diff options
Diffstat (limited to 'drivers/gpu/drm/amd')
390 files changed, 17919 insertions, 25809 deletions
diff --git a/drivers/gpu/drm/amd/amdgpu/Makefile b/drivers/gpu/drm/amd/amdgpu/Makefile index 138cb787d27e..466da5954a68 100644 --- a/drivers/gpu/drm/amd/amdgpu/Makefile +++ b/drivers/gpu/drm/amd/amdgpu/Makefile @@ -53,11 +53,11 @@ amdgpu-y += amdgpu_device.o amdgpu_kms.o \ amdgpu_ucode.o amdgpu_bo_list.o amdgpu_ctx.o amdgpu_sync.o \ amdgpu_gtt_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_xgmi.o + amdgpu_gmc.o amdgpu_xgmi.o amdgpu_csa.o # add asic specific block amdgpu-$(CONFIG_DRM_AMDGPU_CIK)+= cik.o cik_ih.o kv_smc.o kv_dpm.o \ - ci_smc.o ci_dpm.o dce_v8_0.o gfx_v7_0.o cik_sdma.o uvd_v4_2.o vce_v2_0.o + dce_v8_0.o gfx_v7_0.o cik_sdma.o uvd_v4_2.o vce_v2_0.o amdgpu-$(CONFIG_DRM_AMDGPU_SI)+= si.o gmc_v6_0.o gfx_v6_0.o si_ih.o si_dma.o dce_v6_0.o si_dpm.o si_smc.o @@ -105,6 +105,7 @@ amdgpu-y += \ # add GFX block amdgpu-y += \ amdgpu_gfx.o \ + amdgpu_rlc.o \ gfx_v8_0.o \ gfx_v9_0.o diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h b/drivers/gpu/drm/amd/amdgpu/amdgpu.h index b0fc116296cb..8d0d7f3dd5fb 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu.h @@ -75,11 +75,14 @@ #include "amdgpu_sdma.h" #include "amdgpu_dm.h" #include "amdgpu_virt.h" +#include "amdgpu_csa.h" #include "amdgpu_gart.h" #include "amdgpu_debugfs.h" #include "amdgpu_job.h" #include "amdgpu_bo_list.h" #include "amdgpu_gem.h" +#include "amdgpu_doorbell.h" +#include "amdgpu_amdkfd.h" #define MAX_GPU_INSTANCE 16 @@ -161,6 +164,7 @@ extern int amdgpu_si_support; extern int amdgpu_cik_support; #endif +#define AMDGPU_VM_MAX_NUM_CTX 4096 #define AMDGPU_SG_THRESHOLD (256*1024*1024) #define AMDGPU_DEFAULT_GTT_SIZE_MB 3072ULL /* 3GB by default */ #define AMDGPU_WAIT_IDLE_TIMEOUT_IN_MS 3000 @@ -360,123 +364,6 @@ int amdgpu_fence_slab_init(void); void amdgpu_fence_slab_fini(void); /* - * GPU doorbell structures, functions & helpers - */ -typedef enum _AMDGPU_DOORBELL_ASSIGNMENT -{ - AMDGPU_DOORBELL_KIQ = 0x000, - AMDGPU_DOORBELL_HIQ = 0x001, - AMDGPU_DOORBELL_DIQ = 0x002, - AMDGPU_DOORBELL_MEC_RING0 = 0x010, - AMDGPU_DOORBELL_MEC_RING1 = 0x011, - AMDGPU_DOORBELL_MEC_RING2 = 0x012, - AMDGPU_DOORBELL_MEC_RING3 = 0x013, - AMDGPU_DOORBELL_MEC_RING4 = 0x014, - AMDGPU_DOORBELL_MEC_RING5 = 0x015, - AMDGPU_DOORBELL_MEC_RING6 = 0x016, - AMDGPU_DOORBELL_MEC_RING7 = 0x017, - AMDGPU_DOORBELL_GFX_RING0 = 0x020, - AMDGPU_DOORBELL_sDMA_ENGINE0 = 0x1E0, - AMDGPU_DOORBELL_sDMA_ENGINE1 = 0x1E1, - AMDGPU_DOORBELL_IH = 0x1E8, - AMDGPU_DOORBELL_MAX_ASSIGNMENT = 0x3FF, - AMDGPU_DOORBELL_INVALID = 0xFFFF -} AMDGPU_DOORBELL_ASSIGNMENT; - -struct amdgpu_doorbell { - /* doorbell mmio */ - resource_size_t base; - resource_size_t size; - u32 __iomem *ptr; - u32 num_doorbells; /* Number of doorbells actually reserved for amdgpu. */ -}; - -/* - * 64bit doorbell, offset are in QWORD, occupy 2KB doorbell space - */ -typedef enum _AMDGPU_DOORBELL64_ASSIGNMENT -{ - /* - * All compute related doorbells: kiq, hiq, diq, traditional compute queue, user queue, should locate in - * a continues range so that programming CP_MEC_DOORBELL_RANGE_LOWER/UPPER can cover this range. - * Compute related doorbells are allocated from 0x00 to 0x8a - */ - - - /* kernel scheduling */ - AMDGPU_DOORBELL64_KIQ = 0x00, - - /* HSA interface queue and debug queue */ - AMDGPU_DOORBELL64_HIQ = 0x01, - AMDGPU_DOORBELL64_DIQ = 0x02, - - /* Compute engines */ - AMDGPU_DOORBELL64_MEC_RING0 = 0x03, - AMDGPU_DOORBELL64_MEC_RING1 = 0x04, - AMDGPU_DOORBELL64_MEC_RING2 = 0x05, - AMDGPU_DOORBELL64_MEC_RING3 = 0x06, - AMDGPU_DOORBELL64_MEC_RING4 = 0x07, - AMDGPU_DOORBELL64_MEC_RING5 = 0x08, - AMDGPU_DOORBELL64_MEC_RING6 = 0x09, - AMDGPU_DOORBELL64_MEC_RING7 = 0x0a, - - /* User queue doorbell range (128 doorbells) */ - AMDGPU_DOORBELL64_USERQUEUE_START = 0x0b, - AMDGPU_DOORBELL64_USERQUEUE_END = 0x8a, - - /* Graphics engine */ - AMDGPU_DOORBELL64_GFX_RING0 = 0x8b, - - /* - * Other graphics doorbells can be allocated here: from 0x8c to 0xdf - * Graphics voltage island aperture 1 - * default non-graphics QWORD index is 0xe0 - 0xFF inclusive - */ - - /* sDMA engines reserved from 0xe0 -oxef */ - AMDGPU_DOORBELL64_sDMA_ENGINE0 = 0xE0, - AMDGPU_DOORBELL64_sDMA_HI_PRI_ENGINE0 = 0xE1, - AMDGPU_DOORBELL64_sDMA_ENGINE1 = 0xE8, - AMDGPU_DOORBELL64_sDMA_HI_PRI_ENGINE1 = 0xE9, - - /* For vega10 sriov, the sdma doorbell must be fixed as follow - * to keep the same setting with host driver, or it will - * happen conflicts - */ - AMDGPU_VEGA10_DOORBELL64_sDMA_ENGINE0 = 0xF0, - AMDGPU_VEGA10_DOORBELL64_sDMA_HI_PRI_ENGINE0 = 0xF1, - AMDGPU_VEGA10_DOORBELL64_sDMA_ENGINE1 = 0xF2, - AMDGPU_VEGA10_DOORBELL64_sDMA_HI_PRI_ENGINE1 = 0xF3, - - /* Interrupt handler */ - AMDGPU_DOORBELL64_IH = 0xF4, /* For legacy interrupt ring buffer */ - AMDGPU_DOORBELL64_IH_RING1 = 0xF5, /* For page migration request log */ - AMDGPU_DOORBELL64_IH_RING2 = 0xF6, /* For page migration translation/invalidation log */ - - /* VCN engine use 32 bits doorbell */ - AMDGPU_DOORBELL64_VCN0_1 = 0xF8, /* lower 32 bits for VNC0 and upper 32 bits for VNC1 */ - AMDGPU_DOORBELL64_VCN2_3 = 0xF9, - AMDGPU_DOORBELL64_VCN4_5 = 0xFA, - AMDGPU_DOORBELL64_VCN6_7 = 0xFB, - - /* overlap the doorbell assignment with VCN as they are mutually exclusive - * VCE engine's doorbell is 32 bit and two VCE ring share one QWORD - */ - AMDGPU_DOORBELL64_UVD_RING0_1 = 0xF8, - AMDGPU_DOORBELL64_UVD_RING2_3 = 0xF9, - AMDGPU_DOORBELL64_UVD_RING4_5 = 0xFA, - AMDGPU_DOORBELL64_UVD_RING6_7 = 0xFB, - - AMDGPU_DOORBELL64_VCE_RING0_1 = 0xFC, - AMDGPU_DOORBELL64_VCE_RING2_3 = 0xFD, - AMDGPU_DOORBELL64_VCE_RING4_5 = 0xFE, - AMDGPU_DOORBELL64_VCE_RING6_7 = 0xFF, - - AMDGPU_DOORBELL64_MAX_ASSIGNMENT = 0xFF, - AMDGPU_DOORBELL64_INVALID = 0xFFFF -} AMDGPU_DOORBELL64_ASSIGNMENT; - -/* * IRQS. */ @@ -524,6 +411,8 @@ struct amdgpu_fpriv { struct amdgpu_ctx_mgr ctx_mgr; }; +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, struct amdgpu_ib *ib); void amdgpu_ib_free(struct amdgpu_device *adev, struct amdgpu_ib *ib, @@ -653,6 +542,13 @@ struct amdgpu_asic_funcs { struct amdgpu_ring *ring); /* check if the asic needs a full reset of if soft reset will work */ bool (*need_full_reset)(struct amdgpu_device *adev); + /* initialize doorbell layout for specific asic*/ + void (*init_doorbell_index)(struct amdgpu_device *adev); + /* PCIe bandwidth usage */ + void (*get_pcie_usage)(struct amdgpu_device *adev, uint64_t *count0, + uint64_t *count1); + /* do we need to reset the asic at init time (e.g., kexec) */ + bool (*need_reset_on_init)(struct amdgpu_device *adev); }; /* @@ -745,7 +641,7 @@ struct amdgpu_nbio_funcs { void (*hdp_flush)(struct amdgpu_device *adev, struct amdgpu_ring *ring); u32 (*get_memsize)(struct amdgpu_device *adev); void (*sdma_doorbell_range)(struct amdgpu_device *adev, int instance, - bool use_doorbell, int doorbell_index); + bool use_doorbell, int doorbell_index, int doorbell_size); void (*enable_doorbell_aperture)(struct amdgpu_device *adev, bool enable); void (*enable_doorbell_selfring_aperture)(struct amdgpu_device *adev, @@ -831,7 +727,6 @@ struct amdgpu_device { bool need_dma32; bool need_swiotlb; bool accel_working; - struct work_struct reset_work; struct notifier_block acpi_nb; struct amdgpu_i2c_chan *i2c_bus[AMDGPU_MAX_I2C_BUS]; struct amdgpu_debugfs debugfs[AMDGPU_DEBUGFS_MAX_COMPONENTS]; @@ -976,6 +871,9 @@ struct amdgpu_device { /* GDS */ struct amdgpu_gds gds; + /* KFD */ + struct amdgpu_kfd_dev kfd; + /* display related functionality */ struct amdgpu_display_manager dm; @@ -989,9 +887,6 @@ struct amdgpu_device { atomic64_t visible_pin_size; atomic64_t gart_pin_size; - /* amdkfd interface */ - struct kfd_dev *kfd; - /* soc15 register offset based on ip, instance and segment */ uint32_t *reg_offset[MAX_HWIP][HWIP_MAX_INSTANCE]; @@ -1023,6 +918,10 @@ struct amdgpu_device { unsigned long last_mm_index; bool in_gpu_reset; struct mutex lock_reset; + struct amdgpu_doorbell_index doorbell_index; + + int asic_reset_res; + struct work_struct xgmi_reset_work; }; static inline struct amdgpu_device *amdgpu_ttm_adev(struct ttm_bo_device *bdev) @@ -1047,11 +946,6 @@ uint8_t amdgpu_mm_rreg8(struct amdgpu_device *adev, uint32_t offset); u32 amdgpu_io_rreg(struct amdgpu_device *adev, u32 reg); void amdgpu_io_wreg(struct amdgpu_device *adev, u32 reg, u32 v); -u32 amdgpu_mm_rdoorbell(struct amdgpu_device *adev, u32 index); -void amdgpu_mm_wdoorbell(struct amdgpu_device *adev, u32 index, u32 v); -u64 amdgpu_mm_rdoorbell64(struct amdgpu_device *adev, u32 index); -void amdgpu_mm_wdoorbell64(struct amdgpu_device *adev, u32 index, u64 v); - bool amdgpu_device_asic_has_dc_support(enum amd_asic_type asic_type); bool amdgpu_device_has_dc_support(struct amdgpu_device *adev); @@ -1113,11 +1007,6 @@ int emu_soc_asic_init(struct amdgpu_device *adev); #define RREG32_IO(reg) amdgpu_io_rreg(adev, (reg)) #define WREG32_IO(reg, v) amdgpu_io_wreg(adev, (reg), (v)) -#define RDOORBELL32(index) amdgpu_mm_rdoorbell(adev, (index)) -#define WDOORBELL32(index, v) amdgpu_mm_wdoorbell(adev, (index), (v)) -#define RDOORBELL64(index) amdgpu_mm_rdoorbell64(adev, (index)) -#define WDOORBELL64(index, v) amdgpu_mm_wdoorbell64(adev, (index), (v)) - #define REG_FIELD_SHIFT(reg, field) reg##__##field##__SHIFT #define REG_FIELD_MASK(reg, field) reg##__##field##_MASK @@ -1159,6 +1048,9 @@ int emu_soc_asic_init(struct amdgpu_device *adev); #define amdgpu_asic_flush_hdp(adev, r) (adev)->asic_funcs->flush_hdp((adev), (r)) #define amdgpu_asic_invalidate_hdp(adev, r) (adev)->asic_funcs->invalidate_hdp((adev), (r)) #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))) +#define amdgpu_asic_need_reset_on_init(adev) (adev)->asic_funcs->need_reset_on_init((adev)) /* Common functions */ bool amdgpu_device_should_recover_gpu(struct amdgpu_device *adev); @@ -1219,12 +1111,6 @@ void amdgpu_disable_vblank_kms(struct drm_device *dev, unsigned int pipe); long amdgpu_kms_compat_ioctl(struct file *filp, unsigned int cmd, unsigned long arg); - -/* - * functions used by amdgpu_xgmi.c - */ -int amdgpu_xgmi_add_device(struct amdgpu_device *adev); - /* * functions used by amdgpu_encoder.c */ @@ -1252,6 +1138,9 @@ bool amdgpu_acpi_is_pcie_performance_request_supported(struct amdgpu_device *ade int amdgpu_acpi_pcie_performance_request(struct amdgpu_device *adev, u8 perf_req, bool advertise); int amdgpu_acpi_pcie_notify_device_ready(struct amdgpu_device *adev); + +void amdgpu_acpi_get_backlight_caps(struct amdgpu_device *adev, + struct amdgpu_dm_backlight_caps *caps); #else static inline int amdgpu_acpi_init(struct amdgpu_device *adev) { return 0; } static inline void amdgpu_acpi_fini(struct amdgpu_device *adev) { } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_acpi.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_acpi.c index 7f0afc526419..4376b17ca594 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_acpi.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_acpi.c @@ -41,28 +41,21 @@ struct amdgpu_atif_notification_cfg { }; struct amdgpu_atif_notifications { - bool display_switch; - bool expansion_mode_change; bool thermal_state; bool forced_power_state; bool system_power_state; - bool display_conf_change; - bool px_gfx_switch; bool brightness_change; bool dgpu_display_event; + bool gpu_package_power_limit; }; struct amdgpu_atif_functions { bool system_params; bool sbios_requests; - bool select_active_disp; - bool lid_state; - bool get_tv_standard; - bool set_tv_standard; - bool get_panel_expansion_mode; - bool set_panel_expansion_mode; bool temperature_change; - bool graphics_device_types; + bool query_backlight_transfer_characteristics; + bool ready_to_undock; + bool external_gpu_information; }; struct amdgpu_atif { @@ -72,6 +65,7 @@ struct amdgpu_atif { struct amdgpu_atif_functions functions; struct amdgpu_atif_notification_cfg notification_cfg; struct amdgpu_encoder *encoder_for_bl; + struct amdgpu_dm_backlight_caps backlight_caps; }; /* Call the ATIF method @@ -137,15 +131,12 @@ static union acpi_object *amdgpu_atif_call(struct amdgpu_atif *atif, */ static void amdgpu_atif_parse_notification(struct amdgpu_atif_notifications *n, u32 mask) { - n->display_switch = mask & ATIF_DISPLAY_SWITCH_REQUEST_SUPPORTED; - n->expansion_mode_change = mask & ATIF_EXPANSION_MODE_CHANGE_REQUEST_SUPPORTED; n->thermal_state = mask & ATIF_THERMAL_STATE_CHANGE_REQUEST_SUPPORTED; n->forced_power_state = mask & ATIF_FORCED_POWER_STATE_CHANGE_REQUEST_SUPPORTED; n->system_power_state = mask & ATIF_SYSTEM_POWER_SOURCE_CHANGE_REQUEST_SUPPORTED; - n->display_conf_change = mask & ATIF_DISPLAY_CONF_CHANGE_REQUEST_SUPPORTED; - n->px_gfx_switch = mask & ATIF_PX_GFX_SWITCH_REQUEST_SUPPORTED; n->brightness_change = mask & ATIF_PANEL_BRIGHTNESS_CHANGE_REQUEST_SUPPORTED; n->dgpu_display_event = mask & ATIF_DGPU_DISPLAY_EVENT_SUPPORTED; + n->gpu_package_power_limit = mask & ATIF_GPU_PACKAGE_POWER_LIMIT_REQUEST_SUPPORTED; } /** @@ -162,14 +153,11 @@ static void amdgpu_atif_parse_functions(struct amdgpu_atif_functions *f, u32 mas { f->system_params = mask & ATIF_GET_SYSTEM_PARAMETERS_SUPPORTED; f->sbios_requests = mask & ATIF_GET_SYSTEM_BIOS_REQUESTS_SUPPORTED; - f->select_active_disp = mask & ATIF_SELECT_ACTIVE_DISPLAYS_SUPPORTED; - f->lid_state = mask & ATIF_GET_LID_STATE_SUPPORTED; - f->get_tv_standard = mask & ATIF_GET_TV_STANDARD_FROM_CMOS_SUPPORTED; - f->set_tv_standard = mask & ATIF_SET_TV_STANDARD_IN_CMOS_SUPPORTED; - f->get_panel_expansion_mode = mask & ATIF_GET_PANEL_EXPANSION_MODE_FROM_CMOS_SUPPORTED; - f->set_panel_expansion_mode = mask & ATIF_SET_PANEL_EXPANSION_MODE_IN_CMOS_SUPPORTED; f->temperature_change = mask & ATIF_TEMPERATURE_CHANGE_NOTIFICATION_SUPPORTED; - f->graphics_device_types = mask & ATIF_GET_GRAPHICS_DEVICE_TYPES_SUPPORTED; + f->query_backlight_transfer_characteristics = + mask & ATIF_QUERY_BACKLIGHT_TRANSFER_CHARACTERISTICS_SUPPORTED; + f->ready_to_undock = mask & ATIF_READY_TO_UNDOCK_NOTIFICATION_SUPPORTED; + f->external_gpu_information = mask & ATIF_GET_EXTERNAL_GPU_INFORMATION_SUPPORTED; } /** @@ -311,6 +299,65 @@ out: } /** + * amdgpu_atif_query_backlight_caps - get min and max backlight input signal + * + * @handle: acpi handle + * + * Execute the QUERY_BRIGHTNESS_TRANSFER_CHARACTERISTICS ATIF function + * to determine the acceptable range of backlight values + * + * Backlight_caps.caps_valid will be set to true if the query is successful + * + * The input signals are in range 0-255 + * + * This function assumes the display with backlight is the first LCD + * + * Returns 0 on success, error on failure. + */ +static int amdgpu_atif_query_backlight_caps(struct amdgpu_atif *atif) +{ + union acpi_object *info; + struct atif_qbtc_output characteristics; + struct atif_qbtc_arguments arguments; + struct acpi_buffer params; + size_t size; + int err = 0; + + arguments.size = sizeof(arguments); + arguments.requested_display = ATIF_QBTC_REQUEST_LCD1; + + params.length = sizeof(arguments); + params.pointer = (void *)&arguments; + + info = amdgpu_atif_call(atif, + ATIF_FUNCTION_QUERY_BRIGHTNESS_TRANSFER_CHARACTERISTICS, + ¶ms); + if (!info) { + err = -EIO; + goto out; + } + + size = *(u16 *) info->buffer.pointer; + if (size < 10) { + err = -EINVAL; + goto out; + } + + memset(&characteristics, 0, sizeof(characteristics)); + size = min(sizeof(characteristics), size); + memcpy(&characteristics, info->buffer.pointer, size); + + atif->backlight_caps.caps_valid = true; + atif->backlight_caps.min_input_signal = + characteristics.min_input_signal; + atif->backlight_caps.max_input_signal = + characteristics.max_input_signal; +out: + kfree(info); + return err; +} + +/** * amdgpu_atif_get_sbios_requests - get requested sbios event * * @handle: acpi handle @@ -799,6 +846,17 @@ int amdgpu_acpi_init(struct amdgpu_device *adev) } } + if (atif->functions.query_backlight_transfer_characteristics) { + ret = amdgpu_atif_query_backlight_caps(atif); + if (ret) { + DRM_DEBUG_DRIVER("Call to QUERY_BACKLIGHT_TRANSFER_CHARACTERISTICS failed: %d\n", + ret); + atif->backlight_caps.caps_valid = false; + } + } else { + atif->backlight_caps.caps_valid = false; + } + out: adev->acpi_nb.notifier_call = amdgpu_acpi_event; register_acpi_notifier(&adev->acpi_nb); @@ -806,6 +864,18 @@ out: return ret; } +void amdgpu_acpi_get_backlight_caps(struct amdgpu_device *adev, + struct amdgpu_dm_backlight_caps *caps) +{ + if (!adev->atif) { + caps->caps_valid = false; + return; + } + caps->caps_valid = adev->atif->backlight_caps.caps_valid; + caps->min_input_signal = adev->atif->backlight_caps.min_input_signal; + caps->max_input_signal = adev->atif->backlight_caps.max_input_signal; +} + /** * amdgpu_acpi_fini - tear down driver acpi support * @@ -816,6 +886,5 @@ out: void amdgpu_acpi_fini(struct amdgpu_device *adev) { unregister_acpi_notifier(&adev->acpi_nb); - if (adev->atif) - kfree(adev->atif); + kfree(adev->atif); } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c index 1580ec60b89f..fe1d7368c1e6 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c @@ -26,22 +26,28 @@ #include "amdgpu.h" #include "amdgpu_gfx.h" #include <linux/module.h> - -const struct kgd2kfd_calls *kgd2kfd; +#include <linux/dma-buf.h> static const unsigned int compute_vmid_bitmap = 0xFF00; +/* Total memory size in system memory and all GPU VRAM. Used to + * estimate worst case amount of memory to reserve for page tables + */ +uint64_t amdgpu_amdkfd_total_mem_size; + int amdgpu_amdkfd_init(void) { + struct sysinfo si; int ret; + si_meminfo(&si); + amdgpu_amdkfd_total_mem_size = si.totalram - si.totalhigh; + amdgpu_amdkfd_total_mem_size *= si.mem_unit; + #ifdef CONFIG_HSA_AMD - ret = kgd2kfd_init(KFD_INTERFACE_VERSION, &kgd2kfd); - if (ret) - kgd2kfd = NULL; + ret = kgd2kfd_init(); amdgpu_amdkfd_gpuvm_init_mem_limits(); #else - kgd2kfd = NULL; ret = -ENOENT; #endif @@ -50,17 +56,13 @@ int amdgpu_amdkfd_init(void) void amdgpu_amdkfd_fini(void) { - if (kgd2kfd) - kgd2kfd->exit(); + kgd2kfd_exit(); } void amdgpu_amdkfd_device_probe(struct amdgpu_device *adev) { const struct kfd2kgd_calls *kfd2kgd; - if (!kgd2kfd) - return; - switch (adev->asic_type) { #ifdef CONFIG_DRM_AMDGPU_CIK case CHIP_KAVERI: @@ -73,9 +75,11 @@ void amdgpu_amdkfd_device_probe(struct amdgpu_device *adev) case CHIP_FIJI: case CHIP_POLARIS10: case CHIP_POLARIS11: + case CHIP_POLARIS12: kfd2kgd = amdgpu_amdkfd_gfx_8_0_get_functions(); break; case CHIP_VEGA10: + case CHIP_VEGA12: case CHIP_VEGA20: case CHIP_RAVEN: kfd2kgd = amdgpu_amdkfd_gfx_9_0_get_functions(); @@ -85,8 +89,11 @@ void amdgpu_amdkfd_device_probe(struct amdgpu_device *adev) return; } - adev->kfd = kgd2kfd->probe((struct kgd_dev *)adev, - adev->pdev, kfd2kgd); + adev->kfd.dev = kgd2kfd_probe((struct kgd_dev *)adev, + adev->pdev, kfd2kgd); + + if (adev->kfd.dev) + amdgpu_amdkfd_total_mem_size += adev->gmc.real_vram_size; } /** @@ -124,9 +131,10 @@ static void amdgpu_doorbell_get_kfd_info(struct amdgpu_device *adev, void amdgpu_amdkfd_device_init(struct amdgpu_device *adev) { - int i, n; + int i; int last_valid_bit; - if (adev->kfd) { + + if (adev->kfd.dev) { struct kgd2kfd_shared_resources gpu_resources = { .compute_vmid_bitmap = compute_vmid_bitmap, .num_pipe_per_mec = adev->gfx.mec.num_pipe_per_mec, @@ -134,7 +142,9 @@ void amdgpu_amdkfd_device_init(struct amdgpu_device *adev) .gpuvm_size = min(adev->vm_manager.max_pfn << AMDGPU_GPU_PAGE_SHIFT, AMDGPU_GMC_HOLE_START), - .drm_render_minor = adev->ddev->render->index + .drm_render_minor = adev->ddev->render->index, + .sdma_doorbell_idx = adev->doorbell_index.sdma_engine, + }; /* this is going to have a few of the MSBs set that we need to @@ -144,7 +154,7 @@ void amdgpu_amdkfd_device_init(struct amdgpu_device *adev) KGD_MAX_QUEUES); /* remove the KIQ bit as well */ - if (adev->gfx.kiq.ring.ready) + if (adev->gfx.kiq.ring.sched.ready) clear_bit(amdgpu_gfx_queue_to_bit(adev, adev->gfx.kiq.ring.me - 1, adev->gfx.kiq.ring.pipe, @@ -164,78 +174,52 @@ void amdgpu_amdkfd_device_init(struct amdgpu_device *adev) &gpu_resources.doorbell_aperture_size, &gpu_resources.doorbell_start_offset); - if (adev->asic_type < CHIP_VEGA10) { - kgd2kfd->device_init(adev->kfd, &gpu_resources); - return; - } - - n = (adev->asic_type < CHIP_VEGA20) ? 2 : 8; - - for (i = 0; i < n; i += 2) { - /* On SOC15 the BIF is involved in routing - * doorbells using the low 12 bits of the - * address. Communicate the assignments to - * KFD. KFD uses two doorbell pages per - * process in case of 64-bit doorbells so we - * can use each doorbell assignment twice. - */ - if (adev->asic_type == CHIP_VEGA10) { - gpu_resources.sdma_doorbell[0][i] = - AMDGPU_VEGA10_DOORBELL64_sDMA_ENGINE0 + (i >> 1); - gpu_resources.sdma_doorbell[0][i+1] = - AMDGPU_VEGA10_DOORBELL64_sDMA_ENGINE0 + 0x200 + (i >> 1); - gpu_resources.sdma_doorbell[1][i] = - AMDGPU_VEGA10_DOORBELL64_sDMA_ENGINE1 + (i >> 1); - gpu_resources.sdma_doorbell[1][i+1] = - AMDGPU_VEGA10_DOORBELL64_sDMA_ENGINE1 + 0x200 + (i >> 1); - } else { - gpu_resources.sdma_doorbell[0][i] = - AMDGPU_DOORBELL64_sDMA_ENGINE0 + (i >> 1); - gpu_resources.sdma_doorbell[0][i+1] = - AMDGPU_DOORBELL64_sDMA_ENGINE0 + 0x200 + (i >> 1); - gpu_resources.sdma_doorbell[1][i] = - AMDGPU_DOORBELL64_sDMA_ENGINE1 + (i >> 1); - gpu_resources.sdma_doorbell[1][i+1] = - AMDGPU_DOORBELL64_sDMA_ENGINE1 + 0x200 + (i >> 1); - } - } - /* Doorbells 0x0e0-0ff and 0x2e0-2ff are reserved for - * SDMA, IH and VCN. So don't use them for the CP. + /* Since SOC15, BIF starts to statically use the + * lower 12 bits of doorbell addresses for routing + * based on settings in registers like + * SDMA0_DOORBELL_RANGE etc.. + * In order to route a doorbell to CP engine, the lower + * 12 bits of its address has to be outside the range + * set for SDMA, VCN, and IH blocks. */ - gpu_resources.reserved_doorbell_mask = 0x1e0; - gpu_resources.reserved_doorbell_val = 0x0e0; + if (adev->asic_type >= CHIP_VEGA10) { + gpu_resources.non_cp_doorbells_start = + adev->doorbell_index.first_non_cp; + gpu_resources.non_cp_doorbells_end = + adev->doorbell_index.last_non_cp; + } - kgd2kfd->device_init(adev->kfd, &gpu_resources); + kgd2kfd_device_init(adev->kfd.dev, &gpu_resources); } } void amdgpu_amdkfd_device_fini(struct amdgpu_device *adev) { - if (adev->kfd) { - kgd2kfd->device_exit(adev->kfd); - adev->kfd = NULL; + if (adev->kfd.dev) { + kgd2kfd_device_exit(adev->kfd.dev); + adev->kfd.dev = NULL; } } void amdgpu_amdkfd_interrupt(struct amdgpu_device *adev, const void *ih_ring_entry) { - if (adev->kfd) - kgd2kfd->interrupt(adev->kfd, ih_ring_entry); + if (adev->kfd.dev) + kgd2kfd_interrupt(adev->kfd.dev, ih_ring_entry); } void amdgpu_amdkfd_suspend(struct amdgpu_device *adev) { - if (adev->kfd) - kgd2kfd->suspend(adev->kfd); + if (adev->kfd.dev) + kgd2kfd_suspend(adev->kfd.dev); } int amdgpu_amdkfd_resume(struct amdgpu_device *adev) { int r = 0; - if (adev->kfd) - r = kgd2kfd->resume(adev->kfd); + if (adev->kfd.dev) + r = kgd2kfd_resume(adev->kfd.dev); return r; } @@ -244,8 +228,8 @@ int amdgpu_amdkfd_pre_reset(struct amdgpu_device *adev) { int r = 0; - if (adev->kfd) - r = kgd2kfd->pre_reset(adev->kfd); + if (adev->kfd.dev) + r = kgd2kfd_pre_reset(adev->kfd.dev); return r; } @@ -254,8 +238,8 @@ int amdgpu_amdkfd_post_reset(struct amdgpu_device *adev) { int r = 0; - if (adev->kfd) - r = kgd2kfd->post_reset(adev->kfd); + if (adev->kfd.dev) + r = kgd2kfd_post_reset(adev->kfd.dev); return r; } @@ -268,9 +252,9 @@ void amdgpu_amdkfd_gpu_reset(struct kgd_dev *kgd) amdgpu_device_gpu_recover(adev, NULL); } -int alloc_gtt_mem(struct kgd_dev *kgd, size_t size, - void **mem_obj, uint64_t *gpu_addr, - void **cpu_ptr, bool mqd_gfx9) +int amdgpu_amdkfd_alloc_gtt_mem(struct kgd_dev *kgd, size_t size, + void **mem_obj, uint64_t *gpu_addr, + void **cpu_ptr, bool mqd_gfx9) { struct amdgpu_device *adev = (struct amdgpu_device *)kgd; struct amdgpu_bo *bo = NULL; @@ -340,7 +324,7 @@ allocate_mem_reserve_bo_failed: return r; } -void free_gtt_mem(struct kgd_dev *kgd, void *mem_obj) +void amdgpu_amdkfd_free_gtt_mem(struct kgd_dev *kgd, void *mem_obj) { struct amdgpu_bo *bo = (struct amdgpu_bo *) mem_obj; @@ -351,8 +335,8 @@ void free_gtt_mem(struct kgd_dev *kgd, void *mem_obj) amdgpu_bo_unref(&(bo)); } -void get_local_mem_info(struct kgd_dev *kgd, - struct kfd_local_mem_info *mem_info) +void amdgpu_amdkfd_get_local_mem_info(struct kgd_dev *kgd, + struct kfd_local_mem_info *mem_info) { struct amdgpu_device *adev = (struct amdgpu_device *)kgd; uint64_t address_mask = adev->dev->dma_mask ? ~*adev->dev->dma_mask : @@ -383,7 +367,7 @@ void get_local_mem_info(struct kgd_dev *kgd, mem_info->mem_clk_max = 100; } -uint64_t get_gpu_clock_counter(struct kgd_dev *kgd) +uint64_t amdgpu_amdkfd_get_gpu_clock_counter(struct kgd_dev *kgd) { struct amdgpu_device *adev = (struct amdgpu_device *)kgd; @@ -392,7 +376,7 @@ uint64_t get_gpu_clock_counter(struct kgd_dev *kgd) return 0; } -uint32_t get_max_engine_clock_in_mhz(struct kgd_dev *kgd) +uint32_t amdgpu_amdkfd_get_max_engine_clock_in_mhz(struct kgd_dev *kgd) { struct amdgpu_device *adev = (struct amdgpu_device *)kgd; @@ -405,7 +389,7 @@ uint32_t get_max_engine_clock_in_mhz(struct kgd_dev *kgd) return 100; } -void get_cu_info(struct kgd_dev *kgd, struct kfd_cu_info *cu_info) +void amdgpu_amdkfd_get_cu_info(struct kgd_dev *kgd, struct kfd_cu_info *cu_info) { struct amdgpu_device *adev = (struct amdgpu_device *)kgd; struct amdgpu_cu_info acu_info = adev->gfx.cu_info; @@ -428,6 +412,62 @@ void get_cu_info(struct kgd_dev *kgd, struct kfd_cu_info *cu_info) cu_info->lds_size = acu_info.lds_size; } +int amdgpu_amdkfd_get_dmabuf_info(struct kgd_dev *kgd, int dma_buf_fd, + struct kgd_dev **dma_buf_kgd, + uint64_t *bo_size, void *metadata_buffer, + size_t buffer_size, uint32_t *metadata_size, + uint32_t *flags) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)kgd; + struct dma_buf *dma_buf; + struct drm_gem_object *obj; + struct amdgpu_bo *bo; + uint64_t metadata_flags; + int r = -EINVAL; + + dma_buf = dma_buf_get(dma_buf_fd); + if (IS_ERR(dma_buf)) + return PTR_ERR(dma_buf); + + if (dma_buf->ops != &amdgpu_dmabuf_ops) + /* Can't handle non-graphics buffers */ + goto out_put; + + obj = dma_buf->priv; + if (obj->dev->driver != adev->ddev->driver) + /* Can't handle buffers from different drivers */ + goto out_put; + + adev = obj->dev->dev_private; + bo = gem_to_amdgpu_bo(obj); + if (!(bo->preferred_domains & (AMDGPU_GEM_DOMAIN_VRAM | + AMDGPU_GEM_DOMAIN_GTT))) + /* Only VRAM and GTT BOs are supported */ + goto out_put; + + r = 0; + if (dma_buf_kgd) + *dma_buf_kgd = (struct kgd_dev *)adev; + if (bo_size) + *bo_size = amdgpu_bo_size(bo); + if (metadata_size) + *metadata_size = bo->metadata_size; + if (metadata_buffer) + r = amdgpu_bo_get_metadata(bo, metadata_buffer, buffer_size, + metadata_size, &metadata_flags); + if (flags) { + *flags = (bo->preferred_domains & AMDGPU_GEM_DOMAIN_VRAM) ? + ALLOC_MEM_FLAGS_VRAM : ALLOC_MEM_FLAGS_GTT; + + if (bo->flags & AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED) + *flags |= ALLOC_MEM_FLAGS_PUBLIC; + } + +out_put: + dma_buf_put(dma_buf); + return r; +} + uint64_t amdgpu_amdkfd_get_vram_usage(struct kgd_dev *kgd) { struct amdgpu_device *adev = (struct amdgpu_device *)kgd; @@ -510,7 +550,7 @@ void amdgpu_amdkfd_set_compute_idle(struct kgd_dev *kgd, bool idle) bool amdgpu_amdkfd_is_kfd_vmid(struct amdgpu_device *adev, u32 vmid) { - if (adev->kfd) { + if (adev->kfd.dev) { if ((1 << vmid) & compute_vmid_bitmap) return true; } @@ -524,7 +564,7 @@ bool amdkfd_fence_check_mm(struct dma_fence *f, struct mm_struct *mm) return false; } -void amdgpu_amdkfd_unreserve_system_memory_limit(struct amdgpu_bo *bo) +void amdgpu_amdkfd_unreserve_memory_limit(struct amdgpu_bo *bo) { } @@ -557,4 +597,47 @@ struct kfd2kgd_calls *amdgpu_amdkfd_gfx_9_0_get_functions(void) { return NULL; } + +struct kfd_dev *kgd2kfd_probe(struct kgd_dev *kgd, struct pci_dev *pdev, + const struct kfd2kgd_calls *f2g) +{ + return NULL; +} + +bool kgd2kfd_device_init(struct kfd_dev *kfd, + const struct kgd2kfd_shared_resources *gpu_resources) +{ + return false; +} + +void kgd2kfd_device_exit(struct kfd_dev *kfd) +{ +} + +void kgd2kfd_exit(void) +{ +} + +void kgd2kfd_suspend(struct kfd_dev *kfd) +{ +} + +int kgd2kfd_resume(struct kfd_dev *kfd) +{ + return 0; +} + +int kgd2kfd_pre_reset(struct kfd_dev *kfd) +{ + return 0; +} + +int kgd2kfd_post_reset(struct kfd_dev *kfd) +{ + return 0; +} + +void kgd2kfd_interrupt(struct kfd_dev *kfd, const void *ih_ring_entry) +{ +} #endif diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h index 8e0d4f7196b4..0b31a1859023 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h @@ -27,14 +27,13 @@ #include <linux/types.h> #include <linux/mm.h> -#include <linux/mmu_context.h> #include <linux/workqueue.h> #include <kgd_kfd_interface.h> #include <drm/ttm/ttm_execbuf_util.h> #include "amdgpu_sync.h" #include "amdgpu_vm.h" -extern const struct kgd2kfd_calls *kgd2kfd; +extern uint64_t amdgpu_amdkfd_total_mem_size; struct amdgpu_device; @@ -77,6 +76,11 @@ struct amdgpu_amdkfd_fence { char timeline_name[TASK_COMM_LEN]; }; +struct amdgpu_kfd_dev { + struct kfd_dev *dev; + uint64_t vram_used; +}; + struct amdgpu_amdkfd_fence *amdgpu_amdkfd_fence_create(u64 context, struct mm_struct *mm); bool amdkfd_fence_check_mm(struct dma_fence *f, struct mm_struct *mm); @@ -134,16 +138,21 @@ int amdgpu_amdkfd_post_reset(struct amdgpu_device *adev); void amdgpu_amdkfd_gpu_reset(struct kgd_dev *kgd); /* Shared API */ -int alloc_gtt_mem(struct kgd_dev *kgd, size_t size, - void **mem_obj, uint64_t *gpu_addr, - void **cpu_ptr, bool mqd_gfx9); -void free_gtt_mem(struct kgd_dev *kgd, void *mem_obj); -void get_local_mem_info(struct kgd_dev *kgd, - struct kfd_local_mem_info *mem_info); -uint64_t get_gpu_clock_counter(struct kgd_dev *kgd); - -uint32_t get_max_engine_clock_in_mhz(struct kgd_dev *kgd); -void get_cu_info(struct kgd_dev *kgd, struct kfd_cu_info *cu_info); +int amdgpu_amdkfd_alloc_gtt_mem(struct kgd_dev *kgd, size_t size, + void **mem_obj, uint64_t *gpu_addr, + void **cpu_ptr, bool mqd_gfx9); +void amdgpu_amdkfd_free_gtt_mem(struct kgd_dev *kgd, void *mem_obj); +void amdgpu_amdkfd_get_local_mem_info(struct kgd_dev *kgd, + struct kfd_local_mem_info *mem_info); +uint64_t amdgpu_amdkfd_get_gpu_clock_counter(struct kgd_dev *kgd); + +uint32_t amdgpu_amdkfd_get_max_engine_clock_in_mhz(struct kgd_dev *kgd); +void amdgpu_amdkfd_get_cu_info(struct kgd_dev *kgd, struct kfd_cu_info *cu_info); +int amdgpu_amdkfd_get_dmabuf_info(struct kgd_dev *kgd, int dma_buf_fd, + struct kgd_dev **dmabuf_kgd, + 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 kgd_dev *kgd); uint64_t amdgpu_amdkfd_get_hive_id(struct kgd_dev *kgd); @@ -195,7 +204,31 @@ int amdgpu_amdkfd_gpuvm_restore_process_bos(void *process_info, int amdgpu_amdkfd_gpuvm_get_vm_fault_info(struct kgd_dev *kgd, struct kfd_vm_fault_info *info); +int amdgpu_amdkfd_gpuvm_import_dmabuf(struct kgd_dev *kgd, + struct dma_buf *dmabuf, + uint64_t va, void *vm, + struct kgd_mem **mem, uint64_t *size, + uint64_t *mmap_offset); + void amdgpu_amdkfd_gpuvm_init_mem_limits(void); -void amdgpu_amdkfd_unreserve_system_memory_limit(struct amdgpu_bo *bo); +void amdgpu_amdkfd_unreserve_memory_limit(struct amdgpu_bo *bo); + +/* KGD2KFD callbacks */ +int kgd2kfd_init(void); +void kgd2kfd_exit(void); +struct kfd_dev *kgd2kfd_probe(struct kgd_dev *kgd, struct pci_dev *pdev, + const struct kfd2kgd_calls *f2g); +bool kgd2kfd_device_init(struct kfd_dev *kfd, + const struct kgd2kfd_shared_resources *gpu_resources); +void kgd2kfd_device_exit(struct kfd_dev *kfd); +void kgd2kfd_suspend(struct kfd_dev *kfd); +int kgd2kfd_resume(struct kfd_dev *kfd); +int kgd2kfd_pre_reset(struct kfd_dev *kfd); +int kgd2kfd_post_reset(struct kfd_dev *kfd); +void kgd2kfd_interrupt(struct kfd_dev *kfd, const void *ih_ring_entry); +int kgd2kfd_quiesce_mm(struct mm_struct *mm); +int kgd2kfd_resume_mm(struct mm_struct *mm); +int kgd2kfd_schedule_evict_and_restore_process(struct mm_struct *mm, + struct dma_fence *fence); #endif /* AMDGPU_AMDKFD_H_INCLUDED */ diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_fence.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_fence.c index 574c1181ae9a..3107b9575929 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_fence.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_fence.c @@ -122,7 +122,7 @@ static bool amdkfd_fence_enable_signaling(struct dma_fence *f) if (dma_fence_is_signaled(f)) return true; - if (!kgd2kfd->schedule_evict_and_restore_process(fence->mm, f)) + if (!kgd2kfd_schedule_evict_and_restore_process(fence->mm, f)) return true; return false; 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 244d9834a381..ff7fac7df34b 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c @@ -23,6 +23,7 @@ #include <linux/fdtable.h> #include <linux/uaccess.h> #include <linux/firmware.h> +#include <linux/mmu_context.h> #include <drm/drmP.h> #include "amdgpu.h" #include "amdgpu_amdkfd.h" @@ -173,13 +174,6 @@ static int get_tile_config(struct kgd_dev *kgd, } static const struct kfd2kgd_calls kfd2kgd = { - .init_gtt_mem_allocation = alloc_gtt_mem, - .free_gtt_mem = free_gtt_mem, - .get_local_mem_info = get_local_mem_info, - .get_gpu_clock_counter = get_gpu_clock_counter, - .get_max_engine_clock_in_mhz = get_max_engine_clock_in_mhz, - .alloc_pasid = amdgpu_pasid_alloc, - .free_pasid = amdgpu_pasid_free, .program_sh_mem_settings = kgd_program_sh_mem_settings, .set_pasid_vmid_mapping = kgd_set_pasid_vmid_mapping, .init_interrupts = kgd_init_interrupts, @@ -200,28 +194,10 @@ static const struct kfd2kgd_calls kfd2kgd = { .get_fw_version = get_fw_version, .set_scratch_backing_va = set_scratch_backing_va, .get_tile_config = get_tile_config, - .get_cu_info = get_cu_info, - .get_vram_usage = amdgpu_amdkfd_get_vram_usage, - .create_process_vm = amdgpu_amdkfd_gpuvm_create_process_vm, - .acquire_process_vm = amdgpu_amdkfd_gpuvm_acquire_process_vm, - .destroy_process_vm = amdgpu_amdkfd_gpuvm_destroy_process_vm, - .release_process_vm = amdgpu_amdkfd_gpuvm_release_process_vm, - .get_process_page_dir = amdgpu_amdkfd_gpuvm_get_process_page_dir, .set_vm_context_page_table_base = set_vm_context_page_table_base, - .alloc_memory_of_gpu = amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu, - .free_memory_of_gpu = amdgpu_amdkfd_gpuvm_free_memory_of_gpu, - .map_memory_to_gpu = amdgpu_amdkfd_gpuvm_map_memory_to_gpu, - .unmap_memory_to_gpu = amdgpu_amdkfd_gpuvm_unmap_memory_from_gpu, - .sync_memory = amdgpu_amdkfd_gpuvm_sync_memory, - .map_gtt_bo_to_kernel = amdgpu_amdkfd_gpuvm_map_gtt_bo_to_kernel, - .restore_process_bos = amdgpu_amdkfd_gpuvm_restore_process_bos, .invalidate_tlbs = invalidate_tlbs, .invalidate_tlbs_vmid = invalidate_tlbs_vmid, - .submit_ib = amdgpu_amdkfd_submit_ib, - .get_vm_fault_info = amdgpu_amdkfd_gpuvm_get_vm_fault_info, .read_vmid_from_vmfault_reg = read_vmid_from_vmfault_reg, - .gpu_recover = amdgpu_amdkfd_gpu_reset, - .set_compute_idle = amdgpu_amdkfd_set_compute_idle }; struct kfd2kgd_calls *amdgpu_amdkfd_gfx_7_get_functions(void) 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 9f149914ad6c..56ea929f524b 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c @@ -24,6 +24,7 @@ #include <linux/fdtable.h> #include <linux/uaccess.h> #include <linux/firmware.h> +#include <linux/mmu_context.h> #include <drm/drmP.h> #include "amdgpu.h" #include "amdgpu_amdkfd.h" @@ -128,13 +129,6 @@ static int get_tile_config(struct kgd_dev *kgd, } static const struct kfd2kgd_calls kfd2kgd = { - .init_gtt_mem_allocation = alloc_gtt_mem, - .free_gtt_mem = free_gtt_mem, - .get_local_mem_info = get_local_mem_info, - .get_gpu_clock_counter = get_gpu_clock_counter, - .get_max_engine_clock_in_mhz = get_max_engine_clock_in_mhz, - .alloc_pasid = amdgpu_pasid_alloc, - .free_pasid = amdgpu_pasid_free, .program_sh_mem_settings = kgd_program_sh_mem_settings, .set_pasid_vmid_mapping = kgd_set_pasid_vmid_mapping, .init_interrupts = kgd_init_interrupts, @@ -157,27 +151,9 @@ static const struct kfd2kgd_calls kfd2kgd = { .get_fw_version = get_fw_version, .set_scratch_backing_va = set_scratch_backing_va, .get_tile_config = get_tile_config, - .get_cu_info = get_cu_info, - .get_vram_usage = amdgpu_amdkfd_get_vram_usage, - .create_process_vm = amdgpu_amdkfd_gpuvm_create_process_vm, - .acquire_process_vm = amdgpu_amdkfd_gpuvm_acquire_process_vm, - .destroy_process_vm = amdgpu_amdkfd_gpuvm_destroy_process_vm, - .release_process_vm = amdgpu_amdkfd_gpuvm_release_process_vm, - .get_process_page_dir = amdgpu_amdkfd_gpuvm_get_process_page_dir, .set_vm_context_page_table_base = set_vm_context_page_table_base, - .alloc_memory_of_gpu = amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu, - .free_memory_of_gpu = amdgpu_amdkfd_gpuvm_free_memory_of_gpu, - .map_memory_to_gpu = amdgpu_amdkfd_gpuvm_map_memory_to_gpu, - .unmap_memory_to_gpu = amdgpu_amdkfd_gpuvm_unmap_memory_from_gpu, - .sync_memory = amdgpu_amdkfd_gpuvm_sync_memory, - .map_gtt_bo_to_kernel = amdgpu_amdkfd_gpuvm_map_gtt_bo_to_kernel, - .restore_process_bos = amdgpu_amdkfd_gpuvm_restore_process_bos, .invalidate_tlbs = invalidate_tlbs, .invalidate_tlbs_vmid = invalidate_tlbs_vmid, - .submit_ib = amdgpu_amdkfd_submit_ib, - .get_vm_fault_info = amdgpu_amdkfd_gpuvm_get_vm_fault_info, - .gpu_recover = amdgpu_amdkfd_gpu_reset, - .set_compute_idle = amdgpu_amdkfd_set_compute_idle }; struct kfd2kgd_calls *amdgpu_amdkfd_gfx_8_0_get_functions(void) 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 42cb4c4e0929..5c51d4910650 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c @@ -26,6 +26,7 @@ #include <linux/fdtable.h> #include <linux/uaccess.h> #include <linux/firmware.h> +#include <linux/mmu_context.h> #include <drm/drmP.h> #include "amdgpu.h" #include "amdgpu_amdkfd.h" @@ -46,38 +47,9 @@ #include "v9_structs.h" #include "soc15.h" #include "soc15d.h" +#include "mmhub_v1_0.h" +#include "gfxhub_v1_0.h" -/* HACK: MMHUB and GC both have VM-related register with the same - * names but different offsets. Define the MMHUB register we need here - * with a prefix. A proper solution would be to move the functions - * programming these registers into gfx_v9_0.c and mmhub_v1_0.c - * respectively. - */ -#define mmMMHUB_VM_INVALIDATE_ENG16_REQ 0x06f3 -#define mmMMHUB_VM_INVALIDATE_ENG16_REQ_BASE_IDX 0 - -#define mmMMHUB_VM_INVALIDATE_ENG16_ACK 0x0705 -#define mmMMHUB_VM_INVALIDATE_ENG16_ACK_BASE_IDX 0 - -#define mmMMHUB_VM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32 0x072b -#define mmMMHUB_VM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32_BASE_IDX 0 -#define mmMMHUB_VM_CONTEXT0_PAGE_TABLE_BASE_ADDR_HI32 0x072c -#define mmMMHUB_VM_CONTEXT0_PAGE_TABLE_BASE_ADDR_HI32_BASE_IDX 0 - -#define mmMMHUB_VM_CONTEXT0_PAGE_TABLE_START_ADDR_LO32 0x074b -#define mmMMHUB_VM_CONTEXT0_PAGE_TABLE_START_ADDR_LO32_BASE_IDX 0 -#define mmMMHUB_VM_CONTEXT0_PAGE_TABLE_START_ADDR_HI32 0x074c -#define mmMMHUB_VM_CONTEXT0_PAGE_TABLE_START_ADDR_HI32_BASE_IDX 0 - -#define mmMMHUB_VM_CONTEXT0_PAGE_TABLE_END_ADDR_LO32 0x076b -#define mmMMHUB_VM_CONTEXT0_PAGE_TABLE_END_ADDR_LO32_BASE_IDX 0 -#define mmMMHUB_VM_CONTEXT0_PAGE_TABLE_END_ADDR_HI32 0x076c -#define mmMMHUB_VM_CONTEXT0_PAGE_TABLE_END_ADDR_HI32_BASE_IDX 0 - -#define mmMMHUB_VM_INVALIDATE_ENG16_ADDR_RANGE_LO32 0x0727 -#define mmMMHUB_VM_INVALIDATE_ENG16_ADDR_RANGE_LO32_BASE_IDX 0 -#define mmMMHUB_VM_INVALIDATE_ENG16_ADDR_RANGE_HI32 0x0728 -#define mmMMHUB_VM_INVALIDATE_ENG16_ADDR_RANGE_HI32_BASE_IDX 0 #define V9_PIPE_PER_MEC (4) #define V9_QUEUES_PER_PIPE_MEC (8) @@ -167,13 +139,6 @@ static int amdgpu_amdkfd_get_tile_config(struct kgd_dev *kgd, } static const struct kfd2kgd_calls kfd2kgd = { - .init_gtt_mem_allocation = alloc_gtt_mem, - .free_gtt_mem = free_gtt_mem, - .get_local_mem_info = get_local_mem_info, - .get_gpu_clock_counter = get_gpu_clock_counter, - .get_max_engine_clock_in_mhz = get_max_engine_clock_in_mhz, - .alloc_pasid = amdgpu_pasid_alloc, - .free_pasid = amdgpu_pasid_free, .program_sh_mem_settings = kgd_program_sh_mem_settings, .set_pasid_vmid_mapping = kgd_set_pasid_vmid_mapping, .init_interrupts = kgd_init_interrupts, @@ -196,26 +161,9 @@ static const struct kfd2kgd_calls kfd2kgd = { .get_fw_version = get_fw_version, .set_scratch_backing_va = set_scratch_backing_va, .get_tile_config = amdgpu_amdkfd_get_tile_config, - .get_cu_info = get_cu_info, - .get_vram_usage = amdgpu_amdkfd_get_vram_usage, - .create_process_vm = amdgpu_amdkfd_gpuvm_create_process_vm, - .acquire_process_vm = amdgpu_amdkfd_gpuvm_acquire_process_vm, - .destroy_process_vm = amdgpu_amdkfd_gpuvm_destroy_process_vm, - .release_process_vm = amdgpu_amdkfd_gpuvm_release_process_vm, - .get_process_page_dir = amdgpu_amdkfd_gpuvm_get_process_page_dir, .set_vm_context_page_table_base = set_vm_context_page_table_base, - .alloc_memory_of_gpu = amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu, - .free_memory_of_gpu = amdgpu_amdkfd_gpuvm_free_memory_of_gpu, - .map_memory_to_gpu = amdgpu_amdkfd_gpuvm_map_memory_to_gpu, - .unmap_memory_to_gpu = amdgpu_amdkfd_gpuvm_unmap_memory_from_gpu, - .sync_memory = amdgpu_amdkfd_gpuvm_sync_memory, - .map_gtt_bo_to_kernel = amdgpu_amdkfd_gpuvm_map_gtt_bo_to_kernel, - .restore_process_bos = amdgpu_amdkfd_gpuvm_restore_process_bos, .invalidate_tlbs = invalidate_tlbs, .invalidate_tlbs_vmid = invalidate_tlbs_vmid, - .submit_ib = amdgpu_amdkfd_submit_ib, - .gpu_recover = amdgpu_amdkfd_gpu_reset, - .set_compute_idle = amdgpu_amdkfd_set_compute_idle, .get_hive_id = amdgpu_amdkfd_get_hive_id, }; @@ -785,15 +733,6 @@ static uint16_t get_atc_vmid_pasid_mapping_pasid(struct kgd_dev *kgd, static void write_vmid_invalidate_request(struct kgd_dev *kgd, uint8_t vmid) { struct amdgpu_device *adev = (struct amdgpu_device *) kgd; - uint32_t req = (1 << vmid) | - (0 << VM_INVALIDATE_ENG16_REQ__FLUSH_TYPE__SHIFT) | /* legacy */ - VM_INVALIDATE_ENG16_REQ__INVALIDATE_L2_PTES_MASK | - VM_INVALIDATE_ENG16_REQ__INVALIDATE_L2_PDE0_MASK | - VM_INVALIDATE_ENG16_REQ__INVALIDATE_L2_PDE1_MASK | - VM_INVALIDATE_ENG16_REQ__INVALIDATE_L2_PDE2_MASK | - VM_INVALIDATE_ENG16_REQ__INVALIDATE_L1_PTES_MASK; - - mutex_lock(&adev->srbm_mutex); /* Use legacy mode tlb invalidation. * @@ -810,34 +749,7 @@ static void write_vmid_invalidate_request(struct kgd_dev *kgd, uint8_t vmid) * TODO 2: support range-based invalidation, requires kfg2kgd * interface change */ - WREG32(SOC15_REG_OFFSET(GC, 0, mmVM_INVALIDATE_ENG16_ADDR_RANGE_LO32), - 0xffffffff); - WREG32(SOC15_REG_OFFSET(GC, 0, mmVM_INVALIDATE_ENG16_ADDR_RANGE_HI32), - 0x0000001f); - - WREG32(SOC15_REG_OFFSET(MMHUB, 0, - mmMMHUB_VM_INVALIDATE_ENG16_ADDR_RANGE_LO32), - 0xffffffff); - WREG32(SOC15_REG_OFFSET(MMHUB, 0, - mmMMHUB_VM_INVALIDATE_ENG16_ADDR_RANGE_HI32), - 0x0000001f); - - WREG32(SOC15_REG_OFFSET(GC, 0, mmVM_INVALIDATE_ENG16_REQ), req); - - WREG32(SOC15_REG_OFFSET(MMHUB, 0, mmMMHUB_VM_INVALIDATE_ENG16_REQ), - req); - - while (!(RREG32(SOC15_REG_OFFSET(GC, 0, mmVM_INVALIDATE_ENG16_ACK)) & - (1 << vmid))) - cpu_relax(); - - while (!(RREG32(SOC15_REG_OFFSET(MMHUB, 0, - mmMMHUB_VM_INVALIDATE_ENG16_ACK)) & - (1 << vmid))) - cpu_relax(); - - mutex_unlock(&adev->srbm_mutex); - + amdgpu_gmc_flush_gpu_tlb(adev, vmid, 0); } static int invalidate_tlbs_with_kiq(struct amdgpu_device *adev, uint16_t pasid) @@ -876,7 +788,7 @@ static int invalidate_tlbs(struct kgd_dev *kgd, uint16_t pasid) if (adev->in_gpu_reset) return -EIO; - if (ring->ready) + if (ring->sched.ready) return invalidate_tlbs_with_kiq(adev, pasid); for (vmid = 0; vmid < 16; vmid++) { @@ -1016,7 +928,6 @@ static void set_vm_context_page_table_base(struct kgd_dev *kgd, uint32_t vmid, uint64_t page_table_base) { struct amdgpu_device *adev = get_amdgpu_device(kgd); - uint64_t base = page_table_base | AMDGPU_PTE_VALID; if (!amdgpu_amdkfd_is_kfd_vmid(adev, vmid)) { pr_err("trying to set page table base for wrong VMID %u\n", @@ -1028,25 +939,7 @@ static void set_vm_context_page_table_base(struct kgd_dev *kgd, uint32_t vmid, * now, all processes share the same address space size, like * on GFX8 and older. */ - WREG32(SOC15_REG_OFFSET(MMHUB, 0, mmMMHUB_VM_CONTEXT0_PAGE_TABLE_START_ADDR_LO32) + (vmid*2), 0); - WREG32(SOC15_REG_OFFSET(MMHUB, 0, mmMMHUB_VM_CONTEXT0_PAGE_TABLE_START_ADDR_HI32) + (vmid*2), 0); - - WREG32(SOC15_REG_OFFSET(MMHUB, 0, mmMMHUB_VM_CONTEXT0_PAGE_TABLE_END_ADDR_LO32) + (vmid*2), - lower_32_bits(adev->vm_manager.max_pfn - 1)); - WREG32(SOC15_REG_OFFSET(MMHUB, 0, mmMMHUB_VM_CONTEXT0_PAGE_TABLE_END_ADDR_HI32) + (vmid*2), - upper_32_bits(adev->vm_manager.max_pfn - 1)); - - WREG32(SOC15_REG_OFFSET(MMHUB, 0, mmMMHUB_VM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32) + (vmid*2), lower_32_bits(base)); - WREG32(SOC15_REG_OFFSET(MMHUB, 0, mmMMHUB_VM_CONTEXT0_PAGE_TABLE_BASE_ADDR_HI32) + (vmid*2), upper_32_bits(base)); - - WREG32(SOC15_REG_OFFSET(GC, 0, mmVM_CONTEXT0_PAGE_TABLE_START_ADDR_LO32) + (vmid*2), 0); - WREG32(SOC15_REG_OFFSET(GC, 0, mmVM_CONTEXT0_PAGE_TABLE_START_ADDR_HI32) + (vmid*2), 0); - - WREG32(SOC15_REG_OFFSET(GC, 0, mmVM_CONTEXT0_PAGE_TABLE_END_ADDR_LO32) + (vmid*2), - lower_32_bits(adev->vm_manager.max_pfn - 1)); - WREG32(SOC15_REG_OFFSET(GC, 0, mmVM_CONTEXT0_PAGE_TABLE_END_ADDR_HI32) + (vmid*2), - upper_32_bits(adev->vm_manager.max_pfn - 1)); + mmhub_v1_0_setup_vm_pt_regs(adev, vmid, page_table_base); - WREG32(SOC15_REG_OFFSET(GC, 0, mmVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32) + (vmid*2), lower_32_bits(base)); - WREG32(SOC15_REG_OFFSET(GC, 0, mmVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_HI32) + (vmid*2), upper_32_bits(base)); + gfxhub_v1_0_setup_vm_pt_regs(adev, vmid, page_table_base); } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c index df0a059565f9..1921dec3df7a 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c @@ -25,6 +25,7 @@ #include <linux/list.h> #include <linux/pagemap.h> #include <linux/sched/mm.h> +#include <linux/dma-buf.h> #include <drm/drmP.h> #include "amdgpu_object.h" #include "amdgpu_vm.h" @@ -46,9 +47,9 @@ /* Impose limit on how much memory KFD can use */ static struct { uint64_t max_system_mem_limit; - uint64_t max_userptr_mem_limit; + uint64_t max_ttm_mem_limit; int64_t system_mem_used; - int64_t userptr_mem_used; + int64_t ttm_mem_used; spinlock_t mem_limit_lock; } kfd_mem_limit; @@ -90,8 +91,8 @@ static bool check_if_add_bo_to_vm(struct amdgpu_vm *avm, } /* Set memory usage limits. Current, limits are - * System (kernel) memory - 3/8th System RAM - * Userptr memory - 3/4th System RAM + * System (TTM + userptr) memory - 3/4th System RAM + * TTM memory - 3/8th System RAM */ void amdgpu_amdkfd_gpuvm_init_mem_limits(void) { @@ -103,48 +104,61 @@ void amdgpu_amdkfd_gpuvm_init_mem_limits(void) mem *= si.mem_unit; spin_lock_init(&kfd_mem_limit.mem_limit_lock); - kfd_mem_limit.max_system_mem_limit = (mem >> 1) - (mem >> 3); - kfd_mem_limit.max_userptr_mem_limit = mem - (mem >> 2); - pr_debug("Kernel memory limit %lluM, userptr limit %lluM\n", + kfd_mem_limit.max_system_mem_limit = (mem >> 1) + (mem >> 2); + kfd_mem_limit.max_ttm_mem_limit = (mem >> 1) - (mem >> 3); + pr_debug("Kernel memory limit %lluM, TTM limit %lluM\n", (kfd_mem_limit.max_system_mem_limit >> 20), - (kfd_mem_limit.max_userptr_mem_limit >> 20)); + (kfd_mem_limit.max_ttm_mem_limit >> 20)); } -static int amdgpu_amdkfd_reserve_system_mem_limit(struct amdgpu_device *adev, - uint64_t size, u32 domain) +static int amdgpu_amdkfd_reserve_mem_limit(struct amdgpu_device *adev, + uint64_t size, u32 domain, bool sg) { - size_t acc_size; + size_t acc_size, system_mem_needed, ttm_mem_needed, vram_needed; + uint64_t reserved_for_pt = amdgpu_amdkfd_total_mem_size >> 9; int ret = 0; acc_size = ttm_bo_dma_acc_size(&adev->mman.bdev, size, sizeof(struct amdgpu_bo)); - spin_lock(&kfd_mem_limit.mem_limit_lock); + vram_needed = 0; if (domain == AMDGPU_GEM_DOMAIN_GTT) { - if (kfd_mem_limit.system_mem_used + (acc_size + size) > - kfd_mem_limit.max_system_mem_limit) { - ret = -ENOMEM; - goto err_no_mem; - } - kfd_mem_limit.system_mem_used += (acc_size + size); - } else if (domain == AMDGPU_GEM_DOMAIN_CPU) { - if ((kfd_mem_limit.system_mem_used + acc_size > - kfd_mem_limit.max_system_mem_limit) || - (kfd_mem_limit.userptr_mem_used + (size + acc_size) > - kfd_mem_limit.max_userptr_mem_limit)) { - ret = -ENOMEM; - goto err_no_mem; - } - kfd_mem_limit.system_mem_used += acc_size; - kfd_mem_limit.userptr_mem_used += size; + /* TTM GTT memory */ + system_mem_needed = acc_size + size; + ttm_mem_needed = acc_size + size; + } else if (domain == AMDGPU_GEM_DOMAIN_CPU && !sg) { + /* Userptr */ + system_mem_needed = acc_size + size; + ttm_mem_needed = acc_size; + } else { + /* VRAM and SG */ + system_mem_needed = acc_size; + ttm_mem_needed = acc_size; + if (domain == AMDGPU_GEM_DOMAIN_VRAM) + vram_needed = size; + } + + spin_lock(&kfd_mem_limit.mem_limit_lock); + + if ((kfd_mem_limit.system_mem_used + system_mem_needed > + kfd_mem_limit.max_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)) { + ret = -ENOMEM; + } else { + kfd_mem_limit.system_mem_used += system_mem_needed; + kfd_mem_limit.ttm_mem_used += ttm_mem_needed; + adev->kfd.vram_used += vram_needed; } -err_no_mem: + spin_unlock(&kfd_mem_limit.mem_limit_lock); return ret; } -static void unreserve_system_mem_limit(struct amdgpu_device *adev, - uint64_t size, u32 domain) +static void unreserve_mem_limit(struct amdgpu_device *adev, + uint64_t size, u32 domain, bool sg) { size_t acc_size; @@ -154,70 +168,61 @@ static void unreserve_system_mem_limit(struct amdgpu_device *adev, spin_lock(&kfd_mem_limit.mem_limit_lock); if (domain == AMDGPU_GEM_DOMAIN_GTT) { kfd_mem_limit.system_mem_used -= (acc_size + size); - } else if (domain == AMDGPU_GEM_DOMAIN_CPU) { + kfd_mem_limit.ttm_mem_used -= (acc_size + size); + } else if (domain == AMDGPU_GEM_DOMAIN_CPU && !sg) { + kfd_mem_limit.system_mem_used -= (acc_size + size); + kfd_mem_limit.ttm_mem_used -= acc_size; + } else { kfd_mem_limit.system_mem_used -= acc_size; - kfd_mem_limit.userptr_mem_used -= size; + kfd_mem_limit.ttm_mem_used -= acc_size; + if (domain == AMDGPU_GEM_DOMAIN_VRAM) { + adev->kfd.vram_used -= size; + WARN_ONCE(adev->kfd.vram_used < 0, + "kfd VRAM memory accounting unbalanced"); + } } WARN_ONCE(kfd_mem_limit.system_mem_used < 0, "kfd system memory accounting unbalanced"); - WARN_ONCE(kfd_mem_limit.userptr_mem_used < 0, - "kfd userptr memory accounting unbalanced"); + WARN_ONCE(kfd_mem_limit.ttm_mem_used < 0, + "kfd TTM memory accounting unbalanced"); spin_unlock(&kfd_mem_limit.mem_limit_lock); } -void amdgpu_amdkfd_unreserve_system_memory_limit(struct amdgpu_bo *bo) +void amdgpu_amdkfd_unreserve_memory_limit(struct amdgpu_bo *bo) { - spin_lock(&kfd_mem_limit.mem_limit_lock); + struct amdgpu_device *adev = amdgpu_ttm_adev(bo->tbo.bdev); + u32 domain = bo->preferred_domains; + bool sg = (bo->preferred_domains == AMDGPU_GEM_DOMAIN_CPU); if (bo->flags & AMDGPU_AMDKFD_USERPTR_BO) { - kfd_mem_limit.system_mem_used -= bo->tbo.acc_size; - kfd_mem_limit.userptr_mem_used -= amdgpu_bo_size(bo); - } else if (bo->preferred_domains == AMDGPU_GEM_DOMAIN_GTT) { - kfd_mem_limit.system_mem_used -= - (bo->tbo.acc_size + amdgpu_bo_size(bo)); + domain = AMDGPU_GEM_DOMAIN_CPU; + sg = false; } - WARN_ONCE(kfd_mem_limit.system_mem_used < 0, - "kfd system memory accounting unbalanced"); - WARN_ONCE(kfd_mem_limit.userptr_mem_used < 0, - "kfd userptr memory accounting unbalanced"); - spin_unlock(&kfd_mem_limit.mem_limit_lock); + unreserve_mem_limit(adev, amdgpu_bo_size(bo), domain, sg); } -/* amdgpu_amdkfd_remove_eviction_fence - Removes eviction fence(s) from BO's +/* amdgpu_amdkfd_remove_eviction_fence - Removes eviction fence from BO's * reservation object. * * @bo: [IN] Remove eviction fence(s) from this BO - * @ef: [IN] If ef is specified, then this eviction fence is removed if it + * @ef: [IN] This eviction fence is removed if it * is present in the shared list. - * @ef_list: [OUT] Returns list of eviction fences. These fences are removed - * from BO's reservation object shared list. - * @ef_count: [OUT] Number of fences in ef_list. * - * NOTE: If called with ef_list, then amdgpu_amdkfd_add_eviction_fence must be - * called to restore the eviction fences and to avoid memory leak. This is - * useful for shared BOs. * NOTE: Must be called with BO reserved i.e. bo->tbo.resv->lock held. */ static int amdgpu_amdkfd_remove_eviction_fence(struct amdgpu_bo *bo, - struct amdgpu_amdkfd_fence *ef, - struct amdgpu_amdkfd_fence ***ef_list, - unsigned int *ef_count) + struct amdgpu_amdkfd_fence *ef) { struct reservation_object *resv = bo->tbo.resv; struct reservation_object_list *old, *new; unsigned int i, j, k; - if (!ef && !ef_list) + if (!ef) return -EINVAL; - if (ef_list) { - *ef_list = NULL; - *ef_count = 0; - } - old = reservation_object_get_list(resv); if (!old) return 0; @@ -236,8 +241,7 @@ static int amdgpu_amdkfd_remove_eviction_fence(struct amdgpu_bo *bo, f = rcu_dereference_protected(old->shared[i], reservation_object_held(resv)); - if ((ef && f->context == ef->base.context) || - (!ef && to_amdgpu_amdkfd_fence(f))) + if (f->context == ef->base.context) RCU_INIT_POINTER(new->shared[--j], f); else RCU_INIT_POINTER(new->shared[k++], f); @@ -245,21 +249,6 @@ static int amdgpu_amdkfd_remove_eviction_fence(struct amdgpu_bo *bo, new->shared_max = old->shared_max; new->shared_count = k; - if (!ef) { - unsigned int count = old->shared_count - j; - - /* Alloc memory for count number of eviction fence pointers. - * Fill the ef_list array and ef_count - */ - *ef_list = kcalloc(count, sizeof(**ef_list), GFP_KERNEL); - *ef_count = count; - - if (!*ef_list) { - kfree(new); - return -ENOMEM; - } - } - /* Install the new fence list, seqcount provides the barriers */ preempt_disable(); write_seqcount_begin(&resv->seq); @@ -273,46 +262,13 @@ static int amdgpu_amdkfd_remove_eviction_fence(struct amdgpu_bo *bo, f = rcu_dereference_protected(new->shared[i], reservation_object_held(resv)); - if (!ef) - (*ef_list)[k++] = to_amdgpu_amdkfd_fence(f); - else - dma_fence_put(f); + dma_fence_put(f); } kfree_rcu(old, rcu); return 0; } -/* amdgpu_amdkfd_add_eviction_fence - Adds eviction fence(s) back into BO's - * reservation object. - * - * @bo: [IN] Add eviction fences to this BO - * @ef_list: [IN] List of eviction fences to be added - * @ef_count: [IN] Number of fences in ef_list. - * - * NOTE: Must call amdgpu_amdkfd_remove_eviction_fence before calling this - * function. - */ -static void amdgpu_amdkfd_add_eviction_fence(struct amdgpu_bo *bo, - struct amdgpu_amdkfd_fence **ef_list, - unsigned int ef_count) -{ - int i; - - if (!ef_list || !ef_count) - return; - - for (i = 0; i < ef_count; i++) { - amdgpu_bo_fence(bo, &ef_list[i]->base, true); - /* Re-adding the fence takes an additional reference. Drop that - * reference. - */ - dma_fence_put(&ef_list[i]->base); - } - - kfree(ef_list); -} - static int amdgpu_amdkfd_bo_validate(struct amdgpu_bo *bo, uint32_t domain, bool wait) { @@ -328,18 +284,8 @@ static int amdgpu_amdkfd_bo_validate(struct amdgpu_bo *bo, uint32_t domain, ret = ttm_bo_validate(&bo->tbo, &bo->placement, &ctx); if (ret) goto validate_fail; - if (wait) { - struct amdgpu_amdkfd_fence **ef_list; - unsigned int ef_count; - - ret = amdgpu_amdkfd_remove_eviction_fence(bo, NULL, &ef_list, - &ef_count); - if (ret) - goto validate_fail; - - ttm_bo_wait(&bo->tbo, false, false); - amdgpu_amdkfd_add_eviction_fence(bo, ef_list, ef_count); - } + if (wait) + amdgpu_bo_sync_wait(bo, AMDGPU_FENCE_OWNER_KFD, false); validate_fail: return ret; @@ -395,23 +341,6 @@ static int vm_validate_pt_pd_bos(struct amdgpu_vm *vm) return 0; } -static int sync_vm_fence(struct amdgpu_device *adev, struct amdgpu_sync *sync, - struct dma_fence *f) -{ - int ret = amdgpu_sync_fence(adev, sync, f, false); - - /* Sync objects can't handle multiple GPUs (contexts) updating - * sync->last_vm_update. Fortunately we don't need it for - * KFD's purposes, so we can just drop that fence. - */ - if (sync->last_vm_update) { - dma_fence_put(sync->last_vm_update); - sync->last_vm_update = NULL; - } - - return ret; -} - static int vm_update_pds(struct amdgpu_vm *vm, struct amdgpu_sync *sync) { struct amdgpu_bo *pd = vm->root.base.bo; @@ -422,7 +351,7 @@ static int vm_update_pds(struct amdgpu_vm *vm, struct amdgpu_sync *sync) if (ret) return ret; - return sync_vm_fence(adev, sync, vm->last_update); + return amdgpu_sync_fence(NULL, sync, vm->last_update, false); } /* add_bo_to_vm - Add a BO to a VM @@ -443,7 +372,6 @@ static int add_bo_to_vm(struct amdgpu_device *adev, struct kgd_mem *mem, { int ret; struct kfd_bo_va_list *bo_va_entry; - struct amdgpu_bo *pd = vm->root.base.bo; struct amdgpu_bo *bo = mem->bo; uint64_t va = mem->va; struct list_head *list_bo_va = &mem->bo_va_list; @@ -483,14 +411,8 @@ static int add_bo_to_vm(struct amdgpu_device *adev, struct kgd_mem *mem, *p_bo_va_entry = bo_va_entry; /* Allocate new page tables if needed and validate - * them. Clearing of new page tables and validate need to wait - * on move fences. We don't want that to trigger the eviction - * fence, so remove it temporarily. + * them. */ - amdgpu_amdkfd_remove_eviction_fence(pd, - vm->process_info->eviction_fence, - NULL, NULL); - ret = amdgpu_vm_alloc_pts(adev, vm, va, amdgpu_bo_size(bo)); if (ret) { pr_err("Failed to allocate pts, err=%d\n", ret); @@ -503,13 +425,9 @@ static int add_bo_to_vm(struct amdgpu_device *adev, struct kgd_mem *mem, goto err_alloc_pts; } - /* Add the eviction fence back */ - amdgpu_bo_fence(pd, &vm->process_info->eviction_fence->base, true); - return 0; err_alloc_pts: - amdgpu_bo_fence(pd, &vm->process_info->eviction_fence->base, true); amdgpu_vm_bo_rmv(adev, bo_va_entry->bo_va); list_del(&bo_va_entry->bo_list); err_vmadd: @@ -536,7 +454,7 @@ static void add_kgd_mem_to_kfd_bo_list(struct kgd_mem *mem, struct amdgpu_bo *bo = mem->bo; INIT_LIST_HEAD(&entry->head); - entry->shared = true; + entry->num_shared = 1; entry->bo = &bo->tbo; mutex_lock(&process_info->lock); if (userptr) @@ -677,7 +595,7 @@ static int reserve_bo_and_vm(struct kgd_mem *mem, ctx->kfd_bo.priority = 0; ctx->kfd_bo.tv.bo = &bo->tbo; - ctx->kfd_bo.tv.shared = true; + ctx->kfd_bo.tv.num_shared = 1; ctx->kfd_bo.user_pages = NULL; list_add(&ctx->kfd_bo.tv.head, &ctx->list); @@ -741,7 +659,7 @@ static int reserve_bo_and_cond_vms(struct kgd_mem *mem, ctx->kfd_bo.priority = 0; ctx->kfd_bo.tv.bo = &bo->tbo; - ctx->kfd_bo.tv.shared = true; + ctx->kfd_bo.tv.num_shared = 1; ctx->kfd_bo.user_pages = NULL; list_add(&ctx->kfd_bo.tv.head, &ctx->list); @@ -808,25 +726,12 @@ static int unmap_bo_from_gpuvm(struct amdgpu_device *adev, { struct amdgpu_bo_va *bo_va = entry->bo_va; struct amdgpu_vm *vm = bo_va->base.vm; - struct amdgpu_bo *pd = vm->root.base.bo; - /* Remove eviction fence from PD (and thereby from PTs too as - * they share the resv. object). Otherwise during PT update - * job (see amdgpu_vm_bo_update_mapping), eviction fence would - * get added to job->sync object and job execution would - * trigger the eviction fence. - */ - amdgpu_amdkfd_remove_eviction_fence(pd, - vm->process_info->eviction_fence, - NULL, NULL); amdgpu_vm_bo_unmap(adev, bo_va, entry->va); amdgpu_vm_clear_freed(adev, vm, &bo_va->last_pt_update); - /* Add the eviction fence back */ - amdgpu_bo_fence(pd, &vm->process_info->eviction_fence->base, true); - - sync_vm_fence(adev, sync, bo_va->last_pt_update); + amdgpu_sync_fence(NULL, sync, bo_va->last_pt_update, false); return 0; } @@ -851,7 +756,7 @@ static int update_gpuvm_pte(struct amdgpu_device *adev, return ret; } - return sync_vm_fence(adev, sync, bo_va->last_pt_update); + return amdgpu_sync_fence(NULL, sync, bo_va->last_pt_update, false); } static int map_bo_to_gpuvm(struct amdgpu_device *adev, @@ -886,6 +791,24 @@ 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; @@ -901,6 +824,26 @@ static int process_validate_vms(struct amdkfd_process_info *process_info) return 0; } +static int process_sync_pds_resv(struct amdkfd_process_info *process_info, + struct amdgpu_sync *sync) +{ + struct amdgpu_vm *peer_vm; + int ret; + + list_for_each_entry(peer_vm, &process_info->vm_list_head, + vm_list_node) { + struct amdgpu_bo *pd = peer_vm->root.base.bo; + + ret = amdgpu_sync_resv(NULL, + sync, pd->tbo.resv, + AMDGPU_FENCE_OWNER_UNDEFINED, false); + if (ret) + return ret; + } + + return 0; +} + static int process_update_pds(struct amdkfd_process_info *process_info, struct amdgpu_sync *sync) { @@ -963,7 +906,7 @@ static int init_kfd_vm(struct amdgpu_vm *vm, void **process_info, pr_err("validate_pt_pd_bos() failed\n"); goto validate_pd_fail; } - ret = ttm_bo_wait(&vm->root.base.bo->tbo, false, false); + amdgpu_bo_sync_wait(vm->root.base.bo, AMDGPU_FENCE_OWNER_KFD, false); if (ret) goto wait_pd_fail; amdgpu_bo_fence(vm->root.base.bo, @@ -1149,6 +1092,8 @@ int amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu( { struct amdgpu_device *adev = get_amdgpu_device(kgd); struct amdgpu_vm *avm = (struct amdgpu_vm *)vm; + enum ttm_bo_type bo_type = ttm_bo_type_device; + struct sg_table *sg = NULL; uint64_t user_addr = 0; struct amdgpu_bo *bo; struct amdgpu_bo_param bp; @@ -1177,13 +1122,25 @@ int amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu( if (!offset || !*offset) return -EINVAL; user_addr = *offset; + } else if (flags & ALLOC_MEM_FLAGS_DOORBELL) { + domain = AMDGPU_GEM_DOMAIN_GTT; + alloc_domain = AMDGPU_GEM_DOMAIN_CPU; + bo_type = ttm_bo_type_sg; + alloc_flags = 0; + if (size > UINT_MAX) + return -EINVAL; + sg = create_doorbell_sg(*offset, size); + if (!sg) + return -ENOMEM; } else { return -EINVAL; } *mem = kzalloc(sizeof(struct kgd_mem), GFP_KERNEL); - if (!*mem) - return -ENOMEM; + if (!*mem) { + ret = -ENOMEM; + goto err; + } INIT_LIST_HEAD(&(*mem)->bo_va_list); mutex_init(&(*mem)->lock); (*mem)->aql_queue = !!(flags & ALLOC_MEM_FLAGS_AQL_QUEUE_MEM); @@ -1199,7 +1156,8 @@ int amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu( byte_align = (adev->family == AMDGPU_FAMILY_VI && adev->asic_type != CHIP_FIJI && adev->asic_type != CHIP_POLARIS10 && - adev->asic_type != CHIP_POLARIS11) ? + adev->asic_type != CHIP_POLARIS11 && + adev->asic_type != CHIP_POLARIS12) ? VI_BO_SIZE_ALIGN : 1; mapping_flags = AMDGPU_VM_PAGE_READABLE; @@ -1215,10 +1173,10 @@ int amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu( amdgpu_sync_create(&(*mem)->sync); - ret = amdgpu_amdkfd_reserve_system_mem_limit(adev, size, alloc_domain); + ret = amdgpu_amdkfd_reserve_mem_limit(adev, size, alloc_domain, !!sg); if (ret) { pr_debug("Insufficient system memory\n"); - goto err_reserve_system_mem; + goto err_reserve_limit; } pr_debug("\tcreate BO VA 0x%llx size 0x%llx domain %s\n", @@ -1229,7 +1187,7 @@ int amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu( bp.byte_align = byte_align; bp.domain = alloc_domain; bp.flags = alloc_flags; - bp.type = ttm_bo_type_device; + bp.type = bo_type; bp.resv = NULL; ret = amdgpu_bo_create(adev, &bp, &bo); if (ret) { @@ -1237,6 +1195,10 @@ int amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu( domain_string(alloc_domain), ret); goto err_bo_create; } + if (bo_type == ttm_bo_type_sg) { + bo->tbo.sg = sg; + bo->tbo.ttm->sg = sg; + } bo->kfd_bo = *mem; (*mem)->bo = bo; if (user_addr) @@ -1266,12 +1228,17 @@ int amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu( allocate_init_user_pages_failed: amdgpu_bo_unref(&bo); /* Don't unreserve system mem limit twice */ - goto err_reserve_system_mem; + goto err_reserve_limit; err_bo_create: - unreserve_system_mem_limit(adev, size, alloc_domain); -err_reserve_system_mem: + unreserve_mem_limit(adev, size, alloc_domain, !!sg); +err_reserve_limit: mutex_destroy(&(*mem)->lock); kfree(*mem); +err: + if (sg) { + sg_free_table(sg); + kfree(sg); + } return ret; } @@ -1326,8 +1293,7 @@ int amdgpu_amdkfd_gpuvm_free_memory_of_gpu( * attached */ amdgpu_amdkfd_remove_eviction_fence(mem->bo, - process_info->eviction_fence, - NULL, NULL); + process_info->eviction_fence); pr_debug("Release VA 0x%llx - 0x%llx\n", mem->va, mem->va + bo_size * (1 + mem->aql_queue)); @@ -1341,6 +1307,14 @@ int amdgpu_amdkfd_gpuvm_free_memory_of_gpu( /* Free the sync object */ amdgpu_sync_free(&mem->sync); + /* If the SG is not NULL, it's one we created for a doorbell + * BO. We need to free it. + */ + if (mem->bo->tbo.sg) { + sg_free_table(mem->bo->tbo.sg); + kfree(mem->bo->tbo.sg); + } + /* Free the BO*/ amdgpu_bo_unref(&mem->bo); mutex_destroy(&mem->lock); @@ -1405,7 +1379,8 @@ int amdgpu_amdkfd_gpuvm_map_memory_to_gpu( * the queues are still stopped and we can leave mapping for * the next restore worker */ - if (bo->tbo.mem.mem_type == TTM_PL_SYSTEM) + if (amdgpu_ttm_tt_get_usermm(bo->tbo.ttm) && + bo->tbo.mem.mem_type == TTM_PL_SYSTEM) is_invalid_userptr = true; if (check_if_add_bo_to_vm(avm, mem)) { @@ -1545,8 +1520,7 @@ int amdgpu_amdkfd_gpuvm_unmap_memory_from_gpu( if (mem->mapped_to_gpu_memory == 0 && !amdgpu_ttm_tt_get_usermm(mem->bo->tbo.ttm) && !mem->bo->pin_count) amdgpu_amdkfd_remove_eviction_fence(mem->bo, - process_info->eviction_fence, - NULL, NULL); + process_info->eviction_fence); unreserve_out: unreserve_bo_and_vms(&ctx, false, false); @@ -1607,7 +1581,7 @@ int amdgpu_amdkfd_gpuvm_map_gtt_bo_to_kernel(struct kgd_dev *kgd, } amdgpu_amdkfd_remove_eviction_fence( - bo, mem->process_info->eviction_fence, NULL, NULL); + bo, mem->process_info->eviction_fence); list_del_init(&mem->validate_list.head); if (size) @@ -1642,6 +1616,60 @@ int amdgpu_amdkfd_gpuvm_get_vm_fault_info(struct kgd_dev *kgd, return 0; } +int amdgpu_amdkfd_gpuvm_import_dmabuf(struct kgd_dev *kgd, + struct dma_buf *dma_buf, + uint64_t va, void *vm, + struct kgd_mem **mem, uint64_t *size, + uint64_t *mmap_offset) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)kgd; + struct drm_gem_object *obj; + struct amdgpu_bo *bo; + struct amdgpu_vm *avm = (struct amdgpu_vm *)vm; + + if (dma_buf->ops != &amdgpu_dmabuf_ops) + /* Can't handle non-graphics buffers */ + return -EINVAL; + + obj = dma_buf->priv; + if (obj->dev->dev_private != adev) + /* Can't handle buffers from other devices */ + return -EINVAL; + + bo = gem_to_amdgpu_bo(obj); + if (!(bo->preferred_domains & (AMDGPU_GEM_DOMAIN_VRAM | + AMDGPU_GEM_DOMAIN_GTT))) + /* Only VRAM and GTT BOs are supported */ + return -EINVAL; + + *mem = kzalloc(sizeof(struct kgd_mem), GFP_KERNEL); + if (!*mem) + return -ENOMEM; + + if (size) + *size = amdgpu_bo_size(bo); + + if (mmap_offset) + *mmap_offset = amdgpu_bo_mmap_offset(bo); + + INIT_LIST_HEAD(&(*mem)->bo_va_list); + mutex_init(&(*mem)->lock); + (*mem)->mapping_flags = + AMDGPU_VM_PAGE_READABLE | AMDGPU_VM_PAGE_WRITEABLE | + AMDGPU_VM_PAGE_EXECUTABLE | AMDGPU_VM_MTYPE_NC; + + (*mem)->bo = amdgpu_bo_ref(bo); + (*mem)->va = va; + (*mem)->domain = (bo->preferred_domains & AMDGPU_GEM_DOMAIN_VRAM) ? + AMDGPU_GEM_DOMAIN_VRAM : AMDGPU_GEM_DOMAIN_GTT; + (*mem)->mapped_to_gpu_memory = 0; + (*mem)->process_info = avm->process_info; + add_kgd_mem_to_kfd_bo_list(*mem, avm->process_info, false); + amdgpu_sync_create(&(*mem)->sync); + + return 0; +} + /* Evict a userptr BO by stopping the queues if necessary * * Runs in MMU notifier, may be in RECLAIM_FS context. This means it @@ -1664,7 +1692,7 @@ int amdgpu_amdkfd_evict_userptr(struct kgd_mem *mem, 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); if (r) pr_err("Failed to quiesce KFD\n"); schedule_delayed_work(&process_info->restore_userptr_work, @@ -1808,7 +1836,7 @@ static int validate_invalid_user_pages(struct amdkfd_process_info *process_info) validate_list.head) { list_add_tail(&mem->resv_list.head, &resv_list); mem->resv_list.bo = mem->validate_list.bo; - mem->resv_list.shared = mem->validate_list.shared; + mem->resv_list.num_shared = mem->validate_list.num_shared; } /* Reserve all BOs and page tables for validation */ @@ -1819,16 +1847,6 @@ static int validate_invalid_user_pages(struct amdkfd_process_info *process_info) amdgpu_sync_create(&sync); - /* Avoid triggering eviction fences when unmapping invalid - * userptr BOs (waits for all fences, doesn't use - * FENCE_OWNER_VM) - */ - list_for_each_entry(peer_vm, &process_info->vm_list_head, - vm_list_node) - amdgpu_amdkfd_remove_eviction_fence(peer_vm->root.base.bo, - process_info->eviction_fence, - NULL, NULL); - ret = process_validate_vms(process_info); if (ret) goto unreserve_out; @@ -1889,10 +1907,6 @@ static int validate_invalid_user_pages(struct amdkfd_process_info *process_info) ret = process_update_pds(process_info, &sync); unreserve_out: - list_for_each_entry(peer_vm, &process_info->vm_list_head, - vm_list_node) - amdgpu_bo_fence(peer_vm->root.base.bo, - &process_info->eviction_fence->base, true); ttm_eu_backoff_reservation(&ticket, &resv_list); amdgpu_sync_wait(&sync, false); amdgpu_sync_free(&sync); @@ -1956,7 +1970,7 @@ static void amdgpu_amdkfd_restore_userptr_worker(struct work_struct *work) evicted_bos) goto unlock_out; evicted_bos = 0; - if (kgd2kfd->resume_mm(mm)) { + if (kgd2kfd_resume_mm(mm)) { pr_err("%s: Failed to resume KFD\n", __func__); /* No recovery from this failure. Probably the CP is * hanging. No point trying again. @@ -2027,7 +2041,7 @@ int amdgpu_amdkfd_gpuvm_restore_process_bos(void *info, struct dma_fence **ef) list_add_tail(&mem->resv_list.head, &ctx.list); mem->resv_list.bo = mem->validate_list.bo; - mem->resv_list.shared = mem->validate_list.shared; + mem->resv_list.num_shared = mem->validate_list.num_shared; } ret = ttm_eu_reserve_buffers(&ctx.ticket, &ctx.list, @@ -2044,13 +2058,10 @@ int amdgpu_amdkfd_gpuvm_restore_process_bos(void *info, struct dma_fence **ef) if (ret) goto validate_map_fail; - /* Wait for PD/PTs validate to finish */ - /* FIXME: I think this isn't needed */ - list_for_each_entry(peer_vm, &process_info->vm_list_head, - vm_list_node) { - struct amdgpu_bo *bo = peer_vm->root.base.bo; - - ttm_bo_wait(&bo->tbo, false, false); + ret = process_sync_pds_resv(process_info, &sync_obj); + if (ret) { + pr_debug("Memory eviction: Failed to sync to PD BO moving fence. Try again\n"); + goto validate_map_fail; } /* Validate BOs and map them to GPUVM (update VM page tables). */ @@ -2066,7 +2077,11 @@ int amdgpu_amdkfd_gpuvm_restore_process_bos(void *info, struct dma_fence **ef) pr_debug("Memory eviction: Validate BOs failed. Try again\n"); goto validate_map_fail; } - + ret = amdgpu_sync_fence(NULL, &sync_obj, bo->tbo.moving, false); + if (ret) { + pr_debug("Memory eviction: Sync BO fence failed. Try again\n"); + goto validate_map_fail; + } list_for_each_entry(bo_va_entry, &mem->bo_va_list, bo_list) { ret = update_gpuvm_pte((struct amdgpu_device *) @@ -2087,6 +2102,7 @@ int amdgpu_amdkfd_gpuvm_restore_process_bos(void *info, struct dma_fence **ef) goto validate_map_fail; } + /* Wait for validate and PT updates to finish */ amdgpu_sync_wait(&sync_obj, false); /* Release old eviction fence and create new one, because fence only @@ -2105,10 +2121,7 @@ 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); - /* Wait for validate to finish and attach new eviction fence */ - list_for_each_entry(mem, &process_info->kfd_bo_list, - validate_list.head) - ttm_bo_wait(&mem->bo->tbo, false, false); + /* Attach new eviction fence to all BOs */ list_for_each_entry(mem, &process_info->kfd_bo_list, validate_list.head) amdgpu_bo_fence(mem->bo, diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_atpx_handler.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_atpx_handler.c index a028661d9e20..92b11de19581 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_atpx_handler.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_atpx_handler.c @@ -576,6 +576,7 @@ static const struct amdgpu_px_quirk amdgpu_px_quirk_list[] = { { 0x1002, 0x6900, 0x1028, 0x0812, AMDGPU_PX_QUIRK_FORCE_ATPX }, { 0x1002, 0x6900, 0x1028, 0x0813, AMDGPU_PX_QUIRK_FORCE_ATPX }, { 0x1002, 0x6900, 0x1025, 0x125A, AMDGPU_PX_QUIRK_FORCE_ATPX }, + { 0x1002, 0x6900, 0x17AA, 0x3806, AMDGPU_PX_QUIRK_FORCE_ATPX }, { 0, 0, 0, 0, 0 }, }; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_bo_list.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_bo_list.c index 14d2982a47cc..5c79da8e1150 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_bo_list.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_bo_list.c @@ -118,7 +118,6 @@ int amdgpu_bo_list_create(struct amdgpu_device *adev, struct drm_file *filp, entry->priority = min(info[i].bo_priority, AMDGPU_BO_LIST_MAX_PRIORITY); entry->tv.bo = &bo->tbo; - entry->tv.shared = !bo->prime_shared_count; if (bo->preferred_domains == AMDGPU_GEM_DOMAIN_GDS) list->gds_obj = bo; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_connectors.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_connectors.c index 69ad6ec0a4f3..bf04c12bd324 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_connectors.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_connectors.c @@ -25,8 +25,8 @@ */ #include <drm/drmP.h> #include <drm/drm_edid.h> -#include <drm/drm_crtc_helper.h> #include <drm/drm_fb_helper.h> +#include <drm/drm_probe_helper.h> #include <drm/amdgpu_drm.h> #include "amdgpu.h" #include "atom.h" diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c index 0acc8dee2cb8..52a5e4fdc95b 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c @@ -50,7 +50,8 @@ 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; - p->uf_entry.tv.shared = true; + /* One for TTM and one for the CS job */ + p->uf_entry.tv.num_shared = 2; p->uf_entry.user_pages = NULL; drm_gem_object_put_unlocked(gobj); @@ -213,6 +214,7 @@ static int amdgpu_cs_parser_init(struct amdgpu_cs_parser *p, union drm_amdgpu_cs case AMDGPU_CHUNK_ID_DEPENDENCIES: case AMDGPU_CHUNK_ID_SYNCOBJ_IN: case AMDGPU_CHUNK_ID_SYNCOBJ_OUT: + case AMDGPU_CHUNK_ID_SCHEDULED_DEPENDENCIES: break; default: @@ -598,6 +600,10 @@ static int amdgpu_cs_parser_bos(struct amdgpu_cs_parser *p, return r; } + /* One for TTM and one for the CS job */ + amdgpu_bo_list_for_each_entry(e, p->bo_list) + e->tv.num_shared = 2; + amdgpu_bo_list_get_list(p->bo_list, &p->validated); if (p->bo_list->first_userptr != p->bo_list->num_entries) p->mn = amdgpu_mn_get(p->adev, AMDGPU_MN_TYPE_GFX); @@ -717,8 +723,14 @@ static int amdgpu_cs_parser_bos(struct amdgpu_cs_parser *p, gws = p->bo_list->gws_obj; oa = p->bo_list->oa_obj; - amdgpu_bo_list_for_each_entry(e, p->bo_list) - e->bo_va = amdgpu_vm_bo_find(vm, ttm_to_amdgpu_bo(e->tv.bo)); + amdgpu_bo_list_for_each_entry(e, p->bo_list) { + struct amdgpu_bo *bo = ttm_to_amdgpu_bo(e->tv.bo); + + /* Make sure we use the exclusive slot for shared BOs */ + if (bo->prime_shared_count) + e->tv.num_shared = 0; + e->bo_va = amdgpu_vm_bo_find(vm, bo); + } if (gds) { p->job->gds_base = amdgpu_bo_gpu_offset(gds) >> PAGE_SHIFT; @@ -955,10 +967,6 @@ static int amdgpu_cs_vm_handling(struct amdgpu_cs_parser *p) if (r) return r; - r = reservation_object_reserve_shared(vm->root.base.bo->tbo.resv); - if (r) - return r; - p->job->vm_pd_addr = amdgpu_gmc_pd_addr(vm->root.base.bo); if (amdgpu_vm_debug) { @@ -1083,6 +1091,15 @@ static int amdgpu_cs_process_fence_dep(struct amdgpu_cs_parser *p, fence = amdgpu_ctx_get_fence(ctx, entity, deps[i].handle); + + if (chunk->chunk_id == AMDGPU_CHUNK_ID_SCHEDULED_DEPENDENCIES) { + struct drm_sched_fence *s_fence = to_drm_sched_fence(fence); + struct dma_fence *old = fence; + + fence = dma_fence_get(&s_fence->scheduled); + dma_fence_put(old); + } + if (IS_ERR(fence)) { r = PTR_ERR(fence); amdgpu_ctx_put(ctx); @@ -1104,7 +1121,7 @@ static int amdgpu_syncobj_lookup_and_add_to_sync(struct amdgpu_cs_parser *p, { int r; struct dma_fence *fence; - r = drm_syncobj_find_fence(p->filp, handle, 0, &fence); + r = drm_syncobj_find_fence(p->filp, handle, 0, 0, &fence); if (r) return r; @@ -1170,7 +1187,8 @@ static int amdgpu_cs_dependencies(struct amdgpu_device *adev, chunk = &p->chunks[i]; - if (chunk->chunk_id == AMDGPU_CHUNK_ID_DEPENDENCIES) { + if (chunk->chunk_id == AMDGPU_CHUNK_ID_DEPENDENCIES || + chunk->chunk_id == AMDGPU_CHUNK_ID_SCHEDULED_DEPENDENCIES) { r = amdgpu_cs_process_fence_dep(p, chunk); if (r) return r; @@ -1193,7 +1211,7 @@ static void amdgpu_cs_post_dependencies(struct amdgpu_cs_parser *p) int i; for (i = 0; i < p->num_post_dep_syncobjs; ++i) - drm_syncobj_replace_fence(p->post_dep_syncobjs[i], 0, p->fence); + drm_syncobj_replace_fence(p->post_dep_syncobjs[i], p->fence); } static int amdgpu_cs_submit(struct amdgpu_cs_parser *p, @@ -1260,8 +1278,7 @@ static int amdgpu_cs_submit(struct amdgpu_cs_parser *p, return 0; error_abort: - dma_fence_put(&job->base.s_fence->finished); - job->base.s_fence = NULL; + drm_sched_job_cleanup(&job->base); amdgpu_mn_unlock(p->mn); error_unlock: @@ -1285,7 +1302,7 @@ int amdgpu_cs_ioctl(struct drm_device *dev, void *data, struct drm_file *filp) r = amdgpu_cs_parser_init(&parser, data); if (r) { - DRM_ERROR("Failed to initialize parser !\n"); + DRM_ERROR("Failed to initialize parser %d!\n", r); goto out; } @@ -1422,6 +1439,9 @@ int amdgpu_cs_fence_to_handle_ioctl(struct drm_device *dev, void *data, if (IS_ERR(fence)) return PTR_ERR(fence); + if (!fence) + fence = dma_fence_get_stub(); + switch (info->in.what) { case AMDGPU_FENCE_TO_HANDLE_GET_SYNCOBJ: r = drm_syncobj_create(&syncobj, 0, fence); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_csa.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_csa.c new file mode 100644 index 000000000000..7e22be7ca68a --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_csa.c @@ -0,0 +1,117 @@ +/* + * Copyright 2016 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. + + * * Author: Monk.liu@amd.com + */ + +#include "amdgpu.h" + +uint64_t amdgpu_csa_vaddr(struct amdgpu_device *adev) +{ + uint64_t addr = adev->vm_manager.max_pfn << AMDGPU_GPU_PAGE_SHIFT; + + addr -= AMDGPU_VA_RESERVED_SIZE; + addr = amdgpu_gmc_sign_extend(addr); + + return addr; +} + +int amdgpu_allocate_static_csa(struct amdgpu_device *adev, struct amdgpu_bo **bo, + u32 domain, uint32_t size) +{ + int r; + void *ptr; + + r = amdgpu_bo_create_kernel(adev, size, PAGE_SIZE, + domain, bo, + NULL, &ptr); + if (!*bo) + return -ENOMEM; + + memset(ptr, 0, size); + return 0; +} + +void amdgpu_free_static_csa(struct amdgpu_bo **bo) +{ + amdgpu_bo_free_kernel(bo, NULL, NULL); +} + +/* + * amdgpu_map_static_csa should be called during amdgpu_vm_init + * it maps virtual address amdgpu_csa_vaddr() to this VM, and each command + * submission of GFX should use this virtual address within META_DATA init + * package to support SRIOV gfx preemption. + */ +int amdgpu_map_static_csa(struct amdgpu_device *adev, struct amdgpu_vm *vm, + struct amdgpu_bo *bo, struct amdgpu_bo_va **bo_va, + uint64_t csa_addr, uint32_t size) +{ + struct ww_acquire_ctx ticket; + struct list_head list; + struct amdgpu_bo_list_entry pd; + struct ttm_validate_buffer csa_tv; + int r; + + INIT_LIST_HEAD(&list); + INIT_LIST_HEAD(&csa_tv.head); + csa_tv.bo = &bo->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 CSA,PD BOs: err=%d\n", r); + return r; + } + + *bo_va = amdgpu_vm_bo_add(adev, vm, bo); + if (!*bo_va) { + ttm_eu_backoff_reservation(&ticket, &list); + DRM_ERROR("failed to create bo_va for static CSA\n"); + return -ENOMEM; + } + + r = amdgpu_vm_alloc_pts(adev, (*bo_va)->base.vm, csa_addr, + size); + if (r) { + DRM_ERROR("failed to allocate pts for static CSA, err=%d\n", r); + amdgpu_vm_bo_rmv(adev, *bo_va); + ttm_eu_backoff_reservation(&ticket, &list); + return r; + } + + r = amdgpu_vm_bo_map(adev, *bo_va, csa_addr, 0, size, + AMDGPU_PTE_READABLE | AMDGPU_PTE_WRITEABLE | + AMDGPU_PTE_EXECUTABLE); + + if (r) { + DRM_ERROR("failed to do bo_map on static CSA, err=%d\n", r); + amdgpu_vm_bo_rmv(adev, *bo_va); + ttm_eu_backoff_reservation(&ticket, &list); + return r; + } + + ttm_eu_backoff_reservation(&ticket, &list); + return 0; +} diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_csa.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_csa.h new file mode 100644 index 000000000000..524b4437a021 --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_csa.h @@ -0,0 +1,39 @@ +/* + * Copyright 2016 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. + * + * Author: Monk.liu@amd.com + */ + +#ifndef AMDGPU_CSA_MANAGER_H +#define AMDGPU_CSA_MANAGER_H + +#define AMDGPU_CSA_SIZE (128 * 1024) + +uint32_t amdgpu_get_total_csa_size(struct amdgpu_device *adev); +uint64_t amdgpu_csa_vaddr(struct amdgpu_device *adev); +int amdgpu_allocate_static_csa(struct amdgpu_device *adev, struct amdgpu_bo **bo, + u32 domain, uint32_t size); +int amdgpu_map_static_csa(struct amdgpu_device *adev, struct amdgpu_vm *vm, + struct amdgpu_bo *bo, struct amdgpu_bo_va **bo_va, + uint64_t csa_addr, uint32_t size); +void amdgpu_free_static_csa(struct amdgpu_bo **bo); + +#endif diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c index 95f4c4139fc6..7b526593eb77 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c @@ -124,6 +124,7 @@ static int amdgpu_ctx_init(struct amdgpu_device *adev, struct amdgpu_ring *rings[AMDGPU_MAX_RINGS]; struct drm_sched_rq *rqs[AMDGPU_MAX_RINGS]; unsigned num_rings; + unsigned num_rqs = 0; switch (i) { case AMDGPU_HW_IP_GFX: @@ -166,12 +167,16 @@ static int amdgpu_ctx_init(struct amdgpu_device *adev, break; } - for (j = 0; j < num_rings; ++j) - rqs[j] = &rings[j]->sched.sched_rq[priority]; + for (j = 0; j < num_rings; ++j) { + if (!rings[j]->adev) + continue; + + rqs[num_rqs++] = &rings[j]->sched.sched_rq[priority]; + } for (j = 0; j < amdgpu_ctx_num_entities[i]; ++j) r = drm_sched_entity_init(&ctx->entities[i][j].entity, - rqs, num_rings, &ctx->guilty); + rqs, num_rqs, &ctx->guilty); if (r) goto error_cleanup_entities; } @@ -248,7 +253,7 @@ static int amdgpu_ctx_alloc(struct amdgpu_device *adev, return -ENOMEM; mutex_lock(&mgr->lock); - r = idr_alloc(&mgr->ctx_handles, ctx, 1, 0, GFP_KERNEL); + r = idr_alloc(&mgr->ctx_handles, ctx, 1, AMDGPU_VM_MAX_NUM_CTX, GFP_KERNEL); if (r < 0) { mutex_unlock(&mgr->lock); kfree(ctx); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c index dd9a4fb9ce39..4ae3ff9a1d4c 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c @@ -158,9 +158,6 @@ static int amdgpu_debugfs_process_reg_op(bool read, struct file *f, while (size) { uint32_t value; - if (*pos > adev->rmmio_size) - goto end; - if (read) { value = RREG32(*pos >> 2); r = put_user(value, (uint32_t *)buf); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index 30bc345d6fdf..79fb302fb954 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -30,8 +30,8 @@ #include <linux/console.h> #include <linux/slab.h> #include <drm/drmP.h> -#include <drm/drm_crtc_helper.h> #include <drm/drm_atomic_helper.h> +#include <drm/drm_probe_helper.h> #include <drm/amdgpu_drm.h> #include <linux/vgaarb.h> #include <linux/vga_switcheroo.h> @@ -59,6 +59,8 @@ #include "amdgpu_amdkfd.h" #include "amdgpu_pm.h" +#include "amdgpu_xgmi.h" + MODULE_FIRMWARE("amdgpu/vega10_gpu_info.bin"); MODULE_FIRMWARE("amdgpu/vega12_gpu_info.bin"); MODULE_FIRMWARE("amdgpu/raven_gpu_info.bin"); @@ -513,6 +515,7 @@ void amdgpu_device_pci_config_reset(struct amdgpu_device *adev) */ static int amdgpu_device_doorbell_init(struct amdgpu_device *adev) { + /* No doorbell on SI hardware generation */ if (adev->asic_type < CHIP_BONAIRE) { adev->doorbell.base = 0; @@ -525,15 +528,26 @@ static int amdgpu_device_doorbell_init(struct amdgpu_device *adev) if (pci_resource_flags(adev->pdev, 2) & IORESOURCE_UNSET) return -EINVAL; + amdgpu_asic_init_doorbell_index(adev); + /* doorbell bar mapping */ 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), - AMDGPU_DOORBELL_MAX_ASSIGNMENT+1); + 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 * sizeof(u32)); @@ -1631,7 +1645,7 @@ static int amdgpu_device_ip_init(struct amdgpu_device *adev) if (r) { DRM_ERROR("sw_init of IP block <%s> failed %d\n", adev->ip_blocks[i].version->funcs->name, r); - return r; + goto init_failed; } adev->ip_blocks[i].status.sw = true; @@ -1640,26 +1654,28 @@ static int amdgpu_device_ip_init(struct amdgpu_device *adev) r = amdgpu_device_vram_scratch_init(adev); if (r) { DRM_ERROR("amdgpu_vram_scratch_init failed %d\n", r); - return r; + goto init_failed; } r = adev->ip_blocks[i].version->funcs->hw_init((void *)adev); if (r) { DRM_ERROR("hw_init %d failed %d\n", i, r); - return r; + goto init_failed; } r = amdgpu_device_wb_init(adev); if (r) { DRM_ERROR("amdgpu_device_wb_init failed %d\n", r); - return r; + goto init_failed; } adev->ip_blocks[i].status.hw = true; /* right after GMC hw init, we create CSA */ if (amdgpu_sriov_vf(adev)) { - r = amdgpu_allocate_static_csa(adev); + r = amdgpu_allocate_static_csa(adev, &adev->virt.csa_obj, + AMDGPU_GEM_DOMAIN_VRAM, + AMDGPU_CSA_SIZE); if (r) { DRM_ERROR("allocate CSA failed %d\n", r); - return r; + goto init_failed; } } } @@ -1667,27 +1683,32 @@ static int amdgpu_device_ip_init(struct amdgpu_device *adev) r = amdgpu_ucode_create_bo(adev); /* create ucode bo when sw_init complete*/ if (r) - return r; + goto init_failed; r = amdgpu_device_ip_hw_init_phase1(adev); if (r) - return r; + goto init_failed; r = amdgpu_device_fw_loading(adev); if (r) - return r; + goto init_failed; r = amdgpu_device_ip_hw_init_phase2(adev); if (r) - return r; + goto init_failed; - amdgpu_xgmi_add_device(adev); + if (adev->gmc.xgmi.num_physical_nodes > 1) + amdgpu_xgmi_add_device(adev); amdgpu_amdkfd_device_init(adev); - if (amdgpu_sriov_vf(adev)) +init_failed: + if (amdgpu_sriov_vf(adev)) { + if (!r) + amdgpu_virt_init_data_exchange(adev); amdgpu_virt_release_full_gpu(adev, true); + } - return 0; + return r; } /** @@ -1848,6 +1869,9 @@ static int amdgpu_device_ip_fini(struct amdgpu_device *adev) { int i, r; + if (adev->gmc.xgmi.num_physical_nodes > 1) + amdgpu_xgmi_remove_device(adev); + amdgpu_amdkfd_device_fini(adev); amdgpu_device_set_pg_state(adev, AMD_PG_STATE_UNGATE); @@ -1890,7 +1914,7 @@ static int amdgpu_device_ip_fini(struct amdgpu_device *adev) if (adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_GMC) { amdgpu_ucode_free_bo(adev); - amdgpu_free_static_csa(adev); + amdgpu_free_static_csa(&adev->virt.csa_obj); amdgpu_device_wb_fini(adev); amdgpu_device_vram_scratch_fini(adev); } @@ -2111,7 +2135,7 @@ static int amdgpu_device_ip_reinit_early_sriov(struct amdgpu_device *adev) continue; r = block->version->funcs->hw_init(adev); - DRM_INFO("RE-INIT: %s %s\n", block->version->funcs->name, r?"failed":"succeeded"); + DRM_INFO("RE-INIT-early: %s %s\n", block->version->funcs->name, r?"failed":"succeeded"); if (r) return r; } @@ -2145,7 +2169,7 @@ static int amdgpu_device_ip_reinit_late_sriov(struct amdgpu_device *adev) continue; r = block->version->funcs->hw_init(adev); - DRM_INFO("RE-INIT: %s %s\n", block->version->funcs->name, r?"failed":"succeeded"); + DRM_INFO("RE-INIT-late: %s %s\n", block->version->funcs->name, r?"failed":"succeeded"); if (r) return r; } @@ -2337,6 +2361,19 @@ bool amdgpu_device_has_dc_support(struct amdgpu_device *adev) return amdgpu_device_asic_has_dc_support(adev->asic_type); } + +static void amdgpu_device_xgmi_reset_func(struct work_struct *__work) +{ + struct amdgpu_device *adev = + container_of(__work, struct amdgpu_device, xgmi_reset_work); + + adev->asic_reset_res = amdgpu_asic_reset(adev); + if (adev->asic_reset_res) + DRM_WARN("ASIC reset failed with err r, %d for drm dev, %s", + adev->asic_reset_res, adev->ddev->unique); +} + + /** * amdgpu_device_init - initialize the driver * @@ -2435,6 +2472,8 @@ int amdgpu_device_init(struct amdgpu_device *adev, INIT_DELAYED_WORK(&adev->gfx.gfx_off_delay_work, amdgpu_device_delay_enable_gfx_off); + INIT_WORK(&adev->xgmi_reset_work, amdgpu_device_xgmi_reset_func); + adev->gfx.gfx_off_req_count = 1; adev->pm.ac_power = power_supply_is_system_supplied() > 0 ? true : false; @@ -2455,9 +2494,6 @@ int amdgpu_device_init(struct amdgpu_device *adev, DRM_INFO("register mmio base: 0x%08X\n", (uint32_t)adev->rmmio_base); DRM_INFO("register mmio size: %u\n", (unsigned)adev->rmmio_size); - /* doorbell bar mapping */ - amdgpu_device_doorbell_init(adev); - /* io port mapping */ for (i = 0; i < DEVICE_COUNT_RESOURCE; i++) { if (pci_resource_flags(adev->pdev, i) & IORESOURCE_IO) { @@ -2476,6 +2512,9 @@ int amdgpu_device_init(struct amdgpu_device *adev, if (r) return r; + /* doorbell bar mapping and doorbell index init*/ + amdgpu_device_doorbell_init(adev); + /* if we have > 1 VGA cards, then disable the amdgpu VGA resources */ /* this will fail for cards that aren't VGA class devices, just * ignore it */ @@ -2511,6 +2550,17 @@ int amdgpu_device_init(struct amdgpu_device *adev, /* detect if we are with an SRIOV vbios */ amdgpu_device_detect_sriov_bios(adev); + /* check if we need to reset the asic + * E.g., driver was not cleanly unloaded previously, etc. + */ + if (!amdgpu_sriov_vf(adev) && amdgpu_asic_need_reset_on_init(adev)) { + r = amdgpu_asic_reset(adev); + if (r) { + dev_err(adev->dev, "asic reset on init failed\n"); + goto failed; + } + } + /* Post card if necessary */ if (amdgpu_device_need_post(adev)) { if (!adev->bios) { @@ -2575,6 +2625,8 @@ fence_driver_init: } dev_err(adev->dev, "amdgpu_device_ip_init failed\n"); amdgpu_vf_error_put(adev, AMDGIM_ERROR_VF_AMDGPU_INIT_FAIL, 0, 0); + if (amdgpu_virt_request_full_gpu(adev, false)) + amdgpu_virt_release_full_gpu(adev, false); goto failed; } @@ -2597,9 +2649,6 @@ fence_driver_init: goto failed; } - if (amdgpu_sriov_vf(adev)) - amdgpu_virt_init_data_exchange(adev); - amdgpu_fbdev_init(adev); r = amdgpu_pm_sysfs_init(adev); @@ -2673,7 +2722,7 @@ void amdgpu_device_fini(struct amdgpu_device *adev) amdgpu_irq_disable_all(adev); if (adev->mode_info.mode_config_initialized){ if (!amdgpu_device_has_dc_support(adev)) - drm_crtc_force_disable_all(adev->ddev); + drm_helper_force_disable_all(adev->ddev); else drm_atomic_helper_shutdown(adev->ddev); } @@ -2763,7 +2812,7 @@ int amdgpu_device_suspend(struct drm_device *dev, bool suspend, bool fbcon) struct drm_framebuffer *fb = crtc->primary->fb; struct amdgpu_bo *robj; - if (amdgpu_crtc->cursor_bo) { + if (amdgpu_crtc->cursor_bo && !adev->enable_virtual_display) { struct amdgpu_bo *aobj = gem_to_amdgpu_bo(amdgpu_crtc->cursor_bo); r = amdgpu_bo_reserve(aobj, true); if (r == 0) { @@ -2871,7 +2920,7 @@ int amdgpu_device_resume(struct drm_device *dev, bool resume, bool fbcon) list_for_each_entry(crtc, &dev->mode_config.crtc_list, head) { struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc); - if (amdgpu_crtc->cursor_bo) { + if (amdgpu_crtc->cursor_bo && !adev->enable_virtual_display) { struct amdgpu_bo *aobj = gem_to_amdgpu_bo(amdgpu_crtc->cursor_bo); r = amdgpu_bo_reserve(aobj, true); if (r == 0) { @@ -3116,6 +3165,7 @@ static int amdgpu_device_recover_vram(struct amdgpu_device *adev) /* No need to recover an evicted BO */ if (shadow->tbo.mem.mem_type != TTM_PL_TT || + shadow->tbo.mem.start == AMDGPU_BO_INVALID_OFFSET || shadow->parent->tbo.mem.mem_type != TTM_PL_VRAM) continue; @@ -3124,11 +3174,16 @@ static int amdgpu_device_recover_vram(struct amdgpu_device *adev) break; if (fence) { - r = dma_fence_wait_timeout(fence, false, tmo); + tmo = dma_fence_wait_timeout(fence, false, tmo); dma_fence_put(fence); fence = next; - if (r <= 0) + if (tmo == 0) { + r = -ETIMEDOUT; break; + } else if (tmo < 0) { + r = tmo; + break; + } } else { fence = next; } @@ -3139,8 +3194,8 @@ static int amdgpu_device_recover_vram(struct amdgpu_device *adev) tmo = dma_fence_wait_timeout(fence, false, tmo); dma_fence_put(fence); - if (r <= 0 || tmo <= 0) { - DRM_ERROR("recover vram bo from shadow failed\n"); + if (r < 0 || tmo <= 0) { + DRM_ERROR("recover vram bo from shadow failed, r is %ld, tmo is %ld\n", r, tmo); return -EIO; } @@ -3148,86 +3203,6 @@ static int amdgpu_device_recover_vram(struct amdgpu_device *adev) return 0; } -/** - * amdgpu_device_reset - reset ASIC/GPU for bare-metal or passthrough - * - * @adev: amdgpu device pointer - * - * attempt to do soft-reset or full-reset and reinitialize Asic - * return 0 means succeeded otherwise failed - */ -static int amdgpu_device_reset(struct amdgpu_device *adev) -{ - bool need_full_reset, vram_lost = 0; - int r; - - need_full_reset = amdgpu_device_ip_need_full_reset(adev); - - if (!need_full_reset) { - amdgpu_device_ip_pre_soft_reset(adev); - r = amdgpu_device_ip_soft_reset(adev); - amdgpu_device_ip_post_soft_reset(adev); - if (r || amdgpu_device_ip_check_soft_reset(adev)) { - DRM_INFO("soft reset failed, will fallback to full reset!\n"); - need_full_reset = true; - } - } - - if (need_full_reset) { - r = amdgpu_device_ip_suspend(adev); - -retry: - r = amdgpu_asic_reset(adev); - /* post card */ - amdgpu_atom_asic_init(adev->mode_info.atom_context); - - if (!r) { - dev_info(adev->dev, "GPU reset succeeded, trying to resume\n"); - r = amdgpu_device_ip_resume_phase1(adev); - if (r) - goto out; - - vram_lost = amdgpu_device_check_vram_lost(adev); - if (vram_lost) { - DRM_ERROR("VRAM is lost!\n"); - atomic_inc(&adev->vram_lost_counter); - } - - r = amdgpu_gtt_mgr_recover( - &adev->mman.bdev.man[TTM_PL_TT]); - if (r) - goto out; - - r = amdgpu_device_fw_loading(adev); - if (r) - return r; - - r = amdgpu_device_ip_resume_phase2(adev); - if (r) - goto out; - - if (vram_lost) - amdgpu_device_fill_reset_magic(adev); - } - } - -out: - if (!r) { - amdgpu_irq_gpu_reset_resume_helper(adev); - r = amdgpu_ib_ring_tests(adev); - if (r) { - dev_err(adev->dev, "ib ring test failed (%d).\n", r); - r = amdgpu_device_ip_suspend(adev); - need_full_reset = true; - goto retry; - } - } - - if (!r) - r = amdgpu_device_recover_vram(adev); - - return r; -} /** * amdgpu_device_reset_sriov - reset ASIC for SR-IOV vf @@ -3271,6 +3246,7 @@ static int amdgpu_device_reset_sriov(struct amdgpu_device *adev, r = amdgpu_ib_ring_tests(adev); error: + amdgpu_virt_init_data_exchange(adev); amdgpu_virt_release_full_gpu(adev, true); if (!r && adev->virt.gim_feature & AMDGIM_FEATURE_GIM_FLR_VRAMLOST) { atomic_inc(&adev->vram_lost_counter); @@ -3295,40 +3271,46 @@ bool amdgpu_device_should_recover_gpu(struct amdgpu_device *adev) return false; } - if (amdgpu_gpu_recovery == 0 || (amdgpu_gpu_recovery == -1 && - !amdgpu_sriov_vf(adev))) { - DRM_INFO("GPU recovery disabled.\n"); - return false; - } + if (amdgpu_gpu_recovery == 0) + goto disabled; - return true; -} + if (amdgpu_sriov_vf(adev)) + return true; -/** - * amdgpu_device_gpu_recover - reset the asic and recover scheduler - * - * @adev: amdgpu device pointer - * @job: which job trigger hang - * - * Attempt to reset the GPU if it has hung (all asics). - * Returns 0 for success or an error on failure. - */ -int amdgpu_device_gpu_recover(struct amdgpu_device *adev, - struct amdgpu_job *job) -{ - int i, r, resched; + 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: + break; + default: + goto disabled; + } + } - dev_info(adev->dev, "GPU reset begin!\n"); + return true; - mutex_lock(&adev->lock_reset); - atomic_inc(&adev->gpu_reset_counter); - adev->in_gpu_reset = 1; +disabled: + DRM_INFO("GPU recovery disabled.\n"); + return false; +} - /* Block kfd */ - amdgpu_amdkfd_pre_reset(adev); - /* block TTM */ - resched = ttm_bo_lock_delayed_workqueue(&adev->mman.bdev); +static int amdgpu_device_pre_asic_reset(struct amdgpu_device *adev, + struct amdgpu_job *job, + bool *need_full_reset_arg) +{ + int i, r = 0; + bool need_full_reset = *need_full_reset_arg; /* block all schedulers and reset given job's ring */ for (i = 0; i < AMDGPU_MAX_RINGS; ++i) { @@ -3337,21 +3319,153 @@ int amdgpu_device_gpu_recover(struct amdgpu_device *adev, if (!ring || !ring->sched.thread) continue; - kthread_park(ring->sched.thread); - - if (job && job->base.sched != &ring->sched) - continue; - - drm_sched_hw_job_reset(&ring->sched, job ? &job->base : NULL); + drm_sched_stop(&ring->sched); /* after all hw jobs are reset, hw fence is meaningless, so force_completion */ amdgpu_fence_driver_force_completion(ring); } - if (amdgpu_sriov_vf(adev)) - r = amdgpu_device_reset_sriov(adev, job ? false : true); - else - r = amdgpu_device_reset(adev); + if(job) + drm_sched_increase_karma(&job->base); + + + + if (!amdgpu_sriov_vf(adev)) { + + if (!need_full_reset) + need_full_reset = amdgpu_device_ip_need_full_reset(adev); + + if (!need_full_reset) { + amdgpu_device_ip_pre_soft_reset(adev); + r = amdgpu_device_ip_soft_reset(adev); + amdgpu_device_ip_post_soft_reset(adev); + if (r || amdgpu_device_ip_check_soft_reset(adev)) { + DRM_INFO("soft reset failed, will fallback to full reset!\n"); + need_full_reset = true; + } + } + + if (need_full_reset) + r = amdgpu_device_ip_suspend(adev); + + *need_full_reset_arg = need_full_reset; + } + + return r; +} + +static int amdgpu_do_asic_reset(struct amdgpu_hive_info *hive, + struct list_head *device_list_handle, + bool *need_full_reset_arg) +{ + struct amdgpu_device *tmp_adev = NULL; + bool need_full_reset = *need_full_reset_arg, vram_lost = false; + int r = 0; + + /* + * ASIC reset has to be done on all HGMI hive nodes ASAP + * to allow proper links negotiation in FW (within 1 sec) + */ + if (need_full_reset) { + list_for_each_entry(tmp_adev, device_list_handle, gmc.xgmi.head) { + /* 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_highpri_wq, &tmp_adev->xgmi_reset_work)) + r = -EALREADY; + } else + r = amdgpu_asic_reset(tmp_adev); + + if (r) { + DRM_ERROR("ASIC reset failed with err r, %d for drm dev, %s", + r, tmp_adev->ddev->unique); + break; + } + } + + /* For XGMI wait for all PSP resets to complete before proceed */ + if (!r) { + list_for_each_entry(tmp_adev, device_list_handle, + gmc.xgmi.head) { + if (tmp_adev->gmc.xgmi.num_physical_nodes > 1) { + flush_work(&tmp_adev->xgmi_reset_work); + r = tmp_adev->asic_reset_res; + if (r) + break; + } + } + } + } + + + list_for_each_entry(tmp_adev, device_list_handle, gmc.xgmi.head) { + if (need_full_reset) { + /* post card */ + if (amdgpu_atom_asic_init(tmp_adev->mode_info.atom_context)) + DRM_WARN("asic atom init failed!"); + + if (!r) { + dev_info(tmp_adev->dev, "GPU reset succeeded, trying to resume\n"); + r = amdgpu_device_ip_resume_phase1(tmp_adev); + if (r) + goto out; + + vram_lost = amdgpu_device_check_vram_lost(tmp_adev); + if (vram_lost) { + DRM_ERROR("VRAM is lost!\n"); + atomic_inc(&tmp_adev->vram_lost_counter); + } + + r = amdgpu_gtt_mgr_recover( + &tmp_adev->mman.bdev.man[TTM_PL_TT]); + if (r) + goto out; + + r = amdgpu_device_fw_loading(tmp_adev); + if (r) + return r; + + r = amdgpu_device_ip_resume_phase2(tmp_adev); + if (r) + goto out; + + if (vram_lost) + amdgpu_device_fill_reset_magic(tmp_adev); + + /* Update PSP FW topology after reset */ + if (hive && tmp_adev->gmc.xgmi.num_physical_nodes > 1) + r = amdgpu_xgmi_update_topology(hive, tmp_adev); + } + } + + +out: + if (!r) { + 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 = amdgpu_device_ip_suspend(tmp_adev); + need_full_reset = true; + r = -EAGAIN; + goto end; + } + } + + if (!r) + r = amdgpu_device_recover_vram(tmp_adev); + else + tmp_adev->asic_reset_res = r; + } + +end: + *need_full_reset_arg = need_full_reset; + return r; +} + +static void amdgpu_device_post_asic_reset(struct amdgpu_device *adev, + struct amdgpu_job *job) +{ + int i; for (i = 0; i < AMDGPU_MAX_RINGS; ++i) { struct amdgpu_ring *ring = adev->rings[i]; @@ -3359,38 +3473,194 @@ int amdgpu_device_gpu_recover(struct amdgpu_device *adev, if (!ring || !ring->sched.thread) continue; - /* only need recovery sched of the given job's ring - * or all rings (in the case @job is NULL) - * after above amdgpu_reset accomplished - */ - if ((!job || job->base.sched == &ring->sched) && !r) - drm_sched_job_recovery(&ring->sched); + if (!adev->asic_reset_res) + drm_sched_resubmit_jobs(&ring->sched); - kthread_unpark(ring->sched.thread); + drm_sched_start(&ring->sched, !adev->asic_reset_res); } if (!amdgpu_device_has_dc_support(adev)) { drm_helper_resume_force_mode(adev->ddev); } - ttm_bo_unlock_delayed_workqueue(&adev->mman.bdev, resched); + adev->asic_reset_res = 0; +} - if (r) { - /* bad news, how to tell it to userspace ? */ - dev_info(adev->dev, "GPU reset(%d) failed\n", atomic_read(&adev->gpu_reset_counter)); - amdgpu_vf_error_put(adev, AMDGIM_ERROR_VF_GPU_RESET_FAIL, 0, r); - } else { - dev_info(adev->dev, "GPU reset(%d) succeeded!\n",atomic_read(&adev->gpu_reset_counter)); - } +static void amdgpu_device_lock_adev(struct amdgpu_device *adev) +{ + mutex_lock(&adev->lock_reset); + atomic_inc(&adev->gpu_reset_counter); + adev->in_gpu_reset = 1; + /* Block kfd: SRIOV would do it separately */ + if (!amdgpu_sriov_vf(adev)) + amdgpu_amdkfd_pre_reset(adev); +} - /*unlock kfd */ - amdgpu_amdkfd_post_reset(adev); +static void amdgpu_device_unlock_adev(struct amdgpu_device *adev) +{ + /*unlock kfd: SRIOV would do it separately */ + if (!amdgpu_sriov_vf(adev)) + amdgpu_amdkfd_post_reset(adev); amdgpu_vf_error_trans_all(adev); adev->in_gpu_reset = 0; mutex_unlock(&adev->lock_reset); +} + + +/** + * amdgpu_device_gpu_recover - reset the asic and recover scheduler + * + * @adev: amdgpu device pointer + * @job: which job trigger hang + * + * Attempt to reset the GPU if it has hung (all asics). + * Attempt to do soft-reset or full-reset and reinitialize Asic + * Returns 0 for success or an error on failure. + */ + +int amdgpu_device_gpu_recover(struct amdgpu_device *adev, + struct amdgpu_job *job) +{ + int r; + struct amdgpu_hive_info *hive = NULL; + bool need_full_reset = false; + struct amdgpu_device *tmp_adev = NULL; + struct list_head device_list, *device_list_handle = NULL; + + INIT_LIST_HEAD(&device_list); + + dev_info(adev->dev, "GPU reset begin!\n"); + + /* + * In case of XGMI hive disallow concurrent resets to be triggered + * by different nodes. No point also since the one node already executing + * reset will also reset all the other nodes in the hive. + */ + hive = amdgpu_get_xgmi_hive(adev, 0); + if (hive && adev->gmc.xgmi.num_physical_nodes > 1 && + !mutex_trylock(&hive->reset_lock)) + return 0; + + /* Start with adev pre asic reset first for soft reset check.*/ + amdgpu_device_lock_adev(adev); + r = amdgpu_device_pre_asic_reset(adev, + job, + &need_full_reset); + if (r) { + /*TODO Should we stop ?*/ + DRM_ERROR("GPU pre asic reset failed with err, %d for drm dev, %s ", + r, adev->ddev->unique); + adev->asic_reset_res = r; + } + + /* Build list of devices to reset */ + if (need_full_reset && adev->gmc.xgmi.num_physical_nodes > 1) { + if (!hive) { + amdgpu_device_unlock_adev(adev); + return -ENODEV; + } + + /* + * In case we are in XGMI hive mode device reset is done for all the + * nodes in the hive to retrain all XGMI links and hence the reset + * sequence is executed in loop on all nodes. + */ + device_list_handle = &hive->device_list; + } else { + list_add_tail(&adev->gmc.xgmi.head, &device_list); + device_list_handle = &device_list; + } + +retry: /* Rest of adevs pre asic reset from XGMI hive. */ + list_for_each_entry(tmp_adev, device_list_handle, gmc.xgmi.head) { + + if (tmp_adev == adev) + continue; + + amdgpu_device_lock_adev(tmp_adev); + r = amdgpu_device_pre_asic_reset(tmp_adev, + NULL, + &need_full_reset); + /*TODO Should we stop ?*/ + if (r) { + DRM_ERROR("GPU pre asic reset failed with err, %d for drm dev, %s ", + r, tmp_adev->ddev->unique); + tmp_adev->asic_reset_res = r; + } + } + + /* Actual ASIC resets if needed.*/ + /* TODO Implement XGMI hive reset logic for SRIOV */ + if (amdgpu_sriov_vf(adev)) { + r = amdgpu_device_reset_sriov(adev, job ? false : true); + if (r) + adev->asic_reset_res = r; + } else { + r = amdgpu_do_asic_reset(hive, device_list_handle, &need_full_reset); + if (r && r == -EAGAIN) + goto retry; + } + + /* Post ASIC reset for all devs .*/ + list_for_each_entry(tmp_adev, device_list_handle, gmc.xgmi.head) { + amdgpu_device_post_asic_reset(tmp_adev, tmp_adev == adev ? job : NULL); + + if (r) { + /* bad news, how to tell it to userspace ? */ + dev_info(tmp_adev->dev, "GPU reset(%d) failed\n", atomic_read(&adev->gpu_reset_counter)); + amdgpu_vf_error_put(tmp_adev, AMDGIM_ERROR_VF_GPU_RESET_FAIL, 0, r); + } else { + dev_info(tmp_adev->dev, "GPU reset(%d) succeeded!\n", atomic_read(&adev->gpu_reset_counter)); + } + + amdgpu_device_unlock_adev(tmp_adev); + } + + if (hive && adev->gmc.xgmi.num_physical_nodes > 1) + mutex_unlock(&hive->reset_lock); + + if (r) + dev_info(adev->dev, "GPU reset end with ret = %d\n", r); return r; } +static void amdgpu_device_get_min_pci_speed_width(struct amdgpu_device *adev, + enum pci_bus_speed *speed, + enum pcie_link_width *width) +{ + struct pci_dev *pdev = adev->pdev; + enum pci_bus_speed cur_speed; + enum pcie_link_width cur_width; + u32 ret = 1; + + *speed = PCI_SPEED_UNKNOWN; + *width = PCIE_LNK_WIDTH_UNKNOWN; + + while (pdev) { + cur_speed = pcie_get_speed_cap(pdev); + cur_width = pcie_get_width_cap(pdev); + ret = pcie_bandwidth_available(adev->pdev, NULL, + NULL, &cur_width); + if (!ret) + cur_width = PCIE_LNK_WIDTH_RESRV; + + if (cur_speed != PCI_SPEED_UNKNOWN) { + if (*speed == PCI_SPEED_UNKNOWN) + *speed = cur_speed; + else if (cur_speed < *speed) + *speed = cur_speed; + } + + if (cur_width != PCIE_LNK_WIDTH_UNKNOWN) { + if (*width == PCIE_LNK_WIDTH_UNKNOWN) + *width = cur_width; + else if (cur_width < *width) + *width = cur_width; + } + pdev = pci_upstream_bridge(pdev); + } +} + /** * amdgpu_device_get_pcie_info - fence pcie info about the PCIE slot * @@ -3403,8 +3673,8 @@ int amdgpu_device_gpu_recover(struct amdgpu_device *adev, static void amdgpu_device_get_pcie_info(struct amdgpu_device *adev) { struct pci_dev *pdev; - enum pci_bus_speed speed_cap; - enum pcie_link_width link_width; + enum pci_bus_speed speed_cap, platform_speed_cap; + enum pcie_link_width platform_link_width; if (amdgpu_pcie_gen_cap) adev->pm.pcie_gen_mask = amdgpu_pcie_gen_cap; @@ -3421,6 +3691,12 @@ static void amdgpu_device_get_pcie_info(struct amdgpu_device *adev) return; } + if (adev->pm.pcie_gen_mask && adev->pm.pcie_mlw_mask) + return; + + amdgpu_device_get_min_pci_speed_width(adev, &platform_speed_cap, + &platform_link_width); + if (adev->pm.pcie_gen_mask == 0) { /* asic caps */ pdev = adev->pdev; @@ -3446,22 +3722,20 @@ static void amdgpu_device_get_pcie_info(struct amdgpu_device *adev) adev->pm.pcie_gen_mask |= CAIL_ASIC_PCIE_LINK_SPEED_SUPPORT_GEN1; } /* platform caps */ - pdev = adev->ddev->pdev->bus->self; - speed_cap = pcie_get_speed_cap(pdev); - if (speed_cap == PCI_SPEED_UNKNOWN) { + if (platform_speed_cap == PCI_SPEED_UNKNOWN) { adev->pm.pcie_gen_mask |= (CAIL_PCIE_LINK_SPEED_SUPPORT_GEN1 | CAIL_PCIE_LINK_SPEED_SUPPORT_GEN2); } else { - if (speed_cap == PCIE_SPEED_16_0GT) + if (platform_speed_cap == PCIE_SPEED_16_0GT) adev->pm.pcie_gen_mask |= (CAIL_PCIE_LINK_SPEED_SUPPORT_GEN1 | CAIL_PCIE_LINK_SPEED_SUPPORT_GEN2 | CAIL_PCIE_LINK_SPEED_SUPPORT_GEN3 | CAIL_PCIE_LINK_SPEED_SUPPORT_GEN4); - else if (speed_cap == PCIE_SPEED_8_0GT) + else if (platform_speed_cap == PCIE_SPEED_8_0GT) adev->pm.pcie_gen_mask |= (CAIL_PCIE_LINK_SPEED_SUPPORT_GEN1 | CAIL_PCIE_LINK_SPEED_SUPPORT_GEN2 | CAIL_PCIE_LINK_SPEED_SUPPORT_GEN3); - else if (speed_cap == PCIE_SPEED_5_0GT) + else if (platform_speed_cap == PCIE_SPEED_5_0GT) adev->pm.pcie_gen_mask |= (CAIL_PCIE_LINK_SPEED_SUPPORT_GEN1 | CAIL_PCIE_LINK_SPEED_SUPPORT_GEN2); else @@ -3470,12 +3744,10 @@ static void amdgpu_device_get_pcie_info(struct amdgpu_device *adev) } } if (adev->pm.pcie_mlw_mask == 0) { - pdev = adev->ddev->pdev->bus->self; - link_width = pcie_get_width_cap(pdev); - if (link_width == PCIE_LNK_WIDTH_UNKNOWN) { + if (platform_link_width == PCIE_LNK_WIDTH_UNKNOWN) { adev->pm.pcie_mlw_mask |= AMDGPU_DEFAULT_PCIE_MLW_MASK; } else { - switch (link_width) { + switch (platform_link_width) { case PCIE_LNK_X32: adev->pm.pcie_mlw_mask = (CAIL_PCIE_LINK_WIDTH_SUPPORT_X32 | CAIL_PCIE_LINK_WIDTH_SUPPORT_X16 | diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_display.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_display.c index 686a26de50f9..b083b219b1a9 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_display.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_display.c @@ -188,10 +188,12 @@ int amdgpu_display_crtc_page_flip_target(struct drm_crtc *crtc, goto cleanup; } - r = amdgpu_bo_pin(new_abo, amdgpu_display_supported_domains(adev)); - if (unlikely(r != 0)) { - DRM_ERROR("failed to pin new abo buffer before flip\n"); - goto unreserve; + if (!adev->enable_virtual_display) { + r = amdgpu_bo_pin(new_abo, amdgpu_display_supported_domains(adev)); + if (unlikely(r != 0)) { + DRM_ERROR("failed to pin new abo buffer before flip\n"); + goto unreserve; + } } r = amdgpu_ttm_alloc_gart(&new_abo->tbo); @@ -211,7 +213,8 @@ int amdgpu_display_crtc_page_flip_target(struct drm_crtc *crtc, amdgpu_bo_get_tiling_flags(new_abo, &tiling_flags); amdgpu_bo_unreserve(new_abo); - work->base = amdgpu_bo_gpu_offset(new_abo); + if (!adev->enable_virtual_display) + work->base = amdgpu_bo_gpu_offset(new_abo); work->target_vblank = target - (uint32_t)drm_crtc_vblank_count(crtc) + amdgpu_get_vblank_counter_kms(dev, work->crtc_id); @@ -242,9 +245,10 @@ pflip_cleanup: goto cleanup; } unpin: - if (unlikely(amdgpu_bo_unpin(new_abo) != 0)) { - DRM_ERROR("failed to unpin new abo in error path\n"); - } + if (!adev->enable_virtual_display) + if (unlikely(amdgpu_bo_unpin(new_abo) != 0)) + DRM_ERROR("failed to unpin new abo in error path\n"); + unreserve: amdgpu_bo_unreserve(new_abo); @@ -631,6 +635,11 @@ int amdgpu_display_modeset_create_props(struct amdgpu_device *adev) drm_property_create_range(adev->ddev, 0, "max bpc", 8, 16); if (!adev->mode_info.max_bpc_property) return -ENOMEM; + adev->mode_info.abm_level_property = + drm_property_create_range(adev->ddev, 0, + "abm level", 0, 4); + if (!adev->mode_info.abm_level_property) + return -ENOMEM; } return 0; @@ -857,7 +866,12 @@ int amdgpu_display_get_crtc_scanoutpos(struct drm_device *dev, /* Inside "upper part" of vblank area? Apply corrective offset if so: */ if (in_vbl && (*vpos >= vbl_start)) { vtotal = mode->crtc_vtotal; - *vpos = *vpos - vtotal; + + /* With variable refresh rate displays the vpos can exceed + * the vtotal value. Clamp to 0 to return -vbl_end instead + * of guessing the remaining number of lines until scanout. + */ + *vpos = (*vpos < vtotal) ? (*vpos - vtotal) : 0; } /* Correct for shifted end of vbl at vbl_end. */ diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_doorbell.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_doorbell.h new file mode 100644 index 000000000000..68959b923f89 --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_doorbell.h @@ -0,0 +1,247 @@ +/* + * Copyright 2018 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. + * + */ + +/* + * GPU doorbell structures, functions & helpers + */ +struct amdgpu_doorbell { + /* doorbell mmio */ + resource_size_t base; + resource_size_t size; + u32 __iomem *ptr; + u32 num_doorbells; /* Number of doorbells actually reserved for amdgpu. */ +}; + +/* Reserved doorbells for amdgpu (including multimedia). + * KFD can use all the rest in the 2M doorbell bar. + * For asic before vega10, doorbell is 32-bit, so the + * index/offset is in dword. For vega10 and after, doorbell + * can be 64-bit, so the index defined is in qword. + */ +struct amdgpu_doorbell_index { + uint32_t kiq; + uint32_t mec_ring0; + uint32_t mec_ring1; + uint32_t mec_ring2; + uint32_t mec_ring3; + uint32_t mec_ring4; + uint32_t mec_ring5; + uint32_t mec_ring6; + uint32_t mec_ring7; + uint32_t userqueue_start; + uint32_t userqueue_end; + uint32_t gfx_ring0; + uint32_t sdma_engine[8]; + uint32_t ih; + union { + struct { + uint32_t vcn_ring0_1; + uint32_t vcn_ring2_3; + uint32_t vcn_ring4_5; + uint32_t vcn_ring6_7; + } vcn; + struct { + uint32_t uvd_ring0_1; + uint32_t uvd_ring2_3; + uint32_t uvd_ring4_5; + uint32_t uvd_ring6_7; + uint32_t vce_ring0_1; + uint32_t vce_ring2_3; + uint32_t vce_ring4_5; + uint32_t vce_ring6_7; + } uvd_vce; + }; + uint32_t first_non_cp; + uint32_t last_non_cp; + uint32_t max_assignment; + /* Per engine SDMA doorbell size in dword */ + uint32_t sdma_doorbell_range; +}; + +typedef enum _AMDGPU_DOORBELL_ASSIGNMENT +{ + AMDGPU_DOORBELL_KIQ = 0x000, + AMDGPU_DOORBELL_HIQ = 0x001, + AMDGPU_DOORBELL_DIQ = 0x002, + AMDGPU_DOORBELL_MEC_RING0 = 0x010, + AMDGPU_DOORBELL_MEC_RING1 = 0x011, + AMDGPU_DOORBELL_MEC_RING2 = 0x012, + AMDGPU_DOORBELL_MEC_RING3 = 0x013, + AMDGPU_DOORBELL_MEC_RING4 = 0x014, + AMDGPU_DOORBELL_MEC_RING5 = 0x015, + AMDGPU_DOORBELL_MEC_RING6 = 0x016, + AMDGPU_DOORBELL_MEC_RING7 = 0x017, + AMDGPU_DOORBELL_GFX_RING0 = 0x020, + AMDGPU_DOORBELL_sDMA_ENGINE0 = 0x1E0, + AMDGPU_DOORBELL_sDMA_ENGINE1 = 0x1E1, + AMDGPU_DOORBELL_IH = 0x1E8, + AMDGPU_DOORBELL_MAX_ASSIGNMENT = 0x3FF, + AMDGPU_DOORBELL_INVALID = 0xFFFF +} AMDGPU_DOORBELL_ASSIGNMENT; + +typedef enum _AMDGPU_VEGA20_DOORBELL_ASSIGNMENT +{ + /* Compute + GFX: 0~255 */ + AMDGPU_VEGA20_DOORBELL_KIQ = 0x000, + AMDGPU_VEGA20_DOORBELL_HIQ = 0x001, + AMDGPU_VEGA20_DOORBELL_DIQ = 0x002, + AMDGPU_VEGA20_DOORBELL_MEC_RING0 = 0x003, + AMDGPU_VEGA20_DOORBELL_MEC_RING1 = 0x004, + AMDGPU_VEGA20_DOORBELL_MEC_RING2 = 0x005, + AMDGPU_VEGA20_DOORBELL_MEC_RING3 = 0x006, + AMDGPU_VEGA20_DOORBELL_MEC_RING4 = 0x007, + AMDGPU_VEGA20_DOORBELL_MEC_RING5 = 0x008, + AMDGPU_VEGA20_DOORBELL_MEC_RING6 = 0x009, + AMDGPU_VEGA20_DOORBELL_MEC_RING7 = 0x00A, + AMDGPU_VEGA20_DOORBELL_USERQUEUE_START = 0x00B, + AMDGPU_VEGA20_DOORBELL_USERQUEUE_END = 0x08A, + AMDGPU_VEGA20_DOORBELL_GFX_RING0 = 0x08B, + /* SDMA:256~335*/ + AMDGPU_VEGA20_DOORBELL_sDMA_ENGINE0 = 0x100, + AMDGPU_VEGA20_DOORBELL_sDMA_ENGINE1 = 0x10A, + AMDGPU_VEGA20_DOORBELL_sDMA_ENGINE2 = 0x114, + AMDGPU_VEGA20_DOORBELL_sDMA_ENGINE3 = 0x11E, + AMDGPU_VEGA20_DOORBELL_sDMA_ENGINE4 = 0x128, + AMDGPU_VEGA20_DOORBELL_sDMA_ENGINE5 = 0x132, + AMDGPU_VEGA20_DOORBELL_sDMA_ENGINE6 = 0x13C, + AMDGPU_VEGA20_DOORBELL_sDMA_ENGINE7 = 0x146, + /* IH: 376~391 */ + AMDGPU_VEGA20_DOORBELL_IH = 0x178, + /* MMSCH: 392~407 + * overlap the doorbell assignment with VCN as they are mutually exclusive + * VCE engine's doorbell is 32 bit and two VCE ring share one QWORD + */ + AMDGPU_VEGA20_DOORBELL64_VCN0_1 = 0x188, /* lower 32 bits for VNC0 and upper 32 bits for VNC1 */ + AMDGPU_VEGA20_DOORBELL64_VCN2_3 = 0x189, + AMDGPU_VEGA20_DOORBELL64_VCN4_5 = 0x18A, + AMDGPU_VEGA20_DOORBELL64_VCN6_7 = 0x18B, + + AMDGPU_VEGA20_DOORBELL64_UVD_RING0_1 = 0x188, + AMDGPU_VEGA20_DOORBELL64_UVD_RING2_3 = 0x189, + AMDGPU_VEGA20_DOORBELL64_UVD_RING4_5 = 0x18A, + AMDGPU_VEGA20_DOORBELL64_UVD_RING6_7 = 0x18B, + + AMDGPU_VEGA20_DOORBELL64_VCE_RING0_1 = 0x18C, + AMDGPU_VEGA20_DOORBELL64_VCE_RING2_3 = 0x18D, + AMDGPU_VEGA20_DOORBELL64_VCE_RING4_5 = 0x18E, + AMDGPU_VEGA20_DOORBELL64_VCE_RING6_7 = 0x18F, + + AMDGPU_VEGA20_DOORBELL64_FIRST_NON_CP = AMDGPU_VEGA20_DOORBELL_sDMA_ENGINE0, + AMDGPU_VEGA20_DOORBELL64_LAST_NON_CP = AMDGPU_VEGA20_DOORBELL64_VCE_RING6_7, + + AMDGPU_VEGA20_DOORBELL_MAX_ASSIGNMENT = 0x18F, + AMDGPU_VEGA20_DOORBELL_INVALID = 0xFFFF +} AMDGPU_VEGA20_DOORBELL_ASSIGNMENT; + +/* + * 64bit doorbell, offset are in QWORD, occupy 2KB doorbell space + */ +typedef enum _AMDGPU_DOORBELL64_ASSIGNMENT +{ + /* + * All compute related doorbells: kiq, hiq, diq, traditional compute queue, user queue, should locate in + * a continues range so that programming CP_MEC_DOORBELL_RANGE_LOWER/UPPER can cover this range. + * Compute related doorbells are allocated from 0x00 to 0x8a + */ + + + /* kernel scheduling */ + AMDGPU_DOORBELL64_KIQ = 0x00, + + /* HSA interface queue and debug queue */ + AMDGPU_DOORBELL64_HIQ = 0x01, + AMDGPU_DOORBELL64_DIQ = 0x02, + + /* Compute engines */ + AMDGPU_DOORBELL64_MEC_RING0 = 0x03, + AMDGPU_DOORBELL64_MEC_RING1 = 0x04, + AMDGPU_DOORBELL64_MEC_RING2 = 0x05, + AMDGPU_DOORBELL64_MEC_RING3 = 0x06, + AMDGPU_DOORBELL64_MEC_RING4 = 0x07, + AMDGPU_DOORBELL64_MEC_RING5 = 0x08, + AMDGPU_DOORBELL64_MEC_RING6 = 0x09, + AMDGPU_DOORBELL64_MEC_RING7 = 0x0a, + + /* User queue doorbell range (128 doorbells) */ + AMDGPU_DOORBELL64_USERQUEUE_START = 0x0b, + AMDGPU_DOORBELL64_USERQUEUE_END = 0x8a, + + /* Graphics engine */ + AMDGPU_DOORBELL64_GFX_RING0 = 0x8b, + + /* + * Other graphics doorbells can be allocated here: from 0x8c to 0xdf + * Graphics voltage island aperture 1 + * default non-graphics QWORD index is 0xe0 - 0xFF inclusive + */ + + /* For vega10 sriov, the sdma doorbell must be fixed as follow + * to keep the same setting with host driver, or it will + * happen conflicts + */ + AMDGPU_DOORBELL64_sDMA_ENGINE0 = 0xF0, + AMDGPU_DOORBELL64_sDMA_HI_PRI_ENGINE0 = 0xF1, + AMDGPU_DOORBELL64_sDMA_ENGINE1 = 0xF2, + AMDGPU_DOORBELL64_sDMA_HI_PRI_ENGINE1 = 0xF3, + + /* Interrupt handler */ + AMDGPU_DOORBELL64_IH = 0xF4, /* For legacy interrupt ring buffer */ + AMDGPU_DOORBELL64_IH_RING1 = 0xF5, /* For page migration request log */ + AMDGPU_DOORBELL64_IH_RING2 = 0xF6, /* For page migration translation/invalidation log */ + + /* VCN engine use 32 bits doorbell */ + AMDGPU_DOORBELL64_VCN0_1 = 0xF8, /* lower 32 bits for VNC0 and upper 32 bits for VNC1 */ + AMDGPU_DOORBELL64_VCN2_3 = 0xF9, + AMDGPU_DOORBELL64_VCN4_5 = 0xFA, + AMDGPU_DOORBELL64_VCN6_7 = 0xFB, + + /* overlap the doorbell assignment with VCN as they are mutually exclusive + * VCE engine's doorbell is 32 bit and two VCE ring share one QWORD + */ + AMDGPU_DOORBELL64_UVD_RING0_1 = 0xF8, + AMDGPU_DOORBELL64_UVD_RING2_3 = 0xF9, + AMDGPU_DOORBELL64_UVD_RING4_5 = 0xFA, + AMDGPU_DOORBELL64_UVD_RING6_7 = 0xFB, + + AMDGPU_DOORBELL64_VCE_RING0_1 = 0xFC, + AMDGPU_DOORBELL64_VCE_RING2_3 = 0xFD, + AMDGPU_DOORBELL64_VCE_RING4_5 = 0xFE, + AMDGPU_DOORBELL64_VCE_RING6_7 = 0xFF, + + AMDGPU_DOORBELL64_FIRST_NON_CP = AMDGPU_DOORBELL64_sDMA_ENGINE0, + AMDGPU_DOORBELL64_LAST_NON_CP = AMDGPU_DOORBELL64_VCE_RING6_7, + + AMDGPU_DOORBELL64_MAX_ASSIGNMENT = 0xFF, + AMDGPU_DOORBELL64_INVALID = 0xFFFF +} AMDGPU_DOORBELL64_ASSIGNMENT; + +u32 amdgpu_mm_rdoorbell(struct amdgpu_device *adev, u32 index); +void amdgpu_mm_wdoorbell(struct amdgpu_device *adev, u32 index, u32 v); +u64 amdgpu_mm_rdoorbell64(struct amdgpu_device *adev, u32 index); +void amdgpu_mm_wdoorbell64(struct amdgpu_device *adev, u32 index, u64 v); + +#define RDOORBELL32(index) amdgpu_mm_rdoorbell(adev, (index)) +#define WDOORBELL32(index, v) amdgpu_mm_wdoorbell(adev, (index), (v)) +#define RDOORBELL64(index) amdgpu_mm_rdoorbell64(adev, (index)) +#define WDOORBELL64(index, v) amdgpu_mm_wdoorbell64(adev, (index), (v)) + diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_dpm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_dpm.c index 1c4595562f8f..344967df3137 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_dpm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_dpm.c @@ -184,61 +184,6 @@ u32 amdgpu_dpm_get_vrefresh(struct amdgpu_device *adev) return vrefresh; } -void amdgpu_calculate_u_and_p(u32 i, u32 r_c, u32 p_b, - u32 *p, u32 *u) -{ - u32 b_c = 0; - u32 i_c; - u32 tmp; - - i_c = (i * r_c) / 100; - tmp = i_c >> p_b; - - while (tmp) { - b_c++; - tmp >>= 1; - } - - *u = (b_c + 1) / 2; - *p = i_c / (1 << (2 * (*u))); -} - -int amdgpu_calculate_at(u32 t, u32 h, u32 fh, u32 fl, u32 *tl, u32 *th) -{ - u32 k, a, ah, al; - u32 t1; - - if ((fl == 0) || (fh == 0) || (fl > fh)) - return -EINVAL; - - k = (100 * fh) / fl; - t1 = (t * (k - 100)); - a = (1000 * (100 * h + t1)) / (10000 + (t1 / 100)); - a = (a + 5) / 10; - ah = ((a * t) + 5000) / 10000; - al = a - ah; - - *th = t - ah; - *tl = t + al; - - return 0; -} - -bool amdgpu_is_uvd_state(u32 class, u32 class2) -{ - if (class & ATOM_PPLIB_CLASSIFICATION_UVDSTATE) - return true; - if (class & ATOM_PPLIB_CLASSIFICATION_HD2STATE) - return true; - if (class & ATOM_PPLIB_CLASSIFICATION_HDSTATE) - return true; - if (class & ATOM_PPLIB_CLASSIFICATION_SDSTATE) - return true; - if (class2 & ATOM_PPLIB_CLASSIFICATION2_MVC) - return true; - return false; -} - bool amdgpu_is_internal_thermal_sensor(enum amdgpu_int_thermal_type sensor) { switch (sensor) { @@ -949,39 +894,6 @@ enum amdgpu_pcie_gen amdgpu_get_pcie_gen_support(struct amdgpu_device *adev, return AMDGPU_PCIE_GEN1; } -u16 amdgpu_get_pcie_lane_support(struct amdgpu_device *adev, - u16 asic_lanes, - u16 default_lanes) -{ - switch (asic_lanes) { - case 0: - default: - return default_lanes; - case 1: - return 1; - case 2: - return 2; - case 4: - return 4; - case 8: - return 8; - case 12: - return 12; - case 16: - return 16; - } -} - -u8 amdgpu_encode_pci_lane_width(u32 lanes) -{ - u8 encoded_lanes[] = { 0, 1, 2, 0, 3, 0, 0, 0, 4, 0, 0, 0, 5, 0, 0, 0, 6 }; - - if (lanes > 16) - return 0; - - return encoded_lanes[lanes]; -} - struct amd_vce_state* amdgpu_get_vce_clock_state(void *handle, u32 idx) { diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_dpm.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_dpm.h index f972cd156795..e871e022c129 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_dpm.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_dpm.h @@ -364,6 +364,14 @@ enum amdgpu_pcie_gen { ((adev)->powerplay.pp_funcs->enable_mgpu_fan_boost(\ (adev)->powerplay.pp_handle)) +#define amdgpu_dpm_get_ppfeature_status(adev, buf) \ + ((adev)->powerplay.pp_funcs->get_ppfeature_status(\ + (adev)->powerplay.pp_handle, (buf))) + +#define amdgpu_dpm_set_ppfeature_status(adev, ppfeatures) \ + ((adev)->powerplay.pp_funcs->set_ppfeature_status(\ + (adev)->powerplay.pp_handle, (ppfeatures))) + struct amdgpu_dpm { struct amdgpu_ps *ps; /* number of valid power states */ @@ -478,10 +486,6 @@ void amdgpu_dpm_print_ps_status(struct amdgpu_device *adev, u32 amdgpu_dpm_get_vblank_time(struct amdgpu_device *adev); u32 amdgpu_dpm_get_vrefresh(struct amdgpu_device *adev); void amdgpu_dpm_get_active_displays(struct amdgpu_device *adev); -bool amdgpu_is_uvd_state(u32 class, u32 class2); -void amdgpu_calculate_u_and_p(u32 i, u32 r_c, u32 p_b, - u32 *p, u32 *u); -int amdgpu_calculate_at(u32 t, u32 h, u32 fh, u32 fl, u32 *tl, u32 *th); bool amdgpu_is_internal_thermal_sensor(enum amdgpu_int_thermal_type sensor); @@ -497,11 +501,6 @@ enum amdgpu_pcie_gen amdgpu_get_pcie_gen_support(struct amdgpu_device *adev, enum amdgpu_pcie_gen asic_gen, enum amdgpu_pcie_gen default_gen); -u16 amdgpu_get_pcie_lane_support(struct amdgpu_device *adev, - u16 asic_lanes, - u16 default_lanes); -u8 amdgpu_encode_pci_lane_width(u32 lanes); - struct amd_vce_state* amdgpu_get_vce_clock_state(void *handle, u32 idx); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c index 74b611e8a1b1..7419ea8a388b 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c @@ -32,7 +32,7 @@ #include <linux/module.h> #include <linux/pm_runtime.h> #include <linux/vga_switcheroo.h> -#include <drm/drm_crtc_helper.h> +#include <drm/drm_probe_helper.h> #include "amdgpu.h" #include "amdgpu_irq.h" @@ -71,9 +71,12 @@ * - 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. + * - 3.28.0 - 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. */ #define KMS_DRIVER_MAJOR 3 -#define KMS_DRIVER_MINOR 27 +#define KMS_DRIVER_MINOR 30 #define KMS_DRIVER_PATCHLEVEL 0 int amdgpu_vram_limit = 0; @@ -454,9 +457,10 @@ module_param_named(cntl_sb_buf_per_se, amdgpu_cntl_sb_buf_per_se, int, 0444); /** * DOC: param_buf_per_se (int) - * Override the size of Off-Chip Pramater Cache per Shader Engine in Byte. The default is 0 (depending on gfx). + * Override the size of Off-Chip Parameter Cache per Shader Engine in Byte. + * The default is 0 (depending on gfx). */ -MODULE_PARM_DESC(param_buf_per_se, "the size of Off-Chip Pramater Cache per Shader Engine (default depending on gfx)"); +MODULE_PARM_DESC(param_buf_per_se, "the size of Off-Chip Parameter Cache per Shader Engine (default depending on gfx)"); module_param_named(param_buf_per_se, amdgpu_param_buf_per_se, int, 0444); /** @@ -864,6 +868,7 @@ static const struct pci_device_id pciidlist[] = { /* VEGAM */ {0x1002, 0x694C, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_VEGAM}, {0x1002, 0x694E, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_VEGAM}, + {0x1002, 0x694F, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_VEGAM}, /* Vega 10 */ {0x1002, 0x6860, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_VEGA10}, {0x1002, 0x6861, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_VEGA10}, @@ -1174,6 +1179,22 @@ static const struct file_operations amdgpu_driver_kms_fops = { #endif }; +int amdgpu_file_to_fpriv(struct file *filp, struct amdgpu_fpriv **fpriv) +{ + struct drm_file *file; + + if (!filp) + return -EINVAL; + + if (filp->f_op != &amdgpu_driver_kms_fops) { + return -EINVAL; + } + + file = filp->private_data; + *fpriv = file->driver_priv; + return 0; +} + static bool amdgpu_get_crtc_scanout_position(struct drm_device *dev, unsigned int pipe, bool in_vblank_irq, int *vpos, int *hpos, @@ -1187,7 +1208,7 @@ amdgpu_get_crtc_scanout_position(struct drm_device *dev, unsigned int pipe, static struct drm_driver kms_driver = { .driver_features = DRIVER_USE_AGP | DRIVER_ATOMIC | - DRIVER_HAVE_IRQ | DRIVER_IRQ_SHARED | DRIVER_GEM | + DRIVER_GEM | DRIVER_PRIME | DRIVER_RENDER | DRIVER_MODESET | DRIVER_SYNCOBJ, .load = amdgpu_driver_load_kms, .open = amdgpu_driver_open_kms, @@ -1227,9 +1248,6 @@ static struct drm_driver kms_driver = { .patchlevel = KMS_DRIVER_PATCHLEVEL, }; -static struct drm_driver *driver; -static struct pci_driver *pdriver; - static struct pci_driver amdgpu_kms_pci_driver = { .name = DRIVER_NAME, .id_table = pciidlist, @@ -1259,16 +1277,14 @@ static int __init amdgpu_init(void) goto error_fence; DRM_INFO("amdgpu kernel modesetting enabled.\n"); - driver = &kms_driver; - pdriver = &amdgpu_kms_pci_driver; - driver->num_ioctls = amdgpu_max_kms_ioctl; + kms_driver.num_ioctls = amdgpu_max_kms_ioctl; amdgpu_register_atpx_handler(); /* Ignore KFD init failures. Normal when CONFIG_HSA_AMD is not set. */ amdgpu_amdkfd_init(); /* let modprobe override vga console setting */ - return pci_register_driver(pdriver); + return pci_register_driver(&amdgpu_kms_pci_driver); error_fence: amdgpu_sync_fini(); @@ -1280,7 +1296,7 @@ error_sync: static void __exit amdgpu_exit(void) { amdgpu_amdkfd_fini(); - pci_unregister_driver(pdriver); + pci_unregister_driver(&amdgpu_kms_pci_driver); amdgpu_unregister_atpx_handler(); amdgpu_sync_fini(); amdgpu_fence_slab_fini(); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_fence.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_fence.c index 5448cf27654e..ee47c11e92ce 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_fence.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_fence.c @@ -398,9 +398,9 @@ int amdgpu_fence_driver_start_ring(struct amdgpu_ring *ring, ring->fence_drv.irq_type = irq_type; ring->fence_drv.initialized = true; - dev_dbg(adev->dev, "fence driver on ring %d use gpu addr 0x%016llx, " - "cpu addr 0x%p\n", ring->idx, - ring->fence_drv.gpu_addr, ring->fence_drv.cpu_addr); + DRM_DEV_DEBUG(adev->dev, "fence driver on ring %s use gpu addr " + "0x%016llx, cpu addr 0x%p\n", ring->name, + ring->fence_drv.gpu_addr, ring->fence_drv.cpu_addr); return 0; } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gart.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gart.c index 11fea28f8ad3..6d11e1721147 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gart.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gart.c @@ -248,7 +248,7 @@ int amdgpu_gart_unbind(struct amdgpu_device *adev, uint64_t offset, } mb(); amdgpu_asic_flush_hdp(adev, NULL); - amdgpu_gmc_flush_gpu_tlb(adev, 0); + amdgpu_gmc_flush_gpu_tlb(adev, 0, 0); return 0; } @@ -259,6 +259,8 @@ int amdgpu_gart_unbind(struct amdgpu_device *adev, uint64_t offset, * @offset: offset into the GPU's gart aperture * @pages: number of pages to bind * @dma_addr: DMA addresses of pages + * @flags: page table entry flags + * @dst: CPU address of the gart table * * Map the dma_addresses into GART entries (all asics). * Returns 0 for success, -EINVAL for failure. @@ -331,7 +333,7 @@ int amdgpu_gart_bind(struct amdgpu_device *adev, uint64_t offset, mb(); amdgpu_asic_flush_hdp(adev, NULL); - amdgpu_gmc_flush_gpu_tlb(adev, 0); + amdgpu_gmc_flush_gpu_tlb(adev, 0, 0); return 0; } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gart.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_gart.h index 9ff62887e4e3..afa2e2877d87 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gart.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gart.h @@ -41,6 +41,7 @@ struct amdgpu_bo; struct amdgpu_gart { struct amdgpu_bo *bo; + /* CPU kmapped address of gart table */ void *ptr; unsigned num_gpu_pages; unsigned num_cpu_pages; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gds.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_gds.h index ecbcefe49a98..f89f5734d985 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gds.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gds.h @@ -37,6 +37,8 @@ struct amdgpu_gds { struct amdgpu_gds_asic_info mem; struct amdgpu_gds_asic_info gws; struct amdgpu_gds_asic_info oa; + uint32_t gds_compute_max_wave_id; + /* At present, GDS, GWS and OA resources for gfx (graphics) * is always pre-allocated and available for graphics operation. * Such resource is shared between all gfx clients. diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c index 7b3d1ebda9df..d21dd2f369da 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c @@ -54,10 +54,6 @@ int amdgpu_gem_object_create(struct amdgpu_device *adev, unsigned long size, memset(&bp, 0, sizeof(bp)); *obj = NULL; - /* At least align on page size */ - if (alignment < PAGE_SIZE) { - alignment = PAGE_SIZE; - } bp.size = size; bp.byte_align = alignment; @@ -169,7 +165,7 @@ void amdgpu_gem_object_close(struct drm_gem_object *obj, INIT_LIST_HEAD(&duplicates); tv.bo = &bo->tbo; - tv.shared = true; + tv.num_shared = 1; list_add(&tv.head, &list); amdgpu_vm_get_pd_bo(vm, &list, &vm_pd); @@ -244,9 +240,6 @@ int amdgpu_gem_create_ioctl(struct drm_device *dev, void *data, return -EINVAL; } flags |= AMDGPU_GEM_CREATE_NO_CPU_ACCESS; - /* GDS allocations must be DW aligned */ - if (args->in.domains & AMDGPU_GEM_DOMAIN_GDS) - size = ALIGN(size, 4); } if (flags & AMDGPU_GEM_CREATE_VM_ALWAYS_VALID) { @@ -604,7 +597,10 @@ int amdgpu_gem_va_ioctl(struct drm_device *dev, void *data, return -ENOENT; abo = gem_to_amdgpu_bo(gobj); tv.bo = &abo->tbo; - tv.shared = !!(abo->flags & AMDGPU_GEM_CREATE_VM_ALWAYS_VALID); + if (abo->flags & AMDGPU_GEM_CREATE_VM_ALWAYS_VALID) + tv.num_shared = 1; + else + tv.num_shared = 0; list_add(&tv.head, &list); } else { gobj = NULL; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.h index d63daba9b17c..f1ddfc50bcc7 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.h @@ -54,6 +54,8 @@ void *amdgpu_gem_prime_vmap(struct drm_gem_object *obj); void amdgpu_gem_prime_vunmap(struct drm_gem_object *obj, void *vaddr); int amdgpu_gem_prime_mmap(struct drm_gem_object *obj, struct vm_area_struct *vma); +extern const struct dma_buf_ops amdgpu_dmabuf_ops; + /* * GEM objects. */ diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c index 1a656b8657f7..97a60da62004 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c @@ -25,6 +25,7 @@ #include <drm/drmP.h> #include "amdgpu.h" #include "amdgpu_gfx.h" +#include "amdgpu_rlc.h" /* delay 0.1 second to enable gfx off feature */ #define GFX_OFF_DELAY_ENABLE msecs_to_jiffies(100) @@ -249,7 +250,7 @@ int amdgpu_gfx_kiq_init_ring(struct amdgpu_device *adev, ring->adev = NULL; ring->ring_obj = NULL; ring->use_doorbell = true; - ring->doorbell_index = AMDGPU_DOORBELL_KIQ; + ring->doorbell_index = adev->doorbell_index.kiq; r = amdgpu_gfx_kiq_acquire(adev, ring); if (r) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h index b61b5c11aead..f790e15bcd08 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h @@ -29,6 +29,7 @@ */ #include "clearstate_defs.h" #include "amdgpu_ring.h" +#include "amdgpu_rlc.h" /* GFX current status */ #define AMDGPU_GFX_NORMAL_MODE 0x00000000L @@ -37,59 +38,6 @@ #define AMDGPU_GFX_CG_DISABLED_MODE 0x00000004L #define AMDGPU_GFX_LBPW_DISABLED_MODE 0x00000008L - -struct amdgpu_rlc_funcs { - void (*enter_safe_mode)(struct amdgpu_device *adev); - void (*exit_safe_mode)(struct amdgpu_device *adev); -}; - -struct amdgpu_rlc { - /* for power gating */ - struct amdgpu_bo *save_restore_obj; - uint64_t save_restore_gpu_addr; - volatile uint32_t *sr_ptr; - const u32 *reg_list; - u32 reg_list_size; - /* for clear state */ - struct amdgpu_bo *clear_state_obj; - uint64_t clear_state_gpu_addr; - volatile uint32_t *cs_ptr; - const struct cs_section_def *cs_data; - u32 clear_state_size; - /* for cp tables */ - struct amdgpu_bo *cp_table_obj; - uint64_t cp_table_gpu_addr; - volatile uint32_t *cp_table_ptr; - u32 cp_table_size; - - /* safe mode for updating CG/PG state */ - bool in_safe_mode; - const struct amdgpu_rlc_funcs *funcs; - - /* for firmware data */ - u32 save_and_restore_offset; - u32 clear_state_descriptor_offset; - u32 avail_scratch_ram_locations; - u32 reg_restore_list_size; - u32 reg_list_format_start; - u32 reg_list_format_separate_start; - u32 starting_offsets_start; - u32 reg_list_format_size_bytes; - u32 reg_list_size_bytes; - u32 reg_list_format_direct_reg_list_length; - u32 save_restore_list_cntl_size_bytes; - u32 save_restore_list_gpm_size_bytes; - u32 save_restore_list_srm_size_bytes; - - u32 *register_list_format; - u32 *register_restore; - u8 *save_restore_list_cntl; - u8 *save_restore_list_gpm; - u8 *save_restore_list_srm; - - bool is_rlc_v2_1; -}; - #define AMDGPU_MAX_COMPUTE_QUEUES KGD_MAX_QUEUES struct amdgpu_mec { diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.h index 6fa7ef446e46..81e6070d255b 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.h @@ -64,7 +64,7 @@ struct amdgpu_vmhub { struct amdgpu_gmc_funcs { /* flush the vm tlb via mmio */ void (*flush_gpu_tlb)(struct amdgpu_device *adev, - uint32_t vmid); + uint32_t vmid, uint32_t flush_type); /* flush the vm tlb via ring */ uint64_t (*emit_flush_gpu_tlb)(struct amdgpu_ring *ring, unsigned vmid, uint64_t pd_addr); @@ -89,7 +89,7 @@ struct amdgpu_gmc_funcs { struct amdgpu_xgmi { /* from psp */ - u64 device_id; + u64 node_id; u64 hive_id; /* fixed per family */ u64 node_segment_size; @@ -99,6 +99,7 @@ struct amdgpu_xgmi { unsigned num_physical_nodes; /* gpu list in the same hive */ struct list_head head; + bool supported; }; struct amdgpu_gmc { @@ -151,7 +152,7 @@ struct amdgpu_gmc { struct amdgpu_xgmi xgmi; }; -#define amdgpu_gmc_flush_gpu_tlb(adev, vmid) (adev)->gmc.gmc_funcs->flush_gpu_tlb((adev), (vmid)) +#define amdgpu_gmc_flush_gpu_tlb(adev, vmid, type) (adev)->gmc.gmc_funcs->flush_gpu_tlb((adev), (vmid), (type)) #define amdgpu_gmc_emit_flush_gpu_tlb(r, vmid, addr) (r)->adev->gmc.gmc_funcs->emit_flush_gpu_tlb((r), (vmid), (addr)) #define amdgpu_gmc_emit_pasid_mapping(r, vmid, pasid) (r)->adev->gmc.gmc_funcs->emit_pasid_mapping((r), (vmid), (pasid)) #define amdgpu_gmc_set_pte_pde(adev, pt, idx, addr, flags) (adev)->gmc.gmc_funcs->set_pte_pde((adev), (pt), (idx), (addr), (flags)) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c index b8963b725dfa..fe393a46f881 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c @@ -35,6 +35,7 @@ #include "amdgpu_trace.h" #define AMDGPU_IB_TEST_TIMEOUT msecs_to_jiffies(1000) +#define AMDGPU_IB_TEST_GFX_XGMI_TIMEOUT msecs_to_jiffies(2000) /* * IB @@ -146,7 +147,7 @@ int amdgpu_ib_schedule(struct amdgpu_ring *ring, unsigned num_ibs, fence_ctx = 0; } - if (!ring->ready) { + if (!ring->sched.ready) { dev_err(adev->dev, "couldn't schedule ib on ring <%s>\n", ring->name); return -EINVAL; } @@ -202,12 +203,12 @@ int amdgpu_ib_schedule(struct amdgpu_ring *ring, unsigned num_ibs, amdgpu_asic_flush_hdp(adev, ring); } + if (need_ctx_switch) + status |= AMDGPU_HAVE_CTX_SWITCH; + skip_preamble = ring->current_ctx == fence_ctx; if (job && ring->funcs->emit_cntxcntl) { - if (need_ctx_switch) - status |= AMDGPU_HAVE_CTX_SWITCH; status |= job->preamble_status; - amdgpu_ring_emit_cntxcntl(ring, status); } @@ -221,9 +222,8 @@ int amdgpu_ib_schedule(struct amdgpu_ring *ring, unsigned num_ibs, !amdgpu_sriov_vf(adev)) /* for SRIOV preemption, Preamble CE ib must be inserted anyway */ continue; - amdgpu_ring_emit_ib(ring, ib, job ? job->vmid : 0, - need_ctx_switch); - need_ctx_switch = false; + amdgpu_ring_emit_ib(ring, job, ib, status); + status &= ~AMDGPU_HAVE_CTX_SWITCH; } if (ring->funcs->emit_tmz) @@ -345,21 +345,18 @@ int amdgpu_ib_ring_tests(struct amdgpu_device *adev) * cost waiting for it coming back under RUNTIME only */ tmo_gfx = 8 * AMDGPU_IB_TEST_TIMEOUT; + } else if (adev->gmc.xgmi.hive_id) { + tmo_gfx = AMDGPU_IB_TEST_GFX_XGMI_TIMEOUT; } - for (i = 0; i < AMDGPU_MAX_RINGS; ++i) { + for (i = 0; i < adev->num_rings; ++i) { struct amdgpu_ring *ring = adev->rings[i]; long tmo; - if (!ring || !ring->ready) - continue; - - /* skip IB tests for KIQ in general for the below reasons: - * 1. We never submit IBs to the KIQ - * 2. KIQ doesn't use the EOP interrupts, - * we use some other CP interrupt. + /* KIQ rings don't have an IB test because we never submit IBs + * to them and they have no interrupt support. */ - if (ring->funcs->type == AMDGPU_RING_TYPE_KIQ) + if (!ring->sched.ready || !ring->funcs->test_ib) continue; /* MM engine need more time */ @@ -374,20 +371,23 @@ int amdgpu_ib_ring_tests(struct amdgpu_device *adev) tmo = tmo_gfx; r = amdgpu_ring_test_ib(ring, tmo); - if (r) { - ring->ready = false; - - if (ring == &adev->gfx.gfx_ring[0]) { - /* oh, oh, that's really bad */ - DRM_ERROR("amdgpu: failed testing IB on GFX ring (%d).\n", r); - adev->accel_working = false; - return r; - - } else { - /* still not good, but we can live with it */ - DRM_ERROR("amdgpu: failed testing IB on ring %d (%d).\n", i, r); - ret = r; - } + if (!r) { + DRM_DEV_DEBUG(adev->dev, "ib test on %s succeeded\n", + ring->name); + continue; + } + + ring->sched.ready = false; + DRM_DEV_ERROR(adev->dev, "IB test failed on %s (%d).\n", + ring->name, r); + + if (ring == &adev->gfx.gfx_ring[0]) { + /* oh, oh, that's really bad */ + adev->accel_working = false; + return r; + + } else { + ret = r; } } return ret; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ih.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ih.c index 8af67f649660..1c50be3ab8a9 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ih.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ih.c @@ -52,6 +52,8 @@ int amdgpu_ih_ring_init(struct amdgpu_device *adev, struct amdgpu_ih_ring *ih, ih->use_bus_addr = use_bus_addr; if (use_bus_addr) { + dma_addr_t dma_addr; + if (ih->ring) return 0; @@ -59,21 +61,26 @@ int amdgpu_ih_ring_init(struct amdgpu_device *adev, struct amdgpu_ih_ring *ih, * add them to the end of the ring allocation. */ ih->ring = dma_alloc_coherent(adev->dev, ih->ring_size + 8, - &ih->rb_dma_addr, GFP_KERNEL); + &dma_addr, GFP_KERNEL); if (ih->ring == NULL) return -ENOMEM; memset((void *)ih->ring, 0, ih->ring_size + 8); - ih->wptr_offs = (ih->ring_size / 4) + 0; - ih->rptr_offs = (ih->ring_size / 4) + 1; + ih->gpu_addr = dma_addr; + ih->wptr_addr = dma_addr + ih->ring_size; + ih->wptr_cpu = &ih->ring[ih->ring_size / 4]; + ih->rptr_addr = dma_addr + ih->ring_size + 4; + ih->rptr_cpu = &ih->ring[(ih->ring_size / 4) + 1]; } else { - r = amdgpu_device_wb_get(adev, &ih->wptr_offs); + unsigned wptr_offs, rptr_offs; + + r = amdgpu_device_wb_get(adev, &wptr_offs); if (r) return r; - r = amdgpu_device_wb_get(adev, &ih->rptr_offs); + r = amdgpu_device_wb_get(adev, &rptr_offs); if (r) { - amdgpu_device_wb_free(adev, ih->wptr_offs); + amdgpu_device_wb_free(adev, wptr_offs); return r; } @@ -82,10 +89,15 @@ int amdgpu_ih_ring_init(struct amdgpu_device *adev, struct amdgpu_ih_ring *ih, &ih->ring_obj, &ih->gpu_addr, (void **)&ih->ring); if (r) { - amdgpu_device_wb_free(adev, ih->rptr_offs); - amdgpu_device_wb_free(adev, ih->wptr_offs); + amdgpu_device_wb_free(adev, rptr_offs); + amdgpu_device_wb_free(adev, wptr_offs); return r; } + + ih->wptr_addr = adev->wb.gpu_addr + wptr_offs * 4; + ih->wptr_cpu = &adev->wb.wb[wptr_offs]; + ih->rptr_addr = adev->wb.gpu_addr + rptr_offs * 4; + ih->rptr_cpu = &adev->wb.wb[rptr_offs]; } return 0; } @@ -109,13 +121,13 @@ void amdgpu_ih_ring_fini(struct amdgpu_device *adev, struct amdgpu_ih_ring *ih) * add them to the end of the ring allocation. */ dma_free_coherent(adev->dev, ih->ring_size + 8, - (void *)ih->ring, ih->rb_dma_addr); + (void *)ih->ring, ih->gpu_addr); ih->ring = NULL; } else { amdgpu_bo_free_kernel(&ih->ring_obj, &ih->gpu_addr, (void **)&ih->ring); - amdgpu_device_wb_free(adev, ih->wptr_offs); - amdgpu_device_wb_free(adev, ih->rptr_offs); + amdgpu_device_wb_free(adev, (ih->wptr_addr - ih->gpu_addr) / 4); + amdgpu_device_wb_free(adev, (ih->rptr_addr - ih->gpu_addr) / 4); } } @@ -128,16 +140,14 @@ void amdgpu_ih_ring_fini(struct amdgpu_device *adev, struct amdgpu_ih_ring *ih) * Interrupt hander (VI), walk the IH ring. * Returns irq process return code. */ -int amdgpu_ih_process(struct amdgpu_device *adev, struct amdgpu_ih_ring *ih, - void (*callback)(struct amdgpu_device *adev, - struct amdgpu_ih_ring *ih)) +int amdgpu_ih_process(struct amdgpu_device *adev, struct amdgpu_ih_ring *ih) { u32 wptr; if (!ih->enabled || adev->shutdown) return IRQ_NONE; - wptr = amdgpu_ih_get_wptr(adev); + wptr = amdgpu_ih_get_wptr(adev, ih); restart_ih: /* is somebody else already processing irqs? */ @@ -150,15 +160,15 @@ restart_ih: rmb(); while (ih->rptr != wptr) { - callback(adev, ih); + amdgpu_irq_dispatch(adev, ih); ih->rptr &= ih->ptr_mask; } - amdgpu_ih_set_rptr(adev); + amdgpu_ih_set_rptr(adev, ih); atomic_set(&ih->lock, 0); /* make sure wptr hasn't changed while processing */ - wptr = amdgpu_ih_get_wptr(adev); + wptr = amdgpu_ih_get_wptr(adev, ih); if (wptr != ih->rptr) goto restart_ih; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ih.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ih.h index 9ce8c93ec19b..113a1ba13d4a 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ih.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ih.h @@ -31,42 +31,44 @@ struct amdgpu_iv_entry; * R6xx+ IH ring */ struct amdgpu_ih_ring { - struct amdgpu_bo *ring_obj; - volatile uint32_t *ring; - unsigned rptr; unsigned ring_size; - uint64_t gpu_addr; uint32_t ptr_mask; - atomic_t lock; - bool enabled; - unsigned wptr_offs; - unsigned rptr_offs; u32 doorbell_index; bool use_doorbell; bool use_bus_addr; - dma_addr_t rb_dma_addr; /* only used when use_bus_addr = true */ + + struct amdgpu_bo *ring_obj; + volatile uint32_t *ring; + uint64_t gpu_addr; + + uint64_t wptr_addr; + volatile uint32_t *wptr_cpu; + + uint64_t rptr_addr; + volatile uint32_t *rptr_cpu; + + bool enabled; + unsigned rptr; + atomic_t lock; }; /* provided by the ih block */ struct amdgpu_ih_funcs { /* ring read/write ptr handling, called from interrupt context */ - u32 (*get_wptr)(struct amdgpu_device *adev); - bool (*prescreen_iv)(struct amdgpu_device *adev); - void (*decode_iv)(struct amdgpu_device *adev, + u32 (*get_wptr)(struct amdgpu_device *adev, struct amdgpu_ih_ring *ih); + void (*decode_iv)(struct amdgpu_device *adev, struct amdgpu_ih_ring *ih, struct amdgpu_iv_entry *entry); - void (*set_rptr)(struct amdgpu_device *adev); + void (*set_rptr)(struct amdgpu_device *adev, struct amdgpu_ih_ring *ih); }; -#define amdgpu_ih_get_wptr(adev) (adev)->irq.ih_funcs->get_wptr((adev)) -#define amdgpu_ih_prescreen_iv(adev) (adev)->irq.ih_funcs->prescreen_iv((adev)) -#define amdgpu_ih_decode_iv(adev, iv) (adev)->irq.ih_funcs->decode_iv((adev), (iv)) -#define amdgpu_ih_set_rptr(adev) (adev)->irq.ih_funcs->set_rptr((adev)) +#define amdgpu_ih_get_wptr(adev, ih) (adev)->irq.ih_funcs->get_wptr((adev), (ih)) +#define amdgpu_ih_decode_iv(adev, iv) \ + (adev)->irq.ih_funcs->decode_iv((adev), (ih), (iv)) +#define amdgpu_ih_set_rptr(adev, ih) (adev)->irq.ih_funcs->set_rptr((adev), (ih)) int amdgpu_ih_ring_init(struct amdgpu_device *adev, struct amdgpu_ih_ring *ih, unsigned ring_size, bool use_bus_addr); void amdgpu_ih_ring_fini(struct amdgpu_device *adev, struct amdgpu_ih_ring *ih); -int amdgpu_ih_process(struct amdgpu_device *adev, struct amdgpu_ih_ring *ih, - void (*callback)(struct amdgpu_device *adev, - struct amdgpu_ih_ring *ih)); +int amdgpu_ih_process(struct amdgpu_device *adev, struct amdgpu_ih_ring *ih); #endif diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c index 52c17f6219a7..af4c3b1af322 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c @@ -94,23 +94,6 @@ static void amdgpu_hotplug_work_func(struct work_struct *work) } /** - * amdgpu_irq_reset_work_func - execute GPU reset - * - * @work: work struct pointer - * - * Execute scheduled GPU reset (Cayman+). - * This function is called when the IRQ handler thinks we need a GPU reset. - */ -static void amdgpu_irq_reset_work_func(struct work_struct *work) -{ - struct amdgpu_device *adev = container_of(work, struct amdgpu_device, - reset_work); - - if (!amdgpu_sriov_vf(adev) && amdgpu_device_should_recover_gpu(adev)) - amdgpu_device_gpu_recover(adev, NULL); -} - -/** * amdgpu_irq_disable_all - disable *all* interrupts * * @adev: amdgpu device pointer @@ -148,34 +131,6 @@ void amdgpu_irq_disable_all(struct amdgpu_device *adev) } /** - * amdgpu_irq_callback - callback from the IH ring - * - * @adev: amdgpu device pointer - * @ih: amdgpu ih ring - * - * Callback from IH ring processing to handle the entry at the current position - * and advance the read pointer. - */ -static void amdgpu_irq_callback(struct amdgpu_device *adev, - struct amdgpu_ih_ring *ih) -{ - u32 ring_index = ih->rptr >> 2; - struct amdgpu_iv_entry entry; - - /* Prescreening of high-frequency interrupts */ - if (!amdgpu_ih_prescreen_iv(adev)) - return; - - /* Before dispatching irq to IP blocks, send it to amdkfd */ - amdgpu_amdkfd_interrupt(adev, (const void *) &ih->ring[ring_index]); - - entry.iv_entry = (const uint32_t *)&ih->ring[ring_index]; - amdgpu_ih_decode_iv(adev, &entry); - - amdgpu_irq_dispatch(adev, &entry); -} - -/** * amdgpu_irq_handler - IRQ handler * * @irq: IRQ number (unused) @@ -192,13 +147,43 @@ irqreturn_t amdgpu_irq_handler(int irq, void *arg) struct amdgpu_device *adev = dev->dev_private; irqreturn_t ret; - ret = amdgpu_ih_process(adev, &adev->irq.ih, amdgpu_irq_callback); + ret = amdgpu_ih_process(adev, &adev->irq.ih); if (ret == IRQ_HANDLED) pm_runtime_mark_last_busy(dev->dev); return ret; } /** + * amdgpu_irq_handle_ih1 - kick of processing for IH1 + * + * @work: work structure in struct amdgpu_irq + * + * Kick of processing IH ring 1. + */ +static void amdgpu_irq_handle_ih1(struct work_struct *work) +{ + struct amdgpu_device *adev = container_of(work, struct amdgpu_device, + irq.ih1_work); + + amdgpu_ih_process(adev, &adev->irq.ih1); +} + +/** + * amdgpu_irq_handle_ih2 - kick of processing for IH2 + * + * @work: work structure in struct amdgpu_irq + * + * Kick of processing IH ring 2. + */ +static void amdgpu_irq_handle_ih2(struct work_struct *work) +{ + struct amdgpu_device *adev = container_of(work, struct amdgpu_device, + irq.ih2_work); + + amdgpu_ih_process(adev, &adev->irq.ih2); +} + +/** * amdgpu_msi_ok - check whether MSI functionality is enabled * * @adev: amdgpu device pointer (unused) @@ -262,7 +247,8 @@ int amdgpu_irq_init(struct amdgpu_device *adev) amdgpu_hotplug_work_func); } - INIT_WORK(&adev->reset_work, amdgpu_irq_reset_work_func); + INIT_WORK(&adev->irq.ih1_work, amdgpu_irq_handle_ih1); + INIT_WORK(&adev->irq.ih2_work, amdgpu_irq_handle_ih2); adev->irq.installed = true; r = drm_irq_install(adev->ddev, adev->ddev->pdev->irq); @@ -270,7 +256,6 @@ int amdgpu_irq_init(struct amdgpu_device *adev) adev->irq.installed = false; if (!amdgpu_device_has_dc_support(adev)) flush_work(&adev->hotplug_work); - cancel_work_sync(&adev->reset_work); return r; } adev->ddev->max_vblank_count = 0x00ffffff; @@ -299,7 +284,6 @@ void amdgpu_irq_fini(struct amdgpu_device *adev) pci_disable_msi(adev->pdev); if (!amdgpu_device_has_dc_support(adev)) flush_work(&adev->hotplug_work); - cancel_work_sync(&adev->reset_work); } for (i = 0; i < AMDGPU_IRQ_CLIENTID_MAX; ++i) { @@ -387,44 +371,50 @@ int amdgpu_irq_add_id(struct amdgpu_device *adev, * Dispatches IRQ to IP blocks. */ void amdgpu_irq_dispatch(struct amdgpu_device *adev, - struct amdgpu_iv_entry *entry) + struct amdgpu_ih_ring *ih) { - unsigned client_id = entry->client_id; - unsigned src_id = entry->src_id; + u32 ring_index = ih->rptr >> 2; + struct amdgpu_iv_entry entry; + unsigned client_id, src_id; struct amdgpu_irq_src *src; + bool handled = false; int r; - trace_amdgpu_iv(entry); + entry.iv_entry = (const uint32_t *)&ih->ring[ring_index]; + amdgpu_ih_decode_iv(adev, &entry); + + trace_amdgpu_iv(ih - &adev->irq.ih, &entry); + + client_id = entry.client_id; + src_id = entry.src_id; if (client_id >= AMDGPU_IRQ_CLIENTID_MAX) { DRM_DEBUG("Invalid client_id in IV: %d\n", client_id); - return; - } - if (src_id >= AMDGPU_MAX_IRQ_SRC_ID) { + } else if (src_id >= AMDGPU_MAX_IRQ_SRC_ID) { DRM_DEBUG("Invalid src_id in IV: %d\n", src_id); - return; - } - if (adev->irq.virq[src_id]) { + } else if (adev->irq.virq[src_id]) { generic_handle_irq(irq_find_mapping(adev->irq.domain, src_id)); - } else { - if (!adev->irq.client[client_id].sources) { - DRM_DEBUG("Unregistered interrupt client_id: %d src_id: %d\n", - client_id, src_id); - return; - } - src = adev->irq.client[client_id].sources[src_id]; - if (!src) { - DRM_DEBUG("Unhandled interrupt src_id: %d\n", src_id); - return; - } + } else if (!adev->irq.client[client_id].sources) { + DRM_DEBUG("Unregistered interrupt client_id: %d src_id: %d\n", + client_id, src_id); - r = src->funcs->process(adev, src, entry); - if (r) + } else if ((src = adev->irq.client[client_id].sources[src_id])) { + r = src->funcs->process(adev, src, &entry); + if (r < 0) DRM_ERROR("error processing interrupt (%d)\n", r); + else if (r) + handled = true; + + } else { + DRM_DEBUG("Unhandled interrupt src_id: %d\n", src_id); } + + /* Send it to amdkfd as well if it isn't already handled */ + if (!handled) + amdgpu_amdkfd_interrupt(adev, entry.iv_entry); } /** diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.h index f6ce171cb8aa..c718e94a55c9 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.h @@ -87,9 +87,11 @@ struct amdgpu_irq { /* status, etc. */ bool msi_enabled; /* msi enabled */ - /* interrupt ring */ - struct amdgpu_ih_ring ih; - const struct amdgpu_ih_funcs *ih_funcs; + /* interrupt rings */ + struct amdgpu_ih_ring ih, ih1, ih2; + const struct amdgpu_ih_funcs *ih_funcs; + struct work_struct ih1_work, ih2_work; + struct amdgpu_irq_src self_irq; /* gen irq stuff */ struct irq_domain *domain; /* GPU irq controller domain */ @@ -106,7 +108,7 @@ int amdgpu_irq_add_id(struct amdgpu_device *adev, unsigned client_id, unsigned src_id, struct amdgpu_irq_src *source); void amdgpu_irq_dispatch(struct amdgpu_device *adev, - struct amdgpu_iv_entry *entry); + struct amdgpu_ih_ring *ih); int amdgpu_irq_update(struct amdgpu_device *adev, struct amdgpu_irq_src *src, unsigned type); int amdgpu_irq_get(struct amdgpu_device *adev, struct amdgpu_irq_src *src, diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c index 755f733bf0d9..0a17fb1af204 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c @@ -32,6 +32,9 @@ static void amdgpu_job_timedout(struct drm_sched_job *s_job) { struct amdgpu_ring *ring = to_amdgpu_ring(s_job->sched); struct amdgpu_job *job = to_amdgpu_job(s_job); + struct amdgpu_task_info ti; + + memset(&ti, 0, sizeof(struct amdgpu_task_info)); if (amdgpu_ring_soft_recovery(ring, job->vmid, s_job->s_fence->parent)) { DRM_ERROR("ring %s timeout, but soft recovered\n", @@ -39,9 +42,12 @@ static void amdgpu_job_timedout(struct drm_sched_job *s_job) return; } + amdgpu_vm_get_task_info(ring->adev, job->pasid, &ti); DRM_ERROR("ring %s timeout, signaled seq=%u, emitted seq=%u\n", job->base.sched->name, atomic_read(&ring->fence_drv.last_seq), ring->fence_drv.sync_seq); + DRM_ERROR("Process information: process %s pid %d thread %s pid %d\n", + 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); @@ -112,6 +118,8 @@ static void amdgpu_job_free_cb(struct drm_sched_job *s_job) struct amdgpu_ring *ring = to_amdgpu_ring(s_job->sched); struct amdgpu_job *job = to_amdgpu_job(s_job); + drm_sched_job_cleanup(s_job); + amdgpu_ring_priority_put(ring, s_job->s_priority); dma_fence_put(job->fence); amdgpu_sync_free(&job->sync); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_job.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_job.h index 57cfe78a262b..e1b46a6703de 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_job.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_job.h @@ -33,6 +33,8 @@ #define to_amdgpu_job(sched_job) \ container_of((sched_job), struct amdgpu_job, base) +#define AMDGPU_JOB_GET_VMID(job) ((job) ? (job)->vmid : 0) + struct amdgpu_fence; struct amdgpu_job { diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c index 8f3d44e5e787..e860412043bb 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c @@ -207,11 +207,12 @@ int amdgpu_driver_load_kms(struct drm_device *dev, unsigned long flags) if (!r) { acpi_status = amdgpu_acpi_init(adev); if (acpi_status) - dev_dbg(&dev->pdev->dev, + dev_dbg(&dev->pdev->dev, "Error during ACPI methods call\n"); } if (amdgpu_device_is_px(dev)) { + dev_pm_set_driver_flags(dev->dev, DPM_FLAG_NEVER_SKIP); pm_runtime_use_autosuspend(dev->dev); pm_runtime_set_autosuspend_delay(dev->dev, 5000); pm_runtime_set_active(dev->dev); @@ -336,7 +337,7 @@ static int amdgpu_hw_ip_info(struct amdgpu_device *adev, case AMDGPU_HW_IP_GFX: type = AMD_IP_BLOCK_TYPE_GFX; for (i = 0; i < adev->gfx.num_gfx_rings; i++) - if (adev->gfx.gfx_ring[i].ready) + if (adev->gfx.gfx_ring[i].sched.ready) ++num_rings; ib_start_alignment = 32; ib_size_alignment = 32; @@ -344,7 +345,7 @@ static int amdgpu_hw_ip_info(struct amdgpu_device *adev, case AMDGPU_HW_IP_COMPUTE: type = AMD_IP_BLOCK_TYPE_GFX; for (i = 0; i < adev->gfx.num_compute_rings; i++) - if (adev->gfx.compute_ring[i].ready) + if (adev->gfx.compute_ring[i].sched.ready) ++num_rings; ib_start_alignment = 32; ib_size_alignment = 32; @@ -352,7 +353,7 @@ static int amdgpu_hw_ip_info(struct amdgpu_device *adev, case AMDGPU_HW_IP_DMA: type = AMD_IP_BLOCK_TYPE_SDMA; for (i = 0; i < adev->sdma.num_instances; i++) - if (adev->sdma.instance[i].ring.ready) + if (adev->sdma.instance[i].ring.sched.ready) ++num_rings; ib_start_alignment = 256; ib_size_alignment = 4; @@ -363,7 +364,7 @@ static int amdgpu_hw_ip_info(struct amdgpu_device *adev, if (adev->uvd.harvest_config & (1 << i)) continue; - if (adev->uvd.inst[i].ring.ready) + if (adev->uvd.inst[i].ring.sched.ready) ++num_rings; } ib_start_alignment = 64; @@ -372,7 +373,7 @@ static int amdgpu_hw_ip_info(struct amdgpu_device *adev, case AMDGPU_HW_IP_VCE: type = AMD_IP_BLOCK_TYPE_VCE; for (i = 0; i < adev->vce.num_rings; i++) - if (adev->vce.ring[i].ready) + if (adev->vce.ring[i].sched.ready) ++num_rings; ib_start_alignment = 4; ib_size_alignment = 1; @@ -384,7 +385,7 @@ static int amdgpu_hw_ip_info(struct amdgpu_device *adev, continue; for (j = 0; j < adev->uvd.num_enc_rings; j++) - if (adev->uvd.inst[i].ring_enc[j].ready) + if (adev->uvd.inst[i].ring_enc[j].sched.ready) ++num_rings; } ib_start_alignment = 64; @@ -392,7 +393,7 @@ static int amdgpu_hw_ip_info(struct amdgpu_device *adev, break; case AMDGPU_HW_IP_VCN_DEC: type = AMD_IP_BLOCK_TYPE_VCN; - if (adev->vcn.ring_dec.ready) + if (adev->vcn.ring_dec.sched.ready) ++num_rings; ib_start_alignment = 16; ib_size_alignment = 16; @@ -400,14 +401,14 @@ static int amdgpu_hw_ip_info(struct amdgpu_device *adev, case AMDGPU_HW_IP_VCN_ENC: type = AMD_IP_BLOCK_TYPE_VCN; for (i = 0; i < adev->vcn.num_enc_rings; i++) - if (adev->vcn.ring_enc[i].ready) + if (adev->vcn.ring_enc[i].sched.ready) ++num_rings; ib_start_alignment = 64; ib_size_alignment = 1; break; case AMDGPU_HW_IP_VCN_JPEG: type = AMD_IP_BLOCK_TYPE_VCN; - if (adev->vcn.ring_jpeg.ready) + if (adev->vcn.ring_jpeg.sched.ready) ++num_rings; ib_start_alignment = 16; ib_size_alignment = 16; @@ -978,7 +979,10 @@ int amdgpu_driver_open_kms(struct drm_device *dev, struct drm_file *file_priv) } if (amdgpu_sriov_vf(adev)) { - r = amdgpu_map_static_csa(adev, &fpriv->vm, &fpriv->csa_va); + uint64_t csa_addr = amdgpu_csa_vaddr(adev) & AMDGPU_GMC_HOLE_MASK; + + r = amdgpu_map_static_csa(adev, &fpriv->vm, adev->virt.csa_obj, + &fpriv->csa_va, csa_addr, AMDGPU_CSA_SIZE); if (r) goto error_vm; } @@ -1048,8 +1052,8 @@ void amdgpu_driver_postclose_kms(struct drm_device *dev, pasid = fpriv->vm.pasid; pd = amdgpu_bo_ref(fpriv->vm.root.base.bo); - amdgpu_vm_fini(adev, &fpriv->vm); amdgpu_ctx_mgr_fini(&fpriv->ctx_mgr); + amdgpu_vm_fini(adev, &fpriv->vm); if (pasid) amdgpu_pasid_free_delayed(pd->tbo.resv, pasid); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_mn.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_mn.c index e55508b39496..3e6823fdd939 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_mn.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_mn.c @@ -238,44 +238,40 @@ static void amdgpu_mn_invalidate_node(struct amdgpu_mn_node *node, * amdgpu_mn_invalidate_range_start_gfx - callback to notify about mm change * * @mn: our notifier - * @mm: the mm this callback is about - * @start: start of updated range - * @end: end of updated range + * @range: mmu notifier context * * Block for operations on BOs to finish and mark pages as accessed and * potentially dirty. */ static int amdgpu_mn_invalidate_range_start_gfx(struct mmu_notifier *mn, - struct mm_struct *mm, - unsigned long start, - unsigned long end, - bool blockable) + const struct mmu_notifier_range *range) { struct amdgpu_mn *amn = container_of(mn, struct amdgpu_mn, mn); struct interval_tree_node *it; + unsigned long end; /* notification is exclusive, but interval is inclusive */ - end -= 1; + end = range->end - 1; /* TODO we should be able to split locking for interval tree and * amdgpu_mn_invalidate_node */ - if (amdgpu_mn_read_lock(amn, blockable)) + if (amdgpu_mn_read_lock(amn, range->blockable)) return -EAGAIN; - it = interval_tree_iter_first(&amn->objects, start, end); + it = interval_tree_iter_first(&amn->objects, range->start, end); while (it) { struct amdgpu_mn_node *node; - if (!blockable) { + if (!range->blockable) { amdgpu_mn_read_unlock(amn); return -EAGAIN; } node = container_of(it, struct amdgpu_mn_node, it); - it = interval_tree_iter_next(it, start, end); + it = interval_tree_iter_next(it, range->start, end); - amdgpu_mn_invalidate_node(node, start, end); + amdgpu_mn_invalidate_node(node, range->start, end); } return 0; @@ -294,39 +290,38 @@ static int amdgpu_mn_invalidate_range_start_gfx(struct mmu_notifier *mn, * are restorted in amdgpu_mn_invalidate_range_end_hsa. */ static int amdgpu_mn_invalidate_range_start_hsa(struct mmu_notifier *mn, - struct mm_struct *mm, - unsigned long start, - unsigned long end, - bool blockable) + const struct mmu_notifier_range *range) { struct amdgpu_mn *amn = container_of(mn, struct amdgpu_mn, mn); struct interval_tree_node *it; + unsigned long end; /* notification is exclusive, but interval is inclusive */ - end -= 1; + end = range->end - 1; - if (amdgpu_mn_read_lock(amn, blockable)) + if (amdgpu_mn_read_lock(amn, range->blockable)) return -EAGAIN; - it = interval_tree_iter_first(&amn->objects, start, end); + it = interval_tree_iter_first(&amn->objects, range->start, end); while (it) { struct amdgpu_mn_node *node; struct amdgpu_bo *bo; - if (!blockable) { + if (!range->blockable) { amdgpu_mn_read_unlock(amn); return -EAGAIN; } node = container_of(it, struct amdgpu_mn_node, it); - it = interval_tree_iter_next(it, start, end); + it = interval_tree_iter_next(it, range->start, end); list_for_each_entry(bo, &node->bos, mn_list) { struct kgd_mem *mem = bo->kfd_bo; if (amdgpu_ttm_tt_affect_userptr(bo->tbo.ttm, - start, end)) - amdgpu_amdkfd_evict_userptr(mem, mm); + range->start, + end)) + amdgpu_amdkfd_evict_userptr(mem, range->mm); } } @@ -344,9 +339,7 @@ static int amdgpu_mn_invalidate_range_start_hsa(struct mmu_notifier *mn, * Release the lock again to allow new command submissions. */ static void amdgpu_mn_invalidate_range_end(struct mmu_notifier *mn, - struct mm_struct *mm, - unsigned long start, - unsigned long end) + const struct mmu_notifier_range *range) { struct amdgpu_mn *amn = container_of(mn, struct amdgpu_mn, mn); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_mode.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_mode.h index d1b4d9b6aae0..889e443eeee7 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_mode.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_mode.h @@ -38,7 +38,7 @@ #include <drm/drm_crtc_helper.h> #include <drm/drm_fb_helper.h> #include <drm/drm_plane_helper.h> -#include <drm/drm_fb_helper.h> +#include <drm/drm_probe_helper.h> #include <linux/i2c.h> #include <linux/i2c-algo-bit.h> #include <linux/hrtimer.h> @@ -57,7 +57,6 @@ struct amdgpu_hpd; #define to_amdgpu_connector(x) container_of(x, struct amdgpu_connector, base) #define to_amdgpu_encoder(x) container_of(x, struct amdgpu_encoder, base) #define to_amdgpu_framebuffer(x) container_of(x, struct amdgpu_framebuffer, base) -#define to_amdgpu_plane(x) container_of(x, struct amdgpu_plane, base) #define to_dm_plane_state(x) container_of(x, struct dm_plane_state, base); @@ -295,13 +294,6 @@ struct amdgpu_display_funcs { uint16_t connector_object_id, struct amdgpu_hpd *hpd, struct amdgpu_router *router); - /* it is used to enter or exit into free sync mode */ - int (*notify_freesync)(struct drm_device *dev, void *data, - struct drm_file *filp); - /* it is used to allow enablement of freesync mode */ - int (*set_freesync_property)(struct drm_connector *connector, - struct drm_property *property, - uint64_t val); }; @@ -325,7 +317,7 @@ struct amdgpu_mode_info { struct card_info *atom_card_info; bool mode_config_initialized; struct amdgpu_crtc *crtcs[AMDGPU_MAX_CRTCS]; - struct amdgpu_plane *planes[AMDGPU_MAX_PLANES]; + struct drm_plane *planes[AMDGPU_MAX_PLANES]; struct amdgpu_afmt *afmt[AMDGPU_MAX_AFMT_BLOCKS]; /* DVI-I properties */ struct drm_property *coherent_mode_property; @@ -341,6 +333,8 @@ struct amdgpu_mode_info { struct drm_property *dither_property; /* maximum number of bits per channel for monitor color */ struct drm_property *max_bpc_property; + /* Adaptive Backlight Modulation (power feature) */ + struct drm_property *abm_level_property; /* hardcoded DFP edid from BIOS */ struct edid *bios_hardcoded_edid; int bios_hardcoded_edid_size; @@ -412,6 +406,7 @@ struct amdgpu_crtc { struct amdgpu_flip_work *pflip_works; enum amdgpu_flip_status pflip_status; int deferred_flip_completion; + u64 last_flip_vblank; /* pll sharing */ struct amdgpu_atom_ss ss; bool ss_enabled; @@ -436,11 +431,6 @@ struct amdgpu_crtc { struct drm_pending_vblank_event *event; }; -struct amdgpu_plane { - struct drm_plane base; - enum drm_plane_type plane_type; -}; - struct amdgpu_encoder_atom_dig { bool linkb; /* atom dig */ diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c index 904014dc5915..ec9e45004bff 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c @@ -81,7 +81,7 @@ static void amdgpu_bo_destroy(struct ttm_buffer_object *tbo) amdgpu_bo_subtract_pin_size(bo); if (bo->kfd_bo) - amdgpu_amdkfd_unreserve_system_memory_limit(bo); + amdgpu_amdkfd_unreserve_memory_limit(bo); amdgpu_bo_kunmap(bo); @@ -426,12 +426,20 @@ static int amdgpu_bo_do_create(struct amdgpu_device *adev, size_t acc_size; int r; - page_align = roundup(bp->byte_align, PAGE_SIZE) >> PAGE_SHIFT; - if (bp->domain & (AMDGPU_GEM_DOMAIN_GDS | AMDGPU_GEM_DOMAIN_GWS | - AMDGPU_GEM_DOMAIN_OA)) + /* Note that GDS/GWS/OA allocates 1 page per byte/resource. */ + if (bp->domain & (AMDGPU_GEM_DOMAIN_GWS | AMDGPU_GEM_DOMAIN_OA)) { + /* GWS and OA don't need any alignment. */ + page_align = bp->byte_align; size <<= PAGE_SHIFT; - else + } else if (bp->domain & AMDGPU_GEM_DOMAIN_GDS) { + /* Both size and alignment must be a multiple of 4. */ + page_align = ALIGN(bp->byte_align, 4); + size = ALIGN(size, 4) << PAGE_SHIFT; + } else { + /* Memory should be aligned at least to a page size. */ + page_align = ALIGN(bp->byte_align, PAGE_SIZE) >> PAGE_SHIFT; size = ALIGN(size, PAGE_SIZE); + } if (!amdgpu_bo_validate_size(adev, size, bp->domain)) return -ENOMEM; @@ -608,53 +616,6 @@ int amdgpu_bo_create(struct amdgpu_device *adev, } /** - * amdgpu_bo_backup_to_shadow - Backs up an &amdgpu_bo buffer object - * @adev: amdgpu device object - * @ring: amdgpu_ring for the engine handling the buffer operations - * @bo: &amdgpu_bo buffer to be backed up - * @resv: reservation object with embedded fence - * @fence: dma_fence associated with the operation - * @direct: whether to submit the job directly - * - * Copies an &amdgpu_bo buffer object to its shadow object. - * Not used for now. - * - * Returns: - * 0 for success or a negative error code on failure. - */ -int amdgpu_bo_backup_to_shadow(struct amdgpu_device *adev, - struct amdgpu_ring *ring, - struct amdgpu_bo *bo, - struct reservation_object *resv, - struct dma_fence **fence, - bool direct) - -{ - struct amdgpu_bo *shadow = bo->shadow; - uint64_t bo_addr, shadow_addr; - int r; - - if (!shadow) - return -EINVAL; - - bo_addr = amdgpu_bo_gpu_offset(bo); - shadow_addr = amdgpu_bo_gpu_offset(bo->shadow); - - r = reservation_object_reserve_shared(bo->tbo.resv); - if (r) - goto err; - - r = amdgpu_copy_buffer(ring, bo_addr, shadow_addr, - amdgpu_bo_size(bo), resv, fence, - direct, false); - if (!r) - amdgpu_bo_fence(bo, *fence, true); - -err: - return r; -} - -/** * amdgpu_bo_validate - validate an &amdgpu_bo buffer object * @bo: pointer to the buffer object * @@ -959,7 +920,7 @@ int amdgpu_bo_unpin(struct amdgpu_bo *bo) struct ttm_operation_ctx ctx = { false, false }; int r, i; - if (!bo->pin_count) { + if (WARN_ON_ONCE(!bo->pin_count)) { dev_warn(adev->dev, "%p unpin not necessary\n", bo); return 0; } @@ -1324,6 +1285,30 @@ void amdgpu_bo_fence(struct amdgpu_bo *bo, struct dma_fence *fence, } /** + * amdgpu_sync_wait_resv - Wait for BO reservation fences + * + * @bo: buffer object + * @owner: fence owner + * @intr: Whether the wait is interruptible + * + * Returns: + * 0 on success, errno otherwise. + */ +int amdgpu_bo_sync_wait(struct amdgpu_bo *bo, void *owner, bool intr) +{ + struct amdgpu_device *adev = amdgpu_ttm_adev(bo->tbo.bdev); + struct amdgpu_sync sync; + int r; + + amdgpu_sync_create(&sync); + amdgpu_sync_resv(adev, &sync, bo->tbo.resv, owner, false); + r = amdgpu_sync_wait(&sync, intr); + amdgpu_sync_free(&sync); + + return r; +} + +/** * amdgpu_bo_gpu_offset - return GPU offset of bo * @bo: amdgpu object for which we query the offset * diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h index 7d3312d0da11..220a6a7b1bc1 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h @@ -266,12 +266,8 @@ void amdgpu_bo_move_notify(struct ttm_buffer_object *bo, int amdgpu_bo_fault_reserve_notify(struct ttm_buffer_object *bo); void amdgpu_bo_fence(struct amdgpu_bo *bo, struct dma_fence *fence, bool shared); +int amdgpu_bo_sync_wait(struct amdgpu_bo *bo, void *owner, bool intr); u64 amdgpu_bo_gpu_offset(struct amdgpu_bo *bo); -int amdgpu_bo_backup_to_shadow(struct amdgpu_device *adev, - struct amdgpu_ring *ring, - struct amdgpu_bo *bo, - struct reservation_object *resv, - struct dma_fence **fence, bool direct); int amdgpu_bo_validate(struct amdgpu_bo *bo); int amdgpu_bo_restore_shadow(struct amdgpu_bo *shadow, struct dma_fence **fence); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_pm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_pm.c index 59cc678de8c1..a7adb7b6bd98 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_pm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_pm.c @@ -33,6 +33,8 @@ #include <linux/hwmon.h> #include <linux/hwmon-sysfs.h> #include <linux/nospec.h> +#include "hwmgr.h" +#define WIDTH_4K 3840 static int amdgpu_debugfs_pm_init(struct amdgpu_device *adev); @@ -624,11 +626,71 @@ static ssize_t amdgpu_get_pp_od_clk_voltage(struct device *dev, } /** - * DOC: pp_dpm_sclk pp_dpm_mclk pp_dpm_pcie + * DOC: ppfeatures + * + * The amdgpu driver provides a sysfs API for adjusting what powerplay + * features to be enabled. The file ppfeatures is used for this. And + * this is only available for Vega10 and later dGPUs. + * + * Reading back the file will show you the followings: + * - Current ppfeature masks + * - List of the all supported powerplay features with their naming, + * bitmasks and enablement status('Y'/'N' means "enabled"/"disabled"). + * + * To manually enable or disable a specific feature, just set or clear + * the corresponding bit from original ppfeature masks and input the + * new ppfeature masks. + */ +static ssize_t amdgpu_set_ppfeature_status(struct device *dev, + struct device_attribute *attr, + const char *buf, + size_t count) +{ + struct drm_device *ddev = dev_get_drvdata(dev); + struct amdgpu_device *adev = ddev->dev_private; + uint64_t featuremask; + int ret; + + ret = kstrtou64(buf, 0, &featuremask); + if (ret) + return -EINVAL; + + pr_debug("featuremask = 0x%llx\n", featuremask); + + if (adev->powerplay.pp_funcs->set_ppfeature_status) { + ret = amdgpu_dpm_set_ppfeature_status(adev, featuremask); + if (ret) + return -EINVAL; + } + + return count; +} + +static ssize_t amdgpu_get_ppfeature_status(struct device *dev, + struct device_attribute *attr, + char *buf) +{ + struct drm_device *ddev = dev_get_drvdata(dev); + struct amdgpu_device *adev = ddev->dev_private; + + if (adev->powerplay.pp_funcs->get_ppfeature_status) + return amdgpu_dpm_get_ppfeature_status(adev, buf); + + return snprintf(buf, PAGE_SIZE, "\n"); +} + +/** + * DOC: pp_dpm_sclk pp_dpm_mclk pp_dpm_socclk pp_dpm_fclk pp_dpm_dcefclk + * pp_dpm_pcie * * The amdgpu driver provides a sysfs API for adjusting what power levels * are enabled for a given power state. The files pp_dpm_sclk, pp_dpm_mclk, - * and pp_dpm_pcie are used for this. + * pp_dpm_socclk, pp_dpm_fclk, pp_dpm_dcefclk and pp_dpm_pcie are used for + * this. + * + * pp_dpm_socclk and pp_dpm_dcefclk interfaces are only available for + * Vega10 and later ASICs. + * pp_dpm_fclk interface is only available for Vega20 and later ASICs. * * Reading back the files will show you the available power levels within * the power state and the clock information for those levels. @@ -638,6 +700,8 @@ static ssize_t amdgpu_get_pp_od_clk_voltage(struct device *dev, * Secondly,Enter a new value for each level by inputing a string that * contains " echo xx xx xx > pp_dpm_sclk/mclk/pcie" * E.g., echo 4 5 6 to > pp_dpm_sclk will enable sclk levels 4, 5, and 6. + * + * NOTE: change to the dcefclk max dpm level is not supported now */ static ssize_t amdgpu_get_pp_dpm_sclk(struct device *dev, @@ -748,6 +812,114 @@ static ssize_t amdgpu_set_pp_dpm_mclk(struct device *dev, return count; } +static ssize_t amdgpu_get_pp_dpm_socclk(struct device *dev, + struct device_attribute *attr, + char *buf) +{ + struct drm_device *ddev = dev_get_drvdata(dev); + struct amdgpu_device *adev = ddev->dev_private; + + if (adev->powerplay.pp_funcs->print_clock_levels) + return amdgpu_dpm_print_clock_levels(adev, PP_SOCCLK, buf); + else + return snprintf(buf, PAGE_SIZE, "\n"); +} + +static ssize_t amdgpu_set_pp_dpm_socclk(struct device *dev, + struct device_attribute *attr, + const char *buf, + size_t count) +{ + struct drm_device *ddev = dev_get_drvdata(dev); + struct amdgpu_device *adev = ddev->dev_private; + int ret; + uint32_t mask = 0; + + ret = amdgpu_read_mask(buf, count, &mask); + if (ret) + return ret; + + if (adev->powerplay.pp_funcs->force_clock_level) + ret = amdgpu_dpm_force_clock_level(adev, PP_SOCCLK, mask); + + if (ret) + return -EINVAL; + + return count; +} + +static ssize_t amdgpu_get_pp_dpm_fclk(struct device *dev, + struct device_attribute *attr, + char *buf) +{ + struct drm_device *ddev = dev_get_drvdata(dev); + struct amdgpu_device *adev = ddev->dev_private; + + if (adev->powerplay.pp_funcs->print_clock_levels) + return amdgpu_dpm_print_clock_levels(adev, PP_FCLK, buf); + else + return snprintf(buf, PAGE_SIZE, "\n"); +} + +static ssize_t amdgpu_set_pp_dpm_fclk(struct device *dev, + struct device_attribute *attr, + const char *buf, + size_t count) +{ + struct drm_device *ddev = dev_get_drvdata(dev); + struct amdgpu_device *adev = ddev->dev_private; + int ret; + uint32_t mask = 0; + + ret = amdgpu_read_mask(buf, count, &mask); + if (ret) + return ret; + + if (adev->powerplay.pp_funcs->force_clock_level) + ret = amdgpu_dpm_force_clock_level(adev, PP_FCLK, mask); + + if (ret) + return -EINVAL; + + return count; +} + +static ssize_t amdgpu_get_pp_dpm_dcefclk(struct device *dev, + struct device_attribute *attr, + char *buf) +{ + struct drm_device *ddev = dev_get_drvdata(dev); + struct amdgpu_device *adev = ddev->dev_private; + + if (adev->powerplay.pp_funcs->print_clock_levels) + return amdgpu_dpm_print_clock_levels(adev, PP_DCEFCLK, buf); + else + return snprintf(buf, PAGE_SIZE, "\n"); +} + +static ssize_t amdgpu_set_pp_dpm_dcefclk(struct device *dev, + struct device_attribute *attr, + const char *buf, + size_t count) +{ + struct drm_device *ddev = dev_get_drvdata(dev); + struct amdgpu_device *adev = ddev->dev_private; + int ret; + uint32_t mask = 0; + + ret = amdgpu_read_mask(buf, count, &mask); + if (ret) + return ret; + + if (adev->powerplay.pp_funcs->force_clock_level) + ret = amdgpu_dpm_force_clock_level(adev, PP_DCEFCLK, mask); + + if (ret) + return -EINVAL; + + return count; +} + static ssize_t amdgpu_get_pp_dpm_pcie(struct device *dev, struct device_attribute *attr, char *buf) @@ -988,6 +1160,31 @@ static ssize_t amdgpu_get_busy_percent(struct device *dev, return snprintf(buf, PAGE_SIZE, "%d\n", value); } +/** + * DOC: pcie_bw + * + * The amdgpu driver provides a sysfs API for estimating how much data + * has been received and sent by the GPU in the last second through PCIe. + * The file pcie_bw is used for this. + * The Perf counters count the number of received and sent messages and return + * those values, as well as the maximum payload size of a PCIe packet (mps). + * Note that it is not possible to easily and quickly obtain the size of each + * packet transmitted, so we output the max payload size (mps) to allow for + * quick estimation of the PCIe bandwidth usage + */ +static ssize_t amdgpu_get_pcie_bw(struct device *dev, + struct device_attribute *attr, + char *buf) +{ + struct drm_device *ddev = dev_get_drvdata(dev); + struct amdgpu_device *adev = ddev->dev_private; + uint64_t count0, count1; + + amdgpu_asic_get_pcie_usage(adev, &count0, &count1); + return snprintf(buf, PAGE_SIZE, "%llu %llu %i\n", + count0, count1, pcie_get_mps(adev->pdev)); +} + static DEVICE_ATTR(power_dpm_state, S_IRUGO | S_IWUSR, amdgpu_get_dpm_state, amdgpu_set_dpm_state); static DEVICE_ATTR(power_dpm_force_performance_level, S_IRUGO | S_IWUSR, amdgpu_get_dpm_forced_performance_level, @@ -1006,6 +1203,15 @@ static DEVICE_ATTR(pp_dpm_sclk, S_IRUGO | S_IWUSR, static DEVICE_ATTR(pp_dpm_mclk, S_IRUGO | S_IWUSR, amdgpu_get_pp_dpm_mclk, amdgpu_set_pp_dpm_mclk); +static DEVICE_ATTR(pp_dpm_socclk, S_IRUGO | S_IWUSR, + amdgpu_get_pp_dpm_socclk, + amdgpu_set_pp_dpm_socclk); +static DEVICE_ATTR(pp_dpm_fclk, S_IRUGO | S_IWUSR, + amdgpu_get_pp_dpm_fclk, + amdgpu_set_pp_dpm_fclk); +static DEVICE_ATTR(pp_dpm_dcefclk, S_IRUGO | S_IWUSR, + amdgpu_get_pp_dpm_dcefclk, + amdgpu_set_pp_dpm_dcefclk); static DEVICE_ATTR(pp_dpm_pcie, S_IRUGO | S_IWUSR, amdgpu_get_pp_dpm_pcie, amdgpu_set_pp_dpm_pcie); @@ -1023,6 +1229,10 @@ static DEVICE_ATTR(pp_od_clk_voltage, S_IRUGO | S_IWUSR, amdgpu_set_pp_od_clk_voltage); static DEVICE_ATTR(gpu_busy_percent, S_IRUGO, amdgpu_get_busy_percent, NULL); +static DEVICE_ATTR(pcie_bw, S_IRUGO, amdgpu_get_pcie_bw, NULL); +static DEVICE_ATTR(ppfeatures, S_IRUGO | S_IWUSR, + amdgpu_get_ppfeature_status, + amdgpu_set_ppfeature_status); static ssize_t amdgpu_hwmon_show_temp(struct device *dev, struct device_attribute *attr, @@ -1514,6 +1724,75 @@ static ssize_t amdgpu_hwmon_set_power_cap(struct device *dev, return count; } +static ssize_t amdgpu_hwmon_show_sclk(struct device *dev, + struct device_attribute *attr, + char *buf) +{ + struct amdgpu_device *adev = dev_get_drvdata(dev); + struct drm_device *ddev = adev->ddev; + uint32_t sclk; + int r, size = sizeof(sclk); + + /* Can't get voltage when the card is off */ + if ((adev->flags & AMD_IS_PX) && + (ddev->switch_power_state != DRM_SWITCH_POWER_ON)) + return -EINVAL; + + /* sanity check PP is enabled */ + if (!(adev->powerplay.pp_funcs && + adev->powerplay.pp_funcs->read_sensor)) + return -EINVAL; + + /* get the sclk */ + r = amdgpu_dpm_read_sensor(adev, AMDGPU_PP_SENSOR_GFX_SCLK, + (void *)&sclk, &size); + if (r) + return r; + + return snprintf(buf, PAGE_SIZE, "%d\n", sclk * 10 * 1000); +} + +static ssize_t amdgpu_hwmon_show_sclk_label(struct device *dev, + struct device_attribute *attr, + char *buf) +{ + return snprintf(buf, PAGE_SIZE, "sclk\n"); +} + +static ssize_t amdgpu_hwmon_show_mclk(struct device *dev, + struct device_attribute *attr, + char *buf) +{ + struct amdgpu_device *adev = dev_get_drvdata(dev); + struct drm_device *ddev = adev->ddev; + uint32_t mclk; + int r, size = sizeof(mclk); + + /* Can't get voltage when the card is off */ + if ((adev->flags & AMD_IS_PX) && + (ddev->switch_power_state != DRM_SWITCH_POWER_ON)) + return -EINVAL; + + /* sanity check PP is enabled */ + if (!(adev->powerplay.pp_funcs && + adev->powerplay.pp_funcs->read_sensor)) + return -EINVAL; + + /* get the sclk */ + r = amdgpu_dpm_read_sensor(adev, AMDGPU_PP_SENSOR_GFX_MCLK, + (void *)&mclk, &size); + if (r) + return r; + + return snprintf(buf, PAGE_SIZE, "%d\n", mclk * 10 * 1000); +} + +static ssize_t amdgpu_hwmon_show_mclk_label(struct device *dev, + struct device_attribute *attr, + char *buf) +{ + return snprintf(buf, PAGE_SIZE, "mclk\n"); +} /** * DOC: hwmon @@ -1530,6 +1809,10 @@ static ssize_t amdgpu_hwmon_set_power_cap(struct device *dev, * * - GPU fan * + * - GPU gfx/compute engine clock + * + * - GPU memory clock (dGPU only) + * * hwmon interfaces for GPU temperature: * * - temp1_input: the on die GPU temperature in millidegrees Celsius @@ -1574,6 +1857,12 @@ static ssize_t amdgpu_hwmon_set_power_cap(struct device *dev, * * - fan[1-*]_enable: Enable or disable the sensors.1: Enable 0: Disable * + * hwmon interfaces for GPU clocks: + * + * - freq1_input: the gfx/compute clock in hertz + * + * - freq2_input: the memory clock in hertz + * * You can use hwmon tools like sensors to view this information on your system. * */ @@ -1598,6 +1887,10 @@ static SENSOR_DEVICE_ATTR(power1_average, S_IRUGO, amdgpu_hwmon_show_power_avg, static SENSOR_DEVICE_ATTR(power1_cap_max, S_IRUGO, amdgpu_hwmon_show_power_cap_max, NULL, 0); static SENSOR_DEVICE_ATTR(power1_cap_min, S_IRUGO, amdgpu_hwmon_show_power_cap_min, NULL, 0); static SENSOR_DEVICE_ATTR(power1_cap, S_IRUGO | S_IWUSR, amdgpu_hwmon_show_power_cap, amdgpu_hwmon_set_power_cap, 0); +static SENSOR_DEVICE_ATTR(freq1_input, S_IRUGO, amdgpu_hwmon_show_sclk, NULL, 0); +static SENSOR_DEVICE_ATTR(freq1_label, S_IRUGO, amdgpu_hwmon_show_sclk_label, NULL, 0); +static SENSOR_DEVICE_ATTR(freq2_input, S_IRUGO, amdgpu_hwmon_show_mclk, NULL, 0); +static SENSOR_DEVICE_ATTR(freq2_label, S_IRUGO, amdgpu_hwmon_show_mclk_label, NULL, 0); static struct attribute *hwmon_attributes[] = { &sensor_dev_attr_temp1_input.dev_attr.attr, @@ -1620,6 +1913,10 @@ static struct attribute *hwmon_attributes[] = { &sensor_dev_attr_power1_cap_max.dev_attr.attr, &sensor_dev_attr_power1_cap_min.dev_attr.attr, &sensor_dev_attr_power1_cap.dev_attr.attr, + &sensor_dev_attr_freq1_input.dev_attr.attr, + &sensor_dev_attr_freq1_label.dev_attr.attr, + &sensor_dev_attr_freq2_input.dev_attr.attr, + &sensor_dev_attr_freq2_label.dev_attr.attr, NULL }; @@ -1642,6 +1939,19 @@ static umode_t hwmon_attributes_visible(struct kobject *kobj, attr == &sensor_dev_attr_fan1_enable.dev_attr.attr)) return 0; + /* Skip fan attributes on APU */ + if ((adev->flags & AMD_IS_APU) && + (attr == &sensor_dev_attr_pwm1.dev_attr.attr || + attr == &sensor_dev_attr_pwm1_enable.dev_attr.attr || + attr == &sensor_dev_attr_pwm1_max.dev_attr.attr || + attr == &sensor_dev_attr_pwm1_min.dev_attr.attr || + attr == &sensor_dev_attr_fan1_input.dev_attr.attr || + attr == &sensor_dev_attr_fan1_min.dev_attr.attr || + attr == &sensor_dev_attr_fan1_max.dev_attr.attr || + attr == &sensor_dev_attr_fan1_target.dev_attr.attr || + attr == &sensor_dev_attr_fan1_enable.dev_attr.attr)) + return 0; + /* Skip limit attributes if DPM is not enabled */ if (!adev->pm.dpm_enabled && (attr == &sensor_dev_attr_temp1_crit.dev_attr.attr || @@ -1671,7 +1981,8 @@ static umode_t hwmon_attributes_visible(struct kobject *kobj, effective_mode &= ~S_IWUSR; if ((adev->flags & AMD_IS_APU) && - (attr == &sensor_dev_attr_power1_cap_max.dev_attr.attr || + (attr == &sensor_dev_attr_power1_average.dev_attr.attr || + attr == &sensor_dev_attr_power1_cap_max.dev_attr.attr || attr == &sensor_dev_attr_power1_cap_min.dev_attr.attr|| attr == &sensor_dev_attr_power1_cap.dev_attr.attr)) return 0; @@ -1697,6 +2008,12 @@ static umode_t hwmon_attributes_visible(struct kobject *kobj, attr == &sensor_dev_attr_in1_label.dev_attr.attr)) return 0; + /* no mclk on APUs */ + if ((adev->flags & AMD_IS_APU) && + (attr == &sensor_dev_attr_freq2_input.dev_attr.attr || + attr == &sensor_dev_attr_freq2_label.dev_attr.attr)) + return 0; + return effective_mode; } @@ -1956,6 +2273,17 @@ void amdgpu_dpm_enable_uvd(struct amdgpu_device *adev, bool enable) amdgpu_dpm_set_powergating_by_smu(adev, AMD_IP_BLOCK_TYPE_UVD, !enable); mutex_unlock(&adev->pm.mutex); } + /* enable/disable Low Memory PState for UVD (4k videos) */ + if (adev->asic_type == CHIP_STONEY && + adev->uvd.decode_image_width >= WIDTH_4K) { + struct pp_hwmgr *hwmgr = adev->powerplay.pp_handle; + + if (hwmgr && hwmgr->hwmgr_func && + hwmgr->hwmgr_func->update_nbdpm_pstate) + hwmgr->hwmgr_func->update_nbdpm_pstate(hwmgr, + !enable, + true); + } } void amdgpu_dpm_enable_vce(struct amdgpu_device *adev, bool enable) @@ -1982,6 +2310,7 @@ void amdgpu_pm_print_power_states(struct amdgpu_device *adev) int amdgpu_pm_sysfs_init(struct amdgpu_device *adev) { + struct pp_hwmgr *hwmgr = adev->powerplay.pp_handle; int ret; if (adev->pm.sysfs_initialized) @@ -2043,6 +2372,25 @@ int amdgpu_pm_sysfs_init(struct amdgpu_device *adev) DRM_ERROR("failed to create device file pp_dpm_mclk\n"); return ret; } + if (adev->asic_type >= CHIP_VEGA10) { + ret = device_create_file(adev->dev, &dev_attr_pp_dpm_socclk); + if (ret) { + DRM_ERROR("failed to create device file pp_dpm_socclk\n"); + return ret; + } + ret = device_create_file(adev->dev, &dev_attr_pp_dpm_dcefclk); + if (ret) { + DRM_ERROR("failed to create device file pp_dpm_dcefclk\n"); + return ret; + } + } + if (adev->asic_type >= CHIP_VEGA20) { + ret = device_create_file(adev->dev, &dev_attr_pp_dpm_fclk); + if (ret) { + DRM_ERROR("failed to create device file pp_dpm_fclk\n"); + return ret; + } + } ret = device_create_file(adev->dev, &dev_attr_pp_dpm_pcie); if (ret) { DRM_ERROR("failed to create device file pp_dpm_pcie\n"); @@ -2065,12 +2413,14 @@ int amdgpu_pm_sysfs_init(struct amdgpu_device *adev) "pp_power_profile_mode\n"); return ret; } - ret = device_create_file(adev->dev, - &dev_attr_pp_od_clk_voltage); - if (ret) { - DRM_ERROR("failed to create device file " - "pp_od_clk_voltage\n"); - return ret; + if (hwmgr->od_enabled) { + ret = device_create_file(adev->dev, + &dev_attr_pp_od_clk_voltage); + if (ret) { + DRM_ERROR("failed to create device file " + "pp_od_clk_voltage\n"); + return ret; + } } ret = device_create_file(adev->dev, &dev_attr_gpu_busy_percent); @@ -2079,12 +2429,31 @@ int amdgpu_pm_sysfs_init(struct amdgpu_device *adev) "gpu_busy_level\n"); return ret; } + /* PCIe Perf counters won't work on APU nodes */ + if (!(adev->flags & AMD_IS_APU)) { + ret = device_create_file(adev->dev, &dev_attr_pcie_bw); + if (ret) { + DRM_ERROR("failed to create device file pcie_bw\n"); + return ret; + } + } ret = amdgpu_debugfs_pm_init(adev); if (ret) { DRM_ERROR("Failed to register debugfs file for dpm!\n"); return ret; } + if ((adev->asic_type >= CHIP_VEGA10) && + !(adev->flags & AMD_IS_APU)) { + ret = device_create_file(adev->dev, + &dev_attr_ppfeatures); + if (ret) { + DRM_ERROR("failed to create device file " + "ppfeatures\n"); + return ret; + } + } + adev->pm.sysfs_initialized = true; return 0; @@ -2092,6 +2461,8 @@ int amdgpu_pm_sysfs_init(struct amdgpu_device *adev) void amdgpu_pm_sysfs_fini(struct amdgpu_device *adev) { + struct pp_hwmgr *hwmgr = adev->powerplay.pp_handle; + if (adev->pm.dpm_enabled == 0) return; @@ -2107,14 +2478,26 @@ void amdgpu_pm_sysfs_fini(struct amdgpu_device *adev) device_remove_file(adev->dev, &dev_attr_pp_dpm_sclk); device_remove_file(adev->dev, &dev_attr_pp_dpm_mclk); + if (adev->asic_type >= CHIP_VEGA10) { + device_remove_file(adev->dev, &dev_attr_pp_dpm_socclk); + device_remove_file(adev->dev, &dev_attr_pp_dpm_dcefclk); + } device_remove_file(adev->dev, &dev_attr_pp_dpm_pcie); + if (adev->asic_type >= CHIP_VEGA20) + device_remove_file(adev->dev, &dev_attr_pp_dpm_fclk); device_remove_file(adev->dev, &dev_attr_pp_sclk_od); device_remove_file(adev->dev, &dev_attr_pp_mclk_od); device_remove_file(adev->dev, &dev_attr_pp_power_profile_mode); - device_remove_file(adev->dev, - &dev_attr_pp_od_clk_voltage); + if (hwmgr->od_enabled) + device_remove_file(adev->dev, + &dev_attr_pp_od_clk_voltage); device_remove_file(adev->dev, &dev_attr_gpu_busy_percent); + if (!(adev->flags & AMD_IS_APU)) + device_remove_file(adev->dev, &dev_attr_pcie_bw); + if ((adev->asic_type >= CHIP_VEGA10) && + !(adev->flags & AMD_IS_APU)) + device_remove_file(adev->dev, &dev_attr_ppfeatures); } void amdgpu_pm_compute_clocks(struct amdgpu_device *adev) @@ -2129,7 +2512,7 @@ void amdgpu_pm_compute_clocks(struct amdgpu_device *adev) for (i = 0; i < AMDGPU_MAX_RINGS; i++) { struct amdgpu_ring *ring = adev->rings[i]; - if (ring && ring->ready) + if (ring && ring->sched.ready) amdgpu_fence_wait_empty(ring); } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_prime.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_prime.c index e45e929aaab5..a38e0fb4a6fe 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_prime.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_prime.c @@ -38,8 +38,7 @@ #include "amdgpu_gem.h" #include <drm/amdgpu_drm.h> #include <linux/dma-buf.h> - -static const struct dma_buf_ops amdgpu_dmabuf_ops; +#include <linux/dma-fence-array.h> /** * amdgpu_gem_prime_get_sg_table - &drm_driver.gem_prime_get_sg_table @@ -189,6 +188,48 @@ error: return ERR_PTR(ret); } +static int +__reservation_object_make_exclusive(struct reservation_object *obj) +{ + struct dma_fence **fences; + unsigned int count; + int r; + + if (!reservation_object_get_list(obj)) /* no shared fences to convert */ + return 0; + + r = reservation_object_get_fences_rcu(obj, NULL, &count, &fences); + if (r) + return r; + + if (count == 0) { + /* Now that was unexpected. */ + } else if (count == 1) { + reservation_object_add_excl_fence(obj, fences[0]); + dma_fence_put(fences[0]); + kfree(fences); + } else { + struct dma_fence_array *array; + + array = dma_fence_array_create(count, fences, + dma_fence_context_alloc(1), 0, + false); + if (!array) + goto err_fences_put; + + reservation_object_add_excl_fence(obj, &array->base); + dma_fence_put(&array->base); + } + + return 0; + +err_fences_put: + while (count--) + dma_fence_put(fences[count]); + kfree(fences); + return -ENOMEM; +} + /** * amdgpu_gem_map_attach - &dma_buf_ops.attach implementation * @dma_buf: Shared DMA buffer @@ -220,16 +261,16 @@ static int amdgpu_gem_map_attach(struct dma_buf *dma_buf, if (attach->dev->driver != adev->dev->driver) { /* - * Wait for all shared fences to complete before we switch to future - * use of exclusive fence on this prime shared bo. + * We only create shared fences for internal use, but importers + * of the dmabuf rely on exclusive fences for implicitly + * tracking write hazards. As any of the current fences may + * correspond to a write, we need to convert all existing + * fences on the reservation object into a single exclusive + * fence. */ - r = reservation_object_wait_timeout_rcu(bo->tbo.resv, - true, false, - MAX_SCHEDULE_TIMEOUT); - if (unlikely(r < 0)) { - DRM_DEBUG_PRIME("Fence wait failed: %li\n", r); + r = __reservation_object_make_exclusive(bo->tbo.resv); + if (r) goto error_unreserve; - } } /* pin buffer into GTT */ @@ -332,15 +373,13 @@ static int amdgpu_gem_begin_cpu_access(struct dma_buf *dma_buf, return ret; } -static const struct dma_buf_ops amdgpu_dmabuf_ops = { +const struct dma_buf_ops amdgpu_dmabuf_ops = { .attach = amdgpu_gem_map_attach, .detach = amdgpu_gem_map_detach, .map_dma_buf = drm_gem_map_dma_buf, .unmap_dma_buf = drm_gem_unmap_dma_buf, .release = drm_gem_dmabuf_release, .begin_cpu_access = amdgpu_gem_begin_cpu_access, - .map = drm_gem_dmabuf_kmap, - .unmap = drm_gem_dmabuf_kunmap, .mmap = drm_gem_dmabuf_mmap, .vmap = drm_gem_dmabuf_vmap, .vunmap = drm_gem_dmabuf_vunmap, diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c index 25d2f3e757f1..3091488cd8cc 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c @@ -67,9 +67,6 @@ static int psp_sw_init(void *handle) psp->adev = adev; - if (adev->firmware.load_type != AMDGPU_FW_LOAD_PSP) - return 0; - ret = psp_init_microcode(psp); if (ret) { DRM_ERROR("Failed to load psp firmware!\n"); @@ -83,13 +80,14 @@ static int psp_sw_fini(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; - if (adev->firmware.load_type != AMDGPU_FW_LOAD_PSP) - return 0; - release_firmware(adev->psp.sos_fw); adev->psp.sos_fw = NULL; release_firmware(adev->psp.asd_fw); adev->psp.asd_fw = NULL; + if (adev->psp.ta_fw) { + release_firmware(adev->psp.ta_fw); + adev->psp.ta_fw = NULL; + } return 0; } @@ -118,29 +116,44 @@ int psp_wait_for(struct psp_context *psp, uint32_t reg_index, static int psp_cmd_submit_buf(struct psp_context *psp, struct amdgpu_firmware_info *ucode, - struct psp_gfx_cmd_resp *cmd, uint64_t fence_mc_addr, - int index) + struct psp_gfx_cmd_resp *cmd, uint64_t fence_mc_addr) { int ret; + int index; memset(psp->cmd_buf_mem, 0, PSP_CMD_BUFFER_SIZE); memcpy(psp->cmd_buf_mem, cmd, sizeof(struct psp_gfx_cmd_resp)); + index = atomic_inc_return(&psp->fence_value); ret = psp_cmd_submit(psp, ucode, psp->cmd_buf_mc_addr, fence_mc_addr, index); + if (ret) { + atomic_dec(&psp->fence_value); + return ret; + } - while (*((unsigned int *)psp->fence_buf) != index) { + while (*((unsigned int *)psp->fence_buf) != index) msleep(1); - } - /* the status field must be 0 after FW is loaded */ - if (ucode && psp->cmd_buf_mem->resp.status) { - DRM_ERROR("failed loading with status (%d) and ucode id (%d)\n", - psp->cmd_buf_mem->resp.status, ucode->ucode_id); - return -EINVAL; + /* In some cases, psp response status is not 0 even there is no + * problem while the command is submitted. Some version of PSP FW + * doesn't write 0 to that field. + * So here we would like to only print a warning instead of an error + * during psp initialization to avoid breaking hw_init and it doesn't + * return -EINVAL. + */ + if (psp->cmd_buf_mem->resp.status) { + if (ucode) + DRM_WARN("failed to load ucode id (%d) ", + ucode->ucode_id); + DRM_WARN("psp command failed and response status is (%d)\n", + psp->cmd_buf_mem->resp.status); } + /* get xGMI session id from response buffer */ + cmd->resp.session_id = psp->cmd_buf_mem->resp.session_id; + if (ucode) { ucode->tmr_mc_addr_lo = psp->cmd_buf_mem->resp.fw_addr_lo; ucode->tmr_mc_addr_hi = psp->cmd_buf_mem->resp.fw_addr_hi; @@ -149,10 +162,14 @@ psp_cmd_submit_buf(struct psp_context *psp, return ret; } -static void psp_prep_tmr_cmd_buf(struct psp_gfx_cmd_resp *cmd, +static void psp_prep_tmr_cmd_buf(struct psp_context *psp, + struct psp_gfx_cmd_resp *cmd, uint64_t tmr_mc, uint32_t size) { - cmd->cmd_id = GFX_CMD_ID_SETUP_TMR; + if (psp_support_vmr_ring(psp)) + cmd->cmd_id = GFX_CMD_ID_SETUP_VMR; + else + cmd->cmd_id = GFX_CMD_ID_SETUP_TMR; cmd->cmd.cmd_setup_tmr.buf_phy_addr_lo = lower_32_bits(tmr_mc); cmd->cmd.cmd_setup_tmr.buf_phy_addr_hi = upper_32_bits(tmr_mc); cmd->cmd.cmd_setup_tmr.buf_size = size; @@ -186,12 +203,12 @@ static int psp_tmr_load(struct psp_context *psp) if (!cmd) return -ENOMEM; - psp_prep_tmr_cmd_buf(cmd, psp->tmr_mc_addr, PSP_TMR_SIZE); + psp_prep_tmr_cmd_buf(psp, cmd, psp->tmr_mc_addr, PSP_TMR_SIZE); DRM_INFO("reserve 0x%x from 0x%llx for PSP TMR SIZE\n", PSP_TMR_SIZE, psp->tmr_mc_addr); ret = psp_cmd_submit_buf(psp, NULL, cmd, - psp->fence_buf_mc_addr, 1); + psp->fence_buf_mc_addr); if (ret) goto failed; @@ -258,13 +275,197 @@ static int psp_asd_load(struct psp_context *psp) psp->asd_ucode_size, PSP_ASD_SHARED_MEM_SIZE); ret = psp_cmd_submit_buf(psp, NULL, cmd, - psp->fence_buf_mc_addr, 2); + psp->fence_buf_mc_addr); kfree(cmd); return ret; } +static void psp_prep_xgmi_ta_load_cmd_buf(struct psp_gfx_cmd_resp *cmd, + uint64_t xgmi_ta_mc, uint64_t xgmi_mc_shared, + uint32_t xgmi_ta_size, uint32_t shared_size) +{ + cmd->cmd_id = GFX_CMD_ID_LOAD_TA; + cmd->cmd.cmd_load_ta.app_phy_addr_lo = lower_32_bits(xgmi_ta_mc); + cmd->cmd.cmd_load_ta.app_phy_addr_hi = upper_32_bits(xgmi_ta_mc); + cmd->cmd.cmd_load_ta.app_len = xgmi_ta_size; + + cmd->cmd.cmd_load_ta.cmd_buf_phy_addr_lo = lower_32_bits(xgmi_mc_shared); + cmd->cmd.cmd_load_ta.cmd_buf_phy_addr_hi = upper_32_bits(xgmi_mc_shared); + cmd->cmd.cmd_load_ta.cmd_buf_len = shared_size; +} + +static int psp_xgmi_init_shared_buf(struct psp_context *psp) +{ + int ret; + + /* + * Allocate 16k memory aligned to 4k from Frame Buffer (local + * physical) for xgmi ta <-> Driver + */ + ret = amdgpu_bo_create_kernel(psp->adev, PSP_XGMI_SHARED_MEM_SIZE, + PAGE_SIZE, AMDGPU_GEM_DOMAIN_VRAM, + &psp->xgmi_context.xgmi_shared_bo, + &psp->xgmi_context.xgmi_shared_mc_addr, + &psp->xgmi_context.xgmi_shared_buf); + + return ret; +} + +static int psp_xgmi_load(struct psp_context *psp) +{ + int ret; + struct psp_gfx_cmd_resp *cmd; + + /* + * TODO: bypass the loading in sriov for now + */ + if (amdgpu_sriov_vf(psp->adev)) + return 0; + + cmd = kzalloc(sizeof(struct psp_gfx_cmd_resp), GFP_KERNEL); + if (!cmd) + return -ENOMEM; + + memset(psp->fw_pri_buf, 0, PSP_1_MEG); + memcpy(psp->fw_pri_buf, psp->ta_xgmi_start_addr, psp->ta_xgmi_ucode_size); + + psp_prep_xgmi_ta_load_cmd_buf(cmd, psp->fw_pri_mc_addr, + psp->xgmi_context.xgmi_shared_mc_addr, + psp->ta_xgmi_ucode_size, PSP_XGMI_SHARED_MEM_SIZE); + + ret = psp_cmd_submit_buf(psp, NULL, cmd, + psp->fence_buf_mc_addr); + + if (!ret) { + psp->xgmi_context.initialized = 1; + psp->xgmi_context.session_id = cmd->resp.session_id; + } + + kfree(cmd); + + return ret; +} + +static void psp_prep_xgmi_ta_unload_cmd_buf(struct psp_gfx_cmd_resp *cmd, + uint32_t xgmi_session_id) +{ + cmd->cmd_id = GFX_CMD_ID_UNLOAD_TA; + cmd->cmd.cmd_unload_ta.session_id = xgmi_session_id; +} + +static int psp_xgmi_unload(struct psp_context *psp) +{ + int ret; + struct psp_gfx_cmd_resp *cmd; + + /* + * TODO: bypass the unloading in sriov for now + */ + if (amdgpu_sriov_vf(psp->adev)) + return 0; + + cmd = kzalloc(sizeof(struct psp_gfx_cmd_resp), GFP_KERNEL); + if (!cmd) + return -ENOMEM; + + psp_prep_xgmi_ta_unload_cmd_buf(cmd, psp->xgmi_context.session_id); + + ret = psp_cmd_submit_buf(psp, NULL, cmd, + psp->fence_buf_mc_addr); + + kfree(cmd); + + return ret; +} + +static void psp_prep_xgmi_ta_invoke_cmd_buf(struct psp_gfx_cmd_resp *cmd, + uint32_t ta_cmd_id, + uint32_t xgmi_session_id) +{ + cmd->cmd_id = GFX_CMD_ID_INVOKE_CMD; + cmd->cmd.cmd_invoke_cmd.session_id = xgmi_session_id; + cmd->cmd.cmd_invoke_cmd.ta_cmd_id = ta_cmd_id; + /* Note: cmd_invoke_cmd.buf is not used for now */ +} + +int psp_xgmi_invoke(struct psp_context *psp, uint32_t ta_cmd_id) +{ + int ret; + struct psp_gfx_cmd_resp *cmd; + + /* + * TODO: bypass the loading in sriov for now + */ + if (amdgpu_sriov_vf(psp->adev)) + return 0; + + cmd = kzalloc(sizeof(struct psp_gfx_cmd_resp), GFP_KERNEL); + if (!cmd) + return -ENOMEM; + + psp_prep_xgmi_ta_invoke_cmd_buf(cmd, ta_cmd_id, + psp->xgmi_context.session_id); + + ret = psp_cmd_submit_buf(psp, NULL, cmd, + psp->fence_buf_mc_addr); + + kfree(cmd); + + return ret; +} + +static int psp_xgmi_terminate(struct psp_context *psp) +{ + int ret; + + if (!psp->xgmi_context.initialized) + return 0; + + ret = psp_xgmi_unload(psp); + if (ret) + return ret; + + psp->xgmi_context.initialized = 0; + + /* free xgmi shared memory */ + amdgpu_bo_free_kernel(&psp->xgmi_context.xgmi_shared_bo, + &psp->xgmi_context.xgmi_shared_mc_addr, + &psp->xgmi_context.xgmi_shared_buf); + + return 0; +} + +static int psp_xgmi_initialize(struct psp_context *psp) +{ + struct ta_xgmi_shared_memory *xgmi_cmd; + int ret; + + if (!psp->adev->psp.ta_fw) + return -ENOENT; + + if (!psp->xgmi_context.initialized) { + ret = psp_xgmi_init_shared_buf(psp); + if (ret) + return ret; + } + + /* Load XGMI TA */ + ret = psp_xgmi_load(psp); + if (ret) + return ret; + + /* Initialize XGMI session */ + xgmi_cmd = (struct ta_xgmi_shared_memory *)(psp->xgmi_context.xgmi_shared_buf); + memset(xgmi_cmd, 0, sizeof(struct ta_xgmi_shared_memory)); + xgmi_cmd->cmd_id = TA_COMMAND_XGMI__INITIALIZE; + + ret = psp_xgmi_invoke(psp, xgmi_cmd->cmd_id); + + return ret; +} + static int psp_hw_start(struct psp_context *psp) { struct amdgpu_device *adev = psp->adev; @@ -292,9 +493,110 @@ static int psp_hw_start(struct psp_context *psp) if (ret) return ret; + if (adev->gmc.xgmi.num_physical_nodes > 1) { + ret = psp_xgmi_initialize(psp); + /* Warning the XGMI seesion initialize failure + * Instead of stop driver initialization + */ + if (ret) + dev_err(psp->adev->dev, + "XGMI: Failed to initialize XGMI session\n"); + } + return 0; +} + +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_SDMA0: + *type = GFX_FW_TYPE_SDMA0; + break; + case AMDGPU_UCODE_ID_SDMA1: + *type = GFX_FW_TYPE_SDMA1; + break; + case AMDGPU_UCODE_ID_CP_CE: + *type = GFX_FW_TYPE_CP_CE; + break; + case AMDGPU_UCODE_ID_CP_PFP: + *type = GFX_FW_TYPE_CP_PFP; + break; + case AMDGPU_UCODE_ID_CP_ME: + *type = GFX_FW_TYPE_CP_ME; + break; + case AMDGPU_UCODE_ID_CP_MEC1: + *type = GFX_FW_TYPE_CP_MEC; + break; + case AMDGPU_UCODE_ID_CP_MEC1_JT: + *type = GFX_FW_TYPE_CP_MEC_ME1; + break; + case AMDGPU_UCODE_ID_CP_MEC2: + *type = GFX_FW_TYPE_CP_MEC; + break; + case AMDGPU_UCODE_ID_CP_MEC2_JT: + *type = GFX_FW_TYPE_CP_MEC_ME2; + break; + case AMDGPU_UCODE_ID_RLC_G: + *type = GFX_FW_TYPE_RLC_G; + break; + case AMDGPU_UCODE_ID_RLC_RESTORE_LIST_CNTL: + *type = GFX_FW_TYPE_RLC_RESTORE_LIST_SRM_CNTL; + break; + case AMDGPU_UCODE_ID_RLC_RESTORE_LIST_GPM_MEM: + *type = GFX_FW_TYPE_RLC_RESTORE_LIST_GPM_MEM; + break; + case AMDGPU_UCODE_ID_RLC_RESTORE_LIST_SRM_MEM: + *type = GFX_FW_TYPE_RLC_RESTORE_LIST_SRM_MEM; + break; + case AMDGPU_UCODE_ID_SMC: + *type = GFX_FW_TYPE_SMU; + break; + case AMDGPU_UCODE_ID_UVD: + *type = GFX_FW_TYPE_UVD; + break; + case AMDGPU_UCODE_ID_UVD1: + *type = GFX_FW_TYPE_UVD1; + break; + case AMDGPU_UCODE_ID_VCE: + *type = GFX_FW_TYPE_VCE; + break; + case AMDGPU_UCODE_ID_VCN: + *type = GFX_FW_TYPE_VCN; + break; + case AMDGPU_UCODE_ID_DMCU_ERAM: + *type = GFX_FW_TYPE_DMCU_ERAM; + break; + case AMDGPU_UCODE_ID_DMCU_INTV: + *type = GFX_FW_TYPE_DMCU_ISR; + break; + case AMDGPU_UCODE_ID_MAXIMUM: + default: + return -EINVAL; + } + return 0; } +static int psp_prep_load_ip_fw_cmd_buf(struct amdgpu_firmware_info *ucode, + struct psp_gfx_cmd_resp *cmd) +{ + int ret; + uint64_t fw_mem_mc_addr = ucode->mc_addr; + + memset(cmd, 0, sizeof(struct psp_gfx_cmd_resp)); + + cmd->cmd_id = GFX_CMD_ID_LOAD_IP_FW; + cmd->cmd.cmd_load_ip_fw.fw_phy_addr_lo = lower_32_bits(fw_mem_mc_addr); + cmd->cmd.cmd_load_ip_fw.fw_phy_addr_hi = upper_32_bits(fw_mem_mc_addr); + cmd->cmd.cmd_load_ip_fw.fw_size = ucode->ucode_size; + + ret = psp_get_fw_type(ucode, &cmd->cmd.cmd_load_ip_fw.fw_type); + if (ret) + DRM_ERROR("Unknown firmware type\n"); + + return ret; +} + static int psp_np_fw_load(struct psp_context *psp) { int i, ret; @@ -316,12 +618,12 @@ static int psp_np_fw_load(struct psp_context *psp) /*skip ucode loading in SRIOV VF */ continue; - ret = psp_prep_cmd_buf(ucode, psp->cmd); + ret = psp_prep_load_ip_fw_cmd_buf(ucode, psp->cmd); if (ret) return ret; ret = psp_cmd_submit_buf(psp, ucode, psp->cmd, - psp->fence_buf_mc_addr, i + 3); + psp->fence_buf_mc_addr); if (ret) return ret; @@ -340,8 +642,10 @@ static int psp_load_fw(struct amdgpu_device *adev) int ret; struct psp_context *psp = &adev->psp; - if (amdgpu_sriov_vf(adev) && adev->in_gpu_reset != 0) + if (amdgpu_sriov_vf(adev) && adev->in_gpu_reset) { + psp_ring_stop(psp, PSP_RING_TYPE__KM); /* should not destroy ring, only stop */ goto skip_memalloc; + } psp->cmd = kzalloc(sizeof(struct psp_gfx_cmd_resp), GFP_KERNEL); if (!psp->cmd) @@ -416,10 +720,6 @@ static int psp_hw_init(void *handle) int ret; struct amdgpu_device *adev = (struct amdgpu_device *)handle; - - if (adev->firmware.load_type != AMDGPU_FW_LOAD_PSP) - return 0; - mutex_lock(&adev->firmware.mutex); /* * This sequence is just used on hw_init only once, no need on @@ -449,8 +749,9 @@ static int psp_hw_fini(void *handle) struct amdgpu_device *adev = (struct amdgpu_device *)handle; struct psp_context *psp = &adev->psp; - if (adev->firmware.load_type != AMDGPU_FW_LOAD_PSP) - return 0; + if (adev->gmc.xgmi.num_physical_nodes > 1 && + psp->xgmi_context.initialized == 1) + psp_xgmi_terminate(psp); psp_ring_destroy(psp, PSP_RING_TYPE__KM); @@ -476,8 +777,14 @@ static int psp_suspend(void *handle) struct amdgpu_device *adev = (struct amdgpu_device *)handle; struct psp_context *psp = &adev->psp; - if (adev->firmware.load_type != AMDGPU_FW_LOAD_PSP) - return 0; + if (adev->gmc.xgmi.num_physical_nodes > 1 && + psp->xgmi_context.initialized == 1) { + ret = psp_xgmi_terminate(psp); + if (ret) { + DRM_ERROR("Failed to terminate xgmi ta\n"); + return ret; + } + } ret = psp_ring_stop(psp, PSP_RING_TYPE__KM); if (ret) { @@ -494,9 +801,6 @@ static int psp_resume(void *handle) struct amdgpu_device *adev = (struct amdgpu_device *)handle; struct psp_context *psp = &adev->psp; - if (adev->firmware.load_type != AMDGPU_FW_LOAD_PSP) - return 0; - DRM_INFO("PSP is resuming...\n"); mutex_lock(&adev->firmware.mutex); @@ -532,11 +836,6 @@ static bool psp_check_fw_loading_status(struct amdgpu_device *adev, { struct amdgpu_firmware_info *ucode = NULL; - if (adev->firmware.load_type != AMDGPU_FW_LOAD_PSP) { - DRM_INFO("firmware is not loaded by PSP\n"); - return true; - } - if (!adev->firmware.fw_size) return false; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.h index 8b8720e9c3f0..2ef98cc755d6 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.h @@ -27,14 +27,17 @@ #include "amdgpu.h" #include "psp_gfx_if.h" +#include "ta_xgmi_if.h" #define PSP_FENCE_BUFFER_SIZE 0x1000 #define PSP_CMD_BUFFER_SIZE 0x1000 -#define PSP_ASD_SHARED_MEM_SIZE 0x4000 +#define PSP_ASD_SHARED_MEM_SIZE 0x4000 +#define PSP_XGMI_SHARED_MEM_SIZE 0x4000 #define PSP_1_MEG 0x100000 #define PSP_TMR_SIZE 0x400000 struct psp_context; +struct psp_xgmi_node_info; struct psp_xgmi_topology_info; enum psp_ring_type @@ -62,8 +65,6 @@ struct psp_funcs int (*init_microcode)(struct psp_context *psp); int (*bootloader_load_sysdrv)(struct psp_context *psp); int (*bootloader_load_sos)(struct psp_context *psp); - int (*prep_cmd_buf)(struct amdgpu_firmware_info *ucode, - struct psp_gfx_cmd_resp *cmd); int (*ring_init)(struct psp_context *psp, enum psp_ring_type ring_type); int (*ring_create)(struct psp_context *psp, enum psp_ring_type ring_type); @@ -80,12 +81,21 @@ struct psp_funcs enum AMDGPU_UCODE_ID ucode_type); bool (*smu_reload_quirk)(struct psp_context *psp); int (*mode1_reset)(struct psp_context *psp); - uint64_t (*xgmi_get_device_id)(struct psp_context *psp); - uint64_t (*xgmi_get_hive_id)(struct psp_context *psp); + int (*xgmi_get_node_id)(struct psp_context *psp, uint64_t *node_id); + int (*xgmi_get_hive_id)(struct psp_context *psp, uint64_t *hive_id); int (*xgmi_get_topology_info)(struct psp_context *psp, int number_devices, - struct psp_xgmi_topology_info *topology); + struct psp_xgmi_topology_info *topology); int (*xgmi_set_topology_info)(struct psp_context *psp, int number_devices, - struct psp_xgmi_topology_info *topology); + struct psp_xgmi_topology_info *topology); + bool (*support_vmr_ring)(struct psp_context *psp); +}; + +struct psp_xgmi_context { + uint8_t initialized; + uint32_t session_id; + struct amdgpu_bo *xgmi_shared_bo; + uint64_t xgmi_shared_mc_addr; + void *xgmi_shared_buf; }; struct psp_context @@ -96,7 +106,7 @@ struct psp_context const struct psp_funcs *funcs; - /* fence buffer */ + /* firmware buffer */ struct amdgpu_bo *fw_pri_bo; uint64_t fw_pri_mc_addr; void *fw_pri_buf; @@ -134,6 +144,16 @@ struct psp_context struct amdgpu_bo *cmd_buf_bo; uint64_t cmd_buf_mc_addr; struct psp_gfx_cmd_resp *cmd_buf_mem; + + /* fence value associated with cmd buffer */ + atomic_t fence_value; + + /* xgmi ta firmware and buffer */ + const struct firmware *ta_fw; + uint32_t ta_xgmi_ucode_version; + uint32_t ta_xgmi_ucode_size; + uint8_t *ta_xgmi_start_addr; + struct psp_xgmi_context xgmi_context; }; struct amdgpu_psp_funcs { @@ -141,24 +161,19 @@ struct amdgpu_psp_funcs { enum AMDGPU_UCODE_ID); }; +#define AMDGPU_XGMI_MAX_CONNECTED_NODES 64 +struct psp_xgmi_node_info { + uint64_t node_id; + uint8_t num_hops; + uint8_t is_sharing_enabled; + enum ta_xgmi_assigned_sdma_engine sdma_engine; +}; + struct psp_xgmi_topology_info { - /* Generated by PSP to identify the GPU instance within xgmi connection */ - uint64_t device_id; - /* - * If all bits set to 0 , driver indicates it wants to retrieve the xgmi - * connection vector topology, but not access enable the connections - * if some or all bits are set to 1, driver indicates it want to retrieve the - * current xgmi topology and access enable the link to GPU[i] associated - * with the bit position in the vector. - * On return,: bits indicated which xgmi links are present/active depending - * on the value passed in. The relative bit offset for the relative GPU index - * within the hive is always marked active. - */ - uint32_t connection_mask; - uint32_t reserved; /* must be 0 */ + uint32_t num_nodes; + struct psp_xgmi_node_info nodes[AMDGPU_XGMI_MAX_CONNECTED_NODES]; }; -#define psp_prep_cmd_buf(ucode, type) (psp)->funcs->prep_cmd_buf((ucode), (type)) #define psp_ring_init(psp, type) (psp)->funcs->ring_init((psp), (type)) #define psp_ring_create(psp, type) (psp)->funcs->ring_create((psp), (type)) #define psp_ring_stop(psp, type) (psp)->funcs->ring_stop((psp), (type)) @@ -175,12 +190,14 @@ struct psp_xgmi_topology_info { ((psp)->funcs->bootloader_load_sos ? (psp)->funcs->bootloader_load_sos((psp)) : 0) #define psp_smu_reload_quirk(psp) \ ((psp)->funcs->smu_reload_quirk ? (psp)->funcs->smu_reload_quirk((psp)) : false) +#define psp_support_vmr_ring(psp) \ + ((psp)->funcs->support_vmr_ring ? (psp)->funcs->support_vmr_ring((psp)) : false) #define psp_mode1_reset(psp) \ ((psp)->funcs->mode1_reset ? (psp)->funcs->mode1_reset((psp)) : false) -#define psp_xgmi_get_device_id(psp) \ - ((psp)->funcs->xgmi_get_device_id ? (psp)->funcs->xgmi_get_device_id((psp)) : 0) -#define psp_xgmi_get_hive_id(psp) \ - ((psp)->funcs->xgmi_get_hive_id ? (psp)->funcs->xgmi_get_hive_id((psp)) : 0) +#define psp_xgmi_get_node_id(psp, node_id) \ + ((psp)->funcs->xgmi_get_node_id ? (psp)->funcs->xgmi_get_node_id((psp), (node_id)) : -EINVAL) +#define psp_xgmi_get_hive_id(psp, hive_id) \ + ((psp)->funcs->xgmi_get_hive_id ? (psp)->funcs->xgmi_get_hive_id((psp), (hive_id)) : -EINVAL) #define psp_xgmi_get_topology_info(psp, num_device, topology) \ ((psp)->funcs->xgmi_get_topology_info ? \ (psp)->funcs->xgmi_get_topology_info((psp), (num_device), (topology)) : -EINVAL) @@ -199,6 +216,7 @@ extern int psp_wait_for(struct psp_context *psp, uint32_t reg_index, extern const struct amdgpu_ip_block_version psp_v10_0_ip_block; int psp_gpu_reset(struct amdgpu_device *adev); +int psp_xgmi_invoke(struct psp_context *psp, uint32_t ta_cmd_id); extern const struct amdgpu_ip_block_version psp_v11_0_ip_block; #endif diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c index b70e85ec147d..335a0edf114b 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c @@ -338,7 +338,7 @@ int amdgpu_ring_init(struct amdgpu_device *adev, struct amdgpu_ring *ring, */ void amdgpu_ring_fini(struct amdgpu_ring *ring) { - ring->ready = false; + ring->sched.ready = false; /* Not to finish a ring which is not initialized */ if (!(ring->adev) || !(ring->adev->rings[ring->idx])) @@ -397,7 +397,7 @@ bool amdgpu_ring_soft_recovery(struct amdgpu_ring *ring, unsigned int vmid, { ktime_t deadline = ktime_add_us(ktime_get(), 10000); - if (!ring->funcs->soft_recovery) + if (!ring->funcs->soft_recovery || !fence) return false; atomic_inc(&ring->adev->gpu_reset_counter); @@ -500,3 +500,29 @@ static void amdgpu_debugfs_ring_fini(struct amdgpu_ring *ring) debugfs_remove(ring->ent); #endif } + +/** + * amdgpu_ring_test_helper - tests ring and set sched readiness status + * + * @ring: ring to try the recovery on + * + * Tests ring and set sched readiness status + * + * Returns 0 on success, error on failure. + */ +int amdgpu_ring_test_helper(struct amdgpu_ring *ring) +{ + struct amdgpu_device *adev = ring->adev; + int r; + + r = amdgpu_ring_test_ring(ring); + if (r) + DRM_DEV_ERROR(adev->dev, "ring %s test failed (%d)\n", + ring->name, r); + else + DRM_DEV_DEBUG(adev->dev, "ring test on %s succeeded\n", + ring->name); + + ring->sched.ready = !r; + return r; +} diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h index 4caa301ce454..d7fae2676269 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h @@ -29,7 +29,7 @@ #include <drm/drm_print.h> /* max number of rings */ -#define AMDGPU_MAX_RINGS 21 +#define AMDGPU_MAX_RINGS 23 #define AMDGPU_MAX_GFX_RINGS 1 #define AMDGPU_MAX_COMPUTE_RINGS 8 #define AMDGPU_MAX_VCE_RINGS 3 @@ -129,8 +129,9 @@ struct amdgpu_ring_funcs { unsigned emit_ib_size; /* command emit functions */ void (*emit_ib)(struct amdgpu_ring *ring, + struct amdgpu_job *job, struct amdgpu_ib *ib, - unsigned vmid, bool ctx_switch); + uint32_t flags); void (*emit_fence)(struct amdgpu_ring *ring, uint64_t addr, uint64_t seq, unsigned flags); void (*emit_pipeline_sync)(struct amdgpu_ring *ring); @@ -189,7 +190,6 @@ struct amdgpu_ring { uint64_t gpu_addr; uint64_t ptr_mask; uint32_t buf_mask; - bool ready; u32 idx; u32 me; u32 pipe; @@ -229,7 +229,7 @@ struct amdgpu_ring { #define amdgpu_ring_get_rptr(r) (r)->funcs->get_rptr((r)) #define amdgpu_ring_get_wptr(r) (r)->funcs->get_wptr((r)) #define amdgpu_ring_set_wptr(r) (r)->funcs->set_wptr((r)) -#define amdgpu_ring_emit_ib(r, ib, vmid, c) (r)->funcs->emit_ib((r), (ib), (vmid), (c)) +#define amdgpu_ring_emit_ib(r, job, ib, flags) ((r)->funcs->emit_ib((r), (job), (ib), (flags))) #define amdgpu_ring_emit_pipeline_sync(r) (r)->funcs->emit_pipeline_sync((r)) #define amdgpu_ring_emit_vm_flush(r, vmid, addr) (r)->funcs->emit_vm_flush((r), (vmid), (addr)) #define amdgpu_ring_emit_fence(r, addr, seq, flags) (r)->funcs->emit_fence((r), (addr), (seq), (flags)) @@ -313,4 +313,6 @@ static inline void amdgpu_ring_write_multiple(struct amdgpu_ring *ring, ring->count_dw -= count_dw; } +int amdgpu_ring_test_helper(struct amdgpu_ring *ring); + #endif diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_rlc.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_rlc.c new file mode 100644 index 000000000000..c8793e6cc3c5 --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_rlc.c @@ -0,0 +1,282 @@ +/* + * Copyright 2014 Advanced Micro Devices, Inc. + * Copyright 2008 Red Hat Inc. + * Copyright 2009 Jerome Glisse. + * + * 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_gfx.h" +#include "amdgpu_rlc.h" + +/** + * amdgpu_gfx_rlc_enter_safe_mode - Set RLC into safe mode + * + * @adev: amdgpu_device pointer + * + * Set RLC enter into safe mode if RLC is enabled and haven't in safe mode. + */ +void amdgpu_gfx_rlc_enter_safe_mode(struct amdgpu_device *adev) +{ + if (adev->gfx.rlc.in_safe_mode) + return; + + /* if RLC is not enabled, do nothing */ + if (!adev->gfx.rlc.funcs->is_rlc_enabled(adev)) + return; + + if (adev->cg_flags & + (AMD_CG_SUPPORT_GFX_CGCG | AMD_CG_SUPPORT_GFX_MGCG | + AMD_CG_SUPPORT_GFX_3D_CGCG)) { + adev->gfx.rlc.funcs->set_safe_mode(adev); + adev->gfx.rlc.in_safe_mode = true; + } +} + +/** + * amdgpu_gfx_rlc_exit_safe_mode - Set RLC out of safe mode + * + * @adev: amdgpu_device pointer + * + * Set RLC exit safe mode if RLC is enabled and have entered into safe mode. + */ +void amdgpu_gfx_rlc_exit_safe_mode(struct amdgpu_device *adev) +{ + if (!(adev->gfx.rlc.in_safe_mode)) + return; + + /* if RLC is not enabled, do nothing */ + if (!adev->gfx.rlc.funcs->is_rlc_enabled(adev)) + return; + + if (adev->cg_flags & + (AMD_CG_SUPPORT_GFX_CGCG | AMD_CG_SUPPORT_GFX_MGCG | + AMD_CG_SUPPORT_GFX_3D_CGCG)) { + adev->gfx.rlc.funcs->unset_safe_mode(adev); + adev->gfx.rlc.in_safe_mode = false; + } +} + +/** + * amdgpu_gfx_rlc_init_sr - Init save restore block + * + * @adev: amdgpu_device pointer + * @dws: the size of save restore block + * + * Allocate and setup value to save restore block of rlc. + * Returns 0 on succeess or negative error code if allocate failed. + */ +int amdgpu_gfx_rlc_init_sr(struct amdgpu_device *adev, u32 dws) +{ + const u32 *src_ptr; + volatile u32 *dst_ptr; + u32 i; + int r; + + /* allocate save restore block */ + r = amdgpu_bo_create_reserved(adev, dws * 4, PAGE_SIZE, + AMDGPU_GEM_DOMAIN_VRAM, + &adev->gfx.rlc.save_restore_obj, + &adev->gfx.rlc.save_restore_gpu_addr, + (void **)&adev->gfx.rlc.sr_ptr); + if (r) { + dev_warn(adev->dev, "(%d) create RLC sr bo failed\n", r); + amdgpu_gfx_rlc_fini(adev); + return r; + } + + /* write the sr buffer */ + src_ptr = adev->gfx.rlc.reg_list; + dst_ptr = adev->gfx.rlc.sr_ptr; + for (i = 0; i < adev->gfx.rlc.reg_list_size; i++) + dst_ptr[i] = cpu_to_le32(src_ptr[i]); + amdgpu_bo_kunmap(adev->gfx.rlc.save_restore_obj); + amdgpu_bo_unreserve(adev->gfx.rlc.save_restore_obj); + + return 0; +} + +/** + * amdgpu_gfx_rlc_init_csb - Init clear state block + * + * @adev: amdgpu_device pointer + * + * Allocate and setup value to clear state block of rlc. + * Returns 0 on succeess or negative error code if allocate failed. + */ +int amdgpu_gfx_rlc_init_csb(struct amdgpu_device *adev) +{ + volatile u32 *dst_ptr; + u32 dws; + int r; + + /* allocate clear state block */ + adev->gfx.rlc.clear_state_size = dws = adev->gfx.rlc.funcs->get_csb_size(adev); + r = amdgpu_bo_create_reserved(adev, dws * 4, PAGE_SIZE, + AMDGPU_GEM_DOMAIN_VRAM, + &adev->gfx.rlc.clear_state_obj, + &adev->gfx.rlc.clear_state_gpu_addr, + (void **)&adev->gfx.rlc.cs_ptr); + if (r) { + dev_err(adev->dev, "(%d) failed to create rlc csb bo\n", r); + amdgpu_gfx_rlc_fini(adev); + return r; + } + + /* set up the cs buffer */ + dst_ptr = adev->gfx.rlc.cs_ptr; + adev->gfx.rlc.funcs->get_csb_buffer(adev, dst_ptr); + amdgpu_bo_kunmap(adev->gfx.rlc.clear_state_obj); + amdgpu_bo_unpin(adev->gfx.rlc.clear_state_obj); + amdgpu_bo_unreserve(adev->gfx.rlc.clear_state_obj); + + return 0; +} + +/** + * amdgpu_gfx_rlc_init_cpt - Init cp table + * + * @adev: amdgpu_device pointer + * + * Allocate and setup value to cp table of rlc. + * Returns 0 on succeess or negative error code if allocate failed. + */ +int amdgpu_gfx_rlc_init_cpt(struct amdgpu_device *adev) +{ + int r; + + r = amdgpu_bo_create_reserved(adev, adev->gfx.rlc.cp_table_size, + PAGE_SIZE, AMDGPU_GEM_DOMAIN_VRAM, + &adev->gfx.rlc.cp_table_obj, + &adev->gfx.rlc.cp_table_gpu_addr, + (void **)&adev->gfx.rlc.cp_table_ptr); + if (r) { + dev_err(adev->dev, "(%d) failed to create cp table bo\n", r); + amdgpu_gfx_rlc_fini(adev); + return r; + } + + /* set up the cp table */ + amdgpu_gfx_rlc_setup_cp_table(adev); + amdgpu_bo_kunmap(adev->gfx.rlc.cp_table_obj); + amdgpu_bo_unreserve(adev->gfx.rlc.cp_table_obj); + + return 0; +} + +/** + * amdgpu_gfx_rlc_setup_cp_table - setup cp the buffer of cp table + * + * @adev: amdgpu_device pointer + * + * Write cp firmware data into cp table. + */ +void amdgpu_gfx_rlc_setup_cp_table(struct amdgpu_device *adev) +{ + const __le32 *fw_data; + volatile u32 *dst_ptr; + int me, i, max_me; + u32 bo_offset = 0; + u32 table_offset, table_size; + + max_me = adev->gfx.rlc.funcs->get_cp_table_num(adev); + + /* write the cp table buffer */ + dst_ptr = adev->gfx.rlc.cp_table_ptr; + for (me = 0; me < max_me; me++) { + if (me == 0) { + const struct gfx_firmware_header_v1_0 *hdr = + (const struct gfx_firmware_header_v1_0 *)adev->gfx.ce_fw->data; + fw_data = (const __le32 *) + (adev->gfx.ce_fw->data + + le32_to_cpu(hdr->header.ucode_array_offset_bytes)); + table_offset = le32_to_cpu(hdr->jt_offset); + table_size = le32_to_cpu(hdr->jt_size); + } else if (me == 1) { + const struct gfx_firmware_header_v1_0 *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(hdr->header.ucode_array_offset_bytes)); + table_offset = le32_to_cpu(hdr->jt_offset); + table_size = le32_to_cpu(hdr->jt_size); + } else if (me == 2) { + const struct gfx_firmware_header_v1_0 *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(hdr->header.ucode_array_offset_bytes)); + table_offset = le32_to_cpu(hdr->jt_offset); + table_size = le32_to_cpu(hdr->jt_size); + } else if (me == 3) { + const struct gfx_firmware_header_v1_0 *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(hdr->header.ucode_array_offset_bytes)); + table_offset = le32_to_cpu(hdr->jt_offset); + table_size = le32_to_cpu(hdr->jt_size); + } else if (me == 4) { + const struct gfx_firmware_header_v1_0 *hdr = + (const struct gfx_firmware_header_v1_0 *)adev->gfx.mec2_fw->data; + fw_data = (const __le32 *) + (adev->gfx.mec2_fw->data + + le32_to_cpu(hdr->header.ucode_array_offset_bytes)); + table_offset = le32_to_cpu(hdr->jt_offset); + table_size = le32_to_cpu(hdr->jt_size); + } + + for (i = 0; i < table_size; i ++) { + dst_ptr[bo_offset + i] = + cpu_to_le32(le32_to_cpu(fw_data[table_offset + i])); + } + + bo_offset += table_size; + } +} + +/** + * amdgpu_gfx_rlc_fini - Free BO which used for RLC + * + * @adev: amdgpu_device pointer + * + * Free three BO which is used for rlc_save_restore_block, rlc_clear_state_block + * and rlc_jump_table_block. + */ +void amdgpu_gfx_rlc_fini(struct amdgpu_device *adev) +{ + /* save restore block */ + if (adev->gfx.rlc.save_restore_obj) { + amdgpu_bo_free_kernel(&adev->gfx.rlc.save_restore_obj, + &adev->gfx.rlc.save_restore_gpu_addr, + (void **)&adev->gfx.rlc.sr_ptr); + } + + /* 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); +} diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_rlc.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_rlc.h new file mode 100644 index 000000000000..49a8ab52113b --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_rlc.h @@ -0,0 +1,98 @@ +/* + * Copyright 2014 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_RLC_H__ +#define __AMDGPU_RLC_H__ + +#include "clearstate_defs.h" + +struct amdgpu_rlc_funcs { + bool (*is_rlc_enabled)(struct amdgpu_device *adev); + void (*set_safe_mode)(struct amdgpu_device *adev); + void (*unset_safe_mode)(struct amdgpu_device *adev); + int (*init)(struct amdgpu_device *adev); + u32 (*get_csb_size)(struct amdgpu_device *adev); + void (*get_csb_buffer)(struct amdgpu_device *adev, volatile u32 *buffer); + int (*get_cp_table_num)(struct amdgpu_device *adev); + int (*resume)(struct amdgpu_device *adev); + void (*stop)(struct amdgpu_device *adev); + void (*reset)(struct amdgpu_device *adev); + void (*start)(struct amdgpu_device *adev); +}; + +struct amdgpu_rlc { + /* for power gating */ + struct amdgpu_bo *save_restore_obj; + uint64_t save_restore_gpu_addr; + volatile uint32_t *sr_ptr; + const u32 *reg_list; + u32 reg_list_size; + /* for clear state */ + struct amdgpu_bo *clear_state_obj; + uint64_t clear_state_gpu_addr; + volatile uint32_t *cs_ptr; + const struct cs_section_def *cs_data; + u32 clear_state_size; + /* for cp tables */ + struct amdgpu_bo *cp_table_obj; + uint64_t cp_table_gpu_addr; + volatile uint32_t *cp_table_ptr; + u32 cp_table_size; + + /* safe mode for updating CG/PG state */ + bool in_safe_mode; + const struct amdgpu_rlc_funcs *funcs; + + /* for firmware data */ + u32 save_and_restore_offset; + u32 clear_state_descriptor_offset; + u32 avail_scratch_ram_locations; + u32 reg_restore_list_size; + u32 reg_list_format_start; + u32 reg_list_format_separate_start; + u32 starting_offsets_start; + u32 reg_list_format_size_bytes; + u32 reg_list_size_bytes; + u32 reg_list_format_direct_reg_list_length; + u32 save_restore_list_cntl_size_bytes; + u32 save_restore_list_gpm_size_bytes; + u32 save_restore_list_srm_size_bytes; + + u32 *register_list_format; + u32 *register_restore; + u8 *save_restore_list_cntl; + u8 *save_restore_list_gpm; + u8 *save_restore_list_srm; + + bool is_rlc_v2_1; +}; + +void amdgpu_gfx_rlc_enter_safe_mode(struct amdgpu_device *adev); +void amdgpu_gfx_rlc_exit_safe_mode(struct amdgpu_device *adev); +int amdgpu_gfx_rlc_init_sr(struct amdgpu_device *adev, u32 dws); +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); + +#endif diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_sa.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_sa.c index 12f2bf97611f..bfaf5c6323be 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_sa.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_sa.c @@ -388,7 +388,7 @@ void amdgpu_sa_bo_dump_debug_info(struct amdgpu_sa_manager *sa_manager, soffset, eoffset, eoffset - soffset); if (i->fence) - seq_printf(m, " protected by 0x%08x on context %llu", + seq_printf(m, " protected by 0x%016llx on context %llu", i->fence->seqno, i->fence->context); seq_printf(m, "\n"); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_sched.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_sched.c index 1cafe8d83a4d..0767a93e4d91 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_sched.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_sched.c @@ -54,16 +54,20 @@ static int amdgpu_sched_process_priority_override(struct amdgpu_device *adev, enum drm_sched_priority priority) { struct file *filp = fget(fd); - struct drm_file *file; struct amdgpu_fpriv *fpriv; struct amdgpu_ctx *ctx; uint32_t id; + int r; if (!filp) return -EINVAL; - file = filp->private_data; - fpriv = file->driver_priv; + r = amdgpu_file_to_fpriv(filp, &fpriv); + if (r) { + fput(filp); + return r; + } + idr_for_each_entry(&fpriv->ctx_mgr.ctx_handles, ctx, id) amdgpu_ctx_priority_override(ctx, priority); @@ -72,6 +76,39 @@ static int amdgpu_sched_process_priority_override(struct amdgpu_device *adev, return 0; } +static int amdgpu_sched_context_priority_override(struct amdgpu_device *adev, + int fd, + unsigned ctx_id, + enum drm_sched_priority priority) +{ + struct file *filp = fget(fd); + struct amdgpu_fpriv *fpriv; + struct amdgpu_ctx *ctx; + int r; + + if (!filp) + return -EINVAL; + + r = amdgpu_file_to_fpriv(filp, &fpriv); + if (r) { + fput(filp); + return r; + } + + ctx = amdgpu_ctx_get(fpriv, ctx_id); + + if (!ctx) { + fput(filp); + return -EINVAL; + } + + amdgpu_ctx_priority_override(ctx, priority); + amdgpu_ctx_put(ctx); + fput(filp); + + return 0; +} + int amdgpu_sched_ioctl(struct drm_device *dev, void *data, struct drm_file *filp) { @@ -81,7 +118,7 @@ int amdgpu_sched_ioctl(struct drm_device *dev, void *data, int r; priority = amdgpu_to_sched_priority(args->in.priority); - if (args->in.flags || priority == DRM_SCHED_PRIORITY_INVALID) + if (priority == DRM_SCHED_PRIORITY_INVALID) return -EINVAL; switch (args->in.op) { @@ -90,6 +127,12 @@ int amdgpu_sched_ioctl(struct drm_device *dev, void *data, args->in.fd, priority); break; + case AMDGPU_SCHED_OP_CONTEXT_PRIORITY_OVERRIDE: + r = amdgpu_sched_context_priority_override(adev, + args->in.fd, + args->in.ctx_id, + priority); + break; default: DRM_ERROR("Invalid sched op specified: %d\n", args->in.op); r = -EINVAL; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_sdma.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_sdma.c index bc9244b429ef..115bb0c99b0f 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_sdma.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_sdma.c @@ -28,17 +28,31 @@ * GPU SDMA IP block helpers function. */ -struct amdgpu_sdma_instance * amdgpu_get_sdma_instance(struct amdgpu_ring *ring) +struct amdgpu_sdma_instance *amdgpu_sdma_get_instance_from_ring(struct amdgpu_ring *ring) { struct amdgpu_device *adev = ring->adev; int i; for (i = 0; i < adev->sdma.num_instances; i++) - if (&adev->sdma.instance[i].ring == ring) - break; + if (ring == &adev->sdma.instance[i].ring || + ring == &adev->sdma.instance[i].page) + return &adev->sdma.instance[i]; - if (i < AMDGPU_MAX_SDMA_INSTANCES) - return &adev->sdma.instance[i]; - else - return NULL; + return NULL; +} + +int amdgpu_sdma_get_index_from_ring(struct amdgpu_ring *ring, uint32_t *index) +{ + struct amdgpu_device *adev = ring->adev; + int i; + + for (i = 0; i < adev->sdma.num_instances; i++) { + if (ring == &adev->sdma.instance[i].ring || + ring == &adev->sdma.instance[i].page) { + *index = i; + return 0; + } + } + + return -EINVAL; } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_sdma.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_sdma.h index 500113ec65ca..16b1a6ae5ba6 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_sdma.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_sdma.h @@ -41,6 +41,7 @@ struct amdgpu_sdma_instance { uint32_t feature_version; struct amdgpu_ring ring; + struct amdgpu_ring page; bool burst_nop; }; @@ -50,6 +51,7 @@ struct amdgpu_sdma { struct amdgpu_irq_src illegal_inst_irq; int num_instances; uint32_t srbm_soft_reset; + bool has_page_queue; }; /* @@ -92,6 +94,7 @@ struct amdgpu_buffer_funcs { #define amdgpu_emit_fill_buffer(adev, ib, s, d, b) (adev)->mman.buffer_funcs->emit_fill_buffer((ib), (s), (d), (b)) struct amdgpu_sdma_instance * -amdgpu_get_sdma_instance(struct amdgpu_ring *ring); +amdgpu_sdma_get_instance_from_ring(struct amdgpu_ring *ring); +int amdgpu_sdma_get_index_from_ring(struct amdgpu_ring *ring, uint32_t *index); #endif diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_trace.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_trace.h index e9bf70e2ac51..d3ca2424b5fe 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_trace.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_trace.h @@ -76,9 +76,10 @@ TRACE_EVENT(amdgpu_mm_wreg, ); TRACE_EVENT(amdgpu_iv, - TP_PROTO(struct amdgpu_iv_entry *iv), - TP_ARGS(iv), + TP_PROTO(unsigned ih, struct amdgpu_iv_entry *iv), + TP_ARGS(ih, iv), TP_STRUCT__entry( + __field(unsigned, ih) __field(unsigned, client_id) __field(unsigned, src_id) __field(unsigned, ring_id) @@ -90,6 +91,7 @@ TRACE_EVENT(amdgpu_iv, __array(unsigned, src_data, 4) ), TP_fast_assign( + __entry->ih = ih; __entry->client_id = iv->client_id; __entry->src_id = iv->src_id; __entry->ring_id = iv->ring_id; @@ -103,8 +105,9 @@ TRACE_EVENT(amdgpu_iv, __entry->src_data[2] = iv->src_data[2]; __entry->src_data[3] = iv->src_data[3]; ), - TP_printk("client_id:%u src_id:%u ring:%u vmid:%u timestamp: %llu pasid:%u src_data: %08x %08x %08x %08x", - __entry->client_id, __entry->src_id, + TP_printk("ih:%u client_id:%u src_id:%u ring:%u vmid:%u " + "timestamp: %llu pasid:%u src_data: %08x %08x %08x %08x", + __entry->ih, __entry->client_id, __entry->src_id, __entry->ring_id, __entry->vmid, __entry->timestamp, __entry->pasid, __entry->src_data[0], __entry->src_data[1], @@ -218,6 +221,7 @@ TRACE_EVENT(amdgpu_vm_grab_id, TP_ARGS(vm, ring, job), TP_STRUCT__entry( __field(u32, pasid) + __string(ring, ring->name) __field(u32, ring) __field(u32, vmid) __field(u32, vm_hub) @@ -227,14 +231,14 @@ TRACE_EVENT(amdgpu_vm_grab_id, TP_fast_assign( __entry->pasid = vm->pasid; - __entry->ring = ring->idx; + __assign_str(ring, ring->name) __entry->vmid = job->vmid; __entry->vm_hub = ring->funcs->vmhub, __entry->pd_addr = job->vm_pd_addr; __entry->needs_flush = job->vm_needs_flush; ), - TP_printk("pasid=%d, ring=%u, id=%u, hub=%u, pd_addr=%010Lx needs_flush=%u", - __entry->pasid, __entry->ring, __entry->vmid, + TP_printk("pasid=%d, ring=%s, id=%u, hub=%u, pd_addr=%010Lx needs_flush=%u", + __entry->pasid, __get_str(ring), __entry->vmid, __entry->vm_hub, __entry->pd_addr, __entry->needs_flush) ); @@ -366,20 +370,20 @@ TRACE_EVENT(amdgpu_vm_flush, uint64_t pd_addr), TP_ARGS(ring, vmid, pd_addr), TP_STRUCT__entry( - __field(u32, ring) + __string(ring, ring->name) __field(u32, vmid) __field(u32, vm_hub) __field(u64, pd_addr) ), TP_fast_assign( - __entry->ring = ring->idx; + __assign_str(ring, ring->name) __entry->vmid = vmid; __entry->vm_hub = ring->funcs->vmhub; __entry->pd_addr = pd_addr; ), - TP_printk("ring=%u, id=%u, hub=%u, pd_addr=%010Lx", - __entry->ring, __entry->vmid, + TP_printk("ring=%s, id=%u, hub=%u, pd_addr=%010Lx", + __get_str(ring), __entry->vmid, __entry->vm_hub,__entry->pd_addr) ); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c index a44fc12ae1f9..73e71e61dc99 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c @@ -61,100 +61,6 @@ static int amdgpu_map_buffer(struct ttm_buffer_object *bo, static int amdgpu_ttm_debugfs_init(struct amdgpu_device *adev); static void amdgpu_ttm_debugfs_fini(struct amdgpu_device *adev); -/* - * Global memory. - */ - -/** - * amdgpu_ttm_mem_global_init - Initialize and acquire reference to - * memory object - * - * @ref: Object for initialization. - * - * This is called by drm_global_item_ref() when an object is being - * initialized. - */ -static int amdgpu_ttm_mem_global_init(struct drm_global_reference *ref) -{ - return ttm_mem_global_init(ref->object); -} - -/** - * amdgpu_ttm_mem_global_release - Drop reference to a memory object - * - * @ref: Object being removed - * - * This is called by drm_global_item_unref() when an object is being - * released. - */ -static void amdgpu_ttm_mem_global_release(struct drm_global_reference *ref) -{ - ttm_mem_global_release(ref->object); -} - -/** - * amdgpu_ttm_global_init - Initialize global TTM memory reference structures. - * - * @adev: AMDGPU device for which the global structures need to be registered. - * - * This is called as part of the AMDGPU ttm init from amdgpu_ttm_init() - * during bring up. - */ -static int amdgpu_ttm_global_init(struct amdgpu_device *adev) -{ - struct drm_global_reference *global_ref; - int r; - - /* ensure reference is false in case init fails */ - adev->mman.mem_global_referenced = false; - - global_ref = &adev->mman.mem_global_ref; - global_ref->global_type = DRM_GLOBAL_TTM_MEM; - global_ref->size = sizeof(struct ttm_mem_global); - global_ref->init = &amdgpu_ttm_mem_global_init; - global_ref->release = &amdgpu_ttm_mem_global_release; - r = drm_global_item_ref(global_ref); - if (r) { - DRM_ERROR("Failed setting up TTM memory accounting " - "subsystem.\n"); - goto error_mem; - } - - adev->mman.bo_global_ref.mem_glob = - adev->mman.mem_global_ref.object; - global_ref = &adev->mman.bo_global_ref.ref; - global_ref->global_type = DRM_GLOBAL_TTM_BO; - global_ref->size = sizeof(struct ttm_bo_global); - global_ref->init = &ttm_bo_global_init; - global_ref->release = &ttm_bo_global_release; - r = drm_global_item_ref(global_ref); - if (r) { - DRM_ERROR("Failed setting up TTM BO subsystem.\n"); - goto error_bo; - } - - mutex_init(&adev->mman.gtt_window_lock); - - adev->mman.mem_global_referenced = true; - - return 0; - -error_bo: - drm_global_item_unref(&adev->mman.mem_global_ref); -error_mem: - return r; -} - -static void amdgpu_ttm_global_fini(struct amdgpu_device *adev) -{ - if (adev->mman.mem_global_referenced) { - mutex_destroy(&adev->mman.gtt_window_lock); - drm_global_item_unref(&adev->mman.bo_global_ref.ref); - drm_global_item_unref(&adev->mman.mem_global_ref); - adev->mman.mem_global_referenced = false; - } -} - static int amdgpu_invalidate_caches(struct ttm_bo_device *bdev, uint32_t flags) { return 0; @@ -1640,7 +1546,8 @@ static struct ttm_bo_driver amdgpu_bo_driver = { .io_mem_reserve = &amdgpu_ttm_io_mem_reserve, .io_mem_free = &amdgpu_ttm_io_mem_free, .io_mem_pfn = amdgpu_ttm_io_mem_pfn, - .access_memory = &amdgpu_ttm_access_memory + .access_memory = &amdgpu_ttm_access_memory, + .del_from_lru_notify = &amdgpu_vm_del_from_lru_notify }; /* @@ -1758,14 +1665,10 @@ int amdgpu_ttm_init(struct amdgpu_device *adev) int r; u64 vis_vram_limit; - /* initialize global references for vram/gtt */ - r = amdgpu_ttm_global_init(adev); - if (r) { - return r; - } + mutex_init(&adev->mman.gtt_window_lock); + /* No others user of address space so set it to 0 */ r = ttm_bo_device_init(&adev->mman.bdev, - adev->mman.bo_global_ref.ref.object, &amdgpu_bo_driver, adev->ddev->anon_inode->i_mapping, DRM_FILE_PAGE_OFFSET, @@ -1853,7 +1756,7 @@ int amdgpu_ttm_init(struct amdgpu_device *adev) } r = amdgpu_bo_create_kernel(adev, adev->gds.mem.gfx_partition_size, - PAGE_SIZE, AMDGPU_GEM_DOMAIN_GDS, + 4, AMDGPU_GEM_DOMAIN_GDS, &adev->gds.gds_gfx_bo, NULL, NULL); if (r) return r; @@ -1866,7 +1769,7 @@ int amdgpu_ttm_init(struct amdgpu_device *adev) } r = amdgpu_bo_create_kernel(adev, adev->gds.gws.gfx_partition_size, - PAGE_SIZE, AMDGPU_GEM_DOMAIN_GWS, + 1, AMDGPU_GEM_DOMAIN_GWS, &adev->gds.gws_gfx_bo, NULL, NULL); if (r) return r; @@ -1879,7 +1782,7 @@ int amdgpu_ttm_init(struct amdgpu_device *adev) } r = amdgpu_bo_create_kernel(adev, adev->gds.oa.gfx_partition_size, - PAGE_SIZE, AMDGPU_GEM_DOMAIN_OA, + 1, AMDGPU_GEM_DOMAIN_OA, &adev->gds.oa_gfx_bo, NULL, NULL); if (r) return r; @@ -1922,7 +1825,6 @@ void amdgpu_ttm_fini(struct amdgpu_device *adev) ttm_bo_clean_mm(&adev->mman.bdev, AMDGPU_PL_GWS); ttm_bo_clean_mm(&adev->mman.bdev, AMDGPU_PL_OA); ttm_bo_device_release(&adev->mman.bdev); - amdgpu_ttm_global_fini(adev); adev->mman.initialized = false; DRM_INFO("amdgpu: ttm finalized\n"); } @@ -2069,7 +1971,7 @@ int amdgpu_copy_buffer(struct amdgpu_ring *ring, uint64_t src_offset, unsigned i; int r; - if (direct_submit && !ring->ready) { + if (direct_submit && !ring->sched.ready) { DRM_ERROR("Trying to move memory with ring turned off.\n"); return -EINVAL; } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.h index fe8f276e9811..b5b2d101f7db 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.h @@ -39,8 +39,6 @@ #define AMDGPU_GTT_NUM_TRANSFER_WINDOWS 2 struct amdgpu_mman { - struct ttm_bo_global_ref bo_global_ref; - struct drm_global_reference mem_global_ref; struct ttm_bo_device bdev; bool mem_global_referenced; bool initialized; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.h index aa6641b944a0..7ac25a1c7853 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.h @@ -58,6 +58,17 @@ struct psp_firmware_header_v1_0 { }; /* version_major=1, version_minor=0 */ +struct ta_firmware_header_v1_0 { + struct common_firmware_header header; + uint32_t ta_xgmi_ucode_version; + uint32_t ta_xgmi_offset_bytes; + uint32_t ta_xgmi_size_bytes; + uint32_t ta_ras_ucode_version; + uint32_t ta_ras_offset_bytes; + uint32_t ta_ras_size_bytes; +}; + +/* version_major=1, version_minor=0 */ struct gfx_firmware_header_v1_0 { struct common_firmware_header header; uint32_t ucode_feature_version; @@ -170,6 +181,7 @@ union amdgpu_firmware_header { struct mc_firmware_header_v1_0 mc; struct smc_firmware_header_v1_0 smc; struct psp_firmware_header_v1_0 psp; + struct ta_firmware_header_v1_0 ta; struct gfx_firmware_header_v1_0 gfx; struct rlc_firmware_header_v1_0 rlc; struct rlc_firmware_header_v2_0 rlc_v2_0; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c index e5a6db6beab7..4e5d13e41f6a 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c @@ -692,6 +692,8 @@ static int amdgpu_uvd_cs_msg_decode(struct amdgpu_device *adev, uint32_t *msg, buf_sizes[0x1] = dpb_size; buf_sizes[0x2] = image_size; buf_sizes[0x4] = min_ctx_size; + /* store image width to adjust nb memory pstate */ + adev->uvd.decode_image_width = width; return 0; } @@ -1243,30 +1245,20 @@ int amdgpu_uvd_ring_test_ib(struct amdgpu_ring *ring, long timeout) { struct dma_fence *fence; long r; - uint32_t ip_instance = ring->me; r = amdgpu_uvd_get_create_msg(ring, 1, NULL); - if (r) { - DRM_ERROR("amdgpu: (%d)failed to get create msg (%ld).\n", ip_instance, r); + if (r) goto error; - } r = amdgpu_uvd_get_destroy_msg(ring, 1, true, &fence); - if (r) { - DRM_ERROR("amdgpu: (%d)failed to get destroy ib (%ld).\n", ip_instance, r); + if (r) goto error; - } r = dma_fence_wait_timeout(fence, false, timeout); - if (r == 0) { - DRM_ERROR("amdgpu: (%d)IB test timed out.\n", ip_instance); + if (r == 0) r = -ETIMEDOUT; - } else if (r < 0) { - DRM_ERROR("amdgpu: (%d)fence wait failed (%ld).\n", ip_instance, r); - } else { - DRM_DEBUG("ib test on (%d)ring %d succeeded\n", ip_instance, ring->idx); + else if (r > 0) r = 0; - } dma_fence_put(fence); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.h index a3ab1a41060f..5eb63288d157 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.h @@ -65,6 +65,8 @@ struct amdgpu_uvd { struct drm_sched_entity entity; struct delayed_work idle_work; unsigned harvest_config; + /* store image width to adjust nb memory state */ + unsigned decode_image_width; }; int amdgpu_uvd_sw_init(struct amdgpu_device *adev); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c index 5f3f54073818..c021b114c8a4 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c @@ -1032,8 +1032,10 @@ out: * @ib: the IB to execute * */ -void amdgpu_vce_ring_emit_ib(struct amdgpu_ring *ring, struct amdgpu_ib *ib, - unsigned vmid, bool ctx_switch) +void amdgpu_vce_ring_emit_ib(struct amdgpu_ring *ring, + struct amdgpu_job *job, + struct amdgpu_ib *ib, + uint32_t flags) { amdgpu_ring_write(ring, VCE_CMD_IB); amdgpu_ring_write(ring, lower_32_bits(ib->gpu_addr)); @@ -1079,11 +1081,9 @@ int amdgpu_vce_ring_test_ring(struct amdgpu_ring *ring) return 0; r = amdgpu_ring_alloc(ring, 16); - if (r) { - DRM_ERROR("amdgpu: vce failed to lock ring %d (%d).\n", - ring->idx, r); + if (r) return r; - } + amdgpu_ring_write(ring, VCE_CMD_END); amdgpu_ring_commit(ring); @@ -1093,14 +1093,8 @@ int amdgpu_vce_ring_test_ring(struct amdgpu_ring *ring) DRM_UDELAY(1); } - if (i < timeout) { - DRM_DEBUG("ring test on %d succeeded in %d usecs\n", - ring->idx, i); - } else { - DRM_ERROR("amdgpu: ring %d test failed\n", - ring->idx); + if (i >= timeout) r = -ETIMEDOUT; - } return r; } @@ -1121,27 +1115,19 @@ int amdgpu_vce_ring_test_ib(struct amdgpu_ring *ring, long timeout) return 0; r = amdgpu_vce_get_create_msg(ring, 1, NULL); - if (r) { - DRM_ERROR("amdgpu: failed to get create msg (%ld).\n", r); + if (r) goto error; - } r = amdgpu_vce_get_destroy_msg(ring, 1, true, &fence); - if (r) { - DRM_ERROR("amdgpu: failed to get destroy ib (%ld).\n", r); + if (r) goto error; - } r = dma_fence_wait_timeout(fence, false, timeout); - if (r == 0) { - DRM_ERROR("amdgpu: IB test timed out.\n"); + if (r == 0) r = -ETIMEDOUT; - } else if (r < 0) { - DRM_ERROR("amdgpu: fence wait failed (%ld).\n", r); - } else { - DRM_DEBUG("ib test on ring %d succeeded\n", ring->idx); + else if (r > 0) r = 0; - } + error: dma_fence_put(fence); return r; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.h index a1f209eed4c4..30ea54dd9117 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.h @@ -65,8 +65,8 @@ int amdgpu_vce_get_destroy_msg(struct amdgpu_ring *ring, uint32_t handle, 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); -void amdgpu_vce_ring_emit_ib(struct amdgpu_ring *ring, struct amdgpu_ib *ib, - unsigned vmid, bool ctx_switch); +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, unsigned flags); int amdgpu_vce_ring_test_ring(struct amdgpu_ring *ring); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c index 27da13df2f11..ecf6f96df2ad 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c @@ -262,7 +262,7 @@ static int amdgpu_vcn_pause_dpg_mode(struct amdgpu_device *adev, ring = &adev->vcn.ring_dec; WREG32_SOC15(UVD, 0, mmUVD_RBC_RB_WPTR, - RREG32_SOC15(UVD, 0, mmUVD_SCRATCH2)); + RREG32_SOC15(UVD, 0, mmUVD_SCRATCH2) & 0x7FFFFFFF); SOC15_WAIT_ON_RREG(UVD, 0, mmUVD_POWER_STATUS, UVD_PGFSM_CONFIG__UVDM_UVDU_PWR_ON, UVD_POWER_STATUS__UVD_POWER_STATUS_MASK, ret_code); @@ -322,7 +322,7 @@ static int amdgpu_vcn_pause_dpg_mode(struct amdgpu_device *adev, ring = &adev->vcn.ring_dec; WREG32_SOC15(UVD, 0, mmUVD_RBC_RB_WPTR, - RREG32_SOC15(UVD, 0, mmUVD_SCRATCH2)); + RREG32_SOC15(UVD, 0, mmUVD_SCRATCH2) & 0x7FFFFFFF); SOC15_WAIT_ON_RREG(UVD, 0, mmUVD_POWER_STATUS, UVD_PGFSM_CONFIG__UVDM_UVDU_PWR_ON, UVD_POWER_STATUS__UVD_POWER_STATUS_MASK, ret_code); @@ -396,16 +396,26 @@ void amdgpu_vcn_ring_begin_use(struct amdgpu_ring *ring) if (adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG) { struct dpg_pause_state new_state; + unsigned int fences = 0; + unsigned int i; - if (ring->funcs->type == AMDGPU_RING_TYPE_VCN_ENC) + for (i = 0; i < adev->vcn.num_enc_rings; ++i) { + fences += amdgpu_fence_count_emitted(&adev->vcn.ring_enc[i]); + } + if (fences) new_state.fw_based = VCN_DPG_STATE__PAUSE; else - new_state.fw_based = adev->vcn.pause_state.fw_based; + new_state.fw_based = VCN_DPG_STATE__UNPAUSE; - if (ring->funcs->type == AMDGPU_RING_TYPE_VCN_JPEG) + if (amdgpu_fence_count_emitted(&adev->vcn.ring_jpeg)) new_state.jpeg = VCN_DPG_STATE__PAUSE; else - new_state.jpeg = adev->vcn.pause_state.jpeg; + new_state.jpeg = VCN_DPG_STATE__UNPAUSE; + + if (ring->funcs->type == AMDGPU_RING_TYPE_VCN_ENC) + new_state.fw_based = VCN_DPG_STATE__PAUSE; + else if (ring->funcs->type == AMDGPU_RING_TYPE_VCN_JPEG) + new_state.jpeg = VCN_DPG_STATE__PAUSE; amdgpu_vcn_pause_dpg_mode(adev, &new_state); } @@ -425,11 +435,9 @@ int amdgpu_vcn_dec_ring_test_ring(struct amdgpu_ring *ring) WREG32(SOC15_REG_OFFSET(UVD, 0, mmUVD_SCRATCH9), 0xCAFEDEAD); r = amdgpu_ring_alloc(ring, 3); - if (r) { - DRM_ERROR("amdgpu: cp failed to lock ring %d (%d).\n", - ring->idx, r); + if (r) return r; - } + amdgpu_ring_write(ring, PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_SCRATCH9), 0)); amdgpu_ring_write(ring, 0xDEADBEEF); @@ -441,14 +449,9 @@ int amdgpu_vcn_dec_ring_test_ring(struct amdgpu_ring *ring) DRM_UDELAY(1); } - if (i < adev->usec_timeout) { - DRM_DEBUG("ring test on %d succeeded in %d usecs\n", - ring->idx, i); - } else { - DRM_ERROR("amdgpu: ring %d test failed (0x%08X)\n", - ring->idx, tmp); - r = -EINVAL; - } + if (i >= adev->usec_timeout) + r = -ETIMEDOUT; + return r; } @@ -570,30 +573,20 @@ int amdgpu_vcn_dec_ring_test_ib(struct amdgpu_ring *ring, long timeout) long r; r = amdgpu_vcn_dec_get_create_msg(ring, 1, NULL); - if (r) { - DRM_ERROR("amdgpu: failed to get create msg (%ld).\n", r); + if (r) goto error; - } r = amdgpu_vcn_dec_get_destroy_msg(ring, 1, &fence); - if (r) { - DRM_ERROR("amdgpu: failed to get destroy ib (%ld).\n", r); + if (r) goto error; - } r = dma_fence_wait_timeout(fence, false, timeout); - if (r == 0) { - DRM_ERROR("amdgpu: IB test timed out.\n"); + if (r == 0) r = -ETIMEDOUT; - } else if (r < 0) { - DRM_ERROR("amdgpu: fence wait failed (%ld).\n", r); - } else { - DRM_DEBUG("ib test on ring %d succeeded\n", ring->idx); + else if (r > 0) r = 0; - } dma_fence_put(fence); - error: return r; } @@ -606,11 +599,9 @@ int amdgpu_vcn_enc_ring_test_ring(struct amdgpu_ring *ring) int r; r = amdgpu_ring_alloc(ring, 16); - if (r) { - DRM_ERROR("amdgpu: vcn enc failed to lock ring %d (%d).\n", - ring->idx, r); + if (r) return r; - } + amdgpu_ring_write(ring, VCN_ENC_CMD_END); amdgpu_ring_commit(ring); @@ -620,14 +611,8 @@ int amdgpu_vcn_enc_ring_test_ring(struct amdgpu_ring *ring) DRM_UDELAY(1); } - if (i < adev->usec_timeout) { - DRM_DEBUG("ring test on %d succeeded in %d usecs\n", - ring->idx, i); - } else { - DRM_ERROR("amdgpu: ring %d test failed\n", - ring->idx); + if (i >= adev->usec_timeout) r = -ETIMEDOUT; - } return r; } @@ -742,27 +727,19 @@ int amdgpu_vcn_enc_ring_test_ib(struct amdgpu_ring *ring, long timeout) long r; r = amdgpu_vcn_enc_get_create_msg(ring, 1, NULL); - if (r) { - DRM_ERROR("amdgpu: failed to get create msg (%ld).\n", r); + if (r) goto error; - } r = amdgpu_vcn_enc_get_destroy_msg(ring, 1, &fence); - if (r) { - DRM_ERROR("amdgpu: failed to get destroy ib (%ld).\n", r); + if (r) goto error; - } r = dma_fence_wait_timeout(fence, false, timeout); - if (r == 0) { - DRM_ERROR("amdgpu: IB test timed out.\n"); + if (r == 0) r = -ETIMEDOUT; - } else if (r < 0) { - DRM_ERROR("amdgpu: fence wait failed (%ld).\n", r); - } else { - DRM_DEBUG("ib test on ring %d succeeded\n", ring->idx); + else if (r > 0) r = 0; - } + error: dma_fence_put(fence); return r; @@ -778,11 +755,8 @@ int amdgpu_vcn_jpeg_ring_test_ring(struct amdgpu_ring *ring) WREG32(SOC15_REG_OFFSET(UVD, 0, mmUVD_SCRATCH9), 0xCAFEDEAD); r = amdgpu_ring_alloc(ring, 3); - if (r) { - DRM_ERROR("amdgpu: cp failed to lock ring %d (%d).\n", - ring->idx, r); + if (r) return r; - } amdgpu_ring_write(ring, PACKETJ(SOC15_REG_OFFSET(UVD, 0, mmUVD_SCRATCH9), 0, 0, 0)); @@ -796,14 +770,8 @@ int amdgpu_vcn_jpeg_ring_test_ring(struct amdgpu_ring *ring) DRM_UDELAY(1); } - if (i < adev->usec_timeout) { - DRM_DEBUG("ring test on %d succeeded in %d usecs\n", - ring->idx, i); - } else { - DRM_ERROR("amdgpu: ring %d test failed (0x%08X)\n", - ring->idx, tmp); - r = -EINVAL; - } + if (i >= adev->usec_timeout) + r = -ETIMEDOUT; return r; } @@ -856,21 +824,18 @@ int amdgpu_vcn_jpeg_ring_test_ib(struct amdgpu_ring *ring, long timeout) long r = 0; r = amdgpu_vcn_jpeg_set_reg(ring, 1, &fence); - if (r) { - DRM_ERROR("amdgpu: failed to set jpeg register (%ld).