diff options
Diffstat (limited to 'drivers/gpu/drm/i915/gt/uc')
25 files changed, 7185 insertions, 0 deletions
diff --git a/drivers/gpu/drm/i915/gt/uc/Makefile b/drivers/gpu/drm/i915/gt/uc/Makefile new file mode 100644 index 000000000000..bec94d434cb6 --- /dev/null +++ b/drivers/gpu/drm/i915/gt/uc/Makefile @@ -0,0 +1,5 @@ +# For building individual subdir files on the command line +subdir-ccflags-y += -I$(srctree)/$(src)/../.. + +# Extra header tests +header-test-pattern-$(CONFIG_DRM_I915_WERROR) := *.h diff --git a/drivers/gpu/drm/i915/gt/uc/intel_guc.c b/drivers/gpu/drm/i915/gt/uc/intel_guc.c new file mode 100644 index 000000000000..249c747e9756 --- /dev/null +++ b/drivers/gpu/drm/i915/gt/uc/intel_guc.c @@ -0,0 +1,619 @@ +// SPDX-License-Identifier: MIT +/* + * Copyright © 2014-2019 Intel Corporation + */ + +#include "gt/intel_gt.h" +#include "intel_guc.h" +#include "intel_guc_ads.h" +#include "intel_guc_submission.h" +#include "i915_drv.h" + +static void gen8_guc_raise_irq(struct intel_guc *guc) +{ + struct intel_gt *gt = guc_to_gt(guc); + + intel_uncore_write(gt->uncore, GUC_SEND_INTERRUPT, GUC_SEND_TRIGGER); +} + +static void gen11_guc_raise_irq(struct intel_guc *guc) +{ + struct intel_gt *gt = guc_to_gt(guc); + + intel_uncore_write(gt->uncore, GEN11_GUC_HOST_INTERRUPT, 0); +} + +static inline i915_reg_t guc_send_reg(struct intel_guc *guc, u32 i) +{ + GEM_BUG_ON(!guc->send_regs.base); + GEM_BUG_ON(!guc->send_regs.count); + GEM_BUG_ON(i >= guc->send_regs.count); + + return _MMIO(guc->send_regs.base + 4 * i); +} + +void intel_guc_init_send_regs(struct intel_guc *guc) +{ + struct intel_gt *gt = guc_to_gt(guc); + enum forcewake_domains fw_domains = 0; + unsigned int i; + + if (INTEL_GEN(gt->i915) >= 11) { + guc->send_regs.base = + i915_mmio_reg_offset(GEN11_SOFT_SCRATCH(0)); + guc->send_regs.count = GEN11_SOFT_SCRATCH_COUNT; + } else { + guc->send_regs.base = i915_mmio_reg_offset(SOFT_SCRATCH(0)); + guc->send_regs.count = GUC_MAX_MMIO_MSG_LEN; + BUILD_BUG_ON(GUC_MAX_MMIO_MSG_LEN > SOFT_SCRATCH_COUNT); + } + + for (i = 0; i < guc->send_regs.count; i++) { + fw_domains |= intel_uncore_forcewake_for_reg(gt->uncore, + guc_send_reg(guc, i), + FW_REG_READ | FW_REG_WRITE); + } + guc->send_regs.fw_domains = fw_domains; +} + +void intel_guc_init_early(struct intel_guc *guc) +{ + struct drm_i915_private *i915 = guc_to_gt(guc)->i915; + + intel_guc_fw_init_early(guc); + intel_guc_ct_init_early(&guc->ct); + intel_guc_log_init_early(&guc->log); + intel_guc_submission_init_early(guc); + + mutex_init(&guc->send_mutex); + spin_lock_init(&guc->irq_lock); + guc->send = intel_guc_send_nop; + guc->handler = intel_guc_to_host_event_handler_nop; + if (INTEL_GEN(i915) >= 11) { + guc->notify = gen11_guc_raise_irq; + guc->interrupts.reset = gen11_reset_guc_interrupts; + guc->interrupts.enable = gen11_enable_guc_interrupts; + guc->interrupts.disable = gen11_disable_guc_interrupts; + } else { + guc->notify = gen8_guc_raise_irq; + guc->interrupts.reset = gen9_reset_guc_interrupts; + guc->interrupts.enable = gen9_enable_guc_interrupts; + guc->interrupts.disable = gen9_disable_guc_interrupts; + } +} + +static int guc_shared_data_create(struct intel_guc *guc) +{ + struct i915_vma *vma; + void *vaddr; + + vma = intel_guc_allocate_vma(guc, PAGE_SIZE); + if (IS_ERR(vma)) + return PTR_ERR(vma); + + vaddr = i915_gem_object_pin_map(vma->obj, I915_MAP_WB); + if (IS_ERR(vaddr)) { + i915_vma_unpin_and_release(&vma, 0); + return PTR_ERR(vaddr); + } + + guc->shared_data = vma; + guc->shared_data_vaddr = vaddr; + + return 0; +} + +static void guc_shared_data_destroy(struct intel_guc *guc) +{ + i915_vma_unpin_and_release(&guc->shared_data, I915_VMA_RELEASE_MAP); +} + +static u32 guc_ctl_debug_flags(struct intel_guc *guc) +{ + u32 level = intel_guc_log_get_level(&guc->log); + u32 flags = 0; + + if (!GUC_LOG_LEVEL_IS_VERBOSE(level)) + flags |= GUC_LOG_DISABLED; + else + flags |= GUC_LOG_LEVEL_TO_VERBOSITY(level) << + GUC_LOG_VERBOSITY_SHIFT; + + return flags; +} + +static u32 guc_ctl_feature_flags(struct intel_guc *guc) +{ + u32 flags = 0; + + if (!intel_guc_is_submission_supported(guc)) + flags |= GUC_CTL_DISABLE_SCHEDULER; + + return flags; +} + +static u32 guc_ctl_ctxinfo_flags(struct intel_guc *guc) +{ + u32 flags = 0; + + if (intel_guc_is_submission_supported(guc)) { + u32 ctxnum, base; + + base = intel_guc_ggtt_offset(guc, guc->stage_desc_pool); + ctxnum = GUC_MAX_STAGE_DESCRIPTORS / 16; + + base >>= PAGE_SHIFT; + flags |= (base << GUC_CTL_BASE_ADDR_SHIFT) | + (ctxnum << GUC_CTL_CTXNUM_IN16_SHIFT); + } + return flags; +} + +static u32 guc_ctl_log_params_flags(struct intel_guc *guc) +{ + u32 offset = intel_guc_ggtt_offset(guc, guc->log.vma) >> PAGE_SHIFT; + u32 flags; + + #if (((CRASH_BUFFER_SIZE) % SZ_1M) == 0) + #define UNIT SZ_1M + #define FLAG GUC_LOG_ALLOC_IN_MEGABYTE + #else + #define UNIT SZ_4K + #define FLAG 0 + #endif + + BUILD_BUG_ON(!CRASH_BUFFER_SIZE); + BUILD_BUG_ON(!IS_ALIGNED(CRASH_BUFFER_SIZE, UNIT)); + BUILD_BUG_ON(!DPC_BUFFER_SIZE); + BUILD_BUG_ON(!IS_ALIGNED(DPC_BUFFER_SIZE, UNIT)); + BUILD_BUG_ON(!ISR_BUFFER_SIZE); + BUILD_BUG_ON(!IS_ALIGNED(ISR_BUFFER_SIZE, UNIT)); + + BUILD_BUG_ON((CRASH_BUFFER_SIZE / UNIT - 1) > + (GUC_LOG_CRASH_MASK >> GUC_LOG_CRASH_SHIFT)); + BUILD_BUG_ON((DPC_BUFFER_SIZE / UNIT - 1) > + (GUC_LOG_DPC_MASK >> GUC_LOG_DPC_SHIFT)); + BUILD_BUG_ON((ISR_BUFFER_SIZE / UNIT - 1) > + (GUC_LOG_ISR_MASK >> GUC_LOG_ISR_SHIFT)); + + flags = GUC_LOG_VALID | + GUC_LOG_NOTIFY_ON_HALF_FULL | + FLAG | + ((CRASH_BUFFER_SIZE / UNIT - 1) << GUC_LOG_CRASH_SHIFT) | + ((DPC_BUFFER_SIZE / UNIT - 1) << GUC_LOG_DPC_SHIFT) | + ((ISR_BUFFER_SIZE / UNIT - 1) << GUC_LOG_ISR_SHIFT) | + (offset << GUC_LOG_BUF_ADDR_SHIFT); + + #undef UNIT + #undef FLAG + + return flags; +} + +static u32 guc_ctl_ads_flags(struct intel_guc *guc) +{ + u32 ads = intel_guc_ggtt_offset(guc, guc->ads_vma) >> PAGE_SHIFT; + u32 flags = ads << GUC_ADS_ADDR_SHIFT; + + return flags; +} + +/* + * Initialise the GuC parameter block before starting the firmware + * transfer. These parameters are read by the firmware on startup + * and cannot be changed thereafter. + */ +static void guc_init_params(struct intel_guc *guc) +{ + u32 *params = guc->params; + int i; + + BUILD_BUG_ON(sizeof(guc->params) != GUC_CTL_MAX_DWORDS * sizeof(u32)); + + params[GUC_CTL_CTXINFO] = guc_ctl_ctxinfo_flags(guc); + params[GUC_CTL_LOG_PARAMS] = guc_ctl_log_params_flags(guc); + params[GUC_CTL_FEATURE] = guc_ctl_feature_flags(guc); + params[GUC_CTL_DEBUG] = guc_ctl_debug_flags(guc); + params[GUC_CTL_ADS] = guc_ctl_ads_flags(guc); + + for (i = 0; i < GUC_CTL_MAX_DWORDS; i++) + DRM_DEBUG_DRIVER("param[%2d] = %#x\n", i, params[i]); +} + +/* + * Initialise the GuC parameter block before starting the firmware + * transfer. These parameters are read by the firmware on startup + * and cannot be changed thereafter. + */ +void intel_guc_write_params(struct intel_guc *guc) +{ + struct intel_uncore *uncore = guc_to_gt(guc)->uncore; + int i; + + /* + * All SOFT_SCRATCH registers are in FORCEWAKE_BLITTER domain and + * they are power context saved so it's ok to release forcewake + * when we are done here and take it again at xfer time. + */ + intel_uncore_forcewake_get(uncore, FORCEWAKE_BLITTER); + + intel_uncore_write(uncore, SOFT_SCRATCH(0), 0); + + for (i = 0; i < GUC_CTL_MAX_DWORDS; i++) + intel_uncore_write(uncore, SOFT_SCRATCH(1 + i), guc->params[i]); + + intel_uncore_forcewake_put(uncore, FORCEWAKE_BLITTER); +} + +int intel_guc_init(struct intel_guc *guc) +{ + struct intel_gt *gt = guc_to_gt(guc); + int ret; + + ret = intel_uc_fw_init(&guc->fw); + if (ret) + goto err_fetch; + + ret = guc_shared_data_create(guc); + if (ret) + goto err_fw; + GEM_BUG_ON(!guc->shared_data); + + ret = intel_guc_log_create(&guc->log); + if (ret) + goto err_shared; + + ret = intel_guc_ads_create(guc); + if (ret) + goto err_log; + GEM_BUG_ON(!guc->ads_vma); + + ret = intel_guc_ct_init(&guc->ct); + if (ret) + goto err_ads; + + if (intel_guc_is_submission_supported(guc)) { + /* + * This is stuff we need to have available at fw load time + * if we are planning to enable submission later + */ + ret = intel_guc_submission_init(guc); + if (ret) + goto err_ct; + } + + /* now that everything is perma-pinned, initialize the parameters */ + guc_init_params(guc); + + /* We need to notify the guc whenever we change the GGTT */ + i915_ggtt_enable_guc(gt->ggtt); + + return 0; + +err_ct: + intel_guc_ct_fini(&guc->ct); +err_ads: + intel_guc_ads_destroy(guc); +err_log: + intel_guc_log_destroy(&guc->log); +err_shared: + guc_shared_data_destroy(guc); +err_fw: + intel_uc_fw_fini(&guc->fw); +err_fetch: + intel_uc_fw_cleanup_fetch(&guc->fw); + DRM_DEV_DEBUG_DRIVER(gt->i915->drm.dev, "failed with %d\n", ret); + return ret; +} + +void intel_guc_fini(struct intel_guc *guc) +{ + struct intel_gt *gt = guc_to_gt(guc); + + if (!intel_uc_fw_is_available(&guc->fw)) + return; + + i915_ggtt_disable_guc(gt->ggtt); + + if (intel_guc_is_submission_supported(guc)) + intel_guc_submission_fini(guc); + + intel_guc_ct_fini(&guc->ct); + + intel_guc_ads_destroy(guc); + intel_guc_log_destroy(&guc->log); + guc_shared_data_destroy(guc); + intel_uc_fw_fini(&guc->fw); + intel_uc_fw_cleanup_fetch(&guc->fw); +} + +int intel_guc_send_nop(struct intel_guc *guc, const u32 *action, u32 len, + u32 *response_buf, u32 response_buf_size) +{ + WARN(1, "Unexpected send: action=%#x\n", *action); + return -ENODEV; +} + +void intel_guc_to_host_event_handler_nop(struct intel_guc *guc) +{ + WARN(1, "Unexpected event: no suitable handler\n"); +} + +/* + * This function implements the MMIO based host to GuC interface. + */ +int intel_guc_send_mmio(struct intel_guc *guc, const u32 *action, u32 len, + u32 *response_buf, u32 response_buf_size) +{ + struct intel_uncore *uncore = guc_to_gt(guc)->uncore; + u32 status; + int i; + int ret; + + GEM_BUG_ON(!len); + GEM_BUG_ON(len > guc->send_regs.count); + + /* We expect only action code */ + GEM_BUG_ON(*action & ~INTEL_GUC_MSG_CODE_MASK); + + /* If CT is available, we expect to use MMIO only during init/fini */ + GEM_BUG_ON(*action != INTEL_GUC_ACTION_REGISTER_COMMAND_TRANSPORT_BUFFER && + *action != INTEL_GUC_ACTION_DEREGISTER_COMMAND_TRANSPORT_BUFFER); + + mutex_lock(&guc->send_mutex); + intel_uncore_forcewake_get(uncore, guc->send_regs.fw_domains); + + for (i = 0; i < len; i++) + intel_uncore_write(uncore, guc_send_reg(guc, i), action[i]); + + intel_uncore_posting_read(uncore, guc_send_reg(guc, i - 1)); + + intel_guc_notify(guc); + + /* + * No GuC command should ever take longer than 10ms. + * Fast commands should still complete in 10us. + */ + ret = __intel_wait_for_register_fw(uncore, + guc_send_reg(guc, 0), + INTEL_GUC_MSG_TYPE_MASK, + INTEL_GUC_MSG_TYPE_RESPONSE << + INTEL_GUC_MSG_TYPE_SHIFT, + 10, 10, &status); + /* If GuC explicitly returned an error, convert it to -EIO */ + if (!ret && !INTEL_GUC_MSG_IS_RESPONSE_SUCCESS(status)) + ret = -EIO; + + if (ret) { + DRM_ERROR("MMIO: GuC action %#x failed with error %d %#x\n", + action[0], ret, status); + goto out; + } + + if (response_buf) { + int count = min(response_buf_size, guc->send_regs.count - 1); + + for (i = 0; i < count; i++) + response_buf[i] = intel_uncore_read(uncore, + guc_send_reg(guc, i + 1)); + } + + /* Use data from the GuC response as our return value */ + ret = INTEL_GUC_MSG_TO_DATA(status); + +out: + intel_uncore_forcewake_put(uncore, guc->send_regs.fw_domains); + mutex_unlock(&guc->send_mutex); + + return ret; +} + +int intel_guc_to_host_process_recv_msg(struct intel_guc *guc, + const u32 *payload, u32 len) +{ + u32 msg; + + if (unlikely(!len)) + return -EPROTO; + + /* Make sure to handle only enabled messages */ + msg = payload[0] & guc->msg_enabled_mask; + + if (msg & (INTEL_GUC_RECV_MSG_FLUSH_LOG_BUFFER | + INTEL_GUC_RECV_MSG_CRASH_DUMP_POSTED)) + intel_guc_log_handle_flush_event(&guc->log); + + return 0; +} + +int intel_guc_sample_forcewake(struct intel_guc *guc) +{ + struct drm_i915_private *dev_priv = guc_to_gt(guc)->i915; + u32 action[2]; + + action[0] = INTEL_GUC_ACTION_SAMPLE_FORCEWAKE; + /* WaRsDisableCoarsePowerGating:skl,cnl */ + if (!HAS_RC6(dev_priv) || NEEDS_WaRsDisableCoarsePowerGating(dev_priv)) + action[1] = 0; + else + /* bit 0 and 1 are for Render and Media domain separately */ + action[1] = GUC_FORCEWAKE_RENDER | GUC_FORCEWAKE_MEDIA; + + return intel_guc_send(guc, action, ARRAY_SIZE(action)); +} + +/** + * intel_guc_auth_huc() - Send action to GuC to authenticate HuC ucode + * @guc: intel_guc structure + * @rsa_offset: rsa offset w.r.t ggtt base of huc vma + * + * Triggers a HuC firmware authentication request to the GuC via intel_guc_send + * INTEL_GUC_ACTION_AUTHENTICATE_HUC interface. This function is invoked by + * intel_huc_auth(). + * + * Return: non-zero code on error + */ +int intel_guc_auth_huc(struct intel_guc *guc, u32 rsa_offset) +{ + u32 action[] = { + INTEL_GUC_ACTION_AUTHENTICATE_HUC, + rsa_offset + }; + + return intel_guc_send(guc, action, ARRAY_SIZE(action)); +} + +/** + * intel_guc_suspend() - notify GuC entering suspend state + * @guc: the guc + */ +int intel_guc_suspend(struct intel_guc *guc) +{ + struct intel_uncore *uncore = guc_to_gt(guc)->uncore; + int ret; + u32 status; + u32 action[] = { + INTEL_GUC_ACTION_ENTER_S_STATE, + GUC_POWER_D1, /* any value greater than GUC_POWER_D0 */ + }; + + /* + * The ENTER_S_STATE action queues the save/restore operation in GuC FW + * and then returns, so waiting on the H2G is not enough to guarantee + * GuC is done. When all the processing is done, GuC writes + * INTEL_GUC_SLEEP_STATE_SUCCESS to scratch register 14, so we can poll + * on that. Note that GuC does not ensure that the value in the register + * is different from INTEL_GUC_SLEEP_STATE_SUCCESS while the action is + * in progress so we need to take care of that ourselves as well. + */ + + intel_uncore_write(uncore, SOFT_SCRATCH(14), + INTEL_GUC_SLEEP_STATE_INVALID_MASK); + + ret = intel_guc_send(guc, action, ARRAY_SIZE(action)); + if (ret) + return ret; + + ret = __intel_wait_for_register(uncore, SOFT_SCRATCH(14), + INTEL_GUC_SLEEP_STATE_INVALID_MASK, + 0, 0, 10, &status); + if (ret) + return ret; + + if (status != INTEL_GUC_SLEEP_STATE_SUCCESS) { + DRM_ERROR("GuC failed to change sleep state. " + "action=0x%x, err=%u\n", + action[0], status); + return -EIO; + } + + return 0; +} + +/** + * intel_guc_reset_engine() - ask GuC to reset an engine + * @guc: intel_guc structure + * @engine: engine to be reset + */ +int intel_guc_reset_engine(struct intel_guc *guc, + struct intel_engine_cs *engine) +{ + u32 data[7]; + + GEM_BUG_ON(!guc->execbuf_client); + + data[0] = INTEL_GUC_ACTION_REQUEST_ENGINE_RESET; + data[1] = engine->guc_id; + data[2] = 0; + data[3] = 0; + data[4] = 0; + data[5] = guc->execbuf_client->stage_id; + data[6] = intel_guc_ggtt_offset(guc, guc->shared_data); + + return intel_guc_send(guc, data, ARRAY_SIZE(data)); +} + +/** + * intel_guc_resume() - notify GuC resuming from suspend state + * @guc: the guc + */ +int intel_guc_resume(struct intel_guc *guc) +{ + u32 action[] = { + INTEL_GUC_ACTION_EXIT_S_STATE, + GUC_POWER_D0, + }; + + return intel_guc_send(guc, action, ARRAY_SIZE(action)); +} + +/** + * DOC: GuC Address Space + * + * The layout of GuC address space is shown below: + * + * :: + * + * +===========> +====================+ <== FFFF_FFFF + * ^ | Reserved | + * | +====================+ <== GUC_GGTT_TOP + * | | | + * | | DRAM | + * GuC | | + * Address +===> +====================+ <== GuC ggtt_pin_bias + * Space ^ | | + * | | | | + * | GuC | GuC | + * | WOPCM | WOPCM | + * | Size | | + * | | | | + * v v | | + * +=======+===> +====================+ <== 0000_0000 + * + * The lower part of GuC Address Space [0, ggtt_pin_bias) is mapped to GuC WOPCM + * while upper part of GuC Address Space [ggtt_pin_bias, GUC_GGTT_TOP) is mapped + * to DRAM. The value of the GuC ggtt_pin_bias is the GuC WOPCM size. + */ + +/** + * intel_guc_allocate_vma() - Allocate a GGTT VMA for GuC usage + * @guc: the guc + * @size: size of area to allocate (both virtual space and memory) + * + * This is a wrapper to create an object for use with the GuC. In order to + * use it inside the GuC, an object needs to be pinned lifetime, so we allocate + * both some backing storage and a range inside the Global GTT. We must pin + * it in the GGTT somewhere other than than [0, GUC ggtt_pin_bias) because that + * range is reserved inside GuC. + * + * Return: A i915_vma if successful, otherwise an ERR_PTR. + */ +struct i915_vma *intel_guc_allocate_vma(struct intel_guc *guc, u32 size) +{ + struct intel_gt *gt = guc_to_gt(guc); + struct drm_i915_gem_object *obj; + struct i915_vma *vma; + u64 flags; + int ret; + + obj = i915_gem_object_create_shmem(gt->i915, size); + if (IS_ERR(obj)) + return ERR_CAST(obj); + + vma = i915_vma_instance(obj, >->ggtt->vm, NULL); + if (IS_ERR(vma)) + goto err; + + flags = PIN_GLOBAL | PIN_OFFSET_BIAS | i915_ggtt_pin_bias(vma); + ret = i915_vma_pin(vma, 0, 0, flags); + if (ret) { + vma = ERR_PTR(ret); + goto err; + } + + return i915_vma_make_unshrinkable(vma); + +err: + i915_gem_object_put(obj); + return vma; +} diff --git a/drivers/gpu/drm/i915/gt/uc/intel_guc.h b/drivers/gpu/drm/i915/gt/uc/intel_guc.h new file mode 100644 index 000000000000..2b2f046d3cc3 --- /dev/null +++ b/drivers/gpu/drm/i915/gt/uc/intel_guc.h @@ -0,0 +1,200 @@ +/* SPDX-License-Identifier: MIT */ +/* + * Copyright © 2014-2019 Intel Corporation + */ + +#ifndef _INTEL_GUC_H_ +#define _INTEL_GUC_H_ + +#include "intel_uncore.h" +#include "intel_guc_fw.h" +#include "intel_guc_fwif.h" +#include "intel_guc_ct.h" +#include "intel_guc_log.h" +#include "intel_guc_reg.h" +#include "intel_uc_fw.h" +#include "i915_utils.h" +#include "i915_vma.h" + +struct __guc_ads_blob; + +/* + * Top level structure of GuC. It handles firmware loading and manages client + * pool and doorbells. intel_guc owns a intel_guc_client to replace the legacy + * ExecList submission. + */ +struct intel_guc { + struct intel_uc_fw fw; + struct intel_guc_log log; + struct intel_guc_ct ct; + + /* intel_guc_recv interrupt related state */ + spinlock_t irq_lock; + unsigned int msg_enabled_mask; + + struct { + bool enabled; + void (*reset)(struct intel_guc *guc); + void (*enable)(struct intel_guc *guc); + void (*disable)(struct intel_guc *guc); + } interrupts; + + bool submission_supported; + + struct i915_vma *ads_vma; + struct __guc_ads_blob *ads_blob; + + struct i915_vma *stage_desc_pool; + void *stage_desc_pool_vaddr; + struct ida stage_ids; + struct i915_vma *shared_data; + void *shared_data_vaddr; + + struct intel_guc_client *execbuf_client; + + DECLARE_BITMAP(doorbell_bitmap, GUC_NUM_DOORBELLS); + /* Cyclic counter mod pagesize */ + u32 db_cacheline; + + /* Control params for fw initialization */ + u32 params[GUC_CTL_MAX_DWORDS]; + + /* GuC's FW specific registers used in MMIO send */ + struct { + u32 base; + unsigned int count; + enum forcewake_domains fw_domains; + } send_regs; + + /* Store msg (e.g. log flush) that we see while CTBs are disabled */ + u32 mmio_msg; + + /* To serialize the intel_guc_send actions */ + struct mutex send_mutex; + + /* GuC's FW specific send function */ + int (*send)(struct intel_guc *guc, const u32 *data, u32 len, + u32 *response_buf, u32 response_buf_size); + + /* GuC's FW specific event handler function */ + void (*handler)(struct intel_guc *guc); + + /* GuC's FW specific notify function */ + void (*notify)(struct intel_guc *guc); +}; + +static +inline int intel_guc_send(struct intel_guc *guc, const u32 *action, u32 len) +{ + return guc->send(guc, action, len, NULL, 0); +} + +static inline int +intel_guc_send_and_receive(struct intel_guc *guc, const u32 *action, u32 len, + u32 *response_buf, u32 response_buf_size) +{ + return guc->send(guc, action, len, response_buf, response_buf_size); +} + +static inline void intel_guc_notify(struct intel_guc *guc) +{ + guc->notify(guc); +} + +static inline void intel_guc_to_host_event_handler(struct intel_guc *guc) +{ + guc->handler(guc); +} + +/* GuC addresses above GUC_GGTT_TOP also don't map through the GTT */ +#define GUC_GGTT_TOP 0xFEE00000 + +/** + * intel_guc_ggtt_offset() - Get and validate the GGTT offset of @vma + * @guc: intel_guc structure. + * @vma: i915 graphics virtual memory area. + * + * GuC does not allow any gfx GGTT address that falls into range + * [0, ggtt.pin_bias), which is reserved for Boot ROM, SRAM and WOPCM. + * Currently, in order to exclude [0, ggtt.pin_bias) address space from + * GGTT, all gfx objects used by GuC are allocated with intel_guc_allocate_vma() + * and pinned with PIN_OFFSET_BIAS along with the value of ggtt.pin_bias. + * + * Return: GGTT offset of the @vma. + */ +static inline u32 intel_guc_ggtt_offset(struct intel_guc *guc, + struct i915_vma *vma) +{ + u32 offset = i915_ggtt_offset(vma); + + GEM_BUG_ON(offset < i915_ggtt_pin_bias(vma)); + GEM_BUG_ON(range_overflows_t(u64, offset, vma->size, GUC_GGTT_TOP)); + + return offset; +} + +void intel_guc_init_early(struct intel_guc *guc); +void intel_guc_init_send_regs(struct intel_guc *guc); +void intel_guc_write_params(struct intel_guc *guc); +int intel_guc_init(struct intel_guc *guc); +void intel_guc_fini(struct intel_guc *guc); +int intel_guc_send_nop(struct intel_guc *guc, const u32 *action, u32 len, + u32 *response_buf, u32 response_buf_size); +int intel_guc_send_mmio(struct intel_guc *guc, const u32 *action, u32 len, + u32 *response_buf, u32 response_buf_size); +void intel_guc_to_host_event_handler(struct intel_guc *guc); +void intel_guc_to_host_event_handler_nop(struct intel_guc *guc); +int intel_guc_to_host_process_recv_msg(struct intel_guc *guc, + const u32 *payload, u32 len); +int intel_guc_sample_forcewake(struct intel_guc *guc); +int intel_guc_auth_huc(struct intel_guc *guc, u32 rsa_offset); +int intel_guc_suspend(struct intel_guc *guc); +int intel_guc_resume(struct intel_guc *guc); +struct i915_vma *intel_guc_allocate_vma(struct intel_guc *guc, u32 size); + +static inline bool intel_guc_is_supported(struct intel_guc *guc) +{ + return intel_uc_fw_is_supported(&guc->fw); +} + +static inline bool intel_guc_is_enabled(struct intel_guc *guc) +{ + return intel_uc_fw_is_enabled(&guc->fw); +} + +static inline bool intel_guc_is_running(struct intel_guc *guc) +{ + return intel_uc_fw_is_running(&guc->fw); +} + +static inline int intel_guc_sanitize(struct intel_guc *guc) +{ + intel_uc_fw_sanitize(&guc->fw); + guc->mmio_msg = 0; + + return 0; +} + +static inline bool intel_guc_is_submission_supported(struct intel_guc *guc) +{ + return guc->submission_supported; +} + +static inline void intel_guc_enable_msg(struct intel_guc *guc, u32 mask) +{ + spin_lock_irq(&guc->irq_lock); + guc->msg_enabled_mask |= mask; + spin_unlock_irq(&guc->irq_lock); +} + +static inline void intel_guc_disable_msg(struct intel_guc *guc, u32 mask) +{ + spin_lock_irq(&guc->irq_lock); + guc->msg_enabled_mask &= ~mask; + spin_unlock_irq(&guc->irq_lock); +} + +int intel_guc_reset_engine(struct intel_guc *guc, + struct intel_engine_cs *engine); + +#endif diff --git a/drivers/gpu/drm/i915/gt/uc/intel_guc_ads.c b/drivers/gpu/drm/i915/gt/uc/intel_guc_ads.c new file mode 100644 index 000000000000..ca6674b8e00c --- /dev/null +++ b/drivers/gpu/drm/i915/gt/uc/intel_guc_ads.c @@ -0,0 +1,184 @@ +// SPDX-License-Identifier: MIT +/* + * Copyright © 2014-2019 Intel Corporation + */ + +#include "gt/intel_gt.h" +#include "intel_guc_ads.h" +#include "intel_uc.h" +#include "i915_drv.h" + +/* + * The Additional Data Struct (ADS) has pointers for different buffers used by + * the GuC. One single gem object contains the ADS struct itself (guc_ads), the + * scheduling policies (guc_policies), a structure describing a collection of + * register sets (guc_mmio_reg_state) and some extra pages for the GuC to save + * its internal state for sleep. + */ + +static void guc_policy_init(struct guc_policy *policy) +{ + policy->execution_quantum = POLICY_DEFAULT_EXECUTION_QUANTUM_US; + policy->preemption_time = POLICY_DEFAULT_PREEMPTION_TIME_US; + policy->fault_time = POLICY_DEFAULT_FAULT_TIME_US; + policy->policy_flags = 0; +} + +static void guc_policies_init(struct guc_policies *policies) +{ + struct guc_policy *policy; + u32 p, i; + + policies->dpc_promote_time = POLICY_DEFAULT_DPC_PROMOTE_TIME_US; + policies->max_num_work_items = POLICY_MAX_NUM_WI; + + for (p = 0; p < GUC_CLIENT_PRIORITY_NUM; p++) { + for (i = 0; i < GUC_MAX_ENGINE_CLASSES; i++) { + policy = &policies->policy[p][i]; + + guc_policy_init(policy); + } + } + + policies->is_valid = 1; +} + +static void guc_ct_pool_entries_init(struct guc_ct_pool_entry *pool, u32 num) +{ + memset(pool, 0, num * sizeof(*pool)); +} + +/* + * The first 80 dwords of the register state context, containing the + * execlists and ppgtt registers. + */ +#define LR_HW_CONTEXT_SIZE (80 * sizeof(u32)) + +/* The ads obj includes the struct itself and buffers passed to GuC */ +struct __guc_ads_blob { + struct guc_ads ads; + struct guc_policies policies; + struct guc_mmio_reg_state reg_state; + struct guc_gt_system_info system_info; + struct guc_clients_info clients_info; + struct guc_ct_pool_entry ct_pool[GUC_CT_POOL_SIZE]; + u8 reg_state_buffer[GUC_S3_SAVE_SPACE_PAGES * PAGE_SIZE]; +} __packed; + +static void __guc_ads_init(struct intel_guc *guc) +{ + struct drm_i915_private *dev_priv = guc_to_gt(guc)->i915; + struct __guc_ads_blob *blob = guc->ads_blob; + const u32 skipped_size = LRC_PPHWSP_SZ * PAGE_SIZE + LR_HW_CONTEXT_SIZE; + u32 base; + u8 engine_class; + + /* GuC scheduling policies */ + guc_policies_init(&blob->policies); + + /* + * GuC expects a per-engine-class context image and size + * (minus hwsp and ring context). The context image will be + * used to reinitialize engines after