[Intel-gfx] [PATCH] drm/i915/psr: Enable PSR1 by default on gen9+ platforms

[Rodrigo Vivi]

On Wed, Jul 25, 2018 at 09:52:44AM -0700, Dhinakaran Pandiyan wrote:
> On Wed, 2018-07-25 at 09:12 -0700, Rodrigo Vivi wrote:
> > On Wed, Jul 25, 2018 at 12:22:28AM -0700, Dhinakaran Pandiyan wrote:
> > > 
> > > We have merged several fixes, re-written some tests and improved
> > > debug
> > > capability in the past several months, so this is a good time to
> > > give PSR1
> > > another try. PSR1 has not been tested on HSW and BDW recently, so
> > > let's
> > > enable only on gen9+ now.
> > > 
> > > Cc: Rodigo Vivi <rodrigo.vivi at intel.com>
> > > Cc: José Roberto de Souza <jose.souza at intel.com>
> > > Signed-off-by: Dhinakaran Pandiyan <dhinakaran.pandiyan at intel.com>
> > > ---
> > >  drivers/gpu/drm/i915/intel_psr.c | 20 ++++++++++----------
> > >  1 file changed, 10 insertions(+), 10 deletions(-)
> > > 
> > > diff --git a/drivers/gpu/drm/i915/intel_psr.c
> > > b/drivers/gpu/drm/i915/intel_psr.c
> > > index 4bd5768731ee..942db85da6a1 100644
> > > --- a/drivers/gpu/drm/i915/intel_psr.c
> > > +++ b/drivers/gpu/drm/i915/intel_psr.c
> > > @@ -471,10 +471,8 @@ void intel_psr_compute_config(struct intel_dp
> > > *intel_dp,
> > >  	if (!CAN_PSR(dev_priv))
> > >  		return;
> > >  
> > > -	if (!i915_modparams.enable_psr) {
> > > -		DRM_DEBUG_KMS("PSR disable by flag\n");
> > Why are you removing the message?
> > I think it is still useful... and enable_psr == -1 doesn't trigger
> > that.
> > 
> The text was a bit vague to start with, and is confusing when combined
> with this patch. Agreed, it is useful to have a debug message, I'll
> replace it.
> 
> > > 
> > > +	if (!i915_modparams.enable_psr)
> > >  		return;
> > > -	}
> > >  
> > >  	/*
> > >  	 * HSW spec explicitly says PSR is tied to port A.
> > > @@ -516,7 +514,11 @@ void intel_psr_compute_config(struct intel_dp
> > > *intel_dp,
> > >  	}
> > >  
> > >  	crtc_state->has_psr = true;
> > > -	crtc_state->has_psr2 = intel_psr2_config_valid(intel_dp,
> > > crtc_state);
> > > +
> > > +	/* Enable only PSR 1 by default for now */
> > > +	crtc_state->has_psr2 = i915_modparams.enable_psr == 1 &&
> > > +			       intel_psr2_config_valid(intel_dp,
> > > crtc_state);
> > > +
> > this might get confusing...
> > -1 - enable psr1
> > 0 - disable
> > 1 - enable psr2
> > 
> > and far from the variable... Well... I want to kill the parameter
> > anyways
> > so no hard feelings on having this here, but what about some debug
> > messages
> > at least?
> > 
> > /* Enable only PSR1 by default for now */
> > if (i915_modparams.enable_psr == -1) {
> > 	DRM_DEBUG_KMS("Avoiding PSR2 by platform default")
> > 	crtc_state->has_psr2 = 0;
> > } else {
> > 	crtc_state->has_psr2 = intel_psr2_config_valid(intel_dp,
> > crtc_state);
> > }
> > 
> 
> The reason I added a check for i915.enable_psr==1 was to enable PSR2
> only when the user passes the exact value. Otherwise, we should fall
> back to default.

well, it could be == 1 check inverting my block here...
but my main point is to have some kind of debug message ;)

> 
> > > 
> > >  	DRM_DEBUG_KMS("Enabling PSR%s\n", crtc_state->has_psr2 ?
> > > "2" : "");
> > >  }
> > >  
> > > @@ -956,12 +958,10 @@ void intel_psr_init(struct drm_i915_private
> > > *dev_priv)
> > >  	if (!dev_priv->psr.sink_support)
> > >  		return;
> > >  
> > > -	if (i915_modparams.enable_psr == -1) {
> > > -		i915_modparams.enable_psr = dev_priv-
> > > >vbt.psr.enable;
> > > -
> > > -		/* Per platform default: all disabled. */
> > > -		i915_modparams.enable_psr = 0;
> > > -	}
> > > +	/* Enable PSR 1 default only on gen9+ */
> > > +	if (i915_modparams.enable_psr == -1)
> > > +		if (INTEL_GEN(dev_priv) < 9 || !dev_priv-
> > > >vbt.psr.enable)
> > > +			i915_modparams.enable_psr = 0;
> > we talked about this in person, but just for the record:
> > we need to check cnl and icl on CI for psr cases before make this >
> > 9.
> 
> The failures on ICL are due to an unrelated debug warning. The CNL ones
> are interesting, most likely due to us enabling PSR2 by setting the
> module parameter=1 from the IGTs. But, it still should not be failing,
> I'll check.
> 
> 
> 
> > 
> > > 
> > > 
> > >  	/* Set link_standby x link_off defaults */
> > >  	if (IS_HASWELL(dev_priv) || IS_BROADWELL(dev_priv))