[2.6.22.y] {01/17} - do_anonymous_page-race - series for stable kernel

[Oliver Pinter (Pintér Olivér)]

-- 
Thanks,
Oliver
commit 8327e0f191341cc32fb89bf4d7bee7c2524ae4e0
Author: Andrea Arcangeli <andrea@xxxxxxx>
Date:   Mon Jan 28 20:52:29 2008 +0100

    Subject: Race condition in userspace testcase
    References: 46948, LTC11574
    
    
    Additional Comment #103 From Andrea Arcangeli 2004-10-15 19:41
    the last patch I attached is the safest I believe.
    
    I'm not sure if a lock_unlock or lock_unlock is always guaranteed to happen
    after the I/O, and this makes sure no race can happen anymore.

diff --git a/fs/bio.c b/fs/bio.c
index 093345f..984d589 100644
--- a/fs/bio.c
+++ b/fs/bio.c
@@ -1028,6 +1028,15 @@ void bio_endio(struct bio *bio, unsigned int bytes_done, int error)
 	bio->bi_size -= bytes_done;
 	bio->bi_sector += (bytes_done >> 9);
 
+	if (bio_data_dir(bio) == READ)
+		/*
+		 * If the current cpu has written to the page by hand
+		 * without dma, we must enforce ordering to be sure
+		 * this written data will be visible before we expose
+		 * the page contents to other cpus (for example with
+		 * a set_pte).
+		 */
+		smp_wmb();
 	if (bio->bi_end_io)
 		bio->bi_end_io(bio, bytes_done, error);
 }