time.c 33.1 KB
Newer Older
L
Linus Torvalds 已提交
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50
/*
 * Common time routines among all ppc machines.
 *
 * Written by Cort Dougan (cort@cs.nmt.edu) to merge
 * Paul Mackerras' version and mine for PReP and Pmac.
 * MPC8xx/MBX changes by Dan Malek (dmalek@jlc.net).
 * Converted for 64-bit by Mike Corrigan (mikejc@us.ibm.com)
 *
 * First round of bugfixes by Gabriel Paubert (paubert@iram.es)
 * to make clock more stable (2.4.0-test5). The only thing
 * that this code assumes is that the timebases have been synchronized
 * by firmware on SMP and are never stopped (never do sleep
 * on SMP then, nap and doze are OK).
 * 
 * Speeded up do_gettimeofday by getting rid of references to
 * xtime (which required locks for consistency). (mikejc@us.ibm.com)
 *
 * TODO (not necessarily in this file):
 * - improve precision and reproducibility of timebase frequency
 * measurement at boot time. (for iSeries, we calibrate the timebase
 * against the Titan chip's clock.)
 * - for astronomical applications: add a new function to get
 * non ambiguous timestamps even around leap seconds. This needs
 * a new timestamp format and a good name.
 *
 * 1997-09-10  Updated NTP code according to technical memorandum Jan '96
 *             "A Kernel Model for Precision Timekeeping" by Dave Mills
 *
 *      This program is free software; you can redistribute it and/or
 *      modify it under the terms of the GNU General Public License
 *      as published by the Free Software Foundation; either version
 *      2 of the License, or (at your option) any later version.
 */

#include <linux/config.h>
#include <linux/errno.h>
#include <linux/module.h>
#include <linux/sched.h>
#include <linux/kernel.h>
#include <linux/param.h>
#include <linux/string.h>
#include <linux/mm.h>
#include <linux/interrupt.h>
#include <linux/timex.h>
#include <linux/kernel_stat.h>
#include <linux/time.h>
#include <linux/init.h>
#include <linux/profile.h>
#include <linux/cpu.h>
#include <linux/security.h>
51 52
#include <linux/percpu.h>
#include <linux/rtc.h>
53
#include <linux/jiffies.h>
54
#include <linux/posix-timers.h>
L
Linus Torvalds 已提交
55 56 57 58 59 60 61 62 63

#include <asm/io.h>
#include <asm/processor.h>
#include <asm/nvram.h>
#include <asm/cache.h>
#include <asm/machdep.h>
#include <asm/uaccess.h>
#include <asm/time.h>
#include <asm/prom.h>
64 65
#include <asm/irq.h>
#include <asm/div64.h>
P
Paul Mackerras 已提交
66
#include <asm/smp.h>
67
#include <asm/vdso_datapage.h>
68
#ifdef CONFIG_PPC64
69
#include <asm/firmware.h>
70 71
#endif
#ifdef CONFIG_PPC_ISERIES
72
#include <asm/iseries/it_lp_queue.h>
73
#include <asm/iseries/hv_call_xm.h>
74
#endif
75
#include <asm/smp.h>
L
Linus Torvalds 已提交
76 77 78 79 80 81 82 83 84 85

/* keep track of when we need to update the rtc */
time_t last_rtc_update;
extern int piranha_simulator;
#ifdef CONFIG_PPC_ISERIES
unsigned long iSeries_recal_titan = 0;
unsigned long iSeries_recal_tb = 0; 
static unsigned long first_settimeofday = 1;
#endif

86 87 88
/* The decrementer counts down by 128 every 128ns on a 601. */
#define DECREMENTER_COUNT_601	(1000000000 / HZ)

L
Linus Torvalds 已提交
89 90
#define XSEC_PER_SEC (1024*1024)

91 92 93 94 95 96 97
#ifdef CONFIG_PPC64
#define SCALE_XSEC(xsec, max)	(((xsec) * max) / XSEC_PER_SEC)
#else
/* compute ((xsec << 12) * max) >> 32 */
#define SCALE_XSEC(xsec, max)	mulhwu((xsec) << 12, max)
#endif

L
Linus Torvalds 已提交
98 99 100 101
unsigned long tb_ticks_per_jiffy;
unsigned long tb_ticks_per_usec = 100; /* sane default */
EXPORT_SYMBOL(tb_ticks_per_usec);
unsigned long tb_ticks_per_sec;
102
EXPORT_SYMBOL(tb_ticks_per_sec);	/* for cputime_t conversions */
103 104
u64 tb_to_xs;
unsigned tb_to_us;
105 106 107 108 109 110 111 112 113

#define TICKLEN_SCALE	(SHIFT_SCALE - 10)
u64 last_tick_len;	/* units are ns / 2^TICKLEN_SCALE */
u64 ticklen_to_xs;	/* 0.64 fraction */

/* If last_tick_len corresponds to about 1/HZ seconds, then
   last_tick_len << TICKLEN_SHIFT will be about 2^63. */
#define TICKLEN_SHIFT	(63 - 30 - TICKLEN_SCALE + SHIFT_HZ)

L
Linus Torvalds 已提交
114
DEFINE_SPINLOCK(rtc_lock);
115
EXPORT_SYMBOL_GPL(rtc_lock);
L
Linus Torvalds 已提交
116

117 118
u64 tb_to_ns_scale;
unsigned tb_to_ns_shift;
L
Linus Torvalds 已提交
119 120 121 122 123 124

struct gettimeofday_struct do_gtod;

extern unsigned long wall_jiffies;

extern struct timezone sys_tz;
125
static long timezone_offset;
L
Linus Torvalds 已提交
126

127 128 129
unsigned long ppc_proc_freq;
unsigned long ppc_tb_freq;

130 131 132 133 134 135 136 137 138 139
u64 tb_last_jiffy __cacheline_aligned_in_smp;
unsigned long tb_last_stamp;

/*
 * Note that on ppc32 this only stores the bottom 32 bits of
 * the timebase value, but that's enough to tell when a jiffy
 * has passed.
 */
DEFINE_PER_CPU(unsigned long, last_jiffy);

