The ->disable_plane hook always had a return value, but only since the
introduction of primary planes was there any implementation that
actually failed.
So handle such failures correctly.
Note that drm_plane_force_disable is special: In the modeset cleanup
case we first disable all crtc, so primary planes should all be freed
already. And in the fb helper we only reset non-primary planes. Still
better be paranoid and add an early return.
I don't see how this could happen, but it might fix the fb refcount
underrun Thierry is seeing. Matt Roper spotted this issue.
Cc: Thierry Reding <treding@nvidia.com>
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Cc: Matt Roper <matthew.d.roper@intel.com>
Reviewed-by: Matt Roper <matthew.d.roper@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
                return;
 
        ret = plane->funcs->disable_plane(plane);
-       if (ret)
+       if (ret) {
                DRM_ERROR("failed to disable plane with busy fb\n");
+               return;
+       }
        /* disconnect the plane from the fb and crtc: */
        __drm_framebuffer_unreference(old_fb);
        plane->fb = NULL;
        if (!plane_req->fb_id) {
                drm_modeset_lock_all(dev);
                old_fb = plane->fb;
-               plane->funcs->disable_plane(plane);
-               plane->crtc = NULL;
-               plane->fb = NULL;
+               ret = plane->funcs->disable_plane(plane);
+               if (!ret) {
+                       plane->crtc = NULL;
+                       plane->fb = NULL;
+               } else {
+                       old_fb = NULL;
+               }
                drm_modeset_unlock_all(dev);
                goto out;
        }