Long-standing problem partially solved.

Colm Buckley (colm@tuatha.org)
Fri, 11 Jul 1997 15:35:34 +0100

Hi all -

I've just partially fixed something which has been troubling me for a
*long* time - namely, the failure of "e2fsck" to fix errors on a
software RAID-0 filesystem without crashing the machine - hitherto, the
fsck on bootup would reach the md0 filesystem, get partway through the
fsck, and then Linux would crash. This happened with many Linux
versions, mostly in the 2.1 series. The puzzling thing was, when I
booted the same kernel from a floppy, I could fsck the filesystem with
no trouble.

The problem *seems* to lie with the fact that raid0 was loaded as a
module, and that kerneld was running. I *suspect* that, partway through
the fsck, the module would be unloaded (autocleaned), causing the
computer to crash - the solution was to explicitly modprobe the raid0
module before running fsck, and load kerneld after the fsck. If the
module is loaded automatically by kerneld, or if kerneld is running
while the fsck is running, the computer crashed partway through the

I hope this helps someone...


Colm Buckley B.A. B.F. | EMail : colm@tuatha.org or colm@lspace.org
Computer Science       | WWW   : http://isg.cs.tcd.ie/cbuckley/
Trinity College        | Phone : +353 87 469146 (087-469146 within Ireland)
Dublin 2, Ireland      | Microsoft - Where do you want to crash today?