Merge tag 'amd-drm-fixes-6.0-2022-08-17' of https://gitlab.freedeskto…

…p.org/agd5f/linux into drm-fixes amd-drm-fixes-6.0-2022-08-17: amdgpu: - Revert some DML stack changes - Rounding fixes in KFD allocations - atombios vram info table parsing fix - DCN 3.1.4 fixes - Clockgating fixes for various new IPs - SMU 13.0.4 fixes - DCN 3.1.4 FP fixes - TMDS fixes for YCbCr420 4k modes - DCN 3.2.x fixes - USB 4 fixes - SMU 13.0 fixes - SMU driver unload memory leak fixes - Display orientation fix - Regression fix for generic fbdev conversion - SDMA 6.x fixes - SR-IOV fixes - IH 6.x fixes - Use after free fix in bo list handling - Revert pipe1 support - XGMI hive reset fix amdkfd: - Fix potential crach in kfd_create_indirect_link_prop() Signed-off-by: Dave Airlie <airlied@redhat.com> From: Alex Deucher <alexander.deucher@amd.com> Link: https://patchwork.freedesktop.org/patch/msgid/20220818025206.6463-1-alexander.deucher@amd.com
mariux64 · Aug 18, 2022 · b1fb6b8 · b1fb6b8
2 parents 2ae6ab9 + 085292c
commit b1fb6b8
Showing 108 changed files with 1,758 additions and 1,184 deletions.
diff --git a/drivers/gpu/drm/amd/amdgpu/aldebaran.c b/drivers/gpu/drm/amd/amdgpu/aldebaran.c
@@ -148,38 +148,30 @@ aldebaran_mode2_perform_reset(struct amdgpu_reset_control *reset_ctl,
 			      struct amdgpu_reset_context *reset_context)
 {
 	struct amdgpu_device *adev = (struct amdgpu_device *)reset_ctl->handle;
+	struct list_head *reset_device_list = reset_context->reset_device_list;
 	struct amdgpu_device *tmp_adev = NULL;
-	struct list_head reset_device_list;
 	int r = 0;
 
 	dev_dbg(adev->dev, "aldebaran perform hw reset\n");
+
+	if (reset_device_list == NULL)
+		return -EINVAL;
+
 	if (adev->ip_versions[MP1_HWIP][0] == IP_VERSION(13, 0, 2) &&
 	    reset_context->hive == NULL) {
 		/* Wrong context, return error */
 		return -EINVAL;
 	}
 
-	INIT_LIST_HEAD(&reset_device_list);
-	if (reset_context->hive) {
-		list_for_each_entry (tmp_adev,
-				     &reset_context->hive->device_list,
-				     gmc.xgmi.head)
-			list_add_tail(&tmp_adev->reset_list,
-				      &reset_device_list);
-	} else {
-		list_add_tail(&reset_context->reset_req_dev->reset_list,
-			      &reset_device_list);
-	}
-
-	list_for_each_entry (tmp_adev, &reset_device_list, reset_list) {
+	list_for_each_entry(tmp_adev, reset_device_list, reset_list) {
 		mutex_lock(&tmp_adev->reset_cntl->reset_lock);
 		tmp_adev->reset_cntl->active_reset = AMD_RESET_METHOD_MODE2;
 	}
 	/*
 	 * Mode2 reset doesn't need any sync between nodes in XGMI hive, instead launch
 	 * them together so that they can be completed asynchronously on multiple nodes
 	 */
-	list_for_each_entry (tmp_adev, &reset_device_list, reset_list) {
+	list_for_each_entry(tmp_adev, reset_device_list, reset_list) {
 		/* For XGMI run all resets in parallel to speed up the process */
 		if (tmp_adev->gmc.xgmi.num_physical_nodes > 1) {
 			if (!queue_work(system_unbound_wq,
@@ -197,7 +189,7 @@ aldebaran_mode2_perform_reset(struct amdgpu_reset_control *reset_ctl,
 
 	/* For XGMI wait for all resets to complete before proceed */
 	if (!r) {
-		list_for_each_entry (tmp_adev, &reset_device_list, reset_list) {
+		list_for_each_entry(tmp_adev, reset_device_list, reset_list) {
 			if (tmp_adev->gmc.xgmi.num_physical_nodes > 1) {
 				flush_work(&tmp_adev->reset_cntl->reset_work);
 				r = tmp_adev->asic_reset_res;
@@ -207,7 +199,7 @@ aldebaran_mode2_perform_reset(struct amdgpu_reset_control *reset_ctl,
 		}
 	}
 
