linux-kernel - Re: [PATCH] memcg: fix stale swap cache leak v5

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]

Message-Id: <20090502115649.027e4a88.d-nishimura@mtf.biglobe.ne.jp>
Date:	Sat, 2 May 2009 11:56:49 +0900
From:	Daisuke Nishimura <d-nishimura@....biglobe.ne.jp>
To:	balbir@...ux.vnet.ibm.com
Cc:	KAMEZAWA Hiroyuki <kamezawa.hiroyu@...fujitsu.com>,
	"linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
	"linux-mm@...ck.org" <linux-mm@...ck.org>,
	"akpm@...ux-foundation.org" <akpm@...ux-foundation.org>,
	"hugh@...itas.com" <hugh@...itas.com>,
	d-nishimura@....biglobe.ne.jp, nishimura@....nes.nec.co.jp
Subject: Re: [PATCH] memcg: fix stale swap cache leak v5

> > > Daisuke in the race conditions mentioned is (2) significant? Since the
> > > accounting is already fixed during mem_cgroup_uncharge_page()?
> > > 
> > Do you mean type-2 stale swap caches I described before ?
> > 
> > They doesn't pressure mem.usage nor memsw.usage as you say,
> > but consumes swp_entry(of cource, type-1 has this problem too).
> > As a result, all the swap space can be used up and causes OOM.
> > 
> 
> Good point..
> 
> > I've verified it long ago by:
> > 
> > - make swap space small(50MB).
> > - set mem.limit(32MB).
> > - run some programs(allocate, touch sometimes, exit) enough to
> >   exceed mem.limit repeatedly(I used page01 included in ltp and run
> >   5 instances 8MB per each in cpuset with 4cpus.).
> > - wait for a very long time :) (2,30 hours IIRC)
> >   You can see the usage of swap cache(grep SwapCached /proc/meminfo)
> >   increasing gradually.
> > 
> > 
> > BTW, I'm now testing a attached patch to fix type-2 with setting page-cluster
> > to 0 to aboid type-1, and seeing what happens in the usage of swap cache.
> > (I can't test it in large box though, because my office is closed till May 06.)
> > 
In my small box(i386/2CPU/2GB mem), a similar test shows after 12H
about 600MB leak of swap cache before applying this patch,
while no outstanding leak can be seen after it.

(snip)

> Looking through the patch, I have my doubts
> 
>  shrink_page_list() will catch the page - how? It is not on memcg's
> LRU, so if we have a small cgroup with lots of memory, when the cgroup
> is running out of memory, will this page show up in
> shrink_page_list()?
> 
It's just because the page has not been uncharged yet when shrink_page_list()
is called(,more precicely, when shrink_inactive_list() isolates the page),
so it can be handled in memcg's LRU scanning.


Thanks,
Daisuke Nishimura.
--
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/