[Intel-gfx] [PATCH v4 1/2] drm: Add drm_any_plane_has_format()

[Daniel Vetter]

On Tue, Oct 30, 2018 at 04:18:28PM +0200, Ville Syrjälä wrote:
> On Tue, Oct 30, 2018 at 10:35:07AM +0100, Daniel Vetter wrote:
> > On Mon, Oct 29, 2018 at 04:00:04PM -0700, Eric Anholt wrote:
> > > Ville Syrjala <ville.syrjala at linux.intel.com> writes:
> > > 
> > > > From: Ville Syrjälä <ville.syrjala at linux.intel.com>
> > > >
> > > > Add a function to check whether there is at least one plane that
> > > > supports a specific format and modifier combination. Drivers can
> > > > use this to reject unsupported formats/modifiers in .fb_create().
> > > >
> > > > v2: Accept anyformat if the driver doesn't do planes (Eric)
> > > >     s/planes_have_format/any_plane_has_format/ (Eric)
> > > >     Check the modifier as well since we already have a function
> > > >     that does both
> > > > v3: Don't do the check in the core since we may not know the
> > > >     modifier yet, instead export the function and let drivers
> > > >     call it themselves
> > > >
> > > > Cc: Eric Anholt <eric at anholt.net>
> > > > Cc: Dhinakaran Pandiyan <dhinakaran.pandiyan at intel.com>
> > > > Signed-off-by: Ville Syrjälä <ville.syrjala at linux.intel.com>
> > > > Reviewed-by: Dhinakaran Pandiyan <dhinakaran.pandiyan at intel.com>
> > > 
> > > I don't particularly see the point in having FB creation duplicate the
> > > validation that atomic check will eventually do, and it means that FB
> > > creation cost scales with plane count, but if i915's going to do this,
> > > it seems reasonable for them.
> > 
> > atomic_check checks for a given plane only, I do think it makes sense to
> > make sure you can't create framebuffers that are impossible to use on a
> > given driver at addfb time.
> > 
> > In case the overhead is ever critical, we could compile a static map of
> > this at driver load time, and then check that.
> > 
> > Aside: Shouldn't we make this the default for atomic drivers? With
> > atomic drivers we can assume that all planes have valid format lists
> > (because atomic_check checks them already). Only with non-atomic drivers,
> > how might have a faked primary plane is this not a valid assumption ...
> 
> The problem of making this automagic for everyone was the
> legacy tiling->modifier thing.
> 
> https://patchwork.freedesktop.org/patch/210193/ +
> https://patchwork.freedesktop.org/patch/208070/
> is the best I could really come up with when I tried to make this
> entirely automagic. I haven't bothered to revisit those because I
> wasn't entirely happy with needing the extra vfunc, and people
> didn't seem too excited about this idea.

Hm yeah for full format+modifier checking we need a hook. For format-only
checking we can get by with requiring atomic only I think.

Anyway, kinda orthogonal to all this.
-Daniel
-- 
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch