4 * Kernel internal timers, basic process system calls
6 * Copyright (C) 1991, 1992 Linus Torvalds
8 * 1997-01-28 Modified by Finn Arne Gangstad to make timers scale better.
10 * 1997-09-10 Updated NTP code according to technical memorandum Jan '96
11 * "A Kernel Model for Precision Timekeeping" by Dave Mills
12 * 1998-12-24 Fixed a xtime SMP race (we need the xtime_lock rw spinlock to
13 * serialize accesses to xtime/lost_ticks).
14 * Copyright (C) 1998 Andrea Arcangeli
15 * 1999-03-10 Improved NTP compatibility by Ulrich Windl
16 * 2002-05-31 Move sys_sysinfo here and make its locking sane, Robert Love
17 * 2000-10-05 Implemented scalable SMP per-CPU timer handling.
18 * Copyright (C) 2000, 2001, 2002 Ingo Molnar
19 * Designed by David S. Miller, Alexey Kuznetsov and Ingo Molnar
22 #include <linux/kernel_stat.h>
23 #include <linux/export.h>
24 #include <linux/interrupt.h>
25 #include <linux/percpu.h>
26 #include <linux/init.h>
28 #include <linux/swap.h>
29 #include <linux/pid_namespace.h>
30 #include <linux/notifier.h>
31 #include <linux/thread_info.h>
32 #include <linux/time.h>
33 #include <linux/jiffies.h>
34 #include <linux/posix-timers.h>
35 #include <linux/cpu.h>
36 #include <linux/syscalls.h>
37 #include <linux/delay.h>
38 #include <linux/tick.h>
39 #include <linux/kallsyms.h>
40 #include <linux/irq_work.h>
41 #include <linux/sched.h>
42 #include <linux/slab.h>
44 #include <asm/uaccess.h>
45 #include <asm/unistd.h>
46 #include <asm/div64.h>
47 #include <asm/timex.h>
50 #define CREATE_TRACE_POINTS
51 #include <trace/events/timer.h>
53 u64 jiffies_64 __cacheline_aligned_in_smp = INITIAL_JIFFIES;
55 EXPORT_SYMBOL(jiffies_64);
58 * per-CPU timer vector definitions:
60 #define TVN_BITS (CONFIG_BASE_SMALL ? 4 : 6)
61 #define TVR_BITS (CONFIG_BASE_SMALL ? 6 : 8)
62 #define TVN_SIZE (1 << TVN_BITS)
63 #define TVR_SIZE (1 << TVR_BITS)
64 #define TVN_MASK (TVN_SIZE - 1)
65 #define TVR_MASK (TVR_SIZE - 1)
66 #define MAX_TVAL ((unsigned long)((1ULL << (TVR_BITS + 4*TVN_BITS)) - 1))
69 struct list_head vec[TVN_SIZE];
73 struct list_head vec[TVR_SIZE];
78 struct timer_list *running_timer;
79 unsigned long timer_jiffies;
80 unsigned long next_timer;
86 } ____cacheline_aligned;
88 struct tvec_base boot_tvec_bases;
89 EXPORT_SYMBOL(boot_tvec_bases);
90 static DEFINE_PER_CPU(struct tvec_base *, tvec_bases) = &boot_tvec_bases;
92 /* Functions below help us manage 'deferrable' flag */
93 static inline unsigned int tbase_get_deferrable(struct tvec_base *base)
95 return ((unsigned int)(unsigned long)base & TBASE_DEFERRABLE_FLAG);
98 static inline struct tvec_base *tbase_get_base(struct tvec_base *base)
100 return ((struct tvec_base *)((unsigned long)base & ~TBASE_DEFERRABLE_FLAG));
103 static inline void timer_set_deferrable(struct timer_list *timer)
105 timer->base = TBASE_MAKE_DEFERRED(timer->base);
109 timer_set_base(struct timer_list *timer, struct tvec_base *new_base)
111 timer->base = (struct tvec_base *)((unsigned long)(new_base) |
112 tbase_get_deferrable(timer->base));
115 static unsigned long round_jiffies_common(unsigned long j, int cpu,
119 unsigned long original = j;
122 * We don't want all cpus firing their timers at once hitting the
123 * same lock or cachelines, so we skew each extra cpu with an extra
124 * 3 jiffies. This 3 jiffies came originally from the mm/ code which
126 * The skew is done by adding 3*cpunr, then round, then subtract this
127 * extra offset again.
134 * If the target jiffie is just after a whole second (which can happen
135 * due to delays of the timer irq, long irq off times etc etc) then
136 * we should round down to the whole second, not up. Use 1/4th second
137 * as cutoff for this rounding as an extreme upper bound for this.
138 * But never round down if @force_up is set.
140 if (rem < HZ/4 && !force_up) /* round down */
145 /* now that we have rounded, subtract the extra skew again */
148 if (j <= jiffies) /* rounding ate our timeout entirely; */
154 * __round_jiffies - function to round jiffies to a full second
155 * @j: the time in (absolute) jiffies that should be rounded
156 * @cpu: the processor number on which the timeout will happen
158 * __round_jiffies() rounds an absolute time in the future (in jiffies)
159 * up or down to (approximately) full seconds. This is useful for timers
160 * for which the exact time they fire does not matter too much, as long as
161 * they fire approximately every X seconds.
163 * By rounding these timers to whole seconds, all such timers will fire
164 * at the same time, rather than at various times spread out. The goal
165 * of this is to have the CPU wake up less, which saves power.
167 * The exact rounding is skewed for each processor to avoid all
168 * processors firing at the exact same time, which could lead
169 * to lock contention or spurious cache line bouncing.
171 * The return value is the rounded version of the @j parameter.
173 unsigned long __round_jiffies(unsigned long j, int cpu)
175 return round_jiffies_common(j, cpu, false);
177 EXPORT_SYMBOL_GPL(__round_jiffies);
180 * __round_jiffies_relative - function to round jiffies to a full second
181 * @j: the time in (relative) jiffies that should be rounded
182 * @cpu: the processor number on which the timeout will happen
184 * __round_jiffies_relative() rounds a time delta in the future (in jiffies)
185 * up or down to (approximately) full seconds. This is useful for timers
186 * for which the exact time they fire does not matter too much, as long as
187 * they fire approximately every X seconds.
189 * By rounding these timers to whole seconds, all such timers will fire
190 * at the same time, rather than at various times spread out. The goal
191 * of this is to have the CPU wake up less, which saves power.
193 * The exact rounding is skewed for each processor to avoid all
194 * processors firing at the exact same time, which could lead
195 * to lock contention or spurious cache line bouncing.
197 * The return value is the rounded version of the @j parameter.
