Re: [PATCHv5 2/2] memory barrier: adding smp_mb__after_lock

[Mathieu Desnoyers <mathieu.desnoyers@polymtl.ca>]

* Ingo Molnar (mingo@elte.hu) wrote:
> 
> * Jiri Olsa <jolsa@redhat.com> wrote:
> 
> > +++ b/arch/x86/include/asm/spinlock.h
> > @@ -302,4 +302,7 @@ static inline void __raw_write_unlock(raw_rwlock_t *rw)
> >  #define _raw_read_relax(lock)	cpu_relax()
> >  #define _raw_write_relax(lock)	cpu_relax()
> >  
> > +/* The {read|write|spin}_lock() on x86 are full memory barriers. */
> > +#define smp_mb__after_lock() do { } while (0)
> 

Hm. Looking at http://lkml.org/lkml/2009/6/23/192, a very basic question
comes to my mind :

Why don't we create a read_lock without acquire semantic instead (e.g.
read_lock_nomb(), or something with a better name like __read_lock()) ?
On architectures where memory barriers are needed to provide the acquire
semantic, it would be faster to do :

__read_lock();
smp_mb();

than :

read_lock(); <- e.g. lwsync + isync or something like that
smp_mb(); <- full sync.

Second point : __add_wait_queue/waitqueue_active/wake_up_interruptible
would probably benefit from adding comments about their combined use
with other checks and how nice memory barriers are.

Mathieu


> Two small stylistic comments, please make this an inline function:
> 
> static inline void smp_mb__after_lock(void) { }
> #define smp_mb__after_lock
> 
> (untested)
> 
> > +/* The lock does not imply full memory barrier. */
> > +#ifndef smp_mb__after_lock
> > +#define smp_mb__after_lock() smp_mb()
> > +#endif
> 
> ditto.
> 
> 	Ingo

-- 
Mathieu Desnoyers
OpenPGP key fingerprint: 8CD5 52C3 8E3C 4140 715F  BA06 3F25 A8FE 3BAE 9A68