Re: [PATCH 1/5] cgroup: ID notification call back

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

On Tue, 24 Aug 2010 18:35:00 -0700
Paul Menage <menage@google.com> wrote:

> On Tue, Aug 24, 2010 at 6:03 PM, KAMEZAWA Hiroyuki
> <kamezawa.hiroyu@jp.fujitsu.com> wrote:
> >
> > Hmm. How this pseudo code looks like ? This passes "new id" via
> > cgroup->subsys[array] at creation. (Using union will be better, maybe).
> >
> 
> That's rather ugly. I was thinking of something more like this. (Not
> even compiled yet, and the only subsystem updated is cpuset).
> 

Hmm, but placing css and subsystem's its own structure in different cache line
can increase cacheline/TLB miss, I think.

I wonder I should stop this patch series and do small thing.
I prefer to call alloc_css_id() by ->create() call by subsys's its own decistion
is much better and cleaner. (as my original design)

mem_cgroup_create()
{

	cgroup_attach_css_id(ss, cgrp, &mem->css);
}

And then, there will be no difficulty.

Do we have to call alloc_css_id() in kernel/cgroup.c ?

Thanks,
-Kame