[mirror_ubuntu-bionic-kernel.git] / kernel / time / timekeeping.c

/*
 *  linux/kernel/time/timekeeping.c
 *
 *  Kernel timekeeping code and accessor functions
 *
 *  This code was moved from linux/kernel/timer.c.
 *  Please see that file for copyright and history logs.
 *
 */

#include <linux/module.h>
#include <linux/interrupt.h>
#include <linux/percpu.h>
#include <linux/init.h>
#include <linux/mm.h>
#include <linux/sysdev.h>
#include <linux/clocksource.h>
#include <linux/jiffies.h>
#include <linux/time.h>
#include <linux/tick.h>


/*
 * This read-write spinlock protects us from races in SMP while
 * playing with xtime and avenrun.
 */
__cacheline_aligned_in_smp DEFINE_SEQLOCK(xtime_lock);


/*
 * The current time
 * wall_to_monotonic is what we need to add to xtime (or xtime corrected
 * for sub jiffie times) to get to monotonic time.  Monotonic is pegged
 * at zero at system boot time, so wall_to_monotonic will be negative,
 * however, we will ALWAYS keep the tv_nsec part positive so we can use
 * the usual normalization.
 *
 * wall_to_monotonic is moved after resume from suspend for the monotonic
 * time not to jump. We need to add total_sleep_time to wall_to_monotonic
 * to get the real boot based time offset.
 *
 * - wall_to_monotonic is no longer the boot time, getboottime must be
 * used instead.
 */
struct timespec xtime __attribute__ ((aligned (16)));
struct timespec wall_to_monotonic __attribute__ ((aligned (16)));
static unsigned long total_sleep_time;		/* seconds */

static struct timespec xtime_cache __attribute__ ((aligned (16)));
void update_xtime_cache(u64 nsec)
{
	xtime_cache = xtime;
	timespec_add_ns(&xtime_cache, nsec);
}

static struct clocksource *clock; /* pointer to current clocksource */


#ifdef CONFIG_GENERIC_TIME
/**
 * __get_nsec_offset - Returns nanoseconds since last call to periodic_hook
 *
 * private function, must hold xtime_lock lock when being
 * called. Returns the number of nanoseconds since the
 * last call to update_wall_time() (adjusted by NTP scaling)
 */
static inline s64 __get_nsec_offset(void)
{
	cycle_t cycle_now, cycle_delta;
	s64 ns_offset;

	/* read clocksource: */
	cycle_now = clocksource_read(clock);

	/* calculate the delta since the last update_wall_time: */
	cycle_delta = (cycle_now - clock->cycle_last) & clock->mask;

	/* convert to nanoseconds: */
	ns_offset = cyc2ns(clock, cycle_delta);

	return ns_offset;
}

/**
 * getnstimeofday - Returns the time of day in a timespec
 * @ts:		pointer to the timespec to be set
 *
 * Returns the time of day in a timespec.
 */
void getnstimeofday(struct timespec *ts)
{
	unsigned long seq;
	s64 nsecs;

	do {
		seq = read_seqbegin(&xtime_lock);

		*ts = xtime;
		nsecs = __get_nsec_offset();

	} while (read_seqretry(&xtime_lock, seq));

	timespec_add_ns(ts, nsecs);
}

EXPORT_SYMBOL(getnstimeofday);

/**
 * do_gettimeofday - Returns the time of day in a timeval
 * @tv:		pointer to the timeval to be set
 *
 * NOTE: Users should be converted to using getnstimeofday()
 */
void do_gettimeofday(struct timeval *tv)
{
	struct timespec now;

	getnstimeofday(&now);
	tv->tv_sec = now.tv_sec;
	tv->tv_usec = now.tv_nsec/1000;
}

EXPORT_SYMBOL(do_gettimeofday);
/**
 * do_settimeofday - Sets the time of day
 * @tv:		pointer to the timespec variable containing the new time
 *
 * Sets the time of day to the new time and update NTP and notify hrtimers
 */
int do_settimeofday(struct timespec *tv)
{
	unsigned long flags;
	time_t wtm_sec, sec = tv->tv_sec;
	long wtm_nsec, nsec = tv->tv_nsec;

	if ((unsigned long)tv->tv_nsec >= NSEC_PER_SEC)
		return -EINVAL;

	write_seqlock_irqsave(&xtime_lock, flags);

	nsec -= __get_nsec_offset();

	wtm_sec  = wall_to_monotonic.tv_sec + (xtime.tv_sec - sec);
	wtm_nsec = wall_to_monotonic.tv_nsec + (xtime.tv_nsec - nsec);

	set_normalized_timespec(&xtime, sec, nsec);
	set_normalized_timespec(&wall_to_monotonic, wtm_sec, wtm_nsec);
	update_xtime_cache(0);

	clock->error = 0;
	ntp_clear();

	update_vsyscall(&xtime, clock);

	write_sequnlock_irqrestore(&xtime_lock, flags);

	/* signal hrtimers about time change */
	clock_was_set();

	return 0;
}

EXPORT_SYMBOL(do_settimeofday);

/**
 * change_clocksource - Swaps clocksources if a new one is available
 *
 * Accumulates current time interval and initializes new clocksource
 */
static void change_clocksource(void)
{
	struct clocksource *new;
	cycle_t now;
	u64 nsec;

	new = clocksource_get_next();

	if (clock == new)
		return;

	now = clocksource_read(new);
	nsec =  __get_nsec_offset();
	timespec_add_ns(&xtime, nsec);

	clock = new;
	clock->cycle_last = now;

	clock->error = 0;
	clock->xtime_nsec = 0;
	clocksource_calculate_interval(clock,
		(unsigned long)(current_tick_length()>>TICK_LENGTH_SHIFT));

	tick_clock_notify();

	printk(KERN_INFO "Time: %s clocksource has been installed.\n",
	       clock->name);
}
#else
static inline void change_clocksource(void) { }
static inline s64 __get_nsec_offset(void) { return 0; }
#endif

/**
 * timekeeping_valid_for_hres - Check if timekeeping is suitable for hres
 */
int timekeeping_valid_for_hres(void)
{
	unsigned long seq;
	int ret;

	do {
		seq = read_seqbegin(&xtime_lock);

		ret = clock->flags & CLOCK_SOURCE_VALID_FOR_HRES;

	} while (read_seqretry(&xtime_lock, seq));

	return ret;
}

/**
 * read_persistent_clock -  Return time in seconds from the persistent clock.
 *
 * Weak dummy function for arches that do not yet support it.
 * Returns seconds from epoch using the battery backed persistent clock.
 * Returns zero if unsupported.
 *
 *  XXX - Do be sure to remove it once all arches implement it.
 */
unsigned long __attribute__((weak)) read_persistent_clock(void)
{
	return 0;
}

/*
 * timekeeping_init - Initializes the clocksource and common timekeeping values
 */
void __init timekeeping_init(void)
{
	unsigned long flags;
	unsigned long sec = read_persistent_clock();

	write_seqlock_irqsave(&xtime_lock, flags);

	ntp_clear();

	clock = clocksource_get_next();
	clocksource_calculate_interval(clock,
		(unsigned long)(current_tick_length()>>TICK_LENGTH_SHIFT));
	clock->cycle_last = clocksource_read(clock);

	xtime.tv_sec = sec;
	xtime.tv_nsec = 0;
	set_normalized_timespec(&wall_to_monotonic,
		-xtime.tv_sec, -xtime.tv_nsec);
	update_xtime_cache(0);
	total_sleep_time = 0;
	write_sequnlock_irqrestore(&xtime_lock, flags);
}

/* flag for if timekeeping is suspended */
static int timekeeping_suspended;
/* time in seconds when suspend began */
static unsigned long timekeeping_suspend_time;
/* xtime offset when we went into suspend */
static s64 timekeeping_suspend_nsecs;

/**
 * timekeeping_resume - Resumes the generic timekeeping subsystem.
 * @dev:	unused
 *
 * This is for the generic clocksource timekeeping.
 * xtime/wall_to_monotonic/jiffies/etc are
 * still managed by arch specific suspend/resume code.
 */
static int timekeeping_resume(struct sys_device *dev)
{
	unsigned long flags;
	unsigned long now = read_persistent_clock();

	clocksource_resume();

	write_seqlock_irqsave(&xtime_lock, flags);

	if (now && (now > timekeeping_suspend_time)) {
		unsigned long sleep_length = now - timekeeping_suspend_time;

		xtime.tv_sec += sleep_length;
		wall_to_monotonic.tv_sec -= sleep_length;
		total_sleep_time += sleep_length;
	}
	/* Make sure that we have the correct xtime reference */
	timespec_add_ns(&xtime, timekeeping_suspend_nsecs);
	update_xtime_cache(0);
	/* re-base the last cycle value */
	clock->cycle_last = clocksource_read(clock);
	clock->error = 0;
	timekeeping_suspended = 0;
	write_sequnlock_irqrestore(&xtime_lock, flags);

	touch_softlockup_watchdog();

	clockevents_notify(CLOCK_EVT_NOTIFY_RESUME, NULL);

