Skip to contentSkip to conversation

drm/scheduler: rework entity creation

RFC PATCH Superseded Nirmoy Das 4 participants 5 messages Dec 5, 2019

Subsystems (8)

v1 of 3 Superseded Latest is v3 →

lore
Share & tools
lore

Nirmoy Das

author
#0 lore raw

RFC PATCH 1/1 drm/scheduler: rework entity creation

Entity currently keeps a copy of run_queue list and modify it in drm_sched_entity_set_priority(). Entities shouldn't modify run_queue list. Use drm_gpu_scheduler list instead of drm_sched_rq list in drm_sched_entity struct. In this way we can select a runqueue based on entity/ctx's priority for a drm scheduler.

---
 drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c  |  7 +--
 drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c  |  7 +--
 drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c  |  7 +--
 drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c  |  7 +--
 drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c   | 14 +++--
 drivers/gpu/drm/etnaviv/etnaviv_drv.c    |  8 +--
 drivers/gpu/drm/lima/lima_sched.c        |  5 +-
 drivers/gpu/drm/panfrost/panfrost_job.c  |  7 +--
 drivers/gpu/drm/scheduler/sched_entity.c | 65 +++++++++---------------
 drivers/gpu/drm/v3d/v3d_drv.c            |  7 +--
 include/drm/gpu_scheduler.h              |  9 ++--
 11 files changed, 69 insertions(+), 74 deletions(-)
drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c +4 -3
index a0d3d7b756eb..e8f46c13d073 100644
@@ -122,7 +122,7 @@ static int amdgpu_ctx_init(struct amdgpu_device *adev,
 
 	for (i = 0; i < AMDGPU_HW_IP_NUM; ++i) {
 		struct amdgpu_ring *rings[AMDGPU_MAX_RINGS];
-		struct drm_sched_rq *rqs[AMDGPU_MAX_RINGS];
+		struct drm_gpu_scheduler *sched_list[AMDGPU_MAX_RINGS];
 		unsigned num_rings = 0;
 		unsigned num_rqs = 0;
 
@@ -181,12 +181,13 @@ static int amdgpu_ctx_init(struct amdgpu_device *adev,
 			if (!rings[j]->adev)
 				continue;
 
-			rqs[num_rqs++] = &rings[j]->sched.sched_rq[priority];
+			sched_list[num_rqs++] = &rings[j]->sched;
 		}
 
 		for (j = 0; j < amdgpu_ctx_num_entities[i]; ++j)
 			r = drm_sched_entity_init(&ctx->entities[i][j].entity,
-						  rqs, num_rqs, &ctx->guilty);
+						  sched_list, num_rqs,
+						  &ctx->guilty, priority);
 		if (r)
 			goto error_cleanup_entities;
 	}
drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c +4 -3
index 19ffe00d9072..a960dd7c0711 100644
@@ -1957,11 +1957,12 @@ void amdgpu_ttm_set_buffer_funcs_status(struct amdgpu_device *adev, bool enable)
 
 	if (enable) {
 		struct amdgpu_ring *ring;
-		struct drm_sched_rq *rq;
+		struct drm_gpu_scheduler *sched;
 
 		ring = adev->mman.buffer_funcs_ring;
-		rq = &ring->sched.sched_rq[DRM_SCHED_PRIORITY_KERNEL];
-		r = drm_sched_entity_init(&adev->mman.entity, &rq, 1, NULL);
+		sched = &ring->sched;
+		r = drm_sched_entity_init(&adev->mman.entity, &sched,
+					  1, NULL, DRM_SCHED_PRIORITY_KERNEL);
 		if (r) {
 			DRM_ERROR("Failed setting up TTM BO move entity (%d)\n",
 				  r);
drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c +4 -3
index e324bfe6c58f..b803a8882864 100644
@@ -330,12 +330,13 @@ int amdgpu_uvd_sw_fini(struct amdgpu_device *adev)
 int amdgpu_uvd_entity_init(struct amdgpu_device *adev)
 {
 	struct amdgpu_ring *ring;
-	struct drm_sched_rq *rq;
+	struct drm_gpu_scheduler *sched;
 	int r;
 
 	ring = &adev->uvd.inst[0].ring;
-	rq = &ring->sched.sched_rq[DRM_SCHED_PRIORITY_NORMAL];
-	r = drm_sched_entity_init(&adev->uvd.entity, &rq, 1, NULL);
+	sched = &ring->sched;
+	r = drm_sched_entity_init(&adev->uvd.entity, &sched,
+				  1, NULL, DRM_SCHED_PRIORITY_NORMAL);
 	if (r) {
 		DRM_ERROR("Failed setting up UVD kernel entity.\n");
 		return r;
drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c +4 -3
index 46b590af2fd2..b44f28d44fb4 100644
@@ -240,12 +240,13 @@ int amdgpu_vce_sw_fini(struct amdgpu_device *adev)
 int amdgpu_vce_entity_init(struct amdgpu_device *adev)
 {
 	struct amdgpu_ring *ring;
-	struct drm_sched_rq *rq;
+	struct drm_gpu_scheduler *sched;
 	int r;
 
 	ring = &adev->vce.ring[0];
-	rq = &ring->sched.sched_rq[DRM_SCHED_PRIORITY_NORMAL];
-	r = drm_sched_entity_init(&adev->vce.entity, &rq, 1, NULL);
+	sched = &ring->sched;
+	r = drm_sched_entity_init(&adev->vce.entity, &sched,
+				  1, NULL, DRM_SCHED_PRIORITY_NORMAL);
 	if (r != 0) {
 		DRM_ERROR("Failed setting up VCE run queue.\n");
 		return r;
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +10 -4
index a94c4faa5af1..ec6141773a92 100644
@@ -2687,6 +2687,7 @@ int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm,
 {
 	struct amdgpu_bo_param bp;
 	struct amdgpu_bo *root;
+        struct drm_gpu_scheduler *sched_list[AMDGPU_MAX_RINGS];
 	int r, i;
 
 	vm->va = RB_ROOT_CACHED;
@@ -2700,14 +2701,19 @@ int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm,
 	spin_lock_init(&vm->invalidated_lock);
 	INIT_LIST_HEAD(&vm->freed);
 
+        for (i = 0; i < adev->vm_manager.vm_pte_num_rqs; i++)
+	       sched_list[i] = adev->vm_manager.vm_pte_rqs[i]->sched;
+
 	/* create scheduler entities for page table updates */
-	r = drm_sched_entity_init(&vm->direct, adev->vm_manager.vm_pte_rqs,
-				  adev->vm_manager.vm_pte_num_rqs, NULL);
+	r = drm_sched_entity_init(&vm->direct, sched_list,
+				  adev->vm_manager.vm_pte_num_rqs,
+				  NULL, DRM_SCHED_PRIORITY_KERNEL);
 	if (r)
 		return r;
 
-	r = drm_sched_entity_init(&vm->delayed, adev->vm_manager.vm_pte_rqs,
-				  adev->vm_manager.vm_pte_num_rqs, NULL);
+	r = drm_sched_entity_init(&vm->delayed, sched_list,
+				  adev->vm_manager.vm_pte_num_rqs,
+				  NULL, DRM_SCHED_PRIORITY_KERNEL);
 	if (r)
 		goto error_free_direct;
 
drivers/gpu/drm/etnaviv/etnaviv_drv.c +4 -4
index 1f9c01be40d7..a65c1e115e35 100644
@@ -65,12 +65,12 @@ static int etnaviv_open(struct drm_device *dev, struct drm_file *file)
 
 	for (i = 0; i < ETNA_MAX_PIPES; i++) {
 		struct etnaviv_gpu *gpu = priv->gpu[i];
-		struct drm_sched_rq *rq;
+		struct drm_gpu_scheduler *sched;
 
 		if (gpu) {
-			rq = &gpu->sched.sched_rq[DRM_SCHED_PRIORITY_NORMAL];
-			drm_sched_entity_init(&ctx->sched_entity[i],
-					      &rq, 1, NULL);
+			sched = &gpu->sched;
+			drm_sched_entity_init(&ctx->sched_entity[i], &sched,
+					      1, NULL, DRM_SCHED_PRIORITY_NORMAL);
 			}
 	}
 
drivers/gpu/drm/lima/lima_sched.c +3 -2
index f522c5f99729..a7e53878d841 100644
@@ -159,9 +159,10 @@ int lima_sched_context_init(struct lima_sched_pipe *pipe,
 			    struct lima_sched_context *context,
 			    atomic_t *guilty)
 {
-	struct drm_sched_rq *rq = pipe->base.sched_rq + DRM_SCHED_PRIORITY_NORMAL;
+	struct drm_gpu_scheduler *sched = &pipe->base;
 
-	return drm_sched_entity_init(&context->base, &rq, 1, guilty);
+	return drm_sched_entity_init(&context->base, &sched,
+				     1, guilty, DRM_SCHED_PRIORITY_NORMAL);
 }
 
 void lima_sched_context_fini(struct lima_sched_pipe *pipe,
drivers/gpu/drm/panfrost/panfrost_job.c +4 -3
index d411eb6c8eb9..84178bcf35c9 100644
@@ -542,12 +542,13 @@ int panfrost_job_open(struct panfrost_file_priv *panfrost_priv)
 {
 	struct panfrost_device *pfdev = panfrost_priv->pfdev;
 	struct panfrost_job_slot *js = pfdev->js;
-	struct drm_sched_rq *rq;
+	struct drm_gpu_scheduler *sched;
 	int ret, i;
 
 	for (i = 0; i < NUM_JOB_SLOTS; i++) {
-		rq = &js->queue[i].sched.sched_rq[DRM_SCHED_PRIORITY_NORMAL];
-		ret = drm_sched_entity_init(&panfrost_priv->sched_entity[i], &rq, 1, NULL);
+		sched = &js->queue[i].sched;
+		ret = drm_sched_entity_init(&panfrost_priv->sched_entity[i],
+					    &sched, 1, NULL, DRM_SCHED_PRIORITY_NORMAL);
 		if (WARN_ON(ret))
 			return ret;
 	}
drivers/gpu/drm/scheduler/sched_entity.c +23 -42

