From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from mail138.messagelabs.com (mail138.messagelabs.com [216.82.249.35]) by kanga.kvack.org (Postfix) with SMTP id 7F8F56B004F for ; Thu, 20 Aug 2009 08:13:56 -0400 (EDT) Received: by gxk12 with SMTP id 12so6947218gxk.4 for ; Thu, 20 Aug 2009 05:13:59 -0700 (PDT) MIME-Version: 1.0 In-Reply-To: <20090820114933.GB7359@localhost> References: <20090820024929.GA19793@localhost> <20090820121347.8a886e4b.kamezawa.hiroyu@jp.fujitsu.com> <20090820040533.GA27540@localhost> <28c262360908200401t41c03ad3n114b24e03b61de08@mail.gmail.com> <20090820114933.GB7359@localhost> Date: Thu, 20 Aug 2009 21:13:59 +0900 Message-ID: <28c262360908200513y3fee675do4e1f0204ffb8df63@mail.gmail.com> Subject: Re: [PATCH -v2] mm: do batched scans for mem_cgroup From: Minchan Kim Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: quoted-printable Sender: owner-linux-mm@kvack.org To: Wu Fengguang Cc: KAMEZAWA Hiroyuki , Andrew Morton , Balbir Singh , KOSAKI Motohiro , Rik van Riel , Johannes Weiner , Avi Kivity , Andrea Arcangeli , "Dike, Jeffrey G" , Hugh Dickins , Christoph Lameter , Mel Gorman , LKML , linux-mm , "nishimura@mxp.nes.nec.co.jp" , "lizf@cn.fujitsu.com" , "menage@google.com" List-ID: On Thu, Aug 20, 2009 at 8:49 PM, Wu Fengguang wrote= : > On Thu, Aug 20, 2009 at 07:01:21PM +0800, Minchan Kim wrote: >> Hi, Wu. >> >> On Thu, Aug 20, 2009 at 1:05 PM, Wu Fengguang wr= ote: >> > On Thu, Aug 20, 2009 at 11:13:47AM +0800, KAMEZAWA Hiroyuki wrote: >> >> On Thu, 20 Aug 2009 10:49:29 +0800 >> >> Wu Fengguang wrote: >> >> >> >> > For mem_cgroup, shrink_zone() may call shrink_list() with nr_to_sca= n=3D1, >> >> > in which case shrink_list() _still_ calls isolate_pages() with the = much >> >> > larger SWAP_CLUSTER_MAX. =C2=A0It effectively scales up the inactiv= e list >> >> > scan rate by up to 32 times. >> >> > >> >> > For example, with 16k inactive pages and DEF_PRIORITY=3D12, (16k >>= 12)=3D4. >> >> > So when shrink_zone() expects to scan 4 pages in the active/inactiv= e >> >> > list, it will be scanned SWAP_CLUSTER_MAX=3D32 pages in effect. >> >> > >> >> > The accesses to nr_saved_scan are not lock protected and so not 100= % >> >> > accurate, however we can tolerate small errors and the resulted sma= ll >> >> > imbalanced scan rates between zones. >> >> > >> >> > This batching won't blur up the cgroup limits, since it is driven b= y >> >> > "pages reclaimed" rather than "pages scanned". When shrink_zone() >> >> > decides to cancel (and save) one smallish scan, it may well be call= ed >> >> > again to accumulate up nr_saved_scan. >> >> > >> >> > It could possibly be a problem for some tiny mem_cgroup (which may = be >> >> > _full_ scanned too much times in order to accumulate up nr_saved_sc= an). >> >> > >> >> > CC: Rik van Riel >> >> > CC: Minchan Kim >> >> > CC: Balbir Singh >> >> > CC: KOSAKI Motohiro >> >> > CC: KAMEZAWA Hiroyuki >> >> > Signed-off-by: Wu Fengguang >> >> > --- >> >> >> >> Hmm, how about this ? >> >> =3D=3D >> >> Now, nr_saved_scan is tied to zone's LRU. >> >> But, considering how vmscan works, it should be tied to reclaim_stat. >> >> >> >> By this, memcg can make use of nr_saved_scan information seamlessly. >> > >> > Good idea, full patch updated with your signed-off-by :) >> > >> > Thanks, >> > Fengguang >> > --- >> > mm: do batched scans for mem_cgroup >> > >> > For mem_cgroup, shrink_zone() may call shrink_list() with nr_to_scan= =3D1, >> > in which case shrink_list() _still_ calls isolate_pages() with the muc= h >> > larger SWAP_CLUSTER_MAX. =C2=A0It effectively scales up the inactive l= ist >> > scan rate by up to 32 times. >> >> Yes. It can scan 32 times pages in only inactive list, not active list. > > Yes and no ;) > > inactive anon list over scanned =3D> inactive_anon_is_low() =3D=3D TRUE > =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2= =A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0=3D> shrink_active_list() > =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2= =A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0=3D> active anon list over scanned Why inactive anon list is overscanned in case mem_cgroup ? in shrink_zone, 1) The vm doesn't accumulate nr[l]. 