Maybe this is really a kernel bug? I am not sure whether to open a new report for this.
Anyway, here are some more observations:
- The system can still be accessed via ssh, so it isn't really locked up, only screen and input devices do not work.
- it seems that whenever the described bug occurs, there is some stack trace written to /var/log/kern.log:
Nov 2 14:02:55 slaine kernel: [ 3052.385064] BUG: scheduling with irqs disabled: mysqld/0x00000001/4794
Nov 2 14:02:55 slaine kernel: [ 3052.385083] caller is rt_spin_lock_slowlock+0x92/0x250
Nov 2 14:02:55 slaine kernel: [ 3052.385091] Pid: 4794, comm: mysqld Tainted: P 2.6.27-3-rt #1
Nov 2 14:02:55 slaine kernel: [ 3052.385097] [<c03512f6>] ? printk+0x18/0x1a
Nov 2 14:02:55 slaine kernel: [ 3052.385105] [<c0351cb0>] schedule+0xf0/0x100
Nov 2 14:02:55 slaine kernel: [ 3052.385112] [<c0353682>] rt_spin_lock_slowlock+0x92/0x250
Nov 2 14:02:55 slaine kernel: [ 3052.385120] [<c0353e1a>] __rt_spin_lock+0x4a/0x50
(... A full trace is attached as file trace.log.gz ...)
It is not always mysqld, for instance I also saw a similar trace for kaffeine, which also was created exactly at the moment the system (the x session, whatever) locked up.
There are other traces written to the log, but they do not seem to be related, at least there where no *observable* problems.
I am running an up-to-date ubuntu 8.10 installation.
uname -av says
Linux slaine 2.6.27-3-rt #1 PREEMPT RT Mon Oct 27 03:05:19 UTC 2008 i686 GNU/Linux
The processor is actually a x86_64, but my I am running it in 32bit mode.
Version of the NVidia driver is 173.14.12
The graphics adapter is a GeForce FX 5700LE
I will try to switch to the generic kernel image and see if the problem is still there. Maybe it is just related to the PREEMPT/RT-stuff.
Maybe this is really a kernel bug? I am not sure whether to open a new report for this.
Anyway, here are some more observations:
- The system can still be accessed via ssh, so it isn't really locked up, only screen and input devices do not work.
- it seems that whenever the described bug occurs, there is some stack trace written to /var/log/kern.log:
Nov 2 14:02:55 slaine kernel: [ 3052.385064] BUG: scheduling with irqs disabled: mysqld/ 0x00000001/ 4794 lock_slowlock+ 0x92/0x250 lock_slowlock+ 0x92/0x250 lock+0x4a/ 0x50
Nov 2 14:02:55 slaine kernel: [ 3052.385083] caller is rt_spin_
Nov 2 14:02:55 slaine kernel: [ 3052.385091] Pid: 4794, comm: mysqld Tainted: P 2.6.27-3-rt #1
Nov 2 14:02:55 slaine kernel: [ 3052.385097] [<c03512f6>] ? printk+0x18/0x1a
Nov 2 14:02:55 slaine kernel: [ 3052.385105] [<c0351cb0>] schedule+0xf0/0x100
Nov 2 14:02:55 slaine kernel: [ 3052.385112] [<c0353682>] rt_spin_
Nov 2 14:02:55 slaine kernel: [ 3052.385120] [<c0353e1a>] __rt_spin_
(... A full trace is attached as file trace.log.gz ...)
It is not always mysqld, for instance I also saw a similar trace for kaffeine, which also was created exactly at the moment the system (the x session, whatever) locked up.
There are other traces written to the log, but they do not seem to be related, at least there where no *observable* problems.
I am running an up-to-date ubuntu 8.10 installation.
uname -av says
Linux slaine 2.6.27-3-rt #1 PREEMPT RT Mon Oct 27 03:05:19 UTC 2008 i686 GNU/Linux
The processor is actually a x86_64, but my I am running it in 32bit mode.
Version of the NVidia driver is 173.14.12
The graphics adapter is a GeForce FX 5700LE
I will try to switch to the generic kernel image and see if the problem is still there. Maybe it is just related to the PREEMPT/RT-stuff.