On Mon, Jul 25, 2016 at 05:39:13PM +0900, Minchan Kim wrote:
> > @@ -1465,14 +1471,24 @@ static unsigned long isolate_lru_pages(unsigned
> > long nr_to_scan,
> > */
> > if (!list_empty(&pages_skipped)) {
> > int zid;
> > + unsigned long total_skipped = 0;
> >
> > - list_splice(&pages_skipped, src);
> > for (zid = 0; zid < MAX_NR_ZONES; zid++) {
> > if (!nr_skipped[zid])
> > continue;
> >
> > __count_zid_vm_events(PGSCAN_SKIP, zid,
> > nr_skipped[zid]);
> > + total_skipped += nr_skipped[zid];
> > }
> > +
> > + /*
> > + * Account skipped pages as a partial scan as the pgdat may be
> > + * close to unreclaimable. If the LRU list is empty, account
> > + * skipped pages as a full scan.
> > + */
>
> node-lru made OOM detection lengthy because a freeing of any zone will
> reset NR_PAGES_SCANNED easily so that it's hard to meet a situation
> pgdat_reclaimable returns *false*.
>
Your patch should go a long way to addressing that as it checks the zone
counters first before conducting the scan. Remember as well that the longer
detection of OOM only applies to zone-constrained allocations and there
is always the possibility that highmem shrinking of pages frees lowmem
memory if buffers are used.
--
Mel Gorman
SUSE Labs
