From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1755748AbZETBpT (ORCPT ); Tue, 19 May 2009 21:45:19 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S1755310AbZETBpG (ORCPT ); Tue, 19 May 2009 21:45:06 -0400 Received: from mga14.intel.com ([143.182.124.37]:57207 "EHLO mga14.intel.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1755212AbZETBpE (ORCPT ); Tue, 19 May 2009 21:45:04 -0400 X-ExtLoop1: 1 X-IronPort-AV: E=Sophos;i="4.41,218,1241420400"; d="scan'208";a="144771135" Date: Wed, 20 May 2009 09:44:45 +0800 From: Wu Fengguang To: KOSAKI Motohiro Cc: Christoph Lameter , Andrew Morton , LKML , Elladan , Nick Piggin , Johannes Weiner , Peter Zijlstra , Rik van Riel , "tytso@mit.edu" , "linux-mm@kvack.org" , "minchan.kim@gmail.com" Subject: Re: [PATCH 2/3] vmscan: make mapped executable pages the first class citizen Message-ID: <20090520014445.GA7645@localhost> References: <20090519161756.4EE4.A69D9226@jp.fujitsu.com> <20090519074925.GA690@localhost> <20090519170208.742C.A69D9226@jp.fujitsu.com> <20090519085354.GB2121@localhost> <2f11576a0905190528n5eb29e3fme42785a76eed3551@mail.gmail.com> MIME-Version: 1.0 Content-Type: multipart/mixed; boundary="J/dobhs11T7y2rNN" Content-Disposition: inline Content-Transfer-Encoding: 8bit In-Reply-To: <2f11576a0905190528n5eb29e3fme42785a76eed3551@mail.gmail.com> User-Agent: Mutt/1.5.18 (2008-05-17) Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org --J/dobhs11T7y2rNN Content-Type: text/plain; charset=utf-8 Content-Disposition: inline Content-Transfer-Encoding: 8bit On Tue, May 19, 2009 at 08:28:28PM +0800, KOSAKI Motohiro wrote: > Hi > > 2009/5/19 Wu Fengguang : > > On Tue, May 19, 2009 at 04:06:35PM +0800, KOSAKI Motohiro wrote: > >> > > > Like the console mode, the absolute nr_mapped drops considerably - to 1/13 of > >> > > > the original size - during the streaming IO. > >> > > > > >> > > > The delta of pgmajfault is 3 vs 107 during IO, or 236 vs 393 during the whole > >> > > > process. > >> > > > >> > > hmmm. > >> > > > >> > > about 100 page fault don't match Elladan's problem, I think. > >> > > perhaps We missed any addional reproduce condition? > >> > > >> > Elladan's case is not the point of this test. > >> > Elladan's IO is use-once, so probably not a caching problem at all. > >> > > >> > This test case is specifically devised to confirm whether this patch > >> > works as expected. Conclusion: it is. > >> > >> Dejection ;-) > >> > >> The number should address the patch is useful or not. confirming as expected > >> is not so great. > > > > OK, let's make the conclusion in this way: > > > > The changelog analyzed the possible beneficial situation, and this > > test backs that theory with real numbers, ie: it successfully stops > > major faults when the active file list is slowly scanned when there > > are partially cache hot streaming IO. > > > > Another (amazing) finding of the test is, only around 1/10 mapped pages > > are actively referenced in the absence of user activities. > > > > Shall we protect the remaining 9/10 inactive ones? This is a question ;-) > > Unfortunately, I don't reproduce again. > I don't apply your patch yet. but mapped ratio is reduced only very little. mapped ratio or absolute numbers? The ratio wont change much because nr_mapped is already small. > I think smem can show which library evicted. Can you try it? > > download: http://www.selenic.com/smem/ > usage: ./smem -m -r --abbreviate Sure, but I don't see much change in its output (see attachments). smem-console-0 is collected after fresh boot, smem-console-1 is collected after the big IO. > We can't decide 9/10 is important or not. we need know actual evicted file list. Right. But what I measured is the activeness. Almost zero major page faults means the evicted 90% mapped pages are inactive during the long 300 seconds of IO. Thanks, Fengguang > > Or, shall we take the "protect active VM_EXEC mapped pages" approach, > > or Christoph's "protect all mapped pages all time, unless they grow > > too large" attitude?  