a reset. It must exist + * and be pinned in the GGTT, so that the address won't change after + * we have told GuC where to find it. The context size will be used + * to validate that the LRC base + size fall within allowed GGTT. + */ + for (engine_class = 0; engine_class <= MAX_ENGINE_CLASS; ++engine_class) { + if (engine_class == OTHER_CLASS) + continue; + /* + * TODO: Set context pointer to default state to allow + * GuC to re-init guilty contexts after internal reset. + */ + blob->ads.golden_context_lrca[engine_class] = 0; + blob->ads.eng_state_size[engine_class] = + intel_engine_context_size(dev_priv, engine_class) - + skipped_size; + } + + /* System info */ + blob->system_info.slice_enabled = hweight8(RUNTIME_INFO(dev_priv)->sseu.slice_mask); + blob->system_info.rcs_enabled = 1; + blob->system_info.bcs_enabled = 1; + + blob->system_info.vdbox_enable_mask = VDBOX_MASK(dev_priv); + blob->system_info.vebox_enable_mask = VEBOX_MASK(dev_priv); + blob->system_info.vdbox_sfc_support_mask = RUNTIME_INFO(dev_priv)->vdbox_sfc_access; + + base = intel_guc_ggtt_offset(guc, guc->ads_vma); + + /* Clients info */ + guc_ct_pool_entries_init(blob->ct_pool, ARRAY_SIZE(blob->ct_pool)); + + blob->clients_info.clients_num = 1; + blob->clients_info.ct_pool_addr = base + ptr_offset(blob, ct_pool); + blob->clients_info.ct_pool_count = ARRAY_SIZE(blob->ct_pool); + + /* ADS */ + blob->ads.scheduler_policies = base + ptr_offset(blob, policies); + blob->ads.reg_state_buffer = base + ptr_offset(blob, reg_state_buffer); + blob->ads.reg_state_addr = base + ptr_offset(blob, reg_state); + blob->ads.gt_system_info = base + ptr_offset(blob, system_info); + blob->ads.clients_info = base + ptr_offset(blob, clients_info); + + i915_gem_object_flush_map(guc->ads_vma->obj); +} + +/** + * intel_guc_ads_create() - allocates and initializes GuC ADS. + * @guc: intel_guc struct + * + * GuC needs memory block (Additional Data Struct), where it will store + * some data. Allocate and initialize such memory block for GuC use. + */ +int intel_guc_ads_create(struct intel_guc *guc) +{ + const u32 size = PAGE_ALIGN(sizeof(struct __guc_ads_blob)); + struct i915_vma *vma; + void *blob; + int ret; + + GEM_BUG_ON(guc->ads_vma); + + vma = intel_guc_allocate_vma(guc, size); + if (IS_ERR(vma)) + return PTR_ERR(vma); + + blob = i915_gem_object_pin_map(vma->obj, I915_MAP_WB); + if (IS_ERR(blob)) { + ret = PTR_ERR(blob); + goto err_vma; + } + + guc->ads_vma = vma; + guc->ads_blob = blob; + + __guc_ads_init(guc); + + return 0; + +err_vma: + i915_vma_unpin_and_release(&guc->ads_vma, 0); + return ret; +} + +void intel_guc_ads_destroy(struct intel_guc *guc) +{ + i915_vma_unpin_and_release(&guc->ads_vma, I915_VMA_RELEASE_MAP); +} + +/** + * intel_guc_ads_reset() - prepares GuC Additional Data Struct for reuse + * @guc: intel_guc struct + * + * GuC stores some data in ADS, which might be stale after a reset. + * Reinitialize whole ADS in case any part of it was corrupted during + * previous GuC run. + */ +void intel_guc_ads_reset(struct intel_guc *guc) +{ + if (!guc->ads_vma) + return; + __guc_ads_init(guc); +} diff --git a/drivers/gpu/drm/i915/gt/uc/intel_guc_ads.h b/drivers/gpu/drm/i915/gt/uc/intel_guc_ads.h new file mode 100644 index 000000000000..b00d3ae1113a --- /dev/null +++ b/drivers/gpu/drm/i915/gt/uc/intel_guc_ads.h @@ -0,0 +1,15 @@ +/* SPDX-License-Identifier: MIT */ +/* + * Copyright © 2014-2019 Intel Corporation + */ + +#ifndef _INTEL_GUC_ADS_H_ +#define _INTEL_GUC_ADS_H_ + +struct intel_guc; + +int intel_guc_ads_create(struct intel_guc *guc); +void intel_guc_ads_destroy(struct intel_guc *guc); +void intel_guc_ads_reset(struct intel_guc *guc); + +#endif diff --git a/drivers/gpu/drm/i915/gt/uc/intel_guc_ct.c b/drivers/gpu/drm/i915/gt/uc/intel_guc_ct.c new file mode 100644 index 000000000000..b49115517510 --- /dev/null +++ b/drivers/gpu/drm/i915/gt/uc/intel_guc_ct.c @@ -0,0 +1,897 @@ +// SPDX-License-Identifier: MIT +/* + * Copyright © 2016-2019 Intel Corporation + */ + +#include "i915_drv.h" +#include "intel_guc_ct.h" + +#ifdef CONFIG_DRM_I915_DEBUG_GUC +#define CT_DEBUG_DRIVER(...) DRM_DEBUG_DRIVER(__VA_ARGS__) +#else +#define CT_DEBUG_DRIVER(...) do { } while (0) +#endif + +struct ct_request { + struct list_head link; + u32 fence; + u32 status; + u32 response_len; + u32 *response_buf; +}; + +struct ct_incoming_request { + struct list_head link; + u32 msg[]; +}; + +enum { CTB_SEND = 0, CTB_RECV = 1 }; + +enum { CTB_OWNER_HOST = 0 }; + +static void ct_incoming_request_worker_func(struct work_struct *w); + +/** + * intel_guc_ct_init_early - Initialize CT state without requiring device access + * @ct: pointer to CT struct + */ +void intel_guc_ct_init_early(struct intel_guc_ct *ct) +{ + /* we're using static channel owners */ + ct->host_channel.owner = CTB_OWNER_HOST; + + spin_lock_init(&ct->lock); + INIT_LIST_HEAD(&ct->pending_requests); + INIT_LIST_HEAD(&ct->incoming_requests); + INIT_WORK(&ct->worker, ct_incoming_request_worker_func); +} + +static inline struct intel_guc *ct_to_guc(struct intel_guc_ct *ct) +{ + return container_of(ct, struct intel_guc, ct); +} + +static inline const char *guc_ct_buffer_type_to_str(u32 type) +{ + switch (type) { + case INTEL_GUC_CT_BUFFER_TYPE_SEND: + return "SEND"; + case INTEL_GUC_CT_BUFFER_TYPE_RECV: + return "RECV"; + default: + return "<invalid>"; + } +} + +static void guc_ct_buffer_desc_init(struct guc_ct_buffer_desc *desc, + u32 cmds_addr, u32 size, u32 owner) +{ + CT_DEBUG_DRIVER("CT: desc %p init addr=%#x size=%u owner=%u\n", + desc, cmds_addr, size, owner); + memset(desc, 0, sizeof(*desc)); + desc->addr = cmds_addr; + desc->size = size; + desc->owner = owner; +} + +static void guc_ct_buffer_desc_reset(struct guc_ct_buffer_desc *desc) +{ + CT_DEBUG_DRIVER("CT: desc %p reset head=%u tail=%u\n", + desc, desc->head, desc->tail); + desc->head = 0; + desc->tail = 0; + desc->is_in_error = 0; +} + +static int guc_action_register_ct_buffer(struct intel_guc *guc, + u32 desc_addr, + u32 type) +{ + u32 action[] = { + INTEL_GUC_ACTION_REGISTER_COMMAND_TRANSPORT_BUFFER, + desc_addr, + sizeof(struct guc_ct_buffer_desc), + type + }; + int err; + + /* Can't use generic send(), CT registration must go over MMIO */ + err = intel_guc_send_mmio(guc, action, ARRAY_SIZE(action), NULL, 0); + if (err) + DRM_ERROR("CT: register %s buffer failed; err=%d\n", + guc_ct_buffer_type_to_str(type), err); + return err; +} + +static int guc_action_deregister_ct_buffer(struct intel_guc *guc, + u32 owner, + u32 type) +{ + u32 action[] = { + INTEL_GUC_ACTION_DEREGISTER_COMMAND_TRANSPORT_BUFFER, + owner, + type + }; + int err; + + /* Can't use generic send(), CT deregistration must go over MMIO */ + err = intel_guc_send_mmio(guc, action, ARRAY_SIZE(action), NULL, 0); + if (err) + DRM_ERROR("CT: deregister %s buffer failed; owner=%d err=%d\n", + guc_ct_buffer_type_to_str(type), owner, err); + return err; +} + +static int ctch_init(struct intel_guc *guc, + struct intel_guc_ct_channel *ctch) +{ + struct i915_vma *vma; + void *blob; + int err; + int i; + + GEM_BUG_ON(ctch->vma); + + /* We allocate 1 page to hold both descriptors and both buffers. + * ___________..................... + * |desc (SEND)| : + * |___________| PAGE/4 + * :___________....................: + * |desc (RECV)| : + * |___________| PAGE/4 + * :_______________________________: + * |cmds (SEND) | + * | PAGE/4 + * |_______________________________| + * |cmds (RECV) | + * | PAGE/4 + * |_______________________________| + * + * Each message can use a maximum of 32 dwords and we don't expect to + * have more than 1 in flight at any time, so we have enough space. + * Some logic further ahead will rely on the fact that there is only 1 + * page and that it is always mapped, so if the size is changed the + * other code will need updating as well. + */ + + /* allocate vma */ + vma = intel_guc_allocate_vma(guc, PAGE_SIZE); + if (IS_ERR(vma)) { + err = PTR_ERR(vma); + goto err_out; + } + ctch->vma = vma; + + /* map first page */ + blob = i915_gem_object_pin_map(vma->obj, I915_MAP_WB); + if (IS_ERR(blob)) { + err = PTR_ERR(blob); + goto err_vma; + } + CT_DEBUG_DRIVER("CT: vma base=%#x\n", + intel_guc_ggtt_offset(guc, ctch->vma)); + + /* store pointers to desc and cmds */ + for (i = 0; i < ARRAY_SIZE(ctch->ctbs); i++) { + GEM_BUG_ON((i != CTB_SEND) && (i != CTB_RECV)); + ctch->ctbs[i].desc = blob + PAGE_SIZE/4 * i; + ctch->ctbs[i].cmds = blob + PAGE_SIZE/4 * i + PAGE_SIZE/2; + } + + return 0; + +err_vma: + i915_vma_unpin_and_release(&ctch->vma, 0); +err_out: + CT_DEBUG_DRIVER("CT: channel %d initialization failed; err=%d\n", + ctch->owner, err); + return err; +} + +static void ctch_fini(struct intel_guc *guc, + struct intel_guc_ct_channel *ctch) +{ + GEM_BUG_ON(ctch->enabled); + + i915_vma_unpin_and_release(&ctch->vma, I915_VMA_RELEASE_MAP); +} + +static int ctch_enable(struct intel_guc *guc, + struct intel_guc_ct_channel *ctch) +{ + u32 base; + int err; + int i; + + GEM_BUG_ON(!ctch->vma); + + GEM_BUG_ON(ctch->enabled); + + /* vma should be already allocated and map'ed */ + base = intel_guc_ggtt_offset(guc, ctch->vma); + + /* (re)initialize descriptors + * cmds buffers are in the second half of the blob page + */ + for (i = 0; i < ARRAY_SIZE(ctch->ctbs); i++) { + GEM_BUG_ON((i != CTB_SEND) && (i != CTB_RECV)); + guc_ct_buffer_desc_init(ctch->ctbs[i].desc, + base + PAGE_SIZE/4 * i + PAGE_SIZE/2, + PAGE_SIZE/4, + ctch->owner); + } + + /* register buffers, starting wirh RECV buffer + * descriptors are in first half of the blob + */ + err = guc_action_register_ct_buffer(guc, + base + PAGE_SIZE/4 * CTB_RECV, + INTEL_GUC_CT_BUFFER_TYPE_RECV); + if (unlikely(err)) + goto err_out; + + err = guc_action_register_ct_buffer(guc, + base + PAGE_SIZE/4 * CTB_SEND, + INTEL_GUC_CT_BUFFER_TYPE_SEND); + if (unlikely(err)) + goto err_deregister; + + ctch->enabled = true; + + return 0; + +err_deregister: + guc_action_deregister_ct_buffer(guc, + ctch->owner, + INTEL_GUC_CT_BUFFER_TYPE_RECV); +err_out: + DRM_ERROR("CT: can't open channel %d; err=%d\n", ctch->owner, err); + return err; +} + +static void ctch_disable(struct intel_guc *guc, + struct intel_guc_ct_channel *ctch) +{ + GEM_BUG_ON(!ctch->enabled); + + ctch->enabled = false; + + guc_action_deregister_ct_buffer(guc, + ctch->owner, + INTEL_GUC_CT_BUFFER_TYPE_SEND); + guc_action_deregister_ct_buffer(guc, + ctch->owner, + INTEL_GUC_CT_BUFFER_TYPE_RECV); +} + +static u32 ctch_get_next_fence(struct intel_guc_ct_channel *ctch) +{ + /* For now it's trivial */ + return ++ctch->next_fence; +} + +/** + * DOC: CTB Host to GuC request + * + * Format of the CTB Host to GuC request message is as follows:: + * + * +------------+---------+---------+---------+---------+ + * | msg[0] | [1] | [2] | ... | [n-1] | + * +------------+---------+---------+---------+---------+ + * | MESSAGE | MESSAGE PAYLOAD | + * + HEADER +---------+---------+---------+---------+ + * | | 0 | 1 | ... | n | + * +============+=========+=========+=========+=========+ + * | len >= 1 | FENCE | request specific data | + * +------+-----+---------+---------+---------+---------+ + * + * ^-----------------len-------------------^ + */ + +static int ctb_write(struct intel_guc_ct_buffer *ctb, + const u32 *action, + u32 len /* in dwords */, + u32 fence, + bool want_response) +{ + struct guc_ct_buffer_desc *desc = ctb->desc; + u32 head = desc->head / 4; /* in dwords */ + u32 tail = desc->tail / 4; /* in dwords */ + u32 size = desc->size / 4; /* in dwords */ + u32 used; /* in dwords */ + u32 header; + u32 *cmds = ctb->cmds; + unsigned int i; + + GEM_BUG_ON(desc->size % 4); + GEM_BUG_ON(desc->head % 4); + GEM_BUG_ON(desc->tail % 4); + GEM_BUG_ON(tail >= size); + + /* + * tail == head condition indicates empty. GuC FW does not support + * using up the entire buffer to get tail == head meaning full. + */ + if (tail < head) + used = (size - head) + tail; + else + used = tail - head; + + /* make sure there is a space including extra dw for the fence */ + if (unlikely(used + len + 1 >= size)) + return -ENOSPC; + + /* + * Write the message. The format is the following: + * DW0: header (including action code) + * DW1: fence + * DW2+: action data + */ + header = (len << GUC_CT_MSG_LEN_SHIFT) | + (GUC_CT_MSG_WRITE_FENCE_TO_DESC) | + (want_response ? GUC_CT_MSG_SEND_STATUS : 0) | + (action[0] << GUC_CT_MSG_ACTION_SHIFT); + + CT_DEBUG_DRIVER("CT: writing %*ph %*ph %*ph\n", + 4, &header, 4, &fence, + 4 * (len - 1), &action[1]); + + cmds[tail] = header; + tail = (tail + 1) % size; + + cmds[tail] = fence; + tail = (tail + 1) % size; + + for (i = 1; i < len; i++) { + cmds[tail] = action[i]; + tail = (tail + 1) % size; + } + + /* now update desc tail (back in bytes) */ + desc->tail = tail * 4; + GEM_BUG_ON(desc->tail > desc->size); + + return 0; +} + +/** + * wait_for_ctb_desc_update - Wait for the CT buffer descriptor update. + * @desc: buffer descriptor + * @fence: response fence + * @status: placeholder for status + * + * Guc will update CT buffer descriptor with new fence and status + * after processing the command identified by the fence. Wait for + * specified fence and then read from the descriptor status of the + * command. + * + * Return: + * * 0 response received (status is valid) + * * -ETIMEDOUT no response within hardcoded timeout + * * -EPROTO no response, CT buffer is in error + */ +static int wait_for_ctb_desc_update(struct guc_ct_buffer_desc *desc, + u32 fence, + u32 *status) +{ + int err; + + /* + * Fast commands should complete in less than 10us, so sample quickly + * up to that length of time, then switch to a slower sleep-wait loop. + * No GuC command should ever take longer than 10ms. + */ +#define done (READ_ONCE(desc->fence) == fence) + err = wait_for_us(done, 10); + if (err) + err = wait_for(done, 10); +#undef done + + if (unlikely(err)) { + DRM_ERROR("CT: fence %u failed; reported fence=%u\n", + fence, desc->fence); + + if (WARN_ON(desc->is_in_error)) { + /* Something went wrong with the messaging, try to reset + * the buffer and hope for the best + */ + guc_ct_buffer_desc_reset(desc); + err = -EPROTO; + } + } + + *status = desc->status; + return err; +} + +/** + * wait_for_ct_request_update - Wait for CT request state update. + * @req: pointer to pending request + * @status: placeholder for status + * + * For each sent request, Guc shall send bac CT response message. + * Our message handler will update status of tracked request once + * response message with given fence is received. Wait here and + * check for valid response status value. + * + * Return: + * * 0 response received (status is valid) + * * -ETIMEDOUT no response within hardcoded timeout + */ +static int wait_for_ct_request_update(struct ct_request *req, u32 *status) +{ + int err; + + /* + * Fast commands should complete in less than 10us, so sample quickly + * up to that length of time, then switch to a slower sleep-wait loop. + * No GuC command should ever take longer than 10ms. + */ +#define done INTEL_GUC_MSG_IS_RESPONSE(READ_ONCE(req->status)) + err = wait_for_us(done, 10); + if (err) + err = wait_for(done, 10); +#undef done + + if (unlikely(err)) + DRM_ERROR("CT: fence %u err %d\n", req->fence, err); + + *status = req->status; + return err; +} + +static int ctch_send(struct intel_guc_ct *ct, + struct intel_guc_ct_channel *ctch, + const u32 *action, + u32 len, + u32 *response_buf, + u32 response_buf_size, + u32 *status) +{ + struct intel_guc_ct_buffer *ctb = &ctch->ctbs[CTB_SEND]; + struct guc_ct_buffer_desc *desc = ctb->desc; + struct ct_request request; + unsigned long flags; + u32 fence; + int err; + + GEM_BUG_ON(!ctch->enabled); + GEM_BUG_ON(!len); + GEM_BUG_ON(len & ~GUC_CT_MSG_LEN_MASK); + GEM_BUG_ON(!response_buf && response_buf_size); + + fence = ctch_get_next_fence(ctch); + request.fence = fence; + request.status = 0; + request.response_len = response_buf_size; + request.response_buf = response_buf; + + spin_lock_irqsave(&ct->lock, flags); + list_add_tail(&request.link, &ct->pending_requests); + spin_unlock_irqrestore(&ct->lock, flags); + + err = ctb_write(ctb, action, len, fence, !!response_buf); + if (unlikely(err)) + goto unlink; + + intel_guc_notify(ct_to_guc(ct)); + + if (response_buf) + err = wait_for_ct_request_update(&request, status); + else + err = wait_for_ctb_desc_update(desc, fence, status); + if (unlikely(err)) + goto unlink; + + if (!INTEL_GUC_MSG_IS_RESPONSE_SUCCESS(*status)) { + err = -EIO; + goto unlink; + } + + if (response_buf) { + /* There shall be no data in the status */ + WARN_ON(INTEL_GUC_MSG_TO_DATA(request.status)); + /* Return actual response len */ + err = request.response_len; + } else { + /* There shall be no response payload */ + WARN_ON(request.response_len); + /* Return data decoded from the status dword */ + err = INTEL_GUC_MSG_TO_DATA(*status); + } + +unlink: + spin_lock_irqsave(&ct->lock, flags); + list_del(&request.link); + spin_unlock_irqrestore(&ct->lock, flags); + + return err; +} + +/* + * Command Transport (CT) buffer based GuC send function. + */ +int intel_guc_send_ct(struct intel_guc *guc, const u32 *action, u32 len, + u32 *response_buf, u32 response_buf_size) +{ + struct intel_guc_ct *ct = &guc->ct; + struct intel_guc_ct_channel *ctch = &ct->host_channel; + u32 status = ~0; /* undefined */ + int ret; + + mutex_lock(&guc->send_mutex); + + ret = ctch_send(ct, ctch, action, len, response_buf, response_buf_size, + &status); + if (unlikely(ret < 0)) { + DRM_ERROR("CT: send action %#X failed; err=%d status=%#X\n", + action[0], ret, status); + } else if (unlikely(ret)) { + CT_DEBUG_DRIVER("CT: send action %#x returned %d (%#x)\n", + action[0], ret, ret); + } + + mutex_unlock(&guc->send_mutex); + return ret; +} + +static inline unsigned int ct_header_get_len(u32 header) +{ + return (header >> GUC_CT_MSG_LEN_SHIFT) & GUC_CT_MSG_LEN_MASK; +} + +static inline unsigned int ct_header_get_action(u32 header) +{ + return (header >> GUC_CT_MSG_ACTION_SHIFT) & GUC_CT_MSG_ACTION_MASK; +} + +static inline bool ct_header_is_response(u32 header) +{ + return !!(header & GUC_CT_MSG_IS_RESPONSE); +} + +static int ctb_read(struct intel_guc_ct_buffer *ctb, u32 *data) +{ + struct guc_ct_buffer_desc *desc = ctb->desc; + u32 head = desc->head / 4; /* in dwords */ + u32 tail = desc->tail / 4; /* in dwords */ + u32 size = desc->size / 4; /* in dwords */ + u32 *cmds = ctb->cmds; + s32 available; /* in dwords */ + unsigned int len; + unsigned int i; + + GEM_BUG_ON(desc->size % 4); + GEM_BUG_ON(desc->head % 4); + GEM_BUG_ON(desc->tail % 4); + GEM_BUG_ON(tail >= size); + GEM_BUG_ON(head >= size); + + /* tail == head condition indicates empty */ + available = tail - head; + if (unlikely(available == 0)) + return -ENODATA; + + /* beware of buffer wrap case */ + if (unlikely(available < 0)) + available += size; + CT_DEBUG_DRIVER("CT: available %d (%u:%u)\n", available, head, tail); + GEM_BUG_ON(available < 0); + + data[0] = cmds[head]; + head = (head + 1) % size; + + /* message len with header */ + len = ct_header_get_len(data[0]) + 1; + if (unlikely(len > (u32)available)) { + DRM_ERROR("CT: incomplete message %*ph %*ph %*ph\n", + 4, data, + 4 * (head + available - 1 > size ? + size - head : available - 1), &cmds[head], + 4 * (head + available - 1 > size ? + available - 1 - size + head : 0), &cmds[0]); + return -EPROTO; + } + + for (i = 1; i < len; i++) { + data[i] = cmds[head]; + head = (head + 1) % size; + } + CT_DEBUG_DRIVER("CT: received %*ph\n", 4 * len, data); + + desc->head = head * 4; + return 0; +} + +/** + * DOC: CTB GuC to Host response + * + * Format of the CTB GuC to Host response message is as follows:: + * + * +------------+---------+---------+---------+---------+---------+ + * | msg[0] | [1] | [2] | [3] | ... | [n-1] | + * +------------+---------+---------+---------+---------+---------+ + * | MESSAGE | MESSAGE PAYLOAD | + * + HEADER +---------+---------+---------+---------+---------+ + * | | 0 | 1 | 2 | ... | n | + * +============+=========+=========+=========+=========+=========+ + * | len >= 2 | FENCE | STATUS | response specific data | + * +------+-----+---------+---------+---------+---------+---------+ + * + * ^-----------------------len-----------------------^ + */ + +static int ct_handle_response(struct intel_guc_ct *ct, const u32 *msg) +{ + u32 header = msg[0]; + u32 len = ct_header_get_len(header); + u32 msglen = len + 1; /* total message length including header */ + u32 fence; + u32 status; + u32 datalen; + struct ct_request *req; + bool found = false; + + GEM_BUG_ON(!ct_header_is_response(header)); + GEM_BUG_ON(!in_irq()); + + /* Response payload shall at least include fence and status */ + if (unlikely(len < 2)) { + DRM_ERROR("CT: corrupted response %*ph\n", 4 * msglen, msg); + return -EPROTO; + } + + fence = msg[1]; + status = msg[2]; + datalen = len - 2; + + /* Format of the status follows RESPONSE message */ + if (unlikely(!INTEL_GUC_MSG_IS_RESPONSE(status))) { + DRM_ERROR("CT: corrupted response %*ph\n", 4 * msglen, msg); + return -EPROTO; + } + + CT_DEBUG_DRIVER("CT: response fence %u status %#x\n", fence, status); + + spin_lock(&ct->lock); + list_for_each_entry(req, &ct->pending_requests, link) { + if (unlikely(fence != req->fence)) { + CT_DEBUG_DRIVER("CT: request %u awaits response\n", + req->fence); + continue; + } + if (unlikely(datalen > req->response_len)) { + DRM_ERROR("CT: response %u too long %*ph\n", + req->fence, 4 * msglen, msg); + datalen = 0; + } + if (datalen) + memcpy(req->response_buf, msg + 3, 4 * datalen); + req->response_len = datalen; + WRITE_ONCE(req->status, status); + found = true; + break; + } + spin_unlock(&ct->lock); + + if (!found) + DRM_ERROR("CT: unsolicited response %*ph\n", 4 * msglen, msg); + return 0; +} + +static void ct_process_request(struct intel_guc_ct *ct, + u32 action, u32 len, const u32 *payload) +{ + struct intel_guc *guc = ct_to_guc(ct); + int ret; + + CT_DEBUG_DRIVER("CT: request %x %*ph\n", action, 4 * len, payload); + + switch (action) { + case INTEL_GUC_ACTION_DEFAULT: + ret = intel_guc_to_host_process_recv_msg(guc, payload, len); + if (unlikely(ret)) + goto fail_unexpected; + break; + + default: +fail_unexpected: + DRM_ERROR("CT: unexpected request %x %*ph\n", + action, 4 * len, payload); + break; + } +} + +static bool ct_process_incoming_requests(struct intel_guc_ct *ct) +{ + unsigned long flags; + struct ct_incoming_request *request; + u32 header; + u32 *payload; + bool done; + + spin_lock_irqsave(&ct->lock, flags); + request = list_first_entry_or_null(&ct->incoming_requests, + struct ct_incoming_request, link); + if (request) + list_del(&request->link); + done = !!list_empty(&ct->incoming_requests); + spin_unlock_irqrestore(&ct->lock, flags); + + if (!request) + return true; + + header = request->msg[0]; + payload = &request->msg[1]; + ct_process_request(ct, + ct_header_get_action(header), + ct_header_get_len(header), + payload); + + kfree(request); + return done; +} + +static void ct_incoming_request_worker_func(struct work_struct *w) +{ + struct intel_guc_ct *ct = container_of(w, struct intel_guc_ct, worker); + bool done; + + done = ct_process_incoming_requests(ct); + if (!done) + queue_work(system_unbound_wq, &ct->worker); +} + +/** + * DOC: CTB GuC to Host request + * + * Format of the CTB GuC to Host request message is as follows:: + * + * +------------+---------+---------+---------+---------+---------+ + * | msg[0] | [1] | [2] | [3] | ... | [n-1] | + * +------------+---------+---------+---------+---------+---------+ + * | MESSAGE | MESSAGE PAYLOAD | + * + HEADER +---------+---------+---------+---------+---------+ + * | | 0 | 1 | 2 | ... | n | + * +============+=========+=========+=========+=========+=========+ + * | len | request specific data | + * +------+-----+---------+---------+---------+---------+---------+ + * + * ^-----------------------len-----------------------^ + */ + +static int ct_handle_request(struct intel_guc_ct *ct, const u32 *msg) +{ + u32 header = msg[0]; + u32 len = ct_header_get_len(header); + u32 msglen = len + 1; /* total message length including header */ + struct ct_incoming_request *request; + unsigned long flags; + + GEM_BUG_ON(ct_header_is_response(header)); + + request = kmalloc(sizeof(*request) + 4 * msglen, GFP_ATOMIC); + if (unlikely(!request)) { + DRM_ERROR("CT: dropping request %*ph\n", 4 * msglen, msg); + return 0; /* XXX: -ENOMEM ? */ + } + memcpy(request->msg, msg, 4 * msglen); + + spin_lock_irqsave(&ct->lock, flags); + list_add_tail(&request->link, &ct->incoming_requests); + spin_unlock_irqrestore(&ct->lock, flags); + + queue_work(system_unbound_wq, &ct->worker); + return 0; +} + +static void ct_process_host_channel(struct intel_guc_ct *ct) +{ + struct intel_guc_ct_channel *ctch = &ct->host_channel; + struct intel_guc_ct_buffer *ctb = &ctch->ctbs[CTB_RECV]; + u32 msg[GUC_CT_MSG_LEN_MASK + 1]; /* one extra dw for the header */ + int err = 0; + + if (!ctch->enabled) + return; + + do { + err = ctb_read(ctb, msg); + if (err) + break; + + if (ct_header_is_response(msg[0])) + err = ct_handle_response(ct, msg); + else + err = ct_handle_request(ct, msg); + } while (!err); + + if (GEM_WARN_ON(err == -EPROTO)) { + DRM_ERROR("CT: corrupted message detected!