[Intel-gfx] [PATCH v4 5/6] drm/i915/display: Check if pipe fastset is allowed by external dependencies

[Ville Syrjälä]

On Thu, Dec 19, 2019 at 08:39:26PM +0000, Souza, Jose wrote:
> On Wed, 2019-12-18 at 20:33 +0000, Souza, Jose wrote:
> > On Wed, 2019-12-18 at 22:26 +0200, Ville Syrjälä wrote:
> > > On Wed, Dec 18, 2019 at 08:19:15PM +0000, Souza, Jose wrote:
> > > > On Wed, 2019-12-18 at 21:39 +0200, Ville Syrjälä wrote:
> > > > > On Wed, Dec 18, 2019 at 10:59:09AM -0800, José Roberto de Souza
> > > > > wrote:
> > > > > > Check if fastset is allowed by external dependencies like
> > > > > > other
> > > > > > pipes
> > > > > > and transcoders.
> > > > > > 
> > > > > > Right now this patch only forces a fullmodeset in MST slaves
> > > > > > of
> > > > > > MST
> > > > > > masters that needs a fullmodeset but it will be needed for
> > > > > > port
> > > > > > sync
> > > > > > as well.
> > > > > > 
> > > > > > v3:
> > > > > > - moved handling to intel_atomic_check() this way is
> > > > > > guarantee
> > > > > > that
> > > > > > all pipes will have its state computed
> > > > > > 
> > > > > > v4:
> > > > > > - added a function to return if MST master neeeds modeset to
> > > > > > simply
> > > > > > code in intel_atomic_check()
> > > > > > 
> > > > > > 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 | 53
> > > > > > +++++++++++++++-
> > > > > > ----
> > > > > >  drivers/gpu/drm/i915/display/intel_dp_mst.c  | 14 ++++++
> > > > > >  drivers/gpu/drm/i915/display/intel_dp_mst.h  |  3 ++
> > > > > >  3 files changed, 57 insertions(+), 13 deletions(-)
> > > > > > 
> > > > > > diff --git a/drivers/gpu/drm/i915/display/intel_display.c
> > > > > > b/drivers/gpu/drm/i915/display/intel_display.c
> > > > > > index a4f252d05b37..2a406891567b 100644
> > > > > > --- a/drivers/gpu/drm/i915/display/intel_display.c
> > > > > > +++ b/drivers/gpu/drm/i915/display/intel_display.c
> > > > > > @@ -13903,19 +13903,6 @@ static void
> > > > > > intel_crtc_check_fastset(const
> > > > > > struct intel_crtc_state *old_crtc_sta
> > > > > >  
> > > > > >  	new_crtc_state->uapi.mode_changed = false;
> > > > > >  	new_crtc_state->update_pipe = true;
> > > > > > -
> > > > > > -	/*
> > > > > > -	 * If we're not doing the full modeset we want to
> > > > > > -	 * keep the current M/N values as they may be
> > > > > > -	 * sufficiently different to the computed values
> > > > > > -	 * to cause problems.
> > > > > > -	 *
> > > > > > -	 * FIXME: should really copy more fuzzy state here
> > > > > > -	 */
> > > > > > -	new_crtc_state->fdi_m_n = old_crtc_state->fdi_m_n;
> > > > > > -	new_crtc_state->dp_m_n = old_crtc_state->dp_m_n;
> > > > > > -	new_crtc_state->dp_m2_n2 = old_crtc_state->dp_m2_n2;
> > > > > > -	new_crtc_state->has_drrs = old_crtc_state->has_drrs;
> > > > > >  }
> > > > > >  
> > > > > >  static int intel_crtc_add_planes_to_state(struct
> > > > > > intel_atomic_state *state,
> > > > > > @@ -14083,6 +14070,46 @@ static int intel_atomic_check(struct
> > > > > > drm_device *dev,
> > > > > >  			any_ms = true;
> > > > > >  	}
> > > > > >  
> > > > > > +	/**
> > > > > > +	 * 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.
> > > > > > +	 */
> > > > > > +	for_each_new_intel_crtc_in_state(state, crtc,
> > > > > > new_crtc_state,
> > > > > > i) {
> > > > > > +		enum transcoder master = new_crtc_state-
> > > > > > > mst_master_transcoder;
> > > > > > +
> > > > > > +		if
> > > > > > (!intel_dp_mst_is_slave_trans(new_crtc_state) ||
> > > > > > +		    needs_modeset(new_crtc_state))
> > > > > > +			continue;
> > > > > > +
> > > > > > +		if
> > > > > > (intel_dp_mst_master_trans_needs_modeset(state,
> > > > > > master)) {
> > > > > 
> > > > > I think this has the loops the opposite way of what I was
> > > > > thinking,
> > > > > but should work fine I think... OK. I'm convinced your way is
> > > > > in
> > > > > fact
> > > > > better.
> > > > > 
> > > > > > +			new_crtc_state->uapi.mode_changed =
> > > > > > true;
> > > > > > +			new_crtc_state->update_pipe = false;
> > > > > > +		}
> > > > > > +	}
> > > > > > +
> > > > > > +	for_each_oldnew_intel_crtc_in_state(state, crtc,
> > > > > > old_crtc_state,
> > > > > > +					    new_crtc_state, i)
> > > > > > {
> > > > > > +		if (needs_modeset(new_crtc_state))
> > > > > > +			continue;
> > > > > 
> > > > > I suppose there isn't any way we should have crtcs in the state
> > > > > that
> > > > > neither have update_pipe or needs_modeset flagged here. Could
> > > > > maybe
> > > > > WARN_ON(!update_pipe) here if we're being paranoid.
> > > > 
> > > > Adding the WARN_ON
> 
> Ah we can't add this WARN_ON(), is most atomic states the previous CRTC
> states is added even if it dont need any modification.
> 
> Those are handled here:
> 
> 	for_each_oldnew_intel_crtc_in_state(state, crtc,
> old_crtc_state,
> 					    new_crtc_state, i) {
> 		if (!needs_modeset(new_crtc_state)) {
> 			/* Light copy */
> 			intel_crtc_copy_uapi_to_hw_state_nomodeset(new_
> crtc_state);
> 
> 			continue;
> 		}

After further consideration we only want the copy for fastset crtcs
(ie. ones with update_pipe==true). Crtcs added for pure plane updates
should not do the copy (well, in that case the old and new states
really should match, but I still think it's a bit dubious). So
I guess we should check for '!needs_modeset() && update_pipe'.

-- 
Ville Syrjälä
Intel