Merge tag 'amd-drm-next-6.5-2023-06-16' of https://gitlab.freedesktop.org/agd5f/linux into drm-next

amd-drm-next-6.5-2023-06-16:

amdgpu:
- Misc display fixes
- W=1 fixes
- Improve scheduler naming
- DCN 3.1.4 fixes
- kdoc fixes
- Enable W=1
- VCN 4.0 fix
- xgmi fixes
- TOPDOWN fix for large BAR systems
- eDP fix
- PSR fixes
- SubVP fixes
- Freesync fix
- DPIA fix
- SMU 13.0.5 fixes
- vblflash fix
- RAS fixes
- SDMA 4 fix
- BO locking fix
- BO backing store fix
- NBIO 7.9 fixes
- GC 9.4.3 fixes
- GPU reset recovery fixes
- HMM fix

amdkfd:
- Fix NULL check
- Trap fixes
- Queue count fix
- Add event age tracking

radeon:
- fbdev client fix

scheduler:
- Avoid an infinite loop

UAPI:
- Add KFD event age tracking:
  Proposed ROCT-Thunk-Interface:
  ROCm/ROCT-Thunk-Interface@efdbf6c
  ROCm/ROCT-Thunk-Interface@1820ae0
  Proposed ROCR-Runtime:
  ROCm/ROCR-Runtime@master...zhums:ROCR-Runtime:new_event_wait_review
  ROCm/ROCR-Runtime@e1f5bdb
  ROCm/ROCR-Runtime@7d26afd

drm:
- DP MST fix

Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Alex Deucher <alexander.deucher@amd.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20230616163548.7706-1-alexander.deucher@amd.com

Author: airlied

drivers/gpu/drm/amd/amdgpu/Makefile

@@ -40,7 +40,18 @@ ccflags-y := -I$(FULL_AMD_PATH)/include/asic_reg \
 	-I$(FULL_AMD_PATH)/amdkfd
 
 subdir-ccflags-y := -Wextra
-subdir-ccflags-y += $(call cc-option, -Wunused-but-set-variable)
+subdir-ccflags-y += -Wunused
+subdir-ccflags-y += -Wmissing-prototypes
+subdir-ccflags-y += -Wmissing-declarations
+subdir-ccflags-y += -Wmissing-include-dirs
+subdir-ccflags-y += -Wold-style-definition
+subdir-ccflags-y += -Wmissing-format-attribute
+# Need this to avoid recursive variable evaluation issues
+cond-flags := $(call cc-option, -Wunused-but-set-variable) \
+	$(call cc-option, -Wunused-const-variable) \
+	$(call cc-option, -Wstringop-truncation) \
+	$(call cc-option, -Wpacked-not-aligned)
+subdir-ccflags-y += $(cond-flags)
 subdir-ccflags-y += -Wno-unused-parameter
 subdir-ccflags-y += -Wno-type-limits
 subdir-ccflags-y += -Wno-sign-compare

drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c

@@ -309,7 +309,7 @@ static int amdgpu_cs_pass1(struct amdgpu_cs_parser *p,
 	}
 	p->gang_leader = p->jobs[p->gang_leader_idx];
 
-	if (p->ctx->vram_lost_counter != p->gang_leader->vram_lost_counter) {
+	if (p->ctx->generation != p->gang_leader->generation) {
 		ret = -ECANCELED;
 		goto free_all_kdata;
 	}

drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c

@@ -333,7 +333,7 @@ static int amdgpu_ctx_init(struct amdgpu_ctx_mgr *mgr, int32_t priority,
 
 	ctx->reset_counter = atomic_read(&mgr->adev->gpu_reset_counter);
 	ctx->reset_counter_query = ctx->reset_counter;
-	ctx->vram_lost_counter = atomic_read(&mgr->adev->vram_lost_counter);
+	ctx->generation = amdgpu_vm_generation(mgr->adev, &fpriv->vm);
 	ctx->init_priority = priority;
 	ctx->override_priority = AMDGPU_CTX_PRIORITY_UNSET;
 