130 lines not on this page · raw message

drivers/gpu/drm/v3d/v3d_drv.c +4 -3
index 1a07462b4528..c6aff1aedd27 100644
@@ -140,7 +140,7 @@ v3d_open(struct drm_device *dev, struct drm_file *file)
 {
 	struct v3d_dev *v3d = to_v3d_dev(dev);
 	struct v3d_file_priv *v3d_priv;
-	struct drm_sched_rq *rq;
+	struct drm_gpu_scheduler *sched;
 	int i;
 
 	v3d_priv = kzalloc(sizeof(*v3d_priv), GFP_KERNEL);
@@ -150,8 +150,9 @@ v3d_open(struct drm_device *dev, struct drm_file *file)
 	v3d_priv->v3d = v3d;
 
 	for (i = 0; i < V3D_MAX_QUEUES; i++) {
-		rq = &v3d->queue[i].sched.sched_rq[DRM_SCHED_PRIORITY_NORMAL];
-		drm_sched_entity_init(&v3d_priv->sched_entity[i], &rq, 1, NULL);
+		sched = &v3d->queue[i].sched;
+		drm_sched_entity_init(&v3d_priv->sched_entity[i], &sched,
+				      1, NULL, DRM_SCHED_PRIORITY_NORMAL);
 	}
 
 	file->driver_priv = v3d_priv;
include/drm/gpu_scheduler.h +5 -4
index 684692a8ed76..9df322dfac30 100644
@@ -81,8 +81,9 @@ enum drm_sched_priority {
 struct drm_sched_entity {
 	struct list_head		list;
 	struct drm_sched_rq		*rq;
-	struct drm_sched_rq		**rq_list;
-	unsigned int                    num_rq_list;
+	unsigned int                    num_sched_list;
+	struct drm_gpu_scheduler        **sched_list;
+	enum drm_sched_priority         priority;
 	spinlock_t			rq_lock;
 
 	struct spsc_queue		job_queue;
@@ -312,9 +313,9 @@ void drm_sched_rq_remove_entity(struct drm_sched_rq *rq,
 				struct drm_sched_entity *entity);
 
 int drm_sched_entity_init(struct drm_sched_entity *entity,
-			  struct drm_sched_rq **rq_list,
+			  struct drm_gpu_scheduler **sched_list,
 			  unsigned int num_rq_list,
-			  atomic_t *guilty);
+			  atomic_t *guilty, enum drm_sched_priority priority);
 long drm_sched_entity_flush(struct drm_sched_entity *entity, long timeout);
 void drm_sched_entity_fini(struct drm_sched_entity *entity);
 void drm_sched_entity_destroy(struct drm_sched_entity *entity);
Signature · 6 lines
2.23.0

_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx
#1 lore raw

Hi Christian,

I am not exactly sure about drm_sched_entity_set_priority() I wonder if just changing

entity->priority  to ctx->override_priority should work. With this change drm_sched_entity_select_rq()

will chose a rq based on entity->priority which seems to me correct. But is this enough to fix the old bug you were

talking about which mess up already scheduled job on priority change?

okay I just realized I need a lock to make sure

drm_sched_entity_set_priority() and drm_sched_entity_select_rq() shouldn't happen at the same time.

Regards,

Nirmoy

Quoted 392 lines · Nirmoy Das

Entity currently keeps a copy of run_queue list and modify it in drm_sched_entity_set_priority(). Entities shouldn't modify run_queue list. Use drm_gpu_scheduler list instead of drm_sched_rq list in drm_sched_entity struct. In this way we can select a runqueue based on entity/ctx's priority for a drm scheduler.

---
  drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c  |  7 +--
  drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c  |  7 +--
  drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c  |  7 +--
  drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c  |  7 +--
  drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c   | 14 +++--
  drivers/gpu/drm/etnaviv/etnaviv_drv.c    |  8 +--
  drivers/gpu/drm/lima/lima_sched.c        |  5 +-
  drivers/gpu/drm/panfrost/panfrost_job.c  |  7 +--
  drivers/gpu/drm/scheduler/sched_entity.c | 65 +++++++++---------------
  drivers/gpu/drm/v3d/v3d_drv.c            |  7 +--
  include/drm/gpu_scheduler.h              |  9 ++--
  11 files changed, 69 insertions(+), 74 deletions(-)
drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c +4 -3
index a0d3d7b756eb..e8f46c13d073 100644
@@ -122,7 +122,7 @@ static int amdgpu_ctx_init(struct amdgpu_device *adev,
 
  	for (i = 0; i < AMDGPU_HW_IP_NUM; ++i) {
  		struct amdgpu_ring *rings[AMDGPU_MAX_RINGS];
-		struct drm_sched_rq *rqs[AMDGPU_MAX_RINGS];
+		struct drm_gpu_scheduler *sched_list[AMDGPU_MAX_RINGS];
  		unsigned num_rings = 0;
  		unsigned num_rqs = 0;
 
@@ -181,12 +181,13 @@ static int amdgpu_ctx_init(struct amdgpu_device *adev,
  			if (!rings[j]->adev)
  				continue;
 
-			rqs[num_rqs++] = &rings[j]->sched.sched_rq[priority];
+			sched_list[num_rqs++] = &rings[j]->sched;
  		}
 
  		for (j = 0; j < amdgpu_ctx_num_entities[i]; ++j)
  			r = drm_sched_entity_init(&ctx->entities[i][j].entity,
-						  rqs, num_rqs, &ctx->guilty);
+						  sched_list, num_rqs,
+						  &ctx->guilty, priority);
  		if (r)
  			goto error_cleanup_entities;
  	}
drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c +4 -3
index 19ffe00d9072..a960dd7c0711 100644
@@ -1957,11 +1957,12 @@ void amdgpu_ttm_set_buffer_funcs_status(struct amdgpu_device *adev, bool enable)
 
  	if (enable) {
  		struct amdgpu_ring *ring;
-		struct drm_sched_rq *rq;
+		struct drm_gpu_scheduler *sched;
 
  		ring = adev->mman.buffer_funcs_ring;
-		rq = &ring->sched.sched_rq[DRM_SCHED_PRIORITY_KERNEL];
-		r = drm_sched_entity_init(&adev->mman.entity, &rq, 1, NULL);
+		sched = &ring->sched;
+		r = drm_sched_entity_init(&adev->mman.entity, &sched,
+					  1, NULL, DRM_SCHED_PRIORITY_KERNEL);
  		if (r) {
  			DRM_ERROR("Failed setting up TTM BO move entity (%d)\n",
  				  r);
drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c +4 -3
index e324bfe6c58f..b803a8882864 100644
@@ -330,12 +330,13 @@ int amdgpu_uvd_sw_fini(struct amdgpu_device *adev)
  int amdgpu_uvd_entity_init(struct amdgpu_device *adev)
  {
  	struct amdgpu_ring *ring;
-	struct drm_sched_rq *rq;
+	struct drm_gpu_scheduler *sched;
  	int r;
 
  	ring = &adev->uvd.inst[0].ring;
-	rq = &ring->sched.sched_rq[DRM_SCHED_PRIORITY_NORMAL];
-	r = drm_sched_entity_init(&adev->uvd.entity, &rq, 1, NULL);
+	sched = &ring->sched;
+	r = drm_sched_entity_init(&adev->uvd.entity, &sched,
+				  1, NULL, DRM_SCHED_PRIORITY_NORMAL);
  	if (r) {
  		DRM_ERROR("Failed setting up UVD kernel entity.\n");
  		return r;
drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c +4 -3
index 46b590af2fd2..b44f28d44fb4 100644
@@ -240,12 +240,13 @@ int amdgpu_vce_sw_fini(struct amdgpu_device *adev)
  int amdgpu_vce_entity_init(struct amdgpu_device *adev)
  {
  	struct amdgpu_ring *ring;
-	struct drm_sched_rq *rq;
+	struct drm_gpu_scheduler *sched;
  	int r;
 
  	ring = &adev->vce.ring[0];
-	rq = &ring->sched.sched_rq[DRM_SCHED_PRIORITY_NORMAL];
-	r = drm_sched_entity_init(&adev->vce.entity, &rq, 1, NULL);
+	sched = &ring->sched;
+	r = drm_sched_entity_init(&adev->vce.entity, &sched,
+				  1, NULL, DRM_SCHED_PRIORITY_NORMAL);
  	if (r != 0) {
  		DRM_ERROR("Failed setting up VCE run queue.\n");
  		return r;
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +10 -4
index a94c4faa5af1..ec6141773a92 100644
@@ -2687,6 +2687,7 @@ int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm,
  {
  	struct amdgpu_bo_param bp;
  	struct amdgpu_bo *root;
+        struct drm_gpu_scheduler *sched_list[AMDGPU_MAX_RINGS];
  	int r, i;
 
  	vm->va = RB_ROOT_CACHED;
@@ -2700,14 +2701,19 @@ int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm,
  	spin_lock_init(&vm->invalidated_lock);
  	INIT_LIST_HEAD(&vm->freed);
 
+        for (i = 0; i < adev->vm_manager.vm_pte_num_rqs; i++)
+	       sched_list[i] = adev->vm_manager.vm_pte_rqs[i]->sched;
+
  	/* create scheduler entities for page table updates */
-	r = drm_sched_entity_init(&vm->direct, adev->vm_manager.vm_pte_rqs,
-				  adev->vm_manager.vm_pte_num_rqs, NULL);
+	r = drm_sched_entity_init(&vm->direct, sched_list,
+				  adev->vm_manager.vm_pte_num_rqs,
+				  NULL, DRM_SCHED_PRIORITY_KERNEL);
  	if (r)
  		return r;
 
-	r = drm_sched_entity_init(&vm->delayed, adev->vm_manager.vm_pte_rqs,
-				  adev->vm_manager.vm_pte_num_rqs, NULL);
+	r = drm_sched_entity_init(&vm->delayed, sched_list,
+				  adev->vm_manager.vm_pte_num_rqs,
+				  NULL, DRM_SCHED_PRIORITY_KERNEL);
  	if (r)
  		goto error_free_direct;
 
drivers/gpu/drm/etnaviv/etnaviv_drv.c +4 -4
index 1f9c01be40d7..a65c1e115e35 100644
@@ -65,12 +65,12 @@ static int etnaviv_open(struct drm_device *dev, struct drm_file *file)
 
  	for (i = 0; i < ETNA_MAX_PIPES; i++) {
  		struct etnaviv_gpu *gpu = priv->gpu[i];
-		struct drm_sched_rq *rq;
+		struct drm_gpu_scheduler *sched;
 
  		if (gpu) {
-			rq = &gpu->sched.sched_rq[DRM_SCHED_PRIORITY_NORMAL];
-			drm_sched_entity_init(&ctx->sched_entity[i],
-					      &rq, 1, NULL);
+			sched = &gpu->sched;
+			drm_sched_entity_init(&ctx->sched_entity[i], &sched,
+					      1, NULL, DRM_SCHED_PRIORITY_NORMAL);
  			}
  	}
 
drivers/gpu/drm/lima/lima_sched.c +3 -2
index f522c5f99729..a7e53878d841 100644
@@ -159,9 +159,10 @@ int lima_sched_context_init(struct lima_sched_pipe *pipe,
  			    struct lima_sched_context *context,
  			    atomic_t *guilty)
  {
-	struct drm_sched_rq *rq = pipe->base.sched_rq + DRM_SCHED_PRIORITY_NORMAL;
+	struct drm_gpu_scheduler *sched = &pipe->base;
 
