2 * linux/kernel/time/clocksource.c
4 * This file contains the functions which manage clocksource drivers.
6 * Copyright (C) 2004, 2005 IBM, John Stultz (johnstul@us.ibm.com)
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.
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.
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.
23 * o Allow clocksource drivers to be unregistered
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> /* for spin_unlock_irq() using preempt_count() m68k */
31 #include <linux/tick.h>
32 #include <linux/kthread.h>
34 void timecounter_init(struct timecounter *tc,
35 const struct cyclecounter *cc,
39 tc->cycle_last = cc->read(cc);
40 tc->nsec = start_tstamp;
42 EXPORT_SYMBOL(timecounter_init);
45 * timecounter_read_delta - get nanoseconds since last call of this function
46 * @tc: Pointer to time counter
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
52 * The first call to this function for a new time counter initializes
53 * the time tracking and returns an undefined result.
55 static u64 timecounter_read_delta(struct timecounter *tc)
57 cycle_t cycle_now, cycle_delta;
60 /* read cycle counter: */
61 cycle_now = tc->cc->read(tc->cc);
63 /* calculate the delta since the last timecounter_read_delta(): */
64 cycle_delta = (cycle_now - tc->cycle_last) & tc->cc->mask;
66 /* convert to nanoseconds: */
67 ns_offset = cyclecounter_cyc2ns(tc->cc, cycle_delta);
69 /* update time stamp of timecounter_read_delta() call: */
70 tc->cycle_last = cycle_now;
75 u64 timecounter_read(struct timecounter *tc)
79 /* increment time by nanoseconds since last call */
80 nsec = timecounter_read_delta(tc);
86 EXPORT_SYMBOL(timecounter_read);
88 u64 timecounter_cyc2time(struct timecounter *tc,
91 u64 cycle_delta = (cycle_tstamp - tc->cycle_last) & tc->cc->mask;
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.
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);
103 nsec = cyclecounter_cyc2ns(tc->cc, cycle_delta) + tc->nsec;
108 EXPORT_SYMBOL(timecounter_cyc2time);
110 /*[Clocksource internal variables]---------
112 * currently selected clocksource.
114 * linked list with the registered clocksources
116 * protects manipulations to curr_clocksource and the clocksource_list
118 * Name of the user-specified clocksource.
120 static struct clocksource *curr_clocksource;
121 static LIST_HEAD(clocksource_list);
122 static DEFINE_MUTEX(clocksource_mutex);
123 static char override_name[32];
125 #ifdef CONFIG_CLOCKSOURCE_WATCHDOG
126 static LIST_HEAD(watchdog_list);
127 static struct clocksource *watchdog;
128 static struct timer_list watchdog_timer;
129 static struct work_struct watchdog_work;
130 static DEFINE_SPINLOCK(watchdog_lock);
131 static cycle_t watchdog_last;
132 static int watchdog_running;
134 static int clocksource_watchdog_kthread(void *data);
135 static void __clocksource_change_rating(struct clocksource *cs, int rating);
138 * Interval: 0.5sec Threshold: 0.0625s
140 #define WATCHDOG_INTERVAL (HZ >> 1)
141 #define WATCHDOG_THRESHOLD (NSEC_PER_SEC >> 4)
143 static void clocksource_watchdog_work(struct work_struct *work)
146 * If kthread_run fails the next watchdog scan over the
147 * watchdog_list will find the unstable clock again.
