SUMMARY: Problems with NFS automount of Linux filesystem under Solaris 8.

From: Brian Chase <>
Date: Tue Apr 08 2003 - 16:21:20 EDT
On Tue, 8 Apr 2003, Brian Chase wrote:

> I've an odd problem on a Solaris 8 client of mine.  Its unable to
> automount an NFS filesystem served off of a Linux system.  The same
> system doesn't have problems automounting any of the other filesystems
> being served off of our SGI and NetApp file servers.  More puzzling
> still, I'm able to manually NFS mount the filesystem with the `mount'
> command.
> The automount master file entry looks like this:
>   /net /etc/ -intr
> The relevant /etc/ file entry looks like this:
>   vol606	-retry=5000	linuxnfs:/mnt/vol606
> This should automount the Linux server's filesystem under /net/vol606 on
> the Solaris client system.  When I try to `cd /net/vol606', the command
> hangs for a very long time.  The one error message I get is the
> following:
>   Apr  8 12:00:58 client1 nfs: [ID 333984 kern.notice] NFS server linuxnfs
>     not responding still trying
> However, if I run a `mount linuxfs:/mnt/vol606 /mnt2' on the same
> Solaris client, that command succeeds and I can access the contents of
> the filesystem just fine.
> Any ideas?

Thanks go to Hendrik Visage, Wolf Schaefer and Roger Leonard for
pointing me in the correct direction on this one.  The Linux NFS server
seems to be falsely advertising the NFS protocols it supports:

  # rpcinfo -p madmax | grep nfs
      100003    2   udp   2049  nfs
      100003    3   udp   2049  nfs
      100003    2   tcp   2049  nfs
      100003    3   tcp   2049  nfs

I think the automounter tries using NFSv3 tcp and I guess the manual
mount command is smart enough to fall back on (or it defaults to) one
of the udp versions.

I found that if I tried to manually mount the filesystem with either
version of NFS with `-o proto=tcp', the Linux server was unable to
handle the request.  I added a `proto=udp' option to the relevant entry
in my file, and now it works.

In hindsight, I should've realized it was something like this.   I used
to have a very similar problem with using Linux as an NFS server for
diskless NetBSD/vax clients a number of years ago.  It hadn't occurred
to me that Linux would /still/ be broken in this way.

Thanks for all the quick replies.

sunmanagers mailing list
Received on Tue Apr 8 16:34:24 2003

This archive was generated by hypermail 2.1.8 : Thu Mar 03 2016 - 06:43:08 EST