-	list_for_each_entry (tmp_adev, &reset_device_list, reset_list) {
+	list_for_each_entry(tmp_adev, reset_device_list, reset_list) {
 		mutex_unlock(&tmp_adev->reset_cntl->reset_lock);
 		tmp_adev->reset_cntl->active_reset = AMD_RESET_METHOD_NONE;
 	}
@@ -339,30 +331,21 @@ static int
 aldebaran_mode2_restore_hwcontext(struct amdgpu_reset_control *reset_ctl,
 				  struct amdgpu_reset_context *reset_context)
 {
+	struct list_head *reset_device_list = reset_context->reset_device_list;
 	struct amdgpu_device *tmp_adev = NULL;
-	struct list_head reset_device_list;
 	int r;
 
+	if (reset_device_list == NULL)
+		return -EINVAL;
+
 	if (reset_context->reset_req_dev->ip_versions[MP1_HWIP][0] ==
 		    IP_VERSION(13, 0, 2) &&
 	    reset_context->hive == NULL) {
 		/* Wrong context, return error */
 		return -EINVAL;
 	}
 
-	INIT_LIST_HEAD(&reset_device_list);
-	if (reset_context->hive) {
-		list_for_each_entry (tmp_adev,
-				     &reset_context->hive->device_list,
-				     gmc.xgmi.head)
-			list_add_tail(&tmp_adev->reset_list,
-				      &reset_device_list);
-	} else {
-		list_add_tail(&reset_context->reset_req_dev->reset_list,
-			      &reset_device_list);
-	}
-
-	list_for_each_entry (tmp_adev, &reset_device_list, reset_list) {
+	list_for_each_entry(tmp_adev, reset_device_list, reset_list) {
 		dev_info(tmp_adev->dev,
 			 "GPU reset succeeded, trying to resume\n");
 		r = aldebaran_mode2_restore_ip(tmp_adev);

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h b/drivers/gpu/drm/amd/amdgpu/amdgpu.h
@@ -317,7 +317,7 @@ enum amdgpu_kiq_irq {
 	AMDGPU_CP_KIQ_IRQ_DRIVER0 = 0,
 	AMDGPU_CP_KIQ_IRQ_LAST
 };
-
+#define SRIOV_USEC_TIMEOUT  1200000 /* wait 12 * 100ms for SRIOV */
 #define MAX_KIQ_REG_WAIT       5000 /* in usecs, 5ms */
 #define MAX_KIQ_REG_BAILOUT_INTERVAL   5 /* in msecs, 5ms */
 #define MAX_KIQ_REG_TRY 1000

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h
@@ -96,6 +96,7 @@ struct amdgpu_amdkfd_fence {
 struct amdgpu_kfd_dev {
 	struct kfd_dev *dev;
 	uint64_t vram_used;
+	uint64_t vram_used_aligned;
 	bool init_complete;
 	struct work_struct reset_work;
 };

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c
@@ -40,10 +40,10 @@
 #define AMDGPU_USERPTR_RESTORE_DELAY_MS 1
 
 /*
- * Align VRAM allocations to 2MB to avoid fragmentation caused by 4K allocations in the tail 2MB
+ * Align VRAM availability to 2MB to avoid fragmentation caused by 4K allocations in the tail 2MB
  * BO chunk
  */
-#define VRAM_ALLOCATION_ALIGN (1 << 21)
+#define VRAM_AVAILABLITY_ALIGN (1 << 21)
 
 /* Impose limit on how much memory KFD can use */
 static struct {
@@ -149,7 +149,7 @@ int amdgpu_amdkfd_reserve_mem_limit(struct amdgpu_device *adev,
 		 * to avoid fragmentation caused by 4K allocations in the tail
 		 * 2M BO chunk.
 		 */
-		vram_needed = ALIGN(size, VRAM_ALLOCATION_ALIGN);
+		vram_needed = size;
 	} else if (alloc_flag & KFD_IOC_ALLOC_MEM_FLAGS_USERPTR) {
 		system_mem_needed = size;
 	} else if (!(alloc_flag &
@@ -182,8 +182,10 @@ int amdgpu_amdkfd_reserve_mem_limit(struct amdgpu_device *adev,
 	 */
 	WARN_ONCE(vram_needed && !adev,
 		  "adev reference can't be null when vram is used");
-	if (adev)
+	if (adev) {
 		adev->kfd.vram_used += vram_needed;
+		adev->kfd.vram_used_aligned += ALIGN(vram_needed, VRAM_AVAILABLITY_ALIGN);
+	}
 	kfd_mem_limit.system_mem_used += system_mem_needed;
 	kfd_mem_limit.ttm_mem_used += ttm_mem_needed;
 
