Re: [RFC][PATCH] CGroups: Add a per-subsystem hierarchy lock

["Paul Menage" <menage@google.com>]

On Fri, Jun 27, 2008 at 12:59 AM, Paul Menage <menage@google.com> wrote:
>
> This patch adds a hierarchy_mutex to the cgroup_subsys object that
> protects changes to the hierarchy observed by that subsystem. It is
> taken by the cgroup subsystem for the following operations:
>
> - linking a cgroup into that subsystem's cgroup tree
> - unlinking a cgroup from that subsystem's cgroup tree
> - moving the subsystem to/from a hierarchy (including across the
>  bind() callback)
>
> Thus if the subsystem holds its own hierarchy_mutex, it can safely
> traverse its ts own hierarchy.
>

It struck me that there's a small race in this code now that we're not
doing cgroup_lock() in the hotplug path.

- we start to attach a task T to cpuset C, with a single CPU "X" in
its "cpus" list
- cpuset_can_attach() returns successfully since the cpuset has a cpu
- CPU X gets hot-unplugged; any tasks in C are moved to their parent
cpuset and C loses its cpu.
- we update T->cgroups to point to C, which is broken since C has no cpus.

So we'll need some additional locking work on top of this - but I
think this patch is still a step in the right direction.

Paul