time: tick-sched: Split tick_nohz_stop_sched_tick()

In order to address the issue with short idle duration predictions
by the idle governor after the scheduler tick has been stopped, split
tick_nohz_stop_sched_tick() into two separate routines, one computing
the time to the next timer event and the other simply stopping the
tick when the time to the next timer event is known.

Prepare these two routines to be called separately, as one of them
will be called by the idle governor in the cpuidle_select() code
path after subsequent changes.

Update the former callers of tick_nohz_stop_sched_tick() to use
the new routines, tick_nohz_next_event() and tick_nohz_stop_tick(),
instead of it and move the updates of the sleep_length field in
struct tick_sched into __tick_nohz_idle_stop_tick() as it doesn't
need to be updated anywhere else.

There should be no intentional visible changes in functionality
resulting from this change.

Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Frederic Weisbecker <frederic@kernel.org>

2 files changed, 82 insertions, 46 deletions

diff --git a/kernel/time/tick-sched.c b/kernel/time/tick-sched.c
index 69fe113..f56d2c6 100644
--- a/kernel/time/tick-sched.c
+++ b/kernel/time/tick-sched.c
@@ -641,13 +641,10 @@ static inline bool local_timer_softirq_pending(void)
 	return local_softirq_pending() & TIMER_SOFTIRQ;
 }
 
-static ktime_t tick_nohz_stop_sched_tick(struct tick_sched *ts,
-					 ktime_t now, int cpu)
+static ktime_t tick_nohz_next_event(struct tick_sched *ts, int cpu)
 {
-	struct clock_event_device *dev = __this_cpu_read(tick_cpu_device.evtdev);
 	u64 basemono, next_tick, next_tmr, next_rcu, delta, expires;
 	unsigned long seq, basejiff;
-	ktime_t	tick;
 
 	/* Read jiffies and the time when jiffies were updated last */
 	do {
@@ -656,6 +653,7 @@ static ktime_t tick_nohz_stop_sched_tick(struct tick_sched *ts,
 		basejiff = jiffies;
 	} while (read_seqretry(&jiffies_lock, seq));
 	ts->last_jiffies = basejiff;
+	ts->timer_expires_base = basemono;
 
 	/*
 	 * Keep the periodic tick, when RCU, architecture or irq_work
@@ -700,47 +698,63 @@ static ktime_t tick_nohz_stop_sched_tick(struct tick_sched *ts,
 		 * next period, so no point in stopping it either, bail.
 		 */
 		if (!ts->tick_stopped) {
-			tick = 0;
+			ts->timer_expires = 0;
 			goto out;
 		}
 	}
 
 	/*
+	 * If this CPU is the one which had the do_timer() duty last, we limit
+	 * the sleep time to the timekeeping max_deferment value.
+	 * Otherwise we can sleep as long as we want.
+	 */
+	delta = timekeeping_max_deferment();
+	if (cpu != tick_do_timer_cpu &&
+	    (tick_do_timer_cpu != TICK_DO_TIMER_NONE || !ts->do_timer_last))
+		delta = KTIME_MAX;
+
+	/* Calculate the next expiry time */
+	if (delta < (KTIME_MAX - basemono))
+		expires = basemono + delta;
+	else
+		expires = KTIME_MAX;
+
+	ts->timer_expires = min_t(u64, expires, next_tick);
+
+out:
+	return ts->timer_expires;
+}
+
+static void tick_nohz_stop_tick(struct tick_sched *ts, int cpu)
+{
+	struct clock_event_device *dev = __this_cpu_read(tick_cpu_device.evtdev);
+	u64 basemono = ts->timer_expires_base;
+	u64 expires = ts->timer_expires;
+	ktime_t tick = expires;
+
+	/* Make sure we won't be trying to stop it twice in a row. */
+	ts->timer_expires_base = 0;
+
+	/*
 	 * If this CPU is the one which updates jiffies, then give up
 	 * the assignment and let it be taken by the CPU which runs
 	 * the tick timer next, which might be this CPU as well. If we
 	 * don't drop this here the jiffies might be stale and
 	 * do_timer() never invoked. Keep track of the fact that it
-	 * was the one which had the do_timer() duty last. If this CPU
-	 * is the one which had the do_timer() duty last, we limit the
-	 * sleep time to the timekeeping max_deferment value.
-	 * Otherwise we can sleep as long as we want.
+	 * was the one which had the do_timer() duty last.
 	 */
-	delta = timekeeping_max_deferment();
 	if (cpu == tick_do_timer_cpu) {
 		tick_do_timer_cpu = TICK_DO_TIMER_NONE;
 		ts->do_timer_last = 1;
 	} else if (tick_do_timer_cpu != TICK_DO_TIMER_NONE) {
-		delta = KTIME_MAX;
 		ts->do_timer_last = 0;
-	} else if (!ts->do_timer_last) {
-		delta = KTIME_MAX;
 	}
 
-	/* Calculate the next expiry time */
-	if (delta < (KTIME_MAX - basemono))
-		expires = basemono + delta;
-	else
-		expires = KTIME_MAX;
-
-	expires = min_t(u64, expires, next_tick);
-	tick = expires;
-
 	/* Skip reprogram of event if its not changed */
 	if (ts->tick_stopped && (expires == ts->next_tick)) {
 		/* Sanity check: make sure clockevent is actually programmed */
 		if (tick == KTIME_MAX || ts->next_tick == hrtimer_get_expires(&ts->sched_timer))
-			goto out;
+			return;
 
 		WARN_ON_ONCE(1);
 		printk_once("basemono: %llu ts->next_tick: %llu dev->next_event: %llu timer->active: %d timer->expires: %llu\n",
@@ -774,7 +788,7 @@ static ktime_t tick_nohz_stop_sched_tick(struct tick_sched *ts,
 	if (unlikely(expires == KTIME_MAX)) {
 		if (ts->nohz_mode == NOHZ_MODE_HIGHRES)
 			hrtimer_cancel(&ts->sched_timer);
-		goto out;
+		return;
 	}
 
 	hrtimer_set_expires(&ts->sched_timer, tick);
@@ -783,15 +797,23 @@ static ktime_t tick_nohz_stop_sched_tick(struct tick_sched *ts,
 		hrtimer_start_expires(&ts->sched_timer, HRTIMER_MODE_ABS_PINNED);
 	else
 		tick_program_event(tick, 1);
-out:
-	/*
-	 * Update the estimated sleep length until the next timer
-	 * (not only the tick).
-	 */
-	ts->sleep_length = ktime_sub(dev->next_event, now);
-	return tick;
 }
 
+static void tick_nohz_retain_tick(struct tick_sched *ts)
+{
+	ts->timer_expires_base = 0;
+}
+
+#ifdef CONFIG_NO_HZ_FULL
+static void tick_nohz_stop_sched_tick(struct tick_sched *ts, int cpu)
+{
+	if (tick_nohz_next_event(ts, cpu))
+		tick_nohz_stop_tick(ts, cpu);
+	else
+		tick_nohz_retain_tick(ts);
+}
+#endif /* CONFIG_NO_HZ_FULL */
+
 static void tick_nohz_restart_sched_tick(struct tick_sched *ts, ktime_t now)
 {
 	/* Update jiffies first */
@@ -827,7 +849,7 @@ static void tick_nohz_full_update_tick(struct tick_sched *ts)
 		return;
 
 	if (can_stop_full_tick(cpu, ts))
-		tick_nohz_stop_sched_tick(ts, ktime_get(), cpu);
+		tick_nohz_stop_sched_tick(ts, cpu);
 	else if (ts->tick_stopped)
 		tick_nohz_restart_sched_tick(ts, ktime_get());
 #endif
@@ -853,10 +875,8 @@ static bool can_stop_idle_tick(int cpu, struct tick_sched *ts)
 		return false;
 	}
 
-	if (unlikely(ts->nohz_mode == NOHZ_MODE_INACTIVE)) {
-		ts->sleep_length = NSEC_PER_SEC / HZ;
+	if (unlikely(ts->nohz_mode == NOHZ_MODE_INACTIVE))
 		return false;
-	}
 
 	if (need_resched())
 		return false;
@@ -893,29 +913,37 @@ static bool can_stop_idle_tick(int cpu, struct tick_sched *ts)
 
 static void __tick_nohz_idle_stop_tick(struct tick_sched *ts)
 {
+	struct clock_event_device *dev = __this_cpu_read(tick_cpu_device.evtdev);
 	ktime_t expires;
 	int cpu = smp_processor_id();
 
-	if (can_stop_idle_tick(cpu, ts)) {
+	WARN_ON_ONCE(ts->timer_expires_base);
+
+	if (!can_stop_idle_tick(cpu, ts))
+		goto out;
+
+	expires = tick_nohz_next_event(ts, cpu);
+
+	ts->idle_calls++;
+
+	if (expires > 0LL) {
 		int was_stopped = ts->tick_stopped;
 
-		ts->idle_calls++;
+		tick_nohz_stop_tick(ts, cpu);
 
-		/*
-		 * The idle entry time should be a sufficient approximation of
-		 * the current time at this point.
-		 */
-		expires = tick_nohz_stop_sched_tick(ts, ts->idle_entrytime, cpu);
-		if (expires > 0LL) {
-			ts->idle_sleeps++;
-			ts->idle_expires = expires;
-		}
+		ts->idle_sleeps++;
+		ts->idle_expires = expires;
 
 		if (!was_stopped && ts->tick_stopped) {
 			ts->idle_jiffies = ts->last_jiffies;
 			nohz_balance_enter_idle(cpu);
 		}
+	} else {
+		tick_nohz_retain_tick(ts);
 	}
+
+out:
+	ts->sleep_length = ktime_sub(dev->next_event, ts->idle_entrytime);
 }
 
 /**
@@ -942,6 +970,9 @@ void tick_nohz_idle_enter(void)
 	local_irq_disable();
 
 	ts = this_cpu_ptr(&tick_cpu_sched);
+
+	WARN_ON_ONCE(ts->timer_expires_base);
+
 	ts->inidle = 1;
 	tick_nohz_start_idle(ts);
 
@@ -1067,6 +1098,7 @@ void tick_nohz_idle_exit(void)
 	local_irq_disable();
 
 	WARN_ON_ONCE(!ts->inidle);
+	WARN_ON_ONCE(ts->timer_expires_base);
 
 	ts->inidle = 0;
 
diff --git a/kernel/time/tick-sched.h b/kernel/time/tick-sched.h
index 954b43d..53e45a3 100644
--- a/kernel/time/tick-sched.h
+++ b/kernel/time/tick-sched.h
@@ -39,6 +39,8 @@ enum tick_nohz_mode {
  * @idle_sleeptime:	Sum of the time slept in idle with sched tick stopped
  * @iowait_sleeptime:	Sum of the time slept in idle with sched tick stopped, with IO outstanding
  * @sleep_length:	Duration of the current idle sleep
+ * @timer_expires:	Anticipated timer expiration time (in case sched tick is stopped)
+ * @timer_expires_base:	Base time clock monotonic for @timer_expires
  * @do_timer_lst:	CPU was the last one doing do_timer before going idle
  */
 struct tick_sched {
@@ -60,6 +62,8 @@ struct tick_sched {
 	ktime_t				iowait_sleeptime;
 	ktime_t				sleep_length;
 	unsigned long			last_jiffies;
+	u64				timer_expires;
+	u64				timer_expires_base;
 	u64				next_timer;
 	ktime_t				idle_expires;
 	int				do_timer_last;