xfs
[Top] [All Lists]

RE: quota mismatch for NFS mount of XFS filesystem

To: Dave Chinner <david@xxxxxxxxxxxxx>, Liam Forbes <loforbes@xxxxxxxxxx>
Subject: RE: quota mismatch for NFS mount of XFS filesystem
From: "Burbidge, Simon A" <s.burbidge@xxxxxxxxxxxxxx>
Date: Wed, 9 Nov 2011 09:52:06 +0000
Accept-language: en-GB, en-US
Cc: "xfs@xxxxxxxxxxx" <xfs@xxxxxxxxxxx>, "Burbidge, Simon A" <s.burbidge@xxxxxxxxxxxxxx>
In-reply-to: <20111108225513.GO5534@dastard>
References: <CACXvBApxwWUyYNfRL9rS32t7f04XCDwafbR7fSFjmXg8fJRpTg@xxxxxxxxxxxxxx>,<20111108225513.GO5534@dastard>
Thread-index: AQHMnkelpSuKraZqmEiTzOYNg53vg5WjlmOAgAC2HSk=
Thread-topic: quota mismatch for NFS mount of XFS filesystem
Hi Liam,
I believe we've encountered this exact same problem.
There is a 32 bit size field in the protocol used by rpcquota . What you are 
seeing is overflow of that field.
I figure that this should be classed as a classic "not 64-bit ready" bug - but 
the support people from the Linux distros seem to believe
that it is working according to specification.

I'm not even sure that NFS v4 addresses the issue!

Best Regards
Simon Burbidge
Imperial College London

________________________________________
From: xfs-bounces@xxxxxxxxxxx [xfs-bounces@xxxxxxxxxxx] on behalf of Dave 
Chinner [david@xxxxxxxxxxxxx]
Sent: 08 November 2011 22:55
To: Liam Forbes
Cc: xfs@xxxxxxxxxxx
Subject: Re: quota mismatch for NFS mount of XFS filesystem

On Tue, Nov 08, 2011 at 09:31:05AM -0900, Liam Forbes wrote:
> I searched the archives, but didn't find anything about incorrect
> quota information for an NFS mounted XFS filesystem.  I hope somebody
> might have an idea.
>
> We have a RHEL 5.7 server mounting a 15T XFS filesystem and exporting
> it to RHEL 5.6 and 5.7 clients.  quota and xfs_quota on the server
> show a user using ~5.8T, but quota on a client only shows ~1.8T.
>
> Server:
> nfs2:~$ cat /etc/redhat-release
> Red Hat Enterprise Linux Server release 5.7 (Tikanga)
> nfs2:~$ uname -a
> Linux nfs2 2.6.18-274.3.1.el5 #1 SMP Fri Aug 26 18:49:02 EDT 2011 x86_64 
> x86_64
> nfs2:~$ rpm -qa | grep xfsprogs
> xfsprogs-2.9.4-1.el5.centos
> nfs2:~$ sudo quota -s -v -u krieger
> Disk quotas for user krieger (uid 3105):
>      Filesystem  blocks   quota   limit   grace   files   quota   limit   
> grace
> /dev/mapper/primaryfs_vg-primaryfs
>                   5901G       0       0            310k       0       0
> nfs2:~$ sudo /usr/sbin/xfs_quota
> xfs_quota> quota -u -h krieger
> Disk quotas for User krieger (3105)
> Filesystem   Blocks  Quota  Limit Warn/Time    Mounted on
> /dev/primaryfs_vg/primaryfs
>                5.8T      0      0  00 [------] /export/wrkdir20
> xfs_quota> quit
>
> Client:
> pacman1:~$ cat /etc/redhat-release
> Red Hat Enterprise Linux Server release 5.6 (Tikanga)
> pacman1:~$ uname -a
> Linux pacman1 2.6.18-238.19.1.563g0000.el5 #1 SMP Fri Jul 15 20:26:36
> EDT 2011 x86_64 x86_64 x86_64 GNU/Linux
> pacman1:~$ sudo quota -s -v -u krieger
> Disk quotas for user krieger (uid 3105):
>      Filesystem  blocks   quota   limit   grace   files   quota   limit   
> grace
> nfs2:/export/wrkdir20
>                   1806G       0       0            310k       0       0

That looks like a 32bit variable overflow. In 512 byte blocks
(sectors), 2^32 sectors overflows at 2T. And 5.8T ~= 2T + 2T + 1.8T.

This is most likely to be an NFS issue - I'm not familiar with the
Linux NFS quota protocol implementation, so I can't really say much
more than that.

Cheers,

Dave.
--
Dave Chinner
david@xxxxxxxxxxxxx

_______________________________________________
xfs mailing list
xfs@xxxxxxxxxxx
http://oss.sgi.com/mailman/listinfo/xfs

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