On Sun, 01 Jun 2003 05:50:12 -0500,
"Jeffrey E. Hundstad" <jeffrey@xxxxxxxxxxxx> wrote:
>For the record I just has linux-2.4.21-rc6-xfs (latest cvs) crash. It
>was looking over at the linux kernel mailing list and it seem that other
>folks have been having problem with the 2.4.21-rc? series of kernels
>hanging on high i/o with high memory usage. ...and that's when I'm
>crashing also. It does seem like I can make this crash pretty much "at
>will" at this point. ...start a backup... wait an hour... and BOOM.
>Since the theory is that the box that is crashing is a production server
>I'm going to back off to my stable but prone to root exploit 2.4.18-xfs
>kernel. I'll try to set up a box that is just as crashy but
>debugabble. Does anyone have a URL or text location to where I can
>learn to use the kernel debugger or whatever to give good reports after
>an oops, 'cause I think I may have the ideal circumstance to cause the
>crashes. Thanks!
From your comments, it is unlikely to be an XFS problem. OTOH you have
a kernel with the latest kdb, so let's investigate this a bit more.
Are you running SMP or UP, if UP does your motherboard have an IO APIC?
What sort of error are you seeing, is it a straight hang, does the NMI
watchdog kick in, are you running the NMI watchdog? Or do you get an
oops? The answer to those questions will guide the debugging effort.
In any case, you will need a serial console and a second machine to
capture the debug data. Compile with
CONFIG_VT=y
CONFIG_VT_CONSOLE=y
CONFIG_SERIAL=y
CONFIG_SERIAL_CONSOLE=y
Boot with 'console=tty0 console=ttyS0,38400'. Run a null modem from
ttyS0 (COM1) to a second machine and capture the output using your
favourite terminal emulator, I use minicom.
|