	/* Resume hrtimers */
	hres_timers_resume();

	return 0;
}

static int timekeeping_suspend(struct sys_device *dev, pm_message_t state)
{
	unsigned long flags;

	timekeeping_suspend_time = read_persistent_clock();

	write_seqlock_irqsave(&xtime_lock, flags);
	/* Get the current xtime offset */
	timekeeping_suspend_nsecs = __get_nsec_offset();
	timekeeping_suspended = 1;
	write_sequnlock_irqrestore(&xtime_lock, flags);

	clockevents_notify(CLOCK_EVT_NOTIFY_SUSPEND, NULL);

	return 0;
}

/* sysfs resume/suspend bits for timekeeping */
static struct sysdev_class timekeeping_sysclass = {
	.name		= "timekeeping",
	.resume		= timekeeping_resume,
	.suspend	= timekeeping_suspend,
};

static struct sys_device device_timer = {
	.id		= 0,
	.cls		= &timekeeping_sysclass,
};

static int __init timekeeping_init_device(void)
{
	int error = sysdev_class_register(&timekeeping_sysclass);
	if (!error)
		error = sysdev_register(&device_timer);
	return error;
}

device_initcall(timekeeping_init_device);

/*
 * If the error is already larger, we look ahead even further
 * to compensate for late or lost adjustments.
 */
static __always_inline int clocksource_bigadjust(s64 error, s64 *interval,
						 s64 *offset)
{
	s64 tick_error, i;
	u32 look_ahead, adj;
	s32 error2, mult;

	/*
	 * Use the current error value to determine how much to look ahead.
	 * The larger the error the slower we adjust for it to avoid problems
	 * with losing too many ticks, otherwise we would overadjust and
	 * produce an even larger error.  The smaller the adjustment the
	 * faster we try to adjust for it, as lost ticks can do less harm
	 * here.  This is tuned so that an error of about 1 msec is adusted
	 * within about 1 sec (or 2^20 nsec in 2^SHIFT_HZ ticks).
	 */
	error2 = clock->error >> (TICK_LENGTH_SHIFT + 22 - 2 * SHIFT_HZ);
	error2 = abs(error2);
	for (look_ahead = 0; error2 > 0; look_ahead++)
		error2 >>= 2;

	/*
	 * Now calculate the error in (1 << look_ahead) ticks, but first
	 * remove the single look ahead already included in the error.
	 */
	tick_error = current_tick_length() >>
		(TICK_LENGTH_SHIFT - clock->shift + 1);
	tick_error -= clock->xtime_interval >> 1;
	error = ((error - tick_error) >> look_ahead) + tick_error;

	/* Finally calculate the adjustment shift value.  */
	i = *interval;
	mult = 1;
	if (error < 0) {
		error = -error;
		*interval = -*interval;
		*offset = -*offset;
		mult = -1;
	}
	for (adj = 0; error > i; adj++)
		error >>= 1;

	*interval <<= adj;
	*offset <<= adj;
	return mult << adj;
}

/*
 * Adjust the multiplier to reduce the error value,
 * this is optimized for the most common adjustments of -1,0,1,
 * for other values we can do a bit more work.
 */
static void clocksource_adjust(s64 offset)
{
	s64 error, interval = clock->cycle_interval;
	int adj;

	error = clock->error >> (TICK_LENGTH_SHIFT - clock->shift - 1);
	if (error > interval) {
		error >>= 2;
		if (likely(error <= interval))
			adj = 1;
		else
			adj = clocksource_bigadjust(error, &interval, &offset);
	} else if (error < -interval) {
		error >>= 2;
		if (likely(error >= -interval)) {
			adj = -1;
			interval = -interval;
			offset = -offset;
		} else
			adj = clocksource_bigadjust(error, &interval, &offset);
	} else
		return;

	clock->mult += adj;
	clock->xtime_interval += interval;
	clock->xtime_nsec -= offset;
	clock->error -= (interval - offset) <<
			(TICK_LENGTH_SHIFT - clock->shift);
}

/**
 * update_wall_time - Uses the current clocksource to increment the wall time
 *
 * Called from the timer interrupt, must hold a write on xtime_lock.
 */
void update_wall_time(void)
{
	cycle_t offset;

	/* Make sure we're fully resumed: */
	if (unlikely(timekeeping_suspended))
		return;

#ifdef CONFIG_GENERIC_TIME
	offset = (clocksource_read(clock) - clock->cycle_last) & clock->mask;
#else
	offset = clock->cycle_interval;
#endif
	clock->xtime_nsec += (s64)xtime.tv_nsec << clock->shift;

