rcutorture: Add RCU Tasks Trace and SRCU deadlock scenarios
authorPaul E. McKenney <paulmck@kernel.org>
Tue, 24 Jan 2023 17:43:15 +0000 (09:43 -0800)
committerBoqun Feng <boqun.feng@gmail.com>
Mon, 27 Mar 2023 18:16:12 +0000 (11:16 -0700)
Add a test number 3 that creates deadlock cycles involving one RCU
Tasks Trace step and L-1 SRCU steps.  Please note that lockdep will not
detect these deadlocks until synchronize_rcu_tasks_trace() is marked
with lockdep's new "sync" annotation, which will probably not happen
until some time after these markings prove their worth on SRCU.

Please note that these tests are available only in kernels built with
CONFIG_TASKS_TRACE_RCU=y.

Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: Boqun Feng <boqun.feng@gmail.com>
kernel/rcu/rcutorture.c

index 80ff9a7..9efb825 100644 (file)
@@ -3604,12 +3604,46 @@ static void rcu_torture_init_srcu_lockdep(void)
                return;
        }
 
+#ifdef CONFIG_TASKS_TRACE_RCU
+       if (testtype == 3) {
+               pr_info("%s: test_srcu_lockdep = %05d: SRCU and Tasks Trace RCU %d-way %sdeadlock.\n",
+                       __func__, test_srcu_lockdep, cyclelen, deadlock ? "" : "non-");
+               if (deadlock && cyclelen == 1)
+                       pr_info("%s: Expect hang.\n", __func__);
+               for (i = 0; i < cyclelen; i++) {
+                       char *fl = i == 0 ? "rcu_read_lock_trace" : "srcu_read_lock";
+                       char *fs = i == cyclelen - 1 ? "synchronize_rcu_tasks_trace"
+                                                    : "synchronize_srcu";
+                       char *fu = i == 0 ? "rcu_read_unlock_trace" : "srcu_read_unlock";
+
+                       j = srcu_lockdep_next(__func__, fl, fs, fu, i, cyclelen, deadlock);
+                       if (i == 0)
+                               rcu_read_lock_trace();
+                       else
+                               idx = srcu_read_lock(srcus[i]);
+                       if (j >= 0) {
+                               if (i == cyclelen - 1)
+                                       synchronize_rcu_tasks_trace();
+                               else
+                                       synchronize_srcu(srcus[j]);
+                       }
+                       if (i == 0)
+                               rcu_read_unlock_trace();
+                       else
+                               srcu_read_unlock(srcus[i], idx);
+               }
+               return;
+       }
+#endif // #ifdef CONFIG_TASKS_TRACE_RCU
+
 err_out:
        pr_info("%s: test_srcu_lockdep = %05d does nothing.\n", __func__, test_srcu_lockdep);
        pr_info("%s: test_srcu_lockdep = DNNL.\n", __func__);
        pr_info("%s: D: Deadlock if nonzero.\n", __func__);
-       pr_info("%s: NN: Test number, 0=SRCU, 1=SRCU/mutex, 2=SRCU/rwsem.\n", __func__);
+       pr_info("%s: NN: Test number, 0=SRCU, 1=SRCU/mutex, 2=SRCU/rwsem, 3=SRCU/Tasks Trace RCU.\n", __func__);
        pr_info("%s: L: Cycle length.\n", __func__);
+       if (!IS_ENABLED(CONFIG_TASKS_TRACE_RCU))
+               pr_info("%s: NN=3 disallowed because kernel is built with CONFIG_TASKS_TRACE_RCU=n\n", __func__);
 }
 
 static int __init