[PATCH] Allow ext3 pages to be written back by VM pressure

When ext3 refiles journalled buffers for writeback, it is better to set the
page dirty as well as the buffer.  So the page will be written out by
VM pressure rather than by kjournald alone.

fs/jbd/transaction.c

@@ -1548,7 +1548,7 @@ void __journal_unfile_buffer(struct journal_head *jh)
 	__blist_del_buffer(list, jh);
 	jh->b_jlist = BJ_None;
 	if (test_and_clear_bit(BH_JBDDirty, &jh2bh(jh)->b_state))
-		set_buffer_dirty(jh2bh(jh));
+		mark_buffer_dirty(jh2bh(jh));	/* Expose it to the VM */
 }
 
 void journal_unfile_buffer(struct journal_head *jh)