]>
Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * linux/kernel/timer.c | |
3 | * | |
4a22f166 | 4 | * Kernel internal timers |
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> | |
9984de1a | 23 | #include <linux/export.h> |
1da177e4 LT |
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> |
e360adbe | 40 | #include <linux/irq_work.h> |
eea08f32 | 41 | #include <linux/sched.h> |
cf4aebc2 | 42 | #include <linux/sched/sysctl.h> |
5a0e3ad6 | 43 | #include <linux/slab.h> |
1a0df594 | 44 | #include <linux/compat.h> |
1da177e4 LT |
45 | |
46 | #include <asm/uaccess.h> | |
47 | #include <asm/unistd.h> | |
48 | #include <asm/div64.h> | |
49 | #include <asm/timex.h> | |
50 | #include <asm/io.h> | |
51 | ||
c1ad348b TG |
52 | #include "tick-internal.h" |
53 | ||
2b022e3d XG |
54 | #define CREATE_TRACE_POINTS |
55 | #include <trace/events/timer.h> | |
56 | ||
40747ffa | 57 | __visible u64 jiffies_64 __cacheline_aligned_in_smp = INITIAL_JIFFIES; |
ecea8d19 TG |
58 | |
59 | EXPORT_SYMBOL(jiffies_64); | |
60 | ||
1da177e4 LT |
61 | /* |
62 | * per-CPU timer vector definitions: | |
63 | */ | |
1da177e4 LT |
64 | #define TVN_BITS (CONFIG_BASE_SMALL ? 4 : 6) |
65 | #define TVR_BITS (CONFIG_BASE_SMALL ? 6 : 8) | |
66 | #define TVN_SIZE (1 << TVN_BITS) | |
67 | #define TVR_SIZE (1 << TVR_BITS) | |
68 | #define TVN_MASK (TVN_SIZE - 1) | |
69 | #define TVR_MASK (TVR_SIZE - 1) | |
26cff4e2 | 70 | #define MAX_TVAL ((unsigned long)((1ULL << (TVR_BITS + 4*TVN_BITS)) - 1)) |
1da177e4 | 71 | |
a6fa8e5a | 72 | struct tvec { |
1dabbcec | 73 | struct hlist_head vec[TVN_SIZE]; |
a6fa8e5a | 74 | }; |
1da177e4 | 75 | |
a6fa8e5a | 76 | struct tvec_root { |
1dabbcec | 77 | struct hlist_head vec[TVR_SIZE]; |
a6fa8e5a | 78 | }; |
1da177e4 | 79 | |
a6fa8e5a | 80 | struct tvec_base { |
3691c519 ON |
81 | spinlock_t lock; |
82 | struct timer_list *running_timer; | |
1da177e4 | 83 | unsigned long timer_jiffies; |
97fd9ed4 | 84 | unsigned long next_timer; |
99d5f3aa | 85 | unsigned long active_timers; |
fff42158 | 86 | unsigned long all_timers; |
d6f93829 | 87 | int cpu; |
a6fa8e5a PM |
88 | struct tvec_root tv1; |
89 | struct tvec tv2; | |
90 | struct tvec tv3; | |
91 | struct tvec tv4; | |
92 | struct tvec tv5; | |
6e453a67 | 93 | } ____cacheline_aligned; |
1da177e4 | 94 | |
b337a938 PZ |
95 | /* |
96 | * __TIMER_INITIALIZER() needs to set ->base to a valid pointer (because we've | |
97 | * made NULL special, hint: lock_timer_base()) and we cannot get a compile time | |
98 | * pointer to per-cpu entries because we don't know where we'll map the section, | |
99 | * even for the boot cpu. | |
100 | * | |
101 | * And so we use boot_tvec_bases for boot CPU and per-cpu __tvec_bases for the | |
102 | * rest of them. | |
103 | */ | |
a6fa8e5a | 104 | struct tvec_base boot_tvec_bases; |
3691c519 | 105 | EXPORT_SYMBOL(boot_tvec_bases); |
b337a938 | 106 | |
a6fa8e5a | 107 | static DEFINE_PER_CPU(struct tvec_base *, tvec_bases) = &boot_tvec_bases; |
1da177e4 | 108 | |
6e453a67 | 109 | /* Functions below help us manage 'deferrable' flag */ |
a6fa8e5a | 110 | static inline unsigned int tbase_get_deferrable(struct tvec_base *base) |
6e453a67 | 111 | { |
e52b1db3 | 112 | return ((unsigned int)(unsigned long)base & TIMER_DEFERRABLE); |
6e453a67 VP |
113 | } |
114 | ||
c5f66e99 TH |
115 | static inline unsigned int tbase_get_irqsafe(struct tvec_base *base) |
116 | { | |
117 | return ((unsigned int)(unsigned long)base & TIMER_IRQSAFE); | |
118 | } | |
119 | ||
a6fa8e5a | 120 | static inline struct tvec_base *tbase_get_base(struct tvec_base *base) |
6e453a67 | 121 | { |
e52b1db3 | 122 | return ((struct tvec_base *)((unsigned long)base & ~TIMER_FLAG_MASK)); |
6e453a67 VP |
123 | } |
124 | ||
6e453a67 | 125 | static inline void |
a6fa8e5a | 126 | timer_set_base(struct timer_list *timer, struct tvec_base *new_base) |
6e453a67 | 127 | { |
e52b1db3 TH |
128 | unsigned long flags = (unsigned long)timer->base & TIMER_FLAG_MASK; |
129 | ||
130 | timer->base = (struct tvec_base *)((unsigned long)(new_base) | flags); | |
6e453a67 VP |
131 | } |
132 | ||
9c133c46 AS |
133 | static unsigned long round_jiffies_common(unsigned long j, int cpu, |
134 | bool force_up) | |
4c36a5de AV |
135 | { |
136 | int rem; | |
137 | unsigned long original = j; | |
138 | ||
139 | /* | |
140 | * We don't want all cpus firing their timers at once hitting the | |
141 | * same lock or cachelines, so we skew each extra cpu with an extra | |
142 | * 3 jiffies. This 3 jiffies came originally from the mm/ code which | |
143 | * already did this. | |
144 | * The skew is done by adding 3*cpunr, then round, then subtract this | |
145 | * extra offset again. | |
146 | */ | |
147 | j += cpu * 3; | |
148 | ||
149 | rem = j % HZ; | |
150 | ||
151 | /* | |
152 | * If the target jiffie is just after a whole second (which can happen | |
153 | * due to delays of the timer irq, long irq off times etc etc) then | |
154 | * we should round down to the whole second, not up. Use 1/4th second | |
155 | * as cutoff for this rounding as an extreme upper bound for this. | |
9c133c46 | 156 | * But never round down if @force_up is set. |
4c36a5de | 157 | */ |
9c133c46 | 158 | if (rem < HZ/4 && !force_up) /* round down */ |
4c36a5de AV |
159 | j = j - rem; |
160 | else /* round up */ | |
161 | j = j - rem + HZ; | |
162 | ||
163 | /* now that we have rounded, subtract the extra skew again */ | |
164 | j -= cpu * 3; | |
165 | ||
9e04d380 BVA |
166 | /* |
167 | * Make sure j is still in the future. Otherwise return the | |
168 | * unmodified value. | |
169 | */ | |
170 | return time_is_after_jiffies(j) ? j : original; | |
4c36a5de | 171 | } |
9c133c46 AS |
172 | |
173 | /** | |
174 | * __round_jiffies - function to round jiffies to a full second | |
175 | * @j: the time in (absolute) jiffies that should be rounded | |
176 | * @cpu: the processor number on which the timeout will happen | |
177 | * | |
178 | * __round_jiffies() rounds an absolute time in the future (in jiffies) | |
179 | * up or down to (approximately) full seconds. This is useful for timers | |
180 | * for which the exact time they fire does not matter too much, as long as | |
181 | * they fire approximately every X seconds. | |
182 | * | |
183 | * By rounding these timers to whole seconds, all such timers will fire | |
184 | * at the same time, rather than at various times spread out. The goal | |
185 | * of this is to have the CPU wake up less, which saves power. | |
186 | * | |
187 | * The exact rounding is skewed for each processor to avoid all | |
188 | * processors firing at the exact same time, which could lead | |
189 | * to lock contention or spurious cache line bouncing. | |
190 | * | |
191 | * The return value is the rounded version of the @j parameter. | |
192 | */ | |
193 | unsigned long __round_jiffies(unsigned long j, int cpu) | |
194 | { | |
195 | return round_jiffies_common(j, cpu, false); | |
196 | } | |
4c36a5de AV |
197 | EXPORT_SYMBOL_GPL(__round_jiffies); |
198 | ||
199 | /** | |
200 | * __round_jiffies_relative - function to round jiffies to a full second | |
201 | * @j: the time in (relative) jiffies that should be rounded | |
202 | * @cpu: the processor number on which the timeout will happen | |
203 | * | |
72fd4a35 | 204 | * __round_jiffies_relative() rounds a time delta in the future (in jiffies) |
4c36a5de AV |
205 | * up or down to (approximately) full seconds. This is useful for timers |
206 | * for which the exact time they fire does not matter too much, as long as | |
207 | * they fire approximately every X seconds. | |
208 | * | |
209 | * By rounding these timers to whole seconds, all such timers will fire | |
210 | * at the same time, rather than at various times spread out. The goal | |
211 | * of this is to have the CPU wake up less, which saves power. | |
212 | * | |
213 | * The exact rounding is skewed for each processor to avoid all | |
214 | * processors firing at the exact same time, which could lead | |
215 | * to lock contention or spurious cache line bouncing. | |
216 | * | |
72fd4a35 | 217 | * The return value is the rounded version of the @j parameter. |
4c36a5de AV |
218 | */ |
219 | unsigned long __round_jiffies_relative(unsigned long j, int cpu) | |
220 | { | |
9c133c46 AS |
221 | unsigned long j0 = jiffies; |
222 | ||
223 | /* Use j0 because jiffies might change while we run */ | |
224 | return round_jiffies_common(j + j0, cpu, false) - j0; | |
4c36a5de AV |