149 kthread_run(clocksource_watchdog_kthread, NULL, "kwatchdog");
152 static void clocksource_unstable(struct clocksource *cs, int64_t delta)
154 printk(KERN_WARNING "Clocksource %s unstable (delta = %Ld ns)\n",
156 cs->flags &= ~(CLOCK_SOURCE_VALID_FOR_HRES | CLOCK_SOURCE_WATCHDOG);
157 cs->flags |= CLOCK_SOURCE_UNSTABLE;
158 schedule_work(&watchdog_work);
161 static void clocksource_watchdog(unsigned long data)
163 struct clocksource *cs;
164 cycle_t csnow, wdnow;
165 int64_t wd_nsec, cs_nsec;
168 spin_lock(&watchdog_lock);
169 if (!watchdog_running)
172 wdnow = watchdog->read(watchdog);
173 wd_nsec = clocksource_cyc2ns((wdnow - watchdog_last) & watchdog->mask,
174 watchdog->mult, watchdog->shift);
175 watchdog_last = wdnow;
177 list_for_each_entry(cs, &watchdog_list, wd_list) {
179 /* Clocksource already marked unstable? */
180 if (cs->flags & CLOCK_SOURCE_UNSTABLE) {
181 schedule_work(&watchdog_work);
185 csnow = cs->read(cs);
187 /* Clocksource initialized ? */
188 if (!(cs->flags & CLOCK_SOURCE_WATCHDOG)) {
189 cs->flags |= CLOCK_SOURCE_WATCHDOG;
194 /* Check the deviation from the watchdog clocksource. */
195 cs_nsec = clocksource_cyc2ns((csnow - cs->wd_last) &
196 cs->mask, cs->mult, cs->shift);
198 if (abs(cs_nsec - wd_nsec) > WATCHDOG_THRESHOLD) {
199 clocksource_unstable(cs, cs_nsec - wd_nsec);
203 if (!(cs->flags & CLOCK_SOURCE_VALID_FOR_HRES) &&
204 (cs->flags & CLOCK_SOURCE_IS_CONTINUOUS) &&
205 (watchdog->flags & CLOCK_SOURCE_IS_CONTINUOUS)) {
206 cs->flags |= CLOCK_SOURCE_VALID_FOR_HRES;
208 * We just marked the clocksource as highres-capable,
209 * notify the rest of the system as well so that we
210 * transition into high-res mode:
217 * Cycle through CPUs to check if the CPUs stay synchronized
220 next_cpu = cpumask_next(raw_smp_processor_id(), cpu_online_mask);
221 if (next_cpu >= nr_cpu_ids)
222 next_cpu = cpumask_first(cpu_online_mask);
223 watchdog_timer.expires += WATCHDOG_INTERVAL;
224 add_timer_on(&watchdog_timer, next_cpu);
226 spin_unlock(&watchdog_lock);
229 static inline void clocksource_start_watchdog(void)
231 if (watchdog_running || !watchdog || list_empty(&watchdog_list))
233 INIT_WORK(&watchdog_work, clocksource_watchdog_work);
234 init_timer(&watchdog_timer);
235 watchdog_timer.function = clocksource_watchdog;
236 watchdog_last = watchdog->read(watchdog);
237 watchdog_timer.expires = jiffies + WATCHDOG_INTERVAL;
238 add_timer_on(&watchdog_timer, cpumask_first(cpu_online_mask));
239 watchdog_running = 1;
242 static inline void clocksource_stop_watchdog(void)
244 if (!watchdog_running || (watchdog && !list_empty(&watchdog_list)))
246 del_timer(&watchdog_timer);
247 watchdog_running = 0;
250 static inline void clocksource_reset_watchdog(void)
252 struct clocksource *cs;
254 list_for_each_entry(cs, &watchdog_list, wd_list)
255 cs->flags &= ~CLOCK_SOURCE_WATCHDOG;
258 static void clocksource_resume_watchdog(void)
262 spin_lock_irqsave(&watchdog_lock, flags);
263 clocksource_reset_watchdog();
264 spin_unlock_irqrestore(&watchdog_lock, flags);
267 static void clocksource_enqueue_watchdog(struct clocksource *cs)
271 spin_lock_irqsave(&watchdog_lock, flags);
272 if (cs->flags & CLOCK_SOURCE_MUST_VERIFY) {
