[Intel-gfx] [RFC] drm/i915: Restrict legacy color key ioctl to pre-gen12

[Chris Wilson]

Quoting Matt Roper (2020-01-14 23:03:20)
> On Tue, Jan 14, 2020 at 10:54:50PM +0000, Chris Wilson wrote:
> > Quoting Chris Wilson (2020-01-14 22:47:58)
> > > Quoting Matt Roper (2020-01-14 22:45:08)
> > > > Since gen12 platform support isn't finalized yet, let's kill off the
> > > > legacy color key ioctl for this platform; there's no userspace today
> > > > that can run on this platform that utilizes this legacy ioctl, so we can
> > > > safely kill it now before it becomes ABI.
> > > > 
> > > > Color key functionality never got integrated into the property / atomic
> > > > interface, and the only known open source consumer was the Intel DDX
> > > > which was never updated to run on platforms beyond gen9.  If color
> > > > keying is desired going forward, it should really be exposed as a
> > > > property so that it can be applied atomically with other display updates
> > > > (and should probably be standardized in a way all drivers can choose to
> > > > support rather than being i915-specific).
> > > 
> > > But it does run on those platforms and exposes the sprite plane via Xv.
> > 
> 
> Hmm, looks like I overlooked 00184dc03 ("Sync i915_pciids upto
> d0e062ebb3a4") which finally brought in the CNL and ICL pci ids.  As far
> as I can see it still lacks EHL/JSL and TGL as far as I can see.  Are
> there plans to continue support for xf86-video-intel for future
> platforms?  If not, then cutting the support off at TGL would still be
> safe.

It uses a generic binding to 0x8086. That it provides sprites planes is
one of the reasons why it is invaluable.
-Chris