Michael Naumann <mnaumann@giga-stream.de> [2002-12-07 13:01:24 +0100]:
> 06.12.2002 19:25:57, "nate" <debian-user@aphroland.org> wrote:
> >check /etc/mtab see what it says. thats very unusal. anything
> >odd happen before this started happening? crash or anything?
>
> In /etc/mtab it isn't mentioned either.
Then the problem lies there. When the system mounts drives it should
be updating the contents of /etc/mtab. That is a system private file
that the system uses to communicate this information to other parts of
the system, such as 'df'.
> And yes, there was something odd, I forgot to mention.
> It came back to my memory since I now have a second system with
> the very same symptoms.
> Both systems are installed using Knoppix (/usr/local/bin/knx-hdinstall).
Oh, good. Then it is probably just some simple problem with the start
up scripts that mount the disks there. Once you find the right fix it
will cleanly solve the problem.
> What's most strange to me, is that the systems work quite well
> despite this fact.
It is not strange at all. The filesystem in the kernel knows that the
filesystem is mounted. But outside the kernel system programs use
that file to figure out what is mounted. In the modern age it would
probably make sense to use /proc/mounts instead. But prior to /proc
existing one could only use real files on disk for that purpose.
I have often seen times when a disk is completely full and therefore
nothing could be written to mnttab, utmp, etc. and behavior such as
you are seeing would result. Usually at that time the file would be
zero length. But free up some disk space to allow the system to write
the file again and everything would return to normal.
I don't have a KNOPPIX based system to look at this problem. But I
would start looking at /etc/init.d/checkroot.sh and others and make
sure that after the mount command there the system shows up in the
mtab. Perhaps at the time of the command KNOPPIX is used to working
off of a readonly device and has disabled that feature at that moment.
Just guessing. This part of checkroot.sh seems most interesting in
your case.
# If the root filesystem was not marked as read-only in
# /etc/fstab,
# remount the rootfs rw but do not try to change mtab because it
# is on a ro fs until the remount succeeded. Then clean up old
# mtabs
# and finally write the new mtab.
#
mount -n -o remount,$rootmode /
if [ "$rootmode" = rw ]
then
rm -f /etc/mtab~ /etc/nologin
: > /etc/mtab
mount -f -o remount /
mount -f /proc
[ "$devfs" ] && grep -q '^devfs /dev' /proc/mounts && mount -f
"$devfs"
fi
Bob
Attachment:
pgppt9lxqe5Gs.pgp
Description: PGP signature