vppinfra: add atomic macros for __sync builtins

This is first part of addition of atomic macros with only macros for
__sync builtins.

- Based on earlier patch by Damjan (https://gerrit.fd.io/r/#/c/10729/)
Additionally
- clib_atomic_release macro added and used in the absence
of any memory barrier.
- clib_atomic_bool_cmp_and_swap added

Change-Id: Ie4e48c1e184a652018d1d0d87c4be80ddd180a3b
Original-patch-by: Damjan Marion <damarion@cisco.com>
Signed-off-by: Sirshak Das <sirshak.das@arm.com>
Reviewed-by: Honnappa Nagarahalli <honnappa.nagarahalli@arm.com>
Reviewed-by: Ola Liljedahl <ola.liljedahl@arm.com>
Reviewed-by: Steve Capper <steve.capper@arm.com>

4 files changed, 14 insertions, 30 deletions

diff --git a/src/vlib/buffer_funcs.h b/src/vlib/buffer_funcs.h
index d8abdf31d79..438bf7e5ee4 100644
--- a/src/vlib/buffer_funcs.h
+++ b/src/vlib/buffer_funcs.h
@@ -923,7 +923,7 @@ vlib_buffer_attach_clone (vlib_main_t * vm, vlib_buffer_t * head,
     tail->total_length_not_including_first_buffer;
 
 next_segment:
-  __sync_add_and_fetch (&tail->n_add_refs, 1);
+  clib_atomic_add_fetch (&tail->n_add_refs, 1);
 
   if (tail->flags & VLIB_BUFFER_NEXT_PRESENT)
     {
@@ -1153,7 +1153,7 @@ vlib_validate_buffer_in_use (vlib_buffer_t * b, u32 expected)
 
   oldheap = clib_mem_set_heap (vlib_buffer_state_heap);
 
-  while (__sync_lock_test_and_set (vlib_buffer_state_validation_lock, 1))
+  while (clib_atomic_test_and_set (vlib_buffer_state_validation_lock))
     ;
 
   p = hash_get (vlib_buffer_state_validation_hash, b);
@@ -1196,7 +1196,7 @@ vlib_validate_buffer_set_in_use (vlib_buffer_t * b, u32 expected)
 
   oldheap = clib_mem_set_heap (vlib_buffer_state_heap);
 
-  while (__sync_lock_test_and_set (vlib_buffer_state_validation_lock, 1))
+  while (clib_atomic_test_and_set (vlib_buffer_state_validation_lock))
     ;
 
   hash_set (vlib_buffer_state_validation_hash, b, expected);
diff --git a/src/vlib/threads.c b/src/vlib/threads.c
index 981209bf2e7..7f407e92478 100644
--- a/src/vlib/threads.c
+++ b/src/vlib/threads.c
@@ -516,7 +516,7 @@ vlib_frame_queue_enqueue (vlib_main_t * vm, u32 node_runtime_index,
 
   ASSERT (fq);
 
-  new_tail = __sync_add_and_fetch (&fq->tail, 1);
+  new_tail = clib_atomic_add_fetch (&fq->tail, 1);
 
   /* Wait until a ring slot is available */
   while (new_tail >= fq->head + fq->nelts)
@@ -576,12 +576,12 @@ vlib_worker_thread_init (vlib_worker_thread_t * w)
     {
 
       /* Initial barrier sync, for both worker and i/o threads */
-      clib_smp_atomic_add (vlib_worker_threads->workers_at_barrier, 1);
+      clib_atomic_fetch_add (vlib_worker_threads->workers_at_barrier, 1);
 
       while (*vlib_worker_threads->wait_at_barrier)
 	;
 
-      clib_smp_atomic_add (vlib_worker_threads->workers_at_barrier, -1);
+      clib_atomic_fetch_add (vlib_worker_threads->workers_at_barrier, -1);
     }
 }
 
@@ -1310,22 +1310,6 @@ cpu_config (vlib_main_t * vm, unformat_input_t * input)
 
