| 1 | |
| 2 | * linux/kernel/time/clocksource.c
|
| 3 | *
|
| 4 | * This file contains the functions which manage clocksource drivers.
|
| 5 | *
|
| 6 | * Copyright (C) 2004, 2005 IBM, John Stultz ([email protected])
|
| 7 | *
|
| 8 | * This program is free software; you can redistribute it and/or modify
|
| 9 | * it under the terms of the GNU General Public License as published by
|
| 10 | * the Free Software Foundation; either version 2 of the License, or
|
| 11 | * (at your option) any later version.
|
| 12 | *
|
| 13 | * This program is distributed in the hope that it will be useful,
|
| 14 | * but WITHOUT ANY WARRANTY; without even the implied warranty of
|
| 15 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
| 16 | * GNU General Public License for more details.
|
| 17 | *
|
| 18 | * You should have received a copy of the GNU General Public License
|
| 19 | * along with this program; if not, write to the Free Software
|
| 20 | * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
|
| 21 | *
|
| 22 | * TODO WishList:
|
| 23 | * o Allow clocksource drivers to be unregistered
|
| 24 |
|
| 25 |
|
| 26 | #include <linux/clocksource.h>
|
| 27 | #include <linux/sysdev.h>
|
| 28 | #include <linux/init.h>
|
| 29 | #include <linux/module.h>
|
| 30 | #include <linux/sched.h>
|
| 31 | #include <linux/tick.h>
|
| 32 | #include <linux/kthread.h>
|
| 33 |
|
| 34 | void timecounter_init(struct timecounter *tc,
|
| 35 | const struct cyclecounter *cc,
|
| 36 | u64 start_tstamp)
|
| 37 | {
|
| 38 | tc->cc = cc;
|
| 39 | tc->cycle_last = cc->read(cc);
|
| 40 | tc->nsec = start_tstamp;
|
| 41 | }
|
| 42 | EXPORT_SYMBOL_GPL(timecounter_init);
|
| 43 |
|
| 44 | |
| 45 | * timecounter_read_delta - get nanoseconds since last call of this function
|
| 46 | * @tc: Pointer to time counter
|
| 47 | *
|
| 48 | * When the underlying cycle counter runs over, this will be handled
|
| 49 | * correctly as long as it does not run over more than once between
|
| 50 | * calls.
|
| 51 | *
|
| 52 | * The first call to this function for a new time counter initializes
|
| 53 | * the time tracking and returns an undefined result.
|
| 54 |
|
| 55 | static u64 timecounter_read_delta(struct timecounter *tc)
|
| 56 | {
|
| 57 | cycle_t cycle_now, cycle_delta;
|
| 58 | u64 ns_offset;
|
| 59 |
|
| 60 |
|
| 61 | cycle_now = tc->cc->read(tc->cc);
|
| 62 |
|
| 63 |
|
| 64 | cycle_delta = (cycle_now - tc->cycle_last) & tc->cc->mask;
|
| 65 |
|
| 66 |
|
| 67 | ns_offset = cyclecounter_cyc2ns(tc->cc, cycle_delta);
|
| 68 |
|
| 69 |
|
| 70 | tc->cycle_last = cycle_now;
|
| 71 |
|
| 72 | return ns_offset;
|
| 73 | }
|
| 74 |
|
| 75 | u64 timecounter_read(struct timecounter *tc)
|
| 76 | {
|
| 77 | u64 nsec;
|
| 78 |
|
| 79 |
|
| 80 | nsec = timecounter_read_delta(tc);
|
| 81 | nsec += tc->nsec;
|
| 82 | tc->nsec = nsec;
|
| 83 |
|
| 84 | return nsec;
|
| 85 | }
|
| 86 | EXPORT_SYMBOL_GPL(timecounter_read);
|
| 87 |
|
| 88 | u64 timecounter_cyc2time(struct timecounter *tc,
|
| 89 | cycle_t cycle_tstamp)
|
| 90 | {
|
| 91 | u64 cycle_delta = (cycle_tstamp - tc->cycle_last) & tc->cc->mask;
|
| 92 | u64 nsec;
|
| 93 |
|
| 94 | |
| 95 | * Instead of always treating cycle_tstamp as more recent
|
| 96 | * than tc->cycle_last, detect when it is too far in the
|
| 97 | * future and treat it as old time stamp instead.
|
| 98 |
|
| 99 | if (cycle_delta > tc->cc->mask / 2) {
|
| 100 | cycle_delta = (tc->cycle_last - cycle_tstamp) & tc->cc->mask;
|
| 101 | nsec = tc->nsec - cyclecounter_cyc2ns(tc->cc, cycle_delta);
|
| 102 | } else {
|
| 103 | nsec = cyclecounter_cyc2ns(tc->cc, cycle_delta) + tc->nsec;
|
| 104 | }
|
| 105 |
|
| 106 | return nsec;
|
| 107 | }
|
| 108 | EXPORT_SYMBOL_GPL(timecounter_cyc2time);
|
| 109 |
|
| 110 | |
| 111 | * clocks_calc_mult_shift - calculate mult/shift factors for scaled math of clocks
|
| 112 | * @mult: pointer to mult variable
|
| 113 | * @shift: pointer to shift variable
|
| 114 | * @from: frequency to convert from
|
| 115 | * @to: frequency to convert to
|
| 116 | * @maxsec: guaranteed runtime conversion range in seconds
|
| 117 | *
|
| 118 | * The function evaluates the shift/mult pair for the scaled math
|
| 119 | * operations of clocksources and clockevents.
