[PATCH] drm/msm/dp: move link_ready out of HPD event thread

Johan Hovold johan at kernel.org
Tue Mar 12 16:59:56 UTC 2024


On Tue, Mar 12, 2024 at 05:41:23PM +0100, Johan Hovold wrote:
> On Tue, Mar 12, 2024 at 11:09:11AM +0100, Johan Hovold wrote:
> > On Fri, Mar 08, 2024 at 01:45:32PM -0800, Abhinav Kumar wrote:
> 
> > > @@ -466,6 +466,8 @@ static int dp_display_notify_disconnect(struct device *dev)
> > >  {
> > >  	struct dp_display_private *dp = dev_get_dp_display_private(dev);
> > >  
> > > +	dp->dp_display.link_ready = false;
> > 
> > As I also pointed out in the other thread, setting link_ready to false
> > here means that any spurious connect event (during physical disconnect)
> > will always be processed, something which can currently lead to a leaked
> > runtime pm reference.
> > 
> > Wasting some power is of course preferred over crashing the machine, but
> > please take it into consideration anyway.
> > 
> > Especially if your intention with this patch was to address the resets
> > we saw with sc8280xp which are gone since the HPD notify revert (which
> > fixed the hotplug detect issue that left the bridge in a
> > half-initialised state).
> 
> Heh. This is getting ridiculous. I just tried running with this patch
> and it again breaks hotplug detect in a VT console and in X (where I
> could enable a reconnected external display by running xrandr twice
> before).
> 
> So, please, do not apply this one.

To make things worse, I indeed also hit the reset when disconnecting
after such a failed hotplug.

Johan


More information about the dri-devel mailing list