Re: [PATCH] locking/percpu_rwsem: Rewrite to not use rwsem

[Peter Zijlstra <peterz@infradead.org>]

On Wed, Oct 30, 2019 at 03:21:10PM +0100, Oleg Nesterov wrote:
> On 10/29, Peter Zijlstra wrote:
> >
> > I like that symmetry, but see below ...
> 
> ...
> 
> > > void __percpu_up_read(struct percpu_rw_semaphore *sem)
> > > {
> > > 	smp_mb();
> > >
> > > 	__this_cpu_dec(*sem->read_count);
> > >
> > 	preempt_enable();
> > > 	wake_up(&sem->waiters);
> > 	preempt_disable()
> >
> > and this (sadly) means there's a bunch of back-to-back
> > preempt_disable()+preempt_enable() calls.
> 
> Hmm. Where did these enable+disable come from?
> 
> 	void __percpu_up_read(struct percpu_rw_semaphore *sem)
> 	{
> 		smp_mb();
> 
> 		__this_cpu_dec(*sem->read_count);
> 
> 		wake_up(&sem->waiters);
> 	}
> 
> should work just fine?

Not on PREEMPT_RT, because wake_up() will take wait_queue_head::lock,
which is spin_lock_t and turns into a pi_mutex, which we cannot take
with preemption disabled.