140 141 142 143 144 145 146
#ifdef CONFIG_VIRT_CPU_ACCOUNTING
/*
 * Factors for converting from cputime_t (timebase ticks) to
 * jiffies, milliseconds, seconds, and clock_t (1/USER_HZ seconds).
 * These are all stored as 0.64 fixed-point binary fractions.
 */
u64 __cputime_jiffies_factor;
147
EXPORT_SYMBOL(__cputime_jiffies_factor);
148
u64 __cputime_msec_factor;
149
EXPORT_SYMBOL(__cputime_msec_factor);
150
u64 __cputime_sec_factor;
151
EXPORT_SYMBOL(__cputime_sec_factor);
152
u64 __cputime_clockt_factor;
153
EXPORT_SYMBOL(__cputime_clockt_factor);
154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263

static void calc_cputime_factors(void)
{
	struct div_result res;

	div128_by_32(HZ, 0, tb_ticks_per_sec, &res);
	__cputime_jiffies_factor = res.result_low;
	div128_by_32(1000, 0, tb_ticks_per_sec, &res);
	__cputime_msec_factor = res.result_low;
	div128_by_32(1, 0, tb_ticks_per_sec, &res);
	__cputime_sec_factor = res.result_low;
	div128_by_32(USER_HZ, 0, tb_ticks_per_sec, &res);
	__cputime_clockt_factor = res.result_low;
}

/*
 * Read the PURR on systems that have it, otherwise the timebase.
 */
static u64 read_purr(void)
{
	if (cpu_has_feature(CPU_FTR_PURR))
		return mfspr(SPRN_PURR);
	return mftb();
}

/*
 * Account time for a transition between system, hard irq
 * or soft irq state.
 */
void account_system_vtime(struct task_struct *tsk)
{
	u64 now, delta;
	unsigned long flags;

	local_irq_save(flags);
	now = read_purr();
	delta = now - get_paca()->startpurr;
	get_paca()->startpurr = now;
	if (!in_interrupt()) {
		delta += get_paca()->system_time;
		get_paca()->system_time = 0;
	}
	account_system_time(tsk, 0, delta);
	local_irq_restore(flags);
}

/*
 * Transfer the user and system times accumulated in the paca
 * by the exception entry and exit code to the generic process
 * user and system time records.
 * Must be called with interrupts disabled.
 */
void account_process_vtime(struct task_struct *tsk)
{
	cputime_t utime;

	utime = get_paca()->user_time;
	get_paca()->user_time = 0;
	account_user_time(tsk, utime);
}

static void account_process_time(struct pt_regs *regs)
{
	int cpu = smp_processor_id();

	account_process_vtime(current);
	run_local_timers();
	if (rcu_pending(cpu))
		rcu_check_callbacks(cpu, user_mode(regs));
	scheduler_tick();
 	run_posix_cpu_timers(current);
}

#ifdef CONFIG_PPC_SPLPAR
/*
 * Stuff for accounting stolen time.
 */
struct cpu_purr_data {
	int	initialized;			/* thread is running */
	u64	tb0;			/* timebase at origin time */
	u64	purr0;			/* PURR at origin time */
	u64	tb;			/* last TB value read */
	u64	purr;			/* last PURR value read */
	u64	stolen;			/* stolen time so far */
	spinlock_t lock;
};

static DEFINE_PER_CPU(struct cpu_purr_data, cpu_purr_data);

static void snapshot_tb_and_purr(void *data)
{
	struct cpu_purr_data *p = &__get_cpu_var(cpu_purr_data);

	p->tb0 = mftb();
	p->purr0 = mfspr(SPRN_PURR);
	p->tb = p->tb0;
	p->purr = 0;
	wmb();
	p->initialized = 1;
}

/*
 * Called during boot when all cpus have come up.
 */
void snapshot_timebases(void)
{
	int cpu;

	if (!cpu_has_feature(CPU_FTR_PURR))
		return;
264
	for_each_possible_cpu(cpu)
265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357
		spin_lock_init(&per_cpu(cpu_purr_data, cpu).lock);
	on_each_cpu(snapshot_tb_and_purr, NULL, 0, 1);
}

void calculate_steal_time(void)
{
	u64 tb, purr, t0;
	s64 stolen;
	struct cpu_purr_data *p0, *pme, *phim;
	int cpu;

	if (!cpu_has_feature(CPU_FTR_PURR))
		return;
	cpu = smp_processor_id();
	pme = &per_cpu(cpu_purr_data, cpu);
	if (!pme->initialized)
		return;		/* this can happen in early boot */
	p0 = &per_cpu(cpu_purr_data, cpu & ~1);
	phim = &per_cpu(cpu_purr_data, cpu ^ 1);
	spin_lock(&p0->lock);
	tb = mftb();
	purr = mfspr(SPRN_PURR) - pme->purr0;
	if (!phim->initialized || !cpu_online(cpu ^ 1)) {
		stolen = (tb - pme->tb) - (purr - pme->purr);
	} else {
		t0 = pme->tb0;
		if (phim->tb0 < t0)
			t0 = phim->tb0;
		stolen = phim->tb - t0 - phim->purr - purr - p0->stolen;
	}
	if (stolen > 0) {
		account_steal_time(current, stolen);
		p0->stolen += stolen;
	}
	pme->tb = tb;
	pme->purr = purr;
	spin_unlock(&p0->lock);
}

/*
 * Must be called before the cpu is added to the online map when
 * a cpu is being brought up at runtime.
 */
static void snapshot_purr(void)
{
	int cpu;
	u64 purr;
	struct cpu_purr_data *p0, *pme, *phim;
	unsigned long flags;

	if (!cpu_has_feature(CPU_FTR_PURR))
		return;
	cpu = smp_processor_id();
	pme = &per_cpu(cpu_purr_data, cpu);
	p0 = &per_cpu(cpu_purr_data, cpu & ~1);
	phim = &per_cpu(cpu_purr_data, cpu ^ 1);
	spin_lock_irqsave(&p0->lock, flags);
	pme->tb = pme->tb0 = mftb();
	purr = mfspr(SPRN_PURR);
	if (!phim->initialized) {
		pme->purr = 0;
		pme->purr0 = purr;
	} else {
		/* set p->purr and p->purr0 for no change in p0->stolen */
		pme->purr = phim->tb - phim->tb0 - phim->purr - p0->stolen;
		pme->purr0 = purr - pme->purr;
	}
	pme->initialized = 1;
	spin_unlock_irqrestore(&p0->lock, flags);
}

