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> |
1da177e4 LT |
40 | |
41 | #include <asm/uaccess.h> | |
42 | #include <asm/unistd.h> | |
43 | #include <asm/div64.h> | |
44 | #include <asm/timex.h> | |
45 | #include <asm/io.h> | |
46 | ||
ecea8d19 TG |
47 | u64 jiffies_64 __cacheline_aligned_in_smp = INITIAL_JIFFIES; |
48 | ||
49 | EXPORT_SYMBOL(jiffies_64); | |
50 | ||
1da177e4 LT |
51 | /* |
52 | * per-CPU timer vector definitions: | |
53 | */ | |
1da177e4 LT |
54 | #define TVN_BITS (CONFIG_BASE_SMALL ? 4 : 6) |
55 | #define TVR_BITS (CONFIG_BASE_SMALL ? 6 : 8) | |
56 | #define TVN_SIZE (1 << TVN_BITS) | |
57 | #define TVR_SIZE (1 << TVR_BITS) | |
58 | #define TVN_MASK (TVN_SIZE - 1) | |
59 | #define TVR_MASK (TVR_SIZE - 1) | |
60 | ||
a6fa8e5a | 61 | struct tvec { |
1da177e4 | 62 | struct list_head vec[TVN_SIZE]; |
a6fa8e5a | 63 | }; |
1da177e4 | 64 | |
a6fa8e5a | 65 | struct tvec_root { |
1da177e4 | 66 | struct list_head vec[TVR_SIZE]; |
a6fa8e5a | 67 | }; |
1da177e4 | 68 | |
a6fa8e5a | 69 | struct tvec_base { |
3691c519 ON |
70 | spinlock_t lock; |
71 | struct timer_list *running_timer; | |
1da177e4 | 72 | unsigned long timer_jiffies; |
a6fa8e5a PM |
73 | struct tvec_root tv1; |
74 | struct tvec tv2; | |
75 | struct tvec tv3; | |
76 | struct tvec tv4; | |
77 | struct tvec tv5; | |
6e453a67 | 78 | } ____cacheline_aligned; |
1da177e4 | 79 | |
a6fa8e5a | 80 | struct tvec_base boot_tvec_bases; |
3691c519 | 81 | EXPORT_SYMBOL(boot_tvec_bases); |
a6fa8e5a | 82 | static DEFINE_PER_CPU(struct tvec_base *, tvec_bases) = &boot_tvec_bases; |
1da177e4 | 83 | |
6e453a67 | 84 | /* |
a6fa8e5a | 85 | * Note that all tvec_bases are 2 byte aligned and lower bit of |
6e453a67 VP |
86 | * base in timer_list is guaranteed to be zero. Use the LSB for |
87 | * the new flag to indicate whether the timer is deferrable | |
88 | */ | |
89 | #define TBASE_DEFERRABLE_FLAG (0x1) | |
90 | ||
91 | /* Functions below help us manage 'deferrable' flag */ | |
a6fa8e5a | 92 | static inline unsigned int tbase_get_deferrable(struct tvec_base *base) |
6e453a67 | 93 | { |
e9910846 | 94 | return ((unsigned int)(unsigned long)base & TBASE_DEFERRABLE_FLAG); |
6e453a67 VP |
95 | } |
96 | ||
a6fa8e5a | 97 | static inline struct tvec_base *tbase_get_base(struct tvec_base *base) |
6e453a67 | 98 | { |
a6fa8e5a | 99 | return ((struct tvec_base *)((unsigned long)base & ~TBASE_DEFERRABLE_FLAG)); |
6e453a67 VP |
100 | } |
101 | ||
102 | static inline void timer_set_deferrable(struct timer_list *timer) | |
103 | { | |
a6fa8e5a | 104 | timer->base = ((struct tvec_base *)((unsigned long)(timer->base) | |
6819457d | 105 | TBASE_DEFERRABLE_FLAG)); |
6e453a67 VP |
106 | } |
107 | ||
108 | static inline void | |
a6fa8e5a | 109 | timer_set_base(struct timer_list *timer, struct tvec_base *new_base) |
6e453a67 | 110 | { |
a6fa8e5a | 111 | timer->base = (struct tvec_base *)((unsigned long)(new_base) | |
6819457d | 112 | tbase_get_deferrable(timer->base)); |
6e453a67 VP |
113 | } |
114 | ||
9c133c46 AS |
115 | static unsigned long round_jiffies_common(unsigned long j, int cpu, |
116 | bool force_up) | |
4c36a5de AV |
117 | { |
118 | int rem; | |
119 | unsigned long original = j; | |
120 | ||
121 | /* | |
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 | |
125 | * already did this. | |
126 | * The skew is done by adding 3*cpunr, then round, then subtract this | |
127 | * extra offset again. | |
128 | */ | |
129 | j += cpu * 3; | |
130 | ||
131 | rem = j % HZ; | |
132 | ||
133 | /* | |
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. | |
9c133c46 | 138 | * But never round down if @force_up is set. |
4c36a5de | 139 | */ |
9c133c46 | 140 | if (rem < HZ/4 && !force_up) /* round down */ |
4c36a5de AV |
141 | j = j - rem; |
142 | else /* round up */ | |
143 | j = j - rem + HZ; | |
144 | ||
145 | /* now that we have rounded, subtract the extra skew again */ | |
146 | j -= cpu * 3; | |
147 | ||
148 | if (j <= jiffies) /* rounding ate our timeout entirely; */ | |
149 | return original; | |
150 | return j; | |
151 | } | |
9c133c46 AS |
152 | |
153 | /** | |
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 | |
157 | * | |
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. | |
162 | * | |
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. | |
166 | * | |
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. | |
170 | * | |
171 | * The return value is the rounded version of the @j parameter. | |
172 | */ | |
173 | unsigned long __round_jiffies(unsigned long j, int cpu) | |
174 | { | |
175 | return round_jiffies_common(j, cpu, false); | |
176 | } | |
4c36a5de AV |
177 | EXPORT_SYMBOL_GPL(__round_jiffies); |
178 | ||
179 | /** | |
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 | |
183 | * | |
72fd4a35 | 184 | * __round_jiffies_relative() rounds a time delta in the future (in jiffies) |
4c36a5de AV |
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. | |
188 | * | |
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. | |
192 | * | |
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. | |
196 | * | |
72fd4a35 | 197 | * The return value is the rounded version of the @j parameter. |
4c36a5de AV |
198 | */ |
199 | unsigned long __round_jiffies_relative(unsigned long j, int cpu) | |
200 | { | |
9c133c46 AS |
201 | unsigned long j0 = jiffies; |
202 | ||
203 | /* Use j0 because jiffies might change while we run */ | |
204 | return round_jiffies_common(j + j0, cpu, false) - j0; | |
4c36a5de AV |
205 | } |
206 | EXPORT_SYMBOL_GPL(__round_jiffies_relative); | |
207 | ||
208 | /** | |
209 | * round_jiffies - function to round jiffies to a full second | |
210 | * @j: the time in (absolute) jiffies that should be rounded | |
211 | * | |
72fd4a35 | 212 | * round_jiffies() rounds an absolute time in the future (in jiffies) |
4c36a5de AV |
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. | |
216 | * | |
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. | |
220 | * | |
72fd4a35 | 221 | * The return value is the rounded version of the @j parameter. |
4c36a5de AV |
222 | */ |
223 | unsigned long round_jiffies(unsigned long j) | |
224 | { | |
9c133c46 | 225 | return round_jiffies_common(j, raw_smp_processor_id(), false); |
4c36a5de AV |
226 | } |
227 | EXPORT_SYMBOL_GPL(round_jiffies); | |
228 | ||
229 | /** | |
230 | * round_jiffies_relative - function to round jiffies to a full second | |
231 | * @j: the time in (relative) jiffies that should be rounded | |
232 | * | |
72fd4a35 | 233 | * round_jiffies_relative() rounds a time delta in the future (in jiffies) |
4c36a5de AV |
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. | |
237 | * | |
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. | |
241 | * | |
72fd4a35 | 242 | * The return value is the rounded version of the @j parameter. |
4c36a5de AV |
243 | */ |
244 | unsigned long round_jiffies_relative(unsigned long j) | |
245 | { | |
246 | return __round_jiffies_relative(j, raw_smp_processor_id()); | |
247 | } | |
248 | EXPORT_SYMBOL_GPL(round_jiffies_relative); | |
249 | ||
9c133c46 AS |
250 | /** |
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 | |
254 | * | |
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 | |
