XFS problems (hard lockup and oops on startup) with 2.5.5{6,7}
Hello,
The kernel 2.5.56 seems to have changed something that affected the XFS
code as I'm getting hit by one (and possibly two) bugs related to it now. What happens is that the machine suddenly freezes (running Debian/Sid - XFree 4.2.1, glibc 2.3.1, gnome2 - nothing out of ordinary) so that SysRq keys don't have any effect (but Numlock still toggles the led on the keyboard), machine is inaccessible from the net and the only cure is hard reboot. Nothing gets logged on the freeze and on reboot when XFS attempts to check the first filesystem the kernel oopses with Oops code 0002, in the interrupt handler. Nothing gets logged, of course, so I can't provide the full backtrace right now - I'll try to get it logged through the serial console if it happens again with 2.5.58. I have copied some values by hand from the screen (until I lost patience... :)):
Unable to handle kernel paging request at virtual address 000500074 printing EIP
C0138652
*PDE: 00000000
Oops: 0002
EIP: 0060:[<c0138662>]
EFLAGS: 00010016
EAX: 00050070 EBE: EFBB8000 ECX: EFBB8040 EDX: 0A000000
ESI: EFFEFE50 EDI: 00000000 EBP: EFFEFE5C ESP: C0403ECC
DS: 007B ES: 007B SS: 0069
Process: swapper (PID: 0, threadinfo=C0402000 TASK: C03DBC00)
The only way to get back to 2.5.5{6,7} is
to reboot to 2.5.55 or earlier whose XFS is able to recover the filesystem and then go back to .5{6,7}. My machine config is attached. Is anyone else observing such behavior?
TIA,
marek
Quoted 15 lines · Marek Habersack
Hello,
The kernel 2.5.56 seems to have changed something that affected the XFS
code as I'm getting hit by one (and possibly two) bugs related to it now. What happens is that the machine suddenly freezes (running Debian/Sid - XFree 4.2.1, glibc 2.3.1, gnome2 - nothing out of ordinary) so that SysRq keys don't have any effect (but Numlock still toggles the led on the keyboard), machine is inaccessible from the net and the only cure is hard reboot. Nothing gets logged on the freeze and on reboot when XFS attempts to check the first filesystem the kernel oopses with Oops code 0002, in the interrupt handler. Nothing gets logged, of course, so I can't provide the full backtrace right now - I'll try to get it logged through the serial console if it happens again with 2.5.58. I have copied some values by hand from the screen (until I lost patience... :)):
Quoted 2 lines
Unable to handle kernel paging request at virtual address 000500074 printing EIP
Hmm, that's really no much info. And there weren't any XFS changes from 2.5.52 to 2.5.58..
On Wed, Jan 15, 2003 at 08:21:34PM +0000, Christoph Hellwig scribbled:
[snip]
Quoted 11 lines
Quoted 8 lines
check the first filesystem the kernel oopses with Oops code 0002, in the interrupt handler. Nothing gets logged, of course, so I can't provide the full backtrace right now - I'll try to get it logged through the serial console if it happens again with 2.5.58. I have copied some values by hand from the screen (until I lost patience... :)):
Unable to handle kernel paging request at virtual address 000500074 printing EIPHmm, that's really no much info. And there weren't any XFS changes from 2.5.52 to 2.5.58..
Yep, that's why I wrote about something that affected XFS - at first I thought it might have been preemption but then I checked I had it on also on 2.5.55. I'm running 2.5.58 currently and should the oops happen again, I'll either transcribe the screen or get a screenshot via serial console. The only thing I added in my > .55 config was the freebsd slice support and the UFS support (as the module), but I doubt that would be the cause, although, who knows?
thanks,
marek