rcu/nocb: Add bypass callback queueing

Use of the rcu_data structure's segmented ->cblist for no-CBs CPUs
takes advantage of unrelated grace periods, thus reducing the memory
footprint in the face of floods of call_rcu() invocations.  However,
the ->cblist field is a more-complex rcu_segcblist structure which must
be protected via locking.  Even though there are only three entities
which can acquire this lock (the CPU invoking call_rcu(), the no-CBs
grace-period kthread, and the no-CBs callbacks kthread), the contention
on this lock is excessive under heavy stress.

This commit therefore greatly reduces contention by provisioning
an rcu_cblist structure field named ->nocb_bypass within the
rcu_data structure.  Each no-CBs CPU is permitted only a limited
number of enqueues onto the ->cblist per jiffy, controlled by a new
nocb_nobypass_lim_per_jiffy kernel boot parameter that defaults to
about 16 enqueues per millisecond (16 * 1000 / HZ).  When that limit is
exceeded, the CPU instead enqueues onto the new ->nocb_bypass.

The ->nocb_bypass is flushed into the ->cblist every jiffy or when
the number of callbacks on ->nocb_bypass exceeds qhimark, whichever
happens first.  During call_rcu() floods, this flushing is carried out
by the CPU during the course of its call_rcu() invocations.  However,
a CPU could simply stop invoking call_rcu() at any time.  The no-CBs
grace-period kthread therefore carries out less-aggressive flushing
(every few jiffies or when the number of callbacks on ->nocb_bypass
exceeds (2 * qhimark), whichever comes first).  This means that the
no-CBs grace-period kthread cannot be permitted to do unbounded waits
while there are callbacks on ->nocb_bypass.  A ->nocb_bypass_timer is
used to provide the needed wakeups.

[ paulmck: Apply Coverity feedback reported by Colin Ian King. ]
Signed-off-by: Paul E. McKenney <paulmck@linux.ibm.com>

kernel/rcu/rcu_segcblist.c

@@ -36,6 +36,36 @@ void rcu_cblist_enqueue(struct rcu_cblist *rclp, struct rcu_head *rhp)
 	WRITE_ONCE(rclp->len, rclp->len + 1);
 }
 
+/*
+ * Flush the second rcu_cblist structure onto the first one, obliterating
+ * any contents of the first.  If rhp is non-NULL, enqueue it as the sole
+ * element of the second rcu_cblist structure, but ensuring that the second
+ * rcu_cblist structure, if initially non-empty, always appears non-empty
+ * throughout the process.  If rdp is NULL, the second rcu_cblist structure
+ * is instead initialized to empty.
+ */
+void rcu_cblist_flush_enqueue(struct rcu_cblist *drclp,
+			      struct rcu_cblist *srclp,
+			      struct rcu_head *rhp)
+{
+	drclp->head = srclp->head;
+	if (drclp->head)
+		drclp->tail = srclp->tail;
+	else
+		drclp->tail = &drclp->head;
+	drclp->len = srclp->len;
+	drclp->len_lazy = srclp->len_lazy;
+	if (!rhp) {
+		rcu_cblist_init(srclp);
+	} else {
+		rhp->next = NULL;
+		srclp->head = rhp;
+		srclp->tail = &rhp->next;
+		WRITE_ONCE(srclp->len, 1);
+		srclp->len_lazy = 0;
+	}
+}
+
 /*
  * Dequeue the oldest rcu_head structure from the specified callback
  * list.  This function assumes that the callback is non-lazy, but

kernel/rcu/rcu_segcblist.h

@@ -25,6 +25,10 @@ static inline void rcu_cblist_dequeued_lazy(struct rcu_cblist *rclp)
 }
 
 void rcu_cblist_init(struct rcu_cblist *rclp);
+void rcu_cblist_enqueue(struct rcu_cblist *rclp, struct rcu_head *rhp);
+void rcu_cblist_flush_enqueue(struct rcu_cblist *drclp,
+			      struct rcu_cblist *srclp,
+			      struct rcu_head *rhp);
 struct rcu_head *rcu_cblist_dequeue(struct rcu_cblist *rclp);
 
 /*
@@ -92,6 +96,7 @@ static inline bool rcu_segcblist_restempty(struct rcu_segcblist *rsclp, int seg)
 	return !READ_ONCE(*READ_ONCE(rsclp->tails[seg]));
 }
 
+void rcu_segcblist_inc_len(struct rcu_segcblist *rsclp);
 void rcu_segcblist_init(struct rcu_segcblist *rsclp);
 void rcu_segcblist_disable(struct rcu_segcblist *rsclp);
 void rcu_segcblist_offload(struct rcu_segcblist *rsclp);

kernel/rcu/tree.c

@@ -1251,6 +1251,7 @@ static bool rcu_accelerate_cbs(struct rcu_node *rnp, struct rcu_data *rdp)
 	unsigned long gp_seq_req;
 	bool ret = false;
 
