[ubuntu-x] Kernel Call for Testing(CFT): DRM stack

[Rafael Monica]

2010/3/11 Bryce Harrington <bryce at canonical.com>:
> On Thu, Mar 11, 2010 at 03:20:27AM +0100, Rafael Monica wrote:
>> On Tue, 2010-03-09 at 12:51 -0500, Jeremy Foshee wrote:
>> > Hi All,
>> >     The Kernel team would like to issue a Call For Testing of the
>> > backported DRM stack from the .33 tree. For your reference, we have
>> > taken the v2.6.33 DRM backport as it greatly improves stability and
>> > removes the need for the Nouveau LBM module, thus cleaning up installs.
>> > We are, therefore, keenly interested in the results of testing
>> > installation and use of this DRM stack so that we can verify its
>> > stability for the LTS release of Lucid.
>> >
>> > Your feedback is greatly appreciated.
>>
>> Hi, I tested the latest Daily Live CD with the new DRM with two
>> different ATI GPUs.
>>
>> With the first one, a HD3200 IGP, the display turned off during boot.
>> Once X started the display turned back on and there were no further
>> issues. Note that this was also happening with a Live CD with the old
>> DRM. I suspect this issue is caused by Plymouth as more people have
>> reported problems with Plymouth.
>>
>> With the second one, a HD5750, things were exactly the reverse. All
>> went fine during boot, but at the end of the boot the display turned
>> off and there was nothing I could do but reset the computer. There's
>> no support for this GPU in the 2.6.33 DRM, there's in drm-next git.
>
> Okay, neither of these sound like regressions, which is what we're
> looking for in particular here.
>
> In the first case, it sounds like while there is an issue, it works
> about the same as it did before, so at least it's not a regression in
> the drm.  In the second case you didn't mention having tested the old
> drm but I assume it'll be the same issue of needing new hardware
> support, which is not a regression either.
>
> For the second case make sure to file a bug report requesting support so
> it's in our system.  If you can flag the patch(es) needed, that'll
> improve the chance that it'll get looked at for lucid.
>

You are right. Neither of them are regressions. In the meantime I got
both issues resolved. The second case was just a matter of plugging
the monitor into the other DVI port to get it to work with UMS. It's
probably not worth it pulling in the KMS patches for evergreen since
at the moment there's only modesetting support and no acceleration or
power management for evergreen, and most people with these cards will
probably install fglrx anyway.

>> On a side note: Dave Airlie recently fixed a bug in Mesa that made
>> Blender practically unusable with R600+ GPUs. Since Blender is quite a
>> popular application it would be nice if we can have this bug fix in
>> before Lucid is released.
>>
>> http://cgit.freedesktop.org/mesa/mesa/commit/?h=mesa_7_7_branch&id=cf8af9bcf127e170b64112bd548d5d4e79c8e894
>
> For requesting backports of patches, the best procedure to follow is to
> file a bug in launchpad using 'ubuntu-bug xorg' and either link it to
> the corresponding (closed) upstream bug report, or attach the patch to
> the bug report itself.  Either way will ensure it bubbles up to the top
> of our work queue.
>

Done.

https://bugs.launchpad.net/xorg-server/+bug/539373