Ok, sorry, I see now. Yes i can reproduce this with getdents of course.
This has nothing to do with lxc lying, and yes if you need to work
around this you'll need to use blockdev-backed rootfs for the
containers.
The container is set up with the tasks pivot_root'd into the container's
/, which in your case is a bind mount. Resolution of '..' respects the
fact that the /var/lib/lxc/container/rootfs should be root, and returns
itself. However, there is only one actual directory entry for
/var/lib/lxc/container/rootfs in all the kernel, and it's '..'
(d_parent) points to /var/lib/lxc/container. Lxc can't do anything
about that. The kernel's flexible mounts, bind mounts, and mounts
namespaces provide different *views* of the filesystems, don't change
what is in the actual filesystems.
** Changed in: lxc (Ubuntu)
Status: New => Won't Fix
--
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/1094019
Title:
Linux's getdents gets /.. inode number from host system under LXC
To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu/+source/lxc/+bug/1094019/+subscriptions
--
ubuntu-bugs mailing list
[email protected]
https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs
