1 files changed, 164 insertions, 124 deletions

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c
index 5b3e0f63a115..aeee6840e82b 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c
@@ -27,81 +27,10 @@
 #include <linux/pagemap.h>
 #include <drm/drmP.h>
 #include <drm/amdgpu_drm.h>
+#include <drm/drm_syncobj.h>
 #include "amdgpu.h"
 #include "amdgpu_trace.h"
 
-int amdgpu_cs_get_ring(struct amdgpu_device *adev, u32 ip_type,
-		       u32 ip_instance, u32 ring,
-		       struct amdgpu_ring **out_ring)
-{
-	/* Right now all IPs have only one instance - multiple rings. */
-	if (ip_instance != 0) {
-		DRM_ERROR("invalid ip instance: %d\n", ip_instance);
-		return -EINVAL;
-	}
-
-	switch (ip_type) {
-	default:
-		DRM_ERROR("unknown ip type: %d\n", ip_type);
-		return -EINVAL;
-	case AMDGPU_HW_IP_GFX:
-		if (ring < adev->gfx.num_gfx_rings) {
-			*out_ring = &adev->gfx.gfx_ring[ring];
-		} else {
-			DRM_ERROR("only %d gfx rings are supported now\n",
-				  adev->gfx.num_gfx_rings);
-			return -EINVAL;
-		}
-		break;
-	case AMDGPU_HW_IP_COMPUTE:
-		if (ring < adev->gfx.num_compute_rings) {
-			*out_ring = &adev->gfx.compute_ring[ring];
-		} else {
-			DRM_ERROR("only %d compute rings are supported now\n",
-				  adev->gfx.num_compute_rings);
-			return -EINVAL;
-		}
-		break;
-	case AMDGPU_HW_IP_DMA:
-		if (ring < adev->sdma.num_instances) {
-			*out_ring = &adev->sdma.instance[ring].ring;
-		} else {
-			DRM_ERROR("only %d SDMA rings are supported\n",
-				  adev->sdma.num_instances);
-			return -EINVAL;
-		}
-		break;
-	case AMDGPU_HW_IP_UVD:
-		*out_ring = &adev->uvd.ring;
-		break;
-	case AMDGPU_HW_IP_VCE:
-		if (ring < adev->vce.num_rings){
-			*out_ring = &adev->vce.ring[ring];
-		} else {
-			DRM_ERROR("only %d VCE rings are supported\n", adev->vce.num_rings);
-			return -EINVAL;
-		}
-		break;
-	case AMDGPU_HW_IP_UVD_ENC:
-		if (ring < adev->uvd.num_enc_rings){
-			*out_ring = &adev->uvd.ring_enc[ring];
-		} else {
-			DRM_ERROR("only %d UVD ENC rings are supported\n",
-				adev->uvd.num_enc_rings);
-			return -EINVAL;
-		}
-		break;
-	}
-
-	if (!(*out_ring && (*out_ring)->adev)) {
-		DRM_ERROR("Ring %d is not initialized on IP %d\n",
-			  ring, ip_type);
-		return -EINVAL;
-	}
-
-	return 0;
-}
-
 static int amdgpu_cs_user_fence_chunk(struct amdgpu_cs_parser *p,
 				      struct drm_amdgpu_cs_chunk_fence *data,
 				      uint32_t *offset)
@@ -226,6 +155,8 @@ int amdgpu_cs_parser_init(struct amdgpu_cs_parser *p, void *data)
 			break;
 
 		case AMDGPU_CHUNK_ID_DEPENDENCIES:
+		case AMDGPU_CHUNK_ID_SYNCOBJ_IN:
+		case AMDGPU_CHUNK_ID_SYNCOBJ_OUT:
 			break;
 
 		default:
@@ -597,7 +528,7 @@ static int amdgpu_cs_parser_bos(struct amdgpu_cs_parser *p,
 			goto error_free_pages;
 		}
 
-		/* Fill the page arrays for all useptrs. */
+		/* Fill the page arrays for all userptrs. */
 		list_for_each_entry(e, &need_pages, tv.head) {
 			struct ttm_tt *ttm = e->robj->tbo.ttm;
 
