Commit da758d51 authored by Gerd Hoffmann's avatar Gerd Hoffmann
Browse files

drm/virtio: rework virtio_gpu_execbuffer_ioctl fencing 



Rework fencing workflow, starting with virtio_gpu_execbuffer_ioctl.
Stop using ttm helpers, use the virtio_gpu_array_* helpers (which work
on the reservation objects directly) instead.

Also store the object array in struct virtio_gpu_vbuffer, so we
explicitly keep a reference of all buffers used instead of depending
on ttm_bo_put() checking whenever the object is actually idle before
releasing it.

New workflow:

 (1) All gem objects needed by a command are added to a
     virtio_gpu_object_array.
 (2) All reservation objects will be locked (virtio_gpu_array_lock_resv).
 (3) virtio_gpu_fence_emit() completes fence initialization.
 (4) fence gets added to the objects, reservation objects are unlocked
     (virtio_gpu_array_add_fence, virtio_gpu_array_unlock_resv).
 (5) virtio command is submitted to the host.
 (6) The completion callback (virtio_gpu_dequeue_ctrl_func)
     will drop object references and free virtio_gpu_object_array.

v6: rewrite most of the patch.

Signed-off-by: default avatarGerd Hoffmann <kraxel@redhat.com>
Reviewed-by: default avatarChia-I Wu <olvaffe@gmail.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20190829103301.3539-9-kraxel@redhat.com
parent 98abe21d

drivers/gpu/drm/virtio/virtgpu_drv.h

+4 −2
Original line number Diff line number Diff line
@@ -121,9 +121,9 @@ struct virtio_gpu_vbuffer { 


	char *resp_buf;

	int resp_size;



	virtio_gpu_resp_cb resp_cb;



	struct virtio_gpu_object_array *objs;

	struct list_head list;

};
@@ -318,7 +318,9 @@ void virtio_gpu_cmd_context_detach_resource(struct virtio_gpu_device *vgdev, 
					    uint32_t resource_id);

void virtio_gpu_cmd_submit(struct virtio_gpu_device *vgdev,

			   void *data, uint32_t data_size,

			   uint32_t ctx_id, struct virtio_gpu_fence *fence);

			   uint32_t ctx_id,

			   struct virtio_gpu_object_array *objs,

			   struct virtio_gpu_fence *fence);

void virtio_gpu_cmd_transfer_from_host_3d(struct virtio_gpu_device *vgdev,

					  uint32_t resource_id, uint32_t ctx_id,

					  uint64_t offset, uint32_t level,

drivers/gpu/drm/virtio/virtgpu_ioctl.c

+18 −38
Original line number Diff line number Diff line
@@ -107,16 +107,11 @@ static int virtio_gpu_execbuffer_ioctl(struct drm_device *dev, void *data, 
	struct drm_virtgpu_execbuffer *exbuf = data;

	struct virtio_gpu_device *vgdev = dev->dev_private;

	struct virtio_gpu_fpriv *vfpriv = drm_file->driver_priv;

	struct drm_gem_object *gobj;

	struct virtio_gpu_fence *out_fence;

	struct virtio_gpu_object *qobj;

	int ret;

	uint32_t *bo_handles = NULL;

	void __user *user_bo_handles = NULL;

	struct list_head validate_list;

	struct ttm_validate_buffer *buflist = NULL;

	int i;

	struct ww_acquire_ctx ticket;

	struct virtio_gpu_object_array *buflist = NULL;

	struct sync_file *sync_file;

	int in_fence_fd = exbuf->fence_fd;

	int out_fence_fd = -1;
@@ -157,15 +152,10 @@ static int virtio_gpu_execbuffer_ioctl(struct drm_device *dev, void *data, 
			return out_fence_fd;

	}



	INIT_LIST_HEAD(&validate_list);

	if (exbuf->num_bo_handles) {



		bo_handles = kvmalloc_array(exbuf->num_bo_handles,

					    sizeof(uint32_t), GFP_KERNEL);

		buflist = kvmalloc_array(exbuf->num_bo_handles,

					   sizeof(struct ttm_validate_buffer),

					   GFP_KERNEL | __GFP_ZERO);

		if (!bo_handles || !buflist) {

		if (!bo_handles) {

			ret = -ENOMEM;

			goto out_unused_fd;

		}
@@ -177,25 +167,21 @@ static int virtio_gpu_execbuffer_ioctl(struct drm_device *dev, void *data, 
			goto out_unused_fd;

		}



		for (i = 0; i < exbuf->num_bo_handles; i++) {

			gobj = drm_gem_object_lookup(drm_file, bo_handles[i]);

			if (!gobj) {

		buflist = virtio_gpu_array_from_handles(drm_file, bo_handles,

							exbuf->num_bo_handles);

		if (!buflist) {

			ret = -ENOENT;

			goto out_unused_fd;