I still prefer the best effort VM_EXEC heuristics. > > > > 1) the partially cache hot streaming IO is far more likely to happen > >   on (file) servers. For them, evicting the 9/10 inactive mapped > >   pages over night should be acceptable for sysadms. > > > > 2) for use-once IO on desktop, we have Rik's active file list > >   protection heuristics, so nothing to worry at all. > > > > 3) for big working set small memory desktop, the active list will > >   still be scanned, in this situation, why not evict some of the > >   inactive mapped pages? If they have not been accessed for 1 minute, > >   they are not likely be the user focus, and the tight memory > >   constraint can only afford to cache the user focused working set. > > > > Does that make sense? --J/dobhs11T7y2rNN Content-Type: text/plain; charset=us-ascii Content-Disposition: attachment; filename=smem-console-0 Map PIDs AVGPSS PSS [heap] 2 1.5M 2.9M 2 538.0K 1.1M /usr/bin/python2.5 1 1.0M 1.0M /bin/zsh4 1 592.0K 592.0K /usr/lib/zsh/4.3.9/zsh/zle.so 1 216.0K 216.0K /lib/libncursesw.so.5.7 1 156.0K 156.0K /lib/libc-2.9.so 2 65.0K 130.0K [stack] 2 62.0K 124.0K /usr/lib/zsh/4.3.9/zsh/complete.so 1 56.0K 56.0K /usr/lib/locale/locale-archive 2 28.0K 56.0K /usr/lib/python2.5/lib-dynload/operator. 1 32.0K 32.0K /lib/libm-2.9.so 2 15.0K 30.0K /usr/lib/zsh/4.3.9/zsh/complist.so 1 28.0K 28.0K /lib/ld-2.9.so 2 13.0K 26.0K /usr/lib/zsh/4.3.9/zsh/parameter.so 1 24.0K 24.0K /usr/lib/python2.5/lib-dynload/_struct.s 1 24.0K 24.0K /usr/lib/zsh/4.3.9/zsh/zutil.so 1 20.0K 20.0K /usr/lib/python2.5/lib-dynload/time.so 1 20.0K 20.0K /usr/lib/python2.5/lib-dynload/strop.so 1 20.0K 20.0K /usr/lib/python2.5/lib-dynload/_locale.s 1 20.0K 20.0K /lib/libpthread-2.9.so 1 18.0K 18.0K /lib/libdl-2.9.so 2 9.0K 18.0K /usr/lib/zsh/4.3.9/zsh/rlimits.so 1 16.0K 16.0K /lib/libcap.so.2.11 1 16.0K 16.0K /lib/libattr.so.1.1.0 1 16.0K 16.0K /usr/lib/zsh/4.3.9/zsh/terminfo.so 1 12.0K 12.0K /usr/lib/python2.5/lib-dynload/grp.so 1 12.0K 12.0K /lib/libutil-2.9.so 1 11.0K 11.0K /lib/libnss_nis-2.9.so 1 11.0K 11.0K /lib/libnss_files-2.9.so 1 10.0K 10.0K /lib/libnss_compat-2.9.so 1 10.0K 10.0K /lib/libnsl-2.9.so 1 10.0K 10.0K /usr/lib/gconv/gconv-modules.cache 1 5.0K 5.0K [vsyscall] 2 0 0 [vdso] 2 0 0 --J/dobhs11T7y2rNN Content-Type: text/plain; charset=us-ascii Content-Disposition: attachment; filename=smem-console-1 Map PIDs AVGPSS PSS [heap] 2 1.5M 2.9M 2 538.0K 1.1M /usr/bin/python2.5 1 1.0M 1.0M /bin/zsh4 1 496.0K 496.0K /lib/libc-2.9.so 2 180.0K 360.0K [stack] 2 60.0K 120.0K /lib/ld-2.9.so 2 58.0K 116.0K /usr/lib/zsh/4.3.9/zsh/zle.so 1 112.0K 112.0K /lib/libncursesw.so.5.7 1 56.0K 56.0K /usr/lib/locale/locale-archive 2 26.0K 52.0K /lib/libpthread-2.9.so 1 48.0K 48.0K /lib/libm-2.9.so 2 21.0K 42.0K /usr/lib/python2.5/lib-dynload/operator. 1 32.0K 32.0K /usr/lib/zsh/4.3.9/zsh/complete.so 1 24.0K 24.0K /usr/lib/python2.5/lib-dynload/_struct.s 1 24.0K 24.0K /lib/libdl-2.9.so 2 12.0K 24.0K /usr/lib/python2.5/lib-dynload/time.so 1 20.0K 20.0K /usr/lib/python2.5/lib-dynload/strop.so 1 20.0K 20.0K /usr/lib/python2.5/lib-dynload/_locale.s 1 20.0K 20.0K /lib/libutil-2.9.so 1 16.0K 16.0K /usr/lib/python2.5/lib-dynload/grp.so 1 12.0K 12.0K /lib/libcap.so.2.11 1 12.0K 12.0K /lib/libnss_compat-2.9.so 1 9.0K 9.0K /usr/lib/zsh/4.3.9/zsh/zutil.so 1 8.0K 8.0K /usr/lib/zsh/4.3.9/zsh/rlimits.so 1 8.0K 8.0K /lib/libnss_nis-2.9.so 1 8.0K 8.0K /lib/libnss_files-2.9.so 1 8.0K 8.0K /lib/libnsl-2.9.so 1 8.0K 8.0K /usr/lib/zsh/4.3.9/zsh/terminfo.so 1 4.0K 4.0K /usr/lib/zsh/4.3.9/zsh/parameter.so 1 4.0K 4.0K /usr/lib/zsh/4.3.9/zsh/complist.so 1 4.0K 4.0K /lib/libattr.so.1.1.0 1 4.0K 4.0K [vsyscall] 2 0 0 [vdso] 2 0 0 /usr/lib/gconv/gconv-modules.cache 1 0 0 --J/dobhs11T7y2rNN--