258 | * early. | |
259 | */ | |
260 | unsigned long __round_jiffies_up(unsigned long j, int cpu) | |
261 | { | |
262 | return round_jiffies_common(j, cpu, true); | |
263 | } | |
264 | EXPORT_SYMBOL_GPL(__round_jiffies_up); | |
265 | ||
266 | /** | |
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 | |
270 | * | |
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 | |
274 | * early. | |
275 | */ | |
276 | unsigned long __round_jiffies_up_relative(unsigned long j, int cpu) | |
277 | { | |
278 | unsigned long j0 = jiffies; | |
279 | ||
280 | /* Use j0 because jiffies might change while we run */ | |
281 | return round_jiffies_common(j + j0, cpu, true) - j0; | |
282 | } | |
283 | EXPORT_SYMBOL_GPL(__round_jiffies_up_relative); | |
284 | ||
285 | /** | |
286 | * round_jiffies_up - function to round jiffies up to a full second | |
287 | * @j: the time in (absolute) jiffies that should be rounded | |
288 | * | |
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 | |
292 | * early. | |
293 | */ | |
294 | unsigned long round_jiffies_up(unsigned long j) | |
295 | { | |
296 | return round_jiffies_common(j, raw_smp_processor_id(), true); | |
297 | } | |
298 | EXPORT_SYMBOL_GPL(round_jiffies_up); | |
299 | ||
300 | /** | |
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 | |
303 | * | |
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 | |
307 | * early. | |
308 | */ | |
309 | unsigned long round_jiffies_up_relative(unsigned long j) | |
310 | { | |
311 | return __round_jiffies_up_relative(j, raw_smp_processor_id()); | |
312 | } | |
313 | EXPORT_SYMBOL_GPL(round_jiffies_up_relative); | |
314 | ||
4c36a5de | 315 | |
a6fa8e5a | 316 | static inline void set_running_timer(struct tvec_base *base, |
1da177e4 LT |
317 | struct timer_list *timer) |
318 | { | |
319 | #ifdef CONFIG_SMP | |
3691c519 | 320 | base->running_timer = timer; |
1da177e4 LT |
321 | #endif |
322 | } | |
323 | ||
a6fa8e5a | 324 | static void internal_add_timer(struct tvec_base *base, struct timer_list *timer) |
1da177e4 LT |
325 | { |
326 | unsigned long expires = timer->expires; | |
327 | unsigned long idx = expires - base->timer_jiffies; | |
328 | struct list_head *vec; | |
329 | ||
330 | if (idx < TVR_SIZE) { | |
331 | int i = expires & TVR_MASK; | |
332 | vec = base->tv1.vec + i; | |
333 | } else if (idx < 1 << (TVR_BITS + TVN_BITS)) { | |
334 | int i = (expires >> TVR_BITS) & TVN_MASK; | |
335 | vec = base->tv2.vec + i; | |
336 | } else if (idx < 1 << (TVR_BITS + 2 * TVN_BITS)) { | |
337 | int i = (expires >> (TVR_BITS + TVN_BITS)) & TVN_MASK; | |
338 | vec = base->tv3.vec + i; | |
339 | } else if (idx < 1 << (TVR_BITS + 3 * TVN_BITS)) { | |
340 | int i = (expires >> (TVR_BITS + 2 * TVN_BITS)) & TVN_MASK; | |
341 | vec = base->tv4.vec + i; | |
342 | } else if ((signed long) idx < 0) { | |
343 | /* | |
344 | * Can happen if you add a timer with expires == jiffies, | |
345 | * or you set a timer to go off in the past | |
346 | */ | |
347 | vec = base->tv1.vec + (base->timer_jiffies & TVR_MASK); | |
348 | } else { | |
349 | int i; | |
350 | /* If the timeout is larger than 0xffffffff on 64-bit | |
351 | * architectures then we use the maximum timeout: | |
352 | */ | |
353 | if (idx > 0xffffffffUL) { | |
354 | idx = 0xffffffffUL; | |
355 | expires = idx + base->timer_jiffies; | |
356 | } | |
357 | i = (expires >> (TVR_BITS + 3 * TVN_BITS)) & TVN_MASK; | |
358 | vec = base->tv5.vec + i; | |
359 | } | |
360 | /* | |
361 | * Timers are FIFO: | |
362 | */ | |
363 | list_add_tail(&timer->entry, vec); | |
364 | } | |
365 | ||
82f67cd9 IM |
366 | #ifdef CONFIG_TIMER_STATS |
367 | void __timer_stats_timer_set_start_info(struct timer_list *timer, void *addr) | |
368 | { | |
369 | if (timer->start_site) | |
370 | return; | |
371 | ||
372 | timer->start_site = addr; | |
373 | memcpy(timer->start_comm, current->comm, TASK_COMM_LEN); | |
374 | timer->start_pid = current->pid; | |
375 | } | |
c5c061b8 VP |
376 | |
377 | static void timer_stats_account_timer(struct timer_list *timer) | |
378 | { | |
379 | unsigned int flag = 0; | |
380 | ||
381 | if (unlikely(tbase_get_deferrable(timer->base))) | |
382 | flag |= TIMER_STATS_FLAG_DEFERRABLE; | |
383 | ||
384 | timer_stats_update_stats(timer, timer->start_pid, timer->start_site, | |
385 | timer->function, timer->start_comm, flag); | |
386 | } | |
387 | ||
388 | #else | |
389 | static void timer_stats_account_timer(struct timer_list *timer) {} | |
82f67cd9 IM |
390 | #endif |
391 | ||
c6f3a97f TG |
392 | #ifdef CONFIG_DEBUG_OBJECTS_TIMERS |
393 | ||
394 | static struct debug_obj_descr timer_debug_descr; | |
395 | ||
396 | /* | |
397 | * fixup_init is called when: | |
398 | * - an active object is initialized | |
55c888d6 | 399 | */ |
c6f3a97f TG |
400 | static int timer_fixup_init(void *addr, enum debug_obj_state state) |
401 | { | |
402 | struct timer_list *timer = addr; | |
403 | ||
404 | switch (state) { | |
405 | case ODEBUG_STATE_ACTIVE: | |
406 | del_timer_sync(timer); | |
407 | debug_object_init(timer, &timer_debug_descr); | |
408 | return 1; | |
409 | default: | |
410 | return 0; | |
411 | } | |
412 | } | |
413 | ||
414 | /* | |
415 | * fixup_activate is called when: | |
416 | * - an active object is activated | |
417 | * - an unknown object is activated (might be a statically initialized object) | |
418 | */ | |
419 | static int timer_fixup_activate(void *addr, enum debug_obj_state state) | |
420 | { | |
421 | struct timer_list *timer = addr; | |
422 | ||
423 | switch (state) { | |
424 | ||
425 | case ODEBUG_STATE_NOTAVAILABLE: | |
426 | /* | |
427 | * This is not really a fixup. The timer was | |
428 | * statically initialized. We just make sure that it | |
429 | * is tracked in the object tracker. | |
430 | */ | |
431 | if (timer->entry.next == NULL && | |
432 | timer->entry.prev == TIMER_ENTRY_STATIC) { | |
433 | debug_object_init(timer, &timer_debug_descr); | |
434 | debug_object_activate(timer, &timer_debug_descr); | |
435 | return 0; | |
436 | } else { | |
437 | WARN_ON_ONCE(1); | |
438 | } | |
439 | return 0; | |
440 | ||
441 | case ODEBUG_STATE_ACTIVE: | |
442 | WARN_ON(1); | |
443 | ||
444 | default: | |
445 | return 0; | |
446 | } | |
447 | } | |
448 | ||
449 | /* | |
450 | * fixup_free is called when: | |
451 | * - an active object is freed | |
452 | */ | |
453 | static int timer_fixup_free(void *addr, enum debug_obj_state state) | |
454 | { | |
455 | struct timer_list *timer = addr; | |
456 | ||
457 | switch (state) { | |
458 | case ODEBUG_STATE_ACTIVE: | |
459 | del_timer_sync(timer); | |
460 | debug_object_free(timer, &timer_debug_descr); | |
461 | return 1; | |
462 | default: | |
463 | return 0; | |
464 | } | |
465 | } | |
466 | ||
467 | static struct debug_obj_descr timer_debug_descr = { | |
468 | .name = "timer_list", | |
469 | .fixup_init = timer_fixup_init, | |
470 | .fixup_activate = timer_fixup_activate, | |
471 | .fixup_free = timer_fixup_free, | |
472 | }; | |
473 | ||
474 | static inline void debug_timer_init(struct timer_list *timer) | |
475 | { | |
476 | debug_object_init(timer, &timer_debug_descr); | |
477 | } | |
478 | ||
479 | static inline void debug_timer_activate(struct timer_list *timer) | |
480 | { | |
