[PATCH 0/4] Posix Alarm Timers

[PATCH 0/4] Posix Alarm Timers

[John Stultz]

Just wanted to send out my current Posix Alarm Timers queue for 
comment and review in the hopes of getting it into shape for
2.6.40.

Some background on the patches can be found here:
	https://lwn.net/Articles/429925/

New in this series:
* Added timekeeping_inject_sleeptime, which corrects CLOCK_BOOTTIME
  for arches that don't support read_persistent_clock().
* I've added a CAP_WAKE_ALARM capability to limit random applications
  from setting alarms in the last patch. This probably needs some
  careful review as it allows clock_nanosleep and timer_create to
  return -EPERM, which is new behavior.

thanks
-john

CC: Arve Hjønnevåg <arve@android.com>
CC: Thomas Gleixner <tglx@linutronix.de>
CC: Alessandro Zummo <a.zummo@towertech.it>

John Stultz (4):
  time: Add timekeeping_inject_sleeptime
  timers: Add rb_init_node() to allow for stack allocated rb nodes
  timers: Introduce in-kernel alarm-timer interface
  timers: posix interface for alarm-timers

 drivers/rtc/class.c          |   23 +-
 include/linux/alarmtimer.h   |   30 ++
 include/linux/capability.h   |    7 +-
 include/linux/posix-timers.h |    2 +
 include/linux/rbtree.h       |    8 +
 include/linux/time.h         |    3 +
 include/linux/timerqueue.h   |    2 +-
 kernel/time/Makefile         |    2 +-
 kernel/time/alarmtimer.c     |  705 ++++++++++++++++++++++++++++++++++++++++++
 kernel/time/timekeeping.c    |   56 ++++-
 10 files changed, 818 insertions(+), 20 deletions(-)
 create mode 100644 include/linux/alarmtimer.h
 create mode 100644 kernel/time/alarmtimer.c

-- 
1.7.3.2.146.gca209

[PATCH 1/4] time: Add timekeeping_inject_sleeptime

[John Stultz]

Some platforms cannot implement read_persistent_clock, as
their RTC devices are only accessible when interrupts are enabled.
This keeps them from being used by the timekeeping code on resume
to measure the time in suspend.

The RTC layer tries to work around this, by calling do_settimeofday
on resume after irqs are reenabled to set the time properly. However,
this only corrects CLOCK_REALTIME, and does not properly adjust
the sleep time value. This causes btime in /proc/stat to be incorrect
as well as making the new CLOCK_BOTTTIME inaccurate.

This patch resolves the issue by introducing a new timekeeping hook
to allow the RTC layer to inject the sleep time on resume.

The code also checks to make sure that read_persistent_clock is
nonfunctional before setting the sleep time, so that should the RTC's
HCTOSYS option be configured in on a system that does support
read_persistent_clock we will not increase the total_sleep_time twice.

CC: Arve Hjønnevåg <arve@android.com>
CC: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: John Stultz <john.stultz@linaro.org>
---
 drivers/rtc/class.c       |   23 +++++++-----------
 include/linux/time.h      |    1 +
 kernel/time/timekeeping.c |   56 ++++++++++++++++++++++++++++++++++++++++++--
 3 files changed, 63 insertions(+), 17 deletions(-)

diff --git a/drivers/rtc/class.c b/drivers/rtc/class.c
index 3901386..4194e59 100644
--- a/drivers/rtc/class.c
+++ b/drivers/rtc/class.c
@@ -41,26 +41,21 @@ static void rtc_device_release(struct device *dev)
  * system's wall clock; restore it on resume().
  */
 
-static struct timespec	delta;
 static time_t		oldtime;
+static struct timespec	oldts;
 
 static int rtc_suspend(struct device *dev, pm_message_t mesg)
 {
 	struct rtc_device	*rtc = to_rtc_device(dev);
 	struct rtc_time		tm;
-	struct timespec		ts = current_kernel_time();
 
 	if (strcmp(dev_name(&rtc->dev), CONFIG_RTC_HCTOSYS_DEVICE) != 0)
 		return 0;
 
 	rtc_read_time(rtc, &tm);
+	ktime_get_ts(&oldts);
 	rtc_tm_to_time(&tm, &oldtime);
 
-	/* RTC precision is 1 second; adjust delta for avg 1/2 sec err */
-	set_normalized_timespec(&delta,
-				ts.tv_sec - oldtime,
-				ts.tv_nsec - (NSEC_PER_SEC >> 1));
-
 	return 0;
 }
 
@@ -70,10 +65,12 @@ static int rtc_resume(struct device *dev)
 	struct rtc_time		tm;
 	time_t			newtime;
 	struct timespec		time;
+	struct timespec		newts;
 
 	if (strcmp(dev_name(&rtc->dev), CONFIG_RTC_HCTOSYS_DEVICE) != 0)
 		return 0;
 
+	ktime_get_ts(&newts);
 	rtc_read_time(rtc, &tm);
 	if (rtc_valid_tm(&tm) != 0) {
 		pr_debug("%s:  bogus resume time\n", dev_name(&rtc->dev));
@@ -85,15 +82,13 @@ static int rtc_resume(struct device *dev)
 			pr_debug("%s:  time travel!\n", dev_name(&rtc->dev));
 		return 0;
 	}
+	/* calculate the RTC time delta */
+	set_normalized_timespec(&time, newtime - oldtime, 0);
 
-	/* restore wall clock using delta against this RTC;
-	 * adjust again for avg 1/2 second RTC sampling error
-	 */
-	set_normalized_timespec(&time,
-				newtime + delta.tv_sec,
-				(NSEC_PER_SEC >> 1) + delta.tv_nsec);
-	do_settimeofday(&time);
+	/* subtract kernel time between rtc_suspend to rtc_resume */
+	time = timespec_sub(time, timespec_sub(newts, oldts));
 
+	timekeeping_inject_sleeptime(&time);
 	return 0;
 }
 
diff --git a/include/linux/time.h b/include/linux/time.h
index 454a262..4ea5a75 100644
--- a/include/linux/time.h
+++ b/include/linux/time.h
@@ -126,6 +126,7 @@ struct timespec __current_kernel_time(void); /* does not take xtime_lock */
 struct timespec get_monotonic_coarse(void);
 void get_xtime_and_monotonic_and_sleep_offset(struct timespec *xtim,
 				struct timespec *wtom, struct timespec *sleep);
+void timekeeping_inject_sleeptime(struct timespec *delta);
 
 #define CURRENT_TIME		(current_kernel_time())
 #define CURRENT_TIME_SEC	((struct timespec) { get_seconds(), 0 })
diff --git a/kernel/time/timekeeping.c b/kernel/time/timekeeping.c
index 8ad5d57..8e6a05a 100644
--- a/kernel/time/timekeeping.c
+++ b/kernel/time/timekeeping.c
@@ -596,6 +596,58 @@ void __init timekeeping_init(void)
 static struct timespec timekeeping_suspend_time;
 
 /**
+ * __timekeeping_inject_sleeptime - Internal function to add sleep interval
+ * @delta: pointer to a timespec delta value
+ *
+ * Takes a timespec offset measuring a suspend interval and properly
+ * adds the sleep offset to the timekeeping variables.
+ */
+static void __timekeeping_inject_sleeptime(struct timespec *delta)
+{
+	xtime = timespec_add(xtime, *delta);
+	wall_to_monotonic = timespec_sub(wall_to_monotonic, *delta);
+	total_sleep_time = timespec_add(total_sleep_time, *delta);
+}
+
+
+/**
+ * timekeeping_inject_sleeptime - Adds suspend interval to timeekeeping values
+ * @delta: pointer to a timespec delta value
+ *
+ * This hook is for architectures that cannot support read_persistent_clock
+ * because their RTC/persistent clock is only accessible when irqs are enabled.
+ *
+ * This function should only be called by rtc_resume(), and allows
+ * a suspend offset to be injected into the timekeeping values.
+ */
+void timekeeping_inject_sleeptime(struct timespec *delta)
+{
+	unsigned long flags;
+	struct timespec ts;
+
+	/* Make sure we don't set the clock twice */
+	read_persistent_clock(&ts);
+	if (!(ts.tv_sec == 0 && ts.tv_nsec == 0))
+		return;
+
+	write_seqlock_irqsave(&xtime_lock, flags);
+	timekeeping_forward_now();
+
+	__timekeeping_inject_sleeptime(delta);
+
+	timekeeper.ntp_error = 0;
+	ntp_clear();
+	update_vsyscall(&xtime, &wall_to_monotonic, timekeeper.clock,
+				timekeeper.mult);
+
+	write_sequnlock_irqrestore(&xtime_lock, flags);
+
+	/* signal hrtimers about time change */
+	clock_was_set();
+}
+
+
+/**
  * timekeeping_resume - Resumes the generic timekeeping subsystem.
  *
  * This is for the generic clocksource timekeeping.
@@ -615,9 +667,7 @@ static void timekeeping_resume(void)
 
 	if (timespec_compare(&ts, &timekeeping_suspend_time) > 0) {
 		ts = timespec_sub(ts, timekeeping_suspend_time);
-		xtime = timespec_add(xtime, ts);
-		wall_to_monotonic = timespec_sub(wall_to_monotonic, ts);
-		total_sleep_time = timespec_add(total_sleep_time, ts);
+		__timekeeping_inject_sleeptime(&ts);
 	}
 	/* re-base the last cycle value */
 	timekeeper.clock->cycle_last = timekeeper.clock->read(timekeeper.clock);
-- 
1.7.3.2.146.gca209

[PATCH 2/4] timers: Add rb_init_node() to allow for stack allocated rb nodes

[John Stultz]

In cases where a timerqueue_node or some structure that utilizes
a timerqueue_node is allocated on the stack, gcc would give warnings
caused by the timerqueue_init()'s calling RB_CLEAR_NODE, which
self-references the nodes uninitialized data.

The solution is to create an rb_init_node() function that zeros
the rb_node structure out and then calls RB_CLEAR_NODE(), and
then call the new init function from timerqueue_init().

CC: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: John Stultz <john.stultz@linaro.org>
---
 include/linux/rbtree.h     |    8 ++++++++
 include/linux/timerqueue.h |    2 +-
 2 files changed, 9 insertions(+), 1 deletions(-)

diff --git a/include/linux/rbtree.h b/include/linux/rbtree.h
index 7066acb..033b507 100644
--- a/include/linux/rbtree.h
+++ b/include/linux/rbtree.h
@@ -136,6 +136,14 @@ static inline void rb_set_color(struct rb_node *rb, int color)
 #define RB_EMPTY_NODE(node)	(rb_parent(node) == node)
 #define RB_CLEAR_NODE(node)	(rb_set_parent(node, node))
 
+static inline void rb_init_node(struct rb_node *rb)
+{
+	rb->rb_parent_color = 0;
+	rb->rb_right = NULL;
+	rb->rb_left = NULL;
+	RB_CLEAR_NODE(rb);
+}
+
 extern void rb_insert_color(struct rb_node *, struct rb_root *);
 extern void rb_erase(struct rb_node *, struct rb_root *);
 
diff --git a/include/linux/timerqueue.h b/include/linux/timerqueue.h
index a520fd7..5088727 100644
--- a/include/linux/timerqueue.h
+++ b/include/linux/timerqueue.h
@@ -39,7 +39,7 @@ struct timerqueue_node *timerqueue_getnext(struct timerqueue_head *head)
 
 static inline void timerqueue_init(struct timerqueue_node *node)
 {
-	RB_CLEAR_NODE(&node->node);
+	rb_init_node(&node->node);
 }
 
 static inline void timerqueue_init_head(struct timerqueue_head *head)
-- 
1.7.3.2.146.gca209

[PATCH 3/4] timers: Introduce in-kernel alarm-timer interface

[John Stultz]

This provides the in kernel interface and infrastructure for
alarm-timers.

Alarm-timers are a hybrid style timer, similar to hrtimers,
but when the system is suspended, the RTC device is set to
fire and wake the system for when the soonest alarm-timer
expires.

The concept for Alarm-timers was inspired by the Android Alarm
driver (by Arve Hjønnevåg) found in the Android kernel tree.

See: http://android.git.kernel.org/?p=kernel/common.git;a=blob;f=drivers/rtc/alarm.c;h=1250edfbdf3302f5e4ea6194847c6ef4bb7beb1c;hb=android-2.6.36

This in-kernel interface should be fairly compatible with the
Android alarm driver in-kernel interface, but has the advantage
of utilizing the new RTC timerqueue code instead of doing direct
RTC manipulation.

CC: Arve Hjønnevåg <arve@android.com>
CC: Thomas Gleixner <tglx@linutronix.de>
CC: Alessandro Zummo <a.zummo@towertech.it>
Signed-off-by: John Stultz <john.stultz@linaro.org>
---
 include/linux/alarmtimer.h |   30 ++++
 kernel/time/Makefile       |    2 +-
 kernel/time/alarmtimer.c   |  375 ++++++++++++++++++++++++++++++++++++++++++++
 3 files changed, 406 insertions(+), 1 deletions(-)
 create mode 100644 include/linux/alarmtimer.h
 create mode 100644 kernel/time/alarmtimer.c

diff --git a/include/linux/alarmtimer.h b/include/linux/alarmtimer.h
new file mode 100644
index 0000000..6861f28
--- /dev/null
+++ b/include/linux/alarmtimer.h
@@ -0,0 +1,30 @@
+#ifndef _LINUX_ALARMTIMER_H
+#define _LINUX_ALARMTIMER_H
+
+#include <linux/time.h>
+#include <linux/hrtimer.h>
+#include <linux/timerqueue.h>
+#include <linux/rtc.h>
+
+enum alarmtimer_type {
+	ALARM_REALTIME,
+	ALARM_BOOTTIME,
+
+	ALARM_NUMTYPE,
+};
+
+struct alarm {
+	struct timerqueue_node node;
+	ktime_t period;
+	void (*function)(struct alarm *);
+	enum alarmtimer_type type;
+	char enabled;
+	void *data;
+};
+
+void alarm_init(struct alarm *alarm, enum alarmtimer_type type,
+		void (*function)(struct alarm *));
+void alarm_start(struct alarm *alarm, ktime_t start, ktime_t period);
+void alarm_cancel(struct alarm *alarm);
+
+#endif
diff --git a/kernel/time/Makefile b/kernel/time/Makefile
index b042599..e2fd74b 100644
--- a/kernel/time/Makefile
+++ b/kernel/time/Makefile
@@ -1,5 +1,5 @@
 obj-y += timekeeping.o ntp.o clocksource.o jiffies.o timer_list.o timecompare.o
-obj-y += timeconv.o posix-clock.o
+obj-y += timeconv.o posix-clock.o alarmtimer.o
 
 obj-$(CONFIG_GENERIC_CLOCKEVENTS_BUILD)		+= clockevents.o
 obj-$(CONFIG_GENERIC_CLOCKEVENTS)		+= tick-common.o
diff --git a/kernel/time/alarmtimer.c b/kernel/time/alarmtimer.c
new file mode 100644
index 0000000..4d91f0e
--- /dev/null
+++ b/kernel/time/alarmtimer.c
@@ -0,0 +1,375 @@
+/*
+ * Alarmtimer interface
+ *
+ * This interface provides a timer which is similarto hrtimers,
+ * but triggers a RTC alarm if the box is suspend.
+ *
+ * This interface is influenced by the Android RTC Alarm timer
+ * interface.
+ *
+ * Copyright (C) 2010 IBM Corperation
+ *
+ * Author: John Stultz <john.stultz@linaro.org>
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+#include <linux/time.h>
+#include <linux/hrtimer.h>
+#include <linux/timerqueue.h>
+#include <linux/rtc.h>
+#include <linux/alarmtimer.h>
+#include <linux/mutex.h>
+#include <linux/platform_device.h>
+#include <linux/posix-timers.h>
+#include <linux/workqueue.h>
+#include <linux/freezer.h>
+
+
+static struct alarm_base {
+	spinlock_t lock;
+	struct timerqueue_head timerqueue;
+	struct hrtimer timer;
+	ktime_t (*gettime)(void);
+	clockid_t base_clockid;
+	struct work_struct irqwork;
+} alarm_bases[ALARM_NUMTYPE];
+
+static struct rtc_timer rtctimer;
+static struct rtc_device *rtcdev;
+
+static ktime_t freezer_delta;
+static DEFINE_SPINLOCK(freezer_delta_lock);
+
+
+/**************************************************************************
+ * alarmtimer management code
+ */
+
+/*
+ * alarmtimer_enqueue - Adds an alarm timer to an alarm_base timerqueue
+ * @base: pointer to the base where the timer is being run
+ * @alarm: pointer to alarm being enqueued.
+ *
+ * Adds alarm to a alarm_base timerqueue and if necessary sets
+ * an hrtimer to run.
+ *
+ * Must hold base->lock when calling.
+ */
+static void alarmtimer_enqueue(struct alarm_base *base, struct alarm *alarm)
+{
+	timerqueue_add(&base->timerqueue, &alarm->node);
+	if (&alarm->node == timerqueue_getnext(&base->timerqueue)) {
+		hrtimer_try_to_cancel(&base->timer);
+		hrtimer_start(&base->timer, alarm->node.expires,
+				HRTIMER_MODE_ABS);
+	}
+}
+
+/*
+ * alarmtimer_remove - Removes an alarm timer from an alarm_base timerqueue
+ * @base: pointer to the base where the timer is running
+ * @alarm: pointer to alarm being removed
+ *
+ * Removes alarm to a alarm_base timerqueue and if necessary sets
+ * a new timer to run.
+ *
+ * Must hold base->lock when calling.
+ */
+static void alarmtimer_remove(struct alarm_base *base, struct alarm *alarm)
+{
+	struct timerqueue_node *next = timerqueue_getnext(&base->timerqueue);
+
+	timerqueue_del(&base->timerqueue, &alarm->node);
+	if (next == &alarm->node) {
+		hrtimer_try_to_cancel(&base->timer);
+		next = timerqueue_getnext(&base->timerqueue);
+		if (!next)
+			return;
+		hrtimer_start(&base->timer, next->expires, HRTIMER_MODE_ABS);
+	}
+}
+
+/*
+ * alarmtimer_do_work - Handles alarm being fired.
+ * @work: pointer to workqueue being run
+ *
+ * When a timer fires, this runs through the timerqueue to see
+ * which alarm timers, and run those that expired. If there are
+ * more alarm timers queued, we set the hrtimer to fire in the
+ * future.
+ */
+void alarmtimer_do_work(struct work_struct *work)
+{
+	struct alarm_base *base = container_of(work, struct alarm_base,
+						irqwork);
+	struct timerqueue_node *next;
+	unsigned long flags;
+	ktime_t now;
+
+	spin_lock_irqsave(&base->lock, flags);
+	now = base->gettime();
+	while ((next = timerqueue_getnext(&base->timerqueue))) {
+		struct alarm *alarm;
+		ktime_t expired = next->expires;
+
+		if (expired.tv64 >= now.tv64)
+			break;
+
+		alarm = container_of(next, struct alarm, node);
+
+		timerqueue_del(&base->timerqueue, &alarm->node);
+		alarm->enabled = 0;
+		/* Re-add periodic timers */
+		if (alarm->period.tv64) {
+			alarm->node.expires = ktime_add(expired, alarm->period);
+			timerqueue_add(&base->timerqueue, &alarm->node);
+			alarm->enabled = 1;
+		}
+		spin_unlock_irqrestore(&base->lock, flags);
+		if (alarm->function)
+			alarm->function(alarm);
+		spin_lock_irqsave(&base->lock, flags);
+	}
+
+	if (next) {
+		hrtimer_start(&base->timer, next->expires,
+				HRTIMER_MODE_ABS);
+	}
+	spin_unlock_irqrestore(&base->lock, flags);
+}
+
+
+/*
+ * alarmtimer_fired - Handles alarm hrtimer being fired.
+ * @timer: pointer to hrtimer being run
+ *
+ * When a timer fires, this schedules the do_work function to
+ * be run.
+ */
+static enum hrtimer_restart alarmtimer_fired(struct hrtimer *timer)
+{
+	struct alarm_base *base = container_of(timer, struct alarm_base, timer);
+	schedule_work(&base->irqwork);
+	return HRTIMER_NORESTART;
+}
+
+
+/*
+ * alarmtimer_suspend - Suspend time callback
+ * @dev: unused
+ * @state: unused
+ *
+ * When we are going into suspend, we look through the bases
+ * to see which is the soonest timer to expire. We then
+ * set an rtc timer to fire that far into the future, which
+ * will wake us from suspend.
+ */
+static int alarmtimer_suspend(struct device *dev)
+{
+	struct rtc_time tm;
+	ktime_t min, now;
+	unsigned long flags;
+	int i;
+
+	spin_lock_irqsave(&freezer_delta_lock, flags);
+	min = freezer_delta;
+	freezer_delta = ktime_set(0, 0);
+	spin_unlock_irqrestore(&freezer_delta_lock, flags);
+
+	/* If we have no rtcdev, just return */
+	if (!rtcdev)
+		return 0;
+
+	/* Find the soonest timer to expire*/
+	for (i = 0; i < ALARM_NUMTYPE; i++) {
+		struct alarm_base *base = &alarm_bases[i];
+		struct timerqueue_node *next;
+		ktime_t delta;
+
+		spin_lock_irqsave(&base->lock, flags);
+		next = timerqueue_getnext(&base->timerqueue);
+		spin_unlock_irqrestore(&base->lock, flags);
+		if (!next)
+			continue;
+		delta = ktime_sub(next->expires, base->gettime());
+		if (!min.tv64 || (delta.tv64 < min.tv64))
+			min = delta;
+	}
+	if (min.tv64 == 0)
+		return 0;
+
+	/* XXX - Should we enforce a minimum sleep time? */
+	WARN_ON(min.tv64 < NSEC_PER_SEC);
+
+	/* Setup an rtc timer to fire that far in the future */
+	rtc_timer_cancel(rtcdev, &rtctimer);
+	rtc_read_time(rtcdev, &tm);
+	now = rtc_tm_to_ktime(tm);
+	now = ktime_add(now, min);
+
+	rtc_timer_start(rtcdev, &rtctimer, now, ktime_set(0, 0));
+
+	return 0;
+}
+
+
+/**************************************************************************
+ * alarm kernel interface code
+ */
+
+/*
+ * alarm_init - Initialize an alarm structure
+ * @alarm: ptr to alarm to be initialized
+ * @type: the type of the alarm
+ * @function: callback that is run when the alarm fires
+ *
+ * In-kernel interface to initializes the alarm structure.
+ */
+void alarm_init(struct alarm *alarm, enum alarmtimer_type type,
+		void (*function)(struct alarm *))
+{
+	timerqueue_init(&alarm->node);
+	alarm->period = ktime_set(0, 0);
+	alarm->function = function;
+	alarm->type = type;
+	alarm->enabled = 0;
+}
+
+/*
+ * alarm_start - Sets an alarm to fire
+ * @alarm: ptr to alarm to set
+ * @start: time to run the alarm
+ * @period: period at which the alarm will recur
+ *
+ * In-kernel interface set an alarm timer.
+ */
+void alarm_start(struct alarm *alarm, ktime_t start, ktime_t period)
+{
+	struct alarm_base *base = &alarm_bases[alarm->type];
+	unsigned long flags;
+
+	spin_lock_irqsave(&base->lock, flags);
+	if (alarm->enabled)
+		alarmtimer_remove(base, alarm);
+	alarm->node.expires = start;
+	alarm->period = period;
+	alarmtimer_enqueue(base, alarm);
+	alarm->enabled = 1;
+	spin_unlock_irqrestore(&base->lock, flags);
+}
+
+/*
+ * alarm_cancel - Tries to cancel an alarm timer
+ * @alarm: ptr to alarm to be canceled
+ *
+ * In-kernel interface to cancel an alarm timer.
+ */
+void alarm_cancel(struct alarm *alarm)
+{
+	struct alarm_base *base = &alarm_bases[alarm->type];
+	unsigned long flags;
+
+	spin_lock_irqsave(&base->lock, flags);
+	if (alarm->enabled)
+		alarmtimer_remove(base, alarm);
+	alarm->enabled = 0;
+	spin_unlock_irqrestore(&base->lock, flags);
+}
+
+
+
+/**************************************************************************
+ * alarmtimer initialization code
+ */
+
+/* Suspend hook structures */
+static const struct dev_pm_ops alarmtimer_pm_ops = {
+	.suspend = alarmtimer_suspend,
+};
+
+static struct platform_driver alarmtimer_driver = {
+	.driver = {
+		.name = "alarmtimer",
+		.pm = &alarmtimer_pm_ops,
+	}
+};
+
+/**
+ * alarmtimer_init - Initialize alarm timer code
+ *
+ * This function initializes the alarm bases and registers
+ * the posix clock ids.
+ */
+static int __init alarmtimer_init(void)
+{
+	int error = 0;
+	int i;
+
+	/* Initialize alarm bases */
+	alarm_bases[ALARM_REALTIME].base_clockid = CLOCK_REALTIME;
+	alarm_bases[ALARM_REALTIME].gettime = &ktime_get_real;
+	alarm_bases[ALARM_BOOTTIME].base_clockid = CLOCK_BOOTTIME;
+	alarm_bases[ALARM_BOOTTIME].gettime = &ktime_get_boottime;
+	for (i = 0; i < ALARM_NUMTYPE; i++) {
+		timerqueue_init_head(&alarm_bases[i].timerqueue);
+		spin_lock_init(&alarm_bases[i].lock);
+		hrtimer_init(&alarm_bases[i].timer,
+				alarm_bases[i].base_clockid,
+				HRTIMER_MODE_ABS);
+		alarm_bases[i].timer.function = alarmtimer_fired;
+		INIT_WORK(&alarm_bases[i].irqwork, alarmtimer_do_work);
+	}
+	error = platform_driver_register(&alarmtimer_driver);
+	platform_device_register_simple("alarmtimer", -1, NULL, 0);
+
+	return error;
+}
+device_initcall(alarmtimer_init);
+
+/**
+ * has_wakealarm - check rtc device has wakealarm ability
+ * @dev: current device
+ * @name_ptr: name to be returned
+ *
+ * This helper function checks to see if the rtc device can wake
+ * from suspend.
+ */
+static int __init has_wakealarm(struct device *dev, void *name_ptr)
+{
+	struct rtc_device *candidate = to_rtc_device(dev);
+
+	if (!candidate->ops->set_alarm)
+		return 0;
+	if (!device_may_wakeup(candidate->dev.parent))
+		return 0;
+
+	*(const char **)name_ptr = dev_name(dev);
+	return 1;
+}
+
+/**
+ * alarmtimer_init_late - Late initializing of alarmtimer code
+ *
+ * This function locates a rtc device to use for wakealarms.
+ * Run as late_initcall to make sure rtc devices have been
+ * registered.
+ */
+static int __init alarmtimer_init_late(void)
+{
+	char *str;
+
+	/* Find an rtc device and init the rtc_timer */
+	class_find_device(rtc_class, NULL, &str, has_wakealarm);
+	if (str)
+		rtcdev = rtc_class_open(str);
+	if (!rtcdev) {
+		printk(KERN_WARNING "No RTC device found, ALARM timers will"
+			" not wake from suspend");
+	}
+	rtc_timer_init(&rtctimer, NULL, NULL);
+
+	return 0;
+}
+late_initcall(alarmtimer_init_late);
-- 
1.7.3.2.146.gca209

[PATCH 4/4] timers: posix interface for alarm-timers

[John Stultz]

This patch exposes alarm-timers to userland via the posix clock
and timers interface, using two new clockids: CLOCK_REALTIME_ALARM
and CLOCK_BOOTTIME_ALARM. Both clockids behave identically to
CLOCK_REALTIME and CLOCK_BOOTTIME, respectively, but timers
set against the _ALARM suffixed clockids will wake the system if
it is suspended.

Some background can be found here:
	https://lwn.net/Articles/429925/

The concept for Alarm-timers was inspired by the Android Alarm
driver (by Arve Hjønnevåg) found in the Android kernel tree.

See: http://android.git.kernel.org/?p=kernel/common.git;a=blob;f=drivers/rtc/alarm.c;h=1250edfbdf3302f5e4ea6194847c6ef4bb7beb1c;hb=android-2.6.36

While the in-kernel interface is pretty similar between
alarm-timers and Android alarm driver, the user-space interface
for the Android alarm driver is via ioctls to a new char device.
As mentioned above, I've instead chosen to export this functionality
via the posix interface, as it seemed a little simpler and avoids
creating duplicate interfaces to things like CLOCK_REALTIME and
CLOCK_MONOTONIC under alternate names (ie:ANDROID_ALARM_RTC and
ANDROID_ALARM_SYSTEMTIME).

The semantics of the Android alarm driver are different from what
this posix interface provides. For instance, threads other then
the thread waiting on the Android alarm driver are able to modify
the alarm being waited on. Also this interface does not allow
the same wakelock semantics that the Android driver provides
(ie: kernel takes a wakelock on RTC alarm-interupt, and holds it
through process wakeup, and while the process runs, until the
process either closes the char device or calls back in to wait
on a new alarm).

One potential way to implement similar semantics may be via
the timerfd infrastructure, but this needs more research.

There may also need to be some sort of sysfs system level policy
hooks that allow alarm timers to be disabled to keep them
from firing at inappropriate times (ie: laptop in a well insulated
bag, mid-flight).

CC: Arve Hjønnevåg <arve@android.com>
CC: Thomas Gleixner <tglx@linutronix.de>
CC: Alessandro Zummo <a.zummo@towertech.it>
Signed-off-by: John Stultz <john.stultz@linaro.org>
---
 include/linux/capability.h   |    7 +-
 include/linux/posix-timers.h |    2 +
 include/linux/time.h         |    2 +
 kernel/time/alarmtimer.c     |  330 ++++++++++++++++++++++++++++++++++++++++++
 4 files changed, 340 insertions(+), 1 deletions(-)

diff --git a/include/linux/capability.h b/include/linux/capability.h
index 16ee8b4..7cb23ea 100644
--- a/include/linux/capability.h
+++ b/include/linux/capability.h
@@ -355,7 +355,12 @@ struct cpu_vfs_cap_data {
 
 #define CAP_SYSLOG           34
 
-#define CAP_LAST_CAP         CAP_SYSLOG
+/* Allow triggering something that will wake the system */
+
+#define CAP_WAKE_ALARM            35
+
+
+#define CAP_LAST_CAP         CAP_WAKE_ALARM
 
 #define cap_valid(x) ((x) >= 0 && (x) <= CAP_LAST_CAP)
 
diff --git a/include/linux/posix-timers.h b/include/linux/posix-timers.h
index d51243a..808227d 100644
--- a/include/linux/posix-timers.h
+++ b/include/linux/posix-timers.h
@@ -5,6 +5,7 @@
 #include <linux/list.h>
 #include <linux/sched.h>
 #include <linux/timex.h>
+#include <linux/alarmtimer.h>
 
 union cpu_time_count {
 	cputime_t cpu;
@@ -80,6 +81,7 @@ struct k_itimer {
 			unsigned long incr;
 			unsigned long expires;
 		} mmtimer;
+		struct alarm alarmtimer;
 	} it;
 };
 
diff --git a/include/linux/time.h b/include/linux/time.h
index 4ea5a75..b306178 100644
--- a/include/linux/time.h
+++ b/include/linux/time.h
@@ -295,6 +295,8 @@ struct itimerval {
 #define CLOCK_REALTIME_COARSE		5
 #define CLOCK_MONOTONIC_COARSE		6
 #define CLOCK_BOOTTIME			7
+#define CLOCK_REALTIME_ALARM		8
+#define CLOCK_BOOTTIME_ALARM		9
 
 /*
  * The IDs of various hardware clocks:
diff --git a/kernel/time/alarmtimer.c b/kernel/time/alarmtimer.c
index 4d91f0e..5fec899 100644
--- a/kernel/time/alarmtimer.c
+++ b/kernel/time/alarmtimer.c
@@ -215,6 +215,21 @@ static int alarmtimer_suspend(struct device *dev)
 }
 
 
+static void alarmtimer_freezerset(ktime_t absexp, enum alarmtimer_type type)
+{
+	ktime_t delta;
+	unsigned long flags;
+	struct alarm_base *base = &alarm_bases[type];
+
+	delta = ktime_sub(absexp, base->gettime());
+
+	spin_lock_irqsave(&freezer_delta_lock, flags);
+	if (!freezer_delta.tv64 || (delta.tv64 < freezer_delta.tv64))
+		freezer_delta = delta;
+	spin_unlock_irqrestore(&freezer_delta_lock, flags);
+}
+
+
 /**************************************************************************
  * alarm kernel interface code
  */
@@ -279,6 +294,309 @@ void alarm_cancel(struct alarm *alarm)
 }
 
 
+/**************************************************************************
+ * alarm posix interface code
+ */
+
+/*
+ * clock2alarm - helper that converts from clockid to alarmtypes
+ * @clockid: clockid.
+ *
+ * Helper function that converts from clockids to alarmtypes
+ */
+static enum alarmtimer_type clock2alarm(clockid_t clockid)
+{
+	if (clockid == CLOCK_REALTIME_ALARM)
+		return ALARM_REALTIME;
+	if (clockid == CLOCK_BOOTTIME_ALARM)
+		return ALARM_BOOTTIME;
+	return -1;
+}
+
+/*
+ * alarm_handle_timer - Callback for posix timers
+ * @alarm: alarm that fired
+ *
+ * Posix timer callback for expired alarm timers.
+ */
+static void alarm_handle_timer(struct alarm *alarm)
+{
+	struct k_itimer *ptr = container_of(alarm, struct k_itimer,
+						it.alarmtimer);
+	if (posix_timer_event(ptr, 0) != 0)
+		ptr->it_overrun++;
+}
+
+/*
+ * alarm_clock_getres - posix getres interface
+ * @which_clock: clockid
+ * @tp: timespec to fill
+ *
+ * Returns the granularity of underlying alarm base clock
+ */
+static int alarm_clock_getres(const clockid_t which_clock, struct timespec *tp)
+{
+	clockid_t baseid = alarm_bases[clock2alarm(which_clock)].base_clockid;
+
+	return hrtimer_get_res(baseid, tp);
+}
+
+/**
+ * alarm_clock_get - posix clock_get interface
+ * @which_clock: clockid
+ * @tp: timespec to fill.
+ *
+ * Provides the underlying alarm base time.
+ */
+static int alarm_clock_get(clockid_t which_clock, struct timespec *tp)
+{
+	struct alarm_base *base = &alarm_bases[clock2alarm(which_clock)];
+
+	*tp = ktime_to_timespec(base->gettime());
+	return 0;
+}
+
+/**
+ * alarm_timer_create - posix timer_create interface
+ * @new_timer: k_itimer pointer to manage
+ *
+ * Initializes the k_itimer structure.
+ */
+static int alarm_timer_create(struct k_itimer *new_timer)
+{
+	enum  alarmtimer_type type;
+	struct alarm_base *base;
+
+	if (!capable(CAP_WAKE_ALARM))
+		return -EPERM;
+
+	type = clock2alarm(new_timer->it_clock);
+	base = &alarm_bases[type];
+	alarm_init(&new_timer->it.alarmtimer, type, alarm_handle_timer);
+	return 0;
+}
+
+/**
+ * alarm_timer_get - posix timer_get interface
+ * @new_timer: k_itimer pointer
+ * @cur_setting: itimerspec data to fill
+ *
+ * Copies the itimerspec data out from the k_itimer
+ */
+static void alarm_timer_get(struct k_itimer *timr,
+				struct itimerspec *cur_setting)
+{
+	cur_setting->it_interval =
+			ktime_to_timespec(timr->it.alarmtimer.period);
+	cur_setting->it_value =
+			ktime_to_timespec(timr->it.alarmtimer.node.expires);
+	return;
+}
+
+/**
+ * alarm_timer_del - posix timer_del interface
+ * @timr: k_itimer pointer to be deleted
+ *
+ * Cancels any programmed alarms for the given timer.
+ */
+static int alarm_timer_del(struct k_itimer *timr)
+{
+	alarm_cancel(&timr->it.alarmtimer);
+	return 0;
+}
+
+/**
+ * alarm_timer_set - posix timer_set interface
+ * @timr: k_itimer pointer to be deleted
+ * @flags: timer flags
+ * @new_setting: itimerspec to be used
+ * @old_setting: itimerspec being replaced
+ *
+ * Sets the timer to new_setting, and starts the timer.
+ */
+static int alarm_timer_set(struct k_itimer *timr, int flags,
+				struct itimerspec *new_setting,
+				struct itimerspec *old_setting)
+{
+	/* Save old values */
+	old_setting->it_interval =
+			ktime_to_timespec(timr->it.alarmtimer.period);
+	old_setting->it_value =
+			ktime_to_timespec(timr->it.alarmtimer.node.expires);
+
+	/* If the timer was already set, cancel it */
+	alarm_cancel(&timr->it.alarmtimer);
+
+	/* start the timer */
+	alarm_start(&timr->it.alarmtimer,
+			timespec_to_ktime(new_setting->it_value),
+			timespec_to_ktime(new_setting->it_interval));
+	return 0;
+}
+
+/**
+ * alarmtimer_nsleep_wakeup - Wakeup function for alarm_timer_nsleep
+ * @alarm: ptr to alarm that fired
+ *
+ * Wakes up the task that set the alarmtimer
+ */
+static void alarmtimer_nsleep_wakeup(struct alarm *alarm)
+{
+	struct task_struct *task = (struct task_struct *)alarm->data;
+
+	alarm->data = NULL;
+	if (task)
+		wake_up_process(task);
+}
+
+/**
+ * alarmtimer_do_nsleep - Internal alarmtimer nsleep implementation
+ * @alarm: ptr to alarmtimer
+ * @absexp: absolute expiration time
+ *
+ * Sets the alarm timer and sleeps until it is fired or interrupted.
+ */
+static int alarmtimer_do_nsleep(struct alarm *alarm, ktime_t absexp)
+{
+	alarm->data = (void *)current;
+	do {
+		set_current_state(TASK_INTERRUPTIBLE);
+		alarm_start(alarm, absexp, ktime_set(0, 0));
+		if (likely(alarm->data))
+			schedule();
+
+		alarm_cancel(alarm);
+	} while (alarm->data && !signal_pending(current));
+
+	__set_current_state(TASK_RUNNING);
+
+	return (alarm->data == NULL);
+}
+
+
+/**
+ * update_rmtp - Update remaining timespec value
+ * @exp: expiration time
+ * @type: timer type
+ * @rmtp: user pointer to remaining timepsec value
+ *
+ * Helper function that fills in rmtp value with time between
+ * now and the exp value
+ */
+static int update_rmtp(ktime_t exp, enum  alarmtimer_type type,
+			struct timespec __user *rmtp)
+{
+	struct timespec rmt;
+	ktime_t rem;
+
+	rem = ktime_sub(exp, alarm_bases[type].gettime());
+
+	if (rem.tv64 <= 0)
+		return 0;
+	rmt = ktime_to_timespec(rem);
+
+	if (copy_to_user(rmtp, &rmt, sizeof(*rmtp)))
+		return -EFAULT;
+
+	return 1;
+
+}
+
+/**
+ * alarm_timer_nsleep_restart - restartblock alarmtimer nsleep
+ * @restart: ptr to restart block
+ *
+ * Handles restarted clock_nanosleep calls
+ */
+static long __sched alarm_timer_nsleep_restart(struct restart_block *restart)
+{
+	enum  alarmtimer_type type = restart->nanosleep.index;
+	ktime_t exp;
+	struct timespec __user  *rmtp;
+	struct alarm alarm;
+	int ret = 0;
+
+	exp.tv64 = restart->nanosleep.expires;
+	alarm_init(&alarm, type, alarmtimer_nsleep_wakeup);
+
+	if (alarmtimer_do_nsleep(&alarm, exp))
+		goto out;
+
+	if (freezing(current))
+		alarmtimer_freezerset(exp, type);
+
+	rmtp = restart->nanosleep.rmtp;
+	if (rmtp) {
+		ret = update_rmtp(exp, type, rmtp);
+		if (ret <= 0)
+			goto out;
+	}
+
+
+	/* The other values in restart are already filled in */
+	ret = -ERESTART_RESTARTBLOCK;
+out:
+	return ret;
+}
+
+/**
+ * alarm_timer_nsleep - alarmtimer nanosleep
+ * @which_clock: clockid
+ * @flags: determins abstime or relative
+ * @tsreq: requested sleep time (abs or rel)
+ * @rmtp: remaining sleep time saved
+ *
+ * Handles clock_nanosleep calls against _ALARM clockids
+ */
+static int alarm_timer_nsleep(const clockid_t which_clock, int flags,
+		     struct timespec *tsreq, struct timespec __user *rmtp)
+{
+	enum  alarmtimer_type type = clock2alarm(which_clock);
+	struct alarm alarm;
+	ktime_t exp;
+	int ret = 0;
+	struct restart_block *restart;
+
+	if (!capable(CAP_WAKE_ALARM))
+		return -EPERM;
+
+	alarm_init(&alarm, type, alarmtimer_nsleep_wakeup);
+
+	exp = timespec_to_ktime(*tsreq);
+	/* Convert (if necessary) to absolute time */
+	if (flags != TIMER_ABSTIME) {
+		ktime_t now = alarm_bases[type].gettime();
+		exp = ktime_add(now, exp);
+	}
+
+	if (alarmtimer_do_nsleep(&alarm, exp))
+		goto out;
+
+	if (freezing(current))
+		alarmtimer_freezerset(exp, type);
+
+	/* abs timers don't set remaining time or restart */
+	if (flags == TIMER_ABSTIME) {
+		ret = -ERESTARTNOHAND;
+		goto out;
+	}
+
+	if (rmtp) {
+		ret = update_rmtp(exp, type, rmtp);
+		if (ret <= 0)
+			goto out;
+	}
+
+	restart = &current_thread_info()->restart_block;
+	restart->fn = alarm_timer_nsleep_restart;
+	restart->nanosleep.index = type;
+	restart->nanosleep.expires = exp.tv64;
+	restart->nanosleep.rmtp = rmtp;
+	ret = -ERESTART_RESTARTBLOCK;
+
+out:
+	return ret;
+}
 
 /**************************************************************************
  * alarmtimer initialization code
@@ -306,6 +624,18 @@ static int __init alarmtimer_init(void)
 {
 	int error = 0;
 	int i;
+	struct k_clock alarm_clock = {
+		.clock_getres = alarm_clock_getres,
+		.clock_get = alarm_clock_get,
+		.timer_create = alarm_timer_create,
+		.timer_set = alarm_timer_set,
+		.timer_del = alarm_timer_del,
+		.timer_get = alarm_timer_get,
+		.nsleep = alarm_timer_nsleep,
+	};
+
+	posix_timers_register_clock(CLOCK_REALTIME_ALARM, &alarm_clock);
+	posix_timers_register_clock(CLOCK_BOOTTIME_ALARM, &alarm_clock);
 
 	/* Initialize alarm bases */
 	alarm_bases[ALARM_REALTIME].base_clockid = CLOCK_REALTIME;
-- 
1.7.3.2.146.gca209

Re: [PATCH 4/4] timers: posix interface for alarm-timers

[Richard Cochran]

On Tue, Apr 19, 2011 at 04:58:30PM -0700, John Stultz wrote:
> This patch exposes alarm-timers to userland via the posix clock
> and timers interface, using two new clockids: CLOCK_REALTIME_ALARM
> and CLOCK_BOOTTIME_ALARM. Both clockids behave identically to
> CLOCK_REALTIME and CLOCK_BOOTTIME, respectively, but timers
> set against the _ALARM suffixed clockids will wake the system if
> it is suspended.

So, is there reason these cannot or should not be using the dynamic
posix clock interface?

> The semantics of the Android alarm driver are different from what
> this posix interface provides.
...
> One potential way to implement similar semantics may be via
> the timerfd infrastructure, but this needs more research.
> 
> There may also need to be some sort of sysfs system level policy
> hooks that allow alarm timers to be disabled to keep them
> from firing at inappropriate times (ie: laptop in a well insulated
> bag, mid-flight).

I have not put hardly any thought into these issues, but the dynamic
interface gives you a character device at no extra charge.

Richard

Re: [PATCH 4/4] timers: posix interface for alarm-timers

[John Stultz]

On Wed, 2011-04-20 at 08:34 +0200, Richard Cochran wrote:
> On Tue, Apr 19, 2011 at 04:58:30PM -0700, John Stultz wrote:
> > This patch exposes alarm-timers to userland via the posix clock
> > and timers interface, using two new clockids: CLOCK_REALTIME_ALARM
> > and CLOCK_BOOTTIME_ALARM. Both clockids behave identically to
> > CLOCK_REALTIME and CLOCK_BOOTTIME, respectively, but timers
> > set against the _ALARM suffixed clockids will wake the system if
> > it is suspended.
> 
> So, is there reason these cannot or should not be using the dynamic
> posix clock interface?

So in this case, since the clockids are not directly linked to specific
dynamic hardware, I'm not sure if that's the best route.

> > The semantics of the Android alarm driver are different from what
> > this posix interface provides.
> ...
> > One potential way to implement similar semantics may be via
> > the timerfd infrastructure, but this needs more research.
> > 
> > There may also need to be some sort of sysfs system level policy
> > hooks that allow alarm timers to be disabled to keep them
> > from firing at inappropriate times (ie: laptop in a well insulated
> > bag, mid-flight).
> 
> I have not put hardly any thought into these issues, but the dynamic
> interface gives you a character device at no extra charge.

True, but I'd like to first try to extend the existing interface
(timerfd) before we look at adding new ones.

thanks
-john

Re: [PATCH 0/4] Posix Alarm Timers

[Arnd Bergmann]

On Wednesday 20 April 2011 01:58:26 John Stultz wrote:
> 
> Just wanted to send out my current Posix Alarm Timers queue for 
> comment and review in the hopes of getting it into shape for
> 2.6.40.
> 
> Some background on the patches can be found here:
>         https://lwn.net/Articles/429925/
> 
> New in this series:
> * Added timekeeping_inject_sleeptime, which corrects CLOCK_BOOTTIME
>   for arches that don't support read_persistent_clock().
> * I've added a CAP_WAKE_ALARM capability to limit random applications
>   from setting alarms in the last patch. This probably needs some
>   careful review as it allows clock_nanosleep and timer_create to
>   return -EPERM, which is new behavior.
> 
I haven't looked at these patches much before, so I've done a brief
review now. They look very nice, and I can't find anything to complain
about.

Acked-by: Arnd Bergmann <arnd@arndb.de>