oops during nfs4 mount

[Trond Myklebust]

On Sat, 2008-04-05 at 21:10 -0400, Jeff Layton wrote:
> > Hmm... Bruce & Jeff,
> > 
> > Doesn't the line
> > 
> > 	nfs_callback_info.task = NULL;
> > 
> > in nfs_callback_svc() need protection by the BKL?
> > 
> 
> Moving it under the BKL probably won't hurt anything, but isn't it
> already protected by the nfs_callback_mutex?

AFAICS, nfs_callback_svc() doesn't (and in fact, can't) take the mutex. 

The point of the BKL here would be to protect nfs_callback_down against
the race where nfs_callback_svc() terminates early, and clears
nfs_callback_info.task.

Cheers
  Trond