Re: [PATCH 5/8] drm/i915/psr: Clear pr_dpcd as well on disconnect

[Imre Deak <imre.deak@intel.com>]

On Mon, Dec 01, 2025 at 01:23:41PM +0200, Jouni Hogander wrote:
> On Mon, 2025-12-01 at 12:53 +0200, Imre Deak wrote:
> > On Fri, Nov 21, 2025 at 01:16:52PM +0200, Jouni Högander wrote:
> > > Currently we are leaving pr_dpcd containing Panel Replay capability
> > > DPCD registers as it is on disconnect. Clear it as well on disconnect.
> > > 
> > > Signed-off-by: Jouni Högander <jouni.hogander@intel.com>
> > > ---
> > >  drivers/gpu/drm/i915/display/intel_dp.c | 1 +
> > >  1 file changed, 1 insertion(+)
> > > 
> > > diff --git a/drivers/gpu/drm/i915/display/intel_dp.c
> > > b/drivers/gpu/drm/i915/display/intel_dp.c
> > > index 62808cd35f5f2..7195c408d93ab 100644
> > > --- a/drivers/gpu/drm/i915/display/intel_dp.c
> > > +++ b/drivers/gpu/drm/i915/display/intel_dp.c
> > > @@ -6050,6 +6050,7 @@ intel_dp_detect(struct drm_connector *_connector,
> > >  	if (status == connector_status_disconnected) {
> > >  		intel_dp_test_reset(intel_dp);
> > >  		memset(connector->dp.dsc_dpcd, 0, sizeof(connector->dp.dsc_dpcd));
> > > +		memset(connector->dp.pr_dpcd, 0, sizeof(connector->dp.pr_dpcd));
> > 
> > What about psr_dpcd?
> 
> PSR is only for eDP and can't be disconnected.

Ok. Panel Replay used on non-eDP has also Selective Update for instance,
but that has separate PR specific capability registers. The same is true
I presume for any other functionality that could be used both on PSR2
and PR.

It's still a bit strange that intel_dp_detect() -> intel_psr_init_dpcd()
-> _psr_init_dpcd() reads psr_dpcd (aka dp.psr_caps.dpcd after your
patch) for non-eDP as well. Even though the values should not be used
anywhere based on the above, I'd still avoid explicitly reading them out
for non-eDP (only as a follow-up imo, no need to change this now).

> > All these resetting of the caps cause a problem if the connector
> > needs to be modeset after the sink is disconnected (since then the
> > state computation for the connector will fail seeing these caps
> > being reset). Instead the caps should be kept intact here,
> > resetting/reiniting them only when a new sink is connected. Since
> > this is a pre-existing issue, could you add for now a corrsponding
> > FIXME: comment in this patch?
> 
> Thank you for pointing this out. I will add the FIXME.
> 
> BR,
> 
> Jouni Högander
> 
> > 
> > >  		intel_dp->psr.sink_panel_replay_support = false;
> > >  		intel_dp->psr.sink_panel_replay_su_support = false;
> > >  		intel_dp->psr.sink_panel_replay_dsc_support =
> > > -- 
> > > 2.43.0