* David Goulet ([email protected]) wrote:
> In order to not pollute the userspace namespace for application
> using liburcu or any lib/apps linked with urcu, this patch if the
> first of three major refactor for naming convention.
> 
> The cmm_ prefix is a short name for Concurrent Memory Model and was
> suggested by Mathieu Desnoyers and Paul E. Mckenney. Every memory
> primitives such as mb, wmb, rmb, and so on are renamed.

David,

You are my hero! :-)

All three patches committed and pushed.

Version 0.5.0 is on its way.

Thanks!

Mathieu

> 
> Signed-off-by: David Goulet <[email protected]>
> ---
>  compat_futex.c             |    4 +-
>  tests/api_gcc.h            |    8 ++--
>  tests/api_ppc.h            |    8 ++--
>  tests/api_x86.h            |    8 ++--
>  tests/rcutorture.h         |   18 ++++----
>  tests/test_mutex.c         |    6 +-
>  tests/test_perthreadlock.c |    6 +-
>  tests/test_qsbr.c          |    6 +-
>  tests/test_qsbr_gc.c       |    6 +-
>  tests/test_rwlock.c        |    6 +-
>  tests/test_urcu.c          |    6 +-
>  tests/test_urcu_assign.c   |    6 +-
>  tests/test_urcu_bp.c       |    6 +-
>  tests/test_urcu_defer.c    |    6 +-
>  tests/test_urcu_gc.c       |    6 +-
>  tests/test_urcu_lfq.c      |    6 +-
>  tests/test_urcu_lfs.c      |    6 +-
>  tests/test_urcu_wfq.c      |    6 +-
>  tests/test_urcu_wfs.c      |    6 +-
>  urcu-bp-static.h           |    8 ++--
>  urcu-bp.c                  |   16 ++++----
>  urcu-defer.c               |   14 +++---
>  urcu-pointer-static.h      |    8 ++--
>  urcu-pointer.c             |    6 +-
>  urcu-qsbr-static.h         |   16 ++++----
>  urcu-qsbr.c                |   28 ++++++------
>  urcu-static.h              |   12 +++---
>  urcu.c                     |   36 ++++++++--------
>  urcu/arch_alpha.h          |    6 +-
>  urcu/arch_armv7l.h         |    2 +-
>  urcu/arch_generic.h        |   98 
> ++++++++++++++++++++++----------------------
>  urcu/arch_ppc.h            |    6 +-
>  urcu/arch_s390.h           |    2 +-
>  urcu/arch_sparc64.h        |    6 +-
>  urcu/arch_x86.h            |   14 +++---
>  urcu/compiler.h            |    2 +-
>  urcu/rcuhlist.h            |    2 +-
>  urcu/rculist.h             |    2 +-
>  urcu/system.h              |    8 ++--
>  39 files changed, 211 insertions(+), 211 deletions(-)
> 
> diff --git a/compat_futex.c b/compat_futex.c
> index 61bc6c3..9dfcfeb 100644
> --- a/compat_futex.c
> +++ b/compat_futex.c
> @@ -55,7 +55,7 @@ int compat_futex_noasync(int *uaddr, int op, int val,
>       /*
>        * memory barriers to serialize with the previous uaddr modification.
>        */
> -     smp_mb();
> +     cmm_smp_mb();
>  
>       ret = pthread_mutex_lock(&compat_futex_lock);
>       assert(!ret);
> @@ -100,7 +100,7 @@ int compat_futex_async(int *uaddr, int op, int val,
>       /*
>        * Ensure previous memory operations on uaddr have completed.
>        */
> -     smp_mb();
> +     cmm_smp_mb();
>  
>       switch (op) {
>       case FUTEX_WAIT:
> diff --git a/tests/api_gcc.h b/tests/api_gcc.h
> index be437cc..b23110d 100644
> --- a/tests/api_gcc.h
> +++ b/tests/api_gcc.h
> @@ -251,10 +251,10 @@ cmpxchg(volatile long *ptr, long oldval, long newval)
>  #define atomic_dec_return(v)  (atomic_sub_return(1,v))
>  
>  /* Atomic operations are already serializing on x86 */
> -#define smp_mb__before_atomic_dec()  barrier()
> -#define smp_mb__after_atomic_dec()   barrier()
> -#define smp_mb__before_atomic_inc()  barrier()
> -#define smp_mb__after_atomic_inc()   barrier()
> +#define smp_mb__before_atomic_dec()  cmm_barrier()
> +#define smp_mb__after_atomic_dec()   cmm_barrier()
> +#define smp_mb__before_atomic_inc()  cmm_barrier()
> +#define smp_mb__after_atomic_inc()   cmm_barrier()
>  
>  #endif //0 /* duplicate with arch_atomic.h */
>  
> diff --git a/tests/api_ppc.h b/tests/api_ppc.h
> index 6548077..9773500 100644
> --- a/tests/api_ppc.h
> +++ b/tests/api_ppc.h
> @@ -616,10 +616,10 @@ static __inline__ int atomic_add_unless(atomic_t *v, 
> int a, int u)
>  #define atomic_dec_return(v)  (atomic_sub_return(1,v))
>  
>  /* Atomic operations are already serializing on x86 */
> -#define smp_mb__before_atomic_dec()  smp_mb()
> -#define smp_mb__after_atomic_dec()   smp_mb()
> -#define smp_mb__before_atomic_inc()  smp_mb()
> -#define smp_mb__after_atomic_inc()   smp_mb()
> +#define smp_mb__before_atomic_dec()  cmm_smp_mb()
> +#define smp_mb__after_atomic_dec()   cmm_smp_mb()
> +#define smp_mb__before_atomic_inc()  cmm_smp_mb()
> +#define smp_mb__after_atomic_inc()   cmm_smp_mb()
>  
>  #endif //0 /* duplicate with arch_atomic.h */
>  
> diff --git a/tests/api_x86.h b/tests/api_x86.h
> index cdd4667..fe00a35 100644
> --- a/tests/api_x86.h
> +++ b/tests/api_x86.h
> @@ -312,10 +312,10 @@ __asm__ __volatile__(LOCK_PREFIX "orl %0,%1" \
>  : : "r" (mask),"m" (*(addr)) : "memory")
>  
>  /* Atomic operations are already serializing on x86 */
> -#define smp_mb__before_atomic_dec()  barrier()
> -#define smp_mb__after_atomic_dec()   barrier()
> -#define smp_mb__before_atomic_inc()  barrier()
> -#define smp_mb__after_atomic_inc()   barrier()
> +#define smp_mb__before_atomic_dec()  cmm_barrier()
> +#define smp_mb__after_atomic_dec()   cmm_barrier()
> +#define smp_mb__before_atomic_inc()  cmm_barrier()
> +#define smp_mb__after_atomic_inc()   cmm_barrier()
>  
>  #endif //0
>  
> diff --git a/tests/rcutorture.h b/tests/rcutorture.h
> index 181547a..4192bd0 100644
> --- a/tests/rcutorture.h
> +++ b/tests/rcutorture.h
> @@ -170,15 +170,15 @@ void perftestrun(int nthreads, int nreaders, int 
> nupdaters)
>       int t;
>       int duration = 1;
>  
> -     smp_mb();
> +     cmm_smp_mb();
>       while (uatomic_read(&nthreadsrunning) < nthreads)
>               poll(NULL, 0, 1);
>       goflag = GOFLAG_RUN;
> -     smp_mb();
> +     cmm_smp_mb();
>       sleep(duration);
> -     smp_mb();
> +     cmm_smp_mb();
>       goflag = GOFLAG_STOP;
> -     smp_mb();
> +     cmm_smp_mb();
>       wait_all_threads();
>       for_each_thread(t) {
>               n_reads += per_thread(n_reads_pt, t);
> @@ -309,7 +309,7 @@ void *rcu_update_stress_test(void *arg)
>                       i = 0;
>               p = &rcu_stress_array[i];
>               p->mbtest = 0;
> -             smp_mb();
> +             cmm_smp_mb();
>               p->pipe_count = 0;
>               p->mbtest = 1;
>               rcu_assign_pointer(rcu_stress_current, p);
> @@ -355,13 +355,13 @@ void stresstest(int nreaders)
>       create_thread(rcu_update_stress_test, NULL);
>       for (i = 0; i < 5; i++)
>               create_thread(rcu_fake_update_stress_test, NULL);
> -     smp_mb();
> +     cmm_smp_mb();
>       goflag = GOFLAG_RUN;
> -     smp_mb();
> +     cmm_smp_mb();
>       sleep(10);
> -     smp_mb();
> +     cmm_smp_mb();
>       goflag = GOFLAG_STOP;
> -     smp_mb();
> +     cmm_smp_mb();
>       wait_all_threads();
>       for_each_thread(t)
>               n_reads += per_thread(n_reads_pt, t);
> diff --git a/tests/test_mutex.c b/tests/test_mutex.c
> index 7b2d1a5..7dd1e31 100644
> --- a/tests/test_mutex.c
> +++ b/tests/test_mutex.c
> @@ -229,7 +229,7 @@ void *thr_writer(void *data)
>       while (!test_go)
>       {
>       }
> -     smp_mb();
> +     cmm_smp_mb();
>  
>       for (;;) {
>               pthread_mutex_lock(&lock);
> @@ -278,7 +278,7 @@ int main(int argc, char **argv)
>               show_usage(argc, argv);
>               return -1;
>       }
> -     smp_mb();
> +     cmm_smp_mb();
>  
>       err = sscanf(argv[1], "%u", &nr_readers);
>       if (err != 1) {
> @@ -376,7 +376,7 @@ int main(int argc, char **argv)
>                       exit(1);
>       }
>  
> -     smp_mb();
> +     cmm_smp_mb();
>  
>       test_go = 1;
>  
> diff --git a/tests/test_perthreadlock.c b/tests/test_perthreadlock.c
> index 7862f62..9de8ced 100644
> --- a/tests/test_perthreadlock.c
> +++ b/tests/test_perthreadlock.c
> @@ -234,7 +234,7 @@ void *thr_writer(void *data)
>       while (!test_go)
>       {
>       }
> -     smp_mb();
> +     cmm_smp_mb();
>  
>       for (;;) {
>               for (tidx = 0; tidx < nr_readers; tidx++) {
> @@ -287,7 +287,7 @@ int main(int argc, char **argv)
>               show_usage(argc, argv);
>               return -1;
>       }
> -     smp_mb();
> +     cmm_smp_mb();
>  
>       err = sscanf(argv[1], "%u", &nr_readers);
>       if (err != 1) {
> @@ -386,7 +386,7 @@ int main(int argc, char **argv)
>                       exit(1);
>       }
>  
> -     smp_mb();
> +     cmm_smp_mb();
>  
>       test_go = 1;
>  
> diff --git a/tests/test_qsbr.c b/tests/test_qsbr.c
> index 0101907..4f19c72 100644
> --- a/tests/test_qsbr.c
> +++ b/tests/test_qsbr.c
> @@ -230,7 +230,7 @@ void *thr_reader(void *_count)
>       while (!test_go)
>       {
>       }
> -     smp_mb();
> +     cmm_smp_mb();
>  
>       for (;;) {
>               rcu_read_lock();
> @@ -275,7 +275,7 @@ void *thr_writer(void *_count)
>       while (!test_go)
>       {
>       }
> -     smp_mb();
> +     cmm_smp_mb();
>  
>       for (;;) {
>               new = test_array_alloc();
> @@ -424,7 +424,7 @@ int main(int argc, char **argv)
>                       exit(1);
>       }
>  
> -     smp_mb();
> +     cmm_smp_mb();
>  
>       test_go = 1;
>  
> diff --git a/tests/test_qsbr_gc.c b/tests/test_qsbr_gc.c
> index 2b4fa7a..875fd36 100644
> --- a/tests/test_qsbr_gc.c
> +++ b/tests/test_qsbr_gc.c
> @@ -203,7 +203,7 @@ void *thr_reader(void *_count)
>       while (!test_go)
>       {
>       }
> -     smp_mb();
> +     cmm_smp_mb();
>  
>       for (;;) {
>               _rcu_read_lock();
> @@ -279,7 +279,7 @@ void *thr_writer(void *data)
>       while (!test_go)
>       {
>       }
> -     smp_mb();
> +     cmm_smp_mb();
>  
>       for (;;) {
>  #ifndef TEST_LOCAL_GC
> @@ -444,7 +444,7 @@ int main(int argc, char **argv)
>                       exit(1);
>       }
>  
> -     smp_mb();
> +     cmm_smp_mb();
>  
>       test_go = 1;
>  
> diff --git a/tests/test_rwlock.c b/tests/test_rwlock.c
> index 1c65fca..445ce95 100644
> --- a/tests/test_rwlock.c
> +++ b/tests/test_rwlock.c
> @@ -225,7 +225,7 @@ void *thr_writer(void *_count)
>       while (!test_go)
>       {
>       }
> -     smp_mb();
> +     cmm_smp_mb();
>  
>       for (;;) {
>               pthread_rwlock_wrlock(&lock);
> @@ -274,7 +274,7 @@ int main(int argc, char **argv)
>               show_usage(argc, argv);
>               return -1;
>       }
> -     smp_mb();
> +     cmm_smp_mb();
>  
>       err = sscanf(argv[1], "%u", &nr_readers);
>       if (err != 1) {
> @@ -370,7 +370,7 @@ int main(int argc, char **argv)
>                       exit(1);
>       }
>  
> -     smp_mb();
> +     cmm_smp_mb();
>  
>       test_go = 1;
>  
> diff --git a/tests/test_urcu.c b/tests/test_urcu.c
> index 5096415..eeea7f5 100644
> --- a/tests/test_urcu.c
> +++ b/tests/test_urcu.c
> @@ -231,7 +231,7 @@ void *thr_reader(void *_count)
>       while (!test_go)
>       {
>       }
> -     smp_mb();
> +     cmm_smp_mb();
>  
>       for (;;) {
>               rcu_read_lock();
> @@ -273,7 +273,7 @@ void *thr_writer(void *_count)
>       while (!test_go)
>       {
>       }
> -     smp_mb();
> +     cmm_smp_mb();
>  
>       for (;;) {
>               new = test_array_alloc();
> @@ -421,7 +421,7 @@ int main(int argc, char **argv)
>                       exit(1);
>       }
>  
> -     smp_mb();
> +     cmm_smp_mb();
>  
>       test_go = 1;
>  
> diff --git a/tests/test_urcu_assign.c b/tests/test_urcu_assign.c
> index b1298a4..481cdd6 100644
> --- a/tests/test_urcu_assign.c
> +++ b/tests/test_urcu_assign.c
> @@ -231,7 +231,7 @@ void *thr_reader(void *_count)
>       while (!test_go)
>       {
>       }
> -     smp_mb();
> +     cmm_smp_mb();
>  
>       for (;;) {
>               rcu_read_lock();
> @@ -269,7 +269,7 @@ void *thr_writer(void *_count)
>       while (!test_go)
>       {
>       }
> -     smp_mb();
> +     cmm_smp_mb();
>  
>       for (;;) {
>               new = test_array_alloc();
> @@ -420,7 +420,7 @@ int main(int argc, char **argv)
>                       exit(1);
>       }
>  
> -     smp_mb();
> +     cmm_smp_mb();
>  
>       test_go = 1;
>  
> diff --git a/tests/test_urcu_bp.c b/tests/test_urcu_bp.c
> index 239e843..e831824 100644
> --- a/tests/test_urcu_bp.c
> +++ b/tests/test_urcu_bp.c
> @@ -231,7 +231,7 @@ void *thr_reader(void *_count)
>       while (!test_go)
>       {
>       }
> -     smp_mb();
> +     cmm_smp_mb();
>  
>       for (;;) {
>               rcu_read_lock();
> @@ -269,7 +269,7 @@ void *thr_writer(void *_count)
>       while (!test_go)
>       {
>       }
> -     smp_mb();
> +     cmm_smp_mb();
>  
>       for (;;) {
>               new = test_array_alloc();
> @@ -417,7 +417,7 @@ int main(int argc, char **argv)
>                       exit(1);
>       }
>  
> -     smp_mb();
> +     cmm_smp_mb();
>  
>       test_go = 1;
>  
> diff --git a/tests/test_urcu_defer.c b/tests/test_urcu_defer.c
> index 6fae313..2cbb041 100644
> --- a/tests/test_urcu_defer.c
> +++ b/tests/test_urcu_defer.c
> @@ -199,7 +199,7 @@ void *thr_reader(void *_count)
>       while (!test_go)
>       {
>       }
> -     smp_mb();
> +     cmm_smp_mb();
>  
>       for (;;) {
>               rcu_read_lock();
> @@ -247,7 +247,7 @@ void *thr_writer(void *data)
>       while (!test_go)
>       {
>       }
> -     smp_mb();
> +     cmm_smp_mb();
>  
>       for (;;) {
>               new = malloc(sizeof(*new));
> @@ -400,7 +400,7 @@ int main(int argc, char **argv)
>                       exit(1);
>       }
>  
> -     smp_mb();
> +     cmm_smp_mb();
>  
>       test_go = 1;
>  
> diff --git a/tests/test_urcu_gc.c b/tests/test_urcu_gc.c
> index 2c944a5..ddafb87 100644
> --- a/tests/test_urcu_gc.c
> +++ b/tests/test_urcu_gc.c
> @@ -207,7 +207,7 @@ void *thr_reader(void *_count)
>       while (!test_go)
>       {
>       }
> -     smp_mb();
> +     cmm_smp_mb();
>  
>       for (;;) {
>               rcu_read_lock();
> @@ -280,7 +280,7 @@ void *thr_writer(void *data)
>       while (!test_go)
>       {
>       }
> -     smp_mb();
> +     cmm_smp_mb();
>  
>       for (;;) {
>  #ifndef TEST_LOCAL_GC
> @@ -444,7 +444,7 @@ int main(int argc, char **argv)
>                       exit(1);
>       }
>  
> -     smp_mb();
> +     cmm_smp_mb();
>  
>       test_go = 1;
>  
> diff --git a/tests/test_urcu_lfq.c b/tests/test_urcu_lfq.c
> index 87c53b6..901bcae 100644
> --- a/tests/test_urcu_lfq.c
> +++ b/tests/test_urcu_lfq.c
> @@ -170,7 +170,7 @@ void *thr_enqueuer(void *_count)
>       while (!test_go)
>       {
>       }
> -     smp_mb();
> +     cmm_smp_mb();
>  
>       for (;;) {
>               struct rcu_lfq_node *node = malloc(sizeof(*node));
> @@ -223,7 +223,7 @@ void *thr_dequeuer(void *_count)
>       while (!test_go)
>       {
>       }
> -     smp_mb();
> +     cmm_smp_mb();
>  
>       for (;;) {
>               struct rcu_lfq_node *node = rcu_lfq_dequeue(&q,
> @@ -380,7 +380,7 @@ int main(int argc, char **argv)
>                       exit(1);
>       }
>  
> -     smp_mb();
> +     cmm_smp_mb();
>  
>       test_go = 1;
>  
> diff --git a/tests/test_urcu_lfs.c b/tests/test_urcu_lfs.c
> index 8f881e7..8249eba 100644
> --- a/tests/test_urcu_lfs.c
> +++ b/tests/test_urcu_lfs.c
> @@ -170,7 +170,7 @@ void *thr_enqueuer(void *_count)
>       while (!test_go)
>       {
>       }
> -     smp_mb();
> +     cmm_smp_mb();
>  
>       for (;;) {
>               struct rcu_lfs_node *node = malloc(sizeof(*node));
> @@ -215,7 +215,7 @@ void *thr_dequeuer(void *_count)
>       while (!test_go)
>       {
>       }
> -     smp_mb();
> +     cmm_smp_mb();
>  
>       for (;;) {
>               struct rcu_lfs_node *node = rcu_lfs_pop(&s);
> @@ -365,7 +365,7 @@ int main(int argc, char **argv)
>                       exit(1);
>       }
>  
> -     smp_mb();
> +     cmm_smp_mb();
>  
>       test_go = 1;
>  
> diff --git a/tests/test_urcu_wfq.c b/tests/test_urcu_wfq.c
> index b0629d2..d446e47 100644
> --- a/tests/test_urcu_wfq.c
> +++ b/tests/test_urcu_wfq.c
> @@ -167,7 +167,7 @@ void *thr_enqueuer(void *_count)
>       while (!test_go)
>       {
>       }
> -     smp_mb();
> +     cmm_smp_mb();
>  
>       for (;;) {
>               struct wfq_node *node = malloc(sizeof(*node));
> @@ -207,7 +207,7 @@ void *thr_dequeuer(void *_count)
>       while (!test_go)
>       {
>       }
> -     smp_mb();
> +     cmm_smp_mb();
>  
>       for (;;) {
>               struct wfq_node *node = wfq_dequeue_blocking(&q);
> @@ -354,7 +354,7 @@ int main(int argc, char **argv)
>                       exit(1);
>       }
>  
> -     smp_mb();
> +     cmm_smp_mb();
>  
>       test_go = 1;
>  
> diff --git a/tests/test_urcu_wfs.c b/tests/test_urcu_wfs.c
> index f50b140..294e955 100644
> --- a/tests/test_urcu_wfs.c
> +++ b/tests/test_urcu_wfs.c
> @@ -167,7 +167,7 @@ void *thr_enqueuer(void *_count)
>       while (!test_go)
>       {
>       }
> -     smp_mb();
> +     cmm_smp_mb();
>  
>       for (;;) {
>               struct wfs_node *node = malloc(sizeof(*node));
> @@ -207,7 +207,7 @@ void *thr_dequeuer(void *_count)
>       while (!test_go)
>       {
>       }
> -     smp_mb();
> +     cmm_smp_mb();
>  
>       for (;;) {
>               struct wfs_node *node = wfs_pop_blocking(&s);
> @@ -354,7 +354,7 @@ int main(int argc, char **argv)
>                       exit(1);
>       }
>  
> -     smp_mb();
> +     cmm_smp_mb();
>  
>       test_go = 1;
>  
> diff --git a/urcu-bp-static.h b/urcu-bp-static.h
> index 6d7e3c8..394476b 100644
> --- a/urcu-bp-static.h
> +++ b/urcu-bp-static.h
> @@ -175,7 +175,7 @@ static inline void _rcu_read_lock(void)
>       if (unlikely(!rcu_reader))
>               rcu_bp_register();
>  
> -     barrier();      /* Ensure the compiler does not reorder us with mutex */
> +     cmm_barrier();  /* Ensure the compiler does not reorder us with mutex */
>       tmp = rcu_reader->ctr;
>       /*
>        * rcu_gp_ctr is
> @@ -187,7 +187,7 @@ static inline void _rcu_read_lock(void)
>                * Set active readers count for outermost nesting level before
>                * accessing the pointer.
