timekeeping.c 46.6 KB
Newer Older
1 2 3 4 5 6 7 8 9 10
/*
 *  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.
 *
 */

11
#include <linux/timekeeper_internal.h>
12 13 14 15 16
#include <linux/module.h>
#include <linux/interrupt.h>
#include <linux/percpu.h>
#include <linux/init.h>
#include <linux/mm.h>
17
#include <linux/sched.h>
18
#include <linux/syscore_ops.h>
19 20 21 22
#include <linux/clocksource.h>
#include <linux/jiffies.h>
#include <linux/time.h>
#include <linux/tick.h>
23
#include <linux/stop_machine.h>
24
#include <linux/pvclock_gtod.h>
25
#include <linux/compiler.h>
26

27
#include "tick-internal.h"
28
#include "ntp_internal.h"
29
#include "timekeeping_internal.h"
30

31 32
#define TK_CLEAR_NTP		(1 << 0)
#define TK_MIRROR		(1 << 1)
33
#define TK_CLOCK_WAS_SET	(1 << 2)
34

35
static struct timekeeper timekeeper;
36 37
static DEFINE_RAW_SPINLOCK(timekeeper_lock);
static seqcount_t timekeeper_seq;
38
static struct timekeeper shadow_timekeeper;
39

40 41 42
/* flag for if timekeeping is suspended */
int __read_mostly timekeeping_suspended;

43 44 45
/* Flag for if there is a persistent clock on this platform */
bool __read_mostly persistent_clock_exist = false;

46 47 48 49 50 51 52 53
static inline void tk_normalize_xtime(struct timekeeper *tk)
{
	while (tk->xtime_nsec >= ((u64)NSEC_PER_SEC << tk->shift)) {
		tk->xtime_nsec -= (u64)NSEC_PER_SEC << tk->shift;
		tk->xtime_sec++;
	}
}

54 55 56 57 58 59 60 61 62
static inline struct timespec64 tk_xtime(struct timekeeper *tk)
{
	struct timespec64 ts;

	ts.tv_sec = tk->xtime_sec;
	ts.tv_nsec = (long)(tk->xtime_nsec >> tk->shift);
	return ts;
}

63
static void tk_set_xtime(struct timekeeper *tk, const struct timespec64 *ts)
64 65
{
	tk->xtime_sec = ts->tv_sec;
66
	tk->xtime_nsec = (u64)ts->tv_nsec << tk->shift;
67 68
}

69
static void tk_xtime_add(struct timekeeper *tk, const struct timespec64 *ts)
70 71
{
	tk->xtime_sec += ts->tv_sec;
72
	tk->xtime_nsec += (u64)ts->tv_nsec << tk->shift;
73
	tk_normalize_xtime(tk);
74
}
75

76
static void tk_set_wall_to_mono(struct timekeeper *tk, struct timespec64 wtm)
77
{
78
	struct timespec64 tmp;
79 80 81 82 83

	/*
	 * Verify consistency of: offset_real = -wall_to_monotonic
	 * before modifying anything
	 */
84
	set_normalized_timespec64(&tmp, -tk->wall_to_monotonic.tv_sec,
85
					-tk->wall_to_monotonic.tv_nsec);
86
	WARN_ON_ONCE(tk->offs_real.tv64 != timespec64_to_ktime(tmp).tv64);
87
	tk->wall_to_monotonic = wtm;
88 89
	set_normalized_timespec64(&tmp, -wtm.tv_sec, -wtm.tv_nsec);
	tk->offs_real = timespec64_to_ktime(tmp);
90
	tk->offs_tai = ktime_add(tk->offs_real, ktime_set(tk->tai_offset, 0));
91 92
}

93
static void tk_set_sleep_time(struct timekeeper *tk, struct timespec64 t)
94 95
{
	/* Verify consistency before modifying */
96
	WARN_ON_ONCE(tk->offs_boot.tv64 != timespec64_to_ktime(tk->total_sleep_time).tv64);
97 98

	tk->total_sleep_time	= t;
99
	tk->offs_boot		= timespec64_to_ktime(t);
100 101
}

102
/**
103
 * tk_setup_internals - Set up internals to use clocksource clock.
104
 *
105
 * @tk:		The target timekeeper to setup.
106 107 108 109 110 111 112
 * @clock:		Pointer to clocksource.
 *
 * Calculates a fixed cycle/nsec interval for a given clocksource/adjustment
 * pair and interval request.
 *
 * Unless you're the timekeeping code, you should not be using this!
 */
113
static void tk_setup_internals(struct timekeeper *tk, struct clocksource *clock)
114 115
{
	cycle_t interval;
116
	u64 tmp, ntpinterval;
117
	struct clocksource *old_clock;
118

119 120
	old_clock = tk->clock;
	tk->clock = clock;
121
	tk->cycle_last = clock->cycle_last = clock->read(clock);
122 123 124 125

	/* Do the ns -> cycle conversion first, using original mult */
	tmp = NTP_INTERVAL_LENGTH;
	tmp <<= clock->shift;
126
	ntpinterval = tmp;
127 128
	tmp += clock->mult/2;
	do_div(tmp, clock->mult);
129 130 131 132
	if (tmp == 0)
		tmp = 1;

	interval = (cycle_t) tmp;
133
	tk->cycle_interval = interval;
134 135

	/* Go back from cycles -> shifted ns */
136 137 138
	tk->xtime_interval = (u64) interval * clock->mult;
	tk->xtime_remainder = ntpinterval - tk->xtime_interval;
	tk->raw_interval =
139
		((u64) interval * clock->mult) >> clock->shift;
140

141 142 143 144
	 /* if changing clocks, convert xtime_nsec shift units */
	if (old_clock) {
		int shift_change = clock->shift - old_clock->shift;
		if (shift_change < 0)
145
			tk->xtime_nsec >>= -shift_change;
146
		else
147
			tk->xtime_nsec <<= shift_change;
148
	}
149
	tk->shift = clock->shift;
150

151 152
	tk->ntp_error = 0;
	tk->ntp_error_shift = NTP_SCALE_SHIFT - clock->shift;
153 154 155 156 157 158

	/*
	 * The timekeeper keeps its own mult values for the currently
	 * active clocksource. These value will be adjusted via NTP
	 * to counteract clock drifting.
	 */
159
	tk->mult = clock->mult;
160
}
161

162
/* Timekeeper helper functions. */
163 164

#ifdef CONFIG_ARCH_USES_GETTIMEOFFSET
165 166
static u32 default_arch_gettimeoffset(void) { return 0; }
u32 (*arch_gettimeoffset)(void) = default_arch_gettimeoffset;
167
#else
168
static inline u32 arch_gettimeoffset(void) { return 0; }
169 170
#endif

171
static inline s64 timekeeping_get_ns(struct timekeeper *tk)
172 173 174
{
	cycle_t cycle_now, cycle_delta;
	struct clocksource *clock;
175
	s64 nsec;
176 177

	/* read clocksource: */
178
	clock = tk->clock;
179 180 181 182 183
	cycle_now = clock->read(clock);

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

184 185
	nsec = cycle_delta * tk->mult + tk->xtime_nsec;
	nsec >>= tk->shift;
186

187
	/* If arch requires, add in get_arch_timeoffset() */
188
	return nsec + arch_gettimeoffset();
189 190
}

191
static inline s64 timekeeping_get_ns_raw(struct timekeeper *tk)
192 193 194
{
	cycle_t cycle_now, cycle_delta;
	struct clocksource *clock;
195
	s64 nsec;
196 197

	/* read clocksource: */
198
	clock = tk->clock;
199 200 201 202 203
	cycle_now = clock->read(clock);

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

204 205 206
	/* convert delta to nanoseconds. */
	nsec = clocksource_cyc2ns(cycle_delta, clock->mult, clock->shift);

207
	/* If arch requires, add in get_arch_timeoffset() */
208
	return nsec + arch_gettimeoffset();
209 210
}

211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244
#ifdef CONFIG_GENERIC_TIME_VSYSCALL_OLD

static inline void update_vsyscall(struct timekeeper *tk)
{
	struct timespec xt;

	xt = tk_xtime(tk);
	update_vsyscall_old(&xt, &tk->wall_to_monotonic, tk->clock, tk->mult);
}

static inline void old_vsyscall_fixup(struct timekeeper *tk)
{
	s64 remainder;

	/*
	* Store only full nanoseconds into xtime_nsec after rounding
	* it up and add the remainder to the error difference.
	* XXX - This is necessary to avoid small 1ns inconsistnecies caused
	* by truncating the remainder in vsyscalls. However, it causes
	* additional work to be done in timekeeping_adjust(). Once
	* the vsyscall implementations are converted to use xtime_nsec
	* (shifted nanoseconds), and CONFIG_GENERIC_TIME_VSYSCALL_OLD
	* users are removed, this can be killed.
	*/
	remainder = tk->xtime_nsec & ((1ULL << tk->shift) - 1);
	tk->xtime_nsec -= remainder;
	tk->xtime_nsec += 1ULL << tk->shift;
	tk->ntp_error += remainder << tk->ntp_error_shift;
	tk->ntp_error -= (1ULL << tk->shift) << tk->ntp_error_shift;
}
#else
#define old_vsyscall_fixup(tk)
#endif

245 246
static RAW_NOTIFIER_HEAD(pvclock_gtod_chain);

247
static void update_pvclock_gtod(struct timekeeper *tk, bool was_set)
248
{
249
	raw_notifier_call_chain(&pvclock_gtod_chain, was_set, tk);
250 251 252 253 254 255 256 257 258 259 260
}

/**
 * pvclock_gtod_register_notifier - register a pvclock timedata update listener
 */
int pvclock_gtod_register_notifier(struct notifier_block *nb)
{
	struct timekeeper *tk = &timekeeper;
	unsigned long flags;
	int ret;

261
	raw_spin_lock_irqsave(&timekeeper_lock, flags);
262
	ret = raw_notifier_chain_register(&pvclock_gtod_chain, nb);
263
	update_pvclock_gtod(tk, true);
264
	raw_spin_unlock_irqrestore(&timekeeper_lock, flags);
265 266 267 268 269 270 271 272 273 274 275 276 277 278

	return ret;
}
EXPORT_SYMBOL_GPL(pvclock_gtod_register_notifier);

/**
 * pvclock_gtod_unregister_notifier - unregister a pvclock
 * timedata update listener
 */
int pvclock_gtod_unregister_notifier(struct notifier_block *nb)
{
	unsigned long flags;
	int ret;

279
	raw_spin_lock_irqsave(&timekeeper_lock, flags);
280
	ret = raw_notifier_chain_unregister(&pvclock_gtod_chain, nb);
281
	raw_spin_unlock_irqrestore(&timekeeper_lock, flags);
282 283 284 285 286

	return ret;
}
EXPORT_SYMBOL_GPL(pvclock_gtod_unregister_notifier);

287
/* must hold timekeeper_lock */
288
static void timekeeping_update(struct timekeeper *tk, unsigned int action)
289
{
290
	if (action & TK_CLEAR_NTP) {
291
		tk->ntp_error = 0;
292 293
		ntp_clear();
	}
294
	update_vsyscall(tk);
295
	update_pvclock_gtod(tk, action & TK_CLOCK_WAS_SET);
296

297
	if (action & TK_MIRROR)
298
		memcpy(&shadow_timekeeper, &timekeeper, sizeof(timekeeper));
299 300
}

301
/**
302
 * timekeeping_forward_now - update clock to the current time
303
 *
304 305 306
 * Forward the current clock to update its state since the last call to
 * update_wall_time(). This is useful before significant clock changes,
 * as it avoids having to deal with this time offset explicitly.
307
 */
308
static void timekeeping_forward_now(struct timekeeper *tk)
309 310
{
	cycle_t cycle_now, cycle_delta;
311
	struct clocksource *clock;
312
	s64 nsec;
313

314
	clock = tk->clock;
315
	cycle_now = clock->read(clock);
316
	cycle_delta = (cycle_now - clock->cycle_last) & clock->mask;
317
	tk->cycle_last = clock->cycle_last = cycle_now;
318

319
	tk->xtime_nsec += cycle_delta * tk->mult;
320

321
	/* If arch requires, add in get_arch_timeoffset() */
322
	tk->xtime_nsec += (u64)arch_gettimeoffset() << tk->shift;
323

324
	tk_normalize_xtime(tk);
325

326
	nsec = clocksource_cyc2ns(cycle_delta, clock->mult, clock->shift);
327
	timespec64_add_ns(&tk->raw_time, nsec);
328 329 330
}

/**
331
 * __getnstimeofday64 - Returns the time of day in a timespec64.
332 333
 * @ts:		pointer to the timespec to be set
 *
334 335
 * Updates the time of day in the timespec.
 * Returns 0 on success, or -ve when suspended (timespec will be undefined).
336
 */
337
int __getnstimeofday64(struct timespec64 *ts)
338
{
339
	struct timekeeper *tk = &timekeeper;
340
	unsigned long seq;
341
	s64 nsecs = 0;
342 343

	do {
344
		seq = read_seqcount_begin(&timekeeper_seq);
345

346
		ts->tv_sec = tk->xtime_sec;
347
		nsecs = timekeeping_get_ns(tk);
348

349
	} while (read_seqcount_retry(&timekeeper_seq, seq));
350

351
	ts->tv_nsec = 0;
352
	timespec64_add_ns(ts, nsecs);
353 354 355 356 357 358 359 360 361

	/*
	 * Do not bail out early, in case there were callers still using
	 * the value, even in the face of the WARN_ON.
	 */
	if (unlikely(timekeeping_suspended))
		return -EAGAIN;
	return 0;
}
362
EXPORT_SYMBOL(__getnstimeofday64);
363 364

/**
365
 * getnstimeofday64 - Returns the time of day in a timespec64.
366 367 368 369
 * @ts:		pointer to the timespec to be set
 *
 * Returns the time of day in a timespec (WARN if suspended).
 */
370
void getnstimeofday64(struct timespec64 *ts)
371
{
372
	WARN_ON(__getnstimeofday64(ts));
373
}
374
EXPORT_SYMBOL(getnstimeofday64);
375

376 377
ktime_t ktime_get(void)
{
378
	struct timekeeper *tk = &timekeeper;
379 380 381 382 383 384
	unsigned int seq;
	s64 secs, nsecs;

	WARN_ON(timekeeping_suspended);

	do {
385
		seq = read_seqcount_begin(&timekeeper_seq);
386 387
		secs = tk->xtime_sec + tk->wall_to_monotonic.tv_sec;
		nsecs = timekeeping_get_ns(tk) + tk->wall_to_monotonic.tv_nsec;
388

389
	} while (read_seqcount_retry(&timekeeper_seq, seq));
390 391

	return ktime_set(secs, nsecs);
392 393 394 395
}
EXPORT_SYMBOL_GPL(ktime_get);

/**
396
 * ktime_get_ts64 - get the monotonic clock in timespec64 format
397 398 399 400 401 402
 * @ts:		pointer to timespec variable
 *
 * The function calculates the monotonic clock from the realtime
 * clock and the wall_to_monotonic offset and stores the result
 * in normalized timespec format in the variable pointed to by @ts.
 */
403
void ktime_get_ts64(struct timespec64 *ts)
404
{
405
	struct timekeeper *tk = &timekeeper;
406
	struct timespec64 tomono;
407
	s64 nsec;
408 409 410 411 412
	unsigned int seq;

	WARN_ON(timekeeping_suspended);

	do {
413
		seq = read_seqcount_begin(&timekeeper_seq);
414
		ts->tv_sec = tk->xtime_sec;
415
		nsec = timekeeping_get_ns(tk);
416
		tomono = tk->wall_to_monotonic;
417

418
	} while (read_seqcount_retry(&timekeeper_seq, seq));
419

420 421 422
	ts->tv_sec += tomono.tv_sec;
	ts->tv_nsec = 0;
	timespec64_add_ns(ts, nsec + tomono.tv_nsec);
423
}
424
EXPORT_SYMBOL_GPL(ktime_get_ts64);
425

J
John Stultz 已提交
426 427 428 429 430 431 432 433 434 435

/**
 * timekeeping_clocktai - Returns the TAI time of day in a timespec
 * @ts:		pointer to the timespec to be set
 *
 * Returns the time of day in a timespec.
 */
void timekeeping_clocktai(struct timespec *ts)
{
	struct timekeeper *tk = &timekeeper;
436
	struct timespec64 ts64;
J
John Stultz 已提交
437 438 439 440 441 442
	unsigned long seq;
	u64 nsecs;