199 unsigned long __round_jiffies_relative(unsigned long j, int cpu)
201 unsigned long j0 = jiffies;
203 /* Use j0 because jiffies might change while we run */
204 return round_jiffies_common(j + j0, cpu, false) - j0;
206 EXPORT_SYMBOL_GPL(__round_jiffies_relative);
209 * round_jiffies - function to round jiffies to a full second
210 * @j: the time in (absolute) jiffies that should be rounded
212 * round_jiffies() rounds an absolute time in the future (in jiffies)
213 * up or down to (approximately) full seconds. This is useful for timers
214 * for which the exact time they fire does not matter too much, as long as
215 * they fire approximately every X seconds.
217 * By rounding these timers to whole seconds, all such timers will fire
218 * at the same time, rather than at various times spread out. The goal
219 * of this is to have the CPU wake up less, which saves power.
221 * The return value is the rounded version of the @j parameter.
223 unsigned long round_jiffies(unsigned long j)
225 return round_jiffies_common(j, raw_smp_processor_id(), false);
227 EXPORT_SYMBOL_GPL(round_jiffies);
230 * round_jiffies_relative - function to round jiffies to a full second
231 * @j: the time in (relative) jiffies that should be rounded
233 * round_jiffies_relative() rounds a time delta in the future (in jiffies)
234 * up or down to (approximately) full seconds. This is useful for timers
235 * for which the exact time they fire does not matter too much, as long as
236 * they fire approximately every X seconds.
238 * By rounding these timers to whole seconds, all such timers will fire
239 * at the same time, rather than at various times spread out. The goal
240 * of this is to have the CPU wake up less, which saves power.
242 * The return value is the rounded version of the @j parameter.
244 unsigned long round_jiffies_relative(unsigned long j)
246 return __round_jiffies_relative(j, raw_smp_processor_id());
248 EXPORT_SYMBOL_GPL(round_jiffies_relative);
251 * __round_jiffies_up - function to round jiffies up to a full second
252 * @j: the time in (absolute) jiffies that should be rounded
253 * @cpu: the processor number on which the timeout will happen
255 * This is the same as __round_jiffies() except that it will never
256 * round down. This is useful for timeouts for which the exact time
257 * of firing does not matter too much, as long as they don't fire too
260 unsigned long __round_jiffies_up(unsigned long j, int cpu)
262 return round_jiffies_common(j, cpu, true);
264 EXPORT_SYMBOL_GPL(__round_jiffies_up);
267 * __round_jiffies_up_relative - function to round jiffies up to a full second
268 * @j: the time in (relative) jiffies that should be rounded
269 * @cpu: the processor number on which the timeout will happen
271 * This is the same as __round_jiffies_relative() except that it will never
272 * round down. This is useful for timeouts for which the exact time
273 * of firing does not matter too much, as long as they don't fire too
276 unsigned long __round_jiffies_up_relative(unsigned long j, int cpu)
278 unsigned long j0 = jiffies;
280 /* Use j0 because jiffies might change while we run */
281 return round_jiffies_common(j + j0, cpu, true) - j0;
283 EXPORT_SYMBOL_GPL(__round_jiffies_up_relative);
286 * round_jiffies_up - function to round jiffies up to a full second
287 * @j: the time in (absolute) jiffies that should be rounded
289 * This is the same as round_jiffies() except that it will never
290 * round down. This is useful for timeouts for which the exact time
291 * of firing does not matter too much, as long as they don't fire too
294 unsigned long round_jiffies_up(unsigned long j)
296 return round_jiffies_common(j, raw_smp_processor_id(), true);
298 EXPORT_SYMBOL_GPL(round_jiffies_up);
301 * round_jiffies_up_relative - function to round jiffies up to a full second
302 * @j: the time in (relative) jiffies that should be rounded
304 * This is the same as round_jiffies_relative() except that it will never
305 * round down. This is useful for timeouts for which the exact time
306 * of firing does not matter too much, as long as they don't fire too
309 unsigned long round_jiffies_up_relative(unsigned long j)
311 return __round_jiffies_up_relative(j, raw_smp_processor_id());
313 EXPORT_SYMBOL_GPL(round_jiffies_up_relative);
316 * set_timer_slack - set the allowed slack for a timer
317 * @timer: the timer to be modified
318 * @slack_hz: the amount of time (in jiffies) allowed for rounding
320 * Set the amount of time, in jiffies, that a certain timer has
321 * in terms of slack. By setting this value, the timer subsystem
322 * will schedule the actual timer somewhere between
323 * the time mod_timer() asks for, and that time plus the slack.
325 * By setting the slack to -1, a percentage of the delay is used
328 void set_timer_slack(struct timer_list *timer, int slack_hz)
330 timer->slack = slack_hz;
332 EXPORT_SYMBOL_GPL(set_timer_slack);
334 static void internal_add_timer(struct tvec_base *base, struct timer_list *timer)
336 unsigned long expires = timer->expires;
337 unsigned long idx = expires - base->timer_jiffies;
338 struct list_head *vec;
340 if (idx < TVR_SIZE) {
341 int i = expires & TVR_MASK;
342 vec = base->tv1.vec + i;
343 } else if (idx < 1 << (TVR_BITS + TVN_BITS)) {
344 int i = (expires >> TVR_BITS) & TVN_MASK;
345 vec = base->tv2.vec + i;
346 } else if (idx < 1 << (TVR_BITS + 2 * TVN_BITS)) {
347 int i = (expires >> (TVR_BITS + TVN_BITS)) & TVN_MASK;
348 vec = base->tv3.vec + i;
349 } else if (idx < 1 << (TVR_BITS + 3 * TVN_BITS)) {
350 int i = (expires >> (TVR_BITS + 2 * TVN_BITS)) & TVN_MASK;
351 vec = base->tv4.vec + i;
352 } else if ((signed long) idx < 0) {
354 * Can happen if you add a timer with expires == jiffies,
355 * or you set a timer to go off in the past
357 vec = base->tv1.vec + (base->timer_jiffies & TVR_MASK);
360 /* If the timeout is larger than MAX_TVAL (on 64-bit
361 * architectures or with CONFIG_BASE_SMALL=1) then we
362 * use the maximum timeout.
364 if (idx > MAX_TVAL) {
366 expires = idx + base->timer_jiffies;
368 i = (expires >> (TVR_BITS + 3 * TVN_BITS)) & TVN_MASK;
369 vec = base->tv5.vec + i;
374 list_add_tail(&timer->entry, vec);
377 #ifdef CONFIG_TIMER_STATS
378 void __timer_stats_timer_set_start_info(struct timer_list *timer, void *addr)
380 if (timer->start_site)
383 timer->start_site = addr;
384 memcpy(timer->start_comm, current->comm, TASK_COMM_LEN);
385 timer->start_pid = current->pid;
388 static void timer_stats_account_timer(struct timer_list *timer)
390 unsigned int flag = 0;
392 if (likely(!timer->start_site))
394 if (unlikely(tbase_get_deferrable(timer->base)))
395 flag |= TIMER_STATS_FLAG_DEFERRABLE;
397 timer_stats_update_stats(timer, timer->start_pid, timer->start_site,
398 timer->function, timer->start_comm, flag);
402 static void timer_stats_account_timer(struct timer_list *timer) {}
405 #ifdef CONFIG_DEBUG_OBJECTS_TIMERS
407 static struct debug_obj_descr timer_debug_descr;
409 static void *timer_debug_hint(void *addr)
411 return ((struct timer_list *) addr)->function;
415 * fixup_init is called when:
416 * - an active object is initialized
418 static int timer_fixup_init(void *addr, enum debug_obj_state state)
420 struct timer_list *timer = addr;
423 case ODEBUG_STATE_ACTIVE:
424 del_timer_sync(timer);
425 debug_object_init(timer, &timer_debug_descr);
432 /* Stub timer callback for improperly used timers. */
433 static void stub_timer(unsigned long data)
439 * fixup_activate is called when:
440 * - an active object is activated
441 * - an unknown object is activated (might be a statically initialized object)
443 static int timer_fixup_activate(void *addr, enum debug_obj_state state)
445 struct timer_list *timer = addr;
449 case ODEBUG_STATE_NOTAVAILABLE:
451 * This is not really a fixup. The timer was
452 * statically initialized. We just make sure that it
453 * is tracked in the object tracker.
455 if (timer->entry.next == NULL &&
456 timer->entry.prev == TIMER_ENTRY_STATIC) {
457 debug_object_init(timer, &timer_debug_descr);
458 debug_object_activate(timer, &timer_debug_descr);
461 setup_timer(timer, stub_timer, 0);
466 case ODEBUG_STATE_ACTIVE:
475 * fixup_free is called when:
476 * - an active object is freed
478 static int timer_fixup_free(void *addr, enum debug_obj_state state)
480 struct timer_list *timer = addr;
483 case ODEBUG_STATE_ACTIVE:
484 del_timer_sync(timer);
485 debug_object_free(timer, &timer_debug_descr);
493 * fixup_assert_init is called when:
494 * - an untracked/uninit-ed object is found
496 static int timer_fixup_assert_init(void *addr, enum debug_obj_state state)
498 struct timer_list *timer = addr;
501 case ODEBUG_STATE_NOTAVAILABLE:
502 if (timer->entry.prev == TIMER_ENTRY_STATIC) {
504 * This is not really a fixup. The timer was
505 * statically initialized. We just make sure that it
506 * is tracked in the object tracker.
508 debug_object_init(timer, &timer_debug_descr);
511 setup_timer(timer, stub_timer, 0);
519 static struct debug_obj_descr timer_debug_descr = {
520 .name = "timer_list",
521 .debug_hint = timer_debug_hint,
522 .fixup_init = timer_fixup_init,
523 .fixup_activate = timer_fixup_activate,
524 .fixup_free = timer_fixup_free,
525 .fixup_assert_init = timer_fixup_assert_init,
528 static inline void debug_timer_init(struct timer_list *timer)
530 debug_object_init(timer, &timer_debug_descr);
533 static inline void debug_timer_activate(struct timer_list *timer)
535 debug_object_activate(timer, &timer_debug_descr);
538 static inline void debug_timer_deactivate(struct timer_list *timer)
540 debug_object_deactivate(timer, &timer_debug_descr);
543 static inline void debug_timer_free(struct timer_list *timer)
545 debug_object_free(timer, &timer_debug_descr);
548 static inline void debug_timer_assert_init(struct timer_list *timer)
550 debug_object_assert_init(timer, &timer_debug_descr);
553 static void __init_timer(struct timer_list *timer,
555 struct lock_class_key *key);
557 void init_timer_on_stack_key(struct timer_list *timer,
559 struct lock_class_key *key)
561 debug_object_init_on_stack(timer, &timer_debug_descr);
562 __init_timer(timer, name, key);
564 EXPORT_SYMBOL_GPL(init_timer_on_stack_key);
566 void destroy_timer_on_stack(struct timer_list *timer)
568 debug_object_free(timer, &timer_debug_descr);
570 EXPORT_SYMBOL_GPL(destroy_timer_on_stack);
573 static inline void debug_timer_init(struct timer_list *timer) { }
574 static inline void debug_timer_activate(struct timer_list *timer) { }
575 static inline void debug_timer_deactivate(struct timer_list *timer) { }
576 static inline void debug_timer_assert_init(struct timer_list *timer) { }
579 static inline void debug_init(struct timer_list *timer)
581 debug_timer_init(timer);
582 trace_timer_init(timer);
586 debug_activate(struct timer_list *timer, unsigned long expires)
588 debug_timer_activate(timer);
589 trace_timer_start(timer, expires);
592 static inline void debug_deactivate(struct timer_list *timer)
594 debug_timer_deactivate(timer);
595 trace_timer_cancel(timer);
598 static inline void debug_assert_init(struct timer_list *timer)
600 debug_timer_assert_init(timer);
603 static void __init_timer(struct timer_list *timer,
605 struct lock_class_key *key)
607 timer->entry.next = NULL;
608 timer->base = __raw_get_cpu_var(tvec_bases);
610 #ifdef CONFIG_TIMER_STATS
611 timer->start_site = NULL;
612 timer->start_pid = -1;
613 memset(timer->start_comm, 0, TASK_COMM_LEN);
615 lockdep_init_map(&timer->lockdep_map, name, key, 0);
618 void setup_deferrable_timer_on_stack_key(struct timer_list *timer,
620 struct lock_class_key *key,
621 void (*function)(unsigned long),
624 timer->function = function;
626 init_timer_on_stack_key(timer, name, key);
627 timer_set_deferrable(timer);
629 EXPORT_SYMBOL_GPL(setup_deferrable_timer_on_stack_key);
632 * init_timer_key - initialize a timer
633 * @timer: the timer to be initialized
634 * @name: name of the timer
635 * @key: lockdep class key of the fake lock used for tracking timer
636 * sync lock dependencies
638 * init_timer_key() must be done to a timer prior calling *any* of the
639 * other timer functions.
641 void init_timer_key(struct timer_list *timer,
643 struct lock_class_key *key)
646 __init_timer(timer, name, key);
648 EXPORT_SYMBOL(init_timer_key);
650 void init_timer_deferrable_key(struct timer_list *timer,
652 struct lock_class_key *key)
654 init_timer_key(timer, name, key);
655 timer_set_deferrable(timer);
657 EXPORT_SYMBOL(init_timer_deferrable_key);
659 static inline void detach_timer(struct timer_list *timer,
662 struct list_head *entry = &timer->entry;
664 debug_deactivate(timer);
666 __list_del(entry->prev, entry->next);
669 entry->prev = LIST_POISON2;
673 * We are using hashed locking: holding per_cpu(tvec_bases).lock
674 * means that all timers which are tied to this base via timer->base are
675 * locked, and the base itself is locked too.
677 * So __run_timers/migrate_timers can safely modify all timers which could
678 * be found on ->tvX lists.
680 * When the timer's base is locked, and the timer removed from list, it is
681 * possible to set timer->base = NULL and drop the lock: the timer remains
684 static struct tvec_base *lock_timer_base(struct timer_list *timer,
685 unsigned long *flags)
686 __acquires(timer->base->lock)
688 struct tvec_base *base;
691 struct tvec_base *prelock_base = timer->base;
692 base = tbase_get_base(prelock_base);
693 if (likely(base != NULL)) {
694 spin_lock_irqsave(&base->lock, *flags);
695 if (likely(prelock_base == timer->base))
697 /* The timer has migrated to another CPU */
698 spin_unlock_irqrestore(&base->lock, *flags);
705 __mod_timer(struct timer_list *timer, unsigned long expires,
706 bool pending_only, int pinned)
708 struct tvec_base *base, *new_base;
712 timer_stats_timer_set_start_info(timer);
713 BUG_ON(!timer->function);
715 base = lock_timer_base(timer, &flags);
717 if (timer_pending(timer)) {
718 detach_timer(timer, 0);
719 if (timer->expires == base->next_timer &&
720 !tbase_get_deferrable(timer->base))
721 base->next_timer = base->timer_jiffies;
728 debug_activate(timer, expires);
730 cpu = smp_processor_id();
732 #if defined(CONFIG_NO_HZ) && defined(CONFIG_SMP)
733 if (!pinned && get_sysctl_timer_migration() && idle_cpu(cpu))
734 cpu = get_nohz_timer_target();
736 new_base = per_cpu(tvec_bases, cpu);
738 if (base != new_base) {
740 * We are trying to schedule the timer on the local CPU.
741 * However we can't change timer's base while it is running,
742 * otherwise del_timer_sync() can't detect that the timer's
743 * handler yet has not finished. This also guarantees that
744 * the timer is serialized wrt itself.
746 if (likely(base->running_timer != timer)) {
747 /* See the comment in lock_timer_base() */
748 timer_set_base(timer, NULL);
749 spin_unlock(&base->lock);
751 spin_lock(&base->lock);
752 timer_set_base(timer, base);
756 timer->expires = expires;
757 if (time_before(timer->expires, base->next_timer) &&
758 !tbase_get_deferrable(timer->base))
759 base->next_timer = timer->expires;
760 internal_add_timer(base, timer);
763 spin_unlock_irqrestore(&base->lock, flags);
769 * mod_timer_pending - modify a pending timer's timeout
770 * @timer: the pending timer to be modified
771 * @expires: new timeout in jiffies
773 * mod_timer_pending() is the same for pending timers as mod_timer(),
774 * but will not re-activate and modify already deleted timers.
776 * It is useful for unserialized use of timers.
778 int mod_timer_pending(struct timer_list *timer, unsigned long expires)
780 return __mod_timer(timer, expires, true, TIMER_NOT_PINNED);
782 EXPORT_SYMBOL(mod_timer_pending);
785 * Decide where to put the timer while taking the slack into account
788 * 1) calculate the maximum (absolute) time
789 * 2) calculate the highest bit where the expires and new max are different
790 * 3) use this bit to make a mask
791 * 4) use the bitmask to round down the maximum time, so that all last
795 unsigned long apply_slack(struct timer_list *timer, unsigned long expires)
797 unsigned long expires_limit, mask;
800 if (timer->slack >= 0) {
801 expires_limit = expires + timer->slack;
803 long delta = expires - jiffies;
808 expires_limit = expires + delta / 256;
810 mask = expires ^ expires_limit;
814 bit = find_last_bit(&mask, BITS_PER_LONG);
816 mask = (1 << bit) - 1;
818 expires_limit = expires_limit & ~(mask);
820 return expires_limit;
824 * mod_timer - modify a timer's timeout
825 * @timer: the timer to be modified
826 * @expires: new timeout in jiffies
828 * mod_timer() is a more efficient way to update the expire field of an
829 * active timer (if the timer is inactive it will be activated)
831 * mod_timer(timer, expires) is equivalent to:
833 * del_timer(timer); timer->expires = expires; add_timer(timer);
835 * Note that if there are multiple unserialized concurrent users of the
836 * same timer, then mod_timer() is the only safe way to modify the timeout,
837 * since add_timer() cannot modify an already running timer.
839 * The function returns whether it has modified a pending timer or not.
840 * (ie. mod_timer() of an inactive timer returns 0, mod_timer() of an
841 * active timer returns 1.)
843 int mod_timer(struct timer_list *timer, unsigned long expires)
845 expires = apply_slack(timer, expires);
848 * This is a common optimization triggered by the
849 * networking code - if the timer is re-modified
850 * to be the same thing then just return:
852 if (timer_pending(timer) && timer->expires == expires)
855 return __mod_timer(timer, expires, false, TIMER_NOT_PINNED);
857 EXPORT_SYMBOL(mod_timer);
860 * mod_timer_pinned - modify a timer's timeout
861 * @timer: the timer to be modified
862 * @expires: new timeout in jiffies
864 * mod_timer_pinned() is a way to update the expire field of an
865 * active timer (if the timer is inactive it will be activated)
866 * and not allow the timer to be migrated to a different CPU.
868 * mod_timer_pinned(timer, expires) is equivalent to:
870 * del_timer(timer); timer->expires = expires; add_timer(timer);
872 int mod_timer_pinned(struct timer_list *timer, unsigned long expires)
874 if (timer->expires == expires && timer_pending(timer))
877 return __mod_timer(timer, expires, false, TIMER_PINNED);
879 EXPORT_SYMBOL(mod_timer_pinned);
882 * add_timer - start a timer
883 * @timer: the timer to be added
885 * The kernel will do a ->function(->data) callback from the
886 * timer interrupt at the ->expires point in the future. The
887 * current time is 'jiffies'.
889 * The timer's ->expires, ->function (and if the handler uses it, ->data)
890 * fields must be set prior calling this function.
892 * Timers with an ->expires field in the past will be executed in the next
895 void add_timer(struct timer_list *timer)
897 BUG_ON(timer_pending(timer));
898 mod_timer(timer, timer->expires);
900 EXPORT_SYMBOL(add_timer);
903 * add_timer_on - start a timer on a particular CPU
904 * @timer: the timer to be added
905 * @cpu: the CPU to start it on
907 * This is not very scalable on SMP. Double adds are not possible.
909 void add_timer_on(struct timer_list *timer, int cpu)
911 struct tvec_base *base = per_cpu(tvec_bases, cpu);
914 timer_stats_timer_set_start_info(timer);
915 BUG_ON(timer_pending(timer) || !timer->function);
916 spin_lock_irqsave(&base->lock, flags);
917 timer_set_base(timer, base);
918 debug_activate(timer, timer->expires);
919 if (time_before(timer->expires, base->next_timer) &&
920 !tbase_get_deferrable(timer->base))
921 base->next_timer = timer->expires;
922 internal_add_timer(base, timer);
924 * Check whether the other CPU is idle and needs to be
925 * triggered to reevaluate the timer wheel when nohz is
926 * active. We are protected against the other CPU fiddling
927 * with the timer by holding the timer base lock. This also
928 * makes sure that a CPU on the way to idle can not evaluate
931 wake_up_idle_cpu(cpu);
932 spin_unlock_irqrestore(&base->lock, flags);
934 EXPORT_SYMBOL_GPL(add_timer_on);
937 * del_timer - deactive a timer.
938 * @timer: the timer to be deactivated
940 * del_timer() deactivates a timer - this works on both active and inactive
943 * The function returns whether it has deactivated a pending timer or not.
944 * (ie. del_timer() of an inactive timer returns 0, del_timer() of an
945 * active timer returns 1.)
947 int del_timer(struct timer_list *timer)
949 struct tvec_base *base;
953 debug_assert_init(timer);
955 timer_stats_timer_clear_start_info(timer);
956 if (timer_pending(timer)) {
957 base = lock_timer_base(timer, &flags);
958 if (timer_pending(timer)) {
959 detach_timer(timer, 1);
960 if (timer->expires == base->next_timer &&
961 !tbase_get_deferrable(timer->base))
962 base->next_timer = base->timer_jiffies;
965 spin_unlock_irqrestore(&base->lock, flags);
970 EXPORT_SYMBOL(del_timer);
973 * try_to_del_timer_sync - Try to deactivate a timer
974 * @timer: timer do del
976 * This function tries to deactivate a timer. Upon successful (ret >= 0)
977 * exit the timer is not queued and the handler is not running on any CPU.
979 int try_to_del_timer_sync(struct timer_list *timer)
981 struct tvec_base *base;
985 debug_assert_init(timer);
987 base = lock_timer_base(timer, &flags);
989 if (base->running_timer == timer)
992 timer_stats_timer_clear_start_info(timer);
994 if (timer_pending(timer)) {
995 detach_timer(timer, 1);
996 if (timer->expires == base->next_timer &&
997 !tbase_get_deferrable(timer->base))
998 base->next_timer = base->timer_jiffies;
1002 spin_unlock_irqrestore(&base->lock, flags);
1006 EXPORT_SYMBOL(try_to_del_timer_sync);
1010 * del_timer_sync - deactivate a timer and wait for the handler to finish.
1011 * @timer: the timer to be deactivated
1013 * This function only differs from del_timer() on SMP: besides deactivating
1014 * the timer it also makes sure the handler has finished executing on other
1017 * Synchronization rules: Callers must prevent restarting of the timer,
1018 * otherwise this function is meaningless. It must not be called from
1019 * interrupt contexts. The caller must not hold locks which would prevent
1020 * completion of the timer's handler. The timer's handler must not call
1021 * add_timer_on(). Upon exit the timer is not queued and the handler is
1022 * not running on any CPU.
1024 * Note: You must not hold locks that are held in interrupt context
1025 * while calling this function. Even if the lock has nothing to do
1026 * with the timer in question. Here's why:
1032 * base->running_timer = mytimer;
1033 * spin_lock_irq(somelock);
1035 * spin_lock(somelock);
1036 * del_timer_sync(mytimer);
1037 * while (base->running_timer == mytimer);
1039 * Now del_timer_sync() will never return and never release somelock.
1040 * The interrupt on the other CPU is waiting to grab somelock but
1041 * it has interrupted the softirq that CPU0 is waiting to finish.
1043 * The function returns whether it has deactivated a pending timer or not.
1045 int del_timer_sync(struct timer_list *timer)
1047 #ifdef CONFIG_LOCKDEP
1048 unsigned long flags;
1051 * If lockdep gives a backtrace here, please reference
1052 * the synchronization rules above.
1054 local_irq_save(flags);
1055 lock_map_acquire(&timer->lockdep_map);
1056 lock_map_release(&timer->lockdep_map);
1057 local_irq_restore(flags);
1060 * don't use it in hardirq context, because it
1061 * could lead to deadlock.
1065 int ret = try_to_del_timer_sync(timer);
1071 EXPORT_SYMBOL(del_timer_sync);
1074 static int cascade(struct tvec_base *base, struct tvec *tv, int index)
1076 /* cascade all the timers from tv up one level */
1077 struct timer_list *timer, *tmp;
1078 struct list_head tv_list;
1080 list_replace_init(tv->vec + index, &tv_list);
1083 * We are removing _all_ timers from the list, so we
1084 * don't have to detach them individually.
1086 list_for_each_entry_safe(timer, tmp, &tv_list, entry) {
1087 BUG_ON(tbase_get_base(timer->base) != base);
1088 internal_add_timer(base, timer);
1094 static void call_timer_fn(struct timer_list *timer, void (*fn)(unsigned long),
1097 int preempt_count = preempt_count();
1099 #ifdef CONFIG_LOCKDEP
1101 * It is permissible to free the timer from inside the
1102 * function that is called from it, this we need to take into
1103 * account for lockdep too. To avoid bogus "held lock freed"
1104 * warnings as well as problems when looking into
1105 * timer->lockdep_map, make a copy and use that here.
1107 struct lockdep_map lockdep_map = timer->lockdep_map;
1110 * Couple the lock chain with the lock chain at
1111 * del_timer_sync() by acquiring the lock_map around the fn()
1112 * call here and in del_timer_sync().
1114 lock_map_acquire(&lockdep_map);
1116 trace_timer_expire_entry(timer);
1118 trace_timer_expire_exit(timer);
1120 lock_map_release(&lockdep_map);
1122 if (preempt_count != preempt_count()) {
1123 WARN_ONCE(1, "timer: %pF preempt leak: %08x -> %08x\n",
1124 fn, preempt_count, preempt_count());
1126 * Restore the preempt count. That gives us a decent
1127 * chance to survive and extract information. If the
1128 * callback kept a lock held, bad luck, but not worse
1129 * than the BUG() we had.
1131 preempt_count() = preempt_count;
1135 #define INDEX(N) ((base->timer_jiffies >> (TVR_BITS + (N) * TVN_BITS)) & TVN_MASK)
1138 * __run_timers - run all expired timers (if any) on this CPU.
1139 * @base: the timer vector to be processed.
1141 * This function cascades all vectors and executes all expired timer
1144 static inline void __run_timers(struct tvec_base *base)
1146 struct timer_list *timer;
1148 spin_lock_irq(&base->lock);
1149 while (time_after_eq(jiffies, base->timer_jiffies)) {
1150 struct list_head work_list;
1151 struct list_head *head = &work_list;
1152 int index = base->timer_jiffies & TVR_MASK;
1158 (!cascade(base, &base->tv2, INDEX(0))) &&
1159 (!cascade(base, &base->tv3, INDEX(1))) &&
1160 !cascade(base, &base->tv4, INDEX(2)))
1161 cascade(base, &base->tv5, INDEX(3));
1162 ++base->timer_jiffies;
1163 list_replace_init(base->tv1.vec + index, &work_list);
1164 while (!list_empty(head)) {
1165 void (*fn)(unsigned long);
1168 timer = list_first_entry(head, struct timer_list,entry);
1169 fn = timer->function;
1172 timer_stats_account_timer(timer);
1174 base->running_timer = timer;
1175 detach_timer(timer, 1);
1177 spin_unlock_irq(&base->lock);
1178 call_timer_fn(timer, fn, data);
1179 spin_lock_irq(&base->lock);
1182 base->running_timer = NULL;
1183 spin_unlock_irq(&base->lock);
1188 * Find out when the next timer event is due to happen. This
1189 * is used on S/390 to stop all activity when a CPU is idle.
1190 * This function needs to be called with interrupts disabled.
1192 static unsigned long __next_timer_interrupt(struct tvec_base *base)
1194 unsigned long timer_jiffies = base->timer_jiffies;
1195 unsigned long expires = timer_jiffies + NEXT_TIMER_MAX_DELTA;
1196 int index, slot, array, found = 0;
1197 struct timer_list *nte;
1198 struct tvec *varray[4];
1200 /* Look for timer events in tv1. */
1201 index = slot = timer_jiffies & TVR_MASK;
1203 list_for_each_entry(nte, base->tv1.vec + slot, entry) {
1204 if (tbase_get_deferrable(nte->base))
1208 expires = nte->expires;
1209 /* Look at the cascade bucket(s)? */
1210 if (!index || slot < index)
1214 slot = (slot + 1) & TVR_MASK;
1215 } while (slot != index);
1218 /* Calculate the next cascade event */
1220 timer_jiffies += TVR_SIZE - index;
1221 timer_jiffies >>= TVR_BITS;
1223 /* Check tv2-tv5. */
1224 varray[0] = &base->tv2;
1225 varray[1] = &base->tv3;
1226 varray[2] = &base->tv4;
1227 varray[3] = &base->tv5;
1229 for (array = 0; array < 4; array++) {
1230 struct tvec *varp = varray[array];
1232 index = slot = timer_jiffies & TVN_MASK;
1234 list_for_each_entry(nte, varp->vec + slot, entry) {
1235 if (tbase_get_deferrable(nte->base))
1239 if (time_before(nte->expires, expires))
1240 expires = nte->expires;
1243 * Do we still search for the first timer or are
1244 * we looking up the cascade buckets ?
1247 /* Look at the cascade bucket(s)? */
1248 if (!index || slot < index)
1252 slot = (slot + 1) & TVN_MASK;
1253 } while (slot != index);
1256 timer_jiffies += TVN_SIZE - index;
1257 timer_jiffies >>= TVN_BITS;
1263 * Check, if the next hrtimer event is before the next timer wheel
1266 static unsigned long cmp_next_hrtimer_event(unsigned long now,
1267 unsigned long expires)
1269 ktime_t hr_delta = hrtimer_get_next_event();
1270 struct timespec tsdelta;
1271 unsigned long delta;
1273 if (hr_delta.tv64 == KTIME_MAX)
1277 * Expired timer available, let it expire in the next tick
1279 if (hr_delta.tv64 <= 0)
1282 tsdelta = ktime_to_timespec(hr_delta);
1283 delta = timespec_to_jiffies(&tsdelta);
1286 * Limit the delta to the max value, which is checked in
1287 * tick_nohz_stop_sched_tick():
1289 if (delta > NEXT_TIMER_MAX_DELTA)
1290 delta = NEXT_TIMER_MAX_DELTA;
1293 * Take rounding errors in to account and make sure, that it
1294 * expires in the next tick. Otherwise we go into an endless
1295 * ping pong due to tick_nohz_stop_sched_tick() retriggering
1301 if (time_before(now, expires))
1307 * get_next_timer_interrupt - return the jiffy of the next pending timer
1308 * @now: current time (in jiffies)
1310 unsigned long get_next_timer_interrupt(unsigned long now)
1312 struct tvec_base *base = __this_cpu_read(tvec_bases);
1313 unsigned long expires;
1316 * Pretend that there is no timer pending if the cpu is offline.
1317 * Possible pending timers will be migrated later to an active cpu.
1319 if (cpu_is_offline(smp_processor_id()))
1320 return now + NEXT_TIMER_MAX_DELTA;
1321 spin_lock(&base->lock);
1322 if (time_before_eq(base->next_timer, base->timer_jiffies))
1323 base->next_timer = __next_timer_interrupt(base);
1324 expires = base->next_timer;
1325 spin_unlock(&base->lock);
1327 if (time_before_eq(expires, now))
1330 return cmp_next_hrtimer_event(now, expires);
1335 * Called from the timer interrupt handler to charge one tick to the current
1336 * process. user_tick is 1 if the tick is user time, 0 for system.
1338 void update_process_times(int user_tick)
1340 struct task_struct *p = current;
1341 int cpu = smp_processor_id();
1343 /* Note: this timer irq context must be accounted for as well. */
1344 account_process_tick(p, user_tick);
1346 rcu_check_callbacks(cpu, user_tick);
1348 #ifdef CONFIG_IRQ_WORK
1353 run_posix_cpu_timers(p);
1357 * This function runs timers and the timer-tq in bottom half context.
1359 static void run_timer_softirq(struct softirq_action *h)
1361 struct tvec_base *base = __this_cpu_read(tvec_bases);
1363 hrtimer_run_pending();
1365 if (time_after_eq(jiffies, base->timer_jiffies))
1370 * Called by the local, per-CPU timer interrupt on SMP.
1372 void run_local_timers(void)
1374 hrtimer_run_queues();
1375 raise_softirq(TIMER_SOFTIRQ);
1378 #ifdef __ARCH_WANT_SYS_ALARM
1381 * For backwards compatibility? This can be done in libc so Alpha
1382 * and all newer ports shouldn't need it.
1384 SYSCALL_DEFINE1(alarm, unsigned int, seconds)
1386 return alarm_setitimer(seconds);
1394 * The Alpha uses getxpid, getxuid, and getxgid instead. Maybe this
1395 * should be moved into arch/i386 instead?
1399 * sys_getpid - return the thread group id of the current process
1401 * Note, despite the name, this returns the tgid not the pid. The tgid and
1402 * the pid are identical unless CLONE_THREAD was specified on clone() in
1403 * which case the tgid is the same in all threads of the same group.
1405 * This is SMP safe as current->tgid does not change.
1407 SYSCALL_DEFINE0(getpid)
1409 return task_tgid_vnr(current);
1413 * Accessing ->real_parent is not SMP-safe, it could
1414 * change from under us. However, we can use a stale
1415 * value of ->real_parent under rcu_read_lock(), see
1416 * release_task()->call_rcu(delayed_put_task_struct).
1418 SYSCALL_DEFINE0(getppid)
1423 pid = task_tgid_vnr(rcu_dereference(current->real_parent));
1429 SYSCALL_DEFINE0(getuid)
1431 /* Only we change this so SMP safe */
1432 return current_uid();
1435 SYSCALL_DEFINE0(geteuid)
1437 /* Only we change this so SMP safe */
1438 return current_euid();
1441 SYSCALL_DEFINE0(getgid)
1443 /* Only we change this so SMP safe */
1444 return current_gid();
1447 SYSCALL_DEFINE0(getegid)
1449 /* Only we change this so SMP safe */
1450 return current_egid();
1455 static void process_timeout(unsigned long __data)
1457 wake_up_process((struct task_struct *)__data);
1461 * schedule_timeout - sleep until timeout
1462 * @timeout: timeout value in jiffies
1464 * Make the current task sleep until @timeout jiffies have
1465 * elapsed. The routine will return immediately unless
1466 * the current task state has been set (see set_current_state()).
1468 * You can set the task state as follows -
1470 * %TASK_UNINTERRUPTIBLE - at least @timeout jiffies are guaranteed to
1471 * pass before the routine returns. The routine will return 0
1473 * %TASK_INTERRUPTIBLE - the routine may return early if a signal is
1474 * delivered to the current task. In this case the remaining time
1475 * in jiffies will be returned, or 0 if the timer expired in time
1477 * The current task state is guaranteed to be TASK_RUNNING when this
1480 * Specifying a @timeout value of %MAX_SCHEDULE_TIMEOUT will schedule
1481 * the CPU away without a bound on the timeout. In this case the return
1482 * value will be %MAX_SCHEDULE_TIMEOUT.
1484 * In all cases the return value is guaranteed to be non-negative.
1486 signed long __sched schedule_timeout(signed long timeout)
1488 struct timer_list timer;
1489 unsigned long expire;
1493 case MAX_SCHEDULE_TIMEOUT:
1495 * These two special cases are useful to be comfortable
1496 * in the caller. Nothing more. We could take
1497 * MAX_SCHEDULE_TIMEOUT from one of the negative value
1498 * but I' d like to return a valid offset (>=0) to allow
1499 * the caller to do everything it want with the retval.
1505 * Another bit of PARANOID. Note that the retval will be
1506 * 0 since no piece of kernel is supposed to do a check
1507 * for a negative retval of schedule_timeout() (since it
1508 * should never happens anyway). You just have the printk()
1509 * that will tell you if something is gone wrong and where.
1512 printk(KERN_ERR "schedule_timeout: wrong timeout "
1513 "value %lx\n", timeout);
1515 current->state = TASK_RUNNING;
1520 expire = timeout + jiffies;
1522 setup_timer_on_stack(&timer, process_timeout, (unsigned long)current);
1523 __mod_timer(&timer, expire, false, TIMER_NOT_PINNED);
1525 del_singleshot_timer_sync(&timer);
1527 /* Remove the timer from the object tracker */
1528 destroy_timer_on_stack(&timer);
1530 timeout = expire - jiffies;
1533 return timeout < 0 ? 0 : timeout;
1535 EXPORT_SYMBOL(schedule_timeout);
1538 * We can use __set_current_state() here because schedule_timeout() calls
1539 * schedule() unconditionally.
1541 signed long __sched schedule_timeout_interruptible(signed long timeout)
1543 __set_current_state(TASK_INTERRUPTIBLE);
1544 return schedule_timeout(timeout);
1546 EXPORT_SYMBOL(schedule_timeout_interruptible);
1548 signed long __sched schedule_timeout_killable(signed long timeout)
1550 __set_current_state(TASK_KILLABLE);
1551 return schedule_timeout(timeout);
1553 EXPORT_SYMBOL(schedule_timeout_killable);
1555 signed long __sched schedule_timeout_uninterruptible(signed long timeout)
1557 __set_current_state(TASK_UNINTERRUPTIBLE);
1558 return schedule_timeout(timeout);
1560 EXPORT_SYMBOL(schedule_timeout_uninterruptible);
1562 /* Thread ID - the internal kernel "pid" */
1563 SYSCALL_DEFINE0(gettid)
1565 return task_pid_vnr(current);
1569 * do_sysinfo - fill in sysinfo struct
1570 * @info: pointer to buffer to fill
1572 int do_sysinfo(struct sysinfo *info)
1574 unsigned long mem_total, sav_total;
1575 unsigned int mem_unit, bitcount;
1578 memset(info, 0, sizeof(struct sysinfo));
1581 monotonic_to_bootbased(&tp);
1582 info->uptime = tp.tv_sec + (tp.tv_nsec ? 1 : 0);
1584 get_avenrun(info->loads, 0, SI_LOAD_SHIFT - FSHIFT);
1586 info->procs = nr_threads;
1592 * If the sum of all the available memory (i.e. ram + swap)
1593 * is less than can be stored in a 32 bit unsigned long then
1594 * we can be binary compatible with 2.2.x kernels. If not,
1595 * well, in that case 2.2.x was broken anyways...
1597 * -Erik Andersen <andersee@debian.org>
1600 mem_total = info->totalram + info->totalswap;
1601 if (mem_total < info->totalram || mem_total < info->totalswap)
1604 mem_unit = info->mem_unit;
1605 while (mem_unit > 1) {
1608 sav_total = mem_total;
1610 if (mem_total < sav_total)
1615 * If mem_total did not overflow, multiply all memory values by
1616 * info->mem_unit and set it to 1. This leaves things compatible
1617 * with 2.2.x, and also retains compatibility with earlier 2.4.x
1622 info->totalram <<= bitcount;
1623 info->freeram <<= bitcount;
1624 info->sharedram <<= bitcount;
1625 info->bufferram <<= bitcount;
1626 info->totalswap <<= bitcount;
1627 info->freeswap <<= bitcount;
1628 info->totalhigh <<= bitcount;
1629 info->freehigh <<= bitcount;
1635 SYSCALL_DEFINE1(sysinfo, struct sysinfo __user *, info)
1641 if (copy_to_user(info, &val, sizeof(struct sysinfo)))
1647 static int __cpuinit init_timers_cpu(int cpu)
1650 struct tvec_base *base;
1651 static char __cpuinitdata tvec_base_done[NR_CPUS];
1653 if (!tvec_base_done[cpu]) {
1654 static char boot_done;
1658 * The APs use this path later in boot
1660 base = kmalloc_node(sizeof(*base),
1661 GFP_KERNEL | __GFP_ZERO,
1666 /* Make sure that tvec_base is 2 byte aligned */
1667 if (tbase_get_deferrable(base)) {
1672 per_cpu(tvec_bases, cpu) = base;
1675 * This is for the boot CPU - we use compile-time
1676 * static initialisation because per-cpu memory isn't
1677 * ready yet and because the memory allocators are not
1678 * initialised either.
1681 base = &boot_tvec_bases;
1683 tvec_base_done[cpu] = 1;
1685 base = per_cpu(tvec_bases, cpu);
1688 spin_lock_init(&base->lock);
1690 for (j = 0; j < TVN_SIZE; j++) {
1691 INIT_LIST_HEAD(base->tv5.vec + j);
1692 INIT_LIST_HEAD(base->tv4.vec + j);
1693 INIT_LIST_HEAD(base->tv3.vec + j);
1694 INIT_LIST_HEAD(base->tv2.vec + j);
1696 for (j = 0; j < TVR_SIZE; j++)
1697 INIT_LIST_HEAD(base->tv1.vec + j);
1699 base->timer_jiffies = jiffies;
1700 base->next_timer = base->timer_jiffies;
1704 #ifdef CONFIG_HOTPLUG_CPU
1705 static void migrate_timer_list(struct tvec_base *new_base, struct list_head *head)
1707 struct timer_list *timer;
1709 while (!list_empty(head)) {
1710 timer = list_first_entry(head, struct timer_list, entry);
1711 detach_timer(timer, 0);
1712 timer_set_base(timer, new_base);
1713 if (time_before(timer->expires, new_base->next_timer) &&
1714 !tbase_get_deferrable(timer->base))
1715 new_base->next_timer = timer->expires;
1716 internal_add_timer(new_base, timer);
1720 static void __cpuinit migrate_timers(int cpu)
1722 struct tvec_base *old_base;
1723 struct tvec_base *new_base;
1726 BUG_ON(cpu_online(cpu));
1727 old_base = per_cpu(tvec_bases, cpu);
1728 new_base = get_cpu_var(tvec_bases);
1730 * The caller is globally serialized and nobody else
1731 * takes two locks at once, deadlock is not possible.
1733 spin_lock_irq(&new_base->lock);
1734 spin_lock_nested(&old_base->lock, SINGLE_DEPTH_NESTING);
1736 BUG_ON(old_base->running_timer);
1738 for (i = 0; i < TVR_SIZE; i++)
1739 migrate_timer_list(new_base, old_base->tv1.vec + i);
1740 for (i = 0; i < TVN_SIZE; i++) {
1741 migrate_timer_list(new_base, old_base->tv2.vec + i);
1742 migrate_timer_list(new_base, old_base->tv3.vec + i);
1743 migrate_timer_list(new_base, old_base->tv4.vec + i);
1744 migrate_timer_list(new_base, old_base->tv5.vec + i);
1747 spin_unlock(&old_base->lock);
1748 spin_unlock_irq(&new_base->lock);
1749 put_cpu_var(tvec_bases);
1751 #endif /* CONFIG_HOTPLUG_CPU */
1753 static int __cpuinit timer_cpu_notify(struct notifier_block *self,
1754 unsigned long action, void *hcpu)
1756 long cpu = (long)hcpu;
1760 case CPU_UP_PREPARE:
1761 case CPU_UP_PREPARE_FROZEN:
1762 err = init_timers_cpu(cpu);
1764 return notifier_from_errno(err);
1766 #ifdef CONFIG_HOTPLUG_CPU
1768 case CPU_DEAD_FROZEN:
1769 migrate_timers(cpu);
1778 static struct notifier_block __cpuinitdata timers_nb = {
1779 .notifier_call = timer_cpu_notify,
1783 void __init init_timers(void)
1785 int err = timer_cpu_notify(&timers_nb, (unsigned long)CPU_UP_PREPARE,
1786 (void *)(long)smp_processor_id());
1790 BUG_ON(err != NOTIFY_OK);
1791 register_cpu_notifier(&timers_nb);
1792 open_softirq(TIMER_SOFTIRQ, run_timer_softirq);
1796 * msleep - sleep safely even with waitqueue interruptions
1797 * @msecs: Time in milliseconds to sleep for
1799 void msleep(unsigned int msecs)
1801 unsigned long timeout = msecs_to_jiffies(msecs) + 1;
1804 timeout = schedule_timeout_uninterruptible(timeout);
1807 EXPORT_SYMBOL(msleep);
1810 * msleep_interruptible - sleep waiting for signals
1811 * @msecs: Time in milliseconds to sleep for
1813 unsigned long msleep_interruptible(unsigned int msecs)
1815 unsigned long timeout = msecs_to_jiffies(msecs) + 1;
1817 while (timeout && !signal_pending(current))
1818 timeout = schedule_timeout_interruptible(timeout);
1819 return jiffies_to_msecs(timeout);
1822 EXPORT_SYMBOL(msleep_interruptible);
1824 static int __sched do_usleep_range(unsigned long min, unsigned long max)
1827 unsigned long delta;
1829 kmin = ktime_set(0, min * NSEC_PER_USEC);
1830 delta = (max - min) * NSEC_PER_USEC;
1831 return schedule_hrtimeout_range(&kmin, delta, HRTIMER_MODE_REL);
1835 * usleep_range - Drop in replacement for udelay where wakeup is flexible
1836 * @min: Minimum time in usecs to sleep
1837 * @max: Maximum time in usecs to sleep
1839 void usleep_range(unsigned long min, unsigned long max)
1841 __set_current_state(TASK_UNINTERRUPTIBLE);
1842 do_usleep_range(min, max);
1844 EXPORT_SYMBOL(usleep_range);