2) Below routine store min value to nr_to_scan. nr_to_scan =3D min(nr[l], swap_cluster_max); ex) if nr[l] =3D 4, vm calls shrink_active_list with 4 as nr_to_scan. So I think overscan doesn't occur in active list. > So the end result may be > > - anon inactive =C2=A0=3D> over scanned > - anon active =C2=A0 =C2=A0=3D> over scanned (maybe not as much) > - file inactive =C2=A0=3D> over scanned > - file active =C2=A0 =C2=A0=3D> under scanned (relatively) > >> > For example, with 16k inactive pages and DEF_PRIORITY=3D12, (16k >> 12= )=3D4. >> > So when shrink_zone() expects to scan 4 pages in the active/inactive >> > list, it will be scanned SWAP_CLUSTER_MAX=3D32 pages in effect. >> >> Active list scan would be scanned in 4, =C2=A0inactive list =C2=A0is 32. > > Exactly. > >> > >> > The accesses to nr_saved_scan are not lock protected and so not 100% >> > accurate, however we can tolerate small errors and the resulted small >> > imbalanced scan rates between zones. >> >> Yes. >> >> > This batching won't blur up the cgroup limits, since it is driven by >> > "pages reclaimed" rather than "pages scanned". When shrink_zone() >> > decides to cancel (and save) one smallish scan, it may well be called >> > again to accumulate up nr_saved_scan. >> >> You mean nr_scan_try_batch logic ? >> But that logic works for just global reclaim? >> Now am I missing something? >> >> Could you elaborate more? :) > > Sorry for the confusion. The above paragraph originates from Balbir's > concern: > > =C2=A0 =C2=A0 =C2=A0 =C2=A0This might be a concern (although not a big AT= M), since we can't > =C2=A0 =C2=A0 =C2=A0 =C2=A0afford to miss limits by much. If a cgroup is = near its limit and we > =C2=A0 =C2=A0 =C2=A0 =C2=A0drop scanning it. We'll have to work out what = this means for the end Why does mem_cgroup drops scanning ? It's because nr_scan_try_batch? or something ? Sorry. Still, I can't understand your point. :( > =C2=A0 =C2=A0 =C2=A0 =C2=A0user. May be more fundamental look through is = required at the priority > =C2=A0 =C2=A0 =C2=A0 =C2=A0based logic of exposing how much to scan, I do= n't know. > > Thanks, > Fengguang > >> > It could possibly be a problem for some tiny mem_cgroup (which may be >> > _full_ scanned too much times in order to accumulate up nr_saved_scan)= . >> > >> > CC: Rik van Riel >> > CC: Minchan Kim >> > CC: Balbir Singh >> > CC: KOSAKI Motohiro >> > Signed-off-by: KAMEZAWA Hiroyuki >> > Signed-off-by: Wu Fengguang >> > --- >> > =C2=A0include/linux/mmzone.h | =C2=A0 =C2=A06 +++++- >> > =C2=A0mm/page_alloc.c =C2=A0 =C2=A0 =C2=A0 =C2=A0| =C2=A0 =C2=A02 +- >> > =C2=A0mm/vmscan.c =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0| =C2=A0 20= +++++++++++--------- >> > =C2=A03 files changed, 17 insertions(+), 11 deletions(-) >> > >> > --- linux.orig/include/linux/mmzone.h =C2=A0 2009-07-30 10:45:15.00000= 0000 +0800 >> > +++ linux/include/linux/mmzone.h =C2=A0 =C2=A0 =C2=A0 =C2=A02009-08-20= 11:51:08.000000000 +0800 >> > @@ -269,6 +269,11 @@ struct zone_reclaim_stat { >> > =C2=A0 =C2=A0 =C2=A0 =C2=A0 */ >> > =C2=A0 =C2=A0 =C2=A0 =C2=A0unsigned long =C2=A0 =C2=A0 =C2=A0 =C2=A0 = =C2=A0 recent_rotated[2]; >> > =C2=A0 =C2=A0 =C2=A0 =C2=A0unsigned long =C2=A0 =C2=A0 =C2=A0 =C2=A0 = =C2=A0 recent_scanned[2]; >> > + >> > + =C2=A0 =C2=A0 =C2=A0 /* >> > + =C2=A0 =C2=A0 =C2=A0 =C2=A0* accumulated for batching >> > + =C2=A0 =C2=A0 =C2=A0 =C2=A0*/ >> > + =C2=A0 =C2=A0 =C2=A0 unsigned long =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2= =A0 nr_saved_scan[NR_LRU_LISTS]; >> > =C2=A0}; >> > >> > =C2=A0struct zone { >> > @@ -323,7 +328,6 @@ struct zone { >> > =C2=A0 =C2=A0 =C2=A0 =C2=A0spinlock_t =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2= =A0 =C2=A0 =C2=A0lru_lock; >> > =C2=A0 =C2=A0 =C2=A0 =C2=A0struct zone_lru { >> > =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0struct list_hea= d list; >> > - =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 unsigned long nr_sa= ved_scan; =C2=A0 =C2=A0/* accumulated for batching */ >> > =C2=A0 =C2=A0 =C2=A0 =C2=A0} lru[NR_LRU_LISTS]; >> > >> > =C2=A0 =C2=A0 =C2=A0 =C2=A0struct zone_reclaim_stat reclaim_stat; >> > --- linux.orig/mm/vmscan.c =C2=A0 =C2=A0 =C2=A02009-08-20 11:48:46.000= 000000 +0800 >> > +++ linux/mm/vmscan.c =C2=A0 2009-08-20 12:00:55.000000000 +0800 >> > @@ -1521,6 +1521,7 @@ static void shrink_zone(int priority, st >> > =C2=A0 =C2=A0 =C2=A0 =C2=A0enum lru_list l; >> > =C2=A0 =C2=A0 =C2=A0 =C2=A0unsigned long nr_reclaimed =3D sc->nr_recla= imed; >> > =C2=A0 =C2=A0 =C2=A0 =C2=A0unsigned long swap_cluster_max =3D sc->swap= _cluster_max; >> > + =C2=A0 =C2=A0 =C2=A0 struct zone_reclaim_stat *reclaim_stat =3D get_= reclaim_stat(zone, sc); >> > =C2=A0 =C2=A0 =C2=A0 =C2=A0int noswap =3D 0; >> > >> > =C2=A0 =C2=A0 =C2=A0 =C2=A0/* If we have no swap space, do not bother = scanning anon pages. */ >> > @@ -1540,12 +1541,9 @@ static void shrink_zone(int priority, st >> > =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 = =C2=A0 =C2=A0scan >>=3D priority; >> > =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 = =C2=A0 =C2=A0scan =3D (scan * percent[file]) / 100; >> > =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0} >> > - =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 if (scanning_global= _lru(sc)) >> > - =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2= =A0 =C2=A0 nr[l] =3D nr_scan_try_batch(scan, >> > - =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2= =A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 = =C2=A0 =C2=A0 =C2=A0 =C2=A0 &zone->lru[l].nr_saved_scan, >> > - =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2= =A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 = =C2=A0 =C2=A0 =C2=A0 =C2=A0 swap_cluster_max); >> > - =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 else >> > - =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2= =A0 =C2=A0 nr[l] =3D scan; >> > + =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 nr[l] =3D nr_scan_t= ry_batch(scan, >> > + =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2= =A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 &= reclaim_stat->nr_saved_scan[l], >> > + =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2= =A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 s= wap_cluster_max); >> > =C2=A0 =C2=A0 =C2=A0 =C2=A0} >> > >> > =C2=A0 =C2=A0 =C2=A0 =C2=A0while (nr[LRU_INACTIVE_ANON] || nr[LRU_ACTI= VE_FILE] || >> > @@ -2128,6 +2126,7 @@ static void shrink_all_zones(unsigned lo >> > =C2=A0{ >> > =C2=A0 =C2=A0 =C2=A0 =C2=A0struct zone *zone; >> > =C2=A0 =C2=A0 =C2=A0 =C2=A0unsigned long nr_reclaimed =3D 0; >> > + =C2=A0 =C2=A0 =C2=A0 struct zone_reclaim_stat *reclaim_stat; >> > >> > =C2=A0 =C2=A0 =C2=A0 =C2=A0for_each_populated_zone(zone) { >> > =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0enum lru_list l= ; >> > @@ -2144,11 +2143,14 @@ static void shrink_all_zones(unsigned lo >> > =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 = =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2= =A0 =C2=A0 =C2=A0 =C2=A0l =3D=3D LRU_ACTIVE_FILE)) >> > =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 = =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0continue; >> > >> > - =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2= =A0 =C2=A0 zone->lru[l].nr_saved_scan +=3D (lru_pages >> prio) + 1; >> > - =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2= =A0 =C2=A0 if (zone->lru[l].nr_saved_scan >=3D nr_pages || pass > 3) { >> > + =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2= =A0 =C2=A0 reclaim_stat =3D get_reclaim_stat(zone, sc); >> > + =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2= =A0 =C2=A0 reclaim_stat->nr_saved_scan[l] +=3D >> > + =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2= =A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 = =C2=A0 =C2=A0 =C2=A0 (lru_pages >> prio) + 1; >> > + =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2= =A0 =C2=A0 if (reclaim_stat->nr_saved_scan[l] >> > + =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2= =A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 = =C2=A0 =C2=A0 =C2=A0 >=3D nr_pages || pass > 3) { >> > =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 = =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0unsigned long nr_to_scan; >> > >> > - =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2= =A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 zone->lru[l].nr_saved_scan =3D 0; >> > + =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2= =A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 reclaim_stat->nr_saved_scan[l] =3D 0= ; >> > =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 = =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0nr_to_scan =3D min(nr_pages, lru_p= ages); >> > =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 = =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0nr_reclaimed +=3D shrink_list(l, n= r_to_scan, zone, >> > =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 = =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2= =A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 = =C2=A0sc, prio); >> > --- linux.orig/mm/page_alloc.c =C2=A02009-08-20 11:57:54.000000000 +08= 00 >> > +++ linux/mm/page_alloc.c =C2=A0 =C2=A0 =C2=A0 2009-08-20 11:58:39.000= 000000 +0800 >> > @@ -3716,7 +3716,7 @@ static void __paginginit free_area_init_ >> > =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0zone_pcp_init(z= one); >> > =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0for_each_lru(l)= { >> > =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 = =C2=A0 =C2=A0INIT_LIST_HEAD(&zone->lru[l].list); >> > - =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2= =A0 =C2=A0 zone->lru[l].nr_saved_scan =3D 0; >> > + =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2= =A0 =C2=A0 zone->reclaim_stat.nr_saved_scan[l] =3D 0; >> > =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0} >> > =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0zone->reclaim_s= tat.recent_rotated[0] =3D 0; >> > =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0zone->reclaim_s= tat.recent_rotated[1] =3D 0; >> > >> > -- >> > To unsubscribe, send a message with 'unsubscribe linux-mm' in >> > the body to majordomo@kvack.org. =C2=A0For more info on Linux MM, >> > see: http://www.linux-mm.org/ . >> > Don't email: email@kvack.org >> > >> >> >> >> -- >> Kind regards, >> Minchan Kim > --=20 Kind regards, Minchan Kim -- 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/ . Don't email: email@kvack.org