irqwork: push most work into softirq context

Initially we defered all irqwork into softirq because we didn't want the
latency spikes if perf or another user was busy and delayed the RT task.
The NOHZ trigger (nohz_full_kick_work) was the first user that did not work
as expected if it did not run in the original irqwork context so we had to
bring it back somehow for it. push_irq_work_func is the second one that
requires this.

This patch adds the IRQ_WORK_HARD_IRQ which makes sure the callback runs
in raw-irq context. Everything else is defered into softirq context. Without
-RT we have the orignal behavior.

This patch incorporates tglx orignal work which revoked a little bringing back
the arch_irq_work_raise() if possible and a few fixes from Steven Rostedt and
Mike Galbraith,

Signed-off-by: Sebastian Andrzej Siewior <[email protected]>

include/linux/irq_work.h

@@ -16,6 +16,7 @@
 #define IRQ_WORK_BUSY		2UL
 #define IRQ_WORK_FLAGS		3UL
 #define IRQ_WORK_LAZY		4UL /* Doesn't want IPI, wait for tick */
+#define IRQ_WORK_HARD_IRQ	8UL /* Run hard IRQ context, even on RT */
 
 struct irq_work {
 	unsigned long flags;

kernel/irq_work.c

@@ -17,6 +17,7 @@
 #include <linux/cpu.h>
 #include <linux/notifier.h>
 #include <linux/smp.h>
+#include <linux/interrupt.h>
 #include <asm/processor.h>
 
 
@@ -65,6 +66,8 @@ void __weak arch_irq_work_raise(void)
  */
 bool irq_work_queue_on(struct irq_work *work, int cpu)
 {
+	struct llist_head *list;
+
 	/* All work should have been flushed before going offline */
 	WARN_ON_ONCE(cpu_is_offline(cpu));
 
@@ -75,7 +78,12 @@ bool irq_work_queue_on(struct irq_work *work, int cpu)
 	if (!irq_work_claim(work))
 		return false;
 
-	if (llist_add(&work->llnode, &per_cpu(raised_list, cpu)))
+	if (IS_ENABLED(CONFIG_PREEMPT_RT_FULL) && !(work->flags & IRQ_WORK_HARD_IRQ))
+		list = &per_cpu(lazy_list, cpu);
+	else
+		list = &per_cpu(raised_list, cpu);
+
+	if (llist_add(&work->llnode, list))
 		arch_send_call_function_single_ipi(cpu);
 
 	return true;
@@ -86,20 +94,25 @@ EXPORT_SYMBOL_GPL(irq_work_queue_on);
 /* Enqueue the irq work @work on the current CPU */
 bool irq_work_queue(struct irq_work *work)
 {
+	struct llist_head *list;
+	bool lazy_work, realtime = IS_ENABLED(CONFIG_PREEMPT_RT_FULL);
+
 	/* Only queue if not already pending */
 	if (!irq_work_claim(work))
 		return false;
 
 	/* Queue the entry and raise the IPI if needed. */
 	preempt_disable();
 
-	/* If the work is "lazy", handle it from next tick if any */
-	if (work->flags & IRQ_WORK_LAZY) {
-		if (llist_add(&work->llnode, this_cpu_ptr(&lazy_list)) &&
-		    tick_nohz_tick_stopped())
-			arch_irq_work_raise();
-	} else {
-		if (llist_add(&work->llnode, this_cpu_ptr(&raised_list)))
+	lazy_work = work->flags & IRQ_WORK_LAZY;
+
+	if (lazy_work || (realtime && !(work->flags & IRQ_WORK_HARD_IRQ)))
+		list = this_cpu_ptr(&lazy_list);
+	else
+		list = this_cpu_ptr(&raised_list);
+
+	if (llist_add(&work->llnode, list)) {
+		if (!lazy_work || tick_nohz_tick_stopped())
 			arch_irq_work_raise();
 	}
 
@@ -116,9 +129,8 @@ bool irq_work_needs_cpu(void)
 	raised = this_cpu_ptr(&raised_list);
 	lazy = this_cpu_ptr(&lazy_list);
 
-	if (llist_empty(raised) || arch_irq_work_has_interrupt())
-		if (llist_empty(lazy))
-			return false;
+	if (llist_empty(raised) && llist_empty(lazy))
+		return false;
 
 	/* All work should have been flushed before going offline */
 	WARN_ON_ONCE(cpu_is_offline(smp_processor_id()));
@@ -132,7 +144,7 @@ static void irq_work_run_list(struct llist_head *list)
 	struct irq_work *work;
 	struct llist_node *llnode;
 
-	BUG_ON(!irqs_disabled());
+	BUG_ON_NONRT(!irqs_disabled());
 
 	if (llist_empty(list))
 		return;
@@ -169,7 +181,16 @@ static void irq_work_run_list(struct llist_head *list)
 void irq_work_run(void)
 {
 	irq_work_run_list(this_cpu_ptr(&raised_list));
-	irq_work_run_list(this_cpu_ptr(&lazy_list));
+	if (IS_ENABLED(CONFIG_PREEMPT_RT_FULL)) {
+		/*
+		 * NOTE: we raise softirq via IPI for safety,
+		 * and execute in irq_work_tick() to move the
+		 * overhead from hard to soft irq context.
+		 */
+		if (!llist_empty(this_cpu_ptr(&lazy_list)))
+			raise_softirq(TIMER_SOFTIRQ);
+	} else
+		irq_work_run_list(this_cpu_ptr(&lazy_list));
 }
 EXPORT_SYMBOL_GPL(irq_work_run);
 

kernel/sched/core.c

@@ -6046,6 +6046,7 @@ static int init_rootdomain(struct root_domain *rd)
 	rd->rto_cpu = -1;
 	raw_spin_lock_init(&rd->rto_lock);
 	init_irq_work(&rd->rto_push_work, rto_push_irq_work_func);
+	rd->rto_push_work.flags |= IRQ_WORK_HARD_IRQ;
 #endif
 
 	init_dl_bw(&rd->dl_bw);

kernel/time/tick-sched.c

@@ -220,6 +220,7 @@ static void nohz_full_kick_func(struct irq_work *work)
 
 static DEFINE_PER_CPU(struct irq_work, nohz_full_kick_work) = {
 	.func = nohz_full_kick_func,
+	.flags = IRQ_WORK_HARD_IRQ,
 };
 
 /*

kernel/time/timer.c

@@ -1661,7 +1661,7 @@ void update_process_times(int user_tick)
 	scheduler_tick();
 	run_local_timers();
 	rcu_check_callbacks(user_tick);
-#ifdef CONFIG_IRQ_WORK
+#if defined(CONFIG_IRQ_WORK) && !defined(CONFIG_PREEMPT_RT_FULL)
 	if (in_irq())
 		irq_work_tick();
 #endif
@@ -1717,6 +1717,10 @@ static __latent_entropy void run_timer_softirq(struct softirq_action *h)
 {
 	struct timer_base *base = this_cpu_ptr(&timer_bases[BASE_STD]);
 
+#if defined(CONFIG_IRQ_WORK) && defined(CONFIG_PREEMPT_RT_FULL)
+	irq_work_tick();
+#endif
+
 	__run_timers(base);
 	if (IS_ENABLED(CONFIG_NO_HZ_COMMON))
 		__run_timers(this_cpu_ptr(&timer_bases[BASE_DEF]));