 VLIB_EARLY_CONFIG_FUNCTION (cpu_config, "cpu");
 
-#if !defined (__x86_64__) && !defined (__i386__) && !defined (__aarch64__) && !defined (__powerpc64__) && !defined(__arm__)
-void
-__sync_fetch_and_add_8 (void)
-{
-  fformat (stderr, "%s called\n", __FUNCTION__);
-  abort ();
-}
-
-void
-__sync_add_and_fetch_8 (void)
-{
-  fformat (stderr, "%s called\n", __FUNCTION__);
-  abort ();
-}
-#endif
-
 void vnet_main_fixup (vlib_fork_fixup_t which) __attribute__ ((weak));
 void
 vnet_main_fixup (vlib_fork_fixup_t which)
@@ -1493,8 +1477,8 @@ vlib_worker_thread_barrier_release (vlib_main_t * vm)
 
       /* Do per thread rebuilds in parallel */
       refork_needed = 1;
-      clib_smp_atomic_add (vlib_worker_threads->node_reforks_required,
-			   (vec_len (vlib_mains) - 1));
+      clib_atomic_fetch_add (vlib_worker_threads->node_reforks_required,
+			     (vec_len (vlib_mains) - 1));
       now = vlib_time_now (vm);
       t_update_main = now - vm->barrier_epoch;
     }
diff --git a/src/vlib/threads.h b/src/vlib/threads.h
index bb7c164c2e3..0e9cba52103 100644
--- a/src/vlib/threads.h
+++ b/src/vlib/threads.h
@@ -414,7 +414,7 @@ vlib_worker_thread_barrier_check (void)
 	  ed->thread_index = thread_index;
 	}
 
-      clib_smp_atomic_add (vlib_worker_threads->workers_at_barrier, 1);
+      clib_atomic_fetch_add (vlib_worker_threads->workers_at_barrier, 1);
       if (CLIB_DEBUG > 0)
 	{
 	  vm = vlib_get_main ();
@@ -424,7 +424,7 @@ vlib_worker_thread_barrier_check (void)
 	;
       if (CLIB_DEBUG > 0)
 	vm->parked_at_barrier = 0;
-      clib_smp_atomic_add (vlib_worker_threads->workers_at_barrier, -1);
+      clib_atomic_fetch_add (vlib_worker_threads->workers_at_barrier, -1);
 
       if (PREDICT_FALSE (*vlib_worker_threads->node_reforks_required))
 	{
@@ -450,8 +450,8 @@ vlib_worker_thread_barrier_check (void)
 	    }
 
 	  vlib_worker_thread_node_refork ();
-	  clib_smp_atomic_add (vlib_worker_threads->node_reforks_required,
-			       -1);
+	  clib_atomic_fetch_add (vlib_worker_threads->node_reforks_required,
+				 -1);
 	  while (*vlib_worker_threads->node_reforks_required)
 	    ;
 	}
@@ -519,7 +519,7 @@ vlib_get_frame_queue_elt (u32 frame_queue_index, u32 index)
   fq = fqm->vlib_frame_queues[index];
   ASSERT (fq);
 
-  new_tail = __sync_add_and_fetch (&fq->tail, 1);
+  new_tail = clib_atomic_add_fetch (&fq->tail, 1);
 
   /* Wait until a ring slot is available */
   while (new_tail >= fq->head_hint + fq->nelts)
diff --git a/src/vlib/unix/cj.c b/src/vlib/unix/cj.c
index 0232ea2d690..7757146c7a7 100644
--- a/src/vlib/unix/cj.c
+++ b/src/vlib/unix/cj.c
@@ -44,7 +44,7 @@ cj_log (u32 type, void *data0, void *data1)
   if (cjm->enable == 0)
     return;
 
-  new_tail = __sync_add_and_fetch (&cjm->tail, 1);
+  new_tail = clib_atomic_add_fetch (&cjm->tail, 1);
 
   r = (cj_record_t *) & (cjm->records[new_tail & (cjm->num_records - 1)]);
   r->time = vlib_time_now (cjm->vlib_main);