
On 2012-03-06 19:34, David Jablonski wrote:
We would like to mount /dev and /dev/pts correctly using the libvirt xml config file instead of doing a bind mount in a container init script we use.
You will automatically get a private /dev/pts, without any additional configuration. /dev will also get created, but only a very minimal version, and you cannot use udev inside LXC. So any devices not in the minimal /dev that need to be accessible will need to be created using mknod in your script. If you get a permission denied error from mknod, the device permissions for the cgroup for the container need to be modified. libvirt is supposed to set up such permissions for you if you specify you want a device passed through, but I have not been able to get this to work with containers. I could not get the libcgroup-tools service to run on bootup either. So I have been manipulating the cgroup permissions manually in /proc, which is not exactly ideal. -- Robin