rcu: Make rcutorture fakewriters invoke rcu_barrier()
authorPaul E. McKenney <paul.mckenney@linaro.org>
Wed, 30 May 2012 00:50:51 +0000 (17:50 -0700)
committerPaul E. McKenney <paulmck@linux.vnet.ibm.com>
Mon, 2 Jul 2012 19:34:04 +0000 (12:34 -0700)
The current rcutorture rcu_barrier() testing never intentionally runs
more than one instance of rcu_barrier() at a given time.  This fails
to test the the shiny new concurrency features of rcu_barrier().  This
commit therefore modifies the rcutorture fakewriter kthread to randomly
invoke rcu_barrier() rather than the usual synchronize_rcu().

Signed-off-by: Paul E. McKenney <paul.mckenney@linaro.org>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
kernel/rcutorture.c

index 045a3dc233ead63f892f360da343d541523b9158..c279ee9209473dedb9c510c3a584510f588defa5 100644 (file)
@@ -1025,7 +1025,11 @@ rcu_torture_fakewriter(void *arg)
        do {
                schedule_timeout_uninterruptible(1 + rcu_random(&rand)%10);
                udelay(rcu_random(&rand) & 0x3ff);
-               cur_ops->sync();
+               if (cur_ops->cb_barrier != NULL &&
+                   rcu_random(&rand) % (nfakewriters * 8) == 0)
+                       cur_ops->cb_barrier();
+               else
+                       cur_ops->sync();
                rcu_stutter_wait("rcu_torture_fakewriter");
        } while (!kthread_should_stop() && fullstop == FULLSTOP_DONTSTOP);