|
| 120 | *
|
| 121 | * @to and @from are frequency values in HZ. For clock sources @to is
|
| 122 | * NSEC_PER_SEC == 1GHz and @from is the counter frequency. For clock
|
| 123 | * event @to is the counter frequency and @from is NSEC_PER_SEC.
|
| 124 | *
|
| 125 | * The @maxsec conversion range argument controls the time frame in
|
| 126 | * seconds which must be covered by the runtime conversion with the
|
| 127 | * calculated mult and shift factors. This guarantees that no 64bit
|
| 128 | * overflow happens when the input value of the conversion is
|
| 129 | * multiplied with the calculated mult factor. Larger ranges may
|
| 130 | * reduce the conversion accuracy by chosing smaller mult and shift
|
| 131 | * factors.
|
| 132 |
|
| 133 | void
|
| 134 | clocks_calc_mult_shift(u32 *mult, u32 *shift, u32 from, u32 to, u32 maxsec)
|
| 135 | {
|
| 136 | u64 tmp;
|
| 137 | u32 sft, sftacc= 32;
|
| 138 |
|
| 139 | |
| 140 | * Calculate the shift factor which is limiting the conversion
|
| 141 | * range:
|
| 142 |
|
| 143 | tmp = ((u64)maxsec * from) >> 32;
|
| 144 | while (tmp) {
|
| 145 | tmp >>=1;
|
| 146 | sftacc--;
|
| 147 | }
|
| 148 |
|
| 149 | |
| 150 | * Find the conversion shift/mult pair which has the best
|
| 151 | * accuracy and fits the maxsec conversion range:
|
| 152 |
|
| 153 | for (sft = 32; sft > 0; sft--) {
|
| 154 | tmp = (u64) to << sft;
|
| 155 | tmp += from / 2;
|
| 156 | do_div(tmp, from);
|
| 157 | if ((tmp >> sftacc) == 0)
|
| 158 | break;
|
| 159 | }
|
| 160 | *mult = tmp;
|
| 161 | *shift = sft;
|
| 162 | }
|
| 163 |
|
| 164 | |
| 165 | * curr_clocksource:
|
| 166 | * currently selected clocksource.
|
| 167 | * clocksource_list:
|
| 168 | * linked list with the registered clocksources
|
| 169 | * clocksource_mutex:
|
| 170 | * protects manipulations to curr_clocksource and the clocksource_list
|
| 171 | * override_name:
|
| 172 | * Name of the user-specified clocksource.
|
| 173 |
|
| 174 | static struct clocksource *curr_clocksource;
|
| 175 | static LIST_HEAD(clocksource_list);
|
| 176 | static DEFINE_MUTEX(clocksource_mutex);
|
| 177 | static char override_name[32];
|
| 178 | static int finished_booting;
|
| 179 |
|
| 180 | #ifdef CONFIG_CLOCKSOURCE_WATCHDOG
|
| 181 | static void clocksource_watchdog_work(struct work_struct *work);
|
| 182 |
|
| 183 | static LIST_HEAD(watchdog_list);
|
| 184 | static struct clocksource *watchdog;
|
| 185 | static struct timer_list watchdog_timer;
|
| 186 | static DECLARE_WORK(watchdog_work, clocksource_watchdog_work);
|
| 187 | static DEFINE_SPINLOCK(watchdog_lock);
|
| 188 | static cycle_t watchdog_last;
|
| 189 | static int watchdog_running;
|
| 190 |
|
| 191 | static int clocksource_watchdog_kthread(void *data);
|
| 192 | static void __clocksource_change_rating(struct clocksource *cs, int rating);
|
| 193 |
|
| 194 | |
| 195 | * Interval: 0.5sec Threshold: 0.0625s
|
| 196 |
|
| 197 | #define WATCHDOG_INTERVAL (HZ >> 1)
|
| 198 | #define WATCHDOG_THRESHOLD (NSEC_PER_SEC >> 4)
|
| 199 |
|
| 200 | static void clocksource_watchdog_work(struct work_struct *work)
|
| 201 | {
|
| 202 | |
| 203 | * If kthread_run fails the next watchdog scan over the
|
| 204 | * watchdog_list will find the unstable clock again.
|
| 205 |
|
| 206 | kthread_run(clocksource_watchdog_kthread, NULL, "kwatchdog");
|
| 207 | }
|
| 208 |
|
| 209 | static void __clocksource_unstable(struct clocksource *cs)
|
| 210 | {
|
| 211 | cs->flags &= ~(CLOCK_SOURCE_VALID_FOR_HRES | CLOCK_SOURCE_WATCHDOG);
|
| 212 | cs->flags |= CLOCK_SOURCE_UNSTABLE;
|
| 213 | if (finished_booting)
|
| 214 | schedule_work(&watchdog_work);
|
| 215 | }
|
| 216 |
|
| 217 | static void clocksource_unstable(struct clocksource *cs, int64_t delta)
|
| 218 | {
|
| 219 | printk(KERN_WARNING "Clocksource %s unstable (delta = %Ld ns)\n",
|
| 220 | cs->name, delta);
|
| 221 | __clocksource_unstable(cs);
|
| 222 | }
|
| 223 |
|
| 224 | |
| 225 | * clocksource_mark_unstable - mark clocksource unstable via watchdog
|
| 226 | * @cs: clocksource to be marked unstable
|
| 227 | *
|
| 228 | * This function is called instead of clocksource_change_rating from
|
| 229 | * cpu hotplug code to avoid a deadlock between the clocksource mutex
|
| 230 | * and the cpu hotplug mutex. It defers the update of the clocksource
|
| 231 | * to the watchdog thread.
|
| 232 |
|
| 233 | void clocksource_mark_unstable(struct clocksource *cs)
|
| 234 | {
|
| 235 | unsigned long flags;
|
| 236 |
|
| 237 | spin_lock_irqsave(&watchdog_lock, flags);
|
| 238 | if (!(cs->flags & CLOCK_SOURCE_UNSTABLE)) {
|
| 239 | if (list_empty(&cs->wd_list))
|
| 240 | list_add(&cs->wd_list, &watchdog_list);
|
| 241 | __clocksource_unstable(cs);
|
| 242 | }
|
| 243 | spin_unlock_irqrestore(&watchdog_lock, flags);
|
| 244 | }
|
| 245 |
|
| 246 | static void clocksource_watchdog(unsigned long data)
|
| 247 | {
|
| 248 | struct clocksource *cs;
|
| 249 | cycle_t csnow, wdnow;
|
| 250 | int64_t wd_nsec, cs_nsec;
|
| 251 | int next_cpu;
|
| 252 |
|
| 253 | spin_lock(&watchdog_lock);
|
| 254 | if (!watchdog_running)
|
| 255 | goto out;
|
| 256 |
|
| 257 | wdnow = watchdog->read(watchdog);
|
| 258 | wd_nsec = clocksource_cyc2ns((wdnow - watchdog_last) & watchdog->mask,
|
| 259 | watchdog->mult, watchdog->shift);
|
| 260 | watchdog_last = wdnow;
|
| 261 |
|
| 262 | list_for_each_entry(cs, &watchdog_list, wd_list) {
|
| 263 |
|
| 264 |
|
| 265 | if (cs->flags & CLOCK_SOURCE_UNSTABLE) {
|
| 266 | if (finished_booting)
|
| 267 | schedule_work(&watchdog_work);
|
| 268 | continue;
|
| 269 | }
|
| 270 |
|
| 271 | csnow = cs->read(cs);
|
| 272 |
|
| 273 |
|
| 274 | if (!(cs->flags & CLOCK_SOURCE_WATCHDOG)) {
|
| 275 | cs->flags |= CLOCK_SOURCE_WATCHDOG;
|
| 276 | cs->wd_last = csnow;
|
| 277 | continue;
|
| 278 | }
|
| 279 |
|
| 280 |
|
| 281 | cs_nsec = clocksource_cyc2ns((csnow - cs->wd_last) &
|
| 282 | cs->mask, cs->mult, cs->shift);
|
| 283 | cs->wd_last = csnow;
|
| 284 | if (abs(cs_nsec - wd_nsec) > WATCHDOG_THRESHOLD) {
|
| 285 | clocksource_unstable(cs, cs_nsec - wd_nsec);
|
| 286 | continue;
|
| 287 | }
|
| 288 |
|
| 289 | if (!(cs->flags & CLOCK_SOURCE_VALID_FOR_HRES) &&
|
| 290 | (cs->flags & CLOCK_SOURCE_IS_CONTINUOUS) &&
|
| 291 | (watchdog->flags & CLOCK_SOURCE_IS_CONTINUOUS)) {
|
| 292 | cs->flags |= CLOCK_SOURCE_VALID_FOR_HRES;
|
| 293 | |
| 294 | * We just marked the clocksource as highres-capable,
|
| 295 | * notify the rest of the system as well so that we
|
| 296 | * transition into high-res mode:
|
| 297 |
|
| 298 | tick_clock_notify();
|
| 299 | }
|
| 300 | }
|
| 301 |
|
| 302 | |
| 303 | * Cycle through CPUs to check if the CPUs stay synchronized
|
| 304 | * to each other.
|
| 305 |
|
| 306 | next_cpu = cpumask_next(raw_smp_processor_id(), cpu_online_mask);
|
| 307 | if (next_cpu >= nr_cpu_ids)
|
| 308 | next_cpu = cpumask_first(cpu_online_mask);
|
| 309 | watchdog_timer.expires += WATCHDOG_INTERVAL;
|
| 310 | add_timer_on(&watchdog_timer, next_cpu);
|
| 311 | out:
|
| 312 | spin_unlock(&watchdog_lock);
|
| 313 | }
|
| 314 |
|
| 315 | static inline void clocksource_start_watchdog(void)
|
| 316 | {
|
| 317 | if (watchdog_running || !watchdog || list_empty(&watchdog_list))
|
| 318 | return;
|
| 319 | init_timer(&watchdog_timer);
|
| 320 | watchdog_timer.function = clocksource_watchdog;
|
| 321 | watchdog_last = watchdog->read(watchdog);
|
| 322 | watchdog_timer.expires = jiffies + WATCHDOG_INTERVAL;
|
| 323 | add_timer_on(&watchdog_timer, cpumask_first(cpu_online_mask));
|
| 324 | watchdog_running = 1;
|
| 325 | }
|
| 326 |
|
| 327 | static inline void clocksource_stop_watchdog(void)
|
| 328 | {
|
| 329 | if (!watchdog_running || (watchdog && !list_empty(&watchdog_list)))
|
| 330 | return;
|
| 331 | del_timer(&watchdog_timer);
|
| 332 | watchdog_running = 0;
|
| 333 | }
|
| 334 |
|
| 335 | static inline void clocksource_reset_watchdog(void)
|
| 336 | {
|
| 337 | struct clocksource *cs;
|
| 338 |
|
| 339 | list_for_each_entry(cs, &watchdog_list, wd_list)
|
| 340 | cs->flags &= ~CLOCK_SOURCE_WATCHDOG;
|
| 341 | }
|
| 342 |
|
| 343 | static void clocksource_resume_watchdog(void)
|
| 344 | {
|
| 345 | unsigned long flags;
|
| 346 |
|
| 347 | |
| 348 | * We use trylock here to avoid a potential dead lock when
|
| 349 | * kgdb calls this code after the kernel has been stopped with
|
| 350 | * watchdog_lock held. When watchdog_lock is held we just
|
| 351 | * return and accept, that the watchdog might trigger and mark
|
| 352 | * the monitored clock source (usually TSC) unstable.
