Re: please try patch against hibernate breakage

Re: please try patch against hibernate breakage

[Dave Airlie]

[-- Attachment #1: Type: text/plain, Size: 1384 bytes --]

>
> I tested your patch against kernel-desktop-3.1.10-18.1.x86_64 (openSUSE
> 12.1). After the 9. hibernate / resume sequence, I got this gpu-hang,
> which I never saw before while resuming without kms running on the
> resuming kernel:
>
> [ 1140.700604] [drm:i915_hangcheck_elapsed] *ERROR* Hangcheck timer
> elapsed... GPU hung
> [ 1140.700611] [drm] capturing error event; look for more information in
> /debug/dri/0/i915_error_state
> [ 1140.705594] [drm:i915_wait_request] *ERROR* i915_wait_request returns
> -11 (awaiting 2925 at 2924, next 2927)
> [ 1142.495454] [drm:i915_hangcheck_elapsed] *ERROR* Hangcheck timer
> elapsed... GPU hung
> [ 1142.495535] [drm:i915_wait_request] *ERROR* i915_wait_request returns
> -11 (awaiting 2927 at 2924, next 2928)
> [ 1142.495597] [drm:i915_reset] *ERROR* GPU hanging too fast, declaring
> wedged!
> [ 1142.495603] [drm:i915_reset] *ERROR* Failed to reset chip.
>
> I attached /debug/dri/0/i915_error_state.
>
> kwin and some other kde processes crashed, too.
>
> Until this GPU hang, I couldn't see any broken files as mentioned here
> (see attachments):
>
> https://bugzilla.novell.com/show_bug.cgi?id=732908
>
>

I've cc'dd intel-gfx, I forgot the first time, maybe they can decode
the error state.

But no corruptions is what I'm after, I've got my laptop to do
hib/resume cycles for 3.5 hrs now without seeing any wierdness.

Dave.

[-- Attachment #2: i915_error_state.bz2 --]
[-- Type: application/x-bzip, Size: 129765 bytes --]

[-- Attachment #3: Type: text/plain, Size: 159 bytes --]

_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

Re: please try patch against hibernate breakage

[Daniel Vetter]

On Wed, Mar 28, 2012 at 02:43:12PM +0100, Dave Airlie wrote:
> >
> > I tested your patch against kernel-desktop-3.1.10-18.1.x86_64 (openSUSE
> > 12.1). After the 9. hibernate / resume sequence, I got this gpu-hang,
> > which I never saw before while resuming without kms running on the
> > resuming kernel:
> >
> > [ 1140.700604] [drm:i915_hangcheck_elapsed] *ERROR* Hangcheck timer
> > elapsed... GPU hung
> > [ 1140.700611] [drm] capturing error event; look for more information in
> > /debug/dri/0/i915_error_state
> > [ 1140.705594] [drm:i915_wait_request] *ERROR* i915_wait_request returns
> > -11 (awaiting 2925 at 2924, next 2927)
> > [ 1142.495454] [drm:i915_hangcheck_elapsed] *ERROR* Hangcheck timer
> > elapsed... GPU hung
> > [ 1142.495535] [drm:i915_wait_request] *ERROR* i915_wait_request returns
> > -11 (awaiting 2927 at 2924, next 2928)
> > [ 1142.495597] [drm:i915_reset] *ERROR* GPU hanging too fast, declaring
> > wedged!
> > [ 1142.495603] [drm:i915_reset] *ERROR* Failed to reset chip.
> >
> > I attached /debug/dri/0/i915_error_state.
> >
> > kwin and some other kde processes crashed, too.
> >
> > Until this GPU hang, I couldn't see any broken files as mentioned here
> > (see attachments):
> >
> > https://bugzilla.novell.com/show_bug.cgi?id=732908
> >
> >
> 
> I've cc'dd intel-gfx, I forgot the first time, maybe they can decode
> the error state.
> 
> But no corruptions is what I'm after, I've got my laptop to do
> hib/resume cycles for 3.5 hrs now without seeing any wierdness.

The error_state doesn't seem to have any telltales for why the gpu died -
if the active bits are trustworthy it's one EU that got stuck. It could be
that some of the indirect state got corrupted (we don't dump that), but
hard to tell.

Please keep on testing and if it hangs again, grab the error_state again.
-Daniel
-- 
Daniel Vetter
Mail: daniel@ffwll.ch
Mobile: +41 (0)79 365 57 48