T_BROADCAST_DW_7_dimension_shift) + +/*define for mip_max field*/ +#define SDMA_PKT_COPY_L2T_BROADCAST_DW_7_mip_max_offset 7 +#define SDMA_PKT_COPY_L2T_BROADCAST_DW_7_mip_max_mask 0x0000000F +#define SDMA_PKT_COPY_L2T_BROADCAST_DW_7_mip_max_shift 16 +#define SDMA_PKT_COPY_L2T_BROADCAST_DW_7_MIP_MAX(x) (((x) & SDMA_PKT_COPY_L2T_BROADCAST_DW_7_mip_max_mask) << SDMA_PKT_COPY_L2T_BROADCAST_DW_7_mip_max_shift) + +/*define for DW_8 word*/ +/*define for x field*/ +#define SDMA_PKT_COPY_L2T_BROADCAST_DW_8_x_offset 8 +#define SDMA_PKT_COPY_L2T_BROADCAST_DW_8_x_mask 0x00003FFF +#define SDMA_PKT_COPY_L2T_BROADCAST_DW_8_x_shift 0 +#define SDMA_PKT_COPY_L2T_BROADCAST_DW_8_X(x) (((x) & SDMA_PKT_COPY_L2T_BROADCAST_DW_8_x_mask) << SDMA_PKT_COPY_L2T_BROADCAST_DW_8_x_shift) + +/*define for y field*/ +#define SDMA_PKT_COPY_L2T_BROADCAST_DW_8_y_offset 8 +#define SDMA_PKT_COPY_L2T_BROADCAST_DW_8_y_mask 0x00003FFF +#define SDMA_PKT_COPY_L2T_BROADCAST_DW_8_y_shift 16 +#define SDMA_PKT_COPY_L2T_BROADCAST_DW_8_Y(x) (((x) & SDMA_PKT_COPY_L2T_BROADCAST_DW_8_y_mask) << SDMA_PKT_COPY_L2T_BROADCAST_DW_8_y_shift) + +/*define for DW_9 word*/ +/*define for z field*/ +#define SDMA_PKT_COPY_L2T_BROADCAST_DW_9_z_offset 9 +#define SDMA_PKT_COPY_L2T_BROADCAST_DW_9_z_mask 0x00001FFF +#define SDMA_PKT_COPY_L2T_BROADCAST_DW_9_z_shift 0 +#define SDMA_PKT_COPY_L2T_BROADCAST_DW_9_Z(x) (((x) & SDMA_PKT_COPY_L2T_BROADCAST_DW_9_z_mask) << SDMA_PKT_COPY_L2T_BROADCAST_DW_9_z_shift) + +/*define for DW_10 word*/ +/*define for dst2_sw field*/ +#define SDMA_PKT_COPY_L2T_BROADCAST_DW_10_dst2_sw_offset 10 +#define SDMA_PKT_COPY_L2T_BROADCAST_DW_10_dst2_sw_mask 0x00000003 +#define SDMA_PKT_COPY_L2T_BROADCAST_DW_10_dst2_sw_shift 8 +#define SDMA_PKT_COPY_L2T_BROADCAST_DW_10_DST2_SW(x) (((x) & SDMA_PKT_COPY_L2T_BROADCAST_DW_10_dst2_sw_mask) << SDMA_PKT_COPY_L2T_BROADCAST_DW_10_dst2_sw_shift) + +/*define for dst2_cache_policy field*/ +#define SDMA_PKT_COPY_L2T_BROADCAST_DW_10_dst2_cache_policy_offset 10 +#define SDMA_PKT_COPY_L2T_BROADCAST_DW_10_dst2_cache_policy_mask 0x00000007 +#define SDMA_PKT_COPY_L2T_BROADCAST_DW_10_dst2_cache_policy_shift 10 +#define SDMA_PKT_COPY_L2T_BROADCAST_DW_10_DST2_CACHE_POLICY(x) (((x) & SDMA_PKT_COPY_L2T_BROADCAST_DW_10_dst2_cache_policy_mask) << SDMA_PKT_COPY_L2T_BROADCAST_DW_10_dst2_cache_policy_shift) + +/*define for linear_sw field*/ +#define SDMA_PKT_COPY_L2T_BROADCAST_DW_10_linear_sw_offset 10 +#define SDMA_PKT_COPY_L2T_BROADCAST_DW_10_linear_sw_mask 0x00000003 +#define SDMA_PKT_COPY_L2T_BROADCAST_DW_10_linear_sw_shift 16 +#define SDMA_PKT_COPY_L2T_BROADCAST_DW_10_LINEAR_SW(x) (((x) & SDMA_PKT_COPY_L2T_BROADCAST_DW_10_linear_sw_mask) << SDMA_PKT_COPY_L2T_BROADCAST_DW_10_linear_sw_shift) + +/*define for linear_cache_policy field*/ +#define SDMA_PKT_COPY_L2T_BROADCAST_DW_10_linear_cache_policy_offset 10 +#define SDMA_PKT_COPY_L2T_BROADCAST_DW_10_linear_cache_policy_mask 0x00000007 +#define SDMA_PKT_COPY_L2T_BROADCAST_DW_10_linear_cache_policy_shift 18 +#define SDMA_PKT_COPY_L2T_BROADCAST_DW_10_LINEAR_CACHE_POLICY(x) (((x) & SDMA_PKT_COPY_L2T_BROADCAST_DW_10_linear_cache_policy_mask) << SDMA_PKT_COPY_L2T_BROADCAST_DW_10_linear_cache_policy_shift) + +/*define for tile_sw field*/ +#define SDMA_PKT_COPY_L2T_BROADCAST_DW_10_tile_sw_offset 10 +#define SDMA_PKT_COPY_L2T_BROADCAST_DW_10_tile_sw_mask 0x00000003 +#define SDMA_PKT_COPY_L2T_BROADCAST_DW_10_tile_sw_shift 24 +#define SDMA_PKT_COPY_L2T_BROADCAST_DW_10_TILE_SW(x) (((x) & SDMA_PKT_COPY_L2T_BROADCAST_DW_10_tile_sw_mask) << SDMA_PKT_COPY_L2T_BROADCAST_DW_10_tile_sw_shift) + +/*define for tile_cache_policy field*/ +#define SDMA_PKT_COPY_L2T_BROADCAST_DW_10_tile_cache_policy_offset 10 +#define SDMA_PKT_COPY_L2T_BROADCAST_DW_10_tile_cache_policy_mask 0x00000007 +#define SDMA_PKT_COPY_L2T_BROADCAST_DW_10_tile_cache_policy_shift 26 +#define SDMA_PKT_COPY_L2T_BROADCAST_DW_10_TILE_CACHE_POLICY(x) (((x) & SDMA_PKT_COPY_L2T_BROADCAST_DW_10_tile_cache_policy_mask) << SDMA_PKT_COPY_L2T_BROADCAST_DW_10_tile_cache_policy_shift) + +/*define for LINEAR_ADDR_LO word*/ +/*define for linear_addr_31_0 field*/ +#define SDMA_PKT_COPY_L2T_BROADCAST_LINEAR_ADDR_LO_linear_addr_31_0_offset 11 +#define SDMA_PKT_COPY_L2T_BROADCAST_LINEAR_ADDR_LO_linear_addr_31_0_mask 0xFFFFFFFF +#define SDMA_PKT_COPY_L2T_BROADCAST_LINEAR_ADDR_LO_linear_addr_31_0_shift 0 +#define SDMA_PKT_COPY_L2T_BROADCAST_LINEAR_ADDR_LO_LINEAR_ADDR_31_0(x) (((x) & SDMA_PKT_COPY_L2T_BROADCAST_LINEAR_ADDR_LO_linear_addr_31_0_mask) << SDMA_PKT_COPY_L2T_BROADCAST_LINEAR_ADDR_LO_linear_addr_31_0_shift) + +/*define for LINEAR_ADDR_HI word*/ +/*define for linear_addr_63_32 field*/ +#define SDMA_PKT_COPY_L2T_BROADCAST_LINEAR_ADDR_HI_linear_addr_63_32_offset 12 +#define SDMA_PKT_COPY_L2T_BROADCAST_LINEAR_ADDR_HI_linear_addr_63_32_mask 0xFFFFFFFF +#define SDMA_PKT_COPY_L2T_BROADCAST_LINEAR_ADDR_HI_linear_addr_63_32_shift 0 +#define SDMA_PKT_COPY_L2T_BROADCAST_LINEAR_ADDR_HI_LINEAR_ADDR_63_32(x) (((x) & SDMA_PKT_COPY_L2T_BROADCAST_LINEAR_ADDR_HI_linear_addr_63_32_mask) << SDMA_PKT_COPY_L2T_BROADCAST_LINEAR_ADDR_HI_linear_addr_63_32_shift) + +/*define for LINEAR_PITCH word*/ +/*define for linear_pitch field*/ +#define SDMA_PKT_COPY_L2T_BROADCAST_LINEAR_PITCH_linear_pitch_offset 13 +#define SDMA_PKT_COPY_L2T_BROADCAST_LINEAR_PITCH_linear_pitch_mask 0x0007FFFF +#define SDMA_PKT_COPY_L2T_BROADCAST_LINEAR_PITCH_linear_pitch_shift 0 +#define SDMA_PKT_COPY_L2T_BROADCAST_LINEAR_PITCH_LINEAR_PITCH(x) (((x) & SDMA_PKT_COPY_L2T_BROADCAST_LINEAR_PITCH_linear_pitch_mask) << SDMA_PKT_COPY_L2T_BROADCAST_LINEAR_PITCH_linear_pitch_shift) + +/*define for LINEAR_SLICE_PITCH word*/ +/*define for linear_slice_pitch field*/ +#define SDMA_PKT_COPY_L2T_BROADCAST_LINEAR_SLICE_PITCH_linear_slice_pitch_offset 14 +#define SDMA_PKT_COPY_L2T_BROADCAST_LINEAR_SLICE_PITCH_linear_slice_pitch_mask 0xFFFFFFFF +#define SDMA_PKT_COPY_L2T_BROADCAST_LINEAR_SLICE_PITCH_linear_slice_pitch_shift 0 +#define SDMA_PKT_COPY_L2T_BROADCAST_LINEAR_SLICE_PITCH_LINEAR_SLICE_PITCH(x) (((x) & SDMA_PKT_COPY_L2T_BROADCAST_LINEAR_SLICE_PITCH_linear_slice_pitch_mask) << SDMA_PKT_COPY_L2T_BROADCAST_LINEAR_SLICE_PITCH_linear_slice_pitch_shift) + +/*define for COUNT word*/ +/*define for count field*/ +#define SDMA_PKT_COPY_L2T_BROADCAST_COUNT_count_offset 15 +#define SDMA_PKT_COPY_L2T_BROADCAST_COUNT_count_mask 0x3FFFFFFF +#define SDMA_PKT_COPY_L2T_BROADCAST_COUNT_count_shift 0 +#define SDMA_PKT_COPY_L2T_BROADCAST_COUNT_COUNT(x) (((x) & SDMA_PKT_COPY_L2T_BROADCAST_COUNT_count_mask) << SDMA_PKT_COPY_L2T_BROADCAST_COUNT_count_shift) + + +/* +** Definitions for SDMA_PKT_COPY_T2T packet +*/ + +/*define for HEADER word*/ +/*define for op field*/ +#define SDMA_PKT_COPY_T2T_HEADER_op_offset 0 +#define SDMA_PKT_COPY_T2T_HEADER_op_mask 0x000000FF +#define SDMA_PKT_COPY_T2T_HEADER_op_shift 0 +#define SDMA_PKT_COPY_T2T_HEADER_OP(x) (((x) & SDMA_PKT_COPY_T2T_HEADER_op_mask) << SDMA_PKT_COPY_T2T_HEADER_op_shift) + +/*define for sub_op field*/ +#define SDMA_PKT_COPY_T2T_HEADER_sub_op_offset 0 +#define SDMA_PKT_COPY_T2T_HEADER_sub_op_mask 0x000000FF +#define SDMA_PKT_COPY_T2T_HEADER_sub_op_shift 8 +#define SDMA_PKT_COPY_T2T_HEADER_SUB_OP(x) (((x) & SDMA_PKT_COPY_T2T_HEADER_sub_op_mask) << SDMA_PKT_COPY_T2T_HEADER_sub_op_shift) + +/*define for tmz field*/ +#define SDMA_PKT_COPY_T2T_HEADER_tmz_offset 0 +#define SDMA_PKT_COPY_T2T_HEADER_tmz_mask 0x00000001 +#define SDMA_PKT_COPY_T2T_HEADER_tmz_shift 18 +#define SDMA_PKT_COPY_T2T_HEADER_TMZ(x) (((x) & SDMA_PKT_COPY_T2T_HEADER_tmz_mask) << SDMA_PKT_COPY_T2T_HEADER_tmz_shift) + +/*define for dcc field*/ +#define SDMA_PKT_COPY_T2T_HEADER_dcc_offset 0 +#define SDMA_PKT_COPY_T2T_HEADER_dcc_mask 0x00000001 +#define SDMA_PKT_COPY_T2T_HEADER_dcc_shift 19 +#define SDMA_PKT_COPY_T2T_HEADER_DCC(x) (((x) & SDMA_PKT_COPY_T2T_HEADER_dcc_mask) << SDMA_PKT_COPY_T2T_HEADER_dcc_shift) + +/*define for cpv field*/ +#define SDMA_PKT_COPY_T2T_HEADER_cpv_offset 0 +#define SDMA_PKT_COPY_T2T_HEADER_cpv_mask 0x00000001 +#define SDMA_PKT_COPY_T2T_HEADER_cpv_shift 28 +#define SDMA_PKT_COPY_T2T_HEADER_CPV(x) (((x) & SDMA_PKT_COPY_T2T_HEADER_cpv_mask) << SDMA_PKT_COPY_T2T_HEADER_cpv_shift) + +/*define for dcc_dir field*/ +#define SDMA_PKT_COPY_T2T_HEADER_dcc_dir_offset 0 +#define SDMA_PKT_COPY_T2T_HEADER_dcc_dir_mask 0x00000001 +#define SDMA_PKT_COPY_T2T_HEADER_dcc_dir_shift 31 +#define SDMA_PKT_COPY_T2T_HEADER_DCC_DIR(x) (((x) & SDMA_PKT_COPY_T2T_HEADER_dcc_dir_mask) << SDMA_PKT_COPY_T2T_HEADER_dcc_dir_shift) + +/*define for SRC_ADDR_LO word*/ +/*define for src_addr_31_0 field*/ +#define SDMA_PKT_COPY_T2T_SRC_ADDR_LO_src_addr_31_0_offset 1 +#define SDMA_PKT_COPY_T2T_SRC_ADDR_LO_src_addr_31_0_mask 0xFFFFFFFF +#define SDMA_PKT_COPY_T2T_SRC_ADDR_LO_src_addr_31_0_shift 0 +#define SDMA_PKT_COPY_T2T_SRC_ADDR_LO_SRC_ADDR_31_0(x) (((x) & SDMA_PKT_COPY_T2T_SRC_ADDR_LO_src_addr_31_0_mask) << SDMA_PKT_COPY_T2T_SRC_ADDR_LO_src_addr_31_0_shift) + +/*define for SRC_ADDR_HI word*/ +/*define for src_addr_63_32 field*/ +#define SDMA_PKT_COPY_T2T_SRC_ADDR_HI_src_addr_63_32_offset 2 +#define SDMA_PKT_COPY_T2T_SRC_ADDR_HI_src_addr_63_32_mask 0xFFFFFFFF +#define SDMA_PKT_COPY_T2T_SRC_ADDR_HI_src_addr_63_32_shift 0 +#define SDMA_PKT_COPY_T2T_SRC_ADDR_HI_SRC_ADDR_63_32(x) (((x) & SDMA_PKT_COPY_T2T_SRC_ADDR_HI_src_addr_63_32_mask) << SDMA_PKT_COPY_T2T_SRC_ADDR_HI_src_addr_63_32_shift) + +/*define for DW_3 word*/ +/*define for src_x field*/ +#define SDMA_PKT_COPY_T2T_DW_3_src_x_offset 3 +#define SDMA_PKT_COPY_T2T_DW_3_src_x_mask 0x00003FFF +#define SDMA_PKT_COPY_T2T_DW_3_src_x_shift 0 +#define SDMA_PKT_COPY_T2T_DW_3_SRC_X(x) (((x) & SDMA_PKT_COPY_T2T_DW_3_src_x_mask) << SDMA_PKT_COPY_T2T_DW_3_src_x_shift) + +/*define for src_y field*/ +#define SDMA_PKT_COPY_T2T_DW_3_src_y_offset 3 +#define SDMA_PKT_COPY_T2T_DW_3_src_y_mask 0x00003FFF +#define SDMA_PKT_COPY_T2T_DW_3_src_y_shift 16 +#define SDMA_PKT_COPY_T2T_DW_3_SRC_Y(x) (((x) & SDMA_PKT_COPY_T2T_DW_3_src_y_mask) << SDMA_PKT_COPY_T2T_DW_3_src_y_shift) + +/*define for DW_4 word*/ +/*define for src_z field*/ +#define SDMA_PKT_COPY_T2T_DW_4_src_z_offset 4 +#define SDMA_PKT_COPY_T2T_DW_4_src_z_mask 0x00001FFF +#define SDMA_PKT_COPY_T2T_DW_4_src_z_shift 0 +#define SDMA_PKT_COPY_T2T_DW_4_SRC_Z(x) (((x) & SDMA_PKT_COPY_T2T_DW_4_src_z_mask) << SDMA_PKT_COPY_T2T_DW_4_src_z_shift) + +/*define for src_width field*/ +#define SDMA_PKT_COPY_T2T_DW_4_src_width_offset 4 +#define SDMA_PKT_COPY_T2T_DW_4_src_width_mask 0x00003FFF +#define SDMA_PKT_COPY_T2T_DW_4_src_width_shift 16 +#define SDMA_PKT_COPY_T2T_DW_4_SRC_WIDTH(x) (((x) & SDMA_PKT_COPY_T2T_DW_4_src_width_mask) << SDMA_PKT_COPY_T2T_DW_4_src_width_shift) + +/*define for DW_5 word*/ +/*define for src_height field*/ +#define SDMA_PKT_COPY_T2T_DW_5_src_height_offset 5 +#define SDMA_PKT_COPY_T2T_DW_5_src_height_mask 0x00003FFF +#define SDMA_PKT_COPY_T2T_DW_5_src_height_shift 0 +#define SDMA_PKT_COPY_T2T_DW_5_SRC_HEIGHT(x) (((x) & SDMA_PKT_COPY_T2T_DW_5_src_height_mask) << SDMA_PKT_COPY_T2T_DW_5_src_height_shift) + +/*define for src_depth field*/ +#define SDMA_PKT_COPY_T2T_DW_5_src_depth_offset 5 +#define SDMA_PKT_COPY_T2T_DW_5_src_depth_mask 0x00001FFF +#define SDMA_PKT_COPY_T2T_DW_5_src_depth_shift 16 +#define SDMA_PKT_COPY_T2T_DW_5_SRC_DEPTH(x) (((x) & SDMA_PKT_COPY_T2T_DW_5_src_depth_mask) << SDMA_PKT_COPY_T2T_DW_5_src_depth_shift) + +/*define for DW_6 word*/ +/*define for src_element_size field*/ +#define SDMA_PKT_COPY_T2T_DW_6_src_element_size_offset 6 +#define SDMA_PKT_COPY_T2T_DW_6_src_element_size_mask 0x00000007 +#define SDMA_PKT_COPY_T2T_DW_6_src_element_size_shift 0 +#define SDMA_PKT_COPY_T2T_DW_6_SRC_ELEMENT_SIZE(x) (((x) & SDMA_PKT_COPY_T2T_DW_6_src_element_size_mask) << SDMA_PKT_COPY_T2T_DW_6_src_element_size_shift) + +/*define for src_swizzle_mode field*/ +#define SDMA_PKT_COPY_T2T_DW_6_src_swizzle_mode_offset 6 +#define SDMA_PKT_COPY_T2T_DW_6_src_swizzle_mode_mask 0x0000001F +#define SDMA_PKT_COPY_T2T_DW_6_src_swizzle_mode_shift 3 +#define SDMA_PKT_COPY_T2T_DW_6_SRC_SWIZZLE_MODE(x) (((x) & SDMA_PKT_COPY_T2T_DW_6_src_swizzle_mode_mask) << SDMA_PKT_COPY_T2T_DW_6_src_swizzle_mode_shift) + +/*define for src_dimension field*/ +#define SDMA_PKT_COPY_T2T_DW_6_src_dimension_offset 6 +#define SDMA_PKT_COPY_T2T_DW_6_src_dimension_mask 0x00000003 +#define SDMA_PKT_COPY_T2T_DW_6_src_dimension_shift 9 +#define SDMA_PKT_COPY_T2T_DW_6_SRC_DIMENSION(x) (((x) & SDMA_PKT_COPY_T2T_DW_6_src_dimension_mask) << SDMA_PKT_COPY_T2T_DW_6_src_dimension_shift) + +/*define for src_mip_max field*/ +#define SDMA_PKT_COPY_T2T_DW_6_src_mip_max_offset 6 +#define SDMA_PKT_COPY_T2T_DW_6_src_mip_max_mask 0x0000000F +#define SDMA_PKT_COPY_T2T_DW_6_src_mip_max_shift 16 +#define SDMA_PKT_COPY_T2T_DW_6_SRC_MIP_MAX(x) (((x) & SDMA_PKT_COPY_T2T_DW_6_src_mip_max_mask) << SDMA_PKT_COPY_T2T_DW_6_src_mip_max_shift) + +/*define for src_mip_id field*/ +#define SDMA_PKT_COPY_T2T_DW_6_src_mip_id_offset 6 +#define SDMA_PKT_COPY_T2T_DW_6_src_mip_id_mask 0x0000000F +#define SDMA_PKT_COPY_T2T_DW_6_src_mip_id_shift 20 +#define SDMA_PKT_COPY_T2T_DW_6_SRC_MIP_ID(x) (((x) & SDMA_PKT_COPY_T2T_DW_6_src_mip_id_mask) << SDMA_PKT_COPY_T2T_DW_6_src_mip_id_shift) + +/*define for DST_ADDR_LO word*/ +/*define for dst_addr_31_0 field*/ +#define SDMA_PKT_COPY_T2T_DST_ADDR_LO_dst_addr_31_0_offset 7 +#define SDMA_PKT_COPY_T2T_DST_ADDR_LO_dst_addr_31_0_mask 0xFFFFFFFF +#define SDMA_PKT_COPY_T2T_DST_ADDR_LO_dst_addr_31_0_shift 0 +#define SDMA_PKT_COPY_T2T_DST_ADDR_LO_DST_ADDR_31_0(x) (((x) & SDMA_PKT_COPY_T2T_DST_ADDR_LO_dst_addr_31_0_mask) << SDMA_PKT_COPY_T2T_DST_ADDR_LO_dst_addr_31_0_shift) + +/*define for DST_ADDR_HI word*/ +/*define for dst_addr_63_32 field*/ +#define SDMA_PKT_COPY_T2T_DST_ADDR_HI_dst_addr_63_32_offset 8 +#define SDMA_PKT_COPY_T2T_DST_ADDR_HI_dst_addr_63_32_mask 0xFFFFFFFF +#define SDMA_PKT_COPY_T2T_DST_ADDR_HI_dst_addr_63_32_shift 0 +#define SDMA_PKT_COPY_T2T_DST_ADDR_HI_DST_ADDR_63_32(x) (((x) & SDMA_PKT_COPY_T2T_DST_ADDR_HI_dst_addr_63_32_mask) << SDMA_PKT_COPY_T2T_DST_ADDR_HI_dst_addr_63_32_shift) + +/*define for DW_9 word*/ +/*define for dst_x field*/ +#define SDMA_PKT_COPY_T2T_DW_9_dst_x_offset 9 +#define SDMA_PKT_COPY_T2T_DW_9_dst_x_mask 0x00003FFF +#define SDMA_PKT_COPY_T2T_DW_9_dst_x_shift 0 +#define SDMA_PKT_COPY_T2T_DW_9_DST_X(x) (((x) & SDMA_PKT_COPY_T2T_DW_9_dst_x_mask) << SDMA_PKT_COPY_T2T_DW_9_dst_x_shift) + +/*define for dst_y field*/ +#define SDMA_PKT_COPY_T2T_DW_9_dst_y_offset 9 +#define SDMA_PKT_COPY_T2T_DW_9_dst_y_mask 0x00003FFF +#define SDMA_PKT_COPY_T2T_DW_9_dst_y_shift 16 +#define SDMA_PKT_COPY_T2T_DW_9_DST_Y(x) (((x) & SDMA_PKT_COPY_T2T_DW_9_dst_y_mask) << SDMA_PKT_COPY_T2T_DW_9_dst_y_shift) + +/*define for DW_10 word*/ +/*define for dst_z field*/ +#define SDMA_PKT_COPY_T2T_DW_10_dst_z_offset 10 +#define SDMA_PKT_COPY_T2T_DW_10_dst_z_mask 0x00001FFF +#define SDMA_PKT_COPY_T2T_DW_10_dst_z_shift 0 +#define SDMA_PKT_COPY_T2T_DW_10_DST_Z(x) (((x) & SDMA_PKT_COPY_T2T_DW_10_dst_z_mask) << SDMA_PKT_COPY_T2T_DW_10_dst_z_shift) + +/*define for dst_width field*/ +#define SDMA_PKT_COPY_T2T_DW_10_dst_width_offset 10 +#define SDMA_PKT_COPY_T2T_DW_10_dst_width_mask 0x00003FFF +#define SDMA_PKT_COPY_T2T_DW_10_dst_width_shift 16 +#define SDMA_PKT_COPY_T2T_DW_10_DST_WIDTH(x) (((x) & SDMA_PKT_COPY_T2T_DW_10_dst_width_mask) << SDMA_PKT_COPY_T2T_DW_10_dst_width_shift) + +/*define for DW_11 word*/ +/*define for dst_height field*/ +#define SDMA_PKT_COPY_T2T_DW_11_dst_height_offset 11 +#define SDMA_PKT_COPY_T2T_DW_11_dst_height_mask 0x00003FFF +#define SDMA_PKT_COPY_T2T_DW_11_dst_height_shift 0 +#define SDMA_PKT_COPY_T2T_DW_11_DST_HEIGHT(x) (((x) & SDMA_PKT_COPY_T2T_DW_11_dst_height_mask) << SDMA_PKT_COPY_T2T_DW_11_dst_height_shift) + +/*define for dst_depth field*/ +#define SDMA_PKT_COPY_T2T_DW_11_dst_depth_offset 11 +#define SDMA_PKT_COPY_T2T_DW_11_dst_depth_mask 0x00001FFF +#define SDMA_PKT_COPY_T2T_DW_11_dst_depth_shift 16 +#define SDMA_PKT_COPY_T2T_DW_11_DST_DEPTH(x) (((x) & SDMA_PKT_COPY_T2T_DW_11_dst_depth_mask) << SDMA_PKT_COPY_T2T_DW_11_dst_depth_shift) + +/*define for DW_12 word*/ +/*define for dst_element_size field*/ +#define SDMA_PKT_COPY_T2T_DW_12_dst_element_size_offset 12 +#define SDMA_PKT_COPY_T2T_DW_12_dst_element_size_mask 0x00000007 +#define SDMA_PKT_COPY_T2T_DW_12_dst_element_size_shift 0 +#define SDMA_PKT_COPY_T2T_DW_12_DST_ELEMENT_SIZE(x) (((x) & SDMA_PKT_COPY_T2T_DW_12_dst_element_size_mask) << SDMA_PKT_COPY_T2T_DW_12_dst_element_size_shift) + +/*define for dst_swizzle_mode field*/ +#define SDMA_PKT_COPY_T2T_DW_12_dst_swizzle_mode_offset 12 +#define SDMA_PKT_COPY_T2T_DW_12_dst_swizzle_mode_mask 0x0000001F +#define SDMA_PKT_COPY_T2T_DW_12_dst_swizzle_mode_shift 3 +#define SDMA_PKT_COPY_T2T_DW_12_DST_SWIZZLE_MODE(x) (((x) & SDMA_PKT_COPY_T2T_DW_12_dst_swizzle_mode_mask) << SDMA_PKT_COPY_T2T_DW_12_dst_swizzle_mode_shift) + +/*define for dst_dimension field*/ +#define SDMA_PKT_COPY_T2T_DW_12_dst_dimension_offset 12 +#define SDMA_PKT_COPY_T2T_DW_12_dst_dimension_mask 0x00000003 +#define SDMA_PKT_COPY_T2T_DW_12_dst_dimension_shift 9 +#define SDMA_PKT_COPY_T2T_DW_12_DST_DIMENSION(x) (((x) & SDMA_PKT_COPY_T2T_DW_12_dst_dimension_mask) << SDMA_PKT_COPY_T2T_DW_12_dst_dimension_shift) + +/*define for dst_mip_max field*/ +#define SDMA_PKT_COPY_T2T_DW_12_dst_mip_max_offset 12 +#define SDMA_PKT_COPY_T2T_DW_12_dst_mip_max_mask 0x0000000F +#define SDMA_PKT_COPY_T2T_DW_12_dst_mip_max_shift 16 +#define SDMA_PKT_COPY_T2T_DW_12_DST_MIP_MAX(x) (((x) & SDMA_PKT_COPY_T2T_DW_12_dst_mip_max_mask) << SDMA_PKT_COPY_T2T_DW_12_dst_mip_max_shift) + +/*define for dst_mip_id field*/ +#define SDMA_PKT_COPY_T2T_DW_12_dst_mip_id_offset 12 +#define SDMA_PKT_COPY_T2T_DW_12_dst_mip_id_mask 0x0000000F +#define SDMA_PKT_COPY_T2T_DW_12_dst_mip_id_shift 20 +#define SDMA_PKT_COPY_T2T_DW_12_DST_MIP_ID(x) (((x) & SDMA_PKT_COPY_T2T_DW_12_dst_mip_id_mask) << SDMA_PKT_COPY_T2T_DW_12_dst_mip_id_shift) + +/*define for DW_13 word*/ +/*define for rect_x field*/ +#define SDMA_PKT_COPY_T2T_DW_13_rect_x_offset 13 +#define SDMA_PKT_COPY_T2T_DW_13_rect_x_mask 0x00003FFF +#define SDMA_PKT_COPY_T2T_DW_13_rect_x_shift 0 +#define SDMA_PKT_COPY_T2T_DW_13_RECT_X(x) (((x) & SDMA_PKT_COPY_T2T_DW_13_rect_x_mask) << SDMA_PKT_COPY_T2T_DW_13_rect_x_shift) + +/*define for rect_y field*/ +#define SDMA_PKT_COPY_T2T_DW_13_rect_y_offset 13 +#define SDMA_PKT_COPY_T2T_DW_13_rect_y_mask 0x00003FFF +#define SDMA_PKT_COPY_T2T_DW_13_rect_y_shift 16 +#define SDMA_PKT_COPY_T2T_DW_13_RECT_Y(x) (((x) & SDMA_PKT_COPY_T2T_DW_13_rect_y_mask) << SDMA_PKT_COPY_T2T_DW_13_rect_y_shift) + +/*define for DW_14 word*/ +/*define for rect_z field*/ +#define SDMA_PKT_COPY_T2T_DW_14_rect_z_offset 14 +#define SDMA_PKT_COPY_T2T_DW_14_rect_z_mask 0x00001FFF +#define SDMA_PKT_COPY_T2T_DW_14_rect_z_shift 0 +#define SDMA_PKT_COPY_T2T_DW_14_RECT_Z(x) (((x) & SDMA_PKT_COPY_T2T_DW_14_rect_z_mask) << SDMA_PKT_COPY_T2T_DW_14_rect_z_shift) + +/*define for dst_sw field*/ +#define SDMA_PKT_COPY_T2T_DW_14_dst_sw_offset 14 +#define SDMA_PKT_COPY_T2T_DW_14_dst_sw_mask 0x00000003 +#define SDMA_PKT_COPY_T2T_DW_14_dst_sw_shift 16 +#define SDMA_PKT_COPY_T2T_DW_14_DST_SW(x) (((x) & SDMA_PKT_COPY_T2T_DW_14_dst_sw_mask) << SDMA_PKT_COPY_T2T_DW_14_dst_sw_shift) + +/*define for dst_cache_policy field*/ +#define SDMA_PKT_COPY_T2T_DW_14_dst_cache_policy_offset 14 +#define SDMA_PKT_COPY_T2T_DW_14_dst_cache_policy_mask 0x00000007 +#define SDMA_PKT_COPY_T2T_DW_14_dst_cache_policy_shift 18 +#define SDMA_PKT_COPY_T2T_DW_14_DST_CACHE_POLICY(x) (((x) & SDMA_PKT_COPY_T2T_DW_14_dst_cache_policy_mask) << SDMA_PKT_COPY_T2T_DW_14_dst_cache_policy_shift) + +/*define for src_sw field*/ +#define SDMA_PKT_COPY_T2T_DW_14_src_sw_offset 14 +#define SDMA_PKT_COPY_T2T_DW_14_src_sw_mask 0x00000003 +#define SDMA_PKT_COPY_T2T_DW_14_src_sw_shift 24 +#define SDMA_PKT_COPY_T2T_DW_14_SRC_SW(x) (((x) & SDMA_PKT_COPY_T2T_DW_14_src_sw_mask) << SDMA_PKT_COPY_T2T_DW_14_src_sw_shift) + +/*define for src_cache_policy field*/ +#define SDMA_PKT_COPY_T2T_DW_14_src_cache_policy_offset 14 +#define SDMA_PKT_COPY_T2T_DW_14_src_cache_policy_mask 0x00000007 +#define SDMA_PKT_COPY_T2T_DW_14_src_cache_policy_shift 26 +#define SDMA_PKT_COPY_T2T_DW_14_SRC_CACHE_POLICY(x) (((x) & SDMA_PKT_COPY_T2T_DW_14_src_cache_policy_mask) << SDMA_PKT_COPY_T2T_DW_14_src_cache_policy_shift) + +/*define for META_ADDR_LO word*/ +/*define for meta_addr_31_0 field*/ +#define SDMA_PKT_COPY_T2T_META_ADDR_LO_meta_addr_31_0_offset 15 +#define SDMA_PKT_COPY_T2T_META_ADDR_LO_meta_addr_31_0_mask 0xFFFFFFFF +#define SDMA_PKT_COPY_T2T_META_ADDR_LO_meta_addr_31_0_shift 0 +#define SDMA_PKT_COPY_T2T_META_ADDR_LO_META_ADDR_31_0(x) (((x) & SDMA_PKT_COPY_T2T_META_ADDR_LO_meta_addr_31_0_mask) << SDMA_PKT_COPY_T2T_META_ADDR_LO_meta_addr_31_0_shift) + +/*define for META_ADDR_HI word*/ +/*define for meta_addr_63_32 field*/ +#define SDMA_PKT_COPY_T2T_META_ADDR_HI_meta_addr_63_32_offset 16 +#define SDMA_PKT_COPY_T2T_META_ADDR_HI_meta_addr_63_32_mask 0xFFFFFFFF +#define SDMA_PKT_COPY_T2T_META_ADDR_HI_meta_addr_63_32_shift 0 +#define SDMA_PKT_COPY_T2T_META_ADDR_HI_META_ADDR_63_32(x) (((x) & SDMA_PKT_COPY_T2T_META_ADDR_HI_meta_addr_63_32_mask) << SDMA_PKT_COPY_T2T_META_ADDR_HI_meta_addr_63_32_shift) + +/*define for META_CONFIG word*/ +/*define for data_format field*/ +#define SDMA_PKT_COPY_T2T_META_CONFIG_data_format_offset 17 +#define SDMA_PKT_COPY_T2T_META_CONFIG_data_format_mask 0x0000007F +#define SDMA_PKT_COPY_T2T_META_CONFIG_data_format_shift 0 +#define SDMA_PKT_COPY_T2T_META_CONFIG_DATA_FORMAT(x) (((x) & SDMA_PKT_COPY_T2T_META_CONFIG_data_format_mask) << SDMA_PKT_COPY_T2T_META_CONFIG_data_format_shift) + +/*define for color_transform_disable field*/ +#define SDMA_PKT_COPY_T2T_META_CONFIG_color_transform_disable_offset 17 +#define SDMA_PKT_COPY_T2T_META_CONFIG_color_transform_disable_mask 0x00000001 +#define SDMA_PKT_COPY_T2T_META_CONFIG_color_transform_disable_shift 7 +#define SDMA_PKT_COPY_T2T_META_CONFIG_COLOR_TRANSFORM_DISABLE(x) (((x) & SDMA_PKT_COPY_T2T_META_CONFIG_color_transform_disable_mask) << SDMA_PKT_COPY_T2T_META_CONFIG_color_transform_disable_shift) + +/*define for alpha_is_on_msb field*/ +#define SDMA_PKT_COPY_T2T_META_CONFIG_alpha_is_on_msb_offset 17 +#define SDMA_PKT_COPY_T2T_META_CONFIG_alpha_is_on_msb_mask 0x00000001 +#define SDMA_PKT_COPY_T2T_META_CONFIG_alpha_is_on_msb_shift 8 +#define SDMA_PKT_COPY_T2T_META_CONFIG_ALPHA_IS_ON_MSB(x) (((x) & SDMA_PKT_COPY_T2T_META_CONFIG_alpha_is_on_msb_mask) << SDMA_PKT_COPY_T2T_META_CONFIG_alpha_is_on_msb_shift) + +/*define for number_type field*/ +#define SDMA_PKT_COPY_T2T_META_CONFIG_number_type_offset 17 +#define SDMA_PKT_COPY_T2T_META_CONFIG_number_type_mask 0x00000007 +#define SDMA_PKT_COPY_T2T_META_CONFIG_number_type_shift 9 +#define SDMA_PKT_COPY_T2T_META_CONFIG_NUMBER_TYPE(x) (((x) & SDMA_PKT_COPY_T2T_META_CONFIG_number_type_mask) << SDMA_PKT_COPY_T2T_META_CONFIG_number_type_shift) + +/*define for surface_type field*/ +#define SDMA_PKT_COPY_T2T_META_CONFIG_surface_type_offset 17 +#define SDMA_PKT_COPY_T2T_META_CONFIG_surface_type_mask 0x00000003 +#define SDMA_PKT_COPY_T2T_META_CONFIG_surface_type_shift 12 +#define SDMA_PKT_COPY_T2T_META_CONFIG_SURFACE_TYPE(x) (((x) & SDMA_PKT_COPY_T2T_META_CONFIG_surface_type_mask) << SDMA_PKT_COPY_T2T_META_CONFIG_surface_type_shift) + +/*define for meta_llc field*/ +#define SDMA_PKT_COPY_T2T_META_CONFIG_meta_llc_offset 17 +#define SDMA_PKT_COPY_T2T_META_CONFIG_meta_llc_mask 0x00000001 +#define SDMA_PKT_COPY_T2T_META_CONFIG_meta_llc_shift 14 +#define SDMA_PKT_COPY_T2T_META_CONFIG_META_LLC(x) (((x) & SDMA_PKT_COPY_T2T_META_CONFIG_meta_llc_mask) << SDMA_PKT_COPY_T2T_META_CONFIG_meta_llc_shift) + +/*define for max_comp_block_size field*/ +#define SDMA_PKT_COPY_T2T_META_CONFIG_max_comp_block_size_offset 17 +#define SDMA_PKT_COPY_T2T_META_CONFIG_max_comp_block_size_mask 0x00000003 +#define SDMA_PKT_COPY_T2T_META_CONFIG_max_comp_block_size_shift 24 +#define SDMA_PKT_COPY_T2T_META_CONFIG_MAX_COMP_BLOCK_SIZE(x) (((x) & SDMA_PKT_COPY_T2T_META_CONFIG_max_comp_block_size_mask) << SDMA_PKT_COPY_T2T_META_CONFIG_max_comp_block_size_shift) + +/*define for max_uncomp_block_size field*/ +#define SDMA_PKT_COPY_T2T_META_CONFIG_max_uncomp_block_size_offset 17 +#define SDMA_PKT_COPY_T2T_META_CONFIG_max_uncomp_block_size_mask 0x00000003 +#define SDMA_PKT_COPY_T2T_META_CONFIG_max_uncomp_block_size_shift 26 +#define SDMA_PKT_COPY_T2T_META_CONFIG_MAX_UNCOMP_BLOCK_SIZE(x) (((x) & SDMA_PKT_COPY_T2T_META_CONFIG_max_uncomp_block_size_mask) << SDMA_PKT_COPY_T2T_META_CONFIG_max_uncomp_block_size_shift) + +/*define for write_compress_enable field*/ +#define SDMA_PKT_COPY_T2T_META_CONFIG_write_compress_enable_offset 17 +#define SDMA_PKT_COPY_T2T_META_CONFIG_write_compress_enable_mask 0x00000001 +#define SDMA_PKT_COPY_T2T_META_CONFIG_write_compress_enable_shift 28 +#define SDMA_PKT_COPY_T2T_META_CONFIG_WRITE_COMPRESS_ENABLE(x) (((x) & SDMA_PKT_COPY_T2T_META_CONFIG_write_compress_enable_mask) << SDMA_PKT_COPY_T2T_META_CONFIG_write_compress_enable_shift) + +/*define for meta_tmz field*/ +#define SDMA_PKT_COPY_T2T_META_CONFIG_meta_tmz_offset 17 +#define SDMA_PKT_COPY_T2T_META_CONFIG_meta_tmz_mask 0x00000001 +#define SDMA_PKT_COPY_T2T_META_CONFIG_meta_tmz_shift 29 +#define SDMA_PKT_COPY_T2T_META_CONFIG_META_TMZ(x) (((x) & SDMA_PKT_COPY_T2T_META_CONFIG_meta_tmz_mask) << SDMA_PKT_COPY_T2T_META_CONFIG_meta_tmz_shift) + +/*define for pipe_aligned field*/ +#define SDMA_PKT_COPY_T2T_META_CONFIG_pipe_aligned_offset 17 +#define SDMA_PKT_COPY_T2T_META_CONFIG_pipe_aligned_mask 0x00000001 +#define SDMA_PKT_COPY_T2T_META_CONFIG_pipe_aligned_shift 31 +#define SDMA_PKT_COPY_T2T_META_CONFIG_PIPE_ALIGNED(x) (((x) & SDMA_PKT_COPY_T2T_META_CONFIG_pipe_aligned_mask) << SDMA_PKT_COPY_T2T_META_CONFIG_pipe_aligned_shift) + + +/* +** Definitions for SDMA_PKT_COPY_T2T_BC packet +*/ + +/*define for HEADER word*/ +/*define for op field*/ +#define SDMA_PKT_COPY_T2T_BC_HEADER_op_offset 0 +#define SDMA_PKT_COPY_T2T_BC_HEADER_op_mask 0x000000FF +#define SDMA_PKT_COPY_T2T_BC_HEADER_op_shift 0 +#define SDMA_PKT_COPY_T2T_BC_HEADER_OP(x) (((x) & SDMA_PKT_COPY_T2T_BC_HEADER_op_mask) << SDMA_PKT_COPY_T2T_BC_HEADER_op_shift) + +/*define for sub_op field*/ +#define SDMA_PKT_COPY_T2T_BC_HEADER_sub_op_offset 0 +#define SDMA_PKT_COPY_T2T_BC_HEADER_sub_op_mask 0x000000FF +#define SDMA_PKT_COPY_T2T_BC_HEADER_sub_op_shift 8 +#define SDMA_PKT_COPY_T2T_BC_HEADER_SUB_OP(x) (((x) & SDMA_PKT_COPY_T2T_BC_HEADER_sub_op_mask) << SDMA_PKT_COPY_T2T_BC_HEADER_sub_op_shift) + +/*define for SRC_ADDR_LO word*/ +/*define for src_addr_31_0 field*/ +#define SDMA_PKT_COPY_T2T_BC_SRC_ADDR_LO_src_addr_31_0_offset 1 +#define SDMA_PKT_COPY_T2T_BC_SRC_ADDR_LO_src_addr_31_0_mask 0xFFFFFFFF +#define SDMA_PKT_COPY_T2T_BC_SRC_ADDR_LO_src_addr_31_0_shift 0 +#define SDMA_PKT_COPY_T2T_BC_SRC_ADDR_LO_SRC_ADDR_31_0(x) (((x) & SDMA_PKT_COPY_T2T_BC_SRC_ADDR_LO_src_addr_31_0_mask) << SDMA_PKT_COPY_T2T_BC_SRC_ADDR_LO_src_addr_31_0_shift) + +/*define for SRC_ADDR_HI word*/ +/*define for src_addr_63_32 field*/ +#define SDMA_PKT_COPY_T2T_BC_SRC_ADDR_HI_src_addr_63_32_offset 2 +#define SDMA_PKT_COPY_T2T_BC_SRC_ADDR_HI_src_addr_63_32_mask 0xFFFFFFFF +#define SDMA_PKT_COPY_T2T_BC_SRC_ADDR_HI_src_addr_63_32_shift 0 +#define SDMA_PKT_COPY_T2T_BC_SRC_ADDR_HI_SRC_ADDR_63_32(x) (((x) & SDMA_PKT_COPY_T2T_BC_SRC_ADDR_HI_src_addr_63_32_mask) << SDMA_PKT_COPY_T2T_BC_SRC_ADDR_HI_src_addr_63_32_shift) + +/*define for DW_3 word*/ +/*define for src_x field*/ +#define SDMA_PKT_COPY_T2T_BC_DW_3_src_x_offset 3 +#define SDMA_PKT_COPY_T2T_BC_DW_3_src_x_mask 0x00003FFF +#define SDMA_PKT_COPY_T2T_BC_DW_3_src_x_shift 0 +#define SDMA_PKT_COPY_T2T_BC_DW_3_SRC_X(x) (((x) & SDMA_PKT_COPY_T2T_BC_DW_3_src_x_mask) << SDMA_PKT_COPY_T2T_BC_DW_3_src_x_shift) + +/*define for src_y field*/ +#define SDMA_PKT_COPY_T2T_BC_DW_3_src_y_offset 3 +#define SDMA_PKT_COPY_T2T_BC_DW_3_src_y_mask 0x00003FFF +#define SDMA_PKT_COPY_T2T_BC_DW_3_src_y_shift 16 +#define SDMA_PKT_COPY_T2T_BC_DW_3_SRC_Y(x) (((x) & SDMA_PKT_COPY_T2T_BC_DW_3_src_y_mask) << SDMA_PKT_COPY_T2T_BC_DW_3_src_y_shift) + +/*define for DW_4 word*/ +/*define for src_z field*/ +#define SDMA_PKT_COPY_T2T_BC_DW_4_src_z_offset 4 +#define SDMA_PKT_COPY_T2T_BC_DW_4_src_z_mask 0x000007FF +#define SDMA_PKT_COPY_T2T_BC_DW_4_src_z_shift 0 +#define SDMA_PKT_COPY_T2T_BC_DW_4_SRC_Z(x) (((x) & SDMA_PKT_COPY_T2T_BC_DW_4_src_z_mask) << SDMA_PKT_COPY_T2T_BC_DW_4_src_z_shift) + +/*define for src_width field*/ +#define SDMA_PKT_COPY_T2T_BC_DW_4_src_width_offset 4 +#define SDMA_PKT_COPY_T2T_BC_DW_4_src_width_mask 0x00003FFF +#define SDMA_PKT_COPY_T2T_BC_DW_4_src_width_shift 16 +#define SDMA_PKT_COPY_T2T_BC_DW_4_SRC_WIDTH(x) (((x) & SDMA_PKT_COPY_T2T_BC_DW_4_src_width_mask) << SDMA_PKT_COPY_T2T_BC_DW_4_src_width_shift) + +/*define for DW_5 word*/ +/*define for src_height field*/ +#define SDMA_PKT_COPY_T2T_BC_DW_5_src_height_offset 5 +#define SDMA_PKT_COPY_T2T_BC_DW_5_src_height_mask 0x00003FFF +#define SDMA_PKT_COPY_T2T_BC_DW_5_src_height_shift 0 +#define SDMA_PKT_COPY_T2T_BC_DW_5_SRC_HEIGHT(x) (((x) & SDMA_PKT_COPY_T2T_BC_DW_5_src_height_mask) << SDMA_PKT_COPY_T2T_BC_DW_5_src_height_shift) + +/*define for src_depth field*/ +#define SDMA_PKT_COPY_T2T_BC_DW_5_src_depth_offset 5 +#define SDMA_PKT_COPY_T2T_BC_DW_5_src_depth_mask 0x000007FF +#define SDMA_PKT_COPY_T2T_BC_DW_5_src_depth_shift 16 +#define SDMA_PKT_COPY_T2T_BC_DW_5_SRC_DEPTH(x) (((x) & SDMA_PKT_COPY_T2T_BC_DW_5_src_depth_mask) << SDMA_PKT_COPY_T2T_BC_DW_5_src_depth_shift) + +/*define for DW_6 word*/ +/*define for src_element_size field*/ +#define SDMA_PKT_COPY_T2T_BC_DW_6_src_element_size_offset 6 +#define SDMA_PKT_COPY_T2T_BC_DW_6_src_element_size_mask 0x00000007 +#define SDMA_PKT_COPY_T2T_BC_DW_6_src_element_size_shift 0 +#define SDMA_PKT_COPY_T2T_BC_DW_6_SRC_ELEMENT_SIZE(x) (((x) & SDMA_PKT_COPY_T2T_BC_DW_6_src_element_size_mask) << SDMA_PKT_COPY_T2T_BC_DW_6_src_element_size_shift) + +/*define for src_array_mode field*/ +#define SDMA_PKT_COPY_T2T_BC_DW_6_src_array_mode_offset 6 +#define SDMA_PKT_COPY_T2T_BC_DW_6_src_array_mode_mask 0x0000000F +#define SDMA_PKT_COPY_T2T_BC_DW_6_src_array_mode_shift 3 +#define SDMA_PKT_COPY_T2T_BC_DW_6_SRC_ARRAY_MODE(x) (((x) & SDMA_PKT_COPY_T2T_BC_DW_6_src_array_mode_mask) << SDMA_PKT_COPY_T2T_BC_DW_6_src_array_mode_shift) + +/*define for src_mit_mode field*/ +#define SDMA_PKT_COPY_T2T_BC_DW_6_src_mit_mode_offset 6 +#define SDMA_PKT_COPY_T2T_BC_DW_6_src_mit_mode_mask 0x00000007 +#define SDMA_PKT_COPY_T2T_BC_DW_6_src_mit_mode_shift 8 +#define SDMA_PKT_COPY_T2T_BC_DW_6_SRC_MIT_MODE(x) (((x) & SDMA_PKT_COPY_T2T_BC_DW_6_src_mit_mode_mask) << SDMA_PKT_COPY_T2T_BC_DW_6_src_mit_mode_shift) + +/*define for src_tilesplit_size field*/ +#define SDMA_PKT_COPY_T2T_BC_DW_6_src_tilesplit_size_offset 6 +#define SDMA_PKT_COPY_T2T_BC_DW_6_src_tilesplit_size_mask 0x00000007 +#define SDMA_PKT_COPY_T2T_BC_DW_6_src_tilesplit_size_shift 11 +#define SDMA_PKT_COPY_T2T_BC_DW_6_SRC_TILESPLIT_SIZE(x) (((x) & SDMA_PKT_COPY_T2T_BC_DW_6_src_tilesplit_size_mask) << SDMA_PKT_COPY_T2T_BC_DW_6_src_tilesplit_size_shift) + +/*define for src_bank_w field*/ +#define SDMA_PKT_COPY_T2T_BC_DW_6_src_bank_w_offset 6 +#define SDMA_PKT_COPY_T2T_BC_DW_6_src_bank_w_mask 0x00000003 +#define SDMA_PKT_COPY_T2T_BC_DW_6_src_bank_w_shift 15 +#define SDMA_PKT_COPY_T2T_BC_DW_6_SRC_BANK_W(x) (((x) & SDMA_PKT_COPY_T2T_BC_DW_6_src_bank_w_mask) << SDMA_PKT_COPY_T2T_BC_DW_6_src_bank_w_shift) + +/*define for src_bank_h field*/ +#define SDMA_PKT_COPY_T2T_BC_DW_6_src_bank_h_offset 6 +#define SDMA_PKT_COPY_T2T_BC_DW_6_src_bank_h_mask 0x00000003 +#define SDMA_PKT_COPY_T2T_BC_DW_6_src_bank_h_shift 18 +#define SDMA_PKT_COPY_T2T_BC_DW_6_SRC_BANK_H(x) (((x) & SDMA_PKT_COPY_T2T_BC_DW_6_src_bank_h_mask) << SDMA_PKT_COPY_T2T_BC_DW_6_src_bank_h_shift) + +/*define for src_num_bank field*/ +#define SDMA_PKT_COPY_T2T_BC_DW_6_src_num_bank_offset 6 +#define SDMA_PKT_COPY_T2T_BC_DW_6_src_num_bank_mask 0x00000003 +#define SDMA_PKT_COPY_T2T_BC_DW_6_src_num_bank_shift 21 +#define SDMA_PKT_COPY_T2T_BC_DW_6_SRC_NUM_BANK(x) (((x) & SDMA_PKT_COPY_T2T_BC_DW_6_src_num_bank_mask) << SDMA_PKT_COPY_T2T_BC_DW_6_src_num_bank_shift) + +/*define for src_mat_aspt field*/ +#define SDMA_PKT_COPY_T2T_BC_DW_6_src_mat_aspt_offset 6 +#define SDMA_PKT_COPY_T2T_BC_DW_6_src_mat_aspt_mask 0x00000003 +#define SDMA_PKT_COPY_T2T_BC_DW_6_src_mat_aspt_shift 24 +#define SDMA_PKT_COPY_T2T_BC_DW_6_SRC_MAT_ASPT(x) (((x) & SDMA_PKT_COPY_T2T_BC_DW_6_src_mat_aspt_mask) << SDMA_PKT_COPY_T2T_BC_DW_6_src_mat_aspt_shift) + +/*define for src_pipe_config field*/ +#define SDMA_PKT_COPY_T2T_BC_DW_6_src_pipe_config_offset 6 +#define SDMA_PKT_COPY_T2T_BC_DW_6_src_pipe_config_mask 0x0000001F +#define SDMA_PKT_COPY_T2T_BC_DW_6_src_pipe_config_shift 26 +#define SDMA_PKT_COPY_T2T_BC_DW_6_SRC_PIPE_CONFIG(x) (((x) & SDMA_PKT_COPY_T2T_BC_DW_6_src_pipe_config_mask) << SDMA_PKT_COPY_T2T_BC_DW_6_src_pipe_config_shift) + +/*define for DST_ADDR_LO word*/ +/*define for dst_addr_31_0 field*/ +#define SDMA_PKT_COPY_T2T_BC_DST_ADDR_LO_dst_addr_31_0_offset 7 +#define SDMA_PKT_COPY_T2T_BC_DST_ADDR_LO_dst_addr_31_0_mask 0xFFFFFFFF +#define SDMA_PKT_COPY_T2T_BC_DST_ADDR_LO_dst_addr_31_0_shift 0 +#define SDMA_PKT_COPY_T2T_BC_DST_ADDR_LO_DST_ADDR_31_0(x) (((x) & SDMA_PKT_COPY_T2T_BC_DST_ADDR_LO_dst_addr_31_0_mask) << SDMA_PKT_COPY_T2T_BC_DST_ADDR_LO_dst_addr_31_0_shift) + +/*define for DST_ADDR_HI word*/ +/*define for dst_addr_63_32 field*/ +#define SDMA_PKT_COPY_T2T_BC_DST_ADDR_HI_dst_addr_63_32_offset 8 +#define SDMA_PKT_COPY_T2T_BC_DST_ADDR_HI_dst_addr_63_32_mask 0xFFFFFFFF +#define SDMA_PKT_COPY_T2T_BC_DST_ADDR_HI_dst_addr_63_32_shift 0 +#define SDMA_PKT_COPY_T2T_BC_DST_ADDR_HI_DST_ADDR_63_32(x) (((x) & SDMA_PKT_COPY_T2T_BC_DST_ADDR_HI_dst_addr_63_32_mask) << SDMA_PKT_COPY_T2T_BC_DST_ADDR_HI_dst_addr_63_32_shift) + +/*define for DW_9 word*/ +/*define for dst_x field*/ +#define SDMA_PKT_COPY_T2T_BC_DW_9_dst_x_offset 9 +#define SDMA_PKT_COPY_T2T_BC_DW_9_dst_x_mask 0x00003FFF +#define SDMA_PKT_COPY_T2T_BC_DW_9_dst_x_shift 0 +#define SDMA_PKT_COPY_T2T_BC_DW_9_DST_X(x) (((x) & SDMA_PKT_COPY_T2T_BC_DW_9_dst_x_mask) << SDMA_PKT_COPY_T2T_BC_DW_9_dst_x_shift) + +/*define for dst_y field*/ +#define SDMA_PKT_COPY_T2T_BC_DW_9_dst_y_offset 9 +#define SDMA_PKT_COPY_T2T_BC_DW_9_dst_y_mask 0x00003FFF +#define SDMA_PKT_COPY_T2T_BC_DW_9_dst_y_shift 16 +#define SDMA_PKT_COPY_T2T_BC_DW_9_DST_Y(x) (((x) & SDMA_PKT_COPY_T2T_BC_DW_9_dst_y_mask) << SDMA_PKT_COPY_T2T_BC_DW_9_dst_y_shift) + +/*define for DW_10 word*/ +/*define for dst_z field*/ +#define SDMA_PKT_COPY_T2T_BC_DW_10_dst_z_offset 10 +#define SDMA_PKT_COPY_T2T_BC_DW_10_dst_z_mask 0x000007FF +#define SDMA_PKT_COPY_T2T_BC_DW_10_dst_z_shift 0 +#define SDMA_PKT_COPY_T2T_BC_DW_10_DST_Z(x) (((x) & SDMA_PKT_COPY_T2T_BC_DW_10_dst_z_mask) << SDMA_PKT_COPY_T2T_BC_DW_10_dst_z_shift) + +/*define for dst_width field*/ +#define SDMA_PKT_COPY_T2T_BC_DW_10_dst_width_offset 10 +#define SDMA_PKT_COPY_T2T_BC_DW_10_dst_width_mask 0x00003FFF +#define SDMA_PKT_COPY_T2T_BC_DW_10_dst_width_shift 16 +#define SDMA_PKT_COPY_T2T_BC_DW_10_DST_WIDTH(x) (((x) & SDMA_PKT_COPY_T2T_BC_DW_10_dst_width_mask) << SDMA_PKT_COPY_T2T_BC_DW_10_dst_width_shift) + +/*define for DW_11 word*/ +/*define for dst_height field*/ +#define SDMA_PKT_COPY_T2T_BC_DW_11_dst_height_offset 11 +#define SDMA_PKT_COPY_T2T_BC_DW_11_dst_height_mask 0x00003FFF +#define SDMA_PKT_COPY_T2T_BC_DW_11_dst_height_shift 0 +#define SDMA_PKT_COPY_T2T_BC_DW_11_DST_HEIGHT(x) (((x) & SDMA_PKT_COPY_T2T_BC_DW_11_dst_height_mask) << SDMA_PKT_COPY_T2T_BC_DW_11_dst_height_shift) + +/*define for dst_depth field*/ +#define SDMA_PKT_COPY_T2T_BC_DW_11_dst_depth_offset 11 +#define SDMA_PKT_COPY_T2T_BC_DW_11_dst_depth_mask 0x00000FFF +#define SDMA_PKT_COPY_T2T_BC_DW_11_dst_depth_shift 16 +#define SDMA_PKT_COPY_T2T_BC_DW_11_DST_DEPTH(x) (((x) & SDMA_PKT_COPY_T2T_BC_DW_11_dst_depth_mask) << SDMA_PKT_COPY_T2T_BC_DW_11_dst_depth_shift) + +/*define for DW_12 word*/ +/*define for dst_element_size field*/ +#define SDMA_PKT_COPY_T2T_BC_DW_12_dst_element_size_offset 12 +#define SDMA_PKT_COPY_T2T_BC_DW_12_dst_element_size_mask 0x00000007 +#define SDMA_PKT_COPY_T2T_BC_DW_12_dst_element_size_shift 0 +#define SDMA_PKT_COPY_T2T_BC_DW_12_DST_ELEMENT_SIZE(x) (((x) & SDMA_PKT_COPY_T2T_BC_DW_12_dst_element_size_mask) << SDMA_PKT_COPY_T2T_BC_DW_12_dst_element_size_shift) + +/*define for dst_array_mode field*/ +#define SDMA_PKT_COPY_T2T_BC_DW_12_dst_array_mode_offset 12 +#define SDMA_PKT_COPY_T2T_BC_DW_12_dst_array_mode_mask 0x0000000F +#define SDMA_PKT_COPY_T2T_BC_DW_12_dst_array_mode_shift 3 +#define SDMA_PKT_COPY_T2T_BC_DW_12_DST_ARRAY_MODE(x) (((x) & SDMA_PKT_COPY_T2T_BC_DW_12_dst_array_mode_mask) << SDMA_PKT_COPY_T2T_BC_DW_12_dst_array_mode_shift) + +/*define for dst_mit_mode field*/ +#define SDMA_PKT_COPY_T2T_BC_DW_12_dst_mit_mode_offset 12 +#define SDMA_PKT_COPY_T2T_BC_DW_12_dst_mit_mode_mask 0x00000007 +#define SDMA_PKT_COPY_T2T_BC_DW_12_dst_mit_mode_shift 8 +#define SDMA_PKT_COPY_T2T_BC_DW_12_DST_MIT_MODE(x) (((x) & SDMA_PKT_COPY_T2T_BC_DW_12_dst_mit_mode_mask) << SDMA_PKT_COPY_T2T_BC_DW_12_dst_mit_mode_shift) + +/*define for dst_tilesplit_size field*/ +#define SDMA_PKT_COPY_T2T_BC_DW_12_dst_tilesplit_size_offset 12 +#define SDMA_PKT_COPY_T2T_BC_DW_12_dst_tilesplit_size_mask 0x00000007 +#define SDMA_PKT_COPY_T2T_BC_DW_12_dst_tilesplit_size_shift 11 +#define SDMA_PKT_COPY_T2T_BC_DW_12_DST_TILESPLIT_SIZE(x) (((x) & SDMA_PKT_COPY_T2T_BC_DW_12_dst_tilesplit_size_mask) << SDMA_PKT_COPY_T2T_BC_DW_12_dst_tilesplit_size_shift) + +/*define for dst_bank_w field*/ +#define SDMA_PKT_COPY_T2T_BC_DW_12_dst_bank_w_offset 12 +#define SDMA_PKT_COPY_T2T_BC_DW_12_dst_bank_w_mask 0x00000003 +#define SDMA_PKT_COPY_T2T_BC_DW_12_dst_bank_w_shift 15 +#define SDMA_PKT_COPY_T2T_BC_DW_12_DST_BANK_W(x) (((x) & SDMA_PKT_COPY_T2T_BC_DW_12_dst_bank_w_mask) << SDMA_PKT_COPY_T2T_BC_DW_12_dst_bank_w_shift) + +/*define for dst_bank_h field*/ +#define SDMA_PKT_COPY_T2T_BC_DW_12_dst_bank_h_offset 12 +#define SDMA_PKT_COPY_T2T_BC_DW_12_dst_bank_h_mask 0x00000003 +#define SDMA_PKT_COPY_T2T_BC_DW_12_dst_bank_h_shift 18 +#define SDMA_PKT_COPY_T2T_BC_DW_12_DST_BANK_H(x) (((x) & SDMA_PKT_COPY_T2T_BC_DW_12_dst_bank_h_mask) << SDMA_PKT_COPY_T2T_BC_DW_12_dst_bank_h_shift) + +/*define for dst_num_bank field*/ +#define SDMA_PKT_COPY_T2T_BC_DW_12_dst_num_bank_offset 12 +#define SDMA_PKT_COPY_T2T_BC_DW_12_dst_num_bank_mask 0x00000003 +#define SDMA_PKT_COPY_T2T_BC_DW_12_dst_num_bank_shift 21 +#define SDMA_PKT_COPY_T2T_BC_DW_12_DST_NUM_BANK(x) (((x) & SDMA_PKT_COPY_T2T_BC_DW_12_dst_num_bank_mask) << SDMA_PKT_COPY_T2T_BC_DW_12_dst_num_bank_shift) + +/*define for dst_mat_aspt field*/ +#define SDMA_PKT_COPY_T2T_BC_DW_12_dst_mat_aspt_offset 12 +#define SDMA_PKT_COPY_T2T_BC_DW_12_dst_mat_aspt_mask 0x00000003 +#define SDMA_PKT_COPY_T2T_BC_DW_12_dst_mat_aspt_shift 24 +#define SDMA_PKT_COPY_T2T_BC_DW_12_DST_MAT_ASPT(x) (((x) & SDMA_PKT_COPY_T2T_BC_DW_12_dst_mat_aspt_mask) << SDMA_PKT_COPY_T2T_BC_DW_12_dst_mat_aspt_shift) + +/*define for dst_pipe_config field*/ +#define SDMA_PKT_COPY_T2T_BC_DW_12_dst_pipe_config_offset 12 +#define SDMA_PKT_COPY_T2T_BC_DW_12_dst_pipe_config_mask 0x0000001F +#define SDMA_PKT_COPY_T2T_BC_DW_12_dst_pipe_config_shift 26 +#define SDMA_PKT_COPY_T2T_BC_DW_12_DST_PIPE_CONFIG(x) (((x) & SDMA_PKT_COPY_T2T_BC_DW_12_dst_pipe_config_mask) << SDMA_PKT_COPY_T2T_BC_DW_12_dst_pipe_config_shift) + +/*define for DW_13 word*/ +/*define for rect_x field*/ +#define SDMA_PKT_COPY_T2T_BC_DW_13_rect_x_offset 13 +#define SDMA_PKT_COPY_T2T_BC_DW_13_rect_x_mask 0x00003FFF +#define SDMA_PKT_COPY_T2T_BC_DW_13_rect_x_shift 0 +#define SDMA_PKT_COPY_T2T_BC_DW_13_RECT_X(x) (((x) & SDMA_PKT_COPY_T2T_BC_DW_13_rect_x_mask) << SDMA_PKT_COPY_T2T_BC_DW_13_rect_x_shift) + +/*define for rect_y field*/ +#define SDMA_PKT_COPY_T2T_BC_DW_13_rect_y_offset 13 +#define SDMA_PKT_COPY_T2T_BC_DW_13_rect_y_mask 0x00003FFF +#define SDMA_PKT_COPY_T2T_BC_DW_13_rect_y_shift 16 +#define SDMA_PKT_COPY_T2T_BC_DW_13_RECT_Y(x) (((x) & SDMA_PKT_COPY_T2T_BC_DW_13_rect_y_mask) << SDMA_PKT_COPY_T2T_BC_DW_13_rect_y_shift) + +/*define for DW_14 word*/ +/*define for rect_z field*/ +#define SDMA_PKT_COPY_T2T_BC_DW_14_rect_z_offset 14 +#define SDMA_PKT_COPY_T2T_BC_DW_14_rect_z_mask 0x000007FF +#define SDMA_PKT_COPY_T2T_BC_DW_14_rect_z_shift 0 +#define SDMA_PKT_COPY_T2T_BC_DW_14_RECT_Z(x) (((x) & SDMA_PKT_COPY_T2T_BC_DW_14_rect_z_mask) << SDMA_PKT_COPY_T2T_BC_DW_14_rect_z_shift) + +/*define for dst_sw field*/ +#define SDMA_PKT_COPY_T2T_BC_DW_14_dst_sw_offset 14 +#define SDMA_PKT_COPY_T2T_BC_DW_14_dst_sw_mask 0x00000003 +#define SDMA_PKT_COPY_T2T_BC_DW_14_dst_sw_shift 16 +#define SDMA_PKT_COPY_T2T_BC_DW_14_DST_SW(x) (((x) & SDMA_PKT_COPY_T2T_BC_DW_14_dst_sw_mask) << SDMA_PKT_COPY_T2T_BC_DW_14_dst_sw_shift) + +/*define for src_sw field*/ +#define SDMA_PKT_COPY_T2T_BC_DW_14_src_sw_offset 14 +#define SDMA_PKT_COPY_T2T_BC_DW_14_src_sw_mask 0x00000003 +#define SDMA_PKT_COPY_T2T_BC_DW_14_src_sw_shift 24 +#define SDMA_PKT_COPY_T2T_BC_DW_14_SRC_SW(x) (((x) & SDMA_PKT_COPY_T2T_BC_DW_14_src_sw_mask) << SDMA_PKT_COPY_T2T_BC_DW_14_src_sw_shift) + + +/* +** Definitions for SDMA_PKT_COPY_TILED_SUBWIN packet +*/ + +/*define for HEADER word*/ +/*define for op field*/ +#define SDMA_PKT_COPY_TILED_SUBWIN_HEADER_op_offset 0 +#define SDMA_PKT_COPY_TILED_SUBWIN_HEADER_op_mask 0x000000FF +#define SDMA_PKT_COPY_TILED_SUBWIN_HEADER_op_shift 0 +#define SDMA_PKT_COPY_TILED_SUBWIN_HEADER_OP(x) (((x) & SDMA_PKT_COPY_TILED_SUBWIN_HEADER_op_mask) << SDMA_PKT_COPY_TILED_SUBWIN_HEADER_op_shift) + +/*define for sub_op field*/ +#define SDMA_PKT_COPY_TILED_SUBWIN_HEADER_sub_op_offset 0 +#define SDMA_PKT_COPY_TILED_SUBWIN_HEADER_sub_op_mask 0x000000FF +#define SDMA_PKT_COPY_TILED_SUBWIN_HEADER_sub_op_shift 8 +#define SDMA_PKT_COPY_TILED_SUBWIN_HEADER_SUB_OP(x) (((x) & SDMA_PKT_COPY_TILED_SUBWIN_HEADER_sub_op_mask) << SDMA_PKT_COPY_TILED_SUBWIN_HEADER_sub_op_shift) + +/*define for tmz field*/ +#define SDMA_PKT_COPY_TILED_SUBWIN_HEADER_tmz_offset 0 +#define SDMA_PKT_COPY_TILED_SUBWIN_HEADER_tmz_mask 0x00000001 +#define SDMA_PKT_COPY_TILED_SUBWIN_HEADER_tmz_shift 18 +#define SDMA_PKT_COPY_TILED_SUBWIN_HEADER_TMZ(x) (((x) & SDMA_PKT_COPY_TILED_SUBWIN_HEADER_tmz_mask) << SDMA_PKT_COPY_TILED_SUBWIN_HEADER_tmz_shift) + +/*define for dcc field*/ +#define SDMA_PKT_COPY_TILED_SUBWIN_HEADER_dcc_offset 0 +#define SDMA_PKT_COPY_TILED_SUBWIN_HEADER_dcc_mask 0x00000001 +#define SDMA_PKT_COPY_TILED_SUBWIN_HEADER_dcc_shift 19 +#define SDMA_PKT_COPY_TILED_SUBWIN_HEADER_DCC(x) (((x) & SDMA_PKT_COPY_TILED_SUBWIN_HEADER_dcc_mask) << SDMA_PKT_COPY_TILED_SUBWIN_HEADER_dcc_shift) + +/*define for cpv field*/ +#define SDMA_PKT_COPY_TILED_SUBWIN_HEADER_cpv_offset 0 +#define SDMA_PKT_COPY_TILED_SUBWIN_HEADER_cpv_mask 0x00000001 +#define SDMA_PKT_COPY_TILED_SUBWIN_HEADER_cpv_shift 28 +#define SDMA_PKT_COPY_TILED_SUBWIN_HEADER_CPV(x) (((x) & SDMA_PKT_COPY_TILED_SUBWIN_HEADER_cpv_mask) << SDMA_PKT_COPY_TILED_SUBWIN_HEADER_cpv_shift) + +/*define for detile field*/ +#define SDMA_PKT_COPY_TILED_SUBWIN_HEADER_detile_offset 0 +#define SDMA_PKT_COPY_TILED_SUBWIN_HEADER_detile_mask 0x00000001 +#define SDMA_PKT_COPY_TILED_SUBWIN_HEADER_detile_shift 31 +#define SDMA_PKT_COPY_TILED_SUBWIN_HEADER_DETILE(x) (((x) & SDMA_PKT_COPY_TILED_SUBWIN_HEADER_detile_mask) << SDMA_PKT_COPY_TILED_SUBWIN_HEADER_detile_shift) + +/*define for TILED_ADDR_LO word*/ +/*define for tiled_addr_31_0 field*/ +#define SDMA_PKT_COPY_TILED_SUBWIN_TILED_ADDR_LO_tiled_addr_31_0_offset 1 +#define SDMA_PKT_COPY_TILED_SUBWIN_TILED_ADDR_LO_tiled_addr_31_0_mask 0xFFFFFFFF +#define SDMA_PKT_COPY_TILED_SUBWIN_TILED_ADDR_LO_tiled_addr_31_0_shift 0 +#define SDMA_PKT_COPY_TILED_SUBWIN_TILED_ADDR_LO_TILED_ADDR_31_0(x) (((x) & SDMA_PKT_COPY_TILED_SUBWIN_TILED_ADDR_LO_tiled_addr_31_0_mask) << SDMA_PKT_COPY_TILED_SUBWIN_TILED_ADDR_LO_tiled_addr_31_0_shift) + +/*define for TILED_ADDR_HI word*/ +/*define for tiled_addr_63_32 field*/ +#define SDMA_PKT_COPY_TILED_SUBWIN_TILED_ADDR_HI_tiled_addr_63_32_offset 2 +#define SDMA_PKT_COPY_TILED_SUBWIN_TILED_ADDR_HI_tiled_addr_63_32_mask 0xFFFFFFFF +#define SDMA_PKT_COPY_TILED_SUBWIN_TILED_ADDR_HI_tiled_addr_63_32_shift 0 +#define SDMA_PKT_COPY_TILED_SUBWIN_TILED_ADDR_HI_TILED_ADDR_63_32(x) (((x) & SDMA_PKT_COPY_TILED_SUBWIN_TILED_ADDR_HI_tiled_addr_63_32_mask) << SDMA_PKT_COPY_TILED_SUBWIN_TILED_ADDR_HI_tiled_addr_63_32_shift) + +/*define for DW_3 word*/ +/*define for tiled_x field*/ +#define SDMA_PKT_COPY_TILED_SUBWIN_DW_3_tiled_x_offset 3 +#define SDMA_PKT_COPY_TILED_SUBWIN_DW_3_tiled_x_mask 0x00003FFF +#define SDMA_PKT_COPY_TILED_SUBWIN_DW_3_tiled_x_shift 0 +#define SDMA_PKT_COPY_TILED_SUBWIN_DW_3_TILED_X(x) (((x) & SDMA_PKT_COPY_TILED_SUBWIN_DW_3_tiled_x_mask) << SDMA_PKT_COPY_TILED_SUBWIN_DW_3_tiled_x_shift) + +/*define for tiled_y field*/ +#define SDMA_PKT_COPY_TILED_SUBWIN_DW_3_tiled_y_offset 3 +#define SDMA_PKT_COPY_TILED_SUBWIN_DW_3_tiled_y_mask 0x00003FFF +#define SDMA_PKT_COPY_TILED_SUBWIN_DW_3_tiled_y_shift 16 +#define SDMA_PKT_COPY_TILED_SUBWIN_DW_3_TILED_Y(x) (((x) & SDMA_PKT_COPY_TILED_SUBWIN_DW_3_tiled_y_mask) << SDMA_PKT_COPY_TILED_SUBWIN_DW_3_tiled_y_shift) + +/*define for DW_4 word*/ +/*define for tiled_z field*/ +#define SDMA_PKT_COPY_TILED_SUBWIN_DW_4_tiled_z_offset 4 +#define SDMA_PKT_COPY_TILED_SUBWIN_DW_4_tiled_z_mask 0x00001FFF +#define SDMA_PKT_COPY_TILED_SUBWIN_DW_4_tiled_z_shift 0 +#define SDMA_PKT_COPY_TILED_SUBWIN_DW_4_TILED_Z(x) (((x) & SDMA_PKT_COPY_TILED_SUBWIN_DW_4_tiled_z_mask) << SDMA_PKT_COPY_TILED_SUBWIN_DW_4_tiled_z_shift) + +/*define for width field*/ +#define SDMA_PKT_COPY_TILED_SUBWIN_DW_4_width_offset 4 +#define SDMA_PKT_COPY_TILED_SUBWIN_DW_4_width_mask 0x00003FFF +#define SDMA_PKT_COPY_TILED_SUBWIN_DW_4_width_shift 16 +#define SDMA_PKT_COPY_TILED_SUBWIN_DW_4_WIDTH(x) (((x) & SDMA_PKT_COPY_TILED_SUBWIN_DW_4_width_mask) << SDMA_PKT_COPY_TILED_SUBWIN_DW_4_width_shift) + +/*define for DW_5 word*/ +/*define for height field*/ +#define SDMA_PKT_COPY_TILED_SUBWIN_DW_5_height_offset 5 +#define SDMA_PKT_COPY_TILED_SUBWIN_DW_5_height_mask 0x00003FFF +#define SDMA_PKT_COPY_TILED_SUBWIN_DW_5_height_shift 0 +#define SDMA_PKT_COPY_TILED_SUBWIN_DW_5_HEIGHT(x) (((x) & SDMA_PKT_COPY_TILED_SUBWIN_DW_5_height_mask) << SDMA_PKT_COPY_TILED_SUBWIN_DW_5_height_shift) + +/*define for depth field*/ +#define SDMA_PKT_COPY_TILED_SUBWIN_DW_5_depth_offset 5 +#define SDMA_PKT_COPY_TILED_SUBWIN_DW_5_depth_mask 0x00001FFF +#define SDMA_PKT_COPY_TILED_SUBWIN_DW_5_depth_shift 16 +#define SDMA_PKT_COPY_TILED_SUBWIN_DW_5_DEPTH(x) (((x) & SDMA_PKT_COPY_TILED_SUBWIN_DW_5_depth_mask) << SDMA_PKT_COPY_TILED_SUBWIN_DW_5_depth_shift) + +/*define for DW_6 word*/ +/*define for element_size field*/ +#define SDMA_PKT_COPY_TILED_SUBWIN_DW_6_element_size_offset 6 +#define SDMA_PKT_COPY_TILED_SUBWIN_DW_6_element_size_mask 0x00000007 +#define SDMA_PKT_COPY_TILED_SUBWIN_DW_6_element_size_shift 0 +#define SDMA_PKT_COPY_TILED_SUBWIN_DW_6_ELEMENT_SIZE(x) (((x) & SDMA_PKT_COPY_TILED_SUBWIN_DW_6_element_size_mask) << SDMA_PKT_COPY_TILED_SUBWIN_DW_6_element_size_shift) + +/*define for swizzle_mode field*/ +#define SDMA_PKT_COPY_TILED_SUBWIN_DW_6_swizzle_mode_offset 6 +#define SDMA_PKT_COPY_TILED_SUBWIN_DW_6_swizzle_mode_mask 0x0000001F +#define SDMA_PKT_COPY_TILED_SUBWIN_DW_6_swizzle_mode_shift 3 +#define SDMA_PKT_COPY_TILED_SUBWIN_DW_6_SWIZZLE_MODE(x) (((x) & SDMA_PKT_COPY_TILED_SUBWIN_DW_6_swizzle_mode_mask) << SDMA_PKT_COPY_TILED_SUBWIN_DW_6_swizzle_mode_shift) + +/*define for dimension field*/ +#define SDMA_PKT_COPY_TILED_SUBWIN_DW_6_dimension_offset 6 +#define SDMA_PKT_COPY_TILED_SUBWIN_DW_6_dimension_mask 0x00000003 +#define SDMA_PKT_COPY_TILED_SUBWIN_DW_6_dimension_shift 9 +#define SDMA_PKT_COPY_TILED_SUBWIN_DW_6_DIMENSION(x) (((x) & SDMA_PKT_COPY_TILED_SUBWIN_DW_6_dimension_mask) << SDMA_PKT_COPY_TILED_SUBWIN_DW_6_dimension_shift) + +/*define for mip_max field*/ +#define SDMA_PKT_COPY_TILED_SUBWIN_DW_6_mip_max_offset 6 +#define SDMA_PKT_COPY_TILED_SUBWIN_DW_6_mip_max_mask 0x0000000F +#define SDMA_PKT_COPY_TILED_SUBWIN_DW_6_mip_max_shift 16 +#define SDMA_PKT_COPY_TILED_SUBWIN_DW_6_MIP_MAX(x) (((x) & SDMA_PKT_COPY_TILED_SUBWIN_DW_6_mip_max_mask) << SDMA_PKT_COPY_TILED_SUBWIN_DW_6_mip_max_shift) + +/*define for mip_id field*/ +#define SDMA_PKT_COPY_TILED_SUBWIN_DW_6_mip_id_offset 6 +#define SDMA_PKT_COPY_TILED_SUBWIN_DW_6_mip_id_mask 0x0000000F +#define SDMA_PKT_COPY_TILED_SUBWIN_DW_6_mip_id_shift 20 +#define SDMA_PKT_COPY_TILED_SUBWIN_DW_6_MIP_ID(x) (((x) & SDMA_PKT_COPY_TILED_SUBWIN_DW_6_mip_id_mask) << SDMA_PKT_COPY_TILED_SUBWIN_DW_6_mip_id_shift) + +/*define for LINEAR_ADDR_LO word*/ +/*define for linear_addr_31_0 field*/ +#define SDMA_PKT_COPY_TILED_SUBWIN_LINEAR_ADDR_LO_linear_addr_31_0_offset 7 +#define SDMA_PKT_COPY_TILED_SUBWIN_LINEAR_ADDR_LO_linear_addr_31_0_mask 0xFFFFFFFF +#define SDMA_PKT_COPY_TILED_SUBWIN_LINEAR_ADDR_LO_linear_addr_31_0_shift 0 +#define SDMA_PKT_COPY_TILED_SUBWIN_LINEAR_ADDR_LO_LINEAR_ADDR_31_0(x) (((x) & SDMA_PKT_COPY_TILED_SUBWIN_LINEAR_ADDR_LO_linear_addr_31_0_mask) << SDMA_PKT_COPY_TILED_SUBWIN_LINEAR_ADDR_LO_linear_addr_31_0_shift) + +/*define for LINEAR_ADDR_HI word*/ +/*define for linear_addr_63_32 field*/ +#define SDMA_PKT_COPY_TILED_SUBWIN_LINEAR_ADDR_HI_linear_addr_63_32_offset 8 +#define SDMA_PKT_COPY_TILED_SUBWIN_LINEAR_ADDR_HI_linear_addr_63_32_mask 0xFFFFFFFF +#define SDMA_PKT_COPY_TILED_SUBWIN_LINEAR_ADDR_HI_linear_addr_63_32_shift 0 +#define SDMA_PKT_COPY_TILED_SUBWIN_LINEAR_ADDR_HI_LINEAR_ADDR_63_32(x) (((x) & SDMA_PKT_COPY_TILED_SUBWIN_LINEAR_ADDR_HI_linear_addr_63_32_mask) << SDMA_PKT_COPY_TILED_SUBWIN_LINEAR_ADDR_HI_linear_addr_63_32_shift) + +/*define for DW_9 word*/ +/*define for linear_x field*/ +#define SDMA_PKT_COPY_TILED_SUBWIN_DW_9_linear_x_offset 9 +#define SDMA_PKT_COPY_TILED_SUBWIN_DW_9_linear_x_mask 0x00003FFF +#define SDMA_PKT_COPY_TILED_SUBWIN_DW_9_linear_x_shift 0 +#define SDMA_PKT_COPY_TILED_SUBWIN_DW_9_LINEAR_X(x) (((x) & SDMA_PKT_COPY_TILED_SUBWIN_DW_9_linear_x_mask) << SDMA_PKT_COPY_TILED_SUBWIN_DW_9_linear_x_shift) + +/*define for linear_y field*/ +#define SDMA_PKT_COPY_TILED_SUBWIN_DW_9_linear_y_offset 9 +#define SDMA_PKT_COPY_TILED_SUBWIN_DW_9_linear_y_mask 0x00003FFF +#define SDMA_PKT_COPY_TILED_SUBWIN_DW_9_linear_y_shift 16 +#define SDMA_PKT_COPY_TILED_SUBWIN_DW_9_LINEAR_Y(x) (((x) & SDMA_PKT_COPY_TILED_SUBWIN_DW_9_linear_y_mask) << SDMA_PKT_COPY_TILED_SUBWIN_DW_9_linear_y_shift) + +/*define for DW_10 word*/ +/*define for linear_z field*/ +#define SDMA_PKT_COPY_TILED_SUBWIN_DW_10_linear_z_offset 10 +#define SDMA_PKT_COPY_TILED_SUBWIN_DW_10_linear_z_mask 0x00001FFF +#define SDMA_PKT_COPY_TILED_SUBWIN_DW_10_linear_z_shift 0 +#define SDMA_PKT_COPY_TILED_SUBWIN_DW_10_LINEAR_Z(x) (((x) & SDMA_PKT_COPY_TILED_SUBWIN_DW_10_linear_z_mask) << SDMA_PKT_COPY_TILED_SUBWIN_DW_10_linear_z_shift) + +/*define for linear_pitch field*/ +#define SDMA_PKT_COPY_TILED_SUBWIN_DW_10_linear_pitch_offset 10 +#define SDMA_PKT_COPY_TILED_SUBWIN_DW_10_linear_pitch_mask 0x00003FFF +#define SDMA_PKT_COPY_TILED_SUBWIN_DW_10_linear_pitch_shift 16 +#define SDMA_PKT_COPY_TILED_SUBWIN_DW_10_LINEAR_PITCH(x) (((x) & SDMA_PKT_COPY_TILED_SUBWIN_DW_10_linear_pitch_mask) << SDMA_PKT_COPY_TILED_SUBWIN_DW_10_linear_pitch_shift) + +/*define for DW_11 word*/ +/*define for linear_slice_pitch field*/ +#define SDMA_PKT_COPY_TILED_SUBWIN_DW_11_linear_slice_pitch_offset 11 +#define SDMA_PKT_COPY_TILED_SUBWIN_DW_11_linear_slice_pitch_mask 0x0FFFFFFF +#define SDMA_PKT_COPY_TILED_SUBWIN_DW_11_linear_slice_pitch_shift 0 +#define SDMA_PKT_COPY_TILED_SUBWIN_DW_11_LINEAR_SLICE_PITCH(x) (((x) & SDMA_PKT_COPY_TILED_SUBWIN_DW_11_linear_slice_pitch_mask) << SDMA_PKT_COPY_TILED_SUBWIN_DW_11_linear_slice_pitch_shift) + +/*define for DW_12 word*/ +/*define for rect_x field*/ +#define SDMA_PKT_COPY_TILED_SUBWIN_DW_12_rect_x_offset 12 +#define SDMA_PKT_COPY_TILED_SUBWIN_DW_12_rect_x_mask 0x00003FFF +#define SDMA_PKT_COPY_TILED_SUBWIN_DW_12_rect_x_shift 0 +#define SDMA_PKT_COPY_TILED_SUBWIN_DW_12_RECT_X(x) (((x) & SDMA_PKT_COPY_TILED_SUBWIN_DW_12_rect_x_mask) << SDMA_PKT_COPY_TILED_SUBWIN_DW_12_rect_x_shift) + +/*define for rect_y field*/ +#define SDMA_PKT_COPY_TILED_SUBWIN_DW_12_rect_y_offset 12 +#define SDMA_PKT_COPY_TILED_SUBWIN_DW_12_rect_y_mask 0x00003FFF +#define SDMA_PKT_COPY_TILED_SUBWIN_DW_12_rect_y_shift 16 +#define SDMA_PKT_COPY_TILED_SUBWIN_DW_12_RECT_Y(x) (((x) & SDMA_PKT_COPY_TILED_SUBWIN_DW_12_rect_y_mask) << SDMA_PKT_COPY_TILED_SUBWIN_DW_12_rect_y_shift) + +/*define for DW_13 word*/ +/*define for rect_z field*/ +#define SDMA_PKT_COPY_TILED_SUBWIN_DW_13_rect_z_offset 13 +#define SDMA_PKT_COPY_TILED_SUBWIN_DW_13_rect_z_mask 0x00001FFF +#define SDMA_PKT_COPY_TILED_SUBWIN_DW_13_rect_z_shift 0 +#define SDMA_PKT_COPY_TILED_SUBWIN_DW_13_RECT_Z(x) (((x) & SDMA_PKT_COPY_TILED_SUBWIN_DW_13_rect_z_mask) << SDMA_PKT_COPY_TILED_SUBWIN_DW_13_rect_z_shift) + +/*define for linear_sw field*/ +#define SDMA_PKT_COPY_TILED_SUBWIN_DW_13_linear_sw_offset 13 +#define SDMA_PKT_COPY_TILED_SUBWIN_DW_13_linear_sw_mask 0x00000003 +#define SDMA_PKT_COPY_TILED_SUBWIN_DW_13_linear_sw_shift 16 +#define SDMA_PKT_COPY_TILED_SUBWIN_DW_13_LINEAR_SW(x) (((x) & SDMA_PKT_COPY_TILED_SUBWIN_DW_13_linear_sw_mask) << SDMA_PKT_COPY_TILED_SUBWIN_DW_13_linear_sw_shift) + +/*define for linear_cache_policy field*/ +#define SDMA_PKT_COPY_TILED_SUBWIN_DW_13_linear_cache_policy_offset 13 +#define SDMA_PKT_COPY_TILED_SUBWIN_DW_13_linear_cache_policy_mask 0x00000007 +#define SDMA_PKT_COPY_TILED_SUBWIN_DW_13_linear_cache_policy_shift 18 +#define SDMA_PKT_COPY_TILED_SUBWIN_DW_13_LINEAR_CACHE_POLICY(x) (((x) & SDMA_PKT_COPY_TILED_SUBWIN_DW_13_linear_cache_policy_mask) << SDMA_PKT_COPY_TILED_SUBWIN_DW_13_linear_cache_policy_shift) + +/*define for tile_sw field*/ +#define SDMA_PKT_COPY_TILED_SUBWIN_DW_13_tile_sw_offset 13 +#define SDMA_PKT_COPY_TILED_SUBWIN_DW_13_tile_sw_mask 0x00000003 +#define SDMA_PKT_COPY_TILED_SUBWIN_DW_13_tile_sw_shift 24 +#define SDMA_PKT_COPY_TILED_SUBWIN_DW_13_TILE_SW(x) (((x) & SDMA_PKT_COPY_TILED_SUBWIN_DW_13_tile_sw_mask) << SDMA_PKT_COPY_TILED_SUBWIN_DW_13_tile_sw_shift) + +/*define for tile_cache_policy field*/ +#define SDMA_PKT_COPY_TILED_SUBWIN_DW_13_tile_cache_policy_offset 13 +#define SDMA_PKT_COPY_TILED_SUBWIN_DW_13_tile_cache_policy_mask 0x00000007 +#define SDMA_PKT_COPY_TILED_SUBWIN_DW_13_tile_cache_policy_shift 26 +#define SDMA_PKT_COPY_TILED_SUBWIN_DW_13_TILE_CACHE_POLICY(x) (((x) & SDMA_PKT_COPY_TILED_SUBWIN_DW_13_tile_cache_policy_mask) << SDMA_PKT_COPY_TILED_SUBWIN_DW_13_tile_cache_policy_shift) + +/*define for META_ADDR_LO word*/ +/*define for meta_addr_31_0 field*/ +#define SDMA_PKT_COPY_TILED_SUBWIN_META_ADDR_LO_meta_addr_31_0_offset 14 +#define SDMA_PKT_COPY_TILED_SUBWIN_META_ADDR_LO_meta_addr_31_0_mask 0xFFFFFFFF +#define SDMA_PKT_COPY_TILED_SUBWIN_META_ADDR_LO_meta_addr_31_0_shift 0 +#define SDMA_PKT_COPY_TILED_SUBWIN_META_ADDR_LO_META_ADDR_31_0(x) (((x) & SDMA_PKT_COPY_TILED_SUBWIN_META_ADDR_LO_meta_addr_31_0_mask) << SDMA_PKT_COPY_TILED_SUBWIN_META_ADDR_LO_meta_addr_31_0_shift) + +/*define for META_ADDR_HI word*/ +/*define for meta_addr_63_32 field*/ +#define SDMA_PKT_COPY_TILED_SUBWIN_META_ADDR_HI_meta_addr_63_32_offset 15 +#define SDMA_PKT_COPY_TILED_SUBWIN_META_ADDR_HI_meta_addr_63_32_mask 0xFFFFFFFF +#define SDMA_PKT_COPY_TILED_SUBWIN_META_ADDR_HI_meta_addr_63_32_shift 0 +#define SDMA_PKT_COPY_TILED_SUBWIN_META_ADDR_HI_META_ADDR_63_32(x) (((x) & SDMA_PKT_COPY_TILED_SUBWIN_META_ADDR_HI_meta_addr_63_32_mask) << SDMA_PKT_COPY_TILED_SUBWIN_META_ADDR_HI_meta_addr_63_32_shift) + +/*define