\n", r); + if (r) goto error; - } r = dma_fence_wait_timeout(fence, false, timeout); if (r == 0) { - DRM_ERROR("amdgpu: IB test timed out.\n"); r = -ETIMEDOUT; goto error; } else if (r < 0) { - DRM_ERROR("amdgpu: fence wait failed (%ld).\n", r); goto error; - } else + } else { r = 0; + } for (i = 0; i < adev->usec_timeout; i++) { tmp = RREG32(SOC15_REG_OFFSET(UVD, 0, mmUVD_SCRATCH9)); @@ -879,15 +844,10 @@ int amdgpu_vcn_jpeg_ring_test_ib(struct amdgpu_ring *ring, long timeout) DRM_UDELAY(1); } - if (i < adev->usec_timeout) - DRM_DEBUG("ib test on ring %d succeeded\n", ring->idx); - else { - DRM_ERROR("ib test failed (0x%08X)\n", tmp); - r = -EINVAL; - } + if (i >= adev->usec_timeout) + r = -ETIMEDOUT; dma_fence_put(fence); - error: return r; } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.c index f2f358aa0597..462a04e0f5e6 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.c @@ -23,16 +23,6 @@ #include "amdgpu.h" -uint64_t amdgpu_csa_vaddr(struct amdgpu_device *adev) -{ - uint64_t addr = adev->vm_manager.max_pfn << AMDGPU_GPU_PAGE_SHIFT; - - addr -= AMDGPU_VA_RESERVED_SIZE; - addr = amdgpu_gmc_sign_extend(addr); - - return addr; -} - bool amdgpu_virt_mmio_blocked(struct amdgpu_device *adev) { /* By now all MMIO pages except mailbox are blocked */ @@ -41,88 +31,6 @@ bool amdgpu_virt_mmio_blocked(struct amdgpu_device *adev) return RREG32_NO_KIQ(0xc040) == 0xffffffff; } -int amdgpu_allocate_static_csa(struct amdgpu_device *adev) -{ - int r; - void *ptr; - - r = amdgpu_bo_create_kernel(adev, AMDGPU_CSA_SIZE, PAGE_SIZE, - AMDGPU_GEM_DOMAIN_VRAM, &adev->virt.csa_obj, - &adev->virt.csa_vmid0_addr, &ptr); - if (r) - return r; - - memset(ptr, 0, AMDGPU_CSA_SIZE); - return 0; -} - -void amdgpu_free_static_csa(struct amdgpu_device *adev) { - amdgpu_bo_free_kernel(&adev->virt.csa_obj, - &adev->virt.csa_vmid0_addr, - NULL); -} - -/* - * amdgpu_map_static_csa should be called during amdgpu_vm_init - * it maps virtual address amdgpu_csa_vaddr() to this VM, and each command - * submission of GFX should use this virtual address within META_DATA init - * package to support SRIOV gfx preemption. - */ -int amdgpu_map_static_csa(struct amdgpu_device *adev, struct amdgpu_vm *vm, - struct amdgpu_bo_va **bo_va) -{ - uint64_t csa_addr = amdgpu_csa_vaddr(adev) & AMDGPU_GMC_HOLE_MASK; - struct ww_acquire_ctx ticket; - struct list_head list; - struct amdgpu_bo_list_entry pd; - struct ttm_validate_buffer csa_tv; - int r; - - INIT_LIST_HEAD(&list); - INIT_LIST_HEAD(&csa_tv.head); - csa_tv.bo = &adev->virt.csa_obj->tbo; - csa_tv.shared = true; - - 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 CSA,PD BOs: err=%d\n", r); - return r; - } - - *bo_va = amdgpu_vm_bo_add(adev, vm, adev->virt.csa_obj); - if (!*bo_va) { - ttm_eu_backoff_reservation(&ticket, &list); - DRM_ERROR("failed to create bo_va for static CSA\n"); - return -ENOMEM; - } - - r = amdgpu_vm_alloc_pts(adev, (*bo_va)->base.vm, csa_addr, - AMDGPU_CSA_SIZE); - if (r) { - DRM_ERROR("failed to allocate pts for static CSA, err=%d\n", r); - amdgpu_vm_bo_rmv(adev, *bo_va); - ttm_eu_backoff_reservation(&ticket, &list); - return r; - } - - r = amdgpu_vm_bo_map(adev, *bo_va, csa_addr, 0, AMDGPU_CSA_SIZE, - AMDGPU_PTE_READABLE | AMDGPU_PTE_WRITEABLE | - AMDGPU_PTE_EXECUTABLE); - - if (r) { - DRM_ERROR("failed to do bo_map on static CSA, err=%d\n", r); - amdgpu_vm_bo_rmv(adev, *bo_va); - ttm_eu_backoff_reservation(&ticket, &list); - return r; - } - - ttm_eu_backoff_reservation(&ticket, &list); - return 0; -} - void amdgpu_virt_init_setting(struct amdgpu_device *adev) { /* enable virtual display */ @@ -162,9 +70,7 @@ uint32_t amdgpu_virt_kiq_rreg(struct amdgpu_device *adev, uint32_t reg) if (r < 1 && (adev->in_gpu_reset || in_interrupt())) goto failed_kiq_read; - if (in_interrupt()) - might_sleep(); - + might_sleep(); while (r < 1 && cnt++ < MAX_KIQ_REG_TRY) { msleep(MAX_KIQ_REG_BAILOUT_INTERVAL); r = amdgpu_fence_wait_polling(ring, seq, MAX_KIQ_REG_WAIT); @@ -210,9 +116,7 @@ void amdgpu_virt_kiq_wreg(struct amdgpu_device *adev, uint32_t reg, uint32_t v) if (r < 1 && (adev->in_gpu_reset || in_interrupt())) goto failed_kiq_write; - if (in_interrupt()) - might_sleep(); - + might_sleep(); while (r < 1 && cnt++ < MAX_KIQ_REG_TRY) { msleep(MAX_KIQ_REG_BAILOUT_INTERVAL); @@ -228,6 +132,46 @@ failed_kiq_write: pr_err("failed to write reg:%x\n", reg); } +void amdgpu_virt_kiq_reg_write_reg_wait(struct amdgpu_device *adev, + uint32_t reg0, uint32_t reg1, + uint32_t ref, uint32_t mask) +{ + struct amdgpu_kiq *kiq = &adev->gfx.kiq; + struct amdgpu_ring *ring = &kiq->ring; + signed long r, cnt = 0; + unsigned long flags; + uint32_t seq; + + spin_lock_irqsave(&kiq->ring_lock, flags); + amdgpu_ring_alloc(ring, 32); + amdgpu_ring_emit_reg_write_reg_wait(ring, reg0, reg1, + ref, mask); + amdgpu_fence_emit_polling(ring, &seq); + amdgpu_ring_commit(ring); + spin_unlock_irqrestore(&kiq->ring_lock, flags); + + r = amdgpu_fence_wait_polling(ring, seq, MAX_KIQ_REG_WAIT); + + /* don't wait anymore for IRQ context */ + if (r < 1 && in_interrupt()) + goto failed_kiq; + + might_sleep(); + while (r < 1 && cnt++ < MAX_KIQ_REG_TRY) { + + msleep(MAX_KIQ_REG_BAILOUT_INTERVAL); + r = amdgpu_fence_wait_polling(ring, seq, MAX_KIQ_REG_WAIT); + } + + if (cnt > MAX_KIQ_REG_TRY) + goto failed_kiq; + + return; + +failed_kiq: + pr_err("failed to write reg %x wait reg %x\n", reg0, reg1); +} + /** * amdgpu_virt_request_full_gpu() - request full gpu access * @amdgpu: amdgpu device. @@ -390,7 +334,7 @@ void amdgpu_virt_init_data_exchange(struct amdgpu_device *adev) if (adev->fw_vram_usage.va != NULL) { adev->virt.fw_reserve.p_pf2vf = - (struct amdgim_pf2vf_info_header *)( + (struct amd_sriov_msg_pf2vf_info_header *)( adev->fw_vram_usage.va + AMDGIM_DATAEXCHANGE_OFFSET); AMDGPU_FW_VRAM_PF2VF_READ(adev, header.size, &pf2vf_size); AMDGPU_FW_VRAM_PF2VF_READ(adev, checksum, &checksum); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.h index 880ac113a3a9..722deefc0a7e 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.h @@ -63,8 +63,8 @@ struct amdgpu_virt_ops { * Firmware Reserve Frame buffer */ struct amdgpu_virt_fw_reserve { - struct amdgim_pf2vf_info_header *p_pf2vf; - struct amdgim_vf2pf_info_header *p_vf2pf; + struct amd_sriov_msg_pf2vf_info_header *p_pf2vf; + struct amd_sriov_msg_vf2pf_info_header *p_vf2pf; unsigned int checksum_key; }; /* @@ -85,15 +85,17 @@ enum AMDGIM_FEATURE_FLAG { AMDGIM_FEATURE_GIM_FLR_VRAMLOST = 0x4, }; -struct amdgim_pf2vf_info_header { +struct amd_sriov_msg_pf2vf_info_header { /* the total structure size in byte. */ uint32_t size; /* version of this structure, written by the GIM */ uint32_t version; + /* reserved */ + uint32_t reserved[2]; } __aligned(4); struct amdgim_pf2vf_info_v1 { /* header contains size and version */ - struct amdgim_pf2vf_info_header header; + struct amd_sriov_msg_pf2vf_info_header header; /* max_width * max_height */ unsigned int uvd_enc_max_pixels_count; /* 16x16 pixels/sec, codec independent */ @@ -112,7 +114,7 @@ struct amdgim_pf2vf_info_v1 { struct amdgim_pf2vf_info_v2 { /* header contains size and version */ - struct amdgim_pf2vf_info_header header; + struct amd_sriov_msg_pf2vf_info_header header; /* use private key from mailbox 2 to create chueksum */ uint32_t checksum; /* The features flags of the GIM driver supports. */ @@ -137,20 +139,22 @@ struct amdgim_pf2vf_info_v2 { uint64_t vcefw_kboffset; /* VCE FW size in KB */ uint32_t vcefw_ksize; - uint32_t reserved[AMDGIM_GET_STRUCTURE_RESERVED_SIZE(256, 0, 0, (9 + sizeof(struct amdgim_pf2vf_info_header)/sizeof(uint32_t)), 3)]; + uint32_t reserved[AMDGIM_GET_STRUCTURE_RESERVED_SIZE(256, 0, 0, (9 + sizeof(struct amd_sriov_msg_pf2vf_info_header)/sizeof(uint32_t)), 3)]; } __aligned(4); -struct amdgim_vf2pf_info_header { +struct amd_sriov_msg_vf2pf_info_header { /* the total structure size in byte. */ uint32_t size; /*version of this structure, written by the guest */ uint32_t version; + /* reserved */ + uint32_t reserved[2]; } __aligned(4); struct amdgim_vf2pf_info_v1 { /* header contains size and version */ - struct amdgim_vf2pf_info_header header; + struct amd_sriov_msg_vf2pf_info_header header; /* driver version */ char driver_version[64]; /* driver certification, 1=WHQL, 0=None */ @@ -180,7 +184,7 @@ struct amdgim_vf2pf_info_v1 { struct amdgim_vf2pf_info_v2 { /* header contains size and version */ - struct amdgim_vf2pf_info_header header; + struct amd_sriov_msg_vf2pf_info_header header; uint32_t checksum; /* driver version */ uint8_t driver_version[64]; @@ -206,7 +210,7 @@ struct amdgim_vf2pf_info_v2 { uint32_t uvd_enc_usage; /* guest uvd engine usage percentage. 0xffff means N/A. */ uint32_t uvd_enc_health; - uint32_t reserved[AMDGIM_GET_STRUCTURE_RESERVED_SIZE(256, 64, 0, (12 + sizeof(struct amdgim_vf2pf_info_header)/sizeof(uint32_t)), 0)]; + uint32_t reserved[AMDGIM_GET_STRUCTURE_RESERVED_SIZE(256, 64, 0, (12 + sizeof(struct amd_sriov_msg_vf2pf_info_header)/sizeof(uint32_t)), 0)]; } __aligned(4); #define AMDGPU_FW_VRAM_VF2PF_VER 2 @@ -238,7 +242,6 @@ typedef struct amdgim_vf2pf_info_v2 amdgim_vf2pf_info ; struct amdgpu_virt { uint32_t caps; struct amdgpu_bo *csa_obj; - uint64_t csa_vmid0_addr; bool chained_ib_support; uint32_t reg_val_offs; struct amdgpu_irq_src ack_irq; @@ -251,8 +254,6 @@ struct amdgpu_virt { uint32_t gim_feature; }; -#define AMDGPU_CSA_SIZE (8 * 1024) - #define amdgpu_sriov_enabled(adev) \ ((adev)->virt.caps & AMDGPU_SRIOV_CAPS_ENABLE_IOV) @@ -277,17 +278,13 @@ static inline bool is_virtual_machine(void) #endif } -struct amdgpu_vm; - -uint64_t amdgpu_csa_vaddr(struct amdgpu_device *adev); bool amdgpu_virt_mmio_blocked(struct amdgpu_device *adev); -int amdgpu_allocate_static_csa(struct amdgpu_device *adev); -int amdgpu_map_static_csa(struct amdgpu_device *adev, struct amdgpu_vm *vm, - struct amdgpu_bo_va **bo_va); -void amdgpu_free_static_csa(struct amdgpu_device *adev); void amdgpu_virt_init_setting(struct amdgpu_device *adev); uint32_t amdgpu_virt_kiq_rreg(struct amdgpu_device *adev, uint32_t reg); void amdgpu_virt_kiq_wreg(struct amdgpu_device *adev, uint32_t reg, uint32_t v); +void amdgpu_virt_kiq_reg_write_reg_wait(struct amdgpu_device *adev, + uint32_t reg0, uint32_t rreg1, + uint32_t ref, uint32_t mask); int amdgpu_virt_request_full_gpu(struct amdgpu_device *adev, bool init); int amdgpu_virt_release_full_gpu(struct amdgpu_device *adev, bool init); int amdgpu_virt_reset_gpu(struct amdgpu_device *adev); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c index 0877ff9a9594..16fcb56c232b 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c @@ -107,14 +107,6 @@ struct amdgpu_pte_update_params { * DMA addresses to use for mapping, used during VM update by CPU */ dma_addr_t *pages_addr; - - /** - * @kptr: - * - * Kernel pointer of PD/PT BO that needs to be updated, - * used during VM update by CPU - */ - void *kptr; }; /** @@ -617,11 +609,34 @@ void amdgpu_vm_get_pd_bo(struct amdgpu_vm *vm, { entry->priority = 0; entry->tv.bo = &vm->root.base.bo->tbo; - entry->tv.shared = true; + /* One for the VM updates, one for TTM and one for the CS job */ + entry->tv.num_shared = 3; entry->user_pages = NULL; list_add(&entry->tv.head, validated); } +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->mem.placement & TTM_PL_FLAG_NO_EVICT) + 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.resv == vm->root.base.bo->tbo.resv) + vm->bulk_moveable = false; + } + +} /** * amdgpu_vm_move_to_lru_tail - move all BOs to the end of LRU * @@ -637,12 +652,14 @@ void amdgpu_vm_move_to_lru_tail(struct amdgpu_device *adev, struct ttm_bo_global *glob = adev->mman.bdev.glob; struct amdgpu_vm_bo_base *bo_base; +#if 0 if (vm->bulk_moveable) { spin_lock(&glob->lru_lock); ttm_bo_bulk_move_lru_tail(&vm->lru_bulk_move); spin_unlock(&glob->lru_lock); return; } +#endif memset(&vm->lru_bulk_move, 0, sizeof(vm->lru_bulk_move)); @@ -773,10 +790,6 @@ static int amdgpu_vm_clear_bo(struct amdgpu_device *adev, ring = container_of(vm->entity.rq->sched, struct amdgpu_ring, sched); - r = reservation_object_reserve_shared(bo->tbo.resv); - if (r) - return r; - r = ttm_bo_validate(&bo->tbo, &bo->placement, &ctx); if (r) goto error; @@ -802,15 +815,22 @@ static int amdgpu_vm_clear_bo(struct amdgpu_device *adev, addr += ats_entries * 8; } - if (entries) + if (entries) { + uint64_t value = 0; + + /* Workaround for fault priority problem on GMC9 */ + if (level == AMDGPU_VM_PTB && adev->asic_type >= CHIP_VEGA10) + value = AMDGPU_PTE_EXECUTABLE; + amdgpu_vm_set_pte_pde(adev, &job->ibs[0], addr, 0, - entries, 0, 0); + entries, 0, value); + } amdgpu_ring_pad_ib(ring, &job->ibs[0]); WARN_ON(job->ibs[0].length_dw > 64); r = amdgpu_sync_resv(adev, &job->sync, bo->tbo.resv, - AMDGPU_FENCE_OWNER_UNDEFINED, false); + AMDGPU_FENCE_OWNER_KFD, false); if (r) goto error_free; @@ -850,9 +870,6 @@ static void amdgpu_vm_bo_param(struct amdgpu_device *adev, struct amdgpu_vm *vm, bp->size = amdgpu_vm_bo_size(adev, level); bp->byte_align = AMDGPU_GPU_PAGE_SIZE; bp->domain = AMDGPU_GEM_DOMAIN_VRAM; - if (bp->size <= PAGE_SIZE && adev->asic_type >= CHIP_VEGA10 && - adev->flags & AMD_IS_APU) - bp->domain |= AMDGPU_GEM_DOMAIN_GTT; bp->domain = amdgpu_bo_get_preferred_pin_domain(adev, bp->domain); bp->flags = AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS | AMDGPU_GEM_CREATE_CPU_GTT_USWC; @@ -932,10 +949,6 @@ int amdgpu_vm_alloc_pts(struct amdgpu_device *adev, if (r) return r; - r = amdgpu_vm_clear_bo(adev, vm, pt, cursor.level, ats); - if (r) - goto error_free_pt; - if (vm->use_cpu_for_update) { r = amdgpu_bo_kmap(pt, NULL); if (r) @@ -948,6 +961,10 @@ int amdgpu_vm_alloc_pts(struct amdgpu_device *adev, pt->parent = amdgpu_bo_ref(cursor.parent->base.bo); amdgpu_vm_bo_base_init(&entry->base, vm, pt); + + r = amdgpu_vm_clear_bo(adev, vm, pt, cursor.level, ats); + if (r) + goto error_free_pt; } return 0; @@ -1317,31 +1334,6 @@ static void amdgpu_vm_cpu_set_ptes(struct amdgpu_pte_update_params *params, } } - -/** - * amdgpu_vm_wait_pd - Wait for PT BOs to be free. - * - * @adev: amdgpu_device pointer - * @vm: related vm - * @owner: fence owner - * - * Returns: - * 0 on success, errno otherwise. - */ -static int amdgpu_vm_wait_pd(struct amdgpu_device *adev, struct amdgpu_vm *vm, - void *owner) -{ - struct amdgpu_sync sync; - int r; - - amdgpu_sync_create(&sync); - amdgpu_sync_resv(adev, &sync, vm->root.base.bo->tbo.resv, owner, false); - r = amdgpu_sync_wait(&sync, true); - amdgpu_sync_free(&sync); - - return r; -} - /** * amdgpu_vm_update_func - helper to call update function * @@ -1436,7 +1428,8 @@ restart: params.adev = adev; if (vm->use_cpu_for_update) { - r = amdgpu_vm_wait_pd(adev, vm, AMDGPU_FENCE_OWNER_VM); + r = amdgpu_bo_sync_wait(vm->root.base.bo, + AMDGPU_FENCE_OWNER_VM, true); if (unlikely(r)) return r; @@ -1509,20 +1502,27 @@ error: } /** - * amdgpu_vm_update_huge - figure out parameters for PTE updates + * 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_huge(struct amdgpu_pte_update_params *params, - struct amdgpu_bo *bo, unsigned level, - uint64_t pe, uint64_t addr, - unsigned count, uint32_t incr, - uint64_t flags) +static void amdgpu_vm_update_flags(struct amdgpu_pte_update_params *params, + struct amdgpu_bo *bo, unsigned level, + uint64_t pe, uint64_t addr, + unsigned 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; } amdgpu_vm_update_func(params, bo, pe, addr, count, incr, flags); @@ -1679,9 +1679,9 @@ static int amdgpu_vm_update_ptes(struct amdgpu_pte_update_params *params, uint64_t upd_end = min(entry_end, frag_end); unsigned nptes = (upd_end - frag_start) >> shift; - amdgpu_vm_update_huge(params, pt, cursor.level, - pe_start, dst, nptes, incr, - flags | AMDGPU_PTE_FRAG(frag)); + amdgpu_vm_update_flags(params, pt, cursor.level, + pe_start, dst, nptes, incr, + flags | AMDGPU_PTE_FRAG(frag)); pe_start += nptes * 8; dst += (uint64_t)nptes * AMDGPU_GPU_PAGE_SIZE << shift; @@ -1750,22 +1750,29 @@ static int amdgpu_vm_bo_update_mapping(struct amdgpu_device *adev, params.adev = adev; params.vm = vm; - /* sync to everything on unmapping */ + /* sync to everything except eviction fences on unmapping */ if (!(flags & AMDGPU_PTE_VALID)) - owner = AMDGPU_FENCE_OWNER_UNDEFINED; + owner = AMDGPU_FENCE_OWNER_KFD; if (vm->use_cpu_for_update) { /* params.src is used as flag to indicate system Memory */ if (pages_addr) params.src = ~0; - /* Wait for PT BOs to be free. PTs share the same resv. object + /* Wait for PT BOs to be idle. PTs share the same resv. object * as the root PD BO */ - r = amdgpu_vm_wait_pd(adev, vm, owner); + r = amdgpu_bo_sync_wait(vm->root.base.bo, owner, true); if (unlikely(r)) return r; + /* Wait for any BO move to be completed */ + if (exclusive) { + r = dma_fence_wait(exclusive, true); + if (unlikely(r)) + return r; + } + params.func = amdgpu_vm_cpu_set_ptes; params.pages_addr = pages_addr; return amdgpu_vm_update_ptes(¶ms, start, last + 1, @@ -1779,13 +1786,12 @@ static int amdgpu_vm_bo_update_mapping(struct amdgpu_device *adev, /* * reserve space for two commands every (1 << BLOCK_SIZE) * entries or 2k dwords (whatever is smaller) - * - * The second command is for the shadow pagetables. */ + ncmds = ((nptes >> min(adev->vm_manager.block_size, 11u)) + 1); + + /* The second command is for the shadow pagetables. */ if (vm->root.base.bo->shadow) - ncmds = ((nptes >> min(adev->vm_manager.block_size, 11u)) + 1) * 2; - else - ncmds = ((nptes >> min(adev->vm_manager.block_size, 11u)) + 1); + ncmds *= 2; /* padding, etc. */ ndw = 64; @@ -1804,10 +1810,11 @@ static int amdgpu_vm_bo_update_mapping(struct amdgpu_device *adev, ndw += ncmds * 10; /* extra commands for begin/end fragments */ + ncmds = 2 * adev->vm_manager.fragment_size; if (vm->root.base.bo->shadow) - ndw += 2 * 10 * adev->vm_manager.fragment_size * 2; - else - ndw += 2 * 10 * adev->vm_manager.fragment_size; + ncmds *= 2; + + ndw += 10 * ncmds; params.func = amdgpu_vm_do_set_ptes; } @@ -1844,10 +1851,6 @@ static int amdgpu_vm_bo_update_mapping(struct amdgpu_device *adev, if (r) goto error_free; - r = reservation_object_reserve_shared(vm->root.base.bo->tbo.resv); - if (r) - goto error_free; - r = amdgpu_vm_update_ptes(¶ms, start, last + 1, addr, flags); if (r) goto error_free; @@ -3013,7 +3016,7 @@ int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm, } DRM_DEBUG_DRIVER("VM update mode is %s\n", vm->use_cpu_for_update ? "CPU" : "SDMA"); - WARN_ONCE((vm->use_cpu_for_update & !amdgpu_gmc_vram_full_visible(&adev->gmc)), + WARN_ONCE((vm->use_cpu_for_update && !amdgpu_gmc_vram_full_visible(&adev->gmc)), "CPU update of VM recommended only for large BAR system\n"); vm->last_update = NULL; @@ -3028,13 +3031,18 @@ int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm, if (r) goto error_free_root; + r = reservation_object_reserve_shared(root->tbo.resv, 1); + if (r) + goto error_unreserve; + + amdgpu_vm_bo_base_init(&vm->root.base, vm, root); + r = amdgpu_vm_clear_bo(adev, vm, root, adev->vm_manager.root_level, vm->pte_support_ats); if (r) goto error_unreserve; - amdgpu_vm_bo_base_init(&vm->root.base, vm, root); amdgpu_bo_unreserve(vm->root.base.bo); if (pasid) { @@ -3057,7 +3065,6 @@ int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm, } INIT_KFIFO(vm->faults); - vm->fault_credit = 16; return 0; @@ -3140,7 +3147,7 @@ int amdgpu_vm_make_compute(struct amdgpu_device *adev, struct amdgpu_vm *vm, uns vm->pte_support_ats = pte_support_ats; DRM_DEBUG_DRIVER("VM update mode is %s\n", vm->use_cpu_for_update ? "CPU" : "SDMA"); - WARN_ONCE((vm->use_cpu_for_update & !amdgpu_gmc_vram_full_visible(&adev->gmc)), + WARN_ONCE((vm->use_cpu_for_update && !amdgpu_gmc_vram_full_visible(&adev->gmc)), "CPU update of VM recommended only for large BAR system\n"); if (vm->pasid) { @@ -3270,42 +3277,6 @@ void amdgpu_vm_fini(struct amdgpu_device *adev, struct amdgpu_vm *vm) } /** - * amdgpu_vm_pasid_fault_credit - Check fault credit for given PASID - * - * @adev: amdgpu_device pointer - * @pasid: PASID do identify the VM - * - * This function is expected to be called in interrupt context. - * - * Returns: - * True if there was fault credit, false otherwise - */ -bool amdgpu_vm_pasid_fault_credit(struct amdgpu_device *adev, - unsigned int pasid) -{ - struct amdgpu_vm *vm; - - spin_lock(&adev->vm_manager.pasid_lock); - vm = idr_find(&adev->vm_manager.pasid_idr, pasid); - if (!vm) { - /* VM not found, can't track fault credit */ - spin_unlock(&adev->vm_manager.pasid_lock); - return true; - } - - /* No lock needed. only accessed by IRQ handler */ - if (!vm->fault_credit) { - /* Too many faults in this VM */ - spin_unlock(&adev->vm_manager.pasid_lock); - return false; - } - - vm->fault_credit--; - spin_unlock(&adev->vm_manager.pasid_lock); - return true; -} - -/** * amdgpu_vm_manager_init - init the VM manager * * @adev: amdgpu_device pointer @@ -3406,14 +3377,15 @@ void amdgpu_vm_get_task_info(struct amdgpu_device *adev, unsigned int pasid, struct amdgpu_task_info *task_info) { struct amdgpu_vm *vm; + unsigned long flags; - spin_lock(&adev->vm_manager.pasid_lock); + spin_lock_irqsave(&adev->vm_manager.pasid_lock, flags); vm = idr_find(&adev->vm_manager.pasid_idr, pasid); if (vm) *task_info = vm->task_info; - spin_unlock(&adev->vm_manager.pasid_lock); + spin_unlock_irqrestore(&adev->vm_manager.pasid_lock, flags); } /** diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h index 2a8898d19c8b..81ff8177f092 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h @@ -229,9 +229,6 @@ struct amdgpu_vm { /* Up to 128 pending retry page faults */ DECLARE_KFIFO(faults, u64, 128); - /* Limit non-retry fault storms */ - unsigned int fault_credit; - /* Points to the KFD process VM info */ struct amdkfd_process_info *process_info; @@ -299,8 +296,6 @@ int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm, int amdgpu_vm_make_compute(struct amdgpu_device *adev, struct amdgpu_vm *vm, unsigned int pasid); void amdgpu_vm_release_compute(struct amdgpu_device *adev, struct amdgpu_vm *vm); void amdgpu_vm_fini(struct amdgpu_device *adev, struct amdgpu_vm *vm); -bool amdgpu_vm_pasid_fault_credit(struct amdgpu_device *adev, - unsigned int pasid); void amdgpu_vm_get_pd_bo(struct amdgpu_vm *vm, struct list_head *validated, struct amdgpu_bo_list_entry *entry); @@ -368,4 +363,6 @@ int amdgpu_vm_add_fault(struct amdgpu_retryfault_hashtable *fault_hash, u64 key) void amdgpu_vm_clear_fault(struct amdgpu_retryfault_hashtable *fault_hash, u64 key); +void amdgpu_vm_del_from_lru_notify(struct ttm_buffer_object *bo); + #endif diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c index 897afbb348c1..407dd16cc35c 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c @@ -23,7 +23,7 @@ */ #include <linux/list.h> #include "amdgpu.h" -#include "amdgpu_psp.h" +#include "amdgpu_xgmi.h" static DEFINE_MUTEX(xgmi_mutex); @@ -31,89 +31,154 @@ static DEFINE_MUTEX(xgmi_mutex); #define AMDGPU_MAX_XGMI_HIVE 8 #define AMDGPU_MAX_XGMI_DEVICE_PER_HIVE 4 -struct amdgpu_hive_info { - uint64_t hive_id; - struct list_head device_list; -}; - static struct amdgpu_hive_info xgmi_hives[AMDGPU_MAX_XGMI_HIVE]; static unsigned hive_count = 0; -static struct amdgpu_hive_info *amdgpu_get_xgmi_hive(struct amdgpu_device *adev) + +void *amdgpu_xgmi_hive_try_lock(struct amdgpu_hive_info *hive) +{ + return &hive->device_list; +} + +struct amdgpu_hive_info *amdgpu_get_xgmi_hive(struct amdgpu_device *adev, int lock) { int i; struct amdgpu_hive_info *tmp; if (!adev->gmc.xgmi.hive_id) return NULL; + + mutex_lock(&xgmi_mutex); + for (i = 0 ; i < hive_count; ++i) { tmp = &xgmi_hives[i]; - if (tmp->hive_id == adev->gmc.xgmi.hive_id) + if (tmp->hive_id == adev->gmc.xgmi.hive_id) { + if (lock) + mutex_lock(&tmp->hive_lock); + mutex_unlock(&xgmi_mutex); return tmp; + } } - if (i >= AMDGPU_MAX_XGMI_HIVE) + if (i >= AMDGPU_MAX_XGMI_HIVE) { + mutex_unlock(&xgmi_mutex); return NULL; + } /* initialize new hive if not exist */ tmp = &xgmi_hives[hive_count++]; tmp->hive_id = adev->gmc.xgmi.hive_id; INIT_LIST_HEAD(&tmp->device_list); + mutex_init(&tmp->hive_lock); + mutex_init(&tmp->reset_lock); + if (lock) + mutex_lock(&tmp->hive_lock); + + mutex_unlock(&xgmi_mutex); + return tmp; } +int amdgpu_xgmi_update_topology(struct amdgpu_hive_info *hive, struct amdgpu_device *adev) +{ + int ret = -EINVAL; + + /* Each psp need to set the latest topology */ + ret = psp_xgmi_set_topology_info(&adev->psp, + hive->number_devices, + &hive->topology_info); + if (ret) + dev_err(adev->dev, + "XGMI: Set topology failure on device %llx, hive %llx, ret %d", + adev->gmc.xgmi.node_id, + adev->gmc.xgmi.hive_id, ret); + + return ret; +} + int amdgpu_xgmi_add_device(struct amdgpu_device *adev) { - struct psp_xgmi_topology_info tmp_topology[AMDGPU_MAX_XGMI_DEVICE_PER_HIVE]; + struct psp_xgmi_topology_info *hive_topology; struct amdgpu_hive_info *hive; struct amdgpu_xgmi *entry; - struct amdgpu_device *tmp_adev; + struct amdgpu_device *tmp_adev = NULL; int count = 0, ret = -EINVAL; - if ((adev->asic_type < CHIP_VEGA20) || - (adev->flags & AMD_IS_APU) ) + if (!adev->gmc.xgmi.supported) return 0; - adev->gmc.xgmi.device_id = psp_xgmi_get_device_id(&adev->psp); - adev->gmc.xgmi.hive_id = psp_xgmi_get_hive_id(&adev->psp); - - memset(&tmp_topology[0], 0, sizeof(tmp_topology)); - mutex_lock(&xgmi_mutex); - hive = amdgpu_get_xgmi_hive(adev); - if (!hive) - goto exit; - list_add_tail(&adev->gmc.xgmi.head, &hive->device_list); - list_for_each_entry(entry, &hive->device_list, head) - tmp_topology[count++].device_id = entry->device_id; + ret = psp_xgmi_get_node_id(&adev->psp, &adev->gmc.xgmi.node_id); + if (ret) { + dev_err(adev->dev, + "XGMI: Failed to get node id\n"); + return ret; + } - ret = psp_xgmi_get_topology_info(&adev->psp, count, tmp_topology); + ret = psp_xgmi_get_hive_id(&adev->psp, &adev->gmc.xgmi.hive_id); if (ret) { dev_err(adev->dev, - "XGMI: Get topology failure on device %llx, hive %llx, ret %d", - adev->gmc.xgmi.device_id, - adev->gmc.xgmi.hive_id, ret); + "XGMI: Failed to get hive id\n"); + return ret; + } + + hive = amdgpu_get_xgmi_hive(adev, 1); + if (!hive) { + ret = -EINVAL; + dev_err(adev->dev, + "XGMI: node 0x%llx, can not match hive 0x%llx in the hive list.\n", + adev->gmc.xgmi.node_id, adev->gmc.xgmi.hive_id); goto exit; } - /* Each psp need to set the latest topology */ + + hive_topology = &hive->topology_info; + + list_add_tail(&adev->gmc.xgmi.head, &hive->device_list); + list_for_each_entry(entry, &hive->device_list, head) + hive_topology->nodes[count++].node_id = entry->node_id; + hive->number_devices = count; + + /* Each psp need to get the latest topology */ list_for_each_entry(tmp_adev, &hive->device_list, gmc.xgmi.head) { - ret = psp_xgmi_set_topology_info(&tmp_adev->psp, count, tmp_topology); + ret = psp_xgmi_get_topology_info(&tmp_adev->psp, count, hive_topology); if (ret) { dev_err(tmp_adev->dev, - "XGMI: Set topology failure on device %llx, hive %llx, ret %d", - tmp_adev->gmc.xgmi.device_id, + "XGMI: Get topology failure on device %llx, hive %llx, ret %d", + tmp_adev->gmc.xgmi.node_id, tmp_adev->gmc.xgmi.hive_id, ret); - /* To do : continue with some node failed or disable the whole hive */ + /* To do : continue with some node failed or disable the whole hive */ break; } } - if (!ret) - dev_info(adev->dev, "XGMI: Add node %d to hive 0x%llx.\n", - adev->gmc.xgmi.physical_node_id, - adev->gmc.xgmi.hive_id); + list_for_each_entry(tmp_adev, &hive->device_list, gmc.xgmi.head) { + ret = amdgpu_xgmi_update_topology(hive, tmp_adev); + if (ret) + break; + } + + dev_info(adev->dev, "XGMI: Add node %d, hive 0x%llx.\n", + adev->gmc.xgmi.physical_node_id, adev->gmc.xgmi.hive_id); + + mutex_unlock(&hive->hive_lock); exit: - mutex_unlock(&xgmi_mutex); return ret; } +void amdgpu_xgmi_remove_device(struct amdgpu_device *adev) +{ + struct amdgpu_hive_info *hive; + + if (!adev->gmc.xgmi.supported) + return; + + hive = amdgpu_get_xgmi_hive(adev, 1); + if (!hive) + return; + if (!(hive->number_devices--)) { + mutex_destroy(&hive->hive_lock); + mutex_destroy(&hive->reset_lock); + } else { + mutex_unlock(&hive->hive_lock); + } +} diff --git a/drivers/gpu/drm/amd/display/dc/i2caux/engine_base.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.h index 5d155d36d353..14bc60664159 100644 --- a/drivers/gpu/drm/amd/display/dc/i2caux/engine_base.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.h @@ -1,5 +1,5 @@ /* - * Copyright 2012-15 Advanced Micro Devices, Inc. + * Copyright 2016 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"), @@ -18,35 +18,24 @@ * 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: AMD - * - */ - -#include "dm_services.h" - -/* - * Pre-requisites: headers required by header of this unit - */ -#include "include/i2caux_interface.h" - -/* - * Header of this unit */ +#ifndef __AMDGPU_XGMI_H__ +#define __AMDGPU_XGMI_H__ -#include "engine.h" +#include "amdgpu_psp.h" -void dal_i2caux_construct_engine( - struct engine *engine, - struct dc_context *ctx) -{ - engine->ddc = NULL; - engine->ctx = ctx; -} +struct amdgpu_hive_info { + uint64_t hive_id; + struct list_head device_list; + struct psp_xgmi_topology_info topology_info; + int number_devices; + struct mutex hive_lock, + reset_lock; +}; -void dal_i2caux_destruct_engine( - struct engine *engine) -{ - /* nothing to do */ -} +struct amdgpu_hive_info *amdgpu_get_xgmi_hive(struct amdgpu_device *adev, int lock); +int amdgpu_xgmi_update_topology(struct amdgpu_hive_info *hive, struct amdgpu_device *adev); +int amdgpu_xgmi_add_device(struct amdgpu_device *adev); +void amdgpu_xgmi_remove_device(struct amdgpu_device *adev); +#endif diff --git a/drivers/gpu/drm/amd/amdgpu/atom.c b/drivers/gpu/drm/amd/amdgpu/atom.c index e9934de1b9cf..dd30f4e61a8c 100644 --- a/drivers/gpu/drm/amd/amdgpu/atom.c +++ b/drivers/gpu/drm/amd/amdgpu/atom.c @@ -27,6 +27,8 @@ #include <linux/slab.h> #include <asm/unaligned.h> +#include <drm/drm_util.h> + #define ATOM_DEBUG #include "atom.h" diff --git a/drivers/gpu/drm/amd/amdgpu/ci_dpm.c b/drivers/gpu/drm/amd/amdgpu/ci_dpm.c deleted file mode 100644 index 79220a91abe3..000000000000 --- a/drivers/gpu/drm/amd/amdgpu/ci_dpm.c +++ /dev/null @@ -1,6844 +0,0 @@ -/* - * Copyright 2013 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 <drm/drmP.h> -#include "amdgpu.h" -#include "amdgpu_pm.h" -#include "amdgpu_ucode.h" -#include "cikd.h" -#include "amdgpu_dpm.h" -#include "ci_dpm.h" -#include "gfx_v7_0.h" -#include "atom.h" -#include "amd_pcie.h" -#include <linux/seq_file.h> - -#include "smu/smu_7_0_1_d.h" -#include "smu/smu_7_0_1_sh_mask.h" - -#include "dce/dce_8_0_d.h" -#include "dce/dce_8_0_sh_mask.h" - -#include "bif/bif_4_1_d.h" -#include "bif/bif_4_1_sh_mask.h" - -#include "gca/gfx_7_2_d.h" -#include "gca/gfx_7_2_sh_mask.h" - -#include "gmc/gmc_7_1_d.h" -#include "gmc/gmc_7_1_sh_mask.h" - -MODULE_FIRMWARE("amdgpu/bonaire_smc.bin"); -MODULE_FIRMWARE("amdgpu/bonaire_k_smc.bin"); -MODULE_FIRMWARE("amdgpu/hawaii_smc.bin"); -MODULE_FIRMWARE("amdgpu/hawaii_k_smc.bin"); - -#define MC_CG_ARB_FREQ_F0 0x0a -#define MC_CG_ARB_FREQ_F1 0x0b -#define MC_CG_ARB_FREQ_F2 0x0c -#define MC_CG_ARB_FREQ_F3 0x0d - -#define SMC_RAM_END 0x40000 - -#define VOLTAGE_SCALE 4 -#define VOLTAGE_VID_OFFSET_SCALE1 625 -#define VOLTAGE_VID_OFFSET_SCALE2 100 - -static const struct amd_pm_funcs ci_dpm_funcs; - -static const struct ci_pt_defaults defaults_hawaii_xt = -{ - 1, 0xF, 0xFD, 0x19, 5, 0x14, 0, 0xB0000, - { 0x2E, 0x00, 0x00, 0x88, 0x00, 0x00, 0x72, 0x60, 0x51, 0xA7, 0x79, 0x6B, 0x90, 0xBD, 0x79 }, - { 0x217, 0x217, 0x217, 0x242, 0x242, 0x242, 0x269, 0x269, 0x269, 0x2A1, 0x2A1, 0x2A1, 0x2C9, 0x2C9, 0x2C9 } -}; - -static const struct ci_pt_defaults defaults_hawaii_pro = -{ - 1, 0xF, 0xFD, 0x19, 5, 0x14, 0, 0x65062, - { 0x2E, 0x00, 0x00, 0x88, 0x00, 0x00, 0x72, 0x60, 0x51, 0xA7, 0x79, 0x6B, 0x90, 0xBD, 0x79 }, - { 0x217, 0x217, 0x217, 0x242, 0x242, 0x242, 0x269, 0x269, 0x269, 0x2A1, 0x2A1, 0x2A1, 0x2C9, 0x2C9, 0x2C9 } -}; - -static const struct ci_pt_defaults defaults_bonaire_xt = -{ - 1, 0xF, 0xFD, 0x19, 5, 45, 0, 0xB0000, - { 0x79, 0x253, 0x25D, 0xAE, 0x72, 0x80, 0x83, 0x86, 0x6F, 0xC8, 0xC9, 0xC9, 0x2F, 0x4D, 0x61 }, - { 0x17C, 0x172, 0x180, 0x1BC, 0x1B3, 0x1BD, 0x206, 0x200, 0x203, 0x25D, 0x25A, 0x255, 0x2C3, 0x2C5, 0x2B4 } -}; - -#if 0 -static const struct ci_pt_defaults defaults_bonaire_pro = -{ - 1, 0xF, 0xFD, 0x19, 5, 45, 0, 0x65062, - { 0x8C, 0x23F, 0x244, 0xA6, 0x83, 0x85, 0x86, 0x86, 0x83, 0xDB, 0xDB, 0xDA, 0x67, 0x60, 0x5F }, - { 0x187, 0x193, 0x193, 0x1C7, 0x1D1, 0x1D1, 0x210, 0x219, 0x219, 0x266, 0x26C, 0x26C, 0x2C9, 0x2CB, 0x2CB } -}; -#endif - -static const struct ci_pt_defaults defaults_saturn_xt = -{ - 1, 0xF, 0xFD, 0x19, 5, 55, 0, 0x70000, - { 0x8C, 0x247, 0x249, 0xA6, 0x80, 0x81, 0x8B, 0x89, 0x86, 0xC9, 0xCA, 0xC9, 0x4D, 0x4D, 0x4D }, - { 0x187, 0x187, 0x187, 0x1C7, 0x1C7, 0x1C7, 0x210, 0x210, 0x210, 0x266, 0x266, 0x266, 0x2C9, 0x2C9, 0x2C9 } -}; - -#if 0 -static const struct ci_pt_defaults defaults_saturn_pro = -{ - 1, 0xF, 0xFD, 0x19, 5, 55, 0, 0x30000, - { 0x96, 0x21D, 0x23B, 0xA1, 0x85, 0x87, 0x83, 0x84, 0x81, 0xE6, 0xE6, 0xE6, 0x71, 0x6A, 0x6A }, - { 0x193, 0x19E, 0x19E, 0x1D2, 0x1DC, 0x1DC, 0x21A, 0x223, 0x223, 0x26E, 0x27E, 0x274, 0x2CF, 0x2D2, 0x2D2 } -}; -#endif - -static const struct ci_pt_config_reg didt_config_ci[] = -{ - { 0x10, 0x000000ff, 0, 0x0, CISLANDS_CONFIGREG_DIDT_IND }, - { 0x10, 0x0000ff00, 8, 0x0, CISLANDS_CONFIGREG_DIDT_IND }, - { 0x10, 0x00ff0000, 16, 0x0, CISLANDS_CONFIGREG_DIDT_IND }, - { 0x10, 0xff000000, 24, 0x0, CISLANDS_CONFIGREG_DIDT_IND }, - { 0x11, 0x000000ff, 0, 0x0, CISLANDS_CONFIGREG_DIDT_IND }, - { 0x11, 0x0000ff00, 8, 0x0, CISLANDS_CONFIGREG_DIDT_IND }, - { 0x11, 0x00ff0000, 16, 0x0, CISLANDS_CONFIGREG_DIDT_IND }, - { 0x11, 0xff000000, 24, 0x0, CISLANDS_CONFIGREG_DIDT_IND }, - { 0x12, 0x000000ff, 0, 0x0, CISLANDS_CONFIGREG_DIDT_IND }, - { 0x12, 0x0000ff00, 8, 0x0, CISLANDS_CONFIGREG_DIDT_IND }, - { 0x12, 0x00ff0000, 16, 0x0, CISLANDS_CONFIGREG_DIDT_IND }, - { 0x12, 0xff000000, 24, 0x0, CISLANDS_CONFIGREG_DIDT_IND }, - { 0x2, 0x00003fff, 0, 0x4, CISLANDS_CONFIGREG_DIDT_IND }, - { 0x2, 0x03ff0000, 16, 0x80, CISLANDS_CONFIGREG_DIDT_IND }, - { 0x2, 0x78000000, 27, 0x3, CISLANDS_CONFIGREG_DIDT_IND }, - { 0x1, 0x0000ffff, 0, 0x3FFF, CISLANDS_CONFIGREG_DIDT_IND }, - { 0x1, 0xffff0000, 16, 0x3FFF, CISLANDS_CONFIGREG_DIDT_IND }, - { 0x0, 0x00000001, 0, 0x0, CISLANDS_CONFIGREG_DIDT_IND }, - { 0x30, 0x000000ff, 0, 0x0, CISLANDS_CONFIGREG_DIDT_IND }, - { 0x30, 0x0000ff00, 8, 0x0, CISLANDS_CONFIGREG_DIDT_IND }, - { 0x30, 0x00ff0000, 16, 0x0, CISLANDS_CONFIGREG_DIDT_IND }, - { 0x30, 0xff000000, 24, 0x0, CISLANDS_CONFIGREG_DIDT_IND }, - { 0x31, 0x000000ff, 0, 0x0, CISLANDS_CONFIGREG_DIDT_IND }, - { 0x31, 0x0000ff00, 8, 0x0, CISLANDS_CONFIGREG_DIDT_IND }, - { 0x31, 0x00ff0000, 16, 0x0, CISLANDS_CONFIGREG_DIDT_IND }, - { 0x31, 0xff000000, 24, 0x0, CISLANDS_CONFIGREG_DIDT_IND }, - { 0x32, 0x000000ff, 0, 0x0, CISLANDS_CONFIGREG_DIDT_IND }, - { 0x32, 0x0000ff00, 8, 0x0, CISLANDS_CONFIGREG_DIDT_IND }, - { 0x32, 0x00ff0000, 16, 0x0, CISLANDS_CONFIGREG_DIDT_IND }, - { 0x32, 0xff000000, 24, 0x0, CISLANDS_CONFIGREG_DIDT_IND }, - { 0x22, 0x00003fff, 0, 0x4, CISLANDS_CONFIGREG_DIDT_IND }, - { 0x22, 0x03ff0000, 16, 0x80, CISLANDS_CONFIGREG_DIDT_IND }, - { 0x22, 0x78000000, 27, 0x3, CISLANDS_CONFIGREG_DIDT_IND }, - { 0x21, 0x0000ffff, 0, 0x3FFF, CISLANDS_CONFIGREG_DIDT_IND }, - { 0x21, 0xffff0000, 16, 0x3FFF, CISLANDS_CONFIGREG_DIDT_IND }, - { 0x20, 0x00000001, 0, 0x0, CISLANDS_CONFIGREG_DIDT_IND }, - { 0x50, 0x000000ff, 0, 0x0, CISLANDS_CONFIGREG_DIDT_IND }, - { 0x50, 0x0000ff00, 8, 0x0, CISLANDS_CONFIGREG_DIDT_IND }, - { 0x50, 0x00ff0000, 16, 0x0, CISLANDS_CONFIGREG_DIDT_IND }, - { 0x50, 0xff000000, 24, 0x0, CISLANDS_CONFIGREG_DIDT_IND }, - { 0x51, 0x000000ff, 0, 0x0, CISLANDS_CONFIGREG_DIDT_IND }, - { 0x51, 0x0000ff00, 8, 0x0, CISLANDS_CONFIGREG_DIDT_IND }, - { 0x51, 0x00ff0000, 16, 0x0, CISLANDS_CONFIGREG_DIDT_IND }, - { 0x51, 0xff000000, 24, 0x0, CISLANDS_CONFIGREG_DIDT_IND }, - { 0x52, 0x000000ff, 0, 0x0, CISLANDS_CONFIGREG_DIDT_IND }, - { 0x52, 0x0000ff00, 8, 0x0, CISLANDS_CONFIGREG_DIDT_IND }, - { 0x52, 0x00ff0000, 16, 0x0, CISLANDS_CONFIGREG_DIDT_IND }, - { 0x52, 0xff000000, 24, 0x0, CISLANDS_CONFIGREG_DIDT_IND }, - { 0x42, 0x00003fff, 0, 0x4, CISLANDS_CONFIGREG_DIDT_IND }, - { 0x42, 0x03ff0000, 16, 0x80, CISLANDS_CONFIGREG_DIDT_IND }, - { 0x42, 0x78000000, 27, 0x3, CISLANDS_CONFIGREG_DIDT_IND }, - { 0x41, 0x0000ffff, 0, 0x3FFF, CISLANDS_CONFIGREG_DIDT_IND }, - { 0x41, 0xffff0000, 16, 0x3FFF, CISLANDS_CONFIGREG_DIDT_IND }, - { 0x40, 0x00000001, 0, 0x0, CISLANDS_CONFIGREG_DIDT_IND }, - { 0x70, 0x000000ff, 0, 0x0, CISLANDS_CONFIGREG_DIDT_IND }, - { 0x70, 0x0000ff00, 8, 0x0, CISLANDS_CONFIGREG_DIDT_IND }, - { 0x70, 0x00ff0000, 16, 0x0, CISLANDS_CONFIGREG_DIDT_IND }, - { 0x70, 0xff000000, 24, 0x0, CISLANDS_CONFIGREG_DIDT_IND }, - { 0x71, 0x000000ff, 0, 0x0, CISLANDS_CONFIGREG_DIDT_IND }, - { 0x71, 0x0000ff00, 8, 0x0, CISLANDS_CONFIGREG_DIDT_IND }, - { 0x71, 0x00ff0000, 16, 0x0, CISLANDS_CONFIGREG_DIDT_IND }, - { 0x71, 0xff000000, 24, 0x0, CISLANDS_CONFIGREG_DIDT_IND }, - { 0x72, 0x000000ff, 0, 0x0, CISLANDS_CONFIGREG_DIDT_IND }, - { 0x72, 0x0000ff00, 8, 0x0, CISLANDS_CONFIGREG_DIDT_IND }, - { 0x72, 0x00ff0000, 16, 0x0, CISLANDS_CONFIGREG_DIDT_IND }, - { 0x72, 0xff000000, 24, 0x0, CISLANDS_CONFIGREG_DIDT_IND }, - { 0x62, 0x00003fff, 0, 0x4, CISLANDS_CONFIGREG_DIDT_IND }, - { 0x62, 0x03ff0000, 16, 0x80, CISLANDS_CONFIGREG_DIDT_IND }, - { 0x62, 0x78000000, 27, 0x3, CISLANDS_CONFIGREG_DIDT_IND }, - { 0x61, 0x0000ffff, 0, 0x3FFF, CISLANDS_CONFIGREG_DIDT_IND }, - { 0x61, 0xffff0000, 16, 0x3FFF, CISLANDS_CONFIGREG_DIDT_IND }, - { 0x60, 0x00000001, 0, 0x0, CISLANDS_CONFIGREG_DIDT_IND }, - { 0xFFFFFFFF } -}; - -static u8 ci_get_memory_module_index(struct amdgpu_device *adev) -{ - return (u8) ((RREG32(mmBIOS_SCRATCH_4) >> 16) & 0xff); -} - -#define MC_CG_ARB_FREQ_F0 0x0a -#define MC_CG_ARB_FREQ_F1 0x0b -#define MC_CG_ARB_FREQ_F2 0x0c -#define MC_CG_ARB_FREQ_F3 0x0d - -static int ci_copy_and_switch_arb_sets(struct amdgpu_device *adev, - u32 arb_freq_src, u32 arb_freq_dest) -{ - u32 mc_arb_dram_timing; - u32 mc_arb_dram_timing2; - u32 burst_time; - u32 mc_cg_config; - - switch (arb_freq_src) { - case MC_CG_ARB_FREQ_F0: - mc_arb_dram_timing = RREG32(mmMC_ARB_DRAM_TIMING); - mc_arb_dram_timing2 = RREG32(mmMC_ARB_DRAM_TIMING2); - burst_time = (RREG32(mmMC_ARB_BURST_TIME) & MC_ARB_BURST_TIME__STATE0_MASK) >> - MC_ARB_BURST_TIME__STATE0__SHIFT; - break; - case MC_CG_ARB_FREQ_F1: - mc_arb_dram_timing = RREG32(mmMC_ARB_DRAM_TIMING_1); - mc_arb_dram_timing2 = RREG32(mmMC_ARB_DRAM_TIMING2_1); - burst_time = (RREG32(mmMC_ARB_BURST_TIME) & MC_ARB_BURST_TIME__STATE1_MASK) >> - MC_ARB_BURST_TIME__STATE1__SHIFT; - break; - default: - return -EINVAL; - } - - switch (arb_freq_dest) { - case MC_CG_ARB_FREQ_F0: - WREG32(mmMC_ARB_DRAM_TIMING, mc_arb_dram_timing); - WREG32(mmMC_ARB_DRAM_TIMING2, mc_arb_dram_timing2); - WREG32_P(mmMC_ARB_BURST_TIME, (burst_time << MC_ARB_BURST_TIME__STATE0__SHIFT), - ~MC_ARB_BURST_TIME__STATE0_MASK); - break; - case MC_CG_ARB_FREQ_F1: - WREG32(mmMC_ARB_DRAM_TIMING_1, mc_arb_dram_timing); - WREG32(mmMC_ARB_DRAM_TIMING2_1, mc_arb_dram_timing2); - WREG32_P(mmMC_ARB_BURST_TIME, (burst_time << MC_ARB_BURST_TIME__STATE1__SHIFT), - ~MC_ARB_BURST_TIME__STATE1_MASK); - break; - default: - return -EINVAL; - } - - mc_cg_config = RREG32(mmMC_CG_CONFIG) | 0x0000000F; - WREG32(mmMC_CG_CONFIG, mc_cg_config); - WREG32_P(mmMC_ARB_CG, (arb_freq_dest) << MC_ARB_CG__CG_ARB_REQ__SHIFT, - ~MC_ARB_CG__CG_ARB_REQ_MASK); - - return 0; -} - -static u8 ci_get_ddr3_mclk_frequency_ratio(u32 memory_clock) -{ - u8 mc_para_index; - - if (memory_clock < 10000) - mc_para_index = 0; - else if (memory_clock >= 80000) - mc_para_index = 0x0f; - else - mc_para_index = (u8)((memory_clock - 10000) / 5000 + 1); - return mc_para_index; -} - -static u8 ci_get_mclk_frequency_ratio(u32 memory_clock, bool strobe_mode) -{ - u8 mc_para_index; - - if (strobe_mode) { - if (memory_clock < 12500) - mc_para_index = 0x00; - else if (memory_clock > 47500) - mc_para_index = 0x0f; - else - mc_para_index = (u8)((memory_clock - 10000) / 2500); - } else { - if (memory_clock < 65000) - mc_para_index = 0x00; - else if (memory_clock > 135000) - mc_para_index = 0x0f; - else - mc_para_index = (u8)((memory_clock - 60000) / 5000); - } - return mc_para_index; -} - -static void ci_trim_voltage_table_to_fit_state_table(struct amdgpu_device *adev, - u32 max_voltage_steps, - struct atom_voltage_table *voltage_table) -{ - unsigned int i, diff; - - if (voltage_table->count <= max_voltage_steps) - return; - - diff = voltage_table->count - max_voltage_steps; - - for (i = 0; i < max_voltage_steps; i++) - voltage_table->entries[i] = voltage_table->entries[i + diff]; - - voltage_table->count = max_voltage_steps; -} - -static int ci_get_std_voltage_value_sidd(struct amdgpu_device *adev, - struct atom_voltage_table_entry *voltage_table, - u16 *std_voltage_hi_sidd, u16 *std_voltage_lo_sidd); -static int ci_set_power_limit(struct amdgpu_device *adev, u32 n); -static int ci_set_overdrive_target_tdp(struct amdgpu_device *adev, - u32 target_tdp); -static int ci_update_uvd_dpm(struct amdgpu_device *adev, bool gate); -static void ci_dpm_set_irq_funcs(struct amdgpu_device *adev); - -static PPSMC_Result amdgpu_ci_send_msg_to_smc_with_parameter(struct amdgpu_device *adev, - PPSMC_Msg msg, u32 parameter); -static void ci_thermal_start_smc_fan_control(struct amdgpu_device *adev); -static void ci_fan_ctrl_set_default_mode(struct amdgpu_device *adev); - -static struct ci_power_info *ci_get_pi(struct amdgpu_device *adev) -{ - struct ci_power_info *pi = adev->pm.dpm.priv; - - return pi; -} - -static struct ci_ps *ci_get_ps(struct amdgpu_ps *rps) -{ - struct ci_ps *ps = rps->ps_priv; - - return ps; -} - -static void ci_initialize_powertune_defaults(struct amdgpu_device *adev) -{ - struct ci_power_info *pi = ci_get_pi(adev); - - switch (adev->pdev->device) { - case 0x6649: - case 0x6650: - case 0x6651: - case 0x6658: - case 0x665C: - case 0x665D: - default: - pi->powertune_defaults = &defaults_bonaire_xt; - break; - case 0x6640: - case 0x6641: - case 0x6646: - case 0x6647: - pi->powertune_defaults = &defaults_saturn_xt; - break; - case 0x67B8: - case 0x67B0: - pi->powertune_defaults = &defaults_hawaii_xt; - break; - case 0x67BA: - case 0x67B1: - pi->powertune_defaults = &defaults_hawaii_pro; - break; - case 0x67A0: - case 0x67A1: - case 0x67A2: - case 0x67A8: - case 0x67A9: - case 0x67AA: - case 0x67B9: - case 0x67BE: - pi->powertune_defaults = &defaults_bonaire_xt; - break; - } - - pi->dte_tj_offset = 0; - - pi->caps_power_containment = true; - pi->caps_cac = false; - pi->caps_sq_ramping = false; - pi->caps_db_ramping = false; - pi->caps_td_ramping = false; - pi->caps_tcp_ramping = false; - - if (pi->caps_power_containment) { - pi->caps_cac = true; - if (adev->asic_type == CHIP_HAWAII) - pi->enable_bapm_feature = false; - else - pi->enable_bapm_feature = true; - pi->enable_tdc_limit_feature = true; - pi->enable_pkg_pwr_tracking_feature = true; - } -} - -static u8 ci_convert_to_vid(u16 vddc) -{ - return (6200 - (vddc * VOLTAGE_SCALE)) / 25; -} - -static int ci_populate_bapm_vddc_vid_sidd(struct amdgpu_device *adev) -{ - struct ci_power_info *pi = ci_get_pi(adev); - u8 *hi_vid = pi->smc_powertune_table.BapmVddCVidHiSidd; - u8 *lo_vid = pi->smc_powertune_table.BapmVddCVidLoSidd; - u8 *hi2_vid = pi->smc_powertune_table.BapmVddCVidHiSidd2; - u32 i; - - if (adev->pm.dpm.dyn_state.cac_leakage_table.entries == NULL) - return -EINVAL; - if (adev->pm.dpm.dyn_state.cac_leakage_table.count > 8) - return -EINVAL; - if (adev->pm.dpm.dyn_state.cac_leakage_table.count != - adev->pm.dpm.dyn_state.vddc_dependency_on_sclk.count) - return -EINVAL; - - for (i = 0; i < adev->pm.dpm.dyn_state.cac_leakage_table.count; i++) { - if (adev->pm.dpm.platform_caps & ATOM_PP_PLATFORM_CAP_EVV) { - lo_vid[i] = ci_convert_to_vid(adev->pm.dpm.dyn_state.cac_leakage_table.entries[i].vddc1); - hi_vid[i] = ci_convert_to_vid(adev->pm.dpm.dyn_state.cac_leakage_table.entries[i].vddc2); - hi2_vid[i] = ci_convert_to_vid(adev->pm.dpm.dyn_state.cac_leakage_table.entries[i].vddc3); - } else { - lo_vid[i] = ci_convert_to_vid(adev->pm.dpm.dyn_state.cac_leakage_table.entries[i].vddc); - hi_vid[i] = ci_convert_to_vid((u16)adev->pm.dpm.dyn_state.cac_leakage_table.entries[i].leakage); - } - } - return 0; -} - -static int ci_populate_vddc_vid(struct amdgpu_device *adev) -{ - struct ci_power_info *pi = ci_get_pi(adev); - u8 *vid = pi->smc_powertune_table.VddCVid; - u32 i; - - if (pi->vddc_voltage_table.count > 8) - return -EINVAL; - - for (i = 0; i < pi->vddc_voltage_table.count; i++) - vid[i] = ci_convert_to_vid(pi->vddc_voltage_table.entries[i].value); - - return 0; -} - -static int ci_populate_svi_load_line(struct amdgpu_device *adev) -{ - struct ci_power_info *pi = ci_get_pi(adev); - const struct ci_pt_defaults *pt_defaults = pi->powertune_defaults; - - pi->smc_powertune_table.SviLoadLineEn = pt_defaults->svi_load_line_en; - pi->smc_powertune_table.SviLoadLineVddC = pt_defaults->svi_load_line_vddc; - pi->smc_powertune_table.SviLoadLineTrimVddC = 3; - pi->smc_powertune_table.SviLoadLineOffsetVddC = 0; - - return 0; -} - -static int ci_populate_tdc_limit(struct amdgpu_device *adev) -{ - struct ci_power_info *pi = ci_get_pi(adev); - const struct ci_pt_defaults *pt_defaults = pi->powertune_defaults; - u16 tdc_limit; - - tdc_limit = adev->pm.dpm.dyn_state.cac_tdp_table->tdc * 256; - pi->smc_powertune_table.TDC_VDDC_PkgLimit = cpu_to_be16(tdc_limit); - pi->smc_powertune_table.TDC_VDDC_ThrottleReleaseLimitPerc = - pt_defaults->tdc_vddc_throttle_release_limit_perc; - pi->smc_powertune_table.TDC_MAWt = pt_defaults->tdc_mawt; - - return 0; -} - -static int ci_populate_dw8(struct amdgpu_device *adev) -{ - struct ci_power_info *pi = ci_get_pi(adev); - const struct ci_pt_defaults *pt_defaults = pi->powertune_defaults; - int ret; - - ret = amdgpu_ci_read_smc_sram_dword(adev, - SMU7_FIRMWARE_HEADER_LOCATION + - offsetof(SMU7_Firmware_Header, PmFuseTable) + - offsetof(SMU7_Discrete_PmFuses, TdcWaterfallCtl), - (u32 *)&pi->smc_powertune_table.TdcWaterfallCtl, - pi->sram_end); - if (ret) - return -EINVAL; - else - pi->smc_powertune_table.TdcWaterfallCtl = pt_defaults->tdc_waterfall_ctl; - - return 0; -} - -static int ci_populate_fuzzy_fan(struct amdgpu_device *adev) -{ - struct ci_power_info *pi = ci_get_pi(adev); - - if ((adev->pm.dpm.fan.fan_output_sensitivity & (1 << 15)) || - (adev->pm.dpm.fan.fan_output_sensitivity == 0)) - adev->pm.dpm.fan.fan_output_sensitivity = - adev->pm.dpm.fan.default_fan_output_sensitivity; - - pi->smc_powertune_table.FuzzyFan_PwmSetDelta = - cpu_to_be16(adev->pm.dpm.fan.fan_output_sensitivity); - - return 0; -} - -static int ci_min_max_v_gnbl_pm_lid_from_bapm_vddc(struct amdgpu_device *adev) -{ - struct ci_power_info *pi = ci_get_pi(adev); - u8 *hi_vid = pi->smc_powertune_table.BapmVddCVidHiSidd; - u8 *lo_vid = pi->smc_powertune_table.BapmVddCVidLoSidd; - int i, min, max; - - min = max = hi_vid[0]; - for (i = 0; i < 8; i++) { - if (0 != hi_vid[i]) { - if (min > hi_vid[i]) - min = hi_vid[i]; - if (max < hi_vid[i]) - max = hi_vid[i]; - } - - if (0 != lo_vid[i]) { - if (min > lo_vid[i]) - min = lo_vid[i]; - if (max < lo_vid[i]) - max = lo_vid[i]; - } - } - - if ((min == 0) || (max == 0)) - return -EINVAL; - pi->smc_powertune_table.GnbLPMLMaxVid = (u8)max; - pi->smc_powertune_table.GnbLPMLMinVid = (u8)min; - - return 0; -} - -static int ci_populate_bapm_vddc_base_leakage_sidd(struct amdgpu_device *adev) -{ - struct ci_power_info *pi = ci_get_pi(adev); - u16 hi_sidd = pi->smc_powertune_table.BapmVddCBaseLeakageHiSidd; - u16 lo_sidd = pi->smc_powertune_table.BapmVddCBaseLeakageLoSidd; - struct amdgpu_cac_tdp_table *cac_tdp_table = - adev->pm.dpm.dyn_state.cac_tdp_table; - - hi_sidd = cac_tdp_table->high_cac_leakage / 100 * 256; - lo_sidd = cac_tdp_table->low_cac_leakage / 100 * 256; - - pi->smc_powertune_table.BapmVddCBaseLeakageHiSidd = cpu_to_be16(hi_sidd); - pi->smc_powertune_table.BapmVddCBaseLeakageLoSidd = cpu_to_be16(lo_sidd); - - return 0; -} - -static int ci_populate_bapm_parameters_in_dpm_table(struct amdgpu_device *adev) -{ - struct ci_power_info *pi = ci_get_pi(adev); - const struct ci_pt_defaults *pt_defaults = pi->powertune_defaults; - SMU7_Discrete_DpmTable *dpm_table = &pi->smc_state_table; - struct amdgpu_cac_tdp_table *cac_tdp_table = - adev->pm.dpm.dyn_state.cac_tdp_table; - struct amdgpu_ppm_table *ppm = adev->pm.dpm.dyn_state.ppm_table; - int i, j, k; - const u16 *def1; - const u16 *def2; - - dpm_table->DefaultTdp = cac_tdp_table->tdp * 256; - dpm_table->TargetTdp = cac_tdp_table->configurable_tdp * 256; - - dpm_table->DTETjOffset = (u8)pi->dte_tj_offset; - dpm_table->GpuTjMax = - (u8)(pi->thermal_temp_setting.temperature_high / 1000); - dpm_table->GpuTjHyst = 8; - - dpm_table->DTEAmbientTempBase = pt_defaults->dte_ambient_temp_base; - - if (ppm) { - dpm_table->PPM_PkgPwrLimit = cpu_to_be16((u16)ppm->dgpu_tdp * 256 / 1000); - dpm_table->PPM_TemperatureLimit = cpu_to_be16((u16)ppm->tj_max * 256); - } else { - dpm_table->PPM_PkgPwrLimit = cpu_to_be16(0); - dpm_table->PPM_TemperatureLimit = cpu_to_be16(0); - } - - dpm_table->BAPM_TEMP_GRADIENT = cpu_to_be32(pt_defaults->bapm_temp_gradient); - def1 = pt_defaults->bapmti_r; - def2 = pt_defaults->bapmti_rc; - - for (i = 0; i < SMU7_DTE_ITERATIONS; i++) { - for (j = 0; j < SMU7_DTE_SOURCES; j++) { - for (k = 0; k < SMU7_DTE_SINKS; k++) { - dpm_table->BAPMTI_R[i][j][k] = cpu_to_be16(*def1); - dpm_table->BAPMTI_RC[i][j][k] = cpu_to_be16(*def2); - def1++; - def2++; - } - } - } - - return 0; -} - -static int ci_populate_pm_base(struct amdgpu_device *adev) -{ - struct ci_power_info *pi = ci_get_pi(adev); - u32 pm_fuse_table_offset; - int ret; - - if (pi->caps_power_containment) { - ret = amdgpu_ci_read_smc_sram_dword(adev, - SMU7_FIRMWARE_HEADER_LOCATION + - offsetof(SMU7_Firmware_Header, PmFuseTable), - &pm_fuse_table_offset, pi->sram_end); - if (ret) - return ret; - ret = ci_populate_bapm_vddc_vid_sidd(adev); - if (ret) - return ret; - ret = ci_populate_vddc_vid(adev); - if (ret) - return ret; - ret = ci_populate_svi_load_line(adev); - if (ret) - return ret; - ret = ci_populate_tdc_limit(adev); - if (ret) - return ret; - ret = ci_populate_dw8(adev); - if (ret) - return ret; - ret = ci_populate_fuzzy_fan(adev); - if (ret) - return ret; - ret = ci_min_max_v_gnbl_pm_lid_from_bapm_vddc(adev); - if (ret) - return ret; - ret = ci_populate_bapm_vddc_base_leakage_sidd(adev); - if (ret) - return ret; - ret = amdgpu_ci_copy_bytes_to_smc(adev, pm_fuse_table_offset, - (u8 *)&pi->smc_powertune_table, - sizeof(SMU7_Discrete_PmFuses), pi->sram_end); - if (ret) - return ret; - } - - return 0; -} - -static void ci_do_enable_didt(struct amdgpu_device *adev, const bool enable) -{ - struct ci_power_info *pi = ci_get_pi(adev); - u32 data; - - if (pi->caps_sq_ramping) { - data = RREG32_DIDT(ixDIDT_SQ_CTRL0); - if (enable) - data |= DIDT_SQ_CTRL0__DIDT_CTRL_EN_MASK; - else - data &= ~DIDT_SQ_CTRL0__DIDT_CTRL_EN_MASK; - WREG32_DIDT(ixDIDT_SQ_CTRL0, data); - } - - if (pi->caps_db_ramping) { - data = RREG32_DIDT(ixDIDT_DB_CTRL0); - if (enable) - data |= DIDT_DB_CTRL0__DIDT_CTRL_EN_MASK; - else - data &= ~DIDT_DB_CTRL0__DIDT_CTRL_EN_MASK; - WREG32_DIDT(ixDIDT_DB_CTRL0, data); - } - - if (pi->caps_td_ramping) { - data = RREG32_DIDT(ixDIDT_TD_CTRL0); - if (enable) - data |= DIDT_TD_CTRL0__DIDT_CTRL_EN_MASK; - else - data &= ~DIDT_TD_CTRL0__DIDT_CTRL_EN_MASK; - WREG32_DIDT(ixDIDT_TD_CTRL0, data); - } - - if (pi->caps_tcp_ramping) { - data = RREG32_DIDT(ixDIDT_TCP_CTRL0); - if (enable) - data |= DIDT_TCP_CTRL0__DIDT_CTRL_EN_MASK; - else - data &= ~DIDT_TCP_CTRL0__DIDT_CTRL_EN_MASK; - WREG32_DIDT(ixDIDT_TCP_CTRL0, data); - } -} - -static int ci_program_pt_config_registers(struct amdgpu_device *adev, - const struct ci_pt_config_reg *cac_config_regs) -{ - const struct ci_pt_config_reg *config_regs = cac_config_regs; - u32 data; - u32 cache = 0; - - if (config_regs == NULL) - return -EINVAL; - - while (config_regs->offset != 0xFFFFFFFF) { - if (config_regs->type == CISLANDS_CONFIGREG_CACHE) { - cache |= ((config_regs->value << config_regs->shift) & config_regs->mask); - } else { - switch (config_regs->type) { - case CISLANDS_CONFIGREG_SMC_IND: - data = RREG32_SMC(config_regs->offset); - break; - case CISLANDS_CONFIGREG_DIDT_IND: - data = RREG32_DIDT(config_regs->offset); - break; - default: - data = RREG32(config_regs->offset); - break; - } - - data &= ~config_regs->mask; - data |= ((config_regs->value << config_regs->shift) & config_regs->mask); - data |= cache; - - switch (config_regs->type) { - case CISLANDS_CONFIGREG_SMC_IND: - WREG32_SMC(config_regs->offset, data); - break; - case CISLANDS_CONFIGREG_DIDT_IND: - WREG32_DIDT(config_regs->offset, data); - break; - default: - WREG32(config_regs->offset, data); - break; - } - cache = 0; - } - config_regs++; - } - return 0; -} - -static int ci_enable_didt(struct amdgpu_device *adev, bool enable) -{ - struct ci_power_info *pi = ci_get_pi(adev); - int ret; - - if (pi->caps_sq_ramping || pi->caps_db_ramping || - pi->caps_td_ramping || pi->caps_tcp_ramping) { - adev->gfx.rlc.funcs->enter_safe_mode(adev); - - if (enable) { - ret = ci_program_pt_config_registers(adev, didt_config_ci); - if (ret) { - adev->gfx.rlc.funcs->exit_safe_mode(adev); - return ret; - } - } - - ci_do_enable_didt(adev, enable); - - adev->gfx.rlc.funcs->exit_safe_mode(adev); - } - - return 0; -} - -static int ci_enable_power_containment(struct amdgpu_device *adev, bool enable) -{ - struct ci_power_info *pi = ci_get_pi(adev); - PPSMC_Result smc_result; - int ret = 0; - - if (enable) { - pi->power_containment_features = 0; - if (pi->caps_power_containment) { - if (pi->enable_bapm_feature) { - smc_result = amdgpu_ci_send_msg_to_smc(adev, PPSMC_MSG_EnableDTE); - if (smc_result != PPSMC_Result_OK) - ret = -EINVAL; - else - pi->power_containment_features |= POWERCONTAINMENT_FEATURE_BAPM; - } - - if (pi->enable_tdc_limit_feature) { - smc_result = amdgpu_ci_send_msg_to_smc(adev, PPSMC_MSG_TDCLimitEnable); - if (smc_result != PPSMC_Result_OK) - ret = -EINVAL; - else - pi->power_containment_features |= POWERCONTAINMENT_FEATURE_TDCLimit; - } - - if (pi->enable_pkg_pwr_tracking_feature) { - smc_result = amdgpu_ci_send_msg_to_smc(adev, PPSMC_MSG_PkgPwrLimitEnable); - if (smc_result != PPSMC_Result_OK) { - ret = -EINVAL; - } else { - struct amdgpu_cac_tdp_table *cac_tdp_table = - adev->pm.dpm.dyn_state.cac_tdp_table; - u32 default_pwr_limit = - (u32)(cac_tdp_table->maximum_power_delivery_limit * 256); - - pi->power_containment_features |= POWERCONTAINMENT_FEATURE_PkgPwrLimit; - - ci_set_power_limit(adev, default_pwr_limit); - } - } - } - } else { - if (pi->caps_power_containment && pi->power_containment_features) { - if (pi->power_containment_features & POWERCONTAINMENT_FEATURE_TDCLimit) - amdgpu_ci_send_msg_to_smc(adev, PPSMC_MSG_TDCLimitDisable); - - if (pi->power_containment_features & POWERCONTAINMENT_FEATURE_BAPM) - amdgpu_ci_send_msg_to_smc(adev, PPSMC_MSG_DisableDTE); - - if (pi->power_containment_features & POWERCONTAINMENT_FEATURE_PkgPwrLimit) - amdgpu_ci_send_msg_to_smc(adev, PPSMC_MSG_PkgPwrLimitDisable); - pi->power_containment_features = 0; - } - } - - return ret; -} - -static int ci_enable_smc_cac(struct amdgpu_device *adev, bool enable) -{ - struct ci_power_info *pi = ci_get_pi(adev); - PPSMC_Result smc_result; - int ret = 0; - - if (pi->caps_cac) { - if (enable) { - smc_result = amdgpu_ci_send_msg_to_smc(adev, PPSMC_MSG_EnableCac); - if (smc_result != PPSMC_Result_OK) { - ret = -EINVAL; - pi->cac_enabled = false; - } else { - pi->cac_enabled = true; - } - } else if (pi->cac_enabled) { - amdgpu_ci_send_msg_to_smc(adev, PPSMC_MSG_DisableCac); - pi->cac_enabled = false; - } - } - - return ret; -} - -static int ci_enable_thermal_based_sclk_dpm(struct amdgpu_device *adev, - bool enable) -{ - struct ci_power_info *pi = ci_get_pi(adev); - PPSMC_Result smc_result = PPSMC_Result_OK; - - if (pi->thermal_sclk_dpm_enabled) { - if (enable) - smc_result = amdgpu_ci_send_msg_to_smc(adev, PPSMC_MSG_ENABLE_THERMAL_DPM); - else - smc_result = amdgpu_ci_send_msg_to_smc(adev, PPSMC_MSG_DISABLE_THERMAL_DPM); - } - - if (smc_result == PPSMC_Result_OK) - return 0; - else - return -EINVAL; -} - -static int ci_power_control_set_level(struct amdgpu_device *adev) -{ - struct ci_power_info *pi = ci_get_pi(adev); - struct amdgpu_cac_tdp_table *cac_tdp_table = - adev->pm.dpm.dyn_state.cac_tdp_table; - s32 adjust_percent; - s32 target_tdp; - int ret = 0; - bool adjust_polarity = false; /* ??? */ - - if (pi->caps_power_containment) { - adjust_percent = adjust_polarity ? - adev->pm.dpm.tdp_adjustment : (-1 * adev->pm.dpm.tdp_adjustment); - target_tdp = ((100 + adjust_percent) * - (s32)cac_tdp_table->configurable_tdp) / 100; - - ret = ci_set_overdrive_target_tdp(adev, (u32)target_tdp); - } - - return ret; -} - -static void ci_dpm_powergate_uvd(void *handle, bool gate) -{ - struct amdgpu_device *adev = (struct amdgpu_device *)handle; - struct ci_power_info *pi = ci_get_pi(adev); - - pi->uvd_power_gated = gate; - - if (gate) { - /* stop the UVD block */ - amdgpu_device_ip_set_powergating_state(adev, AMD_IP_BLOCK_TYPE_UVD, - AMD_PG_STATE_GATE); - ci_update_uvd_dpm(adev, gate); - } else { - amdgpu_device_ip_set_powergating_state(adev, AMD_IP_BLOCK_TYPE_UVD, - AMD_PG_STATE_UNGATE); - ci_update_uvd_dpm(adev, gate); - } -} - -static bool ci_dpm_vblank_too_short(void *handle) -{ - struct amdgpu_device *adev = (struct amdgpu_device *)handle; - u32 vblank_time = amdgpu_dpm_get_vblank_time(adev); - u32 switch_limit = adev->gmc.vram_type == AMDGPU_VRAM_TYPE_GDDR5 ? 450 : 300; - - /* disable mclk switching if the refresh is >120Hz, even if the - * blanking period would allow it - */ - if (amdgpu_dpm_get_vrefresh(adev) > 120) - return true; - - if (vblank_time < switch_limit) - return true; - else - return false; - -} - -static void ci_apply_state_adjust_rules(struct amdgpu_device *adev, - struct amdgpu_ps *rps) -{ - struct ci_ps *ps = ci_get_ps(rps); - struct ci_power_info *pi = ci_get_pi(adev); - struct amdgpu_clock_and_voltage_limits *max_limits; - bool disable_mclk_switching; - u32 sclk, mclk; - int i; - - if (rps->vce_active) { - rps->evclk = adev->pm.dpm.vce_states[adev->pm.dpm.vce_level].evclk; - rps->ecclk = adev->pm.dpm.vce_states[adev->pm.dpm.vce_level].ecclk; - } else { - rps->evclk = 0; - rps->ecclk = 0; - } - - if ((adev->pm.dpm.new_active_crtc_count > 1) || - ci_dpm_vblank_too_short(adev)) - disable_mclk_switching = true; - else - disable_mclk_switching = false; - - if ((rps->class & ATOM_PPLIB_CLASSIFICATION_UI_MASK) == ATOM_PPLIB_CLASSIFICATION_UI_BATTERY) - pi->battery_state = true; - else - pi->battery_state = false; - - if (adev->pm.ac_power) - max_limits = &adev->pm.dpm.dyn_state.max_clock_voltage_on_ac; - else - max_limits = &adev->pm.dpm.dyn_state.max_clock_voltage_on_dc; - - if (adev->pm.ac_power == false) { - for (i = 0; i < ps->performance_level_count; i++) { - if (ps->performance_levels[i].mclk > max_limits->mclk) - ps->performance_levels[i].mclk = max_limits->mclk; - if (ps->performance_levels[i].sclk > max_limits->sclk) - ps->performance_levels[i].sclk = max_limits->sclk; - } - } - - /* XXX validate the min clocks required for display */ - - if (disable_mclk_switching) { - mclk = ps->performance_levels[ps->performance_level_count - 1].mclk; - sclk = ps->performance_levels[0].sclk; - } else { - mclk = ps->performance_levels[0].mclk; - sclk = ps->performance_levels[0].sclk; - } - - if (adev->pm.pm_display_cfg.min_core_set_clock > sclk) - sclk = adev->pm.pm_display_cfg.min_core_set_clock; - - if (adev->pm.pm_display_cfg.min_mem_set_clock > mclk) - mclk = adev->pm.pm_display_cfg.min_mem_set_clock; - - if (rps->vce_active) { - if (sclk < adev->pm.dpm.vce_states[adev->pm.dpm.vce_level].sclk) - sclk = adev->pm.dpm.vce_states[adev->pm.dpm.vce_level].sclk; - if (mclk < adev->pm.dpm.vce_states[adev->pm.dpm.vce_level].mclk) - mclk = adev->pm.dpm.vce_states[adev->pm.dpm.vce_level].mclk; - } - - ps->performance_levels[0].sclk = sclk; - ps->performance_levels[0].mclk = mclk; - - if (ps->performance_levels[1].sclk < ps->performance_levels[0].sclk) - ps->performance_levels[1].sclk = ps->performance_levels[0].sclk; - - if (disable_mclk_switching) { - if (ps->performance_levels[0].mclk < ps->performance_levels[1].mclk) - ps->performance_levels[0].mclk = ps->performance_levels[1].mclk; - } else { - if (ps->performance_levels[1].mclk < ps->performance_levels[0].mclk) - ps->performance_levels[1].mclk = ps->performance_levels[0].mclk; - } -} - -static int ci_thermal_set_temperature_range(struct amdgpu_device *adev, - int min_temp, int max_temp) -{ - int low_temp = 0 * 1000; - int high_temp = 255 * 1000; - u32 tmp; - - if (low_temp < min_temp) - low_temp = min_temp; - if (high_temp > max_temp) - high_temp = max_temp; - if (high_temp < low_temp) { - DRM_ERROR("invalid thermal range: %d - %d\n", low_temp, high_temp); - return -EINVAL; - } - - tmp = RREG32_SMC(ixCG_THERMAL_INT); - tmp &= ~(CG_THERMAL_INT__DIG_THERM_INTH_MASK | CG_THERMAL_INT__DIG_THERM_INTL_MASK); - tmp |= ((high_temp / 1000) << CG_THERMAL_INT__DIG_THERM_INTH__SHIFT) | - ((low_temp / 1000)) << CG_THERMAL_INT__DIG_THERM_INTL__SHIFT; - WREG32_SMC(ixCG_THERMAL_INT, tmp); - -#if 0 - /* XXX: need to figure out how to handle this properly */ - tmp = RREG32_SMC(ixCG_THERMAL_CTRL); - tmp &= DIG_THERM_DPM_MASK; - tmp |= DIG_THERM_DPM(high_temp / 1000); - WREG32_SMC(ixCG_THERMAL_CTRL, tmp); -#endif - - adev->pm.dpm.thermal.min_temp = low_temp; - adev->pm.dpm.thermal.max_temp = high_temp; - return 0; -} - -static int ci_thermal_enable_alert(struct amdgpu_device *adev, - bool enable) -{ - u32 thermal_int = RREG32_SMC(ixCG_THERMAL_INT); - PPSMC_Result result; - - if (enable) { - thermal_int &= ~(CG_THERMAL_INT_CTRL__THERM_INTH_MASK_MASK | - CG_THERMAL_INT_CTRL__THERM_INTL_MASK_MASK); - WREG32_SMC(ixCG_THERMAL_INT, thermal_int); - result = amdgpu_ci_send_msg_to_smc(adev, PPSMC_MSG_Thermal_Cntl_Enable); - if (result != PPSMC_Result_OK) { - DRM_DEBUG_KMS("Could not enable thermal interrupts.\n"); - return -EINVAL; - } - } else { - thermal_int |= CG_THERMAL_INT_CTRL__THERM_INTH_MASK_MASK | - CG_THERMAL_INT_CTRL__THERM_INTL_MASK_MASK; - WREG32_SMC(ixCG_THERMAL_INT, thermal_int); - result = amdgpu_ci_send_msg_to_smc(adev, PPSMC_MSG_Thermal_Cntl_Disable); - if (result != PPSMC_Result_OK) { - DRM_DEBUG_KMS("Could not disable thermal interrupts.\n"); - return -EINVAL; - } - } - - return 0; -} - -static void ci_fan_ctrl_set_static_mode(struct amdgpu_device *adev, u32 mode) -{ - struct ci_power_info *pi = ci_get_pi(adev); - u32 tmp; - - if (pi->fan_ctrl_is_in_default_mode) { - tmp = (RREG32_SMC(ixCG_FDO_CTRL2) & CG_FDO_CTRL2__FDO_PWM_MODE_MASK) - >> CG_FDO_CTRL2__FDO_PWM_MODE__SHIFT; - pi->fan_ctrl_default_mode = tmp; - tmp = (RREG32_SMC(ixCG_FDO_CTRL2) & CG_FDO_CTRL2__TMIN_MASK) - >> CG_FDO_CTRL2__TMIN__SHIFT; - pi->t_min = tmp; - pi->fan_ctrl_is_in_default_mode = false; - } - - tmp = RREG32_SMC(ixCG_FDO_CTRL2) & ~CG_FDO_CTRL2__TMIN_MASK; - tmp |= 0 << CG_FDO_CTRL2__TMIN__SHIFT; - WREG32_SMC(ixCG_FDO_CTRL2, tmp); - - tmp = RREG32_SMC(ixCG_FDO_CTRL2) & ~CG_FDO_CTRL2__FDO_PWM_MODE_MASK; - tmp |= mode << CG_FDO_CTRL2__FDO_PWM_MODE__SHIFT; - WREG32_SMC(ixCG_FDO_CTRL2, tmp); -} - -static int ci_thermal_setup_fan_table(struct amdgpu_device *adev) -{ - struct ci_power_info *pi = ci_get_pi(adev); - SMU7_Discrete_FanTable fan_table = { FDO_MODE_HARDWARE }; - u32 duty100; - u32 t_diff1, t_diff2, pwm_diff1, pwm_diff2; - u16 fdo_min, slope1, slope2; - u32 reference_clock, tmp; - int ret; - u64 tmp64; - - if (!pi->fan_table_start) { - adev->pm.dpm.fan.ucode_fan_control = false; - return 0; - } - - duty100 = (RREG32_SMC(ixCG_FDO_CTRL1) & CG_FDO_CTRL1__FMAX_DUTY100_MASK) - >> CG_FDO_CTRL1__FMAX_DUTY100__SHIFT; - - if (duty100 == 0) { - adev->pm.dpm.fan.ucode_fan_control = false; - return 0; - } - - tmp64 = (u64)adev->pm.dpm.fan.pwm_min * duty100; - do_div(tmp64, 10000); - fdo_min = (u16)tmp64; - - t_diff1 = adev->pm.dpm.fan.t_med - adev->pm.dpm.fan.t_min; - t_diff2 = adev->pm.dpm.fan.t_high - adev->pm.dpm.fan.t_med; - - pwm_diff1 = adev->pm.dpm.fan.pwm_med - adev->pm.dpm.fan.pwm_min; - pwm_diff2 = adev->pm.dpm.fan.pwm_high - adev->pm.dpm.fan.pwm_med; - - slope1 = (u16)((50 + ((16 * duty100 * pwm_diff1) / t_diff1)) / 100); - slope2 = (u16)((50 + ((16 * duty100 * pwm_diff2) / t_diff2)) / 100); - - fan_table.TempMin = cpu_to_be16((50 + adev->pm.dpm.fan.t_min) / 100); - fan_table.TempMed = cpu_to_be16((50 + adev->pm.dpm.fan.t_med) / 100); - fan_table.TempMax = cpu_to_be16((50 + adev->pm.dpm.fan.t_max) / 100); - - fan_table.Slope1 = cpu_to_be16(slope1); - fan_table.Slope2 = cpu_to_be16(slope2); - - fan_table.FdoMin = cpu_to_be16(fdo_min); - - fan_table.HystDown = cpu_to_be16(adev->pm.dpm.fan.t_hyst); - - fan_table.HystUp = cpu_to_be16(1); - - fan_table.HystSlope = cpu_to_be16(1); - - fan_table.TempRespLim = cpu_to_be16(5); - - reference_clock = amdgpu_asic_get_xclk(adev); - - fan_table.RefreshPeriod = cpu_to_be32((adev->pm.dpm.fan.cycle_delay * - reference_clock) / 1600); - - fan_table.FdoMax = cpu_to_be16((u16)duty100); - - tmp = (RREG32_SMC(ixCG_MULT_THERMAL_CTRL) & CG_MULT_THERMAL_CTRL__TEMP_SEL_MASK) - >> CG_MULT_THERMAL_CTRL__TEMP_SEL__SHIFT; - fan_table.TempSrc = (uint8_t)tmp; - - ret = amdgpu_ci_copy_bytes_to_smc(adev, - pi->fan_table_start, - (u8 *)(&fan_table), - sizeof(fan_table), - pi->sram_end); - - if (ret) { - DRM_ERROR("Failed to load fan table to the SMC."); - adev->pm.dpm.fan.ucode_fan_control = false; - } - - return 0; -} - -static int ci_fan_ctrl_start_smc_fan_control(struct amdgpu_device *adev) -{ - struct ci_power_info *pi = ci_get_pi(adev); - PPSMC_Result ret; - - if (pi->caps_od_fuzzy_fan_control_support) { - ret = amdgpu_ci_send_msg_to_smc_with_parameter(adev, - PPSMC_StartFanControl, - FAN_CONTROL_FUZZY); - if (ret != PPSMC_Result_OK) - return -EINVAL; - ret = amdgpu_ci_send_msg_to_smc_with_parameter(adev, - PPSMC_MSG_SetFanPwmMax, - adev->pm.dpm.fan.default_max_fan_pwm); - if (ret != PPSMC_Result_OK) - return -EINVAL; - } else { - ret = amdgpu_ci_send_msg_to_smc_with_parameter(adev, - PPSMC_StartFanControl, - FAN_CONTROL_TABLE); - if (ret != PPSMC_Result_OK) - return -EINVAL; - } - - pi->fan_is_controlled_by_smc = true; - return 0; -} - - -static int ci_fan_ctrl_stop_smc_fan_control(struct amdgpu_device *adev) -{ - PPSMC_Result ret; - struct ci_power_info *pi = ci_get_pi(adev); - - ret = amdgpu_ci_send_msg_to_smc(adev, PPSMC_StopFanControl); - if (ret == PPSMC_Result_OK) { - pi->fan_is_controlled_by_smc = false; - return 0; - } else { - return -EINVAL; - } -} - -static int ci_dpm_get_fan_speed_percent(void *handle, - u32 *speed) -{ - u32 duty, duty100; - u64 tmp64; - struct amdgpu_device *adev = (struct amdgpu_device *)handle; - - if (adev->pm.no_fan) - return -ENOENT; - - duty100 = (RREG32_SMC(ixCG_FDO_CTRL1) & CG_FDO_CTRL1__FMAX_DUTY100_MASK) - >> CG_FDO_CTRL1__FMAX_DUTY100__SHIFT; - duty = (RREG32_SMC(ixCG_THERMAL_STATUS) & CG_THERMAL_STATUS__FDO_PWM_DUTY_MASK) - >> CG_THERMAL_STATUS__FDO_PWM_DUTY__SHIFT; - - if (duty100 == 0) - return -EINVAL; - - tmp64 = (u64)duty * 100; - do_div(tmp64, duty100); - *speed = (u32)tmp64; - - if (*speed > 100) - *speed = 100; - - return 0; -} - -static int ci_dpm_set_fan_speed_percent(void *handle, - u32 speed) -{ - u32 tmp; - u32 duty, duty100; - u64 tmp64; - struct amdgpu_device *adev = (struct amdgpu_device *)handle; - struct ci_power_info *pi = ci_get_pi(adev); - - if (adev->pm.no_fan) - return -ENOENT; - - if (pi->fan_is_controlled_by_smc) - return -EINVAL; - - if (speed > 100) - return -EINVAL; - - duty100 = (RREG32_SMC(ixCG_FDO_CTRL1) & CG_FDO_CTRL1__FMAX_DUTY100_MASK) - >> CG_FDO_CTRL1__FMAX_DUTY100__SHIFT; - - if (duty100 == 0) - return -EINVAL; - - tmp64 = (u64)speed * duty100; - do_div(tmp64, 100); - duty = (u32)tmp64; - - tmp = RREG32_SMC(ixCG_FDO_CTRL0) & ~CG_FDO_CTRL0__FDO_STATIC_DUTY_MASK; - tmp |= duty << CG_FDO_CTRL0__FDO_STATIC_DUTY__SHIFT; - WREG32_SMC(ixCG_FDO_CTRL0, tmp); - - return 0; -} - -static void ci_dpm_set_fan_control_mode(void *handle, u32 mode) -{ - struct amdgpu_device *adev = (struct amdgpu_device *)handle; - - switch (mode) { - case AMD_FAN_CTRL_NONE: - if (adev->pm.dpm.fan.ucode_fan_control) - ci_fan_ctrl_stop_smc_fan_control(adev); - ci_dpm_set_fan_speed_percent(adev, 100); - break; - case AMD_FAN_CTRL_MANUAL: - if (adev->pm.dpm.fan.ucode_fan_control) - ci_fan_ctrl_stop_smc_fan_control(adev); - break; - case AMD_FAN_CTRL_AUTO: - if (adev->pm.dpm.fan.ucode_fan_control) - ci_thermal_start_smc_fan_control(adev); - break; - default: - break; - } -} - -static u32 ci_dpm_get_fan_control_mode(void *handle) -{ - struct amdgpu_device *adev = (struct amdgpu_device *)handle; - struct ci_power_info *pi = ci_get_pi(adev); - - if (pi->fan_is_controlled_by_smc) - return AMD_FAN_CTRL_AUTO; - else - return AMD_FAN_CTRL_MANUAL; -} - -#if 0 -static int ci_fan_ctrl_get_fan_speed_rpm(struct amdgpu_device *adev, - u32 *speed) -{ - u32 tach_period; - u32 xclk = amdgpu_asic_get_xclk(adev); - - if (adev->pm.no_fan) - return -ENOENT; - - if (adev->pm.fan_pulses_per_revolution == 0) - return -ENOENT; - - tach_period = (RREG32_SMC(ixCG_TACH_STATUS) & CG_TACH_STATUS__TACH_PERIOD_MASK) - >> CG_TACH_STATUS__TACH_PERIOD__SHIFT; - if (tach_period == 0) - return -ENOENT; - - *speed = 60 * xclk * 10000 / tach_period; - - return 0; -} - -static int ci_fan_ctrl_set_fan_speed_rpm(struct amdgpu_device *adev, - u32 speed) -{ - u32 tach_period, tmp; - u32 xclk = amdgpu_asic_get_xclk(adev); - - if (adev->pm.no_fan) - return -ENOENT; - - if (adev->pm.fan_pulses_per_revolution == 0) - return -ENOENT; - - if ((speed < adev->pm.fan_min_rpm) || - (speed > adev->pm.fan_max_rpm)) - return -EINVAL; - - if (adev->pm.dpm.fan.ucode_fan_control) - ci_fan_ctrl_stop_smc_fan_control(adev); - - tach_period = 60 * xclk * 10000 / (8 * speed); - tmp = RREG32_SMC(ixCG_TACH_CTRL) & ~CG_TACH_CTRL__TARGET_PERIOD_MASK; - tmp |= tach_period << CG_TACH_CTRL__TARGET_PERIOD__SHIFT; - WREG32_SMC(CG_TACH_CTRL, tmp); - - ci_fan_ctrl_set_static_mode(adev, FDO_PWM_MODE_STATIC_RPM); - - return 0; -} -#endif - -static void ci_fan_ctrl_set_default_mode(struct amdgpu_device *adev) -{ - struct ci_power_info *pi = ci_get_pi(adev); - u32 tmp; - - if (!pi->fan_ctrl_is_in_default_mode) { - tmp = RREG32_SMC(ixCG_FDO_CTRL2) & ~CG_FDO_CTRL2__FDO_PWM_MODE_MASK; - tmp |= pi->fan_ctrl_default_mode << CG_FDO_CTRL2__FDO_PWM_MODE__SHIFT; - WREG32_SMC(ixCG_FDO_CTRL2, tmp); - - tmp = RREG32_SMC(ixCG_FDO_CTRL2) & ~CG_FDO_CTRL2__TMIN_MASK; - tmp |= pi->t_min << CG_FDO_CTRL2__TMIN__SHIFT; - WREG32_SMC(ixCG_FDO_CTRL2, tmp); - pi->fan_ctrl_is_in_default_mode = true; - } -} - -static void ci_thermal_start_smc_fan_control(struct amdgpu_device *adev) -{ - if (adev->pm.dpm.fan.ucode_fan_control) { - ci_fan_ctrl_start_smc_fan_control(adev); - ci_fan_ctrl_set_static_mode(adev, FDO_PWM_MODE_STATIC); - } -} - -static void ci_thermal_initialize(struct amdgpu_device *adev) -{ - u32 tmp; - - if (adev->pm.fan_pulses_per_revolution) { - tmp = RREG32_SMC(ixCG_TACH_CTRL) & ~CG_TACH_CTRL__EDGE_PER_REV_MASK; - tmp |= (adev->pm.fan_pulses_per_revolution - 1) - << CG_TACH_CTRL__EDGE_PER_REV__SHIFT; - WREG32_SMC(ixCG_TACH_CTRL, tmp); - } - - tmp = RREG32_SMC(ixCG_FDO_CTRL2) & ~CG_FDO_CTRL2__TACH_PWM_RESP_RATE_MASK; - tmp |= 0x28 << CG_FDO_CTRL2__TACH_PWM_RESP_RATE__SHIFT; - WREG32_SMC(ixCG_FDO_CTRL2, tmp); -} - -static int ci_thermal_start_thermal_controller(struct amdgpu_device *adev) -{ - int ret; - - ci_thermal_initialize(adev); - ret = ci_thermal_set_temperature_range(adev, CISLANDS_TEMP_RANGE_MIN, CISLANDS_TEMP_RANGE_MAX); - if (ret) - return ret; - ret = ci_thermal_enable_alert(adev, true); - if (ret) - return ret; - if (adev->pm.dpm.fan.ucode_fan_control) { - ret = ci_thermal_setup_fan_table(adev); - if (ret) - return ret; - ci_thermal_start_smc_fan_control(adev); - } - - return 0; -} - -static void ci_thermal_stop_thermal_controller(struct amdgpu_device *adev) -{ - if (!adev->pm.no_fan) - ci_fan_ctrl_set_default_mode(adev); -} - -static int ci_read_smc_soft_register(struct amdgpu_device *adev, - u16 reg_offset, u32 *value) -{ - struct ci_power_info *pi = ci_get_pi(adev); - - return amdgpu_ci_read_smc_sram_dword(adev, - pi->soft_regs_start + reg_offset, - value, pi->sram_end); -} - -static int ci_write_smc_soft_register(struct amdgpu_device *adev, - u16 reg_offset, u32 value) -{ - struct ci_power_info *pi = ci_get_pi(adev); - - return amdgpu_ci_write_smc_sram_dword(adev, - pi->soft_regs_start + reg_offset, - value, pi->sram_end); -} - -static void ci_init_fps_limits(struct amdgpu_device *adev) -{ - struct ci_power_info *pi = ci_get_pi(adev); - SMU7_Discrete_DpmTable *table = &pi->smc_state_table; - - if (pi->caps_fps) { - u16 tmp; - - tmp = 45; - table->FpsHighT = cpu_to_be16(tmp); - - tmp = 30; - table->FpsLowT = cpu_to_be16(tmp); - } -} - -static int ci_update_sclk_t(struct amdgpu_device *adev) -{ - struct ci_power_info *pi = ci_get_pi(adev); - int ret = 0; - u32 low_sclk_interrupt_t = 0; - - if (pi->caps_sclk_throttle_low_notification) { - low_sclk_interrupt_t = cpu_to_be32(pi->low_sclk_interrupt_t); - - ret = amdgpu_ci_copy_bytes_to_smc(adev, - pi->dpm_table_start + - offsetof(SMU7_Discrete_DpmTable, LowSclkInterruptT), - (u8 *)&low_sclk_interrupt_t, - sizeof(u32), pi->sram_end); - - } - - return ret; -} - -static void ci_get_leakage_voltages(struct amdgpu_device *adev) -{ - struct ci_power_info *pi = ci_get_pi(adev); - u16 leakage_id, virtual_voltage_id; - u16 vddc, vddci; - int i; - - pi->vddc_leakage.count = 0; - pi->vddci_leakage.count = 0; - - if (adev->pm.dpm.platform_caps & ATOM_PP_PLATFORM_CAP_EVV) { - for (i = 0; i < CISLANDS_MAX_LEAKAGE_COUNT; i++) { - virtual_voltage_id = ATOM_VIRTUAL_VOLTAGE_ID0 + i; - if (amdgpu_atombios_get_voltage_evv(adev, virtual_voltage_id, &vddc) != 0) - continue; - if (vddc != 0 && vddc != virtual_voltage_id) { - pi->vddc_leakage.actual_voltage[pi->vddc_leakage.count] = vddc; - pi->vddc_leakage.leakage_id[pi->vddc_leakage.count] = virtual_voltage_id; - pi->vddc_leakage.count++; - } - } - } else if (amdgpu_atombios_get_leakage_id_from_vbios(adev, &leakage_id) == 0) { - for (i = 0; i < CISLANDS_MAX_LEAKAGE_COUNT; i++) { - virtual_voltage_id = ATOM_VIRTUAL_VOLTAGE_ID0 + i; - if (amdgpu_atombios_get_leakage_vddc_based_on_leakage_params(adev, &vddc, &vddci, - virtual_voltage_id, - leakage_id) == 0) { - if (vddc != 0 && vddc != virtual_voltage_id) { - pi->vddc_leakage.actual_voltage[pi->vddc_leakage.count] = vddc; - pi->vddc_leakage.leakage_id[pi->vddc_leakage.count] = virtual_voltage_id; - pi->vddc_leakage.count++; - } - if (vddci != 0 && vddci != virtual_voltage_id) { - pi->vddci_leakage.actual_voltage[pi->vddci_leakage.count] = vddci; - pi->vddci_leakage.leakage_id[pi->vddci_leakage.count] = virtual_voltage_id; - pi->vddci_leakage.count++; - } - } - } - } -} - -static void ci_set_dpm_event_sources(struct amdgpu_device *adev, u32 sources) -{ - struct ci_power_info *pi = ci_get_pi(adev); - bool want_thermal_protection; - enum amdgpu_dpm_event_src dpm_event_src; - u32 tmp; - - switch (sources) { - case 0: - default: - want_thermal_protection = false; - break; - case (1 << AMDGPU_DPM_AUTO_THROTTLE_SRC_THERMAL): - want_thermal_protection = true; - dpm_event_src = AMDGPU_DPM_EVENT_SRC_DIGITAL; - break; - case (1 << AMDGPU_DPM_AUTO_THROTTLE_SRC_EXTERNAL): - want_thermal_protection = true; - dpm_event_src = AMDGPU_DPM_EVENT_SRC_EXTERNAL; - break; - case ((1 << AMDGPU_DPM_AUTO_THROTTLE_SRC_EXTERNAL) | - (1 << AMDGPU_DPM_AUTO_THROTTLE_SRC_THERMAL)): - want_thermal_protection = true; - dpm_event_src = AMDGPU_DPM_EVENT_SRC_DIGIAL_OR_EXTERNAL; - break; - } - - if (want_thermal_protection) { -#if 0 - /* XXX: need to figure out how to handle this properly */ - tmp = RREG32_SMC(ixCG_THERMAL_CTRL); - tmp &= DPM_EVENT_SRC_MASK; - tmp |= DPM_EVENT_SRC(dpm_event_src); - WREG32_SMC(ixCG_THERMAL_CTRL, tmp); -#endif - - tmp = RREG32_SMC(ixGENERAL_PWRMGT); - if (pi->thermal_protection) - tmp &= ~GENERAL_PWRMGT__THERMAL_PROTECTION_DIS_MASK; - else - tmp |= GENERAL_PWRMGT__THERMAL_PROTECTION_DIS_MASK; - WREG32_SMC(ixGENERAL_PWRMGT, tmp); - } else { - tmp = RREG32_SMC(ixGENERAL_PWRMGT); - tmp |= GENERAL_PWRMGT__THERMAL_PROTECTION_DIS_MASK; - WREG32_SMC(ixGENERAL_PWRMGT, tmp); - } -} - -static void ci_enable_auto_throttle_source(struct amdgpu_device *adev, - enum amdgpu_dpm_auto_throttle_src source, - bool enable) -{ - struct ci_power_info *pi = ci_get_pi(adev); - - if (enable) { - if (!(pi->active_auto_throttle_sources & (1 << source))) { - pi->active_auto_throttle_sources |= 1 << source; - ci_set_dpm_event_sources(adev, pi->active_auto_throttle_sources); - } - } else { - if (pi->active_auto_throttle_sources & (1 << source)) { - pi->active_auto_throttle_sources &= ~(1 << source); - ci_set_dpm_event_sources(adev, pi->active_auto_throttle_sources); - } - } -} - -static void ci_enable_vr_hot_gpio_interrupt(struct amdgpu_device *adev) -{ - if (adev->pm.dpm.platform_caps & ATOM_PP_PLATFORM_CAP_REGULATOR_HOT) - amdgpu_ci_send_msg_to_smc(adev, PPSMC_MSG_EnableVRHotGPIOInterrupt); -} - -static int ci_unfreeze_sclk_mclk_dpm(struct amdgpu_device *adev) -{ - struct ci_power_info *pi = ci_get_pi(adev); - PPSMC_Result smc_result; - - if (!pi->need_update_smu7_dpm_table) - return 0; - - if ((!pi->sclk_dpm_key_disabled) && - (pi->need_update_smu7_dpm_table & (DPMTABLE_OD_UPDATE_SCLK | DPMTABLE_UPDATE_SCLK))) { - smc_result = amdgpu_ci_send_msg_to_smc(adev, PPSMC_MSG_SCLKDPM_UnfreezeLevel); - if (smc_result != PPSMC_Result_OK) - return -EINVAL; - } - - if ((!pi->mclk_dpm_key_disabled) && - (pi->need_update_smu7_dpm_table & DPMTABLE_OD_UPDATE_MCLK)) { - smc_result = amdgpu_ci_send_msg_to_smc(adev, PPSMC_MSG_MCLKDPM_UnfreezeLevel); - if (smc_result != PPSMC_Result_OK) - return -EINVAL; - } - - pi->need_update_smu7_dpm_table = 0; - return 0; -} - -static int ci_enable_sclk_mclk_dpm(struct amdgpu_device *adev, bool enable) -{ - struct ci_power_info *pi = ci_get_pi(adev); - PPSMC_Result smc_result; - - if (enable) { - if (!pi->sclk_dpm_key_disabled) { - smc_result = amdgpu_ci_send_msg_to_smc(adev, PPSMC_MSG_DPM_Enable); - if (smc_result != PPSMC_Result_OK) - return -EINVAL; - } - - if (!pi->mclk_dpm_key_disabled) { - smc_result = amdgpu_ci_send_msg_to_smc(adev, PPSMC_MSG_MCLKDPM_Enable); - if (smc_result != PPSMC_Result_OK) - return -EINVAL; - - WREG32_P(mmMC_SEQ_CNTL_3, MC_SEQ_CNTL_3__CAC_EN_MASK, - ~MC_SEQ_CNTL_3__CAC_EN_MASK); - - WREG32_SMC(ixLCAC_MC0_CNTL, 0x05); - WREG32_SMC(ixLCAC_MC1_CNTL, 0x05); - WREG32_SMC(ixLCAC_CPL_CNTL, 0x100005); - - udelay(10); - - WREG32_SMC(ixLCAC_MC0_CNTL, 0x400005); - WREG32_SMC(ixLCAC_MC1_CNTL, 0x400005); - WREG32_SMC(ixLCAC_CPL_CNTL, 0x500005); - } - } else { - if (!pi->sclk_dpm_key_disabled) { - smc_result = amdgpu_ci_send_msg_to_smc(adev, PPSMC_MSG_DPM_Disable); - if (smc_result != PPSMC_Result_OK) - return -EINVAL; - } - - if (!pi->mclk_dpm_key_disabled) { - smc_result = amdgpu_ci_send_msg_to_smc(adev, PPSMC_MSG_MCLKDPM_Disable); - if (smc_result != PPSMC_Result_OK) - return -EINVAL; - } - } - - return 0; -} - -static int ci_start_dpm(struct amdgpu_device *adev) -{ - struct ci_power_info *pi = ci_get_pi(adev); - PPSMC_Result smc_result; - int ret; - u32 tmp; - - tmp = RREG32_SMC(ixGENERAL_PWRMGT); - tmp |= GENERAL_PWRMGT__GLOBAL_PWRMGT_EN_MASK; - WREG32_SMC(ixGENERAL_PWRMGT, tmp); - - tmp = RREG32_SMC(ixSCLK_PWRMGT_CNTL); - tmp |= SCLK_PWRMGT_CNTL__DYNAMIC_PM_EN_MASK; - WREG32_SMC(ixSCLK_PWRMGT_CNTL, tmp); - - ci_write_smc_soft_register(adev, offsetof(SMU7_SoftRegisters, VoltageChangeTimeout), 0x1000); - - WREG32_P(mmBIF_LNCNT_RESET, 0, ~BIF_LNCNT_RESET__RESET_LNCNT_EN_MASK); - - smc_result = amdgpu_ci_send_msg_to_smc(adev, PPSMC_MSG_Voltage_Cntl_Enable); - if (smc_result != PPSMC_Result_OK) - return -EINVAL; - - ret = ci_enable_sclk_mclk_dpm(adev, true); - if (ret) - return ret; - - if (!pi->pcie_dpm_key_disabled) { - smc_result = amdgpu_ci_send_msg_to_smc(adev, PPSMC_MSG_PCIeDPM_Enable); - if (smc_result != PPSMC_Result_OK) - return -EINVAL; - } - - return 0; -} - -static int ci_freeze_sclk_mclk_dpm(struct amdgpu_device *adev) -{ - struct ci_power_info *pi = ci_get_pi(adev); - PPSMC_Result smc_result; - - if (!pi->need_update_smu7_dpm_table) - return 0; - - if ((!pi->sclk_dpm_key_disabled) && - (pi->need_update_smu7_dpm_table & (DPMTABLE_OD_UPDATE_SCLK | DPMTABLE_UPDATE_SCLK))) { - smc_result = amdgpu_ci_send_msg_to_smc(adev, PPSMC_MSG_SCLKDPM_FreezeLevel); - if (smc_result != PPSMC_Result_OK) - return -EINVAL; - } - - if ((!pi->mclk_dpm_key_disabled) && - (pi->need_update_smu7_dpm_table & DPMTABLE_OD_UPDATE_MCLK)) { - smc_result = amdgpu_ci_send_msg_to_smc(adev, PPSMC_MSG_MCLKDPM_FreezeLevel); - if (smc_result != PPSMC_Result_OK) - return -EINVAL; - } - - return 0; -} - -static int ci_stop_dpm(struct amdgpu_device *adev) -{ - struct ci_power_info *pi = ci_get_pi(adev); - PPSMC_Result smc_result; - int ret; - u32 tmp; - - tmp = RREG32_SMC(ixGENERAL_PWRMGT); - tmp &= ~GENERAL_PWRMGT__GLOBAL_PWRMGT_EN_MASK; - WREG32_SMC(ixGENERAL_PWRMGT, tmp); - - tmp = RREG32_SMC(ixSCLK_PWRMGT_CNTL); - tmp &= ~SCLK_PWRMGT_CNTL__DYNAMIC_PM_EN_MASK; - WREG32_SMC(ixSCLK_PWRMGT_CNTL, tmp); - - if (!pi->pcie_dpm_key_disabled) { - smc_result = amdgpu_ci_send_msg_to_smc(adev, PPSMC_MSG_PCIeDPM_Disable); - if (smc_result != PPSMC_Result_OK) - return -EINVAL; - } - - ret = ci_enable_sclk_mclk_dpm(adev, false); - if (ret) - return ret; - - smc_result = amdgpu_ci_send_msg_to_smc(adev, PPSMC_MSG_Voltage_Cntl_Disable); - if (smc_result != PPSMC_Result_OK) - return -EINVAL; - - return 0; -} - -static void ci_enable_sclk_control(struct amdgpu_device *adev, bool enable) -{ - u32 tmp = RREG32_SMC(ixSCLK_PWRMGT_CNTL); - - if (enable) - tmp &= ~SCLK_PWRMGT_CNTL__SCLK_PWRMGT_OFF_MASK; - else - tmp |= SCLK_PWRMGT_CNTL__SCLK_PWRMGT_OFF_MASK; - WREG32_SMC(ixSCLK_PWRMGT_CNTL, tmp); -} - -#if 0 -static int ci_notify_hw_of_power_source(struct amdgpu_device *adev, - bool ac_power) -{ - struct ci_power_info *pi = ci_get_pi(adev); - struct amdgpu_cac_tdp_table *cac_tdp_table = - adev->pm.dpm.dyn_state.cac_tdp_table; - u32 power_limit; - - if (ac_power) - power_limit = (u32)(cac_tdp_table->maximum_power_delivery_limit * 256); - else - power_limit = (u32)(cac_tdp_table->battery_power_limit * 256); - - ci_set_power_limit(adev, power_limit); - - if (pi->caps_automatic_dc_transition) { - if (ac_power) - amdgpu_ci_send_msg_to_smc(adev, PPSMC_MSG_RunningOnAC); - else - amdgpu_ci_send_msg_to_smc(adev, PPSMC_MSG_Remove_DC_Clamp); - } - - return 0; -} -#endif - -static PPSMC_Result amdgpu_ci_send_msg_to_smc_with_parameter(struct amdgpu_device *adev, - PPSMC_Msg msg, u32 parameter) -{ - WREG32(mmSMC_MSG_ARG_0, parameter); - return amdgpu_ci_send_msg_to_smc(adev, msg); -} - -static PPSMC_Result amdgpu_ci_send_msg_to_smc_return_parameter(struct amdgpu_device *adev, - PPSMC_Msg msg, u32 *parameter) -{ - PPSMC_Result smc_result; - - smc_result = amdgpu_ci_send_msg_to_smc(adev, msg); - - if ((smc_result == PPSMC_Result_OK) && parameter) - *parameter = RREG32(mmSMC_MSG_ARG_0); - - return smc_result; -} - -static int ci_dpm_force_state_sclk(struct amdgpu_device *adev, u32 n) -{ - struct ci_power_info *pi = ci_get_pi(adev); - - if (!pi->sclk_dpm_key_disabled) { - PPSMC_Result smc_result = - amdgpu_ci_send_msg_to_smc_with_parameter(adev, PPSMC_MSG_SCLKDPM_SetEnabledMask, 1 << n); - if (smc_result != PPSMC_Result_OK) - return -EINVAL; - } - - return 0; -} - -static int ci_dpm_force_state_mclk(struct amdgpu_device *adev, u32 n) -{ - struct ci_power_info *pi = ci_get_pi(adev); - - if (!pi->mclk_dpm_key_disabled) { - PPSMC_Result smc_result = - amdgpu_ci_send_msg_to_smc_with_parameter(adev, PPSMC_MSG_MCLKDPM_SetEnabledMask, 1 << n); - if (smc_result != PPSMC_Result_OK) - return -EINVAL; - } - - return 0; -} - -static int ci_dpm_force_state_pcie(struct amdgpu_device *adev, u32 n) -{ - struct ci_power_info *pi = ci_get_pi(adev); - - if (!pi->pcie_dpm_key_disabled) { - PPSMC_Result smc_result = - amdgpu_ci_send_msg_to_smc_with_parameter(adev, PPSMC_MSG_PCIeDPM_ForceLevel, n); - if (smc_result != PPSMC_Result_OK) - return -EINVAL; - } - - return 0; -} - -static int ci_set_power_limit(struct amdgpu_device *adev, u32 n) -{ - struct ci_power_info *pi = ci_get_pi(adev); - - if (pi->power_containment_features & POWERCONTAINMENT_FEATURE_PkgPwrLimit) { - PPSMC_Result smc_result = - amdgpu_ci_send_msg_to_smc_with_parameter(adev, PPSMC_MSG_PkgPwrSetLimit, n); - if (smc_result != PPSMC_Result_OK) - return -EINVAL; - } - - return 0; -} - -static int ci_set_overdrive_target_tdp(struct amdgpu_device *adev, - u32 target_tdp) -{ - PPSMC_Result smc_result = - amdgpu_ci_send_msg_to_smc_with_parameter(adev, PPSMC_MSG_OverDriveSetTargetTdp, target_tdp); - if (smc_result != PPSMC_Result_OK) - return -EINVAL; - return 0; -} - -#if 0 -static int ci_set_boot_state(struct amdgpu_device *adev) -{ - return ci_enable_sclk_mclk_dpm(adev, false); -} -#endif - -static u32 ci_get_average_sclk_freq(struct amdgpu_device *adev) -{ - u32 sclk_freq; - PPSMC_Result smc_result = - amdgpu_ci_send_msg_to_smc_return_parameter(adev, - PPSMC_MSG_API_GetSclkFrequency, - &sclk_freq); - if (smc_result != PPSMC_Result_OK) - sclk_freq = 0; - - return sclk_freq; -} - -static u32 ci_get_average_mclk_freq(struct amdgpu_device *adev) -{ - u32 mclk_freq; - PPSMC_Result smc_result = - amdgpu_ci_send_msg_to_smc_return_parameter(adev, - PPSMC_MSG_API_GetMclkFrequency, - &mclk_freq); - if (smc_result != PPSMC_Result_OK) - mclk_freq = 0; - - return mclk_freq; -} - -static void ci_dpm_start_smc(struct amdgpu_device *adev) -{ - int i; - - amdgpu_ci_program_jump_on_start(adev); - amdgpu_ci_start_smc_clock(adev); - amdgpu_ci_start_smc(adev); - for (i = 0; i < adev->usec_timeout; i++) { - if (RREG32_SMC(ixFIRMWARE_FLAGS) & FIRMWARE_FLAGS__INTERRUPTS_ENABLED_MASK) - break; - } -} - -static void ci_dpm_stop_smc(struct amdgpu_device *adev) -{ - amdgpu_ci_reset_smc(adev); - amdgpu_ci_stop_smc_clock(adev); -} - -static int ci_process_firmware_header(struct amdgpu_device *adev) -{ - struct ci_power_info *pi = ci_get_pi(adev); - u32 tmp; - int ret; - - ret = amdgpu_ci_read_smc_sram_dword(adev, - SMU7_FIRMWARE_HEADER_LOCATION + - offsetof(SMU7_Firmware_Header, DpmTable), - &tmp, pi->sram_end); - if (ret) - return ret; - - pi->dpm_table_start = tmp; - - ret = amdgpu_ci_read_smc_sram_dword(adev, - SMU7_FIRMWARE_HEADER_LOCATION + - offsetof(SMU7_Firmware_Header, SoftRegisters), - &tmp, pi->sram_end); - if (ret) - return ret; - - pi->soft_regs_start = tmp; - - ret = amdgpu_ci_read_smc_sram_dword(adev, - SMU7_FIRMWARE_HEADER_LOCATION + - offsetof(SMU7_Firmware_Header, mcRegisterTable), - &tmp, pi->sram_end); - if (ret) - return ret; - - pi->mc_reg_table_start = tmp; - - ret = amdgpu_ci_read_smc_sram_dword(adev, - SMU7_FIRMWARE_HEADER_LOCATION + - offsetof(SMU7_Firmware_Header, FanTable), - &tmp, pi->sram_end); - if (ret) - return ret; - - pi->fan_table_start = tmp; - - ret = amdgpu_ci_read_smc_sram_dword(adev, - SMU7_FIRMWARE_HEADER_LOCATION + - offsetof(SMU7_Firmware_Header, mcArbDramTimingTable), - &tmp, pi->sram_end); - if (ret) - return ret; - - pi->arb_table_start = tmp; - - return 0; -} - -static void ci_read_clock_registers(struct amdgpu_device *adev) -{ - struct ci_power_info *pi = ci_get_pi(adev); - - pi->clock_registers.cg_spll_func_cntl = - RREG32_SMC(ixCG_SPLL_FUNC_CNTL); - pi->clock_registers.cg_spll_func_cntl_2 = - RREG32_SMC(ixCG_SPLL_FUNC_CNTL_2); - pi->clock_registers.cg_spll_func_cntl_3 = - RREG32_SMC(ixCG_SPLL_FUNC_CNTL_3); - pi->clock_registers.cg_spll_func_cntl_4 = - RREG32_SMC(ixCG_SPLL_FUNC_CNTL_4); - pi->clock_registers.cg_spll_spread_spectrum = - RREG32_SMC(ixCG_SPLL_SPREAD_SPECTRUM); - pi->clock_registers.cg_spll_spread_spectrum_2 = - RREG32_SMC(ixCG_SPLL_SPREAD_SPECTRUM_2); - pi->clock_registers.dll_cntl = RREG32(mmDLL_CNTL); - pi->clock_registers.mclk_pwrmgt_cntl = RREG32(mmMCLK_PWRMGT_CNTL); - pi->clock_registers.mpll_ad_func_cntl = RREG32(mmMPLL_AD_FUNC_CNTL); - pi->clock_registers.mpll_dq_func_cntl = RREG32(mmMPLL_DQ_FUNC_CNTL); - pi->clock_registers.mpll_func_cntl = RREG32(mmMPLL_FUNC_CNTL); - pi->clock_registers.mpll_func_cntl_1 = RREG32(mmMPLL_FUNC_CNTL_1); - pi->clock_registers.mpll_func_cntl_2 = RREG32(mmMPLL_FUNC_CNTL_2); - pi->clock_registers.mpll_ss1 = RREG32(mmMPLL_SS1); - pi->clock_registers.mpll_ss2 = RREG32(mmMPLL_SS2); -} - -static void ci_init_sclk_t(struct amdgpu_device *adev) -{ - struct ci_power_info *pi = ci_get_pi(adev); - - pi->low_sclk_interrupt_t = 0; -} - -static void ci_enable_thermal_protection(struct amdgpu_device *adev, - bool enable) -{ - u32 tmp = RREG32_SMC(ixGENERAL_PWRMGT); - - if (enable) - tmp &= ~GENERAL_PWRMGT__THERMAL_PROTECTION_DIS_MASK; - else - tmp |= GENERAL_PWRMGT__THERMAL_PROTECTION_DIS_MASK; - WREG32_SMC(ixGENERAL_PWRMGT, tmp); -} - -static void ci_enable_acpi_power_management(struct amdgpu_device *adev) -{ - u32 tmp = RREG32_SMC(ixGENERAL_PWRMGT); - - tmp |= GENERAL_PWRMGT__STATIC_PM_EN_MASK; - - WREG32_SMC(ixGENERAL_PWRMGT, tmp); -} - -#if 0 -static int ci_enter_ulp_state(struct amdgpu_device *adev) -{ - - WREG32(mmSMC_MESSAGE_0, PPSMC_MSG_SwitchToMinimumPower); - - udelay(25000); - - return 0; -} - -static int ci_exit_ulp_state(struct amdgpu_device *adev) -{ - int i; - - WREG32(mmSMC_MESSAGE_0, PPSMC_MSG_ResumeFromMinimumPower); - - udelay(7000); - - for (i = 0; i < adev->usec_timeout; i++) { - if (RREG32(mmSMC_RESP_0) == 1) - break; - udelay(1000); - } - - return 0; -} -#endif - -static int ci_notify_smc_display_change(struct amdgpu_device *adev, - bool has_display) -{ - PPSMC_Msg msg = has_display ? PPSMC_MSG_HasDisplay : PPSMC_MSG_NoDisplay; - - return (amdgpu_ci_send_msg_to_smc(adev, msg) == PPSMC_Result_OK) ? 0 : -EINVAL; -} - -static int ci_enable_ds_master_switch(struct amdgpu_device *adev, - bool enable) -{ - struct ci_power_info *pi = ci_get_pi(adev); - - if (enable) { - if (pi->caps_sclk_ds) { - if (amdgpu_ci_send_msg_to_smc(adev, PPSMC_MSG_MASTER_DeepSleep_ON) != PPSMC_Result_OK) - return -EINVAL; - } else { - if (amdgpu_ci_send_msg_to_smc(adev, PPSMC_MSG_MASTER_DeepSleep_OFF) != PPSMC_Result_OK) - return -EINVAL; - } - } else { - if (pi->caps_sclk_ds) { - if (amdgpu_ci_send_msg_to_smc(adev, PPSMC_MSG_MASTER_DeepSleep_OFF) != PPSMC_Result_OK) - return -EINVAL; - } - } - - return 0; -} - -static void ci_program_display_gap(struct amdgpu_device *adev) -{ - u32 tmp = RREG32_SMC(ixCG_DISPLAY_GAP_CNTL); - u32 pre_vbi_time_in_us; - u32 frame_time_in_us; - u32 ref_clock = adev->clock.spll.reference_freq; - u32 refresh_rate = amdgpu_dpm_get_vrefresh(adev); - u32 vblank_time = amdgpu_dpm_get_vblank_time(adev); - - tmp &= ~CG_DISPLAY_GAP_CNTL__DISP_GAP_MASK; - if (adev->pm.dpm.new_active_crtc_count > 0) - tmp |= (AMDGPU_PM_DISPLAY_GAP_VBLANK_OR_WM << CG_DISPLAY_GAP_CNTL__DISP_GAP__SHIFT); - else - tmp |= (AMDGPU_PM_DISPLAY_GAP_IGNORE << CG_DISPLAY_GAP_CNTL__DISP_GAP__SHIFT); - WREG32_SMC(ixCG_DISPLAY_GAP_CNTL, tmp); - - if (refresh_rate == 0) - refresh_rate = 60; - if (vblank_time == 0xffffffff) - vblank_time = 500; - frame_time_in_us = 1000000 / refresh_rate; - pre_vbi_time_in_us = - frame_time_in_us - 200 - vblank_time; - tmp = pre_vbi_time_in_us * (ref_clock / 100); - - WREG32_SMC(ixCG_DISPLAY_GAP_CNTL2, tmp); - ci_write_smc_soft_register(adev, offsetof(SMU7_SoftRegisters, PreVBlankGap), 0x64); - ci_write_smc_soft_register(adev, offsetof(SMU7_SoftRegisters, VBlankTimeout), (frame_time_in_us - pre_vbi_time_in_us)); - - - ci_notify_smc_display_change(adev, (adev->pm.dpm.new_active_crtc_count == 1)); - -} - -static void ci_enable_spread_spectrum(struct amdgpu_device *adev, bool enable) -{ - struct ci_power_info *pi = ci_get_pi(adev); - u32 tmp; - - if (enable) { - if (pi->caps_sclk_ss_support) { - tmp = RREG32_SMC(ixGENERAL_PWRMGT); - tmp |= GENERAL_PWRMGT__DYN_SPREAD_SPECTRUM_EN_MASK; - WREG32_SMC(ixGENERAL_PWRMGT, tmp); - } - } else { - tmp = RREG32_SMC(ixCG_SPLL_SPREAD_SPECTRUM); - tmp &= ~CG_SPLL_SPREAD_SPECTRUM__SSEN_MASK; - WREG32_SMC(ixCG_SPLL_SPREAD_SPECTRUM, tmp); - - tmp = RREG32_SMC(ixGENERAL_PWRMGT); - tmp &= ~GENERAL_PWRMGT__DYN_SPREAD_SPECTRUM_EN_MASK; - WREG32_SMC(ixGENERAL_PWRMGT, tmp); - } -} - -static void ci_program_sstp(struct amdgpu_device *adev) -{ - WREG32_SMC(ixCG_STATIC_SCREEN_PARAMETER, - ((CISLANDS_SSTU_DFLT << CG_STATIC_SCREEN_PARAMETER__STATIC_SCREEN_THRESHOLD_UNIT__SHIFT) | - (CISLANDS_SST_DFLT << CG_STATIC_SCREEN_PARAMETER__STATIC_SCREEN_THRESHOLD__SHIFT))); -} - -static void ci_enable_display_gap(struct amdgpu_device *adev) -{ - u32 tmp = RREG32_SMC(ixCG_DISPLAY_GAP_CNTL); - - tmp &= ~(CG_DISPLAY_GAP_CNTL__DISP_GAP_MASK | - CG_DISPLAY_GAP_CNTL__DISP_GAP_MCHG_MASK); - tmp |= ((AMDGPU_PM_DISPLAY_GAP_IGNORE << CG_DISPLAY_GAP_CNTL__DISP_GAP__SHIFT) | - (AMDGPU_PM_DISPLAY_GAP_VBLANK << CG_DISPLAY_GAP_CNTL__DISP_GAP_MCHG__SHIFT)); - - WREG32_SMC(ixCG_DISPLAY_GAP_CNTL, tmp); -} - -static void ci_program_vc(struct amdgpu_device *adev) -{ - u32 tmp; - - tmp = RREG32_SMC(ixSCLK_PWRMGT_CNTL); - tmp &= ~(SCLK_PWRMGT_CNTL__RESET_SCLK_CNT_MASK | SCLK_PWRMGT_CNTL__RESET_BUSY_CNT_MASK); - WREG32_SMC(ixSCLK_PWRMGT_CNTL, tmp); - - WREG32_SMC(ixCG_FREQ_TRAN_VOTING_0, CISLANDS_VRC_DFLT0); - WREG32_SMC(ixCG_FREQ_TRAN_VOTING_1, CISLANDS_VRC_DFLT1); - WREG32_SMC(ixCG_FREQ_TRAN_VOTING_2, CISLANDS_VRC_DFLT2); - WREG32_SMC(ixCG_FREQ_TRAN_VOTING_3, CISLANDS_VRC_DFLT3); - WREG32_SMC(ixCG_FREQ_TRAN_VOTING_4, CISLANDS_VRC_DFLT4); - WREG32_SMC(ixCG_FREQ_TRAN_VOTING_5, CISLANDS_VRC_DFLT5); - WREG32_SMC(ixCG_FREQ_TRAN_VOTING_6, CISLANDS_VRC_DFLT6); - WREG32_SMC(ixCG_FREQ_TRAN_VOTING_7, CISLANDS_VRC_DFLT7); -} - -static void ci_clear_vc(struct amdgpu_device *adev) -{ - u32 tmp; - - tmp = RREG32_SMC(ixSCLK_PWRMGT_CNTL); - tmp |= (SCLK_PWRMGT_CNTL__RESET_SCLK_CNT_MASK | SCLK_PWRMGT_CNTL__RESET_BUSY_CNT_MASK); - WREG32_SMC(ixSCLK_PWRMGT_CNTL, tmp); - - WREG32_SMC(ixCG_FREQ_TRAN_VOTING_0, 0); - WREG32_SMC(ixCG_FREQ_TRAN_VOTING_1, 0); - WREG32_SMC(ixCG_FREQ_TRAN_VOTING_2, 0); - WREG32_SMC(ixCG_FREQ_TRAN_VOTING_3, 0); - WREG32_SMC(ixCG_FREQ_TRAN_VOTING_4, 0); - WREG32_SMC(ixCG_FREQ_TRAN_VOTING_5, 0); - WREG32_SMC(ixCG_FREQ_TRAN_VOTING_6, 0); - WREG32_SMC(ixCG_FREQ_TRAN_VOTING_7, 0); -} - -static int ci_upload_firmware(struct amdgpu_device *adev) -{ - int i, ret; - - if (amdgpu_ci_is_smc_running(adev)) { - DRM_INFO("smc is running, no need to load smc firmware\n"); - return 0; - } - - for (i = 0; i < adev->usec_timeout; i++) { - if (RREG32_SMC(ixRCU_UC_EVENTS) & RCU_UC_EVENTS__boot_seq_done_MASK) - break; - } - WREG32_SMC(ixSMC_SYSCON_MISC_CNTL, 1); - - amdgpu_ci_stop_smc_clock(adev); - amdgpu_ci_reset_smc(adev); - - ret = amdgpu_ci_load_smc_ucode(adev, SMC_RAM_END); - - return ret; - -} - -static int ci_get_svi2_voltage_table(struct amdgpu_device *adev, - struct amdgpu_clock_voltage_dependency_table *voltage_dependency_table, - struct atom_voltage_table *voltage_table) -{ - u32 i; - - if (voltage_dependency_table == NULL) - return -EINVAL; - - voltage_table->mask_low = 0; - voltage_table->phase_delay = 0; - - voltage_table->count = voltage_dependency_table->count; - for (i = 0; i < voltage_table->count; i++) { - voltage_table->entries[i].value = voltage_dependency_table->entries[i].v; - voltage_table->entries[i].smio_low = 0; - } - - return 0; -} - -static int ci_construct_voltage_tables(struct amdgpu_device *adev) -{ - struct ci_power_info *pi = ci_get_pi(adev); - int ret; - - if (pi->voltage_control == CISLANDS_VOLTAGE_CONTROL_BY_GPIO) { - ret = amdgpu_atombios_get_voltage_table(adev, VOLTAGE_TYPE_VDDC, - VOLTAGE_OBJ_GPIO_LUT, - &pi->vddc_voltage_table); - if (ret) - return ret; - } else if (pi->voltage_control == CISLANDS_VOLTAGE_CONTROL_BY_SVID2) { - ret = ci_get_svi2_voltage_table(adev, - &adev->pm.dpm.dyn_state.vddc_dependency_on_mclk, - &pi->vddc_voltage_table); - if (ret) - return ret; - } - - if (pi->vddc_voltage_table.count > SMU7_MAX_LEVELS_VDDC) - ci_trim_voltage_table_to_fit_state_table(adev, SMU7_MAX_LEVELS_VDDC, - &pi->vddc_voltage_table); - - if (pi->vddci_control == CISLANDS_VOLTAGE_CONTROL_BY_GPIO) { - ret = amdgpu_atombios_get_voltage_table(adev, VOLTAGE_TYPE_VDDCI, - VOLTAGE_OBJ_GPIO_LUT, - &pi->vddci_voltage_table); - if (ret) - return ret; - } else if (pi->vddci_control == CISLANDS_VOLTAGE_CONTROL_BY_SVID2) { - ret = ci_get_svi2_voltage_table(adev, - &adev->pm.dpm.dyn_state.vddci_dependency_on_mclk, - &pi->vddci_voltage_table); - if (ret) - return ret; - } - - if (pi->vddci_voltage_table.count > SMU7_MAX_LEVELS_VDDCI) - ci_trim_voltage_table_to_fit_state_table(adev, SMU7_MAX_LEVELS_VDDCI, - &pi->vddci_voltage_table); - - if (pi->mvdd_control == CISLANDS_VOLTAGE_CONTROL_BY_GPIO) { - ret = amdgpu_atombios_get_voltage_table(adev, VOLTAGE_TYPE_MVDDC, - VOLTAGE_OBJ_GPIO_LUT, - &pi->mvdd_voltage_table); - if (ret) - return ret; - } else if (pi->mvdd_control == CISLANDS_VOLTAGE_CONTROL_BY_SVID2) { - ret = ci_get_svi2_voltage_table(adev, - &adev->pm.dpm.dyn_state.mvdd_dependency_on_mclk, - &pi->mvdd_voltage_table); - if (ret) - return ret; - } - - if (pi->mvdd_voltage_table.count > SMU7_MAX_LEVELS_MVDD) - ci_trim_voltage_table_to_fit_state_table(adev, SMU7_MAX_LEVELS_MVDD, - &pi->mvdd_voltage_table); - - return 0; -} - -static void ci_populate_smc_voltage_table(struct amdgpu_device *adev, - struct atom_voltage_table_entry *voltage_table, - SMU7_Discrete_VoltageLevel *smc_voltage_table) -{ - int ret; - - ret = ci_get_std_voltage_value_sidd(adev, voltage_table, - &smc_voltage_table->StdVoltageHiSidd, - &smc_voltage_table->StdVoltageLoSidd); - - if (ret) { - smc_voltage_table->StdVoltageHiSidd = voltage_table->value * VOLTAGE_SCALE; - smc_voltage_table->StdVoltageLoSidd = voltage_table->value * VOLTAGE_SCALE; - } - - smc_voltage_table->Voltage = cpu_to_be16(voltage_table->value * VOLTAGE_SCALE); - smc_voltage_table->StdVoltageHiSidd = - cpu_to_be16(smc_voltage_table->StdVoltageHiSidd); - smc_voltage_table->StdVoltageLoSidd = - cpu_to_be16(smc_voltage_table->StdVoltageLoSidd); -} - -static int ci_populate_smc_vddc_table(struct amdgpu_device *adev, - SMU7_Discrete_DpmTable *table) -{ - struct ci_power_info *pi = ci_get_pi(adev); - unsigned int count; - - table->VddcLevelCount = pi->vddc_voltage_table.count; - for (count = 0; count < table->VddcLevelCount; count++) { - ci_populate_smc_voltage_table(adev, - &pi->vddc_voltage_table.entries[count], - &table->VddcLevel[count]); - - if (pi->voltage_control == CISLANDS_VOLTAGE_CONTROL_BY_GPIO) - table->VddcLevel[count].Smio |= - pi->vddc_voltage_table.entries[count].smio_low; - else - table->VddcLevel[count].Smio = 0; - } - table->VddcLevelCount = cpu_to_be32(table->VddcLevelCount); - - return 0; -} - -static int ci_populate_smc_vddci_table(struct amdgpu_device *adev, - SMU7_Discrete_DpmTable *table) -{ - unsigned int count; - struct ci_power_info *pi = ci_get_pi(adev); - - table->VddciLevelCount = pi->vddci_voltage_table.count; - for (count = 0; count < table->VddciLevelCount; count++) { - ci_populate_smc_voltage_table(adev, - &pi->vddci_voltage_table.entries[count], - &table->VddciLevel[count]); - - if (pi->vddci_control == CISLANDS_VOLTAGE_CONTROL_BY_GPIO) - table->VddciLevel[count].Smio |= - pi->vddci_voltage_table.entries[count].smio_low; - else - table->VddciLevel[count].Smio = 0; - } - table->VddciLevelCount = cpu_to_be32(table->VddciLevelCount); - - return 0; -} - -static int ci_populate_smc_mvdd_table(struct amdgpu_device *adev, - SMU7_Discrete_DpmTable *table) -{ - struct ci_power_info *pi = ci_get_pi(adev); - unsigned int count; - - table->MvddLevelCount = pi->mvdd_voltage_table.count; - for (count = 0; count < table->MvddLevelCount; count++) { - ci_populate_smc_voltage_table(adev, - &pi->mvdd_voltage_table.entries[count], - &table->MvddLevel[count]); - - if (pi->mvdd_control == CISLANDS_VOLTAGE_CONTROL_BY_GPIO) - table->MvddLevel[count].Smio |= - pi->mvdd_voltage_table.entries[count].smio_low; - else - table->MvddLevel[count].Smio = 0; - } - table->MvddLevelCount = cpu_to_be32(table->MvddLevelCount); - - return 0; -} - -static int ci_populate_smc_voltage_tables(struct amdgpu_device *adev, - SMU7_Discrete_DpmTable *table) -{ - int ret; - - ret = ci_populate_smc_vddc_table(adev, table); - if (ret) - return ret; - - ret = ci_populate_smc_vddci_table(adev, table); - if (ret) - return ret; - - ret = ci_populate_smc_mvdd_table(adev, table); - if (ret) - return ret; - - return 0; -} - -static int ci_populate_mvdd_value(struct amdgpu_device *adev, u32 mclk, - SMU7_Discrete_VoltageLevel *voltage) -{ - struct ci_power_info *pi = ci_get_pi(adev); - u32 i = 0; - - if (pi->mvdd_control != CISLANDS_VOLTAGE_CONTROL_NONE) { - for (i = 0; i < adev->pm.dpm.dyn_state.mvdd_dependency_on_mclk.count; i++) { - if (mclk <= adev->pm.dpm.dyn_state.mvdd_dependency_on_mclk.entries[i].clk) { - voltage->Voltage = pi->mvdd_voltage_table.entries[i].value; - break; - } - } - - if (i >= adev->pm.dpm.dyn_state.mvdd_dependency_on_mclk.count) - return -EINVAL; - } - - return -EINVAL; -} - -static int ci_get_std_voltage_value_sidd(struct amdgpu_device *adev, - struct atom_voltage_table_entry *voltage_table, - u16 *std_voltage_hi_sidd, u16 *std_voltage_lo_sidd) -{ - u16 v_index, idx; - bool voltage_found = false; - *std_voltage_hi_sidd = voltage_table->value * VOLTAGE_SCALE; - *std_voltage_lo_sidd = voltage_table->value * VOLTAGE_SCALE; - - if (adev->pm.dpm.dyn_state.vddc_dependency_on_sclk.entries == NULL) - return -EINVAL; - - if (adev->pm.dpm.dyn_state.cac_leakage_table.entries) { - for (v_index = 0; (u32)v_index < adev->pm.dpm.dyn_state.vddc_dependency_on_sclk.count; v_index++) { - if (voltage_table->value == - adev->pm.dpm.dyn_state.vddc_dependency_on_sclk.entries[v_index].v) { - voltage_found = true; - if ((u32)v_index < adev->pm.dpm.dyn_state.cac_leakage_table.count) - idx = v_index; - else - idx = adev->pm.dpm.dyn_state.cac_leakage_table.count - 1; - *std_voltage_lo_sidd = - adev->pm.dpm.dyn_state.cac_leakage_table.entries[idx].vddc * VOLTAGE_SCALE; - *std_voltage_hi_sidd = - adev->pm.dpm.dyn_state.cac_leakage_table.entries[idx].leakage * VOLTAGE_SCALE; - break; - } - } - - if (!voltage_found) { - for (v_index = 0; (u32)v_index < adev->pm.dpm.dyn_state.vddc_dependency_on_sclk.count; v_index++) { - if (voltage_table->value <= - adev->pm.dpm.dyn_state.vddc_dependency_on_sclk.entries[v_index].v) { - voltage_found = true; - if ((u32)v_index < adev->pm.dpm.dyn_state.cac_leakage_table.count) - idx = v_index; - else - idx = adev->pm.dpm.dyn_state.cac_leakage_table.count - 1; - *std_voltage_lo_sidd = - adev->pm.dpm.dyn_state.cac_leakage_table.entries[idx].vddc * VOLTAGE_SCALE; - *std_voltage_hi_sidd = - adev->pm.dpm.dyn_state.cac_leakage_table.entries[idx].leakage * VOLTAGE_SCALE; - break; - } - } - } - } - - return 0; -} - -static void ci_populate_phase_value_based_on_sclk(struct amdgpu_device *adev, - const struct amdgpu_phase_shedding_limits_table *limits, - u32 sclk, - u32 *phase_shedding) -{ - unsigned int i; - - *phase_shedding = 1; - - for (i = 0; i < limits->count; i++) { - if (sclk < limits->entries[i].sclk) { - *phase_shedding = i; - break; - } - } -} - -static void ci_populate_phase_value_based_on_mclk(struct amdgpu_device *adev, - const struct amdgpu_phase_shedding_limits_table *limits, - u32 mclk, - u32 *phase_shedding) -{ - unsigned int i; - - *phase_shedding = 1; - - for (i = 0; i < limits->count; i++) { - if (mclk < limits->entries[i].mclk) { - *phase_shedding = i; - break; - } - } -} - -static int ci_init_arb_table_index(struct amdgpu_device *adev) -{ - struct ci_power_info *pi = ci_get_pi(adev); - u32 tmp; - int ret; - - ret = amdgpu_ci_read_smc_sram_dword(adev, pi->arb_table_start, - &tmp, pi->sram_end); - if (ret) - return ret; - - tmp &= 0x00FFFFFF; - tmp |= MC_CG_ARB_FREQ_F1 << 24; - - return amdgpu_ci_write_smc_sram_dword(adev, pi->arb_table_start, - tmp, pi->sram_end); -} - -static int ci_get_dependency_volt_by_clk(struct amdgpu_device *adev, - struct amdgpu_clock_voltage_dependency_table *allowed_clock_voltage_table, - u32 clock, u32 *voltage) -{ - u32 i = 0; - - if (allowed_clock_voltage_table->count == 0) - return -EINVAL; - - for (i = 0; i < allowed_clock_voltage_table->count; i++) { - if (allowed_clock_voltage_table->entries[i].clk >= clock) { - *voltage = allowed_clock_voltage_table->entries[i].v; - return 0; - } - } - - *voltage = allowed_clock_voltage_table->entries[i-1].v; - - return 0; -} - -static u8 ci_get_sleep_divider_id_from_clock(u32 sclk, u32 min_sclk_in_sr) -{ - u32 i; - u32 tmp; - u32 min = max(min_sclk_in_sr, (u32)CISLAND_MINIMUM_ENGINE_CLOCK); - - if (sclk < min) - return 0; - - for (i = CISLAND_MAX_DEEPSLEEP_DIVIDER_ID; ; i--) { - tmp = sclk >> i; - if (tmp >= min || i == 0) - break; - } - - return (u8)i; -} - -static int ci_initial_switch_from_arb_f0_to_f1(struct amdgpu_device *adev) -{ - return ci_copy_and_switch_arb_sets(adev, MC_CG_ARB_FREQ_F0, MC_CG_ARB_FREQ_F1); -} - -static int ci_reset_to_default(struct amdgpu_device *adev) -{ - return (amdgpu_ci_send_msg_to_smc(adev, PPSMC_MSG_ResetToDefaults) == PPSMC_Result_OK) ? - 0 : -EINVAL; -} - -static int ci_force_switch_to_arb_f0(struct amdgpu_device *adev) -{ - u32 tmp; - - tmp = (RREG32_SMC(ixSMC_SCRATCH9) & 0x0000ff00) >> 8; - - if (tmp == MC_CG_ARB_FREQ_F0) - return 0; - - return ci_copy_and_switch_arb_sets(adev, tmp, MC_CG_ARB_FREQ_F0); -} - -static void ci_register_patching_mc_arb(struct amdgpu_device *adev, - const u32 engine_clock, - const u32 memory_clock, - u32 *dram_timimg2) -{ - bool patch; - u32 tmp, tmp2; - - tmp = RREG32(mmMC_SEQ_MISC0); - patch = ((tmp & 0x0000f00) == 0x300) ? true : false; - - if (patch && - ((adev->pdev->device == 0x67B0) || - (adev->pdev->device == 0x67B1))) { - if ((memory_clock > 100000) && (memory_clock <= 125000)) { - tmp2 = (((0x31 * engine_clock) / 125000) - 1) & 0xff; - *dram_timimg2 &= ~0x00ff0000; - *dram_timimg2 |= tmp2 << 16; - } else if ((memory_clock > 125000) && (memory_clock <= 137500)) { - tmp2 = (((0x36 * engine_clock) / 137500) - 1) & 0xff; - *dram_timimg2 &= ~0x00ff0000; - *dram_timimg2 |= tmp2 << 16; - } - } -} - -static int ci_populate_memory_timing_parameters(struct amdgpu_device *adev, - u32 sclk, - u32 mclk, - SMU7_Discrete_MCArbDramTimingTableEntry *arb_regs) -{ - u32 dram_timing; - u32 dram_timing2; - u32 burst_time; - - amdgpu_atombios_set_engine_dram_timings(adev, sclk, mclk); - - dram_timing = RREG32(mmMC_ARB_DRAM_TIMING); - dram_timing2 = RREG32(mmMC_ARB_DRAM_TIMING2); - burst_time = RREG32(mmMC_ARB_BURST_TIME) & MC_ARB_BURST_TIME__STATE0_MASK; - - ci_register_patching_mc_arb(adev, sclk, mclk, &dram_timing2); - - arb_regs->McArbDramTiming = cpu_to_be32(dram_timing); - arb_regs->McArbDramTiming2 = cpu_to_be32(dram_timing2); - arb_regs->McArbBurstTime = (u8)burst_time; - - return 0; -} - -static int ci_do_program_memory_timing_parameters(struct amdgpu_device *adev) -{ - struct ci_power_info *pi = ci_get_pi(adev); - SMU7_Discrete_MCArbDramTimingTable arb_regs; - u32 i, j; - int ret = 0; - - memset(&arb_regs, 0, sizeof(SMU7_Discrete_MCArbDramTimingTable)); - - for (i = 0; i < pi->dpm_table.sclk_table.count; i++) { - for (j = 0; j < pi->dpm_table.mclk_table.count; j++) { - ret = ci_populate_memory_timing_parameters(adev, - pi->dpm_table.sclk_table.dpm_levels[i].value, - pi->dpm_table.mclk_table.dpm_levels[j].value, - &arb_regs.entries[i][j]); - if (ret) - break; - } - } - - if (ret == 0) - ret = amdgpu_ci_copy_bytes_to_smc(adev, - pi->arb_table_start, - (u8 *)&arb_regs, - sizeof(SMU7_Discrete_MCArbDramTimingTable), - pi->sram_end); - - return ret; -} - -static int ci_program_memory_timing_parameters(struct amdgpu_device *adev) -{ - struct ci_power_info *pi = ci_get_pi(adev); - - if (pi->need_update_smu7_dpm_table == 0) - return 0; - - return ci_do_program_memory_timing_parameters(adev); -} - -static void ci_populate_smc_initial_state(struct amdgpu_device *adev, - struct amdgpu_ps *amdgpu_boot_state) -{ - struct ci_ps *boot_state = ci_get_ps(amdgpu_boot_state); - struct ci_power_info *pi = ci_get_pi(adev); - u32 level = 0; - - for (level = 0; level < adev->pm.dpm.dyn_state.vddc_dependency_on_sclk.count; level++) { - if (adev->pm.dpm.dyn_state.vddc_dependency_on_sclk.entries[level].clk >= - boot_state->performance_levels[0].sclk) { - pi->smc_state_table.GraphicsBootLevel = level; - break; - } - } - - for (level = 0; level < adev->pm.dpm.dyn_state.vddc_dependency_on_mclk.count; level++) { - if (adev->pm.dpm.dyn_state.vddc_dependency_on_mclk.entries[level].clk >= - boot_state->performance_levels[0].mclk) { - pi->smc_state_table.MemoryBootLevel = level; - break; - } - } -} - -static u32 ci_get_dpm_level_enable_mask_value(struct ci_single_dpm_table *dpm_table) -{ - u32 i; - u32 mask_value = 0; - - for (i = dpm_table->count; i > 0; i--) { - mask_value = mask_value << 1; - if (dpm_table->dpm_levels[i-1].enabled) - mask_value |= 0x1; - else - mask_value &= 0xFFFFFFFE; - } - - return mask_value; -} - -static void ci_populate_smc_link_level(struct amdgpu_device *adev, - SMU7_Discrete_DpmTable *table) -{ - struct ci_power_info *pi = ci_get_pi(adev); - struct ci_dpm_table *dpm_table = &pi->dpm_table; - u32 i; - - for (i = 0; i < dpm_table->pcie_speed_table.count; i++) { - table->LinkLevel[i].PcieGenSpeed = - (u8)dpm_table->pcie_speed_table.dpm_levels[i].value; - table->LinkLevel[i].PcieLaneCount = - amdgpu_encode_pci_lane_width(dpm_table->pcie_speed_table.dpm_levels[i].param1); - table->LinkLevel[i].EnabledForActivity = 1; - table->LinkLevel[i].DownT = cpu_to_be32(5); - table->LinkLevel[i].UpT = cpu_to_be32(30); - } - - pi->smc_state_table.LinkLevelCount = (u8)dpm_table->pcie_speed_table.count; - pi->dpm_level_enable_mask.pcie_dpm_enable_mask = - ci_get_dpm_level_enable_mask_value(&dpm_table->pcie_speed_table); -} - -static int ci_populate_smc_uvd_level(struct amdgpu_device *adev, - SMU7_Discrete_DpmTable *table) -{ - u32 count; - struct atom_clock_dividers dividers; - int ret = -EINVAL; - - table->UvdLevelCount = - adev->pm.dpm.dyn_state.uvd_clock_voltage_dependency_table.count; - - for (count = 0; count < table->UvdLevelCount; count++) { - table->UvdLevel[count].VclkFrequency = - adev->pm.dpm.dyn_state.uvd_clock_voltage_dependency_table.entries[count].vclk; - table->UvdLevel[count].DclkFrequency = - adev->pm.dpm.dyn_state.uvd_clock_voltage_dependency_table.entries[count].dclk; - table->UvdLevel[count].MinVddc = - adev->pm.dpm.dyn_state.uvd_clock_voltage_dependency_table.entries[count].v * VOLTAGE_SCALE; - table->UvdLevel[count].MinVddcPhases = 1; - - ret = amdgpu_atombios_get_clock_dividers(adev, - COMPUTE_GPUCLK_INPUT_FLAG_DEFAULT_GPUCLK, - table->UvdLevel[count].VclkFrequency, false, ÷rs); - if (ret) - return ret; - - table->UvdLevel[count].VclkDivider = (u8)dividers.post_divider; - - ret = amdgpu_atombios_get_clock_dividers(adev, - COMPUTE_GPUCLK_INPUT_FLAG_DEFAULT_GPUCLK, - table->UvdLevel[count].DclkFrequency, false, ÷rs); - if (ret) - return ret; - - table->UvdLevel[count].DclkDivider = (u8)dividers.post_divider; - - table->UvdLevel[count].VclkFrequency = cpu_to_be32(table->UvdLevel[count].VclkFrequency); - table->UvdLevel[count].DclkFrequency = cpu_to_be32(table->UvdLevel[count].DclkFrequency); - table->UvdLevel[count].MinVddc = cpu_to_be16(table->UvdLevel[count].MinVddc); - } - - return ret; -} - -static int ci_populate_smc_vce_level(struct amdgpu_device *adev, - SMU7_Discrete_DpmTable *table) -{ - u32 count; - struct atom_clock_dividers dividers; - int ret = -EINVAL; - - table->VceLevelCount = - adev->pm.dpm.dyn_state.vce_clock_voltage_dependency_table.count; - - for (count = 0; count < table->VceLevelCount; count++) { - table->VceLevel[count].Frequency = - adev->pm.dpm.dyn_state.vce_clock_voltage_dependency_table.entries[count].evclk; - table->VceLevel[count].MinVoltage = - (u16)adev->pm.dpm.dyn_state.vce_clock_voltage_dependency_table.entries[count].v * VOLTAGE_SCALE; - table->VceLevel[count].MinPhases = 1; - - ret = amdgpu_atombios_get_clock_dividers(adev, - COMPUTE_GPUCLK_INPUT_FLAG_DEFAULT_GPUCLK, - table->VceLevel[count].Frequency, false, ÷rs); - if (ret) - return ret; - - table->VceLevel[count].Divider = (u8)dividers.post_divider; - - table->VceLevel[count].Frequency = cpu_to_be32(table->VceLevel[count].Frequency); - table->VceLevel[count].MinVoltage = cpu_to_be16(table->VceLevel[count].MinVoltage); - } - - return ret; - -} - -static int ci_populate_smc_acp_level(struct amdgpu_device *adev, - SMU7_Discrete_DpmTable *table) -{ - u32 count; - struct atom_clock_dividers dividers; - int ret = -EINVAL; - - table->AcpLevelCount = (u8) - (adev->pm.dpm.dyn_state.acp_clock_voltage_dependency_table.count); - - for (count = 0; count < table->AcpLevelCount; count++) { - table->AcpLevel[count].Frequency = - adev->pm.dpm.dyn_state.acp_clock_voltage_dependency_table.entries[count].clk; - table->AcpLevel[count].MinVoltage = - adev->pm.dpm.dyn_state.acp_clock_voltage_dependency_table.entries[count].v; - table->AcpLevel[count].MinPhases = 1; - - ret = amdgpu_atombios_get_clock_dividers(adev, - COMPUTE_GPUCLK_INPUT_FLAG_DEFAULT_GPUCLK, - table->AcpLevel[count].Frequency, false, ÷rs); - if (ret) - return ret; - - table->AcpLevel[count].Divider = (u8)dividers.post_divider; - - table->AcpLevel[count].Frequency = cpu_to_be32(table->AcpLevel[count].Frequency); - table->AcpLevel[count].MinVoltage = cpu_to_be16(table->AcpLevel[count].MinVoltage); - } - - return ret; -} - -static int ci_populate_smc_samu_level(struct amdgpu_device *adev, - SMU7_Discrete_DpmTable *table) -{ - u32 count; - struct atom_clock_dividers dividers; - int ret = -EINVAL; - - table->SamuLevelCount = - adev->pm.dpm.dyn_state.samu_clock_voltage_dependency_table.count; - - for (count = 0; count < table->SamuLevelCount; count++) { - table->SamuLevel[count].Frequency = - adev->pm.dpm.dyn_state.samu_clock_voltage_dependency_table.entries[count].clk; - table->SamuLevel[count].MinVoltage = - adev->pm.dpm.dyn_state.samu_clock_voltage_dependency_table.entries[count].v * VOLTAGE_SCALE; - table->SamuLevel[count].MinPhases = 1; - - ret = amdgpu_atombios_get_clock_dividers(adev, - COMPUTE_GPUCLK_INPUT_FLAG_DEFAULT_GPUCLK, - table->SamuLevel[count].Frequency, false, ÷rs); - if (ret) - return ret; - - table->SamuLevel[count].Divider = (u8)dividers.post_divider; - - table->SamuLevel[count].Frequency = cpu_to_be32(table->SamuLevel[count].Frequency); - table->SamuLevel[count].MinVoltage = cpu_to_be16(table->SamuLevel[count].MinVoltage); - } - - return ret; -} - -static int ci_calculate_mclk_params(struct amdgpu_device *adev, - u32 memory_clock, - SMU7_Discrete_MemoryLevel *mclk, - bool strobe_mode, - bool dll_state_on) -{ - struct ci_power_info *pi = ci_get_pi(adev); - u32 dll_cntl = pi->clock_registers.dll_cntl; - u32 mclk_pwrmgt_cntl = pi->clock_registers.mclk_pwrmgt_cntl; - u32 mpll_ad_func_cntl = pi->clock_registers.mpll_ad_func_cntl; - u32 mpll_dq_func_cntl = pi->clock_registers.mpll_dq_func_cntl; - u32 mpll_func_cntl = pi->clock_registers.mpll_func_cntl; - u32 mpll_func_cntl_1 = pi->clock_registers.mpll_func_cntl_1; - u32 mpll_func_cntl_2 = pi->clock_registers.mpll_func_cntl_2; - u32 mpll_ss1 = pi->clock_registers.mpll_ss1; - u32 mpll_ss2 = pi->clock_registers.mpll_ss2; - struct atom_mpll_param mpll_param; - int ret; - - ret = amdgpu_atombios_get_memory_pll_dividers(adev, memory_clock, strobe_mode, &mpll_param); - if (ret) - return ret; - - mpll_func_cntl &= ~MPLL_FUNC_CNTL__BWCTRL_MASK; - mpll_func_cntl |= (mpll_param.bwcntl << MPLL_FUNC_CNTL__BWCTRL__SHIFT); - - mpll_func_cntl_1 &= ~(MPLL_FUNC_CNTL_1__CLKF_MASK | MPLL_FUNC_CNTL_1__CLKFRAC_MASK | - MPLL_FUNC_CNTL_1__VCO_MODE_MASK); - mpll_func_cntl_1 |= (mpll_param.clkf) << MPLL_FUNC_CNTL_1__CLKF__SHIFT | - (mpll_param.clkfrac << MPLL_FUNC_CNTL_1__CLKFRAC__SHIFT) | - (mpll_param.vco_mode << MPLL_FUNC_CNTL_1__VCO_MODE__SHIFT); - - mpll_ad_func_cntl &= ~MPLL_AD_FUNC_CNTL__YCLK_POST_DIV_MASK; - mpll_ad_func_cntl |= (mpll_param.post_div << MPLL_AD_FUNC_CNTL__YCLK_POST_DIV__SHIFT); - - if (adev->gmc.vram_type == AMDGPU_VRAM_TYPE_GDDR5) { - mpll_dq_func_cntl &= ~(MPLL_DQ_FUNC_CNTL__YCLK_SEL_MASK | - MPLL_AD_FUNC_CNTL__YCLK_POST_DIV_MASK); - mpll_dq_func_cntl |= (mpll_param.yclk_sel << MPLL_DQ_FUNC_CNTL__YCLK_SEL__SHIFT) | - (mpll_param.post_div << MPLL_AD_FUNC_CNTL__YCLK_POST_DIV__SHIFT); - } - - if (pi->caps_mclk_ss_support) { - struct amdgpu_atom_ss ss; - u32 freq_nom; - u32 tmp; - u32 reference_clock = adev->clock.mpll.reference_freq; - - if (mpll_param.qdr == 1) - freq_nom = memory_clock * 4 * (1 << mpll_param.post_div); - else - freq_nom = memory_clock * 2 * (1 << mpll_param.post_div); - - tmp = (freq_nom / reference_clock); - tmp = tmp * tmp; - if (amdgpu_atombios_get_asic_ss_info(adev, &ss, - ASIC_INTERNAL_MEMORY_SS, freq_nom)) { - u32 clks = reference_clock * 5 / ss.rate; - u32 clkv = (u32)((((131 * ss.percentage * ss.rate) / 100) * tmp) / freq_nom); - - mpll_ss1 &= ~MPLL_SS1__CLKV_MASK; - mpll_ss1 |= (clkv << MPLL_SS1__CLKV__SHIFT); - - mpll_ss2 &= ~MPLL_SS2__CLKS_MASK; - mpll_ss2 |= (clks << MPLL_SS2__CLKS__SHIFT); - } - } - - mclk_pwrmgt_cntl &= ~MCLK_PWRMGT_CNTL__DLL_SPEED_MASK; - mclk_pwrmgt_cntl |= (mpll_param.dll_speed << MCLK_PWRMGT_CNTL__DLL_SPEED__SHIFT); - - if (dll_state_on) - mclk_pwrmgt_cntl |= MCLK_PWRMGT_CNTL__MRDCK0_PDNB_MASK | - MCLK_PWRMGT_CNTL__MRDCK1_PDNB_MASK; - else - mclk_pwrmgt_cntl &= ~(MCLK_PWRMGT_CNTL__MRDCK0_PDNB_MASK | - MCLK_PWRMGT_CNTL__MRDCK1_PDNB_MASK); - - mclk->MclkFrequency = memory_clock; - mclk->MpllFuncCntl = mpll_func_cntl; - mclk->MpllFuncCntl_1 = mpll_func_cntl_1; - mclk->MpllFuncCntl_2 = mpll_func_cntl_2; - mclk->MpllAdFuncCntl = mpll_ad_func_cntl; - mclk->MpllDqFuncCntl = mpll_dq_func_cntl; - mclk->MclkPwrmgtCntl = mclk_pwrmgt_cntl; - mclk->DllCntl = dll_cntl; - mclk->MpllSs1 = mpll_ss1; - mclk->MpllSs2 = mpll_ss2; - - return 0; -} - -static int ci_populate_single_memory_level(struct amdgpu_device *adev, - u32 memory_clock, - SMU7_Discrete_MemoryLevel *memory_level) -{ - struct ci_power_info *pi = ci_get_pi(adev); - int ret; - bool dll_state_on; - - if (adev->pm.dpm.dyn_state.vddc_dependency_on_mclk.entries) { - ret = ci_get_dependency_volt_by_clk(adev, - &adev->pm.dpm.dyn_state.vddc_dependency_on_mclk, - memory_clock, &memory_level->MinVddc); - if (ret) - return ret; - } - - if (adev->pm.dpm.dyn_state.vddci_dependency_on_mclk.entries) { - ret = ci_get_dependency_volt_by_clk(adev, - &adev->pm.dpm.dyn_state.vddci_dependency_on_mclk, - memory_clock, &memory_level->MinVddci); - if (ret) - return ret; - } - - if (adev->pm.dpm.dyn_state.mvdd_dependency_on_mclk.entries) { - ret = ci_get_dependency_volt_by_clk(adev, - &adev->pm.dpm.dyn_state.mvdd_dependency_on_mclk, - memory_clock, &memory_level->MinMvdd); - if (ret) - return ret; - } - - memory_level->MinVddcPhases = 1; - - if (pi->vddc_phase_shed_control) - ci_populate_phase_value_based_on_mclk(adev, - &adev->pm.dpm.dyn_state.phase_shedding_limits_table, - memory_clock, - &memory_level->MinVddcPhases); - - memory_level->EnabledForActivity = 1; - memory_level->EnabledForThrottle = 1; - memory_level->UpH = 0; - memory_level->DownH = 100; - memory_level->VoltageDownH = 0; - memory_level->ActivityLevel = (u16)pi->mclk_activity_target; - - memory_level->StutterEnable = false; - memory_level->StrobeEnable = false; - memory_level->EdcReadEnable = false; - memory_level->EdcWriteEnable = false; - memory_level->RttEnable = false; - - memory_level->DisplayWatermark = PPSMC_DISPLAY_WATERMARK_LOW; - - if (pi->mclk_stutter_mode_threshold && - (memory_clock <= pi->mclk_stutter_mode_threshold) && - (!pi->uvd_enabled) && - (RREG32(mmDPG_PIPE_STUTTER_CONTROL) & DPG_PIPE_STUTTER_CONTROL__STUTTER_ENABLE_MASK) && - (adev->pm.dpm.new_active_crtc_count <= 2)) - memory_level->StutterEnable = true; - - if (pi->mclk_strobe_mode_threshold && - (memory_clock <= pi->mclk_strobe_mode_threshold)) - memory_level->StrobeEnable = 1; - - if (adev->gmc.vram_type == AMDGPU_VRAM_TYPE_GDDR5) { - memory_level->StrobeRatio = - ci_get_mclk_frequency_ratio(memory_clock, memory_level->StrobeEnable); - if (pi->mclk_edc_enable_threshold && - (memory_clock > pi->mclk_edc_enable_threshold)) - memory_level->EdcReadEnable = true; - - if (pi->mclk_edc_wr_enable_threshold && - (memory_clock > pi->mclk_edc_wr_enable_threshold)) - memory_level->EdcWriteEnable = true; - - if (memory_level->StrobeEnable) { - if (ci_get_mclk_frequency_ratio(memory_clock, true) >= - ((RREG32(mmMC_SEQ_MISC7) >> 16) & 0xf)) - dll_state_on = ((RREG32(mmMC_SEQ_MISC5) >> 1) & 0x1) ? true : false; - else - dll_state_on = ((RREG32(mmMC_SEQ_MISC6) >> 1) & 0x1) ? true : false; - } else { - dll_state_on = pi->dll_default_on; - } - } else { - memory_level->StrobeRatio = ci_get_ddr3_mclk_frequency_ratio(memory_clock); - dll_state_on = ((RREG32(mmMC_SEQ_MISC5) >> 1) & 0x1) ? true : false; - } - - ret = ci_calculate_mclk_params(adev, memory_clock, memory_level, memory_level->StrobeEnable, dll_state_on); - if (ret) - return ret; - - memory_level->MinVddc = cpu_to_be32(memory_level->MinVddc * VOLTAGE_SCALE); - memory_level->MinVddcPhases = cpu_to_be32(memory_level->MinVddcPhases); - memory_level->MinVddci = cpu_to_be32(memory_level->MinVddci * VOLTAGE_SCALE); - memory_level->MinMvdd = cpu_to_be32(memory_level->MinMvdd * VOLTAGE_SCALE); - - memory_level->MclkFrequency = cpu_to_be32(memory_level->MclkFrequency); - memory_level->ActivityLevel = cpu_to_be16(memory_level->ActivityLevel); - memory_level->MpllFuncCntl = cpu_to_be32(memory_level->MpllFuncCntl); - memory_level->MpllFuncCntl_1 = cpu_to_be32(memory_level->MpllFuncCntl_1); - memory_level->MpllFuncCntl_2 = cpu_to_be32(memory_level->MpllFuncCntl_2); - memory_level->MpllAdFuncCntl = cpu_to_be32(memory_level->MpllAdFuncCntl); - memory_level->MpllDqFuncCntl = cpu_to_be32(memory_level->MpllDqFuncCntl); - memory_level->MclkPwrmgtCntl = cpu_to_be32(memory_level->MclkPwrmgtCntl); - memory_level->DllCntl = cpu_to_be32(memory_level->DllCntl); - memory_level->MpllSs1 = cpu_to_be32(memory_level->MpllSs1); - memory_level->MpllSs2 = cpu_to_be32(memory_level->MpllSs2); - - return 0; -} - -static int ci_populate_smc_acpi_level(struct amdgpu_device *adev, - SMU7_Discrete_DpmTable *table) -{ - struct ci_power_info *pi = ci_get_pi(adev); - struct atom_clock_dividers dividers; - SMU7_Discrete_VoltageLevel voltage_level; - u32 spll_func_cntl = pi->clock_registers.cg_spll_func_cntl; - u32 spll_func_cntl_2 = pi->clock_registers.cg_spll_func_cntl_2; - u32 dll_cntl = pi->clock_registers.dll_cntl; - u32 mclk_pwrmgt_cntl = pi->clock_registers.mclk_pwrmgt_cntl; - int ret; - - table->ACPILevel.Flags &= ~PPSMC_SWSTATE_FLAG_DC; - - if (pi->acpi_vddc) - table->ACPILevel.MinVddc = cpu_to_be32(pi->acpi_vddc * VOLTAGE_SCALE); - else - table->ACPILevel.MinVddc = cpu_to_be32(pi->min_vddc_in_pp_table * VOLTAGE_SCALE); - - table->ACPILevel.MinVddcPhases = pi->vddc_phase_shed_control ? 0 : 1; - - table->ACPILevel.SclkFrequency = adev->clock.spll.reference_freq; - - ret = amdgpu_atombios_get_clock_dividers(adev, - COMPUTE_GPUCLK_INPUT_FLAG_SCLK, - table->ACPILevel.SclkFrequency, false, ÷rs); - if (ret) - return ret; - - table->ACPILevel.SclkDid = (u8)dividers.post_divider; - table->ACPILevel.DisplayWatermark = PPSMC_DISPLAY_WATERMARK_LOW; - table->ACPILevel.DeepSleepDivId = 0; - - spll_func_cntl &= ~CG_SPLL_FUNC_CNTL__SPLL_PWRON_MASK; - spll_func_cntl |= CG_SPLL_FUNC_CNTL__SPLL_RESET_MASK; - - spll_func_cntl_2 &= ~CG_SPLL_FUNC_CNTL_2__SCLK_MUX_SEL_MASK; - spll_func_cntl_2 |= (4 << CG_SPLL_FUNC_CNTL_2__SCLK_MUX_SEL__SHIFT); - - table->ACPILevel.CgSpllFuncCntl = spll_func_cntl; - table->ACPILevel.CgSpllFuncCntl2 = spll_func_cntl_2; - table->ACPILevel.CgSpllFuncCntl3 = pi->clock_registers.cg_spll_func_cntl_3; - table->ACPILevel.CgSpllFuncCntl4 = pi->clock_registers.cg_spll_func_cntl_4; - table->ACPILevel.SpllSpreadSpectrum = pi->clock_registers.cg_spll_spread_spectrum; - table->ACPILevel.SpllSpreadSpectrum2 = pi->clock_registers.cg_spll_spread_spectrum_2; - table->ACPILevel.CcPwrDynRm = 0; - table->ACPILevel.CcPwrDynRm1 = 0; - - table->ACPILevel.Flags = cpu_to_be32(table->ACPILevel.Flags); - table->ACPILevel.MinVddcPhases = cpu_to_be32(table->ACPILevel.MinVddcPhases); - table->ACPILevel.SclkFrequency = cpu_to_be32(table->ACPILevel.SclkFrequency); - table->ACPILevel.CgSpllFuncCntl = cpu_to_be32(table->ACPILevel.CgSpllFuncCntl); - table->ACPILevel.CgSpllFuncCntl2 = cpu_to_be32(table->ACPILevel.CgSpllFuncCntl2); - table->ACPILevel.CgSpllFuncCntl3 = cpu_to_be32(table->ACPILevel.CgSpllFuncCntl3); - table->ACPILevel.CgSpllFuncCntl4 = cpu_to_be32(table->ACPILevel.CgSpllFuncCntl4); - table->ACPILevel.SpllSpreadSpectrum = cpu_to_be32(table->ACPILevel.SpllSpreadSpectrum); - table->ACPILevel.SpllSpreadSpectrum2 = cpu_to_be32(table->ACPILevel.SpllSpreadSpectrum2); - table->ACPILevel.CcPwrDynRm = cpu_to_be32(table->ACPILevel.CcPwrDynRm); - table->ACPILevel.CcPwrDynRm1 = cpu_to_be32(table->ACPILevel.CcPwrDynRm1); - - table->MemoryACPILevel.MinVddc = table->ACPILevel.MinVddc; - table->MemoryACPILevel.MinVddcPhases = table->ACPILevel.MinVddcPhases; - - if (pi->vddci_control != CISLANDS_VOLTAGE_CONTROL_NONE) { - if (pi->acpi_vddci) - table->MemoryACPILevel.MinVddci = - cpu_to_be32(pi->acpi_vddci * VOLTAGE_SCALE); - else - table->MemoryACPILevel.MinVddci = - cpu_to_be32(pi->min_vddci_in_pp_table * VOLTAGE_SCALE); - } - - if (ci_populate_mvdd_value(adev, 0, &voltage_level)) - table->MemoryACPILevel.MinMvdd = 0; - else - table->MemoryACPILevel.MinMvdd = - cpu_to_be32(voltage_level.Voltage * VOLTAGE_SCALE); - - mclk_pwrmgt_cntl |= MCLK_PWRMGT_CNTL__MRDCK0_RESET_MASK | - MCLK_PWRMGT_CNTL__MRDCK1_RESET_MASK; - mclk_pwrmgt_cntl &= ~(MCLK_PWRMGT_CNTL__MRDCK0_PDNB_MASK | - MCLK_PWRMGT_CNTL__MRDCK1_PDNB_MASK); - - dll_cntl &= ~(DLL_CNTL__MRDCK0_BYPASS_MASK | DLL_CNTL__MRDCK1_BYPASS_MASK); - - table->MemoryACPILevel.DllCntl = cpu_to_be32(dll_cntl); - table->MemoryACPILevel.MclkPwrmgtCntl = cpu_to_be32(mclk_pwrmgt_cntl); - table->MemoryACPILevel.MpllAdFuncCntl = - cpu_to_be32(pi->clock_registers.mpll_ad_func_cntl); - table->MemoryACPILevel.MpllDqFuncCntl = - cpu_to_be32(pi->clock_registers.mpll_dq_func_cntl); - table->MemoryACPILevel.MpllFuncCntl = - cpu_to_be32(pi->clock_registers.mpll_func_cntl); - table->MemoryACPILevel.MpllFuncCntl_1 = - cpu_to_be32(pi->clock_registers.mpll_func_cntl_1); - table->MemoryACPILevel.MpllFuncCntl_2 = - cpu_to_be32(pi->clock_registers.mpll_func_cntl_2); - table->MemoryACPILevel.MpllSs1 = cpu_to_be32(pi->clock_registers.mpll_ss1); - table->MemoryACPILevel.MpllSs2 = cpu_to_be32(pi->clock_registers.mpll_ss2); - - table->MemoryACPILevel.EnabledForThrottle = 0; - table->MemoryACPILevel.EnabledForActivity = 0; - table->MemoryACPILevel.UpH = 0; - table->MemoryACPILevel.DownH = 100; - table->MemoryACPILevel.VoltageDownH = 0; - table->MemoryACPILevel.ActivityLevel = - cpu_to_be16((u16)pi->mclk_activity_target); - - table->MemoryACPILevel.StutterEnable = false; - table->MemoryACPILevel.StrobeEnable = false; - table->MemoryACPILevel.EdcReadEnable = false; - table->MemoryACPILevel.EdcWriteEnable = false; - table->MemoryACPILevel.RttEnable = false; - - return 0; -} - - -static int ci_enable_ulv(struct amdgpu_device *adev, bool enable) -{ - struct ci_power_info *pi = ci_get_pi(adev); - struct ci_ulv_parm *ulv = &pi->ulv; - - if (ulv->supported) { - if (enable) - return (amdgpu_ci_send_msg_to_smc(adev, PPSMC_MSG_EnableULV) == PPSMC_Result_OK) ? - 0 : -EINVAL; - else - return (amdgpu_ci_send_msg_to_smc(adev, PPSMC_MSG_DisableULV) == PPSMC_Result_OK) ? - 0 : -EINVAL; - } - - return 0; -} - -static int ci_populate_ulv_level(struct amdgpu_device *adev, - SMU7_Discrete_Ulv *state) -{ - struct ci_power_info *pi = ci_get_pi(adev); - u16 ulv_voltage = adev->pm.dpm.backbias_response_time; - - state->CcPwrDynRm = 0; - state->CcPwrDynRm1 = 0; - - if (ulv_voltage == 0) { - pi->ulv.supported = false; - return 0; - } - - if (pi->voltage_control != CISLANDS_VOLTAGE_CONTROL_BY_SVID2) { - if (ulv_voltage > adev->pm.dpm.dyn_state.vddc_dependency_on_sclk.entries[0].v) - state->VddcOffset = 0; - else - state->VddcOffset = - adev->pm.dpm.dyn_state.vddc_dependency_on_sclk.entries[0].v - ulv_voltage; - } else { - if (ulv_voltage > adev->pm.dpm.dyn_state.vddc_dependency_on_sclk.entries[0].v) - state->VddcOffsetVid = 0; - else - state->VddcOffsetVid = (u8) - ((adev->pm.dpm.dyn_state.vddc_dependency_on_sclk.entries[0].v - ulv_voltage) * - VOLTAGE_VID_OFFSET_SCALE2 / VOLTAGE_VID_OFFSET_SCALE1); - } - state->VddcPhase = pi->vddc_phase_shed_control ? 0 : 1; - - state->CcPwrDynRm = cpu_to_be32(state->CcPwrDynRm); - state->CcPwrDynRm1 = cpu_to_be32(state->CcPwrDynRm1); - state->VddcOffset = cpu_to_be16(state->VddcOffset); - - return 0; -} - -static int ci_calculate_sclk_params(struct amdgpu_device *adev, - u32 engine_clock, - SMU7_Discrete_GraphicsLevel *sclk) -{ - struct ci_power_info *pi = ci_get_pi(adev); - struct atom_clock_dividers dividers; - u32 spll_func_cntl_3 = pi->clock_registers.cg_spll_func_cntl_3; - u32 spll_func_cntl_4 = pi->clock_registers.cg_spll_func_cntl_4; - u32 cg_spll_spread_spectrum = pi->clock_registers.cg_spll_spread_spectrum; - u32 cg_spll_spread_spectrum_2 = pi->clock_registers.cg_spll_spread_spectrum_2; - u32 reference_clock = adev->clock.spll.reference_freq; - u32 reference_divider; - u32 fbdiv; - int ret; - - ret = amdgpu_atombios_get_clock_dividers(adev, - COMPUTE_GPUCLK_INPUT_FLAG_SCLK, - engine_clock, false, ÷rs); - if (ret) - return ret; - - reference_divider = 1 + dividers.ref_div; - fbdiv = dividers.fb_div & 0x3FFFFFF; - - spll_func_cntl_3 &= ~CG_SPLL_FUNC_CNTL_3__SPLL_FB_DIV_MASK; - spll_func_cntl_3 |= (fbdiv << CG_SPLL_FUNC_CNTL_3__SPLL_FB_DIV__SHIFT); - spll_func_cntl_3 |= CG_SPLL_FUNC_CNTL_3__SPLL_DITHEN_MASK; - - if (pi->caps_sclk_ss_support) { - struct amdgpu_atom_ss ss; - u32 vco_freq = engine_clock * dividers.post_div; - - if (amdgpu_atombios_get_asic_ss_info(adev, &ss, - ASIC_INTERNAL_ENGINE_SS, vco_freq)) { - u32 clk_s = reference_clock * 5 / (reference_divider * ss.rate); - u32 clk_v = 4 * ss.percentage * fbdiv / (clk_s * 10000); - - cg_spll_spread_spectrum &= ~(CG_SPLL_SPREAD_SPECTRUM__CLKS_MASK | CG_SPLL_SPREAD_SPECTRUM__SSEN_MASK); - cg_spll_spread_spectrum |= (clk_s << CG_SPLL_SPREAD_SPECTRUM__CLKS__SHIFT); - cg_spll_spread_spectrum |= (1 << CG_SPLL_SPREAD_SPECTRUM__SSEN__SHIFT); - - cg_spll_spread_spectrum_2 &= ~CG_SPLL_SPREAD_SPECTRUM_2__CLKV_MASK; - cg_spll_spread_spectrum_2 |= (clk_v << CG_SPLL_SPREAD_SPECTRUM_2__CLKV__SHIFT); - } - } - - sclk->SclkFrequency = engine_clock; - sclk->CgSpllFuncCntl3 = spll_func_cntl_3; - sclk->CgSpllFuncCntl4 = spll_func_cntl_4; - sclk->SpllSpreadSpectrum = cg_spll_spread_spectrum; - sclk->SpllSpreadSpectrum2 = cg_spll_spread_spectrum_2; - sclk->SclkDid = (u8)dividers.post_divider; - - return 0; -} - -static int ci_populate_single_graphic_level(struct amdgpu_device *adev, - u32 engine_clock, - u16 sclk_activity_level_t, - SMU7_Discrete_GraphicsLevel *graphic_level) -{ - struct ci_power_info *pi = ci_get_pi(adev); - int ret; - - ret = ci_calculate_sclk_params(adev, engine_clock, graphic_level); - if (ret) - return ret; - - ret = ci_get_dependency_volt_by_clk(adev, - &adev->pm.dpm.dyn_state.vddc_dependency_on_sclk, - engine_clock, &graphic_level->MinVddc); - if (ret) - return ret; - - graphic_level->SclkFrequency = engine_clock; - - graphic_level->Flags = 0; - graphic_level->MinVddcPhases = 1; - - if (pi->vddc_phase_shed_control) - ci_populate_phase_value_based_on_sclk(adev, - &adev->pm.dpm.dyn_state.phase_shedding_limits_table, - engine_clock, - &graphic_level->MinVddcPhases); - - graphic_level->ActivityLevel = sclk_activity_level_t; - - graphic_level->CcPwrDynRm = 0; - graphic_level->CcPwrDynRm1 = 0; - graphic_level->EnabledForThrottle = 1; - graphic_level->UpH = 0; - graphic_level->DownH = 0; - graphic_level->VoltageDownH = 0; - graphic_level->PowerThrottle = 0; - - if (pi->caps_sclk_ds) - graphic_level->DeepSleepDivId = ci_get_sleep_divider_id_from_clock(engine_clock, - CISLAND_MINIMUM_ENGINE_CLOCK); - - graphic_level->DisplayWatermark = PPSMC_DISPLAY_WATERMARK_LOW; - - graphic_level->Flags = cpu_to_be32(graphic_level->Flags); - graphic_level->MinVddc = cpu_to_be32(graphic_level->MinVddc * VOLTAGE_SCALE); - graphic_level->MinVddcPhases = cpu_to_be32(graphic_level->MinVddcPhases); - graphic_level->SclkFrequency = cpu_to_be32(graphic_level->SclkFrequency); - graphic_level->ActivityLevel = cpu_to_be16(graphic_level->ActivityLevel); - graphic_level->CgSpllFuncCntl3 = cpu_to_be32(graphic_level->CgSpllFuncCntl3); - graphic_level->CgSpllFuncCntl4 = cpu_to_be32(graphic_level->CgSpllFuncCntl4); - graphic_level->SpllSpreadSpectrum = cpu_to_be32(graphic_level->SpllSpreadSpectrum); - graphic_level->SpllSpreadSpectrum2 = cpu_to_be32(graphic_level->SpllSpreadSpectrum2); - graphic_level->CcPwrDynRm = cpu_to_be32(graphic_level->CcPwrDynRm); - graphic_level->CcPwrDynRm1 = cpu_to_be32(graphic_level->CcPwrDynRm1); - - return 0; -} - -static int ci_populate_all_graphic_levels(struct amdgpu_device *adev) -{ - struct ci_power_info *pi = ci_get_pi(adev); - struct ci_dpm_table *dpm_table = &pi->dpm_table; - u32 level_array_address = pi->dpm_table_start + - offsetof(SMU7_Discrete_DpmTable, GraphicsLevel); - u32 level_array_size = sizeof(SMU7_Discrete_GraphicsLevel) * - SMU7_MAX_LEVELS_GRAPHICS; - SMU7_Discrete_GraphicsLevel *levels = pi->smc_state_table.GraphicsLevel; - u32 i, ret; - - memset(levels, 0, level_array_size); - - for (i = 0; i < dpm_table->sclk_table.count; i++) { - ret = ci_populate_single_graphic_level(adev, - dpm_table->sclk_table.dpm_levels[i].value, - (u16)pi->activity_target[i], - &pi->smc_state_table.GraphicsLevel[i]); - if (ret) - return ret; - if (i > 1) - pi->smc_state_table.GraphicsLevel[i].DeepSleepDivId = 0; - if (i == (dpm_table->sclk_table.count - 1)) - pi->smc_state_table.GraphicsLevel[i].DisplayWatermark = - PPSMC_DISPLAY_WATERMARK_HIGH; - } - pi->smc_state_table.GraphicsLevel[0].EnabledForActivity = 1; - - pi->smc_state_table.GraphicsDpmLevelCount = (u8)dpm_table->sclk_table.count; - pi->dpm_level_enable_mask.sclk_dpm_enable_mask = - ci_get_dpm_level_enable_mask_value(&dpm_table->sclk_table); - - ret = amdgpu_ci_copy_bytes_to_smc(adev, level_array_address, - (u8 *)levels, level_array_size, - pi->sram_end); - if (ret) - return ret; - - return 0; -} - -static int ci_populate_ulv_state(struct amdgpu_device *adev, - SMU7_Discrete_Ulv *ulv_level) -{ - return ci_populate_ulv_level(adev, ulv_level); -} - -static int ci_populate_all_memory_levels(struct amdgpu_device *adev) -{ - struct ci_power_info *pi = ci_get_pi(adev); - struct ci_dpm_table *dpm_table = &pi->dpm_table; - u32 level_array_address = pi->dpm_table_start + - offsetof(SMU7_Discrete_DpmTable, MemoryLevel); - u32 level_array_size = sizeof(SMU7_Discrete_MemoryLevel) * - SMU7_MAX_LEVELS_MEMORY; - SMU7_Discrete_MemoryLevel *levels = pi->smc_state_table.MemoryLevel; - u32 i, ret; - - memset(levels, 0, level_array_size); - - for (i = 0; i < dpm_table->mclk_table.count; i++) { - if (dpm_table->mclk_table.dpm_levels[i].value == 0) - return -EINVAL; - ret = ci_populate_single_memory_level(adev, - dpm_table->mclk_table.dpm_levels[i].value, - &pi->smc_state_table.MemoryLevel[i]); - if (ret) - return ret; - } - - if ((dpm_table->mclk_table.count >= 2) && - ((adev->pdev->device == 0x67B0) || (adev->pdev->device == 0x67B1))) { - pi->smc_state_table.MemoryLevel[1].MinVddc = - pi->smc_state_table.MemoryLevel[0].MinVddc; - pi->smc_state_table.MemoryLevel[1].MinVddcPhases = - pi->smc_state_table.MemoryLevel[0].MinVddcPhases; - } - - pi->smc_state_table.MemoryLevel[0].ActivityLevel = cpu_to_be16(0x1F); - - pi->smc_state_table.MemoryDpmLevelCount = (u8)dpm_table->mclk_table.count; - pi->dpm_level_enable_mask.mclk_dpm_enable_mask = - ci_get_dpm_level_enable_mask_value(&dpm_table->mclk_table); - - pi->smc_state_table.MemoryLevel[dpm_table->mclk_table.count - 1].DisplayWatermark = - PPSMC_DISPLAY_WATERMARK_HIGH; - - ret = amdgpu_ci_copy_bytes_to_smc(adev, level_array_address, - (u8 *)levels, level_array_size, - pi->sram_end); - if (ret) - return ret; - - return 0; -} - -static void ci_reset_single_dpm_table(struct amdgpu_device *adev, - struct ci_single_dpm_table* dpm_table, - u32 count) -{ - u32 i; - - dpm_table->count = count; - for (i = 0; i < MAX_REGULAR_DPM_NUMBER; i++) - dpm_table->dpm_levels[i].enabled = false; -} - -static void ci_setup_pcie_table_entry(struct ci_single_dpm_table* dpm_table, - u32 index, u32 pcie_gen, u32 pcie_lanes) -{ - dpm_table->dpm_levels[index].value = pcie_gen; - dpm_table->dpm_levels[index].param1 = pcie_lanes; - dpm_table->dpm_levels[index].enabled = true; -} - -static int ci_setup_default_pcie_tables(struct amdgpu_device *adev) -{ - struct ci_power_info *pi = ci_get_pi(adev); - - if (!pi->use_pcie_performance_levels && !pi->use_pcie_powersaving_levels) - return -EINVAL; - - if (pi->use_pcie_performance_levels && !pi->use_pcie_powersaving_levels) { - pi->pcie_gen_powersaving = pi->pcie_gen_performance; - pi->pcie_lane_powersaving = pi->pcie_lane_performance; - } else if (!pi->use_pcie_performance_levels && pi->use_pcie_powersaving_levels) { - pi->pcie_gen_performance = pi->pcie_gen_powersaving; - pi->pcie_lane_performance = pi->pcie_lane_powersaving; - } - - ci_reset_single_dpm_table(adev, - &pi->dpm_table.pcie_speed_table, - SMU7_MAX_LEVELS_LINK); - - if (adev->asic_type == CHIP_BONAIRE) - ci_setup_pcie_table_entry(&pi->dpm_table.pcie_speed_table, 0, - pi->pcie_gen_powersaving.min, - pi->pcie_lane_powersaving.max); - else - ci_setup_pcie_table_entry(&pi->dpm_table.pcie_speed_table, 0, - pi->pcie_gen_powersaving.min, - pi->pcie_lane_powersaving.min); - ci_setup_pcie_table_entry(&pi->dpm_table.pcie_speed_table, 1, - pi->pcie_gen_performance.min, - pi->pcie_lane_performance.min); - ci_setup_pcie_table_entry(&pi->dpm_table.pcie_speed_table, 2, - pi->pcie_gen_powersaving.min, - pi->pcie_lane_powersaving.max); - ci_setup_pcie_table_entry(&pi->dpm_table.pcie_speed_table, 3, - pi->pcie_gen_performance.min, - pi->pcie_lane_performance.max); - ci_setup_pcie_table_entry(&pi->dpm_table.pcie_speed_table, 4, - pi->pcie_gen_powersaving.max, - pi->pcie_lane_powersaving.max); - ci_setup_pcie_table_entry(&pi->dpm_table.pcie_speed_table, 5, - pi->pcie_gen_performance.max, - pi->pcie_lane_performance.max); - - pi->dpm_table.pcie_speed_table.count = 6; - - return 0; -} - -static int ci_setup_default_dpm_tables(struct amdgpu_device *adev) -{ - struct ci_power_info *pi = ci_get_pi(adev); - struct amdgpu_clock_voltage_dependency_table *allowed_sclk_vddc_table = - &adev->pm.dpm.dyn_state.vddc_dependency_on_sclk; - struct amdgpu_clock_voltage_dependency_table *allowed_mclk_table = - &adev->pm.dpm.dyn_state.vddc_dependency_on_mclk; - struct amdgpu_cac_leakage_table *std_voltage_table = - &adev->pm.dpm.dyn_state.cac_leakage_table; - u32 i; - - if (allowed_sclk_vddc_table == NULL) - return -EINVAL; - if (allowed_sclk_vddc_table->count < 1) - return -EINVAL; - if (allowed_mclk_table == NULL) - return -EINVAL; - if (allowed_mclk_table->count < 1) - return -EINVAL; - - memset(&pi->dpm_table, 0, sizeof(struct ci_dpm_table)); - - ci_reset_single_dpm_table(adev, - &pi->dpm_table.sclk_table, - SMU7_MAX_LEVELS_GRAPHICS); - ci_reset_single_dpm_table(adev, - &pi->dpm_table.mclk_table, - SMU7_MAX_LEVELS_MEMORY); - ci_reset_single_dpm_table(adev, - &pi->dpm_table.vddc_table, - SMU7_MAX_LEVELS_VDDC); - ci_reset_single_dpm_table(adev, - &pi->dpm_table.vddci_table, - SMU7_MAX_LEVELS_VDDCI); - ci_reset_single_dpm_table(adev, - &pi->dpm_table.mvdd_table, - SMU7_MAX_LEVELS_MVDD); - - pi->dpm_table.sclk_table.count = 0; - for (i = 0; i < allowed_sclk_vddc_table->count; i++) { - if ((i == 0) || - (pi->dpm_table.sclk_table.dpm_levels[pi->dpm_table.sclk_table.count-1].value != - allowed_sclk_vddc_table->entries[i].clk)) { - pi->dpm_table.sclk_table.dpm_levels[pi->dpm_table.sclk_table.count].value = - allowed_sclk_vddc_table->entries[i].clk; - pi->dpm_table.sclk_table.dpm_levels[pi->dpm_table.sclk_table.count].enabled = - (i == 0) ? true : false; - pi->dpm_table.sclk_table.count++; - } - } - - pi->dpm_table.mclk_table.count = 0; - for (i = 0; i < allowed_mclk_table->count; i++) { - if ((i == 0) || - (pi->dpm_table.mclk_table.dpm_levels[pi->dpm_table.mclk_table.count-1].value != - allowed_mclk_table->entries[i].clk)) { - pi->dpm_table.mclk_table.dpm_levels[pi->dpm_table.mclk_table.count].value = - allowed_mclk_table->entries[i].clk; - pi->dpm_table.mclk_table.dpm_levels[pi->dpm_table.mclk_table.count].enabled = - (i == 0) ? true : false; - pi->dpm_table.mclk_table.count++; - } - } - - for (i = 0; i < allowed_sclk_vddc_table->count; i++) { - pi->dpm_table.vddc_table.dpm_levels[i].value = - allowed_sclk_vddc_table->entries[i].v; - pi->dpm_table.vddc_table.dpm_levels[i].param1 = - std_voltage_table->entries[i].leakage; - pi->dpm_table.vddc_table.dpm_levels[i].enabled = true; - } - pi->dpm_table.vddc_table.count = allowed_sclk_vddc_table->count; - - allowed_mclk_table = &adev->pm.dpm.dyn_state.vddci_dependency_on_mclk; - if (allowed_mclk_table) { - for (i = 0; i < allowed_mclk_table->count; i++) { - pi->dpm_table.vddci_table.dpm_levels[i].value = - allowed_mclk_table->entries[i].v; - pi->dpm_table.vddci_table.dpm_levels[i].enabled = true; - } - pi->dpm_table.vddci_table.count = allowed_mclk_table->count; - } - - allowed_mclk_table = &adev->pm.dpm.dyn_state.mvdd_dependency_on_mclk; - if (allowed_mclk_table) { - for (i = 0; i < allowed_mclk_table->count; i++) { - pi->dpm_table.mvdd_table.dpm_levels[i].value = - allowed_mclk_table->entries[i].v; - pi->dpm_table.mvdd_table.dpm_levels[i].enabled = true; - } - pi->dpm_table.mvdd_table.count = allowed_mclk_table->count; - } - - ci_setup_default_pcie_tables(adev); - - /* save a copy of the default DPM table */ - memcpy(&(pi->golden_dpm_table), &(pi->dpm_table), - sizeof(struct ci_dpm_table)); - - return 0; -} - -static int ci_find_boot_level(struct ci_single_dpm_table *table, - u32 value, u32 *boot_level) -{ - u32 i; - int ret = -EINVAL; - - for(i = 0; i < table->count; i++) { - if (value == table->dpm_levels[i].value) { - *boot_level = i; - ret = 0; - } - } - - return ret; -} - -static int ci_init_smc_table(struct amdgpu_device *adev) -{ - struct ci_power_info *pi = ci_get_pi(adev); - struct ci_ulv_parm *ulv = &pi->ulv; - struct amdgpu_ps *amdgpu_boot_state = adev->pm.dpm.boot_ps; - SMU7_Discrete_DpmTable *table = &pi->smc_state_table; - int ret; - - ret = ci_setup_default_dpm_tables(adev); - if (ret) - return ret; - - if (pi->voltage_control != CISLANDS_VOLTAGE_CONTROL_NONE) - ci_populate_smc_voltage_tables(adev, table); - - ci_init_fps_limits(adev); - - if (adev->pm.dpm.platform_caps & ATOM_PP_PLATFORM_CAP_HARDWAREDC) - table->SystemFlags |= PPSMC_SYSTEMFLAG_GPIO_DC; - - if (adev->pm.dpm.platform_caps & ATOM_PP_PLATFORM_CAP_STEPVDDC) - table->SystemFlags |= PPSMC_SYSTEMFLAG_STEPVDDC; - - if (adev->gmc.vram_type == AMDGPU_VRAM_TYPE_GDDR5) - table->SystemFlags |= PPSMC_SYSTEMFLAG_GDDR5; - - if (ulv->supported) { - ret = ci_populate_ulv_state(adev, &pi->smc_state_table.Ulv); - if (ret) - return ret; - WREG32_SMC(ixCG_ULV_PARAMETER, ulv->cg_ulv_parameter); - } - - ret = ci_populate_all_graphic_levels(adev); - if (ret) - return ret; - - ret = ci_populate_all_memory_levels(adev); - if (ret) - return ret; - - ci_populate_smc_link_level(adev, table); - - ret = ci_populate_smc_acpi_level(adev, table); - if (ret) - return ret; - - ret = ci_populate_smc_vce_level(adev, table); - if (ret) - return ret; - - ret = ci_populate_smc_acp_level(adev, table); - if (ret) - return ret; - - ret = ci_populate_smc_samu_level(adev, table); - if (ret) - return ret; - - ret = ci_do_program_memory_timing_parameters(adev); - if (ret) - return ret; - - ret = ci_populate_smc_uvd_level(adev, table); - if (ret) - return ret; - - table->UvdBootLevel = 0; - table->VceBootLevel = 0; - table->AcpBootLevel = 0; - table->SamuBootLevel = 0; - table->GraphicsBootLevel = 0; - table->MemoryBootLevel = 0; - - ret = ci_find_boot_level(&pi->dpm_table.sclk_table, - pi->vbios_boot_state.sclk_bootup_value, - (u32 *)&pi->smc_state_table.GraphicsBootLevel); - - ret = ci_find_boot_level(&pi->dpm_table.mclk_table, - pi->vbios_boot_state.mclk_bootup_value, - (u32 *)&pi->smc_state_table.MemoryBootLevel); - - table->BootVddc = pi->vbios_boot_state.vddc_bootup_value; - table->BootVddci = pi->vbios_boot_state.vddci_bootup_value; - table->BootMVdd = pi->vbios_boot_state.mvdd_bootup_value; - - ci_populate_smc_initial_state(adev, amdgpu_boot_state); - - ret = ci_populate_bapm_parameters_in_dpm_table(adev); - if (ret) - return ret; - - table->UVDInterval = 1; - table->VCEInterval = 1; - table->ACPInterval = 1; - table->SAMUInterval = 1; - table->GraphicsVoltageChangeEnable = 1; - table->GraphicsThermThrottleEnable = 1; - table->GraphicsInterval = 1; - table->VoltageInterval = 1; - table->ThermalInterval = 1; - table->TemperatureLimitHigh = (u16)((pi->thermal_temp_setting.temperature_high * - CISLANDS_Q88_FORMAT_CONVERSION_UNIT) / 1000); - table->TemperatureLimitLow = (u16)((pi->thermal_temp_setting.temperature_low * - CISLANDS_Q88_FORMAT_CONVERSION_UNIT) / 1000); - table->MemoryVoltageChangeEnable = 1; - table->MemoryInterval = 1; - table->VoltageResponseTime = 0; - table->VddcVddciDelta = 4000; - table->PhaseResponseTime = 0; - table->MemoryThermThrottleEnable = 1; - table->PCIeBootLinkLevel = pi->dpm_table.pcie_speed_table.count - 1; - table->PCIeGenInterval = 1; - if (pi->voltage_control == CISLANDS_VOLTAGE_CONTROL_BY_SVID2) - table->SVI2Enable = 1; - else - table->SVI2Enable = 0; - - table->ThermGpio = 17; - table->SclkStepSize = 0x4000; - - table->SystemFlags = cpu_to_be32(table->SystemFlags); - table->SmioMaskVddcVid = cpu_to_be32(table->SmioMaskVddcVid); - table->SmioMaskVddcPhase = cpu_to_be32(table->SmioMaskVddcPhase); - table->SmioMaskVddciVid = cpu_to_be32(table->SmioMaskVddciVid); - table->SmioMaskMvddVid = cpu_to_be32(table->SmioMaskMvddVid); - table->SclkStepSize = cpu_to_be32(table->SclkStepSize); - table->TemperatureLimitHigh = cpu_to_be16(table->TemperatureLimitHigh); - table->TemperatureLimitLow = cpu_to_be16(table->TemperatureLimitLow); - table->VddcVddciDelta = cpu_to_be16(table->VddcVddciDelta); - table->VoltageResponseTime = cpu_to_be16(table->VoltageResponseTime); - table->PhaseResponseTime = cpu_to_be16(table->PhaseResponseTime); - table->BootVddc = cpu_to_be16(table->BootVddc * VOLTAGE_SCALE); - table->BootVddci = cpu_to_be16(table->BootVddci * VOLTAGE_SCALE); - table->BootMVdd = cpu_to_be16(table->BootMVdd * VOLTAGE_SCALE); - - ret = amdgpu_ci_copy_bytes_to_smc(adev, - pi->dpm_table_start + - offsetof(SMU7_Discrete_DpmTable, SystemFlags), - (u8 *)&table->SystemFlags, - sizeof(SMU7_Discrete_DpmTable) - 3 * sizeof(SMU7_PIDController), - pi->sram_end); - if (ret) - return ret; - - return 0; -} - -static void ci_trim_single_dpm_states(struct amdgpu_device *adev, - struct ci_single_dpm_table *dpm_table, - u32 low_limit, u32 high_limit) -{ - u32 i; - - for (i = 0; i < dpm_table->count; i++) { - if ((dpm_table->dpm_levels[i].value < low_limit) || - (dpm_table->dpm_levels[i].value > high_limit)) - dpm_table->dpm_levels[i].enabled = false; - else - dpm_table->dpm_levels[i].enabled = true; - } -} - -static void ci_trim_pcie_dpm_states(struct amdgpu_device *adev, - u32 speed_low, u32 lanes_low, - u32 speed_high, u32 lanes_high) -{ - struct ci_power_info *pi = ci_get_pi(adev); - struct ci_single_dpm_table *pcie_table = &pi->dpm_table.pcie_speed_table; - u32 i, j; - - for (i = 0; i < pcie_table->count; i++) { - if ((pcie_table->dpm_levels[i].value < speed_low) || - (pcie_table->dpm_levels[i].param1 < lanes_low) || - (pcie_table->dpm_levels[i].value > speed_high) || - (pcie_table->dpm_levels[i].param1 > lanes_high)) - pcie_table->dpm_levels[i].enabled = false; - else - pcie_table->dpm_levels[i].enabled = true; - } - - for (i = 0; i < pcie_table->count; i++) { - if (pcie_table->dpm_levels[i].enabled) { - for (j = i + 1; j < pcie_table->count; j++) { - if (pcie_table->dpm_levels[j].enabled) { - if ((pcie_table->dpm_levels[i].value == pcie_table->dpm_levels[j].value) && - (pcie_table->dpm_levels[i].param1 == pcie_table->dpm_levels[j].param1)) - pcie_table->dpm_levels[j].enabled = false; - } - } - } - } -} - -static int ci_trim_dpm_states(struct amdgpu_device *adev, - struct amdgpu_ps *amdgpu_state) -{ - struct ci_ps *state = ci_get_ps(amdgpu_state); - struct ci_power_info *pi = ci_get_pi(adev); - u32 high_limit_count; - - if (state->performance_level_count < 1) - return -EINVAL; - - if (state->performance_level_count == 1) - high_limit_count = 0; - else - high_limit_count = 1; - - ci_trim_single_dpm_states(adev, - &pi->dpm_table.sclk_table, - state->performance_levels[0].sclk, - state->performance_levels[high_limit_count].sclk); - - ci_trim_single_dpm_states(adev, - &pi->dpm_table.mclk_table, - state->performance_levels[0].mclk, - state->performance_levels[high_limit_count].mclk); - - ci_trim_pcie_dpm_states(adev, - state->performance_levels[0].pcie_gen, - state->performance_levels[0].pcie_lane, - state->performance_levels[high_limit_count].pcie_gen, - state->performance_levels[high_limit_count].pcie_lane); - - return 0; -} - -static int ci_apply_disp_minimum_voltage_request(struct amdgpu_device *adev) -{ - struct amdgpu_clock_voltage_dependency_table *disp_voltage_table = - &adev->pm.dpm.dyn_state.vddc_dependency_on_dispclk; - struct amdgpu_clock_voltage_dependency_table *vddc_table = - &adev->pm.dpm.dyn_state.vddc_dependency_on_sclk; - u32 requested_voltage = 0; - u32 i; - - if (disp_voltage_table == NULL) - return -EINVAL; - if (!disp_voltage_table->count) - return -EINVAL; - - for (i = 0; i < disp_voltage_table->count; i++) { - if (adev->clock.current_dispclk == disp_voltage_table->entries[i].clk) - requested_voltage = disp_voltage_table->entries[i].v; - } - - for (i = 0; i < vddc_table->count; i++) { - if (requested_voltage <= vddc_table->entries[i].v) { - requested_voltage = vddc_table->entries[i].v; - return (amdgpu_ci_send_msg_to_smc_with_parameter(adev, - PPSMC_MSG_VddC_Request, - requested_voltage * VOLTAGE_SCALE) == PPSMC_Result_OK) ? - 0 : -EINVAL; - } - } - - return -EINVAL; -} - -static int ci_upload_dpm_level_enable_mask(struct amdgpu_device *adev) -{ - struct ci_power_info *pi = ci_get_pi(adev); - PPSMC_Result result; - - ci_apply_disp_minimum_voltage_request(adev); - - if (!pi->sclk_dpm_key_disabled) { - if (pi->dpm_level_enable_mask.sclk_dpm_enable_mask) { - result = amdgpu_ci_send_msg_to_smc_with_parameter(adev, - PPSMC_MSG_SCLKDPM_SetEnabledMask, - pi->dpm_level_enable_mask.sclk_dpm_enable_mask); - if (result != PPSMC_Result_OK) - return -EINVAL; - } - } - - if (!pi->mclk_dpm_key_disabled) { - if (pi->dpm_level_enable_mask.mclk_dpm_enable_mask) { - result = amdgpu_ci_send_msg_to_smc_with_parameter(adev, - PPSMC_MSG_MCLKDPM_SetEnabledMask, - pi->dpm_level_enable_mask.mclk_dpm_enable_mask); - if (result != PPSMC_Result_OK) - return -EINVAL; - } - } - -#if 0 - if (!pi->pcie_dpm_key_disabled) { - if (pi->dpm_level_enable_mask.pcie_dpm_enable_mask) { - result = amdgpu_ci_send_msg_to_smc_with_parameter(adev, - PPSMC_MSG_PCIeDPM_SetEnabledMask, - pi->dpm_level_enable_mask.pcie_dpm_enable_mask); - if (result != PPSMC_Result_OK) - return -EINVAL; - } - } -#endif - - return 0; -} - -static void ci_find_dpm_states_clocks_in_dpm_table(struct amdgpu_device *adev, - struct amdgpu_ps *amdgpu_state) -{ - struct ci_power_info *pi = ci_get_pi(adev); - struct ci_ps *state = ci_get_ps(amdgpu_state); - struct ci_single_dpm_table *sclk_table = &pi->dpm_table.sclk_table; - u32 sclk = state->performance_levels[state->performance_level_count-1].sclk; - struct ci_single_dpm_table *mclk_table = &pi->dpm_table.mclk_table; - u32 mclk = state->performance_levels[state->performance_level_count-1].mclk; - u32 i; - - pi->need_update_smu7_dpm_table = 0; - - for (i = 0; i < sclk_table->count; i++) { - if (sclk == sclk_table->dpm_levels[i].value) - break; - } - - if (i >= sclk_table->count) { - pi->need_update_smu7_dpm_table |= DPMTABLE_OD_UPDATE_SCLK; - } else { - /* XXX check display min clock requirements */ - if (CISLAND_MINIMUM_ENGINE_CLOCK != CISLAND_MINIMUM_ENGINE_CLOCK) - pi->need_update_smu7_dpm_table |= DPMTABLE_UPDATE_SCLK; - } - - for (i = 0; i < mclk_table->count; i++) { - if (mclk == mclk_table->dpm_levels[i].value) - break; - } - - if (i >= mclk_table->count) - pi->need_update_smu7_dpm_table |= DPMTABLE_OD_UPDATE_MCLK; - - if (adev->pm.dpm.current_active_crtc_count != - adev->pm.dpm.new_active_crtc_count) - pi->need_update_smu7_dpm_table |= DPMTABLE_UPDATE_MCLK; -} - -static int ci_populate_and_upload_sclk_mclk_dpm_levels(struct amdgpu_device *adev, - struct amdgpu_ps *amdgpu_state) -{ - struct ci_power_info *pi = ci_get_pi(adev); - struct ci_ps *state = ci_get_ps(amdgpu_state); - u32 sclk = state->performance_levels[state->performance_level_count-1].sclk; - u32 mclk = state->performance_levels[state->performance_level_count-1].mclk; - struct ci_dpm_table *dpm_table = &pi->dpm_table; - int ret; - - if (!pi->need_update_smu7_dpm_table) - return 0; - - if (pi->need_update_smu7_dpm_table & DPMTABLE_OD_UPDATE_SCLK) - dpm_table->sclk_table.dpm_levels[dpm_table->sclk_table.count-1].value = sclk; - - if (pi->need_update_smu7_dpm_table & DPMTABLE_OD_UPDATE_MCLK) - dpm_table->mclk_table.dpm_levels[dpm_table->mclk_table.count-1].value = mclk; - - if (pi->need_update_smu7_dpm_table & (DPMTABLE_OD_UPDATE_SCLK | DPMTABLE_UPDATE_SCLK)) { - ret = ci_populate_all_graphic_levels(adev); - if (ret) - return ret; - } - - if (pi->need_update_smu7_dpm_table & (DPMTABLE_OD_UPDATE_MCLK | DPMTABLE_UPDATE_MCLK)) { - ret = ci_populate_all_memory_levels(adev); - if (ret) - return ret; - } - - return 0; -} - -static int ci_enable_uvd_dpm(struct amdgpu_device *adev, bool enable) -{ - struct ci_power_info *pi = ci_get_pi(adev); - const struct amdgpu_clock_and_voltage_limits *max_limits; - int i; - - if (adev->pm.ac_power) - max_limits = &adev->pm.dpm.dyn_state.max_clock_voltage_on_ac; - else - max_limits = &adev->pm.dpm.dyn_state.max_clock_voltage_on_dc; - - if (enable) { - pi->dpm_level_enable_mask.uvd_dpm_enable_mask = 0; - - for (i = adev->pm.dpm.dyn_state.uvd_clock_voltage_dependency_table.count - 1; i >= 0; i--) { - if (adev->pm.dpm.dyn_state.uvd_clock_voltage_dependency_table.entries[i].v <= max_limits->vddc) { - pi->dpm_level_enable_mask.uvd_dpm_enable_mask |= 1 << i; - - if (!pi->caps_uvd_dpm) - break; - } - } - - amdgpu_ci_send_msg_to_smc_with_parameter(adev, - PPSMC_MSG_UVDDPM_SetEnabledMask, - pi->dpm_level_enable_mask.uvd_dpm_enable_mask); - - if (pi->last_mclk_dpm_enable_mask & 0x1) { - pi->uvd_enabled = true; - pi->dpm_level_enable_mask.mclk_dpm_enable_mask &= 0xFFFFFFFE; - amdgpu_ci_send_msg_to_smc_with_parameter(adev, - PPSMC_MSG_MCLKDPM_SetEnabledMask, - pi->dpm_level_enable_mask.mclk_dpm_enable_mask); - } - } else { - if (pi->uvd_enabled) { - pi->uvd_enabled = false; - pi->dpm_level_enable_mask.mclk_dpm_enable_mask |= 1; - amdgpu_ci_send_msg_to_smc_with_parameter(adev, - PPSMC_MSG_MCLKDPM_SetEnabledMask, - pi->dpm_level_enable_mask.mclk_dpm_enable_mask); - } - } - - return (amdgpu_ci_send_msg_to_smc(adev, enable ? - PPSMC_MSG_UVDDPM_Enable : PPSMC_MSG_UVDDPM_Disable) == PPSMC_Result_OK) ? - 0 : -EINVAL; -} - -static int ci_enable_vce_dpm(struct amdgpu_device *adev, bool enable) -{ - struct ci_power_info *pi = ci_get_pi(adev); - const struct amdgpu_clock_and_voltage_limits *max_limits; - int i; - - if (adev->pm.ac_power) - max_limits = &adev->pm.dpm.dyn_state.max_clock_voltage_on_ac; - else - max_limits = &adev->pm.dpm.dyn_state.max_clock_voltage_on_dc; - - if (enable) { - pi->dpm_level_enable_mask.vce_dpm_enable_mask = 0; - for (i = adev->pm.dpm.dyn_state.vce_clock_voltage_dependency_table.count - 1; i >= 0; i--) { - if (adev->pm.dpm.dyn_state.vce_clock_voltage_dependency_table.entries[i].v <= max_limits->vddc) { - pi->dpm_level_enable_mask.vce_dpm_enable_mask |= 1 << i; - - if (!pi->caps_vce_dpm) - break; - } - } - - amdgpu_ci_send_msg_to_smc_with_parameter(adev, - PPSMC_MSG_VCEDPM_SetEnabledMask, - pi->dpm_level_enable_mask.vce_dpm_enable_mask); - } - - return (amdgpu_ci_send_msg_to_smc(adev, enable ? - PPSMC_MSG_VCEDPM_Enable : PPSMC_MSG_VCEDPM_Disable) == PPSMC_Result_OK) ? - 0 : -EINVAL; -} - -#if 0 -static int ci_enable_samu_dpm(struct amdgpu_device *adev, bool enable) -{ - struct ci_power_info *pi = ci_get_pi(adev); - const struct amdgpu_clock_and_voltage_limits *max_limits; - int i; - - if (adev->pm.ac_power) - max_limits = &adev->pm.dpm.dyn_state.max_clock_voltage_on_ac; - else - max_limits = &adev->pm.dpm.dyn_state.max_clock_voltage_on_dc; - - if (enable) { - pi->dpm_level_enable_mask.samu_dpm_enable_mask = 0; - for (i = adev->pm.dpm.dyn_state.samu_clock_voltage_dependency_table.count - 1; i >= 0; i--) { - if (adev->pm.dpm.dyn_state.samu_clock_voltage_dependency_table.entries[i].v <= max_limits->vddc) { - pi->dpm_level_enable_mask.samu_dpm_enable_mask |= 1 << i; - - if (!pi->caps_samu_dpm) - break; - } - } - - amdgpu_ci_send_msg_to_smc_with_parameter(adev, - PPSMC_MSG_SAMUDPM_SetEnabledMask, - pi->dpm_level_enable_mask.samu_dpm_enable_mask); - } - return (amdgpu_ci_send_msg_to_smc(adev, enable ? - PPSMC_MSG_SAMUDPM_Enable : PPSMC_MSG_SAMUDPM_Disable) == PPSMC_Result_OK) ? - 0 : -EINVAL; -} - -static int ci_enable_acp_dpm(struct amdgpu_device *adev, bool enable) -{ - struct ci_power_info *pi = ci_get_pi(adev); - const struct amdgpu_clock_and_voltage_limits *max_limits; - int i; - - if (adev->pm.ac_power) - max_limits = &adev->pm.dpm.dyn_state.max_clock_voltage_on_ac; - else - max_limits = &adev->pm.dpm.dyn_state.max_clock_voltage_on_dc; - - if (enable) { - pi->dpm_level_enable_mask.acp_dpm_enable_mask = 0; - for (i = adev->pm.dpm.dyn_state.acp_clock_voltage_dependency_table.count - 1; i >= 0; i--) { - if (adev->pm.dpm.dyn_state.acp_clock_voltage_dependency_table.entries[i].v <= max_limits->vddc) { - pi->dpm_level_enable_mask.acp_dpm_enable_mask |= 1 << i; - - if (!pi->caps_acp_dpm) - break; - } - } - - amdgpu_ci_send_msg_to_smc_with_parameter(adev, - PPSMC_MSG_ACPDPM_SetEnabledMask, - pi->dpm_level_enable_mask.acp_dpm_enable_mask); - } - - return (amdgpu_ci_send_msg_to_smc(adev, enable ? - PPSMC_MSG_ACPDPM_Enable : PPSMC_MSG_ACPDPM_Disable) == PPSMC_Result_OK) ? - 0 : -EINVAL; -} -#endif - -static int ci_update_uvd_dpm(struct amdgpu_device *adev, bool gate) -{ - struct ci_power_info *pi = ci_get_pi(adev); - u32 tmp; - int ret = 0; - - if (!gate) { - /* turn the clocks on when decoding */ - if (pi->caps_uvd_dpm || - (adev->pm.dpm.dyn_state.uvd_clock_voltage_dependency_table.count <= 0)) - pi->smc_state_table.UvdBootLevel = 0; - else - pi->smc_state_table.UvdBootLevel = - adev->pm.dpm.dyn_state.uvd_clock_voltage_dependency_table.count - 1; - - tmp = RREG32_SMC(ixDPM_TABLE_475); - tmp &= ~DPM_TABLE_475__UvdBootLevel_MASK; - tmp |= (pi->smc_state_table.UvdBootLevel << DPM_TABLE_475__UvdBootLevel__SHIFT); - WREG32_SMC(ixDPM_TABLE_475, tmp); - ret = ci_enable_uvd_dpm(adev, true); - } else { - ret = ci_enable_uvd_dpm(adev, false); - if (ret) - return ret; - } - - return ret; -} - -static u8 ci_get_vce_boot_level(struct amdgpu_device *adev) -{ - u8 i; - u32 min_evclk = 30000; /* ??? */ - struct amdgpu_vce_clock_voltage_dependency_table *table = - &adev->pm.dpm.dyn_state.vce_clock_voltage_dependency_table; - - for (i = 0; i < table->count; i++) { - if (table->entries[i].evclk >= min_evclk) - return i; - } - - return table->count - 1; -} - -static int ci_update_vce_dpm(struct amdgpu_device *adev, - struct amdgpu_ps *amdgpu_new_state, - struct amdgpu_ps *amdgpu_current_state) -{ - struct ci_power_info *pi = ci_get_pi(adev); - int ret = 0; - u32 tmp; - - if (amdgpu_current_state->evclk != amdgpu_new_state->evclk) { - if (amdgpu_new_state->evclk) { - pi->smc_state_table.VceBootLevel = ci_get_vce_boot_level(adev); - tmp = RREG32_SMC(ixDPM_TABLE_475); - tmp &= ~DPM_TABLE_475__VceBootLevel_MASK; - tmp |= (pi->smc_state_table.VceBootLevel << DPM_TABLE_475__VceBootLevel__SHIFT); - WREG32_SMC(ixDPM_TABLE_475, tmp); - - ret = ci_enable_vce_dpm(adev, true); - } else { - ret = ci_enable_vce_dpm(adev, false); - if (ret) - return ret; - } - } - return ret; -} - -#if 0 -static int ci_update_samu_dpm(struct amdgpu_device *adev, bool gate) -{ - return ci_enable_samu_dpm(adev, gate); -} - -static int ci_update_acp_dpm(struct amdgpu_device *adev, bool gate) -{ - struct ci_power_info *pi = ci_get_pi(adev); - u32 tmp; - - if (!gate) { - pi->smc_state_table.AcpBootLevel = 0; - - tmp = RREG32_SMC(ixDPM_TABLE_475); - tmp &= ~AcpBootLevel_MASK; - tmp |= AcpBootLevel(pi->smc_state_table.AcpBootLevel); - WREG32_SMC(ixDPM_TABLE_475, tmp); - } - - return ci_enable_acp_dpm(adev, !gate); -} -#endif - -static int ci_generate_dpm_level_enable_mask(struct amdgpu_device *adev, - struct amdgpu_ps *amdgpu_state) -{ - struct ci_power_info *pi = ci_get_pi(adev); - int ret; - - ret = ci_trim_dpm_states(adev, amdgpu_state); - if (ret) - return ret; - - pi->dpm_level_enable_mask.sclk_dpm_enable_mask = - ci_get_dpm_level_enable_mask_value(&pi->dpm_table.sclk_table); - pi->dpm_level_enable_mask.mclk_dpm_enable_mask = - ci_get_dpm_level_enable_mask_value(&pi->dpm_table.mclk_table); - pi->last_mclk_dpm_enable_mask = - pi->dpm_level_enable_mask.mclk_dpm_enable_mask; - if (pi->uvd_enabled) { - if (pi->dpm_level_enable_mask.mclk_dpm_enable_mask & 1) - pi->dpm_level_enable_mask.mclk_dpm_enable_mask &= 0xFFFFFFFE; - } - pi->dpm_level_enable_mask.pcie_dpm_enable_mask = - ci_get_dpm_level_enable_mask_value(&pi->dpm_table.pcie_speed_table); - - return 0; -} - -static u32 ci_get_lowest_enabled_level(struct amdgpu_device *adev, - u32 level_mask) -{ - u32 level = 0; - - while ((level_mask & (1 << level)) == 0) - level++; - - return level; -} - - -static int ci_dpm_force_performance_level(void *handle, - enum amd_dpm_forced_level level) -{ - struct amdgpu_device *adev = (struct amdgpu_device *)handle; - struct ci_power_info *pi = ci_get_pi(adev); - u32 tmp, levels, i; - int ret; - - if (level == AMD_DPM_FORCED_LEVEL_HIGH) { - if ((!pi->pcie_dpm_key_disabled) && - pi->dpm_level_enable_mask.pcie_dpm_enable_mask) { - levels = 0; - tmp = pi->dpm_level_enable_mask.pcie_dpm_enable_mask; - while (tmp >>= 1) - levels++; - if (levels) { - ret = ci_dpm_force_state_pcie(adev, level); - if (ret) - return ret; - for (i = 0; i < adev->usec_timeout; i++) { - tmp = (RREG32_SMC(ixTARGET_AND_CURRENT_PROFILE_INDEX_1) & - TARGET_AND_CURRENT_PROFILE_INDEX_1__CURR_PCIE_INDEX_MASK) >> - TARGET_AND_CURRENT_PROFILE_INDEX_1__CURR_PCIE_INDEX__SHIFT; - if (tmp == levels) - break; - udelay(1); - } - } - } - if ((!pi->sclk_dpm_key_disabled) && - pi->dpm_level_enable_mask.sclk_dpm_enable_mask) { - levels = 0; - tmp = pi->dpm_level_enable_mask.sclk_dpm_enable_mask; - while (tmp >>= 1) - levels++; - if (levels) { - ret = ci_dpm_force_state_sclk(adev, levels); - if (ret) - return ret; - for (i = 0; i < adev->usec_timeout; i++) { - tmp = (RREG32_SMC(ixTARGET_AND_CURRENT_PROFILE_INDEX) & - TARGET_AND_CURRENT_PROFILE_INDEX__CURR_SCLK_INDEX_MASK) >> - TARGET_AND_CURRENT_PROFILE_INDEX__CURR_SCLK_INDEX__SHIFT; - if (tmp == levels) - break; - udelay(1); - } - } - } - if ((!pi->mclk_dpm_key_disabled) && - pi->dpm_level_enable_mask.mclk_dpm_enable_mask) { - levels = 0; - tmp = pi->dpm_level_enable_mask.mclk_dpm_enable_mask; - while (tmp >>= 1) - levels++; - if (levels) { - ret = ci_dpm_force_state_mclk(adev, levels); - if (ret) - return ret; - for (i = 0; i < adev->usec_timeout; i++) { - tmp = (RREG32_SMC(ixTARGET_AND_CURRENT_PROFILE_INDEX) & - TARGET_AND_CURRENT_PROFILE_INDEX__CURR_MCLK_INDEX_MASK) >> - TARGET_AND_CURRENT_PROFILE_INDEX__CURR_MCLK_INDEX__SHIFT; - if (tmp == levels) - break; - udelay(1); - } - } - } - } else if (level == AMD_DPM_FORCED_LEVEL_LOW) { - if ((!pi->sclk_dpm_key_disabled) && - pi->dpm_level_enable_mask.sclk_dpm_enable_mask) { - levels = ci_get_lowest_enabled_level(adev, - pi->dpm_level_enable_mask.sclk_dpm_enable_mask); - ret = ci_dpm_force_state_sclk(adev, levels); - if (ret) - return ret; - for (i = 0; i < adev->usec_timeout; i++) { - tmp = (RREG32_SMC(ixTARGET_AND_CURRENT_PROFILE_INDEX) & - TARGET_AND_CURRENT_PROFILE_INDEX__CURR_SCLK_INDEX_MASK) >> - TARGET_AND_CURRENT_PROFILE_INDEX__CURR_SCLK_INDEX__SHIFT; - if (tmp == levels) - break; - udelay(1); - } - } - if ((!pi->mclk_dpm_key_disabled) && - pi->dpm_level_enable_mask.mclk_dpm_enable_mask) { - levels = ci_get_lowest_enabled_level(adev, - pi->dpm_level_enable_mask.mclk_dpm_enable_mask); - ret = ci_dpm_force_state_mclk(adev, levels); - if (ret) - return ret; - for (i = 0; i < adev->usec_timeout; i++) { - tmp = (RREG32_SMC(ixTARGET_AND_CURRENT_PROFILE_INDEX) & - TARGET_AND_CURRENT_PROFILE_INDEX__CURR_MCLK_INDEX_MASK) >> - TARGET_AND_CURRENT_PROFILE_INDEX__CURR_MCLK_INDEX__SHIFT; - if (tmp == levels) - break; - udelay(1); - } - } - if ((!pi->pcie_dpm_key_disabled) && - pi->dpm_level_enable_mask.pcie_dpm_enable_mask) { - levels = ci_get_lowest_enabled_level(adev, - pi->dpm_level_enable_mask.pcie_dpm_enable_mask); - ret = ci_dpm_force_state_pcie(adev, levels); - if (ret) - return ret; - for (i = 0; i < adev->usec_timeout; i++) { - tmp = (RREG32_SMC(ixTARGET_AND_CURRENT_PROFILE_INDEX_1) & - TARGET_AND_CURRENT_PROFILE_INDEX_1__CURR_PCIE_INDEX_MASK) >> - TARGET_AND_CURRENT_PROFILE_INDEX_1__CURR_PCIE_INDEX__SHIFT; - if (tmp == levels) - break; - udelay(1); - } - } - } else if (level == AMD_DPM_FORCED_LEVEL_AUTO) { - if (!pi->pcie_dpm_key_disabled) { - PPSMC_Result smc_result; - - smc_result = amdgpu_ci_send_msg_to_smc(adev, - PPSMC_MSG_PCIeDPM_UnForceLevel); - if (smc_result != PPSMC_Result_OK) - return -EINVAL; - } - ret = ci_upload_dpm_level_enable_mask(adev); - if (ret) - return ret; - } - - adev->pm.dpm.forced_level = level; - - return 0; -} - -static int ci_set_mc_special_registers(struct amdgpu_device *adev, - struct ci_mc_reg_table *table) -{ - u8 i, j, k; - u32 temp_reg; - - for (i = 0, j = table->last; i < table->last; i++) { - if (j >= SMU7_DISCRETE_MC_REGISTER_ARRAY_SIZE) - return -EINVAL; - switch(table->mc_reg_address[i].s1) { - case mmMC_SEQ_MISC1: - temp_reg = RREG32(mmMC_PMG_CMD_EMRS); - table->mc_reg_address[j].s1 = mmMC_PMG_CMD_EMRS; - table->mc_reg_address[j].s0 = mmMC_SEQ_PMG_CMD_EMRS_LP; - for (k = 0; k < table->num_entries; k++) { - table->mc_reg_table_entry[k].mc_data[j] = - ((temp_reg & 0xffff0000)) | ((table->mc_reg_table_entry[k].mc_data[i] & 0xffff0000) >> 16); - } - j++; - - if (j >= SMU7_DISCRETE_MC_REGISTER_ARRAY_SIZE) - return -EINVAL; - temp_reg = RREG32(mmMC_PMG_CMD_MRS); - table->mc_reg_address[j].s1 = mmMC_PMG_CMD_MRS; - table->mc_reg_address[j].s0 = mmMC_SEQ_PMG_CMD_MRS_LP; - for (k = 0; k < table->num_entries; k++) { - table->mc_reg_table_entry[k].mc_data[j] = - (temp_reg & 0xffff0000) | (table->mc_reg_table_entry[k].mc_data[i] & 0x0000ffff); - if (adev->gmc.vram_type != AMDGPU_VRAM_TYPE_GDDR5) - table->mc_reg_table_entry[k].mc_data[j] |= 0x100; - } - j++; - - if (adev->gmc.vram_type != AMDGPU_VRAM_TYPE_GDDR5) { - if (j >= SMU7_DISCRETE_MC_REGISTER_ARRAY_SIZE) - return -EINVAL; - table->mc_reg_address[j].s1 = mmMC_PMG_AUTO_CMD; - table->mc_reg_address[j].s0 = mmMC_PMG_AUTO_CMD; - for (k = 0; k < table->num_entries; k++) { - table->mc_reg_table_entry[k].mc_data[j] = - (table->mc_reg_table_entry[k].mc_data[i] & 0xffff0000) >> 16; - } - j++; - } - break; - case mmMC_SEQ_RESERVE_M: - temp_reg = RREG32(mmMC_PMG_CMD_MRS1); - table->mc_reg_address[j].s1 = mmMC_PMG_CMD_MRS1; - table->mc_reg_address[j].s0 = mmMC_SEQ_PMG_CMD_MRS1_LP; - for (k = 0; k < table->num_entries; k++) { - table->mc_reg_table_entry[k].mc_data[j] = - (temp_reg & 0xffff0000) | (table->mc_reg_table_entry[k].mc_data[i] & 0x0000ffff); - } - j++; - break; - default: - break; - } - - } - - table->last = j; - - return 0; -} - -static bool ci_check_s0_mc_reg_index(u16 in_reg, u16 *out_reg) -{ - bool result = true; - - switch(in_reg) { - case mmMC_SEQ_RAS_TIMING: - *out_reg = mmMC_SEQ_RAS_TIMING_LP; - break; - case mmMC_SEQ_DLL_STBY: - *out_reg = mmMC_SEQ_DLL_STBY_LP; - break; - case mmMC_SEQ_G5PDX_CMD0: - *out_reg = mmMC_SEQ_G5PDX_CMD0_LP; - break; - case mmMC_SEQ_G5PDX_CMD1: - *out_reg = mmMC_SEQ_G5PDX_CMD1_LP; - break; - case mmMC_SEQ_G5PDX_CTRL: - *out_reg = mmMC_SEQ_G5PDX_CTRL_LP; - break; - case mmMC_SEQ_CAS_TIMING: - *out_reg = mmMC_SEQ_CAS_TIMING_LP; - break; - case mmMC_SEQ_MISC_TIMING: - *out_reg = mmMC_SEQ_MISC_TIMING_LP; - break; - case mmMC_SEQ_MISC_TIMING2: - *out_reg = mmMC_SEQ_MISC_TIMING2_LP; - break; - case mmMC_SEQ_PMG_DVS_CMD: - *out_reg = mmMC_SEQ_PMG_DVS_CMD_LP; - break; - case mmMC_SEQ_PMG_DVS_CTL: - *out_reg = mmMC_SEQ_PMG_DVS_CTL_LP; - break; - case mmMC_SEQ_RD_CTL_D0: - *out_reg = mmMC_SEQ_RD_CTL_D0_LP; - break; - case mmMC_SEQ_RD_CTL_D1: - *out_reg = mmMC_SEQ_RD_CTL_D1_LP; - break; - case mmMC_SEQ_WR_CTL_D0: - *out_reg = mmMC_SEQ_WR_CTL_D0_LP; - break; - case mmMC_SEQ_WR_CTL_D1: - *out_reg = mmMC_SEQ_WR_CTL_D1_LP; - break; - case mmMC_PMG_CMD_EMRS: - *out_reg = mmMC_SEQ_PMG_CMD_EMRS_LP; - break; - case mmMC_PMG_CMD_MRS: - *out_reg = mmMC_SEQ_PMG_CMD_MRS_LP; - break; - case mmMC_PMG_CMD_MRS1: - *out_reg = mmMC_SEQ_PMG_CMD_MRS1_LP; - break; - case mmMC_SEQ_PMG_TIMING: - *out_reg = mmMC_SEQ_PMG_TIMING_LP; - break; - case mmMC_PMG_CMD_MRS2: - *out_reg = mmMC_SEQ_PMG_CMD_MRS2_LP; - break; - case mmMC_SEQ_WR_CTL_2: - *out_reg = mmMC_SEQ_WR_CTL_2_LP; - break; - default: - result = false; - break; - } - - return result; -} - -static void ci_set_valid_flag(struct ci_mc_reg_table *table) -{ - u8 i, j; - - for (i = 0; i < table->last; i++) { - for (j = 1; j < table->num_entries; j++) { - if (table->mc_reg_table_entry[j-1].mc_data[i] != - table->mc_reg_table_entry[j].mc_data[i]) { - table->valid_flag |= 1 << i; - break; - } - } - } -} - -static void ci_set_s0_mc_reg_index(struct ci_mc_reg_table *table) -{ - u32 i; - u16 address; - - for (i = 0; i < table->last; i++) { - table->mc_reg_address[i].s0 = - ci_check_s0_mc_reg_index(table->mc_reg_address[i].s1, &address) ? - address : table->mc_reg_address[i].s1; - } -} - -static int ci_copy_vbios_mc_reg_table(const struct atom_mc_reg_table *table, - struct ci_mc_reg_table *ci_table) -{ - u8 i, j; - - if (table->last > SMU7_DISCRETE_MC_REGISTER_ARRAY_SIZE) - return -EINVAL; - if (table->num_entries > MAX_AC_TIMING_ENTRIES) - return -EINVAL; - - for (i = 0; i < table->last; i++) - ci_table->mc_reg_address[i].s1 = table->mc_reg_address[i].s1; - - ci_table->last = table->last; - - for (i = 0; i < table->num_entries; i++) { - ci_table->mc_reg_table_entry[i].mclk_max = - table->mc_reg_table_entry[i].mclk_max; - for (j = 0; j < table->last; j++) - ci_table->mc_reg_table_entry[i].mc_data[j] = - table->mc_reg_table_entry[i].mc_data[j]; - } - ci_table->num_entries = table->num_entries; - - return 0; -} - -static int ci_register_patching_mc_seq(struct amdgpu_device *adev, - struct ci_mc_reg_table *table) -{ - u8 i, k; - u32 tmp; - bool patch; - - tmp = RREG32(mmMC_SEQ_MISC0); - patch = ((tmp & 0x0000f00) == 0x300) ? true : false; - - if (patch && - ((adev->pdev->device == 0x67B0) || - (adev->pdev->device == 0x67B1))) { - for (i = 0; i < table->last; i++) { - if (table->last >= SMU7_DISCRETE_MC_REGISTER_ARRAY_SIZE) - return -EINVAL; - switch (table->mc_reg_address[i].s1) { - case mmMC_SEQ_MISC1: - for (k = 0; k < table->num_entries; k++) { - if ((table->mc_reg_table_entry[k].mclk_max == 125000) || - (table->mc_reg_table_entry[k].mclk_max == 137500)) - table->mc_reg_table_entry[k].mc_data[i] = - (table->mc_reg_table_entry[k].mc_data[i] & 0xFFFFFFF8) | - 0x00000007; - } - break; - case mmMC_SEQ_WR_CTL_D0: - for (k = 0; k < table->num_entries; k++) { - if ((table->mc_reg_table_entry[k].mclk_max == 125000) || - (table->mc_reg_table_entry[k].mclk_max == 137500)) - table->mc_reg_table_entry[k].mc_data[i] = - (table->mc_reg_table_entry[k].mc_data[i] & 0xFFFF0F00) | - 0x0000D0DD; - } - break; - case mmMC_SEQ_WR_CTL_D1: - for (k = 0; k < table->num_entries; k++) { - if ((table->mc_reg_table_entry[k].mclk_max == 125000) || - (table->mc_reg_table_entry[k].mclk_max == 137500)) - table->mc_reg_table_entry[k].mc_data[i] = - (table->mc_reg_table_entry[k].mc_data[i] & 0xFFFF0F00) | - 0x0000D0DD; - } - break; - case mmMC_SEQ_WR_CTL_2: - for (k = 0; k < table->num_entries; k++) { - if ((table->mc_reg_table_entry[k].mclk_max == 125000) || - (table->mc_reg_table_entry[k].mclk_max == 137500)) - table->mc_reg_table_entry[k].mc_data[i] = 0; - } - break; - case mmMC_SEQ_CAS_TIMING: - for (k = 0; k < table->num_entries; k++) { - if (table->mc_reg_table_entry[k].mclk_max == 125000) - table->mc_reg_table_entry[k].mc_data[i] = - (table->mc_reg_table_entry[k].mc_data[i] & 0xFFE0FE0F) | - 0x000C0140; - else if (table->mc_reg_table_entry[k].mclk_max == 137500) - table->mc_reg_table_entry[k].mc_data[i] = - (table->mc_reg_table_entry[k].mc_data[i] & 0xFFE0FE0F) | - 0x000C0150; - } - break; - case mmMC_SEQ_MISC_TIMING: - for (k = 0; k < table->num_entries; k++) { - if (table->mc_reg_table_entry[k].mclk_max == 125000) - table->mc_reg_table_entry[k].mc_data[i] = - (table->mc_reg_table_entry[k].mc_data[i] & 0xFFFFFFE0) | - 0x00000030; - else if (table->mc_reg_table_entry[k].mclk_max == 137500) - table->mc_reg_table_entry[k].mc_data[i] = - (table->mc_reg_table_entry[k].mc_data[i] & 0xFFFFFFE0) | - 0x00000035; - } - break; - default: - break; - } - } - - WREG32(mmMC_SEQ_IO_DEBUG_INDEX, 3); - tmp = RREG32(mmMC_SEQ_IO_DEBUG_DATA); - tmp = (tmp & 0xFFF8FFFF) | (1 << 16); - WREG32(mmMC_SEQ_IO_DEBUG_INDEX, 3); - WREG32(mmMC_SEQ_IO_DEBUG_DATA, tmp); - } - - return 0; -} - -static int ci_initialize_mc_reg_table(struct amdgpu_device *adev) -{ - struct ci_power_info *pi = ci_get_pi(adev); - struct atom_mc_reg_table *table; - struct ci_mc_reg_table *ci_table = &pi->mc_reg_table; - u8 module_index = ci_get_memory_module_index(adev); - int ret; - - table = kzalloc(sizeof(struct atom_mc_reg_table), GFP_KERNEL); - if (!table) - return -ENOMEM; - - WREG32(mmMC_SEQ_RAS_TIMING_LP, RREG32(mmMC_SEQ_RAS_TIMING)); - WREG32(mmMC_SEQ_CAS_TIMING_LP, RREG32(mmMC_SEQ_CAS_TIMING)); - WREG32(mmMC_SEQ_DLL_STBY_LP, RREG32(mmMC_SEQ_DLL_STBY)); - WREG32(mmMC_SEQ_G5PDX_CMD0_LP, RREG32(mmMC_SEQ_G5PDX_CMD0)); - WREG32(mmMC_SEQ_G5PDX_CMD1_LP, RREG32(mmMC_SEQ_G5PDX_CMD1)); - WREG32(mmMC_SEQ_G5PDX_CTRL_LP, RREG32(mmMC_SEQ_G5PDX_CTRL)); - WREG32(mmMC_SEQ_PMG_DVS_CMD_LP, RREG32(mmMC_SEQ_PMG_DVS_CMD)); - WREG32(mmMC_SEQ_PMG_DVS_CTL_LP, RREG32(mmMC_SEQ_PMG_DVS_CTL)); - WREG32(mmMC_SEQ_MISC_TIMING_LP, RREG32(mmMC_SEQ_MISC_TIMING)); - WREG32(mmMC_SEQ_MISC_TIMING2_LP, RREG32(mmMC_SEQ_MISC_TIMING2)); - WREG32(mmMC_SEQ_PMG_CMD_EMRS_LP, RREG32(mmMC_PMG_CMD_EMRS)); - WREG32(mmMC_SEQ_PMG_CMD_MRS_LP, RREG32(mmMC_PMG_CMD_MRS)); - WREG32(mmMC_SEQ_PMG_CMD_MRS1_LP, RREG32(mmMC_PMG_CMD_MRS1)); - WREG32(mmMC_SEQ_WR_CTL_D0_LP, RREG32(mmMC_SEQ_WR_CTL_D0)); - WREG32(mmMC_SEQ_WR_CTL_D1_LP, RREG32(mmMC_SEQ_WR_CTL_D1)); - WREG32(mmMC_SEQ_RD_CTL_D0_LP, RREG32(mmMC_SEQ_RD_CTL_D0)); - WREG32(mmMC_SEQ_RD_CTL_D1_LP, RREG32(mmMC_SEQ_RD_CTL_D1)); - WREG32(mmMC_SEQ_PMG_TIMING_LP, RREG32(mmMC_SEQ_PMG_TIMING)); - WREG32(mmMC_SEQ_PMG_CMD_MRS2_LP, RREG32(mmMC_PMG_CMD_MRS2)); - WREG32(mmMC_SEQ_WR_CTL_2_LP, RREG32(mmMC_SEQ_WR_CTL_2)); - - ret = amdgpu_atombios_init_mc_reg_table(adev, module_index, table); - if (ret) - goto init_mc_done; - - ret = ci_copy_vbios_mc_reg_table(table, ci_table); - if (ret) - goto init_mc_done; - - ci_set_s0_mc_reg_index(ci_table); - - ret = ci_register_patching_mc_seq(adev, ci_table); - if (ret) - goto init_mc_done; - - ret = ci_set_mc_special_registers(adev, ci_table); - if (ret) - goto init_mc_done; - - ci_set_valid_flag(ci_table); - -init_mc_done: - kfree(table); - - return ret; -} - -static int ci_populate_mc_reg_addresses(struct amdgpu_device *adev, - SMU7_Discrete_MCRegisters *mc_reg_table) -{ - struct ci_power_info *pi = ci_get_pi(adev); - u32 i, j; - - for (i = 0, j = 0; j < pi->mc_reg_table.last; j++) { - if (pi->mc_reg_table.valid_flag & (1 << j)) { - if (i >= SMU7_DISCRETE_MC_REGISTER_ARRAY_SIZE) - return -EINVAL; - mc_reg_table->address[i].s0 = cpu_to_be16(pi->mc_reg_table.mc_reg_address[j].s0); - mc_reg_table->address[i].s1 = cpu_to_be16(pi->mc_reg_table.mc_reg_address[j].s1); - i++; - } - } - - mc_reg_table->last = (u8)i; - - return 0; -} - -static void ci_convert_mc_registers(const struct ci_mc_reg_entry *entry, - SMU7_Discrete_MCRegisterSet *data, - u32 num_entries, u32 valid_flag) -{ - u32 i, j; - - for (i = 0, j = 0; j < num_entries; j++) { - if (valid_flag & (1 << j)) { - data->value[i] = cpu_to_be32(entry->mc_data[j]); - i++; - } - } -} - -static void ci_convert_mc_reg_table_entry_to_smc(struct amdgpu_device *adev, - const u32 memory_clock, - SMU7_Discrete_MCRegisterSet *mc_reg_table_data) -{ - struct ci_power_info *pi = ci_get_pi(adev); - u32 i = 0; - - for(i = 0; i < pi->mc_reg_table.num_entries; i++) { - if (memory_clock <= pi->mc_reg_table.mc_reg_table_entry[i].mclk_max) - break; - } - - if ((i == pi->mc_reg_table.num_entries) && (i > 0)) - --i; - - ci_convert_mc_registers(&pi->mc_reg_table.mc_reg_table_entry[i], - mc_reg_table_data, pi->mc_reg_table.last, - pi->mc_reg_table.valid_flag); -} - -static void ci_convert_mc_reg_table_to_smc(struct amdgpu_device *adev, - SMU7_Discrete_MCRegisters *mc_reg_table) -{ - struct ci_power_info *pi = ci_get_pi(adev); - u32 i; - - for (i = 0; i < pi->dpm_table.mclk_table.count; i++) - ci_convert_mc_reg_table_entry_to_smc(adev, - pi->dpm_table.mclk_table.dpm_levels[i].value, - &mc_reg_table->data[i]); -} - -static int ci_populate_initial_mc_reg_table(struct amdgpu_device *adev) -{ - struct ci_power_info *pi = ci_get_pi(adev); - int ret; - - memset(&pi->smc_mc_reg_table, 0, sizeof(SMU7_Discrete_MCRegisters)); - - ret = ci_populate_mc_reg_addresses(adev, &pi->smc_mc_reg_table); - if (ret) - return ret; - ci_convert_mc_reg_table_to_smc(adev, &pi->smc_mc_reg_table); - - return amdgpu_ci_copy_bytes_to_smc(adev, - pi->mc_reg_table_start, - (u8 *)&pi->smc_mc_reg_table, - sizeof(SMU7_Discrete_MCRegisters), - pi->sram_end); -} - -static int ci_update_and_upload_mc_reg_table(struct amdgpu_device *adev) -{ - struct ci_power_info *pi = ci_get_pi(adev); - - if (!(pi->need_update_smu7_dpm_table & DPMTABLE_OD_UPDATE_MCLK)) - return 0; - - memset(&pi->smc_mc_reg_table, 0, sizeof(SMU7_Discrete_MCRegisters)); - - ci_convert_mc_reg_table_to_smc(adev, &pi->smc_mc_reg_table); - - return amdgpu_ci_copy_bytes_to_smc(adev, - pi->mc_reg_table_start + - offsetof(SMU7_Discrete_MCRegisters, data[0]), - (u8 *)&pi->smc_mc_reg_table.data[0], - sizeof(SMU7_Discrete_MCRegisterSet) * - pi->dpm_table.mclk_table.count, - pi->sram_end); -} - -static void ci_enable_voltage_control(struct amdgpu_device *adev) -{ - u32 tmp = RREG32_SMC(ixGENERAL_PWRMGT); - - tmp |= GENERAL_PWRMGT__VOLT_PWRMGT_EN_MASK; - WREG32_SMC(ixGENERAL_PWRMGT, tmp); -} - -static enum amdgpu_pcie_gen ci_get_maximum_link_speed(struct amdgpu_device *adev, - struct amdgpu_ps *amdgpu_state) -{ - struct ci_ps *state = ci_get_ps(amdgpu_state); - int i; - u16 pcie_speed, max_speed = 0; - - for (i = 0; i < state->performance_level_count; i++) { - pcie_speed = state->performance_levels[i].pcie_gen; - if (max_speed < pcie_speed) - max_speed = pcie_speed; - } - - return max_speed; -} - -static u16 ci_get_current_pcie_speed(struct amdgpu_device *adev) -{ - u32 speed_cntl = 0; - - speed_cntl = RREG32_PCIE(ixPCIE_LC_SPEED_CNTL) & - PCIE_LC_SPEED_CNTL__LC_CURRENT_DATA_RATE_MASK; - speed_cntl >>= PCIE_LC_SPEED_CNTL__LC_CURRENT_DATA_RATE__SHIFT; - - return (u16)speed_cntl; -} - -static int ci_get_current_pcie_lane_number(struct amdgpu_device *adev) -{ - u32 link_width = 0; - - link_width = RREG32_PCIE(ixPCIE_LC_LINK_WIDTH_CNTL) & - PCIE_LC_LINK_WIDTH_CNTL__LC_LINK_WIDTH_RD_MASK; - link_width >>= PCIE_LC_LINK_WIDTH_CNTL__LC_LINK_WIDTH_RD__SHIFT; - - switch (link_width) { - case 1: - return 1; - case 2: - return 2; - case 3: - return 4; - case 4: - return 8; - case 0: - case 6: - default: - return 16; - } -} - -static void ci_request_link_speed_change_before_state_change(struct amdgpu_device *adev, - struct amdgpu_ps *amdgpu_new_state, - struct amdgpu_ps *amdgpu_current_state) -{ - struct ci_power_info *pi = ci_get_pi(adev); - enum amdgpu_pcie_gen target_link_speed = - ci_get_maximum_link_speed(adev, amdgpu_new_state); - enum amdgpu_pcie_gen current_link_speed; - - if (pi->force_pcie_gen == AMDGPU_PCIE_GEN_INVALID) - current_link_speed = ci_get_maximum_link_speed(adev, amdgpu_current_state); - else - current_link_speed = pi->force_pcie_gen; - - pi->force_pcie_gen = AMDGPU_PCIE_GEN_INVALID; - pi->pspp_notify_required = false; - if (target_link_speed > current_link_speed) { - switch (target_link_speed) { -#ifdef CONFIG_ACPI - case AMDGPU_PCIE_GEN3: - if (amdgpu_acpi_pcie_performance_request(adev, PCIE_PERF_REQ_PECI_GEN3, false) == 0) - break; - pi->force_pcie_gen = AMDGPU_PCIE_GEN2; - if (current_link_speed == AMDGPU_PCIE_GEN2) - break; - case AMDGPU_PCIE_GEN2: - if (amdgpu_acpi_pcie_performance_request(adev, PCIE_PERF_REQ_PECI_GEN2, false) == 0) - break; -#endif - default: - pi->force_pcie_gen = ci_get_current_pcie_speed(adev); - break; - } - } else { - if (target_link_speed < current_link_speed) - pi->pspp_notify_required = true; - } -} - -static void ci_notify_link_speed_change_after_state_change(struct amdgpu_device *adev, - struct amdgpu_ps *amdgpu_new_state, - struct amdgpu_ps *amdgpu_current_state) -{ - struct ci_power_info *pi = ci_get_pi(adev); - enum amdgpu_pcie_gen target_link_speed = - ci_get_maximum_link_speed(adev, amdgpu_new_state); - u8 request; - - if (pi->pspp_notify_required) { - if (target_link_speed == AMDGPU_PCIE_GEN3) - request = PCIE_PERF_REQ_PECI_GEN3; - else if (target_link_speed == AMDGPU_PCIE_GEN2) - request = PCIE_PERF_REQ_PECI_GEN2; - else - request = PCIE_PERF_REQ_PECI_GEN1; - - if ((request == PCIE_PERF_REQ_PECI_GEN1) && - (ci_get_current_pcie_speed(adev) > 0)) - return; - -#ifdef CONFIG_ACPI - amdgpu_acpi_pcie_performance_request(adev, request, false); -#endif - } -} - -static int ci_set_private_data_variables_based_on_pptable(struct amdgpu_device *adev) -{ - struct ci_power_info *pi = ci_get_pi(adev); - struct amdgpu_clock_voltage_dependency_table *allowed_sclk_vddc_table = - &adev->pm.dpm.dyn_state.vddc_dependency_on_sclk; - struct amdgpu_clock_voltage_dependency_table *allowed_mclk_vddc_table = - &adev->pm.dpm.dyn_state.vddc_dependency_on_mclk; - struct amdgpu_clock_voltage_dependency_table *allowed_mclk_vddci_table = - &adev->pm.dpm.dyn_state.vddci_dependency_on_mclk; - - if (allowed_sclk_vddc_table == NULL) - return -EINVAL; - if (allowed_sclk_vddc_table->count < 1) - return -EINVAL; - if (allowed_mclk_vddc_table == NULL) - return -EINVAL; - if (allowed_mclk_vddc_table->count < 1) - return -EINVAL; - if (allowed_mclk_vddci_table == NULL) - return -EINVAL; - if (allowed_mclk_vddci_table->count < 1) - return -EINVAL; - - pi->min_vddc_in_pp_table = allowed_sclk_vddc_table->entries[0].v; - pi->max_vddc_in_pp_table = - allowed_sclk_vddc_table->entries[allowed_sclk_vddc_table->count - 1].v; - - pi->min_vddci_in_pp_table = allowed_mclk_vddci_table->entries[0].v; - pi->max_vddci_in_pp_table = - allowed_mclk_vddci_table->entries[allowed_mclk_vddci_table->count - 1].v; - - adev->pm.dpm.dyn_state.max_clock_voltage_on_ac.sclk = - allowed_sclk_vddc_table->entries[allowed_sclk_vddc_table->count - 1].clk; - adev->pm.dpm.dyn_state.max_clock_voltage_on_ac.mclk = - allowed_mclk_vddc_table->entries[allowed_sclk_vddc_table->count - 1].clk; - adev->pm.dpm.dyn_state.max_clock_voltage_on_ac.vddc = - allowed_sclk_vddc_table->entries[allowed_sclk_vddc_table->count - 1].v; - adev->pm.dpm.dyn_state.max_clock_voltage_on_ac.vddci = - allowed_mclk_vddci_table->entries[allowed_mclk_vddci_table->count - 1].v; - - return 0; -} - -static void ci_patch_with_vddc_leakage(struct amdgpu_device *adev, u16 *vddc) -{ - struct ci_power_info *pi = ci_get_pi(adev); - struct ci_leakage_voltage *leakage_table = &pi->vddc_leakage; - u32 leakage_index; - - for (leakage_index = 0; leakage_index < leakage_table->count; leakage_index++) { - if (leakage_table->leakage_id[leakage_index] == *vddc) { - *vddc = leakage_table->actual_voltage[leakage_index]; - break; - } - } -} - -static void ci_patch_with_vddci_leakage(struct amdgpu_device *adev, u16 *vddci) -{ - struct ci_power_info *pi = ci_get_pi(adev); - struct ci_leakage_voltage *leakage_table = &pi->vddci_leakage; - u32 leakage_index; - - for (leakage_index = 0; leakage_index < leakage_table->count; leakage_index++) { - if (leakage_table->leakage_id[leakage_index] == *vddci) { - *vddci = leakage_table->actual_voltage[leakage_index]; - break; - } - } -} - -static void ci_patch_clock_voltage_dependency_table_with_vddc_leakage(struct amdgpu_device *adev, - struct amdgpu_clock_voltage_dependency_table *table) -{ - u32 i; - - if (table) { - for (i = 0; i < table->count; i++) - ci_patch_with_vddc_leakage(adev, &table->entries[i].v); - } -} - -static void ci_patch_clock_voltage_dependency_table_with_vddci_leakage(struct amdgpu_device *adev, - struct amdgpu_clock_voltage_dependency_table *table) -{ - u32 i; - - if (table) { - for (i = 0; i < table->count; i++) - ci_patch_with_vddci_leakage(adev, &table->entries[i].v); - } -} - -static void ci_patch_vce_clock_voltage_dependency_table_with_vddc_leakage(struct amdgpu_device *adev, - struct amdgpu_vce_clock_voltage_dependency_table *table) -{ - u32 i; - - if (table) { - for (i = 0; i < table->count; i++) - ci_patch_with_vddc_leakage(adev, &table->entries[i].v); - } -} - -static void ci_patch_uvd_clock_voltage_dependency_table_with_vddc_leakage(struct amdgpu_device *adev, - struct amdgpu_uvd_clock_voltage_dependency_table *table) -{ - u32 i; - - if (table) { - for (i = 0; i < table->count; i++) - ci_patch_with_vddc_leakage(adev, &table->entries[i].v); - } -} - -static void ci_patch_vddc_phase_shed_limit_table_with_vddc_leakage(struct amdgpu_device *adev, - struct amdgpu_phase_shedding_limits_table *table) -{ - u32 i; - - if (table) { - for (i = 0; i < table->count; i++) - ci_patch_with_vddc_leakage(adev, &table->entries[i].voltage); - } -} - -static void ci_patch_clock_voltage_limits_with_vddc_leakage(struct amdgpu_device *adev, - struct amdgpu_clock_and_voltage_limits *table) -{ - if (table) { - ci_patch_with_vddc_leakage(adev, (u16 *)&table->vddc); - ci_patch_with_vddci_leakage(adev, (u16 *)&table->vddci); - } -} - -static void ci_patch_cac_leakage_table_with_vddc_leakage(struct amdgpu_device *adev, - struct amdgpu_cac_leakage_table *table) -{ - u32 i; - - if (table) { - for (i = 0; i < table->count; i++) - ci_patch_with_vddc_leakage(adev, &table->entries[i].vddc); - } -} - -static void ci_patch_dependency_tables_with_leakage(struct amdgpu_device *adev) -{ - - ci_patch_clock_voltage_dependency_table_with_vddc_leakage(adev, - &adev->pm.dpm.dyn_state.vddc_dependency_on_sclk); - ci_patch_clock_voltage_dependency_table_with_vddc_leakage(adev, - &adev->pm.dpm.dyn_state.vddc_dependency_on_mclk); - ci_patch_clock_voltage_dependency_table_with_vddc_leakage(adev, - &adev->pm.dpm.dyn_state.vddc_dependency_on_dispclk); - ci_patch_clock_voltage_dependency_table_with_vddci_leakage(adev, - &adev->pm.dpm.dyn_state.vddci_dependency_on_mclk); - ci_patch_vce_clock_voltage_dependency_table_with_vddc_leakage(adev, - &adev->pm.dpm.dyn_state.vce_clock_voltage_dependency_table); - ci_patch_uvd_clock_voltage_dependency_table_with_vddc_leakage(adev, - &adev->pm.dpm.dyn_state.uvd_clock_voltage_dependency_table); - ci_patch_clock_voltage_dependency_table_with_vddc_leakage(adev, - &adev->pm.dpm.dyn_state.samu_clock_voltage_dependency_table); - ci_patch_clock_voltage_dependency_table_with_vddc_leakage(adev, - &adev->pm.dpm.dyn_state.acp_clock_voltage_dependency_table); - ci_patch_vddc_phase_shed_limit_table_with_vddc_leakage(adev, - &adev->pm.dpm.dyn_state.phase_shedding_limits_table); - ci_patch_clock_voltage_limits_with_vddc_leakage(adev, - &adev->pm.dpm.dyn_state.max_clock_voltage_on_ac); - ci_patch_clock_voltage_limits_with_vddc_leakage(adev, - &adev->pm.dpm.dyn_state.max_clock_voltage_on_dc); - ci_patch_cac_leakage_table_with_vddc_leakage(adev, - &adev->pm.dpm.dyn_state.cac_leakage_table); - -} - -static void ci_update_current_ps(struct amdgpu_device *adev, - struct amdgpu_ps *rps) -{ - struct ci_ps *new_ps = ci_get_ps(rps); - struct ci_power_info *pi = ci_get_pi(adev); - - pi->current_rps = *rps; - pi->current_ps = *new_ps; - pi->current_rps.ps_priv = &pi->current_ps; - adev->pm.dpm.current_ps = &pi->current_rps; -} - -static void ci_update_requested_ps(struct amdgpu_device *adev, - struct amdgpu_ps *rps) -{ - struct ci_ps *new_ps = ci_get_ps(rps); - struct ci_power_info *pi = ci_get_pi(adev); - - pi->requested_rps = *rps; - pi->requested_ps = *new_ps; - pi->requested_rps.ps_priv = &pi->requested_ps; - adev->pm.dpm.requested_ps = &pi->requested_rps; -} - -static int ci_dpm_pre_set_power_state(void *handle) -{ - struct amdgpu_device *adev = (struct amdgpu_device *)handle; - struct ci_power_info *pi = ci_get_pi(adev); - struct amdgpu_ps requested_ps = *adev->pm.dpm.requested_ps; - struct amdgpu_ps *new_ps = &requested_ps; - - ci_update_requested_ps(adev, new_ps); - - ci_apply_state_adjust_rules(adev, &pi->requested_rps); - - return 0; -} - -static void ci_dpm_post_set_power_state(void *handle) -{ - struct amdgpu_device *adev = (struct amdgpu_device *)handle; - struct ci_power_info *pi = ci_get_pi(adev); - struct amdgpu_ps *new_ps = &pi->requested_rps; - - ci_update_current_ps(adev, new_ps); -} - - -static void ci_dpm_setup_asic(struct amdgpu_device *adev) -{ - ci_read_clock_registers(adev); - ci_enable_acpi_power_management(adev); - ci_init_sclk_t(adev); -} - -static int ci_dpm_enable(struct amdgpu_device *adev) -{ - struct ci_power_info *pi = ci_get_pi(adev); - struct amdgpu_ps *boot_ps = adev->pm.dpm.boot_ps; - int ret; - - if (pi->voltage_control != CISLANDS_VOLTAGE_CONTROL_NONE) { - ci_enable_voltage_control(adev); - ret = ci_construct_voltage_tables(adev); - if (ret) { - DRM_ERROR("ci_construct_voltage_tables failed\n"); - return ret; - } - } - if (pi->caps_dynamic_ac_timing) { - ret = ci_initialize_mc_reg_table(adev); - if (ret) - pi->caps_dynamic_ac_timing = false; - } - if (pi->dynamic_ss) - ci_enable_spread_spectrum(adev, true); - if (pi->thermal_protection) - ci_enable_thermal_protection(adev, true); - ci_program_sstp(adev); - ci_enable_display_gap(adev); - ci_program_vc(adev); - ret = ci_upload_firmware(adev); - if (ret) { - DRM_ERROR("ci_upload_firmware failed\n"); - return ret; - } - ret = ci_process_firmware_header(adev); - if (ret) { - DRM_ERROR("ci_process_firmware_header failed\n"); - return ret; - } - ret = ci_initial_switch_from_arb_f0_to_f1(adev); - if (ret) { - DRM_ERROR("ci_initial_switch_from_arb_f0_to_f1 failed\n"); - return ret; - } - ret = ci_init_smc_table(adev); - if (ret) { - DRM_ERROR("ci_init_smc_table failed\n"); - return ret; - } - ret = ci_init_arb_table_index(adev); - if (ret) { - DRM_ERROR("ci_init_arb_table_index failed\n"); - return ret; - } - if (pi->caps_dynamic_ac_timing) { - ret = ci_populate_initial_mc_reg_table(adev); - if (ret) { - DRM_ERROR("ci_populate_initial_mc_reg_table failed\n"); - return ret; - } - } - ret = ci_populate_pm_base(adev); - if (ret) { - DRM_ERROR("ci_populate_pm_base failed\n"); - return ret; - } - ci_dpm_start_smc(adev); - ci_enable_vr_hot_gpio_interrupt(adev); - ret = ci_notify_smc_display_change(adev, false); - if (ret) { - DRM_ERROR("ci_notify_smc_display_change failed\n"); - return ret; - } - ci_enable_sclk_control(adev, true); - ret = ci_enable_ulv(adev, true); - if (ret) { - DRM_ERROR("ci_enable_ulv failed\n"); - return ret; - } - ret = ci_enable_ds_master_switch(adev, true); - if (ret) { - DRM_ERROR("ci_enable_ds_master_switch failed\n"); - return ret; - } - ret = ci_start_dpm(adev); - if (ret) { - DRM_ERROR("ci_start_dpm failed\n"); - return ret; - } - ret = ci_enable_didt(adev, true); - if (ret) { - DRM_ERROR("ci_enable_didt failed\n"); - return ret; - } - ret = ci_enable_smc_cac(adev, true); - if (ret) { - DRM_ERROR("ci_enable_smc_cac failed\n"); - return ret; - } - ret = ci_enable_power_containment(adev, true); - if (ret) { - DRM_ERROR("ci_enable_power_containment failed\n"); - return ret; - } - - ret = ci_power_control_set_level(adev); - if (ret) { - DRM_ERROR("ci_power_control_set_level failed\n"); - return ret; - } - - ci_enable_auto_throttle_source(adev, AMDGPU_DPM_AUTO_THROTTLE_SRC_THERMAL, true); - - ret = ci_enable_thermal_based_sclk_dpm(adev, true); - if (ret) { - DRM_ERROR("ci_enable_thermal_based_sclk_dpm failed\n"); - return ret; - } - - ci_thermal_start_thermal_controller(adev); - - ci_update_current_ps(adev, boot_ps); - - return 0; -} - -static void ci_dpm_disable(struct amdgpu_device *adev) -{ - struct ci_power_info *pi = ci_get_pi(adev); - struct amdgpu_ps *boot_ps = adev->pm.dpm.boot_ps; - - amdgpu_irq_put(adev, &adev->pm.dpm.thermal.irq, - AMDGPU_THERMAL_IRQ_LOW_TO_HIGH); - amdgpu_irq_put(adev, &adev->pm.dpm.thermal.irq, - AMDGPU_THERMAL_IRQ_HIGH_TO_LOW); - - ci_dpm_powergate_uvd(adev, true); - - if (!amdgpu_ci_is_smc_running(adev)) - return; - - ci_thermal_stop_thermal_controller(adev); - - if (pi->thermal_protection) - ci_enable_thermal_protection(adev, false); - ci_enable_power_containment(adev, false); - ci_enable_smc_cac(adev, false); - ci_enable_didt(adev, false); - ci_enable_spread_spectrum(adev, false); - ci_enable_auto_throttle_source(adev, AMDGPU_DPM_AUTO_THROTTLE_SRC_THERMAL, false); - ci_stop_dpm(adev); - ci_enable_ds_master_switch(adev, false); - ci_enable_ulv(adev, false); - ci_clear_vc(adev); - ci_reset_to_default(adev); - ci_dpm_stop_smc(adev); - ci_force_switch_to_arb_f0(adev); - ci_enable_thermal_based_sclk_dpm(adev, false); - - ci_update_current_ps(adev, boot_ps); -} - -static int ci_dpm_set_power_state(void *handle) -{ - struct amdgpu_device *adev = (struct amdgpu_device *)handle; - struct ci_power_info *pi = ci_get_pi(adev); - struct amdgpu_ps *new_ps = &pi->requested_rps; - struct amdgpu_ps *old_ps = &pi->current_rps; - int ret; - - ci_find_dpm_states_clocks_in_dpm_table(adev, new_ps); - if (pi->pcie_performance_request) - ci_request_link_speed_change_before_state_change(adev, new_ps, old_ps); - ret = ci_freeze_sclk_mclk_dpm(adev); - if (ret) { - DRM_ERROR("ci_freeze_sclk_mclk_dpm failed\n"); - return ret; - } - ret = ci_populate_and_upload_sclk_mclk_dpm_levels(adev, new_ps); - if (ret) { - DRM_ERROR("ci_populate_and_upload_sclk_mclk_dpm_levels failed\n"); - return ret; - } - ret = ci_generate_dpm_level_enable_mask(adev, new_ps); - if (ret) { - DRM_ERROR("ci_generate_dpm_level_enable_mask failed\n"); - return ret; - } - - ret = ci_update_vce_dpm(adev, new_ps, old_ps); - if (ret) { - DRM_ERROR("ci_update_vce_dpm failed\n"); - return ret; - } - - ret = ci_update_sclk_t(adev); - if (ret) { - DRM_ERROR("ci_update_sclk_t failed\n"); - return ret; - } - if (pi->caps_dynamic_ac_timing) { - ret = ci_update_and_upload_mc_reg_table(adev); - if (ret) { - DRM_ERROR("ci_update_and_upload_mc_reg_table failed\n"); - return ret; - } - } - ret = ci_program_memory_timing_parameters(adev); - if (ret) { - DRM_ERROR("ci_program_memory_timing_parameters failed\n"); - return ret; - } - ret = ci_unfreeze_sclk_mclk_dpm(adev); - if (ret) { - DRM_ERROR("ci_unfreeze_sclk_mclk_dpm failed\n"); - return ret; - } - ret = ci_upload_dpm_level_enable_mask(adev); - if (ret) { - DRM_ERROR("ci_upload_dpm_level_enable_mask failed\n"); - return ret; - } - if (pi->pcie_performance_request) - ci_notify_link_speed_change_after_state_change(adev, new_ps, old_ps); - - return 0; -} - -#if 0 -static void ci_dpm_reset_asic(struct amdgpu_device *adev) -{ - ci_set_boot_state(adev); -} -#endif - -static void ci_dpm_display_configuration_changed(void *handle) -{ - struct amdgpu_device *adev = (struct amdgpu_device *)handle; - - ci_program_display_gap(adev); -} - -union power_info { - struct _ATOM_POWERPLAY_INFO info; - struct _ATOM_POWERPLAY_INFO_V2 info_2; - struct _ATOM_POWERPLAY_INFO_V3 info_3; - struct _ATOM_PPLIB_POWERPLAYTABLE pplib; - struct _ATOM_PPLIB_POWERPLAYTABLE2 pplib2; - struct _ATOM_PPLIB_POWERPLAYTABLE3 pplib3; -}; - -union pplib_clock_info { - struct _ATOM_PPLIB_R600_CLOCK_INFO r600; - struct _ATOM_PPLIB_RS780_CLOCK_INFO rs780; - struct _ATOM_PPLIB_EVERGREEN_CLOCK_INFO evergreen; - struct _ATOM_PPLIB_SUMO_CLOCK_INFO sumo; - struct _ATOM_PPLIB_SI_CLOCK_INFO si; - struct _ATOM_PPLIB_CI_CLOCK_INFO ci; -}; - -union pplib_power_state { - struct _ATOM_PPLIB_STATE v1; - struct _ATOM_PPLIB_STATE_V2 v2; -}; - -static void ci_parse_pplib_non_clock_info(struct amdgpu_device *adev, - struct amdgpu_ps *rps, - struct _ATOM_PPLIB_NONCLOCK_INFO *non_clock_info, - u8 table_rev) -{ - rps->caps = le32_to_cpu(non_clock_info->ulCapsAndSettings); - rps->class = le16_to_cpu(non_clock_info->usClassification); - rps->class2 = le16_to_cpu(non_clock_info->usClassification2); - - if (ATOM_PPLIB_NONCLOCKINFO_VER1 < table_rev) { - rps->vclk = le32_to_cpu(non_clock_info->ulVCLK); - rps->dclk = le32_to_cpu(non_clock_info->ulDCLK); - } else { - rps->vclk = 0; - rps->dclk = 0; - } - - if (rps->class & ATOM_PPLIB_CLASSIFICATION_BOOT) - adev->pm.dpm.boot_ps = rps; - if (rps->class & ATOM_PPLIB_CLASSIFICATION_UVDSTATE) - adev->pm.dpm.uvd_ps = rps; -} - -static void ci_parse_pplib_clock_info(struct amdgpu_device *adev, - struct amdgpu_ps *rps, int index, - union pplib_clock_info *clock_info) -{ - struct ci_power_info *pi = ci_get_pi(adev); - struct ci_ps *ps = ci_get_ps(rps); - struct ci_pl *pl = &ps->performance_levels[index]; - - ps->performance_level_count = index + 1; - - pl->sclk = le16_to_cpu(clock_info->ci.usEngineClockLow); - pl->sclk |= clock_info->ci.ucEngineClockHigh << 16; - pl->mclk = le16_to_cpu(clock_info->ci.usMemoryClockLow); - pl->mclk |= clock_info->ci.ucMemoryClockHigh << 16; - - pl->pcie_gen = amdgpu_get_pcie_gen_support(adev, - pi->sys_pcie_mask, - pi->vbios_boot_state.pcie_gen_bootup_value, - clock_info->ci.ucPCIEGen); - pl->pcie_lane = amdgpu_get_pcie_lane_support(adev, - pi->vbios_boot_state.pcie_lane_bootup_value, - le16_to_cpu(clock_info->ci.usPCIELane)); - - if (rps->class & ATOM_PPLIB_CLASSIFICATION_ACPI) { - pi->acpi_pcie_gen = pl->pcie_gen; - } - - if (rps->class2 & ATOM_PPLIB_CLASSIFICATION2_ULV) { - pi->ulv.supported = true; - pi->ulv.pl = *pl; - pi->ulv.cg_ulv_parameter = CISLANDS_CGULVPARAMETER_DFLT; - } - - /* patch up boot state */ - if (rps->class & ATOM_PPLIB_CLASSIFICATION_BOOT) { - pl->mclk = pi->vbios_boot_state.mclk_bootup_value; - pl->sclk = pi->vbios_boot_state.sclk_bootup_value; - pl->pcie_gen = pi->vbios_boot_state.pcie_gen_bootup_value; - pl->pcie_lane = pi->vbios_boot_state.pcie_lane_bootup_value; - } - - switch (rps->class & ATOM_PPLIB_CLASSIFICATION_UI_MASK) { - case ATOM_PPLIB_CLASSIFICATION_UI_BATTERY: - pi->use_pcie_powersaving_levels = true; - if (pi->pcie_gen_powersaving.max < pl->pcie_gen) - pi->pcie_gen_powersaving.max = pl->pcie_gen; - if (pi->pcie_gen_powersaving.min > pl->pcie_gen) - pi->pcie_gen_powersaving.min = pl->pcie_gen; - if (pi->pcie_lane_powersaving.max < pl->pcie_lane) - pi->pcie_lane_powersaving.max = pl->pcie_lane; - if (pi->pcie_lane_powersaving.min > pl->pcie_lane) - pi->pcie_lane_powersaving.min = pl->pcie_lane; - break; - case ATOM_PPLIB_CLASSIFICATION_UI_PERFORMANCE: - pi->use_pcie_performance_levels = true; - if (pi->pcie_gen_performance.max < pl->pcie_gen) - pi->pcie_gen_performance.max = pl->pcie_gen; - if (pi->pcie_gen_performance.min > pl->pcie_gen) - pi->pcie_gen_performance.min = pl->pcie_gen; - if (pi->pcie_lane_performance.max < pl->pcie_lane) - pi->pcie_lane_performance.max = pl->pcie_lane; - if (pi->pcie_lane_performance.min > pl->pcie_lane) - pi->pcie_lane_performance.min = pl->pcie_lane; - break; - default: - break; - } -} - -static int ci_parse_power_table(struct amdgpu_device *adev) -{ - struct amdgpu_mode_info *mode_info = &adev->mode_info; - struct _ATOM_PPLIB_NONCLOCK_INFO *non_clock_info; - union pplib_power_state *power_state; - int i, j, k, non_clock_array_index, clock_array_index; - union pplib_clock_info *clock_info; - struct _StateArray *state_array; - struct _ClockInfoArray *clock_info_array; - struct _NonClockInfoArray *non_clock_info_array; - union power_info *power_info; - int index = GetIndexIntoMasterTable(DATA, PowerPlayInfo); - u16 data_offset; - u8 frev, crev; - u8 *power_state_offset; - struct ci_ps *ps; - - if (!amdgpu_atom_parse_data_header(mode_info->atom_context, index, NULL, - &frev, &crev, &data_offset)) - return -EINVAL; - power_info = (union power_info *)(mode_info->atom_context->bios + data_offset); - - amdgpu_add_thermal_controller(adev); - - state_array = (struct _StateArray *) - (mode_info->atom_context->bios + data_offset + - le16_to_cpu(power_info->pplib.usStateArrayOffset)); - clock_info_array = (struct _ClockInfoArray *) - (mode_info->atom_context->bios + data_offset + - le16_to_cpu(power_info->pplib.usClockInfoArrayOffset)); - non_clock_info_array = (struct _NonClockInfoArray *) - (mode_info->atom_context->bios + data_offset + - le16_to_cpu(power_info->pplib.usNonClockInfoArrayOffset)); - - adev->pm.dpm.ps = kcalloc(state_array->ucNumEntries, - sizeof(struct amdgpu_ps), - GFP_KERNEL); - if (!adev->pm.dpm.ps) - return -ENOMEM; - power_state_offset = (u8 *)state_array->states; - for (i = 0; i < state_array->ucNumEntries; i++) { - u8 *idx; - power_state = (union pplib_power_state *)power_state_offset; - non_clock_array_index = power_state->v2.nonClockInfoIndex; - non_clock_info = (struct _ATOM_PPLIB_NONCLOCK_INFO *) - &non_clock_info_array->nonClockInfo[non_clock_array_index]; - ps = kzalloc(sizeof(struct ci_ps), GFP_KERNEL); - if (ps == NULL) { - kfree(adev->pm.dpm.ps); - return -ENOMEM; - } - adev->pm.dpm.ps[i].ps_priv = ps; - ci_parse_pplib_non_clock_info(adev, &adev->pm.dpm.ps[i], - non_clock_info, - non_clock_info_array->ucEntrySize); - k = 0; - idx = (u8 *)&power_state->v2.clockInfoIndex[0]; - for (j = 0; j < power_state->v2.ucNumDPMLevels; j++) { - clock_array_index = idx[j]; - if (clock_array_index >= clock_info_array->ucNumEntries) - continue; - if (k >= CISLANDS_MAX_HARDWARE_POWERLEVELS) - break; - clock_info = (union pplib_clock_info *) - ((u8 *)&clock_info_array->clockInfo[0] + - (clock_array_index * clock_info_array->ucEntrySize)); - ci_parse_pplib_clock_info(adev, - &adev->pm.dpm.ps[i], k, - clock_info); - k++; - } - power_state_offset += 2 + power_state->v2.ucNumDPMLevels; - } - adev->pm.dpm.num_ps = state_array->ucNumEntries; - - /* fill in the vce power states */ - for (i = 0; i < adev->pm.dpm.num_of_vce_states; i++) { - u32 sclk, mclk; - clock_array_index = adev->pm.dpm.vce_states[i].clk_idx; - clock_info = (union pplib_clock_info *) - &clock_info_array->clockInfo[clock_array_index * clock_info_array->ucEntrySize]; - sclk = le16_to_cpu(clock_info->ci.usEngineClockLow); - sclk |= clock_info->ci.ucEngineClockHigh << 16; - mclk = le16_to_cpu(clock_info->ci.usMemoryClockLow); - mclk |= clock_info->ci.ucMemoryClockHigh << 16; - adev->pm.dpm.vce_states[i].sclk = sclk; - adev->pm.dpm.vce_states[i].mclk = mclk; - } - - return 0; -} - -static int ci_get_vbios_boot_values(struct amdgpu_device *adev, - struct ci_vbios_boot_state *boot_state) -{ - struct amdgpu_mode_info *mode_info = &adev->mode_info; - int index = GetIndexIntoMasterTable(DATA, FirmwareInfo); - ATOM_FIRMWARE_INFO_V2_2 *firmware_info; - u8 frev, crev; - u16 data_offset; - - if (amdgpu_atom_parse_data_header(mode_info->atom_context, index, NULL, - &frev, &crev, &data_offset)) { - firmware_info = - (ATOM_FIRMWARE_INFO_V2_2 *)(mode_info->atom_context->bios + - data_offset); - boot_state->mvdd_bootup_value = le16_to_cpu(firmware_info->usBootUpMVDDCVoltage); - boot_state->vddc_bootup_value = le16_to_cpu(firmware_info->usBootUpVDDCVoltage); - boot_state->vddci_bootup_value = le16_to_cpu(firmware_info->usBootUpVDDCIVoltage); - boot_state->pcie_gen_bootup_value = ci_get_current_pcie_speed(adev); - boot_state->pcie_lane_bootup_value = ci_get_current_pcie_lane_number(adev); - boot_state->sclk_bootup_value = le32_to_cpu(firmware_info->ulDefaultEngineClock); - boot_state->mclk_bootup_value = le32_to_cpu(firmware_info->ulDefaultMemoryClock); - - return 0; - } - return -EINVAL; -} - -static void ci_dpm_fini(struct amdgpu_device *adev) -{ - int i; - - for (i = 0; i < adev->pm.dpm.num_ps; i++) { - kfree(adev->pm.dpm.ps[i].ps_priv); - } - kfree(adev->pm.dpm.ps); - kfree(adev->pm.dpm.priv); - kfree(adev->pm.dpm.dyn_state.vddc_dependency_on_dispclk.entries); - amdgpu_free_extended_power_table(adev); -} - -/** - * ci_dpm_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 ci_dpm_init_microcode(struct amdgpu_device *adev) -{ - const char *chip_name; - char fw_name[30]; - int err; - - DRM_DEBUG("\n"); - - switch (adev->asic_type) { - case CHIP_BONAIRE: - if ((adev->pdev->revision == 0x80) || - (adev->pdev->revision == 0x81) || - (adev->pdev->device == 0x665f)) - chip_name = "bonaire_k"; - else - chip_name = "bonaire"; - break; - case CHIP_HAWAII: - if (adev->pdev->revision == 0x80) - chip_name = "hawaii_k"; - else - chip_name = "hawaii"; - break; - case CHIP_KAVERI: - case CHIP_KABINI: - case CHIP_MULLINS: - default: BUG(); - } - - snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_smc.bin", chip_name); - err = request_firmware(&adev->pm.fw, fw_name, adev->dev); - if (err) - goto out; - err = amdgpu_ucode_validate(adev->pm.fw); - -out: - if (err) { - pr_err("cik_smc: Failed to load firmware \"%s\"\n", fw_name); - release_firmware(adev->pm.fw); - adev->pm.fw = NULL; - } - return err; -} - -static int ci_dpm_init(struct amdgpu_device *adev) -{ - int index = GetIndexIntoMasterTable(DATA, ASIC_InternalSS_Info); - SMU7_Discrete_DpmTable *dpm_table; - struct amdgpu_gpio_rec gpio; - u16 data_offset, size; - u8 frev, crev; - struct ci_power_info *pi; - int ret; - - pi = kzalloc(sizeof(struct ci_power_info), GFP_KERNEL); - if (pi == NULL) - return -ENOMEM; - adev->pm.dpm.priv = pi; - - pi->sys_pcie_mask = - adev->pm.pcie_gen_mask & CAIL_PCIE_LINK_SPEED_SUPPORT_MASK; - - pi->force_pcie_gen = AMDGPU_PCIE_GEN_INVALID; - - pi->pcie_gen_performance.max = AMDGPU_PCIE_GEN1; - pi->pcie_gen_performance.min = AMDGPU_PCIE_GEN3; - pi->pcie_gen_powersaving.max = AMDGPU_PCIE_GEN1; - pi->pcie_gen_powersaving.min = AMDGPU_PCIE_GEN3; - - pi->pcie_lane_performance.max = 0; - pi->pcie_lane_performance.min = 16; - pi->pcie_lane_powersaving.max = 0; - pi->pcie_lane_powersaving.min = 16; - - ret = ci_get_vbios_boot_values(adev, &pi->vbios_boot_state); - if (ret) { - ci_dpm_fini(adev); - return ret; - } - - ret = amdgpu_get_platform_caps(adev); - if (ret) { - ci_dpm_fini(adev); - return ret; - } - - ret = amdgpu_parse_extended_power_table(adev); - if (ret) { - ci_dpm_fini(adev); - return ret; - } - - ret = ci_parse_power_table(adev); - if (ret) { - ci_dpm_fini(adev); - return ret; - } - - pi->dll_default_on = false; - pi->sram_end = SMC_RAM_END; - - pi->activity_target[0] = CISLAND_TARGETACTIVITY_DFLT; - pi->activity_target[1] = CISLAND_TARGETACTIVITY_DFLT; - pi->activity_target[2] = CISLAND_TARGETACTIVITY_DFLT; - pi->activity_target[3] = CISLAND_TARGETACTIVITY_DFLT; - pi->activity_target[4] = CISLAND_TARGETACTIVITY_DFLT; - pi->activity_target[5] = CISLAND_TARGETACTIVITY_DFLT; - pi->activity_target[6] = CISLAND_TARGETACTIVITY_DFLT; - pi->activity_target[7] = CISLAND_TARGETACTIVITY_DFLT; - - pi->mclk_activity_target = CISLAND_MCLK_TARGETACTIVITY_DFLT; - - pi->sclk_dpm_key_disabled = 0; - pi->mclk_dpm_key_disabled = 0; - pi->pcie_dpm_key_disabled = 0; - pi->thermal_sclk_dpm_enabled = 0; - - if (adev->powerplay.pp_feature & PP_SCLK_DEEP_SLEEP_MASK) - pi->caps_sclk_ds = true; - else - pi->caps_sclk_ds = false; - - pi->mclk_strobe_mode_threshold = 40000; - pi->mclk_stutter_mode_threshold = 40000; - pi->mclk_edc_enable_threshold = 40000; - pi->mclk_edc_wr_enable_threshold = 40000; - - ci_initialize_powertune_defaults(adev); - - pi->caps_fps = false; - - pi->caps_sclk_throttle_low_notification = false; - - pi->caps_uvd_dpm = true; - pi->caps_vce_dpm = true; - - ci_get_leakage_voltages(adev); - ci_patch_dependency_tables_with_leakage(adev); - ci_set_private_data_variables_based_on_pptable(adev); - - adev->pm.dpm.dyn_state.vddc_dependency_on_dispclk.entries = - kcalloc(4, - sizeof(struct amdgpu_clock_voltage_dependency_entry), - GFP_KERNEL); - if (!adev->pm.dpm.dyn_state.vddc_dependency_on_dispclk.entries) { - ci_dpm_fini(adev); - return -ENOMEM; - } - adev->pm.dpm.dyn_state.vddc_dependency_on_dispclk.count = 4; - adev->pm.dpm.dyn_state.vddc_dependency_on_dispclk.entries[0].clk = 0; - adev->pm.dpm.dyn_state.vddc_dependency_on_dispclk.entries[0].v = 0; - adev->pm.dpm.dyn_state.vddc_dependency_on_dispclk.entries[1].clk = 36000; - adev->pm.dpm.dyn_state.vddc_dependency_on_dispclk.entries[1].v = 720; - adev->pm.dpm.dyn_state.vddc_dependency_on_dispclk.entries[2].clk = 54000; - adev->pm.dpm.dyn_state.vddc_dependency_on_dispclk.entries[2].v = 810; - adev->pm.dpm.dyn_state.vddc_dependency_on_dispclk.entries[3].clk = 72000; - adev->pm.dpm.dyn_state.vddc_dependency_on_dispclk.entries[3].v = 900; - - adev->pm.dpm.dyn_state.mclk_sclk_ratio = 4; - adev->pm.dpm.dyn_state.sclk_mclk_delta = 15000; - adev->pm.dpm.dyn_state.vddc_vddci_delta = 200; - - adev->pm.dpm.dyn_state.valid_sclk_values.count = 0; - adev->pm.dpm.dyn_state.valid_sclk_values.values = NULL; - adev->pm.dpm.dyn_state.valid_mclk_values.count = 0; - adev->pm.dpm.dyn_state.valid_mclk_values.values = NULL; - - if (adev->asic_type == CHIP_HAWAII) { - pi->thermal_temp_setting.temperature_low = 94500; - pi->thermal_temp_setting.temperature_high = 95000; - pi->thermal_temp_setting.temperature_shutdown = 104000; - } else { - pi->thermal_temp_setting.temperature_low = 99500; - pi->thermal_temp_setting.temperature_high = 100000; - pi->thermal_temp_setting.temperature_shutdown = 104000; - } - - pi->uvd_enabled = false; - - dpm_table = &pi->smc_state_table; - - gpio = amdgpu_atombios_lookup_gpio(adev, VDDC_VRHOT_GPIO_PINID); - if (gpio.valid) { - dpm_table->VRHotGpio = gpio.shift; - adev->pm.dpm.platform_caps |= ATOM_PP_PLATFORM_CAP_REGULATOR_HOT; - } else { - dpm_table->VRHotGpio = CISLANDS_UNUSED_GPIO_PIN; - adev->pm.dpm.platform_caps &= ~ATOM_PP_PLATFORM_CAP_REGULATOR_HOT; - } - - gpio = amdgpu_atombios_lookup_gpio(adev, PP_AC_DC_SWITCH_GPIO_PINID); - if (gpio.valid) { - dpm_table->AcDcGpio = gpio.shift; - adev->pm.dpm.platform_caps |= ATOM_PP_PLATFORM_CAP_HARDWAREDC; - } else { - dpm_table->AcDcGpio = CISLANDS_UNUSED_GPIO_PIN; - adev->pm.dpm.platform_caps &= ~ATOM_PP_PLATFORM_CAP_HARDWAREDC; - } - - gpio = amdgpu_atombios_lookup_gpio(adev, VDDC_PCC_GPIO_PINID); - if (gpio.valid) { - u32 tmp = RREG32_SMC(ixCNB_PWRMGT_CNTL); - - switch (gpio.shift) { - case 0: - tmp &= ~CNB_PWRMGT_CNTL__GNB_SLOW_MODE_MASK; - tmp |= 1 << CNB_PWRMGT_CNTL__GNB_SLOW_MODE__SHIFT; - break; - case 1: - tmp &= ~CNB_PWRMGT_CNTL__GNB_SLOW_MODE_MASK; - tmp |= 2 << CNB_PWRMGT_CNTL__GNB_SLOW_MODE__SHIFT; - break; - case 2: - tmp |= CNB_PWRMGT_CNTL__GNB_SLOW_MASK; - break; - case 3: - tmp |= CNB_PWRMGT_CNTL__FORCE_NB_PS1_MASK; - break; - case 4: - tmp |= CNB_PWRMGT_CNTL__DPM_ENABLED_MASK; - break; - default: - DRM_INFO("Invalid PCC GPIO: %u!\n", gpio.shift); - break; - } - WREG32_SMC(ixCNB_PWRMGT_CNTL, tmp); - } - - pi->voltage_control = CISLANDS_VOLTAGE_CONTROL_NONE; - pi->vddci_control = CISLANDS_VOLTAGE_CONTROL_NONE; - pi->mvdd_control = CISLANDS_VOLTAGE_CONTROL_NONE; - if (amdgpu_atombios_is_voltage_gpio(adev, VOLTAGE_TYPE_VDDC, VOLTAGE_OBJ_GPIO_LUT)) - pi->voltage_control = CISLANDS_VOLTAGE_CONTROL_BY_GPIO; - else if (amdgpu_atombios_is_voltage_gpio(adev, VOLTAGE_TYPE_VDDC, VOLTAGE_OBJ_SVID2)) - pi->voltage_control = CISLANDS_VOLTAGE_CONTROL_BY_SVID2; - - if (adev->pm.dpm.platform_caps & ATOM_PP_PLATFORM_CAP_VDDCI_CONTROL) { - if (amdgpu_atombios_is_voltage_gpio(adev, VOLTAGE_TYPE_VDDCI, VOLTAGE_OBJ_GPIO_LUT)) - pi->vddci_control = CISLANDS_VOLTAGE_CONTROL_BY_GPIO; - else if (amdgpu_atombios_is_voltage_gpio(adev, VOLTAGE_TYPE_VDDCI, VOLTAGE_OBJ_SVID2)) - pi->vddci_control = CISLANDS_VOLTAGE_CONTROL_BY_SVID2; - else - adev->pm.dpm.platform_caps &= ~ATOM_PP_PLATFORM_CAP_VDDCI_CONTROL; - } - - if (adev->pm.dpm.platform_caps & ATOM_PP_PLATFORM_CAP_MVDDCONTROL) { - if (amdgpu_atombios_is_voltage_gpio(adev, VOLTAGE_TYPE_MVDDC, VOLTAGE_OBJ_GPIO_LUT)) - pi->mvdd_control = CISLANDS_VOLTAGE_CONTROL_BY_GPIO; - else if (amdgpu_atombios_is_voltage_gpio(adev, VOLTAGE_TYPE_MVDDC, VOLTAGE_OBJ_SVID2)) - pi->mvdd_control = CISLANDS_VOLTAGE_CONTROL_BY_SVID2; - else - adev->pm.dpm.platform_caps &= ~ATOM_PP_PLATFORM_CAP_MVDDCONTROL; - } - - pi->vddc_phase_shed_control = true; - -#if defined(CONFIG_ACPI) - pi->pcie_performance_request = - amdgpu_acpi_is_pcie_performance_request_supported(adev); -#else - pi->pcie_performance_request = false; -#endif - - if (amdgpu_atom_parse_data_header(adev->mode_info.atom_context, index, &size, - &frev, &crev, &data_offset)) { - pi->caps_sclk_ss_support = true; - pi->caps_mclk_ss_support = true; - pi->dynamic_ss = true; - } else { - pi->caps_sclk_ss_support = false; - pi->caps_mclk_ss_support = false; - pi->dynamic_ss = true; - } - - if (adev->pm.int_thermal_type != THERMAL_TYPE_NONE) - pi->thermal_protection = true; - else - pi->thermal_protection = false; - - pi->caps_dynamic_ac_timing = true; - - pi->uvd_power_gated = true; - - /* make sure dc limits are valid */ - if ((adev->pm.dpm.dyn_state.max_clock_voltage_on_dc.sclk == 0) || - (adev->pm.dpm.dyn_state.max_clock_voltage_on_dc.mclk == 0)) - adev->pm.dpm.dyn_state.max_clock_voltage_on_dc = - adev->pm.dpm.dyn_state.max_clock_voltage_on_ac; - - pi->fan_ctrl_is_in_default_mode = true; - - return 0; -} - -static void -ci_dpm_debugfs_print_current_performance_level(void *handle, - struct seq_file *m) -{ - struct amdgpu_device *adev = (struct amdgpu_device *)handle; - struct ci_power_info *pi = ci_get_pi(adev); - struct amdgpu_ps *rps = &pi->current_rps; - u32 sclk = ci_get_average_sclk_freq(adev); - u32 mclk = ci_get_average_mclk_freq(adev); - u32 activity_percent = 50; - int ret; - - ret = ci_read_smc_soft_register(adev, offsetof(SMU7_SoftRegisters, AverageGraphicsA), - &activity_percent); - - if (ret == 0) { - activity_percent += 0x80; - activity_percent >>= 8; - activity_percent = activity_percent > 100 ? 100 : activity_percent; - } - - seq_printf(m, "uvd %sabled\n", pi->uvd_power_gated ? "dis" : "en"); - seq_printf(m, "vce %sabled\n", rps->vce_active ? "en" : "dis"); - seq_printf(m, "power level avg sclk: %u mclk: %u\n", - sclk, mclk); - seq_printf(m, "GPU load: %u %%\n", activity_percent); -} - -static void ci_dpm_print_power_state(void *handle, void *current_ps) -{ - struct amdgpu_ps *rps = (struct amdgpu_ps *)current_ps; - struct ci_ps *ps = ci_get_ps(rps); - struct ci_pl *pl; - int i; - struct amdgpu_device *adev = (struct amdgpu_device *)handle; - - amdgpu_dpm_print_class_info(rps->class, rps->class2); - amdgpu_dpm_print_cap_info(rps->caps); - printk("\tuvd vclk: %d dclk: %d\n", rps->vclk, rps->dclk); - for (i = 0; i < ps->performance_level_count; i++) { - pl = &ps->performance_levels[i]; - printk("\t\tpower level %d sclk: %u mclk: %u pcie gen: %u pcie lanes: %u\n", - i, pl->sclk, pl->mclk, pl->pcie_gen + 1, pl->pcie_lane); - } - amdgpu_dpm_print_ps_status(adev, rps); -} - -static inline bool ci_are_power_levels_equal(const struct ci_pl *ci_cpl1, - const struct ci_pl *ci_cpl2) -{ - return ((ci_cpl1->mclk == ci_cpl2->mclk) && - (ci_cpl1->sclk == ci_cpl2->sclk) && - (ci_cpl1->pcie_gen == ci_cpl2->pcie_gen) && - (ci_cpl1->pcie_lane == ci_cpl2->pcie_lane)); -} - -static int ci_check_state_equal(void *handle, - void *current_ps, - void *request_ps, - bool *equal) -{ - struct ci_ps *ci_cps; - struct ci_ps *ci_rps; - int i; - struct amdgpu_ps *cps = (struct amdgpu_ps *)current_ps; - struct amdgpu_ps *rps = (struct amdgpu_ps *)request_ps; - struct amdgpu_device *adev = (struct amdgpu_device *)handle; - - if (adev == NULL || cps == NULL || rps == NULL || equal == NULL) - return -EINVAL; - - ci_cps = ci_get_ps((struct amdgpu_ps *)cps); - ci_rps = ci_get_ps((struct amdgpu_ps *)rps); - - if (ci_cps == NULL) { - *equal = false; - return 0; - } - - if (ci_cps->performance_level_count != ci_rps->performance_level_count) { - - *equal = false; - return 0; - } - - for (i = 0; i < ci_cps->performance_level_count; i++) { - if (!ci_are_power_levels_equal(&(ci_cps->performance_levels[i]), - &(ci_rps->performance_levels[i]))) { - *equal = false; - return 0; - } - } - - /* If all performance levels are the same try to use the UVD clocks to break the tie.*/ - *equal = ((cps->vclk == rps->vclk) && (cps->dclk == rps->dclk)); - *equal &= ((cps->evclk == rps->evclk) && (cps->ecclk == rps->ecclk)); - - return 0; -} - -static u32 ci_dpm_get_sclk(void *handle, bool low) -{ - struct amdgpu_device *adev = (struct amdgpu_device *)handle; - struct ci_power_info *pi = ci_get_pi(adev); - struct ci_ps *requested_state = ci_get_ps(&pi->requested_rps); - - if (low) - return requested_state->performance_levels[0].sclk; - else - return requested_state->performance_levels[requested_state->performance_level_count - 1].sclk; -} - -static u32 ci_dpm_get_mclk(void *handle, bool low) -{ - struct amdgpu_device *adev = (struct amdgpu_device *)handle; - struct ci_power_info *pi = ci_get_pi(adev); - struct ci_ps *requested_state = ci_get_ps(&pi->requested_rps); - - if (low) - return requested_state->performance_levels[0].mclk; - else - return requested_state->performance_levels[requested_state->performance_level_count - 1].mclk; -} - -/* get temperature in millidegrees */ -static int ci_dpm_get_temp(void *handle) -{ - u32 temp; - int actual_temp = 0; - struct amdgpu_device *adev = (struct amdgpu_device *)handle; - - temp = (RREG32_SMC(ixCG_MULT_THERMAL_STATUS) & CG_MULT_THERMAL_STATUS__CTF_TEMP_MASK) >> - CG_MULT_THERMAL_STATUS__CTF_TEMP__SHIFT; - - if (temp & 0x200) - actual_temp = 255; - else - actual_temp = temp & 0x1ff; - - actual_temp = actual_temp * 1000; - - return actual_temp; -} - -static int ci_set_temperature_range(struct amdgpu_device *adev) -{ - int ret; - - ret = ci_thermal_enable_alert(adev, false); - if (ret) - return ret; - ret = ci_thermal_set_temperature_range(adev, CISLANDS_TEMP_RANGE_MIN, - CISLANDS_TEMP_RANGE_MAX); - if (ret) - return ret; - ret = ci_thermal_enable_alert(adev, true); - if (ret) - return ret; - return ret; -} - -static int ci_dpm_early_init(void *handle) -{ - struct amdgpu_device *adev = (struct amdgpu_device *)handle; - - adev->powerplay.pp_funcs = &ci_dpm_funcs; - adev->powerplay.pp_handle = adev; - ci_dpm_set_irq_funcs(adev); - - return 0; -} - -static int ci_dpm_late_init(void *handle) -{ - int ret; - struct amdgpu_device *adev = (struct amdgpu_device *)handle; - - if (!adev->pm.dpm_enabled) - return 0; - - /* init the sysfs and debugfs files late */ - ret = amdgpu_pm_sysfs_init(adev); - if (ret) - return ret; - - ret = ci_set_temperature_range(adev); - if (ret) - return ret; - - return 0; -} - -static int ci_dpm_sw_init(void *handle) -{ - int ret; - struct amdgpu_device *adev = (struct amdgpu_device *)handle; - - ret = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, 230, - &adev->pm.dpm.thermal.irq); - if (ret) - return ret; - - ret = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, 231, - &adev->pm.dpm.thermal.irq); - if (ret) - return ret; - - /* default to balanced state */ - adev->pm.dpm.state = POWER_STATE_TYPE_BALANCED; - adev->pm.dpm.user_state = POWER_STATE_TYPE_BALANCED; - adev->pm.dpm.forced_level = AMD_DPM_FORCED_LEVEL_AUTO; - adev->pm.default_sclk = adev->clock.default_sclk; - adev->pm.default_mclk = adev->clock.default_mclk; - adev->pm.current_sclk = adev->clock.default_sclk; - adev->pm.current_mclk = adev->clock.default_mclk; - adev->pm.int_thermal_type = THERMAL_TYPE_NONE; - - ret = ci_dpm_init_microcode(adev); - if (ret) - return ret; - - if (amdgpu_dpm == 0) - return 0; - - INIT_WORK(&adev->pm.dpm.thermal.work, amdgpu_dpm_thermal_work_handler); - mutex_lock(&adev->pm.mutex); - ret = ci_dpm_init(adev); - if (ret) - goto dpm_failed; - adev->pm.dpm.current_ps = adev->pm.dpm.requested_ps = adev->pm.dpm.boot_ps; - if (amdgpu_dpm == 1) - amdgpu_pm_print_power_states(adev); - mutex_unlock(&adev->pm.mutex); - DRM_INFO("amdgpu: dpm initialized\n"); - - return 0; - -dpm_failed: - ci_dpm_fini(adev); - mutex_unlock(&adev->pm.mutex); - DRM_ERROR("amdgpu: dpm initialization failed\n"); - return ret; -} - -static int ci_dpm_sw_fini(void *handle) -{ - struct amdgpu_device *adev = (struct amdgpu_device *)handle; - - flush_work(&adev->pm.dpm.thermal.work); - - mutex_lock(&adev->pm.mutex); - ci_dpm_fini(adev); - mutex_unlock(&adev->pm.mutex); - - release_firmware(adev->pm.fw); - adev->pm.fw = NULL; - - return 0; -} - -static int ci_dpm_hw_init(void *handle) -{ - int ret; - - struct amdgpu_device *adev = (struct amdgpu_device *)handle; - - if (!amdgpu_dpm) { - ret = ci_upload_firmware(adev); - if (ret) { - DRM_ERROR("ci_upload_firmware failed\n"); - return ret; - } - ci_dpm_start_smc(adev); - return 0; - } - - mutex_lock(&adev->pm.mutex); - ci_dpm_setup_asic(adev); - ret = ci_dpm_enable(adev); - if (ret) - adev->pm.dpm_enabled = false; - else - adev->pm.dpm_enabled = true; - mutex_unlock(&adev->pm.mutex); - - return ret; -} - -static int ci_dpm_hw_fini(void *handle) -{ - struct amdgpu_device *adev = (struct amdgpu_device *)handle; - - if (adev->pm.dpm_enabled) { - mutex_lock(&adev->pm.mutex); - ci_dpm_disable(adev); - mutex_unlock(&adev->pm.mutex); - } else { - ci_dpm_stop_smc(adev); - } - - return 0; -} - -static int ci_dpm_suspend(void *handle) -{ - struct amdgpu_device *adev = (struct amdgpu_device *)handle; - - if (adev->pm.dpm_enabled) { - mutex_lock(&adev->pm.mutex); - amdgpu_irq_put(adev, &adev->pm.dpm.thermal.irq, - AMDGPU_THERMAL_IRQ_LOW_TO_HIGH); - amdgpu_irq_put(adev, &adev->pm.dpm.thermal.irq, - AMDGPU_THERMAL_IRQ_HIGH_TO_LOW); - adev->pm.dpm.last_user_state = adev->pm.dpm.user_state; - adev->pm.dpm.last_state = adev->pm.dpm.state; - adev->pm.dpm.user_state = POWER_STATE_TYPE_INTERNAL_BOOT; - adev->pm.dpm.state = POWER_STATE_TYPE_INTERNAL_BOOT; - mutex_unlock(&adev->pm.mutex); - amdgpu_pm_compute_clocks(adev); - - } - - return 0; -} - -static int ci_dpm_resume(void *handle) -{ - int ret; - struct amdgpu_device *adev = (struct amdgpu_device *)handle; - - if (adev->pm.dpm_enabled) { - /* asic init will reset to the boot state */ - mutex_lock(&adev->pm.mutex); - ci_dpm_setup_asic(adev); - ret = ci_dpm_enable(adev); - if (ret) - adev->pm.dpm_enabled = false; - else - adev->pm.dpm_enabled = true; - adev->pm.dpm.user_state = adev->pm.dpm.last_user_state; - adev->pm.dpm.state = adev->pm.dpm.last_state; - mutex_unlock(&adev->pm.mutex); - if (adev->pm.dpm_enabled) - amdgpu_pm_compute_clocks(adev); - } - return 0; -} - -static bool ci_dpm_is_idle(void *handle) -{ - /* XXX */ - return true; -} - -static int ci_dpm_wait_for_idle(void *handle) -{ - /* XXX */ - return 0; -} - -static int ci_dpm_soft_reset(void *handle) -{ - return 0; -} - -static int ci_dpm_set_interrupt_state(struct amdgpu_device *adev, - struct amdgpu_irq_src *source, - unsigned type, - enum amdgpu_interrupt_state state) -{ - u32 cg_thermal_int; - - switch (type) { - case AMDGPU_THERMAL_IRQ_LOW_TO_HIGH: - switch (state) { - case AMDGPU_IRQ_STATE_DISABLE: - cg_thermal_int = RREG32_SMC(ixCG_THERMAL_INT); - cg_thermal_int |= CG_THERMAL_INT_CTRL__THERM_INTH_MASK_MASK; - WREG32_SMC(ixCG_THERMAL_INT, cg_thermal_int); - break; - case AMDGPU_IRQ_STATE_ENABLE: - cg_thermal_int = RREG32_SMC(ixCG_THERMAL_INT); - cg_thermal_int &= ~CG_THERMAL_INT_CTRL__THERM_INTH_MASK_MASK; - WREG32_SMC(ixCG_THERMAL_INT, cg_thermal_int); - break; - default: - break; - } - break; - - case AMDGPU_THERMAL_IRQ_HIGH_TO_LOW: - switch (state) { - case AMDGPU_IRQ_STATE_DISABLE: - cg_thermal_int = RREG32_SMC(ixCG_THERMAL_INT); - cg_thermal_int |= CG_THERMAL_INT_CTRL__THERM_INTL_MASK_MASK; - WREG32_SMC(ixCG_THERMAL_INT, cg_thermal_int); - break; - case AMDGPU_IRQ_STATE_ENABLE: - cg_thermal_int = RREG32_SMC(ixCG_THERMAL_INT); - cg_thermal_int &= ~CG_THERMAL_INT_CTRL__THERM_INTL_MASK_MASK; - WREG32_SMC(ixCG_THERMAL_INT, cg_thermal_int); - break; - default: - break; - } - break; - - default: - break; - } - return 0; -} - -static int ci_dpm_process_interrupt(struct amdgpu_device *adev, - struct amdgpu_irq_src *source, - struct amdgpu_iv_entry *entry) -{ - bool queue_thermal = false; - - if (entry == NULL) - return -EINVAL; - - switch (entry->src_id) { - case 230: /* thermal low to high */ - DRM_DEBUG("IH: thermal low to high\n"); - adev->pm.dpm.thermal.high_to_low = false; - queue_thermal = true; - break; - case 231: /* thermal high to low */ - DRM_DEBUG("IH: thermal high to low\n"); - adev->pm.dpm.thermal.high_to_low = true; - queue_thermal = true; - break; - default: - break; - } - - if (queue_thermal) - schedule_work(&adev->pm.dpm.thermal.work); - - return 0; -} - -static int ci_dpm_set_clockgating_state(void *handle, - enum amd_clockgating_state state) -{ - return 0; -} - -static int ci_dpm_set_powergating_state(void *handle, - enum amd_powergating_state state) -{ - return 0; -} - -static int ci_dpm_print_clock_levels(void *handle, - enum pp_clock_type type, char *buf) -{ - struct amdgpu_device *adev = (struct amdgpu_device *)handle; - struct ci_power_info *pi = ci_get_pi(adev); - struct ci_single_dpm_table *sclk_table = &pi->dpm_table.sclk_table; - struct ci_single_dpm_table *mclk_table = &pi->dpm_table.mclk_table; - struct ci_single_dpm_table *pcie_table = &pi->dpm_table.pcie_speed_table; - - int i, now, size = 0; - uint32_t clock, pcie_speed; - - switch (type) { - case PP_SCLK: - amdgpu_ci_send_msg_to_smc(adev, PPSMC_MSG_API_GetSclkFrequency); - clock = RREG32(mmSMC_MSG_ARG_0); - - for (i = 0; i < sclk_table->count; i++) { - if (clock > sclk_table->dpm_levels[i].value) - continue; - break; - } - now = i; - - for (i = 0; i < sclk_table->count; i++) - size += sprintf(buf + size, "%d: %uMhz %s\n", - i, sclk_table->dpm_levels[i].value / 100, - (i == now) ? "*" : ""); - break; - case PP_MCLK: - amdgpu_ci_send_msg_to_smc(adev, PPSMC_MSG_API_GetMclkFrequency); - clock = RREG32(mmSMC_MSG_ARG_0); - - for (i = 0; i < mclk_table->count; i++) { - if (clock > mclk_table->dpm_levels[i].value) - continue; - break; - } - now = i; - - for (i = 0; i < mclk_table->count; i++) - size += sprintf(buf + size, "%d: %uMhz %s\n", - i, mclk_table->dpm_levels[i].value / 100, - (i == now) ? "*" : ""); - break; - case PP_PCIE: - pcie_speed = ci_get_current_pcie_speed(adev); - for (i = 0; i < pcie_table->count; i++) { - if (pcie_speed != pcie_table->dpm_levels[i].value) - continue; - break; - } - now = i; - - for (i = 0; i < pcie_table->count; i++) - size += sprintf(buf + size, "%d: %s %s\n", i, - (pcie_table->dpm_levels[i].value == 0) ? "2.5GT/s, x1" : - (pcie_table->dpm_levels[i].value == 1) ? "5.0GT/s, x16" : - (pcie_table->dpm_levels[i].value == 2) ? "8.0GT/s, x16" : "", - (i == now) ? "*" : ""); - break; - default: - break; - } - - return size; -} - -static int ci_dpm_force_clock_level(void *handle, - enum pp_clock_type type, uint32_t mask) -{ - struct amdgpu_device *adev = (struct amdgpu_device *)handle; - struct ci_power_info *pi = ci_get_pi(adev); - - if (adev->pm.dpm.forced_level != AMD_DPM_FORCED_LEVEL_MANUAL) - return -EINVAL; - - if (mask == 0) - return -EINVAL; - - switch (type) { - case PP_SCLK: - if (!pi->sclk_dpm_key_disabled) - amdgpu_ci_send_msg_to_smc_with_parameter(adev, - PPSMC_MSG_SCLKDPM_SetEnabledMask, - pi->dpm_level_enable_mask.sclk_dpm_enable_mask & mask); - break; - - case PP_MCLK: - if (!pi->mclk_dpm_key_disabled) - amdgpu_ci_send_msg_to_smc_with_parameter(adev, - PPSMC_MSG_MCLKDPM_SetEnabledMask, - pi->dpm_level_enable_mask.mclk_dpm_enable_mask & mask); - break; - - case PP_PCIE: - { - uint32_t tmp = mask & pi->dpm_level_enable_mask.pcie_dpm_enable_mask; - - if (!pi->pcie_dpm_key_disabled) { - if (fls(tmp) != ffs(tmp)) - amdgpu_ci_send_msg_to_smc(adev, PPSMC_MSG_PCIeDPM_UnForceLevel); - else - amdgpu_ci_send_msg_to_smc_with_parameter(adev, - PPSMC_MSG_PCIeDPM_ForceLevel, - fls(tmp) - 1); - } - break; - } - default: - break; - } - - return 0; -} - -static int ci_dpm_get_sclk_od(void *handle) -{ - struct amdgpu_device *adev = (struct amdgpu_device *)handle; - struct ci_power_info *pi = ci_get_pi(adev); - struct ci_single_dpm_table *sclk_table = &(pi->dpm_table.sclk_table); - struct ci_single_dpm_table *golden_sclk_table = - &(pi->golden_dpm_table.sclk_table); - int value; - - value = (sclk_table->dpm_levels[sclk_table->count - 1].value - - golden_sclk_table->dpm_levels[golden_sclk_table->count - 1].value) * - 100 / - golden_sclk_table->dpm_levels[golden_sclk_table->count - 1].value; - - return value; -} - -static int ci_dpm_set_sclk_od(void *handle, uint32_t value) -{ - struct amdgpu_device *adev = (struct amdgpu_device *)handle; - struct ci_power_info *pi = ci_get_pi(adev); - struct ci_ps *ps = ci_get_ps(adev->pm.dpm.requested_ps); - struct ci_single_dpm_table *golden_sclk_table = - &(pi->golden_dpm_table.sclk_table); - - if (value > 20) - value = 20; - - ps->performance_levels[ps->performance_level_count - 1].sclk = - golden_sclk_table->dpm_levels[golden_sclk_table->count - 1].value * - value / 100 + - golden_sclk_table->dpm_levels[golden_sclk_table->count - 1].value; - - return 0; -} - -static int ci_dpm_get_mclk_od(void *handle) -{ - struct amdgpu_device *adev = (struct amdgpu_device *)handle; - struct ci_power_info *pi = ci_get_pi(adev); - struct ci_single_dpm_table *mclk_table = &(pi->dpm_table.mclk_table); - struct ci_single_dpm_table *golden_mclk_table = - &(pi->golden_dpm_table.mclk_table); - int value; - - value = (mclk_table->dpm_levels[mclk_table->count - 1].value - - golden_mclk_table->dpm_levels[golden_mclk_table->count - 1].value) * - 100 / - golden_mclk_table->dpm_levels[golden_mclk_table->count - 1].value; - - return value; -} - -static int ci_dpm_set_mclk_od(void *handle, uint32_t value) -{ - struct amdgpu_device *adev = (struct amdgpu_device *)handle; - struct ci_power_info *pi = ci_get_pi(adev); - struct ci_ps *ps = ci_get_ps(adev->pm.dpm.requested_ps); - struct ci_single_dpm_table *golden_mclk_table = - &(pi->golden_dpm_table.mclk_table); - - if (value > 20) - value = 20; - - ps->performance_levels[ps->performance_level_count - 1].mclk = - golden_mclk_table->dpm_levels[golden_mclk_table->count - 1].value * - value / 100 + - golden_mclk_table->dpm_levels[golden_mclk_table->count - 1].value; - - return 0; -} - -static int ci_dpm_read_sensor(void *handle, int idx, - void *value, int *size) -{ - u32 activity_percent = 50; - int ret; - struct amdgpu_device *adev = (struct amdgpu_device *)handle; - - /* size must be at least 4 bytes for all sensors */ - if (*size < 4) - return -EINVAL; - - switch (idx) { - case AMDGPU_PP_SENSOR_GFX_SCLK: - *((uint32_t *)value) = ci_get_average_sclk_freq(adev); - *size = 4; - return 0; - case AMDGPU_PP_SENSOR_GFX_MCLK: - *((uint32_t *)value) = ci_get_average_mclk_freq(adev); - *size = 4; - return 0; - case AMDGPU_PP_SENSOR_GPU_TEMP: - *((uint32_t *)value) = ci_dpm_get_temp(adev); - *size = 4; - return 0; - case AMDGPU_PP_SENSOR_GPU_LOAD: - ret = ci_read_smc_soft_register(adev, - offsetof(SMU7_SoftRegisters, - AverageGraphicsA), - &activity_percent); - if (ret == 0) { - activity_percent += 0x80; - activity_percent >>= 8; - activity_percent = - activity_percent > 100 ? 100 : activity_percent; - } - *((uint32_t *)value) = activity_percent; - *size = 4; - return 0; - default: - return -EINVAL; - } -} - -static int ci_set_powergating_by_smu(void *handle, - uint32_t block_type, bool gate) -{ - switch (block_type) { - case AMD_IP_BLOCK_TYPE_UVD: - ci_dpm_powergate_uvd(handle, gate); - break; - default: - break; - } - return 0; -} - -static const struct amd_ip_funcs ci_dpm_ip_funcs = { - .name = "ci_dpm", - .early_init = ci_dpm_early_init, - .late_init = ci_dpm_late_init, - .sw_init = ci_dpm_sw_init, - .sw_fini = ci_dpm_sw_fini, - .hw_init = ci_dpm_hw_init, - .hw_fini = ci_dpm_hw_fini, - .suspend = ci_dpm_suspend, - .resume = ci_dpm_resume, - .is_idle = ci_dpm_is_idle, - .wait_for_idle = ci_dpm_wait_for_idle, - .soft_reset = ci_dpm_soft_reset, - .set_clockgating_state = ci_dpm_set_clockgating_state, - .set_powergating_state = ci_dpm_set_powergating_state, -}; - -const struct amdgpu_ip_block_version ci_smu_ip_block = -{ - .type = AMD_IP_BLOCK_TYPE_SMC, - .major = 7, - .minor = 0, - .rev = 0, - .funcs = &ci_dpm_ip_funcs, -}; - -static const struct amd_pm_funcs ci_dpm_funcs = { - .pre_set_power_state = &ci_dpm_pre_set_power_state, - .set_power_state = &ci_dpm_set_power_state, - .post_set_power_state = &ci_dpm_post_set_power_state, - .display_configuration_changed = &ci_dpm_display_configuration_changed, - .get_sclk = &ci_dpm_get_sclk, - .get_mclk = &ci_dpm_get_mclk, - .print_power_state = &ci_dpm_print_power_state, - .debugfs_print_current_performance_level = &ci_dpm_debugfs_print_current_performance_level, - .force_performance_level = &ci_dpm_force_performance_level, - .vblank_too_short = &ci_dpm_vblank_too_short, - .set_powergating_by_smu = &ci_set_powergating_by_smu, - .set_fan_control_mode = &ci_dpm_set_fan_control_mode, - .get_fan_control_mode = &ci_dpm_get_fan_control_mode, - .set_fan_speed_percent = &ci_dpm_set_fan_speed_percent, - .get_fan_speed_percent = &ci_dpm_get_fan_speed_percent, - .print_clock_levels = ci_dpm_print_clock_levels, - .force_clock_level = ci_dpm_force_clock_level, - .get_sclk_od = ci_dpm_get_sclk_od, - .set_sclk_od = ci_dpm_set_sclk_od, - .get_mclk_od = ci_dpm_get_mclk_od, - .set_mclk_od = ci_dpm_set_mclk_od, - .check_state_equal = ci_check_state_equal, - .get_vce_clock_state = amdgpu_get_vce_clock_state, - .read_sensor = ci_dpm_read_sensor, -}; - -static const struct amdgpu_irq_src_funcs ci_dpm_irq_funcs = { - .set = ci_dpm_set_interrupt_state, - .process = ci_dpm_process_interrupt, -}; - -static void ci_dpm_set_irq_funcs(struct amdgpu_device *adev) -{ - adev->pm.dpm.thermal.irq.num_types = AMDGPU_THERMAL_IRQ_LAST; - adev->pm.dpm.thermal.irq.funcs = &ci_dpm_irq_funcs; -} diff --git a/drivers/gpu/drm/amd/amdgpu/ci_dpm.h b/drivers/gpu/drm/amd/amdgpu/ci_dpm.h deleted file mode 100644 index 91be2996ae7c..000000000000 --- a/drivers/gpu/drm/amd/amdgpu/ci_dpm.h +++ /dev/null @@ -1,349 +0,0 @@ -/* - * Copyright 2013 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 __CI_DPM_H__ -#define __CI_DPM_H__ - -#include "amdgpu_atombios.h" -#include "ppsmc.h" - -#define SMU__NUM_SCLK_DPM_STATE 8 -#define SMU__NUM_MCLK_DPM_LEVELS 6 -#define SMU__NUM_LCLK_DPM_LEVELS 8 -#define SMU__NUM_PCIE_DPM_LEVELS 8 -#include "smu7_discrete.h" - -#define CISLANDS_MAX_HARDWARE_POWERLEVELS 2 - -#define CISLANDS_UNUSED_GPIO_PIN 0x7F - -struct ci_pl { - u32 mclk; - u32 sclk; - enum amdgpu_pcie_gen pcie_gen; - u16 pcie_lane; -}; - -struct ci_ps { - u16 performance_level_count; - bool dc_compatible; - u32 sclk_t; - struct ci_pl performance_levels[CISLANDS_MAX_HARDWARE_POWERLEVELS]; -}; - -struct ci_dpm_level { - bool enabled; - u32 value; - u32 param1; -}; - -#define CISLAND_MAX_DEEPSLEEP_DIVIDER_ID 5 -#define MAX_REGULAR_DPM_NUMBER 8 -#define CISLAND_MINIMUM_ENGINE_CLOCK 800 - -struct ci_single_dpm_table { - u32 count; - struct ci_dpm_level dpm_levels[MAX_REGULAR_DPM_NUMBER]; -}; - -struct ci_dpm_table { - struct ci_single_dpm_table sclk_table; - struct ci_single_dpm_table mclk_table; - struct ci_single_dpm_table pcie_speed_table; - struct ci_single_dpm_table vddc_table; - struct ci_single_dpm_table vddci_table; - struct ci_single_dpm_table mvdd_table; -}; - -struct ci_mc_reg_entry { - u32 mclk_max; - u32 mc_data[SMU7_DISCRETE_MC_REGISTER_ARRAY_SIZE]; -}; - -struct ci_mc_reg_table { - u8 last; - u8 num_entries; - u16 valid_flag; - struct ci_mc_reg_entry mc_reg_table_entry[MAX_AC_TIMING_ENTRIES]; - SMU7_Discrete_MCRegisterAddress mc_reg_address[SMU7_DISCRETE_MC_REGISTER_ARRAY_SIZE]; -}; - -struct ci_ulv_parm -{ - bool supported; - u32 cg_ulv_parameter; - u32 volt_change_delay; - struct ci_pl pl; -}; - -#define CISLANDS_MAX_LEAKAGE_COUNT 8 - -struct ci_leakage_voltage { - u16 count; - u16 leakage_id[CISLANDS_MAX_LEAKAGE_COUNT]; - u16 actual_voltage[CISLANDS_MAX_LEAKAGE_COUNT]; -}; - -struct ci_dpm_level_enable_mask { - u32 uvd_dpm_enable_mask; - u32 vce_dpm_enable_mask; - u32 acp_dpm_enable_mask; - u32 samu_dpm_enable_mask; - u32 sclk_dpm_enable_mask; - u32 mclk_dpm_enable_mask; - u32 pcie_dpm_enable_mask; -}; - -struct ci_vbios_boot_state -{ - u16 mvdd_bootup_value; - u16 vddc_bootup_value; - u16 vddci_bootup_value; - u32 sclk_bootup_value; - u32 mclk_bootup_value; - u16 pcie_gen_bootup_value; - u16 pcie_lane_bootup_value; -}; - -struct ci_clock_registers { - u32 cg_spll_func_cntl; - u32 cg_spll_func_cntl_2; - u32 cg_spll_func_cntl_3; - u32 cg_spll_func_cntl_4; - u32 cg_spll_spread_spectrum; - u32 cg_spll_spread_spectrum_2; - u32 dll_cntl; - u32 mclk_pwrmgt_cntl; - u32 mpll_ad_func_cntl; - u32 mpll_dq_func_cntl; - u32 mpll_func_cntl; - u32 mpll_func_cntl_1; - u32 mpll_func_cntl_2; - u32 mpll_ss1; - u32 mpll_ss2; -}; - -struct ci_thermal_temperature_setting { - s32 temperature_low; - s32 temperature_high; - s32 temperature_shutdown; -}; - -struct ci_pcie_perf_range { - u16 max; - u16 min; -}; - -enum ci_pt_config_reg_type { - CISLANDS_CONFIGREG_MMR = 0, - CISLANDS_CONFIGREG_SMC_IND, - CISLANDS_CONFIGREG_DIDT_IND, - CISLANDS_CONFIGREG_CACHE, - CISLANDS_CONFIGREG_MAX -}; - -#define POWERCONTAINMENT_FEATURE_BAPM 0x00000001 -#define POWERCONTAINMENT_FEATURE_TDCLimit 0x00000002 -#define POWERCONTAINMENT_FEATURE_PkgPwrLimit 0x00000004 - -struct ci_pt_config_reg { - u32 offset; - u32 mask; - u32 shift; - u32 value; - enum ci_pt_config_reg_type type; -}; - -struct ci_pt_defaults { - u8 svi_load_line_en; - u8 svi_load_line_vddc; - u8 tdc_vddc_throttle_release_limit_perc; - u8 tdc_mawt; - u8 tdc_waterfall_ctl; - u8 dte_ambient_temp_base; - u32 display_cac; - u32 bapm_temp_gradient; - u16 bapmti_r[SMU7_DTE_ITERATIONS * SMU7_DTE_SOURCES * SMU7_DTE_SINKS]; - u16 bapmti_rc[SMU7_DTE_ITERATIONS * SMU7_DTE_SOURCES * SMU7_DTE_SINKS]; -}; - -#define DPMTABLE_OD_UPDATE_SCLK 0x00000001 -#define DPMTABLE_OD_UPDATE_MCLK 0x00000002 -#define DPMTABLE_UPDATE_SCLK 0x00000004 -#define DPMTABLE_UPDATE_MCLK 0x00000008 - -struct ci_power_info { - struct ci_dpm_table dpm_table; - struct ci_dpm_table golden_dpm_table; - u32 voltage_control; - u32 mvdd_control; - u32 vddci_control; - u32 active_auto_throttle_sources; - struct ci_clock_registers clock_registers; - u16 acpi_vddc; - u16 acpi_vddci; - enum amdgpu_pcie_gen force_pcie_gen; - enum amdgpu_pcie_gen acpi_pcie_gen; - struct ci_leakage_voltage vddc_leakage; - struct ci_leakage_voltage vddci_leakage; - u16 max_vddc_in_pp_table; - u16 min_vddc_in_pp_table; - u16 max_vddci_in_pp_table; - u16 min_vddci_in_pp_table; - u32 mclk_strobe_mode_threshold; - u32 mclk_stutter_mode_threshold; - u32 mclk_edc_enable_threshold; - u32 mclk_edc_wr_enable_threshold; - struct ci_vbios_boot_state vbios_boot_state; - /* smc offsets */ - u32 sram_end; - u32 dpm_table_start; - u32 soft_regs_start; - u32 mc_reg_table_start; - u32 fan_table_start; - u32 arb_table_start; - /* smc tables */ - SMU7_Discrete_DpmTable smc_state_table; - SMU7_Discrete_MCRegisters smc_mc_reg_table; - SMU7_Discrete_PmFuses smc_powertune_table; - /* other stuff */ - struct ci_mc_reg_table mc_reg_table; - struct atom_voltage_table vddc_voltage_table; - struct atom_voltage_table vddci_voltage_table; - struct atom_voltage_table mvdd_voltage_table; - struct ci_ulv_parm ulv; - u32 power_containment_features; - const struct ci_pt_defaults *powertune_defaults; - u32 dte_tj_offset; - bool vddc_phase_shed_control; - struct ci_thermal_temperature_setting thermal_temp_setting; - struct ci_dpm_level_enable_mask dpm_level_enable_mask; - u32 need_update_smu7_dpm_table; - u32 sclk_dpm_key_disabled; - u32 mclk_dpm_key_disabled; - u32 pcie_dpm_key_disabled; - u32 thermal_sclk_dpm_enabled; - struct ci_pcie_perf_range pcie_gen_performance; - struct ci_pcie_perf_range pcie_lane_performance; - struct ci_pcie_perf_range pcie_gen_powersaving; - struct ci_pcie_perf_range pcie_lane_powersaving; - u32 activity_target[SMU7_MAX_LEVELS_GRAPHICS]; - u32 mclk_activity_target; - u32 low_sclk_interrupt_t; - u32 last_mclk_dpm_enable_mask; - u32 sys_pcie_mask; - /* caps */ - bool caps_power_containment; - bool caps_cac; - bool caps_sq_ramping; - bool caps_db_ramping; - bool caps_td_ramping; - bool caps_tcp_ramping; - bool caps_fps; - bool caps_sclk_ds; - bool caps_sclk_ss_support; - bool caps_mclk_ss_support; - bool caps_uvd_dpm; - bool caps_vce_dpm; - bool caps_samu_dpm; - bool caps_acp_dpm; - bool caps_automatic_dc_transition; - bool caps_sclk_throttle_low_notification; - bool caps_dynamic_ac_timing; - bool caps_od_fuzzy_fan_control_support; - /* flags */ - bool thermal_protection; - bool pcie_performance_request; - bool dynamic_ss; - bool dll_default_on; - bool cac_enabled; - bool uvd_enabled; - bool battery_state; - bool pspp_notify_required; - bool enable_bapm_feature; - bool enable_tdc_limit_feature; - bool enable_pkg_pwr_tracking_feature; - bool use_pcie_performance_levels; - bool use_pcie_powersaving_levels; - bool uvd_power_gated; - /* driver states */ - struct amdgpu_ps current_rps; - struct ci_ps current_ps; - struct amdgpu_ps requested_rps; - struct ci_ps requested_ps; - /* fan control */ - bool fan_ctrl_is_in_default_mode; - bool fan_is_controlled_by_smc; - u32 t_min; - u32 fan_ctrl_default_mode; -}; - -#define CISLANDS_VOLTAGE_CONTROL_NONE 0x0 -#define CISLANDS_VOLTAGE_CONTROL_BY_GPIO 0x1 -#define CISLANDS_VOLTAGE_CONTROL_BY_SVID2 0x2 - -#define CISLANDS_Q88_FORMAT_CONVERSION_UNIT 256 - -#define CISLANDS_VRC_DFLT0 0x3FFFC000 -#define CISLANDS_VRC_DFLT1 0x000400 -#define CISLANDS_VRC_DFLT2 0xC00080 -#define CISLANDS_VRC_DFLT3 0xC00200 -#define CISLANDS_VRC_DFLT4 0xC01680 -#define CISLANDS_VRC_DFLT5 0xC00033 -#define CISLANDS_VRC_DFLT6 0xC00033 -#define CISLANDS_VRC_DFLT7 0x3FFFC000 - -#define CISLANDS_CGULVPARAMETER_DFLT 0x00040035 -#define CISLAND_TARGETACTIVITY_DFLT 30 -#define CISLAND_MCLK_TARGETACTIVITY_DFLT 10 - -#define PCIE_PERF_REQ_REMOVE_REGISTRY 0 -#define PCIE_PERF_REQ_FORCE_LOWPOWER 1 -#define PCIE_PERF_REQ_PECI_GEN1 2 -#define PCIE_PERF_REQ_PECI_GEN2 3 -#define PCIE_PERF_REQ_PECI_GEN3 4 - -#define CISLANDS_SSTU_DFLT 0 -#define CISLANDS_SST_DFLT 0x00C8 - -/* XXX are these ok? */ -#define CISLANDS_TEMP_RANGE_MIN (90 * 1000) -#define CISLANDS_TEMP_RANGE_MAX (120 * 1000) - -int amdgpu_ci_copy_bytes_to_smc(struct amdgpu_device *adev, - u32 smc_start_address, - const u8 *src, u32 byte_count, u32 limit); -void amdgpu_ci_start_smc(struct amdgpu_device *adev); -void amdgpu_ci_reset_smc(struct amdgpu_device *adev); -int amdgpu_ci_program_jump_on_start(struct amdgpu_device *adev); -void amdgpu_ci_stop_smc_clock(struct amdgpu_device *adev); -void amdgpu_ci_start_smc_clock(struct amdgpu_device *adev); -bool amdgpu_ci_is_smc_running(struct amdgpu_device *adev); -PPSMC_Result amdgpu_ci_send_msg_to_smc(struct amdgpu_device *adev, PPSMC_Msg msg); -PPSMC_Result amdgpu_ci_wait_for_smc_inactive(struct amdgpu_device *adev); -int amdgpu_ci_load_smc_ucode(struct amdgpu_device *adev, u32 limit); -int amdgpu_ci_read_smc_sram_dword(struct amdgpu_device *adev, - u32 smc_address, u32 *value, u32 limit); -int amdgpu_ci_write_smc_sram_dword(struct amdgpu_device *adev, - u32 smc_address, u32 value, u32 limit); - -#endif diff --git a/drivers/gpu/drm/amd/amdgpu/ci_smc.c b/drivers/gpu/drm/amd/amdgpu/ci_smc.c deleted file mode 100644 index b8ba51e045b5..000000000000 --- a/drivers/gpu/drm/amd/amdgpu/ci_smc.c +++ /dev/null @@ -1,279 +0,0 @@ -/* - * Copyright 2011 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. - * - * Authors: Alex Deucher - */ - -#include <linux/firmware.h> -#include <drm/drmP.h> -#include "amdgpu.h" -#include "cikd.h" -#include "ppsmc.h" -#include "amdgpu_ucode.h" -#include "ci_dpm.h" - -#include "smu/smu_7_0_1_d.h" -#include "smu/smu_7_0_1_sh_mask.h" - -static int ci_set_smc_sram_address(struct amdgpu_device *adev, - u32 smc_address, u32 limit) -{ - if (smc_address & 3) - return -EINVAL; - if ((smc_address + 3) > limit) - return -EINVAL; - - WREG32(mmSMC_IND_INDEX_0, smc_address); - WREG32_P(mmSMC_IND_ACCESS_CNTL, 0, ~SMC_IND_ACCESS_CNTL__AUTO_INCREMENT_IND_0_MASK); - - return 0; -} - -int amdgpu_ci_copy_bytes_to_smc(struct amdgpu_device *adev, - u32 smc_start_address, - const u8 *src, u32 byte_count, u32 limit) -{ - unsigned long flags; - u32 data, original_data; - u32 addr; - u32 extra_shift; - int ret = 0; - - if (smc_start_address & 3) - return -EINVAL; - if ((smc_start_address + byte_count) > limit) - return -EINVAL; - - addr = smc_start_address; - - spin_lock_irqsave(&adev->smc_idx_lock, flags); - while (byte_count >= 4) { - /* SMC address space is BE */ - data = (src[0] << 24) | (src[1] << 16) | (src[2] << 8) | src[3]; - - ret = ci_set_smc_sram_address(adev, addr, limit); - if (ret) - goto done; - - WREG32(mmSMC_IND_DATA_0, data); - - src += 4; - byte_count -= 4; - addr += 4; - } - - /* RMW for the final bytes */ - if (byte_count > 0) { - data = 0; - - ret = ci_set_smc_sram_address(adev, addr, limit); - if (ret) - goto done; - - original_data = RREG32(mmSMC_IND_DATA_0); - - extra_shift = 8 * (4 - byte_count); - - while (byte_count > 0) { - data = (data << 8) + *src++; - byte_count--; - } - - data <<= extra_shift; - - data |= (original_data & ~((~0UL) << extra_shift)); - - ret = ci_set_smc_sram_address(adev, addr, limit); - if (ret) - goto done; - - WREG32(mmSMC_IND_DATA_0, data); - } - -done: - spin_unlock_irqrestore(&adev->smc_idx_lock, flags); - - return ret; -} - -void amdgpu_ci_start_smc(struct amdgpu_device *adev) -{ - u32 tmp = RREG32_SMC(ixSMC_SYSCON_RESET_CNTL); - - tmp &= ~SMC_SYSCON_RESET_CNTL__rst_reg_MASK; - WREG32_SMC(ixSMC_SYSCON_RESET_CNTL, tmp); -} - -void amdgpu_ci_reset_smc(struct amdgpu_device *adev) -{ - u32 tmp = RREG32_SMC(ixSMC_SYSCON_RESET_CNTL); - - tmp |= SMC_SYSCON_RESET_CNTL__rst_reg_MASK; - WREG32_SMC(ixSMC_SYSCON_RESET_CNTL, tmp); -} - -int amdgpu_ci_program_jump_on_start(struct amdgpu_device *adev) -{ - static u8 data[] = { 0xE0, 0x00, 0x80, 0x40 }; - - return amdgpu_ci_copy_bytes_to_smc(adev, 0x0, data, 4, sizeof(data)+1); -} - -void amdgpu_ci_stop_smc_clock(struct amdgpu_device *adev) -{ - u32 tmp = RREG32_SMC(ixSMC_SYSCON_CLOCK_CNTL_0); - - tmp |= SMC_SYSCON_CLOCK_CNTL_0__ck_disable_MASK; - - WREG32_SMC(ixSMC_SYSCON_CLOCK_CNTL_0, tmp); -} - -void amdgpu_ci_start_smc_clock(struct amdgpu_device *adev) -{ - u32 tmp = RREG32_SMC(ixSMC_SYSCON_CLOCK_CNTL_0); - - tmp &= ~SMC_SYSCON_CLOCK_CNTL_0__ck_disable_MASK; - - WREG32_SMC(ixSMC_SYSCON_CLOCK_CNTL_0, tmp); -} - -bool amdgpu_ci_is_smc_running(struct amdgpu_device *adev) -{ - u32 clk = RREG32_SMC(ixSMC_SYSCON_CLOCK_CNTL_0); - u32 pc_c = RREG32_SMC(ixSMC_PC_C); - - if (!(clk & SMC_SYSCON_CLOCK_CNTL_0__ck_disable_MASK) && (0x20100 <= pc_c)) - return true; - - return false; -} - -PPSMC_Result amdgpu_ci_send_msg_to_smc(struct amdgpu_device *adev, PPSMC_Msg msg) -{ - u32 tmp; - int i; - - if (!amdgpu_ci_is_smc_running(adev)) - return PPSMC_Result_Failed; - - WREG32(mmSMC_MESSAGE_0, msg); - - for (i = 0; i < adev->usec_timeout; i++) { - tmp = RREG32(mmSMC_RESP_0); - if (tmp != 0) - break; - udelay(1); - } - tmp = RREG32(mmSMC_RESP_0); - - return (PPSMC_Result)tmp; -} - -PPSMC_Result amdgpu_ci_wait_for_smc_inactive(struct amdgpu_device *adev) -{ - u32 tmp; - int i; - - if (!amdgpu_ci_is_smc_running(adev)) - return PPSMC_Result_OK; - - for (i = 0; i < adev->usec_timeout; i++) { - tmp = RREG32_SMC(ixSMC_SYSCON_CLOCK_CNTL_0); - if ((tmp & SMC_SYSCON_CLOCK_CNTL_0__cken_MASK) == 0) - break; - udelay(1); - } - - return PPSMC_Result_OK; -} - -int amdgpu_ci_load_smc_ucode(struct amdgpu_device *adev, u32 limit) -{ - const struct smc_firmware_header_v1_0 *hdr; - unsigned long flags; - u32 ucode_start_address; - u32 ucode_size; - const u8 *src; - u32 data; - - if (!adev->pm.fw) - return -EINVAL; - - hdr = (const struct smc_firmware_header_v1_0 *)adev->pm.fw->data; - amdgpu_ucode_print_smc_hdr(&hdr->header); - - adev->pm.fw_version = le32_to_cpu(hdr->header.ucode_version); - ucode_start_address = le32_to_cpu(hdr->ucode_start_addr); - ucode_size = le32_to_cpu(hdr->header.ucode_size_bytes); - src = (const u8 *) - (adev->pm.fw->data + le32_to_cpu(hdr->header.ucode_array_offset_bytes)); - - if (ucode_size & 3) - return -EINVAL; - - spin_lock_irqsave(&adev->smc_idx_lock, flags); - WREG32(mmSMC_IND_INDEX_0, ucode_start_address); - WREG32_P(mmSMC_IND_ACCESS_CNTL, SMC_IND_ACCESS_CNTL__AUTO_INCREMENT_IND_0_MASK, - ~SMC_IND_ACCESS_CNTL__AUTO_INCREMENT_IND_0_MASK); - while (ucode_size >= 4) { - /* SMC address space is BE */ - data = (src[0] << 24) | (src[1] << 16) | (src[2] << 8) | src[3]; - - WREG32(mmSMC_IND_DATA_0, data); - - src += 4; - ucode_size -= 4; - } - WREG32_P(mmSMC_IND_ACCESS_CNTL, 0, ~SMC_IND_ACCESS_CNTL__AUTO_INCREMENT_IND_0_MASK); - spin_unlock_irqrestore(&adev->smc_idx_lock, flags); - - return 0; -} - -int amdgpu_ci_read_smc_sram_dword(struct amdgpu_device *adev, - u32 smc_address, u32 *value, u32 limit) -{ - unsigned long flags; - int ret; - - spin_lock_irqsave(&adev->smc_idx_lock, flags); - ret = ci_set_smc_sram_address(adev, smc_address, limit); - if (ret == 0) - *value = RREG32(mmSMC_IND_DATA_0); - spin_unlock_irqrestore(&adev->smc_idx_lock, flags); - - return ret; -} - -int amdgpu_ci_write_smc_sram_dword(struct amdgpu_device *adev, - u32 smc_address, u32 value, u32 limit) -{ - unsigned long flags; - int ret; - - spin_lock_irqsave(&adev->smc_idx_lock, flags); - ret = ci_set_smc_sram_address(adev, smc_address, limit); - if (ret == 0) - WREG32(mmSMC_IND_DATA_0, value); - spin_unlock_irqrestore(&adev->smc_idx_lock, flags); - - return ret; -} diff --git a/drivers/gpu/drm/amd/amdgpu/cik.c b/drivers/gpu/drm/amd/amdgpu/cik.c index f41f5f57e9f3..07c1f239e9c3 100644 --- a/drivers/gpu/drm/amd/amdgpu/cik.c +++ b/drivers/gpu/drm/amd/amdgpu/cik.c @@ -1741,6 +1741,69 @@ static bool cik_need_full_reset(struct amdgpu_device *adev) return true; } +static void cik_get_pcie_usage(struct amdgpu_device *adev, uint64_t *count0, + uint64_t *count1) +{ + uint32_t perfctr = 0; + uint64_t cnt0_of, cnt1_of; + int tmp; + + /* This reports 0 on APUs, so return to avoid writing/reading registers + * that may or may not be different from their GPU counterparts + */ + if (adev->flags & AMD_IS_APU) + return; + + /* Set the 2 events that we wish to watch, defined above */ + /* Reg 40 is # received msgs, Reg 104 is # of posted requests sent */ + perfctr = REG_SET_FIELD(perfctr, PCIE_PERF_CNTL_TXCLK, EVENT0_SEL, 40); + perfctr = REG_SET_FIELD(perfctr, PCIE_PERF_CNTL_TXCLK, EVENT1_SEL, 104); + + /* Write to enable desired perf counters */ + WREG32_PCIE(ixPCIE_PERF_CNTL_TXCLK, perfctr); + /* Zero out and enable the perf counters + * Write 0x5: + * Bit 0 = Start all counters(1) + * Bit 2 = Global counter reset enable(1) + */ + WREG32_PCIE(ixPCIE_PERF_COUNT_CNTL, 0x00000005); + + msleep(1000); + + /* Load the shadow and disable the perf counters + * Write 0x2: + * Bit 0 = Stop counters(0) + * Bit 1 = Load the shadow counters(1) + */ + WREG32_PCIE(ixPCIE_PERF_COUNT_CNTL, 0x00000002); + + /* Read register values to get any >32bit overflow */ + tmp = RREG32_PCIE(ixPCIE_PERF_CNTL_TXCLK); + cnt0_of = REG_GET_FIELD(tmp, PCIE_PERF_CNTL_TXCLK, COUNTER0_UPPER); + cnt1_of = REG_GET_FIELD(tmp, PCIE_PERF_CNTL_TXCLK, COUNTER1_UPPER); + + /* Get the values and add the overflow */ + *count0 = RREG32_PCIE(ixPCIE_PERF_COUNT0_TXCLK) | (cnt0_of << 32); + *count1 = RREG32_PCIE(ixPCIE_PERF_COUNT1_TXCLK) | (cnt1_of << 32); +} + +static bool cik_need_reset_on_init(struct amdgpu_device *adev) +{ + u32 clock_cntl, pc; + + if (adev->flags & AMD_IS_APU) + return false; + + /* check if the SMC is already running */ + clock_cntl = RREG32_SMC(ixSMC_SYSCON_CLOCK_CNTL_0); + pc = RREG32_SMC(ixSMC_PC_C); + if ((0 == REG_GET_FIELD(clock_cntl, SMC_SYSCON_CLOCK_CNTL_0, ck_disable)) && + (0x20100 <= pc)) + return true; + + return false; +} + static const struct amdgpu_asic_funcs cik_asic_funcs = { .read_disabled_bios = &cik_read_disabled_bios, @@ -1755,6 +1818,9 @@ static const struct amdgpu_asic_funcs cik_asic_funcs = .flush_hdp = &cik_flush_hdp, .invalidate_hdp = &cik_invalidate_hdp, .need_full_reset = &cik_need_full_reset, + .init_doorbell_index = &legacy_doorbell_index_init, + .get_pcie_usage = &cik_get_pcie_usage, + .need_reset_on_init = &cik_need_reset_on_init, }; static int cik_common_early_init(void *handle) @@ -2004,10 +2070,7 @@ int cik_set_ip_blocks(struct amdgpu_device *adev) amdgpu_device_ip_block_add(adev, &cik_ih_ip_block); amdgpu_device_ip_block_add(adev, &gfx_v7_2_ip_block); amdgpu_device_ip_block_add(adev, &cik_sdma_ip_block); - if (amdgpu_dpm == -1) - amdgpu_device_ip_block_add(adev, &pp_smu_ip_block); - else - amdgpu_device_ip_block_add(adev, &ci_smu_ip_block); + amdgpu_device_ip_block_add(adev, &pp_smu_ip_block); if (adev->enable_virtual_display) amdgpu_device_ip_block_add(adev, &dce_virtual_ip_block); #if defined(CONFIG_DRM_AMD_DC) @@ -2025,10 +2088,7 @@ int cik_set_ip_blocks(struct amdgpu_device *adev) amdgpu_device_ip_block_add(adev, &cik_ih_ip_block); amdgpu_device_ip_block_add(adev, &gfx_v7_3_ip_block); amdgpu_device_ip_block_add(adev, &cik_sdma_ip_block); - if (amdgpu_dpm == -1) - amdgpu_device_ip_block_add(adev, &pp_smu_ip_block); - else - amdgpu_device_ip_block_add(adev, &ci_smu_ip_block); + amdgpu_device_ip_block_add(adev, &pp_smu_ip_block); if (adev->enable_virtual_display) amdgpu_device_ip_block_add(adev, &dce_virtual_ip_block); #if defined(CONFIG_DRM_AMD_DC) diff --git a/drivers/gpu/drm/amd/amdgpu/cik.h b/drivers/gpu/drm/amd/amdgpu/cik.h index e49c6f15a0a0..54c625a2e570 100644 --- a/drivers/gpu/drm/amd/amdgpu/cik.h +++ b/drivers/gpu/drm/amd/amdgpu/cik.h @@ -30,4 +30,5 @@ void cik_srbm_select(struct amdgpu_device *adev, u32 me, u32 pipe, u32 queue, u32 vmid); int cik_set_ip_blocks(struct amdgpu_device *adev); +void legacy_doorbell_index_init(struct amdgpu_device *adev); #endif diff --git a/drivers/gpu/drm/amd/amdgpu/cik_dpm.h b/drivers/gpu/drm/amd/amdgpu/cik_dpm.h index 2a086610f74d..2fcc4b60153c 100644 --- a/drivers/gpu/drm/amd/amdgpu/cik_dpm.h +++ b/drivers/gpu/drm/amd/amdgpu/cik_dpm.h @@ -24,7 +24,6 @@ #ifndef __CIK_DPM_H__ #define __CIK_DPM_H__ -extern const struct amdgpu_ip_block_version ci_smu_ip_block; extern const struct amdgpu_ip_block_version kv_smu_ip_block; #endif diff --git a/drivers/gpu/drm/amd/amdgpu/cik_ih.c b/drivers/gpu/drm/amd/amdgpu/cik_ih.c index b5775c6a857b..721c757156e8 100644 --- a/drivers/gpu/drm/amd/amdgpu/cik_ih.c +++ b/drivers/gpu/drm/amd/amdgpu/cik_ih.c @@ -103,9 +103,9 @@ static void cik_ih_disable_interrupts(struct amdgpu_device *adev) */ static int cik_ih_irq_init(struct amdgpu_device *adev) { + struct amdgpu_ih_ring *ih = &adev->irq.ih; int rb_bufsz; u32 interrupt_cntl, ih_cntl, ih_rb_cntl; - u64 wptr_off; /* disable irqs */ cik_ih_disable_interrupts(adev); @@ -131,9 +131,8 @@ static int cik_ih_irq_init(struct amdgpu_device *adev) ih_rb_cntl |= IH_RB_CNTL__WPTR_WRITEBACK_ENABLE_MASK; /* set the writeback address whether it's enabled or not */ - wptr_off = adev->wb.gpu_addr + (adev->irq.ih.wptr_offs * 4); - WREG32(mmIH_RB_WPTR_ADDR_LO, lower_32_bits(wptr_off)); - WREG32(mmIH_RB_WPTR_ADDR_HI, upper_32_bits(wptr_off) & 0xFF); + WREG32(mmIH_RB_WPTR_ADDR_LO, lower_32_bits(ih->wptr_addr)); + WREG32(mmIH_RB_WPTR_ADDR_HI, upper_32_bits(ih->wptr_addr) & 0xFF); WREG32(mmIH_RB_CNTL, ih_rb_cntl); @@ -183,11 +182,12 @@ static void cik_ih_irq_disable(struct amdgpu_device *adev) * Used by cik_irq_process(). * Returns the value of the wptr. */ -static u32 cik_ih_get_wptr(struct amdgpu_device *adev) +static u32 cik_ih_get_wptr(struct amdgpu_device *adev, + struct amdgpu_ih_ring *ih) { u32 wptr, tmp; - wptr = le32_to_cpu(adev->wb.wb[adev->irq.ih.wptr_offs]); + wptr = le32_to_cpu(*ih->wptr_cpu); if (wptr & IH_RB_WPTR__RB_OVERFLOW_MASK) { wptr &= ~IH_RB_WPTR__RB_OVERFLOW_MASK; @@ -196,13 +196,13 @@ static u32 cik_ih_get_wptr(struct amdgpu_device *adev) * this should allow us to catchup. */ dev_warn(adev->dev, "IH ring buffer overflow (0x%08X, 0x%08X, 0x%08X)\n", - wptr, adev->irq.ih.rptr, (wptr + 16) & adev->irq.ih.ptr_mask); - adev->irq.ih.rptr = (wptr + 16) & adev->irq.ih.ptr_mask; + wptr, ih->rptr, (wptr + 16) & ih->ptr_mask); + ih->rptr = (wptr + 16) & ih->ptr_mask; tmp = RREG32(mmIH_RB_CNTL); tmp |= IH_RB_CNTL__WPTR_OVERFLOW_CLEAR_MASK; WREG32(mmIH_RB_CNTL, tmp); } - return (wptr & adev->irq.ih.ptr_mask); + return (wptr & ih->ptr_mask); } /* CIK IV Ring @@ -228,34 +228,6 @@ static u32 cik_ih_get_wptr(struct amdgpu_device *adev) * [127:96] - reserved */ -/** - * cik_ih_prescreen_iv - prescreen an interrupt vector - * - * @adev: amdgpu_device pointer - * - * Returns true if the interrupt vector should be further processed. - */ -static bool cik_ih_prescreen_iv(struct amdgpu_device *adev) -{ - u32 ring_index = adev->irq.ih.rptr >> 2; - u16 pasid; - - switch (le32_to_cpu(adev->irq.ih.ring[ring_index]) & 0xff) { - case 146: - case 147: - pasid = le32_to_cpu(adev->irq.ih.ring[ring_index + 2]) >> 16; - if (!pasid || amdgpu_vm_pasid_fault_credit(adev, pasid)) - return true; - break; - default: - /* Not a VM fault */ - return true; - } - - adev->irq.ih.rptr += 16; - return false; -} - /** * cik_ih_decode_iv - decode an interrupt vector * @@ -265,16 +237,17 @@ static bool cik_ih_prescreen_iv(struct amdgpu_device *adev) * position and also advance the position. */ static void cik_ih_decode_iv(struct amdgpu_device *adev, + struct amdgpu_ih_ring *ih, struct amdgpu_iv_entry *entry) { /* wptr/rptr are in bytes! */ - u32 ring_index = adev->irq.ih.rptr >> 2; + u32 ring_index = ih->rptr >> 2; uint32_t dw[4]; - dw[0] = le32_to_cpu(adev->irq.ih.ring[ring_index + 0]); - dw[1] = le32_to_cpu(adev->irq.ih.ring[ring_index + 1]); - dw[2] = le32_to_cpu(adev->irq.ih.ring[ring_index + 2]); - dw[3] = le32_to_cpu(adev->irq.ih.ring[ring_index + 3]); + dw[0] = le32_to_cpu(ih->ring[ring_index + 0]); + dw[1] = le32_to_cpu(ih->ring[ring_index + 1]); + dw[2] = le32_to_cpu(ih->ring[ring_index + 2]); + dw[3] = le32_to_cpu(ih->ring[ring_index + 3]); entry->client_id = AMDGPU_IRQ_CLIENTID_LEGACY; entry->src_id = dw[0] & 0xff; @@ -284,7 +257,7 @@ static void cik_ih_decode_iv(struct amdgpu_device *adev, entry->pasid = (dw[2] >> 16) & 0xffff; /* wptr/rptr are in bytes! */ - adev->irq.ih.rptr += 16; + ih->rptr += 16; } /** @@ -294,9 +267,10 @@ static void cik_ih_decode_iv(struct amdgpu_device *adev, * * Set the IH ring buffer rptr. */ -static void cik_ih_set_rptr(struct amdgpu_device *adev) +static void cik_ih_set_rptr(struct amdgpu_device *adev, + struct amdgpu_ih_ring *ih) { - WREG32(mmIH_RB_RPTR, adev->irq.ih.rptr); + WREG32(mmIH_RB_RPTR, ih->rptr); } static int cik_ih_early_init(void *handle) @@ -461,7 +435,6 @@ static const struct amd_ip_funcs cik_ih_ip_funcs = { static const struct amdgpu_ih_funcs cik_ih_funcs = { .get_wptr = cik_ih_get_wptr, - .prescreen_iv = cik_ih_prescreen_iv, .decode_iv = cik_ih_decode_iv, .set_rptr = cik_ih_set_rptr }; diff --git a/drivers/gpu/drm/amd/amdgpu/cik_sdma.c b/drivers/gpu/drm/amd/amdgpu/cik_sdma.c index b918c8886b75..189599b694e8 100644 --- a/drivers/gpu/drm/amd/amdgpu/cik_sdma.c +++ b/drivers/gpu/drm/amd/amdgpu/cik_sdma.c @@ -198,7 +198,7 @@ static void cik_sdma_ring_set_wptr(struct amdgpu_ring *ring) static void cik_sdma_ring_insert_nop(struct amdgpu_ring *ring, uint32_t count) { - struct amdgpu_sdma_instance *sdma = amdgpu_get_sdma_instance(ring); + struct amdgpu_sdma_instance *sdma = amdgpu_sdma_get_instance_from_ring(ring); int i; for (i = 0; i < count; i++) @@ -218,9 +218,11 @@ static void cik_sdma_ring_insert_nop(struct amdgpu_ring *ring, uint32_t count) * Schedule an IB in the DMA ring (CIK). */ static void cik_sdma_ring_emit_ib(struct amdgpu_ring *ring, + struct amdgpu_job *job, struct amdgpu_ib *ib, - unsigned vmid, bool ctx_switch) + uint32_t flags) { + unsigned vmid = AMDGPU_JOB_GET_VMID(job); u32 extra_bits = vmid & 0xf; /* IB packet must end on a 8 DW boundary */ @@ -316,8 +318,8 @@ static void cik_sdma_gfx_stop(struct amdgpu_device *adev) WREG32(mmSDMA0_GFX_RB_CNTL + sdma_offsets[i], rb_cntl); WREG32(mmSDMA0_GFX_IB_CNTL + sdma_offsets[i], 0); } - sdma0->ready = false; - sdma1->ready = false; + sdma0->sched.ready = false; + sdma1->sched.ready = false; } /** @@ -494,18 +496,16 @@ static int cik_sdma_gfx_resume(struct amdgpu_device *adev) /* enable DMA IBs */ WREG32(mmSDMA0_GFX_IB_CNTL + sdma_offsets[i], ib_cntl); - ring->ready = true; + ring->sched.ready = true; } cik_sdma_enable(adev, true); for (i = 0; i < adev->sdma.num_instances; i++) { ring = &adev->sdma.instance[i].ring; - r = amdgpu_ring_test_ring(ring); - if (r) { - ring->ready = false; + r = amdgpu_ring_test_helper(ring); + if (r) return r; - } if (adev->mman.buffer_funcs_ring == ring) amdgpu_ttm_set_buffer_funcs_status(adev, true); @@ -618,21 +618,17 @@ static int cik_sdma_ring_test_ring(struct amdgpu_ring *ring) u64 gpu_addr; r = amdgpu_device_wb_get(adev, &index); - if (r) { - dev_err(adev->dev, "(%d) failed to allocate wb slot\n", r); + if (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 (r) { - DRM_ERROR("amdgpu: dma failed to lock ring %d (%d).\n", ring->idx, r); - amdgpu_device_wb_free(adev, index); - return r; - } + if (r) + goto error_free_wb; + amdgpu_ring_write(ring, SDMA_PACKET(SDMA_OPCODE_WRITE, SDMA_WRITE_SUB_OPCODE_LINEAR, 0)); amdgpu_ring_write(ring, lower_32_bits(gpu_addr)); amdgpu_ring_write(ring, upper_32_bits(gpu_addr)); @@ -647,15 +643,11 @@ static int cik_sdma_ring_test_ring(struct amdgpu_ring *ring) DRM_UDELAY(1); } - if (i < adev->usec_timeout) { - DRM_DEBUG("ring test on %d succeeded in %d usecs\n", ring->idx, i); - } else { - DRM_ERROR("amdgpu: ring %d test failed (0x%08X)\n", - ring->idx, tmp); - r = -EINVAL; - } - amdgpu_device_wb_free(adev, index); + if (i >= adev->usec_timeout) + r = -ETIMEDOUT; +error_free_wb: + amdgpu_device_wb_free(adev, index); return r; } @@ -678,20 +670,16 @@ static int cik_sdma_ring_test_ib(struct amdgpu_ring *ring, long timeout) long r; r = amdgpu_device_wb_get(adev, &index); - if (r) { - dev_err(adev->dev, "(%ld) failed to allocate wb slot\n", r); + if (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, &ib); - if (r) { - DRM_ERROR("amdgpu: failed to get ib (%ld).\n", r); + if (r) goto err0; - } ib.ptr[0] = SDMA_PACKET(SDMA_OPCODE_WRITE, SDMA_WRITE_SUB_OPCODE_LINEAR, 0); @@ -706,21 +694,16 @@ static int cik_sdma_ring_test_ib(struct amdgpu_ring *ring, long timeout) 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; } tmp = le32_to_cpu(adev->wb.wb[index]); - if (tmp == 0xDEADBEEF) { - DRM_DEBUG("ib test on ring %d succeeded\n", ring->idx); + if (tmp == 0xDEADBEEF) r = 0; - } else { - DRM_ERROR("amdgpu: ib test failed (0x%08X)\n", tmp); + else r = -EINVAL; - } err1: amdgpu_ib_free(adev, &ib, NULL); @@ -822,7 +805,7 @@ static void cik_sdma_vm_set_pte_pde(struct amdgpu_ib *ib, uint64_t pe, */ static void cik_sdma_ring_pad_ib(struct amdgpu_ring *ring, struct amdgpu_ib *ib) { - struct amdgpu_sdma_instance *sdma = amdgpu_get_sdma_instance(ring); + struct amdgpu_sdma_instance *sdma = amdgpu_sdma_get_instance_from_ring(ring); u32 pad_count; int i; @@ -1214,8 +1197,11 @@ static int cik_sdma_process_illegal_inst_irq(struct amdgpu_device *adev, struct amdgpu_irq_src *source, struct amdgpu_iv_entry *entry) { + u8 instance_id; + DRM_ERROR("Illegal instruction in SDMA command stream\n"); - schedule_work(&adev->reset_work); + instance_id = (entry->ring_id & 0x3) >> 0; + drm_sched_fault(&adev->sdma.instance[instance_id].ring.sched); return 0; } diff --git a/drivers/gpu/drm/amd/amdgpu/cz_ih.c b/drivers/gpu/drm/amd/amdgpu/cz_ih.c index df5ac4d85a00..61024b9c7a4b 100644 --- a/drivers/gpu/drm/amd/amdgpu/cz_ih.c +++ b/drivers/gpu/drm/amd/amdgpu/cz_ih.c @@ -103,9 +103,9 @@ static void cz_ih_disable_interrupts(struct amdgpu_device *adev) */ static int cz_ih_irq_init(struct amdgpu_device *adev) { - int rb_bufsz; + struct amdgpu_ih_ring *ih = &adev->irq.ih; u32 interrupt_cntl, ih_cntl, ih_rb_cntl; - u64 wptr_off; + int rb_bufsz; /* disable irqs */ cz_ih_disable_interrupts(adev); @@ -133,9 +133,8 @@ static int cz_ih_irq_init(struct amdgpu_device *adev) ih_rb_cntl = REG_SET_FIELD(ih_rb_cntl, IH_RB_CNTL, WPTR_WRITEBACK_ENABLE, 1); /* set the writeback address whether it's enabled or not */ - wptr_off = adev->wb.gpu_addr + (adev->irq.ih.wptr_offs * 4); - WREG32(mmIH_RB_WPTR_ADDR_LO, lower_32_bits(wptr_off)); - WREG32(mmIH_RB_WPTR_ADDR_HI, upper_32_bits(wptr_off) & 0xFF); + WREG32(mmIH_RB_WPTR_ADDR_LO, lower_32_bits(ih->wptr_addr)); + WREG32(mmIH_RB_WPTR_ADDR_HI, upper_32_bits(ih->wptr_addr) & 0xFF); WREG32(mmIH_RB_CNTL, ih_rb_cntl); @@ -185,11 +184,12 @@ static void cz_ih_irq_disable(struct amdgpu_device *adev) * Used by cz_irq_process(VI). * Returns the value of the wptr. */ -static u32 cz_ih_get_wptr(struct amdgpu_device *adev) +static u32 cz_ih_get_wptr(struct amdgpu_device *adev, + struct amdgpu_ih_ring *ih) { u32 wptr, tmp; - wptr = le32_to_cpu(adev->wb.wb[adev->irq.ih.wptr_offs]); + wptr = le32_to_cpu(*ih->wptr_cpu); if (REG_GET_FIELD(wptr, IH_RB_WPTR, RB_OVERFLOW)) { wptr = REG_SET_FIELD(wptr, IH_RB_WPTR, RB_OVERFLOW, 0); @@ -198,41 +198,13 @@ static u32 cz_ih_get_wptr(struct amdgpu_device *adev) * this should allow us to catchup. */ dev_warn(adev->dev, "IH ring buffer overflow (0x%08X, 0x%08X, 0x%08X)\n", - wptr, adev->irq.ih.rptr, (wptr + 16) & adev->irq.ih.ptr_mask); - adev->irq.ih.rptr = (wptr + 16) & adev->irq.ih.ptr_mask; + wptr, ih->rptr, (wptr + 16) & ih->ptr_mask); + ih->rptr = (wptr + 16) & ih->ptr_mask; tmp = RREG32(mmIH_RB_CNTL); tmp = REG_SET_FIELD(tmp, IH_RB_CNTL, WPTR_OVERFLOW_CLEAR, 1); WREG32(mmIH_RB_CNTL, tmp); } - return (wptr & adev->irq.ih.ptr_mask); -} - -/** - * cz_ih_prescreen_iv - prescreen an interrupt vector - * - * @adev: amdgpu_device pointer - * - * Returns true if the interrupt vector should be further processed. - */ -static bool cz_ih_prescreen_iv(struct amdgpu_device *adev) -{ - u32 ring_index = adev->irq.ih.rptr >> 2; - u16 pasid; - - switch (le32_to_cpu(adev->irq.ih.ring[ring_index]) & 0xff) { - case 146: - case 147: - pasid = le32_to_cpu(adev->irq.ih.ring[ring_index + 2]) >> 16; - if (!pasid || amdgpu_vm_pasid_fault_credit(adev, pasid)) - return true; - break; - default: - /* Not a VM fault */ - return true; - } - - adev->irq.ih.rptr += 16; - return false; + return (wptr & ih->ptr_mask); } /** @@ -244,16 +216,17 @@ static bool cz_ih_prescreen_iv(struct amdgpu_device *adev) * position and also advance the position. */ static void cz_ih_decode_iv(struct amdgpu_device *adev, - struct amdgpu_iv_entry *entry) + struct amdgpu_ih_ring *ih, + struct amdgpu_iv_entry *entry) { /* wptr/rptr are in bytes! */ - u32 ring_index = adev->irq.ih.rptr >> 2; + u32 ring_index = ih->rptr >> 2; uint32_t dw[4]; - dw[0] = le32_to_cpu(adev->irq.ih.ring[ring_index + 0]); - dw[1] = le32_to_cpu(adev->irq.ih.ring[ring_index + 1]); - dw[2] = le32_to_cpu(adev->irq.ih.ring[ring_index + 2]); - dw[3] = le32_to_cpu(adev->irq.ih.ring[ring_index + 3]); + dw[0] = le32_to_cpu(ih->ring[ring_index + 0]); + dw[1] = le32_to_cpu(ih->ring[ring_index + 1]); + dw[2] = le32_to_cpu(ih->ring[ring_index + 2]); + dw[3] = le32_to_cpu(ih->ring[ring_index + 3]); entry->client_id = AMDGPU_IRQ_CLIENTID_LEGACY; entry->src_id = dw[0] & 0xff; @@ -263,7 +236,7 @@ static void cz_ih_decode_iv(struct amdgpu_device *adev, entry->pasid = (dw[2] >> 16) & 0xffff; /* wptr/rptr are in bytes! */ - adev->irq.ih.rptr += 16; + ih->rptr += 16; } /** @@ -273,9 +246,10 @@ static void cz_ih_decode_iv(struct amdgpu_device *adev, * * Set the IH ring buffer rptr. */ -static void cz_ih_set_rptr(struct amdgpu_device *adev) +static void cz_ih_set_rptr(struct amdgpu_device *adev, + struct amdgpu_ih_ring *ih) { - WREG32(mmIH_RB_RPTR, adev->irq.ih.rptr); + WREG32(mmIH_RB_RPTR, ih->rptr); } static int cz_ih_early_init(void *handle) @@ -442,7 +416,6 @@ static const struct amd_ip_funcs cz_ih_ip_funcs = { static const struct amdgpu_ih_funcs cz_ih_funcs = { .get_wptr = cz_ih_get_wptr, - .prescreen_iv = cz_ih_prescreen_iv, .decode_iv = cz_ih_decode_iv, .set_rptr = cz_ih_set_rptr }; diff --git a/drivers/gpu/drm/amd/amdgpu/dce_v10_0.c b/drivers/gpu/drm/amd/amdgpu/dce_v10_0.c index 4cfecdce29a3..1f0426d2fc2a 100644 --- a/drivers/gpu/drm/amd/amdgpu/dce_v10_0.c +++ b/drivers/gpu/drm/amd/amdgpu/dce_v10_0.c @@ -1682,7 +1682,7 @@ static void dce_v10_0_afmt_setmode(struct drm_encoder *encoder, dce_v10_0_audio_write_sad_regs(encoder); dce_v10_0_audio_write_latency_fields(encoder, mode); - err = drm_hdmi_avi_infoframe_from_display_mode(&frame, mode, false); + err = drm_hdmi_avi_infoframe_from_display_mode(&frame, connector, mode); if (err < 0) { DRM_ERROR("failed to setup AVI infoframe: %zd\n", err); return; diff --git a/drivers/gpu/drm/amd/amdgpu/dce_v11_0.c b/drivers/gpu/drm/amd/amdgpu/dce_v11_0.c index 7c868916d90f..2280b971d758 100644 --- a/drivers/gpu/drm/amd/amdgpu/dce_v11_0.c +++ b/drivers/gpu/drm/amd/amdgpu/dce_v11_0.c @@ -1724,7 +1724,7 @@ static void dce_v11_0_afmt_setmode(struct drm_encoder *encoder, dce_v11_0_audio_write_sad_regs(encoder); dce_v11_0_audio_write_latency_fields(encoder, mode); - err = drm_hdmi_avi_infoframe_from_display_mode(&frame, mode, false); + err = drm_hdmi_avi_infoframe_from_display_mode(&frame, connector, mode); if (err < 0) { DRM_ERROR("failed to setup AVI infoframe: %zd\n", err); return; diff --git a/drivers/gpu/drm/amd/amdgpu/dce_v6_0.c b/drivers/gpu/drm/amd/amdgpu/dce_v6_0.c index 17eaaba36017..bea32f076b91 100644 --- a/drivers/gpu/drm/amd/amdgpu/dce_v6_0.c +++ b/drivers/gpu/drm/amd/amdgpu/dce_v6_0.c @@ -1423,6 +1423,7 @@ static void dce_v6_0_audio_set_avi_infoframe(struct drm_encoder *encoder, struct amdgpu_device *adev = dev->dev_private; struct amdgpu_encoder *amdgpu_encoder = to_amdgpu_encoder(encoder); struct amdgpu_encoder_atom_dig *dig = amdgpu_encoder->enc_priv; + struct drm_connector *connector = amdgpu_get_connector_for_encoder(encoder); struct hdmi_avi_infoframe frame; u8 buffer[HDMI_INFOFRAME_HEADER_SIZE + HDMI_AVI_INFOFRAME_SIZE]; uint8_t *payload = buffer + 3; @@ -1430,7 +1431,7 @@ static void dce_v6_0_audio_set_avi_infoframe(struct drm_encoder *encoder, ssize_t err; u32 tmp; - err = drm_hdmi_avi_infoframe_from_display_mode(&frame, mode, false); + err = drm_hdmi_avi_infoframe_from_display_mode(&frame, connector, mode); if (err < 0) { DRM_ERROR("failed to setup AVI infoframe: %zd\n", err); return; @@ -2979,7 +2980,7 @@ static int dce_v6_0_pageflip_irq(struct amdgpu_device *adev, struct amdgpu_irq_src *source, struct amdgpu_iv_entry *entry) { - unsigned long flags; + unsigned long flags; unsigned crtc_id; struct amdgpu_crtc *amdgpu_crtc; struct amdgpu_flip_work *works; diff --git a/drivers/gpu/drm/amd/amdgpu/dce_v8_0.c b/drivers/gpu/drm/amd/amdgpu/dce_v8_0.c index 8c0576978d36..13da915991dd 100644 --- a/drivers/gpu/drm/amd/amdgpu/dce_v8_0.c +++ b/drivers/gpu/drm/amd/amdgpu/dce_v8_0.c @@ -1616,7 +1616,7 @@ static void dce_v8_0_afmt_setmode(struct drm_encoder *encoder, dce_v8_0_audio_write_sad_regs(encoder); dce_v8_0_audio_write_latency_fields(encoder, mode); - err = drm_hdmi_avi_infoframe_from_display_mode(&frame, mode, false); + err = drm_hdmi_avi_infoframe_from_display_mode(&frame, connector, mode); if (err < 0) { DRM_ERROR("failed to setup AVI infoframe: %zd\n", err); return; diff --git a/drivers/gpu/drm/amd/amdgpu/dce_virtual.c b/drivers/gpu/drm/amd/amdgpu/dce_virtual.c index fdace004544d..e4cc1d48eaab 100644 --- a/drivers/gpu/drm/amd/amdgpu/dce_virtual.c +++ b/drivers/gpu/drm/amd/amdgpu/dce_virtual.c @@ -167,19 +167,6 @@ static void dce_virtual_crtc_disable(struct drm_crtc *crtc) struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc); dce_virtual_crtc_dpms(crtc, DRM_MODE_DPMS_OFF); - if (crtc->primary->fb) { - int r; - struct amdgpu_bo *abo; - - abo = gem_to_amdgpu_bo(crtc->primary->fb->obj[0]); - r = amdgpu_bo_reserve(abo, true); - if (unlikely(r)) - DRM_ERROR("failed to reserve abo before unpin\n"); - else { - amdgpu_bo_unpin(abo); - amdgpu_bo_unreserve(abo); - } - } amdgpu_crtc->pll_id = ATOM_PPLL_INVALID; amdgpu_crtc->encoder = NULL; @@ -692,7 +679,9 @@ static int dce_virtual_pageflip(struct amdgpu_device *adev, spin_unlock_irqrestore(&adev->ddev->event_lock, flags); drm_crtc_vblank_put(&amdgpu_crtc->base); - schedule_work(&works->unpin_work); + amdgpu_bo_unref(&works->old_abo); + kfree(works->shared); + kfree(works); return 0; } diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v6_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v6_0.c index d76eb27945dc..305276c7e4bf 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v6_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v6_0.c @@ -1775,18 +1775,15 @@ static int gfx_v6_0_ring_test_ring(struct amdgpu_ring *ring) int r; r = amdgpu_gfx_scratch_get(adev, &scratch); - if (r) { - DRM_ERROR("amdgpu: cp failed to get scratch reg (%d).\n", r); + if (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; - } + if (r) + goto error_free_scratch; + amdgpu_ring_write(ring, PACKET3(PACKET3_SET_CONFIG_REG, 1)); amdgpu_ring_write(ring, (scratch - PACKET3_SET_CONFIG_REG_START)); amdgpu_ring_write(ring, 0xDEADBEEF); @@ -1798,13 +1795,11 @@ static int gfx_v6_0_ring_test_ring(struct amdgpu_ring *ring) break; DRM_UDELAY(1); } - if (i < adev->usec_timeout) { - DRM_DEBUG("ring test on %d succeeded in %d usecs\n", ring->idx, i); - } else { - DRM_ERROR("amdgpu: ring %d test failed (scratch(0x%04X)=0x%08X)\n", - ring->idx, scratch, tmp); - r = -EINVAL; - } + + if (i >= adev->usec_timeout) + r = -ETIMEDOUT; + +error_free_scratch: amdgpu_gfx_scratch_free(adev, scratch); return r; } @@ -1845,13 +1840,15 @@ static void gfx_v6_0_ring_emit_fence(struct amdgpu_ring *ring, u64 addr, } static void gfx_v6_0_ring_emit_ib(struct amdgpu_ring *ring, + struct amdgpu_job *job, struct amdgpu_ib *ib, - unsigned vmid, bool ctx_switch) + uint32_t flags) { + unsigned vmid = AMDGPU_JOB_GET_VMID(job); u32 header, control = 0; /* insert SWITCH_BUFFER packet before first IB in the ring frame */ - if (ctx_switch) { + if (flags & AMDGPU_HAVE_CTX_SWITCH) { amdgpu_ring_write(ring, PACKET3(PACKET3_SWITCH_BUFFER, 0)); amdgpu_ring_write(ring, 0); } @@ -1892,17 +1889,15 @@ static int gfx_v6_0_ring_test_ib(struct amdgpu_ring *ring, long timeout) long r; r = amdgpu_gfx_scratch_get(adev, &scratch); - if (r) { - DRM_ERROR("amdgpu: failed to get scratch reg (%ld).\n", r); + if (r) return r; - } + WREG32(scratch, 0xCAFEDEAD); memset(&ib, 0, sizeof(ib)); r = amdgpu_ib_get(adev, NULL, 256, &ib); - if (r) { - DRM_ERROR("amdgpu: failed to get ib (%ld).\n", r); + if (r) goto err1; - } + ib.ptr[0] = PACKET3(PACKET3_SET_CONFIG_REG, 1); ib.ptr[1] = ((scratch - PACKET3_SET_CONFIG_REG_START)); ib.ptr[2] = 0xDEADBEEF; @@ -1914,22 +1909,16 @@ static int gfx_v6_0_ring_test_ib(struct amdgpu_ring *ring, long timeout) r = dma_fence_wait_timeout(f, false, timeout); if (r == 0) { - DRM_ERROR("amdgpu: IB test timed out\n"); r = -ETIMEDOUT; goto err2; } else if (r < 0) { - DRM_ERROR("amdgpu: fence wait failed (%ld).\n", r); goto err2; } tmp = RREG32(scratch); - if (tmp == 0xDEADBEEF) { - DRM_DEBUG("ib test on ring %d succeeded\n", ring->idx); + if (tmp == 0xDEADBEEF) r = 0; - } else { - DRM_ERROR("amdgpu: ib test failed (scratch(0x%04X)=0x%08X)\n", - scratch, tmp); + else r = -EINVAL; - } err2: amdgpu_ib_free(adev, &ib, NULL); @@ -1950,9 +1939,9 @@ static void gfx_v6_0_cp_gfx_enable(struct amdgpu_device *adev, bool enable) CP_ME_CNTL__CE_HALT_MASK)); WREG32(mmSCRATCH_UMSK, 0); for (i = 0; i < adev->gfx.num_gfx_rings; i++) - adev->gfx.gfx_ring[i].ready = false; + adev->gfx.gfx_ring[i].sched.ready = false; for (i = 0; i < adev->gfx.num_compute_rings; i++) - adev->gfx.compute_ring[i].ready = false; + adev->gfx.compute_ring[i].sched.ready = false; } udelay(50); } @@ -2124,12 +2113,9 @@ static int gfx_v6_0_cp_gfx_resume(struct amdgpu_device *adev) /* start the rings */ gfx_v6_0_cp_gfx_start(adev); - ring->ready = true; - r = amdgpu_ring_test_ring(ring); - if (r) { - ring->ready = false; + r = amdgpu_ring_test_helper(ring); + if (r) return r; - } return 0; } @@ -2227,14 +2213,11 @@ static int gfx_v6_0_cp_compute_resume(struct amdgpu_device *adev) WREG32(mmCP_RB2_CNTL, tmp); WREG32(mmCP_RB2_BASE, ring->gpu_addr >> 8); - adev->gfx.compute_ring[0].ready = false; - adev->gfx.compute_ring[1].ready = false; for (i = 0; i < 2; i++) { - r = amdgpu_ring_test_ring(&adev->gfx.compute_ring[i]); + r = amdgpu_ring_test_helper(&adev->gfx.compute_ring[i]); if (r) return r; - adev->gfx.compute_ring[i].ready = true; } return 0; @@ -2368,18 +2351,11 @@ static void gfx_v6_0_ring_emit_wreg(struct amdgpu_ring *ring, amdgpu_ring_write(ring, val); } -static void gfx_v6_0_rlc_fini(struct amdgpu_device *adev) -{ - amdgpu_bo_free_kernel(&adev->gfx.rlc.save_restore_obj, NULL, NULL); - amdgpu_bo_free_kernel(&adev->gfx.rlc.clear_state_obj, NULL, NULL); - amdgpu_bo_free_kernel(&adev->gfx.rlc.cp_table_obj, NULL, NULL); -} - static int gfx_v6_0_rlc_init(struct amdgpu_device *adev) { const u32 *src_ptr; volatile u32 *dst_ptr; - u32 dws, i; + u32 dws; u64 reg_list_mc_addr; const struct cs_section_def *cs_data; int r; @@ -2394,26 +2370,10 @@ static int gfx_v6_0_rlc_init(struct amdgpu_device *adev) cs_data = adev->gfx.rlc.cs_data; if (src_ptr) { - /* save restore block */ - r = amdgpu_bo_create_reserved(adev, dws * 4, PAGE_SIZE, - AMDGPU_GEM_DOMAIN_VRAM, - &adev->gfx.rlc.save_restore_obj, - &adev->gfx.rlc.save_restore_gpu_addr, - (void **)&adev->gfx.rlc.sr_ptr); - if (r) { - dev_warn(adev->dev, "(%d) create RLC sr bo failed\n", - r); - gfx_v6_0_rlc_fini(adev); + /* init save restore block */ + r = amdgpu_gfx_rlc_init_sr(adev, dws); + if (r) return r; - } - - /* write the sr buffer */ - dst_ptr = adev->gfx.rlc.sr_ptr; - for (i = 0; i < adev->gfx.rlc.reg_list_size; i++) - dst_ptr[i] = cpu_to_le32(src_ptr[i]); - - amdgpu_bo_kunmap(adev->gfx.rlc.save_restore_obj); - amdgpu_bo_unreserve(adev->gfx.rlc.save_restore_obj); } if (cs_data) { @@ -2428,7 +2388,7 @@ static int gfx_v6_0_rlc_init(struct amdgpu_device *adev) (void **)&adev->gfx.rlc.cs_ptr); if (r) { dev_warn(adev->dev, "(%d) create RLC c bo failed\n", r); - gfx_v6_0_rlc_fini(adev); + amdgpu_gfx_rlc_fini(adev); return r; } @@ -2549,8 +2509,8 @@ static int gfx_v6_0_rlc_resume(struct amdgpu_device *adev) if (!adev->gfx.rlc_fw) return -EINVAL; - gfx_v6_0_rlc_stop(adev); - gfx_v6_0_rlc_reset(adev); + adev->gfx.rlc.funcs->stop(adev); + adev->gfx.rlc.funcs->reset(adev); gfx_v6_0_init_pg(adev); gfx_v6_0_init_cg(adev); @@ -2578,7 +2538,7 @@ static int gfx_v6_0_rlc_resume(struct amdgpu_device *adev) WREG32(mmRLC_UCODE_ADDR, 0); gfx_v6_0_enable_lbpw(adev, gfx_v6_0_lbpw_supported(adev)); - gfx_v6_0_rlc_start(adev); + adev->gfx.rlc.funcs->start(adev); return 0; } @@ -3075,6 +3035,14 @@ static const struct amdgpu_gfx_funcs gfx_v6_0_gfx_funcs = { .select_me_pipe_q = &gfx_v6_0_select_me_pipe_q }; +static const struct amdgpu_rlc_funcs gfx_v6_0_rlc_funcs = { + .init = gfx_v6_0_rlc_init, + .resume = gfx_v6_0_rlc_resume, + .stop = gfx_v6_0_rlc_stop, + .reset = gfx_v6_0_rlc_reset, + .start = gfx_v6_0_rlc_start +}; + static int gfx_v6_0_early_init(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; @@ -3082,6 +3050,7 @@ static int gfx_v6_0_early_init(void *handle) adev->gfx.num_gfx_rings = GFX6_NUM_GFX_RINGS; adev->gfx.num_compute_rings = GFX6_NUM_COMPUTE_RINGS; adev->gfx.funcs = &gfx_v6_0_gfx_funcs; + adev->gfx.rlc.funcs = &gfx_v6_0_rlc_funcs; gfx_v6_0_set_ring_funcs(adev); gfx_v6_0_set_irq_funcs(adev); @@ -3114,7 +3083,7 @@ static int gfx_v6_0_sw_init(void *handle) return r; } - r = gfx_v6_0_rlc_init(adev); + r = adev->gfx.rlc.funcs->init(adev); if (r) { DRM_ERROR("Failed to init rlc BOs!\n"); return r; @@ -3165,7 +3134,7 @@ static int gfx_v6_0_sw_fini(void *handle) for (i = 0; i < adev->gfx.num_compute_rings; i++) amdgpu_ring_fini(&adev->gfx.compute_ring[i]); - gfx_v6_0_rlc_fini(adev); + amdgpu_gfx_rlc_fini(adev); return 0; } @@ -3177,7 +3146,7 @@ static int gfx_v6_0_hw_init(void *handle) gfx_v6_0_constants_init(adev); - r = gfx_v6_0_rlc_resume(adev); + r = adev->gfx.rlc.funcs->resume(adev); if (r) return r; @@ -3195,7 +3164,7 @@ static int gfx_v6_0_hw_fini(void *handle) struct amdgpu_device *adev = (struct amdgpu_device *)handle; gfx_v6_0_cp_enable(adev, false); - gfx_v6_0_rlc_stop(adev); + adev->gfx.rlc.funcs->stop(adev); gfx_v6_0_fini_pg(adev); return 0; @@ -3393,12 +3362,31 @@ static int gfx_v6_0_eop_irq(struct amdgpu_device *adev, return 0; } +static void gfx_v6_0_fault(struct amdgpu_device *adev, + struct amdgpu_iv_entry *entry) +{ + struct amdgpu_ring *ring; + + switch (entry->ring_id) { + case 0: + ring = &adev->gfx.gfx_ring[0]; + break; + case 1: + case 2: + ring = &adev->gfx.compute_ring[entry->ring_id - 1]; + break; + default: + return; + } + drm_sched_fault(&ring->sched); +} + static int gfx_v6_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"); - schedule_work(&adev->reset_work); + gfx_v6_0_fault(adev, entry); return 0; } @@ -3407,7 +3395,7 @@ static int gfx_v6_0_priv_inst_irq(struct amdgpu_device *adev, struct amdgpu_iv_entry *entry) { DRM_ERROR("Illegal instruction in command stream\n"); - schedule_work(&adev->reset_work); + gfx_v6_0_fault(adev, entry); return 0; } diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c index 0e72bc09939a..a59e0fdf5a97 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c @@ -882,7 +882,6 @@ static const u32 kalindi_rlc_save_restore_register_list[] = static u32 gfx_v7_0_get_csb_size(struct amdgpu_device *adev); static void gfx_v7_0_get_csb_buffer(struct amdgpu_device *adev, volatile u32 *buffer); -static void gfx_v7_0_init_cp_pg_table(struct amdgpu_device *adev); static void gfx_v7_0_init_pg(struct amdgpu_device *adev); static void gfx_v7_0_get_cu_info(struct amdgpu_device *adev); @@ -2064,17 +2063,14 @@ static int gfx_v7_0_ring_test_ring(struct amdgpu_ring *ring) int r; r = amdgpu_gfx_scratch_get(adev, &scratch); - if (r) { - DRM_ERROR("amdgpu: cp failed to get scratch reg (%d).\n", r); + if (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; - } + if (r) + goto error_free_scratch; + 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); @@ -2086,13 +2082,10 @@ static int gfx_v7_0_ring_test_ring(struct amdgpu_ring *ring) break; DRM_UDELAY(1); } - if (i < adev->usec_timeout) { - DRM_DEBUG("ring test on %d succeeded in %d usecs\n", ring->idx, i); - } else { - DRM_ERROR("amdgpu: ring %d test failed (scratch(0x%04X)=0x%08X)\n", - ring->idx, scratch, tmp); - r = -EINVAL; - } + if (i >= adev->usec_timeout) + r = -ETIMEDOUT; + +error_free_scratch: amdgpu_gfx_scratch_free(adev, scratch); return r; } @@ -2233,13 +2226,15 @@ static void gfx_v7_0_ring_emit_fence_compute(struct amdgpu_ring *ring, * on the gfx ring for execution by the GPU. */ static void gfx_v7_0_ring_emit_ib_gfx(struct amdgpu_ring *ring, - struct amdgpu_ib *ib, - unsigned vmid, bool ctx_switch) + struct amdgpu_job *job, + struct amdgpu_ib *ib, + uint32_t flags) { + unsigned vmid = AMDGPU_JOB_GET_VMID(job); u32 header, control = 0; /* insert SWITCH_BUFFER packet before first IB in the ring frame */ - if (ctx_switch) { + if (flags & AMDGPU_HAVE_CTX_SWITCH) { amdgpu_ring_write(ring, PACKET3(PACKET3_SWITCH_BUFFER, 0)); amdgpu_ring_write(ring, 0); } @@ -2262,11 +2257,29 @@ static void gfx_v7_0_ring_emit_ib_gfx(struct amdgpu_ring *ring, } static void gfx_v7_0_ring_emit_ib_compute(struct amdgpu_ring *ring, + struct amdgpu_job *job, struct amdgpu_ib *ib, - unsigned vmid, bool ctx_switch) + uint32_t flags) { + unsigned vmid = AMDGPU_JOB_GET_VMID(job); u32 control = INDIRECT_BUFFER_VALID | ib->length_dw | (vmid << 24); + /* 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, mmGDS_COMPUTE_MAX_WAVE_ID - PACKET3_SET_CONFIG_REG_START); + amdgpu_ring_write(ring, ring->adev->gds.gds_compute_max_wave_id); + } + amdgpu_ring_write(ring, PACKET3(PACKET3_INDIRECT_BUFFER, 2)); amdgpu_ring_write(ring, #ifdef __BIG_ENDIAN @@ -2316,17 +2329,15 @@ static int gfx_v7_0_ring_test_ib(struct amdgpu_ring *ring, long timeout) long r; r = amdgpu_gfx_scratch_get(adev, &scratch); - if (r) { - DRM_ERROR("amdgpu: failed to get scratch reg (%ld).\n", r); + if (r) return r; - } + WREG32(scratch, 0xCAFEDEAD); memset(&ib, 0, sizeof(ib)); r = amdgpu_ib_get(adev, NULL, 256, &ib); - if (r) { - DRM_ERROR("amdgpu: failed to get ib (%ld).\n", r); + if (r) goto err1; - } + ib.ptr[0] = PACKET3(PACKET3_SET_UCONFIG_REG, 1); ib.ptr[1] = ((scratch - PACKET3_SET_UCONFIG_REG_START)); ib.ptr[2] = 0xDEADBEEF; @@ -2338,22 +2349,16 @@ static int gfx_v7_0_ring_test_ib(struct amdgpu_ring *ring, long timeout) r = dma_fence_wait_timeout(f, false, timeout); if (r == 0) { - DRM_ERROR("amdgpu: IB test timed out\n"); r = -ETIMEDOUT; goto err2; } else if (r < 0) { - DRM_ERROR("amdgpu: fence wait failed (%ld).\n", r); goto err2; } tmp = RREG32(scratch); - if (tmp == 0xDEADBEEF) { - DRM_DEBUG("ib test on ring %d succeeded\n", ring->idx); + if (tmp == 0xDEADBEEF) r = 0; - } else { - DRM_ERROR("amdgpu: ib test failed (scratch(0x%04X)=0x%08X)\n", - scratch, tmp); + else r = -EINVAL; - } err2: amdgpu_ib_free(adev, &ib, NULL); @@ -2403,7 +2408,7 @@ static void gfx_v7_0_cp_gfx_enable(struct amdgpu_device *adev, bool enable) } else { WREG32(mmCP_ME_CNTL, (CP_ME_CNTL__ME_HALT_MASK | CP_ME_CNTL__PFP_HALT_MASK | CP_ME_CNTL__CE_HALT_MASK)); for (i = 0; i < adev->gfx.num_gfx_rings; i++) - adev->gfx.gfx_ring[i].ready = false; + adev->gfx.gfx_ring[i].sched.ready = false; } udelay(50); } @@ -2613,12 +2618,9 @@ static int gfx_v7_0_cp_gfx_resume(struct amdgpu_device *adev) /* start the ring */ gfx_v7_0_cp_gfx_start(adev); - ring->ready = true; - r = amdgpu_ring_test_ring(ring); - if (r) { - ring->ready = false; + r = amdgpu_ring_test_helper(ring); + if (r) return r; - } return 0; } @@ -2675,7 +2677,7 @@ static void gfx_v7_0_cp_compute_enable(struct amdgpu_device *adev, bool enable) } else { WREG32(mmCP_MEC_CNTL, (CP_MEC_CNTL__MEC_ME1_HALT_MASK | CP_MEC_CNTL__MEC_ME2_HALT_MASK)); for (i = 0; i < adev->gfx.num_compute_rings; i++) - adev->gfx.compute_ring[i].ready = false; + adev->gfx.compute_ring[i].sched.ready = false; } udelay(50); } @@ -2781,7 +2783,7 @@ static int gfx_v7_0_mec_init(struct amdgpu_device *adev) * GFX7_MEC_HPD_SIZE * 2; r = amdgpu_bo_create_reserved(adev, mec_hpd_size, PAGE_SIZE, - AMDGPU_GEM_DOMAIN_GTT, + AMDGPU_GEM_DOMAIN_VRAM, &adev->gfx.mec.hpd_eop_obj, &adev->gfx.mec.hpd_eop_gpu_addr, (void **)&hpd); @@ -3106,10 +3108,7 @@ static int gfx_v7_0_cp_compute_resume(struct amdgpu_device *adev) for (i = 0; i < adev->gfx.num_compute_rings; i++) { ring = &adev->gfx.compute_ring[i]; - ring->ready = true; - r = amdgpu_ring_test_ring(ring); - if (r) - ring->ready = false; + amdgpu_ring_test_helper(ring); } return 0; @@ -3268,18 +3267,10 @@ static void gfx_v7_0_ring_emit_wreg(struct amdgpu_ring *ring, * The RLC is a multi-purpose microengine that handles a * variety of functions. */ -static void gfx_v7_0_rlc_fini(struct amdgpu_device *adev) -{ - amdgpu_bo_free_kernel(&adev->gfx.rlc.save_restore_obj, NULL, NULL); - amdgpu_bo_free_kernel(&adev->gfx.rlc.clear_state_obj, NULL, NULL); - amdgpu_bo_free_kernel(&adev->gfx.rlc.cp_table_obj, NULL, NULL); -} - static int gfx_v7_0_rlc_init(struct amdgpu_device *adev) { const u32 *src_ptr; - volatile u32 *dst_ptr; - u32 dws, i; + u32 dws; const struct cs_section_def *cs_data; int r; @@ -3306,66 +3297,23 @@ static int gfx_v7_0_rlc_init(struct amdgpu_device *adev) cs_data = adev->gfx.rlc.cs_data; if (src_ptr) { - /* save restore block */ - r = amdgpu_bo_create_reserved(adev, dws * 4, PAGE_SIZE, - AMDGPU_GEM_DOMAIN_VRAM, - &adev->gfx.rlc.save_restore_obj, - &adev->gfx.rlc.save_restore_gpu_addr, - (void **)&adev->gfx.rlc.sr_ptr); - if (r) { - dev_warn(adev->dev, "(%d) create, pin or map of RLC sr bo failed\n", r); - gfx_v7_0_rlc_fini(adev); + /* init save restore block */ + r = amdgpu_gfx_rlc_init_sr(adev, dws); + if (r) return r; - } - - /* write the sr buffer */ - dst_ptr = adev->gfx.rlc.sr_ptr; - for (i = 0; i < adev->gfx.rlc.reg_list_size; i++) - dst_ptr[i] = cpu_to_le32(src_ptr[i]); - amdgpu_bo_kunmap(adev->gfx.rlc.save_restore_obj); - amdgpu_bo_unreserve(adev->gfx.rlc.save_restore_obj); } if (cs_data) { - /* clear state block */ - adev->gfx.rlc.clear_state_size = dws = gfx_v7_0_get_csb_size(adev); - - r = amdgpu_bo_create_reserved(adev, dws * 4, PAGE_SIZE, - AMDGPU_GEM_DOMAIN_VRAM, - &adev->gfx.rlc.clear_state_obj, - &adev->gfx.rlc.clear_state_gpu_addr, - (void **)&adev->gfx.rlc.cs_ptr); - if (r) { - dev_warn(adev->dev, "(%d) create RLC c bo failed\n", r); - gfx_v7_0_rlc_fini(adev); + /* init clear state block */ + r = amdgpu_gfx_rlc_init_csb(adev); + if (r) return r; - } - - /* set up the cs buffer */ - dst_ptr = adev->gfx.rlc.cs_ptr; - gfx_v7_0_get_csb_buffer(adev, dst_ptr); - amdgpu_bo_kunmap(adev->gfx.rlc.clear_state_obj); - amdgpu_bo_unreserve(adev->gfx.rlc.clear_state_obj); } if (adev->gfx.rlc.cp_table_size) { - - r = amdgpu_bo_create_reserved(adev, adev->gfx.rlc.cp_table_size, - PAGE_SIZE, AMDGPU_GEM_DOMAIN_VRAM, - &adev->gfx.rlc.cp_table_obj, - &adev->gfx.rlc.cp_table_gpu_addr, - (void **)&adev->gfx.rlc.cp_table_ptr); - if (r) { - dev_warn(adev->dev, "(%d) create RLC cp table bo failed\n", r); - gfx_v7_0_rlc_fini(adev); + r = amdgpu_gfx_rlc_init_cpt(adev); + if (r) return r; - } - - gfx_v7_0_init_cp_pg_table(adev); - - amdgpu_bo_kunmap(adev->gfx.rlc.cp_table_obj); - amdgpu_bo_unreserve(adev->gfx.rlc.cp_table_obj); - } return 0; @@ -3446,7 +3394,12 @@ static u32 gfx_v7_0_halt_rlc(struct amdgpu_device *adev) return orig; } -static void gfx_v7_0_enter_rlc_safe_mode(struct amdgpu_device *adev) +static bool gfx_v7_0_is_rlc_enabled(struct amdgpu_device *adev) +{ + return true; +} + +static void gfx_v7_0_set_safe_mode(struct amdgpu_device *adev) { u32 tmp, i, mask; @@ -3468,7 +3421,7 @@ static void gfx_v7_0_enter_rlc_safe_mode(struct amdgpu_device *adev) } } -static void gfx_v7_0_exit_rlc_safe_mode(struct amdgpu_device *adev) +static void gfx_v7_0_unset_safe_mode(struct amdgpu_device *adev) { u32 tmp; @@ -3545,13 +3498,13 @@ static int gfx_v7_0_rlc_resume(struct amdgpu_device *adev) adev->gfx.rlc_feature_version = le32_to_cpu( hdr->ucode_feature_version); - gfx_v7_0_rlc_stop(adev); + adev->gfx.rlc.funcs->stop(adev); /* disable CG */ tmp = RREG32(mmRLC_CGCG_CGLS_CTRL) & 0xfffffffc; WREG32(mmRLC_CGCG_CGLS_CTRL, tmp); - gfx_v7_0_rlc_reset(adev); + adev->gfx.rlc.funcs->reset(adev); gfx_v7_0_init_pg(adev); @@ -3582,7 +3535,7 @@ static int gfx_v7_0_rlc_resume(struct amdgpu_device *adev) if (adev->asic_type == CHIP_BONAIRE) WREG32(mmRLC_DRIVER_CPDMA_STATUS, 0); - gfx_v7_0_rlc_start(adev); + adev->gfx.rlc.funcs->start(adev); return 0; } @@ -3784,72 +3737,12 @@ static void gfx_v7_0_enable_gds_pg(struct amdgpu_device *adev, bool enable) WREG32(mmRLC_PG_CNTL, data); } -static void gfx_v7_0_init_cp_pg_table(struct amdgpu_device *adev) +static int gfx_v7_0_cp_pg_table_num(struct amdgpu_device *adev) { - const __le32 *fw_data; - volatile u32 *dst_ptr; - int me, i, max_me = 4; - u32 bo_offset = 0; - u32 table_offset, table_size; - if (adev->asic_type == CHIP_KAVERI) - max_me = 5; - - if (adev->gfx.rlc.cp_table_ptr == NULL) - return; - - /* write the cp table buffer */ - dst_ptr = adev->gfx.rlc.cp_table_ptr; - for (me = 0; me < max_me; me++) { - if (me == 0) { - const struct gfx_firmware_header_v1_0 *hdr = - (const struct gfx_firmware_header_v1_0 *)adev->gfx.ce_fw->data; - fw_data = (const __le32 *) - (adev->gfx.ce_fw->data + - le32_to_cpu(hdr->header.ucode_array_offset_bytes)); - table_offset = le32_to_cpu(hdr->jt_offset); - table_size = le32_to_cpu(hdr->jt_size); - } else if (me == 1) { - const struct gfx_firmware_header_v1_0 *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(hdr->header.ucode_array_offset_bytes)); - table_offset = le32_to_cpu(hdr->jt_offset); - table_size = le32_to_cpu(hdr->jt_size); - } else if (me == 2) { - const struct gfx_firmware_header_v1_0 *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(hdr->header.ucode_array_offset_bytes)); - table_offset = le32_to_cpu(hdr->jt_offset); - table_size = le32_to_cpu(hdr->jt_size); - } else if (me == 3) { - const struct gfx_firmware_header_v1_0 *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(hdr->header.ucode_array_offset_bytes)); - table_offset = le32_to_cpu(hdr->jt_offset); - table_size = le32_to_cpu(hdr->jt_size); - } else { - const struct gfx_firmware_header_v1_0 *hdr = - (const struct gfx_firmware_header_v1_0 *)adev->gfx.mec2_fw->data; - fw_data = (const __le32 *) - (adev->gfx.mec2_fw->data + - le32_to_cpu(hdr->header.ucode_array_offset_bytes)); - table_offset = le32_to_cpu(hdr->jt_offset); - table_size = le32_to_cpu(hdr->jt_size); - } - - for (i = 0; i < table_size; i ++) { - dst_ptr[bo_offset + i] = - cpu_to_le32(le32_to_cpu(fw_data[table_offset + i])); - } - - bo_offset += table_size; - } + return 5; + else + return 4; } static void gfx_v7_0_enable_gfx_cgpg(struct amdgpu_device *adev, @@ -4288,8 +4181,17 @@ static const struct amdgpu_gfx_funcs gfx_v7_0_gfx_funcs = { }; static const struct amdgpu_rlc_funcs gfx_v7_0_rlc_funcs = { - .enter_safe_mode = gfx_v7_0_enter_rlc_safe_mode, - .exit_safe_mode = gfx_v7_0_exit_rlc_safe_mode + .is_rlc_enabled = gfx_v7_0_is_rlc_enabled, + .set_safe_mode = gfx_v7_0_set_safe_mode, + .unset_safe_mode = gfx_v7_0_unset_safe_mode, + .init = gfx_v7_0_rlc_init, + .get_csb_size = gfx_v7_0_get_csb_size, + .get_csb_buffer = gfx_v7_0_get_csb_buffer, + .get_cp_table_num = gfx_v7_0_cp_pg_table_num, + .resume = gfx_v7_0_rlc_resume, + .stop = gfx_v7_0_rlc_stop, + .reset = gfx_v7_0_rlc_reset, + .start = gfx_v7_0_rlc_start }; static int gfx_v7_0_early_init(void *handle) @@ -4477,7 +4379,7 @@ static int gfx_v7_0_compute_ring_init(struct amdgpu_device *adev, int ring_id, ring->ring_obj = NULL; ring->use_doorbell = true; - ring->doorbell_index = AMDGPU_DOORBELL_MEC_RING0 + ring_id; + ring->doorbell_index = adev->doorbell_index.mec_ring0 + ring_id; sprintf(ring->name, "comp_%d.%d.%d", ring->me, ring->pipe, ring->queue); irq_type = AMDGPU_CP_IRQ_COMPUTE_MEC1_PIPE0_EOP @@ -4540,7 +4442,7 @@ static int gfx_v7_0_sw_init(void *handle) return r; } - r = gfx_v7_0_rlc_init(adev); + r = adev->gfx.rlc.funcs->init(adev); if (r) { DRM_ERROR("Failed to init rlc BOs!\n"); return r; @@ -4604,7 +4506,7 @@ static int gfx_v7_0_sw_fini(void *handle) amdgpu_ring_fini(&adev->gfx.compute_ring[i]); gfx_v7_0_cp_compute_fini(adev); - gfx_v7_0_rlc_fini(adev); + amdgpu_gfx_rlc_fini(adev); gfx_v7_0_mec_fini(adev); amdgpu_bo_free_kernel(&adev->gfx.rlc.clear_state_obj, &adev->gfx.rlc.clear_state_gpu_addr, @@ -4627,7 +4529,7 @@ static int gfx_v7_0_hw_init(void *handle) gfx_v7_0_constants_init(adev); /* init rlc */ - r = gfx_v7_0_rlc_resume(adev); + r = adev->gfx.rlc.funcs->resume(adev); if (r) return r; @@ -4645,7 +4547,7 @@ static int gfx_v7_0_hw_fini(void *handle) amdgpu_irq_put(adev, &adev->gfx.priv_reg_irq, 0); amdgpu_irq_put(adev, &adev->gfx.priv_inst_irq, 0); gfx_v7_0_cp_enable(adev, false); - gfx_v7_0_rlc_stop(adev); + adev->gfx.rlc.funcs->stop(adev); gfx_v7_0_fini_pg(adev); return 0; @@ -4730,7 +4632,7 @@ static int gfx_v7_0_soft_reset(void *handle) gfx_v7_0_update_cg(adev, false); /* stop the rlc */ - gfx_v7_0_rlc_stop(adev); + adev->gfx.rlc.funcs->stop(adev); /* Disable GFX parsing/prefetching */ WREG32(mmCP_ME_CNTL, CP_ME_CNTL__ME_HALT_MASK | CP_ME_CNTL__PFP_HALT_MASK | CP_ME_CNTL__CE_HALT_MASK); @@ -4959,12 +4861,36 @@ static int gfx_v7_0_eop_irq(struct amdgpu_device *adev, return 0; } +static void gfx_v7_0_fault(struct amdgpu_device *adev, + struct amdgpu_iv_entry *entry) +{ + struct amdgpu_ring *ring; + u8 me_id, pipe_id; + int i; + + me_id = (entry->ring_id & 0x0c) >> 2; + pipe_id = (entry->ring_id & 0x03) >> 0; + switch (me_id) { + case 0: + drm_sched_fault(&adev->gfx.gfx_ring[0].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)) + drm_sched_fault(&ring->sched); + } + break; + } +} + static int gfx_v7_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"); - schedule_work(&adev->reset_work); + gfx_v7_0_fault(adev, entry); return 0; } @@ -4974,7 +4900,7 @@ static int gfx_v7_0_priv_inst_irq(struct amdgpu_device *adev, { DRM_ERROR("Illegal instruction in command stream\n"); // XXX soft reset the gfx block only - schedule_work(&adev->reset_work); + gfx_v7_0_fault(adev, entry); return 0; } @@ -5090,7 +5016,7 @@ static const struct amdgpu_ring_funcs gfx_v7_0_ring_funcs_compute = { 7 + /* gfx_v7_0_ring_emit_pipeline_sync */ CIK_FLUSH_GPU_TLB_NUM_WREG * 5 + 7 + /* gfx_v7_0_ring_emit_vm_flush */ 7 + 7 + 7, /* gfx_v7_0_ring_emit_fence_compute x3 for user fence, vm fence */ - .emit_ib_size = 4, /* gfx_v7_0_ring_emit_ib_compute */ + .emit_ib_size = 7, /* gfx_v7_0_ring_emit_ib_compute */ .emit_ib = gfx_v7_0_ring_emit_ib_compute, .emit_fence = gfx_v7_0_ring_emit_fence_compute, .emit_pipeline_sync = gfx_v7_0_ring_emit_pipeline_sync, @@ -5147,6 +5073,7 @@ static void gfx_v7_0_set_gds_init(struct amdgpu_device *adev) adev->gds.mem.total_size = RREG32(mmGDS_VMID0_SIZE); adev->gds.gws.total_size = 64; adev->gds.oa.total_size = 16; + adev->gds.gds_compute_max_wave_id = RREG32(mmGDS_COMPUTE_MAX_WAVE_ID); if (adev->gds.mem.total_size == 64 * 1024) { adev->gds.mem.gfx_partition_size = 4096; diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c index 617b0c8908a3..b8e50a34bdb3 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c @@ -44,7 +44,6 @@ #include "gca/gfx_8_0_d.h" #include "gca/gfx_8_0_enum.h" #include "gca/gfx_8_0_sh_mask.h" -#include "gca/gfx_8_0_enum.h" #include "dce/dce_10_0_d.h" #include "dce/dce_10_0_sh_mask.h" @@ -54,7 +53,7 @@ #include "ivsrcid/ivsrcid_vislands30.h" #define GFX8_NUM_GFX_RINGS 1 -#define GFX8_MEC_HPD_SIZE 2048 +#define GFX8_MEC_HPD_SIZE 4096 #define TOPAZ_GB_ADDR_CONFIG_GOLDEN 0x22010001 #define CARRIZO_GB_ADDR_CONFIG_GOLDEN 0x22010001 @@ -839,18 +838,14 @@ static int gfx_v8_0_ring_test_ring(struct amdgpu_ring *ring) int r; r = amdgpu_gfx_scratch_get(adev, &scratch); - if (r) { - DRM_ERROR("amdgpu: cp failed to get scratch reg (%d).\n", r); + if (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; - } + if (r) + goto error_free_scratch; + 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); @@ -862,14 +857,11 @@ static int gfx_v8_0_ring_test_ring(struct amdgpu_ring *ring) break; DRM_UDELAY(1); } - if (i < adev->usec_timeout) { - DRM_DEBUG("ring test on %d succeeded in %d usecs\n", - ring->idx, i); - } else { - DRM_ERROR("amdgpu: ring %d test failed (scratch(0x%04X)=0x%08X)\n", - ring->idx, scratch, tmp); - r = -EINVAL; - } + + if (i >= adev->usec_timeout) + r = -ETIMEDOUT; + +error_free_scratch: amdgpu_gfx_scratch_free(adev, scratch); return r; } @@ -886,19 +878,16 @@ static int gfx_v8_0_ring_test_ib(struct amdgpu_ring *ring, long timeout) long r; r = amdgpu_device_wb_get(adev, &index); - if (r) { - dev_err(adev->dev, "(%ld) failed to allocate wb slot\n", r); + 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, &ib); - if (r) { - DRM_ERROR("amdgpu: failed to get ib (%ld).\n", r); + if (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); @@ -912,22 +901,17 @@ static int gfx_v8_0_ring_test_ib(struct amdgpu_ring *ring, long timeout) r = dma_fence_wait_timeout(f, false, timeout); if (r == 0) { - DRM_ERROR("amdgpu: IB test timed out.\n"); r = -ETIMEDOUT; goto err2; } else if (r < 0) { - DRM_ERROR("amdgpu: fence wait failed (%ld).\n", r); goto err2; } tmp = adev->wb.wb[index]; - if (tmp == 0xDEADBEEF) { - DRM_DEBUG("ib test on ring %d succeeded\n", ring->idx); + if (tmp == 0xDEADBEEF) r = 0; - } else { - DRM_ERROR("ib test on ring %d failed\n", ring->idx); + else r = -EINVAL; - } err2: amdgpu_ib_free(adev, &ib, NULL); @@ -1298,81 +1282,16 @@ static void gfx_v8_0_get_csb_buffer(struct amdgpu_device *adev, buffer[count++] = cpu_to_le32(0); } -static void cz_init_cp_jump_table(struct amdgpu_device *adev) +static int gfx_v8_0_cp_jump_table_num(struct amdgpu_device *adev) { - const __le32 *fw_data; - volatile u32 *dst_ptr; - int me, i, max_me = 4; - u32 bo_offset = 0; - u32 table_offset, table_size; - if (adev->asic_type == CHIP_CARRIZO) - max_me = 5; - - /* write the cp table buffer */ - dst_ptr = adev->gfx.rlc.cp_table_ptr; - for (me = 0; me < max_me; me++) { - if (me == 0) { - const struct gfx_firmware_header_v1_0 *hdr = - (const struct gfx_firmware_header_v1_0 *)adev->gfx.ce_fw->data; - fw_data = (const __le32 *) - (adev->gfx.ce_fw->data + - le32_to_cpu(hdr->header.ucode_array_offset_bytes)); - table_offset = le32_to_cpu(hdr->jt_offset); - table_size = le32_to_cpu(hdr->jt_size); - } else if (me == 1) { - const struct gfx_firmware_header_v1_0 *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(hdr->header.ucode_array_offset_bytes)); - table_offset = le32_to_cpu(hdr->jt_offset); - table_size = le32_to_cpu(hdr->jt_size); - } else if (me == 2) { - const struct gfx_firmware_header_v1_0 *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(hdr->header.ucode_array_offset_bytes)); - table_offset = le32_to_cpu(hdr->jt_offset); - table_size = le32_to_cpu(hdr->jt_size); - } else if (me == 3) { - const struct gfx_firmware_header_v1_0 *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(hdr->header.ucode_array_offset_bytes)); - table_offset = le32_to_cpu(hdr->jt_offset); - table_size = le32_to_cpu(hdr->jt_size); - } else if (me == 4) { - const struct gfx_firmware_header_v1_0 *hdr = - (const struct gfx_firmware_header_v1_0 *)adev->gfx.mec2_fw->data; - fw_data = (const __le32 *) - (adev->gfx.mec2_fw->data + - le32_to_cpu(hdr->header.ucode_array_offset_bytes)); - table_offset = le32_to_cpu(hdr->jt_offset); - table_size = le32_to_cpu(hdr->jt_size); - } - - for (i = 0; i < table_size; i ++) { - dst_ptr[bo_offset + i] = - cpu_to_le32(le32_to_cpu(fw_data[table_offset + i])); - } - - bo_offset += table_size; - } -} - -static void gfx_v8_0_rlc_fini(struct amdgpu_device *adev) -{ - amdgpu_bo_free_kernel(&adev->gfx.rlc.clear_state_obj, NULL, NULL); - amdgpu_bo_free_kernel(&adev->gfx.rlc.cp_table_obj, NULL, NULL); + return 5; + else + return 4; } static int gfx_v8_0_rlc_init(struct amdgpu_device *adev) { - volatile u32 *dst_ptr; - u32 dws; const struct cs_section_def *cs_data; int r; @@ -1381,44 +1300,18 @@ static int gfx_v8_0_rlc_init(struct amdgpu_device *adev) cs_data = adev->gfx.rlc.cs_data; if (cs_data) { - /* clear state block */ - adev->gfx.rlc.clear_state_size = dws = gfx_v8_0_get_csb_size(adev); - - r = amdgpu_bo_create_reserved(adev, dws * 4, PAGE_SIZE, - AMDGPU_GEM_DOMAIN_VRAM, - &adev->gfx.rlc.clear_state_obj, - &adev->gfx.rlc.clear_state_gpu_addr, - (void **)&adev->gfx.rlc.cs_ptr); - if (r) { - dev_warn(adev->dev, "(%d) create RLC c bo failed\n", r); - gfx_v8_0_rlc_fini(adev); + /* init clear state block */ + r = amdgpu_gfx_rlc_init_csb(adev); + if (r) return r; - } - - /* set up the cs buffer */ - dst_ptr = adev->gfx.rlc.cs_ptr; - gfx_v8_0_get_csb_buffer(adev, dst_ptr); - amdgpu_bo_kunmap(adev->gfx.rlc.clear_state_obj); - amdgpu_bo_unreserve(adev->gfx.rlc.clear_state_obj); } if ((adev->asic_type == CHIP_CARRIZO) || (adev->asic_type == CHIP_STONEY)) { adev->gfx.rlc.cp_table_size = ALIGN(96 * 5 * 4, 2048) + (64 * 1024); /* JT + GDS */ - r = amdgpu_bo_create_reserved(adev, adev->gfx.rlc.cp_table_size, - PAGE_SIZE, AMDGPU_GEM_DOMAIN_VRAM, - &adev->gfx.rlc.cp_table_obj, - &adev->gfx.rlc.cp_table_gpu_addr, - (void **)&adev->gfx.rlc.cp_table_ptr); - if (r) { - dev_warn(adev->dev, "(%d) create RLC cp table bo failed\n", r); + r = amdgpu_gfx_rlc_init_cpt(adev); + if (r) return r; - } - - cz_init_cp_jump_table(adev); - - amdgpu_bo_kunmap(adev->gfx.rlc.cp_table_obj); - amdgpu_bo_unreserve(adev->gfx.rlc.cp_table_obj); } return 0; @@ -1443,7 +1336,7 @@ static int gfx_v8_0_mec_init(struct amdgpu_device *adev) mec_hpd_size = adev->gfx.num_compute_rings * GFX8_MEC_HPD_SIZE; r = amdgpu_bo_create_reserved(adev, mec_hpd_size, PAGE_SIZE, - AMDGPU_GEM_DOMAIN_GTT, + AMDGPU_GEM_DOMAIN_VRAM, &adev->gfx.mec.hpd_eop_obj, &adev->gfx.mec.hpd_eop_gpu_addr, (void **)&hpd); @@ -1629,7 +1522,7 @@ static int gfx_v8_0_do_edc_gpr_workarounds(struct amdgpu_device *adev) return 0; /* bail if the compute ring is not ready */ - if (!ring->ready) + if (!ring->sched.ready) return 0; tmp = RREG32(mmGB_EDC_MODE); @@ -1997,7 +1890,7 @@ static int gfx_v8_0_compute_ring_init(struct amdgpu_device *adev, int ring_id, ring->ring_obj = NULL; ring->use_doorbell = true; - ring->doorbell_index = AMDGPU_DOORBELL_MEC_RING0 + ring_id; + ring->doorbell_index = adev->doorbell_index.mec_ring0 + ring_id; ring->eop_gpu_addr = adev->gfx.mec.hpd_eop_gpu_addr + (ring_id * GFX8_MEC_HPD_SIZE); sprintf(ring->name, "comp_%d.%d.%d", ring->me, ring->pipe, ring->queue); @@ -2088,7 +1981,7 @@ static int gfx_v8_0_sw_init(void *handle) return r; } - r = gfx_v8_0_rlc_init(adev); + r = adev->gfx.rlc.funcs->init(adev); if (r) { DRM_ERROR("Failed to init rlc BOs!\n"); return r; @@ -2108,7 +2001,7 @@ static int gfx_v8_0_sw_init(void *handle) /* no gfx doorbells on iceland */ if (adev->asic_type != CHIP_TOPAZ) { ring->use_doorbell = true; - ring->doorbell_index = AMDGPU_DOORBELL_GFX_RING0; + ring->doorbell_index = adev->doorbell_index.gfx_ring0; } r = amdgpu_ring_init(adev, ring, 1024, &adev->gfx.eop_irq, @@ -2181,7 +2074,7 @@ static int gfx_v8_0_sw_fini(void *handle) amdgpu_gfx_kiq_fini(adev); gfx_v8_0_mec_fini(adev); - gfx_v8_0_rlc_fini(adev); + amdgpu_gfx_rlc_fini(adev); amdgpu_bo_free_kernel(&adev->gfx.rlc.clear_state_obj, &adev->gfx.rlc.clear_state_gpu_addr, (void **)&adev->gfx.rlc.cs_ptr); @@ -4175,10 +4068,15 @@ static void gfx_v8_0_rlc_start(struct amdgpu_device *adev) static int gfx_v8_0_rlc_resume(struct amdgpu_device *adev) { - gfx_v8_0_rlc_stop(adev); - gfx_v8_0_rlc_reset(adev); + if (amdgpu_sriov_vf(adev)) { + gfx_v8_0_init_csb(adev); + return 0; + } + + adev->gfx.rlc.funcs->stop(adev); + adev->gfx.rlc.funcs->reset(adev); gfx_v8_0_init_pg(adev); - gfx_v8_0_rlc_start(adev); + adev->gfx.rlc.funcs->start(adev); return 0; } @@ -4197,7 +4095,7 @@ static void gfx_v8_0_cp_gfx_enable(struct amdgpu_device *adev, bool enable) tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, PFP_HALT, 1); tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, CE_HALT, 1); for (i = 0; i < adev->gfx.num_gfx_rings; i++) - adev->gfx.gfx_ring[i].ready = false; + adev->gfx.gfx_ring[i].sched.ready = false; } WREG32(mmCP_ME_CNTL, tmp); udelay(50); @@ -4322,7 +4220,7 @@ static void gfx_v8_0_set_cpg_door_bell(struct amdgpu_device *adev, struct amdgpu tmp = REG_SET_FIELD(0, CP_RB_DOORBELL_RANGE_LOWER, DOORBELL_RANGE_LOWER, - AMDGPU_DOORBELL_GFX_RING0); + adev->doorbell_index.gfx_ring0); WREG32(mmCP_RB_DOORBELL_RANGE_LOWER, tmp); WREG32(mmCP_RB_DOORBELL_RANGE_UPPER, @@ -4335,7 +4233,6 @@ static int gfx_v8_0_cp_gfx_resume(struct amdgpu_device *adev) u32 tmp; u32 rb_bufsz; u64 rb_addr, rptr_addr, wptr_gpu_addr; - int r; /* Set the write pointer delay */ WREG32(mmCP_RB_WPTR_DELAY, 0); @@ -4379,12 +4276,9 @@ static int gfx_v8_0_cp_gfx_resume(struct amdgpu_device *adev) /* start the ring */ amdgpu_ring_clear_ring(ring); gfx_v8_0_cp_gfx_start(adev); - ring->ready = true; - r = amdgpu_ring_test_ring(ring); - if (r) - ring->ready = false; + ring->sched.ready = true; - return r; + return 0; } static void gfx_v8_0_cp_compute_enable(struct amdgpu_device *adev, bool enable) @@ -4396,8 +4290,8 @@ static void gfx_v8_0_cp_compute_enable(struct amdgpu_device *adev, bool enable) } else { WREG32(mmCP_MEC_CNTL, (CP_MEC_CNTL__MEC_ME1_HALT_MASK | CP_MEC_CNTL__MEC_ME2_HALT_MASK)); for (i = 0; i < adev->gfx.num_compute_rings; i++) - adev->gfx.compute_ring[i].ready = false; - adev->gfx.kiq.ring.ready = false; + adev->gfx.compute_ring[i].sched.ready = false; + adev->gfx.kiq.ring.sched.ready = false; } udelay(50); } @@ -4473,12 +4367,9 @@ static int gfx_v8_0_kiq_kcq_enable(struct amdgpu_device *adev) amdgpu_ring_write(kiq_ring, upper_32_bits(wptr_addr)); } - r = amdgpu_ring_test_ring(kiq_ring); - if (r) { - DRM_ERROR("KCQ enable failed\n"); - kiq_ring->ready = false; - } - return r; + amdgpu_ring_commit(kiq_ring); + + return 0; } static int gfx_v8_0_deactivate_hqd(struct amdgpu_device *adev, u32 req) @@ -4755,8 +4646,8 @@ static int gfx_v8_0_kcq_init_queue(struct amdgpu_ring *ring) static void gfx_v8_0_set_mec_doorbell_range(struct amdgpu_device *adev) { if (adev->asic_type > CHIP_TONGA) { - WREG32(mmCP_MEC_DOORBELL_RANGE_LOWER, AMDGPU_DOORBELL_KIQ << 2); - WREG32(mmCP_MEC_DOORBELL_RANGE_UPPER, AMDGPU_DOORBELL_MEC_RING7 << 2); + WREG32(mmCP_MEC_DOORBELL_RANGE_LOWER, adev->doorbell_index.kiq << 2); + WREG32(mmCP_MEC_DOORBELL_RANGE_UPPER, adev->doorbell_index.mec_ring7 << 2); } /* enable doorbells */ WREG32_FIELD(CP_PQ_STATUS, DOORBELL_ENABLE, 1); @@ -4781,7 +4672,7 @@ static int gfx_v8_0_kiq_resume(struct amdgpu_device *adev) amdgpu_bo_kunmap(ring->mqd_obj); ring->mqd_ptr = NULL; amdgpu_bo_unreserve(ring->mqd_obj); - ring->ready = true; + ring->sched.ready = true; return 0; } @@ -4815,19 +4706,32 @@ static int gfx_v8_0_kcq_resume(struct amdgpu_device *adev) if (r) goto done; - /* Test KCQs - reversing the order of rings seems to fix ring test failure - * after GPU reset - */ - for (i = adev->gfx.num_compute_rings - 1; i >= 0; i--) { +done: + return r; +} + +static int gfx_v8_0_cp_test_all_rings(struct amdgpu_device *adev) +{ + int r, i; + struct amdgpu_ring *ring; + + /* collect all the ring_tests here, gfx, kiq, compute */ + ring = &adev->gfx.gfx_ring[0]; + r = amdgpu_ring_test_helper(ring); + if (r) + return r; + + ring = &adev->gfx.kiq.ring; + 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]; - ring->ready = true; - r = amdgpu_ring_test_ring(ring); - if (r) - ring->ready = false; + amdgpu_ring_test_helper(ring); } -done: - return r; + return 0; } static int gfx_v8_0_cp_resume(struct amdgpu_device *adev) @@ -4848,6 +4752,11 @@ static int gfx_v8_0_cp_resume(struct amdgpu_device *adev) r = gfx_v8_0_kcq_resume(adev); if (r) return r; + + r = gfx_v8_0_cp_test_all_rings(adev); + if (r) + return r; + gfx_v8_0_enable_gui_idle_interrupt(adev, true); return 0; @@ -4867,7 +4776,7 @@ static int gfx_v8_0_hw_init(void *handle) gfx_v8_0_init_golden_registers(adev); gfx_v8_0_constants_init(adev); - r = gfx_v8_0_rlc_resume(adev); + r = adev->gfx.rlc.funcs->resume(adev); if (r) return r; @@ -4899,7 +4808,7 @@ static int gfx_v8_0_kcq_disable(struct amdgpu_device *adev) amdgpu_ring_write(kiq_ring, 0); amdgpu_ring_write(kiq_ring, 0); } - r = amdgpu_ring_test_ring(kiq_ring); + r = amdgpu_ring_test_helper(kiq_ring); if (r) DRM_ERROR("KCQ disable failed\n"); @@ -4973,16 +4882,16 @@ static int gfx_v8_0_hw_fini(void *handle) pr_debug("For SRIOV client, shouldn't do anything.\n"); return 0; } - adev->gfx.rlc.funcs->enter_safe_mode(adev); + amdgpu_gfx_rlc_enter_safe_mode(adev); if (!gfx_v8_0_wait_for_idle(adev)) gfx_v8_0_cp_enable(adev, false); else pr_err("cp is busy, skip halt cp\n"); if (!gfx_v8_0_wait_for_rlc_idle(adev)) - gfx_v8_0_rlc_stop(adev); + adev->gfx.rlc.funcs->stop(adev); else pr_err("rlc is busy, skip halt rlc\n"); - adev->gfx.rlc.funcs->exit_safe_mode(adev); + amdgpu_gfx_rlc_exit_safe_mode(adev); return 0; } @@ -5061,17 +4970,16 @@ static bool gfx_v8_0_check_soft_reset(void *handle) static int gfx_v8_0_pre_soft_reset(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; - u32 grbm_soft_reset = 0, srbm_soft_reset = 0; + u32 grbm_soft_reset = 0; if ((!adev->gfx.grbm_soft_reset) && (!adev->gfx.srbm_soft_reset)) return 0; grbm_soft_reset = adev->gfx.grbm_soft_reset; - srbm_soft_reset = adev->gfx.srbm_soft_reset; /* stop the rlc */ - gfx_v8_0_rlc_stop(adev); + adev->gfx.rlc.funcs->stop(adev); if (REG_GET_FIELD(grbm_soft_reset, GRBM_SOFT_RESET, SOFT_RESET_CP) || REG_GET_FIELD(grbm_soft_reset, GRBM_SOFT_RESET, SOFT_RESET_GFX)) @@ -5165,14 +5073,13 @@ static int gfx_v8_0_soft_reset(void *handle) static int gfx_v8_0_post_soft_reset(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; - u32 grbm_soft_reset = 0, srbm_soft_reset = 0; + u32 grbm_soft_reset = 0; if ((!adev->gfx.grbm_soft_reset) && (!adev->gfx.srbm_soft_reset)) return 0; grbm_soft_reset = adev->gfx.grbm_soft_reset; - srbm_soft_reset = adev->gfx.srbm_soft_reset; if (REG_GET_FIELD(grbm_soft_reset, GRBM_SOFT_RESET, SOFT_RESET_CP) || REG_GET_FIELD(grbm_soft_reset, GRBM_SOFT_RESET, SOFT_RESET_CPF) || @@ -5197,7 +5104,9 @@ static int gfx_v8_0_post_soft_reset(void *handle) REG_GET_FIELD(grbm_soft_reset, GRBM_SOFT_RESET, SOFT_RESET_GFX)) gfx_v8_0_cp_gfx_resume(adev); - gfx_v8_0_rlc_start(adev); + gfx_v8_0_cp_test_all_rings(adev); + + adev->gfx.rlc.funcs->start(adev); return 0; } @@ -5445,7 +5354,7 @@ static int gfx_v8_0_set_powergating_state(void *handle, AMD_PG_SUPPORT_RLC_SMU_HS | AMD_PG_SUPPORT_CP | AMD_PG_SUPPORT_GFX_DMG)) - adev->gfx.rlc.funcs->enter_safe_mode(adev); + amdgpu_gfx_rlc_enter_safe_mode(adev); switch (adev->asic_type) { case CHIP_CARRIZO: case CHIP_STONEY: @@ -5499,7 +5408,7 @@ static int gfx_v8_0_set_powergating_state(void *handle, AMD_PG_SUPPORT_RLC_SMU_HS | AMD_PG_SUPPORT_CP | AMD_PG_SUPPORT_GFX_DMG)) - adev->gfx.rlc.funcs->exit_safe_mode(adev); + amdgpu_gfx_rlc_exit_safe_mode(adev); return 0; } @@ -5593,57 +5502,53 @@ static void gfx_v8_0_send_serdes_cmd(struct amdgpu_device *adev, #define RLC_GPR_REG2__MESSAGE__SHIFT 0x00000001 #define RLC_GPR_REG2__MESSAGE_MASK 0x0000001e -static void iceland_enter_rlc_safe_mode(struct amdgpu_device *adev) +static bool gfx_v8_0_is_rlc_enabled(struct amdgpu_device *adev) { - u32 data; - unsigned i; + uint32_t rlc_setting; - data = RREG32(mmRLC_CNTL); - if (!(data & RLC_CNTL__RLC_ENABLE_F32_MASK)) - return; + rlc_setting = RREG32(mmRLC_CNTL); + if (!(rlc_setting & RLC_CNTL__RLC_ENABLE_F32_MASK)) + return false; - if (adev->cg_flags & (AMD_CG_SUPPORT_GFX_CGCG | AMD_CG_SUPPORT_GFX_MGCG)) { - data |= RLC_SAFE_MODE__CMD_MASK; - data &= ~RLC_SAFE_MODE__MESSAGE_MASK; - data |= (1 << RLC_SAFE_MODE__MESSAGE__SHIFT); - WREG32(mmRLC_SAFE_MODE, data); + return true; +} - for (i = 0; i < adev->usec_timeout; i++) { - if ((RREG32(mmRLC_GPM_STAT) & - (RLC_GPM_STAT__GFX_CLOCK_STATUS_MASK | - RLC_GPM_STAT__GFX_POWER_STATUS_MASK)) == - (RLC_GPM_STAT__GFX_CLOCK_STATUS_MASK | - RLC_GPM_STAT__GFX_POWER_STATUS_MASK)) - break; - udelay(1); - } +static void gfx_v8_0_set_safe_mode(struct amdgpu_device *adev) +{ + uint32_t data; + unsigned i; + data = RREG32(mmRLC_CNTL); + data |= RLC_SAFE_MODE__CMD_MASK; + data &= ~RLC_SAFE_MODE__MESSAGE_MASK; + data |= (1 << RLC_SAFE_MODE__MESSAGE__SHIFT); + WREG32(mmRLC_SAFE_MODE, data); - for (i = 0; i < adev->usec_timeout; i++) { - if (!REG_GET_FIELD(RREG32(mmRLC_SAFE_MODE), RLC_SAFE_MODE, CMD)) - break; - udelay(1); - } - adev->gfx.rlc.in_safe_mode = true; + /* wait for RLC_SAFE_MODE */ + for (i = 0; i < adev->usec_timeout; i++) { + if ((RREG32(mmRLC_GPM_STAT) & + (RLC_GPM_STAT__GFX_CLOCK_STATUS_MASK | + RLC_GPM_STAT__GFX_POWER_STATUS_MASK)) == + (RLC_GPM_STAT__GFX_CLOCK_STATUS_MASK | + RLC_GPM_STAT__GFX_POWER_STATUS_MASK)) + break; + udelay(1); + } + for (i = 0; i < adev->usec_timeout; i++) { + if (!REG_GET_FIELD(RREG32(mmRLC_SAFE_MODE), RLC_SAFE_MODE, CMD)) + break; + udelay(1); } } -static void iceland_exit_rlc_safe_mode(struct amdgpu_device *adev) +static void gfx_v8_0_unset_safe_mode(struct amdgpu_device *adev) { - u32 data = 0; + uint32_t data; unsigned i; data = RREG32(mmRLC_CNTL); - if (!(data & RLC_CNTL__RLC_ENABLE_F32_MASK)) - return; - - if (adev->cg_flags & (AMD_CG_SUPPORT_GFX_CGCG | AMD_CG_SUPPORT_GFX_MGCG)) { - if (adev->gfx.rlc.in_safe_mode) { - data |= RLC_SAFE_MODE__CMD_MASK; - data &= ~RLC_SAFE_MODE__MESSAGE_MASK; - WREG32(mmRLC_SAFE_MODE, data); - adev->gfx.rlc.in_safe_mode = false; - } - } + data |= RLC_SAFE_MODE__CMD_MASK; + data &= ~RLC_SAFE_MODE__MESSAGE_MASK; + WREG32(mmRLC_SAFE_MODE, data); for (i = 0; i < adev->usec_timeout; i++) { if (!REG_GET_FIELD(RREG32(mmRLC_SAFE_MODE), RLC_SAFE_MODE, CMD)) @@ -5653,8 +5558,17 @@ static void iceland_exit_rlc_safe_mode(struct amdgpu_device *adev) } static const struct amdgpu_rlc_funcs iceland_rlc_funcs = { - .enter_safe_mode = iceland_enter_rlc_safe_mode, - .exit_safe_mode = iceland_exit_rlc_safe_mode + .is_rlc_enabled = gfx_v8_0_is_rlc_enabled, + .set_safe_mode = gfx_v8_0_set_safe_mode, + .unset_safe_mode = gfx_v8_0_unset_safe_mode, + .init = gfx_v8_0_rlc_init, + .get_csb_size = gfx_v8_0_get_csb_size, + .get_csb_buffer = gfx_v8_0_get_csb_buffer, + .get_cp_table_num = gfx_v8_0_cp_jump_table_num, + .resume = gfx_v8_0_rlc_resume, + .stop = gfx_v8_0_rlc_stop, + .reset = gfx_v8_0_rlc_reset, + .start = gfx_v8_0_rlc_start }; static void gfx_v8_0_update_medium_grain_clock_gating(struct amdgpu_device *adev, @@ -5662,7 +5576,7 @@ static void gfx_v8_0_update_medium_grain_clock_gating(struct amdgpu_device *adev { uint32_t temp, data; - adev->gfx.rlc.funcs->enter_safe_mode(adev); + amdgpu_gfx_rlc_enter_safe_mode(adev); /* It is disabled by HW by default */ if (enable && (adev->cg_flags & AMD_CG_SUPPORT_GFX_MGCG)) { @@ -5758,7 +5672,7 @@ static void gfx_v8_0_update_medium_grain_clock_gating(struct amdgpu_device *adev gfx_v8_0_wait_for_rlc_serdes(adev); } - adev->gfx.rlc.funcs->exit_safe_mode(adev); + amdgpu_gfx_rlc_exit_safe_mode(adev); } static void gfx_v8_0_update_coarse_grain_clock_gating(struct amdgpu_device *adev, @@ -5768,7 +5682,7 @@ static void gfx_v8_0_update_coarse_grain_clock_gating(struct amdgpu_device *adev temp = data = RREG32(mmRLC_CGCG_CGLS_CTRL); - adev->gfx.rlc.funcs->enter_safe_mode(adev); + amdgpu_gfx_rlc_enter_safe_mode(adev); if (enable && (adev->cg_flags & AMD_CG_SUPPORT_GFX_CGCG)) { temp1 = data1 = RREG32(mmRLC_CGTT_MGCG_OVERRIDE); @@ -5851,7 +5765,7 @@ static void gfx_v8_0_update_coarse_grain_clock_gating(struct amdgpu_device *adev gfx_v8_0_wait_for_rlc_serdes(adev); - adev->gfx.rlc.funcs->exit_safe_mode(adev); + amdgpu_gfx_rlc_exit_safe_mode(adev); } static int gfx_v8_0_update_gfx_clock_gating(struct amdgpu_device *adev, bool enable) @@ -6131,9 +6045,11 @@ static void gfx_v8_0_ring_emit_vgt_flush(struct amdgpu_ring *ring) } static void gfx_v8_0_ring_emit_ib_gfx(struct amdgpu_ring *ring, - struct amdgpu_ib *ib, - unsigned vmid, bool ctx_switch) + struct amdgpu_job *job, + struct amdgpu_ib *ib, + uint32_t flags) { + unsigned vmid = AMDGPU_JOB_GET_VMID(job); u32 header, control = 0; if (ib->flags & AMDGPU_IB_FLAG_CE) @@ -6161,11 +6077,29 @@ static void gfx_v8_0_ring_emit_ib_gfx(struct amdgpu_ring *ring, } static void gfx_v8_0_ring_emit_ib_compute(struct amdgpu_ring *ring, + struct amdgpu_job *job, struct amdgpu_ib *ib, - unsigned vmid, bool ctx_switch) + uint32_t flags) { + unsigned vmid = AMDGPU_JOB_GET_VMID(job); u32 control = INDIRECT_BUFFER_VALID | ib->length_dw | (vmid << 24); + /* 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, mmGDS_COMPUTE_MAX_WAVE_ID - PACKET3_SET_CONFIG_REG_START); + amdgpu_ring_write(ring, ring->adev->gds.gds_compute_max_wave_id); + } + amdgpu_ring_write(ring, PACKET3(PACKET3_INDIRECT_BUFFER, 2)); amdgpu_ring_write(ring, #ifdef __BIG_ENDIAN @@ -6738,12 +6672,39 @@ static int gfx_v8_0_eop_irq(struct amdgpu_device *adev, return 0; } +static void gfx_v8_0_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: + drm_sched_fault(&adev->gfx.gfx_ring[0].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; + } +} + static int gfx_v8_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"); - schedule_work(&adev->reset_work); + gfx_v8_0_fault(adev, entry); return 0; } @@ -6752,7 +6713,7 @@ static int gfx_v8_0_priv_inst_irq(struct amdgpu_device *adev, struct amdgpu_iv_entry *entry) { DRM_ERROR("Illegal instruction in command stream\n"); - schedule_work(&adev->reset_work); + gfx_v8_0_fault(adev, entry); return 0; } @@ -6945,7 +6906,7 @@ static const struct amdgpu_ring_funcs gfx_v8_0_ring_funcs_compute = { 7 + /* gfx_v8_0_ring_emit_pipeline_sync */ VI_FLUSH_GPU_TLB_NUM_WREG * 5 + 7 + /* gfx_v8_0_ring_emit_vm_flush */ 7 + 7 + 7, /* gfx_v8_0_ring_emit_fence_compute x3 for user fence, vm fence */ - .emit_ib_size = 4, /* gfx_v8_0_ring_emit_ib_compute */ + .emit_ib_size = 7, /* gfx_v8_0_ring_emit_ib_compute */ .emit_ib = gfx_v8_0_ring_emit_ib_compute, .emit_fence = gfx_v8_0_ring_emit_fence_compute, .emit_pipeline_sync = gfx_v8_0_ring_emit_pipeline_sync, @@ -6975,11 +6936,9 @@ static const struct amdgpu_ring_funcs gfx_v8_0_ring_funcs_kiq = { 7 + /* gfx_v8_0_ring_emit_pipeline_sync */ 17 + /* gfx_v8_0_ring_emit_vm_flush */ 7 + 7 + 7, /* gfx_v8_0_ring_emit_fence_kiq x3 for user fence, vm fence */ - .emit_ib_size = 4, /* gfx_v8_0_ring_emit_ib_compute */ - .emit_ib = gfx_v8_0_ring_emit_ib_compute, + .emit_ib_size = 7, /* gfx_v8_0_ring_emit_ib_compute */ .emit_fence = gfx_v8_0_ring_emit_fence_kiq, .test_ring = gfx_v8_0_ring_test_ring, - .test_ib = gfx_v8_0_ring_test_ib, .insert_nop = amdgpu_ring_insert_nop, .pad_ib = amdgpu_ring_generic_pad_ib, .emit_rreg = gfx_v8_0_ring_emit_rreg, @@ -7053,6 +7012,7 @@ static void gfx_v8_0_set_gds_init(struct amdgpu_device *adev) adev->gds.mem.total_size = RREG32(mmGDS_VMID0_SIZE); adev->gds.gws.total_size = 64; adev->gds.oa.total_size = 16; + adev->gds.gds_compute_max_wave_id = RREG32(mmGDS_COMPUTE_MAX_WAVE_ID); if (adev->gds.mem.total_size == 64 * 1024) { adev->gds.mem.gfx_partition_size = 4096; diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c index 21363b2b2ee5..a11db2b1a63f 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c @@ -41,7 +41,7 @@ #include "ivsrcid/gfx/irqsrcs_gfx_9_0.h" #define GFX9_NUM_GFX_RINGS 1 -#define GFX9_MEC_HPD_SIZE 2048 +#define GFX9_MEC_HPD_SIZE 4096 #define RLCG_UCODE_LOADING_START_ADDRESS 0x00002000L #define RLC_SAVE_RESTORE_ADDR_STARTING_OFFSET 0x00000000L @@ -86,6 +86,7 @@ MODULE_FIRMWARE("amdgpu/picasso_me.bin"); MODULE_FIRMWARE("amdgpu/picasso_mec.bin"); MODULE_FIRMWARE("amdgpu/picasso_mec2.bin"); MODULE_FIRMWARE("amdgpu/picasso_rlc.bin"); +MODULE_FIRMWARE("amdgpu/picasso_rlc_am4.bin"); MODULE_FIRMWARE("amdgpu/raven2_ce.bin"); MODULE_FIRMWARE("amdgpu/raven2_pfp.bin"); @@ -112,7 +113,10 @@ static const struct soc15_reg_golden golden_settings_gc_9_0[] = SOC15_REG_GOLDEN_VALUE(GC, 0, mmTCP_CHAN_STEER_HI, 0xffffffff, 0x4a2c0e68), SOC15_REG_GOLDEN_VALUE(GC, 0, mmTCP_CHAN_STEER_LO, 0xffffffff, 0xb5d3f197), SOC15_REG_GOLDEN_VALUE(GC, 0, mmVGT_CACHE_INVALIDATION, 0x3fff3af3, 0x19200000), - SOC15_REG_GOLDEN_VALUE(GC, 0, mmVGT_GS_MAX_WAVE_ID, 0x00000fff, 0x000003ff) + SOC15_REG_GOLDEN_VALUE(GC, 0, mmVGT_GS_MAX_WAVE_ID, 0x00000fff, 0x000003ff), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmCP_MEC1_F32_INT_DIS, 0x00000000, 0x00000800), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmCP_MEC2_F32_INT_DIS, 0x00000000, 0x00000800), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmCP_DEBUG, 0x00000000, 0x00008000) }; static const struct soc15_reg_golden golden_settings_gc_9_0_vg10[] = @@ -134,10 +138,7 @@ static const struct soc15_reg_golden golden_settings_gc_9_0_vg10[] = SOC15_REG_GOLDEN_VALUE(GC, 0, mmRMI_UTCL1_CNTL2, 0x00030000, 0x00020000), SOC15_REG_GOLDEN_VALUE(GC, 0, mmSPI_CONFIG_CNTL_1, 0x0000000f, 0x01000107), SOC15_REG_GOLDEN_VALUE(GC, 0, mmTD_CNTL, 0x00001800, 0x00000800), - SOC15_REG_GOLDEN_VALUE(GC, 0, mmWD_UTCL1_CNTL, 0x08000000, 0x08000080), - SOC15_REG_GOLDEN_VALUE(GC, 0, mmCP_MEC1_F32_INT_DIS, 0x00000000, 0x00000800), - SOC15_REG_GOLDEN_VALUE(GC, 0, mmCP_MEC2_F32_INT_DIS, 0x00000000, 0x00000800), - SOC15_REG_GOLDEN_VALUE(GC, 0, mmCP_DEBUG, 0x00000000, 0x00008000) + SOC15_REG_GOLDEN_VALUE(GC, 0, mmWD_UTCL1_CNTL, 0x08000000, 0x08000080) }; static const struct soc15_reg_golden golden_settings_gc_9_0_vg20[] = @@ -219,6 +220,7 @@ static const struct soc15_reg_golden golden_settings_gc_9_1_rv2[] = static const struct soc15_reg_golden golden_settings_gc_9_x_common[] = { + SOC15_REG_GOLDEN_VALUE(GC, 0, mmCP_SD_CNTL, 0xffffffff, 0x000001ff), SOC15_REG_GOLDEN_VALUE(GC, 0, mmGRBM_CAM_INDEX, 0xffffffff, 0x00000000), SOC15_REG_GOLDEN_VALUE(GC, 0, mmGRBM_CAM_DATA, 0xffffffff, 0x2544c382) }; @@ -396,18 +398,14 @@ static int gfx_v9_0_ring_test_ring(struct amdgpu_ring *ring) int r; r = amdgpu_gfx_scratch_get(adev, &scratch); - if (r) { - DRM_ERROR("amdgpu: cp failed to get scratch reg (%d).\n", r); + if (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; - } + if (r) + goto error_free_scratch; + 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); @@ -419,14 +417,11 @@ static int gfx_v9_0_ring_test_ring(struct amdgpu_ring *ring) break; DRM_UDELAY(1); } - if (i < adev->usec_timeout) { - DRM_DEBUG("ring test on %d succeeded in %d usecs\n", - ring->idx, i); - } else { - DRM_ERROR("amdgpu: ring %d test failed (scratch(0x%04X)=0x%08X)\n", - ring->idx, scratch, tmp); - r = -EINVAL; - } + + if (i >= adev->usec_timeout) + r = -ETIMEDOUT; + +error_free_scratch: amdgpu_gfx_scratch_free(adev, scratch); return r; } @@ -443,19 +438,16 @@ static int gfx_v9_0_ring_test_ib(struct amdgpu_ring *ring, long timeout) long r; r = amdgpu_device_wb_get(adev, &index); - if (r) { - dev_err(adev->dev, "(%ld) failed to allocate wb slot\n", r); + 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, &ib); - if (r) { - DRM_ERROR("amdgpu: failed to get ib (%ld).\n", r); + if (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); @@ -469,22 +461,17 @@ static int gfx_v9_0_ring_test_ib(struct amdgpu_ring *ring, long timeout) r = dma_fence_wait_timeout(f, false, timeout); if (r == 0) { - DRM_ERROR("amdgpu: IB test timed out.\n"); - r = -ETIMEDOUT; - goto err2; + r = -ETIMEDOUT; + goto err2; } else if (r < 0) { - DRM_ERROR("amdgpu: fence wait failed (%ld).\n", r); - goto err2; + goto err2; } tmp = adev->wb.wb[index]; - if (tmp == 0xDEADBEEF) { - DRM_DEBUG("ib test on ring %d succeeded\n", ring->idx); - r = 0; - } else { - DRM_ERROR("ib test on ring %d failed\n", ring->idx); - r = -EINVAL; - } + if (tmp == 0xDEADBEEF) + r = 0; + else + r = -EINVAL; err2: amdgpu_ib_free(adev, &ib, NULL); @@ -660,7 +647,20 @@ static int gfx_v9_0_init_microcode(struct amdgpu_device *adev) 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); - snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_rlc.bin", chip_name); + /* + * For Picasso && AM4 SOCKET board, we use picasso_rlc_am4.bin + * instead of picasso_rlc.bin. + * Judgment method: + * PCO AM4: revision >= 0xC8 && revision <= 0xCF + * or revision >= 0xD8 && revision <= 0xDF + * otherwise is PCO FP5 + */ + if (!strcmp(chip_name, "picasso") && + (((adev->pdev->revision >= 0xC8) && (adev->pdev->revision <= 0xCF)) || + ((adev->pdev->revision >= 0xD8) && (adev->pdev->revision <= 0xDF)))) + snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_rlc_am4.bin", chip_name); + else + 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; @@ -1065,85 +1065,13 @@ static void gfx_v9_0_enable_lbpw(struct amdgpu_device *adev, bool enable) WREG32_FIELD15(GC, 0, RLC_LB_CNTL, LOAD_BALANCE_ENABLE, enable ? 1 : 0); } -static void rv_init_cp_jump_table(struct amdgpu_device *adev) -{ - const __le32 *fw_data; - volatile u32 *dst_ptr; - int me, i, max_me = 5; - u32 bo_offset = 0; - u32 table_offset, table_size; - - /* write the cp table buffer */ - dst_ptr = adev->gfx.rlc.cp_table_ptr; - for (me = 0; me < max_me; me++) { - if (me == 0) { - const struct gfx_firmware_header_v1_0 *hdr = - (const struct gfx_firmware_header_v1_0 *)adev->gfx.ce_fw->data; - fw_data = (const __le32 *) - (adev->gfx.ce_fw->data + - le32_to_cpu(hdr->header.ucode_array_offset_bytes)); - table_offset = le32_to_cpu(hdr->jt_offset); - table_size = le32_to_cpu(hdr->jt_size); - } else if (me == 1) { - const struct gfx_firmware_header_v1_0 *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(hdr->header.ucode_array_offset_bytes)); - table_offset = le32_to_cpu(hdr->jt_offset); - table_size = le32_to_cpu(hdr->jt_size); - } else if (me == 2) { - const struct gfx_firmware_header_v1_0 *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(hdr->header.ucode_array_offset_bytes)); - table_offset = le32_to_cpu(hdr->jt_offset); - table_size = le32_to_cpu(hdr->jt_size); - } else if (me == 3) { - const struct gfx_firmware_header_v1_0 *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(hdr->header.ucode_array_offset_bytes)); - table_offset = le32_to_cpu(hdr->jt_offset); - table_size = le32_to_cpu(hdr->jt_size); - } else if (me == 4) { - const struct gfx_firmware_header_v1_0 *hdr = - (const struct gfx_firmware_header_v1_0 *)adev->gfx.mec2_fw->data; - fw_data = (const __le32 *) - (adev->gfx.mec2_fw->data + - le32_to_cpu(hdr->header.ucode_array_offset_bytes)); - table_offset = le32_to_cpu(hdr->jt_offset); - table_size = le32_to_cpu(hdr->jt_size); - } - - for (i = 0; i < table_size; i ++) { - dst_ptr[bo_offset + i] = - cpu_to_le32(le32_to_cpu(fw_data[table_offset + i])); - } - - bo_offset += table_size; - } -} - -static void gfx_v9_0_rlc_fini(struct amdgpu_device *adev) +static int gfx_v9_0_cp_jump_table_num(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); + return 5; } static int gfx_v9_0_rlc_init(struct amdgpu_device *adev) { - volatile u32 *dst_ptr; - u32 dws; const struct cs_section_def *cs_data; int r; @@ -1152,45 +1080,18 @@ static int gfx_v9_0_rlc_init(struct amdgpu_device *adev) cs_data = adev->gfx.rlc.cs_data; if (cs_data) { - /* clear state block */ - adev->gfx.rlc.clear_state_size = dws = gfx_v9_0_get_csb_size(adev); - r = amdgpu_bo_create_reserved(adev, dws * 4, PAGE_SIZE, - AMDGPU_GEM_DOMAIN_VRAM, - &adev->gfx.rlc.clear_state_obj, - &adev->gfx.rlc.clear_state_gpu_addr, - (void **)&adev->gfx.rlc.cs_ptr); - if (r) { - dev_err(adev->dev, "(%d) failed to create rlc csb bo\n", - r); - gfx_v9_0_rlc_fini(adev); + /* init clear state block */ + r = amdgpu_gfx_rlc_init_csb(adev); + if (r) return r; - } - /* set up the cs buffer */ - dst_ptr = adev->gfx.rlc.cs_ptr; - gfx_v9_0_get_csb_buffer(adev, dst_ptr); - amdgpu_bo_kunmap(adev->gfx.rlc.clear_state_obj); - amdgpu_bo_unpin(adev->gfx.rlc.clear_state_obj); - amdgpu_bo_unreserve(adev->gfx.rlc.clear_state_obj); } if (adev->asic_type == CHIP_RAVEN) { /* TODO: double check the cp_table_size for RV */ adev->gfx.rlc.cp_table_size = ALIGN(96 * 5 * 4, 2048) + (64 * 1024); /* JT + GDS */ - r = amdgpu_bo_create_reserved(adev, adev->gfx.rlc.cp_table_size, - PAGE_SIZE, AMDGPU_GEM_DOMAIN_VRAM, - &adev->gfx.rlc.cp_table_obj, - &adev->gfx.rlc.cp_table_gpu_addr, - (void **)&adev->gfx.rlc.cp_table_ptr); - if (r) { - dev_err(adev->dev, - "(%d) failed to create cp table bo\n", r); - gfx_v9_0_rlc_fini(adev); + r = amdgpu_gfx_rlc_init_cpt(adev); + if (r) return r; - } - - rv_init_cp_jump_table(adev); - amdgpu_bo_kunmap(adev->gfx.rlc.cp_table_obj); - amdgpu_bo_unreserve(adev->gfx.rlc.cp_table_obj); } switch (adev->asic_type) { @@ -1264,7 +1165,7 @@ static int gfx_v9_0_mec_init(struct amdgpu_device *adev) mec_hpd_size = adev->gfx.num_compute_rings * GFX9_MEC_HPD_SIZE; r = amdgpu_bo_create_reserved(adev, mec_hpd_size, PAGE_SIZE, - AMDGPU_GEM_DOMAIN_GTT, + AMDGPU_GEM_DOMAIN_VRAM, &adev->gfx.mec.hpd_eop_obj, &adev->gfx.mec.hpd_eop_gpu_addr, (void **)&hpd); @@ -1635,8 +1536,8 @@ static int gfx_v9_0_ngg_en(struct amdgpu_device *adev) /* Clear GDS reserved memory */ r = amdgpu_ring_alloc(ring, 17); if (r) { - DRM_ERROR("amdgpu: NGG failed to lock ring %d (%d).\n", - ring->idx, r); + DRM_ERROR("amdgpu: NGG failed to lock ring %s (%d).\n", + ring->name, r); return r; } @@ -1680,7 +1581,7 @@ static int gfx_v9_0_compute_ring_init(struct amdgpu_device *adev, int ring_id, ring->ring_obj = NULL; ring->use_doorbell = true; - ring->doorbell_index = (AMDGPU_DOORBELL_MEC_RING0 + ring_id) << 1; + ring->doorbell_index = (adev->doorbell_index.mec_ring0 + ring_id) << 1; ring->eop_gpu_addr = adev->gfx.mec.hpd_eop_gpu_addr + (ring_id * GFX9_MEC_HPD_SIZE); sprintf(ring->name, "comp_%d.%d.%d", ring->me, ring->pipe, ring->queue); @@ -1748,7 +1649,7 @@ static int gfx_v9_0_sw_init(void *handle) return r; } - r = gfx_v9_0_rlc_init(adev); + r = adev->gfx.rlc.funcs->init(adev); if (r) { DRM_ERROR("Failed to init rlc BOs!\n"); return r; @@ -1769,7 +1670,7 @@ static int gfx_v9_0_sw_init(void *handle) else sprintf(ring->name, "gfx_%d", i); ring->use_doorbell = true; - ring->doorbell_index = AMDGPU_DOORBELL64_GFX_RING0 << 1; + ring->doorbell_index = adev->doorbell_index.gfx_ring0 << 1; r = amdgpu_ring_init(adev, ring, 1024, &adev->gfx.eop_irq, AMDGPU_CP_IRQ_GFX_EOP); if (r) @@ -2499,13 +2400,11 @@ static int gfx_v9_0_rlc_resume(struct amdgpu_device *adev) return 0; } - gfx_v9_0_rlc_stop(adev); + adev->gfx.rlc.funcs->stop(adev); /* disable CG */ WREG32_SOC15(GC, 0, mmRLC_CGCG_CGLS_CTRL, 0); - gfx_v9_0_rlc_reset(adev); - gfx_v9_0_init_pg(adev); if (adev->firmware.load_type != AMDGPU_FW_LOAD_PSP) { @@ -2515,15 +2414,24 @@ static int gfx_v9_0_rlc_resume(struct amdgpu_device *adev) return r; } - if (adev->asic_type == CHIP_RAVEN || - adev->asic_type == CHIP_VEGA20) { - if (amdgpu_lbpw != 0) + switch (adev->asic_type) { + case CHIP_RAVEN: + if (amdgpu_lbpw == 0) + gfx_v9_0_enable_lbpw(adev, false); + else + gfx_v9_0_enable_lbpw(adev, true); + break; + case CHIP_VEGA20: + if (amdgpu_lbpw > 0) gfx_v9_0_enable_lbpw(adev, true); else gfx_v9_0_enable_lbpw(adev, false); + break; + default: + break; } - gfx_v9_0_rlc_start(adev); + adev->gfx.rlc.funcs->start(adev); return 0; } @@ -2538,7 +2446,7 @@ static void gfx_v9_0_cp_gfx_enable(struct amdgpu_device *adev, bool enable) tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, CE_HALT, enable ? 0 : 1); if (!enable) { for (i = 0; i < adev->gfx.num_gfx_rings; i++) - adev->gfx.gfx_ring[i].ready = false; + adev->gfx.gfx_ring[i].sched.ready = false; } WREG32_SOC15(GC, 0, mmCP_ME_CNTL, tmp); udelay(50); @@ -2728,7 +2636,7 @@ static int gfx_v9_0_cp_gfx_resume(struct amdgpu_device *adev) /* start the ring */ gfx_v9_0_cp_gfx_start(adev); - ring->ready = true; + ring->sched.ready = true; return 0; } @@ -2743,8 +2651,8 @@ static void gfx_v9_0_cp_compute_enable(struct amdgpu_device *adev, bool enable) WREG32_SOC15(GC, 0, mmCP_MEC_CNTL, (CP_MEC_CNTL__MEC_ME1_HALT_MASK | CP_MEC_CNTL__MEC_ME2_HALT_MASK)); for (i = 0; i < adev->gfx.num_compute_rings; i++) - adev->gfx.compute_ring[i].ready = false; - adev->gfx.kiq.ring.ready = false; + adev->gfx.compute_ring[i].sched.ready = false; + adev->gfx.kiq.ring.sched.ready = false; } udelay(50); } @@ -2867,11 +2775,9 @@ static int gfx_v9_0_kiq_kcq_enable(struct amdgpu_device *adev) amdgpu_ring_write(kiq_ring, upper_32_bits(wptr_addr)); } - r = amdgpu_ring_test_ring(kiq_ring); - if (r) { + r = amdgpu_ring_test_helper(kiq_ring); + if (r) DRM_ERROR("KCQ enable failed\n"); - kiq_ring->ready = false; - } return r; } @@ -3089,9 +2995,9 @@ static int gfx_v9_0_kiq_init_register(struct amdgpu_ring *ring) /* enable the doorbell if requested */ if (ring->use_doorbell) { WREG32_SOC15(GC, 0, mmCP_MEC_DOORBELL_RANGE_LOWER, - (AMDGPU_DOORBELL64_KIQ *2) << 2); + (adev->doorbell_index.kiq * 2) << 2); WREG32_SOC15(GC, 0, mmCP_MEC_DOORBELL_RANGE_UPPER, - (AMDGPU_DOORBELL64_USERQUEUE_END * 2) << 2); + (adev->doorbell_index.userqueue_end * 2) << 2); } WREG32_SOC15(GC, 0, mmCP_HQD_PQ_DOORBELL_CONTROL, @@ -3250,7 +3156,7 @@ static int gfx_v9_0_kiq_resume(struct amdgpu_device *adev) amdgpu_bo_kunmap(ring->mqd_obj); ring->mqd_ptr = NULL; amdgpu_bo_unreserve(ring->mqd_obj); - ring->ready = true; + ring->sched.ready = true; return 0; } @@ -3315,19 +3221,13 @@ static int gfx_v9_0_cp_resume(struct amdgpu_device *adev) return r; ring = &adev->gfx.gfx_ring[0]; - r = amdgpu_ring_test_ring(ring); - if (r) { - ring->ready = false; + 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]; - - ring->ready = true; - r = amdgpu_ring_test_ring(ring); - if (r) - ring->ready = false; + amdgpu_ring_test_helper(ring); } gfx_v9_0_enable_gui_idle_interrupt(adev, true); @@ -3354,7 +3254,7 @@ static int gfx_v9_0_hw_init(void *handle) if (r) return r; - r = gfx_v9_0_rlc_resume(adev); + r = adev->gfx.rlc.funcs->resume(adev); if (r) return r; @@ -3392,7 +3292,7 @@ static int gfx_v9_0_kcq_disable(struct amdgpu_device *adev) amdgpu_ring_write(kiq_ring, 0); amdgpu_ring_write(kiq_ring, 0); } - r = amdgpu_ring_test_ring(kiq_ring); + r = amdgpu_ring_test_helper(kiq_ring); if (r) DRM_ERROR("KCQ disable failed\n"); @@ -3434,7 +3334,7 @@ static int gfx_v9_0_hw_fini(void *handle) } gfx_v9_0_cp_enable(adev, false); - gfx_v9_0_rlc_stop(adev); + adev->gfx.rlc.funcs->stop(adev); gfx_v9_0_csb_vram_unpin(adev); @@ -3509,7 +3409,7 @@ static int gfx_v9_0_soft_reset(void *handle) if (grbm_soft_reset) { /* stop the rlc */ - gfx_v9_0_rlc_stop(adev); + adev->gfx.rlc.funcs->stop(adev); /* Disable GFX parsing/prefetching */ gfx_v9_0_cp_gfx_enable(adev, false); @@ -3608,64 +3508,47 @@ static int gfx_v9_0_late_init(void *handle) return 0; } -static void gfx_v9_0_enter_rlc_safe_mode(struct amdgpu_device *adev) +static bool gfx_v9_0_is_rlc_enabled(struct amdgpu_device *adev) { - uint32_t rlc_setting, data; - unsigned i; - - if (adev->gfx.rlc.in_safe_mode) - return; + uint32_t rlc_setting; /* if RLC is not enabled, do nothing */ rlc_setting = RREG32_SOC15(GC, 0, mmRLC_CNTL); if (!(rlc_setting & RLC_CNTL__RLC_ENABLE_F32_MASK)) - return; - - if (adev->cg_flags & - (AMD_CG_SUPPORT_GFX_CGCG | AMD_CG_SUPPORT_GFX_MGCG | - AMD_CG_SUPPORT_GFX_3D_CGCG)) { - data = RLC_SAFE_MODE__CMD_MASK; - data |= (1 << RLC_SAFE_MODE__MESSAGE__SHIFT); - WREG32_SOC15(GC, 0, mmRLC_SAFE_MODE, data); + return false; - /* wait for RLC_SAFE_MODE */ - for (i = 0; i < adev->usec_timeout; i++) { - if (!REG_GET_FIELD(RREG32_SOC15(GC, 0, mmRLC_SAFE_MODE), RLC_SAFE_MODE, CMD)) - break; - udelay(1); - } - adev->gfx.rlc.in_safe_mode = true; - } + return true; } -static void gfx_v9_0_exit_rlc_safe_mode(struct amdgpu_device *adev) +static void gfx_v9_0_set_safe_mode(struct amdgpu_device *adev) { - uint32_t rlc_setting, data; + uint32_t data; + unsigned i; - if (!adev->gfx.rlc.in_safe_mode) - return; + data = RLC_SAFE_MODE__CMD_MASK; + data |= (1 << RLC_SAFE_MODE__MESSAGE__SHIFT); + WREG32_SOC15(GC, 0, mmRLC_SAFE_MODE, data); - /* if RLC is not enabled, do nothing */ - rlc_setting = RREG32_SOC15(GC, 0, mmRLC_CNTL); - if (!(rlc_setting & RLC_CNTL__RLC_ENABLE_F32_MASK)) - return; - - if (adev->cg_flags & - (AMD_CG_SUPPORT_GFX_CGCG | AMD_CG_SUPPORT_GFX_MGCG)) { - /* - * Try to exit safe mode only if it is already in safe - * mode. - */ - data = RLC_SAFE_MODE__CMD_MASK; - WREG32_SOC15(GC, 0, mmRLC_SAFE_MODE, data); - adev->gfx.rlc.in_safe_mode = false; + /* wait for RLC_SAFE_MODE */ + for (i = 0; i < adev->usec_timeout; i++) { + if (!REG_GET_FIELD(RREG32_SOC15(GC, 0, mmRLC_SAFE_MODE), RLC_SAFE_MODE, CMD)) + break; + udelay(1); } } +static void gfx_v9_0_unset_safe_mode(struct amdgpu_device *adev) +{ + uint32_t data; + + data = RLC_SAFE_MODE__CMD_MASK; + WREG32_SOC15(GC, 0, mmRLC_SAFE_MODE, data); +} + static void gfx_v9_0_update_gfx_cg_power_gating(struct amdgpu_device *adev, bool enable) { - gfx_v9_0_enter_rlc_safe_mode(adev); + amdgpu_gfx_rlc_enter_safe_mode(adev); if ((adev->pg_flags & AMD_PG_SUPPORT_GFX_PG) && enable) { gfx_v9_0_enable_gfx_cg_power_gating(adev, true); @@ -3676,7 +3559,7 @@ static void gfx_v9_0_update_gfx_cg_power_gating(struct amdgpu_device *adev, gfx_v9_0_enable_gfx_pipeline_powergating(adev, false); } - gfx_v9_0_exit_rlc_safe_mode(adev); + amdgpu_gfx_rlc_exit_safe_mode(adev); } static void gfx_v9_0_update_gfx_mg_power_gating(struct amdgpu_device *adev, @@ -3703,6 +3586,8 @@ static void gfx_v9_0_update_medium_grain_clock_gating(struct amdgpu_device *adev { uint32_t data, def; + amdgpu_gfx_rlc_enter_safe_mode(adev); + /* It is disabled by HW by default */ if (enable && (adev->cg_flags & AMD_CG_SUPPORT_GFX_MGCG)) { /* 1 - RLC_CGTT_MGCG_OVERRIDE */ @@ -3767,6 +3652,8 @@ static void gfx_v9_0_update_medium_grain_clock_gating(struct amdgpu_device *adev WREG32_SOC15(GC, 0, mmCP_MEM_SLP_CNTL, data); } } + + amdgpu_gfx_rlc_exit_safe_mode(adev); } static void gfx_v9_0_update_3d_clock_gating(struct amdgpu_device *adev, @@ -3774,7 +3661,7 @@ static void gfx_v9_0_update_3d_clock_gating(struct amdgpu_device *adev, { uint32_t data, def; - adev->gfx.rlc.funcs->enter_safe_mode(adev); + amdgpu_gfx_rlc_enter_safe_mode(adev); /* Enable 3D CGCG/CGLS */ if (enable && (adev->cg_flags & AMD_CG_SUPPORT_GFX_3D_CGCG)) { @@ -3814,7 +3701,7 @@ static void gfx_v9_0_update_3d_clock_gating(struct amdgpu_device *adev, WREG32_SOC15(GC, 0, mmRLC_CGCG_CGLS_CTRL_3D, data); } - adev->gfx.rlc.funcs->exit_safe_mode(adev); + amdgpu_gfx_rlc_exit_safe_mode(adev); } static void gfx_v9_0_update_coarse_grain_clock_gating(struct amdgpu_device *adev, @@ -3822,7 +3709,7 @@ static void gfx_v9_0_update_coarse_grain_clock_gating(struct amdgpu_device *adev { uint32_t def, data; - adev->gfx.rlc.funcs->enter_safe_mode(adev); + amdgpu_gfx_rlc_enter_safe_mode(adev); if (enable && (adev->cg_flags & AMD_CG_SUPPORT_GFX_CGCG)) { def = data = RREG32_SOC15(GC, 0, mmRLC_CGTT_MGCG_OVERRIDE); @@ -3862,7 +3749,7 @@ static void gfx_v9_0_update_coarse_grain_clock_gating(struct amdgpu_device *adev WREG32_SOC15(GC, 0, mmRLC_CGCG_CGLS_CTRL, data); } - adev->gfx.rlc.funcs->exit_safe_mode(adev); + amdgpu_gfx_rlc_exit_safe_mode(adev); } static int gfx_v9_0_update_gfx_clock_gating(struct amdgpu_device *adev, @@ -3891,8 +3778,17 @@ static int gfx_v9_0_update_gfx_clock_gating(struct amdgpu_device *adev, } static const struct amdgpu_rlc_funcs gfx_v9_0_rlc_funcs = { - .enter_safe_mode = gfx_v9_0_enter_rlc_safe_mode, - .exit_safe_mode = gfx_v9_0_exit_rlc_safe_mode + .is_rlc_enabled = gfx_v9_0_is_rlc_enabled, + .set_safe_mode = gfx_v9_0_set_safe_mode, + .unset_safe_mode = gfx_v9_0_unset_safe_mode, + .init = gfx_v9_0_rlc_init, + .get_csb_size = gfx_v9_0_get_csb_size, + .get_csb_buffer = gfx_v9_0_get_csb_buffer, + .get_cp_table_num = gfx_v9_0_cp_jump_table_num, + .resume = gfx_v9_0_rlc_resume, + .stop = gfx_v9_0_rlc_stop, + .reset = gfx_v9_0_rlc_reset, + .start = gfx_v9_0_rlc_start }; static int gfx_v9_0_set_powergating_state(void *handle, @@ -4073,9 +3969,11 @@ static void gfx_v9_0_ring_emit_hdp_flush(struct amdgpu_ring *ring) } static void gfx_v9_0_ring_emit_ib_gfx(struct amdgpu_ring *ring, - struct amdgpu_ib *ib, - unsigned vmid, bool ctx_switch) + struct amdgpu_job *job, + struct amdgpu_ib *ib, + uint32_t flags) { + unsigned vmid = AMDGPU_JOB_GET_VMID(job); u32 header, control = 0; if (ib->flags & AMDGPU_IB_FLAG_CE) @@ -4104,20 +4002,38 @@ static void gfx_v9_0_ring_emit_ib_gfx(struct amdgpu_ring *ring, } static void gfx_v9_0_ring_emit_ib_compute(struct amdgpu_ring *ring, - struct amdgpu_ib *ib, - unsigned vmid, bool ctx_switch) -{ - u32 control = INDIRECT_BUFFER_VALID | ib->length_dw | (vmid << 24); + 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); + + /* 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, mmGDS_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)); + amdgpu_ring_write(ring, PACKET3(PACKET3_INDIRECT_BUFFER, 2)); BUG_ON(ib->gpu_addr & 0x3); /* Dword align */ - amdgpu_ring_write(ring, + amdgpu_ring_write(ring, #ifdef __BIG_ENDIAN - (2 << 0) | + (2 << 0) | #endif - lower_32_bits(ib->gpu_addr)); - amdgpu_ring_write(ring, upper_32_bits(ib->gpu_addr)); - amdgpu_ring_write(ring, control); + lower_32_bits(ib->gpu_addr)); + amdgpu_ring_write(ring, upper_32_bits(ib->gpu_addr)); + amdgpu_ring_write(ring, control); } static void gfx_v9_0_ring_emit_fence(struct amdgpu_ring *ring, u64 addr, @@ -4696,12 +4612,39 @@ static int gfx_v9_0_eop_irq(struct amdgpu_device *adev, return 0; } +static void gfx_v9_0_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: + drm_sched_fault(&adev->gfx.gfx_ring[0].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; + } +} + static int gfx_v9_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"); - schedule_work(&adev->reset_work); + gfx_v9_0_fault(adev, entry); return 0; } @@ -4710,7 +4653,7 @@ static int gfx_v9_0_priv_inst_irq(struct amdgpu_device *adev, struct amdgpu_iv_entry *entry) { DRM_ERROR("Illegal instruction in command stream\n"); - schedule_work(&adev->reset_work); + gfx_v9_0_fault(adev, entry); return 0; } @@ -4801,7 +4744,7 @@ static const struct amdgpu_ring_funcs gfx_v9_0_ring_funcs_compute = { SOC15_FLUSH_GPU_TLB_NUM_REG_WAIT * 7 + 2 + /* gfx_v9_0_ring_emit_vm_flush */ 8 + 8 + 8, /* gfx_v9_0_ring_emit_fence x3 for user fence, vm fence */ - .emit_ib_size = 4, /* gfx_v9_0_ring_emit_ib_compute */ + .emit_ib_size = 7, /* gfx_v9_0_ring_emit_ib_compute */ .emit_ib = gfx_v9_0_ring_emit_ib_compute, .emit_fence = gfx_v9_0_ring_emit_fence, .emit_pipeline_sync = gfx_v9_0_ring_emit_pipeline_sync, @@ -4836,11 +4779,9 @@ static const struct amdgpu_ring_funcs gfx_v9_0_ring_funcs_kiq = { SOC15_FLUSH_GPU_TLB_NUM_REG_WAIT * 7 + 2 + /* gfx_v9_0_ring_emit_vm_flush */ 8 + 8 + 8, /* gfx_v9_0_ring_emit_fence_kiq x3 for user fence, vm fence */ - .emit_ib_size = 4, /* gfx_v9_0_ring_emit_ib_compute */ - .emit_ib = gfx_v9_0_ring_emit_ib_compute, + .emit_ib_size = 7, /* gfx_v9_0_ring_emit_ib_compute */ .emit_fence = gfx_v9_0_ring_emit_fence_kiq, .test_ring = gfx_v9_0_ring_test_ring, - .test_ib = gfx_v9_0_ring_test_ib, .insert_nop = amdgpu_ring_insert_nop, .pad_ib = amdgpu_ring_generic_pad_ib, .emit_rreg = gfx_v9_0_ring_emit_rreg, @@ -4920,6 +4861,26 @@ static void gfx_v9_0_set_gds_init(struct amdgpu_device *adev) break; } + switch (adev->asic_type) { + case CHIP_VEGA10: + case CHIP_VEGA20: + adev->gds.gds_compute_max_wave_id = 0x7ff; + break; + case CHIP_VEGA12: + adev->gds.gds_compute_max_wave_id = 0x27f; + break; + case CHIP_RAVEN: + if (adev->rev_id >= 0x8) + adev->gds.gds_compute_max_wave_id = 0x77; /* raven2 */ + else + adev->gds.gds_compute_max_wave_id = 0x15f; /* raven1 */ + break; + default: + /* this really depends on the chip */ + adev->gds.gds_compute_max_wave_id = 0x7ff; + break; + } + adev->gds.gws.total_size = 64; adev->gds.oa.total_size = 16; diff --git a/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_0.c b/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_0.c index bfa317ad20a9..f5edddf3b29d 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_0.c @@ -35,20 +35,25 @@ u64 gfxhub_v1_0_get_mc_fb_offset(struct amdgpu_device *adev) return (u64)RREG32_SOC15(GC, 0, mmMC_VM_FB_OFFSET) << 24; } -static void gfxhub_v1_0_init_gart_pt_regs(struct amdgpu_device *adev) +void gfxhub_v1_0_setup_vm_pt_regs(struct amdgpu_device *adev, uint32_t vmid, + uint64_t page_table_base) { - uint64_t value = amdgpu_gmc_pd_addr(adev->gart.bo); + /* two registers distance between mmVM_CONTEXT0_* to mmVM_CONTEXT1_* */ + int offset = mmVM_CONTEXT1_PAGE_TABLE_BASE_ADDR_LO32 + - mmVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32; - WREG32_SOC15(GC, 0, mmVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32, - lower_32_bits(value)); + WREG32_SOC15_OFFSET(GC, 0, mmVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32, + offset * vmid, lower_32_bits(page_table_base)); - WREG32_SOC15(GC, 0, mmVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_HI32, - upper_32_bits(value)); + WREG32_SOC15_OFFSET(GC, 0, mmVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_HI32, + offset * vmid, upper_32_bits(page_table_base)); } static void gfxhub_v1_0_init_gart_aperture_regs(struct amdgpu_device *adev) { - gfxhub_v1_0_init_gart_pt_regs(adev); + uint64_t pt_base = amdgpu_gmc_pd_addr(adev->gart.bo); + + gfxhub_v1_0_setup_vm_pt_regs(adev, 0, pt_base); WREG32_SOC15(GC, 0, mmVM_CONTEXT0_PAGE_TABLE_START_ADDR_LO32, (u32)(adev->gmc.gart_start >> 12)); diff --git a/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_0.h b/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_0.h index 206e29cad753..92d3a70cd9b1 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_0.h +++ b/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_0.h @@ -30,5 +30,7 @@ void gfxhub_v1_0_set_fault_enable_default(struct amdgpu_device *adev, bool value); void gfxhub_v1_0_init(struct amdgpu_device *adev); u64 gfxhub_v1_0_get_mc_fb_offset(struct amdgpu_device *adev); +void gfxhub_v1_0_setup_vm_pt_regs(struct amdgpu_device *adev, uint32_t vmid, + uint64_t page_table_base); #endif diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v6_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v6_0.c index 73ad02aea2b2..9fc3296592fe 100644 --- a/drivers/gpu/drm/amd/amdgpu/gmc_v6_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gmc_v6_0.c @@ -359,7 +359,8 @@ static int gmc_v6_0_mc_init(struct amdgpu_device *adev) return 0; } -static void gmc_v6_0_flush_gpu_tlb(struct amdgpu_device *adev, uint32_t vmid) +static void gmc_v6_0_flush_gpu_tlb(struct amdgpu_device *adev, + uint32_t vmid, uint32_t flush_type) { WREG32(mmVM_INVALIDATE_REQUEST, 1 << vmid); } @@ -581,7 +582,7 @@ static int gmc_v6_0_gart_enable(struct amdgpu_device *adev) else gmc_v6_0_set_fault_enable_default(adev, true); - gmc_v6_0_flush_gpu_tlb(adev, 0); + gmc_v6_0_flush_gpu_tlb(adev, 0, 0); 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); diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v7_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v7_0.c index 910c4ce19cb3..761dcfb2fec0 100644 --- a/drivers/gpu/drm/amd/amdgpu/gmc_v7_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gmc_v7_0.c @@ -430,7 +430,8 @@ static int gmc_v7_0_mc_init(struct amdgpu_device *adev) * * Flush the TLB for the requested page table (CIK). */ -static void gmc_v7_0_flush_gpu_tlb(struct amdgpu_device *adev, uint32_t vmid) +static void gmc_v7_0_flush_gpu_tlb(struct amdgpu_device *adev, + uint32_t vmid, uint32_t flush_type) { /* bits 0-15 are the VM contexts0-15 */ WREG32(mmVM_INVALIDATE_REQUEST, 1 << vmid); @@ -698,7 +699,7 @@ static int gmc_v7_0_gart_enable(struct amdgpu_device *adev) WREG32(mmCHUB_CONTROL, tmp); } - gmc_v7_0_flush_gpu_tlb(adev, 0); + gmc_v7_0_flush_gpu_tlb(adev, 0, 0); DRM_INFO("PCIE GART of %uM enabled (table at 0x%016llX).\n", (unsigned)(adev->gmc.gart_size >> 20), (unsigned long long)table_addr); diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c index 747c068379dc..34440672f938 100644 --- a/drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c @@ -633,7 +633,7 @@ static int gmc_v8_0_mc_init(struct amdgpu_device *adev) * Flush the TLB for the requested page table (CIK). */ static void gmc_v8_0_flush_gpu_tlb(struct amdgpu_device *adev, - uint32_t vmid) + uint32_t vmid, uint32_t flush_type) { /* bits 0-15 are the VM contexts0-15 */ WREG32(mmVM_INVALIDATE_REQUEST, 1 << vmid); @@ -942,7 +942,7 @@ static int gmc_v8_0_gart_enable(struct amdgpu_device *adev) else gmc_v8_0_set_fault_enable_default(adev, true); - gmc_v8_0_flush_gpu_tlb(adev, 0); + gmc_v8_0_flush_gpu_tlb(adev, 0, 0); DRM_INFO("PCIE GART of %uM enabled (table at 0x%016llX).\n", (unsigned)(adev->gmc.gart_size >> 20), (unsigned long long)table_addr); @@ -1471,8 +1471,9 @@ static int gmc_v8_0_process_interrupt(struct amdgpu_device *adev, gmc_v8_0_set_fault_enable_default(adev, false); if (printk_ratelimit()) { - struct amdgpu_task_info task_info = { 0 }; + 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, "GPU fault detected: %d 0x%08x for process %s pid %d thread %s pid %d\n", diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c index f35d7a554ad5..2fe8397241ea 100644 --- a/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c @@ -244,30 +244,93 @@ static int gmc_v9_0_vm_fault_interrupt_state(struct amdgpu_device *adev, return 0; } +/** + * vega10_ih_prescreen_iv - prescreen an interrupt vector + * + * @adev: amdgpu_device pointer + * + * Returns true if the interrupt vector should be further processed. + */ +static bool gmc_v9_0_prescreen_iv(struct amdgpu_device *adev, + struct amdgpu_iv_entry *entry, + uint64_t addr) +{ + struct amdgpu_vm *vm; + u64 key; + int r; + + /* No PASID, can't identify faulting process */ + if (!entry->pasid) + return true; + + /* Not a retry fault */ + if (!(entry->src_data[1] & 0x80)) + return true; + + /* Track retry faults in per-VM fault FIFO. */ + spin_lock(&adev->vm_manager.pasid_lock); + vm = idr_find(&adev->vm_manager.pasid_idr, entry->pasid); + if (!vm) { + /* VM not found, process it normally */ + spin_unlock(&adev->vm_manager.pasid_lock); + return true; + } + + key = AMDGPU_VM_FAULT(entry->pasid, addr); + r = amdgpu_vm_add_fault(vm->fault_hash, key); + + /* Hash table is full or the fault is already being processed, + * ignore further page faults + */ + if (r != 0) { + spin_unlock(&adev->vm_manager.pasid_lock); + return false; + } + /* No locking required with single writer and single reader */ + r = kfifo_put(&vm->faults, key); + if (!r) { + /* FIFO is full. Ignore it until there is space */ + amdgpu_vm_clear_fault(vm->fault_hash, key); + spin_unlock(&adev->vm_manager.pasid_lock); + return false; + } + + spin_unlock(&adev->vm_manager.pasid_lock); + /* It's the first fault for this address, process it normally */ + return true; +} + static int gmc_v9_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]; + bool retry_fault = !!(entry->src_data[1] & 0x80); uint32_t status = 0; u64 addr; addr = (u64)entry->src_data[0] << 12; addr |= ((u64)entry->src_data[1] & 0xf) << 44; + if (!gmc_v9_0_prescreen_iv(adev, entry, addr)) + return 1; /* This also prevents sending it to KFD */ + if (!amdgpu_sriov_vf(adev)) { 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 = { 0 }; + 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] VMC page fault (src_id:%u ring:%u vmid:%u pasid:%u, for process %s pid %d thread %s pid %d)\n", + "[%s] %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", + retry_fault ? "retry" : "no-retry", entry->src_id, entry->ring_id, entry->vmid, entry->pasid, task_info.process_name, task_info.tgid, task_info.task_name, task_info.pid); @@ -293,14 +356,14 @@ static void gmc_v9_0_set_irq_funcs(struct amdgpu_device *adev) adev->gmc.vm_fault.funcs = &gmc_v9_0_irq_funcs; } -static uint32_t gmc_v9_0_get_invalidate_req(unsigned int vmid) +static uint32_t gmc_v9_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, VM_INVALIDATE_ENG0_REQ, PER_VMID_INVALIDATE_REQ, 1 << vmid); - req = REG_SET_FIELD(req, VM_INVALIDATE_ENG0_REQ, FLUSH_TYPE, 0); + req = REG_SET_FIELD(req, VM_INVALIDATE_ENG0_REQ, FLUSH_TYPE, flush_type); req = REG_SET_FIELD(req, VM_INVALIDATE_ENG0_REQ, INVALIDATE_L2_PTES, 1); req = REG_SET_FIELD(req, VM_INVALIDATE_ENG0_REQ, INVALIDATE_L2_PDE0, 1); req = REG_SET_FIELD(req, VM_INVALIDATE_ENG0_REQ, INVALIDATE_L2_PDE1, 1); @@ -312,48 +375,6 @@ static uint32_t gmc_v9_0_get_invalidate_req(unsigned int vmid) return req; } -static signed long amdgpu_kiq_reg_write_reg_wait(struct amdgpu_device *adev, - uint32_t reg0, uint32_t reg1, - uint32_t ref, uint32_t mask) -{ - signed long r, cnt = 0; - unsigned long flags; - uint32_t seq; - struct amdgpu_kiq *kiq = &adev->gfx.kiq; - struct amdgpu_ring *ring = &kiq->ring; - - spin_lock_irqsave(&kiq->ring_lock, flags); - - amdgpu_ring_alloc(ring, 32); - amdgpu_ring_emit_reg_write_reg_wait(ring, reg0, reg1, - ref, mask); - amdgpu_fence_emit_polling(ring, &seq); - amdgpu_ring_commit(ring); - spin_unlock_irqrestore(&kiq->ring_lock, flags); - - r = amdgpu_fence_wait_polling(ring, seq, MAX_KIQ_REG_WAIT); - - /* don't wait anymore for IRQ context */ - if (r < 1 && in_interrupt()) - goto failed_kiq; - - might_sleep(); - - while (r < 1 && cnt++ < MAX_KIQ_REG_TRY) { - msleep(MAX_KIQ_REG_BAILOUT_INTERVAL); - r = amdgpu_fence_wait_polling(ring, seq, MAX_KIQ_REG_WAIT); - } - - if (cnt > MAX_KIQ_REG_TRY) - goto failed_kiq; - - return 0; - -failed_kiq: - pr_err("failed to invalidate tlb with kiq\n"); - return r; -} - /* * GART * VMID 0 is the physical GPU addresses as used by the kernel. @@ -362,64 +383,50 @@ failed_kiq: */ /** - * gmc_v9_0_flush_gpu_tlb - gart tlb flush callback + * gmc_v9_0_flush_gpu_tlb - tlb flush with certain type * * @adev: amdgpu_device pointer * @vmid: vm instance to flush + * @flush_type: the flush type * - * Flush the TLB for the requested page table. + * Flush the TLB for the requested page table using certain type. */ static void gmc_v9_0_flush_gpu_tlb(struct amdgpu_device *adev, - uint32_t vmid) + uint32_t vmid, uint32_t flush_type) { - /* Use register 17 for GART */ const unsigned eng = 17; unsigned i, j; - int r; for (i = 0; i < AMDGPU_MAX_VMHUBS; ++i) { struct amdgpu_vmhub *hub = &adev->vmhub[i]; - u32 tmp = gmc_v9_0_get_invalidate_req(vmid); + u32 tmp = gmc_v9_0_get_invalidate_req(vmid, flush_type); - if (adev->gfx.kiq.ring.ready && + /* This is necessary for a HW workaround under SRIOV as well + * as GFXOFF under bare metal + */ + if (adev->gfx.kiq.ring.sched.ready && (amdgpu_sriov_runtime(adev) || !amdgpu_sriov_vf(adev)) && !adev->in_gpu_reset) { - r = amdgpu_kiq_reg_write_reg_wait(adev, hub->vm_inv_eng0_req + eng, - hub->vm_inv_eng0_ack + eng, tmp, 1 << vmid); - if (!r) - continue; - } + uint32_t req = hub->vm_inv_eng0_req + eng; + uint32_t ack = hub->vm_inv_eng0_ack + eng; - spin_lock(&adev->gmc.invalidate_lock); - - WREG32_NO_KIQ(hub->vm_inv_eng0_req + eng, tmp); - - /* Busy wait for ACK.*/ - for (j = 0; j < 100; j++) { - tmp = RREG32_NO_KIQ(hub->vm_inv_eng0_ack + eng); - tmp &= 1 << vmid; - if (tmp) - break; - cpu_relax(); - } - if (j < 100) { - spin_unlock(&adev->gmc.invalidate_lock); + amdgpu_virt_kiq_reg_write_reg_wait(adev, req, ack, tmp, + 1 << vmid); continue; } - /* Wait for ACK with a delay.*/ + spin_lock(&adev->gmc.invalidate_lock); + WREG32_NO_KIQ(hub->vm_inv_eng0_req + eng, tmp); for (j = 0; j < adev->usec_timeout; j++) { tmp = RREG32_NO_KIQ(hub->vm_inv_eng0_ack + eng); - tmp &= 1 << vmid; - if (tmp) + if (tmp & (1 << vmid)) break; udelay(1); } - if (j < adev->usec_timeout) { - spin_unlock(&adev->gmc.invalidate_lock); - continue; - } spin_unlock(&adev->gmc.invalidate_lock); + if (j < adev->usec_timeout) + continue; + DRM_ERROR("Timeout waiting for VM flush ACK!\n"); } } @@ -429,7 +436,7 @@ static uint64_t gmc_v9_0_emit_flush_gpu_tlb(struct amdgpu_ring *ring, { struct amdgpu_device *adev = ring->adev; struct amdgpu_vmhub *hub = &adev->vmhub[ring->funcs->vmhub]; - uint32_t req = gmc_v9_0_get_invalidate_req(vmid); + uint32_t req = gmc_v9_0_get_invalidate_req(vmid, 0); unsigned eng = ring->vm_inv_eng; amdgpu_ring_emit_wreg(ring, hub->ctx0_ptb_addr_lo32 + (2 * vmid), @@ -715,38 +722,46 @@ static bool gmc_v9_0_keep_stolen_memory(struct amdgpu_device *adev) } } +static int gmc_v9_0_allocate_vm_inv_eng(struct amdgpu_device *adev) +{ + struct amdgpu_ring *ring; + unsigned vm_inv_engs[AMDGPU_MAX_VMHUBS] = + {GFXHUB_FREE_VM_INV_ENGS_BITMAP, MMHUB_FREE_VM_INV_ENGS_BITMAP}; + unsigned i; + unsigned vmhub, inv_eng; + + for (i = 0; i < adev->num_rings; ++i) { + ring = adev->rings[i]; + vmhub = ring->funcs->vmhub; + + inv_eng = ffs(vm_inv_engs[vmhub]); + if (!inv_eng) { + dev_err(adev->dev, "no VM inv eng for ring %s\n", + ring->name); + return -EINVAL; + } + + ring->vm_inv_eng = inv_eng - 1; + vm_inv_engs[vmhub] &= ~(1 << ring->vm_inv_eng); + + dev_info(adev->dev, "ring %s uses VM inv eng %u on hub %u\n", + ring->name, ring->vm_inv_eng, ring->funcs->vmhub); + } + + return 0; +} + static int gmc_v9_0_late_init(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; - /* - * The latest engine allocation on gfx9 is: - * Engine 0, 1: idle - * Engine 2, 3: firmware - * Engine 4~13: amdgpu ring, subject to change when ring number changes - * Engine 14~15: idle - * Engine 16: kfd tlb invalidation - * Engine 17: Gart flushes - */ - unsigned vm_inv_eng[AMDGPU_MAX_VMHUBS] = { 4, 4 }; - unsigned i; int r; if (!gmc_v9_0_keep_stolen_memory(adev)) amdgpu_bo_late_init(adev); - for(i = 0; i < adev->num_rings; ++i) { - struct amdgpu_ring *ring = adev->rings[i]; - unsigned vmhub = ring->funcs->vmhub; - - ring->vm_inv_eng = vm_inv_eng[vmhub]++; - dev_info(adev->dev, "ring %u(%s) uses VM inv eng %u on hub %u\n", - ring->idx, ring->name, ring->vm_inv_eng, - ring->funcs->vmhub); - } - - /* Engine 16 is used for KFD and 17 for GART flushes */ - for(i = 0; i < AMDGPU_MAX_VMHUBS; ++i) - BUG_ON(vm_inv_eng[i] > 16); + r = gmc_v9_0_allocate_vm_inv_eng(adev); + if (r) + return r; if (adev->asic_type == CHIP_VEGA10 && !amdgpu_sriov_vf(adev)) { r = gmc_v9_0_ecc_available(adev); @@ -950,7 +965,11 @@ static int gmc_v9_0_sw_init(void *handle) * vm size is 256TB (48bit), maximum size of Vega10, * block size 512 (9bit) */ - amdgpu_vm_adjust_size(adev, 256 * 1024, 9, 3, 48); + /* sriov restrict max_pfn below AMDGPU_GMC_HOLE */ + if (amdgpu_sriov_vf(adev)) + amdgpu_vm_adjust_size(adev, 256 * 1024, 9, 3, 47); + else + amdgpu_vm_adjust_size(adev, 256 * 1024, 9, 3, 48); break; default: break; @@ -959,6 +978,9 @@ static int gmc_v9_0_sw_init(void *handle) /* This interrupt is VMC page fault.*/ r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_VMC, VMC_1_0__SRCID__VM_FAULT, &adev->gmc.vm_fault); + if (r) + return r; + r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_UTCL2, UTCL2_1_0__SRCID__FAULT, &adev->gmc.vm_fault); @@ -991,7 +1013,7 @@ static int gmc_v9_0_sw_init(void *handle) } adev->need_swiotlb = drm_get_max_iomem() > ((u64)1 << dma_bits); - if (adev->asic_type == CHIP_VEGA20) { + if (adev->gmc.xgmi.supported) { r = gfxhub_v1_1_get_xgmi_info(adev); if (r) return r; @@ -1122,7 +1144,7 @@ static int gmc_v9_0_gart_enable(struct amdgpu_device *adev) gfxhub_v1_0_set_fault_enable_default(adev, value); mmhub_v1_0_set_fault_enable_default(adev, value); - gmc_v9_0_flush_gpu_tlb(adev, 0); + gmc_v9_0_flush_gpu_tlb(adev, 0, 0); DRM_INFO("PCIE GART of %uM enabled (table at 0x%016llX).\n", (unsigned)(adev->gmc.gart_size >> 20), diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.h b/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.h index b030ca5ea107..5c8deac65580 100644 --- a/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.h +++ b/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.h @@ -24,6 +24,16 @@ #ifndef __GMC_V9_0_H__ #define __GMC_V9_0_H__ + /* + * The latest engine allocation on gfx9 is: + * Engine 2, 3: firmware + * Engine 0, 1, 4~16: amdgpu ring, + * subject to change when ring number changes + * Engine 17: Gart flushes + */ +#define GFXHUB_FREE_VM_INV_ENGS_BITMAP 0x1FFF3 +#define MMHUB_FREE_VM_INV_ENGS_BITMAP 0x1FFF3 + extern const struct amd_ip_funcs gmc_v9_0_ip_funcs; extern const struct amdgpu_ip_block_version gmc_v9_0_ip_block; diff --git a/drivers/gpu/drm/amd/amdgpu/iceland_ih.c b/drivers/gpu/drm/amd/amdgpu/iceland_ih.c index cf0fc61aebe6..b1626e1d2f5d 100644 --- a/drivers/gpu/drm/amd/amdgpu/iceland_ih.c +++ b/drivers/gpu/drm/amd/amdgpu/iceland_ih.c @@ -103,9 +103,9 @@ static void iceland_ih_disable_interrupts(struct amdgpu_device *adev) */ static int iceland_ih_irq_init(struct amdgpu_device *adev) { + struct amdgpu_ih_ring *ih = &adev->irq.ih; int rb_bufsz; u32 interrupt_cntl, ih_cntl, ih_rb_cntl; - u64 wptr_off; /* disable irqs */ iceland_ih_disable_interrupts(adev); @@ -133,9 +133,8 @@ static int iceland_ih_irq_init(struct amdgpu_device *adev) ih_rb_cntl = REG_SET_FIELD(ih_rb_cntl, IH_RB_CNTL, WPTR_WRITEBACK_ENABLE, 1); /* set the writeback address whether it's enabled or not */ - wptr_off = adev->wb.gpu_addr + (adev->irq.ih.wptr_offs * 4); - WREG32(mmIH_RB_WPTR_ADDR_LO, lower_32_bits(wptr_off)); - WREG32(mmIH_RB_WPTR_ADDR_HI, upper_32_bits(wptr_off) & 0xFF); + WREG32(mmIH_RB_WPTR_ADDR_LO, lower_32_bits(ih->wptr_addr)); + WREG32(mmIH_RB_WPTR_ADDR_HI, upper_32_bits(ih->wptr_addr) & 0xFF); WREG32(mmIH_RB_CNTL, ih_rb_cntl); @@ -185,11 +184,12 @@ static void iceland_ih_irq_disable(struct amdgpu_device *adev) * Used by cz_irq_process(VI). * Returns the value of the wptr. */ -static u32 iceland_ih_get_wptr(struct amdgpu_device *adev) +static u32 iceland_ih_get_wptr(struct amdgpu_device *adev, + struct amdgpu_ih_ring *ih) { u32 wptr, tmp; - wptr = le32_to_cpu(adev->wb.wb[adev->irq.ih.wptr_offs]); + wptr = le32_to_cpu(*ih->wptr_cpu); if (REG_GET_FIELD(wptr, IH_RB_WPTR, RB_OVERFLOW)) { wptr = REG_SET_FIELD(wptr, IH_RB_WPTR, RB_OVERFLOW, 0); @@ -198,41 +198,13 @@ static u32 iceland_ih_get_wptr(struct amdgpu_device *adev) * this should allow us to catchup. */ dev_warn(adev->dev, "IH ring buffer overflow (0x%08X, 0x%08X, 0x%08X)\n", - wptr, adev->irq.ih.rptr, (wptr + 16) & adev->irq.ih.ptr_mask); - adev->irq.ih.rptr = (wptr + 16) & adev->irq.ih.ptr_mask; + wptr, ih->rptr, (wptr + 16) & ih->ptr_mask); + ih->rptr = (wptr + 16) & ih->ptr_mask; tmp = RREG32(mmIH_RB_CNTL); tmp = REG_SET_FIELD(tmp, IH_RB_CNTL, WPTR_OVERFLOW_CLEAR, 1); WREG32(mmIH_RB_CNTL, tmp); } - return (wptr & adev->irq.ih.ptr_mask); -} - -/** - * iceland_ih_prescreen_iv - prescreen an interrupt vector - * - * @adev: amdgpu_device pointer - * - * Returns true if the interrupt vector should be further processed. - */ -static bool iceland_ih_prescreen_iv(struct amdgpu_device *adev) -{ - u32 ring_index = adev->irq.ih.rptr >> 2; - u16 pasid; - - switch (le32_to_cpu(adev->irq.ih.ring[ring_index]) & 0xff) { - case 146: - case 147: - pasid = le32_to_cpu(adev->irq.ih.ring[ring_index + 2]) >> 16; - if (!pasid || amdgpu_vm_pasid_fault_credit(adev, pasid)) - return true; - break; - default: - /* Not a VM fault */ - return true; - } - - adev->irq.ih.rptr += 16; - return false; + return (wptr & ih->ptr_mask); } /** @@ -244,16 +216,17 @@ static bool iceland_ih_prescreen_iv(struct amdgpu_device *adev) * position and also advance the position. */ static void iceland_ih_decode_iv(struct amdgpu_device *adev, + struct amdgpu_ih_ring *ih, struct amdgpu_iv_entry *entry) { /* wptr/rptr are in bytes! */ - u32 ring_index = adev->irq.ih.rptr >> 2; + u32 ring_index = ih->rptr >> 2; uint32_t dw[4]; - dw[0] = le32_to_cpu(adev->irq.ih.ring[ring_index + 0]); - dw[1] = le32_to_cpu(adev->irq.ih.ring[ring_index + 1]); - dw[2] = le32_to_cpu(adev->irq.ih.ring[ring_index + 2]); - dw[3] = le32_to_cpu(adev->irq.ih.ring[ring_index + 3]); + dw[0] = le32_to_cpu(ih->ring[ring_index + 0]); + dw[1] = le32_to_cpu(ih->ring[ring_index + 1]); + dw[2] = le32_to_cpu(ih->ring[ring_index + 2]); + dw[3] = le32_to_cpu(ih->ring[ring_index + 3]); entry->client_id = AMDGPU_IRQ_CLIENTID_LEGACY; entry->src_id = dw[0] & 0xff; @@ -263,7 +236,7 @@ static void iceland_ih_decode_iv(struct amdgpu_device *adev, entry->pasid = (dw[2] >> 16) & 0xffff; /* wptr/rptr are in bytes! */ - adev->irq.ih.rptr += 16; + ih->rptr += 16; } /** @@ -273,9 +246,10 @@ static void iceland_ih_decode_iv(struct amdgpu_device *adev, * * Set the IH ring buffer rptr. */ -static void iceland_ih_set_rptr(struct amdgpu_device *adev) +static void iceland_ih_set_rptr(struct amdgpu_device *adev, + struct amdgpu_ih_ring *ih) { - WREG32(mmIH_RB_RPTR, adev->irq.ih.rptr); + WREG32(mmIH_RB_RPTR, ih->rptr); } static int iceland_ih_early_init(void *handle) @@ -440,7 +414,6 @@ static const struct amd_ip_funcs iceland_ih_ip_funcs = { static const struct amdgpu_ih_funcs iceland_ih_funcs = { .get_wptr = iceland_ih_get_wptr, - .prescreen_iv = iceland_ih_prescreen_iv, .decode_iv = iceland_ih_decode_iv, .set_rptr = iceland_ih_set_rptr }; diff --git a/drivers/gpu/drm/amd/amdgpu/kv_dpm.c b/drivers/gpu/drm/amd/amdgpu/kv_dpm.c index d0e478f43443..0c9a2c03504e 100644 --- a/drivers/gpu/drm/amd/amdgpu/kv_dpm.c +++ b/drivers/gpu/drm/amd/amdgpu/kv_dpm.c @@ -508,19 +508,19 @@ static int kv_enable_didt(struct amdgpu_device *adev, bool enable) pi->caps_db_ramping || pi->caps_td_ramping || pi->caps_tcp_ramping) { - adev->gfx.rlc.funcs->enter_safe_mode(adev); + amdgpu_gfx_rlc_enter_safe_mode(adev); if (enable) { ret = kv_program_pt_config_registers(adev, didt_config_kv); if (ret) { - adev->gfx.rlc.funcs->exit_safe_mode(adev); + amdgpu_gfx_rlc_exit_safe_mode(adev); return ret; } } kv_do_enable_didt(adev, enable); - adev->gfx.rlc.funcs->exit_safe_mode(adev); + amdgpu_gfx_rlc_exit_safe_mode(adev); } return 0; diff --git a/drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.c b/drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.c index a0db67adc34c..1696644ec022 100644 --- a/drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.c +++ b/drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.c @@ -52,20 +52,25 @@ u64 mmhub_v1_0_get_fb_location(struct amdgpu_device *adev) return base; } -static void mmhub_v1_0_init_gart_pt_regs(struct amdgpu_device *adev) +void mmhub_v1_0_setup_vm_pt_regs(struct amdgpu_device *adev, uint32_t vmid, + uint64_t page_table_base) { - uint64_t value = amdgpu_gmc_pd_addr(adev->gart.bo); + /* two registers distance between mmVM_CONTEXT0_* to mmVM_CONTEXT1_* */ + int offset = mmVM_CONTEXT1_PAGE_TABLE_BASE_ADDR_LO32 + - mmVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32; - WREG32_SOC15(MMHUB, 0, mmVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32, - lower_32_bits(value)); + WREG32_SOC15_OFFSET(MMHUB, 0, mmVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32, + offset * vmid, lower_32_bits(page_table_base)); - WREG32_SOC15(MMHUB, 0, mmVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_HI32, - upper_32_bits(value)); + WREG32_SOC15_OFFSET(MMHUB, 0, mmVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_HI32, + offset * vmid, upper_32_bits(page_table_base)); } static void mmhub_v1_0_init_gart_aperture_regs(struct amdgpu_device *adev) { - mmhub_v1_0_init_gart_pt_regs(adev); + uint64_t pt_base = amdgpu_gmc_pd_addr(adev->gart.bo); + + mmhub_v1_0_setup_vm_pt_regs(adev, 0, pt_base); WREG32_SOC15(MMHUB, 0, mmVM_CONTEXT0_PAGE_TABLE_START_ADDR_LO32, (u32)(adev->gmc.gart_start >> 12)); @@ -177,6 +182,7 @@ static void mmhub_v1_0_init_cache_regs(struct amdgpu_device *adev) tmp = REG_SET_FIELD(tmp, VM_L2_CNTL3, L2_CACHE_BIGK_FRAGMENT_SIZE, 6); } + WREG32_SOC15(MMHUB, 0, mmVM_L2_CNTL3, tmp); tmp = mmVM_L2_CNTL4_DEFAULT; tmp = REG_SET_FIELD(tmp, VM_L2_CNTL4, VMC_TAP_PDE_REQUEST_PHYSICAL, 0); diff --git a/drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.h b/drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.h index bef3d0c0c117..0de0fdf98c00 100644 --- a/drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.h +++ b/drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.h @@ -34,5 +34,7 @@ int mmhub_v1_0_set_clockgating(struct amdgpu_device *adev, void mmhub_v1_0_get_clockgating(struct amdgpu_device *adev, u32 *flags); void mmhub_v1_0_update_power_gating(struct amdgpu_device *adev, bool enable); +void mmhub_v1_0_setup_vm_pt_regs(struct amdgpu_device *adev, uint32_t vmid, + uint64_t page_table_base); #endif diff --git a/drivers/gpu/drm/amd/amdgpu/mxgpu_ai.c b/drivers/gpu/drm/amd/amdgpu/mxgpu_ai.c index 8cbb4655896a..73851ebb3833 100644 --- a/drivers/gpu/drm/amd/amdgpu/mxgpu_ai.c +++ b/drivers/gpu/drm/amd/amdgpu/mxgpu_ai.c @@ -174,7 +174,7 @@ static int xgpu_ai_send_access_requests(struct amdgpu_device *adev, return r; } /* Retrieve checksum from mailbox2 */ - if (req == IDH_REQ_GPU_INIT_ACCESS) { + if (req == IDH_REQ_GPU_INIT_ACCESS || req == IDH_REQ_GPU_RESET_ACCESS) { adev->virt.fw_reserve.checksum_key = RREG32_NO_KIQ(SOC15_REG_OFFSET(NBIO, 0, mmBIF_BX_PF0_MAILBOX_MSGBUF_RCV_DW2)); @@ -266,7 +266,8 @@ flr_done: } /* Trigger recovery for world switch failure if no TDR */ - if (amdgpu_device_should_recover_gpu(adev)) + if (amdgpu_device_should_recover_gpu(adev) + && amdgpu_lockup_timeout == MAX_SCHEDULE_TIMEOUT) amdgpu_device_gpu_recover(adev, NULL); } diff --git a/drivers/gpu/drm/amd/amdgpu/mxgpu_vi.c b/drivers/gpu/drm/amd/amdgpu/mxgpu_vi.c index 64e875d528dd..6a0fcd67662a 100644 --- a/drivers/gpu/drm/amd/amdgpu/mxgpu_vi.c +++ b/drivers/gpu/drm/amd/amdgpu/mxgpu_vi.c @@ -37,7 +37,6 @@ #include "gmc/gmc_8_2_sh_mask.h" #include "oss/oss_3_0_d.h" #include "oss/oss_3_0_sh_mask.h" -#include "gca/gfx_8_0_sh_mask.h" #include "dce/dce_10_0_d.h" #include "dce/dce_10_0_sh_mask.h" #include "smu/smu_7_1_3_d.h" diff --git a/drivers/gpu/drm/amd/amdgpu/nbio_v6_1.c b/drivers/gpu/drm/amd/amdgpu/nbio_v6_1.c index 6f9c54978cc1..cc967dbfd631 100644 --- a/drivers/gpu/drm/amd/amdgpu/nbio_v6_1.c +++ b/drivers/gpu/drm/amd/amdgpu/nbio_v6_1.c @@ -27,12 +27,9 @@ #include "nbio/nbio_6_1_default.h" #include "nbio/nbio_6_1_offset.h" #include "nbio/nbio_6_1_sh_mask.h" +#include "nbio/nbio_6_1_smn.h" #include "vega10_enum.h" -#define smnCPM_CONTROL 0x11180460 -#define smnPCIE_CNTL2 0x11180070 -#define smnPCIE_CONFIG_CNTL 0x11180044 - static u32 nbio_v6_1_get_rev_id(struct amdgpu_device *adev) { u32 tmp = RREG32_SOC15(NBIO, 0, mmRCC_DEV0_EPF0_STRAP0); @@ -71,7 +68,7 @@ static u32 nbio_v6_1_get_memsize(struct amdgpu_device *adev) } static void nbio_v6_1_sdma_doorbell_range(struct amdgpu_device *adev, int instance, - bool use_doorbell, int doorbell_index) + bool use_doorbell, int doorbell_index, int doorbell_size) { u32 reg = instance == 0 ? SOC15_REG_OFFSET(NBIO, 0, mmBIF_SDMA0_DOORBELL_RANGE) : SOC15_REG_OFFSET(NBIO, 0, mmBIF_SDMA1_DOORBELL_RANGE); @@ -80,7 +77,7 @@ static void nbio_v6_1_sdma_doorbell_range(struct amdgpu_device *adev, int instan if (use_doorbell) { doorbell_range = REG_SET_FIELD(doorbell_range, BIF_SDMA0_DOORBELL_RANGE, OFFSET, doorbell_index); - doorbell_range = REG_SET_FIELD(doorbell_range, BIF_SDMA0_DOORBELL_RANGE, SIZE, 2); + doorbell_range = REG_SET_FIELD(doorbell_range, BIF_SDMA0_DOORBELL_RANGE, SIZE, doorbell_size); } else doorbell_range = REG_SET_FIELD(doorbell_range, BIF_SDMA0_DOORBELL_RANGE, SIZE, 0); @@ -270,6 +267,12 @@ static void nbio_v6_1_init_registers(struct amdgpu_device *adev) if (def != data) WREG32_PCIE(smnPCIE_CONFIG_CNTL, data); + + def = data = RREG32_PCIE(smnPCIE_CI_CNTL); + data = REG_SET_FIELD(data, PCIE_CI_CNTL, CI_SLV_ORDERING_DIS, 1); + + if (def != data) + WREG32_PCIE(smnPCIE_CI_CNTL, data); } 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 df34dc79d444..1cdb98ad2db3 100644 --- a/drivers/gpu/drm/amd/amdgpu/nbio_v7_0.c +++ b/drivers/gpu/drm/amd/amdgpu/nbio_v7_0.c @@ -27,13 +27,11 @@ #include "nbio/nbio_7_0_default.h" #include "nbio/nbio_7_0_offset.h" #include "nbio/nbio_7_0_sh_mask.h" +#include "nbio/nbio_7_0_smn.h" #include "vega10_enum.h" #define smnNBIF_MGCG_CTRL_LCLK 0x1013a05c -#define smnCPM_CONTROL 0x11180460 -#define smnPCIE_CNTL2 0x11180070 - static u32 nbio_v7_0_get_rev_id(struct amdgpu_device *adev) { u32 tmp = RREG32_SOC15(NBIO, 0, mmRCC_DEV0_EPF0_STRAP0); @@ -69,7 +67,7 @@ static u32 nbio_v7_0_get_memsize(struct amdgpu_device *adev) } static void nbio_v7_0_sdma_doorbell_range(struct amdgpu_device *adev, int instance, - bool use_doorbell, int doorbell_index) + bool use_doorbell, int doorbell_index, int doorbell_size) { u32 reg = instance == 0 ? SOC15_REG_OFFSET(NBIO, 0, mmBIF_SDMA0_DOORBELL_RANGE) : SOC15_REG_OFFSET(NBIO, 0, mmBIF_SDMA1_DOORBELL_RANGE); @@ -78,7 +76,7 @@ static void nbio_v7_0_sdma_doorbell_range(struct amdgpu_device *adev, int instan if (use_doorbell) { doorbell_range = REG_SET_FIELD(doorbell_range, BIF_SDMA0_DOORBELL_RANGE, OFFSET, doorbell_index); - doorbell_range = REG_SET_FIELD(doorbell_range, BIF_SDMA0_DOORBELL_RANGE, SIZE, 2); + doorbell_range = REG_SET_FIELD(doorbell_range, BIF_SDMA0_DOORBELL_RANGE, SIZE, doorbell_size); } else doorbell_range = REG_SET_FIELD(doorbell_range, BIF_SDMA0_DOORBELL_RANGE, SIZE, 0); diff --git a/drivers/gpu/drm/amd/amdgpu/nbio_v7_4.c b/drivers/gpu/drm/amd/amdgpu/nbio_v7_4.c index f8cee95d61cc..c69d51598cfe 100644 --- a/drivers/gpu/drm/amd/amdgpu/nbio_v7_4.c +++ b/drivers/gpu/drm/amd/amdgpu/nbio_v7_4.c @@ -26,15 +26,13 @@ #include "nbio/nbio_7_4_offset.h" #include "nbio/nbio_7_4_sh_mask.h" +#include "nbio/nbio_7_4_0_smn.h" #define smnNBIF_MGCG_CTRL_LCLK 0x1013a21c -#define smnCPM_CONTROL 0x11180460 -#define smnPCIE_CNTL2 0x11180070 - static u32 nbio_v7_4_get_rev_id(struct amdgpu_device *adev) { - u32 tmp = RREG32_SOC15(NBIO, 0, mmRCC_DEV0_EPF0_STRAP0); + u32 tmp = RREG32_SOC15(NBIO, 0, mmRCC_DEV0_EPF0_STRAP0); tmp &= RCC_DEV0_EPF0_STRAP0__STRAP_ATI_REV_ID_DEV0_F0_MASK; tmp >>= RCC_DEV0_EPF0_STRAP0__STRAP_ATI_REV_ID_DEV0_F0__SHIFT; @@ -67,7 +65,7 @@ static u32 nbio_v7_4_get_memsize(struct amdgpu_device *adev) } static void nbio_v7_4_sdma_doorbell_range(struct amdgpu_device *adev, int instance, - bool use_doorbell, int doorbell_index) + bool use_doorbell, int doorbell_index, int doorbell_size) { u32 reg = instance == 0 ? SOC15_REG_OFFSET(NBIO, 0, mmBIF_SDMA0_DOORBELL_RANGE) : SOC15_REG_OFFSET(NBIO, 0, mmBIF_SDMA1_DOORBELL_RANGE); @@ -76,7 +74,7 @@ static void nbio_v7_4_sdma_doorbell_range(struct amdgpu_device *adev, int instan if (use_doorbell) { doorbell_range = REG_SET_FIELD(doorbell_range, BIF_SDMA0_DOORBELL_RANGE, OFFSET, doorbell_index); - doorbell_range = REG_SET_FIELD(doorbell_range, BIF_SDMA0_DOORBELL_RANGE, SIZE, 2); + doorbell_range = REG_SET_FIELD(doorbell_range, BIF_SDMA0_DOORBELL_RANGE, SIZE, doorbell_size); } else doorbell_range = REG_SET_FIELD(doorbell_range, BIF_SDMA0_DOORBELL_RANGE, SIZE, 0); @@ -92,7 +90,20 @@ static void nbio_v7_4_enable_doorbell_aperture(struct amdgpu_device *adev, static void nbio_v7_4_enable_doorbell_selfring_aperture(struct amdgpu_device *adev, bool enable) { + u32 tmp = 0; + + if (enable) { + tmp = REG_SET_FIELD(tmp, DOORBELL_SELFRING_GPA_APER_CNTL, DOORBELL_SELFRING_GPA_APER_EN, 1) | + REG_SET_FIELD(tmp, DOORBELL_SELFRING_GPA_APER_CNTL, DOORBELL_SELFRING_GPA_APER_MODE, 1) | + REG_SET_FIELD(tmp, DOORBELL_SELFRING_GPA_APER_CNTL, DOORBELL_SELFRING_GPA_APER_SIZE, 0); + + WREG32_SOC15(NBIO, 0, mmDOORBELL_SELFRING_GPA_APER_BASE_LOW, + lower_32_bits(adev->doorbell.base)); + WREG32_SOC15(NBIO, 0, mmDOORBELL_SELFRING_GPA_APER_BASE_HIGH, + upper_32_bits(adev->doorbell.base)); + } + WREG32_SOC15(NBIO, 0, mmDOORBELL_SELFRING_GPA_APER_CNTL, tmp); } static void nbio_v7_4_ih_doorbell_range(struct amdgpu_device *adev, @@ -222,7 +233,13 @@ static void nbio_v7_4_detect_hw_virt(struct amdgpu_device *adev) static void nbio_v7_4_init_registers(struct amdgpu_device *adev) { + uint32_t def, data; + + def = data = RREG32_PCIE(smnPCIE_CI_CNTL); + data = REG_SET_FIELD(data, PCIE_CI_CNTL, CI_SLV_ORDERING_DIS, 1); + if (def != data) + WREG32_PCIE(smnPCIE_CI_CNTL, data); } const struct amdgpu_nbio_funcs nbio_v7_4_funcs = { diff --git a/drivers/gpu/drm/amd/amdgpu/psp_gfx_if.h b/drivers/gpu/drm/amd/amdgpu/psp_gfx_if.h index 882bd83a28c4..f3a7d207af07 100644 --- a/drivers/gpu/drm/amd/amdgpu/psp_gfx_if.h +++ b/drivers/gpu/drm/amd/amdgpu/psp_gfx_if.h @@ -43,6 +43,8 @@ enum psp_gfx_crtl_cmd_id GFX_CTRL_CMD_ID_ENABLE_INT = 0x00050000, /* enable PSP-to-Gfx interrupt */ GFX_CTRL_CMD_ID_DISABLE_INT = 0x00060000, /* disable PSP-to-Gfx interrupt */ GFX_CTRL_CMD_ID_MODE1_RST = 0x00070000, /* trigger the Mode 1 reset */ + GFX_CTRL_CMD_ID_CONSUME_CMD = 0x000A0000, /* send interrupt to psp for updating write pointer of vf */ + GFX_CTRL_CMD_ID_DESTROY_GPCOM_RING = 0x000C0000, /* destroy GPCOM ring */ GFX_CTRL_CMD_ID_MAX = 0x000F0000, /* max command ID */ }; @@ -89,7 +91,8 @@ enum psp_gfx_cmd_id GFX_CMD_ID_LOAD_IP_FW = 0x00000006, /* load HW IP FW */ GFX_CMD_ID_DESTROY_TMR = 0x00000007, /* destroy TMR region */ GFX_CMD_ID_SAVE_RESTORE = 0x00000008, /* save/restore HW IP FW */ - + GFX_CMD_ID_SETUP_VMR = 0x00000009, /* setup VMR region */ + GFX_CMD_ID_DESTROY_VMR = 0x0000000A, /* destroy VMR region */ }; @@ -188,7 +191,7 @@ enum psp_gfx_fw_type GFX_FW_TYPE_MMSCH = 19, GFX_FW_TYPE_RLC_RESTORE_LIST_GPM_MEM = 20, GFX_FW_TYPE_RLC_RESTORE_LIST_SRM_MEM = 21, - GFX_FW_TYPE_RLC_RESTORE_LIST_CNTL = 22, + GFX_FW_TYPE_RLC_RESTORE_LIST_SRM_CNTL = 22, GFX_FW_TYPE_UVD1 = 23, GFX_FW_TYPE_MAX = 24 }; diff --git a/drivers/gpu/drm/amd/amdgpu/psp_v10_0.c b/drivers/gpu/drm/amd/amdgpu/psp_v10_0.c index 295c2205485a..77c2bc344dfc 100644 --- a/drivers/gpu/drm/amd/amdgpu/psp_v10_0.c +++ b/drivers/gpu/drm/amd/amdgpu/psp_v10_0.c @@ -38,75 +38,6 @@ MODULE_FIRMWARE("amdgpu/raven_asd.bin"); MODULE_FIRMWARE("amdgpu/picasso_asd.bin"); MODULE_FIRMWARE("amdgpu/raven2_asd.bin"); -static int -psp_v10_0_get_fw_type(struct amdgpu_firmware_info *ucode, enum psp_gfx_fw_type *type) -{ - switch(ucode->ucode_id) { - case AMDGPU_UCODE_ID_SDMA0: - *type = GFX_FW_TYPE_SDMA0; - break; - case AMDGPU_UCODE_ID_SDMA1: - *type = GFX_FW_TYPE_SDMA1; - break; - case AMDGPU_UCODE_ID_CP_CE: - *type = GFX_FW_TYPE_CP_CE; - break; - case AMDGPU_UCODE_ID_CP_PFP: - *type = GFX_FW_TYPE_CP_PFP; - break; - case AMDGPU_UCODE_ID_CP_ME: - *type = GFX_FW_TYPE_CP_ME; - break; - case AMDGPU_UCODE_ID_CP_MEC1: - *type = GFX_FW_TYPE_CP_MEC; - break; - case AMDGPU_UCODE_ID_CP_MEC1_JT: - *type = GFX_FW_TYPE_CP_MEC_ME1; - break; - case AMDGPU_UCODE_ID_CP_MEC2: - *type = GFX_FW_TYPE_CP_MEC; - break; - case AMDGPU_UCODE_ID_CP_MEC2_JT: - *type = GFX_FW_TYPE_CP_MEC_ME2; - break; - case AMDGPU_UCODE_ID_RLC_G: - *type = GFX_FW_TYPE_RLC_G; - break; - case AMDGPU_UCODE_ID_RLC_RESTORE_LIST_CNTL: - *type = GFX_FW_TYPE_RLC_RESTORE_LIST_CNTL; - break; - case AMDGPU_UCODE_ID_RLC_RESTORE_LIST_GPM_MEM: - *type = GFX_FW_TYPE_RLC_RESTORE_LIST_GPM_MEM; - break; - case AMDGPU_UCODE_ID_RLC_RESTORE_LIST_SRM_MEM: - *type = GFX_FW_TYPE_RLC_RESTORE_LIST_SRM_MEM; - break; - case AMDGPU_UCODE_ID_SMC: - *type = GFX_FW_TYPE_SMU; - break; - case AMDGPU_UCODE_ID_UVD: - *type = GFX_FW_TYPE_UVD; - break; - case AMDGPU_UCODE_ID_VCE: - *type = GFX_FW_TYPE_VCE; - break; - case AMDGPU_UCODE_ID_VCN: - *type = GFX_FW_TYPE_VCN; - break; - case AMDGPU_UCODE_ID_DMCU_ERAM: - *type = GFX_FW_TYPE_DMCU_ERAM; - break; - case AMDGPU_UCODE_ID_DMCU_INTV: - *type = GFX_FW_TYPE_DMCU_ISR; - break; - case AMDGPU_UCODE_ID_MAXIMUM: - default: - return -EINVAL; - } - - return 0; -} - static int psp_v10_0_init_microcode(struct psp_context *psp) { struct amdgpu_device *adev = psp->adev; @@ -158,26 +89,6 @@ out: return err; } -static int psp_v10_0_prep_cmd_buf(struct amdgpu_firmware_info *ucode, - struct psp_gfx_cmd_resp *cmd) -{ - int ret; - uint64_t fw_mem_mc_addr = ucode->mc_addr; - - memset(cmd, 0, sizeof(struct psp_gfx_cmd_resp)); - - cmd->cmd_id = GFX_CMD_ID_LOAD_IP_FW; - cmd->cmd.cmd_load_ip_fw.fw_phy_addr_lo = lower_32_bits(fw_mem_mc_addr); - cmd->cmd.cmd_load_ip_fw.fw_phy_addr_hi = upper_32_bits(fw_mem_mc_addr); - cmd->cmd.cmd_load_ip_fw.fw_size = ucode->ucode_size; - - ret = psp_v10_0_get_fw_type(ucode, &cmd->cmd.cmd_load_ip_fw.fw_type); - if (ret) - DRM_ERROR("Unknown firmware type\n"); - - return ret; -} - static int psp_v10_0_ring_init(struct psp_context *psp, enum psp_ring_type ring_type) { @@ -240,12 +151,9 @@ static int psp_v10_0_ring_stop(struct psp_context *psp, enum psp_ring_type ring_type) { int ret = 0; - struct psp_ring *ring; unsigned int psp_ring_reg = 0; struct amdgpu_device *adev = psp->adev; - ring = &psp->km_ring; - /* Write the ring destroy command to C2PMSG_64 */ psp_ring_reg = 3 << 16; WREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_64, psp_ring_reg); @@ -457,7 +365,6 @@ static int psp_v10_0_mode1_reset(struct psp_context *psp) static const struct psp_funcs psp_v10_0_funcs = { .init_microcode = psp_v10_0_init_microcode, - .prep_cmd_buf = psp_v10_0_prep_cmd_buf, .ring_init = psp_v10_0_ring_init, .ring_create = psp_v10_0_ring_create, .ring_stop = psp_v10_0_ring_stop, diff --git a/drivers/gpu/drm/amd/amdgpu/psp_v11_0.c b/drivers/gpu/drm/amd/amdgpu/psp_v11_0.c index 3f3fac2d50cd..860b70d80d3c 100644 --- a/drivers/gpu/drm/amd/amdgpu/psp_v11_0.c +++ b/drivers/gpu/drm/amd/amdgpu/psp_v11_0.c @@ -34,71 +34,21 @@ #include "nbio/nbio_7_4_offset.h" MODULE_FIRMWARE("amdgpu/vega20_sos.bin"); +MODULE_FIRMWARE("amdgpu/vega20_asd.bin"); +MODULE_FIRMWARE("amdgpu/vega20_ta.bin"); /* address block */ #define smnMP1_FIRMWARE_FLAGS 0x3010024 -static int -psp_v11_0_get_fw_type(struct amdgpu_firmware_info *ucode, enum psp_gfx_fw_type *type) -{ - switch (ucode->ucode_id) { - case AMDGPU_UCODE_ID_SDMA0: - *type = GFX_FW_TYPE_SDMA0; - break; - case AMDGPU_UCODE_ID_SDMA1: - *type = GFX_FW_TYPE_SDMA1; - break; - case AMDGPU_UCODE_ID_CP_CE: - *type = GFX_FW_TYPE_CP_CE; - break; - case AMDGPU_UCODE_ID_CP_PFP: - *type = GFX_FW_TYPE_CP_PFP; - break; - case AMDGPU_UCODE_ID_CP_ME: - *type = GFX_FW_TYPE_CP_ME; - break; - case AMDGPU_UCODE_ID_CP_MEC1: - *type = GFX_FW_TYPE_CP_MEC; - break; - case AMDGPU_UCODE_ID_CP_MEC1_JT: - *type = GFX_FW_TYPE_CP_MEC_ME1; - break; - case AMDGPU_UCODE_ID_CP_MEC2: - *type = GFX_FW_TYPE_CP_MEC; - break; - case AMDGPU_UCODE_ID_CP_MEC2_JT: - *type = GFX_FW_TYPE_CP_MEC_ME2; - break; - case AMDGPU_UCODE_ID_RLC_G: - *type = GFX_FW_TYPE_RLC_G; - break; - case AMDGPU_UCODE_ID_SMC: - *type = GFX_FW_TYPE_SMU; - break; - case AMDGPU_UCODE_ID_UVD: - *type = GFX_FW_TYPE_UVD; - break; - case AMDGPU_UCODE_ID_VCE: - *type = GFX_FW_TYPE_VCE; - break; - case AMDGPU_UCODE_ID_UVD1: - *type = GFX_FW_TYPE_UVD1; - break; - case AMDGPU_UCODE_ID_MAXIMUM: - default: - return -EINVAL; - } - - return 0; -} - static int psp_v11_0_init_microcode(struct psp_context *psp) { struct amdgpu_device *adev = psp->adev; const char *chip_name; char fw_name[30]; int err = 0; - const struct psp_firmware_header_v1_0 *hdr; + const struct psp_firmware_header_v1_0 *sos_hdr; + const struct psp_firmware_header_v1_0 *asd_hdr; + const struct ta_firmware_header_v1_0 *ta_hdr; DRM_DEBUG("\n"); @@ -119,26 +69,66 @@ static int psp_v11_0_init_microcode(struct psp_context *psp) if (err) goto out; - hdr = (const struct psp_firmware_header_v1_0 *)adev->psp.sos_fw->data; - adev->psp.sos_fw_version = le32_to_cpu(hdr->header.ucode_version); - adev->psp.sos_feature_version = le32_to_cpu(hdr->ucode_feature_version); - adev->psp.sos_bin_size = le32_to_cpu(hdr->sos_size_bytes); - adev->psp.sys_bin_size = le32_to_cpu(hdr->header.ucode_size_bytes) - - le32_to_cpu(hdr->sos_size_bytes); - adev->psp.sys_start_addr = (uint8_t *)hdr + - le32_to_cpu(hdr->header.ucode_array_offset_bytes); + sos_hdr = (const struct psp_firmware_header_v1_0 *)adev->psp.sos_fw->data; + adev->psp.sos_fw_version = le32_to_cpu(sos_hdr->header.ucode_version); + adev->psp.sos_feature_version = le32_to_cpu(sos_hdr->ucode_feature_version); + adev->psp.sos_bin_size = le32_to_cpu(sos_hdr->sos_size_bytes); + adev->psp.sys_bin_size = le32_to_cpu(sos_hdr->header.ucode_size_bytes) - + le32_to_cpu(sos_hdr->sos_size_bytes); + adev->psp.sys_start_addr = (uint8_t *)sos_hdr + + le32_to_cpu(sos_hdr->header.ucode_array_offset_bytes); adev->psp.sos_start_addr = (uint8_t *)adev->psp.sys_start_addr + - le32_to_cpu(hdr->sos_offset_bytes); - return 0; -out: + le32_to_cpu(sos_hdr->sos_offset_bytes); + + snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_asd.bin", chip_name); + err = request_firmware(&adev->psp.asd_fw, fw_name, adev->dev); + if (err) + goto out1; + + err = amdgpu_ucode_validate(adev->psp.asd_fw); + if (err) + goto out1; + + asd_hdr = (const struct psp_firmware_header_v1_0 *)adev->psp.asd_fw->data; + adev->psp.asd_fw_version = le32_to_cpu(asd_hdr->header.ucode_version); + adev->psp.asd_feature_version = le32_to_cpu(asd_hdr->ucode_feature_version); + adev->psp.asd_ucode_size = le32_to_cpu(asd_hdr->header.ucode_size_bytes); + adev->psp.asd_start_addr = (uint8_t *)asd_hdr + + le32_to_cpu(asd_hdr->header.ucode_array_offset_bytes); + + 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) { - dev_err(adev->dev, - "psp v11.0: Failed to load firmware \"%s\"\n", - fw_name); - release_firmware(adev->psp.sos_fw); - adev->psp.sos_fw = NULL; + release_firmware(adev->psp.ta_fw); + adev->psp.ta_fw = NULL; + dev_info(adev->dev, + "psp v11.0: Failed to load firmware \"%s\"\n", fw_name); + } else { + err = amdgpu_ucode_validate(adev->psp.ta_fw); + if (err) + goto out2; + + ta_hdr = (const struct ta_firmware_header_v1_0 *)adev->psp.ta_fw->data; + adev->psp.ta_xgmi_ucode_version = le32_to_cpu(ta_hdr->ta_xgmi_ucode_version); + adev->psp.ta_xgmi_ucode_size = le32_to_cpu(ta_hdr->ta_xgmi_size_bytes); + adev->psp.ta_xgmi_start_addr = (uint8_t *)ta_hdr + + le32_to_cpu(ta_hdr->header.ucode_array_offset_bytes); } + return 0; + +out2: + release_firmware(adev->psp.ta_fw); + adev->psp.ta_fw = NULL; +out1: + release_firmware(adev->psp.asd_fw); + adev->psp.asd_fw = NULL; +out: + dev_err(adev->dev, + "psp v11.0: Failed to load firmware \"%s\"\n", fw_name); + release_firmware(adev->psp.sos_fw); + adev->psp.sos_fw = NULL; + return err; } @@ -153,8 +143,11 @@ static int psp_v11_0_bootloader_load_sysdrv(struct psp_context *psp) * are already been loaded. */ sol_reg = RREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_81); - if (sol_reg) + if (sol_reg) { + psp->sos_fw_version = RREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_58); + printk("sos fw version = 0x%x.\n", psp->sos_fw_version); return 0; + } /* 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, mmMP0_SMN_C2PMSG_35), @@ -167,7 +160,7 @@ static int psp_v11_0_bootloader_load_sysdrv(struct psp_context *psp) /* Copy PSP System Driver binary to memory */ memcpy(psp->fw_pri_buf, psp->sys_start_addr, psp->sys_bin_size); - /* Provide the sys driver to bootrom */ + /* Provide the sys driver to bootloader */ WREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_36, (uint32_t)(psp->fw_pri_mc_addr >> 20)); psp_gfxdrv_command_reg = 1 << 16; @@ -208,7 +201,7 @@ static int psp_v11_0_bootloader_load_sos(struct psp_context *psp) /* Copy Secure OS binary to PSP memory */ memcpy(psp->fw_pri_buf, psp->sos_start_addr, psp->sos_bin_size); - /* Provide the PSP secure OS to bootrom */ + /* Provide the PSP secure OS to bootloader */ WREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_36, (uint32_t)(psp->fw_pri_mc_addr >> 20)); psp_gfxdrv_command_reg = 2 << 16; @@ -224,26 +217,6 @@ static int psp_v11_0_bootloader_load_sos(struct psp_context *psp) return ret; } -static int psp_v11_0_prep_cmd_buf(struct amdgpu_firmware_info *ucode, - struct psp_gfx_cmd_resp *cmd) -{ - int ret; - uint64_t fw_mem_mc_addr = ucode->mc_addr; - - memset(cmd, 0, sizeof(struct psp_gfx_cmd_resp)); - - cmd->cmd_id = GFX_CMD_ID_LOAD_IP_FW; - cmd->cmd.cmd_load_ip_fw.fw_phy_addr_lo = lower_32_bits(fw_mem_mc_addr); - cmd->cmd.cmd_load_ip_fw.fw_phy_addr_hi = upper_32_bits(fw_mem_mc_addr); - cmd->cmd.cmd_load_ip_fw.fw_size = ucode->ucode_size; - - ret = psp_v11_0_get_fw_type(ucode, &cmd->cmd.cmd_load_ip_fw.fw_type); - if (ret) - DRM_ERROR("Unknown firmware type\n"); - - return ret; -} - static int psp_v11_0_ring_init(struct psp_context *psp, enum psp_ring_type ring_type) { @@ -270,6 +243,13 @@ static int psp_v11_0_ring_init(struct psp_context *psp, return 0; } +static bool psp_v11_0_support_vmr_ring(struct psp_context *psp) +{ + if (amdgpu_sriov_vf(psp->adev) && psp->sos_fw_version > 0x80045) + return true; + return false; +} + static int psp_v11_0_ring_create(struct psp_context *psp, enum psp_ring_type ring_type) { @@ -278,26 +258,47 @@ static int psp_v11_0_ring_create(struct psp_context *psp, struct psp_ring *ring = &psp->km_ring; struct amdgpu_device *adev = psp->adev; - /* Write low address of the ring to C2PMSG_69 */ - psp_ring_reg = lower_32_bits(ring->ring_mem_mc_addr); - WREG32_SOC15(MP0, 0, mmMP0_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, mmMP0_SMN_C2PMSG_70, psp_ring_reg); - /* Write size of ring to C2PMSG_71 */ - psp_ring_reg = ring->ring_size; - WREG32_SOC15(MP0, 0, mmMP0_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, mmMP0_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, mmMP0_SMN_C2PMSG_64), - 0x80000000, 0x8000FFFF, false); + if (psp_v11_0_support_vmr_ring(psp)) { + /* Write low address of the ring to C2PMSG_102 */ + psp_ring_reg = lower_32_bits(ring->ring_mem_mc_addr); + WREG32_SOC15(MP0, 0, mmMP0_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, mmMP0_SMN_C2PMSG_103, psp_ring_reg); + + /* Write the ring initialization command to C2PMSG_101 */ + WREG32_SOC15(MP0, 0, mmMP0_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, mmMP0_SMN_C2PMSG_101), + 0x80000000, 0x8000FFFF, false); + + } else { + /* Write low address of the ring to C2PMSG_69 */ + psp_ring_reg = lower_32_bits(ring->ring_mem_mc_addr); + WREG32_SOC15(MP0, 0, mmMP0_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, mmMP0_SMN_C2PMSG_70, psp_ring_reg); + /* Write size of ring to C2PMSG_71 */ + psp_ring_reg = ring->ring_size; + WREG32_SOC15(MP0, 0, mmMP0_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, mmMP0_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, mmMP0_SMN_C2PMSG_64), + 0x80000000, 0x8000FFFF, false); + } return ret; } @@ -308,15 +309,24 @@ static int psp_v11_0_ring_stop(struct psp_context *psp, int ret = 0; struct amdgpu_device *adev = psp->adev; - /* Write the ring destroy command to C2PMSG_64 */ - WREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_64, GFX_CTRL_CMD_ID_DESTROY_RINGS); + /* Write the ring destroy command*/ + if (psp_v11_0_support_vmr_ring(psp)) + WREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_101, + GFX_CTRL_CMD_ID_DESTROY_GPCOM_RING); + else + WREG32_SOC15(MP0, 0, mmMP0_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) in C2PMSG_64 */ - ret = psp_wait_for(psp, SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_64), - 0x80000000, 0x80000000, false); + /* Wait for response flag (bit 31) */ + if (psp_v11_0_support_vmr_ring(psp)) + ret = psp_wait_for(psp, SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_101), + 0x80000000, 0x80000000, false); + else + ret = psp_wait_for(psp, SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_64), + 0x80000000, 0x80000000, false); return ret; } @@ -355,7 +365,10 @@ static int psp_v11_0_cmd_submit(struct psp_context *psp, uint32_t rb_frame_size_dw = sizeof(struct psp_gfx_rb_frame) / 4; /* KM (GPCOM) prepare write pointer */ - psp_write_ptr_reg = RREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_67); + if (psp_v11_0_support_vmr_ring(psp)) + psp_write_ptr_reg = RREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_102); + else + psp_write_ptr_reg = RREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_67); /* Update KM RB frame pointer to new frame */ /* write_frame ptr increments by size of rb_frame in bytes */ @@ -384,7 +397,11 @@ static int psp_v11_0_cmd_submit(struct psp_context *psp, /* Update the write Pointer in DWORDs */ psp_write_ptr_reg = (psp_write_ptr_reg + rb_frame_size_dw) % ring_size_dw; - WREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_67, psp_write_ptr_reg); + if (psp_v11_0_support_vmr_ring(psp)) { + WREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_102, psp_write_ptr_reg); + WREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_101, GFX_CTRL_CMD_ID_CONSUME_CMD); + } else + WREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_67, psp_write_ptr_reg); return 0; } @@ -529,7 +546,7 @@ static int psp_v11_0_mode1_reset(struct psp_context *psp) /*send the mode 1 reset command*/ WREG32(offset, GFX_CTRL_CMD_ID_MODE1_RST); - mdelay(1000); + msleep(500); offset = SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_33); @@ -552,31 +569,120 @@ static int psp_v11_0_mode1_reset(struct psp_context *psp) static int psp_v11_0_xgmi_get_topology_info(struct psp_context *psp, int number_devices, struct psp_xgmi_topology_info *topology) { + struct ta_xgmi_shared_memory *xgmi_cmd; + struct ta_xgmi_cmd_get_topology_info_input *topology_info_input; + struct ta_xgmi_cmd_get_topology_info_output *topology_info_output; + int i; + int ret; + + if (!topology || topology->num_nodes > TA_XGMI__MAX_CONNECTED_NODES) + return -EINVAL; + + xgmi_cmd = (struct ta_xgmi_shared_memory*)psp->xgmi_context.xgmi_shared_buf; + memset(xgmi_cmd, 0, sizeof(struct ta_xgmi_shared_memory)); + + /* Fill in the shared memory with topology information as input */ + topology_info_input = &xgmi_cmd->xgmi_in_message.get_topology_info; + xgmi_cmd->cmd_id = TA_COMMAND_XGMI__GET_GET_TOPOLOGY_INFO; + topology_info_input->num_nodes = number_devices; + + for (i = 0; i < topology_info_input->num_nodes; i++) { + topology_info_input->nodes[i].node_id = topology->nodes[i].node_id; + topology_info_input->nodes[i].num_hops = topology->nodes[i].num_hops; + topology_info_input->nodes[i].is_sharing_enabled = topology->nodes[i].is_sharing_enabled; + topology_info_input->nodes[i].sdma_engine = topology->nodes[i].sdma_engine; + } + + /* Invoke xgmi ta to get the topology information */ + ret = psp_xgmi_invoke(psp, TA_COMMAND_XGMI__GET_GET_TOPOLOGY_INFO); + if (ret) + return ret; + + /* Read the output topology information from the shared memory */ + topology_info_output = &xgmi_cmd->xgmi_out_message.get_topology_info; + topology->num_nodes = xgmi_cmd->xgmi_out_message.get_topology_info.num_nodes; + for (i = 0; i < topology->num_nodes; i++) { + topology->nodes[i].node_id = topology_info_output->nodes[i].node_id; + topology->nodes[i].num_hops = topology_info_output->nodes[i].num_hops; + topology->nodes[i].is_sharing_enabled = topology_info_output->nodes[i].is_sharing_enabled; + topology->nodes[i].sdma_engine = topology_info_output->nodes[i].sdma_engine; + } + return 0; } static int psp_v11_0_xgmi_set_topology_info(struct psp_context *psp, int number_devices, struct psp_xgmi_topology_info *topology) { + struct ta_xgmi_shared_memory *xgmi_cmd; + struct ta_xgmi_cmd_get_topology_info_input *topology_info_input; + int i; + + if (!topology || topology->num_nodes > TA_XGMI__MAX_CONNECTED_NODES) + return -EINVAL; + + xgmi_cmd = (struct ta_xgmi_shared_memory*)psp->xgmi_context.xgmi_shared_buf; + memset(xgmi_cmd, 0, sizeof(struct ta_xgmi_shared_memory)); + + topology_info_input = &xgmi_cmd->xgmi_in_message.get_topology_info; + xgmi_cmd->cmd_id = TA_COMMAND_XGMI__SET_TOPOLOGY_INFO; + topology_info_input->num_nodes = number_devices; + + for (i = 0; i < topology_info_input->num_nodes; i++) { + topology_info_input->nodes[i].node_id = topology->nodes[i].node_id; + topology_info_input->nodes[i].num_hops = topology->nodes[i].num_hops; + topology_info_input->nodes[i].is_sharing_enabled = topology->nodes[i].is_sharing_enabled; + topology_info_input->nodes[i].sdma_engine = topology->nodes[i].sdma_engine; + } + + /* Invoke xgmi ta to set topology information */ + return psp_xgmi_invoke(psp, TA_COMMAND_XGMI__SET_TOPOLOGY_INFO); +} + +static int psp_v11_0_xgmi_get_hive_id(struct psp_context *psp, uint64_t *hive_id) +{ + struct ta_xgmi_shared_memory *xgmi_cmd; + int ret; + + xgmi_cmd = (struct ta_xgmi_shared_memory*)psp->xgmi_context.xgmi_shared_buf; + memset(xgmi_cmd, 0, sizeof(struct ta_xgmi_shared_memory)); + + xgmi_cmd->cmd_id = TA_COMMAND_XGMI__GET_HIVE_ID; + + /* Invoke xgmi ta to get hive id */ + ret = psp_xgmi_invoke(psp, xgmi_cmd->cmd_id); + if (ret) + return ret; + + *hive_id = xgmi_cmd->xgmi_out_message.get_hive_id.hive_id; + return 0; } -static u64 psp_v11_0_xgmi_get_hive_id(struct psp_context *psp) +static int psp_v11_0_xgmi_get_node_id(struct psp_context *psp, uint64_t *node_id) { - u64 hive_id = 0; + struct ta_xgmi_shared_memory *xgmi_cmd; + int ret; - /* Remove me when we can get correct hive_id through PSP */ - if (psp->adev->gmc.xgmi.num_physical_nodes) - hive_id = 0x123456789abcdef; + xgmi_cmd = (struct ta_xgmi_shared_memory*)psp->xgmi_context.xgmi_shared_buf; + memset(xgmi_cmd, 0, sizeof(struct ta_xgmi_shared_memory)); - return hive_id; + xgmi_cmd->cmd_id = TA_COMMAND_XGMI__GET_NODE_ID; + + /* Invoke xgmi ta to get the node id */ + ret = psp_xgmi_invoke(psp, xgmi_cmd->cmd_id); + if (ret) + return ret; + + *node_id = xgmi_cmd->xgmi_out_message.get_node_id.node_id; + + return 0; } static const struct psp_funcs psp_v11_0_funcs = { .init_microcode = psp_v11_0_init_microcode, .bootloader_load_sysdrv = psp_v11_0_bootloader_load_sysdrv, .bootloader_load_sos = psp_v11_0_bootloader_load_sos, - .prep_cmd_buf = psp_v11_0_prep_cmd_buf, .ring_init = psp_v11_0_ring_init, .ring_create = psp_v11_0_ring_create, .ring_stop = psp_v11_0_ring_stop, @@ -587,6 +693,8 @@ static const struct psp_funcs psp_v11_0_funcs = { .xgmi_get_topology_info = psp_v11_0_xgmi_get_topology_info, .xgmi_set_topology_info = psp_v11_0_xgmi_set_topology_info, .xgmi_get_hive_id = psp_v11_0_xgmi_get_hive_id, + .xgmi_get_node_id = psp_v11_0_xgmi_get_node_id, + .support_vmr_ring = psp_v11_0_support_vmr_ring, }; void psp_v11_0_set_psp_funcs(struct psp_context *psp) diff --git a/drivers/gpu/drm/amd/amdgpu/psp_v3_1.c b/drivers/gpu/drm/amd/amdgpu/psp_v3_1.c index e1ebf770c303..0487e3a4e9e7 100644 --- a/drivers/gpu/drm/amd/amdgpu/psp_v3_1.c +++ b/drivers/gpu/drm/amd/amdgpu/psp_v3_1.c @@ -47,57 +47,6 @@ MODULE_FIRMWARE("amdgpu/vega12_asd.bin"); static uint32_t sos_old_versions[] = {1517616, 1510592, 1448594, 1446554}; -static int -psp_v3_1_get_fw_type(struct amdgpu_firmware_info *ucode, enum psp_gfx_fw_type *type) -{ - switch(ucode->ucode_id) { - case AMDGPU_UCODE_ID_SDMA0: - *type = GFX_FW_TYPE_SDMA0; - break; - case AMDGPU_UCODE_ID_SDMA1: - *type = GFX_FW_TYPE_SDMA1; - break; - case AMDGPU_UCODE_ID_CP_CE: - *type = GFX_FW_TYPE_CP_CE; - break; - case AMDGPU_UCODE_ID_CP_PFP: - *type = GFX_FW_TYPE_CP_PFP; - break; - case AMDGPU_UCODE_ID_CP_ME: - *type = GFX_FW_TYPE_CP_ME; - break; - case AMDGPU_UCODE_ID_CP_MEC1: - *type = GFX_FW_TYPE_CP_MEC; - break; - case AMDGPU_UCODE_ID_CP_MEC1_JT: - *type = GFX_FW_TYPE_CP_MEC_ME1; - break; - case AMDGPU_UCODE_ID_CP_MEC2: - *type = GFX_FW_TYPE_CP_MEC; - break; - case AMDGPU_UCODE_ID_CP_MEC2_JT: - *type = GFX_FW_TYPE_CP_MEC_ME2; - break; - case AMDGPU_UCODE_ID_RLC_G: - *type = GFX_FW_TYPE_RLC_G; - break; - case AMDGPU_UCODE_ID_SMC: - *type = GFX_FW_TYPE_SMU; - break; - case AMDGPU_UCODE_ID_UVD: - *type = GFX_FW_TYPE_UVD; - break; - case AMDGPU_UCODE_ID_VCE: - *type = GFX_FW_TYPE_VCE; - break; - case AMDGPU_UCODE_ID_MAXIMUM: - default: - return -EINVAL; - } - - return 0; -} - static int psp_v3_1_init_microcode(struct psp_context *psp) { struct amdgpu_device *adev = psp->adev; @@ -194,7 +143,7 @@ static int psp_v3_1_bootloader_load_sysdrv(struct psp_context *psp) /* Copy PSP System Driver binary to memory */ memcpy(psp->fw_pri_buf, psp->sys_start_addr, psp->sys_bin_size); - /* Provide the sys driver to bootrom */ + /* Provide the sys driver to bootloader */ WREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_36, (uint32_t)(psp->fw_pri_mc_addr >> 20)); psp_gfxdrv_command_reg = 1 << 16; @@ -240,8 +189,11 @@ static int psp_v3_1_bootloader_load_sos(struct psp_context *psp) * are already been loaded. */ sol_reg = RREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_81); - if (sol_reg) + if (sol_reg) { + psp->sos_fw_version = RREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_58); + printk("sos fw version = 0x%x.\n", psp->sos_fw_version); return 0; + } /* 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, mmMP0_SMN_C2PMSG_35), @@ -254,7 +206,7 @@ static int psp_v3_1_bootloader_load_sos(struct psp_context *psp) /* Copy Secure OS binary to PSP memory */ memcpy(psp->fw_pri_buf, psp->sos_start_addr, psp->sos_bin_size); - /* Provide the PSP secure OS to bootrom */ + /* Provide the PSP secure OS to bootloader */ WREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_36, (uint32_t)(psp->fw_pri_mc_addr >> 20)); psp_gfxdrv_command_reg = 2 << 16; @@ -274,26 +226,6 @@ static int psp_v3_1_bootloader_load_sos(struct psp_context *psp) return ret; } -static int psp_v3_1_prep_cmd_buf(struct amdgpu_firmware_info *ucode, - struct psp_gfx_cmd_resp *cmd) -{ - int ret; - uint64_t fw_mem_mc_addr = ucode->mc_addr; - - memset(cmd, 0, sizeof(struct psp_gfx_cmd_resp)); - - cmd->cmd_id = GFX_CMD_ID_LOAD_IP_FW; - cmd->cmd.cmd_load_ip_fw.fw_phy_addr_lo = lower_32_bits(fw_mem_mc_addr); - cmd->cmd.cmd_load_ip_fw.fw_phy_addr_hi = upper_32_bits(fw_mem_mc_addr); - cmd->cmd.cmd_load_ip_fw.fw_size = ucode->ucode_size; - - ret = psp_v3_1_get_fw_type(ucode, &cmd->cmd.cmd_load_ip_fw.fw_type); - if (ret) - DRM_ERROR("Unknown firmware type\n"); - - return ret; -} - static int psp_v3_1_ring_init(struct psp_context *psp, enum psp_ring_type ring_type) { @@ -356,12 +288,9 @@ static int psp_v3_1_ring_stop(struct psp_context *psp, enum psp_ring_type ring_type) { int ret = 0; - struct psp_ring *ring; unsigned int psp_ring_reg = 0; struct amdgpu_device *adev = psp->adev; - ring = &psp->km_ring; - /* Write the ring destroy command to C2PMSG_64 */ psp_ring_reg = 3 << 16; WREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_64, psp_ring_reg); @@ -571,9 +500,7 @@ static bool psp_v3_1_smu_reload_quirk(struct psp_context *psp) struct amdgpu_device *adev = psp->adev; uint32_t reg; - reg = smnMP1_FIRMWARE_FLAGS | 0x03b00000; - WREG32_SOC15(NBIO, 0, mmPCIE_INDEX2, reg); - reg = RREG32_SOC15(NBIO, 0, mmPCIE_DATA2); + reg = RREG32_PCIE(smnMP1_FIRMWARE_FLAGS | 0x03b00000); return (reg & MP1_FIRMWARE_FLAGS__INTERRUPTS_ENABLED_MASK) ? true : false; } @@ -593,9 +520,9 @@ static int psp_v3_1_mode1_reset(struct psp_context *psp) } /*send the mode 1 reset command*/ - WREG32(offset, 0x70000); + WREG32(offset, GFX_CTRL_CMD_ID_MODE1_RST); - mdelay(1000); + msleep(500); offset = SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_33); @@ -615,7 +542,6 @@ static const struct psp_funcs psp_v3_1_funcs = { .init_microcode = psp_v3_1_init_microcode, .bootloader_load_sysdrv = psp_v3_1_bootloader_load_sysdrv, .bootloader_load_sos = psp_v3_1_bootloader_load_sos, - .prep_cmd_buf = psp_v3_1_prep_cmd_buf, .ring_init = psp_v3_1_ring_init, .ring_create = psp_v3_1_ring_create, .ring_stop = psp_v3_1_ring_stop, diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c b/drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c index 2d4770e173dd..cca3552b36ed 100644 --- a/drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c +++ b/drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c @@ -225,7 +225,7 @@ static void sdma_v2_4_ring_set_wptr(struct amdgpu_ring *ring) static void sdma_v2_4_ring_insert_nop(struct amdgpu_ring *ring, uint32_t count) { - struct amdgpu_sdma_instance *sdma = amdgpu_get_sdma_instance(ring); + struct amdgpu_sdma_instance *sdma = amdgpu_sdma_get_instance_from_ring(ring); int i; for (i = 0; i < count; i++) @@ -245,9 +245,12 @@ static void sdma_v2_4_ring_insert_nop(struct amdgpu_ring *ring, uint32_t count) * Schedule an IB in the DMA ring (VI). */ static void sdma_v2_4_ring_emit_ib(struct amdgpu_ring *ring, + struct amdgpu_job *job, struct amdgpu_ib *ib, - unsigned vmid, bool ctx_switch) + uint32_t flags) { + unsigned vmid = AMDGPU_JOB_GET_VMID(job); + /* IB packet must end on a 8 DW boundary */ sdma_v2_4_ring_insert_nop(ring, (10 - (lower_32_bits(ring->wptr) & 7)) % 8); @@ -349,8 +352,8 @@ static void sdma_v2_4_gfx_stop(struct amdgpu_device *adev) ib_cntl = REG_SET_FIELD(ib_cntl, SDMA0_GFX_IB_CNTL, IB_ENABLE, 0); WREG32(mmSDMA0_GFX_IB_CNTL + sdma_offsets[i], ib_cntl); } - sdma0->ready = false; - sdma1->ready = false; + sdma0->sched.ready = false; + sdma1->sched.ready = false; } /** @@ -471,17 +474,15 @@ static int sdma_v2_4_gfx_resume(struct amdgpu_device *adev) /* enable DMA IBs */ WREG32(mmSDMA0_GFX_IB_CNTL + sdma_offsets[i], ib_cntl); - ring->ready = true; + ring->sched.ready = true; } sdma_v2_4_enable(adev, true); for (i = 0; i < adev->sdma.num_instances; i++) { ring = &adev->sdma.instance[i].ring; - r = amdgpu_ring_test_ring(ring); - if (r) { - ring->ready = false; + r = amdgpu_ring_test_helper(ring); + if (r) return r; - } if (adev->mman.buffer_funcs_ring == ring) amdgpu_ttm_set_buffer_funcs_status(adev, true); @@ -550,21 +551,16 @@ static int sdma_v2_4_ring_test_ring(struct amdgpu_ring *ring) u64 gpu_addr; r = amdgpu_device_wb_get(adev, &index); - if (r) { - dev_err(adev->dev, "(%d) failed to allocate wb slot\n", r); + if (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 (r) { - DRM_ERROR("amdgpu: dma failed to lock ring %d (%d).\n", ring->idx, r); - amdgpu_device_wb_free(adev, index); - return r; - } + if (r) + goto error_free_wb; amdgpu_ring_write(ring, SDMA_PKT_HEADER_OP(SDMA_OP_WRITE) | SDMA_PKT_HEADER_SUB_OP(SDMA_SUBOP_WRITE_LINEAR)); @@ -581,15 +577,11 @@ static int sdma_v2_4_ring_test_ring(struct amdgpu_ring *ring) DRM_UDELAY(1); } - if (i < adev->usec_timeout) { - DRM_DEBUG("ring test on %d succeeded in %d usecs\n", ring->idx, i); - } else { - DRM_ERROR("amdgpu: ring %d test failed (0x%08X)\n", - ring->idx, tmp); - r = -EINVAL; - } - amdgpu_device_wb_free(adev, index); + if (i >= adev->usec_timeout) + r = -ETIMEDOUT; +error_free_wb: + amdgpu_device_wb_free(adev, index); return r; } @@ -612,20 +604,16 @@ static int sdma_v2_4_ring_test_ib(struct amdgpu_ring *ring, long timeout) long r; r = amdgpu_device_wb_get(adev, &index); - if (r) { - dev_err(adev->dev, "(%ld) failed to allocate wb slot\n", r); + if (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, &ib); - if (r) { - DRM_ERROR("amdgpu: failed to get ib (%ld).\n", r); + if (r) goto err0; - } ib.ptr[0] = SDMA_PKT_HEADER_OP(SDMA_OP_WRITE) | SDMA_PKT_HEADER_SUB_OP(SDMA_SUBOP_WRITE_LINEAR); @@ -644,21 +632,16 @@ static int sdma_v2_4_ring_test_ib(struct amdgpu_ring *ring, long timeout) 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; } tmp = le32_to_cpu(adev->wb.wb[index]); - if (tmp == 0xDEADBEEF) { - DRM_DEBUG("ib test on ring %d succeeded\n", ring->idx); + if (tmp == 0xDEADBEEF) r = 0; - } else { - DRM_ERROR("amdgpu: ib test failed (0x%08X)\n", tmp); + else r = -EINVAL; - } err1: amdgpu_ib_free(adev, &ib, NULL); @@ -760,7 +743,7 @@ static void sdma_v2_4_vm_set_pte_pde(struct amdgpu_ib *ib, uint64_t pe, */ static void sdma_v2_4_ring_pad_ib(struct amdgpu_ring *ring, struct amdgpu_ib *ib) { - struct amdgpu_sdma_instance *sdma = amdgpu_get_sdma_instance(ring); + struct amdgpu_sdma_instance *sdma = amdgpu_sdma_get_instance_from_ring(ring); u32 pad_count; int i; @@ -1105,8 +1088,14 @@ static int sdma_v2_4_process_illegal_inst_irq(struct amdgpu_device *adev, struct amdgpu_irq_src *source, struct amdgpu_iv_entry *entry) { + u8 instance_id, queue_id; + DRM_ERROR("Illegal instruction in SDMA command stream\n"); - schedule_work(&adev->reset_work); + instance_id = (entry->ring_id & 0x3) >> 0; + queue_id = (entry->ring_id & 0xc) >> 2; + + if (instance_id <= 1 && queue_id == 0) + drm_sched_fault(&adev->sdma.instance[instance_id].ring.sched); return 0; } diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c b/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c index 6fb3edaba0ec..0ce8331baeb2 100644 --- a/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c +++ b/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c @@ -399,7 +399,7 @@ static void sdma_v3_0_ring_set_wptr(struct amdgpu_ring *ring) static void sdma_v3_0_ring_insert_nop(struct amdgpu_ring *ring, uint32_t count) { - struct amdgpu_sdma_instance *sdma = amdgpu_get_sdma_instance(ring); + struct amdgpu_sdma_instance *sdma = amdgpu_sdma_get_instance_from_ring(ring); int i; for (i = 0; i < count; i++) @@ -419,9 +419,12 @@ static void sdma_v3_0_ring_insert_nop(struct amdgpu_ring *ring, uint32_t count) * Schedule an IB in the DMA ring (VI). */ static void sdma_v3_0_ring_emit_ib(struct amdgpu_ring *ring, + struct amdgpu_job *job, struct amdgpu_ib *ib, - unsigned vmid, bool ctx_switch) + uint32_t flags) { + unsigned vmid = AMDGPU_JOB_GET_VMID(job); + /* IB packet must end on a 8 DW boundary */ sdma_v3_0_ring_insert_nop(ring, (10 - (lower_32_bits(ring->wptr) & 7)) % 8); @@ -523,8 +526,8 @@ static void sdma_v3_0_gfx_stop(struct amdgpu_device *adev) ib_cntl = REG_SET_FIELD(ib_cntl, SDMA0_GFX_IB_CNTL, IB_ENABLE, 0); WREG32(mmSDMA0_GFX_IB_CNTL + sdma_offsets[i], ib_cntl); } - sdma0->ready = false; - sdma1->ready = false; + sdma0->sched.ready = false; + sdma1->sched.ready = false; } /** @@ -739,7 +742,7 @@ static int sdma_v3_0_gfx_resume(struct amdgpu_device *adev) /* enable DMA IBs */ WREG32(mmSDMA0_GFX_IB_CNTL + sdma_offsets[i], ib_cntl); - ring->ready = true; + ring->sched.ready = true; } /* unhalt the MEs */ @@ -749,11 +752,9 @@ 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; - r = amdgpu_ring_test_ring(ring); - if (r) { - ring->ready = false; + r = amdgpu_ring_test_helper(ring); + if (r) return r; - } if (adev->mman.buffer_funcs_ring == ring) amdgpu_ttm_set_buffer_funcs_status(adev, true); @@ -822,21 +823,16 @@ static int sdma_v3_0_ring_test_ring(struct amdgpu_ring *ring) u64 gpu_addr; r = amdgpu_device_wb_get(adev, &index); - if (r) { - dev_err(adev->dev, "(%d) failed to allocate wb slot\n", r); + if (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 (r) { - DRM_ERROR("amdgpu: dma failed to lock ring %d (%d).\n", ring->idx, r); - amdgpu_device_wb_free(adev, index); - return r; - } + if (r) + goto error_free_wb; amdgpu_ring_write(ring, SDMA_PKT_HEADER_OP(SDMA_OP_WRITE) | SDMA_PKT_HEADER_SUB_OP(SDMA_SUBOP_WRITE_LINEAR)); @@ -853,15 +849,11 @@ static int sdma_v3_0_ring_test_ring(struct amdgpu_ring *ring) DRM_UDELAY(1); } - if (i < adev->usec_timeout) { - DRM_DEBUG("ring test on %d succeeded in %d usecs\n", ring->idx, i); - } else { - DRM_ERROR("amdgpu: ring %d test failed (0x%08X)\n", - ring->idx, tmp); - r = -EINVAL; - } - amdgpu_device_wb_free(adev, index); + if (i >= adev->usec_timeout) + r = -ETIMEDOUT; +error_free_wb: + amdgpu_device_wb_free(adev, index); return r; } @@ -884,20 +876,16 @@ static int sdma_v3_0_ring_test_ib(struct amdgpu_ring *ring, long timeout) long r; r = amdgpu_device_wb_get(adev, &index); - if (r) { - dev_err(adev->dev, "(%ld) failed to allocate wb slot\n", r); + if (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, &ib); - if (r) { - DRM_ERROR("amdgpu: failed to get ib (%ld).\n", r); + if (r) goto err0; - } ib.ptr[0] = SDMA_PKT_HEADER_OP(SDMA_OP_WRITE) | SDMA_PKT_HEADER_SUB_OP(SDMA_SUBOP_WRITE_LINEAR); @@ -916,21 +904,16 @@ static int sdma_v3_0_ring_test_ib(struct amdgpu_ring *ring, long timeout) 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; } tmp = le32_to_cpu(adev->wb.wb[index]); - if (tmp == 0xDEADBEEF) { - DRM_DEBUG("ib test on ring %d succeeded\n", ring->idx); + if (tmp == 0xDEADBEEF) r = 0; - } else { - DRM_ERROR("amdgpu: ib test failed (0x%08X)\n", tmp); + else r = -EINVAL; - } err1: amdgpu_ib_free(adev, &ib, NULL); dma_fence_put(f); @@ -1031,7 +1014,7 @@ static void sdma_v3_0_vm_set_pte_pde(struct amdgpu_ib *ib, uint64_t pe, */ static void sdma_v3_0_ring_pad_ib(struct amdgpu_ring *ring, struct amdgpu_ib *ib) { - struct amdgpu_sdma_instance *sdma = amdgpu_get_sdma_instance(ring); + struct amdgpu_sdma_instance *sdma = amdgpu_sdma_get_instance_from_ring(ring); u32 pad_count; int i; @@ -1162,8 +1145,7 @@ static int sdma_v3_0_sw_init(void *handle) ring->ring_obj = NULL; if (!amdgpu_sriov_vf(adev)) { ring->use_doorbell = true; - ring->doorbell_index = (i == 0) ? - AMDGPU_DOORBELL_sDMA_ENGINE0 : AMDGPU_DOORBELL_sDMA_ENGINE1; + ring->doorbell_index = adev->doorbell_index.sdma_engine[i]; } else { ring->use_pollmem = true; } @@ -1440,8 +1422,14 @@ static int sdma_v3_0_process_illegal_inst_irq(struct amdgpu_device *adev, struct amdgpu_irq_src *source, struct amdgpu_iv_entry *entry) { + u8 instance_id, queue_id; + DRM_ERROR("Illegal instruction in SDMA command stream\n"); - schedule_work(&adev->reset_work); + instance_id = (entry->ring_id & 0x3) >> 0; + queue_id = (entry->ring_id & 0xc) >> 2; + + if (instance_id <= 1 && queue_id == 0) + drm_sched_fault(&adev->sdma.instance[instance_id].ring.sched); return 0; } diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c b/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c index 7a8c9172d30a..c816e55d43a9 100644 --- a/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c +++ b/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c @@ -54,6 +54,11 @@ MODULE_FIRMWARE("amdgpu/raven2_sdma.bin"); #define SDMA0_POWER_CNTL__ON_OFF_CONDITION_HOLD_TIME_MASK 0x000000F8L #define SDMA0_POWER_CNTL__ON_OFF_STATUS_DURATION_TIME_MASK 0xFC000000L +#define WREG32_SDMA(instance, offset, value) \ + WREG32(sdma_v4_0_get_reg_offset(adev, (instance), (offset)), value) +#define RREG32_SDMA(instance, offset) \ + RREG32(sdma_v4_0_get_reg_offset(adev, (instance), (offset))) + static void sdma_v4_0_set_ring_funcs(struct amdgpu_device *adev); static void sdma_v4_0_set_buffer_funcs(struct amdgpu_device *adev); static void sdma_v4_0_set_vm_pte_funcs(struct amdgpu_device *adev); @@ -73,7 +78,6 @@ static const struct soc15_reg_golden golden_settings_sdma_4[] = { SOC15_REG_GOLDEN_VALUE(SDMA0, 0, mmSDMA0_RLC1_RB_WPTR_POLL_CNTL, 0x0000fff0, 0x00403000), SOC15_REG_GOLDEN_VALUE(SDMA0, 0, mmSDMA0_UTCL1_PAGE, 0x000003ff, 0x000003c0), SOC15_REG_GOLDEN_VALUE(SDMA0, 0, mmSDMA0_UTCL1_WATERMK, 0xfc000000, 0x00000000), - SOC15_REG_GOLDEN_VALUE(SDMA1, 0, mmSDMA1_CHICKEN_BITS, 0xfe931f07, 0x02831f07), SOC15_REG_GOLDEN_VALUE(SDMA1, 0, mmSDMA1_CLK_CTRL, 0xffffffff, 0x3f000100), SOC15_REG_GOLDEN_VALUE(SDMA1, 0, mmSDMA1_GFX_IB_CNTL, 0x800f0100, 0x00000100), SOC15_REG_GOLDEN_VALUE(SDMA1, 0, mmSDMA1_GFX_RB_WPTR_POLL_CNTL, 0x0000fff0, 0x00403000), @@ -91,6 +95,7 @@ static const struct soc15_reg_golden golden_settings_sdma_4[] = { static const struct soc15_reg_golden golden_settings_sdma_vg10[] = { SOC15_REG_GOLDEN_VALUE(SDMA0, 0, mmSDMA0_GB_ADDR_CONFIG, 0x0018773f, 0x00104002), SOC15_REG_GOLDEN_VALUE(SDMA0, 0, mmSDMA0_GB_ADDR_CONFIG_READ, 0x0018773f, 0x00104002), + SOC15_REG_GOLDEN_VALUE(SDMA1, 0, mmSDMA1_CHICKEN_BITS, 0xfe931f07, 0x02831d07), SOC15_REG_GOLDEN_VALUE(SDMA1, 0, mmSDMA1_GB_ADDR_CONFIG, 0x0018773f, 0x00104002), SOC15_REG_GOLDEN_VALUE(SDMA1, 0, mmSDMA1_GB_ADDR_CONFIG_READ, 0x0018773f, 0x00104002) }; @@ -98,6 +103,7 @@ static const struct soc15_reg_golden golden_settings_sdma_vg10[] = { static const struct soc15_reg_golden golden_settings_sdma_vg12[] = { SOC15_REG_GOLDEN_VALUE(SDMA0, 0, mmSDMA0_GB_ADDR_CONFIG, 0x0018773f, 0x00104001), SOC15_REG_GOLDEN_VALUE(SDMA0, 0, mmSDMA0_GB_ADDR_CONFIG_READ, 0x0018773f, 0x00104001), + SOC15_REG_GOLDEN_VALUE(SDMA1, 0, mmSDMA1_CHICKEN_BITS, 0xfe931f07, 0x02831d07), SOC15_REG_GOLDEN_VALUE(SDMA1, 0, mmSDMA1_GB_ADDR_CONFIG, 0x0018773f, 0x00104001), SOC15_REG_GOLDEN_VALUE(SDMA1, 0, mmSDMA1_GB_ADDR_CONFIG_READ, 0x0018773f, 0x00104001) }; @@ -122,7 +128,7 @@ static const struct soc15_reg_golden golden_settings_sdma0_4_2_init[] = { static const struct soc15_reg_golden golden_settings_sdma0_4_2[] = { - SOC15_REG_GOLDEN_VALUE(SDMA0, 0, mmSDMA0_CHICKEN_BITS, 0xfe931f07, 0x02831d07), + SOC15_REG_GOLDEN_VALUE(SDMA0, 0, mmSDMA0_CHICKEN_BITS, 0xfe931f07, 0x02831f07), SOC15_REG_GOLDEN_VALUE(SDMA0, 0, mmSDMA0_CLK_CTRL, 0xffffffff, 0x3f000100), SOC15_REG_GOLDEN_VALUE(SDMA0, 0, mmSDMA0_GB_ADDR_CONFIG, 0x0000773f, 0x00004002), SOC15_REG_GOLDEN_VALUE(SDMA0, 0, mmSDMA0_GB_ADDR_CONFIG_READ, 0x0000773f, 0x00004002), @@ -152,7 +158,7 @@ static const struct soc15_reg_golden golden_settings_sdma0_4_2[] = }; static const struct soc15_reg_golden golden_settings_sdma1_4_2[] = { - SOC15_REG_GOLDEN_VALUE(SDMA1, 0, mmSDMA1_CHICKEN_BITS, 0xfe931f07, 0x02831d07), + SOC15_REG_GOLDEN_VALUE(SDMA1, 0, mmSDMA1_CHICKEN_BITS, 0xfe931f07, 0x02831f07), SOC15_REG_GOLDEN_VALUE(SDMA1, 0, mmSDMA1_CLK_CTRL, 0xffffffff, 0x3f000100), SOC15_REG_GOLDEN_VALUE(SDMA1, 0, mmSDMA1_GB_ADDR_CONFIG, 0x0000773f, 0x00004002), SOC15_REG_GOLDEN_VALUE(SDMA1, 0, mmSDMA1_GB_ADDR_CONFIG_READ, 0x0000773f, 0x00004002), @@ -367,16 +373,11 @@ static uint64_t sdma_v4_0_ring_get_wptr(struct amdgpu_ring *ring) wptr = READ_ONCE(*((u64 *)&adev->wb.wb[ring->wptr_offs])); DRM_DEBUG("wptr/doorbell before shift == 0x%016llx\n", wptr); } else { - u32 lowbit, highbit; - - lowbit = RREG32(sdma_v4_0_get_reg_offset(adev, ring->me, mmSDMA0_GFX_RB_WPTR)) >> 2; - highbit = RREG32(sdma_v4_0_get_reg_offset(adev, ring->me, mmSDMA0_GFX_RB_WPTR_HI)) >> 2; - - DRM_DEBUG("wptr [%i]high== 0x%08x low==0x%08x\n", - ring->me, highbit, lowbit); - wptr = highbit; + wptr = RREG32_SDMA(ring->me, mmSDMA0_GFX_RB_WPTR_HI); wptr = wptr << 32; - wptr |= lowbit; + wptr |= RREG32_SDMA(ring->me, mmSDMA0_GFX_RB_WPTR); + DRM_DEBUG("wptr before shift [%i] wptr == 0x%016llx\n", + ring->me, wptr); } return wptr >> 2; @@ -417,14 +418,67 @@ static void sdma_v4_0_ring_set_wptr(struct amdgpu_ring *ring) lower_32_bits(ring->wptr << 2), ring->me, upper_32_bits(ring->wptr << 2)); - WREG32(sdma_v4_0_get_reg_offset(adev, ring->me, mmSDMA0_GFX_RB_WPTR), lower_32_bits(ring->wptr << 2)); - WREG32(sdma_v4_0_get_reg_offset(adev, ring->me, mmSDMA0_GFX_RB_WPTR_HI), upper_32_bits(ring->wptr << 2)); + WREG32_SDMA(ring->me, mmSDMA0_GFX_RB_WPTR, + lower_32_bits(ring->wptr << 2)); + WREG32_SDMA(ring->me, mmSDMA0_GFX_RB_WPTR_HI, + upper_32_bits(ring->wptr << 2)); + } +} + +/** + * sdma_v4_0_page_ring_get_wptr - get the current write pointer + * + * @ring: amdgpu ring pointer + * + * Get the current wptr from the hardware (VEGA10+). + */ +static uint64_t sdma_v4_0_page_ring_get_wptr(struct amdgpu_ring *ring) +{ + struct amdgpu_device *adev = ring->adev; + u64 wptr; + + if (ring->use_doorbell) { + /* XXX check if swapping is necessary on BE */ + wptr = READ_ONCE(*((u64 *)&adev->wb.wb[ring->wptr_offs])); + } else { + wptr = RREG32_SDMA(ring->me, mmSDMA0_PAGE_RB_WPTR_HI); + wptr = wptr << 32; + wptr |= RREG32_SDMA(ring->me, mmSDMA0_PAGE_RB_WPTR); + } + + return wptr >> 2; +} + +/** + * sdma_v4_0_ring_set_wptr - commit the write pointer + * + * @ring: amdgpu ring pointer + * + * Write the wptr back to the hardware (VEGA10+). + */ +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]; + + /* XXX check if swapping is necessary on BE */ + WRITE_ONCE(*wb, (ring->wptr << 2)); + WDOORBELL64(ring->doorbell_index, ring->wptr << 2); + } else { + uint64_t wptr = ring->wptr << 2; + + WREG32_SDMA(ring->me, mmSDMA0_PAGE_RB_WPTR, + lower_32_bits(wptr)); + WREG32_SDMA(ring->me, mmSDMA0_PAGE_RB_WPTR_HI, + upper_32_bits(wptr)); } } static void sdma_v4_0_ring_insert_nop(struct amdgpu_ring *ring, uint32_t count) { - struct amdgpu_sdma_instance *sdma = amdgpu_get_sdma_instance(ring); + struct amdgpu_sdma_instance *sdma = amdgpu_sdma_get_instance_from_ring(ring); int i; for (i = 0; i < count; i++) @@ -444,9 +498,12 @@ static void sdma_v4_0_ring_insert_nop(struct amdgpu_ring *ring, uint32_t count) * Schedule an IB in the DMA ring (VEGA10). */ static void sdma_v4_0_ring_emit_ib(struct amdgpu_ring *ring, - struct amdgpu_ib *ib, - unsigned vmid, bool ctx_switch) + struct amdgpu_job *job, + struct amdgpu_ib *ib, + uint32_t flags) { + unsigned vmid = AMDGPU_JOB_GET_VMID(job); + /* IB packet must end on a 8 DW boundary */ sdma_v4_0_ring_insert_nop(ring, (10 - (lower_32_bits(ring->wptr) & 7)) % 8); @@ -568,16 +625,16 @@ static void sdma_v4_0_gfx_stop(struct amdgpu_device *adev) amdgpu_ttm_set_buffer_funcs_status(adev, false); for (i = 0; i < adev->sdma.num_instances; i++) { - rb_cntl = RREG32(sdma_v4_0_get_reg_offset(adev, i, mmSDMA0_GFX_RB_CNTL)); + 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_v4_0_get_reg_offset(adev, i, mmSDMA0_GFX_RB_CNTL), rb_cntl); - ib_cntl = RREG32(sdma_v4_0_get_reg_offset(adev, i, mmSDMA0_GFX_IB_CNTL)); + WREG32_SDMA(i, mmSDMA0_GFX_RB_CNTL, rb_cntl); + ib_cntl = RREG32_SDMA(i, mmSDMA0_GFX_IB_CNTL); ib_cntl = REG_SET_FIELD(ib_cntl, SDMA0_GFX_IB_CNTL, IB_ENABLE, 0); - WREG32(sdma_v4_0_get_reg_offset(adev, i, mmSDMA0_GFX_IB_CNTL), ib_cntl); + WREG32_SDMA(i, mmSDMA0_GFX_IB_CNTL, ib_cntl); } - sdma0->ready = false; - sdma1->ready = false; + sdma0->sched.ready = false; + sdma1->sched.ready = false; } /** @@ -593,6 +650,39 @@ static void sdma_v4_0_rlc_stop(struct amdgpu_device *adev) } /** + * sdma_v4_0_page_stop - stop the page async dma engines + * + * @adev: amdgpu_device pointer + * + * Stop the page async dma ring buffers (VEGA10). + */ +static void sdma_v4_0_page_stop(struct amdgpu_device *adev) +{ + struct amdgpu_ring *sdma0 = &adev->sdma.instance[0].page; + struct amdgpu_ring *sdma1 = &adev->sdma.instance[1].page; + 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); + + 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); + WREG32_SDMA(i, mmSDMA0_PAGE_RB_CNTL, rb_cntl); + ib_cntl = RREG32_SDMA(i, mmSDMA0_PAGE_IB_CNTL); + ib_cntl = REG_SET_FIELD(ib_cntl, SDMA0_PAGE_IB_CNTL, + IB_ENABLE, 0); + WREG32_SDMA(i, mmSDMA0_PAGE_IB_CNTL, ib_cntl); + } + + sdma0->sched.ready = false; + sdma1->sched.ready = false; +} + +/** * sdma_v_0_ctx_switch_enable - stop the async dma engines context switch * * @adev: amdgpu_device pointer @@ -630,18 +720,15 @@ static void sdma_v4_0_ctx_switch_enable(struct amdgpu_device *adev, bool enable) } for (i = 0; i < adev->sdma.num_instances; i++) { - f32_cntl = RREG32(sdma_v4_0_get_reg_offset(adev, i, mmSDMA0_CNTL)); + f32_cntl = RREG32_SDMA(i, mmSDMA0_CNTL); f32_cntl = REG_SET_FIELD(f32_cntl, SDMA0_CNTL, AUTO_CTXSW_ENABLE, enable ? 1 : 0); if (enable && amdgpu_sdma_phase_quantum) { - WREG32(sdma_v4_0_get_reg_offset(adev, i, mmSDMA0_PHASE0_QUANTUM), - phase_quantum); - WREG32(sdma_v4_0_get_reg_offset(adev, i, mmSDMA0_PHASE1_QUANTUM), - phase_quantum); - WREG32(sdma_v4_0_get_reg_offset(adev, i, mmSDMA0_PHASE2_QUANTUM), - phase_quantum); + WREG32_SDMA(i, mmSDMA0_PHASE0_QUANTUM, phase_quantum); + WREG32_SDMA(i, mmSDMA0_PHASE1_QUANTUM, phase_quantum); + WREG32_SDMA(i, mmSDMA0_PHASE2_QUANTUM, phase_quantum); } - WREG32(sdma_v4_0_get_reg_offset(adev, i, mmSDMA0_CNTL), f32_cntl); + WREG32_SDMA(i, mmSDMA0_CNTL, f32_cntl); } } @@ -662,156 +749,213 @@ static void sdma_v4_0_enable(struct amdgpu_device *adev, bool enable) if (enable == false) { sdma_v4_0_gfx_stop(adev); sdma_v4_0_rlc_stop(adev); + if (adev->sdma.has_page_queue) + sdma_v4_0_page_stop(adev); } for (i = 0; i < adev->sdma.num_instances; i++) { - f32_cntl = RREG32(sdma_v4_0_get_reg_offset(adev, i, mmSDMA0_F32_CNTL)); + f32_cntl = RREG32_SDMA(i, mmSDMA0_F32_CNTL); f32_cntl = REG_SET_FIELD(f32_cntl, SDMA0_F32_CNTL, HALT, enable ? 0 : 1); - WREG32(sdma_v4_0_get_reg_offset(adev, i, mmSDMA0_F32_CNTL), f32_cntl); + WREG32_SDMA(i, mmSDMA0_F32_CNTL, f32_cntl); } } /** + * sdma_v4_0_rb_cntl - get parameters for rb_cntl + */ +static uint32_t sdma_v4_0_rb_cntl(struct amdgpu_ring *ring, uint32_t rb_cntl) +{ + /* Set ring buffer size in dwords */ + uint32_t rb_bufsz = order_base_2(ring->ring_size / 4); + + rb_cntl = REG_SET_FIELD(rb_cntl, SDMA0_GFX_RB_CNTL, RB_SIZE, rb_bufsz); +#ifdef __BIG_ENDIAN + rb_cntl = REG_SET_FIELD(rb_cntl, SDMA0_GFX_RB_CNTL, RB_SWAP_ENABLE, 1); + rb_cntl = REG_SET_FIELD(rb_cntl, SDMA0_GFX_RB_CNTL, + RPTR_WRITEBACK_SWAP_ENABLE, 1); +#endif + return rb_cntl; +} + +/** * sdma_v4_0_gfx_resume - setup and start the async dma engines * * @adev: amdgpu_device pointer + * @i: instance to resume * * Set up the gfx DMA ring buffers and enable them (VEGA10). * Returns 0 for success, error for failure. */ -static int sdma_v4_0_gfx_resume(struct amdgpu_device *adev) +static void sdma_v4_0_gfx_resume(struct amdgpu_device *adev, unsigned int i) { - struct amdgpu_ring *ring; + struct amdgpu_ring *ring = &adev->sdma.instance[i].ring; u32 rb_cntl, ib_cntl, wptr_poll_cntl; - u32 rb_bufsz; u32 wb_offset; 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; - wb_offset = (ring->rptr_offs * 4); + wb_offset = (ring->rptr_offs * 4); - WREG32(sdma_v4_0_get_reg_offset(adev, i, mmSDMA0_SEM_WAIT_FAIL_TIMER_CNTL), 0); + 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); - /* Set ring buffer size in dwords */ - rb_bufsz = order_base_2(ring->ring_size / 4); - rb_cntl = RREG32(sdma_v4_0_get_reg_offset(adev, i, mmSDMA0_GFX_RB_CNTL)); - rb_cntl = REG_SET_FIELD(rb_cntl, SDMA0_GFX_RB_CNTL, RB_SIZE, rb_bufsz); -#ifdef __BIG_ENDIAN - rb_cntl = REG_SET_FIELD(rb_cntl, SDMA0_GFX_RB_CNTL, RB_SWAP_ENABLE, 1); - rb_cntl = REG_SET_FIELD(rb_cntl, SDMA0_GFX_RB_CNTL, - RPTR_WRITEBACK_SWAP_ENABLE, 1); -#endif - WREG32(sdma_v4_0_get_reg_offset(adev, i, mmSDMA0_GFX_RB_CNTL), rb_cntl); + /* Initialize the ring buffer's read and write pointers */ + WREG32_SDMA(i, mmSDMA0_GFX_RB_RPTR, 0); + WREG32_SDMA(i, mmSDMA0_GFX_RB_RPTR_HI, 0); + WREG32_SDMA(i, mmSDMA0_GFX_RB_WPTR, 0); + WREG32_SDMA(i, mmSDMA0_GFX_RB_WPTR_HI, 0); - /* Initialize the ring buffer's read and write pointers */ - WREG32(sdma_v4_0_get_reg_offset(adev, i, mmSDMA0_GFX_RB_RPTR), 0); - WREG32(sdma_v4_0_get_reg_offset(adev, i, mmSDMA0_GFX_RB_RPTR_HI), 0); - WREG32(sdma_v4_0_get_reg_offset(adev, i, mmSDMA0_GFX_RB_WPTR), 0); - WREG32(sdma_v4_0_get_reg_offset(adev, i, mmSDMA0_GFX_RB_WPTR_HI), 0); + /* 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); + WREG32_SDMA(i, mmSDMA0_GFX_RB_RPTR_ADDR_LO, + lower_32_bits(adev->wb.gpu_addr + wb_offset) & 0xFFFFFFFC); - /* set the wb address whether it's enabled or not */ - WREG32(sdma_v4_0_get_reg_offset(adev, i, mmSDMA0_GFX_RB_RPTR_ADDR_HI), - upper_32_bits(adev->wb.gpu_addr + wb_offset) & 0xFFFFFFFF); - WREG32(sdma_v4_0_get_reg_offset(adev, i, mmSDMA0_GFX_RB_RPTR_ADDR_LO), - lower_32_bits(adev->wb.gpu_addr + wb_offset) & 0xFFFFFFFC); + rb_cntl = REG_SET_FIELD(rb_cntl, SDMA0_GFX_RB_CNTL, + RPTR_WRITEBACK_ENABLE, 1); - rb_cntl = REG_SET_FIELD(rb_cntl, SDMA0_GFX_RB_CNTL, RPTR_WRITEBACK_ENABLE, 1); + WREG32_SDMA(i, mmSDMA0_GFX_RB_BASE, ring->gpu_addr >> 8); + WREG32_SDMA(i, mmSDMA0_GFX_RB_BASE_HI, ring->gpu_addr >> 40); - WREG32(sdma_v4_0_get_reg_offset(adev, i, mmSDMA0_GFX_RB_BASE), ring->gpu_addr >> 8); - WREG32(sdma_v4_0_get_reg_offset(adev, i, mmSDMA0_GFX_RB_BASE_HI), ring->gpu_addr >> 40); + ring->wptr = 0; - ring->wptr = 0; + /* before programing wptr to a less value, need set minor_ptr_update first */ + WREG32_SDMA(i, mmSDMA0_GFX_MINOR_PTR_UPDATE, 1); - /* before programing wptr to a less value, need set minor_ptr_update first */ - WREG32(sdma_v4_0_get_reg_offset(adev, i, mmSDMA0_GFX_MINOR_PTR_UPDATE), 1); + doorbell = RREG32_SDMA(i, mmSDMA0_GFX_DOORBELL); + doorbell_offset = RREG32_SDMA(i, mmSDMA0_GFX_DOORBELL_OFFSET); - if (!amdgpu_sriov_vf(adev)) { /* only bare-metal use register write for wptr */ - WREG32(sdma_v4_0_get_reg_offset(adev, i, mmSDMA0_GFX_RB_WPTR), lower_32_bits(ring->wptr) << 2); - WREG32(sdma_v4_0_get_reg_offset(adev, i, mmSDMA0_GFX_RB_WPTR_HI), upper_32_bits(ring->wptr) << 2); - } - - doorbell = RREG32(sdma_v4_0_get_reg_offset(adev, i, mmSDMA0_GFX_DOORBELL)); - doorbell_offset = RREG32(sdma_v4_0_get_reg_offset(adev, i, mmSDMA0_GFX_DOORBELL_OFFSET)); - - if (ring->use_doorbell) { - doorbell = REG_SET_FIELD(doorbell, SDMA0_GFX_DOORBELL, ENABLE, 1); - doorbell_offset = REG_SET_FIELD(doorbell_offset, SDMA0_GFX_DOORBELL_OFFSET, + doorbell = REG_SET_FIELD(doorbell, SDMA0_GFX_DOORBELL, ENABLE, + ring->use_doorbell); + doorbell_offset = REG_SET_FIELD(doorbell_offset, + SDMA0_GFX_DOORBELL_OFFSET, OFFSET, ring->doorbell_index); - } else { - doorbell = REG_SET_FIELD(doorbell, SDMA0_GFX_DOORBELL, ENABLE, 0); - } - WREG32(sdma_v4_0_get_reg_offset(adev, i, mmSDMA0_GFX_DOORBELL), doorbell); - WREG32(sdma_v4_0_get_reg_offset(adev, i, mmSDMA0_GFX_DOORBELL_OFFSET), doorbell_offset); - adev->nbio_funcs->sdma_doorbell_range(adev, i, ring->use_doorbell, - ring->doorbell_index); + WREG32_SDMA(i, mmSDMA0_GFX_DOORBELL, doorbell); + WREG32_SDMA(i, mmSDMA0_GFX_DOORBELL_OFFSET, doorbell_offset); + + sdma_v4_0_ring_set_wptr(ring); + + /* set minor_ptr_update to 0 after wptr programed */ + 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); + 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, + upper_32_bits(wptr_gpu_addr)); + wptr_poll_cntl = RREG32_SDMA(i, mmSDMA0_GFX_RB_WPTR_POLL_CNTL); + wptr_poll_cntl = REG_SET_FIELD(wptr_poll_cntl, + SDMA0_GFX_RB_WPTR_POLL_CNTL, + F32_POLL_ENABLE, amdgpu_sriov_vf(adev)); + WREG32_SDMA(i, mmSDMA0_GFX_RB_WPTR_POLL_CNTL, wptr_poll_cntl); + + /* enable DMA RB */ + rb_cntl = REG_SET_FIELD(rb_cntl, SDMA0_GFX_RB_CNTL, RB_ENABLE, 1); + WREG32_SDMA(i, mmSDMA0_GFX_RB_CNTL, rb_cntl); + + ib_cntl = RREG32_SDMA(i, mmSDMA0_GFX_IB_CNTL); + ib_cntl = REG_SET_FIELD(ib_cntl, SDMA0_GFX_IB_CNTL, IB_ENABLE, 1); +#ifdef __BIG_ENDIAN + ib_cntl = REG_SET_FIELD(ib_cntl, SDMA0_GFX_IB_CNTL, IB_SWAP_ENABLE, 1); +#endif + /* enable DMA IBs */ + WREG32_SDMA(i, mmSDMA0_GFX_IB_CNTL, ib_cntl); - if (amdgpu_sriov_vf(adev)) - sdma_v4_0_ring_set_wptr(ring); + ring->sched.ready = true; +} - /* set minor_ptr_update to 0 after wptr programed */ - WREG32(sdma_v4_0_get_reg_offset(adev, i, mmSDMA0_GFX_MINOR_PTR_UPDATE), 0); +/** + * sdma_v4_0_page_resume - setup and start the async dma engines + * + * @adev: amdgpu_device pointer + * @i: instance to resume + * + * Set up the page DMA ring buffers and enable them (VEGA10). + * Returns 0 for success, error for failure. + */ +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; - /* set utc l1 enable flag always to 1 */ - temp = RREG32(sdma_v4_0_get_reg_offset(adev, i, mmSDMA0_CNTL)); - temp = REG_SET_FIELD(temp, SDMA0_CNTL, UTC_L1_ENABLE, 1); - WREG32(sdma_v4_0_get_reg_offset(adev, i, mmSDMA0_CNTL), temp); + wb_offset = (ring->rptr_offs * 4); - if (!amdgpu_sriov_vf(adev)) { - /* unhalt engine */ - temp = RREG32(sdma_v4_0_get_reg_offset(adev, i, mmSDMA0_F32_CNTL)); - temp = REG_SET_FIELD(temp, SDMA0_F32_CNTL, HALT, 0); - WREG32(sdma_v4_0_get_reg_offset(adev, i, mmSDMA0_F32_CNTL), temp); - } + 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); - /* setup the wptr shadow polling */ - wptr_gpu_addr = adev->wb.gpu_addr + (ring->wptr_offs * 4); - WREG32(sdma_v4_0_get_reg_offset(adev, i, mmSDMA0_GFX_RB_WPTR_POLL_ADDR_LO), - lower_32_bits(wptr_gpu_addr)); - WREG32(sdma_v4_0_get_reg_offset(adev, i, mmSDMA0_GFX_RB_WPTR_POLL_ADDR_HI), - upper_32_bits(wptr_gpu_addr)); - wptr_poll_cntl = RREG32(sdma_v4_0_get_reg_offset(adev, i, mmSDMA0_GFX_RB_WPTR_POLL_CNTL)); - if (amdgpu_sriov_vf(adev)) - wptr_poll_cntl = REG_SET_FIELD(wptr_poll_cntl, SDMA0_GFX_RB_WPTR_POLL_CNTL, F32_POLL_ENABLE, 1); - else - wptr_poll_cntl = REG_SET_FIELD(wptr_poll_cntl, SDMA0_GFX_RB_WPTR_POLL_CNTL, F32_POLL_ENABLE, 0); - WREG32(sdma_v4_0_get_reg_offset(adev, i, mmSDMA0_GFX_RB_WPTR_POLL_CNTL), wptr_poll_cntl); + /* Initialize the ring buffer's read and write pointers */ + WREG32_SDMA(i, mmSDMA0_PAGE_RB_RPTR, 0); + WREG32_SDMA(i, mmSDMA0_PAGE_RB_RPTR_HI, 0); + WREG32_SDMA(i, mmSDMA0_PAGE_RB_WPTR, 0); + WREG32_SDMA(i, mmSDMA0_PAGE_RB_WPTR_HI, 0); - /* enable DMA RB */ - rb_cntl = REG_SET_FIELD(rb_cntl, SDMA0_GFX_RB_CNTL, RB_ENABLE, 1); - WREG32(sdma_v4_0_get_reg_offset(adev, i, mmSDMA0_GFX_RB_CNTL), rb_cntl); + /* 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); + WREG32_SDMA(i, mmSDMA0_PAGE_RB_RPTR_ADDR_LO, + lower_32_bits(adev->wb.gpu_addr + wb_offset) & 0xFFFFFFFC); - ib_cntl = RREG32(sdma_v4_0_get_reg_offset(adev, i, mmSDMA0_GFX_IB_CNTL)); - ib_cntl = REG_SET_FIELD(ib_cntl, SDMA0_GFX_IB_CNTL, IB_ENABLE, 1); -#ifdef __BIG_ENDIAN - ib_cntl = REG_SET_FIELD(ib_cntl, SDMA0_GFX_IB_CNTL, IB_SWAP_ENABLE, 1); -#endif - /* enable DMA IBs */ - WREG32(sdma_v4_0_get_reg_offset(adev, i, mmSDMA0_GFX_IB_CNTL), ib_cntl); + rb_cntl = REG_SET_FIELD(rb_cntl, SDMA0_PAGE_RB_CNTL, + RPTR_WRITEBACK_ENABLE, 1); - ring->ready = true; + WREG32_SDMA(i, mmSDMA0_PAGE_RB_BASE, ring->gpu_addr >> 8); + WREG32_SDMA(i, mmSDMA0_PAGE_RB_BASE_HI, ring->gpu_addr >> 40); - if (amdgpu_sriov_vf(adev)) { /* bare-metal sequence doesn't need below to lines */ - sdma_v4_0_ctx_switch_enable(adev, true); - sdma_v4_0_enable(adev, true); - } + ring->wptr = 0; - r = amdgpu_ring_test_ring(ring); - if (r) { - ring->ready = false; - return r; - } + /* before programing wptr to a less value, need set minor_ptr_update first */ + WREG32_SDMA(i, mmSDMA0_PAGE_MINOR_PTR_UPDATE, 1); - if (adev->mman.buffer_funcs_ring == ring) - amdgpu_ttm_set_buffer_funcs_status(adev, true); + doorbell = RREG32_SDMA(i, mmSDMA0_PAGE_DOORBELL); + doorbell_offset = RREG32_SDMA(i, mmSDMA0_PAGE_DOORBELL_OFFSET); - } + doorbell = REG_SET_FIELD(doorbell, SDMA0_PAGE_DOORBELL, ENABLE, + ring->use_doorbell); + doorbell_offset = REG_SET_FIELD(doorbell_offset, + SDMA0_PAGE_DOORBELL_OFFSET, + OFFSET, ring->doorbell_index); + WREG32_SDMA(i, mmSDMA0_PAGE_DOORBELL, doorbell); + WREG32_SDMA(i, mmSDMA0_PAGE_DOORBELL_OFFSET, doorbell_offset); + + /* paging queue doorbell range is setup at sdma_v4_0_gfx_resume */ + sdma_v4_0_page_ring_set_wptr(ring); + + /* set minor_ptr_update to 0 after wptr programed */ + 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); + 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, + upper_32_bits(wptr_gpu_addr)); + wptr_poll_cntl = RREG32_SDMA(i, mmSDMA0_PAGE_RB_WPTR_POLL_CNTL); + wptr_poll_cntl = REG_SET_FIELD(wptr_poll_cntl, + SDMA0_PAGE_RB_WPTR_POLL_CNTL, + F32_POLL_ENABLE, amdgpu_sriov_vf(adev)); + WREG32_SDMA(i, mmSDMA0_PAGE_RB_WPTR_POLL_CNTL, wptr_poll_cntl); + + /* enable DMA RB */ + rb_cntl = REG_SET_FIELD(rb_cntl, SDMA0_PAGE_RB_CNTL, RB_ENABLE, 1); + WREG32_SDMA(i, mmSDMA0_PAGE_RB_CNTL, rb_cntl); + + ib_cntl = RREG32_SDMA(i, mmSDMA0_PAGE_IB_CNTL); + ib_cntl = REG_SET_FIELD(ib_cntl, SDMA0_PAGE_IB_CNTL, IB_ENABLE, 1); +#ifdef __BIG_ENDIAN + ib_cntl = REG_SET_FIELD(ib_cntl, SDMA0_PAGE_IB_CNTL, IB_SWAP_ENABLE, 1); +#endif + /* enable DMA IBs */ + WREG32_SDMA(i, mmSDMA0_PAGE_IB_CNTL, ib_cntl); - return 0; + ring->sched.ready = true; } static void @@ -922,12 +1066,14 @@ static int sdma_v4_0_load_microcode(struct amdgpu_device *adev) (adev->sdma.instance[i].fw->data + le32_to_cpu(hdr->header.ucode_array_offset_bytes)); - WREG32(sdma_v4_0_get_reg_offset(adev, i, mmSDMA0_UCODE_ADDR), 0); + WREG32_SDMA(i, mmSDMA0_UCODE_ADDR, 0); for (j = 0; j < fw_size; j++) - WREG32(sdma_v4_0_get_reg_offset(adev, i, mmSDMA0_UCODE_DATA), le32_to_cpup(fw_data++)); + WREG32_SDMA(i, mmSDMA0_UCODE_DATA, + le32_to_cpup(fw_data++)); - WREG32(sdma_v4_0_get_reg_offset(adev, i, mmSDMA0_UCODE_ADDR), adev->sdma.instance[i].fw_version); + WREG32_SDMA(i, mmSDMA0_UCODE_ADDR, + adev->sdma.instance[i].fw_version); } return 0; @@ -943,33 +1089,78 @@ static int sdma_v4_0_load_microcode(struct amdgpu_device *adev) */ static int sdma_v4_0_start(struct amdgpu_device *adev) { - int r = 0; + struct amdgpu_ring *ring; + int i, r; if (amdgpu_sriov_vf(adev)) { sdma_v4_0_ctx_switch_enable(adev, false); sdma_v4_0_enable(adev, false); + } else { - /* set RB registers */ - r = sdma_v4_0_gfx_resume(adev); - return r; + if (adev->firmware.load_type != AMDGPU_FW_LOAD_PSP) { + r = sdma_v4_0_load_microcode(adev); + if (r) + return r; + } + + /* unhalt the MEs */ + sdma_v4_0_enable(adev, true); + /* enable sdma ring preemption */ + sdma_v4_0_ctx_switch_enable(adev, true); + } + + /* start the gfx rings and rlc compute queues */ + for (i = 0; i < adev->sdma.num_instances; i++) { + uint32_t temp; + + WREG32_SDMA(i, mmSDMA0_SEM_WAIT_FAIL_TIMER_CNTL, 0); + sdma_v4_0_gfx_resume(adev, i); + if (adev->sdma.has_page_queue) + sdma_v4_0_page_resume(adev, i); + + /* set utc l1 enable flag always to 1 */ + temp = RREG32_SDMA(i, mmSDMA0_CNTL); + temp = REG_SET_FIELD(temp, SDMA0_CNTL, UTC_L1_ENABLE, 1); + WREG32_SDMA(i, mmSDMA0_CNTL, temp); + + if (!amdgpu_sriov_vf(adev)) { + /* unhalt engine */ + temp = RREG32_SDMA(i, mmSDMA0_F32_CNTL); + temp = REG_SET_FIELD(temp, SDMA0_F32_CNTL, HALT, 0); + WREG32_SDMA(i, mmSDMA0_F32_CNTL, temp); + } } - if (adev->firmware.load_type != AMDGPU_FW_LOAD_PSP) { - r = sdma_v4_0_load_microcode(adev); + if (amdgpu_sriov_vf(adev)) { + sdma_v4_0_ctx_switch_enable(adev, true); + sdma_v4_0_enable(adev, true); + } else { + r = sdma_v4_0_rlc_resume(adev); if (r) return r; } - /* unhalt the MEs */ - sdma_v4_0_enable(adev, true); - /* enable sdma ring preemption */ - sdma_v4_0_ctx_switch_enable(adev, true); + for (i = 0; i < adev->sdma.num_instances; i++) { + ring = &adev->sdma.instance[i].ring; - /* start the gfx rings and rlc compute queues */ - r = sdma_v4_0_gfx_resume(adev); - if (r) - return r; - r = sdma_v4_0_rlc_resume(adev); + r = amdgpu_ring_test_helper(ring); + if (r) + return r; + + if (adev->sdma.has_page_queue) { + struct amdgpu_ring *page = &adev->sdma.instance[i].page; + + r = amdgpu_ring_test_helper(page); + if (r) + return r; + + if (adev->mman.buffer_funcs_ring == page) + amdgpu_ttm_set_buffer_funcs_status(adev, true); + } + + if (adev->mman.buffer_funcs_ring == ring) + amdgpu_ttm_set_buffer_funcs_status(adev, true); + } return r; } @@ -993,21 +1184,16 @@ static int sdma_v4_0_ring_test_ring(struct amdgpu_ring *ring) u64 gpu_addr; r = amdgpu_device_wb_get(adev, &index); - if (r) { - dev_err(adev->dev, "(%d) failed to allocate wb slot\n", r); + if (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 (r) { - DRM_ERROR("amdgpu: dma failed to lock ring %d (%d).\n", ring->idx, r); - amdgpu_device_wb_free(adev, index); - return r; - } + if (r) + goto error_free_wb; amdgpu_ring_write(ring, SDMA_PKT_HEADER_OP(SDMA_OP_WRITE) | SDMA_PKT_HEADER_SUB_OP(SDMA_SUBOP_WRITE_LINEAR)); @@ -1024,15 +1210,11 @@ static int sdma_v4_0_ring_test_ring(struct amdgpu_ring *ring) DRM_UDELAY(1); } - if (i < adev->usec_timeout) { - DRM_DEBUG("ring test on %d succeeded in %d usecs\n", ring->idx, i); - } else { - DRM_ERROR("amdgpu: ring %d test failed (0x%08X)\n", - ring->idx, tmp); - r = -EINVAL; - } - amdgpu_device_wb_free(adev, index); + if (i >= adev->usec_timeout) + r = -ETIMEDOUT; +error_free_wb: + amdgpu_device_wb_free(adev, index); return r; } @@ -1055,20 +1237,16 @@ static int sdma_v4_0_ring_test_ib(struct amdgpu_ring *ring, long timeout) u64 gpu_addr; r = amdgpu_device_wb_get(adev, &index); - if (r) { - dev_err(adev->dev, "(%ld) failed to allocate wb slot\n", r); + if (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, &ib); - if (r) { - DRM_ERROR("amdgpu: failed to get ib (%ld).\n", r); + if (r) goto err0; - } ib.ptr[0] = SDMA_PKT_HEADER_OP(SDMA_OP_WRITE) | SDMA_PKT_HEADER_SUB_OP(SDMA_SUBOP_WRITE_LINEAR); @@ -1087,21 +1265,17 @@ static int sdma_v4_0_ring_test_ib(struct amdgpu_ring *ring, long timeout) 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; } tmp = le32_to_cpu(adev->wb.wb[index]); - if (tmp == 0xDEADBEEF) { - DRM_DEBUG("ib test on ring %d succeeded\n", ring->idx); + if (tmp == 0xDEADBEEF) r = 0; - } else { - DRM_ERROR("amdgpu: ib test failed (0x%08X)\n", tmp); + else r = -EINVAL; - } + err1: amdgpu_ib_free(adev, &ib, NULL); dma_fence_put(f); @@ -1206,7 +1380,7 @@ static void sdma_v4_0_vm_set_pte_pde(struct amdgpu_ib *ib, */ static void sdma_v4_0_ring_pad_ib(struct amdgpu_ring *ring, struct amdgpu_ib *ib) { - struct amdgpu_sdma_instance *sdma = amdgpu_get_sdma_instance(ring); + struct amdgpu_sdma_instance *sdma = amdgpu_sdma_get_instance_from_ring(ring); u32 pad_count; int i; @@ -1272,15 +1446,45 @@ static void sdma_v4_0_ring_emit_reg_wait(struct amdgpu_ring *ring, uint32_t reg, sdma_v4_0_wait_reg_mem(ring, 0, 0, reg, 0, val, mask, 10); } +static bool sdma_v4_0_fw_support_paging_queue(struct amdgpu_device *adev) +{ + uint fw_version = adev->sdma.instance[0].fw_version; + + switch (adev->asic_type) { + case CHIP_VEGA10: + return fw_version >= 430; + case CHIP_VEGA12: + /*return fw_version >= 31;*/ + return false; + case CHIP_VEGA20: + return fw_version >= 123; + default: + return false; + } +} + static int sdma_v4_0_early_init(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; + int r; if (adev->asic_type == CHIP_RAVEN) adev->sdma.num_instances = 1; else adev->sdma.num_instances = 2; + r = sdma_v4_0_init_microcode(adev); + if (r) { + DRM_ERROR("Failed to load sdma firmware!\n"); + return r; + } + + /* TODO: Page queue breaks driver reload under SRIOV */ + if ((adev->asic_type == CHIP_VEGA10) && amdgpu_sriov_vf((adev))) + adev->sdma.has_page_queue = false; + else if (sdma_v4_0_fw_support_paging_queue(adev)) + adev->sdma.has_page_queue = true; + sdma_v4_0_set_ring_funcs(adev); sdma_v4_0_set_buffer_funcs(adev); sdma_v4_0_set_vm_pte_funcs(adev); @@ -1289,7 +1493,6 @@ static int sdma_v4_0_early_init(void *handle) return 0; } - static int sdma_v4_0_sw_init(void *handle) { struct amdgpu_ring *ring; @@ -1308,12 +1511,6 @@ static int sdma_v4_0_sw_init(void *handle) if (r) return r; - r = sdma_v4_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; @@ -1322,15 +1519,8 @@ static int sdma_v4_0_sw_init(void *handle) DRM_INFO("use_doorbell being set to: [%s]\n", ring->use_doorbell?"true":"false"); - if (adev->asic_type == CHIP_VEGA10) - ring->doorbell_index = (i == 0) ? - (AMDGPU_VEGA10_DOORBELL64_sDMA_ENGINE0 << 1) //get DWORD offset - : (AMDGPU_VEGA10_DOORBELL64_sDMA_ENGINE1 << 1); // get DWORD offset - else - ring->doorbell_index = (i == 0) ? - (AMDGPU_DOORBELL64_sDMA_ENGINE0 << 1) //get DWORD offset - : (AMDGPU_DOORBELL64_sDMA_ENGINE1 << 1); // get DWORD offset - + /* doorbell size is 2 dwords, get DWORD offset */ + ring->doorbell_index = adev->doorbell_index.sdma_engine[i] << 1; sprintf(ring->name, "sdma%d", i); r = amdgpu_ring_init(adev, ring, 1024, @@ -1340,6 +1530,27 @@ static int sdma_v4_0_sw_init(void *handle) AMDGPU_SDMA_IRQ_TRAP1); if (r) return r; + + if (adev->sdma.has_page_queue) { + ring = &adev->sdma.instance[i].page; + ring->ring_obj = NULL; + ring->use_doorbell = true; + + /* paging queue use same doorbell index/routing as gfx queue + * with 0x400 (4096 dwords) offset on second doorbell page + */ + ring->doorbell_index = adev->doorbell_index.sdma_engine[i] << 1; + ring->doorbell_index += 0x400; + + sprintf(ring->name, "page%d", i); + r = amdgpu_ring_init(adev, ring, 1024, + &adev->sdma.trap_irq, + (i == 0) ? + AMDGPU_SDMA_IRQ_TRAP0 : + AMDGPU_SDMA_IRQ_TRAP1); + if (r) + return r; + } } return r; @@ -1350,8 +1561,11 @@ static int sdma_v4_0_sw_fini(void *handle) struct amdgpu_device *adev = (struct amdgpu_device *)handle; int i; - for (i = 0; i < adev->sdma.num_instances; i++) + 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); + } for (i = 0; i < adev->sdma.num_instances; i++) { release_firmware(adev->sdma.instance[i].fw); @@ -1414,7 +1628,7 @@ static bool sdma_v4_0_is_idle(void *handle) u32 i; for (i = 0; i < adev->sdma.num_instances; i++) { - u32 tmp = RREG32(sdma_v4_0_get_reg_offset(adev, i, mmSDMA0_STATUS_REG)); + u32 tmp = RREG32_SDMA(i, mmSDMA0_STATUS_REG); if (!(tmp & SDMA0_STATUS_REG__IDLE_MASK)) return false; @@ -1430,8 +1644,8 @@ static int sdma_v4_0_wait_for_idle(void *handle) struct amdgpu_device *adev = (struct amdgpu_device *)handle; for (i = 0; i < adev->usec_timeout; i++) { - sdma0 = RREG32(sdma_v4_0_get_reg_offset(adev, 0, mmSDMA0_STATUS_REG)); - sdma1 = RREG32(sdma_v4_0_get_reg_offset(adev, 1, mmSDMA0_STATUS_REG)); + sdma0 = RREG32_SDMA(0, mmSDMA0_STATUS_REG); + sdma1 = RREG32_SDMA(1, mmSDMA0_STATUS_REG); if (sdma0 & sdma1 & SDMA0_STATUS_REG__IDLE_MASK) return 0; @@ -1452,16 +1666,13 @@ static int sdma_v4_0_set_trap_irq_state(struct amdgpu_device *adev, unsigned type, enum amdgpu_interrupt_state state) { + unsigned int instance = (type == AMDGPU_SDMA_IRQ_TRAP0) ? 0 : 1; u32 sdma_cntl; - u32 reg_offset = (type == AMDGPU_SDMA_IRQ_TRAP0) ? - sdma_v4_0_get_reg_offset(adev, 0, mmSDMA0_CNTL) : - sdma_v4_0_get_reg_offset(adev, 1, mmSDMA0_CNTL); - - sdma_cntl = RREG32(reg_offset); + sdma_cntl = RREG32_SDMA(instance, mmSDMA0_CNTL); sdma_cntl = REG_SET_FIELD(sdma_cntl, SDMA0_CNTL, TRAP_ENABLE, state == AMDGPU_IRQ_STATE_ENABLE ? 1 : 0); - WREG32(reg_offset, sdma_cntl); + WREG32_SDMA(instance, mmSDMA0_CNTL, sdma_cntl); return 0; } @@ -1470,39 +1681,34 @@ static int sdma_v4_0_process_trap_irq(struct amdgpu_device *adev, struct amdgpu_irq_src *source, struct amdgpu_iv_entry *entry) { + uint32_t instance; + DRM_DEBUG("IH: SDMA trap\n"); switch (entry->client_id) { case SOC15_IH_CLIENTID_SDMA0: - switch (entry->ring_id) { - case 0: - amdgpu_fence_process(&adev->sdma.instance[0].ring); - break; - case 1: - /* XXX compute */ - break; - case 2: - /* XXX compute */ - break; - case 3: - /* XXX page queue*/ - break; - } + instance = 0; break; case SOC15_IH_CLIENTID_SDMA1: - switch (entry->ring_id) { - case 0: - amdgpu_fence_process(&adev->sdma.instance[1].ring); - break; - case 1: - /* XXX compute */ - break; - case 2: - /* XXX compute */ - break; - case 3: - /* XXX page queue*/ - break; - } + instance = 1; + break; + default: + return 0; + } + + switch (entry->ring_id) { + case 0: + amdgpu_fence_process(&adev->sdma.instance[instance].ring); + break; + case 1: + if (adev->asic_type == CHIP_VEGA20) + amdgpu_fence_process(&adev->sdma.instance[instance].page); + break; + case 2: + /* XXX compute */ + break; + case 3: + if (adev->asic_type != CHIP_VEGA20) + amdgpu_fence_process(&adev->sdma.instance[instance].page); break; } return 0; @@ -1512,12 +1718,29 @@ static int sdma_v4_0_process_illegal_inst_irq(struct amdgpu_device *adev, struct amdgpu_irq_src *source, struct amdgpu_iv_entry *entry) { + int instance; + DRM_ERROR("Illegal instruction in SDMA command stream\n"); - schedule_work(&adev->reset_work); + + switch (entry->client_id) { + case SOC15_IH_CLIENTID_SDMA0: + instance = 0; + break; + case SOC15_IH_CLIENTID_SDMA1: + instance = 1; + break; + default: + return 0; + } + + switch (entry->ring_id) { + case 0: + drm_sched_fault(&adev->sdma.instance[instance].ring.sched); + break; + } return 0; } - static void sdma_v4_0_update_medium_grain_clock_gating( struct amdgpu_device *adev, bool enable) @@ -1730,6 +1953,38 @@ static const struct amdgpu_ring_funcs sdma_v4_0_ring_funcs = { .emit_reg_write_reg_wait = amdgpu_ring_emit_reg_write_reg_wait_helper, }; +static const struct amdgpu_ring_funcs sdma_v4_0_page_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_MMHUB, + .get_rptr = sdma_v4_0_ring_get_rptr, + .get_wptr = sdma_v4_0_page_ring_get_wptr, + .set_wptr = sdma_v4_0_page_ring_set_wptr, + .emit_frame_size = + 6 + /* sdma_v4_0_ring_emit_hdp_flush */ + 3 + /* hdp invalidate */ + 6 + /* sdma_v4_0_ring_emit_pipeline_sync */ + /* sdma_v4_0_ring_emit_vm_flush */ + SOC15_FLUSH_GPU_TLB_NUM_WREG * 3 + + SOC15_FLUSH_GPU_TLB_NUM_REG_WAIT * 6 + + 10 + 10 + 10, /* sdma_v4_0_ring_emit_fence x3 for user fence, vm fence */ + .emit_ib_size = 7 + 6, /* sdma_v4_0_ring_emit_ib */ + .emit_ib = sdma_v4_0_ring_emit_ib, + .emit_fence = sdma_v4_0_ring_emit_fence, + .emit_pipeline_sync = sdma_v4_0_ring_emit_pipeline_sync, + .emit_vm_flush = sdma_v4_0_ring_emit_vm_flush, + .emit_hdp_flush = sdma_v4_0_ring_emit_hdp_flush, + .test_ring = sdma_v4_0_ring_test_ring, + .test_ib = sdma_v4_0_ring_test_ib, + .insert_nop = sdma_v4_0_ring_insert_nop, + .pad_ib = sdma_v4_0_ring_pad_ib, + .emit_wreg = sdma_v4_0_ring_emit_wreg, + .emit_reg_wait = sdma_v4_0_ring_emit_reg_wait, + .emit_reg_write_reg_wait = amdgpu_ring_emit_reg_write_reg_wait_helper, +}; + static void sdma_v4_0_set_ring_funcs(struct amdgpu_device *adev) { int i; @@ -1737,6 +1992,10 @@ static void sdma_v4_0_set_ring_funcs(struct amdgpu_device *adev) for (i = 0; i < adev->sdma.num_instances; i++) { adev->sdma.instance[i].ring.funcs = &sdma_v4_0_ring_funcs; adev->sdma.instance[i].ring.me = i; + if (adev->sdma.has_page_queue) { + adev->sdma.instance[i].page.funcs = &sdma_v4_0_page_ring_funcs; + adev->sdma.instance[i].page.me = i; + } } } @@ -1818,7 +2077,10 @@ static const struct amdgpu_buffer_funcs sdma_v4_0_buffer_funcs = { static void sdma_v4_0_set_buffer_funcs(struct amdgpu_device *adev) { adev->mman.buffer_funcs = &sdma_v4_0_buffer_funcs; - adev->mman.buffer_funcs_ring = &adev->sdma.instance[0].ring; + if (adev->sdma.has_page_queue) + adev->mman.buffer_funcs_ring = &adev->sdma.instance[0].page; + else + adev->mman.buffer_funcs_ring = &adev->sdma.instance[0].ring; } static const struct amdgpu_vm_pte_funcs sdma_v4_0_vm_pte_funcs = { @@ -1836,7 +2098,10 @@ static void sdma_v4_0_set_vm_pte_funcs(struct amdgpu_device *adev) adev->vm_manager.vm_pte_funcs = &sdma_v4_0_vm_pte_funcs; for (i = 0; i < adev->sdma.num_instances; i++) { - sched = &adev->sdma.instance[i].ring.sched; + if (adev->sdma.has_page_queue) + sched = &adev->sdma.instance[i].page.sched; + else + sched = &adev->sdma.instance[i].ring.sched; adev->vm_manager.vm_pte_rqs[i] = &sched->sched_rq[DRM_SCHED_PRIORITY_KERNEL]; } diff --git a/drivers/gpu/drm/amd/amdgpu/si.c b/drivers/gpu/drm/amd/amdgpu/si.c index f8408f88cd37..9d8df68893b9 100644 --- a/drivers/gpu/drm/amd/amdgpu/si.c +++ b/drivers/gpu/drm/amd/amdgpu/si.c @@ -47,6 +47,7 @@ #include "dce/dce_6_0_d.h" #include "uvd/uvd_4_0_d.h" #include "bif/bif_3_0_d.h" +#include "bif/bif_3_0_sh_mask.h" static const u32 tahiti_golden_registers[] = { @@ -1258,6 +1259,11 @@ static bool si_need_full_reset(struct amdgpu_device *adev) return true; } +static bool si_need_reset_on_init(struct amdgpu_device *adev) +{ + return false; +} + static int si_get_pcie_lanes(struct amdgpu_device *adev) { u32 link_width_cntl; @@ -1323,6 +1329,52 @@ static void si_set_pcie_lanes(struct amdgpu_device *adev, int lanes) WREG32_PCIE_PORT(PCIE_LC_LINK_WIDTH_CNTL, link_width_cntl); } +static void si_get_pcie_usage(struct amdgpu_device *adev, uint64_t *count0, + uint64_t *count1) +{ + uint32_t perfctr = 0; + uint64_t cnt0_of, cnt1_of; + int tmp; + + /* This reports 0 on APUs, so return to avoid writing/reading registers + * that may or may not be different from their GPU counterparts + */ + if (adev->flags & AMD_IS_APU) + return; + + /* Set the 2 events that we wish to watch, defined above */ + /* Reg 40 is # received msgs, Reg 104 is # of posted requests sent */ + perfctr = REG_SET_FIELD(perfctr, PCIE_PERF_CNTL_TXCLK, EVENT0_SEL, 40); + perfctr = REG_SET_FIELD(perfctr, PCIE_PERF_CNTL_TXCLK, EVENT1_SEL, 104); + + /* Write to enable desired perf counters */ + WREG32_PCIE(ixPCIE_PERF_CNTL_TXCLK, perfctr); + /* Zero out and enable the perf counters + * Write 0x5: + * Bit 0 = Start all counters(1) + * Bit 2 = Global counter reset enable(1) + */ + WREG32_PCIE(ixPCIE_PERF_COUNT_CNTL, 0x00000005); + + msleep(1000); + + /* Load the shadow and disable the perf counters + * Write 0x2: + * Bit 0 = Stop counters(0) + * Bit 1 = Load the shadow counters(1) + */ + WREG32_PCIE(ixPCIE_PERF_COUNT_CNTL, 0x00000002); + + /* Read register values to get any >32bit overflow */ + tmp = RREG32_PCIE(ixPCIE_PERF_CNTL_TXCLK); + cnt0_of = REG_GET_FIELD(tmp, PCIE_PERF_CNTL_TXCLK, COUNTER0_UPPER); + cnt1_of = REG_GET_FIELD(tmp, PCIE_PERF_CNTL_TXCLK, COUNTER1_UPPER); + + /* Get the values and add the overflow */ + *count0 = RREG32_PCIE(ixPCIE_PERF_COUNT0_TXCLK) | (cnt0_of << 32); + *count1 = RREG32_PCIE(ixPCIE_PERF_COUNT1_TXCLK) | (cnt1_of << 32); +} + static const struct amdgpu_asic_funcs si_asic_funcs = { .read_disabled_bios = &si_read_disabled_bios, @@ -1339,6 +1391,8 @@ static const struct amdgpu_asic_funcs si_asic_funcs = .flush_hdp = &si_flush_hdp, .invalidate_hdp = &si_invalidate_hdp, .need_full_reset = &si_need_full_reset, + .get_pcie_usage = &si_get_pcie_usage, + .need_reset_on_init = &si_need_reset_on_init, }; static uint32_t si_get_rev_id(struct amdgpu_device *adev) @@ -1382,7 +1436,7 @@ static int si_common_early_init(void *handle) AMD_CG_SUPPORT_UVD_MGCG | AMD_CG_SUPPORT_HDP_LS | AMD_CG_SUPPORT_HDP_MGCG; - adev->pg_flags = 0; + adev->pg_flags = 0; adev->external_rev_id = (adev->rev_id == 0) ? 1 : (adev->rev_id == 1) ? 5 : 6; break; diff --git a/drivers/gpu/drm/amd/amdgpu/si_dma.c b/drivers/gpu/drm/amd/amdgpu/si_dma.c index adbaea6da0d7..f15f196684ba 100644 --- a/drivers/gpu/drm/amd/amdgpu/si_dma.c +++ b/drivers/gpu/drm/amd/amdgpu/si_dma.c @@ -61,9 +61,11 @@ static void si_dma_ring_set_wptr(struct amdgpu_ring *ring) } static void si_dma_ring_emit_ib(struct amdgpu_ring *ring, + struct amdgpu_job *job, struct amdgpu_ib *ib, - unsigned vmid, bool ctx_switch) + uint32_t flags) { + unsigned vmid = AMDGPU_JOB_GET_VMID(job); /* The indirect buffer packet must end on an 8 DW boundary in the DMA ring. * Pad as necessary with NOPs. */ @@ -122,7 +124,7 @@ static void si_dma_stop(struct amdgpu_device *adev) if (adev->mman.buffer_funcs_ring == ring) amdgpu_ttm_set_buffer_funcs_status(adev, false); - ring->ready = false; + ring->sched.ready = false; } } @@ -175,13 +177,11 @@ static int si_dma_start(struct amdgpu_device *adev) WREG32(DMA_RB_WPTR + sdma_offsets[i], lower_32_bits(ring->wptr) << 2); WREG32(DMA_RB_CNTL + sdma_offsets[i], rb_cntl | DMA_RB_ENABLE); - ring->ready = true; + ring->sched.ready = true; - r = amdgpu_ring_test_ring(ring); - if (r) { - ring->ready = false; + r = amdgpu_ring_test_helper(ring); + if (r) return r; - } if (adev->mman.buffer_funcs_ring == ring) amdgpu_ttm_set_buffer_funcs_status(adev, true); @@ -209,21 +209,16 @@ static int si_dma_ring_test_ring(struct amdgpu_ring *ring) u64 gpu_addr; r = amdgpu_device_wb_get(adev, &index); - if (r) { - dev_err(adev->dev, "(%d) failed to allocate wb slot\n", r); + if (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, 4); - if (r) { - DRM_ERROR("amdgpu: dma failed to lock ring %d (%d).\n", ring->idx, r); - amdgpu_device_wb_free(adev, index); - return r; - } + if (r) + goto error_free_wb; amdgpu_ring_write(ring, DMA_PACKET(DMA_PACKET_WRITE, 0, 0, 0, 1)); amdgpu_ring_write(ring, lower_32_bits(gpu_addr)); @@ -238,15 +233,11 @@ static int si_dma_ring_test_ring(struct amdgpu_ring *ring) DRM_UDELAY(1); } - if (i < adev->usec_timeout) { - DRM_DEBUG("ring test on %d succeeded in %d usecs\n", ring->idx, i); - } else { - DRM_ERROR("amdgpu: ring %d test failed (0x%08X)\n", - ring->idx, tmp); - r = -EINVAL; - } - amdgpu_device_wb_free(adev, index); + if (i >= adev->usec_timeout) + r = -ETIMEDOUT; +error_free_wb: + amdgpu_device_wb_free(adev, index); return r; } @@ -269,20 +260,16 @@ static int si_dma_ring_test_ib(struct amdgpu_ring *ring, long timeout) long r; r = amdgpu_device_wb_get(adev, &index); - if (r) { - dev_err(adev->dev, "(%ld) failed to allocate wb slot\n", r); + if (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, &ib); - if (r) { - DRM_ERROR("amdgpu: failed to get ib (%ld).\n", r); + if (r) goto err0; - } ib.ptr[0] = DMA_PACKET(DMA_PACKET_WRITE, 0, 0, 0, 1); ib.ptr[1] = lower_32_bits(gpu_addr); @@ -295,21 +282,16 @@ static int si_dma_ring_test_ib(struct amdgpu_ring *ring, long timeout) 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; } tmp = le32_to_cpu(adev->wb.wb[index]); - if (tmp == 0xDEADBEEF) { - DRM_DEBUG("ib test on ring %d succeeded\n", ring->idx); + if (tmp == 0xDEADBEEF) r = 0; - } else { - DRM_ERROR("amdgpu: ib test failed (0x%08X)\n", tmp); + else r = -EINVAL; - } err1: amdgpu_ib_free(adev, &ib, NULL); @@ -658,15 +640,6 @@ static int si_dma_process_trap_irq(struct amdgpu_device *adev, return 0; } -static int si_dma_process_illegal_inst_irq(struct amdgpu_device *adev, - struct amdgpu_irq_src *source, - struct amdgpu_iv_entry *entry) -{ - DRM_ERROR("Illegal instruction in SDMA command stream\n"); - schedule_work(&adev->reset_work); - return 0; -} - static int si_dma_set_clockgating_state(void *handle, enum amd_clockgating_state state) { @@ -781,15 +754,10 @@ static const struct amdgpu_irq_src_funcs si_dma_trap_irq_funcs = { .process = si_dma_process_trap_irq, }; -static const struct amdgpu_irq_src_funcs si_dma_illegal_inst_irq_funcs = { - .process = si_dma_process_illegal_inst_irq, -}; - static void si_dma_set_irq_funcs(struct amdgpu_device *adev) { adev->sdma.trap_irq.num_types = AMDGPU_SDMA_IRQ_LAST; adev->sdma.trap_irq.funcs = &si_dma_trap_irq_funcs; - adev->sdma.illegal_inst_irq.funcs = &si_dma_illegal_inst_irq_funcs; } /** diff --git a/drivers/gpu/drm/amd/amdgpu/si_dpm.c b/drivers/gpu/drm/amd/amdgpu/si_dpm.c index da58040fdbdc..41e01a7f57a4 100644 --- a/drivers/gpu/drm/amd/amdgpu/si_dpm.c +++ b/drivers/gpu/drm/amd/amdgpu/si_dpm.c @@ -6216,10 +6216,12 @@ static void si_request_link_speed_change_before_state_change(struct amdgpu_devic si_pi->force_pcie_gen = AMDGPU_PCIE_GEN2; if (current_link_speed == AMDGPU_PCIE_GEN2) break; + /* fall through */ case AMDGPU_PCIE_GEN2: if (amdgpu_acpi_pcie_performance_request(adev, PCIE_PERF_REQ_PECI_GEN2, false) == 0) break; #endif + /* fall through */ default: si_pi->force_pcie_gen = si_get_current_pcie_speed(adev); break; diff --git a/drivers/gpu/drm/amd/amdgpu/si_ih.c b/drivers/gpu/drm/amd/amdgpu/si_ih.c index b3d7d9f83202..8c50c9cab455 100644 --- a/drivers/gpu/drm/amd/amdgpu/si_ih.c +++ b/drivers/gpu/drm/amd/amdgpu/si_ih.c @@ -57,9 +57,9 @@ static void si_ih_disable_interrupts(struct amdgpu_device *adev) static int si_ih_irq_init(struct amdgpu_device *adev) { + struct amdgpu_ih_ring *ih = &adev->irq.ih; int rb_bufsz; u32 interrupt_cntl, ih_cntl, ih_rb_cntl; - u64 wptr_off; si_ih_disable_interrupts(adev); WREG32(INTERRUPT_CNTL2, adev->irq.ih.gpu_addr >> 8); @@ -76,9 +76,8 @@ static int si_ih_irq_init(struct amdgpu_device *adev) (rb_bufsz << 1) | IH_WPTR_WRITEBACK_ENABLE; - wptr_off = adev->wb.gpu_addr + (adev->irq.ih.wptr_offs * 4); - WREG32(IH_RB_WPTR_ADDR_LO, lower_32_bits(wptr_off)); - WREG32(IH_RB_WPTR_ADDR_HI, upper_32_bits(wptr_off) & 0xFF); + WREG32(IH_RB_WPTR_ADDR_LO, lower_32_bits(ih->wptr_addr)); + WREG32(IH_RB_WPTR_ADDR_HI, upper_32_bits(ih->wptr_addr) & 0xFF); WREG32(IH_RB_CNTL, ih_rb_cntl); WREG32(IH_RB_RPTR, 0); WREG32(IH_RB_WPTR, 0); @@ -100,47 +99,36 @@ static void si_ih_irq_disable(struct amdgpu_device *adev) mdelay(1); } -static u32 si_ih_get_wptr(struct amdgpu_device *adev) +static u32 si_ih_get_wptr(struct amdgpu_device *adev, + struct amdgpu_ih_ring *ih) { u32 wptr, tmp; - wptr = le32_to_cpu(adev->wb.wb[adev->irq.ih.wptr_offs]); + wptr = le32_to_cpu(*ih->wptr_cpu); if (wptr & IH_RB_WPTR__RB_OVERFLOW_MASK) { wptr &= ~IH_RB_WPTR__RB_OVERFLOW_MASK; dev_warn(adev->dev, "IH ring buffer overflow (0x%08X, 0x%08X, 0x%08X)\n", - wptr, adev->irq.ih.rptr, (wptr + 16) & adev->irq.ih.ptr_mask); - adev->irq.ih.rptr = (wptr + 16) & adev->irq.ih.ptr_mask; + wptr, ih->rptr, (wptr + 16) & ih->ptr_mask); + ih->rptr = (wptr + 16) & ih->ptr_mask; tmp = RREG32(IH_RB_CNTL); tmp |= IH_RB_CNTL__WPTR_OVERFLOW_CLEAR_MASK; WREG32(IH_RB_CNTL, tmp); } - return (wptr & adev->irq.ih.ptr_mask); -} - -/** - * si_ih_prescreen_iv - prescreen an interrupt vector - * - * @adev: amdgpu_device pointer - * - * Returns true if the interrupt vector should be further processed. - */ -static bool si_ih_prescreen_iv(struct amdgpu_device *adev) -{ - /* Process all interrupts */ - return true; + return (wptr & ih->ptr_mask); } static void si_ih_decode_iv(struct amdgpu_device *adev, - struct amdgpu_iv_entry *entry) + struct amdgpu_ih_ring *ih, + struct amdgpu_iv_entry *entry) { - u32 ring_index = adev->irq.ih.rptr >> 2; + u32 ring_index = ih->rptr >> 2; uint32_t dw[4]; - dw[0] = le32_to_cpu(adev->irq.ih.ring[ring_index + 0]); - dw[1] = le32_to_cpu(adev->irq.ih.ring[ring_index + 1]); - dw[2] = le32_to_cpu(adev->irq.ih.ring[ring_index + 2]); - dw[3] = le32_to_cpu(adev->irq.ih.ring[ring_index + 3]); + dw[0] = le32_to_cpu(ih->ring[ring_index + 0]); + dw[1] = le32_to_cpu(ih->ring[ring_index + 1]); + dw[2] = le32_to_cpu(ih->ring[ring_index + 2]); + dw[3] = le32_to_cpu(ih->ring[ring_index + 3]); entry->client_id = AMDGPU_IRQ_CLIENTID_LEGACY; entry->src_id = dw[0] & 0xff; @@ -148,12 +136,13 @@ static void si_ih_decode_iv(struct amdgpu_device *adev, entry->ring_id = dw[2] & 0xff; entry->vmid = (dw[2] >> 8) & 0xff; - adev->irq.ih.rptr += 16; + ih->rptr += 16; } -static void si_ih_set_rptr(struct amdgpu_device *adev) +static void si_ih_set_rptr(struct amdgpu_device *adev, + struct amdgpu_ih_ring *ih) { - WREG32(IH_RB_RPTR, adev->irq.ih.rptr); + WREG32(IH_RB_RPTR, ih->rptr); } static int si_ih_early_init(void *handle) @@ -301,7 +290,6 @@ static const struct amd_ip_funcs si_ih_ip_funcs = { static const struct amdgpu_ih_funcs si_ih_funcs = { .get_wptr = si_ih_get_wptr, - .prescreen_iv = si_ih_prescreen_iv, .decode_iv = si_ih_decode_iv, .set_rptr = si_ih_set_rptr }; diff --git a/drivers/gpu/drm/amd/amdgpu/soc15.c b/drivers/gpu/drm/amd/amdgpu/soc15.c index 4cc0dcb1a187..ed89a101f73f 100644 --- a/drivers/gpu/drm/amd/amdgpu/soc15.c +++ b/drivers/gpu/drm/amd/amdgpu/soc15.c @@ -43,6 +43,10 @@ #include "hdp/hdp_4_0_sh_mask.h" #include "smuio/smuio_9_0_offset.h" #include "smuio/smuio_9_0_sh_mask.h" +#include "nbio/nbio_7_0_default.h" +#include "nbio/nbio_7_0_sh_mask.h" +#include "nbio/nbio_7_0_smn.h" +#include "mp/mp_9_0_offset.h" #include "soc15.h" #include "soc15_common.h" @@ -385,14 +389,13 @@ void soc15_program_register_sequence(struct amdgpu_device *adev, } - -static int soc15_asic_reset(struct amdgpu_device *adev) +static int soc15_asic_mode1_reset(struct amdgpu_device *adev) { u32 i; amdgpu_atombios_scratch_regs_engine_hung(adev, true); - dev_info(adev->dev, "GPU reset\n"); + dev_info(adev->dev, "GPU mode1 reset\n"); /* disable BM */ pci_clear_master(adev->pdev); @@ -417,6 +420,62 @@ static int soc15_asic_reset(struct amdgpu_device *adev) return 0; } +static int soc15_asic_get_baco_capability(struct amdgpu_device *adev, bool *cap) +{ + void *pp_handle = adev->powerplay.pp_handle; + const struct amd_pm_funcs *pp_funcs = adev->powerplay.pp_funcs; + + if (!pp_funcs || !pp_funcs->get_asic_baco_capability) { + *cap = false; + return -ENOENT; + } + + return pp_funcs->get_asic_baco_capability(pp_handle, cap); +} + +static int soc15_asic_baco_reset(struct amdgpu_device *adev) +{ + void *pp_handle = adev->powerplay.pp_handle; + const struct amd_pm_funcs *pp_funcs = adev->powerplay.pp_funcs; + + if (!pp_funcs ||!pp_funcs->get_asic_baco_state ||!pp_funcs->set_asic_baco_state) + return -ENOENT; + + /* enter BACO state */ + if (pp_funcs->set_asic_baco_state(pp_handle, 1)) + return -EIO; + + /* exit BACO state */ + if (pp_funcs->set_asic_baco_state(pp_handle, 0)) + return -EIO; + + dev_info(adev->dev, "GPU BACO reset\n"); + + return 0; +} + +static int soc15_asic_reset(struct amdgpu_device *adev) +{ + int ret; + bool baco_reset; + + switch (adev->asic_type) { + case CHIP_VEGA10: + soc15_asic_get_baco_capability(adev, &baco_reset); + break; + default: + baco_reset = false; + break; + } + + if (baco_reset) + ret = soc15_asic_baco_reset(adev); + else + ret = soc15_asic_mode1_reset(adev); + + return ret; +} + /*static int soc15_set_uvd_clock(struct amdgpu_device *adev, u32 clock, u32 cntl_reg, u32 status_reg) { @@ -507,6 +566,9 @@ int soc15_set_ip_blocks(struct amdgpu_device *adev) return -EINVAL; } + if (adev->asic_type == CHIP_VEGA20) + adev->gmc.xgmi.supported = true; + if (adev->flags & AMD_IS_APU) adev->nbio_funcs = &nbio_v7_0_funcs; else if (adev->asic_type == CHIP_VEGA20) @@ -532,10 +594,12 @@ int soc15_set_ip_blocks(struct amdgpu_device *adev) amdgpu_device_ip_block_add(adev, &vega10_common_ip_block); amdgpu_device_ip_block_add(adev, &gmc_v9_0_ip_block); amdgpu_device_ip_block_add(adev, &vega10_ih_ip_block); - if (adev->asic_type == CHIP_VEGA20) - amdgpu_device_ip_block_add(adev, &psp_v11_0_ip_block); - else - amdgpu_device_ip_block_add(adev, &psp_v3_1_ip_block); + if (likely(adev->firmware.load_type == AMDGPU_FW_LOAD_PSP)) { + if (adev->asic_type == CHIP_VEGA20) + amdgpu_device_ip_block_add(adev, &psp_v11_0_ip_block); + else + amdgpu_device_ip_block_add(adev, &psp_v3_1_ip_block); + } amdgpu_device_ip_block_add(adev, &gfx_v9_0_ip_block); amdgpu_device_ip_block_add(adev, &sdma_v4_0_ip_block); if (!amdgpu_sriov_vf(adev)) @@ -557,7 +621,8 @@ int soc15_set_ip_blocks(struct amdgpu_device *adev) amdgpu_device_ip_block_add(adev, &vega10_common_ip_block); amdgpu_device_ip_block_add(adev, &gmc_v9_0_ip_block); amdgpu_device_ip_block_add(adev, &vega10_ih_ip_block); - amdgpu_device_ip_block_add(adev, &psp_v10_0_ip_block); + if (likely(adev->firmware.load_type == AMDGPU_FW_LOAD_PSP)) + amdgpu_device_ip_block_add(adev, &psp_v10_0_ip_block); amdgpu_device_ip_block_add(adev, &gfx_v9_0_ip_block); amdgpu_device_ip_block_add(adev, &sdma_v4_0_ip_block); amdgpu_device_ip_block_add(adev, &pp_smu_ip_block); @@ -598,6 +663,68 @@ static bool soc15_need_full_reset(struct amdgpu_device *adev) /* change this when we implement soft reset */ return true; } +static void soc15_get_pcie_usage(struct amdgpu_device *adev, uint64_t *count0, + uint64_t *count1) +{ + uint32_t perfctr = 0; + uint64_t cnt0_of, cnt1_of; + int tmp; + + /* This reports 0 on APUs, so return to avoid writing/reading registers + * that may or may not be different from their GPU counterparts + */ + if (adev->flags & AMD_IS_APU) + return; + + /* Set the 2 events that we wish to watch, defined above */ + /* Reg 40 is # received msgs, Reg 104 is # of posted requests sent */ + perfctr = REG_SET_FIELD(perfctr, PCIE_PERF_CNTL_TXCLK, EVENT0_SEL, 40); + perfctr = REG_SET_FIELD(perfctr, PCIE_PERF_CNTL_TXCLK, EVENT1_SEL, 104); + + /* Write to enable desired perf counters */ + WREG32_PCIE(smnPCIE_PERF_CNTL_TXCLK, perfctr); + /* Zero out and enable the perf counters + * Write 0x5: + * Bit 0 = Start all counters(1) + * Bit 2 = Global counter reset enable(1) + */ + WREG32_PCIE(smnPCIE_PERF_COUNT_CNTL, 0x00000005); + + msleep(1000); + + /* Load the shadow and disable the perf counters + * Write 0x2: + * Bit 0 = Stop counters(0) + * Bit 1 = Load the shadow counters(1) + */ + WREG32_PCIE(smnPCIE_PERF_COUNT_CNTL, 0x00000002); + + /* Read register values to get any >32bit overflow */ + tmp = RREG32_PCIE(smnPCIE_PERF_CNTL_TXCLK); + cnt0_of = REG_GET_FIELD(tmp, PCIE_PERF_CNTL_TXCLK, COUNTER0_UPPER); + cnt1_of = REG_GET_FIELD(tmp, PCIE_PERF_CNTL_TXCLK, COUNTER1_UPPER); + + /* Get the values and add the overflow */ + *count0 = RREG32_PCIE(smnPCIE_PERF_COUNT0_TXCLK) | (cnt0_of << 32); + *count1 = RREG32_PCIE(smnPCIE_PERF_COUNT1_TXCLK) | (cnt1_of << 32); +} + +static bool soc15_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, mmMP0_SMN_C2PMSG_81); + if (sol_reg) + return true; + + return false; +} static const struct amdgpu_asic_funcs soc15_asic_funcs = { @@ -613,6 +740,28 @@ static const struct amdgpu_asic_funcs soc15_asic_funcs = .flush_hdp = &soc15_flush_hdp, .invalidate_hdp = &soc15_invalidate_hdp, .need_full_reset = &soc15_need_full_reset, + .init_doorbell_index = &vega10_doorbell_index_init, + .get_pcie_usage = &soc15_get_pcie_usage, + .need_reset_on_init = &soc15_need_reset_on_init, +}; + +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_register = &soc15_read_register, + .reset = &soc15_asic_reset, + .set_vga_state = &soc15_vga_set_state, + .get_xclk = &soc15_get_xclk, + .set_uvd_clocks = &soc15_set_uvd_clocks, + .set_vce_clocks = &soc15_set_vce_clocks, + .get_config_memsize = &soc15_get_config_memsize, + .flush_hdp = &soc15_flush_hdp, + .invalidate_hdp = &soc15_invalidate_hdp, + .need_full_reset = &soc15_need_full_reset, + .init_doorbell_index = &vega20_doorbell_index_init, + .get_pcie_usage = &soc15_get_pcie_usage, + .need_reset_on_init = &soc15_need_reset_on_init, }; static int soc15_common_early_init(void *handle) @@ -632,11 +781,11 @@ static int soc15_common_early_init(void *handle) adev->se_cac_rreg = &soc15_se_cac_rreg; adev->se_cac_wreg = &soc15_se_cac_wreg; - adev->asic_funcs = &soc15_asic_funcs; adev->external_rev_id = 0xFF; switch (adev->asic_type) { case CHIP_VEGA10: + adev->asic_funcs = &soc15_asic_funcs; adev->cg_flags = AMD_CG_SUPPORT_GFX_MGCG | AMD_CG_SUPPORT_GFX_MGLS | AMD_CG_SUPPORT_GFX_RLC_LS | @@ -660,6 +809,7 @@ static int soc15_common_early_init(void *handle) adev->external_rev_id = 0x1; break; case CHIP_VEGA12: + adev->asic_funcs = &soc15_asic_funcs; adev->cg_flags = AMD_CG_SUPPORT_GFX_MGCG | AMD_CG_SUPPORT_GFX_MGLS | AMD_CG_SUPPORT_GFX_CGCG | @@ -682,6 +832,7 @@ static int soc15_common_early_init(void *handle) adev->external_rev_id = adev->rev_id + 0x14; break; case CHIP_VEGA20: + adev->asic_funcs = &vega20_asic_funcs; adev->cg_flags = AMD_CG_SUPPORT_GFX_MGCG | AMD_CG_SUPPORT_GFX_MGLS | AMD_CG_SUPPORT_GFX_CGCG | @@ -704,12 +855,15 @@ static int soc15_common_early_init(void *handle) adev->external_rev_id = adev->rev_id + 0x28; break; case CHIP_RAVEN: + adev->asic_funcs = &soc15_asic_funcs; if (adev->rev_id >= 0x8) - adev->external_rev_id = adev->rev_id + 0x81; + adev->external_rev_id = adev->rev_id + 0x79; else if (adev->pdev->device == 0x15d8) adev->external_rev_id = adev->rev_id + 0x41; + else if (adev->rev_id == 1) + adev->external_rev_id = adev->rev_id + 0x20; else - adev->external_rev_id = 0x1; + adev->external_rev_id = adev->rev_id + 0x01; if (adev->rev_id >= 0x8) { adev->cg_flags = AMD_CG_SUPPORT_GFX_MGCG | @@ -816,6 +970,22 @@ static int soc15_common_sw_fini(void *handle) return 0; } +static void soc15_doorbell_range_init(struct amdgpu_device *adev) +{ + int i; + struct amdgpu_ring *ring; + + 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, + adev->doorbell_index.sdma_doorbell_range); + } + + adev->nbio_funcs->ih_doorbell_range(adev, adev->irq.ih.use_doorbell, + adev->irq.ih.doorbell_index); +} + static int soc15_common_hw_init(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; @@ -828,6 +998,12 @@ static int soc15_common_hw_init(void *handle) adev->nbio_funcs->init_registers(adev); /* enable the doorbell aperture */ 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. + */ + soc15_doorbell_range_init(adev); return 0; } diff --git a/drivers/gpu/drm/amd/amdgpu/soc15.h b/drivers/gpu/drm/amd/amdgpu/soc15.h index f8ad7804dc40..a66c8bfbbaa6 100644 --- a/drivers/gpu/drm/amd/amdgpu/soc15.h +++ b/drivers/gpu/drm/amd/amdgpu/soc15.h @@ -58,4 +58,6 @@ void soc15_program_register_sequence(struct amdgpu_device *adev, int vega10_reg_base_init(struct amdgpu_device *adev); int vega20_reg_base_init(struct amdgpu_device *adev); +void vega10_doorbell_index_init(struct amdgpu_device *adev); +void vega20_doorbell_index_init(struct amdgpu_device *adev); #endif diff --git a/drivers/gpu/drm/amd/amdgpu/soc15_common.h b/drivers/gpu/drm/amd/amdgpu/soc15_common.h index 958b10a57073..49c262540940 100644 --- a/drivers/gpu/drm/amd/amdgpu/soc15_common.h +++ b/drivers/gpu/drm/amd/amdgpu/soc15_common.h @@ -49,14 +49,19 @@ #define SOC15_WAIT_ON_RREG(ip, inst, reg, expected_value, mask, ret) \ do { \ + uint32_t old_ = 0; \ uint32_t tmp_ = RREG32(adev->reg_offset[ip##_HWIP][inst][reg##_BASE_IDX] + reg); \ uint32_t loop = adev->usec_timeout; \ while ((tmp_ & (mask)) != (expected_value)) { \ - udelay(2); \ + if (old_ != tmp_) { \ + loop = adev->usec_timeout; \ + old_ = tmp_; \ + } else \ + udelay(1); \ tmp_ = RREG32(adev->reg_offset[ip##_HWIP][inst][reg##_BASE_IDX] + reg); \ loop--; \ if (!loop) { \ - DRM_ERROR("Register(%d) [%s] failed to reach value 0x%08x != 0x%08x\n", \ + DRM_WARN("Register(%d) [%s] failed to reach value 0x%08x != 0x%08x\n", \ inst, #reg, (unsigned)expected_value, (unsigned)(tmp_ & (mask))); \ ret = -ETIMEDOUT; \ break; \ diff --git a/drivers/gpu/drm/amd/amdgpu/ta_xgmi_if.h b/drivers/gpu/drm/amd/amdgpu/ta_xgmi_if.h new file mode 100644 index 000000000000..ac2c27b7630c --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/ta_xgmi_if.h @@ -0,0 +1,130 @@ +/* + * Copyright 2018 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 _TA_XGMI_IF_H +#define _TA_XGMI_IF_H + +/* Responses have bit 31 set */ +#define RSP_ID_MASK (1U << 31) +#define RSP_ID(cmdId) (((uint32_t)(cmdId)) | RSP_ID_MASK) + +enum ta_command_xgmi { + TA_COMMAND_XGMI__INITIALIZE = 0x00, + TA_COMMAND_XGMI__GET_NODE_ID = 0x01, + TA_COMMAND_XGMI__GET_HIVE_ID = 0x02, + TA_COMMAND_XGMI__GET_GET_TOPOLOGY_INFO = 0x03, + TA_COMMAND_XGMI__SET_TOPOLOGY_INFO = 0x04 +}; + +/* XGMI related enumerations */ +/**********************************************************/; +enum ta_xgmi_connected_nodes { + TA_XGMI__MAX_CONNECTED_NODES = 64 +}; + +enum ta_xgmi_status { + TA_XGMI_STATUS__SUCCESS = 0x00, + TA_XGMI_STATUS__GENERIC_FAILURE = 0x01, + TA_XGMI_STATUS__NULL_POINTER = 0x02, + TA_XGMI_STATUS__INVALID_PARAMETER = 0x03, + TA_XGMI_STATUS__NOT_INITIALIZED = 0x04, + TA_XGMI_STATUS__INVALID_NODE_NUM = 0x05, + TA_XGMI_STATUS__INVALID_NODE_ID = 0x06, + TA_XGMI_STATUS__INVALID_TOPOLOGY = 0x07, + TA_XGMI_STATUS__FAILED_ID_GEN = 0x08, + TA_XGMI_STATUS__FAILED_TOPOLOGY_INIT = 0x09, + TA_XGMI_STATUS__SET_SHARING_ERROR = 0x0A +}; + +enum ta_xgmi_assigned_sdma_engine { + TA_XGMI_ASSIGNED_SDMA_ENGINE__NOT_ASSIGNED = -1, + TA_XGMI_ASSIGNED_SDMA_ENGINE__SDMA0 = 0, + TA_XGMI_ASSIGNED_SDMA_ENGINE__SDMA1 = 1, + TA_XGMI_ASSIGNED_SDMA_ENGINE__SDMA2 = 2, + TA_XGMI_ASSIGNED_SDMA_ENGINE__SDMA3 = 3, + TA_XGMI_ASSIGNED_SDMA_ENGINE__SDMA4 = 4, + TA_XGMI_ASSIGNED_SDMA_ENGINE__SDMA5 = 5 +}; + +/* input/output structures for XGMI commands */ +/**********************************************************/ +struct ta_xgmi_node_info { + uint64_t node_id; + uint8_t num_hops; + uint8_t is_sharing_enabled; + enum ta_xgmi_assigned_sdma_engine sdma_engine; +}; + +struct ta_xgmi_cmd_initialize_output { + uint32_t status; +}; + +struct ta_xgmi_cmd_get_node_id_output { + uint64_t node_id; +}; + +struct ta_xgmi_cmd_get_hive_id_output { + uint64_t hive_id; +}; + +struct ta_xgmi_cmd_get_topology_info_input { + uint32_t num_nodes; + struct ta_xgmi_node_info nodes[TA_XGMI__MAX_CONNECTED_NODES]; +}; + +struct ta_xgmi_cmd_get_topology_info_output { + uint32_t num_nodes; + struct ta_xgmi_node_info nodes[TA_XGMI__MAX_CONNECTED_NODES]; +}; + +struct ta_xgmi_cmd_set_topology_info_input { + uint32_t num_nodes; + struct ta_xgmi_node_info nodes[TA_XGMI__MAX_CONNECTED_NODES]; +}; + +/**********************************************************/ +/* Common input structure for XGMI callbacks */ +union ta_xgmi_cmd_input { + struct ta_xgmi_cmd_get_topology_info_input get_topology_info; + struct ta_xgmi_cmd_set_topology_info_input set_topology_info; +}; + +/* Common output structure for XGMI callbacks */ +union ta_xgmi_cmd_output { + struct ta_xgmi_cmd_initialize_output initialize; + struct ta_xgmi_cmd_get_node_id_output get_node_id; + struct ta_xgmi_cmd_get_hive_id_output get_hive_id; + struct ta_xgmi_cmd_get_topology_info_output get_topology_info; +}; +/**********************************************************/ + +struct ta_xgmi_shared_memory { + uint32_t cmd_id; + uint32_t resp_id; + enum ta_xgmi_status xgmi_status; + uint32_t reserved; + union ta_xgmi_cmd_input xgmi_in_message; + union ta_xgmi_cmd_output xgmi_out_message; +}; + +#endif //_TA_XGMI_IF_H diff --git a/drivers/gpu/drm/amd/amdgpu/tonga_ih.c b/drivers/gpu/drm/amd/amdgpu/tonga_ih.c index 3abffd06b5c7..a20b711a6756 100644 --- a/drivers/gpu/drm/amd/amdgpu/tonga_ih.c +++ b/drivers/gpu/drm/amd/amdgpu/tonga_ih.c @@ -99,9 +99,9 @@ static void tonga_ih_disable_interrupts(struct amdgpu_device *adev) */ static int tonga_ih_irq_init(struct amdgpu_device *adev) { - int rb_bufsz; u32 interrupt_cntl, ih_rb_cntl, ih_doorbell_rtpr; - u64 wptr_off; + struct amdgpu_ih_ring *ih = &adev->irq.ih; + int rb_bufsz; /* disable irqs */ tonga_ih_disable_interrupts(adev); @@ -118,10 +118,7 @@ static int tonga_ih_irq_init(struct amdgpu_device *adev) WREG32(mmINTERRUPT_CNTL, interrupt_cntl); /* Ring Buffer base. [39:8] of 40-bit address of the beginning of the ring buffer*/ - if (adev->irq.ih.use_bus_addr) - WREG32(mmIH_RB_BASE, adev->irq.ih.rb_dma_addr >> 8); - else - WREG32(mmIH_RB_BASE, adev->irq.ih.gpu_addr >> 8); + WREG32(mmIH_RB_BASE, ih->gpu_addr >> 8); rb_bufsz = order_base_2(adev->irq.ih.ring_size / 4); ih_rb_cntl = REG_SET_FIELD(0, IH_RB_CNTL, WPTR_OVERFLOW_CLEAR, 1); @@ -136,12 +133,8 @@ static int tonga_ih_irq_init(struct amdgpu_device *adev) WREG32(mmIH_RB_CNTL, ih_rb_cntl); /* set the writeback address whether it's enabled or not */ - if (adev->irq.ih.use_bus_addr) - wptr_off = adev->irq.ih.rb_dma_addr + (adev->irq.ih.wptr_offs * 4); - else - wptr_off = adev->wb.gpu_addr + (adev->irq.ih.wptr_offs * 4); - WREG32(mmIH_RB_WPTR_ADDR_LO, lower_32_bits(wptr_off)); - WREG32(mmIH_RB_WPTR_ADDR_HI, upper_32_bits(wptr_off) & 0xFF); + WREG32(mmIH_RB_WPTR_ADDR_LO, lower_32_bits(ih->wptr_addr)); + WREG32(mmIH_RB_WPTR_ADDR_HI, upper_32_bits(ih->wptr_addr) & 0xFF); /* set rptr, wptr to 0 */ WREG32(mmIH_RB_RPTR, 0); @@ -193,14 +186,12 @@ static void tonga_ih_irq_disable(struct amdgpu_device *adev) * Used by cz_irq_process(VI). * Returns the value of the wptr. */ -static u32 tonga_ih_get_wptr(struct amdgpu_device *adev) +static u32 tonga_ih_get_wptr(struct amdgpu_device *adev, + struct amdgpu_ih_ring *ih) { u32 wptr, tmp; - if (adev->irq.ih.use_bus_addr) - wptr = le32_to_cpu(adev->irq.ih.ring[adev->irq.ih.wptr_offs]); - else - wptr = le32_to_cpu(adev->wb.wb[adev->irq.ih.wptr_offs]); + wptr = le32_to_cpu(*ih->wptr_cpu); if (REG_GET_FIELD(wptr, IH_RB_WPTR, RB_OVERFLOW)) { wptr = REG_SET_FIELD(wptr, IH_RB_WPTR, RB_OVERFLOW, 0); @@ -209,41 +200,13 @@ static u32 tonga_ih_get_wptr(struct amdgpu_device *adev) * this should allow us to catchup. */ dev_warn(adev->dev, "IH ring buffer overflow (0x%08X, 0x%08X, 0x%08X)\n", - wptr, adev->irq.ih.rptr, (wptr + 16) & adev->irq.ih.ptr_mask); - adev->irq.ih.rptr = (wptr + 16) & adev->irq.ih.ptr_mask; + wptr, ih->rptr, (wptr + 16) & ih->ptr_mask); + ih->rptr = (wptr + 16) & ih->ptr_mask; tmp = RREG32(mmIH_RB_CNTL); tmp = REG_SET_FIELD(tmp, IH_RB_CNTL, WPTR_OVERFLOW_CLEAR, 1); WREG32(mmIH_RB_CNTL, tmp); } - return (wptr & adev->irq.ih.ptr_mask); -} - -/** - * tonga_ih_prescreen_iv - prescreen an interrupt vector - * - * @adev: amdgpu_device pointer - * - * Returns true if the interrupt vector should be further processed. - */ -static bool tonga_ih_prescreen_iv(struct amdgpu_device *adev) -{ - u32 ring_index = adev->irq.ih.rptr >> 2; - u16 pasid; - - switch (le32_to_cpu(adev->irq.ih.ring[ring_index]) & 0xff) { - case 146: - case 147: - pasid = le32_to_cpu(adev->irq.ih.ring[ring_index + 2]) >> 16; - if (!pasid || amdgpu_vm_pasid_fault_credit(adev, pasid)) - return true; - break; - default: - /* Not a VM fault */ - return true; - } - - adev->irq.ih.rptr += 16; - return false; + return (wptr & ih->ptr_mask); } /** @@ -255,16 +218,17 @@ static bool tonga_ih_prescreen_iv(struct amdgpu_device *adev) * position and also advance the position. */ static void tonga_ih_decode_iv(struct amdgpu_device *adev, - struct amdgpu_iv_entry *entry) + struct amdgpu_ih_ring *ih, + struct amdgpu_iv_entry *entry) { /* wptr/rptr are in bytes! */ - u32 ring_index = adev->irq.ih.rptr >> 2; + u32 ring_index = ih->rptr >> 2; uint32_t dw[4]; - dw[0] = le32_to_cpu(adev->irq.ih.ring[ring_index + 0]); - dw[1] = le32_to_cpu(adev->irq.ih.ring[ring_index + 1]); - dw[2] = le32_to_cpu(adev->irq.ih.ring[ring_index + 2]); - dw[3] = le32_to_cpu(adev->irq.ih.ring[ring_index + 3]); + dw[0] = le32_to_cpu(ih->ring[ring_index + 0]); + dw[1] = le32_to_cpu(ih->ring[ring_index + 1]); + dw[2] = le32_to_cpu(ih->ring[ring_index + 2]); + dw[3] = le32_to_cpu(ih->ring[ring_index + 3]); entry->client_id = AMDGPU_IRQ_CLIENTID_LEGACY; entry->src_id = dw[0] & 0xff; @@ -274,7 +238,7 @@ static void tonga_ih_decode_iv(struct amdgpu_device *adev, entry->pasid = (dw[2] >> 16) & 0xffff; /* wptr/rptr are in bytes! */ - adev->irq.ih.rptr += 16; + ih->rptr += 16; } /** @@ -284,17 +248,15 @@ static void tonga_ih_decode_iv(struct amdgpu_device *adev, * * Set the IH ring buffer rptr. */ -static void tonga_ih_set_rptr(struct amdgpu_device *adev) +static void tonga_ih_set_rptr(struct amdgpu_device *adev, + struct amdgpu_ih_ring *ih) { - if (adev->irq.ih.use_doorbell) { + if (ih->use_doorbell) { /* XXX check if swapping is necessary on BE */ - if (adev->irq.ih.use_bus_addr) - adev->irq.ih.ring[adev->irq.ih.rptr_offs] = adev->irq.ih.rptr; - else - adev->wb.wb[adev->irq.ih.rptr_offs] = adev->irq.ih.rptr; - WDOORBELL32(adev->irq.ih.doorbell_index, adev->irq.ih.rptr); + *ih->rptr_cpu = ih->rptr; + WDOORBELL32(ih->doorbell_index, ih->rptr); } else { - WREG32(mmIH_RB_RPTR, adev->irq.ih.rptr); + WREG32(mmIH_RB_RPTR, ih->rptr); } } @@ -322,7 +284,7 @@ static int tonga_ih_sw_init(void *handle) return r; adev->irq.ih.use_doorbell = true; - adev->irq.ih.doorbell_index = AMDGPU_DOORBELL_IH; + adev->irq.ih.doorbell_index = adev->doorbell_index.ih; r = amdgpu_irq_init(adev); @@ -506,7 +468,6 @@ static const struct amd_ip_funcs tonga_ih_ip_funcs = { static const struct amdgpu_ih_funcs tonga_ih_funcs = { .get_wptr = tonga_ih_get_wptr, - .prescreen_iv = tonga_ih_prescreen_iv, .decode_iv = tonga_ih_decode_iv, .set_rptr = tonga_ih_set_rptr }; diff --git a/drivers/gpu/drm/amd/amdgpu/uvd_v4_2.c b/drivers/gpu/drm/amd/amdgpu/uvd_v4_2.c index 1fc17bf39fed..c4fb58667fd4 100644 --- a/drivers/gpu/drm/amd/amdgpu/uvd_v4_2.c +++ b/drivers/gpu/drm/amd/amdgpu/uvd_v4_2.c @@ -116,16 +116,16 @@ static int uvd_v4_2_sw_init(void *handle) if (r) return r; - r = amdgpu_uvd_resume(adev); - if (r) - return r; - ring = &adev->uvd.inst->ring; sprintf(ring->name, "uvd"); r = amdgpu_ring_init(adev, ring, 512, &adev->uvd.inst->irq, 0); if (r) return r; + r = amdgpu_uvd_resume(adev); + if (r) + return r; + r = amdgpu_uvd_entity_init(adev); return r; @@ -162,12 +162,9 @@ static int uvd_v4_2_hw_init(void *handle) uvd_v4_2_enable_mgcg(adev, true); amdgpu_asic_set_uvd_clocks(adev, 10000, 10000); - ring->ready = true; - r = amdgpu_ring_test_ring(ring); - if (r) { - ring->ready = false; + r = amdgpu_ring_test_helper(ring); + if (r) goto done; - } r = amdgpu_ring_alloc(ring, 10); if (r) { @@ -218,7 +215,7 @@ static int uvd_v4_2_hw_fini(void *handle) if (RREG32(mmUVD_STATUS) != 0) uvd_v4_2_stop(adev); - ring->ready = false; + ring->sched.ready = false; return 0; } @@ -484,11 +481,9 @@ static int uvd_v4_2_ring_test_ring(struct amdgpu_ring *ring) WREG32(mmUVD_CONTEXT_ID, 0xCAFEDEAD); r = amdgpu_ring_alloc(ring, 3); - if (r) { - DRM_ERROR("amdgpu: cp failed to lock ring %d (%d).\n", - ring->idx, r); + if (r) return r; - } + amdgpu_ring_write(ring, PACKET0(mmUVD_CONTEXT_ID, 0)); amdgpu_ring_write(ring, 0xDEADBEEF); amdgpu_ring_commit(ring); @@ -499,14 +494,9 @@ static int uvd_v4_2_ring_test_ring(struct amdgpu_ring *ring) DRM_UDELAY(1); } - if (i < adev->usec_timeout) { - DRM_DEBUG("ring test on %d succeeded in %d usecs\n", - ring->idx, i); - } else { - DRM_ERROR("amdgpu: ring %d test failed (0x%08X)\n", - ring->idx, tmp); - r = -EINVAL; - } + if (i >= adev->usec_timeout) + r = -ETIMEDOUT; + return r; } @@ -519,8 +509,9 @@ static int uvd_v4_2_ring_test_ring(struct amdgpu_ring *ring) * Write ring commands to execute the indirect buffer */ static void uvd_v4_2_ring_emit_ib(struct amdgpu_ring *ring, + struct amdgpu_job *job, struct amdgpu_ib *ib, - unsigned vmid, bool ctx_switch) + uint32_t flags) { amdgpu_ring_write(ring, PACKET0(mmUVD_RBC_IB_BASE, 0)); amdgpu_ring_write(ring, ib->gpu_addr); diff --git a/drivers/gpu/drm/amd/amdgpu/uvd_v5_0.c b/drivers/gpu/drm/amd/amdgpu/uvd_v5_0.c index fde6ad5ac9ab..52bd8a654734 100644 --- a/drivers/gpu/drm/amd/amdgpu/uvd_v5_0.c +++ b/drivers/gpu/drm/amd/amdgpu/uvd_v5_0.c @@ -113,16 +113,16 @@ static int uvd_v5_0_sw_init(void *handle) if (r) return r; - r = amdgpu_uvd_resume(adev); - if (r) - return r; - ring = &adev->uvd.inst->ring; sprintf(ring->name, "uvd"); r = amdgpu_ring_init(adev, ring, 512, &adev->uvd.inst->irq, 0); if (r) return r; + r = amdgpu_uvd_resume(adev); + if (r) + return r; + r = amdgpu_uvd_entity_init(adev); return r; @@ -158,12 +158,9 @@ static int uvd_v5_0_hw_init(void *handle) uvd_v5_0_set_clockgating_state(adev, AMD_CG_STATE_UNGATE); uvd_v5_0_enable_mgcg(adev, true); - ring->ready = true; - r = amdgpu_ring_test_ring(ring); - if (r) { - ring->ready = false; + r = amdgpu_ring_test_helper(ring); + if (r) goto done; - } r = amdgpu_ring_alloc(ring, 10); if (r) { @@ -215,7 +212,7 @@ static int uvd_v5_0_hw_fini(void *handle) if (RREG32(mmUVD_STATUS) != 0) uvd_v5_0_stop(adev); - ring->ready = false; + ring->sched.ready = false; return 0; } @@ -500,11 +497,8 @@ static int uvd_v5_0_ring_test_ring(struct amdgpu_ring *ring) WREG32(mmUVD_CONTEXT_ID, 0xCAFEDEAD); r = amdgpu_ring_alloc(ring, 3); - if (r) { - DRM_ERROR("amdgpu: cp failed to lock ring %d (%d).\n", - ring->idx, r); + if (r) return r; - } amdgpu_ring_write(ring, PACKET0(mmUVD_CONTEXT_ID, 0)); amdgpu_ring_write(ring, 0xDEADBEEF); amdgpu_ring_commit(ring); @@ -515,14 +509,9 @@ static int uvd_v5_0_ring_test_ring(struct amdgpu_ring *ring) DRM_UDELAY(1); } - if (i < adev->usec_timeout) { - DRM_DEBUG("ring test on %d succeeded in %d usecs\n", - ring->idx, i); - } else { - DRM_ERROR("amdgpu: ring %d test failed (0x%08X)\n", - ring->idx, tmp); - r = -EINVAL; - } + if (i >= adev->usec_timeout) + r = -ETIMEDOUT; + return r; } @@ -535,8 +524,9 @@ static int uvd_v5_0_ring_test_ring(struct amdgpu_ring *ring) * Write ring commands to execute the indirect buffer */ static void uvd_v5_0_ring_emit_ib(struct amdgpu_ring *ring, + struct amdgpu_job *job, struct amdgpu_ib *ib, - unsigned vmid, bool ctx_switch) + uint32_t flags) { amdgpu_ring_write(ring, PACKET0(mmUVD_LMI_RBC_IB_64BIT_BAR_LOW, 0)); amdgpu_ring_write(ring, lower_32_bits(ib->gpu_addr)); diff --git a/drivers/gpu/drm/amd/amdgpu/uvd_v6_0.c b/drivers/gpu/drm/amd/amdgpu/uvd_v6_0.c index 7a5b40275e8e..c9edddf9f88a 100644 --- a/drivers/gpu/drm/amd/amdgpu/uvd_v6_0.c +++ b/drivers/gpu/drm/amd/amdgpu/uvd_v6_0.c @@ -175,11 +175,8 @@ static int uvd_v6_0_enc_ring_test_ring(struct amdgpu_ring *ring) int r; r = amdgpu_ring_alloc(ring, 16); - if (r) { - DRM_ERROR("amdgpu: uvd enc failed to lock ring %d (%d).\n", - ring->idx, r); + if (r) return r; - } amdgpu_ring_write(ring, HEVC_ENC_CMD_END); amdgpu_ring_commit(ring); @@ -189,14 +186,8 @@ static int uvd_v6_0_enc_ring_test_ring(struct amdgpu_ring *ring) DRM_UDELAY(1); } - if (i < adev->usec_timeout) { - DRM_DEBUG("ring test on %d succeeded in %d usecs\n", - ring->idx, i); - } else { - DRM_ERROR("amdgpu: ring %d test failed\n", - ring->idx); + if (i >= adev->usec_timeout) r = -ETIMEDOUT; - } return r; } @@ -336,31 +327,24 @@ static int uvd_v6_0_enc_ring_test_ib(struct amdgpu_ring *ring, long timeout) long r; r = uvd_v6_0_enc_get_create_msg(ring, 1, NULL); - if (r) { - DRM_ERROR("amdgpu: failed to get create msg (%ld).\n", r); + if (r) goto error; - } r = uvd_v6_0_enc_get_destroy_msg(ring, 1, &fence); - if (r) { - DRM_ERROR("amdgpu: failed to get destroy ib (%ld).\n", r); + if (r) goto error; - } r = dma_fence_wait_timeout(fence, false, timeout); - if (r == 0) { - DRM_ERROR("amdgpu: IB test timed out.\n"); + if (r == 0) r = -ETIMEDOUT; - } else if (r < 0) { - DRM_ERROR("amdgpu: fence wait failed (%ld).\n", r); - } else { - DRM_DEBUG("ib test on ring %d succeeded\n", ring->idx); + else if (r > 0) r = 0; - } + error: dma_fence_put(fence); return r; } + static int uvd_v6_0_early_init(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; @@ -416,16 +400,16 @@ static int uvd_v6_0_sw_init(void *handle) DRM_INFO("UVD ENC is disabled\n"); } - r = amdgpu_uvd_resume(adev); - if (r) - return r; - ring = &adev->uvd.inst->ring; sprintf(ring->name, "uvd"); r = amdgpu_ring_init(adev, ring, 512, &adev->uvd.inst->irq, 0); if (r) return r; + r = amdgpu_uvd_resume(adev); + if (r) + return r; + if (uvd_v6_0_enc_support(adev)) { for (i = 0; i < adev->uvd.num_enc_rings; ++i) { ring = &adev->uvd.inst->ring_enc[i]; @@ -476,12 +460,9 @@ static int uvd_v6_0_hw_init(void *handle) uvd_v6_0_set_clockgating_state(adev, AMD_CG_STATE_UNGATE); uvd_v6_0_enable_mgcg(adev, true); - ring->ready = true; - r = amdgpu_ring_test_ring(ring); - if (r) { - ring->ready = false; + r = amdgpu_ring_test_helper(ring); + if (r) goto done; - } r = amdgpu_ring_alloc(ring, 10); if (r) { @@ -513,12 +494,9 @@ static int uvd_v6_0_hw_init(void *handle) if (uvd_v6_0_enc_support(adev)) { for (i = 0; i < adev->uvd.num_enc_rings; ++i) { ring = &adev->uvd.inst->ring_enc[i]; - ring->ready = true; - r = amdgpu_ring_test_ring(ring); - if (r) { - ring->ready = false; + r = amdgpu_ring_test_helper(ring); + if (r) goto done; - } } } @@ -548,7 +526,7 @@ static int uvd_v6_0_hw_fini(void *handle) if (RREG32(mmUVD_STATUS) != 0) uvd_v6_0_stop(adev); - ring->ready = false; + ring->sched.ready = false; return 0; } @@ -969,11 +947,9 @@ static int uvd_v6_0_ring_test_ring(struct amdgpu_ring *ring) WREG32(mmUVD_CONTEXT_ID, 0xCAFEDEAD); r = amdgpu_ring_alloc(ring, 3); - if (r) { - DRM_ERROR("amdgpu: cp failed to lock ring %d (%d).