[Intel-gfx] [PATCH v2 rebased 10/11] drm/i915/display: Check if pipe fastset is allowed by external dependencies

[Souza, Jose]

On Thu, 2019-12-12 at 13:41 -0800, Manasi Navare wrote:
> On Thu, Dec 12, 2019 at 11:28:30PM +0200, Ville Syrjälä wrote:
> > On Wed, Dec 11, 2019 at 10:45:25AM -0800, José Roberto de Souza
> > wrote:
> > > Check if fastset is allowed by external dependencies like other
> > > pipes
> > > and transcoders.
> > > 
> > > Right now it only forces a fullmodeset when the MST master
> > > transcoder
> > > did not changed but the pipe of the master transcoder needs a
> > > fullmodeset so all slaves also needs to do a fullmodeset.
> > > But it will probably be need for port sync as well.
> 
> But in case of port sync, now I have this function in
> intel_atomic_check()
> that will force a full modeset on all the connectors with has_tile if
> one of
> the tiled connectors needs a full modeset.
> So we wouldnt need to worry about fastset happening for one of the
> tiles in port sync.
> 
> This patch:
> https://patchwork.freedesktop.org/patch/344766/?series=70788&rev=1

This is not enough, test it or read the context of
intel_crtc_check_fastset()

> 
> Manasi
> 
> > > Cc: Ville Syrjälä <ville.syrjala at linux.intel.com>
> > > Cc: Lucas De Marchi <lucas.demarchi at intel.com>
> > > Cc: Maarten Lankhorst <maarten.lankhorst at linux.intel.com>
> > > Cc: Manasi Navare <manasi.d.navare at intel.com>
> > > Signed-off-by: José Roberto de Souza <jose.souza at intel.com>
> > > ---
> > >  drivers/gpu/drm/i915/display/intel_display.c | 41
> > > ++++++++++++++++++++
> > >  1 file changed, 41 insertions(+)
> > > 
> > > diff --git a/drivers/gpu/drm/i915/display/intel_display.c
> > > b/drivers/gpu/drm/i915/display/intel_display.c
> > > index 092412b10d7c..0c24d7dfa152 100644
> > > --- a/drivers/gpu/drm/i915/display/intel_display.c
> > > +++ b/drivers/gpu/drm/i915/display/intel_display.c
> > > @@ -13930,11 +13930,52 @@ static int calc_watermark_data(struct
> > > intel_atomic_state *state)
> > >  	return 0;
> > >  }
> > >  
> > > +/**
> > > + * Check if fastset is allowed by external dependencies like
> > > other pipes and
> > > + * transcoders.
> > > + *
> > > + * Right now it only forces a fullmodeset when the MST master
> > > transcoder did
> > > + * not changed but the pipe of the master transcoder needs a
> > > fullmodeset so
> > > + * all slaves also needs to do a fullmodeset.
> > > + */
> > > +static bool
> > > +intel_crtc_check_external_dependencies_fastset(const struct
> > > intel_crtc_state *old_crtc_state,
> > > +					       struct intel_crtc_state
> > > *new_crtc_state)
> > > +{
> > > +	struct intel_atomic_state *state =
> > > to_intel_atomic_state(new_crtc_state->uapi.state);
> > > +	struct drm_i915_private *dev_priv = to_i915(new_crtc_state-
> > > >uapi.crtc->dev);
> > > +	struct intel_crtc_state *new_crtc_state_iter;
> > > +	struct intel_crtc *crtc_iter;
> > > +	int i;
> > > +
> > > +	if (INTEL_GEN(dev_priv) < 12)
> > > +		return true;
> > > +
> > > +	if (!intel_crtc_has_type(old_crtc_state, INTEL_OUTPUT_DP_MST)
> > > ||
> > > +	    intel_dp_mst_is_master_trans(new_crtc_state))
> > > +		return true;
> > > +
> > > +	for_each_new_intel_crtc_in_state(state, crtc_iter,
> > > new_crtc_state_iter, i) {
> > > +		if (new_crtc_state_iter->cpu_transcoder !=
> > > +		    new_crtc_state->mst_master_transcoder)
> > > +			continue;
> > > +
> > > +		return !needs_modeset(new_crtc_state_iter);
> > > +	}
> > > +
> > > +	DRM_ERROR("Master MST transcoder of pipe not found\n");
> > > +	return false;
> > > +}
> > > +
> > >  static void intel_crtc_check_fastset(const struct
> > > intel_crtc_state *old_crtc_state,
> > >  				     struct intel_crtc_state
> > > *new_crtc_state)
> > >  {
> > >  	if (!intel_pipe_config_compare(old_crtc_state, new_crtc_state,
> > > true))
> > >  		return;
> > > +	if
> > > (!intel_crtc_check_external_dependencies_fastset(old_crtc_state,
> > > +							    new_crtc_st
> > > ate))
> > > +		return;
> > 
> > I don't think this will work. We've not yet .compute_config()'d
> > everything so the master assignments are still up in the air.
> > 
> > I think we need the logic higher up in intel_atomic_check():
> > 
> > for_each_crtc()
> > 	compute_config()
> > 	if (can_fastset())
> > 		needs_modeset=false
> > 		update_pipe=true;
> > 	}
> > }
> > 
> > for_each_crtc()
> > 	if (mst_slave && master.needs_modeset() {
> > 		needs_modeset=true;
> > 		update_pipe=false;
> > 	}
> > 
> > for_each_crtc()
> > 	if (update_pipe)
> > 		copy_over_old_state();
> > 
> > 	
> > + we should probably rename/split update_pipes or add some
> > lkind of needs_fastset() wrapper to make this less confusing.
> > 
> > 
> > > +
> > >  
> > >  	new_crtc_state->uapi.mode_changed = false;
> > >  	new_crtc_state->update_pipe = true;
> > > -- 
> > > 2.24.1
> > 
> > -- 
> > Ville Syrjälä
> > Intel