drm/syncobj: Fix use-after-free

While reviewing Christian's annotation patch I noticed that we have a
user-after-free for the WAIT_FOR_SUBMIT case: We drop the syncobj
reference before we've completed the waiting.

Of course usually there's nothing bad happening here since userspace
keeps the reference, but we can't rely on userspace to play nice here!

Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
Fixes: bc9c80fe01 ("drm/syncobj: use the timeline point in drm_syncobj_find_fence v4")
Reviewed-by: Christian König <christian.koenig@amd.com>
Cc: Christian König <christian.koenig@amd.com>
Cc: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Cc: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Cc: Maxime Ripard <mripard@kernel.org>
Cc: Thomas Zimmermann <tzimmermann@suse.de>
Cc: David Airlie <airlied@linux.ie>
Cc: Daniel Vetter <daniel@ffwll.ch>
Cc: dri-devel@lists.freedesktop.org
Cc: <stable@vger.kernel.org> # v5.2+
Link: https://patchwork.freedesktop.org/patch/msgid/20210119130318.615145-1-daniel.vetter@ffwll.ch

drivers/gpu/drm/drm_syncobj.c

@@ -388,19 +388,18 @@ int drm_syncobj_find_fence(struct drm_file *file_private,
 		return -ENOENT;
 
 	*fence = drm_syncobj_fence_get(syncobj);
-	drm_syncobj_put(syncobj);
 
 	if (*fence) {
 		ret = dma_fence_chain_find_seqno(fence, point);
 		if (!ret)
-			return 0;
+			goto out;
 		dma_fence_put(*fence);
 	} else {
 		ret = -EINVAL;
 	}
 
 	if (!(flags & DRM_SYNCOBJ_WAIT_FLAGS_WAIT_FOR_SUBMIT))
-		return ret;
+		goto out;
 
 	memset(&wait, 0, sizeof(wait));
 	wait.task = current;
@@ -432,6 +431,9 @@ int drm_syncobj_find_fence(struct drm_file *file_private,
 	if (wait.node.next)
 		drm_syncobj_remove_wait(syncobj, &wait);
 
+out:
+	drm_syncobj_put(syncobj);
+
 	return ret;
 }
 EXPORT_SYMBOL(drm_syncobj_find_fence);