md/raid5: Fix possible confusion when multiple write errors occur.
authorNeilBrown <neilb@suse.de>
Mon, 6 Jan 2014 02:19:42 +0000 (13:19 +1100)
committerBen Hutchings <ben@decadent.org.uk>
Sat, 15 Feb 2014 19:20:16 +0000 (19:20 +0000)
commit 1cc03eb93245e63b0b7a7832165efdc52e25b4e6 upstream.

commit 5d8c71f9e5fbdd95650be00294d238e27a363b5c
    md: raid5 crash during degradation

Fixed a crash in an overly simplistic way which could leave
R5_WriteError or R5_MadeGood set in the stripe cache for devices
for which it is no longer relevant.
When those devices are removed and spares added the flags are still
set and can cause incorrect behaviour.

commit 14a75d3e07c784c004b4b44b34af996b8e4ac453
    md/raid5: preferentially read from replacement device if possible.

Fixed the same bug if a more effective way, so we can now revert
the original commit.

Reported-and-tested-by: Alexander Lyakas <alex.bolshoy@gmail.com>
Fixes: 5d8c71f9e5fbdd95650be00294d238e27a363b5c
Signed-off-by: NeilBrown <neilb@suse.de>
[bwh: Backported to 3.2: adjust context]
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
drivers/md/raid5.c

index 26ef63a..fb67833 100644 (file)
@@ -3084,7 +3084,7 @@ static void analyse_stripe(struct stripe_head *sh, struct stripe_head_state *s)
                         */
                        set_bit(R5_Insync, &dev->flags);
 
-               if (rdev && test_bit(R5_WriteError, &dev->flags)) {
+               if (test_bit(R5_WriteError, &dev->flags)) {
                        clear_bit(R5_Insync, &dev->flags);
                        if (!test_bit(Faulty, &rdev->flags)) {
                                s->handle_bad_blocks = 1;
@@ -3092,7 +3092,7 @@ static void analyse_stripe(struct stripe_head *sh, struct stripe_head_state *s)
                        } else
                                clear_bit(R5_WriteError, &dev->flags);
                }
-               if (rdev && test_bit(R5_MadeGood, &dev->flags)) {
+               if (test_bit(R5_MadeGood, &dev->flags)) {
                        if (!test_bit(Faulty, &rdev->flags)) {
                                s->handle_bad_blocks = 1;
                                atomic_inc(&rdev->nr_pending);