for META_CONFIG word*/ +/*define for data_format field*/ +#define SDMA_PKT_COPY_TILED_SUBWIN_META_CONFIG_data_format_offset 16 +#define SDMA_PKT_COPY_TILED_SUBWIN_META_CONFIG_data_format_mask 0x0000007F +#define SDMA_PKT_COPY_TILED_SUBWIN_META_CONFIG_data_format_shift 0 +#define SDMA_PKT_COPY_TILED_SUBWIN_META_CONFIG_DATA_FORMAT(x) (((x) & SDMA_PKT_COPY_TILED_SUBWIN_META_CONFIG_data_format_mask) << SDMA_PKT_COPY_TILED_SUBWIN_META_CONFIG_data_format_shift) + +/*define for color_transform_disable field*/ +#define SDMA_PKT_COPY_TILED_SUBWIN_META_CONFIG_color_transform_disable_offset 16 +#define SDMA_PKT_COPY_TILED_SUBWIN_META_CONFIG_color_transform_disable_mask 0x00000001 +#define SDMA_PKT_COPY_TILED_SUBWIN_META_CONFIG_color_transform_disable_shift 7 +#define SDMA_PKT_COPY_TILED_SUBWIN_META_CONFIG_COLOR_TRANSFORM_DISABLE(x) (((x) & SDMA_PKT_COPY_TILED_SUBWIN_META_CONFIG_color_transform_disable_mask) << SDMA_PKT_COPY_TILED_SUBWIN_META_CONFIG_color_transform_disable_shift) + +/*define for alpha_is_on_msb field*/ +#define SDMA_PKT_COPY_TILED_SUBWIN_META_CONFIG_alpha_is_on_msb_offset 16 +#define SDMA_PKT_COPY_TILED_SUBWIN_META_CONFIG_alpha_is_on_msb_mask 0x00000001 +#define SDMA_PKT_COPY_TILED_SUBWIN_META_CONFIG_alpha_is_on_msb_shift 8 +#define SDMA_PKT_COPY_TILED_SUBWIN_META_CONFIG_ALPHA_IS_ON_MSB(x) (((x) & SDMA_PKT_COPY_TILED_SUBWIN_META_CONFIG_alpha_is_on_msb_mask) << SDMA_PKT_COPY_TILED_SUBWIN_META_CONFIG_alpha_is_on_msb_shift) + +/*define for number_type field*/ +#define SDMA_PKT_COPY_TILED_SUBWIN_META_CONFIG_number_type_offset 16 +#define SDMA_PKT_COPY_TILED_SUBWIN_META_CONFIG_number_type_mask 0x00000007 +#define SDMA_PKT_COPY_TILED_SUBWIN_META_CONFIG_number_type_shift 9 +#define SDMA_PKT_COPY_TILED_SUBWIN_META_CONFIG_NUMBER_TYPE(x) (((x) & SDMA_PKT_COPY_TILED_SUBWIN_META_CONFIG_number_type_mask) << SDMA_PKT_COPY_TILED_SUBWIN_META_CONFIG_number_type_shift) + +/*define for surface_type field*/ +#define SDMA_PKT_COPY_TILED_SUBWIN_META_CONFIG_surface_type_offset 16 +#define SDMA_PKT_COPY_TILED_SUBWIN_META_CONFIG_surface_type_mask 0x00000003 +#define SDMA_PKT_COPY_TILED_SUBWIN_META_CONFIG_surface_type_shift 12 +#define SDMA_PKT_COPY_TILED_SUBWIN_META_CONFIG_SURFACE_TYPE(x) (((x) & SDMA_PKT_COPY_TILED_SUBWIN_META_CONFIG_surface_type_mask) << SDMA_PKT_COPY_TILED_SUBWIN_META_CONFIG_surface_type_shift) + +/*define for meta_llc field*/ +#define SDMA_PKT_COPY_TILED_SUBWIN_META_CONFIG_meta_llc_offset 16 +#define SDMA_PKT_COPY_TILED_SUBWIN_META_CONFIG_meta_llc_mask 0x00000001 +#define SDMA_PKT_COPY_TILED_SUBWIN_META_CONFIG_meta_llc_shift 14 +#define SDMA_PKT_COPY_TILED_SUBWIN_META_CONFIG_META_LLC(x) (((x) & SDMA_PKT_COPY_TILED_SUBWIN_META_CONFIG_meta_llc_mask) << SDMA_PKT_COPY_TILED_SUBWIN_META_CONFIG_meta_llc_shift) + +/*define for max_comp_block_size field*/ +#define SDMA_PKT_COPY_TILED_SUBWIN_META_CONFIG_max_comp_block_size_offset 16 +#define SDMA_PKT_COPY_TILED_SUBWIN_META_CONFIG_max_comp_block_size_mask 0x00000003 +#define SDMA_PKT_COPY_TILED_SUBWIN_META_CONFIG_max_comp_block_size_shift 24 +#define SDMA_PKT_COPY_TILED_SUBWIN_META_CONFIG_MAX_COMP_BLOCK_SIZE(x) (((x) & SDMA_PKT_COPY_TILED_SUBWIN_META_CONFIG_max_comp_block_size_mask) << SDMA_PKT_COPY_TILED_SUBWIN_META_CONFIG_max_comp_block_size_shift) + +/*define for max_uncomp_block_size field*/ +#define SDMA_PKT_COPY_TILED_SUBWIN_META_CONFIG_max_uncomp_block_size_offset 16 +#define SDMA_PKT_COPY_TILED_SUBWIN_META_CONFIG_max_uncomp_block_size_mask 0x00000003 +#define SDMA_PKT_COPY_TILED_SUBWIN_META_CONFIG_max_uncomp_block_size_shift 26 +#define SDMA_PKT_COPY_TILED_SUBWIN_META_CONFIG_MAX_UNCOMP_BLOCK_SIZE(x) (((x) & SDMA_PKT_COPY_TILED_SUBWIN_META_CONFIG_max_uncomp_block_size_mask) << SDMA_PKT_COPY_TILED_SUBWIN_META_CONFIG_max_uncomp_block_size_shift) + +/*define for write_compress_enable field*/ +#define SDMA_PKT_COPY_TILED_SUBWIN_META_CONFIG_write_compress_enable_offset 16 +#define SDMA_PKT_COPY_TILED_SUBWIN_META_CONFIG_write_compress_enable_mask 0x00000001 +#define SDMA_PKT_COPY_TILED_SUBWIN_META_CONFIG_write_compress_enable_shift 28 +#define SDMA_PKT_COPY_TILED_SUBWIN_META_CONFIG_WRITE_COMPRESS_ENABLE(x) (((x) & SDMA_PKT_COPY_TILED_SUBWIN_META_CONFIG_write_compress_enable_mask) << SDMA_PKT_COPY_TILED_SUBWIN_META_CONFIG_write_compress_enable_shift) + +/*define for meta_tmz field*/ +#define SDMA_PKT_COPY_TILED_SUBWIN_META_CONFIG_meta_tmz_offset 16 +#define SDMA_PKT_COPY_TILED_SUBWIN_META_CONFIG_meta_tmz_mask 0x00000001 +#define SDMA_PKT_COPY_TILED_SUBWIN_META_CONFIG_meta_tmz_shift 29 +#define SDMA_PKT_COPY_TILED_SUBWIN_META_CONFIG_META_TMZ(x) (((x) & SDMA_PKT_COPY_TILED_SUBWIN_META_CONFIG_meta_tmz_mask) << SDMA_PKT_COPY_TILED_SUBWIN_META_CONFIG_meta_tmz_shift) + +/*define for pipe_aligned field*/ +#define SDMA_PKT_COPY_TILED_SUBWIN_META_CONFIG_pipe_aligned_offset 16 +#define SDMA_PKT_COPY_TILED_SUBWIN_META_CONFIG_pipe_aligned_mask 0x00000001 +#define SDMA_PKT_COPY_TILED_SUBWIN_META_CONFIG_pipe_aligned_shift 31 +#define SDMA_PKT_COPY_TILED_SUBWIN_META_CONFIG_PIPE_ALIGNED(x) (((x) & SDMA_PKT_COPY_TILED_SUBWIN_META_CONFIG_pipe_aligned_mask) << SDMA_PKT_COPY_TILED_SUBWIN_META_CONFIG_pipe_aligned_shift) + + +/* +** Definitions for SDMA_PKT_COPY_TILED_SUBWIN_BC packet +*/ + +/*define for HEADER word*/ +/*define for op field*/ +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_HEADER_op_offset 0 +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_HEADER_op_mask 0x000000FF +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_HEADER_op_shift 0 +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_HEADER_OP(x) (((x) & SDMA_PKT_COPY_TILED_SUBWIN_BC_HEADER_op_mask) << SDMA_PKT_COPY_TILED_SUBWIN_BC_HEADER_op_shift) + +/*define for sub_op field*/ +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_HEADER_sub_op_offset 0 +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_HEADER_sub_op_mask 0x000000FF +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_HEADER_sub_op_shift 8 +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_HEADER_SUB_OP(x) (((x) & SDMA_PKT_COPY_TILED_SUBWIN_BC_HEADER_sub_op_mask) << SDMA_PKT_COPY_TILED_SUBWIN_BC_HEADER_sub_op_shift) + +/*define for detile field*/ +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_HEADER_detile_offset 0 +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_HEADER_detile_mask 0x00000001 +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_HEADER_detile_shift 31 +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_HEADER_DETILE(x) (((x) & SDMA_PKT_COPY_TILED_SUBWIN_BC_HEADER_detile_mask) << SDMA_PKT_COPY_TILED_SUBWIN_BC_HEADER_detile_shift) + +/*define for TILED_ADDR_LO word*/ +/*define for tiled_addr_31_0 field*/ +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_TILED_ADDR_LO_tiled_addr_31_0_offset 1 +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_TILED_ADDR_LO_tiled_addr_31_0_mask 0xFFFFFFFF +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_TILED_ADDR_LO_tiled_addr_31_0_shift 0 +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_TILED_ADDR_LO_TILED_ADDR_31_0(x) (((x) & SDMA_PKT_COPY_TILED_SUBWIN_BC_TILED_ADDR_LO_tiled_addr_31_0_mask) << SDMA_PKT_COPY_TILED_SUBWIN_BC_TILED_ADDR_LO_tiled_addr_31_0_shift) + +/*define for TILED_ADDR_HI word*/ +/*define for tiled_addr_63_32 field*/ +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_TILED_ADDR_HI_tiled_addr_63_32_offset 2 +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_TILED_ADDR_HI_tiled_addr_63_32_mask 0xFFFFFFFF +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_TILED_ADDR_HI_tiled_addr_63_32_shift 0 +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_TILED_ADDR_HI_TILED_ADDR_63_32(x) (((x) & SDMA_PKT_COPY_TILED_SUBWIN_BC_TILED_ADDR_HI_tiled_addr_63_32_mask) << SDMA_PKT_COPY_TILED_SUBWIN_BC_TILED_ADDR_HI_tiled_addr_63_32_shift) + +/*define for DW_3 word*/ +/*define for tiled_x field*/ +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_3_tiled_x_offset 3 +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_3_tiled_x_mask 0x00003FFF +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_3_tiled_x_shift 0 +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_3_TILED_X(x) (((x) & SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_3_tiled_x_mask) << SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_3_tiled_x_shift) + +/*define for tiled_y field*/ +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_3_tiled_y_offset 3 +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_3_tiled_y_mask 0x00003FFF +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_3_tiled_y_shift 16 +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_3_TILED_Y(x) (((x) & SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_3_tiled_y_mask) << SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_3_tiled_y_shift) + +/*define for DW_4 word*/ +/*define for tiled_z field*/ +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_4_tiled_z_offset 4 +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_4_tiled_z_mask 0x000007FF +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_4_tiled_z_shift 0 +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_4_TILED_Z(x) (((x) & SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_4_tiled_z_mask) << SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_4_tiled_z_shift) + +/*define for width field*/ +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_4_width_offset 4 +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_4_width_mask 0x00003FFF +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_4_width_shift 16 +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_4_WIDTH(x) (((x) & SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_4_width_mask) << SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_4_width_shift) + +/*define for DW_5 word*/ +/*define for height field*/ +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_5_height_offset 5 +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_5_height_mask 0x00003FFF +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_5_height_shift 0 +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_5_HEIGHT(x) (((x) & SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_5_height_mask) << SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_5_height_shift) + +/*define for depth field*/ +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_5_depth_offset 5 +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_5_depth_mask 0x000007FF +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_5_depth_shift 16 +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_5_DEPTH(x) (((x) & SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_5_depth_mask) << SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_5_depth_shift) + +/*define for DW_6 word*/ +/*define for element_size field*/ +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_6_element_size_offset 6 +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_6_element_size_mask 0x00000007 +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_6_element_size_shift 0 +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_6_ELEMENT_SIZE(x) (((x) & SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_6_element_size_mask) << SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_6_element_size_shift) + +/*define for array_mode field*/ +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_6_array_mode_offset 6 +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_6_array_mode_mask 0x0000000F +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_6_array_mode_shift 3 +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_6_ARRAY_MODE(x) (((x) & SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_6_array_mode_mask) << SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_6_array_mode_shift) + +/*define for mit_mode field*/ +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_6_mit_mode_offset 6 +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_6_mit_mode_mask 0x00000007 +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_6_mit_mode_shift 8 +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_6_MIT_MODE(x) (((x) & SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_6_mit_mode_mask) << SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_6_mit_mode_shift) + +/*define for tilesplit_size field*/ +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_6_tilesplit_size_offset 6 +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_6_tilesplit_size_mask 0x00000007 +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_6_tilesplit_size_shift 11 +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_6_TILESPLIT_SIZE(x) (((x) & SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_6_tilesplit_size_mask) << SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_6_tilesplit_size_shift) + +/*define for bank_w field*/ +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_6_bank_w_offset 6 +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_6_bank_w_mask 0x00000003 +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_6_bank_w_shift 15 +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_6_BANK_W(x) (((x) & SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_6_bank_w_mask) << SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_6_bank_w_shift) + +/*define for bank_h field*/ +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_6_bank_h_offset 6 +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_6_bank_h_mask 0x00000003 +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_6_bank_h_shift 18 +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_6_BANK_H(x) (((x) & SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_6_bank_h_mask) << SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_6_bank_h_shift) + +/*define for num_bank field*/ +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_6_num_bank_offset 6 +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_6_num_bank_mask 0x00000003 +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_6_num_bank_shift 21 +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_6_NUM_BANK(x) (((x) & SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_6_num_bank_mask) << SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_6_num_bank_shift) + +/*define for mat_aspt field*/ +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_6_mat_aspt_offset 6 +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_6_mat_aspt_mask 0x00000003 +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_6_mat_aspt_shift 24 +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_6_MAT_ASPT(x) (((x) & SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_6_mat_aspt_mask) << SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_6_mat_aspt_shift) + +/*define for pipe_config field*/ +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_6_pipe_config_offset 6 +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_6_pipe_config_mask 0x0000001F +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_6_pipe_config_shift 26 +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_6_PIPE_CONFIG(x) (((x) & SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_6_pipe_config_mask) << SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_6_pipe_config_shift) + +/*define for LINEAR_ADDR_LO word*/ +/*define for linear_addr_31_0 field*/ +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_LINEAR_ADDR_LO_linear_addr_31_0_offset 7 +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_LINEAR_ADDR_LO_linear_addr_31_0_mask 0xFFFFFFFF +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_LINEAR_ADDR_LO_linear_addr_31_0_shift 0 +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_LINEAR_ADDR_LO_LINEAR_ADDR_31_0(x) (((x) & SDMA_PKT_COPY_TILED_SUBWIN_BC_LINEAR_ADDR_LO_linear_addr_31_0_mask) << SDMA_PKT_COPY_TILED_SUBWIN_BC_LINEAR_ADDR_LO_linear_addr_31_0_shift) + +/*define for LINEAR_ADDR_HI word*/ +/*define for linear_addr_63_32 field*/ +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_LINEAR_ADDR_HI_linear_addr_63_32_offset 8 +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_LINEAR_ADDR_HI_linear_addr_63_32_mask 0xFFFFFFFF +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_LINEAR_ADDR_HI_linear_addr_63_32_shift 0 +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_LINEAR_ADDR_HI_LINEAR_ADDR_63_32(x) (((x) & SDMA_PKT_COPY_TILED_SUBWIN_BC_LINEAR_ADDR_HI_linear_addr_63_32_mask) << SDMA_PKT_COPY_TILED_SUBWIN_BC_LINEAR_ADDR_HI_linear_addr_63_32_shift) + +/*define for DW_9 word*/ +/*define for linear_x field*/ +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_9_linear_x_offset 9 +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_9_linear_x_mask 0x00003FFF +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_9_linear_x_shift 0 +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_9_LINEAR_X(x) (((x) & SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_9_linear_x_mask) << SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_9_linear_x_shift) + +/*define for linear_y field*/ +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_9_linear_y_offset 9 +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_9_linear_y_mask 0x00003FFF +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_9_linear_y_shift 16 +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_9_LINEAR_Y(x) (((x) & SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_9_linear_y_mask) << SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_9_linear_y_shift) + +/*define for DW_10 word*/ +/*define for linear_z field*/ +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_10_linear_z_offset 10 +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_10_linear_z_mask 0x000007FF +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_10_linear_z_shift 0 +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_10_LINEAR_Z(x) (((x) & SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_10_linear_z_mask) << SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_10_linear_z_shift) + +/*define for linear_pitch field*/ +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_10_linear_pitch_offset 10 +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_10_linear_pitch_mask 0x00003FFF +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_10_linear_pitch_shift 16 +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_10_LINEAR_PITCH(x) (((x) & SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_10_linear_pitch_mask) << SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_10_linear_pitch_shift) + +/*define for DW_11 word*/ +/*define for linear_slice_pitch field*/ +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_11_linear_slice_pitch_offset 11 +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_11_linear_slice_pitch_mask 0x0FFFFFFF +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_11_linear_slice_pitch_shift 0 +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_11_LINEAR_SLICE_PITCH(x) (((x) & SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_11_linear_slice_pitch_mask) << SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_11_linear_slice_pitch_shift) + +/*define for DW_12 word*/ +/*define for rect_x field*/ +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_12_rect_x_offset 12 +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_12_rect_x_mask 0x00003FFF +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_12_rect_x_shift 0 +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_12_RECT_X(x) (((x) & SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_12_rect_x_mask) << SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_12_rect_x_shift) + +/*define for rect_y field*/ +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_12_rect_y_offset 12 +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_12_rect_y_mask 0x00003FFF +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_12_rect_y_shift 16 +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_12_RECT_Y(x) (((x) & SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_12_rect_y_mask) << SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_12_rect_y_shift) + +/*define for DW_13 word*/ +/*define for rect_z field*/ +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_13_rect_z_offset 13 +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_13_rect_z_mask 0x000007FF +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_13_rect_z_shift 0 +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_13_RECT_Z(x) (((x) & SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_13_rect_z_mask) << SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_13_rect_z_shift) + +/*define for linear_sw field*/ +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_13_linear_sw_offset 13 +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_13_linear_sw_mask 0x00000003 +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_13_linear_sw_shift 16 +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_13_LINEAR_SW(x) (((x) & SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_13_linear_sw_mask) << SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_13_linear_sw_shift) + +/*define for tile_sw field*/ +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_13_tile_sw_offset 13 +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_13_tile_sw_mask 0x00000003 +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_13_tile_sw_shift 24 +#define SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_13_TILE_SW(x) (((x) & SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_13_tile_sw_mask) << SDMA_PKT_COPY_TILED_SUBWIN_BC_DW_13_tile_sw_shift) + + +/* +** Definitions for SDMA_PKT_COPY_STRUCT packet +*/ + +/*define for HEADER word*/ +/*define for op field*/ +#define SDMA_PKT_COPY_STRUCT_HEADER_op_offset 0 +#define SDMA_PKT_COPY_STRUCT_HEADER_op_mask 0x000000FF +#define SDMA_PKT_COPY_STRUCT_HEADER_op_shift 0 +#define SDMA_PKT_COPY_STRUCT_HEADER_OP(x) (((x) & SDMA_PKT_COPY_STRUCT_HEADER_op_mask) << SDMA_PKT_COPY_STRUCT_HEADER_op_shift) + +/*define for sub_op field*/ +#define SDMA_PKT_COPY_STRUCT_HEADER_sub_op_offset 0 +#define SDMA_PKT_COPY_STRUCT_HEADER_sub_op_mask 0x000000FF +#define SDMA_PKT_COPY_STRUCT_HEADER_sub_op_shift 8 +#define SDMA_PKT_COPY_STRUCT_HEADER_SUB_OP(x) (((x) & SDMA_PKT_COPY_STRUCT_HEADER_sub_op_mask) << SDMA_PKT_COPY_STRUCT_HEADER_sub_op_shift) + +/*define for tmz field*/ +#define SDMA_PKT_COPY_STRUCT_HEADER_tmz_offset 0 +#define SDMA_PKT_COPY_STRUCT_HEADER_tmz_mask 0x00000001 +#define SDMA_PKT_COPY_STRUCT_HEADER_tmz_shift 18 +#define SDMA_PKT_COPY_STRUCT_HEADER_TMZ(x) (((x) & SDMA_PKT_COPY_STRUCT_HEADER_tmz_mask) << SDMA_PKT_COPY_STRUCT_HEADER_tmz_shift) + +/*define for cpv field*/ +#define SDMA_PKT_COPY_STRUCT_HEADER_cpv_offset 0 +#define SDMA_PKT_COPY_STRUCT_HEADER_cpv_mask 0x00000001 +#define SDMA_PKT_COPY_STRUCT_HEADER_cpv_shift 28 +#define SDMA_PKT_COPY_STRUCT_HEADER_CPV(x) (((x) & SDMA_PKT_COPY_STRUCT_HEADER_cpv_mask) << SDMA_PKT_COPY_STRUCT_HEADER_cpv_shift) + +/*define for detile field*/ +#define SDMA_PKT_COPY_STRUCT_HEADER_detile_offset 0 +#define SDMA_PKT_COPY_STRUCT_HEADER_detile_mask 0x00000001 +#define SDMA_PKT_COPY_STRUCT_HEADER_detile_shift 31 +#define SDMA_PKT_COPY_STRUCT_HEADER_DETILE(x) (((x) & SDMA_PKT_COPY_STRUCT_HEADER_detile_mask) << SDMA_PKT_COPY_STRUCT_HEADER_detile_shift) + +/*define for SB_ADDR_LO word*/ +/*define for sb_addr_31_0 field*/ +#define SDMA_PKT_COPY_STRUCT_SB_ADDR_LO_sb_addr_31_0_offset 1 +#define SDMA_PKT_COPY_STRUCT_SB_ADDR_LO_sb_addr_31_0_mask 0xFFFFFFFF +#define SDMA_PKT_COPY_STRUCT_SB_ADDR_LO_sb_addr_31_0_shift 0 +#define SDMA_PKT_COPY_STRUCT_SB_ADDR_LO_SB_ADDR_31_0(x) (((x) & SDMA_PKT_COPY_STRUCT_SB_ADDR_LO_sb_addr_31_0_mask) << SDMA_PKT_COPY_STRUCT_SB_ADDR_LO_sb_addr_31_0_shift) + +/*define for SB_ADDR_HI word*/ +/*define for sb_addr_63_32 field*/ +#define SDMA_PKT_COPY_STRUCT_SB_ADDR_HI_sb_addr_63_32_offset 2 +#define SDMA_PKT_COPY_STRUCT_SB_ADDR_HI_sb_addr_63_32_mask 0xFFFFFFFF +#define SDMA_PKT_COPY_STRUCT_SB_ADDR_HI_sb_addr_63_32_shift 0 +#define SDMA_PKT_COPY_STRUCT_SB_ADDR_HI_SB_ADDR_63_32(x) (((x) & SDMA_PKT_COPY_STRUCT_SB_ADDR_HI_sb_addr_63_32_mask) << SDMA_PKT_COPY_STRUCT_SB_ADDR_HI_sb_addr_63_32_shift) + +/*define for START_INDEX word*/ +/*define for start_index field*/ +#define SDMA_PKT_COPY_STRUCT_START_INDEX_start_index_offset 3 +#define SDMA_PKT_COPY_STRUCT_START_INDEX_start_index_mask 0xFFFFFFFF +#define SDMA_PKT_COPY_STRUCT_START_INDEX_start_index_shift 0 +#define SDMA_PKT_COPY_STRUCT_START_INDEX_START_INDEX(x) (((x) & SDMA_PKT_COPY_STRUCT_START_INDEX_start_index_mask) << SDMA_PKT_COPY_STRUCT_START_INDEX_start_index_shift) + +/*define for COUNT word*/ +/*define for count field*/ +#define SDMA_PKT_COPY_STRUCT_COUNT_count_offset 4 +#define SDMA_PKT_COPY_STRUCT_COUNT_count_mask 0xFFFFFFFF +#define SDMA_PKT_COPY_STRUCT_COUNT_count_shift 0 +#define SDMA_PKT_COPY_STRUCT_COUNT_COUNT(x) (((x) & SDMA_PKT_COPY_STRUCT_COUNT_count_mask) << SDMA_PKT_COPY_STRUCT_COUNT_count_shift) + +/*define for DW_5 word*/ +/*define for stride field*/ +#define SDMA_PKT_COPY_STRUCT_DW_5_stride_offset 5 +#define SDMA_PKT_COPY_STRUCT_DW_5_stride_mask 0x000007FF +#define SDMA_PKT_COPY_STRUCT_DW_5_stride_shift 0 +#define SDMA_PKT_COPY_STRUCT_DW_5_STRIDE(x) (((x) & SDMA_PKT_COPY_STRUCT_DW_5_stride_mask) << SDMA_PKT_COPY_STRUCT_DW_5_stride_shift) + +/*define for linear_sw field*/ +#define SDMA_PKT_COPY_STRUCT_DW_5_linear_sw_offset 5 +#define SDMA_PKT_COPY_STRUCT_DW_5_linear_sw_mask 0x00000003 +#define SDMA_PKT_COPY_STRUCT_DW_5_linear_sw_shift 16 +#define SDMA_PKT_COPY_STRUCT_DW_5_LINEAR_SW(x) (((x) & SDMA_PKT_COPY_STRUCT_DW_5_linear_sw_mask) << SDMA_PKT_COPY_STRUCT_DW_5_linear_sw_shift) + +/*define for linear_cache_policy field*/ +#define SDMA_PKT_COPY_STRUCT_DW_5_linear_cache_policy_offset 5 +#define SDMA_PKT_COPY_STRUCT_DW_5_linear_cache_policy_mask 0x00000007 +#define SDMA_PKT_COPY_STRUCT_DW_5_linear_cache_policy_shift 18 +#define SDMA_PKT_COPY_STRUCT_DW_5_LINEAR_CACHE_POLICY(x) (((x) & SDMA_PKT_COPY_STRUCT_DW_5_linear_cache_policy_mask) << SDMA_PKT_COPY_STRUCT_DW_5_linear_cache_policy_shift) + +/*define for struct_sw field*/ +#define SDMA_PKT_COPY_STRUCT_DW_5_struct_sw_offset 5 +#define SDMA_PKT_COPY_STRUCT_DW_5_struct_sw_mask 0x00000003 +#define SDMA_PKT_COPY_STRUCT_DW_5_struct_sw_shift 24 +#define SDMA_PKT_COPY_STRUCT_DW_5_STRUCT_SW(x) (((x) & SDMA_PKT_COPY_STRUCT_DW_5_struct_sw_mask) << SDMA_PKT_COPY_STRUCT_DW_5_struct_sw_shift) + +/*define for struct_cache_policy field*/ +#define SDMA_PKT_COPY_STRUCT_DW_5_struct_cache_policy_offset 5 +#define SDMA_PKT_COPY_STRUCT_DW_5_struct_cache_policy_mask 0x00000007 +#define SDMA_PKT_COPY_STRUCT_DW_5_struct_cache_policy_shift 26 +#define SDMA_PKT_COPY_STRUCT_DW_5_STRUCT_CACHE_POLICY(x) (((x) & SDMA_PKT_COPY_STRUCT_DW_5_struct_cache_policy_mask) << SDMA_PKT_COPY_STRUCT_DW_5_struct_cache_policy_shift) + +/*define for LINEAR_ADDR_LO word*/ +/*define for linear_addr_31_0 field*/ +#define SDMA_PKT_COPY_STRUCT_LINEAR_ADDR_LO_linear_addr_31_0_offset 6 +#define SDMA_PKT_COPY_STRUCT_LINEAR_ADDR_LO_linear_addr_31_0_mask 0xFFFFFFFF +#define SDMA_PKT_COPY_STRUCT_LINEAR_ADDR_LO_linear_addr_31_0_shift 0 +#define SDMA_PKT_COPY_STRUCT_LINEAR_ADDR_LO_LINEAR_ADDR_31_0(x) (((x) & SDMA_PKT_COPY_STRUCT_LINEAR_ADDR_LO_linear_addr_31_0_mask) << SDMA_PKT_COPY_STRUCT_LINEAR_ADDR_LO_linear_addr_31_0_shift) + +/*define for LINEAR_ADDR_HI word*/ +/*define for linear_addr_63_32 field*/ +#define SDMA_PKT_COPY_STRUCT_LINEAR_ADDR_HI_linear_addr_63_32_offset 7 +#define SDMA_PKT_COPY_STRUCT_LINEAR_ADDR_HI_linear_addr_63_32_mask 0xFFFFFFFF +#define SDMA_PKT_COPY_STRUCT_LINEAR_ADDR_HI_linear_addr_63_32_shift 0 +#define SDMA_PKT_COPY_STRUCT_LINEAR_ADDR_HI_LINEAR_ADDR_63_32(x) (((x) & SDMA_PKT_COPY_STRUCT_LINEAR_ADDR_HI_linear_addr_63_32_mask) << SDMA_PKT_COPY_STRUCT_LINEAR_ADDR_HI_linear_addr_63_32_shift) + + +/* +** Definitions for SDMA_PKT_WRITE_UNTILED packet +*/ + +/*define for HEADER word*/ +/*define for op field*/ +#define SDMA_PKT_WRITE_UNTILED_HEADER_op_offset 0 +#define SDMA_PKT_WRITE_UNTILED_HEADER_op_mask 0x000000FF +#define SDMA_PKT_WRITE_UNTILED_HEADER_op_shift 0 +#define SDMA_PKT_WRITE_UNTILED_HEADER_OP(x) (((x) & SDMA_PKT_WRITE_UNTILED_HEADER_op_mask) << SDMA_PKT_WRITE_UNTILED_HEADER_op_shift) + +/*define for sub_op field*/ +#define SDMA_PKT_WRITE_UNTILED_HEADER_sub_op_offset 0 +#define SDMA_PKT_WRITE_UNTILED_HEADER_sub_op_mask 0x000000FF +#define SDMA_PKT_WRITE_UNTILED_HEADER_sub_op_shift 8 +#define SDMA_PKT_WRITE_UNTILED_HEADER_SUB_OP(x) (((x) & SDMA_PKT_WRITE_UNTILED_HEADER_sub_op_mask) << SDMA_PKT_WRITE_UNTILED_HEADER_sub_op_shift) + +/*define for encrypt field*/ +#define SDMA_PKT_WRITE_UNTILED_HEADER_encrypt_offset 0 +#define SDMA_PKT_WRITE_UNTILED_HEADER_encrypt_mask 0x00000001 +#define SDMA_PKT_WRITE_UNTILED_HEADER_encrypt_shift 16 +#define SDMA_PKT_WRITE_UNTILED_HEADER_ENCRYPT(x) (((x) & SDMA_PKT_WRITE_UNTILED_HEADER_encrypt_mask) << SDMA_PKT_WRITE_UNTILED_HEADER_encrypt_shift) + +/*define for tmz field*/ +#define SDMA_PKT_WRITE_UNTILED_HEADER_tmz_offset 0 +#define SDMA_PKT_WRITE_UNTILED_HEADER_tmz_mask 0x00000001 +#define SDMA_PKT_WRITE_UNTILED_HEADER_tmz_shift 18 +#define SDMA_PKT_WRITE_UNTILED_HEADER_TMZ(x) (((x) & SDMA_PKT_WRITE_UNTILED_HEADER_tmz_mask) << SDMA_PKT_WRITE_UNTILED_HEADER_tmz_shift) + +/*define for cpv field*/ +#define SDMA_PKT_WRITE_UNTILED_HEADER_cpv_offset 0 +#define SDMA_PKT_WRITE_UNTILED_HEADER_cpv_mask 0x00000001 +#define SDMA_PKT_WRITE_UNTILED_HEADER_cpv_shift 28 +#define SDMA_PKT_WRITE_UNTILED_HEADER_CPV(x) (((x) & SDMA_PKT_WRITE_UNTILED_HEADER_cpv_mask) << SDMA_PKT_WRITE_UNTILED_HEADER_cpv_shift) + +/*define for DST_ADDR_LO word*/ +/*define for dst_addr_31_0 field*/ +#define SDMA_PKT_WRITE_UNTILED_DST_ADDR_LO_dst_addr_31_0_offset 1 +#define SDMA_PKT_WRITE_UNTILED_DST_ADDR_LO_dst_addr_31_0_mask 0xFFFFFFFF +#define SDMA_PKT_WRITE_UNTILED_DST_ADDR_LO_dst_addr_31_0_shift 0 +#define SDMA_PKT_WRITE_UNTILED_DST_ADDR_LO_DST_ADDR_31_0(x) (((x) & SDMA_PKT_WRITE_UNTILED_DST_ADDR_LO_dst_addr_31_0_mask) << SDMA_PKT_WRITE_UNTILED_DST_ADDR_LO_dst_addr_31_0_shift) + +/*define for DST_ADDR_HI word*/ +/*define for dst_addr_63_32 field*/ +#define SDMA_PKT_WRITE_UNTILED_DST_ADDR_HI_dst_addr_63_32_offset 2 +#define SDMA_PKT_WRITE_UNTILED_DST_ADDR_HI_dst_addr_63_32_mask 0xFFFFFFFF +#define SDMA_PKT_WRITE_UNTILED_DST_ADDR_HI_dst_addr_63_32_shift 0 +#define SDMA_PKT_WRITE_UNTILED_DST_ADDR_HI_DST_ADDR_63_32(x) (((x) & SDMA_PKT_WRITE_UNTILED_DST_ADDR_HI_dst_addr_63_32_mask) << SDMA_PKT_WRITE_UNTILED_DST_ADDR_HI_dst_addr_63_32_shift) + +/*define for DW_3 word*/ +/*define for count field*/ +#define SDMA_PKT_WRITE_UNTILED_DW_3_count_offset 3 +#define SDMA_PKT_WRITE_UNTILED_DW_3_count_mask 0x000FFFFF +#define SDMA_PKT_WRITE_UNTILED_DW_3_count_shift 0 +#define SDMA_PKT_WRITE_UNTILED_DW_3_COUNT(x) (((x) & SDMA_PKT_WRITE_UNTILED_DW_3_count_mask) << SDMA_PKT_WRITE_UNTILED_DW_3_count_shift) + +/*define for sw field*/ +#define SDMA_PKT_WRITE_UNTILED_DW_3_sw_offset 3 +#define SDMA_PKT_WRITE_UNTILED_DW_3_sw_mask 0x00000003 +#define SDMA_PKT_WRITE_UNTILED_DW_3_sw_shift 24 +#define SDMA_PKT_WRITE_UNTILED_DW_3_SW(x) (((x) & SDMA_PKT_WRITE_UNTILED_DW_3_sw_mask) << SDMA_PKT_WRITE_UNTILED_DW_3_sw_shift) + +/*define for cache_policy field*/ +#define SDMA_PKT_WRITE_UNTILED_DW_3_cache_policy_offset 3 +#define SDMA_PKT_WRITE_UNTILED_DW_3_cache_policy_mask 0x00000007 +#define SDMA_PKT_WRITE_UNTILED_DW_3_cache_policy_shift 26 +#define SDMA_PKT_WRITE_UNTILED_DW_3_CACHE_POLICY(x) (((x) & SDMA_PKT_WRITE_UNTILED_DW_3_cache_policy_mask) << SDMA_PKT_WRITE_UNTILED_DW_3_cache_policy_shift) + +/*define for DATA0 word*/ +/*define for data0 field*/ +#define SDMA_PKT_WRITE_UNTILED_DATA0_data0_offset 4 +#define SDMA_PKT_WRITE_UNTILED_DATA0_data0_mask 0xFFFFFFFF +#define SDMA_PKT_WRITE_UNTILED_DATA0_data0_shift 0 +#define SDMA_PKT_WRITE_UNTILED_DATA0_DATA0(x) (((x) & SDMA_PKT_WRITE_UNTILED_DATA0_data0_mask) << SDMA_PKT_WRITE_UNTILED_DATA0_data0_shift) + + +/* +** Definitions for SDMA_PKT_WRITE_TILED packet +*/ + +/*define for HEADER word*/ +/*define for op field*/ +#define SDMA_PKT_WRITE_TILED_HEADER_op_offset 0 +#define SDMA_PKT_WRITE_TILED_HEADER_op_mask 0x000000FF +#define SDMA_PKT_WRITE_TILED_HEADER_op_shift 0 +#define SDMA_PKT_WRITE_TILED_HEADER_OP(x) (((x) & SDMA_PKT_WRITE_TILED_HEADER_op_mask) << SDMA_PKT_WRITE_TILED_HEADER_op_shift) + +/*define for sub_op field*/ +#define SDMA_PKT_WRITE_TILED_HEADER_sub_op_offset 0 +#define SDMA_PKT_WRITE_TILED_HEADER_sub_op_mask 0x000000FF +#define SDMA_PKT_WRITE_TILED_HEADER_sub_op_shift 8 +#define SDMA_PKT_WRITE_TILED_HEADER_SUB_OP(x) (((x) & SDMA_PKT_WRITE_TILED_HEADER_sub_op_mask) << SDMA_PKT_WRITE_TILED_HEADER_sub_op_shift) + +/*define for encrypt field*/ +#define SDMA_PKT_WRITE_TILED_HEADER_encrypt_offset 0 +#define SDMA_PKT_WRITE_TILED_HEADER_encrypt_mask 0x00000001 +#define SDMA_PKT_WRITE_TILED_HEADER_encrypt_shift 16 +#define SDMA_PKT_WRITE_TILED_HEADER_ENCRYPT(x) (((x) & SDMA_PKT_WRITE_TILED_HEADER_encrypt_mask) << SDMA_PKT_WRITE_TILED_HEADER_encrypt_shift) + +/*define for tmz field*/ +#define SDMA_PKT_WRITE_TILED_HEADER_tmz_offset 0 +#define SDMA_PKT_WRITE_TILED_HEADER_tmz_mask 0x00000001 +#define SDMA_PKT_WRITE_TILED_HEADER_tmz_shift 18 +#define SDMA_PKT_WRITE_TILED_HEADER_TMZ(x) (((x) & SDMA_PKT_WRITE_TILED_HEADER_tmz_mask) << SDMA_PKT_WRITE_TILED_HEADER_tmz_shift) + +/*define for cpv field*/ +#define SDMA_PKT_WRITE_TILED_HEADER_cpv_offset 0 +#define SDMA_PKT_WRITE_TILED_HEADER_cpv_mask 0x00000001 +#define SDMA_PKT_WRITE_TILED_HEADER_cpv_shift 28 +#define SDMA_PKT_WRITE_TILED_HEADER_CPV(x) (((x) & SDMA_PKT_WRITE_TILED_HEADER_cpv_mask) << SDMA_PKT_WRITE_TILED_HEADER_cpv_shift) + +/*define for DST_ADDR_LO word*/ +/*define for dst_addr_31_0 field*/ +#define SDMA_PKT_WRITE_TILED_DST_ADDR_LO_dst_addr_31_0_offset 1 +#define SDMA_PKT_WRITE_TILED_DST_ADDR_LO_dst_addr_31_0_mask 0xFFFFFFFF +#define SDMA_PKT_WRITE_TILED_DST_ADDR_LO_dst_addr_31_0_shift 0 +#define SDMA_PKT_WRITE_TILED_DST_ADDR_LO_DST_ADDR_31_0(x) (((x) & SDMA_PKT_WRITE_TILED_DST_ADDR_LO_dst_addr_31_0_mask) << SDMA_PKT_WRITE_TILED_DST_ADDR_LO_dst_addr_31_0_shift) + +/*define for DST_ADDR_HI word*/ +/*define for dst_addr_63_32 field*/ +#define SDMA_PKT_WRITE_TILED_DST_ADDR_HI_dst_addr_63_32_offset 2 +#define SDMA_PKT_WRITE_TILED_DST_ADDR_HI_dst_addr_63_32_mask 0xFFFFFFFF +#define SDMA_PKT_WRITE_TILED_DST_ADDR_HI_dst_addr_63_32_shift 0 +#define SDMA_PKT_WRITE_TILED_DST_ADDR_HI_DST_ADDR_63_32(x) (((x) & SDMA_PKT_WRITE_TILED_DST_ADDR_HI_dst_addr_63_32_mask) << SDMA_PKT_WRITE_TILED_DST_ADDR_HI_dst_addr_63_32_shift) + +/*define for DW_3 word*/ +/*define for width field*/ +#define SDMA_PKT_WRITE_TILED_DW_3_width_offset 3 +#define SDMA_PKT_WRITE_TILED_DW_3_width_mask 0x00003FFF +#define SDMA_PKT_WRITE_TILED_DW_3_width_shift 0 +#define SDMA_PKT_WRITE_TILED_DW_3_WIDTH(x) (((x) & SDMA_PKT_WRITE_TILED_DW_3_width_mask) << SDMA_PKT_WRITE_TILED_DW_3_width_shift) + +/*define for DW_4 word*/ +/*define for height field*/ +#define SDMA_PKT_WRITE_TILED_DW_4_height_offset 4 +#define SDMA_PKT_WRITE_TILED_DW_4_height_mask 0x00003FFF +#define SDMA_PKT_WRITE_TILED_DW_4_height_shift 0 +#define SDMA_PKT_WRITE_TILED_DW_4_HEIGHT(x) (((x) & SDMA_PKT_WRITE_TILED_DW_4_height_mask) << SDMA_PKT_WRITE_TILED_DW_4_height_shift) + +/*define for depth field*/ +#define SDMA_PKT_WRITE_TILED_DW_4_depth_offset 4 +#define SDMA_PKT_WRITE_TILED_DW_4_depth_mask 0x00001FFF +#define SDMA_PKT_WRITE_TILED_DW_4_depth_shift 16 +#define SDMA_PKT_WRITE_TILED_DW_4_DEPTH(x) (((x) & SDMA_PKT_WRITE_TILED_DW_4_depth_mask) << SDMA_PKT_WRITE_TILED_DW_4_depth_shift) + +/*define for DW_5 word*/ +/*define for element_size field*/ +#define SDMA_PKT_WRITE_TILED_DW_5_element_size_offset 5 +#define SDMA_PKT_WRITE_TILED_DW_5_element_size_mask 0x00000007 +#define SDMA_PKT_WRITE_TILED_DW_5_element_size_shift 0 +#define SDMA_PKT_WRITE_TILED_DW_5_ELEMENT_SIZE(x) (((x) & SDMA_PKT_WRITE_TILED_DW_5_element_size_mask) << SDMA_PKT_WRITE_TILED_DW_5_element_size_shift) + +/*define for swizzle_mode field*/ +#define SDMA_PKT_WRITE_TILED_DW_5_swizzle_mode_offset 5 +#define SDMA_PKT_WRITE_TILED_DW_5_swizzle_mode_mask 0x0000001F +#define SDMA_PKT_WRITE_TILED_DW_5_swizzle_mode_shift 3 +#define SDMA_PKT_WRITE_TILED_DW_5_SWIZZLE_MODE(x) (((x) & SDMA_PKT_WRITE_TILED_DW_5_swizzle_mode_mask) << SDMA_PKT_WRITE_TILED_DW_5_swizzle_mode_shift) + +/*define for dimension field*/ +#define SDMA_PKT_WRITE_TILED_DW_5_dimension_offset 5 +#define SDMA_PKT_WRITE_TILED_DW_5_dimension_mask 0x00000003 +#define SDMA_PKT_WRITE_TILED_DW_5_dimension_shift 9 +#define SDMA_PKT_WRITE_TILED_DW_5_DIMENSION(x) (((x) & SDMA_PKT_WRITE_TILED_DW_5_dimension_mask) << SDMA_PKT_WRITE_TILED_DW_5_dimension_shift) + +/*define for mip_max field*/ +#define SDMA_PKT_WRITE_TILED_DW_5_mip_max_offset 5 +#define SDMA_PKT_WRITE_TILED_DW_5_mip_max_mask 0x0000000F +#define SDMA_PKT_WRITE_TILED_DW_5_mip_max_shift 16 +#define SDMA_PKT_WRITE_TILED_DW_5_MIP_MAX(x) (((x) & SDMA_PKT_WRITE_TILED_DW_5_mip_max_mask) << SDMA_PKT_WRITE_TILED_DW_5_mip_max_shift) + +/*define for DW_6 word*/ +/*define for x field*/ +#define SDMA_PKT_WRITE_TILED_DW_6_x_offset 6 +#define SDMA_PKT_WRITE_TILED_DW_6_x_mask 0x00003FFF +#define SDMA_PKT_WRITE_TILED_DW_6_x_shift 0 +#define SDMA_PKT_WRITE_TILED_DW_6_X(x) (((x) & SDMA_PKT_WRITE_TILED_DW_6_x_mask) << SDMA_PKT_WRITE_TILED_DW_6_x_shift) + +/*define for y field*/ +#define SDMA_PKT_WRITE_TILED_DW_6_y_offset 6 +#define SDMA_PKT_WRITE_TILED_DW_6_y_mask 0x00003FFF +#define SDMA_PKT_WRITE_TILED_DW_6_y_shift 16 +#define SDMA_PKT_WRITE_TILED_DW_6_Y(x) (((x) & SDMA_PKT_WRITE_TILED_DW_6_y_mask) << SDMA_PKT_WRITE_TILED_DW_6_y_shift) + +/*define for DW_7 word*/ +/*define for z field*/ +#define SDMA_PKT_WRITE_TILED_DW_7_z_offset 7 +#define SDMA_PKT_WRITE_TILED_DW_7_z_mask 0x00001FFF +#define SDMA_PKT_WRITE_TILED_DW_7_z_shift 0 +#define SDMA_PKT_WRITE_TILED_DW_7_Z(x) (((x) & SDMA_PKT_WRITE_TILED_DW_7_z_mask) << SDMA_PKT_WRITE_TILED_DW_7_z_shift) + +/*define for sw field*/ +#define SDMA_PKT_WRITE_TILED_DW_7_sw_offset 7 +#define SDMA_PKT_WRITE_TILED_DW_7_sw_mask 0x00000003 +#define SDMA_PKT_WRITE_TILED_DW_7_sw_shift 24 +#define SDMA_PKT_WRITE_TILED_DW_7_SW(x) (((x) & SDMA_PKT_WRITE_TILED_DW_7_sw_mask) << SDMA_PKT_WRITE_TILED_DW_7_sw_shift) + +/*define for cache_policy field*/ +#define SDMA_PKT_WRITE_TILED_DW_7_cache_policy_offset 7 +#define SDMA_PKT_WRITE_TILED_DW_7_cache_policy_mask 0x00000007 +#define SDMA_PKT_WRITE_TILED_DW_7_cache_policy_shift 26 +#define SDMA_PKT_WRITE_TILED_DW_7_CACHE_POLICY(x) (((x) & SDMA_PKT_WRITE_TILED_DW_7_cache_policy_mask) << SDMA_PKT_WRITE_TILED_DW_7_cache_policy_shift) + +/*define for COUNT word*/ +/*define for count field*/ +#define SDMA_PKT_WRITE_TILED_COUNT_count_offset 8 +#define SDMA_PKT_WRITE_TILED_COUNT_count_mask 0x000FFFFF +#define SDMA_PKT_WRITE_TILED_COUNT_count_shift 0 +#define SDMA_PKT_WRITE_TILED_COUNT_COUNT(x) (((x) & SDMA_PKT_WRITE_TILED_COUNT_count_mask) << SDMA_PKT_WRITE_TILED_COUNT_count_shift) + +/*define for DATA0 word*/ +/*define for data0 field*/ +#define SDMA_PKT_WRITE_TILED_DATA0_data0_offset 9 +#define SDMA_PKT_WRITE_TILED_DATA0_data0_mask 0xFFFFFFFF +#define SDMA_PKT_WRITE_TILED_DATA0_data0_shift 0 +#define SDMA_PKT_WRITE_TILED_DATA0_DATA0(x) (((x) & SDMA_PKT_WRITE_TILED_DATA0_data0_mask) << SDMA_PKT_WRITE_TILED_DATA0_data0_shift) + + +/* +** Definitions for SDMA_PKT_WRITE_TILED_BC packet +*/ + +/*define for HEADER word*/ +/*define for op field*/ +#define SDMA_PKT_WRITE_TILED_BC_HEADER_op_offset 0 +#define SDMA_PKT_WRITE_TILED_BC_HEADER_op_mask 0x000000FF +#define SDMA_PKT_WRITE_TILED_BC_HEADER_op_shift 0 +#define SDMA_PKT_WRITE_TILED_BC_HEADER_OP(x) (((x) & SDMA_PKT_WRITE_TILED_BC_HEADER_op_mask) << SDMA_PKT_WRITE_TILED_BC_HEADER_op_shift) + +/*define for sub_op field*/ +#define SDMA_PKT_WRITE_TILED_BC_HEADER_sub_op_offset 0 +#define SDMA_PKT_WRITE_TILED_BC_HEADER_sub_op_mask 0x000000FF +#define SDMA_PKT_WRITE_TILED_BC_HEADER_sub_op_shift 8 +#define SDMA_PKT_WRITE_TILED_BC_HEADER_SUB_OP(x) (((x) & SDMA_PKT_WRITE_TILED_BC_HEADER_sub_op_mask) << SDMA_PKT_WRITE_TILED_BC_HEADER_sub_op_shift) + +/*define for DST_ADDR_LO word*/ +/*define for dst_addr_31_0 field*/ +#define SDMA_PKT_WRITE_TILED_BC_DST_ADDR_LO_dst_addr_31_0_offset 1 +#define SDMA_PKT_WRITE_TILED_BC_DST_ADDR_LO_dst_addr_31_0_mask 0xFFFFFFFF +#define SDMA_PKT_WRITE_TILED_BC_DST_ADDR_LO_dst_addr_31_0_shift 0 +#define SDMA_PKT_WRITE_TILED_BC_DST_ADDR_LO_DST_ADDR_31_0(x) (((x) & SDMA_PKT_WRITE_TILED_BC_DST_ADDR_LO_dst_addr_31_0_mask) << SDMA_PKT_WRITE_TILED_BC_DST_ADDR_LO_dst_addr_31_0_shift) + +/*define for DST_ADDR_HI word*/ +/*define for dst_addr_63_32 field*/ +#define SDMA_PKT_WRITE_TILED_BC_DST_ADDR_HI_dst_addr_63_32_offset 2 +#define SDMA_PKT_WRITE_TILED_BC_DST_ADDR_HI_dst_addr_63_32_mask 0xFFFFFFFF +#define SDMA_PKT_WRITE_TILED_BC_DST_ADDR_HI_dst_addr_63_32_shift 0 +#define SDMA_PKT_WRITE_TILED_BC_DST_ADDR_HI_DST_ADDR_63_32(x) (((x) & SDMA_PKT_WRITE_TILED_BC_DST_ADDR_HI_dst_addr_63_32_mask) << SDMA_PKT_WRITE_TILED_BC_DST_ADDR_HI_dst_addr_63_32_shift) + +/*define for DW_3 word*/ +/*define for width field*/ +#define SDMA_PKT_WRITE_TILED_BC_DW_3_width_offset 3 +#define SDMA_PKT_WRITE_TILED_BC_DW_3_width_mask 0x00003FFF +#define SDMA_PKT_WRITE_TILED_BC_DW_3_width_shift 0 +#define SDMA_PKT_WRITE_TILED_BC_DW_3_WIDTH(x) (((x) & SDMA_PKT_WRITE_TILED_BC_DW_3_width_mask) << SDMA_PKT_WRITE_TILED_BC_DW_3_width_shift) + +/*define for DW_4 word*/ +/*define for height field*/ +#define SDMA_PKT_WRITE_TILED_BC_DW_4_height_offset 4 +#define SDMA_PKT_WRITE_TILED_BC_DW_4_height_mask 0x00003FFF +#define SDMA_PKT_WRITE_TILED_BC_DW_4_height_shift 0 +#define SDMA_PKT_WRITE_TILED_BC_DW_4_HEIGHT(x) (((x) & SDMA_PKT_WRITE_TILED_BC_DW_4_height_mask) << SDMA_PKT_WRITE_TILED_BC_DW_4_height_shift) + +/*define for depth field*/ +#define SDMA_PKT_WRITE_TILED_BC_DW_4_depth_offset 4 +#define SDMA_PKT_WRITE_TILED_BC_DW_4_depth_mask 0x000007FF +#define SDMA_PKT_WRITE_TILED_BC_DW_4_depth_shift 16 +#define SDMA_PKT_WRITE_TILED_BC_DW_4_DEPTH(x) (((x) & SDMA_PKT_WRITE_TILED_BC_DW_4_depth_mask) << SDMA_PKT_WRITE_TILED_BC_DW_4_depth_shift) + +/*define for DW_5 word*/ +/*define for element_size field*/ +#define SDMA_PKT_WRITE_TILED_BC_DW_5_element_size_offset 5 +#define SDMA_PKT_WRITE_TILED_BC_DW_5_element_size_mask 0x00000007 +#define SDMA_PKT_WRITE_TILED_BC_DW_5_element_size_shift 0 +#define SDMA_PKT_WRITE_TILED_BC_DW_5_ELEMENT_SIZE(x) (((x) & SDMA_PKT_WRITE_TILED_BC_DW_5_element_size_mask) << SDMA_PKT_WRITE_TILED_BC_DW_5_element_size_shift) + +/*define for array_mode field*/ +#define SDMA_PKT_WRITE_TILED_BC_DW_5_array_mode_offset 5 +#define SDMA_PKT_WRITE_TILED_BC_DW_5_array_mode_mask 0x0000000F +#define SDMA_PKT_WRITE_TILED_BC_DW_5_array_mode_shift 3 +#define SDMA_PKT_WRITE_TILED_BC_DW_5_ARRAY_MODE(x) (((x) & SDMA_PKT_WRITE_TILED_BC_DW_5_array_mode_mask) << SDMA_PKT_WRITE_TILED_BC_DW_5_array_mode_shift) + +/*define for mit_mode field*/ +#define SDMA_PKT_WRITE_TILED_BC_DW_5_mit_mode_offset 5 +#define SDMA_PKT_WRITE_TILED_BC_DW_5_mit_mode_mask 0x00000007 +#define SDMA_PKT_WRITE_TILED_BC_DW_5_mit_mode_shift 8 +#define SDMA_PKT_WRITE_TILED_BC_DW_5_MIT_MODE(x) (((x) & SDMA_PKT_WRITE_TILED_BC_DW_5_mit_mode_mask) << SDMA_PKT_WRITE_TILED_BC_DW_5_mit_mode_shift) + +/*define for tilesplit_size field*/ +#define SDMA_PKT_WRITE_TILED_BC_DW_5_tilesplit_size_offset 5 +#define SDMA_PKT_WRITE_TILED_BC_DW_5_tilesplit_size_mask 0x00000007 +#define SDMA_PKT_WRITE_TILED_BC_DW_5_tilesplit_size_shift 11 +#define SDMA_PKT_WRITE_TILED_BC_DW_5_TILESPLIT_SIZE(x) (((x) & SDMA_PKT_WRITE_TILED_BC_DW_5_tilesplit_size_mask) << SDMA_PKT_WRITE_TILED_BC_DW_5_tilesplit_size_shift) + +/*define for bank_w field*/ +#define SDMA_PKT_WRITE_TILED_BC_DW_5_bank_w_offset 5 +#define SDMA_PKT_WRITE_TILED_BC_DW_5_bank_w_mask 0x00000003 +#define SDMA_PKT_WRITE_TILED_BC_DW_5_bank_w_shift 15 +#define SDMA_PKT_WRITE_TILED_BC_DW_5_BANK_W(x) (((x) & SDMA_PKT_WRITE_TILED_BC_DW_5_bank_w_mask) << SDMA_PKT_WRITE_TILED_BC_DW_5_bank_w_shift) + +/*define for bank_h field*/ +#define SDMA_PKT_WRITE_TILED_BC_DW_5_bank_h_offset 5 +#define SDMA_PKT_WRITE_TILED_BC_DW_5_bank_h_mask 0x00000003 +#define SDMA_PKT_WRITE_TILED_BC_DW_5_bank_h_shift 18 +#define SDMA_PKT_WRITE_TILED_BC_DW_5_BANK_H(x) (((x) & SDMA_PKT_WRITE_TILED_BC_DW_5_bank_h_mask) << SDMA_PKT_WRITE_TILED_BC_DW_5_bank_h_shift) + +/*define for num_bank field*/ +#define SDMA_PKT_WRITE_TILED_BC_DW_5_num_bank_offset 5 +#define SDMA_PKT_WRITE_TILED_BC_DW_5_num_bank_mask 0x00000003 +#define SDMA_PKT_WRITE_TILED_BC_DW_5_num_bank_shift 21 +#define SDMA_PKT_WRITE_TILED_BC_DW_5_NUM_BANK(x) (((x) & SDMA_PKT_WRITE_TILED_BC_DW_5_num_bank_mask) << SDMA_PKT_WRITE_TILED_BC_DW_5_num_bank_shift) + +/*define for mat_aspt field*/ +#define SDMA_PKT_WRITE_TILED_BC_DW_5_mat_aspt_offset 5 +#define SDMA_PKT_WRITE_TILED_BC_DW_5_mat_aspt_mask 0x00000003 +#define SDMA_PKT_WRITE_TILED_BC_DW_5_mat_aspt_shift 24 +#define SDMA_PKT_WRITE_TILED_BC_DW_5_MAT_ASPT(x) (((x) & SDMA_PKT_WRITE_TILED_BC_DW_5_mat_aspt_mask) << SDMA_PKT_WRITE_TILED_BC_DW_5_mat_aspt_shift) + +/*define for pipe_config field*/ +#define SDMA_PKT_WRITE_TILED_BC_DW_5_pipe_config_offset 5 +#define SDMA_PKT_WRITE_TILED_BC_DW_5_pipe_config_mask 0x0000001F +#define SDMA_PKT_WRITE_TILED_BC_DW_5_pipe_config_shift 26 +#define SDMA_PKT_WRITE_TILED_BC_DW_5_PIPE_CONFIG(x) (((x) & SDMA_PKT_WRITE_TILED_BC_DW_5_pipe_config_mask) << SDMA_PKT_WRITE_TILED_BC_DW_5_pipe_config_shift) + +/*define for DW_6 word*/ +/*define for x field*/ +#define SDMA_PKT_WRITE_TILED_BC_DW_6_x_offset 6 +#define SDMA_PKT_WRITE_TILED_BC_DW_6_x_mask 0x00003FFF +#define SDMA_PKT_WRITE_TILED_BC_DW_6_x_shift 0 +#define SDMA_PKT_WRITE_TILED_BC_DW_6_X(x) (((x) & SDMA_PKT_WRITE_TILED_BC_DW_6_x_mask) << SDMA_PKT_WRITE_TILED_BC_DW_6_x_shift) + +/*define for y field*/ +#define SDMA_PKT_WRITE_TILED_BC_DW_6_y_offset 6 +#define SDMA_PKT_WRITE_TILED_BC_DW_6_y_mask 0x00003FFF +#define SDMA_PKT_WRITE_TILED_BC_DW_6_y_shift 16 +#define SDMA_PKT_WRITE_TILED_BC_DW_6_Y(x) (((x) & SDMA_PKT_WRITE_TILED_BC_DW_6_y_mask) << SDMA_PKT_WRITE_TILED_BC_DW_6_y_shift) + +/*define for DW_7 word*/ +/*define for z field*/ +#define SDMA_PKT_WRITE_TILED_BC_DW_7_z_offset 7 +#define SDMA_PKT_WRITE_TILED_BC_DW_7_z_mask 0x000007FF +#define SDMA_PKT_WRITE_TILED_BC_DW_7_z_shift 0 +#define SDMA_PKT_WRITE_TILED_BC_DW_7_Z(x) (((x) & SDMA_PKT_WRITE_TILED_BC_DW_7_z_mask) << SDMA_PKT_WRITE_TILED_BC_DW_7_z_shift) + +/*define for sw field*/ +#define SDMA_PKT_WRITE_TILED_BC_DW_7_sw_offset 7 +#define SDMA_PKT_WRITE_TILED_BC_DW_7_sw_mask 0x00000003 +#define SDMA_PKT_WRITE_TILED_BC_DW_7_sw_shift 24 +#define SDMA_PKT_WRITE_TILED_BC_DW_7_SW(x) (((x) & SDMA_PKT_WRITE_TILED_BC_DW_7_sw_mask) << SDMA_PKT_WRITE_TILED_BC_DW_7_sw_shift) + +/*define for COUNT word*/ +/*define for count field*/ +#define SDMA_PKT_WRITE_TILED_BC_COUNT_count_offset 8 +#define SDMA_PKT_WRITE_TILED_BC_COUNT_count_mask 0x000FFFFF +#define SDMA_PKT_WRITE_TILED_BC_COUNT_count_shift 2 +#define SDMA_PKT_WRITE_TILED_BC_COUNT_COUNT(x) (((x) & SDMA_PKT_WRITE_TILED_BC_COUNT_count_mask) << SDMA_PKT_WRITE_TILED_BC_COUNT_count_shift) + +/*define for DATA0 word*/ +/*define for data0 field*/ +#define SDMA_PKT_WRITE_TILED_BC_DATA0_data0_offset 9 +#define SDMA_PKT_WRITE_TILED_BC_DATA0_data0_mask 0xFFFFFFFF +#define SDMA_PKT_WRITE_TILED_BC_DATA0_data0_shift 0 +#define SDMA_PKT_WRITE_TILED_BC_DATA0_DATA0(x) (((x) & SDMA_PKT_WRITE_TILED_BC_DATA0_data0_mask) << SDMA_PKT_WRITE_TILED_BC_DATA0_data0_shift) + + +/* +** Definitions for SDMA_PKT_PTEPDE_COPY packet +*/ + +/*define for HEADER word*/ +/*define for op field*/ +#define SDMA_PKT_PTEPDE_COPY_HEADER_op_offset 0 +#define SDMA_PKT_PTEPDE_COPY_HEADER_op_mask 0x000000FF +#define SDMA_PKT_PTEPDE_COPY_HEADER_op_shift 0 +#define SDMA_PKT_PTEPDE_COPY_HEADER_OP(x) (((x) & SDMA_PKT_PTEPDE_COPY_HEADER_op_mask) << SDMA_PKT_PTEPDE_COPY_HEADER_op_shift) + +/*define for sub_op field*/ +#define SDMA_PKT_PTEPDE_COPY_HEADER_sub_op_offset 0 +#define SDMA_PKT_PTEPDE_COPY_HEADER_sub_op_mask 0x000000FF +#define SDMA_PKT_PTEPDE_COPY_HEADER_sub_op_shift 8 +#define SDMA_PKT_PTEPDE_COPY_HEADER_SUB_OP(x) (((x) & SDMA_PKT_PTEPDE_COPY_HEADER_sub_op_mask) << SDMA_PKT_PTEPDE_COPY_HEADER_sub_op_shift) + +/*define for tmz field*/ +#define SDMA_PKT_PTEPDE_COPY_HEADER_tmz_offset 0 +#define SDMA_PKT_PTEPDE_COPY_HEADER_tmz_mask 0x00000001 +#define SDMA_PKT_PTEPDE_COPY_HEADER_tmz_shift 18 +#define SDMA_PKT_PTEPDE_COPY_HEADER_TMZ(x) (((x) & SDMA_PKT_PTEPDE_COPY_HEADER_tmz_mask) << SDMA_PKT_PTEPDE_COPY_HEADER_tmz_shift) + +/*define for cpv field*/ +#define SDMA_PKT_PTEPDE_COPY_HEADER_cpv_offset 0 +#define SDMA_PKT_PTEPDE_COPY_HEADER_cpv_mask 0x00000001 +#define SDMA_PKT_PTEPDE_COPY_HEADER_cpv_shift 28 +#define SDMA_PKT_PTEPDE_COPY_HEADER_CPV(x) (((x) & SDMA_PKT_PTEPDE_COPY_HEADER_cpv_mask) << SDMA_PKT_PTEPDE_COPY_HEADER_cpv_shift) + +/*define for ptepde_op field*/ +#define SDMA_PKT_PTEPDE_COPY_HEADER_ptepde_op_offset 0 +#define SDMA_PKT_PTEPDE_COPY_HEADER_ptepde_op_mask 0x00000001 +#define SDMA_PKT_PTEPDE_COPY_HEADER_ptepde_op_shift 31 +#define SDMA_PKT_PTEPDE_COPY_HEADER_PTEPDE_OP(x) (((x) & SDMA_PKT_PTEPDE_COPY_HEADER_ptepde_op_mask) << SDMA_PKT_PTEPDE_COPY_HEADER_ptepde_op_shift) + +/*define for SRC_ADDR_LO word*/ +/*define for src_addr_31_0 field*/ +#define SDMA_PKT_PTEPDE_COPY_SRC_ADDR_LO_src_addr_31_0_offset 1 +#define SDMA_PKT_PTEPDE_COPY_SRC_ADDR_LO_src_addr_31_0_mask 0xFFFFFFFF +#define SDMA_PKT_PTEPDE_COPY_SRC_ADDR_LO_src_addr_31_0_shift 0 +#define SDMA_PKT_PTEPDE_COPY_SRC_ADDR_LO_SRC_ADDR_31_0(x) (((x) & SDMA_PKT_PTEPDE_COPY_SRC_ADDR_LO_src_addr_31_0_mask) << SDMA_PKT_PTEPDE_COPY_SRC_ADDR_LO_src_addr_31_0_shift) + +/*define for SRC_ADDR_HI word*/ +/*define for src_addr_63_32 field*/ +#define SDMA_PKT_PTEPDE_COPY_SRC_ADDR_HI_src_addr_63_32_offset 2 +#define SDMA_PKT_PTEPDE_COPY_SRC_ADDR_HI_src_addr_63_32_mask 0xFFFFFFFF +#define SDMA_PKT_PTEPDE_COPY_SRC_ADDR_HI_src_addr_63_32_shift 0 +#define SDMA_PKT_PTEPDE_COPY_SRC_ADDR_HI_SRC_ADDR_63_32(x) (((x) & SDMA_PKT_PTEPDE_COPY_SRC_ADDR_HI_src_addr_63_32_mask) << SDMA_PKT_PTEPDE_COPY_SRC_ADDR_HI_src_addr_63_32_shift) + +/*define for DST_ADDR_LO word*/ +/*define for dst_addr_31_0 field*/ +#define SDMA_PKT_PTEPDE_COPY_DST_ADDR_LO_dst_addr_31_0_offset 3 +#define SDMA_PKT_PTEPDE_COPY_DST_ADDR_LO_dst_addr_31_0_mask 0xFFFFFFFF +#define SDMA_PKT_PTEPDE_COPY_DST_ADDR_LO_dst_addr_31_0_shift 0 +#define SDMA_PKT_PTEPDE_COPY_DST_ADDR_LO_DST_ADDR_31_0(x) (((x) & SDMA_PKT_PTEPDE_COPY_DST_ADDR_LO_dst_addr_31_0_mask) << SDMA_PKT_PTEPDE_COPY_DST_ADDR_LO_dst_addr_31_0_shift) + +/*define for DST_ADDR_HI word*/ +/*define for dst_addr_63_32 field*/ +#define SDMA_PKT_PTEPDE_COPY_DST_ADDR_HI_dst_addr_63_32_offset 4 +#define SDMA_PKT_PTEPDE_COPY_DST_ADDR_HI_dst_addr_63_32_mask 0xFFFFFFFF +#define SDMA_PKT_PTEPDE_COPY_DST_ADDR_HI_dst_addr_63_32_shift 0 +#define SDMA_PKT_PTEPDE_COPY_DST_ADDR_HI_DST_ADDR_63_32(x) (((x) & SDMA_PKT_PTEPDE_COPY_DST_ADDR_HI_dst_addr_63_32_mask) << SDMA_PKT_PTEPDE_COPY_DST_ADDR_HI_dst_addr_63_32_shift) + +/*define for MASK_DW0 word*/ +/*define for mask_dw0 field*/ +#define SDMA_PKT_PTEPDE_COPY_MASK_DW0_mask_dw0_offset 5 +#define SDMA_PKT_PTEPDE_COPY_MASK_DW0_mask_dw0_mask 0xFFFFFFFF +#define SDMA_PKT_PTEPDE_COPY_MASK_DW0_mask_dw0_shift 0 +#define SDMA_PKT_PTEPDE_COPY_MASK_DW0_MASK_DW0(x) (((x) & SDMA_PKT_PTEPDE_COPY_MASK_DW0_mask_dw0_mask) << SDMA_PKT_PTEPDE_COPY_MASK_DW0_mask_dw0_shift) + +/*define for MASK_DW1 word*/ +/*define for mask_dw1 field*/ +#define SDMA_PKT_PTEPDE_COPY_MASK_DW1_mask_dw1_offset 6 +#define SDMA_PKT_PTEPDE_COPY_MASK_DW1_mask_dw1_mask 0xFFFFFFFF +#define SDMA_PKT_PTEPDE_COPY_MASK_DW1_mask_dw1_shift 0 +#define SDMA_PKT_PTEPDE_COPY_MASK_DW1_MASK_DW1(x) (((x) & SDMA_PKT_PTEPDE_COPY_MASK_DW1_mask_dw1_mask) << SDMA_PKT_PTEPDE_COPY_MASK_DW1_mask_dw1_shift) + +/*define for COUNT word*/ +/*define for count field*/ +#define SDMA_PKT_PTEPDE_COPY_COUNT_count_offset 7 +#define SDMA_PKT_PTEPDE_COPY_COUNT_count_mask 0x0007FFFF +#define SDMA_PKT_PTEPDE_COPY_COUNT_count_shift 0 +#define SDMA_PKT_PTEPDE_COPY_COUNT_COUNT(x) (((x) & SDMA_PKT_PTEPDE_COPY_COUNT_count_mask) << SDMA_PKT_PTEPDE_COPY_COUNT_count_shift) + +/*define for dst_cache_policy field*/ +#define SDMA_PKT_PTEPDE_COPY_COUNT_dst_cache_policy_offset 7 +#define SDMA_PKT_PTEPDE_COPY_COUNT_dst_cache_policy_mask 0x00000007 +#define SDMA_PKT_PTEPDE_COPY_COUNT_dst_cache_policy_shift 22 +#define SDMA_PKT_PTEPDE_COPY_COUNT_DST_CACHE_POLICY(x) (((x) & SDMA_PKT_PTEPDE_COPY_COUNT_dst_cache_policy_mask) << SDMA_PKT_PTEPDE_COPY_COUNT_dst_cache_policy_shift) + +/*define for src_cache_policy field*/ +#define SDMA_PKT_PTEPDE_COPY_COUNT_src_cache_policy_offset 7 +#define SDMA_PKT_PTEPDE_COPY_COUNT_src_cache_policy_mask 0x00000007 +#define SDMA_PKT_PTEPDE_COPY_COUNT_src_cache_policy_shift 29 +#define SDMA_PKT_PTEPDE_COPY_COUNT_SRC_CACHE_POLICY(x) (((x) & SDMA_PKT_PTEPDE_COPY_COUNT_src_cache_policy_mask) << SDMA_PKT_PTEPDE_COPY_COUNT_src_cache_policy_shift) + + +/* +** Definitions for SDMA_PKT_PTEPDE_COPY_BACKWARDS packet +*/ + +/*define for HEADER word*/ +/*define for op field*/ +#define SDMA_PKT_PTEPDE_COPY_BACKWARDS_HEADER_op_offset 0 +#define SDMA_PKT_PTEPDE_COPY_BACKWARDS_HEADER_op_mask 0x000000FF +#define SDMA_PKT_PTEPDE_COPY_BACKWARDS_HEADER_op_shift 0 +#define SDMA_PKT_PTEPDE_COPY_BACKWARDS_HEADER_OP(x) (((x) & SDMA_PKT_PTEPDE_COPY_BACKWARDS_HEADER_op_mask) << SDMA_PKT_PTEPDE_COPY_BACKWARDS_HEADER_op_shift) + +/*define for sub_op field*/ +#define SDMA_PKT_PTEPDE_COPY_BACKWARDS_HEADER_sub_op_offset 0 +#define SDMA_PKT_PTEPDE_COPY_BACKWARDS_HEADER_sub_op_mask 0x000000FF +#define SDMA_PKT_PTEPDE_COPY_BACKWARDS_HEADER_sub_op_shift 8 +#define SDMA_PKT_PTEPDE_COPY_BACKWARDS_HEADER_SUB_OP(x) (((x) & SDMA_PKT_PTEPDE_COPY_BACKWARDS_HEADER_sub_op_mask) << SDMA_PKT_PTEPDE_COPY_BACKWARDS_HEADER_sub_op_shift) + +/*define for pte_size field*/ +#define SDMA_PKT_PTEPDE_COPY_BACKWARDS_HEADER_pte_size_offset 0 +#define SDMA_PKT_PTEPDE_COPY_BACKWARDS_HEADER_pte_size_mask 0x00000003 +#define SDMA_PKT_PTEPDE_COPY_BACKWARDS_HEADER_pte_size_shift 28 +#define SDMA_PKT_PTEPDE_COPY_BACKWARDS_HEADER_PTE_SIZE(x) (((x) & SDMA_PKT_PTEPDE_COPY_BACKWARDS_HEADER_pte_size_mask) << SDMA_PKT_PTEPDE_COPY_BACKWARDS_HEADER_pte_size_shift) + +/*define for direction field*/ +#define SDMA_PKT_PTEPDE_COPY_BACKWARDS_HEADER_direction_offset 0 +#define SDMA_PKT_PTEPDE_COPY_BACKWARDS_HEADER_direction_mask 0x00000001 +#define SDMA_PKT_PTEPDE_COPY_BACKWARDS_HEADER_direction_shift 30 +#define SDMA_PKT_PTEPDE_COPY_BACKWARDS_HEADER_DIRECTION(x) (((x) & SDMA_PKT_PTEPDE_COPY_BACKWARDS_HEADER_direction_mask) << SDMA_PKT_PTEPDE_COPY_BACKWARDS_HEADER_direction_shift) + +/*define for ptepde_op field*/ +#define SDMA_PKT_PTEPDE_COPY_BACKWARDS_HEADER_ptepde_op_offset 0 +#define SDMA_PKT_PTEPDE_COPY_BACKWARDS_HEADER_ptepde_op_mask 0x00000001 +#define SDMA_PKT_PTEPDE_COPY_BACKWARDS_HEADER_ptepde_op_shift 31 +#define SDMA_PKT_PTEPDE_COPY_BACKWARDS_HEADER_PTEPDE_OP(x) (((x) & SDMA_PKT_PTEPDE_COPY_BACKWARDS_HEADER_ptepde_op_mask) << SDMA_PKT_PTEPDE_COPY_BACKWARDS_HEADER_ptepde_op_shift) + +/*define for SRC_ADDR_LO word*/ +/*define for src_addr_31_0 field*/ +#define SDMA_PKT_PTEPDE_COPY_BACKWARDS_SRC_ADDR_LO_src_addr_31_0_offset 1 +#define SDMA_PKT_PTEPDE_COPY_BACKWARDS_SRC_ADDR_LO_src_addr_31_0_mask 0xFFFFFFFF +#define SDMA_PKT_PTEPDE_COPY_BACKWARDS_SRC_ADDR_LO_src_addr_31_0_shift 0 +#define SDMA_PKT_PTEPDE_COPY_BACKWARDS_SRC_ADDR_LO_SRC_ADDR_31_0(x) (((x) & SDMA_PKT_PTEPDE_COPY_BACKWARDS_SRC_ADDR_LO_src_addr_31_0_mask) << SDMA_PKT_PTEPDE_COPY_BACKWARDS_SRC_ADDR_LO_src_addr_31_0_shift) + +/*define for SRC_ADDR_HI word*/ +/*define for src_addr_63_32 field*/ +#define SDMA_PKT_PTEPDE_COPY_BACKWARDS_SRC_ADDR_HI_src_addr_63_32_offset 2 +#define SDMA_PKT_PTEPDE_COPY_BACKWARDS_SRC_ADDR_HI_src_addr_63_32_mask 0xFFFFFFFF +#define SDMA_PKT_PTEPDE_COPY_BACKWARDS_SRC_ADDR_HI_src_addr_63_32_shift 0 +#define SDMA_PKT_PTEPDE_COPY_BACKWARDS_SRC_ADDR_HI_SRC_ADDR_63_32(x) (((x) & SDMA_PKT_PTEPDE_COPY_BACKWARDS_SRC_ADDR_HI_src_addr_63_32_mask) << SDMA_PKT_PTEPDE_COPY_BACKWARDS_SRC_ADDR_HI_src_addr_63_32_shift) + +/*define for DST_ADDR_LO word*/ +/*define for dst_addr_31_0 field*/ +#define SDMA_PKT_PTEPDE_COPY_BACKWARDS_DST_ADDR_LO_dst_addr_31_0_offset 3 +#define SDMA_PKT_PTEPDE_COPY_BACKWARDS_DST_ADDR_LO_dst_addr_31_0_mask 0xFFFFFFFF +#define SDMA_PKT_PTEPDE_COPY_BACKWARDS_DST_ADDR_LO_dst_addr_31_0_shift 0 +#define SDMA_PKT_PTEPDE_COPY_BACKWARDS_DST_ADDR_LO_DST_ADDR_31_0(x) (((x) & SDMA_PKT_PTEPDE_COPY_BACKWARDS_DST_ADDR_LO_dst_addr_31_0_mask) << SDMA_PKT_PTEPDE_COPY_BACKWARDS_DST_ADDR_LO_dst_addr_31_0_shift) + +/*define for DST_ADDR_HI word*/ +/*define for dst_addr_63_32 field*/ +#define SDMA_PKT_PTEPDE_COPY_BACKWARDS_DST_ADDR_HI_dst_addr_63_32_offset 4 +#define SDMA_PKT_PTEPDE_COPY_BACKWARDS_DST_ADDR_HI_dst_addr_63_32_mask 0xFFFFFFFF +#define SDMA_PKT_PTEPDE_COPY_BACKWARDS_DST_ADDR_HI_dst_addr_63_32_shift 0 +#define SDMA_PKT_PTEPDE_COPY_BACKWARDS_DST_ADDR_HI_DST_ADDR_63_32(x) (((x) & SDMA_PKT_PTEPDE_COPY_BACKWARDS_DST_ADDR_HI_dst_addr_63_32_mask) << SDMA_PKT_PTEPDE_COPY_BACKWARDS_DST_ADDR_HI_dst_addr_63_32_shift) + +/*define for MASK_BIT_FOR_DW word*/ +/*define for mask_first_xfer field*/ +#define SDMA_PKT_PTEPDE_COPY_BACKWARDS_MASK_BIT_FOR_DW_mask_first_xfer_offset 5 +#define SDMA_PKT_PTEPDE_COPY_BACKWARDS_MASK_BIT_FOR_DW_mask_first_xfer_mask 0x000000FF +#define SDMA_PKT_PTEPDE_COPY_BACKWARDS_MASK_BIT_FOR_DW_mask_first_xfer_shift 0 +#define SDMA_PKT_PTEPDE_COPY_BACKWARDS_MASK_BIT_FOR_DW_MASK_FIRST_XFER(x) (((x) & SDMA_PKT_PTEPDE_COPY_BACKWARDS_MASK_BIT_FOR_DW_mask_first_xfer_mask) << SDMA_PKT_PTEPDE_COPY_BACKWARDS_MASK_BIT_FOR_DW_mask_first_xfer_shift) + +/*define for mask_last_xfer field*/ +#define SDMA_PKT_PTEPDE_COPY_BACKWARDS_MASK_BIT_FOR_DW_mask_last_xfer_offset 5 +#define SDMA_PKT_PTEPDE_COPY_BACKWARDS_MASK_BIT_FOR_DW_mask_last_xfer_mask 0x000000FF +#define SDMA_PKT_PTEPDE_COPY_BACKWARDS_MASK_BIT_FOR_DW_mask_last_xfer_shift 8 +#define SDMA_PKT_PTEPDE_COPY_BACKWARDS_MASK_BIT_FOR_DW_MASK_LAST_XFER(x) (((x) & SDMA_PKT_PTEPDE_COPY_BACKWARDS_MASK_BIT_FOR_DW_mask_last_xfer_mask) << SDMA_PKT_PTEPDE_COPY_BACKWARDS_MASK_BIT_FOR_DW_mask_last_xfer_shift) + +/*define for COUNT_IN_32B_XFER word*/ +/*define for count field*/ +#define SDMA_PKT_PTEPDE_COPY_BACKWARDS_COUNT_IN_32B_XFER_count_offset 6 +#define SDMA_PKT_PTEPDE_COPY_BACKWARDS_COUNT_IN_32B_XFER_count_mask 0x0001FFFF +#define SDMA_PKT_PTEPDE_COPY_BACKWARDS_COUNT_IN_32B_XFER_count_shift 0 +#define SDMA_PKT_PTEPDE_COPY_BACKWARDS_COUNT_IN_32B_XFER_COUNT(x) (((x) & SDMA_PKT_PTEPDE_COPY_BACKWARDS_COUNT_IN_32B_XFER_count_mask) << SDMA_PKT_PTEPDE_COPY_BACKWARDS_COUNT_IN_32B_XFER_count_shift) + + +/* +** Definitions for SDMA_PKT_PTEPDE_RMW packet +*/ + +/*define for HEADER word*/ +/*define for op field*/ +#define SDMA_PKT_PTEPDE_RMW_HEADER_op_offset 0 +#define SDMA_PKT_PTEPDE_RMW_HEADER_op_mask 0x000000FF +#define SDMA_PKT_PTEPDE_RMW_HEADER_op_shift 0 +#define SDMA_PKT_PTEPDE_RMW_HEADER_OP(x) (((x) & SDMA_PKT_PTEPDE_RMW_HEADER_op_mask) << SDMA_PKT_PTEPDE_RMW_HEADER_op_shift) + +/*define for sub_op field*/ +#define SDMA_PKT_PTEPDE_RMW_HEADER_sub_op_offset 0 +#define SDMA_PKT_PTEPDE_RMW_HEADER_sub_op_mask 0x000000FF +#define SDMA_PKT_PTEPDE_RMW_HEADER_sub_op_shift 8 +#define SDMA_PKT_PTEPDE_RMW_HEADER_SUB_OP(x) (((x) & SDMA_PKT_PTEPDE_RMW_HEADER_sub_op_mask) << SDMA_PKT_PTEPDE_RMW_HEADER_sub_op_shift) + +/*define for mtype field*/ +#define SDMA_PKT_PTEPDE_RMW_HEADER_mtype_offset 0 +#define SDMA_PKT_PTEPDE_RMW_HEADER_mtype_mask 0x00000007 +#define SDMA_PKT_PTEPDE_RMW_HEADER_mtype_shift 16 +#define SDMA_PKT_PTEPDE_RMW_HEADER_MTYPE(x) (((x) & SDMA_PKT_PTEPDE_RMW_HEADER_mtype_mask) << SDMA_PKT_PTEPDE_RMW_HEADER_mtype_shift) + +/*define for gcc field*/ +#define SDMA_PKT_PTEPDE_RMW_HEADER_gcc_offset 0 +#define SDMA_PKT_PTEPDE_RMW_HEADER_gcc_mask 0x00000001 +#define SDMA_PKT_PTEPDE_RMW_HEADER_gcc_shift 19 +#define SDMA_PKT_PTEPDE_RMW_HEADER_GCC(x) (((x) & SDMA_PKT_PTEPDE_RMW_HEADER_gcc_mask) << SDMA_PKT_PTEPDE_RMW_HEADER_gcc_shift) + +/*define for sys field*/ +#define SDMA_PKT_PTEPDE_RMW_HEADER_sys_offset 0 +#define SDMA_PKT_PTEPDE_RMW_HEADER_sys_mask 0x00000001 +#define SDMA_PKT_PTEPDE_RMW_HEADER_sys_shift 20 +#define SDMA_PKT_PTEPDE_RMW_HEADER_SYS(x) (((x) & SDMA_PKT_PTEPDE_RMW_HEADER_sys_mask) << SDMA_PKT_PTEPDE_RMW_HEADER_sys_shift) + +/*define for snp field*/ +#define SDMA_PKT_PTEPDE_RMW_HEADER_snp_offset 0 +#define SDMA_PKT_PTEPDE_RMW_HEADER_snp_mask 0x00000001 +#define SDMA_PKT_PTEPDE_RMW_HEADER_snp_shift 22 +#define SDMA_PKT_PTEPDE_RMW_HEADER_SNP(x) (((x) & SDMA_PKT_PTEPDE_RMW_HEADER_snp_mask) << SDMA_PKT_PTEPDE_RMW_HEADER_snp_shift) + +/*define for gpa field*/ +#define SDMA_PKT_PTEPDE_RMW_HEADER_gpa_offset 0 +#define SDMA_PKT_PTEPDE_RMW_HEADER_gpa_mask 0x00000001 +#define SDMA_PKT_PTEPDE_RMW_HEADER_gpa_shift 23 +#define SDMA_PKT_PTEPDE_RMW_HEADER_GPA(x) (((x) & SDMA_PKT_PTEPDE_RMW_HEADER_gpa_mask) << SDMA_PKT_PTEPDE_RMW_HEADER_gpa_shift) + +/*define for l2_policy field*/ +#define SDMA_PKT_PTEPDE_RMW_HEADER_l2_policy_offset 0 +#define SDMA_PKT_PTEPDE_RMW_HEADER_l2_policy_mask 0x00000003 +#define SDMA_PKT_PTEPDE_RMW_HEADER_l2_policy_shift 24 +#define SDMA_PKT_PTEPDE_RMW_HEADER_L2_POLICY(x) (((x) & SDMA_PKT_PTEPDE_RMW_HEADER_l2_policy_mask) << SDMA_PKT_PTEPDE_RMW_HEADER_l2_policy_shift) + +/*define for llc_policy field*/ +#define SDMA_PKT_PTEPDE_RMW_HEADER_llc_policy_offset 0 +#define SDMA_PKT_PTEPDE_RMW_HEADER_llc_policy_mask 0x00000001 +#define SDMA_PKT_PTEPDE_RMW_HEADER_llc_policy_shift 26 +#define SDMA_PKT_PTEPDE_RMW_HEADER_LLC_POLICY(x) (((x) & SDMA_PKT_PTEPDE_RMW_HEADER_llc_policy_mask) << SDMA_PKT_PTEPDE_RMW_HEADER_llc_policy_shift) + +/*define for cpv field*/ +#define SDMA_PKT_PTEPDE_RMW_HEADER_cpv_offset 0 +#define SDMA_PKT_PTEPDE_RMW_HEADER_cpv_mask 0x00000001 +#define SDMA_PKT_PTEPDE_RMW_HEADER_cpv_shift 28 +#define SDMA_PKT_PTEPDE_RMW_HEADER_CPV(x) (((x) & SDMA_PKT_PTEPDE_RMW_HEADER_cpv_mask) << SDMA_PKT_PTEPDE_RMW_HEADER_cpv_shift) + +/*define for ADDR_LO word*/ +/*define for addr_31_0 field*/ +#define SDMA_PKT_PTEPDE_RMW_ADDR_LO_addr_31_0_offset 1 +#define SDMA_PKT_PTEPDE_RMW_ADDR_LO_addr_31_0_mask 0xFFFFFFFF +#define SDMA_PKT_PTEPDE_RMW_ADDR_LO_addr_31_0_shift 0 +#define SDMA_PKT_PTEPDE_RMW_ADDR_LO_ADDR_31_0(x) (((x) & SDMA_PKT_PTEPDE_RMW_ADDR_LO_addr_31_0_mask) << SDMA_PKT_PTEPDE_RMW_ADDR_LO_addr_31_0_shift) + +/*define for ADDR_HI word*/ +/*define for addr_63_32 field*/ +#define SDMA_PKT_PTEPDE_RMW_ADDR_HI_addr_63_32_offset 2 +#define SDMA_PKT_PTEPDE_RMW_ADDR_HI_addr_63_32_mask 0xFFFFFFFF +#define SDMA_PKT_PTEPDE_RMW_ADDR_HI_addr_63_32_shift 0 +#define SDMA_PKT_PTEPDE_RMW_ADDR_HI_ADDR_63_32(x) (((x) & SDMA_PKT_PTEPDE_RMW_ADDR_HI_addr_63_32_mask) << SDMA_PKT_PTEPDE_RMW_ADDR_HI_addr_63_32_shift) + +/*define for MASK_LO word*/ +/*define for mask_31_0 field*/ +#define SDMA_PKT_PTEPDE_RMW_MASK_LO_mask_31_0_offset 3 +#define SDMA_PKT_PTEPDE_RMW_MASK_LO_mask_31_0_mask 0xFFFFFFFF +#define SDMA_PKT_PTEPDE_RMW_MASK_LO_mask_31_0_shift 0 +#define SDMA_PKT_PTEPDE_RMW_MASK_LO_MASK_31_0(x) (((x) & SDMA_PKT_PTEPDE_RMW_MASK_LO_mask_31_0_mask) << SDMA_PKT_PTEPDE_RMW_MASK_LO_mask_31_0_shift) + +/*define for MASK_HI word*/ +/*define for mask_63_32 field*/ +#define SDMA_PKT_PTEPDE_RMW_MASK_HI_mask_63_32_offset 4 +#define SDMA_PKT_PTEPDE_RMW_MASK_HI_mask_63_32_mask 0xFFFFFFFF +#define SDMA_PKT_PTEPDE_RMW_MASK_HI_mask_63_32_shift 0 +#define SDMA_PKT_PTEPDE_RMW_MASK_HI_MASK_63_32(x) (((x) & SDMA_PKT_PTEPDE_RMW_MASK_HI_mask_63_32_mask) << SDMA_PKT_PTEPDE_RMW_MASK_HI_mask_63_32_shift) + +/*define for VALUE_LO word*/ +/*define for value_31_0 field*/ +#define SDMA_PKT_PTEPDE_RMW_VALUE_LO_value_31_0_offset 5 +#define SDMA_PKT_PTEPDE_RMW_VALUE_LO_value_31_0_mask 0xFFFFFFFF +#define SDMA_PKT_PTEPDE_RMW_VALUE_LO_value_31_0_shift 0 +#define SDMA_PKT_PTEPDE_RMW_VALUE_LO_VALUE_31_0(x) (((x) & SDMA_PKT_PTEPDE_RMW_VALUE_LO_value_31_0_mask) << SDMA_PKT_PTEPDE_RMW_VALUE_LO_value_31_0_shift) + +/*define for VALUE_HI word*/ +/*define for value_63_32 field*/ +#define SDMA_PKT_PTEPDE_RMW_VALUE_HI_value_63_32_offset 6 +#define SDMA_PKT_PTEPDE_RMW_VALUE_HI_value_63_32_mask 0xFFFFFFFF +#define SDMA_PKT_PTEPDE_RMW_VALUE_HI_value_63_32_shift 0 +#define SDMA_PKT_PTEPDE_RMW_VALUE_HI_VALUE_63_32(x) (((x) & SDMA_PKT_PTEPDE_RMW_VALUE_HI_value_63_32_mask) << SDMA_PKT_PTEPDE_RMW_VALUE_HI_value_63_32_shift) + +/*define for COUNT word*/ +/*define for num_of_pte field*/ +#define SDMA_PKT_PTEPDE_RMW_COUNT_num_of_pte_offset 7 +#define SDMA_PKT_PTEPDE_RMW_COUNT_num_of_pte_mask 0xFFFFFFFF +#define SDMA_PKT_PTEPDE_RMW_COUNT_num_of_pte_shift 0 +#define SDMA_PKT_PTEPDE_RMW_COUNT_NUM_OF_PTE(x) (((x) & SDMA_PKT_PTEPDE_RMW_COUNT_num_of_pte_mask) << SDMA_PKT_PTEPDE_RMW_COUNT_num_of_pte_shift) + + +/* +** Definitions for SDMA_PKT_REGISTER_RMW packet +*/ + +/*define for HEADER word*/ +/*define for op field*/ +#define SDMA_PKT_REGISTER_RMW_HEADER_op_offset 0 +#define SDMA_PKT_REGISTER_RMW_HEADER_op_mask 0x000000FF +#define SDMA_PKT_REGISTER_RMW_HEADER_op_shift 0 +#define SDMA_PKT_REGISTER_RMW_HEADER_OP(x) (((x) & SDMA_PKT_REGISTER_RMW_HEADER_op_mask) << SDMA_PKT_REGISTER_RMW_HEADER_op_shift) + +/*define for sub_op field*/ +#define SDMA_PKT_REGISTER_RMW_HEADER_sub_op_offset 0 +#define SDMA_PKT_REGISTER_RMW_HEADER_sub_op_mask 0x000000FF +#define SDMA_PKT_REGISTER_RMW_HEADER_sub_op_shift 8 +#define SDMA_PKT_REGISTER_RMW_HEADER_SUB_OP(x) (((x) & SDMA_PKT_REGISTER_RMW_HEADER_sub_op_mask) << SDMA_PKT_REGISTER_RMW_HEADER_sub_op_shift) + +/*define for ADDR word*/ +/*define for addr field*/ +#define SDMA_PKT_REGISTER_RMW_ADDR_addr_offset 1 +#define SDMA_PKT_REGISTER_RMW_ADDR_addr_mask 0x000FFFFF +#define SDMA_PKT_REGISTER_RMW_ADDR_addr_shift 0 +#define SDMA_PKT_REGISTER_RMW_ADDR_ADDR(x) (((x) & SDMA_PKT_REGISTER_RMW_ADDR_addr_mask) << SDMA_PKT_REGISTER_RMW_ADDR_addr_shift) + +/*define for aperture_id field*/ +#define SDMA_PKT_REGISTER_RMW_ADDR_aperture_id_offset 1 +#define SDMA_PKT_REGISTER_RMW_ADDR_aperture_id_mask 0x00000FFF +#define SDMA_PKT_REGISTER_RMW_ADDR_aperture_id_shift 20 +#define SDMA_PKT_REGISTER_RMW_ADDR_APERTURE_ID(x) (((x) & SDMA_PKT_REGISTER_RMW_ADDR_aperture_id_mask) << SDMA_PKT_REGISTER_RMW_ADDR_aperture_id_shift) + +/*define for MASK word*/ +/*define for mask field*/ +#define SDMA_PKT_REGISTER_RMW_MASK_mask_offset 2 +#define SDMA_PKT_REGISTER_RMW_MASK_mask_mask 0xFFFFFFFF +#define SDMA_PKT_REGISTER_RMW_MASK_mask_shift 0 +#define SDMA_PKT_REGISTER_RMW_MASK_MASK(x) (((x) & SDMA_PKT_REGISTER_RMW_MASK_mask_mask) << SDMA_PKT_REGISTER_RMW_MASK_mask_shift) + +/*define for VALUE word*/ +/*define for value field*/ +#define SDMA_PKT_REGISTER_RMW_VALUE_value_offset 3 +#define SDMA_PKT_REGISTER_RMW_VALUE_value_mask 0xFFFFFFFF +#define SDMA_PKT_REGISTER_RMW_VALUE_value_shift 0 +#define SDMA_PKT_REGISTER_RMW_VALUE_VALUE(x) (((x) & SDMA_PKT_REGISTER_RMW_VALUE_value_mask) << SDMA_PKT_REGISTER_RMW_VALUE_value_shift) + +/*define for MISC word*/ +/*define for stride field*/ +#define SDMA_PKT_REGISTER_RMW_MISC_stride_offset 4 +#define SDMA_PKT_REGISTER_RMW_MISC_stride_mask 0x000FFFFF +#define SDMA_PKT_REGISTER_RMW_MISC_stride_shift 0 +#define SDMA_PKT_REGISTER_RMW_MISC_STRIDE(x) (((x) & SDMA_PKT_REGISTER_RMW_MISC_stride_mask) << SDMA_PKT_REGISTER_RMW_MISC_stride_shift) + +/*define for num_of_reg field*/ +#define SDMA_PKT_REGISTER_RMW_MISC_num_of_reg_offset 4 +#define SDMA_PKT_REGISTER_RMW_MISC_num_of_reg_mask 0x00000FFF +#define SDMA_PKT_REGISTER_RMW_MISC_num_of_reg_shift 20 +#define SDMA_PKT_REGISTER_RMW_MISC_NUM_OF_REG(x) (((x) & SDMA_PKT_REGISTER_RMW_MISC_num_of_reg_mask) << SDMA_PKT_REGISTER_RMW_MISC_num_of_reg_shift) + + +/* +** Definitions for SDMA_PKT_WRITE_INCR packet +*/ + +/*define for HEADER word*/ +/*define for op field*/ +#define SDMA_PKT_WRITE_INCR_HEADER_op_offset 0 +#define SDMA_PKT_WRITE_INCR_HEADER_op_mask 0x000000FF +#define SDMA_PKT_WRITE_INCR_HEADER_op_shift 0 +#define SDMA_PKT_WRITE_INCR_HEADER_OP(x) (((x) & SDMA_PKT_WRITE_INCR_HEADER_op_mask) << SDMA_PKT_WRITE_INCR_HEADER_op_shift) + +/*define for sub_op field*/ +#define SDMA_PKT_WRITE_INCR_HEADER_sub_op_offset 0 +#define SDMA_PKT_WRITE_INCR_HEADER_sub_op_mask 0x000000FF +#define SDMA_PKT_WRITE_INCR_HEADER_sub_op_shift 8 +#define SDMA_PKT_WRITE_INCR_HEADER_SUB_OP(x) (((x) & SDMA_PKT_WRITE_INCR_HEADER_sub_op_mask) << SDMA_PKT_WRITE_INCR_HEADER_sub_op_shift) + +/*define for cache_policy field*/ +#define SDMA_PKT_WRITE_INCR_HEADER_cache_policy_offset 0 +#define SDMA_PKT_WRITE_INCR_HEADER_cache_policy_mask 0x00000007 +#define SDMA_PKT_WRITE_INCR_HEADER_cache_policy_shift 24 +#define SDMA_PKT_WRITE_INCR_HEADER_CACHE_POLICY(x) (((x) & SDMA_PKT_WRITE_INCR_HEADER_cache_policy_mask) << SDMA_PKT_WRITE_INCR_HEADER_cache_policy_shift) + +/*define for cpv field*/ +#define SDMA_PKT_WRITE_INCR_HEADER_cpv_offset 0 +#define SDMA_PKT_WRITE_INCR_HEADER_cpv_mask 0x00000001 +#define SDMA_PKT_WRITE_INCR_HEADER_cpv_shift 28 +#define SDMA_PKT_WRITE_INCR_HEADER_CPV(x) (((x) & SDMA_PKT_WRITE_INCR_HEADER_cpv_mask) << SDMA_PKT_WRITE_INCR_HEADER_cpv_shift) + +/*define for DST_ADDR_LO word*/ +/*define for dst_addr_31_0 field*/ +#define SDMA_PKT_WRITE_INCR_DST_ADDR_LO_dst_addr_31_0_offset 1 +#define SDMA_PKT_WRITE_INCR_DST_ADDR_LO_dst_addr_31_0_mask 0xFFFFFFFF +#define SDMA_PKT_WRITE_INCR_DST_ADDR_LO_dst_addr_31_0_shift 0 +#define SDMA_PKT_WRITE_INCR_DST_ADDR_LO_DST_ADDR_31_0(x) (((x) & SDMA_PKT_WRITE_INCR_DST_ADDR_LO_dst_addr_31_0_mask) << SDMA_PKT_WRITE_INCR_DST_ADDR_LO_dst_addr_31_0_shift) + +/*define for DST_ADDR_HI word*/ +/*define for dst_addr_63_32 field*/ +#define SDMA_PKT_WRITE_INCR_DST_ADDR_HI_dst_addr_63_32_offset 2 +#define SDMA_PKT_WRITE_INCR_DST_ADDR_HI_dst_addr_63_32_mask 0xFFFFFFFF +#define SDMA_PKT_WRITE_INCR_DST_ADDR_HI_dst_addr_63_32_shift 0 +#define SDMA_PKT_WRITE_INCR_DST_ADDR_HI_DST_ADDR_63_32(x) (((x) & SDMA_PKT_WRITE_INCR_DST_ADDR_HI_dst_addr_63_32_mask) << SDMA_PKT_WRITE_INCR_DST_ADDR_HI_dst_addr_63_32_shift) + +/*define for MASK_DW0 word*/ +/*define for mask_dw0 field*/ +#define SDMA_PKT_WRITE_INCR_MASK_DW0_mask_dw0_offset 3 +#define SDMA_PKT_WRITE_INCR_MASK_DW0_mask_dw0_mask 0xFFFFFFFF +#define SDMA_PKT_WRITE_INCR_MASK_DW0_mask_dw0_shift 0 +#define SDMA_PKT_WRITE_INCR_MASK_DW0_MASK_DW0(x) (((x) & SDMA_PKT_WRITE_INCR_MASK_DW0_mask_dw0_mask) << SDMA_PKT_WRITE_INCR_MASK_DW0_mask_dw0_shift) + +/*define for MASK_DW1 word*/ +/*define for mask_dw1 field*/ +#define SDMA_PKT_WRITE_INCR_MASK_DW1_mask_dw1_offset 4 +#define SDMA_PKT_WRITE_INCR_MASK_DW1_mask_dw1_mask 0xFFFFFFFF +#define SDMA_PKT_WRITE_INCR_MASK_DW1_mask_dw1_shift 0 +#define SDMA_PKT_WRITE_INCR_MASK_DW1_MASK_DW1(x) (((x) & SDMA_PKT_WRITE_INCR_MASK_DW1_mask_dw1_mask) << SDMA_PKT_WRITE_INCR_MASK_DW1_mask_dw1_shift) + +/*define for INIT_DW0 word*/ +/*define for init_dw0 field*/ +#define SDMA_PKT_WRITE_INCR_INIT_DW0_init_dw0_offset 5 +#define SDMA_PKT_WRITE_INCR_INIT_DW0_init_dw0_mask 0xFFFFFFFF +#define SDMA_PKT_WRITE_INCR_INIT_DW0_init_dw0_shift 0 +#define SDMA_PKT_WRITE_INCR_INIT_DW0_INIT_DW0(x) (((x) & SDMA_PKT_WRITE_INCR_INIT_DW0_init_dw0_mask) << SDMA_PKT_WRITE_INCR_INIT_DW0_init_dw0_shift) + +/*define for INIT_DW1 word*/ +/*define for init_dw1 field*/ +#define SDMA_PKT_WRITE_INCR_INIT_DW1_init_dw1_offset 6 +#define SDMA_PKT_WRITE_INCR_INIT_DW1_init_dw1_mask 0xFFFFFFFF +#define SDMA_PKT_WRITE_INCR_INIT_DW1_init_dw1_shift 0 +#define SDMA_PKT_WRITE_INCR_INIT_DW1_INIT_DW1(x) (((x) & SDMA_PKT_WRITE_INCR_INIT_DW1_init_dw1_mask) << SDMA_PKT_WRITE_INCR_INIT_DW1_init_dw1_shift) + +/*define for INCR_DW0 word*/ +/*define for incr_dw0 field*/ +#define SDMA_PKT_WRITE_INCR_INCR_DW0_incr_dw0_offset 7 +#define SDMA_PKT_WRITE_INCR_INCR_DW0_incr_dw0_mask 0xFFFFFFFF +#define SDMA_PKT_WRITE_INCR_INCR_DW0_incr_dw0_shift 0 +#define SDMA_PKT_WRITE_INCR_INCR_DW0_INCR_DW0(x) (((x) & SDMA_PKT_WRITE_INCR_INCR_DW0_incr_dw0_mask) << SDMA_PKT_WRITE_INCR_INCR_DW0_incr_dw0_shift) + +/*define for INCR_DW1 word*/ +/*define for incr_dw1 field*/ +#define SDMA_PKT_WRITE_INCR_INCR_DW1_incr_dw1_offset 8 +#define SDMA_PKT_WRITE_INCR_INCR_DW1_incr_dw1_mask 0xFFFFFFFF +#define SDMA_PKT_WRITE_INCR_INCR_DW1_incr_dw1_shift 0 +#define SDMA_PKT_WRITE_INCR_INCR_DW1_INCR_DW1(x) (((x) & SDMA_PKT_WRITE_INCR_INCR_DW1_incr_dw1_mask) << SDMA_PKT_WRITE_INCR_INCR_DW1_incr_dw1_shift) + +/*define for COUNT word*/ +/*define for count field*/ +#define SDMA_PKT_WRITE_INCR_COUNT_count_offset 9 +#define SDMA_PKT_WRITE_INCR_COUNT_count_mask 0x0007FFFF +#define SDMA_PKT_WRITE_INCR_COUNT_count_shift 0 +#define SDMA_PKT_WRITE_INCR_COUNT_COUNT(x) (((x) & SDMA_PKT_WRITE_INCR_COUNT_count_mask) << SDMA_PKT_WRITE_INCR_COUNT_count_shift) + + +/* +** Definitions for SDMA_PKT_INDIRECT packet +*/ + +/*define for HEADER word*/ +/*define for op field*/ +#define SDMA_PKT_INDIRECT_HEADER_op_offset 0 +#define SDMA_PKT_INDIRECT_HEADER_op_mask 0x000000FF +#define SDMA_PKT_INDIRECT_HEADER_op_shift 0 +#define SDMA_PKT_INDIRECT_HEADER_OP(x) (((x) & SDMA_PKT_INDIRECT_HEADER_op_mask) << SDMA_PKT_INDIRECT_HEADER_op_shift) + +/*define for sub_op field*/ +#define SDMA_PKT_INDIRECT_HEADER_sub_op_offset 0 +#define SDMA_PKT_INDIRECT_HEADER_sub_op_mask 0x000000FF +#define SDMA_PKT_INDIRECT_HEADER_sub_op_shift 8 +#define SDMA_PKT_INDIRECT_HEADER_SUB_OP(x) (((x) & SDMA_PKT_INDIRECT_HEADER_sub_op_mask) << SDMA_PKT_INDIRECT_HEADER_sub_op_shift) + +/*define for vmid field*/ +#define SDMA_PKT_INDIRECT_HEADER_vmid_offset 0 +#define SDMA_PKT_INDIRECT_HEADER_vmid_mask 0x0000000F +#define SDMA_PKT_INDIRECT_HEADER_vmid_shift 16 +#define SDMA_PKT_INDIRECT_HEADER_VMID(x) (((x) & SDMA_PKT_INDIRECT_HEADER_vmid_mask) << SDMA_PKT_INDIRECT_HEADER_vmid_shift) + +/*define for priv field*/ +#define SDMA_PKT_INDIRECT_HEADER_priv_offset 0 +#define SDMA_PKT_INDIRECT_HEADER_priv_mask 0x00000001 +#define SDMA_PKT_INDIRECT_HEADER_priv_shift 31 +#define SDMA_PKT_INDIRECT_HEADER_PRIV(x) (((x) & SDMA_PKT_INDIRECT_HEADER_priv_mask) << SDMA_PKT_INDIRECT_HEADER_priv_shift) + +/*define for BASE_LO word*/ +/*define for ib_base_31_0 field*/ +#define SDMA_PKT_INDIRECT_BASE_LO_ib_base_31_0_offset 1 +#define SDMA_PKT_INDIRECT_BASE_LO_ib_base_31_0_mask 0xFFFFFFFF +#define SDMA_PKT_INDIRECT_BASE_LO_ib_base_31_0_shift 0 +#define SDMA_PKT_INDIRECT_BASE_LO_IB_BASE_31_0(x) (((x) & SDMA_PKT_INDIRECT_BASE_LO_ib_base_31_0_mask) << SDMA_PKT_INDIRECT_BASE_LO_ib_base_31_0_shift) + +/*define for BASE_HI word*/ +/*define for ib_base_63_32 field*/ +#define SDMA_PKT_INDIRECT_BASE_HI_ib_base_63_32_offset 2 +#define SDMA_PKT_INDIRECT_BASE_HI_ib_base_63_32_mask 0xFFFFFFFF +#define SDMA_PKT_INDIRECT_BASE_HI_ib_base_63_32_shift 0 +#define SDMA_PKT_INDIRECT_BASE_HI_IB_BASE_63_32(x) (((x) & SDMA_PKT_INDIRECT_BASE_HI_ib_base_63_32_mask) << SDMA_PKT_INDIRECT_BASE_HI_ib_base_63_32_shift) + +/*define for IB_SIZE word*/ +/*define for ib_size field*/ +#define SDMA_PKT_INDIRECT_IB_SIZE_ib_size_offset 3 +#define SDMA_PKT_INDIRECT_IB_SIZE_ib_size_mask 0x000FFFFF +#define SDMA_PKT_INDIRECT_IB_SIZE_ib_size_shift 0 +#define SDMA_PKT_INDIRECT_IB_SIZE_IB_SIZE(x) (((x) & SDMA_PKT_INDIRECT_IB_SIZE_ib_size_mask) << SDMA_PKT_INDIRECT_IB_SIZE_ib_size_shift) + +/*define for CSA_ADDR_LO word*/ +/*define for csa_addr_31_0 field*/ +#define SDMA_PKT_INDIRECT_CSA_ADDR_LO_csa_addr_31_0_offset 4 +#define SDMA_PKT_INDIRECT_CSA_ADDR_LO_csa_addr_31_0_mask 0xFFFFFFFF +#define SDMA_PKT_INDIRECT_CSA_ADDR_LO_csa_addr_31_0_shift 0 +#define SDMA_PKT_INDIRECT_CSA_ADDR_LO_CSA_ADDR_31_0(x) (((x) & SDMA_PKT_INDIRECT_CSA_ADDR_LO_csa_addr_31_0_mask) << SDMA_PKT_INDIRECT_CSA_ADDR_LO_csa_addr_31_0_shift) + +/*define for CSA_ADDR_HI word*/ +/*define for csa_addr_63_32 field*/ +#define SDMA_PKT_INDIRECT_CSA_ADDR_HI_csa_addr_63_32_offset 5 +#define SDMA_PKT_INDIRECT_CSA_ADDR_HI_csa_addr_63_32_mask 0xFFFFFFFF +#define SDMA_PKT_INDIRECT_CSA_ADDR_HI_csa_addr_63_32_shift 0 +#define SDMA_PKT_INDIRECT_CSA_ADDR_HI_CSA_ADDR_63_32(x) (((x) & SDMA_PKT_INDIRECT_CSA_ADDR_HI_csa_addr_63_32_mask) << SDMA_PKT_INDIRECT_CSA_ADDR_HI_csa_addr_63_32_shift) + + +/* +** Definitions for SDMA_PKT_SEMAPHORE packet +*/ + +/*define for HEADER word*/ +/*define for op field*/ +#define SDMA_PKT_SEMAPHORE_HEADER_op_offset 0 +#define SDMA_PKT_SEMAPHORE_HEADER_op_mask 0x000000FF +#define SDMA_PKT_SEMAPHORE_HEADER_op_shift 0 +#define SDMA_PKT_SEMAPHORE_HEADER_OP(x) (((x) & SDMA_PKT_SEMAPHORE_HEADER_op_mask) << SDMA_PKT_SEMAPHORE_HEADER_op_shift) + +/*define for sub_op field*/ +#define SDMA_PKT_SEMAPHORE_HEADER_sub_op_offset 0 +#define SDMA_PKT_SEMAPHORE_HEADER_sub_op_mask 0x000000FF +#define SDMA_PKT_SEMAPHORE_HEADER_sub_op_shift 8 +#define SDMA_PKT_SEMAPHORE_HEADER_SUB_OP(x) (((x) & SDMA_PKT_SEMAPHORE_HEADER_sub_op_mask) << SDMA_PKT_SEMAPHORE_HEADER_sub_op_shift) + +/*define for write_one field*/ +#define SDMA_PKT_SEMAPHORE_HEADER_write_one_offset 0 +#define SDMA_PKT_SEMAPHORE_HEADER_write_one_mask 0x00000001 +#define SDMA_PKT_SEMAPHORE_HEADER_write_one_shift 29 +#define SDMA_PKT_SEMAPHORE_HEADER_WRITE_ONE(x) (((x) & SDMA_PKT_SEMAPHORE_HEADER_write_one_mask) << SDMA_PKT_SEMAPHORE_HEADER_write_one_shift) + +/*define for signal field*/ +#define SDMA_PKT_SEMAPHORE_HEADER_signal_offset 0 +#define SDMA_PKT_SEMAPHORE_HEADER_signal_mask 0x00000001 +#define SDMA_PKT_SEMAPHORE_HEADER_signal_shift 30 +#define SDMA_PKT_SEMAPHORE_HEADER_SIGNAL(x) (((x) & SDMA_PKT_SEMAPHORE_HEADER_signal_mask) << SDMA_PKT_SEMAPHORE_HEADER_signal_shift) + +/*define for mailbox field*/ +#define SDMA_PKT_SEMAPHORE_HEADER_mailbox_offset 0 +#define SDMA_PKT_SEMAPHORE_HEADER_mailbox_mask 0x00000001 +#define SDMA_PKT_SEMAPHORE_HEADER_mailbox_shift 31 +#define SDMA_PKT_SEMAPHORE_HEADER_MAILBOX(x) (((x) & SDMA_PKT_SEMAPHORE_HEADER_mailbox_mask) << SDMA_PKT_SEMAPHORE_HEADER_mailbox_shift) + +/*define for ADDR_LO word*/ +/*define for addr_31_0 field*/ +#define SDMA_PKT_SEMAPHORE_ADDR_LO_addr_31_0_offset 1 +#define SDMA_PKT_SEMAPHORE_ADDR_LO_addr_31_0_mask 0xFFFFFFFF +#define SDMA_PKT_SEMAPHORE_ADDR_LO_addr_31_0_shift 0 +#define SDMA_PKT_SEMAPHORE_ADDR_LO_ADDR_31_0(x) (((x) & SDMA_PKT_SEMAPHORE_ADDR_LO_addr_31_0_mask) << SDMA_PKT_SEMAPHORE_ADDR_LO_addr_31_0_shift) + +/*define for ADDR_HI word*/ +/*define for addr_63_32 field*/ +#define SDMA_PKT_SEMAPHORE_ADDR_HI_addr_63_32_offset 2 +#define SDMA_PKT_SEMAPHORE_ADDR_HI_addr_63_32_mask 0xFFFFFFFF +#define SDMA_PKT_SEMAPHORE_ADDR_HI_addr_63_32_shift 0 +#define SDMA_PKT_SEMAPHORE_ADDR_HI_ADDR_63_32(x) (((x) & SDMA_PKT_SEMAPHORE_ADDR_HI_addr_63_32_mask) << SDMA_PKT_SEMAPHORE_ADDR_HI_addr_63_32_shift) + + +/* +** Definitions for SDMA_PKT_MEM_INCR packet +*/ + +/*define for HEADER word*/ +/*define for op field*/ +#define SDMA_PKT_MEM_INCR_HEADER_op_offset 0 +#define SDMA_PKT_MEM_INCR_HEADER_op_mask 0x000000FF +#define SDMA_PKT_MEM_INCR_HEADER_op_shift 0 +#define SDMA_PKT_MEM_INCR_HEADER_OP(x) (((x) & SDMA_PKT_MEM_INCR_HEADER_op_mask) << SDMA_PKT_MEM_INCR_HEADER_op_shift) + +/*define for sub_op field*/ +#define SDMA_PKT_MEM_INCR_HEADER_sub_op_offset 0 +#define SDMA_PKT_MEM_INCR_HEADER_sub_op_mask 0x000000FF +#define SDMA_PKT_MEM_INCR_HEADER_sub_op_shift 8 +#define SDMA_PKT_MEM_INCR_HEADER_SUB_OP(x) (((x) & SDMA_PKT_MEM_INCR_HEADER_sub_op_mask) << SDMA_PKT_MEM_INCR_HEADER_sub_op_shift) + +/*define for l2_policy field*/ +#define SDMA_PKT_MEM_INCR_HEADER_l2_policy_offset 0 +#define SDMA_PKT_MEM_INCR_HEADER_l2_policy_mask 0x00000003 +#define SDMA_PKT_MEM_INCR_HEADER_l2_policy_shift 24 +#define SDMA_PKT_MEM_INCR_HEADER_L2_POLICY(x) (((x) & SDMA_PKT_MEM_INCR_HEADER_l2_policy_mask) << SDMA_PKT_MEM_INCR_HEADER_l2_policy_shift) + +/*define for llc_policy field*/ +#define SDMA_PKT_MEM_INCR_HEADER_llc_policy_offset 0 +#define SDMA_PKT_MEM_INCR_HEADER_llc_policy_mask 0x00000001 +#define SDMA_PKT_MEM_INCR_HEADER_llc_policy_shift 26 +#define SDMA_PKT_MEM_INCR_HEADER_LLC_POLICY(x) (((x) & SDMA_PKT_MEM_INCR_HEADER_llc_policy_mask) << SDMA_PKT_MEM_INCR_HEADER_llc_policy_shift) + +/*define for cpv field*/ +#define SDMA_PKT_MEM_INCR_HEADER_cpv_offset 0 +#define SDMA_PKT_MEM_INCR_HEADER_cpv_mask 0x00000001 +#define SDMA_PKT_MEM_INCR_HEADER_cpv_shift 28 +#define SDMA_PKT_MEM_INCR_HEADER_CPV(x) (((x) & SDMA_PKT_MEM_INCR_HEADER_cpv_mask) << SDMA_PKT_MEM_INCR_HEADER_cpv_shift) + +/*define for ADDR_LO word*/ +/*define for addr_31_0 field*/ +#define SDMA_PKT_MEM_INCR_ADDR_LO_addr_31_0_offset 1 +#define SDMA_PKT_MEM_INCR_ADDR_LO_addr_31_0_mask 0xFFFFFFFF +#define SDMA_PKT_MEM_INCR_ADDR_LO_addr_31_0_shift 0 +#define SDMA_PKT_MEM_INCR_ADDR_LO_ADDR_31_0(x) (((x) & SDMA_PKT_MEM_INCR_ADDR_LO_addr_31_0_mask) << SDMA_PKT_MEM_INCR_ADDR_LO_addr_31_0_shift) + +/*define for ADDR_HI word*/ +/*define for addr_63_32 field*/ +#define SDMA_PKT_MEM_INCR_ADDR_HI_addr_63_32_offset 2 +#define SDMA_PKT_MEM_INCR_ADDR_HI_addr_63_32_mask 0xFFFFFFFF +#define SDMA_PKT_MEM_INCR_ADDR_HI_addr_63_32_shift 0 +#define SDMA_PKT_MEM_INCR_ADDR_HI_ADDR_63_32(x) (((x) & SDMA_PKT_MEM_INCR_ADDR_HI_addr_63_32_mask) << SDMA_PKT_MEM_INCR_ADDR_HI_addr_63_32_shift) + + +/* +** Definitions for SDMA_PKT_VM_INVALIDATION packet +*/ + +/*define for HEADER word*/ +/*define for op field*/ +#define SDMA_PKT_VM_INVALIDATION_HEADER_op_offset 0 +#define SDMA_PKT_VM_INVALIDATION_HEADER_op_mask 0x000000FF +#define SDMA_PKT_VM_INVALIDATION_HEADER_op_shift 0 +#define SDMA_PKT_VM_INVALIDATION_HEADER_OP(x) (((x) & SDMA_PKT_VM_INVALIDATION_HEADER_op_mask) << SDMA_PKT_VM_INVALIDATION_HEADER_op_shift) + +/*define for sub_op field*/ +#define SDMA_PKT_VM_INVALIDATION_HEADER_sub_op_offset 0 +#define SDMA_PKT_VM_INVALIDATION_HEADER_sub_op_mask 0x000000FF +#define SDMA_PKT_VM_INVALIDATION_HEADER_sub_op_shift 8 +#define SDMA_PKT_VM_INVALIDATION_HEADER_SUB_OP(x) (((x) & SDMA_PKT_VM_INVALIDATION_HEADER_sub_op_mask) << SDMA_PKT_VM_INVALIDATION_HEADER_sub_op_shift) + +/*define for gfx_eng_id field*/ +#define SDMA_PKT_VM_INVALIDATION_HEADER_gfx_eng_id_offset 0 +#define SDMA_PKT_VM_INVALIDATION_HEADER_gfx_eng_id_mask 0x0000001F +#define SDMA_PKT_VM_INVALIDATION_HEADER_gfx_eng_id_shift 16 +#define SDMA_PKT_VM_INVALIDATION_HEADER_GFX_ENG_ID(x) (((x) & SDMA_PKT_VM_INVALIDATION_HEADER_gfx_eng_id_mask) << SDMA_PKT_VM_INVALIDATION_HEADER_gfx_eng_id_shift) + +/*define for mm_eng_id field*/ +#define SDMA_PKT_VM_INVALIDATION_HEADER_mm_eng_id_offset 0 +#define SDMA_PKT_VM_INVALIDATION_HEADER_mm_eng_id_mask 0x0000001F +#define SDMA_PKT_VM_INVALIDATION_HEADER_mm_eng_id_shift 24 +#define SDMA_PKT_VM_INVALIDATION_HEADER_MM_ENG_ID(x) (((x) & SDMA_PKT_VM_INVALIDATION_HEADER_mm_eng_id_mask) << SDMA_PKT_VM_INVALIDATION_HEADER_mm_eng_id_shift) + +/*define for INVALIDATEREQ word*/ +/*define for invalidatereq field*/ +#define SDMA_PKT_VM_INVALIDATION_INVALIDATEREQ_invalidatereq_offset 1 +#define SDMA_PKT_VM_INVALIDATION_INVALIDATEREQ_invalidatereq_mask 0xFFFFFFFF +#define SDMA_PKT_VM_INVALIDATION_INVALIDATEREQ_invalidatereq_shift 0 +#define SDMA_PKT_VM_INVALIDATION_INVALIDATEREQ_INVALIDATEREQ(x) (((x) & SDMA_PKT_VM_INVALIDATION_INVALIDATEREQ_invalidatereq_mask) << SDMA_PKT_VM_INVALIDATION_INVALIDATEREQ_invalidatereq_shift) + +/*define for ADDRESSRANGELO word*/ +/*define for addressrangelo field*/ +#define SDMA_PKT_VM_INVALIDATION_ADDRESSRANGELO_addressrangelo_offset 2 +#define SDMA_PKT_VM_INVALIDATION_ADDRESSRANGELO_addressrangelo_mask 0xFFFFFFFF +#define SDMA_PKT_VM_INVALIDATION_ADDRESSRANGELO_addressrangelo_shift 0 +#define SDMA_PKT_VM_INVALIDATION_ADDRESSRANGELO_ADDRESSRANGELO(x) (((x) & SDMA_PKT_VM_INVALIDATION_ADDRESSRANGELO_addressrangelo_mask) << SDMA_PKT_VM_INVALIDATION_ADDRESSRANGELO_addressrangelo_shift) + +/*define for ADDRESSRANGEHI word*/ +/*define for invalidateack field*/ +#define SDMA_PKT_VM_INVALIDATION_ADDRESSRANGEHI_invalidateack_offset 3 +#define SDMA_PKT_VM_INVALIDATION_ADDRESSRANGEHI_invalidateack_mask 0x0000FFFF +#define SDMA_PKT_VM_INVALIDATION_ADDRESSRANGEHI_invalidateack_shift 0 +#define SDMA_PKT_VM_INVALIDATION_ADDRESSRANGEHI_INVALIDATEACK(x) (((x) & SDMA_PKT_VM_INVALIDATION_ADDRESSRANGEHI_invalidateack_mask) << SDMA_PKT_VM_INVALIDATION_ADDRESSRANGEHI_invalidateack_shift) + +/*define for addressrangehi field*/ +#define SDMA_PKT_VM_INVALIDATION_ADDRESSRANGEHI_addressrangehi_offset 3 +#define SDMA_PKT_VM_INVALIDATION_ADDRESSRANGEHI_addressrangehi_mask 0x0000001F +#define SDMA_PKT_VM_INVALIDATION_ADDRESSRANGEHI_addressrangehi_shift 16 +#define SDMA_PKT_VM_INVALIDATION_ADDRESSRANGEHI_ADDRESSRANGEHI(x) (((x) & SDMA_PKT_VM_INVALIDATION_ADDRESSRANGEHI_addressrangehi_mask) << SDMA_PKT_VM_INVALIDATION_ADDRESSRANGEHI_addressrangehi_shift) + +/*define for reserved field*/ +#define SDMA_PKT_VM_INVALIDATION_ADDRESSRANGEHI_reserved_offset 3 +#define SDMA_PKT_VM_INVALIDATION_ADDRESSRANGEHI_reserved_mask 0x000001FF +#define SDMA_PKT_VM_INVALIDATION_ADDRESSRANGEHI_reserved_shift 23 +#define SDMA_PKT_VM_INVALIDATION_ADDRESSRANGEHI_RESERVED(x) (((x) & SDMA_PKT_VM_INVALIDATION_ADDRESSRANGEHI_reserved_mask) << SDMA_PKT_VM_INVALIDATION_ADDRESSRANGEHI_reserved_shift) + + +/* +** Definitions for SDMA_PKT_FENCE packet +*/ + +/*define for HEADER word*/ +/*define for op field*/ +#define SDMA_PKT_FENCE_HEADER_op_offset 0 +#define SDMA_PKT_FENCE_HEADER_op_mask 0x000000FF +#define SDMA_PKT_FENCE_HEADER_op_shift 0 +#define SDMA_PKT_FENCE_HEADER_OP(x) (((x) & SDMA_PKT_FENCE_HEADER_op_mask) << SDMA_PKT_FENCE_HEADER_op_shift) + +/*define for sub_op field*/ +#define SDMA_PKT_FENCE_HEADER_sub_op_offset 0 +#define SDMA_PKT_FENCE_HEADER_sub_op_mask 0x000000FF +#define SDMA_PKT_FENCE_HEADER_sub_op_shift 8 +#define SDMA_PKT_FENCE_HEADER_SUB_OP(x) (((x) & SDMA_PKT_FENCE_HEADER_sub_op_mask) << SDMA_PKT_FENCE_HEADER_sub_op_shift) + +/*define for mtype field*/ +#define SDMA_PKT_FENCE_HEADER_mtype_offset 0 +#define SDMA_PKT_FENCE_HEADER_mtype_mask 0x00000007 +#define SDMA_PKT_FENCE_HEADER_mtype_shift 16 +#define SDMA_PKT_FENCE_HEADER_MTYPE(x) (((x) & SDMA_PKT_FENCE_HEADER_mtype_mask) << SDMA_PKT_FENCE_HEADER_mtype_shift) + +/*define for gcc field*/ +#define SDMA_PKT_FENCE_HEADER_gcc_offset 0 +#define SDMA_PKT_FENCE_HEADER_gcc_mask 0x00000001 +#define SDMA_PKT_FENCE_HEADER_gcc_shift 19 +#define SDMA_PKT_FENCE_HEADER_GCC(x) (((x) & SDMA_PKT_FENCE_HEADER_gcc_mask) << SDMA_PKT_FENCE_HEADER_gcc_shift) + +/*define for sys field*/ +#define SDMA_PKT_FENCE_HEADER_sys_offset 0 +#define SDMA_PKT_FENCE_HEADER_sys_mask 0x00000001 +#define SDMA_PKT_FENCE_HEADER_sys_shift 20 +#define SDMA_PKT_FENCE_HEADER_SYS(x) (((x) & SDMA_PKT_FENCE_HEADER_sys_mask) << SDMA_PKT_FENCE_HEADER_sys_shift) + +/*define for snp field*/ +#define SDMA_PKT_FENCE_HEADER_snp_offset 0 +#define SDMA_PKT_FENCE_HEADER_snp_mask 0x00000001 +#define SDMA_PKT_FENCE_HEADER_snp_shift 22 +#define SDMA_PKT_FENCE_HEADER_SNP(x) (((x) & SDMA_PKT_FENCE_HEADER_snp_mask) << SDMA_PKT_FENCE_HEADER_snp_shift) + +/*define for gpa field*/ +#define SDMA_PKT_FENCE_HEADER_gpa_offset 0 +#define SDMA_PKT_FENCE_HEADER_gpa_mask 0x00000001 +#define SDMA_PKT_FENCE_HEADER_gpa_shift 23 +#define SDMA_PKT_FENCE_HEADER_GPA(x) (((x) & SDMA_PKT_FENCE_HEADER_gpa_mask) << SDMA_PKT_FENCE_HEADER_gpa_shift) + +/*define for l2_policy field*/ +#define SDMA_PKT_FENCE_HEADER_l2_policy_offset 0 +#define SDMA_PKT_FENCE_HEADER_l2_policy_mask 0x00000003 +#define SDMA_PKT_FENCE_HEADER_l2_policy_shift 24 +#define SDMA_PKT_FENCE_HEADER_L2_POLICY(x) (((x) & SDMA_PKT_FENCE_HEADER_l2_policy_mask) << SDMA_PKT_FENCE_HEADER_l2_policy_shift) + +/*define for llc_policy field*/ +#define SDMA_PKT_FENCE_HEADER_llc_policy_offset 0 +#define SDMA_PKT_FENCE_HEADER_llc_policy_mask 0x00000001 +#define SDMA_PKT_FENCE_HEADER_llc_policy_shift 26 +#define SDMA_PKT_FENCE_HEADER_LLC_POLICY(x) (((x) & SDMA_PKT_FENCE_HEADER_llc_policy_mask) << SDMA_PKT_FENCE_HEADER_llc_policy_shift) + +/*define for cpv field*/ +#define SDMA_PKT_FENCE_HEADER_cpv_offset 0 +#define SDMA_PKT_FENCE_HEADER_cpv_mask 0x00000001 +#define SDMA_PKT_FENCE_HEADER_cpv_shift 28 +#define SDMA_PKT_FENCE_HEADER_CPV(x) (((x) & SDMA_PKT_FENCE_HEADER_cpv_mask) << SDMA_PKT_FENCE_HEADER_cpv_shift) + +/*define for ADDR_LO word*/ +/*define for addr_31_0 field*/ +#define SDMA_PKT_FENCE_ADDR_LO_addr_31_0_offset 1 +#define SDMA_PKT_FENCE_ADDR_LO_addr_31_0_mask 0xFFFFFFFF +#define SDMA_PKT_FENCE_ADDR_LO_addr_31_0_shift 0 +#define SDMA_PKT_FENCE_ADDR_LO_ADDR_31_0(x) (((x) & SDMA_PKT_FENCE_ADDR_LO_addr_31_0_mask) << SDMA_PKT_FENCE_ADDR_LO_addr_31_0_shift) + +/*define for ADDR_HI word*/ +/*define for addr_63_32 field*/ +#define SDMA_PKT_FENCE_ADDR_HI_addr_63_32_offset 2 +#define SDMA_PKT_FENCE_ADDR_HI_addr_63_32_mask 0xFFFFFFFF +#define SDMA_PKT_FENCE_ADDR_HI_addr_63_32_shift 0 +#define SDMA_PKT_FENCE_ADDR_HI_ADDR_63_32(x) (((x) & SDMA_PKT_FENCE_ADDR_HI_addr_63_32_mask) << SDMA_PKT_FENCE_ADDR_HI_addr_63_32_shift) + +/*define for DATA word*/ +/*define for data field*/ +#define SDMA_PKT_FENCE_DATA_data_offset 3 +#define SDMA_PKT_FENCE_DATA_data_mask 0xFFFFFFFF +#define SDMA_PKT_FENCE_DATA_data_shift 0 +#define SDMA_PKT_FENCE_DATA_DATA(x) (((x) & SDMA_PKT_FENCE_DATA_data_mask) << SDMA_PKT_FENCE_DATA_data_shift) + + +/* +** Definitions for SDMA_PKT_SRBM_WRITE packet +*/ + +/*define for HEADER word*/ +/*define for op field*/ +#define SDMA_PKT_SRBM_WRITE_HEADER_op_offset 0 +#define SDMA_PKT_SRBM_WRITE_HEADER_op_mask 0x000000FF +#define SDMA_PKT_SRBM_WRITE_HEADER_op_shift 0 +#define SDMA_PKT_SRBM_WRITE_HEADER_OP(x) (((x) & SDMA_PKT_SRBM_WRITE_HEADER_op_mask) << SDMA_PKT_SRBM_WRITE_HEADER_op_shift) + +/*define for sub_op field*/ +#define SDMA_PKT_SRBM_WRITE_HEADER_sub_op_offset 0 +#define SDMA_PKT_SRBM_WRITE_HEADER_sub_op_mask 0x000000FF +#define SDMA_PKT_SRBM_WRITE_HEADER_sub_op_shift 8 +#define SDMA_PKT_SRBM_WRITE_HEADER_SUB_OP(x) (((x) & SDMA_PKT_SRBM_WRITE_HEADER_sub_op_mask) << SDMA_PKT_SRBM_WRITE_HEADER_sub_op_shift) + +/*define for byte_en field*/ +#define SDMA_PKT_SRBM_WRITE_HEADER_byte_en_offset 0 +#define SDMA_PKT_SRBM_WRITE_HEADER_byte_en_mask 0x0000000F +#define SDMA_PKT_SRBM_WRITE_HEADER_byte_en_shift 28 +#define SDMA_PKT_SRBM_WRITE_HEADER_BYTE_EN(x) (((x) & SDMA_PKT_SRBM_WRITE_HEADER_byte_en_mask) << SDMA_PKT_SRBM_WRITE_HEADER_byte_en_shift) + +/*define for ADDR word*/ +/*define for addr field*/ +#define SDMA_PKT_SRBM_WRITE_ADDR_addr_offset 1 +#define SDMA_PKT_SRBM_WRITE_ADDR_addr_mask 0x0003FFFF +#define SDMA_PKT_SRBM_WRITE_ADDR_addr_shift 0 +#define SDMA_PKT_SRBM_WRITE_ADDR_ADDR(x) (((x) & SDMA_PKT_SRBM_WRITE_ADDR_addr_mask) << SDMA_PKT_SRBM_WRITE_ADDR_addr_shift) + +/*define for apertureid field*/ +#define SDMA_PKT_SRBM_WRITE_ADDR_apertureid_offset 1 +#define SDMA_PKT_SRBM_WRITE_ADDR_apertureid_mask 0x00000FFF +#define SDMA_PKT_SRBM_WRITE_ADDR_apertureid_shift 20 +#define SDMA_PKT_SRBM_WRITE_ADDR_APERTUREID(x) (((x) & SDMA_PKT_SRBM_WRITE_ADDR_apertureid_mask) << SDMA_PKT_SRBM_WRITE_ADDR_apertureid_shift) + +/*define for DATA word*/ +/*define for data field*/ +#define SDMA_PKT_SRBM_WRITE_DATA_data_offset 2 +#define SDMA_PKT_SRBM_WRITE_DATA_data_mask 0xFFFFFFFF +#define SDMA_PKT_SRBM_WRITE_DATA_data_shift 0 +#define SDMA_PKT_SRBM_WRITE_DATA_DATA(x) (((x) & SDMA_PKT_SRBM_WRITE_DATA_data_mask) << SDMA_PKT_SRBM_WRITE_DATA_data_shift) + + +/* +** Definitions for SDMA_PKT_PRE_EXE packet +*/ + +/*define for HEADER word*/ +/*define for op field*/ +#define SDMA_PKT_PRE_EXE_HEADER_op_offset 0 +#define SDMA_PKT_PRE_EXE_HEADER_op_mask 0x000000FF +#define SDMA_PKT_PRE_EXE_HEADER_op_shift 0 +#define SDMA_PKT_PRE_EXE_HEADER_OP(x) (((x) & SDMA_PKT_PRE_EXE_HEADER_op_mask) << SDMA_PKT_PRE_EXE_HEADER_op_shift) + +/*define for sub_op field*/ +#define SDMA_PKT_PRE_EXE_HEADER_sub_op_offset 0 +#define SDMA_PKT_PRE_EXE_HEADER_sub_op_mask 0x000000FF +#define SDMA_PKT_PRE_EXE_HEADER_sub_op_shift 8 +#define SDMA_PKT_PRE_EXE_HEADER_SUB_OP(x) (((x) & SDMA_PKT_PRE_EXE_HEADER_sub_op_mask) << SDMA_PKT_PRE_EXE_HEADER_sub_op_shift) + +/*define for dev_sel field*/ +#define SDMA_PKT_PRE_EXE_HEADER_dev_sel_offset 0 +#define SDMA_PKT_PRE_EXE_HEADER_dev_sel_mask 0x000000FF +#define SDMA_PKT_PRE_EXE_HEADER_dev_sel_shift 16 +#define SDMA_PKT_PRE_EXE_HEADER_DEV_SEL(x) (((x) & SDMA_PKT_PRE_EXE_HEADER_dev_sel_mask) << SDMA_PKT_PRE_EXE_HEADER_dev_sel_shift) + +/*define for EXEC_COUNT word*/ +/*define for exec_count field*/ +#define SDMA_PKT_PRE_EXE_EXEC_COUNT_exec_count_offset 1 +#define SDMA_PKT_PRE_EXE_EXEC_COUNT_exec_count_mask 0x00003FFF +#define SDMA_PKT_PRE_EXE_EXEC_COUNT_exec_count_shift 0 +#define SDMA_PKT_PRE_EXE_EXEC_COUNT_EXEC_COUNT(x) (((x) & SDMA_PKT_PRE_EXE_EXEC_COUNT_exec_count_mask) << SDMA_PKT_PRE_EXE_EXEC_COUNT_exec_count_shift) + + +/* +** Definitions for SDMA_PKT_COND_EXE packet +*/ + +/*define for HEADER word*/ +/*define for op field*/ +#define SDMA_PKT_COND_EXE_HEADER_op_offset 0 +#define SDMA_PKT_COND_EXE_HEADER_op_mask 0x000000FF +#define SDMA_PKT_COND_EXE_HEADER_op_shift 0 +#define SDMA_PKT_COND_EXE_HEADER_OP(x) (((x) & SDMA_PKT_COND_EXE_HEADER_op_mask) << SDMA_PKT_COND_EXE_HEADER_op_shift) + +/*define for sub_op field*/ +#define SDMA_PKT_COND_EXE_HEADER_sub_op_offset 0 +#define SDMA_PKT_COND_EXE_HEADER_sub_op_mask 0x000000FF +#define SDMA_PKT_COND_EXE_HEADER_sub_op_shift 8 +#define SDMA_PKT_COND_EXE_HEADER_SUB_OP(x) (((x) & SDMA_PKT_COND_EXE_HEADER_sub_op_mask) << SDMA_PKT_COND_EXE_HEADER_sub_op_shift) + +/*define for cache_policy field*/ +#define SDMA_PKT_COND_EXE_HEADER_cache_policy_offset 0 +#define SDMA_PKT_COND_EXE_HEADER_cache_policy_mask 0x00000007 +#define SDMA_PKT_COND_EXE_HEADER_cache_policy_shift 24 +#define SDMA_PKT_COND_EXE_HEADER_CACHE_POLICY(x) (((x) & SDMA_PKT_COND_EXE_HEADER_cache_policy_mask) << SDMA_PKT_COND_EXE_HEADER_cache_policy_shift) + +/*define for cpv field*/ +#define SDMA_PKT_COND_EXE_HEADER_cpv_offset 0 +#define SDMA_PKT_COND_EXE_HEADER_cpv_mask 0x00000001 +#define SDMA_PKT_COND_EXE_HEADER_cpv_shift 28 +#define SDMA_PKT_COND_EXE_HEADER_CPV(x) (((x) & SDMA_PKT_COND_EXE_HEADER_cpv_mask) << SDMA_PKT_COND_EXE_HEADER_cpv_shift) + +/*define for ADDR_LO word*/ +/*define for addr_31_0 field*/ +#define SDMA_PKT_COND_EXE_ADDR_LO_addr_31_0_offset 1 +#define SDMA_PKT_COND_EXE_ADDR_LO_addr_31_0_mask 0xFFFFFFFF +#define SDMA_PKT_COND_EXE_ADDR_LO_addr_31_0_shift 0 +#define SDMA_PKT_COND_EXE_ADDR_LO_ADDR_31_0(x) (((x) & SDMA_PKT_COND_EXE_ADDR_LO_addr_31_0_mask) << SDMA_PKT_COND_EXE_ADDR_LO_addr_31_0_shift) + +/*define for ADDR_HI word*/ +/*define for addr_63_32 field*/ +#define SDMA_PKT_COND_EXE_ADDR_HI_addr_63_32_offset 2 +#define SDMA_PKT_COND_EXE_ADDR_HI_addr_63_32_mask 0xFFFFFFFF +#define SDMA_PKT_COND_EXE_ADDR_HI_addr_63_32_shift 0 +#define SDMA_PKT_COND_EXE_ADDR_HI_ADDR_63_32(x) (((x) & SDMA_PKT_COND_EXE_ADDR_HI_addr_63_32_mask) << SDMA_PKT_COND_EXE_ADDR_HI_addr_63_32_shift) + +/*define for REFERENCE word*/ +/*define for reference field*/ +#define SDMA_PKT_COND_EXE_REFERENCE_reference_offset 3 +#define SDMA_PKT_COND_EXE_REFERENCE_reference_mask 0xFFFFFFFF +#define SDMA_PKT_COND_EXE_REFERENCE_reference_shift 0 +#define SDMA_PKT_COND_EXE_REFERENCE_REFERENCE(x) (((x) & SDMA_PKT_COND_EXE_REFERENCE_reference_mask) << SDMA_PKT_COND_EXE_REFERENCE_reference_shift) + +/*define for EXEC_COUNT word*/ +/*define for exec_count field*/ +#define SDMA_PKT_COND_EXE_EXEC_COUNT_exec_count_offset 4 +#define SDMA_PKT_COND_EXE_EXEC_COUNT_exec_count_mask 0x00003FFF +#define SDMA_PKT_COND_EXE_EXEC_COUNT_exec_count_shift 0 +#define SDMA_PKT_COND_EXE_EXEC_COUNT_EXEC_COUNT(x) (((x) & SDMA_PKT_COND_EXE_EXEC_COUNT_exec_count_mask) << SDMA_PKT_COND_EXE_EXEC_COUNT_exec_count_shift) + + +/* +** Definitions for SDMA_PKT_CONSTANT_FILL packet +*/ + +/*define for HEADER word*/ +/*define for op field*/ +#define SDMA_PKT_CONSTANT_FILL_HEADER_op_offset 0 +#define SDMA_PKT_CONSTANT_FILL_HEADER_op_mask 0x000000FF +#define SDMA_PKT_CONSTANT_FILL_HEADER_op_shift 0 +#define SDMA_PKT_CONSTANT_FILL_HEADER_OP(x) (((x) & SDMA_PKT_CONSTANT_FILL_HEADER_op_mask) << SDMA_PKT_CONSTANT_FILL_HEADER_op_shift) + +/*define for sub_op field*/ +#define SDMA_PKT_CONSTANT_FILL_HEADER_sub_op_offset 0 +#define SDMA_PKT_CONSTANT_FILL_HEADER_sub_op_mask 0x000000FF +#define SDMA_PKT_CONSTANT_FILL_HEADER_sub_op_shift 8 +#define SDMA_PKT_CONSTANT_FILL_HEADER_SUB_OP(x) (((x) & SDMA_PKT_CONSTANT_FILL_HEADER_sub_op_mask) << SDMA_PKT_CONSTANT_FILL_HEADER_sub_op_shift) + +/*define for sw field*/ +#define SDMA_PKT_CONSTANT_FILL_HEADER_sw_offset 0 +#define SDMA_PKT_CONSTANT_FILL_HEADER_sw_mask 0x00000003 +#define SDMA_PKT_CONSTANT_FILL_HEADER_sw_shift 16 +#define SDMA_PKT_CONSTANT_FILL_HEADER_SW(x) (((x) & SDMA_PKT_CONSTANT_FILL_HEADER_sw_mask) << SDMA_PKT_CONSTANT_FILL_HEADER_sw_shift) + +/*define for cache_policy field*/ +#define SDMA_PKT_CONSTANT_FILL_HEADER_cache_policy_offset 0 +#define SDMA_PKT_CONSTANT_FILL_HEADER_cache_policy_mask 0x00000007 +#define SDMA_PKT_CONSTANT_FILL_HEADER_cache_policy_shift 24 +#define SDMA_PKT_CONSTANT_FILL_HEADER_CACHE_POLICY(x) (((x) & SDMA_PKT_CONSTANT_FILL_HEADER_cache_policy_mask) << SDMA_PKT_CONSTANT_FILL_HEADER_cache_policy_shift) + +/*define for cpv field*/ +#define SDMA_PKT_CONSTANT_FILL_HEADER_cpv_offset 0 +#define SDMA_PKT_CONSTANT_FILL_HEADER_cpv_mask 0x00000001 +#define SDMA_PKT_CONSTANT_FILL_HEADER_cpv_shift 28 +#define SDMA_PKT_CONSTANT_FILL_HEADER_CPV(x) (((x) & SDMA_PKT_CONSTANT_FILL_HEADER_cpv_mask) << SDMA_PKT_CONSTANT_FILL_HEADER_cpv_shift) + +/*define for fillsize field*/ +#define SDMA_PKT_CONSTANT_FILL_HEADER_fillsize_offset 0 +#define SDMA_PKT_CONSTANT_FILL_HEADER_fillsize_mask 0x00000003 +#define SDMA_PKT_CONSTANT_FILL_HEADER_fillsize_shift 30 +#define SDMA_PKT_CONSTANT_FILL_HEADER_FILLSIZE(x) (((x) & SDMA_PKT_CONSTANT_FILL_HEADER_fillsize_mask) << SDMA_PKT_CONSTANT_FILL_HEADER_fillsize_shift) + +/*define for DST_ADDR_LO word*/ +/*define for dst_addr_31_0 field*/ +#define SDMA_PKT_CONSTANT_FILL_DST_ADDR_LO_dst_addr_31_0_offset 1 +#define SDMA_PKT_CONSTANT_FILL_DST_ADDR_LO_dst_addr_31_0_mask 0xFFFFFFFF +#define SDMA_PKT_CONSTANT_FILL_DST_ADDR_LO_dst_addr_31_0_shift 0 +#define SDMA_PKT_CONSTANT_FILL_DST_ADDR_LO_DST_ADDR_31_0(x) (((x) & SDMA_PKT_CONSTANT_FILL_DST_ADDR_LO_dst_addr_31_0_mask) << SDMA_PKT_CONSTANT_FILL_DST_ADDR_LO_dst_addr_31_0_shift) + +/*define for DST_ADDR_HI word*/ +/*define for dst_addr_63_32 field*/ +#define SDMA_PKT_CONSTANT_FILL_DST_ADDR_HI_dst_addr_63_32_offset 2 +#define SDMA_PKT_CONSTANT_FILL_DST_ADDR_HI_dst_addr_63_32_mask 0xFFFFFFFF +#define SDMA_PKT_CONSTANT_FILL_DST_ADDR_HI_dst_addr_63_32_shift 0 +#define SDMA_PKT_CONSTANT_FILL_DST_ADDR_HI_DST_ADDR_63_32(x) (((x) & SDMA_PKT_CONSTANT_FILL_DST_ADDR_HI_dst_addr_63_32_mask) << SDMA_PKT_CONSTANT_FILL_DST_ADDR_HI_dst_addr_63_32_shift) + +/*define for DATA word*/ +/*define for src_data_31_0 field*/ +#define SDMA_PKT_CONSTANT_FILL_DATA_src_data_31_0_offset 3 +#define SDMA_PKT_CONSTANT_FILL_DATA_src_data_31_0_mask 0xFFFFFFFF +#define SDMA_PKT_CONSTANT_FILL_DATA_src_data_31_0_shift 0 +#define SDMA_PKT_CONSTANT_FILL_DATA_SRC_DATA_31_0(x) (((x) & SDMA_PKT_CONSTANT_FILL_DATA_src_data_31_0_mask) << SDMA_PKT_CONSTANT_FILL_DATA_src_data_31_0_shift) + +/*define for COUNT word*/ +/*define for count field*/ +#define SDMA_PKT_CONSTANT_FILL_COUNT_count_offset 4 +#define SDMA_PKT_CONSTANT_FILL_COUNT_count_mask 0x3FFFFFFF +#define SDMA_PKT_CONSTANT_FILL_COUNT_count_shift 0 +#define SDMA_PKT_CONSTANT_FILL_COUNT_COUNT(x) (((x) & SDMA_PKT_CONSTANT_FILL_COUNT_count_mask) << SDMA_PKT_CONSTANT_FILL_COUNT_count_shift) + + +/* +** Definitions for SDMA_PKT_DATA_FILL_MULTI packet +*/ + +/*define for HEADER word*/ +/*define for op field*/ +#define SDMA_PKT_DATA_FILL_MULTI_HEADER_op_offset 0 +#define SDMA_PKT_DATA_FILL_MULTI_HEADER_op_mask 0x000000FF +#define SDMA_PKT_DATA_FILL_MULTI_HEADER_op_shift 0 +#define SDMA_PKT_DATA_FILL_MULTI_HEADER_OP(x) (((x) & SDMA_PKT_DATA_FILL_MULTI_HEADER_op_mask) << SDMA_PKT_DATA_FILL_MULTI_HEADER_op_shift) + +/*define for sub_op field*/ +#define SDMA_PKT_DATA_FILL_MULTI_HEADER_sub_op_offset 0 +#define SDMA_PKT_DATA_FILL_MULTI_HEADER_sub_op_mask 0x000000FF +#define SDMA_PKT_DATA_FILL_MULTI_HEADER_sub_op_shift 8 +#define SDMA_PKT_DATA_FILL_MULTI_HEADER_SUB_OP(x) (((x) & SDMA_PKT_DATA_FILL_MULTI_HEADER_sub_op_mask) << SDMA_PKT_DATA_FILL_MULTI_HEADER_sub_op_shift) + +/*define for cache_policy field*/ +#define SDMA_PKT_DATA_FILL_MULTI_HEADER_cache_policy_offset 0 +#define SDMA_PKT_DATA_FILL_MULTI_HEADER_cache_policy_mask 0x00000007 +#define SDMA_PKT_DATA_FILL_MULTI_HEADER_cache_policy_shift 24 +#define SDMA_PKT_DATA_FILL_MULTI_HEADER_CACHE_POLICY(x) (((x) & SDMA_PKT_DATA_FILL_MULTI_HEADER_cache_policy_mask) << SDMA_PKT_DATA_FILL_MULTI_HEADER_cache_policy_shift) + +/*define for cpv field*/ +#define SDMA_PKT_DATA_FILL_MULTI_HEADER_cpv_offset 0 +#define SDMA_PKT_DATA_FILL_MULTI_HEADER_cpv_mask 0x00000001 +#define SDMA_PKT_DATA_FILL_MULTI_HEADER_cpv_shift 28 +#define SDMA_PKT_DATA_FILL_MULTI_HEADER_CPV(x) (((x) & SDMA_PKT_DATA_FILL_MULTI_HEADER_cpv_mask) << SDMA_PKT_DATA_FILL_MULTI_HEADER_cpv_shift) + +/*define for memlog_clr field*/ +#define SDMA_PKT_DATA_FILL_MULTI_HEADER_memlog_clr_offset 0 +#define SDMA_PKT_DATA_FILL_MULTI_HEADER_memlog_clr_mask 0x00000001 +#define SDMA_PKT_DATA_FILL_MULTI_HEADER_memlog_clr_shift 31 +#define SDMA_PKT_DATA_FILL_MULTI_HEADER_MEMLOG_CLR(x) (((x) & SDMA_PKT_DATA_FILL_MULTI_HEADER_memlog_clr_mask) << SDMA_PKT_DATA_FILL_MULTI_HEADER_memlog_clr_shift) + +/*define for BYTE_STRIDE word*/ +/*define for byte_stride field*/ +#define SDMA_PKT_DATA_FILL_MULTI_BYTE_STRIDE_byte_stride_offset 1 +#define SDMA_PKT_DATA_FILL_MULTI_BYTE_STRIDE_byte_stride_mask 0xFFFFFFFF +#define SDMA_PKT_DATA_FILL_MULTI_BYTE_STRIDE_byte_stride_shift 0 +#define SDMA_PKT_DATA_FILL_MULTI_BYTE_STRIDE_BYTE_STRIDE(x) (((x) & SDMA_PKT_DATA_FILL_MULTI_BYTE_STRIDE_byte_stride_mask) << SDMA_PKT_DATA_FILL_MULTI_BYTE_STRIDE_byte_stride_shift) + +/*define for DMA_COUNT word*/ +/*define for dma_count field*/ +#define SDMA_PKT_DATA_FILL_MULTI_DMA_COUNT_dma_count_offset 2 +#define SDMA_PKT_DATA_FILL_MULTI_DMA_COUNT_dma_count_mask 0xFFFFFFFF +#define SDMA_PKT_DATA_FILL_MULTI_DMA_COUNT_dma_count_shift 0 +#define SDMA_PKT_DATA_FILL_MULTI_DMA_COUNT_DMA_COUNT(x) (((x) & SDMA_PKT_DATA_FILL_MULTI_DMA_COUNT_dma_count_mask) << SDMA_PKT_DATA_FILL_MULTI_DMA_COUNT_dma_count_shift) + +/*define for DST_ADDR_LO word*/ +/*define for dst_addr_31_0 field*/ +#define SDMA_PKT_DATA_FILL_MULTI_DST_ADDR_LO_dst_addr_31_0_offset 3 +#define SDMA_PKT_DATA_FILL_MULTI_DST_ADDR_LO_dst_addr_31_0_mask 0xFFFFFFFF +#define SDMA_PKT_DATA_FILL_MULTI_DST_ADDR_LO_dst_addr_31_0_shift 0 +#define SDMA_PKT_DATA_FILL_MULTI_DST_ADDR_LO_DST_ADDR_31_0(x) (((x) & SDMA_PKT_DATA_FILL_MULTI_DST_ADDR_LO_dst_addr_31_0_mask) << SDMA_PKT_DATA_FILL_MULTI_DST_ADDR_LO_dst_addr_31_0_shift) + +/*define for DST_ADDR_HI word*/ +/*define for dst_addr_63_32 field*/ +#define SDMA_PKT_DATA_FILL_MULTI_DST_ADDR_HI_dst_addr_63_32_offset 4 +#define SDMA_PKT_DATA_FILL_MULTI_DST_ADDR_HI_dst_addr_63_32_mask 0xFFFFFFFF +#define SDMA_PKT_DATA_FILL_MULTI_DST_ADDR_HI_dst_addr_63_32_shift 0 +#define SDMA_PKT_DATA_FILL_MULTI_DST_ADDR_HI_DST_ADDR_63_32(x) (((x) & SDMA_PKT_DATA_FILL_MULTI_DST_ADDR_HI_dst_addr_63_32_mask) << SDMA_PKT_DATA_FILL_MULTI_DST_ADDR_HI_dst_addr_63_32_shift) + +/*define for BYTE_COUNT word*/ +/*define for count field*/ +#define SDMA_PKT_DATA_FILL_MULTI_BYTE_COUNT_count_offset 5 +#define SDMA_PKT_DATA_FILL_MULTI_BYTE_COUNT_count_mask 0x03FFFFFF +#define SDMA_PKT_DATA_FILL_MULTI_BYTE_COUNT_count_shift 0 +#define SDMA_PKT_DATA_FILL_MULTI_BYTE_COUNT_COUNT(x) (((x) & SDMA_PKT_DATA_FILL_MULTI_BYTE_COUNT_count_mask) << SDMA_PKT_DATA_FILL_MULTI_BYTE_COUNT_count_shift) + + +/* +** Definitions for SDMA_PKT_POLL_REGMEM packet +*/ + +/*define for HEADER word*/ +/*define for op field*/ +#define SDMA_PKT_POLL_REGMEM_HEADER_op_offset 0 +#define SDMA_PKT_POLL_REGMEM_HEADER_op_mask 0x000000FF +#define SDMA_PKT_POLL_REGMEM_HEADER_op_shift 0 +#define SDMA_PKT_POLL_REGMEM_HEADER_OP(x) (((x) & SDMA_PKT_POLL_REGMEM_HEADER_op_mask) << SDMA_PKT_POLL_REGMEM_HEADER_op_shift) + +/*define for sub_op field*/ +#define SDMA_PKT_POLL_REGMEM_HEADER_sub_op_offset 0 +#define SDMA_PKT_POLL_REGMEM_HEADER_sub_op_mask 0x000000FF +#define SDMA_PKT_POLL_REGMEM_HEADER_sub_op_shift 8 +#define SDMA_PKT_POLL_REGMEM_HEADER_SUB_OP(x) (((x) & SDMA_PKT_POLL_REGMEM_HEADER_sub_op_mask) << SDMA_PKT_POLL_REGMEM_HEADER_sub_op_shift) + +/*define for cache_policy field*/ +#define SDMA_PKT_POLL_REGMEM_HEADER_cache_policy_offset 0 +#define SDMA_PKT_POLL_REGMEM_HEADER_cache_policy_mask 0x00000007 +#define SDMA_PKT_POLL_REGMEM_HEADER_cache_policy_shift 20 +#define SDMA_PKT_POLL_REGMEM_HEADER_CACHE_POLICY(x) (((x) & SDMA_PKT_POLL_REGMEM_HEADER_cache_policy_mask) << SDMA_PKT_POLL_REGMEM_HEADER_cache_policy_shift) + +/*define for cpv field*/ +#define SDMA_PKT_POLL_REGMEM_HEADER_cpv_offset 0 +#define SDMA_PKT_POLL_REGMEM_HEADER_cpv_mask 0x00000001 +#define SDMA_PKT_POLL_REGMEM_HEADER_cpv_shift 24 +#define SDMA_PKT_POLL_REGMEM_HEADER_CPV(x) (((x) & SDMA_PKT_POLL_REGMEM_HEADER_cpv_mask) << SDMA_PKT_POLL_REGMEM_HEADER_cpv_shift) + +/*define for hdp_flush field*/ +#define SDMA_PKT_POLL_REGMEM_HEADER_hdp_flush_offset 0 +#define SDMA_PKT_POLL_REGMEM_HEADER_hdp_flush_mask 0x00000001 +#define SDMA_PKT_POLL_REGMEM_HEADER_hdp_flush_shift 26 +#define SDMA_PKT_POLL_REGMEM_HEADER_HDP_FLUSH(x) (((x) & SDMA_PKT_POLL_REGMEM_HEADER_hdp_flush_mask) << SDMA_PKT_POLL_REGMEM_HEADER_hdp_flush_shift) + +/*define for func field*/ +#define SDMA_PKT_POLL_REGMEM_HEADER_func_offset 0 +#define SDMA_PKT_POLL_REGMEM_HEADER_func_mask 0x00000007 +#define SDMA_PKT_POLL_REGMEM_HEADER_func_shift 28 +#define SDMA_PKT_POLL_REGMEM_HEADER_FUNC(x) (((x) & SDMA_PKT_POLL_REGMEM_HEADER_func_mask) << SDMA_PKT_POLL_REGMEM_HEADER_func_shift) + +/*define for mem_poll field*/ +#define SDMA_PKT_POLL_REGMEM_HEADER_mem_poll_offset 0 +#define SDMA_PKT_POLL_REGMEM_HEADER_mem_poll_mask 0x00000001 +#define SDMA_PKT_POLL_REGMEM_HEADER_mem_poll_shift 31 +#define SDMA_PKT_POLL_REGMEM_HEADER_MEM_POLL(x) (((x) & SDMA_PKT_POLL_REGMEM_HEADER_mem_poll_mask) << SDMA_PKT_POLL_REGMEM_HEADER_mem_poll_shift) + +/*define for ADDR_LO word*/ +/*define for addr_31_0 field*/ +#define SDMA_PKT_POLL_REGMEM_ADDR_LO_addr_31_0_offset 1 +#define SDMA_PKT_POLL_REGMEM_ADDR_LO_addr_31_0_mask 0xFFFFFFFF +#define SDMA_PKT_POLL_REGMEM_ADDR_LO_addr_31_0_shift 0 +#define SDMA_PKT_POLL_REGMEM_ADDR_LO_ADDR_31_0(x) (((x) & SDMA_PKT_POLL_REGMEM_ADDR_LO_addr_31_0_mask) << SDMA_PKT_POLL_REGMEM_ADDR_LO_addr_31_0_shift) + +/*define for ADDR_HI word*/ +/*define for addr_63_32 field*/ +#define SDMA_PKT_POLL_REGMEM_ADDR_HI_addr_63_32_offset 2 +#define SDMA_PKT_POLL_REGMEM_ADDR_HI_addr_63_32_mask 0xFFFFFFFF +#define SDMA_PKT_POLL_REGMEM_ADDR_HI_addr_63_32_shift 0 +#define SDMA_PKT_POLL_REGMEM_ADDR_HI_ADDR_63_32(x) (((x) & SDMA_PKT_POLL_REGMEM_ADDR_HI_addr_63_32_mask) << SDMA_PKT_POLL_REGMEM_ADDR_HI_addr_63_32_shift) + +/*define for VALUE word*/ +/*define for value field*/ +#define SDMA_PKT_POLL_REGMEM_VALUE_value_offset 3 +#define SDMA_PKT_POLL_REGMEM_VALUE_value_mask 0xFFFFFFFF +#define SDMA_PKT_POLL_REGMEM_VALUE_value_shift 0 +#define SDMA_PKT_POLL_REGMEM_VALUE_VALUE(x) (((x) & SDMA_PKT_POLL_REGMEM_VALUE_value_mask) << SDMA_PKT_POLL_REGMEM_VALUE_value_shift) + +/*define for MASK word*/ +/*define for mask field*/ +#define SDMA_PKT_POLL_REGMEM_MASK_mask_offset 4 +#define SDMA_PKT_POLL_REGMEM_MASK_mask_mask 0xFFFFFFFF +#define SDMA_PKT_POLL_REGMEM_MASK_mask_shift 0 +#define SDMA_PKT_POLL_REGMEM_MASK_MASK(x) (((x) & SDMA_PKT_POLL_REGMEM_MASK_mask_mask) << SDMA_PKT_POLL_REGMEM_MASK_mask_shift) + +/*define for DW5 word*/ +/*define for interval field*/ +#define SDMA_PKT_POLL_REGMEM_DW5_interval_offset 5 +#define SDMA_PKT_POLL_REGMEM_DW5_interval_mask 0x0000FFFF +#define SDMA_PKT_POLL_REGMEM_DW5_interval_shift 0 +#define SDMA_PKT_POLL_REGMEM_DW5_INTERVAL(x) (((x) & SDMA_PKT_POLL_REGMEM_DW5_interval_mask) << SDMA_PKT_POLL_REGMEM_DW5_interval_shift) + +/*define for retry_count field*/ +#define SDMA_PKT_POLL_REGMEM_DW5_retry_count_offset 5 +#define SDMA_PKT_POLL_REGMEM_DW5_retry_count_mask 0x00000FFF +#define SDMA_PKT_POLL_REGMEM_DW5_retry_count_shift 16 +#define SDMA_PKT_POLL_REGMEM_DW5_RETRY_COUNT(x) (((x) & SDMA_PKT_POLL_REGMEM_DW5_retry_count_mask) << SDMA_PKT_POLL_REGMEM_DW5_retry_count_shift) + + +/* +** Definitions for SDMA_PKT_POLL_REG_WRITE_MEM packet +*/ + +/*define for HEADER word*/ +/*define for op field*/ +#define SDMA_PKT_POLL_REG_WRITE_MEM_HEADER_op_offset 0 +#define SDMA_PKT_POLL_REG_WRITE_MEM_HEADER_op_mask 0x000000FF +#define SDMA_PKT_POLL_REG_WRITE_MEM_HEADER_op_shift 0 +#define SDMA_PKT_POLL_REG_WRITE_MEM_HEADER_OP(x) (((x) & SDMA_PKT_POLL_REG_WRITE_MEM_HEADER_op_mask) << SDMA_PKT_POLL_REG_WRITE_MEM_HEADER_op_shift) + +/*define for sub_op field*/ +#define SDMA_PKT_POLL_REG_WRITE_MEM_HEADER_sub_op_offset 0 +#define SDMA_PKT_POLL_REG_WRITE_MEM_HEADER_sub_op_mask 0x000000FF +#define SDMA_PKT_POLL_REG_WRITE_MEM_HEADER_sub_op_shift 8 +#define SDMA_PKT_POLL_REG_WRITE_MEM_HEADER_SUB_OP(x) (((x) & SDMA_PKT_POLL_REG_WRITE_MEM_HEADER_sub_op_mask) << SDMA_PKT_POLL_REG_WRITE_MEM_HEADER_sub_op_shift) + +/*define for cache_policy field*/ +#define SDMA_PKT_POLL_REG_WRITE_MEM_HEADER_cache_policy_offset 0 +#define SDMA_PKT_POLL_REG_WRITE_MEM_HEADER_cache_policy_mask 0x00000007 +#define SDMA_PKT_POLL_REG_WRITE_MEM_HEADER_cache_policy_shift 24 +#define SDMA_PKT_POLL_REG_WRITE_MEM_HEADER_CACHE_POLICY(x) (((x) & SDMA_PKT_POLL_REG_WRITE_MEM_HEADER_cache_policy_mask) << SDMA_PKT_POLL_REG_WRITE_MEM_HEADER_cache_policy_shift) + +/*define for cpv field*/ +#define SDMA_PKT_POLL_REG_WRITE_MEM_HEADER_cpv_offset 0 +#define SDMA_PKT_POLL_REG_WRITE_MEM_HEADER_cpv_mask 0x00000001 +#define SDMA_PKT_POLL_REG_WRITE_MEM_HEADER_cpv_shift 28 +#define SDMA_PKT_POLL_REG_WRITE_MEM_HEADER_CPV(x) (((x) & SDMA_PKT_POLL_REG_WRITE_MEM_HEADER_cpv_mask) << SDMA_PKT_POLL_REG_WRITE_MEM_HEADER_cpv_shift) + +/*define for SRC_ADDR word*/ +/*define for addr_31_2 field*/ +#define SDMA_PKT_POLL_REG_WRITE_MEM_SRC_ADDR_addr_31_2_offset 1 +#define SDMA_PKT_POLL_REG_WRITE_MEM_SRC_ADDR_addr_31_2_mask 0x3FFFFFFF +#define SDMA_PKT_POLL_REG_WRITE_MEM_SRC_ADDR_addr_31_2_shift 2 +#define SDMA_PKT_POLL_REG_WRITE_MEM_SRC_ADDR_ADDR_31_2(x) (((x) & SDMA_PKT_POLL_REG_WRITE_MEM_SRC_ADDR_addr_31_2_mask) << SDMA_PKT_POLL_REG_WRITE_MEM_SRC_ADDR_addr_31_2_shift) + +/*define for DST_ADDR_LO word*/ +/*define for addr_31_0 field*/ +#define SDMA_PKT_POLL_REG_WRITE_MEM_DST_ADDR_LO_addr_31_0_offset 2 +#define SDMA_PKT_POLL_REG_WRITE_MEM_DST_ADDR_LO_addr_31_0_mask 0xFFFFFFFF +#define SDMA_PKT_POLL_REG_WRITE_MEM_DST_ADDR_LO_addr_31_0_shift 0 +#define SDMA_PKT_POLL_REG_WRITE_MEM_DST_ADDR_LO_ADDR_31_0(x) (((x) & SDMA_PKT_POLL_REG_WRITE_MEM_DST_ADDR_LO_addr_31_0_mask) << SDMA_PKT_POLL_REG_WRITE_MEM_DST_ADDR_LO_addr_31_0_shift) + +/*define for DST_ADDR_HI word*/ +/*define for addr_63_32 field*/ +#define SDMA_PKT_POLL_REG_WRITE_MEM_DST_ADDR_HI_addr_63_32_offset 3 +#define SDMA_PKT_POLL_REG_WRITE_MEM_DST_ADDR_HI_addr_63_32_mask 0xFFFFFFFF +#define SDMA_PKT_POLL_REG_WRITE_MEM_DST_ADDR_HI_addr_63_32_shift 0 +#define SDMA_PKT_POLL_REG_WRITE_MEM_DST_ADDR_HI_ADDR_63_32(x) (((x) & SDMA_PKT_POLL_REG_WRITE_MEM_DST_ADDR_HI_addr_63_32_mask) << SDMA_PKT_POLL_REG_WRITE_MEM_DST_ADDR_HI_addr_63_32_shift) + + +/* +** Definitions