Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * linux/kernel/timer.c | |
3 | * | |
8524070b | 4 | * Kernel internal timers, basic process system calls |
1da177e4 LT |
5 | * |
6 | * Copyright (C) 1991, 1992 Linus Torvalds | |
7 | * | |
8 | * 1997-01-28 Modified by Finn Arne Gangstad to make timers scale better. | |
9 | * | |
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 | |
20 | */ | |
21 | ||
22 | #include <linux/kernel_stat.h> | |
23 | #include <linux/module.h> | |
24 | #include <linux/interrupt.h> | |
25 | #include <linux/percpu.h> | |
26 | #include <linux/init.h> | |
27 | #include <linux/mm.h> | |
28 | #include <linux/swap.h> | |
b488893a | 29 | #include <linux/pid_namespace.h> |
1da177e4 LT |
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> | |
97a41e26 | 37 | #include <linux/delay.h> |
79bf2bb3 | 38 | #include <linux/tick.h> |
82f67cd9 | 39 | #include <linux/kallsyms.h> |
925d519a | 40 | #include <linux/perf_counter.h> |
1da177e4 LT |
41 | |
42 | #include <asm/uaccess.h> | |
43 | #include <asm/unistd.h> | |
44 | #include <asm/div64.h> | |
45 | #include <asm/timex.h> | |
46 | #include <asm/io.h> | |
47 | ||
ecea8d19 TG |
48 | u64 jiffies_64 __cacheline_aligned_in_smp = INITIAL_JIFFIES; |
49 | ||
50 | EXPORT_SYMBOL(jiffies_64); | |
51 | ||
1da177e4 LT |
52 | /* |
53 | * per-CPU timer vector definitions: | |
54 | */ | |
1da177e4 LT |
55 | #define TVN_BITS (CONFIG_BASE_SMALL ? 4 : 6) |
56 | #define TVR_BITS (CONFIG_BASE_SMALL ? 6 : 8) | |
57 | #define TVN_SIZE (1 << TVN_BITS) | |
58 | #define TVR_SIZE (1 << TVR_BITS) | |
59 | #define TVN_MASK (TVN_SIZE - 1) | |
60 | #define TVR_MASK (TVR_SIZE - 1) | |
61 | ||
a6fa8e5a | 62 | struct tvec { |
1da177e4 | 63 | struct list_head vec[TVN_SIZE]; |
a6fa8e5a | 64 | }; |
1da177e4 | 65 | |
a6fa8e5a | 66 | struct tvec_root { |
1da177e4 | 67 | struct list_head vec[TVR_SIZE]; |
a6fa8e5a | 68 | }; |
1da177e4 | 69 | |
a6fa8e5a | 70 | struct tvec_base { |
3691c519 ON |
71 | spinlock_t lock; |
72 | struct timer_list *running_timer; | |
1da177e4 | 73 | unsigned long timer_jiffies; |
a6fa8e5a PM |
74 | struct tvec_root tv1; |
75 | struct tvec tv2; | |
76 | struct tvec tv3; | |
77 | struct tvec tv4; | |
78 | struct tvec tv5; | |
6e453a67 | 79 | } ____cacheline_aligned; |
1da177e4 | 80 | |
a6fa8e5a | 81 | struct tvec_base boot_tvec_bases; |
3691c519 | 82 | EXPORT_SYMBOL(boot_tvec_bases); |
a6fa8e5a | 83 | static DEFINE_PER_CPU(struct tvec_base *, tvec_bases) = &boot_tvec_bases; |
1da177e4 | 84 | |
6e453a67 | 85 | /* |
a6fa8e5a | 86 | * Note that all tvec_bases are 2 byte aligned and lower bit of |
6e453a67 VP |
87 | * base in timer_list is guaranteed to be zero. Use the LSB for |
88 | * the new flag to indicate whether the timer is deferrable | |
89 | */ | |
90 | #define TBASE_DEFERRABLE_FLAG (0x1) | |
91 | ||
92 | /* Functions below help us manage 'deferrable' flag */ | |
a6fa8e5a | 93 | static inline unsigned int tbase_get_deferrable(struct tvec_base *base) |
6e453a67 | 94 | { |
e9910846 | 95 | return ((unsigned int)(unsigned long)base & TBASE_DEFERRABLE_FLAG); |
6e453a67 VP |
96 | } |
97 | ||
a6fa8e5a | 98 | static inline struct tvec_base *tbase_get_base(struct tvec_base *base) |
6e453a67 | 99 | { |
a6fa8e5a | 100 | return ((struct tvec_base *)((unsigned long)base & ~TBASE_DEFERRABLE_FLAG)); |
6e453a67 VP |
101 | } |
102 | ||
103 | static inline void timer_set_deferrable(struct timer_list *timer) | |
104 | { | |
a6fa8e5a | 105 | timer->base = ((struct tvec_base *)((unsigned long)(timer->base) | |
6819457d | 106 | TBASE_DEFERRABLE_FLAG)); |
6e453a67 VP |
107 | } |
108 | ||
109 | static inline void | |
a6fa8e5a | 110 | timer_set_base(struct timer_list *timer, struct tvec_base *new_base) |
6e453a67 | 111 | { |
a6fa8e5a | 112 | timer->base = (struct tvec_base *)((unsigned long)(new_base) | |
6819457d | 113 | tbase_get_deferrable(timer->base)); |
6e453a67 VP |
114 | } |
115 | ||
9c133c46 AS |
116 | static unsigned long round_jiffies_common(unsigned long j, int cpu, |
117 | bool force_up) | |
4c36a5de AV |
118 | { |
119 | int rem; | |
120 | unsigned long original = j; | |
121 | ||
122 | /* | |
123 | * We don't want all cpus firing their timers at once hitting the | |
124 | * same lock or cachelines, so we skew each extra cpu with an extra | |
125 | * 3 jiffies. This 3 jiffies came originally from the mm/ code which | |
126 | * already did this. | |
127 | * The skew is done by adding 3*cpunr, then round, then subtract this | |
128 | * extra offset again. | |
129 | */ | |
130 | j += cpu * 3; | |
131 | ||
132 | rem = j % HZ; | |
133 | ||
134 | /* | |
135 | * If the target jiffie is just after a whole second (which can happen | |
136 | * due to delays of the timer irq, long irq off times etc etc) then | |
137 | * we should round down to the whole second, not up. Use 1/4th second | |
138 | * as cutoff for this rounding as an extreme upper bound for this. | |
9c133c46 | 139 | * But never round down if @force_up is set. |
4c36a5de | 140 | */ |
9c133c46 | 141 | if (rem < HZ/4 && !force_up) /* round down */ |
4c36a5de AV |
142 | j = j - rem; |
143 | else /* round up */ | |
144 | j = j - rem + HZ; | |
145 | ||
146 | /* now that we have rounded, subtract the extra skew again */ | |
147 | j -= cpu * 3; | |
148 | ||
149 | if (j <= jiffies) /* rounding ate our timeout entirely; */ | |
150 | return original; | |
151 | return j; | |
152 | } | |
9c133c46 AS |
153 | |
154 | /** | |
155 | * __round_jiffies - function to round jiffies to a full second | |
156 | * @j: the time in (absolute) jiffies that should be rounded | |
157 | * @cpu: the processor number on which the timeout will happen | |
158 | * | |
159 | * __round_jiffies() rounds an absolute time in the future (in jiffies) | |
160 | * up or down to (approximately) full seconds. This is useful for timers | |
161 | * for which the exact time they fire does not matter too much, as long as | |
162 | * they fire approximately every X seconds. | |
163 | * | |
164 | * By rounding these timers to whole seconds, all such timers will fire | |
165 | * at the same time, rather than at various times spread out. The goal | |
166 | * of this is to have the CPU wake up less, which saves power. | |
167 | * | |
168 | * The exact rounding is skewed for each processor to avoid all | |
169 | * processors firing at the exact same time, which could lead | |
170 | * to lock contention or spurious cache line bouncing. | |
171 | * | |
172 | * The return value is the rounded version of the @j parameter. | |
173 | */ | |
174 | unsigned long __round_jiffies(unsigned long j, int cpu) | |
175 | { | |
176 | return round_jiffies_common(j, cpu, false); | |
177 | } | |
4c36a5de AV |
178 | EXPORT_SYMBOL_GPL(__round_jiffies); |
179 | ||
180 | /** | |
181 | * __round_jiffies_relative - function to round jiffies to a full second | |
182 | * @j: the time in (relative) jiffies that should be rounded | |
183 | * @cpu: the processor number on which the timeout will happen | |
184 | * | |
72fd4a35 | 185 | * __round_jiffies_relative() rounds a time delta in the future (in jiffies) |
4c36a5de AV |
186 | * up or down to (approximately) full seconds. This is useful for timers |
187 | * for which the exact time they fire does not matter too much, as long as | |
188 | * they fire approximately every X seconds. | |
189 | * | |
190 | * By rounding these timers to whole seconds, all such timers will fire | |
191 | * at the same time, rather than at various times spread out. The goal | |
192 | * of this is to have the CPU wake up less, which saves power. | |
193 | * | |
194 | * The exact rounding is skewed for each processor to avoid all | |
195 | * processors firing at the exact same time, which could lead | |
196 | * to lock contention or spurious cache line bouncing. | |
197 | * | |
72fd4a35 | 198 | * The return value is the rounded version of the @j parameter. |
4c36a5de AV |
199 | */ |
200 | unsigned long __round_jiffies_relative(unsigned long j, int cpu) | |
201 | { | |
9c133c46 AS |
202 | unsigned long j0 = jiffies; |
203 | ||
204 | /* Use j0 because jiffies might change while we run */ | |
205 | return round_jiffies_common(j + j0, cpu, false) - j0; | |
4c36a5de AV |
206 | } |
207 | EXPORT_SYMBOL_GPL(__round_jiffies_relative); | |
208 | ||
209 | /** | |
210 | * round_jiffies - function to round jiffies to a full second | |
211 | * @j: the time in (absolute) jiffies that should be rounded | |
212 | * | |
72fd4a35 | 213 | * round_jiffies() rounds an absolute time in the future (in jiffies) |
4c36a5de AV |
214 | * up or down to (approximately) full seconds. This is useful for timers |
215 | * for which the exact time they fire does not matter too much, as long as | |
216 | * they fire approximately every X seconds. | |
217 | * | |
218 | * By rounding these timers to whole seconds, all such timers will fire | |
219 | * at the same time, rather than at various times spread out. The goal | |
220 | * of this is to have the CPU wake up less, which saves power. | |
221 | * | |
72fd4a35 | 222 | * The return value is the rounded version of the @j parameter. |
4c36a5de AV |
223 | */ |
224 | unsigned long round_jiffies(unsigned long j) | |
225 | { | |
9c133c46 | 226 | return round_jiffies_common(j, raw_smp_processor_id(), false); |
4c36a5de AV |
227 | } |
228 | EXPORT_SYMBOL_GPL(round_jiffies); | |
229 | ||
230 | /** | |
231 | * round_jiffies_relative - function to round jiffies to a full second | |
232 | * @j: the time in (relative) jiffies that should be rounded | |
233 | * | |
72fd4a35 | 234 | * round_jiffies_relative() rounds a time delta in the future (in jiffies) |
4c36a5de AV |
235 | * up or down to (approximately) full seconds. This is useful for timers |
236 | * for which the exact time they fire does not matter too much, as long as | |
237 | * they fire approximately every X seconds. | |
238 | * | |
239 | * By rounding these timers to whole seconds, all such timers will fire | |
240 | * at the same time, rather than at various times spread out. The goal | |
241 | * of this is to have the CPU wake up less, which saves power. | |
242 | * | |
72fd4a35 | 243 | * The return value is the rounded version of the @j parameter. |
4c36a5de AV |
244 | */ |
245 | unsigned long round_jiffies_relative(unsigned long j) | |
246 | { | |
247 | return __round_jiffies_relative(j, raw_smp_processor_id()); | |
248 | } | |
249 | EXPORT_SYMBOL_GPL(round_jiffies_relative); | |
250 | ||
9c133c46 AS |
251 | /** |
252 | * __round_jiffies_up - function to round jiffies up to a full second | |
253 | * @j: the time in (absolute) jiffies that should be rounded | |
254 | * @cpu: the processor number on which the timeout will happen | |
255 | * | |
256 | * This is the same as __round_jiffies() except that it will never | |
257 | * round down. This is useful for timeouts for which the exact time | |
258 | * of firing does not matter too much, as long as they don't fire too | |
259 | * early. | |
260 | */ | |
261 | unsigned long __round_jiffies_up(unsigned long j, int cpu) | |
262 | { | |
263 | return round_jiffies_common(j, cpu, true); | |
264 | } | |
265 | EXPORT_SYMBOL_GPL(__round_jiffies_up); | |
266 | ||
267 | /** | |
268 | * __round_jiffies_up_relative - function to round jiffies up to a full second | |
269 | * @j: the time in (relative) jiffies that should be rounded | |
270 | * @cpu: the processor number on which the timeout will happen | |
271 | * | |
272 | * This is the same as __round_jiffies_relative() except that it will never | |
273 | * round down. This is useful for timeouts for which the exact time | |
274 | * of firing does not matter too much, as long as they don't fire too | |
275 | * early. | |
276 | */ | |
277 | unsigned long __round_jiffies_up_relative(unsigned long j, int cpu) | |
278 | { | |
279 | unsigned long j0 = jiffies; | |
280 | ||
281 | /* Use j0 because jiffies might change while we run */ | |
282 | return round_jiffies_common(j + j0, cpu, true) - j0; | |
283 | } | |
284 | EXPORT_SYMBOL_GPL(__round_jiffies_up_relative); | |
285 | ||
286 | /** | |
287 | * round_jiffies_up - function to round jiffies up to a full second | |
288 | * @j: the time in (absolute) jiffies that should be rounded | |
289 | * | |
290 | * This is the same as round_jiffies() except that it will never | |
291 | * round down. This is useful for timeouts for which the exact time | |
292 | * of firing does not matter too much, as long as they don't fire too | |
293 | * early. | |
294 | */ | |
295 | unsigned long round_jiffies_up(unsigned long j) | |
296 | { | |
297 | return round_jiffies_common(j, raw_smp_processor_id(), true); | |
298 | } | |
299 | EXPORT_SYMBOL_GPL(round_jiffies_up); | |
300 | ||
301 | /** | |
302 | * round_jiffies_up_relative - function to round jiffies up to a full second | |
303 | * @j: the time in (relative) jiffies that should be rounded | |
304 | * | |
305 | * This is the same as round_jiffies_relative() except that it will never | |
306 | * round down. This is useful for timeouts for which the exact time | |
307 | * of firing does not matter too much, as long as they don't fire too | |
308 | * early. | |
309 | */ | |
310 | unsigned long round_jiffies_up_relative(unsigned long j) | |
311 | { | |
312 | return __round_jiffies_up_relative(j, raw_smp_processor_id()); | |
313 | } | |
314 | EXPORT_SYMBOL_GPL(round_jiffies_up_relative); | |
315 | ||
4c36a5de | 316 | |
a6fa8e5a | 317 | static inline void set_running_timer(struct tvec_base *base, |
1da177e4 LT |
318 | struct timer_list *timer) |
319 | { | |
320 | #ifdef CONFIG_SMP | |
3691c519 | 321 | base->running_timer = timer; |
1da177e4 LT |
322 | #endif |
323 | } | |
324 | ||
a6fa8e5a | 325 | static void internal_add_timer(struct tvec_base *base, struct timer_list *timer) |
1da177e4 LT |
326 | { |
327 | unsigned long expires = timer->expires; | |
328 | unsigned long idx = expires - base->timer_jiffies; | |
329 | struct list_head *vec; | |
330 | ||
331 | if (idx < TVR_SIZE) { | |
332 | int i = expires & TVR_MASK; | |
333 | vec = base->tv1.vec + i; | |
334 | } else if (idx < 1 << (TVR_BITS + TVN_BITS)) { | |
335 | int i = (expires >> TVR_BITS) & TVN_MASK; | |
336 | vec = base->tv2.vec + i; | |
337 | } else if (idx < 1 << (TVR_BITS + 2 * TVN_BITS)) { | |
338 | int i = (expires >> (TVR_BITS + TVN_BITS)) & TVN_MASK; | |
339 | vec = base->tv3.vec + i; | |
340 | } else if (idx < 1 << (TVR_BITS + 3 * TVN_BITS)) { | |
341 | int i = (expires >> (TVR_BITS + 2 * TVN_BITS)) & TVN_MASK; | |
342 | vec = base->tv4.vec + i; | |
343 | } else if ((signed long) idx < 0) { | |
344 | /* | |
345 | * Can happen if you add a timer with expires == jiffies, | |
346 | * or you set a timer to go off in the past | |
347 | */ | |
348 | vec = base->tv1.vec + (base->timer_jiffies & TVR_MASK); | |
349 | } else { | |
350 | int i; | |
351 | /* If the timeout is larger than 0xffffffff on 64-bit | |
352 | * architectures then we use the maximum timeout: | |
353 | */ | |
354 | if (idx > 0xffffffffUL) { | |
355 | idx = 0xffffffffUL; | |
356 | expires = idx + base->timer_jiffies; | |
357 | } | |
358 | i = (expires >> (TVR_BITS + 3 * TVN_BITS)) & TVN_MASK; | |
359 | vec = base->tv5.vec + i; | |
360 | } | |
361 | /* | |
362 | * Timers are FIFO: | |
363 | */ | |
364 | list_add_tail(&timer->entry, vec); | |
365 | } | |
366 | ||
82f67cd9 IM |
367 | #ifdef CONFIG_TIMER_STATS |
368 | void __timer_stats_timer_set_start_info(struct timer_list *timer, void *addr) | |
369 | { | |
