2.6.22-rc6 spurious hangs

[Thomas Sattler <tsattler@gmx.de>]

Hi there ...

I'm observing seldom hangs with linux 2.6. I can't tell when exactly it
happened the first time, I think somewhere around 2.6.16 or 2.6.17. I
see it about once or twice a month. With absolutely nothing in the logs.
So far I asked for help:

 - in the -ck list

Mon Sep 4 10:22:06 EST 2006, [ck] ck-patches seem to break DVB-T drivers
(see http://bhhdoa.org.au/pipermail/ck/2006-September/thread.html#6385)

 - in the linux-dvb list

Wed Sep 6 19:02:29 CEST 2006, [linux-dvb] driver problems when using
ck-patchset
(http://www.linuxtv.org/pipermail/linux-dvb/2006-September/thread.html#12649)

 - in the DaLUG (german, currently no archive) 14.09.2006


But nobody could help me so far.

Here is what I do:

I was running different kernels with different patchsets. It happened in
the past on -ck kernels (staircase), vanilla scheduler and cfs. As far
as I can remember the following patches were allways applied: squashfs
and vesa-tng.

Currently I'm running 2.6.22rc6 with cfs-v18, vesa-tng and an
XFS-lockdep patch:

http://people.redhat.com/mingo/cfs-scheduler/sched-cfs-v2.6.22-rc6-v18.patch
http://dev.gentoo.org/~spock/projects/vesafb-tng/archive/vesafb-tng-1.0-rc2-2.6.20-rc2.patch
see http://marc.info/?l=linux-kernel&m=118286232709378&w=2

I also installed these kernel modules via gentoo portage:

ati-drivers-8.37.6-r1
fuse-2.6.4-r1
kqemu-1.3.0_pre11
truecrypt-4.3

kqemu and truecrypt weren't loaded, but ati-drivers and fuse were.

The box I talk about is an IBM T41p with 1.7GHz Pentium M and 512MB RAM.
The distribution in use is gentoo, quite up to date. Attached to the box
is an USB2.0 DVB-T receiver (Cinergy T², Terratec).

In rare cases the keyboard stops working when the T² stops streaming DVB
to the box. It happens when I record the stream to disk as well as when
I stream it to mplayer.

If end of streaming is caused by a keypress, 'q' or 'enter' on mplayer,
that key gets stuck. It's repeated until I reboot the box.

If the recording was scheduled and stops by itself no more keys are
recognized. The keyboard is dead. The laptop's own and the attached
USB-Keyboard. Magic-Sys-Keys are still working.

I can still use the mouse to move windows around, start new xterms via
icewm's panel or copy and past single characters from an xterm to other
xterms.

I can also close most of the open windows, for example firefox and most
xterms. I cannot close an xterm which is started as 'xterm -e top' by
icewm or a vncviewer. Both windows stay open but lose their content.

If a root shell is open I can enter 'reboot' or 'halt' but most of the
time this doesn't reboot or halt. I get the message for an upcoming
shotdown in all xterms but the box doesn't come down.

The systemload continously increases but there is nothing to see in top why.

Ingo Molnar told me to enable CONFIG_PROVE_LOCKING but xfs triggers it
long before the box hangs. I tested the patch mentioned above but it was
triggered by xfs again, see [1] and I didn't reboot between this and the
last hung. [1] http://marc.info/?l=linux-kernel&m=118295294529681&w=2

As Ingo told me I run 'echo t > /proc/sysrq-trigger' this time. The
corresponding part of my syslogs is attached, as well as my kernel config.


Another thing I observed with the T² is that it doesn't work if it's
already connected when the laptop boots up. I need to power off,
disconnect and boot. If I connect the T² after bootup it works. I can
also rmmod it's driver when it's not in use.

If I boot the box with the T² connected I cannot use it, the blue led in
the T² is always off and I cannot rmmod the driver. (I don't know
whether I ever tired to rmmod the driver before I tried to use the T².)


Please CC me as I'm not subscribed to the list.

Thomas

-- 
keep mailinglists in english, feel free to send PM in german

Attachment: messages.gz
Description: application/gzip

Attachment: config.gz
Description: application/gzip