Re: [Intel-gfx] [PATCH 1/6] drm/i915: Fix enabled_planes bitmask

From: "Ville Syrjälä" <ville.syrjala@linux.intel.com>
To: "Navare, Manasi" <manasi.d.navare@intel.com>
Cc: intel-gfx@lists.freedesktop.org
Subject: Re: [Intel-gfx] [PATCH 1/6] drm/i915: Fix enabled_planes bitmask
Date: Fri, 19 Mar 2021 23:20:07 +0200	[thread overview]
Message-ID: <YFUVh6Iujpsxo8Rm@intel.com> (raw)
In-Reply-To: <20210319211710.GB6359@labuser-Z97X-UD5H>

On Fri, Mar 19, 2021 at 02:17:18PM -0700, Navare, Manasi wrote:
> On Fri, Mar 05, 2021 at 05:36:05PM +0200, Ville Syrjala wrote:
> > From: Ville Syrjälä <ville.syrjala@linux.intel.com>
> > 
> > The enabled_planes bitmask was supposed to track logically enabled
> > planes (ie. fb!=NULL and crtc!=NULL), but instead we end up putting
> > even disabled planes into the bitmask since
> > intel_plane_atomic_check_with_state() only takes the early exit
> > if the plane was disabled and stays disabled. I think I misread
> > the early said codepath to exit whenever the plane is logically
> > disabled, which is not true.
> > 
> > So let's fix this up properly and set the bit only when the plane
> > actually is logically enabled.
> 
> Hmm yes makes sense, I guess I added the check in skl_plane_check_fb to return if !fb
> but I think thats a bug there since I return 0 there instead of return a -EINVAL or something.

No. Return 0 is the correct there. It just means we have nothing to
check, and all is well. I suppose we could move the whole thing later
and not have the fb check at all, but I kinda like doing that early 
to make sure no funky stuff leaks into the later checks.

> 
> Because if we return a negative value there then if !fb, skl_plane_check will return that
> and we will return from plane->check_plane so just moving the enabled planes bitmask assignment
> to after the check_plane will do no need to check for if (fb) there again right?
> 
> Manasi
> 
> > 
> > Cc: Manasi Navare <manasi.d.navare@intel.com>
> > Fixes: ee42ec19ca2e ("drm/i915: Track logically enabled planes for hw state")
> > Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
> > ---
> >  drivers/gpu/drm/i915/display/intel_atomic_plane.c | 5 +++--
> >  1 file changed, 3 insertions(+), 2 deletions(-)
> > 
> > diff --git a/drivers/gpu/drm/i915/display/intel_atomic_plane.c b/drivers/gpu/drm/i915/display/intel_atomic_plane.c
> > index 4683f98f7e54..c3f2962aa1eb 100644
> > --- a/drivers/gpu/drm/i915/display/intel_atomic_plane.c
> > +++ b/drivers/gpu/drm/i915/display/intel_atomic_plane.c
> > @@ -317,12 +317,13 @@ int intel_plane_atomic_check_with_state(const struct intel_crtc_state *old_crtc_
> >  	if (!new_plane_state->hw.crtc && !old_plane_state->hw.crtc)
> >  		return 0;
> >  
> > -	new_crtc_state->enabled_planes |= BIT(plane->id);
> > -
> >  	ret = plane->check_plane(new_crtc_state, new_plane_state);
> >  	if (ret)
> >  		return ret;
> >  
> > +	if (fb)
> > +		new_crtc_state->enabled_planes |= BIT(plane->id);
> > +
> >  	/* FIXME pre-g4x don't work like this */
> >  	if (new_plane_state->uapi.visible)
> >  		new_crtc_state->active_planes |= BIT(plane->id);
> > -- 
> > 2.26.2
> > 

-- 
Ville Syrjälä
Intel
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx