[Samba] Samba-3.0.23c kernel lock problems with new Redhat kernel 2.6.9-42.0.8

[Jeremy Allison]

On Fri, Feb 02, 2007 at 04:45:08PM +0100, Ole Holm Nielsen wrote:
> We run samba-3.0.23c on some Redhat RHEL4 servers, and Samba used to work
> like a charm.  But then a couple of days ago we upgraded the kernel
> on the Samba servers to kernel-smp-2.6.9-42.0.8.EL.  All of a sudden
> our Windows users could not use Microsoft Office with files on the
> Samba shares :-(
> 
> Our filesystems are actually NFS-mounted by the Samba server from
> another server, and the Samba server showed these log entries:
> 
> Feb  1 13:27:41 serv309 smbd[11728]: [2007/02/01 13:27:41, 0] 
> locking/posix.c:posix_fcntl_lock(660)
> Feb  1 13:27:41 serv309 smbd[11728]:   posix_fcntl_lock: WARNING: lock 
> request at offset 2147483538, length 1 returned
> Feb  1 13:27:41 serv309 smbd[11728]: [2007/02/01 13:27:41, 0] 
> locking/posix.c:posix_fcntl_lock(661)
> Feb  1 13:27:41 serv309 smbd[11728]:   an No locks available error. This 
> can happen when using 64 bit lock offsets
> Feb  1 13:27:41 serv309 smbd[11728]: [2007/02/01 13:27:41, 0] 
> locking/posix.c:posix_fcntl_lock(662)
> Feb  1 13:27:41 serv309 smbd[11728]:   on 32 bit NFS mounted file systems.
> Feb  1 13:27:41 serv309 smbd[11728]: [2007/02/01 13:27:41, 0] 
> locking/posix.c:posix_fcntl_lock(660)
> Feb  1 13:27:41 serv309 kernel: lockd: cannot monitor 130.225.87.40
> Feb  1 13:27:41 serv309 kernel: lockd: failed to monitor 130.225.87.40
> 
> With the previous kernel kernel-smp-2.6.9-42.0.3.EL there were no
> such problems !  The Redhat kernel Release Notes
>    https://rhn.redhat.com/errata/RHSA-2007-0014.html
> do not mention any changes that seem to be related to Samba or NFS or 
> locking.
> 
> Fortunately I found this article
>    http://lists.samba.org/archive/samba/2006-October/126638.html
> where Jeremy recommends to use "posix locking = no", and indeed
> this fixes the problem !
> 
> Hopefully these observations can help others, but a real solution
> to the problem would be most welcome !

Looks like a kernel issue to me. Have you reported this to Red Hat ?

Jeremy.