tests: test-thread-pool is racy - add some barriers

I imagine there is more to be done to fix the memory consistency
races here, but these added barriers at least let it pass on ppc64le,
whereas before it would fail regularly there.

Signed-off-by: Bruce Rogers <brogers@suse.com>

tests/test-thread-pool.c

@@ -171,6 +171,7 @@ static void do_test_cancel(bool sync)
     /* Cancel the jobs that haven't been started yet.  */
     num_canceled = 0;
     for (i = 0; i < 100; i++) {
+	smp_mb();
         if (atomic_cmpxchg(&data[i].n, 0, 3) == 0) {
             data[i].ret = -ECANCELED;
             if (sync) {
@@ -185,6 +186,7 @@ static void do_test_cancel(bool sync)
     g_assert_cmpint(num_canceled, <, 100);
 
     for (i = 0; i < 100; i++) {
+	smp_mb();
         if (data[i].aiocb && data[i].n != 3) {
             if (sync) {
                 /* Canceling the others will be a blocking operation.  */
@@ -201,6 +203,7 @@ static void do_test_cancel(bool sync)
     }
     g_assert_cmpint(active, ==, 0);
     for (i = 0; i < 100; i++) {
+	smp_mb();
         if (data[i].n == 3) {
             g_assert_cmpint(data[i].ret, ==, -ECANCELED);
             g_assert(data[i].aiocb == NULL);