@@ -432,6 +432,7 @@ int amdgpu_ctx_get_entity(struct amdgpu_ctx *ctx, u32 hw_ip, u32 instance,
 			  u32 ring, struct drm_sched_entity **entity)
 {
 	int r;
+	struct drm_sched_entity *ctx_entity;
 
 	if (hw_ip >= AMDGPU_HW_IP_NUM) {
 		DRM_ERROR("unknown HW IP type: %d\n", hw_ip);
@@ -455,7 +456,14 @@ int amdgpu_ctx_get_entity(struct amdgpu_ctx *ctx, u32 hw_ip, u32 instance,
 			return r;
 	}
 
-	*entity = &ctx->entities[hw_ip][ring]->entity;
+	ctx_entity = &ctx->entities[hw_ip][ring]->entity;
+	r = drm_sched_entity_error(ctx_entity);
+	if (r) {
+		DRM_DEBUG("error entity %p\n", ctx_entity);
+		return r;
+	}
+
+	*entity = ctx_entity;
 	return 0;
 }
 
@@ -586,7 +594,7 @@ static int amdgpu_ctx_query2(struct amdgpu_device *adev,
 	if (ctx->reset_counter != atomic_read(&adev->gpu_reset_counter))
 		out->state.flags |= AMDGPU_CTX_QUERY2_FLAGS_RESET;
 
-	if (ctx->vram_lost_counter != atomic_read(&adev->vram_lost_counter))
+	if (ctx->generation != amdgpu_vm_generation(adev, &fpriv->vm))
 		out->state.flags |= AMDGPU_CTX_QUERY2_FLAGS_VRAMLOST;
 
 	if (atomic_read(&ctx->guilty))

drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.h

@@ -47,7 +47,7 @@ struct amdgpu_ctx {
 	struct amdgpu_ctx_mgr		*mgr;
 	unsigned			reset_counter;
 	unsigned			reset_counter_query;
-	uint32_t			vram_lost_counter;
+	uint64_t			generation;
 	spinlock_t			ring_lock;
 	struct amdgpu_ctx_entity	*entities[AMDGPU_HW_IP_NUM][AMDGPU_MAX_ENTITY_NUM];
 	bool				preamble_presented;

drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c

@@ -1515,6 +1515,7 @@ static int amdgpu_discovery_get_mall_info(struct amdgpu_device *adev)
 				mall_size += mall_size_per_umc;
 		}
 		adev->gmc.mall_size = mall_size;
+		adev->gmc.m_half_use = half_use;
 		break;
 	default:
 		dev_err(adev->dev,
@@ -1896,6 +1897,8 @@ static int amdgpu_discovery_set_gc_ip_blocks(struct amdgpu_device *adev)
 		amdgpu_device_ip_block_add(adev, &gfx_v9_0_ip_block);
 		break;
 	case IP_VERSION(9, 4, 3):
+		if (!amdgpu_exp_hw_support)
+			return -EINVAL;
 		amdgpu_device_ip_block_add(adev, &gfx_v9_4_3_ip_block);
 		break;
 	case IP_VERSION(10, 1, 10):

drivers/gpu/drm/amd/amdgpu/amdgpu_fence.c

@@ -691,6 +691,30 @@ void amdgpu_fence_driver_clear_job_fences(struct amdgpu_ring *ring)
 	}
 }
 
+/**
+ * amdgpu_fence_driver_set_error - set error code on fences
+ * @ring: the ring which contains the fences
+ * @error: the error code to set
+ *
+ * Set an error code to all the fences pending on the ring.
+ */
+void amdgpu_fence_driver_set_error(struct amdgpu_ring *ring, int error)
+{
+	struct amdgpu_fence_driver *drv = &ring->fence_drv;
+	unsigned long flags;
+
+	spin_lock_irqsave(&drv->lock, flags);
+	for (unsigned int i = 0; i <= drv->num_fences_mask; ++i) {
+		struct dma_fence *fence;
+
+		fence = rcu_dereference_protected(drv->fences[i],
+						  lockdep_is_held(&drv->lock));
+		if (fence && !dma_fence_is_signaled_locked(fence))
+			dma_fence_set_error(fence, error);
+	}
+	spin_unlock_irqrestore(&drv->lock, flags);
+}
+
 /**
  * amdgpu_fence_driver_force_completion - force signal latest fence of ring
  *
@@ -699,6 +723,7 @@ void amdgpu_fence_driver_clear_job_fences(struct amdgpu_ring *ring)
  */
 void amdgpu_fence_driver_force_completion(struct amdgpu_ring *ring)
 {
+	amdgpu_fence_driver_set_error(ring, -ECANCELED);
 	amdgpu_fence_write(ring, ring->fence_drv.sync_seq);
 	amdgpu_fence_process(ring);
 }

drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.h

@@ -301,6 +301,8 @@ struct amdgpu_gmc {
 
 	/* MALL size */
 	u64 mall_size;
+	uint32_t m_half_use;
+
 	/* number of UMC instances */
 	int num_umc;
 	/* mode2 save restore */

drivers/gpu/drm/amd/amdgpu/amdgpu_hmm.c

@@ -190,8 +190,8 @@ int amdgpu_hmm_range_get_pages(struct mmu_interval_notifier *notifier,
 		pr_debug("hmm range: start = 0x%lx, end = 0x%lx",
 			hmm_range->start, hmm_range->end);
 
-		/* Assuming 512MB takes maxmium 1 second to fault page address */
-		timeout = max((hmm_range->end - hmm_range->start) >> 29, 1UL);
+		/* Assuming 128MB takes maximum 1 second to fault page address */
+		timeout = max((hmm_range->end - hmm_range->start) >> 27, 1UL);
 		timeout *= HMM_RANGE_DEFAULT_TIMEOUT;
 		timeout = jiffies + msecs_to_jiffies(timeout);
 

drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c

@@ -467,7 +467,8 @@ void amdgpu_irq_dispatch(struct amdgpu_device *adev,
 			handled = true;
 
 	} else {
-		DRM_DEBUG("Unhandled interrupt src_id: %d\n", src_id);
+		DRM_DEBUG("Unregistered interrupt src_id: %d of client_id:%d\n",
+			src_id, client_id);
 	}
 
 	/* Send it to amdkfd as well if it isn't already handled */

drivers/gpu/drm/amd/amdgpu/amdgpu_job.c

@@ -109,7 +109,7 @@ int amdgpu_job_alloc(struct amdgpu_device *adev, struct amdgpu_vm *vm,
 	(*job)->vm = vm;
 
 	amdgpu_sync_create(&(*job)->explicit_sync);
-	(*job)->vram_lost_counter = atomic_read(&adev->vram_lost_counter);
+	(*job)->generation = amdgpu_vm_generation(adev, vm);
 	(*job)->vm_pd_addr = AMDGPU_BO_INVALID_OFFSET;
 
 	if (!entity)
@@ -258,16 +258,27 @@ amdgpu_job_prepare_job(struct drm_sched_job *sched_job,
 	struct dma_fence *fence = NULL;
 	int r;
 
+	/* Ignore soft recovered fences here */
+	r = drm_sched_entity_error(s_entity);
+	if (r && r != -ENODATA)
+		goto error;
+
 	if (!fence && job->gang_submit)
 		fence = amdgpu_device_switch_gang(ring->adev, job->gang_submit);
 
 	while (!fence && job->vm && !job->vmid) {
 		r = amdgpu_vmid_grab(job->vm, ring, job, &fence);
-		if (r)
+		if (r) {
 			DRM_ERROR("Error getting VM ID (%d)\n", r);
+			goto error;
+		}
 	}
 
 	return fence;
+
+error:
+	dma_fence_set_error(&job->base.s_fence->finished, r);
+	return NULL;
 }
 
 static struct dma_fence *amdgpu_job_run(struct drm_sched_job *sched_job)
@@ -284,7 +295,7 @@ static struct dma_fence *amdgpu_job_run(struct drm_sched_job *sched_job)
 	trace_amdgpu_sched_run_job(job);
 
 	/* Skip job if VRAM is lost and never resubmit gangs */
-	if (job->vram_lost_counter != atomic_read(&adev->vram_lost_counter) ||
+	if (job->generation != amdgpu_vm_generation(adev, job->vm) ||
 	    (job->job_run_counter && job->gang_submit))
 		dma_fence_set_error(finished, -ECANCELED);