for SDMA_PKT_POLL_DBIT_WRITE_MEM packet +*/ + +/*define for HEADER word*/ +/*define for op field*/ +#define SDMA_PKT_POLL_DBIT_WRITE_MEM_HEADER_op_offset 0 +#define SDMA_PKT_POLL_DBIT_WRITE_MEM_HEADER_op_mask 0x000000FF +#define SDMA_PKT_POLL_DBIT_WRITE_MEM_HEADER_op_shift 0 +#define SDMA_PKT_POLL_DBIT_WRITE_MEM_HEADER_OP(x) (((x) & SDMA_PKT_POLL_DBIT_WRITE_MEM_HEADER_op_mask) << SDMA_PKT_POLL_DBIT_WRITE_MEM_HEADER_op_shift) + +/*define for sub_op field*/ +#define SDMA_PKT_POLL_DBIT_WRITE_MEM_HEADER_sub_op_offset 0 +#define SDMA_PKT_POLL_DBIT_WRITE_MEM_HEADER_sub_op_mask 0x000000FF +#define SDMA_PKT_POLL_DBIT_WRITE_MEM_HEADER_sub_op_shift 8 +#define SDMA_PKT_POLL_DBIT_WRITE_MEM_HEADER_SUB_OP(x) (((x) & SDMA_PKT_POLL_DBIT_WRITE_MEM_HEADER_sub_op_mask) << SDMA_PKT_POLL_DBIT_WRITE_MEM_HEADER_sub_op_shift) + +/*define for ea field*/ +#define SDMA_PKT_POLL_DBIT_WRITE_MEM_HEADER_ea_offset 0 +#define SDMA_PKT_POLL_DBIT_WRITE_MEM_HEADER_ea_mask 0x00000003 +#define SDMA_PKT_POLL_DBIT_WRITE_MEM_HEADER_ea_shift 16 +#define SDMA_PKT_POLL_DBIT_WRITE_MEM_HEADER_EA(x) (((x) & SDMA_PKT_POLL_DBIT_WRITE_MEM_HEADER_ea_mask) << SDMA_PKT_POLL_DBIT_WRITE_MEM_HEADER_ea_shift) + +/*define for cache_policy field*/ +#define SDMA_PKT_POLL_DBIT_WRITE_MEM_HEADER_cache_policy_offset 0 +#define SDMA_PKT_POLL_DBIT_WRITE_MEM_HEADER_cache_policy_mask 0x00000007 +#define SDMA_PKT_POLL_DBIT_WRITE_MEM_HEADER_cache_policy_shift 24 +#define SDMA_PKT_POLL_DBIT_WRITE_MEM_HEADER_CACHE_POLICY(x) (((x) & SDMA_PKT_POLL_DBIT_WRITE_MEM_HEADER_cache_policy_mask) << SDMA_PKT_POLL_DBIT_WRITE_MEM_HEADER_cache_policy_shift) + +/*define for cpv field*/ +#define SDMA_PKT_POLL_DBIT_WRITE_MEM_HEADER_cpv_offset 0 +#define SDMA_PKT_POLL_DBIT_WRITE_MEM_HEADER_cpv_mask 0x00000001 +#define SDMA_PKT_POLL_DBIT_WRITE_MEM_HEADER_cpv_shift 28 +#define SDMA_PKT_POLL_DBIT_WRITE_MEM_HEADER_CPV(x) (((x) & SDMA_PKT_POLL_DBIT_WRITE_MEM_HEADER_cpv_mask) << SDMA_PKT_POLL_DBIT_WRITE_MEM_HEADER_cpv_shift) + +/*define for DST_ADDR_LO word*/ +/*define for addr_31_0 field*/ +#define SDMA_PKT_POLL_DBIT_WRITE_MEM_DST_ADDR_LO_addr_31_0_offset 1 +#define SDMA_PKT_POLL_DBIT_WRITE_MEM_DST_ADDR_LO_addr_31_0_mask 0xFFFFFFFF +#define SDMA_PKT_POLL_DBIT_WRITE_MEM_DST_ADDR_LO_addr_31_0_shift 0 +#define SDMA_PKT_POLL_DBIT_WRITE_MEM_DST_ADDR_LO_ADDR_31_0(x) (((x) & SDMA_PKT_POLL_DBIT_WRITE_MEM_DST_ADDR_LO_addr_31_0_mask) << SDMA_PKT_POLL_DBIT_WRITE_MEM_DST_ADDR_LO_addr_31_0_shift) + +/*define for DST_ADDR_HI word*/ +/*define for addr_63_32 field*/ +#define SDMA_PKT_POLL_DBIT_WRITE_MEM_DST_ADDR_HI_addr_63_32_offset 2 +#define SDMA_PKT_POLL_DBIT_WRITE_MEM_DST_ADDR_HI_addr_63_32_mask 0xFFFFFFFF +#define SDMA_PKT_POLL_DBIT_WRITE_MEM_DST_ADDR_HI_addr_63_32_shift 0 +#define SDMA_PKT_POLL_DBIT_WRITE_MEM_DST_ADDR_HI_ADDR_63_32(x) (((x) & SDMA_PKT_POLL_DBIT_WRITE_MEM_DST_ADDR_HI_addr_63_32_mask) << SDMA_PKT_POLL_DBIT_WRITE_MEM_DST_ADDR_HI_addr_63_32_shift) + +/*define for START_PAGE word*/ +/*define for addr_31_4 field*/ +#define SDMA_PKT_POLL_DBIT_WRITE_MEM_START_PAGE_addr_31_4_offset 3 +#define SDMA_PKT_POLL_DBIT_WRITE_MEM_START_PAGE_addr_31_4_mask 0x0FFFFFFF +#define SDMA_PKT_POLL_DBIT_WRITE_MEM_START_PAGE_addr_31_4_shift 4 +#define SDMA_PKT_POLL_DBIT_WRITE_MEM_START_PAGE_ADDR_31_4(x) (((x) & SDMA_PKT_POLL_DBIT_WRITE_MEM_START_PAGE_addr_31_4_mask) << SDMA_PKT_POLL_DBIT_WRITE_MEM_START_PAGE_addr_31_4_shift) + +/*define for PAGE_NUM word*/ +/*define for page_num_31_0 field*/ +#define SDMA_PKT_POLL_DBIT_WRITE_MEM_PAGE_NUM_page_num_31_0_offset 4 +#define SDMA_PKT_POLL_DBIT_WRITE_MEM_PAGE_NUM_page_num_31_0_mask 0xFFFFFFFF +#define SDMA_PKT_POLL_DBIT_WRITE_MEM_PAGE_NUM_page_num_31_0_shift 0 +#define SDMA_PKT_POLL_DBIT_WRITE_MEM_PAGE_NUM_PAGE_NUM_31_0(x) (((x) & SDMA_PKT_POLL_DBIT_WRITE_MEM_PAGE_NUM_page_num_31_0_mask) << SDMA_PKT_POLL_DBIT_WRITE_MEM_PAGE_NUM_page_num_31_0_shift) + + +/* +** Definitions for SDMA_PKT_POLL_MEM_VERIFY packet +*/ + +/*define for HEADER word*/ +/*define for op field*/ +#define SDMA_PKT_POLL_MEM_VERIFY_HEADER_op_offset 0 +#define SDMA_PKT_POLL_MEM_VERIFY_HEADER_op_mask 0x000000FF +#define SDMA_PKT_POLL_MEM_VERIFY_HEADER_op_shift 0 +#define SDMA_PKT_POLL_MEM_VERIFY_HEADER_OP(x) (((x) & SDMA_PKT_POLL_MEM_VERIFY_HEADER_op_mask) << SDMA_PKT_POLL_MEM_VERIFY_HEADER_op_shift) + +/*define for sub_op field*/ +#define SDMA_PKT_POLL_MEM_VERIFY_HEADER_sub_op_offset 0 +#define SDMA_PKT_POLL_MEM_VERIFY_HEADER_sub_op_mask 0x000000FF +#define SDMA_PKT_POLL_MEM_VERIFY_HEADER_sub_op_shift 8 +#define SDMA_PKT_POLL_MEM_VERIFY_HEADER_SUB_OP(x) (((x) & SDMA_PKT_POLL_MEM_VERIFY_HEADER_sub_op_mask) << SDMA_PKT_POLL_MEM_VERIFY_HEADER_sub_op_shift) + +/*define for cache_policy field*/ +#define SDMA_PKT_POLL_MEM_VERIFY_HEADER_cache_policy_offset 0 +#define SDMA_PKT_POLL_MEM_VERIFY_HEADER_cache_policy_mask 0x00000007 +#define SDMA_PKT_POLL_MEM_VERIFY_HEADER_cache_policy_shift 24 +#define SDMA_PKT_POLL_MEM_VERIFY_HEADER_CACHE_POLICY(x) (((x) & SDMA_PKT_POLL_MEM_VERIFY_HEADER_cache_policy_mask) << SDMA_PKT_POLL_MEM_VERIFY_HEADER_cache_policy_shift) + +/*define for cpv field*/ +#define SDMA_PKT_POLL_MEM_VERIFY_HEADER_cpv_offset 0 +#define SDMA_PKT_POLL_MEM_VERIFY_HEADER_cpv_mask 0x00000001 +#define SDMA_PKT_POLL_MEM_VERIFY_HEADER_cpv_shift 28 +#define SDMA_PKT_POLL_MEM_VERIFY_HEADER_CPV(x) (((x) & SDMA_PKT_POLL_MEM_VERIFY_HEADER_cpv_mask) << SDMA_PKT_POLL_MEM_VERIFY_HEADER_cpv_shift) + +/*define for mode field*/ +#define SDMA_PKT_POLL_MEM_VERIFY_HEADER_mode_offset 0 +#define SDMA_PKT_POLL_MEM_VERIFY_HEADER_mode_mask 0x00000001 +#define SDMA_PKT_POLL_MEM_VERIFY_HEADER_mode_shift 31 +#define SDMA_PKT_POLL_MEM_VERIFY_HEADER_MODE(x) (((x) & SDMA_PKT_POLL_MEM_VERIFY_HEADER_mode_mask) << SDMA_PKT_POLL_MEM_VERIFY_HEADER_mode_shift) + +/*define for PATTERN word*/ +/*define for pattern field*/ +#define SDMA_PKT_POLL_MEM_VERIFY_PATTERN_pattern_offset 1 +#define SDMA_PKT_POLL_MEM_VERIFY_PATTERN_pattern_mask 0xFFFFFFFF +#define SDMA_PKT_POLL_MEM_VERIFY_PATTERN_pattern_shift 0 +#define SDMA_PKT_POLL_MEM_VERIFY_PATTERN_PATTERN(x) (((x) & SDMA_PKT_POLL_MEM_VERIFY_PATTERN_pattern_mask) << SDMA_PKT_POLL_MEM_VERIFY_PATTERN_pattern_shift) + +/*define for CMP0_ADDR_START_LO word*/ +/*define for cmp0_start_31_0 field*/ +#define SDMA_PKT_POLL_MEM_VERIFY_CMP0_ADDR_START_LO_cmp0_start_31_0_offset 2 +#define SDMA_PKT_POLL_MEM_VERIFY_CMP0_ADDR_START_LO_cmp0_start_31_0_mask 0xFFFFFFFF +#define SDMA_PKT_POLL_MEM_VERIFY_CMP0_ADDR_START_LO_cmp0_start_31_0_shift 0 +#define SDMA_PKT_POLL_MEM_VERIFY_CMP0_ADDR_START_LO_CMP0_START_31_0(x) (((x) & SDMA_PKT_POLL_MEM_VERIFY_CMP0_ADDR_START_LO_cmp0_start_31_0_mask) << SDMA_PKT_POLL_MEM_VERIFY_CMP0_ADDR_START_LO_cmp0_start_31_0_shift) + +/*define for CMP0_ADDR_START_HI word*/ +/*define for cmp0_start_63_32 field*/ +#define SDMA_PKT_POLL_MEM_VERIFY_CMP0_ADDR_START_HI_cmp0_start_63_32_offset 3 +#define SDMA_PKT_POLL_MEM_VERIFY_CMP0_ADDR_START_HI_cmp0_start_63_32_mask 0xFFFFFFFF +#define SDMA_PKT_POLL_MEM_VERIFY_CMP0_ADDR_START_HI_cmp0_start_63_32_shift 0 +#define SDMA_PKT_POLL_MEM_VERIFY_CMP0_ADDR_START_HI_CMP0_START_63_32(x) (((x) & SDMA_PKT_POLL_MEM_VERIFY_CMP0_ADDR_START_HI_cmp0_start_63_32_mask) << SDMA_PKT_POLL_MEM_VERIFY_CMP0_ADDR_START_HI_cmp0_start_63_32_shift) + +/*define for CMP0_ADDR_END_LO word*/ +/*define for cmp0_end_31_0 field*/ +#define SDMA_PKT_POLL_MEM_VERIFY_CMP0_ADDR_END_LO_cmp0_end_31_0_offset 4 +#define SDMA_PKT_POLL_MEM_VERIFY_CMP0_ADDR_END_LO_cmp0_end_31_0_mask 0xFFFFFFFF +#define SDMA_PKT_POLL_MEM_VERIFY_CMP0_ADDR_END_LO_cmp0_end_31_0_shift 0 +#define SDMA_PKT_POLL_MEM_VERIFY_CMP0_ADDR_END_LO_CMP0_END_31_0(x) (((x) & SDMA_PKT_POLL_MEM_VERIFY_CMP0_ADDR_END_LO_cmp0_end_31_0_mask) << SDMA_PKT_POLL_MEM_VERIFY_CMP0_ADDR_END_LO_cmp0_end_31_0_shift) + +/*define for CMP0_ADDR_END_HI word*/ +/*define for cmp0_end_63_32 field*/ +#define SDMA_PKT_POLL_MEM_VERIFY_CMP0_ADDR_END_HI_cmp0_end_63_32_offset 5 +#define SDMA_PKT_POLL_MEM_VERIFY_CMP0_ADDR_END_HI_cmp0_end_63_32_mask 0xFFFFFFFF +#define SDMA_PKT_POLL_MEM_VERIFY_CMP0_ADDR_END_HI_cmp0_end_63_32_shift 0 +#define SDMA_PKT_POLL_MEM_VERIFY_CMP0_ADDR_END_HI_CMP0_END_63_32(x) (((x) & SDMA_PKT_POLL_MEM_VERIFY_CMP0_ADDR_END_HI_cmp0_end_63_32_mask) << SDMA_PKT_POLL_MEM_VERIFY_CMP0_ADDR_END_HI_cmp0_end_63_32_shift) + +/*define for CMP1_ADDR_START_LO word*/ +/*define for cmp1_start_31_0 field*/ +#define SDMA_PKT_POLL_MEM_VERIFY_CMP1_ADDR_START_LO_cmp1_start_31_0_offset 6 +#define SDMA_PKT_POLL_MEM_VERIFY_CMP1_ADDR_START_LO_cmp1_start_31_0_mask 0xFFFFFFFF +#define SDMA_PKT_POLL_MEM_VERIFY_CMP1_ADDR_START_LO_cmp1_start_31_0_shift 0 +#define SDMA_PKT_POLL_MEM_VERIFY_CMP1_ADDR_START_LO_CMP1_START_31_0(x) (((x) & SDMA_PKT_POLL_MEM_VERIFY_CMP1_ADDR_START_LO_cmp1_start_31_0_mask) << SDMA_PKT_POLL_MEM_VERIFY_CMP1_ADDR_START_LO_cmp1_start_31_0_shift) + +/*define for CMP1_ADDR_START_HI word*/ +/*define for cmp1_start_63_32 field*/ +#define SDMA_PKT_POLL_MEM_VERIFY_CMP1_ADDR_START_HI_cmp1_start_63_32_offset 7 +#define SDMA_PKT_POLL_MEM_VERIFY_CMP1_ADDR_START_HI_cmp1_start_63_32_mask 0xFFFFFFFF +#define SDMA_PKT_POLL_MEM_VERIFY_CMP1_ADDR_START_HI_cmp1_start_63_32_shift 0 +#define SDMA_PKT_POLL_MEM_VERIFY_CMP1_ADDR_START_HI_CMP1_START_63_32(x) (((x) & SDMA_PKT_POLL_MEM_VERIFY_CMP1_ADDR_START_HI_cmp1_start_63_32_mask) << SDMA_PKT_POLL_MEM_VERIFY_CMP1_ADDR_START_HI_cmp1_start_63_32_shift) + +/*define for CMP1_ADDR_END_LO word*/ +/*define for cmp1_end_31_0 field*/ +#define SDMA_PKT_POLL_MEM_VERIFY_CMP1_ADDR_END_LO_cmp1_end_31_0_offset 8 +#define SDMA_PKT_POLL_MEM_VERIFY_CMP1_ADDR_END_LO_cmp1_end_31_0_mask 0xFFFFFFFF +#define SDMA_PKT_POLL_MEM_VERIFY_CMP1_ADDR_END_LO_cmp1_end_31_0_shift 0 +#define SDMA_PKT_POLL_MEM_VERIFY_CMP1_ADDR_END_LO_CMP1_END_31_0(x) (((x) & SDMA_PKT_POLL_MEM_VERIFY_CMP1_ADDR_END_LO_cmp1_end_31_0_mask) << SDMA_PKT_POLL_MEM_VERIFY_CMP1_ADDR_END_LO_cmp1_end_31_0_shift) + +/*define for CMP1_ADDR_END_HI word*/ +/*define for cmp1_end_63_32 field*/ +#define SDMA_PKT_POLL_MEM_VERIFY_CMP1_ADDR_END_HI_cmp1_end_63_32_offset 9 +#define SDMA_PKT_POLL_MEM_VERIFY_CMP1_ADDR_END_HI_cmp1_end_63_32_mask 0xFFFFFFFF +#define SDMA_PKT_POLL_MEM_VERIFY_CMP1_ADDR_END_HI_cmp1_end_63_32_shift 0 +#define SDMA_PKT_POLL_MEM_VERIFY_CMP1_ADDR_END_HI_CMP1_END_63_32(x) (((x) & SDMA_PKT_POLL_MEM_VERIFY_CMP1_ADDR_END_HI_cmp1_end_63_32_mask) << SDMA_PKT_POLL_MEM_VERIFY_CMP1_ADDR_END_HI_cmp1_end_63_32_shift) + +/*define for REC_ADDR_LO word*/ +/*define for rec_31_0 field*/ +#define SDMA_PKT_POLL_MEM_VERIFY_REC_ADDR_LO_rec_31_0_offset 10 +#define SDMA_PKT_POLL_MEM_VERIFY_REC_ADDR_LO_rec_31_0_mask 0xFFFFFFFF +#define SDMA_PKT_POLL_MEM_VERIFY_REC_ADDR_LO_rec_31_0_shift 0 +#define SDMA_PKT_POLL_MEM_VERIFY_REC_ADDR_LO_REC_31_0(x) (((x) & SDMA_PKT_POLL_MEM_VERIFY_REC_ADDR_LO_rec_31_0_mask) << SDMA_PKT_POLL_MEM_VERIFY_REC_ADDR_LO_rec_31_0_shift) + +/*define for REC_ADDR_HI word*/ +/*define for rec_63_32 field*/ +#define SDMA_PKT_POLL_MEM_VERIFY_REC_ADDR_HI_rec_63_32_offset 11 +#define SDMA_PKT_POLL_MEM_VERIFY_REC_ADDR_HI_rec_63_32_mask 0xFFFFFFFF +#define SDMA_PKT_POLL_MEM_VERIFY_REC_ADDR_HI_rec_63_32_shift 0 +#define SDMA_PKT_POLL_MEM_VERIFY_REC_ADDR_HI_REC_63_32(x) (((x) & SDMA_PKT_POLL_MEM_VERIFY_REC_ADDR_HI_rec_63_32_mask) << SDMA_PKT_POLL_MEM_VERIFY_REC_ADDR_HI_rec_63_32_shift) + +/*define for RESERVED word*/ +/*define for reserved field*/ +#define SDMA_PKT_POLL_MEM_VERIFY_RESERVED_reserved_offset 12 +#define SDMA_PKT_POLL_MEM_VERIFY_RESERVED_reserved_mask 0xFFFFFFFF +#define SDMA_PKT_POLL_MEM_VERIFY_RESERVED_reserved_shift 0 +#define SDMA_PKT_POLL_MEM_VERIFY_RESERVED_RESERVED(x) (((x) & SDMA_PKT_POLL_MEM_VERIFY_RESERVED_reserved_mask) << SDMA_PKT_POLL_MEM_VERIFY_RESERVED_reserved_shift) + + +/* +** Definitions for SDMA_PKT_ATOMIC packet +*/ + +/*define for HEADER word*/ +/*define for op field*/ +#define SDMA_PKT_ATOMIC_HEADER_op_offset 0 +#define SDMA_PKT_ATOMIC_HEADER_op_mask 0x000000FF +#define SDMA_PKT_ATOMIC_HEADER_op_shift 0 +#define SDMA_PKT_ATOMIC_HEADER_OP(x) (((x) & SDMA_PKT_ATOMIC_HEADER_op_mask) << SDMA_PKT_ATOMIC_HEADER_op_shift) + +/*define for loop field*/ +#define SDMA_PKT_ATOMIC_HEADER_loop_offset 0 +#define SDMA_PKT_ATOMIC_HEADER_loop_mask 0x00000001 +#define SDMA_PKT_ATOMIC_HEADER_loop_shift 16 +#define SDMA_PKT_ATOMIC_HEADER_LOOP(x) (((x) & SDMA_PKT_ATOMIC_HEADER_loop_mask) << SDMA_PKT_ATOMIC_HEADER_loop_shift) + +/*define for tmz field*/ +#define SDMA_PKT_ATOMIC_HEADER_tmz_offset 0 +#define SDMA_PKT_ATOMIC_HEADER_tmz_mask 0x00000001 +#define SDMA_PKT_ATOMIC_HEADER_tmz_shift 18 +#define SDMA_PKT_ATOMIC_HEADER_TMZ(x) (((x) & SDMA_PKT_ATOMIC_HEADER_tmz_mask) << SDMA_PKT_ATOMIC_HEADER_tmz_shift) + +/*define for cache_policy field*/ +#define SDMA_PKT_ATOMIC_HEADER_cache_policy_offset 0 +#define SDMA_PKT_ATOMIC_HEADER_cache_policy_mask 0x00000007 +#define SDMA_PKT_ATOMIC_HEADER_cache_policy_shift 20 +#define SDMA_PKT_ATOMIC_HEADER_CACHE_POLICY(x) (((x) & SDMA_PKT_ATOMIC_HEADER_cache_policy_mask) << SDMA_PKT_ATOMIC_HEADER_cache_policy_shift) + +/*define for cpv field*/ +#define SDMA_PKT_ATOMIC_HEADER_cpv_offset 0 +#define SDMA_PKT_ATOMIC_HEADER_cpv_mask 0x00000001 +#define SDMA_PKT_ATOMIC_HEADER_cpv_shift 24 +#define SDMA_PKT_ATOMIC_HEADER_CPV(x) (((x) & SDMA_PKT_ATOMIC_HEADER_cpv_mask) << SDMA_PKT_ATOMIC_HEADER_cpv_shift) + +/*define for atomic_op field*/ +#define SDMA_PKT_ATOMIC_HEADER_atomic_op_offset 0 +#define SDMA_PKT_ATOMIC_HEADER_atomic_op_mask 0x0000007F +#define SDMA_PKT_ATOMIC_HEADER_atomic_op_shift 25 +#define SDMA_PKT_ATOMIC_HEADER_ATOMIC_OP(x) (((x) & SDMA_PKT_ATOMIC_HEADER_atomic_op_mask) << SDMA_PKT_ATOMIC_HEADER_atomic_op_shift) + +/*define for ADDR_LO word*/ +/*define for addr_31_0 field*/ +#define SDMA_PKT_ATOMIC_ADDR_LO_addr_31_0_offset 1 +#define SDMA_PKT_ATOMIC_ADDR_LO_addr_31_0_mask 0xFFFFFFFF +#define SDMA_PKT_ATOMIC_ADDR_LO_addr_31_0_shift 0 +#define SDMA_PKT_ATOMIC_ADDR_LO_ADDR_31_0(x) (((x) & SDMA_PKT_ATOMIC_ADDR_LO_addr_31_0_mask) << SDMA_PKT_ATOMIC_ADDR_LO_addr_31_0_shift) + +/*define for ADDR_HI word*/ +/*define for addr_63_32 field*/ +#define SDMA_PKT_ATOMIC_ADDR_HI_addr_63_32_offset 2 +#define SDMA_PKT_ATOMIC_ADDR_HI_addr_63_32_mask 0xFFFFFFFF +#define SDMA_PKT_ATOMIC_ADDR_HI_addr_63_32_shift 0 +#define SDMA_PKT_ATOMIC_ADDR_HI_ADDR_63_32(x) (((x) & SDMA_PKT_ATOMIC_ADDR_HI_addr_63_32_mask) << SDMA_PKT_ATOMIC_ADDR_HI_addr_63_32_shift) + +/*define for SRC_DATA_LO word*/ +/*define for src_data_31_0 field*/ +#define SDMA_PKT_ATOMIC_SRC_DATA_LO_src_data_31_0_offset 3 +#define SDMA_PKT_ATOMIC_SRC_DATA_LO_src_data_31_0_mask 0xFFFFFFFF +#define SDMA_PKT_ATOMIC_SRC_DATA_LO_src_data_31_0_shift 0 +#define SDMA_PKT_ATOMIC_SRC_DATA_LO_SRC_DATA_31_0(x) (((x) & SDMA_PKT_ATOMIC_SRC_DATA_LO_src_data_31_0_mask) << SDMA_PKT_ATOMIC_SRC_DATA_LO_src_data_31_0_shift) + +/*define for SRC_DATA_HI word*/ +/*define for src_data_63_32 field*/ +#define SDMA_PKT_ATOMIC_SRC_DATA_HI_src_data_63_32_offset 4 +#define SDMA_PKT_ATOMIC_SRC_DATA_HI_src_data_63_32_mask 0xFFFFFFFF +#define SDMA_PKT_ATOMIC_SRC_DATA_HI_src_data_63_32_shift 0 +#define SDMA_PKT_ATOMIC_SRC_DATA_HI_SRC_DATA_63_32(x) (((x) & SDMA_PKT_ATOMIC_SRC_DATA_HI_src_data_63_32_mask) << SDMA_PKT_ATOMIC_SRC_DATA_HI_src_data_63_32_shift) + +/*define for CMP_DATA_LO word*/ +/*define for cmp_data_31_0 field*/ +#define SDMA_PKT_ATOMIC_CMP_DATA_LO_cmp_data_31_0_offset 5 +#define SDMA_PKT_ATOMIC_CMP_DATA_LO_cmp_data_31_0_mask 0xFFFFFFFF +#define SDMA_PKT_ATOMIC_CMP_DATA_LO_cmp_data_31_0_shift 0 +#define SDMA_PKT_ATOMIC_CMP_DATA_LO_CMP_DATA_31_0(x) (((x) & SDMA_PKT_ATOMIC_CMP_DATA_LO_cmp_data_31_0_mask) << SDMA_PKT_ATOMIC_CMP_DATA_LO_cmp_data_31_0_shift) + +/*define for CMP_DATA_HI word*/ +/*define for cmp_data_63_32 field*/ +#define SDMA_PKT_ATOMIC_CMP_DATA_HI_cmp_data_63_32_offset 6 +#define SDMA_PKT_ATOMIC_CMP_DATA_HI_cmp_data_63_32_mask 0xFFFFFFFF +#define SDMA_PKT_ATOMIC_CMP_DATA_HI_cmp_data_63_32_shift 0 +#define SDMA_PKT_ATOMIC_CMP_DATA_HI_CMP_DATA_63_32(x) (((x) & SDMA_PKT_ATOMIC_CMP_DATA_HI_cmp_data_63_32_mask) << SDMA_PKT_ATOMIC_CMP_DATA_HI_cmp_data_63_32_shift) + +/*define for LOOP_INTERVAL word*/ +/*define for loop_interval field*/ +#define SDMA_PKT_ATOMIC_LOOP_INTERVAL_loop_interval_offset 7 +#define SDMA_PKT_ATOMIC_LOOP_INTERVAL_loop_interval_mask 0x00001FFF +#define SDMA_PKT_ATOMIC_LOOP_INTERVAL_loop_interval_shift 0 +#define SDMA_PKT_ATOMIC_LOOP_INTERVAL_LOOP_INTERVAL(x) (((x) & SDMA_PKT_ATOMIC_LOOP_INTERVAL_loop_interval_mask) << SDMA_PKT_ATOMIC_LOOP_INTERVAL_loop_interval_shift) + + +/* +** Definitions for SDMA_PKT_TIMESTAMP_SET packet +*/ + +/*define for HEADER word*/ +/*define for op field*/ +#define SDMA_PKT_TIMESTAMP_SET_HEADER_op_offset 0 +#define SDMA_PKT_TIMESTAMP_SET_HEADER_op_mask 0x000000FF +#define SDMA_PKT_TIMESTAMP_SET_HEADER_op_shift 0 +#define SDMA_PKT_TIMESTAMP_SET_HEADER_OP(x) (((x) & SDMA_PKT_TIMESTAMP_SET_HEADER_op_mask) << SDMA_PKT_TIMESTAMP_SET_HEADER_op_shift) + +/*define for sub_op field*/ +#define SDMA_PKT_TIMESTAMP_SET_HEADER_sub_op_offset 0 +#define SDMA_PKT_TIMESTAMP_SET_HEADER_sub_op_mask 0x000000FF +#define SDMA_PKT_TIMESTAMP_SET_HEADER_sub_op_shift 8 +#define SDMA_PKT_TIMESTAMP_SET_HEADER_SUB_OP(x) (((x) & SDMA_PKT_TIMESTAMP_SET_HEADER_sub_op_mask) << SDMA_PKT_TIMESTAMP_SET_HEADER_sub_op_shift) + +/*define for INIT_DATA_LO word*/ +/*define for init_data_31_0 field*/ +#define SDMA_PKT_TIMESTAMP_SET_INIT_DATA_LO_init_data_31_0_offset 1 +#define SDMA_PKT_TIMESTAMP_SET_INIT_DATA_LO_init_data_31_0_mask 0xFFFFFFFF +#define SDMA_PKT_TIMESTAMP_SET_INIT_DATA_LO_init_data_31_0_shift 0 +#define SDMA_PKT_TIMESTAMP_SET_INIT_DATA_LO_INIT_DATA_31_0(x) (((x) & SDMA_PKT_TIMESTAMP_SET_INIT_DATA_LO_init_data_31_0_mask) << SDMA_PKT_TIMESTAMP_SET_INIT_DATA_LO_init_data_31_0_shift) + +/*define for INIT_DATA_HI word*/ +/*define for init_data_63_32 field*/ +#define SDMA_PKT_TIMESTAMP_SET_INIT_DATA_HI_init_data_63_32_offset 2 +#define SDMA_PKT_TIMESTAMP_SET_INIT_DATA_HI_init_data_63_32_mask 0xFFFFFFFF +#define SDMA_PKT_TIMESTAMP_SET_INIT_DATA_HI_init_data_63_32_shift 0 +#define SDMA_PKT_TIMESTAMP_SET_INIT_DATA_HI_INIT_DATA_63_32(x) (((x) & SDMA_PKT_TIMESTAMP_SET_INIT_DATA_HI_init_data_63_32_mask) << SDMA_PKT_TIMESTAMP_SET_INIT_DATA_HI_init_data_63_32_shift) + + +/* +** Definitions for SDMA_PKT_TIMESTAMP_GET packet +*/ + +/*define for HEADER word*/ +/*define for op field*/ +#define SDMA_PKT_TIMESTAMP_GET_HEADER_op_offset 0 +#define SDMA_PKT_TIMESTAMP_GET_HEADER_op_mask 0x000000FF +#define SDMA_PKT_TIMESTAMP_GET_HEADER_op_shift 0 +#define SDMA_PKT_TIMESTAMP_GET_HEADER_OP(x) (((x) & SDMA_PKT_TIMESTAMP_GET_HEADER_op_mask) << SDMA_PKT_TIMESTAMP_GET_HEADER_op_shift) + +/*define for sub_op field*/ +#define SDMA_PKT_TIMESTAMP_GET_HEADER_sub_op_offset 0 +#define SDMA_PKT_TIMESTAMP_GET_HEADER_sub_op_mask 0x000000FF +#define SDMA_PKT_TIMESTAMP_GET_HEADER_sub_op_shift 8 +#define SDMA_PKT_TIMESTAMP_GET_HEADER_SUB_OP(x) (((x) & SDMA_PKT_TIMESTAMP_GET_HEADER_sub_op_mask) << SDMA_PKT_TIMESTAMP_GET_HEADER_sub_op_shift) + +/*define for l2_policy field*/ +#define SDMA_PKT_TIMESTAMP_GET_HEADER_l2_policy_offset 0 +#define SDMA_PKT_TIMESTAMP_GET_HEADER_l2_policy_mask 0x00000003 +#define SDMA_PKT_TIMESTAMP_GET_HEADER_l2_policy_shift 24 +#define SDMA_PKT_TIMESTAMP_GET_HEADER_L2_POLICY(x) (((x) & SDMA_PKT_TIMESTAMP_GET_HEADER_l2_policy_mask) << SDMA_PKT_TIMESTAMP_GET_HEADER_l2_policy_shift) + +/*define for llc_policy field*/ +#define SDMA_PKT_TIMESTAMP_GET_HEADER_llc_policy_offset 0 +#define SDMA_PKT_TIMESTAMP_GET_HEADER_llc_policy_mask 0x00000001 +#define SDMA_PKT_TIMESTAMP_GET_HEADER_llc_policy_shift 26 +#define SDMA_PKT_TIMESTAMP_GET_HEADER_LLC_POLICY(x) (((x) & SDMA_PKT_TIMESTAMP_GET_HEADER_llc_policy_mask) << SDMA_PKT_TIMESTAMP_GET_HEADER_llc_policy_shift) + +/*define for cpv field*/ +#define SDMA_PKT_TIMESTAMP_GET_HEADER_cpv_offset 0 +#define SDMA_PKT_TIMESTAMP_GET_HEADER_cpv_mask 0x00000001 +#define SDMA_PKT_TIMESTAMP_GET_HEADER_cpv_shift 28 +#define SDMA_PKT_TIMESTAMP_GET_HEADER_CPV(x) (((x) & SDMA_PKT_TIMESTAMP_GET_HEADER_cpv_mask) << SDMA_PKT_TIMESTAMP_GET_HEADER_cpv_shift) + +/*define for WRITE_ADDR_LO word*/ +/*define for write_addr_31_3 field*/ +#define SDMA_PKT_TIMESTAMP_GET_WRITE_ADDR_LO_write_addr_31_3_offset 1 +#define SDMA_PKT_TIMESTAMP_GET_WRITE_ADDR_LO_write_addr_31_3_mask 0x1FFFFFFF +#define SDMA_PKT_TIMESTAMP_GET_WRITE_ADDR_LO_write_addr_31_3_shift 3 +#define SDMA_PKT_TIMESTAMP_GET_WRITE_ADDR_LO_WRITE_ADDR_31_3(x) (((x) & SDMA_PKT_TIMESTAMP_GET_WRITE_ADDR_LO_write_addr_31_3_mask) << SDMA_PKT_TIMESTAMP_GET_WRITE_ADDR_LO_write_addr_31_3_shift) + +/*define for WRITE_ADDR_HI word*/ +/*define for write_addr_63_32 field*/ +#define SDMA_PKT_TIMESTAMP_GET_WRITE_ADDR_HI_write_addr_63_32_offset 2 +#define SDMA_PKT_TIMESTAMP_GET_WRITE_ADDR_HI_write_addr_63_32_mask 0xFFFFFFFF +#define SDMA_PKT_TIMESTAMP_GET_WRITE_ADDR_HI_write_addr_63_32_shift 0 +#define SDMA_PKT_TIMESTAMP_GET_WRITE_ADDR_HI_WRITE_ADDR_63_32(x) (((x) & SDMA_PKT_TIMESTAMP_GET_WRITE_ADDR_HI_write_addr_63_32_mask) << SDMA_PKT_TIMESTAMP_GET_WRITE_ADDR_HI_write_addr_63_32_shift) + + +/* +** Definitions for SDMA_PKT_TIMESTAMP_GET_GLOBAL packet +*/ + +/*define for HEADER word*/ +/*define for op field*/ +#define SDMA_PKT_TIMESTAMP_GET_GLOBAL_HEADER_op_offset 0 +#define SDMA_PKT_TIMESTAMP_GET_GLOBAL_HEADER_op_mask 0x000000FF +#define SDMA_PKT_TIMESTAMP_GET_GLOBAL_HEADER_op_shift 0 +#define SDMA_PKT_TIMESTAMP_GET_GLOBAL_HEADER_OP(x) (((x) & SDMA_PKT_TIMESTAMP_GET_GLOBAL_HEADER_op_mask) << SDMA_PKT_TIMESTAMP_GET_GLOBAL_HEADER_op_shift) + +/*define for sub_op field*/ +#define SDMA_PKT_TIMESTAMP_GET_GLOBAL_HEADER_sub_op_offset 0 +#define SDMA_PKT_TIMESTAMP_GET_GLOBAL_HEADER_sub_op_mask 0x000000FF +#define SDMA_PKT_TIMESTAMP_GET_GLOBAL_HEADER_sub_op_shift 8 +#define SDMA_PKT_TIMESTAMP_GET_GLOBAL_HEADER_SUB_OP(x) (((x) & SDMA_PKT_TIMESTAMP_GET_GLOBAL_HEADER_sub_op_mask) << SDMA_PKT_TIMESTAMP_GET_GLOBAL_HEADER_sub_op_shift) + +/*define for l2_policy field*/ +#define SDMA_PKT_TIMESTAMP_GET_GLOBAL_HEADER_l2_policy_offset 0 +#define SDMA_PKT_TIMESTAMP_GET_GLOBAL_HEADER_l2_policy_mask 0x00000003 +#define SDMA_PKT_TIMESTAMP_GET_GLOBAL_HEADER_l2_policy_shift 24 +#define SDMA_PKT_TIMESTAMP_GET_GLOBAL_HEADER_L2_POLICY(x) (((x) & SDMA_PKT_TIMESTAMP_GET_GLOBAL_HEADER_l2_policy_mask) << SDMA_PKT_TIMESTAMP_GET_GLOBAL_HEADER_l2_policy_shift) + +/*define for llc_policy field*/ +#define SDMA_PKT_TIMESTAMP_GET_GLOBAL_HEADER_llc_policy_offset 0 +#define SDMA_PKT_TIMESTAMP_GET_GLOBAL_HEADER_llc_policy_mask 0x00000001 +#define SDMA_PKT_TIMESTAMP_GET_GLOBAL_HEADER_llc_policy_shift 26 +#define SDMA_PKT_TIMESTAMP_GET_GLOBAL_HEADER_LLC_POLICY(x) (((x) & SDMA_PKT_TIMESTAMP_GET_GLOBAL_HEADER_llc_policy_mask) << SDMA_PKT_TIMESTAMP_GET_GLOBAL_HEADER_llc_policy_shift) + +/*define for cpv field*/ +#define SDMA_PKT_TIMESTAMP_GET_GLOBAL_HEADER_cpv_offset 0 +#define SDMA_PKT_TIMESTAMP_GET_GLOBAL_HEADER_cpv_mask 0x00000001 +#define SDMA_PKT_TIMESTAMP_GET_GLOBAL_HEADER_cpv_shift 28 +#define SDMA_PKT_TIMESTAMP_GET_GLOBAL_HEADER_CPV(x) (((x) & SDMA_PKT_TIMESTAMP_GET_GLOBAL_HEADER_cpv_mask) << SDMA_PKT_TIMESTAMP_GET_GLOBAL_HEADER_cpv_shift) + +/*define for WRITE_ADDR_LO word*/ +/*define for write_addr_31_3 field*/ +#define SDMA_PKT_TIMESTAMP_GET_GLOBAL_WRITE_ADDR_LO_write_addr_31_3_offset 1 +#define SDMA_PKT_TIMESTAMP_GET_GLOBAL_WRITE_ADDR_LO_write_addr_31_3_mask 0x1FFFFFFF +#define SDMA_PKT_TIMESTAMP_GET_GLOBAL_WRITE_ADDR_LO_write_addr_31_3_shift 3 +#define SDMA_PKT_TIMESTAMP_GET_GLOBAL_WRITE_ADDR_LO_WRITE_ADDR_31_3(x) (((x) & SDMA_PKT_TIMESTAMP_GET_GLOBAL_WRITE_ADDR_LO_write_addr_31_3_mask) << SDMA_PKT_TIMESTAMP_GET_GLOBAL_WRITE_ADDR_LO_write_addr_31_3_shift) + +/*define for WRITE_ADDR_HI word*/ +/*define for write_addr_63_32 field*/ +#define SDMA_PKT_TIMESTAMP_GET_GLOBAL_WRITE_ADDR_HI_write_addr_63_32_offset 2 +#define SDMA_PKT_TIMESTAMP_GET_GLOBAL_WRITE_ADDR_HI_write_addr_63_32_mask 0xFFFFFFFF +#define SDMA_PKT_TIMESTAMP_GET_GLOBAL_WRITE_ADDR_HI_write_addr_63_32_shift 0 +#define SDMA_PKT_TIMESTAMP_GET_GLOBAL_WRITE_ADDR_HI_WRITE_ADDR_63_32(x) (((x) & SDMA_PKT_TIMESTAMP_GET_GLOBAL_WRITE_ADDR_HI_write_addr_63_32_mask) << SDMA_PKT_TIMESTAMP_GET_GLOBAL_WRITE_ADDR_HI_write_addr_63_32_shift) + + +/* +** Definitions for SDMA_PKT_TRAP packet +*/ + +/*define for HEADER word*/ +/*define for op field*/ +#define SDMA_PKT_TRAP_HEADER_op_offset 0 +#define SDMA_PKT_TRAP_HEADER_op_mask 0x000000FF +#define SDMA_PKT_TRAP_HEADER_op_shift 0 +#define SDMA_PKT_TRAP_HEADER_OP(x) (((x) & SDMA_PKT_TRAP_HEADER_op_mask) << SDMA_PKT_TRAP_HEADER_op_shift) + +/*define for sub_op field*/ +#define SDMA_PKT_TRAP_HEADER_sub_op_offset 0 +#define SDMA_PKT_TRAP_HEADER_sub_op_mask 0x000000FF +#define SDMA_PKT_TRAP_HEADER_sub_op_shift 8 +#define SDMA_PKT_TRAP_HEADER_SUB_OP(x) (((x) & SDMA_PKT_TRAP_HEADER_sub_op_mask) << SDMA_PKT_TRAP_HEADER_sub_op_shift) + +/*define for INT_CONTEXT word*/ +/*define for int_context field*/ +#define SDMA_PKT_TRAP_INT_CONTEXT_int_context_offset 1 +#define SDMA_PKT_TRAP_INT_CONTEXT_int_context_mask 0x0FFFFFFF +#define SDMA_PKT_TRAP_INT_CONTEXT_int_context_shift 0 +#define SDMA_PKT_TRAP_INT_CONTEXT_INT_CONTEXT(x) (((x) & SDMA_PKT_TRAP_INT_CONTEXT_int_context_mask) << SDMA_PKT_TRAP_INT_CONTEXT_int_context_shift) + + +/* +** Definitions for SDMA_PKT_DUMMY_TRAP packet +*/ + +/*define for HEADER word*/ +/*define for op field*/ +#define SDMA_PKT_DUMMY_TRAP_HEADER_op_offset 0 +#define SDMA_PKT_DUMMY_TRAP_HEADER_op_mask 0x000000FF +#define SDMA_PKT_DUMMY_TRAP_HEADER_op_shift 0 +#define SDMA_PKT_DUMMY_TRAP_HEADER_OP(x) (((x) & SDMA_PKT_DUMMY_TRAP_HEADER_op_mask) << SDMA_PKT_DUMMY_TRAP_HEADER_op_shift) + +/*define for sub_op field*/ +#define SDMA_PKT_DUMMY_TRAP_HEADER_sub_op_offset 0 +#define SDMA_PKT_DUMMY_TRAP_HEADER_sub_op_mask 0x000000FF +#define SDMA_PKT_DUMMY_TRAP_HEADER_sub_op_shift 8 +#define SDMA_PKT_DUMMY_TRAP_HEADER_SUB_OP(x) (((x) & SDMA_PKT_DUMMY_TRAP_HEADER_sub_op_mask) << SDMA_PKT_DUMMY_TRAP_HEADER_sub_op_shift) + +/*define for INT_CONTEXT word*/ +/*define for int_context field*/ +#define SDMA_PKT_DUMMY_TRAP_INT_CONTEXT_int_context_offset 1 +#define SDMA_PKT_DUMMY_TRAP_INT_CONTEXT_int_context_mask 0x0FFFFFFF +#define SDMA_PKT_DUMMY_TRAP_INT_CONTEXT_int_context_shift 0 +#define SDMA_PKT_DUMMY_TRAP_INT_CONTEXT_INT_CONTEXT(x) (((x) & SDMA_PKT_DUMMY_TRAP_INT_CONTEXT_int_context_mask) << SDMA_PKT_DUMMY_TRAP_INT_CONTEXT_int_context_shift) + + +/* +** Definitions for SDMA_PKT_GPUVM_INV packet +*/ + +/*define for HEADER word*/ +/*define for op field*/ +#define SDMA_PKT_GPUVM_INV_HEADER_op_offset 0 +#define SDMA_PKT_GPUVM_INV_HEADER_op_mask 0x000000FF +#define SDMA_PKT_GPUVM_INV_HEADER_op_shift 0 +#define SDMA_PKT_GPUVM_INV_HEADER_OP(x) (((x) & SDMA_PKT_GPUVM_INV_HEADER_op_mask) << SDMA_PKT_GPUVM_INV_HEADER_op_shift) + +/*define for sub_op field*/ +#define SDMA_PKT_GPUVM_INV_HEADER_sub_op_offset 0 +#define SDMA_PKT_GPUVM_INV_HEADER_sub_op_mask 0x000000FF +#define SDMA_PKT_GPUVM_INV_HEADER_sub_op_shift 8 +#define SDMA_PKT_GPUVM_INV_HEADER_SUB_OP(x) (((x) & SDMA_PKT_GPUVM_INV_HEADER_sub_op_mask) << SDMA_PKT_GPUVM_INV_HEADER_sub_op_shift) + +/*define for PAYLOAD1 word*/ +/*define for per_vmid_inv_req field*/ +#define SDMA_PKT_GPUVM_INV_PAYLOAD1_per_vmid_inv_req_offset 1 +#define SDMA_PKT_GPUVM_INV_PAYLOAD1_per_vmid_inv_req_mask 0x0000FFFF +#define SDMA_PKT_GPUVM_INV_PAYLOAD1_per_vmid_inv_req_shift 0 +#define SDMA_PKT_GPUVM_INV_PAYLOAD1_PER_VMID_INV_REQ(x) (((x) & SDMA_PKT_GPUVM_INV_PAYLOAD1_per_vmid_inv_req_mask) << SDMA_PKT_GPUVM_INV_PAYLOAD1_per_vmid_inv_req_shift) + +/*define for flush_type field*/ +#define SDMA_PKT_GPUVM_INV_PAYLOAD1_flush_type_offset 1 +#define SDMA_PKT_GPUVM_INV_PAYLOAD1_flush_type_mask 0x00000007 +#define SDMA_PKT_GPUVM_INV_PAYLOAD1_flush_type_shift 16 +#define SDMA_PKT_GPUVM_INV_PAYLOAD1_FLUSH_TYPE(x) (((x) & SDMA_PKT_GPUVM_INV_PAYLOAD1_flush_type_mask) << SDMA_PKT_GPUVM_INV_PAYLOAD1_flush_type_shift) + +/*define for l2_ptes field*/ +#define SDMA_PKT_GPUVM_INV_PAYLOAD1_l2_ptes_offset 1 +#define SDMA_PKT_GPUVM_INV_PAYLOAD1_l2_ptes_mask 0x00000001 +#define SDMA_PKT_GPUVM_INV_PAYLOAD1_l2_ptes_shift 19 +#define SDMA_PKT_GPUVM_INV_PAYLOAD1_L2_PTES(x) (((x) & SDMA_PKT_GPUVM_INV_PAYLOAD1_l2_ptes_mask) << SDMA_PKT_GPUVM_INV_PAYLOAD1_l2_ptes_shift) + +/*define for l2_pde0 field*/ +#define SDMA_PKT_GPUVM_INV_PAYLOAD1_l2_pde0_offset 1 +#define SDMA_PKT_GPUVM_INV_PAYLOAD1_l2_pde0_mask 0x00000001 +#define SDMA_PKT_GPUVM_INV_PAYLOAD1_l2_pde0_shift 20 +#define SDMA_PKT_GPUVM_INV_PAYLOAD1_L2_PDE0(x) (((x) & SDMA_PKT_GPUVM_INV_PAYLOAD1_l2_pde0_mask) << SDMA_PKT_GPUVM_INV_PAYLOAD1_l2_pde0_shift) + +/*define for l2_pde1 field*/ +#define SDMA_PKT_GPUVM_INV_PAYLOAD1_l2_pde1_offset 1 +#define SDMA_PKT_GPUVM_INV_PAYLOAD1_l2_pde1_mask 0x00000001 +#define SDMA_PKT_GPUVM_INV_PAYLOAD1_l2_pde1_shift 21 +#define SDMA_PKT_GPUVM_INV_PAYLOAD1_L2_PDE1(x) (((x) & SDMA_PKT_GPUVM_INV_PAYLOAD1_l2_pde1_mask) << SDMA_PKT_GPUVM_INV_PAYLOAD1_l2_pde1_shift) + +/*define for l2_pde2 field*/ +#define SDMA_PKT_GPUVM_INV_PAYLOAD1_l2_pde2_offset 1 +#define SDMA_PKT_GPUVM_INV_PAYLOAD1_l2_pde2_mask 0x00000001 +#define SDMA_PKT_GPUVM_INV_PAYLOAD1_l2_pde2_shift 22 +#define SDMA_PKT_GPUVM_INV_PAYLOAD1_L2_PDE2(x) (((x) & SDMA_PKT_GPUVM_INV_PAYLOAD1_l2_pde2_mask) << SDMA_PKT_GPUVM_INV_PAYLOAD1_l2_pde2_shift) + +/*define for l1_ptes field*/ +#define SDMA_PKT_GPUVM_INV_PAYLOAD1_l1_ptes_offset 1 +#define SDMA_PKT_GPUVM_INV_PAYLOAD1_l1_ptes_mask 0x00000001 +#define SDMA_PKT_GPUVM_INV_PAYLOAD1_l1_ptes_shift 23 +#define SDMA_PKT_GPUVM_INV_PAYLOAD1_L1_PTES(x) (((x) & SDMA_PKT_GPUVM_INV_PAYLOAD1_l1_ptes_mask) << SDMA_PKT_GPUVM_INV_PAYLOAD1_l1_ptes_shift) + +/*define for clr_protection_fault_status_addr field*/ +#define SDMA_PKT_GPUVM_INV_PAYLOAD1_clr_protection_fault_status_addr_offset 1 +#define SDMA_PKT_GPUVM_INV_PAYLOAD1_clr_protection_fault_status_addr_mask 0x00000001 +#define SDMA_PKT_GPUVM_INV_PAYLOAD1_clr_protection_fault_status_addr_shift 24 +#define SDMA_PKT_GPUVM_INV_PAYLOAD1_CLR_PROTECTION_FAULT_STATUS_ADDR(x) (((x) & SDMA_PKT_GPUVM_INV_PAYLOAD1_clr_protection_fault_status_addr_mask) << SDMA_PKT_GPUVM_INV_PAYLOAD1_clr_protection_fault_status_addr_shift) + +/*define for log_request field*/ +#define SDMA_PKT_GPUVM_INV_PAYLOAD1_log_request_offset 1 +#define SDMA_PKT_GPUVM_INV_PAYLOAD1_log_request_mask 0x00000001 +#define SDMA_PKT_GPUVM_INV_PAYLOAD1_log_request_shift 25 +#define SDMA_PKT_GPUVM_INV_PAYLOAD1_LOG_REQUEST(x) (((x) & SDMA_PKT_GPUVM_INV_PAYLOAD1_log_request_mask) << SDMA_PKT_GPUVM_INV_PAYLOAD1_log_request_shift) + +/*define for four_kilobytes field*/ +#define SDMA_PKT_GPUVM_INV_PAYLOAD1_four_kilobytes_offset 1 +#define SDMA_PKT_GPUVM_INV_PAYLOAD1_four_kilobytes_mask 0x00000001 +#define SDMA_PKT_GPUVM_INV_PAYLOAD1_four_kilobytes_shift 26 +#define SDMA_PKT_GPUVM_INV_PAYLOAD1_FOUR_KILOBYTES(x) (((x) & SDMA_PKT_GPUVM_INV_PAYLOAD1_four_kilobytes_mask) << SDMA_PKT_GPUVM_INV_PAYLOAD1_four_kilobytes_shift) + +/*define for PAYLOAD2 word*/ +/*define for s field*/ +#define SDMA_PKT_GPUVM_INV_PAYLOAD2_s_offset 2 +#define SDMA_PKT_GPUVM_INV_PAYLOAD2_s_mask 0x00000001 +#define SDMA_PKT_GPUVM_INV_PAYLOAD2_s_shift 0 +#define SDMA_PKT_GPUVM_INV_PAYLOAD2_S(x) (((x) & SDMA_PKT_GPUVM_INV_PAYLOAD2_s_mask) << SDMA_PKT_GPUVM_INV_PAYLOAD2_s_shift) + +/*define for page_va_42_12 field*/ +#define SDMA_PKT_GPUVM_INV_PAYLOAD2_page_va_42_12_offset 2 +#define SDMA_PKT_GPUVM_INV_PAYLOAD2_page_va_42_12_mask 0x7FFFFFFF +#define SDMA_PKT_GPUVM_INV_PAYLOAD2_page_va_42_12_shift 1 +#define SDMA_PKT_GPUVM_INV_PAYLOAD2_PAGE_VA_42_12(x) (((x) & SDMA_PKT_GPUVM_INV_PAYLOAD2_page_va_42_12_mask) << SDMA_PKT_GPUVM_INV_PAYLOAD2_page_va_42_12_shift) + +/*define for PAYLOAD3 word*/ +/*define for page_va_47_43 field*/ +#define SDMA_PKT_GPUVM_INV_PAYLOAD3_page_va_47_43_offset 3 +#define SDMA_PKT_GPUVM_INV_PAYLOAD3_page_va_47_43_mask 0x0000003F +#define SDMA_PKT_GPUVM_INV_PAYLOAD3_page_va_47_43_shift 0 +#define SDMA_PKT_GPUVM_INV_PAYLOAD3_PAGE_VA_47_43(x) (((x) & SDMA_PKT_GPUVM_INV_PAYLOAD3_page_va_47_43_mask) << SDMA_PKT_GPUVM_INV_PAYLOAD3_page_va_47_43_shift) + + +/* +** Definitions for SDMA_PKT_GCR_REQ packet +*/ + +/*define for HEADER word*/ +/*define for op field*/ +#define SDMA_PKT_GCR_REQ_HEADER_op_offset 0 +#define SDMA_PKT_GCR_REQ_HEADER_op_mask 0x000000FF +#define SDMA_PKT_GCR_REQ_HEADER_op_shift 0 +#define SDMA_PKT_GCR_REQ_HEADER_OP(x) (((x) & SDMA_PKT_GCR_REQ_HEADER_op_mask) << SDMA_PKT_GCR_REQ_HEADER_op_shift) + +/*define for sub_op field*/ +#define SDMA_PKT_GCR_REQ_HEADER_sub_op_offset 0 +#define SDMA_PKT_GCR_REQ_HEADER_sub_op_mask 0x000000FF +#define SDMA_PKT_GCR_REQ_HEADER_sub_op_shift 8 +#define SDMA_PKT_GCR_REQ_HEADER_SUB_OP(x) (((x) & SDMA_PKT_GCR_REQ_HEADER_sub_op_mask) << SDMA_PKT_GCR_REQ_HEADER_sub_op_shift) + +/*define for PAYLOAD1 word*/ +/*define for base_va_31_7 field*/ +#define SDMA_PKT_GCR_REQ_PAYLOAD1_base_va_31_7_offset 1 +#define SDMA_PKT_GCR_REQ_PAYLOAD1_base_va_31_7_mask 0x01FFFFFF +#define SDMA_PKT_GCR_REQ_PAYLOAD1_base_va_31_7_shift 7 +#define SDMA_PKT_GCR_REQ_PAYLOAD1_BASE_VA_31_7(x) (((x) & SDMA_PKT_GCR_REQ_PAYLOAD1_base_va_31_7_mask) << SDMA_PKT_GCR_REQ_PAYLOAD1_base_va_31_7_shift) + +/*define for PAYLOAD2 word*/ +/*define for base_va_47_32 field*/ +#define SDMA_PKT_GCR_REQ_PAYLOAD2_base_va_47_32_offset 2 +#define SDMA_PKT_GCR_REQ_PAYLOAD2_base_va_47_32_mask 0x0000FFFF +#define SDMA_PKT_GCR_REQ_PAYLOAD2_base_va_47_32_shift 0 +#define SDMA_PKT_GCR_REQ_PAYLOAD2_BASE_VA_47_32(x) (((x) & SDMA_PKT_GCR_REQ_PAYLOAD2_base_va_47_32_mask) << SDMA_PKT_GCR_REQ_PAYLOAD2_base_va_47_32_shift) + +/*define for gcr_control_15_0 field*/ +#define SDMA_PKT_GCR_REQ_PAYLOAD2_gcr_control_15_0_offset 2 +#define SDMA_PKT_GCR_REQ_PAYLOAD2_gcr_control_15_0_mask 0x0000FFFF +#define SDMA_PKT_GCR_REQ_PAYLOAD2_gcr_control_15_0_shift 16 +#define SDMA_PKT_GCR_REQ_PAYLOAD2_GCR_CONTROL_15_0(x) (((x) & SDMA_PKT_GCR_REQ_PAYLOAD2_gcr_control_15_0_mask) << SDMA_PKT_GCR_REQ_PAYLOAD2_gcr_control_15_0_shift) + +/*define for PAYLOAD3 word*/ +/*define for gcr_control_18_16 field*/ +#define SDMA_PKT_GCR_REQ_PAYLOAD3_gcr_control_18_16_offset 3 +#define SDMA_PKT_GCR_REQ_PAYLOAD3_gcr_control_18_16_mask 0x00000007 +#define SDMA_PKT_GCR_REQ_PAYLOAD3_gcr_control_18_16_shift 0 +#define SDMA_PKT_GCR_REQ_PAYLOAD3_GCR_CONTROL_18_16(x) (((x) & SDMA_PKT_GCR_REQ_PAYLOAD3_gcr_control_18_16_mask) << SDMA_PKT_GCR_REQ_PAYLOAD3_gcr_control_18_16_shift) + +/*define for limit_va_31_7 field*/ +#define SDMA_PKT_GCR_REQ_PAYLOAD3_limit_va_31_7_offset 3 +#define SDMA_PKT_GCR_REQ_PAYLOAD3_limit_va_31_7_mask 0x01FFFFFF +#define SDMA_PKT_GCR_REQ_PAYLOAD3_limit_va_31_7_shift 7 +#define SDMA_PKT_GCR_REQ_PAYLOAD3_LIMIT_VA_31_7(x) (((x) & SDMA_PKT_GCR_REQ_PAYLOAD3_limit_va_31_7_mask) << SDMA_PKT_GCR_REQ_PAYLOAD3_limit_va_31_7_shift) + +/*define for PAYLOAD4 word*/ +/*define for limit_va_47_32 field*/ +#define SDMA_PKT_GCR_REQ_PAYLOAD4_limit_va_47_32_offset 4 +#define SDMA_PKT_GCR_REQ_PAYLOAD4_limit_va_47_32_mask 0x0000FFFF +#define SDMA_PKT_GCR_REQ_PAYLOAD4_limit_va_47_32_shift 0 +#define SDMA_PKT_GCR_REQ_PAYLOAD4_LIMIT_VA_47_32(x) (((x) & SDMA_PKT_GCR_REQ_PAYLOAD4_limit_va_47_32_mask) << SDMA_PKT_GCR_REQ_PAYLOAD4_limit_va_47_32_shift) + +/*define for vmid field*/ +#define SDMA_PKT_GCR_REQ_PAYLOAD4_vmid_offset 4 +#define SDMA_PKT_GCR_REQ_PAYLOAD4_vmid_mask 0x0000000F +#define SDMA_PKT_GCR_REQ_PAYLOAD4_vmid_shift 24 +#define SDMA_PKT_GCR_REQ_PAYLOAD4_VMID(x) (((x) & SDMA_PKT_GCR_REQ_PAYLOAD4_vmid_mask) << SDMA_PKT_GCR_REQ_PAYLOAD4_vmid_shift) + + +/* +** Definitions for SDMA_PKT_NOP packet +*/ + +/*define for HEADER word*/ +/*define for op field*/ +#define SDMA_PKT_NOP_HEADER_op_offset 0 +#define SDMA_PKT_NOP_HEADER_op_mask 0x000000FF +#define SDMA_PKT_NOP_HEADER_op_shift 0 +#define SDMA_PKT_NOP_HEADER_OP(x) (((x) & SDMA_PKT_NOP_HEADER_op_mask) << SDMA_PKT_NOP_HEADER_op_shift) + +/*define for sub_op field*/ +#define SDMA_PKT_NOP_HEADER_sub_op_offset 0 +#define SDMA_PKT_NOP_HEADER_sub_op_mask 0x000000FF +#define SDMA_PKT_NOP_HEADER_sub_op_shift 8 +#define SDMA_PKT_NOP_HEADER_SUB_OP(x) (((x) & SDMA_PKT_NOP_HEADER_sub_op_mask) << SDMA_PKT_NOP_HEADER_sub_op_shift) + +/*define for count field*/ +#define SDMA_PKT_NOP_HEADER_count_offset 0 +#define SDMA_PKT_NOP_HEADER_count_mask 0x00003FFF +#define SDMA_PKT_NOP_HEADER_count_shift 16 +#define SDMA_PKT_NOP_HEADER_COUNT(x) (((x) & SDMA_PKT_NOP_HEADER_count_mask) << SDMA_PKT_NOP_HEADER_count_shift) + +/*define for DATA0 word*/ +/*define for data0 field*/ +#define SDMA_PKT_NOP_DATA0_data0_offset 1 +#define SDMA_PKT_NOP_DATA0_data0_mask 0xFFFFFFFF +#define SDMA_PKT_NOP_DATA0_data0_shift 0 +#define SDMA_PKT_NOP_DATA0_DATA0(x) (((x) & SDMA_PKT_NOP_DATA0_data0_mask) << SDMA_PKT_NOP_DATA0_data0_shift) + + +/* +** Definitions for SDMA_AQL_PKT_HEADER packet +*/ + +/*define for HEADER word*/ +/*define for format field*/ +#define SDMA_AQL_PKT_HEADER_HEADER_format_offset 0 +#define SDMA_AQL_PKT_HEADER_HEADER_format_mask 0x000000FF +#define SDMA_AQL_PKT_HEADER_HEADER_format_shift 0 +#define SDMA_AQL_PKT_HEADER_HEADER_FORMAT(x) (((x) & SDMA_AQL_PKT_HEADER_HEADER_format_mask) << SDMA_AQL_PKT_HEADER_HEADER_format_shift) + +/*define for barrier field*/ +#define SDMA_AQL_PKT_HEADER_HEADER_barrier_offset 0 +#define SDMA_AQL_PKT_HEADER_HEADER_barrier_mask 0x00000001 +#define SDMA_AQL_PKT_HEADER_HEADER_barrier_shift 8 +#define SDMA_AQL_PKT_HEADER_HEADER_BARRIER(x) (((x) & SDMA_AQL_PKT_HEADER_HEADER_barrier_mask) << SDMA_AQL_PKT_HEADER_HEADER_barrier_shift) + +/*define for acquire_fence_scope field*/ +#define SDMA_AQL_PKT_HEADER_HEADER_acquire_fence_scope_offset 0 +#define SDMA_AQL_PKT_HEADER_HEADER_acquire_fence_scope_mask 0x00000003 +#define SDMA_AQL_PKT_HEADER_HEADER_acquire_fence_scope_shift 9 +#define SDMA_AQL_PKT_HEADER_HEADER_ACQUIRE_FENCE_SCOPE(x) (((x) & SDMA_AQL_PKT_HEADER_HEADER_acquire_fence_scope_mask) << SDMA_AQL_PKT_HEADER_HEADER_acquire_fence_scope_shift) + +/*define for release_fence_scope field*/ +#define SDMA_AQL_PKT_HEADER_HEADER_release_fence_scope_offset 0 +#define SDMA_AQL_PKT_HEADER_HEADER_release_fence_scope_mask 0x00000003 +#define SDMA_AQL_PKT_HEADER_HEADER_release_fence_scope_shift 11 +#define SDMA_AQL_PKT_HEADER_HEADER_RELEASE_FENCE_SCOPE(x) (((x) & SDMA_AQL_PKT_HEADER_HEADER_release_fence_scope_mask) << SDMA_AQL_PKT_HEADER_HEADER_release_fence_scope_shift) + +/*define for reserved field*/ +#define SDMA_AQL_PKT_HEADER_HEADER_reserved_offset 0 +#define SDMA_AQL_PKT_HEADER_HEADER_reserved_mask 0x00000007 +#define SDMA_AQL_PKT_HEADER_HEADER_reserved_shift 13 +#define SDMA_AQL_PKT_HEADER_HEADER_RESERVED(x) (((x) & SDMA_AQL_PKT_HEADER_HEADER_reserved_mask) << SDMA_AQL_PKT_HEADER_HEADER_reserved_shift) + +/*define for op field*/ +#define SDMA_AQL_PKT_HEADER_HEADER_op_offset 0 +#define SDMA_AQL_PKT_HEADER_HEADER_op_mask 0x0000000F +#define SDMA_AQL_PKT_HEADER_HEADER_op_shift 16 +#define SDMA_AQL_PKT_HEADER_HEADER_OP(x) (((x) & SDMA_AQL_PKT_HEADER_HEADER_op_mask) << SDMA_AQL_PKT_HEADER_HEADER_op_shift) + +/*define for subop field*/ +#define SDMA_AQL_PKT_HEADER_HEADER_subop_offset 0 +#define SDMA_AQL_PKT_HEADER_HEADER_subop_mask 0x00000007 +#define SDMA_AQL_PKT_HEADER_HEADER_subop_shift 20 +#define SDMA_AQL_PKT_HEADER_HEADER_SUBOP(x) (((x) & SDMA_AQL_PKT_HEADER_HEADER_subop_mask) << SDMA_AQL_PKT_HEADER_HEADER_subop_shift) + +/*define for cpv field*/ +#define SDMA_AQL_PKT_HEADER_HEADER_cpv_offset 0 +#define SDMA_AQL_PKT_HEADER_HEADER_cpv_mask 0x00000001 +#define SDMA_AQL_PKT_HEADER_HEADER_cpv_shift 28 +#define SDMA_AQL_PKT_HEADER_HEADER_CPV(x) (((x) & SDMA_AQL_PKT_HEADER_HEADER_cpv_mask) << SDMA_AQL_PKT_HEADER_HEADER_cpv_shift) + + +/* +** Definitions for SDMA_AQL_PKT_COPY_LINEAR packet +*/ + +/*define for HEADER word*/ +/*define for format field*/ +#define SDMA_AQL_PKT_COPY_LINEAR_HEADER_format_offset 0 +#define SDMA_AQL_PKT_COPY_LINEAR_HEADER_format_mask 0x000000FF +#define SDMA_AQL_PKT_COPY_LINEAR_HEADER_format_shift 0 +#define SDMA_AQL_PKT_COPY_LINEAR_HEADER_FORMAT(x) (((x) & SDMA_AQL_PKT_COPY_LINEAR_HEADER_format_mask) << SDMA_AQL_PKT_COPY_LINEAR_HEADER_format_shift) + +/*define for barrier field*/ +#define SDMA_AQL_PKT_COPY_LINEAR_HEADER_barrier_offset 0 +#define SDMA_AQL_PKT_COPY_LINEAR_HEADER_barrier_mask 0x00000001 +#define SDMA_AQL_PKT_COPY_LINEAR_HEADER_barrier_shift 8 +#define SDMA_AQL_PKT_COPY_LINEAR_HEADER_BARRIER(x) (((x) & SDMA_AQL_PKT_COPY_LINEAR_HEADER_barrier_mask) << SDMA_AQL_PKT_COPY_LINEAR_HEADER_barrier_shift) + +/*define for acquire_fence_scope field*/ +#define SDMA_AQL_PKT_COPY_LINEAR_HEADER_acquire_fence_scope_offset 0 +#define SDMA_AQL_PKT_COPY_LINEAR_HEADER_acquire_fence_scope_mask 0x00000003 +#define SDMA_AQL_PKT_COPY_LINEAR_HEADER_acquire_fence_scope_shift 9 +#define SDMA_AQL_PKT_COPY_LINEAR_HEADER_ACQUIRE_FENCE_SCOPE(x) (((x) & SDMA_AQL_PKT_COPY_LINEAR_HEADER_acquire_fence_scope_mask) << SDMA_AQL_PKT_COPY_LINEAR_HEADER_acquire_fence_scope_shift) + +/*define for release_fence_scope field*/ +#define SDMA_AQL_PKT_COPY_LINEAR_HEADER_release_fence_scope_offset 0 +#define SDMA_AQL_PKT_COPY_LINEAR_HEADER_release_fence_scope_mask 0x00000003 +#define SDMA_AQL_PKT_COPY_LINEAR_HEADER_release_fence_scope_shift 11 +#define SDMA_AQL_PKT_COPY_LINEAR_HEADER_RELEASE_FENCE_SCOPE(x) (((x) & SDMA_AQL_PKT_COPY_LINEAR_HEADER_release_fence_scope_mask) << SDMA_AQL_PKT_COPY_LINEAR_HEADER_release_fence_scope_shift) + +/*define for reserved field*/ +#define SDMA_AQL_PKT_COPY_LINEAR_HEADER_reserved_offset 0 +#define SDMA_AQL_PKT_COPY_LINEAR_HEADER_reserved_mask 0x00000007 +#define SDMA_AQL_PKT_COPY_LINEAR_HEADER_reserved_shift 13 +#define SDMA_AQL_PKT_COPY_LINEAR_HEADER_RESERVED(x) (((x) & SDMA_AQL_PKT_COPY_LINEAR_HEADER_reserved_mask) << SDMA_AQL_PKT_COPY_LINEAR_HEADER_reserved_shift) + +/*define for op field*/ +#define SDMA_AQL_PKT_COPY_LINEAR_HEADER_op_offset 0 +#define SDMA_AQL_PKT_COPY_LINEAR_HEADER_op_mask 0x0000000F +#define SDMA_AQL_PKT_COPY_LINEAR_HEADER_op_shift 16 +#define SDMA_AQL_PKT_COPY_LINEAR_HEADER_OP(x) (((x) & SDMA_AQL_PKT_COPY_LINEAR_HEADER_op_mask) << SDMA_AQL_PKT_COPY_LINEAR_HEADER_op_shift) + +/*define for subop field*/ +#define SDMA_AQL_PKT_COPY_LINEAR_HEADER_subop_offset 0 +#define SDMA_AQL_PKT_COPY_LINEAR_HEADER_subop_mask 0x00000007 +#define SDMA_AQL_PKT_COPY_LINEAR_HEADER_subop_shift 20 +#define SDMA_AQL_PKT_COPY_LINEAR_HEADER_SUBOP(x) (((x) & SDMA_AQL_PKT_COPY_LINEAR_HEADER_subop_mask) << SDMA_AQL_PKT_COPY_LINEAR_HEADER_subop_shift) + +/*define for cpv field*/ +#define SDMA_AQL_PKT_COPY_LINEAR_HEADER_cpv_offset 0 +#define SDMA_AQL_PKT_COPY_LINEAR_HEADER_cpv_mask 0x00000001 +#define SDMA_AQL_PKT_COPY_LINEAR_HEADER_cpv_shift 28 +#define SDMA_AQL_PKT_COPY_LINEAR_HEADER_CPV(x) (((x) & SDMA_AQL_PKT_COPY_LINEAR_HEADER_cpv_mask) << SDMA_AQL_PKT_COPY_LINEAR_HEADER_cpv_shift) + +/*define for RESERVED_DW1 word*/ +/*define for reserved_dw1 field*/ +#define SDMA_AQL_PKT_COPY_LINEAR_RESERVED_DW1_reserved_dw1_offset 1 +#define SDMA_AQL_PKT_COPY_LINEAR_RESERVED_DW1_reserved_dw1_mask 0xFFFFFFFF +#define SDMA_AQL_PKT_COPY_LINEAR_RESERVED_DW1_reserved_dw1_shift 0 +#define SDMA_AQL_PKT_COPY_LINEAR_RESERVED_DW1_RESERVED_DW1(x) (((x) & SDMA_AQL_PKT_COPY_LINEAR_RESERVED_DW1_reserved_dw1_mask) << SDMA_AQL_PKT_COPY_LINEAR_RESERVED_DW1_reserved_dw1_shift) + +/*define for RETURN_ADDR_LO word*/ +/*define for return_addr_31_0 field*/ +#define SDMA_AQL_PKT_COPY_LINEAR_RETURN_ADDR_LO_return_addr_31_0_offset 2 +#define SDMA_AQL_PKT_COPY_LINEAR_RETURN_ADDR_LO_return_addr_31_0_mask 0xFFFFFFFF +#define SDMA_AQL_PKT_COPY_LINEAR_RETURN_ADDR_LO_return_addr_31_0_shift 0 +#define SDMA_AQL_PKT_COPY_LINEAR_RETURN_ADDR_LO_RETURN_ADDR_31_0(x) (((x) & SDMA_AQL_PKT_COPY_LINEAR_RETURN_ADDR_LO_return_addr_31_0_mask) << SDMA_AQL_PKT_COPY_LINEAR_RETURN_ADDR_LO_return_addr_31_0_shift) + +/*define for RETURN_ADDR_HI word*/ +/*define for return_addr_63_32 field*/ +#define SDMA_AQL_PKT_COPY_LINEAR_RETURN_ADDR_HI_return_addr_63_32_offset 3 +#define SDMA_AQL_PKT_COPY_LINEAR_RETURN_ADDR_HI_return_addr_63_32_mask 0xFFFFFFFF +#define SDMA_AQL_PKT_COPY_LINEAR_RETURN_ADDR_HI_return_addr_63_32_shift 0 +#define SDMA_AQL_PKT_COPY_LINEAR_RETURN_ADDR_HI_RETURN_ADDR_63_32(x) (((x) & SDMA_AQL_PKT_COPY_LINEAR_RETURN_ADDR_HI_return_addr_63_32_mask) << SDMA_AQL_PKT_COPY_LINEAR_RETURN_ADDR_HI_return_addr_63_32_shift) + +/*define for COUNT word*/ +/*define for count field*/ +#define SDMA_AQL_PKT_COPY_LINEAR_COUNT_count_offset 4 +#define SDMA_AQL_PKT_COPY_LINEAR_COUNT_count_mask 0x003FFFFF +#define SDMA_AQL_PKT_COPY_LINEAR_COUNT_count_shift 0 +#define SDMA_AQL_PKT_COPY_LINEAR_COUNT_COUNT(x) (((x) & SDMA_AQL_PKT_COPY_LINEAR_COUNT_count_mask) << SDMA_AQL_PKT_COPY_LINEAR_COUNT_count_shift) + +/*define for PARAMETER word*/ +/*define for dst_sw field*/ +#define SDMA_AQL_PKT_COPY_LINEAR_PARAMETER_dst_sw_offset 5 +#define SDMA_AQL_PKT_COPY_LINEAR_PARAMETER_dst_sw_mask 0x00000003 +#define SDMA_AQL_PKT_COPY_LINEAR_PARAMETER_dst_sw_shift 16 +#define SDMA_AQL_PKT_COPY_LINEAR_PARAMETER_DST_SW(x) (((x) & SDMA_AQL_PKT_COPY_LINEAR_PARAMETER_dst_sw_mask) << SDMA_AQL_PKT_COPY_LINEAR_PARAMETER_dst_sw_shift) + +/*define for dst_cache_policy field*/ +#define SDMA_AQL_PKT_COPY_LINEAR_PARAMETER_dst_cache_policy_offset 5 +#define SDMA_AQL_PKT_COPY_LINEAR_PARAMETER_dst_cache_policy_mask 0x00000007 +#define SDMA_AQL_PKT_COPY_LINEAR_PARAMETER_dst_cache_policy_shift 18 +#define SDMA_AQL_PKT_COPY_LINEAR_PARAMETER_DST_CACHE_POLICY(x) (((x) & SDMA_AQL_PKT_COPY_LINEAR_PARAMETER_dst_cache_policy_mask) << SDMA_AQL_PKT_COPY_LINEAR_PARAMETER_dst_cache_policy_shift) + +/*define for src_sw field*/ +#define SDMA_AQL_PKT_COPY_LINEAR_PARAMETER_src_sw_offset 5 +#define SDMA_AQL_PKT_COPY_LINEAR_PARAMETER_src_sw_mask 0x00000003 +#define SDMA_AQL_PKT_COPY_LINEAR_PARAMETER_src_sw_shift 24 +#define SDMA_AQL_PKT_COPY_LINEAR_PARAMETER_SRC_SW(x) (((x) & SDMA_AQL_PKT_COPY_LINEAR_PARAMETER_src_sw_mask) << SDMA_AQL_PKT_COPY_LINEAR_PARAMETER_src_sw_shift) + +/*define for src_cache_policy field*/ +#define SDMA_AQL_PKT_COPY_LINEAR_PARAMETER_src_cache_policy_offset 5 +#define SDMA_AQL_PKT_COPY_LINEAR_PARAMETER_src_cache_policy_mask 0x00000007 +#define SDMA_AQL_PKT_COPY_LINEAR_PARAMETER_src_cache_policy_shift 26 +#define SDMA_AQL_PKT_COPY_LINEAR_PARAMETER_SRC_CACHE_POLICY(x) (((x) & SDMA_AQL_PKT_COPY_LINEAR_PARAMETER_src_cache_policy_mask) << SDMA_AQL_PKT_COPY_LINEAR_PARAMETER_src_cache_policy_shift) + +/*define for SRC_ADDR_LO word*/ +/*define for src_addr_31_0 field*/ +#define SDMA_AQL_PKT_COPY_LINEAR_SRC_ADDR_LO_src_addr_31_0_offset 6 +#define SDMA_AQL_PKT_COPY_LINEAR_SRC_ADDR_LO_src_addr_31_0_mask 0xFFFFFFFF +#define SDMA_AQL_PKT_COPY_LINEAR_SRC_ADDR_LO_src_addr_31_0_shift 0 +#define SDMA_AQL_PKT_COPY_LINEAR_SRC_ADDR_LO_SRC_ADDR_31_0(x) (((x) & SDMA_AQL_PKT_COPY_LINEAR_SRC_ADDR_LO_src_addr_31_0_mask) << SDMA_AQL_PKT_COPY_LINEAR_SRC_ADDR_LO_src_addr_31_0_shift) + +/*define for SRC_ADDR_HI word*/ +/*define for src_addr_63_32 field*/ +#define SDMA_AQL_PKT_COPY_LINEAR_SRC_ADDR_HI_src_addr_63_32_offset 7 +#define SDMA_AQL_PKT_COPY_LINEAR_SRC_ADDR_HI_src_addr_63_32_mask 0xFFFFFFFF +#define SDMA_AQL_PKT_COPY_LINEAR_SRC_ADDR_HI_src_addr_63_32_shift 0 +#define SDMA_AQL_PKT_COPY_LINEAR_SRC_ADDR_HI_SRC_ADDR_63_32(x) (((x) & SDMA_AQL_PKT_COPY_LINEAR_SRC_ADDR_HI_src_addr_63_32_mask) << SDMA_AQL_PKT_COPY_LINEAR_SRC_ADDR_HI_src_addr_63_32_shift) + +/*define for DST_ADDR_LO word*/ +/*define for dst_addr_31_0 field*/ +#define SDMA_AQL_PKT_COPY_LINEAR_DST_ADDR_LO_dst_addr_31_0_offset 8 +#define SDMA_AQL_PKT_COPY_LINEAR_DST_ADDR_LO_dst_addr_31_0_mask 0xFFFFFFFF +#define SDMA_AQL_PKT_COPY_LINEAR_DST_ADDR_LO_dst_addr_31_0_shift 0 +#define SDMA_AQL_PKT_COPY_LINEAR_DST_ADDR_LO_DST_ADDR_31_0(x) (((x) & SDMA_AQL_PKT_COPY_LINEAR_DST_ADDR_LO_dst_addr_31_0_mask) << SDMA_AQL_PKT_COPY_LINEAR_DST_ADDR_LO_dst_addr_31_0_shift) + +/*define for DST_ADDR_HI word*/ +/*define for dst_addr_63_32 field*/ +#define SDMA_AQL_PKT_COPY_LINEAR_DST_ADDR_HI_dst_addr_63_32_offset 9 +#define SDMA_AQL_PKT_COPY_LINEAR_DST_ADDR_HI_dst_addr_63_32_mask 0xFFFFFFFF +#define SDMA_AQL_PKT_COPY_LINEAR_DST_ADDR_HI_dst_addr_63_32_shift 0 +#define SDMA_AQL_PKT_COPY_LINEAR_DST_ADDR_HI_DST_ADDR_63_32(x) (((x) & SDMA_AQL_PKT_COPY_LINEAR_DST_ADDR_HI_dst_addr_63_32_mask) << SDMA_AQL_PKT_COPY_LINEAR_DST_ADDR_HI_dst_addr_63_32_shift) + +/*define for RESERVED_DW10 word*/ +/*define for reserved_dw10 field*/ +#define SDMA_AQL_PKT_COPY_LINEAR_RESERVED_DW10_reserved_dw10_offset 10 +#define SDMA_AQL_PKT_COPY_LINEAR_RESERVED_DW10_reserved_dw10_mask 0xFFFFFFFF +#define SDMA_AQL_PKT_COPY_LINEAR_RESERVED_DW10_reserved_dw10_shift 0 +#define SDMA_AQL_PKT_COPY_LINEAR_RESERVED_DW10_RESERVED_DW10(x) (((x) & SDMA_AQL_PKT_COPY_LINEAR_RESERVED_DW10_reserved_dw10_mask) << SDMA_AQL_PKT_COPY_LINEAR_RESERVED_DW10_reserved_dw10_shift) + +/*define for RESERVED_DW11 word*/ +/*define for reserved_dw11 field*/ +#define SDMA_AQL_PKT_COPY_LINEAR_RESERVED_DW11_reserved_dw11_offset 11 +#define SDMA_AQL_PKT_COPY_LINEAR_RESERVED_DW11_reserved_dw11_mask 0xFFFFFFFF +#define SDMA_AQL_PKT_COPY_LINEAR_RESERVED_DW11_reserved_dw11_shift 0 +#define SDMA_AQL_PKT_COPY_LINEAR_RESERVED_DW11_RESERVED_DW11(x) (((x) & SDMA_AQL_PKT_COPY_LINEAR_RESERVED_DW11_reserved_dw11_mask) << SDMA_AQL_PKT_COPY_LINEAR_RESERVED_DW11_reserved_dw11_shift) + +/*define for RESERVED_DW12 word*/ +/*define for reserved_dw12 field*/ +#define SDMA_AQL_PKT_COPY_LINEAR_RESERVED_DW12_reserved_dw12_offset 12 +#define SDMA_AQL_PKT_COPY_LINEAR_RESERVED_DW12_reserved_dw12_mask 0xFFFFFFFF +#define SDMA_AQL_PKT_COPY_LINEAR_RESERVED_DW12_reserved_dw12_shift 0 +#define SDMA_AQL_PKT_COPY_LINEAR_RESERVED_DW12_RESERVED_DW12(x) (((x) & SDMA_AQL_PKT_COPY_LINEAR_RESERVED_DW12_reserved_dw12_mask) << SDMA_AQL_PKT_COPY_LINEAR_RESERVED_DW12_reserved_dw12_shift) + +/*define for RESERVED_DW13 word*/ +/*define for reserved_dw13 field*/ +#define SDMA_AQL_PKT_COPY_LINEAR_RESERVED_DW13_reserved_dw13_offset 13 +#define SDMA_AQL_PKT_COPY_LINEAR_RESERVED_DW13_reserved_dw13_mask 0xFFFFFFFF +#define SDMA_AQL_PKT_COPY_LINEAR_RESERVED_DW13_reserved_dw13_shift 0 +#define SDMA_AQL_PKT_COPY_LINEAR_RESERVED_DW13_RESERVED_DW13(x) (((x) & SDMA_AQL_PKT_COPY_LINEAR_RESERVED_DW13_reserved_dw13_mask) << SDMA_AQL_PKT_COPY_LINEAR_RESERVED_DW13_reserved_dw13_shift) + +/*define for COMPLETION_SIGNAL_LO word*/ +/*define for completion_signal_31_0 field*/ +#define SDMA_AQL_PKT_COPY_LINEAR_COMPLETION_SIGNAL_LO_completion_signal_31_0_offset 14 +#define SDMA_AQL_PKT_COPY_LINEAR_COMPLETION_SIGNAL_LO_completion_signal_31_0_mask 0xFFFFFFFF +#define SDMA_AQL_PKT_COPY_LINEAR_COMPLETION_SIGNAL_LO_completion_signal_31_0_shift 0 +#define SDMA_AQL_PKT_COPY_LINEAR_COMPLETION_SIGNAL_LO_COMPLETION_SIGNAL_31_0(x) (((x) & SDMA_AQL_PKT_COPY_LINEAR_COMPLETION_SIGNAL_LO_completion_signal_31_0_mask) << SDMA_AQL_PKT_COPY_LINEAR_COMPLETION_SIGNAL_LO_completion_signal_31_0_shift) + +/*define for COMPLETION_SIGNAL_HI word*/ +/*define for completion_signal_63_32 field*/ +#define SDMA_AQL_PKT_COPY_LINEAR_COMPLETION_SIGNAL_HI_completion_signal_63_32_offset 15 +#define SDMA_AQL_PKT_COPY_LINEAR_COMPLETION_SIGNAL_HI_completion_signal_63_32_mask 0xFFFFFFFF +#define SDMA_AQL_PKT_COPY_LINEAR_COMPLETION_SIGNAL_HI_completion_signal_63_32_shift 0 +#define SDMA_AQL_PKT_COPY_LINEAR_COMPLETION_SIGNAL_HI_COMPLETION_SIGNAL_63_32(x) (((x) & SDMA_AQL_PKT_COPY_LINEAR_COMPLETION_SIGNAL_HI_completion_signal_63_32_mask) << SDMA_AQL_PKT_COPY_LINEAR_COMPLETION_SIGNAL_HI_completion_signal_63_32_shift) + + +/* +** Definitions for SDMA_AQL_PKT_BARRIER_OR packet +*/ + +/*define for HEADER word*/ +/*define for format field*/ +#define SDMA_AQL_PKT_BARRIER_OR_HEADER_format_offset 0 +#define SDMA_AQL_PKT_BARRIER_OR_HEADER_format_mask 0x000000FF +#define SDMA_AQL_PKT_BARRIER_OR_HEADER_format_shift 0 +#define SDMA_AQL_PKT_BARRIER_OR_HEADER_FORMAT(x) (((x) & SDMA_AQL_PKT_BARRIER_OR_HEADER_format_mask) << SDMA_AQL_PKT_BARRIER_OR_HEADER_format_shift) + +/*define for barrier field*/ +#define SDMA_AQL_PKT_BARRIER_OR_HEADER_barrier_offset 0 +#define SDMA_AQL_PKT_BARRIER_OR_HEADER_barrier_mask 0x00000001 +#define SDMA_AQL_PKT_BARRIER_OR_HEADER_barrier_shift 8 +#define SDMA_AQL_PKT_BARRIER_OR_HEADER_BARRIER(x) (((x) & SDMA_AQL_PKT_BARRIER_OR_HEADER_barrier_mask) << SDMA_AQL_PKT_BARRIER_OR_HEADER_barrier_shift) + +/*define for acquire_fence_scope field*/ +#define SDMA_AQL_PKT_BARRIER_OR_HEADER_acquire_fence_scope_offset 0 +#define SDMA_AQL_PKT_BARRIER_OR_HEADER_acquire_fence_scope_mask 0x00000003 +#define SDMA_AQL_PKT_BARRIER_OR_HEADER_acquire_fence_scope_shift 9 +#define SDMA_AQL_PKT_BARRIER_OR_HEADER_ACQUIRE_FENCE_SCOPE(x) (((x) & SDMA_AQL_PKT_BARRIER_OR_HEADER_acquire_fence_scope_mask) << SDMA_AQL_PKT_BARRIER_OR_HEADER_acquire_fence_scope_shift) + +/*define for release_fence_scope field*/ +#define SDMA_AQL_PKT_BARRIER_OR_HEADER_release_fence_scope_offset 0 +#define SDMA_AQL_PKT_BARRIER_OR_HEADER_release_fence_scope_mask 0x00000003 +#define SDMA_AQL_PKT_BARRIER_OR_HEADER_release_fence_scope_shift 11 +#define SDMA_AQL_PKT_BARRIER_OR_HEADER_RELEASE_FENCE_SCOPE(x) (((x) & SDMA_AQL_PKT_BARRIER_OR_HEADER_release_fence_scope_mask) << SDMA_AQL_PKT_BARRIER_OR_HEADER_release_fence_scope_shift) + +/*define for reserved field*/ +#define SDMA_AQL_PKT_BARRIER_OR_HEADER_reserved_offset 0 +#define SDMA_AQL_PKT_BARRIER_OR_HEADER_reserved_mask 0x00000007 +#define SDMA_AQL_PKT_BARRIER_OR_HEADER_reserved_shift 13 +#define SDMA_AQL_PKT_BARRIER_OR_HEADER_RESERVED(x) (((x) & SDMA_AQL_PKT_BARRIER_OR_HEADER_reserved_mask) << SDMA_AQL_PKT_BARRIER_OR_HEADER_reserved_shift) + +/*define for op field*/ +#define SDMA_AQL_PKT_BARRIER_OR_HEADER_op_offset 0 +#define SDMA_AQL_PKT_BARRIER_OR_HEADER_op_mask 0x0000000F +#define SDMA_AQL_PKT_BARRIER_OR_HEADER_op_shift 16 +#define SDMA_AQL_PKT_BARRIER_OR_HEADER_OP(x) (((x) & SDMA_AQL_PKT_BARRIER_OR_HEADER_op_mask) << SDMA_AQL_PKT_BARRIER_OR_HEADER_op_shift) + +/*define for subop field*/ +#define SDMA_AQL_PKT_BARRIER_OR_HEADER_subop_offset 0 +#define SDMA_AQL_PKT_BARRIER_OR_HEADER_subop_mask 0x00000007 +#define SDMA_AQL_PKT_BARRIER_OR_HEADER_subop_shift 20 +#define SDMA_AQL_PKT_BARRIER_OR_HEADER_SUBOP(x) (((x) & SDMA_AQL_PKT_BARRIER_OR_HEADER_subop_mask) << SDMA_AQL_PKT_BARRIER_OR_HEADER_subop_shift) + +/*define for cpv field*/ +#define SDMA_AQL_PKT_BARRIER_OR_HEADER_cpv_offset 0 +#define SDMA_AQL_PKT_BARRIER_OR_HEADER_cpv_mask 0x00000001 +#define SDMA_AQL_PKT_BARRIER_OR_HEADER_cpv_shift 28 +#define SDMA_AQL_PKT_BARRIER_OR_HEADER_CPV(x) (((x) & SDMA_AQL_PKT_BARRIER_OR_HEADER_cpv_mask) << SDMA_AQL_PKT_BARRIER_OR_HEADER_cpv_shift) + +/*define for RESERVED_DW1 word*/ +/*define for reserved_dw1 field*/ +#define SDMA_AQL_PKT_BARRIER_OR_RESERVED_DW1_reserved_dw1_offset 1 +#define SDMA_AQL_PKT_BARRIER_OR_RESERVED_DW1_reserved_dw1_mask 0xFFFFFFFF +#define SDMA_AQL_PKT_BARRIER_OR_RESERVED_DW1_reserved_dw1_shift 0 +#define SDMA_AQL_PKT_BARRIER_OR_RESERVED_DW1_RESERVED_DW1(x) (((x) & SDMA_AQL_PKT_BARRIER_OR_RESERVED_DW1_reserved_dw1_mask) << SDMA_AQL_PKT_BARRIER_OR_RESERVED_DW1_reserved_dw1_shift) + +/*define for DEPENDENT_ADDR_0_LO word*/ +/*define for dependent_addr_0_31_0 field*/ +#define SDMA_AQL_PKT_BARRIER_OR_DEPENDENT_ADDR_0_LO_dependent_addr_0_31_0_offset 2 +#define SDMA_AQL_PKT_BARRIER_OR_DEPENDENT_ADDR_0_LO_dependent_addr_0_31_0_mask 0xFFFFFFFF +#define SDMA_AQL_PKT_BARRIER_OR_DEPENDENT_ADDR_0_LO_dependent_addr_0_31_0_shift 0 +#define SDMA_AQL_PKT_BARRIER_OR_DEPENDENT_ADDR_0_LO_DEPENDENT_ADDR_0_31_0(x) (((x) & SDMA_AQL_PKT_BARRIER_OR_DEPENDENT_ADDR_0_LO_dependent_addr_0_31_0_mask) << SDMA_AQL_PKT_BARRIER_OR_DEPENDENT_ADDR_0_LO_dependent_addr_0_31_0_shift) + +/*define