X still freezing once in a while

[Per Bothner]

Every other day or so, X freezes.  This has been happening for a while,
while keeping Rawhide current.  It's on a Lenovo T400, with the
Intel GM45 Express Chipset.

Lately I've been running with Desktop Effects enabled; I just
disabled them to see if that would make things more stable.

The symptoms:  Screen freezes (I can't say for sure whether things like
the clock get updated); mouse can move but doesn't change icon; mouse
clicks and keyboard inputs are ignored.  I can ssh in, and it shows Xorg
taking about 100% of one CPU.  It seems to happen when I do certain
things in Firefox (most recently I tried to open an image on lwn.net in
a new window).  No messages in Xorg.0.log or /var/log/messages

This seems a lot like this bug:
https://bugzilla.redhat.com/show_bug.cgi?id=473347
However, the comments say that that bug only happens with EXA,
and I'm definitely running UXA (according to Xorg.0.log).

It could be this, but it doesn't quite match:
https://bugzilla.redhat.com/show_bug.cgi?id=496516
I do see these messages, but they happen *hours* before the freeze.
Apr 28 21:21:28 localhost kernel: [drm:i915_gem_object_bind_to_gtt] 
*ERROR* GTT full, but LRU list empty
Apr 28 21:21:28 localhost kernel: [drm:i915_gem_object_pin] *ERROR* 
Failure to bind: -12
Apr 28 21:21:28 localhost kernel: [drm:i915_gem_evict_something] *ERROR* 
inactive empty 1 request empty 1 flushing empty 1
And there are no "bo map failed" messages in Xorg.0.log.

File a new bug?
-- 
	--Per Bothner
per(a)bothner.com   http://per.bothner.com/

[iarly selbir]

I have same problem here.

I'm following this bug

https://bugzilla.redhat.com/show_bug.cgi?id=476743


>  I'm using KDE. 
Regards,

- -
iarly selbir | ski0s

:wq!


On Wed, Apr 29, 2009 at 4:06 PM, Per Bothner <per(a)bothner.com&gt; wrote:

>  Every other day or so, X freezes.  This has been happening for a
> while,
>  while keeping Rawhide current.  It's on a Lenovo T400, with the
>  Intel GM45 Express Chipset.
>
>  Lately I've been running with Desktop Effects enabled; I just
>  disabled them to see if that would make things more stable.
>
>  The symptoms:  Screen freezes (I can't say for sure whether things like
>  the clock get updated); mouse can move but doesn't change icon; mouse
>  clicks and keyboard inputs are ignored.  I can ssh in, and it shows Xorg
>  taking about 100% of one CPU.  It seems to happen when I do certain
>  things in Firefox (most recently I tried to open an image on lwn.net in
>  a new window).  No messages in Xorg.0.log or /var/log/messages
>
>  This seems a lot like this bug:
>  https://bugzilla.redhat.com/show_bug.cgi?id=473347
>  However, the comments say that that bug only happens with EXA,
>  and I'm definitely running UXA (according to Xorg.0.log).
>
>  It could be this, but it doesn't quite match:
>  https://bugzilla.redhat.com/show_bug.cgi?id=496516
>  I do see these messages, but they happen *hours* before the freeze.
>  Apr 28 21:21:28 localhost kernel: [drm:i915_gem_object_bind_to_gtt] *ERROR*
>  GTT full, but LRU list empty
>  Apr 28 21:21:28 localhost kernel: [drm:i915_gem_object_pin] *ERROR* Failure
>  to bind: -12
>  Apr 28 21:21:28 localhost kernel: [drm:i915_gem_evict_something] *ERROR*
>  inactive empty 1 request empty 1 flushing empty 1
>  And there are no "bo map failed" messages in Xorg.0.log.
>
>  File a new bug?
>  --
>         --Per Bothner
>  per(a)bothner.com   http://per.bothner.com/
>
>  --
>  fedora-test-list mailing list
>  fedora-test-list(a)redhat.com
>  To unsubscribe: https://www.redhat.com/mailman/listinfo/fedora-test-list
>

[MartinG]

On Wed, Apr 29, 2009 at 9:14 PM, iarly selbir | ski0s <iarlyy(a)gmail.com&gt; wrote:
>  I have same problem here. 
Me too, on a Lenovo ThinkPad T400 (as thread starter),
Intel Corporation Mobile 4 Series Chipset Integrated Graphics
Controller (rev 07) (prog-if 00 [VGA controller])
Smolt: http://www.smolts.org/client/show/pub_8c3b9b48-bcf7-4f9d-9aab-183e80ba73e4

I also run an updated Rawhide, initially using UXA (the default when
there is no xorg.conf), but due to corrupted display when I started
emacs in graphics mode, I put up an minimalistic xorg.conf and chose
EXA.
I'm not 100% sure if I did have or didn't have the freezes when using
UXA, but I do get freezes when I use EXA.

