vmscan,tmpfs: treat used once pages on tmpfs as used once

When a page has PG_referenced, shrink_page_list() discards it only if it
is not dirty.  This rule works fine if the backing filesystem is a regular
one.  PG_dirty is a good signal that the page was used recently because
the flusher threads clean pages periodically.  In addition, page writeback
is costlier than simple page discard.

However, when a page is on tmpfs this heuristic doesn't work because
flusher threads don't write back tmpfs pages.  Consequently tmpfs pages
always rotate around the lru twice at least and adds unnecessary lru
churn.  Simple tmpfs streaming io shouldn't cause large anonymous page
swap-out.

Remove this unncessary reclaim bonus of tmpfs pages.

Signed-off-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Cc: Hugh Dickins <hughd@google.com>
Reviewed-by: Johannes Weiner <hannes@cmpxchg.org>
Reviewed-by: Rik van Riel <riel@redhat.com>
Cc: Minchan Kim <minchan.kim@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

diff --git a/mm/vmscan.c b/mm/vmscan.c
index 30fd658bb2897076d4fec53a5b0435d095daaec6..b8a6fdc2131258db15321021375117cae27cec0d 100644 (file)
--- a/mm/vmscan.c
+++ b/mm/vmscan.c
@@ -651,7 +651,7 @@ static enum page_references page_check_references(struct page *page,
        }
 
        /* Reclaim if clean, defer dirty pages to writeback */
-       if (referenced_page)
+       if (referenced_page && !PageSwapBacked(page))
                return PAGEREF_RECLAIM_CLEAN;
 
        return PAGEREF_RECLAIM;