481 | debug_object_activate(timer, &timer_debug_descr); | |
482 | } | |
483 | ||
484 | static inline void debug_timer_deactivate(struct timer_list *timer) | |
485 | { | |
486 | debug_object_deactivate(timer, &timer_debug_descr); | |
487 | } | |
488 | ||
489 | static inline void debug_timer_free(struct timer_list *timer) | |
490 | { | |
491 | debug_object_free(timer, &timer_debug_descr); | |
492 | } | |
493 | ||
494 | static void __init_timer(struct timer_list *timer); | |
495 | ||
496 | void init_timer_on_stack(struct timer_list *timer) | |
497 | { | |
498 | debug_object_init_on_stack(timer, &timer_debug_descr); | |
499 | __init_timer(timer); | |
500 | } | |
501 | EXPORT_SYMBOL_GPL(init_timer_on_stack); | |
502 | ||
503 | void destroy_timer_on_stack(struct timer_list *timer) | |
504 | { | |
505 | debug_object_free(timer, &timer_debug_descr); | |
506 | } | |
507 | EXPORT_SYMBOL_GPL(destroy_timer_on_stack); | |
508 | ||
509 | #else | |
510 | static inline void debug_timer_init(struct timer_list *timer) { } | |
511 | static inline void debug_timer_activate(struct timer_list *timer) { } | |
512 | static inline void debug_timer_deactivate(struct timer_list *timer) { } | |
513 | #endif | |
514 | ||
515 | static void __init_timer(struct timer_list *timer) | |
55c888d6 ON |
516 | { |
517 | timer->entry.next = NULL; | |
bfe5d834 | 518 | timer->base = __raw_get_cpu_var(tvec_bases); |
82f67cd9 IM |
519 | #ifdef CONFIG_TIMER_STATS |
520 | timer->start_site = NULL; | |
521 | timer->start_pid = -1; | |
522 | memset(timer->start_comm, 0, TASK_COMM_LEN); | |
523 | #endif | |
55c888d6 | 524 | } |
c6f3a97f TG |
525 | |
526 | /** | |
527 | * init_timer - initialize a timer. | |
528 | * @timer: the timer to be initialized | |
529 | * | |
530 | * init_timer() must be done to a timer prior calling *any* of the | |
531 | * other timer functions. | |
532 | */ | |
533 | void init_timer(struct timer_list *timer) | |
534 | { | |
535 | debug_timer_init(timer); | |
536 | __init_timer(timer); | |
537 | } | |
55c888d6 ON |
538 | EXPORT_SYMBOL(init_timer); |
539 | ||
7ad5b3a5 | 540 | void init_timer_deferrable(struct timer_list *timer) |
6e453a67 VP |
541 | { |
542 | init_timer(timer); | |
543 | timer_set_deferrable(timer); | |
544 | } | |
545 | EXPORT_SYMBOL(init_timer_deferrable); | |
546 | ||
55c888d6 | 547 | static inline void detach_timer(struct timer_list *timer, |
82f67cd9 | 548 | int clear_pending) |
55c888d6 ON |
549 | { |
550 | struct list_head *entry = &timer->entry; | |
551 | ||
c6f3a97f TG |
552 | debug_timer_deactivate(timer); |
553 | ||
55c888d6 ON |
554 | __list_del(entry->prev, entry->next); |
555 | if (clear_pending) | |
556 | entry->next = NULL; | |
557 | entry->prev = LIST_POISON2; | |
558 | } | |
559 | ||
560 | /* | |
3691c519 | 561 | * We are using hashed locking: holding per_cpu(tvec_bases).lock |
55c888d6 ON |
562 | * means that all timers which are tied to this base via timer->base are |
563 | * locked, and the base itself is locked too. | |
564 | * | |
565 | * So __run_timers/migrate_timers can safely modify all timers which could | |
566 | * be found on ->tvX lists. | |
567 | * | |
568 | * When the timer's base is locked, and the timer removed from list, it is | |
569 | * possible to set timer->base = NULL and drop the lock: the timer remains | |
570 | * locked. | |
571 | */ | |
a6fa8e5a | 572 | static struct tvec_base *lock_timer_base(struct timer_list *timer, |
55c888d6 | 573 | unsigned long *flags) |
89e7e374 | 574 | __acquires(timer->base->lock) |
55c888d6 | 575 | { |
a6fa8e5a | 576 | struct tvec_base *base; |
55c888d6 ON |
577 | |
578 | for (;;) { | |
a6fa8e5a | 579 | struct tvec_base *prelock_base = timer->base; |
6e453a67 | 580 | base = tbase_get_base(prelock_base); |
55c888d6 ON |
581 | if (likely(base != NULL)) { |
582 | spin_lock_irqsave(&base->lock, *flags); | |
6e453a67 | 583 | if (likely(prelock_base == timer->base)) |
55c888d6 ON |
584 | return base; |
585 | /* The timer has migrated to another CPU */ | |
586 | spin_unlock_irqrestore(&base->lock, *flags); | |
587 | } | |
588 | cpu_relax(); | |
589 | } | |
590 | } | |
591 | ||
74019224 IM |
592 | static inline int |
593 | __mod_timer(struct timer_list *timer, unsigned long expires, bool pending_only) | |
1da177e4 | 594 | { |
a6fa8e5a | 595 | struct tvec_base *base, *new_base; |
1da177e4 | 596 | unsigned long flags; |
74019224 IM |
597 | int ret; |
598 | ||
599 | ret = 0; | |
1da177e4 | 600 | |
82f67cd9 | 601 | timer_stats_timer_set_start_info(timer); |
1da177e4 | 602 | BUG_ON(!timer->function); |
1da177e4 | 603 | |
55c888d6 ON |
604 | base = lock_timer_base(timer, &flags); |
605 | ||
606 | if (timer_pending(timer)) { | |
607 | detach_timer(timer, 0); | |
608 | ret = 1; | |
74019224 IM |
609 | } else { |
610 | if (pending_only) | |
611 | goto out_unlock; | |
55c888d6 ON |
612 | } |
613 | ||
c6f3a97f TG |
614 | debug_timer_activate(timer); |
615 | ||
a4a6198b | 616 | new_base = __get_cpu_var(tvec_bases); |
1da177e4 | 617 | |
3691c519 | 618 | if (base != new_base) { |
1da177e4 | 619 | /* |
55c888d6 ON |
620 | * We are trying to schedule the timer on the local CPU. |
621 | * However we can't change timer's base while it is running, | |
622 | * otherwise del_timer_sync() can't detect that the timer's | |
623 | * handler yet has not finished. This also guarantees that | |
624 | * the timer is serialized wrt itself. | |
1da177e4 | 625 | */ |
a2c348fe | 626 | if (likely(base->running_timer != timer)) { |
55c888d6 | 627 | /* See the comment in lock_timer_base() */ |
6e453a67 | 628 | timer_set_base(timer, NULL); |
55c888d6 | 629 | spin_unlock(&base->lock); |
a2c348fe ON |
630 | base = new_base; |
631 | spin_lock(&base->lock); | |
6e453a67 | 632 | timer_set_base(timer, base); |
1da177e4 LT |
633 | } |
634 | } | |
635 | ||
1da177e4 | 636 | timer->expires = expires; |
a2c348fe | 637 | internal_add_timer(base, timer); |
74019224 IM |
638 | |
639 | out_unlock: | |
a2c348fe | 640 | spin_unlock_irqrestore(&base->lock, flags); |
1da177e4 LT |
641 | |
642 | return ret; | |
643 | } | |
644 | ||
2aae4a10 | 645 | /** |
74019224 IM |
646 | * mod_timer_pending - modify a pending timer's timeout |
647 | * @timer: the pending timer to be modified | |
648 | * @expires: new timeout in jiffies | |
1da177e4 | 649 | * |
74019224 IM |
650 | * mod_timer_pending() is the same for pending timers as mod_timer(), |
651 | * but will not re-activate and modify already deleted timers. | |
652 | * | |
653 | * It is useful for unserialized use of timers. | |
1da177e4 | 654 | */ |
74019224 | 655 | int mod_timer_pending(struct timer_list *timer, unsigned long expires) |
1da177e4 | 656 | { |
74019224 | 657 | return __mod_timer(timer, expires, true); |
1da177e4 | 658 | } |
74019224 | 659 | EXPORT_SYMBOL(mod_timer_pending); |
1da177e4 | 660 | |
2aae4a10 | 661 | /** |
1da177e4 LT |
662 | * mod_timer - modify a timer's timeout |
663 | * @timer: the timer to be modified | |
2aae4a10 | 664 | * @expires: new timeout in jiffies |
1da177e4 | 665 | * |
72fd4a35 | 666 | * mod_timer() is a more efficient way to update the expire field of an |
1da177e4 LT |
667 | * active timer (if the timer is inactive it will be activated) |
668 | * | |
669 | * mod_timer(timer, expires) is equivalent to: | |
670 | * | |
671 | * del_timer(timer); timer->expires = expires; add_timer(timer); | |
672 | * | |
673 | * Note that if there are multiple unserialized concurrent users of the | |
674 | * same timer, then mod_timer() is the only safe way to modify the timeout, | |
675 | * since add_timer() cannot modify an already running timer. | |
676 | * | |
677 | * The function returns whether it has modified a pending timer or not. | |
678 | * (ie. mod_timer() of an inactive timer returns 0, mod_timer() of an | |
679 | * active timer returns 1.) | |
680 | */ | |
681 | int mod_timer(struct timer_list *timer, unsigned long expires) | |
682 | { | |
1da177e4 LT |
683 | /* |
684 | * This is a common optimization triggered by the | |
685 | * networking code - if the timer is re-modified | |
686 | * to be the same thing then just return: | |
687 | */ | |
688 | if (timer->expires == expires && timer_pending(timer)) | |
689 | return 1; | |
690 | ||
74019224 | 691 | return __mod_timer(timer, expires, false); |
1da177e4 | 692 | } |
1da177e4 LT |
693 | EXPORT_SYMBOL(mod_timer); |
694 | ||
74019224 IM |
695 | /** |
696 | * add_timer - start a timer | |
697 | * @timer: the timer to be added | |
698 | * | |
699 | * The kernel will do a ->function(->data) callback from the | |
700 | * timer interrupt at the ->expires point in the future. The | |
701 | * current time is 'jiffies'. | |
702 | * | |
703 | * The timer's ->expires, ->function (and if the handler uses it, ->data) | |
704 | * fields must be set prior calling this function. | |
705 | * | |
706 | * Timers with an ->expires field in the past will be executed in the next | |
707 | * timer tick. | |
708 | */ | |
709 | void add_timer(struct timer_list *timer) | |
710 | { | |
711 | BUG_ON(timer_pending(timer)); | |
712 | mod_timer(timer, timer->expires); | |
713 | } | |
714 | EXPORT_SYMBOL(add_timer); | |
715 | ||
716 | /** | |
717 | * add_timer_on - start a timer on a particular CPU | |
718 | * @timer: the timer to be added | |
719 | * @cpu: the CPU to start it on | |
720 | * | |
721 | * This is not very scalable on SMP. Double adds are not possible. | |
722 | */ | |
723 | void add_timer_on(struct timer_list *timer, int cpu) | |
724 | { | |
725 | struct tvec_base *base = per_cpu(tvec_bases, cpu); | |
726 | unsigned long flags; | |
727 | ||
728 | timer_stats_timer_set_start_info(timer); | |
729 | BUG_ON(timer_pending(timer) || !timer->function); | |
730 | spin_lock_irqsave(&base->lock, flags); | |
731 | timer_set_base(timer, base); | |
732 | debug_timer_activate(timer); | |
733 | internal_add_timer(base, timer); | |
734 | /* | |
735 | * Check whether the other CPU is idle and needs to be | |
736 | * triggered to reevaluate the timer wheel when nohz is | |
737 | * active. We are protected against the other CPU fiddling | |
738 | * with the timer by holding the timer base lock. This also | |
739 | * makes sure that a CPU on the way to idle can not evaluate | |
740 | * the timer wheel. | |
741 | */ | |
742 | wake_up_idle_cpu(cpu); | |
743 | spin_unlock_irqrestore(&base->lock, flags); | |
744 | } | |
745 | ||
2aae4a10 | 746 | /** |
1da177e4 LT |
747 | * del_timer - deactive a timer. |
748 | * @timer: the timer to be deactivated | |
749 | * | |
750 | * del_timer() deactivates a timer - this works on both active and inactive | |
751 | * timers. | |
752 | * | |
753 | * The function returns whether it has deactivated a pending timer or not. | |
754 | * (ie. del_timer() of an inactive timer returns 0, del_timer() of an | |
755 | * active timer returns 1.) | |
756 | */ | |
757 | int del_timer(struct timer_list *timer) | |
758 | { | |
a6fa8e5a | 759 | struct tvec_base *base; |
1da177e4 | 760 | unsigned long flags; |
55c888d6 | 761 | int ret = 0; |
1da177e4 | 762 | |
82f67cd9 | 763 | timer_stats_timer_clear_start_info(timer); |
55c888d6 ON |
764 | if (timer_pending(timer)) { |
765 | base = lock_timer_base(timer, &flags); | |
766 | if (timer_pending(timer)) { | |
767 | detach_timer(timer, 1); | |
768 | ret = 1; | |
769 | } | |
1da177e4 | 770 | spin_unlock_irqrestore(&base->lock, flags); |
1da177e4 | 771 | } |
1da177e4 | 772 | |
55c888d6 | 773 | return ret; |
1da177e4 | 774 | } |
1da177e4 LT |
775 | EXPORT_SYMBOL(del_timer); |
776 | ||
777 | #ifdef CONFIG_SMP | |
2aae4a10 REB |
778 | /** |
779 | * try_to_del_timer_sync - Try to deactivate a timer | |
780 | * @timer: timer do del | |
781 | * | |
fd450b73 ON |
782 | * This function tries to deactivate a timer. Upon successful (ret >= 0) |
783 | * exit the timer is not queued and the handler is not running on any CPU. | |
784 | * | |
785 | * It must not be called from interrupt contexts. | |
786 | */ | |
787 | int try_to_del_timer_sync(struct timer_list *timer) | |
788 | { | |
a6fa8e5a | 789 | struct tvec_base *base; |
fd450b73 ON |
790 | unsigned long flags; |
791 | int ret = -1; | |
792 | ||
793 | base = lock_timer_base(timer, &flags); | |
794 | ||
795 | if (base->running_timer == timer) | |
796 | goto out; | |
797 | ||
798 | ret = 0; | |
799 | if (timer_pending(timer)) { | |
800 | detach_timer(timer, 1); | |
801 | ret = 1; | |
802 | } | |
803 | out: | |
804 | spin_unlock_irqrestore(&base->lock, flags); | |
805 | ||
806 | return ret; | |
807 | } | |
e19dff1f DH |
808 | EXPORT_SYMBOL(try_to_del_timer_sync); |
809 | ||
2aae4a10 | 810 | /** |
1da177e4 LT |
811 | * del_timer_sync - deactivate a timer and wait for the handler to finish. |
812 | * @timer: the timer to be deactivated | |
813 | * | |
814 | * This function only differs from del_timer() on SMP: besides deactivating | |
815 | * the timer it also makes sure the handler has finished executing on other | |
816 | * CPUs. | |
817 | * | |
72fd4a35 | 818 | * Synchronization rules: Callers must prevent restarting of the timer, |
1da177e4 LT |
819 | * otherwise this function is meaningless. It must not be called from |
820 | * interrupt contexts. The caller must not hold locks which would prevent | |
55c888d6 ON |
821 | * completion of the timer's handler. The timer's handler must not call |
822 | * add_timer_on(). Upon exit the timer is not queued and the handler is | |
823 | * not running on any CPU. | |
1da177e4 LT |
824 | * |
825 | * The function returns whether it has deactivated a pending timer or not. | |
1da177e4 LT |
826 | */ |
827 | int del_timer_sync(struct timer_list *timer) | |
828 | { | |
fd450b73 ON |
829 | for (;;) { |
830 | int ret = try_to_del_timer_sync(timer); | |
831 | if (ret >= 0) | |
832 | return ret; | |
a0009652 | 833 | cpu_relax(); |
fd450b73 | 834 | } |
1da177e4 | 835 | } |
55c888d6 | 836 | EXPORT_SYMBOL(del_timer_sync); |
1da177e4 LT |
837 | #endif |
838 | ||
a6fa8e5a | 839 | static int cascade(struct tvec_base *base, struct tvec *tv, int index) |
1da177e4 LT |
840 | { |
841 | /* cascade all the timers from tv up one level */ | |
3439dd86 P |
842 | struct timer_list *timer, *tmp; |
843 | struct list_head tv_list; | |
844 | ||
845 | list_replace_init(tv->vec + index, &tv_list); | |
1da177e4 | 846 | |
1da177e4 | 847 | /* |
3439dd86 P |
848 | * We are removing _all_ timers from the list, so we |
849 | * don't have to detach them individually. | |
1da177e4 | 850 | */ |
3439dd86 | 851 | list_for_each_entry_safe(timer, tmp, &tv_list, entry) { |
6e453a67 | 852 | BUG_ON(tbase_get_base(timer->base) != base); |
3439dd86 | 853 | internal_add_timer(base, timer); |
1da177e4 | 854 | } |
1da177e4 LT |
855 | |
856 | return index; | |
857 | } | |
858 | ||
2aae4a10 REB |
859 | #define INDEX(N) ((base->timer_jiffies >> (TVR_BITS + (N) * TVN_BITS)) & TVN_MASK) |
860 | ||
861 | /** | |
1da177e4 LT |
862 | * __run_timers - run all expired timers (if any) on this CPU. |
863 | * @base: the timer vector to be processed. | |
864 | * | |
865 | * This function cascades all vectors and executes all expired timer | |
866 | * vectors. | |
867 | */ | |
a6fa8e5a | 868 | static inline void __run_timers(struct tvec_base *base) |
1da177e4 LT |
869 | { |
870 | struct timer_list *timer; | |
871 | ||
3691c519 | 872 | spin_lock_irq(&base->lock); |
1da177e4 | 873 | while (time_after_eq(jiffies, base->timer_jiffies)) { |
626ab0e6 | 874 | struct list_head work_list; |
1da177e4 | 875 | struct list_head *head = &work_list; |
6819457d | 876 | int index = base->timer_jiffies & TVR_MASK; |
626ab0e6 | 877 | |
1da177e4 LT |
878 | /* |
879 | * Cascade timers: | |
880 | */ | |
881 | if (!index && | |
882 | (!cascade(base, &base->tv2, INDEX(0))) && | |
883 | (!cascade(base, &base->tv3, INDEX(1))) && | |
884 | !cascade(base, &base->tv4, INDEX(2))) | |
885 | cascade(base, &base->tv5, INDEX(3)); | |
626ab0e6 ON |
886 | ++base->timer_jiffies; |
887 | list_replace_init(base->tv1.vec + index, &work_list); | |
55c888d6 | 888 | while (!list_empty(head)) { |
1da177e4 LT |
889 | void (*fn)(unsigned long); |
890 | unsigned long data; | |
891 | ||
b5e61818 | 892 | timer = list_first_entry(head, struct timer_list,entry); |
6819457d TG |
893 | fn = timer->function; |
894 | data = timer->data; | |
1da177e4 | 895 | |
82f67cd9 IM |
896 | timer_stats_account_timer(timer); |
897 | ||
1da177e4 | 898 | set_running_timer(base, timer); |
55c888d6 | 899 | detach_timer(timer, 1); |
3691c519 | 900 | spin_unlock_irq(&base->lock); |
1da177e4 | 901 | { |
be5b4fbd | 902 | int preempt_count = preempt_count(); |
1da177e4 LT |
903 | fn(data); |
904 | if (preempt_count != preempt_count()) { | |
4c9dc641 | 905 | printk(KERN_ERR "huh, entered %p " |
be5b4fbd JJ |
906 | "with preempt_count %08x, exited" |
907 | " with %08x?\n", | |
908 | fn, preempt_count, | |
909 | preempt_count()); | |
1da177e4 LT |
910 | BUG(); |
911 | } | |
912 | } | |
3691c519 | 913 | spin_lock_irq(&base->lock); |
1da177e4 LT |
914 | } |
915 | } | |
916 | set_running_timer(base, NULL); | |
3691c519 | 917 | spin_unlock_irq(&base->lock); |
1da177e4 LT |
918 | } |
919 | ||
ee9c5785 | 920 | #ifdef CONFIG_NO_HZ |
1da177e4 LT |
921 | /* |
922 | * Find out when the next timer event is due to happen. This | |
923 | * is used on S/390 to stop all activity when a cpus is idle. | |
924 | * This functions needs to be called disabled. | |
925 | */ | |
a6fa8e5a | 926 | static unsigned long __next_timer_interrupt(struct tvec_base *base) |
1da177e4 | 927 | { |
1cfd6849 | 928 | unsigned long timer_jiffies = base->timer_jiffies; |
eaad084b | 929 | unsigned long expires = timer_jiffies + NEXT_TIMER_MAX_DELTA; |
1cfd6849 | 930 | int index, slot, array, found = 0; |
1da177e4 | 931 | struct timer_list *nte; |
a6fa8e5a | 932 | struct tvec *varray[4]; |
1da177e4 LT |
933 | |
934 | /* Look for timer events in tv1. */ | |
1cfd6849 | 935 | index = slot = timer_jiffies & TVR_MASK; |
1da177e4 | 936 | do { |
1cfd6849 | 937 | list_for_each_entry(nte, base->tv1.vec + slot, entry) { |
6819457d TG |
938 | if (tbase_get_deferrable(nte->base)) |
939 | continue; | |
6e453a67 | 940 | |
1cfd6849 | 941 | found = 1; |
1da177e4 | 942 | expires = nte->expires; |
1cfd6849 TG |
943 | /* Look at the cascade bucket(s)? */ |
944 | if (!index || slot < index) | |
945 | goto cascade; | |
946 | return expires; | |
1da177e4 | 947 | } |
1cfd6849 TG |
948 | slot = (slot + 1) & TVR_MASK; |
949 | } while (slot != index); | |
950 | ||
951 | cascade: | |
952 | /* Calculate the next cascade event */ | |
953 | if (index) | |
954 | timer_jiffies += TVR_SIZE - index; | |
955 | timer_jiffies >>= TVR_BITS; | |
1da177e4 LT |
956 | |
957 | /* Check tv2-tv5. */ | |
958 | varray[0] = &base->tv2; | |
959 | varray[1] = &base->tv3; | |
960 | varray[2] = &base->tv4; | |
961 | varray[3] = &base->tv5; | |
1cfd6849 TG |
962 | |
963 | for (array = 0; array < 4; array++) { | |
a6fa8e5a | 964 | struct tvec *varp = varray[array]; |
1cfd6849 TG |
965 | |
966 | index = slot = timer_jiffies & TVN_MASK; | |
1da177e4 | 967 | do { |
1cfd6849 TG |
968 | list_for_each_entry(nte, varp->vec + slot, entry) { |
969 | found = 1; | |
1da177e4 LT |
970 | if (time_before(nte->expires, expires)) |
971 | expires = nte->expires; | |
1cfd6849 TG |
972 | } |
973 | /* | |
974 | * Do we still search for the first timer or are | |
975 | * we looking up the cascade buckets ? | |
976 | */ | |
977 | if (found) { | |
978 | /* Look at the cascade bucket(s)? */ | |
979 | if (!index || slot < index) | |
980 | break; | |
981 | return expires; | |
982 | } | |
983 | slot = (slot + 1) & TVN_MASK; | |
984 | } while (slot != index); | |
985 | ||
986 | if (index) | |
987 | timer_jiffies += TVN_SIZE - index; | |
988 | timer_jiffies >>= TVN_BITS; | |
1da177e4 | 989 | } |
1cfd6849 TG |
990 | return expires; |
991 | } | |
69239749 | 992 | |
1cfd6849 TG |
993 | /* |
994 | * Check, if the next hrtimer event is before the next timer wheel | |
995 | * event: | |
996 | */ | |
997 | static unsigned long cmp_next_hrtimer_event(unsigned long now, | |
998 | unsigned long expires) | |
999 | { | |
1000 | ktime_t hr_delta = hrtimer_get_next_event(); | |
1001 | struct timespec tsdelta; | |
9501b6cf | 1002 | unsigned long delta; |
1cfd6849 TG |
1003 | |
1004 | if (hr_delta.tv64 == KTIME_MAX) | |
1005 | return expires; | |
0662b713 | 1006 | |
9501b6cf TG |
1007 | /* |
1008 | * Expired timer available, let it expire in the next tick | |
1009 | */ | |
1010 | if (hr_delta.tv64 <= 0) | |
1011 | return now + 1; | |
69239749 | 1012 | |
1cfd6849 | 1013 | tsdelta = ktime_to_timespec(hr_delta); |
9501b6cf | 1014 | delta = timespec_to_jiffies(&tsdelta); |
eaad084b TG |
1015 | |
1016 | /* | |
1017 | * Limit the delta to the max value, which is checked in | |
1018 | * tick_nohz_stop_sched_tick(): | |
1019 | */ | |
1020 | if (delta > NEXT_TIMER_MAX_DELTA) | |
1021 | delta = NEXT_TIMER_MAX_DELTA; | |
1022 | ||
9501b6cf TG |
1023 | /* |
1024 | * Take rounding errors in to account and make sure, that it | |
1025 | * expires in the next tick. Otherwise we go into an endless | |
1026 | * ping pong due to tick_nohz_stop_sched_tick() retriggering | |
1027 | * the timer softirq | |
1028 | */ | |
1029 | if (delta < 1) | |
1030 | delta = 1; | |
1031 | now += delta; | |
1cfd6849 TG |
1032 | if (time_before(now, expires)) |
1033 | return now; | |
1da177e4 LT |
1034 | return expires; |
1035 | } | |
1cfd6849 TG |
1036 | |
1037 | /** | |
8dce39c2 | 1038 | * get_next_timer_interrupt - return the jiffy of the next pending timer |
05fb6bf0 | 1039 | * @now: current time (in jiffies) |
1cfd6849 | 1040 | */ |
fd064b9b | 1041 | unsigned long get_next_timer_interrupt(unsigned long now) |
1cfd6849 | 1042 | { |
a6fa8e5a | 1043 | struct tvec_base *base = __get_cpu_var(tvec_bases); |
fd064b9b | 1044 | unsigned long expires; |
1cfd6849 TG |
1045 | |
1046 | spin_lock(&base->lock); | |
1047 | expires = __next_timer_interrupt(base); | |
1048 | spin_unlock(&base->lock); | |
1049 | ||
1050 | if (time_before_eq(expires, now)) | |
1051 | return now; | |
1052 | ||
1053 | return cmp_next_hrtimer_event(now, expires); | |
1054 | } | |
1da177e4 LT |
1055 | #endif |
1056 | ||
1da177e4 | 1057 | /* |
5b4db0c2 | 1058 | * Called from the timer interrupt handler to charge one tick to the current |
1da177e4 LT |
1059 | * process. user_tick is 1 if the tick is user time, 0 for system. |
1060 | */ | |
1061 | void update_process_times(int user_tick) | |
1062 | { | |
1063 | struct task_struct *p = current; | |
1064 | int cpu = smp_processor_id(); | |
1065 | ||
1066 | /* Note: this timer irq context must be accounted for as well. */ | |
fa13a5a1 | 1067 | account_process_tick(p, user_tick); |
1da177e4 LT |
1068 | run_local_timers(); |
1069 | if (rcu_pending(cpu)) | |
1070 | rcu_check_callbacks(cpu, user_tick); | |
b845b517 | 1071 | printk_tick(); |
1da177e4 | 1072 | scheduler_tick(); |
6819457d | 1073 | run_posix_cpu_timers(p); |
1da177e4 LT |
1074 | } |
1075 | ||
1076 | /* | |
1077 | * Nr of active tasks - counted in fixed-point numbers | |
1078 | */ | |
1079 | static unsigned long count_active_tasks(void) | |
1080 | { | |
db1b1fef | 1081 | return nr_active() * FIXED_1; |
1da177e4 LT |
1082 | } |
1083 | ||
1084 | /* | |
1085 | * Hmm.. Changed this, as the GNU make sources (load.c) seems to | |
1086 | * imply that avenrun[] is the standard name for this kind of thing. | |
1087 | * Nothing else seems to be standardized: the fractional size etc | |
1088 | * all seem to differ on different machines. | |
1089 | * | |
1090 | * Requires xtime_lock to access. | |
1091 | */ | |
1092 | unsigned long avenrun[3]; | |
1093 | ||
1094 | EXPORT_SYMBOL(avenrun); | |
1095 | ||
1096 | /* | |
1097 | * calc_load - given tick count, update the avenrun load estimates. | |
1098 | * This is called while holding a write_lock on xtime_lock. | |
1099 | */ | |
1100 | static inline void calc_load(unsigned long ticks) | |
1101 | { | |
1102 | unsigned long active_tasks; /* fixed-point */ | |
1103 | static int count = LOAD_FREQ; | |
1104 | ||
cd7175ed ED |
1105 | count -= ticks; |
1106 | if (unlikely(count < 0)) { | |
1107 | active_tasks = count_active_tasks(); | |
1108 | do { | |
1109 | CALC_LOAD(avenrun[0], EXP_1, active_tasks); | |
1110 | CALC_LOAD(avenrun[1], EXP_5, active_tasks); | |
1111 | CALC_LOAD(avenrun[2], EXP_15, active_tasks); | |
1112 | count += LOAD_FREQ; | |
1113 | } while (count < 0); | |
1da177e4 LT |
1114 | } |
1115 | } | |
1116 | ||
1da177e4 LT |
1117 | /* |
1118 | * This function runs timers and the timer-tq in bottom half context. | |
1119 | */ | |
1120 | static void run_timer_softirq(struct softirq_action *h) | |
1121 | { | |
a6fa8e5a | 1122 | struct tvec_base *base = __get_cpu_var(tvec_bases); |
1da177e4 | 1123 | |
d3d74453 | 1124 | hrtimer_run_pending(); |
82f67cd9 | 1125 | |
1da177e4 LT |
1126 | if (time_after_eq(jiffies, base->timer_jiffies)) |
1127 | __run_timers(base); | |
1128 | } | |
1129 | ||
1130 | /* | |
1131 | * Called by the local, per-CPU timer interrupt on SMP. | |
1132 | */ | |
1133 | void run_local_timers(void) | |
1134 | { | |
d3d74453 | 1135 | hrtimer_run_queues(); |
1da177e4 | 1136 | raise_softirq(TIMER_SOFTIRQ); |
6687a97d | 1137 | softlockup_tick(); |
1da177e4 LT |
1138 | } |
1139 | ||
1140 | /* | |
1141 | * Called by the timer interrupt. xtime_lock must already be taken | |
1142 | * by the timer IRQ! | |
1143 | */ | |
3171a030 | 1144 | static inline void update_times(unsigned long ticks) |
1da177e4 | 1145 | { |
ad596171 | 1146 | update_wall_time(); |
1da177e4 LT |
1147 | calc_load(ticks); |
1148 | } | |
6819457d | 1149 | |
1da177e4 LT |
1150 | /* |
1151 | * The 64-bit jiffies value is not atomic - you MUST NOT read it | |
1152 | * without sampling the sequence number in xtime_lock. | |
1153 | * jiffies is defined in the linker script... | |
1154 | */ | |
1155 | ||
3171a030 | 1156 | void do_timer(unsigned long ticks) |
1da177e4 | 1157 | { |
3171a030 AN |
1158 | jiffies_64 += ticks; |
1159 | update_times(ticks); | |
1da177e4 LT |
1160 | } |
1161 | ||
1162 | #ifdef __ARCH_WANT_SYS_ALARM | |
1163 | ||
1164 | /* | |
1165 | * For backwards compatibility? This can be done in libc so Alpha | |
1166 | * and all newer ports shouldn't need it. | |
1167 | */ | |
58fd3aa2 | 1168 | SYSCALL_DEFINE1(alarm, unsigned int, seconds) |
1da177e4 | 1169 | { |
c08b8a49 | 1170 | return alarm_setitimer(seconds); |
1da177e4 LT |
1171 | } |
1172 | ||
1173 | #endif | |
1174 | ||
1175 | #ifndef __alpha__ | |
1176 | ||
1177 | /* | |
1178 | * The Alpha uses getxpid, getxuid, and getxgid instead. Maybe this | |
1179 | * should be moved into arch/i386 instead? | |
1180 | */ | |
1181 | ||
1182 | /** | |
1183 | * sys_getpid - return the thread group id of the current process | |
1184 | * | |
1185 | * Note, despite the name, this returns the tgid not the pid. The tgid and | |
1186 | * the pid are identical unless CLONE_THREAD was specified on clone() in | |
1187 | * which case the tgid is the same in all threads of the same group. | |
1188 | * | |
1189 | * This is SMP safe as current->tgid does not change. | |
1190 | */ | |
58fd3aa2 | 1191 | SYSCALL_DEFINE0(getpid) |
1da177e4 | 1192 | { |
b488893a | 1193 | return task_tgid_vnr(current); |
1da177e4 LT |
1194 | } |
1195 | ||
1196 | /* | |
6997a6fa KK |
1197 | * Accessing ->real_parent is not SMP-safe, it could |
1198 | * change from under us. However, we can use a stale | |
1199 | * value of ->real_parent under rcu_read_lock(), see | |
1200 | * release_task()->call_rcu(delayed_put_task_struct). | |
1da177e4 | 1201 | */ |
dbf040d9 | 1202 | SYSCALL_DEFINE0(getppid) |
1da177e4 LT |
1203 | { |
1204 | int pid; | |
1da177e4 | 1205 | |
6997a6fa | 1206 | rcu_read_lock(); |
6c5f3e7b | 1207 | pid = task_tgid_vnr(current->real_parent); |
6997a6fa | 1208 | rcu_read_unlock(); |
1da177e4 | 1209 | |
1da177e4 LT |
1210 | return pid; |
1211 | } | |
1212 | ||
dbf040d9 | 1213 | SYSCALL_DEFINE0(getuid) |
1da177e4 LT |
1214 | { |
1215 | /* Only we change this so SMP safe */ | |
76aac0e9 | 1216 | return current_uid(); |
1da177e4 LT |
1217 | } |
1218 | ||
dbf040d9 | 1219 | SYSCALL_DEFINE0(geteuid) |
1da177e4 LT |
1220 | { |
1221 | /* Only we change this so SMP safe */ | |
76aac0e9 | 1222 | return current_euid(); |
1da177e4 LT |
1223 | } |
1224 | ||
dbf040d9 | 1225 | SYSCALL_DEFINE0(getgid) |
1da177e4 LT |
1226 | { |
1227 | /* Only we change this so SMP safe */ | |
76aac0e9 | 1228 | return current_gid(); |
1da177e4 LT |
1229 | } |
1230 | ||
dbf040d9 | 1231 | SYSCALL_DEFINE0(getegid) |
1da177e4 LT |
1232 | { |
1233 | /* Only we change this so SMP safe */ | |
76aac0e9 | 1234 | return current_egid(); |
1da177e4 LT |
1235 | } |
1236 | ||
1237 | #endif | |
1238 | ||
1239 | static void process_timeout(unsigned long __data) | |
1240 | { | |
36c8b586 | 1241 | wake_up_process((struct task_struct *)__data); |
1da177e4 LT |
1242 | } |
1243 | ||
1244 | /** | |
1245 | * schedule_timeout - sleep until timeout | |
1246 | * @timeout: timeout value in jiffies | |
1247 | * | |
1248 | * Make the current task sleep until @timeout jiffies have | |
1249 | * elapsed. The routine will return immediately unless | |
1250 | * the current task state has been set (see set_current_state()). | |
1251 | * | |
1252 | * You can set the task state as follows - | |
1253 | * | |
1254 | * %TASK_UNINTERRUPTIBLE - at least @timeout jiffies are guaranteed to | |
1255 | * pass before the routine returns. The routine will return 0 | |
1256 | * | |
1257 | * %TASK_INTERRUPTIBLE - the routine may return early if a signal is | |
1258 | * delivered to the current task. In this case the remaining time | |
1259 | * in jiffies will be returned, or 0 if the timer expired in time | |
1260 | * | |
1261 | * The current task state is guaranteed to be TASK_RUNNING when this | |
1262 | * routine returns. | |
1263 | * | |
1264 | * Specifying a @timeout value of %MAX_SCHEDULE_TIMEOUT will schedule | |
1265 | * the CPU away without a bound on the timeout. In this case the return | |
1266 | * value will be %MAX_SCHEDULE_TIMEOUT. | |
1267 | * | |
1268 | * In all cases the return value is guaranteed to be non-negative. | |
1269 | */ | |
7ad5b3a5 | 1270 | signed long __sched schedule_timeout(signed long timeout) |
1da177e4 LT |
1271 | { |
1272 | struct timer_list timer; | |
1273 | unsigned long expire; | |
1274 | ||
1275 | switch (timeout) | |
1276 | { | |
1277 | case MAX_SCHEDULE_TIMEOUT: | |
1278 | /* | |
1279 | * These two special cases are useful to be comfortable | |
1280 | * in the caller. Nothing more. We could take | |
1281 | * MAX_SCHEDULE_TIMEOUT from one of the negative value | |
1282 | * but I' d like to return a valid offset (>=0) to allow | |
1283 | * the caller to do everything it want with the retval. | |
1284 | */ | |
1285 | schedule(); | |
1286 | goto out; | |
1287 | default: | |
1288 | /* | |
1289 | * Another bit of PARANOID. Note that the retval will be | |
1290 | * 0 since no piece of kernel is supposed to do a check | |
1291 | * for a negative retval of schedule_timeout() (since it | |
1292 | * should never happens anyway). You just have the printk() | |
1293 | * that will tell you if something is gone wrong and where. | |
1294 | */ | |
5b149bcc | 1295 | if (timeout < 0) { |
1da177e4 | 1296 | printk(KERN_ERR "schedule_timeout: wrong timeout " |
5b149bcc AM |
1297 | "value %lx\n", timeout); |
1298 | dump_stack(); | |
1da177e4 LT |
1299 | current->state = TASK_RUNNING; |
1300 | goto out; | |
1301 | } | |
1302 | } | |
1303 | ||
1304 | expire = timeout + jiffies; | |
1305 | ||
c6f3a97f | 1306 | setup_timer_on_stack(&timer, process_timeout, (unsigned long)current); |
74019224 | 1307 | __mod_timer(&timer, expire, false); |
1da177e4 LT |
1308 | schedule(); |
1309 | del_singleshot_timer_sync(&timer); | |
1310 | ||
c6f3a97f TG |
1311 | /* Remove the timer from the object tracker */ |
1312 | destroy_timer_on_stack(&timer); | |
1313 | ||
1da177e4 LT |
1314 | timeout = expire - jiffies; |
1315 | ||
1316 | out: | |
1317 | return timeout < 0 ? 0 : timeout; | |
1318 | } | |
1da177e4 LT |
1319 | EXPORT_SYMBOL(schedule_timeout); |
1320 | ||
8a1c1757 AM |
1321 | /* |
1322 | * We can use __set_current_state() here because schedule_timeout() calls | |
1323 | * schedule() unconditionally. | |
1324 | */ | |
64ed93a2 NA |
1325 | signed long __sched schedule_timeout_interruptible(signed long timeout) |
1326 | { | |
a5a0d52c AM |
1327 | __set_current_state(TASK_INTERRUPTIBLE); |
1328 | return schedule_timeout(timeout); | |
64ed93a2 NA |
1329 | } |
1330 | EXPORT_SYMBOL(schedule_timeout_interruptible); | |
1331 | ||
294d5cc2 MW |
1332 | signed long __sched schedule_timeout_killable(signed long timeout) |
1333 | { | |
1334 | __set_current_state(TASK_KILLABLE); | |
1335 | return schedule_timeout(timeout); | |
1336 | } | |
1337 | EXPORT_SYMBOL(schedule_timeout_killable); | |
1338 | ||
64ed93a2 NA |
1339 | signed long __sched schedule_timeout_uninterruptible(signed long timeout) |
1340 | { | |
a5a0d52c AM |
1341 | __set_current_state(TASK_UNINTERRUPTIBLE); |
1342 | return schedule_timeout(timeout); | |
64ed93a2 NA |
1343 | } |
1344 | EXPORT_SYMBOL(schedule_timeout_uninterruptible); | |
1345 | ||
1da177e4 | 1346 | /* Thread ID - the internal kernel "pid" */ |
58fd3aa2 | 1347 | SYSCALL_DEFINE0(gettid) |
1da177e4 | 1348 | { |
b488893a | 1349 | return task_pid_vnr(current); |
1da177e4 LT |
1350 | } |
1351 | ||
2aae4a10 | 1352 | /** |
d4d23add | 1353 | * do_sysinfo - fill in sysinfo struct |
2aae4a10 | 1354 | * @info: pointer to buffer to fill |
6819457d | 1355 | */ |
d4d23add | 1356 | int do_sysinfo(struct sysinfo *info) |
1da177e4 | 1357 | { |
1da177e4 LT |
1358 | unsigned long mem_total, sav_total; |
1359 | unsigned int mem_unit, bitcount; | |
1360 | unsigned long seq; | |
1361 | ||
d4d23add | 1362 | memset(info, 0, sizeof(struct sysinfo)); |
1da177e4 LT |
1363 | |
1364 | do { | |
1365 | struct timespec tp; | |
1366 | seq = read_seqbegin(&xtime_lock); | |
1367 | ||
1368 | /* | |
1369 | * This is annoying. The below is the same thing | |
1370 | * posix_get_clock_monotonic() does, but it wants to | |
1371 | * take the lock which we want to cover the loads stuff | |
1372 | * too. | |
1373 | */ | |
1374 | ||
1375 | getnstimeofday(&tp); | |
1376 | tp.tv_sec += wall_to_monotonic.tv_sec; | |
1377 | tp.tv_nsec += wall_to_monotonic.tv_nsec; | |
d6214141 | 1378 | monotonic_to_bootbased(&tp); |
1da177e4 LT |
1379 | if (tp.tv_nsec - NSEC_PER_SEC >= 0) { |
1380 | tp.tv_nsec = tp.tv_nsec - NSEC_PER_SEC; | |
1381 | tp.tv_sec++; | |
1382 | } | |
d4d23add | 1383 | info->uptime = tp.tv_sec + (tp.tv_nsec ? 1 : 0); |
1da177e4 | 1384 | |
d4d23add KM |
1385 | info->loads[0] = avenrun[0] << (SI_LOAD_SHIFT - FSHIFT); |
1386 | info->loads[1] = avenrun[1] << (SI_LOAD_SHIFT - FSHIFT); | |
1387 | info->loads[2] = avenrun[2] << (SI_LOAD_SHIFT - FSHIFT); | |
1da177e4 | 1388 | |
d4d23add | 1389 | info->procs = nr_threads; |
1da177e4 LT |
1390 | } while (read_seqretry(&xtime_lock, seq)); |
1391 | ||
d4d23add KM |
1392 | si_meminfo(info); |
1393 | si_swapinfo(info); | |
1da177e4 LT |
1394 | |
1395 | /* | |
1396 | * If the sum of all the available memory (i.e. ram + swap) | |
1397 | * is less than can be stored in a 32 bit unsigned long then | |
1398 | * we can be binary compatible with 2.2.x kernels. If not, | |
1399 | * well, in that case 2.2.x was broken anyways... | |
1400 | * | |
1401 | * -Erik Andersen <andersee@debian.org> | |
1402 | */ | |
1403 | ||
d4d23add KM |
1404 | mem_total = info->totalram + info->totalswap; |
1405 | if (mem_total < info->totalram || mem_total < info->totalswap) | |
1da177e4 LT |
1406 | goto out; |
1407 | bitcount = 0; | |
d4d23add | 1408 | mem_unit = info->mem_unit; |
1da177e4 LT |
1409 | while (mem_unit > 1) { |
1410 | bitcount++; | |
1411 | mem_unit >>= 1; | |
1412 | sav_total = mem_total; | |
1413 | mem_total <<= 1; | |
1414 | if (mem_total < sav_total) | |
1415 | goto out; | |
1416 | } | |
1417 | ||
1418 | /* | |
1419 | * If mem_total did not overflow, multiply all memory values by | |
d4d23add | 1420 | * info->mem_unit and set it to 1. This leaves things compatible |
1da177e4 LT |
1421 | * with 2.2.x, and also retains compatibility with earlier 2.4.x |
