On Tue, Jun 14, 2022 at 03:55:04PM +0300, Hogander, Jouni wrote:
> On Tue, 2022-06-14 at 15:22 +0300, Stanislav Lisovskiy wrote:
> > We seem to enable PSR2 and selective fetch even if there are no
> > active
> > planes. That seems to causes FIFO underruns at least for ADLP.
> > Those are gone if we don't do that. Just adding simple check
> > in intel_psr2_sel_fetch_config_valid seems to do the trick.
>
> We are already disabling PSR intel_psr_pre_plane_update if
> active_planes is 0.
>
> We are also checking active_planes in _intel_psr_post_plane_update and
> not enabling PSR if it's 0.
>
> So I'm now wondering what sequence this patch is actually changing?
> I.e. where PSR is currently enabled/not disabled if active_planes == 0?
Good question! Apparently we still do it, because without this change
we get FIFO underruns, while with that one we don't.
I have suspicion that this happens during modeset, however you are right
we need to know for sure.
I will get back here and post the exact call trace.
One thing I can say for sure that we do it somewhere, otherwise adding
this wouldn't have any effect.
Stan
>
> > Signed-off-by: Stanislav Lisovskiy <stanislav.lisovskiy@xxxxxxxxx>
> > ---
> > drivers/gpu/drm/i915/display/intel_psr.c | 6 ++++++
> > 1 file changed, 6 insertions(+)
> >
> > diff --git a/drivers/gpu/drm/i915/display/intel_psr.c
> > b/drivers/gpu/drm/i915/display/intel_psr.c
> > index 7d61c55184e5..03add69cfdca 100644
> > --- a/drivers/gpu/drm/i915/display/intel_psr.c
> > +++ b/drivers/gpu/drm/i915/display/intel_psr.c
> > @@ -747,6 +747,12 @@ static bool
> > intel_psr2_sel_fetch_config_valid(struct intel_dp *intel_dp,
> > return false;
> > }
> >
> > + if (hweight32(crtc_state->active_planes) == 0) {
> > + drm_dbg_kms(&dev_priv->drm,
> > + "PSR2 sel fetch not enabled, no
> > active_planes\n");
> > + return false;
> > + }
> > +
> > /* Wa_14010254185 Wa_14010103792 */
> > if (IS_TGL_DISPLAY_STEP(dev_priv, STEP_A0, STEP_C0)) {
> > drm_dbg_kms(&dev_priv->drm,
>
