Merge tag 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mst/vhost

[linux.git] / drivers / gpu / drm / amd / amdgpu / amdgpu_cs.c

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c
index f09b2cba40ca505649decf23a27b60850b62407a..1f040d85ac47fe336f609a0cf9d39ac59b94f665 100644 (file)
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c
@@ -181,8 +181,6 @@ int amdgpu_cs_parser_init(struct amdgpu_cs_parser *p, void *data)
                }
                p->chunks[i].chunk_id = user_chunk.chunk_id;
                p->chunks[i].length_dw = user_chunk.length_dw;
-               if (p->chunks[i].chunk_id == AMDGPU_CHUNK_ID_IB)
-                       p->num_ibs++;
 
                size = p->chunks[i].length_dw;
                cdata = (void __user *)(unsigned long)user_chunk.chunk_data;
@@ -199,7 +197,12 @@ int amdgpu_cs_parser_init(struct amdgpu_cs_parser *p, void *data)
                        goto out;
                }
 
-               if (p->chunks[i].chunk_id == AMDGPU_CHUNK_ID_FENCE) {
+               switch (p->chunks[i].chunk_id) {
+               case AMDGPU_CHUNK_ID_IB:
+                       p->num_ibs++;
+                       break;
+
+               case AMDGPU_CHUNK_ID_FENCE:
                        size = sizeof(struct drm_amdgpu_cs_chunk_fence);
                        if (p->chunks[i].length_dw * sizeof(uint32_t) >= size) {
                                uint32_t handle;
@@ -221,6 +224,14 @@ int amdgpu_cs_parser_init(struct amdgpu_cs_parser *p, void *data)
                                r = -EINVAL;
                                goto out;
                        }
+                       break;
+
+               case AMDGPU_CHUNK_ID_DEPENDENCIES:
+                       break;
+
+               default:
+                       r = -EINVAL;
+                       goto out;
                }
        }
 
@@ -445,8 +456,9 @@ static void amdgpu_cs_parser_fini(struct amdgpu_cs_parser *parser, int error, bo
        for (i = 0; i < parser->nchunks; i++)
                drm_free_large(parser->chunks[i].kdata);
        kfree(parser->chunks);
-       for (i = 0; i < parser->num_ibs; i++)
-               amdgpu_ib_free(parser->adev, &parser->ibs[i]);
+       if (parser->ibs)
+               for (i = 0; i < parser->num_ibs; i++)
+                       amdgpu_ib_free(parser->adev, &parser->ibs[i]);
        kfree(parser->ibs);
        if (parser->uf.bo)
                drm_gem_object_unreference_unlocked(&parser->uf.bo->gem_base);
@@ -654,6 +666,64 @@ 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)
+{
+       struct amdgpu_fpriv *fpriv = p->filp->driver_priv;
+       struct amdgpu_ib *ib;
+       int i, j, r;
+
+       if (!p->num_ibs)
+               return 0;
+
+       /* Add dependencies to first IB */
+       ib = &p->ibs[0];
+       for (i = 0; i < p->nchunks; ++i) {
+               struct drm_amdgpu_cs_chunk_dep *deps;
+               struct amdgpu_cs_chunk *chunk;
+               unsigned num_deps;
+
+               chunk = &p->chunks[i];
+
+               if (chunk->chunk_id != AMDGPU_CHUNK_ID_DEPENDENCIES)
+                       continue;
+
+               deps = (struct drm_amdgpu_cs_chunk_dep *)chunk->kdata;
+               num_deps = chunk->length_dw * 4 /
+                       sizeof(struct drm_amdgpu_cs_chunk_dep);
+
+               for (j = 0; j < num_deps; ++j) {
+                       struct amdgpu_fence *fence;
+                       struct amdgpu_ring *ring;
+                       struct amdgpu_ctx *ctx;
+
+                       r = amdgpu_cs_get_ring(adev, deps[j].ip_type,
+                                              deps[j].ip_instance,
+                                              deps[j].ring, &ring);
+                       if (r)
+                               return r;
+
+                       ctx = amdgpu_ctx_get(fpriv, deps[j].ctx_id);
+                       if (ctx == NULL)
+                               return -EINVAL;
+
+                       r = amdgpu_fence_recreate(ring, p->filp,
+                                                 deps[j].handle,
+                                                 &fence);
+                       if (r) {
+                               amdgpu_ctx_put(ctx);
+                               return r;
+                       }
+
+                       amdgpu_sync_fence(&ib->sync, fence);
+                       amdgpu_fence_unref(&fence);
+                       amdgpu_ctx_put(ctx);
+               }
+       }
+
+       return 0;
+}
+
 int amdgpu_cs_ioctl(struct drm_device *dev, void *data, struct drm_file *filp)
 {
        struct amdgpu_device *adev = dev->dev_private;
@@ -688,11 +758,16 @@ int amdgpu_cs_ioctl(struct drm_device *dev, void *data, struct drm_file *filp)
                        else
                                DRM_ERROR("Failed to process the buffer list %d!\n", r);
                }
-       } else {
+       }
+
+       if (!r) {
                reserved_buffers = true;
                r = amdgpu_cs_ib_fill(adev, &parser);
        }
 
+       if (!r)
+               r = amdgpu_cs_dependencies(adev, &parser);
+
        if (r) {
                amdgpu_cs_parser_fini(&parser, r, reserved_buffers);
                up_read(&adev->exclusive_lock);
@@ -730,9 +805,9 @@ 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;
-       uint64_t seq[AMDGPU_MAX_RINGS] = {0};
-       struct amdgpu_ring *ring = NULL;
        unsigned long timeout = amdgpu_gem_timeout(wait->in.timeout);
+       struct amdgpu_fence *fence = NULL;
+       struct amdgpu_ring *ring = NULL;
        struct amdgpu_ctx *ctx;
        long r;
 
@@ -742,12 +817,19 @@ int amdgpu_cs_wait_ioctl(struct drm_device *dev, void *data,
 
        r = amdgpu_cs_get_ring(adev, wait->in.ip_type, wait->in.ip_instance,
                               wait->in.ring, &ring);
-       if (r)
+       if (r) {
+               amdgpu_ctx_put(ctx);
                return r;
+       }
 
-       seq[ring->idx] = wait->in.handle;
+       r = amdgpu_fence_recreate(ring, filp, wait->in.handle, &fence);
+       if (r) {
+               amdgpu_ctx_put(ctx);
+               return r;
+       }
 
-       r = amdgpu_fence_wait_seq_timeout(adev, seq, true, timeout);
+       r = fence_wait_timeout(&fence->base, true, timeout);
+       amdgpu_fence_unref(&fence);
        amdgpu_ctx_put(ctx);
        if (r < 0)
                return r;