Re: [PATCH v2] locking/rwsem: Add reader-owned state to the owner field

[Waiman Long]

On 05/09/2016 04:27 AM, Peter Zijlstra wrote:
> On Fri, May 06, 2016 at 08:20:24PM -0400, Waiman Long wrote:
> > @@ -391,9 +386,11 @@ static bool rwsem_optimistic_spin(struct rw_semaphore *sem)
> >   		 * When there's no owner, we might have preempted between the
> >   		 * owner acquiring the lock and setting the owner field. If
> >   		 * we're an RT task that will live-lock because we won't let
> > +		 * the owner complete. We also quit if the lock is owned by
> > +		 * readers.
> Maybe also note why we quit on readers.

Sure. Will do so.

> >   		*/
> > +		if (rwsem_is_reader_owned(owner) ||
> > +		   (!owner&&  (need_resched() || rt_task(current))))
> >   			break;
> >
> >   		/*
>
> > diff --git a/kernel/locking/rwsem.h b/kernel/locking/rwsem.h
> > index 870ed9a..d7fea18 100644
> > --- a/kernel/locking/rwsem.h
> > +++ b/kernel/locking/rwsem.h
> > @@ -1,3 +1,20 @@
> > +/*
> > + * The owner field of the rw_semaphore structure will be set to
> > + * RWSEM_READ_OWNED when a reader grabs the lock. A writer will clear
> > + * the owner field when it unlocks. A reader, on the other hand, will
> > + * not touch the owner field when it unlocks.
> > + *
> > + * In essence, the owner field now has the following 3 states:
> > + *  1) 0
> > + *     - lock is free or the owner hasn't set the field yet
> > + *  2) RWSEM_READER_OWNED
> > + *     - lock is currently or previously owned by readers (lock is free
> > + *       or not set by owner yet)
> > + *  3) Other non-zero value
> > + *     - a writer owns the lock
> > + */
> > +#define RWSEM_READER_OWNED	1UL
> #define RWSEM_READER_OWNED	((struct task_struct *)1UL)

Will make the change.

> > +
> >   #ifdef CONFIG_RWSEM_SPIN_ON_OWNER
> >   static inline void rwsem_set_owner(struct rw_semaphore *sem)
> >   {
> > @@ -9,6 +26,26 @@ static inline void rwsem_clear_owner(struct rw_semaphore *sem)
> >   	sem->owner = NULL;
> >   }
> >
> > +static inline void rwsem_set_reader_owned(struct rw_semaphore *sem)
> > +{
> > +	/*
> > +	 * We check the owner value first to make sure that we will only
> > +	 * do a write to the rwsem cacheline when it is really necessary
> > +	 * to minimize cacheline contention.
> > +	 */
> > +	if (sem->owner != (struct task_struct *)RWSEM_READER_OWNED)
> > +		sem->owner = (struct task_struct *)RWSEM_READER_OWNED;
> How much if anything did this optimization matter?

I hadn't run any performance test to verify the effective of this 
change. For a reader-heavy rwsem, this change should be able to save 
quite a lot of needless write to the rwsem cacheline.

> > +}
> > +
> > +static inline bool rwsem_is_writer_owned(struct task_struct *owner)
> > +{
> > +	return (unsigned long)owner>  RWSEM_READER_OWNED;
> > +}
> Tad too clever that; what does GCC generate if you write the obvious:
>
> 	return owner&&  owner != RWSEM_READER_OWNER;

You are right. GCC is intelligent enough to make the necessary 
optimization. I will revert it to this form which is more obvious.

> > +
> > +static inline bool rwsem_is_reader_owned(struct task_struct *owner)
> > +{
> > +	return owner == (struct task_struct *)RWSEM_READER_OWNED;
> > +}
> So I don't particularly like these names; they read like they take a
> rwsem as argument, but they don't.
>
> Would something like: rwsem_owner_is_{reader,writer}() make more sense?

Yes, these names look good to me.

Cheers,
Longman