SEO

To all,

I am currently testing out whether I can get a Solaris 8 client
to function as a DHCP client in our network environment. I'm using a
Windoze 2000 DHCP server to get the IP address, but the DHCP client
provides the hostname. This appears to be working fine so far except
I'm seeing an annoying problem occur when I attempt to do a proper
shutdown or proper reboot of the client. It hangs up while running
the "/sbin/umountall" command that is in the /etc/rc6 RC script. I'll
get messages about "NFS server ... not responding" at this point and
it just hangs. I don't have this problem if I run /usr/sbin/reboot
since it bypasses a lot of the rc script stuff. The /etc/rc6 script
is linked to /etc/rc5 and /etc/rc0. I've placed a "truss -f
/sbin/umountall" to try and see what happens. It appears that the
truss output stops right after the following system calls are made:

execve("/etc/fs/autofs/umount", .....
sigaction(SIGHUP, ....
sigaction(SIGQUIT, .....
sigaction(SIGINT, ....
close(10)
close(12)
umount2("/apps/src", .... (sleeping ...)
umount2("/ahc/src", .... (sleeping ...)
umount2("/ahep/comm", .... (sleeping ...)
wait() (sleeping ...)
NFS server ..... not responding still trying
NFS server ..... not responding still trying


The above mounts are automount mounts. Its funny that on a non-DHCP
client I have, the /sbin/umountall process continues on without a
problem and the system either reboots or shuts down depending on the
run state I choose. The only difference between the systems is one is
a DHCP client and the other one is not. They both have NFS mounts in
both the /etc/vfstab and /etc/auto_direct which are the automounted
NFS mounts. I have come up with a work around to this in which I
place the following line in the rc6 script.

/sbin/umount -f /etc/mnttab

The above replaces the /sbin/umountall line originally in the file.
The man pages indicate the /etc/mnttab is really a mounted filesystem
and since the /sbin/umount -f command "forcably" unmounts a
filesystem, I figured this was the only way I could get around the
problem. Anyway, is it ok for me to do this or is there a "better"
solution to this problem? Any input would be greatly appreciated.
Thanks in advance.

Wayne Lee
e-mail: leewa2000@yahoo.com