-	return drm_sched_entity_init(&context->base, &rq, 1, guilty);
+	return drm_sched_entity_init(&context->base, &sched,
+				     1, guilty, DRM_SCHED_PRIORITY_NORMAL);
  }
 
  void lima_sched_context_fini(struct lima_sched_pipe *pipe,
drivers/gpu/drm/panfrost/panfrost_job.c +4 -3
index d411eb6c8eb9..84178bcf35c9 100644
@@ -542,12 +542,13 @@ int panfrost_job_open(struct panfrost_file_priv *panfrost_priv)
  {
  	struct panfrost_device *pfdev = panfrost_priv->pfdev;
  	struct panfrost_job_slot *js = pfdev->js;
-	struct drm_sched_rq *rq;
+	struct drm_gpu_scheduler *sched;
  	int ret, i;
 
  	for (i = 0; i < NUM_JOB_SLOTS; i++) {
-		rq = &js->queue[i].sched.sched_rq[DRM_SCHED_PRIORITY_NORMAL];
-		ret = drm_sched_entity_init(&panfrost_priv->sched_entity[i], &rq, 1, NULL);
+		sched = &js->queue[i].sched;
+		ret = drm_sched_entity_init(&panfrost_priv->sched_entity[i],
+					    &sched, 1, NULL, DRM_SCHED_PRIORITY_NORMAL);
  		if (WARN_ON(ret))
  			return ret;
  	}
drivers/gpu/drm/scheduler/sched_entity.c +23 -42
index 461a7a8129f4..e10d37266836 100644
@@ -38,9 +38,9 @@
   * submit to HW ring.
   *
   * @entity: scheduler entity to init
- * @rq_list: the list of run queue on which jobs from this
+ * @sched_list: the list of drm scheds on which jobs from this
   *           entity can be submitted
- * @num_rq_list: number of run queue in rq_list
+ * @num_sched_list: number of drm sched in sched_list
   * @guilty: atomic_t set to 1 when a job on this queue
   *          is found to be guilty causing a timeout
   *
@@ -50,32 +50,34 @@
   * Returns 0 on success or a negative error code on failure.
   */
  int drm_sched_entity_init(struct drm_sched_entity *entity,
-			  struct drm_sched_rq **rq_list,
-			  unsigned int num_rq_list,
-			  atomic_t *guilty)
+			  struct drm_gpu_scheduler **sched_list,
+			  unsigned int num_sched_list,
+			  atomic_t *guilty, enum drm_sched_priority priority)
  {
  	int i;
 
-	if (!(entity && rq_list && (num_rq_list == 0 || rq_list[0])))
+	if (!(entity && sched_list && (num_sched_list == 0 || sched_list[0])))
  		return -EINVAL;
 
  	memset(entity, 0, sizeof(struct drm_sched_entity));
  	INIT_LIST_HEAD(&entity->list);
  	entity->rq = NULL;
  	entity->guilty = guilty;
-	entity->num_rq_list = num_rq_list;
-	entity->rq_list = kcalloc(num_rq_list, sizeof(struct drm_sched_rq *),
-				GFP_KERNEL);
-	if (!entity->rq_list)
+	entity->num_sched_list = num_sched_list;
+	entity->priority = priority;
+	entity->sched_list =  kcalloc(num_sched_list,
+				      sizeof(struct drm_gpu_scheduler *), GFP_KERNEL);
+
+	if(!entity->sched_list)
  		return -ENOMEM;
 
  	init_completion(&entity->entity_idle);
 
-	for (i = 0; i < num_rq_list; ++i)
-		entity->rq_list[i] = rq_list[i];
+	for (i = 0; i < num_sched_list; i++)
+		entity->sched_list[i] = sched_list[i];
 
-	if (num_rq_list)
-		entity->rq = rq_list[0];
+	if (num_sched_list)
+		entity->rq = &entity->sched_list[0]->sched_rq[entity->priority];
 
  	entity->last_scheduled = NULL;
 
@@ -139,10 +141,10 @@ drm_sched_entity_get_free_sched(struct drm_sched_entity *entity)
  	unsigned int min_jobs = UINT_MAX, num_jobs;
  	int i;
 
-	for (i = 0; i < entity->num_rq_list; ++i) {
-		struct drm_gpu_scheduler *sched = entity->rq_list[i]->sched;
+	for (i = 0; i < entity->num_sched_list; ++i) {
+		struct drm_gpu_scheduler *sched = entity->sched_list[i];
 
-		if (!entity->rq_list[i]->sched->ready) {
+		if (!entity->sched_list[i]->ready) {
  			DRM_WARN("sched%s is not ready, skipping", sched->name);
  			continue;
  		}
@@ -150,7 +152,7 @@ drm_sched_entity_get_free_sched(struct drm_sched_entity *entity)
  		num_jobs = atomic_read(&sched->num_jobs);
  		if (num_jobs < min_jobs) {
  			min_jobs = num_jobs;
-			rq = entity->rq_list[i];
+			rq = &entity->sched_list[i]->sched_rq[entity->priority];
  		}
  	}
 
@@ -308,7 +310,7 @@ void drm_sched_entity_fini(struct drm_sched_entity *entity)
 
  	dma_fence_put(entity->last_scheduled);
  	entity->last_scheduled = NULL;
-	kfree(entity->rq_list);
+	kfree(entity->sched_list);
  }
  EXPORT_SYMBOL(drm_sched_entity_fini);
 
@@ -353,15 +355,6 @@ static void drm_sched_entity_wakeup(struct dma_fence *f,
  	drm_sched_wakeup(entity->rq->sched);
  }
 
-/**
- * drm_sched_entity_set_rq_priority - helper for drm_sched_entity_set_priority
- */
-static void drm_sched_entity_set_rq_priority(struct drm_sched_rq **rq,
-					     enum drm_sched_priority priority)
-{
-	*rq = &(*rq)->sched->sched_rq[priority];
-}
-
  /**
   * drm_sched_entity_set_priority - Sets priority of the entity
   *
@@ -373,20 +366,8 @@ static void drm_sched_entity_set_rq_priority(struct drm_sched_rq **rq,
  void drm_sched_entity_set_priority(struct drm_sched_entity *entity,
  				   enum drm_sched_priority priority)
  {
-	unsigned int i;
-
-	spin_lock(&entity->rq_lock);
 
-	for (i = 0; i < entity->num_rq_list; ++i)
-		drm_sched_entity_set_rq_priority(&entity->rq_list[i], priority);
-
-	if (entity->rq) {
-		drm_sched_rq_remove_entity(entity->rq, entity);
-		drm_sched_entity_set_rq_priority(&entity->rq, priority);
-		drm_sched_rq_add_entity(entity->rq, entity);
-	}
-
-	spin_unlock(&entity->rq_lock);
+	entity->priority = priority;
  }
  EXPORT_SYMBOL(drm_sched_entity_set_priority);
 
@@ -490,7 +471,7 @@ void drm_sched_entity_select_rq(struct drm_sched_entity *entity)
  	struct dma_fence *fence;
  	struct drm_sched_rq *rq;
 
-	if (spsc_queue_count(&entity->job_queue) || entity->num_rq_list <= 1)
+	if (spsc_queue_count(&entity->job_queue) || entity->num_sched_list <= 1)
  		return;
 
  	fence = READ_ONCE(entity->last_scheduled);
drivers/gpu/drm/v3d/v3d_drv.c +4 -3
index 1a07462b4528..c6aff1aedd27 100644
@@ -140,7 +140,7 @@ v3d_open(struct drm_device *dev, struct drm_file *file)
  {
  	struct v3d_dev *v3d = to_v3d_dev(dev);
  	struct v3d_file_priv *v3d_priv;
-	struct drm_sched_rq *rq;
+	struct drm_gpu_scheduler *sched;
  	int i;
 
  	v3d_priv = kzalloc(sizeof(*v3d_priv), GFP_KERNEL);
@@ -150,8 +150,9 @@ v3d_open(struct drm_device *dev, struct drm_file *file)
  	v3d_priv->v3d = v3d;
 
  	for (i = 0; i < V3D_MAX_QUEUES; i++) {
-		rq = &v3d->queue[i].sched.sched_rq[DRM_SCHED_PRIORITY_NORMAL];
-		drm_sched_entity_init(&v3d_priv->sched_entity[i], &rq, 1, NULL);
+		sched = &v3d->queue[i].sched;
+		drm_sched_entity_init(&v3d_priv->sched_entity[i], &sched,
+				      1, NULL, DRM_SCHED_PRIORITY_NORMAL);
  	}
 
  	file->driver_priv = v3d_priv;
include/drm/gpu_scheduler.h +3 -2
index 684692a8ed76..9df322dfac30 100644
@@ -81,8 +81,9 @@ enum drm_sched_priority {
  struct drm_sched_entity {
  	struct list_head		list;
  	struct drm_sched_rq		*rq;
-	struct drm_sched_rq		**rq_list;
-	unsigned int                    num_rq_list;
+	unsigned int                    num_sched_list;
+	struct drm_gpu_scheduler        **sched_list;
+	enum drm_sched_priority         priority;
  	spinlock_t			rq_lock;
 
  	struct spsc_queue		job_queue;
@@ -312,9 +313,9 @@ void drm_sched_rq_remove_entity(struct drm_sched_rq *rq,
  				struct drm_sched_entity *entity);

include/drm/gpu_scheduler.h void drm_sched_rq_remove_entity(struct drm_sched_rq *rq, · line 321

Quoted 10 lines
include/drm/gpu_scheduler.h +2 -2
 
  int drm_sched_entity_init(struct drm_sched_entity *entity,
-			  struct drm_sched_rq **rq_list,
+			  struct drm_gpu_scheduler **sched_list,
  			  unsigned int num_rq_list,
-			  atomic_t *guilty);
+			  atomic_t *guilty, enum drm_sched_priority priority);
  long drm_sched_entity_flush(struct drm_sched_entity *entity, long timeout);
  void drm_sched_entity_fini(struct drm_sched_entity *entity);
  void drm_sched_entity_destroy(struct drm_sched_entity *entity);

_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx

maintainer
#2 lore raw

Am 05.12.19 um 12:04 schrieb Nirmoy:

Quoted 9 lines

Hi Christian,

I am not exactly sure about drm_sched_entity_set_priority() I wonder if just changing

entity->priority  to ctx->override_priority should work. With this change drm_sched_entity_select_rq()

will chose a rq based on entity->priority which seems to me correct. But is this enough to fix the old bug you were

talking about which mess up already scheduled job on priority change?

Yes, that should perfectly do it.

Quoted 4 lines

okay I just realized I need a lock to make sure

drm_sched_entity_set_priority() and drm_sched_entity_select_rq() shouldn't happen at the same time.

Yeah, you probably need to grab the lock and make sure that you get the priority to use while holding the lock as well.

Regards,
Christian.

Quoted 381 lines

Regards,

Nirmoy

Quoted 373 lines · Nirmoy Das

Entity currently keeps a copy of run_queue list and modify it in drm_sched_entity_set_priority(). Entities shouldn't modify run_queue list. Use drm_gpu_scheduler list instead of drm_sched_rq list in drm_sched_entity struct. In this way we can select a runqueue based on entity/ctx's priority for a  drm scheduler.

