xfs
[Top] [All Lists]

Re: XFS bug? (was: Red Hat Linux 9 XFS DVD Released)

To: Steve Lord <lord@xxxxxxx>
Subject: Re: XFS bug? (was: Red Hat Linux 9 XFS DVD Released)
From: Stefan Smietanowski <stesmi@xxxxxxxxxx>
Date: Tue, 15 Apr 2003 20:20:11 +0200
Cc: Axel Thimm <Axel.Thimm@xxxxxxxxxxxxxxxxxxx>, Mogens Kjaer <mk@xxxxxx>, linux-xfs@xxxxxxxxxxx
References: <3E94C880.2070003@xxxxxxxxxx> <3E96BB77.5000902@xxxxxx> <1050067526.7662.8.camel@xxxxxxxxxxxxxxxxxxx> <3E96CC6E.7010707@xxxxxx> <3E9BFD51.8060206@xxxxxx> <20030415163859.GB16672@xxxxxxxxxxxxxx> <1050426530.1166.248.camel@xxxxxxxxxxxxxxxxxxxxxxx>
Sender: linux-xfs-bounce@xxxxxxxxxxx
User-agent: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.0.1) Gecko/20021003
Steve Lord wrote:
On Tue, 2003-04-15 at 11:38, Axel Thimm wrote:

So what is the fault there? Why does one need to unmount and remount, and more
important, when does one have to do so?

There have been reports with people upgrading their rpm (xfs enabled) kernels
and crashing the GRUB second load stage. At the first sight this looks like
the same bug the installer sees (these people have obviously installed the new
kernel and rebooted a short while after, like the installer does).

Maybe xfs root (or boot) filesystems don't write dirty buffers back, and
(as a workaround) should always be remounted before shutting them down?


If you have ext3 filesystems active on the same box then a bug in
ext3 can actually prevent kernel threads from flushing anything else
to disk. This is fixed in the latest 2.4.21-pre kernel, but not a
lot of other places. If folks are running with ext3 in the mix this
may be the cause of the problem.

Steve




This bug appears also on pure XFS systems. I even hacked XFS to always mount the filesystem sync and it still didn't help.

// Stefan


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