dm ioctl: fix nested locking in table_clear() to remove deadlock concern
authorMike Snitzer <snitzer@kernel.org>
Mon, 17 Apr 2023 15:59:56 +0000 (11:59 -0400)
committerMike Snitzer <snitzer@kernel.org>
Mon, 17 Apr 2023 16:15:55 +0000 (12:15 -0400)
commit3d32aaa7e66d5c1479a3c31d6c2c5d45dd0d3b89
treed1f9683569456ae774b090da35c7a32da34cc118
parentf7995089c508a5a11b0491c7b348d5c07217a4e8
dm ioctl: fix nested locking in table_clear() to remove deadlock concern

syzkaller found the following problematic rwsem locking (with write
lock already held):

 down_read+0x9d/0x450 kernel/locking/rwsem.c:1509
 dm_get_inactive_table+0x2b/0xc0 drivers/md/dm-ioctl.c:773
 __dev_status+0x4fd/0x7c0 drivers/md/dm-ioctl.c:844
 table_clear+0x197/0x280 drivers/md/dm-ioctl.c:1537

In table_clear, it first acquires a write lock
https://elixir.bootlin.com/linux/v6.2/source/drivers/md/dm-ioctl.c#L1520
down_write(&_hash_lock);

Then before the lock is released at L1539, there is a path shown above:
table_clear -> __dev_status -> dm_get_inactive_table ->  down_read
https://elixir.bootlin.com/linux/v6.2/source/drivers/md/dm-ioctl.c#L773
down_read(&_hash_lock);

It tries to acquire the same read lock again, resulting in the deadlock
problem.

Fix this by moving table_clear()'s __dev_status() call to after its
up_write(&_hash_lock);

Cc: stable@vger.kernel.org
Reported-by: Zheng Zhang <zheng.zhang@email.ucr.edu>
Signed-off-by: Mike Snitzer <snitzer@kernel.org>
drivers/md/dm-ioctl.c