Re: Kernel update 3.5.7 -> 3.6.3 breaks NFS4

["J. Bruce Fields" <bfields@fieldses.org>]

On Mon, Oct 29, 2012 at 05:33:23PM +0100, Sven Geggus wrote:
> J. Bruce Fields schrieb am Montag, den 29. Oktober um 16:02 Uhr:
> 
> > Re-adding linux-nfs to cc
> 
> OK
> 
> > I don't understand why the server is dropping requests instead of
> > returning errors.  I actually would have expected it to return BADVERF
> > to the DESTROY request and then accept the PUTROOTFH normally, which
> > might have allowed the mount to succeed despite the bizarre rpc.gssd
> > behavior.
> 
> Hm rpc.gssd is the one from debian nfs-utils 1.2.6 looking at their custom
> patches inside the debian package there is nothing which could cause this.
> 
> > I'd be curious to understand what changed on the server to make a
> > difference. 
> 
> Nothing but the kernel. I'm currently dual booting either 3.5.7 or 3.6.3
> vanilla kernels on the same system.
> 
> > Looking at a network trace from a successful mount with 3.5.7 might be
> > useful.
> 
> attached.

Thanks!

The sequence of events is pretty much what I described for the "bad"
trace, except that PUTROOTFH gets a succesful reply.

One other odd difference: in the "bad" case, the timing is a little
different: the socket gssd created doesn't get shut down in the same
way, and the PUTROOTFH comes more quickly on the heels of the FIN.
Which shouldn't make a difference.

Hm.

--b.