	WARN_ON(timekeeping_suspended);

	do {
443
		seq = read_seqcount_begin(&timekeeper_seq);
J
John Stultz 已提交
444

445
		ts64.tv_sec = tk->xtime_sec + tk->tai_offset;
J
John Stultz 已提交
446 447
		nsecs = timekeeping_get_ns(tk);

448
	} while (read_seqcount_retry(&timekeeper_seq, seq));
J
John Stultz 已提交
449

450 451 452
	ts64.tv_nsec = 0;
	timespec64_add_ns(&ts64, nsecs);
	*ts = timespec64_to_timespec(ts64);
J
John Stultz 已提交
453 454 455 456 457

}
EXPORT_SYMBOL(timekeeping_clocktai);


458 459 460 461 462 463 464 465 466 467 468 469 470 471
/**
 * ktime_get_clocktai - Returns the TAI time of day in a ktime
 *
 * Returns the time of day in a ktime.
 */
ktime_t ktime_get_clocktai(void)
{
	struct timespec ts;

	timekeeping_clocktai(&ts);
	return timespec_to_ktime(ts);
}
EXPORT_SYMBOL(ktime_get_clocktai);

472 473 474 475 476 477 478 479 480 481 482 483 484
#ifdef CONFIG_NTP_PPS

/**
 * getnstime_raw_and_real - get day and raw monotonic time in timespec format
 * @ts_raw:	pointer to the timespec to be set to raw monotonic time
 * @ts_real:	pointer to the timespec to be set to the time of day
 *
 * This function reads both the time of day and raw monotonic time at the
 * same time atomically and stores the resulting timestamps in timespec
 * format.
 */
void getnstime_raw_and_real(struct timespec *ts_raw, struct timespec *ts_real)
{
485
	struct timekeeper *tk = &timekeeper;
486 487 488 489 490 491
	unsigned long seq;
	s64 nsecs_raw, nsecs_real;

	WARN_ON_ONCE(timekeeping_suspended);

	do {
492
		seq = read_seqcount_begin(&timekeeper_seq);
493

494
		*ts_raw = timespec64_to_timespec(tk->raw_time);
495
		ts_real->tv_sec = tk->xtime_sec;
496
		ts_real->tv_nsec = 0;
497

498 499
		nsecs_raw = timekeeping_get_ns_raw(tk);
		nsecs_real = timekeeping_get_ns(tk);
500

501
	} while (read_seqcount_retry(&timekeeper_seq, seq));
502 503 504 505 506 507 508 509

	timespec_add_ns(ts_raw, nsecs_raw);
	timespec_add_ns(ts_real, nsecs_real);
}
EXPORT_SYMBOL(getnstime_raw_and_real);

#endif /* CONFIG_NTP_PPS */

510 511 512 513
/**
 * do_gettimeofday - Returns the time of day in a timeval
 * @tv:		pointer to the timeval to be set
 *
514
 * NOTE: Users should be converted to using getnstimeofday()
515 516 517
 */
void do_gettimeofday(struct timeval *tv)
{
518
	struct timespec64 now;
519

520
	getnstimeofday64(&now);
521 522 523 524
	tv->tv_sec = now.tv_sec;
	tv->tv_usec = now.tv_nsec/1000;
}
EXPORT_SYMBOL(do_gettimeofday);
525

526 527 528 529 530 531
/**
 * 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
 */
532
int do_settimeofday(const struct timespec *tv)
533
{
534
	struct timekeeper *tk = &timekeeper;
535
	struct timespec64 ts_delta, xt, tmp;
536
	unsigned long flags;
537

538
	if (!timespec_valid_strict(tv))
539 540
		return -EINVAL;

541 542
	raw_spin_lock_irqsave(&timekeeper_lock, flags);
	write_seqcount_begin(&timekeeper_seq);
543

544
	timekeeping_forward_now(tk);
545

546
	xt = tk_xtime(tk);
547 548 549
	ts_delta.tv_sec = tv->tv_sec - xt.tv_sec;
	ts_delta.tv_nsec = tv->tv_nsec - xt.tv_nsec;

550
	tk_set_wall_to_mono(tk, timespec64_sub(tk->wall_to_monotonic, ts_delta));
551

552 553
	tmp = timespec_to_timespec64(*tv);
	tk_set_xtime(tk, &tmp);
554

555
	timekeeping_update(tk, TK_CLEAR_NTP | TK_MIRROR | TK_CLOCK_WAS_SET);
556

557 558
	write_seqcount_end(&timekeeper_seq);
	raw_spin_unlock_irqrestore(&timekeeper_lock, flags);
559 560 561 562 563 564 565 566

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

	return 0;
}
EXPORT_SYMBOL(do_settimeofday);

567 568 569 570 571 572 573 574
/**
 * timekeeping_inject_offset - Adds or subtracts from the current time.
 * @tv:		pointer to the timespec variable containing the offset
 *
 * Adds or subtracts an offset value from the current time.
 */
int timekeeping_inject_offset(struct timespec *ts)
{
575
	struct timekeeper *tk = &timekeeper;
576
	unsigned long flags;
577
	struct timespec64 ts64, tmp;
578
	int ret = 0;
579 580 581 582

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

583 584
	ts64 = timespec_to_timespec64(*ts);

585 586
	raw_spin_lock_irqsave(&timekeeper_lock, flags);
	write_seqcount_begin(&timekeeper_seq);
587

588
	timekeeping_forward_now(tk);
589

590
	/* Make sure the proposed value is valid */
591 592
	tmp = timespec64_add(tk_xtime(tk),  ts64);
	if (!timespec64_valid_strict(&tmp)) {
593 594 595
		ret = -EINVAL;
		goto error;
	}
596

597 598
	tk_xtime_add(tk, &ts64);
	tk_set_wall_to_mono(tk, timespec64_sub(tk->wall_to_monotonic, ts64));
599

600
error: /* even if we error out, we forwarded the time, so call update */
601
	timekeeping_update(tk, TK_CLEAR_NTP | TK_MIRROR | TK_CLOCK_WAS_SET);
602

603 604
	write_seqcount_end(&timekeeper_seq);
	raw_spin_unlock_irqrestore(&timekeeper_lock, flags);
605 606 607 608

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

609
	return ret;
610 611 612
}
EXPORT_SYMBOL(timekeeping_inject_offset);

613 614 615 616 617 618 619 620 621 622 623 624

/**
 * timekeeping_get_tai_offset - Returns current TAI offset from UTC
 *
 */
s32 timekeeping_get_tai_offset(void)
{
	struct timekeeper *tk = &timekeeper;
	unsigned int seq;
	s32 ret;

	do {
625
		seq = read_seqcount_begin(&timekeeper_seq);
626
		ret = tk->tai_offset;
627
	} while (read_seqcount_retry(&timekeeper_seq, seq));
628 629 630 631 632 633 634 635

	return ret;
}

/**
 * __timekeeping_set_tai_offset - Lock free worker function
 *
 */
636
static void __timekeeping_set_tai_offset(struct timekeeper *tk, s32 tai_offset)
637 638
{
	tk->tai_offset = tai_offset;
639
	tk->offs_tai = ktime_add(tk->offs_real, ktime_set(tai_offset, 0));
640 641 642 643 644 645 646 647 648 649 650
}

/**
 * timekeeping_set_tai_offset - Sets the current TAI offset from UTC
 *
 */
void timekeeping_set_tai_offset(s32 tai_offset)
{
	struct timekeeper *tk = &timekeeper;
	unsigned long flags;

651 652
	raw_spin_lock_irqsave(&timekeeper_lock, flags);
	write_seqcount_begin(&timekeeper_seq);
653
	__timekeeping_set_tai_offset(tk, tai_offset);
654
	timekeeping_update(tk, TK_MIRROR | TK_CLOCK_WAS_SET);
655 656
	write_seqcount_end(&timekeeper_seq);
	raw_spin_unlock_irqrestore(&timekeeper_lock, flags);
657
	clock_was_set();
658 659
}

660 661 662 663 664
/**
 * change_clocksource - Swaps clocksources if a new one is available
 *
 * Accumulates current time interval and initializes new clocksource
 */
665
static int change_clocksource(void *data)
666
{
667
	struct timekeeper *tk = &timekeeper;
668
	struct clocksource *new, *old;
669
	unsigned long flags;
670

671
	new = (struct clocksource *) data;
672

673 674
	raw_spin_lock_irqsave(&timekeeper_lock, flags);
	write_seqcount_begin(&timekeeper_seq);
675

