Re: [RFC][PATCH 3/3] memcg: atomic update of memcg pointer and other bits.

From: Michal Hocko <mhocko@suse.cz>
To: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: linux-mm@kvack.org, cgroups@vger.kernel.org,
	Johannes Weiner <hannes@cmpxchg.org>,
	Hugh Dickins <hughd@google.com>, Han Ying <yinghan@google.com>,
	Glauber Costa <glommer@parallels.com>,
	"Aneesh Kumar K.V" <aneesh.kumar@linux.vnet.ibm.com>,
	Andrew Morton <akpm@linux-foundation.org>,
	suleiman@google.com, n-horiguchi@ah.jp.nec.com,
	khlebnikov@openvz.org, Tejun Heo <tj@kernel.org>
Subject: Re: [RFC][PATCH 3/3] memcg: atomic update of memcg pointer and other bits.
Date: Thu, 22 Mar 2012 14:38:20 +0100	[thread overview]
Message-ID: <20120322133820.GE18665@tiehlicka.suse.cz> (raw)
In-Reply-To: <4F66E85E.6030000@jp.fujitsu.com>

On Mon 19-03-12 17:03:42, KAMEZAWA Hiroyuki wrote:
[...]
> @@ -1237,8 +1237,6 @@ mem_cgroup_get_reclaim_stat_from_page(struct page *page)
>  	pc = lookup_page_cgroup(page);
>  	if (!PageCgroupUsed(pc))
>  		return NULL;
> -	/* Ensure pc's mem_cgroup is visible after reading PCG_USED. */
> -	smp_rmb();
>  	mz = page_cgroup_zoneinfo(pc_to_mem_cgroup(pc), page);
>  	return &mz->reclaim_stat;
>  }
> @@ -2491,16 +2489,7 @@ static void __mem_cgroup_commit_charge(struct mem_cgroup *memcg,
>  		}
>  	}
>  
> -	pc_set_mem_cgroup(pc, memcg);
> -	/*
> -	 * We access a page_cgroup asynchronously without lock_page_cgroup().
> -	 * Especially when a page_cgroup is taken from a page, pc's mem_cgroup
> -	 * is accessed after testing USED bit. To make pc's mem_cgroup visible
> -	 * before USED bit, we need memory barrier here.
> -	 * See mem_cgroup_add_lru_list(), etc.
> - 	 */
> -	smp_wmb();
> -	SetPageCgroupUsed(pc);
> +	pc_set_mem_cgroup(pc, memcg, BIT(PCG_USED) | BIT(PCG_LOCK));

This is not nice. Maybe we need two variants (pc_set_mem_cgroup[_flags])?

>  	if (lrucare) {
>  		if (was_on_lru) {
> @@ -2529,7 +2518,6 @@ static void __mem_cgroup_commit_charge(struct mem_cgroup *memcg,
>  
>  #ifdef CONFIG_TRANSPARENT_HUGEPAGE
>  
> -#define PCGF_NOCOPY_AT_SPLIT ((1 << PCG_LOCK) | (1 << PCG_MIGRATION))
>  /*
>   * Because tail pages are not marked as "used", set it. We're under
>   * zone->lru_lock, 'splitting on pmd' and compound_lock.
> @@ -2547,9 +2535,7 @@ void mem_cgroup_split_huge_fixup(struct page *head)
>  		return;
>  	for (i = 1; i < HPAGE_PMD_NR; i++) {
>  		pc = head_pc + i;
> -		pc_set_mem_cgroup(pc, memcg);
> -		smp_wmb();/* see __commit_charge() */
> -		pc->flags = head_pc->flags & ~PCGF_NOCOPY_AT_SPLIT;
> +		pc_set_mem_cgroup(pc, memcg, BIT(PCG_USED));

Maybe it would be cleaner to remove PCGF_NOCOPY_AT_SPLIT in a separate patch with 
VM_BUG_ON(!head_pc->flags & BIT(PCG_USED))?

>  	}
>  }
>  #endif /* CONFIG_TRANSPARENT_HUGEPAGE */
> @@ -2616,7 +2602,7 @@ static int mem_cgroup_move_account(struct page *page,
>  		__mem_cgroup_cancel_charge(from, nr_pages);
>  
>  	/* caller should have done css_get */
> -	pc_set_mem_cgroup(pc, to);
> +	pc_set_mem_cgroup(pc, to, BIT(PCG_USED) | BIT(PCG_LOCK));

Same here.

-- 
Michal Hocko
SUSE Labs
SUSE LINUX s.r.o.
Lihovarska 1060/12
190 00 Praha 9    
Czech Republic

--
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/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>