[RFC][PATCH][BUGFIX] fix rcu_read_lock in page migraton

[KAMEZAWA Hiroyuki]

This is a patch against the problme Shaohua rported.
Just an idea for fix the problem.
How do you think ? dummy vma is better ? (I don't like dummy vma.)

-Kame
==
In migration fallback path, write_page() or lock_page() will be called.
This causes sleep with holding rcu_read_lock().
For avoding that, just do rcu_lock if the page is Anon.(this is enough.)

Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@xxxxxxxxxxxxxx>


---
 mm/migrate.c |   11 +++++++++--
 1 file changed, 9 insertions(+), 2 deletions(-)

Index: linux-2.6.23-rc2-mm2/mm/migrate.c
===================================================================
--- linux-2.6.23-rc2-mm2.orig/mm/migrate.c
+++ linux-2.6.23-rc2-mm2/mm/migrate.c
@@ -611,6 +611,7 @@ static int unmap_and_move(new_page_t get
 	int rc = 0;
 	int *result = NULL;
 	struct page *newpage = get_new_page(page, private, &result);
+	int rcu_locked = 0;
 
 	if (!newpage)
 		return -ENOMEM;
@@ -636,8 +637,13 @@ static int unmap_and_move(new_page_t get
 	 * we cannot notice that anon_vma is freed while we migrates a page.
 	 * This rcu_read_lock() delays freeing anon_vma pointer until the end
 	 * of migration. File cache pages are no problem because of page_lock()
+	 * File Caches may use write_page() or lock_page() in migration, then,
+	 * just care Anon page here.
 	 */
-	rcu_read_lock();
+	if (PageAnon(page)) {
+		rcu_read_lock();
+		rcu_locked = 1;
+	}
 	/*
 	 * This is a corner case handling.
 	 * When a new swap-cache is read into, it is linked to LRU
@@ -656,7 +662,8 @@ static int unmap_and_move(new_page_t get
 	if (rc)
 		remove_migration_ptes(page, page);
 rcu_unlock:
-	rcu_read_unlock();
+	if (rcu_locked)
+		rcu_read_unlock();
 
 unlock:
 

-
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/