---

  drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c  |  7 +--
  drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c  |  7 +--
  drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c  |  7 +--
  drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c  |  7 +--

  drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c   | 14 +++-- drivers/gpu/drm/etnaviv/etnaviv_drv.c    |  8 +-- drivers/gpu/drm/lima/lima_sched.c        |  5 +- drivers/gpu/drm/panfrost/panfrost_job.c  |  7 +-- drivers/gpu/drm/scheduler/sched_entity.c | 65 +++++++++--------------- drivers/gpu/drm/v3d/v3d_drv.c            |  7 +-- include/drm/gpu_scheduler.h              |  9 ++--

  11 files changed, 69 insertions(+), 74 deletions(-)
drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c +1 -1
index a0d3d7b756eb..e8f46c13d073 100644
@@ -122,7 +122,7 @@ static int amdgpu_ctx_init(struct amdgpu_device 
 *adev,
         for (i = 0; i < AMDGPU_HW_IP_NUM; ++i) {
           struct amdgpu_ring *rings[AMDGPU_MAX_RINGS];
-        struct drm_sched_rq *rqs[AMDGPU_MAX_RINGS];
+        struct drm_gpu_scheduler *sched_list[AMDGPU_MAX_RINGS];
           unsigned num_rings = 0;
           unsigned num_rqs = 0;
   @@ -181,12 +181,13 @@ static int amdgpu_ctx_init(struct amdgpu_device *adev,

              if (!rings[j]->adev)
                  continue;
  -            rqs[num_rqs++] = &rings[j]->sched.sched_rq[priority]; +            sched_list[num_rqs++] = &rings[j]->sched;
          }
            for (j = 0; j < amdgpu_ctx_num_entities[i]; ++j) r = drm_sched_entity_init(&ctx->entities[i][j].entity, -                          rqs, num_rqs, &ctx->guilty);
+                          sched_list, num_rqs,
+                          &ctx->guilty, priority);
          if (r)
              goto error_cleanup_entities;
      }

drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c +4 -3
index 19ffe00d9072..a960dd7c0711 100644
@@ -1957,11 +1957,12 @@ void 
 amdgpu_ttm_set_buffer_funcs_status(struct amdgpu_device *adev, bool enable)
         if (enable) {
           struct amdgpu_ring *ring;
-        struct drm_sched_rq *rq;
+        struct drm_gpu_scheduler *sched;
             ring = adev->mman.buffer_funcs_ring;
-        rq = &ring->sched.sched_rq[DRM_SCHED_PRIORITY_KERNEL];
-        r = drm_sched_entity_init(&adev->mman.entity, &rq, 1, NULL);
+        sched = &ring->sched;
+        r = drm_sched_entity_init(&adev->mman.entity, &sched,
+                      1, NULL, DRM_SCHED_PRIORITY_KERNEL);
           if (r) {
               DRM_ERROR("Failed setting up TTM BO move entity (%d)\n",
                     r);
drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c +4 -3
index e324bfe6c58f..b803a8882864 100644
@@ -330,12 +330,13 @@ int amdgpu_uvd_sw_fini(struct amdgpu_device *adev)
   int amdgpu_uvd_entity_init(struct amdgpu_device *adev)
   {
       struct amdgpu_ring *ring;
-    struct drm_sched_rq *rq;
+    struct drm_gpu_scheduler *sched;
       int r;
         ring = &adev->uvd.inst[0].ring;
-    rq = &ring->sched.sched_rq[DRM_SCHED_PRIORITY_NORMAL];
-    r = drm_sched_entity_init(&adev->uvd.entity, &rq, 1, NULL);
+    sched = &ring->sched;
+    r = drm_sched_entity_init(&adev->uvd.entity, &sched,
+                  1, NULL, DRM_SCHED_PRIORITY_NORMAL);
       if (r) {
           DRM_ERROR("Failed setting up UVD kernel entity.\n");
           return r;
drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c +4 -3
index 46b590af2fd2..b44f28d44fb4 100644
@@ -240,12 +240,13 @@ int amdgpu_vce_sw_fini(struct amdgpu_device *adev)
   int amdgpu_vce_entity_init(struct amdgpu_device *adev)
   {
       struct amdgpu_ring *ring;
-    struct drm_sched_rq *rq;
+    struct drm_gpu_scheduler *sched;
       int r;
         ring = &adev->vce.ring[0];
-    rq = &ring->sched.sched_rq[DRM_SCHED_PRIORITY_NORMAL];
-    r = drm_sched_entity_init(&adev->vce.entity, &rq, 1, NULL);
+    sched = &ring->sched;
+    r = drm_sched_entity_init(&adev->vce.entity, &sched,
+                  1, NULL, DRM_SCHED_PRIORITY_NORMAL);
       if (r != 0) {
           DRM_ERROR("Failed setting up VCE run queue.\n");
           return r;
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +9 -3
index a94c4faa5af1..ec6141773a92 100644
@@ -2687,6 +2687,7 @@ int amdgpu_vm_init(struct amdgpu_device *adev, 
 struct amdgpu_vm *vm,
   {
       struct amdgpu_bo_param bp;
       struct amdgpu_bo *root;
+        struct drm_gpu_scheduler *sched_list[AMDGPU_MAX_RINGS];
       int r, i;
         vm->va = RB_ROOT_CACHED;
@@ -2700,14 +2701,19 @@ int amdgpu_vm_init(struct amdgpu_device 
 *adev, struct amdgpu_vm *vm,
       spin_lock_init(&vm->invalidated_lock);
       INIT_LIST_HEAD(&vm->freed);
   +        for (i = 0; i < adev->vm_manager.vm_pte_num_rqs; i++)
+           sched_list[i] = adev->vm_manager.vm_pte_rqs[i]->sched;
+
       /* create scheduler entities for page table updates */
-    r = drm_sched_entity_init(&vm->direct, adev->vm_manager.vm_pte_rqs,
-                  adev->vm_manager.vm_pte_num_rqs, NULL);
+    r = drm_sched_entity_init(&vm->direct, sched_list,
+                  adev->vm_manager.vm_pte_num_rqs,
+                  NULL, DRM_SCHED_PRIORITY_KERNEL);
       if (r)
           return r;
   -    r = drm_sched_entity_init(&vm->delayed, adev->vm_manager.vm_pte_rqs,
-                  adev->vm_manager.vm_pte_num_rqs, NULL);
+    r = drm_sched_entity_init(&vm->delayed, sched_list,
+                  adev->vm_manager.vm_pte_num_rqs,
+                  NULL, DRM_SCHED_PRIORITY_KERNEL);
       if (r)
           goto error_free_direct;

  diff --git a/drivers/gpu/drm/etnaviv/etnaviv_drv.c b/drivers/gpu/drm/etnaviv/etnaviv_drv.c index 1f9c01be40d7..a65c1e115e35 100644

drivers/gpu/drm/etnaviv/etnaviv_drv.c +4 -4
@@ -65,12 +65,12 @@ static int etnaviv_open(struct drm_device *dev, 
 struct drm_file *file)
         for (i = 0; i < ETNA_MAX_PIPES; i++) {
           struct etnaviv_gpu *gpu = priv->gpu[i];
-        struct drm_sched_rq *rq;
+        struct drm_gpu_scheduler *sched;
             if (gpu) {
-            rq = &gpu->sched.sched_rq[DRM_SCHED_PRIORITY_NORMAL];
-            drm_sched_entity_init(&ctx->sched_entity[i],
-                          &rq, 1, NULL);
+            sched = &gpu->sched;
+            drm_sched_entity_init(&ctx->sched_entity[i], &sched,
+                          1, NULL, DRM_SCHED_PRIORITY_NORMAL);
               }
       }

  diff --git a/drivers/gpu/drm/lima/lima_sched.c b/drivers/gpu/drm/lima/lima_sched.c index f522c5f99729..a7e53878d841 100644

drivers/gpu/drm/lima/lima_sched.c +3 -1
@@ -159,9 +159,10 @@ int lima_sched_context_init(struct 
 lima_sched_pipe *pipe,
                   struct lima_sched_context *context,
                   atomic_t *guilty)
   {
-    struct drm_sched_rq *rq = pipe->base.sched_rq + 
 DRM_SCHED_PRIORITY_NORMAL;
+    struct drm_gpu_scheduler *sched = &pipe->base;
   -    return drm_sched_entity_init(&context->base, &rq, 1, guilty);
+    return drm_sched_entity_init(&context->base, &sched,
+                     1, guilty, DRM_SCHED_PRIORITY_NORMAL);
   }

    void lima_sched_context_fini(struct lima_sched_pipe *pipe,

drivers/gpu/drm/panfrost/panfrost_job.c +4 -3
index d411eb6c8eb9..84178bcf35c9 100644
@@ -542,12 +542,13 @@ int panfrost_job_open(struct panfrost_file_priv 
 *panfrost_priv)
   {
       struct panfrost_device *pfdev = panfrost_priv->pfdev;
       struct panfrost_job_slot *js = pfdev->js;
-    struct drm_sched_rq *rq;
+    struct drm_gpu_scheduler *sched;
       int ret, i;
         for (i = 0; i < NUM_JOB_SLOTS; i++) {
-        rq = &js->queue[i].sched.sched_rq[DRM_SCHED_PRIORITY_NORMAL];
-        ret = drm_sched_entity_init(&panfrost_priv->sched_entity[i], 
 &rq, 1, NULL);
+        sched = &js->queue[i].sched;
+        ret = drm_sched_entity_init(&panfrost_priv->sched_entity[i],
+                        &sched, 1, NULL, DRM_SCHED_PRIORITY_NORMAL);
           if (WARN_ON(ret))
               return ret;

      }

drivers/gpu/drm/scheduler/sched_entity.c +16 -11
index 461a7a8129f4..e10d37266836 100644
@@ -38,9 +38,9 @@
    * submit to HW ring.
    *
    * @entity: scheduler entity to init
- * @rq_list: the list of run queue on which jobs from this
+ * @sched_list: the list of drm scheds on which jobs from this
    *           entity can be submitted
- * @num_rq_list: number of run queue in rq_list
+ * @num_sched_list: number of drm sched in sched_list
    * @guilty: atomic_t set to 1 when a job on this queue
    *          is found to be guilty causing a timeout
    *
@@ -50,32 +50,34 @@
    * Returns 0 on success or a negative error code on failure.
    */
   int drm_sched_entity_init(struct drm_sched_entity *entity,
-              struct drm_sched_rq **rq_list,
-              unsigned int num_rq_list,
-              atomic_t *guilty)
+              struct drm_gpu_scheduler **sched_list,
+              unsigned int num_sched_list,
+              atomic_t *guilty, enum drm_sched_priority priority)
   {
       int i;
   -    if (!(entity && rq_list && (num_rq_list == 0 || rq_list[0])))
+    if (!(entity && sched_list && (num_sched_list == 0 || 
 sched_list[0])))
           return -EINVAL;
         memset(entity, 0, sizeof(struct drm_sched_entity));
       INIT_LIST_HEAD(&entity->list);
       entity->rq = NULL;
       entity->guilty = guilty;
-    entity->num_rq_list = num_rq_list;
-    entity->rq_list = kcalloc(num_rq_list, sizeof(struct 
 drm_sched_rq *),
-                GFP_KERNEL);
-    if (!entity->rq_list)
+    entity->num_sched_list = num_sched_list;
+    entity->priority = priority;
+    entity->sched_list =  kcalloc(num_sched_list,
+                      sizeof(struct drm_gpu_scheduler *), GFP_KERNEL);
+
+    if(!entity->sched_list)
           return -ENOMEM;
         init_completion(&entity->entity_idle);
   -    for (i = 0; i < num_rq_list; ++i)
-        entity->rq_list[i] = rq_list[i];
+    for (i = 0; i < num_sched_list; i++)
+        entity->sched_list[i] = sched_list[i];
   -    if (num_rq_list)
-        entity->rq = rq_list[0];
+    if (num_sched_list)
+        entity->rq = 
 &entity->sched_list[0]->sched_rq[entity->priority];
         entity->last_scheduled = NULL;
   @@ -139,10 +141,10 @@ drm_sched_entity_get_free_sched(struct drm_sched_entity *entity)

      unsigned int min_jobs = UINT_MAX, num_jobs;
      int i;
  -    for (i = 0; i < entity->num_rq_list; ++i) {
-        struct drm_gpu_scheduler *sched = entity->rq_list[i]->sched; +    for (i = 0; i < entity->num_sched_list; ++i) {
+        struct drm_gpu_scheduler *sched = entity->sched_list[i]; -        if (!entity->rq_list[i]->sched->ready) {
+        if (!entity->sched_list[i]->ready) { DRM_WARN("sched%s is not ready, skipping", sched->name); continue;
          }

drivers/gpu/drm/scheduler/sched_entity.c +1 -1
@@ -150,7 +152,7 @@ drm_sched_entity_get_free_sched(struct 
 drm_sched_entity *entity)
           num_jobs = atomic_read(&sched->num_jobs);
           if (num_jobs < min_jobs) {
               min_jobs = num_jobs;
-            rq = entity->rq_list[i];
+            rq = &entity->sched_list[i]->sched_rq[entity->priority];
           }
       }

