Re: [Intel-gfx] [PATCH 1/3] drm/i915: Reject bigjoiner if the pipe doesn't support it

["Ville Syrjälä" <ville.syrjala@linux.intel.com>]

On Fri, Jan 21, 2022 at 05:55:24PM +0200, Ville Syrjälä wrote:
> On Fri, Jan 21, 2022 at 04:03:09PM +0200, Jani Nikula wrote:
> > On Fri, 21 Jan 2022, Ville Syrjala <ville.syrjala@linux.intel.com> wrote:
> > > From: Ville Syrjälä <ville.syrjala@linux.intel.com>
> > >
> > > Check that our crtc can in fact be the bigjoiner master before
> > > we let the modeset proceed with bigjoiner enabled.
> > >
> > > Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
> > > ---
> > >  drivers/gpu/drm/i915/display/intel_display.c | 16 ++++++++++++++++
> > >  1 file changed, 16 insertions(+)
> > >
> > > diff --git a/drivers/gpu/drm/i915/display/intel_display.c b/drivers/gpu/drm/i915/display/intel_display.c
> > > index 0964b2403e2d..36e547bd0cbe 100644
> > > --- a/drivers/gpu/drm/i915/display/intel_display.c
> > > +++ b/drivers/gpu/drm/i915/display/intel_display.c
> > > @@ -4102,6 +4102,14 @@ static u8 bigjoiner_pipes(struct drm_i915_private *i915)
> > >  		return 0;
> > >  }
> > >  
> > > +static u8 bigjoiner_master_pipes(struct drm_i915_private *i915)
> > > +{
> > > +	u8 pipes = bigjoiner_pipes(i915);
> > > +
> > > +	/* last pipe can not be master */
> > > +	return pipes & (pipes >> 1);
> > > +}
> > > +
> > >  static bool transcoder_ddi_func_is_enabled(struct drm_i915_private *dev_priv,
> > >  					   enum transcoder cpu_transcoder)
> > >  {
> > > @@ -7600,6 +7608,7 @@ static int intel_atomic_check_bigjoiner(struct intel_atomic_state *state,
> > >  					struct intel_crtc_state *old_crtc_state,
> > >  					struct intel_crtc_state *new_crtc_state)
> > >  {
> > > +	struct drm_i915_private *i915 = to_i915(state->base.dev);
> > >  	struct intel_crtc_state *slave_crtc_state, *master_crtc_state;
> > >  	struct intel_crtc *slave_crtc, *master_crtc;
> > >  
> > > @@ -7615,6 +7624,13 @@ static int intel_atomic_check_bigjoiner(struct intel_atomic_state *state,
> > >  	if (!new_crtc_state->bigjoiner)
> > >  		return 0;
> > >  
> > > +	if ((bigjoiner_master_pipes(i915) & BIT(crtc->pipe)) == 0) {
> > 
> > Feels like the check should be in
> > intel_dsc_get_bigjoiner_{secondary,primary}.
> 
> The master pipe is selected by userspace. intel_dsc_get_bigjoiner_primary()
> is not relevant here.

Hmm. I guess the intel_dsc_get_bigjoiner_secondary()->NULL is
in fact sufficient on current hardware since only pipe A can
be non-joinable, and the dsc code already rejects the transcoder A
case. Although it only claims to do that due to it not supporting
dsc, but luckily that happens to be the same thing as no-joiner since
there is no uncompressed joiner either.

But to be 100% clear should probably do something along the lines
proposed here. I'm actually think we nuke all these crtc pointers/etc.
and just go for some kind of pipe bitmask based approach everywhere.
So this probaly needs a bit more through all around.

-- 
Ville Syrjälä
Intel