> Please let me know if there's anything additional I can try
> besides just tickling this bug. (I can try to trigger it in
> previous kernels as well, if you want.)
Ditto. I am running on a P166, Buslogic BT948 SCSI controller, and
Conner CFP2107S Drive.
After my filesystem was completely destroyed the first time, I feared
a problem with the drive or the controller. I have since turned off
tagged queueing just to insure that I am not demanding to much from my
SCSI chain (not that i know what tagged queueing is or anything :).
I have experienced the following filesystem corruption symptoms:
a) when untarring a large source tree, sometimes one of the source
files will get munged. it appears that either a piece of the file is
missing or something has been inserted in the file. this seems to
happen without any error messages logged. unpacking the source tree
again fixes the problem.
b) sometimes my system will completely lock while ext2-fs error
messages scroll by (ext2_find_entry, some other ones maybe). this
always happens with heavy i/o is taking place (unapcking the kernel
sources while compiling xlock...) when the system finally recovers
from the temporary lock, potentially massive corruption has happened
to the filesystem. two days ago i ended up with tons of files which
had turned into directories (/bin/sh turned into a directory whose
contents were /usr/src!!!).
i am getting fed up of have to reinstall everything :). anything i
can do to help get this problem solved will be my pleasure.
thanks,
-- garth