  @@ -308,7 +310,7 @@ void drm_sched_entity_fini(struct drm_sched_entity *entity) dma_fence_put(entity->last_scheduled); entity->last_scheduled = NULL;
-    kfree(entity->rq_list);
+    kfree(entity->sched_list);
  }
  EXPORT_SYMBOL(drm_sched_entity_fini);
  @@ -353,15 +355,6 @@ static void drm_sched_entity_wakeup(struct dma_fence *f, drm_sched_wakeup(entity->rq->sched);
  }
  -/**
- * drm_sched_entity_set_rq_priority - helper for drm_sched_entity_set_priority
- */

drivers/gpu/drm/scheduler/sched_entity.c +0 -6
-static void drm_sched_entity_set_rq_priority(struct drm_sched_rq **rq,
-                         enum drm_sched_priority priority)
-{
-    *rq = &(*rq)->sched->sched_rq[priority];
-}
-

  /**
   * drm_sched_entity_set_priority - Sets priority of the entity *

drivers/gpu/drm/scheduler/sched_entity.c +1 -12
@@ -373,20 +366,8 @@ static void 
 drm_sched_entity_set_rq_priority(struct drm_sched_rq **rq,
   void drm_sched_entity_set_priority(struct drm_sched_entity *entity,
                      enum drm_sched_priority priority)
   {
-    unsigned int i;
-
-    spin_lock(&entity->rq_lock);
   -    for (i = 0; i < entity->num_rq_list; ++i)
- drm_sched_entity_set_rq_priority(&entity->rq_list[i], priority);
-
-    if (entity->rq) {
-        drm_sched_rq_remove_entity(entity->rq, entity);
-        drm_sched_entity_set_rq_priority(&entity->rq, priority);
-        drm_sched_rq_add_entity(entity->rq, entity);
-    }
-
-    spin_unlock(&entity->rq_lock);
+    entity->priority = priority;
   }
   EXPORT_SYMBOL(drm_sched_entity_set_priority);

  @@ -490,7 +471,7 @@ void drm_sched_entity_select_rq(struct drm_sched_entity *entity) struct dma_fence *fence;
      struct drm_sched_rq *rq;
  -    if (spsc_queue_count(&entity->job_queue) || entity->num_rq_list <= 1) +    if (spsc_queue_count(&entity->job_queue) ||
entity->num_sched_list <= 1)
          return;
        fence = READ_ONCE(entity->last_scheduled);

drivers/gpu/drm/v3d/v3d_drv.c +4 -3
index 1a07462b4528..c6aff1aedd27 100644
@@ -140,7 +140,7 @@ v3d_open(struct drm_device *dev, struct drm_file 
 *file)
   {
       struct v3d_dev *v3d = to_v3d_dev(dev);
       struct v3d_file_priv *v3d_priv;
-    struct drm_sched_rq *rq;
+    struct drm_gpu_scheduler *sched;
       int i;
         v3d_priv = kzalloc(sizeof(*v3d_priv), GFP_KERNEL);
@@ -150,8 +150,9 @@ v3d_open(struct drm_device *dev, struct drm_file 
 *file)
       v3d_priv->v3d = v3d;
         for (i = 0; i < V3D_MAX_QUEUES; i++) {
-        rq = &v3d->queue[i].sched.sched_rq[DRM_SCHED_PRIORITY_NORMAL];
- drm_sched_entity_init(&v3d_priv->sched_entity[i], &rq, 1, NULL);
+        sched = &v3d->queue[i].sched;
+ drm_sched_entity_init(&v3d_priv->sched_entity[i], &sched,
+                      1, NULL, DRM_SCHED_PRIORITY_NORMAL);
       }
         file->driver_priv = v3d_priv;
include/drm/gpu_scheduler.h +3 -2
index 684692a8ed76..9df322dfac30 100644
@@ -81,8 +81,9 @@ enum drm_sched_priority {
   struct drm_sched_entity {
       struct list_head        list;
       struct drm_sched_rq        *rq;
-    struct drm_sched_rq        **rq_list;
-    unsigned int                    num_rq_list;
+    unsigned int                    num_sched_list;
+    struct drm_gpu_scheduler        **sched_list;
+    enum drm_sched_priority         priority;
       spinlock_t            rq_lock;
         struct spsc_queue        job_queue;
@@ -312,9 +313,9 @@ void drm_sched_rq_remove_entity(struct 
 drm_sched_rq *rq,
                   struct drm_sched_entity *entity);
     int drm_sched_entity_init(struct drm_sched_entity *entity,
Quoted 8 lines

include/drm/gpu_scheduler.h void drm_sched_rq_remove_entity(struct drm_sched_rq *rq, · line 321

Quoted 8 lines
include/drm/gpu_scheduler.h +2 -2
-              struct drm_sched_rq **rq_list,
+              struct drm_gpu_scheduler **sched_list,
                 unsigned int num_rq_list,
-              atomic_t *guilty);
+              atomic_t *guilty, enum drm_sched_priority priority);
   long drm_sched_entity_flush(struct drm_sched_entity *entity, long timeout);
   void drm_sched_entity_fini(struct drm_sched_entity *entity);
   void drm_sched_entity_destroy(struct drm_sched_entity *entity);

_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx

#3 lore raw

[AMD Official Use Only - Internal Distribution Only]

Thanks Christian,

I will modify and resend.

Regards,
Nirmoy
________________________________
From: Koenig, Christian <[email protected]>
Sent: Thursday, December 5, 2019 1:29:49 PM
To: Das, Nirmoy <[email protected]>; Nirmoy Das <[email protected]>; Deucher, Alexander <[email protected]>; Ho, Kenny <[email protected]>
Cc: [email protected] <[email protected]>; Das, Nirmoy <[email protected]>
Subject: Re: [RFC PATCH] drm/scheduler: rework entity creation

Am 05.12.19 um 12:04 schrieb Nirmoy:

Quoted 5 lines

Hi Christian,

I am not exactly sure about drm_sched_entity_set_priority() I wonder if just changing

Quoted 7 lines

entity->priority to ctx->override_priority should work. With this change drm_sched_entity_select_rq()

will chose a rq based on entity->priority which seems to me correct. But is this enough to fix the old bug you were

talking about which mess up already scheduled job on priority change?

Yes, that should perfectly do it.

Quoted 5 lines

okay I just realized I need a lock to make sure

drm_sched_entity_set_priority() and drm_sched_entity_select_rq() shouldn't happen at the same time.

Yeah, you probably need to grab the lock and make sure that you get the priority to use while holding the lock as well.

Regards,
Christian.

Quoted 400 lines

Regards,

Nirmoy

Quoted 392 lines · Nirmoy Das

Entity currently keeps a copy of run_queue list and modify it in drm_sched_entity_set_priority(). Entities shouldn't modify run_queue list. Use drm_gpu_scheduler list instead of drm_sched_rq list in drm_sched_entity struct. In this way we can select a runqueue based on entity/ctx's priority for a drm scheduler.

---
  drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c  |  7 +--
  drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c  |  7 +--
  drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c  |  7 +--
  drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c  |  7 +--
  drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c   | 14 +++--
  drivers/gpu/drm/etnaviv/etnaviv_drv.c    |  8 +--
  drivers/gpu/drm/lima/lima_sched.c        |  5 +-
  drivers/gpu/drm/panfrost/panfrost_job.c  |  7 +--
  drivers/gpu/drm/scheduler/sched_entity.c | 65 +++++++++---------------
  drivers/gpu/drm/v3d/v3d_drv.c            |  7 +--
  include/drm/gpu_scheduler.h              |  9 ++--
  11 files changed, 69 insertions(+), 74 deletions(-)
drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c +1 -1
index a0d3d7b756eb..e8f46c13d073 100644
@@ -122,7 +122,7 @@ static int amdgpu_ctx_init(struct amdgpu_device
 *adev,
        for (i = 0; i < AMDGPU_HW_IP_NUM; ++i) {
          struct amdgpu_ring *rings[AMDGPU_MAX_RINGS];
-        struct drm_sched_rq *rqs[AMDGPU_MAX_RINGS];
+        struct drm_gpu_scheduler *sched_list[AMDGPU_MAX_RINGS];
          unsigned num_rings = 0;
          unsigned num_rqs = 0;
  @@ -181,12 +181,13 @@ static int amdgpu_ctx_init(struct

amdgpu_device *adev,

              if (!rings[j]->adev)
                  continue;

- rqs[num_rqs++] = &rings[j]->sched.sched_rq[priority];

drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c +3 -1
+            sched_list[num_rqs++] = &rings[j]->sched;
          }
            for (j = 0; j < amdgpu_ctx_num_entities[i]; ++j)
              r = drm_sched_entity_init(&ctx->entities[i][j].entity,
-                          rqs, num_rqs, &ctx->guilty);
+                          sched_list, num_rqs,
+                          &ctx->guilty, priority);
          if (r)
              goto error_cleanup_entities;
      }
drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c +0 -0
index 19ffe00d9072..a960dd7c0711 100644
@@ -1957,11 +1957,12 @@ void

amdgpu_ttm_set_buffer_funcs_status(struct amdgpu_device *adev, bool enable)

