2 files changed, 120 insertions, 190 deletions

diff --git a/drivers/cpuidle/governors/ladder.c b/drivers/cpuidle/governors/ladder.c
index 9b784051ec1..9f08e8cce1a 100644
--- a/drivers/cpuidle/governors/ladder.c
+++ b/drivers/cpuidle/governors/ladder.c
@@ -192,14 +192,4 @@ static int __init init_ladder(void)
 	return cpuidle_register_governor(&ladder_governor);
 }
 
-/**
- * exit_ladder - exits the governor
- */
-static void __exit exit_ladder(void)
-{
-	cpuidle_unregister_governor(&ladder_governor);
-}
-
-MODULE_LICENSE("GPL");
-module_init(init_ladder);
-module_exit(exit_ladder);
+postcore_initcall(init_ladder);
diff --git a/drivers/cpuidle/governors/menu.c b/drivers/cpuidle/governors/menu.c
index fe343a06b7d..c4f80c15a48 100644
--- a/drivers/cpuidle/governors/menu.c
+++ b/drivers/cpuidle/governors/menu.c
@@ -21,6 +21,15 @@
 #include <linux/math64.h>
 #include <linux/module.h>
 
+/*
+ * Please note when changing the tuning values:
+ * If (MAX_INTERESTING-1) * RESOLUTION > UINT_MAX, the result of
+ * a scaling operation multiplication may overflow on 32 bit platforms.
+ * In that case, #define RESOLUTION as ULL to get 64 bit result:
+ * #define RESOLUTION 1024ULL
+ *
+ * The default values do not overflow.
+ */
 #define BUCKETS 12
 #define INTERVALS 8
 #define RESOLUTION 1024
@@ -28,13 +37,6 @@
 #define MAX_INTERESTING 50000
 #define STDDEV_THRESH 400
 
-/* 60 * 60 > STDDEV_THRESH * INTERVALS = 400 * 8 */
-#define MAX_DEVIATION 60
-
-static DEFINE_PER_CPU(struct hrtimer, menu_hrtimer);
-static DEFINE_PER_CPU(int, hrtimer_status);
-/* menu hrtimer mode */
-enum {MENU_HRTIMER_STOP, MENU_HRTIMER_REPEAT, MENU_HRTIMER_GENERAL};
 
 /*
  * Concepts and ideas behind the menu governor
@@ -116,23 +118,15 @@ enum {MENU_HRTIMER_STOP, MENU_HRTIMER_REPEAT, MENU_HRTIMER_GENERAL};
  *
  */
 
-/*
- * The C-state residency is so long that is is worthwhile to exit
- * from the shallow C-state and re-enter into a deeper C-state.
- */
-static unsigned int perfect_cstate_ms __read_mostly = 30;
-module_param(perfect_cstate_ms, uint, 0000);
-
 struct menu_device {
 	int		last_state_idx;
 	int             needs_update;
 
-	unsigned int	expected_us;
-	u64		predicted_us;
-	unsigned int	exit_us;
+	unsigned int	next_timer_us;
+	unsigned int	predicted_us;
 	unsigned int	bucket;
-	u64		correction_factor[BUCKETS];
-	u32		intervals[INTERVALS];
+	unsigned int	correction_factor[BUCKETS];
+	unsigned int	intervals[INTERVALS];
 	int		interval_ptr;
 };
 
@@ -205,59 +199,28 @@ static u64 div_round64(u64 dividend, u32 divisor)
 	return div_u64(dividend + (divisor / 2), divisor);
 }
 
-/* Cancel the hrtimer if it is not triggered yet */
-void menu_hrtimer_cancel(void)
-{
-	int cpu = smp_processor_id();
-	struct hrtimer *hrtmr = &per_cpu(menu_hrtimer, cpu);
-
-	/* The timer is still not time out*/
-	if (per_cpu(hrtimer_status, cpu)) {
-		hrtimer_cancel(hrtmr);
-		per_cpu(hrtimer_status, cpu) = MENU_HRTIMER_STOP;
-	}
-}
-EXPORT_SYMBOL_GPL(menu_hrtimer_cancel);
-
-/* Call back for hrtimer is triggered */
-static enum hrtimer_restart menu_hrtimer_notify(struct hrtimer *hrtimer)
-{
-	int cpu = smp_processor_id();
-	struct menu_device *data = &per_cpu(menu_devices, cpu);
-
-	/* In general case, the expected residency is much larger than
-	 *  deepest C-state target residency, but prediction logic still
-	 *  predicts a small predicted residency, so the prediction
-	 *  history is totally broken if the timer is triggered.
-	 *  So reset the correction factor.
-	 */
-	if (per_cpu(hrtimer_status, cpu) == MENU_HRTIMER_GENERAL)
-		data->correction_factor[data->bucket] = RESOLUTION * DECAY;
-
-	per_cpu(hrtimer_status, cpu) = MENU_HRTIMER_STOP;
-
-	return HRTIMER_NORESTART;
-}
-
 /*
  * Try detecting repeating patterns by keeping track of the last 8
  * intervals, and checking if the standard deviation of that set
  * of points is below a threshold. If it is... then use the
  * average of these 8 points as the estimated value.
  */
-static u32 get_typical_interval(struct menu_device *data)
+static void get_typical_interval(struct menu_device *data)
 {
-	int i = 0, divisor = 0;
-	uint64_t max = 0, avg = 0, stddev = 0;
-	int64_t thresh = LLONG_MAX; /* Discard outliers above this value. */
-	unsigned int ret = 0;
+	int i, divisor;
+	unsigned int max, thresh;
+	uint64_t avg, stddev;
+
+	thresh = UINT_MAX; /* Discard outliers above this value */
 
 again:
 
-	/* first calculate average and standard deviation of the past */
-	max = avg = divisor = stddev = 0;
+	/* First calculate the average of past intervals */
+	max = 0;
+	avg = 0;
+	divisor = 0;
 	for (i = 0; i < INTERVALS; i++) {
-		int64_t value = data->intervals[i];
+		unsigned int value = data->intervals[i];
 		if (value <= thresh) {
 			avg += value;
 			divisor++;
@@ -267,15 +230,38 @@ again:
 	}
 	do_div(avg, divisor);
 
+	/* Then try to determine standard deviation */
+	stddev = 0;
 	for (i = 0; i < INTERVALS; i++) {
-		int64_t value = data->intervals[i];
+		unsigned int value = data->intervals[i];
 		if (value <= thresh) {
 			int64_t diff = value - avg;
 			stddev += diff * diff;
 		}
 	}
 	do_div(stddev, divisor);
-	stddev = int_sqrt(stddev);
+	/*
+	 * The typical interval is obtained when standard deviation is small
+	 * or standard deviation is small compared to the average interval.
+	 *
+	 * int_sqrt() formal parameter type is unsigned long. When the
+	 * greatest difference to an outlier exceeds ~65 ms * sqrt(divisor)
+	 * the resulting squared standard deviation exceeds the input domain
+	 * of int_sqrt on platforms where unsigned long is 32 bits in size.
+	 * In such case reject the candidate average.
+	 *
+	 * Use this result only if there is no timer to wake us up sooner.
+	 */
+	if (likely(stddev <= ULONG_MAX)) {
+		stddev = int_sqrt(stddev);
+		if (((avg > stddev * 6) && (divisor * 4 >= INTERVALS * 3))
+							|| stddev <= 20) {
+			if (data->next_timer_us > avg)
+				data->predicted_us = avg;
+			return;
+		}
+	}
+
 	/*
 	 * If we have outliers to the upside in our distribution, discard
 	 * those by setting the threshold to exclude these outliers, then
@@ -284,23 +270,12 @@ again:
 	 *
 	 * This can deal with workloads that have long pauses interspersed
 	 * with sporadic activity with a bunch of short pauses.
-	 *
-	 * The typical interval is obtained when standard deviation is small
-	 * or standard deviation is small compared to the average interval.
 	 */
-	if (((avg > stddev * 6) && (divisor * 4 >= INTERVALS * 3))
-							|| stddev <= 20) {
-		data->predicted_us = avg;
-		ret = 1;
-		return ret;
-
-	} else if ((divisor * 4) > INTERVALS * 3) {
-		/* Exclude the max interval */
-		thresh = max - 1;
-		goto again;
-	}
+	if ((divisor * 4) <= INTERVALS * 3)
+		return;
 
-	return ret;
+	thresh = max - 1;
+	goto again;
 }
 
 /**
@@ -313,19 +288,15 @@ static int menu_select(struct cpuidle_driver *drv, struct cpuidle_device *dev)
 	struct menu_device *data = &__get_cpu_var(menu_devices);
 	int latency_req = pm_qos_request(PM_QOS_CPU_DMA_LATENCY);
 	int i;
-	int multiplier;
+	unsigned int interactivity_req;
 	struct timespec t;
-	int repeat = 0, low_predicted = 0;
-	int cpu = smp_processor_id();
-	struct hrtimer *hrtmr = &per_cpu(menu_hrtimer, cpu);
 
 	if (data->needs_update) {
 		menu_update(drv, dev);
 		data->needs_update = 0;
 	}
 
-	data->last_state_idx = 0;
-	data->exit_us = 0;
+	data->last_state_idx = CPUIDLE_DRIVER_STATE_START - 1;
 
 	/* Special case when user has set very strict latency requirement */
 	if (unlikely(latency_req == 0))
@@ -333,32 +304,37 @@ static int menu_select(struct cpuidle_driver *drv, struct cpuidle_device *dev)
 
 	/* determine the expected residency time, round up */
 	t = ktime_to_timespec(tick_nohz_get_sleep_length());
-	data->expected_us =
+	data->next_timer_us =
 		t.tv_sec * USEC_PER_SEC + t.tv_nsec / NSEC_PER_USEC;
 
 
-	data->bucket = which_bucket(data->expected_us);
-
-	multiplier = performance_multiplier();
+	data->bucket = which_bucket(data->next_timer_us);
 
 	/*
-	 * if the correction factor is 0 (eg first time init or cpu hotplug
-	 * etc), we actually want to start out with a unity factor.
+	 * Force the result of multiplication to be 64 bits even if both
+	 * operands are 32 bits.
+	 * Make sure to round up for half microseconds.
 	 */
-	if (data->correction_factor[data->bucket] == 0)
-		data->correction_factor[data->bucket] = RESOLUTION * DECAY;
-
-	/* Make sure to round up for half microseconds */
-	data->predicted_us = div_round64(data->expected_us * data->correction_factor[data->bucket],
+	data->predicted_us = div_round64((uint64_t)data->next_timer_us *
+					 data->correction_factor[data->bucket],
 					 RESOLUTION * DECAY);
 
-	repeat = get_typical_interval(data);
+	get_typical_interval(data);
+
+	/*
+	 * Performance multiplier defines a minimum predicted idle
+	 * duration / latency ratio. Adjust the latency limit if
+	 * necessary.
+	 */
+	interactivity_req = data->predicted_us / performance_multiplier();
+	if (latency_req > interactivity_req)
+		latency_req = interactivity_req;
 
 	/*
 	 * We want to default to C1 (hlt), not to busy polling
 	 * unless the timer is happening really really soon.
 	 */
-	if (data->expected_us > 5 &&
+	if (data->next_timer_us > 5 &&
 	    !drv->states[CPUIDLE_DRIVER_STATE_START].disabled &&
 		dev->states_usage[CPUIDLE_DRIVER_STATE_START].disable == 0)
 		data->last_state_idx = CPUIDLE_DRIVER_STATE_START;
@@ -373,55 +349,12 @@ static int menu_select(struct cpuidle_driver *drv, struct cpuidle_device *dev)
 
 		if (s->disabled || su->disable)
 			continue;
-		if (s->target_residency > data->predicted_us) {
-			low_predicted = 1;
+		if (s->target_residency > data->predicted_us)
 			continue;
-		}
 		if (s->exit_latency > latency_req)
 			continue;
-		if (s->exit_latency * multiplier > data->predicted_us)
-			continue;
 
 		data->last_state_idx = i;
-		data->exit_us = s->exit_latency;
-	}
-
-	/* not deepest C-state chosen for low predicted residency */
-	if (low_predicted) {
-		unsigned int timer_us = 0;
-		unsigned int perfect_us = 0;
-
-		/*
-		 * Set a timer to detect whether this sleep is much
-		 * longer than repeat mode predicted.  If the timer
-		 * triggers, the code will evaluate whether to put
-		 * the CPU into a deeper C-state.
-		 * The timer is cancelled on CPU wakeup.
-		 */
-		timer_us = 2 * (data->predicted_us + MAX_DEVIATION);
-
-		perfect_us = perfect_cstate_ms * 1000;
-
-		if (repeat && (4 * timer_us < data->expected_us)) {
-			RCU_NONIDLE(hrtimer_start(hrtmr,
-				ns_to_ktime(1000 * timer_us),
-				HRTIMER_MODE_REL_PINNED));
-			/* In repeat case, menu hrtimer is started */
-			per_cpu(hrtimer_status, cpu) = MENU_HRTIMER_REPEAT;
-		} else if (perfect_us < data->expected_us) {
-			/*
-			 * The next timer is long. This could be because
-			 * we did not make a useful prediction.
-			 * In that case, it makes sense to re-enter
-			 * into a deeper C-state after some time.
-			 */
-			RCU_NONIDLE(hrtimer_start(hrtmr,
-				ns_to_ktime(1000 * timer_us),
-				HRTIMER_MODE_REL_PINNED));
-			/* In general case, menu hrtimer is started */
-			per_cpu(hrtimer_status, cpu) = MENU_HRTIMER_GENERAL;
-		}
-
 	}
 
 	return data->last_state_idx;
@@ -452,37 +385,47 @@ static void menu_update(struct cpuidle_driver *drv, struct cpuidle_device *dev)
 {
 	struct menu_device *data = &__get_cpu_var(menu_devices);
 	int last_idx = data->last_state_idx;
-	unsigned int last_idle_us = cpuidle_get_last_residency(dev);
 	struct cpuidle_state *target = &drv->states[last_idx];
 	unsigned int measured_us;
-	u64 new_factor;
+	unsigned int new_factor;
 
 	/*
-	 * Ugh, this idle state doesn't support residency measurements, so we
-	 * are basically lost in the dark.  As a compromise, assume we slept
-	 * for the whole expected time.
+	 * Try to figure out how much time passed between entry to low
+	 * power state and occurrence of the wakeup event.
+	 *
+	 * If the entered idle state didn't support residency measurements,
+	 * we are basically lost in the dark how much time passed.
+	 * As a compromise, assume we slept for the whole expected time.
+	 *
+	 * Any measured amount of time will include the exit latency.
+	 * Since we are interested in when the wakeup begun, not when it
+	 * was completed, we must substract the exit latency. However, if
+	 * the measured amount of time is less than the exit latency,
+	 * assume the state was never reached and the exit latency is 0.
 	 */
-	if (unlikely(!(target->flags & CPUIDLE_FLAG_TIME_VALID)))
-		last_idle_us = data->expected_us;
-
-
-	measured_us = last_idle_us;
+	if (unlikely(!(target->flags & CPUIDLE_FLAG_TIME_VALID))) {
+		/* Use timer value as is */
+		measured_us = data->next_timer_us;
 
-	/*
-	 * We correct for the exit latency; we are assuming here that the
-	 * exit latency happens after the event that we're interested in.
-	 */
-	if (measured_us > data->exit_us)
-		measured_us -= data->exit_us;
+	} else {
+		/* Use measured value */
+		measured_us = cpuidle_get_last_residency(dev);
 
+		/* Deduct exit latency */
+		if (measured_us > target->exit_latency)
+			measured_us -= target->exit_latency;
 
-	/* update our correction ratio */
+		/* Make sure our coefficients do not exceed unity */
+		if (measured_us > data->next_timer_us)
+			measured_us = data->next_timer_us;
+	}
 
-	new_factor = data->correction_factor[data->bucket]
-			* (DECAY - 1) / DECAY;
+	/* Update our correction ratio */
+	new_factor = data->correction_factor[data->bucket];
+	new_factor -= new_factor / DECAY;
 
-	if (data->expected_us > 0 && measured_us < MAX_INTERESTING)
-		new_factor += RESOLUTION * measured_us / data->expected_us;
+	if (data->next_timer_us > 0 && measured_us < MAX_INTERESTING)
+		new_factor += RESOLUTION * measured_us / data->next_timer_us;
 	else
 		/*
 		 * we were idle so long that we count it as a perfect
@@ -492,15 +435,17 @@ static void menu_update(struct cpuidle_driver *drv, struct cpuidle_device *dev)
 
 	/*
 	 * We don't want 0 as factor; we always want at least
-	 * a tiny bit of estimated time.
+	 * a tiny bit of estimated time. Fortunately, due to rounding,
+	 * new_factor will stay nonzero regardless of measured_us values
+	 * and the compiler can eliminate this test as long as DECAY > 1.
 	 */
-	if (new_factor == 0)
+	if (DECAY == 1 && unlikely(new_factor == 0))
 		new_factor = 1;
 
 	data->correction_factor[data->bucket] = new_factor;
 
 	/* update the repeating-pattern data */
-	data->intervals[data->interval_ptr++] = last_idle_us;
+	data->intervals[data->interval_ptr++] = measured_us;
 	if (data->interval_ptr >= INTERVALS)
 		data->interval_ptr = 0;
 }
@@ -514,12 +459,17 @@ static int menu_enable_device(struct cpuidle_driver *drv,
 				struct cpuidle_device *dev)
 {
 	struct menu_device *data = &per_cpu(menu_devices, dev->cpu);
-	struct hrtimer *t = &per_cpu(menu_hrtimer, dev->cpu);
-	hrtimer_init(t, CLOCK_MONOTONIC, HRTIMER_MODE_REL);
-	t->function = menu_hrtimer_notify;
+	int i;
 
 	memset(data, 0, sizeof(struct menu_device));
 
+	/*
+	 * if the correction factor is 0 (eg first time init or cpu hotplug
+	 * etc), we actually want to start out with a unity factor.
+	 */
+	for(i = 0; i < BUCKETS; i++)
+		data->correction_factor[i] = RESOLUTION * DECAY;
+
 	return 0;
 }
 
@@ -540,14 +490,4 @@ static int __init init_menu(void)
 	return cpuidle_register_governor(&menu_governor);
 }
 
-/**
- * exit_menu - exits the governor
- */
-static void __exit exit_menu(void)
-{
-	cpuidle_unregister_governor(&menu_governor);
-}
-
-MODULE_LICENSE("GPL");
-module_init(init_menu);
-module_exit(exit_menu);
+postcore_initcall(init_menu);