On Tue, May 15, 2018 at 01:52:06PM -0400, Waiman Long wrote:
> On 05/15/2018 01:48 PM, Peter Zijlstra wrote:
> > On Tue, May 15, 2018 at 01:38:03PM -0400, Waiman Long wrote:
> >> diff --git a/kernel/locking/rwsem.c b/kernel/locking/rwsem.c
> >> index 30465a2..b7208e1 100644
> >> --- a/kernel/locking/rwsem.c
> >> +++ b/kernel/locking/rwsem.c
> >> @@ -130,7 +130,8 @@ void up_read(struct rw_semaphore *sem)
> >> void up_write(struct rw_semaphore *sem)
> >> {
> >> rwsem_release(&sem->dep_map, 1, _RET_IP_);
> >> - DEBUG_RWSEMS_WARN_ON(sem->owner != current);
> >> + DEBUG_RWSEMS_WARN_ON((sem->owner != current) &&
> >> + !rwsem_has_anonymous_owner(sem->owner));
> > Why? Don't we always do percpu_rwsem_acquire() before up?
> >
>
> This is to allow an unlock to happen in the unknown owner state. Yes, it
> is not necessary to fix the percpu-rwsem problem. It is there just in
> case a rwsem will be released in that state in the future.
Let's not allow that until there's a very good reason for it.
