drbd: fix decoding of bitmap vli rle for device sizes > 64 TB
authorLars Ellenberg <lars.ellenberg@linbit.com>
Wed, 23 Oct 2013 08:59:18 +0000 (10:59 +0200)
committerJens Axboe <axboe@kernel.dk>
Fri, 8 Nov 2013 16:10:28 +0000 (09:10 -0700)
Symptoms: disconnect after bitmap exchange due to
bitmap overflow (e:49731075554) while decoding bm RLE packet

In the decoding step of the variable length integer run length encoding
there was potentially an uncatched bitshift by wordsize (variable >> 64).

The result of which is "undefined" :(
(only "sometimes" the result is the desired 0)

Fix: don't do any bit shift magic for shift == 64, just assign.

Signed-off-by: Philipp Reisner <philipp.reisner@linbit.com>
Signed-off-by: Lars Ellenberg <lars.ellenberg@linbit.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
drivers/block/drbd/drbd_receiver.c

index 12c59eb3b1274ef7f7c9fe0ac5f854eea10f722a..6fa6673b36b396765b58142e8e8abcdc4beaae05 100644 (file)
@@ -4125,7 +4125,11 @@ recv_bm_rle_bits(struct drbd_conf *mdev,
                                (unsigned int)bs.buf_len);
                        return -EIO;
                }
-               look_ahead >>= bits;
+               /* if we consumed all 64 bits, assign 0; >> 64 is "undefined"; */
+               if (likely(bits < 64))
+                       look_ahead >>= bits;
+               else
+                       look_ahead = 0;
                have -= bits;
 
                bits = bitstream_get_bits(&bs, &tmp, 64 - have);