memcg: remove unnecessary thp check in page stat accounting
KAMEZAWA Hiroyuki [Wed, 21 Mar 2012 23:34:20 +0000 (16:34 -0700)]
Commit e94c8a9cbce1 ("memcg: make mem_cgroup_split_huge_fixup() more
efficient") removed move_lock_page_cgroup().  So we do not have to check
PageTransHuge in mem_cgroup_update_page_stat() and fallback into the
locked accounting because both move_account() and thp split are done
with compound_lock so they cannot race.

The race between update vs.  move is protected by mem_cgroup_stealed.

PageTransHuge pages shouldn't appear in this code path currently because
we are tracking only file pages at the moment but later we are planning
to track also other pages (e.g.  mlocked ones).

Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Reviewed-by: Acked-by: Michal Hocko <mhocko@suse.cz>
Cc: David Rientjes <rientjes@google.com>
Acked-by: Ying Han<yinghan@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/memcontrol.c

index 43a9ade..69af5d5 100644 (file)
@@ -1898,7 +1898,7 @@ void mem_cgroup_update_page_stat(struct page *page,
        if (unlikely(!memcg || !PageCgroupUsed(pc)))
                goto out;
        /* pc->mem_cgroup is unstable ? */
-       if (unlikely(mem_cgroup_stealed(memcg)) || PageTransHuge(page)) {
+       if (unlikely(mem_cgroup_stealed(memcg))) {
                /* take a lock against to access pc->mem_cgroup */
                move_lock_page_cgroup(pc, &flags);
                need_unlock = true;