linux-kernel - Re: [PATCH RFC 00/15] mm: memory book keeping and lru

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:	Thu, 16 Feb 2012 09:43:52 +0400
From:	Konstantin Khlebnikov <khlebnikov@...nvz.org>
To:	KAMEZAWA Hiroyuki <kamezawa.hiroyu@...fujitsu.com>
CC:	"linux-mm@...ck.org" <linux-mm@...ck.org>,
	"linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
	Hugh Dickins <hughd@...gle.com>,
	"hannes@...xchg.org" <hannes@...xchg.org>
Subject: Re: [PATCH RFC 00/15] mm: memory book keeping and lru_lock splitting

KAMEZAWA Hiroyuki wrote:
> On Thu, 16 Feb 2012 02:57:04 +0400
> Konstantin Khlebnikov<khlebnikov@...nvz.org>  wrote:
>
>> There should be no logic changes in this patchset, this is only tossing bits around.
>> [ This patchset is on top some memcg cleanup/rework patches,
>>    which I sent to linux-mm@ today/yesterday ]
>>
>> Most of things in this patchset are self-descriptive, so here brief plan:
>>
>
> AFAIK, Hugh Dickins said he has per-zone-per-lru-lock and is testing it.
> So, please CC him and Johannes, at least.
>

Ok

>
>> * Transmute struct lruvec into struct book. Like real book this struct will
>>    store set of pages for one zone. It will be working unit for reclaimer code.
>> [ If memcg is disabled in config there will only one book embedded into struct zone ]
>>
>
> Why you need to add new structure rahter than enhancing lruvec ?
> "book" means a binder of pages ?
>

I responded to this in the reply to Hugh Dickins.

>
>> * move page-lru counters to struct book
>> [ this adds extra overhead in add_page_to_lru_list()/del_page_from_lru_list() for
>>    non-memcg case, but I believe it will be invisible, only one non-atomic add/sub
>>    in the same cacheline with lru list ]
>>
>
> This seems straightforward.
>
>> * unify inactive_list_is_low_global() and cleanup reclaimer code
>> * replace struct mem_cgroup_zone with single pointer to struct book
>
> Hm, ok.
>
>> * optimize page to book translations, move it upper in the call stack,
>>    replace some struct zone arguments with struct book pointer.
>>
>
> a page->book transrater from patch 2/15
>
> +struct book *page_book(struct page *page)
> +{
> +	struct mem_cgroup_per_zone *mz;
> +	struct page_cgroup *pc;
> +
> +	if (mem_cgroup_disabled())
> +		return&page_zone(page)->book;
> +
> +	pc = lookup_page_cgroup(page);
> +	if (!PageCgroupUsed(pc))
> +		return&page_zone(page)->book;
> +	/* Ensure pc->mem_cgroup is visible after reading PCG_USED. */
> +	smp_rmb();
> +	mz = mem_cgroup_zoneinfo(pc->mem_cgroup,
> +			page_to_nid(page), page_zonenum(page));
> +	return&mz->book;
> +}
>
> What happens when pc->mem_cgroup is rewritten by move_account() ?
> Where is the guard for lockless access of this ?

Initially this suppose to be protected with lru_lock, in final patch they are protected with rcu.
After final patch all page_book() calls are collected in [__re]lock_page_book[_irq]() functions.
They pick some book reference, lock its lru and recheck page -> book reference in loop till success.

Currently I found there only one potential problem: free_mem_cgroup_per_zone_info() in "mm: memory bookkeeping core"
maybe should call spin_unlock_wait(&zone->lru_lock), because some guy can pick page_book(pfn_to_page(pfn))
and try to isolate this page. But I not sure, how this is possible. In final patch it is totally fixed with rcu.

>
> Thanks,
> -Kame
>

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/