273 /* cs is a clocksource to be watched. */
274 list_add(&cs->wd_list, &watchdog_list);
275 cs->flags &= ~CLOCK_SOURCE_WATCHDOG;
277 /* cs is a watchdog. */
278 if (cs->flags & CLOCK_SOURCE_IS_CONTINUOUS)
279 cs->flags |= CLOCK_SOURCE_VALID_FOR_HRES;
280 /* Pick the best watchdog. */
281 if (!watchdog || cs->rating > watchdog->rating) {
283 /* Reset watchdog cycles */
284 clocksource_reset_watchdog();
287 /* Check if the watchdog timer needs to be started. */
288 clocksource_start_watchdog();
289 spin_unlock_irqrestore(&watchdog_lock, flags);
292 static void clocksource_dequeue_watchdog(struct clocksource *cs)
294 struct clocksource *tmp;
297 spin_lock_irqsave(&watchdog_lock, flags);
298 if (cs->flags & CLOCK_SOURCE_MUST_VERIFY) {
299 /* cs is a watched clocksource. */
300 list_del_init(&cs->wd_list);
301 } else if (cs == watchdog) {
302 /* Reset watchdog cycles */
303 clocksource_reset_watchdog();
304 /* Current watchdog is removed. Find an alternative. */
306 list_for_each_entry(tmp, &clocksource_list, list) {
307 if (tmp == cs || tmp->flags & CLOCK_SOURCE_MUST_VERIFY)
309 if (!watchdog || tmp->rating > watchdog->rating)
313 cs->flags &= ~CLOCK_SOURCE_WATCHDOG;
314 /* Check if the watchdog timer needs to be stopped. */
315 clocksource_stop_watchdog();
316 spin_unlock_irqrestore(&watchdog_lock, flags);
319 static int clocksource_watchdog_kthread(void *data)
321 struct clocksource *cs, *tmp;
325 mutex_lock(&clocksource_mutex);
326 spin_lock_irqsave(&watchdog_lock, flags);
327 list_for_each_entry_safe(cs, tmp, &watchdog_list, wd_list)
328 if (cs->flags & CLOCK_SOURCE_UNSTABLE) {
329 list_del_init(&cs->wd_list);
330 list_add(&cs->wd_list, &unstable);
332 /* Check if the watchdog timer needs to be stopped. */
333 clocksource_stop_watchdog();
334 spin_unlock_irqrestore(&watchdog_lock, flags);
336 /* Needs to be done outside of watchdog lock */
337 list_for_each_entry_safe(cs, tmp, &unstable, wd_list) {
338 list_del_init(&cs->wd_list);
339 __clocksource_change_rating(cs, 0);
341 mutex_unlock(&clocksource_mutex);
345 #else /* CONFIG_CLOCKSOURCE_WATCHDOG */
347 static void clocksource_enqueue_watchdog(struct clocksource *cs)
349 if (cs->flags & CLOCK_SOURCE_IS_CONTINUOUS)
350 cs->flags |= CLOCK_SOURCE_VALID_FOR_HRES;
353 static inline void clocksource_dequeue_watchdog(struct clocksource *cs) { }
354 static inline void clocksource_resume_watchdog(void) { }
356 #endif /* CONFIG_CLOCKSOURCE_WATCHDOG */
359 * clocksource_resume - resume the clocksource(s)
361 void clocksource_resume(void)
363 struct clocksource *cs;
365 mutex_lock(&clocksource_mutex);
367 list_for_each_entry(cs, &clocksource_list, list)
371 clocksource_resume_watchdog();
373 mutex_unlock(&clocksource_mutex);
377 * clocksource_touch_watchdog - Update watchdog
379 * Update the watchdog after exception contexts such as kgdb so as not
380 * to incorrectly trip the watchdog.
383 void clocksource_touch_watchdog(void)
385 clocksource_resume_watchdog();
388 #ifdef CONFIG_GENERIC_TIME
390 static int finished_booting;
393 * clocksource_select - Select the best clocksource available
395 * Private function. Must hold clocksource_mutex when called.
