Re: [patch 0/2] cpusets, cpu_cgroup: disallow attaching kthreadd

[Peter Zijlstra <a.p.zijlstra@chello.nl>]

On Fri, 2012-04-06 at 08:52 -0700, Tejun Heo wrote:
> Hello, David.
> 
> On Thu, Apr 05, 2012 at 04:40:06PM -0700, David Rientjes wrote:
> > Well, I'm fussing over it because the patch being considered unnecessary 
> > requires that kthreadd can't be moved anywhere and I know one company is 
> > trying to move in a direction where nothing is in the root memcg.
> 
> "Nothing in the root memcg" can't be a goal in and of itself.  You
> want that to achieve some functional goal and I'm trying to say this
> specific kthreadd change doesn't necessarily affect the goal - better
> accounting - all that much.  If root group is gonna be completely
> empty otherwise, just combine information from it.  Even if that
> doesn't work, assigning specific kthreads to appropriate cgroups after
> the creation wouldn't be too far off.  I just don't see how relevant
> it actually would be.
> 
> If we want all controlles to play by the same rules, which is
> necessary for having a unified hierarchy, I wanna keep those rules
> simple.  If bound kthreads in !root cgroups cause issues for some and
> there aren't quite strong reasons to do otherwise, I would just
> restrict them in the root.  It's not like those kthreads are
> cgroup-aware in any form anyway.
> 
> I don't know.  Just proceed without kthreadd in the root.  If the
> fallouts are big enough and can't be easily worked around, let's talk
> then.

Furthermore, the whole point of kthreadd's existence is so that we could
create kthreads without context. Placing it in a cgroup will ensure all
subsequently created kthreads do have context (including possible idle
threads). This seems like a particularly bad idea.