370 | if (timer->start_site) | |
371 | return; | |
372 | ||
373 | timer->start_site = addr; | |
374 | memcpy(timer->start_comm, current->comm, TASK_COMM_LEN); | |
375 | timer->start_pid = current->pid; | |
376 | } | |
c5c061b8 VP |
377 | |
378 | static void timer_stats_account_timer(struct timer_list *timer) | |
379 | { | |
380 | unsigned int flag = 0; | |
381 | ||
382 | if (unlikely(tbase_get_deferrable(timer->base))) | |
383 | flag |= TIMER_STATS_FLAG_DEFERRABLE; | |
384 | ||
385 | timer_stats_update_stats(timer, timer->start_pid, timer->start_site, | |
386 | timer->function, timer->start_comm, flag); | |
387 | } | |
388 | ||
389 | #else | |
390 | static void timer_stats_account_timer(struct timer_list *timer) {} | |
82f67cd9 IM |
391 | #endif |
392 | ||
c6f3a97f TG |
393 | #ifdef CONFIG_DEBUG_OBJECTS_TIMERS |
394 | ||
395 | static struct debug_obj_descr timer_debug_descr; | |
396 | ||
397 | /* | |
398 | * fixup_init is called when: | |
399 | * - an active object is initialized | |
55c888d6 | 400 | */ |
c6f3a97f TG |
401 | static int timer_fixup_init(void *addr, enum debug_obj_state state) |
402 | { | |
403 | struct timer_list *timer = addr; | |
404 | ||
405 | switch (state) { | |
406 | case ODEBUG_STATE_ACTIVE: | |
407 | del_timer_sync(timer); | |
408 | debug_object_init(timer, &timer_debug_descr); | |
409 | return 1; | |
410 | default: | |
411 | return 0; | |
412 | } | |
413 | } | |
414 | ||
415 | /* | |
416 | * fixup_activate is called when: | |
417 | * - an active object is activated | |
418 | * - an unknown object is activated (might be a statically initialized object) | |
419 | */ | |
420 | static int timer_fixup_activate(void *addr, enum debug_obj_state state) | |
421 | { | |
422 | struct timer_list *timer = addr; | |
423 | ||
424 | switch (state) { | |
425 | ||
426 | case ODEBUG_STATE_NOTAVAILABLE: | |
427 | /* | |
428 | * This is not really a fixup. The timer was | |
429 | * statically initialized. We just make sure that it | |
430 | * is tracked in the object tracker. | |
431 | */ | |
432 | if (timer->entry.next == NULL && | |
433 | timer->entry.prev == TIMER_ENTRY_STATIC) { | |
434 | debug_object_init(timer, &timer_debug_descr); | |
435 | debug_object_activate(timer, &timer_debug_descr); | |
436 | return 0; | |
437 | } else { | |
438 | WARN_ON_ONCE(1); | |
439 | } | |
440 | return 0; | |
441 | ||
442 | case ODEBUG_STATE_ACTIVE: | |
443 | WARN_ON(1); | |
444 | ||
445 | default: | |
446 | return 0; | |
447 | } | |
448 | } | |
449 | ||
450 | /* | |
451 | * fixup_free is called when: | |
452 | * - an active object is freed | |
453 | */ | |
454 | static int timer_fixup_free(void *addr, enum debug_obj_state state) | |
455 | { | |
456 | struct timer_list *timer = addr; | |
457 | ||
458 | switch (state) { | |
459 | case ODEBUG_STATE_ACTIVE: | |
460 | del_timer_sync(timer); | |
461 | debug_object_free(timer, &timer_debug_descr); | |
462 | return 1; | |
463 | default: | |
464 | return 0; | |
465 | } | |
466 | } | |
467 | ||
468 | static struct debug_obj_descr timer_debug_descr = { | |
469 | .name = "timer_list", | |
470 | .fixup_init = timer_fixup_init, | |
471 | .fixup_activate = timer_fixup_activate, | |
472 | .fixup_free = timer_fixup_free, | |
473 | }; | |
474 | ||
475 | static inline void debug_timer_init(struct timer_list *timer) | |
476 | { | |
477 | debug_object_init(timer, &timer_debug_descr); | |
478 | } | |
479 | ||
480 | static inline void debug_timer_activate(struct timer_list *timer) | |
481 | { | |
482 | debug_object_activate(timer, &timer_debug_descr); | |
483 | } | |
484 | ||
485 | static inline void debug_timer_deactivate(struct timer_list *timer) | |
486 | { | |
487 | debug_object_deactivate(timer, &timer_debug_descr); | |
488 | } | |
489 | ||
490 | static inline void debug_timer_free(struct timer_list *timer) | |
491 | { | |
492 | debug_object_free(timer, &timer_debug_descr); | |
493 | } | |
494 | ||
6f2b9b9a JB |
495 | static void __init_timer(struct timer_list *timer, |
496 | const char *name, | |
497 | struct lock_class_key *key); | |
c6f3a97f | 498 | |
6f2b9b9a JB |
499 | void init_timer_on_stack_key(struct timer_list *timer, |
500 | const char *name, | |
501 | struct lock_class_key *key) | |
c6f3a97f TG |
502 | { |
503 | debug_object_init_on_stack(timer, &timer_debug_descr); | |
6f2b9b9a | 504 | __init_timer(timer, name, key); |
c6f3a97f | 505 | } |
6f2b9b9a | 506 | EXPORT_SYMBOL_GPL(init_timer_on_stack_key); |
c6f3a97f TG |
507 | |
508 | void destroy_timer_on_stack(struct timer_list *timer) | |
509 | { | |
510 | debug_object_free(timer, &timer_debug_descr); | |
511 | } | |
512 | EXPORT_SYMBOL_GPL(destroy_timer_on_stack); | |
513 | ||
514 | #else | |
515 | static inline void debug_timer_init(struct timer_list *timer) { } | |
516 | static inline void debug_timer_activate(struct timer_list *timer) { } | |
517 | static inline void debug_timer_deactivate(struct timer_list *timer) { } | |
518 | #endif | |
519 | ||
6f2b9b9a JB |
520 | static void __init_timer(struct timer_list *timer, |
521 | const char *name, | |
522 | struct lock_class_key *key) | |
55c888d6 ON |
523 | { |
524 | timer->entry.next = NULL; | |
bfe5d834 | 525 | timer->base = __raw_get_cpu_var(tvec_bases); |
82f67cd9 IM |
526 | #ifdef CONFIG_TIMER_STATS |
527 | timer->start_site = NULL; | |
528 | timer->start_pid = -1; | |
529 | memset(timer->start_comm, 0, TASK_COMM_LEN); | |
530 | #endif | |
6f2b9b9a | 531 | lockdep_init_map(&timer->lockdep_map, name, key, 0); |
55c888d6 | 532 | } |
c6f3a97f TG |
533 | |
534 | /** | |
633fe795 | 535 | * init_timer_key - initialize a timer |
c6f3a97f | 536 | * @timer: the timer to be initialized |
633fe795 RD |
537 | * @name: name of the timer |
538 | * @key: lockdep class key of the fake lock used for tracking timer | |
539 | * sync lock dependencies | |
c6f3a97f | 540 | * |
633fe795 | 541 | * init_timer_key() must be done to a timer prior calling *any* of the |
c6f3a97f TG |
542 | * other timer functions. |
543 | */ | |
6f2b9b9a JB |
544 | void init_timer_key(struct timer_list *timer, |
545 | const char *name, | |
546 | struct lock_class_key *key) | |
c6f3a97f TG |
547 | { |
548 | debug_timer_init(timer); | |
6f2b9b9a | 549 | __init_timer(timer, name, key); |
c6f3a97f | 550 | } |
6f2b9b9a | 551 | EXPORT_SYMBOL(init_timer_key); |
55c888d6 | 552 | |
6f2b9b9a JB |
553 | void init_timer_deferrable_key(struct timer_list *timer, |
554 | const char *name, | |
555 | struct lock_class_key *key) | |
6e453a67 | 556 | { |
6f2b9b9a | 557 | init_timer_key(timer, name, key); |
6e453a67 VP |
558 | timer_set_deferrable(timer); |
559 | } | |
6f2b9b9a | 560 | EXPORT_SYMBOL(init_timer_deferrable_key); |
6e453a67 | 561 | |
55c888d6 | 562 | static inline void detach_timer(struct timer_list *timer, |
82f67cd9 | 563 | int clear_pending) |
55c888d6 ON |
564 | { |
565 | struct list_head *entry = &timer->entry; | |
566 | ||
c6f3a97f TG |
567 | debug_timer_deactivate(timer); |
568 | ||
55c888d6 ON |
569 | __list_del(entry->prev, entry->next); |
570 | if (clear_pending) | |
571 | entry->next = NULL; | |
572 | entry->prev = LIST_POISON2; | |
573 | } | |
574 | ||
575 | /* | |
3691c519 | 576 | * We are using hashed locking: holding per_cpu(tvec_bases).lock |
55c888d6 ON |
577 | * means that all timers which are tied to this base via timer->base are |
578 | * locked, and the base itself is locked too. | |
579 | * | |
580 | * So __run_timers/migrate_timers can safely modify all timers which could | |
581 | * be found on ->tvX lists. | |
582 | * | |
583 | * When the timer's base is locked, and the timer removed from list, it is | |
584 | * possible to set timer->base = NULL and drop the lock: the timer remains | |
585 | * locked. | |
586 | */ | |
a6fa8e5a | 587 | static struct tvec_base *lock_timer_base(struct timer_list *timer, |
55c888d6 | 588 | unsigned long *flags) |
89e7e374 | 589 | __acquires(timer->base->lock) |
55c888d6 | 590 | { |
a6fa8e5a | 591 | struct tvec_base *base; |
55c888d6 ON |
592 | |
593 | for (;;) { | |
a6fa8e5a | 594 | struct tvec_base *prelock_base = timer->base; |
6e453a67 | 595 | base = tbase_get_base(prelock_base); |
55c888d6 ON |
596 | if (likely(base != NULL)) { |
597 | spin_lock_irqsave(&base->lock, *flags); | |
6e453a67 | 598 | if (likely(prelock_base == timer->base)) |
55c888d6 ON |
599 | return base; |
600 | /* The timer has migrated to another CPU */ | |
601 | spin_unlock_irqrestore(&base->lock, *flags); | |
602 | } | |
603 | cpu_relax(); | |
604 | } | |
605 | } | |
606 | ||
74019224 IM |
607 | static inline int |
608 | __mod_timer(struct timer_list *timer, unsigned long expires, bool pending_only) | |
1da177e4 | 609 | { |
a6fa8e5a | 610 | struct tvec_base *base, *new_base; |
1da177e4 | 611 | unsigned long flags; |
74019224 IM |
612 | int ret; |
613 | ||
614 | ret = 0; | |
1da177e4 | 615 | |
82f67cd9 | 616 | timer_stats_timer_set_start_info(timer); |
1da177e4 | 617 | BUG_ON(!timer->function); |
1da177e4 | 618 | |
55c888d6 ON |
619 | base = lock_timer_base(timer, &flags); |
620 | ||
621 | if (timer_pending(timer)) { | |
622 | detach_timer(timer, 0); | |
623 | ret = 1; | |
74019224 IM |
624 | } else { |
625 | if (pending_only) | |
626 | goto out_unlock; | |
55c888d6 ON |
627 | } |
628 | ||
c6f3a97f TG |
629 | debug_timer_activate(timer); |
630 | ||
a4a6198b | 631 | new_base = __get_cpu_var(tvec_bases); |
1da177e4 | 632 | |
3691c519 | 633 | if (base != new_base) { |
1da177e4 | 634 | /* |
55c888d6 ON |
635 | * We are trying to schedule the timer on the local CPU. |
636 | * However we can't change timer's base while it is running, | |
637 | * otherwise del_timer_sync() can't detect that the timer's | |
638 | * handler yet has not finished. This also guarantees that | |
639 | * the timer is serialized wrt itself. | |
1da177e4 | 640 | */ |
a2c348fe | 641 | if (likely(base->running_timer != timer)) { |
55c888d6 | 642 | /* See the comment in lock_timer_base() */ |
6e453a67 | 643 | timer_set_base(timer, NULL); |
55c888d6 | 644 | spin_unlock(&base->lock); |
a2c348fe ON |
645 | base = new_base; |
646 | spin_lock(&base->lock); | |
6e453a67 | 647 | timer_set_base(timer, base); |
1da177e4 LT |
648 | } |
649 | } | |
650 | ||
1da177e4 | 651 | timer->expires = expires; |
a2c348fe | 652 | internal_add_timer(base, timer); |
74019224 IM |
653 | |
654 | out_unlock: | |
a2c348fe | 655 | spin_unlock_irqrestore(&base->lock, flags); |
1da177e4 LT |
656 | |
657 | return ret; | |
658 | } | |
659 | ||
2aae4a10 | 660 | /** |
74019224 IM |
661 | * mod_timer_pending - modify a pending timer's timeout |
662 | * @timer: the pending timer to be modified | |
663 | * @expires: new timeout in jiffies | |
1da177e4 | 664 | * |
74019224 IM |
665 | * mod_timer_pending() is the same for pending timers as mod_timer(), |
666 | * but will not re-activate and modify already deleted timers. | |
667 | * | |
668 | * It is useful for unserialized use of timers. | |
1da177e4 | 669 | */ |
74019224 | 670 | int mod_timer_pending(struct timer_list *timer, unsigned long expires) |
1da177e4 | 671 | { |
74019224 | 672 | return __mod_timer(timer, expires, true); |
1da177e4 | 673 | } |
74019224 | 674 | EXPORT_SYMBOL(mod_timer_pending); |
1da177e4 | 675 | |
2aae4a10 | 676 | /** |
1da177e4 LT |
677 | * mod_timer - modify a timer's timeout |
678 | * @timer: the timer to be modified | |
2aae4a10 | 679 | * @expires: new timeout in jiffies |
1da177e4 | 680 | * |
72fd4a35 | 681 | * mod_timer() is a more efficient way to update the expire field of an |
1da177e4 LT |
682 | * active timer (if the timer is inactive it will be activated) |
683 | * | |
684 | * mod_timer(timer, expires) is equivalent to: | |
685 | * | |
686 | * del_timer(timer); timer->expires = expires; add_timer(timer); | |
687 | * | |
688 | * Note that if there are multiple unserialized concurrent users of the | |
689 | * same timer, then mod_timer() is the only safe way to modify the timeout, | |
690 | * since add_timer() cannot modify an already running timer. | |
691 | * | |
692 | * The function returns whether it has modified a pending timer or not. | |
693 | * (ie. mod_timer() of an inactive timer returns 0, mod_timer() of an | |
694 | * active timer returns 1.) | |
695 | */ | |
696 | int mod_timer(struct timer_list *timer, unsigned long expires) | |
697 | { | |
1da177e4 LT |
698 | /* |
699 | * This is a common optimization triggered by the | |
700 | * networking code - if the timer is re-modified | |
701 | * to be the same thing then just return: | |
702 | */ | |
703 | if (timer->expires == expires && timer_pending(timer)) | |
704 | return 1; | |
705 | ||
74019224 | 706 | return __mod_timer(timer, expires, false); |
1da177e4 | 707 | } |
1da177e4 LT |
708 | EXPORT_SYMBOL(mod_timer); |
709 | ||
74019224 IM |
710 | /** |
711 | * add_timer - start a timer | |
712 | * @timer: the timer to be added | |
713 | * | |
714 | * The kernel will do a ->function(->data) callback from the | |
715 | * timer interrupt at the ->expires point in the future. The | |
716 | * current time is 'jiffies'. | |
717 | * | |
718 | * The timer's ->expires, ->function (and if the handler uses it, ->data) | |
719 | * fields must be set prior calling this function. | |
720 | * | |
721 | * Timers with an ->expires field in the past will be executed in the next | |
722 | * timer tick. | |
723 | */ | |
724 | void add_timer(struct timer_list *timer) | |
725 | { | |
726 | BUG_ON(timer_pending(timer)); | |
727 | mod_timer(timer, timer->expires); | |
728 | } | |
729 | EXPORT_SYMBOL(add_timer); | |
730 | ||
731 | /** | |
732 | * add_timer_on - start a timer on a particular CPU | |
733 | * @timer: the timer to be added | |
734 | * @cpu: the CPU to start it on | |
735 | * | |
736 | * This is not very scalable on SMP. Double adds are not possible. | |
737 | */ | |
738 | void add_timer_on(struct timer_list *timer, int cpu) | |
739 | { | |
740 | struct tvec_base *base = per_cpu(tvec_bases, cpu); | |
741 | unsigned long flags; | |
742 | ||
743 | timer_stats_timer_set_start_info(timer); | |
744 | BUG_ON(timer_pending(timer) || !timer->function); | |
745 | spin_lock_irqsave(&base->lock, flags); | |
746 | timer_set_base(timer, base); | |
747 | debug_timer_activate(timer); | |
748 | internal_add_timer(base, timer); | |
749 | /* | |
750 | * Check whether the other CPU is idle and needs to be | |
751 | * triggered to reevaluate the timer wheel when nohz is | |
752 | * active. We are protected against the other CPU fiddling | |
753 | * with the timer by holding the timer base lock. This also | |
754 | * makes sure that a CPU on the way to idle can not evaluate | |
755 | * the timer wheel. | |
756 | */ | |
757 | wake_up_idle_cpu(cpu); | |
758 | spin_unlock_irqrestore(&base->lock, flags); | |
759 | } | |
760 | ||
2aae4a10 | 761 | /** |
1da177e4 LT |
762 | * del_timer - deactive a timer. |
763 | * @timer: the timer to be deactivated | |
764 | * | |
765 | * del_timer() deactivates a timer - this works on both active and inactive | |
766 | * timers. | |
767 | * | |
768 | * The function returns whether it has deactivated a pending timer or not. | |
769 | * (ie. del_timer() of an inactive timer returns 0, del_timer() of an | |
770 | * active timer returns 1.) | |
771 | */ | |
772 | int del_timer(struct timer_list *timer) | |
773 | { | |
a6fa8e5a | 774 | struct tvec_base *base; |
1da177e4 | 775 | unsigned long flags; |
55c888d6 | 776 | int ret = 0; |
1da177e4 | 777 | |
82f67cd9 | 778 | timer_stats_timer_clear_start_info(timer); |
55c888d6 ON |
779 | if (timer_pending(timer)) { |
780 | base = lock_timer_base(timer, &flags); | |
781 | if (timer_pending(timer)) { | |
782 | detach_timer(timer, 1); | |
783 | ret = 1; | |
784 | } | |
1da177e4 | 785 | spin_unlock_irqrestore(&base->lock, flags); |
1da177e4 | 786 | } |
1da177e4 | 787 | |
55c888d6 | 788 | return ret; |
1da177e4 | 789 | } |
1da177e4 LT |
790 | EXPORT_SYMBOL(del_timer); |
791 | ||
792 | #ifdef CONFIG_SMP | |
2aae4a10 REB |
793 | /** |
794 | * try_to_del_timer_sync - Try to deactivate a timer | |
795 | * @timer: timer do del | |
796 | * | |
fd450b73 ON |
797 | * This function tries to deactivate a timer. Upon successful (ret >= 0) |
798 | * exit the timer is not queued and the handler is not running on any CPU. | |
799 | * | |
800 | * It must not be called from interrupt contexts. | |
801 | */ | |
802 | int try_to_del_timer_sync(struct timer_list *timer) | |
803 | { | |
a6fa8e5a | 804 | struct tvec_base *base; |
fd450b73 ON |
805 | unsigned long flags; |
806 | int ret = -1; | |
807 | ||
808 | base = lock_timer_base(timer, &flags); | |
809 | ||
810 | if (base->running_timer == timer) | |
811 | goto out; | |
812 | ||
813 | ret = 0; | |
814 | if (timer_pending(timer)) { | |
815 | detach_timer(timer, 1); | |
816 | ret = 1; | |
817 | } | |
818 | out: | |
819 | spin_unlock_irqrestore(&base->lock, flags); | |
820 | ||
821 | return ret; | |
822 | } | |
e19dff1f DH |
823 | EXPORT_SYMBOL(try_to_del_timer_sync); |
824 | ||
2aae4a10 | 825 | /** |
1da177e4 LT |
826 | * del_timer_sync - deactivate a timer and wait for the handler to finish. |
827 | * @timer: the timer to be deactivated | |
828 | * | |
829 | * This function only differs from del_timer() on SMP: besides deactivating | |
830 | * the timer it also makes sure the handler has finished executing on other | |
831 | * CPUs. | |
832 | * | |
72fd4a35 | 833 | * Synchronization rules: Callers must prevent restarting of the timer, |
1da177e4 LT |
834 | * otherwise this function is meaningless. It must not be called from |
835 | * interrupt contexts. The caller must not hold locks which would prevent | |
55c888d6 ON |
836 | * completion of the timer's handler. The timer's handler must not call |
837 | * add_timer_on(). Upon exit the timer is not queued and the handler is | |
838 | * not running on any CPU. | |
1da177e4 LT |
839 | * |
840 | * The function returns whether it has deactivated a pending timer or not. | |
1da177e4 LT |
841 | */ |
842 | int del_timer_sync(struct timer_list *timer) | |
843 | { | |
6f2b9b9a JB |
844 | #ifdef CONFIG_LOCKDEP |
845 | unsigned long flags; | |
846 | ||
847 | local_irq_save(flags); | |
848 | lock_map_acquire(&timer->lockdep_map); | |
849 | lock_map_release(&timer->lockdep_map); | |
850 | local_irq_restore(flags); | |
851 | #endif | |
852 | ||
fd450b73 ON |
853 | for (;;) { |
854 | int ret = try_to_del_timer_sync(timer); | |
855 | if (ret >= 0) | |
856 | return ret; | |
a0009652 | 857 | cpu_relax(); |
fd450b73 | 858 | } |
1da177e4 | 859 | } |
55c888d6 | 860 | EXPORT_SYMBOL(del_timer_sync); |
1da177e4 LT |
861 | #endif |
862 | ||
a6fa8e5a | 863 | static int cascade(struct tvec_base *base, struct tvec *tv, int index) |
1da177e4 LT |
864 | { |
865 | /* cascade all the timers from tv up one level */ | |
3439dd86 P |
866 | struct timer_list *timer, *tmp; |
867 | struct list_head tv_list; | |
868 | ||
869 | list_replace_init(tv->vec + index, &tv_list); | |
1da177e4 | 870 | |
1da177e4 | 871 | /* |
3439dd86 P |
872 | * We are removing _all_ timers from the list, so we |
873 | * don't have to detach them individually. | |
1da177e4 | 874 | */ |
3439dd86 | 875 | list_for_each_entry_safe(timer, tmp, &tv_list, entry) { |
6e453a67 | 876 | BUG_ON(tbase_get_base(timer->base) != base); |
3439dd86 | 877 | internal_add_timer(base, timer); |
1da177e4 | 878 | } |
1da177e4 LT |
879 | |
880 | return index; | |
881 | } | |
882 | ||
2aae4a10 REB |
883 | #define INDEX(N) ((base->timer_jiffies >> (TVR_BITS + (N) * TVN_BITS)) & TVN_MASK) |
884 | ||
885 | /** | |
1da177e4 LT |
886 | * __run_timers - run all expired timers (if any) on this CPU. |
887 | * @base: the timer vector to be processed. | |
888 | * | |
889 | * This function cascades all vectors and executes all expired timer | |
890 | * vectors. | |
891 | */ | |
a6fa8e5a | 892 | static inline void __run_timers(struct tvec_base *base) |
1da177e4 LT |
893 | { |
894 | struct timer_list *timer; | |
895 | ||
3691c519 | 896 | spin_lock_irq(&base->lock); |
1da177e4 | 897 | while (time_after_eq(jiffies, base->timer_jiffies)) { |
626ab0e6 | 898 | struct list_head work_list; |
1da177e4 | 899 | struct list_head *head = &work_list; |
6819457d | 900 | int index = base->timer_jiffies & TVR_MASK; |
626ab0e6 | 901 | |
1da177e4 LT |
902 | /* |
903 | * Cascade timers: | |
904 | */ | |
905 | if (!index && | |
906 | (!cascade(base, &base->tv2, INDEX(0))) && | |
907 | (!cascade(base, &base->tv3, INDEX(1))) && | |
908 | !cascade(base, &base->tv4, INDEX(2))) | |
909 | cascade(base, &base->tv5, INDEX(3)); | |
626ab0e6 ON |
910 | ++base->timer_jiffies; |
911 | list_replace_init(base->tv1.vec + index, &work_list); | |
55c888d6 | 912 | while (!list_empty(head)) { |
1da177e4 LT |
913 | void (*fn)(unsigned long); |
914 | unsigned long data; | |
915 | ||
b5e61818 | 916 | timer = list_first_entry(head, struct timer_list,entry); |
6819457d TG |
917 | fn = timer->function; |
918 | data = timer->data; | |
1da177e4 | 919 | |
82f67cd9 IM |
920 | timer_stats_account_timer(timer); |
921 | ||
1da177e4 | 922 | set_running_timer(base, timer); |
55c888d6 | 923 | detach_timer(timer, 1); |
6f2b9b9a | 924 | |
3691c519 | 925 | spin_unlock_irq(&base->lock); |
1da177e4 | 926 | { |
be5b4fbd | 927 | int preempt_count = preempt_count(); |
6f2b9b9a JB |
928 | |
929 | #ifdef CONFIG_LOCKDEP | |
930 | /* | |
931 | * It is permissible to free the timer from | |
932 | * inside the function that is called from | |
933 | * it, this we need to take into account for | |
934 | * lockdep too. To avoid bogus "held lock | |
935 | * freed" warnings as well as problems when | |
936 | * looking into timer->lockdep_map, make a | |
937 | * copy and use that here. | |
938 | */ | |
939 | struct lockdep_map lockdep_map = | |
940 | timer->lockdep_map; | |
941 | #endif | |
942 | /* | |
943 | * Couple the lock chain with the lock chain at | |
944 | * del_timer_sync() by acquiring the lock_map | |
945 | * around the fn() call here and in | |
946 | * del_timer_sync(). | |
947 | */ | |
948 | lock_map_acquire(&lockdep_map); | |
949 | ||
1da177e4 | 950 | fn(data); |
6f2b9b9a JB |
951 | |
952 | lock_map_release(&lockdep_map); | |
953 | ||
1da177e4 | 954 | if (preempt_count != preempt_count()) { |
4c9dc641 | 955 | printk(KERN_ERR "huh, entered %p " |
be5b4fbd JJ |
956 | "with preempt_count %08x, exited" |
957 | " with %08x?\n", | |
958 | fn, preempt_count, | |
959 | preempt_count()); | |
1da177e4 LT |
960 | BUG(); |
961 | } | |
962 | } | |
3691c519 | 963 | spin_lock_irq(&base->lock); |
1da177e4 LT |
964 | } |
965 | } | |
966 | set_running_timer(base, NULL); | |
3691c519 | 967 | spin_unlock_irq(&base->lock); |
1da177e4 LT |
968 | } |
969 | ||
ee9c5785 | 970 | #ifdef CONFIG_NO_HZ |
1da177e4 LT |
971 | /* |
972 | * Find out when the next timer event is due to happen. This | |
973 | * is used on S/390 to stop all activity when a cpus is idle. | |
974 | * This functions needs to be called disabled. | |
975 | */ | |
a6fa8e5a | 976 | static unsigned long __next_timer_interrupt(struct tvec_base *base) |
1da177e4 | 977 | { |
1cfd6849 | 978 | unsigned long timer_jiffies = base->timer_jiffies; |
eaad084b | 979 | unsigned long expires = timer_jiffies + NEXT_TIMER_MAX_DELTA; |
1cfd6849 | 980 | int index, slot, array, found = 0; |
1da177e4 | 981 | struct timer_list *nte; |
a6fa8e5a | 982 | struct tvec *varray[4]; |
1da177e4 LT |
983 | |
984 | /* Look for timer events in tv1. */ | |
1cfd6849 | 985 | index = slot = timer_jiffies & TVR_MASK; |
1da177e4 | 986 | do { |
1cfd6849 | 987 | list_for_each_entry(nte, base->tv1.vec + slot, entry) { |
6819457d TG |
988 | if (tbase_get_deferrable(nte->base)) |
989 | continue; | |
6e453a67 | 990 | |
1cfd6849 | 991 | found = 1; |
1da177e4 | 992 | expires = nte->expires; |
1cfd6849 TG |
993 | /* Look at the cascade bucket(s)? */ |
994 | if (!index || slot < index) | |
995 | goto cascade; | |
996 | return expires; | |
1da177e4 | 997 | } |
1cfd6849 TG |
998 | slot = (slot + 1) & TVR_MASK; |
999 | } while (slot != index); | |
1000 | ||
1001 | cascade: | |
1002 | /* Calculate the next cascade event */ | |
1003 | if (index) | |
1004 | timer_jiffies += TVR_SIZE - index; | |
1005 | timer_jiffies >>= TVR_BITS; | |
1da177e4 LT |
1006 | |
1007 | /* Check tv2-tv5. */ | |
1008 | varray[0] = &base->tv2; | |
1009 | varray[1] = &base->tv3; | |
1010 | varray[2] = &base->tv4; | |
1011 | varray[3] = &base->tv5; | |
1cfd6849 TG |
1012 | |
1013 | for (array = 0; array < 4; array++) { | |
a6fa8e5a | 1014 | struct tvec *varp = varray[array]; |
1cfd6849 TG |
1015 | |
1016 | index = slot = timer_jiffies & TVN_MASK; | |
1da177e4 | 1017 | do { |
1cfd6849 TG |
1018 | list_for_each_entry(nte, varp->vec + slot, entry) { |
1019 | found = 1; | |
1da177e4 LT |
1020 | if (time_before(nte->expires, expires)) |
1021 | expires = nte->expires; | |
1cfd6849 TG |
1022 | } |
1023 | /* | |
1024 | * Do we still search for the first timer or are | |
1025 | * we looking up the cascade buckets ? | |
1026 | */ | |
1027 | if (found) { | |
1028 | /* Look at the cascade bucket(s)? */ | |
1029 | if (!index || slot < index) | |
1030 | break; | |
1031 | return expires; | |
1032 | } | |
1033 | slot = (slot + 1) & TVN_MASK; | |
1034 | } while (slot != index); | |
1035 | ||
1036 | if (index) | |
1037 | timer_jiffies += TVN_SIZE - index; | |
1038 | timer_jiffies >>= TVN_BITS; | |
1da177e4 | 1039 | } |
1cfd6849 TG |
1040 | return expires; |
1041 | } | |
69239749 | 1042 | |
1cfd6849 TG |
1043 | /* |
1044 | * Check, if the next hrtimer event is before the next timer wheel | |
1045 | * event: | |
1046 | */ | |
1047 | static unsigned long cmp_next_hrtimer_event(unsigned long now, | |
1048 | unsigned long expires) | |
1049 | { | |
1050 | ktime_t hr_delta = hrtimer_get_next_event(); | |
1051 | struct timespec tsdelta; | |
9501b6cf | 1052 | unsigned long delta; |
1cfd6849 TG |
1053 | |
1054 | if (hr_delta.tv64 == KTIME_MAX) | |
1055 | return expires; | |
0662b713 | 1056 | |
9501b6cf TG |
1057 | /* |
1058 | * Expired timer available, let it expire in the next tick | |
1059 | */ | |
1060 | if (hr_delta.tv64 <= 0) | |
1061 | return now + 1; | |
69239749 | 1062 | |
1cfd6849 | 1063 | tsdelta = ktime_to_timespec(hr_delta); |
9501b6cf | 1064 | delta = timespec_to_jiffies(&tsdelta); |
eaad084b TG |
1065 | |
1066 | /* | |
1067 | * Limit the delta to the max value, which is checked in | |
1068 | * tick_nohz_stop_sched_tick(): | |
1069 | */ | |
1070 | if (delta > NEXT_TIMER_MAX_DELTA) | |
1071 | delta = NEXT_TIMER_MAX_DELTA; | |
1072 | ||
9501b6cf TG |
1073 | /* |
1074 | * Take rounding errors in to account and make sure, that it | |
1075 | * expires in the next tick. Otherwise we go into an endless | |
1076 | * ping pong due to tick_nohz_stop_sched_tick() retriggering | |
1077 | * the timer softirq | |
1078 | */ | |
1079 | if (delta < 1) | |
1080 | delta = 1; | |
1081 | now += delta; | |
1cfd6849 TG |
1082 | if (time_before(now, expires)) |
1083 | return now; | |
1da177e4 LT |
1084 | return expires; |
1085 | } | |
1cfd6849 TG |
1086 | |
1087 | /** | |
8dce39c2 | 1088 | * get_next_timer_interrupt - return the jiffy of the next pending timer |
05fb6bf0 | 1089 | * @now: current time (in jiffies) |
1cfd6849 | 1090 | */ |
fd064b9b | 1091 | unsigned long get_next_timer_interrupt(unsigned long now) |
1cfd6849 | 1092 | { |
a6fa8e5a | 1093 | struct tvec_base *base = __get_cpu_var(tvec_bases); |
fd064b9b | 1094 | unsigned long expires; |
1cfd6849 TG |
1095 | |
1096 | spin_lock(&base->lock); | |
1097 | expires = __next_timer_interrupt(base); | |
1098 | spin_unlock(&base->lock); | |
1099 | ||
1100 | if (time_before_eq(expires, now)) | |
1101 | return now; | |
1102 | ||
1103 | return cmp_next_hrtimer_event(now, expires); | |
1104 | } | |
1da177e4 LT |
1105 | #endif |
1106 | ||
1da177e4 | 1107 | /* |
5b4db0c2 | 1108 | * Called from the timer interrupt handler to charge one tick to the current |
1da177e4 LT |
1109 | * process. user_tick is 1 if the tick is user time, 0 for system. |
1110 | */ | |
1111 | void update_process_times(int user_tick) | |
1112 | { | |
1113 | struct task_struct *p = current; | |
1114 | int cpu = smp_processor_id(); | |
1115 | ||
1116 | /* Note: this timer irq context must be accounted for as well. */ | |
fa13a5a1 | 1117 | account_process_tick(p, user_tick); |
1da177e4 LT |
1118 | run_local_timers(); |
1119 | if (rcu_pending(cpu)) | |
1120 | rcu_check_callbacks(cpu, user_tick); | |
b845b517 | 1121 | printk_tick(); |
1da177e4 | 1122 | scheduler_tick(); |
6819457d | 1123 | run_posix_cpu_timers(p); |
1da177e4 LT |
1124 | } |
1125 | ||
1da177e4 LT |
1126 | /* |
1127 | * This function runs timers and the timer-tq in bottom half context. | |
1128 | */ | |
1129 | static void run_timer_softirq(struct softirq_action *h) | |
1130 | { | |
a6fa8e5a | 1131 | struct tvec_base *base = __get_cpu_var(tvec_bases); |
1da177e4 | 1132 | |
925d519a PZ |
1133 | perf_counter_do_pending(); |
1134 | ||
d3d74453 | 1135 | hrtimer_run_pending(); |
82f67cd9 | 1136 | |
1da177e4 LT |
1137 | if (time_after_eq(jiffies, base->timer_jiffies)) |
1138 | __run_timers(base); | |
1139 | } | |
1140 | ||
1141 | /* | |
1142 | * Called by the local, per-CPU timer interrupt on SMP. | |
1143 | */ | |
1144 | void run_local_timers(void) | |
1145 | { | |
d3d74453 | 1146 | hrtimer_run_queues(); |
1da177e4 | 1147 | raise_softirq(TIMER_SOFTIRQ); |
6687a97d | 1148 | softlockup_tick(); |
1da177e4 LT |
1149 | } |
1150 | ||
1da177e4 LT |
1151 | /* |
1152 | * The 64-bit jiffies value is not atomic - you MUST NOT read it | |
1153 | * without sampling the sequence number in xtime_lock. | |
1154 | * jiffies is defined in the linker script... | |
1155 | */ | |
1156 | ||
3171a030 | 1157 | void do_timer(unsigned long ticks) |
1da177e4 | 1158 | { |
3171a030 | 1159 | jiffies_64 += ticks; |
dce48a84 TG |
1160 | update_wall_time(); |
1161 | calc_global_load(); | |
1da177e4 LT |
1162 | } |
1163 | ||
1164 | #ifdef __ARCH_WANT_SYS_ALARM | |
1165 | ||
1166 | /* | |
1167 | * For backwards compatibility? This can be done in libc so Alpha | |
1168 | * and all newer ports shouldn't need it. | |
1169 | */ | |
58fd3aa2 | 1170 | SYSCALL_DEFINE1(alarm, unsigned int, seconds) |
1da177e4 | 1171 | { |
c08b8a49 | 1172 | return alarm_setitimer(seconds); |
1da177e4 LT |
1173 | } |
1174 | ||
1175 | #endif | |
1176 | ||
1177 | #ifndef __alpha__ | |
1178 | ||
1179 | /* | |
1180 | * The Alpha uses getxpid, getxuid, and getxgid instead. Maybe this | |
1181 | * should be moved into arch/i386 instead? | |
1182 | */ | |
1183 | ||
1184 | /** | |
1185 | * sys_getpid - return the thread group id of the current process | |
1186 | * | |
1187 | * Note, despite the name, this returns the tgid not the pid. The tgid and | |
1188 | * the pid are identical unless CLONE_THREAD was specified on clone() in | |
1189 | * which case the tgid is the same in all threads of the same group. | |
1190 | * | |
1191 | * This is SMP safe as current->tgid does not change. | |
1192 | */ | |
58fd3aa2 | 1193 | SYSCALL_DEFINE0(getpid) |
1da177e4 | 1194 | { |
b488893a | 1195 | return task_tgid_vnr(current); |
1da177e4 LT |
1196 | } |
1197 | ||
1198 | /* | |
6997a6fa KK |
1199 | * Accessing ->real_parent is not SMP-safe, it could |
1200 | * change from under us. However, we can use a stale | |
1201 | * value of ->real_parent under rcu_read_lock(), see | |
1202 | * release_task()->call_rcu(delayed_put_task_struct). | |
1da177e4 | 1203 | */ |
dbf040d9 | 1204 | SYSCALL_DEFINE0(getppid) |
1da177e4 LT |
1205 | { |
1206 | int pid; | |
1da177e4 | 1207 | |
6997a6fa | 1208 | rcu_read_lock(); |
6c5f3e7b | 1209 | pid = task_tgid_vnr(current->real_parent); |
6997a6fa | 1210 | rcu_read_unlock(); |
1da177e4 | 1211 | |
1da177e4 LT |
1212 | return pid; |
1213 | } | |
1214 | ||
dbf040d9 | 1215 | SYSCALL_DEFINE0(getuid) |
1da177e4 LT |
1216 | { |
1217 | /* Only we change this so SMP safe */ | |
76aac0e9 | 1218 | return current_uid(); |
1da177e4 LT |
1219 | } |
1220 | ||
dbf040d9 | 1221 | SYSCALL_DEFINE0(geteuid) |
1da177e4 LT |
1222 | { |
1223 | /* Only we change this so SMP safe */ | |
76aac0e9 | 1224 | return current_euid(); |
1da177e4 LT |
1225 | } |
1226 | ||
dbf040d9 | 1227 | SYSCALL_DEFINE0(getgid) |
1da177e4 LT |
1228 | { |
1229 | /* Only we change this so SMP safe */ | |
76aac0e9 | 1230 | return current_gid(); |
1da177e4 LT |
1231 | } |
1232 | ||
dbf040d9 | 1233 | SYSCALL_DEFINE0(getegid) |
1da177e4 LT |
1234 | { |
1235 | /* Only we change this so SMP safe */ | |
76aac0e9 | 1236 | return current_egid(); |
1da177e4 LT |
1237 | } |
1238 | ||
1239 | #endif | |
1240 | ||
1241 | static void process_timeout(unsigned long __data) | |
1242 | { | |
36c8b586 | 1243 | wake_up_process((struct task_struct *)__data); |
1da177e4 LT |
1244 | } |
1245 | ||
1246 | /** | |
1247 | * schedule_timeout - sleep until timeout | |
1248 | * @timeout: timeout value in jiffies | |
1249 | * | |
1250 | * Make the current task sleep until @timeout jiffies have | |
1251 | * elapsed. The routine will return immediately unless | |
1252 | * the current task state has been set (see set_current_state()). | |
1253 | * | |
1254 | * You can set the task state as follows - | |
1255 | * | |
1256 | * %TASK_UNINTERRUPTIBLE - at least @timeout jiffies are guaranteed to | |
1257 | * pass before the routine returns. The routine will return 0 | |
1258 | * | |
1259 | * %TASK_INTERRUPTIBLE - the routine may return early if a signal is | |
1260 | * delivered to the current task. In this case the remaining time | |
1261 | * in jiffies will be returned, or 0 if the timer expired in time | |
1262 | * | |
1263 | * The current task state is guaranteed to be TASK_RUNNING when this | |
1264 | * routine returns. | |
1265 | * | |
1266 | * Specifying a @timeout value of %MAX_SCHEDULE_TIMEOUT will schedule | |
1267 | * the CPU away without a bound on the timeout. In this case the return | |
1268 | * value will be %MAX_SCHEDULE_TIMEOUT. | |
1269 | * | |
1270 | * In all cases the return value is guaranteed to be non-negative. | |
1271 | */ | |
7ad5b3a5 | 1272 | signed long __sched schedule_timeout(signed long timeout) |
1da177e4 LT |
1273 | { |
1274 | struct timer_list timer; | |
1275 | unsigned long expire; | |
1276 | ||
1277 | switch (timeout) | |
1278 | { | |
1279 | case MAX_SCHEDULE_TIMEOUT: | |
1280 | /* | |
1281 | * These two special cases are useful to be comfortable | |
1282 | * in the caller. Nothing more. We could take | |
1283 | * MAX_SCHEDULE_TIMEOUT from one of the negative value | |
1284 | * but I' d like to return a valid offset (>=0) to allow | |
1285 | * the caller to do everything it want with the retval. | |
1286 | */ | |
1287 | schedule(); | |
1288 | goto out; | |
1289 | default: | |
1290 | /* | |
1291 | * Another bit of PARANOID. Note that the retval will be | |
1292 | * 0 since no piece of kernel is supposed to do a check | |
1293 | * for a negative retval of schedule_timeout() (since it | |
1294 | * should never happens anyway). You just have the printk() | |
1295 | * that will tell you if something is gone wrong and where. | |
1296 | */ | |
5b149bcc | 1297 | if (timeout < 0) { |
1da177e4 | 1298 | printk(KERN_ERR "schedule_timeout: wrong timeout " |
5b149bcc AM |
1299 | "value %lx\n", timeout); |
1300 | dump_stack(); | |
1da177e4 LT |
1301 | current->state = TASK_RUNNING; |
1302 | goto out; | |
1303 | } | |
1304 | } | |
1305 | ||
1306 | expire = timeout + jiffies; | |
1307 | ||
c6f3a97f | 1308 | setup_timer_on_stack(&timer, process_timeout, (unsigned long)current); |
74019224 | 1309 | __mod_timer(&timer, expire, false); |
1da177e4 LT |
1310 | schedule(); |
1311 | del_singleshot_timer_sync(&timer); | |
1312 | ||
c6f3a97f TG |
1313 | /* Remove the timer from the object tracker */ |
1314 | destroy_timer_on_stack(&timer); | |
1315 | ||
1da177e4 LT |
1316 | timeout = expire - jiffies; |
1317 | ||
1318 | out: | |
1319 | return timeout < 0 ? 0 : timeout; | |
1320 | } | |
1da177e4 LT |
1321 | EXPORT_SYMBOL(schedule_timeout); |
1322 | ||
8a1c1757 AM |
1323 | /* |
1324 | * We can use __set_current_state() here because schedule_timeout() calls | |
1325 | * schedule() unconditionally. | |
1326 | */ | |
64ed93a2 NA |
1327 | signed long __sched schedule_timeout_interruptible(signed long timeout) |
1328 | { | |
a5a0d52c AM |
1329 | __set_current_state(TASK_INTERRUPTIBLE); |
1330 | return schedule_timeout(timeout); | |
64ed93a2 NA |
1331 | } |
1332 | EXPORT_SYMBOL(schedule_timeout_interruptible); | |
1333 | ||
294d5cc2 MW |
1334 | signed long __sched schedule_timeout_killable(signed long timeout) |
1335 | { | |
1336 | __set_current_state(TASK_KILLABLE); | |
1337 | return schedule_timeout(timeout); | |
1338 | } | |
1339 | EXPORT_SYMBOL(schedule_timeout_killable); | |
1340 | ||
64ed93a2 NA |
1341 | signed long __sched schedule_timeout_uninterruptible(signed long timeout) |
1342 | { | |
a5a0d52c AM |
1343 | __set_current_state(TASK_UNINTERRUPTIBLE); |
1344 | return schedule_timeout(timeout); | |
64ed93a2 NA |
1345 | } |
1346 | EXPORT_SYMBOL(schedule_timeout_uninterruptible); | |
1347 | ||
1da177e4 | 1348 | /* Thread ID - the internal kernel "pid" */ |
58fd3aa2 | 1349 | SYSCALL_DEFINE0(gettid) |
1da177e4 | 1350 | { |
b488893a | 1351 | return task_pid_vnr(current); |
1da177e4 LT |
1352 | } |
1353 | ||
2aae4a10 | 1354 | /** |
d4d23add | 1355 | * do_sysinfo - fill in sysinfo struct |
2aae4a10 | 1356 | * @info: pointer to buffer to fill |
6819457d | 1357 | */ |
d4d23add | 1358 | int do_sysinfo(struct sysinfo *info) |
1da177e4 | 1359 | { |
1da177e4 LT |
1360 | unsigned long mem_total, sav_total; |
1361 | unsigned int mem_unit, bitcount; | |
2d02494f | 1362 | struct timespec tp; |
1da177e4 | 1363 | |
d4d23add | 1364 | memset(info, 0, sizeof(struct sysinfo)); |
1da177e4 | 1365 | |
2d02494f TG |
1366 | ktime_get_ts(&tp); |
1367 | monotonic_to_bootbased(&tp); | |
1368 | info->uptime = tp.tv_sec + (tp.tv_nsec ? 1 : 0); | |
1da177e4 | 1369 | |
2d02494f | 1370 | get_avenrun(info->loads, 0, SI_LOAD_SHIFT - FSHIFT); |
1da177e4 | 1371 | |
2d02494f | 1372 | info->procs = nr_threads; |
1da177e4 | 1373 | |
d4d23add KM |
1374 | si_meminfo(info); |
1375 | si_swapinfo(info); | |
1da177e4 LT |
1376 | |
1377 | /* | |
1378 | * If the sum of all the available memory (i.e. ram + swap) | |
1379 | * is less than can be stored in a 32 bit unsigned long then | |
1380 | * we can be binary compatible with 2.2.x kernels. If not, | |
1381 | * well, in that case 2.2.x was broken anyways... | |
1382 | * | |
1383 | * -Erik Andersen <andersee@debian.org> | |
1384 | */ | |
1385 | ||
d4d23add KM |
1386 | mem_total = info->totalram + info->totalswap; |
1387 | if (mem_total < info->totalram || mem_total < info->totalswap) | |
1da177e4 LT |
1388 | goto out; |
1389 | bitcount = 0; | |
d4d23add | 1390 | mem_unit = info->mem_unit; |
