Re: flock() over NFS

[Brian Long <brilong cisco com>]

On Fri, 2004-12-03 at 15:16, Tom 'spot' Callaway wrote:
> On Fri, 2004-12-03 at 14:25 -0500, Brian Long wrote:
> > > > FYI, if we mount the NFS filesystem with -nolock, this gets around the
> > > > behaviour, but it also opens up the posibility of problems for programs
> > > > using fcntl().  Any ideas on how to resolve this would be appreciated!
> > 
> > > I think fnctl() for locking is really supposed to be the way going
> > > forward.. If all your makefiles are belong to some flock-a-file util,
> > > then switching that util shouldn't be *too* painful, true?
> > 
> > False; this has been tried and it appears fcntl() is unable to lock
> > files in certain instances when flock() was able to lock them.
> 
> This has been discussed to death here:
> https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=123415
> 
> If fcntl() doesn't work, file bugs on that.

Actually, it appears GFS-related patches were made to the flock/fcntl
code in 2.4.21-15 and higher.  flock() over NFS works on all kernels up
to and including 2.4.21-9.0.3.EL.  This means Red Hat changed the API
(in my opinion) and broke flock() over NFS in RHEL 3.  Of course, I'm
sure Red Hat is unwilling to remove or reconsider any GFS-related
patches even if they broke existing RHEL 3 functionality.  :-(

/Brian/

-- 
       Brian Long                      |         |           |
       IT Data Center Systems          |       .|||.       .|||.
       Cisco Linux Developer           |   ..:|||||||:...:|||||||:..
       Phone: (919) 392-7363           |   C i s c o   S y s t e m s