\n", - ring->idx, r); + if (r) return r; - } + amdgpu_ring_write(ring, PACKET0(mmUVD_CONTEXT_ID, 0)); amdgpu_ring_write(ring, 0xDEADBEEF); amdgpu_ring_commit(ring); @@ -984,14 +960,9 @@ static int uvd_v6_0_ring_test_ring(struct amdgpu_ring *ring) DRM_UDELAY(1); } - if (i < adev->usec_timeout) { - DRM_DEBUG("ring test on %d succeeded in %d usecs\n", - ring->idx, i); - } else { - DRM_ERROR("amdgpu: ring %d test failed (0x%08X)\n", - ring->idx, tmp); - r = -EINVAL; - } + if (i >= adev->usec_timeout) + r = -ETIMEDOUT; + return r; } @@ -1004,9 +975,12 @@ static int uvd_v6_0_ring_test_ring(struct amdgpu_ring *ring) * Write ring commands to execute the indirect buffer */ static void uvd_v6_0_ring_emit_ib(struct amdgpu_ring *ring, + struct amdgpu_job *job, struct amdgpu_ib *ib, - unsigned vmid, bool ctx_switch) + uint32_t flags) { + unsigned vmid = AMDGPU_JOB_GET_VMID(job); + amdgpu_ring_write(ring, PACKET0(mmUVD_LMI_RBC_IB_VMID, 0)); amdgpu_ring_write(ring, vmid); @@ -1027,8 +1001,12 @@ static void uvd_v6_0_ring_emit_ib(struct amdgpu_ring *ring, * Write enc ring commands to execute the indirect buffer */ static void uvd_v6_0_enc_ring_emit_ib(struct amdgpu_ring *ring, - struct amdgpu_ib *ib, unsigned int vmid, bool ctx_switch) + struct amdgpu_job *job, + struct amdgpu_ib *ib, + uint32_t flags) { + unsigned vmid = AMDGPU_JOB_GET_VMID(job); + amdgpu_ring_write(ring, HEVC_ENC_CMD_IB_VM); amdgpu_ring_write(ring, vmid); amdgpu_ring_write(ring, lower_32_bits(ib->gpu_addr)); diff --git a/drivers/gpu/drm/amd/amdgpu/uvd_v7_0.c b/drivers/gpu/drm/amd/amdgpu/uvd_v7_0.c index 58b39afcfb86..dc461df48da0 100644 --- a/drivers/gpu/drm/amd/amdgpu/uvd_v7_0.c +++ b/drivers/gpu/drm/amd/amdgpu/uvd_v7_0.c @@ -183,11 +183,8 @@ static int uvd_v7_0_enc_ring_test_ring(struct amdgpu_ring *ring) return 0; r = amdgpu_ring_alloc(ring, 16); - if (r) { - DRM_ERROR("amdgpu: uvd enc failed to lock (%d)ring %d (%d).\n", - ring->me, ring->idx, r); + if (r) return r; - } amdgpu_ring_write(ring, HEVC_ENC_CMD_END); amdgpu_ring_commit(ring); @@ -197,14 +194,8 @@ static int uvd_v7_0_enc_ring_test_ring(struct amdgpu_ring *ring) DRM_UDELAY(1); } - if (i < adev->usec_timeout) { - DRM_DEBUG("(%d)ring test on %d succeeded in %d usecs\n", - ring->me, ring->idx, i); - } else { - DRM_ERROR("amdgpu: (%d)ring %d test failed\n", - ring->me, ring->idx); + if (i >= adev->usec_timeout) r = -ETIMEDOUT; - } return r; } @@ -343,27 +334,19 @@ static int uvd_v7_0_enc_ring_test_ib(struct amdgpu_ring *ring, long timeout) long r; r = uvd_v7_0_enc_get_create_msg(ring, 1, NULL); - if (r) { - DRM_ERROR("amdgpu: (%d)failed to get create msg (%ld).\n", ring->me, r); + if (r) goto error; - } r = uvd_v7_0_enc_get_destroy_msg(ring, 1, &fence); - if (r) { - DRM_ERROR("amdgpu: (%d)failed to get destroy ib (%ld).\n", ring->me, r); + if (r) goto error; - } r = dma_fence_wait_timeout(fence, false, timeout); - if (r == 0) { - DRM_ERROR("amdgpu: (%d)IB test timed out.\n", ring->me); + if (r == 0) r = -ETIMEDOUT; - } else if (r < 0) { - DRM_ERROR("amdgpu: (%d)fence wait failed (%ld).\n", ring->me, r); - } else { - DRM_DEBUG("ib test on (%d)ring %d succeeded\n", ring->me, ring->idx); + else if (r > 0) r = 0; - } + error: dma_fence_put(fence); return r; @@ -447,16 +430,12 @@ static int uvd_v7_0_sw_init(void *handle) DRM_INFO("PSP loading UVD firmware\n"); } - r = amdgpu_uvd_resume(adev); - if (r) - return r; - for (j = 0; j < adev->uvd.num_uvd_inst; j++) { if (adev->uvd.harvest_config & (1 << j)) continue; if (!amdgpu_sriov_vf(adev)) { ring = &adev->uvd.inst[j].ring; - sprintf(ring->name, "uvd<%d>", j); + sprintf(ring->name, "uvd_%d", ring->me); r = amdgpu_ring_init(adev, ring, 512, &adev->uvd.inst[j].irq, 0); if (r) return r; @@ -464,7 +443,7 @@ static int uvd_v7_0_sw_init(void *handle) for (i = 0; i < adev->uvd.num_enc_rings; ++i) { ring = &adev->uvd.inst[j].ring_enc[i]; - sprintf(ring->name, "uvd_enc%d<%d>", i, j); + sprintf(ring->name, "uvd_enc_%d.%d", ring->me, i); if (amdgpu_sriov_vf(adev)) { ring->use_doorbell = true; @@ -472,9 +451,9 @@ static int uvd_v7_0_sw_init(void *handle) * sriov, so set unused location for other unused rings. */ if (i == 0) - ring->doorbell_index = AMDGPU_DOORBELL64_UVD_RING0_1 * 2; + ring->doorbell_index = adev->doorbell_index.uvd_vce.uvd_ring0_1 * 2; else - ring->doorbell_index = AMDGPU_DOORBELL64_UVD_RING2_3 * 2 + 1; + ring->doorbell_index = adev->doorbell_index.uvd_vce.uvd_ring2_3 * 2 + 1; } r = amdgpu_ring_init(adev, ring, 512, &adev->uvd.inst[j].irq, 0); if (r) @@ -482,6 +461,10 @@ static int uvd_v7_0_sw_init(void *handle) } } + r = amdgpu_uvd_resume(adev); + if (r) + return r; + r = amdgpu_uvd_entity_init(adev); if (r) return r; @@ -540,12 +523,9 @@ static int uvd_v7_0_hw_init(void *handle) ring = &adev->uvd.inst[j].ring; if (!amdgpu_sriov_vf(adev)) { - ring->ready = true; - r = amdgpu_ring_test_ring(ring); - if (r) { - ring->ready = false; + r = amdgpu_ring_test_helper(ring); + if (r) goto done; - } r = amdgpu_ring_alloc(ring, 10); if (r) { @@ -582,12 +562,9 @@ static int uvd_v7_0_hw_init(void *handle) for (i = 0; i < adev->uvd.num_enc_rings; ++i) { ring = &adev->uvd.inst[j].ring_enc[i]; - ring->ready = true; - r = amdgpu_ring_test_ring(ring); - if (r) { - ring->ready = false; + r = amdgpu_ring_test_helper(ring); + if (r) goto done; - } } } done: @@ -619,7 +596,7 @@ static int uvd_v7_0_hw_fini(void *handle) for (i = 0; i < adev->uvd.num_uvd_inst; ++i) { if (adev->uvd.harvest_config & (1 << i)) continue; - adev->uvd.inst[i].ring.ready = false; + adev->uvd.inst[i].ring.sched.ready = false; } return 0; @@ -1235,11 +1212,9 @@ static int uvd_v7_0_ring_test_ring(struct amdgpu_ring *ring) WREG32_SOC15(UVD, ring->me, mmUVD_CONTEXT_ID, 0xCAFEDEAD); r = amdgpu_ring_alloc(ring, 3); - if (r) { - DRM_ERROR("amdgpu: (%d)cp failed to lock ring %d (%d).\n", - ring->me, ring->idx, r); + if (r) return r; - } + amdgpu_ring_write(ring, PACKET0(SOC15_REG_OFFSET(UVD, ring->me, mmUVD_CONTEXT_ID), 0)); amdgpu_ring_write(ring, 0xDEADBEEF); @@ -1251,14 +1226,9 @@ static int uvd_v7_0_ring_test_ring(struct amdgpu_ring *ring) DRM_UDELAY(1); } - if (i < adev->usec_timeout) { - DRM_DEBUG("(%d)ring test on %d succeeded in %d usecs\n", - ring->me, ring->idx, i); - } else { - DRM_ERROR("(%d)amdgpu: ring %d test failed (0x%08X)\n", - ring->me, ring->idx, tmp); - r = -EINVAL; - } + if (i >= adev->usec_timeout) + r = -ETIMEDOUT; + return r; } @@ -1300,10 +1270,12 @@ static int uvd_v7_0_ring_patch_cs_in_place(struct amdgpu_cs_parser *p, * Write ring commands to execute the indirect buffer */ static void uvd_v7_0_ring_emit_ib(struct amdgpu_ring *ring, + struct amdgpu_job *job, struct amdgpu_ib *ib, - unsigned vmid, bool ctx_switch) + uint32_t flags) { struct amdgpu_device *adev = ring->adev; + unsigned vmid = AMDGPU_JOB_GET_VMID(job); amdgpu_ring_write(ring, PACKET0(SOC15_REG_OFFSET(UVD, ring->me, mmUVD_LMI_RBC_IB_VMID), 0)); @@ -1329,8 +1301,12 @@ static void uvd_v7_0_ring_emit_ib(struct amdgpu_ring *ring, * Write enc ring commands to execute the indirect buffer */ static void uvd_v7_0_enc_ring_emit_ib(struct amdgpu_ring *ring, - struct amdgpu_ib *ib, unsigned int vmid, bool ctx_switch) + struct amdgpu_job *job, + struct amdgpu_ib *ib, + uint32_t flags) { + unsigned vmid = AMDGPU_JOB_GET_VMID(job); + amdgpu_ring_write(ring, HEVC_ENC_CMD_IB_VM); amdgpu_ring_write(ring, vmid); amdgpu_ring_write(ring, lower_32_bits(ib->gpu_addr)); diff --git a/drivers/gpu/drm/amd/amdgpu/vce_v2_0.c b/drivers/gpu/drm/amd/amdgpu/vce_v2_0.c index ea28828360d3..bed78a778e3f 100644 --- a/drivers/gpu/drm/amd/amdgpu/vce_v2_0.c +++ b/drivers/gpu/drm/amd/amdgpu/vce_v2_0.c @@ -463,15 +463,11 @@ static int vce_v2_0_hw_init(void *handle) amdgpu_asic_set_vce_clocks(adev, 10000, 10000); vce_v2_0_enable_mgcg(adev, true, false); - for (i = 0; i < adev->vce.num_rings; i++) - adev->vce.ring[i].ready = false; for (i = 0; i < adev->vce.num_rings; i++) { - r = amdgpu_ring_test_ring(&adev->vce.ring[i]); + r = amdgpu_ring_test_helper(&adev->vce.ring[i]); if (r) return r; - else - adev->vce.ring[i].ready = true; } DRM_INFO("VCE initialized successfully.\n"); diff --git a/drivers/gpu/drm/amd/amdgpu/vce_v3_0.c b/drivers/gpu/drm/amd/amdgpu/vce_v3_0.c index 6dbd39730070..6ec65cf11112 100644 --- a/drivers/gpu/drm/amd/amdgpu/vce_v3_0.c +++ b/drivers/gpu/drm/amd/amdgpu/vce_v3_0.c @@ -37,7 +37,6 @@ #include "gca/gfx_8_0_d.h" #include "smu/smu_7_1_2_d.h" #include "smu/smu_7_1_2_sh_mask.h" -#include "gca/gfx_8_0_d.h" #include "gca/gfx_8_0_sh_mask.h" #include "ivsrcid/ivsrcid_vislands30.h" @@ -474,15 +473,10 @@ static int vce_v3_0_hw_init(void *handle) amdgpu_asic_set_vce_clocks(adev, 10000, 10000); - for (i = 0; i < adev->vce.num_rings; i++) - adev->vce.ring[i].ready = false; - for (i = 0; i < adev->vce.num_rings; i++) { - r = amdgpu_ring_test_ring(&adev->vce.ring[i]); + r = amdgpu_ring_test_helper(&adev->vce.ring[i]); if (r) return r; - else - adev->vce.ring[i].ready = true; } DRM_INFO("VCE initialized successfully.\n"); @@ -838,8 +832,12 @@ out: } static void vce_v3_0_ring_emit_ib(struct amdgpu_ring *ring, - struct amdgpu_ib *ib, unsigned int vmid, bool ctx_switch) + struct amdgpu_job *job, + struct amdgpu_ib *ib, + uint32_t flags) { + unsigned vmid = AMDGPU_JOB_GET_VMID(job); + amdgpu_ring_write(ring, VCE_CMD_IB_VM); amdgpu_ring_write(ring, vmid); amdgpu_ring_write(ring, lower_32_bits(ib->gpu_addr)); diff --git a/drivers/gpu/drm/amd/amdgpu/vce_v4_0.c b/drivers/gpu/drm/amd/amdgpu/vce_v4_0.c index 1c9471890bf7..aadc3e66ebd7 100644 --- a/drivers/gpu/drm/amd/amdgpu/vce_v4_0.c +++ b/drivers/gpu/drm/amd/amdgpu/vce_v4_0.c @@ -466,9 +466,9 @@ static int vce_v4_0_sw_init(void *handle) * so set unused location for other unused rings. */ if (i == 0) - ring->doorbell_index = AMDGPU_DOORBELL64_VCE_RING0_1 * 2; + ring->doorbell_index = adev->doorbell_index.uvd_vce.vce_ring0_1 * 2; else - ring->doorbell_index = AMDGPU_DOORBELL64_VCE_RING2_3 * 2 + 1; + ring->doorbell_index = adev->doorbell_index.uvd_vce.vce_ring2_3 * 2 + 1; } r = amdgpu_ring_init(adev, ring, 512, &adev->vce.irq, 0); if (r) @@ -519,15 +519,10 @@ static int vce_v4_0_hw_init(void *handle) if (r) return r; - for (i = 0; i < adev->vce.num_rings; i++) - adev->vce.ring[i].ready = false; - for (i = 0; i < adev->vce.num_rings; i++) { - r = amdgpu_ring_test_ring(&adev->vce.ring[i]); + r = amdgpu_ring_test_helper(&adev->vce.ring[i]); if (r) return r; - else - adev->vce.ring[i].ready = true; } DRM_INFO("VCE initialized successfully.\n"); @@ -549,7 +544,7 @@ static int vce_v4_0_hw_fini(void *handle) } for (i = 0; i < adev->vce.num_rings; i++) - adev->vce.ring[i].ready = false; + adev->vce.ring[i].sched.ready = false; return 0; } @@ -951,9 +946,11 @@ static int vce_v4_0_set_powergating_state(void *handle, } #endif -static void vce_v4_0_ring_emit_ib(struct amdgpu_ring *ring, - struct amdgpu_ib *ib, unsigned int vmid, bool ctx_switch) +static void vce_v4_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); + amdgpu_ring_write(ring, VCE_CMD_IB_VM); amdgpu_ring_write(ring, vmid); amdgpu_ring_write(ring, lower_32_bits(ib->gpu_addr)); diff --git a/drivers/gpu/drm/amd/amdgpu/vcn_v1_0.c b/drivers/gpu/drm/amd/amdgpu/vcn_v1_0.c index 322e09b5b448..3dbc51f9d3b9 100644 --- a/drivers/gpu/drm/amd/amdgpu/vcn_v1_0.c +++ b/drivers/gpu/drm/amd/amdgpu/vcn_v1_0.c @@ -177,30 +177,22 @@ static int vcn_v1_0_hw_init(void *handle) struct amdgpu_ring *ring = &adev->vcn.ring_dec; int i, r; - ring->ready = true; - r = amdgpu_ring_test_ring(ring); - if (r) { - ring->ready = false; + r = amdgpu_ring_test_helper(ring); + if (r) goto done; - } for (i = 0; i < adev->vcn.num_enc_rings; ++i) { ring = &adev->vcn.ring_enc[i]; - ring->ready = true; - r = amdgpu_ring_test_ring(ring); - if (r) { - ring->ready = false; + ring->sched.ready = true; + r = amdgpu_ring_test_helper(ring); + if (r) goto done; - } } ring = &adev->vcn.ring_jpeg; - ring->ready = true; - r = amdgpu_ring_test_ring(ring); - if (r) { - ring->ready = false; + r = amdgpu_ring_test_helper(ring); + if (r) goto done; - } done: if (!r) @@ -222,10 +214,11 @@ static int vcn_v1_0_hw_fini(void *handle) struct amdgpu_device *adev = (struct amdgpu_device *)handle; struct amdgpu_ring *ring = &adev->vcn.ring_dec; - if (RREG32_SOC15(VCN, 0, mmUVD_STATUS)) + if ((adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG) || + RREG32_SOC15(VCN, 0, mmUVD_STATUS)) vcn_v1_0_set_powergating_state(adev, AMD_PG_STATE_GATE); - ring->ready = false; + ring->sched.ready = false; return 0; } @@ -1095,7 +1088,8 @@ static int vcn_v1_0_start_dpg_mode(struct amdgpu_device *adev) WREG32_P(SOC15_REG_OFFSET(UVD, 0, mmUVD_RBC_RB_CNTL), 0, ~UVD_RBC_RB_CNTL__RB_NO_FETCH_MASK); - /* initialize wptr */ + /* initialize JPEG wptr */ + ring = &adev->vcn.ring_jpeg; ring->wptr = RREG32_SOC15(UVD, 0, mmUVD_JRBC_RB_WPTR); /* copy patch commands to the jpeg ring */ @@ -1167,21 +1161,29 @@ static int vcn_v1_0_stop_spg_mode(struct amdgpu_device *adev) static int vcn_v1_0_stop_dpg_mode(struct amdgpu_device *adev) { int ret_code = 0; + uint32_t tmp; /* Wait for power status to be UVD_POWER_STATUS__UVD_POWER_STATUS_TILES_OFF */ SOC15_WAIT_ON_RREG(UVD, 0, mmUVD_POWER_STATUS, UVD_POWER_STATUS__UVD_POWER_STATUS_TILES_OFF, UVD_POWER_STATUS__UVD_POWER_STATUS_MASK, ret_code); - if (!ret_code) { - int tmp = RREG32_SOC15(UVD, 0, mmUVD_RBC_RB_WPTR) & 0x7FFFFFFF; - /* wait for read ptr to be equal to write ptr */ - SOC15_WAIT_ON_RREG(UVD, 0, mmUVD_RBC_RB_RPTR, tmp, 0xFFFFFFFF, ret_code); + /* wait for read ptr to be equal to write ptr */ + tmp = RREG32_SOC15(UVD, 0, mmUVD_RB_WPTR); + SOC15_WAIT_ON_RREG(UVD, 0, mmUVD_RB_RPTR, tmp, 0xFFFFFFFF, ret_code); - SOC15_WAIT_ON_RREG(UVD, 0, mmUVD_POWER_STATUS, - UVD_POWER_STATUS__UVD_POWER_STATUS_TILES_OFF, - UVD_POWER_STATUS__UVD_POWER_STATUS_MASK, ret_code); - } + tmp = RREG32_SOC15(UVD, 0, mmUVD_RB_WPTR2); + SOC15_WAIT_ON_RREG(UVD, 0, mmUVD_RB_RPTR2, tmp, 0xFFFFFFFF, ret_code); + + tmp = RREG32_SOC15(UVD, 0, mmUVD_JRBC_RB_WPTR); + SOC15_WAIT_ON_RREG(UVD, 0, mmUVD_JRBC_RB_RPTR, tmp, 0xFFFFFFFF, ret_code); + + tmp = RREG32_SOC15(UVD, 0, mmUVD_RBC_RB_WPTR) & 0x7FFFFFFF; + SOC15_WAIT_ON_RREG(UVD, 0, mmUVD_RBC_RB_RPTR, tmp, 0xFFFFFFFF, ret_code); + + SOC15_WAIT_ON_RREG(UVD, 0, mmUVD_POWER_STATUS, + UVD_POWER_STATUS__UVD_POWER_STATUS_TILES_OFF, + UVD_POWER_STATUS__UVD_POWER_STATUS_MASK, ret_code); /* disable dynamic power gating mode */ WREG32_P(SOC15_REG_OFFSET(UVD, 0, mmUVD_POWER_STATUS), 0, @@ -1367,10 +1369,12 @@ static void vcn_v1_0_dec_ring_emit_fence(struct amdgpu_ring *ring, u64 addr, u64 * Write ring commands to execute the indirect buffer */ static void vcn_v1_0_dec_ring_emit_ib(struct amdgpu_ring *ring, - struct amdgpu_ib *ib, - unsigned vmid, bool ctx_switch) + struct amdgpu_job *job, + struct amdgpu_ib *ib, + uint32_t flags) { struct amdgpu_device *adev = ring->adev; + unsigned vmid = AMDGPU_JOB_GET_VMID(job); amdgpu_ring_write(ring, PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_LMI_RBC_IB_VMID), 0)); @@ -1525,8 +1529,12 @@ static void vcn_v1_0_enc_ring_insert_end(struct amdgpu_ring *ring) * Write enc ring commands to execute the indirect buffer */ static void vcn_v1_0_enc_ring_emit_ib(struct amdgpu_ring *ring, - struct amdgpu_ib *ib, unsigned int vmid, bool ctx_switch) + struct amdgpu_job *job, + struct amdgpu_ib *ib, + uint32_t flags) { + unsigned vmid = AMDGPU_JOB_GET_VMID(job); + amdgpu_ring_write(ring, VCN_ENC_CMD_IB); amdgpu_ring_write(ring, vmid); amdgpu_ring_write(ring, lower_32_bits(ib->gpu_addr)); @@ -1726,10 +1734,12 @@ static void vcn_v1_0_jpeg_ring_emit_fence(struct amdgpu_ring *ring, u64 addr, u6 * Write ring commands to execute the indirect buffer. */ static void vcn_v1_0_jpeg_ring_emit_ib(struct amdgpu_ring *ring, - struct amdgpu_ib *ib, - unsigned vmid, bool ctx_switch) + struct amdgpu_job *job, + struct amdgpu_ib *ib, + uint32_t flags) { struct amdgpu_device *adev = ring->adev; + unsigned vmid = AMDGPU_JOB_GET_VMID(job); amdgpu_ring_write(ring, PACKETJ(SOC15_REG_OFFSET(UVD, 0, mmUVD_LMI_JRBC_IB_VMID), 0, 0, PACKETJ_TYPE0)); diff --git a/drivers/gpu/drm/amd/amdgpu/vega10_ih.c b/drivers/gpu/drm/amd/amdgpu/vega10_ih.c index a0fda6f9252a..6d1f804277f8 100644 --- a/drivers/gpu/drm/amd/amdgpu/vega10_ih.c +++ b/drivers/gpu/drm/amd/amdgpu/vega10_ih.c @@ -50,6 +50,22 @@ static void vega10_ih_enable_interrupts(struct amdgpu_device *adev) ih_rb_cntl = REG_SET_FIELD(ih_rb_cntl, IH_RB_CNTL, ENABLE_INTR, 1); WREG32_SOC15(OSSSYS, 0, mmIH_RB_CNTL, ih_rb_cntl); adev->irq.ih.enabled = true; + + if (adev->irq.ih1.ring_size) { + ih_rb_cntl = RREG32_SOC15(OSSSYS, 0, mmIH_RB_CNTL_RING1); + ih_rb_cntl = REG_SET_FIELD(ih_rb_cntl, IH_RB_CNTL_RING1, + RB_ENABLE, 1); + WREG32_SOC15(OSSSYS, 0, mmIH_RB_CNTL_RING1, ih_rb_cntl); + adev->irq.ih1.enabled = true; + } + + if (adev->irq.ih2.ring_size) { + ih_rb_cntl = RREG32_SOC15(OSSSYS, 0, mmIH_RB_CNTL_RING2); + ih_rb_cntl = REG_SET_FIELD(ih_rb_cntl, IH_RB_CNTL_RING2, + RB_ENABLE, 1); + WREG32_SOC15(OSSSYS, 0, mmIH_RB_CNTL_RING2, ih_rb_cntl); + adev->irq.ih2.enabled = true; + } } /** @@ -71,6 +87,53 @@ static void vega10_ih_disable_interrupts(struct amdgpu_device *adev) WREG32_SOC15(OSSSYS, 0, mmIH_RB_WPTR, 0); adev->irq.ih.enabled = false; adev->irq.ih.rptr = 0; + + if (adev->irq.ih1.ring_size) { + ih_rb_cntl = RREG32_SOC15(OSSSYS, 0, mmIH_RB_CNTL_RING1); + ih_rb_cntl = REG_SET_FIELD(ih_rb_cntl, IH_RB_CNTL_RING1, + RB_ENABLE, 0); + WREG32_SOC15(OSSSYS, 0, mmIH_RB_CNTL_RING1, ih_rb_cntl); + /* set rptr, wptr to 0 */ + WREG32_SOC15(OSSSYS, 0, mmIH_RB_RPTR_RING1, 0); + WREG32_SOC15(OSSSYS, 0, mmIH_RB_WPTR_RING1, 0); + adev->irq.ih1.enabled = false; + adev->irq.ih1.rptr = 0; + } + + if (adev->irq.ih2.ring_size) { + ih_rb_cntl = RREG32_SOC15(OSSSYS, 0, mmIH_RB_CNTL_RING2); + ih_rb_cntl = REG_SET_FIELD(ih_rb_cntl, IH_RB_CNTL_RING2, + RB_ENABLE, 0); + WREG32_SOC15(OSSSYS, 0, mmIH_RB_CNTL_RING2, ih_rb_cntl); + /* set rptr, wptr to 0 */ + WREG32_SOC15(OSSSYS, 0, mmIH_RB_RPTR_RING2, 0); + WREG32_SOC15(OSSSYS, 0, mmIH_RB_WPTR_RING2, 0); + adev->irq.ih2.enabled = false; + adev->irq.ih2.rptr = 0; + } +} + +static uint32_t vega10_ih_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 ? 1 : 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; } /** @@ -86,50 +149,32 @@ static void vega10_ih_disable_interrupts(struct amdgpu_device *adev) */ static int vega10_ih_irq_init(struct amdgpu_device *adev) { + struct amdgpu_ih_ring *ih; int ret = 0; - int rb_bufsz; u32 ih_rb_cntl, ih_doorbell_rtpr; u32 tmp; - u64 wptr_off; /* disable irqs */ vega10_ih_disable_interrupts(adev); adev->nbio_funcs->ih_control(adev); - ih_rb_cntl = RREG32_SOC15(OSSSYS, 0, mmIH_RB_CNTL); + ih = &adev->irq.ih; /* Ring Buffer base. [39:8] of 40-bit address of the beginning of the ring buffer*/ - if (adev->irq.ih.use_bus_addr) { - WREG32_SOC15(OSSSYS, 0, mmIH_RB_BASE, adev->irq.ih.rb_dma_addr >> 8); - WREG32_SOC15(OSSSYS, 0, mmIH_RB_BASE_HI, ((u64)adev->irq.ih.rb_dma_addr >> 40) & 0xff); - ih_rb_cntl = REG_SET_FIELD(ih_rb_cntl, IH_RB_CNTL, MC_SPACE, 1); - } else { - WREG32_SOC15(OSSSYS, 0, mmIH_RB_BASE, adev->irq.ih.gpu_addr >> 8); - WREG32_SOC15(OSSSYS, 0, mmIH_RB_BASE_HI, (adev->irq.ih.gpu_addr >> 40) & 0xff); - ih_rb_cntl = REG_SET_FIELD(ih_rb_cntl, IH_RB_CNTL, MC_SPACE, 4); - } - rb_bufsz = order_base_2(adev->irq.ih.ring_size / 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); - - if (adev->irq.msi_enabled) - ih_rb_cntl = REG_SET_FIELD(ih_rb_cntl, IH_RB_CNTL, RPTR_REARM, 1); + WREG32_SOC15(OSSSYS, 0, mmIH_RB_BASE, ih->gpu_addr >> 8); + WREG32_SOC15(OSSSYS, 0, mmIH_RB_BASE_HI, (ih->gpu_addr >> 40) & 0xff); + ih_rb_cntl = RREG32_SOC15(OSSSYS, 0, mmIH_RB_CNTL); + ih_rb_cntl = vega10_ih_rb_cntl(ih, ih_rb_cntl); + ih_rb_cntl = REG_SET_FIELD(ih_rb_cntl, IH_RB_CNTL, RPTR_REARM, + !!adev->irq.msi_enabled); WREG32_SOC15(OSSSYS, 0, mmIH_RB_CNTL, ih_rb_cntl); /* set the writeback address whether it's enabled or not */ - if (adev->irq.ih.use_bus_addr) - wptr_off = adev->irq.ih.rb_dma_addr + (adev->irq.ih.wptr_offs * 4); - else - wptr_off = adev->wb.gpu_addr + (adev->irq.ih.wptr_offs * 4); - WREG32_SOC15(OSSSYS, 0, mmIH_RB_WPTR_ADDR_LO, lower_32_bits(wptr_off)); - WREG32_SOC15(OSSSYS, 0, mmIH_RB_WPTR_ADDR_HI, upper_32_bits(wptr_off) & 0xFFFF); + WREG32_SOC15(OSSSYS, 0, mmIH_RB_WPTR_ADDR_LO, + lower_32_bits(ih->wptr_addr)); + WREG32_SOC15(OSSSYS, 0, mmIH_RB_WPTR_ADDR_HI, + upper_32_bits(ih->wptr_addr) & 0xFFFF); /* set rptr, wptr to 0 */ WREG32_SOC15(OSSSYS, 0, mmIH_RB_RPTR, 0); @@ -137,17 +182,48 @@ static int vega10_ih_irq_init(struct amdgpu_device *adev) ih_doorbell_rtpr = RREG32_SOC15(OSSSYS, 0, mmIH_DOORBELL_RPTR); if (adev->irq.ih.use_doorbell) { - ih_doorbell_rtpr = REG_SET_FIELD(ih_doorbell_rtpr, IH_DOORBELL_RPTR, - OFFSET, adev->irq.ih.doorbell_index); - ih_doorbell_rtpr = REG_SET_FIELD(ih_doorbell_rtpr, IH_DOORBELL_RPTR, + ih_doorbell_rtpr = REG_SET_FIELD(ih_doorbell_rtpr, + IH_DOORBELL_RPTR, OFFSET, + adev->irq.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, + ih_doorbell_rtpr = REG_SET_FIELD(ih_doorbell_rtpr, + IH_DOORBELL_RPTR, ENABLE, 0); } WREG32_SOC15(OSSSYS, 0, mmIH_DOORBELL_RPTR, ih_doorbell_rtpr); - adev->nbio_funcs->ih_doorbell_range(adev, adev->irq.ih.use_doorbell, - adev->irq.ih.doorbell_index); + + ih = &adev->irq.ih1; + if (ih->ring_size) { + WREG32_SOC15(OSSSYS, 0, mmIH_RB_BASE_RING1, ih->gpu_addr >> 8); + WREG32_SOC15(OSSSYS, 0, mmIH_RB_BASE_HI_RING1, + (ih->gpu_addr >> 40) & 0xff); + + ih_rb_cntl = RREG32_SOC15(OSSSYS, 0, mmIH_RB_CNTL_RING1); + ih_rb_cntl = vega10_ih_rb_cntl(ih, ih_rb_cntl); + WREG32_SOC15(OSSSYS, 0, mmIH_RB_CNTL_RING1, ih_rb_cntl); + + /* set rptr, wptr to 0 */ + WREG32_SOC15(OSSSYS, 0, mmIH_RB_RPTR_RING1, 0); + WREG32_SOC15(OSSSYS, 0, mmIH_RB_WPTR_RING1, 0); + } + + ih = &adev->irq.ih2; + if (ih->ring_size) { + WREG32_SOC15(OSSSYS, 0, mmIH_RB_BASE_RING2, ih->gpu_addr >> 8); + WREG32_SOC15(OSSSYS, 0, mmIH_RB_BASE_HI_RING2, + (ih->gpu_addr >> 40) & 0xff); + + ih_rb_cntl = RREG32_SOC15(OSSSYS, 0, mmIH_RB_CNTL_RING1); + ih_rb_cntl = vega10_ih_rb_cntl(ih, ih_rb_cntl); + WREG32_SOC15(OSSSYS, 0, mmIH_RB_CNTL_RING2, ih_rb_cntl); + + /* set rptr, wptr to 0 */ + WREG32_SOC15(OSSSYS, 0, mmIH_RB_RPTR_RING2, 0); + WREG32_SOC15(OSSSYS, 0, mmIH_RB_WPTR_RING2, 0); + } tmp = RREG32_SOC15(OSSSYS, 0, mmIH_STORM_CLIENT_LIST_CNTL); tmp = REG_SET_FIELD(tmp, IH_STORM_CLIENT_LIST_CNTL, @@ -191,116 +267,58 @@ static void vega10_ih_irq_disable(struct amdgpu_device *adev) * ring buffer overflow and deal with it. * Returns the value of the wptr. */ -static u32 vega10_ih_get_wptr(struct amdgpu_device *adev) +static u32 vega10_ih_get_wptr(struct amdgpu_device *adev, + struct amdgpu_ih_ring *ih) { - u32 wptr, tmp; + u32 wptr, reg, tmp; - if (adev->irq.ih.use_bus_addr) - wptr = le32_to_cpu(adev->irq.ih.ring[adev->irq.ih.wptr_offs]); - else - wptr = le32_to_cpu(adev->wb.wb[adev->irq.ih.wptr_offs]); - - if (REG_GET_FIELD(wptr, IH_RB_WPTR, RB_OVERFLOW)) { - 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 catchup. - */ - tmp = (wptr + 32) & adev->irq.ih.ptr_mask; - dev_warn(adev->dev, "IH ring buffer overflow (0x%08X, 0x%08X, 0x%08X)\n", - wptr, adev->irq.ih.rptr, tmp); - adev->irq.ih.rptr = tmp; - - tmp = RREG32_NO_KIQ(SOC15_REG_OFFSET(OSSSYS, 0, mmIH_RB_CNTL)); - tmp = REG_SET_FIELD(tmp, IH_RB_CNTL, WPTR_OVERFLOW_CLEAR, 1); - WREG32_NO_KIQ(SOC15_REG_OFFSET(OSSSYS, 0, mmIH_RB_CNTL), tmp); - } - return (wptr & adev->irq.ih.ptr_mask); -} + wptr = le32_to_cpu(*ih->wptr_cpu); -/** - * vega10_ih_prescreen_iv - prescreen an interrupt vector - * - * @adev: amdgpu_device pointer - * - * Returns true if the interrupt vector should be further processed. - */ -static bool vega10_ih_prescreen_iv(struct amdgpu_device *adev) -{ - u32 ring_index = adev->irq.ih.rptr >> 2; - u32 dw0, dw3, dw4, dw5; - u16 pasid; - u64 addr, key; - struct amdgpu_vm *vm; - int r; + if (!REG_GET_FIELD(wptr, IH_RB_WPTR, RB_OVERFLOW)) + goto out; - dw0 = le32_to_cpu(adev->irq.ih.ring[ring_index + 0]); - dw3 = le32_to_cpu(adev->irq.ih.ring[ring_index + 3]); - dw4 = le32_to_cpu(adev->irq.ih.ring[ring_index + 4]); - dw5 = le32_to_cpu(adev->irq.ih.ring[ring_index + 5]); + /* Double check that the overflow wasn't already cleared. */ - /* Filter retry page faults, let only the first one pass. If - * there are too many outstanding faults, ignore them until - * some faults get cleared. - */ - switch (dw0 & 0xff) { - case SOC15_IH_CLIENTID_VMC: - case SOC15_IH_CLIENTID_UTCL2: - break; - default: - /* Not a VM fault */ - return true; - } + if (ih == &adev->irq.ih) + reg = SOC15_REG_OFFSET(OSSSYS, 0, mmIH_RB_WPTR); + else if (ih == &adev->irq.ih1) + reg = SOC15_REG_OFFSET(OSSSYS, 0, mmIH_RB_WPTR_RING1); + else if (ih == &adev->irq.ih2) + reg = SOC15_REG_OFFSET(OSSSYS, 0, mmIH_RB_WPTR_RING2); + else + BUG(); - pasid = dw3 & 0xffff; - /* No PASID, can't identify faulting process */ - if (!pasid) - return true; + wptr = RREG32_NO_KIQ(reg); + if (!REG_GET_FIELD(wptr, IH_RB_WPTR, RB_OVERFLOW)) + goto out; - /* Not a retry fault, check fault credit */ - if (!(dw5 & 0x80)) { - if (!amdgpu_vm_pasid_fault_credit(adev, pasid)) - goto ignore_iv; - return true; - } + wptr = REG_SET_FIELD(wptr, IH_RB_WPTR, RB_OVERFLOW, 0); - /* Track retry faults in per-VM fault FIFO. */ - spin_lock(&adev->vm_manager.pasid_lock); - vm = idr_find(&adev->vm_manager.pasid_idr, pasid); - addr = ((u64)(dw5 & 0xf) << 44) | ((u64)dw4 << 12); - key = AMDGPU_VM_FAULT(pasid, addr); - if (!vm) { - /* VM not found, process it normally */ - spin_unlock(&adev->vm_manager.pasid_lock); - return true; - } else { - r = amdgpu_vm_add_fault(vm->fault_hash, key); - - /* Hash table is full or the fault is already being processed, - * ignore further page faults - */ - if (r != 0) { - spin_unlock(&adev->vm_manager.pasid_lock); - goto ignore_iv; - } - } - /* No locking required with single writer and single reader */ - r = kfifo_put(&vm->faults, key); - if (!r) { - /* FIFO is full. Ignore it until there is space */ - amdgpu_vm_clear_fault(vm->fault_hash, key); - spin_unlock(&adev->vm_manager.pasid_lock); - goto ignore_iv; - } + /* When a ring buffer overflow happen start parsing interrupt + * from the last not overwritten vector (wptr + 32). Hopefully + * this should allow us to catchup. + */ + 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; + + if (ih == &adev->irq.ih) + reg = SOC15_REG_OFFSET(OSSSYS, 0, mmIH_RB_CNTL); + else if (ih == &adev->irq.ih1) + reg = SOC15_REG_OFFSET(OSSSYS, 0, mmIH_RB_CNTL_RING1); + else if (ih == &adev->irq.ih2) + reg = SOC15_REG_OFFSET(OSSSYS, 0, mmIH_RB_CNTL_RING2); + else + BUG(); - spin_unlock(&adev->vm_manager.pasid_lock); - /* It's the first fault for this address, process it normally */ - return true; + tmp = RREG32_NO_KIQ(reg); + tmp = REG_SET_FIELD(tmp, IH_RB_CNTL, WPTR_OVERFLOW_CLEAR, 1); + WREG32_NO_KIQ(reg, tmp); -ignore_iv: - adev->irq.ih.rptr += 32; - return false; +out: + return (wptr & ih->ptr_mask); } /** @@ -312,20 +330,21 @@ ignore_iv: * position and also advance the position. */ static void vega10_ih_decode_iv(struct amdgpu_device *adev, - struct amdgpu_iv_entry *entry) + struct amdgpu_ih_ring *ih, + struct amdgpu_iv_entry *entry) { /* wptr/rptr are in bytes! */ - u32 ring_index = adev->irq.ih.rptr >> 2; + u32 ring_index = ih->rptr >> 2; uint32_t dw[8]; - dw[0] = le32_to_cpu(adev->irq.ih.ring[ring_index + 0]); - dw[1] = le32_to_cpu(adev->irq.ih.ring[ring_index + 1]); - dw[2] = le32_to_cpu(adev->irq.ih.ring[ring_index + 2]); - dw[3] = le32_to_cpu(adev->irq.ih.ring[ring_index + 3]); - dw[4] = le32_to_cpu(adev->irq.ih.ring[ring_index + 4]); - dw[5] = le32_to_cpu(adev->irq.ih.ring[ring_index + 5]); - dw[6] = le32_to_cpu(adev->irq.ih.ring[ring_index + 6]); - dw[7] = le32_to_cpu(adev->irq.ih.ring[ring_index + 7]); + dw[0] = le32_to_cpu(ih->ring[ring_index + 0]); + dw[1] = le32_to_cpu(ih->ring[ring_index + 1]); + dw[2] = le32_to_cpu(ih->ring[ring_index + 2]); + dw[3] = le32_to_cpu(ih->ring[ring_index + 3]); + dw[4] = le32_to_cpu(ih->ring[ring_index + 4]); + dw[5] = le32_to_cpu(ih->ring[ring_index + 5]); + dw[6] = le32_to_cpu(ih->ring[ring_index + 6]); + dw[7] = le32_to_cpu(ih->ring[ring_index + 7]); entry->client_id = dw[0] & 0xff; entry->src_id = (dw[0] >> 8) & 0xff; @@ -341,9 +360,8 @@ static void vega10_ih_decode_iv(struct amdgpu_device *adev, entry->src_data[2] = dw[6]; entry->src_data[3] = dw[7]; - /* wptr/rptr are in bytes! */ - adev->irq.ih.rptr += 32; + ih->rptr += 32; } /** @@ -353,18 +371,59 @@ static void vega10_ih_decode_iv(struct amdgpu_device *adev, * * Set the IH ring buffer rptr. */ -static void vega10_ih_set_rptr(struct amdgpu_device *adev) +static void vega10_ih_set_rptr(struct amdgpu_device *adev, + struct amdgpu_ih_ring *ih) { - if (adev->irq.ih.use_doorbell) { + if (ih->use_doorbell) { /* XXX check if swapping is necessary on BE */ - if (adev->irq.ih.use_bus_addr) - adev->irq.ih.ring[adev->irq.ih.rptr_offs] = adev->irq.ih.rptr; - else - adev->wb.wb[adev->irq.ih.rptr_offs] = adev->irq.ih.rptr; - WDOORBELL32(adev->irq.ih.doorbell_index, adev->irq.ih.rptr); - } else { - WREG32_SOC15(OSSSYS, 0, mmIH_RB_RPTR, adev->irq.ih.rptr); + *ih->rptr_cpu = ih->rptr; + WDOORBELL32(ih->doorbell_index, ih->rptr); + } else if (ih == &adev->irq.ih) { + WREG32_SOC15(OSSSYS, 0, mmIH_RB_RPTR, ih->rptr); + } else if (ih == &adev->irq.ih1) { + WREG32_SOC15(OSSSYS, 0, mmIH_RB_RPTR_RING1, ih->rptr); + } else if (ih == &adev->irq.ih2) { + WREG32_SOC15(OSSSYS, 0, mmIH_RB_RPTR_RING2, ih->rptr); + } +} + +/** + * vega10_ih_self_irq - dispatch work for ring 1 and 2 + * + * @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 vega10_ih_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; + case 2: + *adev->irq.ih2.wptr_cpu = wptr; + schedule_work(&adev->irq.ih2_work); + break; + default: break; } + return 0; +} + +static const struct amdgpu_irq_src_funcs vega10_ih_self_irq_funcs = { + .process = vega10_ih_self_irq, +}; + +static void vega10_ih_set_self_irq_funcs(struct amdgpu_device *adev) +{ + adev->irq.self_irq.num_types = 0; + adev->irq.self_irq.funcs = &vega10_ih_self_irq_funcs; } static int vega10_ih_early_init(void *handle) @@ -372,20 +431,37 @@ static int vega10_ih_early_init(void *handle) struct amdgpu_device *adev = (struct amdgpu_device *)handle; vega10_ih_set_interrupt_funcs(adev); + vega10_ih_set_self_irq_funcs(adev); return 0; } static int vega10_ih_sw_init(void *handle) { - int r; struct amdgpu_device *adev = (struct amdgpu_device *)handle; + int r; + + r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_IH, 0, + &adev->irq.self_irq); + if (r) + return r; r = amdgpu_ih_ring_init(adev, &adev->irq.ih, 256 * 1024, true); if (r) return r; + if (adev->asic_type == CHIP_VEGA10) { + r = amdgpu_ih_ring_init(adev, &adev->irq.ih1, PAGE_SIZE, true); + if (r) + return r; + + r = amdgpu_ih_ring_init(adev, &adev->irq.ih2, PAGE_SIZE, true); + if (r) + return r; + } + + /* TODO add doorbell for IH1 & IH2 as well */ adev->irq.ih.use_doorbell = true; - adev->irq.ih.doorbell_index = AMDGPU_DOORBELL64_IH << 1; + adev->irq.ih.doorbell_index = adev->doorbell_index.ih << 1; r = amdgpu_irq_init(adev); @@ -397,6 +473,8 @@ static int vega10_ih_sw_fini(void *handle) struct amdgpu_device *adev = (struct amdgpu_device *)handle; amdgpu_irq_fini(adev); + amdgpu_ih_ring_fini(adev, &adev->irq.ih2); + amdgpu_ih_ring_fini(adev, &adev->irq.ih1); amdgpu_ih_ring_fini(adev, &adev->irq.ih); return 0; @@ -487,7 +565,6 @@ const struct amd_ip_funcs vega10_ih_ip_funcs = { static const struct amdgpu_ih_funcs vega10_ih_funcs = { .get_wptr = vega10_ih_get_wptr, - .prescreen_iv = vega10_ih_prescreen_iv, .decode_iv = vega10_ih_decode_iv, .set_rptr = vega10_ih_set_rptr }; diff --git a/drivers/gpu/drm/amd/amdgpu/vega10_reg_init.c b/drivers/gpu/drm/amd/amdgpu/vega10_reg_init.c index c5c9b2bc190d..a8e92638a2e8 100644 --- a/drivers/gpu/drm/amd/amdgpu/vega10_reg_init.c +++ b/drivers/gpu/drm/amd/amdgpu/vega10_reg_init.c @@ -56,4 +56,37 @@ int vega10_reg_base_init(struct amdgpu_device *adev) return 0; } +void vega10_doorbell_index_init(struct amdgpu_device *adev) +{ + adev->doorbell_index.kiq = AMDGPU_DOORBELL64_KIQ; + adev->doorbell_index.mec_ring0 = AMDGPU_DOORBELL64_MEC_RING0; + adev->doorbell_index.mec_ring1 = AMDGPU_DOORBELL64_MEC_RING1; + adev->doorbell_index.mec_ring2 = AMDGPU_DOORBELL64_MEC_RING2; + adev->doorbell_index.mec_ring3 = AMDGPU_DOORBELL64_MEC_RING3; + adev->doorbell_index.mec_ring4 = AMDGPU_DOORBELL64_MEC_RING4; + adev->doorbell_index.mec_ring5 = AMDGPU_DOORBELL64_MEC_RING5; + adev->doorbell_index.mec_ring6 = AMDGPU_DOORBELL64_MEC_RING6; + adev->doorbell_index.mec_ring7 = AMDGPU_DOORBELL64_MEC_RING7; + adev->doorbell_index.userqueue_start = AMDGPU_DOORBELL64_USERQUEUE_START; + adev->doorbell_index.userqueue_end = AMDGPU_DOORBELL64_USERQUEUE_END; + adev->doorbell_index.gfx_ring0 = AMDGPU_DOORBELL64_GFX_RING0; + adev->doorbell_index.sdma_engine[0] = AMDGPU_DOORBELL64_sDMA_ENGINE0; + adev->doorbell_index.sdma_engine[1] = AMDGPU_DOORBELL64_sDMA_ENGINE1; + adev->doorbell_index.ih = AMDGPU_DOORBELL64_IH; + adev->doorbell_index.uvd_vce.uvd_ring0_1 = AMDGPU_DOORBELL64_UVD_RING0_1; + adev->doorbell_index.uvd_vce.uvd_ring2_3 = AMDGPU_DOORBELL64_UVD_RING2_3; + adev->doorbell_index.uvd_vce.uvd_ring4_5 = AMDGPU_DOORBELL64_UVD_RING4_5; + adev->doorbell_index.uvd_vce.uvd_ring6_7 = AMDGPU_DOORBELL64_UVD_RING6_7; + adev->doorbell_index.uvd_vce.vce_ring0_1 = AMDGPU_DOORBELL64_VCE_RING0_1; + adev->doorbell_index.uvd_vce.vce_ring2_3 = AMDGPU_DOORBELL64_VCE_RING2_3; + adev->doorbell_index.uvd_vce.vce_ring4_5 = AMDGPU_DOORBELL64_VCE_RING4_5; + adev->doorbell_index.uvd_vce.vce_ring6_7 = AMDGPU_DOORBELL64_VCE_RING6_7; + + adev->doorbell_index.first_non_cp = AMDGPU_DOORBELL64_FIRST_NON_CP; + adev->doorbell_index.last_non_cp = AMDGPU_DOORBELL64_LAST_NON_CP; + + /* In unit of dword doorbell */ + adev->doorbell_index.max_assignment = AMDGPU_DOORBELL64_MAX_ASSIGNMENT << 1; + adev->doorbell_index.sdma_doorbell_range = 4; +} diff --git a/drivers/gpu/drm/amd/amdgpu/vega20_reg_init.c b/drivers/gpu/drm/amd/amdgpu/vega20_reg_init.c index d13fc4fcb517..0db84386252a 100644 --- a/drivers/gpu/drm/amd/amdgpu/vega20_reg_init.c +++ b/drivers/gpu/drm/amd/amdgpu/vega20_reg_init.c @@ -54,4 +54,42 @@ int vega20_reg_base_init(struct amdgpu_device *adev) return 0; } +void vega20_doorbell_index_init(struct amdgpu_device *adev) +{ + adev->doorbell_index.kiq = AMDGPU_VEGA20_DOORBELL_KIQ; + adev->doorbell_index.mec_ring0 = AMDGPU_VEGA20_DOORBELL_MEC_RING0; + adev->doorbell_index.mec_ring1 = AMDGPU_VEGA20_DOORBELL_MEC_RING1; + adev->doorbell_index.mec_ring2 = AMDGPU_VEGA20_DOORBELL_MEC_RING2; + adev->doorbell_index.mec_ring3 = AMDGPU_VEGA20_DOORBELL_MEC_RING3; + adev->doorbell_index.mec_ring4 = AMDGPU_VEGA20_DOORBELL_MEC_RING4; + adev->doorbell_index.mec_ring5 = AMDGPU_VEGA20_DOORBELL_MEC_RING5; + adev->doorbell_index.mec_ring6 = AMDGPU_VEGA20_DOORBELL_MEC_RING6; + adev->doorbell_index.mec_ring7 = AMDGPU_VEGA20_DOORBELL_MEC_RING7; + adev->doorbell_index.userqueue_start = AMDGPU_VEGA20_DOORBELL_USERQUEUE_START; + adev->doorbell_index.userqueue_end = AMDGPU_VEGA20_DOORBELL_USERQUEUE_END; + adev->doorbell_index.gfx_ring0 = AMDGPU_VEGA20_DOORBELL_GFX_RING0; + adev->doorbell_index.sdma_engine[0] = AMDGPU_VEGA20_DOORBELL_sDMA_ENGINE0; + adev->doorbell_index.sdma_engine[1] = AMDGPU_VEGA20_DOORBELL_sDMA_ENGINE1; + adev->doorbell_index.sdma_engine[2] = AMDGPU_VEGA20_DOORBELL_sDMA_ENGINE2; + adev->doorbell_index.sdma_engine[3] = AMDGPU_VEGA20_DOORBELL_sDMA_ENGINE3; + adev->doorbell_index.sdma_engine[4] = AMDGPU_VEGA20_DOORBELL_sDMA_ENGINE4; + adev->doorbell_index.sdma_engine[5] = AMDGPU_VEGA20_DOORBELL_sDMA_ENGINE5; + adev->doorbell_index.sdma_engine[6] = AMDGPU_VEGA20_DOORBELL_sDMA_ENGINE6; + adev->doorbell_index.sdma_engine[7] = AMDGPU_VEGA20_DOORBELL_sDMA_ENGINE7; + adev->doorbell_index.ih = AMDGPU_VEGA20_DOORBELL_IH; + adev->doorbell_index.uvd_vce.uvd_ring0_1 = AMDGPU_VEGA20_DOORBELL64_UVD_RING0_1; + adev->doorbell_index.uvd_vce.uvd_ring2_3 = AMDGPU_VEGA20_DOORBELL64_UVD_RING2_3; + adev->doorbell_index.uvd_vce.uvd_ring4_5 = AMDGPU_VEGA20_DOORBELL64_UVD_RING4_5; + adev->doorbell_index.uvd_vce.uvd_ring6_7 = AMDGPU_VEGA20_DOORBELL64_UVD_RING6_7; + adev->doorbell_index.uvd_vce.vce_ring0_1 = AMDGPU_VEGA20_DOORBELL64_VCE_RING0_1; + adev->doorbell_index.uvd_vce.vce_ring2_3 = AMDGPU_VEGA20_DOORBELL64_VCE_RING2_3; + adev->doorbell_index.uvd_vce.vce_ring4_5 = AMDGPU_VEGA20_DOORBELL64_VCE_RING4_5; + adev->doorbell_index.uvd_vce.vce_ring6_7 = AMDGPU_VEGA20_DOORBELL64_VCE_RING6_7; + + adev->doorbell_index.first_non_cp = AMDGPU_VEGA20_DOORBELL64_FIRST_NON_CP; + adev->doorbell_index.last_non_cp = AMDGPU_VEGA20_DOORBELL64_LAST_NON_CP; + + adev->doorbell_index.max_assignment = AMDGPU_VEGA20_DOORBELL_MAX_ASSIGNMENT << 1; + adev->doorbell_index.sdma_doorbell_range = 20; +} diff --git a/drivers/gpu/drm/amd/amdgpu/vi.c b/drivers/gpu/drm/amd/amdgpu/vi.c index 07880d35e9de..5e5b42a0744a 100644 --- a/drivers/gpu/drm/amd/amdgpu/vi.c +++ b/drivers/gpu/drm/amd/amdgpu/vi.c @@ -87,9 +87,9 @@ static u32 vi_pcie_rreg(struct amdgpu_device *adev, u32 reg) u32 r; spin_lock_irqsave(&adev->pcie_idx_lock, flags); - WREG32(mmPCIE_INDEX, reg); - (void)RREG32(mmPCIE_INDEX); - r = RREG32(mmPCIE_DATA); + WREG32_NO_KIQ(mmPCIE_INDEX, reg); + (void)RREG32_NO_KIQ(mmPCIE_INDEX); + r = RREG32_NO_KIQ(mmPCIE_DATA); spin_unlock_irqrestore(&adev->pcie_idx_lock, flags); return r; } @@ -99,10 +99,10 @@ static void vi_pcie_wreg(struct amdgpu_device *adev, u32 reg, u32 v) unsigned long flags; spin_lock_irqsave(&adev->pcie_idx_lock, flags); - WREG32(mmPCIE_INDEX, reg); - (void)RREG32(mmPCIE_INDEX); - WREG32(mmPCIE_DATA, v); - (void)RREG32(mmPCIE_DATA); + WREG32_NO_KIQ(mmPCIE_INDEX, reg); + (void)RREG32_NO_KIQ(mmPCIE_INDEX); + WREG32_NO_KIQ(mmPCIE_DATA, v); + (void)RREG32_NO_KIQ(mmPCIE_DATA); spin_unlock_irqrestore(&adev->pcie_idx_lock, flags); } @@ -123,8 +123,8 @@ static void vi_smc_wreg(struct amdgpu_device *adev, u32 reg, u32 v) unsigned long flags; spin_lock_irqsave(&adev->smc_idx_lock, flags); - WREG32(mmSMC_IND_INDEX_11, (reg)); - WREG32(mmSMC_IND_DATA_11, (v)); + WREG32_NO_KIQ(mmSMC_IND_INDEX_11, (reg)); + WREG32_NO_KIQ(mmSMC_IND_DATA_11, (v)); spin_unlock_irqrestore(&adev->smc_idx_lock, flags); } @@ -941,6 +941,69 @@ static bool vi_need_full_reset(struct amdgpu_device *adev) } } +static void vi_get_pcie_usage(struct amdgpu_device *adev, uint64_t *count0, + uint64_t *count1) +{ + uint32_t perfctr = 0; + uint64_t cnt0_of, cnt1_of; + int tmp; + + /* This reports 0 on APUs, so return to avoid writing/reading registers + * that may or may not be different from their GPU counterparts + */ + if (adev->flags & AMD_IS_APU) + return; + + /* Set the 2 events that we wish to watch, defined above */ + /* Reg 40 is # received msgs, Reg 104 is # of posted requests sent */ + perfctr = REG_SET_FIELD(perfctr, PCIE_PERF_CNTL_TXCLK, EVENT0_SEL, 40); + perfctr = REG_SET_FIELD(perfctr, PCIE_PERF_CNTL_TXCLK, EVENT1_SEL, 104); + + /* Write to enable desired perf counters */ + WREG32_PCIE(ixPCIE_PERF_CNTL_TXCLK, perfctr); + /* Zero out and enable the perf counters + * Write 0x5: + * Bit 0 = Start all counters(1) + * Bit 2 = Global counter reset enable(1) + */ + WREG32_PCIE(ixPCIE_PERF_COUNT_CNTL, 0x00000005); + + msleep(1000); + + /* Load the shadow and disable the perf counters + * Write 0x2: + * Bit 0 = Stop counters(0) + * Bit 1 = Load the shadow counters(1) + */ + WREG32_PCIE(ixPCIE_PERF_COUNT_CNTL, 0x00000002); + + /* Read register values to get any >32bit overflow */ + tmp = RREG32_PCIE(ixPCIE_PERF_CNTL_TXCLK); + cnt0_of = REG_GET_FIELD(tmp, PCIE_PERF_CNTL_TXCLK, COUNTER0_UPPER); + cnt1_of = REG_GET_FIELD(tmp, PCIE_PERF_CNTL_TXCLK, COUNTER1_UPPER); + + /* Get the values and add the overflow */ + *count0 = RREG32_PCIE(ixPCIE_PERF_COUNT0_TXCLK) | (cnt0_of << 32); + *count1 = RREG32_PCIE(ixPCIE_PERF_COUNT1_TXCLK) | (cnt1_of << 32); +} + +static bool vi_need_reset_on_init(struct amdgpu_device *adev) +{ + u32 clock_cntl, pc; + + if (adev->flags & AMD_IS_APU) + return false; + + /* check if the SMC is already running */ + clock_cntl = RREG32_SMC(ixSMC_SYSCON_CLOCK_CNTL_0); + pc = RREG32_SMC(ixSMC_PC_C); + if ((0 == REG_GET_FIELD(clock_cntl, SMC_SYSCON_CLOCK_CNTL_0, ck_disable)) && + (0x20100 <= pc)) + return true; + + return false; +} + static const struct amdgpu_asic_funcs vi_asic_funcs = { .read_disabled_bios = &vi_read_disabled_bios, @@ -955,6 +1018,9 @@ static const struct amdgpu_asic_funcs vi_asic_funcs = .flush_hdp = &vi_flush_hdp, .invalidate_hdp = &vi_invalidate_hdp, .need_full_reset = &vi_need_full_reset, + .init_doorbell_index = &legacy_doorbell_index_init, + .get_pcie_usage = &vi_get_pcie_usage, + .need_reset_on_init = &vi_need_reset_on_init, }; #define CZ_REV_BRISTOL(rev) \ @@ -1712,3 +1778,21 @@ int vi_set_ip_blocks(struct amdgpu_device *adev) return 0; } + +void legacy_doorbell_index_init(struct amdgpu_device *adev) +{ + adev->doorbell_index.kiq = AMDGPU_DOORBELL_KIQ; + adev->doorbell_index.mec_ring0 = AMDGPU_DOORBELL_MEC_RING0; + adev->doorbell_index.mec_ring1 = AMDGPU_DOORBELL_MEC_RING1; + adev->doorbell_index.mec_ring2 = AMDGPU_DOORBELL_MEC_RING2; + adev->doorbell_index.mec_ring3 = AMDGPU_DOORBELL_MEC_RING3; + adev->doorbell_index.mec_ring4 = AMDGPU_DOORBELL_MEC_RING4; + adev->doorbell_index.mec_ring5 = AMDGPU_DOORBELL_MEC_RING5; + adev->doorbell_index.mec_ring6 = AMDGPU_DOORBELL_MEC_RING6; + adev->doorbell_index.mec_ring7 = AMDGPU_DOORBELL_MEC_RING7; + adev->doorbell_index.gfx_ring0 = AMDGPU_DOORBELL_GFX_RING0; + adev->doorbell_index.sdma_engine[0] = AMDGPU_DOORBELL_sDMA_ENGINE0; + adev->doorbell_index.sdma_engine[1] = AMDGPU_DOORBELL_sDMA_ENGINE1; + adev->doorbell_index.ih = AMDGPU_DOORBELL_IH; + adev->doorbell_index.max_assignment = AMDGPU_DOORBELL_MAX_ASSIGNMENT; +} diff --git a/drivers/gpu/drm/amd/amdgpu/vi.h b/drivers/gpu/drm/amd/amdgpu/vi.h index 0429fe332269..8de0772f986c 100644 --- a/drivers/gpu/drm/amd/amdgpu/vi.h +++ b/drivers/gpu/drm/amd/amdgpu/vi.h @@ -30,4 +30,5 @@ void vi_srbm_select(struct amdgpu_device *adev, u32 me, u32 pipe, u32 queue, u32 vmid); int vi_set_ip_blocks(struct amdgpu_device *adev); +void legacy_doorbell_index_init(struct amdgpu_device *adev); #endif diff --git a/drivers/gpu/drm/amd/amdkfd/Kconfig b/drivers/gpu/drm/amd/amdkfd/Kconfig index fbf0ee5201c3..c3613604a4f8 100644 --- a/drivers/gpu/drm/amd/amdkfd/Kconfig +++ b/drivers/gpu/drm/amd/amdkfd/Kconfig @@ -4,8 +4,8 @@ config HSA_AMD bool "HSA kernel driver for AMD GPU devices" - depends on DRM_AMDGPU && X86_64 - imply AMD_IOMMU_V2 + depends on DRM_AMDGPU && (X86_64 || ARM64) + imply AMD_IOMMU_V2 if X86_64 select MMU_NOTIFIER help Enable this if you want to use HSA features on AMD GPU devices. diff --git a/drivers/gpu/drm/amd/amdkfd/cik_event_interrupt.c b/drivers/gpu/drm/amd/amdkfd/cik_event_interrupt.c index 5d2475d5392c..177d1e5329a5 100644 --- a/drivers/gpu/drm/amd/amdkfd/cik_event_interrupt.c +++ b/drivers/gpu/drm/amd/amdkfd/cik_event_interrupt.c @@ -23,6 +23,7 @@ #include "kfd_priv.h" #include "kfd_events.h" #include "cik_int.h" +#include "amdgpu_amdkfd.h" static bool cik_event_interrupt_isr(struct kfd_dev *dev, const uint32_t *ih_ring_entry, @@ -107,7 +108,7 @@ static void cik_event_interrupt_wq(struct kfd_dev *dev, kfd_process_vm_fault(dev->dqm, pasid); memset(&info, 0, sizeof(info)); - dev->kfd2kgd->get_vm_fault_info(dev->kgd, &info); + amdgpu_amdkfd_gpuvm_get_vm_fault_info(dev->kgd, &info); if (!info.page_addr && !info.status) return; diff --git a/drivers/gpu/drm/amd/amdkfd/cik_regs.h b/drivers/gpu/drm/amd/amdkfd/cik_regs.h index 37ce6dd65391..8e2a1663c4db 100644 --- a/drivers/gpu/drm/amd/amdkfd/cik_regs.h +++ b/drivers/gpu/drm/amd/amdkfd/cik_regs.h @@ -68,6 +68,4 @@ #define GRBM_GFX_INDEX 0x30800 -#define ATC_VMID_PASID_MAPPING_VALID (1U << 31) - #endif diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c b/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c index 14d5b5fa822d..083bd8114db1 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c @@ -33,10 +33,12 @@ #include <linux/time.h> #include <linux/mm.h> #include <linux/mman.h> +#include <linux/dma-buf.h> #include <asm/processor.h> #include "kfd_priv.h" #include "kfd_device_queue_manager.h" #include "kfd_dbgmgr.h" +#include "amdgpu_amdkfd.h" static long kfd_ioctl(struct file *, unsigned int, unsigned long); static int kfd_open(struct inode *, struct file *); @@ -156,8 +158,7 @@ static int set_queue_properties_from_user(struct queue_properties *q_properties, } if ((args->ring_base_address) && - (!access_ok(VERIFY_WRITE, - (const void __user *) args->ring_base_address, + (!access_ok((const void __user *) args->ring_base_address, sizeof(uint64_t)))) { pr_err("Can't access ring base address\n"); return -EFAULT; @@ -168,31 +169,27 @@ static int set_queue_properties_from_user(struct queue_properties *q_properties, return -EINVAL; } - if (!access_ok(VERIFY_WRITE, - (const void __user *) args->read_pointer_address, + if (!access_ok((const void __user *) args->read_pointer_address, sizeof(uint32_t))) { pr_err("Can't access read pointer\n"); return -EFAULT; } - if (!access_ok(VERIFY_WRITE, - (const void __user *) args->write_pointer_address, + if (!access_ok((const void __user *) args->write_pointer_address, sizeof(uint32_t))) { pr_err("Can't access write pointer\n"); return -EFAULT; } if (args->eop_buffer_address && - !access_ok(VERIFY_WRITE, - (const void __user *) args->eop_buffer_address, + !access_ok((const void __user *) args->eop_buffer_address, sizeof(uint32_t))) { pr_debug("Can't access eop buffer"); return -EFAULT; } if (args->ctx_save_restore_address && - !access_ok(VERIFY_WRITE, - (const void __user *) args->ctx_save_restore_address, + !access_ok((const void __user *) args->ctx_save_restore_address, sizeof(uint32_t))) { pr_debug("Can't access ctx save restore buffer"); return -EFAULT; @@ -363,8 +360,7 @@ static int kfd_ioctl_update_queue(struct file *filp, struct kfd_process *p, } if ((args->ring_base_address) && - (!access_ok(VERIFY_WRITE, - (const void __user *) args->ring_base_address, + (!access_ok((const void __user *) args->ring_base_address, sizeof(uint64_t)))) { pr_err("Can't access ring base address\n"); return -EFAULT; @@ -834,8 +830,7 @@ static int kfd_ioctl_get_clock_counters(struct file *filep, dev = kfd_device_by_id(args->gpu_id); if (dev) /* Reading GPU clock counter from KGD */ - args->gpu_clock_counter = - dev->kfd2kgd->get_gpu_clock_counter(dev->kgd); + args->gpu_clock_counter = amdgpu_amdkfd_get_gpu_clock_counter(dev->kgd); else /* Node without GPU resource */ args->gpu_clock_counter = 0; @@ -1042,7 +1037,7 @@ static int kfd_ioctl_create_event(struct file *filp, struct kfd_process *p, } mutex_unlock(&p->mutex); - err = kfd->kfd2kgd->map_gtt_bo_to_kernel(kfd->kgd, + err = amdgpu_amdkfd_gpuvm_map_gtt_bo_to_kernel(kfd->kgd, mem, &kern_addr, &size); if (err) { pr_err("Failed to map event page to kernel\n"); @@ -1240,7 +1235,7 @@ bool kfd_dev_is_large_bar(struct kfd_dev *dev) if (dev->device_info->needs_iommu_device) return false; - dev->kfd2kgd->get_local_mem_info(dev->kgd, &mem_info); + amdgpu_amdkfd_get_local_mem_info(dev->kgd, &mem_info); if (mem_info.local_mem_size_private == 0 && mem_info.local_mem_size_public > 0) return true; @@ -1273,6 +1268,12 @@ static int kfd_ioctl_alloc_memory_of_gpu(struct file *filep, return -EINVAL; } + if (flags & KFD_IOC_ALLOC_MEM_FLAGS_DOORBELL) { + if (args->size != kfd_doorbell_process_slice(dev)) + return -EINVAL; + offset = kfd_get_process_doorbells(dev, p); + } + mutex_lock(&p->mutex); pdd = kfd_bind_process_to_device(dev, p); @@ -1281,7 +1282,7 @@ static int kfd_ioctl_alloc_memory_of_gpu(struct file *filep, goto err_unlock; } - err = dev->kfd2kgd->alloc_memory_of_gpu( + err = amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu( dev->kgd, args->va_addr, args->size, pdd->vm, (struct kgd_mem **) &mem, &offset, flags); @@ -1303,7 +1304,7 @@ static int kfd_ioctl_alloc_memory_of_gpu(struct file *filep, return 0; err_free: - dev->kfd2kgd->free_memory_of_gpu(dev->kgd, (struct kgd_mem *)mem); + amdgpu_amdkfd_gpuvm_free_memory_of_gpu(dev->kgd, (struct kgd_mem *)mem); err_unlock: mutex_unlock(&p->mutex); return err; @@ -1338,7 +1339,8 @@ static int kfd_ioctl_free_memory_of_gpu(struct file *filep, goto err_unlock; } - ret = dev->kfd2kgd->free_memory_of_gpu(dev->kgd, (struct kgd_mem *)mem); + ret = amdgpu_amdkfd_gpuvm_free_memory_of_gpu(dev->kgd, + (struct kgd_mem *)mem); /* If freeing the buffer failed, leave the handle in place for * clean-up during process tear-down. @@ -1418,7 +1420,7 @@ static int kfd_ioctl_map_memory_to_gpu(struct file *filep, err = PTR_ERR(peer_pdd); goto get_mem_obj_from_handle_failed; } - err = peer->kfd2kgd->map_memory_to_gpu( + err = amdgpu_amdkfd_gpuvm_map_memory_to_gpu( peer->kgd, (struct kgd_mem *)mem, peer_pdd->vm); if (err) { pr_err("Failed to map to gpu %d/%d\n", @@ -1430,7 +1432,7 @@ static int kfd_ioctl_map_memory_to_gpu(struct file *filep, mutex_unlock(&p->mutex); - err = dev->kfd2kgd->sync_memory(dev->kgd, (struct kgd_mem *) mem, true); + err = amdgpu_amdkfd_gpuvm_sync_memory(dev->kgd, (struct kgd_mem *) mem, true); if (err) { pr_debug("Sync memory failed, wait interrupted by user signal\n"); goto sync_memory_failed; @@ -1525,7 +1527,7 @@ static int kfd_ioctl_unmap_memory_from_gpu(struct file *filep, err = -ENODEV; goto get_mem_obj_from_handle_failed; } - err = dev->kfd2kgd->unmap_memory_to_gpu( + err = amdgpu_amdkfd_gpuvm_unmap_memory_from_gpu( peer->kgd, (struct kgd_mem *)mem, peer_pdd->vm); if (err) { pr_err("Failed to unmap from gpu %d/%d\n", @@ -1549,6 +1551,115 @@ copy_from_user_failed: return err; } +static int kfd_ioctl_get_dmabuf_info(struct file *filep, + struct kfd_process *p, void *data) +{ + struct kfd_ioctl_get_dmabuf_info_args *args = data; + struct kfd_dev *dev = NULL; + struct kgd_dev *dma_buf_kgd; + void *metadata_buffer = NULL; + uint32_t flags; + unsigned int i; + int r; + + /* Find a KFD GPU device that supports the get_dmabuf_info query */ + for (i = 0; kfd_topology_enum_kfd_devices(i, &dev) == 0; i++) + if (dev) + break; + if (!dev) + return -EINVAL; + + if (args->metadata_ptr) { + metadata_buffer = kzalloc(args->metadata_size, GFP_KERNEL); + if (!metadata_buffer) + return -ENOMEM; + } + + /* Get dmabuf info from KGD */ + r = amdgpu_amdkfd_get_dmabuf_info(dev->kgd, args->dmabuf_fd, + &dma_buf_kgd, &args->size, + metadata_buffer, args->metadata_size, + &args->metadata_size, &flags); + if (r) + goto exit; + + /* Reverse-lookup gpu_id from kgd pointer */ + dev = kfd_device_by_kgd(dma_buf_kgd); + if (!dev) { + r = -EINVAL; + goto exit; + } + args->gpu_id = dev->id; + args->flags = flags; + + /* Copy metadata buffer to user mode */ + if (metadata_buffer) { + r = copy_to_user((void __user *)args->metadata_ptr, + metadata_buffer, args->metadata_size); + if (r != 0) + r = -EFAULT; + } + +exit: + kfree(metadata_buffer); + + return r; +} + +static int kfd_ioctl_import_dmabuf(struct file *filep, + struct kfd_process *p, void *data) +{ + struct kfd_ioctl_import_dmabuf_args *args = data; + struct kfd_process_device *pdd; + struct dma_buf *dmabuf; + struct kfd_dev *dev; + int idr_handle; + uint64_t size; + void *mem; + int r; + + dev = kfd_device_by_id(args->gpu_id); + if (!dev) + return -EINVAL; + + dmabuf = dma_buf_get(args->dmabuf_fd); + if (IS_ERR(dmabuf)) + return PTR_ERR(dmabuf); + + mutex_lock(&p->mutex); + + pdd = kfd_bind_process_to_device(dev, p); + if (IS_ERR(pdd)) { + r = PTR_ERR(pdd); + goto err_unlock; + } + + r = amdgpu_amdkfd_gpuvm_import_dmabuf(dev->kgd, dmabuf, + args->va_addr, pdd->vm, + (struct kgd_mem **)&mem, &size, + NULL); + if (r) + goto err_unlock; + + idr_handle = kfd_process_device_create_obj_handle(pdd, mem); + if (idr_handle < 0) { + r = -EFAULT; + goto err_free; + } + + mutex_unlock(&p->mutex); + + args->handle = MAKE_HANDLE(args->gpu_id, idr_handle); + + return 0; + +err_free: + amdgpu_amdkfd_gpuvm_free_memory_of_gpu(dev->kgd, (struct kgd_mem *)mem); +err_unlock: + mutex_unlock(&p->mutex); + return r; +} + #define AMDKFD_IOCTL_DEF(ioctl, _func, _flags) \ [_IOC_NR(ioctl)] = {.cmd = ioctl, .func = _func, .flags = _flags, \ .cmd_drv = 0, .name = #ioctl} @@ -1634,7 +1745,13 @@ static const struct amdkfd_ioctl_desc amdkfd_ioctls[] = { kfd_ioctl_set_cu_mask, 0), AMDKFD_IOCTL_DEF(AMDKFD_IOC_GET_QUEUE_WAVE_STATE, - kfd_ioctl_get_queue_wave_state, 0) + kfd_ioctl_get_queue_wave_state, 0), + + AMDKFD_IOCTL_DEF(AMDKFD_IOC_GET_DMABUF_INFO, + kfd_ioctl_get_dmabuf_info, 0), + + AMDKFD_IOCTL_DEF(AMDKFD_IOC_IMPORT_DMABUF, + kfd_ioctl_import_dmabuf, 0), }; diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_crat.c b/drivers/gpu/drm/amd/amdkfd/kfd_crat.c index 56412b0e7e1c..2e7c44955f43 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_crat.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_crat.c @@ -26,6 +26,7 @@ #include "kfd_priv.h" #include "kfd_topology.h" #include "kfd_iommu.h" +#include "amdgpu_amdkfd.h" /* GPU Processor ID base for dGPUs for which VCRAT needs to be created. * GPU processor ID are expressed with Bit[31]=1. @@ -132,6 +133,7 @@ static struct kfd_gpu_cache_info carrizo_cache_info[] = { #define fiji_cache_info carrizo_cache_info #define polaris10_cache_info carrizo_cache_info #define polaris11_cache_info carrizo_cache_info +#define polaris12_cache_info carrizo_cache_info /* TODO - check & update Vega10 cache details */ #define vega10_cache_info carrizo_cache_info #define raven_cache_info carrizo_cache_info @@ -646,7 +648,12 @@ static int kfd_fill_gpu_cache_info(struct kfd_dev *kdev, pcache_info = polaris11_cache_info; num_of_cache_types = ARRAY_SIZE(polaris11_cache_info); break; + case CHIP_POLARIS12: + pcache_info = polaris12_cache_info; + num_of_cache_types = ARRAY_SIZE(polaris12_cache_info); + break; case CHIP_VEGA10: + case CHIP_VEGA12: case CHIP_VEGA20: pcache_info = vega10_cache_info; num_of_cache_types = ARRAY_SIZE(vega10_cache_info); @@ -753,12 +760,10 @@ int kfd_create_crat_image_acpi(void **crat_image, size_t *size) return -ENODATA; } - pcrat_image = kmalloc(crat_table->length, GFP_KERNEL); + pcrat_image = kmemdup(crat_table, crat_table->length, GFP_KERNEL); if (!pcrat_image) return -ENOMEM; - memcpy(pcrat_image, crat_table, crat_table->length); - *crat_image = pcrat_image; *size = crat_table->length; @@ -848,7 +853,7 @@ static int kfd_fill_mem_info_for_cpu(int numa_node_id, int *avail_size, */ pgdat = NODE_DATA(numa_node_id); for (zone_type = 0; zone_type < MAX_NR_ZONES; zone_type++) - mem_in_bytes += pgdat->node_zones[zone_type].managed_pages; + mem_in_bytes += zone_managed_pages(&pgdat->node_zones[zone_type]); mem_in_bytes <<= PAGE_SHIFT; sub_type_hdr->length_low = lower_32_bits(mem_in_bytes); @@ -858,6 +863,7 @@ static int kfd_fill_mem_info_for_cpu(int numa_node_id, int *avail_size, return 0; } +#ifdef CONFIG_X86_64 static int kfd_fill_iolink_info_for_cpu(int numa_node_id, int *avail_size, uint32_t *num_entries, struct crat_subtype_iolink *sub_type_hdr) @@ -900,6 +906,7 @@ static int kfd_fill_iolink_info_for_cpu(int numa_node_id, int *avail_size, return 0; } +#endif /* kfd_create_vcrat_image_cpu - Create Virtual CRAT for CPU * @@ -915,7 +922,9 @@ static int kfd_create_vcrat_image_cpu(void *pcrat_image, size_t *size) struct crat_subtype_generic *sub_type_hdr; int avail_size = *size; int numa_node_id; +#ifdef CONFIG_X86_64 uint32_t entries = 0; +#endif int ret = 0; if (!pcrat_image || avail_size < VCRAT_SIZE_FOR_CPU) @@ -977,6 +986,7 @@ static int kfd_create_vcrat_image_cpu(void *pcrat_image, size_t *size) sub_type_hdr->length); /* Fill in Subtype: IO Link */ +#ifdef CONFIG_X86_64 ret = kfd_fill_iolink_info_for_cpu(numa_node_id, &avail_size, &entries, (struct crat_subtype_iolink *)sub_type_hdr); @@ -987,6 +997,9 @@ static int kfd_create_vcrat_image_cpu(void *pcrat_image, size_t *size) sub_type_hdr = (typeof(sub_type_hdr))((char *)sub_type_hdr + sub_type_hdr->length * entries); +#else + pr_info("IO link not available for non x86 platforms\n"); +#endif crat_table->num_domains++; } @@ -1161,7 +1174,7 @@ static int kfd_create_vcrat_image_gpu(void *pcrat_image, cu->flags |= CRAT_CU_FLAGS_GPU_PRESENT; cu->proximity_domain = proximity_domain; - kdev->kfd2kgd->get_cu_info(kdev->kgd, &cu_info); + amdgpu_amdkfd_get_cu_info(kdev->kgd, &cu_info); cu->num_simd_per_cu = cu_info.simd_per_cu; cu->num_simd_cores = cu_info.simd_per_cu * cu_info.cu_active_number; cu->max_waves_simd = cu_info.max_waves_per_simd; @@ -1192,7 +1205,7 @@ static int kfd_create_vcrat_image_gpu(void *pcrat_image, * report the total FB size (public+private) as a single * private heap. */ - kdev->kfd2kgd->get_local_mem_info(kdev->kgd, &local_mem_info); + amdgpu_amdkfd_get_local_mem_info(kdev->kgd, &local_mem_info); sub_type_hdr = (typeof(sub_type_hdr))((char *)sub_type_hdr + sub_type_hdr->length); diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device.c b/drivers/gpu/drm/amd/amdkfd/kfd_device.c index e4ded890b1cb..cf9a49f49d3a 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_device.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_device.c @@ -28,6 +28,7 @@ #include "kfd_pm4_headers_vi.h" #include "cwsr_trap_handler.h" #include "kfd_iommu.h" +#include "amdgpu_amdkfd.h" #define MQD_SIZE_ALIGNED 768 @@ -204,6 +205,22 @@ static const struct kfd_device_info polaris11_device_info = { .num_sdma_queues_per_engine = 2, }; +static const struct kfd_device_info polaris12_device_info = { + .asic_family = CHIP_POLARIS12, + .max_pasid_bits = 16, + .max_no_of_hqd = 24, + .doorbell_size = 4, + .ih_ring_entry_size = 4 * sizeof(uint32_t), + .event_interrupt_class = &event_interrupt_class_cik, + .num_of_watch_points = 4, + .mqd_size_aligned = MQD_SIZE_ALIGNED, + .supports_cwsr = true, + .needs_iommu_device = false, + .needs_pci_atomics = true, + .num_sdma_engines = 2, + .num_sdma_queues_per_engine = 2, +}; + static const struct kfd_device_info vega10_device_info = { .asic_family = CHIP_VEGA10, .max_pasid_bits = 16, @@ -236,6 +253,22 @@ static const struct kfd_device_info vega10_vf_device_info = { .num_sdma_queues_per_engine = 2, }; +static const struct kfd_device_info vega12_device_info = { + .asic_family = CHIP_VEGA12, + .max_pasid_bits = 16, + .max_no_of_hqd = 24, + .doorbell_size = 8, + .ih_ring_entry_size = 8 * sizeof(uint32_t), + .event_interrupt_class = &event_interrupt_class_v9, + .num_of_watch_points = 4, + .mqd_size_aligned = MQD_SIZE_ALIGNED, + .supports_cwsr = true, + .needs_iommu_device = false, + .needs_pci_atomics = false, + .num_sdma_engines = 2, + .num_sdma_queues_per_engine = 2, +}; + static const struct kfd_device_info vega20_device_info = { .asic_family = CHIP_VEGA20, .max_pasid_bits = 16, @@ -287,6 +320,7 @@ static const struct kfd_deviceid supported_devices[] = { { 0x9876, &carrizo_device_info }, /* Carrizo */ { 0x9877, &carrizo_device_info }, /* Carrizo */ { 0x15DD, &raven_device_info }, /* Raven */ + { 0x15D8, &raven_device_info }, /* Raven */ #endif { 0x67A0, &hawaii_device_info }, /* Hawaii */ { 0x67A1, &hawaii_device_info }, /* Hawaii */ @@ -330,6 +364,14 @@ static const struct kfd_deviceid supported_devices[] = { { 0x67EB, &polaris11_device_info }, /* Polaris11 */ { 0x67EF, &polaris11_device_info }, /* Polaris11 */ { 0x67FF, &polaris11_device_info }, /* Polaris11 */ + { 0x6980, &polaris12_device_info }, /* Polaris12 */ + { 0x6981, &polaris12_device_info }, /* Polaris12 */ + { 0x6985, &polaris12_device_info }, /* Polaris12 */ + { 0x6986, &polaris12_device_info }, /* Polaris12 */ + { 0x6987, &polaris12_device_info }, /* Polaris12 */ + { 0x6995, &polaris12_device_info }, /* Polaris12 */ + { 0x6997, &polaris12_device_info }, /* Polaris12 */ + { 0x699F, &polaris12_device_info }, /* Polaris12 */ { 0x6860, &vega10_device_info }, /* Vega10 */ { 0x6861, &vega10_device_info }, /* Vega10 */ { 0x6862, &vega10_device_info }, /* Vega10 */ @@ -345,6 +387,11 @@ static const struct kfd_deviceid supported_devices[] = { { 0x686E, &vega10_device_info }, /* Vega10 */ { 0x686F, &vega10_device_info }, /* Vega10 */ { 0x687F, &vega10_device_info }, /* Vega10 */ + { 0x69A0, &vega12_device_info }, /* Vega12 */ + { 0x69A1, &vega12_device_info }, /* Vega12 */ + { 0x69A2, &vega12_device_info }, /* Vega12 */ + { 0x69A3, &vega12_device_info }, /* Vega12 */ + { 0x69AF, &vega12_device_info }, /* Vega12 */ { 0x66a0, &vega20_device_info }, /* Vega20 */ { 0x66a1, &vega20_device_info }, /* Vega20 */ { 0x66a2, &vega20_device_info }, /* Vega20 */ @@ -485,7 +532,7 @@ bool kgd2kfd_device_init(struct kfd_dev *kfd, /* add another 512KB for all other allocations on gart (HPD, fences) */ size += 512 * 1024; - if (kfd->kfd2kgd->init_gtt_mem_allocation( + if (amdgpu_amdkfd_alloc_gtt_mem( kfd->kgd, size, &kfd->gtt_mem, &kfd->gtt_start_gpu_addr, &kfd->gtt_start_cpu_ptr, false)) { @@ -559,7 +606,7 @@ kfd_topology_add_device_error: kfd_doorbell_error: kfd_gtt_sa_fini(kfd); kfd_gtt_sa_init_error: - kfd->kfd2kgd->free_gtt_mem(kfd->kgd, kfd->gtt_mem); + amdgpu_amdkfd_free_gtt_mem(kfd->kgd, kfd->gtt_mem); dev_err(kfd_device, "device %x:%x NOT added due to errors\n", kfd->pdev->vendor, kfd->pdev->device); @@ -576,7 +623,7 @@ void kgd2kfd_device_exit(struct kfd_dev *kfd) kfd_topology_remove_device(kfd); kfd_doorbell_fini(kfd); kfd_gtt_sa_fini(kfd); - kfd->kfd2kgd->free_gtt_mem(kfd->kgd, kfd->gtt_mem); + amdgpu_amdkfd_free_gtt_mem(kfd->kgd, kfd->gtt_mem); } kfree(kfd); @@ -688,6 +735,7 @@ void kgd2kfd_interrupt(struct kfd_dev *kfd, const void *ih_ring_entry) { uint32_t patched_ihre[KFD_MAX_RING_ENTRY_SIZE]; bool is_patched = false; + unsigned long flags; if (!kfd->init_complete) return; @@ -697,7 +745,7 @@ void kgd2kfd_interrupt(struct kfd_dev *kfd, const void *ih_ring_entry) return; } - spin_lock(&kfd->interrupt_lock); + spin_lock_irqsave(&kfd->interrupt_lock, flags); if (kfd->interrupts_active && interrupt_is_wanted(kfd, ih_ring_entry, @@ -706,7 +754,7 @@ void kgd2kfd_interrupt(struct kfd_dev *kfd, const void *ih_ring_entry) is_patched ? patched_ihre : ih_ring_entry)) queue_work(kfd->ih_wq, &kfd->interrupt_work); - spin_unlock(&kfd->interrupt_lock); + spin_unlock_irqrestore(&kfd->interrupt_lock, flags); } int kgd2kfd_quiesce_mm(struct mm_struct *mm) diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c index a3b933967171..c6c9530e704e 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c @@ -33,6 +33,7 @@ #include "kfd_mqd_manager.h" #include "cik_regs.h" #include "kfd_kernel_queue.h" +#include "amdgpu_amdkfd.h" /* Size of the per-pipe EOP queue */ #define CIK_HPD_EOP_BYTES_LOG2 11 @@ -133,12 +134,18 @@ static int allocate_doorbell(struct qcm_process_device *qpd, struct queue *q) */ q->doorbell_id = q->properties.queue_id; } else if (q->properties.type == KFD_QUEUE_TYPE_SDMA) { - /* For SDMA queues on SOC15, use static doorbell - * assignments based on the engine and queue. + /* For SDMA queues on SOC15 with 8-byte doorbell, use static + * doorbell assignments based on the engine and queue id. + * The doobell index distance between RLC (2*i) and (2*i+1) + * for a SDMA engine is 512. */ - q->doorbell_id = dev->shared_resources.sdma_doorbell - [q->properties.sdma_engine_id] - [q->properties.sdma_queue_id]; + uint32_t *idx_offset = + dev->shared_resources.sdma_doorbell_idx; + + q->doorbell_id = idx_offset[q->properties.sdma_engine_id] + + (q->properties.sdma_queue_id & 1) + * KFD_QUEUE_DOORBELL_MIRROR_OFFSET + + (q->properties.sdma_queue_id >> 1); } else { /* For CP queues on SOC15 reserve a free doorbell ID */ unsigned int found; @@ -219,7 +226,7 @@ static int flush_texture_cache_nocpsch(struct kfd_dev *kdev, if (ret) return ret; - return kdev->kfd2kgd->submit_ib(kdev->kgd, KGD_ENGINE_MEC1, qpd->vmid, + return amdgpu_amdkfd_submit_ib(kdev->kgd, KGD_ENGINE_MEC1, qpd->vmid, qpd->ib_base, (uint32_t *)qpd->ib_kaddr, pmf->release_mem_size / sizeof(uint32_t)); } @@ -672,7 +679,7 @@ static int restore_process_queues_nocpsch(struct device_queue_manager *dqm, pdd = qpd_to_pdd(qpd); /* Retrieve PD base */ - pd_base = dqm->dev->kfd2kgd->get_process_page_dir(pdd->vm); + pd_base = amdgpu_amdkfd_gpuvm_get_process_page_dir(pdd->vm); dqm_lock(dqm); if (WARN_ON_ONCE(!qpd->evicted)) /* already restored, do nothing */ @@ -743,7 +750,7 @@ static int restore_process_queues_cpsch(struct device_queue_manager *dqm, pdd = qpd_to_pdd(qpd); /* Retrieve PD base */ - pd_base = dqm->dev->kfd2kgd->get_process_page_dir(pdd->vm); + pd_base = amdgpu_amdkfd_gpuvm_get_process_page_dir(pdd->vm); dqm_lock(dqm); if (WARN_ON_ONCE(!qpd->evicted)) /* already restored, do nothing */ @@ -793,7 +800,7 @@ static int register_process(struct device_queue_manager *dqm, pdd = qpd_to_pdd(qpd); /* Retrieve PD base */ - pd_base = dqm->dev->kfd2kgd->get_process_page_dir(pdd->vm); + pd_base = amdgpu_amdkfd_gpuvm_get_process_page_dir(pdd->vm); dqm_lock(dqm); list_add(&n->list, &dqm->queues); @@ -805,7 +812,7 @@ static int register_process(struct device_queue_manager *dqm, retval = dqm->asic_ops.update_qpd(dqm, qpd); if (dqm->processes_count++ == 0) - dqm->dev->kfd2kgd->set_compute_idle(dqm->dev->kgd, false); + amdgpu_amdkfd_set_compute_idle(dqm->dev->kgd, false); dqm_unlock(dqm); @@ -829,7 +836,7 @@ static int unregister_process(struct device_queue_manager *dqm, list_del(&cur->list); kfree(cur); if (--dqm->processes_count == 0) - dqm->dev->kfd2kgd->set_compute_idle( + amdgpu_amdkfd_set_compute_idle( dqm->dev->kgd, true); goto out; } @@ -845,15 +852,8 @@ static int set_pasid_vmid_mapping(struct device_queue_manager *dqm, unsigned int pasid, unsigned int vmid) { - uint32_t pasid_mapping; - - pasid_mapping = (pasid == 0) ? 0 : - (uint32_t)pasid | - ATC_VMID_PASID_MAPPING_VALID; - return dqm->dev->kfd2kgd->set_pasid_vmid_mapping( - dqm->dev->kgd, pasid_mapping, - vmid); + dqm->dev->kgd, pasid, vmid); } static void init_interrupts(struct device_queue_manager *dqm) @@ -1553,7 +1553,7 @@ static int get_wave_state(struct device_queue_manager *dqm, u32 *ctl_stack_used_size, u32 *save_area_used_size) { - struct mqd_manager *mqd; + struct mqd_manager *mqd_mgr; int r; dqm_lock(dqm); @@ -1564,19 +1564,19 @@ static int get_wave_state(struct device_queue_manager *dqm, goto dqm_unlock; } - mqd = dqm->ops.get_mqd_manager(dqm, KFD_MQD_TYPE_COMPUTE); - if (!mqd) { + mqd_mgr = dqm->ops.get_mqd_manager(dqm, KFD_MQD_TYPE_COMPUTE); + if (!mqd_mgr) { r = -ENOMEM; goto dqm_unlock; } - if (!mqd->get_wave_state) { + if (!mqd_mgr->get_wave_state) { r = -EINVAL; goto dqm_unlock; } - r = mqd->get_wave_state(mqd, q->mqd, ctl_stack, ctl_stack_used_size, - save_area_used_size); + r = mqd_mgr->get_wave_state(mqd_mgr, q->mqd, ctl_stack, + ctl_stack_used_size, save_area_used_size); dqm_unlock: dqm_unlock(dqm); @@ -1747,10 +1747,12 @@ struct device_queue_manager *device_queue_manager_init(struct kfd_dev *dev) case CHIP_FIJI: case CHIP_POLARIS10: case CHIP_POLARIS11: + case CHIP_POLARIS12: device_queue_manager_init_vi_tonga(&dqm->asic_ops); break; case CHIP_VEGA10: + case CHIP_VEGA12: case CHIP_VEGA20: case CHIP_RAVEN: device_queue_manager_init_v9(&dqm->asic_ops); @@ -1796,7 +1798,7 @@ static void kfd_process_hw_exception(struct work_struct *work) { struct device_queue_manager *dqm = container_of(work, struct device_queue_manager, hw_exception_work); - dqm->dev->kfd2kgd->gpu_recover(dqm->dev->kgd); + amdgpu_amdkfd_gpu_reset(dqm->dev->kgd); } #if defined(CONFIG_DEBUG_FS) diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager_vi.c b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager_vi.c index fd60a116be37..c3a5dcfe877a 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager_vi.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager_vi.c @@ -24,7 +24,6 @@ #include "kfd_device_queue_manager.h" #include "gca/gfx_8_0_enum.h" #include "gca/gfx_8_0_sh_mask.h" -#include "gca/gfx_8_0_enum.h" #include "oss/oss_3_0_sh_mask.h" static bool set_cache_memory_policy_vi(struct device_queue_manager *dqm, diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_flat_memory.c b/drivers/gpu/drm/amd/amdkfd/kfd_flat_memory.c index 3d66cec414af..213ea5454d11 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_flat_memory.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_flat_memory.c @@ -397,9 +397,11 @@ int kfd_init_apertures(struct kfd_process *process) case CHIP_FIJI: case CHIP_POLARIS10: case CHIP_POLARIS11: + case CHIP_POLARIS12: kfd_init_apertures_vi(pdd, id); break; case CHIP_VEGA10: + case CHIP_VEGA12: case CHIP_VEGA20: case CHIP_RAVEN: kfd_init_apertures_v9(pdd, id); diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_int_process_v9.c b/drivers/gpu/drm/amd/amdkfd/kfd_int_process_v9.c index f836897bbf58..a85904ad0d5f 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_int_process_v9.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_int_process_v9.c @@ -23,7 +23,7 @@ #include "kfd_priv.h" #include "kfd_events.h" #include "soc15_int.h" - +#include "kfd_device_queue_manager.h" static bool event_interrupt_isr_v9(struct kfd_dev *dev, const uint32_t *ih_ring_entry, @@ -39,20 +39,39 @@ static bool event_interrupt_isr_v9(struct kfd_dev *dev, vmid > dev->vm_info.last_vmid_kfd) return 0; - /* If there is no valid PASID, it's likely a firmware bug */ - pasid = SOC15_PASID_FROM_IH_ENTRY(ih_ring_entry); - if (WARN_ONCE(pasid == 0, "FW bug: No PASID in KFD interrupt")) - return 0; - source_id = SOC15_SOURCE_ID_FROM_IH_ENTRY(ih_ring_entry); client_id = SOC15_CLIENT_ID_FROM_IH_ENTRY(ih_ring_entry); + pasid = SOC15_PASID_FROM_IH_ENTRY(ih_ring_entry); + + /* This is a known issue for gfx9. Under non HWS, pasid is not set + * in the interrupt payload, so we need to find out the pasid on our + * own. + */ + if (!pasid && dev->dqm->sched_policy == KFD_SCHED_POLICY_NO_HWS) { + const uint32_t pasid_mask = 0xffff; - pr_debug("client id 0x%x, source id %d, pasid 0x%x. raw data:\n", - client_id, source_id, pasid); + *patched_flag = true; + memcpy(patched_ihre, ih_ring_entry, + dev->device_info->ih_ring_entry_size); + + pasid = dev->kfd2kgd->get_atc_vmid_pasid_mapping_pasid( + dev->kgd, vmid); + + /* Patch the pasid field */ + patched_ihre[3] = cpu_to_le32((le32_to_cpu(patched_ihre[3]) + & ~pasid_mask) | pasid); + } + + pr_debug("client id 0x%x, source id %d, vmid %d, pasid 0x%x. raw data:\n", + client_id, source_id, vmid, pasid); pr_debug("%8X, %8X, %8X, %8X, %8X, %8X, %8X, %8X.\n", data[0], data[1], data[2], data[3], data[4], data[5], data[6], data[7]); + /* If there is no valid PASID, it's likely a bug */ + if (WARN_ONCE(pasid == 0, "Bug: No PASID in KFD interrupt")) + return 0; + /* Interrupt types we care about: various signals and faults. * They will be forwarded to a work queue (see below). */ diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_kernel_queue.c b/drivers/gpu/drm/amd/amdkfd/kfd_kernel_queue.c index 6c31f7370193..f1596881f20a 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_kernel_queue.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_kernel_queue.c @@ -313,6 +313,7 @@ struct kernel_queue *kernel_queue_init(struct kfd_dev *dev, case CHIP_FIJI: case CHIP_POLARIS10: case CHIP_POLARIS11: + case CHIP_POLARIS12: kernel_queue_init_vi(&kq->ops_asic_specific); break; @@ -322,6 +323,7 @@ struct kernel_queue *kernel_queue_init(struct kfd_dev *dev, break; case CHIP_VEGA10: + case CHIP_VEGA12: case CHIP_VEGA20: case CHIP_RAVEN: kernel_queue_init_v9(&kq->ops_asic_specific); diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_module.c b/drivers/gpu/drm/amd/amdkfd/kfd_module.c index 8018163414ff..932007eb9168 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_module.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_module.c @@ -23,22 +23,7 @@ #include <linux/sched.h> #include <linux/device.h> #include "kfd_priv.h" - -static const struct kgd2kfd_calls kgd2kfd = { - .exit = kgd2kfd_exit, - .probe = kgd2kfd_probe, - .device_init = kgd2kfd_device_init, - .device_exit = kgd2kfd_device_exit, - .interrupt = kgd2kfd_interrupt, - .suspend = kgd2kfd_suspend, - .resume = kgd2kfd_resume, - .quiesce_mm = kgd2kfd_quiesce_mm, - .resume_mm = kgd2kfd_resume_mm, - .schedule_evict_and_restore_process = - kgd2kfd_schedule_evict_and_restore_process, - .pre_reset = kgd2kfd_pre_reset, - .post_reset = kgd2kfd_post_reset, -}; +#include "amdgpu_amdkfd.h" static int kfd_init(void) { @@ -91,20 +76,10 @@ static void kfd_exit(void) kfd_chardev_exit(); } -int kgd2kfd_init(unsigned int interface_version, - const struct kgd2kfd_calls **g2f) +int kgd2kfd_init() { - int err; - - err = kfd_init(); - if (err) - return err; - - *g2f = &kgd2kfd; - - return 0; + return kfd_init(); } -EXPORT_SYMBOL(kgd2kfd_init); void kgd2kfd_exit(void) { diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager.c b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager.c index e33019a7a883..aed9b9b82213 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager.c @@ -22,6 +22,7 @@ */ #include "kfd_mqd_manager.h" +#include "amdgpu_amdkfd.h" struct mqd_manager *mqd_manager_init(enum KFD_MQD_TYPE type, struct kfd_dev *dev) @@ -37,8 +38,10 @@ struct mqd_manager *mqd_manager_init(enum KFD_MQD_TYPE type, case CHIP_FIJI: case CHIP_POLARIS10: case CHIP_POLARIS11: + case CHIP_POLARIS12: return mqd_manager_init_vi_tonga(type, dev); case CHIP_VEGA10: + case CHIP_VEGA12: case CHIP_VEGA20: case CHIP_RAVEN: return mqd_manager_init_v9(type, dev); @@ -58,7 +61,7 @@ void mqd_symmetrically_map_cu_mask(struct mqd_manager *mm, uint32_t cu_per_sh[4] = {0}; int i, se, cu = 0; - mm->dev->kfd2kgd->get_cu_info(mm->dev->kgd, &cu_info); + amdgpu_amdkfd_get_cu_info(mm->dev->kgd, &cu_info); if (cu_mask_count > cu_info.cu_active_number) cu_mask_count = cu_info.cu_active_number; diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_cik.c b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_cik.c index 47243165a082..ae90a99909ef 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_cik.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_cik.c @@ -323,57 +323,7 @@ static int init_mqd_hiq(struct mqd_manager *mm, void **mqd, struct kfd_mem_obj **mqd_mem_obj, uint64_t *gart_addr, struct queue_properties *q) { - uint64_t addr; - struct cik_mqd *m; - int retval; - - retval = kfd_gtt_sa_allocate(mm->dev, sizeof(struct cik_mqd), - mqd_mem_obj); - - if (retval != 0) - return -ENOMEM; - - m = (struct cik_mqd *) (*mqd_mem_obj)->cpu_ptr; - addr = (*mqd_mem_obj)->gpu_addr; - - memset(m, 0, ALIGN(sizeof(struct cik_mqd), 256)); - - m->header = 0xC0310800; - m->compute_pipelinestat_enable = 1; - m->compute_static_thread_mgmt_se0 = 0xFFFFFFFF; - m->compute_static_thread_mgmt_se1 = 0xFFFFFFFF; - m->compute_static_thread_mgmt_se2 = 0xFFFFFFFF; - m->compute_static_thread_mgmt_se3 = 0xFFFFFFFF; - - m->cp_hqd_persistent_state = DEFAULT_CP_HQD_PERSISTENT_STATE | - PRELOAD_REQ; - m->cp_hqd_quantum = QUANTUM_EN | QUANTUM_SCALE_1MS | - QUANTUM_DURATION(10); - - m->cp_mqd_control = MQD_CONTROL_PRIV_STATE_EN; - m->cp_mqd_base_addr_lo = lower_32_bits(addr); - m->cp_mqd_base_addr_hi = upper_32_bits(addr); - - m->cp_hqd_ib_control = DEFAULT_MIN_IB_AVAIL_SIZE; - - /* - * Pipe Priority - * Identifies the pipe relative priority when this queue is connected - * to the pipeline. The pipe priority is against the GFX pipe and HP3D. - * In KFD we are using a fixed pipe priority set to CS_MEDIUM. - * 0 = CS_LOW (typically below GFX) - * 1 = CS_MEDIUM (typically between HP3D and GFX - * 2 = CS_HIGH (typically above HP3D) - */ - m->cp_hqd_pipe_priority = 1; - m->cp_hqd_queue_priority = 15; - - *mqd = m; - if (gart_addr) - *gart_addr = addr; - retval = mm->update_mqd(mm, m, q); - - return retval; + return init_mqd(mm, mqd, mqd_mem_obj, gart_addr, q); } static int update_mqd_hiq(struct mqd_manager *mm, void *mqd, diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v9.c b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v9.c index f381c1cb27bd..9dbba609450e 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v9.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v9.c @@ -30,6 +30,7 @@ #include "gc/gc_9_0_offset.h" #include "gc/gc_9_0_sh_mask.h" #include "sdma0/sdma0_4_0_sh_mask.h" +#include "amdgpu_amdkfd.h" static inline struct v9_mqd *get_mqd(void *mqd) { @@ -83,7 +84,7 @@ static int init_mqd(struct mqd_manager *mm, void **mqd, *mqd_mem_obj = kzalloc(sizeof(struct kfd_mem_obj), GFP_KERNEL); if (!*mqd_mem_obj) return -ENOMEM; - retval = kfd->kfd2kgd->init_gtt_mem_allocation(kfd->kgd, + retval = amdgpu_amdkfd_alloc_gtt_mem(kfd->kgd, ALIGN(q->ctl_stack_size, PAGE_SIZE) + ALIGN(sizeof(struct v9_mqd), PAGE_SIZE), &((*mqd_mem_obj)->gtt_mem), @@ -250,7 +251,7 @@ static void uninit_mqd(struct mqd_manager *mm, void *mqd, struct kfd_dev *kfd = mm->dev; if (mqd_mem_obj->gtt_mem) { - kfd->kfd2kgd->free_gtt_mem(kfd->kgd, mqd_mem_obj->gtt_mem); + amdgpu_amdkfd_free_gtt_mem(kfd->kgd, mqd_mem_obj->gtt_mem); kfree(mqd_mem_obj); } else { kfd_gtt_sa_free(mm->dev, mqd_mem_obj); diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_packet_manager.c b/drivers/gpu/drm/amd/amdkfd/kfd_packet_manager.c index c6080ed3b6a7..045a229436a0 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_packet_manager.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_packet_manager.c @@ -226,9 +226,11 @@ int pm_init(struct packet_manager *pm, struct device_queue_manager *dqm) case CHIP_FIJI: case CHIP_POLARIS10: case CHIP_POLARIS11: + case CHIP_POLARIS12: pm->pmf = &kfd_vi_pm_funcs; break; case CHIP_VEGA10: + case CHIP_VEGA12: case CHIP_VEGA20: case CHIP_RAVEN: pm->pmf = &kfd_v9_pm_funcs; diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_pasid.c b/drivers/gpu/drm/amd/amdkfd/kfd_pasid.c index 15fff4420e53..33b08ff00b50 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_pasid.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_pasid.c @@ -22,6 +22,7 @@ #include <linux/types.h> #include "kfd_priv.h" +#include "amdgpu_ids.h" static unsigned int pasid_bits = 16; static const struct kfd2kgd_calls *kfd2kgd; @@ -71,7 +72,7 @@ unsigned int kfd_pasid_alloc(void) return false; } - r = kfd2kgd->alloc_pasid(pasid_bits); + r = amdgpu_pasid_alloc(pasid_bits); return r > 0 ? r : 0; } @@ -79,5 +80,5 @@ unsigned int kfd_pasid_alloc(void) void kfd_pasid_free(unsigned int pasid) { if (kfd2kgd) - kfd2kgd->free_pasid(pasid); + amdgpu_pasid_free(pasid); } diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h index 53ff86d45d91..0eeee3c6d6dc 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h +++ b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h @@ -97,17 +97,29 @@ #define KFD_CWSR_TBA_TMA_SIZE (PAGE_SIZE * 2) #define KFD_CWSR_TMA_OFFSET PAGE_SIZE +#define KFD_MAX_NUM_OF_QUEUES_PER_DEVICE \ + (KFD_MAX_NUM_OF_PROCESSES * \ + KFD_MAX_NUM_OF_QUEUES_PER_PROCESS) + +#define KFD_KERNEL_QUEUE_SIZE 2048 + +/* + * 512 = 0x200 + * The doorbell index distance between SDMA RLC (2*i) and (2*i+1) in the + * same SDMA engine on SOC15, which has 8-byte doorbells for SDMA. + * 512 8-byte doorbell distance (i.e. one page away) ensures that SDMA RLC + * (2*i+1) doorbells (in terms of the lower 12 bit address) lie exactly in + * the OFFSET and SIZE set in registers like BIF_SDMA0_DOORBELL_RANGE. + */ +#define KFD_QUEUE_DOORBELL_MIRROR_OFFSET 512 + + /* * Kernel module parameter to specify maximum number of supported queues per * device */ extern int max_num_of_queues_per_device; -#define KFD_MAX_NUM_OF_QUEUES_PER_DEVICE \ - (KFD_MAX_NUM_OF_PROCESSES * \ - KFD_MAX_NUM_OF_QUEUES_PER_PROCESS) - -#define KFD_KERNEL_QUEUE_SIZE 2048 /* Kernel module parameter to specify the scheduling policy */ extern int sched_policy; @@ -266,14 +278,6 @@ struct kfd_dev { bool pci_atomic_requested; }; -/* KGD2KFD callbacks */ -void kgd2kfd_exit(void); -struct kfd_dev *kgd2kfd_probe(struct kgd_dev *kgd, - struct pci_dev *pdev, const struct kfd2kgd_calls *f2g); -bool kgd2kfd_device_init(struct kfd_dev *kfd, - const struct kgd2kfd_shared_resources *gpu_resources); -void kgd2kfd_device_exit(struct kfd_dev *kfd); - enum kfd_mempool { KFD_MEMPOOL_SYSTEM_CACHEABLE = 1, KFD_MEMPOOL_SYSTEM_WRITECOMBINE = 2, @@ -507,6 +511,7 @@ struct qcm_process_device { * All the memory management data should be here too */ uint64_t gds_context_area; + /* Contains page table flags such as AMDGPU_PTE_VALID since gfx9 */ uint64_t page_table_base; uint32_t sh_mem_config; uint32_t sh_mem_bases; @@ -540,11 +545,6 @@ struct qcm_process_device { /* Approx. time before evicting the process again */ #define PROCESS_ACTIVE_TIME_MS 10 -int kgd2kfd_quiesce_mm(struct mm_struct *mm); -int kgd2kfd_resume_mm(struct mm_struct *mm); -int kgd2kfd_schedule_evict_and_restore_process(struct mm_struct *mm, - struct dma_fence *fence); - /* 8 byte handle containing GPU ID in the most significant 4 bytes and * idr_handle in the least significant 4 bytes */ @@ -792,26 +792,18 @@ struct kfd_topology_device *kfd_topology_device_by_proximity_domain( struct kfd_topology_device *kfd_topology_device_by_id(uint32_t gpu_id); struct kfd_dev *kfd_device_by_id(uint32_t gpu_id); struct kfd_dev *kfd_device_by_pci_dev(const struct pci_dev *pdev); +struct kfd_dev *kfd_device_by_kgd(const struct kgd_dev *kgd); int kfd_topology_enum_kfd_devices(uint8_t idx, struct kfd_dev **kdev); int kfd_numa_node_to_apic_id(int numa_node_id); /* Interrupts */ int kfd_interrupt_init(struct kfd_dev *dev); void kfd_interrupt_exit(struct kfd_dev *dev); -void kgd2kfd_interrupt(struct kfd_dev *kfd, const void *ih_ring_entry); bool enqueue_ih_ring_entry(struct kfd_dev *kfd, const void *ih_ring_entry); bool interrupt_is_wanted(struct kfd_dev *dev, const uint32_t *ih_ring_entry, uint32_t *patched_ihre, bool *flag); -/* Power Management */ -void kgd2kfd_suspend(struct kfd_dev *kfd); -int kgd2kfd_resume(struct kfd_dev *kfd); - -/* GPU reset */ -int kgd2kfd_pre_reset(struct kfd_dev *kfd); -int kgd2kfd_post_reset(struct kfd_dev *kfd); - /* amdkfd Apertures */ int kfd_init_apertures(struct kfd_process *process); diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_process.c b/drivers/gpu/drm/amd/amdkfd/kfd_process.c index 0039e451d9af..4bdae78bab8e 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_process.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_process.c @@ -31,6 +31,7 @@ #include <linux/compat.h> #include <linux/mman.h> #include <linux/file.h> +#include "amdgpu_amdkfd.h" struct mm_struct; @@ -100,8 +101,8 @@ static void kfd_process_free_gpuvm(struct kgd_mem *mem, { struct kfd_dev *dev = pdd->dev; - dev->kfd2kgd->unmap_memory_to_gpu(dev->kgd, mem, pdd->vm); - dev->kfd2kgd->free_memory_of_gpu(dev->kgd, mem); + amdgpu_amdkfd_gpuvm_unmap_memory_from_gpu(dev->kgd, mem, pdd->vm); + amdgpu_amdkfd_gpuvm_free_memory_of_gpu(dev->kgd, mem); } /* kfd_process_alloc_gpuvm - Allocate GPU VM for the KFD process @@ -119,16 +120,16 @@ static int kfd_process_alloc_gpuvm(struct kfd_process_device *pdd, int handle; int err; - err = kdev->kfd2kgd->alloc_memory_of_gpu(kdev->kgd, gpu_va, size, + err = amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu(kdev->kgd, gpu_va, size, pdd->vm, &mem, NULL, flags); if (err) goto err_alloc_mem; - err = kdev->kfd2kgd->map_memory_to_gpu(kdev->kgd, mem, pdd->vm); + err = amdgpu_amdkfd_gpuvm_map_memory_to_gpu(kdev->kgd, mem, pdd->vm); if (err) goto err_map_mem; - err = kdev->kfd2kgd->sync_memory(kdev->kgd, mem, true); + err = amdgpu_amdkfd_gpuvm_sync_memory(kdev->kgd, mem, true); if (err) { pr_debug("Sync memory failed, wait interrupted by user signal\n"); goto sync_memory_failed; @@ -147,7 +148,7 @@ static int kfd_process_alloc_gpuvm(struct kfd_process_device *pdd, } if (kptr) { - err = kdev->kfd2kgd->map_gtt_bo_to_kernel(kdev->kgd, + err = amdgpu_amdkfd_gpuvm_map_gtt_bo_to_kernel(kdev->kgd, (struct kgd_mem *)mem, kptr, NULL); if (err) { pr_debug("Map GTT BO to kernel failed\n"); @@ -165,7 +166,7 @@ sync_memory_failed: return err; err_map_mem: - kdev->kfd2kgd->free_memory_of_gpu(kdev->kgd, mem); + amdgpu_amdkfd_gpuvm_free_memory_of_gpu(kdev->kgd, mem); err_alloc_mem: *kptr = NULL; return err; @@ -296,11 +297,11 @@ static void kfd_process_device_free_bos(struct kfd_process_device *pdd) per_device_list) { if (!peer_pdd->vm) continue; - peer_pdd->dev->kfd2kgd->unmap_memory_to_gpu( + amdgpu_amdkfd_gpuvm_unmap_memory_from_gpu( peer_pdd->dev->kgd, mem, peer_pdd->vm); } - pdd->dev->kfd2kgd->free_memory_of_gpu(pdd->dev->kgd, mem); + amdgpu_amdkfd_gpuvm_free_memory_of_gpu(pdd->dev->kgd, mem); kfd_process_device_remove_obj_handle(pdd, id); } } @@ -323,11 +324,12 @@ static void kfd_process_destroy_pdds(struct kfd_process *p) pdd->dev->id, p->pasid); if (pdd->drm_file) { - pdd->dev->kfd2kgd->release_process_vm(pdd->dev->kgd, pdd->vm); + amdgpu_amdkfd_gpuvm_release_process_vm( + pdd->dev->kgd, pdd->vm); fput(pdd->drm_file); } else if (pdd->vm) - pdd->dev->kfd2kgd->destroy_process_vm( + amdgpu_amdkfd_gpuvm_destroy_process_vm( pdd->dev->kgd, pdd->vm); list_del(&pdd->per_device_list); @@ -605,13 +607,17 @@ static int init_doorbell_bitmap(struct qcm_process_device *qpd, if (!qpd->doorbell_bitmap) return -ENOMEM; - /* Mask out any reserved doorbells */ - for (i = 0; i < KFD_MAX_NUM_OF_QUEUES_PER_PROCESS; i++) - if ((dev->shared_resources.reserved_doorbell_mask & i) == - dev->shared_resources.reserved_doorbell_val) { + /* Mask out doorbells reserved for SDMA, IH, and VCN on SOC15. */ + for (i = 0; i < KFD_MAX_NUM_OF_QUEUES_PER_PROCESS / 2; i++) { + if (i >= dev->shared_resources.non_cp_doorbells_start + && i <= dev->shared_resources.non_cp_doorbells_end) { set_bit(i, qpd->doorbell_bitmap); - pr_debug("reserved doorbell 0x%03x\n", i); + set_bit(i + KFD_QUEUE_DOORBELL_MIRROR_OFFSET, + qpd->doorbell_bitmap); + pr_debug("reserved doorbell 0x%03x and 0x%03x\n", i, + i + KFD_QUEUE_DOORBELL_MIRROR_OFFSET); } + } return 0; } @@ -688,12 +694,12 @@ int kfd_process_device_init_vm(struct kfd_process_device *pdd, dev = pdd->dev; if (drm_file) - ret = dev->kfd2kgd->acquire_process_vm( + ret = amdgpu_amdkfd_gpuvm_acquire_process_vm( dev->kgd, drm_file, p->pasid, &pdd->vm, &p->kgd_process_info, &p->ef); else - ret = dev->kfd2kgd->create_process_vm( - dev->kgd, p->pasid, &pdd->vm, &p->kgd_process_info, &p->ef); + ret = amdgpu_amdkfd_gpuvm_create_process_vm(dev->kgd, p->pasid, + &pdd->vm, &p->kgd_process_info, &p->ef); if (ret) { pr_err("Failed to create process VM object\n"); return ret; @@ -714,7 +720,7 @@ err_init_cwsr: err_reserve_ib_mem: kfd_process_device_free_bos(pdd); if (!drm_file) - dev->kfd2kgd->destroy_process_vm(dev->kgd, pdd->vm); + amdgpu_amdkfd_gpuvm_destroy_process_vm(dev->kgd, pdd->vm); pdd->vm = NULL; return ret; @@ -972,7 +978,7 @@ static void restore_process_worker(struct work_struct *work) */ p->last_restore_timestamp = get_jiffies_64(); - ret = pdd->dev->kfd2kgd->restore_process_bos(p->kgd_process_info, + ret = amdgpu_amdkfd_gpuvm_restore_process_bos(p->kgd_process_info, &p->ef); if (ret) { pr_debug("Failed to restore BOs of pasid %d, retry after %d ms\n", diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_topology.c b/drivers/gpu/drm/amd/amdkfd/kfd_topology.c index e3843c5929ed..09da91644f9f 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_topology.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_topology.c @@ -36,6 +36,7 @@ #include "kfd_topology.h" #include "kfd_device_queue_manager.h" #include "kfd_iommu.h" +#include "amdgpu_amdkfd.h" /* topology_device_list - Master list of all topology devices */ static struct list_head topology_device_list; @@ -100,7 +101,25 @@ struct kfd_dev *kfd_device_by_pci_dev(const struct pci_dev *pdev) down_read(&topology_lock); list_for_each_entry(top_dev, &topology_device_list, list) - if (top_dev->gpu->pdev == pdev) { + if (top_dev->gpu && top_dev->gpu->pdev == pdev) { + device = top_dev->gpu; + break; + } + + up_read(&topology_lock); + + return device; +} + +struct kfd_dev *kfd_device_by_kgd(const struct kgd_dev *kgd) +{ + struct kfd_topology_device *top_dev; + struct kfd_dev *device = NULL; + + down_read(&topology_lock); + + list_for_each_entry(top_dev, &topology_device_list, list) + if (top_dev->gpu && top_dev->gpu->kgd == kgd) { device = top_dev->gpu; break; } @@ -1052,7 +1071,7 @@ static uint32_t kfd_generate_gpu_id(struct kfd_dev *gpu) if (!gpu) return 0; - gpu->kfd2kgd->get_local_mem_info(gpu->kgd, &local_mem_info); + amdgpu_amdkfd_get_local_mem_info(gpu->kgd, &local_mem_info); local_mem_size = local_mem_info.local_mem_size_private + local_mem_info.local_mem_size_public; @@ -1074,8 +1093,6 @@ static uint32_t kfd_generate_gpu_id(struct kfd_dev *gpu) * the GPU device is not already present in the topology device * list then return NULL. This means a new topology device has to * be created for this GPU. - * TODO: Rather than assiging @gpu to first topology device withtout - * gpu attached, it will better to have more stringent check. */ static struct kfd_topology_device *kfd_assign_gpu(struct kfd_dev *gpu) { @@ -1083,12 +1100,20 @@ static struct kfd_topology_device *kfd_assign_gpu(struct kfd_dev *gpu) struct kfd_topology_device *out_dev = NULL; down_write(&topology_lock); - list_for_each_entry(dev, &topology_device_list, list) + list_for_each_entry(dev, &topology_device_list, list) { + /* Discrete GPUs need their own topology device list + * entries. Don't assign them to CPU/APU nodes. + */ + if (!gpu->device_info->needs_iommu_device && + dev->node_props.cpu_cores_count) + continue; + if (!dev->gpu && (dev->node_props.simd_count > 0)) { dev->gpu = gpu; out_dev = dev; break; } + } up_write(&topology_lock); return out_dev; } @@ -1118,8 +1143,7 @@ static void kfd_fill_mem_clk_max_info(struct kfd_topology_device *dev) * for APUs - If CRAT from ACPI reports more than one bank, then * all the banks will report the same mem_clk_max information */ - dev->gpu->kfd2kgd->get_local_mem_info(dev->gpu->kgd, - &local_mem_info); + amdgpu_amdkfd_get_local_mem_info(dev->gpu->kgd, &local_mem_info); list_for_each_entry(mem, &dev->mem_props, list) mem->mem_clk_max = local_mem_info.mem_clk_max; @@ -1240,7 +1264,7 @@ int kfd_topology_add_device(struct kfd_dev *gpu) * needed for the topology */ - dev->gpu->kfd2kgd->get_cu_info(dev->gpu->kgd, &cu_info); + amdgpu_amdkfd_get_cu_info(dev->gpu->kgd, &cu_info); dev->node_props.simd_arrays_per_engine = cu_info.num_shader_arrays_per_engine; @@ -1249,7 +1273,7 @@ int kfd_topology_add_device(struct kfd_dev *gpu) dev->node_props.location_id = PCI_DEVID(gpu->pdev->bus->number, gpu->pdev->devfn); dev->node_props.max_engine_clk_fcompute = - dev->gpu->kfd2kgd->get_max_engine_clock_in_mhz(dev->gpu->kgd); + amdgpu_amdkfd_get_max_engine_clock_in_mhz(dev->gpu->kgd); dev->node_props.max_engine_clk_ccompute = cpufreq_quick_get_max(0) / 1000; dev->node_props.drm_render_minor = @@ -1272,12 +1296,14 @@ int kfd_topology_add_device(struct kfd_dev *gpu) case CHIP_FIJI: case CHIP_POLARIS10: case CHIP_POLARIS11: + case CHIP_POLARIS12: pr_debug("Adding doorbell packet type capability\n"); dev->node_props.capability |= ((HSA_CAP_DOORBELL_TYPE_1_0 << HSA_CAP_DOORBELL_TYPE_TOTALBITS_SHIFT) & HSA_CAP_DOORBELL_TYPE_TOTALBITS_MASK); break; case CHIP_VEGA10: + case CHIP_VEGA12: case CHIP_VEGA20: case CHIP_RAVEN: dev->node_props.capability |= ((HSA_CAP_DOORBELL_TYPE_2_0 << @@ -1372,7 +1398,6 @@ int kfd_topology_enum_kfd_devices(uint8_t idx, struct kfd_dev **kdev) static int kfd_cpumask_to_apic_id(const struct cpumask *cpumask) { - const struct cpuinfo_x86 *cpuinfo; int first_cpu_of_numa_node; if (!cpumask || cpumask == cpu_none_mask) @@ -1380,9 +1405,11 @@ static int kfd_cpumask_to_apic_id(const struct cpumask *cpumask) first_cpu_of_numa_node = cpumask_first(cpumask); if (first_cpu_of_numa_node >= nr_cpu_ids) return -1; - cpuinfo = &cpu_data(first_cpu_of_numa_node); - - return cpuinfo->apicid; +#ifdef CONFIG_X86_64 + return cpu_data(first_cpu_of_numa_node).apicid; +#else + return first_cpu_of_numa_node; +#endif } /* kfd_numa_node_to_apic_id - Returns the APIC ID of the first logical processor diff --git a/drivers/gpu/drm/amd/display/Makefile b/drivers/gpu/drm/amd/display/Makefile index c97dc9613325..cfde1568c79a 100644 --- a/drivers/gpu/drm/amd/display/Makefile +++ b/drivers/gpu/drm/amd/display/Makefile @@ -32,11 +32,12 @@ subdir-ccflags-y += -I$(FULL_AMD_DISPLAY_PATH)/modules/inc subdir-ccflags-y += -I$(FULL_AMD_DISPLAY_PATH)/modules/freesync subdir-ccflags-y += -I$(FULL_AMD_DISPLAY_PATH)/modules/color subdir-ccflags-y += -I$(FULL_AMD_DISPLAY_PATH)/modules/info_packet +subdir-ccflags-y += -I$(FULL_AMD_DISPLAY_PATH)/modules/power #TODO: remove when Timing Sync feature is complete subdir-ccflags-y += -DBUILD_FEATURE_TIMING_SYNC=0 -DAL_LIBS = amdgpu_dm dc modules/freesync modules/color modules/info_packet +DAL_LIBS = amdgpu_dm dc modules/freesync modules/color modules/info_packet modules/power AMD_DAL = $(addsuffix /Makefile, $(addprefix $(FULL_AMD_DISPLAY_PATH)/,$(DAL_LIBS))) diff --git a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c index 5a6edf65c9ea..3082b55b1e77 100644 --- a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c +++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c @@ -23,6 +23,9 @@ * */ +/* The caprices of the preprocessor require that this be declared right here */ +#define CREATE_TRACE_POINTS + #include "dm_services_types.h" #include "dc.h" #include "dc/inc/core_types.h" @@ -38,7 +41,6 @@ #include "amd_shared.h" #include "amdgpu_dm_irq.h" #include "dm_helpers.h" -#include "dm_services_types.h" #include "amdgpu_dm_mst_types.h" #if defined(CONFIG_DEBUG_FS) #include "amdgpu_dm_debugfs.h" @@ -55,6 +57,7 @@ #include <drm/drmP.h> #include <drm/drm_atomic.h> +#include <drm/drm_atomic_uapi.h> #include <drm/drm_atomic_helper.h> #include <drm/drm_dp_mst_helper.h> #include <drm/drm_fb_helper.h> @@ -72,10 +75,22 @@ #endif #include "modules/inc/mod_freesync.h" +#include "modules/power/power_helpers.h" +#include "modules/inc/mod_info_packet.h" #define FIRMWARE_RAVEN_DMCU "amdgpu/raven_dmcu.bin" MODULE_FIRMWARE(FIRMWARE_RAVEN_DMCU); +/** + * DOC: overview + * + * The AMDgpu display manager, **amdgpu_dm** (or even simpler, + * **dm**) sits between DRM and DC. It acts as a liason, converting DRM + * requests into DC requests, and DC responses into DRM responses. + * + * The root control structure is &struct amdgpu_display_manager. + */ + /* basic init/fini API */ static int amdgpu_dm_init(struct amdgpu_device *adev); static void amdgpu_dm_fini(struct amdgpu_device *adev); @@ -95,7 +110,7 @@ static void amdgpu_dm_update_connector_after_detect(struct amdgpu_dm_connector *aconnector); static int amdgpu_dm_plane_init(struct amdgpu_display_manager *dm, - struct amdgpu_plane *aplane, + struct drm_plane *plane, unsigned long possible_crtcs); static int amdgpu_dm_crtc_init(struct amdgpu_display_manager *dm, struct drm_plane *plane, @@ -119,6 +134,8 @@ static void amdgpu_dm_atomic_commit_tail(struct drm_atomic_state *state); static int amdgpu_dm_atomic_check(struct drm_device *dev, struct drm_atomic_state *state); +static void handle_cursor_update(struct drm_plane *plane, + struct drm_plane_state *old_plane_state); @@ -286,12 +303,11 @@ static void dm_pflip_high_irq(void *interrupt_params) return; } + /* Update to correct count(s) if racing with vblank irq */ + amdgpu_crtc->last_flip_vblank = drm_crtc_accurate_vblank_count(&amdgpu_crtc->base); /* wake up userspace */ if (amdgpu_crtc->event) { - /* Update to correct count(s) if racing with vblank irq */ - drm_crtc_accurate_vblank_count(&amdgpu_crtc->base); - drm_crtc_send_vblank_event(&amdgpu_crtc->base, amdgpu_crtc->event); /* page flip completed. clean up */ @@ -314,12 +330,29 @@ static void dm_crtc_high_irq(void *interrupt_params) struct common_irq_params *irq_params = interrupt_params; struct amdgpu_device *adev = irq_params->adev; struct amdgpu_crtc *acrtc; + struct dm_crtc_state *acrtc_state; acrtc = get_crtc_by_otg_inst(adev, irq_params->irq_src - IRQ_TYPE_VBLANK); if (acrtc) { drm_crtc_handle_vblank(&acrtc->base); amdgpu_dm_crtc_handle_crc_irq(&acrtc->base); + + acrtc_state = to_dm_crtc_state(acrtc->base.state); + + if (acrtc_state->stream && + acrtc_state->vrr_params.supported && + acrtc_state->freesync_config.state == VRR_STATE_ACTIVE_VARIABLE) { + mod_freesync_handle_v_update( + adev->dm.freesync_module, + acrtc_state->stream, + &acrtc_state->vrr_params); + + dc_stream_adjust_vmin_vmax( + adev->dm.dc, + acrtc_state->stream, + &acrtc_state->vrr_params.adjust); + } } } @@ -379,11 +412,6 @@ static void amdgpu_dm_fbc_init(struct drm_connector *connector) } -/* - * Init display KMS - * - * Returns 0 on success - */ static int amdgpu_dm_init(struct amdgpu_device *adev) { struct dc_init_data init_data; @@ -393,6 +421,8 @@ static int amdgpu_dm_init(struct amdgpu_device *adev) /* Zero all the fields */ memset(&init_data, 0, sizeof(init_data)); + mutex_init(&adev->dm.dc_lock); + if(amdgpu_dm_irq_init(adev)) { DRM_ERROR("amdgpu: failed to initialize DM IRQ support.\n"); goto error; @@ -507,6 +537,9 @@ static void amdgpu_dm_fini(struct amdgpu_device *adev) /* DC Destroy TODO: Replace destroy DAL */ if (adev->dm.dc) dc_destroy(&adev->dm.dc); + + mutex_destroy(&adev->dm.dc_lock); + return; } @@ -638,6 +671,26 @@ static int dm_late_init(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; + struct dmcu_iram_parameters params; + unsigned int linear_lut[16]; + int i; + struct dmcu *dmcu = adev->dm.dc->res_pool->dmcu; + bool ret; + + for (i = 0; i < 16; i++) + linear_lut[i] = 0xFFFF * i / 15; + + params.set = 0; + params.backlight_ramping_start = 0xCCCC; + params.backlight_ramping_reduction = 0xCCCCCCCC; + params.backlight_lut_array_size = 16; + params.backlight_lut_array = linear_lut; + + ret = dmcu_load_iram(dmcu, params); + + if (!ret) + return -EINVAL; + return detect_mst_link_for_all_connectors(adev->ddev); } @@ -645,24 +698,58 @@ static void s3_handle_mst(struct drm_device *dev, bool suspend) { struct amdgpu_dm_connector *aconnector; struct drm_connector *connector; + struct drm_dp_mst_topology_mgr *mgr; + int ret; + bool need_hotplug = false; drm_modeset_lock(&dev->mode_config.connection_mutex, NULL); - list_for_each_entry(connector, &dev->mode_config.connector_list, head) { - aconnector = to_amdgpu_dm_connector(connector); - if (aconnector->dc_link->type == dc_connection_mst_branch && - !aconnector->mst_port) { + list_for_each_entry(connector, &dev->mode_config.connector_list, + head) { + aconnector = to_amdgpu_dm_connector(connector); + if (aconnector->dc_link->type != dc_connection_mst_branch || + aconnector->mst_port) + continue; - if (suspend) - drm_dp_mst_topology_mgr_suspend(&aconnector->mst_mgr); - else - drm_dp_mst_topology_mgr_resume(&aconnector->mst_mgr); - } + mgr = &aconnector->mst_mgr; + + if (suspend) { + drm_dp_mst_topology_mgr_suspend(mgr); + } else { + ret = drm_dp_mst_topology_mgr_resume(mgr); + if (ret < 0) { + drm_dp_mst_topology_mgr_set_mst(mgr, false); + need_hotplug = true; + } + } } drm_modeset_unlock(&dev->mode_config.connection_mutex); + + if (need_hotplug) + drm_kms_helper_hotplug_event(dev); } +/** + * dm_hw_init() - Initialize DC device + * @handle: The base driver device containing the amdpgu_dm device. + * + * Initialize the &struct amdgpu_display_manager device. This involves calling + * the initializers of each DM component, then populating the struct with them. + * + * Although the function implies hardware initialization, both hardware and + * software are initialized here. Splitting them out to their relevant init + * hooks is a future TODO item. + * + * Some notable things that are initialized here: + * + * - Display Core, both software and hardware + * - DC modules that we need (freesync and color management) + * - DRM software states + * - Interrupt sources and handlers + * - Vblank support + * - Debug FS entries, if enabled + */ static int dm_hw_init(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; @@ -673,6 +760,14 @@ static int dm_hw_init(void *handle) return 0; } +/** + * dm_hw_fini() - Teardown DC device + * @handle: The base driver device containing the amdpgu_dm device. + * + * Teardown components within &struct amdgpu_display_manager that require + * cleanup. This involves cleaning up the DRM device, DC, and any modules that + * were loaded. Also flush IRQ workqueues and disable them. + */ static int dm_hw_fini(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; @@ -690,12 +785,13 @@ static int dm_suspend(void *handle) struct amdgpu_display_manager *dm = &adev->dm; int ret = 0; + WARN_ON(adev->dm.cached_state); + adev->dm.cached_state = drm_atomic_helper_suspend(adev->ddev); + s3_handle_mst(adev->ddev, true); amdgpu_dm_irq_suspend(adev); - WARN_ON(adev->dm.cached_state); - adev->dm.cached_state = drm_atomic_helper_suspend(adev->ddev); dc_set_power_state(dm->dc, DC_ACPI_CM_POWER_STATE_D3); @@ -790,6 +886,7 @@ static void emulated_link_detect(struct dc_link *link) return; } + /* dc_sink_create returns a new reference */ link->local_sink = sink; edid_status = dm_helpers_read_local_edid( @@ -816,7 +913,6 @@ static int dm_resume(void *handle) struct drm_plane_state *new_plane_state; struct dm_plane_state *dm_new_plane_state; enum dc_connection_type new_connection_type = dc_connection_none; - int ret; int i; /* power on hardware */ @@ -857,6 +953,8 @@ static int dm_resume(void *handle) if (aconnector->fake_enable && aconnector->dc_link->local_sink) aconnector->fake_enable = false; + if (aconnector->dc_sink) + dc_sink_release(aconnector->dc_sink); aconnector->dc_sink = NULL; amdgpu_dm_update_connector_after_detect(aconnector); mutex_unlock(&aconnector->hpd_lock); @@ -889,15 +987,25 @@ static int dm_resume(void *handle) } } - ret = drm_atomic_helper_resume(ddev, dm->cached_state); + drm_atomic_helper_resume(ddev, dm->cached_state); dm->cached_state = NULL; amdgpu_dm_irq_resume_late(adev); - return ret; + return 0; } +/** + * DOC: DM Lifecycle + * + * DM (and consequently DC) is registered in the amdgpu base driver as a IP + * block. When CONFIG_DRM_AMD_DC is enabled, the DM device IP block is added to + * the base driver's device list to be initialized and torn down accordingly. + * + * The functions to do so are provided as hooks in &struct amd_ip_funcs. + */ + static const struct amd_ip_funcs amdgpu_dm_funcs = { .name = "dm", .early_init = dm_early_init, @@ -926,53 +1034,17 @@ const struct amdgpu_ip_block_version dm_ip_block = }; -static struct drm_atomic_state * -dm_atomic_state_alloc(struct drm_device *dev) -{ - struct dm_atomic_state *state = kzalloc(sizeof(*state), GFP_KERNEL); - - if (!state) - return NULL; - - if (drm_atomic_state_init(dev, &state->base) < 0) - goto fail; - - return &state->base; - -fail: - kfree(state); - return NULL; -} - -static void -dm_atomic_state_clear(struct drm_atomic_state *state) -{ - struct dm_atomic_state *dm_state = to_dm_atomic_state(state); - - if (dm_state->context) { - dc_release_state(dm_state->context); - dm_state->context = NULL; - } - - drm_atomic_state_default_clear(state); -} - -static void -dm_atomic_state_alloc_free(struct drm_atomic_state *state) -{ - struct dm_atomic_state *dm_state = to_dm_atomic_state(state); - drm_atomic_state_default_release(state); - kfree(dm_state); -} +/** + * DOC: atomic + * + * *WIP* + */ static const struct drm_mode_config_funcs amdgpu_dm_mode_funcs = { .fb_create = amdgpu_display_user_framebuffer_create, .output_poll_changed = drm_fb_helper_output_poll_changed, .atomic_check = amdgpu_dm_atomic_check, .atomic_commit = amdgpu_dm_atomic_commit, - .atomic_state_alloc = dm_atomic_state_alloc, - .atomic_state_clear = dm_atomic_state_clear, - .atomic_state_free = dm_atomic_state_alloc_free }; static struct drm_mode_config_helper_funcs amdgpu_dm_mode_config_helperfuncs = { @@ -992,6 +1064,8 @@ amdgpu_dm_update_connector_after_detect(struct amdgpu_dm_connector *aconnector) sink = aconnector->dc_link->local_sink; + if (sink) + dc_sink_retain(sink); /* * Edid mgmt connector gets first update only in mode_valid hook and then @@ -1016,21 +1090,24 @@ amdgpu_dm_update_connector_after_detect(struct amdgpu_dm_connector *aconnector) * to it anymore after disconnect, so on next crtc to connector * reshuffle by UMD we will get into unwanted dc_sink release */ - if (aconnector->dc_sink != aconnector->dc_em_sink) - dc_sink_release(aconnector->dc_sink); + dc_sink_release(aconnector->dc_sink); } aconnector->dc_sink = sink; + dc_sink_retain(aconnector->dc_sink); amdgpu_dm_update_freesync_caps(connector, aconnector->edid); } else { amdgpu_dm_update_freesync_caps(connector, NULL); - if (!aconnector->dc_sink) + if (!aconnector->dc_sink) { aconnector->dc_sink = aconnector->dc_em_sink; - else if (aconnector->dc_sink != aconnector->dc_em_sink) dc_sink_retain(aconnector->dc_sink); + } } mutex_unlock(&dev->mode_config.mutex); + + if (sink) + dc_sink_release(sink); return; } @@ -1038,8 +1115,10 @@ amdgpu_dm_update_connector_after_detect(struct amdgpu_dm_connector *aconnector) * TODO: temporary guard to look for proper fix * if this sink is MST sink, we should not do anything */ - if (sink && sink->sink_signal == SIGNAL_TYPE_DISPLAY_PORT_MST) + if (sink && sink->sink_signal == SIGNAL_TYPE_DISPLAY_PORT_MST) { + dc_sink_release(sink); return; + } if (aconnector->dc_sink == sink) { /* @@ -1048,6 +1127,8 @@ amdgpu_dm_update_connector_after_detect(struct amdgpu_dm_connector *aconnector) */ DRM_DEBUG_DRIVER("DCHPD: connector_id=%d: dc_sink didn't change.\n", aconnector->connector_id); + if (sink) + dc_sink_release(sink); return; } @@ -1069,6 +1150,7 @@ amdgpu_dm_update_connector_after_detect(struct amdgpu_dm_connector *aconnector) amdgpu_dm_update_freesync_caps(connector, NULL); aconnector->dc_sink = sink; + dc_sink_retain(aconnector->dc_sink); if (sink->dc_edid.length == 0) { aconnector->edid = NULL; drm_dp_cec_unset_edid(&aconnector->dm_dp_aux.aux); @@ -1089,11 +1171,15 @@ amdgpu_dm_update_connector_after_detect(struct amdgpu_dm_connector *aconnector) amdgpu_dm_update_freesync_caps(connector, NULL); drm_connector_update_edid_property(connector, NULL); aconnector->num_modes = 0; + dc_sink_release(aconnector->dc_sink); aconnector->dc_sink = NULL; aconnector->edid = NULL; } mutex_unlock(&dev->mode_config.mutex); + + if (sink) + dc_sink_release(sink); } static void handle_hpd_irq(void *param) @@ -1494,8 +1580,117 @@ static int dcn10_register_irq_handlers(struct amdgpu_device *adev) } #endif +/* + * Acquires the lock for the atomic state object and returns + * the new atomic state. + * + * This should only be called during atomic check. + */ +static int dm_atomic_get_state(struct drm_atomic_state *state, + struct dm_atomic_state **dm_state) +{ + struct drm_device *dev = state->dev; + struct amdgpu_device *adev = dev->dev_private; + struct amdgpu_display_manager *dm = &adev->dm; + struct drm_private_state *priv_state; + int ret; + + if (*dm_state) + return 0; + + ret = drm_modeset_lock(&dm->atomic_obj_lock, state->acquire_ctx); + if (ret) + return ret; + + priv_state = drm_atomic_get_private_obj_state(state, &dm->atomic_obj); + if (IS_ERR(priv_state)) + return PTR_ERR(priv_state); + + *dm_state = to_dm_atomic_state(priv_state); + + return 0; +} + +struct dm_atomic_state * +dm_atomic_get_new_state(struct drm_atomic_state *state) +{ + struct drm_device *dev = state->dev; + struct amdgpu_device *adev = dev->dev_private; + struct amdgpu_display_manager *dm = &adev->dm; + struct drm_private_obj *obj; + struct drm_private_state *new_obj_state; + int i; + + for_each_new_private_obj_in_state(state, obj, new_obj_state, i) { + if (obj->funcs == dm->atomic_obj.funcs) + return to_dm_atomic_state(new_obj_state); + } + + return NULL; +} + +struct dm_atomic_state * +dm_atomic_get_old_state(struct drm_atomic_state *state) +{ + struct drm_device *dev = state->dev; + struct amdgpu_device *adev = dev->dev_private; + struct amdgpu_display_manager *dm = &adev->dm; + struct drm_private_obj *obj; + struct drm_private_state *old_obj_state; + int i; + + for_each_old_private_obj_in_state(state, obj, old_obj_state, i) { + if (obj->funcs == dm->atomic_obj.funcs) + return to_dm_atomic_state(old_obj_state); + } + + return NULL; +} + +static struct drm_private_state * +dm_atomic_duplicate_state(struct drm_private_obj *obj) +{ + struct dm_atomic_state *old_state, *new_state; + + new_state = kzalloc(sizeof(*new_state), GFP_KERNEL); + if (!new_state) + return NULL; + + __drm_atomic_helper_private_obj_duplicate_state(obj, &new_state->base); + + new_state->context = dc_create_state(); + if (!new_state->context) { + kfree(new_state); + return NULL; + } + + old_state = to_dm_atomic_state(obj->state); + if (old_state && old_state->context) + dc_resource_state_copy_construct(old_state->context, + new_state->context); + + return &new_state->base; +} + +static void dm_atomic_destroy_state(struct drm_private_obj *obj, + struct drm_private_state *state) +{ + struct dm_atomic_state *dm_state = to_dm_atomic_state(state); + + if (dm_state && dm_state->context) + dc_release_state(dm_state->context); + + kfree(dm_state); +} + +static struct drm_private_state_funcs dm_atomic_state_funcs = { + .atomic_duplicate_state = dm_atomic_duplicate_state, + .atomic_destroy_state = dm_atomic_destroy_state, +}; + static int amdgpu_dm_mode_config_init(struct amdgpu_device *adev) { + struct dm_atomic_state *state; int r; adev->mode_info.mode_config_initialized = true; @@ -1513,6 +1708,25 @@ static int amdgpu_dm_mode_config_init(struct amdgpu_device *adev) adev->ddev->mode_config.fb_base = adev->gmc.aper_base; + drm_modeset_lock_init(&adev->dm.atomic_obj_lock); + + state = kzalloc(sizeof(*state), GFP_KERNEL); + if (!state) + return -ENOMEM; + + state->context = dc_create_state(); + if (!state->context) { + kfree(state); + return -ENOMEM; + } + + dc_resource_state_copy_construct_current(adev->dm.dc, state->context); + + drm_atomic_private_obj_init(adev->ddev, + &adev->dm.atomic_obj, + &state->base, + &dm_atomic_state_funcs); + r = amdgpu_display_modeset_create_props(adev); if (r) return r; @@ -1520,15 +1734,63 @@ static int amdgpu_dm_mode_config_init(struct amdgpu_device *adev) return 0; } +#define AMDGPU_DM_DEFAULT_MIN_BACKLIGHT 12 +#define AMDGPU_DM_DEFAULT_MAX_BACKLIGHT 255 + #if defined(CONFIG_BACKLIGHT_CLASS_DEVICE) ||\ defined(CONFIG_BACKLIGHT_CLASS_DEVICE_MODULE) +static void amdgpu_dm_update_backlight_caps(struct amdgpu_display_manager *dm) +{ +#if defined(CONFIG_ACPI) + struct amdgpu_dm_backlight_caps caps; + + if (dm->backlight_caps.caps_valid) + return; + + amdgpu_acpi_get_backlight_caps(dm->adev, &caps); + if (caps.caps_valid) { + dm->backlight_caps.min_input_signal = caps.min_input_signal; + dm->backlight_caps.max_input_signal = caps.max_input_signal; + dm->backlight_caps.caps_valid = true; + } else { + dm->backlight_caps.min_input_signal = + AMDGPU_DM_DEFAULT_MIN_BACKLIGHT; + dm->backlight_caps.max_input_signal = + AMDGPU_DM_DEFAULT_MAX_BACKLIGHT; + } +#else + dm->backlight_caps.min_input_signal = AMDGPU_DM_DEFAULT_MIN_BACKLIGHT; + dm->backlight_caps.max_input_signal = AMDGPU_DM_DEFAULT_MAX_BACKLIGHT; +#endif +} + static int amdgpu_dm_backlight_update_status(struct backlight_device *bd) { struct amdgpu_display_manager *dm = bl_get_data(bd); + struct amdgpu_dm_backlight_caps caps; + uint32_t brightness = bd->props.brightness; + + amdgpu_dm_update_backlight_caps(dm); + caps = dm->backlight_caps; + /* + * The brightness input is in the range 0-255 + * It needs to be rescaled to be between the + * requested min and max input signal + * + * It also needs to be scaled up by 0x101 to + * match the DC interface which has a range of + * 0 to 0xffff + */ + brightness = + brightness + * 0x101 + * (caps.max_input_signal - caps.min_input_signal) + / AMDGPU_MAX_BL_LEVEL + + caps.min_input_signal * 0x101; if (dc_link_set_backlight_level(dm->backlight_link, - bd->props.brightness, 0, 0)) + brightness, 0)) return 0; else return 1; @@ -1555,6 +1817,8 @@ amdgpu_dm_register_backlight_device(struct amdgpu_display_manager *dm) char bl_name[16]; struct backlight_properties props = { 0 }; + amdgpu_dm_update_backlight_caps(dm); + props.max_brightness = AMDGPU_MAX_BL_LEVEL; props.brightness = AMDGPU_MAX_BL_LEVEL; props.type = BACKLIGHT_RAW; @@ -1580,18 +1844,18 @@ static int initialize_plane(struct amdgpu_display_manager *dm, struct amdgpu_mode_info *mode_info, int plane_id) { - struct amdgpu_plane *plane; + struct drm_plane *plane; unsigned long possible_crtcs; int ret = 0; - plane = kzalloc(sizeof(struct amdgpu_plane), GFP_KERNEL); + plane = kzalloc(sizeof(struct drm_plane), GFP_KERNEL); mode_info->planes[plane_id] = plane; if (!plane) { DRM_ERROR("KMS: Failed to allocate plane\n"); return -ENOMEM; } - plane->base.type = mode_info->plane_type[plane_id]; + plane->type = mode_info->plane_type[plane_id]; /* * HACK: IGT tests expect that each plane can only have @@ -1682,7 +1946,7 @@ static int amdgpu_dm_initialize_drm_device(struct amdgpu_device *adev) } for (i = 0; i < dm->dc->caps.max_streams; i++) - if (amdgpu_dm_crtc_init(dm, &mode_info->planes[i]->base, i)) { + if (amdgpu_dm_crtc_init(dm, mode_info->planes[i], i)) { DRM_ERROR("KMS: Failed to initialize crtc\n"); goto fail; } @@ -1786,6 +2050,7 @@ fail: static void amdgpu_dm_destroy_drm_device(struct amdgpu_display_manager *dm) { drm_mode_config_cleanup(dm->ddev); + drm_atomic_private_obj_fini(&dm->atomic_obj); return; } @@ -1805,73 +2070,6 @@ static void dm_bandwidth_update(struct amdgpu_device *adev) /* TODO: implement later */ } -static int amdgpu_notify_freesync(struct drm_device *dev, void *data, - struct drm_file *filp) -{ - struct drm_atomic_state *state; - struct drm_modeset_acquire_ctx ctx; - struct drm_crtc *crtc; - struct drm_connector *connector; - struct drm_connector_state *old_con_state, *new_con_state; - int ret = 0; - uint8_t i; - bool enable = false; - - drm_modeset_acquire_init(&ctx, 0); - - state = drm_atomic_state_alloc(dev); - if (!state) { - ret = -ENOMEM; - goto out; - } - state->acquire_ctx = &ctx; - -retry: - drm_for_each_crtc(crtc, dev) { - ret = drm_atomic_add_affected_connectors(state, crtc); - if (ret) - goto fail; - - /* TODO rework amdgpu_dm_commit_planes so we don't need this */ - ret = drm_atomic_add_affected_planes(state, crtc); - if (ret) - goto fail; - } - - for_each_oldnew_connector_in_state(state, connector, old_con_state, new_con_state, i) { - struct dm_connector_state *dm_new_con_state = to_dm_connector_state(new_con_state); - struct drm_crtc_state *new_crtc_state; - struct amdgpu_crtc *acrtc = to_amdgpu_crtc(dm_new_con_state->base.crtc); - struct dm_crtc_state *dm_new_crtc_state; - - if (!acrtc) { - ASSERT(0); - continue; - } - - new_crtc_state = drm_atomic_get_new_crtc_state(state, &acrtc->base); - dm_new_crtc_state = to_dm_crtc_state(new_crtc_state); - - dm_new_crtc_state->freesync_enabled = enable; - } - - ret = drm_atomic_commit(state); - -fail: - if (ret == -EDEADLK) { - drm_atomic_state_clear(state); - drm_modeset_backoff(&ctx); - goto retry; - } - - drm_atomic_state_put(state); - -out: - drm_modeset_drop_locks(&ctx); - drm_modeset_acquire_fini(&ctx); - return ret; -} - static const struct amdgpu_display_funcs dm_display_funcs = { .bandwidth_update = dm_bandwidth_update, /* called unconditionally */ .vblank_get_counter = dm_vblank_get_counter,/* called unconditionally */ @@ -1884,8 +2082,6 @@ static const struct amdgpu_display_funcs dm_display_funcs = { dm_crtc_get_scanoutpos,/* called unconditionally */ .add_encoder = NULL, /* VBIOS parsing. DAL does it. */ .add_connector = NULL, /* VBIOS parsing. DAL does it. */ - .notify_freesync = amdgpu_notify_freesync, - }; #if defined(CONFIG_DEBUG_KERNEL_DC) @@ -2118,6 +2314,71 @@ static int get_fb_info(const struct amdgpu_framebuffer *amdgpu_fb, return r; } +static inline uint64_t get_dcc_address(uint64_t address, uint64_t tiling_flags) +{ + uint32_t offset = AMDGPU_TILING_GET(tiling_flags, DCC_OFFSET_256B); + + return offset ? (address + offset * 256) : 0; +} + +static bool fill_plane_dcc_attributes(struct amdgpu_device *adev, + const struct amdgpu_framebuffer *afb, + struct dc_plane_state *plane_state, + uint64_t info) +{ + struct dc *dc = adev->dm.dc; + struct dc_dcc_surface_param input; + struct dc_surface_dcc_cap output; + uint32_t offset = AMDGPU_TILING_GET(info, DCC_OFFSET_256B); + uint32_t i64b = AMDGPU_TILING_GET(info, DCC_INDEPENDENT_64B) != 0; + uint64_t dcc_address; + + memset(&input, 0, sizeof(input)); + memset(&output, 0, sizeof(output)); + + if (!offset) + return false; + + if (!dc->cap_funcs.get_dcc_compression_cap) + return false; + + input.format = plane_state->format; + input.surface_size.width = + plane_state->plane_size.grph.surface_size.width; + input.surface_size.height = + plane_state->plane_size.grph.surface_size.height; + input.swizzle_mode = plane_state->tiling_info.gfx9.swizzle; + + if (plane_state->rotation == ROTATION_ANGLE_0 || + plane_state->rotation == ROTATION_ANGLE_180) + input.scan = SCAN_DIRECTION_HORIZONTAL; + else if (plane_state->rotation == ROTATION_ANGLE_90 || + plane_state->rotation == ROTATION_ANGLE_270) + input.scan = SCAN_DIRECTION_VERTICAL; + + if (!dc->cap_funcs.get_dcc_compression_cap(dc, &input, &output)) + return false; + + if (!output.capable) + return false; + + if (i64b == 0 && output.grph.rgb.independent_64b_blks != 0) + return false; + + plane_state->dcc.enable = 1; + plane_state->dcc.grph.meta_pitch = + AMDGPU_TILING_GET(info, DCC_PITCH_MAX) + 1; + plane_state->dcc.grph.independent_64b_blks = i64b; + + dcc_address = get_dcc_address(afb->address, info); + plane_state->address.grph.meta_addr.low_part = + lower_32_bits(dcc_address); + plane_state->address.grph.meta_addr.high_part = + upper_32_bits(dcc_address); + + return true; +} + static int fill_plane_attributes_from_fb(struct amdgpu_device *adev, struct dc_plane_state *plane_state, const struct amdgpu_framebuffer *amdgpu_fb) @@ -2170,6 +2431,10 @@ static int fill_plane_attributes_from_fb(struct amdgpu_device *adev, return -EINVAL; } + memset(&plane_state->address, 0, sizeof(plane_state->address)); + memset(&plane_state->tiling_info, 0, sizeof(plane_state->tiling_info)); + memset(&plane_state->dcc, 0, sizeof(plane_state->dcc)); + if (plane_state->format < SURFACE_PIXEL_FORMAT_VIDEO_BEGIN) { plane_state->address.type = PLN_ADDR_TYPE_GRAPHICS; plane_state->plane_size.grph.surface_size.x = 0; @@ -2201,8 +2466,6 @@ static int fill_plane_attributes_from_fb(struct amdgpu_device *adev, plane_state->color_space = COLOR_SPACE_YCBCR709; } - memset(&plane_state->tiling_info, 0, sizeof(plane_state->tiling_info)); - /* Fill GFX8 params */ if (AMDGPU_TILING_GET(tiling_flags, ARRAY_MODE) == DC_ARRAY_2D_TILED_THIN1) { unsigned int bankw, bankh, mtaspect, tile_split, num_banks; @@ -2251,6 +2514,9 @@ static int fill_plane_attributes_from_fb(struct amdgpu_device *adev, plane_state->tiling_info.gfx9.swizzle = AMDGPU_TILING_GET(tiling_flags, SWIZZLE_MODE); plane_state->tiling_info.gfx9.shaderEnable = 1; + + fill_plane_dcc_attributes(adev, amdgpu_fb, plane_state, + tiling_flags); } plane_state->visible = true; @@ -2414,7 +2680,7 @@ get_output_color_space(const struct dc_crtc_timing *dc_crtc_timing) * according to HDMI spec, we use YCbCr709 and YCbCr601 * respectively */ - if (dc_crtc_timing->pix_clk_khz > 27030) { + if (dc_crtc_timing->pix_clk_100hz > 270300) { if (dc_crtc_timing->flags.Y_ONLY) color_space = COLOR_SPACE_YCBCR709_LIMITED; @@ -2457,7 +2723,7 @@ static void adjust_colour_depth_from_display_info(struct dc_crtc_timing *timing_ if (timing_out->display_color_depth <= COLOR_DEPTH_888) return; do { - normalized_clk = timing_out->pix_clk_khz; + normalized_clk = timing_out->pix_clk_100hz / 10; /* YCbCr 4:2:0 requires additional adjustment of 1/2 */ if (timing_out->pixel_encoding == PIXEL_ENCODING_YCBCR420) normalized_clk /= 2; @@ -2486,7 +2752,8 @@ static void adjust_colour_depth_from_display_info(struct dc_crtc_timing *timing_ static void fill_stream_properties_from_drm_display_mode(struct dc_stream_state *stream, const struct drm_display_mode *mode_in, - const struct drm_connector *connector) + const struct drm_connector *connector, + const struct dc_stream_state *old_stream) { struct dc_crtc_timing *timing_out = &stream->timing; const struct drm_display_info *info = &connector->display_info; @@ -2499,10 +2766,10 @@ fill_stream_properties_from_drm_display_mode(struct dc_stream_state *stream, timing_out->v_border_bottom = 0; /* TODO: un-hardcode */ if (drm_mode_is_420_only(info, mode_in) - && stream->sink->sink_signal == SIGNAL_TYPE_HDMI_TYPE_A) + && stream->signal == SIGNAL_TYPE_HDMI_TYPE_A) timing_out->pixel_encoding = PIXEL_ENCODING_YCBCR420; else if ((connector->display_info.color_formats & DRM_COLOR_FORMAT_YCRCB444) - && stream->sink->sink_signal == SIGNAL_TYPE_HDMI_TYPE_A) + && stream->signal == SIGNAL_TYPE_HDMI_TYPE_A) timing_out->pixel_encoding = PIXEL_ENCODING_YCBCR444; else timing_out->pixel_encoding = PIXEL_ENCODING_RGB; @@ -2512,7 +2779,18 @@ fill_stream_properties_from_drm_display_mode(struct dc_stream_state *stream, connector); timing_out->scan_type = SCANNING_TYPE_NODATA; timing_out->hdmi_vic = 0; - timing_out->vic = drm_match_cea_mode(mode_in); + + if(old_stream) { + timing_out->vic = old_stream->timing.vic; + timing_out->flags.HSYNC_POSITIVE_POLARITY = old_stream->timing.flags.HSYNC_POSITIVE_POLARITY; + timing_out->flags.VSYNC_POSITIVE_POLARITY = old_stream->timing.flags.VSYNC_POSITIVE_POLARITY; + } else { + timing_out->vic = drm_match_cea_mode(mode_in); + if (mode_in->flags & DRM_MODE_FLAG_PHSYNC) + timing_out->flags.HSYNC_POSITIVE_POLARITY = 1; + if (mode_in->flags & DRM_MODE_FLAG_PVSYNC) + timing_out->flags.VSYNC_POSITIVE_POLARITY = 1; + } timing_out->h_addressable = mode_in->crtc_hdisplay; timing_out->h_total = mode_in->crtc_htotal; @@ -2526,18 +2804,14 @@ fill_stream_properties_from_drm_display_mode(struct dc_stream_state *stream, mode_in->crtc_vsync_start - mode_in->crtc_vdisplay; timing_out->v_sync_width = mode_in->crtc_vsync_end - mode_in->crtc_vsync_start; - timing_out->pix_clk_khz = mode_in->crtc_clock; + timing_out->pix_clk_100hz = mode_in->crtc_clock * 10; timing_out->aspect_ratio = get_aspect_ratio(mode_in); - if (mode_in->flags & DRM_MODE_FLAG_PHSYNC) - timing_out->flags.HSYNC_POSITIVE_POLARITY = 1; - if (mode_in->flags & DRM_MODE_FLAG_PVSYNC) - timing_out->flags.VSYNC_POSITIVE_POLARITY = 1; stream->output_color_space = get_output_color_space(timing_out); stream->out_transfer_func->type = TF_TYPE_PREDEFINED; stream->out_transfer_func->tf = TRANSFER_FUNCTION_SRGB; - if (stream->sink->sink_signal == SIGNAL_TYPE_HDMI_TYPE_A) + if (stream->signal == SIGNAL_TYPE_HDMI_TYPE_A) adjust_colour_depth_from_display_info(timing_out, info); } @@ -2658,7 +2932,7 @@ static void set_master_stream(struct dc_stream_state *stream_set[], if (stream_set[j] && stream_set[j]->triggered_crtc_reset.enabled) { int refresh_rate = 0; - refresh_rate = (stream_set[j]->timing.pix_clk_khz*1000)/ + refresh_rate = (stream_set[j]->timing.pix_clk_100hz*100)/ (stream_set[j]->timing.h_total*stream_set[j]->timing.v_total); if (refresh_rate > highest_rfr) { highest_rfr = refresh_rate; @@ -2694,13 +2968,18 @@ static void dm_enable_per_frame_crtc_master_sync(struct dc_state *context) static struct dc_stream_state * create_stream_for_sink(struct amdgpu_dm_connector *aconnector, const struct drm_display_mode *drm_mode, - const struct dm_connector_state *dm_state) + const struct dm_connector_state *dm_state, + const struct dc_stream_state *old_stream) { struct drm_display_mode *preferred_mode = NULL; struct drm_connector *drm_connector; struct dc_stream_state *stream = NULL; struct drm_display_mode mode = *drm_mode; bool native_mode_found = false; + bool scale = dm_state ? (dm_state->scaling != RMX_OFF) : false; + int mode_refresh; + int preferred_refresh = 0; + struct dc_sink *sink = NULL; if (aconnector == NULL) { DRM_ERROR("aconnector is NULL!\n"); @@ -2710,13 +2989,12 @@ create_stream_for_sink(struct amdgpu_dm_connector *aconnector, drm_connector = &aconnector->base; if (!aconnector->dc_sink) { - if (!aconnector->mst_port) { - sink = create_fake_sink(aconnector); - if (!sink) - return stream; - } + sink = create_fake_sink(aconnector); + if (!sink) + return stream; } else { sink = aconnector->dc_sink; + dc_sink_retain(sink); } stream = dc_create_stream_for_sink(sink); @@ -2726,6 +3004,8 @@ create_stream_for_sink(struct amdgpu_dm_connector *aconnector, goto finish; } + stream->dm_stream_context = aconnector; + list_for_each_entry(preferred_mode, &aconnector->base.modes, head) { /* Search for preferred mode */ if (preferred_mode->type & DRM_MODE_TYPE_PREFERRED) { @@ -2739,6 +3019,8 @@ create_stream_for_sink(struct amdgpu_dm_connector *aconnector, struct drm_display_mode, head); + mode_refresh = drm_mode_vrefresh(&mode); + if (preferred_mode == NULL) { /* * This may not be an error, the use case is when we have no @@ -2751,13 +3033,23 @@ create_stream_for_sink(struct amdgpu_dm_connector *aconnector, decide_crtc_timing_for_drm_display_mode( &mode, preferred_mode, dm_state ? (dm_state->scaling != RMX_OFF) : false); + preferred_refresh = drm_mode_vrefresh(preferred_mode); } if (!dm_state) drm_mode_set_crtcinfo(&mode, 0); - fill_stream_properties_from_drm_display_mode(stream, - &mode, &aconnector->base); + /* + * If scaling is enabled and refresh rate didn't change + * we copy the vic and polarities of the old timings + */ + if (!scale || mode_refresh != preferred_refresh) + fill_stream_properties_from_drm_display_mode(stream, + &mode, &aconnector->base, NULL); + else + fill_stream_properties_from_drm_display_mode(stream, + &mode, &aconnector->base, old_stream); + update_stream_scaling_settings(&mode, dm_state, stream); fill_audio_info( @@ -2765,13 +3057,10 @@ create_stream_for_sink(struct amdgpu_dm_connector *aconnector, drm_connector, sink); - update_stream_signal(stream); + update_stream_signal(stream, sink); - if (dm_state && dm_state->freesync_capable) - stream->ignore_msa_timing_param = true; finish: - if (sink && sink->sink_signal == SIGNAL_TYPE_VIRTUAL && aconnector->base.force != DRM_FORCE_ON) - dc_sink_release(sink); + dc_sink_release(sink); return stream; } @@ -2835,9 +3124,12 @@ dm_crtc_duplicate_state(struct drm_crtc *crtc) dc_stream_retain(state->stream); } - state->adjust = cur->adjust; + state->vrr_params = cur->vrr_params; state->vrr_infopacket = cur->vrr_infopacket; - state->freesync_enabled = cur->freesync_enabled; + state->abm_level = cur->abm_level; + state->vrr_supported = cur->vrr_supported; + state->freesync_config = cur->freesync_config; + state->crc_enabled = cur->crc_enabled; /* TODO Duplicate dc_stream after objects are stream object is flattened */ @@ -2953,6 +3245,9 @@ int amdgpu_dm_connector_atomic_set_property(struct drm_connector *connector, } else if (property == adev->mode_info.max_bpc_property) { dm_new_state->max_bpc = val; ret = 0; + } else if (property == adev->mode_info.abm_level_property) { + dm_new_state->abm_level = val; + ret = 0; } return ret; @@ -2998,7 +3293,11 @@ int amdgpu_dm_connector_atomic_get_property(struct drm_connector *connector, } else if (property == adev->mode_info.max_bpc_property) { *val = dm_state->max_bpc; ret = 0; + } else if (property == adev->mode_info.abm_level_property) { + *val = dm_state->abm_level; + ret = 0; } + return ret; } @@ -3019,6 +3318,14 @@ static void amdgpu_dm_connector_destroy(struct drm_connector *connector) dm->backlight_dev = NULL; } #endif + + if (aconnector->dc_em_sink) + dc_sink_release(aconnector->dc_em_sink); + aconnector->dc_em_sink = NULL; + if (aconnector->dc_sink) + dc_sink_release(aconnector->dc_sink); + aconnector->dc_sink = NULL; + drm_dp_cec_unregister_connector(&aconnector->dm_dp_aux.aux); drm_connector_unregister(connector); drm_connector_cleanup(connector); @@ -3063,7 +3370,11 @@ amdgpu_dm_connector_atomic_duplicate_state(struct drm_connector *connector) __drm_atomic_helper_connector_duplicate_state(connector, &new_state->base); new_state->freesync_capable = state->freesync_capable; - new_state->freesync_enable = state->freesync_enable; + new_state->abm_level = state->abm_level; + new_state->scaling = state->scaling; + new_state->underscan_enable = state->underscan_enable; + new_state->underscan_hborder = state->underscan_hborder; + new_state->underscan_vborder = state->underscan_vborder; new_state->max_bpc = state->max_bpc; return &new_state->base; @@ -3112,10 +3423,12 @@ static void create_eml_sink(struct amdgpu_dm_connector *aconnector) (edid->extensions + 1) * EDID_LENGTH, &init_params); - if (aconnector->base.force == DRM_FORCE_ON) + if (aconnector->base.force == DRM_FORCE_ON) { aconnector->dc_sink = aconnector->dc_link->local_sink ? aconnector->dc_link->local_sink : aconnector->dc_em_sink; + dc_sink_retain(aconnector->dc_sink); + } } static void handle_edid_mgmt(struct amdgpu_dm_connector *aconnector) @@ -3166,7 +3479,7 @@ enum drm_mode_status amdgpu_dm_connector_mode_valid(struct drm_connector *connec goto fail; } - stream = create_stream_for_sink(aconnector, mode, NULL); + stream = create_stream_for_sink(aconnector, mode, NULL, NULL); if (stream == NULL) { DRM_ERROR("Failed to create stream for sink!\n"); goto fail; @@ -3200,7 +3513,6 @@ amdgpu_dm_connector_helper_funcs = { */ .get_modes = get_modes, .mode_valid = amdgpu_dm_connector_mode_valid, - .best_encoder = drm_atomic_helper_best_encoder }; static void dm_crtc_helper_disable(struct drm_crtc *crtc) @@ -3327,6 +3639,7 @@ static int dm_plane_helper_prepare_fb(struct drm_plane *plane, struct amdgpu_bo *rbo; uint64_t chroma_addr = 0; struct dm_plane_state *dm_plane_state_new, *dm_plane_state_old; + uint64_t tiling_flags, dcc_address; unsigned int awidth; uint32_t domain; int r; @@ -3367,6 +3680,9 @@ static int dm_plane_helper_prepare_fb(struct drm_plane *plane, DRM_ERROR("%p bind failed\n", rbo); return r; } + + amdgpu_bo_get_tiling_flags(rbo, &tiling_flags); + amdgpu_bo_unreserve(rbo); afb->address = amdgpu_bo_gpu_offset(rbo); @@ -3380,6 +3696,13 @@ static int dm_plane_helper_prepare_fb(struct drm_plane *plane, if (plane_state->format < SURFACE_PIXEL_FORMAT_VIDEO_BEGIN) { plane_state->address.grph.addr.low_part = lower_32_bits(afb->address); plane_state->address.grph.addr.high_part = upper_32_bits(afb->address); + + dcc_address = + get_dcc_address(afb->address, tiling_flags); + plane_state->address.grph.meta_addr.low_part = + lower_32_bits(dcc_address); + plane_state->address.grph.meta_addr.high_part = + upper_32_bits(dcc_address); } else { awidth = ALIGN(new_state->fb->width, 64); plane_state->address.type = PLN_ADDR_TYPE_VIDEO_PROGRESSIVE; @@ -3438,10 +3761,53 @@ static int dm_plane_atomic_check(struct drm_plane *plane, return -EINVAL; } +static int dm_plane_atomic_async_check(struct drm_plane *plane, + struct drm_plane_state *new_plane_state) +{ + struct drm_plane_state *old_plane_state = + drm_atomic_get_old_plane_state(new_plane_state->state, plane); + + /* Only support async updates on cursor planes. */ + if (plane->type != DRM_PLANE_TYPE_CURSOR) + return -EINVAL; + + /* + * DRM calls prepare_fb and cleanup_fb on new_plane_state for + * async commits so don't allow fb changes. + */ + if (old_plane_state->fb != new_plane_state->fb) + return -EINVAL; + + return 0; +} + +static void dm_plane_atomic_async_update(struct drm_plane *plane, + struct drm_plane_state *new_state) +{ + struct drm_plane_state *old_state = + drm_atomic_get_old_plane_state(new_state->state, plane); + + if (plane->state->fb != new_state->fb) + drm_atomic_set_fb_for_plane(plane->state, new_state->fb); + + plane->state->src_x = new_state->src_x; + plane->state->src_y = new_state->src_y; + plane->state->src_w = new_state->src_w; + plane->state->src_h = new_state->src_h; + plane->state->crtc_x = new_state->crtc_x; + plane->state->crtc_y = new_state->crtc_y; + plane->state->crtc_w = new_state->crtc_w; + plane->state->crtc_h = new_state->crtc_h; + + handle_cursor_update(plane, old_state); +} + static const struct drm_plane_helper_funcs dm_plane_helper_funcs = { .prepare_fb = dm_plane_helper_prepare_fb, .cleanup_fb = dm_plane_helper_cleanup_fb, .atomic_check = dm_plane_atomic_check, + .atomic_async_check = dm_plane_atomic_async_check, + .atomic_async_update = dm_plane_atomic_async_update }; /* @@ -3451,7 +3817,6 @@ static const struct drm_plane_helper_funcs dm_plane_helper_funcs = { * check will succeed, and let DC implement proper check */ static const uint32_t rgb_formats[] = { - DRM_FORMAT_RGB888, DRM_FORMAT_XRGB8888, DRM_FORMAT_ARGB8888, DRM_FORMAT_RGBA8888, @@ -3473,49 +3838,49 @@ static const u32 cursor_formats[] = { }; static int amdgpu_dm_plane_init(struct amdgpu_display_manager *dm, - struct amdgpu_plane *aplane, + struct drm_plane *plane, unsigned long possible_crtcs) { int res = -EPERM; - switch (aplane->base.type) { + switch (plane->type) { case DRM_PLANE_TYPE_PRIMARY: res = drm_universal_plane_init( dm->adev->ddev, - &aplane->base, + plane, possible_crtcs, &dm_plane_funcs, rgb_formats, ARRAY_SIZE(rgb_formats), - NULL, aplane->base.type, NULL); + NULL, plane->type, NULL); break; case DRM_PLANE_TYPE_OVERLAY: res = drm_universal_plane_init( dm->adev->ddev, - &aplane->base, + plane, possible_crtcs, &dm_plane_funcs, yuv_formats, ARRAY_SIZE(yuv_formats), - NULL, aplane->base.type, NULL); + NULL, plane->type, NULL); break; case DRM_PLANE_TYPE_CURSOR: res = drm_universal_plane_init( dm->adev->ddev, - &aplane->base, + plane, possible_crtcs, &dm_plane_funcs, cursor_formats, ARRAY_SIZE(cursor_formats), - NULL, aplane->base.type, NULL); + NULL, plane->type, NULL); break; } - drm_plane_helper_add(&aplane->base, &dm_plane_helper_funcs); + drm_plane_helper_add(plane, &dm_plane_helper_funcs); /* Create (reset) the plane state */ - if (aplane->base.funcs->reset) - aplane->base.funcs->reset(&aplane->base); + if (plane->funcs->reset) + plane->funcs->reset(plane); return res; @@ -3526,7 +3891,7 @@ static int amdgpu_dm_crtc_init(struct amdgpu_display_manager *dm, uint32_t crtc_index) { struct amdgpu_crtc *acrtc = NULL; - struct amdgpu_plane *cursor_plane; + struct drm_plane *cursor_plane; int res = -ENOMEM; @@ -3534,7 +3899,7 @@ static int amdgpu_dm_crtc_init(struct amdgpu_display_manager *dm, if (!cursor_plane) goto fail; - cursor_plane->base.type = DRM_PLANE_TYPE_CURSOR; + cursor_plane->type = DRM_PLANE_TYPE_CURSOR; res = amdgpu_dm_plane_init(dm, cursor_plane, 0); acrtc = kzalloc(sizeof(struct amdgpu_crtc), GFP_KERNEL); @@ -3545,7 +3910,7 @@ static int amdgpu_dm_crtc_init(struct amdgpu_display_manager *dm, dm->ddev, &acrtc->base, plane, - &cursor_plane->base, + cursor_plane, &amdgpu_dm_crtc_funcs, NULL); if (res) @@ -3603,14 +3968,17 @@ static int to_drm_connector_type(enum signal_type st) } } +static struct drm_encoder *amdgpu_dm_connector_to_encoder(struct drm_connector *connector) +{ + return drm_encoder_find(connector->dev, NULL, connector->encoder_ids[0]); +} + static void amdgpu_dm_get_native_mode(struct drm_connector *connector) { - const struct drm_connector_helper_funcs *helper = - connector->helper_private; struct drm_encoder *encoder; struct amdgpu_encoder *amdgpu_encoder; - encoder = helper->best_encoder(connector); + encoder = amdgpu_dm_connector_to_encoder(connector); if (encoder == NULL) return; @@ -3737,14 +4105,12 @@ static void amdgpu_dm_connector_ddc_get_modes(struct drm_connector *connector, static int amdgpu_dm_connector_get_modes(struct drm_connector *connector) { - const struct drm_connector_helper_funcs *helper = - connector->helper_private; struct amdgpu_dm_connector *amdgpu_dm_connector = to_amdgpu_dm_connector(connector); struct drm_encoder *encoder; struct edid *edid = amdgpu_dm_connector->edid; - encoder = helper->best_encoder(connector); + encoder = amdgpu_dm_connector_to_encoder(connector); if (!edid || !drm_edid_is_valid(edid)) { amdgpu_dm_connector->num_modes = @@ -3783,12 +4149,12 @@ void amdgpu_dm_connector_init_helper(struct amdgpu_display_manager *dm, case DRM_MODE_CONNECTOR_HDMIA: aconnector->base.polled = DRM_CONNECTOR_POLL_HPD; aconnector->base.ycbcr_420_allowed = - link->link_enc->features.ycbcr420_supported ? true : false; + link->link_enc->features.hdmi_ycbcr420_supported ? true : false; break; case DRM_MODE_CONNECTOR_DisplayPort: aconnector->base.polled = DRM_CONNECTOR_POLL_HPD; aconnector->base.ycbcr_420_allowed = - link->link_enc->features.ycbcr420_supported ? true : false; + link->link_enc->features.dp_ycbcr420_supported ? true : false; break; case DRM_MODE_CONNECTOR_DVID: aconnector->base.polled = DRM_CONNECTOR_POLL_HPD; @@ -3814,6 +4180,18 @@ void amdgpu_dm_connector_init_helper(struct amdgpu_display_manager *dm, adev->mode_info.max_bpc_property, 0); + if (connector_type == DRM_MODE_CONNECTOR_eDP && + dc_is_dmcu_initialized(adev->dm.dc)) { + drm_object_attach_property(&aconnector->base.base, + adev->mode_info.abm_level_property, 0); + } + + if (connector_type == DRM_MODE_CONNECTOR_HDMIA || + connector_type == DRM_MODE_CONNECTOR_DisplayPort || + connector_type == DRM_MODE_CONNECTOR_eDP) { + drm_connector_attach_vrr_capable_property( + &aconnector->base); + } } static int amdgpu_dm_i2c_xfer(struct i2c_adapter *i2c_adap, @@ -4118,6 +4496,7 @@ static int get_cursor_position(struct drm_plane *plane, struct drm_crtc *crtc, static void handle_cursor_update(struct drm_plane *plane, struct drm_plane_state *old_plane_state) { + struct amdgpu_device *adev = plane->dev->dev_private; struct amdgpu_framebuffer *afb = to_amdgpu_framebuffer(plane->state->fb); struct drm_crtc *crtc = afb ? plane->state->crtc : old_plane_state->crtc; struct dm_crtc_state *crtc_state = crtc ? to_dm_crtc_state(crtc->state) : NULL; @@ -4142,15 +4521,19 @@ static void handle_cursor_update(struct drm_plane *plane, if (!position.enable) { /* turn off cursor */ - if (crtc_state && crtc_state->stream) + if (crtc_state && crtc_state->stream) { + mutex_lock(&adev->dm.dc_lock); dc_stream_set_cursor_position(crtc_state->stream, &position); + mutex_unlock(&adev->dm.dc_lock); + } return; } amdgpu_crtc->cursor_width = plane->state->crtc_w; amdgpu_crtc->cursor_height = plane->state->crtc_h; + memset(&attributes, 0, sizeof(attributes)); attributes.address.high_part = upper_32_bits(address); attributes.address.low_part = lower_32_bits(address); attributes.width = plane->state->crtc_w; @@ -4162,6 +4545,7 @@ static void handle_cursor_update(struct drm_plane *plane, attributes.pitch = attributes.width; if (crtc_state->stream) { + mutex_lock(&adev->dm.dc_lock); if (!dc_stream_set_cursor_attributes(crtc_state->stream, &attributes)) DRM_ERROR("DC failed to set cursor attributes\n"); @@ -4169,6 +4553,7 @@ static void handle_cursor_update(struct drm_plane *plane, if (!dc_stream_set_cursor_position(crtc_state->stream, &position)) DRM_ERROR("DC failed to set cursor position\n"); + mutex_unlock(&adev->dm.dc_lock); } } @@ -4190,241 +4575,141 @@ static void prepare_flip_isr(struct amdgpu_crtc *acrtc) acrtc->crtc_id); } -/* - * Executes flip - * - * Waits on all BO's fences and for proper vblank count - */ -static void amdgpu_dm_do_flip(struct drm_crtc *crtc, - struct drm_framebuffer *fb, - uint32_t target, - struct dc_state *state) +static void update_freesync_state_on_stream( + struct amdgpu_display_manager *dm, + struct dm_crtc_state *new_crtc_state, + struct dc_stream_state *new_stream, + struct dc_plane_state *surface, + u32 flip_timestamp_in_us) { - unsigned long flags; - uint32_t target_vblank; - int r, vpos, hpos; - struct amdgpu_crtc *acrtc = to_amdgpu_crtc(crtc); - struct amdgpu_framebuffer *afb = to_amdgpu_framebuffer(fb); - struct amdgpu_bo *abo = gem_to_amdgpu_bo(fb->obj[0]); - struct amdgpu_device *adev = crtc->dev->dev_private; - bool async_flip = (crtc->state->pageflip_flags & DRM_MODE_PAGE_FLIP_ASYNC) != 0; - struct dc_flip_addrs addr = { {0} }; - /* TODO eliminate or rename surface_update */ - struct dc_surface_update surface_updates[1] = { {0} }; - struct dm_crtc_state *acrtc_state = to_dm_crtc_state(crtc->state); - struct dc_stream_status *stream_status; - - - /* Prepare wait for target vblank early - before the fence-waits */ - target_vblank = target - (uint32_t)drm_crtc_vblank_count(crtc) + - amdgpu_get_vblank_counter_kms(crtc->dev, acrtc->crtc_id); - - /* - * TODO This might fail and hence better not used, wait - * explicitly on fences instead - * and in general should be called for - * blocking commit to as per framework helpers - */ - r = amdgpu_bo_reserve(abo, true); - if (unlikely(r != 0)) { - DRM_ERROR("failed to reserve buffer before flip\n"); - WARN_ON(1); - } - - /* Wait for all fences on this FB */ - WARN_ON(reservation_object_wait_timeout_rcu(abo->tbo.resv, true, false, - MAX_SCHEDULE_TIMEOUT) < 0); + struct mod_vrr_params vrr_params = new_crtc_state->vrr_params; + struct dc_info_packet vrr_infopacket = {0}; + struct mod_freesync_config config = new_crtc_state->freesync_config; - amdgpu_bo_unreserve(abo); + if (!new_stream) + return; /* - * Wait until we're out of the vertical blank period before the one - * targeted by the flip + * TODO: Determine why min/max totals and vrefresh can be 0 here. + * For now it's sufficient to just guard against these conditions. */ - while ((acrtc->enabled && - (amdgpu_display_get_crtc_scanoutpos(adev->ddev, acrtc->crtc_id, - 0, &vpos, &hpos, NULL, - NULL, &crtc->hwmode) - & (DRM_SCANOUTPOS_VALID | DRM_SCANOUTPOS_IN_VBLANK)) == - (DRM_SCANOUTPOS_VALID | DRM_SCANOUTPOS_IN_VBLANK) && - (int)(target_vblank - - amdgpu_get_vblank_counter_kms(adev->ddev, acrtc->crtc_id)) > 0)) { - usleep_range(1000, 1100); - } - - /* Flip */ - spin_lock_irqsave(&crtc->dev->event_lock, flags); - - WARN_ON(acrtc->pflip_status != AMDGPU_FLIP_NONE); - WARN_ON(!acrtc_state->stream); - - addr.address.grph.addr.low_part = lower_32_bits(afb->address); - addr.address.grph.addr.high_part = upper_32_bits(afb->address); - addr.flip_immediate = async_flip; - - - if (acrtc->base.state->event) - prepare_flip_isr(acrtc); - - spin_unlock_irqrestore(&crtc->dev->event_lock, flags); - stream_status = dc_stream_get_status(acrtc_state->stream); - if (!stream_status) { - DRM_ERROR("No stream status for CRTC: id=%d\n", - acrtc->crtc_id); - return; - } - - surface_updates->surface = stream_status->plane_states[0]; - if (!surface_updates->surface) { - DRM_ERROR("No surface for CRTC: id=%d\n", - acrtc->crtc_id); + if (!new_stream->timing.h_total || !new_stream->timing.v_total) return; - } - surface_updates->flip_addr = &addr; - - dc_commit_updates_for_stream(adev->dm.dc, - surface_updates, - 1, - acrtc_state->stream, - NULL, - &surface_updates->surface, - state); - - DRM_DEBUG_DRIVER("%s Flipping to hi: 0x%x, low: 0x%x \n", - __func__, - addr.address.grph.addr.high_part, - addr.address.grph.addr.low_part); -} - -/* - * TODO this whole function needs to go - * - * dc_surface_update is needlessly complex. See if we can just replace this - * with a dc_plane_state and follow the atomic model a bit more closely here. - */ -static bool commit_planes_to_stream( - struct dc *dc, - struct dc_plane_state **plane_states, - uint8_t new_plane_count, - struct dm_crtc_state *dm_new_crtc_state, - struct dm_crtc_state *dm_old_crtc_state, - struct dc_state *state) -{ - /* no need to dynamically allocate this. it's pretty small */ - struct dc_surface_update updates[MAX_SURFACES]; - struct dc_flip_addrs *flip_addr; - struct dc_plane_info *plane_info; - struct dc_scaling_info *scaling_info; - int i; - struct dc_stream_state *dc_stream = dm_new_crtc_state->stream; - struct dc_stream_update *stream_update = - kzalloc(sizeof(struct dc_stream_update), GFP_KERNEL); - if (!stream_update) { - BREAK_TO_DEBUGGER(); - return false; + if (new_crtc_state->vrr_supported && + config.min_refresh_in_uhz && + config.max_refresh_in_uhz) { + config.state = new_crtc_state->base.vrr_enabled ? + VRR_STATE_ACTIVE_VARIABLE : + VRR_STATE_INACTIVE; + } else { + config.state = VRR_STATE_UNSUPPORTED; } - flip_addr = kcalloc(MAX_SURFACES, sizeof(struct dc_flip_addrs), - GFP_KERNEL); - plane_info = kcalloc(MAX_SURFACES, sizeof(struct dc_plane_info), - GFP_KERNEL); - scaling_info = kcalloc(MAX_SURFACES, sizeof(struct dc_scaling_info), - GFP_KERNEL); + mod_freesync_build_vrr_params(dm->freesync_module, + new_stream, + &config, &vrr_params); + + if (surface) { + mod_freesync_handle_preflip( + dm->freesync_module, + surface, + new_stream, + flip_timestamp_in_us, + &vrr_params); + } + + mod_freesync_build_vrr_infopacket( + dm->freesync_module, + new_stream, + &vrr_params, + PACKET_TYPE_VRR, + TRANSFER_FUNC_UNKNOWN, + &vrr_infopacket); + + new_crtc_state->freesync_timing_changed |= + (memcmp(&new_crtc_state->vrr_params.adjust, + &vrr_params.adjust, + sizeof(vrr_params.adjust)) != 0); + + new_crtc_state->freesync_vrr_info_changed |= + (memcmp(&new_crtc_state->vrr_infopacket, + &vrr_infopacket, + sizeof(vrr_infopacket)) != 0); + + new_crtc_state->vrr_params = vrr_params; + new_crtc_state->vrr_infopacket = vrr_infopacket; - if (!flip_addr || !plane_info || !scaling_info) { - kfree(flip_addr); - kfree(plane_info); - kfree(scaling_info); - kfree(stream_update); - return false; - } + new_stream->adjust = new_crtc_state->vrr_params.adjust; + new_stream->vrr_infopacket = vrr_infopacket; - memset(updates, 0, sizeof(updates)); - - stream_update->src = dc_stream->src; - stream_update->dst = dc_stream->dst; - stream_update->out_transfer_func = dc_stream->out_transfer_func; - - if (dm_new_crtc_state->freesync_enabled != dm_old_crtc_state->freesync_enabled) { - stream_update->vrr_infopacket = &dc_stream->vrr_infopacket; - stream_update->adjust = &dc_stream->adjust; - } - - for (i = 0; i < new_plane_count; i++) { - updates[i].surface = plane_states[i]; - updates[i].gamma = - (struct dc_gamma *)plane_states[i]->gamma_correction; - updates[i].in_transfer_func = plane_states[i]->in_transfer_func; - flip_addr[i].address = plane_states[i]->address; - flip_addr[i].flip_immediate = plane_states[i]->flip_immediate; - plane_info[i].color_space = plane_states[i]->color_space; - plane_info[i].format = plane_states[i]->format; - plane_info[i].plane_size = plane_states[i]->plane_size; - plane_info[i].rotation = plane_states[i]->rotation; - plane_info[i].horizontal_mirror = plane_states[i]->horizontal_mirror; - plane_info[i].stereo_format = plane_states[i]->stereo_format; - plane_info[i].tiling_info = plane_states[i]->tiling_info; - plane_info[i].visible = plane_states[i]->visible; - plane_info[i].per_pixel_alpha = plane_states[i]->per_pixel_alpha; - plane_info[i].dcc = plane_states[i]->dcc; - scaling_info[i].scaling_quality = plane_states[i]->scaling_quality; - scaling_info[i].src_rect = plane_states[i]->src_rect; - scaling_info[i].dst_rect = plane_states[i]->dst_rect; - scaling_info[i].clip_rect = plane_states[i]->clip_rect; - - updates[i].flip_addr = &flip_addr[i]; - updates[i].plane_info = &plane_info[i]; - updates[i].scaling_info = &scaling_info[i]; - } - - dc_commit_updates_for_stream( - dc, - updates, - new_plane_count, - dc_stream, stream_update, plane_states, state); - - kfree(flip_addr); - kfree(plane_info); - kfree(scaling_info); - kfree(stream_update); - return true; + if (new_crtc_state->freesync_vrr_info_changed) + DRM_DEBUG_KMS("VRR packet update: crtc=%u enabled=%d state=%d", + new_crtc_state->base.crtc->base.id, + (int)new_crtc_state->base.vrr_enabled, + (int)vrr_params.state); } static void amdgpu_dm_commit_planes(struct drm_atomic_state *state, + struct dc_state *dc_state, struct drm_device *dev, struct amdgpu_display_manager *dm, struct drm_crtc *pcrtc, bool *wait_for_vblank) { - uint32_t i; + uint32_t i, r; + uint64_t timestamp_ns; struct drm_plane *plane; struct drm_plane_state *old_plane_state, *new_plane_state; - struct dc_stream_state *dc_stream_attach; - struct dc_plane_state *plane_states_constructed[MAX_SURFACES]; struct amdgpu_crtc *acrtc_attach = to_amdgpu_crtc(pcrtc); struct drm_crtc_state *new_pcrtc_state = drm_atomic_get_new_crtc_state(state, pcrtc); struct dm_crtc_state *acrtc_state = to_dm_crtc_state(new_pcrtc_state); struct dm_crtc_state *dm_old_crtc_state = to_dm_crtc_state(drm_atomic_get_old_crtc_state(state, pcrtc)); - struct dm_atomic_state *dm_state = to_dm_atomic_state(state); - int planes_count = 0; + int flip_count = 0, planes_count = 0, vpos, hpos; unsigned long flags; + struct amdgpu_bo *abo; + uint64_t tiling_flags, dcc_address; + uint32_t target, target_vblank; + uint64_t last_flip_vblank; + bool vrr_active = acrtc_state->freesync_config.state == VRR_STATE_ACTIVE_VARIABLE; + + struct { + struct dc_surface_update surface_updates[MAX_SURFACES]; + struct dc_flip_addrs flip_addrs[MAX_SURFACES]; + struct dc_stream_update stream_update; + } *flip; + + struct { + struct dc_surface_update surface_updates[MAX_SURFACES]; + struct dc_plane_info plane_infos[MAX_SURFACES]; + struct dc_scaling_info scaling_infos[MAX_SURFACES]; + struct dc_stream_update stream_update; + } *full; + + flip = kzalloc(sizeof(*flip), GFP_KERNEL); + full = kzalloc(sizeof(*full), GFP_KERNEL); + + if (!flip || !full) { + dm_error("Failed to allocate update bundles\n"); + goto cleanup; + } /* update planes when needed */ for_each_oldnew_plane_in_state(state, plane, old_plane_state, new_plane_state, i) { struct drm_crtc *crtc = new_plane_state->crtc; struct drm_crtc_state *new_crtc_state; struct drm_framebuffer *fb = new_plane_state->fb; + struct amdgpu_framebuffer *afb = to_amdgpu_framebuffer(fb); bool pflip_needed; + struct dc_plane_state *dc_plane; struct dm_plane_state *dm_new_plane_state = to_dm_plane_state(new_plane_state); - if (plane->type == DRM_PLANE_TYPE_CURSOR) { - handle_cursor_update(plane, old_plane_state); + /* Cursor plane is handled after stream updates */ + if (plane->type == DRM_PLANE_TYPE_CURSOR) continue; - } if (!fb || !crtc || pcrtc != crtc) continue; @@ -4433,73 +4718,228 @@ static void amdgpu_dm_commit_planes(struct drm_atomic_state *state, if (!new_crtc_state->active) continue; - pflip_needed = !state->allow_modeset; + pflip_needed = old_plane_state->fb && + old_plane_state->fb != new_plane_state->fb; - spin_lock_irqsave(&crtc->dev->event_lock, flags); - if (acrtc_attach->pflip_status != AMDGPU_FLIP_NONE) { - DRM_ERROR("%s: acrtc %d, already busy\n", - __func__, - acrtc_attach->crtc_id); - /* In commit tail framework this cannot happen */ - WARN_ON(1); + dc_plane = dm_new_plane_state->dc_state; + + if (pflip_needed) { + /* + * Assume even ONE crtc with immediate flip means + * entire can't wait for VBLANK + * TODO Check if it's correct + */ + if (new_pcrtc_state->pageflip_flags & DRM_MODE_PAGE_FLIP_ASYNC) + *wait_for_vblank = false; + + /* + * TODO This might fail and hence better not used, wait + * explicitly on fences instead + * and in general should be called for + * blocking commit to as per framework helpers + */ + abo = gem_to_amdgpu_bo(fb->obj[0]); + r = amdgpu_bo_reserve(abo, true); + if (unlikely(r != 0)) + DRM_ERROR("failed to reserve buffer before flip\n"); + + /* + * Wait for all fences on this FB. Do limited wait to avoid + * deadlock during GPU reset when this fence will not signal + * but we hold reservation lock for the BO. + */ + r = reservation_object_wait_timeout_rcu(abo->tbo.resv, + true, false, + msecs_to_jiffies(5000)); + if (unlikely(r == 0)) + DRM_ERROR("Waiting for fences timed out."); + + + + amdgpu_bo_get_tiling_flags(abo, &tiling_flags); + + amdgpu_bo_unreserve(abo); + + flip->flip_addrs[flip_count].address.grph.addr.low_part = lower_32_bits(afb->address); + flip->flip_addrs[flip_count].address.grph.addr.high_part = upper_32_bits(afb->address); + + dcc_address = get_dcc_address(afb->address, tiling_flags); + flip->flip_addrs[flip_count].address.grph.meta_addr.low_part = lower_32_bits(dcc_address); + flip->flip_addrs[flip_count].address.grph.meta_addr.high_part = upper_32_bits(dcc_address); + + flip->flip_addrs[flip_count].flip_immediate = + (crtc->state->pageflip_flags & DRM_MODE_PAGE_FLIP_ASYNC) != 0; + + timestamp_ns = ktime_get_ns(); + flip->flip_addrs[flip_count].flip_timestamp_in_us = div_u64(timestamp_ns, 1000); + flip->surface_updates[flip_count].flip_addr = &flip->flip_addrs[flip_count]; + flip->surface_updates[flip_count].surface = dc_plane; + + if (!flip->surface_updates[flip_count].surface) { + DRM_ERROR("No surface for CRTC: id=%d\n", + acrtc_attach->crtc_id); + continue; + } + + if (plane == pcrtc->primary) + update_freesync_state_on_stream( + dm, + acrtc_state, + acrtc_state->stream, + dc_plane, + flip->flip_addrs[flip_count].flip_timestamp_in_us); + + DRM_DEBUG_DRIVER("%s Flipping to hi: 0x%x, low: 0x%x\n", + __func__, + flip->flip_addrs[flip_count].address.grph.addr.high_part, + flip->flip_addrs[flip_count].address.grph.addr.low_part); + + flip_count += 1; + } + + full->surface_updates[planes_count].surface = dc_plane; + if (new_pcrtc_state->color_mgmt_changed) { + full->surface_updates[planes_count].gamma = dc_plane->gamma_correction; + full->surface_updates[planes_count].in_transfer_func = dc_plane->in_transfer_func; } - spin_unlock_irqrestore(&crtc->dev->event_lock, flags); - if (!pflip_needed || plane->type == DRM_PLANE_TYPE_OVERLAY) { - WARN_ON(!dm_new_plane_state->dc_state); - plane_states_constructed[planes_count] = dm_new_plane_state->dc_state; + full->scaling_infos[planes_count].scaling_quality = dc_plane->scaling_quality; + full->scaling_infos[planes_count].src_rect = dc_plane->src_rect; + full->scaling_infos[planes_count].dst_rect = dc_plane->dst_rect; + full->scaling_infos[planes_count].clip_rect = dc_plane->clip_rect; + full->surface_updates[planes_count].scaling_info = &full->scaling_infos[planes_count]; - dc_stream_attach = acrtc_state->stream; - planes_count++; - } else if (new_crtc_state->planes_changed) { - /* Assume even ONE crtc with immediate flip means - * entire can't wait for VBLANK - * TODO Check if it's correct + full->plane_infos[planes_count].color_space = dc_plane->color_space; + full->plane_infos[planes_count].format = dc_plane->format; + full->plane_infos[planes_count].plane_size = dc_plane->plane_size; + full->plane_infos[planes_count].rotation = dc_plane->rotation; + full->plane_infos[planes_count].horizontal_mirror = dc_plane->horizontal_mirror; + full->plane_infos[planes_count].stereo_format = dc_plane->stereo_format; + full->plane_infos[planes_count].tiling_info = dc_plane->tiling_info; + full->plane_infos[planes_count].visible = dc_plane->visible; + full->plane_infos[planes_count].per_pixel_alpha = dc_plane->per_pixel_alpha; + full->plane_infos[planes_count].dcc = dc_plane->dcc; + full->surface_updates[planes_count].plane_info = &full->plane_infos[planes_count]; + + planes_count += 1; + + } + + /* + * TODO: For proper atomic behaviour, we should be calling into DC once with + * all the changes. However, DC refuses to do pageflips and non-pageflip + * changes in the same call. Change DC to respect atomic behaviour, + * hopefully eliminating dc_*_update structs in their entirety. + */ + if (flip_count) { + if (!vrr_active) { + /* Use old throttling in non-vrr fixed refresh rate mode + * to keep flip scheduling based on target vblank counts + * working in a backwards compatible way, e.g., for + * clients using the GLX_OML_sync_control extension or + * DRI3/Present extension with defined target_msc. + */ + last_flip_vblank = drm_crtc_vblank_count(pcrtc); + } + else { + /* For variable refresh rate mode only: + * Get vblank of last completed flip to avoid > 1 vrr + * flips per video frame by use of throttling, but allow + * flip programming anywhere in the possibly large + * variable vrr vblank interval for fine-grained flip + * timing control and more opportunity to avoid stutter + * on late submission of flips. */ - *wait_for_vblank = - new_pcrtc_state->pageflip_flags & DRM_MODE_PAGE_FLIP_ASYNC ? - false : true; - - /* TODO: Needs rework for multiplane flip */ - if (plane->type == DRM_PLANE_TYPE_PRIMARY) - drm_crtc_vblank_get(crtc); - - amdgpu_dm_do_flip( - crtc, - fb, - (uint32_t)drm_crtc_vblank_count(crtc) + *wait_for_vblank, - dm_state->context); + spin_lock_irqsave(&pcrtc->dev->event_lock, flags); + last_flip_vblank = acrtc_attach->last_flip_vblank; + spin_unlock_irqrestore(&pcrtc->dev->event_lock, flags); } - } + target = (uint32_t)last_flip_vblank + *wait_for_vblank; - if (planes_count) { - unsigned long flags; + /* Prepare wait for target vblank early - before the fence-waits */ + target_vblank = target - (uint32_t)drm_crtc_vblank_count(pcrtc) + + amdgpu_get_vblank_counter_kms(pcrtc->dev, acrtc_attach->crtc_id); - if (new_pcrtc_state->event) { + /* + * Wait until we're out of the vertical blank period before the one + * targeted by the flip + */ + while ((acrtc_attach->enabled && + (amdgpu_display_get_crtc_scanoutpos(dm->ddev, acrtc_attach->crtc_id, + 0, &vpos, &hpos, NULL, + NULL, &pcrtc->hwmode) + & (DRM_SCANOUTPOS_VALID | DRM_SCANOUTPOS_IN_VBLANK)) == + (DRM_SCANOUTPOS_VALID | DRM_SCANOUTPOS_IN_VBLANK) && + (int)(target_vblank - + amdgpu_get_vblank_counter_kms(dm->ddev, acrtc_attach->crtc_id)) > 0)) { + usleep_range(1000, 1100); + } + if (acrtc_attach->base.state->event) { drm_crtc_vblank_get(pcrtc); spin_lock_irqsave(&pcrtc->dev->event_lock, flags); + + WARN_ON(acrtc_attach->pflip_status != AMDGPU_FLIP_NONE); prepare_flip_isr(acrtc_attach); + spin_unlock_irqrestore(&pcrtc->dev->event_lock, flags); } - dc_stream_attach->adjust = acrtc_state->adjust; - dc_stream_attach->vrr_infopacket = acrtc_state->vrr_infopacket; + if (acrtc_state->stream) { - if (false == commit_planes_to_stream(dm->dc, - plane_states_constructed, - planes_count, - acrtc_state, - dm_old_crtc_state, - dm_state->context)) - dm_error("%s: Failed to attach plane!\n", __func__); - } else { - /*TODO BUG Here should go disable planes on CRTC. */ + if (acrtc_state->freesync_timing_changed) + flip->stream_update.adjust = + &acrtc_state->stream->adjust; + + if (acrtc_state->freesync_vrr_info_changed) + flip->stream_update.vrr_infopacket = + &acrtc_state->stream->vrr_infopacket; + } + + mutex_lock(&dm->dc_lock); + dc_commit_updates_for_stream(dm->dc, + flip->surface_updates, + flip_count, + acrtc_state->stream, + &flip->stream_update, + dc_state); + mutex_unlock(&dm->dc_lock); + } + + if (planes_count) { + if (new_pcrtc_state->mode_changed) { + full->stream_update.src = acrtc_state->stream->src; + full->stream_update.dst = acrtc_state->stream->dst; + } + + if (new_pcrtc_state->color_mgmt_changed) + full->stream_update.out_transfer_func = acrtc_state->stream->out_transfer_func; + + acrtc_state->stream->abm_level = acrtc_state->abm_level; + if (acrtc_state->abm_level != dm_old_crtc_state->abm_level) + full->stream_update.abm_level = &acrtc_state->abm_level; + + mutex_lock(&dm->dc_lock); + dc_commit_updates_for_stream(dm->dc, + full->surface_updates, + planes_count, + acrtc_state->stream, + &full->stream_update, + dc_state); + mutex_unlock(&dm->dc_lock); } + + for_each_oldnew_plane_in_state(state, plane, old_plane_state, new_plane_state, i) + if (plane->type == DRM_PLANE_TYPE_CURSOR) + handle_cursor_update(plane, old_plane_state); + +cleanup: + kfree(flip); + kfree(full); } /* @@ -4513,7 +4953,8 @@ static void amdgpu_dm_commit_planes(struct drm_atomic_state *state, static void amdgpu_dm_crtc_copy_transient_flags(struct drm_crtc_state *crtc_state, struct dc_stream_state *stream_state) { - stream_state->mode_changed = crtc_state->mode_changed; + stream_state->mode_changed = + crtc_state->mode_changed || crtc_state->active_changed; } static int amdgpu_dm_atomic_commit(struct drm_device *dev, @@ -4534,10 +4975,25 @@ static int amdgpu_dm_atomic_commit(struct drm_device *dev, */ for_each_oldnew_crtc_in_state(state, crtc, old_crtc_state, new_crtc_state, i) { struct dm_crtc_state *dm_old_crtc_state = to_dm_crtc_state(old_crtc_state); + struct dm_crtc_state *dm_new_crtc_state = to_dm_crtc_state(new_crtc_state); struct amdgpu_crtc *acrtc = to_amdgpu_crtc(crtc); - if (drm_atomic_crtc_needs_modeset(new_crtc_state) && dm_old_crtc_state->stream) + if (drm_atomic_crtc_needs_modeset(new_crtc_state) + && dm_old_crtc_state->stream) { + /* + * If the stream is removed and CRC capture was + * enabled on the CRTC the extra vblank reference + * needs to be dropped since CRC capture will be + * disabled. + */ + if (!dm_new_crtc_state->stream + && dm_new_crtc_state->crc_enabled) { + drm_crtc_vblank_put(crtc); + dm_new_crtc_state->crc_enabled = false; + } + manage_dm_interrupts(adev, acrtc, false); + } } /* * Add check here for SoC's that support hardware cursor plane, to @@ -4549,12 +5005,21 @@ static int amdgpu_dm_atomic_commit(struct drm_device *dev, /*TODO Handle EINTR, reenable IRQ*/ } +/** + * amdgpu_dm_atomic_commit_tail() - AMDgpu DM's commit tail implementation. + * @state: The atomic state to commit + * + * This will tell DC to commit the constructed DC state from atomic_check, + * programming the hardware. Any failures here implies a hardware failure, since + * atomic check should have filtered anything non-kosher. + */ static void amdgpu_dm_atomic_commit_tail(struct drm_atomic_state *state) { struct drm_device *dev = state->dev; struct amdgpu_device *adev = dev->dev_private; struct amdgpu_display_manager *dm = &adev->dm; struct dm_atomic_state *dm_state; + struct dc_state *dc_state = NULL, *dc_state_temp = NULL; uint32_t i, j; struct drm_crtc *crtc; struct drm_crtc_state *old_crtc_state, *new_crtc_state; @@ -4567,7 +5032,16 @@ static void amdgpu_dm_atomic_commit_tail(struct drm_atomic_state *state) drm_atomic_helper_update_legacy_modeset_state(dev, state); - dm_state = to_dm_atomic_state(state); + dm_state = dm_atomic_get_new_state(state); + if (dm_state && dm_state->context) { + dc_state = dm_state->context; + } else { + /* No state changes, retain current state. */ + dc_state_temp = dc_create_state(); + ASSERT(dc_state_temp); + dc_state = dc_state_temp; + dc_resource_state_copy_construct_current(dm->dc, dc_state); + } /* update changed items */ for_each_oldnew_crtc_in_state(state, crtc, old_crtc_state, new_crtc_state, i) { @@ -4640,9 +5114,11 @@ static void amdgpu_dm_atomic_commit_tail(struct drm_atomic_state *state) } } /* for_each_crtc_in_state() */ - if (dm_state->context) { - dm_enable_per_frame_crtc_master_sync(dm_state->context); - WARN_ON(!dc_commit_state(dm->dc, dm_state->context)); + if (dc_state) { + dm_enable_per_frame_crtc_master_sync(dc_state); + mutex_lock(&dm->dc_lock); + WARN_ON(!dc_commit_state(dm->dc, dc_state)); + mutex_unlock(&dm->dc_lock); } for_each_new_crtc_in_state(state, crtc, new_crtc_state, i) { @@ -4655,19 +5131,28 @@ static void amdgpu_dm_atomic_commit_tail(struct drm_atomic_state *state) dc_stream_get_status(dm_new_crtc_state->stream); if (!status) + status = dc_stream_get_status_from_state(dc_state, + dm_new_crtc_state->stream); + + if (!status) DC_ERR("got no status for stream %p on acrtc%p\n", dm_new_crtc_state->stream, acrtc); else acrtc->otg_inst = status->primary_otg_inst; } } - /* Handle scaling and underscan changes*/ + /* Handle connector state changes */ for_each_oldnew_connector_in_state(state, connector, old_con_state, new_con_state, i) { struct dm_connector_state *dm_new_con_state = to_dm_connector_state(new_con_state); struct dm_connector_state *dm_old_con_state = to_dm_connector_state(old_con_state); struct amdgpu_crtc *acrtc = to_amdgpu_crtc(dm_new_con_state->base.crtc); + struct dc_surface_update dummy_updates[MAX_SURFACES]; + struct dc_stream_update stream_update; struct dc_stream_status *status = NULL; + memset(&dummy_updates, 0, sizeof(dummy_updates)); + memset(&stream_update, 0, sizeof(stream_update)); + if (acrtc) { new_crtc_state = drm_atomic_get_new_crtc_state(state, &acrtc->base); old_crtc_state = drm_atomic_get_old_crtc_state(state, &acrtc->base); @@ -4677,34 +5162,48 @@ static void amdgpu_dm_atomic_commit_tail(struct drm_atomic_state *state) if (!acrtc || drm_atomic_crtc_needs_modeset(new_crtc_state)) continue; - /* Skip anything that is not scaling or underscan changes */ - if (!is_scaling_state_different(dm_new_con_state, dm_old_con_state)) + dm_new_crtc_state = to_dm_crtc_state(new_crtc_state); + dm_old_crtc_state = to_dm_crtc_state(old_crtc_state); + + if (!is_scaling_state_different(dm_new_con_state, dm_old_con_state) && + (dm_new_crtc_state->abm_level == dm_old_crtc_state->abm_level)) continue; - dm_new_crtc_state = to_dm_crtc_state(new_crtc_state); + if (is_scaling_state_different(dm_new_con_state, dm_old_con_state)) { + update_stream_scaling_settings(&dm_new_con_state->base.crtc->mode, + dm_new_con_state, (struct dc_stream_state *)dm_new_crtc_state->stream); - update_stream_scaling_settings(&dm_new_con_state->base.crtc->mode, - dm_new_con_state, (struct dc_stream_state *)dm_new_crtc_state->stream); + stream_update.src = dm_new_crtc_state->stream->src; + stream_update.dst = dm_new_crtc_state->stream->dst; + } - if (!dm_new_crtc_state->stream) - continue; + if (dm_new_crtc_state->abm_level != dm_old_crtc_state->abm_level) { + dm_new_crtc_state->stream->abm_level = dm_new_crtc_state->abm_level; + + stream_update.abm_level = &dm_new_crtc_state->abm_level; + } status = dc_stream_get_status(dm_new_crtc_state->stream); WARN_ON(!status); WARN_ON(!status->plane_count); - dm_new_crtc_state->stream->adjust = dm_new_crtc_state->adjust; - dm_new_crtc_state->stream->vrr_infopacket = dm_new_crtc_state->vrr_infopacket; + /* + * TODO: DC refuses to perform stream updates without a dc_surface_update. + * Here we create an empty update on each plane. + * To fix this, DC should permit updating only stream properties. + */ + for (j = 0; j < status->plane_count; j++) + dummy_updates[j].surface = status->plane_states[0]; - /*TODO How it works with MPO ?*/ - if (!commit_planes_to_stream( - dm->dc, - status->plane_states, - status->plane_count, - dm_new_crtc_state, - to_dm_crtc_state(old_crtc_state), - dm_state->context)) - dm_error("%s: Failed to update stream scaling!\n", __func__); + + mutex_lock(&dm->dc_lock); + dc_commit_updates_for_stream(dm->dc, + dummy_updates, + status->plane_count, + dm_new_crtc_state->stream, + &stream_update, + dc_state); + mutex_unlock(&dm->dc_lock); } for_each_oldnew_crtc_in_state(state, crtc, old_crtc_state, @@ -4729,6 +5228,12 @@ static void amdgpu_dm_atomic_commit_tail(struct drm_atomic_state *state) continue; manage_dm_interrupts(adev, acrtc, true); + +#ifdef CONFIG_DEBUG_FS + /* The stream has changed so CRC capture needs to re-enabled. */ + if (dm_new_crtc_state->crc_enabled) + amdgpu_dm_crtc_set_crc_source(crtc, "auto"); +#endif } /* update planes when needed per crtc*/ @@ -4736,7 +5241,8 @@ static void amdgpu_dm_atomic_commit_tail(struct drm_atomic_state *state) dm_new_crtc_state = to_dm_crtc_state(new_crtc_state); if (dm_new_crtc_state->stream) - amdgpu_dm_commit_planes(state, dev, dm, crtc, &wait_for_vblank); + amdgpu_dm_commit_planes(state, dc_state, dev, + dm, crtc, &wait_for_vblank); } @@ -4754,18 +5260,12 @@ static void amdgpu_dm_atomic_commit_tail(struct drm_atomic_state *state) } spin_unlock_irqrestore(&adev->ddev->event_lock, flags); + /* Signal HW programming completion */ + drm_atomic_helper_commit_hw_done(state); if (wait_for_vblank) drm_atomic_helper_wait_for_flip_done(dev, state); - /* - * FIXME: - * Delay hw_done() until flip_done() is signaled. This is to block - * another commit from freeing the CRTC state while we're still - * waiting on flip_done. - */ - drm_atomic_helper_commit_hw_done(state); - drm_atomic_helper_cleanup_planes(dev, state); /* @@ -4776,6 +5276,9 @@ static void amdgpu_dm_atomic_commit_tail(struct drm_atomic_state *state) for (i = 0; i < crtc_disable_count; i++) pm_runtime_put_autosuspend(dev->dev); pm_runtime_mark_last_busy(dev->dev); + + if (dc_state_temp) + dc_release_state(dc_state_temp); } @@ -4919,20 +5422,23 @@ static int do_aquire_global_lock(struct drm_device *dev, return ret < 0 ? ret : 0; } -void set_freesync_on_stream(struct amdgpu_display_manager *dm, - struct dm_crtc_state *new_crtc_state, - struct dm_connector_state *new_con_state, - struct dc_stream_state *new_stream) +static void get_freesync_config_for_crtc( + struct dm_crtc_state *new_crtc_state, + struct dm_connector_state *new_con_state) { struct mod_freesync_config config = {0}; - struct mod_vrr_params vrr = {0}; - struct dc_info_packet vrr_infopacket = {0}; struct amdgpu_dm_connector *aconnector = to_amdgpu_dm_connector(new_con_state->base.connector); + struct drm_display_mode *mode = &new_crtc_state->base.mode; + int vrefresh = drm_mode_vrefresh(mode); - if (new_con_state->freesync_capable && - new_con_state->freesync_enable) { - config.state = new_crtc_state->freesync_enabled ? + new_crtc_state->vrr_supported = new_con_state->freesync_capable && + vrefresh >= aconnector->min_vfreq && + vrefresh <= aconnector->max_vfreq; + + if (new_crtc_state->vrr_supported) { + new_crtc_state->stream->ignore_msa_timing_param = true; + config.state = new_crtc_state->base.vrr_enabled ? VRR_STATE_ACTIVE_VARIABLE : VRR_STATE_INACTIVE; config.min_refresh_in_uhz = @@ -4940,33 +5446,33 @@ void set_freesync_on_stream(struct amdgpu_display_manager *dm, config.max_refresh_in_uhz = aconnector->max_vfreq * 1000000; config.vsif_supported = true; + config.btr = true; } - mod_freesync_build_vrr_params(dm->freesync_module, - new_stream, - &config, &vrr); + new_crtc_state->freesync_config = config; +} - mod_freesync_build_vrr_infopacket(dm->freesync_module, - new_stream, - &vrr, - packet_type_fs1, - NULL, - &vrr_infopacket); +static void reset_freesync_config_for_crtc( + struct dm_crtc_state *new_crtc_state) +{ + new_crtc_state->vrr_supported = false; - new_crtc_state->adjust = vrr.adjust; - new_crtc_state->vrr_infopacket = vrr_infopacket; + memset(&new_crtc_state->vrr_params, 0, + sizeof(new_crtc_state->vrr_params)); + memset(&new_crtc_state->vrr_infopacket, 0, + sizeof(new_crtc_state->vrr_infopacket)); } -static int dm_update_crtcs_state(struct amdgpu_display_manager *dm, - struct drm_atomic_state *state, - bool enable, - bool *lock_and_validation_needed) +static int dm_update_crtc_state(struct amdgpu_display_manager *dm, + struct drm_atomic_state *state, + struct drm_crtc *crtc, + struct drm_crtc_state *old_crtc_state, + struct drm_crtc_state *new_crtc_state, + bool enable, + bool *lock_and_validation_needed) { - struct drm_crtc *crtc; - struct drm_crtc_state *old_crtc_state, *new_crtc_state; - int i; + struct dm_atomic_state *dm_state = NULL; struct dm_crtc_state *dm_old_crtc_state, *dm_new_crtc_state; - struct dm_atomic_state *dm_state = to_dm_atomic_state(state); struct dc_stream_state *new_stream; int ret = 0; @@ -4974,191 +5480,203 @@ static int dm_update_crtcs_state(struct amdgpu_display_manager *dm, * TODO Move this code into dm_crtc_atomic_check once we get rid of dc_validation_set * update changed items */ - for_each_oldnew_crtc_in_state(state, crtc, old_crtc_state, new_crtc_state, i) { - struct amdgpu_crtc *acrtc = NULL; - struct amdgpu_dm_connector *aconnector = NULL; - struct drm_connector_state *drm_new_conn_state = NULL, *drm_old_conn_state = NULL; - struct dm_connector_state *dm_new_conn_state = NULL, *dm_old_conn_state = NULL; - struct drm_plane_state *new_plane_state = NULL; + struct amdgpu_crtc *acrtc = NULL; + struct amdgpu_dm_connector *aconnector = NULL; + struct drm_connector_state *drm_new_conn_state = NULL, *drm_old_conn_state = NULL; + struct dm_connector_state *dm_new_conn_state = NULL, *dm_old_conn_state = NULL; + struct drm_plane_state *new_plane_state = NULL; - new_stream = NULL; + new_stream = NULL; - dm_old_crtc_state = to_dm_crtc_state(old_crtc_state); - dm_new_crtc_state = to_dm_crtc_state(new_crtc_state); - acrtc = to_amdgpu_crtc(crtc); + dm_old_crtc_state = to_dm_crtc_state(old_crtc_state); + dm_new_crtc_state = to_dm_crtc_state(new_crtc_state); + acrtc = to_amdgpu_crtc(crtc); - new_plane_state = drm_atomic_get_new_plane_state(state, new_crtc_state->crtc->primary); + new_plane_state = drm_atomic_get_new_plane_state(state, new_crtc_state->crtc->primary); - if (new_crtc_state->enable && new_plane_state && !new_plane_state->fb) { - ret = -EINVAL; - goto fail; - } - - aconnector = amdgpu_dm_find_first_crtc_matching_connector(state, crtc); + if (new_crtc_state->enable && new_plane_state && !new_plane_state->fb) { + ret = -EINVAL; + goto fail; + } - /* TODO This hack should go away */ - if (aconnector && enable) { - /* Make sure fake sink is created in plug-in scenario */ - drm_new_conn_state = drm_atomic_get_new_connector_state(state, - &aconnector->base); - drm_old_conn_state = drm_atomic_get_old_connector_state(state, - &aconnector->base); + aconnector = amdgpu_dm_find_first_crtc_matching_connector(state, crtc); - if (IS_ERR(drm_new_conn_state)) { - ret = PTR_ERR_OR_ZERO(drm_new_conn_state); - break; - } + /* TODO This hack should go away */ + if (aconnector && enable) { + /* Make sure fake sink is created in plug-in scenario */ + drm_new_conn_state = drm_atomic_get_new_connector_state(state, + &aconnector->base); + drm_old_conn_state = drm_atomic_get_old_connector_state(state, + &aconnector->base); - dm_new_conn_state = to_dm_connector_state(drm_new_conn_state); - dm_old_conn_state = to_dm_connector_state(drm_old_conn_state); + if (IS_ERR(drm_new_conn_state)) { + ret = PTR_ERR_OR_ZERO(drm_new_conn_state); + goto fail; + } - new_stream = create_stream_for_sink(aconnector, - &new_crtc_state->mode, - dm_new_conn_state); + dm_new_conn_state = to_dm_connector_state(drm_new_conn_state); + dm_old_conn_state = to_dm_connector_state(drm_old_conn_state); - /* - * we can have no stream on ACTION_SET if a display - * was disconnected during S3, in this case it is not an - * error, the OS will be updated after detection, and - * will do the right thing on next atomic commit - */ + if (!drm_atomic_crtc_needs_modeset(new_crtc_state)) + goto skip_modeset; - if (!new_stream) { - DRM_DEBUG_DRIVER("%s: Failed to create new stream for crtc %d\n", - __func__, acrtc->base.base.id); - break; - } + new_stream = create_stream_for_sink(aconnector, + &new_crtc_state->mode, + dm_new_conn_state, + dm_old_crtc_state->stream); - set_freesync_on_stream(dm, dm_new_crtc_state, - dm_new_conn_state, new_stream); + /* + * we can have no stream on ACTION_SET if a display + * was disconnected during S3, in this case it is not an + * error, the OS will be updated after detection, and + * will do the right thing on next atomic commit + */ - if (dc_is_stream_unchanged(new_stream, dm_old_crtc_state->stream) && - dc_is_stream_scaling_unchanged(new_stream, dm_old_crtc_state->stream)) { - new_crtc_state->mode_changed = false; - DRM_DEBUG_DRIVER("Mode change not required, setting mode_changed to %d", - new_crtc_state->mode_changed); - } + if (!new_stream) { + DRM_DEBUG_DRIVER("%s: Failed to create new stream for crtc %d\n", + __func__, acrtc->base.base.id); + ret = -ENOMEM; + goto fail; } - if (dm_old_crtc_state->freesync_enabled != dm_new_crtc_state->freesync_enabled) - new_crtc_state->mode_changed = true; + dm_new_crtc_state->abm_level = dm_new_conn_state->abm_level; - if (!drm_atomic_crtc_needs_modeset(new_crtc_state)) - goto next_crtc; - - DRM_DEBUG_DRIVER( - "amdgpu_crtc id:%d crtc_state_flags: enable:%d, active:%d, " - "planes_changed:%d, mode_changed:%d,active_changed:%d," - "connectors_changed:%d\n", - acrtc->crtc_id, - new_crtc_state->enable, - new_crtc_state->active, - new_crtc_state->planes_changed, - new_crtc_state->mode_changed, - new_crtc_state->active_changed, - new_crtc_state->connectors_changed); + if (dc_is_stream_unchanged(new_stream, dm_old_crtc_state->stream) && + dc_is_stream_scaling_unchanged(new_stream, dm_old_crtc_state->stream)) { + new_crtc_state->mode_changed = false; + DRM_DEBUG_DRIVER("Mode change not required, setting mode_changed to %d", + new_crtc_state->mode_changed); + } + } - /* Remove stream for any changed/disabled CRTC */ - if (!enable) { + /* mode_changed flag may get updated above, need to check again */ + if (!drm_atomic_crtc_needs_modeset(new_crtc_state)) + goto skip_modeset; - if (!dm_old_crtc_state->stream) - goto next_crtc; + DRM_DEBUG_DRIVER( + "amdgpu_crtc id:%d crtc_state_flags: enable:%d, active:%d, " + "planes_changed:%d, mode_changed:%d,active_changed:%d," + "connectors_changed:%d\n", + acrtc->crtc_id, + new_crtc_state->enable, + new_crtc_state->active, + new_crtc_state->planes_changed, + new_crtc_state->mode_changed, + new_crtc_state->active_changed, + new_crtc_state->connectors_changed); - DRM_DEBUG_DRIVER("Disabling DRM crtc: %d\n", - crtc->base.id); + /* Remove stream for any changed/disabled CRTC */ + if (!enable) { - /* i.e. reset mode */ - if (dc_remove_stream_from_ctx( - dm->dc, - dm_state->context, - dm_old_crtc_state->stream) != DC_OK) { - ret = -EINVAL; - goto fail; - } + if (!dm_old_crtc_state->stream) + goto skip_modeset; - dc_stream_release(dm_old_crtc_state->stream); - dm_new_crtc_state->stream = NULL; + ret = dm_atomic_get_state(state, &dm_state); + if (ret) + goto fail; - *lock_and_validation_needed = true; + DRM_DEBUG_DRIVER("Disabling DRM crtc: %d\n", + crtc->base.id); - } else {/* Add stream for any updated/enabled CRTC */ - /* - * Quick fix to prevent NULL pointer on new_stream when - * added MST connectors not found in existing crtc_state in the chained mode - * TODO: need to dig out the root cause of that - */ - if (!aconnector || (!aconnector->dc_sink && aconnector->mst_port)) - goto next_crtc; + /* i.e. reset mode */ + if (dc_remove_stream_from_ctx( + dm->dc, + dm_state->context, + dm_old_crtc_state->stream) != DC_OK) { + ret = -EINVAL; + goto fail; + } - if (modereset_required(new_crtc_state)) - goto next_crtc; + dc_stream_release(dm_old_crtc_state->stream); + dm_new_crtc_state->stream = NULL; - if (modeset_required(new_crtc_state, new_stream, - dm_old_crtc_state->stream)) { + reset_freesync_config_for_crtc(dm_new_crtc_state); - WARN_ON(dm_new_crtc_state->stream); + *lock_and_validation_needed = true; - dm_new_crtc_state->stream = new_stream; + } else {/* Add stream for any updated/enabled CRTC */ + /* + * Quick fix to prevent NULL pointer on new_stream when + * added MST connectors not found in existing crtc_state in the chained mode + * TODO: need to dig out the root cause of that + */ + if (!aconnector || (!aconnector->dc_sink && aconnector->mst_port)) + goto skip_modeset; - dc_stream_retain(new_stream); + if (modereset_required(new_crtc_state)) + goto skip_modeset; - DRM_DEBUG_DRIVER("Enabling DRM crtc: %d\n", - crtc->base.id); + if (modeset_required(new_crtc_state, new_stream, + dm_old_crtc_state->stream)) { - if (dc_add_stream_to_ctx( - dm->dc, - dm_state->context, - dm_new_crtc_state->stream) != DC_OK) { - ret = -EINVAL; - goto fail; - } + WARN_ON(dm_new_crtc_state->stream); - *lock_and_validation_needed = true; - } - } + ret = dm_atomic_get_state(state, &dm_state); + if (ret) + goto fail; -next_crtc: - /* Release extra reference */ - if (new_stream) - dc_stream_release(new_stream); + dm_new_crtc_state->stream = new_stream; - /* - * We want to do dc stream updates that do not require a - * full modeset below. - */ - if (!(enable && aconnector && new_crtc_state->enable && - new_crtc_state->active)) - continue; - /* - * Given above conditions, the dc state cannot be NULL because: - * 1. We're in the process of enabling CRTCs (just been added - * to the dc context, or already is on the context) - * 2. Has a valid connector attached, and - * 3. Is currently active and enabled. - * => The dc stream state currently exists. - */ - BUG_ON(dm_new_crtc_state->stream == NULL); + dc_stream_retain(new_stream); - /* Scaling or underscan settings */ - if (is_scaling_state_different(dm_old_conn_state, dm_new_conn_state)) - update_stream_scaling_settings( - &new_crtc_state->mode, dm_new_conn_state, dm_new_crtc_state->stream); + DRM_DEBUG_DRIVER("Enabling DRM crtc: %d\n", + crtc->base.id); - /* - * Color management settings. We also update color properties - * when a modeset is needed, to ensure it gets reprogrammed. - */ - if (dm_new_crtc_state->base.color_mgmt_changed || - drm_atomic_crtc_needs_modeset(new_crtc_state)) { - ret = amdgpu_dm_set_regamma_lut(dm_new_crtc_state); - if (ret) + if (dc_add_stream_to_ctx( + dm->dc, + dm_state->context, + dm_new_crtc_state->stream) != DC_OK) { + ret = -EINVAL; goto fail; - amdgpu_dm_set_ctm(dm_new_crtc_state); + } + + *lock_and_validation_needed = true; } + } +skip_modeset: + /* Release extra reference */ + if (new_stream) + dc_stream_release(new_stream); + /* + * We want to do dc stream updates that do not require a + * full modeset below. + */ + if (!(enable && aconnector && new_crtc_state->enable && + new_crtc_state->active)) + return 0; + /* + * Given above conditions, the dc state cannot be NULL because: + * 1. We're in the process of enabling CRTCs (just been added + * to the dc context, or already is on the context) + * 2. Has a valid connector attached, and + * 3. Is currently active and enabled. + * => The dc stream state currently exists. + */ + BUG_ON(dm_new_crtc_state->stream == NULL); + + /* Scaling or underscan settings */ + if (is_scaling_state_different(dm_old_conn_state, dm_new_conn_state)) + update_stream_scaling_settings( + &new_crtc_state->mode, dm_new_conn_state, dm_new_crtc_state->stream); + + /* + * Color management settings. We also update color properties + * when a modeset is needed, to ensure it gets reprogrammed. + */ + if (dm_new_crtc_state->base.color_mgmt_changed || + drm_atomic_crtc_needs_modeset(new_crtc_state)) { + ret = amdgpu_dm_set_regamma_lut(dm_new_crtc_state); + if (ret) + goto fail; + amdgpu_dm_set_ctm(dm_new_crtc_state); } + /* Update Freesync settings. */ + get_freesync_config_for_crtc(dm_new_crtc_state, + dm_new_conn_state); + return ret; fail: @@ -5167,144 +5685,154 @@ fail: return ret; } -static int dm_update_planes_state(struct dc *dc, - struct drm_atomic_state *state, - bool enable, - bool *lock_and_validation_needed) +static int dm_update_plane_state(struct dc *dc, + struct drm_atomic_state *state, + struct drm_plane *plane, + struct drm_plane_state *old_plane_state, + struct drm_plane_state *new_plane_state, + bool enable, + bool *lock_and_validation_needed) { + + struct dm_atomic_state *dm_state = NULL; struct drm_crtc *new_plane_crtc, *old_plane_crtc; struct drm_crtc_state *old_crtc_state, *new_crtc_state; - struct drm_plane *plane; - struct drm_plane_state *old_plane_state, *new_plane_state; struct dm_crtc_state *dm_new_crtc_state, *dm_old_crtc_state; - struct dm_atomic_state *dm_state = to_dm_atomic_state(state); struct dm_plane_state *dm_new_plane_state, *dm_old_plane_state; - int i ; /* TODO return page_flip_needed() function */ bool pflip_needed = !state->allow_modeset; int ret = 0; - /* Add new planes, in reverse order as DC expectation */ - for_each_oldnew_plane_in_state_reverse(state, plane, old_plane_state, new_plane_state, i) { - new_plane_crtc = new_plane_state->crtc; - old_plane_crtc = old_plane_state->crtc; - dm_new_plane_state = to_dm_plane_state(new_plane_state); - dm_old_plane_state = to_dm_plane_state(old_plane_state); + new_plane_crtc = new_plane_state->crtc; + old_plane_crtc = old_plane_state->crtc; + dm_new_plane_state = to_dm_plane_state(new_plane_state); + dm_old_plane_state = to_dm_plane_state(old_plane_state); - /*TODO Implement atomic check for cursor plane */ - if (plane->type == DRM_PLANE_TYPE_CURSOR) - continue; + /*TODO Implement atomic check for cursor plane */ + if (plane->type == DRM_PLANE_TYPE_CURSOR) + return 0; - /* Remove any changed/removed planes */ - if (!enable) { - if (pflip_needed && - plane->type != DRM_PLANE_TYPE_OVERLAY) - continue; + /* Remove any changed/removed planes */ + if (!enable) { + if (pflip_needed && + plane->type != DRM_PLANE_TYPE_OVERLAY) + return 0; - if (!old_plane_crtc) - continue; + if (!old_plane_crtc) + return 0; - old_crtc_state = drm_atomic_get_old_crtc_state( - state, old_plane_crtc); - dm_old_crtc_state = to_dm_crtc_state(old_crtc_state); + old_crtc_state = drm_atomic_get_old_crtc_state( + state, old_plane_crtc); + dm_old_crtc_state = to_dm_crtc_state(old_crtc_state); - if (!dm_old_crtc_state->stream) - continue; + if (!dm_old_crtc_state->stream) + return 0; - DRM_DEBUG_ATOMIC("Disabling DRM plane: %d on DRM crtc %d\n", - plane->base.id, old_plane_crtc->base.id); + DRM_DEBUG_ATOMIC("Disabling DRM plane: %d on DRM crtc %d\n", + plane->base.id, old_plane_crtc->base.id); - if (!dc_remove_plane_from_context( - dc, - dm_old_crtc_state->stream, - dm_old_plane_state->dc_state, - dm_state->context)) { + ret = dm_atomic_get_state(state, &dm_state); + if (ret) + return ret; - ret = EINVAL; - return ret; - } + if (!dc_remove_plane_from_context( + dc, + dm_old_crtc_state->stream, + dm_old_plane_state->dc_state, + dm_state->context)) { + ret = EINVAL; + return ret; + } - dc_plane_state_release(dm_old_plane_state->dc_state); - dm_new_plane_state->dc_state = NULL; - *lock_and_validation_needed = true; + dc_plane_state_release(dm_old_plane_state->dc_state); + dm_new_plane_state->dc_state = NULL; - } else { /* Add new planes */ - struct dc_plane_state *dc_new_plane_state; + *lock_and_validation_needed = true; - if (drm_atomic_plane_disabling(plane->state, new_plane_state)) - continue; + } else { /* Add new planes */ + struct dc_plane_state *dc_new_plane_state; - if (!new_plane_crtc) - continue; + if (drm_atomic_plane_disabling(plane->state, new_plane_state)) + return 0; - new_crtc_state = drm_atomic_get_new_crtc_state(state, new_plane_crtc); - dm_new_crtc_state = to_dm_crtc_state(new_crtc_state); + if (!new_plane_crtc) + return 0; - if (!dm_new_crtc_state->stream) - continue; + new_crtc_state = drm_atomic_get_new_crtc_state(state, new_plane_crtc); + dm_new_crtc_state = to_dm_crtc_state(new_crtc_state); - if (pflip_needed && - plane->type != DRM_PLANE_TYPE_OVERLAY) - continue; + if (!dm_new_crtc_state->stream) + return 0; - WARN_ON(dm_new_plane_state->dc_state); + if (pflip_needed && plane->type != DRM_PLANE_TYPE_OVERLAY) + return 0; - dc_new_plane_state = dc_create_plane_state(dc); - if (!dc_new_plane_state) - return -ENOMEM; + WARN_ON(dm_new_plane_state->dc_state); - DRM_DEBUG_DRIVER("Enabling DRM plane: %d on DRM crtc %d\n", - plane->base.id, new_plane_crtc->base.id); + dc_new_plane_state = dc_create_plane_state(dc); + if (!dc_new_plane_state) + return -ENOMEM; - ret = fill_plane_attributes( - new_plane_crtc->dev->dev_private, - dc_new_plane_state, - new_plane_state, - new_crtc_state); - if (ret) { - dc_plane_state_release(dc_new_plane_state); - return ret; - } + DRM_DEBUG_DRIVER("Enabling DRM plane: %d on DRM crtc %d\n", + plane->base.id, new_plane_crtc->base.id); - /* - * Any atomic check errors that occur after this will - * not need a release. The plane state will be attached - * to the stream, and therefore part of the atomic - * state. It'll be released when the atomic state is - * cleaned. - */ - if (!dc_add_plane_to_context( - dc, - dm_new_crtc_state->stream, - dc_new_plane_state, - dm_state->context)) { - - dc_plane_state_release(dc_new_plane_state); - return -EINVAL; - } + ret = fill_plane_attributes( + new_plane_crtc->dev->dev_private, + dc_new_plane_state, + new_plane_state, + new_crtc_state); + if (ret) { + dc_plane_state_release(dc_new_plane_state); + return ret; + } - dm_new_plane_state->dc_state = dc_new_plane_state; + ret = dm_atomic_get_state(state, &dm_state); + if (ret) { + dc_plane_state_release(dc_new_plane_state); + return ret; + } - /* Tell DC to do a full surface update every time there - * is a plane change. Inefficient, but works for now. - */ - dm_new_plane_state->dc_state->update_flags.bits.full_update = 1; + /* + * Any atomic check errors that occur after this will + * not need a release. The plane state will be attached + * to the stream, and therefore part of the atomic + * state. It'll be released when the atomic state is + * cleaned. + */ + if (!dc_add_plane_to_context( + dc, + dm_new_crtc_state->stream, + dc_new_plane_state, + dm_state->context)) { - *lock_and_validation_needed = true; + dc_plane_state_release(dc_new_plane_state); + return -EINVAL; } + + dm_new_plane_state->dc_state = dc_new_plane_state; + + /* Tell DC to do a full surface update every time there + * is a plane change. Inefficient, but works for now. + */ + dm_new_plane_state->dc_state->update_flags.bits.full_update = 1; + + *lock_and_validation_needed = true; } return ret; } -enum surface_update_type dm_determine_update_type_for_commit(struct dc *dc, struct drm_atomic_state *state) -{ - - int i, j, num_plane; +static int +dm_determine_update_type_for_commit(struct dc *dc, + struct drm_atomic_state *state, + enum surface_update_type *out_type) +{ + struct dm_atomic_state *dm_state = NULL, *old_dm_state = NULL; + int i, j, num_plane, ret = 0; struct drm_plane_state *old_plane_state, *new_plane_state; struct dm_plane_state *new_dm_plane_state, *old_dm_plane_state; struct drm_crtc *new_plane_crtc, *old_plane_crtc; @@ -5315,96 +5843,154 @@ enum surface_update_type dm_determine_update_type_for_commit(struct dc *dc, stru struct dm_crtc_state *new_dm_crtc_state, *old_dm_crtc_state; struct dc_stream_status *status = NULL; - struct dc_surface_update *updates = kzalloc(MAX_SURFACES * sizeof(struct dc_surface_update), GFP_KERNEL); - struct dc_plane_state *surface = kzalloc(MAX_SURFACES * sizeof(struct dc_plane_state), GFP_KERNEL); - struct dc_stream_update stream_update; + struct dc_surface_update *updates; + struct dc_plane_state *surface; enum surface_update_type update_type = UPDATE_TYPE_FAST; + updates = kcalloc(MAX_SURFACES, sizeof(*updates), GFP_KERNEL); + surface = kcalloc(MAX_SURFACES, sizeof(*surface), GFP_KERNEL); + + if (!updates || !surface) { + DRM_ERROR("Plane or surface update failed to allocate"); + /* Set type to FULL to avoid crashing in DC*/ + update_type = UPDATE_TYPE_FULL; + goto cleanup; + } for_each_oldnew_crtc_in_state(state, crtc, old_crtc_state, new_crtc_state, i) { + struct dc_stream_update stream_update = { 0 }; + new_dm_crtc_state = to_dm_crtc_state(new_crtc_state); old_dm_crtc_state = to_dm_crtc_state(old_crtc_state); num_plane = 0; - if (new_dm_crtc_state->stream) { - - for_each_oldnew_plane_in_state(state, plane, old_plane_state, new_plane_state, j) { - new_plane_crtc = new_plane_state->crtc; - old_plane_crtc = old_plane_state->crtc; - new_dm_plane_state = to_dm_plane_state(new_plane_state); - old_dm_plane_state = to_dm_plane_state(old_plane_state); - - if (plane->type == DRM_PLANE_TYPE_CURSOR) - continue; - - if (!state->allow_modeset) - continue; - - if (crtc == new_plane_crtc) { - updates[num_plane].surface = &surface[num_plane]; - - if (new_crtc_state->mode_changed) { - updates[num_plane].surface->src_rect = - new_dm_plane_state->dc_state->src_rect; - updates[num_plane].surface->dst_rect = - new_dm_plane_state->dc_state->dst_rect; - updates[num_plane].surface->rotation = - new_dm_plane_state->dc_state->rotation; - updates[num_plane].surface->in_transfer_func = - new_dm_plane_state->dc_state->in_transfer_func; - stream_update.dst = new_dm_crtc_state->stream->dst; - stream_update.src = new_dm_crtc_state->stream->src; - } - - if (new_crtc_state->color_mgmt_changed) { - updates[num_plane].gamma = - new_dm_plane_state->dc_state->gamma_correction; - updates[num_plane].in_transfer_func = - new_dm_plane_state->dc_state->in_transfer_func; - stream_update.gamut_remap = - &new_dm_crtc_state->stream->gamut_remap_matrix; - stream_update.out_transfer_func = - new_dm_crtc_state->stream->out_transfer_func; - } - - num_plane++; - } + if (new_dm_crtc_state->stream != old_dm_crtc_state->stream) { + update_type = UPDATE_TYPE_FULL; + goto cleanup; + } + + if (!new_dm_crtc_state->stream) + continue; + + for_each_oldnew_plane_in_state(state, plane, old_plane_state, new_plane_state, j) { + new_plane_crtc = new_plane_state->crtc; + old_plane_crtc = old_plane_state->crtc; + new_dm_plane_state = to_dm_plane_state(new_plane_state); + old_dm_plane_state = to_dm_plane_state(old_plane_state); + + if (plane->type == DRM_PLANE_TYPE_CURSOR) + continue; + + if (new_dm_plane_state->dc_state != old_dm_plane_state->dc_state) { + update_type = UPDATE_TYPE_FULL; + goto cleanup; } - if (num_plane > 0) { - status = dc_stream_get_status(new_dm_crtc_state->stream); - update_type = dc_check_update_surfaces_for_stream(dc, updates, num_plane, - &stream_update, status); + if (!state->allow_modeset) + continue; + + if (crtc != new_plane_crtc) + continue; - if (update_type > UPDATE_TYPE_MED) { - update_type = UPDATE_TYPE_FULL; - goto ret; - } + updates[num_plane].surface = &surface[num_plane]; + + if (new_crtc_state->mode_changed) { + updates[num_plane].surface->src_rect = + new_dm_plane_state->dc_state->src_rect; + updates[num_plane].surface->dst_rect = + new_dm_plane_state->dc_state->dst_rect; + updates[num_plane].surface->rotation = + new_dm_plane_state->dc_state->rotation; + updates[num_plane].surface->in_transfer_func = + new_dm_plane_state->dc_state->in_transfer_func; + stream_update.dst = new_dm_crtc_state->stream->dst; + stream_update.src = new_dm_crtc_state->stream->src; } - } else if (!new_dm_crtc_state->stream && old_dm_crtc_state->stream) { + if (new_crtc_state->color_mgmt_changed) { + updates[num_plane].gamma = + new_dm_plane_state->dc_state->gamma_correction; + updates[num_plane].in_transfer_func = + new_dm_plane_state->dc_state->in_transfer_func; + stream_update.gamut_remap = + &new_dm_crtc_state->stream->gamut_remap_matrix; + stream_update.out_transfer_func = + new_dm_crtc_state->stream->out_transfer_func; + } + + num_plane++; + } + + if (num_plane == 0) + continue; + + ret = dm_atomic_get_state(state, &dm_state); + if (ret) + goto cleanup; + + old_dm_state = dm_atomic_get_old_state(state); + if (!old_dm_state) { + ret = -EINVAL; + goto cleanup; + } + + status = dc_stream_get_status_from_state(old_dm_state->context, + new_dm_crtc_state->stream); + + update_type = dc_check_update_surfaces_for_stream(dc, updates, num_plane, + &stream_update, status); + + if (update_type > UPDATE_TYPE_MED) { update_type = UPDATE_TYPE_FULL; - goto ret; + goto cleanup; } } -ret: +cleanup: kfree(updates); kfree(surface); - return update_type; + *out_type = update_type; + return ret; } +/** + * amdgpu_dm_atomic_check() - Atomic check implementation for AMDgpu DM. + * @dev: The DRM device + * @state: The atomic state to commit + * + * Validate that the given atomic state is programmable by DC into hardware. + * This involves constructing a &struct dc_state reflecting the new hardware + * state we wish to commit, then querying DC to see if it is programmable. It's + * important not to modify the existing DC state. Otherwise, atomic_check + * may unexpectedly commit hardware changes. + * + * When validating the DC state, it's important that the right locks are + * acquired. For full updates case which removes/adds/updates streams on one + * CRTC while flipping on another CRTC, acquiring global lock will guarantee + * that any such full update commit will wait for completion of any outstanding + * flip using DRMs synchronization events. See + * dm_determine_update_type_for_commit() + * + * Note that DM adds the affected connectors for all CRTCs in state, when that + * might not seem necessary. This is because DC stream creation requires the + * DC sink, which is tied to the DRM connector state. Cleaning this up should + * be possible but non-trivial - a possible TODO item. + * + * Return: -Error code if validation failed. + */ static int amdgpu_dm_atomic_check(struct drm_device *dev, struct drm_atomic_state *state) { struct amdgpu_device *adev = dev->dev_private; + struct dm_atomic_state *dm_state = NULL; struct dc *dc = adev->dm.dc; - struct dm_atomic_state *dm_state = to_dm_atomic_state(state); struct drm_connector *connector; struct drm_connector_state *old_con_state, *new_con_state; struct drm_crtc *crtc; struct drm_crtc_state *old_crtc_state, *new_crtc_state; + struct drm_plane *plane; + struct drm_plane_state *old_plane_state, *new_plane_state; enum surface_update_type update_type = UPDATE_TYPE_FAST; enum surface_update_type overall_update_type = UPDATE_TYPE_FAST; @@ -5421,12 +6007,9 @@ static int amdgpu_dm_atomic_check(struct drm_device *dev, goto fail; for_each_oldnew_crtc_in_state(state, crtc, old_crtc_state, new_crtc_state, i) { - struct dm_crtc_state *dm_new_crtc_state = to_dm_crtc_state(new_crtc_state); - struct dm_crtc_state *dm_old_crtc_state = to_dm_crtc_state(old_crtc_state); - if (!drm_atomic_crtc_needs_modeset(new_crtc_state) && !new_crtc_state->color_mgmt_changed && - (dm_old_crtc_state->freesync_enabled == dm_new_crtc_state->freesync_enabled)) + old_crtc_state->vrr_enabled == new_crtc_state->vrr_enabled) continue; if (!new_crtc_state->enable) @@ -5441,32 +6024,84 @@ static int amdgpu_dm_atomic_check(struct drm_device *dev, goto fail; } - dm_state->context = dc_create_state(); - ASSERT(dm_state->context); - dc_resource_state_copy_construct_current(dc, dm_state->context); + /* + * Add all primary and overlay planes on the CRTC to the state + * whenever a plane is enabled to maintain correct z-ordering + * and to enable fast surface updates. + */ + drm_for_each_crtc(crtc, dev) { + bool modified = false; + + for_each_oldnew_plane_in_state(state, plane, old_plane_state, new_plane_state, i) { + if (plane->type == DRM_PLANE_TYPE_CURSOR) + continue; + + if (new_plane_state->crtc == crtc || + old_plane_state->crtc == crtc) { + modified = true; + break; + } + } + + if (!modified) + continue; + + drm_for_each_plane_mask(plane, state->dev, crtc->state->plane_mask) { + if (plane->type == DRM_PLANE_TYPE_CURSOR) + continue; + + new_plane_state = + drm_atomic_get_plane_state(state, plane); + + if (IS_ERR(new_plane_state)) { + ret = PTR_ERR(new_plane_state); + goto fail; + } + } + } /* Remove exiting planes if they are modified */ - ret = dm_update_planes_state(dc, state, false, &lock_and_validation_needed); - if (ret) { - goto fail; + for_each_oldnew_plane_in_state_reverse(state, plane, old_plane_state, new_plane_state, i) { + ret = dm_update_plane_state(dc, state, plane, + old_plane_state, + new_plane_state, + false, + &lock_and_validation_needed); + if (ret) + goto fail; } /* Disable all crtcs which require disable */ - ret = dm_update_crtcs_state(&adev->dm, state, false, &lock_and_validation_needed); - if (ret) { - goto fail; + for_each_oldnew_crtc_in_state(state, crtc, old_crtc_state, new_crtc_state, i) { + ret = dm_update_crtc_state(&adev->dm, state, crtc, + old_crtc_state, + new_crtc_state, + false, + &lock_and_validation_needed); + if (ret) + goto fail; } /* Enable all crtcs which require enable */ - ret = dm_update_crtcs_state(&adev->dm, state, true, &lock_and_validation_needed); - if (ret) { - goto fail; + for_each_oldnew_crtc_in_state(state, crtc, old_crtc_state, new_crtc_state, i) { + ret = dm_update_crtc_state(&adev->dm, state, crtc, + old_crtc_state, + new_crtc_state, + true, + &lock_and_validation_needed); + if (ret) + goto fail; } /* Add new/modified planes */ - ret = dm_update_planes_state(dc, state, true, &lock_and_validation_needed); - if (ret) { - goto fail; + for_each_oldnew_plane_in_state_reverse(state, plane, old_plane_state, new_plane_state, i) { + ret = dm_update_plane_state(dc, state, plane, + old_plane_state, + new_plane_state, + true, + &lock_and_validation_needed); + if (ret) + goto fail; } /* Run this here since we want to validate the streams we created */ @@ -5497,16 +6132,9 @@ static int amdgpu_dm_atomic_check(struct drm_device *dev, lock_and_validation_needed = true; } - /* - * For full updates case when - * removing/adding/updating streams on one CRTC while flipping - * on another CRTC, - * acquiring global lock will guarantee that any such full - * update commit - * will wait for completion of any outstanding flip using DRMs - * synchronization events. - */ - update_type = dm_determine_update_type_for_commit(dc, state); + ret = dm_determine_update_type_for_commit(dc, state, &update_type); + if (ret) + goto fail; if (overall_update_type < update_type) overall_update_type = update_type; @@ -5524,6 +6152,9 @@ static int amdgpu_dm_atomic_check(struct drm_device *dev, if (overall_update_type > UPDATE_TYPE_FAST) { + ret = dm_atomic_get_state(state, &dm_state); + if (ret) + goto fail; ret = do_aquire_global_lock(dev, state); if (ret) @@ -5533,6 +6164,13 @@ static int amdgpu_dm_atomic_check(struct drm_device *dev, ret = -EINVAL; goto fail; } + } else if (state->legacy_cursor_update) { + /* + * This is a fast cursor update coming from the plane update + * helper, check if it can be done asynchronously for better + * performance. + */ + state->async_update = !drm_atomic_helper_async_check(dev, state); } /* Must be success */ @@ -5578,14 +6216,15 @@ void amdgpu_dm_update_freesync_caps(struct drm_connector *connector, struct detailed_data_monitor_range *range; struct amdgpu_dm_connector *amdgpu_dm_connector = to_amdgpu_dm_connector(connector); - struct dm_connector_state *dm_con_state; + struct dm_connector_state *dm_con_state = NULL; struct drm_device *dev = connector->dev; struct amdgpu_device *adev = dev->dev_private; + bool freesync_capable = false; if (!connector->state) { DRM_ERROR("%s - Connector has no state", __func__); - return; + goto update; } if (!edid) { @@ -5595,9 +6234,7 @@ void amdgpu_dm_update_freesync_caps(struct drm_connector *connector, amdgpu_dm_connector->max_vfreq = 0; amdgpu_dm_connector->pixel_clock_mhz = 0; - dm_con_state->freesync_capable = false; - dm_con_state->freesync_enable = false; - return; + goto update; } dm_con_state = to_dm_connector_state(connector->state); @@ -5605,10 +6242,10 @@ void amdgpu_dm_update_freesync_caps(struct drm_connector *connector, edid_check_required = false; if (!amdgpu_dm_connector->dc_sink) { DRM_ERROR("dc_sink NULL, could not add free_sync module.\n"); - return; + goto update; } if (!adev->dm.freesync_module) - return; + goto update; /* * if edid non zero restrict freesync only for dp and edp */ @@ -5620,7 +6257,6 @@ void amdgpu_dm_update_freesync_caps(struct drm_connector *connector, amdgpu_dm_connector); } } - dm_con_state->freesync_capable = false; if (edid_check_required == true && (edid->version > 1 || (edid->version == 1 && edid->revision > 1))) { for (i = 0; i < 4; i++) { @@ -5652,8 +6288,16 @@ void amdgpu_dm_update_freesync_caps(struct drm_connector *connector, if (amdgpu_dm_connector->max_vfreq - amdgpu_dm_connector->min_vfreq > 10) { - dm_con_state->freesync_capable = true; + freesync_capable = true; } } + +update: + if (dm_con_state) + dm_con_state->freesync_capable = freesync_capable; + + if (connector->vrr_capable_property) + drm_connector_set_vrr_capable_property(connector, + freesync_capable); } diff --git a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.h b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.h index 6e069d777ab2..fbd161ddc3f4 100644 --- a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.h +++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.h @@ -59,60 +59,140 @@ struct common_irq_params { enum dc_irq_source irq_src; }; +/** + * struct irq_list_head - Linked-list for low context IRQ handlers. + * + * @head: The list_head within &struct handler_data + * @work: A work_struct containing the deferred handler work + */ struct irq_list_head { struct list_head head; /* In case this interrupt needs post-processing, 'work' will be queued*/ struct work_struct work; }; +/** + * struct dm_compressor_info - Buffer info used by frame buffer compression + * @cpu_addr: MMIO cpu addr + * @bo_ptr: Pointer to the buffer object + * @gpu_addr: MMIO gpu addr + */ struct dm_comressor_info { void *cpu_addr; struct amdgpu_bo *bo_ptr; uint64_t gpu_addr; }; +/** + * struct amdgpu_dm_backlight_caps - Usable range of backlight values from ACPI + * @min_input_signal: minimum possible input in range 0-255 + * @max_input_signal: maximum possible input in range 0-255 + * @caps_valid: true if these values are from the ACPI interface + */ +struct amdgpu_dm_backlight_caps { + int min_input_signal; + int max_input_signal; + bool caps_valid; +}; + +/** + * struct amdgpu_display_manager - Central amdgpu display manager device + * + * @dc: Display Core control structure + * @adev: AMDGPU base driver structure + * @ddev: DRM base driver structure + * @display_indexes_num: Max number of display streams supported + * @irq_handler_list_table_lock: Synchronizes access to IRQ tables + * @backlight_dev: Backlight control device + * @cached_state: Caches device atomic state for suspend/resume + * @compressor: Frame buffer compression buffer. See &struct dm_comressor_info + */ struct amdgpu_display_manager { + struct dc *dc; + + /** + * @cgs_device: + * + * The Common Graphics Services device. It provides an interface for + * accessing registers. + */ struct cgs_device *cgs_device; - struct amdgpu_device *adev; /*AMD base driver*/ - struct drm_device *ddev; /*DRM base driver*/ + struct amdgpu_device *adev; + struct drm_device *ddev; u16 display_indexes_num; - /* - * 'irq_source_handler_table' holds a list of handlers - * per (DAL) IRQ source. + /** + * @atomic_obj + * + * In combination with &dm_atomic_state it helps manage + * global atomic state that doesn't map cleanly into existing + * drm resources, like &dc_context. + */ + struct drm_private_obj atomic_obj; + + struct drm_modeset_lock atomic_obj_lock; + + /** + * @dc_lock: + * + * Guards access to DC functions that can issue register write + * sequences. + */ + struct mutex dc_lock; + + /** + * @irq_handler_list_low_tab: + * + * Low priority IRQ handler table. * - * Each IRQ source may need to be handled at different contexts. - * By 'context' we mean, for example: - * - The ISR context, which is the direct interrupt handler. - * - The 'deferred' context - this is the post-processing of the - * interrupt, but at a lower priority. + * It is a n*m table consisting of n IRQ sources, and m handlers per IRQ + * source. Low priority IRQ handlers are deferred to a workqueue to be + * processed. Hence, they can sleep. * * Note that handlers are called in the same order as they were * registered (FIFO). */ struct irq_list_head irq_handler_list_low_tab[DAL_IRQ_SOURCES_NUMBER]; + + /** + * @irq_handler_list_high_tab: + * + * High priority IRQ handler table. + * + * It is a n*m table, same as &irq_handler_list_low_tab. However, + * handlers in this table are not deferred and are called immediately. + */ struct list_head irq_handler_list_high_tab[DAL_IRQ_SOURCES_NUMBER]; + /** + * @pflip_params: + * + * Page flip IRQ parameters, passed to registered handlers when + * triggered. + */ struct common_irq_params pflip_params[DC_IRQ_SOURCE_PFLIP_LAST - DC_IRQ_SOURCE_PFLIP_FIRST + 1]; + /** + * @vblank_params: + * + * Vertical blanking IRQ parameters, passed to registered handlers when + * triggered. + */ struct common_irq_params vblank_params[DC_IRQ_SOURCE_VBLANK6 - DC_IRQ_SOURCE_VBLANK1 + 1]; - /* this spin lock synchronizes access to 'irq_handler_list_table' */ spinlock_t irq_handler_list_table_lock; struct backlight_device *backlight_dev; const struct dc_link *backlight_link; + struct amdgpu_dm_backlight_caps backlight_caps; struct mod_freesync *freesync_module; - /** - * Caches device atomic state for suspend/resume - */ struct drm_atomic_state *cached_state; struct dm_comressor_info compressor; @@ -183,15 +263,21 @@ struct dm_crtc_state { int crc_skip_count; bool crc_enabled; - bool freesync_enabled; - struct dc_crtc_timing_adjust adjust; + bool freesync_timing_changed; + bool freesync_vrr_info_changed; + + bool vrr_supported; + struct mod_freesync_config freesync_config; + struct mod_vrr_params vrr_params; struct dc_info_packet vrr_infopacket; + + int abm_level; }; #define to_dm_crtc_state(x) container_of(x, struct dm_crtc_state, base) struct dm_atomic_state { - struct drm_atomic_state base; + struct drm_private_state base; struct dc_state *context; }; @@ -206,8 +292,8 @@ struct dm_connector_state { uint8_t underscan_hborder; uint8_t max_bpc; bool underscan_enable; - bool freesync_enable; bool freesync_capable; + uint8_t abm_level; }; #define to_dm_connector_state(x)\ diff --git a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_color.c b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_color.c index be19e6861189..216e48cec716 100644 --- a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_color.c +++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_color.c @@ -164,7 +164,7 @@ int amdgpu_dm_set_regamma_lut(struct dm_crtc_state *crtc) */ stream->out_transfer_func->type = TF_TYPE_DISTRIBUTED_POINTS; ret = mod_color_calculate_regamma_params(stream->out_transfer_func, - gamma, true, adev->asic_type <= CHIP_RAVEN); + gamma, true, adev->asic_type <= CHIP_RAVEN, NULL); dc_gamma_release(&gamma); if (!ret) { stream->out_transfer_func->type = old_type; diff --git a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_crc.c b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_crc.c index 01fc5717b657..a10e3a50d9ef 100644 --- a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_crc.c +++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_crc.c @@ -64,8 +64,10 @@ amdgpu_dm_crtc_verify_crc_source(struct drm_crtc *crtc, const char *src_name, int amdgpu_dm_crtc_set_crc_source(struct drm_crtc *crtc, const char *src_name) { + struct amdgpu_device *adev = crtc->dev->dev_private; struct dm_crtc_state *crtc_state = to_dm_crtc_state(crtc->state); struct dc_stream_state *stream_state = crtc_state->stream; + bool enable; enum amdgpu_dm_pipe_crc_source source = dm_parse_crc_source(src_name); @@ -75,29 +77,38 @@ int amdgpu_dm_crtc_set_crc_source(struct drm_crtc *crtc, const char *src_name) return -EINVAL; } - /* When enabling CRC, we should also disable dithering. */ - if (source == AMDGPU_DM_PIPE_CRC_SOURCE_AUTO) { - if (dc_stream_configure_crc(stream_state->ctx->dc, - stream_state, - true, true)) { - crtc_state->crc_enabled = true; - dc_stream_set_dither_option(stream_state, - DITHER_OPTION_TRUN8); - } - else - return -EINVAL; - } else { - if (dc_stream_configure_crc(stream_state->ctx->dc, - stream_state, - false, false)) { - crtc_state->crc_enabled = false; - dc_stream_set_dither_option(stream_state, - DITHER_OPTION_DEFAULT); - } - else - return -EINVAL; + if (!stream_state) { + DRM_ERROR("No stream state for CRTC%d\n", crtc->index); + return -EINVAL; } + enable = (source == AMDGPU_DM_PIPE_CRC_SOURCE_AUTO); + + mutex_lock(&adev->dm.dc_lock); + if (!dc_stream_configure_crc(stream_state->ctx->dc, stream_state, + enable, enable)) { + mutex_unlock(&adev->dm.dc_lock); + return -EINVAL; + } + + /* When enabling CRC, we should also disable dithering. */ + dc_stream_set_dither_option(stream_state, + enable ? DITHER_OPTION_TRUN8 + : DITHER_OPTION_DEFAULT); + + mutex_unlock(&adev->dm.dc_lock); + + /* + * Reading the CRC requires the vblank interrupt handler to be + * enabled. Keep a reference until CRC capture stops. + */ + if (!crtc_state->crc_enabled && enable) + drm_crtc_vblank_get(crtc); + else if (crtc_state->crc_enabled && !enable) + drm_crtc_vblank_put(crtc); + + crtc_state->crc_enabled = enable; + /* Reset crc_skipped on dm state */ crtc_state->crc_skip_count = 0; return 0; diff --git a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_debugfs.c b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_debugfs.c index 9a7ac58eb18e..4a55cde027cf 100644 --- a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_debugfs.c +++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_debugfs.c @@ -671,6 +671,25 @@ static ssize_t dp_phy_test_pattern_debugfs_write(struct file *f, const char __us return bytes_from_user; } +/* + * Returns the min and max vrr vfreq through the connector's debugfs file. + * Example usage: cat /sys/kernel/debug/dri/0/DP-1/vrr_range + */ +static int vrr_range_show(struct seq_file *m, void *data) +{ + struct drm_connector *connector = m->private; + struct amdgpu_dm_connector *aconnector = to_amdgpu_dm_connector(connector); + + if (connector->status != connector_status_connected) + return -ENODEV; + + seq_printf(m, "Min: %u\n", (unsigned int)aconnector->min_vfreq); + seq_printf(m, "Max: %u\n", (unsigned int)aconnector->max_vfreq); + + return 0; +} +DEFINE_SHOW_ATTRIBUTE(vrr_range); + static const struct file_operations dp_link_settings_debugfs_fops = { .owner = THIS_MODULE, .read = dp_link_settings_read, @@ -697,7 +716,8 @@ static const struct { } dp_debugfs_entries[] = { {"link_settings", &dp_link_settings_debugfs_fops}, {"phy_settings", &dp_phy_settings_debugfs_fop}, - {"test_pattern", &dp_phy_test_pattern_fops} + {"test_pattern", &dp_phy_test_pattern_fops}, + {"vrr_range", &vrr_range_fops} }; int connector_debugfs_init(struct amdgpu_dm_connector *connector) @@ -783,6 +803,45 @@ static ssize_t dtn_log_write( return size; } +/* + * Backlight at this moment. Read only. + * As written to display, taking ABM and backlight lut into account. + * Ranges from 0x0 to 0x10000 (= 100% PWM) + */ +static int current_backlight_read(struct seq_file *m, void *data) +{ + struct drm_info_node *node = (struct drm_info_node *)m->private; + struct drm_device *dev = node->minor->dev; + struct amdgpu_device *adev = dev->dev_private; + struct dc *dc = adev->dm.dc; + unsigned int backlight = dc_get_current_backlight_pwm(dc); + + seq_printf(m, "0x%x\n", backlight); + return 0; +} + +/* + * Backlight value that is being approached. Read only. + * As written to display, taking ABM and backlight lut into account. + * Ranges from 0x0 to 0x10000 (= 100% PWM) + */ +static int target_backlight_read(struct seq_file *m, void *data) +{ + struct drm_info_node *node = (struct drm_info_node *)m->private; + struct drm_device *dev = node->minor->dev; + struct amdgpu_device *adev = dev->dev_private; + struct dc *dc = adev->dm.dc; + unsigned int backlight = dc_get_target_backlight_pwm(dc); + + seq_printf(m, "0x%x\n", backlight); + return 0; +} + +static const struct drm_info_list amdgpu_dm_debugfs_list[] = { + {"amdgpu_current_backlight_pwm", ¤t_backlight_read}, + {"amdgpu_target_backlight_pwm", &target_backlight_read}, +}; + int dtn_debugfs_init(struct amdgpu_device *adev) { static const struct file_operations dtn_log_fops = { @@ -793,9 +852,15 @@ int dtn_debugfs_init(struct amdgpu_device *adev) }; struct drm_minor *minor = adev->ddev->primary; - struct dentry *root = minor->debugfs_root; + struct dentry *ent, *root = minor->debugfs_root; + int ret; + + ret = amdgpu_debugfs_add_files(adev, amdgpu_dm_debugfs_list, + ARRAY_SIZE(amdgpu_dm_debugfs_list)); + if (ret) + return ret; - struct dentry *ent = debugfs_create_file( + ent = debugfs_create_file( "amdgpu_dm_dtn_log", 0644, root, diff --git a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_helpers.c b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_helpers.c index 39997d977efb..b39766bd2840 100644 --- a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_helpers.c +++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_helpers.c @@ -29,7 +29,7 @@ #include <linux/i2c.h> #include <drm/drmP.h> -#include <drm/drm_crtc_helper.h> +#include <drm/drm_probe_helper.h> #include <drm/amdgpu_drm.h> #include <drm/drm_edid.h> @@ -192,7 +192,7 @@ bool dm_helpers_dp_mst_write_payload_allocation_table( int bpp = 0; int pbn = 0; - aconnector = stream->sink->priv; + aconnector = (struct amdgpu_dm_connector *)stream->dm_stream_context; if (!aconnector || !aconnector->mst_port) return false; @@ -205,7 +205,7 @@ bool dm_helpers_dp_mst_write_payload_allocation_table( mst_port = aconnector->port; if (enable) { - clock = stream->timing.pix_clk_khz; + clock = stream->timing.pix_clk_100hz / 10; switch (stream->timing.display_color_depth) { @@ -263,6 +263,13 @@ bool dm_helpers_dp_mst_write_payload_allocation_table( return true; } +/* + * poll pending down reply before clear payload allocation table + */ +void dm_helpers_dp_mst_poll_pending_down_reply( + struct dc_context *ctx, + const struct dc_link *link) +{} /* * Clear payload allocation table before enable MST DP link. @@ -284,7 +291,7 @@ bool dm_helpers_dp_mst_poll_for_allocation_change_trigger( struct drm_dp_mst_topology_mgr *mst_mgr; int ret; - aconnector = stream->sink->priv; + aconnector = (struct amdgpu_dm_connector *)stream->dm_stream_context; if (!aconnector || !aconnector->mst_port) return false; @@ -312,7 +319,7 @@ bool dm_helpers_dp_mst_send_payload_allocation( struct drm_dp_mst_port *mst_port; int ret; - aconnector = stream->sink->priv; + aconnector = (struct amdgpu_dm_connector *)stream->dm_stream_context; if (!aconnector || !aconnector->mst_port) return false; diff --git a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_irq.c b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_irq.c index a212178f2edc..cd10f77cdeb0 100644 --- a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_irq.c +++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_irq.c @@ -32,16 +32,55 @@ #include "amdgpu_dm.h" #include "amdgpu_dm_irq.h" +/** + * DOC: overview + * + * DM provides another layer of IRQ management on top of what the base driver + * already provides. This is something that could be cleaned up, and is a + * future TODO item. + * + * The base driver provides IRQ source registration with DRM, handler + * registration into the base driver's IRQ table, and a handler callback + * amdgpu_irq_handler(), with which DRM calls on interrupts. This generic + * handler looks up the IRQ table, and calls the respective + * &amdgpu_irq_src_funcs.process hookups. + * + * What DM provides on top are two IRQ tables specifically for top-half and + * bottom-half IRQ handling, with the bottom-half implementing workqueues: + * + * - &amdgpu_display_manager.irq_handler_list_high_tab + * - &amdgpu_display_manager.irq_handler_list_low_tab + * + * They override the base driver's IRQ table, and the effect can be seen + * in the hooks that DM provides for &amdgpu_irq_src_funcs.process. They + * are all set to the DM generic handler amdgpu_dm_irq_handler(), which looks up + * DM's IRQ tables. However, in order for base driver to recognize this hook, DM + * still needs to register the IRQ with the base driver. See + * dce110_register_irq_handlers() and dcn10_register_irq_handlers(). + * + * To expose DC's hardware interrupt toggle to the base driver, DM implements + * &amdgpu_irq_src_funcs.set hooks. Base driver calls it through + * amdgpu_irq_update() to enable or disable the interrupt. + */ + /****************************************************************************** * Private declarations. *****************************************************************************/ +/** + * struct amdgpu_dm_irq_handler_data - Data for DM interrupt handlers. + * + * @list: Linked list entry referencing the next/previous handler + * @handler: Handler function + * @handler_arg: Argument passed to the handler when triggered + * @dm: DM which this handler belongs to + * @irq_source: DC interrupt source that this handler is registered for + */ struct amdgpu_dm_irq_handler_data { struct list_head list; interrupt_handler handler; void *handler_arg; - /* DM which this handler belongs to */ struct amdgpu_display_manager *dm; /* DAL irq source which registered for this interrupt. */ enum dc_irq_source irq_source; @@ -68,7 +107,7 @@ static void init_handler_common_data(struct amdgpu_dm_irq_handler_data *hcd, } /** - * dm_irq_work_func - Handle an IRQ outside of the interrupt handler proper. + * dm_irq_work_func() - Handle an IRQ outside of the interrupt handler proper. * * @work: work struct */ @@ -99,8 +138,8 @@ static void dm_irq_work_func(struct work_struct *work) * (The most common use is HPD interrupt) */ } -/** - * Remove a handler and return a pointer to hander list from which the +/* + * Remove a handler and return a pointer to handler list from which the * handler was removed. */ static struct list_head *remove_irq_handler(struct amdgpu_device *adev, @@ -203,6 +242,24 @@ static bool validate_irq_unregistration_params(enum dc_irq_source irq_source, * Note: caller is responsible for input validation. *****************************************************************************/ +/** + * amdgpu_dm_irq_register_interrupt() - Register a handler within DM. + * @adev: The base driver device containing the DM device. + * @int_params: Interrupt parameters containing the source, and handler context + * @ih: Function pointer to the interrupt handler to register + * @handler_args: Arguments passed to the handler when the interrupt occurs + * + * Register an interrupt handler for the given IRQ source, under the given + * context. The context can either be high or low. High context handlers are + * executed directly within ISR context, while low context is executed within a + * workqueue, thereby allowing operations that sleep. + * + * Registered handlers are called in a FIFO manner, i.e. the most recently + * registered handler will be called first. + * + * Return: Handler data &struct amdgpu_dm_irq_handler_data containing the IRQ + * source, handler function, and args + */ void *amdgpu_dm_irq_register_interrupt(struct amdgpu_device *adev, struct dc_interrupt_params *int_params, void (*ih)(void *), @@ -261,6 +318,15 @@ void *amdgpu_dm_irq_register_interrupt(struct amdgpu_device *adev, return handler_data; } +/** + * amdgpu_dm_irq_unregister_interrupt() - Remove a handler from the DM IRQ table + * @adev: The base driver device containing the DM device + * @irq_source: IRQ source to remove the given handler from + * @ih: Function pointer to the interrupt handler to unregister + * + * Go through both low and high context IRQ tables, and find the given handler + * for the given irq source. If found, remove it. Otherwise, do nothing. + */ void amdgpu_dm_irq_unregister_interrupt(struct amdgpu_device *adev, enum dc_irq_source irq_source, void *ih) @@ -295,6 +361,20 @@ void amdgpu_dm_irq_unregister_interrupt(struct amdgpu_device *adev, } } +/** + * amdgpu_dm_irq_init() - Initialize DM IRQ management + * @adev: The base driver device containing the DM device + * + * Initialize DM's high and low context IRQ tables. + * + * The N by M table contains N IRQ sources, with M + * &struct amdgpu_dm_irq_handler_data hooked together in a linked list. The + * list_heads are initialized here. When an interrupt n is triggered, all m + * handlers are called in sequence, FIFO according to registration order. + * + * The low context table requires special steps to initialize, since handlers + * will be deferred to a workqueue. See &struct irq_list_head. + */ int amdgpu_dm_irq_init(struct amdgpu_device *adev) { int src; @@ -317,7 +397,12 @@ int amdgpu_dm_irq_init(struct amdgpu_device *adev) return 0; } -/* DM IRQ and timer resource release */ +/** + * amdgpu_dm_irq_fini() - Tear down DM IRQ management + * @adev: The base driver device containing the DM device + * + * Flush all work within the low context IRQ table. + */ void amdgpu_dm_irq_fini(struct amdgpu_device *adev) { int src; @@ -414,7 +499,7 @@ int amdgpu_dm_irq_resume_late(struct amdgpu_device *adev) return 0; } -/** +/* * amdgpu_dm_irq_schedule_work - schedule all work items registered for the * "irq_source". */ @@ -439,8 +524,9 @@ static void amdgpu_dm_irq_schedule_work(struct amdgpu_device *adev, } -/** amdgpu_dm_irq_immediate_work - * Callback high irq work immediately, don't send to work queue +/* + * amdgpu_dm_irq_immediate_work + * Callback high irq work immediately, don't send to work queue */ static void amdgpu_dm_irq_immediate_work(struct amdgpu_device *adev, enum dc_irq_source irq_source) @@ -467,11 +553,14 @@ static void amdgpu_dm_irq_immediate_work(struct amdgpu_device *adev, DM_IRQ_TABLE_UNLOCK(adev, irq_table_flags); } -/* - * amdgpu_dm_irq_handler +/** + * amdgpu_dm_irq_handler - Generic DM IRQ handler + * @adev: amdgpu base driver device containing the DM device + * @source: Unused + * @entry: Data about the triggered interrupt * - * Generic IRQ handler, calls all registered high irq work immediately, and - * schedules work for low irq + * Calls all registered high irq work immediately, and schedules work for low + * irq. The DM IRQ table is used to find the corresponding handlers. */ static int amdgpu_dm_irq_handler(struct amdgpu_device *adev, struct amdgpu_irq_src *source, @@ -613,7 +702,7 @@ void amdgpu_dm_set_irq_funcs(struct amdgpu_device *adev) adev->hpd_irq.funcs = &dm_hpd_irq_funcs; } -/* +/** * amdgpu_dm_hpd_init - hpd setup callback. * * @adev: amdgpu_device pointer diff --git a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_mst_types.c b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_mst_types.c index 1b0d209d8367..c4ea3a91f17a 100644 --- a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_mst_types.c +++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_mst_types.c @@ -35,6 +35,8 @@ #include "dc_link_ddc.h" +#include "i2caux_interface.h" + /* #define TRACE_DPCD */ #ifdef TRACE_DPCD @@ -81,80 +83,24 @@ static ssize_t dm_dp_aux_transfer(struct drm_dp_aux *aux, struct drm_dp_aux_msg *msg) { ssize_t result = 0; - enum i2caux_transaction_action action; - enum aux_transaction_type type; + struct aux_payload payload; if (WARN_ON(msg->size > 16)) return -E2BIG; - switch (msg->request & ~DP_AUX_I2C_MOT) { - case DP_AUX_NATIVE_READ: - type = AUX_TRANSACTION_TYPE_DP; - action = I2CAUX_TRANSACTION_ACTION_DP_READ; - - result = dc_link_aux_transfer(TO_DM_AUX(aux)->ddc_service, - msg->address, - &msg->reply, - msg->buffer, - msg->size, - type, - action); - break; - case DP_AUX_NATIVE_WRITE: - type = AUX_TRANSACTION_TYPE_DP; - action = I2CAUX_TRANSACTION_ACTION_DP_WRITE; - - dc_link_aux_transfer(TO_DM_AUX(aux)->ddc_service, - msg->address, - &msg->reply, - msg->buffer, - msg->size, - type, - action); - result = msg->size; - break; - case DP_AUX_I2C_READ: - type = AUX_TRANSACTION_TYPE_I2C; - if (msg->request & DP_AUX_I2C_MOT) - action = I2CAUX_TRANSACTION_ACTION_I2C_READ_MOT; - else - action = I2CAUX_TRANSACTION_ACTION_I2C_READ; - - result = dc_link_aux_transfer(TO_DM_AUX(aux)->ddc_service, - msg->address, - &msg->reply, - msg->buffer, - msg->size, - type, - action); - break; - case DP_AUX_I2C_WRITE: - type = AUX_TRANSACTION_TYPE_I2C; - if (msg->request & DP_AUX_I2C_MOT) - action = I2CAUX_TRANSACTION_ACTION_I2C_WRITE_MOT; - else - action = I2CAUX_TRANSACTION_ACTION_I2C_WRITE; - - dc_link_aux_transfer(TO_DM_AUX(aux)->ddc_service, - msg->address, - &msg->reply, - msg->buffer, - msg->size, - type, - action); - result = msg->size; - break; - default: - return -EINVAL; - } + payload.address = msg->address; + payload.data = msg->buffer; + payload.length = msg->size; + payload.reply = &msg->reply; + payload.i2c_over_aux = (msg->request & DP_AUX_NATIVE_WRITE) == 0; + payload.write = (msg->request & DP_AUX_I2C_READ) == 0; + payload.mot = (msg->request & DP_AUX_I2C_MOT) != 0; + payload.defer_delay = 0; -#ifdef TRACE_DPCD - log_dpcd(msg->request, - msg->address, - msg->buffer, - msg->size, - r == DDC_RESULT_SUCESSFULL); -#endif + result = dc_link_aux_transfer(TO_DM_AUX(aux)->ddc_service, &payload); + + if (payload.write) + result = msg->size; if (result < 0) /* DC doesn't know about kernel error codes */ result = -EIO; @@ -191,6 +137,7 @@ dm_dp_mst_connector_destroy(struct drm_connector *connector) drm_encoder_cleanup(&amdgpu_encoder->base); kfree(amdgpu_encoder); drm_connector_cleanup(connector); + drm_dp_mst_put_port_malloc(amdgpu_dm_connector->port); kfree(amdgpu_dm_connector); } @@ -227,6 +174,11 @@ static int dm_dp_mst_get_modes(struct drm_connector *connector) aconnector->edid = edid; } + if (aconnector->dc_sink && aconnector->dc_sink->sink_signal == SIGNAL_TYPE_VIRTUAL) { + dc_sink_release(aconnector->dc_sink); + aconnector->dc_sink = NULL; + } + if (!aconnector->dc_sink) { struct dc_sink *dc_sink; struct dc_sink_init_data init_params = { @@ -239,6 +191,7 @@ static int dm_dp_mst_get_modes(struct drm_connector *connector) &init_params); dc_sink->priv = aconnector; + /* dc_link_add_remote_sink returns a new reference */ aconnector->dc_sink = dc_sink; if (aconnector->dc_sink) @@ -363,7 +316,9 @@ dm_dp_add_mst_connector(struct drm_dp_mst_topology_mgr *mgr, amdgpu_dm_connector_funcs_reset(connector); DRM_INFO("DM_MST: added connector: %p [id: %d] [master: %p]\n", - aconnector, connector->base.id, aconnector->mst_port); + aconnector, connector->base.id, aconnector->mst_port); + + drm_dp_mst_get_port_malloc(port); DRM_DEBUG_KMS(":%d\n", connector->base.id); @@ -379,12 +334,12 @@ static void dm_dp_destroy_mst_connector(struct drm_dp_mst_topology_mgr *mgr, struct amdgpu_dm_connector *aconnector = to_amdgpu_dm_connector(connector); DRM_INFO("DM_MST: Disabling connector: %p [id: %d] [master: %p]\n", - aconnector, connector->base.id, aconnector->mst_port); + aconnector, connector->base.id, aconnector->mst_port); - aconnector->port = NULL; if (aconnector->dc_sink) { amdgpu_dm_update_freesync_caps(connector, NULL); - dc_link_remove_remote_sink(aconnector->dc_link, aconnector->dc_sink); + dc_link_remove_remote_sink(aconnector->dc_link, + aconnector->dc_sink); dc_sink_release(aconnector->dc_sink); aconnector->dc_sink = NULL; } @@ -395,14 +350,6 @@ static void dm_dp_destroy_mst_connector(struct drm_dp_mst_topology_mgr *mgr, drm_connector_put(connector); } -static void dm_dp_mst_hotplug(struct drm_dp_mst_topology_mgr *mgr) -{ - struct amdgpu_dm_connector *master = container_of(mgr, struct amdgpu_dm_connector, mst_mgr); - struct drm_device *dev = master->base.dev; - - drm_kms_helper_hotplug_event(dev); -} - static void dm_dp_mst_register_connector(struct drm_connector *connector) { struct drm_device *dev = connector->dev; @@ -419,7 +366,6 @@ static void dm_dp_mst_register_connector(struct drm_connector *connector) static const struct drm_dp_mst_topology_cbs dm_mst_cbs = { .add_connector = dm_dp_add_mst_connector, .destroy_connector = dm_dp_destroy_mst_connector, - .hotplug = dm_dp_mst_hotplug, .register_connector = dm_dp_mst_register_connector }; diff --git a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_pp_smu.c b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_pp_smu.c index 12001a006b2d..a114954d6a5b 100644 --- a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_pp_smu.c +++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_pp_smu.c @@ -25,7 +25,7 @@ #include <linux/acpi.h> #include <drm/drmP.h> -#include <drm/drm_crtc_helper.h> +#include <drm/drm_probe_helper.h> #include <drm/amdgpu_drm.h> #include "dm_services.h" #include "amdgpu.h" @@ -485,11 +485,11 @@ void pp_rv_set_display_requirement(struct pp_smu *pp, return; clock.clock_type = amd_pp_dcf_clock; - clock.clock_freq_in_khz = req->hard_min_dcefclk_khz; + clock.clock_freq_in_khz = req->hard_min_dcefclk_mhz * 1000; pp_funcs->display_clock_voltage_request(pp_handle, &clock); clock.clock_type = amd_pp_f_clock; - clock.clock_freq_in_khz = req->hard_min_fclk_khz; + clock.clock_freq_in_khz = req->hard_min_fclk_mhz * 1000; pp_funcs->display_clock_voltage_request(pp_handle, &clock); } @@ -518,13 +518,13 @@ void pp_rv_set_wm_ranges(struct pp_smu *pp, wm_dce_clocks[i].wm_set_id = ranges->reader_wm_sets[i].wm_inst; wm_dce_clocks[i].wm_max_dcfclk_clk_in_khz = - ranges->reader_wm_sets[i].max_drain_clk_khz; + ranges->reader_wm_sets[i].max_drain_clk_mhz * 1000; wm_dce_clocks[i].wm_min_dcfclk_clk_in_khz = - ranges->reader_wm_sets[i].min_drain_clk_khz; + ranges->reader_wm_sets[i].min_drain_clk_mhz * 1000; wm_dce_clocks[i].wm_max_mem_clk_in_khz = - ranges->reader_wm_sets[i].max_fill_clk_khz; + ranges->reader_wm_sets[i].max_fill_clk_mhz * 1000; wm_dce_clocks[i].wm_min_mem_clk_in_khz = - ranges->reader_wm_sets[i].min_fill_clk_khz; + ranges->reader_wm_sets[i].min_fill_clk_mhz * 1000; } for (i = 0; i < wm_with_clock_ranges.num_wm_mcif_sets; i++) { @@ -534,13 +534,13 @@ void pp_rv_set_wm_ranges(struct pp_smu *pp, wm_soc_clocks[i].wm_set_id = ranges->writer_wm_sets[i].wm_inst; wm_soc_clocks[i].wm_max_socclk_clk_in_khz = - ranges->writer_wm_sets[i].max_fill_clk_khz; + ranges->writer_wm_sets[i].max_fill_clk_mhz * 1000; wm_soc_clocks[i].wm_min_socclk_clk_in_khz = - ranges->writer_wm_sets[i].min_fill_clk_khz; + ranges->writer_wm_sets[i].min_fill_clk_mhz * 1000; wm_soc_clocks[i].wm_max_mem_clk_in_khz = - ranges->writer_wm_sets[i].max_drain_clk_khz; + ranges->writer_wm_sets[i].max_drain_clk_mhz * 1000; wm_soc_clocks[i].wm_min_mem_clk_in_khz = - ranges->writer_wm_sets[i].min_drain_clk_khz; + ranges->writer_wm_sets[i].min_drain_clk_mhz * 1000; } pp_funcs->set_watermarks_for_clocks_ranges(pp_handle, &wm_with_clock_ranges); @@ -559,6 +559,58 @@ void pp_rv_set_pme_wa_enable(struct pp_smu *pp) pp_funcs->notify_smu_enable_pwe(pp_handle); } +void pp_rv_set_active_display_count(struct pp_smu *pp, int count) +{ + const struct dc_context *ctx = pp->dm; + struct amdgpu_device *adev = ctx->driver_context; + void *pp_handle = adev->powerplay.pp_handle; + const struct amd_pm_funcs *pp_funcs = adev->powerplay.pp_funcs; + + if (!pp_funcs || !pp_funcs->set_active_display_count) + return; + + pp_funcs->set_active_display_count(pp_handle, count); +} + +void pp_rv_set_min_deep_sleep_dcfclk(struct pp_smu *pp, int clock) +{ + const struct dc_context *ctx = pp->dm; + struct amdgpu_device *adev = ctx->driver_context; + void *pp_handle = adev->powerplay.pp_handle; + const struct amd_pm_funcs *pp_funcs = adev->powerplay.pp_funcs; + + if (!pp_funcs || !pp_funcs->set_min_deep_sleep_dcefclk) + return; + + pp_funcs->set_min_deep_sleep_dcefclk(pp_handle, clock); +} + +void pp_rv_set_hard_min_dcefclk_by_freq(struct pp_smu *pp, int clock) +{ + const struct dc_context *ctx = pp->dm; + struct amdgpu_device *adev = ctx->driver_context; + void *pp_handle = adev->powerplay.pp_handle; + const struct amd_pm_funcs *pp_funcs = adev->powerplay.pp_funcs; + + if (!pp_funcs || !pp_funcs->set_hard_min_dcefclk_by_freq) + return; + + pp_funcs->set_hard_min_dcefclk_by_freq(pp_handle, clock); +} + +void pp_rv_set_hard_min_fclk_by_freq(struct pp_smu *pp, int mhz) +{ + const struct dc_context *ctx = pp->dm; + struct amdgpu_device *adev = ctx->driver_context; + void *pp_handle = adev->powerplay.pp_handle; + const struct amd_pm_funcs *pp_funcs = adev->powerplay.pp_funcs; + + if (!pp_funcs || !pp_funcs->set_hard_min_fclk_by_freq) + return; + + pp_funcs->set_hard_min_fclk_by_freq(pp_handle, mhz); +} + void dm_pp_get_funcs_rv( struct dc_context *ctx, struct pp_smu_funcs_rv *funcs) @@ -567,4 +619,9 @@ void dm_pp_get_funcs_rv( funcs->set_display_requirement = pp_rv_set_display_requirement; funcs->set_wm_ranges = pp_rv_set_wm_ranges; funcs->set_pme_wa_enable = pp_rv_set_pme_wa_enable; + funcs->set_display_count = pp_rv_set_active_display_count; + funcs->set_min_deep_sleep_dcfclk = pp_rv_set_min_deep_sleep_dcfclk; + funcs->set_hard_min_dcfclk_by_freq = pp_rv_set_hard_min_dcefclk_by_freq; + funcs->set_hard_min_fclk_by_freq = pp_rv_set_hard_min_fclk_by_freq; } + diff --git a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_services.c b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_services.c index 516795342dd2..d915e8c8769b 100644 --- a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_services.c +++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_services.c @@ -27,7 +27,7 @@ #include <linux/acpi.h> #include <drm/drmP.h> -#include <drm/drm_crtc_helper.h> +#include <drm/drm_probe_helper.h> #include <drm/amdgpu_drm.h> #include "dm_services.h" #include "amdgpu.h" diff --git a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_trace.h b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_trace.h new file mode 100644 index 000000000000..d898981684d5 --- /dev/null +++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_trace.h @@ -0,0 +1,104 @@ +/* + * Copyright 2018 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. + * + * Authors: AMD + * + */ + +#undef TRACE_SYSTEM +#define TRACE_SYSTEM amdgpu_dm + +#if !defined(_AMDGPU_DM_TRACE_H) || defined(TRACE_HEADER_MULTI_READ) +#define _AMDGPU_DM_TRACE_H_ + +#include <linux/tracepoint.h> + +TRACE_EVENT(amdgpu_dc_rreg, + TP_PROTO(unsigned long *read_count, uint32_t reg, uint32_t value), + TP_ARGS(read_count, reg, value), + TP_STRUCT__entry( + __field(uint32_t, reg) + __field(uint32_t, value) + ), + TP_fast_assign( + __entry->reg = reg; + __entry->value = value; + *read_count = *read_count + 1; + ), + TP_printk("reg=0x%08lx, value=0x%08lx", + (unsigned long)__entry->reg, + (unsigned long)__entry->value) +); + +TRACE_EVENT(amdgpu_dc_wreg, + TP_PROTO(unsigned long *write_count, uint32_t reg, uint32_t value), + TP_ARGS(write_count, reg, value), + TP_STRUCT__entry( + __field(uint32_t, reg) + __field(uint32_t, value) + ), + TP_fast_assign( + __entry->reg = reg; + __entry->value = value; + *write_count = *write_count + 1; + ), + TP_printk("reg=0x%08lx, value=0x%08lx", + (unsigned long)__entry->reg, + (unsigned long)__entry->value) +); + + +TRACE_EVENT(amdgpu_dc_performance, + TP_PROTO(unsigned long read_count, unsigned long write_count, + unsigned long *last_read, unsigned long *last_write, + const char *func, unsigned int line), + TP_ARGS(read_count, write_count, last_read, last_write, func, line), + TP_STRUCT__entry( + __field(uint32_t, reads) + __field(uint32_t, writes) + __field(uint32_t, read_delta) + __field(uint32_t, write_delta) + __string(func, func) + __field(uint32_t, line) + ), + TP_fast_assign( + __entry->reads = read_count; + __entry->writes = write_count; + __entry->read_delta = read_count - *last_read; + __entry->write_delta = write_count - *last_write; + __assign_str(func, func); + __entry->line = line; + *last_read = read_count; + *last_write = write_count; + ), + TP_printk("%s:%d reads=%08ld (%08ld total), writes=%08ld (%08ld total)", + __get_str(func), __entry->line, + (unsigned long)__entry->read_delta, + (unsigned long)__entry->reads, + (unsigned long)__entry->write_delta, + (unsigned long)__entry->writes) +); +#endif /* _AMDGPU_DM_TRACE_H_ */ + +#undef TRACE_INCLUDE_PATH +#define TRACE_INCLUDE_PATH . +#define TRACE_INCLUDE_FILE amdgpu_dm_trace +#include <trace/define_trace.h> diff --git a/drivers/gpu/drm/amd/display/dc/Makefile b/drivers/gpu/drm/amd/display/dc/Makefile index aed538a4d1ba..b8ddb4acccdb 100644 --- a/drivers/gpu/drm/amd/display/dc/Makefile +++ b/drivers/gpu/drm/amd/display/dc/Makefile @@ -23,7 +23,7 @@ # Makefile for Display Core (dc) component. # -DC_LIBS = basics bios calcs dce gpio i2caux irq virtual +DC_LIBS = basics bios calcs dce gpio irq virtual ifdef CONFIG_DRM_AMD_DC_DCN1_0 DC_LIBS += dcn10 dml @@ -41,7 +41,8 @@ AMD_DC = $(addsuffix /Makefile, $(addprefix $(FULL_AMD_DISPLAY_PATH)/dc/,$(DC_LI include $(AMD_DC) DISPLAY_CORE = dc.o dc_link.o dc_resource.o dc_hw_sequencer.o dc_sink.o \ -dc_surface.o dc_link_hwss.o dc_link_dp.o dc_link_ddc.o dc_debug.o dc_stream.o +dc_surface.o dc_link_hwss.o dc_link_dp.o dc_link_ddc.o dc_debug.o dc_stream.o \ +dc_vm_helper.o AMD_DISPLAY_CORE = $(addprefix $(AMDDALPATH)/dc/core/,$(DISPLAY_CORE)) diff --git a/drivers/gpu/drm/amd/display/dc/bios/bios_parser.c b/drivers/gpu/drm/amd/display/dc/bios/bios_parser.c index 0e1dc1b1a48d..a4c97d32e751 100644 --- a/drivers/gpu/drm/amd/display/dc/bios/bios_parser.c +++ b/drivers/gpu/drm/amd/display/dc/bios/bios_parser.c @@ -835,18 +835,6 @@ static enum bp_result bios_parser_enable_crtc( return bp->cmd_tbl.enable_crtc(bp, id, enable); } -static enum bp_result bios_parser_crtc_source_select( - struct dc_bios *dcb, - struct bp_crtc_source_select *bp_params) -{ - struct bios_parser *bp = BP_FROM_DCB(dcb); - - if (!bp->cmd_tbl.select_crtc_source) - return BP_RESULT_FAILURE; - - return bp->cmd_tbl.select_crtc_source(bp, bp_params); -} - static enum bp_result bios_parser_enable_disp_power_gating( struct dc_bios *dcb, enum controller_id controller_id, @@ -2030,7 +2018,7 @@ static uint32_t get_src_obj_list(struct bios_parser *bp, ATOM_OBJECT *object, static struct device_id device_type_from_device_id(uint16_t device_id) { - struct device_id result_device_id; + struct device_id result_device_id = {0}; switch (device_id) { case ATOM_DEVICE_LCD1_SUPPORT: @@ -2842,8 +2830,6 @@ static const struct dc_vbios_funcs vbios_funcs = { .program_crtc_timing = bios_parser_program_crtc_timing, /* still use. should probably retire and program directly */ - .crtc_source_select = bios_parser_crtc_source_select, /* still use. should probably retire and program directly */ - .program_display_engine_pll = bios_parser_program_display_engine_pll, .enable_disp_power_gating = bios_parser_enable_disp_power_gating, diff --git a/drivers/gpu/drm/amd/display/dc/bios/bios_parser2.c b/drivers/gpu/drm/amd/display/dc/bios/bios_parser2.c index ff764da21b6f..fd5266a58297 100644 --- a/drivers/gpu/drm/amd/display/dc/bios/bios_parser2.c +++ b/drivers/gpu/drm/amd/display/dc/bios/bios_parser2.c @@ -265,6 +265,7 @@ static struct atom_display_object_path_v2 *get_bios_object( && id.enum_id == obj_id.enum_id) return &bp->object_info_tbl.v1_4->display_path[i]; } + /* fall through */ case OBJECT_TYPE_CONNECTOR: case OBJECT_TYPE_GENERIC: /* Both Generic and Connector Object ID @@ -277,6 +278,7 @@ static struct atom_display_object_path_v2 *get_bios_object( && id.enum_id == obj_id.enum_id) return &bp->object_info_tbl.v1_4->display_path[i]; } + /* fall through */ default: return NULL; } @@ -638,6 +640,7 @@ static enum bp_result get_ss_info_v4_1( { enum bp_result result = BP_RESULT_OK; struct atom_display_controller_info_v4_1 *disp_cntl_tbl = NULL; + struct atom_smu_info_v3_3 *smu_info = NULL; if (!ss_info) return BP_RESULT_BADINPUT; @@ -650,6 +653,7 @@ static enum bp_result get_ss_info_v4_1( if (!disp_cntl_tbl) return BP_RESULT_BADBIOSTABLE; + ss_info->type.STEP_AND_DELAY_INFO = false; ss_info->spread_percentage_divider = 1000; /* BIOS no longer uses target clock. Always enable for now */ @@ -688,6 +692,19 @@ static enum bp_result get_ss_info_v4_1( */ result = BP_RESULT_UNSUPPORTED; break; + case AS_SIGNAL_TYPE_XGMI: + smu_info = GET_IMAGE(struct atom_smu_info_v3_3, + DATA_TABLES(smu_info)); + if (!smu_info) + return BP_RESULT_BADBIOSTABLE; + + ss_info->spread_spectrum_percentage = + smu_info->waflclk_ss_percentage; + ss_info->spread_spectrum_range = + smu_info->gpuclk_ss_rate_10hz * 10; + if (smu_info->waflclk_ss_mode & ATOM_SS_CENTRE_SPREAD_MODE) + ss_info->type.CENTER_MODE = true; + break; default: result = BP_RESULT_UNSUPPORTED; } @@ -1068,18 +1085,6 @@ static enum bp_result bios_parser_enable_crtc( return bp->cmd_tbl.enable_crtc(bp, id, enable); } -static enum bp_result bios_parser_crtc_source_select( - struct dc_bios *dcb, - struct bp_crtc_source_select *bp_params) -{ - struct bios_parser *bp = BP_FROM_DCB(dcb); - - if (!bp->cmd_tbl.select_crtc_source) - return BP_RESULT_FAILURE; - - return bp->cmd_tbl.select_crtc_source(bp, bp_params); -} - static enum bp_result bios_parser_enable_disp_power_gating( struct dc_bios *dcb, enum controller_id controller_id, @@ -1900,8 +1905,6 @@ static const struct dc_vbios_funcs vbios_funcs = { .program_crtc_timing = bios_parser_program_crtc_timing, - .crtc_source_select = bios_parser_crtc_source_select, - .enable_disp_power_gating = bios_parser_enable_disp_power_gating, .bios_parser_destroy = firmware_parser_destroy, diff --git a/drivers/gpu/drm/amd/display/dc/bios/bios_parser_helper.c b/drivers/gpu/drm/amd/display/dc/bios/bios_parser_helper.c index d4589470985c..fce46ab54c54 100644 --- a/drivers/gpu/drm/amd/display/dc/bios/bios_parser_helper.c +++ b/drivers/gpu/drm/amd/display/dc/bios/bios_parser_helper.c @@ -83,8 +83,7 @@ uint32_t bios_get_vga_enabled_displays( { uint32_t active_disp = 1; - if (bios->regs->BIOS_SCRATCH_3) /*follow up with other asic, todo*/ - active_disp = REG_READ(BIOS_SCRATCH_3) & 0XFFFF; + active_disp = REG_READ(BIOS_SCRATCH_3) & 0XFFFF; return active_disp; } diff --git a/drivers/gpu/drm/amd/display/dc/bios/command_table.c b/drivers/gpu/drm/amd/display/dc/bios/command_table.c index 2bd7cd97e00d..5815983caaf8 100644 --- a/drivers/gpu/drm/amd/display/dc/bios/command_table.c +++ b/drivers/gpu/drm/amd/display/dc/bios/command_table.c @@ -55,7 +55,6 @@ static void init_adjust_display_pll(struct bios_parser *bp); static void init_dac_encoder_control(struct bios_parser *bp); static void init_dac_output_control(struct bios_parser *bp); static void init_set_crtc_timing(struct bios_parser *bp); -static void init_select_crtc_source(struct bios_parser *bp); static void init_enable_crtc(struct bios_parser *bp); static void init_enable_crtc_mem_req(struct bios_parser *bp); static void init_external_encoder_control(struct bios_parser *bp); @@ -73,7 +72,6 @@ void dal_bios_parser_init_cmd_tbl(struct bios_parser *bp) init_dac_encoder_control(bp); init_dac_output_control(bp); init_set_crtc_timing(bp); - init_select_crtc_source(bp); init_enable_crtc(bp); init_enable_crtc_mem_req(bp); init_program_clock(bp); @@ -964,9 +962,9 @@ static enum bp_result set_pixel_clock_v3( allocation.sPCLKInput.ucPostDiv = (uint8_t)bp_params->pixel_clock_post_divider; - /* We need to convert from KHz units into 10KHz units */ + /* We need to convert from 100Hz units into 10KHz units */ allocation.sPCLKInput.usPixelClock = - cpu_to_le16((uint16_t)(bp_params->target_pixel_clock / 10)); + cpu_to_le16((uint16_t)(bp_params->target_pixel_clock_100hz / 100)); params = (PIXEL_CLOCK_PARAMETERS_V3 *)&allocation.sPCLKInput; params->ucTransmitterId = @@ -1042,9 +1040,9 @@ static enum bp_result set_pixel_clock_v5( (uint8_t)bp->cmd_helper->encoder_mode_bp_to_atom( bp_params->signal_type, false); - /* We need to convert from KHz units into 10KHz units */ + /* We need to convert from 100Hz units into 10KHz units */ clk.sPCLKInput.usPixelClock = - cpu_to_le16((uint16_t)(bp_params->target_pixel_clock / 10)); + cpu_to_le16((uint16_t)(bp_params->target_pixel_clock_100hz / 100)); if (bp_params->flags.FORCE_PROGRAMMING_OF_PLL) clk.sPCLKInput.ucMiscInfo |= @@ -1118,9 +1116,9 @@ static enum bp_result set_pixel_clock_v6( (uint8_t) bp->cmd_helper->encoder_mode_bp_to_atom( bp_params->signal_type, false); - /* We need to convert from KHz units into 10KHz units */ + /* We need to convert from 100 Hz units into 10KHz units */ clk.sPCLKInput.ulCrtcPclkFreq.ulPixelClock = - cpu_to_le32(bp_params->target_pixel_clock / 10); + cpu_to_le32(bp_params->target_pixel_clock_100hz / 100); if (bp_params->flags.FORCE_PROGRAMMING_OF_PLL) { clk.sPCLKInput.ucMiscInfo |= @@ -1182,8 +1180,7 @@ static enum bp_result set_pixel_clock_v7( clk.ucTransmitterID = bp->cmd_helper->encoder_id_to_atom(dal_graphics_object_id_get_encoder_id(bp_params->encoder_object_id)); clk.ucEncoderMode = (uint8_t) bp->cmd_helper->encoder_mode_bp_to_atom(bp_params->signal_type, false); - /* We need to convert from KHz units into 10KHz units */ - clk.ulPixelClock = cpu_to_le32(bp_params->target_pixel_clock * 10); + clk.ulPixelClock = cpu_to_le32(bp_params->target_pixel_clock_100hz); clk.ucDeepColorRatio = (uint8_t) bp->cmd_helper->transmitter_color_depth_to_atom(bp_params->color_depth); @@ -1899,120 +1896,6 @@ static enum bp_result set_crtc_using_dtd_timing_v3( /******************************************************************************* ******************************************************************************** ** - ** SELECT CRTC SOURCE - ** - ******************************************************************************** - *******************************************************************************/ - -static enum bp_result select_crtc_source_v2( - struct bios_parser *bp, - struct bp_crtc_source_select *bp_params); -static enum bp_result select_crtc_source_v3( - struct bios_parser *bp, - struct bp_crtc_source_select *bp_params); - -static void init_select_crtc_source(struct bios_parser *bp) -{ - switch (BIOS_CMD_TABLE_PARA_REVISION(SelectCRTC_Source)) { - case 2: - bp->cmd_tbl.select_crtc_source = select_crtc_source_v2; - break; - case 3: - bp->cmd_tbl.select_crtc_source = select_crtc_source_v3; - break; - default: - dm_output_to_console("Don't select_crtc_source enable_crtc for v%d\n", - BIOS_CMD_TABLE_PARA_REVISION(SelectCRTC_Source)); - bp->cmd_tbl.select_crtc_source = NULL; - break; - } -} - -static enum bp_result select_crtc_source_v2( - struct bios_parser *bp, - struct bp_crtc_source_select *bp_params) -{ - enum bp_result result = BP_RESULT_FAILURE; - SELECT_CRTC_SOURCE_PARAMETERS_V2 params; - uint8_t atom_controller_id; - uint32_t atom_engine_id; - enum signal_type s = bp_params->signal; - - memset(¶ms, 0, sizeof(params)); - - /* set controller id */ - if (bp->cmd_helper->controller_id_to_atom( - bp_params->controller_id, &atom_controller_id)) - params.ucCRTC = atom_controller_id; - else - return BP_RESULT_FAILURE; - - /* set encoder id */ - if (bp->cmd_helper->engine_bp_to_atom( - bp_params->engine_id, &atom_engine_id)) - params.ucEncoderID = (uint8_t)atom_engine_id; - else - return BP_RESULT_FAILURE; - - if (SIGNAL_TYPE_EDP == s || - (SIGNAL_TYPE_DISPLAY_PORT == s && - SIGNAL_TYPE_LVDS == bp_params->sink_signal)) - s = SIGNAL_TYPE_LVDS; - - params.ucEncodeMode = - (uint8_t)bp->cmd_helper->encoder_mode_bp_to_atom( - s, bp_params->enable_dp_audio); - - if (EXEC_BIOS_CMD_TABLE(SelectCRTC_Source, params)) - result = BP_RESULT_OK; - - return result; -} - -static enum bp_result select_crtc_source_v3( - struct bios_parser *bp, - struct bp_crtc_source_select *bp_params) -{ - bool result = BP_RESULT_FAILURE; - SELECT_CRTC_SOURCE_PARAMETERS_V3 params; - uint8_t atom_controller_id; - uint32_t atom_engine_id; - enum signal_type s = bp_params->signal; - - memset(¶ms, 0, sizeof(params)); - - if (bp->cmd_helper->controller_id_to_atom(bp_params->controller_id, - &atom_controller_id)) - params.ucCRTC = atom_controller_id; - else - return result; - - if (bp->cmd_helper->engine_bp_to_atom(bp_params->engine_id, - &atom_engine_id)) - params.ucEncoderID = (uint8_t)atom_engine_id; - else - return result; - - if (SIGNAL_TYPE_EDP == s || - (SIGNAL_TYPE_DISPLAY_PORT == s && - SIGNAL_TYPE_LVDS == bp_params->sink_signal)) - s = SIGNAL_TYPE_LVDS; - - params.ucEncodeMode = - bp->cmd_helper->encoder_mode_bp_to_atom( - s, bp_params->enable_dp_audio); - /* Needed for VBIOS Random Spatial Dithering feature */ - params.ucDstBpc = (uint8_t)(bp_params->display_output_bit_depth); - - if (EXEC_BIOS_CMD_TABLE(SelectCRTC_Source, params)) - result = BP_RESULT_OK; - - return result; -} - -/******************************************************************************* - ******************************************************************************** - ** ** ENABLE CRTC ** ******************************************************************************** @@ -2164,7 +2047,7 @@ static enum bp_result program_clock_v5( /* We need to convert from KHz units into 10KHz units */ params.sPCLKInput.ucPpll = (uint8_t) atom_pll_id; params.sPCLKInput.usPixelClock = - cpu_to_le16((uint16_t) (bp_params->target_pixel_clock / 10)); + cpu_to_le16((uint16_t) (bp_params->target_pixel_clock_100hz / 100)); params.sPCLKInput.ucCRTC = (uint8_t) ATOM_CRTC_INVALID; if (bp_params->flags.SET_EXTERNAL_REF_DIV_SRC) @@ -2196,7 +2079,7 @@ static enum bp_result program_clock_v6( /* We need to convert from KHz units into 10KHz units */ params.sPCLKInput.ucPpll = (uint8_t)atom_pll_id; params.sPCLKInput.ulDispEngClkFreq = - cpu_to_le32(bp_params->target_pixel_clock / 10); + cpu_to_le32(bp_params->target_pixel_clock_100hz / 100); if (bp_params->flags.SET_EXTERNAL_REF_DIV_SRC) params.sPCLKInput.ucMiscInfo |= PIXEL_CLOCK_MISC_REF_DIV_SRC; diff --git a/drivers/gpu/drm/amd/display/dc/bios/command_table.h b/drivers/gpu/drm/amd/display/dc/bios/command_table.h index 94f3d43a7471..ad533775e724 100644 --- a/drivers/gpu/drm/amd/display/dc/bios/command_table.h +++ b/drivers/gpu/drm/amd/display/dc/bios/command_table.h @@ -71,9 +71,6 @@ struct cmd_tbl { enum bp_result (*set_crtc_timing)( struct bios_parser *bp, struct bp_hw_crtc_timing_parameters *bp_params); - enum bp_result (*select_crtc_source)( - struct bios_parser *bp, - struct bp_crtc_source_select *bp_params); enum bp_result (*enable_crtc)( struct bios_parser *bp, enum controller_id controller_id, diff --git a/drivers/gpu/drm/amd/display/dc/bios/command_table2.c b/drivers/gpu/drm/amd/display/dc/bios/command_table2.c index 2b5dc499a35e..bb2e8105e6ab 100644 --- a/drivers/gpu/drm/amd/display/dc/bios/command_table2.c +++ b/drivers/gpu/drm/amd/display/dc/bios/command_table2.c @@ -301,17 +301,17 @@ static enum bp_result set_pixel_clock_v7( cmd_helper->encoder_mode_bp_to_atom( bp_params->signal_type, false); - /* We need to convert from KHz units into 10KHz units */ - clk.pixclk_100hz = cpu_to_le32(bp_params->target_pixel_clock * - 10); + clk.pixclk_100hz = cpu_to_le32(bp_params->target_pixel_clock_100hz); clk.deep_color_ratio = (uint8_t) bp->cmd_helper-> transmitter_color_depth_to_atom( bp_params->color_depth); - DC_LOG_BIOS("%s:program display clock = %d"\ - "colorDepth = %d\n", __func__,\ - bp_params->target_pixel_clock, bp_params->color_depth); + + DC_LOG_BIOS("%s:program display clock = %d, tg = %d, pll = %d, "\ + "colorDepth = %d\n", __func__, + bp_params->target_pixel_clock_100hz, (int)controller_id, + pll_id, bp_params->color_depth); if (bp_params->flags.FORCE_PROGRAMMING_OF_PLL) clk.miscinfo |= PIXEL_CLOCK_V7_MISC_FORCE_PROG_PPLL; @@ -463,75 +463,6 @@ static enum bp_result set_crtc_using_dtd_timing_v3( /****************************************************************************** ****************************************************************************** ** - ** SELECT CRTC SOURCE - ** - ****************************************************************************** - *****************************************************************************/ - - -static enum bp_result select_crtc_source_v3( - struct bios_parser *bp, - struct bp_crtc_source_select *bp_params); - -static void init_select_crtc_source(struct bios_parser *bp) -{ - switch (BIOS_CMD_TABLE_PARA_REVISION(selectcrtc_source)) { - case 3: - bp->cmd_tbl.select_crtc_source = select_crtc_source_v3; - break; - default: - dm_output_to_console("Don't select_crtc_source enable_crtc for v%d\n", - BIOS_CMD_TABLE_PARA_REVISION(selectcrtc_source)); - bp->cmd_tbl.select_crtc_source = NULL; - break; - } -} - - -static enum bp_result select_crtc_source_v3( - struct bios_parser *bp, - struct bp_crtc_source_select *bp_params) -{ - bool result = BP_RESULT_FAILURE; - struct select_crtc_source_parameters_v2_3 params; - uint8_t atom_controller_id; - uint32_t atom_engine_id; - enum signal_type s = bp_params->signal; - - memset(¶ms, 0, sizeof(params)); - - if (bp->cmd_helper->controller_id_to_atom(bp_params->controller_id, - &atom_controller_id)) - params.crtc_id = atom_controller_id; - else - return result; - - if (bp->cmd_helper->engine_bp_to_atom(bp_params->engine_id, - &atom_engine_id)) - params.encoder_id = (uint8_t)atom_engine_id; - else - return result; - - if (s == SIGNAL_TYPE_EDP || - (s == SIGNAL_TYPE_DISPLAY_PORT && bp_params->sink_signal == - SIGNAL_TYPE_LVDS)) - s = SIGNAL_TYPE_LVDS; - - params.encode_mode = - bp->cmd_helper->encoder_mode_bp_to_atom( - s, bp_params->enable_dp_audio); - /* Needed for VBIOS Random Spatial Dithering feature */ - params.dst_bpc = (uint8_t)(bp_params->display_output_bit_depth); - - if (EXEC_BIOS_CMD_TABLE(selectcrtc_source, params)) - result = BP_RESULT_OK; - - return result; -} - -/****************************************************************************** - ****************************************************************************** - ** ** ENABLE CRTC ** ****************************************************************************** @@ -808,7 +739,6 @@ void dal_firmware_parser_init_cmd_tbl(struct bios_parser *bp) init_set_crtc_timing(bp); - init_select_crtc_source(bp); init_enable_crtc(bp); init_external_encoder_control(bp); diff --git a/drivers/gpu/drm/amd/display/dc/bios/command_table2.h b/drivers/gpu/drm/amd/display/dc/bios/command_table2.h index ec1c0c9f3f1d..7a2af24dfe60 100644 --- a/drivers/gpu/drm/amd/display/dc/bios/command_table2.h +++ b/drivers/gpu/drm/amd/display/dc/bios/command_table2.h @@ -71,9 +71,6 @@ struct cmd_tbl { enum bp_result (*set_crtc_timing)( struct bios_parser *bp, struct bp_hw_crtc_timing_parameters *bp_params); - enum bp_result (*select_crtc_source)( - struct bios_parser *bp, - struct bp_crtc_source_select *bp_params); enum bp_result (*enable_crtc)( struct bios_parser *bp, enum controller_id controller_id, diff --git a/drivers/gpu/drm/amd/display/dc/bios/command_table_helper2.c b/drivers/gpu/drm/amd/display/dc/bios/command_table_helper2.c index 65b006ad372e..8196f3bb10c7 100644 --- a/drivers/gpu/drm/amd/display/dc/bios/command_table_helper2.c +++ b/drivers/gpu/drm/amd/display/dc/bios/command_table_helper2.c @@ -67,6 +67,7 @@ bool dal_bios_parser_init_cmd_tbl_helper2( return true; #endif case DCE_VERSION_12_0: + case DCE_VERSION_12_1: *h = dal_cmd_tbl_helper_dce112_get_table2(); return true; diff --git a/drivers/gpu/drm/amd/display/dc/calcs/dce_calcs.c b/drivers/gpu/drm/amd/display/dc/calcs/dce_calcs.c index 9ebe30ba4dab..f3aa7b53d2aa 100644 --- a/drivers/gpu/drm/amd/display/dc/calcs/dce_calcs.c +++ b/drivers/gpu/drm/amd/display/dc/calcs/dce_calcs.c @@ -2792,7 +2792,7 @@ static void populate_initial_data( data->lpt_en[num_displays + 4] = false; data->h_total[num_displays + 4] = bw_int_to_fixed(pipe[i].stream->timing.h_total); data->v_total[num_displays + 4] = bw_int_to_fixed(pipe[i].stream->timing.v_total); - data->pixel_rate[num_displays + 4] = bw_frc_to_fixed(pipe[i].stream->timing.pix_clk_khz, 1000); + data->pixel_rate[num_displays + 4] = bw_frc_to_fixed(pipe[i].stream->timing.pix_clk_100hz, 10000); data->src_width[num_displays + 4] = bw_int_to_fixed(pipe[i].plane_res.scl_data.viewport.width); data->pitch_in_pixels[num_displays + 4] = data->src_width[num_displays + 4]; data->src_height[num_displays + 4] = bw_int_to_fixed(pipe[i].plane_res.scl_data.viewport.height); @@ -2881,7 +2881,7 @@ static void populate_initial_data( /* Pipes without underlay after */ for (i = 0; i < pipe_count; i++) { - unsigned int pixel_clock_khz; + unsigned int pixel_clock_100hz; if (!pipe[i].stream || pipe[i].bottom_pipe) continue; @@ -2890,10 +2890,10 @@ static void populate_initial_data( data->lpt_en[num_displays + 4] = false; data->h_total[num_displays + 4] = bw_int_to_fixed(pipe[i].stream->timing.h_total); data->v_total[num_displays + 4] = bw_int_to_fixed(pipe[i].stream->timing.v_total); - pixel_clock_khz = pipe[i].stream->timing.pix_clk_khz; + pixel_clock_100hz = pipe[i].stream->timing.pix_clk_100hz; if (pipe[i].stream->timing.timing_3d_format == TIMING_3D_FORMAT_HW_FRAME_PACKING) - pixel_clock_khz *= 2; - data->pixel_rate[num_displays + 4] = bw_frc_to_fixed(pixel_clock_khz, 1000); + pixel_clock_100hz *= 2; + data->pixel_rate[num_displays + 4] = bw_frc_to_fixed(pixel_clock_100hz, 10000); if (pipe[i].plane_state) { data->src_width[num_displays + 4] = bw_int_to_fixed(pipe[i].plane_res.scl_data.viewport.width); data->pitch_in_pixels[num_displays + 4] = data->src_width[num_displays + 4]; diff --git a/drivers/gpu/drm/amd/display/dc/calcs/dcn_calc_auto.c b/drivers/gpu/drm/amd/display/dc/calcs/dcn_calc_auto.c index d0fc54f8fb1c..1ef0074302c5 100644 --- a/drivers/gpu/drm/amd/display/dc/calcs/dcn_calc_auto.c +++ b/drivers/gpu/drm/amd/display/dc/calcs/dcn_calc_auto.c @@ -63,7 +63,7 @@ void scaler_settings_calculation(struct dcn_bw_internal_vars *v) if (v->interlace_output[k] == 1.0) { v->v_ratio[k] = 2.0 * v->v_ratio[k]; } - if ((v->underscan_output[k] == 1.0)) { + if (v->underscan_output[k] == 1.0) { v->h_ratio[k] = v->h_ratio[k] * v->under_scan_factor; v->v_ratio[k] = v->v_ratio[k] * v->under_scan_factor; } @@ -797,9 +797,40 @@ void mode_support_and_system_configuration(struct dcn_bw_internal_vars *v) else { v->maximum_vstartup = v->v_sync_plus_back_porch[k] - 1.0; } - v->line_times_for_prefetch[k] = v->maximum_vstartup - v->urgent_latency / (v->htotal[k] / v->pixel_clock[k]) - (v->time_calc + v->time_setup) / (v->htotal[k] / v->pixel_clock[k]) - (v->dst_y_after_scaler + v->dst_x_after_scaler / v->htotal[k]); - v->line_times_for_prefetch[k] =dcn_bw_floor2(4.0 * (v->line_times_for_prefetch[k] + 0.125), 1.0) / 4; - v->prefetch_bw[k] = (v->meta_pte_bytes_per_frame[k] + 2.0 * v->meta_row_bytes[k] + 2.0 * v->dpte_bytes_per_row[k] + v->prefetch_lines_y[k] * v->swath_width_yper_state[i][j][k] *dcn_bw_ceil2(v->byte_per_pixel_in_dety[k], 1.0) + v->prefetch_lines_c[k] * v->swath_width_yper_state[i][j][k] / 2.0 *dcn_bw_ceil2(v->byte_per_pixel_in_detc[k], 2.0)) / (v->line_times_for_prefetch[k] * v->htotal[k] / v->pixel_clock[k]); + + do { + v->line_times_for_prefetch[k] = v->maximum_vstartup - v->urgent_latency / (v->htotal[k] / v->pixel_clock[k]) - (v->time_calc + v->time_setup) / (v->htotal[k] / v->pixel_clock[k]) - (v->dst_y_after_scaler + v->dst_x_after_scaler / v->htotal[k]); + v->line_times_for_prefetch[k] =dcn_bw_floor2(4.0 * (v->line_times_for_prefetch[k] + 0.125), 1.0) / 4; + v->prefetch_bw[k] = (v->meta_pte_bytes_per_frame[k] + 2.0 * v->meta_row_bytes[k] + 2.0 * v->dpte_bytes_per_row[k] + v->prefetch_lines_y[k] * v->swath_width_yper_state[i][j][k] *dcn_bw_ceil2(v->byte_per_pixel_in_dety[k], 1.0) + v->prefetch_lines_c[k] * v->swath_width_yper_state[i][j][k] / 2.0 *dcn_bw_ceil2(v->byte_per_pixel_in_detc[k], 2.0)) / (v->line_times_for_prefetch[k] * v->htotal[k] / v->pixel_clock[k]); + + if (v->pte_enable == dcn_bw_yes && v->dcc_enable[k] == dcn_bw_yes) { + v->time_for_meta_pte_without_immediate_flip = dcn_bw_max3( + v->meta_pte_bytes_frame[k] / v->prefetch_bandwidth[k], + v->extra_latency, + v->htotal[k] / v->pixel_clock[k] / 4.0); + } else { + v->time_for_meta_pte_without_immediate_flip = v->htotal[k] / v->pixel_clock[k] / 4.0; + } + + if (v->pte_enable == dcn_bw_yes || v->dcc_enable[k] == dcn_bw_yes) { + v->time_for_meta_and_dpte_row_without_immediate_flip = dcn_bw_max3(( + v->meta_row_bytes[k] + v->dpte_bytes_per_row[k]) / v->prefetch_bandwidth[k], + v->htotal[k] / v->pixel_clock[k] - v->time_for_meta_pte_without_immediate_flip, + v->extra_latency); + } else { + v->time_for_meta_and_dpte_row_without_immediate_flip = dcn_bw_max2( + v->htotal[k] / v->pixel_clock[k] - v->time_for_meta_pte_without_immediate_flip, + v->extra_latency - v->time_for_meta_pte_with_immediate_flip); + } + + v->lines_for_meta_pte_without_immediate_flip[k] =dcn_bw_floor2(4.0 * (v->time_for_meta_pte_without_immediate_flip / (v->htotal[k] / v->pixel_clock[k]) + 0.125), 1.0) / 4; + v->lines_for_meta_and_dpte_row_without_immediate_flip[k] =dcn_bw_floor2(4.0 * (v->time_for_meta_and_dpte_row_without_immediate_flip / (v->htotal[k] / v->pixel_clock[k]) + 0.125), 1.0) / 4; + v->maximum_vstartup = v->maximum_vstartup - 1; + + if (v->lines_for_meta_pte_without_immediate_flip[k] < 8.0 && v->lines_for_meta_and_dpte_row_without_immediate_flip[k] < 16.0) + break; + + } while(1); } v->bw_available_for_immediate_flip = v->return_bw_per_state[i]; for (k = 0; k <= v->number_of_active_planes - 1; k++) { @@ -814,24 +845,18 @@ void mode_support_and_system_configuration(struct dcn_bw_internal_vars *v) for (k = 0; k <= v->number_of_active_planes - 1; k++) { if (v->pte_enable == dcn_bw_yes && v->dcc_enable[k] == dcn_bw_yes) { v->time_for_meta_pte_with_immediate_flip =dcn_bw_max5(v->meta_pte_bytes_per_frame[k] / v->prefetch_bw[k], v->meta_pte_bytes_per_frame[k] * v->total_immediate_flip_bytes[k] / (v->bw_available_for_immediate_flip * (v->meta_pte_bytes_per_frame[k] + v->meta_row_bytes[k] + v->dpte_bytes_per_row[k])), v->extra_latency, v->urgent_latency, v->htotal[k] / v->pixel_clock[k] / 4.0); - v->time_for_meta_pte_without_immediate_flip =dcn_bw_max3(v->meta_pte_bytes_per_frame[k] / v->prefetch_bw[k], v->extra_latency, v->htotal[k] / v->pixel_clock[k] / 4.0); } else { v->time_for_meta_pte_with_immediate_flip = v->htotal[k] / v->pixel_clock[k] / 4.0; - v->time_for_meta_pte_without_immediate_flip = v->htotal[k] / v->pixel_clock[k] / 4.0; } if (v->pte_enable == dcn_bw_yes || v->dcc_enable[k] == dcn_bw_yes) { v->time_for_meta_and_dpte_row_with_immediate_flip =dcn_bw_max5((v->meta_row_bytes[k] + v->dpte_bytes_per_row[k]) / v->prefetch_bw[k], (v->meta_row_bytes[k] + v->dpte_bytes_per_row[k]) * v->total_immediate_flip_bytes[k] / (v->bw_available_for_immediate_flip * (v->meta_pte_bytes_per_frame[k] + v->meta_row_bytes[k] + v->dpte_bytes_per_row[k])), v->htotal[k] / v->pixel_clock[k] - v->time_for_meta_pte_with_immediate_flip, v->extra_latency, 2.0 * v->urgent_latency); - v->time_for_meta_and_dpte_row_without_immediate_flip =dcn_bw_max3((v->meta_row_bytes[k] + v->dpte_bytes_per_row[k]) / v->prefetch_bw[k], v->htotal[k] / v->pixel_clock[k] - v->time_for_meta_pte_without_immediate_flip, v->extra_latency); } else { v->time_for_meta_and_dpte_row_with_immediate_flip =dcn_bw_max2(v->htotal[k] / v->pixel_clock[k] - v->time_for_meta_pte_with_immediate_flip, v->extra_latency - v->time_for_meta_pte_with_immediate_flip); - v->time_for_meta_and_dpte_row_without_immediate_flip =dcn_bw_max2(v->htotal[k] / v->pixel_clock[k] - v->time_for_meta_pte_without_immediate_flip, v->extra_latency - v->time_for_meta_pte_without_immediate_flip); } v->lines_for_meta_pte_with_immediate_flip[k] =dcn_bw_floor2(4.0 * (v->time_for_meta_pte_with_immediate_flip / (v->htotal[k] / v->pixel_clock[k]) + 0.125), 1.0) / 4; - v->lines_for_meta_pte_without_immediate_flip[k] =dcn_bw_floor2(4.0 * (v->time_for_meta_pte_without_immediate_flip / (v->htotal[k] / v->pixel_clock[k]) + 0.125), 1.0) / 4; v->lines_for_meta_and_dpte_row_with_immediate_flip[k] =dcn_bw_floor2(4.0 * (v->time_for_meta_and_dpte_row_with_immediate_flip / (v->htotal[k] / v->pixel_clock[k]) + 0.125), 1.0) / 4; - v->lines_for_meta_and_dpte_row_without_immediate_flip[k] =dcn_bw_floor2(4.0 * (v->time_for_meta_and_dpte_row_without_immediate_flip / (v->htotal[k] / v->pixel_clock[k]) + 0.125), 1.0) / 4; v->line_times_to_request_prefetch_pixel_data_with_immediate_flip = v->line_times_for_prefetch[k] - v->lines_for_meta_pte_with_immediate_flip[k] - v->lines_for_meta_and_dpte_row_with_immediate_flip[k]; v->line_times_to_request_prefetch_pixel_data_without_immediate_flip = v->line_times_for_prefetch[k] - v->lines_for_meta_pte_without_immediate_flip[k] - v->lines_for_meta_and_dpte_row_without_immediate_flip[k]; if (v->line_times_to_request_prefetch_pixel_data_with_immediate_flip > 0.0) { diff --git a/drivers/gpu/drm/amd/display/dc/calcs/dcn_calcs.c b/drivers/gpu/drm/amd/display/dc/calcs/dcn_calcs.c index 3208188b7ed4..eb62d10bb65c 100644 --- a/drivers/gpu/drm/amd/display/dc/calcs/dcn_calcs.c +++ b/drivers/gpu/drm/amd/display/dc/calcs/dcn_calcs.c @@ -290,41 +290,34 @@ static void pipe_ctx_to_e2e_pipe_params ( switch (pipe->plane_state->tiling_info.gfx9.swizzle) { /* for 4/8/16 high tiles */ case DC_SW_LINEAR: - input->src.is_display_sw = 1; input->src.macro_tile_size = dm_4k_tile; break; case DC_SW_4KB_S: case DC_SW_4KB_S_X: - input->src.is_display_sw = 0; input->src.macro_tile_size = dm_4k_tile; break; case DC_SW_64KB_S: case DC_SW_64KB_S_X: case DC_SW_64KB_S_T: - input->src.is_display_sw = 0; input->src.macro_tile_size = dm_64k_tile; break; case DC_SW_VAR_S: case DC_SW_VAR_S_X: - input->src.is_display_sw = 0; input->src.macro_tile_size = dm_256k_tile; break; /* For 64bpp 2 high tiles */ case DC_SW_4KB_D: case DC_SW_4KB_D_X: - input->src.is_display_sw = 1; input->src.macro_tile_size = dm_4k_tile; break; case DC_SW_64KB_D: case DC_SW_64KB_D_X: case DC_SW_64KB_D_T: - input->src.is_display_sw = 1; input->src.macro_tile_size = dm_64k_tile; break; case DC_SW_VAR_D: case DC_SW_VAR_D_X: - input->src.is_display_sw = 1; input->src.macro_tile_size = dm_256k_tile; break; @@ -423,7 +416,7 @@ static void pipe_ctx_to_e2e_pipe_params ( - pipe->stream->timing.v_addressable - pipe->stream->timing.v_border_bottom - pipe->stream->timing.v_border_top; - input->dest.pixel_rate_mhz = pipe->stream->timing.pix_clk_khz/1000.0; + input->dest.pixel_rate_mhz = pipe->stream->timing.pix_clk_100hz/10000.0; input->dest.vstartup_start = pipe->pipe_dlg_param.vstartup_start; input->dest.vupdate_offset = pipe->pipe_dlg_param.vupdate_offset; input->dest.vupdate_offset = pipe->pipe_dlg_param.vupdate_offset; @@ -670,9 +663,9 @@ static void hack_disable_optional_pipe_split(struct dcn_bw_internal_vars *v) } static void hack_force_pipe_split(struct dcn_bw_internal_vars *v, - unsigned int pixel_rate_khz) + unsigned int pixel_rate_100hz) { - float pixel_rate_mhz = pixel_rate_khz / 1000; + float pixel_rate_mhz = pixel_rate_100hz / 10000; /* * force enabling pipe split by lower dpp clock for DPM0 to just @@ -695,7 +688,7 @@ static void hack_bounding_box(struct dcn_bw_internal_vars *v, if (context->stream_count == 1 && dbg->force_single_disp_pipe_split) - hack_force_pipe_split(v, context->streams[0]->timing.pix_clk_khz); + hack_force_pipe_split(v, context->streams[0]->timing.pix_clk_100hz); } bool dcn_validate_bandwidth( @@ -852,7 +845,7 @@ bool dcn_validate_bandwidth( v->v_sync_plus_back_porch[input_idx] = pipe->stream->timing.v_total - v->vactive[input_idx] - pipe->stream->timing.v_front_porch; - v->pixel_clock[input_idx] = pipe->stream->timing.pix_clk_khz/1000.0; + v->pixel_clock[input_idx] = pipe->stream->timing.pix_clk_100hz/10000.0; if (pipe->stream->timing.timing_3d_format == TIMING_3D_FORMAT_HW_FRAME_PACKING) v->pixel_clock[input_idx] *= 2; if (!pipe->plane_state) { @@ -961,7 +954,7 @@ bool dcn_validate_bandwidth( v->dcc_rate[input_idx] = 1; /*TODO: Worst case? does this change?