for DEPENDENT_ADDR_0_HI word*/ +/*define for dependent_addr_0_63_32 field*/ +#define SDMA_AQL_PKT_BARRIER_OR_DEPENDENT_ADDR_0_HI_dependent_addr_0_63_32_offset 3 +#define SDMA_AQL_PKT_BARRIER_OR_DEPENDENT_ADDR_0_HI_dependent_addr_0_63_32_mask 0xFFFFFFFF +#define SDMA_AQL_PKT_BARRIER_OR_DEPENDENT_ADDR_0_HI_dependent_addr_0_63_32_shift 0 +#define SDMA_AQL_PKT_BARRIER_OR_DEPENDENT_ADDR_0_HI_DEPENDENT_ADDR_0_63_32(x) (((x) & SDMA_AQL_PKT_BARRIER_OR_DEPENDENT_ADDR_0_HI_dependent_addr_0_63_32_mask) << SDMA_AQL_PKT_BARRIER_OR_DEPENDENT_ADDR_0_HI_dependent_addr_0_63_32_shift) + +/*define for DEPENDENT_ADDR_1_LO word*/ +/*define for dependent_addr_1_31_0 field*/ +#define SDMA_AQL_PKT_BARRIER_OR_DEPENDENT_ADDR_1_LO_dependent_addr_1_31_0_offset 4 +#define SDMA_AQL_PKT_BARRIER_OR_DEPENDENT_ADDR_1_LO_dependent_addr_1_31_0_mask 0xFFFFFFFF +#define SDMA_AQL_PKT_BARRIER_OR_DEPENDENT_ADDR_1_LO_dependent_addr_1_31_0_shift 0 +#define SDMA_AQL_PKT_BARRIER_OR_DEPENDENT_ADDR_1_LO_DEPENDENT_ADDR_1_31_0(x) (((x) & SDMA_AQL_PKT_BARRIER_OR_DEPENDENT_ADDR_1_LO_dependent_addr_1_31_0_mask) << SDMA_AQL_PKT_BARRIER_OR_DEPENDENT_ADDR_1_LO_dependent_addr_1_31_0_shift) + +/*define for DEPENDENT_ADDR_1_HI word*/ +/*define for dependent_addr_1_63_32 field*/ +#define SDMA_AQL_PKT_BARRIER_OR_DEPENDENT_ADDR_1_HI_dependent_addr_1_63_32_offset 5 +#define SDMA_AQL_PKT_BARRIER_OR_DEPENDENT_ADDR_1_HI_dependent_addr_1_63_32_mask 0xFFFFFFFF +#define SDMA_AQL_PKT_BARRIER_OR_DEPENDENT_ADDR_1_HI_dependent_addr_1_63_32_shift 0 +#define SDMA_AQL_PKT_BARRIER_OR_DEPENDENT_ADDR_1_HI_DEPENDENT_ADDR_1_63_32(x) (((x) & SDMA_AQL_PKT_BARRIER_OR_DEPENDENT_ADDR_1_HI_dependent_addr_1_63_32_mask) << SDMA_AQL_PKT_BARRIER_OR_DEPENDENT_ADDR_1_HI_dependent_addr_1_63_32_shift) + +/*define for DEPENDENT_ADDR_2_LO word*/ +/*define for dependent_addr_2_31_0 field*/ +#define SDMA_AQL_PKT_BARRIER_OR_DEPENDENT_ADDR_2_LO_dependent_addr_2_31_0_offset 6 +#define SDMA_AQL_PKT_BARRIER_OR_DEPENDENT_ADDR_2_LO_dependent_addr_2_31_0_mask 0xFFFFFFFF +#define SDMA_AQL_PKT_BARRIER_OR_DEPENDENT_ADDR_2_LO_dependent_addr_2_31_0_shift 0 +#define SDMA_AQL_PKT_BARRIER_OR_DEPENDENT_ADDR_2_LO_DEPENDENT_ADDR_2_31_0(x) (((x) & SDMA_AQL_PKT_BARRIER_OR_DEPENDENT_ADDR_2_LO_dependent_addr_2_31_0_mask) << SDMA_AQL_PKT_BARRIER_OR_DEPENDENT_ADDR_2_LO_dependent_addr_2_31_0_shift) + +/*define for DEPENDENT_ADDR_2_HI word*/ +/*define for dependent_addr_2_63_32 field*/ +#define SDMA_AQL_PKT_BARRIER_OR_DEPENDENT_ADDR_2_HI_dependent_addr_2_63_32_offset 7 +#define SDMA_AQL_PKT_BARRIER_OR_DEPENDENT_ADDR_2_HI_dependent_addr_2_63_32_mask 0xFFFFFFFF +#define SDMA_AQL_PKT_BARRIER_OR_DEPENDENT_ADDR_2_HI_dependent_addr_2_63_32_shift 0 +#define SDMA_AQL_PKT_BARRIER_OR_DEPENDENT_ADDR_2_HI_DEPENDENT_ADDR_2_63_32(x) (((x) & SDMA_AQL_PKT_BARRIER_OR_DEPENDENT_ADDR_2_HI_dependent_addr_2_63_32_mask) << SDMA_AQL_PKT_BARRIER_OR_DEPENDENT_ADDR_2_HI_dependent_addr_2_63_32_shift) + +/*define for DEPENDENT_ADDR_3_LO word*/ +/*define for dependent_addr_3_31_0 field*/ +#define SDMA_AQL_PKT_BARRIER_OR_DEPENDENT_ADDR_3_LO_dependent_addr_3_31_0_offset 8 +#define SDMA_AQL_PKT_BARRIER_OR_DEPENDENT_ADDR_3_LO_dependent_addr_3_31_0_mask 0xFFFFFFFF +#define SDMA_AQL_PKT_BARRIER_OR_DEPENDENT_ADDR_3_LO_dependent_addr_3_31_0_shift 0 +#define SDMA_AQL_PKT_BARRIER_OR_DEPENDENT_ADDR_3_LO_DEPENDENT_ADDR_3_31_0(x) (((x) & SDMA_AQL_PKT_BARRIER_OR_DEPENDENT_ADDR_3_LO_dependent_addr_3_31_0_mask) << SDMA_AQL_PKT_BARRIER_OR_DEPENDENT_ADDR_3_LO_dependent_addr_3_31_0_shift) + +/*define for DEPENDENT_ADDR_3_HI word*/ +/*define for dependent_addr_3_63_32 field*/ +#define SDMA_AQL_PKT_BARRIER_OR_DEPENDENT_ADDR_3_HI_dependent_addr_3_63_32_offset 9 +#define SDMA_AQL_PKT_BARRIER_OR_DEPENDENT_ADDR_3_HI_dependent_addr_3_63_32_mask 0xFFFFFFFF +#define SDMA_AQL_PKT_BARRIER_OR_DEPENDENT_ADDR_3_HI_dependent_addr_3_63_32_shift 0 +#define SDMA_AQL_PKT_BARRIER_OR_DEPENDENT_ADDR_3_HI_DEPENDENT_ADDR_3_63_32(x) (((x) & SDMA_AQL_PKT_BARRIER_OR_DEPENDENT_ADDR_3_HI_dependent_addr_3_63_32_mask) << SDMA_AQL_PKT_BARRIER_OR_DEPENDENT_ADDR_3_HI_dependent_addr_3_63_32_shift) + +/*define for DEPENDENT_ADDR_4_LO word*/ +/*define for dependent_addr_4_31_0 field*/ +#define SDMA_AQL_PKT_BARRIER_OR_DEPENDENT_ADDR_4_LO_dependent_addr_4_31_0_offset 10 +#define SDMA_AQL_PKT_BARRIER_OR_DEPENDENT_ADDR_4_LO_dependent_addr_4_31_0_mask 0xFFFFFFFF +#define SDMA_AQL_PKT_BARRIER_OR_DEPENDENT_ADDR_4_LO_dependent_addr_4_31_0_shift 0 +#define SDMA_AQL_PKT_BARRIER_OR_DEPENDENT_ADDR_4_LO_DEPENDENT_ADDR_4_31_0(x) (((x) & SDMA_AQL_PKT_BARRIER_OR_DEPENDENT_ADDR_4_LO_dependent_addr_4_31_0_mask) << SDMA_AQL_PKT_BARRIER_OR_DEPENDENT_ADDR_4_LO_dependent_addr_4_31_0_shift) + +/*define for DEPENDENT_ADDR_4_HI word*/ +/*define for dependent_addr_4_63_32 field*/ +#define SDMA_AQL_PKT_BARRIER_OR_DEPENDENT_ADDR_4_HI_dependent_addr_4_63_32_offset 11 +#define SDMA_AQL_PKT_BARRIER_OR_DEPENDENT_ADDR_4_HI_dependent_addr_4_63_32_mask 0xFFFFFFFF +#define SDMA_AQL_PKT_BARRIER_OR_DEPENDENT_ADDR_4_HI_dependent_addr_4_63_32_shift 0 +#define SDMA_AQL_PKT_BARRIER_OR_DEPENDENT_ADDR_4_HI_DEPENDENT_ADDR_4_63_32(x) (((x) & SDMA_AQL_PKT_BARRIER_OR_DEPENDENT_ADDR_4_HI_dependent_addr_4_63_32_mask) << SDMA_AQL_PKT_BARRIER_OR_DEPENDENT_ADDR_4_HI_dependent_addr_4_63_32_shift) + +/*define for CACHE_POLICY word*/ +/*define for cache_policy0 field*/ +#define SDMA_AQL_PKT_BARRIER_OR_CACHE_POLICY_cache_policy0_offset 12 +#define SDMA_AQL_PKT_BARRIER_OR_CACHE_POLICY_cache_policy0_mask 0x00000007 +#define SDMA_AQL_PKT_BARRIER_OR_CACHE_POLICY_cache_policy0_shift 0 +#define SDMA_AQL_PKT_BARRIER_OR_CACHE_POLICY_CACHE_POLICY0(x) (((x) & SDMA_AQL_PKT_BARRIER_OR_CACHE_POLICY_cache_policy0_mask) << SDMA_AQL_PKT_BARRIER_OR_CACHE_POLICY_cache_policy0_shift) + +/*define for cache_policy1 field*/ +#define SDMA_AQL_PKT_BARRIER_OR_CACHE_POLICY_cache_policy1_offset 12 +#define SDMA_AQL_PKT_BARRIER_OR_CACHE_POLICY_cache_policy1_mask 0x00000007 +#define SDMA_AQL_PKT_BARRIER_OR_CACHE_POLICY_cache_policy1_shift 5 +#define SDMA_AQL_PKT_BARRIER_OR_CACHE_POLICY_CACHE_POLICY1(x) (((x) & SDMA_AQL_PKT_BARRIER_OR_CACHE_POLICY_cache_policy1_mask) << SDMA_AQL_PKT_BARRIER_OR_CACHE_POLICY_cache_policy1_shift) + +/*define for cache_policy2 field*/ +#define SDMA_AQL_PKT_BARRIER_OR_CACHE_POLICY_cache_policy2_offset 12 +#define SDMA_AQL_PKT_BARRIER_OR_CACHE_POLICY_cache_policy2_mask 0x00000007 +#define SDMA_AQL_PKT_BARRIER_OR_CACHE_POLICY_cache_policy2_shift 10 +#define SDMA_AQL_PKT_BARRIER_OR_CACHE_POLICY_CACHE_POLICY2(x) (((x) & SDMA_AQL_PKT_BARRIER_OR_CACHE_POLICY_cache_policy2_mask) << SDMA_AQL_PKT_BARRIER_OR_CACHE_POLICY_cache_policy2_shift) + +/*define for cache_policy3 field*/ +#define SDMA_AQL_PKT_BARRIER_OR_CACHE_POLICY_cache_policy3_offset 12 +#define SDMA_AQL_PKT_BARRIER_OR_CACHE_POLICY_cache_policy3_mask 0x00000007 +#define SDMA_AQL_PKT_BARRIER_OR_CACHE_POLICY_cache_policy3_shift 15 +#define SDMA_AQL_PKT_BARRIER_OR_CACHE_POLICY_CACHE_POLICY3(x) (((x) & SDMA_AQL_PKT_BARRIER_OR_CACHE_POLICY_cache_policy3_mask) << SDMA_AQL_PKT_BARRIER_OR_CACHE_POLICY_cache_policy3_shift) + +/*define for cache_policy4 field*/ +#define SDMA_AQL_PKT_BARRIER_OR_CACHE_POLICY_cache_policy4_offset 12 +#define SDMA_AQL_PKT_BARRIER_OR_CACHE_POLICY_cache_policy4_mask 0x00000007 +#define SDMA_AQL_PKT_BARRIER_OR_CACHE_POLICY_cache_policy4_shift 20 +#define SDMA_AQL_PKT_BARRIER_OR_CACHE_POLICY_CACHE_POLICY4(x) (((x) & SDMA_AQL_PKT_BARRIER_OR_CACHE_POLICY_cache_policy4_mask) << SDMA_AQL_PKT_BARRIER_OR_CACHE_POLICY_cache_policy4_shift) + +/*define for RESERVED_DW13 word*/ +/*define for reserved_dw13 field*/ +#define SDMA_AQL_PKT_BARRIER_OR_RESERVED_DW13_reserved_dw13_offset 13 +#define SDMA_AQL_PKT_BARRIER_OR_RESERVED_DW13_reserved_dw13_mask 0xFFFFFFFF +#define SDMA_AQL_PKT_BARRIER_OR_RESERVED_DW13_reserved_dw13_shift 0 +#define SDMA_AQL_PKT_BARRIER_OR_RESERVED_DW13_RESERVED_DW13(x) (((x) & SDMA_AQL_PKT_BARRIER_OR_RESERVED_DW13_reserved_dw13_mask) << SDMA_AQL_PKT_BARRIER_OR_RESERVED_DW13_reserved_dw13_shift) + +/*define for COMPLETION_SIGNAL_LO word*/ +/*define for completion_signal_31_0 field*/ +#define SDMA_AQL_PKT_BARRIER_OR_COMPLETION_SIGNAL_LO_completion_signal_31_0_offset 14 +#define SDMA_AQL_PKT_BARRIER_OR_COMPLETION_SIGNAL_LO_completion_signal_31_0_mask 0xFFFFFFFF +#define SDMA_AQL_PKT_BARRIER_OR_COMPLETION_SIGNAL_LO_completion_signal_31_0_shift 0 +#define SDMA_AQL_PKT_BARRIER_OR_COMPLETION_SIGNAL_LO_COMPLETION_SIGNAL_31_0(x) (((x) & SDMA_AQL_PKT_BARRIER_OR_COMPLETION_SIGNAL_LO_completion_signal_31_0_mask) << SDMA_AQL_PKT_BARRIER_OR_COMPLETION_SIGNAL_LO_completion_signal_31_0_shift) + +/*define for COMPLETION_SIGNAL_HI word*/ +/*define for completion_signal_63_32 field*/ +#define SDMA_AQL_PKT_BARRIER_OR_COMPLETION_SIGNAL_HI_completion_signal_63_32_offset 15 +#define SDMA_AQL_PKT_BARRIER_OR_COMPLETION_SIGNAL_HI_completion_signal_63_32_mask 0xFFFFFFFF +#define SDMA_AQL_PKT_BARRIER_OR_COMPLETION_SIGNAL_HI_completion_signal_63_32_shift 0 +#define SDMA_AQL_PKT_BARRIER_OR_COMPLETION_SIGNAL_HI_COMPLETION_SIGNAL_63_32(x) (((x) & SDMA_AQL_PKT_BARRIER_OR_COMPLETION_SIGNAL_HI_completion_signal_63_32_mask) << SDMA_AQL_PKT_BARRIER_OR_COMPLETION_SIGNAL_HI_completion_signal_63_32_shift) + + +#endif /* __SDMA_V6_0_0_PKT_OPEN_H_ */ diff --git a/drivers/gpu/drm/amd/amdgpu/si.c b/drivers/gpu/drm/amd/amdgpu/si.c index e6d2f74a7976..7f99e130acd0 100644 --- a/drivers/gpu/drm/amd/amdgpu/si.c +++ b/drivers/gpu/drm/amd/amdgpu/si.c @@ -2453,7 +2453,7 @@ static void si_program_aspm(struct amdgpu_device *adev) bool disable_l0s = false, disable_l1 = false, disable_plloff_in_l1 = false; bool disable_clkreq = false; - if (amdgpu_aspm == 0) + if (!amdgpu_device_should_use_aspm(adev)) return; if (adev->flags & AMD_IS_APU) diff --git a/drivers/gpu/drm/amd/amdgpu/si_dma.c b/drivers/gpu/drm/amd/amdgpu/si_dma.c index 195b45bcb8ad..4d5e718540aa 100644 --- a/drivers/gpu/drm/amd/amdgpu/si_dma.c +++ b/drivers/gpu/drm/amd/amdgpu/si_dma.c @@ -40,7 +40,7 @@ static void si_dma_set_irq_funcs(struct amdgpu_device *adev); static uint64_t si_dma_ring_get_rptr(struct amdgpu_ring *ring) { - return ring->adev->wb.wb[ring->rptr_offs>>2]; + return *ring->rptr_cpu_addr; } static uint64_t si_dma_ring_get_wptr(struct amdgpu_ring *ring) @@ -56,8 +56,7 @@ static void si_dma_ring_set_wptr(struct amdgpu_ring *ring) struct amdgpu_device *adev = ring->adev; u32 me = (ring == &adev->sdma.instance[0].ring) ? 0 : 1; - WREG32(DMA_RB_WPTR + sdma_offsets[me], - (lower_32_bits(ring->wptr) << 2) & 0x3fffc); + WREG32(DMA_RB_WPTR + sdma_offsets[me], (ring->wptr << 2) & 0x3fffc); } static void si_dma_ring_emit_ib(struct amdgpu_ring *ring, @@ -117,15 +116,14 @@ static void si_dma_stop(struct amdgpu_device *adev) u32 rb_cntl; unsigned i; + amdgpu_sdma_unset_buffer_funcs_helper(adev); + for (i = 0; i < adev->sdma.num_instances; i++) { ring = &adev->sdma.instance[i].ring; /* dma0 */ rb_cntl = RREG32(DMA_RB_CNTL + sdma_offsets[i]); rb_cntl &= ~DMA_RB_ENABLE; WREG32(DMA_RB_CNTL + sdma_offsets[i], rb_cntl); - - if (adev->mman.buffer_funcs_ring == ring) - amdgpu_ttm_set_buffer_funcs_status(adev, false); } } @@ -154,7 +152,7 @@ static int si_dma_start(struct amdgpu_device *adev) WREG32(DMA_RB_RPTR + sdma_offsets[i], 0); WREG32(DMA_RB_WPTR + sdma_offsets[i], 0); - rptr_addr = adev->wb.gpu_addr + (ring->rptr_offs * 4); + rptr_addr = ring->rptr_gpu_addr; WREG32(DMA_RB_RPTR_ADDR_LO + sdma_offsets[i], lower_32_bits(rptr_addr)); WREG32(DMA_RB_RPTR_ADDR_HI + sdma_offsets[i], upper_32_bits(rptr_addr) & 0xFF); @@ -175,7 +173,7 @@ static int si_dma_start(struct amdgpu_device *adev) WREG32(DMA_CNTL + sdma_offsets[i], dma_cntl); ring->wptr = 0; - WREG32(DMA_RB_WPTR + sdma_offsets[i], lower_32_bits(ring->wptr) << 2); + WREG32(DMA_RB_WPTR + sdma_offsets[i], ring->wptr << 2); WREG32(DMA_RB_CNTL + sdma_offsets[i], rb_cntl | DMA_RB_ENABLE); ring->sched.ready = true; diff --git a/drivers/gpu/drm/amd/amdgpu/sienna_cichlid.c b/drivers/gpu/drm/amd/amdgpu/sienna_cichlid.c new file mode 100644 index 000000000000..81a6d5b94987 --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/sienna_cichlid.c @@ -0,0 +1,312 @@ +/* + * Copyright 2021 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#include "sienna_cichlid.h" +#include "amdgpu_reset.h" +#include "amdgpu_amdkfd.h" +#include "amdgpu_dpm.h" +#include "amdgpu_job.h" +#include "amdgpu_ring.h" +#include "amdgpu_ras.h" +#include "amdgpu_psp.h" +#include "amdgpu_xgmi.h" + +static bool sienna_cichlid_is_mode2_default(struct amdgpu_reset_control *reset_ctl) +{ +#if 0 + struct amdgpu_device *adev = (struct amdgpu_device *)reset_ctl->handle; + + if (adev->ip_versions[MP1_HWIP][0] == IP_VERSION(11, 0, 7) && + adev->pm.fw_version >= 0x3a5500 && !amdgpu_sriov_vf(adev)) + return true; +#endif + return false; +} + +static struct amdgpu_reset_handler * +sienna_cichlid_get_reset_handler(struct amdgpu_reset_control *reset_ctl, + struct amdgpu_reset_context *reset_context) +{ + struct amdgpu_reset_handler *handler; + + if (reset_context->method != AMD_RESET_METHOD_NONE) { + list_for_each_entry(handler, &reset_ctl->reset_handlers, + handler_list) { + if (handler->reset_method == reset_context->method) + return handler; + } + } + + if (sienna_cichlid_is_mode2_default(reset_ctl)) { + list_for_each_entry (handler, &reset_ctl->reset_handlers, + handler_list) { + if (handler->reset_method == AMD_RESET_METHOD_MODE2) + return handler; + } + } + + return NULL; +} + +static int sienna_cichlid_mode2_suspend_ip(struct amdgpu_device *adev) +{ + int r, i; + + amdgpu_device_set_pg_state(adev, AMD_PG_STATE_UNGATE); + amdgpu_device_set_cg_state(adev, AMD_CG_STATE_UNGATE); + + for (i = adev->num_ip_blocks - 1; i >= 0; i--) { + if (!(adev->ip_blocks[i].version->type == + AMD_IP_BLOCK_TYPE_GFX || + adev->ip_blocks[i].version->type == + AMD_IP_BLOCK_TYPE_SDMA)) + continue; + + r = adev->ip_blocks[i].version->funcs->suspend(adev); + + if (r) { + dev_err(adev->dev, + "suspend of IP block <%s> failed %d\n", + adev->ip_blocks[i].version->funcs->name, r); + return r; + } + adev->ip_blocks[i].status.hw = false; + } + + return r; +} + +static int +sienna_cichlid_mode2_prepare_hwcontext(struct amdgpu_reset_control *reset_ctl, + struct amdgpu_reset_context *reset_context) +{ + int r = 0; + struct amdgpu_device *adev = (struct amdgpu_device *)reset_ctl->handle; + + if (!amdgpu_sriov_vf(adev)) { + if (adev->gfxhub.funcs->mode2_save_regs) + adev->gfxhub.funcs->mode2_save_regs(adev); + if (adev->gfxhub.funcs->halt) + adev->gfxhub.funcs->halt(adev); + r = sienna_cichlid_mode2_suspend_ip(adev); + } + + return r; +} + +static void sienna_cichlid_async_reset(struct work_struct *work) +{ + struct amdgpu_reset_handler *handler; + struct amdgpu_reset_control *reset_ctl = + container_of(work, struct amdgpu_reset_control, reset_work); + struct amdgpu_device *adev = (struct amdgpu_device *)reset_ctl->handle; + + list_for_each_entry(handler, &reset_ctl->reset_handlers, + handler_list) { + if (handler->reset_method == reset_ctl->active_reset) { + dev_dbg(adev->dev, "Resetting device\n"); + handler->do_reset(adev); + break; + } + } +} + +static int sienna_cichlid_mode2_reset(struct amdgpu_device *adev) +{ + /* disable BM */ + pci_clear_master(adev->pdev); + return amdgpu_dpm_mode2_reset(adev); +} + +static int +sienna_cichlid_mode2_perform_reset(struct amdgpu_reset_control *reset_ctl, + struct amdgpu_reset_context *reset_context) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)reset_ctl->handle; + int r; + + r = sienna_cichlid_mode2_reset(adev); + if (r) { + dev_err(adev->dev, + "ASIC reset failed with error, %d ", r); + } + return r; +} + +static int sienna_cichlid_mode2_restore_ip(struct amdgpu_device *adev) +{ + int i, r; + struct psp_context *psp = &adev->psp; + + r = psp_rlc_autoload_start(psp); + if (r) { + dev_err(adev->dev, "Failed to start rlc autoload\n"); + return r; + } + + /* Reinit GFXHUB */ + if (adev->gfxhub.funcs->mode2_restore_regs) + adev->gfxhub.funcs->mode2_restore_regs(adev); + adev->gfxhub.funcs->init(adev); + r = adev->gfxhub.funcs->gart_enable(adev); + if (r) { + dev_err(adev->dev, "GFXHUB gart reenable failed after reset\n"); + return r; + } + + for (i = 0; i < adev->num_ip_blocks; i++) { + if (adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_IH) { + r = adev->ip_blocks[i].version->funcs->resume(adev); + if (r) { + dev_err(adev->dev, + "resume of IP block <%s> failed %d\n", + adev->ip_blocks[i].version->funcs->name, r); + return r; + } + + adev->ip_blocks[i].status.hw = true; + } + } + + for (i = 0; i < adev->num_ip_blocks; i++) { + if (!(adev->ip_blocks[i].version->type == + AMD_IP_BLOCK_TYPE_GFX || + adev->ip_blocks[i].version->type == + AMD_IP_BLOCK_TYPE_SDMA)) + continue; + r = adev->ip_blocks[i].version->funcs->resume(adev); + if (r) { + dev_err(adev->dev, + "resume of IP block <%s> failed %d\n", + adev->ip_blocks[i].version->funcs->name, r); + return r; + } + + adev->ip_blocks[i].status.hw = true; + } + + for (i = 0; i < adev->num_ip_blocks; i++) { + if (!(adev->ip_blocks[i].version->type == + AMD_IP_BLOCK_TYPE_GFX || + adev->ip_blocks[i].version->type == + AMD_IP_BLOCK_TYPE_SDMA)) + continue; + + if (adev->ip_blocks[i].version->funcs->late_init) { + r = adev->ip_blocks[i].version->funcs->late_init( + (void *)adev); + if (r) { + dev_err(adev->dev, + "late_init of IP block <%s> failed %d after reset\n", + adev->ip_blocks[i].version->funcs->name, + r); + return r; + } + } + adev->ip_blocks[i].status.late_initialized = true; + } + + amdgpu_device_set_cg_state(adev, AMD_CG_STATE_GATE); + amdgpu_device_set_pg_state(adev, AMD_PG_STATE_GATE); + + return r; +} + +static int +sienna_cichlid_mode2_restore_hwcontext(struct amdgpu_reset_control *reset_ctl, + struct amdgpu_reset_context *reset_context) +{ + int r; + struct amdgpu_device *tmp_adev = (struct amdgpu_device *)reset_ctl->handle; + + dev_info(tmp_adev->dev, + "GPU reset succeeded, trying to resume\n"); + r = sienna_cichlid_mode2_restore_ip(tmp_adev); + if (r) + goto end; + + /* + * Add this ASIC as tracked as reset was already + * complete successfully. + */ + amdgpu_register_gpu_instance(tmp_adev); + + /* Resume RAS */ + amdgpu_ras_resume(tmp_adev); + + amdgpu_irq_gpu_reset_resume_helper(tmp_adev); + + r = amdgpu_ib_ring_tests(tmp_adev); + if (r) { + dev_err(tmp_adev->dev, + "ib ring test failed (%d).\n", r); + r = -EAGAIN; + goto end; + } + +end: + if (r) + return -EAGAIN; + else + return r; +} + +static struct amdgpu_reset_handler sienna_cichlid_mode2_handler = { + .reset_method = AMD_RESET_METHOD_MODE2, + .prepare_env = NULL, + .prepare_hwcontext = sienna_cichlid_mode2_prepare_hwcontext, + .perform_reset = sienna_cichlid_mode2_perform_reset, + .restore_hwcontext = sienna_cichlid_mode2_restore_hwcontext, + .restore_env = NULL, + .do_reset = sienna_cichlid_mode2_reset, +}; + +int sienna_cichlid_reset_init(struct amdgpu_device *adev) +{ + struct amdgpu_reset_control *reset_ctl; + + reset_ctl = kzalloc(sizeof(*reset_ctl), GFP_KERNEL); + if (!reset_ctl) + return -ENOMEM; + + reset_ctl->handle = adev; + reset_ctl->async_reset = sienna_cichlid_async_reset; + reset_ctl->active_reset = AMD_RESET_METHOD_NONE; + reset_ctl->get_reset_handler = sienna_cichlid_get_reset_handler; + + INIT_LIST_HEAD(&reset_ctl->reset_handlers); + INIT_WORK(&reset_ctl->reset_work, reset_ctl->async_reset); + /* Only mode2 is handled through reset control now */ + amdgpu_reset_add_handler(reset_ctl, &sienna_cichlid_mode2_handler); + + adev->reset_cntl = reset_ctl; + + return 0; +} + +int sienna_cichlid_reset_fini(struct amdgpu_device *adev) +{ + kfree(adev->reset_cntl); + adev->reset_cntl = NULL; + return 0; +} diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_hdp.c b/drivers/gpu/drm/amd/amdgpu/sienna_cichlid.h index a766e1aad2b9..5213b162dacd 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_hdp.c +++ b/drivers/gpu/drm/amd/amdgpu/sienna_cichlid.h @@ -21,48 +21,12 @@ * */ -#include "amdgpu.h" -#include "amdgpu_ras.h" - -int amdgpu_hdp_ras_late_init(struct amdgpu_device *adev) -{ - int r; - struct ras_ih_if ih_info = { - .cb = NULL, - }; - struct ras_fs_if fs_info = { - .sysfs_name = "hdp_err_count", - }; +#ifndef __SIENNA_CICHLID_H__ +#define __SIENNA_CICHLID_H__ - if (!adev->hdp.ras_if) { - adev->hdp.ras_if = kmalloc(sizeof(struct ras_common_if), GFP_KERNEL); - if (!adev->hdp.ras_if) - return -ENOMEM; - adev->hdp.ras_if->block = AMDGPU_RAS_BLOCK__HDP; - adev->hdp.ras_if->type = AMDGPU_RAS_ERROR__MULTI_UNCORRECTABLE; - adev->hdp.ras_if->sub_block_index = 0; - } - ih_info.head = fs_info.head = *adev->hdp.ras_if; - r = amdgpu_ras_late_init(adev, adev->hdp.ras_if, - &fs_info, &ih_info); - if (r || !amdgpu_ras_is_supported(adev, adev->hdp.ras_if->block)) { - kfree(adev->hdp.ras_if); - adev->hdp.ras_if = NULL; - } - - return r; -} +#include "amdgpu.h" -void amdgpu_hdp_ras_fini(struct amdgpu_device *adev) -{ - if (amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__HDP) && - adev->hdp.ras_if) { - struct ras_common_if *ras_if = adev->hdp.ras_if; - struct ras_ih_if ih_info = { - .cb = NULL, - }; +int sienna_cichlid_reset_init(struct amdgpu_device *adev); +int sienna_cichlid_reset_fini(struct amdgpu_device *adev); - amdgpu_ras_late_fini(adev, ras_if, &ih_info); - kfree(ras_if); - } -} +#endif diff --git a/drivers/gpu/drm/amd/amdgpu/smu_v11_0_i2c.c b/drivers/gpu/drm/amd/amdgpu/smu_v11_0_i2c.c index 73ffa8fde3df..dd2d66090d23 100644 --- a/drivers/gpu/drm/amd/amdgpu/smu_v11_0_i2c.c +++ b/drivers/gpu/drm/amd/amdgpu/smu_v11_0_i2c.c @@ -26,6 +26,7 @@ #include "smu_v11_0_i2c.h" #include "amdgpu.h" +#include "amdgpu_dpm.h" #include "soc15_common.h" #include <drm/drm_fixed.h> #include <drm/drm_drv.h> @@ -43,11 +44,10 @@ #define I2C_X_RESTART BIT(31) -#define to_amdgpu_device(x) (container_of(x, struct amdgpu_device, pm.smu_i2c)) - static void smu_v11_0_i2c_set_clock_gating(struct i2c_adapter *control, bool en) { - struct amdgpu_device *adev = to_amdgpu_device(control); + struct amdgpu_smu_i2c_bus *smu_i2c = i2c_get_adapdata(control); + struct amdgpu_device *adev = smu_i2c->adev; uint32_t reg = RREG32_SOC15(SMUIO, 0, mmSMUIO_PWRMGT); reg = REG_SET_FIELD(reg, SMUIO_PWRMGT, i2c_clk_gate_en, en ? 1 : 0); @@ -75,7 +75,8 @@ static void smu_v11_0_i2c_set_clock_gating(struct i2c_adapter *control, bool en) static int smu_v11_0_i2c_enable(struct i2c_adapter *control, bool enable) { - struct amdgpu_device *adev = to_amdgpu_device(control); + struct amdgpu_smu_i2c_bus *smu_i2c = i2c_get_adapdata(control); + struct amdgpu_device *adev = smu_i2c->adev; WREG32_SOC15(SMUIO, 0, mmCKSVII2C_IC_ENABLE, enable ? 1 : 0); @@ -100,7 +101,8 @@ static int smu_v11_0_i2c_enable(struct i2c_adapter *control, bool enable) static void smu_v11_0_i2c_clear_status(struct i2c_adapter *control) { - struct amdgpu_device *adev = to_amdgpu_device(control); + struct amdgpu_smu_i2c_bus *smu_i2c = i2c_get_adapdata(control); + struct amdgpu_device *adev = smu_i2c->adev; /* do */ { RREG32_SOC15(SMUIO, 0, mmCKSVII2C_IC_CLR_INTR); @@ -110,7 +112,8 @@ static void smu_v11_0_i2c_clear_status(struct i2c_adapter *control) static void smu_v11_0_i2c_configure(struct i2c_adapter *control) { - struct amdgpu_device *adev = to_amdgpu_device(control); + struct amdgpu_smu_i2c_bus *smu_i2c = i2c_get_adapdata(control); + struct amdgpu_device *adev = smu_i2c->adev; uint32_t reg = 0; reg = REG_SET_FIELD(reg, CKSVII2C_IC_CON, IC_SLAVE_DISABLE, 1); @@ -131,7 +134,8 @@ static void smu_v11_0_i2c_configure(struct i2c_adapter *control) static void smu_v11_0_i2c_set_clock(struct i2c_adapter *control) { - struct amdgpu_device *adev = to_amdgpu_device(control); + struct amdgpu_smu_i2c_bus *smu_i2c = i2c_get_adapdata(control); + struct amdgpu_device *adev = smu_i2c->adev; /* * Standard mode speed, These values are taken from SMUIO MAS, @@ -154,7 +158,8 @@ static void smu_v11_0_i2c_set_clock(struct i2c_adapter *control) static void smu_v11_0_i2c_set_address(struct i2c_adapter *control, u16 address) { - struct amdgpu_device *adev = to_amdgpu_device(control); + struct amdgpu_smu_i2c_bus *smu_i2c = i2c_get_adapdata(control); + struct amdgpu_device *adev = smu_i2c->adev; /* The IC_TAR::IC_TAR field is 10-bits wide. * It takes a 7-bit or 10-bit addresses as an address, @@ -165,7 +170,8 @@ static void smu_v11_0_i2c_set_address(struct i2c_adapter *control, u16 address) static uint32_t smu_v11_0_i2c_poll_tx_status(struct i2c_adapter *control) { - struct amdgpu_device *adev = to_amdgpu_device(control); + struct amdgpu_smu_i2c_bus *smu_i2c = i2c_get_adapdata(control); + struct amdgpu_device *adev = smu_i2c->adev; uint32_t ret = I2C_OK; uint32_t reg, reg_c_tx_abrt_source; @@ -216,7 +222,8 @@ static uint32_t smu_v11_0_i2c_poll_tx_status(struct i2c_adapter *control) static uint32_t smu_v11_0_i2c_poll_rx_status(struct i2c_adapter *control) { - struct amdgpu_device *adev = to_amdgpu_device(control); + struct amdgpu_smu_i2c_bus *smu_i2c = i2c_get_adapdata(control); + struct amdgpu_device *adev = smu_i2c->adev; uint32_t ret = I2C_OK; uint32_t reg_ic_status, reg_c_tx_abrt_source; @@ -262,7 +269,8 @@ static uint32_t smu_v11_0_i2c_transmit(struct i2c_adapter *control, u16 address, u8 *data, u32 numbytes, u32 i2c_flag) { - struct amdgpu_device *adev = to_amdgpu_device(control); + struct amdgpu_smu_i2c_bus *smu_i2c = i2c_get_adapdata(control); + struct amdgpu_device *adev = smu_i2c->adev; u32 bytes_sent, reg, ret = I2C_OK; unsigned long timeout_counter; @@ -360,7 +368,8 @@ static uint32_t smu_v11_0_i2c_receive(struct i2c_adapter *control, u16 address, u8 *data, u32 numbytes, u32 i2c_flag) { - struct amdgpu_device *adev = to_amdgpu_device(control); + struct amdgpu_smu_i2c_bus *smu_i2c = i2c_get_adapdata(control); + struct amdgpu_device *adev = smu_i2c->adev; uint32_t bytes_received, ret = I2C_OK; bytes_received = 0; @@ -431,7 +440,8 @@ static uint32_t smu_v11_0_i2c_receive(struct i2c_adapter *control, static void smu_v11_0_i2c_abort(struct i2c_adapter *control) { - struct amdgpu_device *adev = to_amdgpu_device(control); + struct amdgpu_smu_i2c_bus *smu_i2c = i2c_get_adapdata(control); + struct amdgpu_device *adev = smu_i2c->adev; uint32_t reg = 0; /* Enable I2C engine; */ @@ -447,7 +457,8 @@ static void smu_v11_0_i2c_abort(struct i2c_adapter *control) static bool smu_v11_0_i2c_activity_done(struct i2c_adapter *control) { - struct amdgpu_device *adev = to_amdgpu_device(control); + struct amdgpu_smu_i2c_bus *smu_i2c = i2c_get_adapdata(control); + struct amdgpu_device *adev = smu_i2c->adev; const uint32_t IDLE_TIMEOUT = 1024; uint32_t timeout_count = 0; @@ -508,7 +519,8 @@ static void smu_v11_0_i2c_init(struct i2c_adapter *control) static void smu_v11_0_i2c_fini(struct i2c_adapter *control) { - struct amdgpu_device *adev = to_amdgpu_device(control); + struct amdgpu_smu_i2c_bus *smu_i2c = i2c_get_adapdata(control); + struct amdgpu_device *adev = smu_i2c->adev; u32 status, enable, en_stat; int res; @@ -543,7 +555,8 @@ static void smu_v11_0_i2c_fini(struct i2c_adapter *control) static bool smu_v11_0_i2c_bus_lock(struct i2c_adapter *control) { - struct amdgpu_device *adev = to_amdgpu_device(control); + struct amdgpu_smu_i2c_bus *smu_i2c = i2c_get_adapdata(control); + struct amdgpu_device *adev = smu_i2c->adev; /* Send PPSMC_MSG_RequestI2CBus */ if (!amdgpu_dpm_smu_i2c_bus_access(adev, true)) @@ -554,7 +567,8 @@ static bool smu_v11_0_i2c_bus_lock(struct i2c_adapter *control) static bool smu_v11_0_i2c_bus_unlock(struct i2c_adapter *control) { - struct amdgpu_device *adev = to_amdgpu_device(control); + struct amdgpu_smu_i2c_bus *smu_i2c = i2c_get_adapdata(control); + struct amdgpu_device *adev = smu_i2c->adev; /* Send PPSMC_MSG_ReleaseI2CBus */ if (!amdgpu_dpm_smu_i2c_bus_access(adev, false)) @@ -587,16 +601,17 @@ static uint32_t smu_v11_0_i2c_write_data(struct i2c_adapter *control, if (ret != I2C_OK) DRM_ERROR("WriteI2CData() - I2C error occurred :%x", ret); - + return ret; } static void lock_bus(struct i2c_adapter *i2c, unsigned int flags) { - struct amdgpu_device *adev = to_amdgpu_device(i2c); + struct amdgpu_smu_i2c_bus *smu_i2c = i2c_get_adapdata(i2c); + struct amdgpu_device *adev = smu_i2c->adev; - mutex_lock(&adev->pm.smu_i2c_mutex); + mutex_lock(&smu_i2c->mutex); if (!smu_v11_0_i2c_bus_lock(i2c)) DRM_ERROR("Failed to lock the bus from SMU"); else @@ -611,13 +626,14 @@ static int trylock_bus(struct i2c_adapter *i2c, unsigned int flags) static void unlock_bus(struct i2c_adapter *i2c, unsigned int flags) { - struct amdgpu_device *adev = to_amdgpu_device(i2c); + struct amdgpu_smu_i2c_bus *smu_i2c = i2c_get_adapdata(i2c); + struct amdgpu_device *adev = smu_i2c->adev; if (!smu_v11_0_i2c_bus_unlock(i2c)) DRM_ERROR("Failed to unlock the bus from SMU"); else adev->pm.bus_locked = false; - mutex_unlock(&adev->pm.smu_i2c_mutex); + mutex_unlock(&smu_i2c->mutex); } static const struct i2c_lock_operations smu_v11_0_i2c_i2c_lock_ops = { @@ -706,19 +722,26 @@ static const struct i2c_adapter_quirks smu_v11_0_i2c_control_quirks = { .flags = I2C_AQ_NO_ZERO_LEN, }; -int smu_v11_0_i2c_control_init(struct i2c_adapter *control) +int smu_v11_0_i2c_control_init(struct amdgpu_device *adev) { - struct amdgpu_device *adev = to_amdgpu_device(control); + struct amdgpu_smu_i2c_bus *smu_i2c = &adev->pm.smu_i2c[0]; + struct i2c_adapter *control = &smu_i2c->adapter; int res; - mutex_init(&adev->pm.smu_i2c_mutex); + smu_i2c->adev = adev; + smu_i2c->port = 0; + mutex_init(&smu_i2c->mutex); control->owner = THIS_MODULE; control->class = I2C_CLASS_HWMON; control->dev.parent = &adev->pdev->dev; control->algo = &smu_v11_0_i2c_algo; - snprintf(control->name, sizeof(control->name), "AMDGPU SMU"); + snprintf(control->name, sizeof(control->name), "AMDGPU SMU 0"); control->lock_ops = &smu_v11_0_i2c_i2c_lock_ops; control->quirks = &smu_v11_0_i2c_control_quirks; + i2c_set_adapdata(control, smu_i2c); + + adev->pm.ras_eeprom_i2c_bus = &adev->pm.smu_i2c[0].adapter; + adev->pm.fru_eeprom_i2c_bus = &adev->pm.smu_i2c[0].adapter; res = i2c_add_adapter(control); if (res) @@ -727,9 +750,13 @@ int smu_v11_0_i2c_control_init(struct i2c_adapter *control) return res; } -void smu_v11_0_i2c_control_fini(struct i2c_adapter *control) +void smu_v11_0_i2c_control_fini(struct amdgpu_device *adev) { + struct i2c_adapter *control = adev->pm.ras_eeprom_i2c_bus; + i2c_del_adapter(control); + adev->pm.ras_eeprom_i2c_bus = NULL; + adev->pm.fru_eeprom_i2c_bus = NULL; } /* diff --git a/drivers/gpu/drm/amd/amdgpu/smu_v11_0_i2c.h b/drivers/gpu/drm/amd/amdgpu/smu_v11_0_i2c.h index 44467c05f642..96ad14288a0c 100644 --- a/drivers/gpu/drm/amd/amdgpu/smu_v11_0_i2c.h +++ b/drivers/gpu/drm/amd/amdgpu/smu_v11_0_i2c.h @@ -26,9 +26,9 @@ #include <linux/types.h> -struct i2c_adapter; +struct amdgpu_device; -int smu_v11_0_i2c_control_init(struct i2c_adapter *control); -void smu_v11_0_i2c_control_fini(struct i2c_adapter *control); +int smu_v11_0_i2c_control_init(struct amdgpu_device *adev); +void smu_v11_0_i2c_control_fini(struct amdgpu_device *adev); #endif diff --git a/drivers/gpu/drm/amd/amdgpu/smuio_v11_0.c b/drivers/gpu/drm/amd/amdgpu/smuio_v11_0.c index b6f1322f908c..acdc40f99ab3 100644 --- a/drivers/gpu/drm/amd/amdgpu/smuio_v11_0.c +++ b/drivers/gpu/drm/amd/amdgpu/smuio_v11_0.c @@ -59,7 +59,7 @@ static void smuio_v11_0_update_rom_clock_gating(struct amdgpu_device *adev, bool WREG32_SOC15(SMUIO, 0, mmCGTT_ROM_CLK_CTRL0, data); } -static void smuio_v11_0_get_clock_gating_state(struct amdgpu_device *adev, u32 *flags) +static void smuio_v11_0_get_clock_gating_state(struct amdgpu_device *adev, u64 *flags) { u32 data; diff --git a/drivers/gpu/drm/amd/amdgpu/smuio_v11_0_6.c b/drivers/gpu/drm/amd/amdgpu/smuio_v11_0_6.c index 3a18dbb55c32..2afeb8b37f62 100644 --- a/drivers/gpu/drm/amd/amdgpu/smuio_v11_0_6.c +++ b/drivers/gpu/drm/amd/amdgpu/smuio_v11_0_6.c @@ -56,7 +56,7 @@ static void smuio_v11_0_6_update_rom_clock_gating(struct amdgpu_device *adev, bo WREG32_SOC15(SMUIO, 0, mmCGTT_ROM_CLK_CTRL0, data); } -static void smuio_v11_0_6_get_clock_gating_state(struct amdgpu_device *adev, u32 *flags) +static void smuio_v11_0_6_get_clock_gating_state(struct amdgpu_device *adev, u64 *flags) { u32 data; diff --git a/drivers/gpu/drm/amd/amdgpu/smuio_v13_0.c b/drivers/gpu/drm/amd/amdgpu/smuio_v13_0.c index 39b7c206770f..13e905c22592 100644 --- a/drivers/gpu/drm/amd/amdgpu/smuio_v13_0.c +++ b/drivers/gpu/drm/amd/amdgpu/smuio_v13_0.c @@ -58,7 +58,7 @@ static void smuio_v13_0_update_rom_clock_gating(struct amdgpu_device *adev, bool WREG32_SOC15(SMUIO, 0, regCGTT_ROM_CLK_CTRL0, data); } -static void smuio_v13_0_get_clock_gating_state(struct amdgpu_device *adev, u32 *flags) +static void smuio_v13_0_get_clock_gating_state(struct amdgpu_device *adev, u64 *flags) { u32 data; diff --git a/drivers/gpu/drm/amd/amdgpu/smuio_v13_0_6.c b/drivers/gpu/drm/amd/amdgpu/smuio_v13_0_6.c new file mode 100644 index 000000000000..de998e328b08 --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/smuio_v13_0_6.c @@ -0,0 +1,41 @@ +/* + * Copyright 2022 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ +#include "amdgpu.h" +#include "smuio_v13_0_6.h" +#include "smuio/smuio_13_0_6_offset.h" +#include "smuio/smuio_13_0_6_sh_mask.h" + +static u32 smuio_v13_0_6_get_rom_index_offset(struct amdgpu_device *adev) +{ + return SOC15_REG_OFFSET(SMUIO, 0, regROM_INDEX); +} + +static u32 smuio_v13_0_6_get_rom_data_offset(struct amdgpu_device *adev) +{ + return SOC15_REG_OFFSET(SMUIO, 0, regROM_DATA); +} + +const struct amdgpu_smuio_funcs smuio_v13_0_6_funcs = { + .get_rom_index_offset = smuio_v13_0_6_get_rom_index_offset, + .get_rom_data_offset = smuio_v13_0_6_get_rom_data_offset, +}; diff --git a/drivers/gpu/drm/amd/amdgpu/smuio_v13_0_6.h b/drivers/gpu/drm/amd/amdgpu/smuio_v13_0_6.h new file mode 100644 index 000000000000..c75621de5ab5 --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/smuio_v13_0_6.h @@ -0,0 +1,30 @@ +/* + * Copyright 2022 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ +#ifndef __SMUIO_V13_0_6_H__ +#define __SMUIO_V13_0_6_H__ + +#include "soc15_common.h" + +extern const struct amdgpu_smuio_funcs smuio_v13_0_6_funcs; + +#endif /* __SMUIO_V13_0_6_H__ */ diff --git a/drivers/gpu/drm/amd/amdgpu/smuio_v9_0.c b/drivers/gpu/drm/amd/amdgpu/smuio_v9_0.c index 8417890af227..e4e30b9d481b 100644 --- a/drivers/gpu/drm/amd/amdgpu/smuio_v9_0.c +++ b/drivers/gpu/drm/amd/amdgpu/smuio_v9_0.c @@ -56,7 +56,7 @@ static void smuio_v9_0_update_rom_clock_gating(struct amdgpu_device *adev, bool WREG32_SOC15(SMUIO, 0, mmCGTT_ROM_CLK_CTRL0, data); } -static void smuio_v9_0_get_clock_gating_state(struct amdgpu_device *adev, u32 *flags) +static void smuio_v9_0_get_clock_gating_state(struct amdgpu_device *adev, u64 *flags) { u32 data; diff --git a/drivers/gpu/drm/amd/amdgpu/soc15.c b/drivers/gpu/drm/amd/amdgpu/soc15.c index 0fc1747e4a70..e3b2b6b4f1a6 100644 --- a/drivers/gpu/drm/amd/amdgpu/soc15.c +++ b/drivers/gpu/drm/amd/amdgpu/soc15.c @@ -375,39 +375,6 @@ static bool soc15_read_disabled_bios(struct amdgpu_device *adev) return false; } -static bool soc15_read_bios_from_rom(struct amdgpu_device *adev, - u8 *bios, u32 length_bytes) -{ - u32 *dw_ptr; - u32 i, length_dw; - uint32_t rom_index_offset; - uint32_t rom_data_offset; - - if (bios == NULL) - return false; - if (length_bytes == 0) - return false; - /* APU vbios image is part of sbios image */ - if (adev->flags & AMD_IS_APU) - return false; - - dw_ptr = (u32 *)bios; - length_dw = ALIGN(length_bytes, 4) / 4; - - rom_index_offset = - adev->smuio.funcs->get_rom_index_offset(adev); - rom_data_offset = - adev->smuio.funcs->get_rom_data_offset(adev); - - /* set rom index to 0 */ - WREG32(rom_index_offset, 0); - /* read out the rom data */ - for (i = 0; i < length_dw; i++) - dw_ptr[i] = RREG32(rom_data_offset); - - return true; -} - static struct soc15_allowed_register_entry soc15_allowed_read_registers[] = { { SOC15_REG_ENTRY(GC, 0, mmGRBM_STATUS)}, { SOC15_REG_ENTRY(GC, 0, mmGRBM_STATUS2)}, @@ -619,8 +586,8 @@ soc15_asic_reset_method(struct amdgpu_device *adev) static int soc15_asic_reset(struct amdgpu_device *adev) { /* original raven doesn't have full asic reset */ - if ((adev->apu_flags & AMD_APU_IS_RAVEN) && - !(adev->apu_flags & AMD_APU_IS_RAVEN2)) + if ((adev->apu_flags & AMD_APU_IS_RAVEN) || + (adev->apu_flags & AMD_APU_IS_RAVEN2)) return 0; switch (soc15_asic_reset_method(adev)) { @@ -703,7 +670,7 @@ static void soc15_pcie_gen3_enable(struct amdgpu_device *adev) static void soc15_program_aspm(struct amdgpu_device *adev) { - if (!amdgpu_aspm) + if (!amdgpu_device_should_use_aspm(adev)) return; if (!(adev->flags & AMD_IS_APU) && @@ -734,25 +701,12 @@ static uint32_t soc15_get_rev_id(struct amdgpu_device *adev) static void soc15_reg_base_init(struct amdgpu_device *adev) { - int r; - /* Set IP register base before any HW register access */ switch (adev->asic_type) { case CHIP_VEGA10: case CHIP_VEGA12: case CHIP_RAVEN: - vega10_reg_base_init(adev); - break; case CHIP_RENOIR: - /* It's safe to do ip discovery here for Renoir, - * it doesn't support SRIOV. */ - if (amdgpu_discovery) { - r = amdgpu_discovery_reg_base_init(adev); - if (r == 0) - break; - DRM_WARN("failed to init reg base from ip discovery table, " - "fallback to legacy init method\n"); - } vega10_reg_base_init(adev); break; case CHIP_VEGA20: @@ -886,6 +840,10 @@ static bool soc15_need_reset_on_init(struct amdgpu_device *adev) { u32 sol_reg; + /* CP hangs in IGT reloading test on RN, reset to WA */ + if (adev->asic_type == CHIP_RENOIR) + return true; + /* Just return false for soc15 GPUs. Reset does not seem to * be necessary. */ @@ -925,7 +883,7 @@ static void soc15_pre_asic_init(struct amdgpu_device *adev) static const struct amdgpu_asic_funcs soc15_asic_funcs = { .read_disabled_bios = &soc15_read_disabled_bios, - .read_bios_from_rom = &soc15_read_bios_from_rom, + .read_bios_from_rom = &amdgpu_soc15_read_bios_from_rom, .read_register = &soc15_read_register, .reset = &soc15_asic_reset, .reset_method = &soc15_asic_reset_method, @@ -947,7 +905,7 @@ static const struct amdgpu_asic_funcs soc15_asic_funcs = static const struct amdgpu_asic_funcs vega20_asic_funcs = { .read_disabled_bios = &soc15_read_disabled_bios, - .read_bios_from_rom = &soc15_read_bios_from_rom, + .read_bios_from_rom = &amdgpu_soc15_read_bios_from_rom, .read_register = &soc15_read_register, .reset = &soc15_asic_reset, .reset_method = &soc15_asic_reset_method, @@ -1114,8 +1072,11 @@ static int soc15_common_early_init(void *handle) AMD_CG_SUPPORT_SDMA_LS | AMD_CG_SUPPORT_VCN_MGCG; + /* + * MMHUB PG needs to be disabled for Picasso for + * stability reasons. + */ adev->pg_flags = AMD_PG_SUPPORT_SDMA | - AMD_PG_SUPPORT_MMHUB | AMD_PG_SUPPORT_VCN; } else { adev->cg_flags = AMD_CG_SUPPORT_GFX_MGCG | @@ -1219,16 +1180,11 @@ static int soc15_common_early_init(void *handle) static int soc15_common_late_init(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; - int r = 0; if (amdgpu_sriov_vf(adev)) xgpu_ai_mailbox_get_irq(adev); - if (adev->nbio.ras_funcs && - adev->nbio.ras_funcs->ras_late_init) - r = adev->nbio.ras_funcs->ras_late_init(adev); - - return r; + return 0; } static int soc15_common_sw_init(void *handle) @@ -1249,32 +1205,23 @@ static int soc15_common_sw_fini(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; - if (adev->nbio.ras_funcs && - adev->nbio.ras_funcs->ras_fini) - adev->nbio.ras_funcs->ras_fini(adev); - if (adev->df.funcs && adev->df.funcs->sw_fini) adev->df.funcs->sw_fini(adev); return 0; } -static void soc15_doorbell_range_init(struct amdgpu_device *adev) +static void soc15_sdma_doorbell_range_init(struct amdgpu_device *adev) { int i; - struct amdgpu_ring *ring; - /* sdma/ih doorbell range are programed by hypervisor */ + /* sdma doorbell range is programed by hypervisor */ if (!amdgpu_sriov_vf(adev)) { for (i = 0; i < adev->sdma.num_instances; i++) { - ring = &adev->sdma.instance[i].ring; adev->nbio.funcs->sdma_doorbell_range(adev, i, - ring->use_doorbell, ring->doorbell_index, + true, adev->doorbell_index.sdma_engine[i] << 1, adev->doorbell_index.sdma_doorbell_range); } - - adev->nbio.funcs->ih_doorbell_range(adev, adev->irq.ih.use_doorbell, - adev->irq.ih.doorbell_index); } } @@ -1299,10 +1246,11 @@ static int soc15_common_hw_init(void *handle) soc15_enable_doorbell_aperture(adev, true); /* HW doorbell routing policy: doorbell writing not * in SDMA/IH/MM/ACV range will be routed to CP. So - * we need to init SDMA/IH/MM/ACV doorbell range prior - * to CP ip block init and ring test. + * we need to init SDMA doorbell range prior + * to CP ip block init and ring test. IH already + * happens before CP. */ - soc15_doorbell_range_init(adev); + soc15_sdma_doorbell_range_init(adev); return 0; } @@ -1318,11 +1266,11 @@ static int soc15_common_hw_fini(void *handle) if (adev->nbio.ras_if && amdgpu_ras_is_supported(adev, adev->nbio.ras_if->block)) { - if (adev->nbio.ras_funcs && - adev->nbio.ras_funcs->init_ras_controller_interrupt) + if (adev->nbio.ras && + adev->nbio.ras->init_ras_controller_interrupt) amdgpu_irq_put(adev, &adev->nbio.ras_controller_irq, 0); - if (adev->nbio.ras_funcs && - adev->nbio.ras_funcs->init_ras_err_event_athub_interrupt) + if (adev->nbio.ras && + adev->nbio.ras->init_ras_err_event_athub_interrupt) amdgpu_irq_put(adev, &adev->nbio.ras_err_event_athub_irq, 0); } @@ -1454,7 +1402,7 @@ static int soc15_common_set_clockgating_state(void *handle, return 0; } -static void soc15_common_get_clockgating_state(void *handle, u32 *flags) +static void soc15_common_get_clockgating_state(void *handle, u64 *flags) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; int data; diff --git a/drivers/gpu/drm/amd/amdgpu/soc15_common.h b/drivers/gpu/drm/amd/amdgpu/soc15_common.h index 473767e03676..9fefd403e14f 100644 --- a/drivers/gpu/drm/amd/amdgpu/soc15_common.h +++ b/drivers/gpu/drm/amd/amdgpu/soc15_common.h @@ -28,13 +28,13 @@ #define SOC15_REG_OFFSET(ip, inst, reg) (adev->reg_offset[ip##_HWIP][inst][reg##_BASE_IDX] + reg) #define __WREG32_SOC15_RLC__(reg, value, flag, hwip) \ - ((amdgpu_sriov_vf(adev) && adev->gfx.rlc.funcs && adev->gfx.rlc.funcs->sriov_wreg) ? \ - adev->gfx.rlc.funcs->sriov_wreg(adev, reg, value, flag, hwip) : \ + ((amdgpu_sriov_vf(adev) && adev->gfx.rlc.funcs && adev->gfx.rlc.rlcg_reg_access_supported) ? \ + amdgpu_sriov_wreg(adev, reg, value, flag, hwip) : \ WREG32(reg, value)) #define __RREG32_SOC15_RLC__(reg, flag, hwip) \ - ((amdgpu_sriov_vf(adev) && adev->gfx.rlc.funcs && adev->gfx.rlc.funcs->sriov_rreg) ? \ - adev->gfx.rlc.funcs->sriov_rreg(adev, reg, flag, hwip) : \ + ((amdgpu_sriov_vf(adev) && adev->gfx.rlc.funcs && adev->gfx.rlc.rlcg_reg_access_supported) ? \ + amdgpu_sriov_rreg(adev, reg, flag, hwip) : \ RREG32(reg)) #define WREG32_FIELD15(ip, idx, reg, field, val) \ @@ -45,6 +45,14 @@ ~REG_FIELD_MASK(reg, field)) | (val) << REG_FIELD_SHIFT(reg, field), \ 0, ip##_HWIP) +#define WREG32_FIELD15_PREREG(ip, idx, reg_name, field, val) \ + __WREG32_SOC15_RLC__(adev->reg_offset[ip##_HWIP][idx][reg##reg_name##_BASE_IDX] + reg##reg_name, \ + (__RREG32_SOC15_RLC__( \ + adev->reg_offset[ip##_HWIP][idx][reg##reg_name##_BASE_IDX] + reg##reg_name, \ + 0, ip##_HWIP) & \ + ~REG_FIELD_MASK(reg_name, field)) | (val) << REG_FIELD_SHIFT(reg_name, field), \ + 0, ip##_HWIP) + #define RREG32_SOC15(ip, inst, reg) \ __RREG32_SOC15_RLC__(adev->reg_offset[ip##_HWIP][inst][reg##_BASE_IDX] + reg, \ 0, ip##_HWIP) diff --git a/drivers/gpu/drm/amd/amdgpu/soc21.c b/drivers/gpu/drm/amd/amdgpu/soc21.c new file mode 100644 index 000000000000..e08044008186 --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/soc21.c @@ -0,0 +1,806 @@ +/* + * Copyright 2021 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ +#include <linux/firmware.h> +#include <linux/slab.h> +#include <linux/module.h> +#include <linux/pci.h> + +#include "amdgpu.h" +#include "amdgpu_atombios.h" +#include "amdgpu_ih.h" +#include "amdgpu_uvd.h" +#include "amdgpu_vce.h" +#include "amdgpu_ucode.h" +#include "amdgpu_psp.h" +#include "amdgpu_smu.h" +#include "atom.h" +#include "amd_pcie.h" + +#include "gc/gc_11_0_0_offset.h" +#include "gc/gc_11_0_0_sh_mask.h" +#include "mp/mp_13_0_0_offset.h" + +#include "soc15.h" +#include "soc15_common.h" +#include "soc21.h" + +static const struct amd_ip_funcs soc21_common_ip_funcs; + +/* SOC21 */ +static const struct amdgpu_video_codec_info vcn_4_0_0_video_codecs_encode_array[] = +{ + {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG4_AVC, 4096, 2304, 0)}, + {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_HEVC, 4096, 2304, 0)}, +}; + +static const struct amdgpu_video_codecs vcn_4_0_0_video_codecs_encode = +{ + .codec_count = ARRAY_SIZE(vcn_4_0_0_video_codecs_encode_array), + .codec_array = vcn_4_0_0_video_codecs_encode_array, +}; + +static const struct amdgpu_video_codec_info vcn_4_0_0_video_codecs_decode_array[] = +{ + {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG4_AVC, 4096, 4906, 52)}, + {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_HEVC, 8192, 4352, 186)}, + {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_JPEG, 4096, 4096, 0)}, + {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_VP9, 8192, 4352, 0)}, + {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_AV1, 8192, 4352, 0)}, +}; + +static const struct amdgpu_video_codecs vcn_4_0_0_video_codecs_decode = +{ + .codec_count = ARRAY_SIZE(vcn_4_0_0_video_codecs_decode_array), + .codec_array = vcn_4_0_0_video_codecs_decode_array, +}; + +static int soc21_query_video_codecs(struct amdgpu_device *adev, bool encode, + const struct amdgpu_video_codecs **codecs) +{ + switch (adev->ip_versions[UVD_HWIP][0]) { + + case IP_VERSION(4, 0, 0): + case IP_VERSION(4, 0, 2): + if (encode) + *codecs = &vcn_4_0_0_video_codecs_encode; + else + *codecs = &vcn_4_0_0_video_codecs_decode; + return 0; + default: + return -EINVAL; + } +} +/* + * Indirect registers accessor + */ +static u32 soc21_pcie_rreg(struct amdgpu_device *adev, u32 reg) +{ + unsigned long address, data; + address = adev->nbio.funcs->get_pcie_index_offset(adev); + data = adev->nbio.funcs->get_pcie_data_offset(adev); + + return amdgpu_device_indirect_rreg(adev, address, data, reg); +} + +static void soc21_pcie_wreg(struct amdgpu_device *adev, u32 reg, u32 v) +{ + unsigned long address, data; + + address = adev->nbio.funcs->get_pcie_index_offset(adev); + data = adev->nbio.funcs->get_pcie_data_offset(adev); + + amdgpu_device_indirect_wreg(adev, address, data, reg, v); +} + +static u64 soc21_pcie_rreg64(struct amdgpu_device *adev, u32 reg) +{ + unsigned long address, data; + address = adev->nbio.funcs->get_pcie_index_offset(adev); + data = adev->nbio.funcs->get_pcie_data_offset(adev); + + return amdgpu_device_indirect_rreg64(adev, address, data, reg); +} + +static void soc21_pcie_wreg64(struct amdgpu_device *adev, u32 reg, u64 v) +{ + unsigned long address, data; + + address = adev->nbio.funcs->get_pcie_index_offset(adev); + data = adev->nbio.funcs->get_pcie_data_offset(adev); + + amdgpu_device_indirect_wreg64(adev, address, data, reg, v); +} + +static u32 soc21_didt_rreg(struct amdgpu_device *adev, u32 reg) +{ + unsigned long flags, address, data; + u32 r; + + address = SOC15_REG_OFFSET(GC, 0, regDIDT_IND_INDEX); + data = SOC15_REG_OFFSET(GC, 0, regDIDT_IND_DATA); + + spin_lock_irqsave(&adev->didt_idx_lock, flags); + WREG32(address, (reg)); + r = RREG32(data); + spin_unlock_irqrestore(&adev->didt_idx_lock, flags); + return r; +} + +static void soc21_didt_wreg(struct amdgpu_device *adev, u32 reg, u32 v) +{ + unsigned long flags, address, data; + + address = SOC15_REG_OFFSET(GC, 0, regDIDT_IND_INDEX); + data = SOC15_REG_OFFSET(GC, 0, regDIDT_IND_DATA); + + spin_lock_irqsave(&adev->didt_idx_lock, flags); + WREG32(address, (reg)); + WREG32(data, (v)); + spin_unlock_irqrestore(&adev->didt_idx_lock, flags); +} + +static u32 soc21_get_config_memsize(struct amdgpu_device *adev) +{ + return adev->nbio.funcs->get_memsize(adev); +} + +static u32 soc21_get_xclk(struct amdgpu_device *adev) +{ + return adev->clock.spll.reference_freq; +} + + +void soc21_grbm_select(struct amdgpu_device *adev, + u32 me, u32 pipe, u32 queue, u32 vmid) +{ + u32 grbm_gfx_cntl = 0; + grbm_gfx_cntl = REG_SET_FIELD(grbm_gfx_cntl, GRBM_GFX_CNTL, PIPEID, pipe); + grbm_gfx_cntl = REG_SET_FIELD(grbm_gfx_cntl, GRBM_GFX_CNTL, MEID, me); + grbm_gfx_cntl = REG_SET_FIELD(grbm_gfx_cntl, GRBM_GFX_CNTL, VMID, vmid); + grbm_gfx_cntl = REG_SET_FIELD(grbm_gfx_cntl, GRBM_GFX_CNTL, QUEUEID, queue); + + WREG32_SOC15(GC, 0, regGRBM_GFX_CNTL, grbm_gfx_cntl); +} + +static void soc21_vga_set_state(struct amdgpu_device *adev, bool state) +{ + /* todo */ +} + +static bool soc21_read_disabled_bios(struct amdgpu_device *adev) +{ + /* todo */ + return false; +} + +static struct soc15_allowed_register_entry soc21_allowed_read_registers[] = { + { SOC15_REG_ENTRY(GC, 0, regGRBM_STATUS)}, + { SOC15_REG_ENTRY(GC, 0, regGRBM_STATUS2)}, + { SOC15_REG_ENTRY(GC, 0, regGRBM_STATUS_SE0)}, + { SOC15_REG_ENTRY(GC, 0, regGRBM_STATUS_SE1)}, + { SOC15_REG_ENTRY(GC, 0, regGRBM_STATUS_SE2)}, + { SOC15_REG_ENTRY(GC, 0, regGRBM_STATUS_SE3)}, + { SOC15_REG_ENTRY(SDMA0, 0, regSDMA0_STATUS_REG)}, + { SOC15_REG_ENTRY(SDMA1, 0, regSDMA1_STATUS_REG)}, + { SOC15_REG_ENTRY(GC, 0, regCP_STAT)}, + { SOC15_REG_ENTRY(GC, 0, regCP_STALLED_STAT1)}, + { SOC15_REG_ENTRY(GC, 0, regCP_STALLED_STAT2)}, + { SOC15_REG_ENTRY(GC, 0, regCP_STALLED_STAT3)}, + { SOC15_REG_ENTRY(GC, 0, regCP_CPF_BUSY_STAT)}, + { SOC15_REG_ENTRY(GC, 0, regCP_CPF_STALLED_STAT1)}, + { SOC15_REG_ENTRY(GC, 0, regCP_CPF_STATUS)}, + { SOC15_REG_ENTRY(GC, 0, regCP_CPC_BUSY_STAT)}, + { SOC15_REG_ENTRY(GC, 0, regCP_CPC_STALLED_STAT1)}, + { SOC15_REG_ENTRY(GC, 0, regCP_CPC_STATUS)}, + { SOC15_REG_ENTRY(GC, 0, regGB_ADDR_CONFIG)}, +}; + +static uint32_t soc21_read_indexed_register(struct amdgpu_device *adev, u32 se_num, + u32 sh_num, u32 reg_offset) +{ + uint32_t val; + + mutex_lock(&adev->grbm_idx_mutex); + if (se_num != 0xffffffff || sh_num != 0xffffffff) + amdgpu_gfx_select_se_sh(adev, se_num, sh_num, 0xffffffff); + + val = RREG32(reg_offset); + + if (se_num != 0xffffffff || sh_num != 0xffffffff) + amdgpu_gfx_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff); + mutex_unlock(&adev->grbm_idx_mutex); + return val; +} + +static uint32_t soc21_get_register_value(struct amdgpu_device *adev, + bool indexed, u32 se_num, + u32 sh_num, u32 reg_offset) +{ + if (indexed) { + return soc21_read_indexed_register(adev, se_num, sh_num, reg_offset); + } else { + if (reg_offset == SOC15_REG_OFFSET(GC, 0, regGB_ADDR_CONFIG) && adev->gfx.config.gb_addr_config) + return adev->gfx.config.gb_addr_config; + return RREG32(reg_offset); + } +} + +static int soc21_read_register(struct amdgpu_device *adev, u32 se_num, + u32 sh_num, u32 reg_offset, u32 *value) +{ + uint32_t i; + struct soc15_allowed_register_entry *en; + + *value = 0; + for (i = 0; i < ARRAY_SIZE(soc21_allowed_read_registers); i++) { + en = &soc21_allowed_read_registers[i]; + if (adev->reg_offset[en->hwip][en->inst] && + reg_offset != (adev->reg_offset[en->hwip][en->inst][en->seg] + + en->reg_offset)) + continue; + + *value = soc21_get_register_value(adev, + soc21_allowed_read_registers[i].grbm_indexed, + se_num, sh_num, reg_offset); + return 0; + } + return -EINVAL; +} + +#if 0 +static int soc21_asic_mode1_reset(struct amdgpu_device *adev) +{ + u32 i; + int ret = 0; + + amdgpu_atombios_scratch_regs_engine_hung(adev, true); + + /* disable BM */ + pci_clear_master(adev->pdev); + + amdgpu_device_cache_pci_state(adev->pdev); + + if (amdgpu_dpm_is_mode1_reset_supported(adev)) { + dev_info(adev->dev, "GPU smu mode1 reset\n"); + ret = amdgpu_dpm_mode1_reset(adev); + } else { + dev_info(adev->dev, "GPU psp mode1 reset\n"); + ret = psp_gpu_reset(adev); + } + + if (ret) + dev_err(adev->dev, "GPU mode1 reset failed\n"); + amdgpu_device_load_pci_state(adev->pdev); + + /* wait for asic to come out of reset */ + for (i = 0; i < adev->usec_timeout; i++) { + u32 memsize = adev->nbio.funcs->get_memsize(adev); + + if (memsize != 0xffffffff) + break; + udelay(1); + } + + amdgpu_atombios_scratch_regs_engine_hung(adev, false); + + return ret; +} +#endif + +static enum amd_reset_method +soc21_asic_reset_method(struct amdgpu_device *adev) +{ + if (amdgpu_reset_method == AMD_RESET_METHOD_MODE1 || + amdgpu_reset_method == AMD_RESET_METHOD_MODE2 || + amdgpu_reset_method == AMD_RESET_METHOD_BACO) + return amdgpu_reset_method; + + if (amdgpu_reset_method != -1) + dev_warn(adev->dev, "Specified reset method:%d isn't supported, using AUTO instead.\n", + amdgpu_reset_method); + + switch (adev->ip_versions[MP1_HWIP][0]) { + case IP_VERSION(13, 0, 0): + case IP_VERSION(13, 0, 7): + return AMD_RESET_METHOD_MODE1; + case IP_VERSION(13, 0, 4): + return AMD_RESET_METHOD_MODE2; + default: + if (amdgpu_dpm_is_baco_supported(adev)) + return AMD_RESET_METHOD_BACO; + else + return AMD_RESET_METHOD_MODE1; + } +} + +static int soc21_asic_reset(struct amdgpu_device *adev) +{ + int ret = 0; + + switch (soc21_asic_reset_method(adev)) { + case AMD_RESET_METHOD_PCI: + dev_info(adev->dev, "PCI reset\n"); + ret = amdgpu_device_pci_reset(adev); + break; + case AMD_RESET_METHOD_BACO: + dev_info(adev->dev, "BACO reset\n"); + ret = amdgpu_dpm_baco_reset(adev); + break; + case AMD_RESET_METHOD_MODE2: + dev_info(adev->dev, "MODE2 reset\n"); + ret = amdgpu_dpm_mode2_reset(adev); + break; + default: + dev_info(adev->dev, "MODE1 reset\n"); + ret = amdgpu_device_mode1_reset(adev); + break; + } + + return ret; +} + +static int soc21_set_uvd_clocks(struct amdgpu_device *adev, u32 vclk, u32 dclk) +{ + /* todo */ + return 0; +} + +static int soc21_set_vce_clocks(struct amdgpu_device *adev, u32 evclk, u32 ecclk) +{ + /* todo */ + return 0; +} + +static void soc21_pcie_gen3_enable(struct amdgpu_device *adev) +{ + if (pci_is_root_bus(adev->pdev->bus)) + return; + + if (amdgpu_pcie_gen2 == 0) + return; + + if (!(adev->pm.pcie_gen_mask & (CAIL_PCIE_LINK_SPEED_SUPPORT_GEN2 | + CAIL_PCIE_LINK_SPEED_SUPPORT_GEN3))) + return; + + /* todo */ +} + +static void soc21_program_aspm(struct amdgpu_device *adev) +{ + if (!amdgpu_device_should_use_aspm(adev)) + return; + + if (!(adev->flags & AMD_IS_APU) && + (adev->nbio.funcs->program_aspm)) + adev->nbio.funcs->program_aspm(adev); +} + +static void soc21_enable_doorbell_aperture(struct amdgpu_device *adev, + bool enable) +{ + adev->nbio.funcs->enable_doorbell_aperture(adev, enable); + adev->nbio.funcs->enable_doorbell_selfring_aperture(adev, enable); +} + +const struct amdgpu_ip_block_version soc21_common_ip_block = +{ + .type = AMD_IP_BLOCK_TYPE_COMMON, + .major = 1, + .minor = 0, + .rev = 0, + .funcs = &soc21_common_ip_funcs, +}; + +static uint32_t soc21_get_rev_id(struct amdgpu_device *adev) +{ + return adev->nbio.funcs->get_rev_id(adev); +} + +static bool soc21_need_full_reset(struct amdgpu_device *adev) +{ + switch (adev->ip_versions[GC_HWIP][0]) { + case IP_VERSION(11, 0, 0): + return amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__UMC); + case IP_VERSION(11, 0, 2): + case IP_VERSION(11, 0, 3): + return false; + default: + return true; + } +} + +static bool soc21_need_reset_on_init(struct amdgpu_device *adev) +{ + u32 sol_reg; + + if (adev->flags & AMD_IS_APU) + return false; + + /* Check sOS sign of life register to confirm sys driver and sOS + * are already been loaded. + */ + sol_reg = RREG32_SOC15(MP0, 0, regMP0_SMN_C2PMSG_81); + if (sol_reg) + return true; + + return false; +} + +static uint64_t soc21_get_pcie_replay_count(struct amdgpu_device *adev) +{ + + /* TODO + * dummy implement for pcie_replay_count sysfs interface + * */ + + return 0; +} + +static void soc21_init_doorbell_index(struct amdgpu_device *adev) +{ + adev->doorbell_index.kiq = AMDGPU_NAVI10_DOORBELL_KIQ; + adev->doorbell_index.mec_ring0 = AMDGPU_NAVI10_DOORBELL_MEC_RING0; + adev->doorbell_index.mec_ring1 = AMDGPU_NAVI10_DOORBELL_MEC_RING1; + adev->doorbell_index.mec_ring2 = AMDGPU_NAVI10_DOORBELL_MEC_RING2; + adev->doorbell_index.mec_ring3 = AMDGPU_NAVI10_DOORBELL_MEC_RING3; + adev->doorbell_index.mec_ring4 = AMDGPU_NAVI10_DOORBELL_MEC_RING4; + adev->doorbell_index.mec_ring5 = AMDGPU_NAVI10_DOORBELL_MEC_RING5; + adev->doorbell_index.mec_ring6 = AMDGPU_NAVI10_DOORBELL_MEC_RING6; + adev->doorbell_index.mec_ring7 = AMDGPU_NAVI10_DOORBELL_MEC_RING7; + adev->doorbell_index.userqueue_start = AMDGPU_NAVI10_DOORBELL_USERQUEUE_START; + adev->doorbell_index.userqueue_end = AMDGPU_NAVI10_DOORBELL_USERQUEUE_END; + adev->doorbell_index.gfx_ring0 = AMDGPU_NAVI10_DOORBELL_GFX_RING0; + adev->doorbell_index.gfx_ring1 = AMDGPU_NAVI10_DOORBELL_GFX_RING1; + adev->doorbell_index.gfx_userqueue_start = + AMDGPU_NAVI10_DOORBELL_GFX_USERQUEUE_START; + adev->doorbell_index.gfx_userqueue_end = + AMDGPU_NAVI10_DOORBELL_GFX_USERQUEUE_END; + adev->doorbell_index.mes_ring0 = AMDGPU_NAVI10_DOORBELL_MES_RING0; + adev->doorbell_index.mes_ring1 = AMDGPU_NAVI10_DOORBELL_MES_RING1; + adev->doorbell_index.sdma_engine[0] = AMDGPU_NAVI10_DOORBELL_sDMA_ENGINE0; + adev->doorbell_index.sdma_engine[1] = AMDGPU_NAVI10_DOORBELL_sDMA_ENGINE1; + adev->doorbell_index.ih = AMDGPU_NAVI10_DOORBELL_IH; + adev->doorbell_index.vcn.vcn_ring0_1 = AMDGPU_NAVI10_DOORBELL64_VCN0_1; + adev->doorbell_index.vcn.vcn_ring2_3 = AMDGPU_NAVI10_DOORBELL64_VCN2_3; + adev->doorbell_index.vcn.vcn_ring4_5 = AMDGPU_NAVI10_DOORBELL64_VCN4_5; + adev->doorbell_index.vcn.vcn_ring6_7 = AMDGPU_NAVI10_DOORBELL64_VCN6_7; + adev->doorbell_index.first_non_cp = AMDGPU_NAVI10_DOORBELL64_FIRST_NON_CP; + adev->doorbell_index.last_non_cp = AMDGPU_NAVI10_DOORBELL64_LAST_NON_CP; + + adev->doorbell_index.max_assignment = AMDGPU_NAVI10_DOORBELL_MAX_ASSIGNMENT << 1; + adev->doorbell_index.sdma_doorbell_range = 20; +} + +static void soc21_pre_asic_init(struct amdgpu_device *adev) +{ +} + +static int soc21_update_umd_stable_pstate(struct amdgpu_device *adev, + bool enter) +{ + if (enter) + amdgpu_gfx_rlc_enter_safe_mode(adev); + else + amdgpu_gfx_rlc_exit_safe_mode(adev); + + if (adev->gfx.funcs->update_perfmon_mgcg) + adev->gfx.funcs->update_perfmon_mgcg(adev, !enter); + + return 0; +} + +static const struct amdgpu_asic_funcs soc21_asic_funcs = +{ + .read_disabled_bios = &soc21_read_disabled_bios, + .read_bios_from_rom = &amdgpu_soc15_read_bios_from_rom, + .read_register = &soc21_read_register, + .reset = &soc21_asic_reset, + .reset_method = &soc21_asic_reset_method, + .set_vga_state = &soc21_vga_set_state, + .get_xclk = &soc21_get_xclk, + .set_uvd_clocks = &soc21_set_uvd_clocks, + .set_vce_clocks = &soc21_set_vce_clocks, + .get_config_memsize = &soc21_get_config_memsize, + .init_doorbell_index = &soc21_init_doorbell_index, + .need_full_reset = &soc21_need_full_reset, + .need_reset_on_init = &soc21_need_reset_on_init, + .get_pcie_replay_count = &soc21_get_pcie_replay_count, + .supports_baco = &amdgpu_dpm_is_baco_supported, + .pre_asic_init = &soc21_pre_asic_init, + .query_video_codecs = &soc21_query_video_codecs, + .update_umd_stable_pstate = &soc21_update_umd_stable_pstate, +}; + +static int soc21_common_early_init(void *handle) +{ +#define MMIO_REG_HOLE_OFFSET (0x80000 - PAGE_SIZE) + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + + adev->rmmio_remap.reg_offset = MMIO_REG_HOLE_OFFSET; + adev->rmmio_remap.bus_addr = adev->rmmio_base + MMIO_REG_HOLE_OFFSET; + adev->smc_rreg = NULL; + adev->smc_wreg = NULL; + adev->pcie_rreg = &soc21_pcie_rreg; + adev->pcie_wreg = &soc21_pcie_wreg; + adev->pcie_rreg64 = &soc21_pcie_rreg64; + adev->pcie_wreg64 = &soc21_pcie_wreg64; + adev->pciep_rreg = amdgpu_device_pcie_port_rreg; + adev->pciep_wreg = amdgpu_device_pcie_port_wreg; + + /* TODO: will add them during VCN v2 implementation */ + adev->uvd_ctx_rreg = NULL; + adev->uvd_ctx_wreg = NULL; + + adev->didt_rreg = &soc21_didt_rreg; + adev->didt_wreg = &soc21_didt_wreg; + + adev->asic_funcs = &soc21_asic_funcs; + + adev->rev_id = soc21_get_rev_id(adev); + adev->external_rev_id = 0xff; + switch (adev->ip_versions[GC_HWIP][0]) { + case IP_VERSION(11, 0, 0): + adev->cg_flags = AMD_CG_SUPPORT_GFX_CGCG | + AMD_CG_SUPPORT_GFX_CGLS | +#if 0 + AMD_CG_SUPPORT_GFX_3D_CGCG | + AMD_CG_SUPPORT_GFX_3D_CGLS | +#endif + AMD_CG_SUPPORT_GFX_MGCG | + AMD_CG_SUPPORT_REPEATER_FGCG | + AMD_CG_SUPPORT_GFX_FGCG | + AMD_CG_SUPPORT_GFX_PERF_CLK | + AMD_CG_SUPPORT_VCN_MGCG | + AMD_CG_SUPPORT_JPEG_MGCG | + AMD_CG_SUPPORT_ATHUB_MGCG | + AMD_CG_SUPPORT_ATHUB_LS | + AMD_CG_SUPPORT_MC_MGCG | + AMD_CG_SUPPORT_MC_LS | + AMD_CG_SUPPORT_IH_CG | + AMD_CG_SUPPORT_HDP_SD; + adev->pg_flags = AMD_PG_SUPPORT_VCN | + AMD_PG_SUPPORT_VCN_DPG | + AMD_PG_SUPPORT_JPEG | + AMD_PG_SUPPORT_ATHUB | + AMD_PG_SUPPORT_MMHUB; + if (amdgpu_sriov_vf(adev)) { + adev->cg_flags = 0; + adev->pg_flags = 0; + } + adev->external_rev_id = adev->rev_id + 0x1; // TODO: need update + break; + case IP_VERSION(11, 0, 2): + adev->cg_flags = + AMD_CG_SUPPORT_GFX_CGCG | + AMD_CG_SUPPORT_GFX_CGLS | + AMD_CG_SUPPORT_REPEATER_FGCG | + AMD_CG_SUPPORT_VCN_MGCG | + AMD_CG_SUPPORT_JPEG_MGCG | + AMD_CG_SUPPORT_ATHUB_MGCG | + AMD_CG_SUPPORT_ATHUB_LS | + AMD_CG_SUPPORT_IH_CG | + AMD_CG_SUPPORT_HDP_SD; + adev->pg_flags = + AMD_PG_SUPPORT_VCN | + AMD_PG_SUPPORT_VCN_DPG | + AMD_PG_SUPPORT_JPEG | + AMD_PG_SUPPORT_ATHUB | + AMD_PG_SUPPORT_MMHUB; + adev->external_rev_id = adev->rev_id + 0x10; + break; + case IP_VERSION(11, 0, 1): + adev->cg_flags = + AMD_CG_SUPPORT_GFX_CGCG | + AMD_CG_SUPPORT_GFX_CGLS | + AMD_CG_SUPPORT_GFX_MGCG | + AMD_CG_SUPPORT_GFX_FGCG | + AMD_CG_SUPPORT_REPEATER_FGCG | + AMD_CG_SUPPORT_GFX_PERF_CLK | + AMD_CG_SUPPORT_MC_MGCG | + AMD_CG_SUPPORT_MC_LS | + AMD_CG_SUPPORT_HDP_MGCG | + AMD_CG_SUPPORT_HDP_LS | + AMD_CG_SUPPORT_ATHUB_MGCG | + AMD_CG_SUPPORT_ATHUB_LS | + AMD_CG_SUPPORT_IH_CG | + AMD_CG_SUPPORT_BIF_MGCG | + AMD_CG_SUPPORT_BIF_LS | + AMD_CG_SUPPORT_VCN_MGCG | + AMD_CG_SUPPORT_JPEG_MGCG; + adev->pg_flags = + AMD_PG_SUPPORT_GFX_PG | + AMD_PG_SUPPORT_VCN | + AMD_PG_SUPPORT_VCN_DPG | + AMD_PG_SUPPORT_JPEG; + adev->external_rev_id = adev->rev_id + 0x1; + break; + case IP_VERSION(11, 0, 3): + adev->cg_flags = AMD_CG_SUPPORT_VCN_MGCG | + AMD_CG_SUPPORT_JPEG_MGCG | + AMD_CG_SUPPORT_GFX_CGCG | + AMD_CG_SUPPORT_GFX_CGLS | + AMD_CG_SUPPORT_REPEATER_FGCG | + AMD_CG_SUPPORT_GFX_MGCG; + adev->pg_flags = AMD_PG_SUPPORT_VCN | + AMD_PG_SUPPORT_VCN_DPG | + AMD_PG_SUPPORT_JPEG; + if (amdgpu_sriov_vf(adev)) { + /* hypervisor control CG and PG enablement */ + adev->cg_flags = 0; + adev->pg_flags = 0; + } + adev->external_rev_id = adev->rev_id + 0x20; + break; + default: + /* FIXME: not supported yet */ + return -EINVAL; + } + + return 0; +} + +static int soc21_common_late_init(void *handle) +{ + return 0; +} + +static int soc21_common_sw_init(void *handle) +{ + return 0; +} + +static int soc21_common_sw_fini(void *handle) +{ + return 0; +} + +static int soc21_common_hw_init(void *handle) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + + /* enable pcie gen2/3 link */ + soc21_pcie_gen3_enable(adev); + /* enable aspm */ + soc21_program_aspm(adev); + /* setup nbio registers */ + adev->nbio.funcs->init_registers(adev); + /* remap HDP registers to a hole in mmio space, + * for the purpose of expose those registers + * to process space + */ + if (adev->nbio.funcs->remap_hdp_registers) + adev->nbio.funcs->remap_hdp_registers(adev); + /* enable the doorbell aperture */ + soc21_enable_doorbell_aperture(adev, true); + + return 0; +} + +static int soc21_common_hw_fini(void *handle) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + + /* disable the doorbell aperture */ + soc21_enable_doorbell_aperture(adev, false); + + return 0; +} + +static int soc21_common_suspend(void *handle) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + + return soc21_common_hw_fini(adev); +} + +static int soc21_common_resume(void *handle) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + + return soc21_common_hw_init(adev); +} + +static bool soc21_common_is_idle(void *handle) +{ + return true; +} + +static int soc21_common_wait_for_idle(void *handle) +{ + return 0; +} + +static int soc21_common_soft_reset(void *handle) +{ + return 0; +} + +static int soc21_common_set_clockgating_state(void *handle, + enum amd_clockgating_state state) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + + switch (adev->ip_versions[NBIO_HWIP][0]) { + case IP_VERSION(4, 3, 0): + case IP_VERSION(4, 3, 1): + case IP_VERSION(7, 7, 0): + adev->nbio.funcs->update_medium_grain_clock_gating(adev, + state == AMD_CG_STATE_GATE); + adev->nbio.funcs->update_medium_grain_light_sleep(adev, + state == AMD_CG_STATE_GATE); + adev->hdp.funcs->update_clock_gating(adev, + state == AMD_CG_STATE_GATE); + break; + default: + break; + } + return 0; +} + +static int soc21_common_set_powergating_state(void *handle, + enum amd_powergating_state state) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + + switch (adev->ip_versions[LSDMA_HWIP][0]) { + case IP_VERSION(6, 0, 0): + case IP_VERSION(6, 0, 2): + adev->lsdma.funcs->update_memory_power_gating(adev, + state == AMD_PG_STATE_GATE); + break; + default: + break; + } + + return 0; +} + +static void soc21_common_get_clockgating_state(void *handle, u64 *flags) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + + adev->nbio.funcs->get_clockgating_state(adev, flags); + + adev->hdp.funcs->get_clock_gating_state(adev, flags); + + return; +} + +static const struct amd_ip_funcs soc21_common_ip_funcs = { + .name = "soc21_common", + .early_init = soc21_common_early_init, + .late_init = soc21_common_late_init, + .sw_init = soc21_common_sw_init, + .sw_fini = soc21_common_sw_fini, + .hw_init = soc21_common_hw_init, + .hw_fini = soc21_common_hw_fini, + .suspend = soc21_common_suspend, + .resume = soc21_common_resume, + .is_idle = soc21_common_is_idle, + .wait_for_idle = soc21_common_wait_for_idle, + .soft_reset = soc21_common_soft_reset, + .set_clockgating_state = soc21_common_set_clockgating_state, + .set_powergating_state = soc21_common_set_powergating_state, + .get_clockgating_state = soc21_common_get_clockgating_state, +}; diff --git a/drivers/gpu/drm/amd/amdgpu/soc21.h b/drivers/gpu/drm/amd/amdgpu/soc21.h new file mode 100644 index 000000000000..4c8067af1b65 --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/soc21.h @@ -0,0 +1,30 @@ +/* + * Copyright 2021 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ +#ifndef __SOC21_H__ +#define __SOC21_H__ + +extern const struct amdgpu_ip_block_version soc21_common_ip_block; + +void soc21_grbm_select(struct amdgpu_device *adev, + u32 me, u32 pipe, u32 queue, u32 vmid); +#endif diff --git a/drivers/gpu/drm/amd/amdgpu/ta_ras_if.h b/drivers/gpu/drm/amd/amdgpu/ta_ras_if.h index 5093826a43d1..509d8a1945eb 100644 --- a/drivers/gpu/drm/amd/amdgpu/ta_ras_if.h +++ b/drivers/gpu/drm/amd/amdgpu/ta_ras_if.h @@ -64,7 +64,8 @@ enum ta_ras_status { TA_RAS_STATUS__ERROR_PCS_STATE_ERROR = 0xA016, TA_RAS_STATUS__ERROR_PCS_STATE_HANG = 0xA017, TA_RAS_STATUS__ERROR_PCS_STATE_UNKNOWN = 0xA018, - TA_RAS_STATUS__ERROR_UNSUPPORTED_ERROR_INJ = 0xA019 + TA_RAS_STATUS__ERROR_UNSUPPORTED_ERROR_INJ = 0xA019, + TA_RAS_STATUS__TEE_ERROR_ACCESS_DENIED = 0xA01A }; enum ta_ras_block { diff --git a/drivers/gpu/drm/amd/amdgpu/umc_v6_1.c b/drivers/gpu/drm/amd/amdgpu/umc_v6_1.c index 20b44983ac94..f17d297b594b 100644 --- a/drivers/gpu/drm/amd/amdgpu/umc_v6_1.c +++ b/drivers/gpu/drm/amd/amdgpu/umc_v6_1.c @@ -300,7 +300,6 @@ static void umc_v6_1_query_error_address(struct amdgpu_device *adev, { uint32_t lsb, mc_umc_status_addr; uint64_t mc_umc_status, err_addr, retired_page, mc_umc_addrt0; - struct eeprom_table_record *err_rec; uint32_t channel_index = adev->umc.channel_idx_tbl[umc_inst * adev->umc.channel_inst_num + ch_inst]; if (adev->asic_type == CHIP_ARCTURUS) { @@ -328,12 +327,9 @@ static void umc_v6_1_query_error_address(struct amdgpu_device *adev, return; } - err_rec = &err_data->err_addr[err_data->err_addr_cnt]; - - /* calculate error address if ue/ce error is detected */ + /* calculate error address if ue error is detected */ if (REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, Val) == 1 && - (REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, UECC) == 1 || - REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, CECC) == 1)) { + REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, UECC) == 1) { err_addr = RREG64_PCIE((mc_umc_addrt0 + umc_reg_offset) * 4); /* the lowest lsb bits should be ignored */ @@ -346,20 +342,8 @@ static void umc_v6_1_query_error_address(struct amdgpu_device *adev, ADDR_OF_256B_BLOCK(channel_index) | OFFSET_IN_256B_BLOCK(err_addr); - /* we only save ue error information currently, ce is skipped */ - if (REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, UECC) - == 1) { - err_rec->address = err_addr; - /* page frame address is saved */ - err_rec->retired_page = retired_page >> AMDGPU_GPU_PAGE_SHIFT; - err_rec->ts = (uint64_t)ktime_get_real_seconds(); - err_rec->err_type = AMDGPU_RAS_EEPROM_ERR_NON_RECOVERABLE; - err_rec->cu = 0; - err_rec->mem_channel = channel_index; - err_rec->mcumc_id = umc_inst; - - err_data->err_addr_cnt++; - } + amdgpu_umc_fill_error_record(err_data, err_addr, + retired_page, channel_index, umc_inst); } /* clear umc status */ @@ -465,10 +449,14 @@ static void umc_v6_1_err_cnt_init(struct amdgpu_device *adev) umc_v6_1_enable_umc_index_mode(adev); } -const struct amdgpu_umc_ras_funcs umc_v6_1_ras_funcs = { - .err_cnt_init = umc_v6_1_err_cnt_init, - .ras_late_init = amdgpu_umc_ras_late_init, - .ras_fini = amdgpu_umc_ras_fini, +const struct amdgpu_ras_block_hw_ops umc_v6_1_ras_hw_ops = { .query_ras_error_count = umc_v6_1_query_ras_error_count, .query_ras_error_address = umc_v6_1_query_ras_error_address, }; + +struct amdgpu_umc_ras umc_v6_1_ras = { + .ras_block = { + .hw_ops = &umc_v6_1_ras_hw_ops, + }, + .err_cnt_init = umc_v6_1_err_cnt_init, +};
\ No newline at end of file diff --git a/drivers/gpu/drm/amd/amdgpu/umc_v6_1.h b/drivers/gpu/drm/amd/amdgpu/umc_v6_1.h index 5dc36c730bb2..50c632eb4cc6 100644 --- a/drivers/gpu/drm/amd/amdgpu/umc_v6_1.h +++ b/drivers/gpu/drm/amd/amdgpu/umc_v6_1.h @@ -45,7 +45,7 @@ /* umc ce count initial value */ #define UMC_V6_1_CE_CNT_INIT (UMC_V6_1_CE_CNT_MAX - UMC_V6_1_CE_INT_THRESHOLD) -extern const struct amdgpu_umc_ras_funcs umc_v6_1_ras_funcs; +extern struct amdgpu_umc_ras umc_v6_1_ras; extern const uint32_t umc_v6_1_channel_idx_tbl[UMC_V6_1_UMC_INSTANCE_NUM][UMC_V6_1_CHANNEL_INSTANCE_NUM]; diff --git a/drivers/gpu/drm/amd/amdgpu/umc_v6_7.c b/drivers/gpu/drm/amd/amdgpu/umc_v6_7.c index 6dd1e19e8d43..5d5d031c9e7d 100644 --- a/drivers/gpu/drm/amd/amdgpu/umc_v6_7.c +++ b/drivers/gpu/drm/amd/amdgpu/umc_v6_7.c @@ -47,6 +47,13 @@ static inline uint32_t get_umc_v6_7_reg_offset(struct amdgpu_device *adev, uint32_t umc_inst, uint32_t ch_inst) { + uint32_t index = umc_inst * adev->umc.channel_inst_num + ch_inst; + + /* adjust umc and channel index offset, + * the register address is not linear on each umc instace */ + umc_inst = index / 4; + ch_inst = index % 4; + return adev->umc.channel_offs * ch_inst + UMC_V6_7_INST_DIST * umc_inst; } @@ -57,50 +64,107 @@ static inline uint32_t get_umc_v6_7_channel_index(struct amdgpu_device *adev, return adev->umc.channel_idx_tbl[umc_inst * adev->umc.channel_inst_num + ch_inst]; } +static void umc_v6_7_query_error_status_helper(struct amdgpu_device *adev, + uint64_t mc_umc_status, uint32_t umc_reg_offset) +{ + uint32_t mc_umc_addr; + uint64_t reg_value; + + if (REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, Deferred) == 1) + dev_info(adev->dev, "Deferred error, no user action is needed.\n"); + + if (mc_umc_status) + dev_info(adev->dev, "MCA STATUS 0x%llx, umc_reg_offset 0x%x\n", mc_umc_status, umc_reg_offset); + + /* print IPID registers value */ + mc_umc_addr = + SOC15_REG_OFFSET(UMC, 0, regMCA_UMC_UMC0_MCUMC_IPIDT0); + reg_value = RREG64_PCIE((mc_umc_addr + umc_reg_offset) * 4); + if (reg_value) + dev_info(adev->dev, "MCA IPID 0x%llx, umc_reg_offset 0x%x\n", reg_value, umc_reg_offset); + + /* print SYND registers value */ + mc_umc_addr = + SOC15_REG_OFFSET(UMC, 0, regMCA_UMC_UMC0_MCUMC_SYNDT0); + reg_value = RREG64_PCIE((mc_umc_addr + umc_reg_offset) * 4); + if (reg_value) + dev_info(adev->dev, "MCA SYND 0x%llx, umc_reg_offset 0x%x\n", reg_value, umc_reg_offset); + + /* print MISC0 registers value */ + mc_umc_addr = + SOC15_REG_OFFSET(UMC, 0, regMCA_UMC_UMC0_MCUMC_MISC0T0); + reg_value = RREG64_PCIE((mc_umc_addr + umc_reg_offset) * 4); + if (reg_value) + dev_info(adev->dev, "MCA MISC0 0x%llx, umc_reg_offset 0x%x\n", reg_value, umc_reg_offset); +} + static void umc_v6_7_ecc_info_query_correctable_error_count(struct amdgpu_device *adev, - uint32_t channel_index, + uint32_t umc_inst, uint32_t ch_inst, unsigned long *error_count) { - uint32_t ecc_err_cnt; uint64_t mc_umc_status; + uint32_t eccinfo_table_idx; + uint32_t umc_reg_offset; struct amdgpu_ras *ras = amdgpu_ras_get_context(adev); - /* - * select the lower chip and check the error count - * skip add error count, calc error counter only from mca_umc_status - */ - ecc_err_cnt = ras->umc_ecc.ecc[channel_index].ce_count_lo_chip; - - /* - * select the higher chip and check the err counter - * skip add error count, calc error counter only from mca_umc_status - */ - ecc_err_cnt = ras->umc_ecc.ecc[channel_index].ce_count_hi_chip; + umc_reg_offset = get_umc_v6_7_reg_offset(adev, + umc_inst, ch_inst); + eccinfo_table_idx = umc_inst * adev->umc.channel_inst_num + ch_inst; /* check for SRAM correctable error MCUMC_STATUS is a 64 bit register */ - mc_umc_status = ras->umc_ecc.ecc[channel_index].mca_umc_status; + mc_umc_status = ras->umc_ecc.ecc[eccinfo_table_idx].mca_umc_status; if (REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, Val) == 1 && - REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, CECC) == 1) + REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, CECC) == 1) { *error_count += 1; + + umc_v6_7_query_error_status_helper(adev, mc_umc_status, umc_reg_offset); + + if (ras->umc_ecc.record_ce_addr_supported) { + uint64_t err_addr, soc_pa; + uint32_t channel_index = + adev->umc.channel_idx_tbl[umc_inst * adev->umc.channel_inst_num + ch_inst]; + + err_addr = ras->umc_ecc.ecc[eccinfo_table_idx].mca_ceumc_addr; + err_addr = REG_GET_FIELD(err_addr, MCA_UMC_UMC0_MCUMC_ADDRT0, ErrorAddr); + /* translate umc channel address to soc pa, 3 parts are included */ + soc_pa = ADDR_OF_8KB_BLOCK(err_addr) | + ADDR_OF_256B_BLOCK(channel_index) | + OFFSET_IN_256B_BLOCK(err_addr); + + /* The umc channel bits are not original values, they are hashed */ + SET_CHANNEL_HASH(channel_index, soc_pa); + + dev_info(adev->dev, "Error Address(PA): 0x%llx\n", soc_pa); + } + } } static void umc_v6_7_ecc_info_querry_uncorrectable_error_count(struct amdgpu_device *adev, - uint32_t channel_index, + uint32_t umc_inst, uint32_t ch_inst, unsigned long *error_count) { uint64_t mc_umc_status; + uint32_t eccinfo_table_idx; + uint32_t umc_reg_offset; struct amdgpu_ras *ras = amdgpu_ras_get_context(adev); + umc_reg_offset = get_umc_v6_7_reg_offset(adev, + umc_inst, ch_inst); + + eccinfo_table_idx = umc_inst * adev->umc.channel_inst_num + ch_inst; /* check the MCUMC_STATUS */ - mc_umc_status = ras->umc_ecc.ecc[channel_index].mca_umc_status; + mc_umc_status = ras->umc_ecc.ecc[eccinfo_table_idx].mca_umc_status; if ((REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, Val) == 1) && (REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, Deferred) == 1 || REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, UECC) == 1 || REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, PCC) == 1 || REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, UC) == 1 || - REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, TCC) == 1)) + REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, TCC) == 1)) { *error_count += 1; + + umc_v6_7_query_error_status_helper(adev, mc_umc_status, umc_reg_offset); + } } static void umc_v6_7_ecc_info_query_ras_error_count(struct amdgpu_device *adev, @@ -110,42 +174,65 @@ static void umc_v6_7_ecc_info_query_ras_error_count(struct amdgpu_device *adev, uint32_t umc_inst = 0; uint32_t ch_inst = 0; - uint32_t umc_reg_offset = 0; - uint32_t channel_index = 0; /*TODO: driver needs to toggle DF Cstate to ensure * safe access of UMC registers. Will add the protection */ LOOP_UMC_INST_AND_CH(umc_inst, ch_inst) { - umc_reg_offset = get_umc_v6_7_reg_offset(adev, - umc_inst, - ch_inst); - channel_index = get_umc_v6_7_channel_index(adev, - umc_inst, - ch_inst); umc_v6_7_ecc_info_query_correctable_error_count(adev, - channel_index, + umc_inst, ch_inst, &(err_data->ce_count)); umc_v6_7_ecc_info_querry_uncorrectable_error_count(adev, - channel_index, + umc_inst, ch_inst, &(err_data->ue_count)); } } +static void umc_v6_7_convert_error_address(struct amdgpu_device *adev, + struct ras_err_data *err_data, uint64_t err_addr, + uint32_t ch_inst, uint32_t umc_inst) +{ + uint32_t channel_index; + uint64_t soc_pa, retired_page, column; + + channel_index = + adev->umc.channel_idx_tbl[umc_inst * adev->umc.channel_inst_num + ch_inst]; + /* translate umc channel address to soc pa, 3 parts are included */ + soc_pa = ADDR_OF_8KB_BLOCK(err_addr) | + ADDR_OF_256B_BLOCK(channel_index) | + OFFSET_IN_256B_BLOCK(err_addr); + + /* The umc channel bits are not original values, they are hashed */ + SET_CHANNEL_HASH(channel_index, soc_pa); + + /* clear [C4 C3 C2] in soc physical address */ + soc_pa &= ~(0x7ULL << UMC_V6_7_PA_C2_BIT); + + /* loop for all possibilities of [C4 C3 C2] */ + for (column = 0; column < UMC_V6_7_NA_MAP_PA_NUM; column++) { + retired_page = soc_pa | (column << UMC_V6_7_PA_C2_BIT); + dev_info(adev->dev, "Error Address(PA): 0x%llx\n", retired_page); + amdgpu_umc_fill_error_record(err_data, err_addr, + retired_page, channel_index, umc_inst); + + /* shift R14 bit */ + retired_page ^= (0x1ULL << UMC_V6_7_PA_R14_BIT); + dev_info(adev->dev, "Error Address(PA): 0x%llx\n", retired_page); + amdgpu_umc_fill_error_record(err_data, err_addr, + retired_page, channel_index, umc_inst); + } +} + static void umc_v6_7_ecc_info_query_error_address(struct amdgpu_device *adev, struct ras_err_data *err_data, - uint32_t umc_reg_offset, uint32_t ch_inst, uint32_t umc_inst) { - uint64_t mc_umc_status, err_addr, retired_page; - struct eeprom_table_record *err_rec; - uint32_t channel_index; + uint64_t mc_umc_status, err_addr; + uint32_t eccinfo_table_idx; struct amdgpu_ras *ras = amdgpu_ras_get_context(adev); - channel_index = - adev->umc.channel_idx_tbl[umc_inst * adev->umc.channel_inst_num + ch_inst]; - - mc_umc_status = ras->umc_ecc.ecc[channel_index].mca_umc_status; + eccinfo_table_idx = umc_inst * adev->umc.channel_inst_num + ch_inst; + mc_umc_status = ras->umc_ecc.ecc[eccinfo_table_idx].mca_umc_status; if (mc_umc_status == 0) return; @@ -153,35 +240,15 @@ static void umc_v6_7_ecc_info_query_error_address(struct amdgpu_device *adev, if (!err_data->err_addr) return; - err_rec = &err_data->err_addr[err_data->err_addr_cnt]; - - /* calculate error address if ue/ce error is detected */ + /* calculate error address if ue error is detected */ if (REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, Val) == 1 && - (REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, UECC) == 1 || - REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, CECC) == 1)) { + REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, UECC) == 1) { - err_addr = ras->umc_ecc.ecc[channel_index].mca_umc_addr; + err_addr = ras->umc_ecc.ecc[eccinfo_table_idx].mca_umc_addr; err_addr = REG_GET_FIELD(err_addr, MCA_UMC_UMC0_MCUMC_ADDRT0, ErrorAddr); - /* translate umc channel address to soc pa, 3 parts are included */ - retired_page = ADDR_OF_8KB_BLOCK(err_addr) | - ADDR_OF_256B_BLOCK(channel_index) | - OFFSET_IN_256B_BLOCK(err_addr); - - /* we only save ue error information currently, ce is skipped */ - if (REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, UECC) - == 1) { - err_rec->address = err_addr; - /* page frame address is saved */ - err_rec->retired_page = retired_page >> AMDGPU_GPU_PAGE_SHIFT; - err_rec->ts = (uint64_t)ktime_get_real_seconds(); - err_rec->err_type = AMDGPU_RAS_EEPROM_ERR_NON_RECOVERABLE; - err_rec->cu = 0; - err_rec->mem_channel = channel_index; - err_rec->mcumc_id = umc_inst; - - err_data->err_addr_cnt++; - } + umc_v6_7_convert_error_address(adev, err_data, err_addr, + ch_inst, umc_inst); } } @@ -192,18 +259,13 @@ static void umc_v6_7_ecc_info_query_ras_error_address(struct amdgpu_device *adev uint32_t umc_inst = 0; uint32_t ch_inst = 0; - uint32_t umc_reg_offset = 0; /*TODO: driver needs to toggle DF Cstate to ensure * safe access of UMC resgisters. Will add the protection * when firmware interface is ready */ LOOP_UMC_INST_AND_CH(umc_inst, ch_inst) { - umc_reg_offset = get_umc_v6_7_reg_offset(adev, - umc_inst, - ch_inst); umc_v6_7_ecc_info_query_error_address(adev, err_data, - umc_reg_offset, ch_inst, umc_inst); } @@ -211,7 +273,9 @@ static void umc_v6_7_ecc_info_query_ras_error_address(struct amdgpu_device *adev static void umc_v6_7_query_correctable_error_count(struct amdgpu_device *adev, uint32_t umc_reg_offset, - unsigned long *error_count) + unsigned long *error_count, + uint32_t ch_inst, + uint32_t umc_inst) { uint32_t ecc_err_cnt_sel, ecc_err_cnt_sel_addr; uint32_t ecc_err_cnt, ecc_err_cnt_addr; @@ -251,8 +315,36 @@ static void umc_v6_7_query_correctable_error_count(struct amdgpu_device *adev, MCUMC_STATUS is a 64 bit register */ mc_umc_status = RREG64_PCIE((mc_umc_status_addr + umc_reg_offset) * 4); if (REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, Val) == 1 && - REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, CECC) == 1) + REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, CECC) == 1) { *error_count += 1; + + umc_v6_7_query_error_status_helper(adev, mc_umc_status, umc_reg_offset); + + { + uint64_t err_addr, soc_pa; + uint32_t mc_umc_addrt0; + uint32_t channel_index; + + mc_umc_addrt0 = + SOC15_REG_OFFSET(UMC, 0, regMCA_UMC_UMC0_MCUMC_ADDRT0); + + channel_index = + adev->umc.channel_idx_tbl[umc_inst * adev->umc.channel_inst_num + ch_inst]; + + err_addr = RREG64_PCIE((mc_umc_addrt0 + umc_reg_offset) * 4); + err_addr = REG_GET_FIELD(err_addr, MCA_UMC_UMC0_MCUMC_ADDRT0, ErrorAddr); + + /* translate umc channel address to soc pa, 3 parts are included */ + soc_pa = ADDR_OF_8KB_BLOCK(err_addr) | + ADDR_OF_256B_BLOCK(channel_index) | + OFFSET_IN_256B_BLOCK(err_addr); + + /* The umc channel bits are not original values, they are hashed */ + SET_CHANNEL_HASH(channel_index, soc_pa); + + dev_info(adev->dev, "Error Address(PA): 0x%llx\n", soc_pa); + } + } } static void umc_v6_7_querry_uncorrectable_error_count(struct amdgpu_device *adev, @@ -272,8 +364,11 @@ static void umc_v6_7_querry_uncorrectable_error_count(struct amdgpu_device *adev REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, UECC) == 1 || REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, PCC) == 1 || REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, UC) == 1 || - REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, TCC) == 1)) + REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, TCC) == 1)) { *error_count += 1; + + umc_v6_7_query_error_status_helper(adev, mc_umc_status, umc_reg_offset); + } } static void umc_v6_7_reset_error_count_per_channel(struct amdgpu_device *adev, @@ -349,7 +444,8 @@ static void umc_v6_7_query_ras_error_count(struct amdgpu_device *adev, ch_inst); umc_v6_7_query_correctable_error_count(adev, umc_reg_offset, - &(err_data->ce_count)); + &(err_data->ce_count), + ch_inst, umc_inst); umc_v6_7_querry_uncorrectable_error_count(adev, umc_reg_offset, &(err_data->ue_count)); @@ -360,14 +456,11 @@ static void umc_v6_7_query_ras_error_count(struct amdgpu_device *adev, static void umc_v6_7_query_error_address(struct amdgpu_device *adev, struct ras_err_data *err_data, - uint32_t umc_reg_offset, - uint32_t ch_inst, + uint32_t umc_reg_offset, uint32_t ch_inst, uint32_t umc_inst) { uint32_t mc_umc_status_addr; - uint64_t mc_umc_status, err_addr, retired_page, mc_umc_addrt0; - struct eeprom_table_record *err_rec; - uint32_t channel_index; + uint64_t mc_umc_status = 0, mc_umc_addrt0, err_addr; mc_umc_status_addr = SOC15_REG_OFFSET(UMC, 0, regMCA_UMC_UMC0_MCUMC_STATUST0); @@ -385,38 +478,15 @@ static void umc_v6_7_query_error_address(struct amdgpu_device *adev, return; } - err_rec = &err_data->err_addr[err_data->err_addr_cnt]; - - channel_index = - adev->umc.channel_idx_tbl[umc_inst * adev->umc.channel_inst_num + ch_inst]; - - /* calculate error address if ue/ce error is detected */ + /* calculate error address if ue error is detected */ if (REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, Val) == 1 && - (REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, UECC) == 1 || - REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, CECC) == 1)) { - + REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, UECC) == 1) { err_addr = RREG64_PCIE((mc_umc_addrt0 + umc_reg_offset) * 4); - err_addr = REG_GET_FIELD(err_addr, MCA_UMC_UMC0_MCUMC_ADDRT0, ErrorAddr); + err_addr = + REG_GET_FIELD(err_addr, MCA_UMC_UMC0_MCUMC_ADDRT0, ErrorAddr); - /* translate umc channel address to soc pa, 3 parts are included */ - retired_page = ADDR_OF_8KB_BLOCK(err_addr) | - ADDR_OF_256B_BLOCK(channel_index) | - OFFSET_IN_256B_BLOCK(err_addr); - - /* we only save ue error information currently, ce is skipped */ - if (REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, UECC) - == 1) { - err_rec->address = err_addr; - /* page frame address is saved */ - err_rec->retired_page = retired_page >> AMDGPU_GPU_PAGE_SHIFT; - err_rec->ts = (uint64_t)ktime_get_real_seconds(); - err_rec->err_type = AMDGPU_RAS_EEPROM_ERR_NON_RECOVERABLE; - err_rec->cu = 0; - err_rec->mem_channel = channel_index; - err_rec->mcumc_id = umc_inst; - - err_data->err_addr_cnt++; - } + umc_v6_7_convert_error_address(adev, err_data, err_addr, + ch_inst, umc_inst); } /* clear umc status */ @@ -441,8 +511,7 @@ static void umc_v6_7_query_ras_error_address(struct amdgpu_device *adev, ch_inst); umc_v6_7_query_error_address(adev, err_data, - umc_reg_offset, - ch_inst, + umc_reg_offset, ch_inst, umc_inst); } } @@ -463,29 +532,26 @@ static uint32_t umc_v6_7_query_ras_poison_mode_per_channel( static bool umc_v6_7_query_ras_poison_mode(struct amdgpu_device *adev) { - uint32_t umc_inst = 0; - uint32_t ch_inst = 0; uint32_t umc_reg_offset = 0; - LOOP_UMC_INST_AND_CH(umc_inst, ch_inst) { - umc_reg_offset = get_umc_v6_7_reg_offset(adev, - umc_inst, - ch_inst); - /* Enabling fatal error in one channel will be considered - as fatal error mode */ - if (umc_v6_7_query_ras_poison_mode_per_channel(adev, umc_reg_offset)) - return false; - } - - return true; + /* Enabling fatal error in umc instance0 channel0 will be + * considered as fatal error mode + */ + umc_reg_offset = get_umc_v6_7_reg_offset(adev, 0, 0); + return !umc_v6_7_query_ras_poison_mode_per_channel(adev, umc_reg_offset); } -const struct amdgpu_umc_ras_funcs umc_v6_7_ras_funcs = { - .ras_late_init = amdgpu_umc_ras_late_init, - .ras_fini = amdgpu_umc_ras_fini, +const struct amdgpu_ras_block_hw_ops umc_v6_7_ras_hw_ops = { .query_ras_error_count = umc_v6_7_query_ras_error_count, .query_ras_error_address = umc_v6_7_query_ras_error_address, +}; + +struct amdgpu_umc_ras umc_v6_7_ras = { + .ras_block = { + .hw_ops = &umc_v6_7_ras_hw_ops, + }, .query_ras_poison_mode = umc_v6_7_query_ras_poison_mode, .ecc_info_query_ras_error_count = umc_v6_7_ecc_info_query_ras_error_count, .ecc_info_query_ras_error_address = umc_v6_7_ecc_info_query_ras_error_address, + .convert_ras_error_address = umc_v6_7_convert_error_address, }; diff --git a/drivers/gpu/drm/amd/amdgpu/umc_v6_7.h b/drivers/gpu/drm/amd/amdgpu/umc_v6_7.h index 57f2557e7aca..fe41ed2f5945 100644 --- a/drivers/gpu/drm/amd/amdgpu/umc_v6_7.h +++ b/drivers/gpu/drm/amd/amdgpu/umc_v6_7.h @@ -41,9 +41,32 @@ #define UMC_V6_7_CHANNEL_INSTANCE_NUM 8 /* total channel instances in one umc block */ #define UMC_V6_7_TOTAL_CHANNEL_NUM (UMC_V6_7_CHANNEL_INSTANCE_NUM * UMC_V6_7_UMC_INSTANCE_NUM) +/* one piece of normalizing address is mapped to 8 pieces of physical address */ +#define UMC_V6_7_NA_MAP_PA_NUM 8 +/* R14 bit shift should be considered, double the number */ +#define UMC_V6_7_BAD_PAGE_NUM_PER_CHANNEL (UMC_V6_7_NA_MAP_PA_NUM * 2) +/* The CH4 bit in SOC physical address */ +#define UMC_V6_7_PA_CH4_BIT 12 +/* The C2 bit in SOC physical address */ +#define UMC_V6_7_PA_C2_BIT 17 +/* The R14 bit in SOC physical address */ +#define UMC_V6_7_PA_R14_BIT 34 /* UMC regiser per channel offset */ #define UMC_V6_7_PER_CHANNEL_OFFSET 0x400 -extern const struct amdgpu_umc_ras_funcs umc_v6_7_ras_funcs; + +/* XOR bit 20, 25, 34 of PA into CH4 bit (bit 12 of PA), + * hash bit is only effective when related setting is enabled + */ +#define CHANNEL_HASH(channel_idx, pa) (((channel_idx) >> 4) ^ \ + (((pa) >> 20) & 0x1ULL & adev->df.hash_status.hash_64k) ^ \ + (((pa) >> 25) & 0x1ULL & adev->df.hash_status.hash_2m) ^ \ + (((pa) >> 34) & 0x1ULL & adev->df.hash_status.hash_1g)) +#define SET_CHANNEL_HASH(channel_idx, pa) do { \ + (pa) &= ~(0x1ULL << UMC_V6_7_PA_CH4_BIT); \ + (pa) |= (CHANNEL_HASH(channel_idx, pa) << UMC_V6_7_PA_CH4_BIT); \ + } while (0) + +extern struct amdgpu_umc_ras umc_v6_7_ras; extern const uint32_t umc_v6_7_channel_idx_tbl_second[UMC_V6_7_UMC_INSTANCE_NUM][UMC_V6_7_CHANNEL_INSTANCE_NUM]; extern const uint32_t diff --git a/drivers/gpu/drm/amd/amdgpu/umc_v8_10.c b/drivers/gpu/drm/amd/amdgpu/umc_v8_10.c new file mode 100644 index 000000000000..91235df54e22 --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/umc_v8_10.c @@ -0,0 +1,369 @@ +/* + * Copyright 2022 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ +#include "umc_v8_10.h" +#include "amdgpu_ras.h" +#include "amdgpu_umc.h" +#include "amdgpu.h" +#include "umc/umc_8_10_0_offset.h" +#include "umc/umc_8_10_0_sh_mask.h" + +#define UMC_8_NODE_DIST 0x800000 +#define UMC_8_INST_DIST 0x4000 + +struct channelnum_map_colbit { + uint32_t channel_num; + uint32_t col_bit; +}; + +const struct channelnum_map_colbit umc_v8_10_channelnum_map_colbit_table[] = { + {24, 13}, + {20, 13}, + {16, 12}, + {14, 12}, + {12, 12}, + {10, 12}, + {6, 11}, +}; + +const uint32_t + umc_v8_10_channel_idx_tbl[] + [UMC_V8_10_UMC_INSTANCE_NUM] + [UMC_V8_10_CHANNEL_INSTANCE_NUM] = { + {{16, 18}, {17, 19}}, + {{15, 11}, {3, 7}}, + {{1, 5}, {13, 9}}, + {{23, 21}, {22, 20}}, + {{0, 4}, {12, 8}}, + {{14, 10}, {2, 6}} + }; + +static inline uint32_t get_umc_v8_10_reg_offset(struct amdgpu_device *adev, + uint32_t node_inst, + uint32_t umc_inst, + uint32_t ch_inst) +{ + return adev->umc.channel_offs * ch_inst + UMC_8_INST_DIST * umc_inst + + UMC_8_NODE_DIST * node_inst; +} + +static void umc_v8_10_clear_error_count_per_channel(struct amdgpu_device *adev, + uint32_t umc_reg_offset) +{ + uint32_t ecc_err_cnt_addr; + + ecc_err_cnt_addr = + SOC15_REG_OFFSET(UMC, 0, regUMCCH0_0_GeccErrCnt); + + /* clear error count */ + WREG32_PCIE((ecc_err_cnt_addr + umc_reg_offset) * 4, + UMC_V8_10_CE_CNT_INIT); +} + +static void umc_v8_10_clear_error_count(struct amdgpu_device *adev) +{ + uint32_t node_inst = 0; + uint32_t umc_inst = 0; + uint32_t ch_inst = 0; + uint32_t umc_reg_offset = 0; + + LOOP_UMC_EACH_NODE_INST_AND_CH(node_inst, umc_inst, ch_inst) { + umc_reg_offset = get_umc_v8_10_reg_offset(adev, + node_inst, + umc_inst, + ch_inst); + + umc_v8_10_clear_error_count_per_channel(adev, + umc_reg_offset); + } +} + +static void umc_v8_10_query_correctable_error_count(struct amdgpu_device *adev, + uint32_t umc_reg_offset, + unsigned long *error_count) +{ + uint64_t mc_umc_status; + uint32_t mc_umc_status_addr; + + /* UMC 8_10 registers */ + mc_umc_status_addr = + SOC15_REG_OFFSET(UMC, 0, regMCA_UMC_UMC0_MCUMC_STATUST0); + + /* Rely on MCUMC_STATUS for correctable error counter + * MCUMC_STATUS is a 64 bit register + */ + mc_umc_status = RREG64_PCIE((mc_umc_status_addr + umc_reg_offset) * 4); + if (REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, Val) == 1 && + REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, CECC) == 1) + *error_count += 1; +} + +static void umc_v8_10_query_uncorrectable_error_count(struct amdgpu_device *adev, + uint32_t umc_reg_offset, + unsigned long *error_count) +{ + uint64_t mc_umc_status; + uint32_t mc_umc_status_addr; + + mc_umc_status_addr = SOC15_REG_OFFSET(UMC, 0, regMCA_UMC_UMC0_MCUMC_STATUST0); + + /* Check the MCUMC_STATUS. */ + mc_umc_status = RREG64_PCIE((mc_umc_status_addr + umc_reg_offset) * 4); + if ((REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, Val) == 1) && + (REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, Deferred) == 1 || + REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, UECC) == 1 || + REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, PCC) == 1 || + REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, UC) == 1 || + REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, TCC) == 1)) + *error_count += 1; +} + +static void umc_v8_10_query_ras_error_count(struct amdgpu_device *adev, + void *ras_error_status) +{ + struct ras_err_data *err_data = (struct ras_err_data *)ras_error_status; + + uint32_t node_inst = 0; + uint32_t umc_inst = 0; + uint32_t ch_inst = 0; + uint32_t umc_reg_offset = 0; + + LOOP_UMC_EACH_NODE_INST_AND_CH(node_inst, umc_inst, ch_inst) { + umc_reg_offset = get_umc_v8_10_reg_offset(adev, + node_inst, + umc_inst, + ch_inst); + + umc_v8_10_query_correctable_error_count(adev, + umc_reg_offset, + &(err_data->ce_count)); + umc_v8_10_query_uncorrectable_error_count(adev, + umc_reg_offset, + &(err_data->ue_count)); + } + + umc_v8_10_clear_error_count(adev); +} + +static uint32_t umc_v8_10_get_col_bit(uint32_t channel_num) +{ + uint32_t t = 0; + + for (t = 0; t < ARRAY_SIZE(umc_v8_10_channelnum_map_colbit_table); t++) + if (channel_num == umc_v8_10_channelnum_map_colbit_table[t].channel_num) + return umc_v8_10_channelnum_map_colbit_table[t].col_bit; + + /* Failed to get col_bit. */ + return U32_MAX; +} + +/* + * Mapping normal address to soc physical address in swizzle mode. + */ +static int umc_v8_10_swizzle_mode_na_to_pa(struct amdgpu_device *adev, + uint32_t channel_idx, + uint64_t na, uint64_t *soc_pa) +{ + uint32_t channel_num = UMC_V8_10_TOTAL_CHANNEL_NUM(adev); + uint32_t col_bit = umc_v8_10_get_col_bit(channel_num); + uint64_t tmp_addr; + + if (col_bit == U32_MAX) + return -1; + + tmp_addr = SWIZZLE_MODE_TMP_ADDR(na, channel_num, channel_idx); + *soc_pa = SWIZZLE_MODE_ADDR_HI(tmp_addr, col_bit) | + SWIZZLE_MODE_ADDR_MID(na, col_bit) | + SWIZZLE_MODE_ADDR_LOW(tmp_addr, col_bit) | + SWIZZLE_MODE_ADDR_LSB(na); + + return 0; +} + +static void umc_v8_10_query_error_address(struct amdgpu_device *adev, + struct ras_err_data *err_data, + uint32_t umc_reg_offset, + uint32_t node_inst, + uint32_t ch_inst, + uint32_t umc_inst) +{ + uint64_t mc_umc_status_addr; + uint64_t mc_umc_status, err_addr; + uint64_t mc_umc_addrt0, na_err_addr_base; + uint64_t na_err_addr, retired_page_addr; + uint32_t channel_index, addr_lsb, col = 0; + int ret = 0; + + mc_umc_status_addr = + SOC15_REG_OFFSET(UMC, 0, regMCA_UMC_UMC0_MCUMC_STATUST0); + mc_umc_status = RREG64_PCIE((mc_umc_status_addr + umc_reg_offset) * 4); + + if (mc_umc_status == 0) + return; + + if (!err_data->err_addr) { + /* clear umc status */ + WREG64_PCIE((mc_umc_status_addr + umc_reg_offset) * 4, 0x0ULL); + return; + } + + channel_index = + adev->umc.channel_idx_tbl[node_inst * adev->umc.umc_inst_num * + adev->umc.channel_inst_num + + umc_inst * adev->umc.channel_inst_num + + ch_inst]; + + /* calculate error address if ue error is detected */ + if (REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, Val) == 1 && + REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, AddrV) == 1 && + REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, UECC) == 1) { + + mc_umc_addrt0 = SOC15_REG_OFFSET(UMC, 0, regMCA_UMC_UMC0_MCUMC_ADDRT0); + err_addr = RREG64_PCIE((mc_umc_addrt0 + umc_reg_offset) * 4); + err_addr = REG_GET_FIELD(err_addr, MCA_UMC_UMC0_MCUMC_ADDRT0, ErrorAddr); + + /* the lowest lsb bits should be ignored */ + addr_lsb = REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, AddrLsb); + err_addr &= ~((0x1ULL << addr_lsb) - 1); + na_err_addr_base = err_addr & ~(0x3ULL << UMC_V8_10_NA_C5_BIT); + + /* loop for all possibilities of [C6 C5] in normal address. */ + for (col = 0; col < UMC_V8_10_NA_COL_2BITS_POWER_OF_2_NUM; col++) { + na_err_addr = na_err_addr_base | (col << UMC_V8_10_NA_C5_BIT); + + /* Mapping normal error address to retired soc physical address. */ + ret = umc_v8_10_swizzle_mode_na_to_pa(adev, channel_index, + na_err_addr, &retired_page_addr); + if (ret) { + dev_err(adev->dev, "Failed to map pa from umc na.\n"); + break; + } + dev_info(adev->dev, "Error Address(PA): 0x%llx\n", + retired_page_addr); + amdgpu_umc_fill_error_record(err_data, na_err_addr, + retired_page_addr, channel_index, umc_inst); + } + } + + /* clear umc status */ + WREG64_PCIE((mc_umc_status_addr + umc_reg_offset) * 4, 0x0ULL); +} + +static void umc_v8_10_query_ras_error_address(struct amdgpu_device *adev, + void *ras_error_status) +{ + struct ras_err_data *err_data = (struct ras_err_data *)ras_error_status; + uint32_t node_inst = 0; + uint32_t umc_inst = 0; + uint32_t ch_inst = 0; + uint32_t umc_reg_offset = 0; + + LOOP_UMC_EACH_NODE_INST_AND_CH(node_inst, umc_inst, ch_inst) { + umc_reg_offset = get_umc_v8_10_reg_offset(adev, + node_inst, + umc_inst, + ch_inst); + + umc_v8_10_query_error_address(adev, + err_data, + umc_reg_offset, + node_inst, + ch_inst, + umc_inst); + } +} + +static void umc_v8_10_err_cnt_init_per_channel(struct amdgpu_device *adev, + uint32_t umc_reg_offset) +{ + uint32_t ecc_err_cnt_sel, ecc_err_cnt_sel_addr; + uint32_t ecc_err_cnt_addr; + + ecc_err_cnt_sel_addr = + SOC15_REG_OFFSET(UMC, 0, regUMCCH0_0_GeccErrCntSel); + ecc_err_cnt_addr = + SOC15_REG_OFFSET(UMC, 0, regUMCCH0_0_GeccErrCnt); + + ecc_err_cnt_sel = RREG32_PCIE((ecc_err_cnt_sel_addr + umc_reg_offset) * 4); + + /* set ce error interrupt type to APIC based interrupt */ + ecc_err_cnt_sel = REG_SET_FIELD(ecc_err_cnt_sel, UMCCH0_0_GeccErrCntSel, + GeccErrInt, 0x1); + WREG32_PCIE((ecc_err_cnt_sel_addr + umc_reg_offset) * 4, ecc_err_cnt_sel); + /* set error count to initial value */ + WREG32_PCIE((ecc_err_cnt_addr + umc_reg_offset) * 4, UMC_V8_10_CE_CNT_INIT); +} + +static void umc_v8_10_err_cnt_init(struct amdgpu_device *adev) +{ + uint32_t node_inst = 0; + uint32_t umc_inst = 0; + uint32_t ch_inst = 0; + uint32_t umc_reg_offset = 0; + + LOOP_UMC_EACH_NODE_INST_AND_CH(node_inst, umc_inst, ch_inst) { + umc_reg_offset = get_umc_v8_10_reg_offset(adev, + node_inst, + umc_inst, + ch_inst); + + umc_v8_10_err_cnt_init_per_channel(adev, umc_reg_offset); + } +} + +static uint32_t umc_v8_10_query_ras_poison_mode_per_channel( + struct amdgpu_device *adev, + uint32_t umc_reg_offset) +{ + uint32_t ecc_ctrl_addr, ecc_ctrl; + + ecc_ctrl_addr = + SOC15_REG_OFFSET(UMC, 0, regUMCCH0_0_GeccCtrl); + ecc_ctrl = RREG32_PCIE((ecc_ctrl_addr + + umc_reg_offset) * 4); + + return REG_GET_FIELD(ecc_ctrl, UMCCH0_0_GeccCtrl, UCFatalEn); +} + +static bool umc_v8_10_query_ras_poison_mode(struct amdgpu_device *adev) +{ + uint32_t umc_reg_offset = 0; + + /* Enabling fatal error in umc node0 instance0 channel0 will be + * considered as fatal error mode + */ + umc_reg_offset = get_umc_v8_10_reg_offset(adev, 0, 0, 0); + return !umc_v8_10_query_ras_poison_mode_per_channel(adev, umc_reg_offset); +} + +const struct amdgpu_ras_block_hw_ops umc_v8_10_ras_hw_ops = { + .query_ras_error_count = umc_v8_10_query_ras_error_count, + .query_ras_error_address = umc_v8_10_query_ras_error_address, +}; + +struct amdgpu_umc_ras umc_v8_10_ras = { + .ras_block = { + .hw_ops = &umc_v8_10_ras_hw_ops, + }, + .err_cnt_init = umc_v8_10_err_cnt_init, + .query_ras_poison_mode = umc_v8_10_query_ras_poison_mode, +}; diff --git a/drivers/gpu/drm/amd/amdgpu/umc_v8_10.h b/drivers/gpu/drm/amd/amdgpu/umc_v8_10.h new file mode 100644 index 000000000000..849ede88e111 --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/umc_v8_10.h @@ -0,0 +1,70 @@ +/* + * Copyright 2022 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ +#ifndef __UMC_V8_10_H__ +#define __UMC_V8_10_H__ + +#include "soc15_common.h" +#include "amdgpu.h" + +/* number of umc channel instance with memory map register access */ +#define UMC_V8_10_CHANNEL_INSTANCE_NUM 2 +/* number of umc instance with memory map register access */ +#define UMC_V8_10_UMC_INSTANCE_NUM 2 + +/* Total channel instances for all umc nodes */ +#define UMC_V8_10_TOTAL_CHANNEL_NUM(adev) \ + (UMC_V8_10_CHANNEL_INSTANCE_NUM * UMC_V8_10_UMC_INSTANCE_NUM * (adev)->umc.node_inst_num) + +/* UMC regiser per channel offset */ +#define UMC_V8_10_PER_CHANNEL_OFFSET 0x400 + +/* EccErrCnt max value */ +#define UMC_V8_10_CE_CNT_MAX 0xffff +/* umc ce interrupt threshold */ +#define UUMC_V8_10_CE_INT_THRESHOLD 0xffff +/* umc ce count initial value */ +#define UMC_V8_10_CE_CNT_INIT (UMC_V8_10_CE_CNT_MAX - UUMC_V8_10_CE_INT_THRESHOLD) + +#define UMC_V8_10_NA_COL_2BITS_POWER_OF_2_NUM 4 + +/* The C5 bit in NA address */ +#define UMC_V8_10_NA_C5_BIT 14 + +/* Map to swizzle mode address */ +#define SWIZZLE_MODE_TMP_ADDR(na, ch_num, ch_idx) \ + ((((na) >> 10) * (ch_num) + (ch_idx)) << 10) +#define SWIZZLE_MODE_ADDR_HI(addr, col_bit) \ + (((addr) >> ((col_bit) + 2)) << ((col_bit) + 2)) +#define SWIZZLE_MODE_ADDR_MID(na, col_bit) ((((na) >> 8) & 0x3) << (col_bit)) +#define SWIZZLE_MODE_ADDR_LOW(addr, col_bit) \ + ((((addr) >> 10) & ((0x1ULL << (col_bit - 8)) - 1)) << 8) +#define SWIZZLE_MODE_ADDR_LSB(na) ((na) & 0xFF) + +extern struct amdgpu_umc_ras umc_v8_10_ras; +extern const uint32_t + umc_v8_10_channel_idx_tbl[] + [UMC_V8_10_UMC_INSTANCE_NUM] + [UMC_V8_10_CHANNEL_INSTANCE_NUM]; + +#endif + diff --git a/drivers/gpu/drm/amd/amdgpu/umc_v8_7.c b/drivers/gpu/drm/amd/amdgpu/umc_v8_7.c index af59a35788e3..b717fdaa46e4 100644 --- a/drivers/gpu/drm/amd/amdgpu/umc_v8_7.c +++ b/drivers/gpu/drm/amd/amdgpu/umc_v8_7.c @@ -40,13 +40,143 @@ const uint32_t {9, 0}, {15, 6} }; -static inline uint32_t get_umc_8_reg_offset(struct amdgpu_device *adev, +static inline uint32_t get_umc_v8_7_reg_offset(struct amdgpu_device *adev, uint32_t umc_inst, uint32_t ch_inst) { return adev->umc.channel_offs*ch_inst + UMC_8_INST_DIST*umc_inst; } +static void umc_v8_7_ecc_info_query_correctable_error_count(struct amdgpu_device *adev, + uint32_t umc_inst, uint32_t ch_inst, + unsigned long *error_count) +{ + uint64_t mc_umc_status; + uint32_t eccinfo_table_idx; + struct amdgpu_ras *ras = amdgpu_ras_get_context(adev); + + eccinfo_table_idx = umc_inst * adev->umc.channel_inst_num + ch_inst; + + /* check for SRAM correctable error + * MCUMC_STATUS is a 64 bit register + */ + mc_umc_status = ras->umc_ecc.ecc[eccinfo_table_idx].mca_umc_status; + if (REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, Val) == 1 && + REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, CECC) == 1) + *error_count += 1; +} + +static void umc_v8_7_ecc_info_querry_uncorrectable_error_count(struct amdgpu_device *adev, + uint32_t umc_inst, uint32_t ch_inst, + unsigned long *error_count) +{ + uint64_t mc_umc_status; + uint32_t eccinfo_table_idx; + struct amdgpu_ras *ras = amdgpu_ras_get_context(adev); + + eccinfo_table_idx = umc_inst * adev->umc.channel_inst_num + ch_inst; + + /* check the MCUMC_STATUS */ + mc_umc_status = ras->umc_ecc.ecc[eccinfo_table_idx].mca_umc_status; + if ((REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, Val) == 1) && + (REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, Deferred) == 1 || + REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, UECC) == 1 || + REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, PCC) == 1 || + REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, UC) == 1 || + REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, TCC) == 1)) + *error_count += 1; +} + +static void umc_v8_7_ecc_info_query_ras_error_count(struct amdgpu_device *adev, + void *ras_error_status) +{ + struct ras_err_data *err_data = (struct ras_err_data *)ras_error_status; + + uint32_t umc_inst = 0; + uint32_t ch_inst = 0; + + /* TODO: driver needs to toggle DF Cstate to ensure + * safe access of UMC registers. Will add the protection + */ + LOOP_UMC_INST_AND_CH(umc_inst, ch_inst) { + umc_v8_7_ecc_info_query_correctable_error_count(adev, + umc_inst, ch_inst, + &(err_data->ce_count)); + umc_v8_7_ecc_info_querry_uncorrectable_error_count(adev, + umc_inst, ch_inst, + &(err_data->ue_count)); + } +} + +static void umc_v8_7_convert_error_address(struct amdgpu_device *adev, + struct ras_err_data *err_data, uint64_t err_addr, + uint32_t ch_inst, uint32_t umc_inst) +{ + uint64_t retired_page; + uint32_t channel_index; + + channel_index = + adev->umc.channel_idx_tbl[umc_inst * adev->umc.channel_inst_num + ch_inst]; + + /* translate umc channel address to soc pa, 3 parts are included */ + retired_page = ADDR_OF_4KB_BLOCK(err_addr) | + ADDR_OF_256B_BLOCK(channel_index) | + OFFSET_IN_256B_BLOCK(err_addr); + + amdgpu_umc_fill_error_record(err_data, err_addr, + retired_page, channel_index, umc_inst); +} + +static void umc_v8_7_ecc_info_query_error_address(struct amdgpu_device *adev, + struct ras_err_data *err_data, + uint32_t ch_inst, + uint32_t umc_inst) +{ + uint64_t mc_umc_status, err_addr; + uint32_t eccinfo_table_idx; + struct amdgpu_ras *ras = amdgpu_ras_get_context(adev); + + eccinfo_table_idx = umc_inst * adev->umc.channel_inst_num + ch_inst; + mc_umc_status = ras->umc_ecc.ecc[eccinfo_table_idx].mca_umc_status; + + if (mc_umc_status == 0) + return; + + if (!err_data->err_addr) + return; + + /* calculate error address if ue error is detected */ + if (REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, Val) == 1 && + REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, UECC) == 1) { + + err_addr = ras->umc_ecc.ecc[eccinfo_table_idx].mca_umc_addr; + err_addr = REG_GET_FIELD(err_addr, MCA_UMC_UMC0_MCUMC_ADDRT0, ErrorAddr); + + umc_v8_7_convert_error_address(adev, err_data, err_addr, + ch_inst, umc_inst); + } +} + +static void umc_v8_7_ecc_info_query_ras_error_address(struct amdgpu_device *adev, + void *ras_error_status) +{ + struct ras_err_data *err_data = (struct ras_err_data *)ras_error_status; + + uint32_t umc_inst = 0; + uint32_t ch_inst = 0; + + /* TODO: driver needs to toggle DF Cstate to ensure + * safe access of UMC resgisters. Will add the protection + * when firmware interface is ready + */ + LOOP_UMC_INST_AND_CH(umc_inst, ch_inst) { + umc_v8_7_ecc_info_query_error_address(adev, + err_data, + ch_inst, + umc_inst); + } +} + static void umc_v8_7_clear_error_count_per_channel(struct amdgpu_device *adev, uint32_t umc_reg_offset) { @@ -92,7 +222,7 @@ static void umc_v8_7_clear_error_count(struct amdgpu_device *adev) uint32_t umc_reg_offset = 0; LOOP_UMC_INST_AND_CH(umc_inst, ch_inst) { - umc_reg_offset = get_umc_8_reg_offset(adev, + umc_reg_offset = get_umc_v8_7_reg_offset(adev, umc_inst, ch_inst); @@ -178,7 +308,7 @@ static void umc_v8_7_query_ras_error_count(struct amdgpu_device *adev, uint32_t umc_reg_offset = 0; LOOP_UMC_INST_AND_CH(umc_inst, ch_inst) { - umc_reg_offset = get_umc_8_reg_offset(adev, + umc_reg_offset = get_umc_v8_7_reg_offset(adev, umc_inst, ch_inst); @@ -200,15 +330,12 @@ static void umc_v8_7_query_error_address(struct amdgpu_device *adev, uint32_t umc_inst) { uint32_t lsb, mc_umc_status_addr; - uint64_t mc_umc_status, err_addr, retired_page, mc_umc_addrt0; - struct eeprom_table_record *err_rec; - uint32_t channel_index = adev->umc.channel_idx_tbl[umc_inst * adev->umc.channel_inst_num + ch_inst]; + uint64_t mc_umc_status, err_addr, mc_umc_addrt0; mc_umc_status_addr = SOC15_REG_OFFSET(UMC, 0, mmMCA_UMC_UMC0_MCUMC_STATUST0); mc_umc_addrt0 = SOC15_REG_OFFSET(UMC, 0, mmMCA_UMC_UMC0_MCUMC_ADDRT0); - mc_umc_status = RREG64_PCIE((mc_umc_status_addr + umc_reg_offset) * 4); if (mc_umc_status == 0) @@ -220,12 +347,9 @@ static void umc_v8_7_query_error_address(struct amdgpu_device *adev, return; } - err_rec = &err_data->err_addr[err_data->err_addr_cnt]; - - /* calculate error address if ue/ce error is detected */ + /* calculate error address if ue error is detected */ if (REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, Val) == 1 && - (REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, UECC) == 1 || - REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, CECC) == 1)) { + REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, UECC) == 1) { err_addr = RREG64_PCIE((mc_umc_addrt0 + umc_reg_offset) * 4); /* the lowest lsb bits should be ignored */ @@ -233,25 +357,8 @@ static void umc_v8_7_query_error_address(struct amdgpu_device *adev, err_addr = REG_GET_FIELD(err_addr, MCA_UMC_UMC0_MCUMC_ADDRT0, ErrorAddr); err_addr &= ~((0x1ULL << lsb) - 1); - /* translate umc channel address to soc pa, 3 parts are included */ - retired_page = ADDR_OF_4KB_BLOCK(err_addr) | - ADDR_OF_256B_BLOCK(channel_index) | - OFFSET_IN_256B_BLOCK(err_addr); - - /* we only save ue error information currently, ce is skipped */ - if (REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, UECC) - == 1) { - err_rec->address = err_addr; - /* page frame address is saved */ - err_rec->retired_page = retired_page >> AMDGPU_GPU_PAGE_SHIFT; - err_rec->ts = (uint64_t)ktime_get_real_seconds(); - err_rec->err_type = AMDGPU_RAS_EEPROM_ERR_NON_RECOVERABLE; - err_rec->cu = 0; - err_rec->mem_channel = channel_index; - err_rec->mcumc_id = umc_inst; - - err_data->err_addr_cnt++; - } + umc_v8_7_convert_error_address(adev, err_data, err_addr, + ch_inst, umc_inst); } /* clear umc status */ @@ -268,7 +375,7 @@ static void umc_v8_7_query_ras_error_address(struct amdgpu_device *adev, uint32_t umc_reg_offset = 0; LOOP_UMC_INST_AND_CH(umc_inst, ch_inst) { - umc_reg_offset = get_umc_8_reg_offset(adev, + umc_reg_offset = get_umc_v8_7_reg_offset(adev, umc_inst, ch_inst); @@ -316,7 +423,7 @@ static void umc_v8_7_err_cnt_init(struct amdgpu_device *adev) uint32_t umc_reg_offset = 0; LOOP_UMC_INST_AND_CH(umc_inst, ch_inst) { - umc_reg_offset = get_umc_8_reg_offset(adev, + umc_reg_offset = get_umc_v8_7_reg_offset(adev, umc_inst, ch_inst); @@ -324,10 +431,16 @@ static void umc_v8_7_err_cnt_init(struct amdgpu_device *adev) } } -const struct amdgpu_umc_ras_funcs umc_v8_7_ras_funcs = { - .err_cnt_init = umc_v8_7_err_cnt_init, - .ras_late_init = amdgpu_umc_ras_late_init, - .ras_fini = amdgpu_umc_ras_fini, +const struct amdgpu_ras_block_hw_ops umc_v8_7_ras_hw_ops = { .query_ras_error_count = umc_v8_7_query_ras_error_count, .query_ras_error_address = umc_v8_7_query_ras_error_address, }; + +struct amdgpu_umc_ras umc_v8_7_ras = { + .ras_block = { + .hw_ops = &umc_v8_7_ras_hw_ops, + }, + .err_cnt_init = umc_v8_7_err_cnt_init, + .ecc_info_query_ras_error_count = umc_v8_7_ecc_info_query_ras_error_count, + .ecc_info_query_ras_error_address = umc_v8_7_ecc_info_query_ras_error_address, +}; diff --git a/drivers/gpu/drm/amd/amdgpu/umc_v8_7.h b/drivers/gpu/drm/amd/amdgpu/umc_v8_7.h index 37e6dc7c28e0..dd4993f5f78f 100644 --- a/drivers/gpu/drm/amd/amdgpu/umc_v8_7.h +++ b/drivers/gpu/drm/amd/amdgpu/umc_v8_7.h @@ -44,7 +44,7 @@ /* umc ce count initial value */ #define UMC_V8_7_CE_CNT_INIT (UMC_V8_7_CE_CNT_MAX - UMC_V8_7_CE_INT_THRESHOLD) -extern const struct amdgpu_umc_ras_funcs umc_v8_7_ras_funcs; +extern struct amdgpu_umc_ras umc_v8_7_ras; extern const uint32_t umc_v8_7_channel_idx_tbl[UMC_V8_7_UMC_INSTANCE_NUM][UMC_V8_7_CHANNEL_INSTANCE_NUM]; diff --git a/drivers/gpu/drm/amd/amdgpu/uvd_v5_0.c b/drivers/gpu/drm/amd/amdgpu/uvd_v5_0.c index 563493d1f830..d7e31e48a2b8 100644 --- a/drivers/gpu/drm/amd/amdgpu/uvd_v5_0.c +++ b/drivers/gpu/drm/amd/amdgpu/uvd_v5_0.c @@ -833,7 +833,7 @@ out: return ret; } -static void uvd_v5_0_get_clockgating_state(void *handle, u32 *flags) +static void uvd_v5_0_get_clockgating_state(void *handle, u64 *flags) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; int data; diff --git a/drivers/gpu/drm/amd/amdgpu/uvd_v6_0.c b/drivers/gpu/drm/amd/amdgpu/uvd_v6_0.c index 2d558c2f417d..375c440957dc 100644 --- a/drivers/gpu/drm/amd/amdgpu/uvd_v6_0.c +++ b/drivers/gpu/drm/amd/amdgpu/uvd_v6_0.c @@ -1494,7 +1494,7 @@ out: return ret; } -static void uvd_v6_0_get_clockgating_state(void *handle, u32 *flags) +static void uvd_v6_0_get_clockgating_state(void *handle, u64 *flags) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; int data; diff --git a/drivers/gpu/drm/amd/amdgpu/uvd_v7_0.c b/drivers/gpu/drm/amd/amdgpu/uvd_v7_0.c index b483f03b4591..e668b3baa8c6 100644 --- a/drivers/gpu/drm/amd/amdgpu/uvd_v7_0.c +++ b/drivers/gpu/drm/amd/amdgpu/uvd_v7_0.c @@ -25,6 +25,7 @@ #include "amdgpu.h" #include "amdgpu_uvd.h" +#include "amdgpu_cs.h" #include "soc15.h" #include "soc15d.h" #include "soc15_common.h" @@ -117,7 +118,7 @@ static uint64_t uvd_v7_0_enc_ring_get_wptr(struct amdgpu_ring *ring) struct amdgpu_device *adev = ring->adev; if (ring->use_doorbell) - return adev->wb.wb[ring->wptr_offs]; + return *ring->wptr_cpu_addr; if (ring == &adev->uvd.inst[ring->me].ring_enc[0]) return RREG32_SOC15(UVD, ring->me, mmUVD_RB_WPTR); @@ -152,7 +153,7 @@ static void uvd_v7_0_enc_ring_set_wptr(struct amdgpu_ring *ring) if (ring->use_doorbell) { /* XXX check if swapping is necessary on BE */ - adev->wb.wb[ring->wptr_offs] = lower_32_bits(ring->wptr); + *ring->wptr_cpu_addr = lower_32_bits(ring->wptr); WDOORBELL32(ring->doorbell_index, lower_32_bits(ring->wptr)); return; } @@ -753,7 +754,7 @@ static int uvd_v7_0_mmsch_start(struct amdgpu_device *adev, if (adev->uvd.harvest_config & (1 << i)) continue; WDOORBELL32(adev->uvd.inst[i].ring_enc[0].doorbell_index, 0); - adev->wb.wb[adev->uvd.inst[i].ring_enc[0].wptr_offs] = 0; + *adev->uvd.inst[i].ring_enc[0].wptr_cpu_addr = 0; adev->uvd.inst[i].ring_enc[0].wptr = 0; adev->uvd.inst[i].ring_enc[0].wptr_old = 0; } @@ -1275,14 +1276,15 @@ static int uvd_v7_0_ring_test_ring(struct amdgpu_ring *ring) * uvd_v7_0_ring_patch_cs_in_place - Patch the IB for command submission. * * @p: the CS parser with the IBs - * @ib_idx: which IB to patch + * @job: which job this ib is in + * @ib: which IB to patch * */ static int uvd_v7_0_ring_patch_cs_in_place(struct amdgpu_cs_parser *p, - uint32_t ib_idx) + struct amdgpu_job *job, + struct amdgpu_ib *ib) { - struct amdgpu_ring *ring = to_amdgpu_ring(p->entity->rq->sched); - struct amdgpu_ib *ib = &p->job->ibs[ib_idx]; + struct amdgpu_ring *ring = to_amdgpu_ring(job->base.sched); unsigned i; /* No patching necessary for the first instance */ @@ -1290,12 +1292,12 @@ static int uvd_v7_0_ring_patch_cs_in_place(struct amdgpu_cs_parser *p, return 0; for (i = 0; i < ib->length_dw; i += 2) { - uint32_t reg = amdgpu_get_ib_value(p, ib_idx, i); + uint32_t reg = amdgpu_ib_get_value(ib, i); reg -= p->adev->reg_offset[UVD_HWIP][0][1]; reg += p->adev->reg_offset[UVD_HWIP][1][1]; - amdgpu_set_ib_value(p, ib_idx, i, reg); + amdgpu_ib_set_value(ib, i, reg); } return 0; } diff --git a/drivers/gpu/drm/amd/amdgpu/vce_v3_0.c b/drivers/gpu/drm/amd/amdgpu/vce_v3_0.c index 142e291983b4..8def62c83ffd 100644 --- a/drivers/gpu/drm/amd/amdgpu/vce_v3_0.c +++ b/drivers/gpu/drm/amd/amdgpu/vce_v3_0.c @@ -831,7 +831,7 @@ out: return ret; } -static void vce_v3_0_get_clockgating_state(void *handle, u32 *flags) +static void vce_v3_0_get_clockgating_state(void *handle, u64 *flags) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; int data; diff --git a/drivers/gpu/drm/amd/amdgpu/vce_v4_0.c b/drivers/gpu/drm/amd/amdgpu/vce_v4_0.c index d1fc4e0b8265..66cd3d11aa4b 100644 --- a/drivers/gpu/drm/amd/amdgpu/vce_v4_0.c +++ b/drivers/gpu/drm/amd/amdgpu/vce_v4_0.c @@ -83,7 +83,7 @@ static uint64_t vce_v4_0_ring_get_wptr(struct amdgpu_ring *ring) struct amdgpu_device *adev = ring->adev; if (ring->use_doorbell) - return adev->wb.wb[ring->wptr_offs]; + return *ring->wptr_cpu_addr; if (ring->me == 0) return RREG32(SOC15_REG_OFFSET(VCE, 0, mmVCE_RB_WPTR)); @@ -106,7 +106,7 @@ static void vce_v4_0_ring_set_wptr(struct amdgpu_ring *ring) if (ring->use_doorbell) { /* XXX check if swapping is necessary on BE */ - adev->wb.wb[ring->wptr_offs] = lower_32_bits(ring->wptr); + *ring->wptr_cpu_addr = lower_32_bits(ring->wptr); WDOORBELL32(ring->doorbell_index, lower_32_bits(ring->wptr)); return; } @@ -177,7 +177,7 @@ static int vce_v4_0_mmsch_start(struct amdgpu_device *adev, WREG32(SOC15_REG_OFFSET(VCE, 0, mmVCE_MMSCH_VF_MAILBOX_RESP), 0); WDOORBELL32(adev->vce.ring[0].doorbell_index, 0); - adev->wb.wb[adev->vce.ring[0].wptr_offs] = 0; + *adev->vce.ring[0].wptr_cpu_addr = 0; adev->vce.ring[0].wptr = 0; adev->vce.ring[0].wptr_old = 0; diff --git a/drivers/gpu/drm/amd/amdgpu/vcn_sw_ring.c b/drivers/gpu/drm/amd/amdgpu/vcn_sw_ring.c new file mode 100644 index 000000000000..1ceda3d0cd5b --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/vcn_sw_ring.c @@ -0,0 +1,86 @@ +/* + * Copyright 2022 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#include "amdgpu.h" +#include "vcn_sw_ring.h" + +void vcn_dec_sw_ring_emit_fence(struct amdgpu_ring *ring, u64 addr, + u64 seq, uint32_t flags) +{ + WARN_ON(flags & AMDGPU_FENCE_FLAG_64BIT); + + amdgpu_ring_write(ring, VCN_DEC_SW_CMD_FENCE); + amdgpu_ring_write(ring, addr); + amdgpu_ring_write(ring, upper_32_bits(addr)); + amdgpu_ring_write(ring, seq); + amdgpu_ring_write(ring, VCN_DEC_SW_CMD_TRAP); +} + +void vcn_dec_sw_ring_insert_end(struct amdgpu_ring *ring) +{ + amdgpu_ring_write(ring, VCN_DEC_SW_CMD_END); +} + +void vcn_dec_sw_ring_emit_ib(struct amdgpu_ring *ring, struct amdgpu_job *job, + struct amdgpu_ib *ib, uint32_t flags) +{ + uint32_t vmid = AMDGPU_JOB_GET_VMID(job); + + amdgpu_ring_write(ring, VCN_DEC_SW_CMD_IB); + amdgpu_ring_write(ring, vmid); + amdgpu_ring_write(ring, lower_32_bits(ib->gpu_addr)); + amdgpu_ring_write(ring, upper_32_bits(ib->gpu_addr)); + amdgpu_ring_write(ring, ib->length_dw); +} + +void vcn_dec_sw_ring_emit_reg_wait(struct amdgpu_ring *ring, uint32_t reg, + uint32_t val, uint32_t mask) +{ + amdgpu_ring_write(ring, VCN_DEC_SW_CMD_REG_WAIT); + amdgpu_ring_write(ring, reg << 2); + amdgpu_ring_write(ring, mask); + amdgpu_ring_write(ring, val); +} + +void vcn_dec_sw_ring_emit_vm_flush(struct amdgpu_ring *ring, + uint32_t vmid, uint64_t pd_addr) +{ + struct amdgpu_vmhub *hub = &ring->adev->vmhub[ring->funcs->vmhub]; + uint32_t data0, data1, mask; + + pd_addr = amdgpu_gmc_emit_flush_gpu_tlb(ring, vmid, pd_addr); + + /* wait for register write */ + data0 = hub->ctx0_ptb_addr_lo32 + vmid * hub->ctx_addr_distance; + data1 = lower_32_bits(pd_addr); + mask = 0xffffffff; + vcn_dec_sw_ring_emit_reg_wait(ring, data0, data1, mask); +} + +void vcn_dec_sw_ring_emit_wreg(struct amdgpu_ring *ring, uint32_t reg, + uint32_t val) +{ + amdgpu_ring_write(ring, VCN_DEC_SW_CMD_REG_WRITE); + amdgpu_ring_write(ring, reg << 2); + amdgpu_ring_write(ring, val); +} diff --git a/drivers/gpu/drm/amd/amdgpu/vcn_sw_ring.h b/drivers/gpu/drm/amd/amdgpu/vcn_sw_ring.h new file mode 100644 index 000000000000..7e775725f120 --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/vcn_sw_ring.h @@ -0,0 +1,44 @@ +/* + * Copyright 2022 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#ifndef __VCN_SW_RING_H__ +#define __VCN_SW_RING_H__ + +#define VCN_SW_RING_EMIT_FRAME_SIZE \ + (4 + /* vcn_dec_sw_ring_emit_vm_flush */ \ + 5 + 5 + /* vcn_dec_sw_ring_emit_fence x2 vm fence */ \ + 1) /* vcn_dec_sw_ring_insert_end */ + +void vcn_dec_sw_ring_emit_fence(struct amdgpu_ring *ring, u64 addr, + u64 seq, uint32_t flags); +void vcn_dec_sw_ring_insert_end(struct amdgpu_ring *ring); +void vcn_dec_sw_ring_emit_ib(struct amdgpu_ring *ring, struct amdgpu_job *job, + struct amdgpu_ib *ib, uint32_t flags); +void vcn_dec_sw_ring_emit_reg_wait(struct amdgpu_ring *ring, uint32_t reg, + uint32_t val, uint32_t mask); +void vcn_dec_sw_ring_emit_vm_flush(struct amdgpu_ring *ring, + uint32_t vmid, uint64_t pd_addr); +void vcn_dec_sw_ring_emit_wreg(struct amdgpu_ring *ring, uint32_t reg, + uint32_t val); + +#endif /* __VCN_SW_RING_H__ */ diff --git a/drivers/gpu/drm/amd/amdgpu/vcn_v1_0.c b/drivers/gpu/drm/amd/amdgpu/vcn_v1_0.c index 3799226defc0..f0fbcda76f5e 100644 --- a/drivers/gpu/drm/amd/amdgpu/vcn_v1_0.c +++ b/drivers/gpu/drm/amd/amdgpu/vcn_v1_0.c @@ -24,6 +24,7 @@ #include <linux/firmware.h> #include "amdgpu.h" +#include "amdgpu_cs.h" #include "amdgpu_vcn.h" #include "amdgpu_pm.h" #include "soc15.h" @@ -148,6 +149,13 @@ static int vcn_v1_0_sw_init(void *handle) adev->vcn.pause_dpg_mode = vcn_v1_0_pause_dpg_mode; + if (amdgpu_vcnfw_log) { + volatile struct amdgpu_fw_shared *fw_shared = adev->vcn.inst->fw_shared.cpu_addr; + + fw_shared->present_flag_0 = 0; + amdgpu_vcn_fwlog_init(adev->vcn.inst); + } + r = jpeg_v1_0_sw_init(handle); return r; @@ -1095,13 +1103,8 @@ static int vcn_v1_0_start_dpg_mode(struct amdgpu_device *adev) static int vcn_v1_0_start(struct amdgpu_device *adev) { - int r; - - if (adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG) - r = vcn_v1_0_start_dpg_mode(adev); - else - r = vcn_v1_0_start_spg_mode(adev); - return r; + return (adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG) ? + vcn_v1_0_start_dpg_mode(adev) : vcn_v1_0_start_spg_mode(adev); } /** @@ -1898,15 +1901,86 @@ static const struct amd_ip_funcs vcn_v1_0_ip_funcs = { .set_powergating_state = vcn_v1_0_set_powergating_state, }; +/* + * It is a hardware issue that VCN can't handle a GTT TMZ buffer on + * CHIP_RAVEN series ASIC. Move such a GTT TMZ buffer to VRAM domain + * before command submission as a workaround. + */ +static int vcn_v1_0_validate_bo(struct amdgpu_cs_parser *parser, + struct amdgpu_job *job, + uint64_t addr) +{ + struct ttm_operation_ctx ctx = { false, false }; + struct amdgpu_fpriv *fpriv = parser->filp->driver_priv; + struct amdgpu_vm *vm = &fpriv->vm; + struct amdgpu_bo_va_mapping *mapping; + struct amdgpu_bo *bo; + int r; + + addr &= AMDGPU_GMC_HOLE_MASK; + if (addr & 0x7) { + DRM_ERROR("VCN messages must be 8 byte aligned!\n"); + return -EINVAL; + } + + mapping = amdgpu_vm_bo_lookup_mapping(vm, addr/AMDGPU_GPU_PAGE_SIZE); + if (!mapping || !mapping->bo_va || !mapping->bo_va->base.bo) + return -EINVAL; + + bo = mapping->bo_va->base.bo; + if (!(bo->flags & AMDGPU_GEM_CREATE_ENCRYPTED)) + return 0; + + amdgpu_bo_placement_from_domain(bo, AMDGPU_GEM_DOMAIN_VRAM); + r = ttm_bo_validate(&bo->tbo, &bo->placement, &ctx); + if (r) { + DRM_ERROR("Failed to validate the VCN message BO (%d)!\n", r); + return r; + } + + return r; +} + +static int vcn_v1_0_ring_patch_cs_in_place(struct amdgpu_cs_parser *p, + struct amdgpu_job *job, + struct amdgpu_ib *ib) +{ + uint32_t msg_lo = 0, msg_hi = 0; + int i, r; + + if (!(ib->flags & AMDGPU_IB_FLAGS_SECURE)) + return 0; + + for (i = 0; i < ib->length_dw; i += 2) { + uint32_t reg = amdgpu_ib_get_value(ib, i); + uint32_t val = amdgpu_ib_get_value(ib, i + 1); + + if (reg == PACKET0(p->adev->vcn.internal.data0, 0)) { + msg_lo = val; + } else if (reg == PACKET0(p->adev->vcn.internal.data1, 0)) { + msg_hi = val; + } else if (reg == PACKET0(p->adev->vcn.internal.cmd, 0)) { + r = vcn_v1_0_validate_bo(p, job, + ((u64)msg_hi) << 32 | msg_lo); + if (r) + return r; + } + } + + return 0; +} + static const struct amdgpu_ring_funcs vcn_v1_0_dec_ring_vm_funcs = { .type = AMDGPU_RING_TYPE_VCN_DEC, .align_mask = 0xf, .support_64bit_ptrs = false, .no_user_fence = true, + .secure_submission_supported = true, .vmhub = AMDGPU_MMHUB_0, .get_rptr = vcn_v1_0_dec_ring_get_rptr, .get_wptr = vcn_v1_0_dec_ring_get_wptr, .set_wptr = vcn_v1_0_dec_ring_set_wptr, + .patch_cs_in_place = vcn_v1_0_ring_patch_cs_in_place, .emit_frame_size = 6 + 6 + /* hdp invalidate / flush */ SOC15_FLUSH_GPU_TLB_NUM_WREG * 6 + diff --git a/drivers/gpu/drm/amd/amdgpu/vcn_v2_0.c b/drivers/gpu/drm/amd/amdgpu/vcn_v2_0.c index 313fc1b53999..08871bad9994 100644 --- a/drivers/gpu/drm/amd/amdgpu/vcn_v2_0.c +++ b/drivers/gpu/drm/amd/amdgpu/vcn_v2_0.c @@ -37,6 +37,9 @@ #include "vcn/vcn_2_0_0_sh_mask.h" #include "ivsrcid/vcn/irqsrcs_vcn_2_0.h" +#define VCN_VID_SOC_ADDRESS_2_0 0x1fa00 +#define VCN1_VID_SOC_ADDRESS_3_0 0x48200 + #define mmUVD_CONTEXT_ID_INTERNAL_OFFSET 0x1fd #define mmUVD_GPCOM_VCPU_CMD_INTERNAL_OFFSET 0x503 #define mmUVD_GPCOM_VCPU_DATA0_INTERNAL_OFFSET 0x504 @@ -172,8 +175,12 @@ static int vcn_v2_0_sw_init(void *handle) if (r) return r; - fw_shared = adev->vcn.inst->fw_shared_cpu_addr; + fw_shared = adev->vcn.inst->fw_shared.cpu_addr; fw_shared->present_flag_0 = cpu_to_le32(AMDGPU_VCN_MULTI_QUEUE_FLAG); + + if (amdgpu_vcnfw_log) + amdgpu_vcn_fwlog_init(adev->vcn.inst); + return 0; } @@ -188,7 +195,7 @@ static int vcn_v2_0_sw_fini(void *handle) { int r, idx; struct amdgpu_device *adev = (struct amdgpu_device *)handle; - volatile struct amdgpu_fw_shared *fw_shared = adev->vcn.inst->fw_shared_cpu_addr; + volatile struct amdgpu_fw_shared *fw_shared = adev->vcn.inst->fw_shared.cpu_addr; if (drm_dev_enter(adev_to_drm(adev), &idx)) { fw_shared->present_flag_0 = 0; @@ -364,9 +371,9 @@ static void vcn_v2_0_mc_resume(struct amdgpu_device *adev) /* non-cache window */ WREG32_SOC15(UVD, 0, mmUVD_LMI_VCPU_NC0_64BIT_BAR_LOW, - lower_32_bits(adev->vcn.inst->fw_shared_gpu_addr)); + lower_32_bits(adev->vcn.inst->fw_shared.gpu_addr)); WREG32_SOC15(UVD, 0, mmUVD_LMI_VCPU_NC0_64BIT_BAR_HIGH, - upper_32_bits(adev->vcn.inst->fw_shared_gpu_addr)); + upper_32_bits(adev->vcn.inst->fw_shared.gpu_addr)); WREG32_SOC15(UVD, 0, mmUVD_VCPU_NONCACHE_OFFSET0, 0); WREG32_SOC15(UVD, 0, mmUVD_VCPU_NONCACHE_SIZE0, AMDGPU_GPU_PAGE_ALIGN(sizeof(struct amdgpu_fw_shared))); @@ -455,10 +462,10 @@ static void vcn_v2_0_mc_resume_dpg_mode(struct amdgpu_device *adev, bool indirec /* non-cache window */ WREG32_SOC15_DPG_MODE(0, SOC15_DPG_MODE_OFFSET( UVD, 0, mmUVD_LMI_VCPU_NC0_64BIT_BAR_LOW), - lower_32_bits(adev->vcn.inst->fw_shared_gpu_addr), 0, indirect); + lower_32_bits(adev->vcn.inst->fw_shared.gpu_addr), 0, indirect); WREG32_SOC15_DPG_MODE(0, SOC15_DPG_MODE_OFFSET( UVD, 0, mmUVD_LMI_VCPU_NC0_64BIT_BAR_HIGH), - upper_32_bits(adev->vcn.inst->fw_shared_gpu_addr), 0, indirect); + upper_32_bits(adev->vcn.inst->fw_shared.gpu_addr), 0, indirect); WREG32_SOC15_DPG_MODE(0, SOC15_DPG_MODE_OFFSET( UVD, 0, mmUVD_VCPU_NONCACHE_OFFSET0), 0, 0, indirect); WREG32_SOC15_DPG_MODE(0, SOC15_DPG_MODE_OFFSET( @@ -784,7 +791,7 @@ static void vcn_v2_0_enable_static_power_gating(struct amdgpu_device *adev) static int vcn_v2_0_start_dpg_mode(struct amdgpu_device *adev, bool indirect) { - volatile struct amdgpu_fw_shared *fw_shared = adev->vcn.inst->fw_shared_cpu_addr; + volatile struct amdgpu_fw_shared *fw_shared = adev->vcn.inst->fw_shared.cpu_addr; struct amdgpu_ring *ring = &adev->vcn.inst->ring_dec; uint32_t rb_bufsz, tmp; @@ -921,7 +928,7 @@ static int vcn_v2_0_start_dpg_mode(struct amdgpu_device *adev, bool indirect) static int vcn_v2_0_start(struct amdgpu_device *adev) { - volatile struct amdgpu_fw_shared *fw_shared = adev->vcn.inst->fw_shared_cpu_addr; + volatile struct amdgpu_fw_shared *fw_shared = adev->vcn.inst->fw_shared.cpu_addr; struct amdgpu_ring *ring = &adev->vcn.inst->ring_dec; uint32_t rb_bufsz, tmp; uint32_t lmi_swap_cntl; @@ -1094,8 +1101,10 @@ static int vcn_v2_0_start(struct amdgpu_device *adev) static int vcn_v2_0_stop_dpg_mode(struct amdgpu_device *adev) { + struct dpg_pause_state state = {.fw_based = VCN_DPG_STATE__UNPAUSE}; uint32_t tmp; + vcn_v2_0_pause_dpg_mode(adev, 0, &state); /* Wait for power status to be 1 */ SOC15_WAIT_ON_RREG(UVD, 0, mmUVD_POWER_STATUS, 1, UVD_POWER_STATUS__UVD_POWER_STATUS_MASK); @@ -1207,7 +1216,7 @@ static int vcn_v2_0_pause_dpg_mode(struct amdgpu_device *adev, UVD_POWER_STATUS__UVD_POWER_STATUS_MASK); if (!ret_code) { - volatile struct amdgpu_fw_shared *fw_shared = adev->vcn.inst->fw_shared_cpu_addr; + volatile struct amdgpu_fw_shared *fw_shared = adev->vcn.inst->fw_shared.cpu_addr; /* pause DPG */ reg_data |= UVD_DPG_PAUSE__NJ_PAUSE_DPG_REQ_MASK; WREG32_SOC15(UVD, 0, mmUVD_DPG_PAUSE, reg_data); @@ -1330,7 +1339,7 @@ static uint64_t vcn_v2_0_dec_ring_get_wptr(struct amdgpu_ring *ring) struct amdgpu_device *adev = ring->adev; if (ring->use_doorbell) - return adev->wb.wb[ring->wptr_offs]; + return *ring->wptr_cpu_addr; else return RREG32_SOC15(UVD, 0, mmUVD_RBC_RB_WPTR); } @@ -1351,7 +1360,7 @@ static void vcn_v2_0_dec_ring_set_wptr(struct amdgpu_ring *ring) lower_32_bits(ring->wptr) | 0x80000000); if (ring->use_doorbell) { - adev->wb.wb[ring->wptr_offs] = lower_32_bits(ring->wptr); + *ring->wptr_cpu_addr = lower_32_bits(ring->wptr); WDOORBELL32(ring->doorbell_index, lower_32_bits(ring->wptr)); } else { WREG32_SOC15(UVD, 0, mmUVD_RBC_RB_WPTR, lower_32_bits(ring->wptr)); @@ -1559,12 +1568,12 @@ static uint64_t vcn_v2_0_enc_ring_get_wptr(struct amdgpu_ring *ring) if (ring == &adev->vcn.inst->ring_enc[0]) { if (ring->use_doorbell) - return adev->wb.wb[ring->wptr_offs]; + return *ring->wptr_cpu_addr; else return RREG32_SOC15(UVD, 0, mmUVD_RB_WPTR); } else { if (ring->use_doorbell) - return adev->wb.wb[ring->wptr_offs]; + return *ring->wptr_cpu_addr; else return RREG32_SOC15(UVD, 0, mmUVD_RB_WPTR2); } @@ -1583,14 +1592,14 @@ static void vcn_v2_0_enc_ring_set_wptr(struct amdgpu_ring *ring) if (ring == &adev->vcn.inst->ring_enc[0]) { if (ring->use_doorbell) { - adev->wb.wb[ring->wptr_offs] = lower_32_bits(ring->wptr); + *ring->wptr_cpu_addr = lower_32_bits(ring->wptr); WDOORBELL32(ring->doorbell_index, lower_32_bits(ring->wptr)); } else { WREG32_SOC15(UVD, 0, mmUVD_RB_WPTR, lower_32_bits(ring->wptr)); } } else { if (ring->use_doorbell) { - adev->wb.wb[ring->wptr_offs] = lower_32_bits(ring->wptr); + *ring->wptr_cpu_addr = lower_32_bits(ring->wptr); WDOORBELL32(ring->doorbell_index, lower_32_bits(ring->wptr)); } else { WREG32_SOC15(UVD, 0, mmUVD_RB_WPTR2, lower_32_bits(ring->wptr)); @@ -2003,6 +2012,7 @@ static const struct amd_ip_funcs vcn_v2_0_ip_funcs = { static const struct amdgpu_ring_funcs vcn_v2_0_dec_ring_vm_funcs = { .type = AMDGPU_RING_TYPE_VCN_DEC, .align_mask = 0xf, + .secure_submission_supported = true, .vmhub = AMDGPU_MMHUB_0, .get_rptr = vcn_v2_0_dec_ring_get_rptr, .get_wptr = vcn_v2_0_dec_ring_get_wptr, diff --git a/drivers/gpu/drm/amd/amdgpu/vcn_v2_5.c b/drivers/gpu/drm/amd/amdgpu/vcn_v2_5.c index 44fc4c218433..8a7006d62a87 100644 --- a/drivers/gpu/drm/amd/amdgpu/vcn_v2_5.c +++ b/drivers/gpu/drm/amd/amdgpu/vcn_v2_5.c @@ -31,11 +31,15 @@ #include "soc15d.h" #include "vcn_v2_0.h" #include "mmsch_v1_0.h" +#include "vcn_v2_5.h" #include "vcn/vcn_2_5_offset.h" #include "vcn/vcn_2_5_sh_mask.h" #include "ivsrcid/vcn/irqsrcs_vcn_2_0.h" +#define VCN_VID_SOC_ADDRESS_2_0 0x1fa00 +#define VCN1_VID_SOC_ADDRESS_3_0 0x48200 + #define mmUVD_CONTEXT_ID_INTERNAL_OFFSET 0x27 #define mmUVD_GPCOM_VCPU_CMD_INTERNAL_OFFSET 0x0f #define mmUVD_GPCOM_VCPU_DATA0_INTERNAL_OFFSET 0x10 @@ -59,6 +63,7 @@ static int vcn_v2_5_set_powergating_state(void *handle, static int vcn_v2_5_pause_dpg_mode(struct amdgpu_device *adev, int inst_idx, struct dpg_pause_state *new_state); static int vcn_v2_5_sriov_start(struct amdgpu_device *adev); +static void vcn_v2_5_set_ras_funcs(struct amdgpu_device *adev); static int amdgpu_ih_clientid_vcns[] = { SOC15_IH_CLIENTID_VCN, @@ -100,6 +105,7 @@ static int vcn_v2_5_early_init(void *handle) vcn_v2_5_set_dec_ring_funcs(adev); vcn_v2_5_set_enc_ring_funcs(adev); vcn_v2_5_set_irq_funcs(adev); + vcn_v2_5_set_ras_funcs(adev); return 0; } @@ -133,6 +139,12 @@ static int vcn_v2_5_sw_init(void *handle) if (r) return r; } + + /* VCN POISON TRAP */ + r = amdgpu_irq_add_id(adev, amdgpu_ih_clientid_vcns[j], + VCN_2_6__SRCID_UVD_POISON, &adev->vcn.inst[j].irq); + if (r) + return r; } r = amdgpu_vcn_sw_init(adev); @@ -196,8 +208,11 @@ static int vcn_v2_5_sw_init(void *handle) return r; } - fw_shared = adev->vcn.inst[j].fw_shared_cpu_addr; + fw_shared = adev->vcn.inst[j].fw_shared.cpu_addr; fw_shared->present_flag_0 = cpu_to_le32(AMDGPU_VCN_MULTI_QUEUE_FLAG); + + if (amdgpu_vcnfw_log) + amdgpu_vcn_fwlog_init(&adev->vcn.inst[i]); } if (amdgpu_sriov_vf(adev)) { @@ -229,7 +244,7 @@ static int vcn_v2_5_sw_fini(void *handle) for (i = 0; i < adev->vcn.num_vcn_inst; i++) { if (adev->vcn.harvest_config & (1 << i)) continue; - fw_shared = adev->vcn.inst[i].fw_shared_cpu_addr; + fw_shared = adev->vcn.inst[i].fw_shared.cpu_addr; fw_shared->present_flag_0 = 0; } drm_dev_exit(idx); @@ -423,9 +438,9 @@ static void vcn_v2_5_mc_resume(struct amdgpu_device *adev) /* non-cache window */ WREG32_SOC15(VCN, i, mmUVD_LMI_VCPU_NC0_64BIT_BAR_LOW, - lower_32_bits(adev->vcn.inst[i].fw_shared_gpu_addr)); + lower_32_bits(adev->vcn.inst[i].fw_shared.gpu_addr)); WREG32_SOC15(VCN, i, mmUVD_LMI_VCPU_NC0_64BIT_BAR_HIGH, - upper_32_bits(adev->vcn.inst[i].fw_shared_gpu_addr)); + upper_32_bits(adev->vcn.inst[i].fw_shared.gpu_addr)); WREG32_SOC15(VCN, i, mmUVD_VCPU_NONCACHE_OFFSET0, 0); WREG32_SOC15(VCN, i, mmUVD_VCPU_NONCACHE_SIZE0, AMDGPU_GPU_PAGE_ALIGN(sizeof(struct amdgpu_fw_shared))); @@ -513,10 +528,10 @@ static void vcn_v2_5_mc_resume_dpg_mode(struct amdgpu_device *adev, int inst_idx /* non-cache window */ WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET( VCN, 0, mmUVD_LMI_VCPU_NC0_64BIT_BAR_LOW), - lower_32_bits(adev->vcn.inst[inst_idx].fw_shared_gpu_addr), 0, indirect); + lower_32_bits(adev->vcn.inst[inst_idx].fw_shared.gpu_addr), 0, indirect); WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET( VCN, 0, mmUVD_LMI_VCPU_NC0_64BIT_BAR_HIGH), - upper_32_bits(adev->vcn.inst[inst_idx].fw_shared_gpu_addr), 0, indirect); + upper_32_bits(adev->vcn.inst[inst_idx].fw_shared.gpu_addr), 0, indirect); WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET( VCN, 0, mmUVD_VCPU_NONCACHE_OFFSET0), 0, 0, indirect); WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET( @@ -757,7 +772,7 @@ static void vcn_v2_5_enable_clock_gating(struct amdgpu_device *adev) static int vcn_v2_5_start_dpg_mode(struct amdgpu_device *adev, int inst_idx, bool indirect) { - volatile struct amdgpu_fw_shared *fw_shared = adev->vcn.inst[inst_idx].fw_shared_cpu_addr; + volatile struct amdgpu_fw_shared *fw_shared = adev->vcn.inst[inst_idx].fw_shared.cpu_addr; struct amdgpu_ring *ring; uint32_t rb_bufsz, tmp; @@ -981,7 +996,7 @@ static int vcn_v2_5_start(struct amdgpu_device *adev) vcn_v2_5_mc_resume(adev); for (i = 0; i < adev->vcn.num_vcn_inst; ++i) { - volatile struct amdgpu_fw_shared *fw_shared = adev->vcn.inst[i].fw_shared_cpu_addr; + volatile struct amdgpu_fw_shared *fw_shared = adev->vcn.inst[i].fw_shared.cpu_addr; if (adev->vcn.harvest_config & (1 << i)) continue; /* VCN global tiling registers */ @@ -1403,7 +1418,7 @@ static int vcn_v2_5_pause_dpg_mode(struct amdgpu_device *adev, UVD_POWER_STATUS__UVD_POWER_STATUS_MASK); if (!ret_code) { - volatile struct amdgpu_fw_shared *fw_shared = adev->vcn.inst[inst_idx].fw_shared_cpu_addr; + volatile struct amdgpu_fw_shared *fw_shared = adev->vcn.inst[inst_idx].fw_shared.cpu_addr; /* pause DPG */ reg_data |= UVD_DPG_PAUSE__NJ_PAUSE_DPG_REQ_MASK; @@ -1485,7 +1500,7 @@ static uint64_t vcn_v2_5_dec_ring_get_wptr(struct amdgpu_ring *ring) struct amdgpu_device *adev = ring->adev; if (ring->use_doorbell) - return adev->wb.wb[ring->wptr_offs]; + return *ring->wptr_cpu_addr; else return RREG32_SOC15(VCN, ring->me, mmUVD_RBC_RB_WPTR); } @@ -1502,7 +1517,7 @@ static void vcn_v2_5_dec_ring_set_wptr(struct amdgpu_ring *ring) struct amdgpu_device *adev = ring->adev; if (ring->use_doorbell) { - adev->wb.wb[ring->wptr_offs] = lower_32_bits(ring->wptr); + *ring->wptr_cpu_addr = lower_32_bits(ring->wptr); WDOORBELL32(ring->doorbell_index, lower_32_bits(ring->wptr)); } else { WREG32_SOC15(VCN, ring->me, mmUVD_RBC_RB_WPTR, lower_32_bits(ring->wptr)); @@ -1512,6 +1527,7 @@ static void vcn_v2_5_dec_ring_set_wptr(struct amdgpu_ring *ring) static const struct amdgpu_ring_funcs vcn_v2_5_dec_ring_vm_funcs = { .type = AMDGPU_RING_TYPE_VCN_DEC, .align_mask = 0xf, + .secure_submission_supported = true, .vmhub = AMDGPU_MMHUB_1, .get_rptr = vcn_v2_5_dec_ring_get_rptr, .get_wptr = vcn_v2_5_dec_ring_get_wptr, @@ -1542,6 +1558,7 @@ static const struct amdgpu_ring_funcs vcn_v2_5_dec_ring_vm_funcs = { static const struct amdgpu_ring_funcs vcn_v2_6_dec_ring_vm_funcs = { .type = AMDGPU_RING_TYPE_VCN_DEC, .align_mask = 0xf, + .secure_submission_supported = true, .vmhub = AMDGPU_MMHUB_0, .get_rptr = vcn_v2_5_dec_ring_get_rptr, .get_wptr = vcn_v2_5_dec_ring_get_wptr, @@ -1599,12 +1616,12 @@ static uint64_t vcn_v2_5_enc_ring_get_wptr(struct amdgpu_ring *ring) if (ring == &adev->vcn.inst[ring->me].ring_enc[0]) { if (ring->use_doorbell) - return adev->wb.wb[ring->wptr_offs]; + return *ring->wptr_cpu_addr; else return RREG32_SOC15(VCN, ring->me, mmUVD_RB_WPTR); } else { if (ring->use_doorbell) - return adev->wb.wb[ring->wptr_offs]; + return *ring->wptr_cpu_addr; else return RREG32_SOC15(VCN, ring->me, mmUVD_RB_WPTR2); } @@ -1623,14 +1640,14 @@ static void vcn_v2_5_enc_ring_set_wptr(struct amdgpu_ring *ring) if (ring == &adev->vcn.inst[ring->me].ring_enc[0]) { if (ring->use_doorbell) { - adev->wb.wb[ring->wptr_offs] = lower_32_bits(ring->wptr); + *ring->wptr_cpu_addr = lower_32_bits(ring->wptr); WDOORBELL32(ring->doorbell_index, lower_32_bits(ring->wptr)); } else { WREG32_SOC15(VCN, ring->me, mmUVD_RB_WPTR, lower_32_bits(ring->wptr)); } } else { if (ring->use_doorbell) { - adev->wb.wb[ring->wptr_offs] = lower_32_bits(ring->wptr); + *ring->wptr_cpu_addr = lower_32_bits(ring->wptr); WDOORBELL32(ring->doorbell_index, lower_32_bits(ring->wptr)); } else { WREG32_SOC15(VCN, ring->me, mmUVD_RB_WPTR2, lower_32_bits(ring->wptr)); @@ -1844,6 +1861,9 @@ static int vcn_v2_5_process_interrupt(struct amdgpu_device *adev, case VCN_2_0__SRCID__UVD_ENC_LOW_LATENCY: amdgpu_fence_process(&adev->vcn.inst[ip_instance].ring_enc[1]); break; + case VCN_2_6__SRCID_UVD_POISON: + amdgpu_vcn_process_poison_irq(adev, source, entry); + break; default: DRM_ERROR("Unhandled interrupt: %d %d\n", entry->src_id, entry->src_data[0]); @@ -1927,3 +1947,71 @@ const struct amdgpu_ip_block_version vcn_v2_6_ip_block = .rev = 0, .funcs = &vcn_v2_6_ip_funcs, }; + +static uint32_t vcn_v2_6_query_poison_by_instance(struct amdgpu_device *adev, + uint32_t instance, uint32_t sub_block) +{ + uint32_t poison_stat = 0, reg_value = 0; + + switch (sub_block) { + case AMDGPU_VCN_V2_6_VCPU_VCODEC: + reg_value = RREG32_SOC15(VCN, instance, mmUVD_RAS_VCPU_VCODEC_STATUS); + poison_stat = REG_GET_FIELD(reg_value, UVD_RAS_VCPU_VCODEC_STATUS, POISONED_PF); + break; + default: + break; + } + + if (poison_stat) + dev_info(adev->dev, "Poison detected in VCN%d, sub_block%d\n", + instance, sub_block); + + return poison_stat; +} + +static bool vcn_v2_6_query_poison_status(struct amdgpu_device *adev) +{ + uint32_t inst, sub; + uint32_t poison_stat = 0; + + for (inst = 0; inst < adev->vcn.num_vcn_inst; inst++) + for (sub = 0; sub < AMDGPU_VCN_V2_6_MAX_SUB_BLOCK; sub++) + poison_stat += + vcn_v2_6_query_poison_by_instance(adev, inst, sub); + + return !!poison_stat; +} + +const struct amdgpu_ras_block_hw_ops vcn_v2_6_ras_hw_ops = { + .query_poison_status = vcn_v2_6_query_poison_status, +}; + +static struct amdgpu_vcn_ras vcn_v2_6_ras = { + .ras_block = { + .hw_ops = &vcn_v2_6_ras_hw_ops, + }, +}; + +static void vcn_v2_5_set_ras_funcs(struct amdgpu_device *adev) +{ + switch (adev->ip_versions[VCN_HWIP][0]) { + case IP_VERSION(2, 6, 0): + adev->vcn.ras = &vcn_v2_6_ras; + break; + default: + break; + } + + if (adev->vcn.ras) { + amdgpu_ras_register_ras_block(adev, &adev->vcn.ras->ras_block); + + strcpy(adev->vcn.ras->ras_block.ras_comm.name, "vcn"); + adev->vcn.ras->ras_block.ras_comm.block = AMDGPU_RAS_BLOCK__VCN; + adev->vcn.ras->ras_block.ras_comm.type = AMDGPU_RAS_ERROR__POISON; + adev->vcn.ras_if = &adev->vcn.ras->ras_block.ras_comm; + + /* If don't define special ras_late_init function, use default ras_late_init */ + if (!adev->vcn.ras->ras_block.ras_late_init) + adev->vcn.ras->ras_block.ras_late_init = amdgpu_ras_block_late_init; + } +} diff --git a/drivers/gpu/drm/amd/amdgpu/vcn_v2_5.h b/drivers/gpu/drm/amd/amdgpu/vcn_v2_5.h index e72f799ed0fd..1c19af74e4fd 100644 --- a/drivers/gpu/drm/amd/amdgpu/vcn_v2_5.h +++ b/drivers/gpu/drm/amd/amdgpu/vcn_v2_5.h @@ -24,6 +24,12 @@ #ifndef __VCN_V2_5_H__ #define __VCN_V2_5_H__ +enum amdgpu_vcn_v2_6_sub_block { + AMDGPU_VCN_V2_6_VCPU_VCODEC = 0, + + AMDGPU_VCN_V2_6_MAX_SUB_BLOCK, +}; + extern const struct amdgpu_ip_block_version vcn_v2_5_ip_block; extern const struct amdgpu_ip_block_version vcn_v2_6_ip_block; diff --git a/drivers/gpu/drm/amd/amdgpu/vcn_v3_0.c b/drivers/gpu/drm/amd/amdgpu/vcn_v3_0.c index da11ceba0698..9c8b5fd99037 100644 --- a/drivers/gpu/drm/amd/amdgpu/vcn_v3_0.c +++ b/drivers/gpu/drm/amd/amdgpu/vcn_v3_0.c @@ -25,10 +25,12 @@ #include "amdgpu.h" #include "amdgpu_vcn.h" #include "amdgpu_pm.h" +#include "amdgpu_cs.h" #include "soc15.h" #include "soc15d.h" #include "vcn_v2_0.h" #include "mmsch_v3_0.h" +#include "vcn_sw_ring.h" #include "vcn/vcn_3_0_0_offset.h" #include "vcn/vcn_3_0_0_sh_mask.h" @@ -36,6 +38,9 @@ #include <drm/drm_drv.h> +#define VCN_VID_SOC_ADDRESS_2_0 0x1fa00 +#define VCN1_VID_SOC_ADDRESS_3_0 0x48200 + #define mmUVD_CONTEXT_ID_INTERNAL_OFFSET 0x27 #define mmUVD_GPCOM_VCPU_CMD_INTERNAL_OFFSET 0x0f #define mmUVD_GPCOM_VCPU_DATA0_INTERNAL_OFFSET 0x10 @@ -213,11 +218,19 @@ static int vcn_v3_0_sw_init(void *handle) return r; } - fw_shared = adev->vcn.inst[i].fw_shared_cpu_addr; + fw_shared = adev->vcn.inst[i].fw_shared.cpu_addr; fw_shared->present_flag_0 |= cpu_to_le32(AMDGPU_VCN_SW_RING_FLAG) | cpu_to_le32(AMDGPU_VCN_MULTI_QUEUE_FLAG) | cpu_to_le32(AMDGPU_VCN_FW_SHARED_FLAG_0_RB); fw_shared->sw_ring.is_enabled = cpu_to_le32(DEC_SW_RING_ENABLED); + fw_shared->present_flag_0 |= AMDGPU_VCN_SMU_VERSION_INFO_FLAG; + if (adev->ip_versions[UVD_HWIP][0] == IP_VERSION(3, 1, 2)) + fw_shared->smu_interface_info.smu_interface_type = 2; + else if (adev->ip_versions[UVD_HWIP][0] == IP_VERSION(3, 1, 1)) + fw_shared->smu_interface_info.smu_interface_type = 1; + + if (amdgpu_vcnfw_log) + amdgpu_vcn_fwlog_init(&adev->vcn.inst[i]); } if (amdgpu_sriov_vf(adev)) { @@ -249,7 +262,7 @@ static int vcn_v3_0_sw_fini(void *handle) if (adev->vcn.harvest_config & (1 << i)) continue; - fw_shared = adev->vcn.inst[i].fw_shared_cpu_addr; + fw_shared = adev->vcn.inst[i].fw_shared.cpu_addr; fw_shared->present_flag_0 = 0; fw_shared->sw_ring.is_enabled = false; } @@ -295,6 +308,7 @@ static int vcn_v3_0_hw_init(void *handle) ring = &adev->vcn.inst[i].ring_dec; if (amdgpu_vcn_is_disabled_vcn(adev, VCN_DECODE_RING, i)) { ring->sched.ready = false; + ring->no_scheduler = true; dev_info(adev->dev, "ring %s is disabled by hypervisor\n", ring->name); } else { ring->wptr = 0; @@ -307,6 +321,7 @@ static int vcn_v3_0_hw_init(void *handle) ring = &adev->vcn.inst[i].ring_enc[j]; if (amdgpu_vcn_is_disabled_vcn(adev, VCN_ENCODE_RING, i)) { ring->sched.ready = false; + ring->no_scheduler = true; dev_info(adev->dev, "ring %s is disabled by hypervisor\n", ring->name); } else { ring->wptr = 0; @@ -469,9 +484,9 @@ static void vcn_v3_0_mc_resume(struct amdgpu_device *adev, int inst) /* non-cache window */ WREG32_SOC15(VCN, inst, mmUVD_LMI_VCPU_NC0_64BIT_BAR_LOW, - lower_32_bits(adev->vcn.inst[inst].fw_shared_gpu_addr)); + lower_32_bits(adev->vcn.inst[inst].fw_shared.gpu_addr)); WREG32_SOC15(VCN, inst, mmUVD_LMI_VCPU_NC0_64BIT_BAR_HIGH, - upper_32_bits(adev->vcn.inst[inst].fw_shared_gpu_addr)); + upper_32_bits(adev->vcn.inst[inst].fw_shared.gpu_addr)); WREG32_SOC15(VCN, inst, mmUVD_VCPU_NONCACHE_OFFSET0, 0); WREG32_SOC15(VCN, inst, mmUVD_VCPU_NONCACHE_SIZE0, AMDGPU_GPU_PAGE_ALIGN(sizeof(struct amdgpu_fw_shared))); @@ -558,10 +573,10 @@ static void vcn_v3_0_mc_resume_dpg_mode(struct amdgpu_device *adev, int inst_idx /* non-cache window */ WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET( VCN, inst_idx, mmUVD_LMI_VCPU_NC0_64BIT_BAR_LOW), - lower_32_bits(adev->vcn.inst[inst_idx].fw_shared_gpu_addr), 0, indirect); + lower_32_bits(adev->vcn.inst[inst_idx].fw_shared.gpu_addr), 0, indirect); WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET( VCN, inst_idx, mmUVD_LMI_VCPU_NC0_64BIT_BAR_HIGH), - upper_32_bits(adev->vcn.inst[inst_idx].fw_shared_gpu_addr), 0, indirect); + upper_32_bits(adev->vcn.inst[inst_idx].fw_shared.gpu_addr), 0, indirect); WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET( VCN, inst_idx, mmUVD_VCPU_NONCACHE_OFFSET0), 0, 0, indirect); WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET( @@ -569,8 +584,8 @@ static void vcn_v3_0_mc_resume_dpg_mode(struct amdgpu_device *adev, int inst_idx AMDGPU_GPU_PAGE_ALIGN(sizeof(struct amdgpu_fw_shared)), 0, indirect); /* VCN global tiling registers */ - WREG32_SOC15_DPG_MODE(0, SOC15_DPG_MODE_OFFSET( - UVD, 0, mmUVD_GFX10_ADDR_CONFIG), adev->gfx.config.gb_addr_config, 0, indirect); + WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET( + UVD, inst_idx, mmUVD_GFX10_ADDR_CONFIG), adev->gfx.config.gb_addr_config, 0, indirect); } static void vcn_v3_0_disable_static_power_gating(struct amdgpu_device *adev, int inst) @@ -923,7 +938,7 @@ static void vcn_v3_0_enable_clock_gating(struct amdgpu_device *adev, int inst) static int vcn_v3_0_start_dpg_mode(struct amdgpu_device *adev, int inst_idx, bool indirect) { - volatile struct amdgpu_fw_shared *fw_shared = adev->vcn.inst[inst_idx].fw_shared_cpu_addr; + volatile struct amdgpu_fw_shared *fw_shared = adev->vcn.inst[inst_idx].fw_shared.cpu_addr; struct amdgpu_ring *ring; uint32_t rb_bufsz, tmp; @@ -1220,7 +1235,7 @@ static int vcn_v3_0_start(struct amdgpu_device *adev) tmp = REG_SET_FIELD(tmp, UVD_RBC_RB_CNTL, RB_RPTR_WR_EN, 1); WREG32_SOC15(VCN, i, mmUVD_RBC_RB_CNTL, tmp); - fw_shared = adev->vcn.inst[i].fw_shared_cpu_addr; + fw_shared = adev->vcn.inst[i].fw_shared.cpu_addr; fw_shared->multi_queue.decode_queue_mode |= cpu_to_le32(FW_QUEUE_RING_RESET); /* programm the RB_BASE for ring buffer */ @@ -1474,8 +1489,11 @@ static int vcn_v3_0_start_sriov(struct amdgpu_device *adev) static int vcn_v3_0_stop_dpg_mode(struct amdgpu_device *adev, int inst_idx) { + struct dpg_pause_state state = {.fw_based = VCN_DPG_STATE__UNPAUSE}; uint32_t tmp; + vcn_v3_0_pause_dpg_mode(adev, inst_idx, &state); + /* Wait for power status to be 1 */ SOC15_WAIT_ON_RREG(VCN, inst_idx, mmUVD_POWER_STATUS, 1, UVD_POWER_STATUS__UVD_POWER_STATUS_MASK); @@ -1611,7 +1629,7 @@ static int vcn_v3_0_pause_dpg_mode(struct amdgpu_device *adev, if (adev->ip_versions[UVD_HWIP][0] != IP_VERSION(3, 0, 33)) { /* Restore */ - fw_shared = adev->vcn.inst[inst_idx].fw_shared_cpu_addr; + fw_shared = adev->vcn.inst[inst_idx].fw_shared.cpu_addr; fw_shared->multi_queue.encode_generalpurpose_queue_mode |= cpu_to_le32(FW_QUEUE_RING_RESET); ring = &adev->vcn.inst[inst_idx].ring_enc[0]; ring->wptr = 0; @@ -1681,7 +1699,7 @@ static uint64_t vcn_v3_0_dec_ring_get_wptr(struct amdgpu_ring *ring) struct amdgpu_device *adev = ring->adev; if (ring->use_doorbell) - return adev->wb.wb[ring->wptr_offs]; + return *ring->wptr_cpu_addr; else return RREG32_SOC15(VCN, ring->me, mmUVD_RBC_RB_WPTR); } @@ -1700,86 +1718,25 @@ static void vcn_v3_0_dec_ring_set_wptr(struct amdgpu_ring *ring) if (adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG) { /*whenever update RBC_RB_WPTR, we save the wptr in shared rb.wptr and scratch2 */ - fw_shared = adev->vcn.inst[ring->me].fw_shared_cpu_addr; + fw_shared = adev->vcn.inst[ring->me].fw_shared.cpu_addr; fw_shared->rb.wptr = lower_32_bits(ring->wptr); WREG32_SOC15(VCN, ring->me, mmUVD_SCRATCH2, lower_32_bits(ring->wptr)); } if (ring->use_doorbell) { - adev->wb.wb[ring->wptr_offs] = lower_32_bits(ring->wptr); + *ring->wptr_cpu_addr = lower_32_bits(ring->wptr); WDOORBELL32(ring->doorbell_index, lower_32_bits(ring->wptr)); } else { WREG32_SOC15(VCN, ring->me, mmUVD_RBC_RB_WPTR, lower_32_bits(ring->wptr)); } } -static void vcn_v3_0_dec_sw_ring_emit_fence(struct amdgpu_ring *ring, u64 addr, - u64 seq, uint32_t flags) -{ - WARN_ON(flags & AMDGPU_FENCE_FLAG_64BIT); - - amdgpu_ring_write(ring, VCN_DEC_SW_CMD_FENCE); - amdgpu_ring_write(ring, addr); - amdgpu_ring_write(ring, upper_32_bits(addr)); - amdgpu_ring_write(ring, seq); - amdgpu_ring_write(ring, VCN_DEC_SW_CMD_TRAP); -} - -static void vcn_v3_0_dec_sw_ring_insert_end(struct amdgpu_ring *ring) -{ - amdgpu_ring_write(ring, VCN_DEC_SW_CMD_END); -} - -static void vcn_v3_0_dec_sw_ring_emit_ib(struct amdgpu_ring *ring, - struct amdgpu_job *job, - struct amdgpu_ib *ib, - uint32_t flags) -{ - uint32_t vmid = AMDGPU_JOB_GET_VMID(job); - - amdgpu_ring_write(ring, VCN_DEC_SW_CMD_IB); - amdgpu_ring_write(ring, vmid); - amdgpu_ring_write(ring, lower_32_bits(ib->gpu_addr)); - amdgpu_ring_write(ring, upper_32_bits(ib->gpu_addr)); - amdgpu_ring_write(ring, ib->length_dw); -} - -static void vcn_v3_0_dec_sw_ring_emit_reg_wait(struct amdgpu_ring *ring, uint32_t reg, - uint32_t val, uint32_t mask) -{ - amdgpu_ring_write(ring, VCN_DEC_SW_CMD_REG_WAIT); - amdgpu_ring_write(ring, reg << 2); - amdgpu_ring_write(ring, mask); - amdgpu_ring_write(ring, val); -} - -static void vcn_v3_0_dec_sw_ring_emit_vm_flush(struct amdgpu_ring *ring, - uint32_t vmid, uint64_t pd_addr) -{ - struct amdgpu_vmhub *hub = &ring->adev->vmhub[ring->funcs->vmhub]; - uint32_t data0, data1, mask; - - pd_addr = amdgpu_gmc_emit_flush_gpu_tlb(ring, vmid, pd_addr); - - /* wait for register write */ - data0 = hub->ctx0_ptb_addr_lo32 + vmid * hub->ctx_addr_distance; - data1 = lower_32_bits(pd_addr); - mask = 0xffffffff; - vcn_v3_0_dec_sw_ring_emit_reg_wait(ring, data0, data1, mask); -} - -static void vcn_v3_0_dec_sw_ring_emit_wreg(struct amdgpu_ring *ring, uint32_t reg, uint32_t val) -{ - amdgpu_ring_write(ring, VCN_DEC_SW_CMD_REG_WRITE); - amdgpu_ring_write(ring, reg << 2); - amdgpu_ring_write(ring, val); -} - static const struct amdgpu_ring_funcs vcn_v3_0_dec_sw_ring_vm_funcs = { .type = AMDGPU_RING_TYPE_VCN_DEC, .align_mask = 0x3f, .nop = VCN_DEC_SW_CMD_NO_OP, + .secure_submission_supported = true, .vmhub = AMDGPU_MMHUB_0, .get_rptr = vcn_v3_0_dec_ring_get_rptr, .get_wptr = vcn_v3_0_dec_ring_get_wptr, @@ -1787,40 +1744,40 @@ static const struct amdgpu_ring_funcs vcn_v3_0_dec_sw_ring_vm_funcs = { .emit_frame_size = SOC15_FLUSH_GPU_TLB_NUM_WREG * 3 + SOC15_FLUSH_GPU_TLB_NUM_REG_WAIT * 4 + - 4 + /* vcn_v3_0_dec_sw_ring_emit_vm_flush */ - 5 + 5 + /* vcn_v3_0_dec_sw_ring_emit_fdec_swe x2 vm fdec_swe */ - 1, /* vcn_v3_0_dec_sw_ring_insert_end */ - .emit_ib_size = 5, /* vcn_v3_0_dec_sw_ring_emit_ib */ - .emit_ib = vcn_v3_0_dec_sw_ring_emit_ib, - .emit_fence = vcn_v3_0_dec_sw_ring_emit_fence, - .emit_vm_flush = vcn_v3_0_dec_sw_ring_emit_vm_flush, + VCN_SW_RING_EMIT_FRAME_SIZE, + .emit_ib_size = 5, /* vcn_dec_sw_ring_emit_ib */ + .emit_ib = vcn_dec_sw_ring_emit_ib, + .emit_fence = vcn_dec_sw_ring_emit_fence, + .emit_vm_flush = vcn_dec_sw_ring_emit_vm_flush, .test_ring = amdgpu_vcn_dec_sw_ring_test_ring, .test_ib = NULL,//amdgpu_vcn_dec_sw_ring_test_ib, .insert_nop = amdgpu_ring_insert_nop, - .insert_end = vcn_v3_0_dec_sw_ring_insert_end, + .insert_end = vcn_dec_sw_ring_insert_end, .pad_ib = amdgpu_ring_generic_pad_ib, .begin_use = amdgpu_vcn_ring_begin_use, .end_use = amdgpu_vcn_ring_end_use, - .emit_wreg = vcn_v3_0_dec_sw_ring_emit_wreg, - .emit_reg_wait = vcn_v3_0_dec_sw_ring_emit_reg_wait, + .emit_wreg = vcn_dec_sw_ring_emit_wreg, + .emit_reg_wait = vcn_dec_sw_ring_emit_reg_wait, .emit_reg_write_reg_wait = amdgpu_ring_emit_reg_write_reg_wait_helper, }; -static int vcn_v3_0_limit_sched(struct amdgpu_cs_parser *p) +static int vcn_v3_0_limit_sched(struct amdgpu_cs_parser *p, + struct amdgpu_job *job) { struct drm_gpu_scheduler **scheds; /* The create msg must be in the first IB submitted */ - if (atomic_read(&p->entity->fence_seq)) + if (atomic_read(&job->base.entity->fence_seq)) return -EINVAL; scheds = p->adev->gpu_sched[AMDGPU_HW_IP_VCN_DEC] [AMDGPU_RING_PRIO_DEFAULT].sched; - drm_sched_entity_modify_sched(p->entity, scheds, 1); + drm_sched_entity_modify_sched(job->base.entity, scheds, 1); return 0; } -static int vcn_v3_0_dec_msg(struct amdgpu_cs_parser *p, uint64_t addr) +static int vcn_v3_0_dec_msg(struct amdgpu_cs_parser *p, struct amdgpu_job *job, + uint64_t addr) { struct ttm_operation_ctx ctx = { false, false }; struct amdgpu_bo_va_mapping *map; @@ -1891,7 +1848,7 @@ static int vcn_v3_0_dec_msg(struct amdgpu_cs_parser *p, uint64_t addr) if (create[0] == 0x7 || create[0] == 0x10 || create[0] == 0x11) continue; - r = vcn_v3_0_limit_sched(p); + r = vcn_v3_0_limit_sched(p, job); if (r) goto out; } @@ -1902,10 +1859,10 @@ out: } static int vcn_v3_0_ring_patch_cs_in_place(struct amdgpu_cs_parser *p, - uint32_t ib_idx) + struct amdgpu_job *job, + struct amdgpu_ib *ib) { - struct amdgpu_ring *ring = to_amdgpu_ring(p->entity->rq->sched); - struct amdgpu_ib *ib = &p->job->ibs[ib_idx]; + struct amdgpu_ring *ring = amdgpu_job_ring(job); uint32_t msg_lo = 0, msg_hi = 0; unsigned i; int r; @@ -1915,8 +1872,8 @@ static int vcn_v3_0_ring_patch_cs_in_place(struct amdgpu_cs_parser *p, return 0; for (i = 0; i < ib->length_dw; i += 2) { - uint32_t reg = amdgpu_get_ib_value(p, ib_idx, i); - uint32_t val = amdgpu_get_ib_value(p, ib_idx, i + 1); + uint32_t reg = amdgpu_ib_get_value(ib, i); + uint32_t val = amdgpu_ib_get_value(ib, i + 1); if (reg == PACKET0(p->adev->vcn.internal.data0, 0)) { msg_lo = val; @@ -1924,7 +1881,8 @@ static int vcn_v3_0_ring_patch_cs_in_place(struct amdgpu_cs_parser *p, msg_hi = val; } else if (reg == PACKET0(p->adev->vcn.internal.cmd, 0) && val == 0) { - r = vcn_v3_0_dec_msg(p, ((u64)msg_hi) << 32 | msg_lo); + r = vcn_v3_0_dec_msg(p, job, + ((u64)msg_hi) << 32 | msg_lo); if (r) return r; } @@ -1935,6 +1893,7 @@ static int vcn_v3_0_ring_patch_cs_in_place(struct amdgpu_cs_parser *p, static const struct amdgpu_ring_funcs vcn_v3_0_dec_ring_vm_funcs = { .type = AMDGPU_RING_TYPE_VCN_DEC, .align_mask = 0xf, + .secure_submission_supported = true, .vmhub = AMDGPU_MMHUB_0, .get_rptr = vcn_v3_0_dec_ring_get_rptr, .get_wptr = vcn_v3_0_dec_ring_get_wptr, @@ -1993,12 +1952,12 @@ static uint64_t vcn_v3_0_enc_ring_get_wptr(struct amdgpu_ring *ring) if (ring == &adev->vcn.inst[ring->me].ring_enc[0]) { if (ring->use_doorbell) - return adev->wb.wb[ring->wptr_offs]; + return *ring->wptr_cpu_addr; else return RREG32_SOC15(VCN, ring->me, mmUVD_RB_WPTR); } else { if (ring->use_doorbell) - return adev->wb.wb[ring->wptr_offs]; + return *ring->wptr_cpu_addr; else return RREG32_SOC15(VCN, ring->me, mmUVD_RB_WPTR2); } @@ -2017,14 +1976,14 @@ static void vcn_v3_0_enc_ring_set_wptr(struct amdgpu_ring *ring) if (ring == &adev->vcn.inst[ring->me].ring_enc[0]) { if (ring->use_doorbell) { - adev->wb.wb[ring->wptr_offs] = lower_32_bits(ring->wptr); + *ring->wptr_cpu_addr = lower_32_bits(ring->wptr); WDOORBELL32(ring->doorbell_index, lower_32_bits(ring->wptr)); } else { WREG32_SOC15(VCN, ring->me, mmUVD_RB_WPTR, lower_32_bits(ring->wptr)); } } else { if (ring->use_doorbell) { - adev->wb.wb[ring->wptr_offs] = lower_32_bits(ring->wptr); + *ring->wptr_cpu_addr = lower_32_bits(ring->wptr); WDOORBELL32(ring->doorbell_index, lower_32_bits(ring->wptr)); } else { WREG32_SOC15(VCN, ring->me, mmUVD_RB_WPTR2, lower_32_bits(ring->wptr)); diff --git a/drivers/gpu/drm/amd/amdgpu/vcn_v4_0.c b/drivers/gpu/drm/amd/amdgpu/vcn_v4_0.c new file mode 100644 index 000000000000..897a5ce9c9da --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/vcn_v4_0.c @@ -0,0 +1,2003 @@ +/* + * Copyright 2021 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#include <linux/firmware.h> +#include "amdgpu.h" +#include "amdgpu_vcn.h" +#include "amdgpu_pm.h" +#include "amdgpu_cs.h" +#include "soc15.h" +#include "soc15d.h" +#include "soc15_hw_ip.h" +#include "vcn_v2_0.h" +#include "mmsch_v4_0.h" + +#include "vcn/vcn_4_0_0_offset.h" +#include "vcn/vcn_4_0_0_sh_mask.h" +#include "ivsrcid/vcn/irqsrcs_vcn_4_0.h" + +#include <drm/drm_drv.h> + +#define mmUVD_DPG_LMA_CTL regUVD_DPG_LMA_CTL +#define mmUVD_DPG_LMA_CTL_BASE_IDX regUVD_DPG_LMA_CTL_BASE_IDX +#define mmUVD_DPG_LMA_DATA regUVD_DPG_LMA_DATA +#define mmUVD_DPG_LMA_DATA_BASE_IDX regUVD_DPG_LMA_DATA_BASE_IDX + +#define VCN_VID_SOC_ADDRESS_2_0 0x1fb00 +#define VCN1_VID_SOC_ADDRESS_3_0 0x48300 + +#define VCN_HARVEST_MMSCH 0 + +#define RDECODE_MSG_CREATE 0x00000000 +#define RDECODE_MESSAGE_CREATE 0x00000001 + +static int amdgpu_ih_clientid_vcns[] = { + SOC15_IH_CLIENTID_VCN, + SOC15_IH_CLIENTID_VCN1 +}; + +static int vcn_v4_0_start_sriov(struct amdgpu_device *adev); +static void vcn_v4_0_set_unified_ring_funcs(struct amdgpu_device *adev); +static void vcn_v4_0_set_irq_funcs(struct amdgpu_device *adev); +static int vcn_v4_0_set_powergating_state(void *handle, + enum amd_powergating_state state); +static int vcn_v4_0_pause_dpg_mode(struct amdgpu_device *adev, + int inst_idx, struct dpg_pause_state *new_state); +static void vcn_v4_0_unified_ring_set_wptr(struct amdgpu_ring *ring); + +/** + * vcn_v4_0_early_init - set function pointers + * + * @handle: amdgpu_device pointer + * + * Set ring and irq function pointers + */ +static int vcn_v4_0_early_init(void *handle) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + + if (amdgpu_sriov_vf(adev)) + adev->vcn.harvest_config = VCN_HARVEST_MMSCH; + + /* re-use enc ring as unified ring */ + adev->vcn.num_enc_rings = 1; + + vcn_v4_0_set_unified_ring_funcs(adev); + vcn_v4_0_set_irq_funcs(adev); + + return 0; +} + +/** + * vcn_v4_0_sw_init - sw init for VCN block + * + * @handle: amdgpu_device pointer + * + * Load firmware and sw initialization + */ +static int vcn_v4_0_sw_init(void *handle) +{ + struct amdgpu_ring *ring; + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + int i, r; + int vcn_doorbell_index = 0; + + r = amdgpu_vcn_sw_init(adev); + if (r) + return r; + + amdgpu_vcn_setup_ucode(adev); + + r = amdgpu_vcn_resume(adev); + if (r) + return r; + + if (amdgpu_sriov_vf(adev)) { + vcn_doorbell_index = adev->doorbell_index.vcn.vcn_ring0_1 - MMSCH_DOORBELL_OFFSET; + /* get DWORD offset */ + vcn_doorbell_index = vcn_doorbell_index << 1; + } + + for (i = 0; i < adev->vcn.num_vcn_inst; i++) { + volatile struct amdgpu_vcn4_fw_shared *fw_shared; + + if (adev->vcn.harvest_config & (1 << i)) + continue; + + atomic_set(&adev->vcn.inst[i].sched_score, 0); + + /* VCN UNIFIED TRAP */ + r = amdgpu_irq_add_id(adev, amdgpu_ih_clientid_vcns[i], + VCN_4_0__SRCID__UVD_ENC_GENERAL_PURPOSE, &adev->vcn.inst[i].irq); + if (r) + return r; + + ring = &adev->vcn.inst[i].ring_enc[0]; + ring->use_doorbell = true; + if (amdgpu_sriov_vf(adev)) + ring->doorbell_index = vcn_doorbell_index + i * (adev->vcn.num_enc_rings + 1) + 1; + else + ring->doorbell_index = (adev->doorbell_index.vcn.vcn_ring0_1 << 1) + 2 + 8 * i; + + sprintf(ring->name, "vcn_unified_%d", i); + + r = amdgpu_ring_init(adev, ring, 512, &adev->vcn.inst[i].irq, 0, + AMDGPU_RING_PRIO_0, &adev->vcn.inst[i].sched_score); + if (r) + return r; + + fw_shared = adev->vcn.inst[i].fw_shared.cpu_addr; + fw_shared->present_flag_0 = cpu_to_le32(AMDGPU_FW_SHARED_FLAG_0_UNIFIED_QUEUE); + fw_shared->sq.is_enabled = 1; + + fw_shared->present_flag_0 |= cpu_to_le32(AMDGPU_VCN_SMU_DPM_INTERFACE_FLAG); + fw_shared->smu_dpm_interface.smu_interface_type = (adev->flags & AMD_IS_APU) ? + AMDGPU_VCN_SMU_DPM_INTERFACE_APU : AMDGPU_VCN_SMU_DPM_INTERFACE_DGPU; + + if (amdgpu_sriov_vf(adev)) + fw_shared->present_flag_0 |= cpu_to_le32(AMDGPU_VCN_VF_RB_SETUP_FLAG); + + if (amdgpu_vcnfw_log) + amdgpu_vcn_fwlog_init(&adev->vcn.inst[i]); + } + + if (amdgpu_sriov_vf(adev)) { + r = amdgpu_virt_alloc_mm_table(adev); + if (r) + return r; + } + + if (adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG) + adev->vcn.pause_dpg_mode = vcn_v4_0_pause_dpg_mode; + + return 0; +} + +/** + * vcn_v4_0_sw_fini - sw fini for VCN block + * + * @handle: amdgpu_device pointer + * + * VCN suspend and free up sw allocation + */ +static int vcn_v4_0_sw_fini(void *handle) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + int i, r, idx; + + if (drm_dev_enter(adev_to_drm(adev), &idx)) { + for (i = 0; i < adev->vcn.num_vcn_inst; i++) { + volatile struct amdgpu_vcn4_fw_shared *fw_shared; + + if (adev->vcn.harvest_config & (1 << i)) + continue; + + fw_shared = adev->vcn.inst[i].fw_shared.cpu_addr; + fw_shared->present_flag_0 = 0; + fw_shared->sq.is_enabled = 0; + } + + drm_dev_exit(idx); + } + + if (amdgpu_sriov_vf(adev)) + amdgpu_virt_free_mm_table(adev); + + r = amdgpu_vcn_suspend(adev); + if (r) + return r; + + r = amdgpu_vcn_sw_fini(adev); + + return r; +} + +/** + * vcn_v4_0_hw_init - start and test VCN block + * + * @handle: amdgpu_device pointer + * + * Initialize the hardware, boot up the VCPU and do some testing + */ +static int vcn_v4_0_hw_init(void *handle) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + struct amdgpu_ring *ring; + int i, r; + + if (amdgpu_sriov_vf(adev)) { + r = vcn_v4_0_start_sriov(adev); + if (r) + goto done; + + for (i = 0; i < adev->vcn.num_vcn_inst; ++i) { + if (adev->vcn.harvest_config & (1 << i)) + continue; + + ring = &adev->vcn.inst[i].ring_enc[0]; + if (amdgpu_vcn_is_disabled_vcn(adev, VCN_ENCODE_RING, i)) { + ring->sched.ready = false; + ring->no_scheduler = true; + dev_info(adev->dev, "ring %s is disabled by hypervisor\n", ring->name); + } else { + ring->wptr = 0; + ring->wptr_old = 0; + vcn_v4_0_unified_ring_set_wptr(ring); + ring->sched.ready = true; + } + } + } else { + for (i = 0; i < adev->vcn.num_vcn_inst; ++i) { + if (adev->vcn.harvest_config & (1 << i)) + continue; + + ring = &adev->vcn.inst[i].ring_enc[0]; + + adev->nbio.funcs->vcn_doorbell_range(adev, ring->use_doorbell, + ((adev->doorbell_index.vcn.vcn_ring0_1 << 1) + 8 * i), i); + + r = amdgpu_ring_test_helper(ring); + if (r) + goto done; + + } + } + +done: + if (!r) + DRM_INFO("VCN decode and encode initialized successfully(under %s).\n", + (adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG)?"DPG Mode":"SPG Mode"); + + return r; +} + +/** + * vcn_v4_0_hw_fini - stop the hardware block + * + * @handle: amdgpu_device pointer + * + * Stop the VCN block, mark ring as not ready any more + */ +static int vcn_v4_0_hw_fini(void *handle) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + int i; + + cancel_delayed_work_sync(&adev->vcn.idle_work); + + for (i = 0; i < adev->vcn.num_vcn_inst; ++i) { + if (adev->vcn.harvest_config & (1 << i)) + continue; + if (!amdgpu_sriov_vf(adev)) { + if ((adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG) || + (adev->vcn.cur_state != AMD_PG_STATE_GATE && + RREG32_SOC15(VCN, i, regUVD_STATUS))) { + vcn_v4_0_set_powergating_state(adev, AMD_PG_STATE_GATE); + } + } + + } + + return 0; +} + +/** + * vcn_v4_0_suspend - suspend VCN block + * + * @handle: amdgpu_device pointer + * + * HW fini and suspend VCN block + */ +static int vcn_v4_0_suspend(void *handle) +{ + int r; + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + + r = vcn_v4_0_hw_fini(adev); + if (r) + return r; + + r = amdgpu_vcn_suspend(adev); + + return r; +} + +/** + * vcn_v4_0_resume - resume VCN block + * + * @handle: amdgpu_device pointer + * + * Resume firmware and hw init VCN block + */ +static int vcn_v4_0_resume(void *handle) +{ + int r; + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + + r = amdgpu_vcn_resume(adev); + if (r) + return r; + + r = vcn_v4_0_hw_init(adev); + + return r; +} + +/** + * vcn_v4_0_mc_resume - memory controller programming + * + * @adev: amdgpu_device pointer + * @inst: instance number + * + * Let the VCN memory controller know it's offsets + */ +static void vcn_v4_0_mc_resume(struct amdgpu_device *adev, int inst) +{ + uint32_t offset, size; + const struct common_firmware_header *hdr; + + hdr = (const struct common_firmware_header *)adev->vcn.fw->data; + size = AMDGPU_GPU_PAGE_ALIGN(le32_to_cpu(hdr->ucode_size_bytes) + 8); + + /* cache window 0: fw */ + if (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) { + WREG32_SOC15(VCN, inst, regUVD_LMI_VCPU_CACHE_64BIT_BAR_LOW, + (adev->firmware.ucode[AMDGPU_UCODE_ID_VCN + inst].tmr_mc_addr_lo)); + WREG32_SOC15(VCN, inst, regUVD_LMI_VCPU_CACHE_64BIT_BAR_HIGH, + (adev->firmware.ucode[AMDGPU_UCODE_ID_VCN + inst].tmr_mc_addr_hi)); + WREG32_SOC15(VCN, inst, regUVD_VCPU_CACHE_OFFSET0, 0); + offset = 0; + } else { + WREG32_SOC15(VCN, inst, regUVD_LMI_VCPU_CACHE_64BIT_BAR_LOW, + lower_32_bits(adev->vcn.inst[inst].gpu_addr)); + WREG32_SOC15(VCN, inst, regUVD_LMI_VCPU_CACHE_64BIT_BAR_HIGH, + upper_32_bits(adev->vcn.inst[inst].gpu_addr)); + offset = size; + WREG32_SOC15(VCN, inst, regUVD_VCPU_CACHE_OFFSET0, AMDGPU_UVD_FIRMWARE_OFFSET >> 3); + } + WREG32_SOC15(VCN, inst, regUVD_VCPU_CACHE_SIZE0, size); + + /* cache window 1: stack */ + WREG32_SOC15(VCN, inst, regUVD_LMI_VCPU_CACHE1_64BIT_BAR_LOW, + lower_32_bits(adev->vcn.inst[inst].gpu_addr + offset)); + WREG32_SOC15(VCN, inst, regUVD_LMI_VCPU_CACHE1_64BIT_BAR_HIGH, + upper_32_bits(adev->vcn.inst[inst].gpu_addr + offset)); + WREG32_SOC15(VCN, inst, regUVD_VCPU_CACHE_OFFSET1, 0); + WREG32_SOC15(VCN, inst, regUVD_VCPU_CACHE_SIZE1, AMDGPU_VCN_STACK_SIZE); + + /* cache window 2: context */ + WREG32_SOC15(VCN, inst, regUVD_LMI_VCPU_CACHE2_64BIT_BAR_LOW, + lower_32_bits(adev->vcn.inst[inst].gpu_addr + offset + AMDGPU_VCN_STACK_SIZE)); + WREG32_SOC15(VCN, inst, regUVD_LMI_VCPU_CACHE2_64BIT_BAR_HIGH, + upper_32_bits(adev->vcn.inst[inst].gpu_addr + offset + AMDGPU_VCN_STACK_SIZE)); + WREG32_SOC15(VCN, inst, regUVD_VCPU_CACHE_OFFSET2, 0); + WREG32_SOC15(VCN, inst, regUVD_VCPU_CACHE_SIZE2, AMDGPU_VCN_CONTEXT_SIZE); + + /* non-cache window */ + WREG32_SOC15(VCN, inst, regUVD_LMI_VCPU_NC0_64BIT_BAR_LOW, + lower_32_bits(adev->vcn.inst[inst].fw_shared.gpu_addr)); + WREG32_SOC15(VCN, inst, regUVD_LMI_VCPU_NC0_64BIT_BAR_HIGH, + upper_32_bits(adev->vcn.inst[inst].fw_shared.gpu_addr)); + WREG32_SOC15(VCN, inst, regUVD_VCPU_NONCACHE_OFFSET0, 0); + WREG32_SOC15(VCN, inst, regUVD_VCPU_NONCACHE_SIZE0, + AMDGPU_GPU_PAGE_ALIGN(sizeof(struct amdgpu_vcn4_fw_shared))); +} + +/** + * vcn_v4_0_mc_resume_dpg_mode - memory controller programming for dpg mode + * + * @adev: amdgpu_device pointer + * @inst_idx: instance number index + * @indirect: indirectly write sram + * + * Let the VCN memory controller know it's offsets with dpg mode + */ +static void vcn_v4_0_mc_resume_dpg_mode(struct amdgpu_device *adev, int inst_idx, bool indirect) +{ + uint32_t offset, size; + const struct common_firmware_header *hdr; + hdr = (const struct common_firmware_header *)adev->vcn.fw->data; + size = AMDGPU_GPU_PAGE_ALIGN(le32_to_cpu(hdr->ucode_size_bytes) + 8); + + /* cache window 0: fw */ + if (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) { + if (!indirect) { + WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET( + VCN, inst_idx, regUVD_LMI_VCPU_CACHE_64BIT_BAR_LOW), + (adev->firmware.ucode[AMDGPU_UCODE_ID_VCN + inst_idx].tmr_mc_addr_lo), 0, indirect); + WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET( + VCN, inst_idx, regUVD_LMI_VCPU_CACHE_64BIT_BAR_HIGH), + (adev->firmware.ucode[AMDGPU_UCODE_ID_VCN + inst_idx].tmr_mc_addr_hi), 0, indirect); + WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET( + VCN, inst_idx, regUVD_VCPU_CACHE_OFFSET0), 0, 0, indirect); + } else { + WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET( + VCN, inst_idx, regUVD_LMI_VCPU_CACHE_64BIT_BAR_LOW), 0, 0, indirect); + WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET( + VCN, inst_idx, regUVD_LMI_VCPU_CACHE_64BIT_BAR_HIGH), 0, 0, indirect); + WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET( + VCN, inst_idx, regUVD_VCPU_CACHE_OFFSET0), 0, 0, indirect); + } + offset = 0; + } else { + WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET( + VCN, inst_idx, regUVD_LMI_VCPU_CACHE_64BIT_BAR_LOW), + lower_32_bits(adev->vcn.inst[inst_idx].gpu_addr), 0, indirect); + WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET( + VCN, inst_idx, regUVD_LMI_VCPU_CACHE_64BIT_BAR_HIGH), + upper_32_bits(adev->vcn.inst[inst_idx].gpu_addr), 0, indirect); + offset = size; + WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET( + VCN, inst_idx, regUVD_VCPU_CACHE_OFFSET0), + AMDGPU_UVD_FIRMWARE_OFFSET >> 3, 0, indirect); + } + + if (!indirect) + WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET( + VCN, inst_idx, regUVD_VCPU_CACHE_SIZE0), size, 0, indirect); + else + WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET( + VCN, inst_idx, regUVD_VCPU_CACHE_SIZE0), 0, 0, indirect); + + /* cache window 1: stack */ + if (!indirect) { + WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET( + VCN, inst_idx, regUVD_LMI_VCPU_CACHE1_64BIT_BAR_LOW), + lower_32_bits(adev->vcn.inst[inst_idx].gpu_addr + offset), 0, indirect); + WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET( + VCN, inst_idx, regUVD_LMI_VCPU_CACHE1_64BIT_BAR_HIGH), + upper_32_bits(adev->vcn.inst[inst_idx].gpu_addr + offset), 0, indirect); + WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET( + VCN, inst_idx, regUVD_VCPU_CACHE_OFFSET1), 0, 0, indirect); + } else { + WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET( + VCN, inst_idx, regUVD_LMI_VCPU_CACHE1_64BIT_BAR_LOW), 0, 0, indirect); + WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET( + VCN, inst_idx, regUVD_LMI_VCPU_CACHE1_64BIT_BAR_HIGH), 0, 0, indirect); + WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET( + VCN, inst_idx, regUVD_VCPU_CACHE_OFFSET1), 0, 0, indirect); + } + WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET( + VCN, inst_idx, regUVD_VCPU_CACHE_SIZE1), AMDGPU_VCN_STACK_SIZE, 0, indirect); + + /* cache window 2: context */ + WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET( + VCN, inst_idx, regUVD_LMI_VCPU_CACHE2_64BIT_BAR_LOW), + lower_32_bits(adev->vcn.inst[inst_idx].gpu_addr + offset + AMDGPU_VCN_STACK_SIZE), 0, indirect); + WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET( + VCN, inst_idx, regUVD_LMI_VCPU_CACHE2_64BIT_BAR_HIGH), + upper_32_bits(adev->vcn.inst[inst_idx].gpu_addr + offset + AMDGPU_VCN_STACK_SIZE), 0, indirect); + WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET( + VCN, inst_idx, regUVD_VCPU_CACHE_OFFSET2), 0, 0, indirect); + WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET( + VCN, inst_idx, regUVD_VCPU_CACHE_SIZE2), AMDGPU_VCN_CONTEXT_SIZE, 0, indirect); + + /* non-cache window */ + WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET( + VCN, inst_idx, regUVD_LMI_VCPU_NC0_64BIT_BAR_LOW), + lower_32_bits(adev->vcn.inst[inst_idx].fw_shared.gpu_addr), 0, indirect); + WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET( + VCN, inst_idx, regUVD_LMI_VCPU_NC0_64BIT_BAR_HIGH), + upper_32_bits(adev->vcn.inst[inst_idx].fw_shared.gpu_addr), 0, indirect); + WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET( + VCN, inst_idx, regUVD_VCPU_NONCACHE_OFFSET0), 0, 0, indirect); + WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET( + VCN, inst_idx, regUVD_VCPU_NONCACHE_SIZE0), + AMDGPU_GPU_PAGE_ALIGN(sizeof(struct amdgpu_vcn4_fw_shared)), 0, indirect); + + /* VCN global tiling registers */ + WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET( + VCN, 0, regUVD_GFX10_ADDR_CONFIG), adev->gfx.config.gb_addr_config, 0, indirect); +} + +/** + * vcn_v4_0_disable_static_power_gating - disable VCN static power gating + * + * @adev: amdgpu_device pointer + * @inst: instance number + * + * Disable static power gating for VCN block + */ +static void vcn_v4_0_disable_static_power_gating(struct amdgpu_device *adev, int inst) +{ + uint32_t data = 0; + + if (adev->pg_flags & AMD_PG_SUPPORT_VCN) { + data = (1 << UVD_PGFSM_CONFIG__UVDM_PWR_CONFIG__SHIFT + | 1 << UVD_PGFSM_CONFIG__UVDS_PWR_CONFIG__SHIFT + | 1 << UVD_PGFSM_CONFIG__UVDLM_PWR_CONFIG__SHIFT + | 2 << UVD_PGFSM_CONFIG__UVDF_PWR_CONFIG__SHIFT + | 2 << UVD_PGFSM_CONFIG__UVDTC_PWR_CONFIG__SHIFT + | 2 << UVD_PGFSM_CONFIG__UVDB_PWR_CONFIG__SHIFT + | 2 << UVD_PGFSM_CONFIG__UVDTA_PWR_CONFIG__SHIFT + | 2 << UVD_PGFSM_CONFIG__UVDTD_PWR_CONFIG__SHIFT + | 2 << UVD_PGFSM_CONFIG__UVDTE_PWR_CONFIG__SHIFT + | 2 << UVD_PGFSM_CONFIG__UVDE_PWR_CONFIG__SHIFT + | 2 << UVD_PGFSM_CONFIG__UVDAB_PWR_CONFIG__SHIFT + | 2 << UVD_PGFSM_CONFIG__UVDTB_PWR_CONFIG__SHIFT + | 2 << UVD_PGFSM_CONFIG__UVDNA_PWR_CONFIG__SHIFT + | 2 << UVD_PGFSM_CONFIG__UVDNB_PWR_CONFIG__SHIFT); + + WREG32_SOC15(VCN, inst, regUVD_PGFSM_CONFIG, data); + SOC15_WAIT_ON_RREG(VCN, inst, regUVD_PGFSM_STATUS, + UVD_PGFSM_STATUS__UVDM_UVDU_UVDLM_PWR_ON_3_0, 0x3F3FFFFF); + } else { + uint32_t value; + + value = (inst) ? 0x2200800 : 0; + data = (1 << UVD_PGFSM_CONFIG__UVDM_PWR_CONFIG__SHIFT + | 1 << UVD_PGFSM_CONFIG__UVDS_PWR_CONFIG__SHIFT + | 1 << UVD_PGFSM_CONFIG__UVDLM_PWR_CONFIG__SHIFT + | 1 << UVD_PGFSM_CONFIG__UVDF_PWR_CONFIG__SHIFT + | 1 << UVD_PGFSM_CONFIG__UVDTC_PWR_CONFIG__SHIFT + | 1 << UVD_PGFSM_CONFIG__UVDB_PWR_CONFIG__SHIFT + | 1 << UVD_PGFSM_CONFIG__UVDTA_PWR_CONFIG__SHIFT + | 1 << UVD_PGFSM_CONFIG__UVDTD_PWR_CONFIG__SHIFT + | 1 << UVD_PGFSM_CONFIG__UVDTE_PWR_CONFIG__SHIFT + | 1 << UVD_PGFSM_CONFIG__UVDE_PWR_CONFIG__SHIFT + | 1 << UVD_PGFSM_CONFIG__UVDAB_PWR_CONFIG__SHIFT + | 1 << UVD_PGFSM_CONFIG__UVDTB_PWR_CONFIG__SHIFT + | 1 << UVD_PGFSM_CONFIG__UVDNA_PWR_CONFIG__SHIFT + | 1 << UVD_PGFSM_CONFIG__UVDNB_PWR_CONFIG__SHIFT); + + WREG32_SOC15(VCN, inst, regUVD_PGFSM_CONFIG, data); + SOC15_WAIT_ON_RREG(VCN, inst, regUVD_PGFSM_STATUS, value, 0x3F3FFFFF); + } + + data = RREG32_SOC15(VCN, inst, regUVD_POWER_STATUS); + data &= ~0x103; + if (adev->pg_flags & AMD_PG_SUPPORT_VCN) + data |= UVD_PGFSM_CONFIG__UVDM_UVDU_PWR_ON | + UVD_POWER_STATUS__UVD_PG_EN_MASK; + + WREG32_SOC15(VCN, inst, regUVD_POWER_STATUS, data); + + return; +} + +/** + * vcn_v4_0_enable_static_power_gating - enable VCN static power gating + * + * @adev: amdgpu_device pointer + * @inst: instance number + * + * Enable static power gating for VCN block + */ +static void vcn_v4_0_enable_static_power_gating(struct amdgpu_device *adev, int inst) +{ + uint32_t data; + + if (adev->pg_flags & AMD_PG_SUPPORT_VCN) { + /* Before power off, this indicator has to be turned on */ + data = RREG32_SOC15(VCN, inst, regUVD_POWER_STATUS); + data &= ~UVD_POWER_STATUS__UVD_POWER_STATUS_MASK; + data |= UVD_POWER_STATUS__UVD_POWER_STATUS_TILES_OFF; + WREG32_SOC15(VCN, inst, regUVD_POWER_STATUS, data); + + data = (2 << UVD_PGFSM_CONFIG__UVDM_PWR_CONFIG__SHIFT + | 2 << UVD_PGFSM_CONFIG__UVDS_PWR_CONFIG__SHIFT + | 2 << UVD_PGFSM_CONFIG__UVDF_PWR_CONFIG__SHIFT + | 2 << UVD_PGFSM_CONFIG__UVDTC_PWR_CONFIG__SHIFT + | 2 << UVD_PGFSM_CONFIG__UVDB_PWR_CONFIG__SHIFT + | 2 << UVD_PGFSM_CONFIG__UVDTA_PWR_CONFIG__SHIFT + | 2 << UVD_PGFSM_CONFIG__UVDLM_PWR_CONFIG__SHIFT + | 2 << UVD_PGFSM_CONFIG__UVDTD_PWR_CONFIG__SHIFT + | 2 << UVD_PGFSM_CONFIG__UVDTE_PWR_CONFIG__SHIFT + | 2 << UVD_PGFSM_CONFIG__UVDE_PWR_CONFIG__SHIFT + | 2 << UVD_PGFSM_CONFIG__UVDAB_PWR_CONFIG__SHIFT + | 2 << UVD_PGFSM_CONFIG__UVDTB_PWR_CONFIG__SHIFT + | 2 << UVD_PGFSM_CONFIG__UVDNA_PWR_CONFIG__SHIFT + | 2 << UVD_PGFSM_CONFIG__UVDNB_PWR_CONFIG__SHIFT); + WREG32_SOC15(VCN, inst, regUVD_PGFSM_CONFIG, data); + + data = (2 << UVD_PGFSM_STATUS__UVDM_PWR_STATUS__SHIFT + | 2 << UVD_PGFSM_STATUS__UVDS_PWR_STATUS__SHIFT + | 2 << UVD_PGFSM_STATUS__UVDF_PWR_STATUS__SHIFT + | 2 << UVD_PGFSM_STATUS__UVDTC_PWR_STATUS__SHIFT + | 2 << UVD_PGFSM_STATUS__UVDB_PWR_STATUS__SHIFT + | 2 << UVD_PGFSM_STATUS__UVDTA_PWR_STATUS__SHIFT + | 2 << UVD_PGFSM_STATUS__UVDLM_PWR_STATUS__SHIFT + | 2 << UVD_PGFSM_STATUS__UVDTD_PWR_STATUS__SHIFT + | 2 << UVD_PGFSM_STATUS__UVDTE_PWR_STATUS__SHIFT + | 2 << UVD_PGFSM_STATUS__UVDE_PWR_STATUS__SHIFT + | 2 << UVD_PGFSM_STATUS__UVDAB_PWR_STATUS__SHIFT + | 2 << UVD_PGFSM_STATUS__UVDTB_PWR_STATUS__SHIFT + | 2 << UVD_PGFSM_STATUS__UVDNA_PWR_STATUS__SHIFT + | 2 << UVD_PGFSM_STATUS__UVDNB_PWR_STATUS__SHIFT); + SOC15_WAIT_ON_RREG(VCN, inst, regUVD_PGFSM_STATUS, data, 0x3F3FFFFF); + } + + return; +} + +/** + * vcn_v4_0_disable_clock_gating - disable VCN clock gating + * + * @adev: amdgpu_device pointer + * @inst: instance number + * + * Disable clock gating for VCN block + */ +static void vcn_v4_0_disable_clock_gating(struct amdgpu_device *adev, int inst) +{ + uint32_t data; + + if (adev->cg_flags & AMD_CG_SUPPORT_VCN_MGCG) + return; + + /* VCN disable CGC */ + data = RREG32_SOC15(VCN, inst, regUVD_CGC_CTRL); + data &= ~UVD_CGC_CTRL__DYN_CLOCK_MODE_MASK; + data |= 1 << UVD_CGC_CTRL__CLK_GATE_DLY_TIMER__SHIFT; + data |= 4 << UVD_CGC_CTRL__CLK_OFF_DELAY__SHIFT; + WREG32_SOC15(VCN, inst, regUVD_CGC_CTRL, data); + + data = RREG32_SOC15(VCN, inst, regUVD_CGC_GATE); + data &= ~(UVD_CGC_GATE__SYS_MASK + | UVD_CGC_GATE__UDEC_MASK + | UVD_CGC_GATE__MPEG2_MASK + | UVD_CGC_GATE__REGS_MASK + | UVD_CGC_GATE__RBC_MASK + | UVD_CGC_GATE__LMI_MC_MASK + | UVD_CGC_GATE__LMI_UMC_MASK + | UVD_CGC_GATE__IDCT_MASK + | UVD_CGC_GATE__MPRD_MASK + | UVD_CGC_GATE__MPC_MASK + | UVD_CGC_GATE__LBSI_MASK + | UVD_CGC_GATE__LRBBM_MASK + | UVD_CGC_GATE__UDEC_RE_MASK + | UVD_CGC_GATE__UDEC_CM_MASK + | UVD_CGC_GATE__UDEC_IT_MASK + | UVD_CGC_GATE__UDEC_DB_MASK + | UVD_CGC_GATE__UDEC_MP_MASK + | UVD_CGC_GATE__WCB_MASK + | UVD_CGC_GATE__VCPU_MASK + | UVD_CGC_GATE__MMSCH_MASK); + + WREG32_SOC15(VCN, inst, regUVD_CGC_GATE, data); + SOC15_WAIT_ON_RREG(VCN, inst, regUVD_CGC_GATE, 0, 0xFFFFFFFF); + + data = RREG32_SOC15(VCN, inst, regUVD_CGC_CTRL); + data &= ~(UVD_CGC_CTRL__UDEC_RE_MODE_MASK + | UVD_CGC_CTRL__UDEC_CM_MODE_MASK + | UVD_CGC_CTRL__UDEC_IT_MODE_MASK + | UVD_CGC_CTRL__UDEC_DB_MODE_MASK + | UVD_CGC_CTRL__UDEC_MP_MODE_MASK + | UVD_CGC_CTRL__SYS_MODE_MASK + | UVD_CGC_CTRL__UDEC_MODE_MASK + | UVD_CGC_CTRL__MPEG2_MODE_MASK + | UVD_CGC_CTRL__REGS_MODE_MASK + | UVD_CGC_CTRL__RBC_MODE_MASK + | UVD_CGC_CTRL__LMI_MC_MODE_MASK + | UVD_CGC_CTRL__LMI_UMC_MODE_MASK + | UVD_CGC_CTRL__IDCT_MODE_MASK + | UVD_CGC_CTRL__MPRD_MODE_MASK + | UVD_CGC_CTRL__MPC_MODE_MASK + | UVD_CGC_CTRL__LBSI_MODE_MASK + | UVD_CGC_CTRL__LRBBM_MODE_MASK + | UVD_CGC_CTRL__WCB_MODE_MASK + | UVD_CGC_CTRL__VCPU_MODE_MASK + | UVD_CGC_CTRL__MMSCH_MODE_MASK); + WREG32_SOC15(VCN, inst, regUVD_CGC_CTRL, data); + + data = RREG32_SOC15(VCN, inst, regUVD_SUVD_CGC_GATE); + data |= (UVD_SUVD_CGC_GATE__SRE_MASK + | UVD_SUVD_CGC_GATE__SIT_MASK + | UVD_SUVD_CGC_GATE__SMP_MASK + | UVD_SUVD_CGC_GATE__SCM_MASK + | UVD_SUVD_CGC_GATE__SDB_MASK + | UVD_SUVD_CGC_GATE__SRE_H264_MASK + | UVD_SUVD_CGC_GATE__SRE_HEVC_MASK + | UVD_SUVD_CGC_GATE__SIT_H264_MASK + | UVD_SUVD_CGC_GATE__SIT_HEVC_MASK + | UVD_SUVD_CGC_GATE__SCM_H264_MASK + | UVD_SUVD_CGC_GATE__SCM_HEVC_MASK + | UVD_SUVD_CGC_GATE__SDB_H264_MASK + | UVD_SUVD_CGC_GATE__SDB_HEVC_MASK + | UVD_SUVD_CGC_GATE__SCLR_MASK + | UVD_SUVD_CGC_GATE__UVD_SC_MASK + | UVD_SUVD_CGC_GATE__ENT_MASK + | UVD_SUVD_CGC_GATE__SIT_HEVC_DEC_MASK + | UVD_SUVD_CGC_GATE__SIT_HEVC_ENC_MASK + | UVD_SUVD_CGC_GATE__SITE_MASK + | UVD_SUVD_CGC_GATE__SRE_VP9_MASK + | UVD_SUVD_CGC_GATE__SCM_VP9_MASK + | UVD_SUVD_CGC_GATE__SIT_VP9_DEC_MASK + | UVD_SUVD_CGC_GATE__SDB_VP9_MASK + | UVD_SUVD_CGC_GATE__IME_HEVC_MASK); + WREG32_SOC15(VCN, inst, regUVD_SUVD_CGC_GATE, data); + + data = RREG32_SOC15(VCN, inst, regUVD_SUVD_CGC_CTRL); + data &= ~(UVD_SUVD_CGC_CTRL__SRE_MODE_MASK + | UVD_SUVD_CGC_CTRL__SIT_MODE_MASK + | UVD_SUVD_CGC_CTRL__SMP_MODE_MASK + | UVD_SUVD_CGC_CTRL__SCM_MODE_MASK + | UVD_SUVD_CGC_CTRL__SDB_MODE_MASK + | UVD_SUVD_CGC_CTRL__SCLR_MODE_MASK + | UVD_SUVD_CGC_CTRL__UVD_SC_MODE_MASK + | UVD_SUVD_CGC_CTRL__ENT_MODE_MASK + | UVD_SUVD_CGC_CTRL__IME_MODE_MASK + | UVD_SUVD_CGC_CTRL__SITE_MODE_MASK); + WREG32_SOC15(VCN, inst, regUVD_SUVD_CGC_CTRL, data); +} + +/** + * vcn_v4_0_disable_clock_gating_dpg_mode - disable VCN clock gating dpg mode + * + * @adev: amdgpu_device pointer + * @sram_sel: sram select + * @inst_idx: instance number index + * @indirect: indirectly write sram + * + * Disable clock gating for VCN block with dpg mode + */ +static void vcn_v4_0_disable_clock_gating_dpg_mode(struct amdgpu_device *adev, uint8_t sram_sel, + int inst_idx, uint8_t indirect) +{ + uint32_t reg_data = 0; + + if (adev->cg_flags & AMD_CG_SUPPORT_VCN_MGCG) + return; + + /* enable sw clock gating control */ + reg_data = 0 << UVD_CGC_CTRL__DYN_CLOCK_MODE__SHIFT; + reg_data |= 1 << UVD_CGC_CTRL__CLK_GATE_DLY_TIMER__SHIFT; + reg_data |= 4 << UVD_CGC_CTRL__CLK_OFF_DELAY__SHIFT; + reg_data &= ~(UVD_CGC_CTRL__UDEC_RE_MODE_MASK | + UVD_CGC_CTRL__UDEC_CM_MODE_MASK | + UVD_CGC_CTRL__UDEC_IT_MODE_MASK | + UVD_CGC_CTRL__UDEC_DB_MODE_MASK | + UVD_CGC_CTRL__UDEC_MP_MODE_MASK | + UVD_CGC_CTRL__SYS_MODE_MASK | + UVD_CGC_CTRL__UDEC_MODE_MASK | + UVD_CGC_CTRL__MPEG2_MODE_MASK | + UVD_CGC_CTRL__REGS_MODE_MASK | + UVD_CGC_CTRL__RBC_MODE_MASK | + UVD_CGC_CTRL__LMI_MC_MODE_MASK | + UVD_CGC_CTRL__LMI_UMC_MODE_MASK | + UVD_CGC_CTRL__IDCT_MODE_MASK | + UVD_CGC_CTRL__MPRD_MODE_MASK | + UVD_CGC_CTRL__MPC_MODE_MASK | + UVD_CGC_CTRL__LBSI_MODE_MASK | + UVD_CGC_CTRL__LRBBM_MODE_MASK | + UVD_CGC_CTRL__WCB_MODE_MASK | + UVD_CGC_CTRL__VCPU_MODE_MASK); + WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET( + VCN, inst_idx, regUVD_CGC_CTRL), reg_data, sram_sel, indirect); + + /* turn off clock gating */ + WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET( + VCN, inst_idx, regUVD_CGC_GATE), 0, sram_sel, indirect); + + /* turn on SUVD clock gating */ + WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET( + VCN, inst_idx, regUVD_SUVD_CGC_GATE), 1, sram_sel, indirect); + + /* turn on sw mode in UVD_SUVD_CGC_CTRL */ + WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET( + VCN, inst_idx, regUVD_SUVD_CGC_CTRL), 0, sram_sel, indirect); +} + +/** + * vcn_v4_0_enable_clock_gating - enable VCN clock gating + * + * @adev: amdgpu_device pointer + * @inst: instance number + * + * Enable clock gating for VCN block + */ +static void vcn_v4_0_enable_clock_gating(struct amdgpu_device *adev, int inst) +{ + uint32_t data; + + if (adev->cg_flags & AMD_CG_SUPPORT_VCN_MGCG) + return; + + /* enable VCN CGC */ + data = RREG32_SOC15(VCN, inst, regUVD_CGC_CTRL); + data |= 0 << UVD_CGC_CTRL__DYN_CLOCK_MODE__SHIFT; + data |= 1 << UVD_CGC_CTRL__CLK_GATE_DLY_TIMER__SHIFT; + data |= 4 << UVD_CGC_CTRL__CLK_OFF_DELAY__SHIFT; + WREG32_SOC15(VCN, inst, regUVD_CGC_CTRL, data); + + data = RREG32_SOC15(VCN, inst, regUVD_CGC_CTRL); + data |= (UVD_CGC_CTRL__UDEC_RE_MODE_MASK + | UVD_CGC_CTRL__UDEC_CM_MODE_MASK + | UVD_CGC_CTRL__UDEC_IT_MODE_MASK + | UVD_CGC_CTRL__UDEC_DB_MODE_MASK + | UVD_CGC_CTRL__UDEC_MP_MODE_MASK + | UVD_CGC_CTRL__SYS_MODE_MASK + | UVD_CGC_CTRL__UDEC_MODE_MASK + | UVD_CGC_CTRL__MPEG2_MODE_MASK + | UVD_CGC_CTRL__REGS_MODE_MASK + | UVD_CGC_CTRL__RBC_MODE_MASK + | UVD_CGC_CTRL__LMI_MC_MODE_MASK + | UVD_CGC_CTRL__LMI_UMC_MODE_MASK + | UVD_CGC_CTRL__IDCT_MODE_MASK + | UVD_CGC_CTRL__MPRD_MODE_MASK + | UVD_CGC_CTRL__MPC_MODE_MASK + | UVD_CGC_CTRL__LBSI_MODE_MASK + | UVD_CGC_CTRL__LRBBM_MODE_MASK + | UVD_CGC_CTRL__WCB_MODE_MASK + | UVD_CGC_CTRL__VCPU_MODE_MASK + | UVD_CGC_CTRL__MMSCH_MODE_MASK); + WREG32_SOC15(VCN, inst, regUVD_CGC_CTRL, data); + + data = RREG32_SOC15(VCN, inst, regUVD_SUVD_CGC_CTRL); + data |= (UVD_SUVD_CGC_CTRL__SRE_MODE_MASK + | UVD_SUVD_CGC_CTRL__SIT_MODE_MASK + | UVD_SUVD_CGC_CTRL__SMP_MODE_MASK + | UVD_SUVD_CGC_CTRL__SCM_MODE_MASK + | UVD_SUVD_CGC_CTRL__SDB_MODE_MASK + | UVD_SUVD_CGC_CTRL__SCLR_MODE_MASK + | UVD_SUVD_CGC_CTRL__UVD_SC_MODE_MASK + | UVD_SUVD_CGC_CTRL__ENT_MODE_MASK + | UVD_SUVD_CGC_CTRL__IME_MODE_MASK + | UVD_SUVD_CGC_CTRL__SITE_MODE_MASK); + WREG32_SOC15(VCN, inst, regUVD_SUVD_CGC_CTRL, data); + + return; +} + +/** + * vcn_v4_0_start_dpg_mode - VCN start with dpg mode + * + * @adev: amdgpu_device pointer + * @inst_idx: instance number index + * @indirect: indirectly write sram + * + * Start VCN block with dpg mode + */ +static int vcn_v4_0_start_dpg_mode(struct amdgpu_device *adev, int inst_idx, bool indirect) +{ + volatile struct amdgpu_vcn4_fw_shared *fw_shared = adev->vcn.inst[inst_idx].fw_shared.cpu_addr; + struct amdgpu_ring *ring; + uint32_t tmp; + + /* disable register anti-hang mechanism */ + WREG32_P(SOC15_REG_OFFSET(VCN, inst_idx, regUVD_POWER_STATUS), 1, + ~UVD_POWER_STATUS__UVD_POWER_STATUS_MASK); + /* enable dynamic power gating mode */ + tmp = RREG32_SOC15(VCN, inst_idx, regUVD_POWER_STATUS); + tmp |= UVD_POWER_STATUS__UVD_PG_MODE_MASK; + tmp |= UVD_POWER_STATUS__UVD_PG_EN_MASK; + WREG32_SOC15(VCN, inst_idx, regUVD_POWER_STATUS, tmp); + + if (indirect) + adev->vcn.inst[inst_idx].dpg_sram_curr_addr = (uint32_t *)adev->vcn.inst[inst_idx].dpg_sram_cpu_addr; + + /* enable clock gating */ + vcn_v4_0_disable_clock_gating_dpg_mode(adev, 0, inst_idx, indirect); + + /* enable VCPU clock */ + tmp = (0xFF << UVD_VCPU_CNTL__PRB_TIMEOUT_VAL__SHIFT); + tmp |= UVD_VCPU_CNTL__CLK_EN_MASK | UVD_VCPU_CNTL__BLK_RST_MASK; + WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET( + VCN, inst_idx, regUVD_VCPU_CNTL), tmp, 0, indirect); + + /* disable master interupt */ + WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET( + VCN, inst_idx, regUVD_MASTINT_EN), 0, 0, indirect); + + /* setup regUVD_LMI_CTRL */ + tmp = (UVD_LMI_CTRL__WRITE_CLEAN_TIMER_EN_MASK | + UVD_LMI_CTRL__REQ_MODE_MASK | + UVD_LMI_CTRL__CRC_RESET_MASK | + UVD_LMI_CTRL__MASK_MC_URGENT_MASK | + UVD_LMI_CTRL__DATA_COHERENCY_EN_MASK | + UVD_LMI_CTRL__VCPU_DATA_COHERENCY_EN_MASK | + (8 << UVD_LMI_CTRL__WRITE_CLEAN_TIMER__SHIFT) | + 0x00100000L); + WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET( + VCN, inst_idx, regUVD_LMI_CTRL), tmp, 0, indirect); + + WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET( + VCN, inst_idx, regUVD_MPC_CNTL), + 0x2 << UVD_MPC_CNTL__REPLACEMENT_MODE__SHIFT, 0, indirect); + + WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET( + VCN, inst_idx, regUVD_MPC_SET_MUXA0), + ((0x1 << UVD_MPC_SET_MUXA0__VARA_1__SHIFT) | + (0x2 << UVD_MPC_SET_MUXA0__VARA_2__SHIFT) | + (0x3 << UVD_MPC_SET_MUXA0__VARA_3__SHIFT) | + (0x4 << UVD_MPC_SET_MUXA0__VARA_4__SHIFT)), 0, indirect); + + WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET( + VCN, inst_idx, regUVD_MPC_SET_MUXB0), + ((0x1 << UVD_MPC_SET_MUXB0__VARB_1__SHIFT) | + (0x2 << UVD_MPC_SET_MUXB0__VARB_2__SHIFT) | + (0x3 << UVD_MPC_SET_MUXB0__VARB_3__SHIFT) | + (0x4 << UVD_MPC_SET_MUXB0__VARB_4__SHIFT)), 0, indirect); + + WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET( + VCN, inst_idx, regUVD_MPC_SET_MUX), + ((0x0 << UVD_MPC_SET_MUX__SET_0__SHIFT) | + (0x1 << UVD_MPC_SET_MUX__SET_1__SHIFT) | + (0x2 << UVD_MPC_SET_MUX__SET_2__SHIFT)), 0, indirect); + + vcn_v4_0_mc_resume_dpg_mode(adev, inst_idx, indirect); + + tmp = (0xFF << UVD_VCPU_CNTL__PRB_TIMEOUT_VAL__SHIFT); + tmp |= UVD_VCPU_CNTL__CLK_EN_MASK; + WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET( + VCN, inst_idx, regUVD_VCPU_CNTL), tmp, 0, indirect); + + /* enable LMI MC and UMC channels */ + tmp = 0x1f << UVD_LMI_CTRL2__RE_OFLD_MIF_WR_REQ_NUM__SHIFT; + WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET( + VCN, inst_idx, regUVD_LMI_CTRL2), tmp, 0, indirect); + + /* enable master interrupt */ + WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET( + VCN, inst_idx, regUVD_MASTINT_EN), + UVD_MASTINT_EN__VCPU_EN_MASK, 0, indirect); + + + if (indirect) + psp_update_vcn_sram(adev, inst_idx, adev->vcn.inst[inst_idx].dpg_sram_gpu_addr, + (uint32_t)((uintptr_t)adev->vcn.inst[inst_idx].dpg_sram_curr_addr - + (uintptr_t)adev->vcn.inst[inst_idx].dpg_sram_cpu_addr)); + + ring = &adev->vcn.inst[inst_idx].ring_enc[0]; + + WREG32_SOC15(VCN, inst_idx, regUVD_RB_BASE_LO, ring->gpu_addr); + WREG32_SOC15(VCN, inst_idx, regUVD_RB_BASE_HI, upper_32_bits(ring->gpu_addr)); + WREG32_SOC15(VCN, inst_idx, regUVD_RB_SIZE, ring->ring_size / 4); + + tmp = RREG32_SOC15(VCN, inst_idx, regVCN_RB_ENABLE); + tmp &= ~(VCN_RB_ENABLE__RB1_EN_MASK); + WREG32_SOC15(VCN, inst_idx, regVCN_RB_ENABLE, tmp); + fw_shared->sq.queue_mode |= FW_QUEUE_RING_RESET; + WREG32_SOC15(VCN, inst_idx, regUVD_RB_RPTR, 0); + WREG32_SOC15(VCN, inst_idx, regUVD_RB_WPTR, 0); + + tmp = RREG32_SOC15(VCN, inst_idx, regUVD_RB_RPTR); + WREG32_SOC15(VCN, inst_idx, regUVD_RB_WPTR, tmp); + ring->wptr = RREG32_SOC15(VCN, inst_idx, regUVD_RB_WPTR); + + tmp = RREG32_SOC15(VCN, inst_idx, regVCN_RB_ENABLE); + tmp |= VCN_RB_ENABLE__RB1_EN_MASK; + WREG32_SOC15(VCN, inst_idx, regVCN_RB_ENABLE, tmp); + fw_shared->sq.queue_mode &= ~(FW_QUEUE_RING_RESET | FW_QUEUE_DPG_HOLD_OFF); + + WREG32_SOC15(VCN, inst_idx, regVCN_RB1_DB_CTRL, + ring->doorbell_index << VCN_RB1_DB_CTRL__OFFSET__SHIFT | + VCN_RB1_DB_CTRL__EN_MASK); + + return 0; +} + + +/** + * vcn_v4_0_start - VCN start + * + * @adev: amdgpu_device pointer + * + * Start VCN block + */ +static int vcn_v4_0_start(struct amdgpu_device *adev) +{ + volatile struct amdgpu_vcn4_fw_shared *fw_shared; + struct amdgpu_ring *ring; + uint32_t tmp; + int i, j, k, r; + + if (adev->pm.dpm_enabled) + amdgpu_dpm_enable_uvd(adev, true); + + for (i = 0; i < adev->vcn.num_vcn_inst; ++i) { + fw_shared = adev->vcn.inst[i].fw_shared.cpu_addr; + + if (adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG) { + r = vcn_v4_0_start_dpg_mode(adev, i, adev->vcn.indirect_sram); + continue; + } + + /* disable VCN power gating */ + vcn_v4_0_disable_static_power_gating(adev, i); + + /* set VCN status busy */ + tmp = RREG32_SOC15(VCN, i, regUVD_STATUS) | UVD_STATUS__UVD_BUSY; + WREG32_SOC15(VCN, i, regUVD_STATUS, tmp); + + /*SW clock gating */ + vcn_v4_0_disable_clock_gating(adev, i); + + /* enable VCPU clock */ + WREG32_P(SOC15_REG_OFFSET(VCN, i, regUVD_VCPU_CNTL), + UVD_VCPU_CNTL__CLK_EN_MASK, ~UVD_VCPU_CNTL__CLK_EN_MASK); + + /* disable master interrupt */ + WREG32_P(SOC15_REG_OFFSET(VCN, i, regUVD_MASTINT_EN), 0, + ~UVD_MASTINT_EN__VCPU_EN_MASK); + + /* enable LMI MC and UMC channels */ + WREG32_P(SOC15_REG_OFFSET(VCN, i, regUVD_LMI_CTRL2), 0, + ~UVD_LMI_CTRL2__STALL_ARB_UMC_MASK); + + tmp = RREG32_SOC15(VCN, i, regUVD_SOFT_RESET); + tmp &= ~UVD_SOFT_RESET__LMI_SOFT_RESET_MASK; + tmp &= ~UVD_SOFT_RESET__LMI_UMC_SOFT_RESET_MASK; + WREG32_SOC15(VCN, i, regUVD_SOFT_RESET, tmp); + + /* setup regUVD_LMI_CTRL */ + tmp = RREG32_SOC15(VCN, i, regUVD_LMI_CTRL); + WREG32_SOC15(VCN, i, regUVD_LMI_CTRL, tmp | + UVD_LMI_CTRL__WRITE_CLEAN_TIMER_EN_MASK | + UVD_LMI_CTRL__MASK_MC_URGENT_MASK | + UVD_LMI_CTRL__DATA_COHERENCY_EN_MASK | + UVD_LMI_CTRL__VCPU_DATA_COHERENCY_EN_MASK); + + /* setup regUVD_MPC_CNTL */ + tmp = RREG32_SOC15(VCN, i, regUVD_MPC_CNTL); + tmp &= ~UVD_MPC_CNTL__REPLACEMENT_MODE_MASK; + tmp |= 0x2 << UVD_MPC_CNTL__REPLACEMENT_MODE__SHIFT; + WREG32_SOC15(VCN, i, regUVD_MPC_CNTL, tmp); + + /* setup UVD_MPC_SET_MUXA0 */ + WREG32_SOC15(VCN, i, regUVD_MPC_SET_MUXA0, + ((0x1 << UVD_MPC_SET_MUXA0__VARA_1__SHIFT) | + (0x2 << UVD_MPC_SET_MUXA0__VARA_2__SHIFT) | + (0x3 << UVD_MPC_SET_MUXA0__VARA_3__SHIFT) | + (0x4 << UVD_MPC_SET_MUXA0__VARA_4__SHIFT))); + + /* setup UVD_MPC_SET_MUXB0 */ + WREG32_SOC15(VCN, i, regUVD_MPC_SET_MUXB0, + ((0x1 << UVD_MPC_SET_MUXB0__VARB_1__SHIFT) | + (0x2 << UVD_MPC_SET_MUXB0__VARB_2__SHIFT) | + (0x3 << UVD_MPC_SET_MUXB0__VARB_3__SHIFT) | + (0x4 << UVD_MPC_SET_MUXB0__VARB_4__SHIFT))); + + /* setup UVD_MPC_SET_MUX */ + WREG32_SOC15(VCN, i, regUVD_MPC_SET_MUX, + ((0x0 << UVD_MPC_SET_MUX__SET_0__SHIFT) | + (0x1 << UVD_MPC_SET_MUX__SET_1__SHIFT) | + (0x2 << UVD_MPC_SET_MUX__SET_2__SHIFT))); + + vcn_v4_0_mc_resume(adev, i); + + /* VCN global tiling registers */ + WREG32_SOC15(VCN, i, regUVD_GFX10_ADDR_CONFIG, + adev->gfx.config.gb_addr_config); + + /* unblock VCPU register access */ + WREG32_P(SOC15_REG_OFFSET(VCN, i, regUVD_RB_ARB_CTRL), 0, + ~UVD_RB_ARB_CTRL__VCPU_DIS_MASK); + + /* release VCPU reset to boot */ + WREG32_P(SOC15_REG_OFFSET(VCN, i, regUVD_VCPU_CNTL), 0, + ~UVD_VCPU_CNTL__BLK_RST_MASK); + + for (j = 0; j < 10; ++j) { + uint32_t status; + + for (k = 0; k < 100; ++k) { + status = RREG32_SOC15(VCN, i, regUVD_STATUS); + if (status & 2) + break; + mdelay(10); + if (amdgpu_emu_mode==1) + msleep(1); + } + + if (amdgpu_emu_mode==1) { + r = -1; + if (status & 2) { + r = 0; + break; + } + } else { + r = 0; + if (status & 2) + break; + + dev_err(adev->dev, "VCN[%d] is not responding, trying to reset the VCPU!!!\n", i); + WREG32_P(SOC15_REG_OFFSET(VCN, i, regUVD_VCPU_CNTL), + UVD_VCPU_CNTL__BLK_RST_MASK, + ~UVD_VCPU_CNTL__BLK_RST_MASK); + mdelay(10); + WREG32_P(SOC15_REG_OFFSET(VCN, i, regUVD_VCPU_CNTL), 0, + ~UVD_VCPU_CNTL__BLK_RST_MASK); + + mdelay(10); + r = -1; + } + } + + if (r) { + dev_err(adev->dev, "VCN[%d] is not responding, giving up!!!\n", i); + return r; + } + + /* enable master interrupt */ + WREG32_P(SOC15_REG_OFFSET(VCN, i, regUVD_MASTINT_EN), + UVD_MASTINT_EN__VCPU_EN_MASK, + ~UVD_MASTINT_EN__VCPU_EN_MASK); + + /* clear the busy bit of VCN_STATUS */ + WREG32_P(SOC15_REG_OFFSET(VCN, i, regUVD_STATUS), 0, + ~(2 << UVD_STATUS__VCPU_REPORT__SHIFT)); + + ring = &adev->vcn.inst[i].ring_enc[0]; + WREG32_SOC15(VCN, i, regVCN_RB1_DB_CTRL, + ring->doorbell_index << VCN_RB1_DB_CTRL__OFFSET__SHIFT | + VCN_RB1_DB_CTRL__EN_MASK); + + WREG32_SOC15(VCN, i, regUVD_RB_BASE_LO, ring->gpu_addr); + WREG32_SOC15(VCN, i, regUVD_RB_BASE_HI, upper_32_bits(ring->gpu_addr)); + WREG32_SOC15(VCN, i, regUVD_RB_SIZE, ring->ring_size / 4); + + tmp = RREG32_SOC15(VCN, i, regVCN_RB_ENABLE); + tmp &= ~(VCN_RB_ENABLE__RB1_EN_MASK); + WREG32_SOC15(VCN, i, regVCN_RB_ENABLE, tmp); + fw_shared->sq.queue_mode |= FW_QUEUE_RING_RESET; + WREG32_SOC15(VCN, i, regUVD_RB_RPTR, 0); + WREG32_SOC15(VCN, i, regUVD_RB_WPTR, 0); + + tmp = RREG32_SOC15(VCN, i, regUVD_RB_RPTR); + WREG32_SOC15(VCN, i, regUVD_RB_WPTR, tmp); + ring->wptr = RREG32_SOC15(VCN, i, regUVD_RB_WPTR); + + tmp = RREG32_SOC15(VCN, i, regVCN_RB_ENABLE); + tmp |= VCN_RB_ENABLE__RB1_EN_MASK; + WREG32_SOC15(VCN, i, regVCN_RB_ENABLE, tmp); + fw_shared->sq.queue_mode &= ~(FW_QUEUE_RING_RESET | FW_QUEUE_DPG_HOLD_OFF); + } + + return 0; +} + +static int vcn_v4_0_start_sriov(struct amdgpu_device *adev) +{ + int i; + struct amdgpu_ring *ring_enc; + uint64_t cache_addr; + uint64_t rb_enc_addr; + uint64_t ctx_addr; + uint32_t param, resp, expected; + uint32_t offset, cache_size; + uint32_t tmp, timeout; + + struct amdgpu_mm_table *table = &adev->virt.mm_table; + uint32_t *table_loc; + uint32_t table_size; + uint32_t size, size_dw; + uint32_t init_status; + uint32_t enabled_vcn; + + struct mmsch_v4_0_cmd_direct_write + direct_wt = { {0} }; + struct mmsch_v4_0_cmd_direct_read_modify_write + direct_rd_mod_wt = { {0} }; + struct mmsch_v4_0_cmd_end end = { {0} }; + struct mmsch_v4_0_init_header header; + + volatile struct amdgpu_vcn4_fw_shared *fw_shared; + volatile struct amdgpu_fw_shared_rb_setup *rb_setup; + + direct_wt.cmd_header.command_type = + MMSCH_COMMAND__DIRECT_REG_WRITE; + direct_rd_mod_wt.cmd_header.command_type = + MMSCH_COMMAND__DIRECT_REG_READ_MODIFY_WRITE; + end.cmd_header.command_type = + MMSCH_COMMAND__END; + + header.version = MMSCH_VERSION; + header.total_size = sizeof(struct mmsch_v4_0_init_header) >> 2; + for (i = 0; i < AMDGPU_MAX_VCN_INSTANCES; i++) { + header.inst[i].init_status = 0; + header.inst[i].table_offset = 0; + header.inst[i].table_size = 0; + } + + table_loc = (uint32_t *)table->cpu_addr; + table_loc += header.total_size; + for (i = 0; i < adev->vcn.num_vcn_inst; i++) { + if (adev->vcn.harvest_config & (1 << i)) + continue; + + table_size = 0; + + MMSCH_V4_0_INSERT_DIRECT_RD_MOD_WT(SOC15_REG_OFFSET(VCN, i, + regUVD_STATUS), + ~UVD_STATUS__UVD_BUSY, UVD_STATUS__UVD_BUSY); + + cache_size = AMDGPU_GPU_PAGE_ALIGN(adev->vcn.fw->size + 4); + + if (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) { + MMSCH_V4_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCN, i, + regUVD_LMI_VCPU_CACHE_64BIT_BAR_LOW), + adev->firmware.ucode[AMDGPU_UCODE_ID_VCN + i].tmr_mc_addr_lo); + MMSCH_V4_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCN, i, + regUVD_LMI_VCPU_CACHE_64BIT_BAR_HIGH), + adev->firmware.ucode[AMDGPU_UCODE_ID_VCN + i].tmr_mc_addr_hi); + offset = 0; + MMSCH_V4_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCN, i, + regUVD_VCPU_CACHE_OFFSET0), + 0); + } else { + MMSCH_V4_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCN, i, + regUVD_LMI_VCPU_CACHE_64BIT_BAR_LOW), + lower_32_bits(adev->vcn.inst[i].gpu_addr)); + MMSCH_V4_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCN, i, + regUVD_LMI_VCPU_CACHE_64BIT_BAR_HIGH), + upper_32_bits(adev->vcn.inst[i].gpu_addr)); + offset = cache_size; + MMSCH_V4_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCN, i, + regUVD_VCPU_CACHE_OFFSET0), + AMDGPU_UVD_FIRMWARE_OFFSET >> 3); + } + + MMSCH_V4_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCN, i, + regUVD_VCPU_CACHE_SIZE0), + cache_size); + + cache_addr = adev->vcn.inst[i].gpu_addr + offset; + MMSCH_V4_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCN, i, + regUVD_LMI_VCPU_CACHE1_64BIT_BAR_LOW), + lower_32_bits(cache_addr)); + MMSCH_V4_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCN, i, + regUVD_LMI_VCPU_CACHE1_64BIT_BAR_HIGH), + upper_32_bits(cache_addr)); + MMSCH_V4_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCN, i, + regUVD_VCPU_CACHE_OFFSET1), + 0); + MMSCH_V4_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCN, i, + regUVD_VCPU_CACHE_SIZE1), + AMDGPU_VCN_STACK_SIZE); + + cache_addr = adev->vcn.inst[i].gpu_addr + offset + + AMDGPU_VCN_STACK_SIZE; + MMSCH_V4_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCN, i, + regUVD_LMI_VCPU_CACHE2_64BIT_BAR_LOW), + lower_32_bits(cache_addr)); + MMSCH_V4_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCN, i, + regUVD_LMI_VCPU_CACHE2_64BIT_BAR_HIGH), + upper_32_bits(cache_addr)); + MMSCH_V4_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCN, i, + regUVD_VCPU_CACHE_OFFSET2), + 0); + MMSCH_V4_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCN, i, + regUVD_VCPU_CACHE_SIZE2), + AMDGPU_VCN_CONTEXT_SIZE); + + fw_shared = adev->vcn.inst[i].fw_shared.cpu_addr; + rb_setup = &fw_shared->rb_setup; + + ring_enc = &adev->vcn.inst[i].ring_enc[0]; + ring_enc->wptr = 0; + rb_enc_addr = ring_enc->gpu_addr; + + rb_setup->is_rb_enabled_flags |= RB_ENABLED; + rb_setup->rb_addr_lo = lower_32_bits(rb_enc_addr); + rb_setup->rb_addr_hi = upper_32_bits(rb_enc_addr); + rb_setup->rb_size = ring_enc->ring_size / 4; + fw_shared->present_flag_0 |= cpu_to_le32(AMDGPU_VCN_VF_RB_SETUP_FLAG); + + MMSCH_V4_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCN, i, + regUVD_LMI_VCPU_NC0_64BIT_BAR_LOW), + lower_32_bits(adev->vcn.inst[i].fw_shared.gpu_addr)); + MMSCH_V4_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCN, i, + regUVD_LMI_VCPU_NC0_64BIT_BAR_HIGH), + upper_32_bits(adev->vcn.inst[i].fw_shared.gpu_addr)); + MMSCH_V4_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCN, i, + regUVD_VCPU_NONCACHE_SIZE0), + AMDGPU_GPU_PAGE_ALIGN(sizeof(struct amdgpu_vcn4_fw_shared))); + + /* add end packet */ + MMSCH_V4_0_INSERT_END(); + + /* refine header */ + header.inst[i].init_status = 0; + header.inst[i].table_offset = header.total_size; + header.inst[i].table_size = table_size; + header.total_size += table_size; + } + + /* Update init table header in memory */ + size = sizeof(struct mmsch_v4_0_init_header); + table_loc = (uint32_t *)table->cpu_addr; + memcpy((void *)table_loc, &header, size); + + /* message MMSCH (in VCN[0]) to initialize this client + * 1, write to mmsch_vf_ctx_addr_lo/hi register with GPU mc addr + * of memory descriptor location + */ + ctx_addr = table->gpu_addr; + WREG32_SOC15(VCN, 0, regMMSCH_VF_CTX_ADDR_LO, lower_32_bits(ctx_addr)); + WREG32_SOC15(VCN, 0, regMMSCH_VF_CTX_ADDR_HI, upper_32_bits(ctx_addr)); + + /* 2, update vmid of descriptor */ + tmp = RREG32_SOC15(VCN, 0, regMMSCH_VF_VMID); + tmp &= ~MMSCH_VF_VMID__VF_CTX_VMID_MASK; + /* use domain0 for MM scheduler */ + tmp |= (0 << MMSCH_VF_VMID__VF_CTX_VMID__SHIFT); + WREG32_SOC15(VCN, 0, regMMSCH_VF_VMID, tmp); + + /* 3, notify mmsch about the size of this descriptor */ + size = header.total_size; + WREG32_SOC15(VCN, 0, regMMSCH_VF_CTX_SIZE, size); + + /* 4, set resp to zero */ + WREG32_SOC15(VCN, 0, regMMSCH_VF_MAILBOX_RESP, 0); + + /* 5, kick off the initialization and wait until + * MMSCH_VF_MAILBOX_RESP becomes non-zero + */ + param = 0x00000001; + WREG32_SOC15(VCN, 0, regMMSCH_VF_MAILBOX_HOST, param); + tmp = 0; + timeout = 1000; + resp = 0; + expected = MMSCH_VF_MAILBOX_RESP__OK; + while (resp != expected) { + resp = RREG32_SOC15(VCN, 0, regMMSCH_VF_MAILBOX_RESP); + if (resp != 0) + break; + + udelay(10); + tmp = tmp + 10; + if (tmp >= timeout) { + DRM_ERROR("failed to init MMSCH. TIME-OUT after %d usec"\ + " waiting for regMMSCH_VF_MAILBOX_RESP "\ + "(expected=0x%08x, readback=0x%08x)\n", + tmp, expected, resp); + return -EBUSY; + } + } + enabled_vcn = amdgpu_vcn_is_disabled_vcn(adev, VCN_DECODE_RING, 0) ? 1 : 0; + init_status = ((struct mmsch_v4_0_init_header *)(table_loc))->inst[enabled_vcn].init_status; + if (resp != expected && resp != MMSCH_VF_MAILBOX_RESP__INCOMPLETE + && init_status != MMSCH_VF_ENGINE_STATUS__PASS) + DRM_ERROR("MMSCH init status is incorrect! readback=0x%08x, header init "\ + "status for VCN%x: 0x%x\n", resp, enabled_vcn, init_status); + + return 0; +} + +/** + * vcn_v4_0_stop_dpg_mode - VCN stop with dpg mode + * + * @adev: amdgpu_device pointer + * @inst_idx: instance number index + * + * Stop VCN block with dpg mode + */ +static void vcn_v4_0_stop_dpg_mode(struct amdgpu_device *adev, int inst_idx) +{ + uint32_t tmp; + + /* Wait for power status to be 1 */ + SOC15_WAIT_ON_RREG(VCN, inst_idx, regUVD_POWER_STATUS, 1, + UVD_POWER_STATUS__UVD_POWER_STATUS_MASK); + + /* wait for read ptr to be equal to write ptr */ + tmp = RREG32_SOC15(VCN, inst_idx, regUVD_RB_WPTR); + SOC15_WAIT_ON_RREG(VCN, inst_idx, regUVD_RB_RPTR, tmp, 0xFFFFFFFF); + + SOC15_WAIT_ON_RREG(VCN, inst_idx, regUVD_POWER_STATUS, 1, + UVD_POWER_STATUS__UVD_POWER_STATUS_MASK); + + /* disable dynamic power gating mode */ + WREG32_P(SOC15_REG_OFFSET(VCN, inst_idx, regUVD_POWER_STATUS), 0, + ~UVD_POWER_STATUS__UVD_PG_MODE_MASK); +} + +/** + * vcn_v4_0_stop - VCN stop + * + * @adev: amdgpu_device pointer + * + * Stop VCN block + */ +static int vcn_v4_0_stop(struct amdgpu_device *adev) +{ + volatile struct amdgpu_vcn4_fw_shared *fw_shared; + uint32_t tmp; + int i, r = 0; + + for (i = 0; i < adev->vcn.num_vcn_inst; ++i) { + fw_shared = adev->vcn.inst[i].fw_shared.cpu_addr; + fw_shared->sq.queue_mode |= FW_QUEUE_DPG_HOLD_OFF; + + if (adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG) { + vcn_v4_0_stop_dpg_mode(adev, i); + continue; + } + + /* wait for vcn idle */ + r = SOC15_WAIT_ON_RREG(VCN, i, regUVD_STATUS, UVD_STATUS__IDLE, 0x7); + if (r) + return r; + + tmp = UVD_LMI_STATUS__VCPU_LMI_WRITE_CLEAN_MASK | + UVD_LMI_STATUS__READ_CLEAN_MASK | + UVD_LMI_STATUS__WRITE_CLEAN_MASK | + UVD_LMI_STATUS__WRITE_CLEAN_RAW_MASK; + r = SOC15_WAIT_ON_RREG(VCN, i, regUVD_LMI_STATUS, tmp, tmp); + if (r) + return r; + + /* disable LMI UMC channel */ + tmp = RREG32_SOC15(VCN, i, regUVD_LMI_CTRL2); + tmp |= UVD_LMI_CTRL2__STALL_ARB_UMC_MASK; + WREG32_SOC15(VCN, i, regUVD_LMI_CTRL2, tmp); + tmp = UVD_LMI_STATUS__UMC_READ_CLEAN_RAW_MASK | + UVD_LMI_STATUS__UMC_WRITE_CLEAN_RAW_MASK; + r = SOC15_WAIT_ON_RREG(VCN, i, regUVD_LMI_STATUS, tmp, tmp); + if (r) + return r; + + /* block VCPU register access */ + WREG32_P(SOC15_REG_OFFSET(VCN, i, regUVD_RB_ARB_CTRL), + UVD_RB_ARB_CTRL__VCPU_DIS_MASK, + ~UVD_RB_ARB_CTRL__VCPU_DIS_MASK); + + /* reset VCPU */ + WREG32_P(SOC15_REG_OFFSET(VCN, i, regUVD_VCPU_CNTL), + UVD_VCPU_CNTL__BLK_RST_MASK, + ~UVD_VCPU_CNTL__BLK_RST_MASK); + + /* disable VCPU clock */ + WREG32_P(SOC15_REG_OFFSET(VCN, i, regUVD_VCPU_CNTL), 0, + ~(UVD_VCPU_CNTL__CLK_EN_MASK)); + + /* apply soft reset */ + tmp = RREG32_SOC15(VCN, i, regUVD_SOFT_RESET); + tmp |= UVD_SOFT_RESET__LMI_UMC_SOFT_RESET_MASK; + WREG32_SOC15(VCN, i, regUVD_SOFT_RESET, tmp); + tmp = RREG32_SOC15(VCN, i, regUVD_SOFT_RESET); + tmp |= UVD_SOFT_RESET__LMI_SOFT_RESET_MASK; + WREG32_SOC15(VCN, i, regUVD_SOFT_RESET, tmp); + + /* clear status */ + WREG32_SOC15(VCN, i, regUVD_STATUS, 0); + + /* apply HW clock gating */ + vcn_v4_0_enable_clock_gating(adev, i); + + /* enable VCN power gating */ + vcn_v4_0_enable_static_power_gating(adev, i); + } + + if (adev->pm.dpm_enabled) + amdgpu_dpm_enable_uvd(adev, false); + + return 0; +} + +/** + * vcn_v4_0_pause_dpg_mode - VCN pause with dpg mode + * + * @adev: amdgpu_device pointer + * @inst_idx: instance number index + * @new_state: pause state + * + * Pause dpg mode for VCN block + */ +static int vcn_v4_0_pause_dpg_mode(struct amdgpu_device *adev, int inst_idx, + struct dpg_pause_state *new_state) +{ + uint32_t reg_data = 0; + int ret_code; + + /* pause/unpause if state is changed */ + if (adev->vcn.inst[inst_idx].pause_state.fw_based != new_state->fw_based) { + DRM_DEV_DEBUG(adev->dev, "dpg pause state changed %d -> %d", + adev->vcn.inst[inst_idx].pause_state.fw_based, new_state->fw_based); + reg_data = RREG32_SOC15(VCN, inst_idx, regUVD_DPG_PAUSE) & + (~UVD_DPG_PAUSE__NJ_PAUSE_DPG_ACK_MASK); + + if (new_state->fw_based == VCN_DPG_STATE__PAUSE) { + ret_code = SOC15_WAIT_ON_RREG(VCN, inst_idx, regUVD_POWER_STATUS, 0x1, + UVD_POWER_STATUS__UVD_POWER_STATUS_MASK); + + if (!ret_code) { + /* pause DPG */ + reg_data |= UVD_DPG_PAUSE__NJ_PAUSE_DPG_REQ_MASK; + WREG32_SOC15(VCN, inst_idx, regUVD_DPG_PAUSE, reg_data); + + /* wait for ACK */ + SOC15_WAIT_ON_RREG(VCN, inst_idx, regUVD_DPG_PAUSE, + UVD_DPG_PAUSE__NJ_PAUSE_DPG_ACK_MASK, + UVD_DPG_PAUSE__NJ_PAUSE_DPG_ACK_MASK); + + SOC15_WAIT_ON_RREG(VCN, inst_idx, regUVD_POWER_STATUS, + UVD_PGFSM_CONFIG__UVDM_UVDU_PWR_ON, UVD_POWER_STATUS__UVD_POWER_STATUS_MASK); + } + } else { + /* unpause dpg, no need to wait */ + reg_data &= ~UVD_DPG_PAUSE__NJ_PAUSE_DPG_REQ_MASK; + WREG32_SOC15(VCN, inst_idx, regUVD_DPG_PAUSE, reg_data); + } + adev->vcn.inst[inst_idx].pause_state.fw_based = new_state->fw_based; + } + + return 0; +} + +/** + * vcn_v4_0_unified_ring_get_rptr - get unified read pointer + * + * @ring: amdgpu_ring pointer + * + * Returns the current hardware unified read pointer + */ +static uint64_t vcn_v4_0_unified_ring_get_rptr(struct amdgpu_ring *ring) +{ + struct amdgpu_device *adev = ring->adev; + + if (ring != &adev->vcn.inst[ring->me].ring_enc[0]) + DRM_ERROR("wrong ring id is identified in %s", __func__); + + return RREG32_SOC15(VCN, ring->me, regUVD_RB_RPTR); +} + +/** + * vcn_v4_0_unified_ring_get_wptr - get unified write pointer + * + * @ring: amdgpu_ring pointer + * + * Returns the current hardware unified write pointer + */ +static uint64_t vcn_v4_0_unified_ring_get_wptr(struct amdgpu_ring *ring) +{ + struct amdgpu_device *adev = ring->adev; + + if (ring != &adev->vcn.inst[ring->me].ring_enc[0]) + DRM_ERROR("wrong ring id is identified in %s", __func__); + + if (ring->use_doorbell) + return *ring->wptr_cpu_addr; + else + return RREG32_SOC15(VCN, ring->me, regUVD_RB_WPTR); +} + +/** + * vcn_v4_0_unified_ring_set_wptr - set enc write pointer + * + * @ring: amdgpu_ring pointer + * + * Commits the enc write pointer to the hardware + */ +static void vcn_v4_0_unified_ring_set_wptr(struct amdgpu_ring *ring) +{ + struct amdgpu_device *adev = ring->adev; + + if (ring != &adev->vcn.inst[ring->me].ring_enc[0]) + DRM_ERROR("wrong ring id is identified in %s", __func__); + + if (ring->use_doorbell) { + *ring->wptr_cpu_addr = lower_32_bits(ring->wptr); + WDOORBELL32(ring->doorbell_index, lower_32_bits(ring->wptr)); + } else { + WREG32_SOC15(VCN, ring->me, regUVD_RB_WPTR, lower_32_bits(ring->wptr)); + } +} + +static int vcn_v4_0_limit_sched(struct amdgpu_cs_parser *p, + struct amdgpu_job *job) +{ + struct drm_gpu_scheduler **scheds; + + /* The create msg must be in the first IB submitted */ + if (atomic_read(&job->base.entity->fence_seq)) + return -EINVAL; + + scheds = p->adev->gpu_sched[AMDGPU_HW_IP_VCN_ENC] + [AMDGPU_RING_PRIO_0].sched; + drm_sched_entity_modify_sched(job->base.entity, scheds, 1); + return 0; +} + +static int vcn_v4_0_dec_msg(struct amdgpu_cs_parser *p, struct amdgpu_job *job, + uint64_t addr) +{ + struct ttm_operation_ctx ctx = { false, false }; + struct amdgpu_bo_va_mapping *map; + uint32_t *msg, num_buffers; + struct amdgpu_bo *bo; + uint64_t start, end; + unsigned int i; + void *ptr; + int r; + + addr &= AMDGPU_GMC_HOLE_MASK; + r = amdgpu_cs_find_mapping(p, addr, &bo, &map); + if (r) { + DRM_ERROR("Can't find BO for addr 0x%08llx\n", addr); + return r; + } + + start = map->start * AMDGPU_GPU_PAGE_SIZE; + end = (map->last + 1) * AMDGPU_GPU_PAGE_SIZE; + if (addr & 0x7) { + DRM_ERROR("VCN messages must be 8 byte aligned!\n"); + return -EINVAL; + } + + bo->flags |= AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED; + amdgpu_bo_placement_from_domain(bo, bo->allowed_domains); + r = ttm_bo_validate(&bo->tbo, &bo->placement, &ctx); + if (r) { + DRM_ERROR("Failed validating the VCN message BO (%d)!\n", r); + return r; + } + + r = amdgpu_bo_kmap(bo, &ptr); + if (r) { + DRM_ERROR("Failed mapping the VCN message (%d)!\n", r); + return r; + } + + msg = ptr + addr - start; + + /* Check length */ + if (msg[1] > end - addr) { + r = -EINVAL; + goto out; + } + + if (msg[3] != RDECODE_MSG_CREATE) + goto out; + + num_buffers = msg[2]; + for (i = 0, msg = &msg[6]; i < num_buffers; ++i, msg += 4) { + uint32_t offset, size, *create; + + if (msg[0] != RDECODE_MESSAGE_CREATE) + continue; + + offset = msg[1]; + size = msg[2]; + + if (offset + size > end) { + r = -EINVAL; + goto out; + } + + create = ptr + addr + offset - start; + + /* H246, HEVC and VP9 can run on any instance */ + if (create[0] == 0x7 || create[0] == 0x10 || create[0] == 0x11) + continue; + + r = vcn_v4_0_limit_sched(p, job); + if (r) + goto out; + } + +out: + amdgpu_bo_kunmap(bo); + return r; +} + +#define RADEON_VCN_ENGINE_TYPE_DECODE (0x00000003) + +static int vcn_v4_0_ring_patch_cs_in_place(struct amdgpu_cs_parser *p, + struct amdgpu_job *job, + struct amdgpu_ib *ib) +{ + struct amdgpu_ring *ring = amdgpu_job_ring(job); + struct amdgpu_vcn_decode_buffer *decode_buffer; + uint64_t addr; + uint32_t val; + + /* The first instance can decode anything */ + if (!ring->me) + return 0; + + /* unified queue ib header has 8 double words. */ + if (ib->length_dw < 8) + return 0; + + val = amdgpu_ib_get_value(ib, 6); //RADEON_VCN_ENGINE_TYPE + if (val != RADEON_VCN_ENGINE_TYPE_DECODE) + return 0; + + decode_buffer = (struct amdgpu_vcn_decode_buffer *)&ib->ptr[10]; + + if (!(decode_buffer->valid_buf_flag & 0x1)) + return 0; + + addr = ((u64)decode_buffer->msg_buffer_address_hi) << 32 | + decode_buffer->msg_buffer_address_lo; + return vcn_v4_0_dec_msg(p, job, addr); +} + +static const struct amdgpu_ring_funcs vcn_v4_0_unified_ring_vm_funcs = { + .type = AMDGPU_RING_TYPE_VCN_ENC, + .align_mask = 0x3f, + .nop = VCN_ENC_CMD_NO_OP, + .vmhub = AMDGPU_MMHUB_0, + .get_rptr = vcn_v4_0_unified_ring_get_rptr, + .get_wptr = vcn_v4_0_unified_ring_get_wptr, + .set_wptr = vcn_v4_0_unified_ring_set_wptr, + .patch_cs_in_place = vcn_v4_0_ring_patch_cs_in_place, + .emit_frame_size = + SOC15_FLUSH_GPU_TLB_NUM_WREG * 3 + + SOC15_FLUSH_GPU_TLB_NUM_REG_WAIT * 4 + + 4 + /* vcn_v2_0_enc_ring_emit_vm_flush */ + 5 + 5 + /* vcn_v2_0_enc_ring_emit_fence x2 vm fence */ + 1, /* vcn_v2_0_enc_ring_insert_end */ + .emit_ib_size = 5, /* vcn_v2_0_enc_ring_emit_ib */ + .emit_ib = vcn_v2_0_enc_ring_emit_ib, + .emit_fence = vcn_v2_0_enc_ring_emit_fence, + .emit_vm_flush = vcn_v2_0_enc_ring_emit_vm_flush, + .test_ring = amdgpu_vcn_enc_ring_test_ring, + .test_ib = amdgpu_vcn_unified_ring_test_ib, + .insert_nop = amdgpu_ring_insert_nop, + .insert_end = vcn_v2_0_enc_ring_insert_end, + .pad_ib = amdgpu_ring_generic_pad_ib, + .begin_use = amdgpu_vcn_ring_begin_use, + .end_use = amdgpu_vcn_ring_end_use, + .emit_wreg = vcn_v2_0_enc_ring_emit_wreg, + .emit_reg_wait = vcn_v2_0_enc_ring_emit_reg_wait, + .emit_reg_write_reg_wait = amdgpu_ring_emit_reg_write_reg_wait_helper, +}; + +/** + * vcn_v4_0_set_unified_ring_funcs - set unified ring functions + * + * @adev: amdgpu_device pointer + * + * Set unified ring functions + */ +static void vcn_v4_0_set_unified_ring_funcs(struct amdgpu_device *adev) +{ + int i; + + for (i = 0; i < adev->vcn.num_vcn_inst; ++i) { + if (adev->vcn.harvest_config & (1 << i)) + continue; + + adev->vcn.inst[i].ring_enc[0].funcs = &vcn_v4_0_unified_ring_vm_funcs; + adev->vcn.inst[i].ring_enc[0].me = i; + + DRM_INFO("VCN(%d) encode/decode are enabled in VM mode\n", i); + } +} + +/** + * vcn_v4_0_is_idle - check VCN block is idle + * + * @handle: amdgpu_device pointer + * + * Check whether VCN block is idle + */ +static bool vcn_v4_0_is_idle(void *handle) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + int i, ret = 1; + + for (i = 0; i < adev->vcn.num_vcn_inst; ++i) { + if (adev->vcn.harvest_config & (1 << i)) + continue; + + ret &= (RREG32_SOC15(VCN, i, regUVD_STATUS) == UVD_STATUS__IDLE); + } + + return ret; +} + +/** + * vcn_v4_0_wait_for_idle - wait for VCN block idle + * + * @handle: amdgpu_device pointer + * + * Wait for VCN block idle + */ +static int vcn_v4_0_wait_for_idle(void *handle) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + int i, ret = 0; + + for (i = 0; i < adev->vcn.num_vcn_inst; ++i) { + if (adev->vcn.harvest_config & (1 << i)) + continue; + + ret = SOC15_WAIT_ON_RREG(VCN, i, regUVD_STATUS, UVD_STATUS__IDLE, + UVD_STATUS__IDLE); + if (ret) + return ret; + } + + return ret; +} + +/** + * vcn_v4_0_set_clockgating_state - set VCN block clockgating state + * + * @handle: amdgpu_device pointer + * @state: clock gating state + * + * Set VCN block clockgating state + */ +static int vcn_v4_0_set_clockgating_state(void *handle, enum amd_clockgating_state state) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + bool enable = (state == AMD_CG_STATE_GATE) ? true : false; + int i; + + for (i = 0; i < adev->vcn.num_vcn_inst; ++i) { + if (adev->vcn.harvest_config & (1 << i)) + continue; + + if (enable) { + if (RREG32_SOC15(VCN, i, regUVD_STATUS) != UVD_STATUS__IDLE) + return -EBUSY; + vcn_v4_0_enable_clock_gating(adev, i); + } else { + vcn_v4_0_disable_clock_gating(adev, i); + } + } + + return 0; +} + +/** + * vcn_v4_0_set_powergating_state - set VCN block powergating state + * + * @handle: amdgpu_device pointer + * @state: power gating state + * + * Set VCN block powergating state + */ +static int vcn_v4_0_set_powergating_state(void *handle, enum amd_powergating_state state) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + int ret; + + /* for SRIOV, guest should not control VCN Power-gating + * MMSCH FW should control Power-gating and clock-gating + * guest should avoid touching CGC and PG + */ + if (amdgpu_sriov_vf(adev)) { + adev->vcn.cur_state = AMD_PG_STATE_UNGATE; + return 0; + } + + if(state == adev->vcn.cur_state) + return 0; + + if (state == AMD_PG_STATE_GATE) + ret = vcn_v4_0_stop(adev); + else + ret = vcn_v4_0_start(adev); + + if(!ret) + adev->vcn.cur_state = state; + + return ret; +} + +/** + * vcn_v4_0_set_interrupt_state - set VCN block interrupt state + * + * @adev: amdgpu_device pointer + * @source: interrupt sources + * @type: interrupt types + * @state: interrupt states + * + * Set VCN block interrupt state + */ +static int vcn_v4_0_set_interrupt_state(struct amdgpu_device *adev, struct amdgpu_irq_src *source, + unsigned type, enum amdgpu_interrupt_state state) +{ + return 0; +} + +/** + * vcn_v4_0_process_interrupt - process VCN block interrupt + * + * @adev: amdgpu_device pointer + * @source: interrupt sources + * @entry: interrupt entry from clients and sources + * + * Process VCN block interrupt + */ +static int vcn_v4_0_process_interrupt(struct amdgpu_device *adev, struct amdgpu_irq_src *source, + struct amdgpu_iv_entry *entry) +{ + uint32_t ip_instance; + + switch (entry->client_id) { + case SOC15_IH_CLIENTID_VCN: + ip_instance = 0; + break; + case SOC15_IH_CLIENTID_VCN1: + ip_instance = 1; + break; + default: + DRM_ERROR("Unhandled client id: %d\n", entry->client_id); + return 0; + } + + DRM_DEBUG("IH: VCN TRAP\n"); + + switch (entry->src_id) { + case VCN_4_0__SRCID__UVD_ENC_GENERAL_PURPOSE: + amdgpu_fence_process(&adev->vcn.inst[ip_instance].ring_enc[0]); + break; + default: + DRM_ERROR("Unhandled interrupt: %d %d\n", + entry->src_id, entry->src_data[0]); + break; + } + + return 0; +} + +static const struct amdgpu_irq_src_funcs vcn_v4_0_irq_funcs = { + .set = vcn_v4_0_set_interrupt_state, + .process = vcn_v4_0_process_interrupt, +}; + +/** + * vcn_v4_0_set_irq_funcs - set VCN block interrupt irq functions + * + * @adev: amdgpu_device pointer + * + * Set VCN block interrupt irq functions + */ +static void vcn_v4_0_set_irq_funcs(struct amdgpu_device *adev) +{ + int i; + + for (i = 0; i < adev->vcn.num_vcn_inst; ++i) { + if (adev->vcn.harvest_config & (1 << i)) + continue; + + adev->vcn.inst[i].irq.num_types = adev->vcn.num_enc_rings + 1; + adev->vcn.inst[i].irq.funcs = &vcn_v4_0_irq_funcs; + } +} + +static const struct amd_ip_funcs vcn_v4_0_ip_funcs = { + .name = "vcn_v4_0", + .early_init = vcn_v4_0_early_init, + .late_init = NULL, + .sw_init = vcn_v4_0_sw_init, + .sw_fini = vcn_v4_0_sw_fini, + .hw_init = vcn_v4_0_hw_init, + .hw_fini = vcn_v4_0_hw_fini, + .suspend = vcn_v4_0_suspend, + .resume = vcn_v4_0_resume, + .is_idle = vcn_v4_0_is_idle, + .wait_for_idle = vcn_v4_0_wait_for_idle, + .check_soft_reset = NULL, + .pre_soft_reset = NULL, + .soft_reset = NULL, + .post_soft_reset = NULL, + .set_clockgating_state = vcn_v4_0_set_clockgating_state, + .set_powergating_state = vcn_v4_0_set_powergating_state, +}; + +const struct amdgpu_ip_block_version vcn_v4_0_ip_block = +{ + .type = AMD_IP_BLOCK_TYPE_VCN, + .major = 4, + .minor = 0, + .rev = 0, + .funcs = &vcn_v4_0_ip_funcs, +}; diff --git a/drivers/gpu/drm/amd/amdgpu/vcn_v4_0.h b/drivers/gpu/drm/amd/amdgpu/vcn_v4_0.h new file mode 100644 index 000000000000..7c5c9d91bb52 --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/vcn_v4_0.h @@ -0,0 +1,29 @@ +/* + * Copyright 2021 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#ifndef __VCN_V4_0_H__ +#define __VCN_V4_0_H__ + +extern const struct amdgpu_ip_block_version vcn_v4_0_ip_block; + +#endif /* __VCN_V4_0_H__ */ diff --git a/drivers/gpu/drm/amd/amdgpu/vega10_ih.c b/drivers/gpu/drm/amd/amdgpu/vega10_ih.c index 3070466f54e1..1e83db0c5438 100644 --- a/drivers/gpu/drm/amd/amdgpu/vega10_ih.c +++ b/drivers/gpu/drm/amd/amdgpu/vega10_ih.c @@ -289,6 +289,10 @@ static int vega10_ih_irq_init(struct amdgpu_device *adev) } } + if (!amdgpu_sriov_vf(adev)) + adev->nbio.funcs->ih_doorbell_range(adev, adev->irq.ih.use_doorbell, + adev->irq.ih.doorbell_index); + pci_set_master(adev->pdev); /* enable interrupts */ @@ -334,9 +338,11 @@ static u32 vega10_ih_get_wptr(struct amdgpu_device *adev, u32 wptr, tmp; struct amdgpu_ih_regs *ih_regs; - if (ih == &adev->irq.ih) { + if (ih == &adev->irq.ih || ih == &adev->irq.ih_soft) { /* Only ring0 supports writeback. On other rings fall back * to register-based code with overflow checking below. + * ih_soft ring doesn't have any backing hardware registers, + * update wptr and return. */ wptr = le32_to_cpu(*ih->wptr_cpu); @@ -409,6 +415,9 @@ static void vega10_ih_set_rptr(struct amdgpu_device *adev, { struct amdgpu_ih_regs *ih_regs; + if (ih == &adev->irq.ih_soft) + return; + if (ih->use_doorbell) { /* XXX check if swapping is necessary on BE */ *ih->rptr_cpu = ih->rptr; @@ -521,14 +530,9 @@ static int vega10_ih_sw_fini(void *handle) static int vega10_ih_hw_init(void *handle) { - int r; struct amdgpu_device *adev = (struct amdgpu_device *)handle; - r = vega10_ih_irq_init(adev); - if (r) - return r; - - return 0; + return vega10_ih_irq_init(adev); } static int vega10_ih_hw_fini(void *handle) diff --git a/drivers/gpu/drm/amd/amdgpu/vega20_ih.c b/drivers/gpu/drm/amd/amdgpu/vega20_ih.c index 3b4eb8285943..59dfca093155 100644 --- a/drivers/gpu/drm/amd/amdgpu/vega20_ih.c +++ b/drivers/gpu/drm/amd/amdgpu/vega20_ih.c @@ -340,6 +340,10 @@ static int vega20_ih_irq_init(struct amdgpu_device *adev) } } + if (!amdgpu_sriov_vf(adev)) + adev->nbio.funcs->ih_doorbell_range(adev, adev->irq.ih.use_doorbell, + adev->irq.ih.doorbell_index); + pci_set_master(adev->pdev); /* enable interrupts */ @@ -385,9 +389,11 @@ static u32 vega20_ih_get_wptr(struct amdgpu_device *adev, u32 wptr, tmp; struct amdgpu_ih_regs *ih_regs; - if (ih == &adev->irq.ih) { + if (ih == &adev->irq.ih || ih == &adev->irq.ih_soft) { /* Only ring0 supports writeback. On other rings fall back * to register-based code with overflow checking below. + * ih_soft ring doesn't have any backing hardware registers, + * update wptr and return. */ wptr = le32_to_cpu(*ih->wptr_cpu); @@ -461,6 +467,9 @@ static void vega20_ih_set_rptr(struct amdgpu_device *adev, { struct amdgpu_ih_regs *ih_regs; + if (ih == &adev->irq.ih_soft) + return; + if (ih->use_doorbell) { /* XXX check if swapping is necessary on BE */ *ih->rptr_cpu = ih->rptr; diff --git a/drivers/gpu/drm/amd/amdgpu/vi.c b/drivers/gpu/drm/amd/amdgpu/vi.c index fe9a7cc8d9eb..f6ffd7c96ff9 100644 --- a/drivers/gpu/drm/amd/amdgpu/vi.c +++ b/drivers/gpu/drm/amd/amdgpu/vi.c @@ -81,6 +81,10 @@ #include "mxgpu_vi.h" #include "amdgpu_dm.h" +#if IS_ENABLED(CONFIG_X86) +#include <asm/intel-family.h> +#endif + #define ixPCIE_LC_L1_PM_SUBSTATE 0x100100C6 #define PCIE_LC_L1_PM_SUBSTATE__LC_L1_SUBSTATES_OVERRIDE_EN_MASK 0x00000001L #define PCIE_LC_L1_PM_SUBSTATE__LC_PCI_PM_L1_2_OVERRIDE_MASK 0x00000002L @@ -956,6 +960,10 @@ static int vi_asic_reset(struct amdgpu_device *adev) { int r; + /* APUs don't have full asic reset */ + if (adev->flags & AMD_IS_APU) + return 0; + if (vi_asic_reset_method(adev) == AMD_RESET_METHOD_BACO) { dev_info(adev->dev, "BACO reset\n"); r = amdgpu_dpm_baco_reset(adev); @@ -1130,13 +1138,24 @@ static void vi_enable_aspm(struct amdgpu_device *adev) WREG32_PCIE(ixPCIE_LC_CNTL, data); } +static bool aspm_support_quirk_check(void) +{ +#if IS_ENABLED(CONFIG_X86) + struct cpuinfo_x86 *c = &cpu_data(0); + + return !(c->x86 == 6 && c->x86_model == INTEL_FAM6_ALDERLAKE); +#else + return true; +#endif +} + static void vi_program_aspm(struct amdgpu_device *adev) { u32 data, data1, orig; bool bL1SS = false; bool bClkReqSupport = true; - if (!amdgpu_aspm) + if (!amdgpu_device_should_use_aspm(adev) || !aspm_support_quirk_check()) return; if (adev->flags & AMD_IS_APU || @@ -2029,7 +2048,7 @@ static int vi_common_set_powergating_state(void *handle, return 0; } -static void vi_common_get_clockgating_state(void *handle, u32 *flags) +static void vi_common_get_clockgating_state(void *handle, u64 *flags) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; int data; |