Hangcheck does not work on SNB?

Hangcheck does not work on SNB?

[Lukas Hejtmanek]

Hello,

should the hangcheck mechanism work on Sandy Bridge when XVideo causes GPU to
hang? I can repeatedly hang GPU playing video for at least 1 hour
continuously. Mouse cursor moves but nothing else. 

However, I got no message from kernel that GPU hung. The notebook is alive
except graphics.

-- 
Lukáš Hejtmánek

Re: Hangcheck does not work on SNB?

[Chris Wilson]

On Mon, 2 Apr 2012 01:00:46 +0200, Lukas Hejtmanek <xhejtman@ics.muni.cz> wrote:
> Hello,
> 
> should the hangcheck mechanism work on Sandy Bridge when XVideo causes GPU to
> hang? I can repeatedly hang GPU playing video for at least 1 hour
> continuously. Mouse cursor moves but nothing else. 
> 
> However, I got no message from kernel that GPU hung. The notebook is alive
> except graphics.

Even after you stop the video? All it sounds like so far is that it is no
longer rendering video frames. Also check your Xorg.log and video player
for diagnostics.
-Chris

-- 
Chris Wilson, Intel Open Source Technology Centre

Re: Hangcheck does not work on SNB?

[Lukas Hejtmanek]

On Mon, Apr 02, 2012 at 08:32:50AM +0100, Chris Wilson wrote:
> Even after you stop the video? All it sounds like so far is that it is no
> longer rendering video frames. Also check your Xorg.log and video player
> for diagnostics.

Xorg.log contains (however this may be consequence of SAK (sysrq-k)):

[157714.584] [mi] EQ overflowing.  Additional events will be discarded until
existing events are processed.
[157714.584] Backtrace:
[157714.589] 0: /usr/bin/X (xorg_backtrace+0x26) [0x7f57d14622f6]
[157714.589] 1: /usr/bin/X (mieqEnqueue+0x263) [0x7f57d14429d3]
[157714.589] 2: /usr/bin/X (0x7f57d12da000+0x62814) [0x7f57d133c814]
[157714.589] 3: /usr/bin/X (xf86PostMotionEvent+0xd8) [0x7f57d1379fb8]
[157714.589] 4: /usr/lib/xorg/modules/input/synaptics_drv.so (0x7f57cb9f3000+0x3d28) [0x7f57cb9f6d28]
[157714.589] 5: /usr/lib/xorg/modules/input/synaptics_drv.so (0x7f57cb9f3000+0x5f34) [0x7f57cb9f8f34]
[157714.589] 6: /usr/bin/X (0x7f57d12da000+0x8ace7) [0x7f57d1364ce7]
[157714.590] 7: /usr/bin/X (0x7f57d12da000+0xb0bda) [0x7f57d138abda]
[157714.590] 8: /lib/x86_64-linux-gnu/libpthread.so.0 (0x7f57d0600000+0xfcb0) [0x7f57d060fcb0]
[157714.590] 9: /lib/x86_64-linux-gnu/libc.so.6 (ioctl+0x7) [0x7f57cf55ed27]
[157714.590] 10: /usr/lib/x86_64-linux-gnu/libdrm.so.2 (drmIoctl+0x28) [0x7f57cdaf11a8]
[157714.590] 11: /usr/lib/xorg/modules/drivers/intel_drv.so (0x7f57cd5fa000+0x374ee) [0x7f57cd6314ee]
[157714.590] 12: /usr/lib/xorg/modules/drivers/intel_drv.so (0x7f57cd5fa000+0x6789a) [0x7f57cd66189a]
[157714.590] 13: /usr/lib/xorg/modules/drivers/intel_drv.so (0x7f57cd5fa000+0x42bca) [0x7f57cd63cbca]
[157714.590] 14: /usr/lib/xorg/modules/drivers/intel_drv.so (0x7f57cd5fa000+0x55dee) [0x7f57cd64fdee]
[157714.590] 15: /usr/bin/X (BlockHandler+0x4a) [0x7f57d132c51a]
[157714.590] 16: /usr/bin/X (WaitForSomething+0x11d) [0x7f57d145f65d]
[157714.590] 17: /usr/bin/X (0x7f57d12da000+0x4e4b2) [0x7f57d13284b2]
[157714.590] 18: /usr/bin/X (0x7f57d12da000+0x3d68a) [0x7f57d131768a]
[157714.590] 19: /lib/x86_64-linux-gnu/libc.so.6 (__libc_start_main+0xed) [0x7f57cf49576d]
[157714.591] 20: /usr/bin/X (0x7f57d12da000+0x3d97d) [0x7f57d131797d]
[157714.591] [mi] These backtraces from mieqEnqueue may point to a culprit higher up the stack.
[157714.591] [mi] mieq is *NOT* the cause.  It is a victim.

And I am unable to stop the video, X session is completely dead so I don't
even see if and what mplayer complains about.

It seems that the video suddenly slows down and then stops completely but I am
not 100% sure about this.

-- 
Lukáš Hejtmánek

Re: Hangcheck does not work on SNB?

[Chris Wilson]

On Mon, 2 Apr 2012 10:12:31 +0200, Lukas Hejtmanek <xhejtman@ics.muni.cz> wrote:
> On Mon, Apr 02, 2012 at 08:32:50AM +0100, Chris Wilson wrote:
> > Even after you stop the video? All it sounds like so far is that it is no
> > longer rendering video frames. Also check your Xorg.log and video player
> > for diagnostics.
> 
> Xorg.log contains (however this may be consequence of SAK (sysrq-k)):
> 
> [157714.590] 10: /usr/lib/x86_64-linux-gnu/libdrm.so.2 (drmIoctl+0x28) [0x7f57cdaf11a8]
> [157714.590] 11: /usr/lib/xorg/modules/drivers/intel_drv.so (0x7f57cd5fa000+0x374ee) [0x7f57cd6314ee]
> [157714.590] 12: /usr/lib/xorg/modules/drivers/intel_drv.so (0x7f57cd5fa000+0x6789a) [0x7f57cd66189a]
> [157714.590] 13: /usr/lib/xorg/modules/drivers/intel_drv.so (0x7f57cd5fa000+0x42bca) [0x7f57cd63cbca]
> [157714.590] 14: /usr/lib/xorg/modules/drivers/intel_drv.so (0x7f57cd5fa000+0x55dee) [0x7f57cd64fdee]
> [157714.590] 15: /usr/bin/X (BlockHandler+0x4a) [0x7f57d132c51a]
> 
> And I am unable to stop the video, X session is completely dead so I don't
> even see if and what mplayer complains about.
> 
> It seems that the video suddenly slows down and then stops completely but I am
> not 100% sure about this.

Ok, this sounds more like a silent GPU death. We'll have to hit the
manual reset and see whether we can glean any useful info:

  echo 1 > /sys/kernel/debug/dri/0/i915_wedged

But before you do, can you have a quick look with intel_gpu_top and see
if it detects any activity?
-Chris

-- 
Chris Wilson, Intel Open Source Technology Centre