1 files changed, 22 insertions, 13 deletions

diff --git a/drivers/gpu/drm/vc4/vc4_gem.c b/drivers/gpu/drm/vc4/vc4_gem.c
index 6155e8a..47a095f 100644
--- a/drivers/gpu/drm/vc4/vc4_gem.c
+++ b/drivers/gpu/drm/vc4/vc4_gem.c
@@ -419,10 +419,6 @@ again:
 
 	vc4_flush_caches(dev);
 
-	/* Disable the binner's pre-loaded overflow memory address */
-	V3D_WRITE(V3D_BPOA, 0);
-	V3D_WRITE(V3D_BPOS, 0);
-
 	/* Either put the job in the binner if it uses the binner, or
 	 * immediately move it to the to-be-rendered queue.
 	 */
@@ -471,6 +467,11 @@ vc4_update_bo_seqnos(struct vc4_exec_info *exec, uint64_t seqno)
 	list_for_each_entry(bo, &exec->unref_list, unref_head) {
 		bo->seqno = seqno;
 	}
+
+	for (i = 0; i < exec->rcl_write_bo_count; i++) {
+		bo = to_vc4_bo(&exec->rcl_write_bo[i]->base);
+		bo->write_seqno = seqno;
+	}
 }
 
 /* Queues a struct vc4_exec_info for execution.  If no job is
@@ -534,8 +535,8 @@ vc4_cl_lookup_bos(struct drm_device *dev,
 		return -EINVAL;
 	}
 
-	exec->bo = kcalloc(exec->bo_count, sizeof(struct drm_gem_cma_object *),
-			   GFP_KERNEL);
+	exec->bo = drm_calloc_large(exec->bo_count,
+				    sizeof(struct drm_gem_cma_object *));
 	if (!exec->bo) {
 		DRM_ERROR("Failed to allocate validated BO pointers\n");
 		return -ENOMEM;
@@ -572,8 +573,8 @@ vc4_cl_lookup_bos(struct drm_device *dev,
 	spin_unlock(&file_priv->table_lock);
 
 fail:
-	kfree(handles);
-	return 0;
+	drm_free_large(handles);
+	return ret;
 }
 
 static int
@@ -608,7 +609,7 @@ vc4_get_bcl(struct drm_device *dev, struct vc4_exec_info *exec)
 	 * read the contents back for validation, and I think the
 	 * bo->vaddr is uncached access.
 	 */
-	temp = kmalloc(temp_size, GFP_KERNEL);
+	temp = drm_malloc_ab(temp_size, 1);
 	if (!temp) {
 		DRM_ERROR("Failed to allocate storage for copying "
 			  "in bin/render CLs.\n");
@@ -673,9 +674,17 @@ vc4_get_bcl(struct drm_device *dev, struct vc4_exec_info *exec)
 		goto fail;
 
 	ret = vc4_validate_shader_recs(dev, exec);
+	if (ret)
+		goto fail;
+
+	/* Block waiting on any previous rendering into the CS's VBO,
+	 * IB, or textures, so that pixels are actually written by the
+	 * time we try to read them.
+	 */
+	ret = vc4_wait_for_seqno(dev, exec->bin_dep_seqno, ~0ull, true);
 
 fail:
-	kfree(temp);
+	drm_free_large(temp);
 	return ret;
 }
 
@@ -688,7 +697,7 @@ vc4_complete_exec(struct drm_device *dev, struct vc4_exec_info *exec)
 	if (exec->bo) {
 		for (i = 0; i < exec->bo_count; i++)
 			drm_gem_object_unreference_unlocked(&exec->bo[i]->base);
-		kfree(exec->bo);
+		drm_free_large(exec->bo);
 	}
 
 	while (!list_empty(&exec->unref_list)) {
@@ -942,8 +951,8 @@ vc4_gem_destroy(struct drm_device *dev)
 		vc4->overflow_mem = NULL;
 	}
 
-	vc4_bo_cache_destroy(dev);
-
 	if (vc4->hang_state)
 		vc4_free_hang_state(dev, vc4->hang_state);
+
+	vc4_bo_cache_destroy(dev);
 }