Re: [PATCH 5/5] locking/percpu-rwsem: Remove the embedded rwsem

[Peter Zijlstra]

On Mon, Nov 18, 2019 at 03:19:35PM -0800, Davidlohr Bueso wrote:
> Similarly, afaict we can get rid of __percpu_up_read() and put the
> slowpath all into percpu_up_read(). Also explicitly mention the
> single task nature of the writer (which is a better comment for
> the rcuwait_wake_up()).

> static inline void percpu_down_read(struct percpu_rw_semaphore *sem)
> {
> @@ -103,10 +102,23 @@ static inline void percpu_up_read(struct percpu_rw_semaphore *sem)
> 	/*
> 	 * Same as in percpu_down_read().
> 	 */
> +	if (likely(rcu_sync_is_idle(&sem->rss))) {
> 		__this_cpu_dec(*sem->read_count);
> +		goto done;
> +	}
> +
> +	/*
> +	 * slowpath; reader will only ever wake a single blocked writer.
> +	 */
> +	smp_mb(); /* B matches C */
> +	/*
> +	 * In other words, if they see our decrement (presumably to
> +	 * aggregate zero, as that is the only time it matters) they
> +	 * will also see our critical section.
> +	 */
> +	__this_cpu_dec(*sem->read_count);
> +	rcuwait_wake_up(&sem->writer);
> +done:
> 	preempt_enable();
> }

Let me write that as a normal if () { } else { }.

But yes, that's small enough I suppose.