md/raid1,10: Remove use-after-free bug in make_request.
NeilBrown [Sat, 10 Sep 2011 07:21:23 +0000 (17:21 +1000)]
A single request to RAID1 or RAID10 might result in multiple
requests if there are known bad blocks that need to be avoided.

To detect if we need to submit another write request we test:
  if (sectors_handled < (bio->bi_size >> 9)) {

However this is after we call **_write_done() so the 'bio' no longer
belongs to us - the writes could have completed and the bio freed.

So move the **_write_done call until after the test against
bio->bi_size.

This addresses https://bugzilla.kernel.org/show_bug.cgi?id=41862

Reported-by: Bruno Wolff III <bruno@wolff.to>
Tested-by: Bruno Wolff III <bruno@wolff.to>
Signed-off-by: NeilBrown <neilb@suse.de>

drivers/md/raid1.c
drivers/md/raid10.c

index 32323f0..f4622dd 100644 (file)
@@ -1099,12 +1099,11 @@ read_again:
                bio_list_add(&conf->pending_bio_list, mbio);
                spin_unlock_irqrestore(&conf->device_lock, flags);
        }
-       r1_bio_write_done(r1_bio);
-
-       /* In case raid1d snuck in to freeze_array */
-       wake_up(&conf->wait_barrier);
-
+       /* Mustn't call r1_bio_write_done before this next test,
+        * as it could result in the bio being freed.
+        */
        if (sectors_handled < (bio->bi_size >> 9)) {
+               r1_bio_write_done(r1_bio);
                /* We need another r1_bio.  It has already been counted
                 * in bio->bi_phys_segments
                 */
@@ -1117,6 +1116,11 @@ read_again:
                goto retry_write;
        }
 
+       r1_bio_write_done(r1_bio);
+
+       /* In case raid1d snuck in to freeze_array */
+       wake_up(&conf->wait_barrier);
+
        if (do_sync || !bitmap || !plugged)
                md_wakeup_thread(mddev->thread);
 
index f6873fc..d7a8468 100644 (file)
@@ -1132,13 +1132,12 @@ retry_write:
                spin_unlock_irqrestore(&conf->device_lock, flags);
        }
 
-       /* Remove the bias on 'remaining' */
-       one_write_done(r10_bio);
-
-       /* In case raid10d snuck in to freeze_array */
-       wake_up(&conf->wait_barrier);
+       /* Don't remove the bias on 'remaining' (one_write_done) until
+        * after checking if we need to go around again.
+        */
 
        if (sectors_handled < (bio->bi_size >> 9)) {
+               one_write_done(r10_bio);
                /* We need another r10_bio.  It has already been counted
                 * in bio->bi_phys_segments.
                 */
@@ -1152,6 +1151,10 @@ retry_write:
                r10_bio->state = 0;
                goto retry_write;
        }
+       one_write_done(r10_bio);
+
+       /* In case raid10d snuck in to freeze_array */
+       wake_up(&conf->wait_barrier);
 
        if (do_sync || !mddev->bitmap || !plugged)
                md_wakeup_thread(mddev->thread);