[OpenAFS] Problems with fsck on Solaris 9

[Brian Sebby]

I realized that some of this came up back in August, so I apologize if I'm 
going over old ground.  But after searching through the archives, I'm not
entirely sure if this has ever been fixed.

I have a machine that I upgraded from Solaris 8 to Solaris 9.  While we
already have a Solaris 9 AFS server, it's using an older version of the
operating system.  From what I can tell, Sun made a change to the UFS and
fsck systems that makes the AFS fsck break.  At first I thought this was
contained in patch 113073-13 (newest version is -14), and I thought we had
avoided the problem it caused by not installing it on our older Solaris 9
server.  However, when I installed the new server with Solaris 9, I first
used Solaris 9 09/04, which has this patch preinstalled.  After setting
everything up and trying to mount the file systems, I see this error:

# fsck /vicepa
----Open AFS (R) openafs 1.2.11 fsck----
** /dev/rdsk/c0t9d0s0

CANNOT READ: BLK 0
CONTINUE? [yn] y

THE FOLLOWING DISK SECTORS COULD NOT BE READ:
** Last Mounted on
** Phase 1 - Check Blocks and Sizes
** Phase 2 - Check Pathnames
** Phase 3 - Check Connectivity
** Phase 4 - Check Reference Counts
** Phase 5 - Check Cyl groups
FREE BLK COUNT(S) WRONG IN CYL GROUP (SUPERBLK)
SALVAGE? [yn] y

2 files, 9 used, 17413241 free (9 frags, 2176654 blocks, 0.0% fragmentation)

***** FILE SYSTEM WAS MODIFIED *****

This repeated every time I ran fsck.  Thinking that it was because I had
the UFS patch installed, I rebuilt the system from an older version of
Solaris 9 (version 08/03) that did not have that patch installed.  However,
I am now seeing the same behavior, so I'm thinking this could be due to the
kernel patch.  Our older AFS server is at revision 112233-11, while the new
system came preinstalled with 112233-12.  Does anyone know if that is the
source?

It looks like there's been some work on trying to fix this.  Doug Engert
tried putting in some of the code for Solaris 10 into a development vfsck,
but it didn't work either.  Does anyone have a working binary for fsck, or
know what I need to do to solve this problem?  For reference, this first
came up in August, where it was discussed that Solaris 9 had changed some
fields in the UFS file system.  I couldn't find any list postings indicating
a solution.  If this is something that's been fixed in the development code,
please let me know.


Thanks,

Brian Sebby

-- 
Brian Sebby  (sebby@anl.gov)  |  Distributed Computing Administration
Phone: +1 630.252.9935        |  Computing and Instrumentation Solutions
Fax:   +1 630.252.4601        |  Argonne National Laboratory