xfs
[Top] [All Lists]

Re: [PATCH 5/4] xfs: fix nfs export of 64-bit inodes numbers on 32-bit k

To: Christoph Hellwig <hch@xxxxxxxxxxxxx>
Subject: Re: [PATCH 5/4] xfs: fix nfs export of 64-bit inodes numbers on 32-bit kernels
From: Ben Myers <bpm@xxxxxxx>
Date: Tue, 6 Dec 2011 10:39:13 -0600
Cc: xfs@xxxxxxxxxxx
In-reply-to: <20111202172908.GA10803@xxxxxxxxxxxxx>
References: <20111128081732.350228200@xxxxxxxxxxxxxxxxxxxxxx> <20111130085817.GA22471@xxxxxxxxxxxxx> <20111202160712.GO29840@xxxxxxx> <20111202172908.GA10803@xxxxxxxxxxxxx>
User-agent: Mutt/1.5.18 (2008-05-17)
On Fri, Dec 02, 2011 at 12:29:08PM -0500, Christoph Hellwig wrote:
> On Fri, Dec 02, 2011 at 10:07:12AM -0600, Ben Myers wrote:
> > To what extent did you test this one?  Anything in particular I should
> > look for when I test it?
> 
> Create a large filesystem, mount it using -o inode64 on a 32-bit kernel,
> find an inode that actually uses more than 32-bits, and try to access
> it from an nfs client.
> 
> Alternatively you could probably reproduce it using the open by handle
> system calls, I have an idea how to turn that into a test case for
> xfstests using a large loopback filesystem.

I tested this on a pair of i386 boxes with a large loop filesystem and
inode64.  Without the fix I got ESTALE on any inodes > 32 bits, and with
the patch it works great.

Signed-off-by: Ben Myers <bpm@xxxxxxx>

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