Re: RPCPROG_MNT: RPC: Timed out / receiving NFS error when trying to mount NFS file system after make world

From: Rick Macklem <rmacklem_at_uoguelph.ca>
Date: Fri, 5 Jun 2009 11:17:10 -0400 (EDT)
On Thu, 4 Jun 2009, Dmitry Marakasov wrote:

> * O. Hartmann (ohartman_at_zedat.fu-berlin.de) wrote:
>
> Just my 2 cents: the same behaviour, and while mount -o tcp succeeds,
> umounting behaves strangely: if I umount nfs filesystem, umount
> process hangs for some time, then exits with `umount: hive:
> RPCMNT_UMOUNT: RPC: Timed out', while the filesystem is actually
> umounted immediately after umount is called.  This all also breaks
> amd severely (however it already haven't been working on current
> for a long time for me).
>
That would be consistent with the problem being udp specific. An NFS
server for v2,3 (NFSv4 is a different story) is "stateless" and doesn't
know what mounted on it. All the "umount" does is tell the server's
mountd daemon it has unmounted, so the server's mountd can maintain a
table of mounts that can be queried via "showmount". The -current
umount.c always does this via "udp", so it would fail. The only effect
is that "showmount" might reply bogus info.

It seems that the problem is specific to "udp" and some arches. (I can't
reproduce it on i386 and at least one person sees it on amd64.)

Maybe people who see the problem can post to the -current list, noting
what arch they are using and network config (running virtualized, what
net hardware, multiple net interfaces, etc).

Hopefully there is some commonality among them?

Thanks in advance for any help tracking this down, rick
Received on Fri Jun 05 2009 - 13:15:54 UTC

This archive was generated by hypermail 2.4.0 : Wed May 19 2021 - 11:39:49 UTC