/dev/sda1 must be recovered on every boot

Why are you filing this bug against mountall? It's not mountall that's causing the filesystem to fail to unmount cleanly before reboot.

> I'm also wondering why it isn't telling me how many inodes were repaired (I
> remember to have seen this number on "real" recoveries a while ago).

Not in the kernel message. The messages you're citing are the kernel remounting it rw to do journal recovery after an unclean shutdown.

> Why are you filing this bug against mountall? It's not mountall that's causing the filesystem to fail to unmount cleanly before > reboot.

Which package should it be assigned then?

On Tue, Dec 17, 2013 at 07:30:17PM -0000, Sworddragon wrote:
> Which package should it be assigned then?

That depends on what's causing your disk to not be cleanly remounted
read-only on shutdown. Do you get any error messages on shutdown if booted
without 'splash'?

I could debug it down to a testcase. Just create the file /etc/init/test.conf with the following lines:

start on mounted and stopped mounted-tmp
exec mkdir -p /tmp/no-journal

After a reboot the recovery will happen. If you comment out the exec stanza all is fine again. But I don't know how such a simple upstrat script would cause this behavior.

I have figured out a little more: On removing "mounted and " on the start stanza the recovery doesn't appear anymore. But this still leaves the question why this should cause such a behavior.

Status changed to 'Confirmed' because the bug affects multiple users.

I have 14.04 LTS running as a virtual machine which I upgrade from 12.02 LTS and I do now see the error too on every boot.
Thus I think it is not the disk but some other cause which leads to this error.

In 14.10 amd64, mountall keeps running, causing init to keep /var/log/upstart/mountall.log open for writing, which prevents root from being remounted read-only.

These things are easy to examine by adding a /bin/sh line to /etc/init.d/reboot in do_stop() before "reboot -d -f -i". Then when you reboot you can switch to virtual console 1 and use lsof and other commands to explore what is going on. For example this is the line in lsof output showing the problem:
init 1 root 8w REG 8,5 613 1062101 /var/log/upstart/mountall.log

I added a
/usr/bin/killall mountall
line to /etc/init.d/umountroot before root is supposed to be remounted read-only. That seems to fix the problem. You need the full path because /usr/bin isn't in PATH, and don't use -9.