Description of problem:
When recreating bug #190756, I kept getting kernel panics with
or without my patch to fix the problem.

Version-Release number of selected component (if applicable):
RHEL4 U4

How reproducible:
Happens 25% - 50% of my attempts to recreate 190756.

Steps to Reproduce:
1. Export a GFS file system on a 3-node cluster running RHEL4 U4
   In my case, roth-01, roth-02 and roth-03.
2. Create a virtual IP xx.xx.xx.250 on the servers, defined to
   run on two or more of the NFS/GFS servers.
3. NFS mount the file system through the virtual IP on five
   NFS clients.  In my case, trin-12,13,14,15 and 16.
4. Run the genesis QE test case to create a directory for each
   client, and multiple files inside each.
5. While genesis is running, use the following script to move the
   virtual IP from node to node:

#!/bin/bash
for ((a=1; a>0; a++))
do
        clusvcadm -r xx.xx.xx.250 -m roth-02
        sleep 20
        clusvcadm -r xx.xx.xx.250 -m roth-01
        sleep 20
done
exit 0

Actual results:
Around 25 to 50% of the runs, the kernel will panic.

Expected results:
No kernel panic.

Additional info:
See call trace in the attachment for more information.

Also see bz #190756 for additional information.

I realize this is an awkward setup, so please contact me if you
want me to recreate this.  These are only my test clusters, so I can
put debug code on easily and let you have the machines to recreate
this.  Also, let me know if I can help in any way.