|
| 353 | *
|
| 354 | * This does not affect the other caller clocksource_resume()
|
| 355 | * because at this point the kernel is UP, interrupts are
|
| 356 | * disabled and nothing can hold watchdog_lock.
|
| 357 |
|
| 358 | if (!spin_trylock_irqsave(&watchdog_lock, flags))
|
| 359 | return;
|
| 360 | clocksource_reset_watchdog();
|
| 361 | spin_unlock_irqrestore(&watchdog_lock, flags);
|
| 362 | }
|
| 363 |
|
| 364 | static void clocksource_enqueue_watchdog(struct clocksource *cs)
|
| 365 | {
|
| 366 | unsigned long flags;
|
| 367 |
|
| 368 | spin_lock_irqsave(&watchdog_lock, flags);
|
| 369 | if (cs->flags & CLOCK_SOURCE_MUST_VERIFY) {
|
| 370 |
|
| 371 | list_add(&cs->wd_list, &watchdog_list);
|
| 372 | cs->flags &= ~CLOCK_SOURCE_WATCHDOG;
|
| 373 | } else {
|
| 374 |
|
| 375 | if (cs->flags & CLOCK_SOURCE_IS_CONTINUOUS)
|
| 376 | cs->flags |= CLOCK_SOURCE_VALID_FOR_HRES;
|
| 377 |
|
| 378 | if (!watchdog || cs->rating > watchdog->rating) {
|
| 379 | watchdog = cs;
|
| 380 |
|
| 381 | clocksource_reset_watchdog();
|
| 382 | }
|
| 383 | }
|
| 384 |
|
| 385 | clocksource_start_watchdog();
|
| 386 | spin_unlock_irqrestore(&watchdog_lock, flags);
|
| 387 | }
|
| 388 |
|
| 389 | static void clocksource_dequeue_watchdog(struct clocksource *cs)
|
| 390 | {
|
| 391 | struct clocksource *tmp;
|
| 392 | unsigned long flags;
|
| 393 |
|
| 394 | spin_lock_irqsave(&watchdog_lock, flags);
|
| 395 | if (cs->flags & CLOCK_SOURCE_MUST_VERIFY) {
|
| 396 |
|
| 397 | list_del_init(&cs->wd_list);
|
| 398 | } else if (cs == watchdog) {
|
| 399 |
|
| 400 | clocksource_reset_watchdog();
|
| 401 |
|
| 402 | watchdog = NULL;
|
| 403 | list_for_each_entry(tmp, &clocksource_list, list) {
|
| 404 | if (tmp == cs || tmp->flags & CLOCK_SOURCE_MUST_VERIFY)
|
| 405 | continue;
|
| 406 | if (!watchdog || tmp->rating > watchdog->rating)
|
| 407 | watchdog = tmp;
|
| 408 | }
|
| 409 | }
|
| 410 | cs->flags &= ~CLOCK_SOURCE_WATCHDOG;
|
| 411 |
|
| 412 | clocksource_stop_watchdog();
|
| 413 | spin_unlock_irqrestore(&watchdog_lock, flags);
|
| 414 | }
|
| 415 |
|
| 416 | static int clocksource_watchdog_kthread(void *data)
|
| 417 | {
|
| 418 | struct clocksource *cs, *tmp;
|
| 419 | unsigned long flags;
|
| 420 | LIST_HEAD(unstable);
|
| 421 |
|
| 422 | mutex_lock(&clocksource_mutex);
|
| 423 | spin_lock_irqsave(&watchdog_lock, flags);
|
| 424 | list_for_each_entry_safe(cs, tmp, &watchdog_list, wd_list)
|
| 425 | if (cs->flags & CLOCK_SOURCE_UNSTABLE) {
|
| 426 | list_del_init(&cs->wd_list);
|
| 427 | list_add(&cs->wd_list, &unstable);
|
| 428 | }
|
| 429 |
|
| 430 | clocksource_stop_watchdog();
|
| 431 | spin_unlock_irqrestore(&watchdog_lock, flags);
|
| 432 |
|
| 433 |
|
| 434 | list_for_each_entry_safe(cs, tmp, &unstable, wd_list) {
|
| 435 | list_del_init(&cs->wd_list);
|
| 436 | __clocksource_change_rating(cs, 0);
|
| 437 | }
|
| 438 | mutex_unlock(&clocksource_mutex);
|
| 439 | return 0;
|
| 440 | }
|
| 441 |
|
| 442 | #else
|
| 443 |
|
| 444 | static void clocksource_enqueue_watchdog(struct clocksource *cs)
|
| 445 | {
|
| 446 | if (cs->flags & CLOCK_SOURCE_IS_CONTINUOUS)
|
| 447 | cs->flags |= CLOCK_SOURCE_VALID_FOR_HRES;
|
| 448 | }
|
| 449 |
|
| 450 | static inline void clocksource_dequeue_watchdog(struct clocksource *cs) { }
|
| 451 | static inline void clocksource_resume_watchdog(void) { }
|
| 452 | static inline int clocksource_watchdog_kthread(void *data) { return 0; }
|
| 453 |
|
| 454 | #endif
|
| 455 |
|
| 456 | |
| 457 | * clocksource_suspend - suspend the clocksource(s)
|
| 458 |
|
| 459 | void clocksource_suspend(void)
|
| 460 | {
|
| 461 | struct clocksource *cs;
|
| 462 |
|
| 463 | list_for_each_entry_reverse(cs, &clocksource_list, list)
|
| 464 | if (cs->suspend)
|
| 465 | cs->suspend(cs);
|
| 466 | }
|
| 467 |
|
| 468 | |
| 469 | * clocksource_resume - resume the clocksource(s)
|
| 470 |
|
| 471 | void clocksource_resume(void)
|
| 472 | {
|
| 473 | struct clocksource *cs;
|
| 474 |
|
| 475 | list_for_each_entry(cs, &clocksource_list, list)
|
| 476 | if (cs->resume)
|
| 477 | cs->resume(cs);
|
| 478 |
|
| 479 | clocksource_resume_watchdog();
|
| 480 | }
|
| 481 |
|
| 482 | |
| 483 | * clocksource_touch_watchdog - Update watchdog
|
| 484 | *
|
| 485 | * Update the watchdog after exception contexts such as kgdb so as not
|
| 486 | * to incorrectly trip the watchdog. This might fail when the kernel
|
| 487 | * was stopped in code which holds watchdog_lock.
|
| 488 |
|
| 489 | void clocksource_touch_watchdog(void)
|
| 490 | {
|
| 491 | clocksource_resume_watchdog();
|
| 492 | }
|
| 493 |
|
| 494 | |
| 495 | * clocksource_max_deferment - Returns max time the clocksource can be deferred
|
| 496 | * @cs: Pointer to clocksource
|
| 497 | *
|
| 498 |
|
| 499 | static u64 clocksource_max_deferment(struct clocksource *cs)
|
| 500 | {
|
| 501 | u64 max_nsecs, max_cycles;
|
| 502 |
|
| 503 | |
| 504 | * Calculate the maximum number of cycles that we can pass to the
|
| 505 | * cyc2ns function without overflowing a 64-bit signed result. The
|
| 506 | * maximum number of cycles is equal to ULLONG_MAX/cs->mult which
|
| 507 | * is equivalent to the below.
|
| 508 | * max_cycles < (2^63)/cs->mult
|
| 509 | * max_cycles < 2^(log2((2^63)/cs->mult))
|
| 510 | * max_cycles < 2^(log2(2^63) - log2(cs->mult))
|
| 511 | * max_cycles < 2^(63 - log2(cs->mult))
|
| 512 | * max_cycles < 1 << (63 - log2(cs->mult))
|
| 513 | * Please note that we add 1 to the result of the log2 to account for
|
| 514 | * any rounding errors, ensure the above inequality is satisfied and
|
| 515 | * no overflow will occur.
|
| 516 |
|
| 517 | max_cycles = 1ULL << (63 - (ilog2(cs->mult) + 1));
|
| 518 |
|
| 519 | |
| 520 | * The actual maximum number of cycles we can defer the clocksource is
|
| 521 | * determined by the minimum of max_cycles and cs->mask.
|
| 522 |
|
| 523 | max_cycles = min_t(u64, max_cycles, (u64) cs->mask);
|
| 524 | max_nsecs = clocksource_cyc2ns(max_cycles, cs->mult, cs->shift);
|
| 525 |
|
| 526 | |
| 527 | * To ensure that the clocksource does not wrap whilst we are idle,
|
| 528 | * limit the time the clocksource can be deferred by 12.5%. Please
|
| 529 | * note a margin of 12.5% is used because this can be computed with
|
| 530 | * a shift, versus say 10% which would require division.
|
| 531 |
|
| 532 | return max_nsecs - (max_nsecs >> 5);
|
| 533 | }
|
| 534 |
|
| 535 | #ifndef CONFIG_ARCH_USES_GETTIMEOFFSET
|
| 536 |
|
| 537 | |
| 538 | * clocksource_select - Select the best clocksource available
|
| 539 | *
|
| 540 | * Private function. Must hold clocksource_mutex when called.
|
| 541 | *
|
| 542 | * Select the clocksource with the best rating, or the clocksource,
|
| 543 | * which is selected by userspace override.