@@ -754,6 +685,11 @@ static void amdgpu_cs_parser_fini(struct amdgpu_cs_parser *parser, int error, bo
 		ttm_eu_backoff_reservation(&parser->ticket,
 					   &parser->validated);
 	}
+
+	for (i = 0; i < parser->num_post_dep_syncobjs; i++)
+		drm_syncobj_put(parser->post_dep_syncobjs[i]);
+	kfree(parser->post_dep_syncobjs);
+
 	dma_fence_put(parser->fence);
 
 	if (parser->ctx)
@@ -917,9 +853,8 @@ static int amdgpu_cs_ib_fill(struct amdgpu_device *adev,
 				return -EINVAL;
 		}
 
-		r = amdgpu_cs_get_ring(adev, chunk_ib->ip_type,
-				       chunk_ib->ip_instance, chunk_ib->ring,
-				       &ring);
+		r = amdgpu_queue_mgr_map(adev, &parser->ctx->queue_mgr, chunk_ib->ip_type,
+					 chunk_ib->ip_instance, chunk_ib->ring, &ring);
 		if (r)
 			return r;
 
@@ -996,62 +931,150 @@ static int amdgpu_cs_ib_fill(struct amdgpu_device *adev,
 	return 0;
 }
 
-static int amdgpu_cs_dependencies(struct amdgpu_device *adev,
-				  struct amdgpu_cs_parser *p)
+static int amdgpu_cs_process_fence_dep(struct amdgpu_cs_parser *p,
+				       struct amdgpu_cs_chunk *chunk)
 {
 	struct amdgpu_fpriv *fpriv = p->filp->driver_priv;
-	int i, j, r;
-
-	for (i = 0; i < p->nchunks; ++i) {
-		struct drm_amdgpu_cs_chunk_dep *deps;
-		struct amdgpu_cs_chunk *chunk;
-		unsigned num_deps;
+	unsigned num_deps;
+	int i, r;
+	struct drm_amdgpu_cs_chunk_dep *deps;
 
-		chunk = &p->chunks[i];
+	deps = (struct drm_amdgpu_cs_chunk_dep *)chunk->kdata;
+	num_deps = chunk->length_dw * 4 /
+		sizeof(struct drm_amdgpu_cs_chunk_dep);
 
-		if (chunk->chunk_id != AMDGPU_CHUNK_ID_DEPENDENCIES)
-			continue;
+	for (i = 0; i < num_deps; ++i) {
+		struct amdgpu_ring *ring;
+		struct amdgpu_ctx *ctx;
+		struct dma_fence *fence;
 
-		deps = (struct drm_amdgpu_cs_chunk_dep *)chunk->kdata;
-		num_deps = chunk->length_dw * 4 /
-			sizeof(struct drm_amdgpu_cs_chunk_dep);
+		ctx = amdgpu_ctx_get(fpriv, deps[i].ctx_id);
+		if (ctx == NULL)
+			return -EINVAL;
 
-		for (j = 0; j < num_deps; ++j) {
-			struct amdgpu_ring *ring;
-			struct amdgpu_ctx *ctx;
-			struct dma_fence *fence;
+		r = amdgpu_queue_mgr_map(p->adev, &ctx->queue_mgr,
+					 deps[i].ip_type,
+					 deps[i].ip_instance,
+					 deps[i].ring, &ring);
+		if (r) {
+			amdgpu_ctx_put(ctx);
+			return r;
+		}
 
-			r = amdgpu_cs_get_ring(adev, deps[j].ip_type,
-					       deps[j].ip_instance,
-					       deps[j].ring, &ring);
+		fence = amdgpu_ctx_get_fence(ctx, ring,
+					     deps[i].handle);
+		if (IS_ERR(fence)) {
+			r = PTR_ERR(fence);
+			amdgpu_ctx_put(ctx);
+			return r;
+		} else if (fence) {
+			r = amdgpu_sync_fence(p->adev, &p->job->sync,
+					      fence);
+			dma_fence_put(fence);
+			amdgpu_ctx_put(ctx);
 			if (r)
 				return r;
+		}
+	}
+	return 0;
+}
 
-			ctx = amdgpu_ctx_get(fpriv, deps[j].ctx_id);
-			if (ctx == NULL)
-				return -EINVAL;
+static int amdgpu_syncobj_lookup_and_add_to_sync(struct amdgpu_cs_parser *p,
+						 uint32_t handle)
+{
+	int r;
+	struct dma_fence *fence;
+	r = drm_syncobj_fence_get(p->filp, handle, &fence);
+	if (r)
+		return r;
 