225 | } |
226 | EXPORT_SYMBOL_GPL(__round_jiffies_relative); | |
227 | ||
228 | /** | |
229 | * round_jiffies - function to round jiffies to a full second | |
230 | * @j: the time in (absolute) jiffies that should be rounded | |
231 | * | |
72fd4a35 | 232 | * round_jiffies() rounds an absolute time in the future (in jiffies) |
4c36a5de AV |
233 | * up or down to (approximately) full seconds. This is useful for timers |
234 | * for which the exact time they fire does not matter too much, as long as | |
235 | * they fire approximately every X seconds. | |
236 | * | |
237 | * By rounding these timers to whole seconds, all such timers will fire | |
238 | * at the same time, rather than at various times spread out. The goal | |
239 | * of this is to have the CPU wake up less, which saves power. | |
240 | * | |
72fd4a35 | 241 | * The return value is the rounded version of the @j parameter. |
4c36a5de AV |
242 | */ |
243 | unsigned long round_jiffies(unsigned long j) | |
244 | { | |
9c133c46 | 245 | return round_jiffies_common(j, raw_smp_processor_id(), false); |
4c36a5de AV |
246 | } |
247 | EXPORT_SYMBOL_GPL(round_jiffies); | |
248 | ||
249 | /** | |
250 | * round_jiffies_relative - function to round jiffies to a full second | |
251 | * @j: the time in (relative) jiffies that should be rounded | |
252 | * | |
72fd4a35 | 253 | * round_jiffies_relative() rounds a time delta in the future (in jiffies) |
4c36a5de AV |
254 | * up or down to (approximately) full seconds. This is useful for timers |
255 | * for which the exact time they fire does not matter too much, as long as | |
256 | * they fire approximately every X seconds. | |
257 | * | |
258 | * By rounding these timers to whole seconds, all such timers will fire | |
259 | * at the same time, rather than at various times spread out. The goal | |
260 | * of this is to have the CPU wake up less, which saves power. | |
261 | * | |
72fd4a35 | 262 | * The return value is the rounded version of the @j parameter. |
4c36a5de AV |
263 | */ |
264 | unsigned long round_jiffies_relative(unsigned long j) | |
265 | { | |
266 | return __round_jiffies_relative(j, raw_smp_processor_id()); | |
267 | } | |
268 | EXPORT_SYMBOL_GPL(round_jiffies_relative); | |
269 | ||
9c133c46 AS |
270 | /** |
271 | * __round_jiffies_up - function to round jiffies up to a full second | |
272 | * @j: the time in (absolute) jiffies that should be rounded | |
273 | * @cpu: the processor number on which the timeout will happen | |
274 | * | |
275 | * This is the same as __round_jiffies() except that it will never | |
276 | * round down. This is useful for timeouts for which the exact time | |
277 | * of firing does not matter too much, as long as they don't fire too | |
278 | * early. | |
279 | */ | |
280 | unsigned long __round_jiffies_up(unsigned long j, int cpu) | |
281 | { | |
282 | return round_jiffies_common(j, cpu, true); | |
283 | } | |
284 | EXPORT_SYMBOL_GPL(__round_jiffies_up); | |
285 | ||
286 | /** | |
287 | * __round_jiffies_up_relative - function to round jiffies up to a full second | |
288 | * @j: the time in (relative) jiffies that should be rounded | |
289 | * @cpu: the processor number on which the timeout will happen | |
290 | * | |
291 | * This is the same as __round_jiffies_relative() except that it will never | |
292 | * round down. This is useful for timeouts for which the exact time | |
293 | * of firing does not matter too much, as long as they don't fire too | |
294 | * early. | |
295 | */ | |
296 | unsigned long __round_jiffies_up_relative(unsigned long j, int cpu) | |
297 | { | |
298 | unsigned long j0 = jiffies; | |
299 | ||
300 | /* Use j0 because jiffies might change while we run */ | |
301 | return round_jiffies_common(j + j0, cpu, true) - j0; | |
302 | } | |
303 | EXPORT_SYMBOL_GPL(__round_jiffies_up_relative); | |
304 | ||
305 | /** | |
306 | * round_jiffies_up - function to round jiffies up to a full second | |
307 | * @j: the time in (absolute) jiffies that should be rounded | |
308 | * | |
309 | * This is the same as round_jiffies() except that it will never | |
310 | * round down. This is useful for timeouts for which the exact time | |
311 | * of firing does not matter too much, as long as they don't fire too | |
312 | * early. | |
313 | */ | |
314 | unsigned long round_jiffies_up(unsigned long j) | |
315 | { | |
316 | return round_jiffies_common(j, raw_smp_processor_id(), true); | |
317 | } | |
318 | EXPORT_SYMBOL_GPL(round_jiffies_up); | |
319 | ||
320 | /** | |
321 | * round_jiffies_up_relative - function to round jiffies up to a full second | |
322 | * @j: the time in (relative) jiffies that should be rounded | |
323 | * | |
324 | * This is the same as round_jiffies_relative() except that it will never | |
325 | * round down. This is useful for timeouts for which the exact time | |
326 | * of firing does not matter too much, as long as they don't fire too | |
327 | * early. | |
328 | */ | |
329 | unsigned long round_jiffies_up_relative(unsigned long j) | |
330 | { | |
331 | return __round_jiffies_up_relative(j, raw_smp_processor_id()); | |
332 | } | |
333 | EXPORT_SYMBOL_GPL(round_jiffies_up_relative); | |
334 | ||
3bbb9ec9 AV |
335 | /** |
336 | * set_timer_slack - set the allowed slack for a timer | |
0caa6210 | 337 | * @timer: the timer to be modified |
3bbb9ec9 AV |
338 | * @slack_hz: the amount of time (in jiffies) allowed for rounding |
339 | * | |
340 | * Set the amount of time, in jiffies, that a certain timer has | |
341 | * in terms of slack. By setting this value, the timer subsystem | |
342 | * will schedule the actual timer somewhere between | |
343 | * the time mod_timer() asks for, and that time plus the slack. | |
344 | * | |
345 | * By setting the slack to -1, a percentage of the delay is used | |
346 | * instead. | |
347 | */ | |
348 | void set_timer_slack(struct timer_list *timer, int slack_hz) | |
349 | { | |
350 | timer->slack = slack_hz; | |
351 | } | |
352 | EXPORT_SYMBOL_GPL(set_timer_slack); | |
353 | ||
facbb4a7 TG |
354 | static void |
355 | __internal_add_timer(struct tvec_base *base, struct timer_list *timer) | |
1da177e4 LT |
356 | { |
357 | unsigned long expires = timer->expires; | |
358 | unsigned long idx = expires - base->timer_jiffies; | |
1dabbcec | 359 | struct hlist_head *vec; |
1da177e4 LT |
360 | |
361 | if (idx < TVR_SIZE) { | |
362 | int i = expires & TVR_MASK; | |
363 | vec = base->tv1.vec + i; | |
364 | } else if (idx < 1 << (TVR_BITS + TVN_BITS)) { | |
365 | int i = (expires >> TVR_BITS) & TVN_MASK; | |
366 | vec = base->tv2.vec + i; | |
367 | } else if (idx < 1 << (TVR_BITS + 2 * TVN_BITS)) { | |
368 | int i = (expires >> (TVR_BITS + TVN_BITS)) & TVN_MASK; | |
369 | vec = base->tv3.vec + i; | |
370 | } else if (idx < 1 << (TVR_BITS + 3 * TVN_BITS)) { | |
371 | int i = (expires >> (TVR_BITS + 2 * TVN_BITS)) & TVN_MASK; | |
372 | vec = base->tv4.vec + i; | |
373 | } else if ((signed long) idx < 0) { | |
374 | /* | |
375 | * Can happen if you add a timer with expires == jiffies, | |
376 | * or you set a timer to go off in the past | |
377 | */ | |
378 | vec = base->tv1.vec + (base->timer_jiffies & TVR_MASK); | |
379 | } else { | |
380 | int i; | |
26cff4e2 HC |
381 | /* If the timeout is larger than MAX_TVAL (on 64-bit |
382 | * architectures or with CONFIG_BASE_SMALL=1) then we | |
383 | * use the maximum timeout. | |
1da177e4 | 384 | */ |
26cff4e2 HC |
385 | if (idx > MAX_TVAL) { |
386 | idx = MAX_TVAL; | |
1da177e4 LT |
387 | expires = idx + base->timer_jiffies; |
388 | } | |
389 | i = (expires >> (TVR_BITS + 3 * TVN_BITS)) & TVN_MASK; | |
390 | vec = base->tv5.vec + i; | |
391 | } | |
1bd04bf6 | 392 | |
1dabbcec | 393 | hlist_add_head(&timer->entry, vec); |
1da177e4 LT |
394 | } |
395 | ||
facbb4a7 TG |
396 | static void internal_add_timer(struct tvec_base *base, struct timer_list *timer) |
397 | { | |
3bb475a3 TG |
398 | /* Advance base->jiffies, if the base is empty */ |
399 | if (!base->all_timers++) | |
400 | base->timer_jiffies = jiffies; | |
401 | ||
facbb4a7 TG |
402 | __internal_add_timer(base, timer); |
403 | /* | |
99d5f3aa | 404 | * Update base->active_timers and base->next_timer |
facbb4a7 | 405 | */ |
99d5f3aa | 406 | if (!tbase_get_deferrable(timer->base)) { |
aea369b9 ON |
407 | if (!base->active_timers++ || |
408 | time_before(timer->expires, base->next_timer)) | |
99d5f3aa | 409 | base->next_timer = timer->expires; |
99d5f3aa | 410 | } |
9f6d9baa VK |
411 | |
