Re: [PATCH v3 2/7] socket: initial cgroup code.

[Glauber Costa]

On 09/21/2011 03:47 PM, Greg Thelen wrote:
> On Sun, Sep 18, 2011 at 5:56 PM, Glauber Costa<glommer@xxxxxxxxxxxxx>  wrote:
> > We aim to control the amount of kernel memory pinned at any
> > time by tcp sockets. To lay the foundations for this work,
> > this patch adds a pointer to the kmem_cgroup to the socket
> > structure.
> >
> > Signed-off-by: Glauber Costa<glommer@xxxxxxxxxxxxx>
> > CC: David S. Miller<davem@xxxxxxxxxxxxx>
> > CC: Hiroyouki Kamezawa<kamezawa.hiroyu@xxxxxxxxxxxxxx>
> > CC: Eric W. Biederman<ebiederm@xxxxxxxxxxxx>
> ...
> > +void sock_update_memcg(struct sock *sk)
> > +{
> > +       /* right now a socket spends its whole life in the same cgroup */
> > +       BUG_ON(sk->sk_cgrp);
> > +
> > +       rcu_read_lock();
> > +       sk->sk_cgrp = mem_cgroup_from_task(current);
> > +
> > +       /*
> > +        * We don't need to protect against anything task-related, because
> > +        * we are basically stuck with the sock pointer that won't change,
> > +        * even if the task that originated the socket changes cgroups.
> > +        *
> > +        * What we do have to guarantee, is that the chain leading us to
> > +        * the top level won't change under our noses. Incrementing the
> > +        * reference count via cgroup_exclude_rmdir guarantees that.
> > +        */
> > +       cgroup_exclude_rmdir(mem_cgroup_css(sk->sk_cgrp));
>
> This grabs a css_get() reference, which prevents rmdir (will return
> -EBUSY).
Yes.

 How long is this reference held?
For the socket lifetime.

> I wonder about the case
> where a process creates a socket in memcg M1 and later is moved into
> memcg M2.  At that point an admin would expect to be able to 'rmdir
> M1'.  I think this rmdir would return -EBUSY and I suspect it would be
> difficult for the admin to understand why the rmdir of M1 failed.  It
> seems that to rmdir a memcg, an admin would have to kill all processes
> that allocated sockets while in M1.  Such processes may not still be
> in M1.
>
> > +       rcu_read_unlock();
> > +}
I agree. But also, don't see too much ways around it without 
implementing full task migration.

Right now I am working under the assumption that tasks are long lived 
inside the cgroup. Migration potentially introduces some nasty locking 
problems in the mem_schedule path.

Also, unless I am missing something, the memcg already has the policy of
not carrying charges around, probably because of this very same complexity.

True that at least it won't EBUSY you... But I think this is at least a 
way to guarantee that the cgroup under our nose won't disappear in the 
middle of our allocations.

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/