From: Mel Gorman Subject: Re: [PATCH 4/4] mm: vmscan: If kswapd has been running too long, allow it to sleep Date: Tue, 17 May 2011 11:38:40 +0100 Message-ID: <20110517103840.GL5279@suse.de> References: <1305295404-12129-1-git-send-email-mgorman@suse.de> <1305295404-12129-5-git-send-email-mgorman@suse.de> <4DCFAA80.7040109@jp.fujitsu.com> <1305519711.4806.7.camel@mulgrave.site> <20110516084558.GE5279@suse.de> <20110516102753.GF5279@suse.de> Mime-Version: 1.0 Content-Type: text/plain; charset=iso-8859-15 Content-Transfer-Encoding: quoted-printable Cc: James Bottomley , KOSAKI Motohiro , akpm@linux-foundation.org, colin.king@canonical.com, raghu.prabhu13@gmail.com, jack@suse.cz, chris.mason@oracle.com, cl@linux.com, penberg@kernel.org, riel@redhat.com, hannes@cmpxchg.org, linux-fsdevel@vger.kernel.org, linux-mm@kvack.org, linux-kernel@vger.kernel.org, linux-ext4@vger.kernel.org To: Minchan Kim Return-path: Content-Disposition: inline In-Reply-To: Sender: owner-linux-mm@kvack.org List-Id: linux-ext4.vger.kernel.org On Tue, May 17, 2011 at 08:50:44AM +0900, Minchan Kim wrote: > On Mon, May 16, 2011 at 7:27 PM, Mel Gorman wrote: > > On Mon, May 16, 2011 at 05:58:59PM +0900, Minchan Kim wrote: > >> On Mon, May 16, 2011 at 5:45 PM, Mel Gorman wrote: > >> > On Mon, May 16, 2011 at 02:04:00PM +0900, Minchan Kim wrote: > >> >> On Mon, May 16, 2011 at 1:21 PM, James Bottomley > >> >> wrote: > >> >> > On Sun, 2011-05-15 at 19:27 +0900, KOSAKI Motohiro wrote: > >> >> >> (2011/05/13 23:03), Mel Gorman wrote: > >> >> >> > Under constant allocation pressure, kswapd can be in the sit= uation where > >> >> >> > sleeping_prematurely() will always return true even if kswap= d has been > >> >> >> > running a long time. Check if kswapd needs to be scheduled. > >> >> >> > > >> >> >> > Signed-off-by: Mel Gorman > >> >> >> > --- > >> >> >> > =A0 mm/vmscan.c | =A0 =A04 ++++ > >> >> >> > =A0 1 files changed, 4 insertions(+), 0 deletions(-) > >> >> >> > > >> >> >> > diff --git a/mm/vmscan.c b/mm/vmscan.c > >> >> >> > index af24d1e..4d24828 100644 > >> >> >> > --- a/mm/vmscan.c > >> >> >> > +++ b/mm/vmscan.c > >> >> >> > @@ -2251,6 +2251,10 @@ static bool sleeping_prematurely(pg_d= ata_t *pgdat, int order, long remaining, > >> >> >> > =A0 =A0 unsigned long balanced =3D 0; > >> >> >> > =A0 =A0 bool all_zones_ok =3D true; > >> >> >> > > >> >> >> > + =A0 /* If kswapd has been running too long, just sleep */ > >> >> >> > + =A0 if (need_resched()) > >> >> >> > + =A0 =A0 =A0 =A0 =A0 return false; > >> >> >> > + > >> >> >> > >> >> >> Hmm... I don't like this patch so much. because this code does > >> >> >> > >> >> >> - don't sleep if kswapd got context switch at shrink_inactive_= list > >> >> > > >> >> > This isn't entirely true: =A0need_resched() will be false, so w= e'll follow > >> >> > the normal path for determining whether to sleep or not, in eff= ect > >> >> > leaving the current behaviour unchanged. > >> >> > > >> >> >> - sleep if kswapd didn't > >> >> > > >> >> > This also isn't entirely true: whether need_resched() is true a= t this > >> >> > point depends on a whole lot more that whether we did a context= switch > >> >> > in shrink_inactive. It mostly depends on how long we've been ru= nning > >> >> > without giving up the CPU. =A0Generally that will mean we've be= en round > >> >> > the shrinker loop hundreds to thousands of times without sleepi= ng. > >> >> > > >> >> >> It seems to be semi random behavior. > >> >> > > >> >> > Well, we have to do something. =A0Chris Mason first suspected t= he hang was > >> >> > a kswapd rescheduling problem a while ago. =A0We tried putting > >> >> > cond_rescheds() in several places in the vmscan code, but to no= avail. > >> >> > >> >> Is it a result of =A0test with patch of Hannes(ie, !pgdat_balance= d)? > >> >> > >> >> If it isn't, it would be nop regardless of putting cond_reshed at= vmscan.c. > >> >> Because, although we complete zone balancing, kswapd doesn't slee= p as > >> >> pgdat_balance returns wrong result. And at last VM calls > >> >> balance_pgdat. In this case, balance_pgdat returns without any wo= rk as > >> >> kswap couldn't find zones which have not enough free pages and go= to > >> >> out. kswapd could repeat this work infinitely. So you don't have = a > >> >> chance to call cond_resched. > >> >> > >> >> But if your test was with Hanne's patch, I am very curious how co= me > >> >> kswapd consumes CPU a lot. > >> >> > >> >> > The need_resched() in sleeping_prematurely() seems to be about = the best > >> >> > option. =A0The other option might be just to put a cond_resched= () in > >> >> > kswapd_try_to_sleep(), but that will really have about the same= effect. > >> >> > >> >> I don't oppose it but before that, I think we have to know why ks= wapd > >> >> consumes CPU a lot although we applied Hannes' patch. > >> >> > >> > > >> > Because it's still possible for processes to allocate pages at the= same > >> > rate kswapd is freeing them leading to a