rcu: Move __rcu_read_unlock()'s barrier() within if-statement

We only need to constrain the compiler if we are actually exiting
the top-level RCU read-side critical section.  This commit therefore
moves the first barrier() cal in __rcu_read_unlock() to inside the
"if" statement, thus avoiding needless register flushes for inner
rcu_read_unlock() calls.

Signed-off-by: Paul E. McKenney <paul.mckenney@linaro.org>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
diff --git a/kernel/rcutree_plugin.h b/kernel/rcutree_plugin.h
index 4bac5a2..ed70f6b 100644
--- a/kernel/rcutree_plugin.h
+++ b/kernel/rcutree_plugin.h
@@ -415,10 +415,10 @@
 {
 	struct task_struct *t = current;
 
-	barrier();  /* needed if we ever invoke rcu_read_unlock in rcutree.c */
 	if (t->rcu_read_lock_nesting != 1)
 		--t->rcu_read_lock_nesting;
 	else {
+		barrier();  /* critical section before exit code. */
 		t->rcu_read_lock_nesting = INT_MIN;
 		barrier();  /* assign before ->rcu_read_unlock_special load */
 		if (unlikely(ACCESS_ONCE(t->rcu_read_unlock_special)))