DRM range manger test failures

[Chris Wilson]

On Tue, Jan 10, 2017 at 02:51:17PM +0100, Daniel Vetter wrote:
> On Tue, Jan 10, 2017 at 12:45:04PM +0000, Chris Wilson wrote:
> > On Tue, Jan 10, 2017 at 01:41:54PM +0100, Geert Uytterhoeven wrote:
> > > Hi Chris, Laurent,
> > > 
> > > On R-Car Gen2 (Koelsch) and Gen3 (Salvator-X), the new DRM range
> > > manager selftests fail with:
> > > 
> > >     drm_mm: Testing DRM range manger (struct drm_mm), with
> > > random_seed=0x83a9b910 max_iterations=8192 max_prime=128
> > >     drm_mm: igt_sanitycheck - ok!
> > >     drm_mm: node has wrong color, found 0, expected 1
> > >     drm_mm: default insert failed, size 1 step 1
> > > 
> > > I have no idea what this means.
> > 
> > The existing code has bugs. We got as far as applying the testcases to
> > demonstrate the bugs, the actual fix is still queued.
> 
> Hm, I thought I merged them all. What's missing?

The actual fix required some coordination between drm-misc and i915
trees, so was delayed. (Only because I wanted to remove repetition of
the bug from i915 first.)
-Chris

-- 
Chris Wilson, Intel Open Source Technology Centre