#endif /* CONFIG_PPC_SPLPAR */

#else /* ! CONFIG_VIRT_CPU_ACCOUNTING */
#define calc_cputime_factors()
#define account_process_time(regs)	update_process_times(user_mode(regs))
#define calculate_steal_time()		do { } while (0)
#endif

#if !(defined(CONFIG_VIRT_CPU_ACCOUNTING) && defined(CONFIG_PPC_SPLPAR))
#define snapshot_purr()			do { } while (0)
#endif

/*
 * Called when a cpu comes up after the system has finished booting,
 * i.e. as a result of a hotplug cpu action.
 */
void snapshot_timebase(void)
{
	__get_cpu_var(last_jiffy) = get_tb();
	snapshot_purr();
}

358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385
void __delay(unsigned long loops)
{
	unsigned long start;
	int diff;

	if (__USE_RTC()) {
		start = get_rtcl();
		do {
			/* the RTCL register wraps at 1000000000 */
			diff = get_rtcl() - start;
			if (diff < 0)
				diff += 1000000000;
		} while (diff < loops);
	} else {
		start = get_tbl();
		while (get_tbl() - start < loops)
			HMT_low();
		HMT_medium();
	}
}
EXPORT_SYMBOL(__delay);

void udelay(unsigned long usecs)
{
	__delay(tb_ticks_per_usec * usecs);
}
EXPORT_SYMBOL(udelay);

L
Linus Torvalds 已提交
386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403
static __inline__ void timer_check_rtc(void)
{
        /*
         * update the rtc when needed, this should be performed on the
         * right fraction of a second. Half or full second ?
         * Full second works on mk48t59 clocks, others need testing.
         * Note that this update is basically only used through 
         * the adjtimex system calls. Setting the HW clock in
         * any other way is a /dev/rtc and userland business.
         * This is still wrong by -0.5/+1.5 jiffies because of the
         * timer interrupt resolution and possible delay, but here we 
         * hit a quantization limit which can only be solved by higher
         * resolution timers and decoupling time management from timer
         * interrupts. This is also wrong on the clocks
         * which require being written at the half second boundary.
         * We should have an rtc call that only sets the minutes and
         * seconds like on Intel to avoid problems with non UTC clocks.
         */
404
        if (ppc_md.set_rtc_time && ntp_synced() &&
405
	    xtime.tv_sec - last_rtc_update >= 659 &&
406
	    abs((xtime.tv_nsec/1000) - (1000000-1000000/HZ)) < 500000/HZ) {
407 408 409 410 411 412 413 414 415
		struct rtc_time tm;
		to_tm(xtime.tv_sec + 1 + timezone_offset, &tm);
		tm.tm_year -= 1900;
		tm.tm_mon -= 1;
		if (ppc_md.set_rtc_time(&tm) == 0)
			last_rtc_update = xtime.tv_sec + 1;
		else
			/* Try again one minute later */
			last_rtc_update += 60;
L
Linus Torvalds 已提交
416 417 418 419 420 421
        }
}

/*
 * This version of gettimeofday has microsecond resolution.
 */
422
static inline void __do_gettimeofday(struct timeval *tv, u64 tb_val)
L
Linus Torvalds 已提交
423
{
424 425 426 427
	unsigned long sec, usec;
	u64 tb_ticks, xsec;
	struct gettimeofday_vars *temp_varp;
	u64 temp_tb_to_xs, temp_stamp_xsec;
L
Linus Torvalds 已提交
428 429 430 431 432 433 434 435 436 437 438

	/*
	 * These calculations are faster (gets rid of divides)
	 * if done in units of 1/2^20 rather than microseconds.
	 * The conversion to microseconds at the end is done
	 * without a divide (and in fact, without a multiply)
	 */
	temp_varp = do_gtod.varp;
	tb_ticks = tb_val - temp_varp->tb_orig_stamp;
	temp_tb_to_xs = temp_varp->tb_to_xs;
	temp_stamp_xsec = temp_varp->stamp_xsec;
439
	xsec = temp_stamp_xsec + mulhdu(tb_ticks, temp_tb_to_xs);
L
Linus Torvalds 已提交
440
	sec = xsec / XSEC_PER_SEC;
441 442
	usec = (unsigned long)xsec & (XSEC_PER_SEC - 1);
	usec = SCALE_XSEC(usec, 1000000);
L
Linus Torvalds 已提交
443 444 445 446 447 448 449

	tv->tv_sec = sec;
	tv->tv_usec = usec;
}

void do_gettimeofday(struct timeval *tv)
{
450 451 452
	if (__USE_RTC()) {
		/* do this the old way */
		unsigned long flags, seq;
453
		unsigned int sec, nsec, usec;
454 455 456 457 458 459

		do {
			seq = read_seqbegin_irqsave(&xtime_lock, flags);
			sec = xtime.tv_sec;
			nsec = xtime.tv_nsec + tb_ticks_since(tb_last_stamp);
		} while (read_seqretry_irqrestore(&xtime_lock, seq, flags));
460
		usec = nsec / 1000;
461 462 463 464 465 466 467 468
		while (usec >= 1000000) {
			usec -= 1000000;
			++sec;
		}
		tv->tv_sec = sec;
		tv->tv_usec = usec;
		return;
	}
L
Linus Torvalds 已提交
469 470 471 472 473 474
	__do_gettimeofday(tv, get_tb());
}

EXPORT_SYMBOL(do_gettimeofday);

/*
475 476 477 478 479 480
 * There are two copies of tb_to_xs and stamp_xsec so that no
 * lock is needed to access and use these values in
 * do_gettimeofday.  We alternate the copies and as long as a
 * reasonable time elapses between changes, there will never
 * be inconsistent values.  ntpd has a minimum of one minute
 * between updates.
L
Linus Torvalds 已提交
481
 */
482
static inline void update_gtod(u64 new_tb_stamp, u64 new_stamp_xsec,
483
			       u64 new_tb_to_xs)
L
Linus Torvalds 已提交
484 485
{
	unsigned temp_idx;
486
	struct gettimeofday_vars *temp_varp;
L
Linus Torvalds 已提交
487 488 489 490

	temp_idx = (do_gtod.var_idx == 0);
	temp_varp = &do_gtod.vars[temp_idx];