\n"); + ctb->desc->is_in_error = 1; + } +} + +/* + * When we're communicating with the GuC over CT, GuC uses events + * to notify us about new messages being posted on the RECV buffer. + */ +void intel_guc_to_host_event_handler_ct(struct intel_guc *guc) +{ + struct intel_guc_ct *ct = &guc->ct; + + ct_process_host_channel(ct); +} + +/** + * intel_guc_ct_init - Init CT communication + * @ct: pointer to CT struct + * + * Allocate memory required for communication via + * the CT channel. + * + * Return: 0 on success, a negative errno code on failure. + */ +int intel_guc_ct_init(struct intel_guc_ct *ct) +{ + struct intel_guc *guc = ct_to_guc(ct); + struct intel_guc_ct_channel *ctch = &ct->host_channel; + int err; + + err = ctch_init(guc, ctch); + if (unlikely(err)) { + DRM_ERROR("CT: can't open channel %d; err=%d\n", + ctch->owner, err); + return err; + } + + GEM_BUG_ON(!ctch->vma); + return 0; +} + +/** + * intel_guc_ct_fini - Fini CT communication + * @ct: pointer to CT struct + * + * Deallocate memory required for communication via + * the CT channel. + */ +void intel_guc_ct_fini(struct intel_guc_ct *ct) +{ + struct intel_guc *guc = ct_to_guc(ct); + struct intel_guc_ct_channel *ctch = &ct->host_channel; + + ctch_fini(guc, ctch); +} + +/** + * intel_guc_ct_enable - Enable buffer based command transport. + * @ct: pointer to CT struct + * + * Return: 0 on success, a negative errno code on failure. + */ +int intel_guc_ct_enable(struct intel_guc_ct *ct) +{ + struct intel_guc *guc = ct_to_guc(ct); + struct intel_guc_ct_channel *ctch = &ct->host_channel; + + if (ctch->enabled) + return 0; + + return ctch_enable(guc, ctch); +} + +/** + * intel_guc_ct_disable - Disable buffer based command transport. + * @ct: pointer to CT struct + */ +void intel_guc_ct_disable(struct intel_guc_ct *ct) +{ + struct intel_guc *guc = ct_to_guc(ct); + struct intel_guc_ct_channel *ctch = &ct->host_channel; + + if (!ctch->enabled) + return; + + ctch_disable(guc, ctch); +} diff --git a/drivers/gpu/drm/i915/gt/uc/intel_guc_ct.h b/drivers/gpu/drm/i915/gt/uc/intel_guc_ct.h new file mode 100644 index 000000000000..7c24d83f5c24 --- /dev/null +++ b/drivers/gpu/drm/i915/gt/uc/intel_guc_ct.h @@ -0,0 +1,93 @@ +/* SPDX-License-Identifier: MIT */ +/* + * Copyright © 2016-2019 Intel Corporation + */ + +#ifndef _INTEL_GUC_CT_H_ +#define _INTEL_GUC_CT_H_ + +#include <linux/spinlock.h> +#include <linux/workqueue.h> + +#include "intel_guc_fwif.h" + +struct i915_vma; +struct intel_guc; + +/** + * DOC: Command Transport (CT). + * + * Buffer based command transport is a replacement for MMIO based mechanism. + * It can be used to perform both host-2-guc and guc-to-host communication. + */ + +/** Represents single command transport buffer. + * + * A single command transport buffer consists of two parts, the header + * record (command transport buffer descriptor) and the actual buffer which + * holds the commands. + * + * @desc: pointer to the buffer descriptor + * @cmds: pointer to the commands buffer + */ +struct intel_guc_ct_buffer { + struct guc_ct_buffer_desc *desc; + u32 *cmds; +}; + +/** Represents pair of command transport buffers. + * + * Buffers go in pairs to allow bi-directional communication. + * To simplify the code we place both of them in the same vma. + * Buffers from the same pair must share unique owner id. + * + * @vma: pointer to the vma with pair of CT buffers + * @ctbs: buffers for sending(0) and receiving(1) commands + * @owner: unique identifier + * @next_fence: fence to be used with next send command + */ +struct intel_guc_ct_channel { + struct i915_vma *vma; + struct intel_guc_ct_buffer ctbs[2]; + u32 owner; + u32 next_fence; + bool enabled; +}; + +/** Holds all command transport channels. + * + * @host_channel: main channel used by the host + */ +struct intel_guc_ct { + struct intel_guc_ct_channel host_channel; + /* other channels are tbd */ + + /** @lock: protects pending requests list */ + spinlock_t lock; + + /** @pending_requests: list of requests waiting for response */ + struct list_head pending_requests; + + /** @incoming_requests: list of incoming requests */ + struct list_head incoming_requests; + + /** @worker: worker for handling incoming requests */ + struct work_struct worker; +}; + +void intel_guc_ct_init_early(struct intel_guc_ct *ct); +int intel_guc_ct_init(struct intel_guc_ct *ct); +void intel_guc_ct_fini(struct intel_guc_ct *ct); +int intel_guc_ct_enable(struct intel_guc_ct *ct); +void intel_guc_ct_disable(struct intel_guc_ct *ct); + +static inline void intel_guc_ct_stop(struct intel_guc_ct *ct) +{ + ct->host_channel.enabled = false; +} + +int intel_guc_send_ct(struct intel_guc *guc, const u32 *action, u32 len, + u32 *response_buf, u32 response_buf_size); +void intel_guc_to_host_event_handler_ct(struct intel_guc *guc); + +#endif /* _INTEL_GUC_CT_H_ */ diff --git a/drivers/gpu/drm/i915/gt/uc/intel_guc_fw.c b/drivers/gpu/drm/i915/gt/uc/intel_guc_fw.c new file mode 100644 index 000000000000..5528224448f6 --- /dev/null +++ b/drivers/gpu/drm/i915/gt/uc/intel_guc_fw.c @@ -0,0 +1,166 @@ +// SPDX-License-Identifier: MIT +/* + * Copyright © 2014-2019 Intel Corporation + * + * Authors: + * Vinit Azad <vinit.azad@intel.com> + * Ben Widawsky <ben@bwidawsk.net> + * Dave Gordon <david.s.gordon@intel.com> + * Alex Dai <yu.dai@intel.com> + */ + +#include "gt/intel_gt.h" +#include "intel_guc_fw.h" +#include "i915_drv.h" + +/** + * intel_guc_fw_init_early() - initializes GuC firmware struct + * @guc: intel_guc struct + * + * On platforms with GuC selects firmware for uploading + */ +void intel_guc_fw_init_early(struct intel_guc *guc) +{ + struct drm_i915_private *i915 = guc_to_gt(guc)->i915; + + intel_uc_fw_init_early(&guc->fw, INTEL_UC_FW_TYPE_GUC, HAS_GT_UC(i915), + INTEL_INFO(i915)->platform, INTEL_REVID(i915)); +} + +static void guc_prepare_xfer(struct intel_uncore *uncore) +{ + u32 shim_flags = GUC_DISABLE_SRAM_INIT_TO_ZEROES | + GUC_ENABLE_READ_CACHE_LOGIC | + GUC_ENABLE_MIA_CACHING | + GUC_ENABLE_READ_CACHE_FOR_SRAM_DATA | + GUC_ENABLE_READ_CACHE_FOR_WOPCM_DATA | + GUC_ENABLE_MIA_CLOCK_GATING; + + /* Must program this register before loading the ucode with DMA */ + intel_uncore_write(uncore, GUC_SHIM_CONTROL, shim_flags); + + if (IS_GEN9_LP(uncore->i915)) + intel_uncore_write(uncore, GEN9LP_GT_PM_CONFIG, GT_DOORBELL_ENABLE); + else + intel_uncore_write(uncore, GEN9_GT_PM_CONFIG, GT_DOORBELL_ENABLE); + + if (IS_GEN(uncore->i915, 9)) { + /* DOP Clock Gating Enable for GuC clocks */ + intel_uncore_rmw(uncore, GEN7_MISCCPCTL, + 0, GEN8_DOP_CLOCK_GATE_GUC_ENABLE); + + /* allows for 5us (in 10ns units) before GT can go to RC6 */ + intel_uncore_write(uncore, GUC_ARAT_C6DIS, 0x1FF); + } +} + +/* Copy RSA signature from the fw image to HW for verification */ +static void guc_xfer_rsa(struct intel_uc_fw *guc_fw, + struct intel_uncore *uncore) +{ + u32 rsa[UOS_RSA_SCRATCH_COUNT]; + size_t copied; + int i; + + copied = intel_uc_fw_copy_rsa(guc_fw, rsa, sizeof(rsa)); + GEM_BUG_ON(copied < sizeof(rsa)); + + for (i = 0; i < UOS_RSA_SCRATCH_COUNT; i++) + intel_uncore_write(uncore, UOS_RSA_SCRATCH(i), rsa[i]); +} + +/* + * Read the GuC status register (GUC_STATUS) and store it in the + * specified location; then return a boolean indicating whether + * the value matches either of two values representing completion + * of the GuC boot process. + * + * This is used for polling the GuC status in a wait_for() + * loop below. + */ +static inline bool guc_ready(struct intel_uncore *uncore, u32 *status) +{ + u32 val = intel_uncore_read(uncore, GUC_STATUS); + u32 uk_val = val & GS_UKERNEL_MASK; + + *status = val; + return (uk_val == GS_UKERNEL_READY) || + ((val & GS_MIA_CORE_STATE) && (uk_val == GS_UKERNEL_LAPIC_DONE)); +} + +static int guc_wait_ucode(struct intel_uncore *uncore) +{ + u32 status; + int ret; + + /* + * Wait for the GuC to start up. + * NB: Docs recommend not using the interrupt for completion. + * Measurements indicate this should take no more than 20ms, so a + * timeout here indicates that the GuC has failed and is unusable. + * (Higher levels of the driver may decide to reset the GuC and + * attempt the ucode load again if this happens.) + */ + ret = wait_for(guc_ready(uncore, &status), 100); + DRM_DEBUG_DRIVER("GuC status %#x\n", status); + + if ((status & GS_BOOTROM_MASK) == GS_BOOTROM_RSA_FAILED) { + DRM_ERROR("GuC firmware signature verification failed\n"); + ret = -ENOEXEC; + } + + if ((status & GS_UKERNEL_MASK) == GS_UKERNEL_EXCEPTION) { + DRM_ERROR("GuC firmware exception. EIP: %#x\n", + intel_uncore_read(uncore, SOFT_SCRATCH(13))); + ret = -ENXIO; + } + + return ret; +} + +/** + * intel_guc_fw_upload() - load GuC uCode to device + * @guc: intel_guc structure + * + * Called from intel_uc_init_hw() during driver load, resume from sleep and + * after a GPU reset. + * + * The firmware image should have already been fetched into memory, so only + * check that fetch succeeded, and then transfer the image to the h/w. + * + * Return: non-zero code on error + */ +int intel_guc_fw_upload(struct intel_guc *guc) +{ + struct intel_gt *gt = guc_to_gt(guc); + struct intel_uncore *uncore = gt->uncore; + int ret; + + guc_prepare_xfer(uncore); + + /* + * Note that GuC needs the CSS header plus uKernel code to be copied + * by the DMA engine in one operation, whereas the RSA signature is + * loaded via MMIO. + */ + guc_xfer_rsa(&guc->fw, uncore); + + /* + * Current uCode expects the code to be loaded at 8k; locations below + * this are used for the stack. + */ + ret = intel_uc_fw_upload(&guc->fw, gt, 0x2000, UOS_MOVE); + if (ret) + goto out; + + ret = guc_wait_ucode(uncore); + if (ret) + goto out; + + intel_uc_fw_change_status(&guc->fw, INTEL_UC_FIRMWARE_RUNNING); + return 0; + +out: + intel_uc_fw_change_status(&guc->fw, INTEL_UC_FIRMWARE_FAIL); + return ret; +} diff --git a/drivers/gpu/drm/i915/gt/uc/intel_guc_fw.h b/drivers/gpu/drm/i915/gt/uc/intel_guc_fw.h new file mode 100644 index 000000000000..b5ab639d7259 --- /dev/null +++ b/drivers/gpu/drm/i915/gt/uc/intel_guc_fw.h @@ -0,0 +1,14 @@ +/* SPDX-License-Identifier: MIT */ +/* + * Copyright © 2017-2019 Intel Corporation + */ + +#ifndef _INTEL_GUC_FW_H_ +#define _INTEL_GUC_FW_H_ + +struct intel_guc; + +void intel_guc_fw_init_early(struct intel_guc *guc); +int intel_guc_fw_upload(struct intel_guc *guc); + +#endif diff --git a/drivers/gpu/drm/i915/gt/uc/intel_guc_fwif.h b/drivers/gpu/drm/i915/gt/uc/intel_guc_fwif.h new file mode 100644 index 000000000000..1d3cdd67ca2f --- /dev/null +++ b/drivers/gpu/drm/i915/gt/uc/intel_guc_fwif.h @@ -0,0 +1,604 @@ +/* SPDX-License-Identifier: MIT */ +/* + * Copyright © 2014-2019 Intel Corporation + */ + +#ifndef _INTEL_GUC_FWIF_H +#define _INTEL_GUC_FWIF_H + +#include <linux/bits.h> +#include <linux/compiler.h> +#include <linux/types.h> + +#define GUC_CLIENT_PRIORITY_KMD_HIGH 0 +#define GUC_CLIENT_PRIORITY_HIGH 1 +#define GUC_CLIENT_PRIORITY_KMD_NORMAL 2 +#define GUC_CLIENT_PRIORITY_NORMAL 3 +#define GUC_CLIENT_PRIORITY_NUM 4 + +#define GUC_MAX_STAGE_DESCRIPTORS 1024 +#define GUC_INVALID_STAGE_ID GUC_MAX_STAGE_DESCRIPTORS + +#define GUC_RENDER_ENGINE 0 +#define GUC_VIDEO_ENGINE 1 +#define GUC_BLITTER_ENGINE 2 +#define GUC_VIDEOENHANCE_ENGINE 3 +#define GUC_VIDEO_ENGINE2 4 +#define GUC_MAX_ENGINES_NUM (GUC_VIDEO_ENGINE2 + 1) + +#define GUC_MAX_ENGINE_CLASSES 5 +#define GUC_MAX_INSTANCES_PER_CLASS 16 + +#define GUC_DOORBELL_INVALID 256 + +#define GUC_DB_SIZE (PAGE_SIZE) +#define GUC_WQ_SIZE (PAGE_SIZE * 2) + +/* Work queue item header definitions */ +#define WQ_STATUS_ACTIVE 1 +#define WQ_STATUS_SUSPENDED 2 +#define WQ_STATUS_CMD_ERROR 3 +#define WQ_STATUS_ENGINE_ID_NOT_USED 4 +#define WQ_STATUS_SUSPENDED_FROM_RESET 5 +#define WQ_TYPE_SHIFT 0 +#define WQ_TYPE_BATCH_BUF (0x1 << WQ_TYPE_SHIFT) +#define WQ_TYPE_PSEUDO (0x2 << WQ_TYPE_SHIFT) +#define WQ_TYPE_INORDER (0x3 << WQ_TYPE_SHIFT) +#define WQ_TYPE_NOOP (0x4 << WQ_TYPE_SHIFT) +#define WQ_TARGET_SHIFT 10 +#define WQ_LEN_SHIFT 16 +#define WQ_NO_WCFLUSH_WAIT (1 << 27) +#define WQ_PRESENT_WORKLOAD (1 << 28) + +#define WQ_RING_TAIL_SHIFT 20 +#define WQ_RING_TAIL_MAX 0x7FF /* 2^11 QWords */ +#define WQ_RING_TAIL_MASK (WQ_RING_TAIL_MAX << WQ_RING_TAIL_SHIFT) + +#define GUC_STAGE_DESC_ATTR_ACTIVE BIT(0) +#define GUC_STAGE_DESC_ATTR_PENDING_DB BIT(1) +#define GUC_STAGE_DESC_ATTR_KERNEL BIT(2) +#define GUC_STAGE_DESC_ATTR_PREEMPT BIT(3) +#define GUC_STAGE_DESC_ATTR_RESET BIT(4) +#define GUC_STAGE_DESC_ATTR_WQLOCKED BIT(5) +#define GUC_STAGE_DESC_ATTR_PCH BIT(6) +#define GUC_STAGE_DESC_ATTR_TERMINATED BIT(7) + +/* New GuC control data */ +#define GUC_CTL_CTXINFO 0 +#define GUC_CTL_CTXNUM_IN16_SHIFT 0 +#define GUC_CTL_BASE_ADDR_SHIFT 12 + +#define GUC_CTL_LOG_PARAMS 1 +#define GUC_LOG_VALID (1 << 0) +#define GUC_LOG_NOTIFY_ON_HALF_FULL (1 << 1) +#define GUC_LOG_ALLOC_IN_MEGABYTE (1 << 3) +#define GUC_LOG_CRASH_SHIFT 4 +#define GUC_LOG_CRASH_MASK (0x3 << GUC_LOG_CRASH_SHIFT) +#define GUC_LOG_DPC_SHIFT 6 +#define GUC_LOG_DPC_MASK (0x7 << GUC_LOG_DPC_SHIFT) +#define GUC_LOG_ISR_SHIFT 9 +#define GUC_LOG_ISR_MASK (0x7 << GUC_LOG_ISR_SHIFT) +#define GUC_LOG_BUF_ADDR_SHIFT 12 + +#define GUC_CTL_WA 2 +#define GUC_CTL_FEATURE 3 +#define GUC_CTL_DISABLE_SCHEDULER (1 << 14) + +#define GUC_CTL_DEBUG 4 +#define GUC_LOG_VERBOSITY_SHIFT 0 +#define GUC_LOG_VERBOSITY_LOW (0 << GUC_LOG_VERBOSITY_SHIFT) +#define GUC_LOG_VERBOSITY_MED (1 << GUC_LOG_VERBOSITY_SHIFT) +#define GUC_LOG_VERBOSITY_HIGH (2 << GUC_LOG_VERBOSITY_SHIFT) +#define GUC_LOG_VERBOSITY_ULTRA (3 << GUC_LOG_VERBOSITY_SHIFT) +/* Verbosity range-check limits, without the shift */ +#define GUC_LOG_VERBOSITY_MIN 0 +#define GUC_LOG_VERBOSITY_MAX 3 +#define GUC_LOG_VERBOSITY_MASK 0x0000000f +#define GUC_LOG_DESTINATION_MASK (3 << 4) +#define GUC_LOG_DISABLED (1 << 6) +#define GUC_PROFILE_ENABLED (1 << 7) + +#define GUC_CTL_ADS 5 +#define GUC_ADS_ADDR_SHIFT 1 +#define GUC_ADS_ADDR_MASK (0xFFFFF << GUC_ADS_ADDR_SHIFT) + +#define GUC_CTL_MAX_DWORDS (SOFT_SCRATCH_COUNT - 2) /* [1..14] */ + +/* Work item for submitting workloads into work queue of GuC. */ +struct guc_wq_item { + u32 header; + u32 context_desc; + u32 submit_element_info; + u32 fence_id; +} __packed; + +struct guc_process_desc { + u32 stage_id; + u64 db_base_addr; + u32 head; + u32 tail; + u32 error_offset; + u64 wq_base_addr; + u32 wq_size_bytes; + u32 wq_status; + u32 engine_presence; + u32 priority; + u32 reserved[30]; +} __packed; + +/* engine id and context id is packed into guc_execlist_context.context_id*/ +#define GUC_ELC_CTXID_OFFSET 0 +#define GUC_ELC_ENGINE_OFFSET 29 + +/* The execlist context including software and HW information */ +struct guc_execlist_context { + u32 context_desc; + u32 context_id; + u32 ring_status; + u32 ring_lrca; + u32 ring_begin; + u32 ring_end; + u32 ring_next_free_location; + u32 ring_current_tail_pointer_value; + u8 engine_state_submit_value; + u8 engine_state_wait_value; + u16 pagefault_count; + u16 engine_submit_queue_count; +} __packed; + +/* + * This structure describes a stage set arranged for a particular communication + * between uKernel (GuC) and Driver (KMD). Technically, this is known as a + * "GuC Context descriptor" in the specs, but we use the term "stage descriptor" + * to avoid confusion with all the other things already named "context" in the + * driver. A static pool of these descriptors are stored inside a GEM object + * (stage_desc_pool) which is held for the entire lifetime of our interaction + * with the GuC, being allocated before the GuC is loaded with its firmware. + */ +struct guc_stage_desc { + u32 sched_common_area; + u32 stage_id; + u32 pas_id; + u8 engines_used; + u64 db_trigger_cpu; + u32 db_trigger_uk; + u64 db_trigger_phy; + u16 db_id; + + struct guc_execlist_context lrc[GUC_MAX_ENGINES_NUM]; + + u8 attribute; + + u32 priority; + + u32 wq_sampled_tail_offset; + u32 wq_total_submit_enqueues; + + u32 process_desc; + u32 wq_addr; + u32 wq_size; + + u32 engine_presence; + + u8 engine_suspended; + + u8 reserved0[3]; + u64 reserved1[1]; + + u64 desc_private; +} __packed; + +/** + * DOC: CTB based communication + * + * The CTB (command transport buffer) communication between Host and GuC + * is based on u32 data stream written to the shared buffer. One buffer can + * be used to transmit data only in one direction (one-directional channel). + * + * Current status of the each buffer is stored in the buffer descriptor. + * Buffer descriptor holds tail and head fields that represents active data + * stream. The tail field is updated by the data producer (sender), and head + * field is updated by the data consumer (receiver):: + * + * +------------+ + * | DESCRIPTOR | +=================+============+========+ + * +============+ | | MESSAGE(s) | | + * | address |--------->+=================+============+========+ + * +------------+ + * | head | ^-----head--------^ + * +------------+ + * | tail | ^---------tail-----------------^ + * +------------+ + * | size | ^---------------size--------------------^ + * +------------+ + * + * Each message in data stream starts with the single u32 treated as a header, + * followed by optional set of u32 data that makes message specific payload:: + * + * +------------+---------+---------+---------+ + * | MESSAGE | + * +------------+---------+---------+---------+ + * | msg[0] | [1] | ... | [n-1] | + * +------------+---------+---------+---------+ + * | MESSAGE | MESSAGE PAYLOAD | + * + HEADER +---------+---------+---------+ + * | | 0 | ... | n | + * +======+=====+=========+=========+=========+ + * | 31:16| code| | | | + * +------+-----+ | | | + * | 15:5|flags| | | | + * +------+-----+ | | | + * | 4:0| len| | | | + * +------+-----+---------+---------+---------+ + * + * ^-------------len-------------^ + * + * The message header consists of: + * + * - **len**, indicates length of the message payload (in u32) + * - **code**, indicates message code + * - **flags**, holds various bits to control message handling + */ + +/* + * Describes single command transport buffer. + * Used by both guc-master and clients. + */ +struct guc_ct_buffer_desc { + u32 addr; /* gfx address */ + u64 host_private; /* host private data */ + u32 size; /* size in bytes */ + u32 head; /* offset updated by GuC*/ + u32 tail; /* offset updated by owner */ + u32 is_in_error; /* error indicator */ + u32 fence; /* fence updated by GuC */ + u32 status; /* status updated by GuC */ + u32 owner; /* id of the channel owner */ + u32 owner_sub_id; /* owner-defined field for extra tracking */ + u32 reserved[5]; +} __packed; + +/* Type of command transport buffer */ +#define INTEL_GUC_CT_BUFFER_TYPE_SEND 0x0u +#define INTEL_GUC_CT_BUFFER_TYPE_RECV 0x1u + +/* + * Definition of the command transport message header (DW0) + * + * bit[4..0] message len (in dwords) + * bit[7..5] reserved + * bit[8] response (G2H only) + * bit[8] write fence to desc (H2G only) + * bit[9] write status to H2G buff (H2G only) + * bit[10] send status back via G2H (H2G only) + * bit[15..11] reserved + * bit[31..16] action code + */ +#define GUC_CT_MSG_LEN_SHIFT 0 +#define GUC_CT_MSG_LEN_MASK 0x1F +#define GUC_CT_MSG_IS_RESPONSE (1 << 8) +#define GUC_CT_MSG_WRITE_FENCE_TO_DESC (1 << 8) +#define GUC_CT_MSG_WRITE_STATUS_TO_BUFF (1 << 9) +#define GUC_CT_MSG_SEND_STATUS (1 << 10) +#define GUC_CT_MSG_ACTION_SHIFT 16 +#define GUC_CT_MSG_ACTION_MASK 0xFFFF + +#define GUC_FORCEWAKE_RENDER (1 << 0) +#define GUC_FORCEWAKE_MEDIA (1 << 1) + +#define GUC_POWER_UNSPECIFIED 0 +#define GUC_POWER_D0 1 +#define GUC_POWER_D1 2 +#define GUC_POWER_D2 3 +#define GUC_POWER_D3 4 + +/* Scheduling policy settings */ + +/* Reset engine upon preempt failure */ +#define POLICY_RESET_ENGINE (1<<0) +/* Preempt to idle on quantum expiry */ +#define POLICY_PREEMPT_TO_IDLE (1<<1) + +#define POLICY_MAX_NUM_WI 15 +#define POLICY_DEFAULT_DPC_PROMOTE_TIME_US 500000 +#define POLICY_DEFAULT_EXECUTION_QUANTUM_US 1000000 +#define POLICY_DEFAULT_PREEMPTION_TIME_US 500000 +#define POLICY_DEFAULT_FAULT_TIME_US 250000 + +struct guc_policy { + /* Time for one workload to execute. (in micro seconds) */ + u32 execution_quantum; + /* Time to wait for a preemption request to completed before issuing a + * reset. (in micro seconds). */ + u32 preemption_time; + /* How much time to allow to run after the first fault is observed. + * Then preempt afterwards. (in micro seconds) */ + u32 fault_time; + u32 policy_flags; + u32 reserved[8]; +} __packed; + +struct guc_policies { + struct guc_policy policy[GUC_CLIENT_PRIORITY_NUM][GUC_MAX_ENGINE_CLASSES]; + u32 submission_queue_depth[GUC_MAX_ENGINE_CLASSES]; + /* In micro seconds. How much time to allow before DPC processing is + * called back via interrupt (to prevent DPC queue drain starving). + * Typically 1000s of micro seconds (example only, not granularity). */ + u32 dpc_promote_time; + + /* Must be set to take these new values. */ + u32 is_valid; + + /* Max number of WIs to process per call. A large value may keep CS + * idle. */ + u32 max_num_work_items; + + u32 reserved[4]; +} __packed; + +/* GuC MMIO reg state struct */ + + +#define GUC_REGSET_MAX_REGISTERS 64 +#define GUC_S3_SAVE_SPACE_PAGES 10 + +struct guc_mmio_reg { + u32 offset; + u32 value; + u32 flags; +#define GUC_REGSET_MASKED (1 << 0) +} __packed; + +struct guc_mmio_regset { + struct guc_mmio_reg registers[GUC_REGSET_MAX_REGISTERS]; + u32 values_valid; + u32 number_of_registers; +} __packed; + +/* GuC register sets */ +struct guc_mmio_reg_state { + struct guc_mmio_regset engine_reg[GUC_MAX_ENGINE_CLASSES][GUC_MAX_INSTANCES_PER_CLASS]; + u32 reserved[98]; +} __packed; + +/* HW info */ +struct guc_gt_system_info { + u32 slice_enabled; + u32 rcs_enabled; + u32 reserved0; + u32 bcs_enabled; + u32 vdbox_enable_mask; + u32 vdbox_sfc_support_mask; + u32 vebox_enable_mask; + u32 reserved[9]; +} __packed; + +/* Clients info */ +struct guc_ct_pool_entry { + struct guc_ct_buffer_desc desc; + u32 reserved[7]; +} __packed; + +#define GUC_CT_POOL_SIZE 2 + +struct guc_clients_info { + u32 clients_num; + u32 reserved0[13]; + u32 ct_pool_addr; + u32 ct_pool_count; + u32 reserved[4]; +} __packed; + +/* GuC Additional Data Struct */ +struct guc_ads { + u32 reg_state_addr; + u32 reg_state_buffer; + u32 scheduler_policies; + u32 gt_system_info; + u32 clients_info; + u32 control_data; + u32 golden_context_lrca[GUC_MAX_ENGINE_CLASSES]; + u32 eng_state_size[GUC_MAX_ENGINE_CLASSES]; + u32 reserved[16]; +} __packed; + +/* GuC logging structures */ + +enum guc_log_buffer_type { + GUC_ISR_LOG_BUFFER, + GUC_DPC_LOG_BUFFER, + GUC_CRASH_DUMP_LOG_BUFFER, + GUC_MAX_LOG_BUFFER +}; + +/** + * struct guc_log_buffer_state - GuC log buffer state + * + * Below state structure is used for coordination of retrieval of GuC firmware + * logs. Separate state is maintained for each log buffer type. + * read_ptr points to the location where i915 read last in log buffer and + * is read only for GuC firmware. write_ptr is incremented by GuC with number + * of bytes written for each log entry and is read only for i915. + * When any type of log buffer becomes half full, GuC sends a flush interrupt. + * GuC firmware expects that while it is writing to 2nd half of the buffer, + * first half would get consumed by Host and then get a flush completed + * acknowledgment from Host, so that it does not end up doing any overwrite + * causing loss of logs. So when buffer gets half filled & i915 has requested + * for interrupt, GuC will set flush_to_file field, set the sampled_write_ptr + * to the value of write_ptr and raise the interrupt. + * On receiving the interrupt i915 should read the buffer, clear flush_to_file + * field and also update read_ptr with the value of sample_write_ptr, before + * sending an acknowledgment to GuC. marker & version fields are for internal + * usage of GuC and opaque to i915. buffer_full_cnt field is incremented every + * time GuC detects the log buffer overflow. + */ +struct guc_log_buffer_state { + u32 marker[2]; + u32 read_ptr; + u32 write_ptr; + u32 size; + u32 sampled_write_ptr; + union { + struct { + u32 flush_to_file:1; + u32 buffer_full_cnt:4; + u32 reserved:27; + }; + u32 flags; + }; + u32 version; +} __packed; + +struct guc_ctx_report { + u32 report_return_status; + u32 reserved1[64]; + u32 affected_count; + u32 reserved2[2]; +} __packed; + +/* GuC Shared Context Data Struct */ +struct guc_shared_ctx_data { + u32 addr_of_last_preempted_data_low; + u32 addr_of_last_preempted_data_high; + u32 addr_of_last_preempted_data_high_tmp; + u32 padding; + u32 is_mapped_to_proxy; + u32 proxy_ctx_id; + u32 engine_reset_ctx_id; + u32 media_reset_count; + u32 reserved1[8]; + u32 uk_last_ctx_switch_reason; + u32 was_reset; + u32 lrca_gpu_addr; + u64 execlist_ctx; + u32 reserved2[66]; + struct guc_ctx_report preempt_ctx_report[GUC_MAX_ENGINES_NUM]; +} __packed; + +/** + * DOC: MMIO based communication + * + * The MMIO based communication between Host and GuC uses software scratch + * registers, where first register holds data treated as message header, + * and other registers are used to hold message payload. + * + * For Gen9+, GuC uses software scratch registers 0xC180-0xC1B8, + * but no H2G command takes more than 8 parameters and the GuC FW + * itself uses an 8-element array to store the H2G message. + * + * +-----------+---------+---------+---------+ + * | MMIO[0] | MMIO[1] | ... | MMIO[n] | + * +-----------+---------+---------+---------+ + * | header | optional payload | + * +======+====+=========+=========+=========+ + * | 31:28|type| | | | + * +------+----+ | | | + * | 27:16|data| | | | + * +------+----+ | | | + * | 15:0|code| | | | + * +------+----+---------+---------+---------+ + * + * The message header consists of: + * + * - **type**, indicates message type + * - **code**, indicates message code, is specific for **type** + * - **data**, indicates message data, optional, depends on **code** + * + * The following message **types** are supported: + * + * - **REQUEST**, indicates Host-to-GuC request, requested GuC action code + * must be priovided in **code** field. Optional action specific parameters + * can be provided in remaining payload registers or **data** field. + * + * - **RESPONSE**, indicates GuC-to-Host response from earlier GuC request, + * action response status will be provided in **code** field. Optional + * response data can be returned in remaining payload registers or **data** + * field. + */ + +#define GUC_MAX_MMIO_MSG_LEN 8 + +#define INTEL_GUC_MSG_TYPE_SHIFT 28 +#define INTEL_GUC_MSG_TYPE_MASK (0xF << INTEL_GUC_MSG_TYPE_SHIFT) +#define INTEL_GUC_MSG_DATA_SHIFT 16 +#define INTEL_GUC_MSG_DATA_MASK (0xFFF << INTEL_GUC_MSG_DATA_SHIFT) +#define INTEL_GUC_MSG_CODE_SHIFT 0 +#define INTEL_GUC_MSG_CODE_MASK (0xFFFF << INTEL_GUC_MSG_CODE_SHIFT) + +#define __INTEL_GUC_MSG_GET(T, m) \ + (((m) & INTEL_GUC_MSG_ ## T ## _MASK) >> INTEL_GUC_MSG_ ## T ## _SHIFT) +#define INTEL_GUC_MSG_TO_TYPE(m) __INTEL_GUC_MSG_GET(TYPE, m) +#define INTEL_GUC_MSG_TO_DATA(m) __INTEL_GUC_MSG_GET(DATA, m) +#define INTEL_GUC_MSG_TO_CODE(m) __INTEL_GUC_MSG_GET(CODE, m) + +enum intel_guc_msg_type { + INTEL_GUC_MSG_TYPE_REQUEST = 0x0, + INTEL_GUC_MSG_TYPE_RESPONSE = 0xF, +}; + +#define __INTEL_GUC_MSG_TYPE_IS(T, m) \ + (INTEL_GUC_MSG_TO_TYPE(m) == INTEL_GUC_MSG_TYPE_ ## T) +#define INTEL_GUC_MSG_IS_REQUEST(m) __INTEL_GUC_MSG_TYPE_IS(REQUEST, m) +#define INTEL_GUC_MSG_IS_RESPONSE(m) __INTEL_GUC_MSG_TYPE_IS(RESPONSE, m) + +enum intel_guc_action { + INTEL_GUC_ACTION_DEFAULT = 0x0, + INTEL_GUC_ACTION_REQUEST_PREEMPTION = 0x2, + INTEL_GUC_ACTION_REQUEST_ENGINE_RESET = 0x3, + INTEL_GUC_ACTION_ALLOCATE_DOORBELL = 0x10, + INTEL_GUC_ACTION_DEALLOCATE_DOORBELL = 0x20, + INTEL_GUC_ACTION_LOG_BUFFER_FILE_FLUSH_COMPLETE = 0x30, + INTEL_GUC_ACTION_FORCE_LOG_BUFFER_FLUSH = 0x302, + INTEL_GUC_ACTION_ENTER_S_STATE = 0x501, + INTEL_GUC_ACTION_EXIT_S_STATE = 0x502, + INTEL_GUC_ACTION_SLPC_REQUEST = 0x3003, + INTEL_GUC_ACTION_SAMPLE_FORCEWAKE = 0x3005, + INTEL_GUC_ACTION_AUTHENTICATE_HUC = 0x4000, + INTEL_GUC_ACTION_REGISTER_COMMAND_TRANSPORT_BUFFER = 0x4505, + INTEL_GUC_ACTION_DEREGISTER_COMMAND_TRANSPORT_BUFFER = 0x4506, + INTEL_GUC_ACTION_UK_LOG_ENABLE_LOGGING = 0x0E000, + INTEL_GUC_ACTION_LIMIT +}; + +enum intel_guc_preempt_options { + INTEL_GUC_PREEMPT_OPTION_DROP_WORK_Q = 0x4, + INTEL_GUC_PREEMPT_OPTION_DROP_SUBMIT_Q = 0x8, +}; + +enum intel_guc_report_status { + INTEL_GUC_REPORT_STATUS_UNKNOWN = 0x0, + INTEL_GUC_REPORT_STATUS_ACKED = 0x1, + INTEL_GUC_REPORT_STATUS_ERROR = 0x2, + INTEL_GUC_REPORT_STATUS_COMPLETE = 0x4, +}; + +enum intel_guc_sleep_state_status { + INTEL_GUC_SLEEP_STATE_SUCCESS = 0x1, + INTEL_GUC_SLEEP_STATE_PREEMPT_TO_IDLE_FAILED = 0x2, + INTEL_GUC_SLEEP_STATE_ENGINE_RESET_FAILED = 0x3 +#define INTEL_GUC_SLEEP_STATE_INVALID_MASK 0x80000000 +}; + +#define GUC_LOG_CONTROL_LOGGING_ENABLED (1 << 0) +#define GUC_LOG_CONTROL_VERBOSITY_SHIFT 4 +#define GUC_LOG_CONTROL_VERBOSITY_MASK (0xF << GUC_LOG_CONTROL_VERBOSITY_SHIFT) +#define GUC_LOG_CONTROL_DEFAULT_LOGGING (1 << 8) + +enum intel_guc_response_status { + INTEL_GUC_RESPONSE_STATUS_SUCCESS = 0x0, + INTEL_GUC_RESPONSE_STATUS_GENERIC_FAIL = 0xF000, +}; + +#define INTEL_GUC_MSG_IS_RESPONSE_SUCCESS(m) \ + (typecheck(u32, (m)) && \ + ((m) & (INTEL_GUC_MSG_TYPE_MASK | INTEL_GUC_MSG_CODE_MASK)) == \ + ((INTEL_GUC_MSG_TYPE_RESPONSE << INTEL_GUC_MSG_TYPE_SHIFT) | \ + (INTEL_GUC_RESPONSE_STATUS_SUCCESS << INTEL_GUC_MSG_CODE_SHIFT))) + +/* This action will be programmed in C1BC - SOFT_SCRATCH_15_REG */ +enum intel_guc_recv_message { + INTEL_GUC_RECV_MSG_CRASH_DUMP_POSTED = BIT(1), + INTEL_GUC_RECV_MSG_FLUSH_LOG_BUFFER = BIT(3) +}; + +#endif diff --git a/drivers/gpu/drm/i915/gt/uc/intel_guc_log.c b/drivers/gpu/drm/i915/gt/uc/intel_guc_log.c new file mode 100644 index 000000000000..36332064de9c --- /dev/null +++ b/drivers/gpu/drm/i915/gt/uc/intel_guc_log.c @@ -0,0 +1,646 @@ +// SPDX-License-Identifier: MIT +/* + * Copyright © 2014-2019 Intel Corporation + */ + +#include <linux/debugfs.h> + +#include "gt/intel_gt.h" +#include "i915_drv.h" +#include "i915_memcpy.h" +#include "intel_guc_log.h" + +static void guc_log_capture_logs(struct intel_guc_log *log); + +/** + * DOC: GuC firmware log + * + * Firmware log is enabled by setting i915.guc_log_level to the positive level. + * Log data is printed out via reading debugfs i915_guc_log_dump. Reading from + * i915_guc_load_status will print out firmware loading status and scratch + * registers value. + */ + +static int guc_action_flush_log_complete(struct intel_guc *guc) +{ + u32 action[] = { + INTEL_GUC_ACTION_LOG_BUFFER_FILE_FLUSH_COMPLETE + }; + + return intel_guc_send(guc, action, ARRAY_SIZE(action)); +} + +static int guc_action_flush_log(struct intel_guc *guc) +{ + u32 action[] = { + INTEL_GUC_ACTION_FORCE_LOG_BUFFER_FLUSH, + 0 + }; + + return intel_guc_send(guc, action, ARRAY_SIZE(action)); +} + +static int guc_action_control_log(struct intel_guc *guc, bool enable, + bool default_logging, u32 verbosity) +{ + u32 action[] = { + INTEL_GUC_ACTION_UK_LOG_ENABLE_LOGGING, + (enable ? GUC_LOG_CONTROL_LOGGING_ENABLED : 0) | + (verbosity << GUC_LOG_CONTROL_VERBOSITY_SHIFT) | + (default_logging ? GUC_LOG_CONTROL_DEFAULT_LOGGING : 0) + }; + + GEM_BUG_ON(verbosity > GUC_LOG_VERBOSITY_MAX); + + return intel_guc_send(guc, action, ARRAY_SIZE(action)); +} + +static inline struct intel_guc *log_to_guc(struct intel_guc_log *log) +{ + return container_of(log, struct intel_guc, log); +} + +static void guc_log_enable_flush_events(struct intel_guc_log *log) +{ + intel_guc_enable_msg(log_to_guc(log), + INTEL_GUC_RECV_MSG_FLUSH_LOG_BUFFER | + INTEL_GUC_RECV_MSG_CRASH_DUMP_POSTED); +} + +static void guc_log_disable_flush_events(struct intel_guc_log *log) +{ + intel_guc_disable_msg(log_to_guc(log), + INTEL_GUC_RECV_MSG_FLUSH_LOG_BUFFER | + INTEL_GUC_RECV_MSG_CRASH_DUMP_POSTED); +} + +/* + * Sub buffer switch callback. Called whenever relay has to switch to a new + * sub buffer, relay stays on the same sub buffer if 0 is returned. + */ +static int subbuf_start_callback(struct rchan_buf *buf, + void *subbuf, + void *prev_subbuf, + size_t prev_padding) +{ + /* + * Use no-overwrite mode by default, where relay will stop accepting + * new data if there are no empty sub buffers left. + * There is no strict synchronization enforced by relay between Consumer + * and Producer. In overwrite mode, there is a possibility of getting + * inconsistent/garbled data, the producer could be writing on to the + * same sub buffer from which Consumer is reading. This can't be avoided + * unless Consumer is fast enough and can always run in tandem with + * Producer. + */ + if (relay_buf_full(buf)) + return 0; + + return 1; +} + +/* + * file_create() callback. Creates relay file in debugfs. + */ +static struct dentry *create_buf_file_callback(const char *filename, + struct dentry *parent, + umode_t mode, + struct rchan_buf *buf, + int *is_global) +{ + struct dentry *buf_file; + + /* + * This to enable the use of a single buffer for the relay channel and + * correspondingly have a single file exposed to User, through which + * it can collect the logs in order without any post-processing. + * Need to set 'is_global' even if parent is NULL for early logging. + */ + *is_global = 1; + + if (!parent) + return NULL; + + buf_file = debugfs_create_file(filename, mode, + parent, buf, &relay_file_operations); + if (IS_ERR(buf_file)) + return NULL; + + return buf_file; +} + +/* + * file_remove() default callback. Removes relay file in debugfs. + */ +static int remove_buf_file_callback(struct dentry *dentry) +{ + debugfs_remove(dentry); + return 0; +} + +/* relay channel callbacks */ +static struct rchan_callbacks relay_callbacks = { + .subbuf_start = subbuf_start_callback, + .create_buf_file = create_buf_file_callback, + .remove_buf_file = remove_buf_file_callback, +}; + +static void guc_move_to_next_buf(struct intel_guc_log *log) +{ + /* + * Make sure the updates made in the sub buffer are visible when + * Consumer sees the following update to offset inside the sub buffer. + */ + smp_wmb(); + + /* All data has been written, so now move the offset of sub buffer. */ + relay_reserve(log->relay.channel, log->vma->obj->base.size); + + /* Switch to the next sub buffer */ + relay_flush(log->relay.channel); +} + +static void *guc_get_write_buffer(struct intel_guc_log *log) +{ + /* + * Just get the base address of a new sub buffer and copy data into it + * ourselves. NULL will be returned in no-overwrite mode, if all sub + * buffers are full. Could have used the relay_write() to indirectly + * copy the data, but that would have been bit convoluted, as we need to + * write to only certain locations inside a sub buffer which cannot be + * done without using relay_reserve() along with relay_write(). So its + * better to use relay_reserve() alone. + */ + return relay_reserve(log->relay.channel, 0); +} + +static bool guc_check_log_buf_overflow(struct intel_guc_log *log, + enum guc_log_buffer_type type, + unsigned int full_cnt) +{ + unsigned int prev_full_cnt = log->stats[type].sampled_overflow; + bool overflow = false; + + if (full_cnt != prev_full_cnt) { + overflow = true; + + log->stats[type].overflow = full_cnt; + log->stats[type].sampled_overflow += full_cnt - prev_full_cnt; + + if (full_cnt < prev_full_cnt) { + /* buffer_full_cnt is a 4 bit counter */ + log->stats[type].sampled_overflow += 16; + } + + dev_notice_ratelimited(guc_to_gt(log_to_guc(log))->i915->drm.dev, + "GuC log buffer overflow\n"); + } + + return overflow; +} + +static unsigned int guc_get_log_buffer_size(enum guc_log_buffer_type type) +{ + switch (type) { + case GUC_ISR_LOG_BUFFER: + return ISR_BUFFER_SIZE; + case GUC_DPC_LOG_BUFFER: + return DPC_BUFFER_SIZE; + case GUC_CRASH_DUMP_LOG_BUFFER: + return CRASH_BUFFER_SIZE; + default: + MISSING_CASE(type); + } + + return 0; +} + +static void guc_read_update_log_buffer(struct intel_guc_log *log) +{ + unsigned int buffer_size, read_offset, write_offset, bytes_to_copy, full_cnt; + struct guc_log_buffer_state *log_buf_state, *log_buf_snapshot_state; + struct guc_log_buffer_state log_buf_state_local; + enum guc_log_buffer_type type; + void *src_data, *dst_data; + bool new_overflow; + + mutex_lock(&log->relay.lock); + + if (WARN_ON(!intel_guc_log_relay_enabled(log))) + goto out_unlock; + + /* Get the pointer to shared GuC log buffer */ + log_buf_state = src_data = log->relay.buf_addr; + + /* Get the pointer to local buffer to store the logs */ + log_buf_snapshot_state = dst_data = guc_get_write_buffer(log); + + if (unlikely(!log_buf_snapshot_state)) { + /* + * Used rate limited to avoid deluge of messages, logs might be + * getting consumed by User at a slow rate. + */ + DRM_ERROR_RATELIMITED("no sub-buffer to capture logs\n"); + log->relay.full_count++; + + goto out_unlock; + } + + /* Actual logs are present from the 2nd page */ + src_data += PAGE_SIZE; + dst_data += PAGE_SIZE; + + for (type = GUC_ISR_LOG_BUFFER; type < GUC_MAX_LOG_BUFFER; type++) { + /* + * Make a copy of the state structure, inside GuC log buffer + * (which is uncached mapped), on the stack to avoid reading + * from it multiple times. + */ + memcpy(&log_buf_state_local, log_buf_state, + sizeof(struct guc_log_buffer_state)); + buffer_size = guc_get_log_buffer_size(type); + read_offset = log_buf_state_local.read_ptr; + write_offset = log_buf_state_local.sampled_write_ptr; + full_cnt = log_buf_state_local.buffer_full_cnt; + + /* Bookkeeping stuff */ + log->stats[type].flush += log_buf_state_local.flush_to_file; + new_overflow = guc_check_log_buf_overflow(log, type, full_cnt); + + /* Update the state of shared log buffer */ + log_buf_state->read_ptr = write_offset; + log_buf_state->flush_to_file = 0; + log_buf_state++; + + /* First copy the state structure in snapshot buffer */ + memcpy(log_buf_snapshot_state, &log_buf_state_local, + sizeof(struct guc_log_buffer_state)); + + /* + * The write pointer could have been updated by GuC firmware, + * after sending the flush interrupt to Host, for consistency + * set write pointer value to same value of sampled_write_ptr + * in the snapshot buffer. + */ + log_buf_snapshot_state->write_ptr = write_offset; + log_buf_snapshot_state++; + + /* Now copy the actual logs. */ + if (unlikely(new_overflow)) { + /* copy the whole buffer in case of overflow */ + read_offset = 0; + write_offset = buffer_size; + } else if (unlikely((read_offset > buffer_size) || + (write_offset > buffer_size))) { + DRM_ERROR("invalid log buffer state\n"); + /* copy whole buffer as offsets are unreliable */ + read_offset = 0; + write_offset = buffer_size; + } + + /* Just copy the newly written data */ + if (read_offset > write_offset) { + i915_memcpy_from_wc(dst_data, src_data, write_offset); + bytes_to_copy = buffer_size - read_offset; + } else { + bytes_to_copy = write_offset - read_offset; + } + i915_memcpy_from_wc(dst_data + read_offset, + src_data + read_offset, bytes_to_copy); + + src_data += buffer_size; + dst_data += buffer_size; + } + + guc_move_to_next_buf(log); + +out_unlock: + mutex_unlock(&log->relay.lock); +} + +static void capture_logs_work(struct work_struct *work) +{ + struct intel_guc_log *log = + container_of(work, struct intel_guc_log, relay.flush_work); + + guc_log_capture_logs(log); +} + +static int guc_log_map(struct intel_guc_log *log) +{ + void *vaddr; + + lockdep_assert_held(&log->relay.lock); + + if (!log->vma) + return -ENODEV; + + /* + * Create a WC (Uncached for read) vmalloc mapping of log + * buffer pages, so that we can directly get the data + * (up-to-date) from memory. + */ + vaddr = i915_gem_object_pin_map(log->vma->obj, I915_MAP_WC); + if (IS_ERR(vaddr)) + return PTR_ERR(vaddr); + + log->relay.buf_addr = vaddr; + + return 0; +} + +static void guc_log_unmap(struct intel_guc_log *log) +{ + lockdep_assert_held(&log->relay.lock); + + i915_gem_object_unpin_map(log->vma->obj); + log->relay.buf_addr = NULL; +} + +void intel_guc_log_init_early(struct intel_guc_log *log) +{ + mutex_init(&log->relay.lock); + INIT_WORK(&log->relay.flush_work, capture_logs_work); +} + +static int guc_log_relay_create(struct intel_guc_log *log) +{ + struct intel_guc *guc = log_to_guc(log); + struct drm_i915_private *dev_priv = guc_to_gt(guc)->i915; + struct rchan *guc_log_relay_chan; + size_t n_subbufs, subbuf_size; + int ret; + + lockdep_assert_held(&log->relay.lock); + GEM_BUG_ON(!log->vma); + + /* Keep the size of sub buffers same as shared log buffer */ + subbuf_size = log->vma->size; + + /* + * Store up to 8 snapshots, which is large enough to buffer sufficient + * boot time logs and provides enough leeway to User, in terms of + * latency, for consuming the logs from relay. Also doesn't take + * up too much memory. + */ + n_subbufs = 8; + + guc_log_relay_chan = relay_open("guc_log", + dev_priv->drm.primary->debugfs_root, + subbuf_size, n_subbufs, + &relay_callbacks, dev_priv); + if (!guc_log_relay_chan) { + DRM_ERROR("Couldn't create relay chan for GuC logging\n"); + + ret = -ENOMEM; + return ret; + } + + GEM_BUG_ON(guc_log_relay_chan->subbuf_size < subbuf_size); + log->relay.channel = guc_log_relay_chan; + + return 0; +} + +static void guc_log_relay_destroy(struct intel_guc_log *log) +{ + lockdep_assert_held(&log->relay.lock); + + relay_close(log->relay.channel); + log->relay.channel = NULL; +} + +static void guc_log_capture_logs(struct intel_guc_log *log) +{ + struct intel_guc *guc = log_to_guc(log); + struct drm_i915_private *dev_priv = guc_to_gt(guc)->i915; + intel_wakeref_t wakeref; + + guc_read_update_log_buffer(log); + + /* + * Generally device is expected to be active only at this + * time, so get/put should be really quick. + */ + with_intel_runtime_pm(&dev_priv->runtime_pm, wakeref) + guc_action_flush_log_complete(guc); +} + +static u32 __get_default_log_level(struct intel_guc_log *log) +{ + /* A negative value means "use platform/config default" */ + if (i915_modparams.guc_log_level < 0) { + return (IS_ENABLED(CONFIG_DRM_I915_DEBUG) || + IS_ENABLED(CONFIG_DRM_I915_DEBUG_GEM)) ? + GUC_LOG_LEVEL_MAX : GUC_LOG_LEVEL_NON_VERBOSE; + } + + if (i915_modparams.guc_log_level > GUC_LOG_LEVEL_MAX) { + DRM_WARN("Incompatible option detected: %s=%d, %s!\n", + "guc_log_level", i915_modparams.guc_log_level, + "verbosity too high"); + return (IS_ENABLED(CONFIG_DRM_I915_DEBUG) || + IS_ENABLED(CONFIG_DRM_I915_DEBUG_GEM)) ? + GUC_LOG_LEVEL_MAX : GUC_LOG_LEVEL_DISABLED; + } + + GEM_BUG_ON(i915_modparams.guc_log_level < GUC_LOG_LEVEL_DISABLED); + GEM_BUG_ON(i915_modparams.guc_log_level > GUC_LOG_LEVEL_MAX); + return i915_modparams.guc_log_level; +} + +int intel_guc_log_create(struct intel_guc_log *log) +{ + struct intel_guc *guc = log_to_guc(log); + struct i915_vma *vma; + u32 guc_log_size; + int ret; + + GEM_BUG_ON(log->vma); + + /* + * GuC Log buffer Layout + * + * +===============================+ 00B + * | Crash dump state header | + * +-------------------------------+ 32B + * | DPC state header | + * +-------------------------------+ 64B + * | ISR state header | + * +-------------------------------+ 96B + * | | + * +===============================+ PAGE_SIZE (4KB) + * | Crash Dump logs | + * +===============================+ + CRASH_SIZE + * | DPC logs | + * +===============================+ + DPC_SIZE + * | ISR logs | + * +===============================+ + ISR_SIZE + */ + guc_log_size = PAGE_SIZE + CRASH_BUFFER_SIZE + DPC_BUFFER_SIZE + + ISR_BUFFER_SIZE; + + vma = intel_guc_allocate_vma(guc, guc_log_size); + if (IS_ERR(vma)) { + ret = PTR_ERR(vma); + goto err; + } + + log->vma = vma; + + log->level = __get_default_log_level(log); + DRM_DEBUG_DRIVER("guc_log_level=%d (%s, verbose:%s, verbosity:%d)\n", + log->level, enableddisabled(log->level), + yesno(GUC_LOG_LEVEL_IS_VERBOSE(log->level)), + GUC_LOG_LEVEL_TO_VERBOSITY(log->level)); + + return 0; + +err: + DRM_ERROR("Failed to allocate GuC log buffer. %d\n", ret); + return ret; +} + +void intel_guc_log_destroy(struct intel_guc_log *log) +{ + i915_vma_unpin_and_release(&log->vma, 0); +} + +int intel_guc_log_set_level(struct intel_guc_log *log, u32 level) +{ + struct intel_guc *guc = log_to_guc(log); + struct drm_i915_private *dev_priv = guc_to_gt(guc)->i915; + intel_wakeref_t wakeref; + int ret = 0; + + BUILD_BUG_ON(GUC_LOG_VERBOSITY_MIN != 0); + GEM_BUG_ON(!log->vma); + + /* + * GuC is recognizing log levels starting from 0 to max, we're using 0 + * as indication that logging should be disabled. + */ + if (level < GUC_LOG_LEVEL_DISABLED || level > GUC_LOG_LEVEL_MAX) + return -EINVAL; + + mutex_lock(&dev_priv->drm.struct_mutex); + + if (log->level == level) + goto out_unlock; + + with_intel_runtime_pm(&dev_priv->runtime_pm, wakeref) + ret = guc_action_control_log(guc, + GUC_LOG_LEVEL_IS_VERBOSE(level), + GUC_LOG_LEVEL_IS_ENABLED(level), + GUC_LOG_LEVEL_TO_VERBOSITY(level)); + if (ret) { + DRM_DEBUG_DRIVER("guc_log_control action failed %d\n", ret); + goto out_unlock; + } + + log->level = level; + +out_unlock: + mutex_unlock(&dev_priv->drm.struct_mutex); + + return ret; +} + +bool intel_guc_log_relay_enabled(const struct intel_guc_log *log) +{ + return log->relay.buf_addr; +} + +int intel_guc_log_relay_open(struct intel_guc_log *log) +{ + int ret; + + if (!log->vma) + return -ENODEV; + + mutex_lock(&log->relay.lock); + + if (intel_guc_log_relay_enabled(log)) { + ret = -EEXIST; + goto out_unlock; + } + + /* + * We require SSE 4.1 for fast reads from the GuC log buffer and + * it should be present on the chipsets supporting GuC based + * submisssions. + */ + if (!i915_has_memcpy_from_wc()) { + ret = -ENXIO; + goto out_unlock; + } + + ret = guc_log_relay_create(log); + if (ret) + goto out_unlock; + + ret = guc_log_map(log); + if (ret) + goto out_relay; + + mutex_unlock(&log->relay.lock); + + guc_log_enable_flush_events(log); + + /* + * When GuC is logging without us relaying to userspace, we're ignoring + * the flush notification. This means that we need to unconditionally + * flush on relay enabling, since GuC only notifies us once. + */ + queue_work(system_highpri_wq, &log->relay.flush_work); + + return 0; + +out_relay: + guc_log_relay_destroy(log); +out_unlock: + mutex_unlock(&log->relay.lock); + + return ret; +} + +void intel_guc_log_relay_flush(struct intel_guc_log *log) +{ + struct intel_guc *guc = log_to_guc(log); + struct drm_i915_private *i915 = guc_to_gt(guc)->i915; + intel_wakeref_t wakeref; + + /* + * Before initiating the forceful flush, wait for any pending/ongoing + * flush to complete otherwise forceful flush may not actually happen. + */ + flush_work(&log->relay.flush_work); + + with_intel_runtime_pm(&i915->runtime_pm, wakeref) + guc_action_flush_log(guc); + + /* GuC would have updated log buffer by now, so capture it */ + guc_log_capture_logs(log); +} + +void intel_guc_log_relay_close(struct intel_guc_log *log) +{ + struct intel_guc *guc = log_to_guc(log); + struct drm_i915_private *i915 = guc_to_gt(guc)->i915; + + guc_log_disable_flush_events(log); + intel_synchronize_irq(i915); + + flush_work(&log->relay.flush_work); + + mutex_lock(&log->relay.lock); + GEM_BUG_ON(!intel_guc_log_relay_enabled(log)); + guc_log_unmap(log); + guc_log_relay_destroy(log); + mutex_unlock(&log->relay.lock); +} + +void intel_guc_log_handle_flush_event(struct intel_guc_log *log) +{ + queue_work(system_highpri_wq, &log->relay.flush_work); +} diff --git a/drivers/gpu/drm/i915/gt/uc/intel_guc_log.h b/drivers/gpu/drm/i915/gt/uc/intel_guc_log.h new file mode 100644 index 000000000000..6f764879acb1 --- /dev/null +++ b/drivers/gpu/drm/i915/gt/uc/intel_guc_log.h @@ -0,0 +1,80 @@ +/* SPDX-License-Identifier: MIT */ +/* + * Copyright © 2014-2019 Intel Corporation + */ + +#ifndef _INTEL_GUC_LOG_H_ +#define _INTEL_GUC_LOG_H_ + +#include <linux/mutex.h> +#include <linux/relay.h> +#include <linux/workqueue.h> + +#include "intel_guc_fwif.h" +#include "i915_gem.h" + +struct intel_guc; + +#ifdef CONFIG_DRM_I915_DEBUG_GUC +#define CRASH_BUFFER_SIZE SZ_2M +#define DPC_BUFFER_SIZE SZ_8M +#define ISR_BUFFER_SIZE SZ_8M +#else +#define CRASH_BUFFER_SIZE SZ_8K +#define DPC_BUFFER_SIZE SZ_32K +#define ISR_BUFFER_SIZE SZ_32K +#endif + +/* + * While we're using plain log level in i915, GuC controls are much more... + * "elaborate"? We have a couple of bits for verbosity, separate bit for actual + * log enabling, and separate bit for default logging - which "conveniently" + * ignores the enable bit. + */ +#define GUC_LOG_LEVEL_DISABLED 0 +#define GUC_LOG_LEVEL_NON_VERBOSE 1 +#define GUC_LOG_LEVEL_IS_ENABLED(x) ((x) > GUC_LOG_LEVEL_DISABLED) +#define GUC_LOG_LEVEL_IS_VERBOSE(x) ((x) > GUC_LOG_LEVEL_NON_VERBOSE) +#define GUC_LOG_LEVEL_TO_VERBOSITY(x) ({ \ + typeof(x) _x = (x); \ + GUC_LOG_LEVEL_IS_VERBOSE(_x) ? _x - 2 : 0; \ +}) +#define GUC_VERBOSITY_TO_LOG_LEVEL(x) ((x) + 2) +#define GUC_LOG_LEVEL_MAX GUC_VERBOSITY_TO_LOG_LEVEL(GUC_LOG_VERBOSITY_MAX) + +struct intel_guc_log { + u32 level; + struct i915_vma *vma; + struct { + void *buf_addr; + struct work_struct flush_work; + struct rchan *channel; + struct mutex lock; + u32 full_count; + } relay; + /* logging related stats */ + struct { + u32 sampled_overflow; + u32 overflow; + u32 flush; + } stats[GUC_MAX_LOG_BUFFER]; +}; + +void intel_guc_log_init_early(struct intel_guc_log *log); +int intel_guc_log_create(struct intel_guc_log *log); +void intel_guc_log_destroy(struct intel_guc_log *log); + +int intel_guc_log_set_level(struct intel_guc_log *log, u32 level); +bool intel_guc_log_relay_enabled(const struct intel_guc_log *log); +int intel_guc_log_relay_open(struct intel_guc_log *log); +void intel_guc_log_relay_flush(struct intel_guc_log *log); +void intel_guc_log_relay_close(struct intel_guc_log *log); + +void intel_guc_log_handle_flush_event(struct intel_guc_log *log); + +static inline u32 intel_guc_log_get_level(struct intel_guc_log *log) +{ + return log->level; +} + +#endif diff --git a/drivers/gpu/drm/i915/gt/uc/intel_guc_reg.h b/drivers/gpu/drm/i915/gt/uc/intel_guc_reg.h new file mode 100644 index 000000000000..edf194d23c6b --- /dev/null +++ b/drivers/gpu/drm/i915/gt/uc/intel_guc_reg.h @@ -0,0 +1,143 @@ +/* SPDX-License-Identifier: MIT */ +/* + * Copyright © 2014-2019 Intel Corporation + */ + +#ifndef _INTEL_GUC_REG_H_ +#define _INTEL_GUC_REG_H_ + +#include <linux/compiler.h> +#include <linux/types.h> + +#include "i915_reg.h" + +/* Definitions of GuC H/W registers, bits, etc */ + +#define GUC_STATUS _MMIO(0xc000) +#define GS_RESET_SHIFT 0 +#define GS_MIA_IN_RESET (0x01 << GS_RESET_SHIFT) +#define GS_BOOTROM_SHIFT 1 +#define GS_BOOTROM_MASK (0x7F << GS_BOOTROM_SHIFT) +#define GS_BOOTROM_RSA_FAILED (0x50 << GS_BOOTROM_SHIFT) +#define GS_BOOTROM_JUMP_PASSED (0x76 << GS_BOOTROM_SHIFT) +#define GS_UKERNEL_SHIFT 8 +#define GS_UKERNEL_MASK (0xFF << GS_UKERNEL_SHIFT) +#define GS_UKERNEL_LAPIC_DONE (0x30 << GS_UKERNEL_SHIFT) +#define GS_UKERNEL_DPC_ERROR (0x60 << GS_UKERNEL_SHIFT) +#define GS_UKERNEL_EXCEPTION (0x70 << GS_UKERNEL_SHIFT) +#define GS_UKERNEL_READY (0xF0 << GS_UKERNEL_SHIFT) +#define GS_MIA_SHIFT 16 +#define GS_MIA_MASK (0x07 << GS_MIA_SHIFT) +#define GS_MIA_CORE_STATE (0x01 << GS_MIA_SHIFT) +#define GS_MIA_HALT_REQUESTED (0x02 << GS_MIA_SHIFT) +#define GS_MIA_ISR_ENTRY (0x04 << GS_MIA_SHIFT) +#define GS_AUTH_STATUS_SHIFT 30 +#define GS_AUTH_STATUS_MASK (0x03 << GS_AUTH_STATUS_SHIFT) +#define GS_AUTH_STATUS_BAD (0x01 << GS_AUTH_STATUS_SHIFT) +#define GS_AUTH_STATUS_GOOD (0x02 << GS_AUTH_STATUS_SHIFT) + +#define SOFT_SCRATCH(n) _MMIO(0xc180 + (n) * 4) +#define SOFT_SCRATCH_COUNT 16 + +#define GEN11_SOFT_SCRATCH(n) _MMIO(0x190240 + (n) * 4) +#define GEN11_SOFT_SCRATCH_COUNT 4 + +#define UOS_RSA_SCRATCH(i) _MMIO(0xc200 + (i) * 4) +#define UOS_RSA_SCRATCH_COUNT 64 + +#define DMA_ADDR_0_LOW _MMIO(0xc300) +#define DMA_ADDR_0_HIGH _MMIO(0xc304) +#define DMA_ADDR_1_LOW _MMIO(0xc308) +#define DMA_ADDR_1_HIGH _MMIO(0xc30c) +#define DMA_ADDRESS_SPACE_WOPCM (7 << 16) +#define DMA_ADDRESS_SPACE_GTT (8 << 16) +#define DMA_COPY_SIZE _MMIO(0xc310) +#define DMA_CTRL _MMIO(0xc314) +#define HUC_UKERNEL (1<<9) +#define UOS_MOVE (1<<4) +#define START_DMA (1<<0) +#define DMA_GUC_WOPCM_OFFSET _MMIO(0xc340) +#define GUC_WOPCM_OFFSET_VALID (1<<0) +#define