[This is slightly off topic, but since the e2compr list seems essentially
dead, I'll take the risk and continue the thread here. Somebody might be
interested here on l-k.]
On Sun, Jun 11, 2000 at 08:08:06PM -0400, you [Simon Kirby] claimed:
> On Sun, Jun 11, 2000 at 11:53:48PM +0300, Ville Herva wrote:
>
> > Just for the record, I get "Whoops: end_buffer_io_async: async io complete
> > on unlocked page" messages in the logs every now and then with a dual PII
> > 2.2.15pre17. The kernel has ext2compr, ide and raid patches on it, so it
> > could be anything. The box seems very stable despite the log messages.
> > (Well, there is a reliable way to lock up kernel solid via a e2compr bug
> > that I just can't trace down. That's however easy to avoid.)
>
> We have a machine here patched with e2compr used for large NFS remote
> backups. We see the message as well, quite a few times during a backup
We do backups from Windows machines via rsync onto a e2compressed softraid
partition (that is then shared via samba).
> run. I'm fairly sure the "Whoops" message is coming from the ext2
> compression patches, and it's even listed on the e2compr pages as a bug.
It is? I didn't notice...
> It would be nice to get to the bottom of where the message is coming
> from...it's been happening for as long as I can remember, even back in
> the 2.1 days, I think.
Yes. E2compr seems very stable otherwise, so it's a shame this bug still
lingers. I'm not sure, but it could have something to do with the
reproducible lock up. It first showed up when deleting files through
samba, but I have since been able to reduce it into a ten line c program
that reproduces it reliably. It seems that a buffer head is locked
somewhere, but the lock is not released. The kernel then hangs waiting for
the lock. I even set up a test machine to debug it with kdb, but I just
don't have the knowledge about ext2 locking semantics... For a
knowledgeable person, this should be a piece of cake. I posted the kdb
traces and everything I found out to e2compr mailing list (thread starting
from http://e2compr.memalpha.cx/pipermail/e2compr/2000-April/000039.html
(*)).
-- v --
(*) It seems the archiving software mangled some of my messages, sorry for
that. I can mail them if anyone is interested.
-
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/
This archive was generated by hypermail 2b29 : Thu Jun 15 2000 - 21:00:24 EST