md/raid1: fix bug in read_balance introduced by hot-replace
NeilBrown [Tue, 3 Jul 2012 05:58:42 +0000 (15:58 +1000)]
When we added hot_replace we doubled the number of devices
that could be in a RAID1 array.  So we doubled how far read_balance
would search.  Unfortunately we didn't double the point at which
it looped back to the beginning - so it effectively loops over
all non-replacement disks twice.
This doesn't cause bad behaviour, but it pointless and means we
never read from replacement devices.

Signed-off-by: NeilBrown <neilb@suse.de>

drivers/md/raid1.c

index 39b2a8a..34b4665 100644 (file)
@@ -517,8 +517,8 @@ static int read_balance(struct r1conf *conf, struct r1bio *r1_bio, int *max_sect
                int bad_sectors;
 
                int disk = start_disk + i;
-               if (disk >= conf->raid_disks)
-                       disk -= conf->raid_disks;
+               if (disk >= conf->raid_disks * 2)
+                       disk -= conf->raid_disks * 2;
 
                rdev = rcu_dereference(conf->mirrors[disk].rdev);
                if (r1_bio->bios[disk] == IO_BLOCKED