block: fix an integer overflow in logical block size
authorMikulas Patocka <mpatocka@redhat.com>
Wed, 15 Jan 2020 13:35:25 +0000 (08:35 -0500)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Thu, 23 Jan 2020 07:22:32 +0000 (08:22 +0100)
commit6eed26e35cfda2fa6e1a6fcaf5115c6eb587b566
tree39c72098e6122d45105a762bb66a5ff56d765dd3
parentd904824b87dbe09f5d9ac0c53d2beac6be2f68d4
block: fix an integer overflow in logical block size

commit ad6bf88a6c19a39fb3b0045d78ea880325dfcf15 upstream.

Logical block size has type unsigned short. That means that it can be at
most 32768. However, there are architectures that can run with 64k pages
(for example arm64) and on these architectures, it may be possible to
create block devices with 64k block size.

For exmaple (run this on an architecture with 64k pages):

Mount will fail with this error because it tries to read the superblock using 2-sector
access:
  device-mapper: writecache: I/O is not aligned, sector 2, size 1024, block size 65536
  EXT4-fs (dm-0): unable to read superblock

This patch changes the logical block size from unsigned short to unsigned
int to avoid the overflow.

Cc: stable@vger.kernel.org
Reviewed-by: Martin K. Petersen <martin.petersen@oracle.com>
Reviewed-by: Ming Lei <ming.lei@redhat.com>
Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
block/blk-settings.c
drivers/md/dm-snap-persistent.c
drivers/md/raid0.c
include/linux/blkdev.h