From: Herbert Poetzl (herbert_at_13thfloor.at)
Date: Mon 17 Feb 2003 - 22:22:40 GMT
On Mon, Feb 17, 2003 at 11:15:06AM -1000, Warren Togami wrote:
> Herbert Poetzl wrote:
> >On Sat, Feb 15, 2003 at 02:08:26AM -1000, Warren Togami wrote:
> >
> >>I have seen consistent kernel panics on several systems running kernel
> >>2.4.20 ctx 16. This occurred on three Dual Athlon motherboards Tyan
> >>Tiger 760MP, Tyan Tiger 760MPX, and Tyan Thunder 760MPX running Red Hat
> >>Linux 7.3 and 8.0. I compiled my own custom kernel and using ext3 in
> >>all cases. Downgrading to any 2.4.19 ctx seems to make it 100% stable
> >>again.
> >>
> >>Things seem to be stable when load is low, however the kernel panics
> >>readily during high I/O activity. Load with PHP/MySQL transactions or
> >>building several source trees at the same time on these machines seems
> >>to trigger the kernel panic.
> >>
> >>I'm driving down to one of the servers tomorrow in order to read the
> >>kernel panic from the console. Could someone provide exact instructions
> >>for me to collect better debug info?
> >
> >
> >get http://www.kernel.org/pub/linux/utils/kernel/ksymoops/v2.4/
> >read the manpage and use as much additional information
> >as possible, also collect a complete lspci -vvv listing
> >and the dmesg listing of your system boot ...
> >
> >best,
> >Herbert
> >
>
> I am not able to get into the server room until Tuesday =(. In the mean
> time I was reading ksymoops documentation. I'm not completely sure how
> to use this? Do I need a serial console?
not for ksymoops ... although serial console
is a nice thing to have ...
> ksymoops says that it is for kernel OOPSes, how can this decode the
> panic message?
the "panic message" actually is a kernel oops
which includes the stack backtrace as hex values
- if you are able to capture this message, by
reading it on screen and writing it down, or by
copy/paste from serial console, or by ocr from
a screenshot ... you can later feed it through
ksymoops, which, if provided the necessary data
(on the actual system) will "translate" the hex
values into more meaningful stuff like function
symbols and offsets ...
take a look at the recent kernel oops postings
on this list (archive) and compare the message
with and without ksymoops translation ...
best,
Herbert
> Warren Togami
> warren_at_togami.com