On Fri, 20 Mar 1998, MOLNAR Ingo wrote:
[...]
> if the lockups are still there even with the new driver, then could you
> give the NMI-watchdog a try? Try the 2.1.90-B version,i've posted it to
[...]
> - apply the patch
I applied a patch from your post of 11.3.1998 with /nmi-oops.*90-B/ in
the subject. That was the patch you meant, I assume.
> - configure NMI watchdog to yes in Char Dev/Watchdogs
did it, left the IRQ-number at 0. Does it matter?
> - reboot into the new kernel
> - wait for a lockup, a 'double oops' will be generated, one oops
Sometimes I get not the system locked, but only an uninterruptibly sleeping
process, e.g:
newsfeed:/root # ps l 1177
FLAGS UID PID PPID PRI NI SIZE RSS WCHAN STA TTY TIME COMMAND
0 0 1177 151 0 0 992 464 wait_on_pag D p0 0:09 tar -cf - .
newsfeed:/root # ps m 1177
PID TTY MAJFLT MINFLT TRS DRS SIZE SWAP RSS SHRD LIB DT COMMAND
1177 p0 20835 40 56 408 464 0 464 300 0 41 tar -cf - .
e.g. the 20836th pagefault locked that process. There was only one warning in the
logs, again reading "kernel: (scsi0:0:0) Target busy". The nmi-oops patch made no
difference here. I'd guess, a system lock occurs, when an important process for the
system hangs in wait_on_pag....
> per CPU. Write the 'Call Trace' lines down and send them through
This will become clearer to me, when the case actually happens, I hope.
> ksymoops. (register content doesnt matter)
I'll try it.
Looking forward for the next system lockup,
-- Dipl.-Phys. Roland Hautz Evolution Systems GmbH Network Administration Mobil-Oil-Str. 42 mailto:rh@evosys.net http://www.esib.net/ D-84539 Ampfing Tel.:(49)(8636)9830-32 Fax: (49)(8636)9830-99
- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu