Return to BSD News archive
Path: euryale.cc.adfa.oz.au!newshost.anu.edu.au!harbinger.cc.monash.edu.au!simtel!fu-berlin.de!fub46.fddi1.fu-berlin.DE!not-for-mail
From: gusw@zedat.fu-berlin.de (Gunther Schadow)
Newsgroups: comp.unix.bsd.freebsd.misc
Subject: Re: NFS problems...
Date: 29 Sep 1995 20:57:19 GMT
Organization: Freie Universitaet Berlin
Lines: 112
Message-ID: <44hmjf$8vc@fu-berlin.de>
References: <42mu88$e8f@fu-berlin.de> <42ncku$qtp@sol.ctr.columbia.edu>
NNTP-Posting-Host: fub46.fddi1.fu-berlin.de (160.45.1.46)
X-Access: 16 17 19
thank's all who answered my last call for help on the NFS mount. I am
now able to mount an NFS exported from an HP-UX workstation, but still
I have problems with a Novell NFS Server. It is likely that I am not
deeply enough inside RPC to understand what's happening, but I woul
appreciate your help even though the problem might have less to do
with FreeBSD. However, I realized specific differences in the behavior
of the HP-UX to the Novell server versus the FreeBSD system. Again,
it's a 2.1.0-950726-SNAP FreeBSD.
Now, when I ask the Novell machine for the status of RPC services
registration in it's portmapper, i get the following:
> rpcinfo -p anae_iop 2>/dev/null
program vers proto port
100000 2 udp 111 portmapper
100000 2 tcp 111 portmapper
300055 2 udp 4833
100005 1 udp 4835 mountd
100005 2 udp 4836 mountd
100003 2 udp 2049 nfs
with a lot of "clnttcp_create: RPC: Program not registered" messages
sent to stderr. These are not reported on the HPUX and are as well
reported when I do a "rpcinfo -p localhost" on the FreeBSD system. So
I conclude that these `clnttcp_create' errors are a problem of
FreeBSD, but what? Anyway, the `clnttcp_create' problem seems not to
matter here, since it does occur for request to the HPUX system as
well, but does not interfere with a proper mount of an NFS volume from
it.
Now, I will ask the Novell server for specific services by polling the
procedure 0 via the udp channel:
> rpcinfo -u anae_iop nfsd
[no answer, timeout after some time]
but when I specify the port 2049:
> rpcinfo -n 2049 -u anae_iop nfsd
clnttcp_create: RPC: Program not registered
[repeats 7x, then:]
program 100003 version 2 ready and waiting
which is what I expect. The same holds for mountd: no response if port
number isn't supplied. I conclude that the problem is due to the
portmapper, either the Novell portmapper is broken or the FreeBSD's
mode of communication with it. Since this is so, I tried the -P option
to mount_nfs but with no success.
Compared to the HPUX, I found that HPUX seems to have no portmapper
running (even though `ps' shows one, and I started it myself), and
HPUX has the mountd on a different port number than the Novell server:
> rpcinfo -p uks3p 2>/dev/null
program vers proto port
100005 1 udp 978 mountd
100005 1 tcp 980 mountd
100003 2 udp 2049 nfs
100024 1 udp 985 status
100024 1 tcp 987 status
100021 1 tcp 988 nlockmgr
100021 1 udp 2496 nlockmgr
100021 3 tcp 992 nlockmgr
100021 3 udp 2497 nlockmgr
100020 1 udp 2498 llockmgr
100020 1 tcp 997 llockmgr
100021 2 tcp 1000 nlockmgr
Nevertheless, I am able to mount NFS from HPUX even without the
portmapper and without specifying -P to mount_nfs. Amazing, isn't it?
BTW: the polling for nfs and mountd on the Novell server shown above
succeeds without hesitation if issued from the HPUX system.
Finally, to complete the diagnostics, I ask for the status of the
portmapper istself by polling it:
> rpcinfo -n 111 -u anae_iop portmapper
clnttcp_create: RPC: Program
[repeated 5x, then:]
not registered
rpcinfo: RPC: Timed out
program 100000 version 2 is not available
but for the tcp channel, the things seem to work:
> rpcinfo -n 111 -t anae_iop portmapper
clnttcp_create: RPC: Program
[repeated 5x, then:]
program 100000 version 2 ready and waiting
again, both operations work without problems on HPUX using either tcp
or udp and without *having* to specify the port number by -n 111. If I
omit the -n 111 option on the FreeBSD system, the polling fails even
on the tcp channel.
Thus I conclude, that the communication to the Novell portmapper *via
UDP* is broken on FreeBSD. On the other hand, I see no way to specify
tcp communication for the default portmapping call (mopunt_nfs -T
doesn't help).
So, all BSDoctors, what is wrong here? I appreciate any hint (and I
hope the whole thing works out, since the reputation of FreeBSD which
I introduced to my institution is in danger...).
thank you,
-Gunther Schadow
--
Gunther Schadow__________________Windsteiner Weg 54a, Berlin 14165, FR. Germany
Dept. of Anaesthesia and Operative Intensive Care Medicine/
Dept. of Medical Informatics at the |gusw@fub46.zedat.fu-berlin.de
Universitaetsklinikum Steglitz, Berlin, FRG. |#include <usual/disclaimer>