Bug#677097: I know what the problem is, don't know the solution

On Thu, Jun 21, 2012 at 07:18:09PM +0100, Alan Chandler wrote:

I found the problem. Its in /etc/init.d/mountall.sh

This script calls the mount-functions.sh script in /lib/init

One of the things it does is attempt to mount /run as a tmpfs and
just after that the udev messages I already showed started happening

The key to this was in the boot log output where I have verbose set
to "YES" and the line it output was

Thu Jun 21 18:12:43 2012: [] Files under mount point '/run' will
be hidden.

Do you have an entry in /etc/fstab for /run?

The first three fields *must* be

tmpfs /run tmpfs

or it will think the filesystem type doesn't match, and do a new
mount rather than a remount.

I tried commenting out the actual mount of /run but it meant that
even in recovery mode I had no keyboard, and hotpluging it in and
out didn't restore it. I had to use a rescue disk to recover the
system. However when I did comment it out, these messages from
udevd stopped.

Are you using an initramfs? It should already be mounted if so.
And in this situation, it should just result in a remount with
the options from /etc/fstab. If you're not, it should mount it
in mountkernfs and then remount in mountall if there's an fstab
entry.

This is as documented in tmpfs(5).


Regards,
Roger

.''`. Roger Leigh
: :' : Debian GNU/Linux http://people.debian.org/~rleigh/
`. `' schroot and sbuild http://alioth.debian.org/projects/buildd-tools
`- GPG Public Key F33D 281D 470A B443 6756 147C 07B3 C8BC 4083 E800



To UNSUBSCRIBE, email to
with a subject of "unsubscribe". Trouble? Contact

On 21/06/12 20:27, Roger Leigh wrote:

On Thu, Jun 21, 2012 at 07:18:09PM +0100, Alan Chandler wrote:

I found the problem. Its in /etc/init.d/mountall.sh

This script calls the mount-functions.sh script in /lib/init

One of the things it does is attempt to mount /run as a tmpfs and
just after that the udev messages I already showed started happening

The key to this was in the boot log output where I have verbose set
to "YES" and the line it output was

Thu Jun 21 18:12:43 2012: [] Files under mount point '/run' will
be hidden.

Do you have an entry in /etc/fstab for /run?

The first three fields *must* be

tmpfs /run tmpfs

or it will think the filesystem type doesn't match, and do a new
mount rather than a remount.

I don't have an entry in /etc/fstab.

How was it supposed to get there? I have had the same /etc/fstab as
with installation - plus the manually editing by me last february to add
the lvm volumes reported in the boot log.

I also notice there is an /etc/fstab.d directory, but that is empty

any way I will try adding it and rebooting after I have sent this e-mail
to see if that sorts me out.

I tried commenting out the actual mount of /run but it meant that
even in recovery mode I had no keyboard, and hotpluging it in and
out didn't restore it. I had to use a rescue disk to recover the
system. However when I did comment it out, these messages from
udevd stopped.

Are you using an initramfs? It should already be mounted if so.
And in this situation, it should just result in a remount with
the options from /etc/fstab. If you're not, it should mount it
in mountkernfs and then remount in mountall if there's an fstab
entry.

I have an initramfs produced as standard by what ever mechanism debian
uses. I haven't touched it.

This is as documented in tmpfs(5).

My /etc/default/tmpfs was last modified (not be me) on Jun 8 and
contains only comments.

Alan Chandler
http://www.chandlerfamily.org.uk




To UNSUBSCRIBE, email to
with a subject of "unsubscribe". Trouble? Contact

On 21/06/12 21:55, Alan Chandler wrote:

On 21/06/12 20:27, Roger Leigh wrote:

be hidden.
Do you have an entry in /etc/fstab for /run?

The first three fields *must* be

tmpfs /run tmpfs

or it will think the filesystem type doesn't match, and do a new
mount rather than a remount.

I don't have an entry in /etc/fstab.

...

any way I will try adding it and rebooting after I have sent this
e-mail to see if that sorts me out.

Thanks - it did.

