On Thu, Feb 10, 2000 at 12:16:30PM -0500, Mason Loring Bliss wrote:
> Hi, all.
> 
> I haven't paid close attention to this, given my elusively bad RAM over
> the last month, but I share source trees using NFS, and there seem to
> be problems.
> 
> This is from tcpdump, showing attempted mounts:
> 
> 11:25:23.601839 acheron.domain > styx.65533:  3850* 1/1/1 A 0.0
> .115.230 (104)
> 11:25:23.603569 styx.1019 > acheron.sunrpc:  udp 56
> 11:25:23.605127 acheron.sunrpc > styx.1019:  udp 28
> 11:25:23.606883 styx.1018 > acheron.sunrpc:  udp 56
> 11:25:23.615103 acheron.sunrpc > styx.1018:  udp 28
> 11:25:23.617205 styx.1017 > acheron.1022:  udp 104
> 11:25:23.617642 acheron.1022 > styx.1017:  udp 68
> 11:25:23.620794 styx.778506246 > acheron.nfs: 96 getattr [|nfs]
> 11:25:23.620969 acheron.nfs > styx.778506246: reply ok 28 getat
> tr ERROR: 'Input/output error' [|nfs]
> 11:25:23.622199 styx.778506247 > acheron.nfs: 92 fsinfo [|nfs]
> 11:25:23.622300 acheron.nfs > styx.778506247: reply ok 32 fsinf
> o ERROR: 'Stale NFS file handle'
> 11:25:23.623183 styx.778506248 > acheron.nfs: 92 fsstat [|nfs]
> 11:25:23.623276 acheron.nfs > styx.778506248: reply ok 32 fssta
> t ERROR: 'Input/output error' [|nfs]
> 11:25:23.716509 acheron.nfs > styx.778506249: reply ok 32 fsinfo ERROR: 'Stale NFS file handle'
> 11:25:23.717620 styx.778506250 > acheron.nfs: 92 fsstat [|nfs]
> 11:25:23.717718 acheron.nfs > styx.778506250: reply ok 32 fsstat ERROR: 'Input/output error' [|nfs]11:25:23.718691 styx.778506251 > acheron.nfs: 92 fsinfo [|nfs]
> 11:25:23.718792 acheron.nfs > styx.778506251: reply ok 32 fsinfo ERROR: 'Stale NFS file handle'
> 
> Acheron is an i386 running 1.4S. Styx is a SPARC running 1.4P. Styx mounts
> Acheron's /usr/src and a couple other things.
> 
> The port numbers on Styx seem bizarrely large. Is this an RPC thing or
> something? Styx hasn't changed since before NFS stopped working, in any
> event, so it would seem likely that the change is in Acheron, somewhere
> between 1.4P and 1.4S. I can pop a source tree over onto Styx and rebuild
> if it would be useful, but my initial impression is that the problem
> probably isn't there. I'm way open to ideas. :)

I got something similar today, while attempting to mount '/home/cvs.netbsd.org'
from a -current to a solaris box.
Changing the export from /home/cvs.netbsd.org to /home -alldirs fixed
fixed things but it's only a workaroung :)
I didn't have the time to investigate.

--
Manuel Bouyer <bouyer@antioche.eu.org>
--