412 | /* | |
413 | * Check whether the other CPU is in dynticks mode and needs | |
414 | * to be triggered to reevaluate the timer wheel. | |
415 | * We are protected against the other CPU fiddling | |
416 | * with the timer by holding the timer base lock. This also | |
417 | * makes sure that a CPU on the way to stop its tick can not | |
418 | * evaluate the timer wheel. | |
419 | * | |
420 | * Spare the IPI for deferrable timers on idle targets though. | |
421 | * The next busy ticks will take care of it. Except full dynticks | |
422 | * require special care against races with idle_cpu(), lets deal | |
423 | * with that later. | |
424 | */ | |
781978e6 | 425 | if (!tbase_get_deferrable(timer->base) || tick_nohz_full_cpu(base->cpu)) |
9f6d9baa | 426 | wake_up_nohz_cpu(base->cpu); |
facbb4a7 TG |
427 | } |
428 | ||
82f67cd9 IM |
429 | #ifdef CONFIG_TIMER_STATS |
430 | void __timer_stats_timer_set_start_info(struct timer_list *timer, void *addr) | |
431 | { | |
432 | if (timer->start_site) | |
433 | return; | |
434 | ||
435 | timer->start_site = addr; | |
436 | memcpy(timer->start_comm, current->comm, TASK_COMM_LEN); | |
437 | timer->start_pid = current->pid; | |
438 | } | |
c5c061b8 VP |
439 | |
440 | static void timer_stats_account_timer(struct timer_list *timer) | |
441 | { | |
442 | unsigned int flag = 0; | |
443 | ||
507e1231 HC |
444 | if (likely(!timer->start_site)) |
445 | return; | |
c5c061b8 VP |
446 | if (unlikely(tbase_get_deferrable(timer->base))) |
447 | flag |= TIMER_STATS_FLAG_DEFERRABLE; | |
448 | ||
449 | timer_stats_update_stats(timer, timer->start_pid, timer->start_site, | |
450 | timer->function, timer->start_comm, flag); | |
451 | } | |
452 | ||
453 | #else | |
454 | static void timer_stats_account_timer(struct timer_list *timer) {} | |
82f67cd9 IM |
455 | #endif |
456 | ||
c6f3a97f TG |
457 | #ifdef CONFIG_DEBUG_OBJECTS_TIMERS |
458 | ||
459 | static struct debug_obj_descr timer_debug_descr; | |
460 | ||
99777288 SG |
461 | static void *timer_debug_hint(void *addr) |
462 | { | |
463 | return ((struct timer_list *) addr)->function; | |
464 | } | |
465 | ||
c6f3a97f TG |
466 | /* |
467 | * fixup_init is called when: | |
468 | * - an active object is initialized | |
55c888d6 | 469 | */ |
c6f3a97f TG |
470 | static int timer_fixup_init(void *addr, enum debug_obj_state state) |
471 | { | |
472 | struct timer_list *timer = addr; | |
473 | ||
474 | switch (state) { | |
475 | case ODEBUG_STATE_ACTIVE: | |
476 | del_timer_sync(timer); | |
477 | debug_object_init(timer, &timer_debug_descr); | |
478 | return 1; | |
479 | default: | |
480 | return 0; | |
481 | } | |
482 | } | |
483 | ||
fb16b8cf SB |
484 | /* Stub timer callback for improperly used timers. */ |
485 | static void stub_timer(unsigned long data) | |
486 | { | |
487 | WARN_ON(1); | |
488 | } | |
489 | ||
c6f3a97f TG |
490 | /* |
491 | * fixup_activate is called when: | |
492 | * - an active object is activated | |
493 | * - an unknown object is activated (might be a statically initialized object) | |
494 | */ | |
495 | static int timer_fixup_activate(void *addr, enum debug_obj_state state) | |
496 | { | |
497 | struct timer_list *timer = addr; | |
498 | ||
499 | switch (state) { | |
500 | ||
501 | case ODEBUG_STATE_NOTAVAILABLE: | |
502 | /* | |
503 | * This is not really a fixup. The timer was | |
504 | * statically initialized. We just make sure that it | |
505 | * is tracked in the object tracker. | |
506 | */ | |
1dabbcec TG |
507 | if (timer->entry.pprev == NULL && |
508 | timer->entry.next == TIMER_ENTRY_STATIC) { | |
c6f3a97f TG |
509 | debug_object_init(timer, &timer_debug_descr); |
510 | debug_object_activate(timer, &timer_debug_descr); | |
511 | return 0; | |
512 | } else { | |
fb16b8cf SB |
513 | setup_timer(timer, stub_timer, 0); |
514 | return 1; | |
c6f3a97f TG |
515 | } |
516 | return 0; | |
517 | ||
518 | case ODEBUG_STATE_ACTIVE: | |
519 | WARN_ON(1); | |
520 | ||
521 | default: | |
522 | return 0; | |
523 | } | |
524 | } | |
525 | ||
526 | /* | |
527 | * fixup_free is called when: | |
528 | * - an active object is freed | |
529 | */ | |
530 | static int timer_fixup_free(void *addr, enum debug_obj_state state) | |
531 | { | |
532 | struct timer_list *timer = addr; | |
533 | ||
534 | switch (state) { | |
535 | case ODEBUG_STATE_ACTIVE: | |
536 | del_timer_sync(timer); | |
537 | debug_object_free(timer, &timer_debug_descr); | |
538 | return 1; | |
539 | default: | |
540 | return 0; | |
541 | } | |
542 | } | |
543 | ||
dc4218bd CC |
544 | /* |
545 | * fixup_assert_init is called when: | |
546 | * - an untracked/uninit-ed object is found | |
547 | */ | |
548 | static int timer_fixup_assert_init(void *addr, enum debug_obj_state state) | |
549 | { | |
550 | struct timer_list *timer = addr; | |
551 | ||
552 | switch (state) { | |
553 | case ODEBUG_STATE_NOTAVAILABLE: | |
1dabbcec | 554 | if (timer->entry.next == TIMER_ENTRY_STATIC) { |
dc4218bd CC |
555 | /* |
556 | * This is not really a fixup. The timer was | |
557 | * statically initialized. We just make sure that it | |
558 | * is tracked in the object tracker. | |
559 | */ | |
560 | debug_object_init(timer, &timer_debug_descr); | |
561 | return 0; | |
562 | } else { | |
563 | setup_timer(timer, stub_timer, 0); | |
564 | return 1; | |
565 | } | |
566 | default: | |
567 | return 0; | |
568 | } | |
569 | } | |
570 | ||
c6f3a97f | 571 | static struct debug_obj_descr timer_debug_descr = { |
dc4218bd CC |
572 | .name = "timer_list", |
573 | .debug_hint = timer_debug_hint, | |
574 | .fixup_init = timer_fixup_init, | |
575 | .fixup_activate = timer_fixup_activate, | |
576 | .fixup_free = timer_fixup_free, | |
577 | .fixup_assert_init = timer_fixup_assert_init, | |
c6f3a97f TG |
578 | }; |
579 | ||
580 | static inline void debug_timer_init(struct timer_list *timer) | |
581 | { | |
582 | debug_object_init(timer, &timer_debug_descr); | |
583 | } | |
584 | ||
585 | static inline void debug_timer_activate(struct timer_list *timer) | |
586 | { | |
587 | debug_object_activate(timer, &timer_debug_descr); | |
588 | } | |
589 | ||
590 | static inline void debug_timer_deactivate(struct timer_list *timer) | |
591 | { | |
592 | debug_object_deactivate(timer, &timer_debug_descr); | |
593 | } | |
594 | ||
595 | static inline void debug_timer_free(struct timer_list *timer) | |
596 | { | |
597 | debug_object_free(timer, &timer_debug_descr); | |
598 | } | |
599 | ||
dc4218bd CC |
600 | static inline void debug_timer_assert_init(struct timer_list *timer) |
601 | { | |
602 | debug_object_assert_init(timer, &timer_debug_descr); | |
603 | } | |
604 | ||
fc683995 TH |
605 | static void do_init_timer(struct timer_list *timer, unsigned int flags, |
606 | const char *name, struct lock_class_key *key); | |
c6f3a97f | 607 | |
fc683995 TH |
608 | void init_timer_on_stack_key(struct timer_list *timer, unsigned int flags, |
609 | const char *name, struct lock_class_key *key) | |
c6f3a97f TG |
610 | { |
611 | debug_object_init_on_stack(timer, &timer_debug_descr); | |
fc683995 | 612 | do_init_timer(timer, flags, name, key); |
c6f3a97f | 613 | } |
6f2b9b9a | 614 | EXPORT_SYMBOL_GPL(init_timer_on_stack_key); |
c6f3a97f TG |
615 | |
616 | void destroy_timer_on_stack(struct timer_list *timer) | |
617 | { | |
618 | debug_object_free(timer, &timer_debug_descr); | |
619 | } | |
620 | EXPORT_SYMBOL_GPL(destroy_timer_on_stack); | |
621 | ||
622 | #else | |
623 | static inline void debug_timer_init(struct timer_list *timer) { } | |
624 | static inline void debug_timer_activate(struct timer_list *timer) { } | |
625 | static inline void debug_timer_deactivate(struct timer_list *timer) { } | |
dc4218bd | 626 | static inline void debug_timer_assert_init(struct timer_list *timer) { } |
c6f3a97f TG |
627 | #endif |
628 | ||
2b022e3d XG |
629 | static inline void debug_init(struct timer_list *timer) |
630 | { | |
631 | debug_timer_init(timer); | |
632 | trace_timer_init(timer); | |
633 | } | |
634 | ||
635 | static inline void | |
636 | debug_activate(struct timer_list *timer, unsigned long expires) | |
637 | { | |
638 | debug_timer_activate(timer); | |
4e413e85 | 639 | trace_timer_start(timer, expires, tbase_get_deferrable(timer->base)); |
2b022e3d XG |
640 | } |
641 | ||
642 | static inline void debug_deactivate(struct timer_list *timer) | |
643 | { | |
644 | debug_timer_deactivate(timer); | |
645 | trace_timer_cancel(timer); | |
646 | } | |
647 | ||
dc4218bd CC |