+	rcu_lockdep_assert_cblist_protected(rdp);
 	raw_lockdep_assert_held_rcu_node(rnp);
 
 	/* If no pending (not yet ready to invoke) callbacks, nothing to do. */
@@ -1292,7 +1293,7 @@ static void rcu_accelerate_cbs_unlocked(struct rcu_node *rnp,
 	unsigned long c;
 	bool needwake;
 
-	lockdep_assert_irqs_disabled();
+	rcu_lockdep_assert_cblist_protected(rdp);
 	c = rcu_seq_snap(&rcu_state.gp_seq);
 	if (!rdp->gpwrap && ULONG_CMP_GE(rdp->gp_seq_needed, c)) {
 		/* Old request still live, so mark recent callbacks. */
@@ -1318,6 +1319,7 @@ static void rcu_accelerate_cbs_unlocked(struct rcu_node *rnp,
  */
 static bool rcu_advance_cbs(struct rcu_node *rnp, struct rcu_data *rdp)
 {
+	rcu_lockdep_assert_cblist_protected(rdp);
 	raw_lockdep_assert_held_rcu_node(rnp);
 
 	/* If no pending (not yet ready to invoke) callbacks, nothing to do. */
@@ -1341,6 +1343,7 @@ static bool rcu_advance_cbs(struct rcu_node *rnp, struct rcu_data *rdp)
 static void __maybe_unused rcu_advance_cbs_nowake(struct rcu_node *rnp,
 						  struct rcu_data *rdp)
 {
+	rcu_lockdep_assert_cblist_protected(rdp);
 	if (!rcu_seq_state(rcu_seq_current(&rnp->gp_seq)) ||
 	    !raw_spin_trylock_rcu_node(rnp))
 		return;
@@ -2187,7 +2190,9 @@ static void rcu_do_batch(struct rcu_data *rdp)
 	 * The following usually indicates a double call_rcu().  To track
 	 * this down, try building with CONFIG_DEBUG_OBJECTS_RCU_HEAD=y.
 	 */
-	WARN_ON_ONCE(rcu_segcblist_empty(&rdp->cblist) != (count == 0));
+	WARN_ON_ONCE(count == 0 && !rcu_segcblist_empty(&rdp->cblist));
+	WARN_ON_ONCE(!IS_ENABLED(CONFIG_RCU_NOCB_CPU) &&
+		     count != 0 && rcu_segcblist_empty(&rdp->cblist));
 
 	rcu_nocb_unlock_irqrestore(rdp, flags);
 
@@ -2564,8 +2569,9 @@ __call_rcu(struct rcu_head *head, rcu_callback_t func, bool lazy)
 		if (rcu_segcblist_empty(&rdp->cblist))
 			rcu_segcblist_init(&rdp->cblist);
 	}
-	rcu_nocb_lock(rdp);
-	was_alldone = !rcu_segcblist_pend_cbs(&rdp->cblist);
+	if (rcu_nocb_try_bypass(rdp, head, &was_alldone, flags))
+		return; // Enqueued onto ->nocb_bypass, so just leave.
+	/* If we get here, rcu_nocb_try_bypass() acquired ->nocb_lock. */
 	rcu_segcblist_enqueue(&rdp->cblist, head, lazy);
 	if (__is_kfree_rcu_offset((unsigned long)func))
 		trace_rcu_kfree_callback(rcu_state.name, head,
@@ -2839,6 +2845,7 @@ static void rcu_barrier_func(void *unused)
 	rdp->barrier_head.func = rcu_barrier_callback;
 	debug_rcu_head_queue(&rdp->barrier_head);
 	rcu_nocb_lock(rdp);
+	WARN_ON_ONCE(!rcu_nocb_flush_bypass(rdp, NULL, jiffies));
 	if (rcu_segcblist_entrain(&rdp->cblist, &rdp->barrier_head, 0)) {
 		atomic_inc(&rcu_state.barrier_cpu_count);
 	} else {
@@ -3192,6 +3199,7 @@ void rcutree_migrate_callbacks(int cpu)
 	my_rdp = this_cpu_ptr(&rcu_data);
 	my_rnp = my_rdp->mynode;
 	rcu_nocb_lock(my_rdp); /* irqs already disabled. */
+	WARN_ON_ONCE(!rcu_nocb_flush_bypass(my_rdp, NULL, jiffies));
 	raw_spin_lock_rcu_node(my_rnp); /* irqs already disabled. */
 	/* Leverage recent GPs and set GP for new callbacks. */
 	needwake = rcu_advance_cbs(my_rnp, rdp) ||

kernel/rcu/tree.h

@@ -200,18 +200,26 @@ struct rcu_data {
 	atomic_t nocb_lock_contended;	/* Contention experienced. */
 	int nocb_defer_wakeup;		/* Defer wakeup of nocb_kthread. */
 	struct timer_list nocb_timer;	/* Enforce finite deferral. */
+	unsigned long nocb_gp_adv_time;	/* Last call_rcu() CB adv (jiffies). */
+
+	/* The following fields are used by call_rcu, hence own cacheline. */
+	raw_spinlock_t nocb_bypass_lock ____cacheline_internodealigned_in_smp;
+	struct rcu_cblist nocb_bypass;	/* Lock-contention-bypass CB list. */
+	unsigned long nocb_bypass_first; /* Time (jiffies) of first enqueue. */
+	unsigned long nocb_nobypass_last; /* Last ->cblist enqueue (jiffies). */
+	int nocb_nobypass_count;	/* # ->cblist enqueues at ^^^ time. */
 
 	/* The following fields are used by GP kthread, hence own cacheline. */
 	raw_spinlock_t nocb_gp_lock ____cacheline_internodealigned_in_smp;
-	bool nocb_gp_sleep;
-					/* Is the nocb GP thread asleep? */
+	struct timer_list nocb_bypass_timer; /* Force nocb_bypass flush. */
+	bool nocb_gp_sleep;		/* Is the nocb GP thread asleep? */
 	struct swait_queue_head nocb_gp_wq; /* For nocb kthreads to sleep on. */
 	bool nocb_cb_sleep;		/* Is the nocb CB thread asleep? */
 	struct task_struct *nocb_cb_kthread;
 	struct rcu_data *nocb_next_cb_rdp;
 					/* Next rcu_data in wakeup chain. */
 
-	/* The following fields are used by CB kthread, hence new cachline. */
+	/* The following fields are used by CB kthread, hence new cacheline. */
 	struct rcu_data *nocb_gp_rdp ____cacheline_internodealigned_in_smp;
 					/* GP rdp takes GP-end wakeups. */
 #endif /* #ifdef CONFIG_RCU_NOCB_CPU */
@@ -419,6 +427,10 @@ static void zero_cpu_stall_ticks(struct rcu_data *rdp);
 static struct swait_queue_head *rcu_nocb_gp_get(struct rcu_node *rnp);
 static void rcu_nocb_gp_cleanup(struct swait_queue_head *sq);
 static void rcu_init_one_nocb(struct rcu_node *rnp);
+static bool rcu_nocb_flush_bypass(struct rcu_data *rdp, struct rcu_head *rhp,
+				  unsigned long j);
+static bool rcu_nocb_try_bypass(struct rcu_data *rdp, struct rcu_head *rhp,
+				bool *was_alldone, unsigned long flags);
 static void __call_rcu_nocb_wake(struct rcu_data *rdp, bool was_empty,
 				 unsigned long flags);
 static int rcu_nocb_need_deferred_wakeup(struct rcu_data *rdp);
@@ -430,19 +442,15 @@ static void rcu_nocb_lock(struct rcu_data *rdp);
 static void rcu_nocb_unlock(struct rcu_data *rdp);
 static void rcu_nocb_unlock_irqrestore(struct rcu_data *rdp,
 				       unsigned long flags);
+static void rcu_lockdep_assert_cblist_protected(struct rcu_data *rdp);
 #ifdef CONFIG_RCU_NOCB_CPU
 static void __init rcu_organize_nocb_kthreads(void);
 #define rcu_nocb_lock_irqsave(rdp, flags)				\
 do {									\
-	if (!rcu_segcblist_is_offloaded(&(rdp)->cblist)) {		\
+	if (!rcu_segcblist_is_offloaded(&(rdp)->cblist))		\
 		local_irq_save(flags);					\
-	} else if (!raw_spin_trylock_irqsave(&(rdp)->nocb_lock, (flags))) {\
-		atomic_inc(&(rdp)->nocb_lock_contended);		\
-		smp_mb__after_atomic(); /* atomic_inc() before lock. */	\
+	else								\
 		raw_spin_lock_irqsave(&(rdp)->nocb_lock, (flags));	\
-		smp_mb__before_atomic(); /* atomic_dec() after lock. */	\
-		atomic_dec(&(rdp)->nocb_lock_contended);		\
-	}								\
 } while (0)
 #else /* #ifdef CONFIG_RCU_NOCB_CPU */
 #define rcu_nocb_lock_irqsave(rdp, flags) local_irq_save(flags)