The patch titled
Subject: mm, memcg: prevent memory.low load/store tearing
has been removed from the -mm tree. Its filename was
mm-memcg-prevent-memorylow-load-store-tearing.patch
This patch was dropped because it was merged into mainline or a subsystem tree
------------------------------------------------------
From: Chris Down <chris@xxxxxxxxxxxxxx>
Subject: mm, memcg: prevent memory.low load/store tearing
This can be set concurrently with reads, which may cause the wrong value
to be propagated.
Link: http://lkml.kernel.org/r/448206f44b0fa7be9dad2ca2601d2bcb2c0b7844.1584034301.git.chris@xxxxxxxxxxxxxx
Signed-off-by: Chris Down <chris@xxxxxxxxxxxxxx>
Acked-by: Michal Hocko <mhocko@xxxxxxxx>
Cc: Johannes Weiner <hannes@xxxxxxxxxxx>
Cc: Roman Gushchin <guro@xxxxxx>
Cc: Tejun Heo <tj@xxxxxxxxxx>
Signed-off-by: Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx>
---
mm/page_counter.c | 8 +++++---
1 file changed, 5 insertions(+), 3 deletions(-)
--- a/mm/page_counter.c~mm-memcg-prevent-memorylow-load-store-tearing
+++ a/mm/page_counter.c
@@ -17,6 +17,7 @@ static void propagate_protected_usage(st
unsigned long usage)
{
unsigned long protected, old_protected;
+ unsigned long low;
long delta;
if (!c->parent)
@@ -30,8 +31,9 @@ static void propagate_protected_usage(st
atomic_long_add(delta, &c->parent->children_min_usage);
}
- if (c->low || atomic_long_read(&c->low_usage)) {
- protected = min(usage, c->low);
+ low = READ_ONCE(c->low);
+ if (low || atomic_long_read(&c->low_usage)) {
+ protected = min(usage, low);
old_protected = atomic_long_xchg(&c->low_usage, protected);
delta = protected - old_protected;
if (delta)
@@ -222,7 +224,7 @@ void page_counter_set_low(struct page_co
{
struct page_counter *c;
- counter->low = nr_pages;
+ WRITE_ONCE(counter->low, nr_pages);
for (c = counter; c; c = c->parent)
propagate_protected_usage(c, atomic_long_read(&c->usage));
_
Patches currently in -mm which might be from chris@xxxxxxxxxxxxxx are
mm-memcg-bypass-high-reclaim-iteration-for-cgroup-hierarchy-root.patch
