Re: [PATCH -mm 07/16] second chance replacement for anonymous pages

[KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>]

[-- Attachment #1: Type: text/plain, Size: 624 bytes --]

> @@ -1129,7 +1141,11 @@ static unsigned long shrink_list(enum lr
>  {
>  	int file = is_file_lru(lru);
>  
> -	if (lru == LRU_ACTIVE_ANON || lru == LRU_ACTIVE_FILE) {
> +	if (lru == LRU_ACTIVE_FILE) {
> +		shrink_active_list(nr_to_scan, zone, sc, priority, file);
> +		return 0;
> +	}
> +	if (lru == LRU_ACTIVE_ANON && inactive_anon_low(zone)) {
>  		shrink_active_list(nr_to_scan, zone, sc, priority, file);
>  		return 0;
>  	}

I made memcgroup noreclaim infrastructure today.
and, I found slightly odd behaviour.

this condition increase OOM, because prevent active -> inactive moving 
even though non global reclaim.


[-- Attachment #2: rvr-07.1-kosaki-memcg-shrink_list.patch --]
[-- Type: application/octet-stream, Size: 772 bytes --]

When memcgroup is used, OOM killer is happend very freqently.
because inactive_anon_low() prevent move active lru to inactive lru.


Signed-off-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>

---
 mm/vmscan.c |    4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

Index: b/mm/vmscan.c
===================================================================
--- a/mm/vmscan.c
+++ b/mm/vmscan.c
@@ -1145,7 +1145,9 @@ static unsigned long shrink_list(enum lr
 		shrink_active_list(nr_to_scan, zone, sc, priority, file);
 		return 0;
 	}
-	if (lru == LRU_ACTIVE_ANON && inactive_anon_low(zone)) {
+
+	if (lru == LRU_ACTIVE_ANON &&
+	    (!scan_global_lru(sc) || inactive_anon_low(zone))) {
 		shrink_active_list(nr_to_scan, zone, sc, priority, file);
 		return 0;
 	}