*/ v->output_format[input_idx] = pipe->stream->timing.pixel_encoding == PIXEL_ENCODING_YCBCR420 ? dcn_bw_420 : dcn_bw_444; - v->output[input_idx] = pipe->stream->sink->sink_signal == + v->output[input_idx] = pipe->stream->signal == SIGNAL_TYPE_HDMI_TYPE_A ? dcn_bw_hdmi : dcn_bw_dp; v->output_deep_color[input_idx] = dcn_bw_encoder_8bpc; if (v->output[input_idx] == dcn_bw_hdmi) { @@ -1355,12 +1348,12 @@ void dcn_bw_update_from_pplib(struct dc *dc) struct dm_pp_clock_levels_with_voltage fclks = {0}, dcfclks = {0}; bool res; - kernel_fpu_begin(); - /* TODO: This is not the proper way to obtain fabric_and_dram_bandwidth, should be min(fclk, memclk) */ res = dm_pp_get_clock_levels_by_type_with_voltage( ctx, DM_PP_CLOCK_TYPE_FCLK, &fclks); + kernel_fpu_begin(); + if (res) res = verify_clock_values(&fclks); @@ -1379,9 +1372,13 @@ void dcn_bw_update_from_pplib(struct dc *dc) } else BREAK_TO_DEBUGGER(); + kernel_fpu_end(); + res = dm_pp_get_clock_levels_by_type_with_voltage( ctx, DM_PP_CLOCK_TYPE_DCFCLK, &dcfclks); + kernel_fpu_begin(); + if (res) res = verify_clock_values(&dcfclks); @@ -1423,27 +1420,27 @@ void dcn_bw_notify_pplib_of_wm_ranges(struct dc *dc) ranges.num_reader_wm_sets = WM_SET_COUNT; ranges.num_writer_wm_sets = WM_SET_COUNT; ranges.reader_wm_sets[0].wm_inst = WM_A; - ranges.reader_wm_sets[0].min_drain_clk_khz = min_dcfclk_khz; - ranges.reader_wm_sets[0].max_drain_clk_khz = overdrive; - ranges.reader_wm_sets[0].min_fill_clk_khz = min_fclk_khz; - ranges.reader_wm_sets[0].max_fill_clk_khz = overdrive; + ranges.reader_wm_sets[0].min_drain_clk_mhz = min_dcfclk_khz / 1000; + ranges.reader_wm_sets[0].max_drain_clk_mhz = overdrive / 1000; + ranges.reader_wm_sets[0].min_fill_clk_mhz = min_fclk_khz / 1000; + ranges.reader_wm_sets[0].max_fill_clk_mhz = overdrive / 1000; ranges.writer_wm_sets[0].wm_inst = WM_A; - ranges.writer_wm_sets[0].min_fill_clk_khz = socclk_khz; - ranges.writer_wm_sets[0].max_fill_clk_khz = overdrive; - ranges.writer_wm_sets[0].min_drain_clk_khz = min_fclk_khz; - ranges.writer_wm_sets[0].max_drain_clk_khz = overdrive; + ranges.writer_wm_sets[0].min_fill_clk_mhz = socclk_khz / 1000; + ranges.writer_wm_sets[0].max_fill_clk_mhz = overdrive / 1000; + ranges.writer_wm_sets[0].min_drain_clk_mhz = min_fclk_khz / 1000; + ranges.writer_wm_sets[0].max_drain_clk_mhz = overdrive / 1000; if (dc->debug.pplib_wm_report_mode == WM_REPORT_OVERRIDE) { ranges.reader_wm_sets[0].wm_inst = WM_A; - ranges.reader_wm_sets[0].min_drain_clk_khz = 300000; - ranges.reader_wm_sets[0].max_drain_clk_khz = 5000000; - ranges.reader_wm_sets[0].min_fill_clk_khz = 800000; - ranges.reader_wm_sets[0].max_fill_clk_khz = 5000000; + ranges.reader_wm_sets[0].min_drain_clk_mhz = 300; + ranges.reader_wm_sets[0].max_drain_clk_mhz = 5000; + ranges.reader_wm_sets[0].min_fill_clk_mhz = 800; + ranges.reader_wm_sets[0].max_fill_clk_mhz = 5000; ranges.writer_wm_sets[0].wm_inst = WM_A; - ranges.writer_wm_sets[0].min_fill_clk_khz = 200000; - ranges.writer_wm_sets[0].max_fill_clk_khz = 5000000; - ranges.writer_wm_sets[0].min_drain_clk_khz = 800000; - ranges.writer_wm_sets[0].max_drain_clk_khz = 5000000; + ranges.writer_wm_sets[0].min_fill_clk_mhz = 200; + ranges.writer_wm_sets[0].max_fill_clk_mhz = 5000; + ranges.writer_wm_sets[0].min_drain_clk_mhz = 800; + ranges.writer_wm_sets[0].max_drain_clk_mhz = 5000; } ranges.reader_wm_sets[1] = ranges.writer_wm_sets[0]; diff --git a/drivers/gpu/drm/amd/display/dc/core/dc.c b/drivers/gpu/drm/amd/display/dc/core/dc.c index 7c491c91465f..a6cda201c964 100644 --- a/drivers/gpu/drm/amd/display/dc/core/dc.c +++ b/drivers/gpu/drm/amd/display/dc/core/dc.c @@ -62,6 +62,55 @@ const static char DC_BUILD_ID[] = "production-build"; +/** + * DOC: Overview + * + * DC is the OS-agnostic component of the amdgpu DC driver. + * + * DC maintains and validates a set of structs representing the state of the + * driver and writes that state to AMD hardware + * + * Main DC HW structs: + * + * struct dc - The central struct. One per driver. Created on driver load, + * destroyed on driver unload. + * + * struct dc_context - One per driver. + * Used as a backpointer by most other structs in dc. + * + * struct dc_link - One per connector (the physical DP, HDMI, miniDP, or eDP + * plugpoints). Created on driver load, destroyed on driver unload. + * + * struct dc_sink - One per display. Created on boot or hotplug. + * Destroyed on shutdown or hotunplug. A dc_link can have a local sink + * (the display directly attached). It may also have one or more remote + * sinks (in the Multi-Stream Transport case) + * + * struct resource_pool - One per driver. Represents the hw blocks not in the + * main pipeline. Not directly accessible by dm. + * + * Main dc state structs: + * + * These structs can be created and destroyed as needed. There is a full set of + * these structs in dc->current_state representing the currently programmed state. + * + * struct dc_state - The global DC state to track global state information, + * such as bandwidth values. + * + * struct dc_stream_state - Represents the hw configuration for the pipeline from + * a framebuffer to a display. Maps one-to-one with dc_sink. + * + * struct dc_plane_state - Represents a framebuffer. Each stream has at least one, + * and may have more in the Multi-Plane Overlay case. + * + * struct resource_context - Represents the programmable state of everything in + * the resource_pool. Not directly accessible by dm. + * + * struct pipe_ctx - A member of struct resource_context. Represents the + * internal hardware pipeline components. Each dc_plane_state has either + * one or two (in the pipe-split case). + */ + /******************************************************************************* * Private functions ******************************************************************************/ @@ -102,10 +151,6 @@ static bool create_links( return false; } - if (connectors_num == 0 && num_virtual_links == 0) { - dm_error("DC: Number of connectors is zero!\n"); - } - dm_output_to_console( "DC: %s: connectors_num: physical:%d, virtual:%d\n", __func__, @@ -175,6 +220,17 @@ failed_alloc: return false; } +static struct dc_perf_trace *dc_perf_trace_create(void) +{ + return kzalloc(sizeof(struct dc_perf_trace), GFP_KERNEL); +} + +static void dc_perf_trace_destroy(struct dc_perf_trace **perf_trace) +{ + kfree(*perf_trace); + *perf_trace = NULL; +} + /** ***************************************************************************** * Function: dc_stream_adjust_vmin_vmax @@ -240,7 +296,7 @@ bool dc_stream_get_crtc_position(struct dc *dc, } /** - * dc_stream_configure_crc: Configure CRC capture for the given stream. + * dc_stream_configure_crc() - Configure CRC capture for the given stream. * @dc: DC Object * @stream: The stream to configure CRC on. * @enable: Enable CRC if true, disable otherwise. @@ -292,7 +348,7 @@ bool dc_stream_configure_crc(struct dc *dc, struct dc_stream_state *stream, } /** - * dc_stream_get_crc: Get CRC values for the given stream. + * dc_stream_get_crc() - Get CRC values for the given stream. * @dc: DC object * @stream: The DC stream state of the stream to get CRCs from. * @r_cr, g_y, b_cb: CRC values for the three channels are stored here. @@ -328,7 +384,7 @@ void dc_stream_set_dither_option(struct dc_stream_state *stream, enum dc_dither_option option) { struct bit_depth_reduction_params params; - struct dc_link *link = stream->status.link; + struct dc_link *link = stream->link; struct pipe_ctx *pipes = NULL; int i; @@ -391,9 +447,11 @@ bool dc_stream_program_csc_matrix(struct dc *dc, struct dc_stream_state *stream) == stream) { pipes = &dc->current_state->res_ctx.pipe_ctx[i]; - dc->hwss.program_csc_matrix(pipes, - stream->output_color_space, - stream->csc_color_matrix.matrix); + dc->hwss.program_output_csc(dc, + pipes, + stream->output_color_space, + stream->csc_color_matrix.matrix, + pipes->plane_res.hubp ? pipes->plane_res.hubp->opp_id : 0); ret = true; } } @@ -468,9 +526,8 @@ void dc_link_set_preferred_link_settings(struct dc *dc, for (i = 0; i < MAX_PIPES; i++) { pipe = &dc->current_state->res_ctx.pipe_ctx[i]; - if (pipe->stream && pipe->stream->sink - && pipe->stream->sink->link) { - if (pipe->stream->sink->link == link) + if (pipe->stream && pipe->stream->link) { + if (pipe->stream->link == link) break; } } @@ -528,12 +585,11 @@ static void destruct(struct dc *dc) if (dc->ctx->gpio_service) dal_gpio_service_destroy(&dc->ctx->gpio_service); - if (dc->ctx->i2caux) - dal_i2caux_destroy(&dc->ctx->i2caux); - if (dc->ctx->created_bios) dal_bios_parser_destroy(&dc->ctx->dc_bios); + dc_perf_trace_destroy(&dc->ctx->perf_trace); + kfree(dc->ctx); dc->ctx = NULL; @@ -565,7 +621,6 @@ static bool construct(struct dc *dc, #endif enum dce_version dc_version = DCE_VERSION_UNKNOWN; - dc_dceip = kzalloc(sizeof(*dc_dceip), GFP_KERNEL); if (!dc_dceip) { dm_error("%s: failed to create dceip\n", __func__); @@ -610,6 +665,7 @@ static bool construct(struct dc *dc, dc_ctx->dc = dc; dc_ctx->asic_id = init_params->asic_id; dc_ctx->dc_sink_id_count = 0; + dc_ctx->dc_stream_id_count = 0; dc->ctx = dc_ctx; dc->current_state = dc_create_state(); @@ -649,10 +705,8 @@ static bool construct(struct dc *dc, dc_ctx->created_bios = true; } - /* Create I2C AUX */ - dc_ctx->i2caux = dal_i2caux_create(dc_ctx); - - if (!dc_ctx->i2caux) { + dc_ctx->perf_trace = dc_perf_trace_create(); + if (!dc_ctx->perf_trace) { ASSERT_CRITICAL(false); goto fail; } @@ -774,6 +828,11 @@ alloc_fail: return NULL; } +void dc_init_callbacks(struct dc *dc, + const struct dc_callback_init *init_params) +{ +} + void dc_destroy(struct dc **dc) { destruct(*dc); @@ -809,8 +868,9 @@ static void program_timing_sync( struct dc *dc, struct dc_state *ctx) { - int i, j; + int i, j, k; int group_index = 0; + int num_group = 0; int pipe_count = dc->res_pool->pipe_count; struct pipe_ctx *unsynced_pipes[MAX_PIPES] = { NULL }; @@ -847,11 +907,11 @@ static void program_timing_sync( } } - /* set first unblanked pipe as master */ + /* set first pipe with plane as master */ for (j = 0; j < group_size; j++) { struct pipe_ctx *temp; - if (pipe_set[j]->stream_res.tg->funcs->is_blanked && !pipe_set[j]->stream_res.tg->funcs->is_blanked(pipe_set[j]->stream_res.tg)) { + if (pipe_set[j]->plane_state) { if (j == 0) break; @@ -862,9 +922,21 @@ static void program_timing_sync( } } - /* remove any other unblanked pipes as they have already been synced */ + + for (k = 0; k < group_size; k++) { + struct dc_stream_status *status = dc_stream_get_status_from_state(ctx, pipe_set[k]->stream); + + status->timing_sync_info.group_id = num_group; + status->timing_sync_info.group_size = group_size; + if (k == 0) + status->timing_sync_info.master = true; + else + status->timing_sync_info.master = false; + + } + /* remove any other pipes with plane as they have already been synced */ for (j = j + 1; j < group_size; j++) { - if (pipe_set[j]->stream_res.tg->funcs->is_blanked && !pipe_set[j]->stream_res.tg->funcs->is_blanked(pipe_set[j]->stream_res.tg)) { + if (pipe_set[j]->plane_state) { group_size--; pipe_set[j] = pipe_set[group_size]; j--; @@ -876,6 +948,7 @@ static void program_timing_sync( dc, group_index, group_size, pipe_set); group_index++; } + num_group++; } } @@ -896,6 +969,52 @@ static bool context_changed( return false; } +bool dc_validate_seamless_boot_timing(struct dc *dc, + const struct dc_sink *sink, + struct dc_crtc_timing *crtc_timing) +{ + struct timing_generator *tg; + struct dc_link *link = sink->link; + unsigned int inst; + + /* Check for enabled DIG to identify enabled display */ + if (!link->link_enc->funcs->is_dig_enabled(link->link_enc)) + return false; + + /* Check for which front end is used by this encoder. + * Note the inst is 1 indexed, where 0 is undefined. + * Note that DIG_FE can source from different OTG but our + * current implementation always map 1-to-1, so this code makes + * the same assumption and doesn't check OTG source. + */ + inst = link->link_enc->funcs->get_dig_frontend(link->link_enc) - 1; + + /* Instance should be within the range of the pool */ + if (inst >= dc->res_pool->pipe_count) + return false; + + tg = dc->res_pool->timing_generators[inst]; + + if (!tg->funcs->is_matching_timing) + return false; + + if (!tg->funcs->is_matching_timing(tg, crtc_timing)) + return false; + + if (dc_is_dp_signal(link->connector_signal)) { + unsigned int pix_clk_100hz; + + dc->res_pool->dp_clock_source->funcs->get_pixel_clk_frequency_100hz( + dc->res_pool->dp_clock_source, + inst, &pix_clk_100hz); + + if (crtc_timing->pix_clk_100hz != pix_clk_100hz) + return false; + } + + return true; +} + bool dc_enable_stereo( struct dc *dc, struct dc_state *context, @@ -941,7 +1060,7 @@ static enum dc_status dc_commit_state_no_check(struct dc *dc, struct dc_state *c if (!dcb->funcs->is_accelerated_mode(dcb)) dc->hwss.enable_accelerated_mode(dc, context); - dc->hwss.set_bandwidth(dc, context, false); + dc->hwss.prepare_bandwidth(dc, context); /* re-program planes for existing stream, in case we need to * free up plane resource for later use @@ -957,8 +1076,6 @@ static enum dc_status dc_commit_state_no_check(struct dc *dc, struct dc_state *c } /* Program hardware */ - dc->hwss.ready_shared_resources(dc, context); - for (i = 0; i < dc->res_pool->pipe_count; i++) { pipe = &context->res_ctx.pipe_ctx[i]; dc->hwss.wait_for_mpcc_disconnect(dc, dc->res_pool, pipe); @@ -976,7 +1093,11 @@ static enum dc_status dc_commit_state_no_check(struct dc *dc, struct dc_state *c /* Program all planes within new context*/ for (i = 0; i < context->stream_count; i++) { - const struct dc_sink *sink = context->streams[i]->sink; + const struct dc_link *link = context->streams[i]->link; + struct dc_stream_status *status; + + if (context->streams[i]->apply_seamless_boot_optimization) + context->streams[i]->apply_seamless_boot_optimization = false; if (!context->streams[i]->mode_changed) continue; @@ -1001,18 +1122,24 @@ static enum dc_status dc_commit_state_no_check(struct dc *dc, struct dc_state *c } } - CONN_MSG_MODE(sink->link, "{%dx%d, %dx%d@%dKhz}", + status = dc_stream_get_status_from_state(context, context->streams[i]); + context->streams[i]->out.otg_offset = status->primary_otg_inst; + + CONN_MSG_MODE(link, "{%dx%d, %dx%d@%dKhz}", context->streams[i]->timing.h_addressable, context->streams[i]->timing.v_addressable, context->streams[i]->timing.h_total, context->streams[i]->timing.v_total, - context->streams[i]->timing.pix_clk_khz); + context->streams[i]->timing.pix_clk_100hz / 10); } dc_enable_stereo(dc, context, dc_streams, context->stream_count); /* pplib is notified if disp_num changed */ - dc->hwss.set_bandwidth(dc, context, true); + dc->hwss.optimize_bandwidth(dc, context); + + for (i = 0; i < context->stream_count; i++) + context->streams[i]->mode_changed = false; dc_release_state(dc->current_state); @@ -1020,8 +1147,6 @@ static enum dc_status dc_commit_state_no_check(struct dc *dc, struct dc_state *c dc_retain_state(dc->current_state); - dc->hwss.optimize_shared_resources(dc); - return result; } @@ -1052,6 +1177,9 @@ bool dc_post_update_surfaces_to_stream(struct dc *dc) int i; struct dc_state *context = dc->current_state; + if (dc->optimized_required == false) + return true; + post_surface_trace(dc); for (i = 0; i < dc->res_pool->pipe_count; i++) @@ -1063,7 +1191,7 @@ bool dc_post_update_surfaces_to_stream(struct dc *dc) dc->optimized_required = false; - dc->hwss.set_bandwidth(dc, context, true); + dc->hwss.optimize_bandwidth(dc, context); return true; } @@ -1153,6 +1281,12 @@ static enum surface_update_type get_plane_info_update_type(const struct dc_surfa */ update_flags->bits.bpp_change = 1; + if (u->plane_info->plane_size.grph.surface_pitch != u->surface->plane_size.grph.surface_pitch + || u->plane_info->plane_size.video.luma_pitch != u->surface->plane_size.video.luma_pitch + || u->plane_info->plane_size.video.chroma_pitch != u->surface->plane_size.video.chroma_pitch) + update_flags->bits.plane_size_change = 1; + + if (memcmp(&u->plane_info->tiling_info, &u->surface->tiling_info, sizeof(union dc_tiling_info)) != 0) { update_flags->bits.swizzle_change = 1; @@ -1174,7 +1308,7 @@ static enum surface_update_type get_plane_info_update_type(const struct dc_surfa || update_flags->bits.output_tf_change) return UPDATE_TYPE_FULL; - return UPDATE_TYPE_MED; + return update_flags->raw ? UPDATE_TYPE_MED : UPDATE_TYPE_FAST; } static enum surface_update_type get_scaling_info_update_type( @@ -1243,6 +1377,11 @@ static enum surface_update_type det_surface_update(const struct dc *dc, return UPDATE_TYPE_FULL; } + if (u->surface->force_full_update) { + update_flags->bits.full_update = 1; + return UPDATE_TYPE_FULL; + } + type = get_plane_info_update_type(u); elevate_update_type(&overall_type, type); @@ -1331,6 +1470,11 @@ static enum surface_update_type check_update_surfaces_for_stream( return overall_type; } +/** + * dc_check_update_surfaces_for_stream() - Determine update type (fast, med, or full) + * + * See :c:type:`enum surface_update_type <surface_update_type>` for explanation of update types + */ enum surface_update_type dc_check_update_surfaces_for_stream( struct dc *dc, struct dc_surface_update *updates, @@ -1369,33 +1513,99 @@ static struct dc_stream_status *stream_get_status( static const enum surface_update_type update_surface_trace_level = UPDATE_TYPE_FULL; -static void notify_display_count_to_smu( - struct dc *dc, - struct dc_state *context) +static void copy_surface_update_to_plane( + struct dc_plane_state *surface, + struct dc_surface_update *srf_update) { - int i, display_count; - struct pp_smu_funcs_rv *pp_smu = dc->res_pool->pp_smu; - - /* - * if function pointer not set up, this message is - * sent as part of pplib_apply_display_requirements. - * So just return. - */ - if (!pp_smu || !pp_smu->set_display_count) - return; - - display_count = 0; - for (i = 0; i < context->stream_count; i++) { - const struct dc_stream_state *stream = context->streams[i]; - - /* only notify active stream */ - if (stream->dpms_off) - continue; - - display_count++; - } - - pp_smu->set_display_count(&pp_smu->pp_smu, display_count); + if (srf_update->flip_addr) { + surface->address = srf_update->flip_addr->address; + surface->flip_immediate = + srf_update->flip_addr->flip_immediate; + surface->time.time_elapsed_in_us[surface->time.index] = + srf_update->flip_addr->flip_timestamp_in_us - + surface->time.prev_update_time_in_us; + surface->time.prev_update_time_in_us = + srf_update->flip_addr->flip_timestamp_in_us; + surface->time.index++; + if (surface->time.index >= DC_PLANE_UPDATE_TIMES_MAX) + surface->time.index = 0; + } + + if (srf_update->scaling_info) { + surface->scaling_quality = + srf_update->scaling_info->scaling_quality; + surface->dst_rect = + srf_update->scaling_info->dst_rect; + surface->src_rect = + srf_update->scaling_info->src_rect; + surface->clip_rect = + srf_update->scaling_info->clip_rect; + } + + if (srf_update->plane_info) { + surface->color_space = + srf_update->plane_info->color_space; + surface->format = + srf_update->plane_info->format; + surface->plane_size = + srf_update->plane_info->plane_size; + surface->rotation = + srf_update->plane_info->rotation; + surface->horizontal_mirror = + srf_update->plane_info->horizontal_mirror; + surface->stereo_format = + srf_update->plane_info->stereo_format; + surface->tiling_info = + srf_update->plane_info->tiling_info; + surface->visible = + srf_update->plane_info->visible; + surface->per_pixel_alpha = + srf_update->plane_info->per_pixel_alpha; + surface->global_alpha = + srf_update->plane_info->global_alpha; + surface->global_alpha_value = + srf_update->plane_info->global_alpha_value; + surface->dcc = + srf_update->plane_info->dcc; + surface->sdr_white_level = + srf_update->plane_info->sdr_white_level; + } + + if (srf_update->gamma && + (surface->gamma_correction != + srf_update->gamma)) { + memcpy(&surface->gamma_correction->entries, + &srf_update->gamma->entries, + sizeof(struct dc_gamma_entries)); + surface->gamma_correction->is_identity = + srf_update->gamma->is_identity; + surface->gamma_correction->num_entries = + srf_update->gamma->num_entries; + surface->gamma_correction->type = + srf_update->gamma->type; + } + + if (srf_update->in_transfer_func && + (surface->in_transfer_func != + srf_update->in_transfer_func)) { + surface->in_transfer_func->sdr_ref_white_level = + srf_update->in_transfer_func->sdr_ref_white_level; + surface->in_transfer_func->tf = + srf_update->in_transfer_func->tf; + surface->in_transfer_func->type = + srf_update->in_transfer_func->type; + memcpy(&surface->in_transfer_func->tf_pts, + &srf_update->in_transfer_func->tf_pts, + sizeof(struct dc_transfer_func_distributed_points)); + } + + if (srf_update->input_csc_color_matrix) + surface->input_csc_color_matrix = + *srf_update->input_csc_color_matrix; + + if (srf_update->coeff_reduction_factor) + surface->coeff_reduction_factor = + *srf_update->coeff_reduction_factor; } static void commit_planes_do_stream_update(struct dc *dc, @@ -1421,16 +1631,18 @@ static void commit_planes_do_stream_update(struct dc *dc, stream_update->adjust->v_total_min, stream_update->adjust->v_total_max); - if (stream_update->periodic_fn_vsync_delta && - pipe_ctx->stream_res.tg && - pipe_ctx->stream_res.tg->funcs->program_vline_interrupt) - pipe_ctx->stream_res.tg->funcs->program_vline_interrupt( - pipe_ctx->stream_res.tg, &pipe_ctx->stream->timing, - pipe_ctx->stream->periodic_fn_vsync_delta); + if (stream_update->periodic_interrupt0 && + dc->hwss.setup_periodic_interrupt) + dc->hwss.setup_periodic_interrupt(pipe_ctx, VLINE0); + + if (stream_update->periodic_interrupt1 && + dc->hwss.setup_periodic_interrupt) + dc->hwss.setup_periodic_interrupt(pipe_ctx, VLINE1); if ((stream_update->hdr_static_metadata && !stream->use_dynamic_meta) || stream_update->vrr_infopacket || - stream_update->vsc_infopacket) { + stream_update->vsc_infopacket || + stream_update->vsp_infopacket) { resource_build_info_frame(pipe_ctx); dc->hwss.update_info_frame(pipe_ctx); } @@ -1441,6 +1653,14 @@ static void commit_planes_do_stream_update(struct dc *dc, if (stream_update->output_csc_transform) dc_stream_program_csc_matrix(dc, stream); + if (stream_update->dither_option) { + resource_build_bit_depth_reduction_params(pipe_ctx->stream, + &pipe_ctx->stream->bit_depth_params); + pipe_ctx->stream_res.opp->funcs->opp_program_fmt(pipe_ctx->stream_res.opp, + &stream->bit_depth_params, + &stream->clamping); + } + /* Full fe update*/ if (update_type == UPDATE_TYPE_FAST) continue; @@ -1448,19 +1668,13 @@ static void commit_planes_do_stream_update(struct dc *dc, if (stream_update->dpms_off) { if (*stream_update->dpms_off) { core_link_disable_stream(pipe_ctx, KEEP_ACQUIRED_RESOURCE); - dc->hwss.pplib_apply_display_requirements( - dc, dc->current_state); - notify_display_count_to_smu(dc, dc->current_state); + dc->hwss.optimize_bandwidth(dc, dc->current_state); } else { - dc->hwss.pplib_apply_display_requirements( - dc, dc->current_state); - notify_display_count_to_smu(dc, dc->current_state); + dc->hwss.prepare_bandwidth(dc, dc->current_state); core_link_enable_stream(dc->current_state, pipe_ctx); } } - - if (stream_update->abm_level && pipe_ctx->stream_res.abm) { if (pipe_ctx->stream_res.tg->funcs->is_blanked) { // if otg funcs defined check if blanked before programming @@ -1487,7 +1701,7 @@ static void commit_planes_for_stream(struct dc *dc, struct pipe_ctx *top_pipe_to_program = NULL; if (update_type == UPDATE_TYPE_FULL) { - dc->hwss.set_bandwidth(dc, context, false); + dc->hwss.prepare_bandwidth(dc, context); context_clock_trace(dc, context); } @@ -1530,9 +1744,6 @@ static void commit_planes_for_stream(struct dc *dc, } } - if (update_type == UPDATE_TYPE_FULL) - context_timing_trace(dc, &context->res_ctx); - // Update Type FAST, Surface updates if (update_type == UPDATE_TYPE_FAST) { /* Lock the top pipe while updating plane addrs, since freesync requires @@ -1568,7 +1779,6 @@ void dc_commit_updates_for_stream(struct dc *dc, int surface_count, struct dc_stream_state *stream, struct dc_stream_update *stream_update, - struct dc_plane_state **plane_states, struct dc_state *state) { const struct dc_stream_status *stream_status; @@ -1597,20 +1807,21 @@ void dc_commit_updates_for_stream(struct dc *dc, } dc_resource_state_copy_construct(state, context); + + for (i = 0; i < dc->res_pool->pipe_count; i++) { + struct pipe_ctx *new_pipe = &context->res_ctx.pipe_ctx[i]; + struct pipe_ctx *old_pipe = &dc->current_state->res_ctx.pipe_ctx[i]; + + if (new_pipe->plane_state && new_pipe->plane_state != old_pipe->plane_state) + new_pipe->plane_state->force_full_update = true; + } } for (i = 0; i < surface_count; i++) { struct dc_plane_state *surface = srf_updates[i].surface; - /* TODO: On flip we don't build the state, so it still has the - * old address. Which is why we are updating the address here - */ - if (srf_updates[i].flip_addr) { - surface->address = srf_updates[i].flip_addr->address; - surface->flip_immediate = srf_updates[i].flip_addr->flip_immediate; - - } + copy_surface_update_to_plane(surface, &srf_updates[i]); if (update_type >= UPDATE_TYPE_MED) { for (j = 0; j < dc->res_pool->pipe_count; j++) { @@ -1640,6 +1851,12 @@ void dc_commit_updates_for_stream(struct dc *dc, dc->current_state = context; dc_release_state(old); + for (i = 0; i < dc->res_pool->pipe_count; i++) { + struct pipe_ctx *pipe_ctx = &context->res_ctx.pipe_ctx[i]; + + if (pipe_ctx->plane_state && pipe_ctx->stream == stream) + pipe_ctx->plane_state->force_full_update = false; + } } /*let's use current_state to update watermark etc*/ if (update_type >= UPDATE_TYPE_FULL) @@ -1669,6 +1886,9 @@ enum dc_irq_source dc_interrupt_to_irq_source( return dal_irq_service_to_irq_source(dc->res_pool->irqs, src_id, ext_id); } +/** + * dc_interrupt_set() - Enable/disable an AMD hw interrupt source + */ bool dc_interrupt_set(struct dc *dc, enum dc_irq_source src, bool enable) { @@ -1724,6 +1944,35 @@ void dc_resume(struct dc *dc) core_link_resume(dc->links[i]); } +unsigned int dc_get_current_backlight_pwm(struct dc *dc) +{ + struct abm *abm = dc->res_pool->abm; + + if (abm) + return abm->funcs->get_current_backlight(abm); + + return 0; +} + +unsigned int dc_get_target_backlight_pwm(struct dc *dc) +{ + struct abm *abm = dc->res_pool->abm; + + if (abm) + return abm->funcs->get_target_backlight(abm); + + return 0; +} + +bool dc_is_dmcu_initialized(struct dc *dc) +{ + struct dmcu *dmcu = dc->res_pool->dmcu; + + if (dmcu) + return dmcu->funcs->is_dmcu_initialized(dmcu); + return false; +} + bool dc_submit_i2c( struct dc *dc, uint32_t link_index, @@ -1753,6 +2002,11 @@ static bool link_add_remote_sink_helper(struct dc_link *dc_link, struct dc_sink return true; } +/** + * dc_link_add_remote_sink() - Create a sink and attach it to an existing link + * + * EDID length is in bytes + */ struct dc_sink *dc_link_add_remote_sink( struct dc_link *link, const uint8_t *edid, @@ -1811,6 +2065,12 @@ fail_add_sink: return NULL; } +/** + * dc_link_remove_remote_sink() - Remove a remote sink from a dc_link + * + * Note that this just removes the struct dc_sink - it doesn't + * program hardware or alter other members of dc_link + */ void dc_link_remove_remote_sink(struct dc_link *link, struct dc_sink *sink) { int i; @@ -1848,4 +2108,4 @@ void get_clock_requirements_for_state(struct dc_state *state, struct AsicStateEx info->dcfClockDeepSleep = (unsigned int)state->bw.dcn.clk.dcfclk_deep_sleep_khz; info->fClock = (unsigned int)state->bw.dcn.clk.fclk_khz; info->phyClock = (unsigned int)state->bw.dcn.clk.phyclk_khz; -}
\ No newline at end of file +} diff --git a/drivers/gpu/drm/amd/display/dc/core/dc_debug.c b/drivers/gpu/drm/amd/display/dc/core/dc_debug.c index e1ebdf7b5eaf..73d049506618 100644 --- a/drivers/gpu/drm/amd/display/dc/core/dc_debug.c +++ b/drivers/gpu/drm/amd/display/dc/core/dc_debug.c @@ -311,7 +311,7 @@ void context_timing_trace( { int i; struct dc *core_dc = dc; - int h_pos[MAX_PIPES], v_pos[MAX_PIPES]; + int h_pos[MAX_PIPES] = {0}, v_pos[MAX_PIPES] = {0}; struct crtc_position position; unsigned int underlay_idx = core_dc->res_pool->underlay_pipe_index; DC_LOGGER_INIT(dc->ctx->logger); @@ -322,8 +322,7 @@ void context_timing_trace( /* get_position() returns CRTC vertical/horizontal counter * hence not applicable for underlay pipe */ - if (pipe_ctx->stream == NULL - || pipe_ctx->pipe_idx == underlay_idx) + if (pipe_ctx->stream == NULL || pipe_ctx->pipe_idx == underlay_idx) continue; pipe_ctx->stream_res.tg->funcs->get_position(pipe_ctx->stream_res.tg, &position); @@ -333,7 +332,7 @@ void context_timing_trace( for (i = 0; i < core_dc->res_pool->pipe_count; i++) { struct pipe_ctx *pipe_ctx = &res_ctx->pipe_ctx[i]; - if (pipe_ctx->stream == NULL) + if (pipe_ctx->stream == NULL || pipe_ctx->pipe_idx == underlay_idx) continue; TIMING_TRACE("OTG_%d H_tot:%d V_tot:%d H_pos:%d V_pos:%d\n", diff --git a/drivers/gpu/drm/amd/display/dc/core/dc_link.c b/drivers/gpu/drm/amd/display/dc/core/dc_link.c index 5da2186b3615..ea18e9c2d8ce 100644 --- a/drivers/gpu/drm/amd/display/dc/core/dc_link.c +++ b/drivers/gpu/drm/amd/display/dc/core/dc_link.c @@ -43,10 +43,6 @@ #include "dpcd_defs.h" #include "dmcu.h" -#include "dce/dce_11_0_d.h" -#include "dce/dce_11_0_enum.h" -#include "dce/dce_11_0_sh_mask.h" - #define DC_LOGGER_INIT(logger) @@ -80,6 +76,12 @@ static void destruct(struct dc_link *link) { int i; + if (link->hpd_gpio != NULL) { + dal_gpio_close(link->hpd_gpio); + dal_gpio_destroy_irq(&link->hpd_gpio); + link->hpd_gpio = NULL; + } + if (link->ddc) dal_ddc_service_destroy(&link->ddc); @@ -198,6 +200,13 @@ static bool program_hpd_filter( return result; } +/** + * dc_link_detect_sink() - Determine if there is a sink connected + * + * @type: Returned connection type + * Does not detect downstream devices, such as MST sinks + * or display connected through active dongles + */ bool dc_link_detect_sink(struct dc_link *link, enum dc_connection_type *type) { uint32_t is_hpd_high = 0; @@ -208,6 +217,9 @@ bool dc_link_detect_sink(struct dc_link *link, enum dc_connection_type *type) return true; } + if (link->connector_signal == SIGNAL_TYPE_EDP) + link->dc->hwss.edp_wait_for_hpd_ready(link, true); + /* todo: may need to lock gpio access */ hpd_pin = get_hpd_gpio(link->ctx->dc_bios, link->link_id, link->ctx->gpio_service); if (hpd_pin == NULL) @@ -324,15 +336,15 @@ static enum signal_type get_basic_signal_type( return SIGNAL_TYPE_NONE; } -/* - * @brief - * Check whether there is a dongle on DP connector +/** + * dc_link_is_dp_sink_present() - Check if there is a native DP + * or passive DP-HDMI dongle connected */ bool dc_link_is_dp_sink_present(struct dc_link *link) { enum gpio_result gpio_result; uint32_t clock_pin = 0; - + uint8_t retry = 0; struct ddc *ddc; enum connector_id connector_id = @@ -361,11 +373,22 @@ bool dc_link_is_dp_sink_present(struct dc_link *link) return present; } - /* Read GPIO: DP sink is present if both clock and data pins are zero */ - /* [anaumov] in DAL2, there was no check for GPIO failure */ - - gpio_result = dal_gpio_get_value(ddc->pin_clock, &clock_pin); - ASSERT(gpio_result == GPIO_RESULT_OK); + /* + * Read GPIO: DP sink is present if both clock and data pins are zero + * + * [W/A] plug-unplug DP cable, sometimes customer board has + * one short pulse on clk_pin(1V, < 1ms). DP will be config to HDMI/DVI + * then monitor can't br light up. Add retry 3 times + * But in real passive dongle, it need additional 3ms to detect + */ + do { + gpio_result = dal_gpio_get_value(ddc->pin_clock, &clock_pin); + ASSERT(gpio_result == GPIO_RESULT_OK); + if (clock_pin) + udelay(1000); + else + break; + } while (retry++ < 3); present = (gpio_result == GPIO_RESULT_OK) && !clock_pin; @@ -593,6 +616,14 @@ static bool is_same_edid(struct dc_edid *old_edid, struct dc_edid *new_edid) return (memcmp(old_edid->raw_edid, new_edid->raw_edid, new_edid->length) == 0); } +/** + * dc_link_detect() - Detect if a sink is attached to a given link + * + * link->local_sink is created or destroyed as needed. + * + * This does not create remote sinks but will trigger DM + * to start MST detection if a branch is detected. + */ bool dc_link_detect(struct dc_link *link, enum dc_detect_reason reason) { struct dc_sink_init_data sink_init_data = { 0 }; @@ -688,12 +719,26 @@ bool dc_link_detect(struct dc_link *link, enum dc_detect_reason reason) if (memcmp(&link->dpcd_caps, &prev_dpcd_caps, sizeof(struct dpcd_caps))) same_dpcd = false; } - /* Active dongle downstream unplug */ + /* Active dongle plug in without display or downstream unplug*/ if (link->type == dc_connection_active_dongle && link->dpcd_caps.sink_count. bits.SINK_COUNT == 0) { - if (prev_sink != NULL) + if (prev_sink != NULL) { + /* Downstream unplug */ dc_sink_release(prev_sink); + } else { + /* Empty dongle plug in */ + for (i = 0; i < LINK_TRAINING_MAX_VERIFY_RETRY; i++) { + int fail_count = 0; + + dp_verify_link_cap(link, + &link->reported_link_cap, + &fail_count); + + if (fail_count == 0) + break; + } + } return true; } @@ -746,9 +791,10 @@ bool dc_link_detect(struct dc_link *link, enum dc_detect_reason reason) return false; } - sink->dongle_max_pix_clk = sink_caps.max_hdmi_pixel_clock; + sink->link->dongle_max_pix_clk = sink_caps.max_hdmi_pixel_clock; sink->converter_disable_audio = converter_disable_audio; + /* dc_sink_create returns a new reference */ link->local_sink = sink; edid_status = dm_helpers_read_local_edid( @@ -892,18 +938,11 @@ bool dc_link_detect(struct dc_link *link, enum dc_detect_reason reason) bool dc_link_get_hpd_state(struct dc_link *dc_link) { - struct gpio *hpd_pin; uint32_t state; - hpd_pin = get_hpd_gpio(dc_link->ctx->dc_bios, - dc_link->link_id, dc_link->ctx->gpio_service); - if (hpd_pin == NULL) - ASSERT(false); - - dal_gpio_open(hpd_pin, GPIO_MODE_INTERRUPT); - dal_gpio_get_value(hpd_pin, &state); - dal_gpio_close(hpd_pin); - dal_gpio_destroy_irq(&hpd_pin); + dal_gpio_lock_pin(dc_link->hpd_gpio); + dal_gpio_get_value(dc_link->hpd_gpio, &state); + dal_gpio_unlock_pin(dc_link->hpd_gpio); return state; } @@ -1059,7 +1098,6 @@ static bool construct( const struct link_init_data *init_params) { uint8_t i; - struct gpio *hpd_gpio = NULL; struct ddc_service_init_data ddc_service_init_data = { { 0 } }; struct dc_context *dc_ctx = init_params->ctx; struct encoder_init_data enc_init_data = { 0 }; @@ -1089,10 +1127,12 @@ static bool construct( if (link->dc->res_pool->funcs->link_init) link->dc->res_pool->funcs->link_init(link); - hpd_gpio = get_hpd_gpio(link->ctx->dc_bios, link->link_id, link->ctx->gpio_service); - - if (hpd_gpio != NULL) - link->irq_source_hpd = dal_irq_get_source(hpd_gpio); + link->hpd_gpio = get_hpd_gpio(link->ctx->dc_bios, link->link_id, link->ctx->gpio_service); + if (link->hpd_gpio != NULL) { + dal_gpio_open(link->hpd_gpio, GPIO_MODE_INTERRUPT); + dal_gpio_unlock_pin(link->hpd_gpio); + link->irq_source_hpd = dal_irq_get_source(link->hpd_gpio); + } switch (link->link_id.id) { case CONNECTOR_ID_HDMI_TYPE_A: @@ -1110,18 +1150,18 @@ static bool construct( case CONNECTOR_ID_DISPLAY_PORT: link->connector_signal = SIGNAL_TYPE_DISPLAY_PORT; - if (hpd_gpio != NULL) + if (link->hpd_gpio != NULL) link->irq_source_hpd_rx = - dal_irq_get_rx_source(hpd_gpio); + dal_irq_get_rx_source(link->hpd_gpio); break; case CONNECTOR_ID_EDP: link->connector_signal = SIGNAL_TYPE_EDP; - if (hpd_gpio != NULL) { + if (link->hpd_gpio != NULL) { link->irq_source_hpd = DC_IRQ_SOURCE_INVALID; link->irq_source_hpd_rx = - dal_irq_get_rx_source(hpd_gpio); + dal_irq_get_rx_source(link->hpd_gpio); } break; case CONNECTOR_ID_LVDS: @@ -1132,10 +1172,7 @@ static bool construct( goto create_fail; } - if (hpd_gpio != NULL) { - dal_gpio_destroy_irq(&hpd_gpio); - hpd_gpio = NULL; - } + /* TODO: #DAL3 Implement id to str function.*/ LINK_INFO("Connector[%d] description:" @@ -1238,8 +1275,9 @@ link_enc_create_fail: ddc_create_fail: create_fail: - if (hpd_gpio != NULL) { - dal_gpio_destroy_irq(&hpd_gpio); + if (link->hpd_gpio != NULL) { + dal_gpio_destroy_irq(&link->hpd_gpio); + link->hpd_gpio = NULL; } return false; @@ -1329,7 +1367,7 @@ static void dpcd_configure_panel_mode( static void enable_stream_features(struct pipe_ctx *pipe_ctx) { struct dc_stream_state *stream = pipe_ctx->stream; - struct dc_link *link = stream->sink->link; + struct dc_link *link = stream->link; union down_spread_ctrl old_downspread; union down_spread_ctrl new_downspread; @@ -1354,31 +1392,16 @@ static enum dc_status enable_link_dp( struct dc_stream_state *stream = pipe_ctx->stream; enum dc_status status; bool skip_video_pattern; - struct dc_link *link = stream->sink->link; + struct dc_link *link = stream->link; struct dc_link_settings link_settings = {0}; enum dp_panel_mode panel_mode; - enum dc_link_rate max_link_rate = LINK_RATE_HIGH2; /* get link settings for video mode timing */ decide_link_settings(stream, &link_settings); - /* raise clock state for HBR3 if required. Confirmed with HW DCE/DPCS - * logic for HBR3 still needs Nominal (0.8V) on VDDC rail - */ - if (link->link_enc->features.flags.bits.IS_HBR3_CAPABLE) - max_link_rate = LINK_RATE_HIGH3; - - if (link_settings.link_rate == max_link_rate) { - struct dc_clocks clocks = state->bw.dcn.clk; - - /* dce/dcn compat, do not update dispclk */ - clocks.dispclk_khz = 0; - /* 27mhz = 27000000hz= 27000khz */ - clocks.phyclk_khz = link_settings.link_rate * 27000; - - state->dis_clk->funcs->update_clocks( - state->dis_clk, &clocks, false); - } + pipe_ctx->stream_res.pix_clk_params.requested_sym_clk = + link_settings.link_rate * LINK_RATE_REF_FREQ_IN_KHZ; + state->dccg->funcs->update_clocks(state->dccg, state, false); dp_enable_link_phy( link, @@ -1386,8 +1409,8 @@ static enum dc_status enable_link_dp( pipe_ctx->clock_source->id, &link_settings); - if (stream->sink->edid_caps.panel_patch.dppowerup_delay > 0) { - int delay_dp_power_up_in_ms = stream->sink->edid_caps.panel_patch.dppowerup_delay; + if (stream->sink_patches.dppowerup_delay > 0) { + int delay_dp_power_up_in_ms = stream->sink_patches.dppowerup_delay; msleep(delay_dp_power_up_in_ms); } @@ -1411,8 +1434,6 @@ static enum dc_status enable_link_dp( else status = DC_FAIL_DP_LINK_TRAINING; - enable_stream_features(pipe_ctx); - return status; } @@ -1422,7 +1443,7 @@ static enum dc_status enable_link_edp( { enum dc_status status; struct dc_stream_state *stream = pipe_ctx->stream; - struct dc_link *link = stream->sink->link; + struct dc_link *link = stream->link; /*in case it is not on*/ link->dc->hwss.edp_power_control(link, true); link->dc->hwss.edp_wait_for_hpd_ready(link, true); @@ -1437,7 +1458,7 @@ static enum dc_status enable_link_dp_mst( struct dc_state *state, struct pipe_ctx *pipe_ctx) { - struct dc_link *link = pipe_ctx->stream->sink->link; + struct dc_link *link = pipe_ctx->stream->link; /* sink signal type after MST branch is MST. Multiple MST sinks * share one link. Link DP PHY is enable or training only once. @@ -1445,6 +1466,11 @@ static enum dc_status enable_link_dp_mst( if (link->cur_link_settings.lane_count != LANE_COUNT_UNKNOWN) return DC_OK; + /* to make sure the pending down rep can be processed + * before clear payload table + */ + dm_helpers_dp_mst_poll_pending_down_reply(link->ctx, link); + /* clear payload table */ dm_helpers_dp_mst_clear_payload_allocation_table(link->ctx, link); @@ -1571,7 +1597,7 @@ static bool i2c_write(struct pipe_ctx *pipe_ctx, cmd.payloads = &payload; if (dm_helpers_submit_i2c(pipe_ctx->stream->ctx, - pipe_ctx->stream->sink->link, &cmd)) + pipe_ctx->stream->link, &cmd)) return true; return false; @@ -1625,7 +1651,7 @@ static void write_i2c_retimer_setting( else { i2c_success = dal_ddc_service_query_ddc_data( - pipe_ctx->stream->sink->link->ddc, + pipe_ctx->stream->link->ddc, slave_address, &offset, 1, &value, 1); if (!i2c_success) /* Write failure */ @@ -1678,7 +1704,7 @@ static void write_i2c_retimer_setting( else { i2c_success = dal_ddc_service_query_ddc_data( - pipe_ctx->stream->sink->link->ddc, + pipe_ctx->stream->link->ddc, slave_address, &offset, 1, &value, 1); if (!i2c_success) /* Write failure */ @@ -1903,7 +1929,7 @@ static void write_i2c_redriver_setting( static void enable_link_hdmi(struct pipe_ctx *pipe_ctx) { struct dc_stream_state *stream = pipe_ctx->stream; - struct dc_link *link = stream->sink->link; + struct dc_link *link = stream->link; enum dc_color_depth display_color_depth; enum engine_id eng_id; struct ext_hdmi_settings settings = {0}; @@ -1912,12 +1938,12 @@ static void enable_link_hdmi(struct pipe_ctx *pipe_ctx) && (stream->timing.v_addressable == 480); if (stream->phy_pix_clk == 0) - stream->phy_pix_clk = stream->timing.pix_clk_khz; + stream->phy_pix_clk = stream->timing.pix_clk_100hz / 10; if (stream->phy_pix_clk > 340000) is_over_340mhz = true; if (dc_is_hdmi_signal(pipe_ctx->stream->signal)) { - unsigned short masked_chip_caps = pipe_ctx->stream->sink->link->chip_caps & + unsigned short masked_chip_caps = pipe_ctx->stream->link->chip_caps & EXT_DISPLAY_PATH_CAPS__EXT_CHIP_MASK; if (masked_chip_caps == EXT_DISPLAY_PATH_CAPS__HDMI20_TISN65DP159RSBT) { /* DP159, Retimer settings */ @@ -1938,11 +1964,11 @@ static void enable_link_hdmi(struct pipe_ctx *pipe_ctx) if (dc_is_hdmi_signal(pipe_ctx->stream->signal)) dal_ddc_service_write_scdc_data( - stream->sink->link->ddc, + stream->link->ddc, stream->phy_pix_clk, stream->timing.flags.LTE_340MCSC_SCRAMBLE); - memset(&stream->sink->link->cur_link_settings, 0, + memset(&stream->link->cur_link_settings, 0, sizeof(struct dc_link_settings)); display_color_depth = stream->timing.display_color_depth; @@ -1963,12 +1989,12 @@ static void enable_link_hdmi(struct pipe_ctx *pipe_ctx) static void enable_link_lvds(struct pipe_ctx *pipe_ctx) { struct dc_stream_state *stream = pipe_ctx->stream; - struct dc_link *link = stream->sink->link; + struct dc_link *link = stream->link; if (stream->phy_pix_clk == 0) - stream->phy_pix_clk = stream->timing.pix_clk_khz; + stream->phy_pix_clk = stream->timing.pix_clk_100hz / 10; - memset(&stream->sink->link->cur_link_settings, 0, + memset(&stream->link->cur_link_settings, 0, sizeof(struct dc_link_settings)); link->link_enc->funcs->enable_lvds_output( @@ -2012,6 +2038,9 @@ static enum dc_status enable_link( break; } + if (status == DC_OK) + pipe_ctx->stream->link->link_status.link_active = true; + return status; } @@ -2035,13 +2064,21 @@ static void disable_link(struct dc_link *link, enum signal_type signal) dp_disable_link_phy_mst(link, signal); } else link->link_enc->funcs->disable_output(link->link_enc, signal); + + if (signal == SIGNAL_TYPE_DISPLAY_PORT_MST) { + /* MST disable link only when no stream use the link */ + if (link->mst_stream_alloc_table.stream_count <= 0) + link->link_status.link_active = false; + } else { + link->link_status.link_active = false; + } } static bool dp_active_dongle_validate_timing( const struct dc_crtc_timing *timing, const struct dpcd_caps *dpcd_caps) { - unsigned int required_pix_clk = timing->pix_clk_khz; + unsigned int required_pix_clk_100hz = timing->pix_clk_100hz; const struct dc_dongle_caps *dongle_caps = &dpcd_caps->dongle_caps; switch (dpcd_caps->dongle_type) { @@ -2081,9 +2118,9 @@ static bool dp_active_dongle_validate_timing( /* Check Color Depth and Pixel Clock */ if (timing->pixel_encoding == PIXEL_ENCODING_YCBCR420) - required_pix_clk /= 2; + required_pix_clk_100hz /= 2; else if (timing->pixel_encoding == PIXEL_ENCODING_YCBCR422) - required_pix_clk = required_pix_clk * 2 / 3; + required_pix_clk_100hz = required_pix_clk_100hz * 2 / 3; switch (timing->display_color_depth) { case COLOR_DEPTH_666: @@ -2093,12 +2130,12 @@ static bool dp_active_dongle_validate_timing( case COLOR_DEPTH_101010: if (dongle_caps->dp_hdmi_max_bpc < 10) return false; - required_pix_clk = required_pix_clk * 10 / 8; + required_pix_clk_100hz = required_pix_clk_100hz * 10 / 8; break; case COLOR_DEPTH_121212: if (dongle_caps->dp_hdmi_max_bpc < 12) return false; - required_pix_clk = required_pix_clk * 12 / 8; + required_pix_clk_100hz = required_pix_clk_100hz * 12 / 8; break; case COLOR_DEPTH_141414: @@ -2108,7 +2145,7 @@ static bool dp_active_dongle_validate_timing( return false; } - if (required_pix_clk > dongle_caps->dp_hdmi_max_pixel_clk) + if (required_pix_clk_100hz > (dongle_caps->dp_hdmi_max_pixel_clk * 10)) return false; return true; @@ -2119,7 +2156,7 @@ enum dc_status dc_link_validate_mode_timing( struct dc_link *link, const struct dc_crtc_timing *timing) { - uint32_t max_pix_clk = stream->sink->dongle_max_pix_clk; + uint32_t max_pix_clk = stream->link->dongle_max_pix_clk * 10; struct dpcd_caps *dpcd_caps = &link->dpcd_caps; /* A hack to avoid failing any modes for EDID override feature on @@ -2129,7 +2166,7 @@ enum dc_status dc_link_validate_mode_timing( return DC_OK; /* Passive Dongle */ - if (0 != max_pix_clk && timing->pix_clk_khz > max_pix_clk) + if (0 != max_pix_clk && timing->pix_clk_100hz > max_pix_clk) return DC_EXCEED_DONGLE_CAP; /* Active Dongle*/ @@ -2156,14 +2193,15 @@ int dc_link_get_backlight_level(const struct dc_link *link) { struct abm *abm = link->ctx->dc->res_pool->abm; - if (abm == NULL || abm->funcs->get_current_backlight_8_bit == NULL) + if (abm == NULL || abm->funcs->get_current_backlight == NULL) return DC_ERROR_UNEXPECTED; - return (int) abm->funcs->get_current_backlight_8_bit(abm); + return (int) abm->funcs->get_current_backlight(abm); } -bool dc_link_set_backlight_level(const struct dc_link *link, uint32_t level, - uint32_t frame_ramp, const struct dc_stream_state *stream) +bool dc_link_set_backlight_level(const struct dc_link *link, + uint32_t backlight_pwm_u16_16, + uint32_t frame_ramp) { struct dc *core_dc = link->ctx->dc; struct abm *abm = core_dc->res_pool->abm; @@ -2175,26 +2213,20 @@ bool dc_link_set_backlight_level(const struct dc_link *link, uint32_t level, if ((dmcu == NULL) || (abm == NULL) || - (abm->funcs->set_backlight_level == NULL)) + (abm->funcs->set_backlight_level_pwm == NULL)) return false; - if (stream) { - if (stream->bl_pwm_level == EDP_BACKLIGHT_RAMP_DISABLE_LEVEL) - frame_ramp = 0; - - ((struct dc_stream_state *)stream)->bl_pwm_level = level; - } - use_smooth_brightness = dmcu->funcs->is_dmcu_initialized(dmcu); - DC_LOG_BACKLIGHT("New Backlight level: %d (0x%X)\n", level, level); + DC_LOG_BACKLIGHT("New Backlight level: %d (0x%X)\n", + backlight_pwm_u16_16, backlight_pwm_u16_16); if (dc_is_embedded_signal(link->connector_signal)) { - if (stream != NULL) { - for (i = 0; i < MAX_PIPES; i++) { + for (i = 0; i < MAX_PIPES; i++) { + if (core_dc->current_state->res_ctx.pipe_ctx[i].stream) { if (core_dc->current_state->res_ctx. - pipe_ctx[i].stream - == stream) + pipe_ctx[i].stream->link + == link) /* DMCU -1 for all controller id values, * therefore +1 here */ @@ -2204,9 +2236,9 @@ bool dc_link_set_backlight_level(const struct dc_link *link, uint32_t level, 1; } } - abm->funcs->set_backlight_level( + abm->funcs->set_backlight_level_pwm( abm, - level, + backlight_pwm_u16_16, frame_ramp, controller_id, use_smooth_brightness); @@ -2220,7 +2252,7 @@ bool dc_link_set_abm_disable(const struct dc_link *link) struct dc *core_dc = link->ctx->dc; struct abm *abm = core_dc->res_pool->abm; - if ((abm == NULL) || (abm->funcs->set_backlight_level == NULL)) + if ((abm == NULL) || (abm->funcs->set_backlight_level_pwm == NULL)) return false; abm->funcs->set_abm_immediate_disable(abm); @@ -2233,7 +2265,7 @@ bool dc_link_set_psr_enable(const struct dc_link *link, bool enable, bool wait) struct dc *core_dc = link->ctx->dc; struct dmcu *dmcu = core_dc->res_pool->dmcu; - if (dmcu != NULL && link->psr_enabled) + if ((dmcu != NULL && dmcu->funcs->is_dmcu_initialized(dmcu)) && link->psr_enabled) dmcu->funcs->set_psr_enable(dmcu, enable, wait); return true; @@ -2253,7 +2285,7 @@ void core_link_resume(struct dc_link *link) static struct fixed31_32 get_pbn_per_slot(struct dc_stream_state *stream) { struct dc_link_settings *link_settings = - &stream->sink->link->cur_link_settings; + &stream->link->cur_link_settings; uint32_t link_rate_in_mbps = link_settings->link_rate * LINK_RATE_REF_FREQ_IN_MHZ; struct fixed31_32 mbps = dc_fixpt_from_int( @@ -2284,7 +2316,7 @@ static struct fixed31_32 get_pbn_from_timing(struct pipe_ctx *pipe_ctx) uint32_t denominator; bpc = get_color_depth(pipe_ctx->stream_res.pix_clk_params.color_depth); - kbps = pipe_ctx->stream_res.pix_clk_params.requested_pix_clk * bpc * 3; + kbps = pipe_ctx->stream_res.pix_clk_params.requested_pix_clk_100hz / 10 * bpc * 3; /* * margin 5300ppm + 300ppm ~ 0.6% as per spec, factor is 1.006 @@ -2360,7 +2392,7 @@ static void update_mst_stream_alloc_table( static enum dc_status allocate_mst_payload(struct pipe_ctx *pipe_ctx) { struct dc_stream_state *stream = pipe_ctx->stream; - struct dc_link *link = stream->sink->link; + struct dc_link *link = stream->link; struct link_encoder *link_encoder = link->link_enc; struct stream_encoder *stream_encoder = pipe_ctx->stream_res.stream_enc; struct dp_mst_stream_allocation_table proposed_table = {0}; @@ -2440,7 +2472,7 @@ static enum dc_status allocate_mst_payload(struct pipe_ctx *pipe_ctx) static enum dc_status deallocate_mst_payload(struct pipe_ctx *pipe_ctx) { struct dc_stream_state *stream = pipe_ctx->stream; - struct dc_link *link = stream->sink->link; + struct dc_link *link = stream->link; struct link_encoder *link_encoder = link->link_enc; struct stream_encoder *stream_encoder = pipe_ctx->stream_res.stream_enc; struct dp_mst_stream_allocation_table proposed_table = {0}; @@ -2525,8 +2557,8 @@ void core_link_enable_stream( DC_LOGGER_INIT(pipe_ctx->stream->ctx->logger); if (pipe_ctx->stream->signal != SIGNAL_TYPE_VIRTUAL) { - stream->sink->link->link_enc->funcs->setup( - stream->sink->link->link_enc, + stream->link->link_enc->funcs->setup( + stream->link->link_enc, pipe_ctx->stream->signal); pipe_ctx->stream_res.stream_enc->funcs->setup_stereo_sync( pipe_ctx->stream_res.stream_enc, @@ -2560,13 +2592,23 @@ void core_link_enable_stream( &stream->timing); if (!IS_FPGA_MAXIMUS_DC(core_dc->ctx->dce_environment)) { + bool apply_edp_fast_boot_optimization = + pipe_ctx->stream->apply_edp_fast_boot_optimization; + + pipe_ctx->stream->apply_edp_fast_boot_optimization = false; + resource_build_info_frame(pipe_ctx); core_dc->hwss.update_info_frame(pipe_ctx); + /* Do not touch link on seamless boot optimization. */ + if (pipe_ctx->stream->apply_seamless_boot_optimization) { + pipe_ctx->stream->dpms_off = false; + return; + } + /* eDP lit up by bios already, no need to enable again. */ if (pipe_ctx->stream->signal == SIGNAL_TYPE_EDP && - pipe_ctx->stream->apply_edp_fast_boot_optimization) { - pipe_ctx->stream->apply_edp_fast_boot_optimization = false; + apply_edp_fast_boot_optimization) { pipe_ctx->stream->dpms_off = false; return; } @@ -2578,7 +2620,7 @@ void core_link_enable_stream( if (status != DC_OK) { DC_LOG_WARNING("enabling link %u failed: %d\n", - pipe_ctx->stream->sink->link->link_index, + pipe_ctx->stream->link->link_index, status); /* Abort stream enable *unless* the failure was due to @@ -2607,8 +2649,10 @@ void core_link_enable_stream( allocate_mst_payload(pipe_ctx); core_dc->hwss.unblank_stream(pipe_ctx, - &pipe_ctx->stream->sink->link->cur_link_settings); + &pipe_ctx->stream->link->cur_link_settings); + if (dc_is_dp_signal(pipe_ctx->stream->signal)) + enable_stream_features(pipe_ctx); } } @@ -2616,15 +2660,21 @@ void core_link_enable_stream( void core_link_disable_stream(struct pipe_ctx *pipe_ctx, int option) { struct dc *core_dc = pipe_ctx->stream->ctx->dc; + struct dc_stream_state *stream = pipe_ctx->stream; + + core_dc->hwss.blank_stream(pipe_ctx); if (pipe_ctx->stream->signal == SIGNAL_TYPE_DISPLAY_PORT_MST) deallocate_mst_payload(pipe_ctx); - core_dc->hwss.blank_stream(pipe_ctx); + if (dc_is_hdmi_signal(pipe_ctx->stream->signal)) + dal_ddc_service_write_scdc_data( + stream->link->ddc, 0, + stream->timing.flags.LTE_340MCSC_SCRAMBLE); core_dc->hwss.disable_stream(pipe_ctx, option); - disable_link(pipe_ctx->stream->sink->link, pipe_ctx->stream->signal); + disable_link(pipe_ctx->stream->link, pipe_ctx->stream->signal); } void core_link_set_avmute(struct pipe_ctx *pipe_ctx, bool enable) diff --git a/drivers/gpu/drm/amd/display/dc/core/dc_link_ddc.c b/drivers/gpu/drm/amd/display/dc/core/dc_link_ddc.c index 506a97e16956..b7ee63cd8dc7 100644 --- a/drivers/gpu/drm/amd/display/dc/core/dc_link_ddc.c +++ b/drivers/gpu/drm/amd/display/dc/core/dc_link_ddc.c @@ -33,7 +33,7 @@ #include "include/vector.h" #include "core_types.h" #include "dc_link_ddc.h" -#include "aux_engine.h" +#include "dce/dce_aux.h" #define AUX_POWER_UP_WA_DELAY 500 #define I2C_OVER_AUX_DEFER_WA_DELAY 70 @@ -42,7 +42,6 @@ #define CV_SMART_DONGLE_ADDRESS 0x20 /* DVI-HDMI dongle slave address for retrieving dongle signature*/ #define DVI_HDMI_DONGLE_ADDRESS 0x68 -static const int8_t dvi_hdmi_dongle_signature_str[] = "6140063500G"; struct dvi_hdmi_dongle_signature_data { int8_t vendor[3];/* "AMD" */ uint8_t version[2]; @@ -165,43 +164,6 @@ static void dal_ddc_i2c_payloads_destroy(struct i2c_payloads **p) } -static struct aux_payloads *dal_ddc_aux_payloads_create(struct dc_context *ctx, uint32_t count) -{ - struct aux_payloads *payloads; - - payloads = kzalloc(sizeof(struct aux_payloads), GFP_KERNEL); - - if (!payloads) - return NULL; - - if (dal_vector_construct( - &payloads->payloads, ctx, count, sizeof(struct aux_payload))) - return payloads; - - kfree(payloads); - return NULL; -} - -static struct aux_payload *dal_ddc_aux_payloads_get(struct aux_payloads *p) -{ - return (struct aux_payload *)p->payloads.container; -} - -static uint32_t dal_ddc_aux_payloads_get_count(struct aux_payloads *p) -{ - return p->payloads.count; -} - -static void dal_ddc_aux_payloads_destroy(struct aux_payloads **p) -{ - if (!p || !*p) - return; - - dal_vector_destruct(&(*p)->payloads); - kfree(*p); - *p = NULL; -} - #define DDC_MIN(a, b) (((a) < (b)) ? (a) : (b)) void dal_ddc_i2c_payloads_add( @@ -225,27 +187,6 @@ void dal_ddc_i2c_payloads_add( } -void dal_ddc_aux_payloads_add( - struct aux_payloads *payloads, - uint32_t address, - uint32_t len, - uint8_t *data, - bool write) -{ - uint32_t payload_size = DEFAULT_AUX_MAX_DATA_SIZE; - uint32_t pos; - - for (pos = 0; pos < len; pos += payload_size) { - struct aux_payload payload = { - .i2c_over_aux = true, - .write = write, - .address = address, - .length = DDC_MIN(payload_size, len - pos), - .data = data + pos }; - dal_vector_append(&payloads->payloads, &payload); - } -} - static void construct( struct ddc_service *ddc_service, struct ddc_service_init_data *init_data) @@ -574,32 +515,34 @@ bool dal_ddc_service_query_ddc_data( /*TODO: len of payload data for i2c and aux is uint8!!!!, * but we want to read 256 over i2c!!!!*/ if (dal_ddc_service_is_in_aux_transaction_mode(ddc)) { - - struct aux_payloads *payloads = - dal_ddc_aux_payloads_create(ddc->ctx, payloads_num); - - struct aux_command command = { - .payloads = dal_ddc_aux_payloads_get(payloads), - .number_of_payloads = 0, + struct aux_payload write_payload = { + .i2c_over_aux = true, + .write = true, + .mot = true, + .address = address, + .length = write_size, + .data = write_buf, + .reply = NULL, .defer_delay = get_defer_delay(ddc), - .max_defer_write_retry = 0 }; + }; - dal_ddc_aux_payloads_add( - payloads, address, write_size, write_buf, true); - - dal_ddc_aux_payloads_add( - payloads, address, read_size, read_buf, false); - - command.number_of_payloads = - dal_ddc_aux_payloads_get_count(payloads); + struct aux_payload read_payload = { + .i2c_over_aux = true, + .write = false, + .mot = false, + .address = address, + .length = read_size, + .data = read_buf, + .reply = NULL, + .defer_delay = get_defer_delay(ddc), + }; - ret = dal_i2caux_submit_aux_command( - ddc->ctx->i2caux, - ddc->ddc_pin, - &command); + ret = dc_link_aux_transfer_with_retries(ddc, &write_payload); - dal_ddc_aux_payloads_destroy(&payloads); + if (!ret) + return false; + ret = dc_link_aux_transfer_with_retries(ddc, &read_payload); } else { struct i2c_payloads *payloads = dal_ddc_i2c_payloads_create(ddc->ctx, payloads_num); @@ -631,56 +574,15 @@ bool dal_ddc_service_query_ddc_data( } int dc_link_aux_transfer(struct ddc_service *ddc, - unsigned int address, - uint8_t *reply, - void *buffer, - unsigned int size, - enum aux_transaction_type type, - enum i2caux_transaction_action action) + struct aux_payload *payload) { - struct ddc *ddc_pin = ddc->ddc_pin; - struct aux_engine *aux_engine; - enum aux_channel_operation_result operation_result; - struct aux_request_transaction_data aux_req; - struct aux_reply_transaction_data aux_rep; - uint8_t returned_bytes = 0; - int res = -1; - uint32_t status; - - memset(&aux_req, 0, sizeof(aux_req)); - memset(&aux_rep, 0, sizeof(aux_rep)); - - aux_engine = ddc->ctx->dc->res_pool->engines[ddc_pin->pin_data->en]; - aux_engine->funcs->acquire(aux_engine, ddc_pin); - - aux_req.type = type; - aux_req.action = action; - - aux_req.address = address; - aux_req.delay = 0; - aux_req.length = size; - aux_req.data = buffer; - - aux_engine->funcs->submit_channel_request(aux_engine, &aux_req); - operation_result = aux_engine->funcs->get_channel_status(aux_engine, &returned_bytes); - - switch (operation_result) { - case AUX_CHANNEL_OPERATION_SUCCEEDED: - res = aux_engine->funcs->read_channel_reply(aux_engine, size, - buffer, reply, - &status); - break; - case AUX_CHANNEL_OPERATION_FAILED_HPD_DISCON: - res = 0; - break; - case AUX_CHANNEL_OPERATION_FAILED_REASON_UNKNOWN: - case AUX_CHANNEL_OPERATION_FAILED_INVALID_REPLY: - case AUX_CHANNEL_OPERATION_FAILED_TIMEOUT: - res = -1; - break; - } - aux_engine->funcs->release_engine(aux_engine); - return res; + return dce_aux_transfer(ddc, payload); +} + +bool dc_link_aux_transfer_with_retries(struct ddc_service *ddc, + struct aux_payload *payload) +{ + return dce_aux_transfer_with_retries(ddc, payload); } /*test only function*/ diff --git a/drivers/gpu/drm/amd/display/dc/core/dc_link_dp.c b/drivers/gpu/drm/amd/display/dc/core/dc_link_dp.c index d91df5ef0cb3..09d301216076 100644 --- a/drivers/gpu/drm/amd/display/dc/core/dc_link_dp.c +++ b/drivers/gpu/drm/amd/display/dc/core/dc_link_dp.c @@ -49,6 +49,8 @@ static void wait_for_training_aux_rd_interval( { union training_aux_rd_interval training_rd_interval; + memset(&training_rd_interval, 0, sizeof(training_rd_interval)); + /* overwrite the delay if rev > 1.1*/ if (link->dpcd_caps.dpcd_rev.raw >= DPCD_REV_12) { /* DP 1.2 or later - retrieve delay through @@ -117,6 +119,13 @@ static void dpcd_set_link_settings( core_link_write_dpcd(link, DP_DOWNSPREAD_CTRL, &downspread.raw, sizeof(downspread)); + if (link->dpcd_caps.dpcd_rev.raw >= DPCD_REV_14 && + (link->dpcd_caps.link_rate_set >= 1 && + link->dpcd_caps.link_rate_set <= 8)) { + core_link_write_dpcd(link, DP_LINK_RATE_SET, + &link->dpcd_caps.link_rate_set, 1); + } + DC_LOG_HW_LINK_TRAINING("%s\n %x rate = %x\n %x lane = %x\n %x spread = %x\n", __func__, DP_LINK_BW_SET, @@ -1089,6 +1098,121 @@ static struct dc_link_settings get_max_link_cap(struct dc_link *link) return max_link_cap; } +static enum dc_status read_hpd_rx_irq_data( + struct dc_link *link, + union hpd_irq_data *irq_data) +{ + static enum dc_status retval; + + /* The HW reads 16 bytes from 200h on HPD, + * but if we get an AUX_DEFER, the HW cannot retry + * and this causes the CTS tests 4.3.2.1 - 3.2.4 to + * fail, so we now explicitly read 6 bytes which is + * the req from the above mentioned test cases. + * + * For DP 1.4 we need to read those from 2002h range. + */ + if (link->dpcd_caps.dpcd_rev.raw < DPCD_REV_14) + retval = core_link_read_dpcd( + link, + DP_SINK_COUNT, + irq_data->raw, + sizeof(union hpd_irq_data)); + else { + /* Read 14 bytes in a single read and then copy only the required fields. + * This is more efficient than doing it in two separate AUX reads. */ + + uint8_t tmp[DP_SINK_STATUS_ESI - DP_SINK_COUNT_ESI + 1]; + + retval = core_link_read_dpcd( + link, + DP_SINK_COUNT_ESI, + tmp, + sizeof(tmp)); + + if (retval != DC_OK) + return retval; + + irq_data->bytes.sink_cnt.raw = tmp[DP_SINK_COUNT_ESI - DP_SINK_COUNT_ESI]; + irq_data->bytes.device_service_irq.raw = tmp[DP_DEVICE_SERVICE_IRQ_VECTOR_ESI0 - DP_SINK_COUNT_ESI]; + irq_data->bytes.lane01_status.raw = tmp[DP_LANE0_1_STATUS_ESI - DP_SINK_COUNT_ESI]; + irq_data->bytes.lane23_status.raw = tmp[DP_LANE2_3_STATUS_ESI - DP_SINK_COUNT_ESI]; + irq_data->bytes.lane_status_updated.raw = tmp[DP_LANE_ALIGN_STATUS_UPDATED_ESI - DP_SINK_COUNT_ESI]; + irq_data->bytes.sink_status.raw = tmp[DP_SINK_STATUS_ESI - DP_SINK_COUNT_ESI]; + } + + return retval; +} + +static bool hpd_rx_irq_check_link_loss_status( + struct dc_link *link, + union hpd_irq_data *hpd_irq_dpcd_data) +{ + uint8_t irq_reg_rx_power_state = 0; + enum dc_status dpcd_result = DC_ERROR_UNEXPECTED; + union lane_status lane_status; + uint32_t lane; + bool sink_status_changed; + bool return_code; + + sink_status_changed = false; + return_code = false; + + if (link->cur_link_settings.lane_count == 0) + return return_code; + + /*1. Check that Link Status changed, before re-training.*/ + + /*parse lane status*/ + for (lane = 0; lane < link->cur_link_settings.lane_count; lane++) { + /* check status of lanes 0,1 + * changed DpcdAddress_Lane01Status (0x202) + */ + lane_status.raw = get_nibble_at_index( + &hpd_irq_dpcd_data->bytes.lane01_status.raw, + lane); + + if (!lane_status.bits.CHANNEL_EQ_DONE_0 || + !lane_status.bits.CR_DONE_0 || + !lane_status.bits.SYMBOL_LOCKED_0) { + /* if one of the channel equalization, clock + * recovery or symbol lock is dropped + * consider it as (link has been + * dropped) dp sink status has changed + */ + sink_status_changed = true; + break; + } + } + + /* Check interlane align.*/ + if (sink_status_changed || + !hpd_irq_dpcd_data->bytes.lane_status_updated.bits.INTERLANE_ALIGN_DONE) { + + DC_LOG_HW_HPD_IRQ("%s: Link Status changed.\n", __func__); + + return_code = true; + + /*2. Check that we can handle interrupt: Not in FS DOS, + * Not in "Display Timeout" state, Link is trained. + */ + dpcd_result = core_link_read_dpcd(link, + DP_SET_POWER, + &irq_reg_rx_power_state, + sizeof(irq_reg_rx_power_state)); + + if (dpcd_result != DC_OK) { + DC_LOG_HW_HPD_IRQ("%s: DPCD read failed to obtain power state.\n", + __func__); + } else { + if (irq_reg_rx_power_state != DP_SET_POWER_D0) + return_code = false; + } + } + + return return_code; +} + bool dp_verify_link_cap( struct dc_link *link, struct dc_link_settings *known_limit_link_setting, @@ -1104,12 +1228,14 @@ bool dp_verify_link_cap( struct clock_source *dp_cs; enum clock_source_id dp_cs_id = CLOCK_SOURCE_ID_EXTERNAL; enum link_training_result status; + union hpd_irq_data irq_data; if (link->dc->debug.skip_detection_link_training) { link->verified_link_cap = *known_limit_link_setting; return true; } + memset(&irq_data, 0, sizeof(irq_data)); success = false; skip_link_training = false; @@ -1168,9 +1294,15 @@ bool dp_verify_link_cap( (*fail_count)++; } - if (success) + if (success) { link->verified_link_cap = *cur; - + udelay(1000); + if (read_hpd_rx_irq_data(link, &irq_data) == DC_OK) + if (hpd_rx_irq_check_link_loss_status( + link, + &irq_data)) + (*fail_count)++; + } /* always disable the link before trying another * setting or before returning we'll enable it later * based on the actual mode we're driving @@ -1419,7 +1551,7 @@ static uint32_t bandwidth_in_kbps_from_timing( ASSERT(bits_per_channel != 0); - kbps = timing->pix_clk_khz; + kbps = timing->pix_clk_100hz / 10; kbps *= bits_per_channel; if (timing->flags.Y_ONLY != 1) { @@ -1461,7 +1593,7 @@ bool dp_validate_mode_timing( const struct dc_link_settings *link_setting; /*always DP fail safe mode*/ - if (timing->pix_clk_khz == (uint32_t) 25175 && + if ((timing->pix_clk_100hz / 10) == (uint32_t) 25175 && timing->h_addressable == (uint32_t) 640 && timing->v_addressable == (uint32_t) 480) return true; @@ -1511,7 +1643,7 @@ void decide_link_settings(struct dc_stream_state *stream, req_bw = bandwidth_in_kbps_from_timing(&stream->timing); - link = stream->sink->link; + link = stream->link; /* if preferred is specified through AMDDP, use it, if it's enough * to drive the mode @@ -1533,7 +1665,7 @@ void decide_link_settings(struct dc_stream_state *stream, } /* EDP use the link cap setting */ - if (stream->sink->sink_signal == SIGNAL_TYPE_EDP) { + if (link->connector_signal == SIGNAL_TYPE_EDP) { *link_setting = link->verified_link_cap; return; } @@ -1572,122 +1704,6 @@ void decide_link_settings(struct dc_stream_state *stream, } /*************************Short Pulse IRQ***************************/ - -static bool hpd_rx_irq_check_link_loss_status( - struct dc_link *link, - union hpd_irq_data *hpd_irq_dpcd_data) -{ - uint8_t irq_reg_rx_power_state = 0; - enum dc_status dpcd_result = DC_ERROR_UNEXPECTED; - union lane_status lane_status; - uint32_t lane; - bool sink_status_changed; - bool return_code; - - sink_status_changed = false; - return_code = false; - - if (link->cur_link_settings.lane_count == 0) - return return_code; - - /*1. Check that Link Status changed, before re-training.*/ - - /*parse lane status*/ - for (lane = 0; lane < link->cur_link_settings.lane_count; lane++) { - /* check status of lanes 0,1 - * changed DpcdAddress_Lane01Status (0x202) - */ - lane_status.raw = get_nibble_at_index( - &hpd_irq_dpcd_data->bytes.lane01_status.raw, - lane); - - if (!lane_status.bits.CHANNEL_EQ_DONE_0 || - !lane_status.bits.CR_DONE_0 || - !lane_status.bits.SYMBOL_LOCKED_0) { - /* if one of the channel equalization, clock - * recovery or symbol lock is dropped - * consider it as (link has been - * dropped) dp sink status has changed - */ - sink_status_changed = true; - break; - } - } - - /* Check interlane align.*/ - if (sink_status_changed || - !hpd_irq_dpcd_data->bytes.lane_status_updated.bits.INTERLANE_ALIGN_DONE) { - - DC_LOG_HW_HPD_IRQ("%s: Link Status changed.\n", __func__); - - return_code = true; - - /*2. Check that we can handle interrupt: Not in FS DOS, - * Not in "Display Timeout" state, Link is trained. - */ - dpcd_result = core_link_read_dpcd(link, - DP_SET_POWER, - &irq_reg_rx_power_state, - sizeof(irq_reg_rx_power_state)); - - if (dpcd_result != DC_OK) { - DC_LOG_HW_HPD_IRQ("%s: DPCD read failed to obtain power state.\n", - __func__); - } else { - if (irq_reg_rx_power_state != DP_SET_POWER_D0) - return_code = false; - } - } - - return return_code; -} - -static enum dc_status read_hpd_rx_irq_data( - struct dc_link *link, - union hpd_irq_data *irq_data) -{ - static enum dc_status retval; - - /* The HW reads 16 bytes from 200h on HPD, - * but if we get an AUX_DEFER, the HW cannot retry - * and this causes the CTS tests 4.3.2.1 - 3.2.4 to - * fail, so we now explicitly read 6 bytes which is - * the req from the above mentioned test cases. - * - * For DP 1.4 we need to read those from 2002h range. - */ - if (link->dpcd_caps.dpcd_rev.raw < DPCD_REV_14) - retval = core_link_read_dpcd( - link, - DP_SINK_COUNT, - irq_data->raw, - sizeof(union hpd_irq_data)); - else { - /* Read 14 bytes in a single read and then copy only the required fields. - * This is more efficient than doing it in two separate AUX reads. */ - - uint8_t tmp[DP_SINK_STATUS_ESI - DP_SINK_COUNT_ESI + 1]; - - retval = core_link_read_dpcd( - link, - DP_SINK_COUNT_ESI, - tmp, - sizeof(tmp)); - - if (retval != DC_OK) - return retval; - - irq_data->bytes.sink_cnt.raw = tmp[DP_SINK_COUNT_ESI - DP_SINK_COUNT_ESI]; - irq_data->bytes.device_service_irq.raw = tmp[DP_DEVICE_SERVICE_IRQ_VECTOR_ESI0 - DP_SINK_COUNT_ESI]; - irq_data->bytes.lane01_status.raw = tmp[DP_LANE0_1_STATUS_ESI - DP_SINK_COUNT_ESI]; - irq_data->bytes.lane23_status.raw = tmp[DP_LANE2_3_STATUS_ESI - DP_SINK_COUNT_ESI]; - irq_data->bytes.lane_status_updated.raw = tmp[DP_LANE_ALIGN_STATUS_UPDATED_ESI - DP_SINK_COUNT_ESI]; - irq_data->bytes.sink_status.raw = tmp[DP_SINK_STATUS_ESI - DP_SINK_COUNT_ESI]; - } - - return retval; -} - static bool allow_hpd_rx_irq(const struct dc_link *link) { /* @@ -1995,11 +2011,7 @@ static void handle_automated_test(struct dc_link *link) dp_test_send_phy_test_pattern(link); test_response.bits.ACK = 1; } - if (!test_request.raw) - /* no requests, revert all test signals - * TODO: revert all test signals - */ - test_response.bits.ACK = 1; + /* send request acknowledgment */ if (test_response.bits.ACK) core_link_write_dpcd( @@ -2196,7 +2208,7 @@ static void get_active_converter_info( } if (link->dpcd_caps.dpcd_rev.raw >= DPCD_REV_11) { - uint8_t det_caps[4]; + uint8_t det_caps[16]; /* CTS 4.2.2.7 expects source to read Detailed Capabilities Info : 00080h-0008F.*/ union dwnstream_port_caps_byte0 *port_caps = (union dwnstream_port_caps_byte0 *)det_caps; core_link_read_dpcd(link, DP_DOWNSTREAM_PORT_0, @@ -2240,7 +2252,8 @@ static void get_active_converter_info( translate_dpcd_max_bpc( hdmi_color_caps.bits.MAX_BITS_PER_COLOR_COMPONENT); - link->dpcd_caps.dongle_caps.extendedCapValid = true; + if (link->dpcd_caps.dongle_caps.dp_hdmi_max_pixel_clk != 0) + link->dpcd_caps.dongle_caps.extendedCapValid = true; } break; @@ -2371,11 +2384,22 @@ static bool retrieve_link_cap(struct dc_link *link) dpcd_data[DP_TRAINING_AUX_RD_INTERVAL]; if (aux_rd_interval.bits.EXT_RECIEVER_CAP_FIELD_PRESENT == 1) { - core_link_read_dpcd( + uint8_t ext_cap_data[16]; + + memset(ext_cap_data, '\0', sizeof(ext_cap_data)); + for (i = 0; i < read_dpcd_retry_cnt; i++) { + status = core_link_read_dpcd( link, DP_DP13_DPCD_REV, - dpcd_data, - sizeof(dpcd_data)); + ext_cap_data, + sizeof(ext_cap_data)); + if (status == DC_OK) { + memcpy(dpcd_data, ext_cap_data, sizeof(dpcd_data)); + break; + } + } + if (status != DC_OK) + dm_error("%s: Read extend caps data failed, use cap from dpcd 0.\n", __func__); } } @@ -2474,13 +2498,72 @@ bool detect_dp_sink_caps(struct dc_link *link) /* TODO save sink caps in link->sink */ } +enum dc_link_rate linkRateInKHzToLinkRateMultiplier(uint32_t link_rate_in_khz) +{ + enum dc_link_rate link_rate; + // LinkRate is normally stored as a multiplier of 0.27 Gbps per lane. Do the translation. + switch (link_rate_in_khz) { + case 1620000: + link_rate = LINK_RATE_LOW; // Rate_1 (RBR) - 1.62 Gbps/Lane + break; + case 2160000: + link_rate = LINK_RATE_RATE_2; // Rate_2 - 2.16 Gbps/Lane + break; + case 2430000: + link_rate = LINK_RATE_RATE_3; // Rate_3 - 2.43 Gbps/Lane + break; + case 2700000: + link_rate = LINK_RATE_HIGH; // Rate_4 (HBR) - 2.70 Gbps/Lane + break; + case 3240000: + link_rate = LINK_RATE_RBR2; // Rate_5 (RBR2) - 3.24 Gbps/Lane + break; + case 4320000: + link_rate = LINK_RATE_RATE_6; // Rate_6 - 4.32 Gbps/Lane + break; + case 5400000: + link_rate = LINK_RATE_HIGH2; // Rate_7 (HBR2) - 5.40 Gbps/Lane + break; + case 8100000: + link_rate = LINK_RATE_HIGH3; // Rate_8 (HBR3) - 8.10 Gbps/Lane + break; + default: + link_rate = LINK_RATE_UNKNOWN; + break; + } + return link_rate; +} + void detect_edp_sink_caps(struct dc_link *link) { - retrieve_link_cap(link); + uint8_t supported_link_rates[16] = {0}; + uint32_t entry; + uint32_t link_rate_in_khz; + enum dc_link_rate link_rate = LINK_RATE_UNKNOWN; - if (link->reported_link_cap.link_rate == LINK_RATE_UNKNOWN) - link->reported_link_cap.link_rate = LINK_RATE_HIGH2; + retrieve_link_cap(link); + if (link->dpcd_caps.dpcd_rev.raw >= DPCD_REV_14) { + // Read DPCD 00010h - 0001Fh 16 bytes at one shot + core_link_read_dpcd(link, DP_SUPPORTED_LINK_RATES, + supported_link_rates, sizeof(supported_link_rates)); + + link->dpcd_caps.link_rate_set = 0; + for (entry = 0; entry < 16; entry += 2) { + // DPCD register reports per-lane link rate = 16-bit link rate capability + // value X 200 kHz. Need multipler to find link rate in kHz. + link_rate_in_khz = (supported_link_rates[entry+1] * 0x100 + + supported_link_rates[entry]) * 200; + + if (link_rate_in_khz != 0) { + link_rate = linkRateInKHzToLinkRateMultiplier(link_rate_in_khz); + if (link->reported_link_cap.link_rate < link_rate) { + link->reported_link_cap.link_rate = link_rate; + link->dpcd_caps.link_rate_set = entry; + } + } + } + } link->verified_link_cap = link->reported_link_cap; } @@ -2602,7 +2685,7 @@ bool dc_link_dp_set_test_pattern( memset(&training_pattern, 0, sizeof(training_pattern)); for (i = 0; i < MAX_PIPES; i++) { - if (pipes[i].stream->sink->link == link) { + if (pipes[i].stream->link == link) { pipe_ctx = &pipes[i]; break; } diff --git a/drivers/gpu/drm/amd/display/dc/core/dc_link_hwss.c b/drivers/gpu/drm/amd/display/dc/core/dc_link_hwss.c index 82cd1d6e6e59..f7f7515f65f4 100644 --- a/drivers/gpu/drm/amd/display/dc/core/dc_link_hwss.c +++ b/drivers/gpu/drm/amd/display/dc/core/dc_link_hwss.c @@ -70,13 +70,12 @@ void dp_enable_link_phy( */ for (i = 0; i < MAX_PIPES; i++) { if (pipes[i].stream != NULL && - pipes[i].stream->sink != NULL && - pipes[i].stream->sink->link == link) { + pipes[i].stream->link == link) { if (pipes[i].clock_source != NULL && pipes[i].clock_source->id != CLOCK_SOURCE_ID_DP_DTO) { pipes[i].clock_source = dp_cs; - pipes[i].stream_res.pix_clk_params.requested_pix_clk = - pipes[i].stream->timing.pix_clk_khz; + pipes[i].stream_res.pix_clk_params.requested_pix_clk_100hz = + pipes[i].stream->timing.pix_clk_100hz; pipes[i].clock_source->funcs->program_pix_clk( pipes[i].clock_source, &pipes[i].stream_res.pix_clk_params, @@ -96,6 +95,7 @@ void dp_enable_link_phy( link_settings, clock_source); } + link->cur_link_settings = *link_settings; dp_receiver_power_ctrl(link, true); } @@ -119,6 +119,10 @@ bool edp_receiver_ready_T9(struct dc_link *link) break; udelay(100); //MAx T9 } while (++tries < 50); + + if (link->local_sink->edid_caps.panel_patch.extra_delay_backlight_off > 0) + udelay(link->local_sink->edid_caps.panel_patch.extra_delay_backlight_off * 1000); + return result; } bool edp_receiver_ready_T7(struct dc_link *link) @@ -278,10 +282,8 @@ void dp_retrain_link_dp_test(struct dc_link *link, for (i = 0; i < MAX_PIPES; i++) { if (pipes[i].stream != NULL && !pipes[i].top_pipe && - pipes[i].stream->sink != NULL && - pipes[i].stream->sink->link != NULL && - pipes[i].stream_res.stream_enc != NULL && - pipes[i].stream->sink->link == link) { + pipes[i].stream->link != NULL && + pipes[i].stream_res.stream_enc != NULL) { udelay(100); pipes[i].stream_res.stream_enc->funcs->dp_blank( @@ -307,6 +309,7 @@ void dp_retrain_link_dp_test(struct dc_link *link, link->link_enc, link_setting, pipes[i].clock_source->id); + link->cur_link_settings = *link_setting; dp_receiver_power_ctrl(link, true); @@ -316,7 +319,6 @@ void dp_retrain_link_dp_test(struct dc_link *link, skip_video_pattern, LINK_TRAINING_ATTEMPTS); - link->cur_link_settings = *link_setting; link->dc->hwss.enable_stream(&pipes[i]); diff --git a/drivers/gpu/drm/amd/display/dc/core/dc_resource.c b/drivers/gpu/drm/amd/display/dc/core/dc_resource.c index b6fe29b9fb65..349ab8017776 100644 --- a/drivers/gpu/drm/amd/display/dc/core/dc_resource.c +++ b/drivers/gpu/drm/amd/display/dc/core/dc_resource.c @@ -83,7 +83,10 @@ enum dce_version resource_parse_asic_id(struct hw_asic_id asic_id) dc_version = DCE_VERSION_11_22; break; case FAMILY_AI: - dc_version = DCE_VERSION_12_0; + if (ASICREV_IS_VEGA20_P(asic_id.hw_internal_rev)) + dc_version = DCE_VERSION_12_1; + else + dc_version = DCE_VERSION_12_0; break; #if defined(CONFIG_DRM_AMD_DC_DCN1_0) case FAMILY_RV: @@ -136,6 +139,7 @@ struct resource_pool *dc_create_resource_pool( num_virtual_links, dc); break; case DCE_VERSION_12_0: + case DCE_VERSION_12_1: res_pool = dce120_create_resource_pool( num_virtual_links, dc); break; @@ -351,8 +355,8 @@ bool resource_are_streams_timing_synchronizable( != stream2->timing.v_addressable) return false; - if (stream1->timing.pix_clk_khz - != stream2->timing.pix_clk_khz) + if (stream1->timing.pix_clk_100hz + != stream2->timing.pix_clk_100hz) return false; if (stream1->clamping.c_depth != stream2->clamping.c_depth) @@ -478,10 +482,29 @@ static enum pixel_format convert_pixel_format_to_dalsurface( return dal_pixel_format; } -static void rect_swap_helper(struct rect *rect) +static inline void get_vp_scan_direction( + enum dc_rotation_angle rotation, + bool horizontal_mirror, + bool *orthogonal_rotation, + bool *flip_vert_scan_dir, + bool *flip_horz_scan_dir) { - swap(rect->height, rect->width); - swap(rect->x, rect->y); + *orthogonal_rotation = false; + *flip_vert_scan_dir = false; + *flip_horz_scan_dir = false; + if (rotation == ROTATION_ANGLE_180) { + *flip_vert_scan_dir = true; + *flip_horz_scan_dir = true; + } else if (rotation == ROTATION_ANGLE_90) { + *orthogonal_rotation = true; + *flip_horz_scan_dir = true; + } else if (rotation == ROTATION_ANGLE_270) { + *orthogonal_rotation = true; + *flip_vert_scan_dir = true; + } + + if (horizontal_mirror) + *flip_horz_scan_dir = !*flip_horz_scan_dir; } static void calculate_viewport(struct pipe_ctx *pipe_ctx) @@ -490,25 +513,14 @@ static void calculate_viewport(struct pipe_ctx *pipe_ctx) const struct dc_stream_state *stream = pipe_ctx->stream; struct scaler_data *data = &pipe_ctx->plane_res.scl_data; struct rect surf_src = plane_state->src_rect; - struct rect clip = { 0 }; + struct rect clip, dest; int vpc_div = (data->format == PIXEL_FORMAT_420BPP8 || data->format == PIXEL_FORMAT_420BPP10) ? 2 : 1; bool pri_split = pipe_ctx->bottom_pipe && pipe_ctx->bottom_pipe->plane_state == pipe_ctx->plane_state; bool sec_split = pipe_ctx->top_pipe && pipe_ctx->top_pipe->plane_state == pipe_ctx->plane_state; - bool flip_vert_scan_dir = false, flip_horz_scan_dir = false; - - /* - * Need to calculate the scan direction for viewport to properly determine offset - */ - if (pipe_ctx->plane_state->rotation == ROTATION_ANGLE_180) { - flip_vert_scan_dir = true; - flip_horz_scan_dir = true; - } else if (pipe_ctx->plane_state->rotation == ROTATION_ANGLE_90) - flip_vert_scan_dir = true; - else if (pipe_ctx->plane_state->rotation == ROTATION_ANGLE_270) - flip_horz_scan_dir = true; + bool orthogonal_rotation, flip_y_start, flip_x_start; if (stream->view_format == VIEW_3D_FORMAT_SIDE_BY_SIDE || stream->view_format == VIEW_3D_FORMAT_TOP_AND_BOTTOM) { @@ -516,13 +528,10 @@ static void calculate_viewport(struct pipe_ctx *pipe_ctx) sec_split = false; } - if (pipe_ctx->plane_state->rotation == ROTATION_ANGLE_90 || - pipe_ctx->plane_state->rotation == ROTATION_ANGLE_270) - rect_swap_helper(&surf_src); - /* The actual clip is an intersection between stream * source and surface clip */ + dest = plane_state->dst_rect; clip.x = stream->src.x > plane_state->clip_rect.x ? stream->src.x : plane_state->clip_rect.x; @@ -539,84 +548,77 @@ static void calculate_viewport(struct pipe_ctx *pipe_ctx) stream->src.y + stream->src.height - clip.y : plane_state->clip_rect.y + plane_state->clip_rect.height - clip.y ; + /* + * Need to calculate how scan origin is shifted in vp space + * to correctly rotate clip and dst + */ + get_vp_scan_direction( + plane_state->rotation, + plane_state->horizontal_mirror, + &orthogonal_rotation, + &flip_y_start, + &flip_x_start); + + if (orthogonal_rotation) { + swap(clip.x, clip.y); + swap(clip.width, clip.height); + swap(dest.x, dest.y); + swap(dest.width, dest.height); + } + if (flip_x_start) { + clip.x = dest.x + dest.width - clip.x - clip.width; + dest.x = 0; + } + if (flip_y_start) { + clip.y = dest.y + dest.height - clip.y - clip.height; + dest.y = 0; + } + /* offset = surf_src.ofs + (clip.ofs - surface->dst_rect.ofs) * scl_ratio * num_pixels = clip.num_pix * scl_ratio */ - data->viewport.x = surf_src.x + (clip.x - plane_state->dst_rect.x) * - surf_src.width / plane_state->dst_rect.width; - data->viewport.width = clip.width * - surf_src.width / plane_state->dst_rect.width; - - data->viewport.y = surf_src.y + (clip.y - plane_state->dst_rect.y) * - surf_src.height / plane_state->dst_rect.height; - data->viewport.height = clip.height * - surf_src.height / plane_state->dst_rect.height; - - /* To transfer the x, y to correct coordinate on mirror image (camera). - * deg 0 : transfer x, - * deg 90 : don't need to transfer, - * deg180 : transfer y, - * deg270 : transfer x and y. - * To transfer the x, y to correct coordinate on non-mirror image (video). - * deg 0 : don't need to transfer, - * deg 90 : transfer y, - * deg180 : transfer x and y, - * deg270 : transfer x. - */ - if (pipe_ctx->plane_state->horizontal_mirror) { - if (flip_horz_scan_dir && !flip_vert_scan_dir) { - data->viewport.y = surf_src.height - data->viewport.y - data->viewport.height; - data->viewport.x = surf_src.width - data->viewport.x - data->viewport.width; - } else if (flip_horz_scan_dir && flip_vert_scan_dir) - data->viewport.y = surf_src.height - data->viewport.y - data->viewport.height; - else { - if (!flip_horz_scan_dir && !flip_vert_scan_dir) - data->viewport.x = surf_src.width - data->viewport.x - data->viewport.width; + data->viewport.x = surf_src.x + (clip.x - dest.x) * surf_src.width / dest.width; + data->viewport.width = clip.width * surf_src.width / dest.width; + + data->viewport.y = surf_src.y + (clip.y - dest.y) * surf_src.height / dest.height; + data->viewport.height = clip.height * surf_src.height / dest.height; + + /* Handle split */ + if (pri_split || sec_split) { + if (orthogonal_rotation) { + if (flip_y_start != pri_split) + data->viewport.height /= 2; + else { + data->viewport.y += data->viewport.height / 2; + /* Ceil offset pipe */ + data->viewport.height = (data->viewport.height + 1) / 2; + } + } else { + if (flip_x_start != pri_split) + data->viewport.width /= 2; + else { + data->viewport.x += data->viewport.width / 2; + /* Ceil offset pipe */ + data->viewport.width = (data->viewport.width + 1) / 2; + } } - } else { - if (flip_horz_scan_dir) - data->viewport.x = surf_src.width - data->viewport.x - data->viewport.width; - if (flip_vert_scan_dir) - data->viewport.y = surf_src.height - data->viewport.y - data->viewport.height; } /* Round down, compensate in init */ data->viewport_c.x = data->viewport.x / vpc_div; data->viewport_c.y = data->viewport.y / vpc_div; - data->inits.h_c = (data->viewport.x % vpc_div) != 0 ? - dc_fixpt_half : dc_fixpt_zero; - data->inits.v_c = (data->viewport.y % vpc_div) != 0 ? - dc_fixpt_half : dc_fixpt_zero; + data->inits.h_c = (data->viewport.x % vpc_div) != 0 ? dc_fixpt_half : dc_fixpt_zero; + data->inits.v_c = (data->viewport.y % vpc_div) != 0 ? dc_fixpt_half : dc_fixpt_zero; + /* Round up, assume original video size always even dimensions */ data->viewport_c.width = (data->viewport.width + vpc_div - 1) / vpc_div; data->viewport_c.height = (data->viewport.height + vpc_div - 1) / vpc_div; - - /* Handle hsplit */ - if (sec_split) { - data->viewport.x += data->viewport.width / 2; - data->viewport_c.x += data->viewport_c.width / 2; - /* Ceil offset pipe */ - data->viewport.width = (data->viewport.width + 1) / 2; - data->viewport_c.width = (data->viewport_c.width + 1) / 2; - } else if (pri_split) { - if (data->viewport.width > 1) - data->viewport.width /= 2; - if (data->viewport_c.width > 1) - data->viewport_c.width /= 2; - } - - if (plane_state->rotation == ROTATION_ANGLE_90 || - plane_state->rotation == ROTATION_ANGLE_270) { - rect_swap_helper(&data->viewport_c); - rect_swap_helper(&data->viewport); - } } -static void calculate_recout(struct pipe_ctx *pipe_ctx, struct rect *recout_full) +static void calculate_recout(struct pipe_ctx *pipe_ctx) { const struct dc_plane_state *plane_state = pipe_ctx->plane_state; const struct dc_stream_state *stream = pipe_ctx->stream; - struct rect surf_src = plane_state->src_rect; struct rect surf_clip = plane_state->clip_rect; bool pri_split = pipe_ctx->bottom_pipe && pipe_ctx->bottom_pipe->plane_state == pipe_ctx->plane_state; @@ -624,10 +626,6 @@ static void calculate_recout(struct pipe_ctx *pipe_ctx, struct rect *recout_full pipe_ctx->top_pipe->plane_state == pipe_ctx->plane_state; bool top_bottom_split = stream->view_format == VIEW_3D_FORMAT_TOP_AND_BOTTOM; - if (pipe_ctx->plane_state->rotation == ROTATION_ANGLE_90 || - pipe_ctx->plane_state->rotation == ROTATION_ANGLE_270) - rect_swap_helper(&surf_src); - pipe_ctx->plane_res.scl_data.recout.x = stream->dst.x; if (stream->src.x < surf_clip.x) pipe_ctx->plane_res.scl_data.recout.x += (surf_clip.x @@ -656,7 +654,7 @@ static void calculate_recout(struct pipe_ctx *pipe_ctx, struct rect *recout_full stream->dst.y + stream->dst.height - pipe_ctx->plane_res.scl_data.recout.y; - /* Handle h & vsplit */ + /* Handle h & v split, handle rotation using viewport */ if (sec_split && top_bottom_split) { pipe_ctx->plane_res.scl_data.recout.y += pipe_ctx->plane_res.scl_data.recout.height / 2; @@ -665,44 +663,14 @@ static void calculate_recout(struct pipe_ctx *pipe_ctx, struct rect *recout_full (pipe_ctx->plane_res.scl_data.recout.height + 1) / 2; } else if (pri_split && top_bottom_split) pipe_ctx->plane_res.scl_data.recout.height /= 2; - else if (pri_split || sec_split) { - /* HMirror XOR Secondary_pipe XOR Rotation_180 */ - bool right_view = (sec_split != plane_state->horizontal_mirror) != - (plane_state->rotation == ROTATION_ANGLE_180); - - if (plane_state->rotation == ROTATION_ANGLE_90 - || plane_state->rotation == ROTATION_ANGLE_270) - /* Secondary_pipe XOR Rotation_270 */ - right_view = (plane_state->rotation == ROTATION_ANGLE_270) != sec_split; - - if (right_view) { - pipe_ctx->plane_res.scl_data.recout.x += - pipe_ctx->plane_res.scl_data.recout.width / 2; - /* Ceil offset pipe */ - pipe_ctx->plane_res.scl_data.recout.width = - (pipe_ctx->plane_res.scl_data.recout.width + 1) / 2; - } else { - if (pipe_ctx->plane_res.scl_data.recout.width > 1) - pipe_ctx->plane_res.scl_data.recout.width /= 2; - } - } - /* Unclipped recout offset = stream dst offset + ((surf dst offset - stream surf_src offset) - * * 1/ stream scaling ratio) - (surf surf_src offset * 1/ full scl - * ratio) - */ - recout_full->x = stream->dst.x + (plane_state->dst_rect.x - stream->src.x) - * stream->dst.width / stream->src.width - - surf_src.x * plane_state->dst_rect.width / surf_src.width - * stream->dst.width / stream->src.width; - recout_full->y = stream->dst.y + (plane_state->dst_rect.y - stream->src.y) - * stream->dst.height / stream->src.height - - surf_src.y * plane_state->dst_rect.height / surf_src.height - * stream->dst.height / stream->src.height; - - recout_full->width = plane_state->dst_rect.width - * stream->dst.width / stream->src.width; - recout_full->height = plane_state->dst_rect.height - * stream->dst.height / stream->src.height; + else if (sec_split) { + pipe_ctx->plane_res.scl_data.recout.x += + pipe_ctx->plane_res.scl_data.recout.width / 2; + /* Ceil offset pipe */ + pipe_ctx->plane_res.scl_data.recout.width = + (pipe_ctx->plane_res.scl_data.recout.width + 1) / 2; + } else if (pri_split) + pipe_ctx->plane_res.scl_data.recout.width /= 2; } static void calculate_scaling_ratios(struct pipe_ctx *pipe_ctx) @@ -715,9 +683,10 @@ static void calculate_scaling_ratios(struct pipe_ctx *pipe_ctx) const int out_w = stream->dst.width; const int out_h = stream->dst.height; + /*Swap surf_src height and width since scaling ratios are in recout rotation*/ if (pipe_ctx->plane_state->rotation == ROTATION_ANGLE_90 || pipe_ctx->plane_state->rotation == ROTATION_ANGLE_270) - rect_swap_helper(&surf_src); + swap(surf_src.height, surf_src.width); pipe_ctx->plane_res.scl_data.ratios.horz = dc_fixpt_from_fraction( surf_src.width, @@ -754,358 +723,202 @@ static void calculate_scaling_ratios(struct pipe_ctx *pipe_ctx) pipe_ctx->plane_res.scl_data.ratios.vert_c, 19); } -static void calculate_inits_and_adj_vp(struct pipe_ctx *pipe_ctx, struct rect *recout_full) +static inline void adjust_vp_and_init_for_seamless_clip( + bool flip_scan_dir, + int recout_skip, + int src_size, + int taps, + struct fixed31_32 ratio, + struct fixed31_32 *init, + int *vp_offset, + int *vp_size) { - struct scaler_data *data = &pipe_ctx->plane_res.scl_data; - struct rect src = pipe_ctx->plane_state->src_rect; - int vpc_div = (data->format == PIXEL_FORMAT_420BPP8 - || data->format == PIXEL_FORMAT_420BPP10) ? 2 : 1; - bool flip_vert_scan_dir = false, flip_horz_scan_dir = false; - - /* - * Need to calculate the scan direction for viewport to make adjustments - */ - if (pipe_ctx->plane_state->rotation == ROTATION_ANGLE_180) { - flip_vert_scan_dir = true; - flip_horz_scan_dir = true; - } else if (pipe_ctx->plane_state->rotation == ROTATION_ANGLE_90) - flip_vert_scan_dir = true; - else if (pipe_ctx->plane_state->rotation == ROTATION_ANGLE_270) - flip_horz_scan_dir = true; - - if (pipe_ctx->plane_state->rotation == ROTATION_ANGLE_90 || - pipe_ctx->plane_state->rotation == ROTATION_ANGLE_270) { - rect_swap_helper(&src); - rect_swap_helper(&data->viewport_c); - rect_swap_helper(&data->viewport); - - if (pipe_ctx->plane_state->rotation == ROTATION_ANGLE_270 && - pipe_ctx->plane_state->horizontal_mirror) { - flip_vert_scan_dir = true; - } - if (pipe_ctx->plane_state->rotation == ROTATION_ANGLE_90 && - pipe_ctx->plane_state->horizontal_mirror) { - flip_vert_scan_dir = false; - } - } else if (pipe_ctx->plane_state->horizontal_mirror) - flip_horz_scan_dir = !flip_horz_scan_dir; - - /* - * Init calculated according to formula: - * init = (scaling_ratio + number_of_taps + 1) / 2 - * init_bot = init + scaling_ratio - * init_c = init + truncated_vp_c_offset(from calculate viewport) - */ - data->inits.h = dc_fixpt_truncate(dc_fixpt_div_int( - dc_fixpt_add_int(data->ratios.horz, data->taps.h_taps + 1), 2), 19); - - data->inits.h_c = dc_fixpt_truncate(dc_fixpt_add(data->inits.h_c, dc_fixpt_div_int( - dc_fixpt_add_int(data->ratios.horz_c, data->taps.h_taps_c + 1), 2)), 19); - - data->inits.v = dc_fixpt_truncate(dc_fixpt_div_int( - dc_fixpt_add_int(data->ratios.vert, data->taps.v_taps + 1), 2), 19); - - data->inits.v_c = dc_fixpt_truncate(dc_fixpt_add(data->inits.v_c, dc_fixpt_div_int( - dc_fixpt_add_int(data->ratios.vert_c, data->taps.v_taps_c + 1), 2)), 19); - - if (!flip_horz_scan_dir) { + if (!flip_scan_dir) { /* Adjust for viewport end clip-off */ - if ((data->viewport.x + data->viewport.width) < (src.x + src.width)) { - int vp_clip = src.x + src.width - data->viewport.width - data->viewport.x; - int int_part = dc_fixpt_floor( - dc_fixpt_sub(data->inits.h, data->ratios.horz)); + if ((*vp_offset + *vp_size) < src_size) { + int vp_clip = src_size - *vp_size - *vp_offset; + int int_part = dc_fixpt_floor(dc_fixpt_sub(*init, ratio)); int_part = int_part > 0 ? int_part : 0; - data->viewport.width += int_part < vp_clip ? int_part : vp_clip; - } - if ((data->viewport_c.x + data->viewport_c.width) < (src.x + src.width) / vpc_div) { - int vp_clip = (src.x + src.width) / vpc_div - - data->viewport_c.width - data->viewport_c.x; - int int_part = dc_fixpt_floor( - dc_fixpt_sub(data->inits.h_c, data->ratios.horz_c)); - - int_part = int_part > 0 ? int_part : 0; - data->viewport_c.width += int_part < vp_clip ? int_part : vp_clip; + *vp_size += int_part < vp_clip ? int_part : vp_clip; } /* Adjust for non-0 viewport offset */ - if (data->viewport.x) { - int int_part; - - data->inits.h = dc_fixpt_add(data->inits.h, dc_fixpt_mul_int( - data->ratios.horz, data->recout.x - recout_full->x)); - int_part = dc_fixpt_floor(data->inits.h) - data->viewport.x; - if (int_part < data->taps.h_taps) { - int int_adj = data->viewport.x >= (data->taps.h_taps - int_part) ? - (data->taps.h_taps - int_part) : data->viewport.x; - data->viewport.x -= int_adj; - data->viewport.width += int_adj; - int_part += int_adj; - } else if (int_part > data->taps.h_taps) { - data->viewport.x += int_part - data->taps.h_taps; - data->viewport.width -= int_part - data->taps.h_taps; - int_part = data->taps.h_taps; - } - data->inits.h.value &= 0xffffffff; - data->inits.h = dc_fixpt_add_int(data->inits.h, int_part); - } - - if (data->viewport_c.x) { + if (*vp_offset) { int int_part; - data->inits.h_c = dc_fixpt_add(data->inits.h_c, dc_fixpt_mul_int( - data->ratios.horz_c, data->recout.x - recout_full->x)); - int_part = dc_fixpt_floor(data->inits.h_c) - data->viewport_c.x; - if (int_part < data->taps.h_taps_c) { - int int_adj = data->viewport_c.x >= (data->taps.h_taps_c - int_part) ? - (data->taps.h_taps_c - int_part) : data->viewport_c.x; - data->viewport_c.x -= int_adj; - data->viewport_c.width += int_adj; + *init = dc_fixpt_add(*init, dc_fixpt_mul_int(ratio, recout_skip)); + int_part = dc_fixpt_floor(*init) - *vp_offset; + if (int_part < taps) { + int int_adj = *vp_offset >= (taps - int_part) ? + (taps - int_part) : *vp_offset; + *vp_offset -= int_adj; + *vp_size += int_adj; int_part += int_adj; - } else if (int_part > data->taps.h_taps_c) { - data->viewport_c.x += int_part - data->taps.h_taps_c; - data->viewport_c.width -= int_part - data->taps.h_taps_c; - int_part = data->taps.h_taps_c; + } else if (int_part > taps) { + *vp_offset += int_part - taps; + *vp_size -= int_part - taps; + int_part = taps; } - data->inits.h_c.value &= 0xffffffff; - data->inits.h_c = dc_fixpt_add_int(data->inits.h_c, int_part); + init->value &= 0xffffffff; + *init = dc_fixpt_add_int(*init, int_part); } } else { /* Adjust for non-0 viewport offset */ - if (data->viewport.x) { - int int_part = dc_fixpt_floor( - dc_fixpt_sub(data->inits.h, data->ratios.horz)); - - int_part = int_part > 0 ? int_part : 0; - data->viewport.width += int_part < data->viewport.x ? int_part : data->viewport.x; - data->viewport.x -= int_part < data->viewport.x ? int_part : data->viewport.x; - } - if (data->viewport_c.x) { - int int_part = dc_fixpt_floor( - dc_fixpt_sub(data->inits.h_c, data->ratios.horz_c)); + if (*vp_offset) { + int int_part = dc_fixpt_floor(dc_fixpt_sub(*init, ratio)); int_part = int_part > 0 ? int_part : 0; - data->viewport_c.width += int_part < data->viewport_c.x ? int_part : data->viewport_c.x; - data->viewport_c.x -= int_part < data->viewport_c.x ? int_part : data->viewport_c.x; + *vp_size += int_part < *vp_offset ? int_part : *vp_offset; + *vp_offset -= int_part < *vp_offset ? int_part : *vp_offset; } /* Adjust for viewport end clip-off */ - if ((data->viewport.x + data->viewport.width) < (src.x + src.width)) { + if ((*vp_offset + *vp_size) < src_size) { int int_part; - int end_offset = src.x + src.width - - data->viewport.x - data->viewport.width; + int end_offset = src_size - *vp_offset - *vp_size; /* * this is init if vp had no offset, keep in mind this is from the * right side of vp due to scan direction */ - data->inits.h = dc_fixpt_add(data->inits.h, dc_fixpt_mul_int( - data->ratios.horz, data->recout.x - recout_full->x)); + *init = dc_fixpt_add(*init, dc_fixpt_mul_int(ratio, recout_skip)); /* * this is the difference between first pixel of viewport available to read * and init position, takning into account scan direction */ - int_part = dc_fixpt_floor(data->inits.h) - end_offset; - if (int_part < data->taps.h_taps) { - int int_adj = end_offset >= (data->taps.h_taps - int_part) ? - (data->taps.h_taps - int_part) : end_offset; - data->viewport.width += int_adj; + int_part = dc_fixpt_floor(*init) - end_offset; + if (int_part < taps) { + int int_adj = end_offset >= (taps - int_part) ? + (taps - int_part) : end_offset; + *vp_size += int_adj; int_part += int_adj; - } else if (int_part > data->taps.h_taps) { - data->viewport.width += int_part - data->taps.h_taps; - int_part = data->taps.h_taps; + } else if (int_part > taps) { + *vp_size += int_part - taps; + int_part = taps; } - data->inits.h.value &= 0xffffffff; - data->inits.h = dc_fixpt_add_int(data->inits.h, int_part); + init->value &= 0xffffffff; + *init = dc_fixpt_add_int(*init, int_part); } - - if ((data->viewport_c.x + data->viewport_c.width) < (src.x + src.width) / vpc_div) { - int int_part; - int end_offset = (src.x + src.width) / vpc_div - - data->viewport_c.x - data->viewport_c.width; - - /* - * this is init if vp had no offset, keep in mind this is from the - * right side of vp due to scan direction - */ - data->inits.h_c = dc_fixpt_add(data->inits.h_c, dc_fixpt_mul_int( - data->ratios.horz_c, data->recout.x - recout_full->x)); - /* - * this is the difference between first pixel of viewport available to read - * and init position, takning into account scan direction - */ - int_part = dc_fixpt_floor(data->inits.h_c) - end_offset; - if (int_part < data->taps.h_taps_c) { - int int_adj = end_offset >= (data->taps.h_taps_c - int_part) ? - (data->taps.h_taps_c - int_part) : end_offset; - data->viewport_c.width += int_adj; - int_part += int_adj; - } else if (int_part > data->taps.h_taps_c) { - data->viewport_c.width += int_part - data->taps.h_taps_c; - int_part = data->taps.h_taps_c; - } - data->inits.h_c.value &= 0xffffffff; - data->inits.h_c = dc_fixpt_add_int(data->inits.h_c, int_part); - } - } - if (!flip_vert_scan_dir) { - /* Adjust for viewport end clip-off */ - if ((data->viewport.y + data->viewport.height) < (src.y + src.height)) { - int vp_clip = src.y + src.height - data->viewport.height - data->viewport.y; - int int_part = dc_fixpt_floor( - dc_fixpt_sub(data->inits.v, data->ratios.vert)); - - int_part = int_part > 0 ? int_part : 0; - data->viewport.height += int_part < vp_clip ? int_part : vp_clip; - } - if ((data->viewport_c.y + data->viewport_c.height) < (src.y + src.height) / vpc_div) { - int vp_clip = (src.y + src.height) / vpc_div - - data->viewport_c.height - data->viewport_c.y; - int int_part = dc_fixpt_floor( - dc_fixpt_sub(data->inits.v_c, data->ratios.vert_c)); - - int_part = int_part > 0 ? int_part : 0; - data->viewport_c.height += int_part < vp_clip ? int_part : vp_clip; - } - - /* Adjust for non-0 viewport offset */ - if (data->viewport.y) { - int int_part; - - data->inits.v = dc_fixpt_add(data->inits.v, dc_fixpt_mul_int( - data->ratios.vert, data->recout.y - recout_full->y)); - int_part = dc_fixpt_floor(data->inits.v) - data->viewport.y; - if (int_part < data->taps.v_taps) { - int int_adj = data->viewport.y >= (data->taps.v_taps - int_part) ? - (data->taps.v_taps - int_part) : data->viewport.y; - data->viewport.y -= int_adj; - data->viewport.height += int_adj; - int_part += int_adj; - } else if (int_part > data->taps.v_taps) { - data->viewport.y += int_part - data->taps.v_taps; - data->viewport.height -= int_part - data->taps.v_taps; - int_part = data->taps.v_taps; - } - data->inits.v.value &= 0xffffffff; - data->inits.v = dc_fixpt_add_int(data->inits.v, int_part); - } - - if (data->viewport_c.y) { - int int_part; - - data->inits.v_c = dc_fixpt_add(data->inits.v_c, dc_fixpt_mul_int( - data->ratios.vert_c, data->recout.y - recout_full->y)); - int_part = dc_fixpt_floor(data->inits.v_c) - data->viewport_c.y; - if (int_part < data->taps.v_taps_c) { - int int_adj = data->viewport_c.y >= (data->taps.v_taps_c - int_part) ? - (data->taps.v_taps_c - int_part) : data->viewport_c.y; - data->viewport_c.y -= int_adj; - data->viewport_c.height += int_adj; - int_part += int_adj; - } else if (int_part > data->taps.v_taps_c) { - data->viewport_c.y += int_part - data->taps.v_taps_c; - data->viewport_c.height -= int_part - data->taps.v_taps_c; - int_part = data->taps.v_taps_c; - } - data->inits.v_c.value &= 0xffffffff; - data->inits.v_c = dc_fixpt_add_int(data->inits.v_c, int_part); - } - } else { - /* Adjust for non-0 viewport offset */ - if (data->viewport.y) { - int int_part = dc_fixpt_floor( - dc_fixpt_sub(data->inits.v, data->ratios.vert)); +} - int_part = int_part > 0 ? int_part : 0; - data->viewport.height += int_part < data->viewport.y ? int_part : data->viewport.y; - data->viewport.y -= int_part < data->viewport.y ? int_part : data->viewport.y; - } - if (data->viewport_c.y) { - int int_part = dc_fixpt_floor( - dc_fixpt_sub(data->inits.v_c, data->ratios.vert_c)); +static void calculate_inits_and_adj_vp(struct pipe_ctx *pipe_ctx) +{ + const struct dc_plane_state *plane_state = pipe_ctx->plane_state; + const struct dc_stream_state *stream = pipe_ctx->stream; + struct scaler_data *data = &pipe_ctx->plane_res.scl_data; + struct rect src = pipe_ctx->plane_state->src_rect; + int recout_skip_h, recout_skip_v, surf_size_h, surf_size_v; + int vpc_div = (data->format == PIXEL_FORMAT_420BPP8 + || data->format == PIXEL_FORMAT_420BPP10) ? 2 : 1; + bool orthogonal_rotation, flip_vert_scan_dir, flip_horz_scan_dir; - int_part = int_part > 0 ? int_part : 0; - data->viewport_c.height += int_part < data->viewport_c.y ? int_part : data->viewport_c.y; - data->viewport_c.y -= int_part < data->viewport_c.y ? int_part : data->viewport_c.y; - } + /* + * Need to calculate the scan direction for viewport to make adjustments + */ + get_vp_scan_direction( + plane_state->rotation, + plane_state->horizontal_mirror, + &orthogonal_rotation, + &flip_vert_scan_dir, + &flip_horz_scan_dir); + + /* Calculate src rect rotation adjusted to recout space */ + surf_size_h = src.x + src.width; + surf_size_v = src.y + src.height; + if (flip_horz_scan_dir) + src.x = 0; + if (flip_vert_scan_dir) + src.y = 0; + if (orthogonal_rotation) { + swap(src.x, src.y); + swap(src.width, src.height); + } + + /* Recout matching initial vp offset = recout_offset - (stream dst offset + + * ((surf dst offset - stream src offset) * 1/ stream scaling ratio) + * - (surf surf_src offset * 1/ full scl ratio)) + */ + recout_skip_h = data->recout.x - (stream->dst.x + (plane_state->dst_rect.x - stream->src.x) + * stream->dst.width / stream->src.width - + src.x * plane_state->dst_rect.width / src.width + * stream->dst.width / stream->src.width); + recout_skip_v = data->recout.y - (stream->dst.y + (plane_state->dst_rect.y - stream->src.y) + * stream->dst.height / stream->src.height - + src.y * plane_state->dst_rect.height / src.height + * stream->dst.height / stream->src.height); + if (orthogonal_rotation) + swap(recout_skip_h, recout_skip_v); + /* + * Init calculated according to formula: + * init = (scaling_ratio + number_of_taps + 1) / 2 + * init_bot = init + scaling_ratio + * init_c = init + truncated_vp_c_offset(from calculate viewport) + */ + data->inits.h = dc_fixpt_truncate(dc_fixpt_div_int( + dc_fixpt_add_int(data->ratios.horz, data->taps.h_taps + 1), 2), 19); - /* Adjust for viewport end clip-off */ - if ((data->viewport.y + data->viewport.height) < (src.y + src.height)) { - int int_part; - int end_offset = src.y + src.height - - data->viewport.y - data->viewport.height; + data->inits.h_c = dc_fixpt_truncate(dc_fixpt_add(data->inits.h_c, dc_fixpt_div_int( + dc_fixpt_add_int(data->ratios.horz_c, data->taps.h_taps_c + 1), 2)), 19); - /* - * this is init if vp had no offset, keep in mind this is from the - * right side of vp due to scan direction - */ - data->inits.v = dc_fixpt_add(data->inits.v, dc_fixpt_mul_int( - data->ratios.vert, data->recout.y - recout_full->y)); - /* - * this is the difference between first pixel of viewport available to read - * and init position, taking into account scan direction - */ - int_part = dc_fixpt_floor(data->inits.v) - end_offset; - if (int_part < data->taps.v_taps) { - int int_adj = end_offset >= (data->taps.v_taps - int_part) ? - (data->taps.v_taps - int_part) : end_offset; - data->viewport.height += int_adj; - int_part += int_adj; - } else if (int_part > data->taps.v_taps) { - data->viewport.height += int_part - data->taps.v_taps; - int_part = data->taps.v_taps; - } - data->inits.v.value &= 0xffffffff; - data->inits.v = dc_fixpt_add_int(data->inits.v, int_part); - } + data->inits.v = dc_fixpt_truncate(dc_fixpt_div_int( + dc_fixpt_add_int(data->ratios.vert, data->taps.v_taps + 1), 2), 19); - if ((data->viewport_c.y + data->viewport_c.height) < (src.y + src.height) / vpc_div) { - int int_part; - int end_offset = (src.y + src.height) / vpc_div - - data->viewport_c.y - data->viewport_c.height; + data->inits.v_c = dc_fixpt_truncate(dc_fixpt_add(data->inits.v_c, dc_fixpt_div_int( + dc_fixpt_add_int(data->ratios.vert_c, data->taps.v_taps_c + 1), 2)), 19); - /* - * this is init if vp had no offset, keep in mind this is from the - * right side of vp due to scan direction - */ - data->inits.v_c = dc_fixpt_add(data->inits.v_c, dc_fixpt_mul_int( - data->ratios.vert_c, data->recout.y - recout_full->y)); - /* - * this is the difference between first pixel of viewport available to read - * and init position, taking into account scan direction - */ - int_part = dc_fixpt_floor(data->inits.v_c) - end_offset; - if (int_part < data->taps.v_taps_c) { - int int_adj = end_offset >= (data->taps.v_taps_c - int_part) ? - (data->taps.v_taps_c - int_part) : end_offset; - data->viewport_c.height += int_adj; - int_part += int_adj; - } else if (int_part > data->taps.v_taps_c) { - data->viewport_c.height += int_part - data->taps.v_taps_c; - int_part = data->taps.v_taps_c; - } - data->inits.v_c.value &= 0xffffffff; - data->inits.v_c = dc_fixpt_add_int(data->inits.v_c, int_part); - } - } + /* + * Taps, inits and scaling ratios are in recout space need to rotate + * to viewport rotation before adjustment + */ + adjust_vp_and_init_for_seamless_clip( + flip_horz_scan_dir, + recout_skip_h, + surf_size_h, + orthogonal_rotation ? data->taps.v_taps : data->taps.h_taps, + orthogonal_rotation ? data->ratios.vert : data->ratios.horz, + orthogonal_rotation ? &data->inits.v : &data->inits.h, + &data->viewport.x, + &data->viewport.width); + adjust_vp_and_init_for_seamless_clip( + flip_horz_scan_dir, + recout_skip_h, + surf_size_h / vpc_div, + orthogonal_rotation ? data->taps.v_taps_c : data->taps.h_taps_c, + orthogonal_rotation ? data->ratios.vert_c : data->ratios.horz_c, + orthogonal_rotation ? &data->inits.v_c : &data->inits.h_c, + &data->viewport_c.x, + &data->viewport_c.width); + adjust_vp_and_init_for_seamless_clip( + flip_vert_scan_dir, + recout_skip_v, + surf_size_v, + orthogonal_rotation ? data->taps.h_taps : data->taps.v_taps, + orthogonal_rotation ? data->ratios.horz : data->ratios.vert, + orthogonal_rotation ? &data->inits.h : &data->inits.v, + &data->viewport.y, + &data->viewport.height); + adjust_vp_and_init_for_seamless_clip( + flip_vert_scan_dir, + recout_skip_v, + surf_size_v / vpc_div, + orthogonal_rotation ? data->taps.h_taps_c : data->taps.v_taps_c, + orthogonal_rotation ? data->ratios.horz_c : data->ratios.vert_c, + orthogonal_rotation ? &data->inits.h_c : &data->inits.v_c, + &data->viewport_c.y, + &data->viewport_c.height); /* Interlaced inits based on final vert inits */ data->inits.v_bot = dc_fixpt_add(data->inits.v, data->ratios.vert); data->inits.v_c_bot = dc_fixpt_add(data->inits.v_c, data->ratios.vert_c); - if (pipe_ctx->plane_state->rotation == ROTATION_ANGLE_90 || - pipe_ctx->plane_state->rotation == ROTATION_ANGLE_270) { - rect_swap_helper(&data->viewport_c); - rect_swap_helper(&data->viewport); - } } bool resource_build_scaling_params(struct pipe_ctx *pipe_ctx) { const struct dc_plane_state *plane_state = pipe_ctx->plane_state; struct dc_crtc_timing *timing = &pipe_ctx->stream->timing; - struct rect recout_full = { 0 }; bool res = false; DC_LOGGER_INIT(pipe_ctx->stream->ctx->logger); /* Important: scaling ratio calculation requires pixel format, @@ -1115,9 +928,6 @@ bool resource_build_scaling_params(struct pipe_ctx *pipe_ctx) pipe_ctx->plane_res.scl_data.format = convert_pixel_format_to_dalsurface( pipe_ctx->plane_state->format); - if (pipe_ctx->stream->timing.flags.INTERLACE) - pipe_ctx->stream->dst.height *= 2; - calculate_scaling_ratios(pipe_ctx); calculate_viewport(pipe_ctx); @@ -1125,7 +935,7 @@ bool resource_build_scaling_params(struct pipe_ctx *pipe_ctx) if (pipe_ctx->plane_res.scl_data.viewport.height < 16 || pipe_ctx->plane_res.scl_data.viewport.width < 16) return false; - calculate_recout(pipe_ctx, &recout_full); + calculate_recout(pipe_ctx); /** * Setting line buffer pixel depth to 24bpp yields banding @@ -1138,9 +948,6 @@ bool resource_build_scaling_params(struct pipe_ctx *pipe_ctx) pipe_ctx->plane_res.scl_data.h_active = timing->h_addressable + timing->h_border_left + timing->h_border_right; pipe_ctx->plane_res.scl_data.v_active = timing->v_addressable + timing->v_border_top + timing->v_border_bottom; - if (pipe_ctx->stream->timing.flags.INTERLACE) - pipe_ctx->plane_res.scl_data.v_active *= 2; - /* Taps calculations */ if (pipe_ctx->plane_res.xfm != NULL) @@ -1169,7 +976,7 @@ bool resource_build_scaling_params(struct pipe_ctx *pipe_ctx) if (res) /* May need to re-check lb size after this in some obscure scenario */ - calculate_inits_and_adj_vp(pipe_ctx, &recout_full); + calculate_inits_and_adj_vp(pipe_ctx); DC_LOG_SCALER( "%s: Viewport:\nheight:%d width:%d x:%d " @@ -1185,9 +992,6 @@ bool resource_build_scaling_params(struct pipe_ctx *pipe_ctx) plane_state->dst_rect.x, plane_state->dst_rect.y); - if (pipe_ctx->stream->timing.flags.INTERLACE) - pipe_ctx->stream->dst.height /= 2; - return res; } @@ -1382,6 +1186,9 @@ bool dc_add_plane_to_context( return false; } + tail_pipe = resource_get_tail_pipe_for_stream(&context->res_ctx, stream); + ASSERT(tail_pipe); + free_pipe = acquire_free_pipe_for_stream(context, pool, stream); #if defined(CONFIG_DRM_AMD_DC_DCN1_0) @@ -1399,10 +1206,6 @@ bool dc_add_plane_to_context( free_pipe->plane_state = plane_state; if (head_pipe != free_pipe) { - - tail_pipe = resource_get_tail_pipe_for_stream(&context->res_ctx, stream); - ASSERT(tail_pipe); - free_pipe->stream_res.tg = tail_pipe->stream_res.tg; free_pipe->stream_res.abm = tail_pipe->stream_res.abm; free_pipe->stream_res.opp = tail_pipe->stream_res.opp; @@ -1648,6 +1451,14 @@ static bool are_stream_backends_same( return true; } +/** + * dc_is_stream_unchanged() - Compare two stream states for equivalence. + * + * Checks if there a difference between the two states + * that would require a mode change. + * + * Does not compare cursor position or attributes. + */ bool dc_is_stream_unchanged( struct dc_stream_state *old_stream, struct dc_stream_state *stream) { @@ -1658,6 +1469,9 @@ bool dc_is_stream_unchanged( return true; } +/** + * dc_is_stream_scaling_unchanged() - Compare scaling rectangles of two streams. + */ bool dc_is_stream_scaling_unchanged( struct dc_stream_state *old_stream, struct dc_stream_state *stream) { @@ -1745,7 +1559,7 @@ static struct stream_encoder *find_first_free_match_stream_enc_for_link( { int i; int j = -1; - struct dc_link *link = stream->sink->link; + struct dc_link *link = stream->link; for (i = 0; i < pool->stream_enc_count; i++) { if (!res_ctx->is_stream_enc_acquired[i] && @@ -1817,16 +1631,19 @@ bool resource_is_stream_unchanged( return false; } +/** + * dc_add_stream_to_ctx() - Add a new dc_stream_state to a dc_state. + */ enum dc_status dc_add_stream_to_ctx( struct dc *dc, struct dc_state *new_ctx, struct dc_stream_state *stream) { - struct dc_context *dc_ctx = dc->ctx; enum dc_status res; + DC_LOGGER_INIT(dc->ctx->logger); if (new_ctx->stream_count >= dc->res_pool->timing_generator_count) { - DC_ERROR("Max streams reached, can't add stream %p !\n", stream); + DC_LOG_WARNING("Max streams reached, can't add stream %p !\n", stream); return DC_ERROR_UNEXPECTED; } @@ -1836,11 +1653,14 @@ enum dc_status dc_add_stream_to_ctx( res = dc->res_pool->funcs->add_stream_to_ctx(dc, new_ctx, stream); if (res != DC_OK) - DC_ERROR("Adding stream %p to context failed with err %d!\n", stream, res); + DC_LOG_WARNING("Adding stream %p to context failed with err %d!\n", stream, res); return res; } +/** + * dc_remove_stream_from_ctx() - Remove a stream from a dc_state. + */ enum dc_status dc_remove_stream_from_ctx( struct dc *dc, struct dc_state *new_ctx, @@ -1928,7 +1748,7 @@ static struct dc_stream_state *find_pll_sharable_stream( if (resource_are_streams_timing_synchronizable( stream_needs_pll, stream_has_pll) && !dc_is_dp_signal(stream_has_pll->signal) - && stream_has_pll->sink->link->connector_signal + && stream_has_pll->link->connector_signal != SIGNAL_TYPE_VIRTUAL) return stream_has_pll; @@ -1939,7 +1759,7 @@ static struct dc_stream_state *find_pll_sharable_stream( static int get_norm_pix_clk(const struct dc_crtc_timing *timing) { - uint32_t pix_clk = timing->pix_clk_khz; + uint32_t pix_clk = timing->pix_clk_100hz; uint32_t normalized_pix_clk = pix_clk; if (timing->pixel_encoding == PIXEL_ENCODING_YCBCR420) @@ -1971,15 +1791,60 @@ static void calculate_phy_pix_clks(struct dc_stream_state *stream) /* update actual pixel clock on all streams */ if (dc_is_hdmi_signal(stream->signal)) stream->phy_pix_clk = get_norm_pix_clk( - &stream->timing); + &stream->timing) / 10; else stream->phy_pix_clk = - stream->timing.pix_clk_khz; + stream->timing.pix_clk_100hz / 10; if (stream->timing.timing_3d_format == TIMING_3D_FORMAT_HW_FRAME_PACKING) stream->phy_pix_clk *= 2; } +static int acquire_resource_from_hw_enabled_state( + struct resource_context *res_ctx, + const struct resource_pool *pool, + struct dc_stream_state *stream) +{ + struct dc_link *link = stream->link; + unsigned int inst; + + /* Check for enabled DIG to identify enabled display */ + if (!link->link_enc->funcs->is_dig_enabled(link->link_enc)) + return -1; + + /* Check for which front end is used by this encoder. + * Note the inst is 1 indexed, where 0 is undefined. + * Note that DIG_FE can source from different OTG but our + * current implementation always map 1-to-1, so this code makes + * the same assumption and doesn't check OTG source. + */ + inst = link->link_enc->funcs->get_dig_frontend(link->link_enc) - 1; + + /* Instance should be within the range of the pool */ + if (inst >= pool->pipe_count) + return -1; + + if (!res_ctx->pipe_ctx[inst].stream) { + struct pipe_ctx *pipe_ctx = &res_ctx->pipe_ctx[inst]; + + pipe_ctx->stream_res.tg = pool->timing_generators[inst]; + pipe_ctx->plane_res.mi = pool->mis[inst]; + pipe_ctx->plane_res.hubp = pool->hubps[inst]; + pipe_ctx->plane_res.ipp = pool->ipps[inst]; + pipe_ctx->plane_res.xfm = pool->transforms[inst]; + pipe_ctx->plane_res.dpp = pool->dpps[inst]; + pipe_ctx->stream_res.opp = pool->opps[inst]; + if (pool->dpps[inst]) + pipe_ctx->plane_res.mpcc_inst = pool->dpps[inst]->inst; + pipe_ctx->pipe_idx = inst; + + pipe_ctx->stream = stream; + return inst; + } + + return -1; +} + enum dc_status resource_map_pool_resources( const struct dc *dc, struct dc_state *context, @@ -2002,8 +1867,17 @@ enum dc_status resource_map_pool_resources( } */ - /* acquire new resources */ - pipe_idx = acquire_first_free_pipe(&context->res_ctx, pool, stream); + calculate_phy_pix_clks(stream); + + if (stream->apply_seamless_boot_optimization) + pipe_idx = acquire_resource_from_hw_enabled_state( + &context->res_ctx, + pool, + stream); + + if (pipe_idx < 0) + /* acquire new resources */ + pipe_idx = acquire_first_free_pipe(&context->res_ctx, pool, stream); #ifdef CONFIG_DRM_AMD_DC_DCN1_0 if (pipe_idx < 0) @@ -2020,7 +1894,7 @@ enum dc_status resource_map_pool_resources( &context->res_ctx, pool, stream); if (!pipe_ctx->stream_res.stream_enc) - return DC_NO_STREAM_ENG_RESOURCE; + return DC_NO_STREAM_ENC_RESOURCE; update_stream_engine_usage( &context->res_ctx, pool, @@ -2028,7 +1902,7 @@ enum dc_status resource_map_pool_resources( true); /* TODO: Add check if ASIC support and EDID audio */ - if (!stream->sink->converter_disable_audio && + if (!stream->converter_disable_audio && dc_is_audio_capable_signal(pipe_ctx->stream->signal) && stream->audio_info.mode_count) { pipe_ctx->stream_res.audio = find_first_free_audio( @@ -2059,6 +1933,12 @@ enum dc_status resource_map_pool_resources( return DC_ERROR_UNEXPECTED; } +/** + * dc_resource_state_copy_construct_current() - Creates a new dc_state from existing state + * Is a shallow copy. Increments refcounts on existing streams and planes. + * @dc: copy out of dc->current_state + * @dst_ctx: copy into this + */ void dc_resource_state_copy_construct_current( const struct dc *dc, struct dc_state *dst_ctx) @@ -2071,9 +1951,17 @@ void dc_resource_state_construct( const struct dc *dc, struct dc_state *dst_ctx) { - dst_ctx->dis_clk = dc->res_pool->dccg; + dst_ctx->dccg = dc->res_pool->clk_mgr; } +/** + * dc_validate_global_state() - Determine if HW can support a given state + * Checks HW resource availability and bandwidth requirement. + * @dc: dc struct for this driver + * @new_ctx: state to be validated + * + * Return: DC_OK if the result can be programmed. Otherwise, an error code. + */ enum dc_status dc_validate_global_state( struct dc *dc, struct dc_state *new_ctx) @@ -2276,7 +2164,7 @@ static void set_avi_info_frame( itc = true; itc_value = 1; - support = stream->sink->edid_caps.content_support; + support = stream->content_support; if (itc) { if (!support.bits.valid_content_type) { @@ -2315,8 +2203,8 @@ static void set_avi_info_frame( /* TODO : We should handle YCC quantization */ /* but we do not have matrix calculation */ - if (stream->sink->edid_caps.qs_bit == 1 && - stream->sink->edid_caps.qy_bit == 1) { + if (stream->qs_bit == 1 && + stream->qy_bit == 1) { if (color_space == COLOR_SPACE_SRGB || color_space == COLOR_SPACE_2020_RGB_FULLRANGE) { hdmi_info.bits.Q0_Q1 = RGB_QUANTIZATION_FULL_RANGE; @@ -2401,113 +2289,15 @@ static void set_vendor_info_packet( struct dc_info_packet *info_packet, struct dc_stream_state *stream) { - uint32_t length = 0; - bool hdmi_vic_mode = false; - uint8_t checksum = 0; - uint32_t i = 0; - enum dc_timing_3d_format format; - // Can be different depending on packet content /*todo*/ - // unsigned int length = pPathMode->dolbyVision ? 24 : 5; - - info_packet->valid = false; - - format = stream->timing.timing_3d_format; - if (stream->view_format == VIEW_3D_FORMAT_NONE) - format = TIMING_3D_FORMAT_NONE; - - /* Can be different depending on packet content */ - length = 5; - - if (stream->timing.hdmi_vic != 0 - && stream->timing.h_total >= 3840 - && stream->timing.v_total >= 2160) - hdmi_vic_mode = true; - - /* According to HDMI 1.4a CTS, VSIF should be sent - * for both 3D stereo and HDMI VIC modes. - * For all other modes, there is no VSIF sent. */ + /* SPD info packet for FreeSync */ - if (format == TIMING_3D_FORMAT_NONE && !hdmi_vic_mode) + /* Check if Freesync is supported. Return if false. If true, + * set the corresponding bit in the info packet + */ + if (!stream->vsp_infopacket.valid) return; - /* 24bit IEEE Registration identifier (0x000c03). LSB first. */ - info_packet->sb[1] = 0x03; - info_packet->sb[2] = 0x0C; - info_packet->sb[3] = 0x00; - - /*PB4: 5 lower bytes = 0 (reserved). 3 higher bits = HDMI_Video_Format. - * The value for HDMI_Video_Format are: - * 0x0 (0b000) - No additional HDMI video format is presented in this - * packet - * 0x1 (0b001) - Extended resolution format present. 1 byte of HDMI_VIC - * parameter follows - * 0x2 (0b010) - 3D format indication present. 3D_Structure and - * potentially 3D_Ext_Data follows - * 0x3..0x7 (0b011..0b111) - reserved for future use */ - if (format != TIMING_3D_FORMAT_NONE) - info_packet->sb[4] = (2 << 5); - else if (hdmi_vic_mode) - info_packet->sb[4] = (1 << 5); - - /* PB5: If PB4 claims 3D timing (HDMI_Video_Format = 0x2): - * 4 lower bites = 0 (reserved). 4 higher bits = 3D_Structure. - * The value for 3D_Structure are: - * 0x0 - Frame Packing - * 0x1 - Field Alternative - * 0x2 - Line Alternative - * 0x3 - Side-by-Side (full) - * 0x4 - L + depth - * 0x5 - L + depth + graphics + graphics-depth - * 0x6 - Top-and-Bottom - * 0x7 - Reserved for future use - * 0x8 - Side-by-Side (Half) - * 0x9..0xE - Reserved for future use - * 0xF - Not used */ - switch (format) { - case TIMING_3D_FORMAT_HW_FRAME_PACKING: - case TIMING_3D_FORMAT_SW_FRAME_PACKING: - info_packet->sb[5] = (0x0 << 4); - break; - - case TIMING_3D_FORMAT_SIDE_BY_SIDE: - case TIMING_3D_FORMAT_SBS_SW_PACKED: - info_packet->sb[5] = (0x8 << 4); - length = 6; - break; - - case TIMING_3D_FORMAT_TOP_AND_BOTTOM: - case TIMING_3D_FORMAT_TB_SW_PACKED: - info_packet->sb[5] = (0x6 << 4); - break; - - default: - break; - } - - /*PB5: If PB4 is set to 0x1 (extended resolution format) - * fill PB5 with the correct HDMI VIC code */ - if (hdmi_vic_mode) - info_packet->sb[5] = stream->timing.hdmi_vic; - - /* Header */ - info_packet->hb0 = HDMI_INFOFRAME_TYPE_VENDOR; /* VSIF packet type. */ - info_packet->hb1 = 0x01; /* Version */ - - /* 4 lower bits = Length, 4 higher bits = 0 (reserved) */ - info_packet->hb2 = (uint8_t) (length); - - /* Calculate checksum */ - checksum = 0; - checksum += info_packet->hb0; - checksum += info_packet->hb1; - checksum += info_packet->hb2; - - for (i = 1; i <= length; i++) - checksum += info_packet->sb[i]; - - info_packet->sb[0] = (uint8_t) (0x100 - checksum); - - info_packet->valid = true; + *info_packet = stream->vsp_infopacket; } static void set_spd_info_packet( @@ -2563,10 +2353,6 @@ void dc_resource_state_destruct(struct dc_state *context) } } -/* - * Copy src_ctx into dst_ctx and retain all surfaces and streams referenced - * by the src_ctx - */ void dc_resource_state_copy_construct( const struct dc_state *src_ctx, struct dc_state *dst_ctx) @@ -2862,7 +2648,7 @@ void resource_build_bit_depth_reduction_params(struct dc_stream_state *stream, enum dc_status dc_validate_stream(struct dc *dc, struct dc_stream_state *stream) { struct dc *core_dc = dc; - struct dc_link *link = stream->sink->link; + struct dc_link *link = stream->link; struct timing_generator *tg = core_dc->res_pool->timing_generators[0]; enum dc_status res = DC_OK; diff --git a/drivers/gpu/drm/amd/display/dc/core/dc_stream.c b/drivers/gpu/drm/amd/display/dc/core/dc_stream.c index 2ac848a106ba..996298c35f42 100644 --- a/drivers/gpu/drm/amd/display/dc/core/dc_stream.c +++ b/drivers/gpu/drm/amd/display/dc/core/dc_stream.c @@ -35,20 +35,17 @@ /******************************************************************************* * Private functions ******************************************************************************/ -void update_stream_signal(struct dc_stream_state *stream) +void update_stream_signal(struct dc_stream_state *stream, struct dc_sink *sink) { - - struct dc_sink *dc_sink = stream->sink; - - if (dc_sink->sink_signal == SIGNAL_TYPE_NONE) - stream->signal = stream->sink->link->connector_signal; + if (sink->sink_signal == SIGNAL_TYPE_NONE) + stream->signal = stream->link->connector_signal; else - stream->signal = dc_sink->sink_signal; + stream->signal = sink->sink_signal; if (dc_is_dvi_signal(stream->signal)) { if (stream->ctx->dc->caps.dual_link_dvi && - stream->timing.pix_clk_khz > TMDS_MAX_PIXEL_CLOCK && - stream->sink->sink_signal != SIGNAL_TYPE_DVI_SINGLE_LINK) + (stream->timing.pix_clk_100hz / 10) > TMDS_MAX_PIXEL_CLOCK && + sink->sink_signal != SIGNAL_TYPE_DVI_SINGLE_LINK) stream->signal = SIGNAL_TYPE_DVI_DUAL_LINK; else stream->signal = SIGNAL_TYPE_DVI_SINGLE_LINK; @@ -61,10 +58,15 @@ static void construct(struct dc_stream_state *stream, uint32_t i = 0; stream->sink = dc_sink_data; - stream->ctx = stream->sink->ctx; - dc_sink_retain(dc_sink_data); + stream->ctx = dc_sink_data->ctx; + stream->link = dc_sink_data->link; + stream->sink_patches = dc_sink_data->edid_caps.panel_patch; + stream->converter_disable_audio = dc_sink_data->converter_disable_audio; + stream->qs_bit = dc_sink_data->edid_caps.qs_bit; + stream->qy_bit = dc_sink_data->edid_caps.qy_bit; + /* Copy audio modes */ /* TODO - Remove this translation */ for (i = 0; i < (dc_sink_data->edid_caps.audio_mode_count); i++) @@ -100,12 +102,14 @@ static void construct(struct dc_stream_state *stream, /* EDID CAP translation for HDMI 2.0 */ stream->timing.flags.LTE_340MCSC_SCRAMBLE = dc_sink_data->edid_caps.lte_340mcsc_scramble; - stream->status.link = stream->sink->link; - - update_stream_signal(stream); + update_stream_signal(stream, dc_sink_data); stream->out_transfer_func = dc_create_transfer_func(); stream->out_transfer_func->type = TF_TYPE_BYPASS; + stream->out_transfer_func->ctx = stream->ctx; + + stream->stream_id = stream->ctx->dc_stream_id_count; + stream->ctx->dc_stream_id_count++; } static void destruct(struct dc_stream_state *stream) @@ -156,22 +160,44 @@ struct dc_stream_state *dc_create_stream_for_sink( return stream; } -struct dc_stream_status *dc_stream_get_status( +/** + * dc_stream_get_status_from_state - Get stream status from given dc state + * @state: DC state to find the stream status in + * @stream: The stream to get the stream status for + * + * The given stream is expected to exist in the given dc state. Otherwise, NULL + * will be returned. + */ +struct dc_stream_status *dc_stream_get_status_from_state( + struct dc_state *state, struct dc_stream_state *stream) { uint8_t i; - struct dc *dc = stream->ctx->dc; - for (i = 0; i < dc->current_state->stream_count; i++) { - if (stream == dc->current_state->streams[i]) - return &dc->current_state->stream_status[i]; + for (i = 0; i < state->stream_count; i++) { + if (stream == state->streams[i]) + return &state->stream_status[i]; } return NULL; } /** - * Update the cursor attributes and set cursor surface address + * dc_stream_get_status() - Get current stream status of the given stream state + * @stream: The stream to get the stream status for. + * + * The given stream is expected to exist in dc->current_state. Otherwise, NULL + * will be returned. + */ +struct dc_stream_status *dc_stream_get_status( + struct dc_stream_state *stream) +{ + struct dc *dc = stream->ctx->dc; + return dc_stream_get_status_from_state(dc->current_state, stream); +} + +/** + * dc_stream_set_cursor_attributes() - Update cursor attributes and set cursor surface address */ bool dc_stream_set_cursor_attributes( struct dc_stream_state *stream, @@ -335,16 +361,12 @@ void dc_stream_log(const struct dc *dc, const struct dc_stream_state *stream) stream->output_color_space); DC_LOG_DC( "\tpix_clk_khz: %d, h_total: %d, v_total: %d, pixelencoder:%d, displaycolorDepth:%d\n", - stream->timing.pix_clk_khz, + stream->timing.pix_clk_100hz / 10, stream->timing.h_total, stream->timing.v_total, stream->timing.pixel_encoding, stream->timing.display_color_depth); DC_LOG_DC( - "\tsink name: %s, serial: %d\n", - stream->sink->edid_caps.display_name, - stream->sink->edid_caps.serial_number); - DC_LOG_DC( "\tlink: %d\n", - stream->sink->link->link_index); + stream->link->link_index); } diff --git a/drivers/gpu/drm/amd/display/dc/core/dc_surface.c b/drivers/gpu/drm/amd/display/dc/core/dc_surface.c index 8fb3aefd195c..ee6bd50f60b8 100644 --- a/drivers/gpu/drm/amd/display/dc/core/dc_surface.c +++ b/drivers/gpu/drm/amd/display/dc/core/dc_surface.c @@ -40,10 +40,14 @@ static void construct(struct dc_context *ctx, struct dc_plane_state *plane_state plane_state->ctx = ctx; plane_state->gamma_correction = dc_create_gamma(); - plane_state->gamma_correction->is_identity = true; + if (plane_state->gamma_correction != NULL) + plane_state->gamma_correction->is_identity = true; plane_state->in_transfer_func = dc_create_transfer_func(); - plane_state->in_transfer_func->type = TF_TYPE_BYPASS; + if (plane_state->in_transfer_func != NULL) { + plane_state->in_transfer_func->type = TF_TYPE_BYPASS; + plane_state->in_transfer_func->ctx = ctx; + } } static void destruct(struct dc_plane_state *plane_state) diff --git a/drivers/gpu/drm/amd/display/dc/core/dc_vm_helper.c b/drivers/gpu/drm/amd/display/dc/core/dc_vm_helper.c new file mode 100644 index 000000000000..6ce87b682a32 --- /dev/null +++ b/drivers/gpu/drm/amd/display/dc/core/dc_vm_helper.c @@ -0,0 +1,123 @@ +/* + * Copyright 2018 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. + * + * Authors: AMD + * + */ + +#include "vm_helper.h" + +static void mark_vmid_used(struct vm_helper *vm_helper, unsigned int pos, uint8_t hubp_idx) +{ + struct vmid_usage vmids = vm_helper->hubp_vmid_usage[hubp_idx]; + + vmids.vmid_usage[0] = vmids.vmid_usage[1]; + vmids.vmid_usage[1] = 1 << pos; +} + +static void add_ptb_to_table(struct vm_helper *vm_helper, unsigned int vmid, uint64_t ptb) +{ + vm_helper->ptb_assigned_to_vmid[vmid] = ptb; + vm_helper->num_vmids_available--; +} + +static void clear_entry_from_vmid_table(struct vm_helper *vm_helper, unsigned int vmid) +{ + vm_helper->ptb_assigned_to_vmid[vmid] = 0; + vm_helper->num_vmids_available++; +} + +static void evict_vmids(struct vm_helper *vm_helper) +{ + int i; + uint16_t ord = 0; + + for (i = 0; i < vm_helper->num_vmid; i++) + ord |= vm_helper->hubp_vmid_usage[i].vmid_usage[0] | vm_helper->hubp_vmid_usage[i].vmid_usage[1]; + + // At this point any positions with value 0 are unused vmids, evict them + for (i = 1; i < vm_helper->num_vmid; i++) { + if (ord & (1u << i)) + clear_entry_from_vmid_table(vm_helper, i); + } +} + +// Return value of -1 indicates vmid table unitialized or ptb dne in the table +static int get_existing_vmid_for_ptb(struct vm_helper *vm_helper, uint64_t ptb) +{ + int i; + + for (i = 0; i < vm_helper->num_vmid; i++) { + if (vm_helper->ptb_assigned_to_vmid[i] == ptb) + return i; + } + + return -1; +} + +// Expected to be called only when there's an available vmid +static int get_next_available_vmid(struct vm_helper *vm_helper) +{ + int i; + + for (i = 1; i < vm_helper->num_vmid; i++) { + if (vm_helper->ptb_assigned_to_vmid[i] == 0) + return i; + } + + return -1; +} + +uint8_t get_vmid_for_ptb(struct vm_helper *vm_helper, int64_t ptb, uint8_t hubp_idx) +{ + unsigned int vmid = 0; + int vmid_exists = -1; + + // Physical address gets vmid 0 + if (ptb == 0) + return 0; + + vmid_exists = get_existing_vmid_for_ptb(vm_helper, ptb); + + if (vmid_exists != -1) { + mark_vmid_used(vm_helper, vmid_exists, hubp_idx); + vmid = vmid_exists; + } else { + if (vm_helper->num_vmids_available == 0) + evict_vmids(vm_helper); + + vmid = get_next_available_vmid(vm_helper); + mark_vmid_used(vm_helper, vmid, hubp_idx); + add_ptb_to_table(vm_helper, vmid, ptb); + } + + return vmid; +} + +void init_vm_helper(struct vm_helper *vm_helper, unsigned int num_vmid, unsigned int num_hubp) +{ + vm_helper->num_vmid = num_vmid; + vm_helper->num_hubp = num_hubp; + vm_helper->num_vmids_available = num_vmid - 1; + + memset(vm_helper->hubp_vmid_usage, 0, sizeof(vm_helper->hubp_vmid_usage[0]) * MAX_HUBP); + memset(vm_helper->ptb_assigned_to_vmid, 0, sizeof(vm_helper->ptb_assigned_to_vmid[0]) * MAX_VMID); +} diff --git a/drivers/gpu/drm/amd/display/dc/dc.h b/drivers/gpu/drm/amd/display/dc/dc.h index b57fa61b3034..0515095574e7 100644 --- a/drivers/gpu/drm/amd/display/dc/dc.h +++ b/drivers/gpu/drm/amd/display/dc/dc.h @@ -36,9 +36,10 @@ #include "inc/hw_sequencer.h" #include "inc/compressor.h" +#include "inc/hw/dmcu.h" #include "dml/display_mode_lib.h" -#define DC_VER "3.1.68" +#define DC_VER "3.2.17" #define MAX_SURFACES 3 #define MAX_STREAMS 6 @@ -47,13 +48,6 @@ /******************************************************************************* * Display Core Interfaces ******************************************************************************/ -struct dmcu_version { - unsigned int date; - unsigned int month; - unsigned int year; - unsigned int interface_version; -}; - struct dc_versions { const char *dc_ver; struct dmcu_version dmcu_version; @@ -250,8 +244,6 @@ struct dc_debug_options { bool disable_dmcu; bool disable_psr; bool force_abm_enable; - bool disable_hbup_pg; - bool disable_dpp_pg; bool disable_stereo_support; bool vsr_support; bool performance_trace; @@ -263,6 +255,8 @@ struct dc_debug_options { bool scl_reset_length10; bool hdmi20_disable; bool skip_detection_link_training; + unsigned int force_odm_combine; //bit vector based on otg inst + unsigned int force_fclk_khz; }; struct dc_debug_data { @@ -271,7 +265,6 @@ struct dc_debug_data { uint32_t auxErrorCount; }; - struct dc_state; struct resource_pool; struct dce_hwseq; @@ -305,11 +298,6 @@ struct dc { struct hw_sequencer_funcs hwss; struct dce_hwseq *hwseq; - /* temp store of dm_pp_display_configuration - * to compare to see if display config changed - */ - struct dm_pp_display_configuration prev_display_config; - bool optimized_required; /* FBC compressor */ @@ -352,8 +340,13 @@ struct dc_init_data { uint32_t log_mask; }; -struct dc *dc_create(const struct dc_init_data *init_params); +struct dc_callback_init { + uint8_t reserved; +}; +struct dc *dc_create(const struct dc_init_data *init_params); +void dc_init_callbacks(struct dc *dc, + const struct dc_callback_init *init_params); void dc_destroy(struct dc **dc); /******************************************************************************* @@ -453,6 +446,7 @@ union surface_update_flags { uint32_t coeff_reduction_change:1; uint32_t output_tf_change:1; uint32_t pixel_format_change:1; + uint32_t plane_size_change:1; /* Full updates */ uint32_t new_plane:1; @@ -509,6 +503,9 @@ struct dc_plane_state { struct dc_plane_status status; struct dc_context *ctx; + /* HACK: Workaround for forcing full reprogramming under some conditions */ + bool force_full_update; + /* private to dc_surface.c */ enum dc_irq_source irq_source; struct kref refcount; @@ -600,6 +597,10 @@ struct dc_validation_set { uint8_t plane_count; }; +bool dc_validate_seamless_boot_timing(struct dc *dc, + const struct dc_sink *sink, + struct dc_crtc_timing *crtc_timing); + enum dc_status dc_validate_plane(struct dc *dc, const struct dc_plane_state *plane_state); void get_clock_requirements_for_state(struct dc_state *state, struct AsicStateEx *info); @@ -665,6 +666,7 @@ struct dpcd_caps { int8_t branch_dev_name[6]; int8_t branch_hw_revision; int8_t branch_fw_revision[2]; + uint8_t link_rate_set; bool allow_invalid_MSA_timing_param; bool panel_mode_edp; @@ -755,5 +757,9 @@ void dc_set_power_state( struct dc *dc, enum dc_acpi_cm_power_state power_state); void dc_resume(struct dc *dc); +unsigned int dc_get_current_backlight_pwm(struct dc *dc); +unsigned int dc_get_target_backlight_pwm(struct dc *dc); + +bool dc_is_dmcu_initialized(struct dc *dc); #endif /* DC_INTERFACE_H_ */ diff --git a/drivers/gpu/drm/amd/display/dc/dc_bios_types.h b/drivers/gpu/drm/amd/display/dc/dc_bios_types.h index 8130b95ccc53..78c3b300ec45 100644 --- a/drivers/gpu/drm/amd/display/dc/dc_bios_types.h +++ b/drivers/gpu/drm/amd/display/dc/dc_bios_types.h @@ -121,10 +121,6 @@ struct dc_vbios_funcs { enum bp_result (*program_crtc_timing)( struct dc_bios *bios, struct bp_hw_crtc_timing_parameters *bp_params); - - enum bp_result (*crtc_source_select)( - struct dc_bios *bios, - struct bp_crtc_source_select *bp_params); enum bp_result (*program_display_engine_pll)( struct dc_bios *bios, struct bp_pixel_clock_parameters *bp_params); diff --git a/drivers/gpu/drm/amd/display/dc/dc_dp_types.h b/drivers/gpu/drm/amd/display/dc/dc_dp_types.h index da93ab43f2d8..d4eab33c453b 100644 --- a/drivers/gpu/drm/amd/display/dc/dc_dp_types.h +++ b/drivers/gpu/drm/amd/display/dc/dc_dp_types.h @@ -46,11 +46,14 @@ enum dc_lane_count { */ enum dc_link_rate { LINK_RATE_UNKNOWN = 0, - LINK_RATE_LOW = 0x06, - LINK_RATE_HIGH = 0x0A, - LINK_RATE_RBR2 = 0x0C, - LINK_RATE_HIGH2 = 0x14, - LINK_RATE_HIGH3 = 0x1E + LINK_RATE_LOW = 0x06, // Rate_1 (RBR) - 1.62 Gbps/Lane + LINK_RATE_RATE_2 = 0x08, // Rate_2 - 2.16 Gbps/Lane + LINK_RATE_RATE_3 = 0x09, // Rate_3 - 2.43 Gbps/Lane + LINK_RATE_HIGH = 0x0A, // Rate_4 (HBR) - 2.70 Gbps/Lane + LINK_RATE_RBR2 = 0x0C, // Rate_5 (RBR2)- 3.24 Gbps/Lane + LINK_RATE_RATE_6 = 0x10, // Rate_6 - 4.32 Gbps/Lane + LINK_RATE_HIGH2 = 0x14, // Rate_7 (HBR2)- 5.40 Gbps/Lane + LINK_RATE_HIGH3 = 0x1E // Rate_8 (HBR3)- 8.10 Gbps/Lane }; enum dc_link_spread { diff --git a/drivers/gpu/drm/amd/display/dc/dc_helper.c b/drivers/gpu/drm/amd/display/dc/dc_helper.c index fcfd50b5dba0..597d38393379 100644 --- a/drivers/gpu/drm/amd/display/dc/dc_helper.c +++ b/drivers/gpu/drm/amd/display/dc/dc_helper.c @@ -29,31 +29,59 @@ #include "dm_services.h" #include <stdarg.h> +struct dc_reg_value_masks { + uint32_t value; + uint32_t mask; +}; + +struct dc_reg_sequence { + uint32_t addr; + struct dc_reg_value_masks value_masks; +}; + +static inline void set_reg_field_value_masks( + struct dc_reg_value_masks *field_value_mask, + uint32_t value, + uint32_t mask, + uint8_t shift) +{ + ASSERT(mask != 0); + + field_value_mask->value = (field_value_mask->value & ~mask) | (mask & (value << shift)); + field_value_mask->mask = field_value_mask->mask | mask; +} + uint32_t generic_reg_update_ex(const struct dc_context *ctx, uint32_t addr, uint32_t reg_val, int n, uint8_t shift1, uint32_t mask1, uint32_t field_value1, ...) { + struct dc_reg_value_masks field_value_mask = {0}; uint32_t shift, mask, field_value; int i = 1; va_list ap; va_start(ap, field_value1); - reg_val = set_reg_field_value_ex(reg_val, field_value1, mask1, shift1); + /* gather all bits value/mask getting updated in this register */ + set_reg_field_value_masks(&field_value_mask, + field_value1, mask1, shift1); while (i < n) { shift = va_arg(ap, uint32_t); mask = va_arg(ap, uint32_t); field_value = va_arg(ap, uint32_t); - reg_val = set_reg_field_value_ex(reg_val, field_value, mask, shift); + set_reg_field_value_masks(&field_value_mask, + field_value, mask, shift); i++; } - - dm_write_reg(ctx, addr, reg_val); va_end(ap); + + /* mmio write directly */ + reg_val = (reg_val & ~field_value_mask.mask) | field_value_mask.value; + dm_write_reg(ctx, addr, reg_val); return reg_val; } @@ -234,14 +262,14 @@ uint32_t generic_reg_wait(const struct dc_context *ctx, if (field_value == condition_value) { if (i * delay_between_poll_us > 1000 && !IS_FPGA_MAXIMUS_DC(ctx->dce_environment)) - dm_output_to_console("REG_WAIT taking a while: %dms in %s line:%d\n", + DC_LOG_DC("REG_WAIT taking a while: %dms in %s line:%d\n", delay_between_poll_us * i / 1000, func_name, line); return reg_val; } } - dm_error("REG_WAIT timeout %dus * %d tries - %s line:%d\n", + DC_LOG_WARNING("REG_WAIT timeout %dus * %d tries - %s line:%d\n", delay_between_poll_us, time_out_num_tries, func_name, line); diff --git a/drivers/gpu/drm/amd/display/dc/dc_hw_types.h b/drivers/gpu/drm/amd/display/dc/dc_hw_types.h index 7825e4b5e97c..da55d623647a 100644 --- a/drivers/gpu/drm/amd/display/dc/dc_hw_types.h +++ b/drivers/gpu/drm/amd/display/dc/dc_hw_types.h @@ -97,6 +97,8 @@ struct dc_plane_address { union large_integer chroma_dcc_const_color; } video_progressive; }; + + union large_integer page_table_base; }; struct dc_size { @@ -192,7 +194,6 @@ enum surface_pixel_format { /*swaped & float*/ SURFACE_PIXEL_FORMAT_GRPH_ABGR16161616F, /*grow graphics here if necessary */ - SURFACE_PIXEL_FORMAT_VIDEO_AYCrCb8888, SURFACE_PIXEL_FORMAT_VIDEO_BEGIN, SURFACE_PIXEL_FORMAT_VIDEO_420_YCbCr = SURFACE_PIXEL_FORMAT_VIDEO_BEGIN, @@ -200,6 +201,7 @@ enum surface_pixel_format { SURFACE_PIXEL_FORMAT_VIDEO_420_10bpc_YCbCr, SURFACE_PIXEL_FORMAT_VIDEO_420_10bpc_YCrCb, SURFACE_PIXEL_FORMAT_SUBSAMPLE_END, + SURFACE_PIXEL_FORMAT_VIDEO_AYCrCb8888, SURFACE_PIXEL_FORMAT_INVALID /*grow 444 video here if necessary */ @@ -358,15 +360,16 @@ union dc_tiling_info { } gfx8; struct { + enum swizzle_mode_values swizzle; unsigned int num_pipes; - unsigned int num_banks; + unsigned int max_compressed_frags; unsigned int pipe_interleave; + + unsigned int num_banks; unsigned int num_shader_engines; unsigned int num_rb_per_se; - unsigned int max_compressed_frags; bool shaderEnable; - enum swizzle_mode_values swizzle; bool meta_linear; bool rb_aligned; bool pipe_aligned; @@ -729,7 +732,7 @@ struct dc_crtc_timing { uint32_t v_front_porch; uint32_t v_sync_width; - uint32_t pix_clk_khz; + uint32_t pix_clk_100hz; uint32_t vic; uint32_t hdmi_vic; diff --git a/drivers/gpu/drm/amd/display/dc/dc_link.h b/drivers/gpu/drm/amd/display/dc/dc_link.h index 3bfdccceb524..8fc223defed4 100644 --- a/drivers/gpu/drm/amd/display/dc/dc_link.h +++ b/drivers/gpu/drm/amd/display/dc/dc_link.h @@ -30,6 +30,7 @@ #include "grph_object_defs.h" struct dc_link_status { + bool link_active; struct dpcd_caps *dpcd_caps; }; @@ -110,6 +111,7 @@ struct dc_link { union ddi_channel_mapping ddi_channel_mapping; struct connector_device_tag_info device_tag; struct dpcd_caps dpcd_caps; + uint32_t dongle_max_pix_clk; unsigned short chip_caps; unsigned int dpcd_sink_count; enum edp_revision edp_revision; @@ -124,12 +126,15 @@ struct dc_link { struct dc_link_status link_status; struct link_trace link_trace; + struct gpio *hpd_gpio; }; const struct dc_link_status *dc_link_get_status(const struct dc_link *dc_link); -/* - * Return an enumerated dc_link. dc_link order is constant and determined at +/** + * dc_get_link_at_index() - Return an enumerated dc_link. + * + * dc_link order is constant and determined at * boot time. They cannot be created or destroyed. * Use dc_get_caps() to get number of links. */ @@ -138,9 +143,13 @@ static inline struct dc_link *dc_get_link_at_index(struct dc *dc, uint32_t link_ return dc->links[link_index]; } -/* Set backlight level of an embedded panel (eDP, LVDS). */ -bool dc_link_set_backlight_level(const struct dc_link *dc_link, uint32_t level, - uint32_t frame_ramp, const struct dc_stream_state *stream); +/* Set backlight level of an embedded panel (eDP, LVDS). + * backlight_pwm_u16_16 is unsigned 32 bit with 16 bit integer + * and 16 bit fractional, where 1.0 is max backlight value. + */ +bool dc_link_set_backlight_level(const struct dc_link *dc_link, + uint32_t backlight_pwm_u16_16, + uint32_t frame_ramp); int dc_link_get_backlight_level(const struct dc_link *dc_link); diff --git a/drivers/gpu/drm/amd/display/dc/dc_stream.h b/drivers/gpu/drm/amd/display/dc/dc_stream.h index c5bd1fbb6982..5657cb3a2ad3 100644 --- a/drivers/gpu/drm/amd/display/dc/dc_stream.h +++ b/drivers/gpu/drm/amd/display/dc/dc_stream.h @@ -32,17 +32,18 @@ /******************************************************************************* * Stream Interfaces ******************************************************************************/ +struct timing_sync_info { + int group_id; + int group_size; + bool master; +}; struct dc_stream_status { int primary_otg_inst; int stream_enc_inst; int plane_count; + struct timing_sync_info timing_sync_info; struct dc_plane_state *plane_states[MAX_SURFACE_NUM]; - - /* - * link this stream passes through - */ - struct dc_link *link; }; // TODO: References to this needs to be removed.. @@ -50,12 +51,35 @@ struct freesync_context { bool dummy; }; +enum vertical_interrupt_ref_point { + START_V_UPDATE = 0, + START_V_SYNC, + INVALID_POINT + + //For now, only v_update interrupt is used. + //START_V_BLANK, + //START_V_ACTIVE +}; + +struct periodic_interrupt_config { + enum vertical_interrupt_ref_point ref_point; + int lines_offset; +}; + + struct dc_stream_state { + // sink is deprecated, new code should not reference + // this pointer struct dc_sink *sink; + + struct dc_link *link; + struct dc_panel_patch sink_patches; + union display_content_support content_support; struct dc_crtc_timing timing; struct dc_crtc_timing_adjust adjust; struct dc_info_packet vrr_infopacket; struct dc_info_packet vsc_infopacket; + struct dc_info_packet vsp_infopacket; struct rect src; /* composition area */ struct rect dst; /* stream addressable area */ @@ -79,8 +103,9 @@ struct dc_stream_state { enum view_3d_format view_format; bool ignore_msa_timing_param; - - unsigned long long periodic_fn_vsync_delta; + bool converter_disable_audio; + uint8_t qs_bit; + uint8_t qy_bit; /* TODO: custom INFO packets */ /* TODO: ABM info (DMCU) */ @@ -90,7 +115,9 @@ struct dc_stream_state { /* DMCU info */ unsigned int abm_level; - unsigned int bl_pwm_level; + + struct periodic_interrupt_config periodic_interrupt0; + struct periodic_interrupt_config periodic_interrupt1; /* from core_stream struct */ struct dc_context *ctx; @@ -102,9 +129,8 @@ struct dc_stream_state { int phy_pix_clk; enum signal_type signal; bool dpms_off; - bool apply_edp_fast_boot_optimization; - struct dc_stream_status status; + void *dm_stream_context; struct dc_cursor_attributes cursor_attributes; struct dc_cursor_position cursor_position; @@ -118,6 +144,21 @@ struct dc_stream_state { /* Computed state bits */ bool mode_changed : 1; + /* Output from DC when stream state is committed or altered + * DC may only access these values during: + * dc_commit_state, dc_commit_state_no_check, dc_commit_streams + * values may not change outside of those calls + */ + struct { + // For interrupt management, some hardware instance + // offsets need to be exposed to DM + uint8_t otg_offset; + } out; + + bool apply_edp_fast_boot_optimization; + bool apply_seamless_boot_optimization; + + uint32_t stream_id; }; struct dc_stream_update { @@ -127,15 +168,19 @@ struct dc_stream_update { struct dc_info_packet *hdr_static_metadata; unsigned int *abm_level; - unsigned long long *periodic_fn_vsync_delta; + struct periodic_interrupt_config *periodic_interrupt0; + struct periodic_interrupt_config *periodic_interrupt1; + struct dc_crtc_timing_adjust *adjust; struct dc_info_packet *vrr_infopacket; struct dc_info_packet *vsc_infopacket; + struct dc_info_packet *vsp_infopacket; bool *dpms_off; struct colorspace_transform *gamut_remap; enum dc_color_space *output_color_space; + enum dc_dither_option *dither_option; struct dc_csc_transform *output_csc_transform; @@ -162,7 +207,6 @@ void dc_commit_updates_for_stream(struct dc *dc, int surface_count, struct dc_stream_state *stream, struct dc_stream_update *stream_update, - struct dc_plane_state **plane_states, struct dc_state *state); /* * Log the current stream state. @@ -255,11 +299,14 @@ enum surface_update_type dc_check_update_surfaces_for_stream( */ struct dc_stream_state *dc_create_stream_for_sink(struct dc_sink *dc_sink); -void update_stream_signal(struct dc_stream_state *stream); +void update_stream_signal(struct dc_stream_state *stream, struct dc_sink *sink); void dc_stream_retain(struct dc_stream_state *dc_stream); void dc_stream_release(struct dc_stream_state *dc_stream); +struct dc_stream_status *dc_stream_get_status_from_state( + struct dc_state *state, + struct dc_stream_state *stream); struct dc_stream_status *dc_stream_get_status( struct dc_stream_state *dc_stream); diff --git a/drivers/gpu/drm/amd/display/dc/dc_types.h b/drivers/gpu/drm/amd/display/dc/dc_types.h index 6e12d640d020..da2009a108cf 100644 --- a/drivers/gpu/drm/amd/display/dc/dc_types.h +++ b/drivers/gpu/drm/amd/display/dc/dc_types.h @@ -73,10 +73,18 @@ struct hw_asic_id { void *atombios_base_address; }; +struct dc_perf_trace { + unsigned long read_count; + unsigned long write_count; + unsigned long last_entry_read; + unsigned long last_entry_write; +}; + struct dc_context { struct dc *dc; void *driver_context; /* e.g. amdgpu_device */ + struct dc_perf_trace *perf_trace; void *cgs_device; enum dce_environment dce_environment; @@ -89,8 +97,8 @@ struct dc_context { struct dc_bios *dc_bios; bool created_bios; struct gpio_service *gpio_service; - struct i2caux *i2caux; uint32_t dc_sink_id_count; + uint32_t dc_stream_id_count; uint64_t fbc_gpu_addr; }; @@ -191,9 +199,9 @@ union display_content_support { }; struct dc_panel_patch { - unsigned int disconnect_delay; unsigned int dppowerup_delay; unsigned int extra_t12_ms; + unsigned int extra_delay_backlight_off; }; struct dc_edid_caps { diff --git a/drivers/gpu/drm/amd/display/dc/dce/Makefile b/drivers/gpu/drm/amd/display/dc/dce/Makefile index 8f7f0e8b341f..6d7b64a743ca 100644 --- a/drivers/gpu/drm/amd/display/dc/dce/Makefile +++ b/drivers/gpu/drm/amd/display/dc/dce/Makefile @@ -28,7 +28,7 @@ DCE = dce_audio.o dce_stream_encoder.o dce_link_encoder.o dce_hwseq.o \ dce_mem_input.o dce_clock_source.o dce_scl_filters.o dce_transform.o \ -dce_clocks.o dce_opp.o dce_dmcu.o dce_abm.o dce_ipp.o dce_aux.o \ +dce_clk_mgr.o dce_opp.o dce_dmcu.o dce_abm.o dce_ipp.o dce_aux.o \ dce_i2c.o dce_i2c_hw.o dce_i2c_sw.o AMD_DAL_DCE = $(addprefix $(AMDDALPATH)/dc/dce/,$(DCE)) diff --git a/drivers/gpu/drm/amd/display/dc/dce/dce_abm.c b/drivers/gpu/drm/amd/display/dc/dce/dce_abm.c index 29294db1a96b..da96229db53a 100644 --- a/drivers/gpu/drm/amd/display/dc/dce/dce_abm.c +++ b/drivers/gpu/drm/amd/display/dc/dce/dce_abm.c @@ -53,8 +53,29 @@ #define MCP_DISABLE_ABM_IMMEDIATELY 255 +static bool dce_abm_set_pipe(struct abm *abm, uint32_t controller_id) +{ + struct dce_abm *abm_dce = TO_DCE_ABM(abm); + uint32_t rampingBoundary = 0xFFFF; + + REG_WAIT(MASTER_COMM_CNTL_REG, MASTER_COMM_INTERRUPT, 0, + 1, 80000); + + /* set ramping boundary */ + REG_WRITE(MASTER_COMM_DATA_REG1, rampingBoundary); + + /* setDMCUParam_Pipe */ + REG_UPDATE_2(MASTER_COMM_CMD_REG, + MASTER_COMM_CMD_REG_BYTE0, MCP_ABM_PIPE_SET, + MASTER_COMM_CMD_REG_BYTE1, controller_id); -static unsigned int get_current_backlight_16_bit(struct dce_abm *abm_dce) + /* notifyDMCUMsg */ + REG_UPDATE(MASTER_COMM_CNTL_REG, MASTER_COMM_INTERRUPT, 1); + + return true; +} + +static unsigned int calculate_16_bit_backlight_from_pwm(struct dce_abm *abm_dce) { uint64_t current_backlight; uint32_t round_result; @@ -103,45 +124,21 @@ static unsigned int get_current_backlight_16_bit(struct dce_abm *abm_dce) return (uint32_t)(current_backlight); } -static void driver_set_backlight_level(struct dce_abm *abm_dce, uint32_t level) +static void driver_set_backlight_level(struct dce_abm *abm_dce, + uint32_t backlight_pwm_u16_16) { - uint32_t backlight_24bit; - uint32_t backlight_17bit; uint32_t backlight_16bit; uint32_t masked_pwm_period; - uint8_t rounding_bit; uint8_t bit_count; uint64_t active_duty_cycle; uint32_t pwm_period_bitcnt; /* - * 1. Convert 8-bit value to 17 bit U1.16 format - * (1 integer, 16 fractional bits) - */ - - /* 1.1 multiply 8 bit value by 0x10101 to get a 24 bit value, - * effectively multiplying value by 256/255 - * eg. for a level of 0xEF, backlight_24bit = 0xEF * 0x10101 = 0xEFEFEF - */ - backlight_24bit = level * 0x10101; - - /* 1.2 The upper 16 bits of the 24 bit value is the fraction, lower 8 - * used for rounding, take most significant bit of fraction for - * rounding, e.g. for 0xEFEFEF, rounding bit is 1 - */ - rounding_bit = (backlight_24bit >> 7) & 1; - - /* 1.3 Add the upper 16 bits of the 24 bit value with the rounding bit - * resulting in a 17 bit value e.g. 0xEFF0 = (0xEFEFEF >> 8) + 1 - */ - backlight_17bit = (backlight_24bit >> 8) + rounding_bit; - - /* - * 2. Find 16 bit backlight active duty cycle, where 0 <= backlight + * 1. Find 16 bit backlight active duty cycle, where 0 <= backlight * active duty cycle <= backlight period */ - /* 2.1 Apply bitmask for backlight period value based on value of BITCNT + /* 1.1 Apply bitmask for backlight period value based on value of BITCNT */ REG_GET_2(BL_PWM_PERIOD_CNTL, BL_PWM_PERIOD_BITCNT, &pwm_period_bitcnt, @@ -155,13 +152,13 @@ static void driver_set_backlight_level(struct dce_abm *abm_dce, uint32_t level) /* e.g. maskedPwmPeriod = 0x24 when bitCount is 6 */ masked_pwm_period = masked_pwm_period & ((1 << bit_count) - 1); - /* 2.2 Calculate integer active duty cycle required upper 16 bits + /* 1.2 Calculate integer active duty cycle required upper 16 bits * contain integer component, lower 16 bits contain fractional component * of active duty cycle e.g. 0x21BDC0 = 0xEFF0 * 0x24 */ - active_duty_cycle = backlight_17bit * masked_pwm_period; + active_duty_cycle = backlight_pwm_u16_16 * masked_pwm_period; - /* 2.3 Calculate 16 bit active duty cycle from integer and fractional + /* 1.3 Calculate 16 bit active duty cycle from integer and fractional * components shift by bitCount then mask 16 bits and add rounding bit * from MSB of fraction e.g. 0x86F7 = ((0x21BDC0 >> 6) & 0xFFF) + 0 */ @@ -170,23 +167,23 @@ static void driver_set_backlight_level(struct dce_abm *abm_dce, uint32_t level) backlight_16bit += (active_duty_cycle >> (bit_count - 1)) & 0x1; /* - * 3. Program register with updated value + * 2. Program register with updated value */ - /* 3.1 Lock group 2 backlight registers */ + /* 2.1 Lock group 2 backlight registers */ REG_UPDATE_2(BL_PWM_GRP1_REG_LOCK, BL_PWM_GRP1_IGNORE_MASTER_LOCK_EN, 1, BL_PWM_GRP1_REG_LOCK, 1); - // 3.2 Write new active duty cycle + // 2.2 Write new active duty cycle REG_UPDATE(BL_PWM_CNTL, BL_ACTIVE_INT_FRAC_CNT, backlight_16bit); - /* 3.3 Unlock group 2 backlight registers */ + /* 2.3 Unlock group 2 backlight registers */ REG_UPDATE(BL_PWM_GRP1_REG_LOCK, BL_PWM_GRP1_REG_LOCK, 0); - /* 5.4.4 Wait for pending bit to be cleared */ + /* 3 Wait for pending bit to be cleared */ REG_WAIT(BL_PWM_GRP1_REG_LOCK, BL_PWM_GRP1_REG_UPDATE_PENDING, 0, 1, 10000); @@ -194,33 +191,28 @@ static void driver_set_backlight_level(struct dce_abm *abm_dce, uint32_t level) static void dmcu_set_backlight_level( struct dce_abm *abm_dce, - uint32_t level, + uint32_t backlight_pwm_u16_16, uint32_t frame_ramp, uint32_t controller_id) { - unsigned int backlight_16_bit = (level * 0x10101) >> 8; - unsigned int backlight_17_bit = backlight_16_bit + - (((backlight_16_bit & 0x80) >> 7) & 1); - uint32_t rampingBoundary = 0xFFFF; + unsigned int backlight_8_bit = 0; uint32_t s2; - /* set ramping boundary */ - REG_WRITE(MASTER_COMM_DATA_REG1, rampingBoundary); - - /* setDMCUParam_Pipe */ - REG_UPDATE_2(MASTER_COMM_CMD_REG, - MASTER_COMM_CMD_REG_BYTE0, MCP_ABM_PIPE_SET, - MASTER_COMM_CMD_REG_BYTE1, controller_id); + if (backlight_pwm_u16_16 & 0x10000) + // Check for max backlight condition + backlight_8_bit = 0xFF; + else + // Take MSB of fractional part since backlight is not max + backlight_8_bit = (backlight_pwm_u16_16 >> 8) & 0xFF; - /* notifyDMCUMsg */ - REG_UPDATE(MASTER_COMM_CNTL_REG, MASTER_COMM_INTERRUPT, 1); + dce_abm_set_pipe(&abm_dce->base, controller_id); /* waitDMCUReadyForCmd */ REG_WAIT(MASTER_COMM_CNTL_REG, MASTER_COMM_INTERRUPT, 0, 1, 80000); /* setDMCUParam_BL */ - REG_UPDATE(BL1_PWM_USER_LEVEL, BL1_PWM_USER_LEVEL, backlight_17_bit); + REG_UPDATE(BL1_PWM_USER_LEVEL, BL1_PWM_USER_LEVEL, backlight_pwm_u16_16); /* write ramp */ if (controller_id == 0) @@ -237,9 +229,9 @@ static void dmcu_set_backlight_level( s2 = REG_READ(BIOS_SCRATCH_2); s2 &= ~ATOM_S2_CURRENT_BL_LEVEL_MASK; - level &= (ATOM_S2_CURRENT_BL_LEVEL_MASK >> + backlight_8_bit &= (ATOM_S2_CURRENT_BL_LEVEL_MASK >> ATOM_S2_CURRENT_BL_LEVEL_SHIFT); - s2 |= (level << ATOM_S2_CURRENT_BL_LEVEL_SHIFT); + s2 |= (backlight_8_bit << ATOM_S2_CURRENT_BL_LEVEL_SHIFT); REG_WRITE(BIOS_SCRATCH_2, s2); } @@ -247,7 +239,7 @@ static void dmcu_set_backlight_level( static void dce_abm_init(struct abm *abm) { struct dce_abm *abm_dce = TO_DCE_ABM(abm); - unsigned int backlight = get_current_backlight_16_bit(abm_dce); + unsigned int backlight = calculate_16_bit_backlight_from_pwm(abm_dce); REG_WRITE(DC_ABM1_HG_SAMPLE_RATE, 0x103); REG_WRITE(DC_ABM1_HG_SAMPLE_RATE, 0x101); @@ -284,12 +276,26 @@ static void dce_abm_init(struct abm *abm) ABM1_BL_REG_READ_MISSED_FRAME_CLEAR, 1); } -static unsigned int dce_abm_get_current_backlight_8_bit(struct abm *abm) +static unsigned int dce_abm_get_current_backlight(struct abm *abm) { struct dce_abm *abm_dce = TO_DCE_ABM(abm); unsigned int backlight = REG_READ(BL1_PWM_CURRENT_ABM_LEVEL); - return (backlight >> 8); + /* return backlight in hardware format which is unsigned 17 bits, with + * 1 bit integer and 16 bit fractional + */ + return backlight; +} + +static unsigned int dce_abm_get_target_backlight(struct abm *abm) +{ + struct dce_abm *abm_dce = TO_DCE_ABM(abm); + unsigned int backlight = REG_READ(BL1_PWM_TARGET_ABM_LEVEL); + + /* return backlight in hardware format which is unsigned 17 bits, with + * 1 bit integer and 16 bit fractional + */ + return backlight; } static bool dce_abm_set_level(struct abm *abm, uint32_t level) @@ -314,16 +320,7 @@ static bool dce_abm_immediate_disable(struct abm *abm) { struct dce_abm *abm_dce = TO_DCE_ABM(abm); - REG_WAIT(MASTER_COMM_CNTL_REG, MASTER_COMM_INTERRUPT, 0, - 1, 80000); - - /* setDMCUParam_ABMLevel */ - REG_UPDATE_2(MASTER_COMM_CMD_REG, - MASTER_COMM_CMD_REG_BYTE0, MCP_ABM_LEVEL_SET, - MASTER_COMM_CMD_REG_BYTE2, MCP_DISABLE_ABM_IMMEDIATELY); - - /* notifyDMCUMsg */ - REG_UPDATE(MASTER_COMM_CNTL_REG, MASTER_COMM_INTERRUPT, 1); + dce_abm_set_pipe(abm, MCP_DISABLE_ABM_IMMEDIATELY); abm->stored_backlight_registers.BL_PWM_CNTL = REG_READ(BL_PWM_CNTL); @@ -396,9 +393,9 @@ static bool dce_abm_init_backlight(struct abm *abm) return true; } -static bool dce_abm_set_backlight_level( +static bool dce_abm_set_backlight_level_pwm( struct abm *abm, - unsigned int backlight_level, + unsigned int backlight_pwm_u16_16, unsigned int frame_ramp, unsigned int controller_id, bool use_smooth_brightness) @@ -406,16 +403,16 @@ static bool dce_abm_set_backlight_level( struct dce_abm *abm_dce = TO_DCE_ABM(abm); DC_LOG_BACKLIGHT("New Backlight level: %d (0x%X)\n", - backlight_level, backlight_level); + backlight_pwm_u16_16, backlight_pwm_u16_16); /* If DMCU is in reset state, DMCU is uninitialized */ if (use_smooth_brightness) dmcu_set_backlight_level(abm_dce, - backlight_level, + backlight_pwm_u16_16, frame_ramp, controller_id); else - driver_set_backlight_level(abm_dce, backlight_level); + driver_set_backlight_level(abm_dce, backlight_pwm_u16_16); return true; } @@ -424,8 +421,10 @@ static const struct abm_funcs dce_funcs = { .abm_init = dce_abm_init, .set_abm_level = dce_abm_set_level, .init_backlight = dce_abm_init_backlight, - .set_backlight_level = dce_abm_set_backlight_level, - .get_current_backlight_8_bit = dce_abm_get_current_backlight_8_bit, + .set_pipe = dce_abm_set_pipe, + .set_backlight_level_pwm = dce_abm_set_backlight_level_pwm, + .get_current_backlight = dce_abm_get_current_backlight, + .get_target_backlight = dce_abm_get_target_backlight, .set_abm_immediate_disable = dce_abm_immediate_disable }; diff --git a/drivers/gpu/drm/amd/display/dc/dce/dce_aux.c b/drivers/gpu/drm/amd/display/dc/dce/dce_aux.c index aaeb7faac0c4..4fe3664fb495 100644 --- a/drivers/gpu/drm/amd/display/dc/dce/dce_aux.c +++ b/drivers/gpu/drm/amd/display/dc/dce/dce_aux.c @@ -24,6 +24,7 @@ */ #include "dm_services.h" +#include "core_types.h" #include "dce_aux.h" #include "dce/dce_11_0_sh_mask.h" @@ -41,17 +42,17 @@ container_of((ptr), struct aux_engine_dce110, base) #define FROM_ENGINE(ptr) \ - FROM_AUX_ENGINE(container_of((ptr), struct aux_engine, base)) + FROM_AUX_ENGINE(container_of((ptr), struct dce_aux, base)) #define FROM_AUX_ENGINE_ENGINE(ptr) \ - container_of((ptr), struct aux_engine, base) + container_of((ptr), struct dce_aux, base) enum { AUX_INVALID_REPLY_RETRY_COUNTER = 1, AUX_TIMED_OUT_RETRY_COUNTER = 2, AUX_DEFER_RETRY_COUNTER = 6 }; static void release_engine( - struct aux_engine *engine) + struct dce_aux *engine) { struct aux_engine_dce110 *aux110 = FROM_AUX_ENGINE(engine); @@ -66,7 +67,7 @@ static void release_engine( #define DMCU_CAN_ACCESS_AUX 2 static bool is_engine_available( - struct aux_engine *engine) + struct dce_aux *engine) { struct aux_engine_dce110 *aux110 = FROM_AUX_ENGINE(engine); @@ -79,7 +80,7 @@ static bool is_engine_available( return (field != DMCU_CAN_ACCESS_AUX); } static bool acquire_engine( - struct aux_engine *engine) + struct dce_aux *engine) { struct aux_engine_dce110 *aux110 = FROM_AUX_ENGINE(engine); @@ -155,7 +156,7 @@ static bool acquire_engine( (0xFF & (address)) static void submit_channel_request( - struct aux_engine *engine, + struct dce_aux *engine, struct aux_request_transaction_data *request) { struct aux_engine_dce110 *aux110 = FROM_AUX_ENGINE(engine); @@ -189,6 +190,12 @@ static void submit_channel_request( 1, 0); } + + REG_UPDATE(AUX_INTERRUPT_CONTROL, AUX_SW_DONE_ACK, 1); + + REG_WAIT(AUX_SW_STATUS, AUX_SW_DONE, 0, + 10, aux110->timeout_period/10); + /* set the delay and the number of bytes to write */ /* The length include @@ -241,13 +248,10 @@ static void submit_channel_request( } } - REG_UPDATE(AUX_INTERRUPT_CONTROL, AUX_SW_DONE_ACK, 1); - REG_WAIT(AUX_SW_STATUS, AUX_SW_DONE, 0, - 10, aux110->timeout_period/10); REG_UPDATE(AUX_SW_CONTROL, AUX_SW_GO, 1); } -static int read_channel_reply(struct aux_engine *engine, uint32_t size, +static int read_channel_reply(struct dce_aux *engine, uint32_t size, uint8_t *buffer, uint8_t *reply_result, uint32_t *sw_status) { @@ -273,7 +277,8 @@ static int read_channel_reply(struct aux_engine *engine, uint32_t size, REG_GET(AUX_SW_DATA, AUX_SW_DATA, &reply_result_32); reply_result_32 = reply_result_32 >> 4; - *reply_result = (uint8_t)reply_result_32; + if (reply_result != NULL) + *reply_result = (uint8_t)reply_result_32; if (reply_result_32 == 0) { /* ACK */ uint32_t i = 0; @@ -299,61 +304,8 @@ static int read_channel_reply(struct aux_engine *engine, uint32_t size, return 0; } -static void process_channel_reply( - struct aux_engine *engine, - struct aux_reply_transaction_data *reply) -{ - int bytes_replied; - uint8_t reply_result; - uint32_t sw_status; - - bytes_replied = read_channel_reply(engine, reply->length, reply->data, - &reply_result, &sw_status); - - /* in case HPD is LOW, exit AUX transaction */ - if ((sw_status & AUX_SW_STATUS__AUX_SW_HPD_DISCON_MASK)) { - reply->status = AUX_TRANSACTION_REPLY_HPD_DISCON; - return; - } - - if (bytes_replied < 0) { - /* Need to handle an error case... - * Hopefully, upper layer function won't call this function if - * the number of bytes in the reply was 0, because there was - * surely an error that was asserted that should have been - * handled for hot plug case, this could happens - */ - if (!(sw_status & AUX_SW_STATUS__AUX_SW_HPD_DISCON_MASK)) { - reply->status = AUX_TRANSACTION_REPLY_INVALID; - ASSERT_CRITICAL(false); - return; - } - } else { - - switch (reply_result) { - case 0: /* ACK */ - reply->status = AUX_TRANSACTION_REPLY_AUX_ACK; - break; - case 1: /* NACK */ - reply->status = AUX_TRANSACTION_REPLY_AUX_NACK; - break; - case 2: /* DEFER */ - reply->status = AUX_TRANSACTION_REPLY_AUX_DEFER; - break; - case 4: /* AUX ACK / I2C NACK */ - reply->status = AUX_TRANSACTION_REPLY_I2C_NACK; - break; - case 8: /* AUX ACK / I2C DEFER */ - reply->status = AUX_TRANSACTION_REPLY_I2C_DEFER; - break; - default: - reply->status = AUX_TRANSACTION_REPLY_INVALID; - } - } -} - static enum aux_channel_operation_result get_channel_status( - struct aux_engine *engine, + struct dce_aux *engine, uint8_t *returned_bytes) { struct aux_engine_dce110 *aux110 = FROM_AUX_ENGINE(engine); @@ -414,469 +366,22 @@ static enum aux_channel_operation_result get_channel_status( return AUX_CHANNEL_OPERATION_FAILED_TIMEOUT; } } -static void process_read_reply( - struct aux_engine *engine, - struct read_command_context *ctx) -{ - engine->funcs->process_channel_reply(engine, &ctx->reply); - - switch (ctx->reply.status) { - case AUX_TRANSACTION_REPLY_AUX_ACK: - ctx->defer_retry_aux = 0; - if (ctx->returned_byte > ctx->current_read_length) { - ctx->status = - I2CAUX_TRANSACTION_STATUS_FAILED_PROTOCOL_ERROR; - ctx->operation_succeeded = false; - } else if (ctx->returned_byte < ctx->current_read_length) { - ctx->current_read_length -= ctx->returned_byte; - - ctx->offset += ctx->returned_byte; - - ++ctx->invalid_reply_retry_aux_on_ack; - - if (ctx->invalid_reply_retry_aux_on_ack > - AUX_INVALID_REPLY_RETRY_COUNTER) { - ctx->status = - I2CAUX_TRANSACTION_STATUS_FAILED_PROTOCOL_ERROR; - ctx->operation_succeeded = false; - } - } else { - ctx->status = I2CAUX_TRANSACTION_STATUS_SUCCEEDED; - ctx->transaction_complete = true; - ctx->operation_succeeded = true; - } - break; - case AUX_TRANSACTION_REPLY_AUX_NACK: - ctx->status = I2CAUX_TRANSACTION_STATUS_FAILED_NACK; - ctx->operation_succeeded = false; - break; - case AUX_TRANSACTION_REPLY_AUX_DEFER: - ++ctx->defer_retry_aux; - - if (ctx->defer_retry_aux > AUX_DEFER_RETRY_COUNTER) { - ctx->status = I2CAUX_TRANSACTION_STATUS_FAILED_TIMEOUT; - ctx->operation_succeeded = false; - } - break; - case AUX_TRANSACTION_REPLY_I2C_DEFER: - ctx->defer_retry_aux = 0; - - ++ctx->defer_retry_i2c; - - if (ctx->defer_retry_i2c > AUX_DEFER_RETRY_COUNTER) { - ctx->status = I2CAUX_TRANSACTION_STATUS_FAILED_TIMEOUT; - ctx->operation_succeeded = false; - } - break; - case AUX_TRANSACTION_REPLY_HPD_DISCON: - ctx->status = I2CAUX_TRANSACTION_STATUS_FAILED_HPD_DISCON; - ctx->operation_succeeded = false; - break; - default: - ctx->status = I2CAUX_TRANSACTION_STATUS_UNKNOWN; - ctx->operation_succeeded = false; - } -} -static void process_read_request( - struct aux_engine *engine, - struct read_command_context *ctx) -{ - enum aux_channel_operation_result operation_result; - - engine->funcs->submit_channel_request(engine, &ctx->request); - - operation_result = engine->funcs->get_channel_status( - engine, &ctx->returned_byte); - - switch (operation_result) { - case AUX_CHANNEL_OPERATION_SUCCEEDED: - if (ctx->returned_byte > ctx->current_read_length) { - ctx->status = - I2CAUX_TRANSACTION_STATUS_FAILED_PROTOCOL_ERROR; - ctx->operation_succeeded = false; - } else { - ctx->timed_out_retry_aux = 0; - ctx->invalid_reply_retry_aux = 0; - - ctx->reply.length = ctx->returned_byte; - ctx->reply.data = ctx->buffer; - - process_read_reply(engine, ctx); - } - break; - case AUX_CHANNEL_OPERATION_FAILED_INVALID_REPLY: - ++ctx->invalid_reply_retry_aux; - - if (ctx->invalid_reply_retry_aux > - AUX_INVALID_REPLY_RETRY_COUNTER) { - ctx->status = - I2CAUX_TRANSACTION_STATUS_FAILED_PROTOCOL_ERROR; - ctx->operation_succeeded = false; - } else - udelay(400); - break; - case AUX_CHANNEL_OPERATION_FAILED_TIMEOUT: - ++ctx->timed_out_retry_aux; - - if (ctx->timed_out_retry_aux > AUX_TIMED_OUT_RETRY_COUNTER) { - ctx->status = I2CAUX_TRANSACTION_STATUS_FAILED_TIMEOUT; - ctx->operation_succeeded = false; - } else { - /* DP 1.2a, table 2-58: - * "S3: AUX Request CMD PENDING: - * retry 3 times, with 400usec wait on each" - * The HW timeout is set to 550usec, - * so we should not wait here - */ - } - break; - case AUX_CHANNEL_OPERATION_FAILED_HPD_DISCON: - ctx->status = I2CAUX_TRANSACTION_STATUS_FAILED_HPD_DISCON; - ctx->operation_succeeded = false; - break; - default: - ctx->status = I2CAUX_TRANSACTION_STATUS_UNKNOWN; - ctx->operation_succeeded = false; - } -} -static bool read_command( - struct aux_engine *engine, - struct i2caux_transaction_request *request, - bool middle_of_transaction) -{ - struct read_command_context ctx; - - ctx.buffer = request->payload.data; - ctx.current_read_length = request->payload.length; - ctx.offset = 0; - ctx.timed_out_retry_aux = 0; - ctx.invalid_reply_retry_aux = 0; - ctx.defer_retry_aux = 0; - ctx.defer_retry_i2c = 0; - ctx.invalid_reply_retry_aux_on_ack = 0; - ctx.transaction_complete = false; - ctx.operation_succeeded = true; - - if (request->payload.address_space == - I2CAUX_TRANSACTION_ADDRESS_SPACE_DPCD) { - ctx.request.type = AUX_TRANSACTION_TYPE_DP; - ctx.request.action = I2CAUX_TRANSACTION_ACTION_DP_READ; - ctx.request.address = request->payload.address; - } else if (request->payload.address_space == - I2CAUX_TRANSACTION_ADDRESS_SPACE_I2C) { - ctx.request.type = AUX_TRANSACTION_TYPE_I2C; - ctx.request.action = middle_of_transaction ? - I2CAUX_TRANSACTION_ACTION_I2C_READ_MOT : - I2CAUX_TRANSACTION_ACTION_I2C_READ; - ctx.request.address = request->payload.address >> 1; - } else { - /* in DAL2, there was no return in such case */ - BREAK_TO_DEBUGGER(); - return false; - } - - ctx.request.delay = 0; - - do { - memset(ctx.buffer + ctx.offset, 0, ctx.current_read_length); - - ctx.request.data = ctx.buffer + ctx.offset; - ctx.request.length = ctx.current_read_length; - - process_read_request(engine, &ctx); - - request->status = ctx.status; - - if (ctx.operation_succeeded && !ctx.transaction_complete) - if (ctx.request.type == AUX_TRANSACTION_TYPE_I2C) - msleep(engine->delay); - } while (ctx.operation_succeeded && !ctx.transaction_complete); - - if (request->payload.address_space == - I2CAUX_TRANSACTION_ADDRESS_SPACE_DPCD) { - DC_LOG_I2C_AUX("READ: addr:0x%x value:0x%x Result:%d", - request->payload.address, - request->payload.data[0], - ctx.operation_succeeded); - } - - return ctx.operation_succeeded; -} - -static void process_write_reply( - struct aux_engine *engine, - struct write_command_context *ctx) -{ - engine->funcs->process_channel_reply(engine, &ctx->reply); - - switch (ctx->reply.status) { - case AUX_TRANSACTION_REPLY_AUX_ACK: - ctx->operation_succeeded = true; - - if (ctx->returned_byte) { - ctx->request.action = ctx->mot ? - I2CAUX_TRANSACTION_ACTION_I2C_STATUS_REQUEST_MOT : - I2CAUX_TRANSACTION_ACTION_I2C_STATUS_REQUEST; - - ctx->current_write_length = 0; - - ++ctx->ack_m_retry; - - if (ctx->ack_m_retry > AUX_DEFER_RETRY_COUNTER) { - ctx->status = - I2CAUX_TRANSACTION_STATUS_FAILED_TIMEOUT; - ctx->operation_succeeded = false; - } else - udelay(300); - } else { - ctx->status = I2CAUX_TRANSACTION_STATUS_SUCCEEDED; - ctx->defer_retry_aux = 0; - ctx->ack_m_retry = 0; - ctx->transaction_complete = true; - } - break; - case AUX_TRANSACTION_REPLY_AUX_NACK: - ctx->status = I2CAUX_TRANSACTION_STATUS_FAILED_NACK; - ctx->operation_succeeded = false; - break; - case AUX_TRANSACTION_REPLY_AUX_DEFER: - ++ctx->defer_retry_aux; - - if (ctx->defer_retry_aux > ctx->max_defer_retry) { - ctx->status = I2CAUX_TRANSACTION_STATUS_FAILED_TIMEOUT; - ctx->operation_succeeded = false; - } - break; - case AUX_TRANSACTION_REPLY_I2C_DEFER: - ctx->defer_retry_aux = 0; - ctx->current_write_length = 0; - - ctx->request.action = ctx->mot ? - I2CAUX_TRANSACTION_ACTION_I2C_STATUS_REQUEST_MOT : - I2CAUX_TRANSACTION_ACTION_I2C_STATUS_REQUEST; - - ++ctx->defer_retry_i2c; - - if (ctx->defer_retry_i2c > ctx->max_defer_retry) { - ctx->status = I2CAUX_TRANSACTION_STATUS_FAILED_TIMEOUT; - ctx->operation_succeeded = false; - } - break; - case AUX_TRANSACTION_REPLY_HPD_DISCON: - ctx->status = I2CAUX_TRANSACTION_STATUS_FAILED_HPD_DISCON; - ctx->operation_succeeded = false; - break; - default: - ctx->status = I2CAUX_TRANSACTION_STATUS_UNKNOWN; - ctx->operation_succeeded = false; - } -} -static void process_write_request( - struct aux_engine *engine, - struct write_command_context *ctx) -{ - enum aux_channel_operation_result operation_result; - - engine->funcs->submit_channel_request(engine, &ctx->request); - - operation_result = engine->funcs->get_channel_status( - engine, &ctx->returned_byte); - - switch (operation_result) { - case AUX_CHANNEL_OPERATION_SUCCEEDED: - ctx->timed_out_retry_aux = 0; - ctx->invalid_reply_retry_aux = 0; - - ctx->reply.length = ctx->returned_byte; - ctx->reply.data = ctx->reply_data; - - process_write_reply(engine, ctx); - break; - case AUX_CHANNEL_OPERATION_FAILED_INVALID_REPLY: - ++ctx->invalid_reply_retry_aux; - - if (ctx->invalid_reply_retry_aux > - AUX_INVALID_REPLY_RETRY_COUNTER) { - ctx->status = - I2CAUX_TRANSACTION_STATUS_FAILED_PROTOCOL_ERROR; - ctx->operation_succeeded = false; - } else - udelay(400); - break; - case AUX_CHANNEL_OPERATION_FAILED_TIMEOUT: - ++ctx->timed_out_retry_aux; - - if (ctx->timed_out_retry_aux > AUX_TIMED_OUT_RETRY_COUNTER) { - ctx->status = I2CAUX_TRANSACTION_STATUS_FAILED_TIMEOUT; - ctx->operation_succeeded = false; - } else { - /* DP 1.2a, table 2-58: - * "S3: AUX Request CMD PENDING: - * retry 3 times, with 400usec wait on each" - * The HW timeout is set to 550usec, - * so we should not wait here - */ - } - break; - case AUX_CHANNEL_OPERATION_FAILED_HPD_DISCON: - ctx->status = I2CAUX_TRANSACTION_STATUS_FAILED_HPD_DISCON; - ctx->operation_succeeded = false; - break; - default: - ctx->status = I2CAUX_TRANSACTION_STATUS_UNKNOWN; - ctx->operation_succeeded = false; - } -} -static bool write_command( - struct aux_engine *engine, - struct i2caux_transaction_request *request, - bool middle_of_transaction) -{ - struct write_command_context ctx; - - ctx.mot = middle_of_transaction; - ctx.buffer = request->payload.data; - ctx.current_write_length = request->payload.length; - ctx.timed_out_retry_aux = 0; - ctx.invalid_reply_retry_aux = 0; - ctx.defer_retry_aux = 0; - ctx.defer_retry_i2c = 0; - ctx.ack_m_retry = 0; - ctx.transaction_complete = false; - ctx.operation_succeeded = true; - - if (request->payload.address_space == - I2CAUX_TRANSACTION_ADDRESS_SPACE_DPCD) { - ctx.request.type = AUX_TRANSACTION_TYPE_DP; - ctx.request.action = I2CAUX_TRANSACTION_ACTION_DP_WRITE; - ctx.request.address = request->payload.address; - } else if (request->payload.address_space == - I2CAUX_TRANSACTION_ADDRESS_SPACE_I2C) { - ctx.request.type = AUX_TRANSACTION_TYPE_I2C; - ctx.request.action = middle_of_transaction ? - I2CAUX_TRANSACTION_ACTION_I2C_WRITE_MOT : - I2CAUX_TRANSACTION_ACTION_I2C_WRITE; - ctx.request.address = request->payload.address >> 1; - } else { - /* in DAL2, there was no return in such case */ - BREAK_TO_DEBUGGER(); - return false; - } - - ctx.request.delay = 0; - - ctx.max_defer_retry = - (engine->max_defer_write_retry > AUX_DEFER_RETRY_COUNTER) ? - engine->max_defer_write_retry : AUX_DEFER_RETRY_COUNTER; - - do { - ctx.request.data = ctx.buffer; - ctx.request.length = ctx.current_write_length; - - process_write_request(engine, &ctx); - - request->status = ctx.status; - - if (ctx.operation_succeeded && !ctx.transaction_complete) - if (ctx.request.type == AUX_TRANSACTION_TYPE_I2C) - msleep(engine->delay); - } while (ctx.operation_succeeded && !ctx.transaction_complete); - - if (request->payload.address_space == - I2CAUX_TRANSACTION_ADDRESS_SPACE_DPCD) { - DC_LOG_I2C_AUX("WRITE: addr:0x%x value:0x%x Result:%d", - request->payload.address, - request->payload.data[0], - ctx.operation_succeeded); - } - - return ctx.operation_succeeded; -} -static bool end_of_transaction_command( - struct aux_engine *engine, - struct i2caux_transaction_request *request) -{ - struct i2caux_transaction_request dummy_request; - uint8_t dummy_data; - - /* [tcheng] We only need to send the stop (read with MOT = 0) - * for I2C-over-Aux, not native AUX - */ - - if (request->payload.address_space != - I2CAUX_TRANSACTION_ADDRESS_SPACE_I2C) - return false; - - dummy_request.operation = request->operation; - dummy_request.payload.address_space = request->payload.address_space; - dummy_request.payload.address = request->payload.address; - - /* - * Add a dummy byte due to some receiver quirk - * where one byte is sent along with MOT = 0. - * Ideally this should be 0. - */ - - dummy_request.payload.length = 0; - dummy_request.payload.data = &dummy_data; - - if (request->operation == I2CAUX_TRANSACTION_READ) - return read_command(engine, &dummy_request, false); - else - return write_command(engine, &dummy_request, false); - /* according Syed, it does not need now DoDummyMOT */ -} -static bool submit_request( - struct aux_engine *engine, - struct i2caux_transaction_request *request, - bool middle_of_transaction) -{ - - bool result; - bool mot_used = true; - - switch (request->operation) { - case I2CAUX_TRANSACTION_READ: - result = read_command(engine, request, mot_used); - break; - case I2CAUX_TRANSACTION_WRITE: - result = write_command(engine, request, mot_used); - break; - default: - result = false; - } - - /* [tcheng] - * need to send stop for the last transaction to free up the AUX - * if the above command fails, this would be the last transaction - */ - - if (!middle_of_transaction || !result) - end_of_transaction_command(engine, request); - - /* mask AUX interrupt */ - - return result; -} enum i2caux_engine_type get_engine_type( - const struct aux_engine *engine) + const struct dce_aux *engine) { return I2CAUX_ENGINE_TYPE_AUX; } static bool acquire( - struct aux_engine *engine, + struct dce_aux *engine, struct ddc *ddc) { enum gpio_result result; - if (engine->funcs->is_engine_available) { - /*check whether SW could use the engine*/ - if (!engine->funcs->is_engine_available(engine)) - return false; - } + if (!is_engine_available(engine)) + return false; result = dal_ddc_open(ddc, GPIO_MODE_HARDWARE, GPIO_DDC_CONFIG_TYPE_MODE_AUX); @@ -884,7 +389,7 @@ static bool acquire( if (result != GPIO_RESULT_OK) return false; - if (!engine->funcs->acquire_engine(engine)) { + if (!acquire_engine(engine)) { dal_ddc_close(ddc); return false; } @@ -894,21 +399,7 @@ static bool acquire( return true; } -static const struct aux_engine_funcs aux_engine_funcs = { - .acquire_engine = acquire_engine, - .submit_channel_request = submit_channel_request, - .process_channel_reply = process_channel_reply, - .read_channel_reply = read_channel_reply, - .get_channel_status = get_channel_status, - .is_engine_available = is_engine_available, - .release_engine = release_engine, - .destroy_engine = dce110_engine_destroy, - .submit_request = submit_request, - .get_engine_type = get_engine_type, - .acquire = acquire, -}; - -void dce110_engine_destroy(struct aux_engine **engine) +void dce110_engine_destroy(struct dce_aux **engine) { struct aux_engine_dce110 *engine110 = FROM_AUX_ENGINE(*engine); @@ -917,7 +408,7 @@ void dce110_engine_destroy(struct aux_engine **engine) *engine = NULL; } -struct aux_engine *dce110_aux_engine_construct(struct aux_engine_dce110 *aux_engine110, +struct dce_aux *dce110_aux_engine_construct(struct aux_engine_dce110 *aux_engine110, struct dc_context *ctx, uint32_t inst, uint32_t timeout_period, @@ -927,7 +418,6 @@ struct aux_engine *dce110_aux_engine_construct(struct aux_engine_dce110 *aux_eng aux_engine110->base.ctx = ctx; aux_engine110->base.delay = 0; aux_engine110->base.max_defer_write_retry = 0; - aux_engine110->base.funcs = &aux_engine_funcs; aux_engine110->base.inst = inst; aux_engine110->timeout_period = timeout_period; aux_engine110->regs = regs; @@ -935,3 +425,101 @@ struct aux_engine *dce110_aux_engine_construct(struct aux_engine_dce110 *aux_eng return &aux_engine110->base; } +static enum i2caux_transaction_action i2caux_action_from_payload(struct aux_payload *payload) +{ + if (payload->i2c_over_aux) { + if (payload->write) { + if (payload->mot) + return I2CAUX_TRANSACTION_ACTION_I2C_WRITE_MOT; + return I2CAUX_TRANSACTION_ACTION_I2C_WRITE; + } + if (payload->mot) + return I2CAUX_TRANSACTION_ACTION_I2C_READ_MOT; + return I2CAUX_TRANSACTION_ACTION_I2C_READ; + } + if (payload->write) + return I2CAUX_TRANSACTION_ACTION_DP_WRITE; + return I2CAUX_TRANSACTION_ACTION_DP_READ; +} + +int dce_aux_transfer(struct ddc_service *ddc, + struct aux_payload *payload) +{ + struct ddc *ddc_pin = ddc->ddc_pin; + struct dce_aux *aux_engine; + enum aux_channel_operation_result operation_result; + struct aux_request_transaction_data aux_req; + struct aux_reply_transaction_data aux_rep; + uint8_t returned_bytes = 0; + int res = -1; + uint32_t status; + + memset(&aux_req, 0, sizeof(aux_req)); + memset(&aux_rep, 0, sizeof(aux_rep)); + + aux_engine = ddc->ctx->dc->res_pool->engines[ddc_pin->pin_data->en]; + acquire(aux_engine, ddc_pin); + + if (payload->i2c_over_aux) + aux_req.type = AUX_TRANSACTION_TYPE_I2C; + else + aux_req.type = AUX_TRANSACTION_TYPE_DP; + + aux_req.action = i2caux_action_from_payload(payload); + + aux_req.address = payload->address; + aux_req.delay = payload->defer_delay * 10; + aux_req.length = payload->length; + aux_req.data = payload->data; + + submit_channel_request(aux_engine, &aux_req); + operation_result = get_channel_status(aux_engine, &returned_bytes); + + switch (operation_result) { + case AUX_CHANNEL_OPERATION_SUCCEEDED: + res = read_channel_reply(aux_engine, payload->length, + payload->data, payload->reply, + &status); + break; + case AUX_CHANNEL_OPERATION_FAILED_HPD_DISCON: + res = 0; + break; + case AUX_CHANNEL_OPERATION_FAILED_REASON_UNKNOWN: + case AUX_CHANNEL_OPERATION_FAILED_INVALID_REPLY: + case AUX_CHANNEL_OPERATION_FAILED_TIMEOUT: + res = -1; + break; + } + release_engine(aux_engine); + return res; +} + +#define AUX_RETRY_MAX 7 + +bool dce_aux_transfer_with_retries(struct ddc_service *ddc, + struct aux_payload *payload) +{ + int i, ret = 0; + uint8_t reply; + bool payload_reply = true; + + if (!payload->reply) { + payload_reply = false; + payload->reply = &reply; + } + + for (i = 0; i < AUX_RETRY_MAX; i++) { + ret = dce_aux_transfer(ddc, payload); + + if (ret >= 0) { + if (*payload->reply == 0) { + if (!payload_reply) + payload->reply = NULL; + return true; + } + } + + udelay(1000); + } + return false; +} diff --git a/drivers/gpu/drm/amd/display/dc/dce/dce_aux.h b/drivers/gpu/drm/amd/display/dc/dce/dce_aux.h index f7caab85dc80..e28ed6a00ff4 100644 --- a/drivers/gpu/drm/amd/display/dc/dce/dce_aux.h +++ b/drivers/gpu/drm/amd/display/dc/dce/dce_aux.h @@ -25,7 +25,9 @@ #ifndef __DAL_AUX_ENGINE_DCE110_H__ #define __DAL_AUX_ENGINE_DCE110_H__ -#include "aux_engine.h" + +#include "i2caux_interface.h" +#include "inc/hw/aux_engine.h" #define AUX_COMMON_REG_LIST(id)\ SRI(AUX_CONTROL, DP_AUX, id), \ @@ -69,14 +71,26 @@ enum { /* This is the timeout as defined in DP 1.2a, * at most within ~240usec. That means, * increasing this timeout will not affect normal operation, * and we'll timeout after - * SW_AUX_TIMEOUT_PERIOD_MULTIPLIER * AUX_TIMEOUT_PERIOD = 1600usec. + * SW_AUX_TIMEOUT_PERIOD_MULTIPLIER * AUX_TIMEOUT_PERIOD = 2400usec. * This timeout is especially important for - * resume from S3 and CTS. + * converters, resume from S3, and CTS. */ - SW_AUX_TIMEOUT_PERIOD_MULTIPLIER = 4 + SW_AUX_TIMEOUT_PERIOD_MULTIPLIER = 6 +}; + +struct dce_aux { + uint32_t inst; + struct ddc *ddc; + struct dc_context *ctx; + /* following values are expressed in milliseconds */ + uint32_t delay; + uint32_t max_defer_write_retry; + + bool acquire_reset; }; + struct aux_engine_dce110 { - struct aux_engine base; + struct dce_aux base; const struct dce110_aux_registers *regs; struct { uint32_t aux_control; @@ -96,16 +110,22 @@ struct aux_engine_dce110_init_data { const struct dce110_aux_registers *regs; }; -struct aux_engine *dce110_aux_engine_construct( +struct dce_aux *dce110_aux_engine_construct( struct aux_engine_dce110 *aux_engine110, struct dc_context *ctx, uint32_t inst, uint32_t timeout_period, const struct dce110_aux_registers *regs); -void dce110_engine_destroy(struct aux_engine **engine); +void dce110_engine_destroy(struct dce_aux **engine); bool dce110_aux_engine_acquire( - struct aux_engine *aux_engine, + struct dce_aux *aux_engine, struct ddc *ddc); + +int dce_aux_transfer(struct ddc_service *ddc, + struct aux_payload *cmd); + +bool dce_aux_transfer_with_retries(struct ddc_service *ddc, + struct aux_payload *cmd); #endif diff --git a/drivers/gpu/drm/amd/display/dc/dce/dce_clk_mgr.c b/drivers/gpu/drm/amd/display/dc/dce/dce_clk_mgr.c new file mode 100644 index 000000000000..6e142c2db986 --- /dev/null +++ b/drivers/gpu/drm/amd/display/dc/dce/dce_clk_mgr.c @@ -0,0 +1,961 @@ +/* + * Copyright 2012-16 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. + * + * Authors: AMD + * + */ + +#include "dce_clk_mgr.h" + +#include "reg_helper.h" +#include "dmcu.h" +#include "core_types.h" +#include "dal_asic_id.h" + +#define TO_DCE_CLK_MGR(clocks)\ + container_of(clocks, struct dce_clk_mgr, base) + +#define REG(reg) \ + (clk_mgr_dce->regs->reg) + +#undef FN +#define FN(reg_name, field_name) \ + clk_mgr_dce->clk_mgr_shift->field_name, clk_mgr_dce->clk_mgr_mask->field_name + +#define CTX \ + clk_mgr_dce->base.ctx +#define DC_LOGGER \ + clk_mgr->ctx->logger + +/* Max clock values for each state indexed by "enum clocks_state": */ +static const struct state_dependent_clocks dce80_max_clks_by_state[] = { +/* ClocksStateInvalid - should not be used */ +{ .display_clk_khz = 0, .pixel_clk_khz = 0 }, +/* ClocksStateUltraLow - not expected to be used for DCE 8.0 */ +{ .display_clk_khz = 0, .pixel_clk_khz = 0 }, +/* ClocksStateLow */ +{ .display_clk_khz = 352000, .pixel_clk_khz = 330000}, +/* ClocksStateNominal */ +{ .display_clk_khz = 600000, .pixel_clk_khz = 400000 }, +/* ClocksStatePerformance */ +{ .display_clk_khz = 600000, .pixel_clk_khz = 400000 } }; + +static const struct state_dependent_clocks dce110_max_clks_by_state[] = { +/*ClocksStateInvalid - should not be used*/ +{ .display_clk_khz = 0, .pixel_clk_khz = 0 }, +/*ClocksStateUltraLow - currently by HW design team not supposed to be used*/ +{ .display_clk_khz = 352000, .pixel_clk_khz = 330000 }, +/*ClocksStateLow*/ +{ .display_clk_khz = 352000, .pixel_clk_khz = 330000 }, +/*ClocksStateNominal*/ +{ .display_clk_khz = 467000, .pixel_clk_khz = 400000 }, +/*ClocksStatePerformance*/ +{ .display_clk_khz = 643000, .pixel_clk_khz = 400000 } }; + +static const struct state_dependent_clocks dce112_max_clks_by_state[] = { +/*ClocksStateInvalid - should not be used*/ +{ .display_clk_khz = 0, .pixel_clk_khz = 0 }, +/*ClocksStateUltraLow - currently by HW design team not supposed to be used*/ +{ .display_clk_khz = 389189, .pixel_clk_khz = 346672 }, +/*ClocksStateLow*/ +{ .display_clk_khz = 459000, .pixel_clk_khz = 400000 }, +/*ClocksStateNominal*/ +{ .display_clk_khz = 667000, .pixel_clk_khz = 600000 }, +/*ClocksStatePerformance*/ +{ .display_clk_khz = 1132000, .pixel_clk_khz = 600000 } }; + +static const struct state_dependent_clocks dce120_max_clks_by_state[] = { +/*ClocksStateInvalid - should not be used*/ +{ .display_clk_khz = 0, .pixel_clk_khz = 0 }, +/*ClocksStateUltraLow - currently by HW design team not supposed to be used*/ +{ .display_clk_khz = 0, .pixel_clk_khz = 0 }, +/*ClocksStateLow*/ +{ .display_clk_khz = 460000, .pixel_clk_khz = 400000 }, +/*ClocksStateNominal*/ +{ .display_clk_khz = 670000, .pixel_clk_khz = 600000 }, +/*ClocksStatePerformance*/ +{ .display_clk_khz = 1133000, .pixel_clk_khz = 600000 } }; + +int dentist_get_divider_from_did(int did) +{ + if (did < DENTIST_BASE_DID_1) + did = DENTIST_BASE_DID_1; + if (did > DENTIST_MAX_DID) + did = DENTIST_MAX_DID; + + if (did < DENTIST_BASE_DID_2) { + return DENTIST_DIVIDER_RANGE_1_START + DENTIST_DIVIDER_RANGE_1_STEP + * (did - DENTIST_BASE_DID_1); + } else if (did < DENTIST_BASE_DID_3) { + return DENTIST_DIVIDER_RANGE_2_START + DENTIST_DIVIDER_RANGE_2_STEP + * (did - DENTIST_BASE_DID_2); + } else if (did < DENTIST_BASE_DID_4) { + return DENTIST_DIVIDER_RANGE_3_START + DENTIST_DIVIDER_RANGE_3_STEP + * (did - DENTIST_BASE_DID_3); + } else { + return DENTIST_DIVIDER_RANGE_4_START + DENTIST_DIVIDER_RANGE_4_STEP + * (did - DENTIST_BASE_DID_4); + } +} + +/* SW will adjust DP REF Clock average value for all purposes + * (DP DTO / DP Audio DTO and DP GTC) + if clock is spread for all cases: + -if SS enabled on DP Ref clock and HW de-spreading enabled with SW + calculations for DS_INCR/DS_MODULO (this is planned to be default case) + -if SS enabled on DP Ref clock and HW de-spreading enabled with HW + calculations (not planned to be used, but average clock should still + be valid) + -if SS enabled on DP Ref clock and HW de-spreading disabled + (should not be case with CIK) then SW should program all rates + generated according to average value (case as with previous ASICs) + */ +static int clk_mgr_adjust_dp_ref_freq_for_ss(struct dce_clk_mgr *clk_mgr_dce, int dp_ref_clk_khz) +{ + if (clk_mgr_dce->ss_on_dprefclk && clk_mgr_dce->dprefclk_ss_divider != 0) { + struct fixed31_32 ss_percentage = dc_fixpt_div_int( + dc_fixpt_from_fraction(clk_mgr_dce->dprefclk_ss_percentage, + clk_mgr_dce->dprefclk_ss_divider), 200); + struct fixed31_32 adj_dp_ref_clk_khz; + + ss_percentage = dc_fixpt_sub(dc_fixpt_one, ss_percentage); + adj_dp_ref_clk_khz = dc_fixpt_mul_int(ss_percentage, dp_ref_clk_khz); + dp_ref_clk_khz = dc_fixpt_floor(adj_dp_ref_clk_khz); + } + return dp_ref_clk_khz; +} + +static int dce_get_dp_ref_freq_khz(struct clk_mgr *clk_mgr) +{ + struct dce_clk_mgr *clk_mgr_dce = TO_DCE_CLK_MGR(clk_mgr); + int dprefclk_wdivider; + int dprefclk_src_sel; + int dp_ref_clk_khz = 600000; + int target_div; + + /* ASSERT DP Reference Clock source is from DFS*/ + REG_GET(DPREFCLK_CNTL, DPREFCLK_SRC_SEL, &dprefclk_src_sel); + ASSERT(dprefclk_src_sel == 0); + + /* Read the mmDENTIST_DISPCLK_CNTL to get the currently + * programmed DID DENTIST_DPREFCLK_WDIVIDER*/ + REG_GET(DENTIST_DISPCLK_CNTL, DENTIST_DPREFCLK_WDIVIDER, &dprefclk_wdivider); + + /* Convert DENTIST_DPREFCLK_WDIVIDERto actual divider*/ + target_div = dentist_get_divider_from_did(dprefclk_wdivider); + + /* Calculate the current DFS clock, in kHz.*/ + dp_ref_clk_khz = (DENTIST_DIVIDER_RANGE_SCALE_FACTOR + * clk_mgr_dce->dentist_vco_freq_khz) / target_div; + + return clk_mgr_adjust_dp_ref_freq_for_ss(clk_mgr_dce, dp_ref_clk_khz); +} + +int dce12_get_dp_ref_freq_khz(struct clk_mgr *clk_mgr) +{ + struct dce_clk_mgr *clk_mgr_dce = TO_DCE_CLK_MGR(clk_mgr); + + return clk_mgr_adjust_dp_ref_freq_for_ss(clk_mgr_dce, clk_mgr_dce->dprefclk_khz); +} + +/* unit: in_khz before mode set, get pixel clock from context. ASIC register + * may not be programmed yet + */ +static uint32_t get_max_pixel_clock_for_all_paths(struct dc_state *context) +{ + uint32_t max_pix_clk = 0; + int i; + + for (i = 0; i < MAX_PIPES; i++) { + struct pipe_ctx *pipe_ctx = &context->res_ctx.pipe_ctx[i]; + + if (pipe_ctx->stream == NULL) + continue; + + /* do not check under lay */ + if (pipe_ctx->top_pipe) + continue; + + if (pipe_ctx->stream_res.pix_clk_params.requested_pix_clk_100hz / 10 > max_pix_clk) + max_pix_clk = pipe_ctx->stream_res.pix_clk_params.requested_pix_clk_100hz / 10; + + /* raise clock state for HBR3/2 if required. Confirmed with HW DCE/DPCS + * logic for HBR3 still needs Nominal (0.8V) on VDDC rail + */ + if (dc_is_dp_signal(pipe_ctx->stream->signal) && + pipe_ctx->stream_res.pix_clk_params.requested_sym_clk > max_pix_clk) + max_pix_clk = pipe_ctx->stream_res.pix_clk_params.requested_sym_clk; + } + + return max_pix_clk; +} + +static enum dm_pp_clocks_state dce_get_required_clocks_state( + struct clk_mgr *clk_mgr, + struct dc_state *context) +{ + struct dce_clk_mgr *clk_mgr_dce = TO_DCE_CLK_MGR(clk_mgr); + int i; + enum dm_pp_clocks_state low_req_clk; + int max_pix_clk = get_max_pixel_clock_for_all_paths(context); + + /* Iterate from highest supported to lowest valid state, and update + * lowest RequiredState with the lowest state that satisfies + * all required clocks + */ + for (i = clk_mgr_dce->max_clks_state; i >= DM_PP_CLOCKS_STATE_ULTRA_LOW; i--) + if (context->bw.dce.dispclk_khz > + clk_mgr_dce->max_clks_by_state[i].display_clk_khz + || max_pix_clk > + clk_mgr_dce->max_clks_by_state[i].pixel_clk_khz) + break; + + low_req_clk = i + 1; + if (low_req_clk > clk_mgr_dce->max_clks_state) { + /* set max clock state for high phyclock, invalid on exceeding display clock */ + if (clk_mgr_dce->max_clks_by_state[clk_mgr_dce->max_clks_state].display_clk_khz + < context->bw.dce.dispclk_khz) + low_req_clk = DM_PP_CLOCKS_STATE_INVALID; + else + low_req_clk = clk_mgr_dce->max_clks_state; + } + + return low_req_clk; +} + +static int dce_set_clock( + struct clk_mgr *clk_mgr, + int requested_clk_khz) +{ + struct dce_clk_mgr *clk_mgr_dce = TO_DCE_CLK_MGR(clk_mgr); + struct bp_pixel_clock_parameters pxl_clk_params = { 0 }; + struct dc_bios *bp = clk_mgr->ctx->dc_bios; + int actual_clock = requested_clk_khz; + struct dmcu *dmcu = clk_mgr_dce->base.ctx->dc->res_pool->dmcu; + + /* Make sure requested clock isn't lower than minimum threshold*/ + if (requested_clk_khz > 0) + requested_clk_khz = max(requested_clk_khz, + clk_mgr_dce->dentist_vco_freq_khz / 64); + + /* Prepare to program display clock*/ + pxl_clk_params.target_pixel_clock_100hz = requested_clk_khz * 10; + pxl_clk_params.pll_id = CLOCK_SOURCE_ID_DFS; + + if (clk_mgr_dce->dfs_bypass_active) + pxl_clk_params.flags.SET_DISPCLK_DFS_BYPASS = true; + + bp->funcs->program_display_engine_pll(bp, &pxl_clk_params); + + if (clk_mgr_dce->dfs_bypass_active) { + /* Cache the fixed display clock*/ + clk_mgr_dce->dfs_bypass_disp_clk = + pxl_clk_params.dfs_bypass_display_clock; + actual_clock = pxl_clk_params.dfs_bypass_display_clock; + } + + /* from power down, we need mark the clock state as ClocksStateNominal + * from HWReset, so when resume we will call pplib voltage regulator.*/ + if (requested_clk_khz == 0) + clk_mgr_dce->cur_min_clks_state = DM_PP_CLOCKS_STATE_NOMINAL; + + if (dmcu && dmcu->funcs->is_dmcu_initialized(dmcu)) + dmcu->funcs->set_psr_wait_loop(dmcu, actual_clock / 1000 / 7); + + return actual_clock; +} + +int dce112_set_clock(struct clk_mgr *clk_mgr, int requested_clk_khz) +{ + struct dce_clk_mgr *clk_mgr_dce = TO_DCE_CLK_MGR(clk_mgr); + struct bp_set_dce_clock_parameters dce_clk_params; + struct dc_bios *bp = clk_mgr->ctx->dc_bios; + struct dc *core_dc = clk_mgr->ctx->dc; + struct dmcu *dmcu = core_dc->res_pool->dmcu; + int actual_clock = requested_clk_khz; + /* Prepare to program display clock*/ + memset(&dce_clk_params, 0, sizeof(dce_clk_params)); + + /* Make sure requested clock isn't lower than minimum threshold*/ + if (requested_clk_khz > 0) + requested_clk_khz = max(requested_clk_khz, + clk_mgr_dce->dentist_vco_freq_khz / 62); + + dce_clk_params.target_clock_frequency = requested_clk_khz; + dce_clk_params.pll_id = CLOCK_SOURCE_ID_DFS; + dce_clk_params.clock_type = DCECLOCK_TYPE_DISPLAY_CLOCK; + + bp->funcs->set_dce_clock(bp, &dce_clk_params); + actual_clock = dce_clk_params.target_clock_frequency; + + /* from power down, we need mark the clock state as ClocksStateNominal + * from HWReset, so when resume we will call pplib voltage regulator.*/ + if (requested_clk_khz == 0) + clk_mgr_dce->cur_min_clks_state = DM_PP_CLOCKS_STATE_NOMINAL; + + /*Program DP ref Clock*/ + /*VBIOS will determine DPREFCLK frequency, so we don't set it*/ + dce_clk_params.target_clock_frequency = 0; + dce_clk_params.clock_type = DCECLOCK_TYPE_DPREFCLK; + if (!ASICREV_IS_VEGA20_P(clk_mgr->ctx->asic_id.hw_internal_rev)) + dce_clk_params.flags.USE_GENLOCK_AS_SOURCE_FOR_DPREFCLK = + (dce_clk_params.pll_id == + CLOCK_SOURCE_COMBO_DISPLAY_PLL0); + else + dce_clk_params.flags.USE_GENLOCK_AS_SOURCE_FOR_DPREFCLK = false; + + bp->funcs->set_dce_clock(bp, &dce_clk_params); + + if (!IS_FPGA_MAXIMUS_DC(core_dc->ctx->dce_environment)) { + if (dmcu && dmcu->funcs->is_dmcu_initialized(dmcu)) { + if (clk_mgr_dce->dfs_bypass_disp_clk != actual_clock) + dmcu->funcs->set_psr_wait_loop(dmcu, + actual_clock / 1000 / 7); + } + } + + clk_mgr_dce->dfs_bypass_disp_clk = actual_clock; + return actual_clock; +} + +static void dce_clock_read_integrated_info(struct dce_clk_mgr *clk_mgr_dce) +{ + struct dc_debug_options *debug = &clk_mgr_dce->base.ctx->dc->debug; + struct dc_bios *bp = clk_mgr_dce->base.ctx->dc_bios; + struct integrated_info info = { { { 0 } } }; + struct dc_firmware_info fw_info = { { 0 } }; + int i; + + if (bp->integrated_info) + info = *bp->integrated_info; + + clk_mgr_dce->dentist_vco_freq_khz = info.dentist_vco_freq; + if (clk_mgr_dce->dentist_vco_freq_khz == 0) { + bp->funcs->get_firmware_info(bp, &fw_info); + clk_mgr_dce->dentist_vco_freq_khz = + fw_info.smu_gpu_pll_output_freq; + if (clk_mgr_dce->dentist_vco_freq_khz == 0) + clk_mgr_dce->dentist_vco_freq_khz = 3600000; + } + + /*update the maximum display clock for each power state*/ + for (i = 0; i < NUMBER_OF_DISP_CLK_VOLTAGE; ++i) { + enum dm_pp_clocks_state clk_state = DM_PP_CLOCKS_STATE_INVALID; + + switch (i) { + case 0: + clk_state = DM_PP_CLOCKS_STATE_ULTRA_LOW; + break; + + case 1: + clk_state = DM_PP_CLOCKS_STATE_LOW; + break; + + case 2: + clk_state = DM_PP_CLOCKS_STATE_NOMINAL; + break; + + case 3: + clk_state = DM_PP_CLOCKS_STATE_PERFORMANCE; + break; + + default: + clk_state = DM_PP_CLOCKS_STATE_INVALID; + break; + } + + /*Do not allow bad VBIOS/SBIOS to override with invalid values, + * check for > 100MHz*/ + if (info.disp_clk_voltage[i].max_supported_clk >= 100000) + clk_mgr_dce->max_clks_by_state[clk_state].display_clk_khz = + info.disp_clk_voltage[i].max_supported_clk; + } + + if (!debug->disable_dfs_bypass && bp->integrated_info) + if (bp->integrated_info->gpu_cap_info & DFS_BYPASS_ENABLE) + clk_mgr_dce->dfs_bypass_enabled = true; +} + +void dce_clock_read_ss_info(struct dce_clk_mgr *clk_mgr_dce) +{ + struct dc_bios *bp = clk_mgr_dce->base.ctx->dc_bios; + int ss_info_num = bp->funcs->get_ss_entry_number( + bp, AS_SIGNAL_TYPE_GPU_PLL); + + if (ss_info_num) { + struct spread_spectrum_info info = { { 0 } }; + enum bp_result result = bp->funcs->get_spread_spectrum_info( + bp, AS_SIGNAL_TYPE_GPU_PLL, 0, &info); + + /* Based on VBIOS, VBIOS will keep entry for GPU PLL SS + * even if SS not enabled and in that case + * SSInfo.spreadSpectrumPercentage !=0 would be sign + * that SS is enabled + */ + if (result == BP_RESULT_OK && + info.spread_spectrum_percentage != 0) { + clk_mgr_dce->ss_on_dprefclk = true; + clk_mgr_dce->dprefclk_ss_divider = info.spread_percentage_divider; + + if (info.type.CENTER_MODE == 0) { + /* TODO: Currently for DP Reference clock we + * need only SS percentage for + * downspread */ + clk_mgr_dce->dprefclk_ss_percentage = + info.spread_spectrum_percentage; + } + + return; + } + + result = bp->funcs->get_spread_spectrum_info( + bp, AS_SIGNAL_TYPE_DISPLAY_PORT, 0, &info); + + /* Based on VBIOS, VBIOS will keep entry for DPREFCLK SS + * even if SS not enabled and in that case + * SSInfo.spreadSpectrumPercentage !=0 would be sign + * that SS is enabled + */ + if (result == BP_RESULT_OK && + info.spread_spectrum_percentage != 0) { + clk_mgr_dce->ss_on_dprefclk = true; + clk_mgr_dce->dprefclk_ss_divider = info.spread_percentage_divider; + + if (info.type.CENTER_MODE == 0) { + /* Currently for DP Reference clock we + * need only SS percentage for + * downspread */ + clk_mgr_dce->dprefclk_ss_percentage = + info.spread_spectrum_percentage; + } + } + } +} + +/** + * dce121_clock_patch_xgmi_ss_info() - Save XGMI spread spectrum info + * @clk_mgr: clock manager base structure + * + * Reads from VBIOS the XGMI spread spectrum info and saves it within + * the dce clock manager. This operation will overwrite the existing dprefclk + * SS values if the vBIOS query succeeds. Otherwise, it does nothing. It also + * sets the ->xgmi_enabled flag. + */ +void dce121_clock_patch_xgmi_ss_info(struct clk_mgr *clk_mgr) +{ + struct dce_clk_mgr *clk_mgr_dce = TO_DCE_CLK_MGR(clk_mgr); + enum bp_result result; + struct spread_spectrum_info info = { { 0 } }; + struct dc_bios *bp = clk_mgr_dce->base.ctx->dc_bios; + + clk_mgr_dce->xgmi_enabled = false; + + result = bp->funcs->get_spread_spectrum_info(bp, AS_SIGNAL_TYPE_XGMI, + 0, &info); + if (result == BP_RESULT_OK && info.spread_spectrum_percentage != 0) { + clk_mgr_dce->xgmi_enabled = true; + clk_mgr_dce->ss_on_dprefclk = true; + clk_mgr_dce->dprefclk_ss_divider = + info.spread_percentage_divider; + + if (info.type.CENTER_MODE == 0) { + /* Currently for DP Reference clock we + * need only SS percentage for + * downspread */ + clk_mgr_dce->dprefclk_ss_percentage = + info.spread_spectrum_percentage; + } + } +} + +void dce110_fill_display_configs( + const struct dc_state *context, + struct dm_pp_display_configuration *pp_display_cfg) +{ + int j; + int num_cfgs = 0; + + for (j = 0; j < context->stream_count; j++) { + int k; + + const struct dc_stream_state *stream = context->streams[j]; + struct dm_pp_single_disp_config *cfg = + &pp_display_cfg->disp_configs[num_cfgs]; + const struct pipe_ctx *pipe_ctx = NULL; + + for (k = 0; k < MAX_PIPES; k++) + if (stream == context->res_ctx.pipe_ctx[k].stream) { + pipe_ctx = &context->res_ctx.pipe_ctx[k]; + break; + } + + ASSERT(pipe_ctx != NULL); + + /* only notify active stream */ + if (stream->dpms_off) + continue; + + num_cfgs++; + cfg->signal = pipe_ctx->stream->signal; + cfg->pipe_idx = pipe_ctx->stream_res.tg->inst; + cfg->src_height = stream->src.height; + cfg->src_width = stream->src.width; + cfg->ddi_channel_mapping = + stream->link->ddi_channel_mapping.raw; + cfg->transmitter = + stream->link->link_enc->transmitter; + cfg->link_settings.lane_count = + stream->link->cur_link_settings.lane_count; + cfg->link_settings.link_rate = + stream->link->cur_link_settings.link_rate; + cfg->link_settings.link_spread = + stream->link->cur_link_settings.link_spread; + cfg->sym_clock = stream->phy_pix_clk; + /* Round v_refresh*/ + cfg->v_refresh = stream->timing.pix_clk_100hz * 100; + cfg->v_refresh /= stream->timing.h_total; + cfg->v_refresh = (cfg->v_refresh + stream->timing.v_total / 2) + / stream->timing.v_total; + } + + pp_display_cfg->display_count = num_cfgs; +} + +static uint32_t dce110_get_min_vblank_time_us(const struct dc_state *context) +{ + uint8_t j; + uint32_t min_vertical_blank_time = -1; + + for (j = 0; j < context->stream_count; j++) { + struct dc_stream_state *stream = context->streams[j]; + uint32_t vertical_blank_in_pixels = 0; + uint32_t vertical_blank_time = 0; + + vertical_blank_in_pixels = stream->timing.h_total * + (stream->timing.v_total + - stream->timing.v_addressable); + + vertical_blank_time = vertical_blank_in_pixels + * 10000 / stream->timing.pix_clk_100hz; + + if (min_vertical_blank_time > vertical_blank_time) + min_vertical_blank_time = vertical_blank_time; + } + + return min_vertical_blank_time; +} + +static int determine_sclk_from_bounding_box( + const struct dc *dc, + int required_sclk) +{ + int i; + + /* + * Some asics do not give us sclk levels, so we just report the actual + * required sclk + */ + if (dc->sclk_lvls.num_levels == 0) + return required_sclk; + + for (i = 0; i < dc->sclk_lvls.num_levels; i++) { + if (dc->sclk_lvls.clocks_in_khz[i] >= required_sclk) + return dc->sclk_lvls.clocks_in_khz[i]; + } + /* + * even maximum level could not satisfy requirement, this + * is unexpected at this stage, should have been caught at + * validation time + */ + ASSERT(0); + return dc->sclk_lvls.clocks_in_khz[dc->sclk_lvls.num_levels - 1]; +} + +static void dce_pplib_apply_display_requirements( + struct dc *dc, + struct dc_state *context) +{ + struct dm_pp_display_configuration *pp_display_cfg = &context->pp_display_cfg; + + pp_display_cfg->avail_mclk_switch_time_us = dce110_get_min_vblank_time_us(context); + + dce110_fill_display_configs(context, pp_display_cfg); + + if (memcmp(&dc->current_state->pp_display_cfg, pp_display_cfg, sizeof(*pp_display_cfg)) != 0) + dm_pp_apply_display_requirements(dc->ctx, pp_display_cfg); +} + +static void dce11_pplib_apply_display_requirements( + struct dc *dc, + struct dc_state *context) +{ + struct dm_pp_display_configuration *pp_display_cfg = &context->pp_display_cfg; + + pp_display_cfg->all_displays_in_sync = + context->bw.dce.all_displays_in_sync; + pp_display_cfg->nb_pstate_switch_disable = + context->bw.dce.nbp_state_change_enable == false; + pp_display_cfg->cpu_cc6_disable = + context->bw.dce.cpuc_state_change_enable == false; + pp_display_cfg->cpu_pstate_disable = + context->bw.dce.cpup_state_change_enable == false; + pp_display_cfg->cpu_pstate_separation_time = + context->bw.dce.blackout_recovery_time_us; + + pp_display_cfg->min_memory_clock_khz = context->bw.dce.yclk_khz + / MEMORY_TYPE_MULTIPLIER_CZ; + + pp_display_cfg->min_engine_clock_khz = determine_sclk_from_bounding_box( + dc, + context->bw.dce.sclk_khz); + + /* + * As workaround for >4x4K lightup set dcfclock to min_engine_clock value. + * This is not required for less than 5 displays, + * thus don't request decfclk in dc to avoid impact + * on power saving. + * + */ + pp_display_cfg->min_dcfclock_khz = (context->stream_count > 4)? + pp_display_cfg->min_engine_clock_khz : 0; + + pp_display_cfg->min_engine_clock_deep_sleep_khz + = context->bw.dce.sclk_deep_sleep_khz; + + pp_display_cfg->avail_mclk_switch_time_us = + dce110_get_min_vblank_time_us(context); + /* TODO: dce11.2*/ + pp_display_cfg->avail_mclk_switch_time_in_disp_active_us = 0; + + pp_display_cfg->disp_clk_khz = dc->res_pool->clk_mgr->clks.dispclk_khz; + + dce110_fill_display_configs(context, pp_display_cfg); + + /* TODO: is this still applicable?*/ + if (pp_display_cfg->display_count == 1) { + const struct dc_crtc_timing *timing = + &context->streams[0]->timing; + + pp_display_cfg->crtc_index = + pp_display_cfg->disp_configs[0].pipe_idx; + pp_display_cfg->line_time_in_us = timing->h_total * 10000 / timing->pix_clk_100hz; + } + + if (memcmp(&dc->current_state->pp_display_cfg, pp_display_cfg, sizeof(*pp_display_cfg)) != 0) + dm_pp_apply_display_requirements(dc->ctx, pp_display_cfg); +} + +static void dce_update_clocks(struct clk_mgr *clk_mgr, + struct dc_state *context, + bool safe_to_lower) +{ + struct dce_clk_mgr *clk_mgr_dce = TO_DCE_CLK_MGR(clk_mgr); + struct dm_pp_power_level_change_request level_change_req; + int patched_disp_clk = context->bw.dce.dispclk_khz; + + /*TODO: W/A for dal3 linux, investigate why this works */ + if (!clk_mgr_dce->dfs_bypass_active) + patched_disp_clk = patched_disp_clk * 115 / 100; + + level_change_req.power_level = dce_get_required_clocks_state(clk_mgr, context); + /* get max clock state from PPLIB */ + if ((level_change_req.power_level < clk_mgr_dce->cur_min_clks_state && safe_to_lower) + || level_change_req.power_level > clk_mgr_dce->cur_min_clks_state) { + if (dm_pp_apply_power_level_change_request(clk_mgr->ctx, &level_change_req)) + clk_mgr_dce->cur_min_clks_state = level_change_req.power_level; + } + + if (should_set_clock(safe_to_lower, patched_disp_clk, clk_mgr->clks.dispclk_khz)) { + patched_disp_clk = dce_set_clock(clk_mgr, patched_disp_clk); + clk_mgr->clks.dispclk_khz = patched_disp_clk; + } + dce_pplib_apply_display_requirements(clk_mgr->ctx->dc, context); +} + +static void dce11_update_clocks(struct clk_mgr *clk_mgr, + struct dc_state *context, + bool safe_to_lower) +{ + struct dce_clk_mgr *clk_mgr_dce = TO_DCE_CLK_MGR(clk_mgr); + struct dm_pp_power_level_change_request level_change_req; + int patched_disp_clk = context->bw.dce.dispclk_khz; + + /*TODO: W/A for dal3 linux, investigate why this works */ + if (!clk_mgr_dce->dfs_bypass_active) + patched_disp_clk = patched_disp_clk * 115 / 100; + + level_change_req.power_level = dce_get_required_clocks_state(clk_mgr, context); + /* get max clock state from PPLIB */ + if ((level_change_req.power_level < clk_mgr_dce->cur_min_clks_state && safe_to_lower) + || level_change_req.power_level > clk_mgr_dce->cur_min_clks_state) { + if (dm_pp_apply_power_level_change_request(clk_mgr->ctx, &level_change_req)) + clk_mgr_dce->cur_min_clks_state = level_change_req.power_level; + } + + if (should_set_clock(safe_to_lower, patched_disp_clk, clk_mgr->clks.dispclk_khz)) { + context->bw.dce.dispclk_khz = dce_set_clock(clk_mgr, patched_disp_clk); + clk_mgr->clks.dispclk_khz = patched_disp_clk; + } + dce11_pplib_apply_display_requirements(clk_mgr->ctx->dc, context); +} + +static void dce112_update_clocks(struct clk_mgr *clk_mgr, + struct dc_state *context, + bool safe_to_lower) +{ + struct dce_clk_mgr *clk_mgr_dce = TO_DCE_CLK_MGR(clk_mgr); + struct dm_pp_power_level_change_request level_change_req; + int patched_disp_clk = context->bw.dce.dispclk_khz; + + /*TODO: W/A for dal3 linux, investigate why this works */ + if (!clk_mgr_dce->dfs_bypass_active) + patched_disp_clk = patched_disp_clk * 115 / 100; + + level_change_req.power_level = dce_get_required_clocks_state(clk_mgr, context); + /* get max clock state from PPLIB */ + if ((level_change_req.power_level < clk_mgr_dce->cur_min_clks_state && safe_to_lower) + || level_change_req.power_level > clk_mgr_dce->cur_min_clks_state) { + if (dm_pp_apply_power_level_change_request(clk_mgr->ctx, &level_change_req)) + clk_mgr_dce->cur_min_clks_state = level_change_req.power_level; + } + + if (should_set_clock(safe_to_lower, patched_disp_clk, clk_mgr->clks.dispclk_khz)) { + patched_disp_clk = dce112_set_clock(clk_mgr, patched_disp_clk); + clk_mgr->clks.dispclk_khz = patched_disp_clk; + } + dce11_pplib_apply_display_requirements(clk_mgr->ctx->dc, context); +} + +static void dce12_update_clocks(struct clk_mgr *clk_mgr, + struct dc_state *context, + bool safe_to_lower) +{ + struct dce_clk_mgr *clk_mgr_dce = TO_DCE_CLK_MGR(clk_mgr); + struct dm_pp_clock_for_voltage_req clock_voltage_req = {0}; + int max_pix_clk = get_max_pixel_clock_for_all_paths(context); + int patched_disp_clk = context->bw.dce.dispclk_khz; + + /*TODO: W/A for dal3 linux, investigate why this works */ + if (!clk_mgr_dce->dfs_bypass_active) + patched_disp_clk = patched_disp_clk * 115 / 100; + + if (should_set_clock(safe_to_lower, patched_disp_clk, clk_mgr->clks.dispclk_khz)) { + clock_voltage_req.clk_type = DM_PP_CLOCK_TYPE_DISPLAY_CLK; + /* + * When xGMI is enabled, the display clk needs to be adjusted + * with the WAFL link's SS percentage. + */ + if (clk_mgr_dce->xgmi_enabled) + patched_disp_clk = clk_mgr_adjust_dp_ref_freq_for_ss( + clk_mgr_dce, patched_disp_clk); + clock_voltage_req.clocks_in_khz = patched_disp_clk; + clk_mgr->clks.dispclk_khz = dce112_set_clock(clk_mgr, patched_disp_clk); + + dm_pp_apply_clock_for_voltage_request(clk_mgr->ctx, &clock_voltage_req); + } + + if (should_set_clock(safe_to_lower, max_pix_clk, clk_mgr->clks.phyclk_khz)) { + clock_voltage_req.clk_type = DM_PP_CLOCK_TYPE_DISPLAYPHYCLK; + clock_voltage_req.clocks_in_khz = max_pix_clk; + clk_mgr->clks.phyclk_khz = max_pix_clk; + + dm_pp_apply_clock_for_voltage_request(clk_mgr->ctx, &clock_voltage_req); + } + dce11_pplib_apply_display_requirements(clk_mgr->ctx->dc, context); +} + +static const struct clk_mgr_funcs dce120_funcs = { + .get_dp_ref_clk_frequency = dce12_get_dp_ref_freq_khz, + .update_clocks = dce12_update_clocks +}; + +static const struct clk_mgr_funcs dce112_funcs = { + .get_dp_ref_clk_frequency = dce_get_dp_ref_freq_khz, + .update_clocks = dce112_update_clocks +}; + +static const struct clk_mgr_funcs dce110_funcs = { + .get_dp_ref_clk_frequency = dce_get_dp_ref_freq_khz, + .update_clocks = dce11_update_clocks, +}; + +static const struct clk_mgr_funcs dce_funcs = { + .get_dp_ref_clk_frequency = dce_get_dp_ref_freq_khz, + .update_clocks = dce_update_clocks +}; + +static void dce_clk_mgr_construct( + struct dce_clk_mgr *clk_mgr_dce, + struct dc_context *ctx, + const struct clk_mgr_registers *regs, + const struct clk_mgr_shift *clk_shift, + const struct clk_mgr_mask *clk_mask) +{ + struct clk_mgr *base = &clk_mgr_dce->base; + struct dm_pp_static_clock_info static_clk_info = {0}; + + base->ctx = ctx; + base->funcs = &dce_funcs; + + clk_mgr_dce->regs = regs; + clk_mgr_dce->clk_mgr_shift = clk_shift; + clk_mgr_dce->clk_mgr_mask = clk_mask; + + clk_mgr_dce->dfs_bypass_disp_clk = 0; + + clk_mgr_dce->dprefclk_ss_percentage = 0; + clk_mgr_dce->dprefclk_ss_divider = 1000; + clk_mgr_dce->ss_on_dprefclk = false; + + + if (dm_pp_get_static_clocks(ctx, &static_clk_info)) + clk_mgr_dce->max_clks_state = static_clk_info.max_clocks_state; + else + clk_mgr_dce->max_clks_state = DM_PP_CLOCKS_STATE_NOMINAL; + clk_mgr_dce->cur_min_clks_state = DM_PP_CLOCKS_STATE_INVALID; + + dce_clock_read_integrated_info(clk_mgr_dce); + dce_clock_read_ss_info(clk_mgr_dce); +} + +struct clk_mgr *dce_clk_mgr_create( + struct dc_context *ctx, + const struct clk_mgr_registers *regs, + const struct clk_mgr_shift *clk_shift, + const struct clk_mgr_mask *clk_mask) +{ + struct dce_clk_mgr *clk_mgr_dce = kzalloc(sizeof(*clk_mgr_dce), GFP_KERNEL); + + if (clk_mgr_dce == NULL) { + BREAK_TO_DEBUGGER(); + return NULL; + } + + memcpy(clk_mgr_dce->max_clks_by_state, + dce80_max_clks_by_state, + sizeof(dce80_max_clks_by_state)); + + dce_clk_mgr_construct( + clk_mgr_dce, ctx, regs, clk_shift, clk_mask); + + return &clk_mgr_dce->base; +} + +struct clk_mgr *dce110_clk_mgr_create( + struct dc_context *ctx, + const struct clk_mgr_registers *regs, + const struct clk_mgr_shift *clk_shift, + const struct clk_mgr_mask *clk_mask) +{ + struct dce_clk_mgr *clk_mgr_dce = kzalloc(sizeof(*clk_mgr_dce), GFP_KERNEL); + + if (clk_mgr_dce == NULL) { + BREAK_TO_DEBUGGER(); + return NULL; + } + + memcpy(clk_mgr_dce->max_clks_by_state, + dce110_max_clks_by_state, + sizeof(dce110_max_clks_by_state)); + + dce_clk_mgr_construct( + clk_mgr_dce, ctx, regs, clk_shift, clk_mask); + + clk_mgr_dce->base.funcs = &dce110_funcs; + + return &clk_mgr_dce->base; +} + +struct clk_mgr *dce112_clk_mgr_create( + struct dc_context *ctx, + const struct clk_mgr_registers *regs, + const struct clk_mgr_shift *clk_shift, + const struct clk_mgr_mask *clk_mask) +{ + struct dce_clk_mgr *clk_mgr_dce = kzalloc(sizeof(*clk_mgr_dce), GFP_KERNEL); + + if (clk_mgr_dce == NULL) { + BREAK_TO_DEBUGGER(); + return NULL; + } + + memcpy(clk_mgr_dce->max_clks_by_state, + dce112_max_clks_by_state, + sizeof(dce112_max_clks_by_state)); + + dce_clk_mgr_construct( + clk_mgr_dce, ctx, regs, clk_shift, clk_mask); + + clk_mgr_dce->base.funcs = &dce112_funcs; + + return &clk_mgr_dce->base; +} + +struct clk_mgr *dce120_clk_mgr_create(struct dc_context *ctx) +{ + struct dce_clk_mgr *clk_mgr_dce = kzalloc(sizeof(*clk_mgr_dce), GFP_KERNEL); + + if (clk_mgr_dce == NULL) { + BREAK_TO_DEBUGGER(); + return NULL; + } + + memcpy(clk_mgr_dce->max_clks_by_state, + dce120_max_clks_by_state, + sizeof(dce120_max_clks_by_state)); + + dce_clk_mgr_construct( + clk_mgr_dce, ctx, NULL, NULL, NULL); + + clk_mgr_dce->dprefclk_khz = 600000; + clk_mgr_dce->base.funcs = &dce120_funcs; + + return &clk_mgr_dce->base; +} + +struct clk_mgr *dce121_clk_mgr_create(struct dc_context *ctx) +{ + struct dce_clk_mgr *clk_mgr_dce = kzalloc(sizeof(*clk_mgr_dce), + GFP_KERNEL); + + if (clk_mgr_dce == NULL) { + BREAK_TO_DEBUGGER(); + return NULL; + } + + memcpy(clk_mgr_dce->max_clks_by_state, dce120_max_clks_by_state, + sizeof(dce120_max_clks_by_state)); + + dce_clk_mgr_construct(clk_mgr_dce, ctx, NULL, NULL, NULL); + + clk_mgr_dce->dprefclk_khz = 625000; + clk_mgr_dce->base.funcs = &dce120_funcs; + + return &clk_mgr_dce->base; +} + +void dce_clk_mgr_destroy(struct clk_mgr **clk_mgr) +{ + struct dce_clk_mgr *clk_mgr_dce = TO_DCE_CLK_MGR(*clk_mgr); + + kfree(clk_mgr_dce); + *clk_mgr = NULL; +} diff --git a/drivers/gpu/drm/amd/display/dc/dce/dce_clk_mgr.h b/drivers/gpu/drm/amd/display/dc/dce/dce_clk_mgr.h new file mode 100644 index 000000000000..c8f8c442142a --- /dev/null +++ b/drivers/gpu/drm/amd/display/dc/dce/dce_clk_mgr.h @@ -0,0 +1,199 @@ +/* + * Copyright 2012-16 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. + * + * Authors: AMD + * + */ + + +#ifndef _DCE_CLK_MGR_H_ +#define _DCE_CLK_MGR_H_ + +#include "clk_mgr.h" +#include "dccg.h" + +#define MEMORY_TYPE_MULTIPLIER_CZ 4 + +#define CLK_COMMON_REG_LIST_DCE_BASE() \ + .DPREFCLK_CNTL = mmDPREFCLK_CNTL, \ + .DENTIST_DISPCLK_CNTL = mmDENTIST_DISPCLK_CNTL + +#define CLK_COMMON_REG_LIST_DCN_BASE() \ + SR(DENTIST_DISPCLK_CNTL) + +#define CLK_SF(reg_name, field_name, post_fix)\ + .field_name = reg_name ## __ ## field_name ## post_fix + +#define CLK_COMMON_MASK_SH_LIST_DCE_COMMON_BASE(mask_sh) \ + CLK_SF(DPREFCLK_CNTL, DPREFCLK_SRC_SEL, mask_sh), \ + CLK_SF(DENTIST_DISPCLK_CNTL, DENTIST_DPREFCLK_WDIVIDER, mask_sh) + +#define CLK_COMMON_MASK_SH_LIST_DCN_COMMON_BASE(mask_sh) \ + CLK_SF(DENTIST_DISPCLK_CNTL, DENTIST_DISPCLK_WDIVIDER, mask_sh),\ + CLK_SF(DENTIST_DISPCLK_CNTL, DENTIST_DISPCLK_CHG_DONE, mask_sh) + +#define CLK_REG_FIELD_LIST(type) \ + type DPREFCLK_SRC_SEL; \ + type DENTIST_DPREFCLK_WDIVIDER; \ + type DENTIST_DISPCLK_WDIVIDER; \ + type DENTIST_DISPCLK_CHG_DONE; + +struct clk_mgr_shift { + CLK_REG_FIELD_LIST(uint8_t) +}; + +struct clk_mgr_mask { + CLK_REG_FIELD_LIST(uint32_t) +}; + +struct clk_mgr_registers { + uint32_t DPREFCLK_CNTL; + uint32_t DENTIST_DISPCLK_CNTL; +}; + +struct state_dependent_clocks { + int display_clk_khz; + int pixel_clk_khz; +}; + +struct dce_clk_mgr { + struct clk_mgr base; + const struct clk_mgr_registers *regs; + const struct clk_mgr_shift *clk_mgr_shift; + const struct clk_mgr_mask *clk_mgr_mask; + + struct dccg *dccg; + + struct state_dependent_clocks max_clks_by_state[DM_PP_CLOCKS_MAX_STATES]; + + int dentist_vco_freq_khz; + + /* Cache the status of DFS-bypass feature*/ + bool dfs_bypass_enabled; + /* True if the DFS-bypass feature is enabled and active. */ + bool dfs_bypass_active; + /* Cache the display clock returned by VBIOS if DFS-bypass is enabled. + * This is basically "Crystal Frequency In KHz" (XTALIN) frequency */ + int dfs_bypass_disp_clk; + + /** + * @ss_on_dprefclk: + * + * True if spread spectrum is enabled on the DP ref clock. + */ + bool ss_on_dprefclk; + + /** + * @xgmi_enabled: + * + * True if xGMI is enabled. On VG20, both audio and display clocks need + * to be adjusted with the WAFL link's SS info if xGMI is enabled. + */ + bool xgmi_enabled; + + /** + * @dprefclk_ss_percentage: + * + * DPREFCLK SS percentage (if down-spread enabled). + * + * Note that if XGMI is enabled, the SS info (percentage and divider) + * from the WAFL link is used instead. This is decided during + * dce_clk_mgr initialization. + */ + int dprefclk_ss_percentage; + + /** + * @dprefclk_ss_divider: + * + * DPREFCLK SS percentage Divider (100 or 1000). + */ + int dprefclk_ss_divider; + int dprefclk_khz; + + enum dm_pp_clocks_state max_clks_state; + enum dm_pp_clocks_state cur_min_clks_state; +}; + +/* Starting DID for each range */ +enum dentist_base_divider_id { + DENTIST_BASE_DID_1 = 0x08, + DENTIST_BASE_DID_2 = 0x40, + DENTIST_BASE_DID_3 = 0x60, + DENTIST_BASE_DID_4 = 0x7e, + DENTIST_MAX_DID = 0x7f +}; + +/* Starting point and step size for each divider range.