1422 | * kernels... | |
1423 | */ | |
1424 | ||
d4d23add KM |
1425 | info->mem_unit = 1; |
1426 | info->totalram <<= bitcount; | |
1427 | info->freeram <<= bitcount; | |
1428 | info->sharedram <<= bitcount; | |
1429 | info->bufferram <<= bitcount; | |
1430 | info->totalswap <<= bitcount; | |
1431 | info->freeswap <<= bitcount; | |
1432 | info->totalhigh <<= bitcount; | |
1433 | info->freehigh <<= bitcount; | |
1434 | ||
1435 | out: | |
1436 | return 0; | |
1437 | } | |
1438 | ||
1e7bfb21 | 1439 | SYSCALL_DEFINE1(sysinfo, struct sysinfo __user *, info) |
d4d23add KM |
1440 | { |
1441 | struct sysinfo val; | |
1442 | ||
1443 | do_sysinfo(&val); | |
1da177e4 | 1444 | |
1da177e4 LT |
1445 | if (copy_to_user(info, &val, sizeof(struct sysinfo))) |
1446 | return -EFAULT; | |
1447 | ||
1448 | return 0; | |
1449 | } | |
1450 | ||
b4be6258 | 1451 | static int __cpuinit init_timers_cpu(int cpu) |
1da177e4 LT |
1452 | { |
1453 | int j; | |
a6fa8e5a | 1454 | struct tvec_base *base; |
b4be6258 | 1455 | static char __cpuinitdata tvec_base_done[NR_CPUS]; |
55c888d6 | 1456 | |
ba6edfcd | 1457 | if (!tvec_base_done[cpu]) { |
a4a6198b JB |
1458 | static char boot_done; |
1459 | ||
a4a6198b | 1460 | if (boot_done) { |
ba6edfcd AM |
1461 | /* |
1462 | * The APs use this path later in boot | |
1463 | */ | |
94f6030c CL |
1464 | base = kmalloc_node(sizeof(*base), |
1465 | GFP_KERNEL | __GFP_ZERO, | |
a4a6198b JB |
1466 | cpu_to_node(cpu)); |
1467 | if (!base) | |
1468 | return -ENOMEM; | |
6e453a67 VP |
1469 | |
1470 | /* Make sure that tvec_base is 2 byte aligned */ | |
1471 | if (tbase_get_deferrable(base)) { | |
1472 | WARN_ON(1); | |
1473 | kfree(base); | |
1474 | return -ENOMEM; | |
1475 | } | |
ba6edfcd | 1476 | per_cpu(tvec_bases, cpu) = base; |
a4a6198b | 1477 | } else { |
ba6edfcd AM |
1478 | /* |
1479 | * This is for the boot CPU - we use compile-time | |
1480 | * static initialisation because per-cpu memory isn't | |
1481 | * ready yet and because the memory allocators are not | |
1482 | * initialised either. | |
1483 | */ | |
a4a6198b | 1484 | boot_done = 1; |
ba6edfcd | 1485 | base = &boot_tvec_bases; |
a4a6198b | 1486 | } |
ba6edfcd AM |
1487 | tvec_base_done[cpu] = 1; |
1488 | } else { | |
1489 | base = per_cpu(tvec_bases, cpu); | |
a4a6198b | 1490 | } |
ba6edfcd | 1491 | |
3691c519 | 1492 | spin_lock_init(&base->lock); |
d730e882 | 1493 | |
1da177e4 LT |
1494 | for (j = 0; j < TVN_SIZE; j++) { |
1495 | INIT_LIST_HEAD(base->tv5.vec + j); | |
1496 | INIT_LIST_HEAD(base->tv4.vec + j); | |
1497 | INIT_LIST_HEAD(base->tv3.vec + j); | |
1498 | INIT_LIST_HEAD(base->tv2.vec + j); | |
1499 | } | |
1500 | for (j = 0; j < TVR_SIZE; j++) | |
1501 | INIT_LIST_HEAD(base->tv1.vec + j); | |
1502 | ||
1503 | base->timer_jiffies = jiffies; | |
a4a6198b | 1504 | return 0; |
1da177e4 LT |
1505 | } |
1506 | ||
1507 | #ifdef CONFIG_HOTPLUG_CPU | |
a6fa8e5a | 1508 | static void migrate_timer_list(struct tvec_base *new_base, struct list_head *head) |
1da177e4 LT |
1509 | { |
1510 | struct timer_list *timer; | |
1511 | ||
1512 | while (!list_empty(head)) { | |
b5e61818 | 1513 | timer = list_first_entry(head, struct timer_list, entry); |
55c888d6 | 1514 | detach_timer(timer, 0); |
6e453a67 | 1515 | timer_set_base(timer, new_base); |
1da177e4 | 1516 | internal_add_timer(new_base, timer); |
1da177e4 | 1517 | } |
1da177e4 LT |
1518 | } |
1519 | ||
48ccf3da | 1520 | static void __cpuinit migrate_timers(int cpu) |
1da177e4 | 1521 | { |
a6fa8e5a PM |
1522 | struct tvec_base *old_base; |
1523 | struct tvec_base *new_base; | |
1da177e4 LT |
1524 | int i; |
1525 | ||
1526 | BUG_ON(cpu_online(cpu)); | |
a4a6198b JB |
1527 | old_base = per_cpu(tvec_bases, cpu); |
1528 | new_base = get_cpu_var(tvec_bases); | |
d82f0b0f ON |
1529 | /* |
1530 | * The caller is globally serialized and nobody else | |
1531 | * takes two locks at once, deadlock is not possible. | |
1532 | */ | |
1533 | spin_lock_irq(&new_base->lock); | |
0d180406 | 1534 | spin_lock_nested(&old_base->lock, SINGLE_DEPTH_NESTING); |
3691c519 ON |
1535 | |
1536 | BUG_ON(old_base->running_timer); | |
1da177e4 | 1537 | |
1da177e4 | 1538 | for (i = 0; i < TVR_SIZE; i++) |
55c888d6 ON |
1539 | migrate_timer_list(new_base, old_base->tv1.vec + i); |
1540 | for (i = 0; i < TVN_SIZE; i++) { | |
1541 | migrate_timer_list(new_base, old_base->tv2.vec + i); | |
1542 | migrate_timer_list(new_base, old_base->tv3.vec + i); | |
1543 | migrate_timer_list(new_base, old_base->tv4.vec + i); | |
1544 | migrate_timer_list(new_base, old_base->tv5.vec + i); | |
1545 | } | |
1546 | ||
0d180406 | 1547 | spin_unlock(&old_base->lock); |
d82f0b0f | 1548 | spin_unlock_irq(&new_base->lock); |
1da177e4 | 1549 | put_cpu_var(tvec_bases); |
1da177e4 LT |
1550 | } |
1551 | #endif /* CONFIG_HOTPLUG_CPU */ | |
1552 | ||
8c78f307 | 1553 | static int __cpuinit timer_cpu_notify(struct notifier_block *self, |
1da177e4 LT |
1554 | unsigned long action, void *hcpu) |
1555 | { | |
1556 | long cpu = (long)hcpu; | |
1557 | switch(action) { | |
1558 | case CPU_UP_PREPARE: | |
8bb78442 | 1559 | case CPU_UP_PREPARE_FROZEN: |
a4a6198b JB |
1560 | if (init_timers_cpu(cpu) < 0) |
1561 | return NOTIFY_BAD; | |
1da177e4 LT |
1562 | break; |
1563 | #ifdef CONFIG_HOTPLUG_CPU | |
1564 | case CPU_DEAD: | |
8bb78442 | 1565 | case CPU_DEAD_FROZEN: |
1da177e4 LT |
1566 | migrate_timers(cpu); |
1567 | break; | |
1568 | #endif | |
1569 | default: | |
1570 | break; | |
1571 | } | |
1572 | return NOTIFY_OK; | |
1573 | } | |
1574 | ||
8c78f307 | 1575 | static struct notifier_block __cpuinitdata timers_nb = { |
1da177e4 LT |
1576 | .notifier_call = timer_cpu_notify, |
1577 | }; | |
1578 | ||
1579 | ||
1580 | void __init init_timers(void) | |
1581 | { | |
07dccf33 | 1582 | int err = timer_cpu_notify(&timers_nb, (unsigned long)CPU_UP_PREPARE, |
1da177e4 | 1583 | (void *)(long)smp_processor_id()); |
07dccf33 | 1584 | |
82f67cd9 IM |
1585 | init_timer_stats(); |
1586 | ||
07dccf33 | 1587 | BUG_ON(err == NOTIFY_BAD); |
1da177e4 | 1588 | register_cpu_notifier(&timers_nb); |
962cf36c | 1589 | open_softirq(TIMER_SOFTIRQ, run_timer_softirq); |
1da177e4 LT |
1590 | } |
1591 | ||
1da177e4 LT |
1592 | /** |
1593 | * msleep - sleep safely even with waitqueue interruptions | |
1594 | * @msecs: Time in milliseconds to sleep for | |
1595 | */ | |
1596 | void msleep(unsigned int msecs) | |
1597 | { | |
1598 | unsigned long timeout = msecs_to_jiffies(msecs) + 1; | |
1599 | ||
75bcc8c5 NA |
1600 | while (timeout) |
1601 | timeout = schedule_timeout_uninterruptible(timeout); | |
1da177e4 LT |
1602 | } |
1603 | ||
1604 | EXPORT_SYMBOL(msleep); | |
1605 | ||
1606 | /** | |
96ec3efd | 1607 | * msleep_interruptible - sleep waiting for signals |
1da177e4 LT |
1608 | * @msecs: Time in milliseconds to sleep for |
1609 | */ | |
1610 | unsigned long msleep_interruptible(unsigned int msecs) | |
1611 | { | |
1612 | unsigned long timeout = msecs_to_jiffies(msecs) + 1; | |
1613 | ||
75bcc8c5 NA |
1614 | while (timeout && !signal_pending(current)) |
1615 | timeout = schedule_timeout_interruptible(timeout); | |
1da177e4 LT |
1616 | return jiffies_to_msecs(timeout); |
1617 | } | |
1618 | ||
1619 | EXPORT_SYMBOL(msleep_interruptible); |