linux-kernel - Re: [PATCH 1/5] vmscan: remove all_unreclaimable check from direct reclaim path completely

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, 24 Mar 2011 16:25:43 +0900
From:	Minchan Kim <minchan.kim@...il.com>
To:	KOSAKI Motohiro <kosaki.motohiro@...fujitsu.com>
Cc:	linux-kernel@...r.kernel.org,
	Andrew Morton <akpm@...ux-foundation.org>,
	David Rientjes <rientjes@...gle.com>,
	Linus Torvalds <torvalds@...ux-foundation.org>,
	Rik van Riel <riel@...hat.com>,
	Oleg Nesterov <oleg@...hat.com>, linux-mm <linux-mm@...ck.org>,
	Andrey Vagin <avagin@...nvz.org>,
	Hugh Dickins <hughd@...gle.com>,
	KAMEZAWA Hiroyuki <kamezawa.hiroyu@...fujitsu.com>,
	Nick Piggin <npiggin@...nel.dk>,
	Johannes Weiner <hannes@...xchg.org>
Subject: Re: [PATCH 1/5] vmscan: remove all_unreclaimable check from direct
 reclaim path completely

On Thu, Mar 24, 2011 at 4:03 PM, KOSAKI Motohiro
<kosaki.motohiro@...fujitsu.com> wrote:
>> On Thu, Mar 24, 2011 at 3:16 PM, KOSAKI Motohiro
>> <kosaki.motohiro@...fujitsu.com> wrote:
>> > Hi
>> >
>> >> Thanks for your effort, Kosaki.
>> >> But I still doubt this patch is good.
>> >>
>> >> This patch makes early oom killing in hibernation as it skip
>> >> all_unreclaimable check.
>> >> Normally,  hibernation needs many memory so page_reclaim pressure
>> >> would be big in small memory system. So I don't like early give up.
>> >
>> > Wait. When occur big pressure? hibernation reclaim pressure
>> > (sc->nr_to_recliam) depend on physical memory size. therefore
>> > a pressure seems to don't depend on the size.
>>
>> It depends on physical memory size and /sys/power/image_size.
>> If you want to tune image size bigger, reclaim pressure would be big.
>
> Ok, _If_ I want.
> However, I haven't seen desktop people customize it.
>
>
>> >> Do you think my patch has a problem? Personally, I think it's very
>> >> simple and clear. :)
>> >
>> > To be honest, I dislike following parts. It's madness on madness.
>> >
>> >        static bool zone_reclaimable(struct zone *zone)
>> >        {
>> >                if (zone->all_unreclaimable)
>> >                        return false;
>> >
>> >                return zone->pages_scanned < zone_reclaimable_pages(zone) * 6;
>> >        }
>> >
>> >
>> > The function require a reviewer know
>> >
>> >  o pages_scanned and all_unreclaimable are racy
>>
>> Yes. That part should be written down of comment.
>>
>> >  o at hibernation, zone->all_unreclaimable can be false negative,
>> >   but can't be false positive.
>>
>> The comment of all_unreclaimable already does explain it well, I think.
>
> Where is?
>
>
>> > And, a function comment of all_unreclaimable() says
>> >
>> >         /*
>> >          * As hibernation is going on, kswapd is freezed so that it can't mark
>> >          * the zone into all_unreclaimable. It can't handle OOM during hibernation.
>> >          * So let's check zone's unreclaimable in direct reclaim as well as kswapd.
>> >          */
>> >
>> > But, now it is no longer copy of kswapd algorithm.
>>
>> The comment don't say it should be a copy of kswapd.
>
> I meant the comments says
>
>           * So let's check zone's unreclaimable in direct reclaim as well as kswapd.
>
> but now it isn't aswell as kswapd.
>
> I think it's critical important. If people can't understand why the
> algorithm was choosed, anyone will break the code again sooner or later.
>
>
>> > If you strongly prefer this idea even if you hear above explanation,
>> > please consider to add much and much comments. I can't say
>> > current your patch is enough readable/reviewable.
>>
>> My patch isn't a formal patch for merge but just a concept to show.
>> If you agree the idea, of course, I will add more concrete comment
>> when I send formal patch.
>>
>> Before, I would like to get a your agreement. :)
>> If you solve my concern(early give up in hibernation) in your patch, I
>> don't insist on my patch, either.
>
> Ok. Let's try.
>
> Please concern why priority=0 is not enough. zone_reclaimable_pages(zone) * 6
> is a conservative value of worry about multi thread race. While one task
> is reclaiming, others can allocate/free memory concurrently. therefore,
> even after priority=0, we have a chance getting reclaimable pages on lru.
> But, in hibernation case, almost all tasks was freezed before hibernation
> call shrink_all_memory(). therefore, there is no race. priority=0 reclaim
> can cover all lru pages.
>
> Is this enough explanation for you?

For example, In 4G desktop system.
32M full scanning and fail to reclaim a page.
It's under 1% coverage.

Is it enough to give up?
I am not sure.

-- 
Kind regards,
Minchan Kim
--
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/