[PATCH 1/2] drm/amdgpu: individualize amdgpu entity allocation per HW IP

[Nirmoy Das]

Do not allocate all the entity at once. This is required for
dynamic amdgpu_ctx_entity creation.

Signed-off-by: Nirmoy Das <nirmoy.das at amd.com>
---
 drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c | 143 +++++++++++++-----------
 1 file changed, 76 insertions(+), 67 deletions(-)

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c
index 05c2af61e7de..eecbd68db986 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c
@@ -42,16 +42,6 @@ const unsigned int amdgpu_ctx_num_entities[AMDGPU_HW_IP_NUM] = {
 	[AMDGPU_HW_IP_VCN_JPEG]	=	1,
 };
 
-static int amdgpu_ctx_total_num_entities(void)
-{
-	unsigned i, num_entities = 0;
-
-	for (i = 0; i < AMDGPU_HW_IP_NUM; ++i)
-		num_entities += amdgpu_ctx_num_entities[i];
-
-	return num_entities;
-}
-
 static int amdgpu_ctx_priority_permit(struct drm_file *filp,
 				      enum drm_sched_priority priority)
 {
@@ -73,7 +63,6 @@ static int amdgpu_ctx_init(struct amdgpu_device *adev,
 			   struct drm_file *filp,
 			   struct amdgpu_ctx *ctx)
 {
-	unsigned num_entities = amdgpu_ctx_total_num_entities();
 	unsigned i, j;
 	int r;
 
@@ -87,28 +76,30 @@ static int amdgpu_ctx_init(struct amdgpu_device *adev,
 	memset(ctx, 0, sizeof(*ctx));
 	ctx->adev = adev;
 
+	for (i = 0; i < AMDGPU_HW_IP_NUM; ++i) {
+		ctx->entities[i] = kcalloc(amdgpu_ctx_num_entities[i],
+					   sizeof(struct amdgpu_ctx_entity),
+					   GFP_KERNEL);
 
-	ctx->entities[0] = kcalloc(num_entities,
-				   sizeof(struct amdgpu_ctx_entity),
-				   GFP_KERNEL);
-	if (!ctx->entities[0])
-		return -ENOMEM;
-
-
-	for (i = 0; i < num_entities; ++i) {
-		struct amdgpu_ctx_entity *entity = &ctx->entities[0][i];
+		if (!ctx->entities[0]) {
+			r =  -ENOMEM;
+			goto error_cleanup_entity_memory;
+		}
+	}
 
-		entity->sequence = 1;
-		entity->fences = kcalloc(amdgpu_sched_jobs,
-					 sizeof(struct dma_fence*), GFP_KERNEL);
-		if (!entity->fences) {
-			r = -ENOMEM;
-			goto error_cleanup_memory;
+	for (i = 0; i < AMDGPU_HW_IP_NUM; ++i) {
+		for (j = 0; j < amdgpu_ctx_num_entities[i]; ++j) {
+			struct amdgpu_ctx_entity *entity = &ctx->entities[i][j];
+
+			entity->sequence = 1;
+			entity->fences = kcalloc(amdgpu_sched_jobs,
+						 sizeof(struct dma_fence*), GFP_KERNEL);
+			if (!entity->fences) {
+				r = -ENOMEM;
+				goto error_cleanup_memory;
+			}
 		}
 	}
-	for (i = 1; i < AMDGPU_HW_IP_NUM; ++i)
-		ctx->entities[i] = ctx->entities[i - 1] +
-			amdgpu_ctx_num_entities[i - 1];
 
 	kref_init(&ctx->refcount);
 	spin_lock_init(&ctx->ring_lock);
@@ -179,44 +170,55 @@ static int amdgpu_ctx_init(struct amdgpu_device *adev,
 	return 0;
 
 error_cleanup_entities:
-	for (i = 0; i < num_entities; ++i)
-		drm_sched_entity_destroy(&ctx->entities[0][i].entity);
+	for (i = 0; i < AMDGPU_HW_IP_NUM; ++i) {
+		for (j = 0; j < amdgpu_ctx_num_entities[i]; ++j)
+			drm_sched_entity_destroy(&ctx->entities[i][j].entity);
+	}
 
 error_cleanup_memory:
-	for (i = 0; i < num_entities; ++i) {
-		struct amdgpu_ctx_entity *entity = &ctx->entities[0][i];
+	for (i = 0; i < AMDGPU_HW_IP_NUM; ++i) {
+		for (j = 0; j < amdgpu_ctx_num_entities[i]; ++j) {
+			struct amdgpu_ctx_entity *entity = &ctx->entities[i][j];
 
-		kfree(entity->fences);
-		entity->fences = NULL;
+			kfree(entity->fences);
+			entity->fences = NULL;
+		}
+	}
+
+error_cleanup_entity_memory:
+	for (i = 0; i < AMDGPU_HW_IP_NUM; ++i) {
+		kfree(ctx->entities[i]);
+		ctx->entities[i] = NULL;
 	}
 
-	kfree(ctx->entities[0]);
-	ctx->entities[0] = NULL;
 	return r;
 }
 
 static void amdgpu_ctx_fini(struct kref *ref)
 {
 	struct amdgpu_ctx *ctx = container_of(ref, struct amdgpu_ctx, refcount);
-	unsigned num_entities = amdgpu_ctx_total_num_entities();
 	struct amdgpu_device *adev = ctx->adev;
-	unsigned i, j;
+	unsigned i, j, k;
 
 	if (!adev)
 		return;
+	for (i = 0; i < AMDGPU_HW_IP_NUM; ++i) {
+		for (j = 0; j < amdgpu_ctx_num_entities[i]; ++j) {
+			struct amdgpu_ctx_entity *entity = &ctx->entities[i][j];
 
