[igt-dev] [PATCH i-g-t 3/4] lib/kms: Commit primary plane props with COMMIT_LEGACY

[Ville Syrjälä]

On Tue, Apr 20, 2021 at 10:43:03AM +0300, Petri Latvala wrote:
> On Fri, Apr 16, 2021 at 08:48:40PM +0300, Ville Syrjala wrote:
> > From: Ville Syrjälä <ville.syrjala at linux.intel.com>
> > 
> > Currently COMMIT_LEGACY for the primary plane only issues the
> > setcrtc ioctl, leaving all other plane properties unchanged
> > even if they were flagged as needing an update. Let's issue
> > the appropriate setprop ioctls in addition to the setcrtc to
> > make sure the plane state really reflects what was requested.
> > 
> > Without this the prop values can leak between tests. Eg. on
> > skl/derivatives running one of the failing kms_plane_alpha_blend
> > subtests first, and following it up with kms_cursor_crc (which
> > only uses legacy commits) causes crc mismatches since the
> > primary plane alpha is still left at some stale value despite
> > igt_plane_reset() trying to reset it back to 1.0.
> > 
> > Signed-off-by: Ville Syrjälä <ville.syrjala at linux.intel.com>
> 
> The reasoning is sound and this seems to do what the commit message
> explains, but why does kms_flip_tiling fail consistently with it?

Dunno yet. Annoyingly it passes with flying color on my local glk.

-- 
Ville Syrjälä
Intel