        if (enable) {
          struct amdgpu_ring *ring;
drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c +4 -3
-        struct drm_sched_rq *rq;
+        struct drm_gpu_scheduler *sched;
            ring = adev->mman.buffer_funcs_ring;
-        rq = &ring->sched.sched_rq[DRM_SCHED_PRIORITY_KERNEL];
-        r = drm_sched_entity_init(&adev->mman.entity, &rq, 1, NULL);
+        sched = &ring->sched;
+        r = drm_sched_entity_init(&adev->mman.entity, &sched,
+                      1, NULL, DRM_SCHED_PRIORITY_KERNEL);
          if (r) {
              DRM_ERROR("Failed setting up TTM BO move entity (%d)\n",
                    r);
drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c +4 -3
index e324bfe6c58f..b803a8882864 100644
@@ -330,12 +330,13 @@ int amdgpu_uvd_sw_fini(struct amdgpu_device *adev)
  int amdgpu_uvd_entity_init(struct amdgpu_device *adev)
  {
      struct amdgpu_ring *ring;
-    struct drm_sched_rq *rq;
+    struct drm_gpu_scheduler *sched;
      int r;
        ring = &adev->uvd.inst[0].ring;
-    rq = &ring->sched.sched_rq[DRM_SCHED_PRIORITY_NORMAL];
-    r = drm_sched_entity_init(&adev->uvd.entity, &rq, 1, NULL);
+    sched = &ring->sched;
+    r = drm_sched_entity_init(&adev->uvd.entity, &sched,
+                  1, NULL, DRM_SCHED_PRIORITY_NORMAL);
      if (r) {
          DRM_ERROR("Failed setting up UVD kernel entity.\n");
          return r;
drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c +4 -3
index 46b590af2fd2..b44f28d44fb4 100644
@@ -240,12 +240,13 @@ int amdgpu_vce_sw_fini(struct amdgpu_device *adev)
  int amdgpu_vce_entity_init(struct amdgpu_device *adev)
  {
      struct amdgpu_ring *ring;
-    struct drm_sched_rq *rq;
+    struct drm_gpu_scheduler *sched;
      int r;
        ring = &adev->vce.ring[0];
-    rq = &ring->sched.sched_rq[DRM_SCHED_PRIORITY_NORMAL];
-    r = drm_sched_entity_init(&adev->vce.entity, &rq, 1, NULL);
+    sched = &ring->sched;
+    r = drm_sched_entity_init(&adev->vce.entity, &sched,
+                  1, NULL, DRM_SCHED_PRIORITY_NORMAL);
      if (r != 0) {
          DRM_ERROR("Failed setting up VCE run queue.\n");
          return r;
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +9 -3
index a94c4faa5af1..ec6141773a92 100644
@@ -2687,6 +2687,7 @@ int amdgpu_vm_init(struct amdgpu_device *adev,
 struct amdgpu_vm *vm,
  {
      struct amdgpu_bo_param bp;
      struct amdgpu_bo *root;
+        struct drm_gpu_scheduler *sched_list[AMDGPU_MAX_RINGS];
      int r, i;
        vm->va = RB_ROOT_CACHED;
@@ -2700,14 +2701,19 @@ int amdgpu_vm_init(struct amdgpu_device
 *adev, struct amdgpu_vm *vm,
      spin_lock_init(&vm->invalidated_lock);
      INIT_LIST_HEAD(&vm->freed);
  +        for (i = 0; i < adev->vm_manager.vm_pte_num_rqs; i++)
+           sched_list[i] = adev->vm_manager.vm_pte_rqs[i]->sched;
+
      /* create scheduler entities for page table updates */
-    r = drm_sched_entity_init(&vm->direct, adev->vm_manager.vm_pte_rqs,
-                  adev->vm_manager.vm_pte_num_rqs, NULL);
+    r = drm_sched_entity_init(&vm->direct, sched_list,
+                  adev->vm_manager.vm_pte_num_rqs,
+                  NULL, DRM_SCHED_PRIORITY_KERNEL);
      if (r)
          return r;
  -    r = drm_sched_entity_init(&vm->delayed,
 adev->vm_manager.vm_pte_rqs,
-                  adev->vm_manager.vm_pte_num_rqs, NULL);
+    r = drm_sched_entity_init(&vm->delayed, sched_list,
+                  adev->vm_manager.vm_pte_num_rqs,
+                  NULL, DRM_SCHED_PRIORITY_KERNEL);
      if (r)
          goto error_free_direct;

diff --git a/drivers/gpu/drm/etnaviv/etnaviv_drv.c
b/drivers/gpu/drm/etnaviv/etnaviv_drv.c index 1f9c01be40d7..a65c1e115e35 100644

drivers/gpu/drm/etnaviv/etnaviv_drv.c +4 -4
@@ -65,12 +65,12 @@ static int etnaviv_open(struct drm_device *dev,
 struct drm_file *file)
        for (i = 0; i < ETNA_MAX_PIPES; i++) {
          struct etnaviv_gpu *gpu = priv->gpu[i];
-        struct drm_sched_rq *rq;
+        struct drm_gpu_scheduler *sched;
            if (gpu) {
-            rq = &gpu->sched.sched_rq[DRM_SCHED_PRIORITY_NORMAL];
-            drm_sched_entity_init(&ctx->sched_entity[i],
-                          &rq, 1, NULL);
+            sched = &gpu->sched;
+            drm_sched_entity_init(&ctx->sched_entity[i], &sched,
+                          1, NULL, DRM_SCHED_PRIORITY_NORMAL);
              }
      }
  diff --git a/drivers/gpu/drm/lima/lima_sched.c

b/drivers/gpu/drm/lima/lima_sched.c
index f522c5f99729..a7e53878d841 100644

drivers/gpu/drm/lima/lima_sched.c +3 -1
@@ -159,9 +159,10 @@ int lima_sched_context_init(struct
 lima_sched_pipe *pipe,
                  struct lima_sched_context *context,
                  atomic_t *guilty)
  {
-    struct drm_sched_rq *rq = pipe->base.sched_rq +
 DRM_SCHED_PRIORITY_NORMAL;
+    struct drm_gpu_scheduler *sched = &pipe->base;
  -    return drm_sched_entity_init(&context->base, &rq, 1, guilty);
+    return drm_sched_entity_init(&context->base, &sched,
+                     1, guilty, DRM_SCHED_PRIORITY_NORMAL);
  }
    void lima_sched_context_fini(struct lima_sched_pipe *pipe,
drivers/gpu/drm/panfrost/panfrost_job.c +4 -3
index d411eb6c8eb9..84178bcf35c9 100644
@@ -542,12 +542,13 @@ int panfrost_job_open(struct panfrost_file_priv
 *panfrost_priv)
  {
      struct panfrost_device *pfdev = panfrost_priv->pfdev;
      struct panfrost_job_slot *js = pfdev->js;
-    struct drm_sched_rq *rq;
+    struct drm_gpu_scheduler *sched;
      int ret, i;
        for (i = 0; i < NUM_JOB_SLOTS; i++) {
-        rq = &js->queue[i].sched.sched_rq[DRM_SCHED_PRIORITY_NORMAL];
-        ret = drm_sched_entity_init(&panfrost_priv->sched_entity[i],
 &rq, 1, NULL);
+        sched = &js->queue[i].sched;
+        ret = drm_sched_entity_init(&panfrost_priv->sched_entity[i],
+                        &sched, 1, NULL, DRM_SCHED_PRIORITY_NORMAL);
          if (WARN_ON(ret))
              return ret;
      }
drivers/gpu/drm/scheduler/sched_entity.c +16 -11
index 461a7a8129f4..e10d37266836 100644
@@ -38,9 +38,9 @@
   * submit to HW ring.
   *
   * @entity: scheduler entity to init
- * @rq_list: the list of run queue on which jobs from this
+ * @sched_list: the list of drm scheds on which jobs from this
   *           entity can be submitted
- * @num_rq_list: number of run queue in rq_list
+ * @num_sched_list: number of drm sched in sched_list
   * @guilty: atomic_t set to 1 when a job on this queue
   *          is found to be guilty causing a timeout
   *
@@ -50,32 +50,34 @@
   * Returns 0 on success or a negative error code on failure.
   */
  int drm_sched_entity_init(struct drm_sched_entity *entity,
-              struct drm_sched_rq **rq_list,
-              unsigned int num_rq_list,
-              atomic_t *guilty)
+              struct drm_gpu_scheduler **sched_list,
+              unsigned int num_sched_list,
+              atomic_t *guilty, enum drm_sched_priority priority)
  {
      int i;
  -    if (!(entity && rq_list && (num_rq_list == 0 || rq_list[0])))
+    if (!(entity && sched_list && (num_sched_list == 0 ||
 sched_list[0])))
          return -EINVAL;
        memset(entity, 0, sizeof(struct drm_sched_entity));
      INIT_LIST_HEAD(&entity->list);
      entity->rq = NULL;
      entity->guilty = guilty;
-    entity->num_rq_list = num_rq_list;
-    entity->rq_list = kcalloc(num_rq_list, sizeof(struct
 drm_sched_rq *),
-                GFP_KERNEL);
-    if (!entity->rq_list)
+    entity->num_sched_list = num_sched_list;
+    entity->priority = priority;
+    entity->sched_list =  kcalloc(num_sched_list,
+                      sizeof(struct drm_gpu_scheduler *), GFP_KERNEL);
+
+    if(!entity->sched_list)
          return -ENOMEM;
        init_completion(&entity->entity_idle);
  -    for (i = 0; i < num_rq_list; ++i)
-        entity->rq_list[i] = rq_list[i];
+    for (i = 0; i < num_sched_list; i++)
+        entity->sched_list[i] = sched_list[i];
  -    if (num_rq_list)
-        entity->rq = rq_list[0];
+    if (num_sched_list)
+        entity->rq =
 &entity->sched_list[0]->sched_rq[entity->priority];
        entity->last_scheduled = NULL;
  @@ -139,10 +141,10 @@ drm_sched_entity_get_free_sched(struct

drm_sched_entity *entity)

      unsigned int min_jobs = UINT_MAX, num_jobs;
      int i;

- for (i = 0; i < entity->num_rq_list; ++i) {

drivers/gpu/drm/scheduler/sched_entity.c +4 -2
-        struct drm_gpu_scheduler *sched = entity->rq_list[i]->sched;
+    for (i = 0; i < entity->num_sched_list; ++i) {
+        struct drm_gpu_scheduler *sched = entity->sched_list[i];
  -        if (!entity->rq_list[i]->sched->ready) {
+        if (!entity->sched_list[i]->ready) {
              DRM_WARN("sched%s is not ready, skipping", sched->name);
              continue;
          }
@@ -150,7 +152,7 @@ drm_sched_entity_get_free_sched(struct
 drm_sched_entity *entity)
          num_jobs = atomic_read(&sched->num_jobs);
          if (num_jobs < min_jobs) {
              min_jobs = num_jobs;
-            rq = entity->rq_list[i];
+            rq = &entity->sched_list[i]->sched_rq[entity->priority];
          }
      }

@@ -308,7 +310,7 @@ void drm_sched_entity_fini(struct
drm_sched_entity *entity)

        dma_fence_put(entity->last_scheduled);
      entity->last_scheduled = NULL;
drivers/gpu/drm/scheduler/sched_entity.c +1 -1
-    kfree(entity->rq_list);
+    kfree(entity->sched_list);
  }
  EXPORT_SYMBOL(drm_sched_entity_fini);
  @@ -353,15 +355,6 @@ static void drm_sched_entity_wakeup(struct

dma_fence *f,

      drm_sched_wakeup(entity->rq->sched);

}
-/**
- * drm_sched_entity_set_rq_priority - helper for
drm_sched_entity_set_priority
- */

drivers/gpu/drm/scheduler/sched_entity.c +1 -18
-static void drm_sched_entity_set_rq_priority(struct drm_sched_rq **rq,
-                         enum drm_sched_priority priority)
-{
-    *rq = &(*rq)->sched->sched_rq[priority];
-}
-
  /**
   * drm_sched_entity_set_priority - Sets priority of the entity
   *
@@ -373,20 +366,8 @@ static void
 drm_sched_entity_set_rq_priority(struct drm_sched_rq **rq,
  void drm_sched_entity_set_priority(struct drm_sched_entity *entity,
                     enum drm_sched_priority priority)
  {
-    unsigned int i;
-
-    spin_lock(&entity->rq_lock);
  -    for (i = 0; i < entity->num_rq_list; ++i)
- drm_sched_entity_set_rq_priority(&entity->rq_list[i], priority);
-
-    if (entity->rq) {
-        drm_sched_rq_remove_entity(entity->rq, entity);
-        drm_sched_entity_set_rq_priority(&entity->rq, priority);
-        drm_sched_rq_add_entity(entity->rq, entity);
-    }
-
-    spin_unlock(&entity->rq_lock);
+    entity->priority = priority;
  }
  EXPORT_SYMBOL(drm_sched_entity_set_priority);
  @@ -490,7 +471,7 @@ void drm_sched_entity_select_rq(struct

drm_sched_entity *entity)

      struct dma_fence *fence;
      struct drm_sched_rq *rq;