*/ +enum dentist_divider_range { + DENTIST_DIVIDER_RANGE_1_START = 8, /* 2.00 */ + DENTIST_DIVIDER_RANGE_1_STEP = 1, /* 0.25 */ + DENTIST_DIVIDER_RANGE_2_START = 64, /* 16.00 */ + DENTIST_DIVIDER_RANGE_2_STEP = 2, /* 0.50 */ + DENTIST_DIVIDER_RANGE_3_START = 128, /* 32.00 */ + DENTIST_DIVIDER_RANGE_3_STEP = 4, /* 1.00 */ + DENTIST_DIVIDER_RANGE_4_START = 248, /* 62.00 */ + DENTIST_DIVIDER_RANGE_4_STEP = 264, /* 66.00 */ + DENTIST_DIVIDER_RANGE_SCALE_FACTOR = 4 +}; + +static inline bool should_set_clock(bool safe_to_lower, int calc_clk, int cur_clk) +{ + return ((safe_to_lower && calc_clk < cur_clk) || calc_clk > cur_clk); +} + +void dce_clock_read_ss_info(struct dce_clk_mgr *dccg_dce); + +int dce12_get_dp_ref_freq_khz(struct clk_mgr *dccg); + +void dce110_fill_display_configs( + const struct dc_state *context, + struct dm_pp_display_configuration *pp_display_cfg); + +int dce112_set_clock(struct clk_mgr *dccg, int requested_clk_khz); + +struct clk_mgr *dce_clk_mgr_create( + struct dc_context *ctx, + const struct clk_mgr_registers *regs, + const struct clk_mgr_shift *clk_shift, + const struct clk_mgr_mask *clk_mask); + +struct clk_mgr *dce110_clk_mgr_create( + struct dc_context *ctx, + const struct clk_mgr_registers *regs, + const struct clk_mgr_shift *clk_shift, + const struct clk_mgr_mask *clk_mask); + +struct clk_mgr *dce112_clk_mgr_create( + struct dc_context *ctx, + const struct clk_mgr_registers *regs, + const struct clk_mgr_shift *clk_shift, + const struct clk_mgr_mask *clk_mask); + +struct clk_mgr *dce120_clk_mgr_create(struct dc_context *ctx); + +struct clk_mgr *dce121_clk_mgr_create(struct dc_context *ctx); +void dce121_clock_patch_xgmi_ss_info(struct clk_mgr *clk_mgr); + +void dce_clk_mgr_destroy(struct clk_mgr **clk_mgr); + +int dentist_get_divider_from_did(int did); + +#endif /* _DCE_CLK_MGR_H_ */ diff --git a/drivers/gpu/drm/amd/display/dc/dce/dce_clock_source.c b/drivers/gpu/drm/amd/display/dc/dce/dce_clock_source.c index 723ce80ed89c..71d5777de961 100644 --- a/drivers/gpu/drm/amd/display/dc/dce/dce_clock_source.c +++ b/drivers/gpu/drm/amd/display/dc/dce/dce_clock_source.c @@ -108,28 +108,28 @@ static const struct spread_spectrum_data *get_ss_data_entry( } /** -* Function: calculate_fb_and_fractional_fb_divider -* -* * DESCRIPTION: Calculates feedback and fractional feedback dividers values -* -*PARAMETERS: -* targetPixelClock Desired frequency in 10 KHz -* ref_divider Reference divider (already known) -* postDivider Post Divider (already known) -* feedback_divider_param Pointer where to store -* calculated feedback divider value -* fract_feedback_divider_param Pointer where to store -* calculated fract feedback divider value -* -*RETURNS: -* It fills the locations pointed by feedback_divider_param -* and fract_feedback_divider_param -* It returns - true if feedback divider not 0 -* - false should never happen) -*/ + * Function: calculate_fb_and_fractional_fb_divider + * + * * DESCRIPTION: Calculates feedback and fractional feedback dividers values + * + *PARAMETERS: + * targetPixelClock Desired frequency in 100 Hz + * ref_divider Reference divider (already known) + * postDivider Post Divider (already known) + * feedback_divider_param Pointer where to store + * calculated feedback divider value + * fract_feedback_divider_param Pointer where to store + * calculated fract feedback divider value + * + *RETURNS: + * It fills the locations pointed by feedback_divider_param + * and fract_feedback_divider_param + * It returns - true if feedback divider not 0 + * - false should never happen) + */ static bool calculate_fb_and_fractional_fb_divider( struct calc_pll_clock_source *calc_pll_cs, - uint32_t target_pix_clk_khz, + uint32_t target_pix_clk_100hz, uint32_t ref_divider, uint32_t post_divider, uint32_t *feedback_divider_param, @@ -138,11 +138,11 @@ static bool calculate_fb_and_fractional_fb_divider( uint64_t feedback_divider; feedback_divider = - (uint64_t)target_pix_clk_khz * ref_divider * post_divider; + (uint64_t)target_pix_clk_100hz * ref_divider * post_divider; feedback_divider *= 10; /* additional factor, since we divide by 10 afterwards */ feedback_divider *= (uint64_t)(calc_pll_cs->fract_fb_divider_factor); - feedback_divider = div_u64(feedback_divider, calc_pll_cs->ref_freq_khz); + feedback_divider = div_u64(feedback_divider, calc_pll_cs->ref_freq_khz * 10ull); /*Round to the number of precision * The following code replace the old code (ullfeedbackDivider + 5)/10 @@ -195,36 +195,36 @@ static bool calc_fb_divider_checking_tolerance( { uint32_t feedback_divider; uint32_t fract_feedback_divider; - uint32_t actual_calculated_clock_khz; + uint32_t actual_calculated_clock_100hz; uint32_t abs_err; - uint64_t actual_calc_clk_khz; + uint64_t actual_calc_clk_100hz; calculate_fb_and_fractional_fb_divider( calc_pll_cs, - pll_settings->adjusted_pix_clk, + pll_settings->adjusted_pix_clk_100hz, ref_divider, post_divider, &feedback_divider, &fract_feedback_divider); /*Actual calculated value*/ - actual_calc_clk_khz = (uint64_t)feedback_divider * + actual_calc_clk_100hz = (uint64_t)feedback_divider * calc_pll_cs->fract_fb_divider_factor + fract_feedback_divider; - actual_calc_clk_khz *= calc_pll_cs->ref_freq_khz; - actual_calc_clk_khz = - div_u64(actual_calc_clk_khz, + actual_calc_clk_100hz *= calc_pll_cs->ref_freq_khz * 10; + actual_calc_clk_100hz = + div_u64(actual_calc_clk_100hz, ref_divider * post_divider * calc_pll_cs->fract_fb_divider_factor); - actual_calculated_clock_khz = (uint32_t)(actual_calc_clk_khz); + actual_calculated_clock_100hz = (uint32_t)(actual_calc_clk_100hz); - abs_err = (actual_calculated_clock_khz > - pll_settings->adjusted_pix_clk) - ? actual_calculated_clock_khz - - pll_settings->adjusted_pix_clk - : pll_settings->adjusted_pix_clk - - actual_calculated_clock_khz; + abs_err = (actual_calculated_clock_100hz > + pll_settings->adjusted_pix_clk_100hz) + ? actual_calculated_clock_100hz - + pll_settings->adjusted_pix_clk_100hz + : pll_settings->adjusted_pix_clk_100hz - + actual_calculated_clock_100hz; if (abs_err <= tolerance) { /*found good values*/ @@ -233,10 +233,10 @@ static bool calc_fb_divider_checking_tolerance( pll_settings->feedback_divider = feedback_divider; pll_settings->fract_feedback_divider = fract_feedback_divider; pll_settings->pix_clk_post_divider = post_divider; - pll_settings->calculated_pix_clk = - actual_calculated_clock_khz; + pll_settings->calculated_pix_clk_100hz = + actual_calculated_clock_100hz; pll_settings->vco_freq = - actual_calculated_clock_khz * post_divider; + actual_calculated_clock_100hz * post_divider / 10; return true; } return false; @@ -257,8 +257,8 @@ static bool calc_pll_dividers_in_range( /* This is err_tolerance / 10000 = 0.0025 - acceptable error of 0.25% * This is errorTolerance / 10000 = 0.0001 - acceptable error of 0.01%*/ - tolerance = (pll_settings->adjusted_pix_clk * err_tolerance) / - 10000; + tolerance = (pll_settings->adjusted_pix_clk_100hz * err_tolerance) / + 100000; if (tolerance < CALC_PLL_CLK_SRC_ERR_TOLERANCE) tolerance = CALC_PLL_CLK_SRC_ERR_TOLERANCE; @@ -294,7 +294,7 @@ static uint32_t calculate_pixel_clock_pll_dividers( uint32_t min_ref_divider; uint32_t max_ref_divider; - if (pll_settings->adjusted_pix_clk == 0) { + if (pll_settings->adjusted_pix_clk_100hz == 0) { DC_LOG_ERROR( "%s Bad requested pixel clock", __func__); return MAX_PLL_CALC_ERROR; @@ -306,21 +306,21 @@ static uint32_t calculate_pixel_clock_pll_dividers( max_post_divider = pll_settings->pix_clk_post_divider; } else { min_post_divider = calc_pll_cs->min_pix_clock_pll_post_divider; - if (min_post_divider * pll_settings->adjusted_pix_clk < - calc_pll_cs->min_vco_khz) { - min_post_divider = calc_pll_cs->min_vco_khz / - pll_settings->adjusted_pix_clk; + if (min_post_divider * pll_settings->adjusted_pix_clk_100hz < + calc_pll_cs->min_vco_khz * 10) { + min_post_divider = calc_pll_cs->min_vco_khz * 10 / + pll_settings->adjusted_pix_clk_100hz; if ((min_post_divider * - pll_settings->adjusted_pix_clk) < - calc_pll_cs->min_vco_khz) + pll_settings->adjusted_pix_clk_100hz) < + calc_pll_cs->min_vco_khz * 10) min_post_divider++; } max_post_divider = calc_pll_cs->max_pix_clock_pll_post_divider; - if (max_post_divider * pll_settings->adjusted_pix_clk - > calc_pll_cs->max_vco_khz) - max_post_divider = calc_pll_cs->max_vco_khz / - pll_settings->adjusted_pix_clk; + if (max_post_divider * pll_settings->adjusted_pix_clk_100hz + > calc_pll_cs->max_vco_khz * 10) + max_post_divider = calc_pll_cs->max_vco_khz * 10 / + pll_settings->adjusted_pix_clk_100hz; } /* 2) Find Reference divider ranges @@ -392,47 +392,47 @@ static bool pll_adjust_pix_clk( struct pixel_clk_params *pix_clk_params, struct pll_settings *pll_settings) { - uint32_t actual_pix_clk_khz = 0; - uint32_t requested_clk_khz = 0; + uint32_t actual_pix_clk_100hz = 0; + uint32_t requested_clk_100hz = 0; struct bp_adjust_pixel_clock_parameters bp_adjust_pixel_clock_params = { 0 }; enum bp_result bp_result; switch (pix_clk_params->signal_type) { case SIGNAL_TYPE_HDMI_TYPE_A: { - requested_clk_khz = pix_clk_params->requested_pix_clk; + requested_clk_100hz = pix_clk_params->requested_pix_clk_100hz; if (pix_clk_params->pixel_encoding != PIXEL_ENCODING_YCBCR422) { switch (pix_clk_params->color_depth) { case COLOR_DEPTH_101010: - requested_clk_khz = (requested_clk_khz * 5) >> 2; + requested_clk_100hz = (requested_clk_100hz * 5) >> 2; break; /* x1.25*/ case COLOR_DEPTH_121212: - requested_clk_khz = (requested_clk_khz * 6) >> 2; + requested_clk_100hz = (requested_clk_100hz * 6) >> 2; break; /* x1.5*/ case COLOR_DEPTH_161616: - requested_clk_khz = requested_clk_khz * 2; + requested_clk_100hz = requested_clk_100hz * 2; break; /* x2.0*/ default: break; } } - actual_pix_clk_khz = requested_clk_khz; + actual_pix_clk_100hz = requested_clk_100hz; } break; case SIGNAL_TYPE_DISPLAY_PORT: case SIGNAL_TYPE_DISPLAY_PORT_MST: case SIGNAL_TYPE_EDP: - requested_clk_khz = pix_clk_params->requested_sym_clk; - actual_pix_clk_khz = pix_clk_params->requested_pix_clk; + requested_clk_100hz = pix_clk_params->requested_sym_clk * 10; + actual_pix_clk_100hz = pix_clk_params->requested_pix_clk_100hz; break; default: - requested_clk_khz = pix_clk_params->requested_pix_clk; - actual_pix_clk_khz = pix_clk_params->requested_pix_clk; + requested_clk_100hz = pix_clk_params->requested_pix_clk_100hz; + actual_pix_clk_100hz = pix_clk_params->requested_pix_clk_100hz; break; } - bp_adjust_pixel_clock_params.pixel_clock = requested_clk_khz; + bp_adjust_pixel_clock_params.pixel_clock = requested_clk_100hz / 10; bp_adjust_pixel_clock_params. encoder_object_id = pix_clk_params->encoder_object_id; bp_adjust_pixel_clock_params.signal_type = pix_clk_params->signal_type; @@ -441,9 +441,9 @@ static bool pll_adjust_pix_clk( bp_result = clk_src->bios->funcs->adjust_pixel_clock( clk_src->bios, &bp_adjust_pixel_clock_params); if (bp_result == BP_RESULT_OK) { - pll_settings->actual_pix_clk = actual_pix_clk_khz; - pll_settings->adjusted_pix_clk = - bp_adjust_pixel_clock_params.adjusted_pixel_clock; + pll_settings->actual_pix_clk_100hz = actual_pix_clk_100hz; + pll_settings->adjusted_pix_clk_100hz = + bp_adjust_pixel_clock_params.adjusted_pixel_clock * 10; pll_settings->reference_divider = bp_adjust_pixel_clock_params.reference_divider; pll_settings->pix_clk_post_divider = @@ -490,7 +490,7 @@ static uint32_t dce110_get_pix_clk_dividers_helper ( const struct spread_spectrum_data *ss_data = get_ss_data_entry( clk_src, pix_clk_params->signal_type, - pll_settings->adjusted_pix_clk); + pll_settings->adjusted_pix_clk_100hz / 10); if (NULL != ss_data) pll_settings->ss_percentage = ss_data->percentage; @@ -502,13 +502,13 @@ static uint32_t dce110_get_pix_clk_dividers_helper ( * to continue. */ DC_LOG_ERROR( "%s: Failed to adjust pixel clock!!", __func__); - pll_settings->actual_pix_clk = - pix_clk_params->requested_pix_clk; - pll_settings->adjusted_pix_clk = - pix_clk_params->requested_pix_clk; + pll_settings->actual_pix_clk_100hz = + pix_clk_params->requested_pix_clk_100hz; + pll_settings->adjusted_pix_clk_100hz = + pix_clk_params->requested_pix_clk_100hz; if (dc_is_dp_signal(pix_clk_params->signal_type)) - pll_settings->adjusted_pix_clk = 100000; + pll_settings->adjusted_pix_clk_100hz = 1000000; } /* Calculate Dividers */ @@ -533,28 +533,28 @@ static void dce112_get_pix_clk_dividers_helper ( struct pll_settings *pll_settings, struct pixel_clk_params *pix_clk_params) { - uint32_t actualPixelClockInKHz; + uint32_t actual_pixel_clock_100hz; - actualPixelClockInKHz = pix_clk_params->requested_pix_clk; + actual_pixel_clock_100hz = pix_clk_params->requested_pix_clk_100hz; /* Calculate Dividers */ if (pix_clk_params->signal_type == SIGNAL_TYPE_HDMI_TYPE_A) { switch (pix_clk_params->color_depth) { case COLOR_DEPTH_101010: - actualPixelClockInKHz = (actualPixelClockInKHz * 5) >> 2; + actual_pixel_clock_100hz = (actual_pixel_clock_100hz * 5) >> 2; break; case COLOR_DEPTH_121212: - actualPixelClockInKHz = (actualPixelClockInKHz * 6) >> 2; + actual_pixel_clock_100hz = (actual_pixel_clock_100hz * 6) >> 2; break; case COLOR_DEPTH_161616: - actualPixelClockInKHz = actualPixelClockInKHz * 2; + actual_pixel_clock_100hz = actual_pixel_clock_100hz * 2; break; default: break; } } - pll_settings->actual_pix_clk = actualPixelClockInKHz; - pll_settings->adjusted_pix_clk = actualPixelClockInKHz; - pll_settings->calculated_pix_clk = pix_clk_params->requested_pix_clk; + pll_settings->actual_pix_clk_100hz = actual_pixel_clock_100hz; + pll_settings->adjusted_pix_clk_100hz = actual_pixel_clock_100hz; + pll_settings->calculated_pix_clk_100hz = pix_clk_params->requested_pix_clk_100hz; } static uint32_t dce110_get_pix_clk_dividers( @@ -567,7 +567,7 @@ static uint32_t dce110_get_pix_clk_dividers( DC_LOGGER_INIT(); if (pix_clk_params == NULL || pll_settings == NULL - || pix_clk_params->requested_pix_clk == 0) { + || pix_clk_params->requested_pix_clk_100hz == 0) { DC_LOG_ERROR( "%s: Invalid parameters!!\n", __func__); return pll_calc_error; @@ -577,10 +577,10 @@ static uint32_t dce110_get_pix_clk_dividers( if (cs->id == CLOCK_SOURCE_ID_DP_DTO || cs->id == CLOCK_SOURCE_ID_EXTERNAL) { - pll_settings->adjusted_pix_clk = clk_src->ext_clk_khz; - pll_settings->calculated_pix_clk = clk_src->ext_clk_khz; - pll_settings->actual_pix_clk = - pix_clk_params->requested_pix_clk; + pll_settings->adjusted_pix_clk_100hz = clk_src->ext_clk_khz * 10; + pll_settings->calculated_pix_clk_100hz = clk_src->ext_clk_khz * 10; + pll_settings->actual_pix_clk_100hz = + pix_clk_params->requested_pix_clk_100hz; return 0; } @@ -599,7 +599,7 @@ static uint32_t dce112_get_pix_clk_dividers( DC_LOGGER_INIT(); if (pix_clk_params == NULL || pll_settings == NULL - || pix_clk_params->requested_pix_clk == 0) { + || pix_clk_params->requested_pix_clk_100hz == 0) { DC_LOG_ERROR( "%s: Invalid parameters!!\n", __func__); return -1; @@ -609,10 +609,10 @@ static uint32_t dce112_get_pix_clk_dividers( if (cs->id == CLOCK_SOURCE_ID_DP_DTO || cs->id == CLOCK_SOURCE_ID_EXTERNAL) { - pll_settings->adjusted_pix_clk = clk_src->ext_clk_khz; - pll_settings->calculated_pix_clk = clk_src->ext_clk_khz; - pll_settings->actual_pix_clk = - pix_clk_params->requested_pix_clk; + pll_settings->adjusted_pix_clk_100hz = clk_src->ext_clk_khz * 10; + pll_settings->calculated_pix_clk_100hz = clk_src->ext_clk_khz * 10; + pll_settings->actual_pix_clk_100hz = + pix_clk_params->requested_pix_clk_100hz; return -1; } @@ -714,7 +714,7 @@ static bool enable_spread_spectrum( ss_data = get_ss_data_entry( clk_src, signal, - pll_settings->calculated_pix_clk); + pll_settings->calculated_pix_clk_100hz / 10); /* Pixel clock PLL has been programmed to generate desired pixel clock, * now enable SS on pixel clock */ @@ -853,7 +853,7 @@ static bool dce110_program_pix_clk( /*ATOMBIOS expects pixel rate adjusted by deep color ratio)*/ bp_pc_params.controller_id = pix_clk_params->controller_id; bp_pc_params.pll_id = clock_source->id; - bp_pc_params.target_pixel_clock = pll_settings->actual_pix_clk; + bp_pc_params.target_pixel_clock_100hz = pll_settings->actual_pix_clk_100hz; bp_pc_params.encoder_object_id = pix_clk_params->encoder_object_id; bp_pc_params.signal_type = pix_clk_params->signal_type; @@ -903,12 +903,12 @@ static bool dce112_program_pix_clk( #if defined(CONFIG_DRM_AMD_DC_DCN1_0) if (IS_FPGA_MAXIMUS_DC(clock_source->ctx->dce_environment)) { unsigned int inst = pix_clk_params->controller_id - CONTROLLER_ID_D0; - unsigned dp_dto_ref_kHz = 700000; - unsigned clock_kHz = pll_settings->actual_pix_clk; + unsigned dp_dto_ref_100hz = 7000000; + unsigned clock_100hz = pll_settings->actual_pix_clk_100hz; /* Set DTO values: phase = target clock, modulo = reference clock */ - REG_WRITE(PHASE[inst], clock_kHz); - REG_WRITE(MODULO[inst], dp_dto_ref_kHz); + REG_WRITE(PHASE[inst], clock_100hz); + REG_WRITE(MODULO[inst], dp_dto_ref_100hz); /* Enable DTO */ REG_UPDATE(PIXEL_RATE_CNTL[inst], DP_DTO0_ENABLE, 1); @@ -927,7 +927,7 @@ static bool dce112_program_pix_clk( /*ATOMBIOS expects pixel rate adjusted by deep color ratio)*/ bp_pc_params.controller_id = pix_clk_params->controller_id; bp_pc_params.pll_id = clock_source->id; - bp_pc_params.target_pixel_clock = pll_settings->actual_pix_clk; + bp_pc_params.target_pixel_clock_100hz = pll_settings->actual_pix_clk_100hz; bp_pc_params.encoder_object_id = pix_clk_params->encoder_object_id; bp_pc_params.signal_type = pix_clk_params->signal_type; @@ -977,6 +977,28 @@ static bool dce110_clock_source_power_down( return bp_result == BP_RESULT_OK; } +static bool get_pixel_clk_frequency_100hz( + struct clock_source *clock_source, + unsigned int inst, + unsigned int *pixel_clk_khz) +{ + struct dce110_clk_src *clk_src = TO_DCE110_CLK_SRC(clock_source); + unsigned int clock_hz = 0; + + if (clock_source->id == CLOCK_SOURCE_ID_DP_DTO) { + clock_hz = REG_READ(PHASE[inst]); + + /* NOTE: There is agreement with VBIOS here that MODULO is + * programmed equal to DPREFCLK, in which case PHASE will be + * equivalent to pixel clock. + */ + *pixel_clk_khz = clock_hz / 100; + return true; + } + + return false; +} + /*****************************************/ /* Constructor */ /*****************************************/ @@ -984,12 +1006,14 @@ static bool dce110_clock_source_power_down( static const struct clock_source_funcs dce112_clk_src_funcs = { .cs_power_down = dce110_clock_source_power_down, .program_pix_clk = dce112_program_pix_clk, - .get_pix_clk_dividers = dce112_get_pix_clk_dividers + .get_pix_clk_dividers = dce112_get_pix_clk_dividers, + .get_pixel_clk_frequency_100hz = get_pixel_clk_frequency_100hz }; static const struct clock_source_funcs dce110_clk_src_funcs = { .cs_power_down = dce110_clock_source_power_down, .program_pix_clk = dce110_program_pix_clk, - .get_pix_clk_dividers = dce110_get_pix_clk_dividers + .get_pix_clk_dividers = dce110_get_pix_clk_dividers, + .get_pixel_clk_frequency_100hz = get_pixel_clk_frequency_100hz }; diff --git a/drivers/gpu/drm/amd/display/dc/dce/dce_clocks.c b/drivers/gpu/drm/amd/display/dc/dce/dce_clocks.c deleted file mode 100644 index d89a097ba936..000000000000 --- a/drivers/gpu/drm/amd/display/dc/dce/dce_clocks.c +++ /dev/null @@ -1,947 +0,0 @@ -/* - * Copyright 2012-16 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. - * - * Authors: AMD - * - */ - -#include "dce_clocks.h" -#include "dm_services.h" -#include "reg_helper.h" -#include "fixed31_32.h" -#include "bios_parser_interface.h" -#include "dc.h" -#include "dmcu.h" -#if defined(CONFIG_DRM_AMD_DC_DCN1_0) -#include "dcn_calcs.h" -#endif -#include "core_types.h" -#include "dc_types.h" -#include "dal_asic_id.h" - -#define TO_DCE_CLOCKS(clocks)\ - container_of(clocks, struct dce_dccg, base) - -#define REG(reg) \ - (clk_dce->regs->reg) - -#undef FN -#define FN(reg_name, field_name) \ - clk_dce->clk_shift->field_name, clk_dce->clk_mask->field_name - -#define CTX \ - clk_dce->base.ctx -#define DC_LOGGER \ - clk->ctx->logger - -/* Max clock values for each state indexed by "enum clocks_state": */ -static const struct state_dependent_clocks dce80_max_clks_by_state[] = { -/* ClocksStateInvalid - should not be used */ -{ .display_clk_khz = 0, .pixel_clk_khz = 0 }, -/* ClocksStateUltraLow - not expected to be used for DCE 8.0 */ -{ .display_clk_khz = 0, .pixel_clk_khz = 0 }, -/* ClocksStateLow */ -{ .display_clk_khz = 352000, .pixel_clk_khz = 330000}, -/* ClocksStateNominal */ -{ .display_clk_khz = 600000, .pixel_clk_khz = 400000 }, -/* ClocksStatePerformance */ -{ .display_clk_khz = 600000, .pixel_clk_khz = 400000 } }; - -static const struct state_dependent_clocks dce110_max_clks_by_state[] = { -/*ClocksStateInvalid - should not be used*/ -{ .display_clk_khz = 0, .pixel_clk_khz = 0 }, -/*ClocksStateUltraLow - currently by HW design team not supposed to be used*/ -{ .display_clk_khz = 352000, .pixel_clk_khz = 330000 }, -/*ClocksStateLow*/ -{ .display_clk_khz = 352000, .pixel_clk_khz = 330000 }, -/*ClocksStateNominal*/ -{ .display_clk_khz = 467000, .pixel_clk_khz = 400000 }, -/*ClocksStatePerformance*/ -{ .display_clk_khz = 643000, .pixel_clk_khz = 400000 } }; - -static const struct state_dependent_clocks dce112_max_clks_by_state[] = { -/*ClocksStateInvalid - should not be used*/ -{ .display_clk_khz = 0, .pixel_clk_khz = 0 }, -/*ClocksStateUltraLow - currently by HW design team not supposed to be used*/ -{ .display_clk_khz = 389189, .pixel_clk_khz = 346672 }, -/*ClocksStateLow*/ -{ .display_clk_khz = 459000, .pixel_clk_khz = 400000 }, -/*ClocksStateNominal*/ -{ .display_clk_khz = 667000, .pixel_clk_khz = 600000 }, -/*ClocksStatePerformance*/ -{ .display_clk_khz = 1132000, .pixel_clk_khz = 600000 } }; - -static const struct state_dependent_clocks dce120_max_clks_by_state[] = { -/*ClocksStateInvalid - should not be used*/ -{ .display_clk_khz = 0, .pixel_clk_khz = 0 }, -/*ClocksStateUltraLow - currently by HW design team not supposed to be used*/ -{ .display_clk_khz = 0, .pixel_clk_khz = 0 }, -/*ClocksStateLow*/ -{ .display_clk_khz = 460000, .pixel_clk_khz = 400000 }, -/*ClocksStateNominal*/ -{ .display_clk_khz = 670000, .pixel_clk_khz = 600000 }, -/*ClocksStatePerformance*/ -{ .display_clk_khz = 1133000, .pixel_clk_khz = 600000 } }; - -/* Starting DID for each range */ -enum dentist_base_divider_id { - DENTIST_BASE_DID_1 = 0x08, - DENTIST_BASE_DID_2 = 0x40, - DENTIST_BASE_DID_3 = 0x60, - DENTIST_BASE_DID_4 = 0x7e, - DENTIST_MAX_DID = 0x7f -}; - -/* Starting point and step size for each divider range.*/ -enum dentist_divider_range { - DENTIST_DIVIDER_RANGE_1_START = 8, /* 2.00 */ - DENTIST_DIVIDER_RANGE_1_STEP = 1, /* 0.25 */ - DENTIST_DIVIDER_RANGE_2_START = 64, /* 16.00 */ - DENTIST_DIVIDER_RANGE_2_STEP = 2, /* 0.50 */ - DENTIST_DIVIDER_RANGE_3_START = 128, /* 32.00 */ - DENTIST_DIVIDER_RANGE_3_STEP = 4, /* 1.00 */ - DENTIST_DIVIDER_RANGE_4_START = 248, /* 62.00 */ - DENTIST_DIVIDER_RANGE_4_STEP = 264, /* 66.00 */ - DENTIST_DIVIDER_RANGE_SCALE_FACTOR = 4 -}; - -static int dentist_get_divider_from_did(int did) -{ - if (did < DENTIST_BASE_DID_1) - did = DENTIST_BASE_DID_1; - if (did > DENTIST_MAX_DID) - did = DENTIST_MAX_DID; - - if (did < DENTIST_BASE_DID_2) { - return DENTIST_DIVIDER_RANGE_1_START + DENTIST_DIVIDER_RANGE_1_STEP - * (did - DENTIST_BASE_DID_1); - } else if (did < DENTIST_BASE_DID_3) { - return DENTIST_DIVIDER_RANGE_2_START + DENTIST_DIVIDER_RANGE_2_STEP - * (did - DENTIST_BASE_DID_2); - } else if (did < DENTIST_BASE_DID_4) { - return DENTIST_DIVIDER_RANGE_3_START + DENTIST_DIVIDER_RANGE_3_STEP - * (did - DENTIST_BASE_DID_3); - } else { - return DENTIST_DIVIDER_RANGE_4_START + DENTIST_DIVIDER_RANGE_4_STEP - * (did - DENTIST_BASE_DID_4); - } -} - -/* SW will adjust DP REF Clock average value for all purposes - * (DP DTO / DP Audio DTO and DP GTC) - if clock is spread for all cases: - -if SS enabled on DP Ref clock and HW de-spreading enabled with SW - calculations for DS_INCR/DS_MODULO (this is planned to be default case) - -if SS enabled on DP Ref clock and HW de-spreading enabled with HW - calculations (not planned to be used, but average clock should still - be valid) - -if SS enabled on DP Ref clock and HW de-spreading disabled - (should not be case with CIK) then SW should program all rates - generated according to average value (case as with previous ASICs) - */ -static int dccg_adjust_dp_ref_freq_for_ss(struct dce_dccg *clk_dce, int dp_ref_clk_khz) -{ - if (clk_dce->ss_on_dprefclk && clk_dce->dprefclk_ss_divider != 0) { - struct fixed31_32 ss_percentage = dc_fixpt_div_int( - dc_fixpt_from_fraction(clk_dce->dprefclk_ss_percentage, - clk_dce->dprefclk_ss_divider), 200); - struct fixed31_32 adj_dp_ref_clk_khz; - - ss_percentage = dc_fixpt_sub(dc_fixpt_one, ss_percentage); - adj_dp_ref_clk_khz = dc_fixpt_mul_int(ss_percentage, dp_ref_clk_khz); - dp_ref_clk_khz = dc_fixpt_floor(adj_dp_ref_clk_khz); - } - return dp_ref_clk_khz; -} - -static int dce_get_dp_ref_freq_khz(struct dccg *clk) -{ - struct dce_dccg *clk_dce = TO_DCE_CLOCKS(clk); - int dprefclk_wdivider; - int dprefclk_src_sel; - int dp_ref_clk_khz = 600000; - int target_div; - - /* ASSERT DP Reference Clock source is from DFS*/ - REG_GET(DPREFCLK_CNTL, DPREFCLK_SRC_SEL, &dprefclk_src_sel); - ASSERT(dprefclk_src_sel == 0); - - /* Read the mmDENTIST_DISPCLK_CNTL to get the currently - * programmed DID DENTIST_DPREFCLK_WDIVIDER*/ - REG_GET(DENTIST_DISPCLK_CNTL, DENTIST_DPREFCLK_WDIVIDER, &dprefclk_wdivider); - - /* Convert DENTIST_DPREFCLK_WDIVIDERto actual divider*/ - target_div = dentist_get_divider_from_did(dprefclk_wdivider); - - /* Calculate the current DFS clock, in kHz.*/ - dp_ref_clk_khz = (DENTIST_DIVIDER_RANGE_SCALE_FACTOR - * clk_dce->dentist_vco_freq_khz) / target_div; - - return dccg_adjust_dp_ref_freq_for_ss(clk_dce, dp_ref_clk_khz); -} - -static int dce12_get_dp_ref_freq_khz(struct dccg *clk) -{ - struct dce_dccg *clk_dce = TO_DCE_CLOCKS(clk); - - return dccg_adjust_dp_ref_freq_for_ss(clk_dce, clk_dce->dprefclk_khz); -} - -static enum dm_pp_clocks_state dce_get_required_clocks_state( - struct dccg *clk, - struct dc_clocks *req_clocks) -{ - struct dce_dccg *clk_dce = TO_DCE_CLOCKS(clk); - int i; - enum dm_pp_clocks_state low_req_clk; - - /* Iterate from highest supported to lowest valid state, and update - * lowest RequiredState with the lowest state that satisfies - * all required clocks - */ - for (i = clk->max_clks_state; i >= DM_PP_CLOCKS_STATE_ULTRA_LOW; i--) - if (req_clocks->dispclk_khz > - clk_dce->max_clks_by_state[i].display_clk_khz - || req_clocks->phyclk_khz > - clk_dce->max_clks_by_state[i].pixel_clk_khz) - break; - - low_req_clk = i + 1; - if (low_req_clk > clk->max_clks_state) { - /* set max clock state for high phyclock, invalid on exceeding display clock */ - if (clk_dce->max_clks_by_state[clk->max_clks_state].display_clk_khz - < req_clocks->dispclk_khz) - low_req_clk = DM_PP_CLOCKS_STATE_INVALID; - else - low_req_clk = clk->max_clks_state; - } - - return low_req_clk; -} - -static int dce_set_clock( - struct dccg *clk, - int requested_clk_khz) -{ - struct dce_dccg *clk_dce = TO_DCE_CLOCKS(clk); - struct bp_pixel_clock_parameters pxl_clk_params = { 0 }; - struct dc_bios *bp = clk->ctx->dc_bios; - int actual_clock = requested_clk_khz; - - /* Make sure requested clock isn't lower than minimum threshold*/ - if (requested_clk_khz > 0) - requested_clk_khz = max(requested_clk_khz, - clk_dce->dentist_vco_freq_khz / 64); - - /* Prepare to program display clock*/ - pxl_clk_params.target_pixel_clock = requested_clk_khz; - pxl_clk_params.pll_id = CLOCK_SOURCE_ID_DFS; - - if (clk_dce->dfs_bypass_active) - pxl_clk_params.flags.SET_DISPCLK_DFS_BYPASS = true; - - bp->funcs->program_display_engine_pll(bp, &pxl_clk_params); - - if (clk_dce->dfs_bypass_active) { - /* Cache the fixed display clock*/ - clk_dce->dfs_bypass_disp_clk = - pxl_clk_params.dfs_bypass_display_clock; - actual_clock = pxl_clk_params.dfs_bypass_display_clock; - } - - /* from power down, we need mark the clock state as ClocksStateNominal - * from HWReset, so when resume we will call pplib voltage regulator.*/ - if (requested_clk_khz == 0) - clk->cur_min_clks_state = DM_PP_CLOCKS_STATE_NOMINAL; - return actual_clock; -} - -static int dce_psr_set_clock( - struct dccg *clk, - int requested_clk_khz) -{ - struct dce_dccg *clk_dce = TO_DCE_CLOCKS(clk); - struct dc_context *ctx = clk_dce->base.ctx; - struct dc *core_dc = ctx->dc; - struct dmcu *dmcu = core_dc->res_pool->dmcu; - int actual_clk_khz = requested_clk_khz; - - actual_clk_khz = dce_set_clock(clk, requested_clk_khz); - - dmcu->funcs->set_psr_wait_loop(dmcu, actual_clk_khz / 1000 / 7); - return actual_clk_khz; -} - -static int dce112_set_clock( - struct dccg *clk, - int requested_clk_khz) -{ - struct dce_dccg *clk_dce = TO_DCE_CLOCKS(clk); - struct bp_set_dce_clock_parameters dce_clk_params; - struct dc_bios *bp = clk->ctx->dc_bios; - struct dc *core_dc = clk->ctx->dc; - struct dmcu *dmcu = core_dc->res_pool->dmcu; - int actual_clock = requested_clk_khz; - /* Prepare to program display clock*/ - memset(&dce_clk_params, 0, sizeof(dce_clk_params)); - - /* Make sure requested clock isn't lower than minimum threshold*/ - if (requested_clk_khz > 0) - requested_clk_khz = max(requested_clk_khz, - clk_dce->dentist_vco_freq_khz / 62); - - dce_clk_params.target_clock_frequency = requested_clk_khz; - dce_clk_params.pll_id = CLOCK_SOURCE_ID_DFS; - dce_clk_params.clock_type = DCECLOCK_TYPE_DISPLAY_CLOCK; - - bp->funcs->set_dce_clock(bp, &dce_clk_params); - actual_clock = dce_clk_params.target_clock_frequency; - - /* from power down, we need mark the clock state as ClocksStateNominal - * from HWReset, so when resume we will call pplib voltage regulator.*/ - if (requested_clk_khz == 0) - clk->cur_min_clks_state = DM_PP_CLOCKS_STATE_NOMINAL; - - /*Program DP ref Clock*/ - /*VBIOS will determine DPREFCLK frequency, so we don't set it*/ - dce_clk_params.target_clock_frequency = 0; - dce_clk_params.clock_type = DCECLOCK_TYPE_DPREFCLK; - if (!ASICREV_IS_VEGA20_P(clk->ctx->asic_id.hw_internal_rev)) - dce_clk_params.flags.USE_GENLOCK_AS_SOURCE_FOR_DPREFCLK = - (dce_clk_params.pll_id == - CLOCK_SOURCE_COMBO_DISPLAY_PLL0); - else - dce_clk_params.flags.USE_GENLOCK_AS_SOURCE_FOR_DPREFCLK = false; - - bp->funcs->set_dce_clock(bp, &dce_clk_params); - - if (!IS_FPGA_MAXIMUS_DC(core_dc->ctx->dce_environment)) { - if (clk_dce->dfs_bypass_disp_clk != actual_clock) - dmcu->funcs->set_psr_wait_loop(dmcu, - actual_clock / 1000 / 7); - } - - clk_dce->dfs_bypass_disp_clk = actual_clock; - return actual_clock; -} - -static void dce_clock_read_integrated_info(struct dce_dccg *clk_dce) -{ - struct dc_debug_options *debug = &clk_dce->base.ctx->dc->debug; - struct dc_bios *bp = clk_dce->base.ctx->dc_bios; - struct integrated_info info = { { { 0 } } }; - struct dc_firmware_info fw_info = { { 0 } }; - int i; - - if (bp->integrated_info) - info = *bp->integrated_info; - - clk_dce->dentist_vco_freq_khz = info.dentist_vco_freq; - if (clk_dce->dentist_vco_freq_khz == 0) { - bp->funcs->get_firmware_info(bp, &fw_info); - clk_dce->dentist_vco_freq_khz = - fw_info.smu_gpu_pll_output_freq; - if (clk_dce->dentist_vco_freq_khz == 0) - clk_dce->dentist_vco_freq_khz = 3600000; - } - - /*update the maximum display clock for each power state*/ - for (i = 0; i < NUMBER_OF_DISP_CLK_VOLTAGE; ++i) { - enum dm_pp_clocks_state clk_state = DM_PP_CLOCKS_STATE_INVALID; - - switch (i) { - case 0: - clk_state = DM_PP_CLOCKS_STATE_ULTRA_LOW; - break; - - case 1: - clk_state = DM_PP_CLOCKS_STATE_LOW; - break; - - case 2: - clk_state = DM_PP_CLOCKS_STATE_NOMINAL; - break; - - case 3: - clk_state = DM_PP_CLOCKS_STATE_PERFORMANCE; - break; - - default: - clk_state = DM_PP_CLOCKS_STATE_INVALID; - break; - } - - /*Do not allow bad VBIOS/SBIOS to override with invalid values, - * check for > 100MHz*/ - if (info.disp_clk_voltage[i].max_supported_clk >= 100000) - clk_dce->max_clks_by_state[clk_state].display_clk_khz = - info.disp_clk_voltage[i].max_supported_clk; - } - - if (!debug->disable_dfs_bypass && bp->integrated_info) - if (bp->integrated_info->gpu_cap_info & DFS_BYPASS_ENABLE) - clk_dce->dfs_bypass_enabled = true; -} - -static void dce_clock_read_ss_info(struct dce_dccg *clk_dce) -{ - struct dc_bios *bp = clk_dce->base.ctx->dc_bios; - int ss_info_num = bp->funcs->get_ss_entry_number( - bp, AS_SIGNAL_TYPE_GPU_PLL); - - if (ss_info_num) { - struct spread_spectrum_info info = { { 0 } }; - enum bp_result result = bp->funcs->get_spread_spectrum_info( - bp, AS_SIGNAL_TYPE_GPU_PLL, 0, &info); - - /* Based on VBIOS, VBIOS will keep entry for GPU PLL SS - * even if SS not enabled and in that case - * SSInfo.spreadSpectrumPercentage !=0 would be sign - * that SS is enabled - */ - if (result == BP_RESULT_OK && - info.spread_spectrum_percentage != 0) { - clk_dce->ss_on_dprefclk = true; - clk_dce->dprefclk_ss_divider = info.spread_percentage_divider; - - if (info.type.CENTER_MODE == 0) { - /* TODO: Currently for DP Reference clock we - * need only SS percentage for - * downspread */ - clk_dce->dprefclk_ss_percentage = - info.spread_spectrum_percentage; - } - - return; - } - - result = bp->funcs->get_spread_spectrum_info( - bp, AS_SIGNAL_TYPE_DISPLAY_PORT, 0, &info); - - /* Based on VBIOS, VBIOS will keep entry for DPREFCLK SS - * even if SS not enabled and in that case - * SSInfo.spreadSpectrumPercentage !=0 would be sign - * that SS is enabled - */ - if (result == BP_RESULT_OK && - info.spread_spectrum_percentage != 0) { - clk_dce->ss_on_dprefclk = true; - clk_dce->dprefclk_ss_divider = info.spread_percentage_divider; - - if (info.type.CENTER_MODE == 0) { - /* Currently for DP Reference clock we - * need only SS percentage for - * downspread */ - clk_dce->dprefclk_ss_percentage = - info.spread_spectrum_percentage; - } - } - } -} - -static inline bool should_set_clock(bool safe_to_lower, int calc_clk, int cur_clk) -{ - return ((safe_to_lower && calc_clk < cur_clk) || calc_clk > cur_clk); -} - -static void dce12_update_clocks(struct dccg *dccg, - struct dc_clocks *new_clocks, - bool safe_to_lower) -{ - struct dm_pp_clock_for_voltage_req clock_voltage_req = {0}; - - /* TODO: Investigate why this is needed to fix display corruption. */ - new_clocks->dispclk_khz = new_clocks->dispclk_khz * 115 / 100; - - if (should_set_clock(safe_to_lower, new_clocks->dispclk_khz, dccg->clks.dispclk_khz)) { - clock_voltage_req.clk_type = DM_PP_CLOCK_TYPE_DISPLAY_CLK; - clock_voltage_req.clocks_in_khz = new_clocks->dispclk_khz; - new_clocks->dispclk_khz = dccg->funcs->set_dispclk(dccg, new_clocks->dispclk_khz); - dccg->clks.dispclk_khz = new_clocks->dispclk_khz; - - dm_pp_apply_clock_for_voltage_request(dccg->ctx, &clock_voltage_req); - } - - if (should_set_clock(safe_to_lower, new_clocks->phyclk_khz, dccg->clks.phyclk_khz)) { - clock_voltage_req.clk_type = DM_PP_CLOCK_TYPE_DISPLAYPHYCLK; - clock_voltage_req.clocks_in_khz = new_clocks->phyclk_khz; - dccg->clks.phyclk_khz = new_clocks->phyclk_khz; - - dm_pp_apply_clock_for_voltage_request(dccg->ctx, &clock_voltage_req); - } -} - -#ifdef CONFIG_DRM_AMD_DC_DCN1_0 -static int dcn1_determine_dppclk_threshold(struct dccg *dccg, struct dc_clocks *new_clocks) -{ - bool request_dpp_div = new_clocks->dispclk_khz > new_clocks->dppclk_khz; - bool dispclk_increase = new_clocks->dispclk_khz > dccg->clks.dispclk_khz; - int disp_clk_threshold = new_clocks->max_supported_dppclk_khz; - bool cur_dpp_div = dccg->clks.dispclk_khz > dccg->clks.dppclk_khz; - - /* increase clock, looking for div is 0 for current, request div is 1*/ - if (dispclk_increase) { - /* already divided by 2, no need to reach target clk with 2 steps*/ - if (cur_dpp_div) - return new_clocks->dispclk_khz; - - /* request disp clk is lower than maximum supported dpp clk, - * no need to reach target clk with two steps. - */ - if (new_clocks->dispclk_khz <= disp_clk_threshold) - return new_clocks->dispclk_khz; - - /* target dpp clk not request divided by 2, still within threshold */ - if (!request_dpp_div) - return new_clocks->dispclk_khz; - - } else { - /* decrease clock, looking for current dppclk divided by 2, - * request dppclk not divided by 2. - */ - - /* current dpp clk not divided by 2, no need to ramp*/ - if (!cur_dpp_div) - return new_clocks->dispclk_khz; - - /* current disp clk is lower than current maximum dpp clk, - * no need to ramp - */ - if (dccg->clks.dispclk_khz <= disp_clk_threshold) - return new_clocks->dispclk_khz; - - /* request dpp clk need to be divided by 2 */ - if (request_dpp_div) - return new_clocks->dispclk_khz; - } - - return disp_clk_threshold; -} - -static void dcn1_ramp_up_dispclk_with_dpp(struct dccg *dccg, struct dc_clocks *new_clocks) -{ - struct dc *dc = dccg->ctx->dc; - int dispclk_to_dpp_threshold = dcn1_determine_dppclk_threshold(dccg, new_clocks); - bool request_dpp_div = new_clocks->dispclk_khz > new_clocks->dppclk_khz; - int i; - - /* set disp clk to dpp clk threshold */ - dccg->funcs->set_dispclk(dccg, dispclk_to_dpp_threshold); - - /* update request dpp clk division option */ - for (i = 0; i < dc->res_pool->pipe_count; i++) { - struct pipe_ctx *pipe_ctx = &dc->current_state->res_ctx.pipe_ctx[i]; - - if (!pipe_ctx->plane_state) - continue; - - pipe_ctx->plane_res.dpp->funcs->dpp_dppclk_control( - pipe_ctx->plane_res.dpp, - request_dpp_div, - true); - } - - /* If target clk not same as dppclk threshold, set to target clock */ - if (dispclk_to_dpp_threshold != new_clocks->dispclk_khz) - dccg->funcs->set_dispclk(dccg, new_clocks->dispclk_khz); - - dccg->clks.dispclk_khz = new_clocks->dispclk_khz; - dccg->clks.dppclk_khz = new_clocks->dppclk_khz; - dccg->clks.max_supported_dppclk_khz = new_clocks->max_supported_dppclk_khz; -} - -static void dcn1_update_clocks(struct dccg *dccg, - struct dc_clocks *new_clocks, - bool safe_to_lower) -{ - struct dc *dc = dccg->ctx->dc; - struct pp_smu_display_requirement_rv *smu_req_cur = - &dc->res_pool->pp_smu_req; - struct pp_smu_display_requirement_rv smu_req = *smu_req_cur; - struct pp_smu_funcs_rv *pp_smu = dc->res_pool->pp_smu; - struct dm_pp_clock_for_voltage_req clock_voltage_req = {0}; - bool send_request_to_increase = false; - bool send_request_to_lower = false; - - if (new_clocks->phyclk_khz) - smu_req.display_count = 1; - else - smu_req.display_count = 0; - - if (new_clocks->dispclk_khz > dccg->clks.dispclk_khz - || new_clocks->phyclk_khz > dccg->clks.phyclk_khz - || new_clocks->fclk_khz > dccg->clks.fclk_khz - || new_clocks->dcfclk_khz > dccg->clks.dcfclk_khz) - send_request_to_increase = true; - - if (should_set_clock(safe_to_lower, new_clocks->phyclk_khz, dccg->clks.phyclk_khz)) { - dccg->clks.phyclk_khz = new_clocks->phyclk_khz; - - send_request_to_lower = true; - } - - if (should_set_clock(safe_to_lower, new_clocks->fclk_khz, dccg->clks.fclk_khz)) { - dccg->clks.fclk_khz = new_clocks->fclk_khz; - clock_voltage_req.clk_type = DM_PP_CLOCK_TYPE_FCLK; - clock_voltage_req.clocks_in_khz = new_clocks->fclk_khz; - smu_req.hard_min_fclk_khz = new_clocks->fclk_khz; - - dm_pp_apply_clock_for_voltage_request(dccg->ctx, &clock_voltage_req); - send_request_to_lower = true; - } - - if (should_set_clock(safe_to_lower, new_clocks->dcfclk_khz, dccg->clks.dcfclk_khz)) { - dccg->clks.dcfclk_khz = new_clocks->dcfclk_khz; - smu_req.hard_min_dcefclk_khz = new_clocks->dcfclk_khz; - - send_request_to_lower = true; - } - - if (should_set_clock(safe_to_lower, - new_clocks->dcfclk_deep_sleep_khz, dccg->clks.dcfclk_deep_sleep_khz)) { - dccg->clks.dcfclk_deep_sleep_khz = new_clocks->dcfclk_deep_sleep_khz; - smu_req.min_deep_sleep_dcefclk_mhz = new_clocks->dcfclk_deep_sleep_khz; - - send_request_to_lower = true; - } - - /* make sure dcf clk is before dpp clk to - * make sure we have enough voltage to run dpp clk - */ - if (send_request_to_increase) { - /*use dcfclk to request voltage*/ - clock_voltage_req.clk_type = DM_PP_CLOCK_TYPE_DCFCLK; - clock_voltage_req.clocks_in_khz = dcn_find_dcfclk_suits_all(dc, new_clocks); - dm_pp_apply_clock_for_voltage_request(dccg->ctx, &clock_voltage_req); - if (pp_smu->set_display_requirement) - pp_smu->set_display_requirement(&pp_smu->pp_smu, &smu_req); - } - - /* dcn1 dppclk is tied to dispclk */ - /* program dispclk on = as a w/a for sleep resume clock ramping issues */ - if (should_set_clock(safe_to_lower, new_clocks->dispclk_khz, dccg->clks.dispclk_khz) - || new_clocks->dispclk_khz == dccg->clks.dispclk_khz) { - dcn1_ramp_up_dispclk_with_dpp(dccg, new_clocks); - dccg->clks.dispclk_khz = new_clocks->dispclk_khz; - - send_request_to_lower = true; - } - - if (!send_request_to_increase && send_request_to_lower) { - /*use dcfclk to request voltage*/ - clock_voltage_req.clk_type = DM_PP_CLOCK_TYPE_DCFCLK; - clock_voltage_req.clocks_in_khz = dcn_find_dcfclk_suits_all(dc, new_clocks); - dm_pp_apply_clock_for_voltage_request(dccg->ctx, &clock_voltage_req); - if (pp_smu->set_display_requirement) - pp_smu->set_display_requirement(&pp_smu->pp_smu, &smu_req); - } - - - *smu_req_cur = smu_req; -} -#endif - -static void dce_update_clocks(struct dccg *dccg, - struct dc_clocks *new_clocks, - bool safe_to_lower) -{ - struct dm_pp_power_level_change_request level_change_req; - struct dce_dccg *clk_dce = TO_DCE_CLOCKS(dccg); - - /* TODO: Investigate why this is needed to fix display corruption. */ - if (!clk_dce->dfs_bypass_active) - new_clocks->dispclk_khz = new_clocks->dispclk_khz * 115 / 100; - - level_change_req.power_level = dce_get_required_clocks_state(dccg, new_clocks); - /* get max clock state from PPLIB */ - if ((level_change_req.power_level < dccg->cur_min_clks_state && safe_to_lower) - || level_change_req.power_level > dccg->cur_min_clks_state) { - if (dm_pp_apply_power_level_change_request(dccg->ctx, &level_change_req)) - dccg->cur_min_clks_state = level_change_req.power_level; - } - - if (should_set_clock(safe_to_lower, new_clocks->dispclk_khz, dccg->clks.dispclk_khz)) { - new_clocks->dispclk_khz = dccg->funcs->set_dispclk(dccg, new_clocks->dispclk_khz); - dccg->clks.dispclk_khz = new_clocks->dispclk_khz; - } -} - -static bool dce_update_dfs_bypass( - struct dccg *dccg, - struct dc *dc, - struct dc_state *context, - int requested_clock_khz) -{ - struct dce_dccg *clk_dce = TO_DCE_CLOCKS(dccg); - struct resource_context *res_ctx = &context->res_ctx; - enum signal_type signal_type = SIGNAL_TYPE_NONE; - bool was_active = clk_dce->dfs_bypass_active; - int i; - - /* Disable DFS bypass by default. */ - clk_dce->dfs_bypass_active = false; - - /* Check that DFS bypass is available. */ - if (!clk_dce->dfs_bypass_enabled) - goto update; - - /* Check if the requested display clock is below the threshold. */ - if (requested_clock_khz >= 400000) - goto update; - - /* DFS-bypass should only be enabled on single stream setups */ - if (context->stream_count != 1) - goto update; - - /* Check that the stream's signal type is an embedded panel */ - for (i = 0; i < dc->res_pool->pipe_count; i++) { - if (res_ctx->pipe_ctx[i].stream) { - struct pipe_ctx *pipe_ctx = &res_ctx->pipe_ctx[i]; - - signal_type = pipe_ctx->stream->sink->link->connector_signal; - break; - } - } - - if (signal_type == SIGNAL_TYPE_EDP || - signal_type == SIGNAL_TYPE_LVDS) - clk_dce->dfs_bypass_active = true; - -update: - /* Update the clock state. We don't need to respect safe_to_lower - * because DFS bypass should always be greater than the current - * display clock frequency. - */ - if (was_active != clk_dce->dfs_bypass_active) { - dccg->clks.dispclk_khz = - dccg->funcs->set_dispclk(dccg, dccg->clks.dispclk_khz); - return true; - } - - return false; -} - -#ifdef CONFIG_DRM_AMD_DC_DCN1_0 -static const struct display_clock_funcs dcn1_funcs = { - .get_dp_ref_clk_frequency = dce12_get_dp_ref_freq_khz, - .set_dispclk = dce112_set_clock, - .update_clocks = dcn1_update_clocks -}; -#endif - -static const struct display_clock_funcs dce120_funcs = { - .get_dp_ref_clk_frequency = dce12_get_dp_ref_freq_khz, - .set_dispclk = dce112_set_clock, - .update_clocks = dce12_update_clocks -}; - -static const struct display_clock_funcs dce112_funcs = { - .get_dp_ref_clk_frequency = dce_get_dp_ref_freq_khz, - .set_dispclk = dce112_set_clock, - .update_clocks = dce_update_clocks -}; - -static const struct display_clock_funcs dce110_funcs = { - .get_dp_ref_clk_frequency = dce_get_dp_ref_freq_khz, - .set_dispclk = dce_psr_set_clock, - .update_clocks = dce_update_clocks, - .update_dfs_bypass = dce_update_dfs_bypass -}; - -static const struct display_clock_funcs dce_funcs = { - .get_dp_ref_clk_frequency = dce_get_dp_ref_freq_khz, - .set_dispclk = dce_set_clock, - .update_clocks = dce_update_clocks -}; - -static void dce_dccg_construct( - struct dce_dccg *clk_dce, - struct dc_context *ctx, - const struct dccg_registers *regs, - const struct dccg_shift *clk_shift, - const struct dccg_mask *clk_mask) -{ - struct dccg *base = &clk_dce->base; - - base->ctx = ctx; - base->funcs = &dce_funcs; - - clk_dce->regs = regs; - clk_dce->clk_shift = clk_shift; - clk_dce->clk_mask = clk_mask; - - clk_dce->dfs_bypass_disp_clk = 0; - - clk_dce->dprefclk_ss_percentage = 0; - clk_dce->dprefclk_ss_divider = 1000; - clk_dce->ss_on_dprefclk = false; - - base->max_clks_state = DM_PP_CLOCKS_STATE_NOMINAL; - base->cur_min_clks_state = DM_PP_CLOCKS_STATE_INVALID; - - dce_clock_read_integrated_info(clk_dce); - dce_clock_read_ss_info(clk_dce); -} - -struct dccg *dce_dccg_create( - struct dc_context *ctx, - const struct dccg_registers *regs, - const struct dccg_shift *clk_shift, - const struct dccg_mask *clk_mask) -{ - struct dce_dccg *clk_dce = kzalloc(sizeof(*clk_dce), GFP_KERNEL); - - if (clk_dce == NULL) { - BREAK_TO_DEBUGGER(); - return NULL; - } - - memcpy(clk_dce->max_clks_by_state, - dce80_max_clks_by_state, - sizeof(dce80_max_clks_by_state)); - - dce_dccg_construct( - clk_dce, ctx, regs, clk_shift, clk_mask); - - return &clk_dce->base; -} - -struct dccg *dce110_dccg_create( - struct dc_context *ctx, - const struct dccg_registers *regs, - const struct dccg_shift *clk_shift, - const struct dccg_mask *clk_mask) -{ - struct dce_dccg *clk_dce = kzalloc(sizeof(*clk_dce), GFP_KERNEL); - - if (clk_dce == NULL) { - BREAK_TO_DEBUGGER(); - return NULL; - } - - memcpy(clk_dce->max_clks_by_state, - dce110_max_clks_by_state, - sizeof(dce110_max_clks_by_state)); - - dce_dccg_construct( - clk_dce, ctx, regs, clk_shift, clk_mask); - - clk_dce->base.funcs = &dce110_funcs; - - return &clk_dce->base; -} - -struct dccg *dce112_dccg_create( - struct dc_context *ctx, - const struct dccg_registers *regs, - const struct dccg_shift *clk_shift, - const struct dccg_mask *clk_mask) -{ - struct dce_dccg *clk_dce = kzalloc(sizeof(*clk_dce), GFP_KERNEL); - - if (clk_dce == NULL) { - BREAK_TO_DEBUGGER(); - return NULL; - } - - memcpy(clk_dce->max_clks_by_state, - dce112_max_clks_by_state, - sizeof(dce112_max_clks_by_state)); - - dce_dccg_construct( - clk_dce, ctx, regs, clk_shift, clk_mask); - - clk_dce->base.funcs = &dce112_funcs; - - return &clk_dce->base; -} - -struct dccg *dce120_dccg_create(struct dc_context *ctx) -{ - struct dce_dccg *clk_dce = kzalloc(sizeof(*clk_dce), GFP_KERNEL); - - if (clk_dce == NULL) { - BREAK_TO_DEBUGGER(); - return NULL; - } - - memcpy(clk_dce->max_clks_by_state, - dce120_max_clks_by_state, - sizeof(dce120_max_clks_by_state)); - - dce_dccg_construct( - clk_dce, ctx, NULL, NULL, NULL); - - clk_dce->dprefclk_khz = 600000; - clk_dce->base.funcs = &dce120_funcs; - - return &clk_dce->base; -} - -#ifdef CONFIG_DRM_AMD_DC_DCN1_0 -struct dccg *dcn1_dccg_create(struct dc_context *ctx) -{ - struct dc_debug_options *debug = &ctx->dc->debug; - struct dc_bios *bp = ctx->dc_bios; - struct dc_firmware_info fw_info = { { 0 } }; - struct dce_dccg *clk_dce = kzalloc(sizeof(*clk_dce), GFP_KERNEL); - - if (clk_dce == NULL) { - BREAK_TO_DEBUGGER(); - return NULL; - } - - clk_dce->base.ctx = ctx; - clk_dce->base.funcs = &dcn1_funcs; - - clk_dce->dfs_bypass_disp_clk = 0; - - clk_dce->dprefclk_ss_percentage = 0; - clk_dce->dprefclk_ss_divider = 1000; - clk_dce->ss_on_dprefclk = false; - - clk_dce->dprefclk_khz = 600000; - if (bp->integrated_info) - clk_dce->dentist_vco_freq_khz = bp->integrated_info->dentist_vco_freq; - if (clk_dce->dentist_vco_freq_khz == 0) { - bp->funcs->get_firmware_info(bp, &fw_info); - clk_dce->dentist_vco_freq_khz = fw_info.smu_gpu_pll_output_freq; - if (clk_dce->dentist_vco_freq_khz == 0) - clk_dce->dentist_vco_freq_khz = 3600000; - } - - if (!debug->disable_dfs_bypass && bp->integrated_info) - if (bp->integrated_info->gpu_cap_info & DFS_BYPASS_ENABLE) - clk_dce->dfs_bypass_enabled = true; - - dce_clock_read_ss_info(clk_dce); - - return &clk_dce->base; -} -#endif - -void dce_dccg_destroy(struct dccg **dccg) -{ - struct dce_dccg *clk_dce = TO_DCE_CLOCKS(*dccg); - - kfree(clk_dce); - *dccg = NULL; -} diff --git a/drivers/gpu/drm/amd/display/dc/dce/dce_clocks.h b/drivers/gpu/drm/amd/display/dc/dce/dce_clocks.h deleted file mode 100644 index 34fdb386c884..000000000000 --- a/drivers/gpu/drm/amd/display/dc/dce/dce_clocks.h +++ /dev/null @@ -1,123 +0,0 @@ -/* - * Copyright 2012-16 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. - * - * Authors: AMD - * - */ - - -#ifndef _DCE_CLOCKS_H_ -#define _DCE_CLOCKS_H_ - -#include "display_clock.h" - -#define CLK_COMMON_REG_LIST_DCE_BASE() \ - .DPREFCLK_CNTL = mmDPREFCLK_CNTL, \ - .DENTIST_DISPCLK_CNTL = mmDENTIST_DISPCLK_CNTL - -#define CLK_COMMON_REG_LIST_DCN_BASE() \ - SR(DENTIST_DISPCLK_CNTL) - -#define CLK_SF(reg_name, field_name, post_fix)\ - .field_name = reg_name ## __ ## field_name ## post_fix - -#define CLK_COMMON_MASK_SH_LIST_DCE_COMMON_BASE(mask_sh) \ - CLK_SF(DPREFCLK_CNTL, DPREFCLK_SRC_SEL, mask_sh), \ - CLK_SF(DENTIST_DISPCLK_CNTL, DENTIST_DPREFCLK_WDIVIDER, mask_sh) - -#define CLK_COMMON_MASK_SH_LIST_DCN_COMMON_BASE(mask_sh) \ - CLK_SF(DENTIST_DISPCLK_CNTL, DENTIST_DISPCLK_WDIVIDER, mask_sh),\ - CLK_SF(DENTIST_DISPCLK_CNTL, DENTIST_DISPCLK_CHG_DONE, mask_sh) - -#define CLK_REG_FIELD_LIST(type) \ - type DPREFCLK_SRC_SEL; \ - type DENTIST_DPREFCLK_WDIVIDER; \ - type DENTIST_DISPCLK_WDIVIDER; \ - type DENTIST_DISPCLK_CHG_DONE; - -struct dccg_shift { - CLK_REG_FIELD_LIST(uint8_t) -}; - -struct dccg_mask { - CLK_REG_FIELD_LIST(uint32_t) -}; - -struct dccg_registers { - uint32_t DPREFCLK_CNTL; - uint32_t DENTIST_DISPCLK_CNTL; -}; - -struct dce_dccg { - struct dccg base; - const struct dccg_registers *regs; - const struct dccg_shift *clk_shift; - const struct dccg_mask *clk_mask; - - struct state_dependent_clocks max_clks_by_state[DM_PP_CLOCKS_MAX_STATES]; - - int dentist_vco_freq_khz; - - /* Cache the status of DFS-bypass feature*/ - bool dfs_bypass_enabled; - /* True if the DFS-bypass feature is enabled and active. */ - bool dfs_bypass_active; - /* Cache the display clock returned by VBIOS if DFS-bypass is enabled. - * This is basically "Crystal Frequency In KHz" (XTALIN) frequency */ - int dfs_bypass_disp_clk; - - /* Flag for Enabled SS on DPREFCLK */ - bool ss_on_dprefclk; - /* DPREFCLK SS percentage (if down-spread enabled) */ - int dprefclk_ss_percentage; - /* DPREFCLK SS percentage Divider (100 or 1000) */ - int dprefclk_ss_divider; - int dprefclk_khz; -}; - - -struct dccg *dce_dccg_create( - struct dc_context *ctx, - const struct dccg_registers *regs, - const struct dccg_shift *clk_shift, - const struct dccg_mask *clk_mask); - -struct dccg *dce110_dccg_create( - struct dc_context *ctx, - const struct dccg_registers *regs, - const struct dccg_shift *clk_shift, - const struct dccg_mask *clk_mask); - -struct dccg *dce112_dccg_create( - struct dc_context *ctx, - const struct dccg_registers *regs, - const struct dccg_shift *clk_shift, - const struct dccg_mask *clk_mask); - -struct dccg *dce120_dccg_create(struct dc_context *ctx); - -#ifdef CONFIG_DRM_AMD_DC_DCN1_0 -struct dccg *dcn1_dccg_create(struct dc_context *ctx); -#endif - -void dce_dccg_destroy(struct dccg **dccg); - -#endif /* _DCE_CLOCKS_H_ */ diff --git a/drivers/gpu/drm/amd/display/dc/dce/dce_dmcu.c b/drivers/gpu/drm/amd/display/dc/dce/dce_dmcu.c index dea40b322191..c2926cf19dee 100644 --- a/drivers/gpu/drm/amd/display/dc/dce/dce_dmcu.c +++ b/drivers/gpu/drm/amd/display/dc/dce/dce_dmcu.c @@ -51,7 +51,6 @@ #define PSR_SET_WAITLOOP 0x31 #define MCP_INIT_DMCU 0x88 #define MCP_INIT_IRAM 0x89 -#define MCP_DMCU_VERSION 0x90 #define MASTER_COMM_CNTL_REG__MASTER_COMM_INTERRUPT_MASK 0x00000001L static bool dce_dmcu_init(struct dmcu *dmcu) @@ -317,38 +316,11 @@ static void dce_get_psr_wait_loop( } #if defined(CONFIG_DRM_AMD_DC_DCN1_0) -static void dcn10_get_dmcu_state(struct dmcu *dmcu) -{ - struct dce_dmcu *dmcu_dce = TO_DCE_DMCU(dmcu); - uint32_t dmcu_state_offset = 0xf6; - - /* Enable write access to IRAM */ - REG_UPDATE_2(DMCU_RAM_ACCESS_CTRL, - IRAM_HOST_ACCESS_EN, 1, - IRAM_RD_ADDR_AUTO_INC, 1); - - REG_WAIT(DMU_MEM_PWR_CNTL, DMCU_IRAM_MEM_PWR_STATE, 0, 2, 10); - - /* Write address to IRAM_RD_ADDR in DMCU_IRAM_RD_CTRL */ - REG_WRITE(DMCU_IRAM_RD_CTRL, dmcu_state_offset); - - /* Read data from IRAM_RD_DATA in DMCU_IRAM_RD_DATA*/ - dmcu->dmcu_state = REG_READ(DMCU_IRAM_RD_DATA); - - /* Disable write access to IRAM to allow dynamic sleep state */ - REG_UPDATE_2(DMCU_RAM_ACCESS_CTRL, - IRAM_HOST_ACCESS_EN, 0, - IRAM_RD_ADDR_AUTO_INC, 0); -} - static void dcn10_get_dmcu_version(struct dmcu *dmcu) { struct dce_dmcu *dmcu_dce = TO_DCE_DMCU(dmcu); uint32_t dmcu_version_offset = 0xf1; - /* Clear scratch */ - REG_WRITE(DC_DMCU_SCRATCH, 0); - /* Enable write access to IRAM */ REG_UPDATE_2(DMCU_RAM_ACCESS_CTRL, IRAM_HOST_ACCESS_EN, 1, @@ -359,85 +331,74 @@ static void dcn10_get_dmcu_version(struct dmcu *dmcu) /* Write address to IRAM_RD_ADDR and read from DATA register */ REG_WRITE(DMCU_IRAM_RD_CTRL, dmcu_version_offset); dmcu->dmcu_version.interface_version = REG_READ(DMCU_IRAM_RD_DATA); - dmcu->dmcu_version.year = ((REG_READ(DMCU_IRAM_RD_DATA) << 8) | + dmcu->dmcu_version.abm_version = REG_READ(DMCU_IRAM_RD_DATA); + dmcu->dmcu_version.psr_version = REG_READ(DMCU_IRAM_RD_DATA); + dmcu->dmcu_version.build_version = ((REG_READ(DMCU_IRAM_RD_DATA) << 8) | REG_READ(DMCU_IRAM_RD_DATA)); - dmcu->dmcu_version.month = REG_READ(DMCU_IRAM_RD_DATA); - dmcu->dmcu_version.date = REG_READ(DMCU_IRAM_RD_DATA); /* Disable write access to IRAM to allow dynamic sleep state */ REG_UPDATE_2(DMCU_RAM_ACCESS_CTRL, IRAM_HOST_ACCESS_EN, 0, IRAM_RD_ADDR_AUTO_INC, 0); - - /* Send MCP command message to DMCU to get version reply from FW. - * We expect this version should match the one in IRAM, otherwise - * something is wrong with DMCU and we should fail and disable UC. - */ - REG_WAIT(MASTER_COMM_CNTL_REG, MASTER_COMM_INTERRUPT, 0, 100, 800); - - /* Set command to get DMCU version from microcontroller */ - REG_UPDATE(MASTER_COMM_CMD_REG, MASTER_COMM_CMD_REG_BYTE0, - MCP_DMCU_VERSION); - - /* Notify microcontroller of new command */ - REG_UPDATE(MASTER_COMM_CNTL_REG, MASTER_COMM_INTERRUPT, 1); - - /* Ensure command has been executed before continuing */ - REG_WAIT(MASTER_COMM_CNTL_REG, MASTER_COMM_INTERRUPT, 0, 100, 800); - - /* Somehow version does not match, so fail and return version 0 */ - if (dmcu->dmcu_version.interface_version != REG_READ(DC_DMCU_SCRATCH)) - dmcu->dmcu_version.interface_version = 0; } static bool dcn10_dmcu_init(struct dmcu *dmcu) { struct dce_dmcu *dmcu_dce = TO_DCE_DMCU(dmcu); + bool status = false; - /* DMCU FW should populate the scratch register if running */ - if (REG_READ(DC_DMCU_SCRATCH) == 0) - return false; - - /* Check state is uninitialized */ - dcn10_get_dmcu_state(dmcu); - - /* If microcontroller is already initialized, do nothing */ - if (dmcu->dmcu_state == DMCU_RUNNING) - return true; - - /* Retrieve and cache the DMCU firmware version. */ - dcn10_get_dmcu_version(dmcu); - - /* Check interface version to confirm firmware is loaded and running */ - if (dmcu->dmcu_version.interface_version == 0) - return false; + /* Definition of DC_DMCU_SCRATCH + * 0 : firmare not loaded + * 1 : PSP load DMCU FW but not initialized + * 2 : Firmware already initialized + */ + dmcu->dmcu_state = REG_READ(DC_DMCU_SCRATCH); - /* Wait until microcontroller is ready to process interrupt */ - REG_WAIT(MASTER_COMM_CNTL_REG, MASTER_COMM_INTERRUPT, 0, 100, 800); + switch (dmcu->dmcu_state) { + case DMCU_UNLOADED: + status = false; + break; + case DMCU_LOADED_UNINITIALIZED: + /* Wait until microcontroller is ready to process interrupt */ + REG_WAIT(MASTER_COMM_CNTL_REG, MASTER_COMM_INTERRUPT, 0, 100, 800); - /* Set initialized ramping boundary value */ - REG_WRITE(MASTER_COMM_DATA_REG1, 0xFFFF); + /* Set initialized ramping boundary value */ + REG_WRITE(MASTER_COMM_DATA_REG1, 0xFFFF); - /* Set command to initialize microcontroller */ - REG_UPDATE(MASTER_COMM_CMD_REG, MASTER_COMM_CMD_REG_BYTE0, + /* Set command to initialize microcontroller */ + REG_UPDATE(MASTER_COMM_CMD_REG, MASTER_COMM_CMD_REG_BYTE0, MCP_INIT_DMCU); - /* Notify microcontroller of new command */ - REG_UPDATE(MASTER_COMM_CNTL_REG, MASTER_COMM_INTERRUPT, 1); + /* Notify microcontroller of new command */ + REG_UPDATE(MASTER_COMM_CNTL_REG, MASTER_COMM_INTERRUPT, 1); - /* Ensure command has been executed before continuing */ - REG_WAIT(MASTER_COMM_CNTL_REG, MASTER_COMM_INTERRUPT, 0, 100, 800); + /* Ensure command has been executed before continuing */ + REG_WAIT(MASTER_COMM_CNTL_REG, MASTER_COMM_INTERRUPT, 0, 100, 800); - // Check state is initialized - dcn10_get_dmcu_state(dmcu); + // Check state is initialized + dmcu->dmcu_state = REG_READ(DC_DMCU_SCRATCH); - // If microcontroller is not in running state, fail - if (dmcu->dmcu_state != DMCU_RUNNING) - return false; + // If microcontroller is not in running state, fail + if (dmcu->dmcu_state == DMCU_RUNNING) { + /* Retrieve and cache the DMCU firmware version. */ + dcn10_get_dmcu_version(dmcu); + status = true; + } else + status = false; - return true; + break; + case DMCU_RUNNING: + status = true; + break; + default: + status = false; + break; + } + + return status; } + static bool dcn10_dmcu_load_iram(struct dmcu *dmcu, unsigned int start_offset, const char *src, diff --git a/drivers/gpu/drm/amd/display/dc/dce/dce_hwseq.h b/drivers/gpu/drm/amd/display/dc/dce/dce_hwseq.h index 64dc75378541..956bdf14503f 100644 --- a/drivers/gpu/drm/amd/display/dc/dce/dce_hwseq.h +++ b/drivers/gpu/drm/amd/display/dc/dce/dce_hwseq.h @@ -133,6 +133,10 @@ SR(DCHUB_AGP_TOP), \ BL_REG_LIST() +#define HWSEQ_VG20_REG_LIST() \ + HWSEQ_DCE120_REG_LIST(),\ + MMHUB_SR(MC_VM_XGMI_LFB_CNTL) + #define HWSEQ_DCE112_REG_LIST() \ HWSEQ_DCE10_REG_LIST(), \ HWSEQ_PIXEL_RATE_REG_LIST(CRTC), \ @@ -233,6 +237,16 @@ struct dce_hwseq_registers { uint32_t DOMAIN5_PG_CONFIG; uint32_t DOMAIN6_PG_CONFIG; uint32_t DOMAIN7_PG_CONFIG; + uint32_t DOMAIN8_PG_CONFIG; + uint32_t DOMAIN9_PG_CONFIG; + uint32_t DOMAIN10_PG_CONFIG; + uint32_t DOMAIN11_PG_CONFIG; + uint32_t DOMAIN16_PG_CONFIG; + uint32_t DOMAIN17_PG_CONFIG; + uint32_t DOMAIN18_PG_CONFIG; + uint32_t DOMAIN19_PG_CONFIG; + uint32_t DOMAIN20_PG_CONFIG; + uint32_t DOMAIN21_PG_CONFIG; uint32_t DOMAIN0_PG_STATUS; uint32_t DOMAIN1_PG_STATUS; uint32_t DOMAIN2_PG_STATUS; @@ -241,6 +255,16 @@ struct dce_hwseq_registers { uint32_t DOMAIN5_PG_STATUS; uint32_t DOMAIN6_PG_STATUS; uint32_t DOMAIN7_PG_STATUS; + uint32_t DOMAIN8_PG_STATUS; + uint32_t DOMAIN9_PG_STATUS; + uint32_t DOMAIN10_PG_STATUS; + uint32_t DOMAIN11_PG_STATUS; + uint32_t DOMAIN16_PG_STATUS; + uint32_t DOMAIN17_PG_STATUS; + uint32_t DOMAIN18_PG_STATUS; + uint32_t DOMAIN19_PG_STATUS; + uint32_t DOMAIN20_PG_STATUS; + uint32_t DOMAIN21_PG_STATUS; uint32_t DIO_MEM_PWR_CTRL; uint32_t DCCG_GATE_DISABLE_CNTL; uint32_t DCCG_GATE_DISABLE_CNTL2; @@ -262,6 +286,8 @@ struct dce_hwseq_registers { uint32_t D2VGA_CONTROL; uint32_t D3VGA_CONTROL; uint32_t D4VGA_CONTROL; + uint32_t D5VGA_CONTROL; + uint32_t D6VGA_CONTROL; uint32_t VGA_TEST_CONTROL; /* MMHUB registers. read only. temporary hack */ uint32_t VM_CONTEXT0_PAGE_TABLE_BASE_ADDR_HI32; @@ -276,6 +302,7 @@ struct dce_hwseq_registers { uint32_t MC_VM_SYSTEM_APERTURE_DEFAULT_ADDR_LSB; uint32_t MC_VM_SYSTEM_APERTURE_LOW_ADDR; uint32_t MC_VM_SYSTEM_APERTURE_HIGH_ADDR; + uint32_t MC_VM_XGMI_LFB_CNTL; uint32_t AZALIA_AUDIO_DTO; uint32_t AZALIA_CONTROLLER_CLOCK_GATING; }; @@ -360,6 +387,11 @@ struct dce_hwseq_registers { HWS_SF(, LVTMA_PWRSEQ_CNTL, LVTMA_BLON, mask_sh), \ HWS_SF(, LVTMA_PWRSEQ_STATE, LVTMA_PWRSEQ_TARGET_STATE_R, mask_sh) +#define HWSEQ_VG20_MASK_SH_LIST(mask_sh)\ + HWSEQ_DCE12_MASK_SH_LIST(mask_sh),\ + HWS_SF(, MC_VM_XGMI_LFB_CNTL, PF_LFB_REGION, mask_sh),\ + HWS_SF(, MC_VM_XGMI_LFB_CNTL, PF_MAX_REGION, mask_sh) + #define HWSEQ_DCN_MASK_SH_LIST(mask_sh)\ HWSEQ_PIXEL_RATE_MASK_SH_LIST(mask_sh, OTG0_),\ HWS_SF1(OTG0_, PHYPLL_PIXEL_RATE_CNTL, PHYPLL_PIXEL_RATE_SOURCE, mask_sh), \ @@ -448,6 +480,8 @@ struct dce_hwseq_registers { type PHYSICAL_PAGE_NUMBER_MSB;\ type PHYSICAL_PAGE_NUMBER_LSB;\ type LOGICAL_ADDR; \ + type PF_LFB_REGION;\ + type PF_MAX_REGION;\ type ENABLE_L1_TLB;\ type SYSTEM_ACCESS_MODE;\ type LVTMA_BLON;\ @@ -489,6 +523,26 @@ struct dce_hwseq_registers { type DOMAIN6_POWER_GATE; \ type DOMAIN7_POWER_FORCEON; \ type DOMAIN7_POWER_GATE; \ + type DOMAIN8_POWER_FORCEON; \ + type DOMAIN8_POWER_GATE; \ + type DOMAIN9_POWER_FORCEON; \ + type DOMAIN9_POWER_GATE; \ + type DOMAIN10_POWER_FORCEON; \ + type DOMAIN10_POWER_GATE; \ + type DOMAIN11_POWER_FORCEON; \ + type DOMAIN11_POWER_GATE; \ + type DOMAIN16_POWER_FORCEON; \ + type DOMAIN16_POWER_GATE; \ + type DOMAIN17_POWER_FORCEON; \ + type DOMAIN17_POWER_GATE; \ + type DOMAIN18_POWER_FORCEON; \ + type DOMAIN18_POWER_GATE; \ + type DOMAIN19_POWER_FORCEON; \ + type DOMAIN19_POWER_GATE; \ + type DOMAIN20_POWER_FORCEON; \ + type DOMAIN20_POWER_GATE; \ + type DOMAIN21_POWER_FORCEON; \ + type DOMAIN21_POWER_GATE; \ type DOMAIN0_PGFSM_PWR_STATUS; \ type DOMAIN1_PGFSM_PWR_STATUS; \ type DOMAIN2_PGFSM_PWR_STATUS; \ @@ -497,6 +551,16 @@ struct dce_hwseq_registers { type DOMAIN5_PGFSM_PWR_STATUS; \ type DOMAIN6_PGFSM_PWR_STATUS; \ type DOMAIN7_PGFSM_PWR_STATUS; \ + type DOMAIN8_PGFSM_PWR_STATUS; \ + type DOMAIN9_PGFSM_PWR_STATUS; \ + type DOMAIN10_PGFSM_PWR_STATUS; \ + type DOMAIN11_PGFSM_PWR_STATUS; \ + type DOMAIN16_PGFSM_PWR_STATUS; \ + type DOMAIN17_PGFSM_PWR_STATUS; \ + type DOMAIN18_PGFSM_PWR_STATUS; \ + type DOMAIN19_PGFSM_PWR_STATUS; \ + type DOMAIN20_PGFSM_PWR_STATUS; \ + type DOMAIN21_PGFSM_PWR_STATUS; \ type DCFCLK_GATE_DIS; \ type DCHUBBUB_GLOBAL_TIMER_REFDIV; \ type VGA_TEST_ENABLE; \ diff --git a/drivers/gpu/drm/amd/display/dc/dce/dce_link_encoder.c b/drivers/gpu/drm/amd/display/dc/dce/dce_link_encoder.c index 366bc8c2c643..314c04a915d2 100644 --- a/drivers/gpu/drm/amd/display/dc/dce/dce_link_encoder.c +++ b/drivers/gpu/drm/amd/display/dc/dce/dce_link_encoder.c @@ -599,12 +599,12 @@ bool dce110_link_encoder_validate_dvi_output( if ((connector_signal == SIGNAL_TYPE_DVI_SINGLE_LINK || connector_signal == SIGNAL_TYPE_HDMI_TYPE_A) && signal != SIGNAL_TYPE_HDMI_TYPE_A && - crtc_timing->pix_clk_khz > TMDS_MAX_PIXEL_CLOCK) + crtc_timing->pix_clk_100hz > (TMDS_MAX_PIXEL_CLOCK * 10)) return false; - if (crtc_timing->pix_clk_khz < TMDS_MIN_PIXEL_CLOCK) + if (crtc_timing->pix_clk_100hz < (TMDS_MIN_PIXEL_CLOCK * 10)) return false; - if (crtc_timing->pix_clk_khz > max_pixel_clock) + if (crtc_timing->pix_clk_100hz > (max_pixel_clock * 10)) return false; /* DVI supports 6/8bpp single-link and 10/16bpp dual-link */ @@ -645,7 +645,7 @@ static bool dce110_link_encoder_validate_hdmi_output( return false; /* DCE11 HW does not support 420 */ - if (!enc110->base.features.ycbcr420_supported && + if (!enc110->base.features.hdmi_ycbcr420_supported && crtc_timing->pixel_encoding == PIXEL_ENCODING_YCBCR420) return false; @@ -788,7 +788,7 @@ bool dce110_link_encoder_validate_output_with_stream( case SIGNAL_TYPE_DVI_DUAL_LINK: is_valid = dce110_link_encoder_validate_dvi_output( enc110, - stream->sink->link->connector_signal, + stream->link->connector_signal, stream->signal, &stream->timing); break; diff --git a/drivers/gpu/drm/amd/display/dc/dce/dce_mem_input.c b/drivers/gpu/drm/amd/display/dc/dce/dce_mem_input.c index 85686d917636..a24a2bda8656 100644 --- a/drivers/gpu/drm/amd/display/dc/dce/dce_mem_input.c +++ b/drivers/gpu/drm/amd/display/dc/dce/dce_mem_input.c @@ -479,7 +479,7 @@ static void program_grph_pixel_format( case SURFACE_PIXEL_FORMAT_GRPH_ABGR16161616F: sign = 1; floating = 1; - /* no break */ + /* fall through */ case SURFACE_PIXEL_FORMAT_GRPH_ARGB16161616F: /* shouldn't this get float too? */ case SURFACE_PIXEL_FORMAT_GRPH_ARGB16161616: grph_depth = 3; diff --git a/drivers/gpu/drm/amd/display/dc/dce/dce_stream_encoder.c b/drivers/gpu/drm/amd/display/dc/dce/dce_stream_encoder.c index c47c81883d3c..1fa2d4fd7a35 100644 --- a/drivers/gpu/drm/amd/display/dc/dce/dce_stream_encoder.c +++ b/drivers/gpu/drm/amd/display/dc/dce/dce_stream_encoder.c @@ -288,9 +288,18 @@ static void dce110_stream_encoder_dp_set_stream_attribute( #endif struct dce110_stream_encoder *enc110 = DCE110STRENC_FROM_STRENC(enc); - + struct dc_crtc_timing hw_crtc_timing = *crtc_timing; + if (hw_crtc_timing.flags.INTERLACE) { + /*the input timing is in VESA spec format with Interlace flag =1*/ + hw_crtc_timing.v_total /= 2; + hw_crtc_timing.v_border_top /= 2; + hw_crtc_timing.v_addressable /= 2; + hw_crtc_timing.v_border_bottom /= 2; + hw_crtc_timing.v_front_porch /= 2; + hw_crtc_timing.v_sync_width /= 2; + } /* set pixel encoding */ - switch (crtc_timing->pixel_encoding) { + switch (hw_crtc_timing.pixel_encoding) { case PIXEL_ENCODING_YCBCR422: REG_UPDATE(DP_PIXEL_FORMAT, DP_PIXEL_ENCODING, DP_PIXEL_ENCODING_TYPE_YCBCR422); @@ -299,8 +308,8 @@ static void dce110_stream_encoder_dp_set_stream_attribute( REG_UPDATE(DP_PIXEL_FORMAT, DP_PIXEL_ENCODING, DP_PIXEL_ENCODING_TYPE_YCBCR444); - if (crtc_timing->flags.Y_ONLY) - if (crtc_timing->display_color_depth != COLOR_DEPTH_666) + if (hw_crtc_timing.flags.Y_ONLY) + if (hw_crtc_timing.display_color_depth != COLOR_DEPTH_666) /* HW testing only, no use case yet. * Color depth of Y-only could be * 8, 10, 12, 16 bits */ @@ -335,7 +344,7 @@ static void dce110_stream_encoder_dp_set_stream_attribute( /* set color depth */ - switch (crtc_timing->display_color_depth) { + switch (hw_crtc_timing.display_color_depth) { case COLOR_DEPTH_666: REG_UPDATE(DP_PIXEL_FORMAT, DP_COMPONENT_DEPTH, 0); @@ -363,7 +372,7 @@ static void dce110_stream_encoder_dp_set_stream_attribute( #if defined(CONFIG_DRM_AMD_DC_DCN1_0) - switch (crtc_timing->display_color_depth) { + switch (hw_crtc_timing.display_color_depth) { case COLOR_DEPTH_666: colorimetry_bpc = 0; break; @@ -401,9 +410,9 @@ static void dce110_stream_encoder_dp_set_stream_attribute( misc0 = misc0 | 0x8; /* bit3=1, bit4=0 */ misc1 = misc1 & ~0x80; /* bit7 = 0*/ dynamic_range_ycbcr = 0; /*bt601*/ - if (crtc_timing->pixel_encoding == PIXEL_ENCODING_YCBCR422) + if (hw_crtc_timing.pixel_encoding == PIXEL_ENCODING_YCBCR422) misc0 = misc0 | 0x2; /* bit2=0, bit1=1 */ - else if (crtc_timing->pixel_encoding == PIXEL_ENCODING_YCBCR444) + else if (hw_crtc_timing.pixel_encoding == PIXEL_ENCODING_YCBCR444) misc0 = misc0 | 0x4; /* bit2=1, bit1=0 */ break; case COLOR_SPACE_YCBCR709: @@ -411,9 +420,9 @@ static void dce110_stream_encoder_dp_set_stream_attribute( misc0 = misc0 | 0x18; /* bit3=1, bit4=1 */ misc1 = misc1 & ~0x80; /* bit7 = 0*/ dynamic_range_ycbcr = 1; /*bt709*/ - if (crtc_timing->pixel_encoding == PIXEL_ENCODING_YCBCR422) + if (hw_crtc_timing.pixel_encoding == PIXEL_ENCODING_YCBCR422) misc0 = misc0 | 0x2; /* bit2=0, bit1=1 */ - else if (crtc_timing->pixel_encoding == PIXEL_ENCODING_YCBCR444) + else if (hw_crtc_timing.pixel_encoding == PIXEL_ENCODING_YCBCR444) misc0 = misc0 | 0x4; /* bit2=1, bit1=0 */ break; case COLOR_SPACE_2020_RGB_LIMITEDRANGE: @@ -453,27 +462,27 @@ static void dce110_stream_encoder_dp_set_stream_attribute( */ if (REG(DP_MSA_TIMING_PARAM1)) REG_SET_2(DP_MSA_TIMING_PARAM1, 0, - DP_MSA_HTOTAL, crtc_timing->h_total, - DP_MSA_VTOTAL, crtc_timing->v_total); + DP_MSA_HTOTAL, hw_crtc_timing.h_total, + DP_MSA_VTOTAL, hw_crtc_timing.v_total); #endif /* calcuate from vesa timing parameters * h_active_start related to leading edge of sync */ - h_blank = crtc_timing->h_total - crtc_timing->h_border_left - - crtc_timing->h_addressable - crtc_timing->h_border_right; + h_blank = hw_crtc_timing.h_total - hw_crtc_timing.h_border_left - + hw_crtc_timing.h_addressable - hw_crtc_timing.h_border_right; - h_back_porch = h_blank - crtc_timing->h_front_porch - - crtc_timing->h_sync_width; + h_back_porch = h_blank - hw_crtc_timing.h_front_porch - + hw_crtc_timing.h_sync_width; /* start at begining of left border */ - h_active_start = crtc_timing->h_sync_width + h_back_porch; + h_active_start = hw_crtc_timing.h_sync_width + h_back_porch; - v_active_start = crtc_timing->v_total - crtc_timing->v_border_top - - crtc_timing->v_addressable - crtc_timing->v_border_bottom - - crtc_timing->v_front_porch; + v_active_start = hw_crtc_timing.v_total - hw_crtc_timing.v_border_top - + hw_crtc_timing.v_addressable - hw_crtc_timing.v_border_bottom - + hw_crtc_timing.v_front_porch; #if defined(CONFIG_DRM_AMD_DC_DCN1_0) @@ -486,21 +495,21 @@ static void dce110_stream_encoder_dp_set_stream_attribute( if (REG(DP_MSA_TIMING_PARAM3)) REG_SET_4(DP_MSA_TIMING_PARAM3, 0, DP_MSA_HSYNCWIDTH, - crtc_timing->h_sync_width, + hw_crtc_timing.h_sync_width, DP_MSA_HSYNCPOLARITY, - !crtc_timing->flags.HSYNC_POSITIVE_POLARITY, + !hw_crtc_timing.flags.HSYNC_POSITIVE_POLARITY, DP_MSA_VSYNCWIDTH, - crtc_timing->v_sync_width, + hw_crtc_timing.v_sync_width, DP_MSA_VSYNCPOLARITY, - !crtc_timing->flags.VSYNC_POSITIVE_POLARITY); + !hw_crtc_timing.flags.VSYNC_POSITIVE_POLARITY); /* HWDITH include border or overscan */ if (REG(DP_MSA_TIMING_PARAM4)) REG_SET_2(DP_MSA_TIMING_PARAM4, 0, - DP_MSA_HWIDTH, crtc_timing->h_border_left + - crtc_timing->h_addressable + crtc_timing->h_border_right, - DP_MSA_VHEIGHT, crtc_timing->v_border_top + - crtc_timing->v_addressable + crtc_timing->v_border_bottom); + DP_MSA_HWIDTH, hw_crtc_timing.h_border_left + + hw_crtc_timing.h_addressable + hw_crtc_timing.h_border_right, + DP_MSA_VHEIGHT, hw_crtc_timing.v_border_top + + hw_crtc_timing.v_addressable + hw_crtc_timing.v_border_bottom); #endif } #endif @@ -662,7 +671,7 @@ static void dce110_stream_encoder_dvi_set_stream_attribute( cntl.signal = is_dual_link ? SIGNAL_TYPE_DVI_DUAL_LINK : SIGNAL_TYPE_DVI_SINGLE_LINK; cntl.enable_dp_audio = false; - cntl.pixel_clock = crtc_timing->pix_clk_khz; + cntl.pixel_clock = crtc_timing->pix_clk_100hz / 10; cntl.lanes_number = (is_dual_link) ? LANE_COUNT_EIGHT : LANE_COUNT_FOUR; if (enc110->base.bp->funcs->encoder_control( @@ -686,7 +695,7 @@ static void dce110_stream_encoder_lvds_set_stream_attribute( cntl.engine_id = enc110->base.id; cntl.signal = SIGNAL_TYPE_LVDS; cntl.enable_dp_audio = false; - cntl.pixel_clock = crtc_timing->pix_clk_khz; + cntl.pixel_clock = crtc_timing->pix_clk_100hz / 10; cntl.lanes_number = LANE_COUNT_FOUR; if (enc110->base.bp->funcs->encoder_control( @@ -908,7 +917,6 @@ static void dce110_stream_encoder_dp_blank( struct stream_encoder *enc) { struct dce110_stream_encoder *enc110 = DCE110STRENC_FROM_STRENC(enc); - uint32_t retries = 0; uint32_t reg1 = 0; uint32_t max_retries = DP_BLANK_MAX_RETRY * 10; @@ -926,30 +934,28 @@ static void dce110_stream_encoder_dp_blank( * (2 = start of the next vertical blank) */ REG_UPDATE(DP_VID_STREAM_CNTL, DP_VID_STREAM_DIS_DEFER, 2); /* Larger delay to wait until VBLANK - use max retry of - * 10us*3000=30ms. This covers 16.6ms of typical 60 Hz mode + - * a little more because we may not trust delay accuracy. - */ + * 10us*3000=30ms. This covers 16.6ms of typical 60 Hz mode + + * a little more because we may not trust delay accuracy. + */ max_retries = DP_BLANK_MAX_RETRY * 150; /* disable DP stream */ REG_UPDATE(DP_VID_STREAM_CNTL, DP_VID_STREAM_ENABLE, 0); /* the encoder stops sending the video stream - * at the start of the vertical blanking. - * Poll for DP_VID_STREAM_STATUS == 0 - */ + * at the start of the vertical blanking. + * Poll for DP_VID_STREAM_STATUS == 0 + */ REG_WAIT(DP_VID_STREAM_CNTL, DP_VID_STREAM_STATUS, 0, 10, max_retries); - ASSERT(retries <= max_retries); - /* Tell the DP encoder to ignore timing from CRTC, must be done after - * the polling. If we set DP_STEER_FIFO_RESET before DP stream blank is - * complete, stream status will be stuck in video stream enabled state, - * i.e. DP_VID_STREAM_STATUS stuck at 1. - */ + * the polling. If we set DP_STEER_FIFO_RESET before DP stream blank is + * complete, stream status will be stuck in video stream enabled state, + * i.e. DP_VID_STREAM_STATUS stuck at 1. + */ REG_UPDATE(DP_STEER_FIFO, DP_STEER_FIFO_RESET, true); } @@ -1578,6 +1584,14 @@ static void setup_stereo_sync( REG_UPDATE(DIG_FE_CNTL, DIG_STEREOSYNC_GATE_EN, !enable); } +static void dig_connect_to_otg( + struct stream_encoder *enc, + int tg_inst) +{ + struct dce110_stream_encoder *enc110 = DCE110STRENC_FROM_STRENC(enc); + + REG_UPDATE(DIG_FE_CNTL, DIG_SOURCE_SELECT, tg_inst); +} static const struct stream_encoder_funcs dce110_str_enc_funcs = { .dp_set_stream_attribute = @@ -1612,7 +1626,7 @@ static const struct stream_encoder_funcs dce110_str_enc_funcs = { .hdmi_audio_disable = dce110_se_hdmi_audio_disable, .setup_stereo_sync = setup_stereo_sync, .set_avmute = dce110_stream_encoder_set_avmute, - + .dig_connect_to_otg = dig_connect_to_otg, }; void dce110_stream_encoder_construct( diff --git a/drivers/gpu/drm/amd/display/dc/dce/dce_stream_encoder.h b/drivers/gpu/drm/amd/display/dc/dce/dce_stream_encoder.h index 6c28229c76eb..f9cdf2b5242c 100644 --- a/drivers/gpu/drm/amd/display/dc/dce/dce_stream_encoder.h +++ b/drivers/gpu/drm/amd/display/dc/dce/dce_stream_encoder.h @@ -199,7 +199,8 @@ SE_SF(DP_SEC_CNTL, DP_SEC_ATP_ENABLE, mask_sh),\ SE_SF(DP_SEC_CNTL, DP_SEC_AIP_ENABLE, mask_sh),\ SE_SF(DP_SEC_CNTL, DP_SEC_ACM_ENABLE, mask_sh),\ - SE_SF(AFMT_AUDIO_PACKET_CONTROL, AFMT_AUDIO_SAMPLE_SEND, mask_sh) + SE_SF(AFMT_AUDIO_PACKET_CONTROL, AFMT_AUDIO_SAMPLE_SEND, mask_sh),\ + SE_SF(DIG_FE_CNTL, DIG_SOURCE_SELECT, mask_sh) #define SE_COMMON_MASK_SH_LIST_DCE_COMMON(mask_sh)\ SE_COMMON_MASK_SH_LIST_DCE_COMMON_BASE(mask_sh) @@ -284,7 +285,8 @@ SE_SF(DIG0_DIG_FE_CNTL, TMDS_PIXEL_ENCODING, mask_sh),\ SE_SF(DIG0_DIG_FE_CNTL, TMDS_COLOR_FORMAT, mask_sh),\ SE_SF(DIG0_DIG_FE_CNTL, DIG_STEREOSYNC_SELECT, mask_sh),\ - SE_SF(DIG0_DIG_FE_CNTL, DIG_STEREOSYNC_GATE_EN, mask_sh) + SE_SF(DIG0_DIG_FE_CNTL, DIG_STEREOSYNC_GATE_EN, mask_sh),\ + SE_SF(DIG0_DIG_FE_CNTL, DIG_SOURCE_SELECT, mask_sh) #define SE_COMMON_MASK_SH_LIST_SOC(mask_sh)\ SE_COMMON_MASK_SH_LIST_SOC_BASE(mask_sh) @@ -494,6 +496,7 @@ struct dce_stream_encoder_shift { uint8_t HDMI_DB_DISABLE; uint8_t DP_VID_N_MUL; uint8_t DP_VID_M_DOUBLE_VALUE_EN; + uint8_t DIG_SOURCE_SELECT; }; struct dce_stream_encoder_mask { @@ -624,6 +627,7 @@ struct dce_stream_encoder_mask { uint32_t HDMI_DB_DISABLE; uint32_t DP_VID_N_MUL; uint32_t DP_VID_M_DOUBLE_VALUE_EN; + uint32_t DIG_SOURCE_SELECT; }; struct dce110_stream_enc_registers { diff --git a/drivers/gpu/drm/amd/display/dc/dce100/dce100_hw_sequencer.c b/drivers/gpu/drm/amd/display/dc/dce100/dce100_hw_sequencer.c index 74c05e878807..87771676acac 100644 --- a/drivers/gpu/drm/amd/display/dc/dce100/dce100_hw_sequencer.c +++ b/drivers/gpu/drm/amd/display/dc/dce100/dce100_hw_sequencer.c @@ -105,74 +105,30 @@ bool dce100_enable_display_power_gating( return false; } -static void dce100_pplib_apply_display_requirements( - struct dc *dc, - struct dc_state *context) -{ - struct dm_pp_display_configuration *pp_display_cfg = &context->pp_display_cfg; - - pp_display_cfg->avail_mclk_switch_time_us = - dce110_get_min_vblank_time_us(context); - /*pp_display_cfg->min_memory_clock_khz = context->bw.dce.yclk_khz - / MEMORY_TYPE_MULTIPLIER;*/ - - dce110_fill_display_configs(context, pp_display_cfg); - - if (memcmp(&dc->prev_display_config, pp_display_cfg, sizeof( - struct dm_pp_display_configuration)) != 0) - dm_pp_apply_display_requirements(dc->ctx, pp_display_cfg); - - dc->prev_display_config = *pp_display_cfg; -} - -/* unit: in_khz before mode set, get pixel clock from context. ASIC register - * may not be programmed yet - */ -static uint32_t get_max_pixel_clock_for_all_paths( - struct dc *dc, - struct dc_state *context) +void dce100_prepare_bandwidth( + struct dc *dc, + struct dc_state *context) { - uint32_t max_pix_clk = 0; - int i; - - for (i = 0; i < MAX_PIPES; i++) { - struct pipe_ctx *pipe_ctx = &context->res_ctx.pipe_ctx[i]; - - if (pipe_ctx->stream == NULL) - continue; - - /* do not check under lay */ - if (pipe_ctx->top_pipe) - continue; + dce110_set_safe_displaymarks(&context->res_ctx, dc->res_pool); - if (pipe_ctx->stream_res.pix_clk_params.requested_pix_clk > max_pix_clk) - max_pix_clk = - pipe_ctx->stream_res.pix_clk_params.requested_pix_clk; - } - return max_pix_clk; + dc->res_pool->clk_mgr->funcs->update_clocks( + dc->res_pool->clk_mgr, + context, + false); } -void dce100_set_bandwidth( +void dce100_optimize_bandwidth( struct dc *dc, - struct dc_state *context, - bool decrease_allowed) + struct dc_state *context) { - struct dc_clocks req_clks; - - req_clks.dispclk_khz = context->bw.dce.dispclk_khz * 115 / 100; - req_clks.phyclk_khz = get_max_pixel_clock_for_all_paths(dc, context); - dce110_set_safe_displaymarks(&context->res_ctx, dc->res_pool); - dc->res_pool->dccg->funcs->update_clocks( - dc->res_pool->dccg, - &req_clks, - decrease_allowed); - - dce100_pplib_apply_display_requirements(dc, context); + dc->res_pool->clk_mgr->funcs->update_clocks( + dc->res_pool->clk_mgr, + context, + true); } - /**************************************************************************/ void dce100_hw_sequencer_construct(struct dc *dc) @@ -180,8 +136,7 @@ void dce100_hw_sequencer_construct(struct dc *dc) dce110_hw_sequencer_construct(dc); dc->hwss.enable_display_power_gating = dce100_enable_display_power_gating; - dc->hwss.set_bandwidth = dce100_set_bandwidth; - dc->hwss.pplib_apply_display_requirements = - dce100_pplib_apply_display_requirements; + dc->hwss.prepare_bandwidth = dce100_prepare_bandwidth; + dc->hwss.optimize_bandwidth = dce100_optimize_bandwidth; } diff --git a/drivers/gpu/drm/amd/display/dc/dce100/dce100_hw_sequencer.h b/drivers/gpu/drm/amd/display/dc/dce100/dce100_hw_sequencer.h index c6ec0ed6ec3d..a6b80fdaa666 100644 --- a/drivers/gpu/drm/amd/display/dc/dce100/dce100_hw_sequencer.h +++ b/drivers/gpu/drm/amd/display/dc/dce100/dce100_hw_sequencer.h @@ -33,10 +33,13 @@ struct dc_state; void dce100_hw_sequencer_construct(struct dc *dc); -void dce100_set_bandwidth( +void dce100_prepare_bandwidth( struct dc *dc, - struct dc_state *context, - bool decrease_allowed); + struct dc_state *context); + +void dce100_optimize_bandwidth( + struct dc *dc, + struct dc_state *context); bool dce100_enable_display_power_gating(struct dc *dc, uint8_t controller_id, struct dc_bios *dcb, diff --git a/drivers/gpu/drm/amd/display/dc/dce100/dce100_resource.c b/drivers/gpu/drm/amd/display/dc/dce100/dce100_resource.c index 14754a87156c..23044e6723e8 100644 --- a/drivers/gpu/drm/amd/display/dc/dce100/dce100_resource.c +++ b/drivers/gpu/drm/amd/display/dc/dce100/dce100_resource.c @@ -36,11 +36,11 @@ #include "dce/dce_link_encoder.h" #include "dce/dce_stream_encoder.h" +#include "dce/dce_clk_mgr.h" #include "dce/dce_mem_input.h" #include "dce/dce_ipp.h" #include "dce/dce_transform.h" #include "dce/dce_opp.h" -#include "dce/dce_clocks.h" #include "dce/dce_clock_source.h" #include "dce/dce_audio.h" #include "dce/dce_hwseq.h" @@ -76,6 +76,7 @@ #ifndef mmBIOS_SCRATCH_2 #define mmBIOS_SCRATCH_2 0x05CB + #define mmBIOS_SCRATCH_3 0x05CC #define mmBIOS_SCRATCH_6 0x05CF #endif @@ -137,15 +138,15 @@ static const struct dce110_timing_generator_offsets dce100_tg_offsets[] = { .reg_name = mm ## block ## id ## _ ## reg_name -static const struct dccg_registers disp_clk_regs = { +static const struct clk_mgr_registers disp_clk_regs = { CLK_COMMON_REG_LIST_DCE_BASE() }; -static const struct dccg_shift disp_clk_shift = { +static const struct clk_mgr_shift disp_clk_shift = { CLK_COMMON_MASK_SH_LIST_DCE_COMMON_BASE(__SHIFT) }; -static const struct dccg_mask disp_clk_mask = { +static const struct clk_mgr_mask disp_clk_mask = { CLK_COMMON_MASK_SH_LIST_DCE_COMMON_BASE(_MASK) }; @@ -365,6 +366,7 @@ static const struct dce_abm_mask abm_mask = { #define DCFE_MEM_PWR_CTRL_REG_BASE 0x1b03 static const struct bios_registers bios_regs = { + .BIOS_SCRATCH_3 = mmBIOS_SCRATCH_3, .BIOS_SCRATCH_6 = mmBIOS_SCRATCH_6 }; @@ -587,7 +589,7 @@ struct output_pixel_processor *dce100_opp_create( return &opp->base; } -struct aux_engine *dce100_aux_engine_create( +struct dce_aux *dce100_aux_engine_create( struct dc_context *ctx, uint32_t inst) { @@ -722,8 +724,8 @@ static void destruct(struct dce110_resource_pool *pool) dce_aud_destroy(&pool->base.audios[i]); } - if (pool->base.dccg != NULL) - dce_dccg_destroy(&pool->base.dccg); + if (pool->base.clk_mgr != NULL) + dce_clk_mgr_destroy(&pool->base.clk_mgr); if (pool->base.abm != NULL) dce_abm_destroy(&pool->base.abm); @@ -767,7 +769,7 @@ bool dce100_validate_bandwidth( if (at_least_one_pipe) { /* TODO implement when needed but for now hardcode max value*/ context->bw.dce.dispclk_khz = 681000; - context->bw.dce.yclk_khz = 250000 * MEMORY_TYPE_MULTIPLIER; + context->bw.dce.yclk_khz = 250000 * MEMORY_TYPE_MULTIPLIER_CZ; } else { context->bw.dce.dispclk_khz = 0; context->bw.dce.yclk_khz = 0; @@ -860,7 +862,6 @@ static bool construct( struct dc_context *ctx = dc->ctx; struct dc_firmware_info info; struct dc_bios *bp; - struct dm_pp_static_clock_info static_clk_info = {0}; ctx->dc_bios->regs = &bios_regs; @@ -908,11 +909,11 @@ static bool construct( } } - pool->base.dccg = dce_dccg_create(ctx, + pool->base.clk_mgr = dce_clk_mgr_create(ctx, &disp_clk_regs, &disp_clk_shift, &disp_clk_mask); - if (pool->base.dccg == NULL) { + if (pool->base.clk_mgr == NULL) { dm_error("DC: failed to create display clock!\n"); BREAK_TO_DEBUGGER(); goto res_create_fail; @@ -938,12 +939,6 @@ static bool construct( goto res_create_fail; } - /* get static clock information for PPLIB or firmware, save - * max_clock_state - */ - if (dm_pp_get_static_clocks(ctx, &static_clk_info)) - pool->base.dccg->max_clks_state = - static_clk_info.max_clocks_state; { struct irq_service_init_data init_data; init_data.ctx = dc->ctx; diff --git a/drivers/gpu/drm/amd/display/dc/dce110/dce110_compressor.c b/drivers/gpu/drm/amd/display/dc/dce110/dce110_compressor.c index 1f7f25013217..7b23239d33fe 100644 --- a/drivers/gpu/drm/amd/display/dc/dce110/dce110_compressor.c +++ b/drivers/gpu/drm/amd/display/dc/dce110/dce110_compressor.c @@ -62,67 +62,37 @@ static const struct dce110_compressor_reg_offsets reg_offsets[] = { } }; -static const uint32_t dce11_one_lpt_channel_max_resolution = 2560 * 1600; - -enum fbc_idle_force { - /* Bit 0 - Display registers updated */ - FBC_IDLE_FORCE_DISPLAY_REGISTER_UPDATE = 0x00000001, - - /* Bit 2 - FBC_GRPH_COMP_EN register updated */ - FBC_IDLE_FORCE_GRPH_COMP_EN = 0x00000002, - /* Bit 3 - FBC_SRC_SEL register updated */ - FBC_IDLE_FORCE_SRC_SEL_CHANGE = 0x00000004, - /* Bit 4 - FBC_MIN_COMPRESSION register updated */ - FBC_IDLE_FORCE_MIN_COMPRESSION_CHANGE = 0x00000008, - /* Bit 5 - FBC_ALPHA_COMP_EN register updated */ - FBC_IDLE_FORCE_ALPHA_COMP_EN = 0x00000010, - /* Bit 6 - FBC_ZERO_ALPHA_CHUNK_SKIP_EN register updated */ - FBC_IDLE_FORCE_ZERO_ALPHA_CHUNK_SKIP_EN = 0x00000020, - /* Bit 7 - FBC_FORCE_COPY_TO_COMP_BUF register updated */ - FBC_IDLE_FORCE_FORCE_COPY_TO_COMP_BUF = 0x00000040, - - /* Bit 24 - Memory write to region 0 defined by MC registers. */ - FBC_IDLE_FORCE_MEMORY_WRITE_TO_REGION0 = 0x01000000, - /* Bit 25 - Memory write to region 1 defined by MC registers */ - FBC_IDLE_FORCE_MEMORY_WRITE_TO_REGION1 = 0x02000000, - /* Bit 26 - Memory write to region 2 defined by MC registers */ - FBC_IDLE_FORCE_MEMORY_WRITE_TO_REGION2 = 0x04000000, - /* Bit 27 - Memory write to region 3 defined by MC registers. */ - FBC_IDLE_FORCE_MEMORY_WRITE_TO_REGION3 = 0x08000000, - - /* Bit 28 - Memory write from any client other than MCIF */ - FBC_IDLE_FORCE_MEMORY_WRITE_OTHER_THAN_MCIF = 0x10000000, - /* Bit 29 - CG statics screen signal is inactive */ - FBC_IDLE_FORCE_CG_STATIC_SCREEN_IS_INACTIVE = 0x20000000, -}; - - static uint32_t align_to_chunks_number_per_line(uint32_t pixels) { return 256 * ((pixels + 255) / 256); } -static void reset_lb_on_vblank(struct dc_context *ctx) +static void reset_lb_on_vblank(struct compressor *compressor, uint32_t crtc_inst) { - uint32_t value, frame_count; + uint32_t value; + uint32_t frame_count; + uint32_t status_pos; uint32_t retry = 0; - uint32_t status_pos = - dm_read_reg(ctx, mmCRTC_STATUS_POSITION); + struct dce110_compressor *cp110 = TO_DCE110_COMPRESSOR(compressor); + + cp110->offsets = reg_offsets[crtc_inst]; + + status_pos = dm_read_reg(compressor->ctx, DCP_REG(mmCRTC_STATUS_POSITION)); /* Only if CRTC is enabled and counter is moving we wait for one frame. */ - if (status_pos != dm_read_reg(ctx, mmCRTC_STATUS_POSITION)) { + if (status_pos != dm_read_reg(compressor->ctx, DCP_REG(mmCRTC_STATUS_POSITION))) { /* Resetting LB on VBlank */ - value = dm_read_reg(ctx, mmLB_SYNC_RESET_SEL); + value = dm_read_reg(compressor->ctx, DCP_REG(mmLB_SYNC_RESET_SEL)); set_reg_field_value(value, 3, LB_SYNC_RESET_SEL, LB_SYNC_RESET_SEL); set_reg_field_value(value, 1, LB_SYNC_RESET_SEL, LB_SYNC_RESET_SEL2); - dm_write_reg(ctx, mmLB_SYNC_RESET_SEL, value); + dm_write_reg(compressor->ctx, DCP_REG(mmLB_SYNC_RESET_SEL), value); - frame_count = dm_read_reg(ctx, mmCRTC_STATUS_FRAME_COUNT); + frame_count = dm_read_reg(compressor->ctx, DCP_REG(mmCRTC_STATUS_FRAME_COUNT)); for (retry = 10000; retry > 0; retry--) { - if (frame_count != dm_read_reg(ctx, mmCRTC_STATUS_FRAME_COUNT)) + if (frame_count != dm_read_reg(compressor->ctx, DCP_REG(mmCRTC_STATUS_FRAME_COUNT))) break; udelay(10); } @@ -130,13 +100,11 @@ static void reset_lb_on_vblank(struct dc_context *ctx) dm_error("Frame count did not increase for 100ms.\n"); /* Resetting LB on VBlank */ - value = dm_read_reg(ctx, mmLB_SYNC_RESET_SEL); + value = dm_read_reg(compressor->ctx, DCP_REG(mmLB_SYNC_RESET_SEL)); set_reg_field_value(value, 2, LB_SYNC_RESET_SEL, LB_SYNC_RESET_SEL); set_reg_field_value(value, 0, LB_SYNC_RESET_SEL, LB_SYNC_RESET_SEL2); - dm_write_reg(ctx, mmLB_SYNC_RESET_SEL, value); - + dm_write_reg(compressor->ctx, DCP_REG(mmLB_SYNC_RESET_SEL), value); } - } static void wait_for_fbc_state_changed( @@ -226,10 +194,10 @@ void dce110_compressor_enable_fbc( uint32_t addr; uint32_t value, misc_value; - addr = mmFBC_CNTL; value = dm_read_reg(compressor->ctx, addr); set_reg_field_value(value, 1, FBC_CNTL, FBC_GRPH_COMP_EN); + /* params->inst is valid HW CRTC instance start from 0 */ set_reg_field_value( value, params->inst, @@ -238,8 +206,10 @@ void dce110_compressor_enable_fbc( /* Keep track of enum controller_id FBC is attached to */ compressor->is_enabled = true; - compressor->attached_inst = params->inst; - cp110->offsets = reg_offsets[params->inst]; + /* attached_inst is SW CRTC instance start from 1 + * 0 = CONTROLLER_ID_UNDEFINED means not attached crtc + */ + compressor->attached_inst = params->inst + CONTROLLER_ID_D0; /* Toggle it as there is bug in HW */ set_reg_field_value(value, 0, FBC_CNTL, FBC_GRPH_COMP_EN); @@ -268,9 +238,10 @@ void dce110_compressor_enable_fbc( void dce110_compressor_disable_fbc(struct compressor *compressor) { struct dce110_compressor *cp110 = TO_DCE110_COMPRESSOR(compressor); + uint32_t crtc_inst = 0; if (compressor->options.bits.FBC_SUPPORT) { - if (dce110_compressor_is_fbc_enabled_in_hw(compressor, NULL)) { + if (dce110_compressor_is_fbc_enabled_in_hw(compressor, &crtc_inst)) { uint32_t reg_data; /* Turn off compression */ reg_data = dm_read_reg(compressor->ctx, mmFBC_CNTL); @@ -284,8 +255,10 @@ void dce110_compressor_disable_fbc(struct compressor *compressor) wait_for_fbc_state_changed(cp110, false); } - /* Sync line buffer - dce100/110 only*/ - reset_lb_on_vblank(compressor->ctx); + /* Sync line buffer which fbc was attached to dce100/110 only */ + if (crtc_inst > CONTROLLER_ID_UNDEFINED && crtc_inst < CONTROLLER_ID_D3) + reset_lb_on_vblank(compressor, + crtc_inst - CONTROLLER_ID_D0); } } @@ -328,6 +301,8 @@ void dce110_compressor_program_compressed_surface_address_and_pitch( uint32_t compressed_surf_address_low_part = compressor->compr_surface_address.addr.low_part; + cp110->offsets = reg_offsets[params->inst]; + /* Clear content first. */ dm_write_reg( compressor->ctx, @@ -410,13 +385,7 @@ void dce110_compressor_set_fbc_invalidation_triggers( value = dm_read_reg(compressor->ctx, addr); set_reg_field_value( value, - fbc_trigger | - FBC_IDLE_FORCE_GRPH_COMP_EN | - FBC_IDLE_FORCE_SRC_SEL_CHANGE | - FBC_IDLE_FORCE_MIN_COMPRESSION_CHANGE | - FBC_IDLE_FORCE_ALPHA_COMP_EN | - FBC_IDLE_FORCE_ZERO_ALPHA_CHUNK_SKIP_EN | - FBC_IDLE_FORCE_FORCE_COPY_TO_COMP_BUF, + fbc_trigger, FBC_IDLE_FORCE_CLEAR_MASK, FBC_IDLE_FORCE_CLEAR_MASK); dm_write_reg(compressor->ctx, addr, value); @@ -549,7 +518,7 @@ void dce110_compressor_construct(struct dce110_compressor *compressor, compressor->base.channel_interleave_size = 0; compressor->base.dram_channels_num = 0; compressor->base.lpt_channels_num = 0; - compressor->base.attached_inst = 0; + compressor->base.attached_inst = CONTROLLER_ID_UNDEFINED; compressor->base.is_enabled = false; compressor->base.funcs = &dce110_compressor_funcs; diff --git a/drivers/gpu/drm/amd/display/dc/dce110/dce110_hw_sequencer.c b/drivers/gpu/drm/amd/display/dc/dce110/dce110_hw_sequencer.c index a6bcb90e8419..5e4db3712eef 100644 --- a/drivers/gpu/drm/amd/display/dc/dce110/dce110_hw_sequencer.c +++ b/drivers/gpu/drm/amd/display/dc/dce110/dce110_hw_sequencer.c @@ -548,14 +548,14 @@ dce110_translate_regamma_to_hw_format(const struct dc_transfer_func *output_tf, regamma_params->hw_points_num = hw_points; - i = 1; - for (k = 0; k < 16 && i < 16; k++) { + k = 0; + for (i = 1; i < 16; i++) { if (seg_distr[k] != -1) { regamma_params->arr_curve_points[k].segments_num = seg_distr[k]; regamma_params->arr_curve_points[i].offset = regamma_params->arr_curve_points[k].offset + (1 << seg_distr[k]); } - i++; + k++; } if (seg_distr[k] != -1) @@ -614,55 +614,6 @@ dce110_set_output_transfer_func(struct pipe_ctx *pipe_ctx, return true; } -static enum dc_status bios_parser_crtc_source_select( - struct pipe_ctx *pipe_ctx) -{ - struct dc_bios *dcb; - /* call VBIOS table to set CRTC source for the HW - * encoder block - * note: video bios clears all FMT setting here. */ - struct bp_crtc_source_select crtc_source_select = {0}; - const struct dc_sink *sink = pipe_ctx->stream->sink; - - crtc_source_select.engine_id = pipe_ctx->stream_res.stream_enc->id; - crtc_source_select.controller_id = pipe_ctx->stream_res.tg->inst + 1; - /*TODO: Need to un-hardcode color depth, dp_audio and account for - * the case where signal and sink signal is different (translator - * encoder)*/ - crtc_source_select.signal = pipe_ctx->stream->signal; - crtc_source_select.enable_dp_audio = false; - crtc_source_select.sink_signal = pipe_ctx->stream->signal; - - switch (pipe_ctx->stream->timing.display_color_depth) { - case COLOR_DEPTH_666: - crtc_source_select.display_output_bit_depth = PANEL_6BIT_COLOR; - break; - case COLOR_DEPTH_888: - crtc_source_select.display_output_bit_depth = PANEL_8BIT_COLOR; - break; - case COLOR_DEPTH_101010: - crtc_source_select.display_output_bit_depth = PANEL_10BIT_COLOR; - break; - case COLOR_DEPTH_121212: - crtc_source_select.display_output_bit_depth = PANEL_12BIT_COLOR; - break; - default: - BREAK_TO_DEBUGGER(); - crtc_source_select.display_output_bit_depth = PANEL_8BIT_COLOR; - break; - } - - dcb = sink->ctx->dc_bios; - - if (BP_RESULT_OK != dcb->funcs->crtc_source_select( - dcb, - &crtc_source_select)) { - return DC_ERROR_UNEXPECTED; - } - - return DC_OK; -} - void dce110_update_info_frame(struct pipe_ctx *pipe_ctx) { bool is_hdmi; @@ -692,10 +643,10 @@ void dce110_update_info_frame(struct pipe_ctx *pipe_ctx) void dce110_enable_stream(struct pipe_ctx *pipe_ctx) { enum dc_lane_count lane_count = - pipe_ctx->stream->sink->link->cur_link_settings.lane_count; + pipe_ctx->stream->link->cur_link_settings.lane_count; struct dc_crtc_timing *timing = &pipe_ctx->stream->timing; - struct dc_link *link = pipe_ctx->stream->sink->link; + struct dc_link *link = pipe_ctx->stream->link; uint32_t active_total_with_borders; @@ -1000,7 +951,7 @@ void dce110_enable_audio_stream(struct pipe_ctx *pipe_ctx) pipe_ctx->stream_res.audio->funcs->az_enable(pipe_ctx->stream_res.audio); - if (num_audio == 1 && pp_smu != NULL && pp_smu->set_pme_wa_enable != NULL) + if (num_audio >= 1 && pp_smu != NULL && pp_smu->set_pme_wa_enable != NULL) /*this is the first audio. apply the PME w/a in order to wake AZ from D3*/ pp_smu->set_pme_wa_enable(&pp_smu->pp_smu); /* un-mute audio */ @@ -1017,6 +968,8 @@ void dce110_disable_audio_stream(struct pipe_ctx *pipe_ctx, int option) pipe_ctx->stream_res.stream_enc->funcs->audio_mute_control( pipe_ctx->stream_res.stream_enc, true); if (pipe_ctx->stream_res.audio) { + struct pp_smu_funcs_rv *pp_smu = dc->res_pool->pp_smu; + if (option != KEEP_ACQUIRED_RESOURCE || !dc->debug.az_endpoint_mute_only) { /*only disalbe az_endpoint if power down or free*/ @@ -1036,6 +989,9 @@ void dce110_disable_audio_stream(struct pipe_ctx *pipe_ctx, int option) update_audio_usage(&dc->current_state->res_ctx, dc->res_pool, pipe_ctx->stream_res.audio, false); pipe_ctx->stream_res.audio = NULL; } + if (pp_smu != NULL && pp_smu->set_pme_wa_enable != NULL) + /*this is the first audio. apply the PME w/a in order to wake AZ from D3*/ + pp_smu->set_pme_wa_enable(&pp_smu->pp_smu); /* TODO: notify audio driver for if audio modes list changed * add audio mode list change flag */ @@ -1048,7 +1004,7 @@ void dce110_disable_audio_stream(struct pipe_ctx *pipe_ctx, int option) void dce110_disable_stream(struct pipe_ctx *pipe_ctx, int option) { struct dc_stream_state *stream = pipe_ctx->stream; - struct dc_link *link = stream->sink->link; + struct dc_link *link = stream->link; struct dc *dc = pipe_ctx->stream->ctx->dc; if (dc_is_hdmi_signal(pipe_ctx->stream->signal)) @@ -1073,11 +1029,10 @@ void dce110_unblank_stream(struct pipe_ctx *pipe_ctx, { struct encoder_unblank_param params = { { 0 } }; struct dc_stream_state *stream = pipe_ctx->stream; - struct dc_link *link = stream->sink->link; + struct dc_link *link = stream->link; /* only 3 items below are used by unblank */ - params.pixel_clk_khz = - pipe_ctx->stream->timing.pix_clk_khz; + params.pixel_clk_khz = pipe_ctx->stream->timing.pix_clk_100hz / 10; params.link_settings.link_rate = link_settings->link_rate; if (dc_is_dp_signal(pipe_ctx->stream->signal)) @@ -1085,13 +1040,13 @@ void dce110_unblank_stream(struct pipe_ctx *pipe_ctx, if (link->local_sink && link->local_sink->sink_signal == SIGNAL_TYPE_EDP) { link->dc->hwss.edp_backlight_control(link, true); - stream->bl_pwm_level = EDP_BACKLIGHT_RAMP_DISABLE_LEVEL; } } + void dce110_blank_stream(struct pipe_ctx *pipe_ctx) { struct dc_stream_state *stream = pipe_ctx->stream; - struct dc_link *link = stream->sink->link; + struct dc_link *link = stream->link; if (link->local_sink && link->local_sink->sink_signal == SIGNAL_TYPE_EDP) { link->dc->hwss.edp_backlight_control(link, false); @@ -1164,27 +1119,27 @@ static void build_audio_output( stream->timing.flags.INTERLACE; audio_output->crtc_info.refresh_rate = - (stream->timing.pix_clk_khz*1000)/ + (stream->timing.pix_clk_100hz*10000)/ (stream->timing.h_total*stream->timing.v_total); audio_output->crtc_info.color_depth = stream->timing.display_color_depth; audio_output->crtc_info.requested_pixel_clock = - pipe_ctx->stream_res.pix_clk_params.requested_pix_clk; + pipe_ctx->stream_res.pix_clk_params.requested_pix_clk_100hz / 10; audio_output->crtc_info.calculated_pixel_clock = - pipe_ctx->stream_res.pix_clk_params.requested_pix_clk; + pipe_ctx->stream_res.pix_clk_params.requested_pix_clk_100hz / 10; /*for HDMI, audio ACR is with deep color ratio factor*/ if (dc_is_hdmi_signal(pipe_ctx->stream->signal) && audio_output->crtc_info.requested_pixel_clock == - stream->timing.pix_clk_khz) { + (stream->timing.pix_clk_100hz / 10)) { if (pipe_ctx->stream_res.pix_clk_params.pixel_encoding == PIXEL_ENCODING_YCBCR420) { audio_output->crtc_info.requested_pixel_clock = audio_output->crtc_info.requested_pixel_clock/2; audio_output->crtc_info.calculated_pixel_clock = - pipe_ctx->stream_res.pix_clk_params.requested_pix_clk/2; + pipe_ctx->stream_res.pix_clk_params.requested_pix_clk_100hz/20; } } @@ -1192,8 +1147,8 @@ static void build_audio_output( if (pipe_ctx->stream->signal == SIGNAL_TYPE_DISPLAY_PORT || pipe_ctx->stream->signal == SIGNAL_TYPE_DISPLAY_PORT_MST) { audio_output->pll_info.dp_dto_source_clock_in_khz = - state->dis_clk->funcs->get_dp_ref_clk_frequency( - state->dis_clk); + state->dccg->funcs->get_dp_ref_clk_frequency( + state->dccg); } audio_output->pll_info.feed_back_divider = @@ -1268,10 +1223,19 @@ static void program_scaler(const struct dc *dc, pipe_ctx->plane_res.scl_data.lb_params.depth, &pipe_ctx->stream->bit_depth_params); - if (pipe_ctx->stream_res.tg->funcs->set_overscan_blank_color) + if (pipe_ctx->stream_res.tg->funcs->set_overscan_blank_color) { + /* + * The way 420 is packed, 2 channels carry Y component, 1 channel + * alternate between Cb and Cr, so both channels need the pixel + * value for Y + */ + if (pipe_ctx->stream->timing.pixel_encoding == PIXEL_ENCODING_YCBCR420) + color.color_r_cr = color.color_g_y; + pipe_ctx->stream_res.tg->funcs->set_overscan_blank_color( pipe_ctx->stream_res.tg, &color); + } pipe_ctx->plane_res.xfm->funcs->transform_set_scaler(pipe_ctx->plane_res.xfm, &pipe_ctx->plane_res.scl_data); @@ -1286,8 +1250,6 @@ static enum dc_status dce110_enable_stream_timing( struct pipe_ctx *pipe_ctx_old = &dc->current_state->res_ctx. pipe_ctx[pipe_ctx->pipe_idx]; struct tg_color black_color = {0}; - struct drr_params params = {0}; - unsigned int event_triggers = 0; if (!pipe_ctx_old->stream) { @@ -1316,20 +1278,6 @@ static enum dc_status dce110_enable_stream_timing( pipe_ctx->stream_res.tg, &stream->timing, true); - - params.vertical_total_min = stream->adjust.v_total_min; - params.vertical_total_max = stream->adjust.v_total_max; - if (pipe_ctx->stream_res.tg->funcs->set_drr) - pipe_ctx->stream_res.tg->funcs->set_drr( - pipe_ctx->stream_res.tg, ¶ms); - - // DRR should set trigger event to monitor surface update event - if (stream->adjust.v_total_min != 0 && - stream->adjust.v_total_max != 0) - event_triggers = 0x80; - if (pipe_ctx->stream_res.tg->funcs->set_static_screen_control) - pipe_ctx->stream_res.tg->funcs->set_static_screen_control( - pipe_ctx->stream_res.tg, event_triggers); } if (!pipe_ctx_old->stream) { @@ -1349,6 +1297,12 @@ static enum dc_status apply_single_controller_ctx_to_hw( struct dc *dc) { struct dc_stream_state *stream = pipe_ctx->stream; + struct drr_params params = {0}; + unsigned int event_triggers = 0; + + if (dc->hwss.disable_stream_gating) { + dc->hwss.disable_stream_gating(dc, pipe_ctx); + } if (pipe_ctx->stream_res.audio != NULL) { struct audio_output audio_output; @@ -1375,14 +1329,30 @@ static enum dc_status apply_single_controller_ctx_to_hw( } /* */ - dc->hwss.enable_stream_timing(pipe_ctx, context, dc); + /* Do not touch stream timing on seamless boot optimization. */ + if (!pipe_ctx->stream->apply_seamless_boot_optimization) + dc->hwss.enable_stream_timing(pipe_ctx, context, dc); + + if (dc->hwss.setup_vupdate_interrupt) + dc->hwss.setup_vupdate_interrupt(pipe_ctx); + + params.vertical_total_min = stream->adjust.v_total_min; + params.vertical_total_max = stream->adjust.v_total_max; + if (pipe_ctx->stream_res.tg->funcs->set_drr) + pipe_ctx->stream_res.tg->funcs->set_drr( + pipe_ctx->stream_res.tg, ¶ms); + + // DRR should set trigger event to monitor surface update event + if (stream->adjust.v_total_min != 0 && stream->adjust.v_total_max != 0) + event_triggers = 0x80; + if (pipe_ctx->stream_res.tg->funcs->set_static_screen_control) + pipe_ctx->stream_res.tg->funcs->set_static_screen_control( + pipe_ctx->stream_res.tg, event_triggers); - /* TODO: move to stream encoder */ if (pipe_ctx->stream->signal != SIGNAL_TYPE_VIRTUAL) - if (DC_OK != bios_parser_crtc_source_select(pipe_ctx)) { - BREAK_TO_DEBUGGER(); - return DC_ERROR_UNEXPECTED; - } + pipe_ctx->stream_res.stream_enc->funcs->dig_connect_to_otg( + pipe_ctx->stream_res.stream_enc, + pipe_ctx->stream_res.tg->inst); pipe_ctx->stream_res.opp->funcs->opp_set_dyn_expansion( pipe_ctx->stream_res.opp, @@ -1400,7 +1370,7 @@ static enum dc_status apply_single_controller_ctx_to_hw( pipe_ctx->plane_res.scl_data.lb_params.alpha_en = pipe_ctx->bottom_pipe != 0; - pipe_ctx->stream->sink->link->psr_enabled = false; + pipe_ctx->stream->link->psr_enabled = false; return DC_OK; } @@ -1510,7 +1480,7 @@ static struct dc_link *get_link_for_edp(struct dc *dc) return NULL; } -static struct dc_link *get_link_for_edp_not_in_use( +static struct dc_link *get_link_for_edp_to_turn_off( struct dc *dc, struct dc_state *context) { @@ -1519,8 +1489,12 @@ static struct dc_link *get_link_for_edp_not_in_use( /* check if eDP panel is suppose to be set mode, if yes, no need to disable */ for (i = 0; i < context->stream_count; i++) { - if (context->streams[i]->signal == SIGNAL_TYPE_EDP) - return NULL; + if (context->streams[i]->signal == SIGNAL_TYPE_EDP) { + if (context->streams[i]->dpms_off == true) + return context->streams[i]->sink->link; + else + return NULL; + } } /* check if there is an eDP panel not in use */ @@ -1549,6 +1523,14 @@ void dce110_enable_accelerated_mode(struct dc *dc, struct dc_state *context) struct dc_link *edp_link = get_link_for_edp(dc); bool can_edp_fast_boot_optimize = false; bool apply_edp_fast_boot_optimization = false; + bool can_apply_seamless_boot = false; + + for (i = 0; i < context->stream_count; i++) { + if (context->streams[i]->apply_seamless_boot_optimization) { + can_apply_seamless_boot = true; + break; + } + } if (edp_link) { /* this seems to cause blank screens on DCE8 */ @@ -1562,7 +1544,7 @@ void dce110_enable_accelerated_mode(struct dc *dc, struct dc_state *context) } if (can_edp_fast_boot_optimize) - edp_link_to_turnoff = get_link_for_edp_not_in_use(dc, context); + edp_link_to_turnoff = get_link_for_edp_to_turn_off(dc, context); /* if OS doesn't light up eDP and eDP link is available, we want to disable * If resume from S4/S5, should optimization. @@ -1577,7 +1559,7 @@ void dce110_enable_accelerated_mode(struct dc *dc, struct dc_state *context) } } - if (!apply_edp_fast_boot_optimization) { + if (!apply_edp_fast_boot_optimization && !can_apply_seamless_boot) { if (edp_link_to_turnoff) { /*turn off backlight before DP_blank and encoder powered down*/ dc->hwss.edp_backlight_control(edp_link_to_turnoff, false); @@ -1601,8 +1583,8 @@ static uint32_t compute_pstate_blackout_duration( pstate_blackout_duration_ns = 1000 * blackout_duration.value >> 24; total_dest_line_time_ns = 1000000UL * - stream->timing.h_total / - stream->timing.pix_clk_khz + + (stream->timing.h_total * 10) / + stream->timing.pix_clk_100hz + pstate_blackout_duration_ns; return total_dest_line_time_ns; @@ -1748,44 +1730,17 @@ static void set_static_screen_control(struct pipe_ctx **pipe_ctx, set_static_screen_control(pipe_ctx[i]->stream_res.tg, value); } -/* unit: in_khz before mode set, get pixel clock from context. ASIC register - * may not be programmed yet - */ -static uint32_t get_max_pixel_clock_for_all_paths( - struct dc *dc, - struct dc_state *context) -{ - uint32_t max_pix_clk = 0; - int i; - - for (i = 0; i < MAX_PIPES; i++) { - struct pipe_ctx *pipe_ctx = &context->res_ctx.pipe_ctx[i]; - - if (pipe_ctx->stream == NULL) - continue; - - /* do not check under lay */ - if (pipe_ctx->top_pipe) - continue; - - if (pipe_ctx->stream_res.pix_clk_params.requested_pix_clk > max_pix_clk) - max_pix_clk = - pipe_ctx->stream_res.pix_clk_params.requested_pix_clk; - } - - return max_pix_clk; -} - /* * Check if FBC can be enabled */ static bool should_enable_fbc(struct dc *dc, - struct dc_state *context, - uint32_t *pipe_idx) + struct dc_state *context, + uint32_t *pipe_idx) { uint32_t i; struct pipe_ctx *pipe_ctx = NULL; struct resource_context *res_ctx = &context->res_ctx; + unsigned int underlay_idx = dc->res_pool->underlay_pipe_index; ASSERT(dc->fbc_compressor); @@ -1800,21 +1755,32 @@ static bool should_enable_fbc(struct dc *dc, for (i = 0; i < dc->res_pool->pipe_count; i++) { if (res_ctx->pipe_ctx[i].stream) { + pipe_ctx = &res_ctx->pipe_ctx[i]; - *pipe_idx = i; - break; + + if (!pipe_ctx) + continue; + + /* fbc not applicable on underlay pipe */ + if (pipe_ctx->pipe_idx != underlay_idx) { + *pipe_idx = i; + break; + } } } - /* Pipe context should be found */ - ASSERT(pipe_ctx); + if (i == dc->res_pool->pipe_count) + return false; + + if (!pipe_ctx->stream->link) + return false; /* Only supports eDP */ - if (pipe_ctx->stream->sink->link->connector_signal != SIGNAL_TYPE_EDP) + if (pipe_ctx->stream->link->connector_signal != SIGNAL_TYPE_EDP) return false; /* PSR should not be enabled */ - if (pipe_ctx->stream->sink->link->psr_enabled) + if (pipe_ctx->stream->link->psr_enabled) return false; /* Nothing to compress */ @@ -1831,8 +1797,9 @@ static bool should_enable_fbc(struct dc *dc, /* * Enable FBC */ -static void enable_fbc(struct dc *dc, - struct dc_state *context) +static void enable_fbc( + struct dc *dc, + struct dc_state *context) { uint32_t pipe_idx = 0; @@ -1842,10 +1809,9 @@ static void enable_fbc(struct dc *dc, struct compressor *compr = dc->fbc_compressor; struct pipe_ctx *pipe_ctx = &context->res_ctx.pipe_ctx[pipe_idx]; - params.source_view_width = pipe_ctx->stream->timing.h_addressable; params.source_view_height = pipe_ctx->stream->timing.v_addressable; - + params.inst = pipe_ctx->stream_res.tg->inst; compr->compr_surface_address.quad_part = dc->ctx->fbc_gpu_addr; compr->funcs->surface_address_and_pitch(compr, ¶ms); @@ -2060,10 +2026,10 @@ enum dc_status dce110_apply_ctx_to_hw( return status; } - dcb->funcs->set_scratch_critical_state(dcb, false); - if (dc->fbc_compressor) - enable_fbc(dc, context); + enable_fbc(dc, dc->current_state); + + dcb->funcs->set_scratch_critical_state(dcb, false); return DC_OK; } @@ -2296,7 +2262,7 @@ static void dce110_enable_per_frame_crtc_position_reset( int i; gsl_params.gsl_group = 0; - gsl_params.gsl_master = grouped_pipes[0]->stream->triggered_crtc_reset.event_source->status.primary_otg_inst; + gsl_params.gsl_master = 0; for (i = 0; i < group_size; i++) grouped_pipes[i]->stream_res.tg->funcs->setup_global_swap_lock( @@ -2319,6 +2285,11 @@ static void dce110_enable_per_frame_crtc_position_reset( } +static void init_pipes(struct dc *dc, struct dc_state *context) +{ + // Do nothing +} + static void init_hw(struct dc *dc) { int i; @@ -2385,193 +2356,33 @@ static void init_hw(struct dc *dc) } -void dce110_fill_display_configs( - const struct dc_state *context, - struct dm_pp_display_configuration *pp_display_cfg) -{ - int j; - int num_cfgs = 0; - - for (j = 0; j < context->stream_count; j++) { - int k; - - const struct dc_stream_state *stream = context->streams[j]; - struct dm_pp_single_disp_config *cfg = - &pp_display_cfg->disp_configs[num_cfgs]; - const struct pipe_ctx *pipe_ctx = NULL; - for (k = 0; k < MAX_PIPES; k++) - if (stream == context->res_ctx.pipe_ctx[k].stream) { - pipe_ctx = &context->res_ctx.pipe_ctx[k]; - break; - } - - ASSERT(pipe_ctx != NULL); - - /* only notify active stream */ - if (stream->dpms_off) - continue; - - num_cfgs++; - cfg->signal = pipe_ctx->stream->signal; - cfg->pipe_idx = pipe_ctx->stream_res.tg->inst; - cfg->src_height = stream->src.height; - cfg->src_width = stream->src.width; - cfg->ddi_channel_mapping = - stream->sink->link->ddi_channel_mapping.raw; - cfg->transmitter = - stream->sink->link->link_enc->transmitter; - cfg->link_settings.lane_count = - stream->sink->link->cur_link_settings.lane_count; - cfg->link_settings.link_rate = - stream->sink->link->cur_link_settings.link_rate; - cfg->link_settings.link_spread = - stream->sink->link->cur_link_settings.link_spread; - cfg->sym_clock = stream->phy_pix_clk; - /* Round v_refresh*/ - cfg->v_refresh = stream->timing.pix_clk_khz * 1000; - cfg->v_refresh /= stream->timing.h_total; - cfg->v_refresh = (cfg->v_refresh + stream->timing.v_total / 2) - / stream->timing.v_total; - } - - pp_display_cfg->display_count = num_cfgs; -} - -uint32_t dce110_get_min_vblank_time_us(const struct dc_state *context) -{ - uint8_t j; - uint32_t min_vertical_blank_time = -1; - - for (j = 0; j < context->stream_count; j++) { - struct dc_stream_state *stream = context->streams[j]; - uint32_t vertical_blank_in_pixels = 0; - uint32_t vertical_blank_time = 0; - - vertical_blank_in_pixels = stream->timing.h_total * - (stream->timing.v_total - - stream->timing.v_addressable); - - vertical_blank_time = vertical_blank_in_pixels - * 1000 / stream->timing.pix_clk_khz; - - if (min_vertical_blank_time > vertical_blank_time) - min_vertical_blank_time = vertical_blank_time; - } - - return min_vertical_blank_time; -} - -static int determine_sclk_from_bounding_box( - const struct dc *dc, - int required_sclk) -{ - int i; - - /* - * Some asics do not give us sclk levels, so we just report the actual - * required sclk - */ - if (dc->sclk_lvls.num_levels == 0) - return required_sclk; - - for (i = 0; i < dc->sclk_lvls.num_levels; i++) { - if (dc->sclk_lvls.clocks_in_khz[i] >= required_sclk) - return dc->sclk_lvls.clocks_in_khz[i]; - } - /* - * even maximum level could not satisfy requirement, this - * is unexpected at this stage, should have been caught at - * validation time - */ - ASSERT(0); - return dc->sclk_lvls.clocks_in_khz[dc->sclk_lvls.num_levels - 1]; -} - -static void pplib_apply_display_requirements( - struct dc *dc, - struct dc_state *context) +void dce110_prepare_bandwidth( + struct dc *dc, + struct dc_state *context) { - struct dm_pp_display_configuration *pp_display_cfg = &context->pp_display_cfg; - - pp_display_cfg->all_displays_in_sync = - context->bw.dce.all_displays_in_sync; - pp_display_cfg->nb_pstate_switch_disable = - context->bw.dce.nbp_state_change_enable == false; - pp_display_cfg->cpu_cc6_disable = - context->bw.dce.cpuc_state_change_enable == false; - pp_display_cfg->cpu_pstate_disable = - context->bw.dce.cpup_state_change_enable == false; - pp_display_cfg->cpu_pstate_separation_time = - context->bw.dce.blackout_recovery_time_us; - - pp_display_cfg->min_memory_clock_khz = context->bw.dce.yclk_khz - / MEMORY_TYPE_MULTIPLIER; - - pp_display_cfg->min_engine_clock_khz = determine_sclk_from_bounding_box( - dc, - context->bw.dce.sclk_khz); - - pp_display_cfg->min_dcfclock_khz = pp_display_cfg->min_engine_clock_khz; + struct clk_mgr *dccg = dc->res_pool->clk_mgr; - pp_display_cfg->min_engine_clock_deep_sleep_khz - = context->bw.dce.sclk_deep_sleep_khz; + dce110_set_safe_displaymarks(&context->res_ctx, dc->res_pool); - pp_display_cfg->avail_mclk_switch_time_us = - dce110_get_min_vblank_time_us(context); - /* TODO: dce11.2*/ - pp_display_cfg->avail_mclk_switch_time_in_disp_active_us = 0; - - pp_display_cfg->disp_clk_khz = dc->res_pool->dccg->clks.dispclk_khz; - - dce110_fill_display_configs(context, pp_display_cfg); - - /* TODO: is this still applicable?*/ - if (pp_display_cfg->display_count == 1) { - const struct dc_crtc_timing *timing = - &context->streams[0]->timing; - - pp_display_cfg->crtc_index = - pp_display_cfg->disp_configs[0].pipe_idx; - pp_display_cfg->line_time_in_us = timing->h_total * 1000 - / timing->pix_clk_khz; - } - - if (memcmp(&dc->prev_display_config, pp_display_cfg, sizeof( - struct dm_pp_display_configuration)) != 0) - dm_pp_apply_display_requirements(dc->ctx, pp_display_cfg); - - dc->prev_display_config = *pp_display_cfg; + dccg->funcs->update_clocks( + dccg, + context, + false); } -static void dce110_set_bandwidth( +void dce110_optimize_bandwidth( struct dc *dc, - struct dc_state *context, - bool decrease_allowed) + struct dc_state *context) { - struct dc_clocks req_clks; - struct dccg *dccg = dc->res_pool->dccg; - - req_clks.dispclk_khz = context->bw.dce.dispclk_khz; - req_clks.phyclk_khz = get_max_pixel_clock_for_all_paths(dc, context); - - if (decrease_allowed) - dce110_set_displaymarks(dc, context); - else - dce110_set_safe_displaymarks(&context->res_ctx, dc->res_pool); + struct clk_mgr *dccg = dc->res_pool->clk_mgr; - if (dccg->funcs->update_dfs_bypass) - dccg->funcs->update_dfs_bypass( - dccg, - dc, - context, - req_clks.dispclk_khz); + dce110_set_displaymarks(dc, context); dccg->funcs->update_clocks( dccg, - &req_clks, - decrease_allowed); - pplib_apply_display_requirements(dc, context); + context, + true); } static void dce110_program_front_end_for_pipe( @@ -2582,7 +2393,6 @@ static void dce110_program_front_end_for_pipe( struct dc_plane_state *plane_state = pipe_ctx->plane_state; struct xfm_grph_csc_adjustment adjust; struct out_csc_color_matrix tbl_entry; - unsigned int underlay_idx = dc->res_pool->underlay_pipe_index; unsigned int i; DC_LOGGER_INIT(); memset(&tbl_entry, 0, sizeof(tbl_entry)); @@ -2623,15 +2433,6 @@ static void dce110_program_front_end_for_pipe( program_scaler(dc, pipe_ctx); - /* fbc not applicable on Underlay pipe */ - if (dc->fbc_compressor && old_pipe->stream && - pipe_ctx->pipe_idx != underlay_idx) { - if (plane_state->tiling_info.gfx8.array_mode == DC_ARRAY_LINEAR_GENERAL) - dc->fbc_compressor->funcs->disable_fbc(dc->fbc_compressor); - else - enable_fbc(dc, dc->current_state); - } - mi->funcs->mem_input_program_surface_config( mi, plane_state->format, @@ -2708,6 +2509,9 @@ static void dce110_apply_ctx_for_surface( if (num_planes == 0) return; + if (dc->fbc_compressor) + dc->fbc_compressor->funcs->disable_fbc(dc->fbc_compressor); + for (i = 0; i < dc->res_pool->pipe_count; i++) { struct pipe_ctx *pipe_ctx = &context->res_ctx.pipe_ctx[i]; struct pipe_ctx *old_pipe_ctx = &dc->current_state->res_ctx.pipe_ctx[i]; @@ -2730,7 +2534,7 @@ static void dce110_apply_ctx_for_surface( pipe_ctx->plane_res.mi, pipe_ctx->stream->timing.h_total, pipe_ctx->stream->timing.v_total, - pipe_ctx->stream->timing.pix_clk_khz, + pipe_ctx->stream->timing.pix_clk_100hz / 10, context->stream_count); dce110_program_front_end_for_pipe(dc, pipe_ctx); @@ -2750,6 +2554,9 @@ static void dce110_apply_ctx_for_surface( (pipe_ctx->plane_state || old_pipe_ctx->plane_state)) dc->hwss.pipe_control_lock(dc, pipe_ctx, false); } + + if (dc->fbc_compressor) + enable_fbc(dc, context); } static void dce110_power_down_fe(struct dc *dc, struct pipe_ctx *pipe_ctx) @@ -2776,25 +2583,25 @@ static void dce110_wait_for_mpcc_disconnect( /* do nothing*/ } -static void program_csc_matrix(struct pipe_ctx *pipe_ctx, +static void program_output_csc(struct dc *dc, + struct pipe_ctx *pipe_ctx, enum dc_color_space colorspace, - uint16_t *matrix) + uint16_t *matrix, + int opp_id) { int i; struct out_csc_color_matrix tbl_entry; - if (pipe_ctx->stream->csc_color_matrix.enable_adjustment - == true) { - enum dc_color_space color_space = - pipe_ctx->stream->output_color_space; + if (pipe_ctx->stream->csc_color_matrix.enable_adjustment == true) { + enum dc_color_space color_space = pipe_ctx->stream->output_color_space; + + for (i = 0; i < 12; i++) + tbl_entry.regval[i] = pipe_ctx->stream->csc_color_matrix.matrix[i]; - //uint16_t matrix[12]; - for (i = 0; i < 12; i++) - tbl_entry.regval[i] = pipe_ctx->stream->csc_color_matrix.matrix[i]; + tbl_entry.color_space = color_space; - tbl_entry.color_space = color_space; - //tbl_entry.regval = matrix; - pipe_ctx->plane_res.xfm->funcs->opp_set_csc_adjustment(pipe_ctx->plane_res.xfm, &tbl_entry); + pipe_ctx->plane_res.xfm->funcs->opp_set_csc_adjustment( + pipe_ctx->plane_res.xfm, &tbl_entry); } } @@ -2804,7 +2611,7 @@ void dce110_set_cursor_position(struct pipe_ctx *pipe_ctx) struct input_pixel_processor *ipp = pipe_ctx->plane_res.ipp; struct mem_input *mi = pipe_ctx->plane_res.mi; struct dc_cursor_mi_param param = { - .pixel_clk_khz = pipe_ctx->stream->timing.pix_clk_khz, + .pixel_clk_khz = pipe_ctx->stream->timing.pix_clk_100hz / 10, .ref_clk_khz = pipe_ctx->stream->ctx->dc->res_pool->ref_clock_inKhz, .viewport = pipe_ctx->plane_res.scl_data.viewport, .h_scale_ratio = pipe_ctx->plane_res.scl_data.ratios.horz, @@ -2846,14 +2653,11 @@ void dce110_set_cursor_attribute(struct pipe_ctx *pipe_ctx) pipe_ctx->plane_res.xfm, attributes); } -static void ready_shared_resources(struct dc *dc, struct dc_state *context) {} - -static void optimize_shared_resources(struct dc *dc) {} - static const struct hw_sequencer_funcs dce110_funcs = { .program_gamut_remap = program_gamut_remap, - .program_csc_matrix = program_csc_matrix, + .program_output_csc = program_output_csc, .init_hw = init_hw, + .init_pipes = init_pipes, .apply_ctx_to_hw = dce110_apply_ctx_to_hw, .apply_ctx_for_surface = dce110_apply_ctx_for_surface, .update_plane_addr = update_plane_addr, @@ -2875,18 +2679,18 @@ static const struct hw_sequencer_funcs dce110_funcs = { .enable_display_power_gating = dce110_enable_display_power_gating, .disable_plane = dce110_power_down_fe, .pipe_control_lock = dce_pipe_control_lock, - .set_bandwidth = dce110_set_bandwidth, + .prepare_bandwidth = dce110_prepare_bandwidth, + .optimize_bandwidth = dce110_optimize_bandwidth, .set_drr = set_drr, .get_position = get_position, .set_static_screen_control = set_static_screen_control, .reset_hw_ctx_wrap = dce110_reset_hw_ctx_wrap, .enable_stream_timing = dce110_enable_stream_timing, + .disable_stream_gating = NULL, + .enable_stream_gating = NULL, .setup_stereo = NULL, .set_avmute = dce110_set_avmute, .wait_for_mpcc_disconnect = dce110_wait_for_mpcc_disconnect, - .ready_shared_resources = ready_shared_resources, - .optimize_shared_resources = optimize_shared_resources, - .pplib_apply_display_requirements = pplib_apply_display_requirements, .edp_backlight_control = hwss_edp_backlight_control, .edp_power_control = hwss_edp_power_control, .edp_wait_for_hpd_ready = hwss_edp_wait_for_hpd_ready, diff --git a/drivers/gpu/drm/amd/display/dc/dce110/dce110_hw_sequencer.h b/drivers/gpu/drm/amd/display/dc/dce110/dce110_hw_sequencer.h index d6db3dbd9015..cd3e36d52a52 100644 --- a/drivers/gpu/drm/amd/display/dc/dce110/dce110_hw_sequencer.h +++ b/drivers/gpu/drm/amd/display/dc/dce110/dce110_hw_sequencer.h @@ -40,7 +40,6 @@ enum dc_status dce110_apply_ctx_to_hw( struct dc_state *context); - void dce110_enable_stream(struct pipe_ctx *pipe_ctx); void dce110_disable_stream(struct pipe_ctx *pipe_ctx, int option); @@ -64,11 +63,13 @@ void dce110_set_safe_displaymarks( struct resource_context *res_ctx, const struct resource_pool *pool); -void dce110_fill_display_configs( - const struct dc_state *context, - struct dm_pp_display_configuration *pp_display_cfg); +void dce110_prepare_bandwidth( + struct dc *dc, + struct dc_state *context); -uint32_t dce110_get_min_vblank_time_us(const struct dc_state *context); +void dce110_optimize_bandwidth( + struct dc *dc, + struct dc_state *context); void dp_receiver_power_ctrl(struct dc_link *link, bool on); diff --git a/drivers/gpu/drm/amd/display/dc/dce110/dce110_resource.c b/drivers/gpu/drm/amd/display/dc/dce110/dce110_resource.c index 7c9fd9052ee2..7549adaa1542 100644 --- a/drivers/gpu/drm/amd/display/dc/dce110/dce110_resource.c +++ b/drivers/gpu/drm/amd/display/dc/dce110/dce110_resource.c @@ -31,6 +31,7 @@ #include "resource.h" #include "dce110/dce110_resource.h" +#include "dce/dce_clk_mgr.h" #include "include/irq_service_interface.h" #include "dce/dce_audio.h" #include "dce110/dce110_timing_generator.h" @@ -45,7 +46,6 @@ #include "dce110/dce110_transform_v.h" #include "dce/dce_opp.h" #include "dce110/dce110_opp_v.h" -#include "dce/dce_clocks.h" #include "dce/dce_clock_source.h" #include "dce/dce_hwseq.h" #include "dce110/dce110_hw_sequencer.h" @@ -84,6 +84,7 @@ #ifndef mmBIOS_SCRATCH_2 #define mmBIOS_SCRATCH_2 0x05CB + #define mmBIOS_SCRATCH_3 0x05CC #define mmBIOS_SCRATCH_6 0x05CF #endif @@ -148,15 +149,15 @@ static const struct dce110_timing_generator_offsets dce110_tg_offsets[] = { #define SRI(reg_name, block, id)\ .reg_name = mm ## block ## id ## _ ## reg_name -static const struct dccg_registers disp_clk_regs = { +static const struct clk_mgr_registers disp_clk_regs = { CLK_COMMON_REG_LIST_DCE_BASE() }; -static const struct dccg_shift disp_clk_shift = { +static const struct clk_mgr_shift disp_clk_shift = { CLK_COMMON_MASK_SH_LIST_DCE_COMMON_BASE(__SHIFT) }; -static const struct dccg_mask disp_clk_mask = { +static const struct clk_mgr_mask disp_clk_mask = { CLK_COMMON_MASK_SH_LIST_DCE_COMMON_BASE(_MASK) }; @@ -369,6 +370,7 @@ static const struct dce110_clk_src_mask cs_mask = { }; static const struct bios_registers bios_regs = { + .BIOS_SCRATCH_3 = mmBIOS_SCRATCH_3, .BIOS_SCRATCH_6 = mmBIOS_SCRATCH_6 }; @@ -606,7 +608,7 @@ static struct output_pixel_processor *dce110_opp_create( return &opp->base; } -struct aux_engine *dce110_aux_engine_create( +struct dce_aux *dce110_aux_engine_create( struct dc_context *ctx, uint32_t inst) { @@ -760,8 +762,8 @@ static void destruct(struct dce110_resource_pool *pool) if (pool->base.dmcu != NULL) dce_dmcu_destroy(&pool->base.dmcu); - if (pool->base.dccg != NULL) - dce_dccg_destroy(&pool->base.dccg); + if (pool->base.clk_mgr != NULL) + dce_clk_mgr_destroy(&pool->base.clk_mgr); if (pool->base.irqs != NULL) { dal_irq_service_destroy(&pool->base.irqs); @@ -779,8 +781,8 @@ static void get_pixel_clock_parameters( * the pixel clock normalization for hdmi up to here instead of doing it * in pll_adjust_pix_clk */ - pixel_clk_params->requested_pix_clk = stream->timing.pix_clk_khz; - pixel_clk_params->encoder_object_id = stream->sink->link->link_enc->id; + pixel_clk_params->requested_pix_clk_100hz = stream->timing.pix_clk_100hz; + pixel_clk_params->encoder_object_id = stream->link->link_enc->id; pixel_clk_params->signal_type = pipe_ctx->stream->signal; pixel_clk_params->controller_id = pipe_ctx->stream_res.tg->inst + 1; /* TODO: un-hardcode*/ @@ -797,10 +799,10 @@ static void get_pixel_clock_parameters( pixel_clk_params->color_depth = COLOR_DEPTH_888; } if (stream->timing.pixel_encoding == PIXEL_ENCODING_YCBCR420) { - pixel_clk_params->requested_pix_clk = pixel_clk_params->requested_pix_clk / 2; + pixel_clk_params->requested_pix_clk_100hz = pixel_clk_params->requested_pix_clk_100hz / 2; } if (stream->timing.timing_3d_format == TIMING_3D_FORMAT_HW_FRAME_PACKING) - pixel_clk_params->requested_pix_clk *= 2; + pixel_clk_params->requested_pix_clk_100hz *= 2; } @@ -874,7 +876,7 @@ static bool dce110_validate_bandwidth( __func__, context->streams[0]->timing.h_addressable, context->streams[0]->timing.v_addressable, - context->streams[0]->timing.pix_clk_khz); + context->streams[0]->timing.pix_clk_100hz / 10); if (memcmp(&dc->current_state->bw.dce, &context->bw.dce, sizeof(context->bw.dce))) { @@ -1055,7 +1057,7 @@ static struct pipe_ctx *dce110_acquire_underlay( pipe_ctx->plane_res.mi->funcs->allocate_mem_input(pipe_ctx->plane_res.mi, stream->timing.h_total, stream->timing.v_total, - stream->timing.pix_clk_khz, + stream->timing.pix_clk_100hz / 10, context->stream_count); color_space_to_black_color(dc, @@ -1173,12 +1175,12 @@ static void bw_calcs_data_update_from_pplib(struct dc *dc) &clks); dc->bw_vbios->low_yclk = bw_frc_to_fixed( - clks.clocks_in_khz[0] * MEMORY_TYPE_MULTIPLIER, 1000); + clks.clocks_in_khz[0] * MEMORY_TYPE_MULTIPLIER_CZ, 1000); dc->bw_vbios->mid_yclk = bw_frc_to_fixed( - clks.clocks_in_khz[clks.num_levels>>1] * MEMORY_TYPE_MULTIPLIER, + clks.clocks_in_khz[clks.num_levels>>1] * MEMORY_TYPE_MULTIPLIER_CZ, 1000); dc->bw_vbios->high_yclk = bw_frc_to_fixed( - clks.clocks_in_khz[clks.num_levels-1] * MEMORY_TYPE_MULTIPLIER, + clks.clocks_in_khz[clks.num_levels-1] * MEMORY_TYPE_MULTIPLIER_CZ, 1000); } @@ -1201,7 +1203,6 @@ static bool construct( struct dc_context *ctx = dc->ctx; struct dc_firmware_info info; struct dc_bios *bp; - struct dm_pp_static_clock_info static_clk_info = {0}; ctx->dc_bios->regs = &bios_regs; @@ -1257,11 +1258,11 @@ static bool construct( } } - pool->base.dccg = dce110_dccg_create(ctx, + pool->base.clk_mgr = dce110_clk_mgr_create(ctx, &disp_clk_regs, &disp_clk_shift, &disp_clk_mask); - if (pool->base.dccg == NULL) { + if (pool->base.clk_mgr == NULL) { dm_error("DC: failed to create display clock!\n"); BREAK_TO_DEBUGGER(); goto res_create_fail; @@ -1287,13 +1288,6 @@ static bool construct( goto res_create_fail; } - /* get static clock information for PPLIB or firmware, save - * max_clock_state - */ - if (dm_pp_get_static_clocks(ctx, &static_clk_info)) - pool->base.dccg->max_clks_state = - static_clk_info.max_clocks_state; - { struct irq_service_init_data init_data; init_data.ctx = dc->ctx; diff --git a/drivers/gpu/drm/amd/display/dc/dce112/dce112_resource.c b/drivers/gpu/drm/amd/display/dc/dce112/dce112_resource.c index 3ce79c208ddf..ea3065d63372 100644 --- a/drivers/gpu/drm/amd/display/dc/dce112/dce112_resource.c +++ b/drivers/gpu/drm/amd/display/dc/dce112/dce112_resource.c @@ -35,6 +35,7 @@ #include "irq/dce110/irq_service_dce110.h" +#include "dce/dce_clk_mgr.h" #include "dce/dce_mem_input.h" #include "dce/dce_transform.h" #include "dce/dce_link_encoder.h" @@ -42,7 +43,6 @@ #include "dce/dce_audio.h" #include "dce/dce_opp.h" #include "dce/dce_ipp.h" -#include "dce/dce_clocks.h" #include "dce/dce_clock_source.h" #include "dce/dce_hwseq.h" @@ -76,6 +76,7 @@ #ifndef mmBIOS_SCRATCH_2 #define mmBIOS_SCRATCH_2 0x05CB + #define mmBIOS_SCRATCH_3 0x05CC #define mmBIOS_SCRATCH_6 0x05CF #endif @@ -148,15 +149,15 @@ static const struct dce110_timing_generator_offsets dce112_tg_offsets[] = { .reg_name = mm ## block ## id ## _ ## reg_name -static const struct dccg_registers disp_clk_regs = { +static const struct clk_mgr_registers disp_clk_regs = { CLK_COMMON_REG_LIST_DCE_BASE() }; -static const struct dccg_shift disp_clk_shift = { +static const struct clk_mgr_shift disp_clk_shift = { CLK_COMMON_MASK_SH_LIST_DCE_COMMON_BASE(__SHIFT) }; -static const struct dccg_mask disp_clk_mask = { +static const struct clk_mgr_mask disp_clk_mask = { CLK_COMMON_MASK_SH_LIST_DCE_COMMON_BASE(_MASK) }; @@ -376,6 +377,7 @@ static const struct dce110_clk_src_mask cs_mask = { }; static const struct bios_registers bios_regs = { + .BIOS_SCRATCH_3 = mmBIOS_SCRATCH_3, .BIOS_SCRATCH_6 = mmBIOS_SCRATCH_6 }; @@ -551,7 +553,8 @@ static struct transform *dce112_transform_create( static const struct encoder_feature_support link_enc_feature = { .max_hdmi_deep_color = COLOR_DEPTH_121212, .max_hdmi_pixel_clock = 600000, - .ycbcr420_supported = true, + .hdmi_ycbcr420_supported = true, + .dp_ycbcr420_supported = false, .flags.bits.IS_HBR2_CAPABLE = true, .flags.bits.IS_HBR3_CAPABLE = true, .flags.bits.IS_TPS3_CAPABLE = true, @@ -606,7 +609,7 @@ struct output_pixel_processor *dce112_opp_create( return &opp->base; } -struct aux_engine *dce112_aux_engine_create( +struct dce_aux *dce112_aux_engine_create( struct dc_context *ctx, uint32_t inst) { @@ -749,8 +752,8 @@ static void destruct(struct dce110_resource_pool *pool) if (pool->base.dmcu != NULL) dce_dmcu_destroy(&pool->base.dmcu); - if (pool->base.dccg != NULL) - dce_dccg_destroy(&pool->base.dccg); + if (pool->base.clk_mgr != NULL) + dce_clk_mgr_destroy(&pool->base.clk_mgr); if (pool->base.irqs != NULL) { dal_irq_service_destroy(&pool->base.irqs); @@ -762,7 +765,7 @@ static struct clock_source *find_matching_pll( const struct resource_pool *pool, const struct dc_stream_state *const stream) { - switch (stream->sink->link->link_enc->transmitter) { + switch (stream->link->link_enc->transmitter) { case TRANSMITTER_UNIPHY_A: return pool->clock_sources[DCE112_CLK_SRC_PLL0]; case TRANSMITTER_UNIPHY_B: @@ -1015,12 +1018,12 @@ static void bw_calcs_data_update_from_pplib(struct dc *dc) &clks); dc->bw_vbios->low_yclk = bw_frc_to_fixed( - clks.clocks_in_khz[0] * MEMORY_TYPE_MULTIPLIER, 1000); + clks.clocks_in_khz[0] * MEMORY_TYPE_MULTIPLIER_CZ, 1000); dc->bw_vbios->mid_yclk = bw_frc_to_fixed( - clks.clocks_in_khz[clks.num_levels>>1] * MEMORY_TYPE_MULTIPLIER, + clks.clocks_in_khz[clks.num_levels>>1] * MEMORY_TYPE_MULTIPLIER_CZ, 1000); dc->bw_vbios->high_yclk = bw_frc_to_fixed( - clks.clocks_in_khz[clks.num_levels-1] * MEMORY_TYPE_MULTIPLIER, + clks.clocks_in_khz[clks.num_levels-1] * MEMORY_TYPE_MULTIPLIER_CZ, 1000); return; @@ -1056,12 +1059,12 @@ static void bw_calcs_data_update_from_pplib(struct dc *dc) * YCLK = UMACLK*m_memoryTypeMultiplier */ dc->bw_vbios->low_yclk = bw_frc_to_fixed( - mem_clks.data[0].clocks_in_khz * MEMORY_TYPE_MULTIPLIER, 1000); + mem_clks.data[0].clocks_in_khz * MEMORY_TYPE_MULTIPLIER_CZ, 1000); dc->bw_vbios->mid_yclk = bw_frc_to_fixed( - mem_clks.data[mem_clks.num_levels>>1].clocks_in_khz * MEMORY_TYPE_MULTIPLIER, + mem_clks.data[mem_clks.num_levels>>1].clocks_in_khz * MEMORY_TYPE_MULTIPLIER_CZ, 1000); dc->bw_vbios->high_yclk = bw_frc_to_fixed( - mem_clks.data[mem_clks.num_levels-1].clocks_in_khz * MEMORY_TYPE_MULTIPLIER, + mem_clks.data[mem_clks.num_levels-1].clocks_in_khz * MEMORY_TYPE_MULTIPLIER_CZ, 1000); /* Now notify PPLib/SMU about which Watermarks sets they should select @@ -1131,7 +1134,6 @@ static bool construct( { unsigned int i; struct dc_context *ctx = dc->ctx; - struct dm_pp_static_clock_info static_clk_info = {0}; ctx->dc_bios->regs = &bios_regs; @@ -1199,11 +1201,11 @@ static bool construct( } } - pool->base.dccg = dce112_dccg_create(ctx, + pool->base.clk_mgr = dce112_clk_mgr_create(ctx, &disp_clk_regs, &disp_clk_shift, &disp_clk_mask); - if (pool->base.dccg == NULL) { + if (pool->base.clk_mgr == NULL) { dm_error("DC: failed to create display clock!\n"); BREAK_TO_DEBUGGER(); goto res_create_fail; @@ -1229,13 +1231,6 @@ static bool construct( goto res_create_fail; } - /* get static clock information for PPLIB or firmware, save - * max_clock_state - */ - if (dm_pp_get_static_clocks(ctx, &static_clk_info)) - pool->base.dccg->max_clks_state = - static_clk_info.max_clocks_state; - { struct irq_service_init_data init_data; init_data.ctx = dc->ctx; diff --git a/drivers/gpu/drm/amd/display/dc/dce120/dce120_hw_sequencer.c b/drivers/gpu/drm/amd/display/dc/dce120/dce120_hw_sequencer.c index eb0f5f9a973b..1ca30928025e 100644 --- a/drivers/gpu/drm/amd/display/dc/dce120/dce120_hw_sequencer.c +++ b/drivers/gpu/drm/amd/display/dc/dce120/dce120_hw_sequencer.c @@ -244,6 +244,21 @@ static void dce120_update_dchub( dh_data->dchub_info_valid = false; } +/** + * dce121_xgmi_enabled() - Check if xGMI is enabled + * @hws: DCE hardware sequencer object + * + * Return true if xGMI is enabled. False otherwise. + */ +bool dce121_xgmi_enabled(struct dce_hwseq *hws) +{ + uint32_t pf_max_region; + + REG_GET(MC_VM_XGMI_LFB_CNTL, PF_MAX_REGION, &pf_max_region); + /* PF_MAX_REGION == 0 means xgmi is disabled */ + return !!pf_max_region; +} + void dce120_hw_sequencer_construct(struct dc *dc) { /* All registers used by dce11.2 match those in dce11 in offset and diff --git a/drivers/gpu/drm/amd/display/dc/dce120/dce120_hw_sequencer.h b/drivers/gpu/drm/amd/display/dc/dce120/dce120_hw_sequencer.h index 77a6b86d7606..c51afbd0b012 100644 --- a/drivers/gpu/drm/amd/display/dc/dce120/dce120_hw_sequencer.h +++ b/drivers/gpu/drm/amd/display/dc/dce120/dce120_hw_sequencer.h @@ -30,6 +30,7 @@ struct dc; +bool dce121_xgmi_enabled(struct dce_hwseq *hws); void dce120_hw_sequencer_construct(struct dc *dc); #endif /* __DC_HWSS_DCE112_H__ */ diff --git a/drivers/gpu/drm/amd/display/dc/dce120/dce120_resource.c b/drivers/gpu/drm/amd/display/dc/dce120/dce120_resource.c index 79ab5f9f9115..312a0aebf91f 100644 --- a/drivers/gpu/drm/amd/display/dc/dce120/dce120_resource.c +++ b/drivers/gpu/drm/amd/display/dc/dce120/dce120_resource.c @@ -31,6 +31,7 @@ #include "resource.h" #include "include/irq_service_interface.h" #include "dce120_resource.h" + #include "dce112/dce112_resource.h" #include "dce110/dce110_resource.h" @@ -39,7 +40,6 @@ #include "irq/dce120/irq_service_dce120.h" #include "dce/dce_opp.h" #include "dce/dce_clock_source.h" -#include "dce/dce_clocks.h" #include "dce/dce_ipp.h" #include "dce/dce_mem_input.h" @@ -47,6 +47,7 @@ #include "dce120/dce120_hw_sequencer.h" #include "dce/dce_transform.h" +#include "dce/dce_clk_mgr.h" #include "dce/dce_audio.h" #include "dce/dce_link_encoder.h" #include "dce/dce_stream_encoder.h" @@ -61,6 +62,8 @@ #include "soc15_hw_ip.h" #include "vega10_ip_offset.h" #include "nbio/nbio_6_1_offset.h" +#include "mmhub/mmhub_9_4_0_offset.h" +#include "mmhub/mmhub_9_4_0_sh_mask.h" #include "reg_helper.h" #include "dce100/dce100_resource.h" @@ -138,6 +141,17 @@ static const struct dce110_timing_generator_offsets dce120_tg_offsets[] = { .reg_name = BASE(mm ## block ## id ## _ ## reg_name ## _BASE_IDX) + \ mm ## block ## id ## _ ## reg_name +/* MMHUB */ +#define MMHUB_BASE_INNER(seg) \ + MMHUB_BASE__INST0_SEG ## seg + +#define MMHUB_BASE(seg) \ + MMHUB_BASE_INNER(seg) + +#define MMHUB_SR(reg_name)\ + .reg_name = MMHUB_BASE(mm ## reg_name ## _BASE_IDX) + \ + mm ## reg_name + /* macros to expend register list macro defined in HW object header file * end *********************/ @@ -377,7 +391,7 @@ struct output_pixel_processor *dce120_opp_create( ctx, inst, &opp_regs[inst], &opp_shift, &opp_mask); return &opp->base; } -struct aux_engine *dce120_aux_engine_create( +struct dce_aux *dce120_aux_engine_create( struct dc_context *ctx, uint32_t inst) { @@ -428,6 +442,7 @@ struct dce_i2c_hw *dce120_i2c_hw_create( return dce_i2c_hw; } static const struct bios_registers bios_regs = { + .BIOS_SCRATCH_3 = mmBIOS_SCRATCH_3 + NBIO_BASE(mmBIOS_SCRATCH_3_BASE_IDX), .BIOS_SCRATCH_6 = mmBIOS_SCRATCH_6 + NBIO_BASE(mmBIOS_SCRATCH_6_BASE_IDX) }; @@ -573,8 +588,8 @@ static void destruct(struct dce110_resource_pool *pool) if (pool->base.dmcu != NULL) dce_dmcu_destroy(&pool->base.dmcu); - if (pool->base.dccg != NULL) - dce_dccg_destroy(&pool->base.dccg); + if (pool->base.clk_mgr != NULL) + dce_clk_mgr_destroy(&pool->base.clk_mgr); } static void read_dce_straps( @@ -606,7 +621,8 @@ static struct audio *create_audio( static const struct encoder_feature_support link_enc_feature = { .max_hdmi_deep_color = COLOR_DEPTH_121212, .max_hdmi_pixel_clock = 600000, - .ycbcr420_supported = true, + .hdmi_ycbcr420_supported = true, + .dp_ycbcr420_supported = false, .flags.bits.IS_HBR2_CAPABLE = true, .flags.bits.IS_HBR3_CAPABLE = true, .flags.bits.IS_TPS3_CAPABLE = true, @@ -679,6 +695,19 @@ static const struct dce_hwseq_mask hwseq_mask = { HWSEQ_DCE12_MASK_SH_LIST(_MASK) }; +/* HWSEQ regs for VG20 */ +static const struct dce_hwseq_registers dce121_hwseq_reg = { + HWSEQ_VG20_REG_LIST() +}; + +static const struct dce_hwseq_shift dce121_hwseq_shift = { + HWSEQ_VG20_MASK_SH_LIST(__SHIFT) +}; + +static const struct dce_hwseq_mask dce121_hwseq_mask = { + HWSEQ_VG20_MASK_SH_LIST(_MASK) +}; + static struct dce_hwseq *dce120_hwseq_create( struct dc_context *ctx) { @@ -693,6 +722,20 @@ static struct dce_hwseq *dce120_hwseq_create( return hws; } +static struct dce_hwseq *dce121_hwseq_create( + struct dc_context *ctx) +{ + struct dce_hwseq *hws = kzalloc(sizeof(struct dce_hwseq), GFP_KERNEL); + + if (hws) { + hws->ctx = ctx; + hws->regs = &dce121_hwseq_reg; + hws->shifts = &dce121_hwseq_shift; + hws->masks = &dce121_hwseq_mask; + } + return hws; +} + static const struct resource_create_funcs res_create_funcs = { .read_dce_straps = read_dce_straps, .create_audio = create_audio, @@ -700,6 +743,14 @@ static const struct resource_create_funcs res_create_funcs = { .create_hwseq = dce120_hwseq_create, }; +static const struct resource_create_funcs dce121_res_create_funcs = { + .read_dce_straps = read_dce_straps, + .create_audio = create_audio, + .create_stream_encoder = dce120_stream_encoder_create, + .create_hwseq = dce121_hwseq_create, +}; + + #define mi_inst_regs(id) { MI_DCE12_REG_LIST(id) } static const struct dce_mem_input_registers mi_regs[] = { mi_inst_regs(0), @@ -834,12 +885,12 @@ static void bw_calcs_data_update_from_pplib(struct dc *dc) * YCLK = UMACLK*m_memoryTypeMultiplier */ dc->bw_vbios->low_yclk = bw_frc_to_fixed( - mem_clks.data[0].clocks_in_khz * MEMORY_TYPE_MULTIPLIER, 1000); + mem_clks.data[0].clocks_in_khz * MEMORY_TYPE_MULTIPLIER_CZ, 1000); dc->bw_vbios->mid_yclk = bw_frc_to_fixed( - mem_clks.data[mem_clks.num_levels>>1].clocks_in_khz * MEMORY_TYPE_MULTIPLIER, + mem_clks.data[mem_clks.num_levels>>1].clocks_in_khz * MEMORY_TYPE_MULTIPLIER_CZ, 1000); dc->bw_vbios->high_yclk = bw_frc_to_fixed( - mem_clks.data[mem_clks.num_levels-1].clocks_in_khz * MEMORY_TYPE_MULTIPLIER, + mem_clks.data[mem_clks.num_levels-1].clocks_in_khz * MEMORY_TYPE_MULTIPLIER_CZ, 1000); /* Now notify PPLib/SMU about which Watermarks sets they should select @@ -909,7 +960,8 @@ static bool construct( int j; struct dc_context *ctx = dc->ctx; struct irq_service_init_data irq_init_data; - bool harvest_enabled = ASICREV_IS_VEGA20_P(ctx->asic_id.hw_internal_rev); + static const struct resource_create_funcs *res_funcs; + bool is_vg20 = ASICREV_IS_VEGA20_P(ctx->asic_id.hw_internal_rev); uint32_t pipe_fuses; ctx->dc_bios->regs = &bios_regs; @@ -973,8 +1025,12 @@ static bool construct( } } - pool->base.dccg = dce120_dccg_create(ctx); - if (pool->base.dccg == NULL) { + if (is_vg20) + pool->base.clk_mgr = dce121_clk_mgr_create(ctx); + else + pool->base.clk_mgr = dce120_clk_mgr_create(ctx); + + if (pool->base.clk_mgr == NULL) { dm_error("DC: failed to create display clock!\n"); BREAK_TO_DEBUGGER(); goto dccg_create_fail; @@ -1006,14 +1062,14 @@ static bool construct( if (!pool->base.irqs) goto irqs_create_fail; - /* retrieve valid pipe fuses */ - if (harvest_enabled) + /* VG20: Pipe harvesting enabled, retrieve valid pipe fuses */ + if (is_vg20) pipe_fuses = read_pipe_fuses(ctx); /* index to valid pipe resource */ j = 0; for (i = 0; i < pool->base.pipe_count; i++) { - if (harvest_enabled) { + if (is_vg20) { if ((pipe_fuses & (1 << i)) != 0) { dm_error("DC: skip invalid pipe %d!\n", i); continue; @@ -1091,10 +1147,24 @@ static bool construct( pool->base.pipe_count = j; pool->base.timing_generator_count = j; - if (!resource_construct(num_virtual_links, dc, &pool->base, - &res_create_funcs)) + if (is_vg20) + res_funcs = &dce121_res_create_funcs; + else + res_funcs = &res_create_funcs; + + if (!resource_construct(num_virtual_links, dc, &pool->base, res_funcs)) goto res_create_fail; + /* + * This is a bit of a hack. The xGMI enabled info is used to determine + * if audio and display clocks need to be adjusted with the WAFL link's + * SS info. This is a responsiblity of the clk_mgr. But since MMHUB is + * under hwseq, and the relevant register is in MMHUB, we have to do it + * here. + */ + if (is_vg20 && dce121_xgmi_enabled(dc->hwseq)) + dce121_clock_patch_xgmi_ss_info(pool->base.clk_mgr); + /* Create hardware sequencer */ if (!dce120_hw_sequencer_create(dc)) goto controller_create_fail; diff --git a/drivers/gpu/drm/amd/display/dc/dce80/dce80_hw_sequencer.c b/drivers/gpu/drm/amd/display/dc/dce80/dce80_hw_sequencer.c index 6c6a1a16af19..c4543178ba20 100644 --- a/drivers/gpu/drm/amd/display/dc/dce80/dce80_hw_sequencer.c +++ b/drivers/gpu/drm/amd/display/dc/dce80/dce80_hw_sequencer.c @@ -76,6 +76,7 @@ void dce80_hw_sequencer_construct(struct dc *dc) dc->hwss.enable_display_power_gating = dce100_enable_display_power_gating; dc->hwss.pipe_control_lock = dce_pipe_control_lock; - dc->hwss.set_bandwidth = dce100_set_bandwidth; + dc->hwss.prepare_bandwidth = dce100_prepare_bandwidth; + dc->hwss.optimize_bandwidth = dce100_optimize_bandwidth; } diff --git a/drivers/gpu/drm/amd/display/dc/dce80/dce80_resource.c b/drivers/gpu/drm/amd/display/dc/dce80/dce80_resource.c index d68f951f9869..c109ace96be9 100644 --- a/drivers/gpu/drm/amd/display/dc/dce80/dce80_resource.c +++ b/drivers/gpu/drm/amd/display/dc/dce80/dce80_resource.c @@ -37,14 +37,13 @@ #include "dce110/dce110_timing_generator.h" #include "dce110/dce110_resource.h" #include "dce80/dce80_timing_generator.h" +#include "dce/dce_clk_mgr.h" #include "dce/dce_mem_input.h" #include "dce/dce_link_encoder.h" #include "dce/dce_stream_encoder.h" -#include "dce/dce_mem_input.h" #include "dce/dce_ipp.h" #include "dce/dce_transform.h" #include "dce/dce_opp.h" -#include "dce/dce_clocks.h" #include "dce/dce_clock_source.h" #include "dce/dce_audio.h" #include "dce/dce_hwseq.h" @@ -78,6 +77,7 @@ #ifndef mmBIOS_SCRATCH_2 #define mmBIOS_SCRATCH_2 0x05CB + #define mmBIOS_SCRATCH_3 0x05CC #define mmBIOS_SCRATCH_6 0x05CF #endif @@ -155,15 +155,15 @@ static const struct dce110_timing_generator_offsets dce80_tg_offsets[] = { .reg_name = mm ## block ## id ## _ ## reg_name -static const struct dccg_registers disp_clk_regs = { +static const struct clk_mgr_registers disp_clk_regs = { CLK_COMMON_REG_LIST_DCE_BASE() }; -static const struct dccg_shift disp_clk_shift = { +static const struct clk_mgr_shift disp_clk_shift = { CLK_COMMON_MASK_SH_LIST_DCE_COMMON_BASE(__SHIFT) }; -static const struct dccg_mask disp_clk_mask = { +static const struct clk_mgr_mask disp_clk_mask = { CLK_COMMON_MASK_SH_LIST_DCE_COMMON_BASE(_MASK) }; @@ -359,6 +359,7 @@ static const struct dce110_clk_src_mask cs_mask = { }; static const struct bios_registers bios_regs = { + .BIOS_SCRATCH_3 = mmBIOS_SCRATCH_3, .BIOS_SCRATCH_6 = mmBIOS_SCRATCH_6 }; @@ -468,7 +469,7 @@ static struct output_pixel_processor *dce80_opp_create( return &opp->base; } -struct aux_engine *dce80_aux_engine_create( +struct dce_aux *dce80_aux_engine_create( struct dc_context *ctx, uint32_t inst) { @@ -779,8 +780,8 @@ static void destruct(struct dce110_resource_pool *pool) } } - if (pool->base.dccg != NULL) - dce_dccg_destroy(&pool->base.dccg); + if (pool->base.clk_mgr != NULL) + dce_clk_mgr_destroy(&pool->base.clk_mgr); if (pool->base.irqs != NULL) { dal_irq_service_destroy(&pool->base.irqs); @@ -791,9 +792,22 @@ bool dce80_validate_bandwidth( struct dc *dc, struct dc_state *context) { - /* TODO implement when needed but for now hardcode max value*/ - context->bw.dce.dispclk_khz = 681000; - context->bw.dce.yclk_khz = 250000 * MEMORY_TYPE_MULTIPLIER; + int i; + bool at_least_one_pipe = false; + + for (i = 0; i < dc->res_pool->pipe_count; i++) { + if (context->res_ctx.pipe_ctx[i].stream) + at_least_one_pipe = true; + } + + if (at_least_one_pipe) { + /* TODO implement when needed but for now hardcode max value*/ + context->bw.dce.dispclk_khz = 681000; + context->bw.dce.yclk_khz = 250000 * MEMORY_TYPE_MULTIPLIER_CZ; + } else { + context->bw.dce.dispclk_khz = 0; + context->bw.dce.yclk_khz = 0; + } return true; } @@ -855,7 +869,6 @@ static bool dce80_construct( struct dc_context *ctx = dc->ctx; struct dc_firmware_info info; struct dc_bios *bp; - struct dm_pp_static_clock_info static_clk_info = {0}; ctx->dc_bios->regs = &bios_regs; @@ -918,11 +931,11 @@ static bool dce80_construct( } } - pool->base.dccg = dce_dccg_create(ctx, + pool->base.clk_mgr = dce_clk_mgr_create(ctx, &disp_clk_regs, &disp_clk_shift, &disp_clk_mask); - if (pool->base.dccg == NULL) { + if (pool->base.clk_mgr == NULL) { dm_error("DC: failed to create display clock!\n"); BREAK_TO_DEBUGGER(); goto res_create_fail; @@ -948,10 +961,6 @@ static bool dce80_construct( goto res_create_fail; } - if (dm_pp_get_static_clocks(ctx, &static_clk_info)) - pool->base.dccg->max_clks_state = - static_clk_info.max_clocks_state; - { struct irq_service_init_data init_data; init_data.ctx = dc->ctx; @@ -1065,7 +1074,6 @@ static bool dce81_construct( struct dc_context *ctx = dc->ctx; struct dc_firmware_info info; struct dc_bios *bp; - struct dm_pp_static_clock_info static_clk_info = {0}; ctx->dc_bios->regs = &bios_regs; @@ -1128,11 +1136,11 @@ static bool dce81_construct( } } - pool->base.dccg = dce_dccg_create(ctx, + pool->base.clk_mgr = dce_clk_mgr_create(ctx, &disp_clk_regs, &disp_clk_shift, &disp_clk_mask); - if (pool->base.dccg == NULL) { + if (pool->base.clk_mgr == NULL) { dm_error("DC: failed to create display clock!\n"); BREAK_TO_DEBUGGER(); goto res_create_fail; @@ -1158,10 +1166,6 @@ static bool dce81_construct( goto res_create_fail; } - if (dm_pp_get_static_clocks(ctx, &static_clk_info)) - pool->base.dccg->max_clks_state = - static_clk_info.max_clocks_state; - { struct irq_service_init_data init_data; init_data.ctx = dc->ctx; @@ -1275,7 +1279,6 @@ static bool dce83_construct( struct dc_context *ctx = dc->ctx; struct dc_firmware_info info; struct dc_bios *bp; - struct dm_pp_static_clock_info static_clk_info = {0}; ctx->dc_bios->regs = &bios_regs; @@ -1334,11 +1337,11 @@ static bool dce83_construct( } } - pool->base.dccg = dce_dccg_create(ctx, + pool->base.clk_mgr = dce_clk_mgr_create(ctx, &disp_clk_regs, &disp_clk_shift, &disp_clk_mask); - if (pool->base.dccg == NULL) { + if (pool->base.clk_mgr == NULL) { dm_error("DC: failed to create display clock!\n"); BREAK_TO_DEBUGGER(); goto res_create_fail; @@ -1364,10 +1367,6 @@ static bool dce83_construct( goto res_create_fail; } - if (dm_pp_get_static_clocks(ctx, &static_clk_info)) - pool->base.dccg->max_clks_state = - static_clk_info.max_clocks_state; - { struct irq_service_init_data init_data; init_data.ctx = dc->ctx; diff --git a/drivers/gpu/drm/amd/display/dc/dce80/dce80_timing_generator.c b/drivers/gpu/drm/amd/display/dc/dce80/dce80_timing_generator.c index 3ba4712a35ab..8b5ce557ee71 100644 --- a/drivers/gpu/drm/amd/display/dc/dce80/dce80_timing_generator.c +++ b/drivers/gpu/drm/amd/display/dc/dce80/dce80_timing_generator.c @@ -84,17 +84,17 @@ static const struct dce110_timing_generator_offsets reg_offsets[] = { #define DCP_REG(reg) (reg + tg110->offsets.dcp) #define DMIF_REG(reg) (reg + tg110->offsets.dmif) -static void program_pix_dur(struct timing_generator *tg, uint32_t pix_clk_khz) +static void program_pix_dur(struct timing_generator *tg, uint32_t pix_clk_100hz) { uint64_t pix_dur; uint32_t addr = mmDMIF_PG0_DPG_PIPE_ARBITRATION_CONTROL1 + DCE110TG_FROM_TG(tg)->offsets.dmif; uint32_t value = dm_read_reg(tg->ctx, addr); - if (pix_clk_khz == 0) + if (pix_clk_100hz == 0) return; - pix_dur = 1000000000 / pix_clk_khz; + pix_dur = div_u64(10000000000ull, pix_clk_100hz); set_reg_field_value( value, @@ -110,7 +110,7 @@ static void program_timing(struct timing_generator *tg, bool use_vbios) { if (!use_vbios) - program_pix_dur(tg, timing->pix_clk_khz); + program_pix_dur(tg, timing->pix_clk_100hz); dce110_tg_program_timing(tg, timing, use_vbios); } diff --git a/drivers/gpu/drm/amd/display/dc/dcn10/Makefile b/drivers/gpu/drm/amd/display/dc/dcn10/Makefile index 032f872be89c..55f293c8a3c0 100644 --- a/drivers/gpu/drm/amd/display/dc/dcn10/Makefile +++ b/drivers/gpu/drm/amd/display/dc/dcn10/Makefile @@ -24,7 +24,7 @@ DCN10 = dcn10_resource.o dcn10_ipp.o dcn10_hw_sequencer.o dcn10_hw_sequencer_debug.o \ dcn10_dpp.o dcn10_opp.o dcn10_optc.o \ - dcn10_hubp.o dcn10_mpc.o \ + dcn10_hubp.o dcn10_mpc.o dcn10_clk_mgr.o \ dcn10_dpp_dscl.o dcn10_dpp_cm.o dcn10_cm_common.o \ dcn10_hubbub.o dcn10_stream_encoder.o dcn10_link_encoder.o diff --git a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_clk_mgr.c b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_clk_mgr.c new file mode 100644 index 000000000000..afe8c42211cd --- /dev/null +++ b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_clk_mgr.c @@ -0,0 +1,328 @@ +/* + * Copyright 2018 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. + * + * Authors: AMD + * + */ + +#include "dcn10_clk_mgr.h" + +#include "reg_helper.h" +#include "core_types.h" + +#define TO_DCE_CLK_MGR(clocks)\ + container_of(clocks, struct dce_clk_mgr, base) + +#define REG(reg) \ + (clk_mgr_dce->regs->reg) + +#undef FN +#define FN(reg_name, field_name) \ + clk_mgr_dce->clk_mgr_shift->field_name, clk_mgr_dce->clk_mgr_mask->field_name + +#define CTX \ + clk_mgr_dce->base.ctx +#define DC_LOGGER \ + clk_mgr->ctx->logger + +void dcn1_pplib_apply_display_requirements( + struct dc *dc, + struct dc_state *context) +{ + struct dm_pp_display_configuration *pp_display_cfg = &context->pp_display_cfg; + + pp_display_cfg->min_engine_clock_khz = dc->res_pool->clk_mgr->clks.dcfclk_khz; + pp_display_cfg->min_memory_clock_khz = dc->res_pool->clk_mgr->clks.fclk_khz; + pp_display_cfg->min_engine_clock_deep_sleep_khz = dc->res_pool->clk_mgr->clks.dcfclk_deep_sleep_khz; + pp_display_cfg->min_dcfc_deep_sleep_clock_khz = dc->res_pool->clk_mgr->clks.dcfclk_deep_sleep_khz; + pp_display_cfg->min_dcfclock_khz = dc->res_pool->clk_mgr->clks.dcfclk_khz; + pp_display_cfg->disp_clk_khz = dc->res_pool->clk_mgr->clks.dispclk_khz; + dce110_fill_display_configs(context, pp_display_cfg); + + dm_pp_apply_display_requirements(dc->ctx, pp_display_cfg); +} + +static int dcn1_determine_dppclk_threshold(struct clk_mgr *clk_mgr, struct dc_clocks *new_clocks) +{ + bool request_dpp_div = new_clocks->dispclk_khz > new_clocks->dppclk_khz; + bool dispclk_increase = new_clocks->dispclk_khz > clk_mgr->clks.dispclk_khz; + int disp_clk_threshold = new_clocks->max_supported_dppclk_khz; + bool cur_dpp_div = clk_mgr->clks.dispclk_khz > clk_mgr->clks.dppclk_khz; + + /* increase clock, looking for div is 0 for current, request div is 1*/ + if (dispclk_increase) { + /* already divided by 2, no need to reach target clk with 2 steps*/ + if (cur_dpp_div) + return new_clocks->dispclk_khz; + + /* request disp clk is lower than maximum supported dpp clk, + * no need to reach target clk with two steps. + */ + if (new_clocks->dispclk_khz <= disp_clk_threshold) + return new_clocks->dispclk_khz; + + /* target dpp clk not request divided by 2, still within threshold */ + if (!request_dpp_div) + return new_clocks->dispclk_khz; + + } else { + /* decrease clock, looking for current dppclk divided by 2, + * request dppclk not divided by 2. + */ + + /* current dpp clk not divided by 2, no need to ramp*/ + if (!cur_dpp_div) + return new_clocks->dispclk_khz; + + /* current disp clk is lower than current maximum dpp clk, + * no need to ramp + */ + if (clk_mgr->clks.dispclk_khz <= disp_clk_threshold) + return new_clocks->dispclk_khz; + + /* request dpp clk need to be divided by 2 */ + if (request_dpp_div) + return new_clocks->dispclk_khz; + } + + return disp_clk_threshold; +} + +static void dcn1_ramp_up_dispclk_with_dpp(struct clk_mgr *clk_mgr, struct dc_clocks *new_clocks) +{ + struct dc *dc = clk_mgr->ctx->dc; + int dispclk_to_dpp_threshold = dcn1_determine_dppclk_threshold(clk_mgr, new_clocks); + bool request_dpp_div = new_clocks->dispclk_khz > new_clocks->dppclk_khz; + int i; + + /* set disp clk to dpp clk threshold */ + dce112_set_clock(clk_mgr, dispclk_to_dpp_threshold); + + /* update request dpp clk division option */ + for (i = 0; i < dc->res_pool->pipe_count; i++) { + struct pipe_ctx *pipe_ctx = &dc->current_state->res_ctx.pipe_ctx[i]; + + if (!pipe_ctx->plane_state) + continue; + + pipe_ctx->plane_res.dpp->funcs->dpp_dppclk_control( + pipe_ctx->plane_res.dpp, + request_dpp_div, + true); + } + + /* If target clk not same as dppclk threshold, set to target clock */ + if (dispclk_to_dpp_threshold != new_clocks->dispclk_khz) + dce112_set_clock(clk_mgr, new_clocks->dispclk_khz); + + clk_mgr->clks.dispclk_khz = new_clocks->dispclk_khz; + clk_mgr->clks.dppclk_khz = new_clocks->dppclk_khz; + clk_mgr->clks.max_supported_dppclk_khz = new_clocks->max_supported_dppclk_khz; +} + +static int get_active_display_cnt( + struct dc *dc, + struct dc_state *context) +{ + int i, display_count; + + display_count = 0; + for (i = 0; i < context->stream_count; i++) { + const struct dc_stream_state *stream = context->streams[i]; + + /* + * Only notify active stream or virtual stream. + * Need to notify virtual stream to work around + * headless case. HPD does not fire when system is in + * S0i2. + */ + if (!stream->dpms_off || stream->signal == SIGNAL_TYPE_VIRTUAL) + display_count++; + } + + return display_count; +} + +static void dcn1_update_clocks(struct clk_mgr *clk_mgr, + struct dc_state *context, + bool safe_to_lower) +{ + struct dc *dc = clk_mgr->ctx->dc; + struct dc_debug_options *debug = &dc->debug; + struct dc_clocks *new_clocks = &context->bw.dcn.clk; + struct pp_smu_display_requirement_rv *smu_req_cur = + &dc->res_pool->pp_smu_req; + struct pp_smu_display_requirement_rv smu_req = *smu_req_cur; + struct pp_smu_funcs_rv *pp_smu = dc->res_pool->pp_smu; + bool send_request_to_increase = false; + bool send_request_to_lower = false; + int display_count; + + bool enter_display_off = false; + + display_count = get_active_display_cnt(dc, context); + + if (display_count == 0) + enter_display_off = true; + + if (enter_display_off == safe_to_lower) { + /* + * Notify SMU active displays + * if function pointer not set up, this message is + * sent as part of pplib_apply_display_requirements. + */ + if (pp_smu->set_display_count) + pp_smu->set_display_count(&pp_smu->pp_smu, display_count); + + smu_req.display_count = display_count; + } + + if (new_clocks->dispclk_khz > clk_mgr->clks.dispclk_khz + || new_clocks->phyclk_khz > clk_mgr->clks.phyclk_khz + || new_clocks->fclk_khz > clk_mgr->clks.fclk_khz + || new_clocks->dcfclk_khz > clk_mgr->clks.dcfclk_khz) + send_request_to_increase = true; + + if (should_set_clock(safe_to_lower, new_clocks->phyclk_khz, clk_mgr->clks.phyclk_khz)) { + clk_mgr->clks.phyclk_khz = new_clocks->phyclk_khz; + + send_request_to_lower = true; + } + + // F Clock + if (debug->force_fclk_khz != 0) + new_clocks->fclk_khz = debug->force_fclk_khz; + + if (should_set_clock(safe_to_lower, new_clocks->fclk_khz, clk_mgr->clks.fclk_khz)) { + clk_mgr->clks.fclk_khz = new_clocks->fclk_khz; + smu_req.hard_min_fclk_mhz = new_clocks->fclk_khz / 1000; + + send_request_to_lower = true; + } + + //DCF Clock + if (should_set_clock(safe_to_lower, new_clocks->dcfclk_khz, clk_mgr->clks.dcfclk_khz)) { + clk_mgr->clks.dcfclk_khz = new_clocks->dcfclk_khz; + smu_req.hard_min_dcefclk_mhz = new_clocks->dcfclk_khz / 1000; + + send_request_to_lower = true; + } + + if (should_set_clock(safe_to_lower, + new_clocks->dcfclk_deep_sleep_khz, clk_mgr->clks.dcfclk_deep_sleep_khz)) { + clk_mgr->clks.dcfclk_deep_sleep_khz = new_clocks->dcfclk_deep_sleep_khz; + smu_req.min_deep_sleep_dcefclk_mhz = (new_clocks->dcfclk_deep_sleep_khz + 999) / 1000; + + send_request_to_lower = true; + } + + /* make sure dcf clk is before dpp clk to + * make sure we have enough voltage to run dpp clk + */ + if (send_request_to_increase) { + /*use dcfclk to request voltage*/ + if (pp_smu->set_hard_min_fclk_by_freq && + pp_smu->set_hard_min_dcfclk_by_freq && + pp_smu->set_min_deep_sleep_dcfclk) { + + pp_smu->set_hard_min_fclk_by_freq(&pp_smu->pp_smu, smu_req.hard_min_fclk_mhz); + pp_smu->set_hard_min_dcfclk_by_freq(&pp_smu->pp_smu, smu_req.hard_min_dcefclk_mhz); + pp_smu->set_min_deep_sleep_dcfclk(&pp_smu->pp_smu, smu_req.min_deep_sleep_dcefclk_mhz); + } else { + if (pp_smu->set_display_requirement) + pp_smu->set_display_requirement(&pp_smu->pp_smu, &smu_req); + dcn1_pplib_apply_display_requirements(dc, context); + } + } + + /* dcn1 dppclk is tied to dispclk */ + /* program dispclk on = as a w/a for sleep resume clock ramping issues */ + if (should_set_clock(safe_to_lower, new_clocks->dispclk_khz, clk_mgr->clks.dispclk_khz) + || new_clocks->dispclk_khz == clk_mgr->clks.dispclk_khz) { + dcn1_ramp_up_dispclk_with_dpp(clk_mgr, new_clocks); + clk_mgr->clks.dispclk_khz = new_clocks->dispclk_khz; + + send_request_to_lower = true; + } + + if (!send_request_to_increase && send_request_to_lower) { + /*use dcfclk to request voltage*/ + if (pp_smu->set_hard_min_fclk_by_freq && + pp_smu->set_hard_min_dcfclk_by_freq && + pp_smu->set_min_deep_sleep_dcfclk) { + + pp_smu->set_hard_min_fclk_by_freq(&pp_smu->pp_smu, smu_req.hard_min_fclk_mhz); + pp_smu->set_hard_min_dcfclk_by_freq(&pp_smu->pp_smu, smu_req.hard_min_dcefclk_mhz); + pp_smu->set_min_deep_sleep_dcfclk(&pp_smu->pp_smu, smu_req.min_deep_sleep_dcefclk_mhz); + } else { + if (pp_smu->set_display_requirement) + pp_smu->set_display_requirement(&pp_smu->pp_smu, &smu_req); + dcn1_pplib_apply_display_requirements(dc, context); + } + } + + *smu_req_cur = smu_req; +} +static const struct clk_mgr_funcs dcn1_funcs = { + .get_dp_ref_clk_frequency = dce12_get_dp_ref_freq_khz, + .update_clocks = dcn1_update_clocks +}; +struct clk_mgr *dcn1_clk_mgr_create(struct dc_context *ctx) +{ + struct dc_debug_options *debug = &ctx->dc->debug; + struct dc_bios *bp = ctx->dc_bios; + struct dc_firmware_info fw_info = { { 0 } }; + struct dce_clk_mgr *clk_mgr_dce = kzalloc(sizeof(*clk_mgr_dce), GFP_KERNEL); + + if (clk_mgr_dce == NULL) { + BREAK_TO_DEBUGGER(); + return NULL; + } + + clk_mgr_dce->base.ctx = ctx; + clk_mgr_dce->base.funcs = &dcn1_funcs; + + clk_mgr_dce->dfs_bypass_disp_clk = 0; + + clk_mgr_dce->dprefclk_ss_percentage = 0; + clk_mgr_dce->dprefclk_ss_divider = 1000; + clk_mgr_dce->ss_on_dprefclk = false; + + clk_mgr_dce->dprefclk_khz = 600000; + if (bp->integrated_info) + clk_mgr_dce->dentist_vco_freq_khz = bp->integrated_info->dentist_vco_freq; + if (clk_mgr_dce->dentist_vco_freq_khz == 0) { + bp->funcs->get_firmware_info(bp, &fw_info); + clk_mgr_dce->dentist_vco_freq_khz = fw_info.smu_gpu_pll_output_freq; + if (clk_mgr_dce->dentist_vco_freq_khz == 0) + clk_mgr_dce->dentist_vco_freq_khz = 3600000; + } + + if (!debug->disable_dfs_bypass && bp->integrated_info) + if (bp->integrated_info->gpu_cap_info & DFS_BYPASS_ENABLE) + clk_mgr_dce->dfs_bypass_enabled = true; + + dce_clock_read_ss_info(clk_mgr_dce); + + return &clk_mgr_dce->base; +} + + diff --git a/drivers/gpu/drm/amd/display/dc/i2caux/dce112/i2caux_dce112.h b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_clk_mgr.h index 8d35453c25b6..a995eda443a3 100644 --- a/drivers/gpu/drm/amd/display/dc/i2caux/dce112/i2caux_dce112.h +++ b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_clk_mgr.h @@ -1,5 +1,5 @@ /* - * Copyright 2012-15 Advanced Micro Devices, Inc. + * Copyright 2018 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"), @@ -23,10 +23,21 @@ * */ -#ifndef __DAL_I2C_AUX_DCE112_H__ -#define __DAL_I2C_AUX_DCE112_H__ +#ifndef __DCN10_CLK_MGR_H__ +#define __DCN10_CLK_MGR_H__ -struct i2caux *dal_i2caux_dce112_create( - struct dc_context *ctx); +#include "../dce/dce_clk_mgr.h" -#endif /* __DAL_I2C_AUX_DCE112_H__ */ +struct clk_bypass { + uint32_t dcfclk_bypass; + uint32_t dispclk_pypass; + uint32_t dprefclk_bypass; +}; + +void dcn1_pplib_apply_display_requirements( + struct dc *dc, + struct dc_state *context); + +struct clk_mgr *dcn1_clk_mgr_create(struct dc_context *ctx); + +#endif //__DCN10_CLK_MGR_H__ diff --git a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_cm_common.c b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_cm_common.c index 5d95a997fd9f..7469333a2c8a 100644 --- a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_cm_common.c +++ b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_cm_common.c @@ -71,39 +71,39 @@ void cm_helper_program_xfer_func( unsigned int i = 0; REG_SET_2(reg->start_cntl_b, 0, - exp_region_start, params->arr_points[0].custom_float_x, + exp_region_start, params->corner_points[0].blue.custom_float_x, exp_resion_start_segment, 0); REG_SET_2(reg->start_cntl_g, 0, - exp_region_start, params->arr_points[0].custom_float_x, + exp_region_start, params->corner_points[0].green.custom_float_x, exp_resion_start_segment, 0); REG_SET_2(reg->start_cntl_r, 0, - exp_region_start, params->arr_points[0].custom_float_x, + exp_region_start, params->corner_points[0].red.custom_float_x, exp_resion_start_segment, 0); REG_SET(reg->start_slope_cntl_b, 0, - field_region_linear_slope, params->arr_points[0].custom_float_slope); + field_region_linear_slope, params->corner_points[0].blue.custom_float_slope); REG_SET(reg->start_slope_cntl_g, 0, - field_region_linear_slope, params->arr_points[0].custom_float_slope); + field_region_linear_slope, params->corner_points[0].green.custom_float_slope); REG_SET(reg->start_slope_cntl_r, 0, - field_region_linear_slope, params->arr_points[0].custom_float_slope); + field_region_linear_slope, params->corner_points[0].red.custom_float_slope); REG_SET(reg->start_end_cntl1_b, 0, - field_region_end, params->arr_points[1].custom_float_x); + field_region_end, params->corner_points[1].blue.custom_float_x); REG_SET_2(reg->start_end_cntl2_b, 0, - field_region_end_slope, params->arr_points[1].custom_float_slope, - field_region_end_base, params->arr_points[1].custom_float_y); + field_region_end_slope, params->corner_points[1].blue.custom_float_slope, + field_region_end_base, params->corner_points[1].blue.custom_float_y); REG_SET(reg->start_end_cntl1_g, 0, - field_region_end, params->arr_points[1].custom_float_x); + field_region_end, params->corner_points[1].green.custom_float_x); REG_SET_2(reg->start_end_cntl2_g, 0, - field_region_end_slope, params->arr_points[1].custom_float_slope, - field_region_end_base, params->arr_points[1].custom_float_y); + field_region_end_slope, params->corner_points[1].green.custom_float_slope, + field_region_end_base, params->corner_points[1].green.custom_float_y); REG_SET(reg->start_end_cntl1_r, 0, - field_region_end, params->arr_points[1].custom_float_x); + field_region_end, params->corner_points[1].red.custom_float_x); REG_SET_2(reg->start_end_cntl2_r, 0, - field_region_end_slope, params->arr_points[1].custom_float_slope, - field_region_end_base, params->arr_points[1].custom_float_y); + field_region_end_slope, params->corner_points[1].red.custom_float_slope, + field_region_end_base, params->corner_points[1].red.custom_float_y); for (reg_region_cur = reg->region_start; reg_region_cur <= reg->region_end; @@ -127,7 +127,7 @@ void cm_helper_program_xfer_func( bool cm_helper_convert_to_custom_float( struct pwl_result_data *rgb_resulted, - struct curve_points *arr_points, + struct curve_points3 *corner_points, uint32_t hw_points_num, bool fixpoint) { @@ -141,20 +141,53 @@ bool cm_helper_convert_to_custom_float( fmt.mantissa_bits = 12; fmt.sign = false; - if (!convert_to_custom_float_format(arr_points[0].x, &fmt, - &arr_points[0].custom_float_x)) { + /* corner_points[0] - beginning base, slope offset for R,G,B + * corner_points[1] - end base, slope offset for R,G,B + */ + if (!convert_to_custom_float_format(corner_points[0].red.x, &fmt, + &corner_points[0].red.custom_float_x)) { + BREAK_TO_DEBUGGER(); + return false; + } + if (!convert_to_custom_float_format(corner_points[0].green.x, &fmt, + &corner_points[0].green.custom_float_x)) { + BREAK_TO_DEBUGGER(); + return false; + } + if (!convert_to_custom_float_format(corner_points[0].blue.x, &fmt, + &corner_points[0].blue.custom_float_x)) { BREAK_TO_DEBUGGER(); return false; } - if (!convert_to_custom_float_format(arr_points[0].offset, &fmt, - &arr_points[0].custom_float_offset)) { + if (!convert_to_custom_float_format(corner_points[0].red.offset, &fmt, + &corner_points[0].red.custom_float_offset)) { + BREAK_TO_DEBUGGER(); + return false; + } + if (!convert_to_custom_float_format(corner_points[0].green.offset, &fmt, + &corner_points[0].green.custom_float_offset)) { + BREAK_TO_DEBUGGER(); + return false; + } + if (!convert_to_custom_float_format(corner_points[0].blue.offset, &fmt, + &corner_points[0].blue.custom_float_offset)) { BREAK_TO_DEBUGGER(); return false; } - if (!convert_to_custom_float_format(arr_points[0].slope, &fmt, - &arr_points[0].custom_float_slope)) { + if (!convert_to_custom_float_format(corner_points[0].red.slope, &fmt, + &corner_points[0].red.custom_float_slope)) { + BREAK_TO_DEBUGGER(); + return false; + } + if (!convert_to_custom_float_format(corner_points[0].green.slope, &fmt, + &corner_points[0].green.custom_float_slope)) { + BREAK_TO_DEBUGGER(); + return false; + } + if (!convert_to_custom_float_format(corner_points[0].blue.slope, &fmt, + &corner_points[0].blue.custom_float_slope)) { BREAK_TO_DEBUGGER(); return false; } @@ -162,22 +195,59 @@ bool cm_helper_convert_to_custom_float( fmt.mantissa_bits = 10; fmt.sign = false; - if (!convert_to_custom_float_format(arr_points[1].x, &fmt, - &arr_points[1].custom_float_x)) { + if (!convert_to_custom_float_format(corner_points[1].red.x, &fmt, + &corner_points[1].red.custom_float_x)) { BREAK_TO_DEBUGGER(); return false; } - - if (fixpoint == true) - arr_points[1].custom_float_y = dc_fixpt_clamp_u0d14(arr_points[1].y); - else if (!convert_to_custom_float_format(arr_points[1].y, &fmt, - &arr_points[1].custom_float_y)) { + if (!convert_to_custom_float_format(corner_points[1].green.x, &fmt, + &corner_points[1].green.custom_float_x)) { + BREAK_TO_DEBUGGER(); + return false; + } + if (!convert_to_custom_float_format(corner_points[1].blue.x, &fmt, + &corner_points[1].blue.custom_float_x)) { BREAK_TO_DEBUGGER(); return false; } - if (!convert_to_custom_float_format(arr_points[1].slope, &fmt, - &arr_points[1].custom_float_slope)) { + if (fixpoint == true) { + corner_points[1].red.custom_float_y = + dc_fixpt_clamp_u0d14(corner_points[1].red.y); + corner_points[1].green.custom_float_y = + dc_fixpt_clamp_u0d14(corner_points[1].green.y); + corner_points[1].blue.custom_float_y = + dc_fixpt_clamp_u0d14(corner_points[1].blue.y); + } else { + if (!convert_to_custom_float_format(corner_points[1].red.y, + &fmt, &corner_points[1].red.custom_float_y)) { + BREAK_TO_DEBUGGER(); + return false; + } + if (!convert_to_custom_float_format(corner_points[1].green.y, + &fmt, &corner_points[1].green.custom_float_y)) { + BREAK_TO_DEBUGGER(); + return false; + } + if (!convert_to_custom_float_format(corner_points[1].blue.y, + &fmt, &corner_points[1].blue.custom_float_y)) { + BREAK_TO_DEBUGGER(); + return false; + } + } + + if (!convert_to_custom_float_format(corner_points[1].red.slope, &fmt, + &corner_points[1].red.custom_float_slope)) { + BREAK_TO_DEBUGGER(); + return false; + } + if (!convert_to_custom_float_format(corner_points[1].green.slope, &fmt, + &corner_points[1].green.custom_float_slope)) { + BREAK_TO_DEBUGGER(); + return false; + } + if (!convert_to_custom_float_format(corner_points[1].blue.slope, &fmt, + &corner_points[1].blue.custom_float_slope)) { BREAK_TO_DEBUGGER(); return false; } @@ -242,15 +312,10 @@ bool cm_helper_translate_curve_to_hw_format( const struct dc_transfer_func *output_tf, struct pwl_params *lut_params, bool fixpoint) { - struct curve_points *arr_points; + struct curve_points3 *corner_points; struct pwl_result_data *rgb_resulted; struct pwl_result_data *rgb; struct pwl_result_data *rgb_plus_1; - struct fixed31_32 y_r; - struct fixed31_32 y_g; - struct fixed31_32 y_b; - struct fixed31_32 y1_min; - struct fixed31_32 y3_max; int32_t region_start, region_end; int32_t i; @@ -259,16 +324,16 @@ bool cm_helper_translate_curve_to_hw_format( if (output_tf == NULL || lut_params == NULL || output_tf->type == TF_TYPE_BYPASS) return false; - PERF_TRACE(); + PERF_TRACE_CTX(output_tf->ctx); - arr_points = lut_params->arr_points; + corner_points = lut_params->corner_points; rgb_resulted = lut_params->rgb_resulted; hw_points = 0; memset(lut_params, 0, sizeof(struct pwl_params)); memset(seg_distr, 0, sizeof(seg_distr)); - if (output_tf->tf == TRANSFER_FUNCTION_PQ) { + if (output_tf->tf == TRANSFER_FUNCTION_PQ || output_tf->tf == TRANSFER_FUNCTION_GAMMA22) { /* 32 segments * segments are from 2^-25 to 2^7 */ @@ -327,31 +392,37 @@ bool cm_helper_translate_curve_to_hw_format( rgb_resulted[hw_points - 1].green = output_tf->tf_pts.green[start_index]; rgb_resulted[hw_points - 1].blue = output_tf->tf_pts.blue[start_index]; - arr_points[0].x = dc_fixpt_pow(dc_fixpt_from_int(2), + // All 3 color channels have same x + corner_points[0].red.x = dc_fixpt_pow(dc_fixpt_from_int(2), dc_fixpt_from_int(region_start)); - arr_points[1].x = dc_fixpt_pow(dc_fixpt_from_int(2), - dc_fixpt_from_int(region_end)); + corner_points[0].green.x = corner_points[0].red.x; + corner_points[0].blue.x = corner_points[0].red.x; - y_r = rgb_resulted[0].red; - y_g = rgb_resulted[0].green; - y_b = rgb_resulted[0].blue; + corner_points[1].red.x = dc_fixpt_pow(dc_fixpt_from_int(2), + dc_fixpt_from_int(region_end)); + corner_points[1].green.x = corner_points[1].red.x; + corner_points[1].blue.x = corner_points[1].red.x; - y1_min = dc_fixpt_min(y_r, dc_fixpt_min(y_g, y_b)); + corner_points[0].red.y = rgb_resulted[0].red; + corner_points[0].green.y = rgb_resulted[0].green; + corner_points[0].blue.y = rgb_resulted[0].blue; - arr_points[0].y = y1_min; - arr_points[0].slope = dc_fixpt_div(arr_points[0].y, arr_points[0].x); - y_r = rgb_resulted[hw_points - 1].red; - y_g = rgb_resulted[hw_points - 1].green; - y_b = rgb_resulted[hw_points - 1].blue; + corner_points[0].red.slope = dc_fixpt_div(corner_points[0].red.y, + corner_points[0].red.x); + corner_points[0].green.slope = dc_fixpt_div(corner_points[0].green.y, + corner_points[0].green.x); + corner_points[0].blue.slope = dc_fixpt_div(corner_points[0].blue.y, + corner_points[0].blue.x); /* see comment above, m_arrPoints[1].y should be the Y value for the * region end (m_numOfHwPoints), not last HW point(m_numOfHwPoints - 1) */ - y3_max = dc_fixpt_max(y_r, dc_fixpt_max(y_g, y_b)); - - arr_points[1].y = y3_max; - - arr_points[1].slope = dc_fixpt_zero; + corner_points[1].red.y = rgb_resulted[hw_points - 1].red; + corner_points[1].green.y = rgb_resulted[hw_points - 1].green; + corner_points[1].blue.y = rgb_resulted[hw_points - 1].blue; + corner_points[1].red.slope = dc_fixpt_zero; + corner_points[1].green.slope = dc_fixpt_zero; + corner_points[1].blue.slope = dc_fixpt_zero; if (output_tf->tf == TRANSFER_FUNCTION_PQ) { /* for PQ, we want to have a straight line from last HW X point, @@ -360,9 +431,15 @@ bool cm_helper_translate_curve_to_hw_format( const struct fixed31_32 end_value = dc_fixpt_from_int(125); - arr_points[1].slope = dc_fixpt_div( - dc_fixpt_sub(dc_fixpt_one, arr_points[1].y), - dc_fixpt_sub(end_value, arr_points[1].x)); + corner_points[1].red.slope = dc_fixpt_div( + dc_fixpt_sub(dc_fixpt_one, corner_points[1].red.y), + dc_fixpt_sub(end_value, corner_points[1].red.x)); + corner_points[1].green.slope = dc_fixpt_div( + dc_fixpt_sub(dc_fixpt_one, corner_points[1].green.y), + dc_fixpt_sub(end_value, corner_points[1].green.x)); + corner_points[1].blue.slope = dc_fixpt_div( + dc_fixpt_sub(dc_fixpt_one, corner_points[1].blue.y), + dc_fixpt_sub(end_value, corner_points[1].blue.x)); } lut_params->hw_points_num = hw_points; @@ -411,7 +488,7 @@ bool cm_helper_translate_curve_to_hw_format( ++i; } cm_helper_convert_to_custom_float(rgb_resulted, - lut_params->arr_points, + lut_params->corner_points, hw_points, fixpoint); return true; @@ -424,15 +501,10 @@ bool cm_helper_translate_curve_to_degamma_hw_format( const struct dc_transfer_func *output_tf, struct pwl_params *lut_params) { - struct curve_points *arr_points; + struct curve_points3 *corner_points; struct pwl_result_data *rgb_resulted; struct pwl_result_data *rgb; struct pwl_result_data *rgb_plus_1; - struct fixed31_32 y_r; - struct fixed31_32 y_g; - struct fixed31_32 y_b; - struct fixed31_32 y1_min; - struct fixed31_32 y3_max; int32_t region_start, region_end; int32_t i; @@ -441,9 +513,9 @@ bool cm_helper_translate_curve_to_degamma_hw_format( if (output_tf == NULL || lut_params == NULL || output_tf->type == TF_TYPE_BYPASS) return false; - PERF_TRACE(); + PERF_TRACE_CTX(output_tf->ctx); - arr_points = lut_params->arr_points; + corner_points = lut_params->corner_points; rgb_resulted = lut_params->rgb_resulted; hw_points = 0; @@ -489,31 +561,28 @@ bool cm_helper_translate_curve_to_degamma_hw_format( rgb_resulted[hw_points - 1].green = output_tf->tf_pts.green[start_index]; rgb_resulted[hw_points - 1].blue = output_tf->tf_pts.blue[start_index]; - arr_points[0].x = dc_fixpt_pow(dc_fixpt_from_int(2), + corner_points[0].red.x = dc_fixpt_pow(dc_fixpt_from_int(2), dc_fixpt_from_int(region_start)); - arr_points[1].x = dc_fixpt_pow(dc_fixpt_from_int(2), + corner_points[0].green.x = corner_points[0].red.x; + corner_points[0].blue.x = corner_points[0].red.x; + corner_points[1].red.x = dc_fixpt_pow(dc_fixpt_from_int(2), dc_fixpt_from_int(region_end)); + corner_points[1].green.x = corner_points[1].red.x; + corner_points[1].blue.x = corner_points[1].red.x; - y_r = rgb_resulted[0].red; - y_g = rgb_resulted[0].green; - y_b = rgb_resulted[0].blue; - - y1_min = dc_fixpt_min(y_r, dc_fixpt_min(y_g, y_b)); - - arr_points[0].y = y1_min; - arr_points[0].slope = dc_fixpt_div(arr_points[0].y, arr_points[0].x); - y_r = rgb_resulted[hw_points - 1].red; - y_g = rgb_resulted[hw_points - 1].green; - y_b = rgb_resulted[hw_points - 1].blue; + corner_points[0].red.y = rgb_resulted[0].red; + corner_points[0].green.y = rgb_resulted[0].green; + corner_points[0].blue.y = rgb_resulted[0].blue; /* see comment above, m_arrPoints[1].y should be the Y value for the * region end (m_numOfHwPoints), not last HW point(m_numOfHwPoints - 1) */ - y3_max = dc_fixpt_max(y_r, dc_fixpt_max(y_g, y_b)); - - arr_points[1].y = y3_max; - - arr_points[1].slope = dc_fixpt_zero; + corner_points[1].red.y = rgb_resulted[hw_points - 1].red; + corner_points[1].green.y = rgb_resulted[hw_points - 1].green; + corner_points[1].blue.y = rgb_resulted[hw_points - 1].blue; + corner_points[1].red.slope = dc_fixpt_zero; + corner_points[1].green.slope = dc_fixpt_zero; + corner_points[1].blue.slope = dc_fixpt_zero; if (output_tf->tf == TRANSFER_FUNCTION_PQ) { /* for PQ, we want to have a straight line from last HW X point, @@ -522,9 +591,15 @@ bool cm_helper_translate_curve_to_degamma_hw_format( const struct fixed31_32 end_value = dc_fixpt_from_int(125); - arr_points[1].slope = dc_fixpt_div( - dc_fixpt_sub(dc_fixpt_one, arr_points[1].y), - dc_fixpt_sub(end_value, arr_points[1].x)); + corner_points[1].red.slope = dc_fixpt_div( + dc_fixpt_sub(dc_fixpt_one, corner_points[1].red.y), + dc_fixpt_sub(end_value, corner_points[1].red.x)); + corner_points[1].green.slope = dc_fixpt_div( + dc_fixpt_sub(dc_fixpt_one, corner_points[1].green.y), + dc_fixpt_sub(end_value, corner_points[1].green.x)); + corner_points[1].blue.slope = dc_fixpt_div( + dc_fixpt_sub(dc_fixpt_one, corner_points[1].blue.y), + dc_fixpt_sub(end_value, corner_points[1].blue.x)); } lut_params->hw_points_num = hw_points; @@ -564,7 +639,7 @@ bool cm_helper_translate_curve_to_degamma_hw_format( ++i; } cm_helper_convert_to_custom_float(rgb_resulted, - lut_params->arr_points, + lut_params->corner_points, hw_points, false); return true; diff --git a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_cm_common.h b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_cm_common.h index 7a531b02871f..5ae4d69391a5 100644 --- a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_cm_common.h +++ b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_cm_common.h @@ -98,7 +98,7 @@ void cm_helper_program_xfer_func( bool cm_helper_convert_to_custom_float( struct pwl_result_data *rgb_resulted, - struct curve_points *arr_points, + struct curve_points3 *corner_points, uint32_t hw_points_num, bool fixpoint); diff --git a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_dpp.c b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_dpp.c index dcb3c5530236..cd1ebe57ed59 100644 --- a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_dpp.c +++ b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_dpp.c @@ -463,7 +463,7 @@ void dpp1_set_cursor_position( if (src_y_offset >= (int)param->viewport.height) cur_en = 0; /* not visible beyond bottom edge*/ - if (src_y_offset < 0) + if (src_y_offset + (int)height <= 0) cur_en = 0; /* not visible beyond top edge*/ REG_UPDATE(CURSOR0_CONTROL, diff --git a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_dpp_cm.c b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_dpp_cm.c index 116977eb24e2..41f0f4c912e7 100644 --- a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_dpp_cm.c +++ b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_dpp_cm.c @@ -51,10 +51,6 @@ #define NUM_ELEMENTS(a) (sizeof(a) / sizeof((a)[0])) -struct dcn10_input_csc_matrix { - enum dc_color_space color_space; - uint16_t regval[12]; -}; enum dcn10_coef_filter_type_sel { SCL_COEF_LUMA_VERT_FILTER = 0, @@ -99,7 +95,7 @@ enum gamut_remap_select { GAMUT_REMAP_COMB_COEFF }; -static const struct dcn10_input_csc_matrix dcn10_input_csc_matrix[] = { +static const struct dpp_input_csc_matrix dpp_input_csc_matrix[] = { {COLOR_SPACE_SRGB, {0x2000, 0, 0, 0, 0, 0x2000, 0, 0, 0, 0, 0x2000, 0} }, {COLOR_SPACE_SRGB_LIMITED, @@ -454,7 +450,7 @@ void dpp1_program_input_csc( { struct dcn10_dpp *dpp = TO_DCN10_DPP(dpp_base); int i; - int arr_size = sizeof(dcn10_input_csc_matrix)/sizeof(struct dcn10_input_csc_matrix); + int arr_size = sizeof(dpp_input_csc_matrix)/sizeof(struct dpp_input_csc_matrix); const uint16_t *regval = NULL; uint32_t cur_select = 0; enum dcn10_input_csc_select select; @@ -467,8 +463,8 @@ void dpp1_program_input_csc( if (tbl_entry == NULL) { for (i = 0; i < arr_size; i++) - if (dcn10_input_csc_matrix[i].color_space == color_space) { - regval = dcn10_input_csc_matrix[i].regval; + if (dpp_input_csc_matrix[i].color_space == color_space) { + regval = dpp_input_csc_matrix[i].regval; break; } diff --git a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_dpp_dscl.c b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_dpp_dscl.c index 4a863a5dab41..c7642e748297 100644 --- a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_dpp_dscl.c +++ b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_dpp_dscl.c @@ -597,11 +597,13 @@ static void dpp1_dscl_set_manual_ratio_init( SCL_V_INIT_FRAC, init_frac, SCL_V_INIT_INT, init_int); - init_frac = dc_fixpt_u0d19(data->inits.v_bot) << 5; - init_int = dc_fixpt_floor(data->inits.v_bot); - REG_SET_2(SCL_VERT_FILTER_INIT_BOT, 0, - SCL_V_INIT_FRAC_BOT, init_frac, - SCL_V_INIT_INT_BOT, init_int); + if (REG(SCL_VERT_FILTER_INIT_BOT)) { + init_frac = dc_fixpt_u0d19(data->inits.v_bot) << 5; + init_int = dc_fixpt_floor(data->inits.v_bot); + REG_SET_2(SCL_VERT_FILTER_INIT_BOT, 0, + SCL_V_INIT_FRAC_BOT, init_frac, + SCL_V_INIT_INT_BOT, init_int); + } init_frac = dc_fixpt_u0d19(data->inits.v_c) << 5; init_int = dc_fixpt_floor(data->inits.v_c); @@ -609,11 +611,13 @@ static void dpp1_dscl_set_manual_ratio_init( SCL_V_INIT_FRAC_C, init_frac, SCL_V_INIT_INT_C, init_int); - init_frac = dc_fixpt_u0d19(data->inits.v_c_bot) << 5; - init_int = dc_fixpt_floor(data->inits.v_c_bot); - REG_SET_2(SCL_VERT_FILTER_INIT_BOT_C, 0, - SCL_V_INIT_FRAC_BOT_C, init_frac, - SCL_V_INIT_INT_BOT_C, init_int); + if (REG(SCL_VERT_FILTER_INIT_BOT_C)) { + init_frac = dc_fixpt_u0d19(data->inits.v_c_bot) << 5; + init_int = dc_fixpt_floor(data->inits.v_c_bot); + REG_SET_2(SCL_VERT_FILTER_INIT_BOT_C, 0, + SCL_V_INIT_FRAC_BOT_C, init_frac, + SCL_V_INIT_INT_BOT_C, init_int); + } } @@ -688,15 +692,17 @@ void dpp1_dscl_set_scaler_manual_scale( return; /* Black offsets */ - if (ycbcr) - REG_SET_2(SCL_BLACK_OFFSET, 0, - SCL_BLACK_OFFSET_RGB_Y, BLACK_OFFSET_RGB_Y, - SCL_BLACK_OFFSET_CBCR, BLACK_OFFSET_CBCR); - else + if (REG(SCL_BLACK_OFFSET)) { + if (ycbcr) + REG_SET_2(SCL_BLACK_OFFSET, 0, + SCL_BLACK_OFFSET_RGB_Y, BLACK_OFFSET_RGB_Y, + SCL_BLACK_OFFSET_CBCR, BLACK_OFFSET_CBCR); + else - REG_SET_2(SCL_BLACK_OFFSET, 0, - SCL_BLACK_OFFSET_RGB_Y, BLACK_OFFSET_RGB_Y, - SCL_BLACK_OFFSET_CBCR, BLACK_OFFSET_RGB_Y); + REG_SET_2(SCL_BLACK_OFFSET, 0, + SCL_BLACK_OFFSET_RGB_Y, BLACK_OFFSET_RGB_Y, + SCL_BLACK_OFFSET_CBCR, BLACK_OFFSET_RGB_Y); + } /* Manually calculate scale ratio and init values */ dpp1_dscl_set_manual_ratio_init(dpp, scl_data); diff --git a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_hubbub.c b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_hubbub.c index 4254e7e1a509..e161ad836812 100644 --- a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_hubbub.c +++ b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_hubbub.c @@ -29,19 +29,20 @@ #include "reg_helper.h" #define CTX \ - hubbub->ctx + hubbub1->base.ctx #define DC_LOGGER \ - hubbub->ctx->logger + hubbub1->base.ctx->logger #define REG(reg)\ - hubbub->regs->reg + hubbub1->regs->reg #undef FN #define FN(reg_name, field_name) \ - hubbub->shifts->field_name, hubbub->masks->field_name + hubbub1->shifts->field_name, hubbub1->masks->field_name void hubbub1_wm_read_state(struct hubbub *hubbub, struct dcn_hubbub_wm *wm) { + struct dcn10_hubbub *hubbub1 = TO_DCN10_HUBBUB(hubbub); struct dcn_hubbub_wm_set *s; memset(wm, 0, sizeof(struct dcn_hubbub_wm)); @@ -87,26 +88,37 @@ void hubbub1_wm_read_state(struct hubbub *hubbub, s->dram_clk_chanage = REG_READ(DCHUBBUB_ARB_ALLOW_DRAM_CLK_CHANGE_WATERMARK_D); } -void hubbub1_disable_allow_self_refresh(struct hubbub *hubbub) +void hubbub1_allow_self_refresh_control(struct hubbub *hubbub, bool allow) { - REG_UPDATE(DCHUBBUB_ARB_DRAM_STATE_CNTL, - DCHUBBUB_ARB_ALLOW_SELF_REFRESH_FORCE_ENABLE, 0); + struct dcn10_hubbub *hubbub1 = TO_DCN10_HUBBUB(hubbub); + + /* + * DCHUBBUB_ARB_ALLOW_SELF_REFRESH_FORCE_ENABLE = 1 means do not allow stutter + * DCHUBBUB_ARB_ALLOW_SELF_REFRESH_FORCE_ENABLE = 0 means allow stutter + */ + + REG_UPDATE_2(DCHUBBUB_ARB_DRAM_STATE_CNTL, + DCHUBBUB_ARB_ALLOW_SELF_REFRESH_FORCE_VALUE, 0, + DCHUBBUB_ARB_ALLOW_SELF_REFRESH_FORCE_ENABLE, !allow); } bool hububu1_is_allow_self_refresh_enabled(struct hubbub *hubbub) { + struct dcn10_hubbub *hubbub1 = TO_DCN10_HUBBUB(hubbub); uint32_t enable = 0; REG_GET(DCHUBBUB_ARB_DRAM_STATE_CNTL, DCHUBBUB_ARB_ALLOW_SELF_REFRESH_FORCE_ENABLE, &enable); - return true ? false : enable; + return enable ? true : false; } bool hubbub1_verify_allow_pstate_change_high( struct hubbub *hubbub) { + struct dcn10_hubbub *hubbub1 = TO_DCN10_HUBBUB(hubbub); + /* pstate latency is ~20us so if we wait over 40us and pstate allow * still not asserted, we are probably stuck and going to hang * @@ -193,7 +205,7 @@ bool hubbub1_verify_allow_pstate_change_high( * 31: SOC pstate change request */ - REG_WRITE(DCHUBBUB_TEST_DEBUG_INDEX, hubbub->debug_test_index_pstate); + REG_WRITE(DCHUBBUB_TEST_DEBUG_INDEX, hubbub1->debug_test_index_pstate); for (i = 0; i < pstate_wait_timeout_us; i++) { debug_data = REG_READ(DCHUBBUB_TEST_DEBUG_DATA); @@ -244,6 +256,8 @@ static uint32_t convert_and_clamp( void hubbub1_wm_change_req_wa(struct hubbub *hubbub) { + struct dcn10_hubbub *hubbub1 = TO_DCN10_HUBBUB(hubbub); + REG_UPDATE_SEQ(DCHUBBUB_ARB_WATERMARK_CHANGE_CNTL, DCHUBBUB_ARB_WATERMARK_CHANGE_REQUEST, 0, 1); } @@ -254,7 +268,7 @@ void hubbub1_program_watermarks( unsigned int refclk_mhz, bool safe_to_lower) { - uint32_t force_en = hubbub->ctx->dc->debug.disable_stutter ? 1 : 0; + struct dcn10_hubbub *hubbub1 = TO_DCN10_HUBBUB(hubbub); /* * Need to clamp to max of the register values (i.e. no wrap) * for dcn1, all wm registers are 21-bit wide @@ -264,8 +278,8 @@ void hubbub1_program_watermarks( /* Repeat for water mark set A, B, C and D. */ /* clock state A */ - if (safe_to_lower || watermarks->a.urgent_ns > hubbub->watermarks.a.urgent_ns) { - hubbub->watermarks.a.urgent_ns = watermarks->a.urgent_ns; + if (safe_to_lower || watermarks->a.urgent_ns > hubbub1->watermarks.a.urgent_ns) { + hubbub1->watermarks.a.urgent_ns = watermarks->a.urgent_ns; prog_wm_value = convert_and_clamp(watermarks->a.urgent_ns, refclk_mhz, 0x1fffff); REG_WRITE(DCHUBBUB_ARB_DATA_URGENCY_WATERMARK_A, prog_wm_value); @@ -275,20 +289,22 @@ void hubbub1_program_watermarks( watermarks->a.urgent_ns, prog_wm_value); } - if (safe_to_lower || watermarks->a.pte_meta_urgent_ns > hubbub->watermarks.a.pte_meta_urgent_ns) { - hubbub->watermarks.a.pte_meta_urgent_ns = watermarks->a.pte_meta_urgent_ns; - prog_wm_value = convert_and_clamp(watermarks->a.pte_meta_urgent_ns, - refclk_mhz, 0x1fffff); - REG_WRITE(DCHUBBUB_ARB_PTE_META_URGENCY_WATERMARK_A, prog_wm_value); - DC_LOG_BANDWIDTH_CALCS("PTE_META_URGENCY_WATERMARK_A calculated =%d\n" - "HW register value = 0x%x\n", - watermarks->a.pte_meta_urgent_ns, prog_wm_value); + if (REG(DCHUBBUB_ARB_PTE_META_URGENCY_WATERMARK_A)) { + if (safe_to_lower || watermarks->a.pte_meta_urgent_ns > hubbub1->watermarks.a.pte_meta_urgent_ns) { + hubbub1->watermarks.a.pte_meta_urgent_ns = watermarks->a.pte_meta_urgent_ns; + prog_wm_value = convert_and_clamp(watermarks->a.pte_meta_urgent_ns, + refclk_mhz, 0x1fffff); + REG_WRITE(DCHUBBUB_ARB_PTE_META_URGENCY_WATERMARK_A, prog_wm_value); + DC_LOG_BANDWIDTH_CALCS("PTE_META_URGENCY_WATERMARK_A calculated =%d\n" + "HW register value = 0x%x\n", + watermarks->a.pte_meta_urgent_ns, prog_wm_value); + } } if (REG(DCHUBBUB_ARB_ALLOW_SR_ENTER_WATERMARK_A)) { if (safe_to_lower || watermarks->a.cstate_pstate.cstate_enter_plus_exit_ns - > hubbub->watermarks.a.cstate_pstate.cstate_enter_plus_exit_ns) { - hubbub->watermarks.a.cstate_pstate.cstate_enter_plus_exit_ns = + > hubbub1->watermarks.a.cstate_pstate.cstate_enter_plus_exit_ns) { + hubbub1->watermarks.a.cstate_pstate.cstate_enter_plus_exit_ns = watermarks->a.cstate_pstate.cstate_enter_plus_exit_ns; prog_wm_value = convert_and_clamp( watermarks->a.cstate_pstate.cstate_enter_plus_exit_ns, @@ -300,8 +316,8 @@ void hubbub1_program_watermarks( } if (safe_to_lower || watermarks->a.cstate_pstate.cstate_exit_ns - > hubbub->watermarks.a.cstate_pstate.cstate_exit_ns) { - hubbub->watermarks.a.cstate_pstate.cstate_exit_ns = + > hubbub1->watermarks.a.cstate_pstate.cstate_exit_ns) { + hubbub1->watermarks.a.cstate_pstate.cstate_exit_ns = watermarks->a.cstate_pstate.cstate_exit_ns; prog_wm_value = convert_and_clamp( watermarks->a.cstate_pstate.cstate_exit_ns, @@ -314,8 +330,8 @@ void hubbub1_program_watermarks( } if (safe_to_lower || watermarks->a.cstate_pstate.pstate_change_ns - > hubbub->watermarks.a.cstate_pstate.pstate_change_ns) { - hubbub->watermarks.a.cstate_pstate.pstate_change_ns = + > hubbub1->watermarks.a.cstate_pstate.pstate_change_ns) { + hubbub1->watermarks.a.cstate_pstate.pstate_change_ns = watermarks->a.cstate_pstate.pstate_change_ns; prog_wm_value = convert_and_clamp( watermarks->a.cstate_pstate.pstate_change_ns, @@ -327,8 +343,8 @@ void hubbub1_program_watermarks( } /* clock state B */ - if (safe_to_lower || watermarks->b.urgent_ns > hubbub->watermarks.b.urgent_ns) { - hubbub->watermarks.b.urgent_ns = watermarks->b.urgent_ns; + if (safe_to_lower || watermarks->b.urgent_ns > hubbub1->watermarks.b.urgent_ns) { + hubbub1->watermarks.b.urgent_ns = watermarks->b.urgent_ns; prog_wm_value = convert_and_clamp(watermarks->b.urgent_ns, refclk_mhz, 0x1fffff); REG_WRITE(DCHUBBUB_ARB_DATA_URGENCY_WATERMARK_B, prog_wm_value); @@ -338,20 +354,22 @@ void hubbub1_program_watermarks( watermarks->b.urgent_ns, prog_wm_value); } - if (safe_to_lower || watermarks->b.pte_meta_urgent_ns > hubbub->watermarks.b.pte_meta_urgent_ns) { - hubbub->watermarks.b.pte_meta_urgent_ns = watermarks->b.pte_meta_urgent_ns; - prog_wm_value = convert_and_clamp(watermarks->b.pte_meta_urgent_ns, - refclk_mhz, 0x1fffff); - REG_WRITE(DCHUBBUB_ARB_PTE_META_URGENCY_WATERMARK_B, prog_wm_value); - DC_LOG_BANDWIDTH_CALCS("PTE_META_URGENCY_WATERMARK_B calculated =%d\n" - "HW register value = 0x%x\n", - watermarks->b.pte_meta_urgent_ns, prog_wm_value); + if (REG(DCHUBBUB_ARB_PTE_META_URGENCY_WATERMARK_B)) { + if (safe_to_lower || watermarks->b.pte_meta_urgent_ns > hubbub1->watermarks.b.pte_meta_urgent_ns) { + hubbub1->watermarks.b.pte_meta_urgent_ns = watermarks->b.pte_meta_urgent_ns; + prog_wm_value = convert_and_clamp(watermarks->b.pte_meta_urgent_ns, + refclk_mhz, 0x1fffff); + REG_WRITE(DCHUBBUB_ARB_PTE_META_URGENCY_WATERMARK_B, prog_wm_value); + DC_LOG_BANDWIDTH_CALCS("PTE_META_URGENCY_WATERMARK_B calculated =%d\n" + "HW register value = 0x%x\n", + watermarks->b.pte_meta_urgent_ns, prog_wm_value); + } } if (REG(DCHUBBUB_ARB_ALLOW_SR_ENTER_WATERMARK_B)) { if (safe_to_lower || watermarks->b.cstate_pstate.cstate_enter_plus_exit_ns - > hubbub->watermarks.b.cstate_pstate.cstate_enter_plus_exit_ns) { - hubbub->watermarks.b.cstate_pstate.cstate_enter_plus_exit_ns = + > hubbub1->watermarks.b.cstate_pstate.cstate_enter_plus_exit_ns) { + hubbub1->watermarks.b.cstate_pstate.cstate_enter_plus_exit_ns = watermarks->b.cstate_pstate.cstate_enter_plus_exit_ns; prog_wm_value = convert_and_clamp( watermarks->b.cstate_pstate.cstate_enter_plus_exit_ns, @@ -363,8 +381,8 @@ void hubbub1_program_watermarks( } if (safe_to_lower || watermarks->b.cstate_pstate.cstate_exit_ns - > hubbub->watermarks.b.cstate_pstate.cstate_exit_ns) { - hubbub->watermarks.b.cstate_pstate.cstate_exit_ns = + > hubbub1->watermarks.b.cstate_pstate.cstate_exit_ns) { + hubbub1->watermarks.b.cstate_pstate.cstate_exit_ns = watermarks->b.cstate_pstate.cstate_exit_ns; prog_wm_value = convert_and_clamp( watermarks->b.cstate_pstate.cstate_exit_ns, @@ -377,8 +395,8 @@ void hubbub1_program_watermarks( } if (safe_to_lower || watermarks->b.cstate_pstate.pstate_change_ns - > hubbub->watermarks.b.cstate_pstate.pstate_change_ns) { - hubbub->watermarks.b.cstate_pstate.pstate_change_ns = + > hubbub1->watermarks.b.cstate_pstate.pstate_change_ns) { + hubbub1->watermarks.b.cstate_pstate.pstate_change_ns = watermarks->b.cstate_pstate.pstate_change_ns; prog_wm_value = convert_and_clamp( watermarks->b.cstate_pstate.pstate_change_ns, @@ -390,8 +408,8 @@ void hubbub1_program_watermarks( } /* clock state C */ - if (safe_to_lower || watermarks->c.urgent_ns > hubbub->watermarks.c.urgent_ns) { - hubbub->watermarks.c.urgent_ns = watermarks->c.urgent_ns; + if (safe_to_lower || watermarks->c.urgent_ns > hubbub1->watermarks.c.urgent_ns) { + hubbub1->watermarks.c.urgent_ns = watermarks->c.urgent_ns; prog_wm_value = convert_and_clamp(watermarks->c.urgent_ns, refclk_mhz, 0x1fffff); REG_WRITE(DCHUBBUB_ARB_DATA_URGENCY_WATERMARK_C, prog_wm_value); @@ -401,20 +419,22 @@ void hubbub1_program_watermarks( watermarks->c.urgent_ns, prog_wm_value); } - if (safe_to_lower || watermarks->c.pte_meta_urgent_ns > hubbub->watermarks.c.pte_meta_urgent_ns) { - hubbub->watermarks.c.pte_meta_urgent_ns = watermarks->c.pte_meta_urgent_ns; - prog_wm_value = convert_and_clamp(watermarks->c.pte_meta_urgent_ns, - refclk_mhz, 0x1fffff); - REG_WRITE(DCHUBBUB_ARB_PTE_META_URGENCY_WATERMARK_C, prog_wm_value); - DC_LOG_BANDWIDTH_CALCS("PTE_META_URGENCY_WATERMARK_C calculated =%d\n" - "HW register value = 0x%x\n", - watermarks->c.pte_meta_urgent_ns, prog_wm_value); + if (REG(DCHUBBUB_ARB_PTE_META_URGENCY_WATERMARK_C)) { + if (safe_to_lower || watermarks->c.pte_meta_urgent_ns > hubbub1->watermarks.c.pte_meta_urgent_ns) { + hubbub1->watermarks.c.pte_meta_urgent_ns = watermarks->c.pte_meta_urgent_ns; + prog_wm_value = convert_and_clamp(watermarks->c.pte_meta_urgent_ns, + refclk_mhz, 0x1fffff); + REG_WRITE(DCHUBBUB_ARB_PTE_META_URGENCY_WATERMARK_C, prog_wm_value); + DC_LOG_BANDWIDTH_CALCS("PTE_META_URGENCY_WATERMARK_C calculated =%d\n" + "HW register value = 0x%x\n", + watermarks->c.pte_meta_urgent_ns, prog_wm_value); + } } if (REG(DCHUBBUB_ARB_ALLOW_SR_ENTER_WATERMARK_C)) { if (safe_to_lower || watermarks->c.cstate_pstate.cstate_enter_plus_exit_ns - > hubbub->watermarks.c.cstate_pstate.cstate_enter_plus_exit_ns) { - hubbub->watermarks.c.cstate_pstate.cstate_enter_plus_exit_ns = + > hubbub1->watermarks.c.cstate_pstate.cstate_enter_plus_exit_ns) { + hubbub1->watermarks.c.cstate_pstate.cstate_enter_plus_exit_ns = watermarks->c.cstate_pstate.cstate_enter_plus_exit_ns; prog_wm_value = convert_and_clamp( watermarks->c.cstate_pstate.cstate_enter_plus_exit_ns, @@ -426,8 +446,8 @@ void hubbub1_program_watermarks( } if (safe_to_lower || watermarks->c.cstate_pstate.cstate_exit_ns - > hubbub->watermarks.c.cstate_pstate.cstate_exit_ns) { - hubbub->watermarks.c.cstate_pstate.cstate_exit_ns = + > hubbub1->watermarks.c.cstate_pstate.cstate_exit_ns) { + hubbub1->watermarks.c.cstate_pstate.cstate_exit_ns = watermarks->c.cstate_pstate.cstate_exit_ns; prog_wm_value = convert_and_clamp( watermarks->c.cstate_pstate.cstate_exit_ns, @@ -440,8 +460,8 @@ void hubbub1_program_watermarks( } if (safe_to_lower || watermarks->c.cstate_pstate.pstate_change_ns - > hubbub->watermarks.c.cstate_pstate.pstate_change_ns) { - hubbub->watermarks.c.cstate_pstate.pstate_change_ns = + > hubbub1->watermarks.c.cstate_pstate.pstate_change_ns) { + hubbub1->watermarks.c.cstate_pstate.pstate_change_ns = watermarks->c.cstate_pstate.pstate_change_ns; prog_wm_value = convert_and_clamp( watermarks->c.cstate_pstate.pstate_change_ns, @@ -453,8 +473,8 @@ void hubbub1_program_watermarks( } /* clock state D */ - if (safe_to_lower || watermarks->d.urgent_ns > hubbub->watermarks.d.urgent_ns) { - hubbub->watermarks.d.urgent_ns = watermarks->d.urgent_ns; + if (safe_to_lower || watermarks->d.urgent_ns > hubbub1->watermarks.d.urgent_ns) { + hubbub1->watermarks.d.urgent_ns = watermarks->d.urgent_ns; prog_wm_value = convert_and_clamp(watermarks->d.urgent_ns, refclk_mhz, 0x1fffff); REG_WRITE(DCHUBBUB_ARB_DATA_URGENCY_WATERMARK_D, prog_wm_value); @@ -464,20 +484,22 @@ void hubbub1_program_watermarks( watermarks->d.urgent_ns, prog_wm_value); } - if (safe_to_lower || watermarks->d.pte_meta_urgent_ns > hubbub->watermarks.d.pte_meta_urgent_ns) { - hubbub->watermarks.d.pte_meta_urgent_ns = watermarks->d.pte_meta_urgent_ns; - prog_wm_value = convert_and_clamp(watermarks->d.pte_meta_urgent_ns, - refclk_mhz, 0x1fffff); - REG_WRITE(DCHUBBUB_ARB_PTE_META_URGENCY_WATERMARK_D, prog_wm_value); - DC_LOG_BANDWIDTH_CALCS("PTE_META_URGENCY_WATERMARK_D calculated =%d\n" - "HW register value = 0x%x\n", - watermarks->d.pte_meta_urgent_ns, prog_wm_value); + if (REG(DCHUBBUB_ARB_PTE_META_URGENCY_WATERMARK_D)) { + if (safe_to_lower || watermarks->d.pte_meta_urgent_ns > hubbub1->watermarks.d.pte_meta_urgent_ns) { + hubbub1->watermarks.d.pte_meta_urgent_ns = watermarks->d.pte_meta_urgent_ns; + prog_wm_value = convert_and_clamp(watermarks->d.pte_meta_urgent_ns, + refclk_mhz, 0x1fffff); + REG_WRITE(DCHUBBUB_ARB_PTE_META_URGENCY_WATERMARK_D, prog_wm_value); + DC_LOG_BANDWIDTH_CALCS("PTE_META_URGENCY_WATERMARK_D calculated =%d\n" + "HW register value = 0x%x\n", + watermarks->d.pte_meta_urgent_ns, prog_wm_value); + } } if (REG(DCHUBBUB_ARB_ALLOW_SR_ENTER_WATERMARK_D)) { if (safe_to_lower || watermarks->d.cstate_pstate.cstate_enter_plus_exit_ns - > hubbub->watermarks.d.cstate_pstate.cstate_enter_plus_exit_ns) { - hubbub->watermarks.d.cstate_pstate.cstate_enter_plus_exit_ns = + > hubbub1->watermarks.d.cstate_pstate.cstate_enter_plus_exit_ns) { + hubbub1->watermarks.d.cstate_pstate.cstate_enter_plus_exit_ns = watermarks->d.cstate_pstate.cstate_enter_plus_exit_ns; prog_wm_value = convert_and_clamp( watermarks->d.cstate_pstate.cstate_enter_plus_exit_ns, @@ -489,8 +511,8 @@ void hubbub1_program_watermarks( } if (safe_to_lower || watermarks->d.cstate_pstate.cstate_exit_ns - > hubbub->watermarks.d.cstate_pstate.cstate_exit_ns) { - hubbub->watermarks.d.cstate_pstate.cstate_exit_ns = + > hubbub1->watermarks.d.cstate_pstate.cstate_exit_ns) { + hubbub1->watermarks.d.cstate_pstate.cstate_exit_ns = watermarks->d.cstate_pstate.cstate_exit_ns; prog_wm_value = convert_and_clamp( watermarks->d.cstate_pstate.cstate_exit_ns, @@ -503,8 +525,8 @@ void hubbub1_program_watermarks( } if (safe_to_lower || watermarks->d.cstate_pstate.pstate_change_ns - > hubbub->watermarks.d.cstate_pstate.pstate_change_ns) { - hubbub->watermarks.d.cstate_pstate.pstate_change_ns = + > hubbub1->watermarks.d.cstate_pstate.pstate_change_ns) { + hubbub1->watermarks.d.cstate_pstate.pstate_change_ns = watermarks->d.cstate_pstate.pstate_change_ns; prog_wm_value = convert_and_clamp( watermarks->d.cstate_pstate.pstate_change_ns, @@ -520,9 +542,7 @@ void hubbub1_program_watermarks( REG_UPDATE(DCHUBBUB_ARB_DF_REQ_OUTSTAND, DCHUBBUB_ARB_MIN_REQ_OUTSTAND, 68); - REG_UPDATE_2(DCHUBBUB_ARB_DRAM_STATE_CNTL, - DCHUBBUB_ARB_ALLOW_SELF_REFRESH_FORCE_VALUE, 0, - DCHUBBUB_ARB_ALLOW_SELF_REFRESH_FORCE_ENABLE, force_en); + hubbub1_allow_self_refresh_control(hubbub, !hubbub->ctx->dc->debug.disable_stutter); #if 0 REG_UPDATE_2(DCHUBBUB_ARB_WATERMARK_CHANGE_CNTL, @@ -535,6 +555,8 @@ void hubbub1_update_dchub( struct hubbub *hubbub, struct dchub_init_data *dh_data) { + struct dcn10_hubbub *hubbub1 = TO_DCN10_HUBBUB(hubbub); + if (REG(DCHUBBUB_SDPIF_FB_TOP) == 0) { ASSERT(false); /*should not come here*/ @@ -594,6 +616,8 @@ void hubbub1_update_dchub( void hubbub1_toggle_watermark_change_req(struct hubbub *hubbub) { + struct dcn10_hubbub *hubbub1 = TO_DCN10_HUBBUB(hubbub); + uint32_t watermark_change_req; REG_GET(DCHUBBUB_ARB_WATERMARK_CHANGE_CNTL, @@ -610,6 +634,8 @@ void hubbub1_toggle_watermark_change_req(struct hubbub *hubbub) void hubbub1_soft_reset(struct hubbub *hubbub, bool reset) { + struct dcn10_hubbub *hubbub1 = TO_DCN10_HUBBUB(hubbub); + uint32_t reset_en = reset ? 1 : 0; REG_UPDATE(DCHUBBUB_SOFT_RESET, @@ -752,7 +778,9 @@ static bool hubbub1_get_dcc_compression_cap(struct hubbub *hubbub, const struct dc_dcc_surface_param *input, struct dc_surface_dcc_cap *output) { - struct dc *dc = hubbub->ctx->dc; + struct dcn10_hubbub *hubbub1 = TO_DCN10_HUBBUB(hubbub); + struct dc *dc = hubbub1->base.ctx->dc; + /* implement section 1.6.2.1 of DCN1_Programming_Guide.docx */ enum dcc_control dcc_control; unsigned int bpe; @@ -764,10 +792,10 @@ static bool hubbub1_get_dcc_compression_cap(struct hubbub *hubbub, if (dc->debug.disable_dcc == DCC_DISABLE) return false; - if (!hubbub->funcs->dcc_support_pixel_format(input->format, &bpe)) + if (!hubbub1->base.funcs->dcc_support_pixel_format(input->format, &bpe)) return false; - if (!hubbub->funcs->dcc_support_swizzle(input->swizzle_mode, bpe, + if (!hubbub1->base.funcs->dcc_support_swizzle(input->swizzle_mode, bpe, &segment_order_horz, &segment_order_vert)) return false; @@ -837,6 +865,7 @@ static const struct hubbub_funcs hubbub1_funcs = { .dcc_support_swizzle = hubbub1_dcc_support_swizzle, .dcc_support_pixel_format = hubbub1_dcc_support_pixel_format, .get_dcc_compression_cap = hubbub1_get_dcc_compression_cap, + .wm_read_state = hubbub1_wm_read_state, }; void hubbub1_construct(struct hubbub *hubbub, @@ -845,18 +874,20 @@ void hubbub1_construct(struct hubbub *hubbub, const struct dcn_hubbub_shift *hubbub_shift, const struct dcn_hubbub_mask *hubbub_mask) { - hubbub->ctx = ctx; + struct dcn10_hubbub *hubbub1 = TO_DCN10_HUBBUB(hubbub); + + hubbub1->base.ctx = ctx; - hubbub->funcs = &hubbub1_funcs; + hubbub1->base.funcs = &hubbub1_funcs; - hubbub->regs = hubbub_regs; - hubbub->shifts = hubbub_shift; - hubbub->masks = hubbub_mask; + hubbub1->regs = hubbub_regs; + hubbub1->shifts = hubbub_shift; + hubbub1->masks = hubbub_mask; - hubbub->debug_test_index_pstate = 0x7; + hubbub1->debug_test_index_pstate = 0x7; #if defined(CONFIG_DRM_AMD_DC_DCN1_01) if (ctx->dce_version == DCN_VERSION_1_01) - hubbub->debug_test_index_pstate = 0xB; + hubbub1->debug_test_index_pstate = 0xB; #endif } diff --git a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_hubbub.h b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_hubbub.h index d0f03d152913..9cd4a5194154 100644 --- a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_hubbub.h +++ b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_hubbub.h @@ -29,6 +29,9 @@ #include "core_types.h" #include "dchubbub.h" +#define TO_DCN10_HUBBUB(hubbub)\ + container_of(hubbub, struct dcn10_hubbub, base) + #define HUBHUB_REG_LIST_DCN()\ SR(DCHUBBUB_ARB_DATA_URGENCY_WATERMARK_A),\ SR(DCHUBBUB_ARB_PTE_META_URGENCY_WATERMARK_A),\ @@ -107,6 +110,12 @@ struct dcn_hubbub_registers { uint32_t DCHUBBUB_SDPIF_AGP_TOP; uint32_t DCHUBBUB_CRC_CTRL; uint32_t DCHUBBUB_SOFT_RESET; + uint32_t DCN_VM_FB_LOCATION_BASE; + uint32_t DCN_VM_FB_LOCATION_TOP; + uint32_t DCN_VM_FB_OFFSET; + uint32_t DCN_VM_AGP_BOT; + uint32_t DCN_VM_AGP_TOP; + uint32_t DCN_VM_AGP_BASE; }; /* set field name */ @@ -152,7 +161,13 @@ struct dcn_hubbub_registers { type SDPIF_FB_OFFSET;\ type SDPIF_AGP_BASE;\ type SDPIF_AGP_BOT;\ - type SDPIF_AGP_TOP + type SDPIF_AGP_TOP;\ + type FB_BASE;\ + type FB_TOP;\ + type FB_OFFSET;\ + type AGP_BOT;\ + type AGP_TOP;\ + type AGP_BASE struct dcn_hubbub_shift { @@ -165,22 +180,8 @@ struct dcn_hubbub_mask { struct dc; -struct dcn_hubbub_wm_set { - uint32_t wm_set; - uint32_t data_urgent; - uint32_t pte_meta_urgent; - uint32_t sr_enter; - uint32_t sr_exit; - uint32_t dram_clk_chanage; -}; - -struct dcn_hubbub_wm { - struct dcn_hubbub_wm_set sets[4]; -}; - -struct hubbub { - const struct hubbub_funcs *funcs; - struct dc_context *ctx; +struct dcn10_hubbub { + struct hubbub base; const struct dcn_hubbub_registers *regs; const struct dcn_hubbub_shift *shifts; const struct dcn_hubbub_mask *masks; @@ -203,7 +204,7 @@ void hubbub1_program_watermarks( unsigned int refclk_mhz, bool safe_to_lower); -void hubbub1_disable_allow_self_refresh(struct hubbub *hubbub); +void hubbub1_allow_self_refresh_control(struct hubbub *hubbub, bool allow); bool hububu1_is_allow_self_refresh_enabled(struct hubbub *hubub); diff --git a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_hubp.c b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_hubp.c index 74132a1f3046..0ba68d41b9c3 100644 --- a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_hubp.c +++ b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_hubp.c @@ -99,6 +99,14 @@ static unsigned int hubp1_get_underflow_status(struct hubp *hubp) return hubp_underflow; } + +void hubp1_clear_underflow(struct hubp *hubp) +{ + struct dcn10_hubp *hubp1 = TO_DCN10_HUBP(hubp); + + REG_UPDATE(DCHUBP_CNTL, HUBP_UNDERFLOW_CLEAR, 1); +} + static void hubp1_set_hubp_blank_en(struct hubp *hubp, bool blank) { struct dcn10_hubp *hubp1 = TO_DCN10_HUBP(hubp); @@ -107,7 +115,7 @@ static void hubp1_set_hubp_blank_en(struct hubp *hubp, bool blank) REG_UPDATE(DCHUBP_CNTL, HUBP_BLANK_EN, blank_en); } -static void hubp1_vready_workaround(struct hubp *hubp, +void hubp1_vready_workaround(struct hubp *hubp, struct _vcs_dpi_display_pipe_dest_params_st *pipe_dest) { uint32_t value = 0; @@ -309,7 +317,8 @@ void hubp1_program_pixel_format( bool hubp1_program_surface_flip_and_addr( struct hubp *hubp, const struct dc_plane_address *address, - bool flip_immediate) + bool flip_immediate, + uint8_t vmid) { struct dcn10_hubp *hubp1 = TO_DCN10_HUBP(hubp); @@ -565,19 +574,6 @@ void hubp1_program_deadline( REFCYC_X_AFTER_SCALER, dlg_attr->refcyc_x_after_scaler, DST_Y_AFTER_SCALER, dlg_attr->dst_y_after_scaler); - if (REG(PREFETCH_SETTINS)) - REG_SET_2(PREFETCH_SETTINS, 0, - DST_Y_PREFETCH, dlg_attr->dst_y_prefetch, - VRATIO_PREFETCH, dlg_attr->vratio_prefetch); - else - REG_SET_2(PREFETCH_SETTINGS, 0, - DST_Y_PREFETCH, dlg_attr->dst_y_prefetch, - VRATIO_PREFETCH, dlg_attr->vratio_prefetch); - - REG_SET_2(VBLANK_PARAMETERS_0, 0, - DST_Y_PER_VM_VBLANK, dlg_attr->dst_y_per_vm_vblank, - DST_Y_PER_ROW_VBLANK, dlg_attr->dst_y_per_row_vblank); - REG_SET(REF_FREQ_TO_PIX_FREQ, 0, REF_FREQ_TO_PIX_FREQ, dlg_attr->ref_freq_to_pix_freq); @@ -585,9 +581,6 @@ void hubp1_program_deadline( REG_SET(VBLANK_PARAMETERS_1, 0, REFCYC_PER_PTE_GROUP_VBLANK_L, dlg_attr->refcyc_per_pte_group_vblank_l); - REG_SET(VBLANK_PARAMETERS_3, 0, - REFCYC_PER_META_CHUNK_VBLANK_L, dlg_attr->refcyc_per_meta_chunk_vblank_l); - if (REG(NOM_PARAMETERS_0)) REG_SET(NOM_PARAMETERS_0, 0, DST_Y_PER_PTE_ROW_NOM_L, dlg_attr->dst_y_per_pte_row_nom_l); @@ -602,27 +595,13 @@ void hubp1_program_deadline( REG_SET(NOM_PARAMETERS_5, 0, REFCYC_PER_META_CHUNK_NOM_L, dlg_attr->refcyc_per_meta_chunk_nom_l); - REG_SET_2(PER_LINE_DELIVERY_PRE, 0, - REFCYC_PER_LINE_DELIVERY_PRE_L, dlg_attr->refcyc_per_line_delivery_pre_l, - REFCYC_PER_LINE_DELIVERY_PRE_C, dlg_attr->refcyc_per_line_delivery_pre_c); - REG_SET_2(PER_LINE_DELIVERY, 0, REFCYC_PER_LINE_DELIVERY_L, dlg_attr->refcyc_per_line_delivery_l, REFCYC_PER_LINE_DELIVERY_C, dlg_attr->refcyc_per_line_delivery_c); - if (REG(PREFETCH_SETTINS_C)) - REG_SET(PREFETCH_SETTINS_C, 0, - VRATIO_PREFETCH_C, dlg_attr->vratio_prefetch_c); - else - REG_SET(PREFETCH_SETTINGS_C, 0, - VRATIO_PREFETCH_C, dlg_attr->vratio_prefetch_c); - REG_SET(VBLANK_PARAMETERS_2, 0, REFCYC_PER_PTE_GROUP_VBLANK_C, dlg_attr->refcyc_per_pte_group_vblank_c); - REG_SET(VBLANK_PARAMETERS_4, 0, - REFCYC_PER_META_CHUNK_VBLANK_C, dlg_attr->refcyc_per_meta_chunk_vblank_c); - if (REG(NOM_PARAMETERS_2)) REG_SET(NOM_PARAMETERS_2, 0, DST_Y_PER_PTE_ROW_NOM_C, dlg_attr->dst_y_per_pte_row_nom_c); @@ -642,10 +621,6 @@ void hubp1_program_deadline( QoS_LEVEL_LOW_WM, ttu_attr->qos_level_low_wm, QoS_LEVEL_HIGH_WM, ttu_attr->qos_level_high_wm); - REG_SET_2(DCN_GLOBAL_TTU_CNTL, 0, - MIN_TTU_VBLANK, ttu_attr->min_ttu_vblank, - QoS_LEVEL_FLIP, ttu_attr->qos_level_flip); - /* TTU - per luma/chroma */ /* Assumed surf0 is luma and 1 is chroma */ @@ -654,25 +629,15 @@ void hubp1_program_deadline( QoS_LEVEL_FIXED, ttu_attr->qos_level_fixed_l, QoS_RAMP_DISABLE, ttu_attr->qos_ramp_disable_l); - REG_SET(DCN_SURF0_TTU_CNTL1, 0, - REFCYC_PER_REQ_DELIVERY_PRE, - ttu_attr->refcyc_per_req_delivery_pre_l); - REG_SET_3(DCN_SURF1_TTU_CNTL0, 0, REFCYC_PER_REQ_DELIVERY, ttu_attr->refcyc_per_req_delivery_c, QoS_LEVEL_FIXED, ttu_attr->qos_level_fixed_c, QoS_RAMP_DISABLE, ttu_attr->qos_ramp_disable_c); - REG_SET(DCN_SURF1_TTU_CNTL1, 0, - REFCYC_PER_REQ_DELIVERY_PRE, - ttu_attr->refcyc_per_req_delivery_pre_c); - REG_SET_3(DCN_CUR0_TTU_CNTL0, 0, REFCYC_PER_REQ_DELIVERY, ttu_attr->refcyc_per_req_delivery_cur0, QoS_LEVEL_FIXED, ttu_attr->qos_level_fixed_cur0, QoS_RAMP_DISABLE, ttu_attr->qos_ramp_disable_cur0); - REG_SET(DCN_CUR0_TTU_CNTL1, 0, - REFCYC_PER_REQ_DELIVERY_PRE, ttu_attr->refcyc_per_req_delivery_pre_cur0); } static void hubp1_setup( @@ -690,6 +655,48 @@ static void hubp1_setup( hubp1_vready_workaround(hubp, pipe_dest); } +static void hubp1_setup_interdependent( + struct hubp *hubp, + struct _vcs_dpi_display_dlg_regs_st *dlg_attr, + struct _vcs_dpi_display_ttu_regs_st *ttu_attr) +{ + struct dcn10_hubp *hubp1 = TO_DCN10_HUBP(hubp); + + REG_SET_2(PREFETCH_SETTINS, 0, + DST_Y_PREFETCH, dlg_attr->dst_y_prefetch, + VRATIO_PREFETCH, dlg_attr->vratio_prefetch); + + REG_SET(PREFETCH_SETTINS_C, 0, + VRATIO_PREFETCH_C, dlg_attr->vratio_prefetch_c); + + REG_SET_2(VBLANK_PARAMETERS_0, 0, + DST_Y_PER_VM_VBLANK, dlg_attr->dst_y_per_vm_vblank, + DST_Y_PER_ROW_VBLANK, dlg_attr->dst_y_per_row_vblank); + + REG_SET(VBLANK_PARAMETERS_3, 0, + REFCYC_PER_META_CHUNK_VBLANK_L, dlg_attr->refcyc_per_meta_chunk_vblank_l); + + REG_SET(VBLANK_PARAMETERS_4, 0, + REFCYC_PER_META_CHUNK_VBLANK_C, dlg_attr->refcyc_per_meta_chunk_vblank_c); + + REG_SET_2(PER_LINE_DELIVERY_PRE, 0, + REFCYC_PER_LINE_DELIVERY_PRE_L, dlg_attr->refcyc_per_line_delivery_pre_l, + REFCYC_PER_LINE_DELIVERY_PRE_C, dlg_attr->refcyc_per_line_delivery_pre_c); + + REG_SET(DCN_SURF0_TTU_CNTL1, 0, + REFCYC_PER_REQ_DELIVERY_PRE, + ttu_attr->refcyc_per_req_delivery_pre_l); + REG_SET(DCN_SURF1_TTU_CNTL1, 0, + REFCYC_PER_REQ_DELIVERY_PRE, + ttu_attr->refcyc_per_req_delivery_pre_c); + REG_SET(DCN_CUR0_TTU_CNTL1, 0, + REFCYC_PER_REQ_DELIVERY_PRE, ttu_attr->refcyc_per_req_delivery_pre_cur0); + + REG_SET_2(DCN_GLOBAL_TTU_CNTL, 0, + MIN_TTU_VBLANK, ttu_attr->min_ttu_vblank, + QoS_LEVEL_FLIP, ttu_attr->qos_level_flip); +} + bool hubp1_is_flip_pending(struct hubp *hubp) { uint32_t flip_pending = 0; @@ -1134,7 +1141,7 @@ void hubp1_cursor_set_position( if (src_y_offset >= (int)param->viewport.height) cur_en = 0; /* not visible beyond bottom edge*/ - if (src_y_offset < 0) //+ (int)hubp->curs_attr.height + if (src_y_offset + (int)hubp->curs_attr.height <= 0) cur_en = 0; /* not visible beyond top edge*/ if (cur_en && REG_READ(CURSOR_SURFACE_ADDRESS) == 0) @@ -1178,6 +1185,7 @@ static const struct hubp_funcs dcn10_hubp_funcs = { hubp1_program_surface_config, .hubp_is_flip_pending = hubp1_is_flip_pending, .hubp_setup = hubp1_setup, + .hubp_setup_interdependent = hubp1_setup_interdependent, .hubp_set_vm_system_aperture_settings = hubp1_set_vm_system_aperture_settings, .hubp_set_vm_context0_settings = hubp1_set_vm_context0_settings, .set_blank = hubp1_set_blank, @@ -1190,6 +1198,7 @@ static const struct hubp_funcs dcn10_hubp_funcs = { .hubp_clk_cntl = hubp1_clk_cntl, .hubp_vtg_sel = hubp1_vtg_sel, .hubp_read_state = hubp1_read_state, + .hubp_clear_underflow = hubp1_clear_underflow, .hubp_disable_control = hubp1_disable_control, .hubp_get_underflow_status = hubp1_get_underflow_status, diff --git a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_hubp.h b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_hubp.h index 4890273b632b..a6d6dfe00617 100644 --- a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_hubp.h +++ b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_hubp.h @@ -251,6 +251,7 @@ HUBP_SF(HUBP0_DCHUBP_CNTL, HUBP_BLANK_EN, mask_sh),\ HUBP_SF(HUBP0_DCHUBP_CNTL, HUBP_TTU_DISABLE, mask_sh),\ HUBP_SF(HUBP0_DCHUBP_CNTL, HUBP_UNDERFLOW_STATUS, mask_sh),\ + HUBP_SF(HUBP0_DCHUBP_CNTL, HUBP_UNDERFLOW_CLEAR, mask_sh),\ HUBP_SF(HUBP0_DCHUBP_CNTL, HUBP_NO_OUTSTANDING_REQ, mask_sh),\ HUBP_SF(HUBP0_DCHUBP_CNTL, HUBP_VTG_SEL, mask_sh),\ HUBP_SF(HUBP0_DCHUBP_CNTL, HUBP_DISABLE, mask_sh),\ @@ -435,6 +436,7 @@ type HUBP_NO_OUTSTANDING_REQ;\ type HUBP_VTG_SEL;\ type HUBP_UNDERFLOW_STATUS;\ + type HUBP_UNDERFLOW_CLEAR;\ type NUM_PIPES;\ type NUM_BANKS;\ type PIPE_INTERLEAVE;\ @@ -705,11 +707,6 @@ void hubp1_dcc_control(struct hubp *hubp, bool enable, bool independent_64b_blks); -bool hubp1_program_surface_flip_and_addr( - struct hubp *hubp, - const struct dc_plane_address *address, - bool flip_immediate); - bool hubp1_is_flip_pending(struct hubp *hubp); void hubp1_cursor_set_attributes( @@ -739,8 +736,11 @@ void dcn10_hubp_construct( const struct dcn_mi_mask *hubp_mask); void hubp1_read_state(struct hubp *hubp); +void hubp1_clear_underflow(struct hubp *hubp); enum cursor_pitch hubp1_get_cursor_pitch(unsigned int pitch); +void hubp1_vready_workaround(struct hubp *hubp, + struct _vcs_dpi_display_pipe_dest_params_st *pipe_dest); #endif diff --git a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_hw_sequencer.c b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_hw_sequencer.c index 193184affefb..d1a8f1c302a9 100644 --- a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_hw_sequencer.c +++ b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_hw_sequencer.c @@ -40,11 +40,11 @@ #include "ipp.h" #include "mpc.h" #include "reg_helper.h" -#include "custom_float.h" #include "dcn10_hubp.h" #include "dcn10_hubbub.h" #include "dcn10_cm_common.h" #include "dc_link_dp.h" +#include "dccg.h" #define DC_LOGGER_INIT(logger) @@ -91,10 +91,11 @@ static void log_mpc_crc(struct dc *dc, void dcn10_log_hubbub_state(struct dc *dc, struct dc_log_buffer_ctx *log_ctx) { struct dc_context *dc_ctx = dc->ctx; - struct dcn_hubbub_wm wm = {0}; + struct dcn_hubbub_wm wm; int i; - hubbub1_wm_read_state(dc->res_pool->hubbub, &wm); + memset(&wm, 0, sizeof(struct dcn_hubbub_wm)); + dc->res_pool->hubbub->funcs->wm_read_state(dc->res_pool->hubbub, &wm); DTN_INFO("HUBBUB WM: data_urgent pte_meta_urgent" " sr_enter sr_exit dram_clk_change\n"); @@ -635,8 +636,6 @@ static enum dc_status dcn10_enable_stream_timing( struct dc_stream_state *stream = pipe_ctx->stream; enum dc_color_space color_space; struct tg_color black_color = {0}; - struct drr_params params = {0}; - unsigned int event_triggers = 0; /* by upper caller loop, pipe0 is parent pipe and be called first. * back end is set up by for pipe0. Other children pipe share back end @@ -704,19 +703,6 @@ static enum dc_status dcn10_enable_stream_timing( return DC_ERROR_UNEXPECTED; } - params.vertical_total_min = stream->adjust.v_total_min; - params.vertical_total_max = stream->adjust.v_total_max; - if (pipe_ctx->stream_res.tg->funcs->set_drr) - pipe_ctx->stream_res.tg->funcs->set_drr( - pipe_ctx->stream_res.tg, ¶ms); - - // DRR should set trigger event to monitor surface update event - if (stream->adjust.v_total_min != 0 && stream->adjust.v_total_max != 0) - event_triggers = 0x80; - if (pipe_ctx->stream_res.tg->funcs->set_static_screen_control) - pipe_ctx->stream_res.tg->funcs->set_static_screen_control( - pipe_ctx->stream_res.tg, event_triggers); - /* TODO program crtc source select for non-virtual signal*/ /* TODO program FMT */ /* TODO setup link_enc */ @@ -786,7 +772,7 @@ static bool dcn10_hw_wa_force_recovery(struct dc *dc) &dc->current_state->res_ctx.pipe_ctx[i]; if (pipe_ctx != NULL) { hubp = pipe_ctx->plane_res.hubp; - if (hubp != NULL) { + if (hubp != NULL && hubp->funcs->hubp_get_underflow_status) { if (hubp->funcs->hubp_get_underflow_status(hubp) != 0) { /* one pipe underflow, we will reset all the pipes*/ need_recover = true; @@ -812,7 +798,7 @@ static bool dcn10_hw_wa_force_recovery(struct dc *dc) if (pipe_ctx != NULL) { hubp = pipe_ctx->plane_res.hubp; /*DCHUBP_CNTL:HUBP_BLANK_EN=1*/ - if (hubp != NULL) + if (hubp != NULL && hubp->funcs->set_hubp_blank_en) hubp->funcs->set_hubp_blank_en(hubp, true); } } @@ -825,7 +811,7 @@ static bool dcn10_hw_wa_force_recovery(struct dc *dc) if (pipe_ctx != NULL) { hubp = pipe_ctx->plane_res.hubp; /*DCHUBP_CNTL:HUBP_DISABLE=1*/ - if (hubp != NULL) + if (hubp != NULL && hubp->funcs->hubp_disable_control) hubp->funcs->hubp_disable_control(hubp, true); } } @@ -835,7 +821,7 @@ static bool dcn10_hw_wa_force_recovery(struct dc *dc) if (pipe_ctx != NULL) { hubp = pipe_ctx->plane_res.hubp; /*DCHUBP_CNTL:HUBP_DISABLE=0*/ - if (hubp != NULL) + if (hubp != NULL && hubp->funcs->hubp_disable_control) hubp->funcs->hubp_disable_control(hubp, true); } } @@ -847,7 +833,7 @@ static bool dcn10_hw_wa_force_recovery(struct dc *dc) if (pipe_ctx != NULL) { hubp = pipe_ctx->plane_res.hubp; /*DCHUBP_CNTL:HUBP_BLANK_EN=0*/ - if (hubp != NULL) + if (hubp != NULL && hubp->funcs->set_hubp_blank_en) hubp->funcs->set_hubp_blank_en(hubp, true); } } @@ -970,92 +956,62 @@ static void dcn10_disable_plane(struct dc *dc, struct pipe_ctx *pipe_ctx) pipe_ctx->pipe_idx); } -static void dcn10_init_hw(struct dc *dc) +static void dcn10_init_pipes(struct dc *dc, struct dc_state *context) { int i; - struct abm *abm = dc->res_pool->abm; - struct dmcu *dmcu = dc->res_pool->dmcu; - struct dce_hwseq *hws = dc->hwseq; - struct dc_bios *dcb = dc->ctx->dc_bios; - struct dc_state *context = dc->current_state; - - if (IS_FPGA_MAXIMUS_DC(dc->ctx->dce_environment)) { - REG_WRITE(REFCLK_CNTL, 0); - REG_UPDATE(DCHUBBUB_GLOBAL_TIMER_CNTL, DCHUBBUB_GLOBAL_TIMER_ENABLE, 1); - REG_WRITE(DIO_MEM_PWR_CTRL, 0); - - if (!dc->debug.disable_clock_gate) { - /* enable all DCN clock gating */ - REG_WRITE(DCCG_GATE_DISABLE_CNTL, 0); - - REG_WRITE(DCCG_GATE_DISABLE_CNTL2, 0); - - REG_UPDATE(DCFCLK_CNTL, DCFCLK_GATE_DIS, 0); - } + bool can_apply_seamless_boot = false; - enable_power_gating_plane(dc->hwseq, true); - } else { - - if (!dcb->funcs->is_accelerated_mode(dcb)) { - bool allow_self_fresh_force_enable = - hububu1_is_allow_self_refresh_enabled(dc->res_pool->hubbub); - - bios_golden_init(dc); - - /* WA for making DF sleep when idle after resume from S0i3. - * DCHUBBUB_ARB_ALLOW_SELF_REFRESH_FORCE_ENABLE is set to 1 by - * command table, if DCHUBBUB_ARB_ALLOW_SELF_REFRESH_FORCE_ENABLE = 0 - * before calling command table and it changed to 1 after, - * it should be set back to 0. - */ - if (allow_self_fresh_force_enable == false && - hububu1_is_allow_self_refresh_enabled(dc->res_pool->hubbub)) - hubbub1_disable_allow_self_refresh(dc->res_pool->hubbub); - - disable_vga(dc->hwseq); - } - - for (i = 0; i < dc->link_count; i++) { - /* Power up AND update implementation according to the - * required signal (which may be different from the - * default signal on connector). - */ - struct dc_link *link = dc->links[i]; - - if (link->link_enc->connector.id == CONNECTOR_ID_EDP) - dc->hwss.edp_power_control(link, true); - - link->link_enc->funcs->hw_init(link->link_enc); + for (i = 0; i < context->stream_count; i++) { + if (context->streams[i]->apply_seamless_boot_optimization) { + can_apply_seamless_boot = true; + break; } } for (i = 0; i < dc->res_pool->pipe_count; i++) { struct timing_generator *tg = dc->res_pool->timing_generators[i]; + struct pipe_ctx *pipe_ctx = &context->res_ctx.pipe_ctx[i]; + + /* There is assumption that pipe_ctx is not mapping irregularly + * to non-preferred front end. If pipe_ctx->stream is not NULL, + * we will use the pipe, so don't disable + */ + if (pipe_ctx->stream != NULL) + continue; if (tg->funcs->is_tg_enabled(tg)) tg->funcs->lock(tg); - } - - /* Blank controller using driver code instead of - * command table. - */ - for (i = 0; i < dc->res_pool->pipe_count; i++) { - struct timing_generator *tg = dc->res_pool->timing_generators[i]; + /* Blank controller using driver code instead of + * command table. + */ if (tg->funcs->is_tg_enabled(tg)) { tg->funcs->set_blank(tg, true); hwss_wait_for_blank_complete(tg); } } - /* Reset all MPCC muxes */ - dc->res_pool->mpc->funcs->mpc_init(dc->res_pool->mpc); + /* Cannot reset the MPC mux if seamless boot */ + if (!can_apply_seamless_boot) + dc->res_pool->mpc->funcs->mpc_init(dc->res_pool->mpc); - for (i = 0; i < dc->res_pool->timing_generator_count; i++) { + for (i = 0; i < dc->res_pool->pipe_count; i++) { struct timing_generator *tg = dc->res_pool->timing_generators[i]; - struct pipe_ctx *pipe_ctx = &context->res_ctx.pipe_ctx[i]; struct hubp *hubp = dc->res_pool->hubps[i]; struct dpp *dpp = dc->res_pool->dpps[i]; + struct pipe_ctx *pipe_ctx = &context->res_ctx.pipe_ctx[i]; + + // W/A for issue with dc_post_update_surfaces_to_stream + hubp->power_gated = true; + + /* There is assumption that pipe_ctx is not mapping irregularly + * to non-preferred front end. If pipe_ctx->stream is not NULL, + * we will use the pipe, so don't disable + */ + if (pipe_ctx->stream != NULL) + continue; + + dpp->funcs->dpp_reset(dpp); pipe_ctx->stream_res.tg = tg; pipe_ctx->pipe_idx = i; @@ -1073,18 +1029,9 @@ static void dcn10_init_hw(struct dc *dc) pipe_ctx->stream_res.opp = dc->res_pool->opps[i]; hwss1_plane_atomic_disconnect(dc, pipe_ctx); - } - - for (i = 0; i < dc->res_pool->pipe_count; i++) { - struct timing_generator *tg = dc->res_pool->timing_generators[i]; if (tg->funcs->is_tg_enabled(tg)) tg->funcs->unlock(tg); - } - - for (i = 0; i < dc->res_pool->pipe_count; i++) { - struct timing_generator *tg = dc->res_pool->timing_generators[i]; - struct pipe_ctx *pipe_ctx = &context->res_ctx.pipe_ctx[i]; dcn10_disable_plane(dc, pipe_ctx); @@ -1093,10 +1040,73 @@ static void dcn10_init_hw(struct dc *dc) tg->funcs->tg_init(tg); } +} + +static void dcn10_init_hw(struct dc *dc) +{ + int i; + struct abm *abm = dc->res_pool->abm; + struct dmcu *dmcu = dc->res_pool->dmcu; + struct dce_hwseq *hws = dc->hwseq; + struct dc_bios *dcb = dc->ctx->dc_bios; + + if (IS_FPGA_MAXIMUS_DC(dc->ctx->dce_environment)) { + REG_WRITE(REFCLK_CNTL, 0); + REG_UPDATE(DCHUBBUB_GLOBAL_TIMER_CNTL, DCHUBBUB_GLOBAL_TIMER_ENABLE, 1); + REG_WRITE(DIO_MEM_PWR_CTRL, 0); - /* end of FPGA. Below if real ASIC */ - if (IS_FPGA_MAXIMUS_DC(dc->ctx->dce_environment)) + if (!dc->debug.disable_clock_gate) { + /* enable all DCN clock gating */ + REG_WRITE(DCCG_GATE_DISABLE_CNTL, 0); + + REG_WRITE(DCCG_GATE_DISABLE_CNTL2, 0); + + REG_UPDATE(DCFCLK_CNTL, DCFCLK_GATE_DIS, 0); + } + + enable_power_gating_plane(dc->hwseq, true); + + /* end of FPGA. Below if real ASIC */ return; + } + + if (!dcb->funcs->is_accelerated_mode(dcb)) { + bool allow_self_fresh_force_enable = + hububu1_is_allow_self_refresh_enabled( + dc->res_pool->hubbub); + + bios_golden_init(dc); + + /* WA for making DF sleep when idle after resume from S0i3. + * DCHUBBUB_ARB_ALLOW_SELF_REFRESH_FORCE_ENABLE is set to 1 by + * command table, if DCHUBBUB_ARB_ALLOW_SELF_REFRESH_FORCE_ENABLE = 0 + * before calling command table and it changed to 1 after, + * it should be set back to 0. + */ + if (allow_self_fresh_force_enable == false && + hububu1_is_allow_self_refresh_enabled(dc->res_pool->hubbub)) + hubbub1_allow_self_refresh_control(dc->res_pool->hubbub, true); + + disable_vga(dc->hwseq); + } + + for (i = 0; i < dc->link_count; i++) { + /* Power up AND update implementation according to the + * required signal (which may be different from the + * default signal on connector). + */ + struct dc_link *link = dc->links[i]; + + if (link->link_enc->connector.id == CONNECTOR_ID_EDP) + dc->hwss.edp_power_control(link, true); + + link->link_enc->funcs->hw_init(link->link_enc); + + /* Check for enabled DIG to identify enabled display */ + if (link->link_enc->funcs->is_dig_enabled && + link->link_enc->funcs->is_dig_enabled(link->link_enc)) + link->link_status.link_active = true; + } for (i = 0; i < dc->res_pool->audio_count; i++) { struct audio *audio = dc->res_pool->audios[i]; @@ -1126,7 +1136,10 @@ static void dcn10_init_hw(struct dc *dc) enable_power_gating_plane(dc->hwseq, true); - memset(&dc->res_pool->dccg->clks, 0, sizeof(dc->res_pool->dccg->clks)); + memset(&dc->res_pool->clk_mgr->clks, 0, sizeof(dc->res_pool->clk_mgr->clks)); + + if (dc->hwss.init_pipes) + dc->hwss.init_pipes(dc, dc->current_state); } static void reset_hw_ctx_wrap( @@ -1152,11 +1165,13 @@ static void reset_hw_ctx_wrap( struct clock_source *old_clk = pipe_ctx_old->clock_source; reset_back_end_for_pipe(dc, pipe_ctx_old, dc->current_state); + if (dc->hwss.enable_stream_gating) { + dc->hwss.enable_stream_gating(dc, pipe_ctx); + } if (old_clk) old_clk->funcs->cs_power_down(old_clk); } } - } static bool patch_address_for_sbs_tb_stereo( @@ -1201,7 +1216,8 @@ static void dcn10_update_plane_addr(const struct dc *dc, struct pipe_ctx *pipe_c pipe_ctx->plane_res.hubp->funcs->hubp_program_surface_flip_and_addr( pipe_ctx->plane_res.hubp, &plane_state->address, - plane_state->flip_immediate); + plane_state->flip_immediate, + 0); plane_state->status.requested_address = plane_state->address; @@ -1226,7 +1242,8 @@ static bool dcn10_set_input_transfer_func(struct pipe_ctx *pipe_ctx, tf = plane_state->in_transfer_func; if (plane_state->gamma_correction && - !plane_state->gamma_correction->is_identity + !dpp_base->ctx->dc->debug.always_use_regamma + && !plane_state->gamma_correction->is_identity && dce_use_lut(plane_state->format)) dpp_base->funcs->dpp_program_input_lut(dpp_base, plane_state->gamma_correction); @@ -1399,7 +1416,7 @@ static void dcn10_enable_per_frame_crtc_position_reset( if (grouped_pipes[i]->stream_res.tg->funcs->enable_crtc_reset) grouped_pipes[i]->stream_res.tg->funcs->enable_crtc_reset( grouped_pipes[i]->stream_res.tg, - grouped_pipes[i]->stream->triggered_crtc_reset.event_source->status.primary_otg_inst, + 0, &grouped_pipes[i]->stream->triggered_crtc_reset); DC_SYNC_INFO("Waiting for trigger\n"); @@ -1603,7 +1620,7 @@ static void mmhub_read_vm_context0_settings(struct dcn10_hubp *hubp1, } -static void dcn10_program_pte_vm(struct dce_hwseq *hws, struct hubp *hubp) +void dcn10_program_pte_vm(struct dce_hwseq *hws, struct hubp *hubp) { struct dcn10_hubp *hubp1 = TO_DCN10_HUBP(hubp); struct vm_system_aperture_param apt = { {{ 0 } } }; @@ -1703,33 +1720,22 @@ static void program_gamut_remap(struct pipe_ctx *pipe_ctx) pipe_ctx->plane_res.dpp->funcs->dpp_set_gamut_remap(pipe_ctx->plane_res.dpp, &adjust); } - -static void program_csc_matrix(struct pipe_ctx *pipe_ctx, +static void dcn10_program_output_csc(struct dc *dc, + struct pipe_ctx *pipe_ctx, enum dc_color_space colorspace, - uint16_t *matrix) + uint16_t *matrix, + int opp_id) { if (pipe_ctx->stream->csc_color_matrix.enable_adjustment == true) { - if (pipe_ctx->plane_res.dpp->funcs->dpp_set_csc_adjustment != NULL) - pipe_ctx->plane_res.dpp->funcs->dpp_set_csc_adjustment(pipe_ctx->plane_res.dpp, matrix); + if (pipe_ctx->plane_res.dpp->funcs->dpp_set_csc_adjustment != NULL) + pipe_ctx->plane_res.dpp->funcs->dpp_set_csc_adjustment(pipe_ctx->plane_res.dpp, matrix); } else { if (pipe_ctx->plane_res.dpp->funcs->dpp_set_csc_default != NULL) pipe_ctx->plane_res.dpp->funcs->dpp_set_csc_default(pipe_ctx->plane_res.dpp, colorspace); } } -static void dcn10_program_output_csc(struct dc *dc, - struct pipe_ctx *pipe_ctx, - enum dc_color_space colorspace, - uint16_t *matrix, - int opp_id) -{ - if (pipe_ctx->plane_res.dpp->funcs->dpp_set_csc_adjustment != NULL) - program_csc_matrix(pipe_ctx, - colorspace, - matrix); -} - -static bool is_lower_pipe_tree_visible(struct pipe_ctx *pipe_ctx) +bool is_lower_pipe_tree_visible(struct pipe_ctx *pipe_ctx) { if (pipe_ctx->plane_state->visible) return true; @@ -1738,7 +1744,7 @@ static bool is_lower_pipe_tree_visible(struct pipe_ctx *pipe_ctx) return false; } -static bool is_upper_pipe_tree_visible(struct pipe_ctx *pipe_ctx) +bool is_upper_pipe_tree_visible(struct pipe_ctx *pipe_ctx) { if (pipe_ctx->plane_state->visible) return true; @@ -1747,7 +1753,7 @@ static bool is_upper_pipe_tree_visible(struct pipe_ctx *pipe_ctx) return false; } -static bool is_pipe_tree_visible(struct pipe_ctx *pipe_ctx) +bool is_pipe_tree_visible(struct pipe_ctx *pipe_ctx) { if (pipe_ctx->plane_state->visible) return true; @@ -1780,7 +1786,7 @@ bool is_rgb_cspace(enum dc_color_space output_color_space) } } -static void dcn10_get_surface_visual_confirm_color( +void dcn10_get_surface_visual_confirm_color( const struct pipe_ctx *pipe_ctx, struct tg_color *color) { @@ -1816,7 +1822,7 @@ static void dcn10_get_surface_visual_confirm_color( } } -static void dcn10_get_hdr_visual_confirm_color( +void dcn10_get_hdr_visual_confirm_color( struct pipe_ctx *pipe_ctx, struct tg_color *color) { @@ -1943,10 +1949,6 @@ static void dcn10_update_mpcc(struct dc *dc, struct pipe_ctx *pipe_ctx) struct mpc *mpc = dc->res_pool->mpc; struct mpc_tree *mpc_tree_params = &(pipe_ctx->stream_res.opp->mpc_tree_params); - - - /* TODO: proper fix once fpga works */ - if (dc->debug.visual_confirm == VISUAL_CONFIRM_HDR) { dcn10_get_hdr_visual_confirm_color( pipe_ctx, &blnd_cfg.black_color); @@ -2026,8 +2028,6 @@ static void update_scaler(struct pipe_ctx *pipe_ctx) bool per_pixel_alpha = pipe_ctx->plane_state->per_pixel_alpha && pipe_ctx->bottom_pipe; - /* TODO: proper fix once fpga works */ - pipe_ctx->plane_res.scl_data.lb_params.alpha_en = per_pixel_alpha; pipe_ctx->plane_res.scl_data.lb_params.depth = LB_PIXEL_DEPTH_30BPP; /* scaler configuration */ @@ -2035,7 +2035,7 @@ static void update_scaler(struct pipe_ctx *pipe_ctx) pipe_ctx->plane_res.dpp, &pipe_ctx->plane_res.scl_data); } -static void update_dchubp_dpp( +void update_dchubp_dpp( struct dc *dc, struct pipe_ctx *pipe_ctx, struct dc_state *context) @@ -2052,16 +2052,22 @@ static void update_dchubp_dpp( */ if (plane_state->update_flags.bits.full_update) { bool should_divided_by_2 = context->bw.dcn.clk.dppclk_khz <= - dc->res_pool->dccg->clks.dispclk_khz / 2; + dc->res_pool->clk_mgr->clks.dispclk_khz / 2; dpp->funcs->dpp_dppclk_control( dpp, should_divided_by_2, true); - dc->res_pool->dccg->clks.dppclk_khz = should_divided_by_2 ? - dc->res_pool->dccg->clks.dispclk_khz / 2 : - dc->res_pool->dccg->clks.dispclk_khz; + if (dc->res_pool->dccg) + dc->res_pool->dccg->funcs->update_dpp_dto( + dc->res_pool->dccg, + dpp->inst, + pipe_ctx->plane_res.bw.dppclk_khz); + else + dc->res_pool->clk_mgr->clks.dppclk_khz = should_divided_by_2 ? + dc->res_pool->clk_mgr->clks.dispclk_khz / 2 : + dc->res_pool->clk_mgr->clks.dispclk_khz; } /* TODO: Need input parameter to tell current DCHUB pipe tie to which OTG @@ -2077,6 +2083,10 @@ static void update_dchubp_dpp( &pipe_ctx->ttu_regs, &pipe_ctx->rq_regs, &pipe_ctx->pipe_dlg_param); + hubp->funcs->hubp_setup_interdependent( + hubp, + &pipe_ctx->dlg_regs, + &pipe_ctx->ttu_regs); } size.grph.surface_size = pipe_ctx->plane_res.scl_data.viewport; @@ -2130,7 +2140,8 @@ static void update_dchubp_dpp( plane_state->update_flags.bits.swizzle_change || plane_state->update_flags.bits.dcc_change || plane_state->update_flags.bits.bpp_change || - plane_state->update_flags.bits.scaling_change) { + plane_state->update_flags.bits.scaling_change || + plane_state->update_flags.bits.plane_size_change) { hubp->funcs->hubp_program_surface_config( hubp, plane_state->format, @@ -2164,6 +2175,15 @@ static void dcn10_blank_pixel_data( color_space = stream->output_color_space; color_space_to_black_color(dc, color_space, &black_color); + /* + * The way 420 is packed, 2 channels carry Y component, 1 channel + * alternate between Cb and Cr, so both channels need the pixel + * value for Y + */ + if (stream->timing.pixel_encoding == PIXEL_ENCODING_YCBCR420) + black_color.color_r_cr = black_color.color_g_y; + + if (stream_res->tg->funcs->set_blank_color) stream_res->tg->funcs->set_blank_color( stream_res->tg, @@ -2172,8 +2192,10 @@ static void dcn10_blank_pixel_data( if (!blank) { if (stream_res->tg->funcs->set_blank) stream_res->tg->funcs->set_blank(stream_res->tg, blank); - if (stream_res->abm) + if (stream_res->abm) { + stream_res->abm->funcs->set_pipe(stream_res->abm, stream_res->tg->inst + 1); stream_res->abm->funcs->set_abm_level(stream_res->abm, stream->abm_level); + } } else if (blank) { if (stream_res->abm) stream_res->abm->funcs->set_abm_immediate_disable(stream_res->abm); @@ -2182,7 +2204,7 @@ static void dcn10_blank_pixel_data( } } -static void set_hdr_multiplier(struct pipe_ctx *pipe_ctx) +void set_hdr_multiplier(struct pipe_ctx *pipe_ctx) { struct fixed31_32 multiplier = dc_fixpt_from_fraction( pipe_ctx->plane_state->sdr_white_level, 80); @@ -2248,56 +2270,14 @@ static void program_all_pipe_in_tree( } - if (pipe_ctx->plane_state != NULL) { + if (pipe_ctx->plane_state != NULL) dcn10_program_pipe(dc, pipe_ctx, context); - } - if (pipe_ctx->bottom_pipe != NULL && pipe_ctx->bottom_pipe != pipe_ctx) { + if (pipe_ctx->bottom_pipe != NULL && pipe_ctx->bottom_pipe != pipe_ctx) program_all_pipe_in_tree(dc, pipe_ctx->bottom_pipe, context); - } -} - -static void dcn10_pplib_apply_display_requirements( - struct dc *dc, - struct dc_state *context) -{ - struct dm_pp_display_configuration *pp_display_cfg = &context->pp_display_cfg; - - pp_display_cfg->min_engine_clock_khz = dc->res_pool->dccg->clks.dcfclk_khz; - pp_display_cfg->min_memory_clock_khz = dc->res_pool->dccg->clks.fclk_khz; - pp_display_cfg->min_engine_clock_deep_sleep_khz = dc->res_pool->dccg->clks.dcfclk_deep_sleep_khz; - pp_display_cfg->min_dcfc_deep_sleep_clock_khz = dc->res_pool->dccg->clks.dcfclk_deep_sleep_khz; - pp_display_cfg->min_dcfclock_khz = dc->res_pool->dccg->clks.dcfclk_khz; - pp_display_cfg->disp_clk_khz = dc->res_pool->dccg->clks.dispclk_khz; - dce110_fill_display_configs(context, pp_display_cfg); - - if (memcmp(&dc->prev_display_config, pp_display_cfg, sizeof( - struct dm_pp_display_configuration)) != 0) - dm_pp_apply_display_requirements(dc->ctx, pp_display_cfg); - - dc->prev_display_config = *pp_display_cfg; -} - -static void optimize_shared_resources(struct dc *dc) -{ - if (dc->current_state->stream_count == 0) { - /* S0i2 message */ - dcn10_pplib_apply_display_requirements(dc, dc->current_state); - } - - if (dc->debug.pplib_wm_report_mode == WM_REPORT_OVERRIDE) - dcn_bw_notify_pplib_of_wm_ranges(dc); } -static void ready_shared_resources(struct dc *dc, struct dc_state *context) -{ - /* S0i2 message */ - if (dc->current_state->stream_count == 0 && - context->stream_count != 0) - dcn10_pplib_apply_display_requirements(dc, context); -} - -static struct pipe_ctx *find_top_pipe_for_stream( +struct pipe_ctx *find_top_pipe_for_stream( struct dc *dc, struct dc_state *context, const struct dc_stream_state *stream) @@ -2370,9 +2350,10 @@ static void dcn10_apply_ctx_for_surface( } } - if (!pipe_ctx->plane_state && - old_pipe_ctx->plane_state && - old_pipe_ctx->stream_res.tg == tg) { + if ((!pipe_ctx->plane_state || + pipe_ctx->stream_res.tg != old_pipe_ctx->stream_res.tg) && + old_pipe_ctx->plane_state && + old_pipe_ctx->stream_res.tg == tg) { dc->hwss.plane_atomic_disconnect(dc, old_pipe_ctx); removed_pipe[i] = true; @@ -2387,6 +2368,27 @@ static void dcn10_apply_ctx_for_surface( dcn10_pipe_control_lock(dc, top_pipe_to_program, false); + if (top_pipe_to_program->plane_state && + top_pipe_to_program->plane_state->update_flags.bits.full_update) + for (i = 0; i < dc->res_pool->pipe_count; i++) { + struct pipe_ctx *pipe_ctx = &context->res_ctx.pipe_ctx[i]; + tg = pipe_ctx->stream_res.tg; + /* Skip inactive pipes and ones already updated */ + if (!pipe_ctx->stream || pipe_ctx->stream == stream + || !pipe_ctx->plane_state + || !tg->funcs->is_tg_enabled(tg)) + continue; + + tg->funcs->lock(tg); + + pipe_ctx->plane_res.hubp->funcs->hubp_setup_interdependent( + pipe_ctx->plane_res.hubp, + &pipe_ctx->dlg_regs, + &pipe_ctx->ttu_regs); + + tg->funcs->unlock(tg); + } + if (num_planes == 0) false_optc_underflow_wa(dc, stream, tg); @@ -2398,10 +2400,25 @@ static void dcn10_apply_ctx_for_surface( hubbub1_wm_change_req_wa(dc->res_pool->hubbub); } -static void dcn10_set_bandwidth( +static void dcn10_stereo_hw_frame_pack_wa(struct dc *dc, struct dc_state *context) +{ + uint8_t i; + + for (i = 0; i < context->stream_count; i++) { + if (context->streams[i]->timing.timing_3d_format + == TIMING_3D_FORMAT_HW_FRAME_PACKING) { + /* + * Disable stutter + */ + hubbub1_allow_self_refresh_control(dc->res_pool->hubbub, false); + break; + } + } +} + +static void dcn10_prepare_bandwidth( struct dc *dc, - struct dc_state *context, - bool safe_to_lower) + struct dc_state *context) { if (dc->debug.sanity_checks) dcn10_verify_allow_pstate_change_high(dc); @@ -2410,18 +2427,50 @@ static void dcn10_set_bandwidth( if (context->stream_count == 0) context->bw.dcn.clk.phyclk_khz = 0; - dc->res_pool->dccg->funcs->update_clocks( - dc->res_pool->dccg, - &context->bw.dcn.clk, - safe_to_lower); + dc->res_pool->clk_mgr->funcs->update_clocks( + dc->res_pool->clk_mgr, + context, + false); + } + + hubbub1_program_watermarks(dc->res_pool->hubbub, + &context->bw.dcn.watermarks, + dc->res_pool->ref_clock_inKhz / 1000, + true); + dcn10_stereo_hw_frame_pack_wa(dc, context); + + if (dc->debug.pplib_wm_report_mode == WM_REPORT_OVERRIDE) + dcn_bw_notify_pplib_of_wm_ranges(dc); + + if (dc->debug.sanity_checks) + dcn10_verify_allow_pstate_change_high(dc); +} + +static void dcn10_optimize_bandwidth( + struct dc *dc, + struct dc_state *context) +{ + if (dc->debug.sanity_checks) + dcn10_verify_allow_pstate_change_high(dc); + + if (!IS_FPGA_MAXIMUS_DC(dc->ctx->dce_environment)) { + if (context->stream_count == 0) + context->bw.dcn.clk.phyclk_khz = 0; - dcn10_pplib_apply_display_requirements(dc, context); + dc->res_pool->clk_mgr->funcs->update_clocks( + dc->res_pool->clk_mgr, + context, + true); } hubbub1_program_watermarks(dc->res_pool->hubbub, &context->bw.dcn.watermarks, dc->res_pool->ref_clock_inKhz / 1000, true); + dcn10_stereo_hw_frame_pack_wa(dc, context); + + if (dc->debug.pplib_wm_report_mode == WM_REPORT_OVERRIDE) + dcn_bw_notify_pplib_of_wm_ranges(dc); if (dc->debug.sanity_checks) dcn10_verify_allow_pstate_change_high(dc); @@ -2504,7 +2553,7 @@ static void dcn10_config_stereo_parameters( timing_3d_format == TIMING_3D_FORMAT_DP_HDMI_INBAND_FA || timing_3d_format == TIMING_3D_FORMAT_SIDEBAND_FA) { enum display_dongle_type dongle = \ - stream->sink->link->ddc->dongle_type; + stream->link->ddc->dongle_type; if (dongle == DISPLAY_DONGLE_DP_VGA_CONVERTER || dongle == DISPLAY_DONGLE_DP_DVI_CONVERTER || dongle == DISPLAY_DONGLE_DP_HDMI_CONVERTER) @@ -2635,7 +2684,7 @@ static void dcn10_set_cursor_position(struct pipe_ctx *pipe_ctx) struct hubp *hubp = pipe_ctx->plane_res.hubp; struct dpp *dpp = pipe_ctx->plane_res.dpp; struct dc_cursor_mi_param param = { - .pixel_clk_khz = pipe_ctx->stream->timing.pix_clk_khz, + .pixel_clk_khz = pipe_ctx->stream->timing.pix_clk_100hz / 10, .ref_clk_khz = pipe_ctx->stream->ctx->dc->res_pool->ref_clock_inKhz, .viewport = pipe_ctx->plane_res.scl_data.viewport, .h_scale_ratio = pipe_ctx->plane_res.scl_data.ratios.horz, @@ -2644,8 +2693,8 @@ static void dcn10_set_cursor_position(struct pipe_ctx *pipe_ctx) .mirror = pipe_ctx->plane_state->horizontal_mirror }; - pos_cpy.x -= pipe_ctx->plane_state->dst_rect.x; - pos_cpy.y -= pipe_ctx->plane_state->dst_rect.y; + pos_cpy.x_hotspot += pipe_ctx->plane_state->dst_rect.x; + pos_cpy.y_hotspot += pipe_ctx->plane_state->dst_rect.y; if (pipe_ctx->plane_state->address.type == PLN_ADDR_TYPE_VIDEO_PROGRESSIVE) @@ -2692,10 +2741,151 @@ static void dcn10_set_cursor_sdr_white_level(struct pipe_ctx *pipe_ctx) pipe_ctx->plane_res.dpp, &opt_attr); } +/** +* apply_front_porch_workaround TODO FPGA still need? +* +* This is a workaround for a bug that has existed since R5xx and has not been +* fixed keep Front porch at minimum 2 for Interlaced mode or 1 for progressive. +*/ +static void apply_front_porch_workaround( + struct dc_crtc_timing *timing) +{ + if (timing->flags.INTERLACE == 1) { + if (timing->v_front_porch < 2) + timing->v_front_porch = 2; + } else { + if (timing->v_front_porch < 1) + timing->v_front_porch = 1; + } +} + +int get_vupdate_offset_from_vsync(struct pipe_ctx *pipe_ctx) +{ + struct timing_generator *optc = pipe_ctx->stream_res.tg; + const struct dc_crtc_timing *dc_crtc_timing = &pipe_ctx->stream->timing; + struct dc_crtc_timing patched_crtc_timing; + int vesa_sync_start; + int asic_blank_end; + int interlace_factor; + int vertical_line_start; + + patched_crtc_timing = *dc_crtc_timing; + apply_front_porch_workaround(&patched_crtc_timing); + + interlace_factor = patched_crtc_timing.flags.INTERLACE ? 2 : 1; + + vesa_sync_start = patched_crtc_timing.v_addressable + + patched_crtc_timing.v_border_bottom + + patched_crtc_timing.v_front_porch; + + asic_blank_end = (patched_crtc_timing.v_total - + vesa_sync_start - + patched_crtc_timing.v_border_top) + * interlace_factor; + + vertical_line_start = asic_blank_end - + optc->dlg_otg_param.vstartup_start + 1; + + return vertical_line_start; +} + +static void calc_vupdate_position( + struct pipe_ctx *pipe_ctx, + uint32_t *start_line, + uint32_t *end_line) +{ + const struct dc_crtc_timing *dc_crtc_timing = &pipe_ctx->stream->timing; + int vline_int_offset_from_vupdate = + pipe_ctx->stream->periodic_interrupt0.lines_offset; + int vupdate_offset_from_vsync = get_vupdate_offset_from_vsync(pipe_ctx); + int start_position; + + if (vline_int_offset_from_vupdate > 0) + vline_int_offset_from_vupdate--; + else if (vline_int_offset_from_vupdate < 0) + vline_int_offset_from_vupdate++; + + start_position = vline_int_offset_from_vupdate + vupdate_offset_from_vsync; + + if (start_position >= 0) + *start_line = start_position; + else + *start_line = dc_crtc_timing->v_total + start_position - 1; + + *end_line = *start_line + 2; + + if (*end_line >= dc_crtc_timing->v_total) + *end_line = 2; +} + +static void cal_vline_position( + struct pipe_ctx *pipe_ctx, + enum vline_select vline, + uint32_t *start_line, + uint32_t *end_line) +{ + enum vertical_interrupt_ref_point ref_point = INVALID_POINT; + + if (vline == VLINE0) + ref_point = pipe_ctx->stream->periodic_interrupt0.ref_point; + else if (vline == VLINE1) + ref_point = pipe_ctx->stream->periodic_interrupt1.ref_point; + + switch (ref_point) { + case START_V_UPDATE: + calc_vupdate_position( + pipe_ctx, + start_line, + end_line); + break; + case START_V_SYNC: + // Suppose to do nothing because vsync is 0; + break; + default: + ASSERT(0); + break; + } +} + +static void dcn10_setup_periodic_interrupt( + struct pipe_ctx *pipe_ctx, + enum vline_select vline) +{ + struct timing_generator *tg = pipe_ctx->stream_res.tg; + + if (vline == VLINE0) { + uint32_t start_line = 0; + uint32_t end_line = 0; + + cal_vline_position(pipe_ctx, vline, &start_line, &end_line); + + tg->funcs->setup_vertical_interrupt0(tg, start_line, end_line); + + } else if (vline == VLINE1) { + pipe_ctx->stream_res.tg->funcs->setup_vertical_interrupt1( + tg, + pipe_ctx->stream->periodic_interrupt1.lines_offset); + } +} + +static void dcn10_setup_vupdate_interrupt(struct pipe_ctx *pipe_ctx) +{ + struct timing_generator *tg = pipe_ctx->stream_res.tg; + int start_line = get_vupdate_offset_from_vsync(pipe_ctx); + + if (start_line < 0) { + ASSERT(0); + start_line = 0; + } + + if (tg->funcs->setup_vertical_interrupt2) + tg->funcs->setup_vertical_interrupt2(tg, start_line); +} + static const struct hw_sequencer_funcs dcn10_funcs = { .program_gamut_remap = program_gamut_remap, - .program_csc_matrix = program_csc_matrix, .init_hw = dcn10_init_hw, + .init_pipes = dcn10_init_pipes, .apply_ctx_to_hw = dce110_apply_ctx_to_hw, .apply_ctx_for_surface = dcn10_apply_ctx_for_surface, .update_plane_addr = dcn10_update_plane_addr, @@ -2721,7 +2911,8 @@ static const struct hw_sequencer_funcs dcn10_funcs = { .disable_plane = dcn10_disable_plane, .blank_pixel_data = dcn10_blank_pixel_data, .pipe_control_lock = dcn10_pipe_control_lock, - .set_bandwidth = dcn10_set_bandwidth, + .prepare_bandwidth = dcn10_prepare_bandwidth, + .optimize_bandwidth = dcn10_optimize_bandwidth, .reset_hw_ctx_wrap = reset_hw_ctx_wrap, .enable_stream_timing = dcn10_enable_stream_timing, .set_drr = set_drr, @@ -2731,17 +2922,18 @@ static const struct hw_sequencer_funcs dcn10_funcs = { .set_avmute = dce110_set_avmute, .log_hw_state = dcn10_log_hw_state, .get_hw_state = dcn10_get_hw_state, + .clear_status_bits = dcn10_clear_status_bits, .wait_for_mpcc_disconnect = dcn10_wait_for_mpcc_disconnect, - .ready_shared_resources = ready_shared_resources, - .optimize_shared_resources = optimize_shared_resources, - .pplib_apply_display_requirements = - dcn10_pplib_apply_display_requirements, .edp_backlight_control = hwss_edp_backlight_control, .edp_power_control = hwss_edp_power_control, .edp_wait_for_hpd_ready = hwss_edp_wait_for_hpd_ready, .set_cursor_position = dcn10_set_cursor_position, .set_cursor_attribute = dcn10_set_cursor_attribute, - .set_cursor_sdr_white_level = dcn10_set_cursor_sdr_white_level + .set_cursor_sdr_white_level = dcn10_set_cursor_sdr_white_level, + .disable_stream_gating = NULL, + .enable_stream_gating = NULL, + .setup_periodic_interrupt = dcn10_setup_periodic_interrupt, + .setup_vupdate_interrupt = dcn10_setup_vupdate_interrupt }; diff --git a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_hw_sequencer.h b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_hw_sequencer.h index 84d461e0ed3e..6d66084df55f 100644 --- a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_hw_sequencer.h +++ b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_hw_sequencer.h @@ -51,4 +51,36 @@ void dcn10_get_hw_state( char *pBuf, unsigned int bufSize, unsigned int mask); +void dcn10_clear_status_bits(struct dc *dc, unsigned int mask); + +bool is_lower_pipe_tree_visible(struct pipe_ctx *pipe_ctx); + +bool is_upper_pipe_tree_visible(struct pipe_ctx *pipe_ctx); + +bool is_pipe_tree_visible(struct pipe_ctx *pipe_ctx); + +void dcn10_program_pte_vm(struct dce_hwseq *hws, struct hubp *hubp); + +void set_hdr_multiplier(struct pipe_ctx *pipe_ctx); + +void dcn10_get_surface_visual_confirm_color( + const struct pipe_ctx *pipe_ctx, + struct tg_color *color); + +void dcn10_get_hdr_visual_confirm_color( + struct pipe_ctx *pipe_ctx, + struct tg_color *color); + +void update_dchubp_dpp( + struct dc *dc, + struct pipe_ctx *pipe_ctx, + struct dc_state *context); + +struct pipe_ctx *find_top_pipe_for_stream( + struct dc *dc, + struct dc_state *context, + const struct dc_stream_state *stream); + +int get_vupdate_offset_from_vsync(struct pipe_ctx *pipe_ctx); + #endif /* __DC_HWSS_DCN10_H__ */ diff --git a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_hw_sequencer_debug.c b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_hw_sequencer_debug.c index 64158900730f..98f41d250978 100644 --- a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_hw_sequencer_debug.c +++ b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_hw_sequencer_debug.c @@ -40,10 +40,10 @@ #include "ipp.h" #include "mpc.h" #include "reg_helper.h" -#include "custom_float.h" #include "dcn10_hubp.h" #include "dcn10_hubbub.h" #include "dcn10_cm_common.h" +#include "dcn10_clk_mgr.h" static unsigned int snprintf_count(char *pBuf, unsigned int bufSize, char *fmt, ...) { @@ -71,7 +71,7 @@ static unsigned int snprintf_count(char *pBuf, unsigned int bufSize, char *fmt, static unsigned int dcn10_get_hubbub_state(struct dc *dc, char *pBuf, unsigned int bufSize) { struct dc_context *dc_ctx = dc->ctx; - struct dcn_hubbub_wm wm = {0}; + struct dcn_hubbub_wm wm; int i; unsigned int chars_printed = 0; @@ -80,7 +80,8 @@ static unsigned int dcn10_get_hubbub_state(struct dc *dc, char *pBuf, unsigned i const uint32_t ref_clk_mhz = dc_ctx->dc->res_pool->ref_clock_inKhz / 1000; static const unsigned int frac = 1000; - hubbub1_wm_read_state(dc->res_pool->hubbub, &wm); + memset(&wm, 0, sizeof(struct dcn_hubbub_wm)); + dc->res_pool->hubbub->funcs->wm_read_state(dc->res_pool->hubbub, &wm); chars_printed = snprintf_count(pBuf, remaining_buffer, "wm_set_index,data_urgent,pte_meta_urgent,sr_enter,sr_exit,dram_clk_chanage\n"); remaining_buffer -= chars_printed; @@ -418,20 +419,22 @@ static unsigned int dcn10_get_otg_states(struct dc *dc, char *pBuf, unsigned int unsigned int remaining_buffer = bufSize; chars_printed = snprintf_count(pBuf, remaining_buffer, "instance,v_bs,v_be,v_ss,v_se,vpol,vmax,vmin,vmax_sel,vmin_sel," - "h_bs,h_be,h_ss,h_se,hpol,htot,vtot,underflow\n"); + "h_bs,h_be,h_ss,h_se,hpol,htot,vtot,underflow,pixelclk[khz]\n"); remaining_buffer -= chars_printed; pBuf += chars_printed; for (i = 0; i < pool->timing_generator_count; i++) { struct timing_generator *tg = pool->timing_generators[i]; struct dcn_otg_state s = {0}; + int pix_clk = 0; optc1_read_otg_state(DCN10TG_FROM_TG(tg), &s); + pix_clk = dc->current_state->res_ctx.pipe_ctx[i].stream_res.pix_clk_params.requested_pix_clk_100hz / 10; //only print if OTG master is enabled if (s.otg_enabled & 1) { chars_printed = snprintf_count(pBuf, remaining_buffer, "%x,%d,%d,%d,%d,%d,%d,%d,%d,%d," - "%d,%d,%d,%d,%d,%d,%d,%d" + "%d,%d,%d,%d,%d,%d,%d,%d,%d" "\n", tg->inst, s.v_blank_start, @@ -450,16 +453,11 @@ static unsigned int dcn10_get_otg_states(struct dc *dc, char *pBuf, unsigned int s.h_sync_a_pol, s.h_total, s.v_total, - s.underflow_occurred_status); + s.underflow_occurred_status, + pix_clk); remaining_buffer -= chars_printed; pBuf += chars_printed; - - // Clear underflow for debug purposes - // We want to keep underflow sticky bit on for the longevity tests outside of test environment. - // This function is called only from Windows or Diags test environment, hence it's safe to clear - // it from here without affecting the original intent. - tg->funcs->clear_optc_underflow(tg); } } @@ -469,19 +467,75 @@ static unsigned int dcn10_get_otg_states(struct dc *dc, char *pBuf, unsigned int static unsigned int dcn10_get_clock_states(struct dc *dc, char *pBuf, unsigned int bufSize) { unsigned int chars_printed = 0; + unsigned int remaining_buffer = bufSize; - chars_printed = snprintf_count(pBuf, bufSize, "dcfclk_khz,dcfclk_deep_sleep_khz,dispclk_khz," - "dppclk_khz,max_supported_dppclk_khz,fclk_khz,socclk_khz\n" - "%d,%d,%d,%d,%d,%d,%d\n", + chars_printed = snprintf_count(pBuf, bufSize, "dcfclk,dcfclk_deep_sleep,dispclk," + "dppclk,fclk,socclk\n" + "%d,%d,%d,%d,%d,%d\n", dc->current_state->bw.dcn.clk.dcfclk_khz, dc->current_state->bw.dcn.clk.dcfclk_deep_sleep_khz, dc->current_state->bw.dcn.clk.dispclk_khz, dc->current_state->bw.dcn.clk.dppclk_khz, - dc->current_state->bw.dcn.clk.max_supported_dppclk_khz, dc->current_state->bw.dcn.clk.fclk_khz, dc->current_state->bw.dcn.clk.socclk_khz); - return chars_printed; + remaining_buffer -= chars_printed; + pBuf += chars_printed; + + return bufSize - remaining_buffer; +} + +static void dcn10_clear_otpc_underflow(struct dc *dc) +{ + struct resource_pool *pool = dc->res_pool; + int i; + + for (i = 0; i < pool->timing_generator_count; i++) { + struct timing_generator *tg = pool->timing_generators[i]; + struct dcn_otg_state s = {0}; + + optc1_read_otg_state(DCN10TG_FROM_TG(tg), &s); + + if (s.otg_enabled & 1) + tg->funcs->clear_optc_underflow(tg); + } +} + +static void dcn10_clear_hubp_underflow(struct dc *dc) +{ + struct resource_pool *pool = dc->res_pool; + int i; + + for (i = 0; i < pool->pipe_count; i++) { + struct hubp *hubp = pool->hubps[i]; + struct dcn_hubp_state *s = &(TO_DCN10_HUBP(hubp)->state); + + hubp->funcs->hubp_read_state(hubp); + + if (!s->blank_en) + hubp->funcs->hubp_clear_underflow(hubp); + } +} + +void dcn10_clear_status_bits(struct dc *dc, unsigned int mask) +{ + /* + * Mask Format + * Bit 0 - 31: Status bit to clear + * + * Mask = 0x0 means clear all status bits + */ + const unsigned int DC_HW_STATE_MASK_HUBP_UNDERFLOW = 0x1; + const unsigned int DC_HW_STATE_MASK_OTPC_UNDERFLOW = 0x2; + + if (mask == 0x0) + mask = 0xFFFFFFFF; + + if (mask & DC_HW_STATE_MASK_HUBP_UNDERFLOW) + dcn10_clear_hubp_underflow(dc); + + if (mask & DC_HW_STATE_MASK_OTPC_UNDERFLOW) + dcn10_clear_otpc_underflow(dc); } void dcn10_get_hw_state(struct dc *dc, char *pBuf, unsigned int bufSize, unsigned int mask) @@ -491,16 +545,16 @@ void dcn10_get_hw_state(struct dc *dc, char *pBuf, unsigned int bufSize, unsigne * Bit 0 - 15: Hardware block mask * Bit 15: 1 = Invariant Only, 0 = All */ - const unsigned int DC_HW_STATE_MASK_HUBBUB = 0x1; - const unsigned int DC_HW_STATE_MASK_HUBP = 0x2; - const unsigned int DC_HW_STATE_MASK_RQ = 0x4; - const unsigned int DC_HW_STATE_MASK_DLG = 0x8; - const unsigned int DC_HW_STATE_MASK_TTU = 0x10; - const unsigned int DC_HW_STATE_MASK_CM = 0x20; - const unsigned int DC_HW_STATE_MASK_MPCC = 0x40; - const unsigned int DC_HW_STATE_MASK_OTG = 0x80; - const unsigned int DC_HW_STATE_MASK_CLOCKS = 0x100; - const unsigned int DC_HW_STATE_INVAR_ONLY = 0x8000; + const unsigned int DC_HW_STATE_MASK_HUBBUB = 0x1; + const unsigned int DC_HW_STATE_MASK_HUBP = 0x2; + const unsigned int DC_HW_STATE_MASK_RQ = 0x4; + const unsigned int DC_HW_STATE_MASK_DLG = 0x8; + const unsigned int DC_HW_STATE_MASK_TTU = 0x10; + const unsigned int DC_HW_STATE_MASK_CM = 0x20; + const unsigned int DC_HW_STATE_MASK_MPCC = 0x40; + const unsigned int DC_HW_STATE_MASK_OTG = 0x80; + const unsigned int DC_HW_STATE_MASK_CLOCKS = 0x100; + const unsigned int DC_HW_STATE_INVAR_ONLY = 0x8000; unsigned int chars_printed = 0; unsigned int remaining_buf_size = bufSize; @@ -556,6 +610,9 @@ void dcn10_get_hw_state(struct dc *dc, char *pBuf, unsigned int bufSize, unsigne remaining_buf_size -= chars_printed; } - if ((mask & DC_HW_STATE_MASK_CLOCKS) && remaining_buf_size > 0) + if ((mask & DC_HW_STATE_MASK_CLOCKS) && remaining_buf_size > 0) { chars_printed = dcn10_get_clock_states(dc, pBuf, remaining_buf_size); + pBuf += chars_printed; + remaining_buf_size -= chars_printed; + } } diff --git a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_link_encoder.c b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_link_encoder.c index ba6a8686062f..a9db372688ff 100644 --- a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_link_encoder.c +++ b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_link_encoder.c @@ -85,6 +85,7 @@ static const struct link_encoder_funcs dcn10_lnk_enc_funcs = { .enable_hpd = dcn10_link_encoder_enable_hpd, .disable_hpd = dcn10_link_encoder_disable_hpd, .is_dig_enabled = dcn10_is_dig_enabled, + .get_dig_frontend = dcn10_get_dig_frontend, .destroy = dcn10_link_encoder_destroy }; @@ -440,7 +441,7 @@ static uint8_t get_frontend_source( } } -void configure_encoder( +void enc1_configure_encoder( struct dcn10_link_encoder *enc10, const struct dc_link_settings *link_settings) { @@ -495,6 +496,15 @@ bool dcn10_is_dig_enabled(struct link_encoder *enc) return value; } +unsigned int dcn10_get_dig_frontend(struct link_encoder *enc) +{ + struct dcn10_link_encoder *enc10 = TO_DCN10_LINK_ENC(enc); + uint32_t value; + + REG_GET(DIG_BE_CNTL, DIG_FE_SOURCE_SELECT, &value); + return value; +} + static void link_encoder_disable(struct dcn10_link_encoder *enc10) { /* reset training pattern */ @@ -543,12 +553,12 @@ bool dcn10_link_encoder_validate_dvi_output( if ((connector_signal == SIGNAL_TYPE_DVI_SINGLE_LINK || connector_signal == SIGNAL_TYPE_HDMI_TYPE_A) && signal != SIGNAL_TYPE_HDMI_TYPE_A && - crtc_timing->pix_clk_khz > TMDS_MAX_PIXEL_CLOCK) + crtc_timing->pix_clk_100hz > (TMDS_MAX_PIXEL_CLOCK * 10)) return false; - if (crtc_timing->pix_clk_khz < TMDS_MIN_PIXEL_CLOCK) + if (crtc_timing->pix_clk_100hz < (TMDS_MIN_PIXEL_CLOCK * 10)) return false; - if (crtc_timing->pix_clk_khz > max_pixel_clock) + if (crtc_timing->pix_clk_100hz > (max_pixel_clock * 10)) return false; /* DVI supports 6/8bpp single-link and 10/16bpp dual-link */ @@ -571,7 +581,7 @@ bool dcn10_link_encoder_validate_dvi_output( static bool dcn10_link_encoder_validate_hdmi_output( const struct dcn10_link_encoder *enc10, const struct dc_crtc_timing *crtc_timing, - int adjusted_pix_clk_khz) + int adjusted_pix_clk_100hz) { enum dc_color_depth max_deep_color = enc10->base.features.max_hdmi_deep_color; @@ -581,20 +591,20 @@ static bool dcn10_link_encoder_validate_hdmi_output( if (crtc_timing->display_color_depth < COLOR_DEPTH_888) return false; - if (adjusted_pix_clk_khz < TMDS_MIN_PIXEL_CLOCK) + if (adjusted_pix_clk_100hz < (TMDS_MIN_PIXEL_CLOCK * 10)) return false; - if ((adjusted_pix_clk_khz == 0) || - (adjusted_pix_clk_khz > enc10->base.features.max_hdmi_pixel_clock)) + if ((adjusted_pix_clk_100hz == 0) || + (adjusted_pix_clk_100hz > (enc10->base.features.max_hdmi_pixel_clock * 10))) return false; /* DCE11 HW does not support 420 */ - if (!enc10->base.features.ycbcr420_supported && + if (!enc10->base.features.hdmi_ycbcr420_supported && crtc_timing->pixel_encoding == PIXEL_ENCODING_YCBCR420) return false; if (!enc10->base.features.flags.bits.HDMI_6GB_EN && - adjusted_pix_clk_khz >= 300000) + adjusted_pix_clk_100hz >= 3000000) return false; if (enc10->base.ctx->dc->debug.hdmi20_disable && crtc_timing->pixel_encoding == PIXEL_ENCODING_YCBCR420) @@ -606,8 +616,10 @@ bool dcn10_link_encoder_validate_dp_output( const struct dcn10_link_encoder *enc10, const struct dc_crtc_timing *crtc_timing) { - if (crtc_timing->pixel_encoding == PIXEL_ENCODING_YCBCR420) - return false; + if (crtc_timing->pixel_encoding == PIXEL_ENCODING_YCBCR420) { + if (!enc10->base.features.dp_ycbcr420_supported) + return false; + } return true; } @@ -736,7 +748,7 @@ bool dcn10_link_encoder_validate_output_with_stream( case SIGNAL_TYPE_DVI_DUAL_LINK: is_valid = dcn10_link_encoder_validate_dvi_output( enc10, - stream->sink->link->connector_signal, + stream->link->connector_signal, stream->signal, &stream->timing); break; @@ -744,7 +756,7 @@ bool dcn10_link_encoder_validate_output_with_stream( is_valid = dcn10_link_encoder_validate_hdmi_output( enc10, &stream->timing, - stream->phy_pix_clk); + stream->phy_pix_clk * 10); break; case SIGNAL_TYPE_DISPLAY_PORT: case SIGNAL_TYPE_DISPLAY_PORT_MST: @@ -908,7 +920,7 @@ void dcn10_link_encoder_enable_dp_output( * but it's not passed to asic_control. * We need to set number of lanes manually. */ - configure_encoder(enc10, link_settings); + enc1_configure_encoder(enc10, link_settings); cntl.action = TRANSMITTER_CONTROL_ENABLE; cntl.engine_id = enc->preferred_engine; @@ -947,7 +959,7 @@ void dcn10_link_encoder_enable_dp_mst_output( * but it's not passed to asic_control. * We need to set number of lanes manually. */ - configure_encoder(enc10, link_settings); + enc1_configure_encoder(enc10, link_settings); cntl.action = TRANSMITTER_CONTROL_ENABLE; cntl.engine_id = ENGINE_ID_UNKNOWN; diff --git a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_link_encoder.h b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_link_encoder.h index 49ead12b2532..b74b80a247ec 100644 --- a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_link_encoder.h +++ b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_link_encoder.h @@ -271,7 +271,7 @@ void dcn10_link_encoder_setup( struct link_encoder *enc, enum signal_type signal); -void configure_encoder( +void enc1_configure_encoder( struct dcn10_link_encoder *enc10, const struct dc_link_settings *link_settings); @@ -336,6 +336,8 @@ void dcn10_psr_program_secondary_packet(struct link_encoder *enc, bool dcn10_is_dig_enabled(struct link_encoder *enc); +unsigned int dcn10_get_dig_frontend(struct link_encoder *enc); + void dcn10_aux_initialize(struct dcn10_link_encoder *enc10); #endif /* __DC_LINK_ENCODER__DCN10_H__ */ diff --git a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_optc.c b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_optc.c index 54626682bab2..0345d51e9d6f 100644 --- a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_optc.c +++ b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_optc.c @@ -87,81 +87,41 @@ static void optc1_disable_stereo(struct timing_generator *optc) REG_SET(OTG_STEREO_CONTROL, 0, OTG_STEREO_EN, 0); - REG_SET_3(OTG_3D_STRUCTURE_CONTROL, 0, + REG_SET_2(OTG_3D_STRUCTURE_CONTROL, 0, OTG_3D_STRUCTURE_EN, 0, - OTG_3D_STRUCTURE_V_UPDATE_MODE, 0, OTG_3D_STRUCTURE_STEREO_SEL_OVR, 0); } -static uint32_t get_start_vline(struct timing_generator *optc, const struct dc_crtc_timing *dc_crtc_timing) +void optc1_setup_vertical_interrupt0( + struct timing_generator *optc, + uint32_t start_line, + uint32_t end_line) { - struct dc_crtc_timing patched_crtc_timing; - int vesa_sync_start; - int asic_blank_end; - int vertical_line_start; - - patched_crtc_timing = *dc_crtc_timing; - optc1_apply_front_porch_workaround(optc, &patched_crtc_timing); - - vesa_sync_start = patched_crtc_timing.h_addressable + - patched_crtc_timing.h_border_right + - patched_crtc_timing.h_front_porch; - - asic_blank_end = patched_crtc_timing.h_total - - vesa_sync_start - - patched_crtc_timing.h_border_left; - - vesa_sync_start = patched_crtc_timing.v_addressable + - patched_crtc_timing.v_border_bottom + - patched_crtc_timing.v_front_porch; - - asic_blank_end = (patched_crtc_timing.v_total - - vesa_sync_start - - patched_crtc_timing.v_border_top); - - vertical_line_start = asic_blank_end - optc->dlg_otg_param.vstartup_start + 1; - if (vertical_line_start < 0) { - ASSERT(0); - vertical_line_start = 0; - } + struct optc *optc1 = DCN10TG_FROM_TG(optc); - return vertical_line_start; + REG_SET_2(OTG_VERTICAL_INTERRUPT0_POSITION, 0, + OTG_VERTICAL_INTERRUPT0_LINE_START, start_line, + OTG_VERTICAL_INTERRUPT0_LINE_END, end_line); } -void optc1_program_vline_interrupt( +void optc1_setup_vertical_interrupt1( struct timing_generator *optc, - const struct dc_crtc_timing *dc_crtc_timing, - unsigned long long vsync_delta) + uint32_t start_line) { - struct optc *optc1 = DCN10TG_FROM_TG(optc); - unsigned long long req_delta_tens_of_usec = div64_u64((vsync_delta + 9999), 10000); - unsigned long long pix_clk_hundreds_khz = div64_u64((dc_crtc_timing->pix_clk_khz + 99), 100); - uint32_t req_delta_lines = (uint32_t) div64_u64( - (req_delta_tens_of_usec * pix_clk_hundreds_khz + dc_crtc_timing->h_total - 1), - dc_crtc_timing->h_total); - - uint32_t vsync_line = get_start_vline(optc, dc_crtc_timing); - uint32_t start_line = 0; - uint32_t endLine = 0; - - if (req_delta_lines != 0) - req_delta_lines--; - - if (req_delta_lines > vsync_line) - start_line = dc_crtc_timing->v_total - (req_delta_lines - vsync_line) + 2; - else - start_line = vsync_line - req_delta_lines; - - endLine = start_line + 2; + REG_SET(OTG_VERTICAL_INTERRUPT1_POSITION, 0, + OTG_VERTICAL_INTERRUPT1_LINE_START, start_line); +} - if (endLine >= dc_crtc_timing->v_total) - endLine = 2; +void optc1_setup_vertical_interrupt2( + struct timing_generator *optc, + uint32_t start_line) +{ + struct optc *optc1 = DCN10TG_FROM_TG(optc); - REG_SET_2(OTG_VERTICAL_INTERRUPT0_POSITION, 0, - OTG_VERTICAL_INTERRUPT0_LINE_START, start_line, - OTG_VERTICAL_INTERRUPT0_LINE_END, endLine); + REG_SET(OTG_VERTICAL_INTERRUPT2_POSITION, 0, + OTG_VERTICAL_INTERRUPT2_LINE_START, start_line); } /** @@ -266,21 +226,15 @@ void optc1_program_timing( patched_crtc_timing.v_addressable + patched_crtc_timing.v_border_bottom); + vertical_line_start = asic_blank_end - optc->dlg_otg_param.vstartup_start + 1; + v_fp2 = 0; + if (vertical_line_start < 0) + v_fp2 = -vertical_line_start; + REG_UPDATE_2(OTG_V_BLANK_START_END, OTG_V_BLANK_START, asic_blank_start, OTG_V_BLANK_END, asic_blank_end); - /* Use OTG_VERTICAL_INTERRUPT2 replace VUPDATE interrupt, - * program the reg for interrupt postition. - */ - vertical_line_start = asic_blank_end - optc->dlg_otg_param.vstartup_start + 1; - if (vertical_line_start < 0) { - ASSERT(0); - vertical_line_start = 0; - } - REG_SET(OTG_VERTICAL_INTERRUPT2_POSITION, 0, - OTG_VERTICAL_INTERRUPT2_LINE_START, vertical_line_start); - /* v_sync polarity */ v_sync_polarity = patched_crtc_timing.flags.VSYNC_POSITIVE_POLARITY ? 