[RFC] memory hotremove patch take 2 [10/10] (retry swap-in page)

[Yasunori Goto]

There is a race condition between swap-in and unmap_and_move().
When swap-in occur, page_mapped might be not set yet.
So, unmap_and_move() gives up at once, and tries later.



Signed-off-by: Yasunori Goto <y-goto@xxxxxxxxxxxxxx>


 mm/migrate.c |    5 +++++
 1 files changed, 5 insertions(+)

Index: current_test/mm/migrate.c
===================================================================
--- current_test.orig/mm/migrate.c	2007-05-08 15:08:09.000000000 +0900
+++ current_test/mm/migrate.c	2007-05-08 15:08:09.000000000 +0900
@@ -670,6 +670,11 @@ static int unmap_and_move(new_page_t get
 		/* hold this anon_vma until remove_migration_ptes() finishes */
 		anon_vma_hold(page);
 	}
+
+	if (PageSwapCache(page) && !page_mapped(page))
+		/* swap in now. try lator*/
+		goto unlock;
+
 	/*
 	 * Establish migration ptes or remove ptes
 	 */

-- 
Yasunori Goto 


-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/