491 492
	temp_varp->tb_to_xs = new_tb_to_xs;
	temp_varp->tb_orig_stamp = new_tb_stamp;
L
Linus Torvalds 已提交
493
	temp_varp->stamp_xsec = new_stamp_xsec;
494
	smp_mb();
L
Linus Torvalds 已提交
495 496 497
	do_gtod.varp = temp_varp;
	do_gtod.var_idx = temp_idx;

498 499 500 501 502 503 504 505
	/*
	 * tb_update_count is used to allow the userspace gettimeofday code
	 * to assure itself that it sees a consistent view of the tb_to_xs and
	 * stamp_xsec variables.  It reads the tb_update_count, then reads
	 * tb_to_xs and stamp_xsec and then reads tb_update_count again.  If
	 * the two values of tb_update_count match and are even then the
	 * tb_to_xs and stamp_xsec values are consistent.  If not, then it
	 * loops back and reads them again until this criteria is met.
506 507
	 * We expect the caller to have done the first increment of
	 * vdso_data->tb_update_count already.
508
	 */
509 510 511 512 513
	vdso_data->tb_orig_stamp = new_tb_stamp;
	vdso_data->stamp_xsec = new_stamp_xsec;
	vdso_data->tb_to_xs = new_tb_to_xs;
	vdso_data->wtom_clock_sec = wall_to_monotonic.tv_sec;
	vdso_data->wtom_clock_nsec = wall_to_monotonic.tv_nsec;
514
	smp_wmb();
515
	++(vdso_data->tb_update_count);
516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531
}

/*
 * When the timebase - tb_orig_stamp gets too big, we do a manipulation
 * between tb_orig_stamp and stamp_xsec. The goal here is to keep the
 * difference tb - tb_orig_stamp small enough to always fit inside a
 * 32 bits number. This is a requirement of our fast 32 bits userland
 * implementation in the vdso. If we "miss" a call to this function
 * (interrupt latency, CPU locked in a spinlock, ...) and we end up
 * with a too big difference, then the vdso will fallback to calling
 * the syscall
 */
static __inline__ void timer_recalc_offset(u64 cur_tb)
{
	unsigned long offset;
	u64 new_stamp_xsec;
532
	u64 tlen, t2x;
533 534
	u64 tb, xsec_old, xsec_new;
	struct gettimeofday_vars *varp;
535

536 537
	if (__USE_RTC())
		return;
538
	tlen = current_tick_length();
539
	offset = cur_tb - do_gtod.varp->tb_orig_stamp;
540 541
	if (tlen == last_tick_len && offset < 0x80000000u)
		return;
542 543 544 545 546 547 548 549
	if (tlen != last_tick_len) {
		t2x = mulhdu(tlen << TICKLEN_SHIFT, ticklen_to_xs);
		last_tick_len = tlen;
	} else
		t2x = do_gtod.varp->tb_to_xs;
	new_stamp_xsec = (u64) xtime.tv_nsec * XSEC_PER_SEC;
	do_div(new_stamp_xsec, 1000000000);
	new_stamp_xsec += (u64) xtime.tv_sec * XSEC_PER_SEC;
550 551 552 553 554 555 556 557 558 559 560 561 562 563 564

	++vdso_data->tb_update_count;
	smp_mb();

	/*
	 * Make sure time doesn't go backwards for userspace gettimeofday.
	 */
	tb = get_tb();
	varp = do_gtod.varp;
	xsec_old = mulhdu(tb - varp->tb_orig_stamp, varp->tb_to_xs)
		+ varp->stamp_xsec;
	xsec_new = mulhdu(tb - cur_tb, t2x) + new_stamp_xsec;
	if (xsec_new < xsec_old)
		new_stamp_xsec += xsec_old - xsec_new;

565
	update_gtod(cur_tb, new_stamp_xsec, t2x);
L
Linus Torvalds 已提交
566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614
}

#ifdef CONFIG_SMP
unsigned long profile_pc(struct pt_regs *regs)
{
	unsigned long pc = instruction_pointer(regs);

	if (in_lock_functions(pc))
		return regs->link;

	return pc;
}
EXPORT_SYMBOL(profile_pc);
#endif

#ifdef CONFIG_PPC_ISERIES

/* 
 * This function recalibrates the timebase based on the 49-bit time-of-day
 * value in the Titan chip.  The Titan is much more accurate than the value
 * returned by the service processor for the timebase frequency.  
 */

static void iSeries_tb_recal(void)
{
	struct div_result divres;
	unsigned long titan, tb;
	tb = get_tb();
	titan = HvCallXm_loadTod();
	if ( iSeries_recal_titan ) {
		unsigned long tb_ticks = tb - iSeries_recal_tb;
		unsigned long titan_usec = (titan - iSeries_recal_titan) >> 12;
		unsigned long new_tb_ticks_per_sec   = (tb_ticks * USEC_PER_SEC)/titan_usec;
		unsigned long new_tb_ticks_per_jiffy = (new_tb_ticks_per_sec+(HZ/2))/HZ;
		long tick_diff = new_tb_ticks_per_jiffy - tb_ticks_per_jiffy;
		char sign = '+';		
		/* make sure tb_ticks_per_sec and tb_ticks_per_jiffy are consistent */
		new_tb_ticks_per_sec = new_tb_ticks_per_jiffy * HZ;

		if ( tick_diff < 0 ) {
			tick_diff = -tick_diff;
			sign = '-';
		}
		if ( tick_diff ) {
			if ( tick_diff < tb_ticks_per_jiffy/25 ) {
				printk( "Titan recalibrate: new tb_ticks_per_jiffy = %lu (%c%ld)\n",
						new_tb_ticks_per_jiffy, sign, tick_diff );
				tb_ticks_per_jiffy = new_tb_ticks_per_jiffy;
				tb_ticks_per_sec   = new_tb_ticks_per_sec;
615
				calc_cputime_factors();
L
Linus Torvalds 已提交
616 617 618 619
				div128_by_32( XSEC_PER_SEC, 0, tb_ticks_per_sec, &divres );
				do_gtod.tb_ticks_per_sec = tb_ticks_per_sec;
				tb_to_xs = divres.result_low;
				do_gtod.varp->tb_to_xs = tb_to_xs;
620 621
				vdso_data->tb_ticks_per_sec = tb_ticks_per_sec;
				vdso_data->tb_to_xs = tb_to_xs;
L
Linus Torvalds 已提交
622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649
			}
			else {
				printk( "Titan recalibrate: FAILED (difference > 4 percent)\n"
					"                   new tb_ticks_per_jiffy = %lu\n"
					"                   old tb_ticks_per_jiffy = %lu\n",
					new_tb_ticks_per_jiffy, tb_ticks_per_jiffy );
			}
		}
	}
	iSeries_recal_titan = titan;
	iSeries_recal_tb = tb;
}
#endif

