clone 635267 -1
reassign -1 sysvinit
severity -1 normal
found -1 2.88dsf-13.10
thanks

Hello Christian,
On Mon, Jul 25, 2011 at 06:26:07PM +0200, Christian Kastner wrote:
> On 07/25/2011 01:40 PM, Helge Kreutzmann wrote:
> > On Sun, Jul 24, 2011 at 11:12:41PM +0200, Christian Kastner wrote:
> >> One possibility would be that something went wrong during the recent
> >> /var/run -> /run transition (I see the above mounts under /run)
> > 
> > /run is a symlink to /var/run on that machine.
> 
> Well, that's the first weird thing -- AFAIK it should be the other way
> around. Looking at sysvinit's changelog, the above could be a result of
> the transition process, but I'm not sure.

Well looking at the changelog, the transition happend before my
install. (Which happend on July 9th, cf. #633306 for the details). So
somehow the install created the wrong symlinks and sysvinit did not
change them because it assumed the transition to be over, I assume?

Cloning to sysvinit to get the best course of action. Simply turn the
symlinks around?

> >> you can spot something fishy?
> > 
> > /etc/fstab:
> > /dev/mapper/system_vg-var_vg /var            ext3 nodev,nosuid,data=ordered 
> > 0       2
> > /dev/mapper/system_vg-aptcache_lv /var/cache/apt-cacher-ng
> > ext3    nodev,nosuid,noexec,data=ordered    0       2
> > 
> > Those tmpfs file systems are not mentioned.
> 
> From a brief look, it appears as though /etc/init.d/mountkernfs.sh
> automatically takes care of them.

That's where I grepped their existance, too.

> > syslog.1:Jul 24 18:31:11 sneo kernel: tmpfs: Bad mount option data
> > 
> > So somehow the tmpfs-entries inherit the mount options from /var?
> 
> Nope

I don't think so neither, but I'm totally lost where they come from. I
only specified them explicitly for the file systems in /etc/fstab.

> > I removed the "data=ordered" from /var and together with the following
> > changes, the "Bad mount option" error message is gone.
> 
> I'd put that back in. It's irrelevant to this case, but there was some
> controversy regarding data safety around it (although I really don't
> recall what it was)

I turned it on for syslog and the error remains absent. Probably
related to my other changes.

(Btw. I picked it up on lwn some time ago and added it as a safety
measure).

> > Next I tried to get tmpfs mounted on /var/run
> > 
> > I disabled the other two tmpfs entries in /etc/default/rcS, i.e.
> > /run/lock and /runshm.
> > 
> > I added /var/run explicitly into /etc/fstab
> > tmpfs   /var/run        tmpfs   nodev,nosuid,size=5242880,mode=755
> > 0      0
> > 
> > Now, after a reboot, tmpfs is supposedly mounted (i.e. mount|grep
> > tmpfs). However, the files in /var/run look like before. When I umount
> > /var/run, I get an error message that the file system is not mounted.
> > Now, in mount|grep tmpfs /var/run is no longer listed. When I mount
> > /var/run explicitly afterwards, /var/run is empty as expected (and
> > tmpfs mounted).
> > 
> > I'm clearly running out of ideas how to get /var/run mounted as tmpfs.
> > Actually, I probably don't understand which script mounts what and
> > when; ideally, /var/run would simply be zeroed at reboot...
> 
> Again, to me it looks like something went wrong during the /run
> transition. For example, this is an excerpt from my /etc/fstab:
> 
> # /var
> UUID=<foo> /var            jfs     defaults        0       2
> 
> # This mount for /run replaces the default configured in
> # /etc/default/tmpfs
> tmpfs /run tmpfs nosuid,noexec,size=1m,mode=755 0 0
> 
> # This mount for /run/lock replaces the default configured in
> # /etc/default/tmpfs
> tmpfs /run/lock tmpfs nodev,noexec,nosuid,size=1m,mode=1777 0 0

I don't have those comments and prior to adding it myself, /var/run
was not mentioned in /etc/fstab.

> I'm not sure I can help any further because I followed the transition
> only in passing, but perhaps you can find some hints in the changelog
> entries[0] for that transition?
> 
> Christian
> 
> 
> [0] see 2.88dsf-13.3, 2.88dsf-13.5
> http://packages.debian.org/changelogs/pool/main/s/sysvinit/current/changelog

I'll clone the bug, maybe some d-i/sysvinit interaction is broken.
Unless I hear otherwise, I'll try to manually switch the symlinks
around and see how it goes.

Thanks again *very much* for your help!

Greetings

             Helge


-- 
      Dr. Helge Kreutzmann                     deb...@helgefjell.de
           Dipl.-Phys.                   http://www.helgefjell.de/debian.php
        64bit GNU powered                     gpg signed mail preferred
           Help keep free software "libre": http://www.ffii.de/

Attachment: signature.asc
Description: Digital signature

Reply via email to