>                */
> -             smp_mb();
> +             cmm_smp_mb();
>       } else {
>               _STORE_SHARED(rcu_reader->ctr, tmp + RCU_GP_COUNT);
>       }
> @@ -198,9 +198,9 @@ static inline void _rcu_read_unlock(void)
>       /*
>        * Finish using rcu before decrementing the pointer.
>        */
> -     smp_mb();
> +     cmm_smp_mb();
>       _STORE_SHARED(rcu_reader->ctr, rcu_reader->ctr - RCU_GP_COUNT);
> -     barrier();      /* Ensure the compiler does not reorder us with mutex */
> +     cmm_barrier();  /* Ensure the compiler does not reorder us with mutex */
>  }
>  
>  #ifdef __cplusplus 
> diff --git a/urcu-bp.c b/urcu-bp.c
> index 61d42fa..33352c2 100644
> --- a/urcu-bp.c
> +++ b/urcu-bp.c
> @@ -96,9 +96,9 @@ static void mutex_lock(pthread_mutex_t *mutex)
>                       exit(-1);
>               }
>               if (rcu_reader.need_mb) {
> -                     smp_mb();
> +                     cmm_smp_mb();
>                       rcu_reader.need_mb = 0;
> -                     smp_mb();
> +                     cmm_smp_mb();
>               }
>               poll(NULL,0,10);
>       }
> @@ -133,11 +133,11 @@ void update_counter_and_wait(void)
>        */
>  
>       /*
> -      * Adding a smp_mb() which is _not_ formally required, but makes the
> +      * Adding a cmm_smp_mb() which is _not_ formally required, but makes the
>        * model easier to understand. It does not have a big performance impact
>        * anyway, given this is the write-side.
>        */
> -     smp_mb();
> +     cmm_smp_mb();
>  
>       /*
>        * Wait for each thread rcu_reader.ctr count to become 0.
> @@ -180,7 +180,7 @@ void synchronize_rcu(void)
>       /* All threads should read qparity before accessing data structure
>        * where new ptr points to. */
>       /* Write new ptr before changing the qparity */
> -     smp_mb();
> +     cmm_smp_mb();
>  
>       /* Remove old registry elements */
>       rcu_gc_registry();
> @@ -191,11 +191,11 @@ void synchronize_rcu(void)
>       update_counter_and_wait();      /* 0 -> 1, wait readers in parity 0 */
>  
>       /*
> -      * Adding a smp_mb() which is _not_ formally required, but makes the
> +      * Adding a cmm_smp_mb() which is _not_ formally required, but makes the
>        * model easier to understand. It does not have a big performance impact
>        * anyway, given this is the write-side.
>        */
> -     smp_mb();
> +     cmm_smp_mb();
>  
>       /*
>        * Wait for previous parity to be empty of readers.
> @@ -206,7 +206,7 @@ void synchronize_rcu(void)
>        * Finish waiting for reader threads before letting the old ptr being
>        * freed.
>        */
> -     smp_mb();
> +     cmm_smp_mb();
>  out:
>       mutex_unlock(&rcu_gp_lock);
>       ret = pthread_sigmask(SIG_SETMASK, &oldmask, NULL);
> diff --git a/urcu-defer.c b/urcu-defer.c
> index e241302..6dc08a3 100644
> --- a/urcu-defer.c
> +++ b/urcu-defer.c
> @@ -123,13 +123,13 @@ static unsigned long rcu_defer_num_callbacks(void)
>  static void wait_defer(void)
>  {
>       uatomic_dec(&defer_thread_futex);
> -     smp_mb();       /* Write futex before read queue */
> +     cmm_smp_mb();   /* Write futex before read queue */
>       if (rcu_defer_num_callbacks()) {
> -             smp_mb();       /* Read queue before write futex */
> +             cmm_smp_mb();   /* Read queue before write futex */
>               /* Callbacks are queued, don't wait. */
>               uatomic_set(&defer_thread_futex, 0);
>       } else {
> -             smp_rmb();      /* Read queue before read futex */
> +             cmm_smp_rmb();  /* Read queue before read futex */
>               if (uatomic_read(&defer_thread_futex) == -1)
>                       futex_noasync(&defer_thread_futex, FUTEX_WAIT, -1,
>                             NULL, NULL, 0);
> @@ -152,7 +152,7 @@ static void rcu_defer_barrier_queue(struct defer_queue 
> *queue,
>        */
>  
>       for (i = queue->tail; i != head;) {
> -             smp_rmb();       /* read head before q[]. */
> +             cmm_smp_rmb();       /* read head before q[]. */
>               p = LOAD_SHARED(queue->q[i++ & DEFER_QUEUE_MASK]);
>               if (unlikely(DQ_IS_FCT_BIT(p))) {
>                       DQ_CLEAR_FCT_BIT(p);
> @@ -166,7 +166,7 @@ static void rcu_defer_barrier_queue(struct defer_queue 
> *queue,
>               fct = queue->last_fct_out;
>               fct(p);
>       }
> -     smp_mb();       /* push tail after having used q[] */
> +     cmm_smp_mb();   /* push tail after having used q[] */
>       STORE_SHARED(queue->tail, i);
>  }
>  
> @@ -283,10 +283,10 @@ void _defer_rcu(void (*fct)(void *p), void *p)
>               }
>       }
>       _STORE_SHARED(defer_queue.q[head++ & DEFER_QUEUE_MASK], p);
> -     smp_wmb();      /* Publish new pointer before head */
> +     cmm_smp_wmb();  /* Publish new pointer before head */
>                       /* Write q[] before head. */
>       STORE_SHARED(defer_queue.head, head);
> -     smp_mb();       /* Write queue head before read futex */
> +     cmm_smp_mb();   /* Write queue head before read futex */
>       /*
>        * Wake-up any waiting defer thread.
