[PATCH] fix a writeback race

Fixes a bug in generic_writepages() and its cut-n-paste-cousin,
mpage_writepages().

The code was clearing PageDirty and then baling out if it discovered
the page was nder writeback.  Which would cause the dirty bit to be
lost.

It's a very small window, but reversing the order so PageDirty is only
cleared when we know for-sure that IO will be started fixes it up.

fs/mpage.c

@@ -515,8 +515,8 @@ mpage_writepages(struct address_space *mapping,
 
 		lock_page(page);
 
-		if (page->mapping && TestClearPageDirty(page) &&
-					!PageWriteback(page)) {
+		if (page->mapping && !PageWriteback(page) &&
+					TestClearPageDirty(page)) {
 			/* FIXME: batch this up */
 			if (!PageActive(page) && PageLRU(page)) {
 				spin_lock(&pagemap_lru_lock);

mm/page-writeback.c

@@ -383,8 +383,8 @@ int generic_writepages(struct address_space *mapping, int *nr_to_write)
 		lock_page(page);
 
 		/* It may have been removed from swapcache: check ->mapping */
-		if (page->mapping && TestClearPageDirty(page) &&
-					!PageWriteback(page)) {
+		if (page->mapping && !PageWriteback(page) &&
+					TestClearPageDirty(page)) {
 			/* FIXME: batch this up */
 			if (!PageActive(page) && PageLRU(page)) {
 				spin_lock(&pagemap_lru_lock);