On Fri, Mar 19, 2021 at 5:09 AM Fabio Estevam <festevam@xxxxxxxxx> wrote: > > Hi Dmitry, > > On Mon, Mar 1, 2021 at 6:41 PM Dmitry Baryshkov > <dmitry.baryshkov@xxxxxxxxxx> wrote: > > > diff --git a/drivers/gpu/drm/msm/msm_atomic.c b/drivers/gpu/drm/msm/msm_atomic.c > > index 6a326761dc4a..2fd0cf6421ad 100644 > > --- a/drivers/gpu/drm/msm/msm_atomic.c > > +++ b/drivers/gpu/drm/msm/msm_atomic.c > > @@ -207,7 +207,12 @@ void msm_atomic_commit_tail(struct drm_atomic_state *state) > > struct msm_kms *kms = priv->kms; > > struct drm_crtc *async_crtc = NULL; > > unsigned crtc_mask = get_crtc_mask(state); > > - bool async = kms->funcs->vsync_time && > > + bool async; > > + > > + if (!kms) > > + return; > > + > > + async = kms->funcs->vsync_time && > > can_do_async(state, &async_crtc); > > I also see the same issue on a i.MX53: > https://lists.freedesktop.org/archives/freedreno/2021-January/009369.html > > Then I got a different suggestion from Rob. Please check: > > https://www.spinics.net/lists/dri-devel/msg286648.html > and > https://www.spinics.net/lists/dri-devel/msg286649.html > > Does this series fix the issue in your platform too? I think that might not help if something fails to probe due to (for example) a missing dependency, so !priv->kms is probably a better check to cover both cases. But the 2nd patch makes a good point, that the suspend/resume path probably needs the same treatment BR, -R