situation where kswapd do= es not > >> > consider the zone balanced for prolonged periods of time. > >> > >> We have cond_resched in shrink_page_list, shrink_slab and balance_pg= dat. > >> So I think kswapd can be scheduled out although it's scheduled in > >> after a short time as task scheduled also need page reclaim. Althoug= h > >> all task in system need reclaim, kswapd cpu 99% consumption is a > >> natural result, I think. > >> Do I miss something? > >> > > > > Lets see; > > > > shrink_page_list() only applies if inactive pages were isolated > > =A0 =A0 =A0 =A0which in turn may not happen if all_unreclaimable is s= et in > > =A0 =A0 =A0 =A0shrink_zones(). If for whatver reason, all_unreclaimab= le is > > =A0 =A0 =A0 =A0set on all zones, we can miss calling cond_resched(). > > > > shrink_slab only applies if we are reclaiming slab pages. If the firs= t > > =A0 =A0 =A0 =A0shrinker returns -1, we do not call cond_resched(). If= that > > =A0 =A0 =A0 =A0first shrinker is dcache and __GFP_FS is not set, dire= ct > > =A0 =A0 =A0 =A0reclaimers will not shrink at all. However, if there a= re > > =A0 =A0 =A0 =A0enough of them running or if one of the other shrinker= s > > =A0 =A0 =A0 =A0is running for a very long time, kswapd could be starv= ed > > =A0 =A0 =A0 =A0acquiring the shrinker_rwsem and never reaching the > > =A0 =A0 =A0 =A0cond_resched(). >=20 > Don't we have to move cond_resched? >=20 > diff --git a/mm/vmscan.c b/mm/vmscan.c > index 292582c..633e761 100644 > --- a/mm/vmscan.c > +++ b/mm/vmscan.c > @@ -231,8 +231,10 @@ unsigned long shrink_slab(struct shrink_control *s= hrink, > if (scanned =3D=3D 0) > scanned =3D SWAP_CLUSTER_MAX; >=20 > - if (!down_read_trylock(&shrinker_rwsem)) > - return 1; /* Assume we'll be able to shrink next = time */ > + if (!down_read_trylock(&shrinker_rwsem)) { > + ret =3D 1; > + goto out; /* Assume we'll be able to shrink next time *= / > + } >=20 > list_for_each_entry(shrinker, &shrinker_list, list) { > unsigned long long delta; > @@ -280,12 +282,14 @@ unsigned long shrink_slab(struct shrink_control *= shrink, > count_vm_events(SLABS_SCANNED, this_scan); > total_scan -=3D this_scan; >=20 > - cond_resched(); > } >=20 > shrinker->nr +=3D total_scan; > + cond_resched(); > } > up_read(&shrinker_rwsem); > +out: > + cond_resched(); > return ret; > } >=20 This makes some sense for the exit path but if one or more of the shrinkers takes a very long time without sleeping (extremely long list searches for example) then kswapd will not call cond_resched() between shrinkers and still consume a lot of CPU. > > > > balance_pgdat() only calls cond_resched if the zones are not > > =A0 =A0 =A0 =A0balanced. For a high-order allocation that is balanced= , it > > =A0 =A0 =A0 =A0checks order-0 again. During that window, order-0 migh= t have > > =A0 =A0 =A0 =A0become unbalanced so it loops again for order-0 and re= turns > > =A0 =A0 =A0 =A0that was reclaiming for order-0 to kswapd(). It can th= en find > > =A0 =A0 =A0 =A0that a caller has rewoken kswapd for a high-order and = re-enters > > =A0 =A0 =A0 =A0balance_pgdat() without ever have called cond_resched(= ). >=20 > If kswapd reclaims order-o followed by high order, it would have a > chance to call cond_resched in shrink_page_list. But if all zones are > all_unreclaimable is set, balance_pgdat could return any work. Okay. > It does make sense. > By your scenario, someone wakes up kswapd with higher order, again. > So re-enters balance_pgdat without ever have called cond_resched. > But if someone wakes up higher order again, we can't have a chance to > call kswapd_try_to_sleep. So your patch effect would be nop, too. >=20 > It would be better to put cond_resched after balance_pgdat? >=20 Which will leave kswapd runnable instead of going to sleep but guarantees a scheduling point. Lets see if the problem is that cond_resched is being missed although if this was the case then patch 4 would truly be a no-op but Colin has already reported that patch 1 on its own didn't fix his problem. If the problem is sandybridge-specific where kswapd remains runnable and consuming large amounts of CPU in turbo mode then we know that there are other cond_resched() decisions that will need to be revisited. Colin or James, would you be willing to test with patch 1 from this series and Minchan's patch below? Thanks. > diff --git a/mm/vmscan.c b/mm/vmscan.c > index 292582c..61c45d0 100644 > --- a/mm/vmscan.c > +++ b/mm/vmscan.c > @@ -2753,6 +2753,7 @@ static int kswapd(void *p) > if (!ret) { > trace_mm_vmscan_kswapd_wake(pgdat->node_id, ord= er); > order =3D balance_pgdat(pgdat, order, &classzon= e_idx); > + cond_resched(); > } > } > return 0; >=20 > > > > While it appears unlikely, there are bad conditions which can result > > in cond_resched() being avoided. >=20 > > > > -- > > Mel Gorman > > SUSE Labs > > >=20 >=20 >=20 > --=20 > Kind regards, > Minchan Kim --=20 Mel Gorman SUSE Labs -- 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