-			fence = amdgpu_ctx_get_fence(ctx, ring,
-						     deps[j].handle);
-			if (IS_ERR(fence)) {
-				r = PTR_ERR(fence);
-				amdgpu_ctx_put(ctx);
-				return r;
+	r = amdgpu_sync_fence(p->adev, &p->job->sync, fence);
+	dma_fence_put(fence);
 
-			} else if (fence) {
-				r = amdgpu_sync_fence(adev, &p->job->sync,
-						      fence);
-				dma_fence_put(fence);
-				amdgpu_ctx_put(ctx);
-				if (r)
-					return r;
-			}
+	return r;
+}
+
+static int amdgpu_cs_process_syncobj_in_dep(struct amdgpu_cs_parser *p,
+					    struct amdgpu_cs_chunk *chunk)
+{
+	unsigned num_deps;
+	int i, r;
+	struct drm_amdgpu_cs_chunk_sem *deps;
+
+	deps = (struct drm_amdgpu_cs_chunk_sem *)chunk->kdata;
+	num_deps = chunk->length_dw * 4 /
+		sizeof(struct drm_amdgpu_cs_chunk_sem);
+
+	for (i = 0; i < num_deps; ++i) {
+		r = amdgpu_syncobj_lookup_and_add_to_sync(p, deps[i].handle);
+		if (r)
+			return r;
+	}
+	return 0;
+}
+
+static int amdgpu_cs_process_syncobj_out_dep(struct amdgpu_cs_parser *p,
+					     struct amdgpu_cs_chunk *chunk)
+{
+	unsigned num_deps;
+	int i;
+	struct drm_amdgpu_cs_chunk_sem *deps;
+	deps = (struct drm_amdgpu_cs_chunk_sem *)chunk->kdata;
+	num_deps = chunk->length_dw * 4 /
+		sizeof(struct drm_amdgpu_cs_chunk_sem);
+
+	p->post_dep_syncobjs = kmalloc_array(num_deps,
+					     sizeof(struct drm_syncobj *),
+					     GFP_KERNEL);
+	p->num_post_dep_syncobjs = 0;
+
+	for (i = 0; i < num_deps; ++i) {
+		p->post_dep_syncobjs[i] = drm_syncobj_find(p->filp, deps[i].handle);
+		if (!p->post_dep_syncobjs[i])
+			return -EINVAL;
+		p->num_post_dep_syncobjs++;
+	}
+	return 0;
+}
+
+static int amdgpu_cs_dependencies(struct amdgpu_device *adev,
+				  struct amdgpu_cs_parser *p)
+{
+	int i, r;
+
+	for (i = 0; i < p->nchunks; ++i) {
+		struct amdgpu_cs_chunk *chunk;
+
+		chunk = &p->chunks[i];
+
+		if (chunk->chunk_id == AMDGPU_CHUNK_ID_DEPENDENCIES) {
+			r = amdgpu_cs_process_fence_dep(p, chunk);
+			if (r)
+				return r;
+		} else if (chunk->chunk_id == AMDGPU_CHUNK_ID_SYNCOBJ_IN) {
+			r = amdgpu_cs_process_syncobj_in_dep(p, chunk);
+			if (r)
+				return r;
+		} else if (chunk->chunk_id == AMDGPU_CHUNK_ID_SYNCOBJ_OUT) {
+			r = amdgpu_cs_process_syncobj_out_dep(p, chunk);
+			if (r)
+				return r;
 		}
 	}
 
 	return 0;
 }
 
+static void amdgpu_cs_post_dependencies(struct amdgpu_cs_parser *p)
+{
+	int i;
+
+	for (i = 0; i < p->num_post_dep_syncobjs; ++i) {
+		drm_syncobj_replace_fence(p->filp, p->post_dep_syncobjs[i],
+					  p->fence);
+	}
+}
+
 static int amdgpu_cs_submit(struct amdgpu_cs_parser *p,
 			    union drm_amdgpu_cs *cs)
 {
@@ -1072,6 +1095,9 @@ static int amdgpu_cs_submit(struct amdgpu_cs_parser *p,
 	job->owner = p->filp;
 	job->fence_ctx = entity->fence_context;
 	p->fence = dma_fence_get(&job->base.s_fence->finished);
+
+	amdgpu_cs_post_dependencies(p);
+
 	cs->out.handle = amdgpu_ctx_add_fence(p->ctx, ring, p->fence);
 	job->uf_sequence = cs->out.handle;
 	amdgpu_job_free_resources(job);
@@ -1079,13 +1105,13 @@ static int amdgpu_cs_submit(struct amdgpu_cs_parser *p,
 
