setclientid: string in use by client

[Trond Myklebust]

On Mon, 2007-03-05 at 21:41 -0500, J. Bruce Fields wrote:
> On Mon, Mar 05, 2007 at 04:56:31PM -0500, Trond Myklebust wrote:
> > Normally, the client in question should be able to work around this by
> > changing its clientid. Unless you are seeing a case where the clients
> > are actually unable to open files etc, this message can, and should be
> > ignored.
> > 
> > Bruce, could you please convert that to a dprintk()? There is absolutely
> > no reason to be printing out these "error messages" in stable kernels
> > unless there are known bugs that need to be identified and fixed.
> 
> Yeah, I guess you're right.  Will any real problems caused by the
> clientid in use result in obvious errors on the client, or could locks
> be silently lost?

If the client is unable to successfully set a client id, then it will be
unable even to open() a file (resulting in a hang and/or EIO error). In
addition, a printk will hit the syslog to the effect that "state
recovery failed on NFSv4 server %u.%u.%u.%u with error %d\n".

Cheers
  Trond