|
| 544 |
|
| 545 | static void clocksource_select(void)
|
| 546 | {
|
| 547 | struct clocksource *best, *cs;
|
| 548 |
|
| 549 | if (!finished_booting || list_empty(&clocksource_list))
|
| 550 | return;
|
| 551 |
|
| 552 | best = list_first_entry(&clocksource_list, struct clocksource, list);
|
| 553 |
|
| 554 | list_for_each_entry(cs, &clocksource_list, list) {
|
| 555 | if (strcmp(cs->name, override_name) != 0)
|
| 556 | continue;
|
| 557 | |
| 558 | * Check to make sure we don't switch to a non-highres
|
| 559 | * capable clocksource if the tick code is in oneshot
|
| 560 | * mode (highres or nohz)
|
| 561 |
|
| 562 | if (!(cs->flags & CLOCK_SOURCE_VALID_FOR_HRES) &&
|
| 563 | tick_oneshot_mode_active()) {
|
| 564 |
|
| 565 | printk(KERN_WARNING "Override clocksource %s is not "
|
| 566 | "HRT compatible. Cannot switch while in "
|
| 567 | "HRT/NOHZ mode\n", cs->name);
|
| 568 | override_name[0] = 0;
|
| 569 | } else
|
| 570 |
|
| 571 | best = cs;
|
| 572 | break;
|
| 573 | }
|
| 574 | if (curr_clocksource != best) {
|
| 575 | printk(KERN_INFO "Switching to clocksource %s\n", best->name);
|
| 576 | curr_clocksource = best;
|
| 577 | timekeeping_notify(curr_clocksource);
|
| 578 | }
|
| 579 | }
|
| 580 |
|
| 581 | #else
|
| 582 |
|
| 583 | static inline void clocksource_select(void) { }
|
| 584 |
|
| 585 | #endif
|
| 586 |
|
| 587 | |
| 588 | * clocksource_done_booting - Called near the end of core bootup
|
| 589 | *
|
| 590 | * Hack to avoid lots of clocksource churn at boot time.
|
| 591 | * We use fs_initcall because we want this to start before
|
| 592 | * device_initcall but after subsys_initcall.
|
| 593 |
|
| 594 | static int __init clocksource_done_booting(void)
|
| 595 | {
|
| 596 | mutex_lock(&clocksource_mutex);
|
| 597 | curr_clocksource = clocksource_default_clock();
|
| 598 | mutex_unlock(&clocksource_mutex);
|
| 599 |
|
| 600 | finished_booting = 1;
|
| 601 |
|
| 602 | |
| 603 | * Run the watchdog first to eliminate unstable clock sources
|
| 604 |
|
| 605 | clocksource_watchdog_kthread(NULL);
|
| 606 |
|
| 607 | mutex_lock(&clocksource_mutex);
|
| 608 | clocksource_select();
|
| 609 | mutex_unlock(&clocksource_mutex);
|
| 610 | return 0;
|
| 611 | }
|
| 612 | fs_initcall(clocksource_done_booting);
|
| 613 |
|
| 614 | |
| 615 | * Enqueue the clocksource sorted by rating
|
| 616 |
|
| 617 | static void clocksource_enqueue(struct clocksource *cs)
|
| 618 | {
|
| 619 | struct list_head *entry = &clocksource_list;
|
| 620 | struct clocksource *tmp;
|
| 621 |
|
| 622 | list_for_each_entry(tmp, &clocksource_list, list)
|
| 623 |
|
| 624 | if (tmp->rating >= cs->rating)
|
| 625 | entry = &tmp->list;
|
| 626 | list_add(&cs->list, entry);
|
| 627 | }
|
| 628 |
|
| 629 |
|
| 630 | |
| 631 | * Maximum time we expect to go between ticks. This includes idle
|
| 632 | * tickless time. It provides the trade off between selecting a
|
| 633 | * mult/shift pair that is very precise but can only handle a short
|
| 634 | * period of time, vs. a mult/shift pair that can handle long periods
|
| 635 | * of time but isn't as precise.
|
| 636 | *
|
| 637 | * This is a subsystem constant, and actual hardware limitations
|
| 638 | * may override it (ie: clocksources that wrap every 3 seconds).
|
| 639 |
|
| 640 | #define MAX_UPDATE_LENGTH 5
|
| 641 |
|
| 642 | |
| 643 | * __clocksource_updatefreq_scale - Used update clocksource with new freq
|
| 644 | * @t: clocksource to be registered
|
| 645 | * @scale: Scale factor multiplied against freq to get clocksource hz
|
| 646 | * @freq: clocksource frequency (cycles per second) divided by scale
|
| 647 | *
|
| 648 | * This should only be called from the clocksource->enable() method.
|
| 649 | *
|
| 650 | * This *SHOULD NOT* be called directly! Please use the
|
| 651 | * clocksource_updatefreq_hz() or clocksource_updatefreq_khz helper functions.
|
| 652 |
|
| 653 | void __clocksource_updatefreq_scale(struct clocksource *cs, u32 scale, u32 freq)
|
| 654 | {
|
| 655 | |
| 656 | * Ideally we want to use some of the limits used in
|
| 657 | * clocksource_max_deferment, to provide a more informed
|
| 658 | * MAX_UPDATE_LENGTH. But for now this just gets the
|
| 659 | * register interface working properly.
|
| 660 |
|
| 661 | clocks_calc_mult_shift(&cs->mult, &cs->shift, freq,
|
| 662 | NSEC_PER_SEC/scale,
|
| 663 | MAX_UPDATE_LENGTH*scale);
|
| 664 | cs->max_idle_ns = clocksource_max_deferment(cs);
|
| 665 | }
|
| 666 | EXPORT_SYMBOL_GPL(__clocksource_updatefreq_scale);
|
| 667 |
|
| 668 | |
| 669 | * __clocksource_register_scale - Used to install new clocksources
|
| 670 | * @t: clocksource to be registered
|
| 671 | * @scale: Scale factor multiplied against freq to get clocksource hz
|
| 672 | * @freq: clocksource frequency (cycles per second) divided by scale
|
| 673 | *
|
| 674 | * Returns -EBUSY if registration fails, zero otherwise.