-	for (i = 0; i < num_entities; ++i) {
-		struct amdgpu_ctx_entity *entity = &ctx->entities[0][i];
+			for (k = 0; k < amdgpu_sched_jobs; ++k)
+				dma_fence_put(entity->fences[k]);
 
-		for (j = 0; j < amdgpu_sched_jobs; ++j)
-			dma_fence_put(entity->fences[j]);
+			kfree(entity->fences);
+		}
+	}
 
-		kfree(entity->fences);
+	for (i = 0; i < AMDGPU_HW_IP_NUM; ++i) {
+		kfree(ctx->entities[i]);
+		ctx->entities[i] = NULL;
 	}
 
-	kfree(ctx->entities[0]);
 	mutex_destroy(&ctx->lock);
-
 	kfree(ctx);
 }
 
@@ -279,14 +281,13 @@ static int amdgpu_ctx_alloc(struct amdgpu_device *adev,
 static void amdgpu_ctx_do_release(struct kref *ref)
 {
 	struct amdgpu_ctx *ctx;
-	unsigned num_entities;
-	u32 i;
+	u32 i, j;
 
 	ctx = container_of(ref, struct amdgpu_ctx, refcount);
-
-	num_entities = amdgpu_ctx_total_num_entities();
-	for (i = 0; i < num_entities; i++)
-		drm_sched_entity_destroy(&ctx->entities[0][i].entity);
+	for (i = 0; i < AMDGPU_HW_IP_NUM; ++i) {
+		for (j = 0; j < amdgpu_ctx_num_entities[i]; ++j)
+			drm_sched_entity_destroy(&ctx->entities[i][j].entity);
+	}
 
 	amdgpu_ctx_fini(ref);
 }
@@ -516,19 +517,21 @@ struct dma_fence *amdgpu_ctx_get_fence(struct amdgpu_ctx *ctx,
 void amdgpu_ctx_priority_override(struct amdgpu_ctx *ctx,
 				  enum drm_sched_priority priority)
 {
-	unsigned num_entities = amdgpu_ctx_total_num_entities();
 	enum drm_sched_priority ctx_prio;
-	unsigned i;
+	unsigned i, j;
 
 	ctx->override_priority = priority;
 
 	ctx_prio = (ctx->override_priority == DRM_SCHED_PRIORITY_UNSET) ?
 			ctx->init_priority : ctx->override_priority;
+	for (i = 0; i < AMDGPU_HW_IP_NUM; ++i) {
+		for (j = 0; j < amdgpu_ctx_num_entities[i]; ++j) {
 
-	for (i = 0; i < num_entities; i++) {
-		struct drm_sched_entity *entity = &ctx->entities[0][i].entity;
+			struct drm_sched_entity *entity =
+				&ctx->entities[i][j].entity;
 
-		drm_sched_entity_set_priority(entity, ctx_prio);
+			drm_sched_entity_set_priority(entity, ctx_prio);
+		}
 	}
 }
 
@@ -564,20 +567,21 @@ void amdgpu_ctx_mgr_init(struct amdgpu_ctx_mgr *mgr)
 
 long amdgpu_ctx_mgr_entity_flush(struct amdgpu_ctx_mgr *mgr, long timeout)
 {
-	unsigned num_entities = amdgpu_ctx_total_num_entities();
 	struct amdgpu_ctx *ctx;
 	struct idr *idp;
-	uint32_t id, i;
+	uint32_t id, i, j;
 
 	idp = &mgr->ctx_handles;
 
 	mutex_lock(&mgr->lock);
 	idr_for_each_entry(idp, ctx, id) {
-		for (i = 0; i < num_entities; i++) {
-			struct drm_sched_entity *entity;
+		for (i = 0; i < AMDGPU_HW_IP_NUM; ++i) {
+			for (j = 0; j < amdgpu_ctx_num_entities[i]; ++j) {
+				struct drm_sched_entity *entity;
 
-			entity = &ctx->entities[0][i].entity;
-			timeout = drm_sched_entity_flush(entity, timeout);
+				entity = &ctx->entities[i][j].entity;
+				timeout = drm_sched_entity_flush(entity, timeout);
+			}
 		}
 	}
 	mutex_unlock(&mgr->lock);
@@ -586,10 +590,9 @@ long amdgpu_ctx_mgr_entity_flush(struct amdgpu_ctx_mgr *mgr, long timeout)
 
 void amdgpu_ctx_mgr_entity_fini(struct amdgpu_ctx_mgr *mgr)
 {
-	unsigned num_entities = amdgpu_ctx_total_num_entities();
 	struct amdgpu_ctx *ctx;
 	struct idr *idp;
-	uint32_t id, i;
+	uint32_t id, i, j;
 
 	idp = &mgr->ctx_handles;
 
@@ -599,8 +602,14 @@ void amdgpu_ctx_mgr_entity_fini(struct amdgpu_ctx_mgr *mgr)
 			continue;
 		}
 
-		for (i = 0; i < num_entities; i++)
-			drm_sched_entity_fini(&ctx->entities[0][i].entity);
+		for (i = 0; i < AMDGPU_HW_IP_NUM; ++i) {
+			for (j = 0; j < amdgpu_ctx_num_entities[i]; ++j) {
+				struct drm_sched_entity *entity;
+
+				entity = &ctx->entities[i][j].entity;
+				drm_sched_entity_fini(entity);
+			}
+		}
 	}
 }
 
-- 
2.24.1