	/* normally this loop will run just once, however in the
	 * case of lost or late ticks, it will accumulate correctly.
	 */
	while (offset >= clock->cycle_interval) {
		/* accumulate one interval */
		clock->xtime_nsec += clock->xtime_interval;
		clock->cycle_last += clock->cycle_interval;
		offset -= clock->cycle_interval;

		if (clock->xtime_nsec >= (u64)NSEC_PER_SEC << clock->shift) {
			clock->xtime_nsec -= (u64)NSEC_PER_SEC << clock->shift;
			xtime.tv_sec++;
			second_overflow();
		}

		/* accumulate error between NTP and clock interval */
		clock->error += current_tick_length();
		clock->error -= clock->xtime_interval << (TICK_LENGTH_SHIFT - clock->shift);
	}

	/* correct the clock when NTP error is too big */
	clocksource_adjust(offset);

	/* store full nanoseconds into xtime */
	xtime.tv_nsec = (s64)clock->xtime_nsec >> clock->shift;
	clock->xtime_nsec -= (s64)xtime.tv_nsec << clock->shift;

	update_xtime_cache(cyc2ns(clock, offset));

	/* check to see if there is a new clocksource to use */
	change_clocksource();
	update_vsyscall(&xtime, clock);
}

/**
 * getboottime - Return the real time of system boot.
 * @ts:		pointer to the timespec to be set
 *
 * Returns the time of day in a timespec.
 *
 * This is based on the wall_to_monotonic offset and the total suspend
 * time. Calls to settimeofday will affect the value returned (which
 * basically means that however wrong your real time clock is at boot time,
 * you get the right time here).
 */
void getboottime(struct timespec *ts)
{
	set_normalized_timespec(ts,
		- (wall_to_monotonic.tv_sec + total_sleep_time),
		- wall_to_monotonic.tv_nsec);
}

/**
 * monotonic_to_bootbased - Convert the monotonic time to boot based.
 * @ts:		pointer to the timespec to be converted
 */
void monotonic_to_bootbased(struct timespec *ts)
{
	ts->tv_sec += total_sleep_time;
}

unsigned long get_seconds(void)
{
	return xtime_cache.tv_sec;
}
EXPORT_SYMBOL(get_seconds);


struct timespec current_kernel_time(void)
{
	struct timespec now;
	unsigned long seq;

	do {
		seq = read_seqbegin(&xtime_lock);

		now = xtime_cache;
	} while (read_seqretry(&xtime_lock, seq));

