From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from mail144.messagelabs.com (mail144.messagelabs.com [216.82.254.51]) by kanga.kvack.org (Postfix) with ESMTP id B274B6B0011 for ; Sun, 22 May 2011 08:22:46 -0400 (EDT) Received: by pzk4 with SMTP id 4so2966745pzk.14 for ; Sun, 22 May 2011 05:22:42 -0700 (PDT) MIME-Version: 1.0 In-Reply-To: References: <4DD5DC06.6010204@jp.fujitsu.com> <20110520140856.fdf4d1c8.kamezawa.hiroyu@jp.fujitsu.com> <20110520101120.GC11729@random.random> <20110520153346.GA1843@barrios-desktop> <20110520161934.GA2386@barrios-desktop> From: Andrew Lutomirski Date: Sun, 22 May 2011 08:22:22 -0400 Message-ID: Subject: Re: Kernel falls apart under light memory pressure (i.e. linking vmlinux) Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: quoted-printable Sender: owner-linux-mm@kvack.org List-ID: To: Minchan Kim Cc: KOSAKI Motohiro , Andrea Arcangeli , KAMEZAWA Hiroyuki , fengguang.wu@intel.com, andi@firstfloor.org, linux-mm@kvack.org, linux-kernel@vger.kernel.org, mgorman@suse.de, hannes@cmpxchg.org, riel@redhat.com On Sat, May 21, 2011 at 10:44 AM, Minchan Kim wrote= : > Hi Andrew. > > On Sat, May 21, 2011 at 10:34 PM, Andrew Lutomirski wrote: >> On Sat, May 21, 2011 at 8:04 AM, KOSAKI Motohiro >> wrote: >>>> diff --git a/mm/vmscan.c b/mm/vmscan.c >>>> index 3f44b81..d1dabc9 100644 >>>> @@ -1426,8 +1437,13 @@ shrink_inactive_list(unsigned long nr_to_scan, >>>> struct zone *zone, >>>> >>>> =A0 =A0 =A0 =A0/* Check if we should syncronously wait for writeback *= / >>>> =A0 =A0 =A0 =A0if (should_reclaim_stall(nr_taken, nr_reclaimed, priori= ty, sc)) { >>>> + =A0 =A0 =A0 =A0 =A0 =A0 =A0 unsigned long nr_active, old_nr_scanned; >>>> =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0set_reclaim_mode(priority, sc, true); >>>> + =A0 =A0 =A0 =A0 =A0 =A0 =A0 nr_active =3D clear_active_flags(&page_l= ist, NULL); >>>> + =A0 =A0 =A0 =A0 =A0 =A0 =A0 count_vm_events(PGDEACTIVATE, nr_active)= ; >>>> + =A0 =A0 =A0 =A0 =A0 =A0 =A0 old_nr_scanned =3D sc->nr_scanned; >>>> =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0nr_reclaimed +=3D shrink_page_list(&pag= e_list, zone, sc); >>>> + =A0 =A0 =A0 =A0 =A0 =A0 =A0 sc->nr_scanned =3D old_nr_scanned; >>>> =A0 =A0 =A0 =A0} >>>> >>>> =A0 =A0 =A0 =A0local_irq_disable(); >>>> >>>> I just tested 2.6.38.6 with the attached patch. =A0It survived dirty_r= am >>>> and test_mempressure without any problems other than slowness, but >>>> when I hit ctrl-c to stop test_mempressure, I got the attached oom. >>> >>> Minchan, >>> >>> I'm confused now. >>> If pages got SetPageActive(), should_reclaim_stall() should never retur= n true. >>> Can you please explain which bad scenario was happen? >>> >>> -----------------------------------------------------------------------= ------------------------------ >>> static void reset_reclaim_mode(struct scan_control *sc) >>> { >>> =A0 =A0 =A0 =A0sc->reclaim_mode =3D RECLAIM_MODE_SINGLE | RECLAIM_MODE_= ASYNC; >>> } >>> >>> shrink_page_list() >>> { >>> =A0(snip) >>> =A0activate_locked: >>> =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0SetPageActive(page); >>> =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0pgactivate++; >>> =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0unlock_page(page); >>> =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0reset_reclaim_mode(sc); =A0 =A0 =A0 =A0 = =A0 =A0 =A0 =A0 =A0/// here >>> =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0list_add(&page->lru, &ret_pages); >>> =A0 =A0 =A0 =A0} >>> -----------------------------------------------------------------------= ------------------------------ >>> >>> >>> -----------------------------------------------------------------------= ------------------------------ >>> bool should_reclaim_stall() >>> { >>> =A0(snip) >>> >>> =A0 =A0 =A0 =A0/* Only stall on lumpy reclaim */ >>> =A0 =A0 =A0 =A0if (sc->reclaim_mode & RECLAIM_MODE_SINGLE) =A0 /// and = here >>> =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0return false; >>> -----------------------------------------------------------------------= ------------------------------ >>> >> >> I did some tracing and the oops happens from the second call to >> shrink_page_list after should_reclaim_stall returns true and it hits >> the same pages in the same order that the earlier call just finished >> calling SetPageActive on. =A0I have *not* confirmed that the two calls >> happened from the same call to shrink_inactive_list, but something's >> certainly wrong in there. >> >> This is very easy to reproduce on my laptop. > > I would like to confirm this problem. > Could you show the diff of 2.6.38.6 with current your 2.6.38.6 + alpha? > (ie, I would like to know that what patches you add up on vanilla > 2.6.38.6 to reproduce this problem) > I believe you added my crap below patch. Right? > > diff --git a/mm/vmscan.c b/mm/vmscan.c > index 292582c..69d317e 100644 > --- a/mm/vmscan.c > +++ b/mm/vmscan.c > @@ -311,7 +311,8 @@ static void set_reclaim_mode(int priority, struct > scan_control *sc, > =A0 =A0 =A0 =A0*/ > =A0 =A0 =A0 if (sc->order > PAGE_ALLOC_COSTLY_ORDER) > =A0 =A0 =A0 =A0 =A0 =A0 =A0 sc->reclaim_mode |=3D syncmode; > - =A0 =A0 =A0 else if (sc->order && priority < DEF_PRIORITY - 2) > + =A0 =A0 =A0 else if ((sc->order && priority < DEF_PRIORITY - 2) || > + =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 prioiry <= =3D DEF_PRIORITY / 3) > =A0 =A0 =A0 =A0 =A0 =A0 =A0 sc->reclaim_mode |=3D syncmode; > =A0 =A0 =A0 else > =A0 =A0 =A0 =A0 =A0 =A0 =A0 sc->reclaim_mode =3D RECLAIM_MODE_SINGLE | RE= CLAIM_MODE_ASYNC; > @@ -1349,10 +1350,6 @@ static inline bool > should_reclaim_stall(unsigned long nr_taken, > =A0 =A0 =A0 if (current_is_kswapd()) > =A0 =A0 =A0 =A0 =A0 =A0 =A0 return false; > > - =A0 =A0 =A0 /* Only stall on lumpy reclaim */ > - =A0 =A0 =A0 if (sc->reclaim_mode & RECLAIM_MODE_SINGLE) > - =A0 =A0 =A0 =A0 =A0 =A0 =A0 return false; > - Bah. It's this last hunk. Without this I can't reproduce the oops. With this hunk, the reset_reclaim_mode doesn't work and shrink_page_list is incorrectly called twice. So we're back to the original problem... --Andy -- 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/ . Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/ Don't email: email@kvack.org