Re: -mm merge plans for 2.6.26 (memcgroup)

[Peter Zijlstra <peterz@infradead.org>]

On Mon, 2008-04-21 at 12:17 +0530, Balbir Singh wrote:
> * Balbir Singh <balbir@linux.vnet.ibm.com> [2008-04-21 12:14:28]:
> 
> > Andrew Morton wrote:
> > >> On Mon, 21 Apr 2008 09:30:59 +0900 KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com> wrote:
> > >> On Mon, 21 Apr 2008 00:51:30 +0100 (BST)
> > >> Hugh Dickins <hugh@veritas.com> wrote:
> > >>>>  disable-the-memory-controller-by-default-v3.patch
> > >>>>  disable-the-memory-controller-by-default-v3-fix.patch
> > >>> If those are to go in, then the sooner the better, yes.
> > >>>
> > >>> But though I argued for cgroup_disable=memory (or some such),
> > >>> I think myself that taking it even further now (requiring an
> > >>> additional cgroup_enable=memory at boottime to get the memcg
> > >>> stuff you chose with CGROUP_MEM_RES_CTLR=y at build time) is
> > >>> confusing overkill, just messing around.
> > > 
> > > Yes, it does sound a bit silly.  I'd say just enable it, and provide a
> > > cgroup_disable.
> > > 
> > 
> OK, fair enough. Andi Kleen spoke about the overhead and how distros would
> be impacted if they enabled CONFIG_MEM_RES_CTLR and it was not disabled by
> default. I think the enable/disable is good. I just need to turn on/off
> mem_cgroup_subsys.disabled. The patch is as simple as
>  
> Signed-off-by: Balbir Singh <balbir@linux.vnet.ibm.com>
> ---
> 
>  mm/memcontrol.c |    1 -
>  1 file changed, 1 deletion(-)
> 
> diff -puN mm/memcontrol.c~memcg-dont-disable-by-default mm/memcontrol.c
> --- linux-2.6.25/mm/memcontrol.c~memcg-dont-disable-by-default	2008-04-21 12:11:31.000000000 +0530
> +++ linux-2.6.25-balbir/mm/memcontrol.c	2008-04-21 12:11:40.000000000 +0530
> @@ -1108,5 +1108,4 @@ struct cgroup_subsys mem_cgroup_subsys =
>  	.populate = mem_cgroup_populate,
>  	.attach = mem_cgroup_move_task,
>  	.early_init = 0,
> -	.disabled = 1,
>  };
> _
> 
> This would enable the memory controller by default, it can be disabled
> using cgroup_disable=memory at boot time.

how about working on a runtime switch that only enables the whole
memcgroup machinery when an actual cgroup gets created - and disable it
again once the last cgroup goes.