mm: don't waste swap on locked pages
Hugh Dickins [Tue, 5 Feb 2008 06:29:23 +0000 (22:29 -0800)]
try_to_unmap always fails on a page found in a VM_LOCKED vma (unless
migrating), and recycles it back to the active list.  But if it's an
anonymous page, we've already allocated swap to it: just wasting swap.
Spot locked pages in page_referenced_one and treat them as referenced.

Signed-off-by: Hugh Dickins <hugh@veritas.com>
Tested-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Ethan Solomita <solo@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/rmap.c

index 0334c8f..57ad276 100644 (file)
--- a/mm/rmap.c
+++ b/mm/rmap.c
@@ -283,7 +283,10 @@ static int page_referenced_one(struct page *page,
        if (!pte)
                goto out;
 
-       if (ptep_clear_flush_young(vma, address, pte))
+       if (vma->vm_flags & VM_LOCKED) {
+               referenced++;
+               *mapcount = 1;  /* break early from loop */
+       } else if (ptep_clear_flush_young(vma, address, pte))
                referenced++;
 
        /* Pretend the page is referenced if the task has the