Re: [RFC][PATCH 4/4] cgroup-memcg fix frequent EBUSY at rmdir

[KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>]

On Tue, 13 Jan 2009 18:48:43 -0800
Paul Menage <menage@google.com> wrote:

> On Thu, Jan 8, 2009 at 1:35 AM, KAMEZAWA Hiroyuki
> <kamezawa.hiroyu@jp.fujitsu.com> wrote:
> > +       if (ret == -EAGAIN) { /* subsys asks us to retry later */
> > +               mutex_unlock(&cgroup_mutex);
> > +               cond_resched();
> > +               goto retry;
> > +       }
> 
> This spinning worries me a bit. It might be better to do an
> interruptible sleep until the relevant CSS's refcount goes down to
> zero. 

Hmm, add wait_queue to css and wake it up at css_put() ?

like this ?
==
__css_put()
{
	if (atomi_dec_return(&css->refcnt) == 1) {
		if (notify_on_release(cgrp) {
			.....
		}
		if (someone_waiting_rmdir(css)) {
			wake_up_him().
		}
	}
}
==

> And is there no way that the memory controller can hang on to a
> reference indefinitely, if the cgroup still has some pages charged to
> it?
> 
pre_destroy() is for that.  Now, If there are still references from "page"
after pre_destroy(), it's bug.
swap-in after pre_destory() may add new refs from pages.
(I implemented reference from "swap" to be memcg internal refcnt not to css.)

Allowing Ctrl-C/alarm() here by signal_pending() will be better, anyway.

Thanks,
-Kame

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>