[Top] [All Lists]

Re: [] assfail in xfs_attr_quiesce() during umount

To: Paul Smith <paul@xxxxxxxxxxxxxxxxx>
Subject: Re: [] assfail in xfs_attr_quiesce() during umount
From: Christoph Hellwig <hch@xxxxxxxxxxxxx>
Date: Tue, 7 Jul 2009 15:06:43 -0400
Cc: Linux XFS <linux-xfs@xxxxxxxxxxx>, christoph hellwig <hch@xxxxxxxxxxxx>, david chinner <dgc@xxxxxxx>, linux-kernel@xxxxxxxxxxxxxxx
In-reply-to: <MAIL10OSN4RHlLx5ERz0002f238@xxxxxxxxxxxxxxxxx>
References: <MAIL10OSN4RHlLx5ERz0002f238@xxxxxxxxxxxxxxxxx>
User-agent: Mutt/1.5.18 (2008-05-17)
On Tue, Jul 07, 2009 at 02:42:16PM -0400, Paul Smith wrote:
> According to the thread for the former, it was intended to be resolved
> by "proper per-vfsmount write count in 2.6.25 as part of the per-mount
> r/o patches" (C.Hellwig)
> Did those fixes indeed go into 2.6.25 (or 2.6.27 or below)?  Was there
> something else that needed to be done for XFS to be able take advantage
> of that enhancement?

We haven't fixed the VFS issues yet.  In some later kernel we just
changed the assert to a warn on for now until the issue is sorted out.

> And, is the patch posted by DavidC here:
> http://marc.info/?l=linux-kernel&m=120106649923499&w=2

It looks somehwat dangerous for me.  For now I would just comment
out the assert or change it to a WARN_ON as in the newer kernels.

I've recently started looking at that area in the VFS again and I'll
hopefully make progress on a solution that sets MS_RDONLY at the
correct time.

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