There's no need to lock, we can just use an atomic_t.
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>