Re: [PATCH v12 8/8] cgroup: implement the PIDs subsystem

From: Thomas Gleixner <tglx@linutronix.de>
To: Aleksa Sarai <cyphar@cyphar.com>
Cc: "Peter Zijlstra" <peterz@infradead.org>,
	"Tejun Heo" <tj@kernel.org>,
	lizefan@huawei.com, mingo@redhat.com, richard@nod.at,
	"Frédéric Weisbecker" <fweisbec@gmail.com>,
	linux-kernel@vger.kernel.org, cgroups@vger.kernel.org
Subject: Re: [PATCH v12 8/8] cgroup: implement the PIDs subsystem
Date: Tue, 19 May 2015 15:11:34 +0200 (CEST)	[thread overview]
Message-ID: <alpine.DEB.2.11.1505191508530.4225@nanos> (raw)
In-Reply-To: <CAOviyaij2bays4aYQ_5HcopcBOfj4M_gKE7oQ_FsV38skK6vWA@mail.gmail.com>

On Tue, 19 May 2015, Aleksa Sarai wrote:

> >> However, it should be noted that organisational operations (adding and
> >> removing tasks from a PIDs hierarchy) will *not* be prevented.
> >
> > This is how you spell: broken controller.
> 
> This has been discussed before. Organisational operations (i.e.
> attaching to a cgroup) are not to be blocked by a cgroup controller in
> the unified hierarchy. You simply can't escape out of a parent
> cgroup's limit through attaching to a child cgroup (because you will
> attach either before the fork checks against the cgroup [in which case
> the child's limit is followed -- which means you also follow the
> parent's limit] or after it checks [which means you'll hit the
> parent's limit and won't be able to fork]).

That's complete and utter nonsense. What has the parent limit to do
with the overflow of the child limit?

parent:	 limit 100   usecnt 80
child:	 limit 10    usecnt 10

So moving anything into child is violating the constraints and has to
be refused. Anything else is just dirty hackery.

Thanks,

	tglx