 	trace_amdgpu_cs_ioctl(job);
 	amd_sched_entity_push_job(&job->base);
-
 	return 0;
 }
 
 int amdgpu_cs_ioctl(struct drm_device *dev, void *data, struct drm_file *filp)
 {
 	struct amdgpu_device *adev = dev->dev_private;
+	struct amdgpu_fpriv *fpriv = filp->driver_priv;
 	union drm_amdgpu_cs *cs = data;
 	struct amdgpu_cs_parser parser = {};
 	bool reserved_buffers = false;
@@ -1093,6 +1119,8 @@ int amdgpu_cs_ioctl(struct drm_device *dev, void *data, struct drm_file *filp)
 
 	if (!adev->accel_working)
 		return -EBUSY;
+	if (amdgpu_kms_vram_lost(adev, fpriv))
+		return -ENODEV;
 
 	parser.adev = adev;
 	parser.filp = filp;
@@ -1154,21 +1182,28 @@ int amdgpu_cs_wait_ioctl(struct drm_device *dev, void *data,
 {
 	union drm_amdgpu_wait_cs *wait = data;
 	struct amdgpu_device *adev = dev->dev_private;
+	struct amdgpu_fpriv *fpriv = filp->driver_priv;
 	unsigned long timeout = amdgpu_gem_timeout(wait->in.timeout);
 	struct amdgpu_ring *ring = NULL;
 	struct amdgpu_ctx *ctx;
 	struct dma_fence *fence;
 	long r;
 
-	r = amdgpu_cs_get_ring(adev, wait->in.ip_type, wait->in.ip_instance,
-			       wait->in.ring, &ring);
-	if (r)
-		return r;
+	if (amdgpu_kms_vram_lost(adev, fpriv))
+		return -ENODEV;
 
 	ctx = amdgpu_ctx_get(filp->driver_priv, wait->in.ctx_id);
 	if (ctx == NULL)
 		return -EINVAL;
 
+	r = amdgpu_queue_mgr_map(adev, &ctx->queue_mgr,
+				 wait->in.ip_type, wait->in.ip_instance,
+				 wait->in.ring, &ring);
+	if (r) {
+		amdgpu_ctx_put(ctx);
+		return r;
+	}
+
 	fence = amdgpu_ctx_get_fence(ctx, ring, wait->in.handle);
 	if (IS_ERR(fence))
 		r = PTR_ERR(fence);
@@ -1204,15 +1239,17 @@ static struct dma_fence *amdgpu_cs_get_fence(struct amdgpu_device *adev,
 	struct dma_fence *fence;
 	int r;
 
-	r = amdgpu_cs_get_ring(adev, user->ip_type, user->ip_instance,
-			       user->ring, &ring);
-	if (r)
-		return ERR_PTR(r);
-
 	ctx = amdgpu_ctx_get(filp->driver_priv, user->ctx_id);
 	if (ctx == NULL)
 		return ERR_PTR(-EINVAL);
 
+	r = amdgpu_queue_mgr_map(adev, &ctx->queue_mgr, user->ip_type,
+				 user->ip_instance, user->ring, &ring);
+	if (r) {
+		amdgpu_ctx_put(ctx);
+		return ERR_PTR(r);
+	}
+
 	fence = amdgpu_ctx_get_fence(ctx, ring, user->seq_no);
 	amdgpu_ctx_put(ctx);
 
@@ -1333,12 +1370,15 @@ int amdgpu_cs_wait_fences_ioctl(struct drm_device *dev, void *data,
 				struct drm_file *filp)
 {
 	struct amdgpu_device *adev = dev->dev_private;
+	struct amdgpu_fpriv *fpriv = filp->driver_priv;
 	union drm_amdgpu_wait_fences *wait = data;
 	uint32_t fence_count = wait->in.fence_count;
 	struct drm_amdgpu_fence *fences_user;
 	struct drm_amdgpu_fence *fences;
 	int r;
 
+	if (amdgpu_kms_vram_lost(adev, fpriv))
+		return -ENODEV;
 	/* Get the fences from userspace */
 	fences = kmalloc_array(fence_count, sizeof(struct drm_amdgpu_fence),
 			GFP_KERNEL);