	return now;
}
EXPORT_SYMBOL(current_kernel_time);
Commit	Line	Data
8524070b JS	1	/*
	2	* linux/kernel/time/timekeeping.c
	3	*
	4	* Kernel timekeeping code and accessor functions
	5	*
	6	* This code was moved from linux/kernel/timer.c.
	7	* Please see that file for copyright and history logs.
	8	*
	9	*/
	10
	11	#include <linux/module.h>
	12	#include <linux/interrupt.h>
	13	#include <linux/percpu.h>
	14	#include <linux/init.h>
	15	#include <linux/mm.h>
	16	#include <linux/sysdev.h>
	17	#include <linux/clocksource.h>
	18	#include <linux/jiffies.h>
	19	#include <linux/time.h>
	20	#include <linux/tick.h>
	21
	22
	23	/*
	24	* This read-write spinlock protects us from races in SMP while
	25	* playing with xtime and avenrun.
	26	*/
ba2a631b	27	__cacheline_aligned_in_smp DEFINE_SEQLOCK(xtime_lock);
8524070b JS	28
	29
	30	/*
	31	* The current time
	32	* wall_to_monotonic is what we need to add to xtime (or xtime corrected
	33	* for sub jiffie times) to get to monotonic time. Monotonic is pegged
	34	* at zero at system boot time, so wall_to_monotonic will be negative,
	35	* however, we will ALWAYS keep the tv_nsec part positive so we can use
	36	* the usual normalization.
7c3f1a57 TJ	37	*
	38	* wall_to_monotonic is moved after resume from suspend for the monotonic
	39	* time not to jump. We need to add total_sleep_time to wall_to_monotonic
	40	* to get the real boot based time offset.
	41	*
	42	* - wall_to_monotonic is no longer the boot time, getboottime must be
	43	* used instead.
8524070b JS	44	*/
	45	struct timespec xtime __attribute__ ((aligned (16)));
	46	struct timespec wall_to_monotonic __attribute__ ((aligned (16)));
7c3f1a57	47	static unsigned long total_sleep_time; /* seconds */
8524070b	48
17c38b74	49	static struct timespec xtime_cache __attribute__ ((aligned (16)));
1001d0a9	50	void update_xtime_cache(u64 nsec)
17c38b74 JS	51	{
	52	xtime_cache = xtime;
	53	timespec_add_ns(&xtime_cache, nsec);
	54	}
17c38b74	55
8524070b JS	56	static struct clocksource clock; / pointer to current clocksource */
	57
	58
	59	#ifdef CONFIG_GENERIC_TIME
	60	/**
	61	* __get_nsec_offset - Returns nanoseconds since last call to periodic_hook
	62	*
	63	* private function, must hold xtime_lock lock when being
	64	* called. Returns the number of nanoseconds since the
	65	* last call to update_wall_time() (adjusted by NTP scaling)
	66	*/
	67	static inline s64 __get_nsec_offset(void)
	68	{
	69	cycle_t cycle_now, cycle_delta;
	70	s64 ns_offset;
	71
	72	/* read clocksource: */
	73	cycle_now = clocksource_read(clock);
	74
	75	/* calculate the delta since the last update_wall_time: */
	76	cycle_delta = (cycle_now - clock->cycle_last) & clock->mask;
	77
	78	/* convert to nanoseconds: */
	79	ns_offset = cyc2ns(clock, cycle_delta);
	80
	81	return ns_offset;
	82	}
	83
	84	/**
efd9ac86	85	* getnstimeofday - Returns the time of day in a timespec
8524070b JS	86	* @ts: pointer to the timespec to be set
8524070b JS	87	*
efd9ac86	88	* Returns the time of day in a timespec.
8524070b	89	*/
efd9ac86	90	void getnstimeofday(struct timespec *ts)
8524070b JS	91	{
	92	unsigned long seq;
	93	s64 nsecs;
	94
	95	do {
	96	seq = read_seqbegin(&xtime_lock);
	97
	98	*ts = xtime;
	99	nsecs = __get_nsec_offset();
	100
	101	} while (read_seqretry(&xtime_lock, seq));
	102
	103	timespec_add_ns(ts, nsecs);
	104	}
	105
8524070b JS	106	EXPORT_SYMBOL(getnstimeofday);
	107
	108	/**
	109	* do_gettimeofday - Returns the time of day in a timeval
	110	* @tv: pointer to the timeval to be set
	111	*
efd9ac86	112	* NOTE: Users should be converted to using getnstimeofday()
8524070b JS	113	*/
	114	void do_gettimeofday(struct timeval *tv)
	115	{
	116	struct timespec now;
	117
efd9ac86	118	getnstimeofday(&now);
8524070b JS	119	tv->tv_sec = now.tv_sec;
	120	tv->tv_usec = now.tv_nsec/1000;
	121	}
	122
	123	EXPORT_SYMBOL(do_gettimeofday);
	124	/**
	125	* do_settimeofday - Sets the time of day
	126	* @tv: pointer to the timespec variable containing the new time
	127	*
	128	* Sets the time of day to the new time and update NTP and notify hrtimers
	129	*/
	130	int do_settimeofday(struct timespec *tv)
	131	{
	132	unsigned long flags;
	133	time_t wtm_sec, sec = tv->tv_sec;
	134	long wtm_nsec, nsec = tv->tv_nsec;
	135
	136	if ((unsigned long)tv->tv_nsec >= NSEC_PER_SEC)
	137	return -EINVAL;
	138
	139	write_seqlock_irqsave(&xtime_lock, flags);
	140
	141	nsec -= __get_nsec_offset();
	142
	143	wtm_sec = wall_to_monotonic.tv_sec + (xtime.tv_sec - sec);
	144	wtm_nsec = wall_to_monotonic.tv_nsec + (xtime.tv_nsec - nsec);
	145
	146	set_normalized_timespec(&xtime, sec, nsec);
	147	set_normalized_timespec(&wall_to_monotonic, wtm_sec, wtm_nsec);
1001d0a9	148	update_xtime_cache(0);
8524070b JS	149
	150	clock->error = 0;
	151	ntp_clear();
	152
	153	update_vsyscall(&xtime, clock);
	154
	155	write_sequnlock_irqrestore(&xtime_lock, flags);
	156
	157	/* signal hrtimers about time change */
	158	clock_was_set();
	159
	160	return 0;
	161	}
	162
	163	EXPORT_SYMBOL(do_settimeofday);
	164
	165	/**
	166	* change_clocksource - Swaps clocksources if a new one is available
	167	*
	168	* Accumulates current time interval and initializes new clocksource
	169	*/
	170	static void change_clocksource(void)
	171	{
	172	struct clocksource *new;
	173	cycle_t now;
	174	u64 nsec;
	175
	176	new = clocksource_get_next();
	177
	178	if (clock == new)
	179	return;
	180
	181	now = clocksource_read(new);
	182	nsec = __get_nsec_offset();
	183	timespec_add_ns(&xtime, nsec);
	184
	185	clock = new;
	186	clock->cycle_last = now;
	187
	188	clock->error = 0;
	189	clock->xtime_nsec = 0;
bbe4d18a JS	190	clocksource_calculate_interval(clock,
bbe4d18a JS	191	(unsigned long)(current_tick_length()>>TICK_LENGTH_SHIFT));
8524070b JS	192
	193	tick_clock_notify();
	194
	195	printk(KERN_INFO "Time: %s clocksource has been installed.\n",
	196	clock->name);
	197	}
	198	#else
	199	static inline void change_clocksource(void) { }
6a669ee8	200	static inline s64 __get_nsec_offset(void) { return 0; }
8524070b JS	201	#endif
	202
	203	/**
cf4fc6cb	204	* timekeeping_valid_for_hres - Check if timekeeping is suitable for hres
8524070b	205	*/
cf4fc6cb	206	int timekeeping_valid_for_hres(void)
8524070b JS	207	{
	208	unsigned long seq;
	209	int ret;
	210
	211	do {
	212	seq = read_seqbegin(&xtime_lock);
	213
	214	ret = clock->flags & CLOCK_SOURCE_VALID_FOR_HRES;
	215
	216	} while (read_seqretry(&xtime_lock, seq));
	217
	218	return ret;
	219	}
	220
	221	/**
	222	* read_persistent_clock - Return time in seconds from the persistent clock.
	223	*
	224	* Weak dummy function for arches that do not yet support it.
	225	* Returns seconds from epoch using the battery backed persistent clock.
	226	* Returns zero if unsupported.
	227	*
	228	* XXX - Do be sure to remove it once all arches implement it.
	229	*/
	230	unsigned long __attribute__((weak)) read_persistent_clock(void)
	231	{
	232	return 0;
	233	}
	234
	235	/*
	236	* timekeeping_init - Initializes the clocksource and common timekeeping values
	237	*/
	238	void __init timekeeping_init(void)
	239	{
	240	unsigned long flags;
	241	unsigned long sec = read_persistent_clock();
	242
	243	write_seqlock_irqsave(&xtime_lock, flags);
	244
	245	ntp_clear();
	246
	247	clock = clocksource_get_next();
bbe4d18a JS	248	clocksource_calculate_interval(clock,
bbe4d18a JS	249	(unsigned long)(current_tick_length()>>TICK_LENGTH_SHIFT));
8524070b JS	250	clock->cycle_last = clocksource_read(clock);
	251
	252	xtime.tv_sec = sec;
	253	xtime.tv_nsec = 0;
	254	set_normalized_timespec(&wall_to_monotonic,
	255	-xtime.tv_sec, -xtime.tv_nsec);
1001d0a9	256	update_xtime_cache(0);
7c3f1a57	257	total_sleep_time = 0;
8524070b JS	258	write_sequnlock_irqrestore(&xtime_lock, flags);
	259	}
	260
	261	/* flag for if timekeeping is suspended */
	262	static int timekeeping_suspended;
	263	/* time in seconds when suspend began */
	264	static unsigned long timekeeping_suspend_time;
6a669ee8 TG	265	/* xtime offset when we went into suspend */
6a669ee8 TG	266	static s64 timekeeping_suspend_nsecs;
8524070b JS	267
	268	/**
	269	* timekeeping_resume - Resumes the generic timekeeping subsystem.
	270	* @dev: unused
	271	*
	272	* This is for the generic clocksource timekeeping.
	273	* xtime/wall_to_monotonic/jiffies/etc are
	274	* still managed by arch specific suspend/resume code.
	275	*/
	276	static int timekeeping_resume(struct sys_device *dev)
	277	{
	278	unsigned long flags;
	279	unsigned long now = read_persistent_clock();
	280
d10ff3fb TG	281	clocksource_resume();
d10ff3fb TG	282
8524070b JS	283	write_seqlock_irqsave(&xtime_lock, flags);
	284
	285	if (now && (now > timekeeping_suspend_time)) {
	286	unsigned long sleep_length = now - timekeeping_suspend_time;
	287
	288	xtime.tv_sec += sleep_length;
	289	wall_to_monotonic.tv_sec -= sleep_length;
7c3f1a57	290	total_sleep_time += sleep_length;
8524070b	291	}
6a669ee8 TG	292	/* Make sure that we have the correct xtime reference */
6a669ee8 TG	293	timespec_add_ns(&xtime, timekeeping_suspend_nsecs);
1001d0a9	294	update_xtime_cache(0);
8524070b JS	295	/* re-base the last cycle value */
	296	clock->cycle_last = clocksource_read(clock);
	297	clock->error = 0;
	298	timekeeping_suspended = 0;
	299	write_sequnlock_irqrestore(&xtime_lock, flags);
	300
	301	touch_softlockup_watchdog();
	302
	303	clockevents_notify(CLOCK_EVT_NOTIFY_RESUME, NULL);
	304
	305	/* Resume hrtimers */
	306	hres_timers_resume();
	307
	308	return 0;
	309	}
	310
	311	static int timekeeping_suspend(struct sys_device *dev, pm_message_t state)
	312	{
	313	unsigned long flags;
	314
3be90950 TG	315	timekeeping_suspend_time = read_persistent_clock();
3be90950 TG	316
8524070b	317	write_seqlock_irqsave(&xtime_lock, flags);
6a669ee8 TG	318	/* Get the current xtime offset */
6a669ee8 TG	319	timekeeping_suspend_nsecs = __get_nsec_offset();
8524070b	320	timekeeping_suspended = 1;
8524070b JS	321	write_sequnlock_irqrestore(&xtime_lock, flags);
	322
	323	clockevents_notify(CLOCK_EVT_NOTIFY_SUSPEND, NULL);
	324
	325	return 0;
	326	}
	327
	328	/* sysfs resume/suspend bits for timekeeping */
	329	static struct sysdev_class timekeeping_sysclass = {
af5ca3f4	330	.name = "timekeeping",
8524070b JS	331	.resume = timekeeping_resume,
8524070b JS	332	.suspend = timekeeping_suspend,
8524070b JS	333	};
	334
	335	static struct sys_device device_timer = {
	336	.id = 0,
	337	.cls = &timekeeping_sysclass,
	338	};
	339
	340	static int __init timekeeping_init_device(void)
	341	{
	342	int error = sysdev_class_register(&timekeeping_sysclass);
	343	if (!error)
	344	error = sysdev_register(&device_timer);
	345	return error;
	346	}
	347
	348	device_initcall(timekeeping_init_device);
	349
	350	/*
	351	* If the error is already larger, we look ahead even further
	352	* to compensate for late or lost adjustments.
	353	*/
	354	static __always_inline int clocksource_bigadjust(s64 error, s64 *interval,
	355	s64 *offset)
	356	{
	357	s64 tick_error, i;
	358	u32 look_ahead, adj;
	359	s32 error2, mult;
	360
	361	/*
	362	* Use the current error value to determine how much to look ahead.
	363	* The larger the error the slower we adjust for it to avoid problems
	364	* with losing too many ticks, otherwise we would overadjust and
	365	* produce an even larger error. The smaller the adjustment the
	366	* faster we try to adjust for it, as lost ticks can do less harm
	367	* here. This is tuned so that an error of about 1 msec is adusted
	368	* within about 1 sec (or 2^20 nsec in 2^SHIFT_HZ ticks).
	369	*/
	370	error2 = clock->error >> (TICK_LENGTH_SHIFT + 22 - 2 * SHIFT_HZ);
	371	error2 = abs(error2);
	372	for (look_ahead = 0; error2 > 0; look_ahead++)
	373	error2 >>= 2;
	374
	375	/*
	376	* Now calculate the error in (1 << look_ahead) ticks, but first
	377	* remove the single look ahead already included in the error.
	378	*/
	379	tick_error = current_tick_length() >>
	380	(TICK_LENGTH_SHIFT - clock->shift + 1);
	381	tick_error -= clock->xtime_interval >> 1;
	382	error = ((error - tick_error) >> look_ahead) + tick_error;
	383
	384	/* Finally calculate the adjustment shift value. */
	385	i = *interval;
	386	mult = 1;
	387	if (error < 0) {
	388	error = -error;
	389	interval = -interval;
	390	offset = -offset;
	391	mult = -1;
	392	}
	393	for (adj = 0; error > i; adj++)
	394	error >>= 1;
	395
	396	*interval <<= adj;
397	*offset <<= adj;
398	return mult << adj;
399	}
400
401	/*
402	* Adjust the multiplier to reduce the error value,
403	* this is optimized for the most common adjustments of -1,0,1,
