Hello,
I'm writing to report a problem with autofs in linux.  We have some machines
running Red Hat 7.1 and we have configured them to join one of our NIS
domains run by Solaris 2.6 servers.  Other Solaris 2.6 servers provide home
directories and other filesystems via NFSv2 and v3.  The problem is that
autofs (with default options) hangs when trying to auto-mount home
directories off of the Solaris NFS servers, but it works fine when
auto-mounting filesystems from linux NFS servers.

If I force autofs to use tcp mode (ie. localoptions='tcp' in
/etc/init.d/autofs) I can automount filesystems off of the Solaris servers
with no problem, but I can't automount filesystems from the linux NFS
servers anymore.  This makes sense because Linux nfsd does not support TCP
mode in NFSv3 yet. (any idea when that will happen?)

So basically I can have autofs working with Solaris, or working with Linux,
but not both at the same time.  I don't think it's a problem
with NFS itself, because I can mount filesystems manually without any
problem using tcp or udp (see transcript below).  There seems to be some
kind of problem with the way autofs is using mount option arguments.  I
noticed that when I added options to local automaps, all options were
cumulative... in other words, if I specify 'tcp' as a mount option in the
autofs script, and then I specify 'udp' in /etc/auto.master, autofs happily
uses both options even though they are mutually exclusive.  I tried this in
an attempt to get all the systems working with each other; since I need the
tcp option to be able to mount filesystems from Solaris, I thought I would
specify 'udp' on the linux automount entries to override the global options,
but because the options are cumulative it doesn't work.

Is there anything that can be done to get autofs working with both Solaris
and linux?  Are there any patches that I could apply?  Configuration changes
to be made?  At present I have implemented a work-around by mounting the
linux filesystems in /etc/fstab, but I still have to force autofs to use the
tcp option so that I can mount filesystems from Solaris NFS servers.  Since
the Solaris servers support NFSv3 in either tcp or udp mode I'm not sure why
this should be necessary, I only know that tcp mode is the only way for use
to use autofs with the Solaris servers.

Please write back soon if you have any suggestions.  I would be happy to
provide further information if that would help.  

Thanks in advance!
Aaron Ogden
Unix/Linux System Administrator

Note: the machine 'ray' mentioned below is a Solaris 2.6 NFS server.  As you
can see, no problems mounting filesystems from it using either transport
protocol.
[root@dw1 /root]# mkdir /mnt/tmp/tcp
[root@dw1 /root]# mkdir /mnt/tmp/udp
[root@dw1 /root]# mount -t nfs -o nfsvers=3,tcp ray:/fb1 /mnt/tmp/tcp
[root@dw1 /root]# mount -t nfs -o nfsvers=3,udp ray:/fb1 /mnt/tmp/udp
[root@dw1 /root]# grep ray /etc/mtab
ray:/fb1 /mnt/tmp/tcp nfs rw,nfsvers=3,tcp,addr=148.89.144.169 0 0
ray:/fb1 /mnt/tmp/udp nfs rw,nfsvers=3,udp,addr=148.89.144.169 0 0

Reply via email to