rawhide report: 20090528 changes

[Adam Williamson]

On Fri, 2009-05-29 at 10:53 +0530, Rahul Sundaram wrote:
> On 05/28/2009 07:35 PM, Rawhide Report wrote:
> > Compose started at Thu May 28 06:15:03 UTC 2009
> > 
> > Updated Packages:
> > 
> > kernel-2.6.29.4-167.fc11
> > ------------------------
> > * Wed May 27 2009 Kyle McMartin <kyle at redhat.com> 2.6.29.4-164
> > - drm-intel-disable-kms-i8xx.patch: disable KMS by default on 845, 855,
> >   and 865. It can be forced on with i915.modeset=1 boot parameter.
> > 
> > * Wed May 27 2009 Kyle McMartin <kyle at redhat.com> 2.6.29.4-165
> > - Enable KMS/gem on I865.
> > - drm-no-gem-on-i8xx.patch: Remove I865 so GEM will be enabled.
> > - drm-intel-disable-kms-i8xx.patch: Enable KMS on I865. 
> > - Two fixes from Eric Anholt to fix i8x5:
> >    drm-i915-apply-a-big-hammer-to-865-gem-object.patch
> >    drm-i915-fix-tiling-pitch.patch
> > 
> > * Wed May 27 2009 Kristian Høgsberg <krh at redhat.com> - 2.6.29.4-166
> > - Add drm-intel-set-domain-on-fault.patch to fix random gem object
> >   corruption when swapping (495323 and probably others).
> > - Enable kms on 845 and 855 as well, Erics tiling patch should fix
> >   those too.
> > 
> > * Wed May 27 2009 Kristian Høgsberg <krh at madara.bos.redhat.com> - 2.6.4.167
> > - Actually disable drm-intel-disable-kms-i8xx.patch.
> 
> Seems a lot of back and forth. What's going on? Can someone tell me the
> final state of expected behaviour is various cards? It would be useful
> to point users to this.

I was going to explain yesterday and then decided I shouldn't - guess I
should have :)

This mostly relates to bug 502077, filed on May 21st (so quite late in
cycle). Intel i865 cards were experiencing random hangs with kernel
modesetting enabled (which was the default). There was a single
unconfirmed report of the same issue affect i845, and another single
unconfirmed report for i855.

Since it was so late in the cycle and it felt like every time we zapped
one KMS bug, another popped up to take its place, QA team / RelEng, at a
blocker meeting, recommended disabling KMS for i865 chips. Kyle came in
to make this change, and he (along with the other intel/kernel devs)
decided to disable KMS for i845 and i855 as well because of the other
reports (I think they were under the impression it was more certain that
i845 / i855 also suffered from this issue). That's -164.

Kristian wasn't happy with this approach, and preferred to try and fix
the real bug. Together with Kyle, he identified a potential fix, and
built a kernel with that fix included and KMS re-enabled on i865 (that's
-165) and got testing for it. A couple of i865 users confirmed that it
fixed their issue.

We then discussed it a bit further and I explained that we'd only one
one unconfirmed report of hangs for both i845 and i855, and I'd actually
actively asked other users of those chips if they were experiencing
hangs and they'd said no. So we decided that, as it was unclear whether
i845 and i855 had really ever suffered from this exact issue, and the
patch should fix it for them as well as i865 even if they did, we should
go ahead and re-enable KMS for all three chipsets (that's -166 and
-167).

So, from 163 to 167, the sum total of changes was really just the
addition of the patch to fix the hanging problem (and also
drm-intel-set-domain-on-fault.patch , which is for another bug that we
considered critical).
-- 
Adam Williamson
Fedora QA Community Monkey
IRC: adamw | Fedora Talk: adamwill AT fedoraproject DOT org
http://www.happyassassin.net