On Tue, Jul 19, 2016 at 05:48:57PM +0100, Mel Gorman wrote: > On Wed, Jul 20, 2016 at 12:50:32AM +0900, Minchan Kim wrote: > > While I did stress test with hackbench, I got OOM message frequently > > which didn't ever happen in zone-lru. > > > > This one also showed pgdat going unreclaimable early. Have you tried any > of the three oom-related patches I sent to Joonsoo to see what impact, > if any, it had? Before the result, I want to say goal of this patch, again. Without per-zone lru stat, it's really hard to debug OOM problem in multiple zones system so regardless of solving the problem, we should add per-zone lru stat for debuggability of OOM which has been never perfect solution, ever. You sent 3 patches in that thread and first one was same I had applied when I found this problem firstly. It didn't solve the problem. So I tested last one diff --git a/mm/vmscan.c b/mm/vmscan.c index a6f31617a08c..0dc443b52228 100644 --- a/mm/vmscan.c +++ b/mm/vmscan.c @@ -1415,7 +1415,7 @@ static unsigned long isolate_lru_pages(unsigned long nr_to_scan, LIST_HEAD(pages_skipped); for (scan = 0; scan < nr_to_scan && nr_taken < nr_to_scan && - !list_empty(src); scan++) { + !list_empty(src);) { struct page *page; page = lru_to_page(src); @@ -1428,6 +1428,9 @@ static unsigned long isolate_lru_pages(unsigned long nr_to_scan, nr_skipped[page_zonenum(page)]++; continue; } +` + /* Pages skipped do not contribute to scan */ + scan++; switch (__isolate_lru_page(page, mode)) { case 0: The result is not OOM but hackbench stalls forever. When I parse vmstat for every 2sec, I found pgskip_high velocity is too high(i.e., 100000000 pages per 2 sec) while pgscan_direct and pgdeactiation is really low(i.e., 30 pages per 2 sec). The reason why it doesn't trigger OOM is a small amout of pages(i.e. 20 pages per sec) are freed so NR_PAGES_SCANNED is always reset to zero. > > -- > Mel Gorman > SUSE Labs -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@xxxxxxxxx. For more info on Linux MM, see: http://www.linux-mm.org/ . Don't email: <a href=mailto:"dont@xxxxxxxxx"> email@xxxxxxxxx </a>