>        */
> diff --git a/urcu-pointer-static.h b/urcu-pointer-static.h
> index 983d0d5..c8ac7f0 100644
> --- a/urcu-pointer-static.h
> +++ b/urcu-pointer-static.h
> @@ -63,7 +63,7 @@ extern "C" {
>  
>  #define _rcu_dereference(p)     ({                                   \
>                               typeof(p) _________p1 = LOAD_SHARED(p); \
> -                             smp_read_barrier_depends();             \
> +                             cmm_smp_read_barrier_depends();         \
>                               (_________p1);                          \
>                               })
>  
> @@ -81,7 +81,7 @@ extern "C" {
>               typeof(*p) _________pnew = (_new);                      \
>               if (!__builtin_constant_p(_new) ||                      \
>                   ((_new) != NULL))                                   \
> -                     wmb();                                          \
> +                     cmm_wmb();                                              
> \
>               uatomic_cmpxchg(p, _________pold, _________pnew);       \
>       })
>  
> @@ -96,7 +96,7 @@ extern "C" {
>               typeof(*p) _________pv = (v);           \
>               if (!__builtin_constant_p(v) ||         \
>                   ((v) != NULL))                      \
> -                     wmb();                          \
> +                     cmm_wmb();                              \
>               uatomic_xchg(p, _________pv);           \
>       })
>  
> @@ -106,7 +106,7 @@ extern "C" {
>               typeof(*p) _________pv = (v);           \
>               if (!__builtin_constant_p(v) ||         \
>                   ((v) != NULL))                      \
> -                     wmb();                          \
> +                     cmm_wmb();                              \
>               uatomic_set(p, _________pv);            \
>       })
>  
> diff --git a/urcu-pointer.c b/urcu-pointer.c
> index f5c9811..180c834 100644
> --- a/urcu-pointer.c
> +++ b/urcu-pointer.c
> @@ -39,18 +39,18 @@ void *rcu_dereference_sym(void *p)
>  
>  void *rcu_set_pointer_sym(void **p, void *v)
>  {
> -     wmb();
> +     cmm_wmb();
>       return uatomic_set(p, v);
>  }
>  
>  void *rcu_xchg_pointer_sym(void **p, void *v)
>  {
> -     wmb();
> +     cmm_wmb();
>       return uatomic_xchg(p, v);
>  }
>  
>  void *rcu_cmpxchg_pointer_sym(void **p, void *old, void *_new)
>  {
> -     wmb();
> +     cmm_wmb();
>       return uatomic_cmpxchg(p, old, _new);
>  }
> diff --git a/urcu-qsbr-static.h b/urcu-qsbr-static.h
> index 147eb18..108ef6e 100644
> --- a/urcu-qsbr-static.h
> +++ b/urcu-qsbr-static.h
> @@ -174,27 +174,27 @@ static inline void _rcu_read_unlock(void)
>  
>  static inline void _rcu_quiescent_state(void)
>  {
> -     smp_mb();       
> +     cmm_smp_mb();
>       _STORE_SHARED(rcu_reader.ctr, _LOAD_SHARED(rcu_gp_ctr));
> -     smp_mb();       /* write rcu_reader.ctr before read futex */
> +     cmm_smp_mb();   /* write rcu_reader.ctr before read futex */
>       wake_up_gp();
> -     smp_mb();
> +     cmm_smp_mb();
>  }
>  
>  static inline void _rcu_thread_offline(void)
>  {
> -     smp_mb();
> +     cmm_smp_mb();
>       STORE_SHARED(rcu_reader.ctr, 0);
> -     smp_mb();       /* write rcu_reader.ctr before read futex */
> +     cmm_smp_mb();   /* write rcu_reader.ctr before read futex */
>       wake_up_gp();
> -     barrier();      /* Ensure the compiler does not reorder us with mutex */
> +     cmm_barrier();  /* Ensure the compiler does not reorder us with mutex */
>  }
>  
>  static inline void _rcu_thread_online(void)
>  {
> -     barrier();      /* Ensure the compiler does not reorder us with mutex */
> +     cmm_barrier();  /* Ensure the compiler does not reorder us with mutex */
>       _STORE_SHARED(rcu_reader.ctr, LOAD_SHARED(rcu_gp_ctr));
> -     smp_mb();
> +     cmm_smp_mb();
>  }
>  
>  #ifdef __cplusplus 
> diff --git a/urcu-qsbr.c b/urcu-qsbr.c
> index 5e8b960..2cf73d5 100644
> --- a/urcu-qsbr.c
> +++ b/urcu-qsbr.c
> @@ -100,7 +100,7 @@ static void mutex_unlock(pthread_mutex_t *mutex)
>  static void wait_gp(void)
>  {
>       /* Read reader_gp before read futex */
> -     smp_rmb();
> +     cmm_smp_rmb();
>       if (uatomic_read(&gp_futex) == -1)
>               futex_noasync(&gp_futex, FUTEX_WAIT, -1,
>                     NULL, NULL, 0);
> @@ -126,14 +126,14 @@ static void update_counter_and_wait(void)
>        * while new readers are always accessing data (no progress). Enforce
>        * compiler-order of store to rcu_gp_ctr before load rcu_reader ctr.
>        */
> -     barrier();
> +     cmm_barrier();
>  
>       /*
> -      * Adding a smp_mb() which is _not_ formally required, but makes the
> +      * Adding a cmm_smp_mb() which is _not_ formally required, but makes the
>        * model easier to understand. It does not have a big performance impact
>        * anyway, given this is the write-side.
>        */
> -     smp_mb();
> +     cmm_smp_mb();
>  
>       /*
>        * Wait for each thread rcu_reader_qs_gp count to become 0.
> @@ -143,7 +143,7 @@ static void update_counter_and_wait(void)
>               if (wait_loops == RCU_QS_ACTIVE_ATTEMPTS) {
>                       uatomic_dec(&gp_futex);
>                       /* Write futex before read reader_gp */
> -                     smp_mb();
> +                     cmm_smp_mb();
>               }
>  
>               list_for_each_entry_safe(index, tmp, &registry, node) {
> @@ -154,7 +154,7 @@ static void update_counter_and_wait(void)
>               if (list_empty(&registry)) {
>                       if (wait_loops == RCU_QS_ACTIVE_ATTEMPTS) {
>                               /* Read reader_gp before write futex */
> -                             smp_mb();
> +                             cmm_smp_mb();
>                               uatomic_set(&gp_futex, 0);
>                       }
>                       break;
> @@ -165,7 +165,7 @@ static void update_counter_and_wait(void)
>  #ifndef HAS_INCOHERENT_CACHES
>                               cpu_relax();
>  #else /* #ifndef HAS_INCOHERENT_CACHES */
> -                             smp_mb();
> +                             cmm_smp_mb();
>  #endif /* #else #ifndef HAS_INCOHERENT_CACHES */
>                       }
>               }
> @@ -190,7 +190,7 @@ void synchronize_rcu(void)
>        * where new ptr points to.
>        */
>       /* Write new ptr before changing the qparity */
> -     smp_mb();
> +     cmm_smp_mb();
>  
>       /*
>        * Mark the writer thread offline to make sure we don't wait for
> @@ -217,14 +217,14 @@ void synchronize_rcu(void)
>        * accessing data (no progress).  Enforce compiler-order of load
>        * rcu_reader ctr before store to rcu_gp_ctr.
>        */
> -     barrier();
> +     cmm_barrier();
>  
>       /*
> -      * Adding a smp_mb() which is _not_ formally required, but makes the
> +      * Adding a cmm_smp_mb() which is _not_ formally required, but makes the
>        * model easier to understand. It does not have a big performance impact
>        * anyway, given this is the write-side.
>        */
> -     smp_mb();
> +     cmm_smp_mb();
>  
>       /*
>        * Wait for previous parity to be empty of readers.