/*
 * For iSeries shared processors, we have to let the hypervisor
 * set the hardware decrementer.  We set a virtual decrementer
 * in the lppaca and call the hypervisor if the virtual
 * decrementer is less than the current value in the hardware
 * decrementer. (almost always the new decrementer value will
 * be greater than the current hardware decementer so the hypervisor
 * call will not be needed)
 */

/*
 * timer_interrupt - gets called when the decrementer overflows,
 * with interrupts disabled.
 */
650
void timer_interrupt(struct pt_regs * regs)
L
Linus Torvalds 已提交
651 652
{
	int next_dec;
653 654 655 656 657 658 659
	int cpu = smp_processor_id();
	unsigned long ticks;

#ifdef CONFIG_PPC32
	if (atomic_read(&ppc_n_lost_interrupts) != 0)
		do_IRQ(regs);
#endif
L
Linus Torvalds 已提交
660 661 662 663

	irq_enter();

	profile_tick(CPU_PROFILING, regs);
664
	calculate_steal_time();
L
Linus Torvalds 已提交
665

666
#ifdef CONFIG_PPC_ISERIES
667
	get_lppaca()->int_dword.fields.decr_int = 0;
668 669 670 671 672 673 674 675 676
#endif

	while ((ticks = tb_ticks_since(per_cpu(last_jiffy, cpu)))
	       >= tb_ticks_per_jiffy) {
		/* Update last_jiffy */
		per_cpu(last_jiffy, cpu) += tb_ticks_per_jiffy;
		/* Handle RTCL overflow on 601 */
		if (__USE_RTC() && per_cpu(last_jiffy, cpu) >= 1000000000)
			per_cpu(last_jiffy, cpu) -= 1000000000;
L
Linus Torvalds 已提交
677 678 679 680 681 682 683 684 685

		/*
		 * We cannot disable the decrementer, so in the period
		 * between this cpu's being marked offline in cpu_online_map
		 * and calling stop-self, it is taking timer interrupts.
		 * Avoid calling into the scheduler rebalancing code if this
		 * is the case.
		 */
		if (!cpu_is_offline(cpu))
686
			account_process_time(regs);
687

L
Linus Torvalds 已提交
688 689 690 691
		/*
		 * No need to check whether cpu is offline here; boot_cpuid
		 * should have been fixed up by now.
		 */
692 693 694 695
		if (cpu != boot_cpuid)
			continue;

		write_seqlock(&xtime_lock);
696 697
		tb_last_jiffy += tb_ticks_per_jiffy;
		tb_last_stamp = per_cpu(last_jiffy, cpu);
698
		do_timer(regs);
699
		timer_recalc_offset(tb_last_jiffy);
700 701
		timer_check_rtc();
		write_sequnlock(&xtime_lock);
L
Linus Torvalds 已提交
702 703
	}
	
704
	next_dec = tb_ticks_per_jiffy - ticks;
L
Linus Torvalds 已提交
705 706 707
	set_dec(next_dec);

#ifdef CONFIG_PPC_ISERIES
708
	if (hvlpevent_is_pending())
709
		process_hvlpevents(regs);
L
Linus Torvalds 已提交
710 711
#endif

712
#ifdef CONFIG_PPC64
713
	/* collect purr register values often, for accurate calculations */
714
	if (firmware_has_feature(FW_FEATURE_SPLPAR)) {
L
Linus Torvalds 已提交
715 716 717
		struct cpu_usage *cu = &__get_cpu_var(cpu_usage_array);
		cu->current_tb = mfspr(SPRN_PURR);
	}
718
#endif
L
Linus Torvalds 已提交
719 720 721 722

	irq_exit();
}

723 724
void wakeup_decrementer(void)
{
725
	unsigned long ticks;
726 727

	/*
728 729
	 * The timebase gets saved on sleep and restored on wakeup,
	 * so all we need to do is to reset the decrementer.
730
	 */
731 732 733 734 735 736
	ticks = tb_ticks_since(__get_cpu_var(last_jiffy));
	if (ticks < tb_ticks_per_jiffy)
		ticks = tb_ticks_per_jiffy - ticks;
	else
		ticks = 1;
	set_dec(ticks);
737 738
}

739
#ifdef CONFIG_SMP
740 741 742
void __init smp_space_timers(unsigned int max_cpus)
{
	int i;
743
	unsigned long half = tb_ticks_per_jiffy / 2;
744 745 746
	unsigned long offset = tb_ticks_per_jiffy / max_cpus;
	unsigned long previous_tb = per_cpu(last_jiffy, boot_cpuid);

747 748
	/* make sure tb > per_cpu(last_jiffy, cpu) for all cpus always */
	previous_tb -= tb_ticks_per_jiffy;
749 750 751 752 753
	/*
	 * The stolen time calculation for POWER5 shared-processor LPAR
	 * systems works better if the two threads' timebase interrupts
	 * are staggered by half a jiffy with respect to each other.
	 */
754
	for_each_possible_cpu(i) {
755 756 757 758 759 760 761 762 763
		if (i == boot_cpuid)
			continue;
		if (i == (boot_cpuid ^ 1))
			per_cpu(last_jiffy, i) =
				per_cpu(last_jiffy, boot_cpuid) - half;
		else if (i & 1)
			per_cpu(last_jiffy, i) =
				per_cpu(last_jiffy, i ^ 1) + half;
		else {
764 765 766 767 768 769 770
			previous_tb += offset;
			per_cpu(last_jiffy, i) = previous_tb;
		}
	}
}
#endif

