[Intel-gfx] Errors while resuming from hibernate in 2.6.35-rc5

[Pedro Ribeiro]

On 20 July 2010 08:31, Chris Wilson <chris at chris-wilson.co.uk> wrote:
> On Tue, 20 Jul 2010 01:56:01 +0100, Pedro Ribeiro <pedrib at gmail.com> wrote:
>> with kernel 2.6.35-rc5 I'm seeing a problem which happens every few
>> hibernate/resume cycles.
>> When I resume my computer, X is black (only the mouse cursor is
>> visible) and the X server outputs a stack trace to the VT. Changing to
>> the VT and back restores xorg correctly. dmesg says:
>>
>> [118132.523397] video LNXVIDEO:00: Restoring backlight state
>> [118132.523808] render error detected, EIR: 0x00000010
>> [118132.523811]   IPEIR: 0x00000000
>> [118132.523812]   IPEHR: 0x00000000
>> [118132.523814]   INSTDONE: 0xfffffffe
>> [118132.523815]   INSTPS: 0x00000000
>> [118132.523816]   INSTDONE1: 0xffffffff
>> [118132.523818]   ACTHD: 0x02000000
>> [118132.523819] page table error
>> [118132.523821]   PGTBL_ER: 0x00040000
>> [118132.523823] [drm:i915_report_and_clear_eir] *ERROR* EIR stuck:
>> 0x00000010, masking
>>
>> This would not happen with 2.6.34. My hardware is a GM45.
>>
>> any idea what this might be?
>
> Looks like we got the ordering wrong. My guess is that we enabled the
> backlight controller whilst the pipe was turned off. Can you try and
> capture the error/dmesg with drm.debug=0xf? Meanwhile, it's time to hit
> the docs and see if they have any advice.
>
> --
> Chris Wilson, Intel Open Source Technology Centre
>

I just hit the bug!

Unfortunately there is so much output that dmesg does not show
anything useful - the log is so big that it gets cropped before I get
a chance to see it.
I've stopped rsyslog since it kept filing up my root partition to
100%... Should I reduce the debug output?

Regards,
Pedro