HUC_LOADING_AGENT_VCR (0<<1) +#define HUC_LOADING_AGENT_GUC (1<<1) +#define GUC_WOPCM_OFFSET_SHIFT 14 +#define GUC_WOPCM_OFFSET_MASK (0x3ffff << GUC_WOPCM_OFFSET_SHIFT) +#define GUC_MAX_IDLE_COUNT _MMIO(0xC3E4) + +#define HUC_STATUS2 _MMIO(0xD3B0) +#define HUC_FW_VERIFIED (1<<7) + +#define GEN11_HUC_KERNEL_LOAD_INFO _MMIO(0xC1DC) +#define HUC_LOAD_SUCCESSFUL (1 << 0) + +#define GUC_WOPCM_SIZE _MMIO(0xc050) +#define GUC_WOPCM_SIZE_LOCKED (1<<0) +#define GUC_WOPCM_SIZE_SHIFT 12 +#define GUC_WOPCM_SIZE_MASK (0xfffff << GUC_WOPCM_SIZE_SHIFT) + +#define GEN8_GT_PM_CONFIG _MMIO(0x138140) +#define GEN9LP_GT_PM_CONFIG _MMIO(0x138140) +#define GEN9_GT_PM_CONFIG _MMIO(0x13816c) +#define GT_DOORBELL_ENABLE (1<<0) + +#define GEN8_GTCR _MMIO(0x4274) +#define GEN8_GTCR_INVALIDATE (1<<0) + +#define GUC_ARAT_C6DIS _MMIO(0xA178) + +#define GUC_SHIM_CONTROL _MMIO(0xc064) +#define GUC_DISABLE_SRAM_INIT_TO_ZEROES (1<<0) +#define GUC_ENABLE_READ_CACHE_LOGIC (1<<1) +#define GUC_ENABLE_MIA_CACHING (1<<2) +#define GUC_GEN10_MSGCH_ENABLE (1<<4) +#define GUC_ENABLE_READ_CACHE_FOR_SRAM_DATA (1<<9) +#define GUC_ENABLE_READ_CACHE_FOR_WOPCM_DATA (1<<10) +#define GUC_ENABLE_MIA_CLOCK_GATING (1<<15) +#define GUC_GEN10_SHIM_WC_ENABLE (1<<21) + +#define GUC_SEND_INTERRUPT _MMIO(0xc4c8) +#define GUC_SEND_TRIGGER (1<<0) +#define GEN11_GUC_HOST_INTERRUPT _MMIO(0x1901f0) + +#define GUC_NUM_DOORBELLS 256 + +/* format of the HW-monitored doorbell cacheline */ +struct guc_doorbell_info { + u32 db_status; +#define GUC_DOORBELL_DISABLED 0 +#define GUC_DOORBELL_ENABLED 1 + + u32 cookie; + u32 reserved[14]; +} __packed; + +#define GEN8_DRBREGL(x) _MMIO(0x1000 + (x) * 8) +#define GEN8_DRB_VALID (1<<0) +#define GEN8_DRBREGU(x) _MMIO(0x1000 + (x) * 8 + 4) + +#define DE_GUCRMR _MMIO(0x44054) + +#define GUC_BCS_RCS_IER _MMIO(0xC550) +#define GUC_VCS2_VCS1_IER _MMIO(0xC554) +#define GUC_WD_VECS_IER _MMIO(0xC558) +#define GUC_PM_P24C_IER _MMIO(0xC55C) + +/* GuC Interrupt Vector */ +#define GUC_INTR_GUC2HOST BIT(15) +#define GUC_INTR_EXEC_ERROR BIT(14) +#define GUC_INTR_DISPLAY_EVENT BIT(13) +#define GUC_INTR_SEM_SIG BIT(12) +#define GUC_INTR_IOMMU2GUC BIT(11) +#define GUC_INTR_DOORBELL_RANG BIT(10) +#define GUC_INTR_DMA_DONE BIT(9) +#define GUC_INTR_FATAL_ERROR BIT(8) +#define GUC_INTR_NOTIF_ERROR BIT(7) +#define GUC_INTR_SW_INT_6 BIT(6) +#define GUC_INTR_SW_INT_5 BIT(5) +#define GUC_INTR_SW_INT_4 BIT(4) +#define GUC_INTR_SW_INT_3 BIT(3) +#define GUC_INTR_SW_INT_2 BIT(2) +#define GUC_INTR_SW_INT_1 BIT(1) +#define GUC_INTR_SW_INT_0 BIT(0) + +#endif diff --git a/drivers/gpu/drm/i915/gt/uc/intel_guc_submission.c b/drivers/gpu/drm/i915/gt/uc/intel_guc_submission.c new file mode 100644 index 000000000000..f325d3dd564f --- /dev/null +++ b/drivers/gpu/drm/i915/gt/uc/intel_guc_submission.c @@ -0,0 +1,1184 @@ +// SPDX-License-Identifier: MIT +/* + * Copyright © 2014 Intel Corporation + */ + +#include <linux/circ_buf.h> + +#include "gem/i915_gem_context.h" + +#include "gt/intel_context.h" +#include "gt/intel_engine_pm.h" +#include "gt/intel_gt.h" +#include "gt/intel_gt_pm.h" +#include "gt/intel_lrc_reg.h" +#include "intel_guc_submission.h" + +#include "i915_drv.h" +#include "i915_trace.h" + +enum { + GUC_PREEMPT_NONE = 0, + GUC_PREEMPT_INPROGRESS, + GUC_PREEMPT_FINISHED, +}; +#define GUC_PREEMPT_BREADCRUMB_DWORDS 0x8 +#define GUC_PREEMPT_BREADCRUMB_BYTES \ + (sizeof(u32) * GUC_PREEMPT_BREADCRUMB_DWORDS) + +/** + * DOC: GuC-based command submission + * + * GuC client: + * A intel_guc_client refers to a submission path through GuC. Currently, there + * is only one client, which is charged with all submissions to the GuC. This + * struct is the owner of a doorbell, a process descriptor and a workqueue (all + * of them inside a single gem object that contains all required pages for these + * elements). + * + * GuC stage descriptor: + * During initialization, the driver allocates a static pool of 1024 such + * descriptors, and shares them with the GuC. + * Currently, there exists a 1:1 mapping between a intel_guc_client and a + * guc_stage_desc (via the client's stage_id), so effectively only one + * gets used. This stage descriptor lets the GuC know about the doorbell, + * workqueue and process descriptor. Theoretically, it also lets the GuC + * know about our HW contexts (context ID, etc...), but we actually + * employ a kind of submission where the GuC uses the LRCA sent via the work + * item instead (the single guc_stage_desc associated to execbuf client + * contains information about the default kernel context only, but this is + * essentially unused). This is called a "proxy" submission. + * + * The Scratch registers: + * There are 16 MMIO-based registers start from 0xC180. The kernel driver writes + * a value to the action register (SOFT_SCRATCH_0) along with any data. It then + * triggers an interrupt on the GuC via another register write (0xC4C8). + * Firmware writes a success/fail code back to the action register after + * processes the request. The kernel driver polls waiting for this update and + * then proceeds. + * See intel_guc_send() + * + * Doorbells: + * Doorbells are interrupts to uKernel. A doorbell is a single cache line (QW) + * mapped into process space. + * + * Work Items: + * There are several types of work items that the host may place into a + * workqueue, each with its own requirements and limitations. Currently only + * WQ_TYPE_INORDER is needed to support legacy submission via GuC, which + * represents in-order queue. The kernel driver packs ring tail pointer and an + * ELSP context descriptor dword into Work Item. + * See guc_add_request() + * + */ + +static inline struct i915_priolist *to_priolist(struct rb_node *rb) +{ + return rb_entry(rb, struct i915_priolist, node); +} + +static inline bool is_high_priority(struct intel_guc_client *client) +{ + return (client->priority == GUC_CLIENT_PRIORITY_KMD_HIGH || + client->priority == GUC_CLIENT_PRIORITY_HIGH); +} + +static int reserve_doorbell(struct intel_guc_client *client) +{ + unsigned long offset; + unsigned long end; + u16 id; + + GEM_BUG_ON(client->doorbell_id != GUC_DOORBELL_INVALID); + + /* + * The bitmap tracks which doorbell registers are currently in use. + * It is split into two halves; the first half is used for normal + * priority contexts, the second half for high-priority ones. + */ + offset = 0; + end = GUC_NUM_DOORBELLS / 2; + if (is_high_priority(client)) { + offset = end; + end += offset; + } + + id = find_next_zero_bit(client->guc->doorbell_bitmap, end, offset); + if (id == end) + return -ENOSPC; + + __set_bit(id, client->guc->doorbell_bitmap); + client->doorbell_id = id; + DRM_DEBUG_DRIVER("client %u (high prio=%s) reserved doorbell: %d\n", + client->stage_id, yesno(is_high_priority(client)), + id); + return 0; +} + +static bool has_doorbell(struct intel_guc_client *client) +{ + if (client->doorbell_id == GUC_DOORBELL_INVALID) + return false; + + return test_bit(client->doorbell_id, client->guc->doorbell_bitmap); +} + +static void unreserve_doorbell(struct intel_guc_client *client) +{ + GEM_BUG_ON(!has_doorbell(client)); + + __clear_bit(client->doorbell_id, client->guc->doorbell_bitmap); + client->doorbell_id = GUC_DOORBELL_INVALID; +} + +/* + * Tell the GuC to allocate or deallocate a specific doorbell + */ + +static int __guc_allocate_doorbell(struct intel_guc *guc, u32 stage_id) +{ + u32 action[] = { + INTEL_GUC_ACTION_ALLOCATE_DOORBELL, + stage_id + }; + + return intel_guc_send(guc, action, ARRAY_SIZE(action)); +} + +static int __guc_deallocate_doorbell(struct intel_guc *guc, u32 stage_id) +{ + u32 action[] = { + INTEL_GUC_ACTION_DEALLOCATE_DOORBELL, + stage_id + }; + + return intel_guc_send(guc, action, ARRAY_SIZE(action)); +} + +static struct guc_stage_desc *__get_stage_desc(struct intel_guc_client *client) +{ + struct guc_stage_desc *base = client->guc->stage_desc_pool_vaddr; + + return &base[client->stage_id]; +} + +/* + * Initialise, update, or clear doorbell data shared with the GuC + * + * These functions modify shared data and so need access to the mapped + * client object which contains the page being used for the doorbell + */ + +static void __update_doorbell_desc(struct intel_guc_client *client, u16 new_id) +{ + struct guc_stage_desc *desc; + + /* Update the GuC's idea of the doorbell ID */ + desc = __get_stage_desc(client); + desc->db_id = new_id; +} + +static struct guc_doorbell_info *__get_doorbell(struct intel_guc_client *client) +{ + return client->vaddr + client->doorbell_offset; +} + +static bool __doorbell_valid(struct intel_guc *guc, u16 db_id) +{ + struct intel_uncore *uncore = guc_to_gt(guc)->uncore; + + GEM_BUG_ON(db_id >= GUC_NUM_DOORBELLS); + return intel_uncore_read(uncore, GEN8_DRBREGL(db_id)) & GEN8_DRB_VALID; +} + +static void __init_doorbell(struct intel_guc_client *client) +{ + struct guc_doorbell_info *doorbell; + + doorbell = __get_doorbell(client); + doorbell->db_status = GUC_DOORBELL_ENABLED; + doorbell->cookie = 0; +} + +static void __fini_doorbell(struct intel_guc_client *client) +{ + struct guc_doorbell_info *doorbell; + u16 db_id = client->doorbell_id; + + doorbell = __get_doorbell(client); + doorbell->db_status = GUC_DOORBELL_DISABLED; + + /* Doorbell release flow requires that we wait for GEN8_DRB_VALID bit + * to go to zero after updating db_status before we call the GuC to + * release the doorbell + */ + if (wait_for_us(!__doorbell_valid(client->guc, db_id), 10)) + WARN_ONCE(true, "Doorbell never became invalid after disable\n"); +} + +static int create_doorbell(struct intel_guc_client *client) +{ + int ret; + + if (WARN_ON(!has_doorbell(client))) + return -ENODEV; /* internal setup error, should never happen */ + + __update_doorbell_desc(client, client->doorbell_id); + __init_doorbell(client); + + ret = __guc_allocate_doorbell(client->guc, client->stage_id); + if (ret) { + __fini_doorbell(client); + __update_doorbell_desc(client, GUC_DOORBELL_INVALID); + DRM_DEBUG_DRIVER("Couldn't create client %u doorbell: %d\n", + client->stage_id, ret); + return ret; + } + + return 0; +} + +static int destroy_doorbell(struct intel_guc_client *client) +{ + int ret; + + GEM_BUG_ON(!has_doorbell(client)); + + __fini_doorbell(client); + ret = __guc_deallocate_doorbell(client->guc, client->stage_id); + if (ret) + DRM_ERROR("Couldn't destroy client %u doorbell: %d\n", + client->stage_id, ret); + + __update_doorbell_desc(client, GUC_DOORBELL_INVALID); + + return ret; +} + +static unsigned long __select_cacheline(struct intel_guc *guc) +{ + unsigned long offset; + + /* Doorbell uses a single cache line within a page */ + offset = offset_in_page(guc->db_cacheline); + + /* Moving to next cache line to reduce contention */ + guc->db_cacheline += cache_line_size(); + + DRM_DEBUG_DRIVER("reserved cacheline 0x%lx, next 0x%x, linesize %u\n", + offset, guc->db_cacheline, cache_line_size()); + return offset; +} + +static inline struct guc_process_desc * +__get_process_desc(struct intel_guc_client *client) +{ + return client->vaddr + client->proc_desc_offset; +} + +/* + * Initialise the process descriptor shared with the GuC firmware. + */ +static void guc_proc_desc_init(struct intel_guc_client *client) +{ + struct guc_process_desc *desc; + + desc = memset(__get_process_desc(client), 0, sizeof(*desc)); + + /* + * XXX: pDoorbell and WQVBaseAddress are pointers in process address + * space for ring3 clients (set them as in mmap_ioctl) or kernel + * space for kernel clients (map on demand instead? May make debug + * easier to have it mapped). + */ + desc->wq_base_addr = 0; + desc->db_base_addr = 0; + + desc->stage_id = client->stage_id; + desc->wq_size_bytes = GUC_WQ_SIZE; + desc->wq_status = WQ_STATUS_ACTIVE; + desc->priority = client->priority; +} + +static void guc_proc_desc_fini(struct intel_guc_client *client) +{ + struct guc_process_desc *desc; + + desc = __get_process_desc(client); + memset(desc, 0, sizeof(*desc)); +} + +static int guc_stage_desc_pool_create(struct intel_guc *guc) +{ + struct i915_vma *vma; + void *vaddr; + + vma = intel_guc_allocate_vma(guc, + PAGE_ALIGN(sizeof(struct guc_stage_desc) * + GUC_MAX_STAGE_DESCRIPTORS)); + if (IS_ERR(vma)) + return PTR_ERR(vma); + + vaddr = i915_gem_object_pin_map(vma->obj, I915_MAP_WB); + if (IS_ERR(vaddr)) { + i915_vma_unpin_and_release(&vma, 0); + return PTR_ERR(vaddr); + } + + guc->stage_desc_pool = vma; + guc->stage_desc_pool_vaddr = vaddr; + ida_init(&guc->stage_ids); + + return 0; +} + +static void guc_stage_desc_pool_destroy(struct intel_guc *guc) +{ + ida_destroy(&guc->stage_ids); + i915_vma_unpin_and_release(&guc->stage_desc_pool, I915_VMA_RELEASE_MAP); +} + +/* + * Initialise/clear the stage descriptor shared with the GuC firmware. + * + * This descriptor tells the GuC where (in GGTT space) to find the important + * data structures relating to this client (doorbell, process descriptor, + * write queue, etc). + */ +static void guc_stage_desc_init(struct intel_guc_client *client) +{ + struct intel_guc *guc = client->guc; + struct guc_stage_desc *desc; + u32 gfx_addr; + + desc = __get_stage_desc(client); + memset(desc, 0, sizeof(*desc)); + + desc->attribute = GUC_STAGE_DESC_ATTR_ACTIVE | + GUC_STAGE_DESC_ATTR_KERNEL; + if (is_high_priority(client)) + desc->attribute |= GUC_STAGE_DESC_ATTR_PREEMPT; + desc->stage_id = client->stage_id; + desc->priority = client->priority; + desc->db_id = client->doorbell_id; + + /* + * The doorbell, process descriptor, and workqueue are all parts + * of the client object, which the GuC will reference via the GGTT + */ + gfx_addr = intel_guc_ggtt_offset(guc, client->vma); + desc->db_trigger_phy = sg_dma_address(client->vma->pages->sgl) + + client->doorbell_offset; + desc->db_trigger_cpu = ptr_to_u64(__get_doorbell(client)); + desc->db_trigger_uk = gfx_addr + client->doorbell_offset; + desc->process_desc = gfx_addr + client->proc_desc_offset; + desc->wq_addr = gfx_addr + GUC_DB_SIZE; + desc->wq_size = GUC_WQ_SIZE; + + desc->desc_private = ptr_to_u64(client); +} + +static void guc_stage_desc_fini(struct intel_guc_client *client) +{ + struct guc_stage_desc *desc; + + desc = __get_stage_desc(client); + memset(desc, 0, sizeof(*desc)); +} + +/* Construct a Work Item and append it to the GuC's Work Queue */ +static void guc_wq_item_append(struct intel_guc_client *client, + u32 target_engine, u32 context_desc, + u32 ring_tail, u32 fence_id) +{ + /* wqi_len is in DWords, and does not include the one-word header */ + const size_t wqi_size = sizeof(struct guc_wq_item); + const u32 wqi_len = wqi_size / sizeof(u32) - 1; + struct guc_process_desc *desc = __get_process_desc(client); + struct guc_wq_item *wqi; + u32 wq_off; + + lockdep_assert_held(&client->wq_lock); + + /* For now workqueue item is 4 DWs; workqueue buffer is 2 pages. So we + * should not have the case where structure wqi is across page, neither + * wrapped to the beginning. This simplifies the implementation below. + * + * XXX: if not the case, we need save data to a temp wqi and copy it to + * workqueue buffer dw by dw. + */ + BUILD_BUG_ON(wqi_size != 16); + + /* We expect the WQ to be active if we're appending items to it */ + GEM_BUG_ON(desc->wq_status != WQ_STATUS_ACTIVE); + + /* Free space is guaranteed. */ + wq_off = READ_ONCE(desc->tail); + GEM_BUG_ON(CIRC_SPACE(wq_off, READ_ONCE(desc->head), + GUC_WQ_SIZE) < wqi_size); + GEM_BUG_ON(wq_off & (wqi_size - 1)); + + /* WQ starts from the page after doorbell / process_desc */ + wqi = client->vaddr + wq_off + GUC_DB_SIZE; + + if (I915_SELFTEST_ONLY(client->use_nop_wqi)) { + wqi->header = WQ_TYPE_NOOP | (wqi_len << WQ_LEN_SHIFT); + } else { + /* Now fill in the 4-word work queue item */ + wqi->header = WQ_TYPE_INORDER | + (wqi_len << WQ_LEN_SHIFT) | + (target_engine << WQ_TARGET_SHIFT) | + WQ_NO_WCFLUSH_WAIT; + wqi->context_desc = context_desc; + wqi->submit_element_info = ring_tail << WQ_RING_TAIL_SHIFT; + GEM_BUG_ON(ring_tail > WQ_RING_TAIL_MAX); + wqi->fence_id = fence_id; + } + + /* Make the update visible to GuC */ + WRITE_ONCE(desc->tail, (wq_off + wqi_size) & (GUC_WQ_SIZE - 1)); +} + +static void guc_ring_doorbell(struct intel_guc_client *client) +{ + struct guc_doorbell_info *db; + u32 cookie; + + lockdep_assert_held(&client->wq_lock); + + /* pointer of current doorbell cacheline */ + db = __get_doorbell(client); + + /* + * We're not expecting the doorbell cookie to change behind our back, + * we also need to treat 0 as a reserved value. + */ + cookie = READ_ONCE(db->cookie); + WARN_ON_ONCE(xchg(&db->cookie, cookie + 1 ?: cookie + 2) != cookie); + + /* XXX: doorbell was lost and need to acquire it again */ + GEM_BUG_ON(db->db_status != GUC_DOORBELL_ENABLED); +} + +static void guc_add_request(struct intel_guc *guc, struct i915_request *rq) +{ + struct intel_guc_client *client = guc->execbuf_client; + struct intel_engine_cs *engine = rq->engine; + u32 ctx_desc = lower_32_bits(rq->hw_context->lrc_desc); + u32 ring_tail = intel_ring_set_tail(rq->ring, rq->tail) / sizeof(u64); + + guc_wq_item_append(client, engine->guc_id, ctx_desc, + ring_tail, rq->fence.seqno); + guc_ring_doorbell(client); +} + +/* + * When we're doing submissions using regular execlists backend, writing to + * ELSP from CPU side is enough to make sure that writes to ringbuffer pages + * pinned in mappable aperture portion of GGTT are visible to command streamer. + * Writes done by GuC on our behalf are not guaranteeing such ordering, + * therefore, to ensure the flush, we're issuing a POSTING READ. + */ +static void flush_ggtt_writes(struct i915_vma *vma) +{ + struct drm_i915_private *i915 = vma->vm->i915; + + if (i915_vma_is_map_and_fenceable(vma)) + intel_uncore_posting_read_fw(&i915->uncore, GUC_STATUS); +} + +static void guc_submit(struct intel_engine_cs *engine, + struct i915_request **out, + struct i915_request **end) +{ + struct intel_guc *guc = &engine->gt->uc.guc; + struct intel_guc_client *client = guc->execbuf_client; + + spin_lock(&client->wq_lock); + + do { + struct i915_request *rq = *out++; + + flush_ggtt_writes(rq->ring->vma); + guc_add_request(guc, rq); + } while (out != end); + + spin_unlock(&client->wq_lock); +} + +static inline int rq_prio(const struct i915_request *rq) +{ + return rq->sched.attr.priority | __NO_PREEMPTION; +} + +static struct i915_request *schedule_in(struct i915_request *rq, int idx) +{ + trace_i915_request_in(rq, idx); + + /* + * Currently we are not tracking the rq->context being inflight + * (ce->inflight = rq->engine). It is only used by the execlists + * backend at the moment, a similar counting strategy would be + * required if we generalise the inflight tracking. + */ + + intel_gt_pm_get(rq->engine->gt); + return i915_request_get(rq); +} + +static void schedule_out(struct i915_request *rq) +{ + trace_i915_request_out(rq); + + intel_gt_pm_put(rq->engine->gt); + i915_request_put(rq); +} + +static void __guc_dequeue(struct intel_engine_cs *engine) +{ + struct intel_engine_execlists * const execlists = &engine->execlists; + struct i915_request **first = execlists->inflight; + struct i915_request ** const last_port = first + execlists->port_mask; + struct i915_request *last = first[0]; + struct i915_request **port; + bool submit = false; + struct rb_node *rb; + + lockdep_assert_held(&engine->active.lock); + + if (last) { + if (*++first) + return; + + last = NULL; + } + + /* + * We write directly into the execlists->inflight queue and don't use + * the execlists->pending queue, as we don't have a distinct switch + * event. + */ + port = first; + while ((rb = rb_first_cached(&execlists->queue))) { + struct i915_priolist *p = to_priolist(rb); + struct i915_request *rq, *rn; + int i; + + priolist_for_each_request_consume(rq, rn, p, i) { + if (last && rq->hw_context != last->hw_context) { + if (port == last_port) + goto done; + + *port = schedule_in(last, + port - execlists->inflight); + port++; + } + + list_del_init(&rq->sched.link); + __i915_request_submit(rq); + submit = true; + last = rq; + } + + rb_erase_cached(&p->node, &execlists->queue); + i915_priolist_free(p); + } +done: + execlists->queue_priority_hint = + rb ? to_priolist(rb)->priority : INT_MIN; + if (submit) { + *port = schedule_in(last, port - execlists->inflight); + *++port = NULL; + guc_submit(engine, first, port); + } + execlists->active = execlists->inflight; +} + +static void guc_submission_tasklet(unsigned long data) +{ + struct intel_engine_cs * const engine = (struct intel_engine_cs *)data; + struct intel_engine_execlists * const execlists = &engine->execlists; + struct i915_request **port, *rq; + unsigned long flags; + + spin_lock_irqsave(&engine->active.lock, flags); + + for (port = execlists->inflight; (rq = *port); port++) { + if (!i915_request_completed(rq)) + break; + + schedule_out(rq); + } + if (port != execlists->inflight) { + int idx = port - execlists->inflight; + int rem = ARRAY_SIZE(execlists->inflight) - idx; + memmove(execlists->inflight, port, rem * sizeof(*port)); + } + + __guc_dequeue(engine); + + spin_unlock_irqrestore(&engine->active.lock, flags); +} + +static void guc_reset_prepare(struct intel_engine_cs *engine) +{ + struct intel_engine_execlists * const execlists = &engine->execlists; + + GEM_TRACE("%s\n", engine->name); + + /* + * Prevent request submission to the hardware until we have + * completed the reset in i915_gem_reset_finish(). If a request + * is completed by one engine, it may then queue a request + * to a second via its execlists->tasklet *just* as we are + * calling engine->init_hw() and also writing the ELSP. + * Turning off the execlists->tasklet until the reset is over + * prevents the race. + */ + __tasklet_disable_sync_once(&execlists->tasklet); +} + +static void +cancel_port_requests(struct intel_engine_execlists * const execlists) +{ + struct i915_request * const *port, *rq; + + /* Note we are only using the inflight and not the pending queue */ + + for (port = execlists->active; (rq = *port); port++) + schedule_out(rq); + execlists->active = + memset(execlists->inflight, 0, sizeof(execlists->inflight)); +} + +static void guc_reset(struct intel_engine_cs *engine, bool stalled) +{ + struct intel_engine_execlists * const execlists = &engine->execlists; + struct i915_request *rq; + unsigned long flags; + + spin_lock_irqsave(&engine->active.lock, flags); + + cancel_port_requests(execlists); + + /* Push back any incomplete requests for replay after the reset. */ + rq = execlists_unwind_incomplete_requests(execlists); + if (!rq) + goto out_unlock; + + if (!i915_request_started(rq)) + stalled = false; + + __i915_request_reset(rq, stalled); + intel_lr_context_reset(engine, rq->hw_context, rq->head, stalled); + +out_unlock: + spin_unlock_irqrestore(&engine->active.lock, flags); +} + +static void guc_cancel_requests(struct intel_engine_cs *engine) +{ + struct intel_engine_execlists * const execlists = &engine->execlists; + struct i915_request *rq, *rn; + struct rb_node *rb; + unsigned long flags; + + GEM_TRACE("%s\n", engine->name); + + /* + * Before we call engine->cancel_requests(), we should have exclusive + * access to the submission state. This is arranged for us by the + * caller disabling the interrupt generation, the tasklet and other + * threads that may then access the same state, giving us a free hand + * to reset state. However, we still need to let lockdep be aware that + * we know this state may be accessed in hardirq context, so we + * disable the irq around this manipulation and we want to keep + * the spinlock focused on its duties and not accidentally conflate + * coverage to the submission's irq state. (Similarly, although we + * shouldn't need to disable irq around the manipulation of the + * submission's irq state, we also wish to remind ourselves that + * it is irq state.) + */ + spin_lock_irqsave(&engine->active.lock, flags); + + /* Cancel the requests on the HW and clear the ELSP tracker. */ + cancel_port_requests(execlists); + + /* Mark all executing requests as skipped. */ + list_for_each_entry(rq, &engine->active.requests, sched.link) { + if (!i915_request_signaled(rq)) + dma_fence_set_error(&rq->fence, -EIO); + + i915_request_mark_complete(rq); + } + + /* Flush the queued requests to the timeline list (for retiring). */ + while ((rb = rb_first_cached(&execlists->queue))) { + struct i915_priolist *p = to_priolist(rb); + int i; + + priolist_for_each_request_consume(rq, rn, p, i) { + list_del_init(&rq->sched.link); + __i915_request_submit(rq); + dma_fence_set_error(&rq->fence, -EIO); + i915_request_mark_complete(rq); + } + + rb_erase_cached(&p->node, &execlists->queue); + i915_priolist_free(p); + } + + /* Remaining _unready_ requests will be nop'ed when submitted */ + + execlists->queue_priority_hint = INT_MIN; + execlists->queue = RB_ROOT_CACHED; + + spin_unlock_irqrestore(&engine->active.lock, flags); +} + +static void guc_reset_finish(struct intel_engine_cs *engine) +{ + struct intel_engine_execlists * const execlists = &engine->execlists; + + if (__tasklet_enable(&execlists->tasklet)) + /* And kick in case we missed a new request submission. */ + tasklet_hi_schedule(&execlists->tasklet); + + GEM_TRACE("%s: depth->%d\n", engine->name, + atomic_read(&execlists->tasklet.count)); +} + +/* + * Everything below here is concerned with setup & teardown, and is + * therefore not part of the somewhat time-critical batch-submission + * path of guc_submit() above. + */ + +/* Check that a doorbell register is in the expected state */ +static bool doorbell_ok(struct intel_guc *guc, u16 db_id) +{ + bool valid; + + GEM_BUG_ON(db_id >= GUC_NUM_DOORBELLS); + + valid = __doorbell_valid(guc, db_id); + + if (test_bit(db_id, guc->doorbell_bitmap) == valid) + return true; + + DRM_DEBUG_DRIVER("Doorbell %u has unexpected state: valid=%s\n", + db_id, yesno(valid)); + + return false; +} + +static bool guc_verify_doorbells(struct intel_guc *guc) +{ + bool doorbells_ok = true; + u16 db_id; + + for (db_id = 0; db_id < GUC_NUM_DOORBELLS; ++db_id) + if (!doorbell_ok(guc, db_id)) + doorbells_ok = false; + + return doorbells_ok; +} + +/** + * guc_client_alloc() - Allocate an intel_guc_client + * @guc: the intel_guc structure + * @priority: four levels priority _CRITICAL, _HIGH, _NORMAL and _LOW + * The kernel client to replace ExecList submission is created with + * NORMAL priority. Priority of a client for scheduler can be HIGH, + * while a preemption context can use CRITICAL. + * + * Return: An intel_guc_client object if success, else NULL. + */ +static struct intel_guc_client * +guc_client_alloc(struct intel_guc *guc, u32 priority) +{ + struct intel_guc_client *client; + struct i915_vma *vma; + void *vaddr; + int ret; + + client = kzalloc(sizeof(*client), GFP_KERNEL); + if (!client) + return ERR_PTR(-ENOMEM); + + client->guc = guc; + client->priority = priority; + client->doorbell_id = GUC_DOORBELL_INVALID; + spin_lock_init(&client->wq_lock); + + ret = ida_simple_get(&guc->stage_ids, 0, GUC_MAX_STAGE_DESCRIPTORS, + GFP_KERNEL); + if (ret < 0) + goto err_client; + + client->stage_id = ret; + + /* The first page is doorbell/proc_desc. Two followed pages are wq. */ + vma = intel_guc_allocate_vma(guc, GUC_DB_SIZE + GUC_WQ_SIZE); + if (IS_ERR(vma)) { + ret = PTR_ERR(vma); + goto err_id; + } + + /* We'll keep just the first (doorbell/proc) page permanently kmap'd. */ + client->vma = vma; + + vaddr = i915_gem_object_pin_map(vma->obj, I915_MAP_WB); + if (IS_ERR(vaddr)) { + ret = PTR_ERR(vaddr); + goto err_vma; + } + client->vaddr = vaddr; + + ret = reserve_doorbell(client); + if (ret) + goto err_vaddr; + + client->doorbell_offset = __select_cacheline(guc); + + /* + * Since the doorbell only requires a single cacheline, we can save + * space by putting the application process descriptor in the same + * page. Use the half of the page that doesn't include the doorbell. + */ + if (client->doorbell_offset >= (GUC_DB_SIZE / 2)) + client->proc_desc_offset = 0; + else + client->proc_desc_offset = (GUC_DB_SIZE / 2); + + DRM_DEBUG_DRIVER("new priority %u client %p: stage_id %u\n", + priority, client, client->stage_id); + DRM_DEBUG_DRIVER("doorbell id %u, cacheline offset 0x%lx\n", + client->doorbell_id, client->doorbell_offset); + + return client; + +err_vaddr: + i915_gem_object_unpin_map(client->vma->obj); +err_vma: + i915_vma_unpin_and_release(&client->vma, 0); +err_id: + ida_simple_remove(&guc->stage_ids, client->stage_id); +err_client: + kfree(client); + return ERR_PTR(ret); +} + +static void guc_client_free(struct intel_guc_client *client) +{ + unreserve_doorbell(client); + i915_vma_unpin_and_release(&client->vma, I915_VMA_RELEASE_MAP); + ida_simple_remove(&client->guc->stage_ids, client->stage_id); + kfree(client); +} + +static inline bool ctx_save_restore_disabled(struct intel_context *ce) +{ + u32 sr = ce->lrc_reg_state[CTX_CONTEXT_CONTROL + 1]; + +#define SR_DISABLED \ + _MASKED_BIT_ENABLE(CTX_CTRL_ENGINE_CTX_RESTORE_INHIBIT | \ + CTX_CTRL_ENGINE_CTX_SAVE_INHIBIT) + + return (sr & SR_DISABLED) == SR_DISABLED; + +#undef SR_DISABLED +} + +static int guc_clients_create(struct intel_guc *guc) +{ + struct intel_guc_client *client; + + GEM_BUG_ON(guc->execbuf_client); + + client = guc_client_alloc(guc, GUC_CLIENT_PRIORITY_KMD_NORMAL); + if (IS_ERR(client)) { + DRM_ERROR("Failed to create GuC client for submission!