@@ -203,8 +205,10 @@ void amdgpu_amdkfd_unreserve_mem_limit(struct amdgpu_device *adev,
 	} else if (alloc_flag & KFD_IOC_ALLOC_MEM_FLAGS_VRAM) {
 		WARN_ONCE(!adev,
 			  "adev reference can't be null when alloc mem flags vram is set");
-		if (adev)
-			adev->kfd.vram_used -= ALIGN(size, VRAM_ALLOCATION_ALIGN);
+		if (adev) {
+			adev->kfd.vram_used -= size;
+			adev->kfd.vram_used_aligned -= ALIGN(size, VRAM_AVAILABLITY_ALIGN);
+		}
 	} else if (alloc_flag & KFD_IOC_ALLOC_MEM_FLAGS_USERPTR) {
 		kfd_mem_limit.system_mem_used -= size;
 	} else if (!(alloc_flag &
@@ -1608,15 +1612,14 @@ size_t amdgpu_amdkfd_get_available_memory(struct amdgpu_device *adev)
 	uint64_t reserved_for_pt =
 		ESTIMATE_PT_SIZE(amdgpu_amdkfd_total_mem_size);
 	size_t available;
-
 	spin_lock(&kfd_mem_limit.mem_limit_lock);
 	available = adev->gmc.real_vram_size
-		- adev->kfd.vram_used
+		- adev->kfd.vram_used_aligned
 		- atomic64_read(&adev->vram_pin_size)
 		- reserved_for_pt;
 	spin_unlock(&kfd_mem_limit.mem_limit_lock);
 
-	return ALIGN_DOWN(available, VRAM_ALLOCATION_ALIGN);
+	return ALIGN_DOWN(available, VRAM_AVAILABLITY_ALIGN);
 }
 
 int amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu(

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_atomfirmware.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_atomfirmware.c
@@ -314,7 +314,7 @@ amdgpu_atomfirmware_get_vram_info(struct amdgpu_device *adev,
 					mem_channel_number = vram_info->v30.channel_num;
 					mem_channel_width = vram_info->v30.channel_width;
 					if (vram_width)
-						*vram_width = mem_channel_number * mem_channel_width;
+						*vram_width = mem_channel_number * (1 << mem_channel_width);
 					break;
 				default:
 					return -EINVAL;

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c
@@ -837,16 +837,12 @@ static int amdgpu_cs_vm_handling(struct amdgpu_cs_parser *p)
 			continue;
 
 		r = amdgpu_vm_bo_update(adev, bo_va, false);
-		if (r) {
-			mutex_unlock(&p->bo_list->bo_list_mutex);
+		if (r)
 			return r;
-		}
 
 		r = amdgpu_sync_fence(&p->job->sync, bo_va->last_pt_update);
-		if (r) {
-			mutex_unlock(&p->bo_list->bo_list_mutex);
+		if (r)
 			return r;
-		}
 	}
 
 	r = amdgpu_vm_handle_moved(adev, vm);

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c
@@ -1705,7 +1705,7 @@ static ssize_t amdgpu_reset_dump_register_list_write(struct file *f,
 {
 	struct amdgpu_device *adev = (struct amdgpu_device *)file_inode(f)->i_private;
 	char reg_offset[11];
-	uint32_t *new, *tmp = NULL;
+	uint32_t *new = NULL, *tmp = NULL;
 	int ret, i = 0, len = 0;
 
 	do {
@@ -1747,7 +1747,8 @@ static ssize_t amdgpu_reset_dump_register_list_write(struct file *f,
 	ret = size;
 
 error_free:
-	kfree(tmp);
+	if (tmp != new)
+		kfree(tmp);
 	kfree(new);
 	return ret;
 }

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
@@ -4742,6 +4742,8 @@ int amdgpu_do_asic_reset(struct list_head *device_list_handle,
 	tmp_adev = list_first_entry(device_list_handle, struct amdgpu_device,
 				    reset_list);
 	amdgpu_reset_reg_dumps(tmp_adev);
+
+	reset_context->reset_device_list = device_list_handle;
 	r = amdgpu_reset_perform_reset(tmp_adev, reset_context);
 	/* If reset handler not implemented, continue; otherwise return */
 	if (r == -ENOSYS)

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c
@@ -272,10 +272,6 @@ void amdgpu_job_stop_all_jobs_on_sched(struct drm_gpu_scheduler *sched)
 	/* Signal all jobs not yet scheduled */
 	for (i = DRM_SCHED_PRIORITY_COUNT - 1; i >= DRM_SCHED_PRIORITY_MIN; i--) {
 		struct drm_sched_rq *rq = &sched->sched_rq[i];
-
-		if (!rq)
-			continue;
-
 		spin_lock(&rq->lock);
 		list_for_each_entry(s_entity, &rq->entities, list) {
 			while ((s_job = to_drm_sched_job(spsc_queue_pop(&s_entity->job_queue)))) {

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_reset.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_reset.h
@@ -37,6 +37,7 @@ struct amdgpu_reset_context {
 	struct amdgpu_device *reset_req_dev;
 	struct amdgpu_job *job;
 	struct amdgpu_hive_info *hive;
+	struct list_head *reset_device_list;
 	unsigned long flags;
 };