648 | static inline void debug_assert_init(struct timer_list *timer) |
649 | { | |
650 | debug_timer_assert_init(timer); | |
651 | } | |
652 | ||
fc683995 TH |
653 | static void do_init_timer(struct timer_list *timer, unsigned int flags, |
654 | const char *name, struct lock_class_key *key) | |
55c888d6 | 655 | { |
22127e93 | 656 | struct tvec_base *base = raw_cpu_read(tvec_bases); |
fc683995 | 657 | |
1dabbcec | 658 | timer->entry.pprev = NULL; |
fc683995 | 659 | timer->base = (void *)((unsigned long)base | flags); |
3bbb9ec9 | 660 | timer->slack = -1; |
82f67cd9 IM |
661 | #ifdef CONFIG_TIMER_STATS |
662 | timer->start_site = NULL; | |
663 | timer->start_pid = -1; | |
664 | memset(timer->start_comm, 0, TASK_COMM_LEN); | |
665 | #endif | |
6f2b9b9a | 666 | lockdep_init_map(&timer->lockdep_map, name, key, 0); |
55c888d6 | 667 | } |
c6f3a97f TG |
668 | |
669 | /** | |
633fe795 | 670 | * init_timer_key - initialize a timer |
c6f3a97f | 671 | * @timer: the timer to be initialized |
fc683995 | 672 | * @flags: timer flags |
633fe795 RD |
673 | * @name: name of the timer |
674 | * @key: lockdep class key of the fake lock used for tracking timer | |
675 | * sync lock dependencies | |
c6f3a97f | 676 | * |
633fe795 | 677 | * init_timer_key() must be done to a timer prior calling *any* of the |
c6f3a97f TG |
678 | * other timer functions. |
679 | */ | |
fc683995 TH |
680 | void init_timer_key(struct timer_list *timer, unsigned int flags, |
681 | const char *name, struct lock_class_key *key) | |
c6f3a97f | 682 | { |
2b022e3d | 683 | debug_init(timer); |
fc683995 | 684 | do_init_timer(timer, flags, name, key); |
c6f3a97f | 685 | } |
6f2b9b9a | 686 | EXPORT_SYMBOL(init_timer_key); |
55c888d6 | 687 | |
ec44bc7a | 688 | static inline void detach_timer(struct timer_list *timer, bool clear_pending) |
55c888d6 | 689 | { |
1dabbcec | 690 | struct hlist_node *entry = &timer->entry; |
55c888d6 | 691 | |
2b022e3d | 692 | debug_deactivate(timer); |
c6f3a97f | 693 | |
1dabbcec | 694 | __hlist_del(entry); |
55c888d6 | 695 | if (clear_pending) |
1dabbcec TG |
696 | entry->pprev = NULL; |
697 | entry->next = LIST_POISON2; | |
55c888d6 ON |
698 | } |
699 | ||
99d5f3aa TG |
700 | static inline void |
701 | detach_expired_timer(struct timer_list *timer, struct tvec_base *base) | |
702 | { | |
703 | detach_timer(timer, true); | |
704 | if (!tbase_get_deferrable(timer->base)) | |
e52b1db3 | 705 | base->active_timers--; |
fff42158 | 706 | base->all_timers--; |
99d5f3aa TG |
707 | } |
708 | ||
ec44bc7a TG |
709 | static int detach_if_pending(struct timer_list *timer, struct tvec_base *base, |
710 | bool clear_pending) | |
711 | { | |
712 | if (!timer_pending(timer)) | |
713 | return 0; | |
714 | ||
715 | detach_timer(timer, clear_pending); | |
99d5f3aa | 716 | if (!tbase_get_deferrable(timer->base)) { |
e52b1db3 | 717 | base->active_timers--; |
99d5f3aa TG |
718 | if (timer->expires == base->next_timer) |
719 | base->next_timer = base->timer_jiffies; | |
720 | } | |
3bb475a3 TG |
721 | /* If this was the last timer, advance base->jiffies */ |
722 | if (!--base->all_timers) | |
723 | base->timer_jiffies = jiffies; | |
ec44bc7a TG |
724 | return 1; |
725 | } | |
726 | ||
55c888d6 | 727 | /* |
3691c519 | 728 | * We are using hashed locking: holding per_cpu(tvec_bases).lock |
55c888d6 ON |
729 | * means that all timers which are tied to this base via timer->base are |
730 | * locked, and the base itself is locked too. | |
731 | * | |
732 | * So __run_timers/migrate_timers can safely modify all timers which could | |
733 | * be found on ->tvX lists. | |
734 | * | |
735 | * When the timer's base is locked, and the timer removed from list, it is | |
736 | * possible to set timer->base = NULL and drop the lock: the timer remains | |
737 | * locked. | |
738 | */ | |
a6fa8e5a | 739 | static struct tvec_base *lock_timer_base(struct timer_list *timer, |
55c888d6 | 740 | unsigned long *flags) |
89e7e374 | 741 | __acquires(timer->base->lock) |
55c888d6 | 742 | { |
a6fa8e5a | 743 | struct tvec_base *base; |
55c888d6 ON |
744 | |
745 | for (;;) { | |
a6fa8e5a | 746 | struct tvec_base *prelock_base = timer->base; |
6e453a67 | 747 | base = tbase_get_base(prelock_base); |
55c888d6 ON |
748 | if (likely(base != NULL)) { |
749 | spin_lock_irqsave(&base->lock, *flags); | |
6e453a67 | 750 | if (likely(prelock_base == timer->base)) |
55c888d6 ON |
751 | return base; |
752 | /* The timer has migrated to another CPU */ | |
753 | spin_unlock_irqrestore(&base->lock, *flags); | |
754 | } | |
755 | cpu_relax(); | |
756 | } | |
757 | } | |
758 | ||
74019224 | 759 | static inline int |
597d0275 AB |
760 | __mod_timer(struct timer_list *timer, unsigned long expires, |
761 | bool pending_only, int pinned) | |
1da177e4 | 762 | { |
a6fa8e5a | 763 | struct tvec_base *base, *new_base; |
1da177e4 | 764 | unsigned long flags; |
eea08f32 | 765 | int ret = 0 , cpu; |
1da177e4 | 766 | |
82f67cd9 | 767 | timer_stats_timer_set_start_info(timer); |
1da177e4 | 768 | BUG_ON(!timer->function); |
1da177e4 | 769 | |
55c888d6 ON |
770 | base = lock_timer_base(timer, &flags); |
771 | ||
ec44bc7a TG |
772 | ret = detach_if_pending(timer, base, false); |
773 | if (!ret && pending_only) | |
774 | goto out_unlock; | |
55c888d6 | 775 | |
2b022e3d | 776 | debug_activate(timer, expires); |
c6f3a97f | 777 | |
6201b4d6 | 778 | cpu = get_nohz_timer_target(pinned); |
eea08f32 AB |
779 | new_base = per_cpu(tvec_bases, cpu); |
780 | ||
3691c519 | 781 | if (base != new_base) { |
1da177e4 | 782 | /* |
55c888d6 ON |
783 | * We are trying to schedule the timer on the local CPU. |
784 | * However we can't change timer's base while it is running, | |
785 | * otherwise del_timer_sync() can't detect that the timer's | |
786 | * handler yet has not finished. This also guarantees that | |
787 | * the timer is serialized wrt itself. | |
1da177e4 | 788 | */ |
a2c348fe | 789 | if (likely(base->running_timer != timer)) { |
55c888d6 | 790 | /* See the comment in lock_timer_base() */ |
6e453a67 | 791 | timer_set_base(timer, NULL); |
55c888d6 | 792 | spin_unlock(&base->lock); |
a2c348fe ON |
793 | base = new_base; |
794 | spin_lock(&base->lock); | |
6e453a67 | 795 | timer_set_base(timer, base); |
1da177e4 LT |
796 | } |
797 | } | |
798 | ||
1da177e4 | 799 | timer->expires = expires; |
a2c348fe | 800 | internal_add_timer(base, timer); |
74019224 IM |
801 | |
802 | out_unlock: | |
a2c348fe | 803 | spin_unlock_irqrestore(&base->lock, flags); |
1da177e4 LT |
804 | |
805 | return ret; | |
806 | } | |
807 | ||
2aae4a10 | 808 | /** |
74019224 IM |
809 | * mod_timer_pending - modify a pending timer's timeout |
810 | * @timer: the pending timer to be modified | |
811 | * @expires: new timeout in jiffies | |
1da177e4 | 812 | * |
74019224 IM |
813 | * mod_timer_pending() is the same for pending timers as mod_timer(), |
814 | * but will not re-activate and modify already deleted timers. | |
815 | * | |
816 | * It is useful for unserialized use of timers. | |
1da177e4 | 817 | */ |
74019224 | 818 | int mod_timer_pending(struct timer_list *timer, unsigned long expires) |
1da177e4 | 819 | { |
597d0275 | 820 | return __mod_timer(timer, expires, true, TIMER_NOT_PINNED); |
1da177e4 | 821 | } |
74019224 | 822 | EXPORT_SYMBOL(mod_timer_pending); |
1da177e4 | 823 | |
3bbb9ec9 AV |
824 | /* |
825 | * Decide where to put the timer while taking the slack into account | |
826 | * | |
827 | * Algorithm: | |
828 | * 1) calculate the maximum (absolute) time | |
829 | * 2) calculate the highest bit where the expires and new max are different | |
830 | * 3) use this bit to make a mask | |
831 | * 4) use the bitmask to round down the maximum time, so that all last | |
832 | * bits are zeros | |
833 | */ | |
834 | static inline | |
835 | unsigned long apply_slack(struct timer_list *timer, unsigned long expires) | |
836 | { | |
837 | unsigned long expires_limit, mask; | |
838 | int bit; | |
839 | ||
8e63d779 | 840 | if (timer->slack >= 0) { |
f00e047e | 841 | expires_limit = expires + timer->slack; |
8e63d779 | 842 | } else { |
1c3cc116 SAS |
843 | long delta = expires - jiffies; |
844 | ||
845 | if (delta < 256) | |
846 | return expires; | |
3bbb9ec9 | 847 | |
1c3cc116 | 848 | expires_limit = expires + delta / 256; |
8e63d779 | 849 | } |
3bbb9ec9 | 850 | mask = expires ^ expires_limit; |
3bbb9ec9 AV |
851 | if (mask == 0) |
852 | return expires; | |
853 | ||
854 | bit = find_last_bit(&mask, BITS_PER_LONG); | |
855 | ||
98a01e77 | 856 | mask = (1UL << bit) - 1; |
