Bug 26298

Summary: FreeBSD 4.3-RC #0 amd fails to automount nfs filesystems
Product: Base System Reporter: galen.menzel <galen.menzel>
Component: kernAssignee: freebsd-bugs (Nobody) <bugs>
Status: Closed FIXED    
Severity: Affects Only Me    
Priority: Normal    
Version: Unspecified   
Hardware: Any   
OS: Any   

Description galen.menzel 2001-04-02 20:10:00 UTC
I am using amd to automount two nfs filesystems, both living on the same
Network Appliance filer.  I am using /etc/amd/source as an amd config
file.  The contents of /etc/amd/source are as follows:

public  type:=nfs;opts:=rw,hard,nosuid;rhost:=<servername>;\
        rfs:=/export/public/source
private type:=nfs;opts:=rw,hard,nosuid;rhost:=<servername>;\
        rfs:=/export/unix-source

with <servername> replaced with the net app filer's service name.  The
following lines are in /etc/rc.conf:

amd_enable="YES"
amd_flags="/usr/share/src /etc/amd/source"

and 'ps auxw | grep amd' yeilds:

amd -p /usr/share/src /etc/amd/source

This configuration works with the FreeBSD 4.2 kernel.  However, with the
current 4.x-stable kernel, as of 2 April 2001, if I try to access
/usr/share/src/public or /usr/share/src/private (via ls, cd, or
whatever), the access hangs for maybe a minute, and then quits with an
"operation timed out" error. After the timeout error has occured,
attempts to access the directory fail with the timeour error
immediately.

Before attempting to access the automount directories, df works fine.
If access is attempted, but canceled before the timeout error is
received, further attempts to access the automount directories,
including df, hang until the error is received.  After the error is
received, df works and shows that the automount filesystems are not
mounted.

I am able to mount both of these filesystems by hand as root.

How-To-Repeat: Start amd 'amd -p /usr/share/src /etc/amd/source' and attempt to access
the automount filesystems.
Comment 1 dd freebsd_committer freebsd_triage 2001-04-04 01:43:16 UTC
State Changed
From-To: open->closed

Originator reports that the problem was likely caused by having the 
kernel and userland out of sync, which is definitely a no-no.