Re: [PATCH v2] drm/i915/psr: Force fastset on debugfs write for each connector supporting PSR

On Tue, 2026-06-23 at 06:24 +0000, Garg, Nemesa wrote:
> 
> 
> > -----Original Message-----
> > From: Garg, Nemesa
> > Sent: Tuesday, June 23, 2026 11:18 AM
> > To: 'Jouni Högander' <[email protected]>; intel-
> > [email protected]; [email protected]
> > Cc: Hogander, Jouni <[email protected]>
> > Subject: RE: [PATCH v2] drm/i915/psr: Force fastset on debugfs
> > write for each
> > connector supporting PSR
> > 
> > 
> > 
> > > -----Original Message-----
> > > From: Intel-gfx <[email protected]> On
> > > Behalf Of
> > > Jouni Högander
> > > Sent: Thursday, June 18, 2026 3:22 PM
> > > To: [email protected];
> > > [email protected]
> > > Cc: Hogander, Jouni <[email protected]>
> > > Subject: [PATCH v2] drm/i915/psr: Force fastset on debugfs write
> > > for
> > > each connector supporting PSR
> > > 
> > > Currently intel_psr_fast_set_force is forcing fastset only for
> > > eDP connectors.
> > > Panel Replay is supported on DP as well and writing
> > > i915_edp_psr_debug
> > > is setting debug parameters for each DP transcoder (struct
> > intel_dp:psr:debug).
> > > Force fastset for each connector supporting PSR or Panel Replay.
> > > While
> > > modifying the code also switch to for_each_intel_connector_iter.
> > > 
> > > v2:
> > >   - use for_each_intel_connector_iter
> > >   - update and improve commit message
> > > 
> > > Signed-off-by: Jouni Högander <[email protected]>
> > > ---
> > >  drivers/gpu/drm/i915/display/intel_psr.c | 9 +++++----
> > >  1 file changed, 5 insertions(+), 4 deletions(-)
> > > 
> > > diff --git a/drivers/gpu/drm/i915/display/intel_psr.c
> > > b/drivers/gpu/drm/i915/display/intel_psr.c
> > > index 911afb9cb24e3..622edfb6a7fb2 100644
> > > --- a/drivers/gpu/drm/i915/display/intel_psr.c
> > > +++ b/drivers/gpu/drm/i915/display/intel_psr.c
> > > @@ -3375,7 +3375,7 @@ static int intel_psr_fastset_force(struct
> > > intel_display
> > > *display)
> > >  	struct drm_connector_list_iter conn_iter;
> > >  	struct drm_modeset_acquire_ctx ctx;
> > >  	struct drm_atomic_commit *state;
> > > -	struct drm_connector *conn;
> > > +	struct intel_connector *connector;
> > >  	int err = 0;
> > > 
> > >  	state = drm_atomic_commit_alloc(display->drm);
> > > @@ -3389,14 +3389,15 @@ static int intel_psr_fastset_force(struct
> > > intel_display *display)
> > > 
> > >  retry:
> > >  	drm_connector_list_iter_begin(display->drm, &conn_iter);
> > > -	drm_for_each_connector_iter(conn, &conn_iter) {
> > > +	for_each_intel_connector_iter(connector, &conn_iter) {
> > >  		struct drm_connector_state *conn_state;
> > >  		struct drm_crtc_state *crtc_state;
> > > 
> > > -		if (conn->connector_type !=
> > > DRM_MODE_CONNECTOR_eDP)
> > > +		if (!connector->dp.psr_caps.support &&
> > > +		    !connector->dp.panel_replay_caps.support)
> > >  			continue;
> > > 
> > > -		conn_state =
> > > drm_atomic_get_connector_state(state, conn);
> > > +		conn_state =
> > > drm_atomic_get_connector_state(state,
> > > &connector->base);
> > >  		if (IS_ERR(conn_state)) {
> > >  			err = PTR_ERR(conn_state);
> > >  			break;
> > > --
> > LGTM,
> > Reviewed-by: Nemesa Garg <[email protected]>
> > 
> Just a thought should we add a link training check before fastset to
> ensure the link is established.

I think lock taken by drm_atomic_get_connector_state is ensuring there
isn't ongoing link training.

BR,
Jouni Högander

> 
> > > 2.43.0
>