| 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
| ||
|
Message-ID: <20180424105238.6quxqngkmtok7ghn@esperanza>
Date: Tue, 24 Apr 2018 13:56:20 +0300
From: Vladimir Davydov <vdavydov.dev@...il.com>
To: Kirill Tkhai <ktkhai@...tuozzo.com>
Cc: akpm@...ux-foundation.org, shakeelb@...gle.com,
viro@...iv.linux.org.uk, hannes@...xchg.org, mhocko@...nel.org,
tglx@...utronix.de, pombredanne@...b.com, stummala@...eaurora.org,
gregkh@...uxfoundation.org, sfr@...b.auug.org.au, guro@...com,
mka@...omium.org, penguin-kernel@...ove.SAKURA.ne.jp,
chris@...is-wilson.co.uk, longman@...hat.com, minchan@...nel.org,
hillf.zj@...baba-inc.com, ying.huang@...el.com,
mgorman@...hsingularity.net, jbacik@...com, linux@...ck-us.net,
linux-kernel@...r.kernel.org, linux-mm@...ck.org,
willy@...radead.org, lirongqing@...du.com, aryabinin@...tuozzo.com
Subject: Re: [PATCH v2 12/12] mm: Clear shrinker bit if there are no objects
related to memcg
On Mon, Apr 23, 2018 at 01:01:08PM +0300, Kirill Tkhai wrote:
> On 22.04.2018 21:21, Vladimir Davydov wrote:
> > On Tue, Apr 17, 2018 at 09:54:51PM +0300, Kirill Tkhai wrote:
> >> To avoid further unneed calls of do_shrink_slab()
> >> for shrinkers, which already do not have any charged
> >> objects in a memcg, their bits have to be cleared.
> >>
> >> This patch introduces a lockless mechanism to do that
> >> without races without parallel list lru add. After
> >> do_shrink_slab() returns SHRINK_EMPTY the first time,
> >> we clear the bit and call it once again. Then we restore
> >> the bit, if the new return value is different.
> >>
> >> Note, that single smp_mb__after_atomic() in shrink_slab_memcg()
> >> covers two situations:
> >>
> >> 1)list_lru_add() shrink_slab_memcg
> >> list_add_tail() for_each_set_bit() <--- read bit
> >> do_shrink_slab() <--- missed list update (no barrier)
> >> <MB> <MB>
> >> set_bit() do_shrink_slab() <--- seen list update
> >>
> >> This situation, when the first do_shrink_slab() sees set bit,
> >> but it doesn't see list update (i.e., race with the first element
> >> queueing), is rare. So we don't add <MB> before the first call
> >> of do_shrink_slab() instead of this to do not slow down generic
> >> case. Also, it's need the second call as seen in below in (2).
> >>
> >> 2)list_lru_add() shrink_slab_memcg()
> >> list_add_tail() ...
> >> set_bit() ...
> >> ... for_each_set_bit()
> >> do_shrink_slab() do_shrink_slab()
> >> clear_bit() ...
> >> ... ...
> >> list_lru_add() ...
> >> list_add_tail() clear_bit()
> >> <MB> <MB>
> >> set_bit() do_shrink_slab()
> >>
> >> The barriers guarantees, the second do_shrink_slab()
> >> in the right side task sees list update if really
> >> cleared the bit. This case is drawn in the code comment.