- if (spsc_queue_count(&entity->job_queue) ||
entity->num_rq_list <= 1)
+ if (spsc_queue_count(&entity->job_queue) ||
entity->num_sched_list <= 1)

          return;
        fence = READ_ONCE(entity->last_scheduled);
drivers/gpu/drm/v3d/v3d_drv.c +4 -3
index 1a07462b4528..c6aff1aedd27 100644
@@ -140,7 +140,7 @@ v3d_open(struct drm_device *dev, struct drm_file
 *file)
  {
      struct v3d_dev *v3d = to_v3d_dev(dev);
      struct v3d_file_priv *v3d_priv;
-    struct drm_sched_rq *rq;
+    struct drm_gpu_scheduler *sched;
      int i;
        v3d_priv = kzalloc(sizeof(*v3d_priv), GFP_KERNEL);
@@ -150,8 +150,9 @@ v3d_open(struct drm_device *dev, struct drm_file
 *file)
      v3d_priv->v3d = v3d;
        for (i = 0; i < V3D_MAX_QUEUES; i++) {
-        rq = &v3d->queue[i].sched.sched_rq[DRM_SCHED_PRIORITY_NORMAL];
- drm_sched_entity_init(&v3d_priv->sched_entity[i], &rq, 1, NULL);
+        sched = &v3d->queue[i].sched;
+ drm_sched_entity_init(&v3d_priv->sched_entity[i], &sched,
+                      1, NULL, DRM_SCHED_PRIORITY_NORMAL);
      }
        file->driver_priv = v3d_priv;
include/drm/gpu_scheduler.h +3 -3
index 684692a8ed76..9df322dfac30 100644
@@ -81,8 +81,9 @@ enum drm_sched_priority {
  struct drm_sched_entity {
      struct list_head        list;
      struct drm_sched_rq        *rq;
-    struct drm_sched_rq        **rq_list;
-    unsigned int                    num_rq_list;
+    unsigned int                    num_sched_list;
+    struct drm_gpu_scheduler        **sched_list;
+    enum drm_sched_priority         priority;
      spinlock_t            rq_lock;
        struct spsc_queue        job_queue;
@@ -312,9 +313,9 @@ void drm_sched_rq_remove_entity(struct
 drm_sched_rq *rq,
                  struct drm_sched_entity *entity);
    int drm_sched_entity_init(struct drm_sched_entity *entity,
-              struct drm_sched_rq **rq_list,
Quoted 8 lines
Quoted 8 lines
include/drm/gpu_scheduler.h +2 -1
+              struct drm_gpu_scheduler **sched_list,
                unsigned int num_rq_list,
-              atomic_t *guilty);
+              atomic_t *guilty, enum drm_sched_priority priority);
  long drm_sched_entity_flush(struct drm_sched_entity *entity, long
 timeout);
  void drm_sched_entity_fini(struct drm_sched_entity *entity);
  void drm_sched_entity_destroy(struct drm_sched_entity *entity);

_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx

maintainer
#4 lore raw
In reply to Nirmoy Das

Am 05.12.19 um 11:52 schrieb Nirmoy Das:

Quoted 5 lines

Entity currently keeps a copy of run_queue list and modify it in drm_sched_entity_set_priority(). Entities shouldn't modify run_queue list. Use drm_gpu_scheduler list instead of drm_sched_rq list in drm_sched_entity struct. In this way we can select a runqueue based on entity/ctx's priority for a drm scheduler.

Looks good to me in general only a few nit picks below.

Quoted 120 lines
---
  drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c  |  7 +--
  drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c  |  7 +--
  drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c  |  7 +--
  drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c  |  7 +--
  drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c   | 14 +++--
  drivers/gpu/drm/etnaviv/etnaviv_drv.c    |  8 +--
  drivers/gpu/drm/lima/lima_sched.c        |  5 +-
  drivers/gpu/drm/panfrost/panfrost_job.c  |  7 +--
  drivers/gpu/drm/scheduler/sched_entity.c | 65 +++++++++---------------
  drivers/gpu/drm/v3d/v3d_drv.c            |  7 +--
  include/drm/gpu_scheduler.h              |  9 ++--
  11 files changed, 69 insertions(+), 74 deletions(-)
drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c +4 -3
index a0d3d7b756eb..e8f46c13d073 100644
@@ -122,7 +122,7 @@ static int amdgpu_ctx_init(struct amdgpu_device *adev,
 
  	for (i = 0; i < AMDGPU_HW_IP_NUM; ++i) {
  		struct amdgpu_ring *rings[AMDGPU_MAX_RINGS];
-		struct drm_sched_rq *rqs[AMDGPU_MAX_RINGS];
+		struct drm_gpu_scheduler *sched_list[AMDGPU_MAX_RINGS];
  		unsigned num_rings = 0;
  		unsigned num_rqs = 0;
 
@@ -181,12 +181,13 @@ static int amdgpu_ctx_init(struct amdgpu_device *adev,
  			if (!rings[j]->adev)
  				continue;
 
-			rqs[num_rqs++] = &rings[j]->sched.sched_rq[priority];
+			sched_list[num_rqs++] = &rings[j]->sched;
  		}
 
  		for (j = 0; j < amdgpu_ctx_num_entities[i]; ++j)
  			r = drm_sched_entity_init(&ctx->entities[i][j].entity,
-						  rqs, num_rqs, &ctx->guilty);
+						  sched_list, num_rqs,
+						  &ctx->guilty, priority);
  		if (r)
  			goto error_cleanup_entities;
  	}
drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c +4 -3
index 19ffe00d9072..a960dd7c0711 100644
@@ -1957,11 +1957,12 @@ void amdgpu_ttm_set_buffer_funcs_status(struct amdgpu_device *adev, bool enable)
 
  	if (enable) {
  		struct amdgpu_ring *ring;
-		struct drm_sched_rq *rq;
+		struct drm_gpu_scheduler *sched;
 
  		ring = adev->mman.buffer_funcs_ring;
-		rq = &ring->sched.sched_rq[DRM_SCHED_PRIORITY_KERNEL];
-		r = drm_sched_entity_init(&adev->mman.entity, &rq, 1, NULL);
+		sched = &ring->sched;
+		r = drm_sched_entity_init(&adev->mman.entity, &sched,
+					  1, NULL, DRM_SCHED_PRIORITY_KERNEL);
  		if (r) {
  			DRM_ERROR("Failed setting up TTM BO move entity (%d)\n",
  				  r);
drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c +4 -3
index e324bfe6c58f..b803a8882864 100644
@@ -330,12 +330,13 @@ int amdgpu_uvd_sw_fini(struct amdgpu_device *adev)
  int amdgpu_uvd_entity_init(struct amdgpu_device *adev)
  {
  	struct amdgpu_ring *ring;
-	struct drm_sched_rq *rq;
+	struct drm_gpu_scheduler *sched;
  	int r;
 
  	ring = &adev->uvd.inst[0].ring;
-	rq = &ring->sched.sched_rq[DRM_SCHED_PRIORITY_NORMAL];
-	r = drm_sched_entity_init(&adev->uvd.entity, &rq, 1, NULL);
+	sched = &ring->sched;
+	r = drm_sched_entity_init(&adev->uvd.entity, &sched,
+				  1, NULL, DRM_SCHED_PRIORITY_NORMAL);
  	if (r) {
  		DRM_ERROR("Failed setting up UVD kernel entity.\n");
  		return r;
drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c +4 -3
index 46b590af2fd2..b44f28d44fb4 100644
@@ -240,12 +240,13 @@ int amdgpu_vce_sw_fini(struct amdgpu_device *adev)
  int amdgpu_vce_entity_init(struct amdgpu_device *adev)
  {
  	struct amdgpu_ring *ring;
-	struct drm_sched_rq *rq;
+	struct drm_gpu_scheduler *sched;
  	int r;
 
  	ring = &adev->vce.ring[0];
-	rq = &ring->sched.sched_rq[DRM_SCHED_PRIORITY_NORMAL];
-	r = drm_sched_entity_init(&adev->vce.entity, &rq, 1, NULL);
+	sched = &ring->sched;
+	r = drm_sched_entity_init(&adev->vce.entity, &sched,
+				  1, NULL, DRM_SCHED_PRIORITY_NORMAL);
  	if (r != 0) {
  		DRM_ERROR("Failed setting up VCE run queue.\n");
  		return r;
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +1 -0
index a94c4faa5af1..ec6141773a92 100644
@@ -2687,6 +2687,7 @@ int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm,
  {
  	struct amdgpu_bo_param bp;
  	struct amdgpu_bo *root;
+        struct drm_gpu_scheduler *sched_list[AMDGPU_MAX_RINGS];
  	int r, i;
 
  	vm->va = RB_ROOT_CACHED;

drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm, · line 2706

Quoted 6 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +3 -0
@@ -2700,14 +2701,19 @@ int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm,
  	spin_lock_init(&vm->invalidated_lock);
  	INIT_LIST_HEAD(&vm->freed);
 
+        for (i = 0; i < adev->vm_manager.vm_pte_num_rqs; i++)
+	       sched_list[i] = adev->vm_manager.vm_pte_rqs[i]->sched;
+

It would be better to turn the vm_pte_rqs array into a vm_pte_sched array instead.

But can be done in a follow up cleanup as well.

Quoted 93 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +6 -4
  	/* create scheduler entities for page table updates */
-	r = drm_sched_entity_init(&vm->direct, adev->vm_manager.vm_pte_rqs,
-				  adev->vm_manager.vm_pte_num_rqs, NULL);
+	r = drm_sched_entity_init(&vm->direct, sched_list,
+				  adev->vm_manager.vm_pte_num_rqs,
+				  NULL, DRM_SCHED_PRIORITY_KERNEL);
  	if (r)
  		return r;
 
-	r = drm_sched_entity_init(&vm->delayed, adev->vm_manager.vm_pte_rqs,
-				  adev->vm_manager.vm_pte_num_rqs, NULL);
+	r = drm_sched_entity_init(&vm->delayed, sched_list,
+				  adev->vm_manager.vm_pte_num_rqs,
+				  NULL, DRM_SCHED_PRIORITY_KERNEL);
  	if (r)
  		goto error_free_direct;
 
drivers/gpu/drm/etnaviv/etnaviv_drv.c +4 -4
index 1f9c01be40d7..a65c1e115e35 100644
@@ -65,12 +65,12 @@ static int etnaviv_open(struct drm_device *dev, struct drm_file *file)
 
  	for (i = 0; i < ETNA_MAX_PIPES; i++) {
  		struct etnaviv_gpu *gpu = priv->gpu[i];
-		struct drm_sched_rq *rq;
+		struct drm_gpu_scheduler *sched;
 
  		if (gpu) {
-			rq = &gpu->sched.sched_rq[DRM_SCHED_PRIORITY_NORMAL];
-			drm_sched_entity_init(&ctx->sched_entity[i],
-					      &rq, 1, NULL);
+			sched = &gpu->sched;
+			drm_sched_entity_init(&ctx->sched_entity[i], &sched,
+					      1, NULL, DRM_SCHED_PRIORITY_NORMAL);
  			}
  	}
 
drivers/gpu/drm/lima/lima_sched.c +3 -2
index f522c5f99729..a7e53878d841 100644
@@ -159,9 +159,10 @@ int lima_sched_context_init(struct lima_sched_pipe *pipe,
  			    struct lima_sched_context *context,
  			    atomic_t *guilty)
  {
-	struct drm_sched_rq *rq = pipe->base.sched_rq + DRM_SCHED_PRIORITY_NORMAL;
+	struct drm_gpu_scheduler *sched = &pipe->base;
 