3bbb9ec9 AV |
857 | |
858 | expires_limit = expires_limit & ~(mask); | |
859 | ||
860 | return expires_limit; | |
861 | } | |
862 | ||
2aae4a10 | 863 | /** |
1da177e4 LT |
864 | * mod_timer - modify a timer's timeout |
865 | * @timer: the timer to be modified | |
2aae4a10 | 866 | * @expires: new timeout in jiffies |
1da177e4 | 867 | * |
72fd4a35 | 868 | * mod_timer() is a more efficient way to update the expire field of an |
1da177e4 LT |
869 | * active timer (if the timer is inactive it will be activated) |
870 | * | |
871 | * mod_timer(timer, expires) is equivalent to: | |
872 | * | |
873 | * del_timer(timer); timer->expires = expires; add_timer(timer); | |
874 | * | |
875 | * Note that if there are multiple unserialized concurrent users of the | |
876 | * same timer, then mod_timer() is the only safe way to modify the timeout, | |
877 | * since add_timer() cannot modify an already running timer. | |
878 | * | |
879 | * The function returns whether it has modified a pending timer or not. | |
880 | * (ie. mod_timer() of an inactive timer returns 0, mod_timer() of an | |
881 | * active timer returns 1.) | |
882 | */ | |
883 | int mod_timer(struct timer_list *timer, unsigned long expires) | |
884 | { | |
1c3cc116 SAS |
885 | expires = apply_slack(timer, expires); |
886 | ||
1da177e4 LT |
887 | /* |
888 | * This is a common optimization triggered by the | |
889 | * networking code - if the timer is re-modified | |
890 | * to be the same thing then just return: | |
891 | */ | |
4841158b | 892 | if (timer_pending(timer) && timer->expires == expires) |
1da177e4 LT |
893 | return 1; |
894 | ||
597d0275 | 895 | return __mod_timer(timer, expires, false, TIMER_NOT_PINNED); |
1da177e4 | 896 | } |
1da177e4 LT |
897 | EXPORT_SYMBOL(mod_timer); |
898 | ||
597d0275 AB |
899 | /** |
900 | * mod_timer_pinned - modify a timer's timeout | |
901 | * @timer: the timer to be modified | |
902 | * @expires: new timeout in jiffies | |
903 | * | |
904 | * mod_timer_pinned() is a way to update the expire field of an | |
905 | * active timer (if the timer is inactive it will be activated) | |
048a0e8f PM |
906 | * and to ensure that the timer is scheduled on the current CPU. |
907 | * | |
908 | * Note that this does not prevent the timer from being migrated | |
909 | * when the current CPU goes offline. If this is a problem for | |
910 | * you, use CPU-hotplug notifiers to handle it correctly, for | |
911 | * example, cancelling the timer when the corresponding CPU goes | |
912 | * offline. | |
597d0275 AB |
913 | * |
914 | * mod_timer_pinned(timer, expires) is equivalent to: | |
915 | * | |
916 | * del_timer(timer); timer->expires = expires; add_timer(timer); | |
917 | */ | |
918 | int mod_timer_pinned(struct timer_list *timer, unsigned long expires) | |
919 | { | |
920 | if (timer->expires == expires && timer_pending(timer)) | |
921 | return 1; | |
922 | ||
923 | return __mod_timer(timer, expires, false, TIMER_PINNED); | |
924 | } | |
925 | EXPORT_SYMBOL(mod_timer_pinned); | |
926 | ||
74019224 IM |
927 | /** |
928 | * add_timer - start a timer | |
929 | * @timer: the timer to be added | |
930 | * | |
931 | * The kernel will do a ->function(->data) callback from the | |
932 | * timer interrupt at the ->expires point in the future. The | |
933 | * current time is 'jiffies'. | |
934 | * | |
935 | * The timer's ->expires, ->function (and if the handler uses it, ->data) | |
936 | * fields must be set prior calling this function. | |
937 | * | |
938 | * Timers with an ->expires field in the past will be executed in the next | |
939 | * timer tick. | |
940 | */ | |
941 | void add_timer(struct timer_list *timer) | |
942 | { | |
943 | BUG_ON(timer_pending(timer)); | |
944 | mod_timer(timer, timer->expires); | |
945 | } | |
946 | EXPORT_SYMBOL(add_timer); | |
947 | ||
948 | /** | |
949 | * add_timer_on - start a timer on a particular CPU | |
950 | * @timer: the timer to be added | |
951 | * @cpu: the CPU to start it on | |
952 | * | |
953 | * This is not very scalable on SMP. Double adds are not possible. | |
954 | */ | |
955 | void add_timer_on(struct timer_list *timer, int cpu) | |
956 | { | |
957 | struct tvec_base *base = per_cpu(tvec_bases, cpu); | |
958 | unsigned long flags; | |
959 | ||
960 | timer_stats_timer_set_start_info(timer); | |
961 | BUG_ON(timer_pending(timer) || !timer->function); | |
962 | spin_lock_irqsave(&base->lock, flags); | |
963 | timer_set_base(timer, base); | |
2b022e3d | 964 | debug_activate(timer, timer->expires); |
74019224 | 965 | internal_add_timer(base, timer); |
74019224 IM |
966 | spin_unlock_irqrestore(&base->lock, flags); |
967 | } | |
a9862e05 | 968 | EXPORT_SYMBOL_GPL(add_timer_on); |
74019224 | 969 | |
2aae4a10 | 970 | /** |
1da177e4 LT |
971 | * del_timer - deactive a timer. |
972 | * @timer: the timer to be deactivated | |
973 | * | |
974 | * del_timer() deactivates a timer - this works on both active and inactive | |
975 | * timers. | |
976 | * | |
977 | * The function returns whether it has deactivated a pending timer or not. | |
978 | * (ie. del_timer() of an inactive timer returns 0, del_timer() of an | |
979 | * active timer returns 1.) | |
980 | */ | |
981 | int del_timer(struct timer_list *timer) | |
982 | { | |
a6fa8e5a | 983 | struct tvec_base *base; |
1da177e4 | 984 | unsigned long flags; |
55c888d6 | 985 | int ret = 0; |
1da177e4 | 986 | |
dc4218bd CC |
987 | debug_assert_init(timer); |
988 | ||
82f67cd9 | 989 | timer_stats_timer_clear_start_info(timer); |
55c888d6 ON |
990 | if (timer_pending(timer)) { |
991 | base = lock_timer_base(timer, &flags); | |
ec44bc7a | 992 | ret = detach_if_pending(timer, base, true); |
1da177e4 | 993 | spin_unlock_irqrestore(&base->lock, flags); |
1da177e4 | 994 | } |
1da177e4 | 995 | |
55c888d6 | 996 | return ret; |
1da177e4 | 997 | } |
1da177e4 LT |
998 | EXPORT_SYMBOL(del_timer); |
999 | ||
2aae4a10 REB |
1000 | /** |
1001 | * try_to_del_timer_sync - Try to deactivate a timer | |
1002 | * @timer: timer do del | |
1003 | * | |
fd450b73 ON |
1004 | * This function tries to deactivate a timer. Upon successful (ret >= 0) |
1005 | * exit the timer is not queued and the handler is not running on any CPU. | |
fd450b73 ON |
1006 | */ |
1007 | int try_to_del_timer_sync(struct timer_list *timer) | |
1008 | { | |
a6fa8e5a | 1009 | struct tvec_base *base; |
fd450b73 ON |
1010 | unsigned long flags; |
1011 | int ret = -1; | |
1012 | ||
dc4218bd CC |
1013 | debug_assert_init(timer); |
1014 | ||
fd450b73 ON |
1015 | base = lock_timer_base(timer, &flags); |
1016 | ||
ec44bc7a TG |
1017 | if (base->running_timer != timer) { |
1018 | timer_stats_timer_clear_start_info(timer); | |
1019 | ret = detach_if_pending(timer, base, true); | |
fd450b73 | 1020 | } |
fd450b73 ON |
1021 | spin_unlock_irqrestore(&base->lock, flags); |
1022 | ||
1023 | return ret; | |
1024 | } | |
e19dff1f DH |
1025 | EXPORT_SYMBOL(try_to_del_timer_sync); |
1026 | ||
6f1bc451 | 1027 | #ifdef CONFIG_SMP |
3650b57f PZ |
1028 | static DEFINE_PER_CPU(struct tvec_base, __tvec_bases); |
1029 | ||
2aae4a10 | 1030 | /** |
1da177e4 LT |
1031 | * del_timer_sync - deactivate a timer and wait for the handler to finish. |
1032 | * @timer: the timer to be deactivated | |
1033 | * | |
1034 | * This function only differs from del_timer() on SMP: besides deactivating | |
1035 | * the timer it also makes sure the handler has finished executing on other | |
1036 | * CPUs. | |
1037 | * | |
72fd4a35 | 1038 | * Synchronization rules: Callers must prevent restarting of the timer, |
1da177e4 | 1039 | * otherwise this function is meaningless. It must not be called from |
c5f66e99 TH |
1040 | * interrupt contexts unless the timer is an irqsafe one. The caller must |
1041 | * not hold locks which would prevent completion of the timer's | |
1042 | * handler. The timer's handler must not call add_timer_on(). Upon exit the | |
1043 | * timer is not queued and the handler is not running on any CPU. | |
1da177e4 | 1044 | * |
c5f66e99 TH |
1045 | * Note: For !irqsafe timers, you must not hold locks that are held in |
1046 | * interrupt context while calling this function. Even if the lock has | |
1047 | * nothing to do with the timer in question. Here's why: | |
48228f7b SR |
1048 | * |
1049 | * CPU0 CPU1 | |
1050 | * ---- ---- | |
1051 | * <SOFTIRQ> | |
1052 | * call_timer_fn(); | |
1053 | * base->running_timer = mytimer; | |
1054 | * spin_lock_irq(somelock); | |
1055 | * <IRQ> | |
1056 | * spin_lock(somelock); | |
1057 | * del_timer_sync(mytimer); | |
1058 | * while (base->running_timer == mytimer); | |
1059 | * | |
1060 | * Now del_timer_sync() will never return and never release somelock. | |
1061 | * The interrupt on the other CPU is waiting to grab somelock but | |
1062 | * it has interrupted the softirq that CPU0 is waiting to finish. | |
1063 | * | |
1da177e4 | 1064 | * The function returns whether it has deactivated a pending timer or not. |
1da177e4 LT |
1065 | */ |
1066 | int del_timer_sync(struct timer_list *timer) | |
1067 | { | |
6f2b9b9a | 1068 | #ifdef CONFIG_LOCKDEP |
f266a511 PZ |
1069 | unsigned long flags; |
1070 | ||
48228f7b SR |
1071 | /* |
1072 | * If lockdep gives a backtrace here, please reference | |
1073 | * the synchronization rules above. | |
1074 | */ | |
7ff20792 | 1075 | local_irq_save(flags); |
6f2b9b9a JB |
1076 | lock_map_acquire(&timer->lockdep_map); |
1077 | lock_map_release(&timer->lockdep_map); | |
7ff20792 | 1078 | local_irq_restore(flags); |
6f2b9b9a | 1079 | #endif |
466bd303 YZ |
1080 | /* |
1081 | * don't use it in hardirq context, because it | |
1082 | * could lead to deadlock. | |
1083 | */ | |
c5f66e99 | 1084 | WARN_ON(in_irq() && !tbase_get_irqsafe(timer->base)); |
fd450b73 ON |
1085 | for (;;) { |
1086 | int ret = try_to_del_timer_sync(timer); | |
1087 | if (ret >= 0) | |
1088 | return ret; | |
a0009652 | 1089 | cpu_relax(); |
fd450b73 | 1090 | } |
1da177e4 | 1091 | } |
55c888d6 | 1092 | EXPORT_SYMBOL(del_timer_sync); |
1da177e4 LT |
1093 | #endif |
1094 | ||
a6fa8e5a | 1095 | static int cascade(struct tvec_base *base, struct tvec *tv, int index) |
1da177e4 LT |
1096 | { |
1097 | /* cascade all the timers from tv up one level */ | |
1dabbcec TG |
1098 | struct timer_list *timer; |
1099 | struct hlist_node *tmp; | |
1100 | struct hlist_head tv_list; | |
3439dd86 | 1101 | |
1dabbcec | 1102 | hlist_move_list(tv->vec + index, &tv_list); |
1da177e4 | 1103 | |
1da177e4 | 1104 | /* |
3439dd86 P |
1105 | * We are removing _all_ timers from the list, so we |
1106 | * don't have to detach them individually. | |
1da177e4 | 1107 | */ |
1dabbcec | 1108 | hlist_for_each_entry_safe(timer, tmp, &tv_list, entry) { |
6e453a67 | 1109 | BUG_ON(tbase_get_base(timer->base) != base); |
facbb4a7 TG |
1110 | /* No accounting, while moving them */ |
1111 | __internal_add_timer(base, timer); | |
1da177e4 | 1112 | } |
1da177e4 LT |
1113 | |
1114 | return index; | |
1115 | } | |
1116 | ||
576da126 TG |
1117 | static void call_timer_fn(struct timer_list *timer, void (*fn)(unsigned long), |
1118 | unsigned long data) | |
1119 | { | |
4a2b4b22 | 1120 | int count = preempt_count(); |
576da126 TG |
1121 | |
1122 | #ifdef CONFIG_LOCKDEP | |
1123 | /* | |
1124 | * It is permissible to free the timer from inside the | |
1125 | * function that is called from it, this we need to take into | |
1126 | * account for lockdep too. To avoid bogus "held lock freed" | |
1127 | * warnings as well as problems when looking into | |
1128 | * timer->lockdep_map, make a copy and use that here. | |
1129 | */ | |
4d82a1de PZ |
1130 | struct lockdep_map lockdep_map; |
1131 | ||
1132 | lockdep_copy_map(&lockdep_map, &timer->lockdep_map); | |
576da126 TG |
1133 | #endif |
1134 | /* | |
1135 | * Couple the lock chain with the lock chain at | |
1136 | * del_timer_sync() by acquiring the lock_map around the fn() | |
1137 | * call here and in del_timer_sync(). | |
1138 | */ | |
1139 | lock_map_acquire(&lockdep_map); | |
1140 | ||
1141 | trace_timer_expire_entry(timer); | |
1142 | fn(data); | |
1143 | trace_timer_expire_exit(timer); | |
1144 | ||
1145 | lock_map_release(&lockdep_map); | |
1146 | ||
4a2b4b22 | 1147 | if (count != preempt_count()) { |
802702e0 | 1148 | WARN_ONCE(1, "timer: %pF preempt leak: %08x -> %08x\n", |
4a2b4b22 | 1149 | fn, count, preempt_count()); |
802702e0 TG |
1150 | /* |
1151 | * Restore the preempt count. That gives us a decent | |
1152 | * chance to survive and extract information. If the | |
1153 | * callback kept a lock held, bad luck, but not worse | |
1154 | * than the BUG() we had. | |
1155 | */ | |
4a2b4b22 | 1156 | preempt_count_set(count); |
576da126 TG |
1157 | } |
1158 | } | |
1159 | ||
2aae4a10 REB |
1160 | #define INDEX(N) ((base->timer_jiffies >> (TVR_BITS + (N) * TVN_BITS)) & TVN_MASK) |
1161 | ||
1162 | /** | |
1da177e4 LT |
1163 | * __run_timers - run all expired timers (if any) on this CPU. |
1164 | * @base: the timer vector to be processed. | |
1165 | * | |
1166 | * This function cascades all vectors and executes all expired timer | |
1167 | * vectors. | |
1168 | */ | |
a6fa8e5a | 1169 | static inline void __run_timers(struct tvec_base *base) |
1da177e4 LT |
1170 | { |
1171 | struct timer_list *timer; | |
1172 | ||
3691c519 | 1173 | spin_lock_irq(&base->lock); |
3bb475a3 | 1174 | |
1da177e4 | 1175 | while (time_after_eq(jiffies, base->timer_jiffies)) { |
1dabbcec TG |
1176 | struct hlist_head work_list; |
1177 | struct hlist_head *head = &work_list; | |
3bb475a3 TG |
1178 | int index; |
1179 | ||
1180 | if (!base->all_timers) { | |
1181 | base->timer_jiffies = jiffies; | |
1182 | break; | |
1183 | } | |
1184 | ||
1185 | index = base->timer_jiffies & TVR_MASK; | |
626ab0e6 | 1186 | |
1da177e4 LT |
1187 | /* |
1188 | * Cascade timers: | |
1189 | */ | |
1190 | if (!index && | |
1191 | (!cascade(base, &base->tv2, INDEX(0))) && | |
1192 | (!cascade(base, &base->tv3, INDEX(1))) && | |
1193 | !cascade(base, &base->tv4, INDEX(2))) | |
1194 | cascade(base, &base->tv5, INDEX(3)); | |
626ab0e6 | 1195 | ++base->timer_jiffies; |
1dabbcec TG |
1196 | hlist_move_list(base->tv1.vec + index, head); |
1197 | while (!hlist_empty(head)) { | |
1da177e4 LT |
1198 | void (*fn)(unsigned long); |
1199 | unsigned long data; | |
c5f66e99 | 1200 | bool irqsafe; |
1da177e4 | 1201 | |
1dabbcec | 1202 | timer = hlist_entry(head->first, struct timer_list, entry); |
6819457d TG |
1203 | fn = timer->function; |
1204 | data = timer->data; | |
c5f66e99 | 1205 | irqsafe = tbase_get_irqsafe(timer->base); |
1da177e4 | 1206 | |
82f67cd9 IM |
1207 | timer_stats_account_timer(timer); |
1208 | ||
6f1bc451 | 1209 | base->running_timer = timer; |
99d5f3aa | 1210 | detach_expired_timer(timer, base); |
6f2b9b9a | 1211 | |
c5f66e99 TH |
1212 | if (irqsafe) { |
1213 | spin_unlock(&base->lock); | |
1214 | call_timer_fn(timer, fn, data); | |
1215 | spin_lock(&base->lock); | |
1216 | } else { | |
1217 | spin_unlock_irq(&base->lock); | |
1218 | call_timer_fn(timer, fn, data); | |
1219 | spin_lock_irq(&base->lock); | |
1220 | } | |
1da177e4 LT |
1221 | } |
1222 | } | |
6f1bc451 | 1223 | base->running_timer = NULL; |
3691c519 | 1224 | spin_unlock_irq(&base->lock); |
1da177e4 LT |
1225 | } |
1226 | ||
3451d024 | 1227 | #ifdef CONFIG_NO_HZ_COMMON |
1da177e4 LT |
1228 | /* |
1229 | * Find out when the next timer event is due to happen. This | |
90cba64a RD |
1230 | * is used on S/390 to stop all activity when a CPU is idle. |
1231 | * This function needs to be called with interrupts disabled. | |
1da177e4 | 1232 | */ |
a6fa8e5a | 1233 | static unsigned long __next_timer_interrupt(struct tvec_base *base) |
1da177e4 | 1234 | { |
1cfd6849 | 1235 | unsigned long timer_jiffies = base->timer_jiffies; |
eaad084b | 1236 | unsigned long expires = timer_jiffies + NEXT_TIMER_MAX_DELTA; |
1cfd6849 | 1237 | int index, slot, array, found = 0; |
1da177e4 | 1238 | struct timer_list *nte; |
a6fa8e5a | 1239 | struct tvec *varray[4]; |
1da177e4 LT |
1240 | |
1241 | /* Look for timer events in tv1. */ | |
1cfd6849 | 1242 | index = slot = timer_jiffies & TVR_MASK; |
1da177e4 | 1243 | do { |
1dabbcec | 1244 | hlist_for_each_entry(nte, base->tv1.vec + slot, entry) { |
6819457d TG |
1245 | if (tbase_get_deferrable(nte->base)) |
1246 | continue; | |
6e453a67 | 1247 | |
1cfd6849 | 1248 | found = 1; |
1da177e4 | 1249 | expires = nte->expires; |
1cfd6849 TG |
1250 | /* Look at the cascade bucket(s)? */ |
1251 | if (!index || slot < index) | |
1252 | goto cascade; | |
1253 | return expires; | |
1da177e4 | 1254 | } |
1cfd6849 TG |
1255 | slot = (slot + 1) & TVR_MASK; |
1256 | } while (slot != index); | |
1257 | ||
1258 | cascade: | |
1259 | /* Calculate the next cascade event */ | |
1260 | if (index) | |
1261 | timer_jiffies += TVR_SIZE - index; | |
1262 | timer_jiffies >>= TVR_BITS; | |
1da177e4 LT |
1263 | |
1264 | /* Check tv2-tv5. */ | |
1265 | varray[0] = &base->tv2; | |
1266 | varray[1] = &base->tv3; | |
1267 | varray[2] = &base->tv4; | |
1268 | varray[3] = &base->tv5; | |
1cfd6849 TG |
1269 | |
1270 | for (array = 0; array < 4; array++) { | |
a6fa8e5a | 1271 | struct tvec *varp = varray[array]; |
1cfd6849 TG |
1272 | |
1273 | index = slot = timer_jiffies & TVN_MASK; | |
1da177e4 | 1274 | do { |
1dabbcec | 1275 | hlist_for_each_entry(nte, varp->vec + slot, entry) { |
a0419888 JH |
1276 | if (tbase_get_deferrable(nte->base)) |
1277 | continue; | |
1278 | ||
1cfd6849 | 1279 | found = 1; |
1da177e4 LT |
1280 | if (time_before(nte->expires, expires)) |
1281 | expires = nte->expires; | |
1cfd6849 TG |
1282 | } |
1283 | /* | |
1284 | * Do we still search for the first timer or are | |
1285 | * we looking up the cascade buckets ? | |
1286 | */ | |
1287 | if (found) { | |
1288 | /* Look at the cascade bucket(s)? */ | |
1289 | if (!index || slot < index) | |
1290 | break; | |
1291 | return expires; | |
1292 | } | |
1293 | slot = (slot + 1) & TVN_MASK; | |
1294 | } while (slot != index); | |
1295 | ||
1296 | if (index) | |
1297 | timer_jiffies += TVN_SIZE - index; | |
1298 | timer_jiffies >>= TVN_BITS; | |
1da177e4 | 1299 | } |
1cfd6849 TG |
1300 | return expires; |
1301 | } | |
69239749 | 1302 | |
1cfd6849 TG |
1303 | /* |
1304 | * Check, if the next hrtimer event is before the next timer wheel | |
1305 | * event: | |
1306 | */ | |
c1ad348b | 1307 | static u64 cmp_next_hrtimer_event(u64 basem, u64 expires) |
1cfd6849 | 1308 | { |
c1ad348b | 1309 | u64 nextevt = hrtimer_get_next_event(); |
0662b713 | 1310 | |
9501b6cf | 1311 | /* |
c1ad348b TG |
1312 | * If high resolution timers are enabled |
1313 | * hrtimer_get_next_event() returns KTIME_MAX. | |
9501b6cf | 1314 | */ |
c1ad348b TG |
1315 | if (expires <= nextevt) |
1316 | return expires; | |
eaad084b TG |
1317 | |
1318 | /* | |
c1ad348b TG |
1319 | * If the next timer is already expired, return the tick base |
1320 | * time so the tick is fired immediately. | |
eaad084b | 1321 | */ |
c1ad348b TG |
1322 | if (nextevt <= basem) |
1323 | return basem; | |
eaad084b | 1324 | |
9501b6cf | 1325 | /* |
c1ad348b TG |
1326 | * Round up to the next jiffie. High resolution timers are |
1327 | * off, so the hrtimers are expired in the tick and we need to | |
1328 | * make sure that this tick really expires the timer to avoid | |
1329 | * a ping pong of the nohz stop code. | |
1330 | * | |
1331 | * Use DIV_ROUND_UP_ULL to prevent gcc calling __divdi3 | |
9501b6cf | 1332 | */ |
c1ad348b | 1333 | return DIV_ROUND_UP_ULL(nextevt, TICK_NSEC) * TICK_NSEC; |
1da177e4 | 1334 | } |
1cfd6849 TG |
1335 | |
1336 | /** | |
c1ad348b TG |
1337 | * get_next_timer_interrupt - return the time (clock mono) of the next timer |
1338 | * @basej: base time jiffies | |
1339 | * @basem: base time clock monotonic | |
1340 | * | |
1341 | * Returns the tick aligned clock monotonic time of the next pending | |
1342 | * timer or KTIME_MAX if no timer is pending. | |
1cfd6849 | 1343 | */ |
c1ad348b | 1344 | u64 get_next_timer_interrupt(unsigned long basej, u64 basem) |
1cfd6849 | 1345 | { |
7496351a | 1346 | struct tvec_base *base = __this_cpu_read(tvec_bases); |
c1ad348b TG |
1347 | u64 expires = KTIME_MAX; |
1348 | unsigned long nextevt; | |
1cfd6849 | 1349 | |
dbd87b5a HC |
1350 | /* |
1351 | * Pretend that there is no timer pending if the cpu is offline. | |
1352 | * Possible pending timers will be migrated later to an active cpu. | |
1353 | */ | |
1354 | if (cpu_is_offline(smp_processor_id())) | |
e40468a5 TG |
1355 | return expires; |
1356 | ||
1cfd6849 | 1357 | spin_lock(&base->lock); |
e40468a5 TG |
1358 | if (base->active_timers) { |
1359 | if (time_before_eq(base->next_timer, base->timer_jiffies)) | |
1360 | base->next_timer = __next_timer_interrupt(base); | |
c1ad348b TG |
1361 | nextevt = base->next_timer; |
1362 | if (time_before_eq(nextevt, basej)) | |
1363 | expires = basem; | |
1364 | else | |
1365 | expires = basem + (nextevt - basej) * TICK_NSEC; | |
e40468a5 | 1366 | } |
1cfd6849 TG |
1367 | spin_unlock(&base->lock); |
1368 | ||
c1ad348b | 1369 | return cmp_next_hrtimer_event(basem, expires); |
1cfd6849 | 1370 | } |
1da177e4 LT |
1371 | #endif |
1372 | ||
1da177e4 | 1373 | /* |
5b4db0c2 | 1374 | * Called from the timer interrupt handler to charge one tick to the current |
1da177e4 LT |
1375 | * process. user_tick is 1 if the tick is user time, 0 for system. |
1376 | */ | |
1377 | void update_process_times(int user_tick) | |
1378 | { | |
1379 | struct task_struct *p = current; | |
1da177e4 LT |
1380 | |
1381 | /* Note: this timer irq context must be accounted for as well. */ | |
fa13a5a1 | 1382 | account_process_tick(p, user_tick); |
1da177e4 | 1383 | run_local_timers(); |
c3377c2d | 1384 | rcu_check_callbacks(user_tick); |
e360adbe PZ |
1385 | #ifdef CONFIG_IRQ_WORK |
1386 | if (in_irq()) | |
76a33061 | 1387 | irq_work_tick(); |
e360adbe | 1388 | #endif |
1da177e4 | 1389 | scheduler_tick(); |
6819457d | 1390 | run_posix_cpu_timers(p); |
1da177e4 LT |
1391 | } |
1392 | ||
1da177e4 LT |
1393 | /* |
1394 | * This function runs timers and the timer-tq in bottom half context. | |
1395 | */ | |
1396 | static void run_timer_softirq(struct softirq_action *h) | |
1397 | { | |
7496351a | 1398 | struct tvec_base *base = __this_cpu_read(tvec_bases); |
1da177e4 LT |
1399 | |
1400 | if (time_after_eq(jiffies, base->timer_jiffies)) | |
1401 | __run_timers(base); | |
1402 | } | |
1403 | ||
1404 | /* | |
1405 | * Called by the local, per-CPU timer interrupt on SMP. | |
1406 | */ | |
1407 | void run_local_timers(void) | |
1408 | { | |
d3d74453 | 1409 | hrtimer_run_queues(); |
1da177e4 LT |
1410 | raise_softirq(TIMER_SOFTIRQ); |
1411 | } | |
1412 | ||
1da177e4 LT |
1413 | #ifdef __ARCH_WANT_SYS_ALARM |
1414 | ||
1415 | /* | |
1416 | * For backwards compatibility? This can be done in libc so Alpha | |
1417 | * and all newer ports shouldn't need it. | |
1418 | */ | |
58fd3aa2 | 1419 | SYSCALL_DEFINE1(alarm, unsigned int, seconds) |
1da177e4 | 1420 | { |
c08b8a49 | 1421 | return alarm_setitimer(seconds); |
1da177e4 LT |
1422 | } |
1423 | ||
1424 | #endif | |
1425 | ||
1da177e4 LT |
1426 | static void process_timeout(unsigned long __data) |
1427 | { | |
36c8b586 | 1428 | wake_up_process((struct task_struct *)__data); |
1da177e4 LT |
1429 | } |
1430 | ||
1431 | /** | |
1432 | * schedule_timeout - sleep until timeout | |
1433 | * @timeout: timeout value in jiffies | |
1434 | * | |
1435 | * Make the current task sleep until @timeout jiffies have | |
1436 | * elapsed. The routine will return immediately unless | |
1437 | * the current task state has been set (see set_current_state()). | |
1438 | * | |
1439 | * You can set the task state as follows - | |
1440 | * | |
1441 | * %TASK_UNINTERRUPTIBLE - at least @timeout jiffies are guaranteed to | |
1442 | * pass before the routine returns. The routine will return 0 | |
1443 | * | |
1444 | * %TASK_INTERRUPTIBLE - the routine may return early if a signal is | |
1445 | * delivered to the current task. In this case the remaining time | |
1446 | * in jiffies will be returned, or 0 if the timer expired in time | |
1447 | * | |
1448 | * The current task state is guaranteed to be TASK_RUNNING when this | |
1449 | * routine returns. | |
1450 | * | |
1451 | * Specifying a @timeout value of %MAX_SCHEDULE_TIMEOUT will schedule | |
1452 | * the CPU away without a bound on the timeout. In this case the return | |
1453 | * value will be %MAX_SCHEDULE_TIMEOUT. | |
1454 | * | |
1455 | * In all cases the return value is guaranteed to be non-negative. | |
1456 | */ | |
7ad5b3a5 | 1457 | signed long __sched schedule_timeout(signed long timeout) |
1da177e4 LT |
1458 | { |
1459 | struct timer_list timer; | |
1460 | unsigned long expire; | |
1461 | ||
1462 | switch (timeout) | |
1463 | { | |
1464 | case MAX_SCHEDULE_TIMEOUT: | |
1465 | /* | |
1466 | * These two special cases are useful to be comfortable | |
1467 | * in the caller. Nothing more. We could take | |
1468 | * MAX_SCHEDULE_TIMEOUT from one of the negative value | |
1469 | * but I' d like to return a valid offset (>=0) to allow | |
1470 | * the caller to do everything it want with the retval. | |
1471 | */ | |
1472 | schedule(); | |
1473 | goto out; | |
1474 | default: | |
1475 | /* | |
1476 | * Another bit of PARANOID. Note that the retval will be | |
1477 | * 0 since no piece of kernel is supposed to do a check | |
1478 | * for a negative retval of schedule_timeout() (since it | |
1479 | * should never happens anyway). You just have the printk() | |
1480 | * that will tell you if something is gone wrong and where. | |
1481 | */ | |
5b149bcc | 1482 | if (timeout < 0) { |
1da177e4 | 1483 | printk(KERN_ERR "schedule_timeout: wrong timeout " |
5b149bcc AM |
1484 | "value %lx\n", timeout); |
1485 | dump_stack(); | |
1da177e4 LT |
1486 | current->state = TASK_RUNNING; |
1487 | goto out; | |
1488 | } | |
1489 | } | |
1490 | ||
1491 | expire = timeout + jiffies; | |
1492 | ||
c6f3a97f | 1493 | setup_timer_on_stack(&timer, process_timeout, (unsigned long)current); |
597d0275 | 1494 | __mod_timer(&timer, expire, false, TIMER_NOT_PINNED); |
1da177e4 LT |
1495 | schedule(); |
1496 | del_singleshot_timer_sync(&timer); | |
1497 | ||
c6f3a97f TG |
1498 | /* Remove the timer from the object tracker */ |
1499 | destroy_timer_on_stack(&timer); | |
1500 | ||
1da177e4 LT |
1501 | timeout = expire - jiffies; |
1502 | ||
1503 | out: | |
1504 | return timeout < 0 ? 0 : timeout; | |
1505 | } | |
1da177e4 LT |
1506 | EXPORT_SYMBOL(schedule_timeout); |
1507 | ||
8a1c1757 AM |
1508 | /* |
1509 | * We can use __set_current_state() here because schedule_timeout() calls | |
1510 | * schedule() unconditionally. | |
1511 | */ | |
64ed93a2 NA |
1512 | signed long __sched schedule_timeout_interruptible(signed long timeout) |
1513 | { | |
a5a0d52c AM |
1514 | __set_current_state(TASK_INTERRUPTIBLE); |
1515 | return schedule_timeout(timeout); | |
64ed93a2 NA |
1516 | } |
1517 | EXPORT_SYMBOL(schedule_timeout_interruptible); | |
1518 | ||
294d5cc2 MW |
1519 | signed long __sched schedule_timeout_killable(signed long timeout) |
1520 | { | |
1521 | __set_current_state(TASK_KILLABLE); | |
1522 | return schedule_timeout(timeout); | |
1523 | } | |
1524 | EXPORT_SYMBOL(schedule_timeout_killable); | |
1525 | ||
64ed93a2 NA |
1526 | signed long __sched schedule_timeout_uninterruptible(signed long timeout) |
1527 | { | |
a5a0d52c AM |
1528 | __set_current_state(TASK_UNINTERRUPTIBLE); |
1529 | return schedule_timeout(timeout); | |
64ed93a2 NA |
1530 | } |
1531 | EXPORT_SYMBOL(schedule_timeout_uninterruptible); | |
1532 | ||
1da177e4 | 1533 | #ifdef CONFIG_HOTPLUG_CPU |
1dabbcec | 1534 | static void migrate_timer_list(struct tvec_base *new_base, struct hlist_head *head) |
1da177e4 LT |
1535 | { |
1536 | struct timer_list *timer; | |
1537 | ||
1dabbcec TG |
1538 | while (!hlist_empty(head)) { |
1539 | timer = hlist_entry(head->first, struct timer_list, entry); | |
99d5f3aa | 1540 | /* We ignore the accounting on the dying cpu */ |
ec44bc7a | 1541 | detach_timer(timer, false); |
6e453a67 | 1542 | timer_set_base(timer, new_base); |
1da177e4 | 1543 | internal_add_timer(new_base, timer); |
1da177e4 | 1544 | } |
1da177e4 LT |
1545 | } |
1546 | ||
0db0628d | 1547 | static void migrate_timers(int cpu) |
1da177e4 | 1548 | { |
a6fa8e5a PM |
1549 | struct tvec_base *old_base; |
1550 | struct tvec_base *new_base; | |
1da177e4 LT |
1551 | int i; |
1552 | ||
1553 | BUG_ON(cpu_online(cpu)); | |
a4a6198b JB |
1554 | old_base = per_cpu(tvec_bases, cpu); |
1555 | new_base = get_cpu_var(tvec_bases); | |
d82f0b0f ON |
1556 | /* |
1557 | * The caller is globally serialized and nobody else | |
1558 | * takes two locks at once, deadlock is not possible. | |
1559 | */ | |
1560 | spin_lock_irq(&new_base->lock); | |
0d180406 | 1561 | spin_lock_nested(&old_base->lock, SINGLE_DEPTH_NESTING); |
3691c519 ON |
1562 | |
1563 | BUG_ON(old_base->running_timer); | |
1da177e4 | 1564 | |
1da177e4 | 1565 | for (i = 0; i < TVR_SIZE; i++) |
55c888d6 ON |
1566 | migrate_timer_list(new_base, old_base->tv1.vec + i); |
1567 | for (i = 0; i < TVN_SIZE; i++) { | |
1568 | migrate_timer_list(new_base, old_base->tv2.vec + i); | |
1569 | migrate_timer_list(new_base, old_base->tv3.vec + i); | |
1570 | migrate_timer_list(new_base, old_base->tv4.vec + i); | |
1571 | migrate_timer_list(new_base, old_base->tv5.vec + i); | |
1572 | } | |
1573 | ||
8def9060 VK |
1574 | old_base->active_timers = 0; |
1575 | old_base->all_timers = 0; | |
1576 | ||
0d180406 | 1577 | spin_unlock(&old_base->lock); |
d82f0b0f | 1578 | spin_unlock_irq(&new_base->lock); |
1da177e4 | 1579 | put_cpu_var(tvec_bases); |
1da177e4 | 1580 | } |
1da177e4 | 1581 | |
0db0628d | 1582 | static int timer_cpu_notify(struct notifier_block *self, |
1da177e4 LT |
1583 | unsigned long action, void *hcpu) |
1584 | { | |
8def9060 | 1585 | switch (action) { |
1da177e4 | 1586 | case CPU_DEAD: |
8bb78442 | 1587 | case CPU_DEAD_FROZEN: |
8def9060 | 1588 | migrate_timers((long)hcpu); |
1da177e4 | 1589 | break; |
1da177e4 LT |
1590 | default: |
1591 | break; | |
1592 | } | |
3650b57f | 1593 | |
1da177e4 LT |
1594 | return NOTIFY_OK; |
1595 | } | |
1596 | ||
3650b57f PZ |
1597 | static inline void timer_register_cpu_notifier(void) |
1598 | { | |
1599 | cpu_notifier(timer_cpu_notify, 0); | |
1600 | } | |
1601 | #else | |
1602 | static inline void timer_register_cpu_notifier(void) { } | |
1603 | #endif /* CONFIG_HOTPLUG_CPU */ | |
1da177e4 | 1604 | |
8def9060 VK |
1605 | static void __init init_timer_cpu(struct tvec_base *base, int cpu) |
1606 | { | |
3650b57f PZ |
1607 | BUG_ON(base != tbase_get_base(base)); |
1608 | ||
8def9060 VK |
1609 | base->cpu = cpu; |
1610 | per_cpu(tvec_bases, cpu) = base; | |
1611 | spin_lock_init(&base->lock); | |
1612 | ||
8def9060 VK |
1613 | base->timer_jiffies = jiffies; |
1614 | base->next_timer = base->timer_jiffies; | |
1615 | } | |
1616 | ||
1617 | static void __init init_timer_cpus(void) | |
1da177e4 | 1618 | { |
8def9060 VK |
1619 | struct tvec_base *base; |
1620 | int local_cpu = smp_processor_id(); | |
1621 | int cpu; | |
1622 | ||
1623 | for_each_possible_cpu(cpu) { | |
1624 | if (cpu == local_cpu) | |
1625 | base = &boot_tvec_bases; | |
3650b57f | 1626 | #ifdef CONFIG_SMP |
8def9060 VK |
1627 | else |
1628 | base = per_cpu_ptr(&__tvec_bases, cpu); | |
3650b57f | 1629 | #endif |
8def9060 VK |
1630 | |
1631 | init_timer_cpu(base, cpu); | |
1632 | } | |
1633 | } | |
e52b1db3 | 1634 | |
8def9060 VK |
1635 | void __init init_timers(void) |
1636 | { | |
e52b1db3 TH |
1637 | /* ensure there are enough low bits for flags in timer->base pointer */ |
1638 | BUILD_BUG_ON(__alignof__(struct tvec_base) & TIMER_FLAG_MASK); | |
07dccf33 | 1639 | |
8def9060 | 1640 | init_timer_cpus(); |
c24a4a36 | 1641 | init_timer_stats(); |
3650b57f | 1642 | timer_register_cpu_notifier(); |
962cf36c | 1643 | open_softirq(TIMER_SOFTIRQ, run_timer_softirq); |
1da177e4 LT |
1644 | } |
1645 | ||
1da177e4 LT |
1646 | /** |
1647 | * msleep - sleep safely even with waitqueue interruptions | |
1648 | * @msecs: Time in milliseconds to sleep for | |
1649 | */ | |
1650 | void msleep(unsigned int msecs) | |
1651 | { | |
1652 | unsigned long timeout = msecs_to_jiffies(msecs) + 1; | |
1653 | ||
75bcc8c5 NA |
1654 | while (timeout) |
1655 | timeout = schedule_timeout_uninterruptible(timeout); | |
1da177e4 LT |
1656 | } |
1657 | ||
1658 | EXPORT_SYMBOL(msleep); | |
1659 | ||
1660 | /** | |
96ec3efd | 1661 | * msleep_interruptible - sleep waiting for signals |
1da177e4 LT |
1662 | * @msecs: Time in milliseconds to sleep for |
1663 | */ | |
1664 | unsigned long msleep_interruptible(unsigned int msecs) | |
1665 | { | |
1666 | unsigned long timeout = msecs_to_jiffies(msecs) + 1; | |
1667 | ||
75bcc8c5 NA |
1668 | while (timeout && !signal_pending(current)) |
1669 | timeout = schedule_timeout_interruptible(timeout); | |
1da177e4 LT |
1670 | return jiffies_to_msecs(timeout); |
1671 | } | |
1672 | ||
1673 | EXPORT_SYMBOL(msleep_interruptible); | |
5e7f5a17 | 1674 | |
6deba083 | 1675 | static void __sched do_usleep_range(unsigned long min, unsigned long max) |
5e7f5a17 PP |
1676 | { |
1677 | ktime_t kmin; | |
1678 | unsigned long delta; | |
1679 | ||
1680 | kmin = ktime_set(0, min * NSEC_PER_USEC); | |
1681 | delta = (max - min) * NSEC_PER_USEC; | |
6deba083 | 1682 | schedule_hrtimeout_range(&kmin, delta, HRTIMER_MODE_REL); |
5e7f5a17 PP |
1683 | } |
1684 | ||
1685 | /** | |
1686 | * usleep_range - Drop in replacement for udelay where wakeup is flexible | |
1687 | * @min: Minimum time in usecs to sleep | |
1688 | * @max: Maximum time in usecs to sleep | |
1689 | */ | |
2ad5d327 | 1690 | void __sched usleep_range(unsigned long min, unsigned long max) |
5e7f5a17 PP |
1691 | { |
1692 | __set_current_state(TASK_UNINTERRUPTIBLE); | |
1693 | do_usleep_range(min, max); | |
1694 | } | |
1695 | EXPORT_SYMBOL(usleep_range); |