> @@ -239,7 +239,7 @@ out:
>        */
>       if (was_online)
>               _STORE_SHARED(rcu_reader.ctr, LOAD_SHARED(rcu_gp_ctr));
> -     smp_mb();
> +     cmm_smp_mb();
>  }
>  #else /* !(BITS_PER_LONG < 64) */
>  void synchronize_rcu(void)
> @@ -253,7 +253,7 @@ void synchronize_rcu(void)
>        * our own quiescent state. This allows using synchronize_rcu() in
>        * threads registered as readers.
>        */
> -     smp_mb();
> +     cmm_smp_mb();
>       if (was_online)
>               STORE_SHARED(rcu_reader.ctr, 0);
>  
> @@ -266,7 +266,7 @@ out:
>  
>       if (was_online)
>               _STORE_SHARED(rcu_reader.ctr, LOAD_SHARED(rcu_gp_ctr));
> -     smp_mb();
> +     cmm_smp_mb();
>  }
>  #endif  /* !(BITS_PER_LONG < 64) */
>  
> diff --git a/urcu-static.h b/urcu-static.h
> index eea39bb..ad415ac 100644
> --- a/urcu-static.h
> +++ b/urcu-static.h
> @@ -181,23 +181,23 @@ extern int has_sys_membarrier;
>  static inline void smp_mb_slave(int group)
>  {
>       if (likely(has_sys_membarrier))
> -             barrier();
> +             cmm_barrier();
>       else
> -             smp_mb();
> +             cmm_smp_mb();
>  }
>  #endif
>  
>  #ifdef RCU_MB
>  static inline void smp_mb_slave(int group)
>  {
> -     smp_mb();
> +     cmm_smp_mb();
>  }
>  #endif
>  
>  #ifdef RCU_SIGNAL
>  static inline void smp_mb_slave(int group)
>  {
> -     barrier();
> +     cmm_barrier();
>  }
>  #endif
>  
> @@ -259,7 +259,7 @@ static inline void _rcu_read_lock(void)
>  {
>       unsigned long tmp;
>  
> -     barrier();      /* Ensure the compiler does not reorder us with mutex */
> +     cmm_barrier();  /* Ensure the compiler does not reorder us with mutex */
>       tmp = rcu_reader.ctr;
>       /*
>        * rcu_gp_ctr is
> @@ -295,7 +295,7 @@ static inline void _rcu_read_unlock(void)
>       } else {
>               _STORE_SHARED(rcu_reader.ctr, rcu_reader.ctr - RCU_GP_COUNT);
>       }
> -     barrier();      /* Ensure the compiler does not reorder us with mutex */
> +     cmm_barrier();  /* Ensure the compiler does not reorder us with mutex */
>  }
>  
>  #ifdef __cplusplus
> diff --git a/urcu.c b/urcu.c
> index 918d274..9c556aa 100644
> --- a/urcu.c
> +++ b/urcu.c
> @@ -100,9 +100,9 @@ static void mutex_lock(pthread_mutex_t *mutex)
>                       exit(-1);
>               }
>               if (LOAD_SHARED(rcu_reader.need_mb)) {
> -                     smp_mb();
> +                     cmm_smp_mb();
>                       _STORE_SHARED(rcu_reader.need_mb, 0);
> -                     smp_mb();
> +                     cmm_smp_mb();
>               }
>               poll(NULL,0,10);
>       }
> @@ -126,14 +126,14 @@ static void smp_mb_master(int group)
>       if (likely(has_sys_membarrier))
>               membarrier(MEMBARRIER_EXPEDITED);
>       else
> -             smp_mb();
> +             cmm_smp_mb();
>  }
>  #endif
>  
>  #ifdef RCU_MB
>  static void smp_mb_master(int group)
>  {
> -     smp_mb();
> +     cmm_smp_mb();
>  }
>  #endif
>  
> @@ -143,15 +143,15 @@ static void force_mb_all_readers(void)
>       struct rcu_reader *index;
>  
>       /*
> -      * Ask for each threads to execute a smp_mb() so we can consider the
> +      * Ask for each threads to execute a cmm_smp_mb() so we can consider the
>        * compiler barriers around rcu read lock as real memory barriers.
>        */
>       if (list_empty(&registry))
>               return;
>       /*
> -      * pthread_kill has a smp_mb(). But beware, we assume it performs
> +      * pthread_kill has a cmm_smp_mb(). But beware, we assume it performs
>        * a cache flush on architectures with non-coherent cache. Let's play
> -      * safe and don't assume anything : we use smp_mc() to make sure the
> +      * safe and don't assume anything : we use cmm_smp_mc() to make sure the
>        * cache flush is enforced.
>        */
>       list_for_each_entry(index, &registry, node) {
> @@ -177,7 +177,7 @@ static void force_mb_all_readers(void)
>                       poll(NULL, 0, 1);
>               }
>       }
> -     smp_mb();       /* read ->need_mb before ending the barrier */
> +     cmm_smp_mb();   /* read ->need_mb before ending the barrier */
>  }
>  
>  static void smp_mb_master(int group)
> @@ -213,15 +213,15 @@ void update_counter_and_wait(void)
>        * while new readers are always accessing data (no progress). Enforce
>        * compiler-order of store to rcu_gp_ctr before load rcu_reader ctr.
>        */
> -     barrier();
> +     cmm_barrier();
>  
>       /*
>        *
> -      * Adding a smp_mb() which is _not_ formally required, but makes the
> +      * Adding a cmm_smp_mb() which is _not_ formally required, but makes the
>        * model easier to understand. It does not have a big performance impact
>        * anyway, given this is the write-side.
>        */
> -     smp_mb();
> +     cmm_smp_mb();
>  
>       /*
>        * Wait for each thread rcu_reader.ctr count to become 0.
> @@ -309,14 +309,14 @@ void synchronize_rcu(void)
>        * accessing data (no progress).  Enforce compiler-order of load
>        * rcu_reader ctr before store to rcu_gp_ctr.
>        */
> -     barrier();
> +     cmm_barrier();
>  
>       /*
> -      * Adding a smp_mb() which is _not_ formally required, but makes the
> +      * Adding a cmm_smp_mb() which is _not_ formally required, but makes the
>        * model easier to understand. It does not have a big performance impact
>        * anyway, given this is the write-side.
>        */
> -     smp_mb();
> +     cmm_smp_mb();
>  
>       /*
>        * Wait for previous parity to be empty of readers.
> @@ -379,13 +379,13 @@ void rcu_init(void)
>  static void sigrcu_handler(int signo, siginfo_t *siginfo, void *context)
>  {
>       /*
> -      * Executing this smp_mb() is the only purpose of this signal handler.
> -      * It punctually promotes barrier() into smp_mb() on every thread it is
> +      * Executing this cmm_smp_mb() is the only purpose of this signal 
> handler.
> +      * It punctually promotes cmm_barrier() into cmm_smp_mb() on every 
> thread it is
>        * executed on.