0 : 1; @@ -296,21 +250,19 @@ void optc1_program_timing( if (patched_crtc_timing.flags.INTERLACE == 1) field_num = 1; } - v_fp2 = 0; - if (optc->dlg_otg_param.vstartup_start > asic_blank_end) - v_fp2 = optc->dlg_otg_param.vstartup_start > asic_blank_end; /* Interlace */ - if (patched_crtc_timing.flags.INTERLACE == 1) { - REG_UPDATE(OTG_INTERLACE_CONTROL, - OTG_INTERLACE_ENABLE, 1); - v_init = v_init / 2; - if ((optc->dlg_otg_param.vstartup_start/2)*2 > asic_blank_end) - v_fp2 = v_fp2 / 2; - } else - REG_UPDATE(OTG_INTERLACE_CONTROL, - OTG_INTERLACE_ENABLE, 0); - + if (REG(OTG_INTERLACE_CONTROL)) { + if (patched_crtc_timing.flags.INTERLACE == 1) { + REG_UPDATE(OTG_INTERLACE_CONTROL, + OTG_INTERLACE_ENABLE, 1); + v_init = v_init / 2; + if ((optc->dlg_otg_param.vstartup_start/2)*2 > asic_blank_end) + v_fp2 = v_fp2 / 2; + } else + REG_UPDATE(OTG_INTERLACE_CONTROL, + OTG_INTERLACE_ENABLE, 0); + } /* VTG enable set to 0 first VInit */ REG_UPDATE(CONTROL, @@ -337,11 +289,10 @@ void optc1_program_timing( /* Enable stereo - only when we need to pack 3D frame. Other types * of stereo handled in explicit call */ - h_div_2 = (dc_crtc_timing->pixel_encoding == PIXEL_ENCODING_YCBCR420) ? - 1 : 0; + h_div_2 = optc1_is_two_pixels_per_containter(&patched_crtc_timing); REG_UPDATE(OTG_H_TIMING_CNTL, - OTG_H_TIMING_DIV_BY2, h_div_2); + OTG_H_TIMING_DIV_BY2, h_div_2 || optc1->comb_opp_id != 0xf); } @@ -362,20 +313,19 @@ void optc1_set_blank_data_double_buffer(struct timing_generator *optc, bool enab static void optc1_unblank_crtc(struct timing_generator *optc) { struct optc *optc1 = DCN10TG_FROM_TG(optc); - uint32_t vertical_interrupt_enable = 0; - - REG_GET(OTG_VERTICAL_INTERRUPT2_CONTROL, - OTG_VERTICAL_INTERRUPT2_INT_ENABLE, &vertical_interrupt_enable); - - /* temporary work around for vertical interrupt, once vertical interrupt enabled, - * this check will be removed. - */ - if (vertical_interrupt_enable) - optc1_set_blank_data_double_buffer(optc, true); REG_UPDATE_2(OTG_BLANK_CONTROL, OTG_BLANK_DATA_EN, 0, OTG_BLANK_DE_MODE, 0); + + /* W/A for automated testing + * Automated testing will fail underflow test as there + * sporadic underflows which occur during the optc blank + * sequence. As a w/a, clear underflow on unblank. + * This prevents the failure, but will not mask actual + * underflow that affect real use cases. + */ + optc1_clear_optc_underflow(optc); } /** @@ -1155,9 +1105,8 @@ static void optc1_enable_stereo(struct timing_generator *optc, OTG_DISABLE_STEREOSYNC_OUTPUT_FOR_DP, 1); if (flags->PROGRAM_STEREO) - REG_UPDATE_3(OTG_3D_STRUCTURE_CONTROL, + REG_UPDATE_2(OTG_3D_STRUCTURE_CONTROL, OTG_3D_STRUCTURE_EN, flags->FRAME_PACKED, - OTG_3D_STRUCTURE_V_UPDATE_MODE, flags->FRAME_PACKED, OTG_3D_STRUCTURE_STEREO_SEL_OVR, flags->FRAME_PACKED); } @@ -1189,6 +1138,64 @@ bool optc1_is_stereo_left_eye(struct timing_generator *optc) return ret; } +bool optc1_is_matching_timing(struct timing_generator *tg, + const struct dc_crtc_timing *otg_timing) +{ + struct dc_crtc_timing hw_crtc_timing = {0}; + struct dcn_otg_state s = {0}; + + if (tg == NULL || otg_timing == NULL) + return false; + + optc1_read_otg_state(DCN10TG_FROM_TG(tg), &s); + + hw_crtc_timing.h_total = s.h_total + 1; + hw_crtc_timing.h_addressable = s.h_total - ((s.h_total - s.h_blank_start) + s.h_blank_end); + hw_crtc_timing.h_front_porch = s.h_total + 1 - s.h_blank_start; + hw_crtc_timing.h_sync_width = s.h_sync_a_end - s.h_sync_a_start; + + hw_crtc_timing.v_total = s.v_total + 1; + hw_crtc_timing.v_addressable = s.v_total - ((s.v_total - s.v_blank_start) + s.v_blank_end); + hw_crtc_timing.v_front_porch = s.v_total + 1 - s.v_blank_start; + hw_crtc_timing.v_sync_width = s.v_sync_a_end - s.v_sync_a_start; + + if (otg_timing->h_total != hw_crtc_timing.h_total) + return false; + + if (otg_timing->h_border_left != hw_crtc_timing.h_border_left) + return false; + + if (otg_timing->h_addressable != hw_crtc_timing.h_addressable) + return false; + + if (otg_timing->h_border_right != hw_crtc_timing.h_border_right) + return false; + + if (otg_timing->h_front_porch != hw_crtc_timing.h_front_porch) + return false; + + if (otg_timing->h_sync_width != hw_crtc_timing.h_sync_width) + return false; + + if (otg_timing->v_total != hw_crtc_timing.v_total) + return false; + + if (otg_timing->v_border_top != hw_crtc_timing.v_border_top) + return false; + + if (otg_timing->v_addressable != hw_crtc_timing.v_addressable) + return false; + + if (otg_timing->v_border_bottom != hw_crtc_timing.v_border_bottom) + return false; + + if (otg_timing->v_sync_width != hw_crtc_timing.v_sync_width) + return false; + + return true; +} + + void optc1_read_otg_state(struct optc *optc1, struct dcn_otg_state *s) { @@ -1375,7 +1382,9 @@ bool optc1_get_crc(struct timing_generator *optc, static const struct timing_generator_funcs dcn10_tg_funcs = { .validate_timing = optc1_validate_timing, .program_timing = optc1_program_timing, - .program_vline_interrupt = optc1_program_vline_interrupt, + .setup_vertical_interrupt0 = optc1_setup_vertical_interrupt0, + .setup_vertical_interrupt1 = optc1_setup_vertical_interrupt1, + .setup_vertical_interrupt2 = optc1_setup_vertical_interrupt2, .program_global_sync = optc1_program_global_sync, .enable_crtc = optc1_enable_crtc, .disable_crtc = optc1_disable_crtc, @@ -1385,6 +1394,7 @@ static const struct timing_generator_funcs dcn10_tg_funcs = { .get_frame_count = optc1_get_vblank_counter, .get_scanoutpos = optc1_get_crtc_scanoutpos, .get_otg_active_size = optc1_get_otg_active_size, + .is_matching_timing = optc1_is_matching_timing, .set_early_control = optc1_set_early_control, /* used by enable_timing_synchronization. Not need for FPGA */ .wait_for_state = optc1_wait_for_state, @@ -1424,4 +1434,13 @@ void dcn10_timing_generator_init(struct optc *optc1) optc1->min_v_blank_interlace = 5; optc1->min_h_sync_width = 8; optc1->min_v_sync_width = 1; + optc1->comb_opp_id = 0xf; } + +bool optc1_is_two_pixels_per_containter(const struct dc_crtc_timing *timing) +{ + bool two_pix = timing->pixel_encoding == PIXEL_ENCODING_YCBCR420; + + return two_pix; +} + diff --git a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_optc.h b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_optc.h index c1b114209fe8..4eb9a898c237 100644 --- a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_optc.h +++ b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_optc.h @@ -67,6 +67,8 @@ SRI(OTG_CLOCK_CONTROL, OTG, inst),\ SRI(OTG_VERTICAL_INTERRUPT0_CONTROL, OTG, inst),\ SRI(OTG_VERTICAL_INTERRUPT0_POSITION, OTG, inst),\ + SRI(OTG_VERTICAL_INTERRUPT1_CONTROL, OTG, inst),\ + SRI(OTG_VERTICAL_INTERRUPT1_POSITION, OTG, inst),\ SRI(OTG_VERTICAL_INTERRUPT2_CONTROL, OTG, inst),\ SRI(OTG_VERTICAL_INTERRUPT2_POSITION, OTG, inst),\ SRI(OPTC_INPUT_CLOCK_CONTROL, ODM, inst),\ @@ -135,6 +137,8 @@ struct dcn_optc_registers { uint32_t OTG_CLOCK_CONTROL; uint32_t OTG_VERTICAL_INTERRUPT0_CONTROL; uint32_t OTG_VERTICAL_INTERRUPT0_POSITION; + uint32_t OTG_VERTICAL_INTERRUPT1_CONTROL; + uint32_t OTG_VERTICAL_INTERRUPT1_POSITION; uint32_t OTG_VERTICAL_INTERRUPT2_CONTROL; uint32_t OTG_VERTICAL_INTERRUPT2_POSITION; uint32_t OPTC_INPUT_CLOCK_CONTROL; @@ -227,6 +231,8 @@ struct dcn_optc_registers { SF(OTG0_OTG_VERTICAL_INTERRUPT0_CONTROL, OTG_VERTICAL_INTERRUPT0_INT_ENABLE, mask_sh),\ SF(OTG0_OTG_VERTICAL_INTERRUPT0_POSITION, OTG_VERTICAL_INTERRUPT0_LINE_START, mask_sh),\ SF(OTG0_OTG_VERTICAL_INTERRUPT0_POSITION, OTG_VERTICAL_INTERRUPT0_LINE_END, mask_sh),\ + SF(OTG0_OTG_VERTICAL_INTERRUPT1_CONTROL, OTG_VERTICAL_INTERRUPT1_INT_ENABLE, mask_sh),\ + SF(OTG0_OTG_VERTICAL_INTERRUPT1_POSITION, OTG_VERTICAL_INTERRUPT1_LINE_START, mask_sh),\ SF(OTG0_OTG_VERTICAL_INTERRUPT2_CONTROL, OTG_VERTICAL_INTERRUPT2_INT_ENABLE, mask_sh),\ SF(OTG0_OTG_VERTICAL_INTERRUPT2_POSITION, OTG_VERTICAL_INTERRUPT2_LINE_START, mask_sh),\ SF(ODM0_OPTC_INPUT_CLOCK_CONTROL, OPTC_INPUT_CLK_EN, mask_sh),\ @@ -361,6 +367,8 @@ struct dcn_optc_registers { type OTG_VERTICAL_INTERRUPT0_INT_ENABLE;\ type OTG_VERTICAL_INTERRUPT0_LINE_START;\ type OTG_VERTICAL_INTERRUPT0_LINE_END;\ + type OTG_VERTICAL_INTERRUPT1_INT_ENABLE;\ + type OTG_VERTICAL_INTERRUPT1_LINE_START;\ type OTG_VERTICAL_INTERRUPT2_INT_ENABLE;\ type OTG_VERTICAL_INTERRUPT2_LINE_START;\ type OPTC_INPUT_CLK_EN;\ @@ -427,7 +435,7 @@ struct optc { const struct dcn_optc_shift *tg_shift; const struct dcn_optc_mask *tg_mask; - enum controller_id controller_id; + int comb_opp_id; uint32_t max_h_total; uint32_t max_v_total; @@ -475,9 +483,16 @@ void optc1_program_timing( const struct dc_crtc_timing *dc_crtc_timing, bool use_vbios); -void optc1_program_vline_interrupt(struct timing_generator *optc, - const struct dc_crtc_timing *dc_crtc_timing, - unsigned long long vsync_delta); +void optc1_setup_vertical_interrupt0( + struct timing_generator *optc, + uint32_t start_line, + uint32_t end_line); +void optc1_setup_vertical_interrupt1( + struct timing_generator *optc, + uint32_t start_line); +void optc1_setup_vertical_interrupt2( + struct timing_generator *optc, + uint32_t start_line); void optc1_program_global_sync( struct timing_generator *optc); @@ -565,4 +580,6 @@ bool optc1_configure_crc(struct timing_generator *optc, bool optc1_get_crc(struct timing_generator *optc, uint32_t *r_cr, uint32_t *g_y, uint32_t *b_cb); +bool optc1_is_two_pixels_per_containter(const struct dc_crtc_timing *timing); + #endif /* __DC_TIMING_GENERATOR_DCN10_H__ */ diff --git a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_resource.c b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_resource.c index a71453a15ae3..09d74070a49b 100644 --- a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_resource.c +++ b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_resource.c @@ -28,23 +28,23 @@ #include "resource.h" #include "include/irq_service_interface.h" -#include "dcn10/dcn10_resource.h" +#include "dcn10_resource.h" -#include "dcn10/dcn10_ipp.h" -#include "dcn10/dcn10_mpc.h" +#include "dcn10_ipp.h" +#include "dcn10_mpc.h" #include "irq/dcn10/irq_service_dcn10.h" -#include "dcn10/dcn10_dpp.h" +#include "dcn10_dpp.h" #include "dcn10_optc.h" -#include "dcn10/dcn10_hw_sequencer.h" +#include "dcn10_hw_sequencer.h" #include "dce110/dce110_hw_sequencer.h" -#include "dcn10/dcn10_opp.h" -#include "dcn10/dcn10_link_encoder.h" -#include "dcn10/dcn10_stream_encoder.h" -#include "dce/dce_clocks.h" +#include "dcn10_opp.h" +#include "dcn10_link_encoder.h" +#include "dcn10_stream_encoder.h" +#include "dcn10_clk_mgr.h" #include "dce/dce_clock_source.h" #include "dce/dce_audio.h" #include "dce/dce_hwseq.h" -#include "../virtual/virtual_stream_encoder.h" +#include "virtual/virtual_stream_encoder.h" #include "dce110/dce110_resource.h" #include "dce112/dce112_resource.h" #include "dcn10_hubp.h" @@ -70,7 +70,7 @@ const struct _vcs_dpi_ip_params_st dcn1_0_ip = { .rob_buffer_size_kbytes = 64, .det_buffer_size_kbytes = 164, - .dpte_buffer_size_in_pte_reqs = 42, + .dpte_buffer_size_in_pte_reqs_luma = 42, .dpp_output_buffer_pixels = 2560, .opp_output_buffer_lines = 1, .pixel_chunk_size_kbytes = 8, @@ -202,7 +202,6 @@ enum dcn10_clk_src_array_id { #define MMHUB_SR(reg_name)\ .reg_name = MMHUB_BASE(mm ## reg_name ## _BASE_IDX) + \ mm ## reg_name - /* macros to expend register list macro defined in HW object header file * end *********************/ @@ -436,7 +435,6 @@ static const struct dcn_optc_mask tg_mask = { TG_COMMON_MASK_SH_LIST_DCN1_0(_MASK) }; - static const struct bios_registers bios_regs = { NBIO_SR(BIOS_SCRATCH_3), NBIO_SR(BIOS_SCRATCH_6) @@ -496,7 +494,6 @@ static const struct dce110_clk_src_mask cs_mask = { CS_COMMON_MASK_SH_LIST_DCN1_0(_MASK) }; - static const struct resource_caps res_cap = { .num_timing_generator = 4, .num_opp = 4, @@ -611,7 +608,7 @@ static struct output_pixel_processor *dcn10_opp_create( return &opp->base; } -struct aux_engine *dcn10_aux_engine_create( +struct dce_aux *dcn10_aux_engine_create( struct dc_context *ctx, uint32_t inst) { @@ -680,18 +677,18 @@ static struct mpc *dcn10_mpc_create(struct dc_context *ctx) static struct hubbub *dcn10_hubbub_create(struct dc_context *ctx) { - struct hubbub *hubbub = kzalloc(sizeof(struct hubbub), + struct dcn10_hubbub *dcn10_hubbub = kzalloc(sizeof(struct dcn10_hubbub), GFP_KERNEL); - if (!hubbub) + if (!dcn10_hubbub) return NULL; - hubbub1_construct(hubbub, ctx, + hubbub1_construct(&dcn10_hubbub->base, ctx, &hubbub_reg, &hubbub_shift, &hubbub_mask); - return hubbub; + return &dcn10_hubbub->base; } static struct timing_generator *dcn10_timing_generator_create( @@ -719,7 +716,8 @@ static struct timing_generator *dcn10_timing_generator_create( static const struct encoder_feature_support link_enc_feature = { .max_hdmi_deep_color = COLOR_DEPTH_121212, .max_hdmi_pixel_clock = 600000, - .ycbcr420_supported = true, + .hdmi_ycbcr420_supported = true, + .dp_ycbcr420_supported = false, .flags.bits.IS_HBR2_CAPABLE = true, .flags.bits.IS_HBR3_CAPABLE = true, .flags.bits.IS_TPS3_CAPABLE = true, @@ -912,7 +910,7 @@ static void destruct(struct dcn10_resource_pool *pool) for (i = 0; i < pool->base.res_cap->num_ddc; i++) { if (pool->base.engines[i] != NULL) - pool->base.engines[i]->funcs->destroy_engine(&pool->base.engines[i]); + dce110_engine_destroy(&pool->base.engines[i]); if (pool->base.hw_i2cs[i] != NULL) { kfree(pool->base.hw_i2cs[i]); pool->base.hw_i2cs[i] = NULL; @@ -949,8 +947,8 @@ static void destruct(struct dcn10_resource_pool *pool) if (pool->base.dmcu != NULL) dce_dmcu_destroy(&pool->base.dmcu); - if (pool->base.dccg != NULL) - dce_dccg_destroy(&pool->base.dccg); + if (pool->base.clk_mgr != NULL) + dce_clk_mgr_destroy(&pool->base.clk_mgr); kfree(pool->base.pp_smu); } @@ -975,8 +973,8 @@ static void get_pixel_clock_parameters( struct pixel_clk_params *pixel_clk_params) { const struct dc_stream_state *stream = pipe_ctx->stream; - pixel_clk_params->requested_pix_clk = stream->timing.pix_clk_khz; - pixel_clk_params->encoder_object_id = stream->sink->link->link_enc->id; + pixel_clk_params->requested_pix_clk_100hz = stream->timing.pix_clk_100hz; + pixel_clk_params->encoder_object_id = stream->link->link_enc->id; pixel_clk_params->signal_type = pipe_ctx->stream->signal; pixel_clk_params->controller_id = pipe_ctx->stream_res.tg->inst + 1; /* TODO: un-hardcode*/ @@ -992,9 +990,9 @@ static void get_pixel_clock_parameters( pixel_clk_params->color_depth = COLOR_DEPTH_888; if (stream->timing.pixel_encoding == PIXEL_ENCODING_YCBCR420) - pixel_clk_params->requested_pix_clk /= 2; + pixel_clk_params->requested_pix_clk_100hz /= 2; if (stream->timing.timing_3d_format == TIMING_3D_FORMAT_HW_FRAME_PACKING) - pixel_clk_params->requested_pix_clk *= 2; + pixel_clk_params->requested_pix_clk_100hz *= 2; } @@ -1132,6 +1130,56 @@ static enum dc_status dcn10_validate_plane(const struct dc_plane_state *plane_st return DC_OK; } +static enum dc_status dcn10_validate_global(struct dc *dc, struct dc_state *context) +{ + int i, j; + bool video_down_scaled = false; + bool video_large = false; + bool desktop_large = false; + bool dcc_disabled = false; + + for (i = 0; i < context->stream_count; i++) { + if (context->stream_status[i].plane_count == 0) + continue; + + if (context->stream_status[i].plane_count > 2) + return false; + + for (j = 0; j < context->stream_status[i].plane_count; j++) { + struct dc_plane_state *plane = + context->stream_status[i].plane_states[j]; + + + if (plane->format >= SURFACE_PIXEL_FORMAT_VIDEO_BEGIN) { + + if (plane->src_rect.width > plane->dst_rect.width || + plane->src_rect.height > plane->dst_rect.height) + video_down_scaled = true; + + if (plane->src_rect.width >= 3840) + video_large = true; + + } else { + if (plane->src_rect.width >= 3840) + desktop_large = true; + if (!plane->dcc.enable) + dcc_disabled = true; + } + } + } + + /* + * Workaround: On DCN10 there is UMC issue that causes underflow when + * playing 4k video on 4k desktop with video downscaled and single channel + * memory + */ + if (video_large && desktop_large && video_down_scaled && dcc_disabled && + dc->dcn_soc->number_of_channels == 1) + return DC_FAIL_SURFACE_VALIDATE; + + return DC_OK; +} + static enum dc_status dcn10_get_default_swizzle_mode(struct dc_plane_state *plane_state) { enum dc_status result = DC_OK; @@ -1160,6 +1208,7 @@ static const struct resource_funcs dcn10_res_pool_funcs = { .validate_bandwidth = dcn_validate_bandwidth, .acquire_idle_pipe_for_layer = dcn10_acquire_idle_pipe_for_layer, .validate_plane = dcn10_validate_plane, + .validate_global = dcn10_validate_global, .add_stream_to_ctx = dcn10_add_stream_to_ctx, .get_default_swizzle_mode = dcn10_get_default_swizzle_mode }; @@ -1275,9 +1324,8 @@ static bool construct( goto fail; } } - - pool->base.dccg = dcn1_dccg_create(ctx); - if (pool->base.dccg == NULL) { + pool->base.clk_mgr = dcn1_clk_mgr_create(ctx); + if (pool->base.clk_mgr == NULL) { dm_error("DC: failed to create display clock!\n"); BREAK_TO_DEBUGGER(); goto fail; diff --git a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_stream_encoder.c b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_stream_encoder.c index 6f9078f3c4d3..b08254121251 100644 --- a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_stream_encoder.c +++ b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_stream_encoder.c @@ -261,17 +261,29 @@ void enc1_stream_encoder_dp_set_stream_attribute( uint8_t dp_component_depth = 0; struct dcn10_stream_encoder *enc1 = DCN10STRENC_FROM_STRENC(enc); + struct dc_crtc_timing hw_crtc_timing = *crtc_timing; + + if (hw_crtc_timing.flags.INTERLACE) { + /*the input timing is in VESA spec format with Interlace flag =1*/ + hw_crtc_timing.v_total /= 2; + hw_crtc_timing.v_border_top /= 2; + hw_crtc_timing.v_addressable /= 2; + hw_crtc_timing.v_border_bottom /= 2; + hw_crtc_timing.v_front_porch /= 2; + hw_crtc_timing.v_sync_width /= 2; + } + /* set pixel encoding */ - switch (crtc_timing->pixel_encoding) { + switch (hw_crtc_timing.pixel_encoding) { case PIXEL_ENCODING_YCBCR422: dp_pixel_encoding = DP_PIXEL_ENCODING_TYPE_YCBCR422; break; case PIXEL_ENCODING_YCBCR444: dp_pixel_encoding = DP_PIXEL_ENCODING_TYPE_YCBCR444; - if (crtc_timing->flags.Y_ONLY) - if (crtc_timing->display_color_depth != COLOR_DEPTH_666) + if (hw_crtc_timing.flags.Y_ONLY) + if (hw_crtc_timing.display_color_depth != COLOR_DEPTH_666) /* HW testing only, no use case yet. * Color depth of Y-only could be * 8, 10, 12, 16 bits @@ -299,7 +311,7 @@ void enc1_stream_encoder_dp_set_stream_attribute( * Pixel Encoding/Colorimetry Format and that a Sink device shall ignore MISC1, bit 7, * and MISC0, bits 7:1 (MISC1, bit 7, and MISC0, bits 7:1, become "don't care"). */ - if ((crtc_timing->pixel_encoding == PIXEL_ENCODING_YCBCR420) || + if ((hw_crtc_timing.pixel_encoding == PIXEL_ENCODING_YCBCR420) || (output_color_space == COLOR_SPACE_2020_YCBCR) || (output_color_space == COLOR_SPACE_2020_RGB_FULLRANGE) || (output_color_space == COLOR_SPACE_2020_RGB_LIMITEDRANGE)) @@ -308,7 +320,7 @@ void enc1_stream_encoder_dp_set_stream_attribute( misc1 = misc1 & ~0x40; /* set color depth */ - switch (crtc_timing->display_color_depth) { + switch (hw_crtc_timing.display_color_depth) { case COLOR_DEPTH_666: dp_component_depth = DP_COMPONENT_PIXEL_DEPTH_6BPC; break; @@ -336,7 +348,7 @@ void enc1_stream_encoder_dp_set_stream_attribute( /* set dynamic range and YCbCr range */ - switch (crtc_timing->display_color_depth) { + switch (hw_crtc_timing.display_color_depth) { case COLOR_DEPTH_666: colorimetry_bpc = 0; break; @@ -372,9 +384,9 @@ void enc1_stream_encoder_dp_set_stream_attribute( misc0 = misc0 | 0x8; /* bit3=1, bit4=0 */ misc1 = misc1 & ~0x80; /* bit7 = 0*/ dynamic_range_ycbcr = 0; /*bt601*/ - if (crtc_timing->pixel_encoding == PIXEL_ENCODING_YCBCR422) + if (hw_crtc_timing.pixel_encoding == PIXEL_ENCODING_YCBCR422) misc0 = misc0 | 0x2; /* bit2=0, bit1=1 */ - else if (crtc_timing->pixel_encoding == PIXEL_ENCODING_YCBCR444) + else if (hw_crtc_timing.pixel_encoding == PIXEL_ENCODING_YCBCR444) misc0 = misc0 | 0x4; /* bit2=1, bit1=0 */ break; case COLOR_SPACE_YCBCR709: @@ -382,9 +394,9 @@ void enc1_stream_encoder_dp_set_stream_attribute( misc0 = misc0 | 0x18; /* bit3=1, bit4=1 */ misc1 = misc1 & ~0x80; /* bit7 = 0*/ dynamic_range_ycbcr = 1; /*bt709*/ - if (crtc_timing->pixel_encoding == PIXEL_ENCODING_YCBCR422) + if (hw_crtc_timing.pixel_encoding == PIXEL_ENCODING_YCBCR422) misc0 = misc0 | 0x2; /* bit2=0, bit1=1 */ - else if (crtc_timing->pixel_encoding == PIXEL_ENCODING_YCBCR444) + else if (hw_crtc_timing.pixel_encoding == PIXEL_ENCODING_YCBCR444) misc0 = misc0 | 0x4; /* bit2=1, bit1=0 */ break; case COLOR_SPACE_2020_RGB_LIMITEDRANGE: @@ -414,26 +426,26 @@ void enc1_stream_encoder_dp_set_stream_attribute( * dc_crtc_timing is vesa dmt struct. data from edid */ REG_SET_2(DP_MSA_TIMING_PARAM1, 0, - DP_MSA_HTOTAL, crtc_timing->h_total, - DP_MSA_VTOTAL, crtc_timing->v_total); + DP_MSA_HTOTAL, hw_crtc_timing.h_total, + DP_MSA_VTOTAL, hw_crtc_timing.v_total); /* calculate from vesa timing parameters * h_active_start related to leading edge of sync */ - h_blank = crtc_timing->h_total - crtc_timing->h_border_left - - crtc_timing->h_addressable - crtc_timing->h_border_right; + h_blank = hw_crtc_timing.h_total - hw_crtc_timing.h_border_left - + hw_crtc_timing.h_addressable - hw_crtc_timing.h_border_right; - h_back_porch = h_blank - crtc_timing->h_front_porch - - crtc_timing->h_sync_width; + h_back_porch = h_blank - hw_crtc_timing.h_front_porch - + hw_crtc_timing.h_sync_width; /* start at beginning of left border */ - h_active_start = crtc_timing->h_sync_width + h_back_porch; + h_active_start = hw_crtc_timing.h_sync_width + h_back_porch; - v_active_start = crtc_timing->v_total - crtc_timing->v_border_top - - crtc_timing->v_addressable - crtc_timing->v_border_bottom - - crtc_timing->v_front_porch; + v_active_start = hw_crtc_timing.v_total - hw_crtc_timing.v_border_top - + hw_crtc_timing.v_addressable - hw_crtc_timing.v_border_bottom - + hw_crtc_timing.v_front_porch; /* start at beginning of left border */ @@ -443,20 +455,20 @@ void enc1_stream_encoder_dp_set_stream_attribute( REG_SET_4(DP_MSA_TIMING_PARAM3, 0, DP_MSA_HSYNCWIDTH, - crtc_timing->h_sync_width, + hw_crtc_timing.h_sync_width, DP_MSA_HSYNCPOLARITY, - !crtc_timing->flags.HSYNC_POSITIVE_POLARITY, + !hw_crtc_timing.flags.HSYNC_POSITIVE_POLARITY, DP_MSA_VSYNCWIDTH, - crtc_timing->v_sync_width, + hw_crtc_timing.v_sync_width, DP_MSA_VSYNCPOLARITY, - !crtc_timing->flags.VSYNC_POSITIVE_POLARITY); + !hw_crtc_timing.flags.VSYNC_POSITIVE_POLARITY); /* HWDITH include border or overscan */ REG_SET_2(DP_MSA_TIMING_PARAM4, 0, - DP_MSA_HWIDTH, crtc_timing->h_border_left + - crtc_timing->h_addressable + crtc_timing->h_border_right, - DP_MSA_VHEIGHT, crtc_timing->v_border_top + - crtc_timing->v_addressable + crtc_timing->v_border_bottom); + DP_MSA_HWIDTH, hw_crtc_timing.h_border_left + + hw_crtc_timing.h_addressable + hw_crtc_timing.h_border_right, + DP_MSA_VHEIGHT, hw_crtc_timing.v_border_top + + hw_crtc_timing.v_addressable + hw_crtc_timing.v_border_bottom); } static void enc1_stream_encoder_set_stream_attribute_helper( @@ -594,7 +606,7 @@ void enc1_stream_encoder_dvi_set_stream_attribute( cntl.signal = is_dual_link ? SIGNAL_TYPE_DVI_DUAL_LINK : SIGNAL_TYPE_DVI_SINGLE_LINK; cntl.enable_dp_audio = false; - cntl.pixel_clock = crtc_timing->pix_clk_khz; + cntl.pixel_clock = crtc_timing->pix_clk_100hz / 10; cntl.lanes_number = (is_dual_link) ? LANE_COUNT_EIGHT : LANE_COUNT_FOUR; if (enc1->base.bp->funcs->encoder_control( @@ -766,7 +778,6 @@ void enc1_stream_encoder_dp_blank( struct stream_encoder *enc) { struct dcn10_stream_encoder *enc1 = DCN10STRENC_FROM_STRENC(enc); - uint32_t retries = 0; uint32_t reg1 = 0; uint32_t max_retries = DP_BLANK_MAX_RETRY * 10; @@ -803,8 +814,6 @@ void enc1_stream_encoder_dp_blank( 0, 10, max_retries); - ASSERT(retries <= max_retries); - /* Tell the DP encoder to ignore timing from CRTC, must be done after * the polling. If we set DP_STEER_FIFO_RESET before DP stream blank is * complete, stream status will be stuck in video stream enabled state, @@ -1416,6 +1425,14 @@ void enc1_setup_stereo_sync( REG_UPDATE(DIG_FE_CNTL, DIG_STEREOSYNC_GATE_EN, !enable); } +void enc1_dig_connect_to_otg( + struct stream_encoder *enc, + int tg_inst) +{ + struct dcn10_stream_encoder *enc1 = DCN10STRENC_FROM_STRENC(enc); + + REG_UPDATE(DIG_FE_CNTL, DIG_SOURCE_SELECT, tg_inst); +} static const struct stream_encoder_funcs dcn10_str_enc_funcs = { .dp_set_stream_attribute = @@ -1448,6 +1465,7 @@ static const struct stream_encoder_funcs dcn10_str_enc_funcs = { .hdmi_audio_disable = enc1_se_hdmi_audio_disable, .setup_stereo_sync = enc1_setup_stereo_sync, .set_avmute = enc1_stream_encoder_set_avmute, + .dig_connect_to_otg = enc1_dig_connect_to_otg, }; void dcn10_stream_encoder_construct( diff --git a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_stream_encoder.h b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_stream_encoder.h index 67f3e4dd95c1..b7c800e10a32 100644 --- a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_stream_encoder.h +++ b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_stream_encoder.h @@ -274,7 +274,8 @@ struct dcn10_stream_enc_registers { SE_SF(DP0_DP_MSA_TIMING_PARAM4, DP_MSA_HWIDTH, mask_sh),\ SE_SF(DP0_DP_MSA_TIMING_PARAM4, DP_MSA_VHEIGHT, mask_sh),\ SE_SF(DIG0_HDMI_DB_CONTROL, HDMI_DB_DISABLE, mask_sh),\ - SE_SF(DP0_DP_VID_TIMING, DP_VID_N_MUL, mask_sh) + SE_SF(DP0_DP_VID_TIMING, DP_VID_N_MUL, mask_sh),\ + SE_SF(DIG0_DIG_FE_CNTL, DIG_SOURCE_SELECT, mask_sh) #define SE_COMMON_MASK_SH_LIST_SOC(mask_sh)\ SE_COMMON_MASK_SH_LIST_SOC_BASE(mask_sh) @@ -426,7 +427,8 @@ struct dcn10_stream_enc_registers { type DP_MSA_VHEIGHT;\ type HDMI_DB_DISABLE;\ type DP_VID_N_MUL;\ - type DP_VID_M_DOUBLE_VALUE_EN + type DP_VID_M_DOUBLE_VALUE_EN;\ + type DIG_SOURCE_SELECT struct dcn10_stream_encoder_shift { SE_REG_FIELD_LIST_DCN1_0(uint8_t); @@ -523,4 +525,8 @@ void enc1_se_hdmi_audio_setup( void enc1_se_hdmi_audio_disable( struct stream_encoder *enc); +void enc1_dig_connect_to_otg( + struct stream_encoder *enc, + int tg_inst); + #endif /* __DC_STREAM_ENCODER_DCN10_H__ */ diff --git a/drivers/gpu/drm/amd/display/dc/dm_event_log.h b/drivers/gpu/drm/amd/display/dc/dm_event_log.h index 34a701ca879e..65663f4d93e1 100644 --- a/drivers/gpu/drm/amd/display/dc/dm_event_log.h +++ b/drivers/gpu/drm/amd/display/dc/dm_event_log.h @@ -33,6 +33,7 @@ #define EVENT_LOG_AUX_REQ(ddc, type, action, address, len, data) #define EVENT_LOG_AUX_REP(ddc, type, replyStatus, len, data) +#define EVENT_LOG_CUST_MSG(tag, a, ...) #endif diff --git a/drivers/gpu/drm/amd/display/dc/dm_helpers.h b/drivers/gpu/drm/amd/display/dc/dm_helpers.h index 5d4527d03045..e81b24374bcb 100644 --- a/drivers/gpu/drm/amd/display/dc/dm_helpers.h +++ b/drivers/gpu/drm/amd/display/dc/dm_helpers.h @@ -58,6 +58,13 @@ bool dm_helpers_dp_mst_write_payload_allocation_table( bool enable); /* + * poll pending down reply before clear payload allocation table + */ +void dm_helpers_dp_mst_poll_pending_down_reply( + struct dc_context *ctx, + const struct dc_link *link); + +/* * Clear payload allocation table before enable MST DP link. */ void dm_helpers_dp_mst_clear_payload_allocation_table( diff --git a/drivers/gpu/drm/amd/display/dc/dm_pp_smu.h b/drivers/gpu/drm/amd/display/dc/dm_pp_smu.h index f2ea8452d48f..14bed5b1fa97 100644 --- a/drivers/gpu/drm/amd/display/dc/dm_pp_smu.h +++ b/drivers/gpu/drm/amd/display/dc/dm_pp_smu.h @@ -38,7 +38,8 @@ enum pp_smu_ver { * of interface sharing between families of ASIcs. */ PP_SMU_UNSUPPORTED, - PP_SMU_VER_RV + PP_SMU_VER_RV, + PP_SMU_VER_MAX }; struct pp_smu { @@ -55,10 +56,10 @@ struct pp_smu { struct pp_smu_wm_set_range { unsigned int wm_inst; - uint32_t min_fill_clk_khz; - uint32_t max_fill_clk_khz; - uint32_t min_drain_clk_khz; - uint32_t max_drain_clk_khz; + uint32_t min_fill_clk_mhz; + uint32_t max_fill_clk_mhz; + uint32_t min_drain_clk_mhz; + uint32_t max_drain_clk_mhz; }; #define MAX_WATERMARK_SETS 4 @@ -77,15 +78,15 @@ struct pp_smu_display_requirement_rv { */ unsigned int display_count; - /* PPSMC_MSG_SetHardMinFclkByFreq: khz + /* PPSMC_MSG_SetHardMinFclkByFreq: mhz * FCLK will vary with DPM, but never below requested hard min */ - unsigned int hard_min_fclk_khz; + unsigned int hard_min_fclk_mhz; - /* PPSMC_MSG_SetHardMinDcefclkByFreq: khz + /* PPSMC_MSG_SetHardMinDcefclkByFreq: mhz * fixed clock at requested freq, either from FCH bypass or DFS */ - unsigned int hard_min_dcefclk_khz; + unsigned int hard_min_dcefclk_mhz; /* PPSMC_MSG_SetMinDeepSleepDcefclk: mhz * when DF is in cstate, dcf clock is further divided down @@ -102,14 +103,20 @@ struct pp_smu_funcs_rv { */ void (*set_display_count)(struct pp_smu *pp, int count); - /* which SMU message? are reader and writer WM separate SMU msg? */ + /* reader and writer WM's are sent together as part of one table*/ + /* + * PPSMC_MSG_SetDriverDramAddrHigh + * PPSMC_MSG_SetDriverDramAddrLow + * PPSMC_MSG_TransferTableDram2Smu + * + * */ void (*set_wm_ranges)(struct pp_smu *pp, struct pp_smu_wm_range_sets *ranges); /* PPSMC_MSG_SetHardMinDcfclkByFreq * fixed clock at requested freq, either from FCH bypass or DFS */ - void (*set_hard_min_dcfclk_by_freq)(struct pp_smu *pp, int khz); + void (*set_hard_min_dcfclk_by_freq)(struct pp_smu *pp, int mhz); /* PPSMC_MSG_SetMinDeepSleepDcfclk * when DF is in cstate, dcf clock is further divided down @@ -120,12 +127,12 @@ struct pp_smu_funcs_rv { /* PPSMC_MSG_SetHardMinFclkByFreq * FCLK will vary with DPM, but never below requested hard min */ - void (*set_hard_min_fclk_by_freq)(struct pp_smu *pp, int khz); + void (*set_hard_min_fclk_by_freq)(struct pp_smu *pp, int mhz); /* PPSMC_MSG_SetHardMinSocclkByFreq * Needed for DWB support */ - void (*set_hard_min_socclk_by_freq)(struct pp_smu *pp, int khz); + void (*set_hard_min_socclk_by_freq)(struct pp_smu *pp, int mhz); /* PME w/a */ void (*set_pme_wa_enable)(struct pp_smu *pp); diff --git a/drivers/gpu/drm/amd/display/dc/dm_services.h b/drivers/gpu/drm/amd/display/dc/dm_services.h index 28128c02de00..1961cc6d9143 100644 --- a/drivers/gpu/drm/amd/display/dc/dm_services.h +++ b/drivers/gpu/drm/amd/display/dc/dm_services.h @@ -31,6 +31,8 @@ #define __DM_SERVICES_H__ +#include "amdgpu_dm_trace.h" + /* TODO: remove when DC is complete. */ #include "dm_services_types.h" #include "logger_interface.h" @@ -70,6 +72,7 @@ static inline uint32_t dm_read_reg_func( } #endif value = cgs_read_register(ctx->cgs_device, address); + trace_amdgpu_dc_rreg(&ctx->perf_trace->read_count, address, value); return value; } @@ -90,6 +93,7 @@ static inline void dm_write_reg_func( } #endif cgs_write_register(ctx->cgs_device, address, value); + trace_amdgpu_dc_wreg(&ctx->perf_trace->write_count, address, value); } static inline uint32_t dm_read_index_reg( @@ -351,8 +355,12 @@ unsigned long long dm_get_elapse_time_in_ns(struct dc_context *ctx, /* * performance tracing */ -void dm_perf_trace_timestamp(const char *func_name, unsigned int line); -#define PERF_TRACE() dm_perf_trace_timestamp(__func__, __LINE__) +#define PERF_TRACE() trace_amdgpu_dc_performance(CTX->perf_trace->read_count,\ + CTX->perf_trace->write_count, &CTX->perf_trace->last_entry_read,\ + &CTX->perf_trace->last_entry_write, __func__, __LINE__) +#define PERF_TRACE_CTX(__CTX) trace_amdgpu_dc_performance(__CTX->perf_trace->read_count,\ + __CTX->perf_trace->write_count, &__CTX->perf_trace->last_entry_read,\ + &__CTX->perf_trace->last_entry_write, __func__, __LINE__) /* diff --git a/drivers/gpu/drm/amd/display/dc/dm_services_types.h b/drivers/gpu/drm/amd/display/dc/dm_services_types.h index 2b83f922ac02..77200711abbe 100644 --- a/drivers/gpu/drm/amd/display/dc/dm_services_types.h +++ b/drivers/gpu/drm/amd/display/dc/dm_services_types.h @@ -82,9 +82,17 @@ enum dm_pp_clock_type { #define DC_DECODE_PP_CLOCK_TYPE(clk_type) \ (clk_type) == DM_PP_CLOCK_TYPE_DISPLAY_CLK ? "Display" : \ (clk_type) == DM_PP_CLOCK_TYPE_ENGINE_CLK ? "Engine" : \ - (clk_type) == DM_PP_CLOCK_TYPE_MEMORY_CLK ? "Memory" : "Invalid" - -#define DM_PP_MAX_CLOCK_LEVELS 8 + (clk_type) == DM_PP_CLOCK_TYPE_MEMORY_CLK ? "Memory" : \ + (clk_type) == DM_PP_CLOCK_TYPE_DCFCLK ? "DCF" : \ + (clk_type) == DM_PP_CLOCK_TYPE_DCEFCLK ? "DCEF" : \ + (clk_type) == DM_PP_CLOCK_TYPE_SOCCLK ? "SoC" : \ + (clk_type) == DM_PP_CLOCK_TYPE_PIXELCLK ? "Pixel" : \ + (clk_type) == DM_PP_CLOCK_TYPE_DISPLAYPHYCLK ? "Display PHY" : \ + (clk_type) == DM_PP_CLOCK_TYPE_DPPCLK ? "DPP" : \ + (clk_type) == DM_PP_CLOCK_TYPE_FCLK ? "F" : \ + "Invalid" + +#define DM_PP_MAX_CLOCK_LEVELS 16 struct dm_pp_clock_levels { uint32_t num_levels; @@ -208,22 +216,20 @@ struct dm_bl_data_point { /* Brightness level as effective value in range 0-255, * corresponding to above percentage */ - uint8_t signalLevel; + uint8_t signal_level; }; /* Total size of the structure should not exceed 256 bytes */ struct dm_acpi_atif_backlight_caps { - - uint16_t size; /* Bytes 0-1 (2 bytes) */ uint16_t flags; /* Byted 2-3 (2 bytes) */ - uint8_t errorCode; /* Byte 4 */ - uint8_t acLevelPercentage; /* Byte 5 */ - uint8_t dcLevelPercentage; /* Byte 6 */ - uint8_t minInputSignal; /* Byte 7 */ - uint8_t maxInputSignal; /* Byte 8 */ - uint8_t numOfDataPoints; /* Byte 9 */ - struct dm_bl_data_point dataPoints[99]; /* Bytes 10-207 (198 bytes)*/ + uint8_t error_code; /* Byte 4 */ + uint8_t ac_level_percentage; /* Byte 5 */ + uint8_t dc_level_percentage; /* Byte 6 */ + uint8_t min_input_signal; /* Byte 7 */ + uint8_t max_input_signal; /* Byte 8 */ + uint8_t num_data_points; /* Byte 9 */ + struct dm_bl_data_point data_points[99]; /* Bytes 10-207 (198 bytes)*/ }; enum dm_acpi_display_type { diff --git a/drivers/gpu/drm/amd/display/dc/dml/display_mode_enums.h b/drivers/gpu/drm/amd/display/dc/dml/display_mode_enums.h index bea4e61b94c7..c59e582c1f40 100644 --- a/drivers/gpu/drm/amd/display/dc/dml/display_mode_enums.h +++ b/drivers/gpu/drm/amd/display/dc/dml/display_mode_enums.h @@ -121,4 +121,30 @@ enum self_refresh_affinity { dm_neither_self_refresh_nor_mclk_switch }; +enum dm_validation_status { + DML_VALIDATION_OK, + DML_FAIL_SCALE_RATIO_TAP, + DML_FAIL_SOURCE_PIXEL_FORMAT, + DML_FAIL_VIEWPORT_SIZE, + DML_FAIL_TOTAL_V_ACTIVE_BW, + DML_FAIL_DIO_SUPPORT, + DML_FAIL_NOT_ENOUGH_DSC, + DML_FAIL_DSC_CLK_REQUIRED, + DML_FAIL_URGENT_LATENCY, + DML_FAIL_REORDERING_BUFFER, + DML_FAIL_DISPCLK_DPPCLK, + DML_FAIL_TOTAL_AVAILABLE_PIPES, + DML_FAIL_NUM_OTG, + DML_FAIL_WRITEBACK_MODE, + DML_FAIL_WRITEBACK_LATENCY, + DML_FAIL_WRITEBACK_SCALE_RATIO_TAP, + DML_FAIL_CURSOR_SUPPORT, + DML_FAIL_PITCH_SUPPORT, + DML_FAIL_PTE_BUFFER_SIZE, + DML_FAIL_HOST_VM_IMMEDIATE_FLIP, + DML_FAIL_DSC_INPUT_BPC, + DML_FAIL_PREFETCH_SUPPORT, + DML_FAIL_V_RATIO_PREFETCH, +}; + #endif diff --git a/drivers/gpu/drm/amd/display/dc/dml/display_mode_lib.c b/drivers/gpu/drm/amd/display/dc/dml/display_mode_lib.c index dddeb0d4db8f..d303b789adfe 100644 --- a/drivers/gpu/drm/amd/display/dc/dml/display_mode_lib.c +++ b/drivers/gpu/drm/amd/display/dc/dml/display_mode_lib.c @@ -62,3 +62,31 @@ void dml_init_instance(struct display_mode_lib *lib, enum dml_project project) } } +const char *dml_get_status_message(enum dm_validation_status status) +{ + switch (status) { + case DML_VALIDATION_OK: return "Validation OK"; + case DML_FAIL_SCALE_RATIO_TAP: return "Scale ratio/tap"; + case DML_FAIL_SOURCE_PIXEL_FORMAT: return "Source pixel format"; + case DML_FAIL_VIEWPORT_SIZE: return "Viewport size"; + case DML_FAIL_TOTAL_V_ACTIVE_BW: return "Total vertical active bandwidth"; + case DML_FAIL_DIO_SUPPORT: return "DIO support"; + case DML_FAIL_NOT_ENOUGH_DSC: return "Not enough DSC Units"; + case DML_FAIL_DSC_CLK_REQUIRED: return "DSC clock required"; + case DML_FAIL_URGENT_LATENCY: return "Urgent latency"; + case DML_FAIL_REORDERING_BUFFER: return "Re-ordering buffer"; + case DML_FAIL_DISPCLK_DPPCLK: return "Dispclk and Dppclk"; + case DML_FAIL_TOTAL_AVAILABLE_PIPES: return "Total available pipes"; + case DML_FAIL_NUM_OTG: return "Number of OTG"; + case DML_FAIL_WRITEBACK_MODE: return "Writeback mode"; + case DML_FAIL_WRITEBACK_LATENCY: return "Writeback latency"; + case DML_FAIL_WRITEBACK_SCALE_RATIO_TAP: return "Writeback scale ratio/tap"; + case DML_FAIL_CURSOR_SUPPORT: return "Cursor support"; + case DML_FAIL_PITCH_SUPPORT: return "Pitch support"; + case DML_FAIL_PTE_BUFFER_SIZE: return "PTE buffer size"; + case DML_FAIL_DSC_INPUT_BPC: return "DSC input bpc"; + case DML_FAIL_PREFETCH_SUPPORT: return "Prefetch support"; + case DML_FAIL_V_RATIO_PREFETCH: return "Vertical ratio prefetch"; + default: return "Unknown Status"; + } +} diff --git a/drivers/gpu/drm/amd/display/dc/dml/display_mode_lib.h b/drivers/gpu/drm/amd/display/dc/dml/display_mode_lib.h index 635206248889..a730e0209c05 100644 --- a/drivers/gpu/drm/amd/display/dc/dml/display_mode_lib.h +++ b/drivers/gpu/drm/amd/display/dc/dml/display_mode_lib.h @@ -43,4 +43,6 @@ struct display_mode_lib { void dml_init_instance(struct display_mode_lib *lib, enum dml_project project); +const char *dml_get_status_message(enum dm_validation_status status); + #endif diff --git a/drivers/gpu/drm/amd/display/dc/dml/display_mode_structs.h b/drivers/gpu/drm/amd/display/dc/dml/display_mode_structs.h index cbafce649e33..391183e3428f 100644 --- a/drivers/gpu/drm/amd/display/dc/dml/display_mode_structs.h +++ b/drivers/gpu/drm/amd/display/dc/dml/display_mode_structs.h @@ -30,22 +30,15 @@ typedef struct _vcs_dpi_soc_bounding_box_st soc_bounding_box_st; typedef struct _vcs_dpi_ip_params_st ip_params_st; typedef struct _vcs_dpi_display_pipe_source_params_st display_pipe_source_params_st; typedef struct _vcs_dpi_display_output_params_st display_output_params_st; -typedef struct _vcs_dpi_display_bandwidth_st display_bandwidth_st; typedef struct _vcs_dpi_scaler_ratio_depth_st scaler_ratio_depth_st; typedef struct _vcs_dpi_scaler_taps_st scaler_taps_st; typedef struct _vcs_dpi_display_pipe_dest_params_st display_pipe_dest_params_st; typedef struct _vcs_dpi_display_pipe_params_st display_pipe_params_st; typedef struct _vcs_dpi_display_clocks_and_cfg_st display_clocks_and_cfg_st; typedef struct _vcs_dpi_display_e2e_pipe_params_st display_e2e_pipe_params_st; -typedef struct _vcs_dpi_dchub_buffer_sizing_st dchub_buffer_sizing_st; -typedef struct _vcs_dpi_watermarks_perf_st watermarks_perf_st; -typedef struct _vcs_dpi_cstate_pstate_watermarks_st cstate_pstate_watermarks_st; -typedef struct _vcs_dpi_wm_calc_pipe_params_st wm_calc_pipe_params_st; -typedef struct _vcs_dpi_vratio_pre_st vratio_pre_st; typedef struct _vcs_dpi_display_data_rq_misc_params_st display_data_rq_misc_params_st; typedef struct _vcs_dpi_display_data_rq_sizing_params_st display_data_rq_sizing_params_st; typedef struct _vcs_dpi_display_data_rq_dlg_params_st display_data_rq_dlg_params_st; -typedef struct _vcs_dpi_display_cur_rq_dlg_params_st display_cur_rq_dlg_params_st; typedef struct _vcs_dpi_display_rq_dlg_params_st display_rq_dlg_params_st; typedef struct _vcs_dpi_display_rq_sizing_params_st display_rq_sizing_params_st; typedef struct _vcs_dpi_display_rq_misc_params_st display_rq_misc_params_st; @@ -55,8 +48,6 @@ typedef struct _vcs_dpi_display_ttu_regs_st display_ttu_regs_st; typedef struct _vcs_dpi_display_data_rq_regs_st display_data_rq_regs_st; typedef struct _vcs_dpi_display_rq_regs_st display_rq_regs_st; typedef struct _vcs_dpi_display_dlg_sys_params_st display_dlg_sys_params_st; -typedef struct _vcs_dpi_display_dlg_prefetch_param_st display_dlg_prefetch_param_st; -typedef struct _vcs_dpi_display_pipe_clock_st display_pipe_clock_st; typedef struct _vcs_dpi_display_arb_params_st display_arb_params_st; struct _vcs_dpi_voltage_scaling_st { @@ -111,9 +102,8 @@ struct _vcs_dpi_soc_bounding_box_st { double xfc_bus_transport_time_us; double xfc_xbuf_latency_tolerance_us; int use_urgent_burst_bw; - double max_hscl_ratio; - double max_vscl_ratio; - struct _vcs_dpi_voltage_scaling_st clock_limits[7]; + unsigned int num_states; + struct _vcs_dpi_voltage_scaling_st clock_limits[8]; }; struct _vcs_dpi_ip_params_st { @@ -128,7 +118,8 @@ struct _vcs_dpi_ip_params_st { unsigned int odm_capable; unsigned int rob_buffer_size_kbytes; unsigned int det_buffer_size_kbytes; - unsigned int dpte_buffer_size_in_pte_reqs; + unsigned int dpte_buffer_size_in_pte_reqs_luma; + unsigned int dpte_buffer_size_in_pte_reqs_chroma; unsigned int pde_proc_buffer_size_64k_reqs; unsigned int dpp_output_buffer_pixels; unsigned int opp_output_buffer_lines; @@ -191,7 +182,6 @@ struct _vcs_dpi_display_xfc_params_st { struct _vcs_dpi_display_pipe_source_params_st { int source_format; unsigned char dcc; - unsigned int dcc_override; unsigned int dcc_rate; unsigned char dcc_use_global; unsigned char vm; @@ -204,7 +194,6 @@ struct _vcs_dpi_display_pipe_source_params_st { int source_scan; int sw_mode; int macro_tile_size; - unsigned char is_display_sw; unsigned int viewport_width; unsigned int viewport_height; unsigned int viewport_y_y; @@ -251,16 +240,10 @@ struct _vcs_dpi_display_output_params_st { int output_bpc; int output_type; int output_format; - int output_standard; int dsc_slices; struct writeback_st wb; }; -struct _vcs_dpi_display_bandwidth_st { - double total_bw_consumed_gbps; - double guaranteed_urgent_return_bw_gbps; -}; - struct _vcs_dpi_scaler_ratio_depth_st { double hscl_ratio; double vscl_ratio; @@ -299,11 +282,9 @@ struct _vcs_dpi_display_pipe_dest_params_st { unsigned int vupdate_width; unsigned int vready_offset; unsigned char interlaced; - unsigned char underscan; double pixel_rate_mhz; unsigned char synchronized_vblank_all_planes; unsigned char otg_inst; - unsigned char odm_split_cnt; unsigned char odm_combine; unsigned char use_maximum_vstartup; }; @@ -330,65 +311,6 @@ struct _vcs_dpi_display_e2e_pipe_params_st { display_clocks_and_cfg_st clks_cfg; }; -struct _vcs_dpi_dchub_buffer_sizing_st { - unsigned int swath_width_y; - unsigned int swath_height_y; - unsigned int swath_height_c; - unsigned int detail_buffer_size_y; -}; - -struct _vcs_dpi_watermarks_perf_st { - double stutter_eff_in_active_region_percent; - double urgent_latency_supported_us; - double non_urgent_latency_supported_us; - double dram_clock_change_margin_us; - double dram_access_eff_percent; -}; - -struct _vcs_dpi_cstate_pstate_watermarks_st { - double cstate_exit_us; - double cstate_enter_plus_exit_us; - double pstate_change_us; -}; - -struct _vcs_dpi_wm_calc_pipe_params_st { - unsigned int num_dpp; - int voltage; - int output_type; - double dcfclk_mhz; - double socclk_mhz; - double dppclk_mhz; - double pixclk_mhz; - unsigned char interlace_en; - unsigned char pte_enable; - unsigned char dcc_enable; - double dcc_rate; - double bytes_per_pixel_c; - double bytes_per_pixel_y; - unsigned int swath_width_y; - unsigned int swath_height_y; - unsigned int swath_height_c; - unsigned int det_buffer_size_y; - double h_ratio; - double v_ratio; - unsigned int h_taps; - unsigned int h_total; - unsigned int v_total; - unsigned int v_active; - unsigned int e2e_index; - double display_pipe_line_delivery_time; - double read_bw; - unsigned int lines_in_det_y; - unsigned int lines_in_det_y_rounded_down_to_swath; - double full_det_buffering_time; - double dcfclk_deepsleep_mhz_per_plane; -}; - -struct _vcs_dpi_vratio_pre_st { - double vratio_pre_l; - double vratio_pre_c; -}; - struct _vcs_dpi_display_data_rq_misc_params_st { unsigned int full_swath_bytes; unsigned int stored_swath_bytes; @@ -422,16 +344,9 @@ struct _vcs_dpi_display_data_rq_dlg_params_st { unsigned int meta_bytes_per_row_ub; }; -struct _vcs_dpi_display_cur_rq_dlg_params_st { - unsigned char enable; - unsigned int swath_height; - unsigned int req_per_line; -}; - struct _vcs_dpi_display_rq_dlg_params_st { display_data_rq_dlg_params_st rq_l; display_data_rq_dlg_params_st rq_c; - display_cur_rq_dlg_params_st rq_cur0; }; struct _vcs_dpi_display_rq_sizing_params_st { @@ -497,6 +412,10 @@ struct _vcs_dpi_display_dlg_regs_st { unsigned int xfc_reg_remote_surface_flip_latency; unsigned int xfc_reg_prefetch_margin; unsigned int dst_y_delta_drq_limit; + unsigned int refcyc_per_vm_group_vblank; + unsigned int refcyc_per_vm_group_flip; + unsigned int refcyc_per_vm_req_vblank; + unsigned int refcyc_per_vm_req_flip; }; struct _vcs_dpi_display_ttu_regs_st { @@ -555,19 +474,6 @@ struct _vcs_dpi_display_dlg_sys_params_st { unsigned int total_flip_bytes; }; -struct _vcs_dpi_display_dlg_prefetch_param_st { - double prefetch_bw; - unsigned int flip_bytes; -}; - -struct _vcs_dpi_display_pipe_clock_st { - double dcfclk_mhz; - double dispclk_mhz; - double socclk_mhz; - double dscclk_mhz[6]; - double dppclk_mhz[6]; -}; - struct _vcs_dpi_display_arb_params_st { int max_req_outstanding; int min_req_outstanding; diff --git a/drivers/gpu/drm/amd/display/dc/dml/dml1_display_rq_dlg_calc.c b/drivers/gpu/drm/amd/display/dc/dml/dml1_display_rq_dlg_calc.c index c2037daa8e66..ad8571f5a142 100644 --- a/drivers/gpu/drm/amd/display/dc/dml/dml1_display_rq_dlg_calc.c +++ b/drivers/gpu/drm/amd/display/dc/dml/dml1_display_rq_dlg_calc.c @@ -459,7 +459,7 @@ static void dml1_rq_dlg_get_row_heights( /* dpte */ /* ------ */ log2_vmpg_bytes = dml_log2(mode_lib->soc.vmm_page_size_bytes); - dpte_buf_in_pte_reqs = mode_lib->ip.dpte_buffer_size_in_pte_reqs; + dpte_buf_in_pte_reqs = mode_lib->ip.dpte_buffer_size_in_pte_reqs_luma; log2_vmpg_height = 0; log2_vmpg_width = 0; @@ -776,7 +776,7 @@ static void get_surf_rq_param( /* dpte */ /* ------ */ log2_vmpg_bytes = dml_log2(mode_lib->soc.vmm_page_size_bytes); - dpte_buf_in_pte_reqs = mode_lib->ip.dpte_buffer_size_in_pte_reqs; + dpte_buf_in_pte_reqs = mode_lib->ip.dpte_buffer_size_in_pte_reqs_luma; log2_vmpg_height = 0; log2_vmpg_width = 0; @@ -881,7 +881,7 @@ static void get_surf_rq_param( /* the dpte_group_bytes is reduced for the specific case of vertical * access of a tile surface that has dpte request of 8x1 ptes. */ - if (!surf_linear & (log2_dpte_req_height_ptes == 0) & surf_vert) /*reduced, in this case, will have page fault within a group */ + if (!surf_linear && (log2_dpte_req_height_ptes == 0) && surf_vert) /*reduced, in this case, will have page fault within a group */ rq_sizing_param->dpte_group_bytes = 512; else /*full size */ diff --git a/drivers/gpu/drm/amd/display/dc/gpio/gpio_base.c b/drivers/gpu/drm/amd/display/dc/gpio/gpio_base.c index 1d1efd72b291..cf76ea2d9f5a 100644 --- a/drivers/gpu/drm/amd/display/dc/gpio/gpio_base.c +++ b/drivers/gpu/drm/amd/display/dc/gpio/gpio_base.c @@ -101,6 +101,18 @@ enum gpio_mode dal_gpio_get_mode( return gpio->mode; } +enum gpio_result dal_gpio_lock_pin( + struct gpio *gpio) +{ + return dal_gpio_service_lock(gpio->service, gpio->id, gpio->en); +} + +enum gpio_result dal_gpio_unlock_pin( + struct gpio *gpio) +{ + return dal_gpio_service_unlock(gpio->service, gpio->id, gpio->en); +} + enum gpio_result dal_gpio_change_mode( struct gpio *gpio, enum gpio_mode mode) diff --git a/drivers/gpu/drm/amd/display/dc/gpio/gpio_service.c b/drivers/gpu/drm/amd/display/dc/gpio/gpio_service.c index f20161c5706d..3c63a3c04dbb 100644 --- a/drivers/gpu/drm/amd/display/dc/gpio/gpio_service.c +++ b/drivers/gpu/drm/amd/display/dc/gpio/gpio_service.c @@ -56,7 +56,6 @@ struct gpio_service *dal_gpio_service_create( struct dc_context *ctx) { struct gpio_service *service; - uint32_t index_of_id; service = kzalloc(sizeof(struct gpio_service), GFP_KERNEL); @@ -78,44 +77,33 @@ struct gpio_service *dal_gpio_service_create( goto failure_1; } - /* allocate and initialize business storage */ + /* allocate and initialize busyness storage */ { - const uint32_t bits_per_uint = sizeof(uint32_t) << 3; - index_of_id = 0; service->ctx = ctx; do { uint32_t number_of_bits = service->factory.number_of_pins[index_of_id]; + uint32_t i = 0; - uint32_t number_of_uints = - (number_of_bits + bits_per_uint - 1) / - bits_per_uint; - - uint32_t *slot; - - if (number_of_bits) { - uint32_t index_of_uint = 0; - - slot = kcalloc(number_of_uints, - sizeof(uint32_t), - GFP_KERNEL); + if (number_of_bits) { + service->busyness[index_of_id] = + kcalloc(number_of_bits, sizeof(char), + GFP_KERNEL); - if (!slot) { + if (!service->busyness[index_of_id]) { BREAK_TO_DEBUGGER(); goto failure_2; } do { - slot[index_of_uint] = 0; - - ++index_of_uint; - } while (index_of_uint < number_of_uints); - } else - slot = NULL; - - service->busyness[index_of_id] = slot; + service->busyness[index_of_id][i] = 0; + ++i; + } while (i < number_of_bits); + } else { + service->busyness[index_of_id] = NULL; + } ++index_of_id; } while (index_of_id < GPIO_ID_COUNT); @@ -125,13 +113,8 @@ struct gpio_service *dal_gpio_service_create( failure_2: while (index_of_id) { - uint32_t *slot; - --index_of_id; - - slot = service->busyness[index_of_id]; - - kfree(slot); + kfree(service->busyness[index_of_id]); } failure_1: @@ -169,9 +152,7 @@ void dal_gpio_service_destroy( uint32_t index_of_id = 0; do { - uint32_t *slot = (*ptr)->busyness[index_of_id]; - - kfree(slot); + kfree((*ptr)->busyness[index_of_id]); ++index_of_id; } while (index_of_id < GPIO_ID_COUNT); @@ -192,33 +173,51 @@ static bool is_pin_busy( enum gpio_id id, uint32_t en) { - const uint32_t bits_per_uint = sizeof(uint32_t) << 3; + return service->busyness[id][en]; +} - const uint32_t *slot = service->busyness[id] + (en / bits_per_uint); +static void set_pin_busy( + struct gpio_service *service, + enum gpio_id id, + uint32_t en) +{ + service->busyness[id][en] = true; +} - return 0 != (*slot & (1 << (en % bits_per_uint))); +static void set_pin_free( + struct gpio_service *service, + enum gpio_id id, + uint32_t en) +{ + service->busyness[id][en] = false; } -static void set_pin_busy( +enum gpio_result dal_gpio_service_lock( struct gpio_service *service, enum gpio_id id, uint32_t en) { - const uint32_t bits_per_uint = sizeof(uint32_t) << 3; + if (!service->busyness[id]) { + ASSERT_CRITICAL(false); + return GPIO_RESULT_OPEN_FAILED; + } - service->busyness[id][en / bits_per_uint] |= - (1 << (en % bits_per_uint)); + set_pin_busy(service, id, en); + return GPIO_RESULT_OK; } -static void set_pin_free( +enum gpio_result dal_gpio_service_unlock( struct gpio_service *service, enum gpio_id id, uint32_t en) { - const uint32_t bits_per_uint = sizeof(uint32_t) << 3; + if (!service->busyness[id]) { + ASSERT_CRITICAL(false); + return GPIO_RESULT_OPEN_FAILED; + } - service->busyness[id][en / bits_per_uint] &= - ~(1 << (en % bits_per_uint)); + set_pin_free(service, id, en); + return GPIO_RESULT_OK; } enum gpio_result dal_gpio_service_open( diff --git a/drivers/gpu/drm/amd/display/dc/gpio/gpio_service.h b/drivers/gpu/drm/amd/display/dc/gpio/gpio_service.h index c7f3081f59cc..0c678af75331 100644 --- a/drivers/gpu/drm/amd/display/dc/gpio/gpio_service.h +++ b/drivers/gpu/drm/amd/display/dc/gpio/gpio_service.h @@ -36,10 +36,9 @@ struct gpio_service { /* * @brief * Business storage. - * For each member of 'enum gpio_id', - * store array of bits (packed into uint32_t slots), - * index individual bit by 'en' value */ - uint32_t *busyness[GPIO_ID_COUNT]; + * one byte For each member of 'enum gpio_id' + */ + char *busyness[GPIO_ID_COUNT]; }; enum gpio_result dal_gpio_service_open( @@ -53,4 +52,14 @@ void dal_gpio_service_close( struct gpio_service *service, struct hw_gpio_pin **ptr); +enum gpio_result dal_gpio_service_lock( + struct gpio_service *service, + enum gpio_id id, + uint32_t en); + +enum gpio_result dal_gpio_service_unlock( + struct gpio_service *service, + enum gpio_id id, + uint32_t en); + #endif diff --git a/drivers/gpu/drm/amd/display/dc/gpio/hw_factory.c b/drivers/gpu/drm/amd/display/dc/gpio/hw_factory.c index a683f4102e65..c2028c4744a6 100644 --- a/drivers/gpu/drm/amd/display/dc/gpio/hw_factory.c +++ b/drivers/gpu/drm/amd/display/dc/gpio/hw_factory.c @@ -79,6 +79,7 @@ bool dal_hw_factory_init( dal_hw_factory_dce110_init(factory); return true; case DCE_VERSION_12_0: + case DCE_VERSION_12_1: dal_hw_factory_dce120_init(factory); return true; #if defined(CONFIG_DRM_AMD_DC_DCN1_0) diff --git a/drivers/gpu/drm/amd/display/dc/gpio/hw_translate.c b/drivers/gpu/drm/amd/display/dc/gpio/hw_translate.c index 096f45628630..236ca28784a9 100644 --- a/drivers/gpu/drm/amd/display/dc/gpio/hw_translate.c +++ b/drivers/gpu/drm/amd/display/dc/gpio/hw_translate.c @@ -76,6 +76,7 @@ bool dal_hw_translate_init( dal_hw_translate_dce110_init(translate); return true; case DCE_VERSION_12_0: + case DCE_VERSION_12_1: dal_hw_translate_dce120_init(translate); return true; #if defined(CONFIG_DRM_AMD_DC_DCN1_0) diff --git a/drivers/gpu/drm/amd/display/dc/i2caux/Makefile b/drivers/gpu/drm/amd/display/dc/i2caux/Makefile deleted file mode 100644 index 352885cb4d07..000000000000 --- a/drivers/gpu/drm/amd/display/dc/i2caux/Makefile +++ /dev/null @@ -1,99 +0,0 @@ -# -# Copyright 2017 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. -# -# -# Makefile for the 'i2c' sub-component of DAL. -# It provides the control and status of HW i2c engine of the adapter. - -I2CAUX = aux_engine.o engine_base.o i2caux.o i2c_engine.o \ - i2c_generic_hw_engine.o i2c_hw_engine.o i2c_sw_engine.o - -AMD_DAL_I2CAUX = $(addprefix $(AMDDALPATH)/dc/i2caux/,$(I2CAUX)) - -AMD_DISPLAY_FILES += $(AMD_DAL_I2CAUX) - -############################################################################### -# DCE 8x family -############################################################################### -I2CAUX_DCE80 = i2caux_dce80.o i2c_hw_engine_dce80.o \ - i2c_sw_engine_dce80.o - -AMD_DAL_I2CAUX_DCE80 = $(addprefix $(AMDDALPATH)/dc/i2caux/dce80/,$(I2CAUX_DCE80)) - -AMD_DISPLAY_FILES += $(AMD_DAL_I2CAUX_DCE80) - -############################################################################### -# DCE 100 family -############################################################################### -I2CAUX_DCE100 = i2caux_dce100.o - -AMD_DAL_I2CAUX_DCE100 = $(addprefix $(AMDDALPATH)/dc/i2caux/dce100/,$(I2CAUX_DCE100)) - -AMD_DISPLAY_FILES += $(AMD_DAL_I2CAUX_DCE100) - -############################################################################### -# DCE 110 family -############################################################################### -I2CAUX_DCE110 = i2caux_dce110.o i2c_sw_engine_dce110.o i2c_hw_engine_dce110.o \ - aux_engine_dce110.o - -AMD_DAL_I2CAUX_DCE110 = $(addprefix $(AMDDALPATH)/dc/i2caux/dce110/,$(I2CAUX_DCE110)) - -AMD_DISPLAY_FILES += $(AMD_DAL_I2CAUX_DCE110) - -############################################################################### -# DCE 112 family -############################################################################### -I2CAUX_DCE112 = i2caux_dce112.o - -AMD_DAL_I2CAUX_DCE112 = $(addprefix $(AMDDALPATH)/dc/i2caux/dce112/,$(I2CAUX_DCE112)) - -AMD_DISPLAY_FILES += $(AMD_DAL_I2CAUX_DCE112) - -############################################################################### -# DCN 1.0 family -############################################################################### -ifdef CONFIG_DRM_AMD_DC_DCN1_0 -I2CAUX_DCN1 = i2caux_dcn10.o - -AMD_DAL_I2CAUX_DCN1 = $(addprefix $(AMDDALPATH)/dc/i2caux/dcn10/,$(I2CAUX_DCN1)) - -AMD_DISPLAY_FILES += $(AMD_DAL_I2CAUX_DCN1) -endif - -############################################################################### -# DCE 120 family -############################################################################### -I2CAUX_DCE120 = i2caux_dce120.o - -AMD_DAL_I2CAUX_DCE120 = $(addprefix $(AMDDALPATH)/dc/i2caux/dce120/,$(I2CAUX_DCE120)) - -AMD_DISPLAY_FILES += $(AMD_DAL_I2CAUX_DCE120) - -############################################################################### -# Diagnostics on FPGA -############################################################################### -I2CAUX_DIAG = i2caux_diag.o - -AMD_DAL_I2CAUX_DIAG = $(addprefix $(AMDDALPATH)/dc/i2caux/diagnostics/,$(I2CAUX_DIAG)) - -AMD_DISPLAY_FILES += $(AMD_DAL_I2CAUX_DIAG) - diff --git a/drivers/gpu/drm/amd/display/dc/i2caux/aux_engine.c b/drivers/gpu/drm/amd/display/dc/i2caux/aux_engine.c deleted file mode 100644 index 8cbf38b2470d..000000000000 --- a/drivers/gpu/drm/amd/display/dc/i2caux/aux_engine.c +++ /dev/null @@ -1,606 +0,0 @@ -/* - * Copyright 2012-15 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. - * - * Authors: AMD - * - */ - -#include "dm_services.h" -#include "dm_event_log.h" - -/* - * Pre-requisites: headers required by header of this unit - */ -#include "include/i2caux_interface.h" -#include "engine.h" - -/* - * Header of this unit - */ - -#include "aux_engine.h" - -/* - * Post-requisites: headers required by this unit - */ - -#include "include/link_service_types.h" - -/* - * This unit - */ - -enum { - AUX_INVALID_REPLY_RETRY_COUNTER = 1, - AUX_TIMED_OUT_RETRY_COUNTER = 2, - AUX_DEFER_RETRY_COUNTER = 6 -}; - -#define FROM_ENGINE(ptr) \ - container_of((ptr), struct aux_engine, base) -#define DC_LOGGER \ - engine->base.ctx->logger - -enum i2caux_engine_type dal_aux_engine_get_engine_type( - const struct engine *engine) -{ - return I2CAUX_ENGINE_TYPE_AUX; -} - -bool dal_aux_engine_acquire( - struct engine *engine, - struct ddc *ddc) -{ - struct aux_engine *aux_engine = FROM_ENGINE(engine); - - enum gpio_result result; - if (aux_engine->funcs->is_engine_available) { - /*check whether SW could use the engine*/ - if (!aux_engine->funcs->is_engine_available(aux_engine)) { - return false; - } - } - - result = dal_ddc_open(ddc, GPIO_MODE_HARDWARE, - GPIO_DDC_CONFIG_TYPE_MODE_AUX); - - if (result != GPIO_RESULT_OK) - return false; - - if (!aux_engine->funcs->acquire_engine(aux_engine)) { - dal_ddc_close(ddc); - return false; - } - - engine->ddc = ddc; - - return true; -} - -struct read_command_context { - uint8_t *buffer; - uint32_t current_read_length; - uint32_t offset; - enum i2caux_transaction_status status; - - struct aux_request_transaction_data request; - struct aux_reply_transaction_data reply; - - uint8_t returned_byte; - - uint32_t timed_out_retry_aux; - uint32_t invalid_reply_retry_aux; - uint32_t defer_retry_aux; - uint32_t defer_retry_i2c; - uint32_t invalid_reply_retry_aux_on_ack; - - bool transaction_complete; - bool operation_succeeded; -}; - -static void process_read_reply( - struct aux_engine *engine, - struct read_command_context *ctx) -{ - engine->funcs->process_channel_reply(engine, &ctx->reply); - - switch (ctx->reply.status) { - case AUX_TRANSACTION_REPLY_AUX_ACK: - ctx->defer_retry_aux = 0; - if (ctx->returned_byte > ctx->current_read_length) { - ctx->status = - I2CAUX_TRANSACTION_STATUS_FAILED_PROTOCOL_ERROR; - ctx->operation_succeeded = false; - } else if (ctx->returned_byte < ctx->current_read_length) { - ctx->current_read_length -= ctx->returned_byte; - - ctx->offset += ctx->returned_byte; - - ++ctx->invalid_reply_retry_aux_on_ack; - - if (ctx->invalid_reply_retry_aux_on_ack > - AUX_INVALID_REPLY_RETRY_COUNTER) { - ctx->status = - I2CAUX_TRANSACTION_STATUS_FAILED_PROTOCOL_ERROR; - ctx->operation_succeeded = false; - } - } else { - ctx->status = I2CAUX_TRANSACTION_STATUS_SUCCEEDED; - ctx->transaction_complete = true; - ctx->operation_succeeded = true; - } - break; - case AUX_TRANSACTION_REPLY_AUX_NACK: - ctx->status = I2CAUX_TRANSACTION_STATUS_FAILED_NACK; - ctx->operation_succeeded = false; - break; - case AUX_TRANSACTION_REPLY_AUX_DEFER: - ++ctx->defer_retry_aux; - - if (ctx->defer_retry_aux > AUX_DEFER_RETRY_COUNTER) { - ctx->status = I2CAUX_TRANSACTION_STATUS_FAILED_TIMEOUT; - ctx->operation_succeeded = false; - } - break; - case AUX_TRANSACTION_REPLY_I2C_DEFER: - ctx->defer_retry_aux = 0; - - ++ctx->defer_retry_i2c; - - if (ctx->defer_retry_i2c > AUX_DEFER_RETRY_COUNTER) { - ctx->status = I2CAUX_TRANSACTION_STATUS_FAILED_TIMEOUT; - ctx->operation_succeeded = false; - } - break; - case AUX_TRANSACTION_REPLY_HPD_DISCON: - ctx->status = I2CAUX_TRANSACTION_STATUS_FAILED_HPD_DISCON; - ctx->operation_succeeded = false; - break; - default: - ctx->status = I2CAUX_TRANSACTION_STATUS_UNKNOWN; - ctx->operation_succeeded = false; - } -} - -static void process_read_request( - struct aux_engine *engine, - struct read_command_context *ctx) -{ - enum aux_channel_operation_result operation_result; - - engine->funcs->submit_channel_request(engine, &ctx->request); - - operation_result = engine->funcs->get_channel_status( - engine, &ctx->returned_byte); - - switch (operation_result) { - case AUX_CHANNEL_OPERATION_SUCCEEDED: - if (ctx->returned_byte > ctx->current_read_length) { - ctx->status = - I2CAUX_TRANSACTION_STATUS_FAILED_PROTOCOL_ERROR; - ctx->operation_succeeded = false; - } else { - ctx->timed_out_retry_aux = 0; - ctx->invalid_reply_retry_aux = 0; - - ctx->reply.length = ctx->returned_byte; - ctx->reply.data = ctx->buffer; - - process_read_reply(engine, ctx); - } - break; - case AUX_CHANNEL_OPERATION_FAILED_INVALID_REPLY: - ++ctx->invalid_reply_retry_aux; - - if (ctx->invalid_reply_retry_aux > - AUX_INVALID_REPLY_RETRY_COUNTER) { - ctx->status = - I2CAUX_TRANSACTION_STATUS_FAILED_PROTOCOL_ERROR; - ctx->operation_succeeded = false; - } else - udelay(400); - break; - case AUX_CHANNEL_OPERATION_FAILED_TIMEOUT: - ++ctx->timed_out_retry_aux; - - if (ctx->timed_out_retry_aux > AUX_TIMED_OUT_RETRY_COUNTER) { - ctx->status = I2CAUX_TRANSACTION_STATUS_FAILED_TIMEOUT; - ctx->operation_succeeded = false; - } else { - /* DP 1.2a, table 2-58: - * "S3: AUX Request CMD PENDING: - * retry 3 times, with 400usec wait on each" - * The HW timeout is set to 550usec, - * so we should not wait here */ - } - break; - case AUX_CHANNEL_OPERATION_FAILED_HPD_DISCON: - ctx->status = I2CAUX_TRANSACTION_STATUS_FAILED_HPD_DISCON; - ctx->operation_succeeded = false; - break; - default: - ctx->status = I2CAUX_TRANSACTION_STATUS_UNKNOWN; - ctx->operation_succeeded = false; - } -} - -static bool read_command( - struct aux_engine *engine, - struct i2caux_transaction_request *request, - bool middle_of_transaction) -{ - struct read_command_context ctx; - - ctx.buffer = request->payload.data; - ctx.current_read_length = request->payload.length; - ctx.offset = 0; - ctx.timed_out_retry_aux = 0; - ctx.invalid_reply_retry_aux = 0; - ctx.defer_retry_aux = 0; - ctx.defer_retry_i2c = 0; - ctx.invalid_reply_retry_aux_on_ack = 0; - ctx.transaction_complete = false; - ctx.operation_succeeded = true; - - if (request->payload.address_space == - I2CAUX_TRANSACTION_ADDRESS_SPACE_DPCD) { - ctx.request.type = AUX_TRANSACTION_TYPE_DP; - ctx.request.action = I2CAUX_TRANSACTION_ACTION_DP_READ; - ctx.request.address = request->payload.address; - } else if (request->payload.address_space == - I2CAUX_TRANSACTION_ADDRESS_SPACE_I2C) { - ctx.request.type = AUX_TRANSACTION_TYPE_I2C; - ctx.request.action = middle_of_transaction ? - I2CAUX_TRANSACTION_ACTION_I2C_READ_MOT : - I2CAUX_TRANSACTION_ACTION_I2C_READ; - ctx.request.address = request->payload.address >> 1; - } else { - /* in DAL2, there was no return in such case */ - BREAK_TO_DEBUGGER(); - return false; - } - - ctx.request.delay = 0; - - do { - memset(ctx.buffer + ctx.offset, 0, ctx.current_read_length); - - ctx.request.data = ctx.buffer + ctx.offset; - ctx.request.length = ctx.current_read_length; - - process_read_request(engine, &ctx); - - request->status = ctx.status; - - if (ctx.operation_succeeded && !ctx.transaction_complete) - if (ctx.request.type == AUX_TRANSACTION_TYPE_I2C) - msleep(engine->delay); - } while (ctx.operation_succeeded && !ctx.transaction_complete); - - if (request->payload.address_space == - I2CAUX_TRANSACTION_ADDRESS_SPACE_DPCD) { - DC_LOG_I2C_AUX("READ: addr:0x%x value:0x%x Result:%d", - request->payload.address, - request->payload.data[0], - ctx.operation_succeeded); - } - - return ctx.operation_succeeded; -} - -struct write_command_context { - bool mot; - - uint8_t *buffer; - uint32_t current_write_length; - enum i2caux_transaction_status status; - - struct aux_request_transaction_data request; - struct aux_reply_transaction_data reply; - - uint8_t returned_byte; - - uint32_t timed_out_retry_aux; - uint32_t invalid_reply_retry_aux; - uint32_t defer_retry_aux; - uint32_t defer_retry_i2c; - uint32_t max_defer_retry; - uint32_t ack_m_retry; - - uint8_t reply_data[DEFAULT_AUX_MAX_DATA_SIZE]; - - bool transaction_complete; - bool operation_succeeded; -}; - -static void process_write_reply( - struct aux_engine *engine, - struct write_command_context *ctx) -{ - engine->funcs->process_channel_reply(engine, &ctx->reply); - - switch (ctx->reply.status) { - case AUX_TRANSACTION_REPLY_AUX_ACK: - ctx->operation_succeeded = true; - - if (ctx->returned_byte) { - ctx->request.action = ctx->mot ? - I2CAUX_TRANSACTION_ACTION_I2C_STATUS_REQUEST_MOT : - I2CAUX_TRANSACTION_ACTION_I2C_STATUS_REQUEST; - - ctx->current_write_length = 0; - - ++ctx->ack_m_retry; - - if (ctx->ack_m_retry > AUX_DEFER_RETRY_COUNTER) { - ctx->status = - I2CAUX_TRANSACTION_STATUS_FAILED_TIMEOUT; - ctx->operation_succeeded = false; - } else - udelay(300); - } else { - ctx->status = I2CAUX_TRANSACTION_STATUS_SUCCEEDED; - ctx->defer_retry_aux = 0; - ctx->ack_m_retry = 0; - ctx->transaction_complete = true; - } - break; - case AUX_TRANSACTION_REPLY_AUX_NACK: - ctx->status = I2CAUX_TRANSACTION_STATUS_FAILED_NACK; - ctx->operation_succeeded = false; - break; - case AUX_TRANSACTION_REPLY_AUX_DEFER: - ++ctx->defer_retry_aux; - - if (ctx->defer_retry_aux > ctx->max_defer_retry) { - ctx->status = I2CAUX_TRANSACTION_STATUS_FAILED_TIMEOUT; - ctx->operation_succeeded = false; - } - break; - case AUX_TRANSACTION_REPLY_I2C_DEFER: - ctx->defer_retry_aux = 0; - ctx->current_write_length = 0; - - ctx->request.action = ctx->mot ? - I2CAUX_TRANSACTION_ACTION_I2C_STATUS_REQUEST_MOT : - I2CAUX_TRANSACTION_ACTION_I2C_STATUS_REQUEST; - - ++ctx->defer_retry_i2c; - - if (ctx->defer_retry_i2c > ctx->max_defer_retry) { - ctx->status = I2CAUX_TRANSACTION_STATUS_FAILED_TIMEOUT; - ctx->operation_succeeded = false; - } - break; - case AUX_TRANSACTION_REPLY_HPD_DISCON: - ctx->status = I2CAUX_TRANSACTION_STATUS_FAILED_HPD_DISCON; - ctx->operation_succeeded = false; - break; - default: - ctx->status = I2CAUX_TRANSACTION_STATUS_UNKNOWN; - ctx->operation_succeeded = false; - } -} - -static void process_write_request( - struct aux_engine *engine, - struct write_command_context *ctx) -{ - enum aux_channel_operation_result operation_result; - - engine->funcs->submit_channel_request(engine, &ctx->request); - - operation_result = engine->funcs->get_channel_status( - engine, &ctx->returned_byte); - - switch (operation_result) { - case AUX_CHANNEL_OPERATION_SUCCEEDED: - ctx->timed_out_retry_aux = 0; - ctx->invalid_reply_retry_aux = 0; - - ctx->reply.length = ctx->returned_byte; - ctx->reply.data = ctx->reply_data; - - process_write_reply(engine, ctx); - break; - case AUX_CHANNEL_OPERATION_FAILED_INVALID_REPLY: - ++ctx->invalid_reply_retry_aux; - - if (ctx->invalid_reply_retry_aux > - AUX_INVALID_REPLY_RETRY_COUNTER) { - ctx->status = - I2CAUX_TRANSACTION_STATUS_FAILED_PROTOCOL_ERROR; - ctx->operation_succeeded = false; - } else - udelay(400); - break; - case AUX_CHANNEL_OPERATION_FAILED_TIMEOUT: - ++ctx->timed_out_retry_aux; - - if (ctx->timed_out_retry_aux > AUX_TIMED_OUT_RETRY_COUNTER) { - ctx->status = I2CAUX_TRANSACTION_STATUS_FAILED_TIMEOUT; - ctx->operation_succeeded = false; - } else { - /* DP 1.2a, table 2-58: - * "S3: AUX Request CMD PENDING: - * retry 3 times, with 400usec wait on each" - * The HW timeout is set to 550usec, - * so we should not wait here */ - } - break; - case AUX_CHANNEL_OPERATION_FAILED_HPD_DISCON: - ctx->status = I2CAUX_TRANSACTION_STATUS_FAILED_HPD_DISCON; - ctx->operation_succeeded = false; - break; - default: - ctx->status = I2CAUX_TRANSACTION_STATUS_UNKNOWN; - ctx->operation_succeeded = false; - } -} - -static bool write_command( - struct aux_engine *engine, - struct i2caux_transaction_request *request, - bool middle_of_transaction) -{ - struct write_command_context ctx; - - ctx.mot = middle_of_transaction; - ctx.buffer = request->payload.data; - ctx.current_write_length = request->payload.length; - ctx.timed_out_retry_aux = 0; - ctx.invalid_reply_retry_aux = 0; - ctx.defer_retry_aux = 0; - ctx.defer_retry_i2c = 0; - ctx.ack_m_retry = 0; - ctx.transaction_complete = false; - ctx.operation_succeeded = true; - - if (request->payload.address_space == - I2CAUX_TRANSACTION_ADDRESS_SPACE_DPCD) { - ctx.request.type = AUX_TRANSACTION_TYPE_DP; - ctx.request.action = I2CAUX_TRANSACTION_ACTION_DP_WRITE; - ctx.request.address = request->payload.address; - } else if (request->payload.address_space == - I2CAUX_TRANSACTION_ADDRESS_SPACE_I2C) { - ctx.request.type = AUX_TRANSACTION_TYPE_I2C; - ctx.request.action = middle_of_transaction ? - I2CAUX_TRANSACTION_ACTION_I2C_WRITE_MOT : - I2CAUX_TRANSACTION_ACTION_I2C_WRITE; - ctx.request.address = request->payload.address >> 1; - } else { - /* in DAL2, there was no return in such case */ - BREAK_TO_DEBUGGER(); - return false; - } - - ctx.request.delay = 0; - - ctx.max_defer_retry = - (engine->max_defer_write_retry > AUX_DEFER_RETRY_COUNTER) ? - engine->max_defer_write_retry : AUX_DEFER_RETRY_COUNTER; - - do { - ctx.request.data = ctx.buffer; - ctx.request.length = ctx.current_write_length; - - process_write_request(engine, &ctx); - - request->status = ctx.status; - - if (ctx.operation_succeeded && !ctx.transaction_complete) - if (ctx.request.type == AUX_TRANSACTION_TYPE_I2C) - msleep(engine->delay); - } while (ctx.operation_succeeded && !ctx.transaction_complete); - - if (request->payload.address_space == - I2CAUX_TRANSACTION_ADDRESS_SPACE_DPCD) { - DC_LOG_I2C_AUX("WRITE: addr:0x%x value:0x%x Result:%d", - request->payload.address, - request->payload.data[0], - ctx.operation_succeeded); - } - - return ctx.operation_succeeded; -} - -static bool end_of_transaction_command( - struct aux_engine *engine, - struct i2caux_transaction_request *request) -{ - struct i2caux_transaction_request dummy_request; - uint8_t dummy_data; - - /* [tcheng] We only need to send the stop (read with MOT = 0) - * for I2C-over-Aux, not native AUX */ - - if (request->payload.address_space != - I2CAUX_TRANSACTION_ADDRESS_SPACE_I2C) - return false; - - dummy_request.operation = request->operation; - dummy_request.payload.address_space = request->payload.address_space; - dummy_request.payload.address = request->payload.address; - - /* - * Add a dummy byte due to some receiver quirk - * where one byte is sent along with MOT = 0. - * Ideally this should be 0. - */ - - dummy_request.payload.length = 0; - dummy_request.payload.data = &dummy_data; - - if (request->operation == I2CAUX_TRANSACTION_READ) - return read_command(engine, &dummy_request, false); - else - return write_command(engine, &dummy_request, false); - - /* according Syed, it does not need now DoDummyMOT */ -} - -bool dal_aux_engine_submit_request( - struct engine *engine, - struct i2caux_transaction_request *request, - bool middle_of_transaction) -{ - struct aux_engine *aux_engine = FROM_ENGINE(engine); - - bool result; - bool mot_used = true; - - switch (request->operation) { - case I2CAUX_TRANSACTION_READ: - result = read_command(aux_engine, request, mot_used); - break; - case I2CAUX_TRANSACTION_WRITE: - result = write_command(aux_engine, request, mot_used); - break; - default: - result = false; - } - - /* [tcheng] - * need to send stop for the last transaction to free up the AUX - * if the above command fails, this would be the last transaction */ - - if (!middle_of_transaction || !result) - end_of_transaction_command(aux_engine, request); - - /* mask AUX interrupt */ - - return result; -} - -void dal_aux_engine_construct( - struct aux_engine *engine, - struct dc_context *ctx) -{ - dal_i2caux_construct_engine(&engine->base, ctx); - engine->delay = 0; - engine->max_defer_write_retry = 0; -} - -void dal_aux_engine_destruct( - struct aux_engine *engine) -{ - dal_i2caux_destruct_engine(&engine->base); -} diff --git a/drivers/gpu/drm/amd/display/dc/i2caux/aux_engine.h b/drivers/gpu/drm/amd/display/dc/i2caux/aux_engine.h deleted file mode 100644 index c33a2898d967..000000000000 --- a/drivers/gpu/drm/amd/display/dc/i2caux/aux_engine.h +++ /dev/null @@ -1,86 +0,0 @@ -/* - * Copyright 2012-15 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. - * - * Authors: AMD - * - */ - -#ifndef __DAL_AUX_ENGINE_H__ -#define __DAL_AUX_ENGINE_H__ - -#include "dc_ddc_types.h" - -struct aux_engine; - -struct aux_engine_funcs { - void (*destroy)( - struct aux_engine **ptr); - bool (*acquire_engine)( - struct aux_engine *engine); - void (*configure)( - struct aux_engine *engine, - union aux_config cfg); - void (*submit_channel_request)( - struct aux_engine *engine, - struct aux_request_transaction_data *request); - void (*process_channel_reply)( - struct aux_engine *engine, - struct aux_reply_transaction_data *reply); - int (*read_channel_reply)( - struct aux_engine *engine, - uint32_t size, - uint8_t *buffer, - uint8_t *reply_result, - uint32_t *sw_status); - enum aux_channel_operation_result (*get_channel_status)( - struct aux_engine *engine, - uint8_t *returned_bytes); - bool (*is_engine_available) ( - struct aux_engine *engine); -}; - -struct aux_engine { - struct engine base; - const struct aux_engine_funcs *funcs; - /* following values are expressed in milliseconds */ - uint32_t delay; - uint32_t max_defer_write_retry; - - bool acquire_reset; -}; - -void dal_aux_engine_construct( - struct aux_engine *engine, - struct dc_context *ctx); - -void dal_aux_engine_destruct( - struct aux_engine *engine); -bool dal_aux_engine_submit_request( - struct engine *ptr, - struct i2caux_transaction_request *request, - bool middle_of_transaction); -bool dal_aux_engine_acquire( - struct engine *ptr, - struct ddc *ddc); -enum i2caux_engine_type dal_aux_engine_get_engine_type( - const struct engine *engine); - -#endif diff --git a/drivers/gpu/drm/amd/display/dc/i2caux/dce100/i2caux_dce100.c b/drivers/gpu/drm/amd/display/dc/i2caux/dce100/i2caux_dce100.c deleted file mode 100644 index 8b704ab0471c..000000000000 --- a/drivers/gpu/drm/amd/display/dc/i2caux/dce100/i2caux_dce100.c +++ /dev/null @@ -1,106 +0,0 @@ -/* - * Copyright 2012-15 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. - * - * Authors: AMD - * - */ - -#include "dm_services.h" - -#include "include/i2caux_interface.h" -#include "../i2caux.h" -#include "../engine.h" -#include "../i2c_engine.h" -#include "../i2c_sw_engine.h" -#include "../i2c_hw_engine.h" - -#include "../dce110/aux_engine_dce110.h" -#include "../dce110/i2c_hw_engine_dce110.h" -#include "../dce110/i2caux_dce110.h" - -#include "dce/dce_10_0_d.h" -#include "dce/dce_10_0_sh_mask.h" - -/* set register offset */ -#define SR(reg_name)\ - .reg_name = mm ## reg_name - -/* set register offset with instance */ -#define SRI(reg_name, block, id)\ - .reg_name = mm ## block ## id ## _ ## reg_name - -#define aux_regs(id)\ -[id] = {\ - AUX_COMMON_REG_LIST(id), \ - .AUX_RESET_MASK = 0 \ -} - -#define hw_engine_regs(id)\ -{\ - I2C_HW_ENGINE_COMMON_REG_LIST(id) \ -} - -static const struct dce110_aux_registers dce100_aux_regs[] = { - aux_regs(0), - aux_regs(1), - aux_regs(2), - aux_regs(3), - aux_regs(4), - aux_regs(5), -}; - -static const struct dce110_i2c_hw_engine_registers dce100_hw_engine_regs[] = { - hw_engine_regs(1), - hw_engine_regs(2), - hw_engine_regs(3), - hw_engine_regs(4), - hw_engine_regs(5), - hw_engine_regs(6) -}; - -static const struct dce110_i2c_hw_engine_shift i2c_shift = { - I2C_COMMON_MASK_SH_LIST_DCE100(__SHIFT) -}; - -static const struct dce110_i2c_hw_engine_mask i2c_mask = { - I2C_COMMON_MASK_SH_LIST_DCE100(_MASK) -}; - -struct i2caux *dal_i2caux_dce100_create( - struct dc_context *ctx) -{ - struct i2caux_dce110 *i2caux_dce110 = - kzalloc(sizeof(struct i2caux_dce110), GFP_KERNEL); - - if (!i2caux_dce110) { - ASSERT_CRITICAL(false); - return NULL; - } - - dal_i2caux_dce110_construct(i2caux_dce110, - ctx, - ARRAY_SIZE(dce100_aux_regs), - dce100_aux_regs, - dce100_hw_engine_regs, - &i2c_shift, - &i2c_mask); - return &i2caux_dce110->base; -} diff --git a/drivers/gpu/drm/amd/display/dc/i2caux/dce110/aux_engine_dce110.c b/drivers/gpu/drm/amd/display/dc/i2caux/dce110/aux_engine_dce110.c deleted file mode 100644 index 59c3ed43d609..000000000000 --- a/drivers/gpu/drm/amd/display/dc/i2caux/dce110/aux_engine_dce110.c +++ /dev/null @@ -1,505 +0,0 @@ -/* - * Copyright 2012-15 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. - * - * Authors: AMD - * - */ - -#include "dm_services.h" -#include "dm_event_log.h" - -/* - * Pre-requisites: headers required by header of this unit - */ -#include "include/i2caux_interface.h" -#include "../engine.h" -#include "../aux_engine.h" - -/* - * Header of this unit - */ - -#include "aux_engine_dce110.h" - -/* - * Post-requisites: headers required by this unit - */ -#include "dce/dce_11_0_sh_mask.h" - -#define CTX \ - aux110->base.base.ctx -#define REG(reg_name)\ - (aux110->regs->reg_name) -#include "reg_helper.h" - -/* - * This unit - */ - -/* - * @brief - * Cast 'struct aux_engine *' - * to 'struct aux_engine_dce110 *' - */ -#define FROM_AUX_ENGINE(ptr) \ - container_of((ptr), struct aux_engine_dce110, base) - -/* - * @brief - * Cast 'struct engine *' - * to 'struct aux_engine_dce110 *' - */ -#define FROM_ENGINE(ptr) \ - FROM_AUX_ENGINE(container_of((ptr), struct aux_engine, base)) - -static void release_engine( - struct engine *engine) -{ - struct aux_engine_dce110 *aux110 = FROM_ENGINE(engine); - - REG_UPDATE(AUX_ARB_CONTROL, AUX_SW_DONE_USING_AUX_REG, 1); -} - -static void destruct( - struct aux_engine_dce110 *engine); - -static void destroy( - struct aux_engine **aux_engine) -{ - struct aux_engine_dce110 *engine = FROM_AUX_ENGINE(*aux_engine); - - destruct(engine); - - kfree(engine); - - *aux_engine = NULL; -} - -#define SW_CAN_ACCESS_AUX 1 -#define DMCU_CAN_ACCESS_AUX 2 - -static bool is_engine_available( - struct aux_engine *engine) -{ - struct aux_engine_dce110 *aux110 = FROM_AUX_ENGINE(engine); - - uint32_t value = REG_READ(AUX_ARB_CONTROL); - uint32_t field = get_reg_field_value( - value, - AUX_ARB_CONTROL, - AUX_REG_RW_CNTL_STATUS); - - return (field != DMCU_CAN_ACCESS_AUX); -} -static bool acquire_engine( - struct aux_engine *engine) -{ - struct aux_engine_dce110 *aux110 = FROM_AUX_ENGINE(engine); - - uint32_t value = REG_READ(AUX_ARB_CONTROL); - uint32_t field = get_reg_field_value( - value, - AUX_ARB_CONTROL, - AUX_REG_RW_CNTL_STATUS); - if (field == DMCU_CAN_ACCESS_AUX) - return false; - /* enable AUX before request SW to access AUX */ - value = REG_READ(AUX_CONTROL); - field = get_reg_field_value(value, - AUX_CONTROL, - AUX_EN); - - if (field == 0) { - set_reg_field_value( - value, - 1, - AUX_CONTROL, - AUX_EN); - - if (REG(AUX_RESET_MASK)) { - /*DP_AUX block as part of the enable sequence*/ - set_reg_field_value( - value, - 1, - AUX_CONTROL, - AUX_RESET); - } - - REG_WRITE(AUX_CONTROL, value); - - if (REG(AUX_RESET_MASK)) { - /*poll HW to make sure reset it done*/ - - REG_WAIT(AUX_CONTROL, AUX_RESET_DONE, 1, - 1, 11); - - set_reg_field_value( - value, - 0, - AUX_CONTROL, - AUX_RESET); - - REG_WRITE(AUX_CONTROL, value); - - REG_WAIT(AUX_CONTROL, AUX_RESET_DONE, 0, - 1, 11); - } - } /*if (field)*/ - - /* request SW to access AUX */ - REG_UPDATE(AUX_ARB_CONTROL, AUX_SW_USE_AUX_REG_REQ, 1); - - value = REG_READ(AUX_ARB_CONTROL); - field = get_reg_field_value( - value, - AUX_ARB_CONTROL, - AUX_REG_RW_CNTL_STATUS); - - return (field == SW_CAN_ACCESS_AUX); -} - -#define COMPOSE_AUX_SW_DATA_16_20(command, address) \ - ((command) | ((0xF0000 & (address)) >> 16)) - -#define COMPOSE_AUX_SW_DATA_8_15(address) \ - ((0xFF00 & (address)) >> 8) - -#define COMPOSE_AUX_SW_DATA_0_7(address) \ - (0xFF & (address)) - -static void submit_channel_request( - struct aux_engine *engine, - struct aux_request_transaction_data *request) -{ - struct aux_engine_dce110 *aux110 = FROM_AUX_ENGINE(engine); - uint32_t value; - uint32_t length; - - bool is_write = - ((request->type == AUX_TRANSACTION_TYPE_DP) && - (request->action == I2CAUX_TRANSACTION_ACTION_DP_WRITE)) || - ((request->type == AUX_TRANSACTION_TYPE_I2C) && - ((request->action == I2CAUX_TRANSACTION_ACTION_I2C_WRITE) || - (request->action == I2CAUX_TRANSACTION_ACTION_I2C_WRITE_MOT))); - if (REG(AUXN_IMPCAL)) { - /* clear_aux_error */ - REG_UPDATE_SEQ(AUXN_IMPCAL, AUXN_CALOUT_ERROR_AK, - 1, - 0); - - REG_UPDATE_SEQ(AUXP_IMPCAL, AUXP_CALOUT_ERROR_AK, - 1, - 0); - - /* force_default_calibrate */ - REG_UPDATE_1BY1_2(AUXN_IMPCAL, - AUXN_IMPCAL_ENABLE, 1, - AUXN_IMPCAL_OVERRIDE_ENABLE, 0); - - /* bug? why AUXN update EN and OVERRIDE_EN 1 by 1 while AUX P toggles OVERRIDE? */ - - REG_UPDATE_SEQ(AUXP_IMPCAL, AUXP_IMPCAL_OVERRIDE_ENABLE, - 1, - 0); - } - /* set the delay and the number of bytes to write */ - - /* The length include - * the 4 bit header and the 20 bit address - * (that is 3 byte). - * If the requested length is non zero this means - * an addition byte specifying the length is required. */ - - length = request->length ? 4 : 3; - if (is_write) - length += request->length; - - REG_UPDATE_2(AUX_SW_CONTROL, - AUX_SW_START_DELAY, request->delay, - AUX_SW_WR_BYTES, length); - - /* program action and address and payload data (if 'is_write') */ - value = REG_UPDATE_4(AUX_SW_DATA, - AUX_SW_INDEX, 0, - AUX_SW_DATA_RW, 0, - AUX_SW_AUTOINCREMENT_DISABLE, 1, - AUX_SW_DATA, COMPOSE_AUX_SW_DATA_16_20(request->action, request->address)); - - value = REG_SET_2(AUX_SW_DATA, value, - AUX_SW_AUTOINCREMENT_DISABLE, 0, - AUX_SW_DATA, COMPOSE_AUX_SW_DATA_8_15(request->address)); - - value = REG_SET(AUX_SW_DATA, value, - AUX_SW_DATA, COMPOSE_AUX_SW_DATA_0_7(request->address)); - - if (request->length) { - value = REG_SET(AUX_SW_DATA, value, - AUX_SW_DATA, request->length - 1); - } - - if (is_write) { - /* Load the HW buffer with the Data to be sent. - * This is relevant for write operation. - * For read, the data recived data will be - * processed in process_channel_reply(). */ - uint32_t i = 0; - - while (i < request->length) { - value = REG_SET(AUX_SW_DATA, value, - AUX_SW_DATA, request->data[i]); - - ++i; - } - } - - REG_UPDATE(AUX_INTERRUPT_CONTROL, AUX_SW_DONE_ACK, 1); - REG_WAIT(AUX_SW_STATUS, AUX_SW_DONE, 0, - 10, aux110->timeout_period/10); - REG_UPDATE(AUX_SW_CONTROL, AUX_SW_GO, 1); - EVENT_LOG_AUX_REQ(engine->base.ddc->pin_data->en, EVENT_LOG_AUX_ORIGIN_NATIVE, - request->action, request->address, request->length, request->data); -} - -static int read_channel_reply(struct aux_engine *engine, uint32_t size, - uint8_t *buffer, uint8_t *reply_result, - uint32_t *sw_status) -{ - struct aux_engine_dce110 *aux110 = FROM_AUX_ENGINE(engine); - uint32_t bytes_replied; - uint32_t reply_result_32; - - *sw_status = REG_GET(AUX_SW_STATUS, AUX_SW_REPLY_BYTE_COUNT, - &bytes_replied); - - /* In case HPD is LOW, exit AUX transaction */ - if ((*sw_status & AUX_SW_STATUS__AUX_SW_HPD_DISCON_MASK)) - return -1; - - /* Need at least the status byte */ - if (!bytes_replied) - return -1; - - REG_UPDATE_1BY1_3(AUX_SW_DATA, - AUX_SW_INDEX, 0, - AUX_SW_AUTOINCREMENT_DISABLE, 1, - AUX_SW_DATA_RW, 1); - - REG_GET(AUX_SW_DATA, AUX_SW_DATA, &reply_result_32); - reply_result_32 = reply_result_32 >> 4; - *reply_result = (uint8_t)reply_result_32; - - if (reply_result_32 == 0) { /* ACK */ - uint32_t i = 0; - - /* First byte was already used to get the command status */ - --bytes_replied; - - /* Do not overflow buffer */ - if (bytes_replied > size) - return -1; - - while (i < bytes_replied) { - uint32_t aux_sw_data_val; - - REG_GET(AUX_SW_DATA, AUX_SW_DATA, &aux_sw_data_val); - buffer[i] = aux_sw_data_val; - ++i; - } - - return i; - } - - return 0; -} - -static void process_channel_reply( - struct aux_engine *engine, - struct aux_reply_transaction_data *reply) -{ - int bytes_replied; - uint8_t reply_result; - uint32_t sw_status; - - bytes_replied = read_channel_reply(engine, reply->length, reply->data, - &reply_result, &sw_status); - EVENT_LOG_AUX_REP(engine->base.ddc->pin_data->en, - EVENT_LOG_AUX_ORIGIN_NATIVE, reply_result, - bytes_replied, reply->data); - - /* in case HPD is LOW, exit AUX transaction */ - if ((sw_status & AUX_SW_STATUS__AUX_SW_HPD_DISCON_MASK)) { - reply->status = AUX_TRANSACTION_REPLY_HPD_DISCON; - return; - } - - if (bytes_replied < 0) { - /* Need to handle an error case... - * Hopefully, upper layer function won't call this function if - * the number of bytes in the reply was 0, because there was - * surely an error that was asserted that should have been - * handled for hot plug case, this could happens - */ - if (!(sw_status & AUX_SW_STATUS__AUX_SW_HPD_DISCON_MASK)) { - reply->status = AUX_TRANSACTION_REPLY_INVALID; - ASSERT_CRITICAL(false); - return; - } - } else { - - switch (reply_result) { - case 0: /* ACK */ - reply->status = AUX_TRANSACTION_REPLY_AUX_ACK; - break; - case 1: /* NACK */ - reply->status = AUX_TRANSACTION_REPLY_AUX_NACK; - break; - case 2: /* DEFER */ - reply->status = AUX_TRANSACTION_REPLY_AUX_DEFER; - break; - case 4: /* AUX ACK / I2C NACK */ - reply->status = AUX_TRANSACTION_REPLY_I2C_NACK; - break; - case 8: /* AUX ACK / I2C DEFER */ - reply->status = AUX_TRANSACTION_REPLY_I2C_DEFER; - break; - default: - reply->status = AUX_TRANSACTION_REPLY_INVALID; - } - } -} - -static enum aux_channel_operation_result get_channel_status( - struct aux_engine *engine, - uint8_t *returned_bytes) -{ - struct aux_engine_dce110 *aux110 = FROM_AUX_ENGINE(engine); - - uint32_t value; - - if (returned_bytes == NULL) { - /*caller pass NULL pointer*/ - ASSERT_CRITICAL(false); - return AUX_CHANNEL_OPERATION_FAILED_REASON_UNKNOWN; - } - *returned_bytes = 0; - - /* poll to make sure that SW_DONE is asserted */ - value = REG_WAIT(AUX_SW_STATUS, AUX_SW_DONE, 1, - 10, aux110->timeout_period/10); - - /* in case HPD is LOW, exit AUX transaction */ - if ((value & AUX_SW_STATUS__AUX_SW_HPD_DISCON_MASK)) - return AUX_CHANNEL_OPERATION_FAILED_HPD_DISCON; - - /* Note that the following bits are set in 'status.bits' - * during CTS 4.2.1.2 (FW 3.3.1): - * AUX_SW_RX_MIN_COUNT_VIOL, AUX_SW_RX_INVALID_STOP, - * AUX_SW_RX_RECV_NO_DET, AUX_SW_RX_RECV_INVALID_H. - * - * AUX_SW_RX_MIN_COUNT_VIOL is an internal, - * HW debugging bit and should be ignored. */ - if (value & AUX_SW_STATUS__AUX_SW_DONE_MASK) { - if ((value & AUX_SW_STATUS__AUX_SW_RX_TIMEOUT_STATE_MASK) || - (value & AUX_SW_STATUS__AUX_SW_RX_TIMEOUT_MASK)) - return AUX_CHANNEL_OPERATION_FAILED_TIMEOUT; - - else if ((value & AUX_SW_STATUS__AUX_SW_RX_INVALID_STOP_MASK) || - (value & AUX_SW_STATUS__AUX_SW_RX_RECV_NO_DET_MASK) || - (value & - AUX_SW_STATUS__AUX_SW_RX_RECV_INVALID_H_MASK) || - (value & AUX_SW_STATUS__AUX_SW_RX_RECV_INVALID_L_MASK)) - return AUX_CHANNEL_OPERATION_FAILED_INVALID_REPLY; - - *returned_bytes = get_reg_field_value(value, - AUX_SW_STATUS, - AUX_SW_REPLY_BYTE_COUNT); - - if (*returned_bytes == 0) - return - AUX_CHANNEL_OPERATION_FAILED_INVALID_REPLY; - else { - *returned_bytes -= 1; - return AUX_CHANNEL_OPERATION_SUCCEEDED; - } - } else { - /*time_elapsed >= aux_engine->timeout_period - * AUX_SW_STATUS__AUX_SW_HPD_DISCON = at this point - */ - ASSERT_CRITICAL(false); - return AUX_CHANNEL_OPERATION_FAILED_TIMEOUT; - } -} - -static const struct aux_engine_funcs aux_engine_funcs = { - .destroy = destroy, - .acquire_engine = acquire_engine, - .submit_channel_request = submit_channel_request, - .process_channel_reply = process_channel_reply, - .read_channel_reply = read_channel_reply, - .get_channel_status = get_channel_status, - .is_engine_available = is_engine_available, -}; - -static const struct engine_funcs engine_funcs = { - .release_engine = release_engine, - .submit_request = dal_aux_engine_submit_request, - .get_engine_type = dal_aux_engine_get_engine_type, - .acquire = dal_aux_engine_acquire, -}; - -static void construct( - struct aux_engine_dce110 *engine, - const struct aux_engine_dce110_init_data *aux_init_data) -{ - dal_aux_engine_construct(&engine->base, aux_init_data->ctx); - engine->base.base.funcs = &engine_funcs; - engine->base.funcs = &aux_engine_funcs; - - engine->timeout_period = aux_init_data->timeout_period; - engine->regs = aux_init_data->regs; -} - -static void destruct( - struct aux_engine_dce110 *engine) -{ - dal_aux_engine_destruct(&engine->base); -} - -struct aux_engine *dal_aux_engine_dce110_create( - const struct aux_engine_dce110_init_data *aux_init_data) -{ - struct aux_engine_dce110 *engine; - - if (!aux_init_data) { - ASSERT_CRITICAL(false); - return NULL; - } - - engine = kzalloc(sizeof(*engine), GFP_KERNEL); - - if (!engine) { - ASSERT_CRITICAL(false); - return NULL; - } - - construct(engine, aux_init_data); - return &engine->base; -} diff --git a/drivers/gpu/drm/amd/display/dc/i2caux/dce110/aux_engine_dce110.h b/drivers/gpu/drm/amd/display/dc/i2caux/dce110/aux_engine_dce110.h deleted file mode 100644 index 85ee82162590..000000000000 --- a/drivers/gpu/drm/amd/display/dc/i2caux/dce110/aux_engine_dce110.h +++ /dev/null @@ -1,78 +0,0 @@ -/* - * Copyright 2012-15 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. - * - * Authors: AMD - * - */ - -#ifndef __DAL_AUX_ENGINE_DCE110_H__ -#define __DAL_AUX_ENGINE_DCE110_H__ - -#include "../aux_engine.h" - -#define AUX_COMMON_REG_LIST(id)\ - SRI(AUX_CONTROL, DP_AUX, id), \ - SRI(AUX_ARB_CONTROL, DP_AUX, id), \ - SRI(AUX_SW_DATA, DP_AUX, id), \ - SRI(AUX_SW_CONTROL, DP_AUX, id), \ - SRI(AUX_INTERRUPT_CONTROL, DP_AUX, id), \ - SRI(AUX_SW_STATUS, DP_AUX, id), \ - SR(AUXN_IMPCAL), \ - SR(AUXP_IMPCAL) - -struct dce110_aux_registers { - uint32_t AUX_CONTROL; - uint32_t AUX_ARB_CONTROL; - uint32_t AUX_SW_DATA; - uint32_t AUX_SW_CONTROL; - uint32_t AUX_INTERRUPT_CONTROL; - uint32_t AUX_SW_STATUS; - uint32_t AUXN_IMPCAL; - uint32_t AUXP_IMPCAL; - - uint32_t AUX_RESET_MASK; -}; - -struct aux_engine_dce110 { - struct aux_engine base; - const struct dce110_aux_registers *regs; - struct { - uint32_t aux_control; - uint32_t aux_arb_control; - uint32_t aux_sw_data; - uint32_t aux_sw_control; - uint32_t aux_interrupt_control; - uint32_t aux_sw_status; - } addr; - uint32_t timeout_period; -}; - -struct aux_engine_dce110_init_data { - uint32_t engine_id; - uint32_t timeout_period; - struct dc_context *ctx; - const struct dce110_aux_registers *regs; -}; - -struct aux_engine *dal_aux_engine_dce110_create( - const struct aux_engine_dce110_init_data *aux_init_data); - -#endif diff --git a/drivers/gpu/drm/amd/display/dc/i2caux/dce110/i2c_hw_engine_dce110.c b/drivers/gpu/drm/amd/display/dc/i2caux/dce110/i2c_hw_engine_dce110.c deleted file mode 100644 index 9cbe1a7a6bcb..000000000000 --- a/drivers/gpu/drm/amd/display/dc/i2caux/dce110/i2c_hw_engine_dce110.c +++ /dev/null @@ -1,574 +0,0 @@ -/* - * Copyright 2012-15 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. - * - * Authors: AMD - * - */ - -#include "dm_services.h" -#include "include/logger_interface.h" -/* - * Pre-requisites: headers required by header of this unit - */ - -#include "include/i2caux_interface.h" -#include "../engine.h" -#include "../i2c_engine.h" -#include "../i2c_hw_engine.h" -#include "../i2c_generic_hw_engine.h" -/* - * Header of this unit - */ - -#include "i2c_hw_engine_dce110.h" - -/* - * Post-requisites: headers required by this unit - */ -#include "reg_helper.h" - -/* - * This unit - */ -#define DC_LOGGER \ - hw_engine->base.base.base.ctx->logger - -enum dc_i2c_status { - DC_I2C_STATUS__DC_I2C_STATUS_IDLE, - DC_I2C_STATUS__DC_I2C_STATUS_USED_BY_SW, - DC_I2C_STATUS__DC_I2C_STATUS_USED_BY_HW -}; - -enum dc_i2c_arbitration { - DC_I2C_ARBITRATION__DC_I2C_SW_PRIORITY_NORMAL, - DC_I2C_ARBITRATION__DC_I2C_SW_PRIORITY_HIGH -}; - - - -/* - * @brief - * Cast pointer to 'struct i2c_hw_engine *' - * to pointer 'struct i2c_hw_engine_dce110 *' - */ -#define FROM_I2C_HW_ENGINE(ptr) \ - container_of((ptr), struct i2c_hw_engine_dce110, base) -/* - * @brief - * Cast pointer to 'struct i2c_engine *' - * to pointer to 'struct i2c_hw_engine_dce110 *' - */ -#define FROM_I2C_ENGINE(ptr) \ - FROM_I2C_HW_ENGINE(container_of((ptr), struct i2c_hw_engine, base)) - -/* - * @brief - * Cast pointer to 'struct engine *' - * to 'pointer to struct i2c_hw_engine_dce110 *' - */ -#define FROM_ENGINE(ptr) \ - FROM_I2C_ENGINE(container_of((ptr), struct i2c_engine, base)) - -#define CTX \ - hw_engine->base.base.base.ctx - -#define REG(reg_name)\ - (hw_engine->regs->reg_name) - -#undef FN -#define FN(reg_name, field_name) \ - hw_engine->i2c_shift->field_name, hw_engine->i2c_mask->field_name - -#include "reg_helper.h" - -static void disable_i2c_hw_engine( - struct i2c_hw_engine_dce110 *hw_engine) -{ - REG_UPDATE_N(SETUP, 1, FN(DC_I2C_DDC1_SETUP, DC_I2C_DDC1_ENABLE), 0); -} - -static void release_engine( - struct engine *engine) -{ - struct i2c_hw_engine_dce110 *hw_engine = FROM_ENGINE(engine); - - struct i2c_engine *base = NULL; - bool safe_to_reset; - - base = &hw_engine->base.base; - - /* Restore original HW engine speed */ - - base->funcs->set_speed(base, hw_engine->base.original_speed); - - /* Release I2C */ - REG_UPDATE(DC_I2C_ARBITRATION, DC_I2C_SW_DONE_USING_I2C_REG, 1); - - /* Reset HW engine */ - { - uint32_t i2c_sw_status = 0; - REG_GET(DC_I2C_SW_STATUS, DC_I2C_SW_STATUS, &i2c_sw_status); - /* if used by SW, safe to reset */ - safe_to_reset = (i2c_sw_status == 1); - } - - if (safe_to_reset) - REG_UPDATE_2( - DC_I2C_CONTROL, - DC_I2C_SOFT_RESET, 1, - DC_I2C_SW_STATUS_RESET, 1); - else - REG_UPDATE(DC_I2C_CONTROL, DC_I2C_SW_STATUS_RESET, 1); - - /* HW I2c engine - clock gating feature */ - if (!hw_engine->engine_keep_power_up_count) - disable_i2c_hw_engine(hw_engine); -} - -static bool setup_engine( - struct i2c_engine *i2c_engine) -{ - struct i2c_hw_engine_dce110 *hw_engine = FROM_I2C_ENGINE(i2c_engine); - uint32_t i2c_setup_limit = I2C_SETUP_TIME_LIMIT_DCE; - uint32_t reset_length = 0; - - if (hw_engine->base.base.setup_limit != 0) - i2c_setup_limit = hw_engine->base.base.setup_limit; - - /* Program pin select */ - REG_UPDATE_6( - DC_I2C_CONTROL, - DC_I2C_GO, 0, - DC_I2C_SOFT_RESET, 0, - DC_I2C_SEND_RESET, 0, - DC_I2C_SW_STATUS_RESET, 1, - DC_I2C_TRANSACTION_COUNT, 0, - DC_I2C_DDC_SELECT, hw_engine->engine_id); - - /* Program time limit */ - if (hw_engine->base.base.send_reset_length == 0) { - /*pre-dcn*/ - REG_UPDATE_N( - SETUP, 2, - FN(DC_I2C_DDC1_SETUP, DC_I2C_DDC1_TIME_LIMIT), i2c_setup_limit, - FN(DC_I2C_DDC1_SETUP, DC_I2C_DDC1_ENABLE), 1); - } else { - reset_length = hw_engine->base.base.send_reset_length; - } - /* Program HW priority - * set to High - interrupt software I2C at any time - * Enable restart of SW I2C that was interrupted by HW - * disable queuing of software while I2C is in use by HW */ - REG_UPDATE_2( - DC_I2C_ARBITRATION, - DC_I2C_NO_QUEUED_SW_GO, 0, - DC_I2C_SW_PRIORITY, DC_I2C_ARBITRATION__DC_I2C_SW_PRIORITY_NORMAL); - - return true; -} - -static uint32_t get_speed( - const struct i2c_engine *i2c_engine) -{ - const struct i2c_hw_engine_dce110 *hw_engine = FROM_I2C_ENGINE(i2c_engine); - uint32_t pre_scale = 0; - - REG_GET(SPEED, DC_I2C_DDC1_PRESCALE, &pre_scale); - - /* [anaumov] it seems following is unnecessary */ - /*ASSERT(value.bits.DC_I2C_DDC1_PRESCALE);*/ - return pre_scale ? - hw_engine->reference_frequency / pre_scale : - hw_engine->base.default_speed; -} - -static void set_speed( - struct i2c_engine *i2c_engine, - uint32_t speed) -{ - struct i2c_hw_engine_dce110 *hw_engine = FROM_I2C_ENGINE(i2c_engine); - - if (speed) { - if (hw_engine->i2c_mask->DC_I2C_DDC1_START_STOP_TIMING_CNTL) - REG_UPDATE_N( - SPEED, 3, - FN(DC_I2C_DDC1_SPEED, DC_I2C_DDC1_PRESCALE), hw_engine->reference_frequency / speed, - FN(DC_I2C_DDC1_SPEED, DC_I2C_DDC1_THRESHOLD), 2, - FN(DC_I2C_DDC1_SPEED, DC_I2C_DDC1_START_STOP_TIMING_CNTL), speed > 50 ? 2:1); - else - REG_UPDATE_N( - SPEED, 2, - FN(DC_I2C_DDC1_SPEED, DC_I2C_DDC1_PRESCALE), hw_engine->reference_frequency / speed, - FN(DC_I2C_DDC1_SPEED, DC_I2C_DDC1_THRESHOLD), 2); - } -} - -static inline void reset_hw_engine(struct engine *engine) -{ - struct i2c_hw_engine_dce110 *hw_engine = FROM_ENGINE(engine); - - REG_UPDATE_2( - DC_I2C_CONTROL, - DC_I2C_SW_STATUS_RESET, 1, - DC_I2C_SW_STATUS_RESET, 1); -} - -static bool is_hw_busy(struct engine *engine) -{ - struct i2c_hw_engine_dce110 *hw_engine = FROM_ENGINE(engine); - uint32_t i2c_sw_status = 0; - - REG_GET(DC_I2C_SW_STATUS, DC_I2C_SW_STATUS, &i2c_sw_status); - if (i2c_sw_status == DC_I2C_STATUS__DC_I2C_STATUS_IDLE) - return false; - - reset_hw_engine(engine); - - REG_GET(DC_I2C_SW_STATUS, DC_I2C_SW_STATUS, &i2c_sw_status); - return i2c_sw_status != DC_I2C_STATUS__DC_I2C_STATUS_IDLE; -} - - -#define STOP_TRANS_PREDICAT \ - ((hw_engine->transaction_count == 3) || \ - (request->action == I2CAUX_TRANSACTION_ACTION_I2C_WRITE) || \ - (request->action & I2CAUX_TRANSACTION_ACTION_I2C_READ)) - -#define SET_I2C_TRANSACTION(id) \ - do { \ - REG_UPDATE_N(DC_I2C_TRANSACTION##id, 5, \ - FN(DC_I2C_TRANSACTION0, DC_I2C_STOP_ON_NACK0), 1, \ - FN(DC_I2C_TRANSACTION0, DC_I2C_START0), 1, \ - FN(DC_I2C_TRANSACTION0, DC_I2C_STOP0), STOP_TRANS_PREDICAT ? 1:0, \ - FN(DC_I2C_TRANSACTION0, DC_I2C_RW0), (0 != (request->action & I2CAUX_TRANSACTION_ACTION_I2C_READ)), \ - FN(DC_I2C_TRANSACTION0, DC_I2C_COUNT0), length); \ - if (STOP_TRANS_PREDICAT) \ - last_transaction = true; \ - } while (false) - - -static bool process_transaction( - struct i2c_hw_engine_dce110 *hw_engine, - struct i2c_request_transaction_data *request) -{ - uint32_t length = request->length; - uint8_t *buffer = request->data; - uint32_t value = 0; - - bool last_transaction = false; - - struct dc_context *ctx = NULL; - - ctx = hw_engine->base.base.base.ctx; - - - - switch (hw_engine->transaction_count) { - case 0: - SET_I2C_TRANSACTION(0); - break; - case 1: - SET_I2C_TRANSACTION(1); - break; - case 2: - SET_I2C_TRANSACTION(2); - break; - case 3: - SET_I2C_TRANSACTION(3); - break; - default: - /* TODO Warning ? */ - break; - } - - - /* Write the I2C address and I2C data - * into the hardware circular buffer, one byte per entry. - * As an example, the 7-bit I2C slave address for CRT monitor - * for reading DDC/EDID information is 0b1010001. - * For an I2C send operation, the LSB must be programmed to 0; - * for I2C receive operation, the LSB must be programmed to 1. */ - if (hw_engine->transaction_count == 0) { - value = REG_SET_4(DC_I2C_DATA, 0, - DC_I2C_DATA_RW, false, - DC_I2C_DATA, request->address, - DC_I2C_INDEX, 0, - DC_I2C_INDEX_WRITE, 1); - hw_engine->buffer_used_write = 0; - } else - value = REG_SET_2(DC_I2C_DATA, 0, - DC_I2C_DATA_RW, false, - DC_I2C_DATA, request->address); - - hw_engine->buffer_used_write++; - - if (!(request->action & I2CAUX_TRANSACTION_ACTION_I2C_READ)) { - while (length) { - REG_SET_2(DC_I2C_DATA, value, - DC_I2C_INDEX_WRITE, 0, - DC_I2C_DATA, *buffer++); - hw_engine->buffer_used_write++; - --length; - } - } - - ++hw_engine->transaction_count; - hw_engine->buffer_used_bytes += length + 1; - - return last_transaction; -} - -static void execute_transaction( - struct i2c_hw_engine_dce110 *hw_engine) -{ - REG_UPDATE_N(SETUP, 5, - FN(DC_I2C_DDC1_SETUP, DC_I2C_DDC1_DATA_DRIVE_EN), 0, - FN(DC_I2C_DDC1_SETUP, DC_I2C_DDC1_CLK_DRIVE_EN), 0, - FN(DC_I2C_DDC1_SETUP, DC_I2C_DDC1_DATA_DRIVE_SEL), 0, - FN(DC_I2C_DDC1_SETUP, DC_I2C_DDC1_INTRA_TRANSACTION_DELAY), 0, - FN(DC_I2C_DDC1_SETUP, DC_I2C_DDC1_INTRA_BYTE_DELAY), 0); - - - REG_UPDATE_5(DC_I2C_CONTROL, - DC_I2C_SOFT_RESET, 0, - DC_I2C_SW_STATUS_RESET, 0, - DC_I2C_SEND_RESET, 0, - DC_I2C_GO, 0, - DC_I2C_TRANSACTION_COUNT, hw_engine->transaction_count - 1); - - /* start I2C transfer */ - REG_UPDATE(DC_I2C_CONTROL, DC_I2C_GO, 1); - - /* all transactions were executed and HW buffer became empty - * (even though it actually happens when status becomes DONE) */ - hw_engine->transaction_count = 0; - hw_engine->buffer_used_bytes = 0; -} - -static void submit_channel_request( - struct i2c_engine *engine, - struct i2c_request_transaction_data *request) -{ - request->status = I2C_CHANNEL_OPERATION_SUCCEEDED; - - if (!process_transaction(FROM_I2C_ENGINE(engine), request)) - return; - - if (is_hw_busy(&engine->base)) { - request->status = I2C_CHANNEL_OPERATION_ENGINE_BUSY; - return; - } - - execute_transaction(FROM_I2C_ENGINE(engine)); -} - -static void process_channel_reply( - struct i2c_engine *engine, - struct i2c_reply_transaction_data *reply) -{ - uint32_t length = reply->length; - uint8_t *buffer = reply->data; - - struct i2c_hw_engine_dce110 *hw_engine = - FROM_I2C_ENGINE(engine); - - - REG_SET_3(DC_I2C_DATA, 0, - DC_I2C_INDEX, hw_engine->buffer_used_write, - DC_I2C_DATA_RW, 1, - DC_I2C_INDEX_WRITE, 1); - - while (length) { - /* after reading the status, - * if the I2C operation executed successfully - * (i.e. DC_I2C_STATUS_DONE = 1) then the I2C controller - * should read data bytes from I2C circular data buffer */ - - uint32_t i2c_data; - - REG_GET(DC_I2C_DATA, DC_I2C_DATA, &i2c_data); - *buffer++ = i2c_data; - - --length; - } -} - -static enum i2c_channel_operation_result get_channel_status( - struct i2c_engine *i2c_engine, - uint8_t *returned_bytes) -{ - uint32_t i2c_sw_status = 0; - struct i2c_hw_engine_dce110 *hw_engine = FROM_I2C_ENGINE(i2c_engine); - uint32_t value = - REG_GET(DC_I2C_SW_STATUS, DC_I2C_SW_STATUS, &i2c_sw_status); - - if (i2c_sw_status == DC_I2C_STATUS__DC_I2C_STATUS_USED_BY_SW) - return I2C_CHANNEL_OPERATION_ENGINE_BUSY; - else if (value & hw_engine->i2c_mask->DC_I2C_SW_STOPPED_ON_NACK) - return I2C_CHANNEL_OPERATION_NO_RESPONSE; - else if (value & hw_engine->i2c_mask->DC_I2C_SW_TIMEOUT) - return I2C_CHANNEL_OPERATION_TIMEOUT; - else if (value & hw_engine->i2c_mask->DC_I2C_SW_ABORTED) - return I2C_CHANNEL_OPERATION_FAILED; - else if (value & hw_engine->i2c_mask->DC_I2C_SW_DONE) - return I2C_CHANNEL_OPERATION_SUCCEEDED; - - /* - * this is the case when HW used for communication, I2C_SW_STATUS - * could be zero - */ - return I2C_CHANNEL_OPERATION_SUCCEEDED; -} - -static uint32_t get_hw_buffer_available_size( - const struct i2c_hw_engine *engine) -{ - return I2C_HW_BUFFER_SIZE - - FROM_I2C_HW_ENGINE(engine)->buffer_used_bytes; -} - -static uint32_t get_transaction_timeout( - const struct i2c_hw_engine *engine, - uint32_t length) -{ - uint32_t speed = engine->base.funcs->get_speed(&engine->base); - - uint32_t period_timeout; - uint32_t num_of_clock_stretches; - - if (!speed) - return 0; - - period_timeout = (1000 * TRANSACTION_TIMEOUT_IN_I2C_CLOCKS) / speed; - - num_of_clock_stretches = 1 + (length << 3) + 1; - num_of_clock_stretches += - (FROM_I2C_HW_ENGINE(engine)->buffer_used_bytes << 3) + - (FROM_I2C_HW_ENGINE(engine)->transaction_count << 1); - - return period_timeout * num_of_clock_stretches; -} - -static void destroy( - struct i2c_engine **i2c_engine) -{ - struct i2c_hw_engine_dce110 *engine_dce110 = - FROM_I2C_ENGINE(*i2c_engine); - - dal_i2c_hw_engine_destruct(&engine_dce110->base); - - kfree(engine_dce110); - - *i2c_engine = NULL; -} - -static const struct i2c_engine_funcs i2c_engine_funcs = { - .destroy = destroy, - .get_speed = get_speed, - .set_speed = set_speed, - .setup_engine = setup_engine, - .submit_channel_request = submit_channel_request, - .process_channel_reply = process_channel_reply, - .get_channel_status = get_channel_status, - .acquire_engine = dal_i2c_hw_engine_acquire_engine, -}; - -static const struct engine_funcs engine_funcs = { - .release_engine = release_engine, - .get_engine_type = dal_i2c_hw_engine_get_engine_type, - .acquire = dal_i2c_engine_acquire, - .submit_request = dal_i2c_hw_engine_submit_request, -}; - -static const struct i2c_hw_engine_funcs i2c_hw_engine_funcs = { - .get_hw_buffer_available_size = get_hw_buffer_available_size, - .get_transaction_timeout = get_transaction_timeout, - .wait_on_operation_result = dal_i2c_hw_engine_wait_on_operation_result, -}; - -static void construct( - struct i2c_hw_engine_dce110 *hw_engine, - const struct i2c_hw_engine_dce110_create_arg *arg) -{ - uint32_t xtal_ref_div = 0; - - dal_i2c_hw_engine_construct(&hw_engine->base, arg->ctx); - - hw_engine->base.base.base.funcs = &engine_funcs; - hw_engine->base.base.funcs = &i2c_engine_funcs; - hw_engine->base.funcs = &i2c_hw_engine_funcs; - hw_engine->base.default_speed = arg->default_speed; - - hw_engine->regs = arg->regs; - hw_engine->i2c_shift = arg->i2c_shift; - hw_engine->i2c_mask = arg->i2c_mask; - - hw_engine->engine_id = arg->engine_id; - - hw_engine->buffer_used_bytes = 0; - hw_engine->transaction_count = 0; - hw_engine->engine_keep_power_up_count = 1; - - - REG_GET(MICROSECOND_TIME_BASE_DIV, XTAL_REF_DIV, &xtal_ref_div); - - if (xtal_ref_div == 0) { - DC_LOG_WARNING("Invalid base timer divider [%s]\n", - __func__); - xtal_ref_div = 2; - } - - /*Calculating Reference Clock by divding original frequency by - * XTAL_REF_DIV. - * At upper level, uint32_t reference_frequency = - * dal_i2caux_get_reference_clock(as) >> 1 - * which already divided by 2. So we need x2 to get original - * reference clock from ppll_info - */ - hw_engine->reference_frequency = - (arg->reference_frequency * 2) / xtal_ref_div; -} - -struct i2c_engine *dal_i2c_hw_engine_dce110_create( - const struct i2c_hw_engine_dce110_create_arg *arg) -{ - struct i2c_hw_engine_dce110 *engine_dce10; - - if (!arg) { - ASSERT_CRITICAL(false); - return NULL; - } - if (!arg->reference_frequency) { - ASSERT_CRITICAL(false); - return NULL; - } - - engine_dce10 = kzalloc(sizeof(struct i2c_hw_engine_dce110), - GFP_KERNEL); - - if (!engine_dce10) { - ASSERT_CRITICAL(false); - return NULL; - } - - construct(engine_dce10, arg); - return &engine_dce10->base.base; -} diff --git a/drivers/gpu/drm/amd/display/dc/i2caux/dce110/i2c_hw_engine_dce110.h b/drivers/gpu/drm/amd/display/dc/i2caux/dce110/i2c_hw_engine_dce110.h deleted file mode 100644 index fea2946906ed..000000000000 --- a/drivers/gpu/drm/amd/display/dc/i2caux/dce110/i2c_hw_engine_dce110.h +++ /dev/null @@ -1,218 +0,0 @@ -/* - * Copyright 2012-15 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. - * - * Authors: AMD - * - */ - -#ifndef __DAL_I2C_HW_ENGINE_DCE110_H__ -#define __DAL_I2C_HW_ENGINE_DCE110_H__ - -#define I2C_HW_ENGINE_COMMON_REG_LIST(id)\ - SRI(SETUP, DC_I2C_DDC, id),\ - SRI(SPEED, DC_I2C_DDC, id),\ - SR(DC_I2C_ARBITRATION),\ - SR(DC_I2C_CONTROL),\ - SR(DC_I2C_SW_STATUS),\ - SR(DC_I2C_TRANSACTION0),\ - SR(DC_I2C_TRANSACTION1),\ - SR(DC_I2C_TRANSACTION2),\ - SR(DC_I2C_TRANSACTION3),\ - SR(DC_I2C_DATA),\ - SR(MICROSECOND_TIME_BASE_DIV) - -#define I2C_SF(reg_name, field_name, post_fix)\ - .field_name = reg_name ## __ ## field_name ## post_fix - -#define I2C_COMMON_MASK_SH_LIST_DCE_COMMON_BASE(mask_sh)\ - I2C_SF(DC_I2C_DDC1_SETUP, DC_I2C_DDC1_ENABLE, mask_sh),\ - I2C_SF(DC_I2C_DDC1_SETUP, DC_I2C_DDC1_TIME_LIMIT, mask_sh),\ - I2C_SF(DC_I2C_DDC1_SETUP, DC_I2C_DDC1_DATA_DRIVE_EN, mask_sh),\ - I2C_SF(DC_I2C_DDC1_SETUP, DC_I2C_DDC1_CLK_DRIVE_EN, mask_sh),\ - I2C_SF(DC_I2C_DDC1_SETUP, DC_I2C_DDC1_DATA_DRIVE_SEL, mask_sh),\ - I2C_SF(DC_I2C_DDC1_SETUP, DC_I2C_DDC1_INTRA_TRANSACTION_DELAY, mask_sh),\ - I2C_SF(DC_I2C_DDC1_SETUP, DC_I2C_DDC1_INTRA_BYTE_DELAY, mask_sh),\ - I2C_SF(DC_I2C_ARBITRATION, DC_I2C_SW_DONE_USING_I2C_REG, mask_sh),\ - I2C_SF(DC_I2C_ARBITRATION, DC_I2C_NO_QUEUED_SW_GO, mask_sh),\ - I2C_SF(DC_I2C_ARBITRATION, DC_I2C_SW_PRIORITY, mask_sh),\ - I2C_SF(DC_I2C_CONTROL, DC_I2C_SOFT_RESET, mask_sh),\ - I2C_SF(DC_I2C_CONTROL, DC_I2C_SW_STATUS_RESET, mask_sh),\ - I2C_SF(DC_I2C_CONTROL, DC_I2C_GO, mask_sh),\ - I2C_SF(DC_I2C_CONTROL, DC_I2C_SEND_RESET, mask_sh),\ - I2C_SF(DC_I2C_CONTROL, DC_I2C_TRANSACTION_COUNT, mask_sh),\ - I2C_SF(DC_I2C_CONTROL, DC_I2C_DDC_SELECT, mask_sh),\ - I2C_SF(DC_I2C_DDC1_SPEED, DC_I2C_DDC1_PRESCALE, mask_sh),\ - I2C_SF(DC_I2C_DDC1_SPEED, DC_I2C_DDC1_THRESHOLD, mask_sh),\ - I2C_SF(DC_I2C_SW_STATUS, DC_I2C_SW_STOPPED_ON_NACK, mask_sh),\ - I2C_SF(DC_I2C_SW_STATUS, DC_I2C_SW_TIMEOUT, mask_sh),\ - I2C_SF(DC_I2C_SW_STATUS, DC_I2C_SW_ABORTED, mask_sh),\ - I2C_SF(DC_I2C_SW_STATUS, DC_I2C_SW_DONE, mask_sh),\ - I2C_SF(DC_I2C_SW_STATUS, DC_I2C_SW_STATUS, mask_sh),\ - I2C_SF(DC_I2C_TRANSACTION0, DC_I2C_STOP_ON_NACK0, mask_sh),\ - I2C_SF(DC_I2C_TRANSACTION0, DC_I2C_START0, mask_sh),\ - I2C_SF(DC_I2C_TRANSACTION0, DC_I2C_RW0, mask_sh),\ - I2C_SF(DC_I2C_TRANSACTION0, DC_I2C_STOP0, mask_sh),\ - I2C_SF(DC_I2C_TRANSACTION0, DC_I2C_COUNT0, mask_sh),\ - I2C_SF(DC_I2C_DATA, DC_I2C_DATA_RW, mask_sh),\ - I2C_SF(DC_I2C_DATA, DC_I2C_DATA, mask_sh),\ - I2C_SF(DC_I2C_DATA, DC_I2C_INDEX, mask_sh),\ - I2C_SF(DC_I2C_DATA, DC_I2C_INDEX_WRITE, mask_sh),\ - I2C_SF(MICROSECOND_TIME_BASE_DIV, XTAL_REF_DIV, mask_sh) - -#define I2C_COMMON_MASK_SH_LIST_DCE100(mask_sh)\ - I2C_COMMON_MASK_SH_LIST_DCE_COMMON_BASE(mask_sh) - -#define I2C_COMMON_MASK_SH_LIST_DCE110(mask_sh)\ - I2C_COMMON_MASK_SH_LIST_DCE_COMMON_BASE(mask_sh),\ - I2C_SF(DC_I2C_DDC1_SPEED, DC_I2C_DDC1_START_STOP_TIMING_CNTL, mask_sh) - -struct dce110_i2c_hw_engine_shift { - uint8_t DC_I2C_DDC1_ENABLE; - uint8_t DC_I2C_DDC1_TIME_LIMIT; - uint8_t DC_I2C_DDC1_DATA_DRIVE_EN; - uint8_t DC_I2C_DDC1_CLK_DRIVE_EN; - uint8_t DC_I2C_DDC1_DATA_DRIVE_SEL; - uint8_t DC_I2C_DDC1_INTRA_TRANSACTION_DELAY; - uint8_t DC_I2C_DDC1_INTRA_BYTE_DELAY; - uint8_t DC_I2C_SW_DONE_USING_I2C_REG; - uint8_t DC_I2C_NO_QUEUED_SW_GO; - uint8_t DC_I2C_SW_PRIORITY; - uint8_t DC_I2C_SOFT_RESET; - uint8_t DC_I2C_SW_STATUS_RESET; - uint8_t DC_I2C_GO; - uint8_t DC_I2C_SEND_RESET; - uint8_t DC_I2C_TRANSACTION_COUNT; - uint8_t DC_I2C_DDC_SELECT; - uint8_t DC_I2C_DDC1_PRESCALE; - uint8_t DC_I2C_DDC1_THRESHOLD; - uint8_t DC_I2C_DDC1_START_STOP_TIMING_CNTL; - uint8_t DC_I2C_SW_STOPPED_ON_NACK; - uint8_t DC_I2C_SW_TIMEOUT; - uint8_t DC_I2C_SW_ABORTED; - uint8_t DC_I2C_SW_DONE; - uint8_t DC_I2C_SW_STATUS; - uint8_t DC_I2C_STOP_ON_NACK0; - uint8_t DC_I2C_START0; - uint8_t DC_I2C_RW0; - uint8_t DC_I2C_STOP0; - uint8_t DC_I2C_COUNT0; - uint8_t DC_I2C_DATA_RW; - uint8_t DC_I2C_DATA; - uint8_t DC_I2C_INDEX; - uint8_t DC_I2C_INDEX_WRITE; - uint8_t XTAL_REF_DIV; -}; - -struct dce110_i2c_hw_engine_mask { - uint32_t DC_I2C_DDC1_ENABLE; - uint32_t DC_I2C_DDC1_TIME_LIMIT; - uint32_t DC_I2C_DDC1_DATA_DRIVE_EN; - uint32_t DC_I2C_DDC1_CLK_DRIVE_EN; - uint32_t DC_I2C_DDC1_DATA_DRIVE_SEL; - uint32_t DC_I2C_DDC1_INTRA_TRANSACTION_DELAY; - uint32_t DC_I2C_DDC1_INTRA_BYTE_DELAY; - uint32_t DC_I2C_SW_DONE_USING_I2C_REG; - uint32_t DC_I2C_NO_QUEUED_SW_GO; - uint32_t DC_I2C_SW_PRIORITY; - uint32_t DC_I2C_SOFT_RESET; - uint32_t DC_I2C_SW_STATUS_RESET; - uint32_t DC_I2C_GO; - uint32_t DC_I2C_SEND_RESET; - uint32_t DC_I2C_TRANSACTION_COUNT; - uint32_t DC_I2C_DDC_SELECT; - uint32_t DC_I2C_DDC1_PRESCALE; - uint32_t DC_I2C_DDC1_THRESHOLD; - uint32_t DC_I2C_DDC1_START_STOP_TIMING_CNTL; - uint32_t DC_I2C_SW_STOPPED_ON_NACK; - uint32_t DC_I2C_SW_TIMEOUT; - uint32_t DC_I2C_SW_ABORTED; - uint32_t DC_I2C_SW_DONE; - uint32_t DC_I2C_SW_STATUS; - uint32_t DC_I2C_STOP_ON_NACK0; - uint32_t DC_I2C_START0; - uint32_t DC_I2C_RW0; - uint32_t DC_I2C_STOP0; - uint32_t DC_I2C_COUNT0; - uint32_t DC_I2C_DATA_RW; - uint32_t DC_I2C_DATA; - uint32_t DC_I2C_INDEX; - uint32_t DC_I2C_INDEX_WRITE; - uint32_t XTAL_REF_DIV; -}; - -struct dce110_i2c_hw_engine_registers { - uint32_t SETUP; - uint32_t SPEED; - uint32_t DC_I2C_ARBITRATION; - uint32_t DC_I2C_CONTROL; - uint32_t DC_I2C_SW_STATUS; - uint32_t DC_I2C_TRANSACTION0; - uint32_t DC_I2C_TRANSACTION1; - uint32_t DC_I2C_TRANSACTION2; - uint32_t DC_I2C_TRANSACTION3; - uint32_t DC_I2C_DATA; - uint32_t MICROSECOND_TIME_BASE_DIV; -}; - -struct i2c_hw_engine_dce110 { - struct i2c_hw_engine base; - const struct dce110_i2c_hw_engine_registers *regs; - const struct dce110_i2c_hw_engine_shift *i2c_shift; - const struct dce110_i2c_hw_engine_mask *i2c_mask; - struct { - uint32_t DC_I2C_DDCX_SETUP; - uint32_t DC_I2C_DDCX_SPEED; - } addr; - uint32_t engine_id; - /* expressed in kilohertz */ - uint32_t reference_frequency; - /* number of bytes currently used in HW buffer */ - uint32_t buffer_used_bytes; - /* number of bytes used for write transaction in HW buffer - * - this will be used as the index to read from*/ - uint32_t buffer_used_write; - /* number of pending transactions (before GO) */ - uint32_t transaction_count; - uint32_t engine_keep_power_up_count; - uint32_t i2_setup_time_limit; -}; - -struct i2c_hw_engine_dce110_create_arg { - uint32_t engine_id; - uint32_t reference_frequency; - uint32_t default_speed; - struct dc_context *ctx; - const struct dce110_i2c_hw_engine_registers *regs; - const struct dce110_i2c_hw_engine_shift *i2c_shift; - const struct dce110_i2c_hw_engine_mask *i2c_mask; -}; - -struct i2c_engine *dal_i2c_hw_engine_dce110_create( - const struct i2c_hw_engine_dce110_create_arg *arg); - -enum { - I2C_SETUP_TIME_LIMIT_DCE = 255, - I2C_SETUP_TIME_LIMIT_DCN = 3, - I2C_HW_BUFFER_SIZE = 538, - I2C_SEND_RESET_LENGTH_9 = 9, - I2C_SEND_RESET_LENGTH_10 = 10, -}; -#endif diff --git a/drivers/gpu/drm/amd/display/dc/i2caux/dce110/i2c_sw_engine_dce110.c b/drivers/gpu/drm/amd/display/dc/i2caux/dce110/i2c_sw_engine_dce110.c deleted file mode 100644 index 3aa7f791e523..000000000000 --- a/drivers/gpu/drm/amd/display/dc/i2caux/dce110/i2c_sw_engine_dce110.c +++ /dev/null @@ -1,160 +0,0 @@ -/* - * Copyright 2012-15 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. - * - * Authors: AMD - * - */ - -#include "dm_services.h" - -/* - * Pre-requisites: headers required by header of this unit - */ -#include "include/i2caux_interface.h" -#include "../engine.h" -#include "../i2c_engine.h" -#include "../i2c_sw_engine.h" - -/* - * Header of this unit - */ - -#include "i2c_sw_engine_dce110.h" - -/* - * Post-requisites: headers required by this unit - */ - -/* - * This unit - */ - -/* - * @brief - * Cast 'struct i2c_sw_engine *' - * to 'struct i2c_sw_engine_dce110 *' - */ -#define FROM_I2C_SW_ENGINE(ptr) \ - container_of((ptr), struct i2c_sw_engine_dce110, base) -/* - * @brief - * Cast 'struct i2c_engine *' - * to 'struct i2c_sw_engine_dce80 *' - */ -#define FROM_I2C_ENGINE(ptr) \ - FROM_I2C_SW_ENGINE(container_of((ptr), struct i2c_sw_engine, base)) - -/* - * @brief - * Cast 'struct engine *' - * to 'struct i2c_sw_engine_dce80 *' - */ -#define FROM_ENGINE(ptr) \ - FROM_I2C_ENGINE(container_of((ptr), struct i2c_engine, base)) - -static void release_engine( - struct engine *engine) -{ -} - -static void destruct( - struct i2c_sw_engine_dce110 *engine) -{ - dal_i2c_sw_engine_destruct(&engine->base); -} - -static void destroy( - struct i2c_engine **engine) -{ - struct i2c_sw_engine_dce110 *sw_engine = FROM_I2C_ENGINE(*engine); - - destruct(sw_engine); - - kfree(sw_engine); - - *engine = NULL; -} - -static bool acquire_engine( - struct i2c_engine *engine, - struct ddc *ddc_handle) -{ - return dal_i2caux_i2c_sw_engine_acquire_engine(engine, ddc_handle); -} - -static const struct i2c_engine_funcs i2c_engine_funcs = { - .acquire_engine = acquire_engine, - .destroy = destroy, - .get_speed = dal_i2c_sw_engine_get_speed, - .set_speed = dal_i2c_sw_engine_set_speed, - .setup_engine = dal_i2c_engine_setup_i2c_engine, - .submit_channel_request = dal_i2c_sw_engine_submit_channel_request, - .process_channel_reply = dal_i2c_engine_process_channel_reply, - .get_channel_status = dal_i2c_sw_engine_get_channel_status, -}; - -static const struct engine_funcs engine_funcs = { - .release_engine = release_engine, - .get_engine_type = dal_i2c_sw_engine_get_engine_type, - .acquire = dal_i2c_engine_acquire, - .submit_request = dal_i2c_sw_engine_submit_request, -}; - -static void construct( - struct i2c_sw_engine_dce110 *engine_dce110, - const struct i2c_sw_engine_dce110_create_arg *arg_dce110) -{ - struct i2c_sw_engine_create_arg arg_base; - - arg_base.ctx = arg_dce110->ctx; - arg_base.default_speed = arg_dce110->default_speed; - - dal_i2c_sw_engine_construct(&engine_dce110->base, &arg_base); - - /*struct engine struct engine_funcs*/ - engine_dce110->base.base.base.funcs = &engine_funcs; - /*struct i2c_engine struct i2c_engine_funcs*/ - engine_dce110->base.base.funcs = &i2c_engine_funcs; - engine_dce110->base.default_speed = arg_dce110->default_speed; - engine_dce110->engine_id = arg_dce110->engine_id; -} - -struct i2c_engine *dal_i2c_sw_engine_dce110_create( - const struct i2c_sw_engine_dce110_create_arg *arg) -{ - struct i2c_sw_engine_dce110 *engine_dce110; - - if (!arg) { - ASSERT_CRITICAL(false); - return NULL; - } - - engine_dce110 = kzalloc(sizeof(struct i2c_sw_engine_dce110), - GFP_KERNEL); - - if (!engine_dce110) { - ASSERT_CRITICAL(false); - return NULL; - } - - construct(engine_dce110, arg); - return &engine_dce110->base.base; -} diff --git a/drivers/gpu/drm/amd/display/dc/i2caux/dce110/i2caux_dce110.c b/drivers/gpu/drm/amd/display/dc/i2caux/dce110/i2caux_dce110.c deleted file mode 100644 index 1d748ac1d6d6..000000000000 --- a/drivers/gpu/drm/amd/display/dc/i2caux/dce110/i2caux_dce110.c +++ /dev/null @@ -1,329 +0,0 @@ -/* - * Copyright 2012-15 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. - * - * Authors: AMD - * - */ - -#include "dm_services.h" - -/* - * Pre-requisites: headers required by header of this unit - */ -#include "include/i2caux_interface.h" -#include "../i2caux.h" -#include "../engine.h" -#include "../i2c_engine.h" -#include "../i2c_sw_engine.h" -#include "../i2c_hw_engine.h" - -/* - * Header of this unit - */ -#include "i2caux_dce110.h" - -#include "i2c_sw_engine_dce110.h" -#include "i2c_hw_engine_dce110.h" -#include "aux_engine_dce110.h" -#include "../../dc.h" -#include "dc_types.h" - - -/* - * Post-requisites: headers required by this unit - */ - -/* - * This unit - */ -/*cast pointer to struct i2caux TO pointer to struct i2caux_dce110*/ -#define FROM_I2C_AUX(ptr) \ - container_of((ptr), struct i2caux_dce110, base) - -static void destruct( - struct i2caux_dce110 *i2caux_dce110) -{ - dal_i2caux_destruct(&i2caux_dce110->base); -} - -static void destroy( - struct i2caux **i2c_engine) -{ - struct i2caux_dce110 *i2caux_dce110 = FROM_I2C_AUX(*i2c_engine); - - destruct(i2caux_dce110); - - kfree(i2caux_dce110); - - *i2c_engine = NULL; -} - -static struct i2c_engine *acquire_i2c_hw_engine( - struct i2caux *i2caux, - struct ddc *ddc) -{ - struct i2caux_dce110 *i2caux_dce110 = FROM_I2C_AUX(i2caux); - - struct i2c_engine *engine = NULL; - /* generic hw engine is not used for EDID read - * It may be needed for external i2c device, like thermal chip, - * TODO will be implemented when needed. - * check dce80 bool non_generic for generic hw engine; - */ - - if (!ddc) - return NULL; - - if (ddc->hw_info.hw_supported) { - enum gpio_ddc_line line = dal_ddc_get_line(ddc); - - if (line < GPIO_DDC_LINE_COUNT) - engine = i2caux->i2c_hw_engines[line]; - } - - if (!engine) - return NULL; - - if (!i2caux_dce110->i2c_hw_buffer_in_use && - engine->base.funcs->acquire(&engine->base, ddc)) { - i2caux_dce110->i2c_hw_buffer_in_use = true; - return engine; - } - - return NULL; -} - -static void release_engine( - struct i2caux *i2caux, - struct engine *engine) -{ - struct i2caux_dce110 *i2caux_dce110 = FROM_I2C_AUX(i2caux); - - if (engine->funcs->get_engine_type(engine) == - I2CAUX_ENGINE_TYPE_I2C_DDC_HW) - i2caux_dce110->i2c_hw_buffer_in_use = false; - - dal_i2caux_release_engine(i2caux, engine); -} - -static const enum gpio_ddc_line hw_ddc_lines[] = { - GPIO_DDC_LINE_DDC1, - GPIO_DDC_LINE_DDC2, - GPIO_DDC_LINE_DDC3, - GPIO_DDC_LINE_DDC4, - GPIO_DDC_LINE_DDC5, - GPIO_DDC_LINE_DDC6, -}; - -static const enum gpio_ddc_line hw_aux_lines[] = { - GPIO_DDC_LINE_DDC1, - GPIO_DDC_LINE_DDC2, - GPIO_DDC_LINE_DDC3, - GPIO_DDC_LINE_DDC4, - GPIO_DDC_LINE_DDC5, - GPIO_DDC_LINE_DDC6, -}; - -/* function table */ -static const struct i2caux_funcs i2caux_funcs = { - .destroy = destroy, - .acquire_i2c_hw_engine = acquire_i2c_hw_engine, - .release_engine = release_engine, - .acquire_i2c_sw_engine = dal_i2caux_acquire_i2c_sw_engine, - .acquire_aux_engine = dal_i2caux_acquire_aux_engine, -}; - -#include "dce/dce_11_0_d.h" -#include "dce/dce_11_0_sh_mask.h" - -/* set register offset */ -#define SR(reg_name)\ - .reg_name = mm ## reg_name - -/* set register offset with instance */ -#define SRI(reg_name, block, id)\ - .reg_name = mm ## block ## id ## _ ## reg_name - -#define aux_regs(id)\ -[id] = {\ - AUX_COMMON_REG_LIST(id), \ - .AUX_RESET_MASK = AUX_CONTROL__AUX_RESET_MASK \ -} - -#define hw_engine_regs(id)\ -{\ - I2C_HW_ENGINE_COMMON_REG_LIST(id) \ -} - -static const struct dce110_aux_registers dce110_aux_regs[] = { - aux_regs(0), - aux_regs(1), - aux_regs(2), - aux_regs(3), - aux_regs(4), - aux_regs(5) -}; - -static const struct dce110_i2c_hw_engine_registers i2c_hw_engine_regs[] = { - hw_engine_regs(1), - hw_engine_regs(2), - hw_engine_regs(3), - hw_engine_regs(4), - hw_engine_regs(5), - hw_engine_regs(6) -}; - -static const struct dce110_i2c_hw_engine_shift i2c_shift = { - I2C_COMMON_MASK_SH_LIST_DCE110(__SHIFT) -}; - -static const struct dce110_i2c_hw_engine_mask i2c_mask = { - I2C_COMMON_MASK_SH_LIST_DCE110(_MASK) -}; - -void dal_i2caux_dce110_construct( - struct i2caux_dce110 *i2caux_dce110, - struct dc_context *ctx, - unsigned int num_i2caux_inst, - const struct dce110_aux_registers aux_regs[], - const struct dce110_i2c_hw_engine_registers i2c_hw_engine_regs[], - const struct dce110_i2c_hw_engine_shift *i2c_shift, - const struct dce110_i2c_hw_engine_mask *i2c_mask) -{ - uint32_t i = 0; - uint32_t reference_frequency = 0; - bool use_i2c_sw_engine = false; - struct i2caux *base = NULL; - /*TODO: For CZ bring up, if dal_i2caux_get_reference_clock - * does not return 48KHz, we need hard coded for 48Khz. - * Some BIOS setting incorrect cause this - * For production, we always get value from BIOS*/ - reference_frequency = - dal_i2caux_get_reference_clock(ctx->dc_bios) >> 1; - - base = &i2caux_dce110->base; - - dal_i2caux_construct(base, ctx); - - i2caux_dce110->base.funcs = &i2caux_funcs; - i2caux_dce110->i2c_hw_buffer_in_use = false; - /* Create I2C engines (DDC lines per connector) - * different I2C/AUX usage cases, DDC, Generic GPIO, AUX. - */ - do { - enum gpio_ddc_line line_id = hw_ddc_lines[i]; - - struct i2c_hw_engine_dce110_create_arg hw_arg_dce110; - - if (use_i2c_sw_engine) { - struct i2c_sw_engine_dce110_create_arg sw_arg; - - sw_arg.engine_id = i; - sw_arg.default_speed = base->default_i2c_sw_speed; - sw_arg.ctx = ctx; - base->i2c_sw_engines[line_id] = - dal_i2c_sw_engine_dce110_create(&sw_arg); - } - - hw_arg_dce110.engine_id = i; - hw_arg_dce110.reference_frequency = reference_frequency; - hw_arg_dce110.default_speed = base->default_i2c_hw_speed; - hw_arg_dce110.ctx = ctx; - hw_arg_dce110.regs = &i2c_hw_engine_regs[i]; - hw_arg_dce110.i2c_shift = i2c_shift; - hw_arg_dce110.i2c_mask = i2c_mask; - - base->i2c_hw_engines[line_id] = - dal_i2c_hw_engine_dce110_create(&hw_arg_dce110); - if (base->i2c_hw_engines[line_id] != NULL) { - switch (ctx->dce_version) { - case DCN_VERSION_1_0: - base->i2c_hw_engines[line_id]->setup_limit = - I2C_SETUP_TIME_LIMIT_DCN; - base->i2c_hw_engines[line_id]->send_reset_length = 0; - break; - default: - base->i2c_hw_engines[line_id]->setup_limit = - I2C_SETUP_TIME_LIMIT_DCE; - base->i2c_hw_engines[line_id]->send_reset_length = 0; - break; - } - } - ++i; - } while (i < num_i2caux_inst); - - /* Create AUX engines for all lines which has assisted HW AUX - * 'i' (loop counter) used as DDC/AUX engine_id */ - - i = 0; - - do { - enum gpio_ddc_line line_id = hw_aux_lines[i]; - - struct aux_engine_dce110_init_data aux_init_data; - - aux_init_data.engine_id = i; - aux_init_data.timeout_period = base->aux_timeout_period; - aux_init_data.ctx = ctx; - aux_init_data.regs = &aux_regs[i]; - - base->aux_engines[line_id] = - dal_aux_engine_dce110_create(&aux_init_data); - - ++i; - } while (i < num_i2caux_inst); - - /*TODO Generic I2C SW and HW*/ -} - -/* - * dal_i2caux_dce110_create - * - * @brief - * public interface to allocate memory for DCE11 I2CAUX - * - * @param - * struct adapter_service *as - [in] - * struct dc_context *ctx - [in] - * - * @return - * pointer to the base struct of DCE11 I2CAUX - */ -struct i2caux *dal_i2caux_dce110_create( - struct dc_context *ctx) -{ - struct i2caux_dce110 *i2caux_dce110 = - kzalloc(sizeof(struct i2caux_dce110), GFP_KERNEL); - - if (!i2caux_dce110) { - ASSERT_CRITICAL(false); - return NULL; - } - - dal_i2caux_dce110_construct(i2caux_dce110, - ctx, - ARRAY_SIZE(dce110_aux_regs), - dce110_aux_regs, - i2c_hw_engine_regs, - &i2c_shift, - &i2c_mask); - return &i2caux_dce110->base; -} diff --git a/drivers/gpu/drm/amd/display/dc/i2caux/dce110/i2caux_dce110.h b/drivers/gpu/drm/amd/display/dc/i2caux/dce110/i2caux_dce110.h deleted file mode 100644 index d3d8cc58666a..000000000000 --- a/drivers/gpu/drm/amd/display/dc/i2caux/dce110/i2caux_dce110.h +++ /dev/null @@ -1,54 +0,0 @@ -/* - * Copyright 2012-15 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. - * - * Authors: AMD - * - */ - -#ifndef __DAL_I2C_AUX_DCE110_H__ -#define __DAL_I2C_AUX_DCE110_H__ - -#include "../i2caux.h" - -struct i2caux_dce110 { - struct i2caux base; - /* indicate the I2C HW circular buffer is in use */ - bool i2c_hw_buffer_in_use; -}; - -struct dce110_aux_registers; -struct dce110_i2c_hw_engine_registers; -struct dce110_i2c_hw_engine_shift; -struct dce110_i2c_hw_engine_mask; - -struct i2caux *dal_i2caux_dce110_create( - struct dc_context *ctx); - -void dal_i2caux_dce110_construct( - struct i2caux_dce110 *i2caux_dce110, - struct dc_context *ctx, - unsigned int num_i2caux_inst, - const struct dce110_aux_registers *aux_regs, - const struct dce110_i2c_hw_engine_registers *i2c_hw_engine_regs, - const struct dce110_i2c_hw_engine_shift *i2c_shift, - const struct dce110_i2c_hw_engine_mask *i2c_mask); - -#endif /* __DAL_I2C_AUX_DCE110_H__ */ diff --git a/drivers/gpu/drm/amd/display/dc/i2caux/dce112/i2caux_dce112.c b/drivers/gpu/drm/amd/display/dc/i2caux/dce112/i2caux_dce112.c deleted file mode 100644 index a9db04738724..000000000000 --- a/drivers/gpu/drm/amd/display/dc/i2caux/dce112/i2caux_dce112.c +++ /dev/null @@ -1,129 +0,0 @@ -/* - * Copyright 2012-15 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. - * - * Authors: AMD - * - */ - -#include "dm_services.h" - -#include "include/i2caux_interface.h" -#include "../i2caux.h" -#include "../engine.h" -#include "../i2c_engine.h" -#include "../i2c_sw_engine.h" -#include "../i2c_hw_engine.h" - -#include "../dce110/i2caux_dce110.h" -#include "i2caux_dce112.h" - -#include "../dce110/aux_engine_dce110.h" - -#include "../dce110/i2c_hw_engine_dce110.h" - -#include "dce/dce_11_2_d.h" -#include "dce/dce_11_2_sh_mask.h" - -/* set register offset */ -#define SR(reg_name)\ - .reg_name = mm ## reg_name - -/* set register offset with instance */ -#define SRI(reg_name, block, id)\ - .reg_name = mm ## block ## id ## _ ## reg_name - -#define aux_regs(id)\ -[id] = {\ - AUX_COMMON_REG_LIST(id), \ - .AUX_RESET_MASK = AUX_CONTROL__AUX_RESET_MASK \ -} - -#define hw_engine_regs(id)\ -{\ - I2C_HW_ENGINE_COMMON_REG_LIST(id) \ -} - -static const struct dce110_aux_registers dce112_aux_regs[] = { - aux_regs(0), - aux_regs(1), - aux_regs(2), - aux_regs(3), - aux_regs(4), - aux_regs(5), -}; - -static const struct dce110_i2c_hw_engine_registers dce112_hw_engine_regs[] = { - hw_engine_regs(1), - hw_engine_regs(2), - hw_engine_regs(3), - hw_engine_regs(4), - hw_engine_regs(5), - hw_engine_regs(6) -}; - -static const struct dce110_i2c_hw_engine_shift i2c_shift = { - I2C_COMMON_MASK_SH_LIST_DCE110(__SHIFT) -}; - -static const struct dce110_i2c_hw_engine_mask i2c_mask = { - I2C_COMMON_MASK_SH_LIST_DCE110(_MASK) -}; - -static void construct( - struct i2caux_dce110 *i2caux_dce110, - struct dc_context *ctx) -{ - dal_i2caux_dce110_construct(i2caux_dce110, - ctx, - ARRAY_SIZE(dce112_aux_regs), - dce112_aux_regs, - dce112_hw_engine_regs, - &i2c_shift, - &i2c_mask); -} - -/* - * dal_i2caux_dce110_create - * - * @brief - * public interface to allocate memory for DCE11 I2CAUX - * - * @param - * struct adapter_service *as - [in] - * struct dc_context *ctx - [in] - * - * @return - * pointer to the base struct of DCE11 I2CAUX - */ -struct i2caux *dal_i2caux_dce112_create( - struct dc_context *ctx) -{ - struct i2caux_dce110 *i2caux_dce110 = - kzalloc(sizeof(struct i2caux_dce110), GFP_KERNEL); - - if (!i2caux_dce110) { - ASSERT_CRITICAL(false); - return NULL; - } - - construct(i2caux_dce110, ctx); - return &i2caux_dce110->base; -} diff --git a/drivers/gpu/drm/amd/display/dc/i2caux/dce120/i2caux_dce120.c b/drivers/gpu/drm/amd/display/dc/i2caux/dce120/i2caux_dce120.c deleted file mode 100644 index 6a4f344c1db4..000000000000 --- a/drivers/gpu/drm/amd/display/dc/i2caux/dce120/i2caux_dce120.c +++ /dev/null @@ -1,120 +0,0 @@ -/* - * Copyright 2012-16 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. - * - * Authors: AMD - * - */ - -#include "dm_services.h" - -#include "include/i2caux_interface.h" -#include "../i2caux.h" -#include "../engine.h" -#include "../i2c_engine.h" -#include "../i2c_sw_engine.h" -#include "../i2c_hw_engine.h" - -#include "../dce110/i2c_hw_engine_dce110.h" -#include "../dce110/aux_engine_dce110.h" -#include "../dce110/i2caux_dce110.h" - -#include "dce/dce_12_0_offset.h" -#include "dce/dce_12_0_sh_mask.h" -#include "soc15_hw_ip.h" -#include "vega10_ip_offset.h" - -/* begin ********************* - * macros to expend register list macro defined in HW object header file */ - -#define BASE_INNER(seg) \ - DCE_BASE__INST0_SEG ## seg - -/* compile time expand base address. */ -#define BASE(seg) \ - BASE_INNER(seg) - -#define SR(reg_name)\ - .reg_name = BASE(mm ## reg_name ## _BASE_IDX) + \ - mm ## reg_name - -#define SRI(reg_name, block, id)\ - .reg_name = BASE(mm ## block ## id ## _ ## reg_name ## _BASE_IDX) + \ - mm ## block ## id ## _ ## reg_name -/* macros to expend register list macro defined in HW object header file - * end *********************/ - -#define aux_regs(id)\ -[id] = {\ - AUX_COMMON_REG_LIST(id), \ - .AUX_RESET_MASK = DP_AUX0_AUX_CONTROL__AUX_RESET_MASK \ -} - -static const struct dce110_aux_registers dce120_aux_regs[] = { - aux_regs(0), - aux_regs(1), - aux_regs(2), - aux_regs(3), - aux_regs(4), - aux_regs(5), -}; - -#define hw_engine_regs(id)\ -{\ - I2C_HW_ENGINE_COMMON_REG_LIST(id) \ -} - -static const struct dce110_i2c_hw_engine_registers dce120_hw_engine_regs[] = { - hw_engine_regs(1), - hw_engine_regs(2), - hw_engine_regs(3), - hw_engine_regs(4), - hw_engine_regs(5), - hw_engine_regs(6) -}; - -static const struct dce110_i2c_hw_engine_shift i2c_shift = { - I2C_COMMON_MASK_SH_LIST_DCE110(__SHIFT) -}; - -static const struct dce110_i2c_hw_engine_mask i2c_mask = { - I2C_COMMON_MASK_SH_LIST_DCE110(_MASK) -}; - -struct i2caux *dal_i2caux_dce120_create( - struct dc_context *ctx) -{ - struct i2caux_dce110 *i2caux_dce110 = - kzalloc(sizeof(struct i2caux_dce110), GFP_KERNEL); - - if (!i2caux_dce110) { - ASSERT_CRITICAL(false); - return NULL; - } - - dal_i2caux_dce110_construct(i2caux_dce110, - ctx, - ARRAY_SIZE(dce120_aux_regs), - dce120_aux_regs, - dce120_hw_engine_regs, - &i2c_shift, - &i2c_mask); - return &i2caux_dce110->base; -} diff --git a/drivers/gpu/drm/amd/display/dc/i2caux/dce80/i2c_hw_engine_dce80.c b/drivers/gpu/drm/amd/display/dc/i2caux/dce80/i2c_hw_engine_dce80.c deleted file mode 100644 index fd0832dd2c75..000000000000 --- a/drivers/gpu/drm/amd/display/dc/i2caux/dce80/i2c_hw_engine_dce80.c +++ /dev/null @@ -1,875 +0,0 @@ -/* - * Copyright 2012-15 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. - * - * Authors: AMD - * - */ - -#include "dm_services.h" - -/* - * Pre-requisites: headers required by header of this unit - */ -#include "include/i2caux_interface.h" -#include "../engine.h" -#include "../i2c_engine.h" -#include "../i2c_hw_engine.h" -#include "../i2c_generic_hw_engine.h" -/* - * Header of this unit - */ - -#include "i2c_hw_engine_dce80.h" - -/* - * Post-requisites: headers required by this unit - */ - -#include "dce/dce_8_0_d.h" -#include "dce/dce_8_0_sh_mask.h" -/* - * This unit - */ - -enum dc_i2c_status { - DC_I2C_STATUS__DC_I2C_STATUS_IDLE, - DC_I2C_STATUS__DC_I2C_STATUS_USED_BY_SW, - DC_I2C_STATUS__DC_I2C_STATUS_USED_BY_HW -}; - -enum dc_i2c_arbitration { - DC_I2C_ARBITRATION__DC_I2C_SW_PRIORITY_NORMAL, - DC_I2C_ARBITRATION__DC_I2C_SW_PRIORITY_HIGH -}; - -enum { - /* No timeout in HW - * (timeout implemented in SW by querying status) */ - I2C_SETUP_TIME_LIMIT = 255, - I2C_HW_BUFFER_SIZE = 144 -}; - -/* - * @brief - * Cast 'struct i2c_hw_engine *' - * to 'struct i2c_hw_engine_dce80 *' - */ -#define FROM_I2C_HW_ENGINE(ptr) \ - container_of((ptr), struct i2c_hw_engine_dce80, base) - -/* - * @brief - * Cast pointer to 'struct i2c_engine *' - * to pointer to 'struct i2c_hw_engine_dce80 *' - */ -#define FROM_I2C_ENGINE(ptr) \ - FROM_I2C_HW_ENGINE(container_of((ptr), struct i2c_hw_engine, base)) - -/* - * @brief - * Cast pointer to 'struct engine *' - * to 'pointer to struct i2c_hw_engine_dce80 *' - */ -#define FROM_ENGINE(ptr) \ - FROM_I2C_ENGINE(container_of((ptr), struct i2c_engine, base)) - -static void disable_i2c_hw_engine( - struct i2c_hw_engine_dce80 *engine) -{ - const uint32_t addr = engine->addr.DC_I2C_DDCX_SETUP; - uint32_t value = 0; - - struct dc_context *ctx = NULL; - - ctx = engine->base.base.base.ctx; - - value = dm_read_reg(ctx, addr); - - set_reg_field_value( - value, - 0, - DC_I2C_DDC1_SETUP, - DC_I2C_DDC1_ENABLE); - - dm_write_reg(ctx, addr, value); -} - -static void release_engine( - struct engine *engine) -{ - struct i2c_hw_engine_dce80 *hw_engine = FROM_ENGINE(engine); - - struct i2c_engine *base = NULL; - bool safe_to_reset; - uint32_t value = 0; - - base = &hw_engine->base.base; - - /* Restore original HW engine speed */ - - base->funcs->set_speed(base, hw_engine->base.original_speed); - - /* Release I2C */ - { - value = dm_read_reg(engine->ctx, mmDC_I2C_ARBITRATION); - - set_reg_field_value( - value, - 1, - DC_I2C_ARBITRATION, - DC_I2C_SW_DONE_USING_I2C_REG); - - dm_write_reg(engine->ctx, mmDC_I2C_ARBITRATION, value); - } - - /* Reset HW engine */ - { - uint32_t i2c_sw_status = 0; - - value = dm_read_reg(engine->ctx, mmDC_I2C_SW_STATUS); - - i2c_sw_status = get_reg_field_value( - value, - DC_I2C_SW_STATUS, - DC_I2C_SW_STATUS); - /* if used by SW, safe to reset */ - safe_to_reset = (i2c_sw_status == 1); - } - { - value = dm_read_reg(engine->ctx, mmDC_I2C_CONTROL); - - if (safe_to_reset) - set_reg_field_value( - value, - 1, - DC_I2C_CONTROL, - DC_I2C_SOFT_RESET); - - set_reg_field_value( - value, - 1, - DC_I2C_CONTROL, - DC_I2C_SW_STATUS_RESET); - - dm_write_reg(engine->ctx, mmDC_I2C_CONTROL, value); - } - - /* HW I2c engine - clock gating feature */ - if (!hw_engine->engine_keep_power_up_count) - disable_i2c_hw_engine(hw_engine); -} - -static void destruct( - struct i2c_hw_engine_dce80 *engine) -{ - dal_i2c_hw_engine_destruct(&engine->base); -} - -static void destroy( - struct i2c_engine **i2c_engine) -{ - struct i2c_hw_engine_dce80 *engine = FROM_I2C_ENGINE(*i2c_engine); - - destruct(engine); - - kfree(engine); - - *i2c_engine = NULL; -} - -static bool setup_engine( - struct i2c_engine *i2c_engine) -{ - uint32_t value = 0; - struct i2c_hw_engine_dce80 *engine = FROM_I2C_ENGINE(i2c_engine); - - /* Program pin select */ - { - const uint32_t addr = mmDC_I2C_CONTROL; - - value = dm_read_reg(i2c_engine->base.ctx, addr); - - set_reg_field_value( - value, - 0, - DC_I2C_CONTROL, - DC_I2C_GO); - - set_reg_field_value( - value, - 0, - DC_I2C_CONTROL, - DC_I2C_SOFT_RESET); - - set_reg_field_value( - value, - 0, - DC_I2C_CONTROL, - DC_I2C_SEND_RESET); - - set_reg_field_value( - value, - 0, - DC_I2C_CONTROL, - DC_I2C_SW_STATUS_RESET); - - set_reg_field_value( - value, - 0, - DC_I2C_CONTROL, - DC_I2C_TRANSACTION_COUNT); - - set_reg_field_value( - value, - engine->engine_id, - DC_I2C_CONTROL, - DC_I2C_DDC_SELECT); - - dm_write_reg(i2c_engine->base.ctx, addr, value); - } - - /* Program time limit */ - { - const uint32_t addr = engine->addr.DC_I2C_DDCX_SETUP; - - value = dm_read_reg(i2c_engine->base.ctx, addr); - - set_reg_field_value( - value, - I2C_SETUP_TIME_LIMIT, - DC_I2C_DDC1_SETUP, - DC_I2C_DDC1_TIME_LIMIT); - - set_reg_field_value( - value, - 1, - DC_I2C_DDC1_SETUP, - DC_I2C_DDC1_ENABLE); - - dm_write_reg(i2c_engine->base.ctx, addr, value); - } - - /* Program HW priority - * set to High - interrupt software I2C at any time - * Enable restart of SW I2C that was interrupted by HW - * disable queuing of software while I2C is in use by HW */ - { - value = dm_read_reg(i2c_engine->base.ctx, - mmDC_I2C_ARBITRATION); - - set_reg_field_value( - value, - 0, - DC_I2C_ARBITRATION, - DC_I2C_NO_QUEUED_SW_GO); - - set_reg_field_value( - value, - DC_I2C_ARBITRATION__DC_I2C_SW_PRIORITY_NORMAL, - DC_I2C_ARBITRATION, - DC_I2C_SW_PRIORITY); - - dm_write_reg(i2c_engine->base.ctx, - mmDC_I2C_ARBITRATION, value); - } - - return true; -} - -static uint32_t get_speed( - const struct i2c_engine *i2c_engine) -{ - const struct i2c_hw_engine_dce80 *engine = FROM_I2C_ENGINE(i2c_engine); - - const uint32_t addr = engine->addr.DC_I2C_DDCX_SPEED; - - uint32_t pre_scale = 0; - - uint32_t value = dm_read_reg(i2c_engine->base.ctx, addr); - - pre_scale = get_reg_field_value( - value, - DC_I2C_DDC1_SPEED, - DC_I2C_DDC1_PRESCALE); - - /* [anaumov] it seems following is unnecessary */ - /*ASSERT(value.bits.DC_I2C_DDC1_PRESCALE);*/ - - return pre_scale ? - engine->reference_frequency / pre_scale : - engine->base.default_speed; -} - -static void set_speed( - struct i2c_engine *i2c_engine, - uint32_t speed) -{ - struct i2c_hw_engine_dce80 *engine = FROM_I2C_ENGINE(i2c_engine); - - if (speed) { - const uint32_t addr = engine->addr.DC_I2C_DDCX_SPEED; - - uint32_t value = dm_read_reg(i2c_engine->base.ctx, addr); - - set_reg_field_value( - value, - engine->reference_frequency / speed, - DC_I2C_DDC1_SPEED, - DC_I2C_DDC1_PRESCALE); - - set_reg_field_value( - value, - 2, - DC_I2C_DDC1_SPEED, - DC_I2C_DDC1_THRESHOLD); - - dm_write_reg(i2c_engine->base.ctx, addr, value); - } -} - -static inline void reset_hw_engine(struct engine *engine) -{ - uint32_t value = dm_read_reg(engine->ctx, mmDC_I2C_CONTROL); - - set_reg_field_value( - value, - 1, - DC_I2C_CONTROL, - DC_I2C_SOFT_RESET); - - set_reg_field_value( - value, - 1, - DC_I2C_CONTROL, - DC_I2C_SW_STATUS_RESET); - - dm_write_reg(engine->ctx, mmDC_I2C_CONTROL, value); -} - -static bool is_hw_busy(struct engine *engine) -{ - uint32_t i2c_sw_status = 0; - - uint32_t value = dm_read_reg(engine->ctx, mmDC_I2C_SW_STATUS); - - i2c_sw_status = get_reg_field_value( - value, - DC_I2C_SW_STATUS, - DC_I2C_SW_STATUS); - - if (i2c_sw_status == DC_I2C_STATUS__DC_I2C_STATUS_IDLE) - return false; - - reset_hw_engine(engine); - - value = dm_read_reg(engine->ctx, mmDC_I2C_SW_STATUS); - - i2c_sw_status = get_reg_field_value( - value, - DC_I2C_SW_STATUS, - DC_I2C_SW_STATUS); - - return i2c_sw_status != DC_I2C_STATUS__DC_I2C_STATUS_IDLE; -} - -/* - * @brief - * DC_GPIO_DDC MM register offsets - */ -static const uint32_t transaction_addr[] = { - mmDC_I2C_TRANSACTION0, - mmDC_I2C_TRANSACTION1, - mmDC_I2C_TRANSACTION2, - mmDC_I2C_TRANSACTION3 -}; - -static bool process_transaction( - struct i2c_hw_engine_dce80 *engine, - struct i2c_request_transaction_data *request) -{ - uint32_t length = request->length; - uint8_t *buffer = request->data; - - bool last_transaction = false; - uint32_t value = 0; - - struct dc_context *ctx = NULL; - - ctx = engine->base.base.base.ctx; - - { - const uint32_t addr = - transaction_addr[engine->transaction_count]; - - value = dm_read_reg(ctx, addr); - - set_reg_field_value( - value, - 1, - DC_I2C_TRANSACTION0, - DC_I2C_STOP_ON_NACK0); - - set_reg_field_value( - value, - 1, - DC_I2C_TRANSACTION0, - DC_I2C_START0); - - if ((engine->transaction_count == 3) || - (request->action == I2CAUX_TRANSACTION_ACTION_I2C_WRITE) || - (request->action & I2CAUX_TRANSACTION_ACTION_I2C_READ)) { - - set_reg_field_value( - value, - 1, - DC_I2C_TRANSACTION0, - DC_I2C_STOP0); - - last_transaction = true; - } else - set_reg_field_value( - value, - 0, - DC_I2C_TRANSACTION0, - DC_I2C_STOP0); - - set_reg_field_value( - value, - (0 != (request->action & - I2CAUX_TRANSACTION_ACTION_I2C_READ)), - DC_I2C_TRANSACTION0, - DC_I2C_RW0); - - set_reg_field_value( - value, - length, - DC_I2C_TRANSACTION0, - DC_I2C_COUNT0); - - dm_write_reg(ctx, addr, value); - } - - /* Write the I2C address and I2C data - * into the hardware circular buffer, one byte per entry. - * As an example, the 7-bit I2C slave address for CRT monitor - * for reading DDC/EDID information is 0b1010001. - * For an I2C send operation, the LSB must be programmed to 0; - * for I2C receive operation, the LSB must be programmed to 1. */ - - { - value = 0; - - set_reg_field_value( - value, - false, - DC_I2C_DATA, - DC_I2C_DATA_RW); - - set_reg_field_value( - value, - request->address, - DC_I2C_DATA, - DC_I2C_DATA); - - if (engine->transaction_count == 0) { - set_reg_field_value( - value, - 0, - DC_I2C_DATA, - DC_I2C_INDEX); - - /*enable index write*/ - set_reg_field_value( - value, - 1, - DC_I2C_DATA, - DC_I2C_INDEX_WRITE); - } - - dm_write_reg(ctx, mmDC_I2C_DATA, value); - - if (!(request->action & I2CAUX_TRANSACTION_ACTION_I2C_READ)) { - - set_reg_field_value( - value, - 0, - DC_I2C_DATA, - DC_I2C_INDEX_WRITE); - - while (length) { - - set_reg_field_value( - value, - *buffer++, - DC_I2C_DATA, - DC_I2C_DATA); - - dm_write_reg(ctx, mmDC_I2C_DATA, value); - --length; - } - } - } - - ++engine->transaction_count; - engine->buffer_used_bytes += length + 1; - - return last_transaction; -} - -static void execute_transaction( - struct i2c_hw_engine_dce80 *engine) -{ - uint32_t value = 0; - struct dc_context *ctx = NULL; - - ctx = engine->base.base.base.ctx; - - { - const uint32_t addr = engine->addr.DC_I2C_DDCX_SETUP; - - value = dm_read_reg(ctx, addr); - - set_reg_field_value( - value, - 0, - DC_I2C_DDC1_SETUP, - DC_I2C_DDC1_DATA_DRIVE_EN); - - set_reg_field_value( - value, - 0, - DC_I2C_DDC1_SETUP, - DC_I2C_DDC1_CLK_DRIVE_EN); - - set_reg_field_value( - value, - 0, - DC_I2C_DDC1_SETUP, - DC_I2C_DDC1_DATA_DRIVE_SEL); - - set_reg_field_value( - value, - 0, - DC_I2C_DDC1_SETUP, - DC_I2C_DDC1_INTRA_TRANSACTION_DELAY); - - set_reg_field_value( - value, - 0, - DC_I2C_DDC1_SETUP, - DC_I2C_DDC1_INTRA_BYTE_DELAY); - - dm_write_reg(ctx, addr, value); - } - - { - const uint32_t addr = mmDC_I2C_CONTROL; - - value = dm_read_reg(ctx, addr); - - set_reg_field_value( - value, - 0, - DC_I2C_CONTROL, - DC_I2C_SOFT_RESET); - - set_reg_field_value( - value, - 0, - DC_I2C_CONTROL, - DC_I2C_SW_STATUS_RESET); - - set_reg_field_value( - value, - 0, - DC_I2C_CONTROL, - DC_I2C_SEND_RESET); - - set_reg_field_value( - value, - 0, - DC_I2C_CONTROL, - DC_I2C_GO); - - set_reg_field_value( - value, - engine->transaction_count - 1, - DC_I2C_CONTROL, - DC_I2C_TRANSACTION_COUNT); - - dm_write_reg(ctx, addr, value); - } - - /* start I2C transfer */ - { - const uint32_t addr = mmDC_I2C_CONTROL; - - value = dm_read_reg(ctx, addr); - - set_reg_field_value( - value, - 1, - DC_I2C_CONTROL, - DC_I2C_GO); - - dm_write_reg(ctx, addr, value); - } - - /* all transactions were executed and HW buffer became empty - * (even though it actually happens when status becomes DONE) */ - engine->transaction_count = 0; - engine->buffer_used_bytes = 0; -} - -static void submit_channel_request( - struct i2c_engine *engine, - struct i2c_request_transaction_data *request) -{ - request->status = I2C_CHANNEL_OPERATION_SUCCEEDED; - - if (!process_transaction(FROM_I2C_ENGINE(engine), request)) - return; - - if (is_hw_busy(&engine->base)) { - request->status = I2C_CHANNEL_OPERATION_ENGINE_BUSY; - return; - } - - execute_transaction(FROM_I2C_ENGINE(engine)); -} - -static void process_channel_reply( - struct i2c_engine *engine, - struct i2c_reply_transaction_data *reply) -{ - uint32_t length = reply->length; - uint8_t *buffer = reply->data; - - uint32_t value = 0; - - /*set index*/ - set_reg_field_value( - value, - length - 1, - DC_I2C_DATA, - DC_I2C_INDEX); - - set_reg_field_value( - value, - 1, - DC_I2C_DATA, - DC_I2C_DATA_RW); - - set_reg_field_value( - value, - 1, - DC_I2C_DATA, - DC_I2C_INDEX_WRITE); - - dm_write_reg(engine->base.ctx, mmDC_I2C_DATA, value); - - while (length) { - /* after reading the status, - * if the I2C operation executed successfully - * (i.e. DC_I2C_STATUS_DONE = 1) then the I2C controller - * should read data bytes from I2C circular data buffer */ - - value = dm_read_reg(engine->base.ctx, mmDC_I2C_DATA); - - *buffer++ = get_reg_field_value( - value, - DC_I2C_DATA, - DC_I2C_DATA); - - --length; - } -} - -static enum i2c_channel_operation_result get_channel_status( - struct i2c_engine *engine, - uint8_t *returned_bytes) -{ - uint32_t i2c_sw_status = 0; - uint32_t value = dm_read_reg(engine->base.ctx, mmDC_I2C_SW_STATUS); - - i2c_sw_status = get_reg_field_value( - value, - DC_I2C_SW_STATUS, - DC_I2C_SW_STATUS); - - if (i2c_sw_status == DC_I2C_STATUS__DC_I2C_STATUS_USED_BY_SW) - return I2C_CHANNEL_OPERATION_ENGINE_BUSY; - else if (value & DC_I2C_SW_STATUS__DC_I2C_SW_STOPPED_ON_NACK_MASK) - return I2C_CHANNEL_OPERATION_NO_RESPONSE; - else if (value & DC_I2C_SW_STATUS__DC_I2C_SW_TIMEOUT_MASK) - return I2C_CHANNEL_OPERATION_TIMEOUT; - else if (value & DC_I2C_SW_STATUS__DC_I2C_SW_ABORTED_MASK) - return I2C_CHANNEL_OPERATION_FAILED; - else if (value & DC_I2C_SW_STATUS__DC_I2C_SW_DONE_MASK) - return I2C_CHANNEL_OPERATION_SUCCEEDED; - - /* - * this is the case when HW used for communication, I2C_SW_STATUS - * could be zero - */ - return I2C_CHANNEL_OPERATION_SUCCEEDED; -} - -static uint32_t get_hw_buffer_available_size( - const struct i2c_hw_engine *engine) -{ - return I2C_HW_BUFFER_SIZE - - FROM_I2C_HW_ENGINE(engine)->buffer_used_bytes; -} - -static uint32_t get_transaction_timeout( - const struct i2c_hw_engine *engine, - uint32_t length) -{ - uint32_t speed = engine->base.funcs->get_speed(&engine->base); - - uint32_t period_timeout; - uint32_t num_of_clock_stretches; - - if (!speed) - return 0; - - period_timeout = (1000 * TRANSACTION_TIMEOUT_IN_I2C_CLOCKS) / speed; - - num_of_clock_stretches = 1 + (length << 3) + 1; - num_of_clock_stretches += - (FROM_I2C_HW_ENGINE(engine)->buffer_used_bytes << 3) + - (FROM_I2C_HW_ENGINE(engine)->transaction_count << 1); - - return period_timeout * num_of_clock_stretches; -} - -/* - * @brief - * DC_I2C_DDC1_SETUP MM register offsets - * - * @note - * The indices of this offset array are DDC engine IDs - */ -static const int32_t ddc_setup_offset[] = { - - mmDC_I2C_DDC1_SETUP - mmDC_I2C_DDC1_SETUP, /* DDC Engine 1 */ - mmDC_I2C_DDC2_SETUP - mmDC_I2C_DDC1_SETUP, /* DDC Engine 2 */ - mmDC_I2C_DDC3_SETUP - mmDC_I2C_DDC1_SETUP, /* DDC Engine 3 */ - mmDC_I2C_DDC4_SETUP - mmDC_I2C_DDC1_SETUP, /* DDC Engine 4 */ - mmDC_I2C_DDC5_SETUP - mmDC_I2C_DDC1_SETUP, /* DDC Engine 5 */ - mmDC_I2C_DDC6_SETUP - mmDC_I2C_DDC1_SETUP, /* DDC Engine 6 */ - mmDC_I2C_DDCVGA_SETUP - mmDC_I2C_DDC1_SETUP /* DDC Engine 7 */ -}; - -/* - * @brief - * DC_I2C_DDC1_SPEED MM register offsets - * - * @note - * The indices of this offset array are DDC engine IDs - */ -static const int32_t ddc_speed_offset[] = { - mmDC_I2C_DDC1_SPEED - mmDC_I2C_DDC1_SPEED, /* DDC Engine 1 */ - mmDC_I2C_DDC2_SPEED - mmDC_I2C_DDC1_SPEED, /* DDC Engine 2 */ - mmDC_I2C_DDC3_SPEED - mmDC_I2C_DDC1_SPEED, /* DDC Engine 3 */ - mmDC_I2C_DDC4_SPEED - mmDC_I2C_DDC1_SPEED, /* DDC Engine 4 */ - mmDC_I2C_DDC5_SPEED - mmDC_I2C_DDC1_SPEED, /* DDC Engine 5 */ - mmDC_I2C_DDC6_SPEED - mmDC_I2C_DDC1_SPEED, /* DDC Engine 6 */ - mmDC_I2C_DDCVGA_SPEED - mmDC_I2C_DDC1_SPEED /* DDC Engine 7 */ -}; - -static const struct i2c_engine_funcs i2c_engine_funcs = { - .destroy = destroy, - .get_speed = get_speed, - .set_speed = set_speed, - .setup_engine = setup_engine, - .submit_channel_request = submit_channel_request, - .process_channel_reply = process_channel_reply, - .get_channel_status = get_channel_status, - .acquire_engine = dal_i2c_hw_engine_acquire_engine, -}; - -static const struct engine_funcs engine_funcs = { - .release_engine = release_engine, - .get_engine_type = dal_i2c_hw_engine_get_engine_type, - .acquire = dal_i2c_engine_acquire, - .submit_request = dal_i2c_hw_engine_submit_request, -}; - -static const struct i2c_hw_engine_funcs i2c_hw_engine_funcs = { - .get_hw_buffer_available_size = - get_hw_buffer_available_size, - .get_transaction_timeout = - get_transaction_timeout, - .wait_on_operation_result = - dal_i2c_hw_engine_wait_on_operation_result, -}; - -static void construct( - struct i2c_hw_engine_dce80 *engine, - const struct i2c_hw_engine_dce80_create_arg *arg) -{ - dal_i2c_hw_engine_construct(&engine->base, arg->ctx); - - engine->base.base.base.funcs = &engine_funcs; - engine->base.base.funcs = &i2c_engine_funcs; - engine->base.funcs = &i2c_hw_engine_funcs; - engine->base.default_speed = arg->default_speed; - engine->addr.DC_I2C_DDCX_SETUP = - mmDC_I2C_DDC1_SETUP + ddc_setup_offset[arg->engine_id]; - engine->addr.DC_I2C_DDCX_SPEED = - mmDC_I2C_DDC1_SPEED + ddc_speed_offset[arg->engine_id]; - - engine->engine_id = arg->engine_id; - engine->reference_frequency = arg->reference_frequency; - engine->buffer_used_bytes = 0; - engine->transaction_count = 0; - engine->engine_keep_power_up_count = 1; -} - -struct i2c_engine *dal_i2c_hw_engine_dce80_create( - const struct i2c_hw_engine_dce80_create_arg *arg) -{ - struct i2c_hw_engine_dce80 *engine; - - if (!arg) { - BREAK_TO_DEBUGGER(); - return NULL; - } - - if ((arg->engine_id >= sizeof(ddc_setup_offset) / sizeof(int32_t)) || - (arg->engine_id >= sizeof(ddc_speed_offset) / sizeof(int32_t)) || - !arg->reference_frequency) { - BREAK_TO_DEBUGGER(); - return NULL; - } - - engine = kzalloc(sizeof(struct i2c_hw_engine_dce80), GFP_KERNEL); - - if (!engine) { - BREAK_TO_DEBUGGER(); - return NULL; - } - - construct(engine, arg); - return &engine->base.base; -} diff --git a/drivers/gpu/drm/amd/display/dc/i2caux/dce80/i2c_hw_engine_dce80.h b/drivers/gpu/drm/amd/display/dc/i2caux/dce80/i2c_hw_engine_dce80.h deleted file mode 100644 index 5c6116fb5479..000000000000 --- a/drivers/gpu/drm/amd/display/dc/i2caux/dce80/i2c_hw_engine_dce80.h +++ /dev/null @@ -1,54 +0,0 @@ -/* - * Copyright 2012-15 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. - * - * Authors: AMD - * - */ - -#ifndef __DAL_I2C_HW_ENGINE_DCE80_H__ -#define __DAL_I2C_HW_ENGINE_DCE80_H__ - -struct i2c_hw_engine_dce80 { - struct i2c_hw_engine base; - struct { - uint32_t DC_I2C_DDCX_SETUP; - uint32_t DC_I2C_DDCX_SPEED; - } addr; - uint32_t engine_id; - /* expressed in kilohertz */ - uint32_t reference_frequency; - /* number of bytes currently used in HW buffer */ - uint32_t buffer_used_bytes; - /* number of pending transactions (before GO) */ - uint32_t transaction_count; - uint32_t engine_keep_power_up_count; -}; - -struct i2c_hw_engine_dce80_create_arg { - uint32_t engine_id; - uint32_t reference_frequency; - uint32_t default_speed; - struct dc_context *ctx; -}; - -struct i2c_engine *dal_i2c_hw_engine_dce80_create( - const struct i2c_hw_engine_dce80_create_arg *arg); -#endif diff --git a/drivers/gpu/drm/amd/display/dc/i2caux/dce80/i2c_sw_engine_dce80.c b/drivers/gpu/drm/amd/display/dc/i2caux/dce80/i2c_sw_engine_dce80.c deleted file mode 100644 index 4853ee26096a..000000000000 --- a/drivers/gpu/drm/amd/display/dc/i2caux/dce80/i2c_sw_engine_dce80.c +++ /dev/null @@ -1,173 +0,0 @@ -/* - * Copyright 2012-15 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. - * - * Authors: AMD - * - */ - -#include "dm_services.h" - -/* - * Pre-requisites: headers required by header of this unit - */ -#include "include/i2caux_interface.h" -#include "../engine.h" -#include "../i2c_engine.h" -#include "../i2c_sw_engine.h" - -/* - * Header of this unit - */ - -#include "i2c_sw_engine_dce80.h" - -/* - * Post-requisites: headers required by this unit - */ - -#include "dce/dce_8_0_d.h" -#include "dce/dce_8_0_sh_mask.h" - -/* - * This unit - */ - -static const uint32_t ddc_hw_status_addr[] = { - mmDC_I2C_DDC1_HW_STATUS, - mmDC_I2C_DDC2_HW_STATUS, - mmDC_I2C_DDC3_HW_STATUS, - mmDC_I2C_DDC4_HW_STATUS, - mmDC_I2C_DDC5_HW_STATUS, - mmDC_I2C_DDC6_HW_STATUS, - mmDC_I2C_DDCVGA_HW_STATUS -}; - -/* - * @brief - * Cast 'struct i2c_sw_engine *' - * to 'struct i2c_sw_engine_dce80 *' - */ -#define FROM_I2C_SW_ENGINE(ptr) \ - container_of((ptr), struct i2c_sw_engine_dce80, base) - -/* - * @brief - * Cast 'struct i2c_engine *' - * to 'struct i2c_sw_engine_dce80 *' - */ -#define FROM_I2C_ENGINE(ptr) \ - FROM_I2C_SW_ENGINE(container_of((ptr), struct i2c_sw_engine, base)) - -/* - * @brief - * Cast 'struct engine *' - * to 'struct i2c_sw_engine_dce80 *' - */ -#define FROM_ENGINE(ptr) \ - FROM_I2C_ENGINE(container_of((ptr), struct i2c_engine, base)) - -static void release_engine( - struct engine *engine) -{ - -} - -static void destruct( - struct i2c_sw_engine_dce80 *engine) -{ - dal_i2c_sw_engine_destruct(&engine->base); -} - -static void destroy( - struct i2c_engine **engine) -{ - struct i2c_sw_engine_dce80 *sw_engine = FROM_I2C_ENGINE(*engine); - - destruct(sw_engine); - - kfree(sw_engine); - - *engine = NULL; -} - -static bool acquire_engine( - struct i2c_engine *engine, - struct ddc *ddc_handle) -{ - return dal_i2caux_i2c_sw_engine_acquire_engine(engine, ddc_handle); -} - -static const struct i2c_engine_funcs i2c_engine_funcs = { - .acquire_engine = acquire_engine, - .destroy = destroy, - .get_speed = dal_i2c_sw_engine_get_speed, - .set_speed = dal_i2c_sw_engine_set_speed, - .setup_engine = dal_i2c_engine_setup_i2c_engine, - .submit_channel_request = dal_i2c_sw_engine_submit_channel_request, - .process_channel_reply = dal_i2c_engine_process_channel_reply, - .get_channel_status = dal_i2c_sw_engine_get_channel_status, -}; - -static const struct engine_funcs engine_funcs = { - .release_engine = release_engine, - .get_engine_type = dal_i2c_sw_engine_get_engine_type, - .acquire = dal_i2c_engine_acquire, - .submit_request = dal_i2c_sw_engine_submit_request, -}; - -static void construct( - struct i2c_sw_engine_dce80 *engine, - const struct i2c_sw_engine_dce80_create_arg *arg) -{ - struct i2c_sw_engine_create_arg arg_base; - - arg_base.ctx = arg->ctx; - arg_base.default_speed = arg->default_speed; - - dal_i2c_sw_engine_construct(&engine->base, &arg_base); - - engine->base.base.base.funcs = &engine_funcs; - engine->base.base.funcs = &i2c_engine_funcs; - engine->base.default_speed = arg->default_speed; - engine->engine_id = arg->engine_id; -} - -struct i2c_engine *dal_i2c_sw_engine_dce80_create( - const struct i2c_sw_engine_dce80_create_arg *arg) -{ - struct i2c_sw_engine_dce80 *engine; - - if (!arg) { - BREAK_TO_DEBUGGER(); - return NULL; - } - - engine = kzalloc(sizeof(struct i2c_sw_engine_dce80), GFP_KERNEL); - - if (!engine) { - BREAK_TO_DEBUGGER(); - return NULL; - } - - construct(engine, arg); - return &engine->base.base; -} - diff --git a/drivers/gpu/drm/amd/display/dc/i2caux/dce80/i2caux_dce80.c b/drivers/gpu/drm/amd/display/dc/i2caux/dce80/i2caux_dce80.c deleted file mode 100644 index ed48596dd2a5..000000000000 --- a/drivers/gpu/drm/amd/display/dc/i2caux/dce80/i2caux_dce80.c +++ /dev/null @@ -1,284 +0,0 @@ -/* - * Copyright 2012-15 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. - * - * Authors: AMD - * - */ - -#include "dm_services.h" - -/* - * Pre-requisites: headers required by header of this unit - */ -#include "include/i2caux_interface.h" -#include "../i2caux.h" - -/* - * Header of this unit - */ - -#include "i2caux_dce80.h" - -/* - * Post-requisites: headers required by this unit - */ - -#include "../engine.h" -#include "../i2c_engine.h" -#include "../i2c_sw_engine.h" -#include "i2c_sw_engine_dce80.h" -#include "../i2c_hw_engine.h" -#include "i2c_hw_engine_dce80.h" -#include "../i2c_generic_hw_engine.h" -#include "../aux_engine.h" - - -#include "../dce110/aux_engine_dce110.h" -#include "../dce110/i2caux_dce110.h" - -#include "dce/dce_8_0_d.h" -#include "dce/dce_8_0_sh_mask.h" - - -/* set register offset */ -#define SR(reg_name)\ - .reg_name = mm ## reg_name - -/* set register offset with instance */ -#define SRI(reg_name, block, id)\ - .reg_name = mm ## block ## id ## _ ## reg_name - -#define aux_regs(id)\ -[id] = {\ - AUX_COMMON_REG_LIST(id), \ - .AUX_RESET_MASK = 0 \ -} - -static const struct dce110_aux_registers dce80_aux_regs[] = { - aux_regs(0), - aux_regs(1), - aux_regs(2), - aux_regs(3), - aux_regs(4), - aux_regs(5) -}; - -/* - * This unit - */ - -#define FROM_I2C_AUX(ptr) \ - container_of((ptr), struct i2caux_dce80, base) - -static void destruct( - struct i2caux_dce80 *i2caux_dce80) -{ - dal_i2caux_destruct(&i2caux_dce80->base); -} - -static void destroy( - struct i2caux **i2c_engine) -{ - struct i2caux_dce80 *i2caux_dce80 = FROM_I2C_AUX(*i2c_engine); - - destruct(i2caux_dce80); - - kfree(i2caux_dce80); - - *i2c_engine = NULL; -} - -static struct i2c_engine *acquire_i2c_hw_engine( - struct i2caux *i2caux, - struct ddc *ddc) -{ - struct i2caux_dce80 *i2caux_dce80 = FROM_I2C_AUX(i2caux); - - struct i2c_engine *engine = NULL; - bool non_generic; - - if (!ddc) - return NULL; - - if (ddc->hw_info.hw_supported) { - enum gpio_ddc_line line = dal_ddc_get_line(ddc); - - if (line < GPIO_DDC_LINE_COUNT) { - non_generic = true; - engine = i2caux->i2c_hw_engines[line]; - } - } - - if (!engine) { - non_generic = false; - engine = i2caux->i2c_generic_hw_engine; - } - - if (!engine) - return NULL; - - if (non_generic) { - if (!i2caux_dce80->i2c_hw_buffer_in_use && - engine->base.funcs->acquire(&engine->base, ddc)) { - i2caux_dce80->i2c_hw_buffer_in_use = true; - return engine; - } - } else { - if (engine->base.funcs->acquire(&engine->base, ddc)) - return engine; - } - - return NULL; -} - -static void release_engine( - struct i2caux *i2caux, - struct engine *engine) -{ - if (engine->funcs->get_engine_type(engine) == - I2CAUX_ENGINE_TYPE_I2C_DDC_HW) - FROM_I2C_AUX(i2caux)->i2c_hw_buffer_in_use = false; - - dal_i2caux_release_engine(i2caux, engine); -} - -static const enum gpio_ddc_line hw_ddc_lines[] = { - GPIO_DDC_LINE_DDC1, - GPIO_DDC_LINE_DDC2, - GPIO_DDC_LINE_DDC3, - GPIO_DDC_LINE_DDC4, - GPIO_DDC_LINE_DDC5, - GPIO_DDC_LINE_DDC6, - GPIO_DDC_LINE_DDC_VGA -}; - -static const enum gpio_ddc_line hw_aux_lines[] = { - GPIO_DDC_LINE_DDC1, - GPIO_DDC_LINE_DDC2, - GPIO_DDC_LINE_DDC3, - GPIO_DDC_LINE_DDC4, - GPIO_DDC_LINE_DDC5, - GPIO_DDC_LINE_DDC6 -}; - -static const struct i2caux_funcs i2caux_funcs = { - .destroy = destroy, - .acquire_i2c_hw_engine = acquire_i2c_hw_engine, - .release_engine = release_engine, - .acquire_i2c_sw_engine = dal_i2caux_acquire_i2c_sw_engine, - .acquire_aux_engine = dal_i2caux_acquire_aux_engine, -}; - -static void construct( - struct i2caux_dce80 *i2caux_dce80, - struct dc_context *ctx) -{ - /* Entire family have I2C engine reference clock frequency - * changed from XTALIN (27) to XTALIN/2 (13.5) */ - - struct i2caux *base = &i2caux_dce80->base; - - uint32_t reference_frequency = - dal_i2caux_get_reference_clock(ctx->dc_bios) >> 1; - - /*bool use_i2c_sw_engine = dal_adapter_service_is_feature_supported(as, - FEATURE_RESTORE_USAGE_I2C_SW_ENGINE);*/ - - /* Use SWI2C for dce8 currently, sicne we have bug with hwi2c */ - bool use_i2c_sw_engine = true; - - uint32_t i; - - dal_i2caux_construct(base, ctx); - - i2caux_dce80->base.funcs = &i2caux_funcs; - i2caux_dce80->i2c_hw_buffer_in_use = false; - - /* Create I2C HW engines (HW + SW pairs) - * for all lines which has assisted HW DDC - * 'i' (loop counter) used as DDC/AUX engine_id */ - - i = 0; - - do { - enum gpio_ddc_line line_id = hw_ddc_lines[i]; - - struct i2c_hw_engine_dce80_create_arg hw_arg; - - if (use_i2c_sw_engine) { - struct i2c_sw_engine_dce80_create_arg sw_arg; - - sw_arg.engine_id = i; - sw_arg.default_speed = base->default_i2c_sw_speed; - sw_arg.ctx = ctx; - base->i2c_sw_engines[line_id] = - dal_i2c_sw_engine_dce80_create(&sw_arg); - } - - hw_arg.engine_id = i; - hw_arg.reference_frequency = reference_frequency; - hw_arg.default_speed = base->default_i2c_hw_speed; - hw_arg.ctx = ctx; - - base->i2c_hw_engines[line_id] = - dal_i2c_hw_engine_dce80_create(&hw_arg); - - ++i; - } while (i < ARRAY_SIZE(hw_ddc_lines)); - - /* Create AUX engines for all lines which has assisted HW AUX - * 'i' (loop counter) used as DDC/AUX engine_id */ - - i = 0; - - do { - enum gpio_ddc_line line_id = hw_aux_lines[i]; - - struct aux_engine_dce110_init_data arg; - - arg.engine_id = i; - arg.timeout_period = base->aux_timeout_period; - arg.ctx = ctx; - arg.regs = &dce80_aux_regs[i]; - - base->aux_engines[line_id] = - dal_aux_engine_dce110_create(&arg); - - ++i; - } while (i < ARRAY_SIZE(hw_aux_lines)); - - /* TODO Generic I2C SW and HW */ -} - -struct i2caux *dal_i2caux_dce80_create( - struct dc_context *ctx) -{ - struct i2caux_dce80 *i2caux_dce80 = - kzalloc(sizeof(struct i2caux_dce80), GFP_KERNEL); - - if (!i2caux_dce80) { - BREAK_TO_DEBUGGER(); - return NULL; - } - - construct(i2caux_dce80, ctx); - return &i2caux_dce80->base; -} diff --git a/drivers/gpu/drm/amd/display/dc/i2caux/dcn10/i2caux_dcn10.c b/drivers/gpu/drm/amd/display/dc/i2caux/dcn10/i2caux_dcn10.c deleted file mode 100644 index a59c1f50c1e8..000000000000 --- a/drivers/gpu/drm/amd/display/dc/i2caux/dcn10/i2caux_dcn10.c +++ /dev/null @@ -1,120 +0,0 @@ -/* - * Copyright 2012-15 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. - * - * Authors: AMD - * - */ - -#include "dm_services.h" - -#include "include/i2caux_interface.h" -#include "../i2caux.h" -#include "../engine.h" -#include "../i2c_engine.h" -#include "../i2c_sw_engine.h" -#include "../i2c_hw_engine.h" - -#include "../dce110/aux_engine_dce110.h" -#include "../dce110/i2c_hw_engine_dce110.h" -#include "../dce110/i2caux_dce110.h" - -#include "dcn/dcn_1_0_offset.h" -#include "dcn/dcn_1_0_sh_mask.h" -#include "soc15_hw_ip.h" -#include "vega10_ip_offset.h" - -/* begin ********************* - * macros to expend register list macro defined in HW object header file */ - -#define BASE_INNER(seg) \ - DCE_BASE__INST0_SEG ## seg - -/* compile time expand base address. */ -#define BASE(seg) \ - BASE_INNER(seg) - -#define SR(reg_name)\ - .reg_name = BASE(mm ## reg_name ## _BASE_IDX) + \ - mm ## reg_name - -#define SRI(reg_name, block, id)\ - .reg_name = BASE(mm ## block ## id ## _ ## reg_name ## _BASE_IDX) + \ - mm ## block ## id ## _ ## reg_name -/* macros to expend register list macro defined in HW object header file - * end *********************/ - -#define aux_regs(id)\ -[id] = {\ - AUX_COMMON_REG_LIST(id), \ - .AUX_RESET_MASK = DP_AUX0_AUX_CONTROL__AUX_RESET_MASK \ -} - -#define hw_engine_regs(id)\ -{\ - I2C_HW_ENGINE_COMMON_REG_LIST(id) \ -} - -static const struct dce110_aux_registers dcn10_aux_regs[] = { - aux_regs(0), - aux_regs(1), - aux_regs(2), - aux_regs(3), - aux_regs(4), - aux_regs(5), -}; - -static const struct dce110_i2c_hw_engine_registers dcn10_hw_engine_regs[] = { - hw_engine_regs(1), - hw_engine_regs(2), - hw_engine_regs(3), - hw_engine_regs(4), - hw_engine_regs(5), - hw_engine_regs(6) -}; - -static const struct dce110_i2c_hw_engine_shift i2c_shift = { - I2C_COMMON_MASK_SH_LIST_DCE110(__SHIFT) -}; - -static const struct dce110_i2c_hw_engine_mask i2c_mask = { - I2C_COMMON_MASK_SH_LIST_DCE110(_MASK) -}; - -struct i2caux *dal_i2caux_dcn10_create( - struct dc_context *ctx) -{ - struct i2caux_dce110 *i2caux_dce110 = - kzalloc(sizeof(struct i2caux_dce110), GFP_KERNEL); - - if (!i2caux_dce110) { - ASSERT_CRITICAL(false); - return NULL; - } - - dal_i2caux_dce110_construct(i2caux_dce110, - ctx, - ARRAY_SIZE(dcn10_aux_regs), - dcn10_aux_regs, - dcn10_hw_engine_regs, - &i2c_shift, - &i2c_mask); - return &i2caux_dce110->base; -} diff --git a/drivers/gpu/drm/amd/display/dc/i2caux/dcn10/i2caux_dcn10.h b/drivers/gpu/drm/amd/display/dc/i2caux/dcn10/i2caux_dcn10.h deleted file mode 100644 index aeb4a86463d4..000000000000 --- a/drivers/gpu/drm/amd/display/dc/i2caux/dcn10/i2caux_dcn10.h +++ /dev/null @@ -1,32 +0,0 @@ -/* - * Copyright 2012-15 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. - * - * Authors: AMD - * - */ - -#ifndef __DAL_I2C_AUX_DCN10_H__ -#define __DAL_I2C_AUX_DCN10_H__ - -struct i2caux *dal_i2caux_dcn10_create( - struct dc_context *ctx); - -#endif /* __DAL_I2C_AUX_DCN10_H__ */ diff --git a/drivers/gpu/drm/amd/display/dc/i2caux/diagnostics/i2caux_diag.c b/drivers/gpu/drm/amd/display/dc/i2caux/diagnostics/i2caux_diag.c deleted file mode 100644 index e6408f644086..000000000000 --- a/drivers/gpu/drm/amd/display/dc/i2caux/diagnostics/i2caux_diag.c +++ /dev/null @@ -1,97 +0,0 @@ -/* - * Copyright 2012-16 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. - * - * Authors: AMD - * - */ - -#include "dm_services.h" - -/* - * Pre-requisites: headers required by header of this unit - */ -#include "include/i2caux_interface.h" -#include "../i2caux.h" -#include "../engine.h" -#include "../i2c_engine.h" -#include "../i2c_sw_engine.h" -#include "../i2c_hw_engine.h" - -/* - * Header of this unit - */ -#include "i2caux_diag.h" - -/* - * Post-requisites: headers required by this unit - */ - -/* - * This unit - */ - -static void destruct( - struct i2caux *i2caux) -{ - dal_i2caux_destruct(i2caux); -} - -static void destroy( - struct i2caux **i2c_engine) -{ - destruct(*i2c_engine); - - kfree(*i2c_engine); - - *i2c_engine = NULL; -} - -/* function table */ -static const struct i2caux_funcs i2caux_funcs = { - .destroy = destroy, - .acquire_i2c_hw_engine = NULL, - .release_engine = NULL, - .acquire_i2c_sw_engine = NULL, - .acquire_aux_engine = NULL, -}; - -static void construct( - struct i2caux *i2caux, - struct dc_context *ctx) -{ - dal_i2caux_construct(i2caux, ctx); - i2caux->funcs = &i2caux_funcs; -} - -struct i2caux *dal_i2caux_diag_fpga_create( - struct dc_context *ctx) -{ - struct i2caux *i2caux = kzalloc(sizeof(struct i2caux), - GFP_KERNEL); - - if (!i2caux) { - ASSERT_CRITICAL(false); - return NULL; - } - - construct(i2caux, ctx); - return i2caux; -} diff --git a/drivers/gpu/drm/amd/display/dc/i2caux/engine.h b/drivers/gpu/drm/amd/display/dc/i2caux/engine.h deleted file mode 100644 index b16fb1ff687d..000000000000 --- a/drivers/gpu/drm/amd/display/dc/i2caux/engine.h +++ /dev/null @@ -1,111 +0,0 @@ -/* - * Copyright 2012-15 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. - * - * Authors: AMD - * - */ - -#ifndef __DAL_ENGINE_H__ -#define __DAL_ENGINE_H__ - -#include "dc_ddc_types.h" - -enum i2caux_transaction_operation { - I2CAUX_TRANSACTION_READ, - I2CAUX_TRANSACTION_WRITE -}; - -enum i2caux_transaction_address_space { - I2CAUX_TRANSACTION_ADDRESS_SPACE_I2C = 1, - I2CAUX_TRANSACTION_ADDRESS_SPACE_DPCD -}; - -struct i2caux_transaction_payload { - enum i2caux_transaction_address_space address_space; - uint32_t address; - uint32_t length; - uint8_t *data; -}; - -enum i2caux_transaction_status { - I2CAUX_TRANSACTION_STATUS_UNKNOWN = (-1L), - I2CAUX_TRANSACTION_STATUS_SUCCEEDED, - I2CAUX_TRANSACTION_STATUS_FAILED_CHANNEL_BUSY, - I2CAUX_TRANSACTION_STATUS_FAILED_TIMEOUT, - I2CAUX_TRANSACTION_STATUS_FAILED_PROTOCOL_ERROR, - I2CAUX_TRANSACTION_STATUS_FAILED_NACK, - I2CAUX_TRANSACTION_STATUS_FAILED_INCOMPLETE, - I2CAUX_TRANSACTION_STATUS_FAILED_OPERATION, - I2CAUX_TRANSACTION_STATUS_FAILED_INVALID_OPERATION, - I2CAUX_TRANSACTION_STATUS_FAILED_BUFFER_OVERFLOW, - I2CAUX_TRANSACTION_STATUS_FAILED_HPD_DISCON -}; - -struct i2caux_transaction_request { - enum i2caux_transaction_operation operation; - struct i2caux_transaction_payload payload; - enum i2caux_transaction_status status; -}; - -enum i2caux_engine_type { - I2CAUX_ENGINE_TYPE_UNKNOWN = (-1L), - I2CAUX_ENGINE_TYPE_AUX, - I2CAUX_ENGINE_TYPE_I2C_DDC_HW, - I2CAUX_ENGINE_TYPE_I2C_GENERIC_HW, - I2CAUX_ENGINE_TYPE_I2C_SW -}; - -enum i2c_default_speed { - I2CAUX_DEFAULT_I2C_HW_SPEED = 50, - I2CAUX_DEFAULT_I2C_SW_SPEED = 50 -}; - -struct engine; - -struct engine_funcs { - enum i2caux_engine_type (*get_engine_type)( - const struct engine *engine); - bool (*acquire)( - struct engine *engine, - struct ddc *ddc); - bool (*submit_request)( - struct engine *engine, - struct i2caux_transaction_request *request, - bool middle_of_transaction); - void (*release_engine)( - struct engine *engine); -}; - -struct engine { - const struct engine_funcs *funcs; - uint32_t inst; - struct ddc *ddc; - struct dc_context *ctx; -}; - -void dal_i2caux_construct_engine( - struct engine *engine, - struct dc_context *ctx); - -void dal_i2caux_destruct_engine( - struct engine *engine); - -#endif diff --git a/drivers/gpu/drm/amd/display/dc/i2caux/i2c_engine.c b/drivers/gpu/drm/amd/display/dc/i2caux/i2c_engine.c deleted file mode 100644 index 70e20bd47ce4..000000000000 --- a/drivers/gpu/drm/amd/display/dc/i2caux/i2c_engine.c +++ /dev/null @@ -1,118 +0,0 @@ -/* - * Copyright 2012-15 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. - * - * Authors: AMD - * - */ - -#include "dm_services.h" - -/* - * Pre-requisites: headers required by header of this unit - */ -#include "include/i2caux_interface.h" -#include "engine.h" - -/* - * Header of this unit - */ - -#include "i2c_engine.h" - -/* - * Post-requisites: headers required by this unit - */ - -/* - * This unit - */ - -#define FROM_ENGINE(ptr) \ - container_of((ptr), struct i2c_engine, base) - -bool dal_i2c_engine_acquire( - struct engine *engine, - struct ddc *ddc_handle) -{ - struct i2c_engine *i2c_engine = FROM_ENGINE(engine); - - uint32_t counter = 0; - bool result; - - do { - result = i2c_engine->funcs->acquire_engine( - i2c_engine, ddc_handle); - - if (result) - break; - - /* i2c_engine is busy by VBios, lets wait and retry */ - - udelay(10); - - ++counter; - } while (counter < 2); - - if (result) { - if (!i2c_engine->funcs->setup_engine(i2c_engine)) { - engine->funcs->release_engine(engine); - result = false; - } - } - - return result; -} - -bool dal_i2c_engine_setup_i2c_engine( - struct i2c_engine *engine) -{ - /* Derivative classes do not have to override this */ - - return true; -} - -void dal_i2c_engine_submit_channel_request( - struct i2c_engine *engine, - struct i2c_request_transaction_data *request) -{ - -} - -void dal_i2c_engine_process_channel_reply( - struct i2c_engine *engine, - struct i2c_reply_transaction_data *reply) -{ - -} - -void dal_i2c_engine_construct( - struct i2c_engine *engine, - struct dc_context *ctx) -{ - dal_i2caux_construct_engine(&engine->base, ctx); - engine->timeout_delay = 0; -} - -void dal_i2c_engine_destruct( - struct i2c_engine *engine) -{ - dal_i2caux_destruct_engine(&engine->base); -} diff --git a/drivers/gpu/drm/amd/display/dc/i2caux/i2c_engine.h b/drivers/gpu/drm/amd/display/dc/i2caux/i2c_engine.h deleted file mode 100644 index ded6ea34b714..000000000000 --- a/drivers/gpu/drm/amd/display/dc/i2caux/i2c_engine.h +++ /dev/null @@ -1,115 +0,0 @@ -/* - * Copyright 2012-15 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. - * - * Authors: AMD - * - */ - -#ifndef __DAL_I2C_ENGINE_H__ -#define __DAL_I2C_ENGINE_H__ - -enum i2c_channel_operation_result { - I2C_CHANNEL_OPERATION_SUCCEEDED, - I2C_CHANNEL_OPERATION_FAILED, - I2C_CHANNEL_OPERATION_NOT_GRANTED, - I2C_CHANNEL_OPERATION_IS_BUSY, - I2C_CHANNEL_OPERATION_NO_HANDLE_PROVIDED, - I2C_CHANNEL_OPERATION_CHANNEL_IN_USE, - I2C_CHANNEL_OPERATION_CHANNEL_CLIENT_MAX_ALLOWED, - I2C_CHANNEL_OPERATION_ENGINE_BUSY, - I2C_CHANNEL_OPERATION_TIMEOUT, - I2C_CHANNEL_OPERATION_NO_RESPONSE, - I2C_CHANNEL_OPERATION_HW_REQUEST_I2C_BUS, - I2C_CHANNEL_OPERATION_WRONG_PARAMETER, - I2C_CHANNEL_OPERATION_OUT_NB_OF_RETRIES, - I2C_CHANNEL_OPERATION_NOT_STARTED -}; - -struct i2c_request_transaction_data { - enum i2caux_transaction_action action; - enum i2c_channel_operation_result status; - uint8_t address; - uint32_t length; - uint8_t *data; -}; - -struct i2c_reply_transaction_data { - uint32_t length; - uint8_t *data; -}; - -struct i2c_engine; - -struct i2c_engine_funcs { - void (*destroy)( - struct i2c_engine **ptr); - uint32_t (*get_speed)( - const struct i2c_engine *engine); - void (*set_speed)( - struct i2c_engine *engine, - uint32_t speed); - bool (*acquire_engine)( - struct i2c_engine *engine, - struct ddc *ddc); - bool (*setup_engine)( - struct i2c_engine *engine); - void (*submit_channel_request)( - struct i2c_engine *engine, - struct i2c_request_transaction_data *request); - void (*process_channel_reply)( - struct i2c_engine *engine, - struct i2c_reply_transaction_data *reply); - enum i2c_channel_operation_result (*get_channel_status)( - struct i2c_engine *engine, - uint8_t *returned_bytes); -}; - -struct i2c_engine { - struct engine base; - const struct i2c_engine_funcs *funcs; - uint32_t timeout_delay; - uint32_t setup_limit; - uint32_t send_reset_length; -}; - -void dal_i2c_engine_construct( - struct i2c_engine *engine, - struct dc_context *ctx); - -void dal_i2c_engine_destruct( - struct i2c_engine *engine); - -bool dal_i2c_engine_setup_i2c_engine( - struct i2c_engine *engine); - -void dal_i2c_engine_submit_channel_request( - struct i2c_engine *engine, - struct i2c_request_transaction_data *request); - -void dal_i2c_engine_process_channel_reply( - struct i2c_engine *engine, - struct i2c_reply_transaction_data *reply); - -bool dal_i2c_engine_acquire( - struct engine *ptr, - struct ddc *ddc_handle); - -#endif diff --git a/drivers/gpu/drm/amd/display/dc/i2caux/i2c_generic_hw_engine.c b/drivers/gpu/drm/amd/display/dc/i2caux/i2c_generic_hw_engine.c deleted file mode 100644 index 5a4295e0fae5..000000000000 --- a/drivers/gpu/drm/amd/display/dc/i2caux/i2c_generic_hw_engine.c +++ /dev/null @@ -1,284 +0,0 @@ -/* - * Copyright 2012-15 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. - * - * Authors: AMD - * - */ - -#include "dm_services.h" - -/* - * Pre-requisites: headers required by header of this unit - */ -#include "include/i2caux_interface.h" -#include "engine.h" -#include "i2c_engine.h" -#include "i2c_hw_engine.h" - -/* - * Header of this unit - */ - -#include "i2c_generic_hw_engine.h" - -/* - * Post-requisites: headers required by this unit - */ - -/* - * This unit - */ - -/* - * @brief - * Cast 'struct i2c_hw_engine *' - * to 'struct i2c_generic_hw_engine *' - */ -#define FROM_I2C_HW_ENGINE(ptr) \ - container_of((ptr), struct i2c_generic_hw_engine, base) - -/* - * @brief - * Cast 'struct i2c_engine *' - * to 'struct i2c_generic_hw_engine *' - */ -#define FROM_I2C_ENGINE(ptr) \ - FROM_I2C_HW_ENGINE(container_of((ptr), struct i2c_hw_engine, base)) - -/* - * @brief - * Cast 'struct engine *' - * to 'struct i2c_generic_hw_engine *' - */ -#define FROM_ENGINE(ptr) \ - FROM_I2C_ENGINE(container_of((ptr), struct i2c_engine, base)) - -enum i2caux_engine_type dal_i2c_generic_hw_engine_get_engine_type( - const struct engine *engine) -{ - return I2CAUX_ENGINE_TYPE_I2C_GENERIC_HW; -} - -/* - * @brief - * Single transaction handling. - * Since transaction may be bigger than HW buffer size, - * it divides transaction to sub-transactions - * and uses batch transaction feature of the engine. - */ -bool dal_i2c_generic_hw_engine_submit_request( - struct engine *engine, - struct i2caux_transaction_request *i2caux_request, - bool middle_of_transaction) -{ - struct i2c_generic_hw_engine *hw_engine = FROM_ENGINE(engine); - - struct i2c_hw_engine *base = &hw_engine->base; - - uint32_t max_payload_size = - base->funcs->get_hw_buffer_available_size(base); - - bool initial_stop_bit = !middle_of_transaction; - - struct i2c_generic_transaction_attributes attributes; - - enum i2c_channel_operation_result operation_result = - I2C_CHANNEL_OPERATION_FAILED; - - bool result = false; - - /* setup transaction initial properties */ - - uint8_t address = i2caux_request->payload.address; - uint8_t *current_payload = i2caux_request->payload.data; - uint32_t remaining_payload_size = i2caux_request->payload.length; - - bool first_iteration = true; - - if (i2caux_request->operation == I2CAUX_TRANSACTION_READ) - attributes.action = I2CAUX_TRANSACTION_ACTION_I2C_READ; - else if (i2caux_request->operation == I2CAUX_TRANSACTION_WRITE) - attributes.action = I2CAUX_TRANSACTION_ACTION_I2C_WRITE; - else { - i2caux_request->status = - I2CAUX_TRANSACTION_STATUS_FAILED_INVALID_OPERATION; - return false; - } - - /* Do batch transaction. - * Divide read/write data into payloads which fit HW buffer size. - * 1. Single transaction: - * start_bit = 1, stop_bit depends on session state, ack_on_read = 0; - * 2. Start of batch transaction: - * start_bit = 1, stop_bit = 0, ack_on_read = 1; - * 3. Middle of batch transaction: - * start_bit = 0, stop_bit = 0, ack_on_read = 1; - * 4. End of batch transaction: - * start_bit = 0, stop_bit depends on session state, ack_on_read = 0. - * Session stop bit is set if 'middle_of_transaction' = 0. */ - - while (remaining_payload_size) { - uint32_t current_transaction_size; - uint32_t current_payload_size; - - bool last_iteration; - bool stop_bit; - - /* Calculate current transaction size and payload size. - * Transaction size = total number of bytes in transaction, - * including slave's address; - * Payload size = number of data bytes in transaction. */ - - if (first_iteration) { - /* In the first sub-transaction we send slave's address - * thus we need to reserve one byte for it */ - current_transaction_size = - (remaining_payload_size > max_payload_size - 1) ? - max_payload_size : - remaining_payload_size + 1; - - current_payload_size = current_transaction_size - 1; - } else { - /* Second and further sub-transactions will have - * entire buffer reserved for data */ - current_transaction_size = - (remaining_payload_size > max_payload_size) ? - max_payload_size : - remaining_payload_size; - - current_payload_size = current_transaction_size; - } - - last_iteration = - (remaining_payload_size == current_payload_size); - - stop_bit = last_iteration ? initial_stop_bit : false; - - /* write slave device address */ - - if (first_iteration) - hw_engine->funcs->write_address(hw_engine, address); - - /* write current portion of data, if requested */ - - if (i2caux_request->operation == I2CAUX_TRANSACTION_WRITE) - hw_engine->funcs->write_data( - hw_engine, - current_payload, - current_payload_size); - - /* execute transaction */ - - attributes.start_bit = first_iteration; - attributes.stop_bit = stop_bit; - attributes.last_read = last_iteration; - attributes.transaction_size = current_transaction_size; - - hw_engine->funcs->execute_transaction(hw_engine, &attributes); - - /* wait until transaction is processed; if it fails - quit */ - - operation_result = base->funcs->wait_on_operation_result( - base, - base->funcs->get_transaction_timeout( - base, current_transaction_size), - I2C_CHANNEL_OPERATION_ENGINE_BUSY); - - if (operation_result != I2C_CHANNEL_OPERATION_SUCCEEDED) - break; - - /* read current portion of data, if requested */ - - /* the read offset should be 1 for first sub-transaction, - * and 0 for any next one */ - - if (i2caux_request->operation == I2CAUX_TRANSACTION_READ) - hw_engine->funcs->read_data(hw_engine, current_payload, - current_payload_size, first_iteration ? 1 : 0); - - /* update loop variables */ - - first_iteration = false; - current_payload += current_payload_size; - remaining_payload_size -= current_payload_size; - } - - /* update transaction status */ - - switch (operation_result) { - case I2C_CHANNEL_OPERATION_SUCCEEDED: - i2caux_request->status = - I2CAUX_TRANSACTION_STATUS_SUCCEEDED; - result = true; - break; - case I2C_CHANNEL_OPERATION_NO_RESPONSE: - i2caux_request->status = - I2CAUX_TRANSACTION_STATUS_FAILED_NACK; - break; - case I2C_CHANNEL_OPERATION_TIMEOUT: - i2caux_request->status = - I2CAUX_TRANSACTION_STATUS_FAILED_TIMEOUT; - break; - case I2C_CHANNEL_OPERATION_FAILED: - i2caux_request->status = - I2CAUX_TRANSACTION_STATUS_FAILED_INCOMPLETE; - break; - default: - i2caux_request->status = - I2CAUX_TRANSACTION_STATUS_FAILED_OPERATION; - } - - return result; -} - -/* - * @brief - * Returns number of microseconds to wait until timeout to be considered - */ -uint32_t dal_i2c_generic_hw_engine_get_transaction_timeout( - const struct i2c_hw_engine *engine, - uint32_t length) -{ - const struct i2c_engine *base = &engine->base; - - uint32_t speed = base->funcs->get_speed(base); - - if (!speed) - return 0; - - /* total timeout = period_timeout * (start + data bits count + stop) */ - - return ((1000 * TRANSACTION_TIMEOUT_IN_I2C_CLOCKS) / speed) * - (1 + (length << 3) + 1); -} - -void dal_i2c_generic_hw_engine_construct( - struct i2c_generic_hw_engine *engine, - struct dc_context *ctx) -{ - dal_i2c_hw_engine_construct(&engine->base, ctx); -} - -void dal_i2c_generic_hw_engine_destruct( - struct i2c_generic_hw_engine *engine) -{ - dal_i2c_hw_engine_destruct(&engine->base); -} diff --git a/drivers/gpu/drm/amd/display/dc/i2caux/i2c_generic_hw_engine.h b/drivers/gpu/drm/amd/display/dc/i2caux/i2c_generic_hw_engine.h deleted file mode 100644 index 1da0397b04a2..000000000000 --- a/drivers/gpu/drm/amd/display/dc/i2caux/i2c_generic_hw_engine.h +++ /dev/null @@ -1,77 +0,0 @@ -/* - * Copyright 2012-15 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. - * - * Authors: AMD - * - */ - -#ifndef __DAL_I2C_GENERIC_HW_ENGINE_H__ -#define __DAL_I2C_GENERIC_HW_ENGINE_H__ - -struct i2c_generic_transaction_attributes { - enum i2caux_transaction_action action; - uint32_t transaction_size; - bool start_bit; - bool stop_bit; - bool last_read; -}; - -struct i2c_generic_hw_engine; - -struct i2c_generic_hw_engine_funcs { - void (*write_address)( - struct i2c_generic_hw_engine *engine, - uint8_t address); - void (*write_data)( - struct i2c_generic_hw_engine *engine, - const uint8_t *buffer, - uint32_t length); - void (*read_data)( - struct i2c_generic_hw_engine *engine, - uint8_t *buffer, - uint32_t length, - uint32_t offset); - void (*execute_transaction)( - struct i2c_generic_hw_engine *engine, - struct i2c_generic_transaction_attributes *attributes); -}; - -struct i2c_generic_hw_engine { - struct i2c_hw_engine base; - const struct i2c_generic_hw_engine_funcs *funcs; -}; - -void dal_i2c_generic_hw_engine_construct( - struct i2c_generic_hw_engine *engine, - struct dc_context *ctx); - -void dal_i2c_generic_hw_engine_destruct( - struct i2c_generic_hw_engine *engine); -enum i2caux_engine_type dal_i2c_generic_hw_engine_get_engine_type( - const struct engine *engine); -bool dal_i2c_generic_hw_engine_submit_request( - struct engine *ptr, - struct i2caux_transaction_request *i2caux_request, - bool middle_of_transaction); -uint32_t dal_i2c_generic_hw_engine_get_transaction_timeout( - const struct i2c_hw_engine *engine, - uint32_t length); -#endif diff --git a/drivers/gpu/drm/amd/display/dc/i2caux/i2c_hw_engine.c b/drivers/gpu/drm/amd/display/dc/i2caux/i2c_hw_engine.c deleted file mode 100644 index 141898533e8e..000000000000 --- a/drivers/gpu/drm/amd/display/dc/i2caux/i2c_hw_engine.c +++ /dev/null @@ -1,251 +0,0 @@ -/* - * Copyright 2012-15 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. - * - * Authors: AMD - * - */ - -#include "dm_services.h" -#include "dm_event_log.h" - -/* - * Pre-requisites: headers required by header of this unit - */ -#include "include/i2caux_interface.h" -#include "engine.h" -#include "i2c_engine.h" - -/* - * Header of this unit - */ - -#include "i2c_hw_engine.h" - -/* - * Post-requisites: headers required by this unit - */ - -/* - * This unit - */ - -/* - * @brief - * Cast 'struct i2c_engine *' - * to 'struct i2c_hw_engine *' - */ -#define FROM_I2C_ENGINE(ptr) \ - container_of((ptr), struct i2c_hw_engine, base) - -/* - * @brief - * Cast 'struct engine *' - * to 'struct i2c_hw_engine *' - */ -#define FROM_ENGINE(ptr) \ - FROM_I2C_ENGINE(container_of((ptr), struct i2c_engine, base)) - -enum i2caux_engine_type dal_i2c_hw_engine_get_engine_type( - const struct engine *engine) -{ - return I2CAUX_ENGINE_TYPE_I2C_DDC_HW; -} - -bool dal_i2c_hw_engine_submit_request( - struct engine *engine, - struct i2caux_transaction_request *i2caux_request, - bool middle_of_transaction) -{ - struct i2c_hw_engine *hw_engine = FROM_ENGINE(engine); - - struct i2c_request_transaction_data request; - - uint32_t transaction_timeout; - - enum i2c_channel_operation_result operation_result; - - bool result = false; - - /* We need following: - * transaction length will not exceed - * the number of free bytes in HW buffer (minus one for address)*/ - - if (i2caux_request->payload.length >= - hw_engine->funcs->get_hw_buffer_available_size(hw_engine)) { - i2caux_request->status = - I2CAUX_TRANSACTION_STATUS_FAILED_BUFFER_OVERFLOW; - return false; - } - - if (i2caux_request->operation == I2CAUX_TRANSACTION_READ) - request.action = middle_of_transaction ? - I2CAUX_TRANSACTION_ACTION_I2C_READ_MOT : - I2CAUX_TRANSACTION_ACTION_I2C_READ; - else if (i2caux_request->operation == I2CAUX_TRANSACTION_WRITE) - request.action = middle_of_transaction ? - I2CAUX_TRANSACTION_ACTION_I2C_WRITE_MOT : - I2CAUX_TRANSACTION_ACTION_I2C_WRITE; - else { - i2caux_request->status = - I2CAUX_TRANSACTION_STATUS_FAILED_INVALID_OPERATION; - /* [anaumov] in DAL2, there was no "return false" */ - return false; - } - - request.address = (uint8_t)i2caux_request->payload.address; - request.length = i2caux_request->payload.length; - request.data = i2caux_request->payload.data; - - /* obtain timeout value before submitting request */ - - transaction_timeout = hw_engine->funcs->get_transaction_timeout( - hw_engine, i2caux_request->payload.length + 1); - - hw_engine->base.funcs->submit_channel_request( - &hw_engine->base, &request); - /* EVENT_LOG_AUX_REQ(engine->ddc->pin_data->en, EVENT_LOG_AUX_ORIGIN_I2C, */ - /* request.action, request.address, request.length, request.data); */ - - if ((request.status == I2C_CHANNEL_OPERATION_FAILED) || - (request.status == I2C_CHANNEL_OPERATION_ENGINE_BUSY)) { - i2caux_request->status = - I2CAUX_TRANSACTION_STATUS_FAILED_CHANNEL_BUSY; - return false; - } - - /* wait until transaction proceed */ - - operation_result = hw_engine->funcs->wait_on_operation_result( - hw_engine, - transaction_timeout, - I2C_CHANNEL_OPERATION_ENGINE_BUSY); - - /* update transaction status */ - - switch (operation_result) { - case I2C_CHANNEL_OPERATION_SUCCEEDED: - i2caux_request->status = - I2CAUX_TRANSACTION_STATUS_SUCCEEDED; - result = true; - break; - case I2C_CHANNEL_OPERATION_NO_RESPONSE: - i2caux_request->status = - I2CAUX_TRANSACTION_STATUS_FAILED_NACK; - break; - case I2C_CHANNEL_OPERATION_TIMEOUT: - i2caux_request->status = - I2CAUX_TRANSACTION_STATUS_FAILED_TIMEOUT; - break; - case I2C_CHANNEL_OPERATION_FAILED: - i2caux_request->status = - I2CAUX_TRANSACTION_STATUS_FAILED_INCOMPLETE; - break; - default: - i2caux_request->status = - I2CAUX_TRANSACTION_STATUS_FAILED_OPERATION; - } - - if (result && (i2caux_request->operation == I2CAUX_TRANSACTION_READ)) { - struct i2c_reply_transaction_data reply; - - reply.data = i2caux_request->payload.data; - reply.length = i2caux_request->payload.length; - - hw_engine->base.funcs-> - process_channel_reply(&hw_engine->base, &reply); - /* EVENT_LOG_AUX_REP(engine->ddc->pin_data->en, EVENT_LOG_AUX_ORIGIN_I2C, */ - /* AUX_TRANSACTION_REPLY_I2C_ACK, reply.length, reply.data); */ - } - - - - return result; -} - -bool dal_i2c_hw_engine_acquire_engine( - struct i2c_engine *engine, - struct ddc *ddc) -{ - enum gpio_result result; - uint32_t current_speed; - - result = dal_ddc_open(ddc, GPIO_MODE_HARDWARE, - GPIO_DDC_CONFIG_TYPE_MODE_I2C); - - if (result != GPIO_RESULT_OK) - return false; - - engine->base.ddc = ddc; - - current_speed = engine->funcs->get_speed(engine); - - if (current_speed) - FROM_I2C_ENGINE(engine)->original_speed = current_speed; - - return true; -} -/* - * @brief - * Queries in a loop for current engine status - * until retrieved status matches 'expected_result', or timeout occurs. - * Timeout given in microseconds - * and the status query frequency is also one per microsecond. - */ -enum i2c_channel_operation_result dal_i2c_hw_engine_wait_on_operation_result( - struct i2c_hw_engine *engine, - uint32_t timeout, - enum i2c_channel_operation_result expected_result) -{ - enum i2c_channel_operation_result result; - uint32_t i = 0; - - if (!timeout) - return I2C_CHANNEL_OPERATION_SUCCEEDED; - - do { - result = engine->base.funcs->get_channel_status( - &engine->base, NULL); - - if (result != expected_result) - break; - - udelay(1); - - ++i; - } while (i < timeout); - - return result; -} - -void dal_i2c_hw_engine_construct( - struct i2c_hw_engine *engine, - struct dc_context *ctx) -{ - dal_i2c_engine_construct(&engine->base, ctx); - engine->original_speed = I2CAUX_DEFAULT_I2C_HW_SPEED; - engine->default_speed = I2CAUX_DEFAULT_I2C_HW_SPEED; -} - -void dal_i2c_hw_engine_destruct( - struct i2c_hw_engine *engine) -{ - dal_i2c_engine_destruct(&engine->base); -} diff --git a/drivers/gpu/drm/amd/display/dc/i2caux/i2c_hw_engine.h b/drivers/gpu/drm/amd/display/dc/i2caux/i2c_hw_engine.h deleted file mode 100644 index 8936a994804a..000000000000 --- a/drivers/gpu/drm/amd/display/dc/i2caux/i2c_hw_engine.h +++ /dev/null @@ -1,80 +0,0 @@ -/* - * Copyright 2012-15 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. - * - * Authors: AMD - * - */ - -#ifndef __DAL_I2C_HW_ENGINE_H__ -#define __DAL_I2C_HW_ENGINE_H__ - -enum { - TRANSACTION_TIMEOUT_IN_I2C_CLOCKS = 32 -}; - -struct i2c_hw_engine; - -struct i2c_hw_engine_funcs { - uint32_t (*get_hw_buffer_available_size)( - const struct i2c_hw_engine *engine); - enum i2c_channel_operation_result (*wait_on_operation_result)( - struct i2c_hw_engine *engine, - uint32_t timeout, - enum i2c_channel_operation_result expected_result); - uint32_t (*get_transaction_timeout)( - const struct i2c_hw_engine *engine, - uint32_t length); -}; - -struct i2c_hw_engine { - struct i2c_engine base; - const struct i2c_hw_engine_funcs *funcs; - - /* Values below are in kilohertz */ - uint32_t original_speed; - uint32_t default_speed; -}; - -void dal_i2c_hw_engine_construct( - struct i2c_hw_engine *engine, - struct dc_context *ctx); - -void dal_i2c_hw_engine_destruct( - struct i2c_hw_engine *engine); - -enum i2c_channel_operation_result dal_i2c_hw_engine_wait_on_operation_result( - struct i2c_hw_engine *engine, - uint32_t timeout, - enum i2c_channel_operation_result expected_result); - -bool dal_i2c_hw_engine_acquire_engine( - struct i2c_engine *engine, - struct ddc *ddc); - -bool dal_i2c_hw_engine_submit_request( - struct engine *ptr, - struct i2caux_transaction_request *i2caux_request, - bool middle_of_transaction); - -enum i2caux_engine_type dal_i2c_hw_engine_get_engine_type( - const struct engine *engine); - -#endif diff --git a/drivers/gpu/drm/amd/display/dc/i2caux/i2c_sw_engine.c b/drivers/gpu/drm/amd/display/dc/i2caux/i2c_sw_engine.c deleted file mode 100644 index 8e19bb629394..000000000000 --- a/drivers/gpu/drm/amd/display/dc/i2caux/i2c_sw_engine.c +++ /dev/null @@ -1,601 +0,0 @@ -/* - * Copyright 2012-15 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. - * - * Authors: AMD - * - */ - -#include "dm_services.h" - -/* - * Pre-requisites: headers required by header of this unit - */ -#include "include/i2caux_interface.h" -#include "engine.h" -#include "i2c_engine.h" - -/* - * Header of this unit - */ - -#include "i2c_sw_engine.h" - -/* - * Post-requisites: headers required by this unit - */ - -/* - * This unit - */ - -#define SCL false -#define SDA true - -static inline bool read_bit_from_ddc( - struct ddc *ddc, - bool data_nor_clock) -{ - uint32_t value = 0; - - if (data_nor_clock) - dal_gpio_get_value(ddc->pin_data, &value); - else - dal_gpio_get_value(ddc->pin_clock, &value); - - return (value != 0); -} - -static inline void write_bit_to_ddc( - struct ddc *ddc, - bool data_nor_clock, - bool bit) -{ - uint32_t value = bit ? 1 : 0; - - if (data_nor_clock) - dal_gpio_set_value(ddc->pin_data, value); - else - dal_gpio_set_value(ddc->pin_clock, value); -} - -static bool wait_for_scl_high( - struct dc_context *ctx, - struct ddc *ddc, - uint16_t clock_delay_div_4) -{ - uint32_t scl_retry = 0; - uint32_t scl_retry_max = I2C_SW_TIMEOUT_DELAY / clock_delay_div_4; - - udelay(clock_delay_div_4); - - /* 3 milliseconds delay - * to wake up some displays from "low power" state. - */ - - do { - if (read_bit_from_ddc(ddc, SCL)) - return true; - - udelay(clock_delay_div_4); - - ++scl_retry; - } while (scl_retry <= scl_retry_max); - - return false; -} - -static bool start_sync( - struct dc_context *ctx, - struct ddc *ddc_handle, - uint16_t clock_delay_div_4) -{ - uint32_t retry = 0; - - /* The I2C communications start signal is: - * the SDA going low from high, while the SCL is high. */ - - write_bit_to_ddc(ddc_handle, SCL, true); - - udelay(clock_delay_div_4); - - do { - write_bit_to_ddc(ddc_handle, SDA, true); - - if (!read_bit_from_ddc(ddc_handle, SDA)) { - ++retry; - continue; - } - - udelay(clock_delay_div_4); - - write_bit_to_ddc(ddc_handle, SCL, true); - - if (!wait_for_scl_high(ctx, ddc_handle, clock_delay_div_4)) - break; - - write_bit_to_ddc(ddc_handle, SDA, false); - - udelay(clock_delay_div_4); - - write_bit_to_ddc(ddc_handle, SCL, false); - - udelay(clock_delay_div_4); - - return true; - } while (retry <= I2C_SW_RETRIES); - - return false; -} - -static bool stop_sync( - struct dc_context *ctx, - struct ddc *ddc_handle, - uint16_t clock_delay_div_4) -{ - uint32_t retry = 0; - - /* The I2C communications stop signal is: - * the SDA going high from low, while the SCL is high. */ - - write_bit_to_ddc(ddc_handle, SCL, false); - - udelay(clock_delay_div_4); - - write_bit_to_ddc(ddc_handle, SDA, false); - - udelay(clock_delay_div_4); - - write_bit_to_ddc(ddc_handle, SCL, true); - - if (!wait_for_scl_high(ctx, ddc_handle, clock_delay_div_4)) - return false; - - write_bit_to_ddc(ddc_handle, SDA, true); - - do { - udelay(clock_delay_div_4); - - if (read_bit_from_ddc(ddc_handle, SDA)) - return true; - - ++retry; - } while (retry <= 2); - - return false; -} - -static bool write_byte( - struct dc_context *ctx, - struct ddc *ddc_handle, - uint16_t clock_delay_div_4, - uint8_t byte) -{ - int32_t shift = 7; - bool ack; - - /* bits are transmitted serially, starting from MSB */ - - do { - udelay(clock_delay_div_4); - - write_bit_to_ddc(ddc_handle, SDA, (byte >> shift) & 1); - - udelay(clock_delay_div_4); - - write_bit_to_ddc(ddc_handle, SCL, true); - - if (!wait_for_scl_high(ctx, ddc_handle, clock_delay_div_4)) - return false; - - write_bit_to_ddc(ddc_handle, SCL, false); - - --shift; - } while (shift >= 0); - - /* The display sends ACK by preventing the SDA from going high - * after the SCL pulse we use to send our last data bit. - * If the SDA goes high after that bit, it's a NACK */ - - udelay(clock_delay_div_4); - - write_bit_to_ddc(ddc_handle, SDA, true); - - udelay(clock_delay_div_4); - - write_bit_to_ddc(ddc_handle, SCL, true); - - if (!wait_for_scl_high(ctx, ddc_handle, clock_delay_div_4)) - return false; - - /* read ACK bit */ - - ack = !read_bit_from_ddc(ddc_handle, SDA); - - udelay(clock_delay_div_4 << 1); - - write_bit_to_ddc(ddc_handle, SCL, false); - - udelay(clock_delay_div_4 << 1); - - return ack; -} - -static bool read_byte( - struct dc_context *ctx, - struct ddc *ddc_handle, - uint16_t clock_delay_div_4, - uint8_t *byte, - bool more) -{ - int32_t shift = 7; - - uint8_t data = 0; - - /* The data bits are read from MSB to LSB; - * bit is read while SCL is high */ - - do { - write_bit_to_ddc(ddc_handle, SCL, true); - - if (!wait_for_scl_high(ctx, ddc_handle, clock_delay_div_4)) - return false; - - if (read_bit_from_ddc(ddc_handle, SDA)) - data |= (1 << shift); - - write_bit_to_ddc(ddc_handle, SCL, false); - - udelay(clock_delay_div_4 << 1); - - --shift; - } while (shift >= 0); - - /* read only whole byte */ - - *byte = data; - - udelay(clock_delay_div_4); - - /* send the acknowledge bit: - * SDA low means ACK, SDA high means NACK */ - - write_bit_to_ddc(ddc_handle, SDA, !more); - - udelay(clock_delay_div_4); - - write_bit_to_ddc(ddc_handle, SCL, true); - - if (!wait_for_scl_high(ctx, ddc_handle, clock_delay_div_4)) - return false; - - write_bit_to_ddc(ddc_handle, SCL, false); - - udelay(clock_delay_div_4); - - write_bit_to_ddc(ddc_handle, SDA, true); - - udelay(clock_delay_div_4); - - return true; -} - -static bool i2c_write( - struct dc_context *ctx, - struct ddc *ddc_handle, - uint16_t clock_delay_div_4, - uint8_t address, - uint32_t length, - const uint8_t *data) -{ - uint32_t i = 0; - - if (!write_byte(ctx, ddc_handle, clock_delay_div_4, address)) - return false; - - while (i < length) { - if (!write_byte(ctx, ddc_handle, clock_delay_div_4, data[i])) - return false; - ++i; - } - - return true; -} - -static bool i2c_read( - struct dc_context *ctx, - struct ddc *ddc_handle, - uint16_t clock_delay_div_4, - uint8_t address, - uint32_t length, - uint8_t *data) -{ - uint32_t i = 0; - - if (!write_byte(ctx, ddc_handle, clock_delay_div_4, address)) - return false; - - while (i < length) { - if (!read_byte(ctx, ddc_handle, clock_delay_div_4, data + i, - i < length - 1)) - return false; - ++i; - } - - return true; -} - -/* - * @brief - * Cast 'struct i2c_engine *' - * to 'struct i2c_sw_engine *' - */ -#define FROM_I2C_ENGINE(ptr) \ - container_of((ptr), struct i2c_sw_engine, base) - -/* - * @brief - * Cast 'struct engine *' - * to 'struct i2c_sw_engine *' - */ -#define FROM_ENGINE(ptr) \ - FROM_I2C_ENGINE(container_of((ptr), struct i2c_engine, base)) - -enum i2caux_engine_type dal_i2c_sw_engine_get_engine_type( - const struct engine *engine) -{ - return I2CAUX_ENGINE_TYPE_I2C_SW; -} - -bool dal_i2c_sw_engine_submit_request( - struct engine *engine, - struct i2caux_transaction_request *i2caux_request, - bool middle_of_transaction) -{ - struct i2c_sw_engine *sw_engine = FROM_ENGINE(engine); - - struct i2c_engine *base = &sw_engine->base; - - struct i2c_request_transaction_data request; - bool operation_succeeded = false; - - if (i2caux_request->operation == I2CAUX_TRANSACTION_READ) - request.action = middle_of_transaction ? - I2CAUX_TRANSACTION_ACTION_I2C_READ_MOT : - I2CAUX_TRANSACTION_ACTION_I2C_READ; - else if (i2caux_request->operation == I2CAUX_TRANSACTION_WRITE) - request.action = middle_of_transaction ? - I2CAUX_TRANSACTION_ACTION_I2C_WRITE_MOT : - I2CAUX_TRANSACTION_ACTION_I2C_WRITE; - else { - i2caux_request->status = - I2CAUX_TRANSACTION_STATUS_FAILED_INVALID_OPERATION; - /* in DAL2, there was no "return false" */ - return false; - } - - request.address = (uint8_t)i2caux_request->payload.address; - request.length = i2caux_request->payload.length; - request.data = i2caux_request->payload.data; - - base->funcs->submit_channel_request(base, &request); - - if ((request.status == I2C_CHANNEL_OPERATION_ENGINE_BUSY) || - (request.status == I2C_CHANNEL_OPERATION_FAILED)) - i2caux_request->status = - I2CAUX_TRANSACTION_STATUS_FAILED_CHANNEL_BUSY; - else { - enum i2c_channel_operation_result operation_result; - - do { - operation_result = - base->funcs->get_channel_status(base, NULL); - - switch (operation_result) { - case I2C_CHANNEL_OPERATION_SUCCEEDED: - i2caux_request->status = - I2CAUX_TRANSACTION_STATUS_SUCCEEDED; - operation_succeeded = true; - break; - case I2C_CHANNEL_OPERATION_NO_RESPONSE: - i2caux_request->status = - I2CAUX_TRANSACTION_STATUS_FAILED_NACK; - break; - case I2C_CHANNEL_OPERATION_TIMEOUT: - i2caux_request->status = - I2CAUX_TRANSACTION_STATUS_FAILED_TIMEOUT; - break; - case I2C_CHANNEL_OPERATION_FAILED: - i2caux_request->status = - I2CAUX_TRANSACTION_STATUS_FAILED_INCOMPLETE; - break; - default: - i2caux_request->status = - I2CAUX_TRANSACTION_STATUS_FAILED_OPERATION; - break; - } - } while (operation_result == I2C_CHANNEL_OPERATION_ENGINE_BUSY); - } - - return operation_succeeded; -} - -uint32_t dal_i2c_sw_engine_get_speed( - const struct i2c_engine *engine) -{ - return FROM_I2C_ENGINE(engine)->speed; -} - -void dal_i2c_sw_engine_set_speed( - struct i2c_engine *engine, - uint32_t speed) -{ - struct i2c_sw_engine *sw_engine = FROM_I2C_ENGINE(engine); - - ASSERT(speed); - - sw_engine->speed = speed ? speed : I2CAUX_DEFAULT_I2C_SW_SPEED; - - sw_engine->clock_delay = 1000 / sw_engine->speed; - - if (sw_engine->clock_delay < 12) - sw_engine->clock_delay = 12; -} - -bool dal_i2caux_i2c_sw_engine_acquire_engine( - struct i2c_engine *engine, - struct ddc *ddc) -{ - enum gpio_result result; - - result = dal_ddc_open(ddc, GPIO_MODE_FAST_OUTPUT, - GPIO_DDC_CONFIG_TYPE_MODE_I2C); - - if (result != GPIO_RESULT_OK) - return false; - - engine->base.ddc = ddc; - - return true; -} - -void dal_i2c_sw_engine_submit_channel_request( - struct i2c_engine *engine, - struct i2c_request_transaction_data *req) -{ - struct i2c_sw_engine *sw_engine = FROM_I2C_ENGINE(engine); - - struct ddc *ddc = engine->base.ddc; - uint16_t clock_delay_div_4 = sw_engine->clock_delay >> 2; - - /* send sync (start / repeated start) */ - - bool result = start_sync(engine->base.ctx, ddc, clock_delay_div_4); - - /* process payload */ - - if (result) { - switch (req->action) { - case I2CAUX_TRANSACTION_ACTION_I2C_WRITE: - case I2CAUX_TRANSACTION_ACTION_I2C_WRITE_MOT: - result = i2c_write(engine->base.ctx, ddc, clock_delay_div_4, - req->address, req->length, req->data); - break; - case I2CAUX_TRANSACTION_ACTION_I2C_READ: - case I2CAUX_TRANSACTION_ACTION_I2C_READ_MOT: - result = i2c_read(engine->base.ctx, ddc, clock_delay_div_4, - req->address, req->length, req->data); - break; - default: - result = false; - break; - } - } - - /* send stop if not 'mot' or operation failed */ - - if (!result || - (req->action == I2CAUX_TRANSACTION_ACTION_I2C_WRITE) || - (req->action == I2CAUX_TRANSACTION_ACTION_I2C_READ)) - if (!stop_sync(engine->base.ctx, ddc, clock_delay_div_4)) - result = false; - - req->status = result ? - I2C_CHANNEL_OPERATION_SUCCEEDED : - I2C_CHANNEL_OPERATION_FAILED; -} - -enum i2c_channel_operation_result dal_i2c_sw_engine_get_channel_status( - struct i2c_engine *engine, - uint8_t *returned_bytes) -{ - /* No arbitration with VBIOS is performed since DCE 6.0 */ - return I2C_CHANNEL_OPERATION_SUCCEEDED; -} - -void dal_i2c_sw_engine_destruct( - struct i2c_sw_engine *engine) -{ - dal_i2c_engine_destruct(&engine->base); -} - -static void destroy( - struct i2c_engine **ptr) -{ - dal_i2c_sw_engine_destruct(FROM_I2C_ENGINE(*ptr)); - - kfree(*ptr); - *ptr = NULL; -} - -static const struct i2c_engine_funcs i2c_engine_funcs = { - .acquire_engine = dal_i2caux_i2c_sw_engine_acquire_engine, - .destroy = destroy, - .get_speed = dal_i2c_sw_engine_get_speed, - .set_speed = dal_i2c_sw_engine_set_speed, - .setup_engine = dal_i2c_engine_setup_i2c_engine, - .submit_channel_request = dal_i2c_sw_engine_submit_channel_request, - .process_channel_reply = dal_i2c_engine_process_channel_reply, - .get_channel_status = dal_i2c_sw_engine_get_channel_status, -}; - -static void release_engine( - struct engine *engine) -{ - -} - -static const struct engine_funcs engine_funcs = { - .release_engine = release_engine, - .get_engine_type = dal_i2c_sw_engine_get_engine_type, - .acquire = dal_i2c_engine_acquire, - .submit_request = dal_i2c_sw_engine_submit_request, -}; - -void dal_i2c_sw_engine_construct( - struct i2c_sw_engine *engine, - const struct i2c_sw_engine_create_arg *arg) -{ - dal_i2c_engine_construct(&engine->base, arg->ctx); - dal_i2c_sw_engine_set_speed(&engine->base, arg->default_speed); - engine->base.funcs = &i2c_engine_funcs; - engine->base.base.funcs = &engine_funcs; -} - -struct i2c_engine *dal_i2c_sw_engine_create( - const struct i2c_sw_engine_create_arg *arg) -{ - struct i2c_sw_engine *engine; - - if (!arg) { - BREAK_TO_DEBUGGER(); - return NULL; - } - - engine = kzalloc(sizeof(struct i2c_sw_engine), GFP_KERNEL); - - if (!engine) { - BREAK_TO_DEBUGGER(); - return NULL; - } - - dal_i2c_sw_engine_construct(engine, arg); - return &engine->base; -} diff --git a/drivers/gpu/drm/amd/display/dc/i2caux/i2c_sw_engine.h b/drivers/gpu/drm/amd/display/dc/i2caux/i2c_sw_engine.h deleted file mode 100644 index 546f15b0d3f1..000000000000 --- a/drivers/gpu/drm/amd/display/dc/i2caux/i2c_sw_engine.h +++ /dev/null @@ -1,81 +0,0 @@ -/* - * Copyright 2012-15 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. - * - * Authors: AMD - * - */ - -#ifndef __DAL_I2C_SW_ENGINE_H__ -#define __DAL_I2C_SW_ENGINE_H__ - -enum { - I2C_SW_RETRIES = 10, - I2C_SW_SCL_READ_RETRIES = 128, - /* following value is in microseconds */ - I2C_SW_TIMEOUT_DELAY = 3000 -}; - -struct i2c_sw_engine; - -struct i2c_sw_engine { - struct i2c_engine base; - uint32_t clock_delay; - /* Values below are in KHz */ - uint32_t speed; - uint32_t default_speed; -}; - -struct i2c_sw_engine_create_arg { - uint32_t default_speed; - struct dc_context *ctx; -}; - -void dal_i2c_sw_engine_construct( - struct i2c_sw_engine *engine, - const struct i2c_sw_engine_create_arg *arg); - -bool dal_i2caux_i2c_sw_engine_acquire_engine( - struct i2c_engine *engine, - struct ddc *ddc_handle); - -void dal_i2c_sw_engine_destruct( - struct i2c_sw_engine *engine); - -struct i2c_engine *dal_i2c_sw_engine_create( - const struct i2c_sw_engine_create_arg *arg); -enum i2caux_engine_type dal_i2c_sw_engine_get_engine_type( - const struct engine *engine); -bool dal_i2c_sw_engine_submit_request( - struct engine *ptr, - struct i2caux_transaction_request *i2caux_request, - bool middle_of_transaction); -uint32_t dal_i2c_sw_engine_get_speed( - const struct i2c_engine *engine); -void dal_i2c_sw_engine_set_speed( - struct i2c_engine *ptr, - uint32_t speed); -void dal_i2c_sw_engine_submit_channel_request( - struct i2c_engine *ptr, - struct i2c_request_transaction_data *req); -enum i2c_channel_operation_result dal_i2c_sw_engine_get_channel_status( - struct i2c_engine *engine, - uint8_t *returned_bytes); -#endif diff --git a/drivers/gpu/drm/amd/display/dc/i2caux/i2caux.c b/drivers/gpu/drm/amd/display/dc/i2caux/i2caux.c deleted file mode 100644 index e56093f26eed..000000000000 --- a/drivers/gpu/drm/amd/display/dc/i2caux/i2caux.c +++ /dev/null @@ -1,490 +0,0 @@ -/* - * Copyright 2012-15 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. - * - * Authors: AMD - * - */ - -#include "dm_services.h" - -/* - * Pre-requisites: headers required by header of this unit - */ -#include "include/i2caux_interface.h" -#include "dc_bios_types.h" - -/* - * Header of this unit - */ - -#include "i2caux.h" - -/* - * Post-requisites: headers required by this unit - */ - -#include "engine.h" -#include "i2c_engine.h" -#include "aux_engine.h" - -/* - * This unit - */ - -#include "dce80/i2caux_dce80.h" - -#include "dce100/i2caux_dce100.h" - -#include "dce110/i2caux_dce110.h" - -#include "dce112/i2caux_dce112.h" - -#include "dce120/i2caux_dce120.h" - -#if defined(CONFIG_DRM_AMD_DC_DCN1_0) -#include "dcn10/i2caux_dcn10.h" -#endif - -#include "diagnostics/i2caux_diag.h" - -/* - * @brief - * Plain API, available publicly - */ - -struct i2caux *dal_i2caux_create( - struct dc_context *ctx) -{ - if (IS_FPGA_MAXIMUS_DC(ctx->dce_environment)) { - return dal_i2caux_diag_fpga_create(ctx); - } - - switch (ctx->dce_version) { - case DCE_VERSION_8_0: - case DCE_VERSION_8_1: - case DCE_VERSION_8_3: - return dal_i2caux_dce80_create(ctx); - case DCE_VERSION_11_2: - case DCE_VERSION_11_22: - return dal_i2caux_dce112_create(ctx); - case DCE_VERSION_11_0: - return dal_i2caux_dce110_create(ctx); - case DCE_VERSION_10_0: - return dal_i2caux_dce100_create(ctx); - case DCE_VERSION_12_0: - return dal_i2caux_dce120_create(ctx); -#if defined(CONFIG_DRM_AMD_DC_DCN1_0) - case DCN_VERSION_1_0: - return dal_i2caux_dcn10_create(ctx); -#endif - -#if defined(CONFIG_DRM_AMD_DC_DCN1_01) - case DCN_VERSION_1_01: - return dal_i2caux_dcn10_create(ctx); -#endif - default: - BREAK_TO_DEBUGGER(); - return NULL; - } -} - -bool dal_i2caux_submit_i2c_command( - struct i2caux *i2caux, - struct ddc *ddc, - struct i2c_command *cmd) -{ - struct i2c_engine *engine; - uint8_t index_of_payload = 0; - bool result; - - if (!ddc) { - BREAK_TO_DEBUGGER(); - return false; - } - - if (!cmd) { - BREAK_TO_DEBUGGER(); - return false; - } - - /* - * default will be SW, however there is a feature flag in adapter - * service that determines whether SW i2c_engine will be available or - * not, if sw i2c is not available we will fallback to hw. This feature - * flag is set to not creating sw i2c engine for every dce except dce80 - * currently - */ - switch (cmd->engine) { - case I2C_COMMAND_ENGINE_DEFAULT: - case I2C_COMMAND_ENGINE_SW: - /* try to acquire SW engine first, - * acquire HW engine if SW engine not available */ - engine = i2caux->funcs->acquire_i2c_sw_engine(i2caux, ddc); - - if (!engine) - engine = i2caux->funcs->acquire_i2c_hw_engine( - i2caux, ddc); - break; - case I2C_COMMAND_ENGINE_HW: - default: - /* try to acquire HW engine first, - * acquire SW engine if HW engine not available */ - engine = i2caux->funcs->acquire_i2c_hw_engine(i2caux, ddc); - - if (!engine) - engine = i2caux->funcs->acquire_i2c_sw_engine( - i2caux, ddc); - } - - if (!engine) - return false; - - engine->funcs->set_speed(engine, cmd->speed); - - result = true; - - while (index_of_payload < cmd->number_of_payloads) { - bool mot = (index_of_payload != cmd->number_of_payloads - 1); - - struct i2c_payload *payload = cmd->payloads + index_of_payload; - - struct i2caux_transaction_request request = { 0 }; - - request.operation = payload->write ? - I2CAUX_TRANSACTION_WRITE : - I2CAUX_TRANSACTION_READ; - - request.payload.address_space = - I2CAUX_TRANSACTION_ADDRESS_SPACE_I2C; - request.payload.address = (payload->address << 1) | - !payload->write; - request.payload.length = payload->length; - request.payload.data = payload->data; - - if (!engine->base.funcs->submit_request( - &engine->base, &request, mot)) { - result = false; - break; - } - - ++index_of_payload; - } - - i2caux->funcs->release_engine(i2caux, &engine->base); - - return result; -} - -bool dal_i2caux_submit_aux_command( - struct i2caux *i2caux, - struct ddc *ddc, - struct aux_command *cmd) -{ - struct aux_engine *engine; - uint8_t index_of_payload = 0; - bool result; - bool mot; - - if (!ddc) { - BREAK_TO_DEBUGGER(); - return false; - } - - if (!cmd) { - BREAK_TO_DEBUGGER(); - return false; - } - - engine = i2caux->funcs->acquire_aux_engine(i2caux, ddc); - - if (!engine) - return false; - - engine->delay = cmd->defer_delay; - engine->max_defer_write_retry = cmd->max_defer_write_retry; - - result = true; - - while (index_of_payload < cmd->number_of_payloads) { - struct aux_payload *payload = cmd->payloads + index_of_payload; - struct i2caux_transaction_request request = { 0 }; - - if (cmd->mot == I2C_MOT_UNDEF) - mot = (index_of_payload != cmd->number_of_payloads - 1); - else - mot = (cmd->mot == I2C_MOT_TRUE); - - request.operation = payload->write ? - I2CAUX_TRANSACTION_WRITE : - I2CAUX_TRANSACTION_READ; - - if (payload->i2c_over_aux) { - request.payload.address_space = - I2CAUX_TRANSACTION_ADDRESS_SPACE_I2C; - - request.payload.address = (payload->address << 1) | - !payload->write; - } else { - request.payload.address_space = - I2CAUX_TRANSACTION_ADDRESS_SPACE_DPCD; - - request.payload.address = payload->address; - } - - request.payload.length = payload->length; - request.payload.data = payload->data; - - if (!engine->base.funcs->submit_request( - &engine->base, &request, mot)) { - result = false; - break; - } - - ++index_of_payload; - } - - i2caux->funcs->release_engine(i2caux, &engine->base); - - return result; -} - -static bool get_hw_supported_ddc_line( - struct ddc *ddc, - enum gpio_ddc_line *line) -{ - enum gpio_ddc_line line_found; - - *line = GPIO_DDC_LINE_UNKNOWN; - - if (!ddc) { - BREAK_TO_DEBUGGER(); - return false; - } - - if (!ddc->hw_info.hw_supported) - return false; - - line_found = dal_ddc_get_line(ddc); - - if (line_found >= GPIO_DDC_LINE_COUNT) - return false; - - *line = line_found; - - return true; -} - -void dal_i2caux_configure_aux( - struct i2caux *i2caux, - struct ddc *ddc, - union aux_config cfg) -{ - struct aux_engine *engine = - i2caux->funcs->acquire_aux_engine(i2caux, ddc); - - if (!engine) - return; - - engine->funcs->configure(engine, cfg); - - i2caux->funcs->release_engine(i2caux, &engine->base); -} - -void dal_i2caux_destroy( - struct i2caux **i2caux) -{ - if (!i2caux || !