\n"); + return PTR_ERR(client); + } + guc->execbuf_client = client; + + return 0; +} + +static void guc_clients_destroy(struct intel_guc *guc) +{ + struct intel_guc_client *client; + + client = fetch_and_zero(&guc->execbuf_client); + if (client) + guc_client_free(client); +} + +static int __guc_client_enable(struct intel_guc_client *client) +{ + int ret; + + guc_proc_desc_init(client); + guc_stage_desc_init(client); + + ret = create_doorbell(client); + if (ret) + goto fail; + + return 0; + +fail: + guc_stage_desc_fini(client); + guc_proc_desc_fini(client); + return ret; +} + +static void __guc_client_disable(struct intel_guc_client *client) +{ + /* + * By the time we're here, GuC may have already been reset. if that is + * the case, instead of trying (in vain) to communicate with it, let's + * just cleanup the doorbell HW and our internal state. + */ + if (intel_guc_is_running(client->guc)) + destroy_doorbell(client); + else + __fini_doorbell(client); + + guc_stage_desc_fini(client); + guc_proc_desc_fini(client); +} + +static int guc_clients_enable(struct intel_guc *guc) +{ + return __guc_client_enable(guc->execbuf_client); +} + +static void guc_clients_disable(struct intel_guc *guc) +{ + if (guc->execbuf_client) + __guc_client_disable(guc->execbuf_client); +} + +/* + * Set up the memory resources to be shared with the GuC (via the GGTT) + * at firmware loading time. + */ +int intel_guc_submission_init(struct intel_guc *guc) +{ + int ret; + + if (guc->stage_desc_pool) + return 0; + + ret = guc_stage_desc_pool_create(guc); + if (ret) + return ret; + /* + * Keep static analysers happy, let them know that we allocated the + * vma after testing that it didn't exist earlier. + */ + GEM_BUG_ON(!guc->stage_desc_pool); + + WARN_ON(!guc_verify_doorbells(guc)); + ret = guc_clients_create(guc); + if (ret) + goto err_pool; + + return 0; + +err_pool: + guc_stage_desc_pool_destroy(guc); + return ret; +} + +void intel_guc_submission_fini(struct intel_guc *guc) +{ + guc_clients_destroy(guc); + WARN_ON(!guc_verify_doorbells(guc)); + + if (guc->stage_desc_pool) + guc_stage_desc_pool_destroy(guc); +} + +static void guc_interrupts_capture(struct intel_gt *gt) +{ + struct intel_rps *rps = >->i915->gt_pm.rps; + struct intel_uncore *uncore = gt->uncore; + struct intel_engine_cs *engine; + enum intel_engine_id id; + int irqs; + + /* tell all command streamers to forward interrupts (but not vblank) + * to GuC + */ + irqs = _MASKED_BIT_ENABLE(GFX_INTERRUPT_STEERING); + for_each_engine(engine, gt->i915, id) + ENGINE_WRITE(engine, RING_MODE_GEN7, irqs); + + /* route USER_INTERRUPT to Host, all others are sent to GuC. */ + irqs = GT_RENDER_USER_INTERRUPT << GEN8_RCS_IRQ_SHIFT | + GT_RENDER_USER_INTERRUPT << GEN8_BCS_IRQ_SHIFT; + /* These three registers have the same bit definitions */ + intel_uncore_write(uncore, GUC_BCS_RCS_IER, ~irqs); + intel_uncore_write(uncore, GUC_VCS2_VCS1_IER, ~irqs); + intel_uncore_write(uncore, GUC_WD_VECS_IER, ~irqs); + + /* + * The REDIRECT_TO_GUC bit of the PMINTRMSK register directs all + * (unmasked) PM interrupts to the GuC. All other bits of this + * register *disable* generation of a specific interrupt. + * + * 'pm_intrmsk_mbz' indicates bits that are NOT to be set when + * writing to the PM interrupt mask register, i.e. interrupts + * that must not be disabled. + * + * If the GuC is handling these interrupts, then we must not let + * the PM code disable ANY interrupt that the GuC is expecting. + * So for each ENABLED (0) bit in this register, we must SET the + * bit in pm_intrmsk_mbz so that it's left enabled for the GuC. + * GuC needs ARAT expired interrupt unmasked hence it is set in + * pm_intrmsk_mbz. + * + * Here we CLEAR REDIRECT_TO_GUC bit in pm_intrmsk_mbz, which will + * result in the register bit being left SET! + */ + rps->pm_intrmsk_mbz |= ARAT_EXPIRED_INTRMSK; + rps->pm_intrmsk_mbz &= ~GEN8_PMINTR_DISABLE_REDIRECT_TO_GUC; +} + +static void guc_interrupts_release(struct intel_gt *gt) +{ + struct intel_rps *rps = >->i915->gt_pm.rps; + struct intel_uncore *uncore = gt->uncore; + struct intel_engine_cs *engine; + enum intel_engine_id id; + int irqs; + + /* + * tell all command streamers NOT to forward interrupts or vblank + * to GuC. + */ + irqs = _MASKED_FIELD(GFX_FORWARD_VBLANK_MASK, GFX_FORWARD_VBLANK_NEVER); + irqs |= _MASKED_BIT_DISABLE(GFX_INTERRUPT_STEERING); + for_each_engine(engine, gt->i915, id) + ENGINE_WRITE(engine, RING_MODE_GEN7, irqs); + + /* route all GT interrupts to the host */ + intel_uncore_write(uncore, GUC_BCS_RCS_IER, 0); + intel_uncore_write(uncore, GUC_VCS2_VCS1_IER, 0); + intel_uncore_write(uncore, GUC_WD_VECS_IER, 0); + + rps->pm_intrmsk_mbz |= GEN8_PMINTR_DISABLE_REDIRECT_TO_GUC; + rps->pm_intrmsk_mbz &= ~ARAT_EXPIRED_INTRMSK; +} + +static void guc_set_default_submission(struct intel_engine_cs *engine) +{ + /* + * We inherit a bunch of functions from execlists that we'd like + * to keep using: + * + * engine->submit_request = execlists_submit_request; + * engine->cancel_requests = execlists_cancel_requests; + * engine->schedule = execlists_schedule; + * + * But we need to override the actual submission backend in order + * to talk to the GuC. + */ + intel_execlists_set_default_submission(engine); + + engine->execlists.tasklet.func = guc_submission_tasklet; + + /* do not use execlists park/unpark */ + engine->park = engine->unpark = NULL; + + engine->reset.prepare = guc_reset_prepare; + engine->reset.reset = guc_reset; + engine->reset.finish = guc_reset_finish; + + engine->cancel_requests = guc_cancel_requests; + + engine->flags &= ~I915_ENGINE_SUPPORTS_STATS; + engine->flags |= I915_ENGINE_NEEDS_BREADCRUMB_TASKLET; + + /* + * For the breadcrumb irq to work we need the interrupts to stay + * enabled. However, on all platforms on which we'll have support for + * GuC submission we don't allow disabling the interrupts at runtime, so + * we're always safe with the current flow. + */ + GEM_BUG_ON(engine->irq_enable || engine->irq_disable); +} + +int intel_guc_submission_enable(struct intel_guc *guc) +{ + struct intel_gt *gt = guc_to_gt(guc); + struct intel_engine_cs *engine; + enum intel_engine_id id; + int err; + + err = i915_inject_load_error(gt->i915, -ENXIO); + if (err) + return err; + + /* + * We're using GuC work items for submitting work through GuC. Since + * we're coalescing multiple requests from a single context into a + * single work item prior to assigning it to execlist_port, we can + * never have more work items than the total number of ports (for all + * engines). The GuC firmware is controlling the HEAD of work queue, + * and it is guaranteed that it will remove the work item from the + * queue before our request is completed. + */ + BUILD_BUG_ON(ARRAY_SIZE(engine->execlists.inflight) * + sizeof(struct guc_wq_item) * + I915_NUM_ENGINES > GUC_WQ_SIZE); + + GEM_BUG_ON(!guc->execbuf_client); + + err = guc_clients_enable(guc); + if (err) + return err; + + /* Take over from manual control of ELSP (execlists) */ + guc_interrupts_capture(gt); + + for_each_engine(engine, gt->i915, id) { + engine->set_default_submission = guc_set_default_submission; + engine->set_default_submission(engine); + } + + return 0; +} + +void intel_guc_submission_disable(struct intel_guc *guc) +{ + struct intel_gt *gt = guc_to_gt(guc); + + GEM_BUG_ON(gt->awake); /* GT should be parked first */ + + guc_interrupts_release(gt); + guc_clients_disable(guc); +} + +static bool __guc_submission_support(struct intel_guc *guc) +{ + /* XXX: GuC submission is unavailable for now */ + return false; + + if (!intel_guc_is_supported(guc)) + return false; + + return i915_modparams.enable_guc & ENABLE_GUC_SUBMISSION; +} + +void intel_guc_submission_init_early(struct intel_guc *guc) +{ + guc->submission_supported = __guc_submission_support(guc); +} + +#if IS_ENABLED(CONFIG_DRM_I915_SELFTEST) +#include "selftest_guc.c" +#endif diff --git a/drivers/gpu/drm/i915/gt/uc/intel_guc_submission.h b/drivers/gpu/drm/i915/gt/uc/intel_guc_submission.h new file mode 100644 index 000000000000..54d716828352 --- /dev/null +++ b/drivers/gpu/drm/i915/gt/uc/intel_guc_submission.h @@ -0,0 +1,67 @@ +/* SPDX-License-Identifier: MIT */ +/* + * Copyright © 2014-2019 Intel Corporation + */ + +#ifndef _INTEL_GUC_SUBMISSION_H_ +#define _INTEL_GUC_SUBMISSION_H_ + +#include <linux/spinlock.h> + +#include "gt/intel_engine_types.h" + +#include "i915_gem.h" +#include "i915_selftest.h" + +struct drm_i915_private; + +/* + * This structure primarily describes the GEM object shared with the GuC. + * The specs sometimes refer to this object as a "GuC context", but we use + * the term "client" to avoid confusion with hardware contexts. This + * GEM object is held for the entire lifetime of our interaction with + * the GuC, being allocated before the GuC is loaded with its firmware. + * Because there's no way to update the address used by the GuC after + * initialisation, the shared object must stay pinned into the GGTT as + * long as the GuC is in use. We also keep the first page (only) mapped + * into kernel address space, as it includes shared data that must be + * updated on every request submission. + * + * The single GEM object described here is actually made up of several + * separate areas, as far as the GuC is concerned. The first page (kept + * kmap'd) includes the "process descriptor" which holds sequence data for + * the doorbell, and one cacheline which actually *is* the doorbell; a + * write to this will "ring the doorbell" (i.e. send an interrupt to the + * GuC). The subsequent pages of the client object constitute the work + * queue (a circular array of work items), again described in the process + * descriptor. Work queue pages are mapped momentarily as required. + */ +struct intel_guc_client { + struct i915_vma *vma; + void *vaddr; + struct intel_guc *guc; + + /* bitmap of (host) engine ids */ + u32 priority; + u32 stage_id; + u32 proc_desc_offset; + + u16 doorbell_id; + unsigned long doorbell_offset; + + /* Protects GuC client's WQ access */ + spinlock_t wq_lock; + + /* For testing purposes, use nop WQ items instead of real ones */ + I915_SELFTEST_DECLARE(bool use_nop_wqi); +}; + +void intel_guc_submission_init_early(struct intel_guc *guc); +int intel_guc_submission_init(struct intel_guc *guc); +int intel_guc_submission_enable(struct intel_guc *guc); +void intel_guc_submission_disable(struct intel_guc *guc); +void intel_guc_submission_fini(struct intel_guc *guc); +int intel_guc_preempt_work_create(struct intel_guc *guc); +void intel_guc_preempt_work_destroy(struct intel_guc *guc); + +#endif diff --git a/drivers/gpu/drm/i915/gt/uc/intel_huc.c b/drivers/gpu/drm/i915/gt/uc/intel_huc.c new file mode 100644 index 000000000000..d4625c97b4f9 --- /dev/null +++ b/drivers/gpu/drm/i915/gt/uc/intel_huc.c @@ -0,0 +1,192 @@ +// SPDX-License-Identifier: MIT +/* + * Copyright © 2016-2019 Intel Corporation + */ + +#include <linux/types.h> + +#include "gt/intel_gt.h" +#include "intel_huc.h" +#include "i915_drv.h" + +void intel_huc_init_early(struct intel_huc *huc) +{ + struct drm_i915_private *i915 = huc_to_gt(huc)->i915; + + intel_huc_fw_init_early(huc); + + if (INTEL_GEN(i915) >= 11) { + huc->status.reg = GEN11_HUC_KERNEL_LOAD_INFO; + huc->status.mask = HUC_LOAD_SUCCESSFUL; + huc->status.value = HUC_LOAD_SUCCESSFUL; + } else { + huc->status.reg = HUC_STATUS2; + huc->status.mask = HUC_FW_VERIFIED; + huc->status.value = HUC_FW_VERIFIED; + } +} + +static int intel_huc_rsa_data_create(struct intel_huc *huc) +{ + struct intel_gt *gt = huc_to_gt(huc); + struct intel_guc *guc = >->uc.guc; + struct i915_vma *vma; + size_t copied; + void *vaddr; + int err; + + err = i915_inject_load_error(gt->i915, -ENXIO); + if (err) + return err; + + /* + * HuC firmware will sit above GUC_GGTT_TOP and will not map + * through GTT. Unfortunately, this means GuC cannot perform + * the HuC auth. as the rsa offset now falls within the GuC + * inaccessible range. We resort to perma-pinning an additional + * vma within the accessible range that only contains the rsa + * signature. The GuC can use this extra pinning to perform + * the authentication since its GGTT offset will be GuC + * accessible. + */ + GEM_BUG_ON(huc->fw.rsa_size > PAGE_SIZE); + vma = intel_guc_allocate_vma(guc, PAGE_SIZE); + if (IS_ERR(vma)) + return PTR_ERR(vma); + + vaddr = i915_gem_object_pin_map(vma->obj, I915_MAP_WB); + if (IS_ERR(vaddr)) { + i915_vma_unpin_and_release(&vma, 0); + return PTR_ERR(vaddr); + } + + copied = intel_uc_fw_copy_rsa(&huc->fw, vaddr, vma->size); + GEM_BUG_ON(copied < huc->fw.rsa_size); + + i915_gem_object_unpin_map(vma->obj); + + huc->rsa_data = vma; + + return 0; +} + +static void intel_huc_rsa_data_destroy(struct intel_huc *huc) +{ + i915_vma_unpin_and_release(&huc->rsa_data, 0); +} + +int intel_huc_init(struct intel_huc *huc) +{ + struct drm_i915_private *i915 = huc_to_gt(huc)->i915; + int err; + + err = intel_uc_fw_init(&huc->fw); + if (err) + goto out; + + /* + * HuC firmware image is outside GuC accessible range. + * Copy the RSA signature out of the image into + * a perma-pinned region set aside for it + */ + err = intel_huc_rsa_data_create(huc); + if (err) + goto out_fini; + + return 0; + +out_fini: + intel_uc_fw_fini(&huc->fw); +out: + intel_uc_fw_cleanup_fetch(&huc->fw); + DRM_DEV_DEBUG_DRIVER(i915->drm.dev, "failed with %d\n", err); + return err; +} + +void intel_huc_fini(struct intel_huc *huc) +{ + if (!intel_uc_fw_is_available(&huc->fw)) + return; + + intel_huc_rsa_data_destroy(huc); + intel_uc_fw_fini(&huc->fw); +} + +/** + * intel_huc_auth() - Authenticate HuC uCode + * @huc: intel_huc structure + * + * Called after HuC and GuC firmware loading during intel_uc_init_hw(). + * + * This function pins HuC firmware image object into GGTT. + * Then it invokes GuC action to authenticate passing the offset to RSA + * signature through intel_guc_auth_huc(). It then waits for 50ms for + * firmware verification ACK and unpins the object. + */ +int intel_huc_auth(struct intel_huc *huc) +{ + struct intel_gt *gt = huc_to_gt(huc); + struct intel_guc *guc = >->uc.guc; + int ret; + + GEM_BUG_ON(intel_huc_is_authenticated(huc)); + + if (!intel_uc_fw_is_loaded(&huc->fw)) + return -ENOEXEC; + + ret = i915_inject_load_error(gt->i915, -ENXIO); + if (ret) + goto fail; + + ret = intel_guc_auth_huc(guc, + intel_guc_ggtt_offset(guc, huc->rsa_data)); + if (ret) { + DRM_ERROR("HuC: GuC did not ack Auth request %d\n", ret); + goto fail; + } + + /* Check authentication status, it should be done by now */ + ret = __intel_wait_for_register(gt->uncore, + huc->status.reg, + huc->status.mask, + huc->status.value, + 2, 50, NULL); + if (ret) { + DRM_ERROR("HuC: Firmware not verified %d\n", ret); + goto fail; + } + + intel_uc_fw_change_status(&huc->fw, INTEL_UC_FIRMWARE_RUNNING); + return 0; + +fail: + i915_probe_error(gt->i915, "HuC: Authentication failed %d\n", ret); + intel_uc_fw_change_status(&huc->fw, INTEL_UC_FIRMWARE_FAIL); + return ret; +} + +/** + * intel_huc_check_status() - check HuC status + * @huc: intel_huc structure + * + * This function reads status register to verify if HuC + * firmware was successfully loaded. + * + * Returns: 1 if HuC firmware is loaded and verified, + * 0 if HuC firmware is not loaded and -ENODEV if HuC + * is not present on this platform. + */ +int intel_huc_check_status(struct intel_huc *huc) +{ + struct intel_gt *gt = huc_to_gt(huc); + intel_wakeref_t wakeref; + u32 status = 0; + + if (!intel_huc_is_supported(huc)) + return -ENODEV; + + with_intel_runtime_pm(>->i915->runtime_pm, wakeref) + status = intel_uncore_read(gt->uncore, huc->status.reg); + + return (status & huc->status.mask) == huc->status.value; +} diff --git a/drivers/gpu/drm/i915/gt/uc/intel_huc.h b/drivers/gpu/drm/i915/gt/uc/intel_huc.h new file mode 100644 index 000000000000..644c059fe01d --- /dev/null +++ b/drivers/gpu/drm/i915/gt/uc/intel_huc.h @@ -0,0 +1,54 @@ +/* SPDX-License-Identifier: MIT */ +/* + * Copyright © 2014-2019 Intel Corporation + */ + +#ifndef _INTEL_HUC_H_ +#define _INTEL_HUC_H_ + +#include "i915_reg.h" +#include "intel_uc_fw.h" +#include "intel_huc_fw.h" + +struct intel_huc { + /* Generic uC firmware management */ + struct intel_uc_fw fw; + + /* HuC-specific additions */ + struct i915_vma *rsa_data; + + struct { + i915_reg_t reg; + u32 mask; + u32 value; + } status; +}; + +void intel_huc_init_early(struct intel_huc *huc); +int intel_huc_init(struct intel_huc *huc); +void intel_huc_fini(struct intel_huc *huc); +int intel_huc_auth(struct intel_huc *huc); +int intel_huc_check_status(struct intel_huc *huc); + +static inline int intel_huc_sanitize(struct intel_huc *huc) +{ + intel_uc_fw_sanitize(&huc->fw); + return 0; +} + +static inline bool intel_huc_is_supported(struct intel_huc *huc) +{ + return intel_uc_fw_is_supported(&huc->fw); +} + +static inline bool intel_huc_is_enabled(struct intel_huc *huc) +{ + return intel_uc_fw_is_enabled(&huc->fw); +} + +static inline bool intel_huc_is_authenticated(struct intel_huc *huc) +{ + return intel_uc_fw_is_running(&huc->fw); +} + +#endif diff --git a/drivers/gpu/drm/i915/gt/uc/intel_huc_fw.c b/drivers/gpu/drm/i915/gt/uc/intel_huc_fw.c new file mode 100644 index 000000000000..74602487ed67 --- /dev/null +++ b/drivers/gpu/drm/i915/gt/uc/intel_huc_fw.c @@ -0,0 +1,58 @@ +// SPDX-License-Identifier: MIT +/* + * Copyright © 2014-2019 Intel Corporation + */ + +#include "gt/intel_gt.h" +#include "intel_huc_fw.h" +#include "i915_drv.h" + +/** + * DOC: HuC Firmware + * + * Motivation: + * GEN9 introduces a new dedicated firmware for usage in media HEVC (High + * Efficiency Video Coding) operations. Userspace can use the firmware + * capabilities by adding HuC specific commands to batch buffers. + * + * Implementation: + * The same firmware loader is used as the GuC. However, the actual + * loading to HW is deferred until GEM initialization is done. + * + * Note that HuC firmware loading must be done before GuC loading. + */ + +/** + * intel_huc_fw_init_early() - initializes HuC firmware struct + * @huc: intel_huc struct + * + * On platforms with HuC selects firmware for uploading + */ +void intel_huc_fw_init_early(struct intel_huc *huc) +{ + struct intel_gt *gt = huc_to_gt(huc); + struct intel_uc *uc = >->uc; + struct drm_i915_private *i915 = gt->i915; + + intel_uc_fw_init_early(&huc->fw, INTEL_UC_FW_TYPE_HUC, + intel_uc_uses_guc(uc), + INTEL_INFO(i915)->platform, INTEL_REVID(i915)); +} + +/** + * intel_huc_fw_upload() - load HuC uCode to device + * @huc: intel_huc structure + * + * Called from intel_uc_init_hw() during driver load, resume from sleep and + * after a GPU reset. Note that HuC must be loaded before GuC. + * + * The firmware image should have already been fetched into memory, so only + * check that fetch succeeded, and then transfer the image to the h/w. + * + * Return: non-zero code on error + */ +int intel_huc_fw_upload(struct intel_huc *huc) +{ + /* HW doesn't look at destination address for HuC, so set it to 0 */ + return intel_uc_fw_upload(&huc->fw, huc_to_gt(huc), 0, HUC_UKERNEL); +} diff --git a/drivers/gpu/drm/i915/gt/uc/intel_huc_fw.h b/drivers/gpu/drm/i915/gt/uc/intel_huc_fw.h new file mode 100644 index 000000000000..b791269ce923 --- /dev/null +++ b/drivers/gpu/drm/i915/gt/uc/intel_huc_fw.h @@ -0,0 +1,14 @@ +/* SPDX-License-Identifier: MIT */ +/* + * Copyright © 2014-2019 Intel Corporation + */ + +#ifndef _INTEL_HUC_FW_H_ +#define _INTEL_HUC_FW_H_ + +struct intel_huc; + +void intel_huc_fw_init_early(struct intel_huc *huc); +int intel_huc_fw_upload(struct intel_huc *huc); + +#endif diff --git a/drivers/gpu/drm/i915/gt/uc/intel_uc.c b/drivers/gpu/drm/i915/gt/uc/intel_uc.c new file mode 100644 index 000000000000..71ee7ab035cc --- /dev/null +++ b/drivers/gpu/drm/i915/gt/uc/intel_uc.c @@ -0,0 +1,627 @@ +// SPDX-License-Identifier: MIT +/* + * Copyright © 2016-2019 Intel Corporation + */ + +#include "gt/intel_gt.h" +#include "gt/intel_reset.h" +#include "intel_guc.h" +#include "intel_guc_ads.h" +#include "intel_guc_submission.h" +#include "intel_uc.h" + +#include "i915_drv.h" + +/* Reset GuC providing us with fresh state for both GuC and HuC. + */ +static int __intel_uc_reset_hw(struct intel_uc *uc) +{ + struct intel_gt *gt = uc_to_gt(uc); + int ret; + u32 guc_status; + + ret = i915_inject_load_error(gt->i915, -ENXIO); + if (ret) + return ret; + + ret = intel_reset_guc(gt); + if (ret) { + DRM_ERROR("Failed to reset GuC, ret = %d\n", ret); + return ret; + } + + guc_status = intel_uncore_read(gt->uncore, GUC_STATUS); + WARN(!