676
	timekeeping_forward_now(tk);
677 678 679 680 681 682 683 684 685 686 687 688 689 690
	/*
	 * If the cs is in module, get a module reference. Succeeds
	 * for built-in code (owner == NULL) as well.
	 */
	if (try_module_get(new->owner)) {
		if (!new->enable || new->enable(new) == 0) {
			old = tk->clock;
			tk_setup_internals(tk, new);
			if (old->disable)
				old->disable(old);
			module_put(old->owner);
		} else {
			module_put(new->owner);
		}
691
	}
692
	timekeeping_update(tk, TK_CLEAR_NTP | TK_MIRROR | TK_CLOCK_WAS_SET);
693

694 695
	write_seqcount_end(&timekeeper_seq);
	raw_spin_unlock_irqrestore(&timekeeper_lock, flags);
696

697 698
	return 0;
}
699

700 701 702 703 704 705 706
/**
 * timekeeping_notify - Install a new clock source
 * @clock:		pointer to the clock source
 *
 * This function is called from clocksource.c after a new, better clock
 * source has been registered. The caller holds the clocksource_mutex.
 */
707
int timekeeping_notify(struct clocksource *clock)
708
{
709 710 711
	struct timekeeper *tk = &timekeeper;

	if (tk->clock == clock)
712
		return 0;
713
	stop_machine(change_clocksource, clock, NULL);
714
	tick_clock_notify();
715
	return tk->clock == clock ? 0 : -1;
716
}
717

718 719 720 721 722 723 724
/**
 * ktime_get_real - get the real (wall-) time in ktime_t format
 *
 * returns the time in ktime_t format
 */
ktime_t ktime_get_real(void)
{
725
	struct timespec64 now;
726

727
	getnstimeofday64(&now);
728

729
	return timespec64_to_ktime(now);
730 731
}
EXPORT_SYMBOL_GPL(ktime_get_real);
732

733 734 735 736 737 738 739 740
/**
 * getrawmonotonic - Returns the raw monotonic time in a timespec
 * @ts:		pointer to the timespec to be set
 *
 * Returns the raw monotonic time (completely un-modified by ntp)
 */
void getrawmonotonic(struct timespec *ts)
{
741
	struct timekeeper *tk = &timekeeper;
742
	struct timespec64 ts64;
743 744 745 746
	unsigned long seq;
	s64 nsecs;

	do {
747
		seq = read_seqcount_begin(&timekeeper_seq);
748
		nsecs = timekeeping_get_ns_raw(tk);
749
		ts64 = tk->raw_time;
750

751
	} while (read_seqcount_retry(&timekeeper_seq, seq));
752

753 754
	timespec64_add_ns(&ts64, nsecs);
	*ts = timespec64_to_timespec(ts64);
755 756 757
}
EXPORT_SYMBOL(getrawmonotonic);

758
/**
759
 * timekeeping_valid_for_hres - Check if timekeeping is suitable for hres
760
 */
761
int timekeeping_valid_for_hres(void)
762
{
763
	struct timekeeper *tk = &timekeeper;
764 765 766 767
	unsigned long seq;
	int ret;

	do {
768
		seq = read_seqcount_begin(&timekeeper_seq);
769

770
		ret = tk->clock->flags & CLOCK_SOURCE_VALID_FOR_HRES;
771

772
	} while (read_seqcount_retry(&timekeeper_seq, seq));
773 774 775 776

	return ret;
}

777 778 779 780 781
/**
 * timekeeping_max_deferment - Returns max time the clocksource can be deferred
 */
u64 timekeeping_max_deferment(void)
{
782
	struct timekeeper *tk = &timekeeper;
J
John Stultz 已提交
783 784
	unsigned long seq;
	u64 ret;
785

J
John Stultz 已提交
786
	do {
787
		seq = read_seqcount_begin(&timekeeper_seq);
J
John Stultz 已提交
788

789
		ret = tk->clock->max_idle_ns;
J
John Stultz 已提交
790

791
	} while (read_seqcount_retry(&timekeeper_seq, seq));
J
John Stultz 已提交
792 793

	return ret;
794 795
}

796
/**
797
 * read_persistent_clock -  Return time from the persistent clock.
798 799
 *
 * Weak dummy function for arches that do not yet support it.
800 801
 * Reads the time from the battery backed persistent clock.
 * Returns a timespec with tv_sec=0 and tv_nsec=0 if unsupported.
802 803 804
 *
 *  XXX - Do be sure to remove it once all arches implement it.
 */
805
void __weak read_persistent_clock(struct timespec *ts)
806
{
807 808
	ts->tv_sec = 0;
	ts->tv_nsec = 0;
809 810
}

811 812 813 814 815 816 817 818 819
/**
 * read_boot_clock -  Return time of the system start.
 *
 * Weak dummy function for arches that do not yet support it.
 * Function to read the exact time the system has been started.
 * Returns a timespec with tv_sec=0 and tv_nsec=0 if unsupported.
 *
 *  XXX - Do be sure to remove it once all arches implement it.
 */
820
void __weak read_boot_clock(struct timespec *ts)
821 822 823 824 825
{
	ts->tv_sec = 0;
	ts->tv_nsec = 0;
}

826 827 828 829 830
/*
 * timekeeping_init - Initializes the clocksource and common timekeeping values
 */
void __init timekeeping_init(void)
{
831
	struct timekeeper *tk = &timekeeper;
832
	struct clocksource *clock;
833
	unsigned long flags;
834 835
	struct timespec64 now, boot, tmp;
	struct timespec ts;
836

837 838 839
	read_persistent_clock(&ts);
	now = timespec_to_timespec64(ts);
	if (!timespec64_valid_strict(&now)) {
840 841 842 843
		pr_warn("WARNING: Persistent clock returned invalid value!\n"
			"         Check your CMOS/BIOS settings.\n");
		now.tv_sec = 0;
		now.tv_nsec = 0;
844 845
	} else if (now.tv_sec || now.tv_nsec)
		persistent_clock_exist = true;
846

847 848 849
	read_boot_clock(&ts);
	boot = timespec_to_timespec64(ts);
	if (!timespec64_valid_strict(&boot)) {
850 851 852 853 854
		pr_warn("WARNING: Boot clock returned invalid value!\n"
			"         Check your CMOS/BIOS settings.\n");
		boot.tv_sec = 0;
		boot.tv_nsec = 0;
	}
855

856 857
	raw_spin_lock_irqsave(&timekeeper_lock, flags);
	write_seqcount_begin(&timekeeper_seq);
858 859
	ntp_init();

860
	clock = clocksource_default_clock();
861 862
	if (clock->enable)
		clock->enable(clock);
863
	tk_setup_internals(tk, clock);
864

865 866 867
	tk_set_xtime(tk, &now);
	tk->raw_time.tv_sec = 0;
	tk->raw_time.tv_nsec = 0;
868
	if (boot.tv_sec == 0 && boot.tv_nsec == 0)
869
		boot = tk_xtime(tk);
870

871
	set_normalized_timespec64(&tmp, -boot.tv_sec, -boot.tv_nsec);
872
	tk_set_wall_to_mono(tk, tmp);
873 874 875

	tmp.tv_sec = 0;
	tmp.tv_nsec = 0;
876
	tk_set_sleep_time(tk, tmp);
877

878 879
	memcpy(&shadow_timekeeper, &timekeeper, sizeof(timekeeper));

880 881
	write_seqcount_end(&timekeeper_seq);
	raw_spin_unlock_irqrestore(&timekeeper_lock, flags);
882 883 884
}

/* time in seconds when suspend began */
885
static struct timespec64 timekeeping_suspend_time;
886

887 888 889 890 891 892 893
/**
 * __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.
 */
894
static void __timekeeping_inject_sleeptime(struct timekeeper *tk,
895
					   struct timespec64 *delta)
896
{
897
	if (!timespec64_valid_strict(delta)) {
898 899 900
		printk_deferred(KERN_WARNING
				"__timekeeping_inject_sleeptime: Invalid "
				"sleep delta value!\n");
901 902
		return;
	}
903
	tk_xtime_add(tk, delta);
904 905
	tk_set_wall_to_mono(tk, timespec64_sub(tk->wall_to_monotonic, *delta));
	tk_set_sleep_time(tk, timespec64_add(tk->total_sleep_time, *delta));
906
	tk_debug_account_sleep_time(delta);
907 908 909 910 911 912 913 914 915 916 917 918 919 920
}

