[RFC 1/2] memcg: hierarchy support core (yet another one)

[KAMEZAWA Hiroyuki]

On Thu, 29 May 2008 14:11:04 +0900 (JST)
yamamoto at valinux.co.jp (YAMAMOTO Takashi) wrote:

> > @@ -39,6 +39,18 @@ struct res_counter {
> >  	 */
> >  	unsigned long long failcnt;
> >  	/*
> > +	 * the amount of resource comes from parenet cgroup. Should be
> > +	 * returned to the parent at destroying/resizing this res_counter.
> > +	 */
> > +	unsigned long long borrow;
> 
> why do you need this in addition to the limit?
> ie. aren't their values always equal except the root cgroup?
> 
yes, except the root group. that's a reason....no,no

To be honest, I thought of different concept of hierarchy when I started this
and borrow != limit in first version. But it was complicated and big..
Finally, I set borrow=limit but I didn't remove "borrrow" because it seems
to help a man to undetstand the whole logic.

I'm now retrying borrow != limit version, again. (but no good progress ;)

Thanks,
-Kame