(guc_status & GS_MIA_IN_RESET), + "GuC status: 0x%x, MIA core expected to be in reset\n", + guc_status); + + return ret; +} + +static void __confirm_options(struct intel_uc *uc) +{ + struct drm_i915_private *i915 = uc_to_gt(uc)->i915; + + DRM_DEV_DEBUG_DRIVER(i915->drm.dev, + "enable_guc=%d (guc:%s submission:%s huc:%s)\n", + i915_modparams.enable_guc, + yesno(intel_uc_uses_guc(uc)), + yesno(intel_uc_uses_guc_submission(uc)), + yesno(intel_uc_uses_huc(uc))); + + if (i915_modparams.enable_guc == -1) + return; + + if (i915_modparams.enable_guc == 0) { + GEM_BUG_ON(intel_uc_uses_guc(uc)); + GEM_BUG_ON(intel_uc_uses_guc_submission(uc)); + GEM_BUG_ON(intel_uc_uses_huc(uc)); + return; + } + + if (!intel_uc_supports_guc(uc)) + dev_info(i915->drm.dev, + "Incompatible option enable_guc=%d - %s\n", + i915_modparams.enable_guc, "GuC is not supported!"); + + if (i915_modparams.enable_guc & ENABLE_GUC_LOAD_HUC && + !intel_uc_supports_huc(uc)) + dev_info(i915->drm.dev, + "Incompatible option enable_guc=%d - %s\n", + i915_modparams.enable_guc, "HuC is not supported!"); + + if (i915_modparams.enable_guc & ENABLE_GUC_SUBMISSION && + !intel_uc_supports_guc_submission(uc)) + dev_info(i915->drm.dev, + "Incompatible option enable_guc=%d - %s\n", + i915_modparams.enable_guc, "GuC submission is N/A"); + + if (i915_modparams.enable_guc & ~(ENABLE_GUC_SUBMISSION | + ENABLE_GUC_LOAD_HUC)) + dev_info(i915->drm.dev, + "Incompatible option enable_guc=%d - %s\n", + i915_modparams.enable_guc, "undocumented flag"); +} + +void intel_uc_init_early(struct intel_uc *uc) +{ + intel_guc_init_early(&uc->guc); + intel_huc_init_early(&uc->huc); + + __confirm_options(uc); +} + +void intel_uc_driver_late_release(struct intel_uc *uc) +{ +} + +/** + * intel_uc_init_mmio - setup uC MMIO access + * @uc: the intel_uc structure + * + * Setup minimal state necessary for MMIO accesses later in the + * initialization sequence. + */ +void intel_uc_init_mmio(struct intel_uc *uc) +{ + intel_guc_init_send_regs(&uc->guc); +} + +static void __uc_capture_load_err_log(struct intel_uc *uc) +{ + struct intel_guc *guc = &uc->guc; + + if (guc->log.vma && !uc->load_err_log) + uc->load_err_log = i915_gem_object_get(guc->log.vma->obj); +} + +static void __uc_free_load_err_log(struct intel_uc *uc) +{ + struct drm_i915_gem_object *log = fetch_and_zero(&uc->load_err_log); + + if (log) + i915_gem_object_put(log); +} + +/* + * Events triggered while CT buffers are disabled are logged in the SCRATCH_15 + * register using the same bits used in the CT message payload. Since our + * communication channel with guc is turned off at this point, we can save the + * message and handle it after we turn it back on. + */ +static void guc_clear_mmio_msg(struct intel_guc *guc) +{ + intel_uncore_write(guc_to_gt(guc)->uncore, SOFT_SCRATCH(15), 0); +} + +static void guc_get_mmio_msg(struct intel_guc *guc) +{ + u32 val; + + spin_lock_irq(&guc->irq_lock); + + val = intel_uncore_read(guc_to_gt(guc)->uncore, SOFT_SCRATCH(15)); + guc->mmio_msg |= val & guc->msg_enabled_mask; + + /* + * clear all events, including the ones we're not currently servicing, + * to make sure we don't try to process a stale message if we enable + * handling of more events later. + */ + guc_clear_mmio_msg(guc); + + spin_unlock_irq(&guc->irq_lock); +} + +static void guc_handle_mmio_msg(struct intel_guc *guc) +{ + struct drm_i915_private *i915 = guc_to_gt(guc)->i915; + + /* we need communication to be enabled to reply to GuC */ + GEM_BUG_ON(guc->handler == intel_guc_to_host_event_handler_nop); + + if (!guc->mmio_msg) + return; + + spin_lock_irq(&i915->irq_lock); + intel_guc_to_host_process_recv_msg(guc, &guc->mmio_msg, 1); + spin_unlock_irq(&i915->irq_lock); + + guc->mmio_msg = 0; +} + +static void guc_reset_interrupts(struct intel_guc *guc) +{ + guc->interrupts.reset(guc); +} + +static void guc_enable_interrupts(struct intel_guc *guc) +{ + guc->interrupts.enable(guc); +} + +static void guc_disable_interrupts(struct intel_guc *guc) +{ + guc->interrupts.disable(guc); +} + +static inline bool guc_communication_enabled(struct intel_guc *guc) +{ + return guc->send != intel_guc_send_nop; +} + +static int guc_enable_communication(struct intel_guc *guc) +{ + struct drm_i915_private *i915 = guc_to_gt(guc)->i915; + int ret; + + GEM_BUG_ON(guc_communication_enabled(guc)); + + ret = i915_inject_load_error(i915, -ENXIO); + if (ret) + return ret; + + ret = intel_guc_ct_enable(&guc->ct); + if (ret) + return ret; + + guc->send = intel_guc_send_ct; + guc->handler = intel_guc_to_host_event_handler_ct; + + /* check for mmio messages received before/during the CT enable */ + guc_get_mmio_msg(guc); + guc_handle_mmio_msg(guc); + + guc_enable_interrupts(guc); + + /* check for CT messages received before we enabled interrupts */ + spin_lock_irq(&i915->irq_lock); + intel_guc_to_host_event_handler_ct(guc); + spin_unlock_irq(&i915->irq_lock); + + DRM_INFO("GuC communication enabled\n"); + + return 0; +} + +static void guc_stop_communication(struct intel_guc *guc) +{ + intel_guc_ct_stop(&guc->ct); + + guc->send = intel_guc_send_nop; + guc->handler = intel_guc_to_host_event_handler_nop; + + guc_clear_mmio_msg(guc); +} + +static void guc_disable_communication(struct intel_guc *guc) +{ + /* + * Events generated during or after CT disable are logged by guc in + * via mmio. Make sure the register is clear before disabling CT since + * all events we cared about have already been processed via CT. + */ + guc_clear_mmio_msg(guc); + + guc_disable_interrupts(guc); + + guc->send = intel_guc_send_nop; + guc->handler = intel_guc_to_host_event_handler_nop; + + intel_guc_ct_disable(&guc->ct); + + /* + * Check for messages received during/after the CT disable. We do not + * expect any messages to have arrived via CT between the interrupt + * disable and the CT disable because GuC should've been idle until we + * triggered the CT disable protocol. + */ + guc_get_mmio_msg(guc); + + DRM_INFO("GuC communication disabled\n"); +} + +void intel_uc_fetch_firmwares(struct intel_uc *uc) +{ + struct drm_i915_private *i915 = uc_to_gt(uc)->i915; + int err; + + if (!intel_uc_uses_guc(uc)) + return; + + err = intel_uc_fw_fetch(&uc->guc.fw, i915); + if (err) + return; + + if (intel_uc_uses_huc(uc)) + intel_uc_fw_fetch(&uc->huc.fw, i915); +} + +void intel_uc_cleanup_firmwares(struct intel_uc *uc) +{ + if (!intel_uc_uses_guc(uc)) + return; + + if (intel_uc_uses_huc(uc)) + intel_uc_fw_cleanup_fetch(&uc->huc.fw); + + intel_uc_fw_cleanup_fetch(&uc->guc.fw); +} + +void intel_uc_init(struct intel_uc *uc) +{ + struct intel_guc *guc = &uc->guc; + struct intel_huc *huc = &uc->huc; + int ret; + + if (!intel_uc_uses_guc(uc)) + return; + + /* XXX: GuC submission is unavailable for now */ + GEM_BUG_ON(intel_uc_supports_guc_submission(uc)); + + ret = intel_guc_init(guc); + if (ret) { + intel_uc_fw_cleanup_fetch(&huc->fw); + return; + } + + if (intel_uc_uses_huc(uc)) + intel_huc_init(huc); +} + +void intel_uc_fini(struct intel_uc *uc) +{ + struct intel_guc *guc = &uc->guc; + + if (!intel_uc_uses_guc(uc)) + return; + + if (intel_uc_uses_huc(uc)) + intel_huc_fini(&uc->huc); + + intel_guc_fini(guc); + + __uc_free_load_err_log(uc); +} + +static int __uc_sanitize(struct intel_uc *uc) +{ + struct intel_guc *guc = &uc->guc; + struct intel_huc *huc = &uc->huc; + + GEM_BUG_ON(!intel_uc_supports_guc(uc)); + + intel_huc_sanitize(huc); + intel_guc_sanitize(guc); + + return __intel_uc_reset_hw(uc); +} + +void intel_uc_sanitize(struct intel_uc *uc) +{ + if (!intel_uc_supports_guc(uc)) + return; + + __uc_sanitize(uc); +} + +/* Initialize and verify the uC regs related to uC positioning in WOPCM */ +static int uc_init_wopcm(struct intel_uc *uc) +{ + struct intel_gt *gt = uc_to_gt(uc); + struct intel_uncore *uncore = gt->uncore; + u32 base = intel_wopcm_guc_base(>->i915->wopcm); + u32 size = intel_wopcm_guc_size(>->i915->wopcm); + u32 huc_agent = intel_uc_uses_huc(uc) ? HUC_LOADING_AGENT_GUC : 0; + u32 mask; + int err; + + if (unlikely(!base || !size)) { + i915_probe_error(gt->i915, "Unsuccessful WOPCM partitioning\n"); + return -E2BIG; + } + + GEM_BUG_ON(!intel_uc_supports_guc(uc)); + GEM_BUG_ON(!(base & GUC_WOPCM_OFFSET_MASK)); + GEM_BUG_ON(base & ~GUC_WOPCM_OFFSET_MASK); + GEM_BUG_ON(!(size & GUC_WOPCM_SIZE_MASK)); + GEM_BUG_ON(size & ~GUC_WOPCM_SIZE_MASK); + + err = i915_inject_load_error(gt->i915, -ENXIO); + if (err) + return err; + + mask = GUC_WOPCM_SIZE_MASK | GUC_WOPCM_SIZE_LOCKED; + err = intel_uncore_write_and_verify(uncore, GUC_WOPCM_SIZE, size, mask, + size | GUC_WOPCM_SIZE_LOCKED); + if (err) + goto err_out; + + mask = GUC_WOPCM_OFFSET_MASK | GUC_WOPCM_OFFSET_VALID | huc_agent; + err = intel_uncore_write_and_verify(uncore, DMA_GUC_WOPCM_OFFSET, + base | huc_agent, mask, + base | huc_agent | + GUC_WOPCM_OFFSET_VALID); + if (err) + goto err_out; + + return 0; + +err_out: + i915_probe_error(gt->i915, "Failed to init uC WOPCM registers!\n"); + i915_probe_error(gt->i915, "%s(%#x)=%#x\n", "DMA_GUC_WOPCM_OFFSET", + i915_mmio_reg_offset(DMA_GUC_WOPCM_OFFSET), + intel_uncore_read(uncore, DMA_GUC_WOPCM_OFFSET)); + i915_probe_error(gt->i915, "%s(%#x)=%#x\n", "GUC_WOPCM_SIZE", + i915_mmio_reg_offset(GUC_WOPCM_SIZE), + intel_uncore_read(uncore, GUC_WOPCM_SIZE)); + + return err; +} + +static bool uc_is_wopcm_locked(struct intel_uc *uc) +{ + struct intel_gt *gt = uc_to_gt(uc); + struct intel_uncore *uncore = gt->uncore; + + return (intel_uncore_read(uncore, GUC_WOPCM_SIZE) & GUC_WOPCM_SIZE_LOCKED) || + (intel_uncore_read(uncore, DMA_GUC_WOPCM_OFFSET) & GUC_WOPCM_OFFSET_VALID); +} + +int intel_uc_init_hw(struct intel_uc *uc) +{ + struct drm_i915_private *i915 = uc_to_gt(uc)->i915; + struct intel_guc *guc = &uc->guc; + struct intel_huc *huc = &uc->huc; + int ret, attempts; + + if (!intel_uc_supports_guc(uc)) + return 0; + + /* + * We can silently continue without GuC only if it was never enabled + * before on this system after reboot, otherwise we risk GPU hangs. + * To check if GuC was loaded before we look at WOPCM registers. + */ + if (!intel_uc_uses_guc(uc) && !uc_is_wopcm_locked(uc)) + return 0; + + if (!intel_uc_fw_is_available(&guc->fw)) { + ret = uc_is_wopcm_locked(uc) || + intel_uc_fw_is_overridden(&guc->fw) || + intel_uc_supports_guc_submission(uc) ? + intel_uc_fw_status_to_error(guc->fw.status) : 0; + goto err_out; + } + + ret = uc_init_wopcm(uc); + if (ret) + goto err_out; + + guc_reset_interrupts(guc); + + /* WaEnableuKernelHeaderValidFix:skl */ + /* WaEnableGuCBootHashCheckNotSet:skl,bxt,kbl */ + if (IS_GEN(i915, 9)) + attempts = 3; + else + attempts = 1; + + while (attempts--) { + /* + * Always reset the GuC just before (re)loading, so + * that the state and timing are fairly predictable + */ + ret = __uc_sanitize(uc); + if (ret) + goto err_out; + + intel_huc_fw_upload(huc); + intel_guc_ads_reset(guc); + intel_guc_write_params(guc); + ret = intel_guc_fw_upload(guc); + if (ret == 0) + break; + + DRM_DEBUG_DRIVER("GuC fw load failed: %d; will reset and " + "retry %d more time(s)\n", ret, attempts); + } + + /* Did we succeded or run out of retries? */ + if (ret) + goto err_log_capture; + + ret = guc_enable_communication(guc); + if (ret) + goto err_log_capture; + + intel_huc_auth(huc); + + ret = intel_guc_sample_forcewake(guc); + if (ret) + goto err_communication; + + if (intel_uc_supports_guc_submission(uc)) { + ret = intel_guc_submission_enable(guc); + if (ret) + goto err_communication; + } + + dev_info(i915->drm.dev, "%s firmware %s version %u.%u %s:%s\n", + intel_uc_fw_type_repr(INTEL_UC_FW_TYPE_GUC), guc->fw.path, + guc->fw.major_ver_found, guc->fw.minor_ver_found, + "submission", + enableddisabled(intel_uc_supports_guc_submission(uc))); + + if (intel_uc_uses_huc(uc)) { + dev_info(i915->drm.dev, "%s firmware %s version %u.%u %s:%s\n", + intel_uc_fw_type_repr(INTEL_UC_FW_TYPE_HUC), + huc->fw.path, + huc->fw.major_ver_found, huc->fw.minor_ver_found, + "authenticated", + yesno(intel_huc_is_authenticated(huc))); + } + + return 0; + + /* + * We've failed to load the firmware :( + */ +err_communication: + guc_disable_communication(guc); +err_log_capture: + __uc_capture_load_err_log(uc); +err_out: + __uc_sanitize(uc); + + if (!ret) { + dev_notice(i915->drm.dev, "GuC is uninitialized\n"); + /* We want to run without GuC submission */ + return 0; + } + + i915_probe_error(i915, "GuC initialization failed %d\n", ret); + + /* We want to keep KMS alive */ + return -EIO; +} + +void intel_uc_fini_hw(struct intel_uc *uc) +{ + struct intel_guc *guc = &uc->guc; + + if (!intel_guc_is_running(guc)) + return; + + if (intel_uc_supports_guc_submission(uc)) + intel_guc_submission_disable(guc); + + guc_disable_communication(guc); + __uc_sanitize(uc); +} + +/** + * intel_uc_reset_prepare - Prepare for reset + * @uc: the intel_uc structure + * + * Preparing for full gpu reset. + */ +void intel_uc_reset_prepare(struct intel_uc *uc) +{ + struct intel_guc *guc = &uc->guc; + + if (!intel_guc_is_running(guc)) + return; + + guc_stop_communication(guc); + __uc_sanitize(uc); +} + +void intel_uc_runtime_suspend(struct intel_uc *uc) +{ + struct intel_guc *guc = &uc->guc; + int err; + + if (!intel_guc_is_running(guc)) + return; + + err = intel_guc_suspend(guc); + if (err) + DRM_DEBUG_DRIVER("Failed to suspend GuC, err=%d", err); + + guc_disable_communication(guc); +} + +void intel_uc_suspend(struct intel_uc *uc) +{ + struct intel_guc *guc = &uc->guc; + intel_wakeref_t wakeref; + + if (!intel_guc_is_running(guc)) + return; + + with_intel_runtime_pm(&uc_to_gt(uc)->i915->runtime_pm, wakeref) + intel_uc_runtime_suspend(uc); +} + +static int __uc_resume(struct intel_uc *uc, bool enable_communication) +{ + struct intel_guc *guc = &uc->guc; + int err; + + if (!intel_guc_is_running(guc)) + return 0; + + /* Make sure we enable communication if and only if it's disabled */ + GEM_BUG_ON(enable_communication == guc_communication_enabled(guc)); + + if (enable_communication) + guc_enable_communication(guc); + + err = intel_guc_resume(guc); + if (err) { + DRM_DEBUG_DRIVER("Failed to resume GuC, err=%d", err); + return err; + } + + return 0; +} + +int intel_uc_resume(struct intel_uc *uc) +{ + /* + * When coming out of S3/S4 we sanitize and re-init the HW, so + * communication is already re-enabled at this point. + */ + return __uc_resume(uc, false); +} + +int intel_uc_runtime_resume(struct intel_uc *uc) +{ + /* + * During runtime resume we don't sanitize, so we need to re-init + * communication as well. + */ + return __uc_resume(uc, true); +} diff --git a/drivers/gpu/drm/i915/gt/uc/intel_uc.h b/drivers/gpu/drm/i915/gt/uc/intel_uc.h new file mode 100644 index 000000000000..527995c21196 --- /dev/null +++ b/drivers/gpu/drm/i915/gt/uc/intel_uc.h @@ -0,0 +1,67 @@ +/* SPDX-License-Identifier: MIT */ +/* + * Copyright © 2014-2019 Intel Corporation + */ + +#ifndef _INTEL_UC_H_ +#define _INTEL_UC_H_ + +#include "intel_guc.h" +#include "intel_huc.h" +#include "i915_params.h" + +struct intel_uc { + struct intel_guc guc; + struct intel_huc huc; + + /* Snapshot of GuC log from last failed load */ + struct drm_i915_gem_object *load_err_log; +}; + +void intel_uc_init_early(struct intel_uc *uc); +void intel_uc_driver_late_release(struct intel_uc *uc); +void intel_uc_init_mmio(struct intel_uc *uc); +void intel_uc_fetch_firmwares(struct intel_uc *uc); +void intel_uc_cleanup_firmwares(struct intel_uc *uc); +void intel_uc_sanitize(struct intel_uc *uc); +void intel_uc_init(struct intel_uc *uc); +int intel_uc_init_hw(struct intel_uc *uc); +void intel_uc_fini_hw(struct intel_uc *uc); +void intel_uc_fini(struct intel_uc *uc); +void intel_uc_reset_prepare(struct intel_uc *uc); +void intel_uc_suspend(struct intel_uc *uc); +void intel_uc_runtime_suspend(struct intel_uc *uc); +int intel_uc_resume(struct intel_uc *uc); +int intel_uc_runtime_resume(struct intel_uc *uc); + +static inline bool intel_uc_supports_guc(struct intel_uc *uc) +{ + return intel_guc_is_supported(&uc->guc); +} + +static inline bool intel_uc_uses_guc(struct intel_uc *uc) +{ + return intel_guc_is_enabled(&uc->guc); +} + +static inline bool intel_uc_supports_guc_submission(struct intel_uc *uc) +{ + return intel_guc_is_submission_supported(&uc->guc); +} + +static inline bool intel_uc_uses_guc_submission(struct intel_uc *uc) +{ + return intel_guc_is_submission_supported(&uc->guc); +} + +static inline bool intel_uc_supports_huc(struct intel_uc *uc) +{ + return intel_uc_supports_guc(uc); +} + +static inline bool intel_uc_uses_huc(struct intel_uc *uc) +{ + return intel_huc_is_enabled(&uc->huc); +} + +#endif diff --git a/drivers/gpu/drm/i915/gt/uc/intel_uc_fw.c b/drivers/gpu/drm/i915/gt/uc/intel_uc_fw.c new file mode 100644 index 000000000000..bd22bf11adad --- /dev/null +++ b/drivers/gpu/drm/i915/gt/uc/intel_uc_fw.c @@ -0,0 +1,616 @@ +// SPDX-License-Identifier: MIT +/* + * Copyright © 2016-2019 Intel Corporation + */ + +#include <linux/bitfield.h> +#include <linux/firmware.h> +#include <drm/drm_print.h> + +#include "intel_uc_fw.h" +#include "intel_uc_fw_abi.h" +#include "i915_drv.h" + +#ifdef CONFIG_DRM_I915_DEBUG_GUC +static inline struct intel_gt *__uc_fw_to_gt(struct intel_uc_fw *uc_fw) +{ + GEM_BUG_ON(uc_fw->status == INTEL_UC_FIRMWARE_UNINITIALIZED); + if (uc_fw->type == INTEL_UC_FW_TYPE_GUC) + return container_of(uc_fw, struct intel_gt, uc.guc.fw); + + GEM_BUG_ON(uc_fw->type != INTEL_UC_FW_TYPE_HUC); + return container_of(uc_fw, struct intel_gt, uc.huc.fw); +} + +void intel_uc_fw_change_status(struct intel_uc_fw *uc_fw, + enum intel_uc_fw_status status) +{ + uc_fw->__status = status; + DRM_DEV_DEBUG_DRIVER(__uc_fw_to_gt(uc_fw)->i915->drm.dev, + "%s firmware -> %s\n", + intel_uc_fw_type_repr(uc_fw->type), + status == INTEL_UC_FIRMWARE_SELECTED ? + uc_fw->path : intel_uc_fw_status_repr(status)); +} +#endif + +/* + * List of required GuC and HuC binaries per-platform. + * Must be ordered based on platform + revid, from newer to older. + */ +#define INTEL_UC_FIRMWARE_DEFS(fw_def, guc_def, huc_def) \ + fw_def(ICELAKE, 0, guc_def(icl, 33, 0, 0), huc_def(icl, 8, 4, 3238)) \ + fw_def(COFFEELAKE, 0, guc_def(kbl, 33, 0, 0), huc_def(kbl, 02, 00, 1810)) \ + fw_def(GEMINILAKE, 0, guc_def(glk, 33, 0, 0), huc_def(glk, 03, 01, 2893)) \ + fw_def(KABYLAKE, 0, guc_def(kbl, 33, 0, 0), huc_def(kbl, 02, 00, 1810)) \ + fw_def(BROXTON, 0, guc_def(bxt, 33, 0, 0), huc_def(bxt, 01, 8, 2893)) \ + fw_def(SKYLAKE, 0, guc_def(skl, 33, 0, 0), huc_def(skl, 01, 07, 1398)) + +#define __MAKE_UC_FW_PATH(prefix_, name_, separator_, major_, minor_, patch_) \ + "i915/" \ + __stringify(prefix_) name_ \ + __stringify(major_) separator_ \ + __stringify(minor_) separator_ \ + __stringify(patch_) ".bin" + +#define MAKE_GUC_FW_PATH(prefix_, major_, minor_, patch_) \ + __MAKE_UC_FW_PATH(prefix_, "_guc_", ".", major_, minor_, patch_) + +#define MAKE_HUC_FW_PATH(prefix_, major_, minor_, bld_num_) \ + __MAKE_UC_FW_PATH(prefix_, "_huc_ver", "_", major_, minor_, bld_num_) + +/* All blobs need to be declared via MODULE_FIRMWARE() */ +#define INTEL_UC_MODULE_FW(platform_, revid_, guc_, huc_) \ + MODULE_FIRMWARE(guc_); \ + MODULE_FIRMWARE(huc_); + +INTEL_UC_FIRMWARE_DEFS(INTEL_UC_MODULE_FW, MAKE_GUC_FW_PATH, MAKE_HUC_FW_PATH) + +/* The below structs and macros are used to iterate across the list of blobs */ +struct __packed uc_fw_blob { + u8 major; + u8 minor; + const char *path; +}; + +#define UC_FW_BLOB(major_, minor_, path_) \ + { .major = major_, .minor = minor_, .path = path_ } + +#define GUC_FW_BLOB(prefix_, major_, minor_, patch_) \ + UC_FW_BLOB(major_, minor_, \ + MAKE_GUC_FW_PATH(prefix_, major_, minor_, patch_)) + +#define HUC_FW_BLOB(prefix_, major_, minor_, bld_num_) \ + UC_FW_BLOB(major_, minor_, \ + MAKE_HUC_FW_PATH(prefix_, major_, minor_, bld_num_)) + +struct __packed uc_fw_platform_requirement { + enum intel_platform p; + u8 rev; /* first platform rev using this FW */ + const struct uc_fw_blob blobs[INTEL_UC_FW_NUM_TYPES]; +}; + +#define MAKE_FW_LIST(platform_, revid_, guc_, huc_) \ +{ \ + .p = INTEL_##platform_, \ + .rev = revid_, \ + .blobs[INTEL_UC_FW_TYPE_GUC] = guc_, \ + .blobs[INTEL_UC_FW_TYPE_HUC] = huc_, \ +}, + +static void +__uc_fw_auto_select(struct intel_uc_fw *uc_fw, enum intel_platform p, u8 rev) +{ + static const struct uc_fw_platform_requirement fw_blobs[] = { + INTEL_UC_FIRMWARE_DEFS(MAKE_FW_LIST, GUC_FW_BLOB, HUC_FW_BLOB) + }; + int i; + + for (i = 0; i < ARRAY_SIZE(fw_blobs) && p <= fw_blobs[i].p; i++) { + if (p == fw_blobs[i].p && rev >= fw_blobs[i].rev) { + const struct uc_fw_blob *blob = + &fw_blobs[i].blobs[uc_fw->type]; + uc_fw->path = blob->path; + uc_fw->major_ver_wanted = blob->major; + uc_fw->minor_ver_wanted = blob->minor; + break; + } + } + + /* make sure the list is ordered as expected */ + if (IS_ENABLED(CONFIG_DRM_I915_SELFTEST)) { + for (i = 1; i < ARRAY_SIZE(fw_blobs); i++) { + if (fw_blobs[i].p < fw_blobs[i - 1].p) + continue; + + if (fw_blobs[i].p == fw_blobs[i - 1].p && + fw_blobs[i].rev < fw_blobs[i - 1].rev) + continue; + + pr_err("invalid FW blob order: %s r%u comes before %s r%u\n", + intel_platform_name(fw_blobs[i - 1].p), + fw_blobs[i - 1].rev, + intel_platform_name(fw_blobs[i].p), + fw_blobs[i].rev); + + uc_fw->path = NULL; + } + } + + /* We don't want to enable GuC/HuC on pre-Gen11 by default */ + if (i915_modparams.enable_guc == -1 && p < INTEL_ICELAKE) + uc_fw->path = NULL; +} + +static const char *__override_guc_firmware_path(void) +{ + if (i915_modparams.enable_guc & (ENABLE_GUC_SUBMISSION | + ENABLE_GUC_LOAD_HUC)) + return i915_modparams.guc_firmware_path; + return ""; +} + +static const char *__override_huc_firmware_path(void) +{ + if (i915_modparams.enable_guc & ENABLE_GUC_LOAD_HUC) + return i915_modparams.huc_firmware_path; + return ""; +} + +static void __uc_fw_user_override(struct intel_uc_fw *uc_fw) +{ + const char *path = NULL; + + switch (uc_fw->type) { + case INTEL_UC_FW_TYPE_GUC: + path = __override_guc_firmware_path(); + break; + case INTEL_UC_FW_TYPE_HUC: + path = __override_huc_firmware_path(); + break; + } + + if (unlikely(path)) { + uc_fw->path = path; + uc_fw->user_overridden = true; + } +} + +/** + * intel_uc_fw_init_early - initialize the uC object and select the firmware + * @uc_fw: uC firmware + * @type: type of uC + * @supported: is uC support possible + * @platform: platform identifier + * @rev: hardware revision + * + * Initialize the state of our uC object and relevant tracking and select the + * firmware to fetch and load. + */ +void intel_uc_fw_init_early(struct intel_uc_fw *uc_fw, + enum intel_uc_fw_type type, bool supported, + enum intel_platform platform, u8 rev) +{ + /* + * we use FIRMWARE_UNINITIALIZED to detect checks against uc_fw->status + * before we're looked at the HW caps to see if we have uc support + */ + BUILD_BUG_ON(INTEL_UC_FIRMWARE_UNINITIALIZED); + GEM_BUG_ON(uc_fw->status); + GEM_BUG_ON(uc_fw->path); + + uc_fw->type = type; + + if (supported) { + __uc_fw_auto_select(uc_fw, platform, rev); + __uc_fw_user_override(uc_fw); + } + + intel_uc_fw_change_status(uc_fw, uc_fw->path ? *uc_fw->path ? + INTEL_UC_FIRMWARE_SELECTED : + INTEL_UC_FIRMWARE_DISABLED : + INTEL_UC_FIRMWARE_NOT_SUPPORTED); +} + +static void __force_fw_fetch_failures(struct intel_uc_fw *uc_fw, + struct drm_i915_private *i915, + int e) +{ + bool user = e == -EINVAL; + + if (i915_inject_load_error(i915, e)) { + /* non-existing blob */ + uc_fw->path = "<invalid>"; + uc_fw->user_overridden = user; + } else if (i915_inject_load_error(i915, e)) { + /* require next major version */ + uc_fw->major_ver_wanted += 1; + uc_fw->minor_ver_wanted = 0; + uc_fw->user_overridden = user; + } else if (i915_inject_load_error(i915, e)) { + /* require next minor version */ + uc_fw->minor_ver_wanted += 1; + uc_fw->user_overridden = user; + } else if (uc_fw->major_ver_wanted && i915_inject_load_error(i915, e)) { + /* require prev major version */ + uc_fw->major_ver_wanted -= 1; + uc_fw->minor_ver_wanted = 0; + uc_fw->user_overridden = user; + } else if (uc_fw->minor_ver_wanted && i915_inject_load_error(i915, e)) { + /* require prev minor version - hey, this should work! */ + uc_fw->minor_ver_wanted -= 1; + uc_fw->user_overridden = user; + } else if (user && i915_inject_load_error(i915, e)) { + /* officially unsupported platform */ + uc_fw->major_ver_wanted = 0; + uc_fw->minor_ver_wanted = 0; + uc_fw->user_overridden = true; + } +} + +/** + * intel_uc_fw_fetch - fetch uC firmware + * @uc_fw: uC firmware + * @i915: device private + * + * Fetch uC firmware into GEM obj. + * + * Return: 0 on success, a negative errno code on failure. + */ +int intel_uc_fw_fetch(struct intel_uc_fw *uc_fw, struct drm_i915_private *i915) +{ + struct device *dev = i915->drm.dev; + struct drm_i915_gem_object *obj; + const struct firmware *fw = NULL; + struct uc_css_header *css; + size_t size; + int err; + + GEM_BUG_ON(!i915->wopcm.size); + GEM_BUG_ON(!intel_uc_fw_is_enabled(uc_fw)); + + err = i915_inject_load_error(i915, -ENXIO); + if (err) + return err; + + __force_fw_fetch_failures(uc_fw, i915, -EINVAL); + __force_fw_fetch_failures(uc_fw, i915, -ESTALE); + + err = request_firmware(&fw, uc_fw->path, dev); + if (err) + goto fail; + + /* Check the size of the blob before examining buffer contents */ + if (unlikely(fw->size < sizeof(struct uc_css_header))) { + dev_warn(dev, "%s firmware %s: invalid size: %zu < %zu\n", + intel_uc_fw_type_repr(uc_fw->type), uc_fw->path, + fw->size, sizeof(struct uc_css_header)); + err = -ENODATA; + goto fail; + } + + css = (struct uc_css_header *)fw->data; + + /* Check integrity of size values inside CSS header */ + size = (css->header_size_dw - css->key_size_dw - css->modulus_size_dw - + css->exponent_size_dw) * sizeof(u32); + if (unlikely(size != sizeof(struct uc_css_header))) { + dev_warn(dev, + "%s firmware %s: unexpected header size: %zu != %zu\n", + intel_uc_fw_type_repr(uc_fw->type), uc_fw->path, + fw->size, sizeof(struct uc_css_header)); + err = -EPROTO; + goto fail; + } + + /* uCode size must calculated from other sizes */ + uc_fw->ucode_size = (css->size_dw - css->header_size_dw) * sizeof(u32); + + /* now RSA */ + if (unlikely(css->key_size_dw != UOS_RSA_SCRATCH_COUNT)) { + dev_warn(dev, "%s firmware %s: unexpected key size: %u != %u\n", + intel_uc_fw_type_repr(uc_fw->type), uc_fw->path, + css->key_size_dw, UOS_RSA_SCRATCH_COUNT); + err = -EPROTO; + goto fail; + } + uc_fw->rsa_size = css->key_size_dw * sizeof(u32); + + /* At least, it should have header, uCode and RSA. Size of all three. */ + size = sizeof(struct uc_css_header) + uc_fw->ucode_size + uc_fw->rsa_size; + if (unlikely(fw->size < size)) { + dev_warn(dev, "%s firmware %s: invalid size: %zu < %zu\n", + intel_uc_fw_type_repr(uc_fw->type), uc_fw->path, + fw->size, size); + err = -ENOEXEC; + goto fail; + } + + /* Sanity check whether this fw is not larger than whole WOPCM memory */ + size = __intel_uc_fw_get_upload_size(uc_fw); + if (unlikely(size >= i915->wopcm.size)) { + dev_warn(dev, "%s firmware %s: invalid size: %zu > %zu\n", + intel_uc_fw_type_repr(uc_fw->type), uc_fw->path, + size, (size_t)i915->wopcm.size); + err = -E2BIG; + goto fail; + } + + /* Get version numbers from the CSS header */ + switch (uc_fw->type) { + case INTEL_UC_FW_TYPE_GUC: + uc_fw->major_ver_found = FIELD_GET(CSS_SW_VERSION_GUC_MAJOR, + css->sw_version); + uc_fw->minor_ver_found = FIELD_GET(CSS_SW_VERSION_GUC_MINOR, + css->sw_version); + break; + + case INTEL_UC_FW_TYPE_HUC: + uc_fw->major_ver_found = FIELD_GET(CSS_SW_VERSION_HUC_MAJOR, + css->sw_version); + uc_fw->minor_ver_found = FIELD_GET(CSS_SW_VERSION_HUC_MINOR, + css->sw_version); + break; + + default: + MISSING_CASE(uc_fw->type); + break; + } + + if (uc_fw->major_ver_found != uc_fw->major_ver_wanted || + uc_fw->minor_ver_found < uc_fw->minor_ver_wanted) { + dev_notice(dev, "%s firmware %s: unexpected version: %u.