[Intel-gfx] [PATCH 1/2] drm/i915: re-enable semaphores by default

[Nicolas Kalkhof]

Hi Daniel,

ok after an rlogin from a remote machine I've managed to find exzessive syslog dumps (10+MB) repeating the following message:

Nov 18 19:26:45 gonzo kernel: [   81.266864] DRHD: handling fault status reg 3
Nov 18 19:26:45 gonzo kernel: [   81.266876] DMAR:[DMA Write] Request device [00:02.0] fault addr 2101f0000
Nov 18 19:26:45 gonzo kernel: [   81.266879] DMAR:[fault reason 05] PTE Write access is not set

More detailed log attached. Maybe this helps? I've also attached my kernel .config in case I did something stupid there ;-)

Too bad that bugzilla on kernel.org is still down. lkml is quite busy :(

Regards
Nic

-----Ursprüngliche Nachricht-----
Von: "Daniel Vetter" <daniel.vetter at ffwll.ch>
Gesendet: Nov 18, 2011 4:26:35 PM
An: "Nicolas Kalkhof" <nkalkhof at web.de>
Betreff: Re: [Intel-gfx] [PATCH 1/2] drm/i915: re-enable semaphores by default

>On Thu, Nov 17, 2011 at 19:31, Nicolas Kalkhof <nkalkhof at web.de> wrote:
>> Hi Daniel,
>>
>> no problem :). Your patches look good! No hangs or stalls so far with semaphores and rc6 switched on. However I had to disable intel_iommu since it freezes my box shortly after firing up X :(.
>
>Meh, so ppgtt doesn't seem to help rc6 hangs caused by dmar. Has
>anything changed wrt how the machine hangs, i.e. can you gather more
>information via ssh/logs/netconsole/error_state?
>
>> One other thing still remains: Since the merges for kernel 3.2 came in late october, my cpu/gpu seems to run hotter (approx 10 °C) when idling and after some minutes the cpu load goes up on all cores but the cpu frequency stays low (800 MHz) as well as the gpu (650MHz) and the system is responding well. Yet still additional 5-8 watts are drained from the power supply. Something has to throttle up :(. Oprofile shows that the kernel somewhere hogs the cpu (see dump below). This happens with every kernel from 3.1 up to latest linux-next. I wasn't able to track the bug down yet so I guess I'll better file a bug report on lkml to get someone to verify this.
>
>Iirc not the first time I've heard of this. Reporting a bug to lkml
>sounds like the right thing to do (kernel bugzilla is still down).
>
>Thanks for testing, Daniel
>--
>Daniel Vetter
>daniel.vetter at ffwll.ch - +41 (0) 79 365 57 48 - http://blog.ffwll.ch


___________________________________________________________
SMS schreiben mit WEB.DE FreeMail - einfach, schnell und
kostenguenstig. Jetzt gleich testen! http://f.web.de/?mc=021192
-------------- next part --------------
A non-text attachment was scrubbed...
Name: dmar_write_issue.txt
Type: application/octet-stream
Size: 295076 bytes
Desc: not available
URL: <http://lists.freedesktop.org/archives/intel-gfx/attachments/20111118/cf7cfac8/attachment.obj>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: .config
Type: application/octet-stream
Size: 64286 bytes
Desc: not available
URL: <http://lists.freedesktop.org/archives/intel-gfx/attachments/20111118/cf7cfac8/attachment-0001.obj>