>        */
> -     smp_mb();
> +     cmm_smp_mb();
>       _STORE_SHARED(rcu_reader.need_mb, 0);
> -     smp_mb();
> +     cmm_smp_mb();
>  }
>  
>  /*
> diff --git a/urcu/arch_alpha.h b/urcu/arch_alpha.h
> index 0596939..51a1fff 100644
> --- a/urcu/arch_alpha.h
> +++ b/urcu/arch_alpha.h
> @@ -28,9 +28,9 @@
>  extern "C" {
>  #endif
>  
> -#define mb()                 asm volatile("mb":::"memory")
> -#define wmb()                        asm volatile("wmb":::"memory")
> -#define read_barrier_depends()       asm volatile("mb":::"memory")
> +#define cmm_mb()                     asm volatile("mb":::"memory")
> +#define cmm_wmb()                    asm volatile("wmb":::"memory")
> +#define cmm_read_barrier_depends()   asm volatile("mb":::"memory")
>  
>  typedef unsigned long long cycles_t;
>  
> diff --git a/urcu/arch_armv7l.h b/urcu/arch_armv7l.h
> index e63ad3c..6d58bb7 100644
> --- a/urcu/arch_armv7l.h
> +++ b/urcu/arch_armv7l.h
> @@ -29,7 +29,7 @@
>  extern "C" {
>  #endif 
>  
> -#define mb()    asm volatile("dmb":::"memory")
> +#define cmm_mb()    asm volatile("dmb":::"memory")
>  
>  #include <stdlib.h>
>  #include <sys/time.h>
> diff --git a/urcu/arch_generic.h b/urcu/arch_generic.h
> index 36a0003..25a9f7a 100644
> --- a/urcu/arch_generic.h
> +++ b/urcu/arch_generic.h
> @@ -32,93 +32,93 @@ extern "C" {
>  #define CACHE_LINE_SIZE      64
>  #endif
>  
> -#if !defined(mc) && !defined(rmc) && !defined(wmc)
> +#if !defined(cmm_mc) && !defined(cmm_rmc) && !defined(cmm_wmc)
>  #define CONFIG_HAVE_MEM_COHERENCY
>  /*
> - * Architectures with cache coherency must _not_ define mc/rmc/wmc.
> + * Architectures with cache coherency must _not_ define 
> cmm_mc/cmm_rmc/cmm_wmc.
>   *
> - * For them, mc/rmc/wmc are implemented with a * simple compiler barrier;
> - * in addition, we provide defaults for mb (using GCC builtins) as well as
> - * rmb and wmb (defaulting to mb).
> + * For them, cmm_mc/cmm_rmc/cmm_wmc are implemented with a * simple compiler 
> barrier;
> + * in addition, we provide defaults for cmm_mb (using GCC builtins) as well 
> as
> + * cmm_rmb and cmm_wmb (defaulting to cmm_mb).
>   */
>  
> -#ifndef mb
> -#define mb()    __sync_synchronize()
> +#ifndef cmm_mb
> +#define cmm_mb()    __sync_synchronize()
>  #endif
>  
> -#ifndef rmb
> -#define rmb()    mb()
> +#ifndef cmm_rmb
> +#define cmm_rmb()    cmm_mb()
>  #endif
>  
> -#ifndef wmb
> -#define wmb()    mb()
> +#ifndef cmm_wmb
> +#define cmm_wmb()    cmm_mb()
>  #endif
>  
> -#define mc()  barrier()
> -#define rmc()         barrier()
> -#define wmc()         barrier()
> +#define cmm_mc()     cmm_barrier()
> +#define cmm_rmc()    cmm_barrier()
> +#define cmm_wmc()    cmm_barrier()
>  #else
>  /*
>   * Architectures without cache coherency need something like the following:
>   *
> - * #define mc()              arch_cache_flush() 
> - * #define rmc()     arch_cache_flush_read()
> - * #define wmc()     arch_cache_flush_write()
> + * #define cmm_mc()          arch_cache_flush() 
> + * #define cmm_rmc() arch_cache_flush_read()
> + * #define cmm_wmc() arch_cache_flush_write()
>   *
> - * Of these, only mc is mandatory.  rmc and wmc default to mc.  mb/rmb/wmb
> - * use these definitions by default:
> + * Of these, only cmm_mc is mandatory.  cmm_rmc and cmm_wmc default to 
> cmm_mc.  
> + * cmm_mb/cmm_rmb/cmm_wmb use these definitions by default:
>   *
> - * #define mb()              mc()
> - * #define rmb()     rmc()
> - * #define wmb()     wmc()
> + * #define cmm_mb()          cmm_mc()
> + * #define cmm_rmb() cmm_rmc()
> + * #define cmm_wmb() cmm_wmc()
>   */
>  
> -#ifndef mb
> -#define mb()    mc()
> +#ifndef cmm_mb
> +#define cmm_mb()     cmm_mc()
>  #endif
>  
> -#ifndef rmb
> -#define rmb()    rmc()
> +#ifndef cmm_rmb
> +#define cmm_rmb()    cmm_rmc()
>  #endif
>  
> -#ifndef wmb
> -#define wmb()    wmc()
> +#ifndef cmm_wmb
> +#define cmm_wmb()    cmm_wmc()
>  #endif
>  
> -#ifndef rmc
> -#define rmc()    mc()
> +#ifndef cmm_rmc
> +#define cmm_rmc()    cmm_mc()
>  #endif
>  
> -#ifndef wmc
> -#define wmc()    mc()
> +#ifndef cmm_wmc
> +#define cmm_wmc()    cmm_mc()
>  #endif
>  #endif
>  
>  /* Nop everywhere except on alpha. */
> -#ifndef read_barrier_depends
> -#define read_barrier_depends()
> +#ifndef cmm_read_barrier_depends
> +#define cmm_read_barrier_depends()
>  #endif
>  
>  #ifdef CONFIG_RCU_SMP
> -#define smp_mb()     mb()
> -#define smp_rmb()    rmb()
> -#define smp_wmb()    wmb()
> -#define smp_mc()     mc()
> -#define smp_rmc()    rmc()
> -#define smp_wmc()    wmc()
> -#define smp_read_barrier_depends()   read_barrier_depends()
> +#define cmm_smp_mb() cmm_mb()
> +#define cmm_smp_rmb()        cmm_rmb()
> +#define cmm_smp_wmb()        cmm_wmb()
> +#define cmm_smp_mc() cmm_mc()
> +#define cmm_smp_rmc()        cmm_rmc()
> +#define cmm_smp_wmc()        cmm_wmc()
> +#define cmm_smp_read_barrier_depends()       cmm_read_barrier_depends()
>  #else
> -#define smp_mb()     barrier()
> -#define smp_rmb()    barrier()
> -#define smp_wmb()    barrier()
> -#define smp_mc()     barrier()
> -#define smp_rmc()    barrier()
> -#define smp_wmc()    barrier()
> -#define smp_read_barrier_depends()
> +#define cmm_smp_mb() cmm_barrier()
> +#define cmm_smp_rmb()        cmm_barrier()
> +#define cmm_smp_wmb()        cmm_barrier()
> +#define cmm_smp_mc() cmm_barrier()
> +#define cmm_smp_rmc()        cmm_barrier()
> +#define cmm_smp_wmc()        cmm_barrier()
> +#define cmm_smp_read_barrier_depends()
>  #endif
>  
>  #ifndef cpu_relax
> -#define cpu_relax()   barrier()
> +#define cpu_relax()          cmm_barrier()
>  #endif
>  
>  #ifdef __cplusplus
> diff --git a/urcu/arch_ppc.h b/urcu/arch_ppc.h
> index 7a217c8..93aed2a 100644
> --- a/urcu/arch_ppc.h
> +++ b/urcu/arch_ppc.h
> @@ -32,7 +32,7 @@ extern "C" {
>  /* Include size of POWER5+ L3 cache lines: 256 bytes */
>  #define CACHE_LINE_SIZE      256
>  
> -#define mb()    asm volatile("sync":::"memory")
