Re: NFS locking and statd
Bill Hawes (whawes@star.net)
Tue, 16 Dec 1997 08:43:09 -0500
okir@monad.swb.de wrote:
> The problem is that lockd is a kernel-mode process (for a good reason),
> but statd shouldn't be.
>
> As I said, we can get rid of the kernel RPC upcalls and fold the
> statd calls into the mount(8) and rpc.mountd programs. This would make
> use of statd user-controllable via command line options on both apps.
>
> The only problem I see is that when a remote non-Linux machine reboots,
> its statd may sit there forever trying to contact the Linux statd, which
> isn't there.
>
> However, I'm not too fond of running lockd without statd. It complicates
> things a lot in lockd, I'm afraid.
OK, it sounds like for now we should just document that statd is
required to use locking under NFS. The present kernel lockd code returns
error if the SM_MON call fails anyway. Then at some point if enough
people want statd to be optional we can think about changing it.
Regards,
Bill