/**
 * 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)
{
921
	struct timekeeper *tk = &timekeeper;
922
	struct timespec64 tmp;
923
	unsigned long flags;
924

925 926 927 928 929
	/*
	 * Make sure we don't set the clock twice, as timekeeping_resume()
	 * already did it
	 */
	if (has_persistent_clock())
930 931
		return;

932 933
	raw_spin_lock_irqsave(&timekeeper_lock, flags);
	write_seqcount_begin(&timekeeper_seq);
J
John Stultz 已提交
934

935
	timekeeping_forward_now(tk);
936

937 938
	tmp = timespec_to_timespec64(*delta);
	__timekeeping_inject_sleeptime(tk, &tmp);
939

940
	timekeeping_update(tk, TK_CLEAR_NTP | TK_MIRROR | TK_CLOCK_WAS_SET);
941

942 943
	write_seqcount_end(&timekeeper_seq);
	raw_spin_unlock_irqrestore(&timekeeper_lock, flags);
944 945 946 947 948

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

949 950 951 952 953 954 955
/**
 * timekeeping_resume - Resumes the generic timekeeping subsystem.
 *
 * This is for the generic clocksource timekeeping.
 * xtime/wall_to_monotonic/jiffies/etc are
 * still managed by arch specific suspend/resume code.
 */
956
static void timekeeping_resume(void)
957
{
958
	struct timekeeper *tk = &timekeeper;
959
	struct clocksource *clock = tk->clock;
960
	unsigned long flags;
961 962
	struct timespec64 ts_new, ts_delta;
	struct timespec tmp;
963 964
	cycle_t cycle_now, cycle_delta;
	bool suspendtime_found = false;
965

966 967
	read_persistent_clock(&tmp);
	ts_new = timespec_to_timespec64(tmp);
968

969
	clockevents_resume();
970 971
	clocksource_resume();

972 973
	raw_spin_lock_irqsave(&timekeeper_lock, flags);
	write_seqcount_begin(&timekeeper_seq);
974

975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009
	/*
	 * After system resumes, we need to calculate the suspended time and
	 * compensate it for the OS time. There are 3 sources that could be
	 * used: Nonstop clocksource during suspend, persistent clock and rtc
	 * device.
	 *
	 * One specific platform may have 1 or 2 or all of them, and the
	 * preference will be:
	 *	suspend-nonstop clocksource -> persistent clock -> rtc
	 * The less preferred source will only be tried if there is no better
	 * usable source. The rtc part is handled separately in rtc core code.
	 */
	cycle_now = clock->read(clock);
	if ((clock->flags & CLOCK_SOURCE_SUSPEND_NONSTOP) &&
		cycle_now > clock->cycle_last) {
		u64 num, max = ULLONG_MAX;
		u32 mult = clock->mult;
		u32 shift = clock->shift;
		s64 nsec = 0;

		cycle_delta = (cycle_now - clock->cycle_last) & clock->mask;

		/*
		 * "cycle_delta * mutl" may cause 64 bits overflow, if the
		 * suspended time is too long. In that case we need do the
		 * 64 bits math carefully
		 */
		do_div(max, mult);
		if (cycle_delta > max) {
			num = div64_u64(cycle_delta, max);
			nsec = (((u64) max * mult) >> shift) * num;
			cycle_delta -= num * max;
		}
		nsec += ((u64) cycle_delta * mult) >> shift;

1010
		ts_delta = ns_to_timespec64(nsec);
1011
		suspendtime_found = true;
1012 1013
	} else if (timespec64_compare(&ts_new, &timekeeping_suspend_time) > 0) {
		ts_delta = timespec64_sub(ts_new, timekeeping_suspend_time);
1014
		suspendtime_found = true;
1015
	}
1016 1017 1018 1019 1020

	if (suspendtime_found)
		__timekeeping_inject_sleeptime(tk, &ts_delta);

	/* Re-base the last cycle value */
1021
	tk->cycle_last = clock->cycle_last = cycle_now;
1022
	tk->ntp_error = 0;
1023
	timekeeping_suspended = 0;
1024
	timekeeping_update(tk, TK_MIRROR | TK_CLOCK_WAS_SET);
1025 1026
	write_seqcount_end(&timekeeper_seq);
	raw_spin_unlock_irqrestore(&timekeeper_lock, flags);
1027 1028 1029 1030 1031 1032

	touch_softlockup_watchdog();

	clockevents_notify(CLOCK_EVT_NOTIFY_RESUME, NULL);

	/* Resume hrtimers */
1033
	hrtimers_resume();
1034 1035
}

1036
static int timekeeping_suspend(void)
1037
{
1038
	struct timekeeper *tk = &timekeeper;
1039
	unsigned long flags;
1040 1041 1042
	struct timespec64		delta, delta_delta;
	static struct timespec64	old_delta;
	struct timespec tmp;
1043

1044 1045
	read_persistent_clock(&tmp);
	timekeeping_suspend_time = timespec_to_timespec64(tmp);
1046

1047 1048 1049 1050 1051 1052 1053 1054
	/*
	 * On some systems the persistent_clock can not be detected at
	 * timekeeping_init by its return value, so if we see a valid
	 * value returned, update the persistent_clock_exists flag.
	 */
	if (timekeeping_suspend_time.tv_sec || timekeeping_suspend_time.tv_nsec)
		persistent_clock_exist = true;

1055 1056
	raw_spin_lock_irqsave(&timekeeper_lock, flags);
	write_seqcount_begin(&timekeeper_seq);
1057
	timekeeping_forward_now(tk);
1058
	timekeeping_suspended = 1;
1059 1060 1061 1062 1063 1064 1065

	/*
	 * To avoid drift caused by repeated suspend/resumes,
	 * which each can add ~1 second drift error,
	 * try to compensate so the difference in system time
	 * and persistent_clock time stays close to constant.
	 */
1066 1067
	delta = timespec64_sub(tk_xtime(tk), timekeeping_suspend_time);
	delta_delta = timespec64_sub(delta, old_delta);
1068 1069 1070 1071 1072 1073 1074 1075 1076
	if (abs(delta_delta.tv_sec)  >= 2) {
		/*
		 * if delta_delta is too large, assume time correction
		 * has occured and set old_delta to the current delta.
		 */
		old_delta = delta;
	} else {
		/* Otherwise try to adjust old_system to compensate */
		timekeeping_suspend_time =
1077
			timespec64_add(timekeeping_suspend_time, delta_delta);
1078
	}
1079 1080

	timekeeping_update(tk, TK_MIRROR);
1081 1082
	write_seqcount_end(&timekeeper_seq);
	raw_spin_unlock_irqrestore(&timekeeper_lock, flags);
1083 1084

	clockevents_notify(CLOCK_EVT_NOTIFY_SUSPEND, NULL);
M
Magnus Damm 已提交
1085
	clocksource_suspend();
1086
	clockevents_suspend();
1087 1088 1089 1090 1091

	return 0;
}

/* sysfs resume/suspend bits for timekeeping */
1092
static struct syscore_ops timekeeping_syscore_ops = {
1093 1094 1095 1096
	.resume		= timekeeping_resume,
	.suspend	= timekeeping_suspend,
};

1097
static int __init timekeeping_init_ops(void)
1098
{
1099 1100
	register_syscore_ops(&timekeeping_syscore_ops);
	return 0;
1101 1102
}

1103
device_initcall(timekeeping_init_ops);
1104 1105 1106 1107 1108

/*
 * If the error is already larger, we look ahead even further
 * to compensate for late or lost adjustments.
 */
1109 1110
static __always_inline int timekeeping_bigadjust(struct timekeeper *tk,
						 s64 error, s64 *interval,
1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122
						 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
L
Li Zefan 已提交
1123
	 * here.  This is tuned so that an error of about 1 msec is adjusted
1124 1125
	 * within about 1 sec (or 2^20 nsec in 2^SHIFT_HZ ticks).
	 */
1126
	error2 = tk->ntp_error >> (NTP_SCALE_SHIFT + 22 - 2 * SHIFT_HZ);
1127 1128 1129 1130 1131 1132 1133 1134
	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.
	 */
1135 1136
	tick_error = ntp_tick_length() >> (tk->ntp_error_shift + 1);
	tick_error -= tk->xtime_interval >> 1;
1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160
	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.
 */
1161
static void timekeeping_adjust(struct timekeeper *tk, s64 offset)
1162
{
1163
	s64 error, interval = tk->cycle_interval;
1164 1165
	int adj;