%u != %u.%u\n", + intel_uc_fw_type_repr(uc_fw->type), uc_fw->path, + uc_fw->major_ver_found, uc_fw->minor_ver_found, + uc_fw->major_ver_wanted, uc_fw->minor_ver_wanted); + if (!intel_uc_fw_is_overridden(uc_fw)) { + err = -ENOEXEC; + goto fail; + } + } + + obj = i915_gem_object_create_shmem_from_data(i915, fw->data, fw->size); + if (IS_ERR(obj)) { + err = PTR_ERR(obj); + goto fail; + } + + uc_fw->obj = obj; + uc_fw->size = fw->size; + intel_uc_fw_change_status(uc_fw, INTEL_UC_FIRMWARE_AVAILABLE); + + release_firmware(fw); + return 0; + +fail: + intel_uc_fw_change_status(uc_fw, err == -ENOENT ? + INTEL_UC_FIRMWARE_MISSING : + INTEL_UC_FIRMWARE_ERROR); + + dev_notice(dev, "%s firmware %s: fetch failed with error %d\n", + intel_uc_fw_type_repr(uc_fw->type), uc_fw->path, err); + dev_info(dev, "%s firmware(s) can be downloaded from %s\n", + intel_uc_fw_type_repr(uc_fw->type), INTEL_UC_FIRMWARE_URL); + + release_firmware(fw); /* OK even if fw is NULL */ + return err; +} + +static u32 uc_fw_ggtt_offset(struct intel_uc_fw *uc_fw, struct i915_ggtt *ggtt) +{ + struct drm_mm_node *node = &ggtt->uc_fw; + + GEM_BUG_ON(!node->allocated); + GEM_BUG_ON(upper_32_bits(node->start)); + GEM_BUG_ON(upper_32_bits(node->start + node->size - 1)); + + return lower_32_bits(node->start); +} + +static void intel_uc_fw_ggtt_bind(struct intel_uc_fw *uc_fw, + struct intel_gt *gt) +{ + struct drm_i915_gem_object *obj = uc_fw->obj; + struct i915_ggtt *ggtt = gt->ggtt; + struct i915_vma dummy = { + .node.start = uc_fw_ggtt_offset(uc_fw, ggtt), + .node.size = obj->base.size, + .pages = obj->mm.pages, + .vm = &ggtt->vm, + }; + + GEM_BUG_ON(!i915_gem_object_has_pinned_pages(obj)); + GEM_BUG_ON(dummy.node.size > ggtt->uc_fw.size); + + /* uc_fw->obj cache domains were not controlled across suspend */ + drm_clflush_sg(dummy.pages); + + ggtt->vm.insert_entries(&ggtt->vm, &dummy, I915_CACHE_NONE, 0); +} + +static void intel_uc_fw_ggtt_unbind(struct intel_uc_fw *uc_fw, + struct intel_gt *gt) +{ + struct drm_i915_gem_object *obj = uc_fw->obj; + struct i915_ggtt *ggtt = gt->ggtt; + u64 start = uc_fw_ggtt_offset(uc_fw, ggtt); + + ggtt->vm.clear_range(&ggtt->vm, start, obj->base.size); +} + +static int uc_fw_xfer(struct intel_uc_fw *uc_fw, struct intel_gt *gt, + u32 wopcm_offset, u32 dma_flags) +{ + struct intel_uncore *uncore = gt->uncore; + u64 offset; + int ret; + + ret = i915_inject_load_error(gt->i915, -ETIMEDOUT); + if (ret) + return ret; + + intel_uncore_forcewake_get(uncore, FORCEWAKE_ALL); + + /* Set the source address for the uCode */ + offset = uc_fw_ggtt_offset(uc_fw, gt->ggtt); + GEM_BUG_ON(upper_32_bits(offset) & 0xFFFF0000); + intel_uncore_write_fw(uncore, DMA_ADDR_0_LOW, lower_32_bits(offset)); + intel_uncore_write_fw(uncore, DMA_ADDR_0_HIGH, upper_32_bits(offset)); + + /* Set the DMA destination */ + intel_uncore_write_fw(uncore, DMA_ADDR_1_LOW, wopcm_offset); + intel_uncore_write_fw(uncore, DMA_ADDR_1_HIGH, DMA_ADDRESS_SPACE_WOPCM); + + /* + * Set the transfer size. The header plus uCode will be copied to WOPCM + * via DMA, excluding any other components + */ + intel_uncore_write_fw(uncore, DMA_COPY_SIZE, + sizeof(struct uc_css_header) + uc_fw->ucode_size); + + /* Start the DMA */ + intel_uncore_write_fw(uncore, DMA_CTRL, + _MASKED_BIT_ENABLE(dma_flags | START_DMA)); + + /* Wait for DMA to finish */ + ret = intel_wait_for_register_fw(uncore, DMA_CTRL, START_DMA, 0, 100); + if (ret) + dev_err(gt->i915->drm.dev, "DMA for %s fw failed, DMA_CTRL=%u\n", + intel_uc_fw_type_repr(uc_fw->type), + intel_uncore_read_fw(uncore, DMA_CTRL)); + + /* Disable the bits once DMA is over */ + intel_uncore_write_fw(uncore, DMA_CTRL, _MASKED_BIT_DISABLE(dma_flags)); + + intel_uncore_forcewake_put(uncore, FORCEWAKE_ALL); + + return ret; +} + +/** + * intel_uc_fw_upload - load uC firmware using custom loader + * @uc_fw: uC firmware + * @gt: the intel_gt structure + * @wopcm_offset: destination offset in wopcm + * @dma_flags: flags for flags for dma ctrl + * + * Loads uC firmware and updates internal flags. + * + * Return: 0 on success, non-zero on failure. + */ +int intel_uc_fw_upload(struct intel_uc_fw *uc_fw, struct intel_gt *gt, + u32 wopcm_offset, u32 dma_flags) +{ + int err; + + /* make sure the status was cleared the last time we reset the uc */ + GEM_BUG_ON(intel_uc_fw_is_loaded(uc_fw)); + + err = i915_inject_load_error(gt->i915, -ENOEXEC); + if (err) + return err; + + if (!intel_uc_fw_is_available(uc_fw)) + return -ENOEXEC; + + /* Call custom loader */ + intel_uc_fw_ggtt_bind(uc_fw, gt); + err = uc_fw_xfer(uc_fw, gt, wopcm_offset, dma_flags); + intel_uc_fw_ggtt_unbind(uc_fw, gt); + if (err) + goto fail; + + intel_uc_fw_change_status(uc_fw, INTEL_UC_FIRMWARE_TRANSFERRED); + return 0; + +fail: + i915_probe_error(gt->i915, "Failed to load %s firmware %s (%d)\n", + intel_uc_fw_type_repr(uc_fw->type), uc_fw->path, + err); + intel_uc_fw_change_status(uc_fw, INTEL_UC_FIRMWARE_FAIL); + return err; +} + +int intel_uc_fw_init(struct intel_uc_fw *uc_fw) +{ + int err; + + /* this should happen before the load! */ + GEM_BUG_ON(intel_uc_fw_is_loaded(uc_fw)); + + if (!intel_uc_fw_is_available(uc_fw)) + return -ENOEXEC; + + err = i915_gem_object_pin_pages(uc_fw->obj); + if (err) { + DRM_DEBUG_DRIVER("%s fw pin-pages err=%d\n", + intel_uc_fw_type_repr(uc_fw->type), err); + intel_uc_fw_change_status(uc_fw, INTEL_UC_FIRMWARE_FAIL); + } + + return err; +} + +void intel_uc_fw_fini(struct intel_uc_fw *uc_fw) +{ + if (!intel_uc_fw_is_available(uc_fw)) + return; + + i915_gem_object_unpin_pages(uc_fw->obj); +} + +/** + * intel_uc_fw_cleanup_fetch - cleanup uC firmware + * @uc_fw: uC firmware + * + * Cleans up uC firmware by releasing the firmware GEM obj. + */ +void intel_uc_fw_cleanup_fetch(struct intel_uc_fw *uc_fw) +{ + if (!intel_uc_fw_is_available(uc_fw)) + return; + + i915_gem_object_put(fetch_and_zero(&uc_fw->obj)); + + intel_uc_fw_change_status(uc_fw, INTEL_UC_FIRMWARE_SELECTED); +} + +/** + * intel_uc_fw_copy_rsa - copy fw RSA to buffer + * + * @uc_fw: uC firmware + * @dst: dst buffer + * @max_len: max number of bytes to copy + * + * Return: number of copied bytes. + */ +size_t intel_uc_fw_copy_rsa(struct intel_uc_fw *uc_fw, void *dst, u32 max_len) +{ + struct sg_table *pages = uc_fw->obj->mm.pages; + u32 size = min_t(u32, uc_fw->rsa_size, max_len); + u32 offset = sizeof(struct uc_css_header) + uc_fw->ucode_size; + + GEM_BUG_ON(!intel_uc_fw_is_available(uc_fw)); + + return sg_pcopy_to_buffer(pages->sgl, pages->nents, dst, size, offset); +} + +/** + * intel_uc_fw_dump - dump information about uC firmware + * @uc_fw: uC firmware + * @p: the &drm_printer + * + * Pretty printer for uC firmware. + */ +void intel_uc_fw_dump(const struct intel_uc_fw *uc_fw, struct drm_printer *p) +{ + drm_printf(p, "%s firmware: %s\n", + intel_uc_fw_type_repr(uc_fw->type), uc_fw->path); + drm_printf(p, "\tstatus: %s\n", + intel_uc_fw_status_repr(uc_fw->status)); + drm_printf(p, "\tversion: wanted %u.%u, found %u.%u\n", + uc_fw->major_ver_wanted, uc_fw->minor_ver_wanted, + uc_fw->major_ver_found, uc_fw->minor_ver_found); + drm_printf(p, "\tuCode: %u bytes\n", uc_fw->ucode_size); + drm_printf(p, "\tRSA: %u bytes\n", uc_fw->rsa_size); +} diff --git a/drivers/gpu/drm/i915/gt/uc/intel_uc_fw.h b/drivers/gpu/drm/i915/gt/uc/intel_uc_fw.h new file mode 100644 index 000000000000..7a0a5989afc9 --- /dev/null +++ b/drivers/gpu/drm/i915/gt/uc/intel_uc_fw.h @@ -0,0 +1,241 @@ +/* SPDX-License-Identifier: MIT */ +/* + * Copyright © 2014-2019 Intel Corporation + */ + +#ifndef _INTEL_UC_FW_H_ +#define _INTEL_UC_FW_H_ + +#include <linux/types.h> +#include "intel_uc_fw_abi.h" +#include "intel_device_info.h" +#include "i915_gem.h" + +struct drm_printer; +struct drm_i915_private; +struct intel_gt; + +/* Home of GuC, HuC and DMC firmwares */ +#define INTEL_UC_FIRMWARE_URL "https://git.kernel.org/pub/scm/linux/kernel/git/firmware/linux-firmware.git/tree/i915" + +/* + * +------------+---------------------------------------------------+ + * | PHASE | FIRMWARE STATUS TRANSITIONS | + * +============+===================================================+ + * | | UNINITIALIZED | + * +------------+- / | \ -+ + * | | DISABLED <--/ | \--> NOT_SUPPORTED | + * | init_early | V | + * | | SELECTED | + * +------------+- / | \ -+ + * | | MISSING <--/ | \--> ERROR | + * | fetch | | | + * | | /------> AVAILABLE <---<-----------\ | + * +------------+- \ / \ \ \ -+ + * | | FAIL <--< \--> TRANSFERRED \ | + * | upload | \ / \ / | + * | | \---------/ \--> RUNNING | + * +------------+---------------------------------------------------+ + */ + +enum intel_uc_fw_status { + INTEL_UC_FIRMWARE_NOT_SUPPORTED = -1, /* no uc HW */ + INTEL_UC_FIRMWARE_UNINITIALIZED = 0, /* used to catch checks done too early */ + INTEL_UC_FIRMWARE_DISABLED, /* disabled */ + INTEL_UC_FIRMWARE_SELECTED, /* selected the blob we want to load */ + INTEL_UC_FIRMWARE_MISSING, /* blob not found on the system */ + INTEL_UC_FIRMWARE_ERROR, /* invalid format or version */ + INTEL_UC_FIRMWARE_AVAILABLE, /* blob found and copied in mem */ + INTEL_UC_FIRMWARE_FAIL, /* failed to xfer or init/auth the fw */ + INTEL_UC_FIRMWARE_TRANSFERRED, /* dma xfer done */ + INTEL_UC_FIRMWARE_RUNNING /* init/auth done */ +}; + +enum intel_uc_fw_type { + INTEL_UC_FW_TYPE_GUC = 0, + INTEL_UC_FW_TYPE_HUC +}; +#define INTEL_UC_FW_NUM_TYPES 2 + +/* + * This structure encapsulates all the data needed during the process + * of fetching, caching, and loading the firmware image into the uC. + */ +struct intel_uc_fw { + enum intel_uc_fw_type type; + union { + const enum intel_uc_fw_status status; + enum intel_uc_fw_status __status; /* no accidental overwrites */ + }; + const char *path; + bool user_overridden; + size_t size; + struct drm_i915_gem_object *obj; + + /* + * The firmware build process will generate a version header file with major and + * minor version defined. The versions are built into CSS header of firmware. + * i915 kernel driver set the minimal firmware version required per platform. + */ + u16 major_ver_wanted; + u16 minor_ver_wanted; + u16 major_ver_found; + u16 minor_ver_found; + + u32 rsa_size; + u32 ucode_size; +}; + +#ifdef CONFIG_DRM_I915_DEBUG_GUC +void intel_uc_fw_change_status(struct intel_uc_fw *uc_fw, + enum intel_uc_fw_status status); +#else +static inline void intel_uc_fw_change_status(struct intel_uc_fw *uc_fw, + enum intel_uc_fw_status status) +{ + uc_fw->__status = status; +} +#endif + +static inline +const char *intel_uc_fw_status_repr(enum intel_uc_fw_status status) +{ + switch (status) { + case INTEL_UC_FIRMWARE_NOT_SUPPORTED: + return "N/A"; + case INTEL_UC_FIRMWARE_UNINITIALIZED: + return "UNINITIALIZED"; + case INTEL_UC_FIRMWARE_DISABLED: + return "DISABLED"; + case INTEL_UC_FIRMWARE_SELECTED: + return "SELECTED"; + case INTEL_UC_FIRMWARE_MISSING: + return "MISSING"; + case INTEL_UC_FIRMWARE_ERROR: + return "ERROR"; + case INTEL_UC_FIRMWARE_AVAILABLE: + return "AVAILABLE"; + case INTEL_UC_FIRMWARE_FAIL: + return "FAIL"; + case INTEL_UC_FIRMWARE_TRANSFERRED: + return "TRANSFERRED"; + case INTEL_UC_FIRMWARE_RUNNING: + return "RUNNING"; + } + return "<invalid>"; +} + +static inline int intel_uc_fw_status_to_error(enum intel_uc_fw_status status) +{ + switch (status) { + case INTEL_UC_FIRMWARE_NOT_SUPPORTED: + return -ENODEV; + case INTEL_UC_FIRMWARE_UNINITIALIZED: + return -EACCES; + case INTEL_UC_FIRMWARE_DISABLED: + return -EPERM; + case INTEL_UC_FIRMWARE_MISSING: + return -ENOENT; + case INTEL_UC_FIRMWARE_ERROR: + return -ENOEXEC; + case INTEL_UC_FIRMWARE_FAIL: + return -EIO; + case INTEL_UC_FIRMWARE_SELECTED: + return -ESTALE; + case INTEL_UC_FIRMWARE_AVAILABLE: + case INTEL_UC_FIRMWARE_TRANSFERRED: + case INTEL_UC_FIRMWARE_RUNNING: + return 0; + } + return -EINVAL; +} + +static inline const char *intel_uc_fw_type_repr(enum intel_uc_fw_type type) +{ + switch (type) { + case INTEL_UC_FW_TYPE_GUC: + return "GuC"; + case INTEL_UC_FW_TYPE_HUC: + return "HuC"; + } + return "uC"; +} + +static inline enum intel_uc_fw_status +__intel_uc_fw_status(struct intel_uc_fw *uc_fw) +{ + /* shouldn't call this before checking hw/blob availability */ + GEM_BUG_ON(uc_fw->status == INTEL_UC_FIRMWARE_UNINITIALIZED); + return uc_fw->status; +} + +static inline bool intel_uc_fw_is_supported(struct intel_uc_fw *uc_fw) +{ + return __intel_uc_fw_status(uc_fw) != INTEL_UC_FIRMWARE_NOT_SUPPORTED; +} + +static inline bool intel_uc_fw_is_enabled(struct intel_uc_fw *uc_fw) +{ + return __intel_uc_fw_status(uc_fw) > INTEL_UC_FIRMWARE_DISABLED; +} + +static inline bool intel_uc_fw_is_available(struct intel_uc_fw *uc_fw) +{ + return __intel_uc_fw_status(uc_fw) >= INTEL_UC_FIRMWARE_AVAILABLE; +} + +static inline bool intel_uc_fw_is_loaded(struct intel_uc_fw *uc_fw) +{ + return __intel_uc_fw_status(uc_fw) >= INTEL_UC_FIRMWARE_TRANSFERRED; +} + +static inline bool intel_uc_fw_is_running(struct intel_uc_fw *uc_fw) +{ + return __intel_uc_fw_status(uc_fw) == INTEL_UC_FIRMWARE_RUNNING; +} + +static inline bool intel_uc_fw_is_overridden(const struct intel_uc_fw *uc_fw) +{ + return uc_fw->user_overridden; +} + +static inline void intel_uc_fw_sanitize(struct intel_uc_fw *uc_fw) +{ + if (intel_uc_fw_is_loaded(uc_fw)) + intel_uc_fw_change_status(uc_fw, INTEL_UC_FIRMWARE_AVAILABLE); +} + +static inline u32 __intel_uc_fw_get_upload_size(struct intel_uc_fw *uc_fw) +{ + return sizeof(struct uc_css_header) + uc_fw->ucode_size; +} + +/** + * intel_uc_fw_get_upload_size() - Get size of firmware needed to be uploaded. + * @uc_fw: uC firmware. + * + * Get the size of the firmware and header that will be uploaded to WOPCM. + * + * Return: Upload firmware size, or zero on firmware fetch failure. + */ +static inline u32 intel_uc_fw_get_upload_size(struct intel_uc_fw *uc_fw) +{ + if (!intel_uc_fw_is_available(uc_fw)) + return 0; + + return __intel_uc_fw_get_upload_size(uc_fw); +} + +void intel_uc_fw_init_early(struct intel_uc_fw *uc_fw, + enum intel_uc_fw_type type, bool supported, + enum intel_platform platform, u8 rev); +int intel_uc_fw_fetch(struct intel_uc_fw *uc_fw, struct drm_i915_private *i915); +void intel_uc_fw_cleanup_fetch(struct intel_uc_fw *uc_fw); +int intel_uc_fw_upload(struct intel_uc_fw *uc_fw, struct intel_gt *gt, + u32 wopcm_offset, u32 dma_flags); +int intel_uc_fw_init(struct intel_uc_fw *uc_fw); +void intel_uc_fw_fini(struct intel_uc_fw *uc_fw); +size_t intel_uc_fw_copy_rsa(struct intel_uc_fw *uc_fw, void *dst, u32 max_len); +void intel_uc_fw_dump(const struct intel_uc_fw *uc_fw, struct drm_printer *p); + +#endif diff --git a/drivers/gpu/drm/i915/gt/uc/intel_uc_fw_abi.h b/drivers/gpu/drm/i915/gt/uc/intel_uc_fw_abi.h new file mode 100644 index 000000000000..ae58e8a8c53b --- /dev/null +++ b/drivers/gpu/drm/i915/gt/uc/intel_uc_fw_abi.h @@ -0,0 +1,82 @@ +/* SPDX-License-Identifier: MIT */ +/* + * Copyright © 2019 Intel Corporation + */ + +#ifndef _INTEL_UC_FW_ABI_H +#define _INTEL_UC_FW_ABI_H + +#include <linux/types.h> +#include <linux/build_bug.h> + +/** + * DOC: Firmware Layout + * + * The GuC/HuC firmware layout looks like this:: + * + * +======================================================================+ + * | Firmware blob | + * +===============+===============+============+============+============+ + * | CSS header | uCode | RSA key | modulus | exponent | + * +===============+===============+============+============+============+ + * <-header size-> <---header size continued -----------> + * <--- size -----------------------------------------------------------> + * <-key size-> + * <-mod size-> + * <-exp size-> + * + * The firmware may or may not have modulus key and exponent data. The header, + * uCode and RSA signature are must-have components that will be used by driver. + * Length of each components, which is all in dwords, can be found in header. + * In the case that modulus and exponent are not present in fw, a.k.a truncated + * image, the length value still appears in header. + * + * Driver will do some basic fw size validation based on the following rules: + * + * 1. Header, uCode and RSA are must-have components. + * 2. All firmware components, if they present, are in the sequence illustrated + * in the layout table above. + * 3. Length info of each component can be found in header, in dwords. + * 4. Modulus and exponent key are not required by driver. They may not appear + * in fw. So driver will load a truncated firmware in this case. + * + * The only difference between GuC and HuC firmwares is how the version + * information is saved. + */ + +struct uc_css_header { + u32 module_type; + /* + * header_size includes all non-uCode bits, including css_header, rsa + * key, modulus key and exponent data. + */ + u32 header_size_dw; + u32 header_version; + u32 module_id; + u32 module_vendor; + u32 date; +#define CSS_DATE_DAY (0xFF << 0) +#define CSS_DATE_MONTH (0xFF << 8) +#define CSS_DATE_YEAR (0xFFFF << 16) + u32 size_dw; /* uCode plus header_size_dw */ + u32 key_size_dw; + u32 modulus_size_dw; + u32 exponent_size_dw; + u32 time; +#define CSS_TIME_HOUR (0xFF << 0) +#define CSS_DATE_MIN (0xFF << 8) +#define CSS_DATE_SEC (0xFFFF << 16) + char username[8]; + char buildnumber[12]; + u32 sw_version; +#define CSS_SW_VERSION_GUC_MAJOR (0xFF << 16) +#define CSS_SW_VERSION_GUC_MINOR (0xFF << 8) +#define CSS_SW_VERSION_GUC_PATCH (0xFF << 0) +#define CSS_SW_VERSION_HUC_MAJOR (0xFFFF << 16) +#define CSS_SW_VERSION_HUC_MINOR (0xFFFF << 0) + u32 reserved[14]; + u32 header_info; +} __packed; +static_assert(sizeof(struct uc_css_header) == 128); + +#endif /* _INTEL_UC_FW_ABI_H */ diff --git a/drivers/gpu/drm/i915/gt/uc/selftest_guc.c b/drivers/gpu/drm/i915/gt/uc/selftest_guc.c new file mode 100644 index 000000000000..bba0eafe1cdb --- /dev/null +++ b/drivers/gpu/drm/i915/gt/uc/selftest_guc.c @@ -0,0 +1,317 @@ +// SPDX-License-Identifier: MIT +/* + * Copyright © 2017 Intel Corporation + */ + +#include "i915_selftest.h" +#include "gem/i915_gem_pm.h" + +/* max doorbell number + negative test for each client type */ +#define ATTEMPTS (GUC_NUM_DOORBELLS + GUC_CLIENT_PRIORITY_NUM) + +static struct intel_guc_client *clients[ATTEMPTS]; + +static bool available_dbs(struct intel_guc *guc, u32 priority) +{ + unsigned long offset; + unsigned long end; + u16 id; + + /* first half is used for normal priority, second half for high */ + offset = 0; + end = GUC_NUM_DOORBELLS / 2; + if (priority <= GUC_CLIENT_PRIORITY_HIGH) { + offset = end; + end += offset; + } + + id = find_next_zero_bit(guc->doorbell_bitmap, end, offset); + if (id < end) + return true; + + return false; +} + +static int check_all_doorbells(struct intel_guc *guc) +{ + u16 db_id; + + pr_info_once("Max number of doorbells: %d", GUC_NUM_DOORBELLS); + for (db_id = 0; db_id < GUC_NUM_DOORBELLS; ++db_id) { + if (!doorbell_ok(guc, db_id)) { + pr_err("doorbell %d, not ok\n", db_id); + return -EIO; + } + } + + return 0; +} + +static int ring_doorbell_nop(struct intel_guc_client *client) +{ + struct guc_process_desc *desc = __get_process_desc(client); + int err; + + client->use_nop_wqi = true; + + spin_lock_irq(&client->wq_lock); + + guc_wq_item_append(client, 0, 0, 0, 0); + guc_ring_doorbell(client); + + spin_unlock_irq(&client->wq_lock); + + client->use_nop_wqi = false; + + /* if there are no issues GuC will update the WQ head and keep the + * WQ in active status + */ + err = wait_for(READ_ONCE(desc->head) == READ_ONCE(desc->tail), 10); + if (err) { + pr_err("doorbell %u ring failed!\n", client->doorbell_id); + return -EIO; + } + + if (desc->wq_status != WQ_STATUS_ACTIVE) { + pr_err("doorbell %u ring put WQ in bad state (%u)!\n", + client->doorbell_id, desc->wq_status); + return -EIO; + } + + return 0; +} + +/* + * Basic client sanity check, handy to validate create_clients. + */ +static int validate_client(struct intel_guc_client *client, int client_priority) +{ + if (client->priority != client_priority || + client->doorbell_id == GUC_DOORBELL_INVALID) + return -EINVAL; + else + return 0; +} + +static bool client_doorbell_in_sync(struct intel_guc_client *client) +{ + return !client || doorbell_ok(client->guc, client->doorbell_id); +} + +/* + * Check that we're able to synchronize guc_clients with their doorbells + * + * We're creating clients and reserving doorbells once, at module load. During + * module lifetime, GuC, doorbell HW, and i915 state may go out of sync due to + * GuC being reset. In other words - GuC clients are still around, but the + * status of their doorbells may be incorrect. This is the reason behind + * validating that the doorbells status expected by the driver matches what the + * GuC/HW have. + */ +static int igt_guc_clients(void *args) +{ + struct drm_i915_private *dev_priv = args; + intel_wakeref_t wakeref; + struct intel_guc *guc; + int err = 0; + + GEM_BUG_ON(!HAS_GT_UC(dev_priv)); + mutex_lock(&dev_priv->drm.struct_mutex); + wakeref = intel_runtime_pm_get(&dev_priv->runtime_pm); + + guc = &dev_priv->gt.uc.guc; + if (!guc) { + pr_err("No guc object!\n"); + err = -EINVAL; + goto unlock; + } + + err = check_all_doorbells(guc); + if (err) + goto unlock; + + /* + * Get rid of clients created during driver load because the test will + * recreate them. + */ + guc_clients_disable(guc); + guc_clients_destroy(guc); + if (guc->execbuf_client) { + pr_err("guc_clients_destroy lied!\n"); + err = -EINVAL; + goto unlock; + } + + err = guc_clients_create(guc); + if (err) { + pr_err("Failed to create clients\n"); + goto unlock; + } + GEM_BUG_ON(!guc->execbuf_client); + + err = validate_client(guc->execbuf_client, + GUC_CLIENT_PRIORITY_KMD_NORMAL); + if (err) { + pr_err("execbug client validation failed\n"); + goto out; + } + + /* the client should now have reserved a doorbell */ + if (!has_doorbell(guc->execbuf_client)) { + pr_err("guc_clients_create didn't reserve doorbells\n"); + err = -EINVAL; + goto out; + } + + /* Now enable the clients */ + guc_clients_enable(guc); + + /* each client should now have received a doorbell */ + if (!client_doorbell_in_sync(guc->execbuf_client)) { + pr_err("failed to initialize the doorbells\n"); + err = -EINVAL; + goto out; + } + + /* + * Basic test - an attempt to reallocate a valid doorbell to the + * client it is currently assigned should not cause a failure. + */ + err = create_doorbell(guc->execbuf_client); + +out: + /* + * Leave clean state for other test, plus the driver always destroy the + * clients during unload. + */ + guc_clients_disable(guc); + guc_clients_destroy(guc); + guc_clients_create(guc); + guc_clients_enable(guc); +unlock: + intel_runtime_pm_put(&dev_priv->runtime_pm, wakeref); + mutex_unlock(&dev_priv->drm.struct_mutex); + return err; +} + +/* + * Create as many clients as number of doorbells. Note that there's already + * client(s)/doorbell(s) created during driver load, but this test creates + * its own and do not interact with the existing ones. + */ +static int igt_guc_doorbells(void *arg) +{ + struct drm_i915_private *dev_priv = arg; + intel_wakeref_t wakeref; + struct intel_guc *guc; + int i, err = 0; + u16 db_id; + + GEM_BUG_ON(!HAS_GT_UC(dev_priv)); + mutex_lock(&dev_priv->drm.struct_mutex); + wakeref = intel_runtime_pm_get(&dev_priv->runtime_pm); + + guc = &dev_priv->gt.uc.guc; + if (!guc) { + pr_err("No guc object!\n"); + err = -EINVAL; + goto unlock; + } + + err = check_all_doorbells(guc); + if (err) + goto unlock; + + for (i = 0; i < ATTEMPTS; i++) { + clients[i] = guc_client_alloc(guc, i % GUC_CLIENT_PRIORITY_NUM); + + if (!clients[i]) { + pr_err("[%d] No guc client\n", i); + err = -EINVAL; + goto out; + } + + if (IS_ERR(clients[i])) { + if (PTR_ERR(clients[i]) != -ENOSPC) { + pr_err("[%d] unexpected error\n", i); + err = PTR_ERR(clients[i]); + goto out; + } + + if (available_dbs(guc, i % GUC_CLIENT_PRIORITY_NUM)) { + pr_err("[%d] non-db related alloc fail\n", i); + err = -EINVAL; + goto out; + } + + /* expected, ran out of dbs for this client type */ + continue; + } + + /* + * The check below is only valid because we keep a doorbell + * assigned during the whole life of the client. + */ + if (clients[i]->stage_id >= GUC_NUM_DOORBELLS) { + pr_err("[%d] more clients than doorbells (%d >= %d)\n", + i, clients[i]->stage_id, GUC_NUM_DOORBELLS); + err = -EINVAL; + goto out; + } + + err = validate_client(clients[i], i % GUC_CLIENT_PRIORITY_NUM); + if (err) { + pr_err("[%d] client_alloc sanity check failed!\n", i); + err = -EINVAL; + goto out; + } + + db_id = clients[i]->doorbell_id; + + err = __guc_client_enable(clients[i]); + if (err) { + pr_err("[%d] Failed to create a doorbell\n", i); + goto out; + } + + /* doorbell id shouldn't change, we are holding the mutex */ + if (db_id != clients[i]->doorbell_id) { + pr_err("[%d] doorbell id changed (%d != %d)\n", + i, db_id, clients[i]->doorbell_id); + err = -EINVAL; + goto out; + } + + err = check_all_doorbells(guc); + if (err) + goto out; + + err = ring_doorbell_nop(clients[i]); + if (err) + goto out; + } + +out: + for (i = 0; i < ATTEMPTS; i++) + if (!IS_ERR_OR_NULL(clients[i])) { + __guc_client_disable(clients[i]); + guc_client_free(clients[i]); + } +unlock: + intel_runtime_pm_put(&dev_priv->runtime_pm, wakeref); + mutex_unlock(&dev_priv->drm.struct_mutex); + return err; +} + +int intel_guc_live_selftest(struct drm_i915_private *dev_priv) +{ + static const struct i915_subtest tests[] = { + SUBTEST(igt_guc_clients), + SUBTEST(igt_guc_doorbells), + }; + + if (!USES_GUC_SUBMISSION(dev_priv)) + return 0; + + return i915_subtests(tests, dev_priv); +} |