xfs
[Top] [All Lists]

Re: The mmap() problem is back

To: "Keith Owens" <kaos@xxxxxxx>, "Florian Weimer" <Weimer@xxxxxxxxxxxxxxxxxxxxx>
Subject: Re: The mmap() problem is back
From: "Sander Steffann" <sander@xxxxxxxxxxx>
Date: Mon, 5 Aug 2002 21:14:15 +0200
Cc: "Ethan Benson" <erbenson@xxxxxxxxxx>, <linux-xfs@xxxxxxxxxxx>
References: <3674.1028557285@ocs3.intra.ocs.com.au> <87znw1yz7g.fsf@CERT.Uni-Stuttgart.DE>
Sender: owner-linux-xfs@xxxxxxxxxxx
> I've encountered the same problem with ReiserFS ages ago.  In this
> case, the page cache had the correct view, but the data on disk was
> wrong.  Unless the page was somehow flushed, the real disk contents
> didn't show up until the next reboot.

Hmmm. This sounds like a problem we had several months ago... Only we ran
RedHat 6.2 with a 2.2 kernel and an ext2 filesystem. We noticed something
was wrong when rsync started to hang. Even a kill -9 couldn't stop it. At
that moment, the filesystem seemed damaged, but not too bad. After a reboot
the main superblock was gone, and even fsck.ext2 couldn't make much of the
filesystem... We had to restore a backup from tape :(

This happen twice in two weeks. Both times it was the same filesystem that
was messed up, but since it was also the most active filesystem that makes
sense.

We suspected bad hardware. Mostly because the airconditioning had broken,
and another mainboard had already died from the heat... Mainboard, processor
and memory were replaced, and we never had that problem again.

I hope this helps. Funny that exactly the same thing can happen on three
different filesystem-types.

Good luck,
Sander



<Prev in Thread] Current Thread [Next in Thread>