Re: [PATCH v3 3/4] rcu/nocb: Protect NOCB state via local_lock() under PREEMPT_RT

On 17/08/21 17:36, Sebastian Andrzej Siewior wrote:

quoted

+static inline int rcu_lockdep_is_held_nocb_local(struct rcu_data *rdp)
+{
+	return lockdep_is_held(
+#ifdef CONFIG_PREEMPT_RT
+		&rdp->nocb_local_lock.lock
+#else
+		&rdp->nocb_local_lock
+#endif
+	);
+}

Now that I see it and Paul asked for it, please just use !RT version.
      return lockdep_is_held(&rdp->nocb_local_lock);

and RT will work, too.

The above was required due to the (previous) definition of local_lock_t:

  typedef struct {
          spinlock_t		lock;
  } local_lock_t;

On -rt11 I see this is now:

  typedef spinlock_t local_lock_t;

which indeed means the iffdefery can (actually it *has* to) go.

quoted

 static inline bool rcu_current_is_nocb_kthread(struct rcu_data *rdp)
 {
     /* Race on early boot between thread creation and assignment */

@@ -1629,6 +1664,22 @@ static void rcu_nocb_unlock_irqrestore(struct rcu_data *rdp,
     }
 }

+/*
+ * The invocation of rcu_core() within the RCU core kthreads remains preemptible
+ * under PREEMPT_RT, thus the offload state of a CPU could change while
+ * said kthreads are preempted. Prevent this from happening by protecting the
+ * offload state with a local_lock().
+ */
+static void rcu_nocb_local_lock(struct rcu_data *rdp)
+{
+	local_lock(&rcu_data.nocb_local_lock);
+}
+
+static void rcu_nocb_local_unlock(struct rcu_data *rdp)
+{
+	local_unlock(&rcu_data.nocb_local_lock);
+}
+

Do you need to pass rdp given that it is not used?

Not anymore, you're right.

quoted

 /* Lockdep check that ->cblist may be safely accessed. */
 static void rcu_lockdep_assert_cblist_protected(struct rcu_data *rdp)
 {

Sebastian