Re: [RFC][ for -mm] memory controller enhancements for NUMA [1/10] record nid/zid on page_cgroup

[yamamoto-jCdQPDEk3idL9jVzuh4AOg@public.gmane.org (YAMAMOTO Takashi)]

> This patch adds nid/zoneid value to page cgroup.
> This helps per-zone accounting for memory cgroup and reclaim routine.
> 
> Signed-off-by: KAMEZAWA Hiroyuki <kmaezawa.hiroyu-+CUm20s59erQFUHtdCDX3A@public.gmane.org>
> 
>  mm/memcontrol.c |    6 ++++++
>  1 file changed, 6 insertions(+)
> 
> Index: linux-2.6.24-rc2-mm1/mm/memcontrol.c
> ===================================================================
> --- linux-2.6.24-rc2-mm1.orig/mm/memcontrol.c
> +++ linux-2.6.24-rc2-mm1/mm/memcontrol.c
> @@ -131,6 +131,8 @@ struct page_cgroup {
>  	atomic_t ref_cnt;		/* Helpful when pages move b/w  */
>  					/* mapped and cached states     */
>  	int	 flags;
> +	short	nid;
> +	short	zid;
>  };
>  #define PAGE_CGROUP_FLAG_CACHE	(0x1)	/* charged as cache */
>  #define PAGE_CGROUP_FLAG_ACTIVE (0x2)	/* page is active in this cgroup */
> @@ -216,6 +218,10 @@ void page_assign_page_cgroup(struct page
>  		VM_BUG_ON(!page_cgroup_locked(page));
>  	locked = (page->page_cgroup & PAGE_CGROUP_LOCK);
>  	page->page_cgroup = ((unsigned long)pc | locked);
> +	if (pc) {
> +		pc->nid = page_to_nid(page);
> +		pc->zid = page_zonenum(page);
> +	}
>  }
>  
>  struct page_cgroup *page_get_page_cgroup(struct page *page)

are they worth to be cached?
can't you use page_zonenum(pc->page)?

YAMAMOTO Takashi