rcu: apply TINY_PREEMPT_RCU read-side speedup to TREE_PREEMPT_RCU

Replace one of the ACCESS_ONCE() calls in each of __rcu_read_lock()
and __rcu_read_unlock() with barrier() as suggested by Steve Rostedt in
order to avoid the potential compiler-optimization-induced bug noted by
Mathieu Desnoyers.
Located-by: Mathieu Desnoyers <mathieu.desnoyers@polymtl.ca>
Suggested-by: Steven Rostedt <rostedt@goodmis.org>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>

kernel/rcutree_plugin.h

@@ -201,7 +201,7 @@ static void rcu_preempt_note_context_switch(int cpu)
  */
 void __rcu_read_lock(void)
 {
-	ACCESS_ONCE(current->rcu_read_lock_nesting)++;
+	current->rcu_read_lock_nesting++;
 	barrier();  /* needed if we ever invoke rcu_read_lock in rcutree.c */
 }
 EXPORT_SYMBOL_GPL(__rcu_read_lock);
@@ -344,7 +344,9 @@ void __rcu_read_unlock(void)
 	struct task_struct *t = current;
 
 	barrier();  /* needed if we ever invoke rcu_read_unlock in rcutree.c */
-	if (--ACCESS_ONCE(t->rcu_read_lock_nesting) == 0 &&
+	--t->rcu_read_lock_nesting;
+	barrier();  /* decrement before load of ->rcu_read_unlock_special */
+	if (t->rcu_read_lock_nesting == 0 &&
 	    unlikely(ACCESS_ONCE(t->rcu_read_unlock_special)))
 		rcu_read_unlock_special(t);
 #ifdef CONFIG_PROVE_LOCKING