|
| 675 | *
|
| 676 | * This *SHOULD NOT* be called directly! Please use the
|
| 677 | * clocksource_register_hz() or clocksource_register_khz helper functions.
|
| 678 |
|
| 679 | int __clocksource_register_scale(struct clocksource *cs, u32 scale, u32 freq)
|
| 680 | {
|
| 681 |
|
| 682 |
|
| 683 | __clocksource_updatefreq_scale(cs, scale, freq);
|
| 684 |
|
| 685 |
|
| 686 | mutex_lock(&clocksource_mutex);
|
| 687 | clocksource_enqueue(cs);
|
| 688 | clocksource_enqueue_watchdog(cs);
|
| 689 | clocksource_select();
|
| 690 | mutex_unlock(&clocksource_mutex);
|
| 691 | return 0;
|
| 692 | }
|
| 693 | EXPORT_SYMBOL_GPL(__clocksource_register_scale);
|
| 694 |
|
| 695 |
|
| 696 | |
| 697 | * clocksource_register - Used to install new clocksources
|
| 698 | * @t: clocksource to be registered
|
| 699 | *
|
| 700 | * Returns -EBUSY if registration fails, zero otherwise.
|
| 701 |
|
| 702 | int clocksource_register(struct clocksource *cs)
|
| 703 | {
|
| 704 |
|
| 705 | cs->max_idle_ns = clocksource_max_deferment(cs);
|
| 706 |
|
| 707 | mutex_lock(&clocksource_mutex);
|
| 708 | clocksource_enqueue(cs);
|
| 709 | clocksource_enqueue_watchdog(cs);
|
| 710 | clocksource_select();
|
| 711 | mutex_unlock(&clocksource_mutex);
|
| 712 | return 0;
|
| 713 | }
|
| 714 | EXPORT_SYMBOL(clocksource_register);
|
| 715 |
|
| 716 | static void __clocksource_change_rating(struct clocksource *cs, int rating)
|
| 717 | {
|
| 718 | list_del(&cs->list);
|
| 719 | cs->rating = rating;
|
| 720 | clocksource_enqueue(cs);
|
| 721 | clocksource_select();
|
| 722 | }
|
| 723 |
|
| 724 | |
| 725 | * clocksource_change_rating - Change the rating of a registered clocksource
|
| 726 |
|
| 727 | void clocksource_change_rating(struct clocksource *cs, int rating)
|
| 728 | {
|
| 729 | mutex_lock(&clocksource_mutex);
|
| 730 | __clocksource_change_rating(cs, rating);
|
| 731 | mutex_unlock(&clocksource_mutex);
|
| 732 | }
|
| 733 | EXPORT_SYMBOL(clocksource_change_rating);
|
| 734 |
|
| 735 | |
| 736 | * clocksource_unregister - remove a registered clocksource
|
| 737 |
|
| 738 | void clocksource_unregister(struct clocksource *cs)
|
| 739 | {
|
| 740 | mutex_lock(&clocksource_mutex);
|
| 741 | clocksource_dequeue_watchdog(cs);
|
| 742 | list_del(&cs->list);
|
| 743 | clocksource_select();
|
| 744 | mutex_unlock(&clocksource_mutex);
|
| 745 | }
|
| 746 | EXPORT_SYMBOL(clocksource_unregister);
|
| 747 |
|
| 748 | #ifdef CONFIG_SYSFS
|
| 749 | |
| 750 | * sysfs_show_current_clocksources - sysfs interface for current clocksource
|
| 751 | * @dev: unused
|
| 752 | * @buf: char buffer to be filled with clocksource list
|
| 753 | *
|
| 754 | * Provides sysfs interface for listing current clocksource.
|
| 755 |
|
| 756 | static ssize_t
|
| 757 | sysfs_show_current_clocksources(struct sys_device *dev,
|
| 758 | struct sysdev_attribute *attr, char *buf)
|
| 759 | {
|
| 760 | ssize_t count = 0;
|
| 761 |
|
| 762 | mutex_lock(&clocksource_mutex);
|
| 763 | count = snprintf(buf, PAGE_SIZE, "%s\n", curr_clocksource->name);
|
| 764 | mutex_unlock(&clocksource_mutex);
|
| 765 |
|
| 766 | return count;
|
| 767 | }
|
| 768 |
|
| 769 | |
| 770 | * sysfs_override_clocksource - interface for manually overriding clocksource
|
| 771 | * @dev: unused
|
| 772 | * @buf: name of override clocksource
|
| 773 | * @count: length of buffer
|
| 774 | *
|
| 775 | * Takes input from sysfs interface for manually overriding the default
|
| 776 | * clocksource selection.
