Re: [PATCH v3] drm/msm/dp: check hpd_state before push idle pattern at dp_bridge_disable()

[Kuogee Hsieh]

On 8/10/2022 3:22 PM, Stephen Boyd wrote:
> Quoting Kuogee Hsieh (2022-08-10 12:25:51)
> > diff --git a/drivers/gpu/drm/msm/dp/dp_display.c b/drivers/gpu/drm/msm/dp/dp_display.c
> > index b36f8b6..678289a 100644
> > --- a/drivers/gpu/drm/msm/dp/dp_display.c
> > +++ b/drivers/gpu/drm/msm/dp/dp_display.c
> > @@ -1729,10 +1729,20 @@ void dp_bridge_disable(struct drm_bridge *drm_bridge)
> >          struct msm_dp_bridge *dp_bridge = to_dp_bridge(drm_bridge);
> >          struct msm_dp *dp = dp_bridge->dp_display;
> >          struct dp_display_private *dp_display;
> > +       u32 state;
> >
> >          dp_display = container_of(dp, struct dp_display_private, dp_display);
> >
> > +       mutex_lock(&dp_display->event_mutex);
> > +
> > +       state = dp_display->hpd_state;
> > +       if (state != ST_DISCONNECT_PENDING && state != ST_CONNECTED) {
> It's concerning that we have to check this at all. Are we still
> interjecting into the disable path when the cable is disconnected?

yes,

The problem is not from cable disconnected.

There is a corner case that this function is called at drm shutdown 
(drm_release).

At that time, mainlink is not enabled, hence dp_ctrl_push_idle() will 
cause system crash.



> > +               mutex_unlock(&dp_display->event_mutex);
> > +               return;
> > +       }
> > +
> >          dp_ctrl_push_idle(dp_display->ctrl);
> > +       mutex_unlock(&dp_display->event_mutex);