mm: fix inactive list balancing between NUMA nodes and cgroups
authorJohannes Weiner <hannes@cmpxchg.org>
Fri, 19 Apr 2019 00:50:34 +0000 (17:50 -0700)
committerLinus Torvalds <torvalds@linux-foundation.org>
Fri, 19 Apr 2019 16:46:05 +0000 (09:46 -0700)
commit3b991208b897f52507168374033771a984b947b1
treeb2a8c30bb8ee998e0d98437fb6d8bdc8e3436121
parent1a9f219157b22d0ffb340a9c5f431afd02cd2cf3
mm: fix inactive list balancing between NUMA nodes and cgroups

During !CONFIG_CGROUP reclaim, we expand the inactive list size if it's
thrashing on the node that is about to be reclaimed.  But when cgroups
are enabled, we suddenly ignore the node scope and use the cgroup scope
only.  The result is that pressure bleeds between NUMA nodes depending
on whether cgroups are merely compiled into Linux.  This behavioral
difference is unexpected and undesirable.

When the refault adaptivity of the inactive list was first introduced,
there were no statistics at the lruvec level - the intersection of node
and memcg - so it was better than nothing.

But now that we have that infrastructure, use lruvec_page_state() to
make the list balancing decision always NUMA aware.

[hannes@cmpxchg.org: fix bisection hole]
Link: http://lkml.kernel.org/r/20190417155241.GB23013@cmpxchg.org
Link: http://lkml.kernel.org/r/20190412144438.2645-1-hannes@cmpxchg.org
Fixes: 2a2e48854d70 ("mm: vmscan: fix IO/refault regression in cache workingset transition")
Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Reviewed-by: Shakeel Butt <shakeelb@google.com>
Cc: Roman Gushchin <guro@fb.com>
Cc: Michal Hocko <mhocko@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
mm/vmscan.c