Bug#670505: [3.2.12 -> 3.2.14 regression] BUG: unable to handle kernel paging request in nfs_mark_delegation_referenced

[Rik Theys <Rik.Theys@esat.kuleuven.be>]

Hi Jonathan,

On Thu, 26 Apr 2012, Jonathan Nieder wrote:
> Rik Theys wrote:
>
> > It seems stable update 3.2.14 or 3.2.15 introduced a regression regarding NFS4 delegations.
> > It seems to be the same issue as reported here[1] against the 3.3 kernel. It was also intruduced
> > there between 3.3.0 and 3.3.1.
> [...]
> > BUG: unable to handle kernel paging request at ffffffffffffffb8
> > IP: [<ffffffffa048f016>] nfs_mark_delegation_referenced+0x6/0x6 [nfs]
>
> In [1], Jeff Layton says it looks like a regression introduced by
>
>  3114ea7a24d3 NFSv4: Return the delegation if the server returns
>               NFS4ERR_OPENMODE
>
> and suggests a fix.  Does this patch help?
>
> -- >8 --
> From: Trond Myklebust <Trond.Myklebust@netapp.com>
> Date: Tue, 27 Mar 2012 18:31:25 -0400
> Subject: NFSv4: Minor cleanups for nfs4_handle_exception and nfs4_async_handle_error
>
> commit 14977489ffdb80d4caf5a184ba41b23b02fbacd9 upstream.

If I read [1] correctly this patch was pushed in a Fedora update, and results 
in the following message being spewed at syslog by the thousands (filling the 
disk):

NFS: nfs4_reclaim_open_state: Lock reclaim failed!

I will try applying the patch to the Debian kernel next week, but I assume it 
will have the same result. If so, I believe it would be better to revert the 
bad commit for now, and apply a complete fix later when it becomes available.

Regards,

Rik