ksm: fix mlockfreed to munlocked

2.6.33-rc1 commit 73848b4684e84a84cfd1555af78d41158f31e16b, adjusted
to include 31e855ea7173bdb0520f9684580423a9560f66e0's movement of
the unlock_page(oldpage), but omit other intervening cleanups.

When KSM merges an mlocked page, it has been forgetting to munlock it:
that's been left to free_page_mlock(), which reports it in /proc/vmstat
as unevictable_pgs_mlockfreed instead of unevictable_pgs_munlocked,
which indicates that such pages _might_ be left unevictable for long
after they should be evictable.  Call munlock_vma_page() to fix that.

Signed-off-by: Hugh Dickins <hugh.dickins@tiscali.co.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

3 files changed, 11 insertions, 10 deletions

diff --git a/mm/internal.h b/mm/internal.h
index 22ec8d2b0fb8..17bc0df273bb 100644
--- a/mm/internal.h
+++ b/mm/internal.h
@@ -107,9 +107,10 @@ static inline int is_mlocked_vma(struct vm_area_struct *vma, struct page *page)
 }
 
 /*
- * must be called with vma's mmap_sem held for read, and page locked.
+ * must be called with vma's mmap_sem held for read or write, and page locked.
  */
 extern void mlock_vma_page(struct page *page);
+extern void munlock_vma_page(struct page *page);
 
 /*
  * Clear the page's PageMlocked().  This can be useful in a situation where
diff --git a/mm/ksm.c b/mm/ksm.c
index 5575f8628fef..e9501f833374 100644
--- a/mm/ksm.c
+++ b/mm/ksm.c
@@ -34,6 +34,7 @@
 #include <linux/ksm.h>
 
 #include <asm/tlbflush.h>
+#include "internal.h"
 
 /*
  * A few notes about the KSM scanning process,
@@ -767,15 +768,14 @@ static int try_to_merge_one_page(struct vm_area_struct *vma,
 	 * ptes are necessarily already write-protected.  But in either
 	 * case, we need to lock and check page_count is not raised.
 	 */
-	if (write_protect_page(vma, oldpage, &orig_pte)) {
-		unlock_page(oldpage);
-		goto out_putpage;
-	}
-	unlock_page(oldpage);
-
-	if (pages_identical(oldpage, newpage))
+	if (write_protect_page(vma, oldpage, &orig_pte) == 0 &&
+	    pages_identical(oldpage, newpage))
 		err = replace_page(vma, oldpage, newpage, orig_pte);
 
+	if ((vma->vm_flags & VM_LOCKED) && !err)
+		munlock_vma_page(oldpage);
+
+	unlock_page(oldpage);
 out_putpage:
 	put_page(oldpage);
 	put_page(newpage);
diff --git a/mm/mlock.c b/mm/mlock.c
index bd6f0e466f6c..2e05c97b04f5 100644
--- a/mm/mlock.c
+++ b/mm/mlock.c
@@ -99,14 +99,14 @@ void mlock_vma_page(struct page *page)
  * not get another chance to clear PageMlocked.  If we successfully
  * isolate the page and try_to_munlock() detects other VM_LOCKED vmas
  * mapping the page, it will restore the PageMlocked state, unless the page
- * is mapped in a non-linear vma.  So, we go ahead and SetPageMlocked(),
+ * is mapped in a non-linear vma.  So, we go ahead and ClearPageMlocked(),
  * perhaps redundantly.
  * If we lose the isolation race, and the page is mapped by other VM_LOCKED
  * vmas, we'll detect this in vmscan--via try_to_munlock() or try_to_unmap()
  * either of which will restore the PageMlocked state by calling
  * mlock_vma_page() above, if it can grab the vma's mmap sem.
  */
-static void munlock_vma_page(struct page *page)
+void munlock_vma_page(struct page *page)
 {
 	BUG_ON(!PageLocked(page));