> >>
> >> [Results/performance of the patchset]
> >>
> >> After the whole patchset applied the below test shows signify
> >> increase of performance:
> >>
> >> $echo 1 > /sys/fs/cgroup/memory/memory.use_hierarchy
> >> $mkdir /sys/fs/cgroup/memory/ct
> >> $echo 4000M > /sys/fs/cgroup/memory/ct/memory.kmem.limit_in_bytes
> >> $for i in `seq 0 4000`; do mkdir /sys/fs/cgroup/memory/ct/$i; echo $$ > /sys/fs/cgroup/memory/ct/$i/cgroup.procs; mkdir -p s/$i; mount -t tmpfs $i s/$i; touch s/$i/file; done
> >>
> >> Then, 4 sequential calls of drop caches:
> >> $time echo 3 > /proc/sys/vm/drop_caches
> >>
> >> 1)Before:
> >> 0.00user 8.99system 0:08.99elapsed 99%CPU
> >> 0.00user 5.97system 0:05.97elapsed 100%CPU
> >> 0.00user 5.97system 0:05.97elapsed 100%CPU
> >> 0.00user 5.85system 0:05.85elapsed 100%CPU
> >>
> >> 2)After
> >> 0.00user 1.11system 0:01.12elapsed 99%CPU
> >> 0.00user 0.00system 0:00.00elapsed 100%CPU
> >> 0.00user 0.00system 0:00.00elapsed 100%CPU
> >> 0.00user 0.00system 0:00.00elapsed 100%CPU
> >>
> >> Even if we round 0:00.00 up to 0:00.01, the results shows
> >> the performance increases at least in 585 times.
> >>
> >> Signed-off-by: Kirill Tkhai <ktkhai@...tuozzo.com>
> >> ---
> >> include/linux/memcontrol.h | 2 ++
> >> mm/vmscan.c | 19 +++++++++++++++++--
> >> 2 files changed, 19 insertions(+), 2 deletions(-)
> >>
> >> diff --git a/include/linux/memcontrol.h b/include/linux/memcontrol.h
> >> index e1c1fa8e417a..1c5c68550e2f 100644
> >> --- a/include/linux/memcontrol.h
> >> +++ b/include/linux/memcontrol.h
> >> @@ -1245,6 +1245,8 @@ static inline void set_shrinker_bit(struct mem_cgroup *memcg, int nid, int nr)
> >>
> >> rcu_read_lock();
> >> map = SHRINKERS_MAP(memcg, nid);
> >> + /* Pairs with smp mb in shrink_slab() */
> >> + smp_mb__before_atomic();
> >> set_bit(nr, map->map);
> >> rcu_read_unlock();
> >> }
> >> diff --git a/mm/vmscan.c b/mm/vmscan.c
> >> index 3be9b4d81c13..a8733bc5377b 100644
> >> --- a/mm/vmscan.c
> >> +++ b/mm/vmscan.c
> >> @@ -579,8 +579,23 @@ static unsigned long shrink_slab_memcg(gfp_t gfp_mask, int nid,
> >> }
> >>
> >> ret = do_shrink_slab(&sc, shrinker, priority);
> >> - if (ret == SHRINK_EMPTY)
> >> - ret = 0;
> >> + if (ret == SHRINK_EMPTY) {
> >> + clear_bit(i, map->map);
> >> + /*
> >> + * Pairs with mb in set_shrinker_bit():
> >> + *
> >> + * list_lru_add() shrink_slab_memcg()
> >> + * list_add_tail() clear_bit()
> >> + * <MB> <MB>
> >> + * set_bit() do_shrink_slab()
> >> + */
> >> + smp_mb__after_atomic();
> >> + ret = do_shrink_slab(&sc, shrinker, priority);
> >> + if (ret == SHRINK_EMPTY)
> >> + ret = 0;
> >> + else
> >> + set_shrinker_bit(memcg, nid, i);
> >> + }
> >
> > This is mind-boggling. Are there any alternatives? For instance, can't
> > we clear the bit in list_lru_del, when we hold the list lock?
>
> Since a single shrinker may iterate over several lru lists, we can't do that.
> Otherwise, we would have to probe another shrinker's lru list from a lru list,
> which became empty in list_lru_del().
>
> The solution I suggested, is generic, and it does not depend on low-level
> structure type, used by shrinker. This even doesn't have to be a lru list.
Fair enough. I guess this is the best way to go after all. Please try to
improve the comment so that it isn't just a pure diagram.
Also, please prefix all memcg-related function names (such as
set_shrinker_bit) with memcg_ (or mem_cgroup_) in this and all
other patches.
Powered by blists - more mailing lists