On Tue, Jan 07, 2003 at 08:34:27AM +0100, T. Ribbrock wrote:
[U5 OpenBSD 3.2 NFS server <-> Linux 2.1.18 NFS client]
[...]
> All programs that use locking (e.g. mutt -> fcntl) throw errors ("Can't get
> lock", "lock failed" and similar) and I am getting error messages on the
> Linux box as well each time a program tries to set a lock:
>
> Jan 6 21:53:12 esme kernel: call_verify: server accept status: 2
> Jan 6 21:53:12 esme last message repeated 2 times
> Jan 6 21:53:12 esme kernel: RPC: garbage, exit EIO
>
> On the OpenBSD box I cannot see any error messages.
[...]
Further archive and google research together with some testing on my
part has resulted in the following:
Apparently, OpenBSD 3.2's 'rpc.lockd' only support NLM 1 and 3 (this
can be seen using rpcinfo - 'man rpc.lockd' actually mentions the
versions 1, 2 and 3), which - according to RFC1813 [0] - corresponds
to NFS V2. NLM 4 (corresponding to NFS V3) does not seem to be
supported. OpenBSD's NFS server, however, does support NFS v3.
An ethereal session of the traffic between the Linux client and the
OpenBSD server does indeed show the exchange of NFS V3 traffic.
Whenever locking is tried on the Linux client, one can see the OpenBSD
box complain about the NLM4 protocol - which is doesn't know.
I then went ahead and mounted the exported directory using
"-o nfsvers=2" on the Linux client. Et voila: Locking suddenly works!
Ethereal confirms that NFS traffic is now V2 and the NLM traffic V1.
Hence, the solution seems to be: If one wants locking via NFS with a
Linux client and an OpenBSD server, one has to use NFS V2.
Cheerio,
Thomas
--
-----------------------------------------------------------------------------
Thomas Ribbrock http://www.ribbrock.org ICQ#: 15839919
"You have to live on the edge of reality - to make your dreams come true!"
--
redhat-list mailing list
unsubscribe mailto:[EMAIL PROTECTED]?subject=unsubscribe
https://listman.redhat.com/mailman/listinfo/redhat-list
