As 'array_size' is a 'sector_t', it may overflow inappropriately when shifted
10 bits.  So We should cast it to a loff_t first.
There are two places with this problem, but the second (in update_raid_disks)
isn't needed so just remove it:
  The only personality that handles ->reshape currently is raid1,
  and it doesn't change the size of the array.
  When added for raid5/6, reshape again won't change the size of the array,
  at least not straight away.
  This code might be need for reshaping 'linear' but linear->shape,
  if implemented, should probably do the i_size_write itself.
Signed-off-by: Neil Brown <neilb@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
                bdev = bdget_disk(mddev->gendisk, 0);
                if (bdev) {
                        mutex_lock(&bdev->bd_inode->i_mutex);
-                       i_size_write(bdev->bd_inode, mddev->array_size << 10);
+                       i_size_write(bdev->bd_inode, (loff_t)mddev->array_size << 10);
                        mutex_unlock(&bdev->bd_inode->i_mutex);
                        bdput(bdev);
                }
        if (mddev->sync_thread)
                return -EBUSY;
        rv = mddev->pers->reshape(mddev, raid_disks);
-       if (!rv) {
-               struct block_device *bdev;
-
-               bdev = bdget_disk(mddev->gendisk, 0);
-               if (bdev) {
-                       mutex_lock(&bdev->bd_inode->i_mutex);
-                       i_size_write(bdev->bd_inode, mddev->array_size << 10);
-                       mutex_unlock(&bdev->bd_inode->i_mutex);
-                       bdput(bdev);
-               }
-       }
        return rv;
 }