Bug 25576 - kernel 2.6.31 (UMS), intel 2.9.1 X.org hang (and system freeze when reading /proc/dri
Summary: kernel 2.6.31 (UMS), intel 2.9.1 X.org hang (and system freeze when reading ...
Status: RESOLVED WONTFIX
Alias: None
Product: xorg
Classification: Unclassified
Component: Driver/intel (show other bugs)
Version: 7.4 (2008.09)
Hardware: x86 (IA32) Linux (All)
: medium normal
Assignee: Carl Worth
QA Contact: Xorg Project Team
URL:
Whiteboard:
Keywords: NEEDINFO
Depends on:
Blocks:
 
Reported: 2009-12-10 20:27 UTC by Daniel Kahn Gillmor
Modified: 2010-05-11 10:36 UTC (History)
0 users

See Also:
i915 platform:
i915 features:


Attachments
script used from acpi trigger to capture state during hang (725 bytes, text/plain)
2009-12-10 20:27 UTC, Daniel Kahn Gillmor
no flags Details
captured data from recent EeePC 900 X.org hang with intel driver (329.54 KB, application/octetstream)
2009-12-10 20:30 UTC, Daniel Kahn Gillmor
no flags Details
dump capture from another hang (338.09 KB, application/octet-stream)
2009-12-17 07:47 UTC, Daniel Kahn Gillmor
no flags Details
dump from system when not hanging (287.89 KB, application/octetstream)
2009-12-17 07:49 UTC, Daniel Kahn Gillmor
no flags Details

Description Daniel Kahn Gillmor 2009-12-10 20:27:50 UTC
Created attachment 31967 [details]
script used from acpi trigger to capture state during hang

I just had an X.org hang, running 2.6.31 (linux-image-2.6.31-1-686 version 2.6.31-2 from debian unstable) and X.org's intel drivers (2:2.9.1-1, also from debian unstable).  i was not using KMS, and the system would not switch to a text-mode console with Ctrl-Alt-F1 :(

this is on an Asus EeePC 900 with the following hardware:

0 dkg@pip:~$ lspci -s 00:02
00:02.0 VGA compatible controller: Intel Corporation Mobile 915GM/GMS/910GML Express Graphics Controller (rev 04)
00:02.1 Display controller: Intel Corporation Mobile 915GM/GMS/910GML Express Graphics Controller (rev 04)
0 dkg@pip:~$ 


I triggered my acpi script (see attachment) to record the state of the system by pressing the power button, but it looks like it caused the system to hang.  I waited for for several minutes, and then i had to press and hold down the power button for 15 seconds to hard-kill the machine.

Interestingly, it recorded everything *except* for the data from /proc/dri -- maybe it hung the system trying to read /proc/dri, as that is the last thing copied out by my script.  maybe that's a clue?

I'll attach a tarball of the dump next.
Comment 1 Daniel Kahn Gillmor 2009-12-10 20:30:58 UTC
Created attachment 31968 [details]
captured data from recent EeePC 900 X.org hang with intel driver

here is the data captured from the acpi trigger
Comment 2 Daniel Kahn Gillmor 2009-12-17 07:47:37 UTC
Created attachment 32148 [details]
dump capture from another hang

I had another hang. Attached is a dump from that hang -- again it appears that trying to copy /proc/dri locked up the script, and the dump did not complete.
Comment 3 Daniel Kahn Gillmor 2009-12-17 07:49:11 UTC
Created attachment 32149 [details]
dump from system when not hanging

just for reference, here is the output of my dump script when the system is *not* hung.  (i captured this after i rebooted the machine this morning)
Comment 4 Carl Worth 2010-02-17 09:32:56 UTC
Hi Daniel,

I don't have any solid theory as to the bug yet.

But, is there a specific reason you're not using KMS?

I'd be very interested to know if using KMS helps avoid the bug.

And if there's something that KMS doesn't let you do, then that's something
we need to fix. (With more recent drivers, KMS is mandatory.)

-Carl
Comment 5 Daniel Kahn Gillmor 2010-02-17 09:50:23 UTC
i'm now using KMS, and more recent versions of all of these tools in an attempt to resolve this bug (see bug 25765), so i don't have this configuration available for testing any more :(
Comment 6 Chris Wilson 2010-05-11 10:36:08 UTC
If the bug is not reproducible under KMS, then there is no way we can fix it...


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.