This is a follow up to my previous message.
I've been able to kill 2.2.8 by raising the level of write activity
to a mix of IDE and SCSI drives. By untaring a kernel tarball into space
on my SCSI drive and resyncing my elm mailbox on the IDE drive, 2.2.8
has a nasty habit of dying.
I've now performed this combination 6 times on 2.2.8 and killed
it 4 of the 6 times. I've done the same thing, intentionally (doesn't
count the dozens of times I've done it routinely over the last month or so),
on 2.2.7 and not killed it once. Of the four times it crashed, twice
fsck gave me unexpected disk errors that required manual fsck. Each time,
the guilty party was an inode with illegal blocks. Each time, that inode
was associated with a coreball in my home directory, which should not have
been there. I had not gotten any recent segfaults in my home directory and
there was no coreball prior to running the crash test.
Another point on the curve... It has only occured while elm was
writing to its temp file (also on an IDE partition). If it gets to the
point of rereading the mail spool file, I seem to be safe. It has
caused significant disk corruption and loss of message out of my mailbox,
however... Enough playing with that box of matches... :-(
Over all... Looks like 2.2.8 might rate a one "bag over head".
I'm not sure where or how to proceed from here so I've just dumped
2.2.8 and dropped back to 2.2.7 unless someone can come up with some
constructive suggestions as to how to debug this sucker (twice it has
left me were I couldn't even toggle the NumLock light).
Back to hacking on Computone serial drivers... :-/
Mike
-- Michael H. Warfield | (770) 985-6132 | mhw@WittsEnd.com (The Mad Wizard) | (770) 925-8248 | http://www.wittsend.com/mhw/ NIC whois: MHW9 | An optimist believes we live in the best of all PGP Key: 0xDF1DD471 | possible worlds. A pessimist is sure of it!
- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/