397 * Select the clocksource with the best rating, or the clocksource,
398 * which is selected by userspace override.
400 static void clocksource_select(void)
402 struct clocksource *best, *cs;
404 if (!finished_booting || list_empty(&clocksource_list))
406 /* First clocksource on the list has the best rating. */
407 best = list_first_entry(&clocksource_list, struct clocksource, list);
408 /* Check for the override clocksource. */
409 list_for_each_entry(cs, &clocksource_list, list) {
410 if (strcmp(cs->name, override_name) != 0)
413 * Check to make sure we don't switch to a non-highres
414 * capable clocksource if the tick code is in oneshot
415 * mode (highres or nohz)
417 if (!(cs->flags & CLOCK_SOURCE_VALID_FOR_HRES) &&
418 tick_oneshot_mode_active()) {
419 /* Override clocksource cannot be used. */
420 printk(KERN_WARNING "Override clocksource %s is not "
421 "HRT compatible. Cannot switch while in "
422 "HRT/NOHZ mode\n", cs->name);
423 override_name[0] = 0;
425 /* Override clocksource can be used. */
429 if (curr_clocksource != best) {
430 printk(KERN_INFO "Switching to clocksource %s\n", best->name);
431 curr_clocksource = best;
432 timekeeping_notify(curr_clocksource);
437 * clocksource_done_booting - Called near the end of core bootup
439 * Hack to avoid lots of clocksource churn at boot time.
440 * We use fs_initcall because we want this to start before
441 * device_initcall but after subsys_initcall.
443 static int __init clocksource_done_booting(void)
445 finished_booting = 1;
446 clocksource_select();
449 fs_initcall(clocksource_done_booting);
451 #else /* CONFIG_GENERIC_TIME */
453 static inline void clocksource_select(void) { }
458 * Enqueue the clocksource sorted by rating
460 static void clocksource_enqueue(struct clocksource *cs)
462 struct list_head *entry = &clocksource_list;
463 struct clocksource *tmp;
465 list_for_each_entry(tmp, &clocksource_list, list)
466 /* Keep track of the place, where to insert */
467 if (tmp->rating >= cs->rating)
469 list_add(&cs->list, entry);
473 * clocksource_register - Used to install new clocksources
474 * @t: clocksource to be registered
476 * Returns -EBUSY if registration fails, zero otherwise.
478 int clocksource_register(struct clocksource *cs)
480 mutex_lock(&clocksource_mutex);
481 clocksource_enqueue(cs);
482 clocksource_select();
483 clocksource_enqueue_watchdog(cs);
484 mutex_unlock(&clocksource_mutex);
487 EXPORT_SYMBOL(clocksource_register);
489 static void __clocksource_change_rating(struct clocksource *cs, int rating)
493 clocksource_enqueue(cs);
494 clocksource_select();
498 * clocksource_change_rating - Change the rating of a registered clocksource
500 void clocksource_change_rating(struct clocksource *cs, int rating)
502 mutex_lock(&clocksource_mutex);
503 __clocksource_change_rating(cs, rating);
504 mutex_unlock(&clocksource_mutex);
506 EXPORT_SYMBOL(clocksource_change_rating);
509 * clocksource_unregister - remove a registered clocksource
511 void clocksource_unregister(struct clocksource *cs)
513 mutex_lock(&clocksource_mutex);
514 clocksource_dequeue_watchdog(cs);
516 clocksource_select();
517 mutex_unlock(&clocksource_mutex);
519 EXPORT_SYMBOL(clocksource_unregister);
523 * sysfs_show_current_clocksources - sysfs interface for current clocksource
525 * @buf: char buffer to be filled with clocksource list
527 * Provides sysfs interface for listing current clocksource.
