Bug 25681 - [945GM KMS] GPU crash with kernel modesetting enabled
Summary: [945GM KMS] GPU crash with kernel modesetting enabled
Status: RESOLVED INVALID
Alias: None
Product: xorg
Classification: Unclassified
Component: Driver/intel (show other bugs)
Version: 7.4 (2008.09)
Hardware: Other All
: medium critical
Assignee: Carl Worth
QA Contact: Xorg Project Team
URL:
Whiteboard:
Keywords:
Depends on:
Blocks:
 
Reported: 2009-12-16 16:02 UTC by celejar
Modified: 2010-07-22 08:26 UTC (History)
1 user (show)

See Also:
i915 platform:
i915 features:


Attachments
Xorg.0.log from a system with KMS enabled (pre-crash) (43.57 KB, application/octet-stream)
2009-12-16 16:02 UTC, celejar
no flags Details
intel_gpu_dump from a system without KMS (76.45 KB, application/octet-stream)
2009-12-16 16:04 UTC, celejar
no flags Details
Example of how to check if the driver supports GEM. (1.10 KB, patch)
2010-07-21 13:48 UTC, Chris Wilson
no flags Details | Splinter Review

Description celejar 2009-12-16 16:02:36 UTC
Created attachment 32133 [details]
Xorg.0.log from a system with KMS enabled (pre-crash)

My kernels consistently crash when KMS is enabled.  I don't know if there's a trigger, but it seems to occur consistently within about ten minutes, while X is running.  I've seen this with a couple of different recent kernels, currently 2.6.32-lizzie-08080-g5ac4d63 (self-built from mainline).  No problem without KMS, with it, c
onsole freezes, no response to keyboard / mouse.  I have not yet been able to try ssh access from another host.

Debian Sid, i686 on a Celeron M 420
Acer Aspire 3690 - lspci excerpt:

00:00.0 Host bridge: Intel Corporation Mobile 945GM/PM/GMS, 943/940GML and 945GT Express Memory Controller Hub (rev 03)
00:02.0 VGA compatible controller: Intel Corporation Mobile 945GM/GMS, 943/940GML Express Integrated Graphics Controller (rev 03)
00:02.1 Display controller: Intel Corporation Mobile 945GM/GMS/GME, 943/940GML Express Integrated Graphics Controller (rev 03)
00:1b.0 Audio device: Intel Corporation 82801G (ICH7 Family) High Definition Audio Controller (rev 02)
00:1c.0 PCI bridge: Intel Corporation 82801G (ICH7 Family) PCI Express Port 1 (rev 02)
00:1c.1 PCI bridge: Intel Corporation 82801G (ICH7 Family) PCI Express Port 2 (rev 02)
00:1c.2 PCI bridge: Intel Corporation 82801G (ICH7 Family) PCI Express Port 3 (rev 02)
00:1c.3 PCI bridge: Intel Corporation 82801G (ICH7 Family) PCI Express Port 4 (rev 02)

I'm attaching a gpu dump without KMS enabled; I can try to get a dump of a crashed gpu if I can get in via ssh, if it would be helpful.
Comment 1 celejar 2009-12-16 16:04:45 UTC
Created attachment 32134 [details]
intel_gpu_dump from a system without KMS
Comment 2 celejar 2009-12-16 16:28:02 UTC
Ok, I tried enabling KMS again, waited for the crash, and then turned to an ssh connection.  I still had shell access, so I tried gpu_dump - which promptly segfaulted, leaving the machine totally dead (at least, it would no longer respond to my attempt to reestablish the (dead) ssh connection).
Comment 3 Chris Wilson 2010-07-10 04:20:27 UTC
Timeout. Recent kernels [2.6.34 and later] automatically store the error state in /sys/kernel/debug/dri/0/i915_error_state. That [if it contains an error] and the Xorg.0.log which *contains* the log from the crash are a minimum.
Comment 4 celejar 2010-07-11 06:32:22 UTC
Okay.  I'm going to re-enable KMS and report back, with the requested information if it still crashes.
Comment 5 Chris Wilson 2010-07-11 06:39:25 UTC
Thanks, I don't like closing bugs unresolved.
Comment 6 celejar 2010-07-21 12:53:31 UTC
Okay, I *think* that I'm successfully running now with KMS enabled:
~$ uname -a
Linux lizzie 2.6.35-rc5-lizzie-00064-ga9f7f2e #19 Mon Jul 19 13:55:56 EDT 2010 i686 GNU/Linux

~$ cat /etc/modprobe.d/i915-kms.conf 
#options i915 modeset=0

I haven't seen any problems through several boot cycles and multiple suspend-to-disk / resume cycles.  If you'll pardon a simple question, is there any way for me to absolutely verify that KMS is actually active?
Comment 7 Chris Wilson 2010-07-21 13:08:39 UTC
> --- Comment #6 from celejar@gmail.com 2010-07-21 12:53:31 PDT ---
> I haven't seen any problems through several boot cycles and multiple
> suspend-to-disk / resume cycles.  If you'll pardon a simple question, is there
> any way for me to absolutely verify that KMS is actually active?

The conclusive test is to query the driver as to whether it supports GEM.
However, circumstantial evidence can be found if the kernel loaded the
fbcon (dmesg | grep fbcon), then the driver is using KMS and GEM.
Comment 8 celejar 2010-07-21 13:24:28 UTC
$ dmesg | grep fbcon
$ 

But this is with an uptime of more than a day, and several suspend / resume cycles.  I suppose that I'll need to try it shortly after boot?

How do I query for GEM support?
Comment 9 Chris Wilson 2010-07-21 13:48:43 UTC
Created attachment 37281 [details] [review]
Example of how to check if the driver supports GEM.

This little bit of code will simply query whether the driver supports GEM or not. Note that is a drm-master (i.e. X) is already running it will require root privilege in order to successfully query the device.
Comment 10 celejar 2010-07-22 08:26:33 UTC
~# /home/celejar/tmp/a.out 
/dev/dri/card0: GEM enabled

So I guess we're good!  If I have trouble in the future, I guess I'll either reopen this, or file a new bug.

Thanks.


Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.