[PATCH cgroup/for-3.11] cgroup: disallow rename(2) if sane_behavior

[Tejun Heo]

Hello, Aristeu.

On Mon, Jun 17, 2013 at 09:00:29AM -0400, Aristeu Rozanski wrote:
> sure, I'm on it. that sounds like one of the versions of the hierarchy
> patchset - having the local configuration saved but only applying what's
> possible based on parent's state.

Yes, basically, any configuration should be allowed but its
application would be confined by its ancestors and a parent changing
configuration should propagate the effective limits down the hierarchy
but shouldn't affect the configurations directly - ie. if an ancestor
adds a device to the list of disallowed devices and then removes it
back, the whole subtree should return to the original state.

As we're introducing an incompatible interface with
cgroup_sane_behavior() anyway, if necessary, devcg can also change
behaviors dependent on the flag.  Not encouraged but given the breadth
of the changes we're making with the flag, I don't think devcg would
stand out too much by doing so.

Thanks.

-- 
tejun