404	* for other values we can do a bit more work.
405	*/
71120f18	406	static void clocksource_adjust(s64 offset)
8524070b JS	407	{
	408	s64 error, interval = clock->cycle_interval;
	409	int adj;
	410
	411	error = clock->error >> (TICK_LENGTH_SHIFT - clock->shift - 1);
	412	if (error > interval) {
	413	error >>= 2;
	414	if (likely(error <= interval))
	415	adj = 1;
	416	else
	417	adj = clocksource_bigadjust(error, &interval, &offset);
	418	} else if (error < -interval) {
	419	error >>= 2;
	420	if (likely(error >= -interval)) {
	421	adj = -1;
	422	interval = -interval;
	423	offset = -offset;
	424	} else
	425	adj = clocksource_bigadjust(error, &interval, &offset);
	426	} else
	427	return;
	428
	429	clock->mult += adj;
	430	clock->xtime_interval += interval;
	431	clock->xtime_nsec -= offset;
	432	clock->error -= (interval - offset) <<
	433	(TICK_LENGTH_SHIFT - clock->shift);
	434	}
	435
	436	/**
	437	* update_wall_time - Uses the current clocksource to increment the wall time
	438	*
	439	* Called from the timer interrupt, must hold a write on xtime_lock.
	440	*/
	441	void update_wall_time(void)
	442	{
	443	cycle_t offset;
	444
	445	/* Make sure we're fully resumed: */
	446	if (unlikely(timekeeping_suspended))
	447	return;
	448
	449	#ifdef CONFIG_GENERIC_TIME
	450	offset = (clocksource_read(clock) - clock->cycle_last) & clock->mask;
	451	#else
	452	offset = clock->cycle_interval;
	453	#endif
	454	clock->xtime_nsec += (s64)xtime.tv_nsec << clock->shift;
	455
	456	/* normally this loop will run just once, however in the
	457	* case of lost or late ticks, it will accumulate correctly.
	458	*/
	459	while (offset >= clock->cycle_interval) {
	460	/* accumulate one interval */
	461	clock->xtime_nsec += clock->xtime_interval;
	462	clock->cycle_last += clock->cycle_interval;
	463	offset -= clock->cycle_interval;
	464
	465	if (clock->xtime_nsec >= (u64)NSEC_PER_SEC << clock->shift) {
	466	clock->xtime_nsec -= (u64)NSEC_PER_SEC << clock->shift;
	467	xtime.tv_sec++;
	468	second_overflow();
	469	}
	470
8524070b JS	471	/* accumulate error between NTP and clock interval */
	472	clock->error += current_tick_length();
	473	clock->error -= clock->xtime_interval << (TICK_LENGTH_SHIFT - clock->shift);
	474	}
	475
	476	/* correct the clock when NTP error is too big */
71120f18	477	clocksource_adjust(offset);
8524070b JS	478
	479	/* store full nanoseconds into xtime */
	480	xtime.tv_nsec = (s64)clock->xtime_nsec >> clock->shift;
	481	clock->xtime_nsec -= (s64)xtime.tv_nsec << clock->shift;
	482
17c38b74 JS	483	update_xtime_cache(cyc2ns(clock, offset));
17c38b74 JS	484
8524070b JS	485	/* check to see if there is a new clocksource to use */
	486	change_clocksource();
	487	update_vsyscall(&xtime, clock);
	488	}
7c3f1a57 TJ	489
	490	/**
	491	* getboottime - Return the real time of system boot.
	492	* @ts: pointer to the timespec to be set
	493	*
	494	* Returns the time of day in a timespec.
	495	*
	496	* This is based on the wall_to_monotonic offset and the total suspend
	497	* time. Calls to settimeofday will affect the value returned (which
	498	* basically means that however wrong your real time clock is at boot time,
	499	* you get the right time here).
	500	*/
	501	void getboottime(struct timespec *ts)
	502	{
	503	set_normalized_timespec(ts,
	504	- (wall_to_monotonic.tv_sec + total_sleep_time),
	505	- wall_to_monotonic.tv_nsec);
	506	}
	507
	508	/**
	509	* monotonic_to_bootbased - Convert the monotonic time to boot based.
	510	* @ts: pointer to the timespec to be converted
	511	*/
	512	void monotonic_to_bootbased(struct timespec *ts)
	513	{
	514	ts->tv_sec += total_sleep_time;
	515	}
2c6b47de	516
17c38b74 JS	517	unsigned long get_seconds(void)
	518	{
	519	return xtime_cache.tv_sec;
	520	}
	521	EXPORT_SYMBOL(get_seconds);
	522
	523
2c6b47de JS	524	struct timespec current_kernel_time(void)
	525	{
	526	struct timespec now;
	527	unsigned long seq;
	528
	529	do {
	530	seq = read_seqbegin(&xtime_lock);
	531
17c38b74	532	now = xtime_cache;
2c6b47de JS	533	} while (read_seqretry(&xtime_lock, seq));
	534
	535	return now;
	536	}
2c6b47de	537	EXPORT_SYMBOL(current_kernel_time);