-	return drm_sched_entity_init(&context->base, &rq, 1, guilty);
+	return drm_sched_entity_init(&context->base, &sched,
+				     1, guilty, DRM_SCHED_PRIORITY_NORMAL);
  }
 
  void lima_sched_context_fini(struct lima_sched_pipe *pipe,
drivers/gpu/drm/panfrost/panfrost_job.c +4 -3
index d411eb6c8eb9..84178bcf35c9 100644
@@ -542,12 +542,13 @@ int panfrost_job_open(struct panfrost_file_priv *panfrost_priv)
  {
  	struct panfrost_device *pfdev = panfrost_priv->pfdev;
  	struct panfrost_job_slot *js = pfdev->js;
-	struct drm_sched_rq *rq;
+	struct drm_gpu_scheduler *sched;
  	int ret, i;
 
  	for (i = 0; i < NUM_JOB_SLOTS; i++) {
-		rq = &js->queue[i].sched.sched_rq[DRM_SCHED_PRIORITY_NORMAL];
-		ret = drm_sched_entity_init(&panfrost_priv->sched_entity[i], &rq, 1, NULL);
+		sched = &js->queue[i].sched;
+		ret = drm_sched_entity_init(&panfrost_priv->sched_entity[i],
+					    &sched, 1, NULL, DRM_SCHED_PRIORITY_NORMAL);
  		if (WARN_ON(ret))
  			return ret;
  	}
drivers/gpu/drm/scheduler/sched_entity.c +2 -2
index 461a7a8129f4..e10d37266836 100644
@@ -38,9 +38,9 @@
   * submit to HW ring.
   *
   * @entity: scheduler entity to init
- * @rq_list: the list of run queue on which jobs from this
+ * @sched_list: the list of drm scheds on which jobs from this
   *           entity can be submitted
- * @num_rq_list: number of run queue in rq_list
+ * @num_sched_list: number of drm sched in sched_list
   * @guilty: atomic_t set to 1 when a job on this queue
   *          is found to be guilty causing a timeout
   *

drivers/gpu/drm/scheduler/sched_entity.c line 55

Quoted 9 lines
drivers/gpu/drm/scheduler/sched_entity.c +3 -3
@@ -50,32 +50,34 @@
   * Returns 0 on success or a negative error code on failure.
   */
  int drm_sched_entity_init(struct drm_sched_entity *entity,
-			  struct drm_sched_rq **rq_list,
-			  unsigned int num_rq_list,
-			  atomic_t *guilty)
+			  struct drm_gpu_scheduler **sched_list,
+			  unsigned int num_sched_list,
+			  atomic_t *guilty, enum drm_sched_priority priority)

The priority needs to be mentioned in the kerneldoc as well and maybe add the parameter as the first one (after entity).

Quoted 9 lines
drivers/gpu/drm/scheduler/sched_entity.c +1 -1
  {
  	int i;
 
-	if (!(entity && rq_list && (num_rq_list == 0 || rq_list[0])))
+	if (!(entity && sched_list && (num_sched_list == 0 || sched_list[0])))
  		return -EINVAL;
 
  	memset(entity, 0, sizeof(struct drm_sched_entity));
  	INIT_LIST_HEAD(&entity->list);

drivers/gpu/drm/scheduler/sched_entity.c line 69

Quoted 10 lines
drivers/gpu/drm/scheduler/sched_entity.c +4 -4
  	entity->rq = NULL;
  	entity->guilty = guilty;
-	entity->num_rq_list = num_rq_list;
-	entity->rq_list = kcalloc(num_rq_list, sizeof(struct drm_sched_rq *),
-				GFP_KERNEL);
-	if (!entity->rq_list)
+	entity->num_sched_list = num_sched_list;
+	entity->priority = priority;
+	entity->sched_list =  kcalloc(num_sched_list,
+				      sizeof(struct drm_gpu_scheduler *), GFP_KERNEL);

As a follow up we want to get rid of this as well, but keep it for now to make the patch simpler.

Quoted 75 lines
drivers/gpu/drm/scheduler/sched_entity.c +11 -22
+
+	if(!entity->sched_list)
  		return -ENOMEM;
 
  	init_completion(&entity->entity_idle);
 
-	for (i = 0; i < num_rq_list; ++i)
-		entity->rq_list[i] = rq_list[i];
+	for (i = 0; i < num_sched_list; i++)
+		entity->sched_list[i] = sched_list[i];
 
-	if (num_rq_list)
-		entity->rq = rq_list[0];
+	if (num_sched_list)
+		entity->rq = &entity->sched_list[0]->sched_rq[entity->priority];
 
  	entity->last_scheduled = NULL;
 
@@ -139,10 +141,10 @@ drm_sched_entity_get_free_sched(struct drm_sched_entity *entity)
  	unsigned int min_jobs = UINT_MAX, num_jobs;
  	int i;
 
-	for (i = 0; i < entity->num_rq_list; ++i) {
-		struct drm_gpu_scheduler *sched = entity->rq_list[i]->sched;
+	for (i = 0; i < entity->num_sched_list; ++i) {
+		struct drm_gpu_scheduler *sched = entity->sched_list[i];
 
-		if (!entity->rq_list[i]->sched->ready) {
+		if (!entity->sched_list[i]->ready) {
  			DRM_WARN("sched%s is not ready, skipping", sched->name);
  			continue;
  		}
@@ -150,7 +152,7 @@ drm_sched_entity_get_free_sched(struct drm_sched_entity *entity)
  		num_jobs = atomic_read(&sched->num_jobs);
  		if (num_jobs < min_jobs) {
  			min_jobs = num_jobs;
-			rq = entity->rq_list[i];
+			rq = &entity->sched_list[i]->sched_rq[entity->priority];
  		}
  	}
 
@@ -308,7 +310,7 @@ void drm_sched_entity_fini(struct drm_sched_entity *entity)
 
  	dma_fence_put(entity->last_scheduled);
  	entity->last_scheduled = NULL;
-	kfree(entity->rq_list);
+	kfree(entity->sched_list);
  }
  EXPORT_SYMBOL(drm_sched_entity_fini);
 
@@ -353,15 +355,6 @@ static void drm_sched_entity_wakeup(struct dma_fence *f,
  	drm_sched_wakeup(entity->rq->sched);
  }
 
-/**
- * drm_sched_entity_set_rq_priority - helper for drm_sched_entity_set_priority
- */
-static void drm_sched_entity_set_rq_priority(struct drm_sched_rq **rq,
-					     enum drm_sched_priority priority)
-{
-	*rq = &(*rq)->sched->sched_rq[priority];
-}
-
  /**
   * drm_sched_entity_set_priority - Sets priority of the entity
   *
@@ -373,20 +366,8 @@ static void drm_sched_entity_set_rq_priority(struct drm_sched_rq **rq,
  void drm_sched_entity_set_priority(struct drm_sched_entity *entity,
  				   enum drm_sched_priority priority)
  {
-	unsigned int i;
-
-	spin_lock(&entity->rq_lock);
 
-	for (i = 0; i < entity->num_rq_list; ++i)

drivers/gpu/drm/scheduler/sched_entity.c static void drm_sched_entity_set_rq_priority(struct drm_sched_rq **rq, · line 370

Quoted 10 lines
drivers/gpu/drm/scheduler/sched_entity.c +1 -9
-		drm_sched_entity_set_rq_priority(&entity->rq_list[i], priority);
-
-	if (entity->rq) {
-		drm_sched_rq_remove_entity(entity->rq, entity);
-		drm_sched_entity_set_rq_priority(&entity->rq, priority);
-		drm_sched_rq_add_entity(entity->rq, entity);
-	}
-
-	spin_unlock(&entity->rq_lock);
+	entity->priority = priority;

You probably still want to grab the lock.

Regards,
Christian.

Quoted 55 lines
drivers/gpu/drm/scheduler/sched_entity.c +1 -1
  }
  EXPORT_SYMBOL(drm_sched_entity_set_priority);
 
@@ -490,7 +471,7 @@ void drm_sched_entity_select_rq(struct drm_sched_entity *entity)
  	struct dma_fence *fence;
  	struct drm_sched_rq *rq;
 
-	if (spsc_queue_count(&entity->job_queue) || entity->num_rq_list <= 1)
+	if (spsc_queue_count(&entity->job_queue) || entity->num_sched_list <= 1)
  		return;
 
  	fence = READ_ONCE(entity->last_scheduled);
drivers/gpu/drm/v3d/v3d_drv.c +4 -3
index 1a07462b4528..c6aff1aedd27 100644
@@ -140,7 +140,7 @@ v3d_open(struct drm_device *dev, struct drm_file *file)
  {
  	struct v3d_dev *v3d = to_v3d_dev(dev);
  	struct v3d_file_priv *v3d_priv;
-	struct drm_sched_rq *rq;
+	struct drm_gpu_scheduler *sched;
  	int i;
 
  	v3d_priv = kzalloc(sizeof(*v3d_priv), GFP_KERNEL);
@@ -150,8 +150,9 @@ v3d_open(struct drm_device *dev, struct drm_file *file)
  	v3d_priv->v3d = v3d;
 
  	for (i = 0; i < V3D_MAX_QUEUES; i++) {
-		rq = &v3d->queue[i].sched.sched_rq[DRM_SCHED_PRIORITY_NORMAL];
-		drm_sched_entity_init(&v3d_priv->sched_entity[i], &rq, 1, NULL);
+		sched = &v3d->queue[i].sched;
+		drm_sched_entity_init(&v3d_priv->sched_entity[i], &sched,
+				      1, NULL, DRM_SCHED_PRIORITY_NORMAL);
  	}
 
  	file->driver_priv = v3d_priv;
include/drm/gpu_scheduler.h +3 -2
index 684692a8ed76..9df322dfac30 100644
@@ -81,8 +81,9 @@ enum drm_sched_priority {
  struct drm_sched_entity {
  	struct list_head		list;
  	struct drm_sched_rq		*rq;
-	struct drm_sched_rq		**rq_list;
-	unsigned int                    num_rq_list;
+	unsigned int                    num_sched_list;
+	struct drm_gpu_scheduler        **sched_list;
+	enum drm_sched_priority         priority;
  	spinlock_t			rq_lock;
 
  	struct spsc_queue		job_queue;
@@ -312,9 +313,9 @@ void drm_sched_rq_remove_entity(struct drm_sched_rq *rq,
  				struct drm_sched_entity *entity);

include/drm/gpu_scheduler.h void drm_sched_rq_remove_entity(struct drm_sched_rq *rq, · line 321

Quoted 10 lines
include/drm/gpu_scheduler.h +2 -2
 
  int drm_sched_entity_init(struct drm_sched_entity *entity,
-			  struct drm_sched_rq **rq_list,
+			  struct drm_gpu_scheduler **sched_list,
  			  unsigned int num_rq_list,
-			  atomic_t *guilty);
+			  atomic_t *guilty, enum drm_sched_priority priority);
  long drm_sched_entity_flush(struct drm_sched_entity *entity, long timeout);
  void drm_sched_entity_fini(struct drm_sched_entity *entity);
  void drm_sched_entity_destroy(struct drm_sched_entity *entity);

_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx