xfs
[Top] [All Lists]

Re: Filesystem Consistency Issues

To: David Brown <xfs@xxxxxxxxxx>
Subject: Re: Filesystem Consistency Issues
From: Andi Kleen <ak@xxxxxxx>
Date: 05 Dec 2005 04:41:05 -0700
Cc: linux-xfs@xxxxxxxxxxx
In-reply-to: <20051205014137.GA7685@xxxxxxxxxxxxxx>
References: <20051204142506.GE2605@xxxxxxxxxxx> <20051205014137.GA7685@xxxxxxxxxxxxxx>
Sender: linux-xfs-bounce@xxxxxxxxxxx
User-agent: Gnus/5.09 (Gnus v5.9.0) Emacs/21.3
David Brown <xfs@xxxxxxxxxx> writes:

> On Sun, Dec 04, 2005 at 10:25:06PM +0800, Federico Sevilla III wrote:
> 
> > When I investigated one of the machines, mounting the filesystem in
> > read/write mode after having booted from a rescue CD automatically fixed
> > part of it, freeing about 60% of the filesystem. Running xfs_repair
> > further freed up space by moving disconnected inodes to lost+found.
> 
> Most likely, the problem is that the root filesystem is being mounted
> readonly by the kernel.  This is default, and I believe Debian leaves that
> as the default.  The problem is that the readonly root mount prevents the
> recovery when the filesystem is mounted.  Apparently, remounting as
> read/write doesn't cause this recovery to happen.

That sounds like a bug. Shouldn't log replay happening on ro->rw mount? 

BTW I think ext3/reiser solve the problem by replaying the logs even
when the mount is ro. That would be another alternative.

> Perhaps this should be added to the XFS FAQ, rather than just saying that
> "Yes", XFS can be used for the root FS.

It's a nasty trap that I bet most people and distributions get wrong.

-Andi


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