> +#define cmm_mb()    asm volatile("sync":::"memory")
>  
>  #define mftbl()                                              \
>       ({                                              \
> @@ -56,9 +56,9 @@ static inline cycles_t get_cycles (void)
>  
>       for (;;) {
>               h = mftbu();
> -             barrier();
> +             cmm_barrier();
>               l = mftbl();
> -             barrier();
> +             cmm_barrier();
>               if (mftbu() == h)
>                       return (((cycles_t) h) << 32) + l;
>       }
> diff --git a/urcu/arch_s390.h b/urcu/arch_s390.h
> index 1a55e0d..8a33e20 100644
> --- a/urcu/arch_s390.h
> +++ b/urcu/arch_s390.h
> @@ -37,7 +37,7 @@ extern "C" {
>  
>  #define CACHE_LINE_SIZE      128
>  
> -#define mb()    __asm__ __volatile__("bcr 15,0" : : : "memory")
> +#define cmm_mb()    __asm__ __volatile__("bcr 15,0" : : : "memory")
>  
>  typedef unsigned long long cycles_t;
>  
> diff --git a/urcu/arch_sparc64.h b/urcu/arch_sparc64.h
> index 8e991b1..39f27c7 100644
> --- a/urcu/arch_sparc64.h
> +++ b/urcu/arch_sparc64.h
> @@ -40,9 +40,9 @@ __asm__ __volatile__("ba,pt %%xcc, 1f\n\t"  \
>                    "1:\n"                     \
>                    : : : "memory")
>  
> -#define mb()    membar_safe("#LoadLoad | #LoadStore | #StoreStore | 
> #StoreLoad")
> -#define rmb()    membar_safe("#LoadLoad")
> -#define wmb()    membar_safe("#StoreStore")
> +#define cmm_mb()     membar_safe("#LoadLoad | #LoadStore | #StoreStore | 
> #StoreLoad")
> +#define cmm_rmb()    membar_safe("#LoadLoad")
> +#define cmm_wmb()    membar_safe("#StoreStore")
>  
>  typedef unsigned long long cycles_t;
>  
> diff --git a/urcu/arch_x86.h b/urcu/arch_x86.h
> index aad541e..d0a58e8 100644
> --- a/urcu/arch_x86.h
> +++ b/urcu/arch_x86.h
> @@ -32,17 +32,17 @@ extern "C" {
>  #define CACHE_LINE_SIZE      128
>  
>  #ifdef CONFIG_RCU_HAVE_FENCE
> -#define mb()    asm volatile("mfence":::"memory")
> -#define rmb()   asm volatile("lfence":::"memory")
> -#define wmb()   asm volatile("sfence"::: "memory")
> +#define cmm_mb()    asm volatile("mfence":::"memory")
> +#define cmm_rmb()   asm volatile("lfence":::"memory")
> +#define cmm_wmb()   asm volatile("sfence"::: "memory")
>  #else
>  /*
> - * Some non-Intel clones support out of order store. wmb() ceases to be a
> + * Some non-Intel clones support out of order store. cmm_wmb() ceases to be a
>   * nop for these.
>   */
> -#define mb()    asm volatile("lock; addl $0,0(%%esp)":::"memory")
> -#define rmb()   asm volatile("lock; addl $0,0(%%esp)":::"memory")
> -#define wmb()   asm volatile("lock; addl $0,0(%%esp)"::: "memory")
> +#define cmm_mb()    asm volatile("lock; addl $0,0(%%esp)":::"memory")
> +#define cmm_rmb()   asm volatile("lock; addl $0,0(%%esp)":::"memory")
> +#define cmm_wmb()   asm volatile("lock; addl $0,0(%%esp)"::: "memory")
>  #endif
>  
>  #define cpu_relax()  asm volatile("rep; nop" : : : "memory");
> diff --git a/urcu/compiler.h b/urcu/compiler.h
> index fb8b829..d67e6c5 100644
> --- a/urcu/compiler.h
> +++ b/urcu/compiler.h
> @@ -23,7 +23,7 @@
>  #define likely(x)    __builtin_expect(!!(x), 1)
>  #define unlikely(x)  __builtin_expect(!!(x), 0)
>  
> -#define      barrier()       asm volatile("" : : : "memory")
> +#define      cmm_barrier()   asm volatile("" : : : "memory")
>  
>  /*
>   * Instruct the compiler to perform only a single access to a variable
> diff --git a/urcu/rcuhlist.h b/urcu/rcuhlist.h
> index 89f7679..cce80aa 100644
> --- a/urcu/rcuhlist.h
> +++ b/urcu/rcuhlist.h
> @@ -35,7 +35,7 @@ static inline void hlist_add_head_rcu(struct hlist_node 
> *newp,
>  {
>       newp->next = head->next;
>       newp->prev = (struct hlist_node *)head;
> -     smp_wmb();
> +     cmm_smp_wmb();
>       if (head->next)
>               head->next->prev = newp;
>       head->next = newp;
> diff --git a/urcu/rculist.h b/urcu/rculist.h
> index b0c4f1a..3a4e335 100644
> --- a/urcu/rculist.h
> +++ b/urcu/rculist.h
> @@ -34,7 +34,7 @@ static inline void list_add_rcu(list_t *newp, list_t *head)
>  {
>       newp->next = head->next;
>       newp->prev = head;
> -     smp_wmb();
> +     cmm_smp_wmb();
>       head->next->prev = newp;
>       head->next = newp;
>  }
> diff --git a/urcu/system.h b/urcu/system.h
> index 0de7dd2..11a499e 100644
> --- a/urcu/system.h
> +++ b/urcu/system.h
> @@ -22,7 +22,7 @@
>  #include <urcu/arch.h>
>  
>  /*
> - * Identify a shared load. A smp_rmc() or smp_mc() should come before the 
> load.
> + * Identify a shared load. A cmm_smp_rmc() or cmm_smp_mc() should come 
> before the load.
>   */
>  #define _LOAD_SHARED(p)             ACCESS_ONCE(p)
>  
> @@ -31,12 +31,12 @@
>   */
>  #define LOAD_SHARED(p)                       \
>       ({                              \
> -             smp_rmc();              \
> +             cmm_smp_rmc();          \
>               _LOAD_SHARED(p);        \
>       })
>  
>  /*
> - * Identify a shared store. A smp_wmc() or smp_mc() should follow the store.
> + * Identify a shared store. A cmm_smp_wmc() or cmm_smp_mc() should follow 
> the store.
>   */
>  #define _STORE_SHARED(x, v)  ({ ACCESS_ONCE(x) = (v); })
>  
> @@ -47,7 +47,7 @@
>  #define STORE_SHARED(x, v)           \
>       ({                              \
>               typeof(x) _v = _STORE_SHARED(x, v);     \
> -             smp_wmc();              \
> +             cmm_smp_wmc();          \
>               _v;                     \
>       })
>  
> -- 
> 1.7.3.2
> 
> 
> _______________________________________________
> ltt-dev mailing list
> [email protected]
> http://lists.casi.polymtl.ca/cgi-bin/mailman/listinfo/ltt-dev
> 

-- 
Mathieu Desnoyers
Operating System Efficiency R&D Consultant
EfficiOS Inc.
http://www.efficios.com

_______________________________________________
ltt-dev mailing list
[email protected]
http://lists.casi.polymtl.ca/cgi-bin/mailman/listinfo/ltt-dev

Reply via email to