L
Linus Torvalds 已提交
771 772 773 774 775 776 777 778 779
/*
 * Scheduler clock - returns current time in nanosec units.
 *
 * Note: mulhdu(a, b) (multiply high double unsigned) returns
 * the high 64 bits of a * b, i.e. (a * b) >> 64, where a and b
 * are 64-bit unsigned numbers.
 */
unsigned long long sched_clock(void)
{
780 781
	if (__USE_RTC())
		return get_rtc();
L
Linus Torvalds 已提交
782 783 784 785 786 787 788 789
	return mulhdu(get_tb(), tb_to_ns_scale) << tb_to_ns_shift;
}

int do_settimeofday(struct timespec *tv)
{
	time_t wtm_sec, new_sec = tv->tv_sec;
	long wtm_nsec, new_nsec = tv->tv_nsec;
	unsigned long flags;
790 791
	u64 new_xsec;
	unsigned long tb_delta;
L
Linus Torvalds 已提交
792 793 794 795 796

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

	write_seqlock_irqsave(&xtime_lock, flags);
797 798 799 800 801

	/*
	 * Updating the RTC is not the job of this code. If the time is
	 * stepped under NTP, the RTC will be updated after STA_UNSYNC
	 * is cleared.  Tools like clock/hwclock either copy the RTC
L
Linus Torvalds 已提交
802 803 804 805 806
	 * to the system time, in which case there is no point in writing
	 * to the RTC again, or write to the RTC but then they don't call
	 * settimeofday to perform this operation.
	 */
#ifdef CONFIG_PPC_ISERIES
807
	if (first_settimeofday) {
L
Linus Torvalds 已提交
808 809 810 811
		iSeries_tb_recal();
		first_settimeofday = 0;
	}
#endif
812

813 814 815 816
	/* Make userspace gettimeofday spin until we're done. */
	++vdso_data->tb_update_count;
	smp_mb();

817 818 819 820 821 822 823 824 825
	/*
	 * Subtract off the number of nanoseconds since the
	 * beginning of the last tick.
	 * Note that since we don't increment jiffies_64 anywhere other
	 * than in do_timer (since we don't have a lost tick problem),
	 * wall_jiffies will always be the same as jiffies,
	 * and therefore the (jiffies - wall_jiffies) computation
	 * has been removed.
	 */
L
Linus Torvalds 已提交
826
	tb_delta = tb_ticks_since(tb_last_stamp);
827 828
	tb_delta = mulhdu(tb_delta, do_gtod.varp->tb_to_xs); /* in xsec */
	new_nsec -= SCALE_XSEC(tb_delta, 1000000000);
L
Linus Torvalds 已提交
829 830 831 832 833 834 835 836 837 838 839 840

	wtm_sec  = wall_to_monotonic.tv_sec + (xtime.tv_sec - new_sec);
	wtm_nsec = wall_to_monotonic.tv_nsec + (xtime.tv_nsec - new_nsec);

 	set_normalized_timespec(&xtime, new_sec, new_nsec);
	set_normalized_timespec(&wall_to_monotonic, wtm_sec, wtm_nsec);

	/* In case of a large backwards jump in time with NTP, we want the 
	 * clock to be updated as soon as the PLL is again in lock.
	 */
	last_rtc_update = new_sec - 658;

J
john stultz 已提交
841
	ntp_clear();
L
Linus Torvalds 已提交
842

843 844 845
	new_xsec = xtime.tv_nsec;
	if (new_xsec != 0) {
		new_xsec *= XSEC_PER_SEC;
846 847
		do_div(new_xsec, NSEC_PER_SEC);
	}
848
	new_xsec += (u64)xtime.tv_sec * XSEC_PER_SEC;
849
	update_gtod(tb_last_jiffy, new_xsec, do_gtod.varp->tb_to_xs);
L
Linus Torvalds 已提交
850

851 852
	vdso_data->tz_minuteswest = sys_tz.tz_minuteswest;
	vdso_data->tz_dsttime = sys_tz.tz_dsttime;
L
Linus Torvalds 已提交
853 854 855 856 857 858 859 860

	write_sequnlock_irqrestore(&xtime_lock, flags);
	clock_was_set();
	return 0;
}

EXPORT_SYMBOL(do_settimeofday);

861 862 863 864 865 866 867 868 869 870 871 872 873 874
void __init generic_calibrate_decr(void)
{
	struct device_node *cpu;
	unsigned int *fp;
	int node_found;

	/*
	 * The cpu node should have a timebase-frequency property
	 * to tell us the rate at which the decrementer counts.
	 */
	cpu = of_find_node_by_type(NULL, "cpu");

	ppc_tb_freq = DEFAULT_TB_FREQ;		/* hardcoded default */
	node_found = 0;
875
	if (cpu) {
876 877
		fp = (unsigned int *)get_property(cpu, "timebase-frequency",
						  NULL);
878
		if (fp) {
879 880 881 882 883 884 885 886 887 888
			node_found = 1;
			ppc_tb_freq = *fp;
		}
	}
	if (!node_found)
		printk(KERN_ERR "WARNING: Estimating decrementer frequency "
				"(not found)\n");

	ppc_proc_freq = DEFAULT_PROC_FREQ;
	node_found = 0;
889
	if (cpu) {
890 891
		fp = (unsigned int *)get_property(cpu, "clock-frequency",
						  NULL);
892
		if (fp) {
893 894 895 896
			node_found = 1;
			ppc_proc_freq = *fp;
		}
	}
897 898 899 900 901 902 903 904 905 906 907
#ifdef CONFIG_BOOKE
	/* Set the time base to zero */
	mtspr(SPRN_TBWL, 0);
	mtspr(SPRN_TBWU, 0);

	/* Clear any pending timer interrupts */
	mtspr(SPRN_TSR, TSR_ENW | TSR_WIS | TSR_DIS | TSR_FIS);

	/* Enable decrementer interrupt */
	mtspr(SPRN_TCR, TCR_DIE);
#endif
908 909 910 911 912 913 914
	if (!node_found)
		printk(KERN_ERR "WARNING: Estimating processor frequency "
				"(not found)\n");

	of_node_put(cpu);
}

915 916 917 918 919 920 921 922 923 924 925 926 927 928
unsigned long get_boot_time(void)
{
	struct rtc_time tm;

	if (ppc_md.get_boot_time)
		return ppc_md.get_boot_time();
	if (!ppc_md.get_rtc_time)
		return 0;
	ppc_md.get_rtc_time(&tm);
	return mktime(tm.tm_year+1900, tm.tm_mon+1, tm.tm_mday,
		      tm.tm_hour, tm.tm_min, tm.tm_sec);
}