|
| 777 |
|
| 778 | static ssize_t sysfs_override_clocksource(struct sys_device *dev,
|
| 779 | struct sysdev_attribute *attr,
|
| 780 | const char *buf, size_t count)
|
| 781 | {
|
| 782 | size_t ret = count;
|
| 783 |
|
| 784 |
|
| 785 | if (count >= sizeof(override_name))
|
| 786 | return -EINVAL;
|
| 787 |
|
| 788 |
|
| 789 | if (buf[count-1] == '\n')
|
| 790 | count--;
|
| 791 |
|
| 792 | mutex_lock(&clocksource_mutex);
|
| 793 |
|
| 794 | if (count > 0)
|
| 795 | memcpy(override_name, buf, count);
|
| 796 | override_name[count] = 0;
|
| 797 | clocksource_select();
|
| 798 |
|
| 799 | mutex_unlock(&clocksource_mutex);
|
| 800 |
|
| 801 | return ret;
|
| 802 | }
|
| 803 |
|
| 804 | |
| 805 | * sysfs_show_available_clocksources - sysfs interface for listing clocksource
|
| 806 | * @dev: unused
|
| 807 | * @buf: char buffer to be filled with clocksource list
|
| 808 | *
|
| 809 | * Provides sysfs interface for listing registered clocksources
|
| 810 |
|
| 811 | static ssize_t
|
| 812 | sysfs_show_available_clocksources(struct sys_device *dev,
|
| 813 | struct sysdev_attribute *attr,
|
| 814 | char *buf)
|
| 815 | {
|
| 816 | struct clocksource *src;
|
| 817 | ssize_t count = 0;
|
| 818 |
|
| 819 | mutex_lock(&clocksource_mutex);
|
| 820 | list_for_each_entry(src, &clocksource_list, list) {
|
| 821 | |
| 822 | * Don't show non-HRES clocksource if the tick code is
|
| 823 | * in one shot mode (highres=on or nohz=on)
|
| 824 |
|
| 825 | if (!tick_oneshot_mode_active() ||
|
| 826 | (src->flags & CLOCK_SOURCE_VALID_FOR_HRES))
|
| 827 | count += snprintf(buf + count,
|
| 828 | max((ssize_t)PAGE_SIZE - count, (ssize_t)0),
|
| 829 | "%s ", src->name);
|
| 830 | }
|
| 831 | mutex_unlock(&clocksource_mutex);
|
| 832 |
|
| 833 | count += snprintf(buf + count,
|
| 834 | max((ssize_t)PAGE_SIZE - count, (ssize_t)0), "\n");
|
| 835 |
|
| 836 | return count;
|
| 837 | }
|
| 838 |
|
| 839 | |
| 840 | * Sysfs setup bits:
|
| 841 |
|
| 842 | static SYSDEV_ATTR(current_clocksource, 0644, sysfs_show_current_clocksources,
|
| 843 | sysfs_override_clocksource);
|
| 844 |
|
| 845 | static SYSDEV_ATTR(available_clocksource, 0444,
|
| 846 | sysfs_show_available_clocksources, NULL);
|
| 847 |
|
| 848 | static struct sysdev_class clocksource_sysclass = {
|
| 849 | .name = "clocksource",
|
| 850 | };
|
| 851 |
|
| 852 | static struct sys_device device_clocksource = {
|
| 853 | .id = 0,
|
| 854 | .cls = &clocksource_sysclass,
|
| 855 | };
|
| 856 |
|
| 857 | static int __init init_clocksource_sysfs(void)
|
| 858 | {
|
| 859 | int error = sysdev_class_register(&clocksource_sysclass);
|
| 860 |
|
| 861 | if (!error)
|
| 862 | error = sysdev_register(&device_clocksource);
|
| 863 | if (!error)
|
| 864 | error = sysdev_create_file(
|
| 865 | &device_clocksource,
|
| 866 | &attr_current_clocksource);
|
| 867 | if (!error)
|
| 868 | error = sysdev_create_file(
|
| 869 | &device_clocksource,
|
| 870 | &attr_available_clocksource);
|
| 871 | return error;
|
| 872 | }
|
| 873 |
|
| 874 | device_initcall(init_clocksource_sysfs);
|
| 875 | #endif
|
| 876 |
|
| 877 | |
| 878 | * boot_override_clocksource - boot clock override
|
| 879 | * @str: override name
|
| 880 | *
|
| 881 | * Takes a clocksource= boot argument and uses it
|
| 882 | * as the clocksource override name.
|
| 883 |
|
| 884 | static int __init boot_override_clocksource(char* str)
|
| 885 | {
|
| 886 | mutex_lock(&clocksource_mutex);
|
| 887 | if (str)
|
| 888 | strlcpy(override_name, str, sizeof(override_name));
|
| 889 | mutex_unlock(&clocksource_mutex);
|
| 890 | return 1;
|
| 891 | }
|
| 892 |
|
| 893 | __setup("clocksource=", boot_override_clocksource);
|
| 894 |
|
| 895 | |
| 896 | * boot_override_clock - Compatibility layer for deprecated boot option
|
| 897 | * @str: override name
|
| 898 | *
|
| 899 | * DEPRECATED! Takes a clock= boot argument and uses it
|
| 900 | * as the clocksource override name
|
| 901 |
|
| 902 | static int __init boot_override_clock(char* str)
|
| 903 | {
|
| 904 | if (!strcmp(str, "pmtmr")) {
|
| 905 | printk("Warning: clock=pmtmr is deprecated. "
|
| 906 | "Use clocksource=acpi_pm.\n");
|
| 907 | return boot_override_clocksource("acpi_pm");
|
| 908 | }
|
| 909 | printk("Warning! clock= boot option is deprecated. "
|
| 910 | "Use clocksource=xyz\n");
|
| 911 | return boot_override_clocksource(str);
|
| 912 | }
|
| 913 |
|
| 914 | __setup("clock=", boot_override_clock);
|