I had to also delete the symlink and find a way to shutdown without a
/run, but after I restarted, it worked great and has resolved the issue

Alan Chandler
http://www.chandlerfamily.org.uk




To UNSUBSCRIBE, email to
with a subject of "unsubscribe". Trouble? Contact

On Thu, Jun 21, 2012 at 10:08:16PM +0100, Alan Chandler wrote:

On 21/06/12 21:55, Alan Chandler wrote:
>On 21/06/12 20:27, Roger Leigh wrote:
>>be hidden.
>>Do you have an entry in /etc/fstab for /run?
>>
>>The first three fields *must* be
>>
>>tmpfs /run tmpfs
>>
>>or it will think the filesystem type doesn't match, and do a new
>>mount rather than a remount.
>
>I don't have an entry in /etc/fstab.
>
>...
>
>any way I will try adding it and rebooting after I have sent this
>e-mail to see if that sorts me out.


Thanks - it did.

I had to also delete the symlink and find a way to shutdown without
a /run, but after I restarted, it worked great and has resolved the
issue

I was not suggesting that you add an entry to /etc/fstab. You
don't need one unless you want to override the defaults. I was
just asking if you had one in case it was added incorrectly,
which would have been one way you could confuse things at boot.

So the reason why this is causing problems is still unexplained.
Could you please remove the entry, reboot and then let me know
what the contents of /proc/mounts are? Are there two entries
for /run, if one is being masked by a second mount?


Thanks,
Roger

.''`. Roger Leigh
: :' : Debian GNU/Linux http://people.debian.org/~rleigh/
`. `' schroot and sbuild http://alioth.debian.org/projects/buildd-tools
`- GPG Public Key F33D 281D 470A B443 6756 147C 07B3 C8BC 4083 E800



To UNSUBSCRIBE, email to
with a subject of "unsubscribe". Trouble? Contact

On 21/06/12 23:10, Roger Leigh wrote:

On Thu, Jun 21, 2012 at 10:08:16PM +0100, Alan Chandler wrote:

On 21/06/12 21:55, Alan Chandler wrote:

On 21/06/12 20:27, Roger Leigh wrote:

be hidden.
Do you have an entry in /etc/fstab for /run?

The first three fields *must* be

tmpfs /run tmpfs

or it will think the filesystem type doesn't match, and do a new
mount rather than a remount.

I don't have an entry in /etc/fstab.

...

any way I will try adding it and rebooting after I have sent this
e-mail to see if that sorts me out.

Thanks - it did.

I had to also delete the symlink and find a way to shutdown without
a /run, but after I restarted, it worked great and has resolved the
issue

I was not suggesting that you add an entry to /etc/fstab. You
don't need one unless you want to override the defaults. I was
just asking if you had one in case it was added incorrectly,
which would have been one way you could confuse things at boot.

So the reason why this is causing problems is still unexplained.
Could you please remove the entry, reboot and then let me know
what the contents of /proc/mounts are? Are there two entries
for /run, if one is being masked by a second mount?

I have removed the entry in /etc/fstab and I am still loading perfectly.

The other thing that is left, is the fact that I had a symlink from /run
to /var/run (which I didn't put there).

I conducted a little experiment, and with a recovery disk I altered my
system as follows.

1) removed the symlink in /var/run -> /run
2) save the contents of /run (I moved it to /oldrun
3) made a new /var/run which contained a copy of the contents of /oldrun
4) symlinked /run -> /var/run

(I think this is as close to how it was before - the only issue I never
checked before was whether /var/run had anything in it before booting)

When I booted like this, I had the exact same symptoms as before.

I checked /proc/mounts in this state, and there were two copies of a
tmpfs mounted at /var/run. (and two copies of /var/run/lock and
/var/run/shm)

I've put it all back now (not the /etc/fstab entry - that is still
commented out) and rebooted - and I am back working again.

SO I think the only issue is really how did that symlink get there in
the first place. Was there some intermediate stage when this migration
to /run began that meant it got put there.

I have been running debian sid for years - on this hardware for about a
year. I use aptitude and do an update every several days, although
occasionally I might let it go for a couple of weeks if I am busy (which
I have been recently). I use a standard configuration (I don't try to
change anything) except that I am running VirtualBox with a Windows 7
environment within it and that does add its own modules.







Alan Chandler
http://www.chandlerfamily.org.uk




To UNSUBSCRIBE, email to
with a subject of "unsubscribe". Trouble? Contact