*i2caux) { - BREAK_TO_DEBUGGER(); - return; - } - - (*i2caux)->funcs->destroy(i2caux); - - *i2caux = NULL; -} - -/* - * @brief - * An utility function used by 'struct i2caux' and its descendants - */ - -uint32_t dal_i2caux_get_reference_clock( - struct dc_bios *bios) -{ - struct dc_firmware_info info = { { 0 } }; - - if (bios->funcs->get_firmware_info(bios, &info) != BP_RESULT_OK) - return 0; - - return info.pll_info.crystal_frequency; -} - -/* - * @brief - * i2caux - */ - -enum { - /* following are expressed in KHz */ - DEFAULT_I2C_SW_SPEED = 50, - DEFAULT_I2C_HW_SPEED = 50, - - DEFAULT_I2C_SW_SPEED_100KHZ = 100, - DEFAULT_I2C_HW_SPEED_100KHZ = 100, - - /* This is the timeout as defined in DP 1.2a, - * 2.3.4 "Detailed uPacket TX AUX CH State Description". */ - AUX_TIMEOUT_PERIOD = 400, - - /* Ideally, the SW timeout should be just above 550usec - * which is programmed in HW. - * But the SW timeout of 600usec is not reliable, - * because on some systems, delay_in_microseconds() - * returns faster than it should. - * EPR #379763: by trial-and-error on different systems, - * 700usec is the minimum reliable SW timeout for polling - * the AUX_SW_STATUS.AUX_SW_DONE bit. - * This timeout expires *only* when there is - * AUX Error or AUX Timeout conditions - not during normal operation. - * During normal operation, AUX_SW_STATUS.AUX_SW_DONE bit is set - * at most within ~240usec. That means, - * increasing this timeout will not affect normal operation, - * and we'll timeout after - * SW_AUX_TIMEOUT_PERIOD_MULTIPLIER * AUX_TIMEOUT_PERIOD = 1600usec. - * This timeout is especially important for - * resume from S3 and CTS. */ - SW_AUX_TIMEOUT_PERIOD_MULTIPLIER = 4 -}; - -struct i2c_engine *dal_i2caux_acquire_i2c_sw_engine( - struct i2caux *i2caux, - struct ddc *ddc) -{ - enum gpio_ddc_line line; - struct i2c_engine *engine = NULL; - - if (get_hw_supported_ddc_line(ddc, &line)) - engine = i2caux->i2c_sw_engines[line]; - - if (!engine) - engine = i2caux->i2c_generic_sw_engine; - - if (!engine) - return NULL; - - if (!engine->base.funcs->acquire(&engine->base, ddc)) - return NULL; - - return engine; -} - -struct aux_engine *dal_i2caux_acquire_aux_engine( - struct i2caux *i2caux, - struct ddc *ddc) -{ - enum gpio_ddc_line line; - struct aux_engine *engine; - - if (!get_hw_supported_ddc_line(ddc, &line)) - return NULL; - - engine = i2caux->aux_engines[line]; - - if (!engine) - return NULL; - - if (!engine->base.funcs->acquire(&engine->base, ddc)) - return NULL; - - return engine; -} - -void dal_i2caux_release_engine( - struct i2caux *i2caux, - struct engine *engine) -{ - engine->funcs->release_engine(engine); - - dal_ddc_close(engine->ddc); - - engine->ddc = NULL; -} - -void dal_i2caux_construct( - struct i2caux *i2caux, - struct dc_context *ctx) -{ - uint32_t i = 0; - - i2caux->ctx = ctx; - do { - i2caux->i2c_sw_engines[i] = NULL; - i2caux->i2c_hw_engines[i] = NULL; - i2caux->aux_engines[i] = NULL; - - ++i; - } while (i < GPIO_DDC_LINE_COUNT); - - i2caux->i2c_generic_sw_engine = NULL; - i2caux->i2c_generic_hw_engine = NULL; - - i2caux->aux_timeout_period = - SW_AUX_TIMEOUT_PERIOD_MULTIPLIER * AUX_TIMEOUT_PERIOD; - - if (ctx->dce_version >= DCE_VERSION_11_2) { - i2caux->default_i2c_hw_speed = DEFAULT_I2C_HW_SPEED_100KHZ; - i2caux->default_i2c_sw_speed = DEFAULT_I2C_SW_SPEED_100KHZ; - } else { - i2caux->default_i2c_hw_speed = DEFAULT_I2C_HW_SPEED; - i2caux->default_i2c_sw_speed = DEFAULT_I2C_SW_SPEED; - } -} - -void dal_i2caux_destruct( - struct i2caux *i2caux) -{ - uint32_t i = 0; - - if (i2caux->i2c_generic_hw_engine) - i2caux->i2c_generic_hw_engine->funcs->destroy( - &i2caux->i2c_generic_hw_engine); - - if (i2caux->i2c_generic_sw_engine) - i2caux->i2c_generic_sw_engine->funcs->destroy( - &i2caux->i2c_generic_sw_engine); - - do { - if (i2caux->aux_engines[i]) - i2caux->aux_engines[i]->funcs->destroy( - &i2caux->aux_engines[i]); - - if (i2caux->i2c_hw_engines[i]) - i2caux->i2c_hw_engines[i]->funcs->destroy( - &i2caux->i2c_hw_engines[i]); - - if (i2caux->i2c_sw_engines[i]) - i2caux->i2c_sw_engines[i]->funcs->destroy( - &i2caux->i2c_sw_engines[i]); - - ++i; - } while (i < GPIO_DDC_LINE_COUNT); -} - diff --git a/drivers/gpu/drm/amd/display/dc/i2caux/i2caux.h b/drivers/gpu/drm/amd/display/dc/i2caux/i2caux.h deleted file mode 100644 index 64f51bb06915..000000000000 --- a/drivers/gpu/drm/amd/display/dc/i2caux/i2caux.h +++ /dev/null @@ -1,122 +0,0 @@ -/* - * Copyright 2012-15 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. - * - * Authors: AMD - * - */ - -#ifndef __DAL_I2C_AUX_H__ -#define __DAL_I2C_AUX_H__ - -uint32_t dal_i2caux_get_reference_clock( - struct dc_bios *bios); - -struct i2caux; - -struct engine; - -struct i2caux_funcs { - void (*destroy)(struct i2caux **ptr); - struct i2c_engine * (*acquire_i2c_sw_engine)( - struct i2caux *i2caux, - struct ddc *ddc); - struct i2c_engine * (*acquire_i2c_hw_engine)( - struct i2caux *i2caux, - struct ddc *ddc); - struct aux_engine * (*acquire_aux_engine)( - struct i2caux *i2caux, - struct ddc *ddc); - void (*release_engine)( - struct i2caux *i2caux, - struct engine *engine); -}; - -struct i2c_engine; -struct aux_engine; - -struct i2caux { - struct dc_context *ctx; - const struct i2caux_funcs *funcs; - /* On ASIC we have certain amount of lines with HW DDC engine - * (4, 6, or maybe more in the future). - * For every such line, we create separate HW DDC engine - * (since we have these engines in HW) and separate SW DDC engine - * (to allow concurrent use of few lines). - * In similar way we have AUX engines. */ - - /* I2C SW engines, per DDC line. - * Only lines with HW DDC support will be initialized */ - struct i2c_engine *i2c_sw_engines[GPIO_DDC_LINE_COUNT]; - - /* I2C HW engines, per DDC line. - * Only lines with HW DDC support will be initialized */ - struct i2c_engine *i2c_hw_engines[GPIO_DDC_LINE_COUNT]; - - /* AUX engines, per DDC line. - * Only lines with HW AUX support will be initialized */ - struct aux_engine *aux_engines[GPIO_DDC_LINE_COUNT]; - - /* For all other lines, we can use - * single instance of generic I2C HW engine - * (since in HW, there is single instance of it) - * or single instance of generic I2C SW engine. - * AUX is not supported for other lines. */ - - /* General-purpose I2C SW engine. - * Can be assigned dynamically to any line per transaction */ - struct i2c_engine *i2c_generic_sw_engine; - - /* General-purpose I2C generic HW engine. - * Can be assigned dynamically to almost any line per transaction */ - struct i2c_engine *i2c_generic_hw_engine; - - /* [anaumov] in DAL2, there is a Mutex */ - - uint32_t aux_timeout_period; - - /* expressed in KHz */ - uint32_t default_i2c_sw_speed; - uint32_t default_i2c_hw_speed; -}; - -void dal_i2caux_construct( - struct i2caux *i2caux, - struct dc_context *ctx); - -void dal_i2caux_release_engine( - struct i2caux *i2caux, - struct engine *engine); - -void dal_i2caux_destruct( - struct i2caux *i2caux); - -void dal_i2caux_destroy( - struct i2caux **ptr); - -struct i2c_engine *dal_i2caux_acquire_i2c_sw_engine( - struct i2caux *i2caux, - struct ddc *ddc); - -struct aux_engine *dal_i2caux_acquire_aux_engine( - struct i2caux *i2caux, - struct ddc *ddc); - -#endif diff --git a/drivers/gpu/drm/amd/display/dc/inc/bw_fixed.h b/drivers/gpu/drm/amd/display/dc/inc/bw_fixed.h index 39ee8eba3c31..d1656c9d50df 100644 --- a/drivers/gpu/drm/amd/display/dc/inc/bw_fixed.h +++ b/drivers/gpu/drm/amd/display/dc/inc/bw_fixed.h @@ -126,7 +126,7 @@ static inline struct bw_fixed bw_div(const struct bw_fixed arg1, const struct bw static inline struct bw_fixed bw_mod(const struct bw_fixed arg1, const struct bw_fixed arg2) { struct bw_fixed res; - div64_u64_rem(arg1.value, arg2.value, &res.value); + div64_u64_rem(arg1.value, arg2.value, (uint64_t *)&res.value); return res; } diff --git a/drivers/gpu/drm/amd/display/dc/inc/clock_source.h b/drivers/gpu/drm/amd/display/dc/inc/clock_source.h index 47ef90495376..fe6301cb8681 100644 --- a/drivers/gpu/drm/amd/display/dc/inc/clock_source.h +++ b/drivers/gpu/drm/amd/display/dc/inc/clock_source.h @@ -78,7 +78,7 @@ struct csdp_ref_clk_ds_params { }; struct pixel_clk_params { - uint32_t requested_pix_clk; /* in KHz */ + uint32_t requested_pix_clk_100hz; /*> Requested Pixel Clock * (based on Video Timing standard used for requested mode)*/ uint32_t requested_sym_clk; /* in KHz */ @@ -104,9 +104,9 @@ struct pixel_clk_params { * with actually calculated Clock and reference Crystal frequency */ struct pll_settings { - uint32_t actual_pix_clk; - uint32_t adjusted_pix_clk; - uint32_t calculated_pix_clk; + uint32_t actual_pix_clk_100hz; + uint32_t adjusted_pix_clk_100hz; + uint32_t calculated_pix_clk_100hz; uint32_t vco_freq; uint32_t reference_freq; uint32_t reference_divider; @@ -166,6 +166,10 @@ struct clock_source_funcs { struct clock_source *, struct pixel_clk_params *, struct pll_settings *); + bool (*get_pixel_clk_frequency_100hz)( + struct clock_source *clock_source, + unsigned int inst, + unsigned int *pixel_clk_khz); }; struct clock_source { diff --git a/drivers/gpu/drm/amd/display/dc/inc/compressor.h b/drivers/gpu/drm/amd/display/dc/inc/compressor.h index bcb18f5e1e60..7a147a9762a0 100644 --- a/drivers/gpu/drm/amd/display/dc/inc/compressor.h +++ b/drivers/gpu/drm/amd/display/dc/inc/compressor.h @@ -77,6 +77,7 @@ struct compressor_funcs { }; struct compressor { struct dc_context *ctx; + /* CONTROLLER_ID_D0 + instance, CONTROLLER_ID_UNDEFINED = 0 */ uint32_t attached_inst; bool is_enabled; const struct compressor_funcs *funcs; diff --git a/drivers/gpu/drm/amd/display/dc/inc/core_status.h b/drivers/gpu/drm/amd/display/dc/inc/core_status.h index 94fc31080fda..2e61a22ef4b2 100644 --- a/drivers/gpu/drm/amd/display/dc/inc/core_status.h +++ b/drivers/gpu/drm/amd/display/dc/inc/core_status.h @@ -30,7 +30,7 @@ enum dc_status { DC_OK = 1, DC_NO_CONTROLLER_RESOURCE = 2, - DC_NO_STREAM_ENG_RESOURCE = 3, + DC_NO_STREAM_ENC_RESOURCE = 3, DC_NO_CLOCK_SOURCE_RESOURCE = 4, DC_FAIL_CONTROLLER_VALIDATE = 5, DC_FAIL_ENC_VALIDATE = 6, diff --git a/drivers/gpu/drm/amd/display/dc/inc/core_types.h b/drivers/gpu/drm/amd/display/dc/inc/core_types.h index c1976c175b57..986ed1728644 100644 --- a/drivers/gpu/drm/amd/display/dc/inc/core_types.h +++ b/drivers/gpu/drm/amd/display/dc/inc/core_types.h @@ -82,7 +82,7 @@ void core_link_disable_stream(struct pipe_ctx *pipe_ctx, int option); void core_link_set_avmute(struct pipe_ctx *pipe_ctx, bool enable); /********** DAL Core*********************/ -#include "display_clock.h" +#include "hw/clk_mgr.h" #include "transform.h" #include "dpp.h" @@ -146,7 +146,7 @@ struct resource_pool { struct mpc *mpc; struct pp_smu_funcs_rv *pp_smu; struct pp_smu_display_requirement_rv pp_smu_req; - struct aux_engine *engines[MAX_PIPES]; + struct dce_aux *engines[MAX_PIPES]; struct dce_i2c_hw *hw_i2cs[MAX_PIPES]; struct dce_i2c_sw *sw_i2cs[MAX_PIPES]; bool i2c_hw_buffer_in_use; @@ -169,6 +169,7 @@ struct resource_pool { unsigned int audio_count; struct audio_support audio_support; + struct clk_mgr *clk_mgr; struct dccg *dccg; struct irq_service *irqs; @@ -179,13 +180,8 @@ struct resource_pool { const struct resource_caps *res_cap; }; -struct dcn_fe_clocks { - int dppclk_khz; -}; - struct dcn_fe_bandwidth { - struct dcn_fe_clocks calc; - struct dcn_fe_clocks cur; + int dppclk_khz; }; struct stream_resource { @@ -271,6 +267,17 @@ union bw_context { struct dce_bw_output dce; }; +/** + * struct dc_state - The full description of a state requested by a user + * + * @streams: Stream properties + * @stream_status: The planes on a given stream + * @res_ctx: Persistent state of resources + * @bw: The output from bandwidth and watermark calculations + * @pp_display_cfg: PowerPlay clocks and settings + * @dcn_bw_vars: non-stack memory to support bandwidth calculations + * + */ struct dc_state { struct dc_stream_state *streams[MAX_PIPES]; struct dc_stream_status stream_status[MAX_PIPES]; @@ -278,7 +285,6 @@ struct dc_state { struct resource_context res_ctx; - /* The output from BW and WM calculations. */ union bw_context bw; /* Note: these are big structures, do *not* put on stack! */ @@ -287,7 +293,7 @@ struct dc_state { struct dcn_bw_internal_vars dcn_bw_vars; #endif - struct dccg *dis_clk; + struct clk_mgr *dccg; struct kref refcount; }; diff --git a/drivers/gpu/drm/amd/display/dc/inc/dc_link_ddc.h b/drivers/gpu/drm/amd/display/dc/inc/dc_link_ddc.h index 538b83303b86..16fd4dc6c4dd 100644 --- a/drivers/gpu/drm/amd/display/dc/inc/dc_link_ddc.h +++ b/drivers/gpu/drm/amd/display/dc/inc/dc_link_ddc.h @@ -64,13 +64,6 @@ void dal_ddc_i2c_payloads_add( uint8_t *data, bool write); -void dal_ddc_aux_payloads_add( - struct aux_payloads *payloads, - uint32_t address, - uint32_t len, - uint8_t *data, - bool write); - struct ddc_service_init_data { struct graphics_object_id id; struct dc_context *ctx; @@ -103,12 +96,10 @@ bool dal_ddc_service_query_ddc_data( uint32_t read_size); int dc_link_aux_transfer(struct ddc_service *ddc, - unsigned int address, - uint8_t *reply, - void *buffer, - unsigned int size, - enum aux_transaction_type type, - enum i2caux_transaction_action action); + struct aux_payload *payload); + +bool dc_link_aux_transfer_with_retries(struct ddc_service *ddc, + struct aux_payload *payload); void dal_ddc_service_write_scdc_data( struct ddc_service *ddc_service, diff --git a/drivers/gpu/drm/amd/display/dc/inc/dcn_calcs.h b/drivers/gpu/drm/amd/display/dc/inc/dcn_calcs.h index e688eb9b975c..ece954a40a8e 100644 --- a/drivers/gpu/drm/amd/display/dc/inc/dcn_calcs.h +++ b/drivers/gpu/drm/amd/display/dc/inc/dcn_calcs.h @@ -31,8 +31,8 @@ #define __DCN_CALCS_H__ #include "bw_fixed.h" -#include "display_clock.h" #include "../dml/display_mode_lib.h" +#include "hw/clk_mgr.h" struct dc; struct dc_state; diff --git a/drivers/gpu/drm/amd/display/dc/inc/hw/abm.h b/drivers/gpu/drm/amd/display/dc/inc/hw/abm.h index a83a48494613..86dc39a02408 100644 --- a/drivers/gpu/drm/amd/display/dc/inc/hw/abm.h +++ b/drivers/gpu/drm/amd/display/dc/inc/hw/abm.h @@ -46,13 +46,20 @@ struct abm_funcs { void (*abm_init)(struct abm *abm); bool (*set_abm_level)(struct abm *abm, unsigned int abm_level); bool (*set_abm_immediate_disable)(struct abm *abm); + bool (*set_pipe)(struct abm *abm, unsigned int controller_id); bool (*init_backlight)(struct abm *abm); - bool (*set_backlight_level)(struct abm *abm, - unsigned int backlight_level, + + /* backlight_pwm_u16_16 is unsigned 32 bit, + * 16 bit integer + 16 fractional, where 1.0 is max backlight value. + */ + bool (*set_backlight_level_pwm)(struct abm *abm, + unsigned int backlight_pwm_u16_16, unsigned int frame_ramp, unsigned int controller_id, bool use_smooth_brightness); - unsigned int (*get_current_backlight_8_bit)(struct abm *abm); + + unsigned int (*get_current_backlight)(struct abm *abm); + unsigned int (*get_target_backlight)(struct abm *abm); }; #endif diff --git a/drivers/gpu/drm/amd/display/dc/i2caux/diagnostics/i2caux_diag.h b/drivers/gpu/drm/amd/display/dc/inc/hw/clk_mgr.h index a83eeb748283..23a4b18e5fee 100644 --- a/drivers/gpu/drm/amd/display/dc/i2caux/diagnostics/i2caux_diag.h +++ b/drivers/gpu/drm/amd/display/dc/inc/hw/clk_mgr.h @@ -23,10 +23,25 @@ * */ -#ifndef __DAL_I2C_AUX_DIAG_FPGA_H__ -#define __DAL_I2C_AUX_DIAG_FPGA_H__ +#ifndef __DAL_CLK_MGR_H__ +#define __DAL_CLK_MGR_H__ -struct i2caux *dal_i2caux_diag_fpga_create( - struct dc_context *ctx); +#include "dm_services_types.h" +#include "dc.h" -#endif /* __DAL_I2C_AUX_DIAG_FPGA_H__ */ +struct clk_mgr { + struct dc_context *ctx; + const struct clk_mgr_funcs *funcs; + + struct dc_clocks clks; +}; + +struct clk_mgr_funcs { + void (*update_clocks)(struct clk_mgr *clk_mgr, + struct dc_state *context, + bool safe_to_lower); + + int (*get_dp_ref_clk_frequency)(struct clk_mgr *clk_mgr); +}; + +#endif /* __DAL_CLK_MGR_H__ */ diff --git a/drivers/gpu/drm/amd/display/dc/i2caux/dce80/i2caux_dce80.h b/drivers/gpu/drm/amd/display/dc/inc/hw/dccg.h index 21908629e973..95a56d012626 100644 --- a/drivers/gpu/drm/amd/display/dc/i2caux/dce80/i2caux_dce80.h +++ b/drivers/gpu/drm/amd/display/dc/inc/hw/dccg.h @@ -1,5 +1,5 @@ /* - * Copyright 2012-15 Advanced Micro Devices, Inc. + * Copyright 2018 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"), @@ -23,16 +23,22 @@ * */ -#ifndef __DAL_I2C_AUX_DCE80_H__ -#define __DAL_I2C_AUX_DCE80_H__ +#ifndef __DAL_DCCG_H__ +#define __DAL_DCCG_H__ -struct i2caux_dce80 { - struct i2caux base; - /* indicate the I2C HW circular buffer is in use */ - bool i2c_hw_buffer_in_use; +#include "dc_types.h" + +struct dccg { + struct dc_context *ctx; + const struct dccg_funcs *funcs; + + int ref_dppclk; }; -struct i2caux *dal_i2caux_dce80_create( - struct dc_context *ctx); +struct dccg_funcs { + void (*update_dpp_dto)(struct dccg *dccg, + int dpp_inst, + int req_dppclk); +}; -#endif +#endif //__DAL_DCCG_H__ diff --git a/drivers/gpu/drm/amd/display/dc/inc/hw/dchubbub.h b/drivers/gpu/drm/amd/display/dc/inc/hw/dchubbub.h index 02f757dd70d4..9d2d8e51306c 100644 --- a/drivers/gpu/drm/amd/display/dc/inc/hw/dchubbub.h +++ b/drivers/gpu/drm/amd/display/dc/inc/hw/dchubbub.h @@ -39,6 +39,18 @@ enum segment_order { segment_order__non_contiguous, }; +struct dcn_hubbub_wm_set { + uint32_t wm_set; + uint32_t data_urgent; + uint32_t pte_meta_urgent; + uint32_t sr_enter; + uint32_t sr_exit; + uint32_t dram_clk_chanage; +}; + +struct dcn_hubbub_wm { + struct dcn_hubbub_wm_set sets[4]; +}; struct hubbub_funcs { void (*update_dchub)( @@ -58,7 +70,14 @@ struct hubbub_funcs { bool (*dcc_support_pixel_format)( enum surface_pixel_format format, unsigned int *bytes_per_element); + + void (*wm_read_state)(struct hubbub *hubbub, + struct dcn_hubbub_wm *wm); }; +struct hubbub { + const struct hubbub_funcs *funcs; + struct dc_context *ctx; +}; #endif diff --git a/drivers/gpu/drm/amd/display/dc/inc/hw/display_clock.h b/drivers/gpu/drm/amd/display/dc/inc/hw/display_clock.h deleted file mode 100644 index 689faa16c0ae..000000000000 --- a/drivers/gpu/drm/amd/display/dc/inc/hw/display_clock.h +++ /dev/null @@ -1,63 +0,0 @@ -/* - * Copyright 2012-16 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. - * - * Authors: AMD - * - */ - -#ifndef __DISPLAY_CLOCK_H__ -#define __DISPLAY_CLOCK_H__ - -#include "dm_services_types.h" -#include "dc.h" - -/* Structure containing all state-dependent clocks - * (dependent on "enum clocks_state") */ -struct state_dependent_clocks { - int display_clk_khz; - int pixel_clk_khz; -}; - -struct dccg { - struct dc_context *ctx; - const struct display_clock_funcs *funcs; - - enum dm_pp_clocks_state max_clks_state; - enum dm_pp_clocks_state cur_min_clks_state; - struct dc_clocks clks; -}; - -struct display_clock_funcs { - void (*update_clocks)(struct dccg *dccg, - struct dc_clocks *new_clocks, - bool safe_to_lower); - int (*set_dispclk)(struct dccg *dccg, - int requested_clock_khz); - - int (*get_dp_ref_clk_frequency)(struct dccg *dccg); - - bool (*update_dfs_bypass)(struct dccg *dccg, - struct dc *dc, - struct dc_state *context, - int requested_clock_khz); -}; - -#endif /* __DISPLAY_CLOCK_H__ */ diff --git a/drivers/gpu/drm/amd/display/dc/inc/hw/dmcu.h b/drivers/gpu/drm/amd/display/dc/inc/hw/dmcu.h index 4550747fb61c..cbaa43853611 100644 --- a/drivers/gpu/drm/amd/display/dc/inc/hw/dmcu.h +++ b/drivers/gpu/drm/amd/display/dc/inc/hw/dmcu.h @@ -27,9 +27,22 @@ #include "dm_services_types.h" +/* If HW itself ever powered down it will be 0. + * fwDmcuInit will write to 1. + * Driver will only call MCP init if current state is 1, + * and the MCP command will transition this to 2. + */ enum dmcu_state { - DMCU_NOT_INITIALIZED = 0, - DMCU_RUNNING = 1 + DMCU_UNLOADED = 0, + DMCU_LOADED_UNINITIALIZED = 1, + DMCU_RUNNING = 2, +}; + +struct dmcu_version { + unsigned int interface_version; + unsigned int abm_version; + unsigned int psr_version; + unsigned int build_version; }; struct dmcu { diff --git a/drivers/gpu/drm/amd/display/dc/inc/hw/dpp.h b/drivers/gpu/drm/amd/display/dc/inc/hw/dpp.h index e894e649ce5a..fb7967b39edb 100644 --- a/drivers/gpu/drm/amd/display/dc/inc/hw/dpp.h +++ b/drivers/gpu/drm/amd/display/dc/inc/hw/dpp.h @@ -39,6 +39,11 @@ struct dpp { }; +struct dpp_input_csc_matrix { + enum dc_color_space color_space; + uint16_t regval[12]; +}; + struct dpp_grph_csc_adjustment { struct fixed31_32 temperature_matrix[CSC_TEMPERATURE_MATRIX_SIZE]; enum graphics_gamut_adjust_type gamut_adjust_type; diff --git a/drivers/gpu/drm/amd/display/dc/inc/hw/hubp.h b/drivers/gpu/drm/amd/display/dc/inc/hw/hubp.h index 334c48cdafdc..1cd07e94ee63 100644 --- a/drivers/gpu/drm/amd/display/dc/inc/hw/hubp.h +++ b/drivers/gpu/drm/amd/display/dc/inc/hw/hubp.h @@ -63,6 +63,11 @@ struct hubp_funcs { struct _vcs_dpi_display_rq_regs_st *rq_regs, struct _vcs_dpi_display_pipe_dest_params_st *pipe_dest); + void (*hubp_setup_interdependent)( + struct hubp *hubp, + struct _vcs_dpi_display_dlg_regs_st *dlg_regs, + struct _vcs_dpi_display_ttu_regs_st *ttu_regs); + void (*dcc_control)(struct hubp *hubp, bool enable, bool independent_64b_blks); void (*mem_program_viewport)( @@ -73,7 +78,8 @@ struct hubp_funcs { bool (*hubp_program_surface_flip_and_addr)( struct hubp *hubp, const struct dc_plane_address *address, - bool flip_immediate); + bool flip_immediate, + uint8_t vmid); void (*hubp_program_pte_vm)( struct hubp *hubp, @@ -121,6 +127,7 @@ struct hubp_funcs { void (*hubp_clk_cntl)(struct hubp *hubp, bool enable); void (*hubp_vtg_sel)(struct hubp *hubp, uint32_t otg_inst); void (*hubp_read_state)(struct hubp *hubp); + void (*hubp_clear_underflow)(struct hubp *hubp); void (*hubp_disable_control)(struct hubp *hubp, bool disable_hubp); unsigned int (*hubp_get_underflow_status)(struct hubp *hubp); diff --git a/drivers/gpu/drm/amd/display/dc/inc/hw/hw_shared.h b/drivers/gpu/drm/amd/display/dc/inc/hw/hw_shared.h index cf7433ebf91a..da85537a4488 100644 --- a/drivers/gpu/drm/amd/display/dc/inc/hw/hw_shared.h +++ b/drivers/gpu/drm/amd/display/dc/inc/hw/hw_shared.h @@ -53,6 +53,12 @@ struct curve_points { uint32_t custom_float_slope; }; +struct curve_points3 { + struct curve_points red; + struct curve_points green; + struct curve_points blue; +}; + struct pwl_result_data { struct fixed31_32 red; struct fixed31_32 green; @@ -71,9 +77,17 @@ struct pwl_result_data { uint32_t delta_blue_reg; }; +/* arr_curve_points - regamma regions/segments specification + * arr_points - beginning and end point specified separately (only one on DCE) + * corner_points - beginning and end point for all 3 colors (DCN) + * rgb_resulted - final curve + */ struct pwl_params { struct gamma_curve arr_curve_points[34]; - struct curve_points arr_points[2]; + union { + struct curve_points arr_points[2]; + struct curve_points3 corner_points[2]; + }; struct pwl_result_data rgb_resulted[256 + 3]; uint32_t hw_points_num; }; diff --git a/drivers/gpu/drm/amd/display/dc/inc/hw/link_encoder.h b/drivers/gpu/drm/amd/display/dc/inc/hw/link_encoder.h index e28e9770e0a3..c9d3e37e9531 100644 --- a/drivers/gpu/drm/amd/display/dc/inc/hw/link_encoder.h +++ b/drivers/gpu/drm/amd/display/dc/inc/hw/link_encoder.h @@ -65,7 +65,8 @@ struct encoder_feature_support { enum dc_color_depth max_hdmi_deep_color; unsigned int max_hdmi_pixel_clock; - bool ycbcr420_supported; + bool hdmi_ycbcr420_supported; + bool dp_ycbcr420_supported; }; union dpcd_psr_configuration { @@ -152,6 +153,7 @@ struct link_encoder_funcs { void (*enable_hpd)(struct link_encoder *enc); void (*disable_hpd)(struct link_encoder *enc); bool (*is_dig_enabled)(struct link_encoder *enc); + unsigned int (*get_dig_frontend)(struct link_encoder *enc); void (*destroy)(struct link_encoder **enc); }; diff --git a/drivers/gpu/drm/amd/display/dc/inc/hw/stream_encoder.h b/drivers/gpu/drm/amd/display/dc/inc/hw/stream_encoder.h index 53a9b64df11a..4051493557bc 100644 --- a/drivers/gpu/drm/amd/display/dc/inc/hw/stream_encoder.h +++ b/drivers/gpu/drm/amd/display/dc/inc/hw/stream_encoder.h @@ -161,6 +161,10 @@ struct stream_encoder_funcs { void (*set_avmute)( struct stream_encoder *enc, bool enable); + void (*dig_connect_to_otg)( + struct stream_encoder *enc, + int tg_inst); + }; #endif /* STREAM_ENCODER_H_ */ diff --git a/drivers/gpu/drm/amd/display/dc/inc/hw/timing_generator.h b/drivers/gpu/drm/amd/display/dc/inc/hw/timing_generator.h index af700c7dac50..c25f7df7b5e3 100644 --- a/drivers/gpu/drm/amd/display/dc/inc/hw/timing_generator.h +++ b/drivers/gpu/drm/amd/display/dc/inc/hw/timing_generator.h @@ -134,15 +134,24 @@ struct dc_crtc_timing; struct drr_params; + struct timing_generator_funcs { bool (*validate_timing)(struct timing_generator *tg, const struct dc_crtc_timing *timing); void (*program_timing)(struct timing_generator *tg, const struct dc_crtc_timing *timing, bool use_vbios); - void (*program_vline_interrupt)(struct timing_generator *optc, - const struct dc_crtc_timing *dc_crtc_timing, - unsigned long long vsync_delta); + void (*setup_vertical_interrupt0)( + struct timing_generator *optc, + uint32_t start_line, + uint32_t end_line); + void (*setup_vertical_interrupt1)( + struct timing_generator *optc, + uint32_t start_line); + void (*setup_vertical_interrupt2)( + struct timing_generator *optc, + uint32_t start_line); + bool (*enable_crtc)(struct timing_generator *tg); bool (*disable_crtc)(struct timing_generator *tg); bool (*is_counter_moving)(struct timing_generator *tg); @@ -159,6 +168,8 @@ struct timing_generator_funcs { bool (*get_otg_active_size)(struct timing_generator *optc, uint32_t *otg_active_width, uint32_t *otg_active_height); + bool (*is_matching_timing)(struct timing_generator *tg, + const struct dc_crtc_timing *otg_timing); void (*set_early_control)(struct timing_generator *tg, uint32_t early_cntl); void (*wait_for_state)(struct timing_generator *tg, diff --git a/drivers/gpu/drm/amd/display/dc/inc/hw/vmid.h b/drivers/gpu/drm/amd/display/dc/inc/hw/vmid.h new file mode 100644 index 000000000000..037beb0a2a27 --- /dev/null +++ b/drivers/gpu/drm/amd/display/dc/inc/hw/vmid.h @@ -0,0 +1,49 @@ +/* + * Copyright 2018 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. + * + * Authors: AMD + * + */ + +#ifndef DAL_DC_INC_HW_VMID_H_ +#define DAL_DC_INC_HW_VMID_H_ + +#include "core_types.h" +#include "dchubbub.h" + +struct dcn_vmid_registers { + uint32_t CNTL; + uint32_t PAGE_TABLE_BASE_ADDR_HI32; + uint32_t PAGE_TABLE_BASE_ADDR_LO32; + uint32_t PAGE_TABLE_START_ADDR_HI32; + uint32_t PAGE_TABLE_START_ADDR_LO32; + uint32_t PAGE_TABLE_END_ADDR_HI32; + uint32_t PAGE_TABLE_END_ADDR_LO32; +}; + +struct dcn_vmid_page_table_config { + uint64_t page_table_start_addr; + uint64_t page_table_end_addr; + enum dcn_hubbub_page_table_depth depth; + enum dcn_hubbub_page_table_block_size block_size; +}; + +#endif /* DAL_DC_INC_HW_VMID_H_ */ diff --git a/drivers/gpu/drm/amd/display/dc/inc/hw_sequencer.h b/drivers/gpu/drm/amd/display/dc/inc/hw_sequencer.h index 26f29d5da3d8..7676f25216b1 100644 --- a/drivers/gpu/drm/amd/display/dc/inc/hw_sequencer.h +++ b/drivers/gpu/drm/amd/display/dc/inc/hw_sequencer.h @@ -32,14 +32,17 @@ #include "inc/hw/link_encoder.h" #include "core_status.h" -#define EDP_BACKLIGHT_RAMP_DISABLE_LEVEL 0xFFFFFFFF - enum pipe_gating_control { PIPE_GATING_CONTROL_DISABLE = 0, PIPE_GATING_CONTROL_ENABLE, PIPE_GATING_CONTROL_INIT }; +enum vline_select { + VLINE0, + VLINE1 +}; + struct dce_hwseq_wa { bool blnd_crtc_trigger; bool DEGVIDCN10_253; @@ -70,8 +73,14 @@ struct stream_resource; struct hw_sequencer_funcs { + void (*disable_stream_gating)(struct dc *dc, struct pipe_ctx *pipe_ctx); + + void (*enable_stream_gating)(struct dc *dc, struct pipe_ctx *pipe_ctx); + void (*init_hw)(struct dc *dc); + void (*init_pipes)(struct dc *dc, struct dc_state *context); + enum dc_status (*apply_ctx_to_hw)( struct dc *dc, struct dc_state *context); @@ -87,11 +96,6 @@ struct hw_sequencer_funcs { void (*program_gamut_remap)( struct pipe_ctx *pipe_ctx); - void (*program_csc_matrix)( - struct pipe_ctx *pipe_ctx, - enum dc_color_space colorspace, - uint16_t *matrix); - void (*program_output_csc)(struct dc *dc, struct pipe_ctx *pipe_ctx, enum dc_color_space colorspace, @@ -177,10 +181,12 @@ struct hw_sequencer_funcs { struct pipe_ctx *pipe_ctx, bool blank); - void (*set_bandwidth)( + void (*prepare_bandwidth)( struct dc *dc, - struct dc_state *context, - bool safe_to_lower); + struct dc_state *context); + void (*optimize_bandwidth)( + struct dc *dc, + struct dc_state *context); void (*set_drr)(struct pipe_ctx **pipe_ctx, int num_pipes, int vmin, int vmax); @@ -205,16 +211,12 @@ struct hw_sequencer_funcs { void (*log_hw_state)(struct dc *dc, struct dc_log_buffer_ctx *log_ctx); void (*get_hw_state)(struct dc *dc, char *pBuf, unsigned int bufSize, unsigned int mask); + void (*clear_status_bits)(struct dc *dc, unsigned int mask); void (*wait_for_mpcc_disconnect)(struct dc *dc, struct resource_pool *res_pool, struct pipe_ctx *pipe_ctx); - void (*ready_shared_resources)(struct dc *dc, struct dc_state *context); - void (*optimize_shared_resources)(struct dc *dc); - void (*pplib_apply_display_requirements)( - struct dc *dc, - struct dc_state *context); void (*edp_power_control)( struct dc_link *link, bool enable); @@ -227,6 +229,9 @@ struct hw_sequencer_funcs { void (*set_cursor_attribute)(struct pipe_ctx *pipe); void (*set_cursor_sdr_white_level)(struct pipe_ctx *pipe); + void (*setup_periodic_interrupt)(struct pipe_ctx *pipe_ctx, enum vline_select vline); + void (*setup_vupdate_interrupt)(struct pipe_ctx *pipe_ctx); + }; void color_space_to_black_color( diff --git a/drivers/gpu/drm/amd/display/dc/inc/resource.h b/drivers/gpu/drm/amd/display/dc/inc/resource.h index 33b99e3ab10d..0086a2f1d21a 100644 --- a/drivers/gpu/drm/amd/display/dc/inc/resource.h +++ b/drivers/gpu/drm/amd/display/dc/inc/resource.h @@ -30,9 +30,6 @@ #include "dal_asic_id.h" #include "dm_pp_smu.h" -/* TODO unhardcode, 4 for CZ*/ -#define MEMORY_TYPE_MULTIPLIER 4 - enum dce_version resource_parse_asic_id( struct hw_asic_id asic_id); diff --git a/drivers/gpu/drm/amd/display/dc/inc/vm_helper.h b/drivers/gpu/drm/amd/display/dc/inc/vm_helper.h new file mode 100644 index 000000000000..193407f76a80 --- /dev/null +++ b/drivers/gpu/drm/amd/display/dc/inc/vm_helper.h @@ -0,0 +1,56 @@ +/* + * Copyright 2018 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. + * + * Authors: AMD + * + */ + +#ifndef DC_INC_VM_HELPER_H_ +#define DC_INC_VM_HELPER_H_ + +#include "dc_types.h" + +#define MAX_VMID 16 +#define MAX_HUBP 6 + +struct vmid_usage { + uint16_t vmid_usage[2]; +}; + +struct vm_helper { + unsigned int num_vmid; + unsigned int num_hubp; + unsigned int num_vmids_available; + uint64_t ptb_assigned_to_vmid[MAX_VMID]; + struct vmid_usage hubp_vmid_usage[MAX_HUBP]; +}; + +uint8_t get_vmid_for_ptb( + struct vm_helper *vm_helper, + int64_t ptb, + uint8_t pipe_idx); + +void init_vm_helper( + struct vm_helper *vm_helper, + unsigned int num_vmid, + unsigned int num_hubp); + +#endif /* DC_INC_VM_HELPER_H_ */ diff --git a/drivers/gpu/drm/amd/display/dc/irq_types.h b/drivers/gpu/drm/amd/display/dc/irq_types.h index 0b5f3a278c22..d0ccd81ad5b4 100644 --- a/drivers/gpu/drm/amd/display/dc/irq_types.h +++ b/drivers/gpu/drm/amd/display/dc/irq_types.h @@ -144,6 +144,14 @@ enum dc_irq_source { DC_IRQ_SOURCE_DC5_VLINE0, DC_IRQ_SOURCE_DC6_VLINE0, + DC_IRQ_SOURCE_DC1_VLINE1, + DC_IRQ_SOURCE_DC2_VLINE1, + DC_IRQ_SOURCE_DC3_VLINE1, + DC_IRQ_SOURCE_DC4_VLINE1, + DC_IRQ_SOURCE_DC5_VLINE1, + DC_IRQ_SOURCE_DC6_VLINE1, + + DAL_IRQ_SOURCES_NUMBER }; diff --git a/drivers/gpu/drm/amd/display/include/bios_parser_types.h b/drivers/gpu/drm/amd/display/include/bios_parser_types.h index f8dbfa5b89f2..01bf01a34a08 100644 --- a/drivers/gpu/drm/amd/display/include/bios_parser_types.h +++ b/drivers/gpu/drm/amd/display/include/bios_parser_types.h @@ -41,6 +41,7 @@ enum as_signal_type { AS_SIGNAL_TYPE_LVDS, AS_SIGNAL_TYPE_DISPLAY_PORT, AS_SIGNAL_TYPE_GPU_PLL, + AS_SIGNAL_TYPE_XGMI, AS_SIGNAL_TYPE_UNKNOWN }; @@ -210,8 +211,8 @@ struct bp_pixel_clock_parameters { /* signal_type -> Encoder Mode - needed by VBIOS Exec table */ enum signal_type signal_type; /* Adjusted Pixel Clock (after VBIOS exec table) - * that becomes Target Pixel Clock (KHz) */ - uint32_t target_pixel_clock; + * that becomes Target Pixel Clock (100 Hz units) */ + uint32_t target_pixel_clock_100hz; /* Calculated Reference divider of Display PLL */ uint32_t reference_divider; /* Calculated Feedback divider of Display PLL */ diff --git a/drivers/gpu/drm/amd/display/include/dal_asic_id.h b/drivers/gpu/drm/amd/display/include/dal_asic_id.h index 4f501ddcfb8d..34d6fdcb32e2 100644 --- a/drivers/gpu/drm/amd/display/include/dal_asic_id.h +++ b/drivers/gpu/drm/amd/display/include/dal_asic_id.h @@ -131,6 +131,7 @@ #define INTERNAL_REV_RAVEN_A0 0x00 /* First spin of Raven */ #define RAVEN_A0 0x01 #define RAVEN_B0 0x21 +#define PICASSO_A0 0x41 #if defined(CONFIG_DRM_AMD_DC_DCN1_01) /* DCN1_01 */ #define RAVEN2_A0 0x81 @@ -165,4 +166,6 @@ #define FAMILY_UNKNOWN 0xFF + + #endif /* __DAL_ASIC_ID_H__ */ diff --git a/drivers/gpu/drm/amd/display/include/dal_types.h b/drivers/gpu/drm/amd/display/include/dal_types.h index 89627133e188..f5bd869d4320 100644 --- a/drivers/gpu/drm/amd/display/include/dal_types.h +++ b/drivers/gpu/drm/amd/display/include/dal_types.h @@ -42,6 +42,7 @@ enum dce_version { DCE_VERSION_11_2, DCE_VERSION_11_22, DCE_VERSION_12_0, + DCE_VERSION_12_1, DCE_VERSION_MAX, DCN_VERSION_1_0, #if defined(CONFIG_DRM_AMD_DC_DCN1_01) diff --git a/drivers/gpu/drm/amd/display/include/gpio_interface.h b/drivers/gpu/drm/amd/display/include/gpio_interface.h index e4fd31024b92..7de64195dc33 100644 --- a/drivers/gpu/drm/amd/display/include/gpio_interface.h +++ b/drivers/gpu/drm/amd/display/include/gpio_interface.h @@ -59,6 +59,14 @@ enum gpio_result dal_gpio_change_mode( struct gpio *gpio, enum gpio_mode mode); +/* Lock Pin */ +enum gpio_result dal_gpio_lock_pin( + struct gpio *gpio); + +/* Unlock Pin */ +enum gpio_result dal_gpio_unlock_pin( + struct gpio *gpio); + /* Get the GPIO id */ enum gpio_id dal_gpio_get_id( const struct gpio *gpio); diff --git a/drivers/gpu/drm/amd/display/include/i2caux_interface.h b/drivers/gpu/drm/amd/display/include/i2caux_interface.h index 13a3c82d118f..bb012cb1a9f5 100644 --- a/drivers/gpu/drm/amd/display/include/i2caux_interface.h +++ b/drivers/gpu/drm/amd/display/include/i2caux_interface.h @@ -40,9 +40,19 @@ struct aux_payload { /* set following flag to write data, * reset it to read data */ bool write; + bool mot; uint32_t address; uint8_t length; uint8_t *data; + /* + * used to return the reply type of the transaction + * ignored if NULL + */ + uint8_t *reply; + /* expressed in milliseconds + * zero means "use default value" + */ + uint32_t defer_delay; }; struct aux_command { @@ -66,27 +76,4 @@ union aux_config { uint32_t raw; }; -struct i2caux; - -struct i2caux *dal_i2caux_create( - struct dc_context *ctx); - -bool dal_i2caux_submit_i2c_command( - struct i2caux *i2caux, - struct ddc *ddc, - struct i2c_command *cmd); - -bool dal_i2caux_submit_aux_command( - struct i2caux *i2caux, - struct ddc *ddc, - struct aux_command *cmd); - -void dal_i2caux_configure_aux( - struct i2caux *i2caux, - struct ddc *ddc, - union aux_config cfg); - -void dal_i2caux_destroy( - struct i2caux **ptr); - #endif diff --git a/drivers/gpu/drm/amd/display/modules/color/color_gamma.c b/drivers/gpu/drm/amd/display/modules/color/color_gamma.c index cdcefd087487..0fbc8fbc3541 100644 --- a/drivers/gpu/drm/amd/display/modules/color/color_gamma.c +++ b/drivers/gpu/drm/amd/display/modules/color/color_gamma.c @@ -306,6 +306,18 @@ static struct fixed31_32 translate_from_linear_space( a1); } +static struct fixed31_32 calculate_gamma22(struct fixed31_32 arg) +{ + struct fixed31_32 gamma = dc_fixpt_from_fraction(22, 10); + + return translate_from_linear_space(arg, + dc_fixpt_zero, + dc_fixpt_zero, + dc_fixpt_zero, + dc_fixpt_zero, + gamma); +} + static struct fixed31_32 translate_to_linear_space( struct fixed31_32 arg, struct fixed31_32 a0, @@ -709,6 +721,175 @@ static void build_regamma(struct pwl_float_data_ex *rgb_regamma, } } +static void hermite_spline_eetf(struct fixed31_32 input_x, + struct fixed31_32 max_display, + struct fixed31_32 min_display, + struct fixed31_32 max_content, + struct fixed31_32 *out_x) +{ + struct fixed31_32 min_lum_pq; + struct fixed31_32 max_lum_pq; + struct fixed31_32 max_content_pq; + struct fixed31_32 ks; + struct fixed31_32 E1; + struct fixed31_32 E2; + struct fixed31_32 E3; + struct fixed31_32 t; + struct fixed31_32 t2; + struct fixed31_32 t3; + struct fixed31_32 two; + struct fixed31_32 three; + struct fixed31_32 temp1; + struct fixed31_32 temp2; + struct fixed31_32 a = dc_fixpt_from_fraction(15, 10); + struct fixed31_32 b = dc_fixpt_from_fraction(5, 10); + struct fixed31_32 epsilon = dc_fixpt_from_fraction(1, 1000000); // dc_fixpt_epsilon is a bit too small + + if (dc_fixpt_eq(max_content, dc_fixpt_zero)) { + *out_x = dc_fixpt_zero; + return; + } + + compute_pq(input_x, &E1); + compute_pq(dc_fixpt_div(min_display, max_content), &min_lum_pq); + compute_pq(dc_fixpt_div(max_display, max_content), &max_lum_pq); + compute_pq(dc_fixpt_one, &max_content_pq); // always 1? DAL2 code is weird + a = dc_fixpt_div(dc_fixpt_add(dc_fixpt_one, b), max_content_pq); // (1+b)/maxContent + ks = dc_fixpt_sub(dc_fixpt_mul(a, max_lum_pq), b); // a * max_lum_pq - b + + if (dc_fixpt_lt(E1, ks)) + E2 = E1; + else if (dc_fixpt_le(ks, E1) && dc_fixpt_le(E1, dc_fixpt_one)) { + if (dc_fixpt_lt(epsilon, dc_fixpt_sub(dc_fixpt_one, ks))) + // t = (E1 - ks) / (1 - ks) + t = dc_fixpt_div(dc_fixpt_sub(E1, ks), + dc_fixpt_sub(dc_fixpt_one, ks)); + else + t = dc_fixpt_zero; + + two = dc_fixpt_from_int(2); + three = dc_fixpt_from_int(3); + + t2 = dc_fixpt_mul(t, t); + t3 = dc_fixpt_mul(t2, t); + temp1 = dc_fixpt_mul(two, t3); + temp2 = dc_fixpt_mul(three, t2); + + // (2t^3 - 3t^2 + 1) * ks + E2 = dc_fixpt_mul(ks, dc_fixpt_add(dc_fixpt_one, + dc_fixpt_sub(temp1, temp2))); + + // (-2t^3 + 3t^2) * max_lum_pq + E2 = dc_fixpt_add(E2, dc_fixpt_mul(max_lum_pq, + dc_fixpt_sub(temp2, temp1))); + + temp1 = dc_fixpt_mul(two, t2); + temp2 = dc_fixpt_sub(dc_fixpt_one, ks); + + // (t^3 - 2t^2 + t) * (1-ks) + E2 = dc_fixpt_add(E2, dc_fixpt_mul(temp2, + dc_fixpt_add(t, dc_fixpt_sub(t3, temp1)))); + } else + E2 = dc_fixpt_one; + + temp1 = dc_fixpt_sub(dc_fixpt_one, E2); + temp2 = dc_fixpt_mul(temp1, temp1); + temp2 = dc_fixpt_mul(temp2, temp2); + // temp2 = (1-E2)^4 + + E3 = dc_fixpt_add(E2, dc_fixpt_mul(min_lum_pq, temp2)); + compute_de_pq(E3, out_x); + + *out_x = dc_fixpt_div(*out_x, dc_fixpt_div(max_display, max_content)); +} + +static bool build_freesync_hdr(struct pwl_float_data_ex *rgb_regamma, + uint32_t hw_points_num, + const struct hw_x_point *coordinate_x, + const struct freesync_hdr_tf_params *fs_params) +{ + uint32_t i; + struct pwl_float_data_ex *rgb = rgb_regamma; + const struct hw_x_point *coord_x = coordinate_x; + struct fixed31_32 scaledX = dc_fixpt_zero; + struct fixed31_32 scaledX1 = dc_fixpt_zero; + struct fixed31_32 max_display; + struct fixed31_32 min_display; + struct fixed31_32 max_content; + struct fixed31_32 min_content; + struct fixed31_32 clip = dc_fixpt_one; + struct fixed31_32 output; + bool use_eetf = false; + bool is_clipped = false; + struct fixed31_32 sdr_white_level; + + if (fs_params->max_content == 0 || + fs_params->max_display == 0) + return false; + + max_display = dc_fixpt_from_int(fs_params->max_display); + min_display = dc_fixpt_from_fraction(fs_params->min_display, 10000); + max_content = dc_fixpt_from_int(fs_params->max_content); + min_content = dc_fixpt_from_fraction(fs_params->min_content, 10000); + sdr_white_level = dc_fixpt_from_int(fs_params->sdr_white_level); + + if (fs_params->min_display > 1000) // cap at 0.1 at the bottom + min_display = dc_fixpt_from_fraction(1, 10); + if (fs_params->max_display < 100) // cap at 100 at the top + max_display = dc_fixpt_from_int(100); + + if (fs_params->min_content < fs_params->min_display) + use_eetf = true; + else + min_content = min_display; + + if (fs_params->max_content > fs_params->max_display) + use_eetf = true; + else + max_content = max_display; + + rgb += 32; // first 32 points have problems with fixed point, too small + coord_x += 32; + for (i = 32; i <= hw_points_num; i++) { + if (!is_clipped) { + if (use_eetf) { + /*max content is equal 1 */ + scaledX1 = dc_fixpt_div(coord_x->x, + dc_fixpt_div(max_content, sdr_white_level)); + hermite_spline_eetf(scaledX1, max_display, min_display, + max_content, &scaledX); + } else + scaledX = dc_fixpt_div(coord_x->x, + dc_fixpt_div(max_display, sdr_white_level)); + + if (dc_fixpt_lt(scaledX, clip)) { + if (dc_fixpt_lt(scaledX, dc_fixpt_zero)) + output = dc_fixpt_zero; + else + output = calculate_gamma22(scaledX); + + rgb->r = output; + rgb->g = output; + rgb->b = output; + } else { + is_clipped = true; + rgb->r = clip; + rgb->g = clip; + rgb->b = clip; + } + } else { + rgb->r = clip; + rgb->g = clip; + rgb->b = clip; + } + + ++coord_x; + ++rgb; + } + + return true; +} + static void build_degamma(struct pwl_float_data_ex *curve, uint32_t hw_points_num, const struct hw_x_point *coordinate_x, bool is_2_4) @@ -1327,7 +1508,7 @@ static bool map_regamma_hw_to_x_user( struct hw_x_point *coords = coords_x; const struct pwl_float_data_ex *regamma = rgb_regamma; - if (mapUserRamp) { + if (ramp && mapUserRamp) { copy_rgb_regamma_to_coordinates_x(coords, hw_points_num, rgb_regamma); @@ -1356,14 +1537,15 @@ static bool map_regamma_hw_to_x_user( #define _EXTRA_POINTS 3 bool mod_color_calculate_regamma_params(struct dc_transfer_func *output_tf, - const struct dc_gamma *ramp, bool mapUserRamp, bool canRomBeUsed) + const struct dc_gamma *ramp, bool mapUserRamp, bool canRomBeUsed, + const struct freesync_hdr_tf_params *fs_params) { struct dc_transfer_func_distributed_points *tf_pts = &output_tf->tf_pts; struct dividers dividers; struct pwl_float_data *rgb_user = NULL; struct pwl_float_data_ex *rgb_regamma = NULL; - struct gamma_pixel *axix_x = NULL; + struct gamma_pixel *axis_x = NULL; struct pixel_gamma_point *coeff = NULL; enum dc_transfer_func_predefined tf = TRANSFER_FUNCTION_SRGB; bool ret = false; @@ -1373,47 +1555,54 @@ bool mod_color_calculate_regamma_params(struct dc_transfer_func *output_tf, /* we can use hardcoded curve for plain SRGB TF */ if (output_tf->type == TF_TYPE_PREDEFINED && canRomBeUsed == true && - output_tf->tf == TRANSFER_FUNCTION_SRGB && - (!mapUserRamp && ramp->type == GAMMA_RGB_256)) - return true; + output_tf->tf == TRANSFER_FUNCTION_SRGB) { + if (ramp == NULL) + return true; + if (ramp->is_identity || (!mapUserRamp && ramp->type == GAMMA_RGB_256)) + return true; + } output_tf->type = TF_TYPE_DISTRIBUTED_POINTS; - rgb_user = kvcalloc(ramp->num_entries + _EXTRA_POINTS, + if (ramp && (mapUserRamp || ramp->type != GAMMA_RGB_256)) { + rgb_user = kvcalloc(ramp->num_entries + _EXTRA_POINTS, sizeof(*rgb_user), GFP_KERNEL); - if (!rgb_user) - goto rgb_user_alloc_fail; + if (!rgb_user) + goto rgb_user_alloc_fail; + + axis_x = kvcalloc(ramp->num_entries + 3, sizeof(*axis_x), + GFP_KERNEL); + if (!axis_x) + goto axis_x_alloc_fail; + + dividers.divider1 = dc_fixpt_from_fraction(3, 2); + dividers.divider2 = dc_fixpt_from_int(2); + dividers.divider3 = dc_fixpt_from_fraction(5, 2); + + build_evenly_distributed_points( + axis_x, + ramp->num_entries, + dividers); + + if (ramp->type == GAMMA_RGB_256 && mapUserRamp) + scale_gamma(rgb_user, ramp, dividers); + else if (ramp->type == GAMMA_RGB_FLOAT_1024) + scale_gamma_dx(rgb_user, ramp, dividers); + } + rgb_regamma = kvcalloc(MAX_HW_POINTS + _EXTRA_POINTS, sizeof(*rgb_regamma), GFP_KERNEL); if (!rgb_regamma) goto rgb_regamma_alloc_fail; - axix_x = kvcalloc(ramp->num_entries + 3, sizeof(*axix_x), - GFP_KERNEL); - if (!axix_x) - goto axix_x_alloc_fail; + coeff = kvcalloc(MAX_HW_POINTS + _EXTRA_POINTS, sizeof(*coeff), GFP_KERNEL); if (!coeff) goto coeff_alloc_fail; - dividers.divider1 = dc_fixpt_from_fraction(3, 2); - dividers.divider2 = dc_fixpt_from_int(2); - dividers.divider3 = dc_fixpt_from_fraction(5, 2); - tf = output_tf->tf; - - build_evenly_distributed_points( - axix_x, - ramp->num_entries, - dividers); - - if (ramp->type == GAMMA_RGB_256 && mapUserRamp) - scale_gamma(rgb_user, ramp, dividers); - else if (ramp->type == GAMMA_RGB_FLOAT_1024) - scale_gamma_dx(rgb_user, ramp, dividers); - if (tf == TRANSFER_FUNCTION_PQ) { tf_pts->end_exponent = 7; tf_pts->x_point_at_y1_red = 125; @@ -1424,6 +1613,12 @@ bool mod_color_calculate_regamma_params(struct dc_transfer_func *output_tf, MAX_HW_POINTS, coordinates_x, output_tf->sdr_ref_white_level); + } else if (tf == TRANSFER_FUNCTION_GAMMA22 && + fs_params != NULL) { + build_freesync_hdr(rgb_regamma, + MAX_HW_POINTS, + coordinates_x, + fs_params); } else { tf_pts->end_exponent = 0; tf_pts->x_point_at_y1_red = 1; @@ -1435,22 +1630,22 @@ bool mod_color_calculate_regamma_params(struct dc_transfer_func *output_tf, coordinates_x, tf == TRANSFER_FUNCTION_SRGB ? true:false); } map_regamma_hw_to_x_user(ramp, coeff, rgb_user, - coordinates_x, axix_x, rgb_regamma, + coordinates_x, axis_x, rgb_regamma, MAX_HW_POINTS, tf_pts, - (mapUserRamp || ramp->type != GAMMA_RGB_256) && - ramp->type != GAMMA_CS_TFM_1D); + (mapUserRamp || (ramp && ramp->type != GAMMA_RGB_256)) && + (ramp && ramp->type != GAMMA_CS_TFM_1D)); - if (ramp->type == GAMMA_CS_TFM_1D) + if (ramp && ramp->type == GAMMA_CS_TFM_1D) apply_lut_1d(ramp, MAX_HW_POINTS, tf_pts); ret = true; kvfree(coeff); coeff_alloc_fail: - kvfree(axix_x); -axix_x_alloc_fail: kvfree(rgb_regamma); rgb_regamma_alloc_fail: + kvfree(axis_x); +axis_x_alloc_fail: kvfree(rgb_user); rgb_user_alloc_fail: return ret; @@ -1570,69 +1765,85 @@ bool mod_color_calculate_degamma_params(struct dc_transfer_func *input_tf, { struct dc_transfer_func_distributed_points *tf_pts = &input_tf->tf_pts; struct dividers dividers; - struct pwl_float_data *rgb_user = NULL; struct pwl_float_data_ex *curve = NULL; - struct gamma_pixel *axix_x = NULL; + struct gamma_pixel *axis_x = NULL; struct pixel_gamma_point *coeff = NULL; enum dc_transfer_func_predefined tf = TRANSFER_FUNCTION_SRGB; + uint32_t i; bool ret = false; if (input_tf->type == TF_TYPE_BYPASS) return false; - /* we can use hardcoded curve for plain SRGB TF */ + /* we can use hardcoded curve for plain SRGB TF + * If linear, it's bypass if on user ramp + */ if (input_tf->type == TF_TYPE_PREDEFINED && - input_tf->tf == TRANSFER_FUNCTION_SRGB && - (!mapUserRamp && - (ramp->type == GAMMA_RGB_256 || ramp->num_entries == 0))) + (input_tf->tf == TRANSFER_FUNCTION_SRGB || + input_tf->tf == TRANSFER_FUNCTION_LINEAR) && + !mapUserRamp) return true; input_tf->type = TF_TYPE_DISTRIBUTED_POINTS; - rgb_user = kvcalloc(ramp->num_entries + _EXTRA_POINTS, - sizeof(*rgb_user), - GFP_KERNEL); - if (!rgb_user) - goto rgb_user_alloc_fail; + if (mapUserRamp && ramp && ramp->type == GAMMA_RGB_256) { + rgb_user = kvcalloc(ramp->num_entries + _EXTRA_POINTS, + sizeof(*rgb_user), + GFP_KERNEL); + if (!rgb_user) + goto rgb_user_alloc_fail; + + axis_x = kvcalloc(ramp->num_entries + _EXTRA_POINTS, sizeof(*axis_x), + GFP_KERNEL); + if (!axis_x) + goto axis_x_alloc_fail; + + dividers.divider1 = dc_fixpt_from_fraction(3, 2); + dividers.divider2 = dc_fixpt_from_int(2); + dividers.divider3 = dc_fixpt_from_fraction(5, 2); + + build_evenly_distributed_points( + axis_x, + ramp->num_entries, + dividers); + + scale_gamma(rgb_user, ramp, dividers); + } + curve = kvcalloc(MAX_HW_POINTS + _EXTRA_POINTS, sizeof(*curve), - GFP_KERNEL); + GFP_KERNEL); if (!curve) goto curve_alloc_fail; - axix_x = kvcalloc(ramp->num_entries + _EXTRA_POINTS, sizeof(*axix_x), - GFP_KERNEL); - if (!axix_x) - goto axix_x_alloc_fail; + coeff = kvcalloc(MAX_HW_POINTS + _EXTRA_POINTS, sizeof(*coeff), - GFP_KERNEL); + GFP_KERNEL); if (!coeff) goto coeff_alloc_fail; - dividers.divider1 = dc_fixpt_from_fraction(3, 2); - dividers.divider2 = dc_fixpt_from_int(2); - dividers.divider3 = dc_fixpt_from_fraction(5, 2); - tf = input_tf->tf; - build_evenly_distributed_points( - axix_x, - ramp->num_entries, - dividers); - - if (ramp->type == GAMMA_RGB_256 && mapUserRamp) - scale_gamma(rgb_user, ramp, dividers); - else if (ramp->type == GAMMA_RGB_FLOAT_1024) - scale_gamma_dx(rgb_user, ramp, dividers); - if (tf == TRANSFER_FUNCTION_PQ) build_de_pq(curve, MAX_HW_POINTS, coordinates_x); - else + else if (tf == TRANSFER_FUNCTION_SRGB || + tf == TRANSFER_FUNCTION_BT709) build_degamma(curve, MAX_HW_POINTS, coordinates_x, - tf == TRANSFER_FUNCTION_SRGB ? true:false); + tf == TRANSFER_FUNCTION_SRGB ? true : false); + else if (tf == TRANSFER_FUNCTION_LINEAR) { + // just copy coordinates_x into curve + i = 0; + while (i != MAX_HW_POINTS + 1) { + curve[i].r = coordinates_x[i].x; + curve[i].g = curve[i].r; + curve[i].b = curve[i].r; + i++; + } + } else + goto invalid_tf_fail; tf_pts->end_exponent = 0; tf_pts->x_point_at_y1_red = 1; @@ -1640,25 +1851,23 @@ bool mod_color_calculate_degamma_params(struct dc_transfer_func *input_tf, tf_pts->x_point_at_y1_blue = 1; map_regamma_hw_to_x_user(ramp, coeff, rgb_user, - coordinates_x, axix_x, curve, + coordinates_x, axis_x, curve, MAX_HW_POINTS, tf_pts, - mapUserRamp && ramp->type != GAMMA_CUSTOM); - if (ramp->type == GAMMA_CUSTOM) - apply_lut_1d(ramp, MAX_HW_POINTS, tf_pts); + mapUserRamp && ramp && ramp->type == GAMMA_RGB_256); ret = true; +invalid_tf_fail: kvfree(coeff); coeff_alloc_fail: - kvfree(axix_x); -axix_x_alloc_fail: kvfree(curve); curve_alloc_fail: + kvfree(axis_x); +axis_x_alloc_fail: kvfree(rgb_user); rgb_user_alloc_fail: return ret; - } diff --git a/drivers/gpu/drm/amd/display/modules/color/color_gamma.h b/drivers/gpu/drm/amd/display/modules/color/color_gamma.h index 63ccb9c91224..a6e164df090a 100644 --- a/drivers/gpu/drm/amd/display/modules/color/color_gamma.h +++ b/drivers/gpu/drm/amd/display/modules/color/color_gamma.h @@ -73,12 +73,21 @@ struct regamma_lut { }; }; +struct freesync_hdr_tf_params { + unsigned int sdr_white_level; + unsigned int min_content; // luminance in 1/10000 nits + unsigned int max_content; // luminance in nits + unsigned int min_display; // luminance in 1/10000 nits + unsigned int max_display; // luminance in nits +}; + void setup_x_points_distribution(void); void precompute_pq(void); void precompute_de_pq(void); bool mod_color_calculate_regamma_params(struct dc_transfer_func *output_tf, - const struct dc_gamma *ramp, bool mapUserRamp, bool canRomBeUsed); + const struct dc_gamma *ramp, bool mapUserRamp, bool canRomBeUsed, + const struct freesync_hdr_tf_params *fs_params); bool mod_color_calculate_degamma_params(struct dc_transfer_func *output_tf, const struct dc_gamma *ramp, bool mapUserRamp); diff --git a/drivers/gpu/drm/amd/display/modules/freesync/freesync.c b/drivers/gpu/drm/amd/display/modules/freesync/freesync.c index 4018c7180d00..bfd27f10879e 100644 --- a/drivers/gpu/drm/amd/display/modules/freesync/freesync.c +++ b/drivers/gpu/drm/amd/display/modules/freesync/freesync.c @@ -37,6 +37,8 @@ #define RENDER_TIMES_MAX_COUNT 10 /* Threshold to exit BTR (to avoid frequent enter-exits at the lower limit) */ #define BTR_EXIT_MARGIN 2000 +/*Threshold to exit fixed refresh rate*/ +#define FIXED_REFRESH_EXIT_MARGIN_IN_HZ 4 /* Number of consecutive frames to check before entering/exiting fixed refresh*/ #define FIXED_REFRESH_ENTER_FRAME_COUNT 5 #define FIXED_REFRESH_EXIT_FRAME_COUNT 5 @@ -106,8 +108,8 @@ static unsigned int calc_duration_in_us_from_v_total( { unsigned int duration_in_us = (unsigned int)(div64_u64(((unsigned long long)(v_total) - * 1000) * stream->timing.h_total, - stream->timing.pix_clk_khz)); + * 10000) * stream->timing.h_total, + stream->timing.pix_clk_100hz)); return duration_in_us; } @@ -124,7 +126,7 @@ static unsigned int calc_v_total_from_refresh( refresh_in_uhz))); v_total = div64_u64(div64_u64(((unsigned long long)( - frame_duration_in_ns) * stream->timing.pix_clk_khz), + frame_duration_in_ns) * (stream->timing.pix_clk_100hz / 10)), stream->timing.h_total), 1000000); /* v_total cannot be less than nominal */ @@ -150,7 +152,7 @@ static unsigned int calc_v_total_from_duration( duration_in_us = vrr->max_duration_in_us; v_total = div64_u64(div64_u64(((unsigned long long)( - duration_in_us) * stream->timing.pix_clk_khz), + duration_in_us) * (stream->timing.pix_clk_100hz / 10)), stream->timing.h_total), 1000); /* v_total cannot be less than nominal */ @@ -225,7 +227,7 @@ static void update_v_total_for_static_ramp( } v_total = div64_u64(div64_u64(((unsigned long long)( - current_duration_in_us) * stream->timing.pix_clk_khz), + current_duration_in_us) * (stream->timing.pix_clk_100hz / 10)), stream->timing.h_total), 1000); in_out_vrr->adjust.v_total_min = v_total; @@ -257,40 +259,14 @@ static void apply_below_the_range(struct core_freesync *core_freesync, if (in_out_vrr->btr.btr_active) { in_out_vrr->btr.frame_counter = 0; in_out_vrr->btr.btr_active = false; - - /* Exit Fixed Refresh mode */ - } else if (in_out_vrr->fixed.fixed_active) { - - in_out_vrr->fixed.frame_counter++; - - if (in_out_vrr->fixed.frame_counter > - FIXED_REFRESH_EXIT_FRAME_COUNT) { - in_out_vrr->fixed.frame_counter = 0; - in_out_vrr->fixed.fixed_active = false; - } } } else if (last_render_time_in_us > max_render_time_in_us) { /* Enter Below the Range */ - if (!in_out_vrr->btr.btr_active && - in_out_vrr->btr.btr_enabled) { - in_out_vrr->btr.btr_active = true; - - /* Enter Fixed Refresh mode */ - } else if (!in_out_vrr->fixed.fixed_active && - !in_out_vrr->btr.btr_enabled) { - in_out_vrr->fixed.frame_counter++; - - if (in_out_vrr->fixed.frame_counter > - FIXED_REFRESH_ENTER_FRAME_COUNT) { - in_out_vrr->fixed.frame_counter = 0; - in_out_vrr->fixed.fixed_active = true; - } - } + in_out_vrr->btr.btr_active = true; } /* BTR set to "not active" so disengage */ if (!in_out_vrr->btr.btr_active) { - in_out_vrr->btr.btr_active = false; in_out_vrr->btr.inserted_duration_in_us = 0; in_out_vrr->btr.frames_to_insert = 0; in_out_vrr->btr.frame_counter = 0; @@ -375,7 +351,12 @@ static void apply_fixed_refresh(struct core_freesync *core_freesync, bool update = false; unsigned int max_render_time_in_us = in_out_vrr->max_duration_in_us; - if (last_render_time_in_us + BTR_EXIT_MARGIN < max_render_time_in_us) { + //Compute the exit refresh rate and exit frame duration + unsigned int exit_refresh_rate_in_milli_hz = ((1000000000/max_render_time_in_us) + + (1000*FIXED_REFRESH_EXIT_MARGIN_IN_HZ)); + unsigned int exit_frame_duration_in_us = 1000000000/exit_refresh_rate_in_milli_hz; + + if (last_render_time_in_us < exit_frame_duration_in_us) { /* Exit Fixed Refresh mode */ if (in_out_vrr->fixed.fixed_active) { in_out_vrr->fixed.frame_counter++; @@ -480,6 +461,26 @@ bool mod_freesync_get_v_position(struct mod_freesync *mod_freesync, return false; } +static void build_vrr_infopacket_header_vtem(enum signal_type signal, + struct dc_info_packet *infopacket) +{ + // HEADER + + // HB0, HB1, HB2 indicates PacketType VTEMPacket + infopacket->hb0 = 0x7F; + infopacket->hb1 = 0xC0; + infopacket->hb2 = 0x00; + /* HB3 Bit Fields + * Reserved :1 = 0 + * Sync :1 = 0 + * VFR :1 = 1 + * Ds_Type :2 = 0 + * End :1 = 0 + * New :1 = 0 + */ + infopacket->hb3 = 0x20; +} + static void build_vrr_infopacket_header_v1(enum signal_type signal, struct dc_info_packet *infopacket, unsigned int *payload_size) @@ -578,6 +579,54 @@ static void build_vrr_infopacket_header_v2(enum signal_type signal, } } +static void build_vrr_vtem_infopacket_data(const struct dc_stream_state *stream, + const struct mod_vrr_params *vrr, + struct dc_info_packet *infopacket) +{ + /* dc_info_packet to VtemPacket Translation of Bit-fields, + * SB[6] + * unsigned char VRR_EN :1 + * unsigned char M_CONST :1 + * unsigned char Reserved2 :2 + * unsigned char FVA_Factor_M1 :4 + * SB[7] + * unsigned char Base_Vfront :8 + * SB[8] + * unsigned char Base_Refresh_Rate_98 :2 + * unsigned char RB :1 + * unsigned char Reserved3 :5 + * SB[9] + * unsigned char Base_RefreshRate_07 :8 + */ + unsigned int fieldRateInHz; + + if (vrr->state == VRR_STATE_ACTIVE_VARIABLE || + vrr->state == VRR_STATE_ACTIVE_FIXED){ + infopacket->sb[6] |= 0x80; //VRR_EN Bit = 1 + } else { + infopacket->sb[6] &= 0x7F; //VRR_EN Bit = 0 + } + + if (!stream->timing.vic) { + infopacket->sb[7] = stream->timing.v_front_porch; + + /* TODO: In dal2, we check mode flags for a reduced blanking timing. + * Need a way to relay that information to this function. + * if("ReducedBlanking") + * { + * infopacket->sb[8] |= 0x20; //Set 3rd bit to 1 + * } + */ + fieldRateInHz = (stream->timing.pix_clk_100hz * 100)/ + (stream->timing.h_total * stream->timing.v_total); + + infopacket->sb[8] |= ((fieldRateInHz & 0x300) >> 2); + infopacket->sb[9] |= fieldRateInHz & 0xFF; + + } + infopacket->valid = true; +} + static void build_vrr_infopacket_data(const struct mod_vrr_params *vrr, struct dc_info_packet *infopacket) { @@ -627,12 +676,12 @@ static void build_vrr_infopacket_data(const struct mod_vrr_params *vrr, static void build_vrr_infopacket_fs2_data(enum color_transfer_func app_tf, struct dc_info_packet *infopacket) { - if (app_tf != transfer_func_unknown) { + if (app_tf != TRANSFER_FUNC_UNKNOWN) { infopacket->valid = true; infopacket->sb[6] |= 0x08; // PB6 = [Bit 3 = Native Color Active] - if (app_tf == transfer_func_gamma_22) { + if (app_tf == TRANSFER_FUNC_GAMMA_22) { infopacket->sb[9] |= 0x04; // PB6 = [Bit 2 = Gamma 2.2 EOTF Active] } } @@ -675,7 +724,7 @@ static void build_vrr_infopacket_v1(enum signal_type signal, static void build_vrr_infopacket_v2(enum signal_type signal, const struct mod_vrr_params *vrr, - const enum color_transfer_func *app_tf, + enum color_transfer_func app_tf, struct dc_info_packet *infopacket) { unsigned int payload_size = 0; @@ -683,35 +732,50 @@ static void build_vrr_infopacket_v2(enum signal_type signal, build_vrr_infopacket_header_v2(signal, infopacket, &payload_size); build_vrr_infopacket_data(vrr, infopacket); - if (app_tf != NULL) - build_vrr_infopacket_fs2_data(*app_tf, infopacket); + build_vrr_infopacket_fs2_data(app_tf, infopacket); build_vrr_infopacket_checksum(&payload_size, infopacket); infopacket->valid = true; } +static void build_vrr_infopacket_vtem(const struct dc_stream_state *stream, + const struct mod_vrr_params *vrr, + struct dc_info_packet *infopacket) +{ + //VTEM info packet for HdmiVrr + + //VTEM Packet is structured differently + build_vrr_infopacket_header_vtem(stream->signal, infopacket); + build_vrr_vtem_infopacket_data(stream, vrr, infopacket); + + infopacket->valid = true; +} + void mod_freesync_build_vrr_infopacket(struct mod_freesync *mod_freesync, const struct dc_stream_state *stream, const struct mod_vrr_params *vrr, enum vrr_packet_type packet_type, - const enum color_transfer_func *app_tf, + enum color_transfer_func app_tf, struct dc_info_packet *infopacket) { - /* SPD info packet for FreeSync */ - - /* Check if Freesync is supported. Return if false. If true, + /* SPD info packet for FreeSync + * VTEM info packet for HdmiVRR + * Check if Freesync is supported. Return if false. If true, * set the corresponding bit in the info packet */ - if (!vrr->supported || !vrr->send_vsif) + if (!vrr->supported || (!vrr->send_info_frame && packet_type != PACKET_TYPE_VTEM)) return; switch (packet_type) { - case packet_type_fs2: + case PACKET_TYPE_FS2: build_vrr_infopacket_v2(stream->signal, vrr, app_tf, infopacket); break; - case packet_type_vrr: - case packet_type_fs1: + case PACKET_TYPE_VTEM: + build_vrr_infopacket_vtem(stream, vrr, infopacket); + break; + case PACKET_TYPE_VRR: + case PACKET_TYPE_FS1: default: build_vrr_infopacket_v1(stream->signal, vrr, infopacket); } @@ -758,7 +822,7 @@ void mod_freesync_build_vrr_params(struct mod_freesync *mod_freesync, return; in_out_vrr->state = in_config->state; - in_out_vrr->send_vsif = in_config->vsif_supported; + in_out_vrr->send_info_frame = in_config->vsif_supported; if (in_config->state == VRR_STATE_UNSUPPORTED) { in_out_vrr->state = VRR_STATE_UNSUPPORTED; @@ -991,7 +1055,7 @@ unsigned long long mod_freesync_calc_nominal_field_rate( unsigned long long nominal_field_rate_in_uhz = 0; /* Calculate nominal field rate for stream */ - nominal_field_rate_in_uhz = stream->timing.pix_clk_khz; + nominal_field_rate_in_uhz = stream->timing.pix_clk_100hz / 10; nominal_field_rate_in_uhz *= 1000ULL * 1000ULL * 1000ULL; nominal_field_rate_in_uhz = div_u64(nominal_field_rate_in_uhz, stream->timing.h_total); diff --git a/drivers/gpu/drm/amd/display/modules/inc/mod_freesync.h b/drivers/gpu/drm/amd/display/modules/inc/mod_freesync.h index 949a8b62aa98..dcef85994c45 100644 --- a/drivers/gpu/drm/amd/display/modules/inc/mod_freesync.h +++ b/drivers/gpu/drm/amd/display/modules/inc/mod_freesync.h @@ -104,7 +104,7 @@ struct mod_vrr_params_fixed_refresh { struct mod_vrr_params { bool supported; - bool send_vsif; + bool send_info_frame; enum mod_vrr_state state; uint32_t min_refresh_in_uhz; @@ -145,7 +145,7 @@ void mod_freesync_build_vrr_infopacket(struct mod_freesync *mod_freesync, const struct dc_stream_state *stream, const struct mod_vrr_params *vrr, enum vrr_packet_type packet_type, - const enum color_transfer_func *app_tf, + enum color_transfer_func app_tf, struct dc_info_packet *infopacket); void mod_freesync_build_vrr_params(struct mod_freesync *mod_freesync, diff --git a/drivers/gpu/drm/amd/display/modules/inc/mod_info_packet.h b/drivers/gpu/drm/amd/display/modules/inc/mod_info_packet.h index 786b34380f85..5b1c9a4c7643 100644 --- a/drivers/gpu/drm/amd/display/modules/inc/mod_info_packet.h +++ b/drivers/gpu/drm/amd/display/modules/inc/mod_info_packet.h @@ -26,15 +26,13 @@ #ifndef MOD_INFO_PACKET_H_ #define MOD_INFO_PACKET_H_ -struct info_packet_inputs { - const struct dc_stream_state *pStream; -}; +#include "mod_shared.h" -struct info_packets { - struct dc_info_packet *pVscInfoPacket; -}; +//Forward Declarations +struct dc_stream_state; +struct dc_info_packet; -void mod_build_infopackets(struct info_packet_inputs *inputs, - struct info_packets *info_packets); +void mod_build_vsc_infopacket(const struct dc_stream_state *stream, + struct dc_info_packet *info_packet); #endif diff --git a/drivers/gpu/drm/amd/display/modules/inc/mod_shared.h b/drivers/gpu/drm/amd/display/modules/inc/mod_shared.h index 238c431ae483..b711e7e6c204 100644 --- a/drivers/gpu/drm/amd/display/modules/inc/mod_shared.h +++ b/drivers/gpu/drm/amd/display/modules/inc/mod_shared.h @@ -23,27 +23,27 @@ * */ - #ifndef MOD_SHARED_H_ #define MOD_SHARED_H_ enum color_transfer_func { - transfer_func_unknown, - transfer_func_srgb, - transfer_func_bt709, - transfer_func_pq2084, - transfer_func_pq2084_interim, - transfer_func_linear_0_1, - transfer_func_linear_0_125, - transfer_func_dolbyvision, - transfer_func_gamma_22, - transfer_func_gamma_26 + TRANSFER_FUNC_UNKNOWN, + TRANSFER_FUNC_SRGB, + TRANSFER_FUNC_BT709, + TRANSFER_FUNC_PQ2084, + TRANSFER_FUNC_PQ2084_INTERIM, + TRANSFER_FUNC_LINEAR_0_1, + TRANSFER_FUNC_LINEAR_0_125, + TRANSFER_FUNC_GAMMA_22, + TRANSFER_FUNC_GAMMA_26 }; enum vrr_packet_type { - packet_type_vrr, - packet_type_fs1, - packet_type_fs2 + PACKET_TYPE_VRR, + PACKET_TYPE_FS1, + PACKET_TYPE_FS2, + PACKET_TYPE_VTEM }; + #endif /* MOD_SHARED_H_ */ diff --git a/drivers/gpu/drm/amd/display/modules/info_packet/info_packet.c b/drivers/gpu/drm/amd/display/modules/info_packet/info_packet.c index ff8bfb9b43b0..db06fab2ad5c 100644 --- a/drivers/gpu/drm/amd/display/modules/info_packet/info_packet.c +++ b/drivers/gpu/drm/amd/display/modules/info_packet/info_packet.c @@ -25,6 +25,10 @@ #include "mod_info_packet.h" #include "core_types.h" +#include "dc_types.h" +#include "mod_shared.h" + +#define HDMI_INFOFRAME_TYPE_VENDOR 0x81 enum ColorimetryRGBDP { ColorimetryRGB_DP_sRGB = 0, @@ -41,7 +45,7 @@ enum ColorimetryYCCDP { ColorimetryYCC_DP_ITU2020YCbCr = 7, }; -static void mod_build_vsc_infopacket(const struct dc_stream_state *stream, +void mod_build_vsc_infopacket(const struct dc_stream_state *stream, struct dc_info_packet *info_packet) { unsigned int vscPacketRevision = 0; @@ -159,7 +163,7 @@ static void mod_build_vsc_infopacket(const struct dc_stream_state *stream, * DPCD register is exposed in the new Extended Receiver Capability field for DPCD Rev. 1.4 * (and higher). When MISC1. bit 6. is Set to 1, a Source device uses a VSC SDP to indicate * the Pixel Encoding/Colorimetry Format and that a Sink device must ignore MISC1, bit 7, and - * MISC0, bits 7:1 (MISC1, bit 7. and MISC0, bits 7:1 become “don’t care”).) + * MISC0, bits 7:1 (MISC1, bit 7. and MISC0, bits 7:1 become "don't care").) */ if (vscPacketRevision == 0x5) { /* Secondary-data Packet ID = 0 */ @@ -320,10 +324,3 @@ static void mod_build_vsc_infopacket(const struct dc_stream_state *stream, } -void mod_build_infopackets(struct info_packet_inputs *inputs, - struct info_packets *info_packets) -{ - if (info_packets->pVscInfoPacket != NULL) - mod_build_vsc_infopacket(inputs->pStream, info_packets->pVscInfoPacket); -} - diff --git a/drivers/gpu/drm/amd/display/modules/power/Makefile b/drivers/gpu/drm/amd/display/modules/power/Makefile new file mode 100644 index 000000000000..87851f892a52 --- /dev/null +++ b/drivers/gpu/drm/amd/display/modules/power/Makefile @@ -0,0 +1,31 @@ +# +# Copyright 2017 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. +# +# +# Makefile for the 'power' sub-module of DAL. +# + +MOD_POWER = power_helpers.o + +AMD_DAL_MOD_POWER = $(addprefix $(AMDDALPATH)/modules/power/,$(MOD_POWER)) +#$(info ************ DAL POWER MODULE MAKEFILE ************) + +AMD_DISPLAY_FILES += $(AMD_DAL_MOD_POWER)
\ No newline at end of file diff --git a/drivers/gpu/drm/amd/display/modules/power/power_helpers.c b/drivers/gpu/drm/amd/display/modules/power/power_helpers.c new file mode 100644 index 000000000000..038b88221c5f --- /dev/null +++ b/drivers/gpu/drm/amd/display/modules/power/power_helpers.c @@ -0,0 +1,580 @@ +/* Copyright 2018 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. + * + * Authors: AMD + * + */ + +#include "power_helpers.h" +#include "dc/inc/hw/dmcu.h" + +#define DIV_ROUNDUP(a, b) (((a)+((b)/2))/(b)) + +/* Possible Min Reduction config from least aggressive to most aggressive + * 0 1 2 3 4 5 6 7 8 9 10 11 12 + * 100 98.0 94.1 94.1 85.1 80.3 75.3 69.4 60.0 57.6 50.2 49.8 40.0 % + */ +static const unsigned char min_reduction_table[13] = { +0xff, 0xfa, 0xf0, 0xf0, 0xd9, 0xcd, 0xc0, 0xb1, 0x99, 0x93, 0x80, 0x82, 0x66}; + +/* Possible Max Reduction configs from least aggressive to most aggressive + * 0 1 2 3 4 5 6 7 8 9 10 11 12 + * 96.1 89.8 85.1 80.3 69.4 64.7 64.7 50.2 39.6 30.2 30.2 30.2 19.6 % + */ +static const unsigned char max_reduction_table[13] = { +0xf5, 0xe5, 0xd9, 0xcd, 0xb1, 0xa5, 0xa5, 0x80, 0x65, 0x4d, 0x4d, 0x4d, 0x32}; + +/* ABM 2.2 Min Reduction effectively disabled (100% for all configs)*/ +static const unsigned char min_reduction_table_v_2_2[13] = { +0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff}; + +/* Possible ABM 2.2 Max Reduction configs from least aggressive to most aggressive + * 0 1 2 3 4 5 6 7 8 9 10 11 12 + * 96.1 89.8 74.9 69.4 64.7 52.2 48.6 39.6 30.2 25.1 19.6 12.5 12.5 % + */ +static const unsigned char max_reduction_table_v_2_2[13] = { +0xf5, 0xe5, 0xbf, 0xb1, 0xa5, 0x85, 0x7c, 0x65, 0x4d, 0x40, 0x32, 0x20, 0x20}; + +/* Predefined ABM configuration sets. We may have different configuration sets + * in order to satisfy different power/quality requirements. + */ +static const unsigned char abm_config[abm_defines_max_config][abm_defines_max_level] = { +/* ABM Level 1, ABM Level 2, ABM Level 3, ABM Level 4 */ +{ 2, 5, 7, 8 }, /* Default - Medium aggressiveness */ +{ 2, 5, 8, 11 }, /* Alt #1 - Increased aggressiveness */ +{ 0, 2, 4, 8 }, /* Alt #2 - Minimal aggressiveness */ +{ 3, 6, 10, 12 }, /* Alt #3 - Super aggressiveness */ +}; + +#define NUM_AMBI_LEVEL 5 +#define NUM_AGGR_LEVEL 4 +#define NUM_POWER_FN_SEGS 8 +#define NUM_BL_CURVE_SEGS 16 +#define IRAM_SIZE 256 + +#define IRAM_RESERVE_AREA_START_V2 0xF0 // reserve 0xF0~0xF6 are write by DMCU only +#define IRAM_RESERVE_AREA_END_V2 0xF6 // reserve 0xF0~0xF6 are write by DMCU only + +#define IRAM_RESERVE_AREA_START_V2_2 0xF0 // reserve 0xF0~0xFF are write by DMCU only +#define IRAM_RESERVE_AREA_END_V2_2 0xFF // reserve 0xF0~0xFF are write by DMCU only + +#pragma pack(push, 1) +/* NOTE: iRAM is 256B in size */ +struct iram_table_v_2 { + /* flags */ + uint16_t flags; /* 0x00 U16 */ + + /* parameters for ABM2.0 algorithm */ + uint8_t min_reduction[NUM_AMBI_LEVEL][NUM_AGGR_LEVEL]; /* 0x02 U0.8 */ + uint8_t max_reduction[NUM_AMBI_LEVEL][NUM_AGGR_LEVEL]; /* 0x16 U0.8 */ + uint8_t bright_pos_gain[NUM_AMBI_LEVEL][NUM_AGGR_LEVEL]; /* 0x2a U2.6 */ + uint8_t bright_neg_gain[NUM_AMBI_LEVEL][NUM_AGGR_LEVEL]; /* 0x3e U2.6 */ + uint8_t dark_pos_gain[NUM_AMBI_LEVEL][NUM_AGGR_LEVEL]; /* 0x52 U2.6 */ + uint8_t dark_neg_gain[NUM_AMBI_LEVEL][NUM_AGGR_LEVEL]; /* 0x66 U2.6 */ + uint8_t iir_curve[NUM_AMBI_LEVEL]; /* 0x7a U0.8 */ + uint8_t deviation_gain; /* 0x7f U0.8 */ + + /* parameters for crgb conversion */ + uint16_t crgb_thresh[NUM_POWER_FN_SEGS]; /* 0x80 U3.13 */ + uint16_t crgb_offset[NUM_POWER_FN_SEGS]; /* 0x90 U1.15 */ + uint16_t crgb_slope[NUM_POWER_FN_SEGS]; /* 0xa0 U4.12 */ + + /* parameters for custom curve */ + /* thresholds for brightness --> backlight */ + uint16_t backlight_thresholds[NUM_BL_CURVE_SEGS]; /* 0xb0 U16.0 */ + /* offsets for brightness --> backlight */ + uint16_t backlight_offsets[NUM_BL_CURVE_SEGS]; /* 0xd0 U16.0 */ + + /* For reading PSR State directly from IRAM */ + uint8_t psr_state; /* 0xf0 */ + uint8_t dmcu_mcp_interface_version; /* 0xf1 */ + uint8_t dmcu_abm_feature_version; /* 0xf2 */ + uint8_t dmcu_psr_feature_version; /* 0xf3 */ + uint16_t dmcu_version; /* 0xf4 */ + uint8_t dmcu_state; /* 0xf6 */ + + uint16_t blRampReduction; /* 0xf7 */ + uint16_t blRampStart; /* 0xf9 */ + uint8_t dummy5; /* 0xfb */ + uint8_t dummy6; /* 0xfc */ + uint8_t dummy7; /* 0xfd */ + uint8_t dummy8; /* 0xfe */ + uint8_t dummy9; /* 0xff */ +}; + +struct iram_table_v_2_2 { + /* flags */ + uint16_t flags; /* 0x00 U16 */ + + /* parameters for ABM2.2 algorithm */ + uint8_t min_reduction[NUM_AMBI_LEVEL][NUM_AGGR_LEVEL]; /* 0x02 U0.8 */ + uint8_t max_reduction[NUM_AMBI_LEVEL][NUM_AGGR_LEVEL]; /* 0x16 U0.8 */ + uint8_t bright_pos_gain[NUM_AMBI_LEVEL][NUM_AGGR_LEVEL]; /* 0x2a U2.6 */ + uint8_t dark_pos_gain[NUM_AMBI_LEVEL][NUM_AGGR_LEVEL]; /* 0x3e U2.6 */ + uint8_t hybridFactor[NUM_AGGR_LEVEL]; /* 0x52 U0.8 */ + uint8_t contrastFactor[NUM_AGGR_LEVEL]; /* 0x56 U0.8 */ + uint8_t deviation_gain[NUM_AGGR_LEVEL]; /* 0x5a U0.8 */ + uint8_t iir_curve[NUM_AMBI_LEVEL]; /* 0x5e U0.8 */ + uint8_t pad[29]; /* 0x63 U0.8 */ + + /* parameters for crgb conversion */ + uint16_t crgb_thresh[NUM_POWER_FN_SEGS]; /* 0x80 U3.13 */ + uint16_t crgb_offset[NUM_POWER_FN_SEGS]; /* 0x90 U1.15 */ + uint16_t crgb_slope[NUM_POWER_FN_SEGS]; /* 0xa0 U4.12 */ + + /* parameters for custom curve */ + /* thresholds for brightness --> backlight */ + uint16_t backlight_thresholds[NUM_BL_CURVE_SEGS]; /* 0xb0 U16.0 */ + /* offsets for brightness --> backlight */ + uint16_t backlight_offsets[NUM_BL_CURVE_SEGS]; /* 0xd0 U16.0 */ + + /* For reading PSR State directly from IRAM */ + uint8_t psr_state; /* 0xf0 */ + uint8_t dmcu_mcp_interface_version; /* 0xf1 */ + uint8_t dmcu_abm_feature_version; /* 0xf2 */ + uint8_t dmcu_psr_feature_version; /* 0xf3 */ + uint16_t dmcu_version; /* 0xf4 */ + uint8_t dmcu_state; /* 0xf6 */ + + uint8_t dummy1; /* 0xf7 */ + uint8_t dummy2; /* 0xf8 */ + uint8_t dummy3; /* 0xf9 */ + uint8_t dummy4; /* 0xfa */ + uint8_t dummy5; /* 0xfb */ + uint8_t dummy6; /* 0xfc */ + uint8_t dummy7; /* 0xfd */ + uint8_t dummy8; /* 0xfe */ + uint8_t dummy9; /* 0xff */ +}; +#pragma pack(pop) + +static void fill_backlight_transform_table(struct dmcu_iram_parameters params, + struct iram_table_v_2 *table) +{ + unsigned int i; + unsigned int num_entries = NUM_BL_CURVE_SEGS; + unsigned int lut_index; + + table->backlight_thresholds[0] = 0; + table->backlight_offsets[0] = params.backlight_lut_array[0]; + table->backlight_thresholds[num_entries-1] = 0xFFFF; + table->backlight_offsets[num_entries-1] = + params.backlight_lut_array[params.backlight_lut_array_size - 1]; + + /* Setup all brightness levels between 0% and 100% exclusive + * Fills brightness-to-backlight transform table. Backlight custom curve + * describes transform from brightness to backlight. It will be defined + * as set of thresholds and set of offsets, together, implying + * extrapolation of custom curve into 16 uniformly spanned linear + * segments. Each threshold/offset represented by 16 bit entry in + * format U4.10. + */ + for (i = 1; i+1 < num_entries; i++) { + lut_index = (params.backlight_lut_array_size - 1) * i / (num_entries - 1); + ASSERT(lut_index < params.backlight_lut_array_size); + + table->backlight_thresholds[i] = + cpu_to_be16(DIV_ROUNDUP((i * 65536), num_entries)); + table->backlight_offsets[i] = + cpu_to_be16(params.backlight_lut_array[lut_index]); + } +} + +static void fill_backlight_transform_table_v_2_2(struct dmcu_iram_parameters params, + struct iram_table_v_2_2 *table) +{ + unsigned int i; + unsigned int num_entries = NUM_BL_CURVE_SEGS; + unsigned int lut_index; + + table->backlight_thresholds[0] = 0; + table->backlight_offsets[0] = params.backlight_lut_array[0]; + table->backlight_thresholds[num_entries-1] = 0xFFFF; + table->backlight_offsets[num_entries-1] = + params.backlight_lut_array[params.backlight_lut_array_size - 1]; + + /* Setup all brightness levels between 0% and 100% exclusive + * Fills brightness-to-backlight transform table. Backlight custom curve + * describes transform from brightness to backlight. It will be defined + * as set of thresholds and set of offsets, together, implying + * extrapolation of custom curve into 16 uniformly spanned linear + * segments. Each threshold/offset represented by 16 bit entry in + * format U4.10. + */ + for (i = 1; i+1 < num_entries; i++) { + lut_index = (params.backlight_lut_array_size - 1) * i / (num_entries - 1); + ASSERT(lut_index < params.backlight_lut_array_size); + + table->backlight_thresholds[i] = + cpu_to_be16(DIV_ROUNDUP((i * 65536), num_entries)); + table->backlight_offsets[i] = + cpu_to_be16(params.backlight_lut_array[lut_index]); + } +} + +void fill_iram_v_2(struct iram_table_v_2 *ram_table, struct dmcu_iram_parameters params) +{ + unsigned int set = params.set; + + ram_table->flags = 0x0; + ram_table->deviation_gain = 0xb3; + + ram_table->blRampReduction = + cpu_to_be16(params.backlight_ramping_reduction); + ram_table->blRampStart = + cpu_to_be16(params.backlight_ramping_start); + + ram_table->min_reduction[0][0] = min_reduction_table[abm_config[set][0]]; + ram_table->min_reduction[1][0] = min_reduction_table[abm_config[set][0]]; + ram_table->min_reduction[2][0] = min_reduction_table[abm_config[set][0]]; + ram_table->min_reduction[3][0] = min_reduction_table[abm_config[set][0]]; + ram_table->min_reduction[4][0] = min_reduction_table[abm_config[set][0]]; + ram_table->max_reduction[0][0] = max_reduction_table[abm_config[set][0]]; + ram_table->max_reduction[1][0] = max_reduction_table[abm_config[set][0]]; + ram_table->max_reduction[2][0] = max_reduction_table[abm_config[set][0]]; + ram_table->max_reduction[3][0] = max_reduction_table[abm_config[set][0]]; + ram_table->max_reduction[4][0] = max_reduction_table[abm_config[set][0]]; + + ram_table->min_reduction[0][1] = min_reduction_table[abm_config[set][1]]; + ram_table->min_reduction[1][1] = min_reduction_table[abm_config[set][1]]; + ram_table->min_reduction[2][1] = min_reduction_table[abm_config[set][1]]; + ram_table->min_reduction[3][1] = min_reduction_table[abm_config[set][1]]; + ram_table->min_reduction[4][1] = min_reduction_table[abm_config[set][1]]; + ram_table->max_reduction[0][1] = max_reduction_table[abm_config[set][1]]; + ram_table->max_reduction[1][1] = max_reduction_table[abm_config[set][1]]; + ram_table->max_reduction[2][1] = max_reduction_table[abm_config[set][1]]; + ram_table->max_reduction[3][1] = max_reduction_table[abm_config[set][1]]; + ram_table->max_reduction[4][1] = max_reduction_table[abm_config[set][1]]; + + ram_table->min_reduction[0][2] = min_reduction_table[abm_config[set][2]]; + ram_table->min_reduction[1][2] = min_reduction_table[abm_config[set][2]]; + ram_table->min_reduction[2][2] = min_reduction_table[abm_config[set][2]]; + ram_table->min_reduction[3][2] = min_reduction_table[abm_config[set][2]]; + ram_table->min_reduction[4][2] = min_reduction_table[abm_config[set][2]]; + ram_table->max_reduction[0][2] = max_reduction_table[abm_config[set][2]]; + ram_table->max_reduction[1][2] = max_reduction_table[abm_config[set][2]]; + ram_table->max_reduction[2][2] = max_reduction_table[abm_config[set][2]]; + ram_table->max_reduction[3][2] = max_reduction_table[abm_config[set][2]]; + ram_table->max_reduction[4][2] = max_reduction_table[abm_config[set][2]]; + + ram_table->min_reduction[0][3] = min_reduction_table[abm_config[set][3]]; + ram_table->min_reduction[1][3] = min_reduction_table[abm_config[set][3]]; + ram_table->min_reduction[2][3] = min_reduction_table[abm_config[set][3]]; + ram_table->min_reduction[3][3] = min_reduction_table[abm_config[set][3]]; + ram_table->min_reduction[4][3] = min_reduction_table[abm_config[set][3]]; + ram_table->max_reduction[0][3] = max_reduction_table[abm_config[set][3]]; + ram_table->max_reduction[1][3] = max_reduction_table[abm_config[set][3]]; + ram_table->max_reduction[2][3] = max_reduction_table[abm_config[set][3]]; + ram_table->max_reduction[3][3] = max_reduction_table[abm_config[set][3]]; + ram_table->max_reduction[4][3] = max_reduction_table[abm_config[set][3]]; + + ram_table->bright_pos_gain[0][0] = 0x20; + ram_table->bright_pos_gain[0][1] = 0x20; + ram_table->bright_pos_gain[0][2] = 0x20; + ram_table->bright_pos_gain[0][3] = 0x20; + ram_table->bright_pos_gain[1][0] = 0x20; + ram_table->bright_pos_gain[1][1] = 0x20; + ram_table->bright_pos_gain[1][2] = 0x20; + ram_table->bright_pos_gain[1][3] = 0x20; + ram_table->bright_pos_gain[2][0] = 0x20; + ram_table->bright_pos_gain[2][1] = 0x20; + ram_table->bright_pos_gain[2][2] = 0x20; + ram_table->bright_pos_gain[2][3] = 0x20; + ram_table->bright_pos_gain[3][0] = 0x20; + ram_table->bright_pos_gain[3][1] = 0x20; + ram_table->bright_pos_gain[3][2] = 0x20; + ram_table->bright_pos_gain[3][3] = 0x20; + ram_table->bright_pos_gain[4][0] = 0x20; + ram_table->bright_pos_gain[4][1] = 0x20; + ram_table->bright_pos_gain[4][2] = 0x20; + ram_table->bright_pos_gain[4][3] = 0x20; + ram_table->bright_neg_gain[0][1] = 0x00; + ram_table->bright_neg_gain[0][2] = 0x00; + ram_table->bright_neg_gain[0][3] = 0x00; + ram_table->bright_neg_gain[1][0] = 0x00; + ram_table->bright_neg_gain[1][1] = 0x00; + ram_table->bright_neg_gain[1][2] = 0x00; + ram_table->bright_neg_gain[1][3] = 0x00; + ram_table->bright_neg_gain[2][0] = 0x00; + ram_table->bright_neg_gain[2][1] = 0x00; + ram_table->bright_neg_gain[2][2] = 0x00; + ram_table->bright_neg_gain[2][3] = 0x00; + ram_table->bright_neg_gain[3][0] = 0x00; + ram_table->bright_neg_gain[3][1] = 0x00; + ram_table->bright_neg_gain[3][2] = 0x00; + ram_table->bright_neg_gain[3][3] = 0x00; + ram_table->bright_neg_gain[4][0] = 0x00; + ram_table->bright_neg_gain[4][1] = 0x00; + ram_table->bright_neg_gain[4][2] = 0x00; + ram_table->bright_neg_gain[4][3] = 0x00; + ram_table->dark_pos_gain[0][0] = 0x00; + ram_table->dark_pos_gain[0][1] = 0x00; + ram_table->dark_pos_gain[0][2] = 0x00; + ram_table->dark_pos_gain[0][3] = 0x00; + ram_table->dark_pos_gain[1][0] = 0x00; + ram_table->dark_pos_gain[1][1] = 0x00; + ram_table->dark_pos_gain[1][2] = 0x00; + ram_table->dark_pos_gain[1][3] = 0x00; + ram_table->dark_pos_gain[2][0] = 0x00; + ram_table->dark_pos_gain[2][1] = 0x00; + ram_table->dark_pos_gain[2][2] = 0x00; + ram_table->dark_pos_gain[2][3] = 0x00; + ram_table->dark_pos_gain[3][0] = 0x00; + ram_table->dark_pos_gain[3][1] = 0x00; + ram_table->dark_pos_gain[3][2] = 0x00; + ram_table->dark_pos_gain[3][3] = 0x00; + ram_table->dark_pos_gain[4][0] = 0x00; + ram_table->dark_pos_gain[4][1] = 0x00; + ram_table->dark_pos_gain[4][2] = 0x00; + ram_table->dark_pos_gain[4][3] = 0x00; + ram_table->dark_neg_gain[0][0] = 0x00; + ram_table->dark_neg_gain[0][1] = 0x00; + ram_table->dark_neg_gain[0][2] = 0x00; + ram_table->dark_neg_gain[0][3] = 0x00; + ram_table->dark_neg_gain[1][0] = 0x00; + ram_table->dark_neg_gain[1][1] = 0x00; + ram_table->dark_neg_gain[1][2] = 0x00; + ram_table->dark_neg_gain[1][3] = 0x00; + ram_table->dark_neg_gain[2][0] = 0x00; + ram_table->dark_neg_gain[2][1] = 0x00; + ram_table->dark_neg_gain[2][2] = 0x00; + ram_table->dark_neg_gain[2][3] = 0x00; + ram_table->dark_neg_gain[3][0] = 0x00; + ram_table->dark_neg_gain[3][1] = 0x00; + ram_table->dark_neg_gain[3][2] = 0x00; + ram_table->dark_neg_gain[3][3] = 0x00; + ram_table->dark_neg_gain[4][0] = 0x00; + ram_table->dark_neg_gain[4][1] = 0x00; + ram_table->dark_neg_gain[4][2] = 0x00; + ram_table->dark_neg_gain[4][3] = 0x00; + + ram_table->iir_curve[0] = 0x65; + ram_table->iir_curve[1] = 0x65; + ram_table->iir_curve[2] = 0x65; + ram_table->iir_curve[3] = 0x65; + ram_table->iir_curve[4] = 0x65; + + //Gamma 2.4 + ram_table->crgb_thresh[0] = cpu_to_be16(0x13b6); + ram_table->crgb_thresh[1] = cpu_to_be16(0x1648); + ram_table->crgb_thresh[2] = cpu_to_be16(0x18e3); + ram_table->crgb_thresh[3] = cpu_to_be16(0x1b41); + ram_table->crgb_thresh[4] = cpu_to_be16(0x1d46); + ram_table->crgb_thresh[5] = cpu_to_be16(0x1f21); + ram_table->crgb_thresh[6] = cpu_to_be16(0x2167); + ram_table->crgb_thresh[7] = cpu_to_be16(0x2384); + ram_table->crgb_offset[0] = cpu_to_be16(0x2999); + ram_table->crgb_offset[1] = cpu_to_be16(0x3999); + ram_table->crgb_offset[2] = cpu_to_be16(0x4666); + ram_table->crgb_offset[3] = cpu_to_be16(0x5999); + ram_table->crgb_offset[4] = cpu_to_be16(0x6333); + ram_table->crgb_offset[5] = cpu_to_be16(0x7800); + ram_table->crgb_offset[6] = cpu_to_be16(0x8c00); + ram_table->crgb_offset[7] = cpu_to_be16(0xa000); + ram_table->crgb_slope[0] = cpu_to_be16(0x3147); + ram_table->crgb_slope[1] = cpu_to_be16(0x2978); + ram_table->crgb_slope[2] = cpu_to_be16(0x23a2); + ram_table->crgb_slope[3] = cpu_to_be16(0x1f55); + ram_table->crgb_slope[4] = cpu_to_be16(0x1c63); + ram_table->crgb_slope[5] = cpu_to_be16(0x1a0f); + ram_table->crgb_slope[6] = cpu_to_be16(0x178d); + ram_table->crgb_slope[7] = cpu_to_be16(0x15ab); + + fill_backlight_transform_table( + params, ram_table); +} + +void fill_iram_v_2_2(struct iram_table_v_2_2 *ram_table, struct dmcu_iram_parameters params) +{ + unsigned int set = params.set; + + ram_table->flags = 0x0; + + ram_table->deviation_gain[0] = 0xb3; + ram_table->deviation_gain[1] = 0xb3; + ram_table->deviation_gain[2] = 0xb3; + ram_table->deviation_gain[3] = 0xb3; + + ram_table->min_reduction[0][0] = min_reduction_table_v_2_2[abm_config[set][0]]; + ram_table->min_reduction[1][0] = min_reduction_table_v_2_2[abm_config[set][0]]; + ram_table->min_reduction[2][0] = min_reduction_table_v_2_2[abm_config[set][0]]; + ram_table->min_reduction[3][0] = min_reduction_table_v_2_2[abm_config[set][0]]; + ram_table->min_reduction[4][0] = min_reduction_table_v_2_2[abm_config[set][0]]; + ram_table->max_reduction[0][0] = max_reduction_table_v_2_2[abm_config[set][0]]; + ram_table->max_reduction[1][0] = max_reduction_table_v_2_2[abm_config[set][0]]; + ram_table->max_reduction[2][0] = max_reduction_table_v_2_2[abm_config[set][0]]; + ram_table->max_reduction[3][0] = max_reduction_table_v_2_2[abm_config[set][0]]; + ram_table->max_reduction[4][0] = max_reduction_table_v_2_2[abm_config[set][0]]; + + ram_table->min_reduction[0][1] = min_reduction_table_v_2_2[abm_config[set][1]]; + ram_table->min_reduction[1][1] = min_reduction_table_v_2_2[abm_config[set][1]]; + ram_table->min_reduction[2][1] = min_reduction_table_v_2_2[abm_config[set][1]]; + ram_table->min_reduction[3][1] = min_reduction_table_v_2_2[abm_config[set][1]]; + ram_table->min_reduction[4][1] = min_reduction_table_v_2_2[abm_config[set][1]]; + ram_table->max_reduction[0][1] = max_reduction_table_v_2_2[abm_config[set][1]]; + ram_table->max_reduction[1][1] = max_reduction_table_v_2_2[abm_config[set][1]]; + ram_table->max_reduction[2][1] = max_reduction_table_v_2_2[abm_config[set][1]]; + ram_table->max_reduction[3][1] = max_reduction_table_v_2_2[abm_config[set][1]]; + ram_table->max_reduction[4][1] = max_reduction_table_v_2_2[abm_config[set][1]]; + + ram_table->min_reduction[0][2] = min_reduction_table_v_2_2[abm_config[set][2]]; + ram_table->min_reduction[1][2] = min_reduction_table_v_2_2[abm_config[set][2]]; + ram_table->min_reduction[2][2] = min_reduction_table_v_2_2[abm_config[set][2]]; + ram_table->min_reduction[3][2] = min_reduction_table_v_2_2[abm_config[set][2]]; + ram_table->min_reduction[4][2] = min_reduction_table_v_2_2[abm_config[set][2]]; + ram_table->max_reduction[0][2] = max_reduction_table_v_2_2[abm_config[set][2]]; + ram_table->max_reduction[1][2] = max_reduction_table_v_2_2[abm_config[set][2]]; + ram_table->max_reduction[2][2] = max_reduction_table_v_2_2[abm_config[set][2]]; + ram_table->max_reduction[3][2] = max_reduction_table_v_2_2[abm_config[set][2]]; + ram_table->max_reduction[4][2] = max_reduction_table_v_2_2[abm_config[set][2]]; + + ram_table->min_reduction[0][3] = min_reduction_table_v_2_2[abm_config[set][3]]; + ram_table->min_reduction[1][3] = min_reduction_table_v_2_2[abm_config[set][3]]; + ram_table->min_reduction[2][3] = min_reduction_table_v_2_2[abm_config[set][3]]; + ram_table->min_reduction[3][3] = min_reduction_table_v_2_2[abm_config[set][3]]; + ram_table->min_reduction[4][3] = min_reduction_table_v_2_2[abm_config[set][3]]; + ram_table->max_reduction[0][3] = max_reduction_table_v_2_2[abm_config[set][3]]; + ram_table->max_reduction[1][3] = max_reduction_table_v_2_2[abm_config[set][3]]; + ram_table->max_reduction[2][3] = max_reduction_table_v_2_2[abm_config[set][3]]; + ram_table->max_reduction[3][3] = max_reduction_table_v_2_2[abm_config[set][3]]; + ram_table->max_reduction[4][3] = max_reduction_table_v_2_2[abm_config[set][3]]; + + ram_table->bright_pos_gain[0][0] = 0x20; + ram_table->bright_pos_gain[0][1] = 0x20; + ram_table->bright_pos_gain[0][2] = 0x20; + ram_table->bright_pos_gain[0][3] = 0x20; + ram_table->bright_pos_gain[1][0] = 0x20; + ram_table->bright_pos_gain[1][1] = 0x20; + ram_table->bright_pos_gain[1][2] = 0x20; + ram_table->bright_pos_gain[1][3] = 0x20; + ram_table->bright_pos_gain[2][0] = 0x20; + ram_table->bright_pos_gain[2][1] = 0x20; + ram_table->bright_pos_gain[2][2] = 0x20; + ram_table->bright_pos_gain[2][3] = 0x20; + ram_table->bright_pos_gain[3][0] = 0x20; + ram_table->bright_pos_gain[3][1] = 0x20; + ram_table->bright_pos_gain[3][2] = 0x20; + ram_table->bright_pos_gain[3][3] = 0x20; + ram_table->bright_pos_gain[4][0] = 0x20; + ram_table->bright_pos_gain[4][1] = 0x20; + ram_table->bright_pos_gain[4][2] = 0x20; + ram_table->bright_pos_gain[4][3] = 0x20; + + ram_table->dark_pos_gain[0][0] = 0x00; + ram_table->dark_pos_gain[0][1] = 0x00; + ram_table->dark_pos_gain[0][2] = 0x00; + ram_table->dark_pos_gain[0][3] = 0x00; + ram_table->dark_pos_gain[1][0] = 0x00; + ram_table->dark_pos_gain[1][1] = 0x00; + ram_table->dark_pos_gain[1][2] = 0x00; + ram_table->dark_pos_gain[1][3] = 0x00; + ram_table->dark_pos_gain[2][0] = 0x00; + ram_table->dark_pos_gain[2][1] = 0x00; + ram_table->dark_pos_gain[2][2] = 0x00; + ram_table->dark_pos_gain[2][3] = 0x00; + ram_table->dark_pos_gain[3][0] = 0x00; + ram_table->dark_pos_gain[3][1] = 0x00; + ram_table->dark_pos_gain[3][2] = 0x00; + ram_table->dark_pos_gain[3][3] = 0x00; + ram_table->dark_pos_gain[4][0] = 0x00; + ram_table->dark_pos_gain[4][1] = 0x00; + ram_table->dark_pos_gain[4][2] = 0x00; + ram_table->dark_pos_gain[4][3] = 0x00; + + ram_table->hybridFactor[0] = 0xff; + ram_table->hybridFactor[1] = 0xff; + ram_table->hybridFactor[2] = 0xff; + ram_table->hybridFactor[3] = 0xc0; + + ram_table->contrastFactor[0] = 0x99; + ram_table->contrastFactor[1] = 0x99; + ram_table->contrastFactor[2] = 0x99; + ram_table->contrastFactor[3] = 0x80; + + ram_table->iir_curve[0] = 0x65; + ram_table->iir_curve[1] = 0x65; + ram_table->iir_curve[2] = 0x65; + ram_table->iir_curve[3] = 0x65; + ram_table->iir_curve[4] = 0x65; + + //Gamma 2.2 + ram_table->crgb_thresh[0] = cpu_to_be16(0x127c); + ram_table->crgb_thresh[1] = cpu_to_be16(0x151b); + ram_table->crgb_thresh[2] = cpu_to_be16(0x17d5); + ram_table->crgb_thresh[3] = cpu_to_be16(0x1a56); + ram_table->crgb_thresh[4] = cpu_to_be16(0x1c83); + ram_table->crgb_thresh[5] = cpu_to_be16(0x1e72); + ram_table->crgb_thresh[6] = cpu_to_be16(0x20f0); + ram_table->crgb_thresh[7] = cpu_to_be16(0x232b); + ram_table->crgb_offset[0] = cpu_to_be16(0x2999); + ram_table->crgb_offset[1] = cpu_to_be16(0x3999); + ram_table->crgb_offset[2] = cpu_to_be16(0x4666); + ram_table->crgb_offset[3] = cpu_to_be16(0x5999); + ram_table->crgb_offset[4] = cpu_to_be16(0x6333); + ram_table->crgb_offset[5] = cpu_to_be16(0x7800); + ram_table->crgb_offset[6] = cpu_to_be16(0x8c00); + ram_table->crgb_offset[7] = cpu_to_be16(0xa000); + ram_table->crgb_slope[0] = cpu_to_be16(0x3609); + ram_table->crgb_slope[1] = cpu_to_be16(0x2dfa); + ram_table->crgb_slope[2] = cpu_to_be16(0x27ea); + ram_table->crgb_slope[3] = cpu_to_be16(0x235d); + ram_table->crgb_slope[4] = cpu_to_be16(0x2042); + ram_table->crgb_slope[5] = cpu_to_be16(0x1dc3); + ram_table->crgb_slope[6] = cpu_to_be16(0x1b1a); + ram_table->crgb_slope[7] = cpu_to_be16(0x1910); + + fill_backlight_transform_table_v_2_2( + params, ram_table); +} + +bool dmcu_load_iram(struct dmcu *dmcu, + struct dmcu_iram_parameters params) +{ + unsigned char ram_table[IRAM_SIZE]; + bool result = false; + + if (dmcu == NULL) + return false; + + if (!dmcu->funcs->is_dmcu_initialized(dmcu)) + return true; + + memset(&ram_table, 0, sizeof(ram_table)); + + if (dmcu->dmcu_version.abm_version == 0x22) { + fill_iram_v_2_2((struct iram_table_v_2_2 *)ram_table, params); + + result = dmcu->funcs->load_iram( + dmcu, 0, (char *)(&ram_table), IRAM_RESERVE_AREA_START_V2_2); + } else { + fill_iram_v_2((struct iram_table_v_2 *)ram_table, params); + + result = dmcu->funcs->load_iram( + dmcu, 0, (char *)(&ram_table), IRAM_RESERVE_AREA_START_V2); + + if (result) + result = dmcu->funcs->load_iram( + dmcu, IRAM_RESERVE_AREA_END_V2 + 1, + (char *)(&ram_table) + IRAM_RESERVE_AREA_END_V2 + 1, + sizeof(ram_table) - IRAM_RESERVE_AREA_END_V2 - 1); + } + + return result; +} diff --git a/drivers/gpu/drm/amd/display/dc/i2caux/dce80/i2c_sw_engine_dce80.h b/drivers/gpu/drm/amd/display/modules/power/power_helpers.h index 26355c088746..da5df00fedce 100644 --- a/drivers/gpu/drm/amd/display/dc/i2caux/dce80/i2c_sw_engine_dce80.h +++ b/drivers/gpu/drm/amd/display/modules/power/power_helpers.h @@ -1,5 +1,4 @@ -/* - * Copyright 2012-15 Advanced Micro Devices, Inc. +/* Copyright 2018 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"), @@ -23,21 +22,26 @@ * */ -#ifndef __DAL_I2C_SW_ENGINE_DCE80_H__ -#define __DAL_I2C_SW_ENGINE_DCE80_H__ +#ifndef MODULES_POWER_POWER_HELPERS_H_ +#define MODULES_POWER_POWER_HELPERS_H_ -struct i2c_sw_engine_dce80 { - struct i2c_sw_engine base; - uint32_t engine_id; +#include "dc/inc/hw/dmcu.h" + + +enum abm_defines { + abm_defines_max_level = 4, + abm_defines_max_config = 4, }; -struct i2c_sw_engine_dce80_create_arg { - uint32_t engine_id; - uint32_t default_speed; - struct dc_context *ctx; +struct dmcu_iram_parameters { + unsigned int *backlight_lut_array; + unsigned int backlight_lut_array_size; + unsigned int backlight_ramping_reduction; + unsigned int backlight_ramping_start; + unsigned int set; }; -struct i2c_engine *dal_i2c_sw_engine_dce80_create( - const struct i2c_sw_engine_dce80_create_arg *arg); +bool dmcu_load_iram(struct dmcu *dmcu, + struct dmcu_iram_parameters params); -#endif +#endif /* MODULES_POWER_POWER_HELPERS_H_ */ diff --git a/drivers/gpu/drm/amd/include/amd_acpi.h b/drivers/gpu/drm/amd/include/amd_acpi.h index 9b9699fc433f..c72cbfe8f684 100644 --- a/drivers/gpu/drm/amd/include/amd_acpi.h +++ b/drivers/gpu/drm/amd/include/amd_acpi.h @@ -52,6 +52,30 @@ struct atif_sbios_requests { u8 backlight_level; /* panel backlight level (0-255) */ } __packed; +struct atif_qbtc_arguments { + u16 size; /* structure size in bytes (includes size field) */ + u8 requested_display; /* which display is requested */ +} __packed; + +#define ATIF_QBTC_MAX_DATA_POINTS 99 + +struct atif_qbtc_data_point { + u8 luminance; /* luminance in percent */ + u8 ipnut_signal; /* input signal in range 0-255 */ +} __packed; + +struct atif_qbtc_output { + u16 size; /* structure size in bytes (includes size field) */ + u16 flags; /* all zeroes */ + u8 error_code; /* error code */ + u8 ac_level; /* default brightness on AC power */ + u8 dc_level; /* default brightness on DC power */ + u8 min_input_signal; /* max input signal in range 0-255 */ + u8 max_input_signal; /* min input signal in range 0-255 */ + u8 number_of_points; /* number of data points */ + struct atif_qbtc_data_point data_points[ATIF_QBTC_MAX_DATA_POINTS]; +} __packed; + #define ATIF_NOTIFY_MASK 0x3 #define ATIF_NOTIFY_NONE 0 #define ATIF_NOTIFY_81 1 @@ -126,26 +150,18 @@ struct atcs_pref_req_output { * DWORD - supported functions bit vector */ /* Notifications mask */ -# define ATIF_DISPLAY_SWITCH_REQUEST_SUPPORTED (1 << 0) -# define ATIF_EXPANSION_MODE_CHANGE_REQUEST_SUPPORTED (1 << 1) # define ATIF_THERMAL_STATE_CHANGE_REQUEST_SUPPORTED (1 << 2) # define ATIF_FORCED_POWER_STATE_CHANGE_REQUEST_SUPPORTED (1 << 3) # define ATIF_SYSTEM_POWER_SOURCE_CHANGE_REQUEST_SUPPORTED (1 << 4) -# define ATIF_DISPLAY_CONF_CHANGE_REQUEST_SUPPORTED (1 << 5) -# define ATIF_PX_GFX_SWITCH_REQUEST_SUPPORTED (1 << 6) # define ATIF_PANEL_BRIGHTNESS_CHANGE_REQUEST_SUPPORTED (1 << 7) # define ATIF_DGPU_DISPLAY_EVENT_SUPPORTED (1 << 8) +# define ATIF_GPU_PACKAGE_POWER_LIMIT_REQUEST_SUPPORTED (1 << 12) /* supported functions vector */ # define ATIF_GET_SYSTEM_PARAMETERS_SUPPORTED (1 << 0) # define ATIF_GET_SYSTEM_BIOS_REQUESTS_SUPPORTED (1 << 1) -# define ATIF_SELECT_ACTIVE_DISPLAYS_SUPPORTED (1 << 2) -# define ATIF_GET_LID_STATE_SUPPORTED (1 << 3) -# define ATIF_GET_TV_STANDARD_FROM_CMOS_SUPPORTED (1 << 4) -# define ATIF_SET_TV_STANDARD_IN_CMOS_SUPPORTED (1 << 5) -# define ATIF_GET_PANEL_EXPANSION_MODE_FROM_CMOS_SUPPORTED (1 << 6) -# define ATIF_SET_PANEL_EXPANSION_MODE_IN_CMOS_SUPPORTED (1 << 7) # define ATIF_TEMPERATURE_CHANGE_NOTIFICATION_SUPPORTED (1 << 12) -# define ATIF_GET_GRAPHICS_DEVICE_TYPES_SUPPORTED (1 << 14) +# define ATIF_QUERY_BACKLIGHT_TRANSFER_CHARACTERISTICS_SUPPORTED (1 << 15) +# define ATIF_READY_TO_UNDOCK_NOTIFICATION_SUPPORTED (1 << 16) # define ATIF_GET_EXTERNAL_GPU_INFORMATION_SUPPORTED (1 << 20) #define ATIF_FUNCTION_GET_SYSTEM_PARAMETERS 0x1 /* ARG0: ATIF_FUNCTION_GET_SYSTEM_PARAMETERS @@ -170,6 +186,10 @@ struct atcs_pref_req_output { * n (0xd0-0xd9) is specified in notify command code. * bit 2: * 1 - lid changes not reported though int10 + * bit 3: + * 1 - system bios controls overclocking + * bit 4: + * 1 - enable overclocking */ #define ATIF_FUNCTION_GET_SYSTEM_BIOS_REQUESTS 0x2 /* ARG0: ATIF_FUNCTION_GET_SYSTEM_BIOS_REQUESTS @@ -177,28 +197,23 @@ struct atcs_pref_req_output { * OUTPUT: * WORD - structure size in bytes (includes size field) * DWORD - pending sbios requests - * BYTE - panel expansion mode + * BYTE - reserved (all zeroes) * BYTE - thermal state: target gfx controller * BYTE - thermal state: state id (0: exit state, non-0: state) * BYTE - forced power state: target gfx controller - * BYTE - forced power state: state id + * BYTE - forced power state: state id (0: forced state, non-0: state) * BYTE - system power source * BYTE - panel backlight level (0-255) + * BYTE - GPU package power limit: target gfx controller + * DWORD - GPU package power limit: value (24:8 fractional format, Watts) */ /* pending sbios requests */ -# define ATIF_DISPLAY_SWITCH_REQUEST (1 << 0) -# define ATIF_EXPANSION_MODE_CHANGE_REQUEST (1 << 1) # define ATIF_THERMAL_STATE_CHANGE_REQUEST (1 << 2) # define ATIF_FORCED_POWER_STATE_CHANGE_REQUEST (1 << 3) # define ATIF_SYSTEM_POWER_SOURCE_CHANGE_REQUEST (1 << 4) -# define ATIF_DISPLAY_CONF_CHANGE_REQUEST (1 << 5) -# define ATIF_PX_GFX_SWITCH_REQUEST (1 << 6) # define ATIF_PANEL_BRIGHTNESS_CHANGE_REQUEST (1 << 7) # define ATIF_DGPU_DISPLAY_EVENT (1 << 8) -/* panel expansion mode */ -# define ATIF_PANEL_EXPANSION_DISABLE 0 -# define ATIF_PANEL_EXPANSION_FULL 1 -# define ATIF_PANEL_EXPANSION_ASPECT 2 +# define ATIF_GPU_PACKAGE_POWER_LIMIT_REQUEST (1 << 12) /* target gfx controller */ # define ATIF_TARGET_GFX_SINGLE 0 # define ATIF_TARGET_GFX_PX_IGPU 1 @@ -208,76 +223,6 @@ struct atcs_pref_req_output { # define ATIF_POWER_SOURCE_DC 2 # define ATIF_POWER_SOURCE_RESTRICTED_AC_1 3 # define ATIF_POWER_SOURCE_RESTRICTED_AC_2 4 -#define ATIF_FUNCTION_SELECT_ACTIVE_DISPLAYS 0x3 -/* ARG0: ATIF_FUNCTION_SELECT_ACTIVE_DISPLAYS - * ARG1: - * WORD - structure size in bytes (includes size field) - * WORD - selected displays - * WORD - connected displays - * OUTPUT: - * WORD - structure size in bytes (includes size field) - * WORD - selected displays - */ -# define ATIF_LCD1 (1 << 0) -# define ATIF_CRT1 (1 << 1) -# define ATIF_TV (1 << 2) -# define ATIF_DFP1 (1 << 3) -# define ATIF_CRT2 (1 << 4) -# define ATIF_LCD2 (1 << 5) -# define ATIF_DFP2 (1 << 7) -# define ATIF_CV (1 << 8) -# define ATIF_DFP3 (1 << 9) -# define ATIF_DFP4 (1 << 10) -# define ATIF_DFP5 (1 << 11) -# define ATIF_DFP6 (1 << 12) -#define ATIF_FUNCTION_GET_LID_STATE 0x4 -/* ARG0: ATIF_FUNCTION_GET_LID_STATE - * ARG1: none - * OUTPUT: - * WORD - structure size in bytes (includes size field) - * BYTE - lid state (0: open, 1: closed) - * - * GET_LID_STATE only works at boot and resume, for general lid - * status, use the kernel provided status - */ -#define ATIF_FUNCTION_GET_TV_STANDARD_FROM_CMOS 0x5 -/* ARG0: ATIF_FUNCTION_GET_TV_STANDARD_FROM_CMOS - * ARG1: none - * OUTPUT: - * WORD - structure size in bytes (includes size field) - * BYTE - 0 - * BYTE - TV standard - */ -# define ATIF_TV_STD_NTSC 0 -# define ATIF_TV_STD_PAL 1 -# define ATIF_TV_STD_PALM 2 -# define ATIF_TV_STD_PAL60 3 -# define ATIF_TV_STD_NTSCJ 4 -# define ATIF_TV_STD_PALCN 5 -# define ATIF_TV_STD_PALN 6 -# define ATIF_TV_STD_SCART_RGB 9 -#define ATIF_FUNCTION_SET_TV_STANDARD_IN_CMOS 0x6 -/* ARG0: ATIF_FUNCTION_SET_TV_STANDARD_IN_CMOS - * ARG1: - * WORD - structure size in bytes (includes size field) - * BYTE - 0 - * BYTE - TV standard - * OUTPUT: none - */ -#define ATIF_FUNCTION_GET_PANEL_EXPANSION_MODE_FROM_CMOS 0x7 -/* ARG0: ATIF_FUNCTION_GET_PANEL_EXPANSION_MODE_FROM_CMOS - * ARG1: none - * OUTPUT: - * WORD - structure size in bytes (includes size field) - * BYTE - panel expansion mode - */ -#define ATIF_FUNCTION_SET_PANEL_EXPANSION_MODE_IN_CMOS 0x8 -/* ARG0: ATIF_FUNCTION_SET_PANEL_EXPANSION_MODE_IN_CMOS - * ARG1: - * WORD - structure size in bytes (includes size field) - * BYTE - panel expansion mode - * OUTPUT: none - */ #define ATIF_FUNCTION_TEMPERATURE_CHANGE_NOTIFICATION 0xD /* ARG0: ATIF_FUNCTION_TEMPERATURE_CHANGE_NOTIFICATION * ARG1: @@ -286,21 +231,43 @@ struct atcs_pref_req_output { * BYTE - current temperature (degress Celsius) * OUTPUT: none */ -#define ATIF_FUNCTION_GET_GRAPHICS_DEVICE_TYPES 0xF -/* ARG0: ATIF_FUNCTION_GET_GRAPHICS_DEVICE_TYPES - * ARG1: none +#define ATIF_FUNCTION_QUERY_BRIGHTNESS_TRANSFER_CHARACTERISTICS 0x10 +/* ARG0: ATIF_FUNCTION_QUERY_BRIGHTNESS_TRANSFER_CHARACTERISTICS + * ARG1: + * WORD - structure size in bytes (includes size field) + * BYTE - requested display * OUTPUT: - * WORD - number of gfx devices - * WORD - device structure size in bytes (excludes device size field) - * DWORD - flags \ - * WORD - bus number } repeated structure - * WORD - device number / + * WORD - structure size in bytes (includes size field) + * WORD - flags (currently all 16 bits are reserved) + * BYTE - error code (on failure, disregard all below fields) + * BYTE - AC level (default brightness in percent when machine has full power) + * BYTE - DC level (default brightness in percent when machine is on battery) + * BYTE - min input signal, in range 0-255, corresponding to 0% backlight + * BYTE - max input signal, in range 0-255, corresponding to 100% backlight + * BYTE - number of reported data points + * BYTE - luminance level in percent \ repeated structure + * BYTE - input signal in range 0-255 / does not have entries for 0% and 100% + */ +/* requested display */ +# define ATIF_QBTC_REQUEST_LCD1 0 +# define ATIF_QBTC_REQUEST_CRT1 1 +# define ATIF_QBTC_REQUEST_DFP1 3 +# define ATIF_QBTC_REQUEST_CRT2 4 +# define ATIF_QBTC_REQUEST_LCD2 5 +# define ATIF_QBTC_REQUEST_DFP2 7 +# define ATIF_QBTC_REQUEST_DFP3 9 +# define ATIF_QBTC_REQUEST_DFP4 10 +# define ATIF_QBTC_REQUEST_DFP5 11 +# define ATIF_QBTC_REQUEST_DFP6 12 +/* error code */ +# define ATIF_QBTC_ERROR_CODE_SUCCESS 0 +# define ATIF_QBTC_ERROR_CODE_FAILURE 1 +# define ATIF_QBTC_ERROR_CODE_DEVICE_NOT_SUPPORTED 2 +#define ATIF_FUNCTION_READY_TO_UNDOCK_NOTIFICATION 0x11 +/* ARG0: ATIF_FUNCTION_READY_TO_UNDOCK_NOTIFICATION + * ARG1: none + * OUTPUT: none */ -/* flags */ -# define ATIF_PX_REMOVABLE_GRAPHICS_DEVICE (1 << 0) -# define ATIF_XGP_PORT (1 << 1) -# define ATIF_VGA_ENABLED_GRAPHICS_DEVICE (1 << 2) -# define ATIF_XGP_PORT_IN_DOCK (1 << 3) #define ATIF_FUNCTION_GET_EXTERNAL_GPU_INFORMATION 0x15 /* ARG0: ATIF_FUNCTION_GET_EXTERNAL_GPU_INFORMATION * ARG1: none diff --git a/drivers/gpu/drm/amd/include/asic_reg/mmhub/mmhub_9_4_0_offset.h b/drivers/gpu/drm/amd/include/asic_reg/mmhub/mmhub_9_4_0_offset.h new file mode 100644 index 000000000000..8f515875a34d --- /dev/null +++ b/drivers/gpu/drm/amd/include/asic_reg/mmhub/mmhub_9_4_0_offset.h @@ -0,0 +1,32 @@ +/* + * Copyright (C) 2018 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) 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_9_4_0_OFFSET_HEADER +#define _mmhub_9_4_0_OFFSET_HEADER + + +// addressBlock: mmhub_utcl2_vmsharedpfdec +// base address: 0x6a040 +#define mmMC_VM_XGMI_LFB_CNTL 0x0823 +#define mmMC_VM_XGMI_LFB_CNTL_BASE_IDX 0 +#define mmMC_VM_XGMI_LFB_SIZE 0x0824 +#define mmMC_VM_XGMI_LFB_SIZE_BASE_IDX 0 + +#endif diff --git a/drivers/gpu/drm/amd/include/asic_reg/mmhub/mmhub_9_4_0_sh_mask.h b/drivers/gpu/drm/amd/include/asic_reg/mmhub/mmhub_9_4_0_sh_mask.h new file mode 100644 index 000000000000..0a6b072d191e --- /dev/null +++ b/drivers/gpu/drm/amd/include/asic_reg/mmhub/mmhub_9_4_0_sh_mask.h @@ -0,0 +1,35 @@ +/* + * Copyright (C) 2018 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) 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_9_4_0_SH_MASK_HEADER +#define _mmhub_9_4_0_SH_MASK_HEADER + + +// addressBlock: mmhub_utcl2_vmsharedpfdec +//MC_VM_XGMI_LFB_CNTL +#define MC_VM_XGMI_LFB_CNTL__PF_LFB_REGION__SHIFT 0x0 +#define MC_VM_XGMI_LFB_CNTL__PF_MAX_REGION__SHIFT 0x4 +#define MC_VM_XGMI_LFB_CNTL__PF_LFB_REGION_MASK 0x00000007L +#define MC_VM_XGMI_LFB_CNTL__PF_MAX_REGION_MASK 0x00000070L +//MC_VM_XGMI_LFB_SIZE +#define MC_VM_XGMI_LFB_SIZE__PF_LFB_SIZE__SHIFT 0x0 +#define MC_VM_XGMI_LFB_SIZE__PF_LFB_SIZE_MASK 0x0000FFFFL + +#endif diff --git a/drivers/gpu/drm/amd/include/asic_reg/nbio/nbio_6_1_offset.h b/drivers/gpu/drm/amd/include/asic_reg/nbio/nbio_6_1_offset.h index 13d4de645190..d8e0dd192fdd 100644 --- a/drivers/gpu/drm/amd/include/asic_reg/nbio/nbio_6_1_offset.h +++ b/drivers/gpu/drm/amd/include/asic_reg/nbio/nbio_6_1_offset.h @@ -2247,6 +2247,8 @@ // addressBlock: nbio_nbif_rcc_strap_BIFDEC1[13440..14975] // base address: 0x3480 +#define mmRCC_BIF_STRAP0 0x0000 +#define mmRCC_BIF_STRAP0_BASE_IDX 2 #define mmRCC_DEV0_EPF0_STRAP0 0x000f #define mmRCC_DEV0_EPF0_STRAP0_BASE_IDX 2 diff --git a/drivers/gpu/drm/amd/include/asic_reg/nbio/nbio_6_1_sh_mask.h b/drivers/gpu/drm/amd/include/asic_reg/nbio/nbio_6_1_sh_mask.h index a02b67943372..29af5167cd00 100644 --- a/drivers/gpu/drm/amd/include/asic_reg/nbio/nbio_6_1_sh_mask.h +++ b/drivers/gpu/drm/amd/include/asic_reg/nbio/nbio_6_1_sh_mask.h @@ -16838,6 +16838,10 @@ // addressBlock: nbio_nbif_rcc_strap_BIFDEC1[13440..14975] +//RCC_BIF_STRAP0 +#define RCC_BIF_STRAP0__STRAP_PX_CAPABLE__SHIFT 0x7 +#define RCC_BIF_STRAP0__STRAP_PX_CAPABLE_MASK 0x00000080L + //RCC_DEV0_EPF0_STRAP0 #define RCC_DEV0_EPF0_STRAP0__STRAP_DEVICE_ID_DEV0_F0__SHIFT 0x0 #define RCC_DEV0_EPF0_STRAP0__STRAP_MAJOR_REV_ID_DEV0_F0__SHIFT 0x10 diff --git a/drivers/gpu/drm/amd/include/asic_reg/nbio/nbio_6_1_smn.h b/drivers/gpu/drm/amd/include/asic_reg/nbio/nbio_6_1_smn.h new file mode 100644 index 000000000000..8c75669eb500 --- /dev/null +++ b/drivers/gpu/drm/amd/include/asic_reg/nbio/nbio_6_1_smn.h @@ -0,0 +1,58 @@ +/* + * Copyright (C) 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) 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_6_1_SMN_HEADER +#define _nbio_6_1_SMN_HEADER + + +#define smnCPM_CONTROL 0x11180460 +#define smnPCIE_CNTL2 0x11180070 +#define smnPCIE_CONFIG_CNTL 0x11180044 +#define smnPCIE_CI_CNTL 0x11180080 + + +#define smnPCIE_PERF_COUNT_CNTL 0x11180200 +#define smnPCIE_PERF_CNTL_TXCLK 0x11180204 +#define smnPCIE_PERF_COUNT0_TXCLK 0x11180208 +#define smnPCIE_PERF_COUNT1_TXCLK 0x1118020c +#define smnPCIE_PERF_CNTL_MST_R_CLK 0x11180210 +#define smnPCIE_PERF_COUNT0_MST_R_CLK 0x11180214 +#define smnPCIE_PERF_COUNT1_MST_R_CLK 0x11180218 +#define smnPCIE_PERF_CNTL_MST_C_CLK 0x1118021c +#define smnPCIE_PERF_COUNT0_MST_C_CLK 0x11180220 +#define smnPCIE_PERF_COUNT1_MST_C_CLK 0x11180224 +#define smnPCIE_PERF_CNTL_SLV_R_CLK 0x11180228 +#define smnPCIE_PERF_COUNT0_SLV_R_CLK 0x1118022c +#define smnPCIE_PERF_COUNT1_SLV_R_CLK 0x11180230 +#define smnPCIE_PERF_CNTL_SLV_S_C_CLK 0x11180234 +#define smnPCIE_PERF_COUNT0_SLV_S_C_CLK 0x11180238 +#define smnPCIE_PERF_COUNT1_SLV_S_C_CLK 0x1118023c +#define smnPCIE_PERF_CNTL_SLV_NS_C_CLK 0x11180240 +#define smnPCIE_PERF_COUNT0_SLV_NS_C_CLK 0x11180244 +#define smnPCIE_PERF_COUNT1_SLV_NS_C_CLK 0x11180248 +#define smnPCIE_PERF_CNTL_EVENT0_PORT_SEL 0x1118024c +#define smnPCIE_PERF_CNTL_EVENT1_PORT_SEL 0x11180250 +#define smnPCIE_PERF_CNTL_TXCLK2 0x11180254 +#define smnPCIE_PERF_COUNT0_TXCLK2 0x11180258 +#define smnPCIE_PERF_COUNT1_TXCLK2 0x1118025c + +#endif // _nbio_6_1_SMN_HEADER + diff --git a/drivers/gpu/drm/amd/include/asic_reg/nbio/nbio_7_0_smn.h b/drivers/gpu/drm/amd/include/asic_reg/nbio/nbio_7_0_smn.h new file mode 100644 index 000000000000..5563f0715896 --- /dev/null +++ b/drivers/gpu/drm/amd/include/asic_reg/nbio/nbio_7_0_smn.h @@ -0,0 +1,54 @@ +/* + * Copyright (C) 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) 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_7_0_SMN_HEADER +#define _nbio_7_0_SMN_HEADER + + +#define smnCPM_CONTROL 0x11180460 +#define smnPCIE_CNTL2 0x11180070 + +#define smnPCIE_PERF_COUNT_CNTL 0x11180200 +#define smnPCIE_PERF_CNTL_TXCLK 0x11180204 +#define smnPCIE_PERF_COUNT0_TXCLK 0x11180208 +#define smnPCIE_PERF_COUNT1_TXCLK 0x1118020c +#define smnPCIE_PERF_CNTL_MST_R_CLK 0x11180210 +#define smnPCIE_PERF_COUNT0_MST_R_CLK 0x11180214 +#define smnPCIE_PERF_COUNT1_MST_R_CLK 0x11180218 +#define smnPCIE_PERF_CNTL_MST_C_CLK 0x1118021c +#define smnPCIE_PERF_COUNT0_MST_C_CLK 0x11180220 +#define smnPCIE_PERF_COUNT1_MST_C_CLK 0x11180224 +#define smnPCIE_PERF_CNTL_SLV_R_CLK 0x11180228 +#define smnPCIE_PERF_COUNT0_SLV_R_CLK 0x1118022c +#define smnPCIE_PERF_COUNT1_SLV_R_CLK 0x11180230 +#define smnPCIE_PERF_CNTL_SLV_S_C_CLK 0x11180234 +#define smnPCIE_PERF_COUNT0_SLV_S_C_CLK 0x11180238 +#define smnPCIE_PERF_COUNT1_SLV_S_C_CLK 0x1118023c +#define smnPCIE_PERF_CNTL_SLV_NS_C_CLK 0x11180240 +#define smnPCIE_PERF_COUNT0_SLV_NS_C_CLK 0x11180244 +#define smnPCIE_PERF_COUNT1_SLV_NS_C_CLK 0x11180248 +#define smnPCIE_PERF_CNTL_EVENT0_PORT_SEL 0x1118024c +#define smnPCIE_PERF_CNTL_EVENT1_PORT_SEL 0x11180250 +#define smnPCIE_PERF_CNTL_TXCLK2 0x11180254 +#define smnPCIE_PERF_COUNT0_TXCLK2 0x11180258 +#define smnPCIE_PERF_COUNT1_TXCLK2 0x1118025c + +#endif // _nbio_7_0_SMN_HEADER diff --git a/drivers/gpu/drm/amd/include/asic_reg/nbio/nbio_7_4_0_smn.h b/drivers/gpu/drm/amd/include/asic_reg/nbio/nbio_7_4_0_smn.h new file mode 100644 index 000000000000..c1457d880c4d --- /dev/null +++ b/drivers/gpu/drm/amd/include/asic_reg/nbio/nbio_7_4_0_smn.h @@ -0,0 +1,53 @@ +/* + * Copyright (C) 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) 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_7_4_0_SMN_HEADER +#define _nbio_7_4_0_SMN_HEADER + + +#define smnNBIF_MGCG_CTRL_LCLK 0x1013a21c +#define smnCPM_CONTROL 0x11180460 +#define smnPCIE_CNTL2 0x11180070 +#define smnPCIE_CI_CNTL 0x11180080 + +#define smnPCIE_PERF_COUNT_CNTL 0x11180200 +#define smnPCIE_PERF_CNTL_TXCLK1 0x11180204 +#define smnPCIE_PERF_COUNT0_TXCLK1 0x11180208 +#define smnPCIE_PERF_COUNT1_TXCLK1 0x1118020c +#define smnPCIE_PERF_CNTL_TXCLK2 0x11180210 +#define smnPCIE_PERF_COUNT0_TXCLK2 0x11180214 +#define smnPCIE_PERF_COUNT1_TXCLK2 0x11180218 +#define smnPCIE_PERF_CNTL_TXCLK3 0x1118021c +#define smnPCIE_PERF_COUNT0_TXCLK3 0x11180220 +#define smnPCIE_PERF_COUNT1_TXCLK3 0x11180224 +#define smnPCIE_PERF_CNTL_TXCLK4 0x11180228 +#define smnPCIE_PERF_COUNT0_TXCLK4 0x1118022c +#define smnPCIE_PERF_COUNT1_TXCLK4 0x11180230 +#define smnPCIE_PERF_CNTL_SCLK1 0x11180234 +#define smnPCIE_PERF_COUNT0_SCLK1 0x11180238 +#define smnPCIE_PERF_COUNT1_SCLK1 0x1118023c +#define smnPCIE_PERF_CNTL_SCLK2 0x11180240 +#define smnPCIE_PERF_COUNT0_SCLK2 0x11180244 +#define smnPCIE_PERF_COUNT1_SCLK2 0x11180248 +#define smnPCIE_PERF_CNTL_EVENT_LC_PORT_SEL 0x1118024c +#define smnPCIE_PERF_CNTL_EVENT_CI_PORT_SEL 0x11180250 + +#endif // _nbio_7_4_0_SMN_HEADER diff --git a/drivers/gpu/drm/amd/include/asic_reg/nbio/nbio_7_4_offset.h b/drivers/gpu/drm/amd/include/asic_reg/nbio/nbio_7_4_offset.h index e932213f87f0..994e796a28d7 100644 --- a/drivers/gpu/drm/amd/include/asic_reg/nbio/nbio_7_4_offset.h +++ b/drivers/gpu/drm/amd/include/asic_reg/nbio/nbio_7_4_offset.h @@ -2567,6 +2567,8 @@ // addressBlock: nbio_nbif0_rcc_strap_BIFDEC1 // base address: 0x0 +#define mmRCC_BIF_STRAP0 0x0000 +#define mmRCC_BIF_STRAP0_BASE_IDX 2 #define mmRCC_DEV0_EPF0_STRAP0 0x0011 #define mmRCC_DEV0_EPF0_STRAP0_BASE_IDX 2 diff --git a/drivers/gpu/drm/amd/include/asic_reg/nbio/nbio_7_4_sh_mask.h b/drivers/gpu/drm/amd/include/asic_reg/nbio/nbio_7_4_sh_mask.h index d3704b438f2d..d467b939c971 100644 --- a/drivers/gpu/drm/amd/include/asic_reg/nbio/nbio_7_4_sh_mask.h +++ b/drivers/gpu/drm/amd/include/asic_reg/nbio/nbio_7_4_sh_mask.h @@ -19690,6 +19690,9 @@ // addressBlock: nbio_nbif0_rcc_strap_BIFDEC1 +//RCC_BIF_STRAP0 +#define RCC_BIF_STRAP0__STRAP_PX_CAPABLE__SHIFT 0x7 +#define RCC_BIF_STRAP0__STRAP_PX_CAPABLE_MASK 0x00000080L //RCC_DEV0_EPF0_STRAP0 #define RCC_DEV0_EPF0_STRAP0__STRAP_DEVICE_ID_DEV0_F0__SHIFT 0x0 #define RCC_DEV0_EPF0_STRAP0__STRAP_MAJOR_REV_ID_DEV0_F0__SHIFT 0x10 diff --git a/drivers/gpu/drm/amd/include/asic_reg/thm/thm_11_0_2_offset.h b/drivers/gpu/drm/amd/include/asic_reg/thm/thm_11_0_2_offset.h index a9eb57a53e59..a485526f3a51 100644 --- a/drivers/gpu/drm/amd/include/asic_reg/thm/thm_11_0_2_offset.h +++ b/drivers/gpu/drm/amd/include/asic_reg/thm/thm_11_0_2_offset.h @@ -46,4 +46,7 @@ #define mmTHM_TCON_THERM_TRIP 0x0002 #define mmTHM_TCON_THERM_TRIP_BASE_IDX 0 +#define mmTHM_BACO_CNTL 0x0081 +#define mmTHM_BACO_CNTL_BASE_IDX 0 + #endif diff --git a/drivers/gpu/drm/amd/include/atombios.h b/drivers/gpu/drm/amd/include/atombios.h index 7931502fa54f..8ba21747b40a 100644 --- a/drivers/gpu/drm/amd/include/atombios.h +++ b/drivers/gpu/drm/amd/include/atombios.h @@ -4106,7 +4106,7 @@ typedef struct _ATOM_LCD_MODE_CONTROL_CAP typedef struct _ATOM_FAKE_EDID_PATCH_RECORD { UCHAR ucRecordType; - UCHAR ucFakeEDIDLength; // = 128 means EDID lenght is 128 bytes, otherwise the EDID length = ucFakeEDIDLength*128 + UCHAR ucFakeEDIDLength; // = 128 means EDID length is 128 bytes, otherwise the EDID length = ucFakeEDIDLength*128 UCHAR ucFakeEDIDString[1]; // This actually has ucFakeEdidLength elements. } ATOM_FAKE_EDID_PATCH_RECORD; diff --git a/drivers/gpu/drm/amd/include/kgd_kfd_interface.h b/drivers/gpu/drm/amd/include/kgd_kfd_interface.h index 64ecffd52126..5f3c10ebff08 100644 --- a/drivers/gpu/drm/amd/include/kgd_kfd_interface.h +++ b/drivers/gpu/drm/amd/include/kgd_kfd_interface.h @@ -34,7 +34,6 @@ struct pci_dev; -#define KFD_INTERFACE_VERSION 2 #define KGD_MAX_QUEUES 128 struct kfd_dev; @@ -138,20 +137,17 @@ struct kgd2kfd_shared_resources { /* Bit n == 1 means Queue n is available for KFD */ DECLARE_BITMAP(queue_bitmap, KGD_MAX_QUEUES); - /* Doorbell assignments (SOC15 and later chips only). Only + /* SDMA doorbell assignments (SOC15 and later chips only). Only * specific doorbells are routed to each SDMA engine. Others * are routed to IH and VCN. They are not usable by the CP. - * - * Any doorbell number D that satisfies the following condition - * is reserved: (D & reserved_doorbell_mask) == reserved_doorbell_val - * - * KFD currently uses 1024 (= 0x3ff) doorbells per process. If - * doorbells 0x0e0-0x0ff and 0x2e0-0x2ff are reserved, that means - * mask would be set to 0x1e0 and val set to 0x0e0. */ - unsigned int sdma_doorbell[2][8]; - unsigned int reserved_doorbell_mask; - unsigned int reserved_doorbell_val; + uint32_t *sdma_doorbell_idx; + + /* From SOC15 onward, the doorbell index range not usable for CP + * queues. + */ + uint32_t non_cp_doorbells_start; + uint32_t non_cp_doorbells_end; /* Base address of doorbell aperture. */ phys_addr_t doorbell_physical_address; @@ -188,8 +184,8 @@ struct tile_config { */ #define ALLOC_MEM_FLAGS_VRAM (1 << 0) #define ALLOC_MEM_FLAGS_GTT (1 << 1) -#define ALLOC_MEM_FLAGS_USERPTR (1 << 2) /* TODO */ -#define ALLOC_MEM_FLAGS_DOORBELL (1 << 3) /* TODO */ +#define ALLOC_MEM_FLAGS_USERPTR (1 << 2) +#define ALLOC_MEM_FLAGS_DOORBELL (1 << 3) /* * Allocation flags attributes/access options. @@ -205,20 +201,6 @@ struct tile_config { /** * struct kfd2kgd_calls * - * @init_gtt_mem_allocation: Allocate a buffer on the gart aperture. - * The buffer can be used for mqds, hpds, kernel queue, fence and runlists - * - * @free_gtt_mem: Frees a buffer that was allocated on the gart aperture - * - * @get_local_mem_info: Retrieves information about GPU local memory - * - * @get_gpu_clock_counter: Retrieves GPU clock counter - * - * @get_max_engine_clock_in_mhz: Retrieves maximum GPU clock in MHz - * - * @alloc_pasid: Allocate a PASID - * @free_pasid: Free a PASID - * * @program_sh_mem_settings: A function that should initiate the memory * properties such as main aperture memory type (cache / non cached) and * secondary aperture base address, size and memory type. @@ -255,64 +237,16 @@ struct tile_config { * * @get_tile_config: Returns GPU-specific tiling mode information * - * @get_cu_info: Retrieves activated cu info - * - * @get_vram_usage: Returns current VRAM usage - * - * @create_process_vm: Create a VM address space for a given process and GPU - * - * @destroy_process_vm: Destroy a VM - * - * @get_process_page_dir: Get physical address of a VM page directory - * * @set_vm_context_page_table_base: Program page table base for a VMID * - * @alloc_memory_of_gpu: Allocate GPUVM memory - * - * @free_memory_of_gpu: Free GPUVM memory - * - * @map_memory_to_gpu: Map GPUVM memory into a specific VM address - * space. Allocates and updates page tables and page directories as - * needed. This function may return before all page table updates have - * completed. This allows multiple map operations (on multiple GPUs) - * to happen concurrently. Use sync_memory to synchronize with all - * pending updates. - * - * @unmap_memor_to_gpu: Unmap GPUVM memory from a specific VM address space - * - * @sync_memory: Wait for pending page table updates to complete - * - * @map_gtt_bo_to_kernel: Map a GTT BO for kernel access - * Pins the BO, maps it to kernel address space. Such BOs are never evicted. - * The kernel virtual address remains valid until the BO is freed. - * - * @restore_process_bos: Restore all BOs that belong to the - * process. This is intended for restoring memory mappings after a TTM - * eviction. - * * @invalidate_tlbs: Invalidate TLBs for a specific PASID * * @invalidate_tlbs_vmid: Invalidate TLBs for a specific VMID * - * @submit_ib: Submits an IB to the engine specified by inserting the - * IB to the corresponding ring (ring type). The IB is executed with the - * specified VMID in a user mode context. - * - * @get_vm_fault_info: Return information about a recent VM fault on - * GFXv7 and v8. If multiple VM faults occurred since the last call of - * this function, it will return information about the first of those - * faults. On GFXv9 VM fault information is fully contained in the IH - * packet and this function is not needed. - * * @read_vmid_from_vmfault_reg: On Hawaii the VMID is not set in the * IH ring entry. This function allows the KFD ISR to get the VMID * from the fault status register as early as possible. * - * @gpu_recover: let kgd reset gpu after kfd detect CPC hang - * - * @set_compute_idle: Indicates that compute is idle on a device. This - * can be used to change power profiles depending on compute activity. - * * @get_hive_id: Returns hive id of current device, 0 if xgmi is not enabled * * This structure contains function pointers to services that the kgd driver @@ -320,21 +254,6 @@ struct tile_config { * */ struct kfd2kgd_calls { - int (*init_gtt_mem_allocation)(struct kgd_dev *kgd, size_t size, - void **mem_obj, uint64_t *gpu_addr, - void **cpu_ptr, bool mqd_gfx9); - - void (*free_gtt_mem)(struct kgd_dev *kgd, void *mem_obj); - - void (*get_local_mem_info)(struct kgd_dev *kgd, - struct kfd_local_mem_info *mem_info); - uint64_t (*get_gpu_clock_counter)(struct kgd_dev *kgd); - - uint32_t (*get_max_engine_clock_in_mhz)(struct kgd_dev *kgd); - - int (*alloc_pasid)(unsigned int bits); - void (*free_pasid)(unsigned int pasid); - /* Register access functions */ void (*program_sh_mem_settings)(struct kgd_dev *kgd, uint32_t vmid, uint32_t sh_mem_config, uint32_t sh_mem_ape1_base, @@ -398,103 +317,13 @@ struct kfd2kgd_calls { uint64_t va, uint32_t vmid); int (*get_tile_config)(struct kgd_dev *kgd, struct tile_config *config); - void (*get_cu_info)(struct kgd_dev *kgd, - struct kfd_cu_info *cu_info); - uint64_t (*get_vram_usage)(struct kgd_dev *kgd); - - int (*create_process_vm)(struct kgd_dev *kgd, unsigned int pasid, void **vm, - void **process_info, struct dma_fence **ef); - int (*acquire_process_vm)(struct kgd_dev *kgd, struct file *filp, - unsigned int pasid, void **vm, void **process_info, - struct dma_fence **ef); - void (*destroy_process_vm)(struct kgd_dev *kgd, void *vm); - void (*release_process_vm)(struct kgd_dev *kgd, void *vm); - uint64_t (*get_process_page_dir)(void *vm); void (*set_vm_context_page_table_base)(struct kgd_dev *kgd, uint32_t vmid, uint64_t page_table_base); - int (*alloc_memory_of_gpu)(struct kgd_dev *kgd, uint64_t va, - uint64_t size, void *vm, - struct kgd_mem **mem, uint64_t *offset, - uint32_t flags); - int (*free_memory_of_gpu)(struct kgd_dev *kgd, struct kgd_mem *mem); - int (*map_memory_to_gpu)(struct kgd_dev *kgd, struct kgd_mem *mem, - void *vm); - int (*unmap_memory_to_gpu)(struct kgd_dev *kgd, struct kgd_mem *mem, - void *vm); - int (*sync_memory)(struct kgd_dev *kgd, struct kgd_mem *mem, bool intr); - int (*map_gtt_bo_to_kernel)(struct kgd_dev *kgd, struct kgd_mem *mem, - void **kptr, uint64_t *size); - int (*restore_process_bos)(void *process_info, struct dma_fence **ef); - int (*invalidate_tlbs)(struct kgd_dev *kgd, uint16_t pasid); int (*invalidate_tlbs_vmid)(struct kgd_dev *kgd, uint16_t vmid); - - int (*submit_ib)(struct kgd_dev *kgd, enum kgd_engine_type engine, - uint32_t vmid, uint64_t gpu_addr, - uint32_t *ib_cmd, uint32_t ib_len); - - int (*get_vm_fault_info)(struct kgd_dev *kgd, - struct kfd_vm_fault_info *info); uint32_t (*read_vmid_from_vmfault_reg)(struct kgd_dev *kgd); - - void (*gpu_recover)(struct kgd_dev *kgd); - - void (*set_compute_idle)(struct kgd_dev *kgd, bool idle); - uint64_t (*get_hive_id)(struct kgd_dev *kgd); }; -/** - * struct kgd2kfd_calls - * - * @exit: Notifies amdkfd that kgd module is unloaded - * - * @probe: Notifies amdkfd about a probe done on a device in the kgd driver. - * - * @device_init: Initialize the newly probed device (if it is a device that - * amdkfd supports) - * - * @device_exit: Notifies amdkfd about a removal of a kgd device - * - * @suspend: Notifies amdkfd about a suspend action done to a kgd device - * - * @resume: Notifies amdkfd about a resume action done to a kgd device - * - * @quiesce_mm: Quiesce all user queue access to specified MM address space - * - * @resume_mm: Resume user queue access to specified MM address space - * - * @schedule_evict_and_restore_process: Schedules work queue that will prepare - * for safe eviction of KFD BOs that belong to the specified process. - * - * @pre_reset: Notifies amdkfd that amdgpu about to reset the gpu - * - * @post_reset: Notify amdkfd that amgpu successfully reseted the gpu - * - * This structure contains function callback pointers so the kgd driver - * will notify to the amdkfd about certain status changes. - * - */ -struct kgd2kfd_calls { - void (*exit)(void); - struct kfd_dev* (*probe)(struct kgd_dev *kgd, struct pci_dev *pdev, - const struct kfd2kgd_calls *f2g); - bool (*device_init)(struct kfd_dev *kfd, - const struct kgd2kfd_shared_resources *gpu_resources); - void (*device_exit)(struct kfd_dev *kfd); - void (*interrupt)(struct kfd_dev *kfd, const void *ih_ring_entry); - void (*suspend)(struct kfd_dev *kfd); - int (*resume)(struct kfd_dev *kfd); - int (*quiesce_mm)(struct mm_struct *mm); - int (*resume_mm)(struct mm_struct *mm); - int (*schedule_evict_and_restore_process)(struct mm_struct *mm, - struct dma_fence *fence); - int (*pre_reset)(struct kfd_dev *kfd); - int (*post_reset)(struct kfd_dev *kfd); -}; - -int kgd2kfd_init(unsigned interface_version, - const struct kgd2kfd_calls **g2f); - #endif /* KGD_KFD_INTERFACE_H_INCLUDED */ diff --git a/drivers/gpu/drm/amd/include/kgd_pp_interface.h b/drivers/gpu/drm/amd/include/kgd_pp_interface.h index 980e696989b1..2b579ba9b685 100644 --- a/drivers/gpu/drm/amd/include/kgd_pp_interface.h +++ b/drivers/gpu/drm/amd/include/kgd_pp_interface.h @@ -92,6 +92,9 @@ enum pp_clock_type { PP_SCLK, PP_MCLK, PP_PCIE, + PP_SOCCLK, + PP_FCLK, + PP_DCEFCLK, OD_SCLK, OD_MCLK, OD_VDDC_CURVE, @@ -127,12 +130,13 @@ enum amd_pp_task { }; enum PP_SMC_POWER_PROFILE { - PP_SMC_POWER_PROFILE_FULLSCREEN3D = 0x0, - PP_SMC_POWER_PROFILE_POWERSAVING = 0x1, - PP_SMC_POWER_PROFILE_VIDEO = 0x2, - PP_SMC_POWER_PROFILE_VR = 0x3, - PP_SMC_POWER_PROFILE_COMPUTE = 0x4, - PP_SMC_POWER_PROFILE_CUSTOM = 0x5, + PP_SMC_POWER_PROFILE_BOOTUP_DEFAULT = 0x0, + PP_SMC_POWER_PROFILE_FULLSCREEN3D = 0x1, + PP_SMC_POWER_PROFILE_POWERSAVING = 0x2, + PP_SMC_POWER_PROFILE_VIDEO = 0x3, + PP_SMC_POWER_PROFILE_VR = 0x4, + PP_SMC_POWER_PROFILE_COMPUTE = 0x5, + PP_SMC_POWER_PROFILE_CUSTOM = 0x6, }; enum { @@ -276,6 +280,15 @@ struct amd_pm_funcs { struct amd_pp_simple_clock_info *clocks); int (*notify_smu_enable_pwe)(void *handle); int (*enable_mgpu_fan_boost)(void *handle); + int (*set_active_display_count)(void *handle, uint32_t count); + int (*set_hard_min_dcefclk_by_freq)(void *handle, uint32_t clock); + int (*set_hard_min_fclk_by_freq)(void *handle, uint32_t clock); + int (*set_min_deep_sleep_dcefclk)(void *handle, uint32_t clock); + int (*get_asic_baco_capability)(void *handle, bool *cap); + int (*get_asic_baco_state)(void *handle, int *state); + int (*set_asic_baco_state)(void *handle, int state); + int (*get_ppfeature_status)(void *handle, char *buf); + int (*set_ppfeature_status)(void *handle, uint64_t ppfeature_masks); }; #endif diff --git a/drivers/gpu/drm/amd/powerplay/amd_powerplay.c b/drivers/gpu/drm/amd/powerplay/amd_powerplay.c index d6aa1d414320..3f73f7cd18b9 100644 --- a/drivers/gpu/drm/amd/powerplay/amd_powerplay.c +++ b/drivers/gpu/drm/amd/powerplay/amd_powerplay.c @@ -300,7 +300,7 @@ static int pp_set_clockgating_by_smu(void *handle, uint32_t msg_id) return -EINVAL; if (hwmgr->hwmgr_func->update_clock_gatings == NULL) { - pr_info("%s was not implemented.\n", __func__); + pr_info_ratelimited("%s was not implemented.\n", __func__); return 0; } @@ -387,7 +387,7 @@ static uint32_t pp_dpm_get_sclk(void *handle, bool low) return 0; if (hwmgr->hwmgr_func->get_sclk == NULL) { - pr_info("%s was not implemented.\n", __func__); + pr_info_ratelimited("%s was not implemented.\n", __func__); return 0; } mutex_lock(&hwmgr->smu_lock); @@ -405,7 +405,7 @@ static uint32_t pp_dpm_get_mclk(void *handle, bool low) return 0; if (hwmgr->hwmgr_func->get_mclk == NULL) { - pr_info("%s was not implemented.\n", __func__); + pr_info_ratelimited("%s was not implemented.\n", __func__); return 0; } mutex_lock(&hwmgr->smu_lock); @@ -422,7 +422,7 @@ static void pp_dpm_powergate_vce(void *handle, bool gate) return; if (hwmgr->hwmgr_func->powergate_vce == NULL) { - pr_info("%s was not implemented.\n", __func__); + pr_info_ratelimited("%s was not implemented.\n", __func__); return; } mutex_lock(&hwmgr->smu_lock); @@ -438,7 +438,7 @@ static void pp_dpm_powergate_uvd(void *handle, bool gate) return; if (hwmgr->hwmgr_func->powergate_uvd == NULL) { - pr_info("%s was not implemented.\n", __func__); + pr_info_ratelimited("%s was not implemented.\n", __func__); return; } mutex_lock(&hwmgr->smu_lock); @@ -505,7 +505,7 @@ static void pp_dpm_set_fan_control_mode(void *handle, uint32_t mode) return; if (hwmgr->hwmgr_func->set_fan_control_mode == NULL) { - pr_info("%s was not implemented.\n", __func__); + pr_info_ratelimited("%s was not implemented.\n", __func__); return; } mutex_lock(&hwmgr->smu_lock); @@ -522,7 +522,7 @@ static uint32_t pp_dpm_get_fan_control_mode(void *handle) return 0; if (hwmgr->hwmgr_func->get_fan_control_mode == NULL) { - pr_info("%s was not implemented.\n", __func__); + pr_info_ratelimited("%s was not implemented.\n", __func__); return 0; } mutex_lock(&hwmgr->smu_lock); @@ -540,7 +540,7 @@ static int pp_dpm_set_fan_speed_percent(void *handle, uint32_t percent) return -EINVAL; if (hwmgr->hwmgr_func->set_fan_speed_percent == NULL) { - pr_info("%s was not implemented.\n", __func__); + pr_info_ratelimited("%s was not implemented.\n", __func__); return 0; } mutex_lock(&hwmgr->smu_lock); @@ -558,7 +558,7 @@ static int pp_dpm_get_fan_speed_percent(void *handle, uint32_t *speed) return -EINVAL; if (hwmgr->hwmgr_func->get_fan_speed_percent == NULL) { - pr_info("%s was not implemented.\n", __func__); + pr_info_ratelimited("%s was not implemented.\n", __func__); return 0; } @@ -594,7 +594,7 @@ static int pp_dpm_set_fan_speed_rpm(void *handle, uint32_t rpm) return -EINVAL; if (hwmgr->hwmgr_func->set_fan_speed_rpm == NULL) { - pr_info("%s was not implemented.\n", __func__); + pr_info_ratelimited("%s was not implemented.\n", __func__); return 0; } mutex_lock(&hwmgr->smu_lock); @@ -720,12 +720,12 @@ static int pp_dpm_force_clock_level(void *handle, return -EINVAL; if (hwmgr->hwmgr_func->force_clock_level == NULL) { - pr_info("%s was not implemented.\n", __func__); + pr_info_ratelimited("%s was not implemented.\n", __func__); return 0; } if (hwmgr->dpm_level != AMD_DPM_FORCED_LEVEL_MANUAL) { - pr_info("force clock level is for dpm manual mode only.\n"); + pr_debug("force clock level is for dpm manual mode only.\n"); return -EINVAL; } @@ -745,7 +745,7 @@ static int pp_dpm_print_clock_levels(void *handle, return -EINVAL; if (hwmgr->hwmgr_func->print_clock_levels == NULL) { - pr_info("%s was not implemented.\n", __func__); + pr_info_ratelimited("%s was not implemented.\n", __func__); return 0; } mutex_lock(&hwmgr->smu_lock); @@ -763,7 +763,7 @@ static int pp_dpm_get_sclk_od(void *handle) return -EINVAL; if (hwmgr->hwmgr_func->get_sclk_od == NULL) { - pr_info("%s was not implemented.\n", __func__); + pr_info_ratelimited("%s was not implemented.\n", __func__); return 0; } mutex_lock(&hwmgr->smu_lock); @@ -781,7 +781,7 @@ static int pp_dpm_set_sclk_od(void *handle, uint32_t value) return -EINVAL; if (hwmgr->hwmgr_func->set_sclk_od == NULL) { - pr_info("%s was not implemented.\n", __func__); + pr_info_ratelimited("%s was not implemented.\n", __func__); return 0; } @@ -800,7 +800,7 @@ static int pp_dpm_get_mclk_od(void *handle) return -EINVAL; if (hwmgr->hwmgr_func->get_mclk_od == NULL) { - pr_info("%s was not implemented.\n", __func__); + pr_info_ratelimited("%s was not implemented.\n", __func__); return 0; } mutex_lock(&hwmgr->smu_lock); @@ -818,7 +818,7 @@ static int pp_dpm_set_mclk_od(void *handle, uint32_t value) return -EINVAL; if (hwmgr->hwmgr_func->set_mclk_od == NULL) { - pr_info("%s was not implemented.\n", __func__); + pr_info_ratelimited("%s was not implemented.\n", __func__); return 0; } mutex_lock(&hwmgr->smu_lock); @@ -878,7 +878,7 @@ static int pp_get_power_profile_mode(void *handle, char *buf) return -EINVAL; if (hwmgr->hwmgr_func->get_power_profile_mode == NULL) { - pr_info("%s was not implemented.\n", __func__); + pr_info_ratelimited("%s was not implemented.\n", __func__); return snprintf(buf, PAGE_SIZE, "\n"); } @@ -894,12 +894,12 @@ static int pp_set_power_profile_mode(void *handle, long *input, uint32_t size) return ret; if (hwmgr->hwmgr_func->set_power_profile_mode == NULL) { - pr_info("%s was not implemented.\n", __func__); + pr_info_ratelimited("%s was not implemented.\n", __func__); return ret; } if (hwmgr->dpm_level != AMD_DPM_FORCED_LEVEL_MANUAL) { - pr_info("power profile setting is for manual dpm mode only.\n"); + pr_debug("power profile setting is for manual dpm mode only.\n"); return ret; } @@ -917,7 +917,7 @@ static int pp_odn_edit_dpm_table(void *handle, uint32_t type, long *input, uint3 return -EINVAL; if (hwmgr->hwmgr_func->odn_edit_dpm_table == NULL) { - pr_info("%s was not implemented.\n", __func__); + pr_info_ratelimited("%s was not implemented.\n", __func__); return -EINVAL; } @@ -935,7 +935,7 @@ static int pp_dpm_switch_power_profile(void *handle, return -EINVAL; if (hwmgr->hwmgr_func->set_power_profile_mode == NULL) { - pr_info("%s was not implemented.\n", __func__); + pr_info_ratelimited("%s was not implemented.\n", __func__); return -EINVAL; } @@ -972,7 +972,7 @@ static int pp_set_power_limit(void *handle, uint32_t limit) return -EINVAL; if (hwmgr->hwmgr_func->set_power_limit == NULL) { - pr_info("%s was not implemented.\n", __func__); + pr_info_ratelimited("%s was not implemented.\n", __func__); return -EINVAL; } @@ -1072,7 +1072,7 @@ static int pp_get_current_clocks(void *handle, &hw_clocks, PHM_PerformanceLevelDesignation_Activity); if (ret) { - pr_info("Error in phm_get_clock_info \n"); + pr_debug("Error in phm_get_clock_info \n"); mutex_unlock(&hwmgr->smu_lock); return -EINVAL; } @@ -1212,7 +1212,7 @@ static int pp_dpm_powergate_mmhub(void *handle) return -EINVAL; if (hwmgr->hwmgr_func->powergate_mmhub == NULL) { - pr_info("%s was not implemented.\n", __func__); + pr_info_ratelimited("%s was not implemented.\n", __func__); return 0; } @@ -1227,7 +1227,7 @@ static int pp_dpm_powergate_gfx(void *handle, bool gate) return 0; if (hwmgr->hwmgr_func->powergate_gfx == NULL) { - pr_info("%s was not implemented.\n", __func__); + pr_info_ratelimited("%s was not implemented.\n", __func__); return 0; } @@ -1242,7 +1242,7 @@ static void pp_dpm_powergate_acp(void *handle, bool gate) return; if (hwmgr->hwmgr_func->powergate_acp == NULL) { - pr_info("%s was not implemented.\n", __func__); + pr_info_ratelimited("%s was not implemented.\n", __func__); return; } @@ -1257,7 +1257,7 @@ static void pp_dpm_powergate_sdma(void *handle, bool gate) return; if (hwmgr->hwmgr_func->powergate_sdma == NULL) { - pr_info("%s was not implemented.\n", __func__); + pr_info_ratelimited("%s was not implemented.\n", __func__); return; } @@ -1303,7 +1303,7 @@ static int pp_notify_smu_enable_pwe(void *handle) return -EINVAL; if (hwmgr->hwmgr_func->smus_notify_pwe == NULL) { - pr_info("%s was not implemented.\n", __func__); + pr_info_ratelimited("%s was not implemented.\n", __func__); return -EINVAL;; } @@ -1332,6 +1332,169 @@ static int pp_enable_mgpu_fan_boost(void *handle) return 0; } +static int pp_set_min_deep_sleep_dcefclk(void *handle, uint32_t clock) +{ + struct pp_hwmgr *hwmgr = handle; + + if (!hwmgr || !hwmgr->pm_en) + return -EINVAL; + + if (hwmgr->hwmgr_func->set_min_deep_sleep_dcefclk == NULL) { + pr_debug("%s was not implemented.\n", __func__); + return -EINVAL;; + } + + mutex_lock(&hwmgr->smu_lock); + hwmgr->hwmgr_func->set_min_deep_sleep_dcefclk(hwmgr, clock); + mutex_unlock(&hwmgr->smu_lock); + + return 0; +} + +static int pp_set_hard_min_dcefclk_by_freq(void *handle, uint32_t clock) +{ + struct pp_hwmgr *hwmgr = handle; + + if (!hwmgr || !hwmgr->pm_en) + return -EINVAL; + + if (hwmgr->hwmgr_func->set_hard_min_dcefclk_by_freq == NULL) { + pr_debug("%s was not implemented.\n", __func__); + return -EINVAL;; + } + + mutex_lock(&hwmgr->smu_lock); + hwmgr->hwmgr_func->set_hard_min_dcefclk_by_freq(hwmgr, clock); + mutex_unlock(&hwmgr->smu_lock); + + return 0; +} + +static int pp_set_hard_min_fclk_by_freq(void *handle, uint32_t clock) +{ + struct pp_hwmgr *hwmgr = handle; + + if (!hwmgr || !hwmgr->pm_en) + return -EINVAL; + + if (hwmgr->hwmgr_func->set_hard_min_fclk_by_freq == NULL) { + pr_debug("%s was not implemented.\n", __func__); + return -EINVAL;; + } + + mutex_lock(&hwmgr->smu_lock); + hwmgr->hwmgr_func->set_hard_min_fclk_by_freq(hwmgr, clock); + mutex_unlock(&hwmgr->smu_lock); + + return 0; +} + +static int pp_set_active_display_count(void *handle, uint32_t count) +{ + struct pp_hwmgr *hwmgr = handle; + int ret = 0; + + if (!hwmgr || !hwmgr->pm_en) + return -EINVAL; + + mutex_lock(&hwmgr->smu_lock); + ret = phm_set_active_display_count(hwmgr, count); + mutex_unlock(&hwmgr->smu_lock); + + return ret; +} + +static int pp_get_asic_baco_capability(void *handle, bool *cap) +{ + struct pp_hwmgr *hwmgr = handle; + + if (!hwmgr) + return -EINVAL; + + if (!hwmgr->pm_en || !hwmgr->hwmgr_func->get_asic_baco_capability) + return 0; + + mutex_lock(&hwmgr->smu_lock); + hwmgr->hwmgr_func->get_asic_baco_capability(hwmgr, cap); + mutex_unlock(&hwmgr->smu_lock); + + return 0; +} + +static int pp_get_asic_baco_state(void *handle, int *state) +{ + struct pp_hwmgr *hwmgr = handle; + + if (!hwmgr) + return -EINVAL; + + if (!hwmgr->pm_en || !hwmgr->hwmgr_func->get_asic_baco_state) + return 0; + + mutex_lock(&hwmgr->smu_lock); + hwmgr->hwmgr_func->get_asic_baco_state(hwmgr, (enum BACO_STATE *)state); + mutex_unlock(&hwmgr->smu_lock); + + return 0; +} + +static int pp_set_asic_baco_state(void *handle, int state) +{ + struct pp_hwmgr *hwmgr = handle; + + if (!hwmgr) + return -EINVAL; + + if (!hwmgr->pm_en || !hwmgr->hwmgr_func->set_asic_baco_state) + return 0; + + mutex_lock(&hwmgr->smu_lock); + hwmgr->hwmgr_func->set_asic_baco_state(hwmgr, (enum BACO_STATE)state); + mutex_unlock(&hwmgr->smu_lock); + + return 0; +} + +static int pp_get_ppfeature_status(void *handle, char *buf) +{ + struct pp_hwmgr *hwmgr = handle; + int ret = 0; + + if (!hwmgr || !hwmgr->pm_en || !buf) + return -EINVAL; + + if (hwmgr->hwmgr_func->get_ppfeature_status == NULL) { + pr_info_ratelimited("%s was not implemented.\n", __func__); + return -EINVAL; + } + + mutex_lock(&hwmgr->smu_lock); + ret = hwmgr->hwmgr_func->get_ppfeature_status(hwmgr, buf); + mutex_unlock(&hwmgr->smu_lock); + + return ret; +} + +static int pp_set_ppfeature_status(void *handle, uint64_t ppfeature_masks) +{ + struct pp_hwmgr *hwmgr = handle; + int ret = 0; + + if (!hwmgr || !hwmgr->pm_en) + return -EINVAL; + + if (hwmgr->hwmgr_func->set_ppfeature_status == NULL) { + pr_info_ratelimited("%s was not implemented.\n", __func__); + return -EINVAL; + } + + mutex_lock(&hwmgr->smu_lock); + ret = hwmgr->hwmgr_func->set_ppfeature_status(hwmgr, ppfeature_masks); + mutex_unlock(&hwmgr->smu_lock); + + return ret; +} + static const struct amd_pm_funcs pp_dpm_funcs = { .load_firmware = pp_dpm_load_fw, .wait_for_fw_loading_complete = pp_dpm_fw_loading_complete, @@ -1378,4 +1541,13 @@ static const struct amd_pm_funcs pp_dpm_funcs = { .get_display_mode_validation_clocks = pp_get_display_mode_validation_clocks, .notify_smu_enable_pwe = pp_notify_smu_enable_pwe, .enable_mgpu_fan_boost = pp_enable_mgpu_fan_boost, + .set_active_display_count = pp_set_active_display_count, + .set_min_deep_sleep_dcefclk = pp_set_min_deep_sleep_dcefclk, + .set_hard_min_dcefclk_by_freq = pp_set_hard_min_dcefclk_by_freq, + .set_hard_min_fclk_by_freq = pp_set_hard_min_fclk_by_freq, + .get_asic_baco_capability = pp_get_asic_baco_capability, + .get_asic_baco_state = pp_get_asic_baco_state, + .set_asic_baco_state = pp_set_asic_baco_state, + .get_ppfeature_status = pp_get_ppfeature_status, + .set_ppfeature_status = pp_set_ppfeature_status, }; diff --git a/drivers/gpu/drm/amd/powerplay/hwmgr/Makefile b/drivers/gpu/drm/amd/powerplay/hwmgr/Makefile index ade8973b6f4d..0b3c6d1d52e4 100644 --- a/drivers/gpu/drm/amd/powerplay/hwmgr/Makefile +++ b/drivers/gpu/drm/amd/powerplay/hwmgr/Makefile @@ -35,7 +35,7 @@ HARDWARE_MGR = hwmgr.o processpptables.o \ vega12_thermal.o \ pp_overdriver.o smu_helper.o \ vega20_processpptables.o vega20_hwmgr.o vega20_powertune.o \ - vega20_thermal.o + vega20_thermal.o common_baco.o vega10_baco.o vega20_baco.o AMD_PP_HWMGR = $(addprefix $(AMD_PP_PATH)/hwmgr/,$(HARDWARE_MGR)) diff --git a/drivers/gpu/drm/amd/powerplay/hwmgr/common_baco.c b/drivers/gpu/drm/amd/powerplay/hwmgr/common_baco.c new file mode 100644 index 000000000000..9c57c1f67749 --- /dev/null +++ b/drivers/gpu/drm/amd/powerplay/hwmgr/common_baco.c @@ -0,0 +1,101 @@ +/* + * Copyright 2018 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 "common_baco.h" + + +static bool baco_wait_register(struct pp_hwmgr *hwmgr, u32 reg, u32 mask, u32 value) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)(hwmgr->adev); + u32 timeout = 5000, data; + + do { + msleep(1); + data = RREG32(reg); + timeout--; + } while (value != (data & mask) && (timeout != 0)); + + if (timeout == 0) + return false; + + return true; +} + +static bool baco_cmd_handler(struct pp_hwmgr *hwmgr, u32 command, u32 reg, u32 mask, + u32 shift, u32 value, u32 timeout) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)(hwmgr->adev); + u32 data; + bool ret = true; + + switch (command) { + case CMD_WRITE: + WREG32(reg, value << shift); + break; + case CMD_READMODIFYWRITE: + data = RREG32(reg); + data = (data & (~mask)) | (value << shift); + WREG32(reg, data); + break; + case CMD_WAITFOR: + ret = baco_wait_register(hwmgr, reg, mask, value); + break; + case CMD_DELAY_MS: + if (timeout) + /* Delay in milli Seconds */ + msleep(timeout); + break; + case CMD_DELAY_US: + if (timeout) + /* Delay in micro Seconds */ + udelay(timeout); + break; + + default: + dev_warn(adev->dev, "Invalid BACO command.\n"); + ret = false; + } + + return ret; +} + +bool soc15_baco_program_registers(struct pp_hwmgr *hwmgr, + const struct soc15_baco_cmd_entry *entry, + const u32 array_size) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)(hwmgr->adev); + u32 i, reg = 0; + + for (i = 0; i < array_size; i++) { + if ((entry[i].cmd == CMD_WRITE) || + (entry[i].cmd == CMD_READMODIFYWRITE) || + (entry[i].cmd == CMD_WAITFOR)) + reg = adev->reg_offset[entry[i].hwip][entry[i].inst][entry[i].seg] + + entry[i].reg_offset; + if (!baco_cmd_handler(hwmgr, entry[i].cmd, reg, entry[i].mask, + entry[i].shift, entry[i].val, entry[i].timeout)) + return false; + } + + return true; +} diff --git a/drivers/gpu/drm/amd/display/dc/i2caux/dce110/i2c_sw_engine_dce110.h b/drivers/gpu/drm/amd/powerplay/hwmgr/common_baco.h index c48c61f540a8..95296c916f4e 100644 --- a/drivers/gpu/drm/amd/display/dc/i2caux/dce110/i2c_sw_engine_dce110.h +++ b/drivers/gpu/drm/amd/powerplay/hwmgr/common_baco.h @@ -1,5 +1,5 @@ /* - * Copyright 2012-15 Advanced Micro Devices, Inc. + * Copyright 2018 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"), @@ -19,25 +19,32 @@ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR * OTHER DEALINGS IN THE SOFTWARE. * - * Authors: AMD - * */ +#ifndef __COMMON_BOCO_H__ +#define __COMMON_BOCO_H__ +#include "hwmgr.h" -#ifndef __DAL_I2C_SW_ENGINE_DCE110_H__ -#define __DAL_I2C_SW_ENGINE_DCE110_H__ -struct i2c_sw_engine_dce110 { - struct i2c_sw_engine base; - uint32_t engine_id; +enum baco_cmd_type { + CMD_WRITE = 0, + CMD_READMODIFYWRITE, + CMD_WAITFOR, + CMD_DELAY_MS, + CMD_DELAY_US, }; -struct i2c_sw_engine_dce110_create_arg { - uint32_t engine_id; - uint32_t default_speed; - struct dc_context *ctx; +struct soc15_baco_cmd_entry { + enum baco_cmd_type cmd; + uint32_t hwip; + uint32_t inst; + uint32_t seg; + uint32_t reg_offset; + uint32_t mask; + uint32_t shift; + uint32_t timeout; + uint32_t val; }; - -struct i2c_engine *dal_i2c_sw_engine_dce110_create( - const struct i2c_sw_engine_dce110_create_arg *arg); - +extern bool soc15_baco_program_registers(struct pp_hwmgr *hwmgr, + const struct soc15_baco_cmd_entry *entry, + const u32 array_size); #endif diff --git a/drivers/gpu/drm/amd/powerplay/hwmgr/hardwaremanager.c b/drivers/gpu/drm/amd/powerplay/hwmgr/hardwaremanager.c index a2a7e0e94aa6..c1c51c115e57 100644 --- a/drivers/gpu/drm/amd/powerplay/hwmgr/hardwaremanager.c +++ b/drivers/gpu/drm/amd/powerplay/hwmgr/hardwaremanager.c @@ -154,15 +154,6 @@ int phm_powerdown_uvd(struct pp_hwmgr *hwmgr) return 0; } -int phm_enable_clock_power_gatings(struct pp_hwmgr *hwmgr) -{ - PHM_FUNC_CHECK(hwmgr); - - if (NULL != hwmgr->hwmgr_func->enable_clock_power_gating) - return hwmgr->hwmgr_func->enable_clock_power_gating(hwmgr); - - return 0; -} int phm_disable_clock_power_gatings(struct pp_hwmgr *hwmgr) { @@ -288,8 +279,8 @@ int phm_store_dal_configuration_data(struct pp_hwmgr *hwmgr, if (display_config == NULL) return -EINVAL; - if (NULL != hwmgr->hwmgr_func->set_deep_sleep_dcefclk) - hwmgr->hwmgr_func->set_deep_sleep_dcefclk(hwmgr, display_config->min_dcef_deep_sleep_set_clk); + if (NULL != hwmgr->hwmgr_func->set_min_deep_sleep_dcefclk) + hwmgr->hwmgr_func->set_min_deep_sleep_dcefclk(hwmgr, display_config->min_dcef_deep_sleep_set_clk); for (index = 0; index < display_config->num_path_including_non_display; index++) { if (display_config->displays[index].controller_id != 0) @@ -480,3 +471,44 @@ int phm_disable_smc_firmware_ctf(struct pp_hwmgr *hwmgr) return hwmgr->hwmgr_func->disable_smc_firmware_ctf(hwmgr); } + +int phm_set_active_display_count(struct pp_hwmgr *hwmgr, uint32_t count) +{ + PHM_FUNC_CHECK(hwmgr); + + if (!hwmgr->hwmgr_func->set_active_display_count) + return -EINVAL; + + return hwmgr->hwmgr_func->set_active_display_count(hwmgr, count); +} + +int phm_set_min_deep_sleep_dcefclk(struct pp_hwmgr *hwmgr, uint32_t clock) +{ + PHM_FUNC_CHECK(hwmgr); + + if (!hwmgr->hwmgr_func->set_min_deep_sleep_dcefclk) + return -EINVAL; + + return hwmgr->hwmgr_func->set_min_deep_sleep_dcefclk(hwmgr, clock); +} + +int phm_set_hard_min_dcefclk_by_freq(struct pp_hwmgr *hwmgr, uint32_t clock) +{ + PHM_FUNC_CHECK(hwmgr); + + if (!hwmgr->hwmgr_func->set_hard_min_dcefclk_by_freq) + return -EINVAL; + + return hwmgr->hwmgr_func->set_hard_min_dcefclk_by_freq(hwmgr, clock); +} + +int phm_set_hard_min_fclk_by_freq(struct pp_hwmgr *hwmgr, uint32_t clock) +{ + PHM_FUNC_CHECK(hwmgr); + + if (!hwmgr->hwmgr_func->set_hard_min_fclk_by_freq) + return -EINVAL; + + return hwmgr->hwmgr_func->set_hard_min_fclk_by_freq(hwmgr, clock); +} + diff --git a/drivers/gpu/drm/amd/powerplay/hwmgr/hwmgr.c b/drivers/gpu/drm/amd/powerplay/hwmgr/hwmgr.c index 0173d0480024..6cd6497c6fc2 100644 --- a/drivers/gpu/drm/amd/powerplay/hwmgr/hwmgr.c +++ b/drivers/gpu/drm/amd/powerplay/hwmgr/hwmgr.c @@ -64,17 +64,19 @@ static int ci_set_asic_special_caps(struct pp_hwmgr *hwmgr); static void hwmgr_init_workload_prority(struct pp_hwmgr *hwmgr) { - hwmgr->workload_prority[PP_SMC_POWER_PROFILE_FULLSCREEN3D] = 2; - hwmgr->workload_prority[PP_SMC_POWER_PROFILE_POWERSAVING] = 0; - hwmgr->workload_prority[PP_SMC_POWER_PROFILE_VIDEO] = 1; - hwmgr->workload_prority[PP_SMC_POWER_PROFILE_VR] = 3; - hwmgr->workload_prority[PP_SMC_POWER_PROFILE_COMPUTE] = 4; - - hwmgr->workload_setting[0] = PP_SMC_POWER_PROFILE_POWERSAVING; - hwmgr->workload_setting[1] = PP_SMC_POWER_PROFILE_VIDEO; - hwmgr->workload_setting[2] = PP_SMC_POWER_PROFILE_FULLSCREEN3D; - hwmgr->workload_setting[3] = PP_SMC_POWER_PROFILE_VR; - hwmgr->workload_setting[4] = PP_SMC_POWER_PROFILE_COMPUTE; + hwmgr->workload_prority[PP_SMC_POWER_PROFILE_BOOTUP_DEFAULT] = 0; + hwmgr->workload_prority[PP_SMC_POWER_PROFILE_FULLSCREEN3D] = 1; + hwmgr->workload_prority[PP_SMC_POWER_PROFILE_POWERSAVING] = 2; + hwmgr->workload_prority[PP_SMC_POWER_PROFILE_VIDEO] = 3; + hwmgr->workload_prority[PP_SMC_POWER_PROFILE_VR] = 4; + hwmgr->workload_prority[PP_SMC_POWER_PROFILE_COMPUTE] = 5; + + hwmgr->workload_setting[0] = PP_SMC_POWER_PROFILE_BOOTUP_DEFAULT; + hwmgr->workload_setting[1] = PP_SMC_POWER_PROFILE_FULLSCREEN3D; + hwmgr->workload_setting[2] = PP_SMC_POWER_PROFILE_POWERSAVING; + hwmgr->workload_setting[3] = PP_SMC_POWER_PROFILE_VIDEO; + hwmgr->workload_setting[4] = PP_SMC_POWER_PROFILE_VR; + hwmgr->workload_setting[5] = PP_SMC_POWER_PROFILE_COMPUTE; } int hwmgr_early_init(struct pp_hwmgr *hwmgr) @@ -271,7 +273,7 @@ int hwmgr_hw_fini(struct pp_hwmgr *hwmgr) phm_stop_thermal_controller(hwmgr); psm_set_boot_states(hwmgr); - psm_adjust_power_state_dynamic(hwmgr, false, NULL); + psm_adjust_power_state_dynamic(hwmgr, true, NULL); phm_disable_dynamic_state_management(hwmgr); phm_disable_clock_power_gatings(hwmgr); @@ -293,7 +295,7 @@ int hwmgr_suspend(struct pp_hwmgr *hwmgr) ret = psm_set_boot_states(hwmgr); if (ret) return ret; - ret = psm_adjust_power_state_dynamic(hwmgr, false, NULL); + ret = psm_adjust_power_state_dynamic(hwmgr, true, NULL); if (ret) return ret; ret = phm_power_down_asic(hwmgr); @@ -323,7 +325,7 @@ int hwmgr_resume(struct pp_hwmgr *hwmgr) if (ret) return ret; - ret = psm_adjust_power_state_dynamic(hwmgr, false, NULL); + ret = psm_adjust_power_state_dynamic(hwmgr, true, NULL); return ret; } @@ -377,12 +379,12 @@ int hwmgr_handle_task(struct pp_hwmgr *hwmgr, enum amd_pp_task task_id, ret = psm_set_user_performance_state(hwmgr, requested_ui_label, &requested_ps); if (ret) return ret; - ret = psm_adjust_power_state_dynamic(hwmgr, false, requested_ps); + ret = psm_adjust_power_state_dynamic(hwmgr, true, requested_ps); break; } case AMD_PP_TASK_COMPLETE_INIT: case AMD_PP_TASK_READJUST_POWER_STATE: - ret = psm_adjust_power_state_dynamic(hwmgr, false, NULL); + ret = psm_adjust_power_state_dynamic(hwmgr, true, NULL); break; default: break; diff --git a/drivers/gpu/drm/amd/powerplay/hwmgr/pp_psm.c b/drivers/gpu/drm/amd/powerplay/hwmgr/pp_psm.c index 56437866d120..6bf48934fdc4 100644 --- a/drivers/gpu/drm/amd/powerplay/hwmgr/pp_psm.c +++ b/drivers/gpu/drm/amd/powerplay/hwmgr/pp_psm.c @@ -256,16 +256,14 @@ static void power_state_management(struct pp_hwmgr *hwmgr, } } -int psm_adjust_power_state_dynamic(struct pp_hwmgr *hwmgr, bool skip, +int psm_adjust_power_state_dynamic(struct pp_hwmgr *hwmgr, bool skip_display_settings, struct pp_power_state *new_ps) { uint32_t index; long workload; - if (skip) - return 0; - - phm_display_configuration_changed(hwmgr); + if (!skip_display_settings) + phm_display_configuration_changed(hwmgr); if (hwmgr->ps) power_state_management(hwmgr, new_ps); @@ -276,7 +274,8 @@ int psm_adjust_power_state_dynamic(struct pp_hwmgr *hwmgr, bool skip, */ phm_apply_clock_adjust_rules(hwmgr); - phm_notify_smc_display_config_after_ps_adjustment(hwmgr); + if (!skip_display_settings) + phm_notify_smc_display_config_after_ps_adjustment(hwmgr); if (!phm_force_dpm_levels(hwmgr, hwmgr->request_dpm_level)) hwmgr->dpm_level = hwmgr->request_dpm_level; diff --git a/drivers/gpu/drm/amd/powerplay/hwmgr/pp_psm.h b/drivers/gpu/drm/amd/powerplay/hwmgr/pp_psm.h index fa1b6825036a..b62d55f1f289 100644 --- a/drivers/gpu/drm/amd/powerplay/hwmgr/pp_psm.h +++ b/drivers/gpu/drm/amd/powerplay/hwmgr/pp_psm.h @@ -34,7 +34,7 @@ int psm_set_user_performance_state(struct pp_hwmgr *hwmgr, enum PP_StateUILabel label_id, struct pp_power_state **state); int psm_adjust_power_state_dynamic(struct pp_hwmgr *hwmgr, - bool skip, + bool skip_display_settings, struct pp_power_state *new_ps); #endif diff --git a/drivers/gpu/drm/amd/powerplay/hwmgr/ppatomfwctrl.c b/drivers/gpu/drm/amd/powerplay/hwmgr/ppatomfwctrl.c index 4588bddf8b33..615cf2c09e54 100644 --- a/drivers/gpu/drm/amd/powerplay/hwmgr/ppatomfwctrl.c +++ b/drivers/gpu/drm/amd/powerplay/hwmgr/ppatomfwctrl.c @@ -489,15 +489,16 @@ int pp_atomfwctrl_get_gpio_information(struct pp_hwmgr *hwmgr, } int pp_atomfwctrl_get_clk_information_by_clkid(struct pp_hwmgr *hwmgr, - uint8_t id, uint32_t *frequency) + uint8_t clk_id, uint8_t syspll_id, + uint32_t *frequency) { struct amdgpu_device *adev = hwmgr->adev; struct atom_get_smu_clock_info_parameters_v3_1 parameters; struct atom_get_smu_clock_info_output_parameters_v3_1 *output; uint32_t ix; - parameters.clk_id = id; - parameters.syspll_id = 0; + parameters.clk_id = clk_id; + parameters.syspll_id = syspll_id; parameters.command = GET_SMU_CLOCK_INFO_V3_1_GET_CLOCK_FREQ; parameters.dfsdid = 0; @@ -530,20 +531,23 @@ static void pp_atomfwctrl_copy_vbios_bootup_values_3_2(struct pp_hwmgr *hwmgr, boot_values->ulSocClk = 0; boot_values->ulDCEFClk = 0; - if (!pp_atomfwctrl_get_clk_information_by_clkid(hwmgr, SMU11_SYSPLL0_SOCCLK_ID, &frequency)) + if (!pp_atomfwctrl_get_clk_information_by_clkid(hwmgr, SMU11_SYSPLL0_SOCCLK_ID, SMU11_SYSPLL0_ID, &frequency)) boot_values->ulSocClk = frequency; - if (!pp_atomfwctrl_get_clk_information_by_clkid(hwmgr, SMU11_SYSPLL0_DCEFCLK_ID, &frequency)) + if (!pp_atomfwctrl_get_clk_information_by_clkid(hwmgr, SMU11_SYSPLL0_DCEFCLK_ID, SMU11_SYSPLL0_ID, &frequency)) boot_values->ulDCEFClk = frequency; - if (!pp_atomfwctrl_get_clk_information_by_clkid(hwmgr, SMU11_SYSPLL0_ECLK_ID, &frequency)) + if (!pp_atomfwctrl_get_clk_information_by_clkid(hwmgr, SMU11_SYSPLL0_ECLK_ID, SMU11_SYSPLL0_ID, &frequency)) boot_values->ulEClk = frequency; - if (!pp_atomfwctrl_get_clk_information_by_clkid(hwmgr, SMU11_SYSPLL0_VCLK_ID, &frequency)) + if (!pp_atomfwctrl_get_clk_information_by_clkid(hwmgr, SMU11_SYSPLL0_VCLK_ID, SMU11_SYSPLL0_ID, &frequency)) boot_values->ulVClk = frequency; - if (!pp_atomfwctrl_get_clk_information_by_clkid(hwmgr, SMU11_SYSPLL0_DCLK_ID, &frequency)) + if (!pp_atomfwctrl_get_clk_information_by_clkid(hwmgr, SMU11_SYSPLL0_DCLK_ID, SMU11_SYSPLL0_ID, &frequency)) boot_values->ulDClk = frequency; + + if (!pp_atomfwctrl_get_clk_information_by_clkid(hwmgr, SMU11_SYSPLL1_0_FCLK_ID, SMU11_SYSPLL1_2_ID, &frequency)) + boot_values->ulFClk = frequency; } static void pp_atomfwctrl_copy_vbios_bootup_values_3_1(struct pp_hwmgr *hwmgr, @@ -563,19 +567,19 @@ static void pp_atomfwctrl_copy_vbios_bootup_values_3_1(struct pp_hwmgr *hwmgr, boot_values->ulSocClk = 0; boot_values->ulDCEFClk = 0; - if (!pp_atomfwctrl_get_clk_information_by_clkid(hwmgr, SMU9_SYSPLL0_SOCCLK_ID, &frequency)) + if (!pp_atomfwctrl_get_clk_information_by_clkid(hwmgr, SMU9_SYSPLL0_SOCCLK_ID, 0, &frequency)) boot_values->ulSocClk = frequency; - if (!pp_atomfwctrl_get_clk_information_by_clkid(hwmgr, SMU9_SYSPLL0_DCEFCLK_ID, &frequency)) + if (!pp_atomfwctrl_get_clk_information_by_clkid(hwmgr, SMU9_SYSPLL0_DCEFCLK_ID, 0, &frequency)) boot_values->ulDCEFClk = frequency; - if (!pp_atomfwctrl_get_clk_information_by_clkid(hwmgr, SMU9_SYSPLL0_ECLK_ID, &frequency)) + if (!pp_atomfwctrl_get_clk_information_by_clkid(hwmgr, SMU9_SYSPLL0_ECLK_ID, 0, &frequency)) boot_values->ulEClk = frequency; - if (!pp_atomfwctrl_get_clk_information_by_clkid(hwmgr, SMU9_SYSPLL0_VCLK_ID, &frequency)) + if (!pp_atomfwctrl_get_clk_information_by_clkid(hwmgr, SMU9_SYSPLL0_VCLK_ID, 0, &frequency)) boot_values->ulVClk = frequency; - if (!pp_atomfwctrl_get_clk_information_by_clkid(hwmgr, SMU9_SYSPLL0_DCLK_ID, &frequency)) + if (!pp_atomfwctrl_get_clk_information_by_clkid(hwmgr, SMU9_SYSPLL0_DCLK_ID, 0, &frequency)) boot_values->ulDClk = frequency; } diff --git a/drivers/gpu/drm/amd/powerplay/hwmgr/ppatomfwctrl.h b/drivers/gpu/drm/amd/powerplay/hwmgr/ppatomfwctrl.h index fe9e8ceef50e..b7e2651b570b 100644 --- a/drivers/gpu/drm/amd/powerplay/hwmgr/ppatomfwctrl.h +++ b/drivers/gpu/drm/amd/powerplay/hwmgr/ppatomfwctrl.h @@ -139,6 +139,7 @@ struct pp_atomfwctrl_bios_boot_up_values { uint32_t ulEClk; uint32_t ulVClk; uint32_t ulDClk; + uint32_t ulFClk; uint16_t usVddc; uint16_t usVddci; uint16_t usMvddc; @@ -236,7 +237,8 @@ int pp_atomfwctrl_get_vbios_bootup_values(struct pp_hwmgr *hwmgr, int pp_atomfwctrl_get_smc_dpm_information(struct pp_hwmgr *hwmgr, struct pp_atomfwctrl_smc_dpm_parameters *param); int pp_atomfwctrl_get_clk_information_by_clkid(struct pp_hwmgr *hwmgr, - uint8_t id, uint32_t *frequency); + uint8_t clk_id, uint8_t syspll_id, + uint32_t *frequency); #endif diff --git a/drivers/gpu/drm/amd/powerplay/hwmgr/smu10_hwmgr.c b/drivers/gpu/drm/amd/powerplay/hwmgr/smu10_hwmgr.c index dd18cb710391..0ad8fe4a6277 100644 --- a/drivers/gpu/drm/amd/powerplay/hwmgr/smu10_hwmgr.c +++ b/drivers/gpu/drm/amd/powerplay/hwmgr/smu10_hwmgr.c @@ -139,12 +139,10 @@ static int smu10_construct_max_power_limits_table(struct pp_hwmgr *hwmgr, static int smu10_init_dynamic_state_adjustment_rule_settings( struct pp_hwmgr *hwmgr) { - uint32_t table_size = - sizeof(struct phm_clock_voltage_dependency_table) + - (7 * sizeof(struct phm_clock_voltage_dependency_record)); + struct phm_clock_voltage_dependency_table *table_clk_vlt; - struct phm_clock_voltage_dependency_table *table_clk_vlt = - kzalloc(table_size, GFP_KERNEL); + table_clk_vlt = kzalloc(struct_size(table_clk_vlt, entries, 7), + GFP_KERNEL); if (NULL == table_clk_vlt) { pr_err("Can not allocate memory!\n"); @@ -216,12 +214,12 @@ static inline uint32_t convert_10k_to_mhz(uint32_t clock) return (clock + 99) / 100; } -static int smu10_set_deep_sleep_dcefclk(struct pp_hwmgr *hwmgr, uint32_t clock) +static int smu10_set_min_deep_sleep_dcefclk(struct pp_hwmgr *hwmgr, uint32_t clock) { struct smu10_hwmgr *smu10_data = (struct smu10_hwmgr *)(hwmgr->backend); if (smu10_data->need_min_deep_sleep_dcefclk && - smu10_data->deep_sleep_dcefclk != convert_10k_to_mhz(clock)) { + smu10_data->deep_sleep_dcefclk != convert_10k_to_mhz(clock)) { smu10_data->deep_sleep_dcefclk = convert_10k_to_mhz(clock); smum_send_msg_to_smc_with_parameter(hwmgr, PPSMC_MSG_SetMinDeepSleepDcefclk, @@ -230,6 +228,34 @@ static int smu10_set_deep_sleep_dcefclk(struct pp_hwmgr *hwmgr, uint32_t clock) return 0; } +static int smu10_set_hard_min_dcefclk_by_freq(struct pp_hwmgr *hwmgr, uint32_t clock) +{ + struct smu10_hwmgr *smu10_data = (struct smu10_hwmgr *)(hwmgr->backend); + + if (smu10_data->dcf_actual_hard_min_freq && + smu10_data->dcf_actual_hard_min_freq != convert_10k_to_mhz(clock)) { + smu10_data->dcf_actual_hard_min_freq = convert_10k_to_mhz(clock); + smum_send_msg_to_smc_with_parameter(hwmgr, + PPSMC_MSG_SetHardMinDcefclkByFreq, + smu10_data->dcf_actual_hard_min_freq); + } + return 0; +} + +static int smu10_set_hard_min_fclk_by_freq(struct pp_hwmgr *hwmgr, uint32_t clock) +{ + struct smu10_hwmgr *smu10_data = (struct smu10_hwmgr *)(hwmgr->backend); + + if (smu10_data->f_actual_hard_min_freq && + smu10_data->f_actual_hard_min_freq != convert_10k_to_mhz(clock)) { + smu10_data->f_actual_hard_min_freq = convert_10k_to_mhz(clock); + smum_send_msg_to_smc_with_parameter(hwmgr, + PPSMC_MSG_SetHardMinFclkByFreq, + smu10_data->f_actual_hard_min_freq); + } + return 0; +} + static int smu10_set_active_display_count(struct pp_hwmgr *hwmgr, uint32_t count) { struct smu10_hwmgr *smu10_data = (struct smu10_hwmgr *)(hwmgr->backend); @@ -1005,6 +1031,7 @@ static int smu10_get_clock_by_type_with_latency(struct pp_hwmgr *hwmgr, break; case amd_pp_dpp_clock: pclk_vol_table = pinfo->vdd_dep_on_dppclk; + break; default: return -EINVAL; } @@ -1206,7 +1233,7 @@ static const struct pp_hwmgr_func smu10_hwmgr_funcs = { .get_max_high_clocks = smu10_get_max_high_clocks, .read_sensor = smu10_read_sensor, .set_active_display_count = smu10_set_active_display_count, - .set_deep_sleep_dcefclk = smu10_set_deep_sleep_dcefclk, + .set_min_deep_sleep_dcefclk = smu10_set_min_deep_sleep_dcefclk, .dynamic_state_management_enable = smu10_enable_dpm_tasks, .power_off_asic = smu10_power_off_asic, .asic_setup = smu10_setup_asic_task, @@ -1217,6 +1244,8 @@ static const struct pp_hwmgr_func smu10_hwmgr_funcs = { .display_clock_voltage_request = smu10_display_clock_voltage_request, .powergate_gfx = smu10_gfx_off_control, .powergate_sdma = smu10_powergate_sdma, + .set_hard_min_dcefclk_by_freq = smu10_set_hard_min_dcefclk_by_freq, + .set_hard_min_fclk_by_freq = smu10_set_hard_min_fclk_by_freq, }; int smu10_init_function_pointers(struct pp_hwmgr *hwmgr) diff --git a/drivers/gpu/drm/amd/powerplay/hwmgr/smu7_hwmgr.c b/drivers/gpu/drm/amd/powerplay/hwmgr/smu7_hwmgr.c index b61a01f55284..83d3d935f3ac 100644 --- a/drivers/gpu/drm/amd/powerplay/hwmgr/smu7_hwmgr.c +++ b/drivers/gpu/drm/amd/powerplay/hwmgr/smu7_hwmgr.c @@ -77,8 +77,9 @@ #define PCIE_BUS_CLK 10000 #define TCLK (PCIE_BUS_CLK / 10) -static const struct profile_mode_setting smu7_profiling[6] = - {{1, 0, 100, 30, 1, 0, 100, 10}, +static const struct profile_mode_setting smu7_profiling[7] = + {{0, 0, 0, 0, 0, 0, 0, 0}, + {1, 0, 100, 30, 1, 0, 100, 10}, {1, 10, 0, 30, 0, 0, 0, 0}, {0, 0, 0, 0, 1, 10, 16, 31}, {1, 0, 11, 50, 1, 0, 100, 10}, @@ -269,7 +270,7 @@ static int smu7_construct_voltage_tables(struct pp_hwmgr *hwmgr) hwmgr->dyn_state.mvdd_dependency_on_mclk); PP_ASSERT_WITH_CODE((0 == result), - "Failed to retrieve SVI2 MVDD table from dependancy table.", + "Failed to retrieve SVI2 MVDD table from dependency table.", return result;); } @@ -288,7 +289,7 @@ static int smu7_construct_voltage_tables(struct pp_hwmgr *hwmgr) result = phm_get_svi2_voltage_table_v0(&(data->vddci_voltage_table), hwmgr->dyn_state.vddci_dependency_on_mclk); PP_ASSERT_WITH_CODE((0 == result), - "Failed to retrieve SVI2 VDDCI table from dependancy table.", + "Failed to retrieve SVI2 VDDCI table from dependency table.", return result); } @@ -317,7 +318,7 @@ static int smu7_construct_voltage_tables(struct pp_hwmgr *hwmgr) table_info->vddc_lookup_table); PP_ASSERT_WITH_CODE((0 == result), - "Failed to retrieve SVI2 VDDC table from dependancy table.", return result;); + "Failed to retrieve SVI2 VDDC table from dependency table.", return result;); } tmp = smum_get_mac_definition(hwmgr, SMU_MAX_LEVELS_VDDC); @@ -2859,7 +2860,10 @@ static int smu7_vblank_too_short(struct pp_hwmgr *hwmgr, case CHIP_POLARIS10: case CHIP_POLARIS11: case CHIP_POLARIS12: - switch_limit_us = data->is_memory_gddr5 ? 190 : 150; + if (hwmgr->is_kicker) + switch_limit_us = data->is_memory_gddr5 ? 450 : 150; + else + switch_limit_us = data->is_memory_gddr5 ? 190 : 150; break; case CHIP_VEGAM: switch_limit_us = 30; @@ -3487,14 +3491,14 @@ static int smu7_get_gpu_power(struct pp_hwmgr *hwmgr, u32 *query) smum_send_msg_to_smc(hwmgr, PPSMC_MSG_PmStatusLogStart); cgs_write_ind_register(hwmgr->device, CGS_IND_REG__SMC, - ixSMU_PM_STATUS_94, 0); + ixSMU_PM_STATUS_95, 0); for (i = 0; i < 10; i++) { - mdelay(1); + mdelay(500); smum_send_msg_to_smc(hwmgr, PPSMC_MSG_PmStatusLogSample); tmp = cgs_read_ind_register(hwmgr->device, CGS_IND_REG__SMC, - ixSMU_PM_STATUS_94); + ixSMU_PM_STATUS_95); if (tmp != 0) break; } @@ -3677,10 +3681,12 @@ static int smu7_request_link_speed_change_before_state_change( data->force_pcie_gen = PP_PCIEGen2; if (current_link_speed == PP_PCIEGen2) break; + /* fall through */ case PP_PCIEGen2: if (0 == amdgpu_acpi_pcie_performance_request(hwmgr->adev, PCIE_PERF_REQ_GEN2, false)) break; #endif + /* fall through */ default: data->force_pcie_gen = smu7_get_current_pcie_speed(hwmgr); break; @@ -4223,9 +4229,17 @@ static int smu7_check_mc_firmware(struct pp_hwmgr *hwmgr) if (tmp & (1 << 23)) { data->mem_latency_high = MEM_LATENCY_HIGH; data->mem_latency_low = MEM_LATENCY_LOW; + if ((hwmgr->chip_id == CHIP_POLARIS10) || + (hwmgr->chip_id == CHIP_POLARIS11) || + (hwmgr->chip_id == CHIP_POLARIS12)) + smum_send_msg_to_smc(hwmgr, PPSMC_MSG_EnableFFC); } else { data->mem_latency_high = 330; data->mem_latency_low = 330; + if ((hwmgr->chip_id == CHIP_POLARIS10) || + (hwmgr->chip_id == CHIP_POLARIS11) || + (hwmgr->chip_id == CHIP_POLARIS12)) + smum_send_msg_to_smc(hwmgr, PPSMC_MSG_DisableFFC); } return 0; @@ -4878,7 +4892,8 @@ static int smu7_get_power_profile_mode(struct pp_hwmgr *hwmgr, char *buf) uint32_t i, size = 0; uint32_t len; - static const char *profile_name[6] = {"3D_FULL_SCREEN", + static const char *profile_name[7] = {"BOOTUP_DEFAULT", + "3D_FULL_SCREEN", "POWER_SAVING", "VIDEO", "VR", diff --git a/drivers/gpu/drm/amd/powerplay/hwmgr/smu7_powertune.c b/drivers/gpu/drm/amd/powerplay/hwmgr/smu7_powertune.c index 5e19f5977eb1..58f5589aaf12 100644 --- a/drivers/gpu/drm/amd/powerplay/hwmgr/smu7_powertune.c +++ b/drivers/gpu/drm/amd/powerplay/hwmgr/smu7_powertune.c @@ -967,7 +967,7 @@ int smu7_enable_didt_config(struct pp_hwmgr *hwmgr) PP_CAP(PHM_PlatformCaps_TDRamping) || PP_CAP(PHM_PlatformCaps_TCPRamping)) { - adev->gfx.rlc.funcs->enter_safe_mode(adev); + amdgpu_gfx_rlc_enter_safe_mode(adev); mutex_lock(&adev->grbm_idx_mutex); value = 0; value2 = cgs_read_register(hwmgr->device, mmGRBM_GFX_INDEX); @@ -1014,13 +1014,13 @@ int smu7_enable_didt_config(struct pp_hwmgr *hwmgr) "Failed to enable DPM DIDT.", goto error); } mutex_unlock(&adev->grbm_idx_mutex); - adev->gfx.rlc.funcs->exit_safe_mode(adev); + amdgpu_gfx_rlc_exit_safe_mode(adev); } return 0; error: mutex_unlock(&adev->grbm_idx_mutex); - adev->gfx.rlc.funcs->exit_safe_mode(adev); + amdgpu_gfx_rlc_exit_safe_mode(adev); return result; } @@ -1034,7 +1034,7 @@ int smu7_disable_didt_config(struct pp_hwmgr *hwmgr) PP_CAP(PHM_PlatformCaps_TDRamping) || PP_CAP(PHM_PlatformCaps_TCPRamping)) { - adev->gfx.rlc.funcs->enter_safe_mode(adev); + amdgpu_gfx_rlc_enter_safe_mode(adev); result = smu7_enable_didt(hwmgr, false); PP_ASSERT_WITH_CODE((result == 0), @@ -1046,12 +1046,12 @@ int smu7_disable_didt_config(struct pp_hwmgr *hwmgr) PP_ASSERT_WITH_CODE((0 == result), "Failed to disable DPM DIDT.", goto error); } - adev->gfx.rlc.funcs->exit_safe_mode(adev); + amdgpu_gfx_rlc_exit_safe_mode(adev); } return 0; error: - adev->gfx.rlc.funcs->exit_safe_mode(adev); + amdgpu_gfx_rlc_exit_safe_mode(adev); return result; } @@ -1211,7 +1211,7 @@ int smu7_power_control_set_level(struct pp_hwmgr *hwmgr) hwmgr->platform_descriptor.TDPAdjustment : (-1 * hwmgr->platform_descriptor.TDPAdjustment); - if (hwmgr->chip_id > CHIP_TONGA) + if (hwmgr->chip_id > CHIP_TONGA) target_tdp = ((100 + adjust_percent) * (int)(cac_table->usTDP * 256)) / 100; else target_tdp = ((100 + adjust_percent) * (int)(cac_table->usConfigurableTDP * 256)) / 100; diff --git a/drivers/gpu/drm/amd/powerplay/hwmgr/smu8_hwmgr.c b/drivers/gpu/drm/amd/powerplay/hwmgr/smu8_hwmgr.c index fef111ddb736..019d6a206492 100644 --- a/drivers/gpu/drm/amd/powerplay/hwmgr/smu8_hwmgr.c +++ b/drivers/gpu/drm/amd/powerplay/hwmgr/smu8_hwmgr.c @@ -272,12 +272,10 @@ static int smu8_init_dynamic_state_adjustment_rule_settings( struct pp_hwmgr *hwmgr, ATOM_CLK_VOLT_CAPABILITY *disp_voltage_table) { - uint32_t table_size = - sizeof(struct phm_clock_voltage_dependency_table) + - (7 * sizeof(struct phm_clock_voltage_dependency_record)); + struct phm_clock_voltage_dependency_table *table_clk_vlt; - struct phm_clock_voltage_dependency_table *table_clk_vlt = - kzalloc(table_size, GFP_KERNEL); + table_clk_vlt = kzalloc(struct_size(table_clk_vlt, entries, 7), + GFP_KERNEL); if (NULL == table_clk_vlt) { pr_err("Can not allocate memory!\n"); @@ -1228,17 +1226,14 @@ static int smu8_dpm_force_dpm_level(struct pp_hwmgr *hwmgr, static int smu8_dpm_powerdown_uvd(struct pp_hwmgr *hwmgr) { - if (PP_CAP(PHM_PlatformCaps_UVDPowerGating)) { - smu8_nbdpm_pstate_enable_disable(hwmgr, true, true); + if (PP_CAP(PHM_PlatformCaps_UVDPowerGating)) return smum_send_msg_to_smc(hwmgr, PPSMC_MSG_UVDPowerOFF); - } return 0; } static int smu8_dpm_powerup_uvd(struct pp_hwmgr *hwmgr) { if (PP_CAP(PHM_PlatformCaps_UVDPowerGating)) { - smu8_nbdpm_pstate_enable_disable(hwmgr, false, true); return smum_send_msg_to_smc_with_parameter( hwmgr, PPSMC_MSG_UVDPowerON, @@ -1995,6 +1990,7 @@ static const struct pp_hwmgr_func smu8_hwmgr_funcs = { .power_state_set = smu8_set_power_state_tasks, .dynamic_state_management_disable = smu8_disable_dpm_tasks, .notify_cac_buffer_info = smu8_notify_cac_buffer_info, + .update_nbdpm_pstate = smu8_nbdpm_pstate_enable_disable, .get_thermal_temperature_range = smu8_get_thermal_temperature_range, }; diff --git a/drivers/gpu/drm/amd/powerplay/hwmgr/vega10_baco.c b/drivers/gpu/drm/amd/powerplay/hwmgr/vega10_baco.c new file mode 100644 index 000000000000..7337be5602e4 --- /dev/null +++ b/drivers/gpu/drm/amd/powerplay/hwmgr/vega10_baco.c @@ -0,0 +1,158 @@ +/* + * Copyright 2018 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 "soc15.h" +#include "soc15_hw_ip.h" +#include "vega10_ip_offset.h" +#include "soc15_common.h" +#include "vega10_inc.h" +#include "vega10_ppsmc.h" +#include "vega10_baco.h" + + + +static const struct soc15_baco_cmd_entry pre_baco_tbl[] = +{ + {CMD_READMODIFYWRITE, SOC15_REG_ENTRY(NBIF, 0, mmBIF_DOORBELL_CNTL), BIF_DOORBELL_CNTL__DOORBELL_MONITOR_EN_MASK, BIF_DOORBELL_CNTL__DOORBELL_MONITOR_EN__SHIFT, 0, 1}, + {CMD_WRITE, SOC15_REG_ENTRY(NBIF, 0, mmBIF_FB_EN), 0, 0, 0, 0}, + {CMD_READMODIFYWRITE, SOC15_REG_ENTRY(NBIF, 0, mmBACO_CNTL), BACO_CNTL__BACO_DSTATE_BYPASS_MASK, BACO_CNTL__BACO_DSTATE_BYPASS__SHIFT, 0, 1}, + {CMD_READMODIFYWRITE, SOC15_REG_ENTRY(NBIF, 0, mmBACO_CNTL), BACO_CNTL__BACO_RST_INTR_MASK_MASK, BACO_CNTL__BACO_RST_INTR_MASK__SHIFT, 0, 1} +}; + +static const struct soc15_baco_cmd_entry enter_baco_tbl[] = +{ + {CMD_WAITFOR, SOC15_REG_ENTRY(THM, 0, mmTHM_BACO_CNTL), THM_BACO_CNTL__SOC_DOMAIN_IDLE_MASK, THM_BACO_CNTL__SOC_DOMAIN_IDLE__SHIFT, 0xffffffff, 0x80000000}, + {CMD_READMODIFYWRITE, SOC15_REG_ENTRY(NBIF, 0, mmBACO_CNTL), BACO_CNTL__BACO_EN_MASK, BACO_CNTL__BACO_EN__SHIFT, 0, 1}, + {CMD_READMODIFYWRITE, SOC15_REG_ENTRY(NBIF, 0, mmBACO_CNTL), BACO_CNTL__BACO_BIF_LCLK_SWITCH_MASK, BACO_CNTL__BACO_BIF_LCLK_SWITCH__SHIFT, 0, 1}, + {CMD_READMODIFYWRITE, SOC15_REG_ENTRY(NBIF, 0, mmBACO_CNTL), BACO_CNTL__BACO_DUMMY_EN_MASK, BACO_CNTL__BACO_DUMMY_EN__SHIFT, 0, 1}, + {CMD_READMODIFYWRITE, SOC15_REG_ENTRY(THM, 0, mmTHM_BACO_CNTL), THM_BACO_CNTL__BACO_SOC_VDCI_RESET_MASK, THM_BACO_CNTL__BACO_SOC_VDCI_RESET__SHIFT, 0, 1}, + {CMD_READMODIFYWRITE, SOC15_REG_ENTRY(THM, 0, mmTHM_BACO_CNTL), THM_BACO_CNTL__BACO_SMNCLK_MUX_MASK, THM_BACO_CNTL__BACO_SMNCLK_MUX__SHIFT,0, 1}, + {CMD_READMODIFYWRITE, SOC15_REG_ENTRY(THM, 0, mmTHM_BACO_CNTL), THM_BACO_CNTL__BACO_ISO_EN_MASK, THM_BACO_CNTL__BACO_ISO_EN__SHIFT, 0, 1}, + {CMD_READMODIFYWRITE, SOC15_REG_ENTRY(THM, 0, mmTHM_BACO_CNTL), THM_BACO_CNTL__BACO_AEB_ISO_EN_MASK, THM_BACO_CNTL__BACO_AEB_ISO_EN__SHIFT,0, 1}, + {CMD_READMODIFYWRITE, SOC15_REG_ENTRY(THM, 0, mmTHM_BACO_CNTL), THM_BACO_CNTL__BACO_ANA_ISO_EN_MASK, THM_BACO_CNTL__BACO_ANA_ISO_EN__SHIFT, 0, 1}, + {CMD_READMODIFYWRITE, SOC15_REG_ENTRY(THM, 0, mmTHM_BACO_CNTL), THM_BACO_CNTL__BACO_SOC_REFCLK_OFF_MASK, THM_BACO_CNTL__BACO_SOC_REFCLK_OFF__SHIFT, 0, 1}, + {CMD_READMODIFYWRITE, SOC15_REG_ENTRY(NBIF, 0, mmBACO_CNTL), BACO_CNTL__BACO_POWER_OFF_MASK, BACO_CNTL__BACO_POWER_OFF__SHIFT, 0, 1}, + {CMD_DELAY_MS, 0, 0, 0, 0, 0, 0, 5, 0}, + {CMD_READMODIFYWRITE, SOC15_REG_ENTRY(THM, 0, mmTHM_BACO_CNTL), THM_BACO_CNTL__BACO_RESET_EN_MASK, THM_BACO_CNTL__BACO_RESET_EN__SHIFT, 0, 1}, + {CMD_READMODIFYWRITE, SOC15_REG_ENTRY(THM, 0, mmTHM_BACO_CNTL), THM_BACO_CNTL__BACO_PWROKRAW_CNTL_MASK, THM_BACO_CNTL__BACO_PWROKRAW_CNTL__SHIFT, 0, 0}, + {CMD_WAITFOR, SOC15_REG_ENTRY(NBIF, 0, mmBACO_CNTL), BACO_CNTL__BACO_MODE_MASK, BACO_CNTL__BACO_MODE__SHIFT, 0xffffffff, 0x100} +}; + +static const struct soc15_baco_cmd_entry exit_baco_tbl[] = +{ + {CMD_READMODIFYWRITE, SOC15_REG_ENTRY(NBIF, 0, mmBACO_CNTL), BACO_CNTL__BACO_POWER_OFF_MASK, BACO_CNTL__BACO_POWER_OFF__SHIFT, 0, 0}, + {CMD_DELAY_MS, 0, 0, 0, 0, 0, 0, 10,0}, + {CMD_READMODIFYWRITE, SOC15_REG_ENTRY(THM, 0, mmTHM_BACO_CNTL), THM_BACO_CNTL__BACO_SOC_REFCLK_OFF_MASK, THM_BACO_CNTL__BACO_SOC_REFCLK_OFF__SHIFT, 0,0}, + {CMD_READMODIFYWRITE, SOC15_REG_ENTRY(THM, 0, mmTHM_BACO_CNTL), THM_BACO_CNTL__BACO_ANA_ISO_EN_MASK, THM_BACO_CNTL__BACO_ANA_ISO_EN__SHIFT, 0, 0}, + {CMD_READMODIFYWRITE, SOC15_REG_ENTRY(THM, 0, mmTHM_BACO_CNTL), THM_BACO_CNTL__BACO_AEB_ISO_EN_MASK, THM_BACO_CNTL__BACO_AEB_ISO_EN__SHIFT,0, 0}, + {CMD_READMODIFYWRITE, SOC15_REG_ENTRY(THM, 0, mmTHM_BACO_CNTL), THM_BACO_CNTL__BACO_ISO_EN_MASK, THM_BACO_CNTL__BACO_ISO_EN__SHIFT, 0, 0}, + {CMD_READMODIFYWRITE, SOC15_REG_ENTRY(THM, 0, mmTHM_BACO_CNTL), THM_BACO_CNTL__BACO_PWROKRAW_CNTL_MASK, THM_BACO_CNTL__BACO_PWROKRAW_CNTL__SHIFT, 0, 1}, + {CMD_READMODIFYWRITE, SOC15_REG_ENTRY(THM, 0, mmTHM_BACO_CNTL), THM_BACO_CNTL__BACO_SMNCLK_MUX_MASK, THM_BACO_CNTL__BACO_SMNCLK_MUX__SHIFT, 0, 0}, + {CMD_READMODIFYWRITE, SOC15_REG_ENTRY(THM, 0, mmTHM_BACO_CNTL), THM_BACO_CNTL__BACO_SOC_VDCI_RESET_MASK, THM_BACO_CNTL__BACO_SOC_VDCI_RESET__SHIFT, 0, 0}, + {CMD_READMODIFYWRITE, SOC15_REG_ENTRY(THM, 0, mmTHM_BACO_CNTL), THM_BACO_CNTL__BACO_EXIT_MASK, THM_BACO_CNTL__BACO_EXIT__SHIFT, 0, 1}, + {CMD_READMODIFYWRITE, SOC15_REG_ENTRY(THM, 0, mmTHM_BACO_CNTL), THM_BACO_CNTL__BACO_RESET_EN_MASK, THM_BACO_CNTL__BACO_RESET_EN__SHIFT, 0, 0}, + {CMD_WAITFOR, SOC15_REG_ENTRY(THM, 0, mmTHM_BACO_CNTL), THM_BACO_CNTL__BACO_EXIT_MASK, 0, 0xffffffff, 0}, + {CMD_READMODIFYWRITE, SOC15_REG_ENTRY(THM, 0, mmTHM_BACO_CNTL), THM_BACO_CNTL__BACO_SB_AXI_FENCE_MASK, THM_BACO_CNTL__BACO_SB_AXI_FENCE__SHIFT, 0, 0}, + {CMD_READMODIFYWRITE, SOC15_REG_ENTRY(NBIF, 0, mmBACO_CNTL), BACO_CNTL__BACO_DUMMY_EN_MASK, BACO_CNTL__BACO_DUMMY_EN__SHIFT, 0, 0}, + {CMD_READMODIFYWRITE, SOC15_REG_ENTRY(NBIF, 0, mmBACO_CNTL), BACO_CNTL__BACO_BIF_LCLK_SWITCH_MASK ,BACO_CNTL__BACO_BIF_LCLK_SWITCH__SHIFT, 0, 0}, + {CMD_READMODIFYWRITE, SOC15_REG_ENTRY(NBIF, 0, mmBACO_CNTL), BACO_CNTL__BACO_EN_MASK , BACO_CNTL__BACO_EN__SHIFT, 0,0}, + {CMD_WAITFOR, SOC15_REG_ENTRY(NBIF, 0, mmBACO_CNTL), BACO_CNTL__BACO_MODE_MASK, 0, 0xffffffff, 0} + }; + +static const struct soc15_baco_cmd_entry clean_baco_tbl[] = +{ + {CMD_WRITE, SOC15_REG_ENTRY(NBIF, 0, mmBIOS_SCRATCH_6), 0, 0, 0, 0}, + {CMD_WRITE, SOC15_REG_ENTRY(NBIF, 0, mmBIOS_SCRATCH_7), 0, 0, 0, 0}, +}; + +int vega10_baco_get_capability(struct pp_hwmgr *hwmgr, bool *cap) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)(hwmgr->adev); + uint32_t reg, data; + + *cap = false; + if (!phm_cap_enabled(hwmgr->platform_descriptor.platformCaps, PHM_PlatformCaps_BACO)) + return 0; + + WREG32(0x12074, 0xFFF0003B); + data = RREG32(0x12075); + + if (data == 0x1) { + reg = RREG32_SOC15(NBIF, 0, mmRCC_BIF_STRAP0); + + if (reg & RCC_BIF_STRAP0__STRAP_PX_CAPABLE_MASK) + *cap = true; + } + + return 0; +} + +int vega10_baco_get_state(struct pp_hwmgr *hwmgr, enum BACO_STATE *state) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)(hwmgr->adev); + uint32_t reg; + + reg = RREG32_SOC15(NBIF, 0, mmBACO_CNTL); + + if (reg & BACO_CNTL__BACO_MODE_MASK) + /* gfx has already entered BACO state */ + *state = BACO_STATE_IN; + else + *state = BACO_STATE_OUT; + return 0; +} + +int vega10_baco_set_state(struct pp_hwmgr *hwmgr, enum BACO_STATE state) +{ + enum BACO_STATE cur_state; + + vega10_baco_get_state(hwmgr, &cur_state); + + if (cur_state == state) + /* aisc already in the target state */ + return 0; + + if (state == BACO_STATE_IN) { + if (soc15_baco_program_registers(hwmgr, pre_baco_tbl, + ARRAY_SIZE(pre_baco_tbl))) { + if (smum_send_msg_to_smc(hwmgr, PPSMC_MSG_EnterBaco)) + return -EINVAL; + + if (soc15_baco_program_registers(hwmgr, enter_baco_tbl, + ARRAY_SIZE(enter_baco_tbl))) + return 0; + } + } else if (state == BACO_STATE_OUT) { + /* HW requires at least 20ms between regulator off and on */ + msleep(20); + /* Execute Hardware BACO exit sequence */ + if (soc15_baco_program_registers(hwmgr, exit_baco_tbl, + ARRAY_SIZE(exit_baco_tbl))) { + if (soc15_baco_program_registers(hwmgr, clean_baco_tbl, + ARRAY_SIZE(clean_baco_tbl))) + return 0; + } + } + + return -EINVAL; +} diff --git a/drivers/gpu/drm/amd/display/dc/i2caux/dce120/i2caux_dce120.h b/drivers/gpu/drm/amd/powerplay/hwmgr/vega10_baco.h index b6ac47617c70..f7a3ffa744b3 100644 --- a/drivers/gpu/drm/amd/display/dc/i2caux/dce120/i2caux_dce120.h +++ b/drivers/gpu/drm/amd/powerplay/hwmgr/vega10_baco.h @@ -1,5 +1,5 @@ /* - * Copyright 2012-16 Advanced Micro Devices, Inc. + * Copyright 2018 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"), @@ -19,14 +19,14 @@ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR * OTHER DEALINGS IN THE SOFTWARE. * - * Authors: AMD - * */ +#ifndef __VEGA10_BACO_H__ +#define __VEGA10_BACO_H__ +#include "hwmgr.h" +#include "common_baco.h" -#ifndef __DAL_I2C_AUX_DCE120_H__ -#define __DAL_I2C_AUX_DCE120_H__ - -struct i2caux *dal_i2caux_dce120_create( - struct dc_context *ctx); +extern int vega10_baco_get_capability(struct pp_hwmgr *hwmgr, bool *cap); +extern int vega10_baco_get_state(struct pp_hwmgr *hwmgr, enum BACO_STATE *state); +extern int vega10_baco_set_state(struct pp_hwmgr *hwmgr, enum BACO_STATE state); -#endif /* __DAL_I2C_AUX_DCE120_H__ */ +#endif diff --git a/drivers/gpu/drm/amd/powerplay/hwmgr/vega10_hwmgr.c b/drivers/gpu/drm/amd/powerplay/hwmgr/vega10_hwmgr.c index 79c86247d0ac..5c4f701939ea 100644 --- a/drivers/gpu/drm/amd/powerplay/hwmgr/vega10_hwmgr.c +++ b/drivers/gpu/drm/amd/powerplay/hwmgr/vega10_hwmgr.c @@ -48,6 +48,7 @@ #include "ppinterrupt.h" #include "pp_overdriver.h" #include "pp_thermal.h" +#include "vega10_baco.h" #include "smuio/smuio_9_0_offset.h" #include "smuio/smuio_9_0_sh_mask.h" @@ -71,6 +72,21 @@ static const uint32_t channel_number[] = {1, 2, 0, 4, 0, 8, 0, 16, 2}; #define DF_CS_AON0_DramBaseAddress0__IntLvAddrSel_MASK 0x00000700L #define DF_CS_AON0_DramBaseAddress0__DramBaseAddr_MASK 0xFFFFF000L +typedef enum { + CLK_SMNCLK = 0, + CLK_SOCCLK, + CLK_MP0CLK, + CLK_MP1CLK, + CLK_LCLK, + CLK_DCEFCLK, + CLK_VCLK, + CLK_DCLK, + CLK_ECLK, + CLK_UCLK, + CLK_GFXCLK, + CLK_COUNT, +} CLOCK_ID_e; + static const ULONG PhwVega10_Magic = (ULONG)(PHM_VIslands_Magic); struct vega10_power_state *cast_phw_vega10_power_state( @@ -804,9 +820,9 @@ static int vega10_hwmgr_backend_init(struct pp_hwmgr *hwmgr) hwmgr->backend = data; - hwmgr->workload_mask = 1 << hwmgr->workload_prority[PP_SMC_POWER_PROFILE_VIDEO]; - hwmgr->power_profile_mode = PP_SMC_POWER_PROFILE_VIDEO; - hwmgr->default_power_profile_mode = PP_SMC_POWER_PROFILE_VIDEO; + hwmgr->workload_mask = 1 << hwmgr->workload_prority[PP_SMC_POWER_PROFILE_BOOTUP_DEFAULT]; + hwmgr->power_profile_mode = PP_SMC_POWER_PROFILE_BOOTUP_DEFAULT; + hwmgr->default_power_profile_mode = PP_SMC_POWER_PROFILE_BOOTUP_DEFAULT; vega10_set_default_registry_data(hwmgr); data->disable_dpm_mask = 0xff; @@ -2559,10 +2575,10 @@ static int vega10_init_smc_table(struct pp_hwmgr *hwmgr) data->vbios_boot_state.gfx_clock = boot_up_values.ulGfxClk; data->vbios_boot_state.mem_clock = boot_up_values.ulUClk; pp_atomfwctrl_get_clk_information_by_clkid(hwmgr, - SMU9_SYSPLL0_SOCCLK_ID, &boot_up_values.ulSocClk); + SMU9_SYSPLL0_SOCCLK_ID, 0, &boot_up_values.ulSocClk); pp_atomfwctrl_get_clk_information_by_clkid(hwmgr, - SMU9_SYSPLL0_DCEFCLK_ID, &boot_up_values.ulDCEFClk); + SMU9_SYSPLL0_DCEFCLK_ID, 0, &boot_up_values.ulDCEFClk); data->vbios_boot_state.soc_clock = boot_up_values.ulSocClk; data->vbios_boot_state.dcef_clock = boot_up_values.ulDCEFClk; @@ -3485,6 +3501,17 @@ static int vega10_upload_dpm_bootup_level(struct pp_hwmgr *hwmgr) } } + if (!data->registry_data.socclk_dpm_key_disabled) { + if (data->smc_state_table.soc_boot_level != + data->dpm_table.soc_table.dpm_state.soft_min_level) { + smum_send_msg_to_smc_with_parameter(hwmgr, + PPSMC_MSG_SetSoftMinSocclkByIndex, + data->smc_state_table.soc_boot_level); + data->dpm_table.soc_table.dpm_state.soft_min_level = + data->smc_state_table.soc_boot_level; + } + } + return 0; } @@ -3516,6 +3543,17 @@ static int vega10_upload_dpm_max_level(struct pp_hwmgr *hwmgr) } } + if (!data->registry_data.socclk_dpm_key_disabled) { + if (data->smc_state_table.soc_max_level != + data->dpm_table.soc_table.dpm_state.soft_max_level) { + smum_send_msg_to_smc_with_parameter(hwmgr, + PPSMC_MSG_SetSoftMaxSocclkByIndex, + data->smc_state_table.soc_max_level); + data->dpm_table.soc_table.dpm_state.soft_max_level = + data->smc_state_table.soc_max_level; + } + } + return 0; } @@ -3541,6 +3579,10 @@ static int vega10_generate_dpm_level_enable_mask( vega10_find_lowest_dpm_level(&(data->dpm_table.mem_table)); data->smc_state_table.mem_max_level = vega10_find_highest_dpm_level(&(data->dpm_table.mem_table)); + data->smc_state_table.soc_boot_level = + vega10_find_lowest_dpm_level(&(data->dpm_table.soc_table)); + data->smc_state_table.soc_max_level = + vega10_find_highest_dpm_level(&(data->dpm_table.soc_table)); PP_ASSERT_WITH_CODE(!vega10_upload_dpm_bootup_level(hwmgr), "Attempt to upload DPM Bootup Levels Failed!", @@ -3555,6 +3597,9 @@ static int vega10_generate_dpm_level_enable_mask( for(i = data->smc_state_table.mem_boot_level; i < data->smc_state_table.mem_max_level; i++) data->dpm_table.mem_table.dpm_levels[i].enabled = true; + for (i = data->smc_state_table.soc_boot_level; i < data->smc_state_table.soc_max_level; i++) + data->dpm_table.soc_table.dpm_levels[i].enabled = true; + return 0; } @@ -4028,6 +4073,24 @@ static int vega10_force_clock_level(struct pp_hwmgr *hwmgr, break; + case PP_SOCCLK: + data->smc_state_table.soc_boot_level = mask ? (ffs(mask) - 1) : 0; + data->smc_state_table.soc_max_level = mask ? (fls(mask) - 1) : 0; + + PP_ASSERT_WITH_CODE(!vega10_upload_dpm_bootup_level(hwmgr), + "Failed to upload boot level to lowest!", + return -EINVAL); + + PP_ASSERT_WITH_CODE(!vega10_upload_dpm_max_level(hwmgr), + "Failed to upload dpm max level to highest!", + return -EINVAL); + + break; + + case PP_DCEFCLK: + pr_info("Setting DCEFCLK min/max dpm level is not supported!\n"); + break; + case PP_PCIE: default: break; @@ -4267,12 +4330,113 @@ static int vega10_set_watermarks_for_clocks_ranges(struct pp_hwmgr *hwmgr, return result; } +static int vega10_get_ppfeature_status(struct pp_hwmgr *hwmgr, char *buf) +{ + static const char *ppfeature_name[] = { + "DPM_PREFETCHER", + "GFXCLK_DPM", + "UCLK_DPM", + "SOCCLK_DPM", + "UVD_DPM", + "VCE_DPM", + "ULV", + "MP0CLK_DPM", + "LINK_DPM", + "DCEFCLK_DPM", + "AVFS", + "GFXCLK_DS", + "SOCCLK_DS", + "LCLK_DS", + "PPT", + "TDC", + "THERMAL", + "GFX_PER_CU_CG", + "RM", + "DCEFCLK_DS", + "ACDC", + "VR0HOT", + "VR1HOT", + "FW_CTF", + "LED_DISPLAY", + "FAN_CONTROL", + "FAST_PPT", + "DIDT", + "ACG", + "PCC_LIMIT"}; + static const char *output_title[] = { + "FEATURES", + "BITMASK", + "ENABLEMENT"}; + uint64_t features_enabled; + int i; + int ret = 0; + int size = 0; + + ret = vega10_get_enabled_smc_features(hwmgr, &features_enabled); + PP_ASSERT_WITH_CODE(!ret, + "[EnableAllSmuFeatures] Failed to get enabled smc features!", + return ret); + + size += sprintf(buf + size, "Current ppfeatures: 0x%016llx\n", features_enabled); + size += sprintf(buf + size, "%-19s %-22s %s\n", + output_title[0], + output_title[1], + output_title[2]); + for (i = 0; i < GNLD_FEATURES_MAX; i++) { + size += sprintf(buf + size, "%-19s 0x%016llx %6s\n", + ppfeature_name[i], + 1ULL << i, + (features_enabled & (1ULL << i)) ? "Y" : "N"); + } + + return size; +} + +static int vega10_set_ppfeature_status(struct pp_hwmgr *hwmgr, uint64_t new_ppfeature_masks) +{ + uint64_t features_enabled; + uint64_t features_to_enable; + uint64_t features_to_disable; + int ret = 0; + + if (new_ppfeature_masks >= (1ULL << GNLD_FEATURES_MAX)) + return -EINVAL; + + ret = vega10_get_enabled_smc_features(hwmgr, &features_enabled); + if (ret) + return ret; + + features_to_disable = + features_enabled & ~new_ppfeature_masks; + features_to_enable = + ~features_enabled & new_ppfeature_masks; + + pr_debug("features_to_disable 0x%llx\n", features_to_disable); + pr_debug("features_to_enable 0x%llx\n", features_to_enable); + + if (features_to_disable) { + ret = vega10_enable_smc_features(hwmgr, false, features_to_disable); + if (ret) + return ret; + } + + if (features_to_enable) { + ret = vega10_enable_smc_features(hwmgr, true, features_to_enable); + if (ret) + return ret; + } + + return 0; +} + static int vega10_print_clock_levels(struct pp_hwmgr *hwmgr, enum pp_clock_type type, char *buf) { struct vega10_hwmgr *data = hwmgr->backend; struct vega10_single_dpm_table *sclk_table = &(data->dpm_table.gfx_table); struct vega10_single_dpm_table *mclk_table = &(data->dpm_table.mem_table); + struct vega10_single_dpm_table *soc_table = &(data->dpm_table.soc_table); + struct vega10_single_dpm_table *dcef_table = &(data->dpm_table.dcef_table); struct vega10_pcie_table *pcie_table = &(data->dpm_table.pcie_table); struct vega10_odn_clock_voltage_dependency_table *podn_vdd_dep = NULL; @@ -4303,6 +4467,32 @@ static int vega10_print_clock_levels(struct pp_hwmgr *hwmgr, i, mclk_table->dpm_levels[i].value / 100, (i == now) ? "*" : ""); break; + case PP_SOCCLK: + if (data->registry_data.socclk_dpm_key_disabled) + break; + + smum_send_msg_to_smc(hwmgr, PPSMC_MSG_GetCurrentSocclkIndex); + now = smum_get_argument(hwmgr); + + for (i = 0; i < soc_table->count; i++) + size += sprintf(buf + size, "%d: %uMhz %s\n", + i, soc_table->dpm_levels[i].value / 100, + (i == now) ? "*" : ""); + break; + case PP_DCEFCLK: + if (data->registry_data.dcefclk_dpm_key_disabled) + break; + + smum_send_msg_to_smc_with_parameter(hwmgr, + PPSMC_MSG_GetClockFreqMHz, CLK_DCEFCLK); + now = smum_get_argument(hwmgr); + + for (i = 0; i < dcef_table->count; i++) + size += sprintf(buf + size, "%d: %uMhz %s\n", + i, dcef_table->dpm_levels[i].value / 100, + (dcef_table->dpm_levels[i].value / 100 == now) ? + "*" : ""); + break; case PP_PCIE: smum_send_msg_to_smc(hwmgr, PPSMC_MSG_GetCurrentLinkIndex); now = smum_get_argument(hwmgr); @@ -4668,13 +4858,15 @@ static int vega10_get_power_profile_mode(struct pp_hwmgr *hwmgr, char *buf) { struct vega10_hwmgr *data = hwmgr->backend; uint32_t i, size = 0; - static const uint8_t profile_mode_setting[5][4] = {{70, 60, 1, 3,}, + static const uint8_t profile_mode_setting[6][4] = {{70, 60, 0, 0,}, + {70, 60, 1, 3,}, {90, 60, 0, 0,}, {70, 60, 0, 0,}, {70, 90, 0, 0,}, {30, 60, 0, 6,}, }; - static const char *profile_name[6] = {"3D_FULL_SCREEN", + static const char *profile_name[7] = {"BOOTUP_DEFAULT", + "3D_FULL_SCREEN", "POWER_SAVING", "VIDEO", "VR", @@ -4978,6 +5170,12 @@ static const struct pp_hwmgr_func vega10_hwmgr_funcs = { .set_power_limit = vega10_set_power_limit, .odn_edit_dpm_table = vega10_odn_edit_dpm_table, .get_performance_level = vega10_get_performance_level, + .get_asic_baco_capability = vega10_baco_get_capability, + .get_asic_baco_state = vega10_baco_get_state, + .set_asic_baco_state = vega10_baco_set_state, + .enable_mgpu_fan_boost = vega10_enable_mgpu_fan_boost, + .get_ppfeature_status = vega10_get_ppfeature_status, + .set_ppfeature_status = vega10_set_ppfeature_status, }; int vega10_hwmgr_init(struct pp_hwmgr *hwmgr) diff --git a/drivers/gpu/drm/amd/powerplay/hwmgr/vega10_hwmgr.h b/drivers/gpu/drm/amd/powerplay/hwmgr/vega10_hwmgr.h index 89870556de1b..f752b4ad0c8a 100644 --- a/drivers/gpu/drm/amd/powerplay/hwmgr/vega10_hwmgr.h +++ b/drivers/gpu/drm/amd/powerplay/hwmgr/vega10_hwmgr.h @@ -199,6 +199,7 @@ struct vega10_smc_state_table { uint32_t vce_boot_level; uint32_t gfx_max_level; uint32_t mem_max_level; + uint32_t soc_max_level; uint8_t vr_hot_gpio; uint8_t ac_dc_gpio; uint8_t therm_out_gpio; diff --git a/drivers/gpu/drm/amd/powerplay/hwmgr/vega10_powertune.c b/drivers/gpu/drm/amd/powerplay/hwmgr/vega10_powertune.c index 2d88abf97e7b..6f26cb241ecc 100644 --- a/drivers/gpu/drm/amd/powerplay/hwmgr/vega10_powertune.c +++ b/drivers/gpu/drm/amd/powerplay/hwmgr/vega10_powertune.c @@ -937,7 +937,7 @@ static int vega10_enable_cac_driving_se_didt_config(struct pp_hwmgr *hwmgr) num_se = adev->gfx.config.max_shader_engines; - adev->gfx.rlc.funcs->enter_safe_mode(adev); + amdgpu_gfx_rlc_enter_safe_mode(adev); mutex_lock(&adev->grbm_idx_mutex); for (count = 0; count < num_se; count++) { @@ -962,7 +962,7 @@ static int vega10_enable_cac_driving_se_didt_config(struct pp_hwmgr *hwmgr) vega10_didt_set_mask(hwmgr, true); - adev->gfx.rlc.funcs->exit_safe_mode(adev); + amdgpu_gfx_rlc_exit_safe_mode(adev); return 0; } @@ -971,11 +971,11 @@ static int vega10_disable_cac_driving_se_didt_config(struct pp_hwmgr *hwmgr) { struct amdgpu_device *adev = hwmgr->adev; - adev->gfx.rlc.funcs->enter_safe_mode(adev); + amdgpu_gfx_rlc_enter_safe_mode(adev); vega10_didt_set_mask(hwmgr, false); - adev->gfx.rlc.funcs->exit_safe_mode(adev); + amdgpu_gfx_rlc_exit_safe_mode(adev); return 0; } @@ -988,7 +988,7 @@ static int vega10_enable_psm_gc_didt_config(struct pp_hwmgr *hwmgr) num_se = adev->gfx.config.max_shader_engines; - adev->gfx.rlc.funcs->enter_safe_mode(adev); + amdgpu_gfx_rlc_enter_safe_mode(adev); mutex_lock(&adev->grbm_idx_mutex); for (count = 0; count < num_se; count++) { @@ -1007,7 +1007,7 @@ static int vega10_enable_psm_gc_didt_config(struct pp_hwmgr *hwmgr) vega10_didt_set_mask(hwmgr, true); - adev->gfx.rlc.funcs->exit_safe_mode(adev); + amdgpu_gfx_rlc_exit_safe_mode(adev); vega10_program_gc_didt_config_registers(hwmgr, GCDiDtDroopCtrlConfig_vega10); if (PP_CAP(PHM_PlatformCaps_GCEDC)) @@ -1024,11 +1024,11 @@ static int vega10_disable_psm_gc_didt_config(struct pp_hwmgr *hwmgr) struct amdgpu_device *adev = hwmgr->adev; uint32_t data; - adev->gfx.rlc.funcs->enter_safe_mode(adev); + amdgpu_gfx_rlc_enter_safe_mode(adev); vega10_didt_set_mask(hwmgr, false); - adev->gfx.rlc.funcs->exit_safe_mode(adev); + amdgpu_gfx_rlc_exit_safe_mode(adev); if (PP_CAP(PHM_PlatformCaps_GCEDC)) { data = 0x00000000; @@ -1049,7 +1049,7 @@ static int vega10_enable_se_edc_config(struct pp_hwmgr *hwmgr) num_se = adev->gfx.config.max_shader_engines; - adev->gfx.rlc.funcs->enter_safe_mode(adev); + amdgpu_gfx_rlc_enter_safe_mode(adev); mutex_lock(&adev->grbm_idx_mutex); for (count = 0; count < num_se; count++) { @@ -1070,7 +1070,7 @@ static int vega10_enable_se_edc_config(struct pp_hwmgr *hwmgr) vega10_didt_set_mask(hwmgr, true); - adev->gfx.rlc.funcs->exit_safe_mode(adev); + amdgpu_gfx_rlc_exit_safe_mode(adev); return 0; } @@ -1079,11 +1079,11 @@ static int vega10_disable_se_edc_config(struct pp_hwmgr *hwmgr) { struct amdgpu_device *adev = hwmgr->adev; - adev->gfx.rlc.funcs->enter_safe_mode(adev); + amdgpu_gfx_rlc_enter_safe_mode(adev); vega10_didt_set_mask(hwmgr, false); - adev->gfx.rlc.funcs->exit_safe_mode(adev); + amdgpu_gfx_rlc_exit_safe_mode(adev); return 0; } @@ -1097,7 +1097,7 @@ static int vega10_enable_psm_gc_edc_config(struct pp_hwmgr *hwmgr) num_se = adev->gfx.config.max_shader_engines; - adev->gfx.rlc.funcs->enter_safe_mode(adev); + amdgpu_gfx_rlc_enter_safe_mode(adev); vega10_program_gc_didt_config_registers(hwmgr, AvfsPSMResetConfig_vega10); @@ -1118,7 +1118,7 @@ static int vega10_enable_psm_gc_edc_config(struct pp_hwmgr *hwmgr) vega10_didt_set_mask(hwmgr, true); - adev->gfx.rlc.funcs->exit_safe_mode(adev); + amdgpu_gfx_rlc_exit_safe_mode(adev); vega10_program_gc_didt_config_registers(hwmgr, PSMGCEDCDroopCtrlConfig_vega10); @@ -1138,11 +1138,11 @@ static int vega10_disable_psm_gc_edc_config(struct pp_hwmgr *hwmgr) struct amdgpu_device *adev = hwmgr->adev; uint32_t data; - adev->gfx.rlc.funcs->enter_safe_mode(adev); + amdgpu_gfx_rlc_enter_safe_mode(adev); vega10_didt_set_mask(hwmgr, false); - adev->gfx.rlc.funcs->exit_safe_mode(adev); + amdgpu_gfx_rlc_exit_safe_mode(adev); if (PP_CAP(PHM_PlatformCaps_GCEDC)) { data = 0x00000000; @@ -1160,7 +1160,7 @@ static int vega10_enable_se_edc_force_stall_config(struct pp_hwmgr *hwmgr) struct amdgpu_device *adev = hwmgr->adev; int result; - adev->gfx.rlc.funcs->enter_safe_mode(adev); + amdgpu_gfx_rlc_enter_safe_mode(adev); mutex_lock(&adev->grbm_idx_mutex); WREG32_SOC15(GC, 0, mmGRBM_GFX_INDEX, 0xE0000000); @@ -1173,7 +1173,7 @@ static int vega10_enable_se_edc_force_stall_config(struct pp_hwmgr *hwmgr) vega10_didt_set_mask(hwmgr, false); - adev->gfx.rlc.funcs->exit_safe_mode(adev); + amdgpu_gfx_rlc_exit_safe_mode(adev); return 0; } diff --git a/drivers/gpu/drm/amd/powerplay/hwmgr/vega10_pptable.h b/drivers/gpu/drm/amd/powerplay/hwmgr/vega10_pptable.h index b3e63003a789..c934e9612c1b 100644 --- a/drivers/gpu/drm/amd/powerplay/hwmgr/vega10_pptable.h +++ b/drivers/gpu/drm/amd/powerplay/hwmgr/vega10_pptable.h @@ -282,6 +282,30 @@ typedef struct _ATOM_Vega10_Fan_Table_V2 { UCHAR ucFanMaxRPM; } ATOM_Vega10_Fan_Table_V2; +typedef struct _ATOM_Vega10_Fan_Table_V3 { + UCHAR ucRevId; + USHORT usFanOutputSensitivity; + USHORT usFanAcousticLimitRpm; + USHORT usThrottlingRPM; + USHORT usTargetTemperature; + USHORT usMinimumPWMLimit; + USHORT usTargetGfxClk; + USHORT usFanGainEdge; + USHORT usFanGainHotspot; + USHORT usFanGainLiquid; + USHORT usFanGainVrVddc; + USHORT usFanGainVrMvdd; + USHORT usFanGainPlx; + USHORT usFanGainHbm; + UCHAR ucEnableZeroRPM; + USHORT usFanStopTemperature; + USHORT usFanStartTemperature; + UCHAR ucFanParameters; + UCHAR ucFanMinRPM; + UCHAR ucFanMaxRPM; + USHORT usMGpuThrottlingRPM; +} ATOM_Vega10_Fan_Table_V3; + typedef struct _ATOM_Vega10_Thermal_Controller { UCHAR ucRevId; UCHAR ucType; /* one of ATOM_VEGA10_PP_THERMALCONTROLLER_*/ diff --git a/drivers/gpu/drm/amd/powerplay/hwmgr/vega10_processpptables.c b/drivers/gpu/drm/amd/powerplay/hwmgr/vega10_processpptables.c index b8747a5c9204..b6767d74dc85 100644 --- a/drivers/gpu/drm/amd/powerplay/hwmgr/vega10_processpptables.c +++ b/drivers/gpu/drm/amd/powerplay/hwmgr/vega10_processpptables.c @@ -32,6 +32,7 @@ #include "vega10_pptable.h" #define NUM_DSPCLK_LEVELS 8 +#define VEGA10_ENGINECLOCK_HARDMAX 198000 static void set_hw_cap(struct pp_hwmgr *hwmgr, bool enable, enum phm_platform_caps cap) @@ -122,6 +123,7 @@ static int init_thermal_controller( const Vega10_PPTable_Generic_SubTable_Header *header; const ATOM_Vega10_Fan_Table *fan_table_v1; const ATOM_Vega10_Fan_Table_V2 *fan_table_v2; + const ATOM_Vega10_Fan_Table_V3 *fan_table_v3; thermal_controller = (ATOM_Vega10_Thermal_Controller *) (((unsigned long)powerplay_table) + @@ -206,7 +208,7 @@ static int init_thermal_controller( le16_to_cpu(fan_table_v1->usFanStopTemperature); hwmgr->thermal_controller.advanceFanControlParameters.usZeroRPMStartTemperature = le16_to_cpu(fan_table_v1->usFanStartTemperature); - } else if (header->ucRevId > 10) { + } else if (header->ucRevId == 0xb) { fan_table_v2 = (ATOM_Vega10_Fan_Table_V2 *)header; hwmgr->thermal_controller.fanInfo.ucTachometerPulsesPerRevolution = @@ -250,7 +252,54 @@ static int init_thermal_controller( le16_to_cpu(fan_table_v2->usFanStopTemperature); hwmgr->thermal_controller.advanceFanControlParameters.usZeroRPMStartTemperature = le16_to_cpu(fan_table_v2->usFanStartTemperature); + } else if (header->ucRevId > 0xb) { + fan_table_v3 = (ATOM_Vega10_Fan_Table_V3 *)header; + + hwmgr->thermal_controller.fanInfo.ucTachometerPulsesPerRevolution = + fan_table_v3->ucFanParameters & ATOM_VEGA10_PP_FANPARAMETERS_TACHOMETER_PULSES_PER_REVOLUTION_MASK; + hwmgr->thermal_controller.fanInfo.ulMinRPM = fan_table_v3->ucFanMinRPM * 100UL; + hwmgr->thermal_controller.fanInfo.ulMaxRPM = fan_table_v3->ucFanMaxRPM * 100UL; + phm_cap_set(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_MicrocodeFanControl); + hwmgr->thermal_controller.advanceFanControlParameters.usFanOutputSensitivity = + le16_to_cpu(fan_table_v3->usFanOutputSensitivity); + hwmgr->thermal_controller.advanceFanControlParameters.usMaxFanRPM = + fan_table_v3->ucFanMaxRPM * 100UL; + hwmgr->thermal_controller.advanceFanControlParameters.usFanRPMMaxLimit = + le16_to_cpu(fan_table_v3->usThrottlingRPM); + hwmgr->thermal_controller.advanceFanControlParameters.ulMinFanSCLKAcousticLimit = + le16_to_cpu(fan_table_v3->usFanAcousticLimitRpm); + hwmgr->thermal_controller.advanceFanControlParameters.usTMax = + le16_to_cpu(fan_table_v3->usTargetTemperature); + hwmgr->thermal_controller.advanceFanControlParameters.usPWMMin = + le16_to_cpu(fan_table_v3->usMinimumPWMLimit); + hwmgr->thermal_controller.advanceFanControlParameters.ulTargetGfxClk = + le16_to_cpu(fan_table_v3->usTargetGfxClk); + hwmgr->thermal_controller.advanceFanControlParameters.usFanGainEdge = + le16_to_cpu(fan_table_v3->usFanGainEdge); + hwmgr->thermal_controller.advanceFanControlParameters.usFanGainHotspot = + le16_to_cpu(fan_table_v3->usFanGainHotspot); + hwmgr->thermal_controller.advanceFanControlParameters.usFanGainLiquid = + le16_to_cpu(fan_table_v3->usFanGainLiquid); + hwmgr->thermal_controller.advanceFanControlParameters.usFanGainVrVddc = + le16_to_cpu(fan_table_v3->usFanGainVrVddc); + hwmgr->thermal_controller.advanceFanControlParameters.usFanGainVrMvdd = + le16_to_cpu(fan_table_v3->usFanGainVrMvdd); + hwmgr->thermal_controller.advanceFanControlParameters.usFanGainPlx = + le16_to_cpu(fan_table_v3->usFanGainPlx); + hwmgr->thermal_controller.advanceFanControlParameters.usFanGainHbm = + le16_to_cpu(fan_table_v3->usFanGainHbm); + + hwmgr->thermal_controller.advanceFanControlParameters.ucEnableZeroRPM = + fan_table_v3->ucEnableZeroRPM; + hwmgr->thermal_controller.advanceFanControlParameters.usZeroRPMStopTemperature = + le16_to_cpu(fan_table_v3->usFanStopTemperature); + hwmgr->thermal_controller.advanceFanControlParameters.usZeroRPMStartTemperature = + le16_to_cpu(fan_table_v3->usFanStartTemperature); + hwmgr->thermal_controller.advanceFanControlParameters.usMGpuThrottlingRPMLimit = + le16_to_cpu(fan_table_v3->usMGpuThrottlingRPM); } + return 0; } @@ -258,7 +307,26 @@ static int init_over_drive_limits( struct pp_hwmgr *hwmgr, const ATOM_Vega10_POWERPLAYTABLE *powerplay_table) { - hwmgr->platform_descriptor.overdriveLimit.engineClock = + const ATOM_Vega10_GFXCLK_Dependency_Table *gfxclk_dep_table = + (const ATOM_Vega10_GFXCLK_Dependency_Table *) + (((unsigned long) powerplay_table) + + le16_to_cpu(powerplay_table->usGfxclkDependencyTableOffset)); + bool is_acg_enabled = false; + ATOM_Vega10_GFXCLK_Dependency_Record_V2 *patom_record_v2; + + if (gfxclk_dep_table->ucRevId == 1) { + patom_record_v2 = + (ATOM_Vega10_GFXCLK_Dependency_Record_V2 *)gfxclk_dep_table->entries; + is_acg_enabled = + (bool)patom_record_v2[gfxclk_dep_table->ucNumEntries-1].ucACGEnable; + } + + if (powerplay_table->ulMaxODEngineClock > VEGA10_ENGINECLOCK_HARDMAX && + !is_acg_enabled) + hwmgr->platform_descriptor.overdriveLimit.engineClock = + VEGA10_ENGINECLOCK_HARDMAX; + else + hwmgr->platform_descriptor.overdriveLimit.engineClock = le32_to_cpu(powerplay_table->ulMaxODEngineClock); hwmgr->platform_descriptor.overdriveLimit.memoryClock = le32_to_cpu(powerplay_table->ulMaxODMemoryClock); diff --git a/drivers/gpu/drm/amd/powerplay/hwmgr/vega10_thermal.c b/drivers/gpu/drm/amd/powerplay/hwmgr/vega10_thermal.c index 3f807d6c95ce..ba8763daa380 100644 --- a/drivers/gpu/drm/amd/powerplay/hwmgr/vega10_thermal.c +++ b/drivers/gpu/drm/amd/powerplay/hwmgr/vega10_thermal.c @@ -556,6 +556,43 @@ int vega10_thermal_setup_fan_table(struct pp_hwmgr *hwmgr) return ret; } +int vega10_enable_mgpu_fan_boost(struct pp_hwmgr *hwmgr) +{ + struct vega10_hwmgr *data = hwmgr->backend; + PPTable_t *table = &(data->smc_state_table.pp_table); + int ret; + + if (!data->smu_features[GNLD_FAN_CONTROL].supported) + return 0; + + if (!hwmgr->thermal_controller.advanceFanControlParameters. + usMGpuThrottlingRPMLimit) + return 0; + + table->FanThrottlingRpm = hwmgr->thermal_controller. + advanceFanControlParameters.usMGpuThrottlingRPMLimit; + + ret = smum_smc_table_manager(hwmgr, + (uint8_t *)(&(data->smc_state_table.pp_table)), + PPTABLE, false); + if (ret) { + pr_info("Failed to update fan control table in pptable!"); + return ret; + } + + ret = vega10_disable_fan_control_feature(hwmgr); + if (ret) { + pr_info("Attempt to disable SMC fan control feature failed!"); + return ret; + } + + ret = vega10_enable_fan_control_feature(hwmgr); + if (ret) + pr_info("Attempt to enable SMC fan control feature failed!"); + + return ret; +} + /** * Start the fan control on the SMC. * @param hwmgr the address of the powerplay hardware manager. diff --git a/drivers/gpu/drm/amd/powerplay/hwmgr/vega10_thermal.h b/drivers/gpu/drm/amd/powerplay/hwmgr/vega10_thermal.h index 21e7c4dfa2ca..4a0ede7c1f07 100644 --- a/drivers/gpu/drm/amd/powerplay/hwmgr/vega10_thermal.h +++ b/drivers/gpu/drm/amd/powerplay/hwmgr/vega10_thermal.h @@ -73,6 +73,7 @@ extern int vega10_thermal_disable_alert(struct pp_hwmgr *hwmgr); extern int vega10_fan_ctrl_start_smc_fan_control(struct pp_hwmgr *hwmgr); extern int vega10_start_thermal_controller(struct pp_hwmgr *hwmgr, struct PP_TemperatureRange *range); +extern int vega10_enable_mgpu_fan_boost(struct pp_hwmgr *hwmgr); #endif diff --git a/drivers/gpu/drm/amd/powerplay/hwmgr/vega12_hwmgr.c b/drivers/gpu/drm/amd/powerplay/hwmgr/vega12_hwmgr.c index 54364444ecd1..bdb48e94eff6 100644 --- a/drivers/gpu/drm/amd/powerplay/hwmgr/vega12_hwmgr.c +++ b/drivers/gpu/drm/amd/powerplay/hwmgr/vega12_hwmgr.c @@ -753,6 +753,22 @@ static int vega12_init_smc_table(struct pp_hwmgr *hwmgr) return 0; } +static int vega12_run_acg_btc(struct pp_hwmgr *hwmgr) +{ + uint32_t result; + + PP_ASSERT_WITH_CODE( + smum_send_msg_to_smc(hwmgr, PPSMC_MSG_RunAcgBtc) == 0, + "[Run_ACG_BTC] Attempt to run ACG BTC failed!", + return -EINVAL); + + result = smum_get_argument(hwmgr); + PP_ASSERT_WITH_CODE(result == 1, + "Failed to run ACG BTC!", return -EINVAL); + + return 0; +} + static int vega12_set_allowed_featuresmask(struct pp_hwmgr *hwmgr) { struct vega12_hwmgr *data = @@ -931,6 +947,11 @@ static int vega12_enable_dpm_tasks(struct pp_hwmgr *hwmgr) "Failed to initialize SMC table!", result = tmp_result); + tmp_result = vega12_run_acg_btc(hwmgr); + PP_ASSERT_WITH_CODE(!tmp_result, + "Failed to run ACG BTC!", + result = tmp_result); + result = vega12_enable_all_smu_features(hwmgr); PP_ASSERT_WITH_CODE(!result, "Failed to enable all smu features!", @@ -1072,6 +1093,16 @@ static int vega12_upload_dpm_min_level(struct pp_hwmgr *hwmgr) return ret); } + if (data->smu_features[GNLD_DPM_DCEFCLK].enabled) { + min_freq = data->dpm_table.dcef_table.dpm_state.hard_min_level; + + PP_ASSERT_WITH_CODE(!(ret = smum_send_msg_to_smc_with_parameter( + hwmgr, PPSMC_MSG_SetHardMinByFreq, + (PPCLK_DCEFCLK << 16) | (min_freq & 0xffff))), + "Failed to set hard min dcefclk!", + return ret); + } + return ret; } @@ -1797,7 +1828,7 @@ static int vega12_force_clock_level(struct pp_hwmgr *hwmgr, enum pp_clock_type type, uint32_t mask) { struct vega12_hwmgr *data = (struct vega12_hwmgr *)(hwmgr->backend); - uint32_t soft_min_level, soft_max_level; + uint32_t soft_min_level, soft_max_level, hard_min_level; int ret = 0; switch (type) { @@ -1842,6 +1873,56 @@ static int vega12_force_clock_level(struct pp_hwmgr *hwmgr, break; + case PP_SOCCLK: + soft_min_level = mask ? (ffs(mask) - 1) : 0; + soft_max_level = mask ? (fls(mask) - 1) : 0; + + if (soft_max_level >= data->dpm_table.soc_table.count) { + pr_err("Clock level specified %d is over max allowed %d\n", + soft_max_level, + data->dpm_table.soc_table.count - 1); + return -EINVAL; + } + + data->dpm_table.soc_table.dpm_state.soft_min_level = + data->dpm_table.soc_table.dpm_levels[soft_min_level].value; + data->dpm_table.soc_table.dpm_state.soft_max_level = + data->dpm_table.soc_table.dpm_levels[soft_max_level].value; + + ret = vega12_upload_dpm_min_level(hwmgr); + PP_ASSERT_WITH_CODE(!ret, + "Failed to upload boot level to lowest!", + return ret); + + ret = vega12_upload_dpm_max_level(hwmgr); + PP_ASSERT_WITH_CODE(!ret, + "Failed to upload dpm max level to highest!", + return ret); + + break; + + case PP_DCEFCLK: + hard_min_level = mask ? (ffs(mask) - 1) : 0; + + if (hard_min_level >= data->dpm_table.dcef_table.count) { + pr_err("Clock level specified %d is over max allowed %d\n", + hard_min_level, + data->dpm_table.dcef_table.count - 1); + return -EINVAL; + } + + data->dpm_table.dcef_table.dpm_state.hard_min_level = + data->dpm_table.dcef_table.dpm_levels[hard_min_level].value; + + ret = vega12_upload_dpm_min_level(hwmgr); + PP_ASSERT_WITH_CODE(!ret, + "Failed to upload boot level to lowest!", + return ret); + + //TODO: Setting DCEFCLK max dpm level is not supported + + break; + case PP_PCIE: break; @@ -1852,6 +1933,104 @@ static int vega12_force_clock_level(struct pp_hwmgr *hwmgr, return 0; } +static int vega12_get_ppfeature_status(struct pp_hwmgr *hwmgr, char *buf) +{ + static const char *ppfeature_name[] = { + "DPM_PREFETCHER", + "GFXCLK_DPM", + "UCLK_DPM", + "SOCCLK_DPM", + "UVD_DPM", + "VCE_DPM", + "ULV", + "MP0CLK_DPM", + "LINK_DPM", + "DCEFCLK_DPM", + "GFXCLK_DS", + "SOCCLK_DS", + "LCLK_DS", + "PPT", + "TDC", + "THERMAL", + "GFX_PER_CU_CG", + "RM", + "DCEFCLK_DS", + "ACDC", + "VR0HOT", + "VR1HOT", + "FW_CTF", + "LED_DISPLAY", + "FAN_CONTROL", + "DIDT", + "GFXOFF", + "CG", + "ACG"}; + static const char *output_title[] = { + "FEATURES", + "BITMASK", + "ENABLEMENT"}; + uint64_t features_enabled; + int i; + int ret = 0; + int size = 0; + + ret = vega12_get_enabled_smc_features(hwmgr, &features_enabled); + PP_ASSERT_WITH_CODE(!ret, + "[EnableAllSmuFeatures] Failed to get enabled smc features!", + return ret); + + size += sprintf(buf + size, "Current ppfeatures: 0x%016llx\n", features_enabled); + size += sprintf(buf + size, "%-19s %-22s %s\n", + output_title[0], + output_title[1], + output_title[2]); + for (i = 0; i < GNLD_FEATURES_MAX; i++) { + size += sprintf(buf + size, "%-19s 0x%016llx %6s\n", + ppfeature_name[i], + 1ULL << i, + (features_enabled & (1ULL << i)) ? "Y" : "N"); + } + + return size; +} + +static int vega12_set_ppfeature_status(struct pp_hwmgr *hwmgr, uint64_t new_ppfeature_masks) +{ + uint64_t features_enabled; + uint64_t features_to_enable; + uint64_t features_to_disable; + int ret = 0; + + if (new_ppfeature_masks >= (1ULL << GNLD_FEATURES_MAX)) + return -EINVAL; + + ret = vega12_get_enabled_smc_features(hwmgr, &features_enabled); + if (ret) + return ret; + + features_to_disable = + features_enabled & ~new_ppfeature_masks; + features_to_enable = + ~features_enabled & new_ppfeature_masks; + + pr_debug("features_to_disable 0x%llx\n", features_to_disable); + pr_debug("features_to_enable 0x%llx\n", features_to_enable); + + if (features_to_disable) { + ret = vega12_enable_smc_features(hwmgr, false, features_to_disable); + if (ret) + return ret; + } + + if (features_to_enable) { + ret = vega12_enable_smc_features(hwmgr, true, features_to_enable); + if (ret) + return ret; + } + + return 0; +} + static int vega12_print_clock_levels(struct pp_hwmgr *hwmgr, enum pp_clock_type type, char *buf) { @@ -1891,6 +2070,42 @@ static int vega12_print_clock_levels(struct pp_hwmgr *hwmgr, (clocks.data[i].clocks_in_khz / 1000 == now / 100) ? "*" : ""); break; + case PP_SOCCLK: + PP_ASSERT_WITH_CODE( + smum_send_msg_to_smc_with_parameter(hwmgr, + PPSMC_MSG_GetDpmClockFreq, (PPCLK_SOCCLK << 16)) == 0, + "Attempt to get Current SOCCLK Frequency Failed!", + return -EINVAL); + now = smum_get_argument(hwmgr); + + PP_ASSERT_WITH_CODE( + vega12_get_socclocks(hwmgr, &clocks) == 0, + "Attempt to get soc clk levels Failed!", + return -1); + for (i = 0; i < clocks.num_levels; i++) + size += sprintf(buf + size, "%d: %uMhz %s\n", + i, clocks.data[i].clocks_in_khz / 1000, + (clocks.data[i].clocks_in_khz / 1000 == now) ? "*" : ""); + break; + + case PP_DCEFCLK: + PP_ASSERT_WITH_CODE( + smum_send_msg_to_smc_with_parameter(hwmgr, + PPSMC_MSG_GetDpmClockFreq, (PPCLK_DCEFCLK << 16)) == 0, + "Attempt to get Current DCEFCLK Frequency Failed!", + return -EINVAL); + now = smum_get_argument(hwmgr); + + PP_ASSERT_WITH_CODE( + vega12_get_dcefclocks(hwmgr, &clocks) == 0, + "Attempt to get dcef clk levels Failed!", + return -1); + for (i = 0; i < clocks.num_levels; i++) + size += sprintf(buf + size, "%d: %uMhz %s\n", + i, clocks.data[i].clocks_in_khz / 1000, + (clocks.data[i].clocks_in_khz / 1000 == now) ? "*" : ""); + break; + case PP_PCIE: break; @@ -2411,6 +2626,8 @@ static const struct pp_hwmgr_func vega12_hwmgr_funcs = { .start_thermal_controller = vega12_start_thermal_controller, .powergate_gfx = vega12_gfx_off_control, .get_performance_level = vega12_get_performance_level, + .get_ppfeature_status = vega12_get_ppfeature_status, + .set_ppfeature_status = vega12_set_ppfeature_status, }; int vega12_hwmgr_init(struct pp_hwmgr *hwmgr) diff --git a/drivers/gpu/drm/amd/powerplay/hwmgr/vega20_baco.c b/drivers/gpu/drm/amd/powerplay/hwmgr/vega20_baco.c new file mode 100644 index 000000000000..5e8602a79b1c --- /dev/null +++ b/drivers/gpu/drm/amd/powerplay/hwmgr/vega20_baco.c @@ -0,0 +1,103 @@ +/* + * Copyright 2018 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 "soc15.h" +#include "soc15_hw_ip.h" +#include "soc15_common.h" +#include "vega20_inc.h" +#include "vega20_ppsmc.h" +#include "vega20_baco.h" + + + +static const struct soc15_baco_cmd_entry clean_baco_tbl[] = +{ + {CMD_WRITE, SOC15_REG_ENTRY(NBIF, 0, mmBIOS_SCRATCH_6), 0, 0, 0, 0}, + {CMD_WRITE, SOC15_REG_ENTRY(NBIF, 0, mmBIOS_SCRATCH_7), 0, 0, 0, 0}, +}; + +int vega20_baco_get_capability(struct pp_hwmgr *hwmgr, bool *cap) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)(hwmgr->adev); + uint32_t reg; + + *cap = false; + if (!phm_cap_enabled(hwmgr->platform_descriptor.platformCaps, PHM_PlatformCaps_BACO)) + return 0; + + if (((RREG32(0x17569) & 0x20000000) >> 29) == 0x1) { + reg = RREG32_SOC15(NBIF, 0, mmRCC_BIF_STRAP0); + + if (reg & RCC_BIF_STRAP0__STRAP_PX_CAPABLE_MASK) + *cap = true; + } + + return 0; +} + +int vega20_baco_get_state(struct pp_hwmgr *hwmgr, enum BACO_STATE *state) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)(hwmgr->adev); + uint32_t reg; + + reg = RREG32_SOC15(NBIF, 0, mmBACO_CNTL); + + if (reg & BACO_CNTL__BACO_MODE_MASK) + /* gfx has already entered BACO state */ + *state = BACO_STATE_IN; + else + *state = BACO_STATE_OUT; + return 0; +} + +int vega20_baco_set_state(struct pp_hwmgr *hwmgr, enum BACO_STATE state) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)(hwmgr->adev); + enum BACO_STATE cur_state; + uint32_t data; + + vega20_baco_get_state(hwmgr, &cur_state); + + if (cur_state == state) + /* aisc already in the target state */ + return 0; + + if (state == BACO_STATE_IN) { + data = RREG32_SOC15(THM, 0, mmTHM_BACO_CNTL); + data |= 0x80000000; + WREG32_SOC15(THM, 0, mmTHM_BACO_CNTL, data); + + + if(smum_send_msg_to_smc_with_parameter(hwmgr, PPSMC_MSG_EnterBaco, 0)) + return -EINVAL; + + } else if (state == BACO_STATE_OUT) { + if (smum_send_msg_to_smc(hwmgr, PPSMC_MSG_ExitBaco)) + return -EINVAL; + if (!soc15_baco_program_registers(hwmgr, clean_baco_tbl, + ARRAY_SIZE(clean_baco_tbl))) + return -EINVAL; + } + + return 0; +} diff --git a/drivers/gpu/drm/amd/display/dc/i2caux/dce100/i2caux_dce100.h b/drivers/gpu/drm/amd/powerplay/hwmgr/vega20_baco.h index 2b508d3e0ef4..51c7f8392925 100644 --- a/drivers/gpu/drm/amd/display/dc/i2caux/dce100/i2caux_dce100.h +++ b/drivers/gpu/drm/amd/powerplay/hwmgr/vega20_baco.h @@ -1,5 +1,5 @@ /* - * Copyright 2012-15 Advanced Micro Devices, Inc. + * Copyright 2018 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"), @@ -19,14 +19,14 @@ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR * OTHER DEALINGS IN THE SOFTWARE. * - * Authors: AMD - * */ +#ifndef __VEGA20_BACO_H__ +#define __VEGA20_BACO_H__ +#include "hwmgr.h" +#include "common_baco.h" -#ifndef __DAL_I2C_AUX_DCE100_H__ -#define __DAL_I2C_AUX_DCE100_H__ - -struct i2caux *dal_i2caux_dce100_create( - struct dc_context *ctx); +extern int vega20_baco_get_capability(struct pp_hwmgr *hwmgr, bool *cap); +extern int vega20_baco_get_state(struct pp_hwmgr *hwmgr, enum BACO_STATE *state); +extern int vega20_baco_set_state(struct pp_hwmgr *hwmgr, enum BACO_STATE state); -#endif /* __DAL_I2C_AUX_DCE100_H__ */ +#endif diff --git a/drivers/gpu/drm/amd/powerplay/hwmgr/vega20_hwmgr.c b/drivers/gpu/drm/amd/powerplay/hwmgr/vega20_hwmgr.c index 3b7fce5d7258..23b5b94a4939 100644 --- a/drivers/gpu/drm/amd/powerplay/hwmgr/vega20_hwmgr.c +++ b/drivers/gpu/drm/amd/powerplay/hwmgr/vega20_hwmgr.c @@ -47,8 +47,13 @@ #include "pp_overdriver.h" #include "pp_thermal.h" #include "soc15_common.h" +#include "vega20_baco.h" #include "smuio/smuio_9_0_offset.h" #include "smuio/smuio_9_0_sh_mask.h" +#include "nbio/nbio_7_4_sh_mask.h" + +#define smnPCIE_LC_SPEED_CNTL 0x11140290 +#define smnPCIE_LC_LINK_WIDTH_CNTL 0x11140288 static void vega20_set_default_registry_data(struct pp_hwmgr *hwmgr) { @@ -86,6 +91,12 @@ static void vega20_set_default_registry_data(struct pp_hwmgr *hwmgr) * MP0CLK DS */ data->registry_data.disallowed_features = 0xE0041C00; + /* ECC feature should be disabled on old SMUs */ + smum_send_msg_to_smc(hwmgr, PPSMC_MSG_GetSmuVersion); + hwmgr->smu_version = smum_get_argument(hwmgr); + if (hwmgr->smu_version < 0x282100) + data->registry_data.disallowed_features |= FEATURE_ECC_MASK; + data->registry_data.od_state_in_dc_support = 0; data->registry_data.thermal_support = 1; data->registry_data.skip_baco_hardware = 0; @@ -352,6 +363,7 @@ static void vega20_init_dpm_defaults(struct pp_hwmgr *hwmgr) data->smu_features[GNLD_DS_MP1CLK].smu_feature_id = FEATURE_DS_MP1CLK_BIT; data->smu_features[GNLD_DS_MP0CLK].smu_feature_id = FEATURE_DS_MP0CLK_BIT; data->smu_features[GNLD_XGMI].smu_feature_id = FEATURE_XGMI_BIT; + data->smu_features[GNLD_ECC].smu_feature_id = FEATURE_ECC_BIT; for (i = 0; i < GNLD_FEATURES_MAX; i++) { data->smu_features[i].smu_feature_bitmap = @@ -386,9 +398,9 @@ static int vega20_hwmgr_backend_init(struct pp_hwmgr *hwmgr) hwmgr->backend = data; - hwmgr->workload_mask = 1 << hwmgr->workload_prority[PP_SMC_POWER_PROFILE_VIDEO]; - hwmgr->power_profile_mode = PP_SMC_POWER_PROFILE_VIDEO; - hwmgr->default_power_profile_mode = PP_SMC_POWER_PROFILE_VIDEO; + hwmgr->workload_mask = 1 << hwmgr->workload_prority[PP_SMC_POWER_PROFILE_BOOTUP_DEFAULT]; + hwmgr->power_profile_mode = PP_SMC_POWER_PROFILE_BOOTUP_DEFAULT; + hwmgr->default_power_profile_mode = PP_SMC_POWER_PROFILE_BOOTUP_DEFAULT; vega20_set_default_registry_data(hwmgr); @@ -458,9 +470,9 @@ static int vega20_setup_asic_task(struct pp_hwmgr *hwmgr) static void vega20_init_dpm_state(struct vega20_dpm_state *dpm_state) { dpm_state->soft_min_level = 0x0; - dpm_state->soft_max_level = 0xffff; + dpm_state->soft_max_level = VG20_CLOCK_MAX_DEFAULT; dpm_state->hard_min_level = 0x0; - dpm_state->hard_max_level = 0xffff; + dpm_state->hard_max_level = VG20_CLOCK_MAX_DEFAULT; } static int vega20_get_number_of_dpm_level(struct pp_hwmgr *hwmgr, @@ -706,8 +718,10 @@ static int vega20_setup_default_dpm_tables(struct pp_hwmgr *hwmgr) PP_ASSERT_WITH_CODE(!ret, "[SetupDefaultDpmTable] failed to get fclk dpm levels!", return ret); - } else - dpm_table->count = 0; + } else { + dpm_table->count = 1; + dpm_table->dpm_levels[0].value = data->vbios_boot_state.fclock / 100; + } vega20_init_dpm_state(&(dpm_table->dpm_state)); /* save a copy of the default DPM table */ @@ -749,6 +763,7 @@ static int vega20_init_smc_table(struct pp_hwmgr *hwmgr) data->vbios_boot_state.eclock = boot_up_values.ulEClk; data->vbios_boot_state.vclock = boot_up_values.ulVClk; data->vbios_boot_state.dclock = boot_up_values.ulDClk; + data->vbios_boot_state.fclock = boot_up_values.ulFClk; data->vbios_boot_state.uc_cooling_id = boot_up_values.ucCoolingID; smum_send_msg_to_smc_with_parameter(hwmgr, @@ -766,6 +781,60 @@ static int vega20_init_smc_table(struct pp_hwmgr *hwmgr) return 0; } +/* + * Override PCIe link speed and link width for DPM Level 1. PPTable entries + * reflect the ASIC capabilities and not the system capabilities. For e.g. + * Vega20 board in a PCI Gen3 system. In this case, when SMU's tries to switch + * to DPM1, it fails as system doesn't support Gen4. + */ +static int vega20_override_pcie_parameters(struct pp_hwmgr *hwmgr) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)(hwmgr->adev); + struct vega20_hwmgr *data = + (struct vega20_hwmgr *)(hwmgr->backend); + uint32_t pcie_gen = 0, pcie_width = 0, smu_pcie_arg; + int ret; + + if (adev->pm.pcie_gen_mask & CAIL_PCIE_LINK_SPEED_SUPPORT_GEN4) + pcie_gen = 3; + else if (adev->pm.pcie_gen_mask & CAIL_PCIE_LINK_SPEED_SUPPORT_GEN3) + pcie_gen = 2; + else if (adev->pm.pcie_gen_mask & CAIL_PCIE_LINK_SPEED_SUPPORT_GEN2) + pcie_gen = 1; + else if (adev->pm.pcie_gen_mask & CAIL_PCIE_LINK_SPEED_SUPPORT_GEN1) + pcie_gen = 0; + + if (adev->pm.pcie_mlw_mask & CAIL_PCIE_LINK_WIDTH_SUPPORT_X16) + pcie_width = 6; + else if (adev->pm.pcie_mlw_mask & CAIL_PCIE_LINK_WIDTH_SUPPORT_X12) + pcie_width = 5; + else if (adev->pm.pcie_mlw_mask & CAIL_PCIE_LINK_WIDTH_SUPPORT_X8) + pcie_width = 4; + else if (adev->pm.pcie_mlw_mask & CAIL_PCIE_LINK_WIDTH_SUPPORT_X4) + pcie_width = 3; + else if (adev->pm.pcie_mlw_mask & CAIL_PCIE_LINK_WIDTH_SUPPORT_X2) + pcie_width = 2; + else if (adev->pm.pcie_mlw_mask & CAIL_PCIE_LINK_WIDTH_SUPPORT_X1) + pcie_width = 1; + + /* Bit 31:16: LCLK DPM level. 0 is DPM0, and 1 is DPM1 + * Bit 15:8: PCIE GEN, 0 to 3 corresponds to GEN1 to GEN4 + * Bit 7:0: PCIE lane width, 1 to 7 corresponds is x1 to x32 + */ + smu_pcie_arg = (1 << 16) | (pcie_gen << 8) | pcie_width; + ret = smum_send_msg_to_smc_with_parameter(hwmgr, + PPSMC_MSG_OverridePcieParameters, smu_pcie_arg); + PP_ASSERT_WITH_CODE(!ret, + "[OverridePcieParameters] Attempt to override pcie params failed!", + return ret); + + data->pcie_parameters_override = 1; + data->pcie_gen_level1 = pcie_gen; + data->pcie_width_level1 = pcie_width; + + return 0; +} + static int vega20_set_allowed_featuresmask(struct pp_hwmgr *hwmgr) { struct vega20_hwmgr *data = @@ -799,6 +868,11 @@ static int vega20_set_allowed_featuresmask(struct pp_hwmgr *hwmgr) return 0; } +static int vega20_run_btc(struct pp_hwmgr *hwmgr) +{ + return smum_send_msg_to_smc(hwmgr, PPSMC_MSG_RunBtc); +} + static int vega20_run_btc_afll(struct pp_hwmgr *hwmgr) { return smum_send_msg_to_smc(hwmgr, PPSMC_MSG_RunAfllBtc); @@ -921,6 +995,8 @@ static int vega20_od8_set_feature_capabilities( } if (data->smu_features[GNLD_DPM_UCLK].enabled) { + pptable_information->od_settings_min[OD8_SETTING_UCLK_FMAX] = + data->dpm_table.mem_table.dpm_levels[data->dpm_table.mem_table.count - 2].value; if (pptable_information->od_feature_capabilities[ATOM_VEGA20_ODFEATURE_UCLK_MAX] && pptable_information->od_settings_min[OD8_SETTING_UCLK_FMAX] > 0 && pptable_information->od_settings_max[OD8_SETTING_UCLK_FMAX] > 0 && @@ -976,6 +1052,9 @@ static int vega20_od8_set_feature_capabilities( pp_table->FanZeroRpmEnable) od_settings->overdrive8_capabilities |= OD8_FAN_ZERO_RPM_CONTROL; + if (!od_settings->overdrive8_capabilities) + hwmgr->od_enabled = false; + return 0; } @@ -1557,6 +1636,11 @@ static int vega20_enable_dpm_tasks(struct pp_hwmgr *hwmgr) "[EnableDPMTasks] Failed to initialize SMC table!", return result); + result = vega20_run_btc(hwmgr); + PP_ASSERT_WITH_CODE(!result, + "[EnableDPMTasks] Failed to run btc!", + return result); + result = vega20_run_btc_afll(hwmgr); PP_ASSERT_WITH_CODE(!result, "[EnableDPMTasks] Failed to run btc afll!", @@ -1567,6 +1651,11 @@ static int vega20_enable_dpm_tasks(struct pp_hwmgr *hwmgr) "[EnableDPMTasks] Failed to enable all smu features!", return result); + result = vega20_override_pcie_parameters(hwmgr); + PP_ASSERT_WITH_CODE(!result, + "[EnableDPMTasks] Failed to override pcie parameters!", + return result); + result = vega20_notify_smc_display_change(hwmgr); PP_ASSERT_WITH_CODE(!result, "[EnableDPMTasks] Failed to notify smc display change!", @@ -1685,13 +1774,6 @@ static int vega20_upload_dpm_min_level(struct pp_hwmgr *hwmgr, uint32_t feature_ (PPCLK_UCLK << 16) | (min_freq & 0xffff))), "Failed to set soft min memclk !", return ret); - - min_freq = data->dpm_table.mem_table.dpm_state.hard_min_level; - PP_ASSERT_WITH_CODE(!(ret = smum_send_msg_to_smc_with_parameter( - hwmgr, PPSMC_MSG_SetHardMinByFreq, - (PPCLK_UCLK << 16) | (min_freq & 0xffff))), - "Failed to set hard min memclk !", - return ret); } if (data->smu_features[GNLD_DPM_UVD].enabled && @@ -1735,6 +1817,28 @@ static int vega20_upload_dpm_min_level(struct pp_hwmgr *hwmgr, uint32_t feature_ return ret); } + if (data->smu_features[GNLD_DPM_FCLK].enabled && + (feature_mask & FEATURE_DPM_FCLK_MASK)) { + min_freq = data->dpm_table.fclk_table.dpm_state.soft_min_level; + + PP_ASSERT_WITH_CODE(!(ret = smum_send_msg_to_smc_with_parameter( + hwmgr, PPSMC_MSG_SetSoftMinByFreq, + (PPCLK_FCLK << 16) | (min_freq & 0xffff))), + "Failed to set soft min fclk!", + return ret); + } + + if (data->smu_features[GNLD_DPM_DCEFCLK].enabled && + (feature_mask & FEATURE_DPM_DCEFCLK_MASK)) { + min_freq = data->dpm_table.dcef_table.dpm_state.hard_min_level; + + PP_ASSERT_WITH_CODE(!(ret = smum_send_msg_to_smc_with_parameter( + hwmgr, PPSMC_MSG_SetHardMinByFreq, + (PPCLK_DCEFCLK << 16) | (min_freq & 0xffff))), + "Failed to set hard min dcefclk!", + return ret); + } + return ret; } @@ -1807,6 +1911,17 @@ static int vega20_upload_dpm_max_level(struct pp_hwmgr *hwmgr, uint32_t feature_ return ret); } + if (data->smu_features[GNLD_DPM_FCLK].enabled && + (feature_mask & FEATURE_DPM_FCLK_MASK)) { + max_freq = data->dpm_table.fclk_table.dpm_state.soft_max_level; + + PP_ASSERT_WITH_CODE(!(ret = smum_send_msg_to_smc_with_parameter( + hwmgr, PPSMC_MSG_SetSoftMaxByFreq, + (PPCLK_FCLK << 16) | (max_freq & 0xffff))), + "Failed to set soft max fclk!", + return ret); + } + return ret; } @@ -1914,16 +2029,36 @@ static uint32_t vega20_dpm_get_mclk(struct pp_hwmgr *hwmgr, bool low) return (mem_clk * 100); } +static int vega20_get_metrics_table(struct pp_hwmgr *hwmgr, SmuMetrics_t *metrics_table) +{ + struct vega20_hwmgr *data = + (struct vega20_hwmgr *)(hwmgr->backend); + int ret = 0; + + if (!data->metrics_time || time_after(jiffies, data->metrics_time + HZ / 2)) { + ret = smum_smc_table_manager(hwmgr, (uint8_t *)metrics_table, + TABLE_SMU_METRICS, true); + if (ret) { + pr_info("Failed to export SMU metrics table!\n"); + return ret; + } + memcpy(&data->metrics_table, metrics_table, sizeof(SmuMetrics_t)); + data->metrics_time = jiffies; + } else + memcpy(metrics_table, &data->metrics_table, sizeof(SmuMetrics_t)); + + return ret; +} + static int vega20_get_gpu_power(struct pp_hwmgr *hwmgr, uint32_t *query) { int ret = 0; SmuMetrics_t metrics_table; - ret = smum_smc_table_manager(hwmgr, (uint8_t *)&metrics_table, TABLE_SMU_METRICS, true); - PP_ASSERT_WITH_CODE(!ret, - "Failed to export SMU METRICS table!", - return ret); + ret = vega20_get_metrics_table(hwmgr, &metrics_table); + if (ret) + return ret; *query = metrics_table.CurrSocketPower << 8; @@ -1954,10 +2089,9 @@ static int vega20_get_current_activity_percent(struct pp_hwmgr *hwmgr, int ret = 0; SmuMetrics_t metrics_table; - ret = smum_smc_table_manager(hwmgr, (uint8_t *)&metrics_table, TABLE_SMU_METRICS, true); - PP_ASSERT_WITH_CODE(!ret, - "Failed to export SMU METRICS table!", - return ret); + ret = vega20_get_metrics_table(hwmgr, &metrics_table); + if (ret) + return ret; *activity_percent = metrics_table.AverageGfxActivity; @@ -1969,16 +2103,18 @@ static int vega20_read_sensor(struct pp_hwmgr *hwmgr, int idx, { struct vega20_hwmgr *data = (struct vega20_hwmgr *)(hwmgr->backend); struct amdgpu_device *adev = hwmgr->adev; + SmuMetrics_t metrics_table; uint32_t val_vid; int ret = 0; switch (idx) { case AMDGPU_PP_SENSOR_GFX_SCLK: - ret = vega20_get_current_clk_freq(hwmgr, - PPCLK_GFXCLK, - (uint32_t *)value); - if (!ret) - *size = 4; + ret = vega20_get_metrics_table(hwmgr, &metrics_table); + if (ret) + return ret; + + *((uint32_t *)value) = metrics_table.AverageGfxclkFrequency * 100; + *size = 4; break; case AMDGPU_PP_SENSOR_GFX_MCLK: ret = vega20_get_current_clk_freq(hwmgr, @@ -2136,6 +2272,12 @@ static int vega20_force_dpm_highest(struct pp_hwmgr *hwmgr) data->dpm_table.mem_table.dpm_state.soft_max_level = data->dpm_table.mem_table.dpm_levels[soft_level].value; + soft_level = vega20_find_highest_dpm_level(&(data->dpm_table.soc_table)); + + data->dpm_table.soc_table.dpm_state.soft_min_level = + data->dpm_table.soc_table.dpm_state.soft_max_level = + data->dpm_table.soc_table.dpm_levels[soft_level].value; + ret = vega20_upload_dpm_min_level(hwmgr, 0xFFFFFFFF); PP_ASSERT_WITH_CODE(!ret, "Failed to upload boot level to highest!", @@ -2168,6 +2310,12 @@ static int vega20_force_dpm_lowest(struct pp_hwmgr *hwmgr) data->dpm_table.mem_table.dpm_state.soft_max_level = data->dpm_table.mem_table.dpm_levels[soft_level].value; + soft_level = vega20_find_lowest_dpm_level(&(data->dpm_table.soc_table)); + + data->dpm_table.soc_table.dpm_state.soft_min_level = + data->dpm_table.soc_table.dpm_state.soft_max_level = + data->dpm_table.soc_table.dpm_levels[soft_level].value; + ret = vega20_upload_dpm_min_level(hwmgr, 0xFFFFFFFF); PP_ASSERT_WITH_CODE(!ret, "Failed to upload boot level to highest!", @@ -2236,7 +2384,7 @@ static int vega20_force_clock_level(struct pp_hwmgr *hwmgr, enum pp_clock_type type, uint32_t mask) { struct vega20_hwmgr *data = (struct vega20_hwmgr *)(hwmgr->backend); - uint32_t soft_min_level, soft_max_level; + uint32_t soft_min_level, soft_max_level, hard_min_level; int ret = 0; switch (type) { @@ -2244,6 +2392,13 @@ static int vega20_force_clock_level(struct pp_hwmgr *hwmgr, soft_min_level = mask ? (ffs(mask) - 1) : 0; soft_max_level = mask ? (fls(mask) - 1) : 0; + if (soft_max_level >= data->dpm_table.gfx_table.count) { + pr_err("Clock level specified %d is over max allowed %d\n", + soft_max_level, + data->dpm_table.gfx_table.count - 1); + return -EINVAL; + } + data->dpm_table.gfx_table.dpm_state.soft_min_level = data->dpm_table.gfx_table.dpm_levels[soft_min_level].value; data->dpm_table.gfx_table.dpm_state.soft_max_level = @@ -2264,6 +2419,13 @@ static int vega20_force_clock_level(struct pp_hwmgr *hwmgr, soft_min_level = mask ? (ffs(mask) - 1) : 0; soft_max_level = mask ? (fls(mask) - 1) : 0; + if (soft_max_level >= data->dpm_table.mem_table.count) { + pr_err("Clock level specified %d is over max allowed %d\n", + soft_max_level, + data->dpm_table.mem_table.count - 1); + return -EINVAL; + } + data->dpm_table.mem_table.dpm_state.soft_min_level = data->dpm_table.mem_table.dpm_levels[soft_min_level].value; data->dpm_table.mem_table.dpm_state.soft_max_level = @@ -2281,7 +2443,97 @@ static int vega20_force_clock_level(struct pp_hwmgr *hwmgr, break; + case PP_SOCCLK: + soft_min_level = mask ? (ffs(mask) - 1) : 0; + soft_max_level = mask ? (fls(mask) - 1) : 0; + + if (soft_max_level >= data->dpm_table.soc_table.count) { + pr_err("Clock level specified %d is over max allowed %d\n", + soft_max_level, + data->dpm_table.soc_table.count - 1); + return -EINVAL; + } + + data->dpm_table.soc_table.dpm_state.soft_min_level = + data->dpm_table.soc_table.dpm_levels[soft_min_level].value; + data->dpm_table.soc_table.dpm_state.soft_max_level = + data->dpm_table.soc_table.dpm_levels[soft_max_level].value; + + ret = vega20_upload_dpm_min_level(hwmgr, FEATURE_DPM_SOCCLK_MASK); + PP_ASSERT_WITH_CODE(!ret, + "Failed to upload boot level to lowest!", + return ret); + + ret = vega20_upload_dpm_max_level(hwmgr, FEATURE_DPM_SOCCLK_MASK); + PP_ASSERT_WITH_CODE(!ret, + "Failed to upload dpm max level to highest!", + return ret); + + break; + + case PP_FCLK: + soft_min_level = mask ? (ffs(mask) - 1) : 0; + soft_max_level = mask ? (fls(mask) - 1) : 0; + + if (soft_max_level >= data->dpm_table.fclk_table.count) { + pr_err("Clock level specified %d is over max allowed %d\n", + soft_max_level, + data->dpm_table.fclk_table.count - 1); + return -EINVAL; + } + + data->dpm_table.fclk_table.dpm_state.soft_min_level = + data->dpm_table.fclk_table.dpm_levels[soft_min_level].value; + data->dpm_table.fclk_table.dpm_state.soft_max_level = + data->dpm_table.fclk_table.dpm_levels[soft_max_level].value; + + ret = vega20_upload_dpm_min_level(hwmgr, FEATURE_DPM_FCLK_MASK); + PP_ASSERT_WITH_CODE(!ret, + "Failed to upload boot level to lowest!", + return ret); + + ret = vega20_upload_dpm_max_level(hwmgr, FEATURE_DPM_FCLK_MASK); + PP_ASSERT_WITH_CODE(!ret, + "Failed to upload dpm max level to highest!", + return ret); + + break; + + case PP_DCEFCLK: + hard_min_level = mask ? (ffs(mask) - 1) : 0; + + if (hard_min_level >= data->dpm_table.dcef_table.count) { + pr_err("Clock level specified %d is over max allowed %d\n", + hard_min_level, + data->dpm_table.dcef_table.count - 1); + return -EINVAL; + } + + data->dpm_table.dcef_table.dpm_state.hard_min_level = + data->dpm_table.dcef_table.dpm_levels[hard_min_level].value; + + ret = vega20_upload_dpm_min_level(hwmgr, FEATURE_DPM_DCEFCLK_MASK); + PP_ASSERT_WITH_CODE(!ret, + "Failed to upload boot level to lowest!", + return ret); + + //TODO: Setting DCEFCLK max dpm level is not supported + + break; + case PP_PCIE: + soft_min_level = mask ? (ffs(mask) - 1) : 0; + soft_max_level = mask ? (fls(mask) - 1) : 0; + if (soft_min_level >= NUM_LINK_LEVELS || + soft_max_level >= NUM_LINK_LEVELS) + return -EINVAL; + + ret = smum_send_msg_to_smc_with_parameter(hwmgr, + PPSMC_MSG_SetMinLinkDpmByIndex, soft_min_level); + PP_ASSERT_WITH_CODE(!ret, + "Failed to set min link dpm level!", + return ret); + break; default: @@ -2319,6 +2571,7 @@ static int vega20_dpm_force_dpm_level(struct pp_hwmgr *hwmgr, return ret; vega20_force_clock_level(hwmgr, PP_SCLK, 1 << sclk_mask); vega20_force_clock_level(hwmgr, PP_MCLK, 1 << mclk_mask); + vega20_force_clock_level(hwmgr, PP_SOCCLK, 1 << soc_mask); break; case AMD_DPM_FORCED_LEVEL_MANUAL: @@ -2382,9 +2635,8 @@ static int vega20_get_sclks(struct pp_hwmgr *hwmgr, struct vega20_single_dpm_table *dpm_table = &(data->dpm_table.gfx_table); int i, count; - PP_ASSERT_WITH_CODE(data->smu_features[GNLD_DPM_GFXCLK].enabled, - "[GetSclks]: gfxclk dpm not enabled!\n", - return -EPERM); + if (!data->smu_features[GNLD_DPM_GFXCLK].enabled) + return -1; count = (dpm_table->count > MAX_NUM_CLOCKS) ? MAX_NUM_CLOCKS : dpm_table->count; clocks->num_levels = count; @@ -2411,9 +2663,8 @@ static int vega20_get_memclocks(struct pp_hwmgr *hwmgr, struct vega20_single_dpm_table *dpm_table = &(data->dpm_table.mem_table); int i, count; - PP_ASSERT_WITH_CODE(data->smu_features[GNLD_DPM_UCLK].enabled, - "[GetMclks]: uclk dpm not enabled!\n", - return -EPERM); + if (!data->smu_features[GNLD_DPM_UCLK].enabled) + return -1; count = (dpm_table->count > MAX_NUM_CLOCKS) ? MAX_NUM_CLOCKS : dpm_table->count; clocks->num_levels = data->mclk_latency_table.count = count; @@ -2437,9 +2688,8 @@ static int vega20_get_dcefclocks(struct pp_hwmgr *hwmgr, struct vega20_single_dpm_table *dpm_table = &(data->dpm_table.dcef_table); int i, count; - PP_ASSERT_WITH_CODE(data->smu_features[GNLD_DPM_DCEFCLK].enabled, - "[GetDcfclocks]: dcefclk dpm not enabled!\n", - return -EPERM); + if (!data->smu_features[GNLD_DPM_DCEFCLK].enabled) + return -1; count = (dpm_table->count > MAX_NUM_CLOCKS) ? MAX_NUM_CLOCKS : dpm_table->count; clocks->num_levels = count; @@ -2460,9 +2710,8 @@ static int vega20_get_socclocks(struct pp_hwmgr *hwmgr, struct vega20_single_dpm_table *dpm_table = &(data->dpm_table.soc_table); int i, count; - PP_ASSERT_WITH_CODE(data->smu_features[GNLD_DPM_SOCCLK].enabled, - "[GetSocclks]: socclk dpm not enabled!\n", - return -EPERM); + if (!data->smu_features[GNLD_DPM_SOCCLK].enabled) + return -1; count = (dpm_table->count > MAX_NUM_CLOCKS) ? MAX_NUM_CLOCKS : dpm_table->count; clocks->num_levels = count; @@ -2540,7 +2789,6 @@ static int vega20_odn_edit_dpm_table(struct pp_hwmgr *hwmgr, data->od8_settings.od8_settings_array; OverDriveTable_t *od_table = &(data->smc_state_table.overdrive_table); - struct pp_clock_levels_with_latency clocks; int32_t input_index, input_clk, input_vol, i; int od8_id; int ret; @@ -2599,11 +2847,6 @@ static int vega20_odn_edit_dpm_table(struct pp_hwmgr *hwmgr, return -EOPNOTSUPP; } - ret = vega20_get_memclocks(hwmgr, &clocks); - PP_ASSERT_WITH_CODE(!ret, - "Attempt to get memory clk levels failed!", - return ret); - for (i = 0; i < size; i += 2) { if (i + 2 > size) { pr_info("invalid number of input parameters %d\n", @@ -2620,11 +2863,11 @@ static int vega20_odn_edit_dpm_table(struct pp_hwmgr *hwmgr, return -EINVAL; } - if (input_clk < clocks.data[0].clocks_in_khz / 1000 || + if (input_clk < od8_settings[OD8_SETTING_UCLK_FMAX].min_value || input_clk > od8_settings[OD8_SETTING_UCLK_FMAX].max_value) { pr_info("clock freq %d is not within allowed range [%d - %d]\n", input_clk, - clocks.data[0].clocks_in_khz / 1000, + od8_settings[OD8_SETTING_UCLK_FMAX].min_value, od8_settings[OD8_SETTING_UCLK_FMAX].max_value); return -EINVAL; } @@ -2749,6 +2992,109 @@ static int vega20_odn_edit_dpm_table(struct pp_hwmgr *hwmgr, return 0; } +static int vega20_get_ppfeature_status(struct pp_hwmgr *hwmgr, char *buf) +{ + static const char *ppfeature_name[] = { + "DPM_PREFETCHER", + "GFXCLK_DPM", + "UCLK_DPM", + "SOCCLK_DPM", + "UVD_DPM", + "VCE_DPM", + "ULV", + "MP0CLK_DPM", + "LINK_DPM", + "DCEFCLK_DPM", + "GFXCLK_DS", + "SOCCLK_DS", + "LCLK_DS", + "PPT", + "TDC", + "THERMAL", + "GFX_PER_CU_CG", + "RM", + "DCEFCLK_DS", + "ACDC", + "VR0HOT", + "VR1HOT", + "FW_CTF", + "LED_DISPLAY", + "FAN_CONTROL", + "GFX_EDC", + "GFXOFF", + "CG", + "FCLK_DPM", + "FCLK_DS", + "MP1CLK_DS", + "MP0CLK_DS", + "XGMI", + "ECC"}; + static const char *output_title[] = { + "FEATURES", + "BITMASK", + "ENABLEMENT"}; + uint64_t features_enabled; + int i; + int ret = 0; + int size = 0; + + ret = vega20_get_enabled_smc_features(hwmgr, &features_enabled); + PP_ASSERT_WITH_CODE(!ret, + "[EnableAllSmuFeatures] Failed to get enabled smc features!", + return ret); + + size += sprintf(buf + size, "Current ppfeatures: 0x%016llx\n", features_enabled); + size += sprintf(buf + size, "%-19s %-22s %s\n", + output_title[0], + output_title[1], + output_title[2]); + for (i = 0; i < GNLD_FEATURES_MAX; i++) { + size += sprintf(buf + size, "%-19s 0x%016llx %6s\n", + ppfeature_name[i], + 1ULL << i, + (features_enabled & (1ULL << i)) ? "Y" : "N"); + } + + return size; +} + +static int vega20_set_ppfeature_status(struct pp_hwmgr *hwmgr, uint64_t new_ppfeature_masks) +{ + uint64_t features_enabled; + uint64_t features_to_enable; + uint64_t features_to_disable; + int ret = 0; + + if (new_ppfeature_masks >= (1ULL << GNLD_FEATURES_MAX)) + return -EINVAL; + + ret = vega20_get_enabled_smc_features(hwmgr, &features_enabled); + if (ret) + return ret; + + features_to_disable = + features_enabled & ~new_ppfeature_masks; + features_to_enable = + ~features_enabled & new_ppfeature_masks; + + pr_debug("features_to_disable 0x%llx\n", features_to_disable); + pr_debug("features_to_enable 0x%llx\n", features_to_enable); + + if (features_to_disable) { + ret = vega20_enable_smc_features(hwmgr, false, features_to_disable); + if (ret) + return ret; + } + + if (features_to_enable) { + ret = vega20_enable_smc_features(hwmgr, true, features_to_enable); + if (ret) + return ret; + } + + return 0; +} + static int vega20_print_clock_levels(struct pp_hwmgr *hwmgr, enum pp_clock_type type, char *buf) { @@ -2758,9 +3104,16 @@ static int vega20_print_clock_levels(struct pp_hwmgr *hwmgr, data->od8_settings.od8_settings_array; OverDriveTable_t *od_table = &(data->smc_state_table.overdrive_table); + struct phm_ppt_v3_information *pptable_information = + (struct phm_ppt_v3_information *)hwmgr->pptable; + PPTable_t *pptable = (PPTable_t *)pptable_information->smc_pptable; + struct amdgpu_device *adev = hwmgr->adev; struct pp_clock_levels_with_latency clocks; + struct vega20_single_dpm_table *fclk_dpm_table = + &(data->dpm_table.fclk_table); int i, now, size = 0; int ret = 0; + uint32_t gen_speed, lane_width, current_gen_speed, current_lane_width; switch (type) { case PP_SCLK: @@ -2769,15 +3122,16 @@ static int vega20_print_clock_levels(struct pp_hwmgr *hwmgr, "Attempt to get current gfx clk Failed!", return ret); - ret = vega20_get_sclks(hwmgr, &clocks); - PP_ASSERT_WITH_CODE(!ret, - "Attempt to get gfx clk levels Failed!", - return ret); + if (vega20_get_sclks(hwmgr, &clocks)) { + size += sprintf(buf + size, "0: %uMhz * (DPM disabled)\n", + now / 100); + break; + } for (i = 0; i < clocks.num_levels; i++) size += sprintf(buf + size, "%d: %uMhz %s\n", i, clocks.data[i].clocks_in_khz / 1000, - (clocks.data[i].clocks_in_khz == now) ? "*" : ""); + (clocks.data[i].clocks_in_khz == now * 10) ? "*" : ""); break; case PP_MCLK: @@ -2786,18 +3140,97 @@ static int vega20_print_clock_levels(struct pp_hwmgr *hwmgr, "Attempt to get current mclk freq Failed!", return ret); - ret = vega20_get_memclocks(hwmgr, &clocks); + if (vega20_get_memclocks(hwmgr, &clocks)) { + size += sprintf(buf + size, "0: %uMhz * (DPM disabled)\n", + now / 100); + break; + } + + for (i = 0; i < clocks.num_levels; i++) + size += sprintf(buf + size, "%d: %uMhz %s\n", + i, clocks.data[i].clocks_in_khz / 1000, + (clocks.data[i].clocks_in_khz == now * 10) ? "*" : ""); + break; + + case PP_SOCCLK: + ret = vega20_get_current_clk_freq(hwmgr, PPCLK_SOCCLK, &now); + PP_ASSERT_WITH_CODE(!ret, + "Attempt to get current socclk freq Failed!", + return ret); + + if (vega20_get_socclocks(hwmgr, &clocks)) { + size += sprintf(buf + size, "0: %uMhz * (DPM disabled)\n", + now / 100); + break; + } + + for (i = 0; i < clocks.num_levels; i++) + size += sprintf(buf + size, "%d: %uMhz %s\n", + i, clocks.data[i].clocks_in_khz / 1000, + (clocks.data[i].clocks_in_khz == now * 10) ? "*" : ""); + break; + + case PP_FCLK: + ret = vega20_get_current_clk_freq(hwmgr, PPCLK_FCLK, &now); PP_ASSERT_WITH_CODE(!ret, - "Attempt to get memory clk levels Failed!", + "Attempt to get current fclk freq Failed!", return ret); + for (i = 0; i < fclk_dpm_table->count; i++) + size += sprintf(buf + size, "%d: %uMhz %s\n", + i, fclk_dpm_table->dpm_levels[i].value, + fclk_dpm_table->dpm_levels[i].value == (now / 100) ? "*" : ""); + break; + + case PP_DCEFCLK: + ret = vega20_get_current_clk_freq(hwmgr, PPCLK_DCEFCLK, &now); + PP_ASSERT_WITH_CODE(!ret, + "Attempt to get current dcefclk freq Failed!", + return ret); + + if (vega20_get_dcefclocks(hwmgr, &clocks)) { + size += sprintf(buf + size, "0: %uMhz * (DPM disabled)\n", + now / 100); + break; + } + for (i = 0; i < clocks.num_levels; i++) size += sprintf(buf + size, "%d: %uMhz %s\n", i, clocks.data[i].clocks_in_khz / 1000, - (clocks.data[i].clocks_in_khz == now) ? "*" : ""); + (clocks.data[i].clocks_in_khz == now * 10) ? "*" : ""); break; case PP_PCIE: + current_gen_speed = (RREG32_PCIE(smnPCIE_LC_SPEED_CNTL) & + PSWUSP0_PCIE_LC_SPEED_CNTL__LC_CURRENT_DATA_RATE_MASK) + >> PSWUSP0_PCIE_LC_SPEED_CNTL__LC_CURRENT_DATA_RATE__SHIFT; + current_lane_width = (RREG32_PCIE(smnPCIE_LC_LINK_WIDTH_CNTL) & + PCIE_LC_LINK_WIDTH_CNTL__LC_LINK_WIDTH_RD_MASK) + >> PCIE_LC_LINK_WIDTH_CNTL__LC_LINK_WIDTH_RD__SHIFT; + for (i = 0; i < NUM_LINK_LEVELS; i++) { + if (i == 1 && data->pcie_parameters_override) { + gen_speed = data->pcie_gen_level1; + lane_width = data->pcie_width_level1; + } else { + gen_speed = pptable->PcieGenSpeed[i]; + lane_width = pptable->PcieLaneCount[i]; + } + size += sprintf(buf + size, "%d: %s %s %dMhz %s\n", i, + (gen_speed == 0) ? "2.5GT/s," : + (gen_speed == 1) ? "5.0GT/s," : + (gen_speed == 2) ? "8.0GT/s," : + (gen_speed == 3) ? "16.0GT/s," : "", + (lane_width == 1) ? "x1" : + (lane_width == 2) ? "x2" : + (lane_width == 3) ? "x4" : + (lane_width == 4) ? "x8" : + (lane_width == 5) ? "x12" : + (lane_width == 6) ? "x16" : "", + pptable->LclkFreq[i], + (current_gen_speed == gen_speed) && + (current_lane_width == lane_width) ? + "*" : ""); + } break; case OD_SCLK: @@ -2852,13 +3285,8 @@ static int vega20_print_clock_levels(struct pp_hwmgr *hwmgr, } if (od8_settings[OD8_SETTING_UCLK_FMAX].feature_id) { - ret = vega20_get_memclocks(hwmgr, &clocks); - PP_ASSERT_WITH_CODE(!ret, - "Fail to get memory clk levels!", - return ret); - size += sprintf(buf + size, "MCLK: %7uMhz %10uMhz\n", - clocks.data[0].clocks_in_khz / 1000, + od8_settings[OD8_SETTING_UCLK_FMAX].min_value, od8_settings[OD8_SETTING_UCLK_FMAX].max_value); } @@ -2920,6 +3348,31 @@ static int vega20_set_uclk_to_highest_dpm_level(struct pp_hwmgr *hwmgr, return ret; } +static int vega20_set_fclk_to_highest_dpm_level(struct pp_hwmgr *hwmgr) +{ + struct vega20_hwmgr *data = (struct vega20_hwmgr *)(hwmgr->backend); + struct vega20_single_dpm_table *dpm_table = &(data->dpm_table.fclk_table); + int ret = 0; + + if (data->smu_features[GNLD_DPM_FCLK].enabled) { + PP_ASSERT_WITH_CODE(dpm_table->count > 0, + "[SetFclkToHightestDpmLevel] Dpm table has no entry!", + return -EINVAL); + PP_ASSERT_WITH_CODE(dpm_table->count <= NUM_FCLK_DPM_LEVELS, + "[SetFclkToHightestDpmLevel] Dpm table has too many entries!", + return -EINVAL); + + dpm_table->dpm_state.soft_min_level = dpm_table->dpm_levels[dpm_table->count - 1].value; + PP_ASSERT_WITH_CODE(!(ret = smum_send_msg_to_smc_with_parameter(hwmgr, + PPSMC_MSG_SetSoftMinByFreq, + (PPCLK_FCLK << 16 ) | dpm_table->dpm_state.soft_min_level)), + "[SetFclkToHightestDpmLevel] Set soft min fclk failed!", + return ret); + } + + return ret; +} + static int vega20_pre_display_configuration_changed_task(struct pp_hwmgr *hwmgr) { struct vega20_hwmgr *data = (struct vega20_hwmgr *)(hwmgr->backend); @@ -2930,8 +3383,10 @@ static int vega20_pre_display_configuration_changed_task(struct pp_hwmgr *hwmgr) ret = vega20_set_uclk_to_highest_dpm_level(hwmgr, &data->dpm_table.mem_table); + if (ret) + return ret; - return ret; + return vega20_set_fclk_to_highest_dpm_level(hwmgr); } static int vega20_display_configuration_changed_task(struct pp_hwmgr *hwmgr) @@ -3015,19 +3470,20 @@ static int vega20_apply_clocks_adjust_rules(struct pp_hwmgr *hwmgr) struct vega20_single_dpm_table *dpm_table; bool vblank_too_short = false; bool disable_mclk_switching; + bool disable_fclk_switching; uint32_t i, latency; disable_mclk_switching = ((1 < hwmgr->display_config->num_display) && !hwmgr->display_config->multi_monitor_in_sync) || vblank_too_short; - latency = hwmgr->display_config->dce_tolerable_mclk_in_active_latency; + latency = hwmgr->display_config->dce_tolerable_mclk_in_active_latency; /* gfxclk */ dpm_table = &(data->dpm_table.gfx_table); dpm_table->dpm_state.soft_min_level = dpm_table->dpm_levels[0].value; - dpm_table->dpm_state.soft_max_level = dpm_table->dpm_levels[dpm_table->count - 1].value; + dpm_table->dpm_state.soft_max_level = VG20_CLOCK_MAX_DEFAULT; dpm_table->dpm_state.hard_min_level = dpm_table->dpm_levels[0].value; - dpm_table->dpm_state.hard_max_level = dpm_table->dpm_levels[dpm_table->count - 1].value; + dpm_table->dpm_state.hard_max_level = VG20_CLOCK_MAX_DEFAULT; if (PP_CAP(PHM_PlatformCaps_UMDPState)) { if (VEGA20_UMD_PSTATE_GFXCLK_LEVEL < dpm_table->count) { @@ -3049,9 +3505,9 @@ static int vega20_apply_clocks_adjust_rules(struct pp_hwmgr *hwmgr) /* memclk */ dpm_table = &(data->dpm_table.mem_table); dpm_table->dpm_state.soft_min_level = dpm_table->dpm_levels[0].value; - dpm_table->dpm_state.soft_max_level = dpm_table->dpm_levels[dpm_table->count - 1].value; + dpm_table->dpm_state.soft_max_level = VG20_CLOCK_MAX_DEFAULT; dpm_table->dpm_state.hard_min_level = dpm_table->dpm_levels[0].value; - dpm_table->dpm_state.hard_max_level = dpm_table->dpm_levels[dpm_table->count - 1].value; + dpm_table->dpm_state.hard_max_level = VG20_CLOCK_MAX_DEFAULT; if (PP_CAP(PHM_PlatformCaps_UMDPState)) { if (VEGA20_UMD_PSTATE_MCLK_LEVEL < dpm_table->count) { @@ -3090,12 +3546,28 @@ static int vega20_apply_clocks_adjust_rules(struct pp_hwmgr *hwmgr) if (hwmgr->display_config->nb_pstate_switch_disable) dpm_table->dpm_state.hard_min_level = dpm_table->dpm_levels[dpm_table->count - 1].value; + if ((disable_mclk_switching && + (dpm_table->dpm_state.hard_min_level == dpm_table->dpm_levels[dpm_table->count - 1].value)) || + hwmgr->display_config->min_mem_set_clock / 100 >= dpm_table->dpm_levels[dpm_table->count - 1].value) + disable_fclk_switching = true; + else + disable_fclk_switching = false; + + /* fclk */ + dpm_table = &(data->dpm_table.fclk_table); + dpm_table->dpm_state.soft_min_level = dpm_table->dpm_levels[0].value; + dpm_table->dpm_state.soft_max_level = VG20_CLOCK_MAX_DEFAULT; + dpm_table->dpm_state.hard_min_level = dpm_table->dpm_levels[0].value; + dpm_table->dpm_state.hard_max_level = VG20_CLOCK_MAX_DEFAULT; + if (hwmgr->display_config->nb_pstate_switch_disable || disable_fclk_switching) + dpm_table->dpm_state.soft_min_level = dpm_table->dpm_levels[dpm_table->count - 1].value; + /* vclk */ dpm_table = &(data->dpm_table.vclk_table); dpm_table->dpm_state.soft_min_level = dpm_table->dpm_levels[0].value; - dpm_table->dpm_state.soft_max_level = dpm_table->dpm_levels[dpm_table->count - 1].value; + dpm_table->dpm_state.soft_max_level = VG20_CLOCK_MAX_DEFAULT; dpm_table->dpm_state.hard_min_level = dpm_table->dpm_levels[0].value; - dpm_table->dpm_state.hard_max_level = dpm_table->dpm_levels[dpm_table->count - 1].value; + dpm_table->dpm_state.hard_max_level = VG20_CLOCK_MAX_DEFAULT; if (PP_CAP(PHM_PlatformCaps_UMDPState)) { if (VEGA20_UMD_PSTATE_UVDCLK_LEVEL < dpm_table->count) { @@ -3112,9 +3584,9 @@ static int vega20_apply_clocks_adjust_rules(struct pp_hwmgr *hwmgr) /* dclk */ dpm_table = &(data->dpm_table.dclk_table); dpm_table->dpm_state.soft_min_level = dpm_table->dpm_levels[0].value; - dpm_table->dpm_state.soft_max_level = dpm_table->dpm_levels[dpm_table->count - 1].value; + dpm_table->dpm_state.soft_max_level = VG20_CLOCK_MAX_DEFAULT; dpm_table->dpm_state.hard_min_level = dpm_table->dpm_levels[0].value; - dpm_table->dpm_state.hard_max_level = dpm_table->dpm_levels[dpm_table->count - 1].value; + dpm_table->dpm_state.hard_max_level = VG20_CLOCK_MAX_DEFAULT; if (PP_CAP(PHM_PlatformCaps_UMDPState)) { if (VEGA20_UMD_PSTATE_UVDCLK_LEVEL < dpm_table->count) { @@ -3131,9 +3603,9 @@ static int vega20_apply_clocks_adjust_rules(struct pp_hwmgr *hwmgr) /* socclk */ dpm_table = &(data->dpm_table.soc_table); dpm_table->dpm_state.soft_min_level = dpm_table->dpm_levels[0].value; - dpm_table->dpm_state.soft_max_level = dpm_table->dpm_levels[dpm_table->count - 1].value; + dpm_table->dpm_state.soft_max_level = VG20_CLOCK_MAX_DEFAULT; dpm_table->dpm_state.hard_min_level = dpm_table->dpm_levels[0].value; - dpm_table->dpm_state.hard_max_level = dpm_table->dpm_levels[dpm_table->count - 1].value; + dpm_table->dpm_state.hard_max_level = VG20_CLOCK_MAX_DEFAULT; if (PP_CAP(PHM_PlatformCaps_UMDPState)) { if (VEGA20_UMD_PSTATE_SOCCLK_LEVEL < dpm_table->count) { @@ -3150,9 +3622,9 @@ static int vega20_apply_clocks_adjust_rules(struct pp_hwmgr *hwmgr) /* eclk */ dpm_table = &(data->dpm_table.eclk_table); dpm_table->dpm_state.soft_min_level = dpm_table->dpm_levels[0].value; - dpm_table->dpm_state.soft_max_level = dpm_table->dpm_levels[dpm_table->count - 1].value; + dpm_table->dpm_state.soft_max_level = VG20_CLOCK_MAX_DEFAULT; dpm_table->dpm_state.hard_min_level = dpm_table->dpm_levels[0].value; - dpm_table->dpm_state.hard_max_level = dpm_table->dpm_levels[dpm_table->count - 1].value; + dpm_table->dpm_state.hard_max_level = VG20_CLOCK_MAX_DEFAULT; if (PP_CAP(PHM_PlatformCaps_UMDPState)) { if (VEGA20_UMD_PSTATE_VCEMCLK_LEVEL < dpm_table->count) { @@ -3218,6 +3690,9 @@ static int conv_power_profile_to_pplib_workload(int power_profile) int pplib_workload = 0; switch (power_profile) { + case PP_SMC_POWER_PROFILE_BOOTUP_DEFAULT: + pplib_workload = WORKLOAD_DEFAULT_BIT; + break; case PP_SMC_POWER_PROFILE_FULLSCREEN3D: pplib_workload = WORKLOAD_PPLIB_FULL_SCREEN_3D_BIT; break; @@ -3247,6 +3722,7 @@ static int vega20_get_power_profile_mode(struct pp_hwmgr *hwmgr, char *buf) uint32_t i, size = 0; uint16_t workload_type = 0; static const char *profile_name[] = { + "BOOTUP_DEFAULT", "3D_FULL_SCREEN", "POWER_SAVING", "VIDEO", @@ -3476,109 +3952,70 @@ static int vega20_get_thermal_temperature_range(struct pp_hwmgr *hwmgr, static const struct pp_hwmgr_func vega20_hwmgr_funcs = { /* init/fini related */ - .backend_init = - vega20_hwmgr_backend_init, - .backend_fini = - vega20_hwmgr_backend_fini, - .asic_setup = - vega20_setup_asic_task, - .power_off_asic = - vega20_power_off_asic, - .dynamic_state_management_enable = - vega20_enable_dpm_tasks, - .dynamic_state_management_disable = - vega20_disable_dpm_tasks, + .backend_init = vega20_hwmgr_backend_init, + .backend_fini = vega20_hwmgr_backend_fini, + .asic_setup = vega20_setup_asic_task, + .power_off_asic = vega20_power_off_asic, + .dynamic_state_management_enable = vega20_enable_dpm_tasks, + .dynamic_state_management_disable = vega20_disable_dpm_tasks, /* power state related */ - .apply_clocks_adjust_rules = - vega20_apply_clocks_adjust_rules, - .pre_display_config_changed = - vega20_pre_display_configuration_changed_task, - .display_config_changed = - vega20_display_configuration_changed_task, + .apply_clocks_adjust_rules = vega20_apply_clocks_adjust_rules, + .pre_display_config_changed = vega20_pre_display_configuration_changed_task, + .display_config_changed = vega20_display_configuration_changed_task, .check_smc_update_required_for_display_configuration = vega20_check_smc_update_required_for_display_configuration, .notify_smc_display_config_after_ps_adjustment = vega20_notify_smc_display_config_after_ps_adjustment, /* export to DAL */ - .get_sclk = - vega20_dpm_get_sclk, - .get_mclk = - vega20_dpm_get_mclk, - .get_dal_power_level = - vega20_get_dal_power_level, - .get_clock_by_type_with_latency = - vega20_get_clock_by_type_with_latency, - .get_clock_by_type_with_voltage = - vega20_get_clock_by_type_with_voltage, - .set_watermarks_for_clocks_ranges = - vega20_set_watermarks_for_clocks_ranges, - .display_clock_voltage_request = - vega20_display_clock_voltage_request, - .get_performance_level = - vega20_get_performance_level, + .get_sclk = vega20_dpm_get_sclk, + .get_mclk = vega20_dpm_get_mclk, + .get_dal_power_level = vega20_get_dal_power_level, + .get_clock_by_type_with_latency = vega20_get_clock_by_type_with_latency, + .get_clock_by_type_with_voltage = vega20_get_clock_by_type_with_voltage, + .set_watermarks_for_clocks_ranges = vega20_set_watermarks_for_clocks_ranges, + .display_clock_voltage_request = vega20_display_clock_voltage_request, + .get_performance_level = vega20_get_performance_level, /* UMD pstate, profile related */ - .force_dpm_level = - vega20_dpm_force_dpm_level, - .get_power_profile_mode = - vega20_get_power_profile_mode, - .set_power_profile_mode = - vega20_set_power_profile_mode, + .force_dpm_level = vega20_dpm_force_dpm_level, + .get_power_profile_mode = vega20_get_power_profile_mode, + .set_power_profile_mode = vega20_set_power_profile_mode, /* od related */ - .set_power_limit = - vega20_set_power_limit, - .get_sclk_od = - vega20_get_sclk_od, - .set_sclk_od = - vega20_set_sclk_od, - .get_mclk_od = - vega20_get_mclk_od, - .set_mclk_od = - vega20_set_mclk_od, - .odn_edit_dpm_table = - vega20_odn_edit_dpm_table, + .set_power_limit = vega20_set_power_limit, + .get_sclk_od = vega20_get_sclk_od, + .set_sclk_od = vega20_set_sclk_od, + .get_mclk_od = vega20_get_mclk_od, + .set_mclk_od = vega20_set_mclk_od, + .odn_edit_dpm_table = vega20_odn_edit_dpm_table, /* for sysfs to retrive/set gfxclk/memclk */ - .force_clock_level = - vega20_force_clock_level, - .print_clock_levels = - vega20_print_clock_levels, - .read_sensor = - vega20_read_sensor, + .force_clock_level = vega20_force_clock_level, + .print_clock_levels = vega20_print_clock_levels, + .read_sensor = vega20_read_sensor, + .get_ppfeature_status = vega20_get_ppfeature_status, + .set_ppfeature_status = vega20_set_ppfeature_status, /* powergate related */ - .powergate_uvd = - vega20_power_gate_uvd, - .powergate_vce = - vega20_power_gate_vce, + .powergate_uvd = vega20_power_gate_uvd, + .powergate_vce = vega20_power_gate_vce, /* thermal related */ - .start_thermal_controller = - vega20_start_thermal_controller, - .stop_thermal_controller = - vega20_thermal_stop_thermal_controller, - .get_thermal_temperature_range = - vega20_get_thermal_temperature_range, - .register_irq_handlers = - smu9_register_irq_handlers, - .disable_smc_firmware_ctf = - vega20_thermal_disable_alert, + .start_thermal_controller = vega20_start_thermal_controller, + .stop_thermal_controller = vega20_thermal_stop_thermal_controller, + .get_thermal_temperature_range = vega20_get_thermal_temperature_range, + .register_irq_handlers = smu9_register_irq_handlers, + .disable_smc_firmware_ctf = vega20_thermal_disable_alert, /* fan control related */ - .get_fan_speed_percent = - vega20_fan_ctrl_get_fan_speed_percent, - .set_fan_speed_percent = - vega20_fan_ctrl_set_fan_speed_percent, - .get_fan_speed_info = - vega20_fan_ctrl_get_fan_speed_info, - .get_fan_speed_rpm = - vega20_fan_ctrl_get_fan_speed_rpm, - .set_fan_speed_rpm = - vega20_fan_ctrl_set_fan_speed_rpm, - .get_fan_control_mode = - vega20_get_fan_control_mode, - .set_fan_control_mode = - vega20_set_fan_control_mode, + .get_fan_speed_percent = vega20_fan_ctrl_get_fan_speed_percent, + .set_fan_speed_percent = vega20_fan_ctrl_set_fan_speed_percent, + .get_fan_speed_info = vega20_fan_ctrl_get_fan_speed_info, + .get_fan_speed_rpm = vega20_fan_ctrl_get_fan_speed_rpm, + .set_fan_speed_rpm = vega20_fan_ctrl_set_fan_speed_rpm, + .get_fan_control_mode = vega20_get_fan_control_mode, + .set_fan_control_mode = vega20_set_fan_control_mode, /* smu memory related */ - .notify_cac_buffer_info = - vega20_notify_cac_buffer_info, - .enable_mgpu_fan_boost = - vega20_enable_mgpu_fan_boost, + .notify_cac_buffer_info = vega20_notify_cac_buffer_info, + .enable_mgpu_fan_boost = vega20_enable_mgpu_fan_boost, + /* BACO related */ + .get_asic_baco_capability = vega20_baco_get_capability, + .get_asic_baco_state = vega20_baco_get_state, + .set_asic_baco_state = vega20_baco_set_state, }; int vega20_hwmgr_init(struct pp_hwmgr *hwmgr) diff --git a/drivers/gpu/drm/amd/powerplay/hwmgr/vega20_hwmgr.h b/drivers/gpu/drm/amd/powerplay/hwmgr/vega20_hwmgr.h index 25faaa5c5b10..ac2a3118a0ae 100644 --- a/drivers/gpu/drm/amd/powerplay/hwmgr/vega20_hwmgr.h +++ b/drivers/gpu/drm/amd/powerplay/hwmgr/vega20_hwmgr.h @@ -42,6 +42,8 @@ #define AVFS_CURVE 0 #define OD8_HOTCURVE_TEMPERATURE 85 +#define VG20_CLOCK_MAX_DEFAULT 0xFFFF + typedef uint32_t PP_Clock; enum { @@ -78,6 +80,7 @@ enum { GNLD_DS_MP1CLK, GNLD_DS_MP0CLK, GNLD_XGMI, + GNLD_ECC, GNLD_FEATURES_MAX }; @@ -219,6 +222,7 @@ struct vega20_vbios_boot_state { uint32_t eclock; uint32_t dclock; uint32_t vclock; + uint32_t fclock; }; #define DPMTABLE_OD_UPDATE_SCLK 0x00000001 @@ -520,6 +524,13 @@ struct vega20_hwmgr { /* ---- Gfxoff ---- */ bool gfxoff_allowed; uint32_t counter_gfxoff; + + unsigned long metrics_time; + SmuMetrics_t metrics_table; + + bool pcie_parameters_override; + uint32_t pcie_gen_level1; + uint32_t pcie_width_level1; }; #define VEGA20_DPM2_NEAR_TDP_DEC 10 diff --git a/drivers/gpu/drm/amd/powerplay/hwmgr/vega20_inc.h b/drivers/gpu/drm/amd/powerplay/hwmgr/vega20_inc.h index 6738bad53602..613cb1989b3d 100644 --- a/drivers/gpu/drm/amd/powerplay/hwmgr/vega20_inc.h +++ b/drivers/gpu/drm/amd/powerplay/hwmgr/vega20_inc.h @@ -31,5 +31,6 @@ #include "asic_reg/mp/mp_9_0_sh_mask.h" #include "asic_reg/nbio/nbio_7_4_offset.h" +#include "asic_reg/nbio/nbio_7_4_sh_mask.h" #endif diff --git a/drivers/gpu/drm/amd/powerplay/hwmgr/vega20_processpptables.c b/drivers/gpu/drm/amd/powerplay/hwmgr/vega20_processpptables.c index 97f8a1a970c3..7a7f15d0c53a 100644 --- a/drivers/gpu/drm/amd/powerplay/hwmgr/vega20_processpptables.c +++ b/drivers/gpu/drm/amd/powerplay/hwmgr/vega20_processpptables.c @@ -32,6 +32,8 @@ #include "cgs_common.h" #include "vega20_pptable.h" +#define VEGA20_FAN_TARGET_TEMPERATURE_OVERRIDE 105 + static void set_hw_cap(struct pp_hwmgr *hwmgr, bool enable, enum phm_platform_caps cap) { @@ -798,6 +800,17 @@ static int append_vbios_pptable(struct pp_hwmgr *hwmgr, PPTable_t *ppsmc_pptable return 0; } +static int override_powerplay_table_fantargettemperature(struct pp_hwmgr *hwmgr) +{ + struct phm_ppt_v3_information *pptable_information = + (struct phm_ppt_v3_information *)hwmgr->pptable; + PPTable_t *ppsmc_pptable = (PPTable_t *)(pptable_information->smc_pptable); + + ppsmc_pptable->FanTargetTemperature = VEGA20_FAN_TARGET_TEMPERATURE_OVERRIDE; + + return 0; +} + #define VEGA20_ENGINECLOCK_HARDMAX 198000 static int init_powerplay_table_information( struct pp_hwmgr *hwmgr, @@ -887,6 +900,10 @@ static int init_powerplay_table_information( result = append_vbios_pptable(hwmgr, (pptable_information->smc_pptable)); + if (result) + return result; + + result = override_powerplay_table_fantargettemperature(hwmgr); return result; } diff --git a/drivers/gpu/drm/amd/powerplay/inc/hardwaremanager.h b/drivers/gpu/drm/amd/powerplay/inc/hardwaremanager.h index 54fd0125d9cf..6e0be6027705 100644 --- a/drivers/gpu/drm/amd/powerplay/inc/hardwaremanager.h +++ b/drivers/gpu/drm/amd/powerplay/inc/hardwaremanager.h @@ -397,7 +397,6 @@ struct phm_odn_clock_levels { }; extern int phm_disable_clock_power_gatings(struct pp_hwmgr *hwmgr); -extern int phm_enable_clock_power_gatings(struct pp_hwmgr *hwmgr); extern int phm_powerdown_uvd(struct pp_hwmgr *hwmgr); extern int phm_setup_asic(struct pp_hwmgr *hwmgr); extern int phm_enable_dynamic_state_management(struct pp_hwmgr *hwmgr); @@ -463,5 +462,8 @@ extern int phm_display_clock_voltage_request(struct pp_hwmgr *hwmgr, extern int phm_get_max_high_clocks(struct pp_hwmgr *hwmgr, struct amd_pp_simple_clock_info *clocks); extern int phm_disable_smc_firmware_ctf(struct pp_hwmgr *hwmgr); + +extern int phm_set_active_display_count(struct pp_hwmgr *hwmgr, uint32_t count); + #endif /* _HARDWARE_MANAGER_H_ */ diff --git a/drivers/gpu/drm/amd/powerplay/inc/hwmgr.h b/drivers/gpu/drm/amd/powerplay/inc/hwmgr.h index e5a60aa44b5d..bac3d85e3b82 100644 --- a/drivers/gpu/drm/amd/powerplay/inc/hwmgr.h +++ b/drivers/gpu/drm/amd/powerplay/inc/hwmgr.h @@ -28,7 +28,6 @@ #include "hardwaremanager.h" #include "hwmgr_ppt.h" #include "ppatomctrl.h" -#include "hwmgr_ppt.h" #include "power_state.h" #include "smu_helper.h" @@ -48,6 +47,11 @@ enum DISPLAY_GAP { }; typedef enum DISPLAY_GAP DISPLAY_GAP; +enum BACO_STATE { + BACO_STATE_OUT = 0, + BACO_STATE_IN, +}; + struct vi_dpm_level { bool enabled; uint32_t value; @@ -252,7 +256,6 @@ struct pp_hwmgr_func { uint32_t (*get_sclk)(struct pp_hwmgr *hwmgr, bool low); int (*power_state_set)(struct pp_hwmgr *hwmgr, const void *state); - int (*enable_clock_power_gating)(struct pp_hwmgr *hwmgr); int (*notify_smc_display_config_after_ps_adjustment)(struct pp_hwmgr *hwmgr); int (*pre_display_config_changed)(struct pp_hwmgr *hwmgr); int (*display_config_changed)(struct pp_hwmgr *hwmgr); @@ -310,7 +313,7 @@ struct pp_hwmgr_func { int (*avfs_control)(struct pp_hwmgr *hwmgr, bool enable); int (*disable_smc_firmware_ctf)(struct pp_hwmgr *hwmgr); int (*set_active_display_count)(struct pp_hwmgr *hwmgr, uint32_t count); - int (*set_deep_sleep_dcefclk)(struct pp_hwmgr *hwmgr, uint32_t clock); + int (*set_min_deep_sleep_dcefclk)(struct pp_hwmgr *hwmgr, uint32_t clock); int (*start_thermal_controller)(struct pp_hwmgr *hwmgr, struct PP_TemperatureRange *range); int (*notify_cac_buffer_info)(struct pp_hwmgr *hwmgr, uint32_t virtual_addr_low, @@ -318,6 +321,9 @@ struct pp_hwmgr_func { uint32_t mc_addr_low, uint32_t mc_addr_hi, uint32_t size); + int (*update_nbdpm_pstate)(struct pp_hwmgr *hwmgr, + bool enable, + bool lock); int (*get_thermal_temperature_range)(struct pp_hwmgr *hwmgr, struct PP_TemperatureRange *range); int (*get_power_profile_mode)(struct pp_hwmgr *hwmgr, char *buf); @@ -330,6 +336,13 @@ struct pp_hwmgr_func { int (*smus_notify_pwe)(struct pp_hwmgr *hwmgr); int (*powergate_sdma)(struct pp_hwmgr *hwmgr, bool bgate); int (*enable_mgpu_fan_boost)(struct pp_hwmgr *hwmgr); + int (*set_hard_min_dcefclk_by_freq)(struct pp_hwmgr *hwmgr, uint32_t clock); + int (*set_hard_min_fclk_by_freq)(struct pp_hwmgr *hwmgr, uint32_t clock); + int (*get_asic_baco_capability)(struct pp_hwmgr *hwmgr, bool *cap); + int (*get_asic_baco_state)(struct pp_hwmgr *hwmgr, enum BACO_STATE *state); + int (*set_asic_baco_state)(struct pp_hwmgr *hwmgr, enum BACO_STATE state); + int (*get_ppfeature_status)(struct pp_hwmgr *hwmgr, char *buf); + int (*set_ppfeature_status)(struct pp_hwmgr *hwmgr, uint64_t ppfeature_masks); }; struct pp_table_func { @@ -674,6 +687,7 @@ struct pp_advance_fan_control_parameters { uint32_t ulTargetGfxClk; uint16_t usZeroRPMStartTemperature; uint16_t usZeroRPMStopTemperature; + uint16_t usMGpuThrottlingRPMLimit; }; struct pp_thermal_controller_info { @@ -701,7 +715,7 @@ enum PP_TABLE_VERSION { /** * The main hardware manager structure. */ -#define Workload_Policy_Max 5 +#define Workload_Policy_Max 6 struct pp_hwmgr { void *adev; diff --git a/drivers/gpu/drm/amd/powerplay/inc/smu11_driver_if.h b/drivers/gpu/drm/amd/powerplay/inc/smu11_driver_if.h index 63d5cf691549..195c4ae67058 100644 --- a/drivers/gpu/drm/amd/powerplay/inc/smu11_driver_if.h +++ b/drivers/gpu/drm/amd/powerplay/inc/smu11_driver_if.h @@ -99,7 +99,7 @@ #define FEATURE_DS_MP1CLK_BIT 30 #define FEATURE_DS_MP0CLK_BIT 31 #define FEATURE_XGMI_BIT 32 -#define FEATURE_SPARE_33_BIT 33 +#define FEATURE_ECC_BIT 33 #define FEATURE_SPARE_34_BIT 34 #define FEATURE_SPARE_35_BIT 35 #define FEATURE_SPARE_36_BIT 36 @@ -165,7 +165,8 @@ #define FEATURE_DS_FCLK_MASK (1 << FEATURE_DS_FCLK_BIT ) #define FEATURE_DS_MP1CLK_MASK (1 << FEATURE_DS_MP1CLK_BIT ) #define FEATURE_DS_MP0CLK_MASK (1 << FEATURE_DS_MP0CLK_BIT ) -#define FEATURE_XGMI_MASK (1 << FEATURE_XGMI_BIT ) +#define FEATURE_XGMI_MASK (1ULL << FEATURE_XGMI_BIT ) +#define FEATURE_ECC_MASK (1ULL << FEATURE_ECC_BIT ) #define DPM_OVERRIDE_DISABLE_SOCCLK_PID 0x00000001 #define DPM_OVERRIDE_DISABLE_UCLK_PID 0x00000002 diff --git a/drivers/gpu/drm/amd/powerplay/inc/smu7_common.h b/drivers/gpu/drm/amd/powerplay/inc/smu7_common.h index 65eb630bfea3..94bf7b649c20 100644 --- a/drivers/gpu/drm/amd/powerplay/inc/smu7_common.h +++ b/drivers/gpu/drm/amd/powerplay/inc/smu7_common.h @@ -40,10 +40,6 @@ #include "bif/bif_5_0_d.h" #include "bif/bif_5_0_sh_mask.h" - -#include "bif/bif_5_0_d.h" -#include "bif/bif_5_0_sh_mask.h" - #include "dce/dce_10_0_d.h" #include "dce/dce_10_0_sh_mask.h" diff --git a/drivers/gpu/drm/amd/powerplay/inc/smu7_ppsmc.h b/drivers/gpu/drm/amd/powerplay/inc/smu7_ppsmc.h index c1a99dfe4913..6e19f4c7cf8f 100644 --- a/drivers/gpu/drm/amd/powerplay/inc/smu7_ppsmc.h +++ b/drivers/gpu/drm/amd/powerplay/inc/smu7_ppsmc.h @@ -397,6 +397,9 @@ typedef uint16_t PPSMC_Result; #define PPSMC_MSG_SetVBITimeout ((uint16_t) 0x306) +#define PPSMC_MSG_EnableFFC ((uint16_t) 0x307) +#define PPSMC_MSG_DisableFFC ((uint16_t) 0x308) + #define PPSMC_MSG_EnableDpmDidt ((uint16_t) 0x309) #define PPSMC_MSG_DisableDpmDidt ((uint16_t) 0x30A) diff --git a/drivers/gpu/drm/amd/powerplay/smumgr/polaris10_smumgr.c b/drivers/gpu/drm/amd/powerplay/smumgr/polaris10_smumgr.c index a1e0ac9ae248..2d4cfe14f72e 100644 --- a/drivers/gpu/drm/amd/powerplay/smumgr/polaris10_smumgr.c +++ b/drivers/gpu/drm/amd/powerplay/smumgr/polaris10_smumgr.c @@ -44,7 +44,6 @@ #include "smu7_hwmgr.h" #include "hardwaremanager.h" -#include "ppatomctrl.h" #include "atombios.h" #include "pppcielanes.h" @@ -1529,8 +1528,21 @@ static int polaris10_populate_clock_stretcher_data_table(struct pp_hwmgr *hwmgr) efuse = efuse >> 24; if (hwmgr->chip_id == CHIP_POLARIS10) { - min = 1000; - max = 2300; + if (hwmgr->is_kicker) { + min = 1200; + max = 2500; + } else { + min = 1000; + max = 2300; + } + } else if (hwmgr->chip_id == CHIP_POLARIS11) { + if (hwmgr->is_kicker) { + min = 900; + max = 2100; + } else { + min = 1100; + max = 2100; + } } else { min = 1100; max = 2100; @@ -1627,6 +1639,7 @@ static int polaris10_populate_avfs_parameters(struct pp_hwmgr *hwmgr) { struct smu7_hwmgr *data = (struct smu7_hwmgr *)(hwmgr->backend); struct polaris10_smumgr *smu_data = (struct polaris10_smumgr *)(hwmgr->smu_backend); + struct amdgpu_device *adev = hwmgr->adev; SMU74_Discrete_DpmTable *table = &(smu_data->smc_state_table); int result = 0; @@ -1647,6 +1660,59 @@ static int polaris10_populate_avfs_parameters(struct pp_hwmgr *hwmgr) result = atomctrl_get_avfs_information(hwmgr, &avfs_params); if (0 == result) { + if (((adev->pdev->device == 0x67ef) && + ((adev->pdev->revision == 0xe0) || + (adev->pdev->revision == 0xe5))) || + ((adev->pdev->device == 0x67ff) && + ((adev->pdev->revision == 0xcf) || + (adev->pdev->revision == 0xef) || + (adev->pdev->revision == 0xff)))) { + avfs_params.ucEnableApplyAVFS_CKS_OFF_Voltage = 1; + if ((adev->pdev->device == 0x67ef && adev->pdev->revision == 0xe5) || + (adev->pdev->device == 0x67ff && adev->pdev->revision == 0xef)) { + if ((avfs_params.ulGB_VDROOP_TABLE_CKSOFF_a0 == 0xEA522DD3) && + (avfs_params.ulGB_VDROOP_TABLE_CKSOFF_a1 == 0x5645A) && + (avfs_params.ulGB_VDROOP_TABLE_CKSOFF_a2 == 0x33F9E) && + (avfs_params.ulAVFSGB_FUSE_TABLE_CKSOFF_m1 == 0xFFFFC5CC) && + (avfs_params.usAVFSGB_FUSE_TABLE_CKSOFF_m2 == 0x1B1A) && + (avfs_params.ulAVFSGB_FUSE_TABLE_CKSOFF_b == 0xFFFFFCED)) { + avfs_params.ulGB_VDROOP_TABLE_CKSOFF_a0 = 0xF718F1D4; + avfs_params.ulGB_VDROOP_TABLE_CKSOFF_a1 = 0x323FD; + avfs_params.ulGB_VDROOP_TABLE_CKSOFF_a2 = 0x1E455; + avfs_params.ulAVFSGB_FUSE_TABLE_CKSOFF_m1 = 0; + avfs_params.usAVFSGB_FUSE_TABLE_CKSOFF_m2 = 0; + avfs_params.ulAVFSGB_FUSE_TABLE_CKSOFF_b = 0x23; + } + } + } else if (hwmgr->chip_id == CHIP_POLARIS12 && !hwmgr->is_kicker) { + avfs_params.ucEnableApplyAVFS_CKS_OFF_Voltage = 1; + avfs_params.ulGB_VDROOP_TABLE_CKSOFF_a0 = 0xF6B024DD; + avfs_params.ulGB_VDROOP_TABLE_CKSOFF_a1 = 0x3005E; + avfs_params.ulGB_VDROOP_TABLE_CKSOFF_a2 = 0x18A5F; + avfs_params.ulAVFSGB_FUSE_TABLE_CKSOFF_m1 = 0x315; + avfs_params.usAVFSGB_FUSE_TABLE_CKSOFF_m2 = 0xFED1; + avfs_params.ulAVFSGB_FUSE_TABLE_CKSOFF_b = 0x3B; + } else if (((adev->pdev->device == 0x67df) && + ((adev->pdev->revision == 0xe0) || + (adev->pdev->revision == 0xe3) || + (adev->pdev->revision == 0xe4) || + (adev->pdev->revision == 0xe5) || + (adev->pdev->revision == 0xe7) || + (adev->pdev->revision == 0xef))) || + ((adev->pdev->device == 0x6fdf) && + ((adev->pdev->revision == 0xef) || + (adev->pdev->revision == 0xff)))) { + avfs_params.ucEnableApplyAVFS_CKS_OFF_Voltage = 1; + avfs_params.ulGB_VDROOP_TABLE_CKSOFF_a0 = 0xF843B66B; + avfs_params.ulGB_VDROOP_TABLE_CKSOFF_a1 = 0x59CB5; + avfs_params.ulGB_VDROOP_TABLE_CKSOFF_a2 = 0xFFFF287F; + avfs_params.ulAVFSGB_FUSE_TABLE_CKSOFF_m1 = 0; + avfs_params.usAVFSGB_FUSE_TABLE_CKSOFF_m2 = 0xFF23; + avfs_params.ulAVFSGB_FUSE_TABLE_CKSOFF_b = 0x58; + } + } + + if (0 == result) { table->BTCGB_VDROOP_TABLE[0].a0 = PP_HOST_TO_SMC_UL(avfs_params.ulGB_VDROOP_TABLE_CKSON_a0); table->BTCGB_VDROOP_TABLE[0].a1 = PP_HOST_TO_SMC_UL(avfs_params.ulGB_VDROOP_TABLE_CKSON_a1); table->BTCGB_VDROOP_TABLE[0].a2 = PP_HOST_TO_SMC_UL(avfs_params.ulGB_VDROOP_TABLE_CKSON_a2); @@ -2264,6 +2330,7 @@ static uint32_t polaris10_get_offsetof(uint32_t type, uint32_t member) case DRAM_LOG_BUFF_SIZE: return offsetof(SMU74_SoftRegisters, DRAM_LOG_BUFF_SIZE); } + break; case SMU_Discrete_DpmTable: switch (member) { case UvdBootLevel: @@ -2273,6 +2340,7 @@ static uint32_t polaris10_get_offsetof(uint32_t type, uint32_t member) case LowSclkInterruptThreshold: return offsetof(SMU74_Discrete_DpmTable, LowSclkInterruptThreshold); } + break; } pr_warn("can't get the offset of type %x member %x\n", type, member); return 0; diff --git a/drivers/gpu/drm/amd/powerplay/smumgr/smu10_smumgr.c b/drivers/gpu/drm/amd/powerplay/smumgr/smu10_smumgr.c index d0eb8ab50148..d111dd4e03d7 100644 --- a/drivers/gpu/drm/amd/powerplay/smumgr/smu10_smumgr.c +++ b/drivers/gpu/drm/amd/powerplay/smumgr/smu10_smumgr.c @@ -29,7 +29,6 @@ #include "rv_ppsmc.h" #include "smu10_driver_if.h" #include "smu10.h" -#include "ppatomctrl.h" #include "pp_debug.h" diff --git a/drivers/gpu/drm/amd/powerplay/smumgr/smu8_smumgr.c b/drivers/gpu/drm/amd/powerplay/smumgr/smu8_smumgr.c index 09b844ec3eab..e2787e14a500 100644 --- a/drivers/gpu/drm/amd/powerplay/smumgr/smu8_smumgr.c +++ b/drivers/gpu/drm/amd/powerplay/smumgr/smu8_smumgr.c @@ -24,6 +24,7 @@ #include <linux/delay.h> #include <linux/gfp.h> #include <linux/kernel.h> +#include <linux/ktime.h> #include <linux/slab.h> #include <linux/types.h> @@ -61,9 +62,13 @@ static uint32_t smu8_get_argument(struct pp_hwmgr *hwmgr) mmSMU_MP1_SRBM2P_ARG_0); } -static int smu8_send_msg_to_smc_async(struct pp_hwmgr *hwmgr, uint16_t msg) +/* Send a message to the SMC, and wait for its response.*/ +static int smu8_send_msg_to_smc_with_parameter(struct pp_hwmgr *hwmgr, + uint16_t msg, uint32_t parameter) { int result = 0; + ktime_t t_start; + s64 elapsed_us; if (hwmgr == NULL || hwmgr->device == NULL) return -EINVAL; @@ -74,28 +79,31 @@ static int smu8_send_msg_to_smc_async(struct pp_hwmgr *hwmgr, uint16_t msg) /* Read the last message to SMU, to report actual cause */ uint32_t val = cgs_read_register(hwmgr->device, mmSMU_MP1_SRBM2P_MSG_0); - pr_err("smu8_send_msg_to_smc_async (0x%04x) failed\n", msg); - pr_err("SMU still servicing msg (0x%04x)\n", val); + pr_err("%s(0x%04x) aborted; SMU still servicing msg (0x%04x)\n", + __func__, msg, val); return result; } + t_start = ktime_get(); + + cgs_write_register(hwmgr->device, mmSMU_MP1_SRBM2P_ARG_0, parameter); cgs_write_register(hwmgr->device, mmSMU_MP1_SRBM2P_RESP_0, 0); cgs_write_register(hwmgr->device, mmSMU_MP1_SRBM2P_MSG_0, msg); - return 0; + result = PHM_WAIT_FIELD_UNEQUAL(hwmgr, + SMU_MP1_SRBM2P_RESP_0, CONTENT, 0); + + elapsed_us = ktime_us_delta(ktime_get(), t_start); + + WARN(result, "%s(0x%04x, %#x) timed out after %lld us\n", + __func__, msg, parameter, elapsed_us); + + return result; } -/* Send a message to the SMC, and wait for its response.*/ static int smu8_send_msg_to_smc(struct pp_hwmgr *hwmgr, uint16_t msg) { - int result = 0; - - result = smu8_send_msg_to_smc_async(hwmgr, msg); - if (result != 0) - return result; - - return PHM_WAIT_FIELD_UNEQUAL(hwmgr, - SMU_MP1_SRBM2P_RESP_0, CONTENT, 0); + return smu8_send_msg_to_smc_with_parameter(hwmgr, msg, 0); } static int smu8_set_smc_sram_address(struct pp_hwmgr *hwmgr, @@ -135,17 +143,6 @@ static int smu8_write_smc_sram_dword(struct pp_hwmgr *hwmgr, return result; } -static int smu8_send_msg_to_smc_with_parameter(struct pp_hwmgr *hwmgr, - uint16_t msg, uint32_t parameter) -{ - if (hwmgr == NULL || hwmgr->device == NULL) - return -EINVAL; - - cgs_write_register(hwmgr->device, mmSMU_MP1_SRBM2P_ARG_0, parameter); - - return smu8_send_msg_to_smc(hwmgr, msg); -} - static int smu8_check_fw_load_finish(struct pp_hwmgr *hwmgr, uint32_t firmware) { @@ -737,6 +734,10 @@ static int smu8_start_smu(struct pp_hwmgr *hwmgr) cgs_write_register(hwmgr->device, mmMP0PUB_IND_INDEX, index); hwmgr->smu_version = cgs_read_register(hwmgr->device, mmMP0PUB_IND_DATA); + pr_info("smu version %02d.%02d.%02d\n", + ((hwmgr->smu_version >> 16) & 0xFF), + ((hwmgr->smu_version >> 8) & 0xFF), + (hwmgr->smu_version & 0xFF)); adev->pm.fw_version = hwmgr->smu_version >> 8; return smu8_request_smu_load_fw(hwmgr); diff --git a/drivers/gpu/drm/amd/powerplay/smumgr/smu9_smumgr.c b/drivers/gpu/drm/amd/powerplay/smumgr/smu9_smumgr.c index 079fc8e8f709..742b3dc1f6cb 100644 --- a/drivers/gpu/drm/amd/powerplay/smumgr/smu9_smumgr.c +++ b/drivers/gpu/drm/amd/powerplay/smumgr/smu9_smumgr.c @@ -40,10 +40,8 @@ bool smu9_is_smc_ram_running(struct pp_hwmgr *hwmgr) struct amdgpu_device *adev = hwmgr->adev; uint32_t mp1_fw_flags; - WREG32_SOC15(NBIF, 0, mmPCIE_INDEX2, - (MP1_Public | (smnMP1_FIRMWARE_FLAGS & 0xffffffff))); - - mp1_fw_flags = RREG32_SOC15(NBIF, 0, mmPCIE_DATA2); + mp1_fw_flags = RREG32_PCIE(MP1_Public | + (smnMP1_FIRMWARE_FLAGS & 0xffffffff)); if (mp1_fw_flags & MP1_FIRMWARE_FLAGS__INTERRUPTS_ENABLED_MASK) return true; diff --git a/drivers/gpu/drm/amd/powerplay/smumgr/smumgr.c b/drivers/gpu/drm/amd/powerplay/smumgr/smumgr.c index a6edd5df33b0..4240aeec9000 100644 --- a/drivers/gpu/drm/amd/powerplay/smumgr/smumgr.c +++ b/drivers/gpu/drm/amd/powerplay/smumgr/smumgr.c @@ -29,6 +29,10 @@ #include <drm/amdgpu_drm.h> #include "smumgr.h" +MODULE_FIRMWARE("amdgpu/bonaire_smc.bin"); +MODULE_FIRMWARE("amdgpu/bonaire_k_smc.bin"); +MODULE_FIRMWARE("amdgpu/hawaii_smc.bin"); +MODULE_FIRMWARE("amdgpu/hawaii_k_smc.bin"); MODULE_FIRMWARE("amdgpu/topaz_smc.bin"); MODULE_FIRMWARE("amdgpu/topaz_k_smc.bin"); MODULE_FIRMWARE("amdgpu/tonga_smc.bin"); diff --git a/drivers/gpu/drm/amd/powerplay/smumgr/vega20_smumgr.c b/drivers/gpu/drm/amd/powerplay/smumgr/vega20_smumgr.c index b7ff7d4d6f44..ba00744c3413 100644 --- a/drivers/gpu/drm/amd/powerplay/smumgr/vega20_smumgr.c +++ b/drivers/gpu/drm/amd/powerplay/smumgr/vega20_smumgr.c @@ -49,10 +49,8 @@ static bool vega20_is_smc_ram_running(struct pp_hwmgr *hwmgr) struct amdgpu_device *adev = hwmgr->adev; uint32_t mp1_fw_flags; - WREG32_SOC15(NBIF, 0, mmPCIE_INDEX2, - (MP1_Public | (smnMP1_FIRMWARE_FLAGS & 0xffffffff))); - - mp1_fw_flags = RREG32_SOC15(NBIF, 0, mmPCIE_DATA2); + mp1_fw_flags = RREG32_PCIE(MP1_Public | + (smnMP1_FIRMWARE_FLAGS & 0xffffffff)); if ((mp1_fw_flags & MP1_FIRMWARE_FLAGS__INTERRUPTS_ENABLED_MASK) >> MP1_FIRMWARE_FLAGS__INTERRUPTS_ENABLED__SHIFT) diff --git a/drivers/gpu/drm/amd/powerplay/smumgr/vegam_smumgr.c b/drivers/gpu/drm/amd/powerplay/smumgr/vegam_smumgr.c index 9f71512b2510..1e69300f6175 100644 --- a/drivers/gpu/drm/amd/powerplay/smumgr/vegam_smumgr.c +++ b/drivers/gpu/drm/amd/powerplay/smumgr/vegam_smumgr.c @@ -40,7 +40,6 @@ #include "smu7_hwmgr.h" #include "hardwaremanager.h" -#include "ppatomctrl.h" #include "atombios.h" #include "pppcielanes.h" |