Re: [PATCH] cgroup: Don't drop the cgroup_mutex in cgroup_rmdir

[Tejun Heo <htejun-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>]

On Thu, Jul 19, 2012 at 07:39:32PM +0530, Aneesh Kumar K.V wrote:
> From: "Aneesh Kumar K.V" <aneesh.kumar-23VcF4HTsmIX0ybBhKVfKdBPR1lH4CV8@public.gmane.org>
> 
> We dropped cgroup mutex, because of a deadlock between memcg and cpuset.
> cpuset took hotplug lock followed by cgroup_mutex, where as memcg pre_destroy
> did lru_add_drain_all() which took hotplug lock while already holding
> cgroup_mutex. The deadlock is explained in 3fa59dfbc3b223f02c26593be69ce6fc9a940405
> But dropping cgroup_mutex in cgroup_rmdir also means tasks could get
> added to cgroup while we are in pre_destroy. This makes error handling in
> pre_destroy complex. So move the unlock/lock to memcg pre_destroy callback.
> Core cgroup will now call pre_destroy with cgroup_mutex held.
> 
> Signed-off-by: Aneesh Kumar K.V <aneesh.kumar-23VcF4HTsmIX0ybBhKVfKdBPR1lH4CV8@public.gmane.org>

So, umm, let's not do this at this point.  Please just fix memcg such
that it doesn't fail ->pre_destroy() and drop
subsys->__DEPRECATED_clear_css_refs.  cgroup core won't give away new
references during or after pre_destroy that way and memcg is the ONLY
subsystem needing the deprecated behavior so it's rather
counter-productive to implement work-around at this point.

 Nacked-And-Please-Drop-The-DEPRECATED-Behavior-by: Tejun Heo <tj-DgEjT+Ai2ygdnm+yROfE0A@public.gmane.org>

Thanks.

-- 
tejun