At one point I thought it was due to java (jre-1.6.0_13-fcs.x86_64),
so I shut down all java stuff (killall java, turn off java globally in
konqueror, etc). Things didn't stabilize, so now I have gcaldaemon
running (java app), and use EXA, but no desktop effects, just plain
kwin.
What I've experienced is:
- X just freeze, I can move the mouse, but keyboard is dead.
- X just crash, I'm kicked out to the login screen
- When I resume, it seems I am given a vt that doesn't actually work,
but has an static image of the graphical screen, ie. I cannot move the
mouse, and no keyboard input is recorded. However, if I press
ctrl-alt-f1, I'm back in X, and things works.
- The lockup behavior might be dependent on whether I have been
through a suspend/resume cycle.
- Using EXA, I have seen corrupted glyphs on the screen (eg. "f" and
"t" are a bit shortened, and have an extra dash in them). I've also
seen random colored lines at random places on the screen.

I have:
kernel-2.6.29.1-102.fc11.x86_64
xorg-x11-drv-intel-2.7.0-2.fc11.x86_64
xorg-x11-server-Xorg-1.6.1-6.fc11.x86_64

>  I'm following this bug
>  https://bugzilla.redhat.com/show_bug.cgi?id=476743 
I've seen this too, but *only* when using desktop effects or compiz.

Here's a report on KMS+UXA:
https://bugs.freedesktop.org/show_bug.cgi?id=21181

It might very well be two or more different bugs (EXA vs UXA etc), but
equally annoying all of them...


-MartinG
(messerting on irc@freenode)

[Adam Jackson]

On Wed, 2009-04-29 at 12:06 -0700, Per Bothner wrote:
>  Every other day or so, X freezes.  This has been happening for a
> while,
>  while keeping Rawhide current.  It's on a Lenovo T400, with the
>  Intel GM45 Express Chipset.
>  
>  Lately I've been running with Desktop Effects enabled; I just
>  disabled them to see if that would make things more stable.
>  
>  The symptoms:  Screen freezes (I can't say for sure whether things like
>  the clock get updated); mouse can move but doesn't change icon; mouse
>  clicks and keyboard inputs are ignored.  I can ssh in, and it shows Xorg
>  taking about 100% of one CPU.  It seems to happen when I do certain
>  things in Firefox (most recently I tried to open an image on lwn.net in
>  a new window).  No messages in Xorg.0.log or /var/log/messages 
When this happens, where does gdb or profile show the server spending
its time?

- ajax

[Per Bothner]

On 04/29/2009 01:59 PM, Adam Jackson wrote:
>  When this happens, where does gdb or profile show the server
> spending
>  its time? 
It's been a while since I've used gdb or done any sort of profiling
on (non-Java) applications, so if you can give me instructions what
to do (when it next happens) that would be appreciated.

I see a request for using intel_gpu_dump; I can also have that
set up in readiness ...
-- 
	--Per Bothner
per(a)bothner.com   http://per.bothner.com/

[Adam Jackson]

On Wed, 2009-04-29 at 14:04 -0700, Per Bothner wrote:
>  On 04/29/2009 01:59 PM, Adam Jackson wrote:
>  > When this happens, where does gdb or profile show the server spending
>  > its time?
>  
>  It's been a while since I've used gdb or done any sort of profiling
>  on (non-Java) applications, so if you can give me instructions what
>  to do (when it next happens) that would be appreciated. 
http://wiki.x.org/wiki/Development/Documentation/ServerDebugging

Has instructions for how to attach with gdb.  Yes, this requires two
machines.

- ajax

[Matej Cepl]

On 2009-04-30, 13:47 GMT, Adam Jackson wrote:
>  http://wiki.x.org/wiki/Development/Documentation/ServerDebugging
>
>  Has instructions for how to attach with gdb.  Yes, this requires two
>  machines. 
That's not much helpful, right?

loaded symbols for /lib64/libnss_files.so.2
0x00000034c58d9043 in __select_nocancel () from /lib64/libc.so.6
(gdb) bt f
#0  0x00000034c58d9043 in __select_nocancel () from 
/lib64/libc.so.6
No symbol table info available.
#1  0x00000000004e77ca in WaitForSomething ()
No symbol table info available.
#2  0x0000000000446ec2 in Dispatch ()
No symbol table info available.
#3  0x000000000042d0e5 in main ()
No symbol table info available.
(gdb) thread apply all backtrace