530 sysfs_show_current_clocksources(struct sys_device *dev,
531 struct sysdev_attribute *attr, char *buf)
535 mutex_lock(&clocksource_mutex);
536 count = snprintf(buf, PAGE_SIZE, "%s\n", curr_clocksource->name);
537 mutex_unlock(&clocksource_mutex);
543 * sysfs_override_clocksource - interface for manually overriding clocksource
545 * @buf: name of override clocksource
546 * @count: length of buffer
548 * Takes input from sysfs interface for manually overriding the default
549 * clocksource selction.
551 static ssize_t sysfs_override_clocksource(struct sys_device *dev,
552 struct sysdev_attribute *attr,
553 const char *buf, size_t count)
557 /* strings from sysfs write are not 0 terminated! */
558 if (count >= sizeof(override_name))
562 if (buf[count-1] == '\n')
565 mutex_lock(&clocksource_mutex);
568 memcpy(override_name, buf, count);
569 override_name[count] = 0;
570 clocksource_select();
572 mutex_unlock(&clocksource_mutex);
578 * sysfs_show_available_clocksources - sysfs interface for listing clocksource
580 * @buf: char buffer to be filled with clocksource list
582 * Provides sysfs interface for listing registered clocksources
585 sysfs_show_available_clocksources(struct sys_device *dev,
586 struct sysdev_attribute *attr,
589 struct clocksource *src;
592 mutex_lock(&clocksource_mutex);
593 list_for_each_entry(src, &clocksource_list, list) {
595 * Don't show non-HRES clocksource if the tick code is
596 * in one shot mode (highres=on or nohz=on)
598 if (!tick_oneshot_mode_active() ||
599 (src->flags & CLOCK_SOURCE_VALID_FOR_HRES))
600 count += snprintf(buf + count,
601 max((ssize_t)PAGE_SIZE - count, (ssize_t)0),
604 mutex_unlock(&clocksource_mutex);
606 count += snprintf(buf + count,
607 max((ssize_t)PAGE_SIZE - count, (ssize_t)0), "\n");
615 static SYSDEV_ATTR(current_clocksource, 0644, sysfs_show_current_clocksources,
616 sysfs_override_clocksource);
618 static SYSDEV_ATTR(available_clocksource, 0444,
619 sysfs_show_available_clocksources, NULL);
621 static struct sysdev_class clocksource_sysclass = {
622 .name = "clocksource",
625 static struct sys_device device_clocksource = {
627 .cls = &clocksource_sysclass,
630 static int __init init_clocksource_sysfs(void)
632 int error = sysdev_class_register(&clocksource_sysclass);
635 error = sysdev_register(&device_clocksource);
637 error = sysdev_create_file(
639 &attr_current_clocksource);
641 error = sysdev_create_file(
643 &attr_available_clocksource);
647 device_initcall(init_clocksource_sysfs);
648 #endif /* CONFIG_SYSFS */
651 * boot_override_clocksource - boot clock override
652 * @str: override name
654 * Takes a clocksource= boot argument and uses it
655 * as the clocksource override name.
657 static int __init boot_override_clocksource(char* str)
659 mutex_lock(&clocksource_mutex);
661 strlcpy(override_name, str, sizeof(override_name));
662 mutex_unlock(&clocksource_mutex);
666 __setup("clocksource=", boot_override_clocksource);
669 * boot_override_clock - Compatibility layer for deprecated boot option
670 * @str: override name
672 * DEPRECATED! Takes a clock= boot argument and uses it
673 * as the clocksource override name
675 static int __init boot_override_clock(char* str)
677 if (!strcmp(str, "pmtmr")) {
678 printk("Warning: clock=pmtmr is deprecated. "
679 "Use clocksource=acpi_pm.\n");
680 return boot_override_clocksource("acpi_pm");
682 printk("Warning! clock= boot option is deprecated. "
683 "Use clocksource=xyz\n");
684 return boot_override_clocksource(str);
687 __setup("clock=", boot_override_clock);