1166
	/*
1167
	 * The point of this is to check if the error is greater than half
1168 1169 1170 1171 1172
	 * an interval.
	 *
	 * First we shift it down from NTP_SHIFT to clocksource->shifted nsecs.
	 *
	 * Note we subtract one in the shift, so that error is really error*2.
1173 1174
	 * This "saves" dividing(shifting) interval twice, but keeps the
	 * (error > interval) comparison as still measuring if error is
1175
	 * larger than half an interval.
1176
	 *
1177
	 * Note: It does not "save" on aggravation when reading the code.
1178
	 */
1179
	error = tk->ntp_error >> (tk->ntp_error_shift - 1);
1180
	if (error > interval) {
1181 1182
		/*
		 * We now divide error by 4(via shift), which checks if
1183
		 * the error is greater than twice the interval.
1184 1185 1186
		 * If it is greater, we need a bigadjust, if its smaller,
		 * we can adjust by 1.
		 */
1187 1188 1189 1190
		error >>= 2;
		if (likely(error <= interval))
			adj = 1;
		else
1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206
			adj = timekeeping_bigadjust(tk, error, &interval, &offset);
	} else {
		if (error < -interval) {
			/* See comment above, this is just switched for the negative */
			error >>= 2;
			if (likely(error >= -interval)) {
				adj = -1;
				interval = -interval;
				offset = -offset;
			} else {
				adj = timekeeping_bigadjust(tk, error, &interval, &offset);
			}
		} else {
			goto out_adjust;
		}
	}
1207

1208 1209
	if (unlikely(tk->clock->maxadj &&
		(tk->mult + adj > tk->clock->mult + tk->clock->maxadj))) {
1210
		printk_deferred_once(KERN_WARNING
1211
			"Adjusting %s more than 11%% (%ld vs %ld)\n",
1212 1213
			tk->clock->name, (long)tk->mult + adj,
			(long)tk->clock->mult + tk->clock->maxadj);
1214
	}
1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262 1263
	/*
	 * So the following can be confusing.
	 *
	 * To keep things simple, lets assume adj == 1 for now.
	 *
	 * When adj != 1, remember that the interval and offset values
	 * have been appropriately scaled so the math is the same.
	 *
	 * The basic idea here is that we're increasing the multiplier
	 * by one, this causes the xtime_interval to be incremented by
	 * one cycle_interval. This is because:
	 *	xtime_interval = cycle_interval * mult
	 * So if mult is being incremented by one:
	 *	xtime_interval = cycle_interval * (mult + 1)
	 * Its the same as:
	 *	xtime_interval = (cycle_interval * mult) + cycle_interval
	 * Which can be shortened to:
	 *	xtime_interval += cycle_interval
	 *
	 * So offset stores the non-accumulated cycles. Thus the current
	 * time (in shifted nanoseconds) is:
	 *	now = (offset * adj) + xtime_nsec
	 * Now, even though we're adjusting the clock frequency, we have
	 * to keep time consistent. In other words, we can't jump back
	 * in time, and we also want to avoid jumping forward in time.
	 *
	 * So given the same offset value, we need the time to be the same
	 * both before and after the freq adjustment.
	 *	now = (offset * adj_1) + xtime_nsec_1
	 *	now = (offset * adj_2) + xtime_nsec_2
	 * So:
	 *	(offset * adj_1) + xtime_nsec_1 =
	 *		(offset * adj_2) + xtime_nsec_2
	 * And we know:
	 *	adj_2 = adj_1 + 1
	 * So:
	 *	(offset * adj_1) + xtime_nsec_1 =
	 *		(offset * (adj_1+1)) + xtime_nsec_2
	 *	(offset * adj_1) + xtime_nsec_1 =
	 *		(offset * adj_1) + offset + xtime_nsec_2
	 * Canceling the sides:
	 *	xtime_nsec_1 = offset + xtime_nsec_2
	 * Which gives us:
	 *	xtime_nsec_2 = xtime_nsec_1 - offset
	 * Which simplfies to:
	 *	xtime_nsec -= offset
	 *
	 * XXX - TODO: Doc ntp_error calculation.
	 */
1264 1265 1266 1267
	tk->mult += adj;
	tk->xtime_interval += interval;
	tk->xtime_nsec -= offset;
	tk->ntp_error -= (interval - offset) << tk->ntp_error_shift;
1268

1269
out_adjust:
1270 1271 1272 1273 1274 1275 1276 1277 1278 1279 1280 1281 1282 1283
	/*
	 * It may be possible that when we entered this function, xtime_nsec
	 * was very small.  Further, if we're slightly speeding the clocksource
	 * in the code above, its possible the required corrective factor to
	 * xtime_nsec could cause it to underflow.
	 *
	 * Now, since we already accumulated the second, cannot simply roll
	 * the accumulated second back, since the NTP subsystem has been
	 * notified via second_overflow. So instead we push xtime_nsec forward
	 * by the amount we underflowed, and add that amount into the error.
	 *
	 * We'll correct this error next time through this function, when
	 * xtime_nsec is not as small.
	 */
1284 1285 1286 1287
	if (unlikely((s64)tk->xtime_nsec < 0)) {
		s64 neg = -(s64)tk->xtime_nsec;
		tk->xtime_nsec = 0;
		tk->ntp_error += neg << tk->ntp_error_shift;
1288 1289
	}

1290 1291
}

1292 1293 1294 1295 1296 1297 1298 1299
/**
 * accumulate_nsecs_to_secs - Accumulates nsecs into secs
 *
 * Helper function that accumulates a the nsecs greater then a second
 * from the xtime_nsec field to the xtime_secs field.
 * It also calls into the NTP code to handle leapsecond processing.
 *
 */
1300
static inline unsigned int accumulate_nsecs_to_secs(struct timekeeper *tk)
1301 1302
{
	u64 nsecps = (u64)NSEC_PER_SEC << tk->shift;
1303
	unsigned int clock_set = 0;
1304 1305 1306 1307 1308 1309 1310 1311 1312

	while (tk->xtime_nsec >= nsecps) {
		int leap;

		tk->xtime_nsec -= nsecps;
		tk->xtime_sec++;

		/* Figure out if its a leap sec and apply if needed */
		leap = second_overflow(tk->xtime_sec);
1313
		if (unlikely(leap)) {
1314
			struct timespec64 ts;
1315 1316

			tk->xtime_sec += leap;
1317

1318 1319 1320
			ts.tv_sec = leap;
			ts.tv_nsec = 0;
			tk_set_wall_to_mono(tk,
1321
				timespec64_sub(tk->wall_to_monotonic, ts));
1322

1323 1324
			__timekeeping_set_tai_offset(tk, tk->tai_offset - leap);

1325
			clock_set = TK_CLOCK_WAS_SET;
1326
		}
1327
	}
1328
	return clock_set;
1329 1330
}

1331 1332 1333 1334 1335 1336 1337 1338 1339
/**
 * logarithmic_accumulation - shifted accumulation of cycles
 *
 * This functions accumulates a shifted interval of cycles into
 * into a shifted interval nanoseconds. Allows for O(log) accumulation
 * loop.
 *
 * Returns the unconsumed cycles.
 */
1340
static cycle_t logarithmic_accumulation(struct timekeeper *tk, cycle_t offset,
1341 1342
						u32 shift,
						unsigned int *clock_set)
1343
{
T
Thomas Gleixner 已提交
1344
	cycle_t interval = tk->cycle_interval << shift;
1345
	u64 raw_nsecs;
1346

1347
	/* If the offset is smaller then a shifted interval, do nothing */
T
Thomas Gleixner 已提交
1348
	if (offset < interval)
1349 1350 1351
		return offset;

	/* Accumulate one shifted interval */
T
Thomas Gleixner 已提交
1352
	offset -= interval;
1353
	tk->cycle_last += interval;
1354

1355
	tk->xtime_nsec += tk->xtime_interval << shift;
1356
	*clock_set |= accumulate_nsecs_to_secs(tk);
1357

1358
	/* Accumulate raw time */
1359
	raw_nsecs = (u64)tk->raw_interval << shift;
1360
	raw_nsecs += tk->raw_time.tv_nsec;
1361 1362 1363
	if (raw_nsecs >= NSEC_PER_SEC) {
		u64 raw_secs = raw_nsecs;
		raw_nsecs = do_div(raw_secs, NSEC_PER_SEC);
1364
		tk->raw_time.tv_sec += raw_secs;
1365
	}
1366
	tk->raw_time.tv_nsec = raw_nsecs;
1367 1368

