[128] in Info-AFS_Redistribution
Local files corrupted under AIX 3.1 and SunOS 4.0.3
daemon@ATHENA.MIT.EDU (Keith Gorlen)
Tue May 28 18:36:20 1991
Date: Tue, 28 May 1991 17:33:22 -0400 (EDT)
From: Keith Gorlen <kgorlen+@alw.nih.gov>
To: Info-AFS@transarc.com
Cc: jane small <jane@alw.nih.gov>, sandy orlow <sandy@alw.nih.gov>,
We've had three incidents at NIH where files on the local disks of
RS/6000 clients running AIX 3.1 have been corrupted, and one incident
where files were corrupted on the local disk of a SunOS 4.0.3 client.
In all of these incidents but one, AFS 3.1 was installed. We're
wondering if some of these problems may be AFS related, and if others
have had similar experiences.
Incident report #1:
The problem I had with the RS6000 was as follows:
1) I logged in as root;
2) I had not started the window manager yet;
3) I was probing around the directories when I typed "X", being in the
directory that "X" is in (I don't remember off hand-- maybe it was
/usr/lib/X11);
4) the X background came up, but nothing else;
5) when I killed X, I noticed that system files such as "xterm" and various
other X executables were missing.
6) when I did a subsequent xwin, nothing worked. X windows had to be
re-installed from the distribution kit.
AFS was *not* yet installed on this system.
Incident report #2:
We had just gotten farside (an RS/6000 with AIX 3.1) installed for the
first time, and I'd gotten AFS to run on it. I then copied over the
rc.afs file, as well as the other "boot" files which you'd prepared on
arsenic (another RS/6000 with AIX 3.1 that had been successfully running
AFS 3.1). At this point when I rebooted, the number 553 came up in the
LCD. It never booted up beyond this point. It was then that we found
out that /bin/sh had been corrupted. The checksums differed, although
the lengths of both the corrupted and actual files were equal. When
trying to execute the corrupted file, the message is:
Could not load program ./sh.orig
./sh.orig is not executable or not in correct XCOFF format
Error was: Exec format error
Checksums reveal:
40240 251 ./sh.orig
38917 251 /bin/sh
Incident report #3:
After our link to the AFS cell went down over night, my system (an
RS/6000 running AIX 3.1) began acting strangely so I rebooted. When the
system came back up, I got a 551 error on the display (Can't load
software). When I looked to see what the problem was, inode 2 on / was
gone. After an fsck, it was semi-recreated, with lost+found put in it,
but no "." and "..". The rest of the filesystem was linked in
underneath. I recreated the / directory, with fsdb and everything was
fine.
Incident report #4:
My sparcstation (larkspur) was rebooted last Friday morning and changed
over (via package) to AFS3.1. Since then I get the following error
message with a number of commands:
ld.so: __DYNAMIC version 0 not supported
In particular, this happens with afs commands such as 'fs'. It is not
just a warning. The command does not work at all. I also see several
of these messages when I login; I assume this is related to commands in
my startup files.
Upon investigation, we discovered that several executables in the
/usr/afs/bin directory, including fs, had been corrupted, although the
lengths and modification times of the files were correct. One
executable contained a user's .history file. /usr/afs/bin resides on
the local disk, and had been updated to AFS 3.1 by a package
configuration file that had worked correctly on many prior occasions.
-----
Keith Gorlen phone: (301) 496-1111
Building 12A, Room 2033 FAX: (301) 402-0007
National Institutes of Health uucp: uunet!kgorlen%alw.nih.gov
Bethesda, MD 20892 Internet: kgorlen@alw.nih.gov