oneiric container problem: missing /run/lock
Bug #880030 reported by
xlyz
This bug affects 1 person
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
lxc (Ubuntu) |
Fix Released
|
Medium
|
Unassigned | ||
Oneiric |
Fix Released
|
Undecided
|
Unassigned |
Bug Description
host: oneiric with server minimal install + lxc + bridge_utils (bridge manually set up)
guest: oneiric
every time the container is started /run/lock is missing.
I then create it manually, and next time is gone again.
=======
SRU Justification:
1. Impact: /run/lock is missing when starting a container.
2. Development fix: create it at container startup
3. Stable fix: same as development fix
4. Test Case: start container, check for /run/lock
5. Regression potential: we simply create a directory at startup, so it's hard to imagine a regression. If there were one it should affect container startup.
=======
description: | updated |
tags: |
added: verification-done removed: verification-needed |
To post a comment you must log in.
Thanks for reporting this bug.
This is another side effect stemming from the fact that the kernel doesn't provide a good way to detect poweroff.vs.reboot in containers yet. The workaround is for the container parent to watch the container's /var/run/utmp. As a result of that, we can't let the container mount tmpfs on /var/run, because then the utmp used by the container would be hidden from the parent. As a result of that, /etc/init/ mounted- varrun. conf does not run, and that is where /var/run/lock is created.
This should be fixed in lxcguest.conf, which does the other ugly manual recreation of things we want in {/var}/run.