On Fri 25-03-16 12:22:37, Andrew Morton wrote:
[...]
> And can someone please explain mem_cgroup_select_victim_node() to me?
> How can we hit the "node = numa_node_id()" path? Only if
> memcg->scan_nodes is empty?
Yes, this seems to be the primary motivation.
mem_cgroup_may_update_nodemask might have seen all the pages on
unevictable LRU last time it checked something.
> is that even valid?
I suspect it is really rare but it seems possible
> The comment seems to have not much to do with the code?
I guess the comment tries to say that the code path is triggered when we
charge the page which happens _before_ it is added to the LRU list and
so last_scanned_node might contain the stale data. Would something like
the following be more clear?
---
diff --git a/mm/memcontrol.c b/mm/memcontrol.c
index 17a847c96618..cff095318950 100644
--- a/mm/memcontrol.c
+++ b/mm/memcontrol.c
@@ -1390,10 +1390,9 @@ int mem_cgroup_select_victim_node(struct mem_cgroup
*memcg)
node = next_node_in(node, memcg->scan_nodes);
/*
- * We call this when we hit limit, not when pages are added to LRU.
- * No LRU may hold pages because all pages are UNEVICTABLE or
- * memcg is too small and all pages are not on LRU. In that case,
- * we use curret node.
+ * mem_cgroup_may_update_nodemask might have seen no reclaimmable pages
+ * last time it really checked all the LRUs due to rate limiting.
+ * Fallback to the current node in that case for simplicity.
*/
if (unlikely(node == MAX_NUMNODES))
node = numa_node_id();
--
Michal Hocko
SUSE Labs
