Page 1of2
                                                 January 22, 1991

This document is provided AS IS.  IBM M&S Defect Support cannot guarantee
that this will work nor can we assist the customer with it's contents.

SUBJECT: LED 552 Workaround For Users Who Did Not Reboot After CorSvc 3003
         Was Applied.
--------------------------------------------------------------------------

Problem: In an effort to assist customers who have encountered the above 
         problem after forgetting to reboot their RISC System/6000 and
         instead, immediately applied PTF #U401273 (commonly referred to
         as Additional Updates 3003), we have come up with a possible
         workaround.

Requirements: This workaround requires that you copy some files from 
         another machine.  If you do not have another machine at
         CorSvc 3003 (bos.obj 03.01.0003.0013+) then you will
         not be able to do this.  If such is the case, then we
         can only suggest that you re-install your system.

Problem History: Customers who participated in early releases of CorSvc 3003
         and failed to reboot their systems as instructed are the
         only ones expected to report this problem.  We have
         painstakingly modified the documentation to make the 
         future possibility of reoccurrance of the problem highly
         unlikely.


BACKGROUND
Since the new lvm commands will not work until after a reboot, the update
attempts to leave the old commands and library in place until a reboot of
the machine.  During the reboot, the new files are moved to the right place.
To accomplish this, the update script first saves off the old lvm files, then
it restores all of the new files, and copies the lvm files to a directory
call /etc/.lvm.hold and compresses the files.  The old files are then moved
back to their original location.  Note that until a system reboot, the only
place that the new files exist on the system is in the /etc/.lvm.hold
directory.

THE PROBLEM
During the apply of PTF #U401273, the update script follows the procedure 
described above with one exception: the new lvm files are not on the update.
When the update script thinks it is copying the new files into the 
/etc/.lvm.hold directory, it is actually overwriting the 3003 versions with 
the original files.  This causes the boot image (created by bosboot) to use 
the old liblvm.a, and that is why the system won't boot.  At this point the 
new version of the files are no longer present anywhere on the system.


                                                 Page 2of2

This document is provided AS IS.  IBM M&S Defect Support cannot guarantee
that this will work nor can we assist the customer with it's contents.

SUBJECT: LED 552 Workaround For Users Who Did Not Reboot After CorSvc 3003
         Was Applied.
--------------------------------------------------------------------------

THE SOLUTION
The contents of the /etc/.lvm.hold directory must be restored to the proper
state.  This is only possible by getting these files from another machine
or the update media.  The easiest way to get them is from a machine that has
had the 3003 update applied and has not been rebooted.  If such a machine is
available, one could simply back up the files and restore them on the broken
machine.  To back up these files from the good machine do the following:

         cd /etc/.lvm.hold
         find . -print | backup -iv

If no machine is available in the above state, you can generate the needed
files from a machine that is at or above CorSvc 3003.  Do the following:

         Create a file called lvm_cmds that contains:

 /etc/chlv
 /etc/extendlv
 /etc/lextendlv
 /etc/lquerylv
 /etc/lquerypv
 /lib/liblvm.a
 /lib/libsm.a

         There should be no blank lines or spaces in the file.

mkdir /tmp/lvm
cd /tmp/lvm

         Put the lvm_cmds file in this directory.

for i in `cat lvm_cmds` ; do
cp $i .
compress `basename $i`
done

cp /lib/liblvm.a liblvm.a.strip
cp /usr/bin/blv.proto .

find . -print | backup -ivq

Once we get a diskette containing the right files, this is how we fix our
broken machine:

         Boot from maintenance diskettes.
         Run /etc/continue hdisk0 # or whatever the hdisk should be
         cd /etc/.lvm.hold
         restore -x # with the diskette we created above
         bosboot -a

You should now be able to shutdown and reboot.