> Well, just prior to the oops, I got two messages from umount, saying that
> /dev/hda2 (an ext2 filesystem) and /proc were still busy. On reboot, fsck
> said /dev/hdb2 (the root filesystem, another ext) was clean, so I presume
> it was unmounted correctly. This leaves /dev/hda1 and /dev/hdb1, msdos
> partitions, as possible culprits. If so, this wouldn't be the first time
> I've had msdos filesystem-related problems... in previous 2.1.5x kernels,
> I would frequently get "VFS: Busy inodes after unmount. Self-destruct in 5
> seconds. Bye-bye.." messages on umount'ing msdos filesystems.
OK, I'm especially interested in msdos fs problems. I fixed a number of
errors that would cause the "busy inodes" problem, but there may be
other errors as well.
Could you attempt to recreate the problem, and if possible umount the
msdos partitions first (while syslog is still running) to try to capture
any messages? The most helpful report would be a specific operation that
causes an oops at umount time, e.g. "If I mount msdos, do xxx, and then
umount, I get this oops ...".
Regards,
Bill