On 3/27/22 20:57, Waiman Long wrote:
Muchun Song found out there could be a race between list_lru_add()
and memcg_reparent_list_lru_node() causing the later function to miss
reparenting of a lru entry as shown below:
CPU0: CPU1:
list_lru_add()
spin_lock(&nlru->lock)
l = list_lru_from_kmem(memcg)
memcg_reparent_objcgs(memcg)
memcg_reparent_list_lrus(memcg)
memcg_reparent_list_lru()
memcg_reparent_list_lru_node()
if (!READ_ONCE(nlru->nr_items))
// Miss reparenting
return
// Assume 0->1
l->nr_items++
// Assume 0->1
nlru->nr_items++
Though it is not likely that a list_lru_node that has 0 item suddenly
has a newly added lru entry at the end of its life. The race is still
theoretically possible.
Adding a spin_is_locked() check will likely be enough for x86, but it
is less certain for other arches with a more relaxed memory semantics
like arcm64 and ppc. To avoid race, this patch moves the nr_items check
to within the lock critical section.
Fixes: 405cc51fc104 ("mm/list_lru: optimize memcg_reparent_list_lru_node()")
Sorry, I should have added
Reported-by: Muchun Song <songmuchun@xxxxxxxxxxxxx>
Signed-off-by: Waiman Long <longman@xxxxxxxxxx>
---
mm/list_lru.c | 12 ++++++------
1 file changed, 6 insertions(+), 6 deletions(-)
diff --git a/mm/list_lru.c b/mm/list_lru.c
index c669d87001a6..8aec8ebd5995 100644
--- a/mm/list_lru.c
+++ b/mm/list_lru.c
@@ -394,18 +394,18 @@ static void memcg_reparent_list_lru_node(struct list_lru *lru, int nid,
int dst_idx = dst_memcg->kmemcg_id;
struct list_lru_one *src, *dst;
- /*
- * If there is no lru entry in this nlru, we can skip it immediately.
- */
- if (!READ_ONCE(nlru->nr_items))
- return;
-
/*
* Since list_lru_{add,del} may be called under an IRQ-safe lock,
* we have to use IRQ-safe primitives here to avoid deadlock.
*/
spin_lock_irq(&nlru->lock);
+ /*
+ * If there is no lru entry in this nlru, we can skip it immediately.
+ */
+ if (!nlru->nr_items)
+ goto out;
+
src = list_lru_from_memcg_idx(lru, nid, src_idx);
if (!src)
goto out;
Cheers,
Longman
