Mysterious Unmounts

Charles R. Dennett dennett at rochester.rr.com
Sun Apr 30 12:39:25 UTC 2006


I'm running FC5 and earlier this week one of the two hard drives in my
system died.  It was the disk with FC5 on it.  I bought a new disk,
installed it and reinstalled FC5. (Previously, I had done upgrade
installations starting with RH9->FC3->FC4->FC5 so this was really the
first time I had done a clean FC5 install.)  I then did a "yum update".

I logged in as my normal non-root user and su'ed to root.  As I was
logged in as root and pulling various files from my backups to get my
web server, mail server and other tools running again I happened to
notice that some disk partitions were no longer mounted.  The disk I
replaced contains partitions mounted on /, /boot, /var, /space1 and a 1
GB swap partition.  /boot and /space1 were missing.  I remounted them.
Some time later I noticed the same thing.  I remounted them.  This kept
happening.  Sometimes the /home partition from the second disk would be
missing.  This seemed to be happening whenever root logged off or I
exited the kyum (a GUI for yum) application.  I was using that to add
additional packages I needed. The / and /var partitions never
disappeared.  It looked like any partition with no open files were being
unmounted.

Here's what I did to try to figure out what was happening.  I renamed
/bin/umount to /bin/umount-real. I then wrote a quick script for
/bin/umount that would append to a file the time and date and the output
from "ps -ef".  Then it would call /bin/umount-real with whatever
arguments had been passed to it.  I forced the problem to happen again
and then looked at the file my script had written.  I caught a umount
from the ps output.  Here are the entries tracking parent and child PIDs
 back to the hald daemon:


68        1850     1  0 Apr28 ?        00:00:03 hald
root      1851  1850  0 Apr28 ?        00:00:00 hald-runner
root     23576  1851  0 22:33 ?        00:00:00 /bin/bash
/usr/share/hal/scripts/hal-system-storage-unmount
root     23577 23576  0 22:33 ?        00:00:00 /bin/bash
/usr/share/hal/scripts/hal-system-storage-unmount
root     23578 23577  0 22:33 ?        00:00:00 /bin/sh /bin/umount /home
root     23580 23578  0 22:33 ?        00:00:00 ps -ef

(The 68 as the UID for the first line is because the username is 9
characters long - haldaemon.  Apparently that's a known problem with ps
and ls when the username is >8 characters.)

I tried googling for this and looking through the archives of this list
but did not find anything (yet).  Does anyone know what is happening
here and how to fix it?  These filesystems are mounted at boot time.
Why is hal trying to unmount them?  They are not removable media.

I'm sure there is more information needed that I have not supplied so
just ask and I'll respond.  If this is a known problem with a known fix,
just point me in the right direction.  If something I've said above is
not clear, let me know and I'll clarify.

Thanks for any help.

Charlie Dennett




More information about the fedora-list mailing list