On 9 February 2011 09:40, Chris Wilson <chris@chris-wilson.co.uk> wrote:
On Wed, 9 Feb 2011 00:27:52 +0000, Pedro Ribeiro <pedrib@gmail.com> wrote:
> Hi,
>
> I've been having unreliable operation when hibernating ever since 2.6.37-rc5
> (the first 2.6.37 I tried).
> With 2.6.36 I never had a problem.

Wrong, you just never saw it.

commit 9334ef755f060e251f3f395caeda1a58b6834ea3
Author: Chris Wilson <chris@chris-wilson.co.uk>
Date:   Fri Jan 28 11:53:03 2011 +0000

   drm: Don't switch fb when disabling an output

   In drm_crtc_helper_set_config, we call drm_crtc_helper_set_mode which
   may return early and do no operation if the crtc is to be disabled. In
   this case we merrily swap to the new fb, discarding the old_fb believing
   that it has been cleaned up. However, due to the early return, the
   old_fb was not presented to the backend for correct reaping, and nor was
   the new one - which is about to be reaped via the
   drm_helper_disable_unused_functions(), leading to incorrect refcounting
   of the pinned objects.

   Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=27722
   Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=29857
   Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=29230
   Tested-by: Takashi Iwai <tiwai@suse.de>
   Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

--
Chris Wilson, Intel Open Source Technology Centre

Thanks for the prompt reply.

I assume this will be included in stable?

For the moment I'll just apply the patch to my build.

Regards,
Pedro