	/* Accumulate error between NTP and clock interval */
1369 1370 1371
	tk->ntp_error += ntp_tick_length() << shift;
	tk->ntp_error -= (tk->xtime_interval + tk->xtime_remainder) <<
						(tk->ntp_error_shift + shift);
1372 1373 1374 1375

	return offset;
}

1376 1377 1378 1379
/**
 * update_wall_time - Uses the current clocksource to increment the wall time
 *
 */
1380
void update_wall_time(void)
1381
{
1382
	struct clocksource *clock;
1383 1384
	struct timekeeper *real_tk = &timekeeper;
	struct timekeeper *tk = &shadow_timekeeper;
1385
	cycle_t offset;
1386
	int shift = 0, maxshift;
1387
	unsigned int clock_set = 0;
J
John Stultz 已提交
1388 1389
	unsigned long flags;

1390
	raw_spin_lock_irqsave(&timekeeper_lock, flags);
1391 1392 1393

	/* Make sure we're fully resumed: */
	if (unlikely(timekeeping_suspended))
J
John Stultz 已提交
1394
		goto out;
1395

1396
	clock = real_tk->clock;
J
John Stultz 已提交
1397 1398

#ifdef CONFIG_ARCH_USES_GETTIMEOFFSET
1399
	offset = real_tk->cycle_interval;
J
John Stultz 已提交
1400 1401
#else
	offset = (clock->read(clock) - clock->cycle_last) & clock->mask;
1402 1403
#endif

1404
	/* Check if there's really nothing to do */
1405
	if (offset < real_tk->cycle_interval)
1406 1407
		goto out;

1408 1409 1410 1411
	/*
	 * With NO_HZ we may have to accumulate many cycle_intervals
	 * (think "ticks") worth of time at once. To do this efficiently,
	 * we calculate the largest doubling multiple of cycle_intervals
1412
	 * that is smaller than the offset.  We then accumulate that
1413 1414
	 * chunk in one go, and then try to consume the next smaller
	 * doubled multiple.
1415
	 */
1416
	shift = ilog2(offset) - ilog2(tk->cycle_interval);
1417
	shift = max(0, shift);
1418
	/* Bound shift to one less than what overflows tick_length */
1419
	maxshift = (64 - (ilog2(ntp_tick_length())+1)) - 1;
1420
	shift = min(shift, maxshift);
1421
	while (offset >= tk->cycle_interval) {
1422 1423
		offset = logarithmic_accumulation(tk, offset, shift,
							&clock_set);
1424
		if (offset < tk->cycle_interval<<shift)
1425
			shift--;
1426 1427 1428
	}

	/* correct the clock when NTP error is too big */
1429
	timekeeping_adjust(tk, offset);
1430

J
John Stultz 已提交
1431
	/*
1432 1433 1434 1435
	 * XXX This can be killed once everyone converts
	 * to the new update_vsyscall.
	 */
	old_vsyscall_fixup(tk);
1436

J
John Stultz 已提交
1437 1438
	/*
	 * Finally, make sure that after the rounding
1439
	 * xtime_nsec isn't larger than NSEC_PER_SEC
J
John Stultz 已提交
1440
	 */
1441
	clock_set |= accumulate_nsecs_to_secs(tk);
L
Linus Torvalds 已提交
1442

1443
	write_seqcount_begin(&timekeeper_seq);
1444 1445
	/* Update clock->cycle_last with the new value */
	clock->cycle_last = tk->cycle_last;
1446 1447 1448 1449 1450 1451 1452 1453 1454 1455 1456
	/*
	 * Update the real timekeeper.
	 *
	 * We could avoid this memcpy by switching pointers, but that
	 * requires changes to all other timekeeper usage sites as
	 * well, i.e. move the timekeeper pointer getter into the
	 * spinlocked/seqcount protected sections. And we trade this
	 * memcpy under the timekeeper_seq against one before we start
	 * updating.
	 */
	memcpy(real_tk, tk, sizeof(*tk));
1457
	timekeeping_update(real_tk, clock_set);
1458
	write_seqcount_end(&timekeeper_seq);
1459
out:
1460
	raw_spin_unlock_irqrestore(&timekeeper_lock, flags);
1461
	if (clock_set)
1462 1463
		/* Have to call _delayed version, since in irq context*/
		clock_was_set_delayed();
1464
}
T
Tomas Janousek 已提交
1465 1466 1467 1468 1469

/**
 * getboottime - Return the real time of system boot.
 * @ts:		pointer to the timespec to be set
 *
1470
 * Returns the wall-time of boot in a timespec.
T
Tomas Janousek 已提交
1471 1472 1473 1474 1475 1476 1477 1478
 *
 * 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)
{
1479
	struct timekeeper *tk = &timekeeper;
1480
	struct timespec boottime = {
1481 1482 1483 1484
		.tv_sec = tk->wall_to_monotonic.tv_sec +
				tk->total_sleep_time.tv_sec,
		.tv_nsec = tk->wall_to_monotonic.tv_nsec +
				tk->total_sleep_time.tv_nsec
1485
	};
1486 1487

	set_normalized_timespec(ts, -boottime.tv_sec, -boottime.tv_nsec);
T
Tomas Janousek 已提交
1488
}
1489
EXPORT_SYMBOL_GPL(getboottime);
T
Tomas Janousek 已提交
1490

1491 1492 1493 1494 1495 1496 1497 1498 1499 1500 1501
/**
 * get_monotonic_boottime - Returns monotonic time since boot
 * @ts:		pointer to the timespec to be set
 *
 * Returns the monotonic time since boot in a timespec.
 *
 * This is similar to CLOCK_MONTONIC/ktime_get_ts, but also
 * includes the time spent in suspend.
 */
void get_monotonic_boottime(struct timespec *ts)
{
1502
	struct timekeeper *tk = &timekeeper;
1503
	struct timespec64 tomono, sleep, ret;
1504
	s64 nsec;
1505 1506 1507 1508 1509
	unsigned int seq;

	WARN_ON(timekeeping_suspended);

	do {
1510
		seq = read_seqcount_begin(&timekeeper_seq);
1511
		ret.tv_sec = tk->xtime_sec;
1512
		nsec = timekeeping_get_ns(tk);
1513 1514
		tomono = tk->wall_to_monotonic;
		sleep = tk->total_sleep_time;
1515

1516
	} while (read_seqcount_retry(&timekeeper_seq, seq));
1517

1518 1519 1520 1521
	ret.tv_sec += tomono.tv_sec + sleep.tv_sec;
	ret.tv_nsec = 0;
	timespec64_add_ns(&ret, nsec + tomono.tv_nsec + sleep.tv_nsec);
	*ts = timespec64_to_timespec(ret);
1522 1523 1524 1525 1526 1527 1528 1529 1530 1531 1532 1533 1534 1535 1536 1537 1538 1539 1540 1541
}
EXPORT_SYMBOL_GPL(get_monotonic_boottime);

/**
 * ktime_get_boottime - Returns monotonic time since boot in a ktime
 *
 * Returns the monotonic time since boot in a ktime
 *
 * This is similar to CLOCK_MONTONIC/ktime_get, but also
 * includes the time spent in suspend.
 */
ktime_t ktime_get_boottime(void)
{
	struct timespec ts;

	get_monotonic_boottime(&ts);
	return timespec_to_ktime(ts);
}
EXPORT_SYMBOL_GPL(ktime_get_boottime);

T
Tomas Janousek 已提交
1542 1543 1544 1545 1546 1547
/**
 * 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)
{
1548
	struct timekeeper *tk = &timekeeper;
1549
	struct timespec64 ts64;
1550

1551 1552 1553
	ts64 = timespec_to_timespec64(*ts);
	ts64 = timespec64_add(ts64, tk->total_sleep_time);
	*ts = timespec64_to_timespec(ts64);
T
Tomas Janousek 已提交
1554
}
1555
EXPORT_SYMBOL_GPL(monotonic_to_bootbased);
1556

1557 1558
unsigned long get_seconds(void)
{
1559 1560 1561
	struct timekeeper *tk = &timekeeper;

	return tk->xtime_sec;
1562 1563 1564
}
EXPORT_SYMBOL(get_seconds);

1565 1566
struct timespec __current_kernel_time(void)
{
1567 1568
	struct timekeeper *tk = &timekeeper;