1da177e4 LT |
1391 | while (mem_unit > 1) { |
1392 | bitcount++; | |
1393 | mem_unit >>= 1; | |
1394 | sav_total = mem_total; | |
1395 | mem_total <<= 1; | |
1396 | if (mem_total < sav_total) | |
1397 | goto out; | |
1398 | } | |
1399 | ||
1400 | /* | |
1401 | * If mem_total did not overflow, multiply all memory values by | |
d4d23add | 1402 | * info->mem_unit and set it to 1. This leaves things compatible |
1da177e4 LT |
1403 | * with 2.2.x, and also retains compatibility with earlier 2.4.x |
1404 | * kernels... | |
1405 | */ | |
1406 | ||
d4d23add KM |
1407 | info->mem_unit = 1; |
1408 | info->totalram <<= bitcount; | |
1409 | info->freeram <<= bitcount; | |
1410 | info->sharedram <<= bitcount; | |
1411 | info->bufferram <<= bitcount; | |
1412 | info->totalswap <<= bitcount; | |
1413 | info->freeswap <<= bitcount; | |
1414 | info->totalhigh <<= bitcount; | |
1415 | info->freehigh <<= bitcount; | |
1416 | ||
1417 | out: | |
1418 | return 0; | |
1419 | } | |
1420 | ||
1e7bfb21 | 1421 | SYSCALL_DEFINE1(sysinfo, struct sysinfo __user *, info) |
d4d23add KM |
1422 | { |
1423 | struct sysinfo val; | |
1424 | ||
1425 | do_sysinfo(&val); | |
1da177e4 | 1426 | |
1da177e4 LT |
1427 | if (copy_to_user(info, &val, sizeof(struct sysinfo))) |
1428 | return -EFAULT; | |
1429 | ||
1430 | return 0; | |
1431 | } | |
1432 | ||
b4be6258 | 1433 | static int __cpuinit init_timers_cpu(int cpu) |
1da177e4 LT |
1434 | { |
1435 | int j; | |
a6fa8e5a | 1436 | struct tvec_base *base; |
b4be6258 | 1437 | static char __cpuinitdata tvec_base_done[NR_CPUS]; |
55c888d6 | 1438 | |
ba6edfcd | 1439 | if (!tvec_base_done[cpu]) { |
a4a6198b JB |
1440 | static char boot_done; |
1441 | ||
a4a6198b | 1442 | if (boot_done) { |
ba6edfcd AM |
1443 | /* |
1444 | * The APs use this path later in boot | |
1445 | */ | |
94f6030c CL |
1446 | base = kmalloc_node(sizeof(*base), |
1447 | GFP_KERNEL | __GFP_ZERO, | |
a4a6198b JB |
1448 | cpu_to_node(cpu)); |
1449 | if (!base) | |
1450 | return -ENOMEM; | |
6e453a67 VP |
1451 | |
1452 | /* Make sure that tvec_base is 2 byte aligned */ | |
1453 | if (tbase_get_deferrable(base)) { | |
1454 | WARN_ON(1); | |
1455 | kfree(base); | |
1456 | return -ENOMEM; | |
1457 | } | |
ba6edfcd | 1458 | per_cpu(tvec_bases, cpu) = base; |
a4a6198b | 1459 | } else { |
ba6edfcd AM |
1460 | /* |
1461 | * This is for the boot CPU - we use compile-time | |
1462 | * static initialisation because per-cpu memory isn't | |
1463 | * ready yet and because the memory allocators are not | |
1464 | * initialised either. | |
1465 | */ | |
a4a6198b | 1466 | boot_done = 1; |
ba6edfcd | 1467 | base = &boot_tvec_bases; |
a4a6198b | 1468 | } |
ba6edfcd AM |
1469 | tvec_base_done[cpu] = 1; |
1470 | } else { | |
1471 | base = per_cpu(tvec_bases, cpu); | |
a4a6198b | 1472 | } |
ba6edfcd | 1473 | |
3691c519 | 1474 | spin_lock_init(&base->lock); |
d730e882 | 1475 | |
1da177e4 LT |
1476 | for (j = 0; j < TVN_SIZE; j++) { |
1477 | INIT_LIST_HEAD(base->tv5.vec + j); | |
1478 | INIT_LIST_HEAD(base->tv4.vec + j); | |
1479 | INIT_LIST_HEAD(base->tv3.vec + j); | |
1480 | INIT_LIST_HEAD(base->tv2.vec + j); | |
1481 | } | |
1482 | for (j = 0; j < TVR_SIZE; j++) | |
1483 | INIT_LIST_HEAD(base->tv1.vec + j); | |
1484 | ||
1485 | base->timer_jiffies = jiffies; | |
a4a6198b | 1486 | return 0; |
1da177e4 LT |
1487 | } |
1488 | ||
1489 | #ifdef CONFIG_HOTPLUG_CPU | |
a6fa8e5a | 1490 | static void migrate_timer_list(struct tvec_base *new_base, struct list_head *head) |
1da177e4 LT |
1491 | { |
1492 | struct timer_list *timer; | |
1493 | ||
1494 | while (!list_empty(head)) { | |
b5e61818 | 1495 | timer = list_first_entry(head, struct timer_list, entry); |
55c888d6 | 1496 | detach_timer(timer, 0); |
6e453a67 | 1497 | timer_set_base(timer, new_base); |
1da177e4 | 1498 | internal_add_timer(new_base, timer); |
1da177e4 | 1499 | } |
1da177e4 LT |
1500 | } |
1501 | ||
48ccf3da | 1502 | static void __cpuinit migrate_timers(int cpu) |
1da177e4 | 1503 | { |
a6fa8e5a PM |
1504 | struct tvec_base *old_base; |
1505 | struct tvec_base *new_base; | |
1da177e4 LT |
1506 | int i; |
1507 | ||
1508 | BUG_ON(cpu_online(cpu)); | |
a4a6198b JB |
1509 | old_base = per_cpu(tvec_bases, cpu); |
1510 | new_base = get_cpu_var(tvec_bases); | |
d82f0b0f ON |
1511 | /* |
1512 | * The caller is globally serialized and nobody else | |
1513 | * takes two locks at once, deadlock is not possible. | |
1514 | */ | |
1515 | spin_lock_irq(&new_base->lock); | |
0d180406 | 1516 | spin_lock_nested(&old_base->lock, SINGLE_DEPTH_NESTING); |
3691c519 ON |
1517 | |
1518 | BUG_ON(old_base->running_timer); | |
1da177e4 | 1519 | |
1da177e4 | 1520 | for (i = 0; i < TVR_SIZE; i++) |
55c888d6 ON |
1521 | migrate_timer_list(new_base, old_base->tv1.vec + i); |
1522 | for (i = 0; i < TVN_SIZE; i++) { | |
1523 | migrate_timer_list(new_base, old_base->tv2.vec + i); | |
1524 | migrate_timer_list(new_base, old_base->tv3.vec + i); | |
1525 | migrate_timer_list(new_base, old_base->tv4.vec + i); | |
1526 | migrate_timer_list(new_base, old_base->tv5.vec + i); | |
1527 | } | |
1528 | ||
0d180406 | 1529 | spin_unlock(&old_base->lock); |
d82f0b0f | 1530 | spin_unlock_irq(&new_base->lock); |
1da177e4 | 1531 | put_cpu_var(tvec_bases); |
1da177e4 LT |
1532 | } |
1533 | #endif /* CONFIG_HOTPLUG_CPU */ | |
1534 | ||
8c78f307 | 1535 | static int __cpuinit timer_cpu_notify(struct notifier_block *self, |
1da177e4 LT |
1536 | unsigned long action, void *hcpu) |
1537 | { | |
1538 | long cpu = (long)hcpu; | |
1539 | switch(action) { | |
1540 | case CPU_UP_PREPARE: | |
8bb78442 | 1541 | case CPU_UP_PREPARE_FROZEN: |
a4a6198b JB |
1542 | if (init_timers_cpu(cpu) < 0) |
1543 | return NOTIFY_BAD; | |
1da177e4 LT |
1544 | break; |
1545 | #ifdef CONFIG_HOTPLUG_CPU | |
1546 | case CPU_DEAD: | |
8bb78442 | 1547 | case CPU_DEAD_FROZEN: |
1da177e4 LT |
1548 | migrate_timers(cpu); |
1549 | break; | |
1550 | #endif | |
1551 | default: | |
1552 | break; | |
1553 | } | |
1554 | return NOTIFY_OK; | |
1555 | } | |
1556 | ||
8c78f307 | 1557 | static struct notifier_block __cpuinitdata timers_nb = { |
1da177e4 LT |
1558 | .notifier_call = timer_cpu_notify, |
1559 | }; | |
1560 | ||
1561 | ||
1562 | void __init init_timers(void) | |
1563 | { | |
07dccf33 | 1564 | int err = timer_cpu_notify(&timers_nb, (unsigned long)CPU_UP_PREPARE, |
1da177e4 | 1565 | (void *)(long)smp_processor_id()); |
07dccf33 | 1566 | |
82f67cd9 IM |
1567 | init_timer_stats(); |
1568 | ||
07dccf33 | 1569 | BUG_ON(err == NOTIFY_BAD); |
1da177e4 | 1570 | register_cpu_notifier(&timers_nb); |
962cf36c | 1571 | open_softirq(TIMER_SOFTIRQ, run_timer_softirq); |
1da177e4 LT |
1572 | } |
1573 | ||
1da177e4 LT |
1574 | /** |
1575 | * msleep - sleep safely even with waitqueue interruptions | |
1576 | * @msecs: Time in milliseconds to sleep for | |
1577 | */ | |
1578 | void msleep(unsigned int msecs) | |
1579 | { | |
1580 | unsigned long timeout = msecs_to_jiffies(msecs) + 1; | |
1581 | ||
75bcc8c5 NA |
1582 | while (timeout) |
1583 | timeout = schedule_timeout_uninterruptible(timeout); | |
1da177e4 LT |
1584 | } |
1585 | ||
1586 | EXPORT_SYMBOL(msleep); | |
1587 | ||
1588 | /** | |
96ec3efd | 1589 | * msleep_interruptible - sleep waiting for signals |
1da177e4 LT |
1590 | * @msecs: Time in milliseconds to sleep for |
1591 | */ | |
1592 | unsigned long msleep_interruptible(unsigned int msecs) | |
1593 | { | |
1594 | unsigned long timeout = msecs_to_jiffies(msecs) + 1; | |
1595 | ||
75bcc8c5 NA |
1596 | while (timeout && !signal_pending(current)) |
1597 | timeout = schedule_timeout_interruptible(timeout); | |
1da177e4 LT |
1598 | return jiffies_to_msecs(timeout); |
1599 | } | |
1600 | ||
1601 | EXPORT_SYMBOL(msleep_interruptible); |