Bug 15417 - intel i915 segfaults xorg
Summary: intel i915 segfaults xorg
Status: RESOLVED CODE_FIX
Alias: None
Product: Drivers
Classification: Unclassified
Component: Video(DRI - Intel) (show other bugs)
Hardware: All Linux
: P1 normal
Assignee: drivers_video-dri-intel@kernel-bugs.osdl.org
URL: http://bugs.archlinux.org/task/18319
Keywords:
Depends on:
Blocks:
 
Reported: 2010-03-01 01:15 UTC by John Doe
Modified: 2010-07-14 15:06 UTC (History)
2 users (show)

See Also:
Kernel Version: 2.6.32.8 and up
Subsystem:
Regression: No
Bisected commit-id:


Attachments
Xorg log (15.18 KB, text/x-log)
2010-03-02 23:56 UTC, John Doe
Details

Description John Doe 2010-03-01 01:15:52 UTC
Hi,
as I've posted on http://bugs.archlinux.org/task/18319 when you use kernel 2.6.32.8 or 2.6.33 sometimes Xorg crashes (when switching from tty to X) or doesn't detect the resolution (800x600 instead of 1280x800). 
My video card is an Intel 4500MHD and I'm using:

KDE 4.4
xorg-server 1.7.5
xf86-video-intel 2.10.0
hal 0.5.14 (I don't use xorg.conf)

My actual OS is Gentoo testing amd64, but the same situation happens in ArchLinux (amd64 too).
I've moved to 2.6.32.7 and everything works fine. 
PS: I haven't tested kernel 2.6.32.9

Thanks.
Comment 1 Jesse Barnes 2010-03-02 02:48:03 UTC
Can you bisect to the failing kernel commit?
Comment 2 John Doe 2010-03-02 23:55:13 UTC
Well, I've tried (again) kernel 2.6.33 and I have to apologise. The description in #0 was ok for kernel 2.6.32.8 but no for 2.6.33
In Kernel 2.6.33 X doesn't crash (or at least, hasn't happened yet) but sometimes the resolution isn't detected correctly (I asumed that if resolution wasn't ok, X will crash too, but I've spent a while, and after insert "/etc/init.d/xdm restart" 10 times, 9 of them were ok and the rest failed on the detection of the resolution). In the attachment is the X log. Thank you (and sorry for my English).
Comment 3 John Doe 2010-03-02 23:56:18 UTC
Created attachment 25328 [details]
Xorg log
Comment 4 Karol Szuster 2010-03-07 17:32:35 UTC
Hi,

I'm experiencing the same bug. The affected kernel version is 2.6.32.8, 2.6.32.9 and 2.6.33.
The resolution is unproperly set (both in text console and X) and Xorg ocasionally crashes.
I've bisected it and found that this bug have been introduced by commit 612e99b2d1ba8839896c6ea78d4329782d4e16b8 (b01f2c3a4a37d09a47ad73ccbb46d554d21cfeb0 upstream)
I've reverted this patch on my local tree and everything works fine now.

Regards.

Ps. sorry for my English :)
Comment 5 Jesse Barnes 2010-03-08 18:12:13 UTC
How does X crash?  Can you get a backtrace with gdb?  The attached log doesn't have one...
Comment 6 John Doe 2010-03-25 01:11:46 UTC
Thank you very much, Karol. Reverting the patch in kernel 2.6.32.10 solved this.
@Jesse: X seems to restart (I get back to the kdm login screen), I'm sorry about gdb, I don't know how to do what you want, but if you help me I'll do it :)
Comment 7 Jesse Barnes 2010-03-25 16:02:54 UTC
Oh that's weird...  Yeah I'll need to see the backtrace.

The easiest way to do this is to attach to X using gdb from an ssh session.  Or if it crashes at startup, start X under gdb to get the backtrace, e.g.
  $ gdb --args /usr/bin/Xorg -verbose -retro
  (gdb) run
  *crash*
  (gdb) bt

that should get you the backtrace
Comment 8 John Doe 2010-05-13 20:07:58 UTC
Sorry for the delay, I've forgotten about this. With 2.6.33.4 the problem with the resolution still exists (I don't know if it hangs), and the patch cannot be reverted anymore. I've tried gdb, but when I switch to tty from gdb + Xorg, system hangs.
Comment 9 John Doe 2010-05-18 18:16:29 UTC
2.6.34 is affected too :(
Comment 10 John Doe 2010-05-31 17:02:48 UTC
2.6.35-rc1 too
Comment 11 John Doe 2010-06-09 22:01:43 UTC
Seems that this has been solved in 2.6.35-rc2. I,'ll try it some more time to get sure that there aren't freezes, but at least, the resolution is correctly handled. If this works, is there a chance that the changes are ported to 2.6.34 until we wait for 2.6.35? Thanks.
Comment 12 Jesse Barnes 2010-07-02 23:20:31 UTC
If you can identify the fix that went into 2.6.35-rc2 we can try to get it backported into the stable release.
Comment 13 Karol Szuster 2010-07-14 15:06:10 UTC
commit Fb8b5a39b6310379d7b54c0c7113703a8eaf4a57 corrects this bug but was reverted lately

Note You need to log in before you can comment on or make changes to this bug.