/* This function is only called on the boot processor */
L
Linus Torvalds 已提交
929 930 931
void __init time_init(void)
{
	unsigned long flags;
932
	unsigned long tm = 0;
L
Linus Torvalds 已提交
933
	struct div_result res;
934
	u64 scale, x;
935 936 937 938
	unsigned shift;

        if (ppc_md.time_init != NULL)
                timezone_offset = ppc_md.time_init();
L
Linus Torvalds 已提交
939

940 941 942 943 944 945 946 947
	if (__USE_RTC()) {
		/* 601 processor: dec counts down by 128 every 128ns */
		ppc_tb_freq = 1000000000;
		tb_last_stamp = get_rtcl();
		tb_last_jiffy = tb_last_stamp;
	} else {
		/* Normal PowerPC with timebase register */
		ppc_md.calibrate_decr();
948
		printk(KERN_DEBUG "time_init: decrementer frequency = %lu.%.6lu MHz\n",
949
		       ppc_tb_freq / 1000000, ppc_tb_freq % 1000000);
950
		printk(KERN_DEBUG "time_init: processor frequency   = %lu.%.6lu MHz\n",
951 952 953
		       ppc_proc_freq / 1000000, ppc_proc_freq % 1000000);
		tb_last_stamp = tb_last_jiffy = get_tb();
	}
954 955

	tb_ticks_per_jiffy = ppc_tb_freq / HZ;
956
	tb_ticks_per_sec = ppc_tb_freq;
957 958
	tb_ticks_per_usec = ppc_tb_freq / 1000000;
	tb_to_us = mulhwu_scale_factor(ppc_tb_freq, 1000000);
959
	calc_cputime_factors();
960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977

	/*
	 * Calculate the length of each tick in ns.  It will not be
	 * exactly 1e9/HZ unless ppc_tb_freq is divisible by HZ.
	 * We compute 1e9 * tb_ticks_per_jiffy / ppc_tb_freq,
	 * rounded up.
	 */
	x = (u64) NSEC_PER_SEC * tb_ticks_per_jiffy + ppc_tb_freq - 1;
	do_div(x, ppc_tb_freq);
	tick_nsec = x;
	last_tick_len = x << TICKLEN_SCALE;

	/*
	 * Compute ticklen_to_xs, which is a factor which gets multiplied
	 * by (last_tick_len << TICKLEN_SHIFT) to get a tb_to_xs value.
	 * It is computed as:
	 * ticklen_to_xs = 2^N / (tb_ticks_per_jiffy * 1e9)
	 * where N = 64 + 20 - TICKLEN_SCALE - TICKLEN_SHIFT
978 979 980 981 982 983 984
	 * which turns out to be N = 51 - SHIFT_HZ.
	 * This gives the result as a 0.64 fixed-point fraction.
	 * That value is reduced by an offset amounting to 1 xsec per
	 * 2^31 timebase ticks to avoid problems with time going backwards
	 * by 1 xsec when we do timer_recalc_offset due to losing the
	 * fractional xsec.  That offset is equal to ppc_tb_freq/2^51
	 * since there are 2^20 xsec in a second.
985
	 */
986 987
	div128_by_32((1ULL << 51) - ppc_tb_freq, 0,
		     tb_ticks_per_jiffy << SHIFT_HZ, &res);
988 989 990 991 992
	div128_by_32(res.result_high, res.result_low, NSEC_PER_SEC, &res);
	ticklen_to_xs = res.result_low;

	/* Compute tb_to_xs from tick_nsec */
	tb_to_xs = mulhdu(last_tick_len << TICKLEN_SHIFT, ticklen_to_xs);
993

L
Linus Torvalds 已提交
994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015
	/*
	 * Compute scale factor for sched_clock.
	 * The calibrate_decr() function has set tb_ticks_per_sec,
	 * which is the timebase frequency.
	 * We compute 1e9 * 2^64 / tb_ticks_per_sec and interpret
	 * the 128-bit result as a 64.64 fixed-point number.
	 * We then shift that number right until it is less than 1.0,
	 * giving us the scale factor and shift count to use in
	 * sched_clock().
	 */
	div128_by_32(1000000000, 0, tb_ticks_per_sec, &res);
	scale = res.result_low;
	for (shift = 0; res.result_high != 0; ++shift) {
		scale = (scale >> 1) | (res.result_high << 63);
		res.result_high >>= 1;
	}
	tb_to_ns_scale = scale;
	tb_to_ns_shift = shift;

#ifdef CONFIG_PPC_ISERIES
	if (!piranha_simulator)
#endif
1016
		tm = get_boot_time();
L
Linus Torvalds 已提交
1017 1018

	write_seqlock_irqsave(&xtime_lock, flags);
1019 1020 1021 1022 1023 1024 1025 1026

	/* If platform provided a timezone (pmac), we correct the time */
        if (timezone_offset) {
		sys_tz.tz_minuteswest = -timezone_offset / 60;
		sys_tz.tz_dsttime = 0;
		tm -= timezone_offset;
        }

1027 1028
	xtime.tv_sec = tm;
	xtime.tv_nsec = 0;
L
Linus Torvalds 已提交
1029 1030
	do_gtod.varp = &do_gtod.vars[0];
	do_gtod.var_idx = 0;
1031
	do_gtod.varp->tb_orig_stamp = tb_last_jiffy;
1032 1033
	__get_cpu_var(last_jiffy) = tb_last_stamp;
	do_gtod.varp->stamp_xsec = (u64) xtime.tv_sec * XSEC_PER_SEC;
L
Linus Torvalds 已提交
1034 1035 1036
	do_gtod.tb_ticks_per_sec = tb_ticks_per_sec;
	do_gtod.varp->tb_to_xs = tb_to_xs;
	do_gtod.tb_to_us = tb_to_us;
1037 1038 1039 1040

	vdso_data->tb_orig_stamp = tb_last_jiffy;
	vdso_data->tb_update_count = 0;
	vdso_data->tb_ticks_per_sec = tb_ticks_per_sec;
1041
	vdso_data->stamp_xsec = (u64) xtime.tv_sec * XSEC_PER_SEC;
1042
	vdso_data->tb_to_xs = tb_to_xs;
L
Linus Torvalds 已提交
1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059

	time_freq = 0;

	last_rtc_update = xtime.tv_sec;
	set_normalized_timespec(&wall_to_monotonic,
	                        -xtime.tv_sec, -xtime.tv_nsec);
	write_sequnlock_irqrestore(&xtime_lock, flags);