Thread 1 (Thread 0x7ff06b5e37b0 (LWP 19954)):
#0  0x00000034c58d9043 in __select_nocancel () from 
/lib64/libc.so.6
#1  0x00000000004e77ca in WaitForSomething ()
#2  0x0000000000446ec2 in Dispatch ()
#3  0x000000000042d0e5 in main ()
(gdb) list
1  /* Protocols file parser in nss_files module.
2     Copyright (C) 1996, 1997 Free Software Foundation, Inc.
3     This file is part of the GNU C Library.
4
5     The GNU C Library is free software; you can redistribute it and/or
6     modify it under the terms of the GNU Lesser General Public
7     License as published by the Free Software Foundation; either
8     version 2.1 of the License, or (at your option) any later version.
9
10    The GNU C Library is distributed in the hope that it will be useful,
Warning: the current language does not match this frame.
(gdb)

Using 
xorg-x11-apps-7.3-8.fc11.x86_64
xorg-x11-drv-ati-debuginfo-6.12.2-7.fc11.x86_64
xorg-x11-drv-ati-6.12.2-7.fc11.x86_64
xorg-x11-drv-evdev-2.2.1-3.fc11.x86_64
xorg-x11-drv-mouse-1.4.0-2.fc11.x86_64
xorg-x11-drv-synaptics-1.1.0-3.fc11.x86_64
xorg-x11-server-common-1.6.1-6.fc11.x86_64
xorg-x11-server-utils-7.4-7.fc11.x86_64
xorg-x11-server-Xorg-1.6.1-6.fc11.x86_64

Matěj

[Adam Jackson]

On Thu, 2009-04-30 at 17:32 +0200, Matej Cepl wrote:
>  On 2009-04-30, 13:47 GMT, Adam Jackson wrote:
>  > http://wiki.x.org/wiki/Development/Documentation/ServerDebugging
>  >
>  > Has instructions for how to attach with gdb.  Yes, this requires two
>  > machines.
>  
>  That's not much helpful, right? 
The (omitted) backtrace shows a server waiting for more requests,
correct.  Meaning it's not really hung, and also not like the OP's bug
report.  Per's report said that the mouse would not change icon when
moving among windows, indicating that the main loop wasn't being
processed.

Yours, if I had to guess from the thread on fedora-devel, was metacity
getting stuck, so gdbing metacity would probably be more enlightening.

- ajax

[Chuck Anderson]

On Thu, Apr 30, 2009 at 03:23:00PM -0400, Adam Jackson wrote:
>  On Thu, 2009-04-30 at 17:32 +0200, Matej Cepl wrote:
>  > On 2009-04-30, 13:47 GMT, Adam Jackson wrote:
>  > > http://wiki.x.org/wiki/Development/Documentation/ServerDebugging
>  > >
>  > > Has instructions for how to attach with gdb.  Yes, this requires two
>  > > machines.
>  > 
>  > That's not much helpful, right?
>  
>  The (omitted) backtrace shows a server waiting for more requests,
>  correct.  Meaning it's not really hung, and also not like the OP's bug
>  report.  Per's report said that the mouse would not change icon when
>  moving among windows, indicating that the main loop wasn't being
>  processed.
>  
>  Yours, if I had to guess from the thread on fedora-devel, was metacity
>  getting stuck, so gdbing metacity would probably be more enlightening. 
Same symptoms here, but on a Radeon X600 (RV380).

Bugzilla:

https://bugzilla.redhat.com/show_bug.cgi?id=498525

Backtrace:

pixmanBltsse2 (__P=<value optimized out>) at pixman-sse2.c:4501
4501             xmm3 = load128Unaligned ((__m128i*)(s+48));
(gdb) thread apply all bt

Thread 1 (Thread 0x7fa8ab9f37b0 (LWP 1916)):
#0  0x000000346763171b in pixmanBltsse2 (__P=<value optimized out>)
   from /usr/lib64/libpixman-1.so.0
#1  0x0000003467620acb in pixman_walk_composite_region (
    compositeRect=<value optimized out>, maskRepeat=<value optimized
out>,
    srcRepeat=<value optimized out>, height=<value optimized out>,
    width=<value optimized out>, yDst=<value optimized out>,
    xDst=<value optimized out>, yMask=<value optimized out>,
    xMask=<value optimized out>, ySrc=<value optimized out>,
    xSrc=<value optimized out>, pDst=<value optimized out>,
    pMask=<value optimized out>, pSrc=<value optimized out>,
    op=<value optimized out>) at pixman-pict.c:1371
#2  pixman_image_composite (compositeRect=<value optimized out>,
    maskRepeat=<value optimized out>, srcRepeat=<value optimized out>,
    height=<value optimized out>, width=<value optimized out>,
    yDst=<value optimized out>, xDst=<value optimized out>,
    yMask=<value optimized out>, xMask=<value optimized out>,
    ySrc=<value optimized out>, xSrc=<value optimized out>,
    pDst=<value optimized out>, pMask=<value optimized out>,
    pSrc=<value optimized out>, op=<value optimized out>) at
pixman-pict.c:2072