		}



			qobj = gem_to_virtio_gpu_obj(gobj);

			buflist[i].bo = &qobj->tbo;



			list_add(&buflist[i].head, &validate_list);

		}

		kvfree(bo_handles);

		bo_handles = NULL;

	}



	ret = virtio_gpu_object_list_validate(&ticket, &validate_list);

	if (buflist) {

		ret = virtio_gpu_array_lock_resv(buflist);

		if (ret)

		goto out_free;

			goto out_unused_fd;

	}



	buf = memdup_user(u64_to_user_ptr(exbuf->command), exbuf->size);

	if (IS_ERR(buf)) {
@@ -222,24 +208,18 @@ static int virtio_gpu_execbuffer_ioctl(struct drm_device *dev, void *data, 
	}



	virtio_gpu_cmd_submit(vgdev, buf, exbuf->size,

			      vfpriv->ctx_id, out_fence);



	ttm_eu_fence_buffer_objects(&ticket, &validate_list, &out_fence->f);



	/* fence the command bo */

	virtio_gpu_unref_list(&validate_list);

	kvfree(buflist);

			      vfpriv->ctx_id, buflist, out_fence);

	return 0;



out_memdup:

	kfree(buf);

out_unresv:

	ttm_eu_backoff_reservation(&ticket, &validate_list);

out_free:

	virtio_gpu_unref_list(&validate_list);

	if (buflist)

		virtio_gpu_array_unlock_resv(buflist);

out_unused_fd:

	kvfree(bo_handles);

	kvfree(buflist);

	if (buflist)

		virtio_gpu_array_put_free(buflist);



	if (out_fence_fd >= 0)

		put_unused_fd(out_fence_fd);

drivers/gpu/drm/virtio/virtgpu_vq.c

+16 −5
Original line number Diff line number Diff line
@@ -192,7 +192,7 @@ void virtio_gpu_dequeue_ctrl_func(struct work_struct *work) 
	} while (!virtqueue_enable_cb(vgdev->ctrlq.vq));

	spin_unlock(&vgdev->ctrlq.qlock);



	list_for_each_entry_safe(entry, tmp, &reclaim_list, list) {

	list_for_each_entry(entry, &reclaim_list, list) {

		resp = (struct virtio_gpu_ctrl_hdr *)entry->resp_buf;



		trace_virtio_gpu_cmd_response(vgdev->ctrlq.vq, resp);
@@ -219,14 +219,18 @@ void virtio_gpu_dequeue_ctrl_func(struct work_struct *work) 
		}

		if (entry->resp_cb)

			entry->resp_cb(vgdev, entry);



		list_del(&entry->list);

		free_vbuf(vgdev, entry);

	}

	wake_up(&vgdev->ctrlq.ack_queue);



	if (fence_id)

		virtio_gpu_fence_event_process(vgdev, fence_id);



	list_for_each_entry_safe(entry, tmp, &reclaim_list, list) {

		if (entry->objs)

			virtio_gpu_array_put_free(entry->objs);

		list_del(&entry->list);

		free_vbuf(vgdev, entry);

	}

}



void virtio_gpu_dequeue_cursor_func(struct work_struct *work)
@@ -337,6 +341,10 @@ again: 


	if (fence)

		virtio_gpu_fence_emit(vgdev, hdr, fence);

	if (vbuf->objs) {

		virtio_gpu_array_add_fence(vbuf->objs, &fence->f);

		virtio_gpu_array_unlock_resv(vbuf->objs);

	}

	notify = virtio_gpu_queue_ctrl_buffer_locked(vgdev, vbuf);

	spin_unlock(&vgdev->ctrlq.qlock);

	if (notify)
@@ -940,7 +948,9 @@ void virtio_gpu_cmd_transfer_from_host_3d(struct virtio_gpu_device *vgdev, 


void virtio_gpu_cmd_submit(struct virtio_gpu_device *vgdev,

			   void *data, uint32_t data_size,

			   uint32_t ctx_id, struct virtio_gpu_fence *fence)

			   uint32_t ctx_id,

			   struct virtio_gpu_object_array *objs,

			   struct virtio_gpu_fence *fence)

{

	struct virtio_gpu_cmd_submit *cmd_p;

	struct virtio_gpu_vbuffer *vbuf;
@@ -950,6 +960,7 @@ void virtio_gpu_cmd_submit(struct virtio_gpu_device *vgdev, 


	vbuf->data_buf = data;

	vbuf->data_size = data_size;

	vbuf->objs = objs;



	cmd_p->hdr.type = cpu_to_le32(VIRTIO_GPU_CMD_SUBMIT_3D);

	cmd_p->hdr.ctx_id = cpu_to_le32(ctx_id);

CRA Git | Maintained and supported by SUSTech CRA and CCSE