1569
	return timespec64_to_timespec(tk_xtime(tk));
1570
}
1571

1572 1573
struct timespec current_kernel_time(void)
{
1574
	struct timekeeper *tk = &timekeeper;
1575
	struct timespec64 now;
1576 1577 1578
	unsigned long seq;

	do {
1579
		seq = read_seqcount_begin(&timekeeper_seq);
L
Linus Torvalds 已提交
1580

1581
		now = tk_xtime(tk);
1582
	} while (read_seqcount_retry(&timekeeper_seq, seq));
1583

1584
	return timespec64_to_timespec(now);
1585 1586
}
EXPORT_SYMBOL(current_kernel_time);
1587 1588 1589

struct timespec get_monotonic_coarse(void)
{
1590
	struct timekeeper *tk = &timekeeper;
1591
	struct timespec64 now, mono;
1592 1593 1594
	unsigned long seq;

	do {
1595
		seq = read_seqcount_begin(&timekeeper_seq);
L
Linus Torvalds 已提交
1596

1597 1598
		now = tk_xtime(tk);
		mono = tk->wall_to_monotonic;
1599
	} while (read_seqcount_retry(&timekeeper_seq, seq));
1600

1601
	set_normalized_timespec64(&now, now.tv_sec + mono.tv_sec,
1602
				now.tv_nsec + mono.tv_nsec);
1603 1604

	return timespec64_to_timespec(now);
1605
}
1606 1607

/*
1608
 * Must hold jiffies_lock
1609 1610 1611 1612 1613 1614
 */
void do_timer(unsigned long ticks)
{
	jiffies_64 += ticks;
	calc_global_load(ticks);
}
1615 1616

/**
1617 1618 1619 1620 1621 1622
 * ktime_get_update_offsets_tick - hrtimer helper
 * @offs_real:	pointer to storage for monotonic -> realtime offset
 * @offs_boot:	pointer to storage for monotonic -> boottime offset
 * @offs_tai:	pointer to storage for monotonic -> clock tai offset
 *
 * Returns monotonic time at last tick and various offsets
1623
 */
1624 1625
ktime_t ktime_get_update_offsets_tick(ktime_t *offs_real, ktime_t *offs_boot,
							ktime_t *offs_tai)
1626
{
1627
	struct timekeeper *tk = &timekeeper;
1628
	struct timespec64 ts;
1629 1630
	ktime_t now;
	unsigned int seq;
1631 1632

	do {
1633
		seq = read_seqcount_begin(&timekeeper_seq);
1634 1635 1636 1637 1638

		ts = tk_xtime(tk);
		*offs_real = tk->offs_real;
		*offs_boot = tk->offs_boot;
		*offs_tai = tk->offs_tai;
1639
	} while (read_seqcount_retry(&timekeeper_seq, seq));
1640 1641 1642 1643

	now = ktime_set(ts.tv_sec, ts.tv_nsec);
	now = ktime_sub(now, *offs_real);
	return now;
1644
}
T
Torben Hohn 已提交
1645

1646 1647
#ifdef CONFIG_HIGH_RES_TIMERS
/**
1648
 * ktime_get_update_offsets_now - hrtimer helper
1649 1650
 * @offs_real:	pointer to storage for monotonic -> realtime offset
 * @offs_boot:	pointer to storage for monotonic -> boottime offset
1651
 * @offs_tai:	pointer to storage for monotonic -> clock tai offset
1652 1653
 *
 * Returns current monotonic time and updates the offsets
1654
 * Called from hrtimer_interrupt() or retrigger_next_event()
1655
 */
1656
ktime_t ktime_get_update_offsets_now(ktime_t *offs_real, ktime_t *offs_boot,
1657
							ktime_t *offs_tai)
1658
{
1659
	struct timekeeper *tk = &timekeeper;
1660 1661 1662 1663 1664
	ktime_t now;
	unsigned int seq;
	u64 secs, nsecs;

	do {
1665
		seq = read_seqcount_begin(&timekeeper_seq);
1666

1667 1668
		secs = tk->xtime_sec;
		nsecs = timekeeping_get_ns(tk);
1669

1670 1671
		*offs_real = tk->offs_real;
		*offs_boot = tk->offs_boot;
1672
		*offs_tai = tk->offs_tai;
1673
	} while (read_seqcount_retry(&timekeeper_seq, seq));
1674 1675 1676 1677 1678 1679 1680

	now = ktime_add_ns(ktime_set(secs, 0), nsecs);
	now = ktime_sub(now, *offs_real);
	return now;
}
#endif

1681 1682 1683 1684 1685
/**
 * ktime_get_monotonic_offset() - get wall_to_monotonic in ktime_t format
 */
ktime_t ktime_get_monotonic_offset(void)
{
1686
	struct timekeeper *tk = &timekeeper;
1687
	unsigned long seq;
1688
	struct timespec64 wtom;
1689 1690

	do {
1691
		seq = read_seqcount_begin(&timekeeper_seq);
1692
		wtom = tk->wall_to_monotonic;
1693
	} while (read_seqcount_retry(&timekeeper_seq, seq));
J
John Stultz 已提交
1694

1695
	return timespec64_to_ktime(wtom);
1696
}
1697 1698
EXPORT_SYMBOL_GPL(ktime_get_monotonic_offset);

1699 1700 1701 1702 1703
/**
 * do_adjtimex() - Accessor function to NTP __do_adjtimex function
 */
int do_adjtimex(struct timex *txc)
{
1704
	struct timekeeper *tk = &timekeeper;
1705
	unsigned long flags;
1706
	struct timespec64 ts;
1707
	s32 orig_tai, tai;
1708 1709 1710 1711 1712 1713 1714
	int ret;

	/* Validate the data before disabling interrupts */
	ret = ntp_validate_timex(txc);
	if (ret)
		return ret;

1715 1716 1717 1718 1719 1720 1721 1722 1723 1724 1725
	if (txc->modes & ADJ_SETOFFSET) {
		struct timespec delta;
		delta.tv_sec  = txc->time.tv_sec;
		delta.tv_nsec = txc->time.tv_usec;
		if (!(txc->modes & ADJ_NANO))
			delta.tv_nsec *= 1000;
		ret = timekeeping_inject_offset(&delta);
		if (ret)
			return ret;
	}

1726
	getnstimeofday64(&ts);
1727

1728 1729 1730
	raw_spin_lock_irqsave(&timekeeper_lock, flags);
	write_seqcount_begin(&timekeeper_seq);

1731
	orig_tai = tai = tk->tai_offset;
1732
	ret = __do_adjtimex(txc, &ts, &tai);
1733

1734 1735
	if (tai != orig_tai) {
		__timekeeping_set_tai_offset(tk, tai);
1736
		timekeeping_update(tk, TK_MIRROR | TK_CLOCK_WAS_SET);
1737
	}
1738 1739 1740
	write_seqcount_end(&timekeeper_seq);
	raw_spin_unlock_irqrestore(&timekeeper_lock, flags);

1741 1742 1743
	if (tai != orig_tai)
		clock_was_set();

1744 1745
	ntp_notify_cmos_timer();

1746 1747
	return ret;
}
1748 1749 1750 1751 1752 1753 1754

#ifdef CONFIG_NTP_PPS
/**
 * hardpps() - Accessor function to NTP __hardpps function
 */
void hardpps(const struct timespec *phase_ts, const struct timespec *raw_ts)
{
1755 1756 1757 1758 1759
	unsigned long flags;

	raw_spin_lock_irqsave(&timekeeper_lock, flags);
	write_seqcount_begin(&timekeeper_seq);

1760
	__hardpps(phase_ts, raw_ts);
1761 1762 1763

	write_seqcount_end(&timekeeper_seq);
	raw_spin_unlock_irqrestore(&timekeeper_lock, flags);
1764 1765 1766 1767
}
EXPORT_SYMBOL(hardpps);
#endif

T
Torben Hohn 已提交
1768 1769 1770 1771 1772 1773 1774 1775
/**
 * xtime_update() - advances the timekeeping infrastructure
 * @ticks:	number of ticks, that have elapsed since the last call.
 *
 * Must be called with interrupts disabled.
 */
void xtime_update(unsigned long ticks)
{
1776
	write_seqlock(&jiffies_lock);
T
Torben Hohn 已提交
1777
	do_timer(ticks);
1778
	write_sequnlock(&jiffies_lock);
1779
	update_wall_time();
T
Torben Hohn 已提交
1780
}