	/* Not exact, but the timer interrupt takes care of this */
	set_dec(tb_ticks_per_jiffy);
}


#define FEBRUARY	2
#define	STARTOFTIME	1970
#define SECDAY		86400L
#define SECYR		(SECDAY * 365)
1060 1061
#define	leapyear(year)		((year) % 4 == 0 && \
				 ((year) % 100 != 0 || (year) % 400 == 0))
L
Linus Torvalds 已提交
1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078
#define	days_in_year(a) 	(leapyear(a) ? 366 : 365)
#define	days_in_month(a) 	(month_days[(a) - 1])

static int month_days[12] = {
	31, 28, 31, 30, 31, 30, 31, 31, 30, 31, 30, 31
};

/*
 * This only works for the Gregorian calendar - i.e. after 1752 (in the UK)
 */
void GregorianDay(struct rtc_time * tm)
{
	int leapsToDate;
	int lastYear;
	int day;
	int MonthOffset[] = { 0, 31, 59, 90, 120, 151, 181, 212, 243, 273, 304, 334 };

1079
	lastYear = tm->tm_year - 1;
L
Linus Torvalds 已提交
1080 1081 1082 1083

	/*
	 * Number of leap corrections to apply up to end of last year
	 */
1084
	leapsToDate = lastYear / 4 - lastYear / 100 + lastYear / 400;
L
Linus Torvalds 已提交
1085 1086 1087 1088 1089

	/*
	 * This year is a leap year if it is divisible by 4 except when it is
	 * divisible by 100 unless it is divisible by 400
	 *
1090
	 * e.g. 1904 was a leap year, 1900 was not, 1996 is, and 2000 was
L
Linus Torvalds 已提交
1091
	 */
1092
	day = tm->tm_mon > 2 && leapyear(tm->tm_year);
L
Linus Torvalds 已提交
1093 1094 1095 1096

	day += lastYear*365 + leapsToDate + MonthOffset[tm->tm_mon-1] +
		   tm->tm_mday;

1097
	tm->tm_wday = day % 7;
L
Linus Torvalds 已提交
1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142
}

void to_tm(int tim, struct rtc_time * tm)
{
	register int    i;
	register long   hms, day;

	day = tim / SECDAY;
	hms = tim % SECDAY;

	/* Hours, minutes, seconds are easy */
	tm->tm_hour = hms / 3600;
	tm->tm_min = (hms % 3600) / 60;
	tm->tm_sec = (hms % 3600) % 60;

	/* Number of years in days */
	for (i = STARTOFTIME; day >= days_in_year(i); i++)
		day -= days_in_year(i);
	tm->tm_year = i;

	/* Number of months in days left */
	if (leapyear(tm->tm_year))
		days_in_month(FEBRUARY) = 29;
	for (i = 1; day >= days_in_month(i); i++)
		day -= days_in_month(i);
	days_in_month(FEBRUARY) = 28;
	tm->tm_mon = i;

	/* Days are what is left over (+1) from all that. */
	tm->tm_mday = day + 1;

	/*
	 * Determine the day of week
	 */
	GregorianDay(tm);
}

/* Auxiliary function to compute scaling factors */
/* Actually the choice of a timebase running at 1/4 the of the bus
 * frequency giving resolution of a few tens of nanoseconds is quite nice.
 * It makes this computation very precise (27-28 bits typically) which
 * is optimistic considering the stability of most processor clock
 * oscillators and the precision with which the timebase frequency
 * is measured but does not harm.
 */
1143 1144
unsigned mulhwu_scale_factor(unsigned inscale, unsigned outscale)
{
L
Linus Torvalds 已提交
1145 1146 1147 1148 1149 1150
        unsigned mlt=0, tmp, err;
        /* No concern for performance, it's done once: use a stupid
         * but safe and compact method to find the multiplier.
         */
  
        for (tmp = 1U<<31; tmp != 0; tmp >>= 1) {
1151 1152
                if (mulhwu(inscale, mlt|tmp) < outscale)
			mlt |= tmp;
L
Linus Torvalds 已提交
1153 1154 1155 1156 1157 1158 1159 1160 1161
        }
  
        /* We might still be off by 1 for the best approximation.
         * A side effect of this is that if outscale is too large
         * the returned value will be zero.
         * Many corner cases have been checked and seem to work,
         * some might have been forgotten in the test however.
         */
  
1162 1163 1164
        err = inscale * (mlt+1);
        if (err <= inscale/2)
		mlt++;
L
Linus Torvalds 已提交
1165
        return mlt;
1166
}
L
Linus Torvalds 已提交
1167 1168 1169 1170 1171

/*
 * Divide a 128-bit dividend by a 32-bit divisor, leaving a 128 bit
 * result.
 */
1172 1173
void div128_by_32(u64 dividend_high, u64 dividend_low,
		  unsigned divisor, struct div_result *dr)
L
Linus Torvalds 已提交
1174
{
1175 1176 1177
	unsigned long a, b, c, d;
	unsigned long w, x, y, z;
	u64 ra, rb, rc;
L
Linus Torvalds 已提交
1178 1179 1180 1181 1182 1183

	a = dividend_high >> 32;
	b = dividend_high & 0xffffffff;
	c = dividend_low >> 32;
	d = dividend_low & 0xffffffff;

1184 1185 1186 1187 1188
	w = a / divisor;
	ra = ((u64)(a - (w * divisor)) << 32) + b;

	rb = ((u64) do_div(ra, divisor) << 32) + c;
	x = ra;
L
Linus Torvalds 已提交
1189

1190 1191 1192 1193 1194
	rc = ((u64) do_div(rb, divisor) << 32) + d;
	y = rb;

	do_div(rc, divisor);
	z = rc;
L
Linus Torvalds 已提交
1195

1196 1197
	dr->result_high = ((u64)w << 32) + x;
	dr->result_low  = ((u64)y << 32) + z;
L
Linus Torvalds 已提交
1198 1199

}