]>
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> |
cdd6c482 | 40 | #include <linux/perf_event.h> |
eea08f32 | 41 | #include <linux/sched.h> |
1da177e4 LT |
42 | |
43 | #include <asm/uaccess.h> | |
44 | #include <asm/unistd.h> | |
45 | #include <asm/div64.h> | |
46 | #include <asm/timex.h> | |
47 | #include <asm/io.h> | |
48 | ||
2b022e3d XG |
49 | #define CREATE_TRACE_POINTS |
50 | #include <trace/events/timer.h> | |
51 | ||
ecea8d19 TG |
52 | u64 jiffies_64 __cacheline_aligned_in_smp = INITIAL_JIFFIES; |
53 | ||
54 | EXPORT_SYMBOL(jiffies_64); | |
55 | ||
1da177e4 LT |
56 | /* |
57 | * per-CPU timer vector definitions: | |
58 | */ | |
1da177e4 LT |
59 | #define TVN_BITS (CONFIG_BASE_SMALL ? 4 : 6) |
60 | #define TVR_BITS (CONFIG_BASE_SMALL ? 6 : 8) | |
61 | #define TVN_SIZE (1 << TVN_BITS) | |
62 | #define TVR_SIZE (1 << TVR_BITS) | |
63 | #define TVN_MASK (TVN_SIZE - 1) | |
64 | #define TVR_MASK (TVR_SIZE - 1) | |
65 | ||
a6fa8e5a | 66 | struct tvec { |
1da177e4 | 67 | struct list_head vec[TVN_SIZE]; |
a6fa8e5a | 68 | }; |
1da177e4 | 69 | |
a6fa8e5a | 70 | struct tvec_root { |
1da177e4 | 71 | struct list_head vec[TVR_SIZE]; |
a6fa8e5a | 72 | }; |
1da177e4 | 73 | |
a6fa8e5a | 74 | struct tvec_base { |
3691c519 ON |
75 | spinlock_t lock; |
76 | struct timer_list *running_timer; | |
1da177e4 | 77 | unsigned long timer_jiffies; |
97fd9ed4 | 78 | unsigned long next_timer; |
a6fa8e5a PM |
79 | struct tvec_root tv1; |
80 | struct tvec tv2; | |
81 | struct tvec tv3; | |
82 | struct tvec tv4; | |
83 | struct tvec tv5; | |
6e453a67 | 84 | } ____cacheline_aligned; |
1da177e4 | 85 | |
a6fa8e5a | 86 | struct tvec_base boot_tvec_bases; |
3691c519 | 87 | EXPORT_SYMBOL(boot_tvec_bases); |
a6fa8e5a | 88 | static DEFINE_PER_CPU(struct tvec_base *, tvec_bases) = &boot_tvec_bases; |
1da177e4 | 89 | |
6e453a67 | 90 | /* |
a6fa8e5a | 91 | * Note that all tvec_bases are 2 byte aligned and lower bit of |
6e453a67 VP |
92 | * base in timer_list is guaranteed to be zero. Use the LSB for |
93 | * the new flag to indicate whether the timer is deferrable | |
94 | */ | |
95 | #define TBASE_DEFERRABLE_FLAG (0x1) | |
96 | ||
97 | /* Functions below help us manage 'deferrable' flag */ | |
a6fa8e5a | 98 | static inline unsigned int tbase_get_deferrable(struct tvec_base *base) |
6e453a67 | 99 | { |
e9910846 | 100 | return ((unsigned int)(unsigned long)base & TBASE_DEFERRABLE_FLAG); |
6e453a67 VP |
101 | } |
102 | ||
a6fa8e5a | 103 | static inline struct tvec_base *tbase_get_base(struct tvec_base *base) |
6e453a67 | 104 | { |
a6fa8e5a | 105 | return ((struct tvec_base *)((unsigned long)base & ~TBASE_DEFERRABLE_FLAG)); |
6e453a67 VP |
106 | } |
107 | ||
108 | static inline void timer_set_deferrable(struct timer_list *timer) | |
109 | { | |
a6fa8e5a | 110 | timer->base = ((struct tvec_base *)((unsigned long)(timer->base) | |
6819457d | 111 | TBASE_DEFERRABLE_FLAG)); |
6e453a67 VP |
112 | } |
113 | ||
114 | static inline void | |
a6fa8e5a | 115 | timer_set_base(struct timer_list *timer, struct tvec_base *new_base) |
6e453a67 | 116 | { |
a6fa8e5a | 117 | timer->base = (struct tvec_base *)((unsigned long)(new_base) | |
6819457d | 118 | tbase_get_deferrable(timer->base)); |
6e453a67 VP |
119 | } |
120 | ||
9c133c46 AS |
121 | static unsigned long round_jiffies_common(unsigned long j, int cpu, |
122 | bool force_up) | |
4c36a5de AV |
123 | { |
124 | int rem; | |
125 | unsigned long original = j; | |
126 | ||
127 | /* | |
128 | * We don't want all cpus firing their timers at once hitting the | |
129 | * same lock or cachelines, so we skew each extra cpu with an extra | |
130 | * 3 jiffies. This 3 jiffies came originally from the mm/ code which | |
131 | * already did this. | |
132 | * The skew is done by adding 3*cpunr, then round, then subtract this | |
133 | * extra offset again. | |
134 | */ | |
135 | j += cpu * 3; | |
136 | ||
137 | rem = j % HZ; | |
138 | ||
139 | /* | |
140 | * If the target jiffie is just after a whole second (which can happen | |
141 | * due to delays of the timer irq, long irq off times etc etc) then | |
142 | * we should round down to the whole second, not up. Use 1/4th second | |
143 | * as cutoff for this rounding as an extreme upper bound for this. | |
9c133c46 | 144 | * But never round down if @force_up is set. |
4c36a5de | 145 | */ |
9c133c46 | 146 | if (rem < HZ/4 && !force_up) /* round down */ |
4c36a5de AV |
147 | j = j - rem; |
148 | else /* round up */ | |
149 | j = j - rem + HZ; | |
150 | ||
151 | /* now that we have rounded, subtract the extra skew again */ | |
152 | j -= cpu * 3; | |
153 | ||
154 | if (j <= jiffies) /* rounding ate our timeout entirely; */ | |
155 | return original; | |
156 | return j; | |
157 | } | |
9c133c46 AS |
158 | |
159 | /** | |
160 | * __round_jiffies - function to round jiffies to a full second | |
161 | * @j: the time in (absolute) jiffies that should be rounded | |
162 | * @cpu: the processor number on which the timeout will happen | |
163 | * | |
164 | * __round_jiffies() rounds an absolute time in the future (in jiffies) | |
165 | * up or down to (approximately) full seconds. This is useful for timers | |
166 | * for which the exact time they fire does not matter too much, as long as | |
167 | * they fire approximately every X seconds. | |
168 | * | |
169 | * By rounding these timers to whole seconds, all such timers will fire | |
170 | * at the same time, rather than at various times spread out. The goal | |
171 | * of this is to have the CPU wake up less, which saves power. | |
172 | * | |
173 | * The exact rounding is skewed for each processor to avoid all | |
174 | * processors firing at the exact same time, which could lead | |
175 | * to lock contention or spurious cache line bouncing. | |
176 | * | |
177 | * The return value is the rounded version of the @j parameter. | |
178 | */ | |
179 | unsigned long __round_jiffies(unsigned long j, int cpu) | |
180 | { | |
181 | return round_jiffies_common(j, cpu, false); | |
182 | } | |
4c36a5de AV |
183 | EXPORT_SYMBOL_GPL(__round_jiffies); |
184 | ||
185 | /** | |
186 | * __round_jiffies_relative - function to round jiffies to a full second | |
187 | * @j: the time in (relative) jiffies that should be rounded | |
188 | * @cpu: the processor number on which the timeout will happen | |
189 | * | |
72fd4a35 | 190 | * __round_jiffies_relative() rounds a time delta in the future (in jiffies) |
4c36a5de AV |
191 | * up or down to (approximately) full seconds. This is useful for timers |
192 | * for which the exact time they fire does not matter too much, as long as | |
193 | * they fire approximately every X seconds. | |
194 | * | |
195 | * By rounding these timers to whole seconds, all such timers will fire | |
196 | * at the same time, rather than at various times spread out. The goal | |
197 | * of this is to have the CPU wake up less, which saves power. | |
198 | * | |
199 | * The exact rounding is skewed for each processor to avoid all | |
200 | * processors firing at the exact same time, which could lead | |
201 | * to lock contention or spurious cache line bouncing. | |
202 | * | |
72fd4a35 | 203 | * The return value is the rounded version of the @j parameter. |
4c36a5de AV |
204 | */ |
205 | unsigned long __round_jiffies_relative(unsigned long j, int cpu) | |
206 | { | |
9c133c46 AS |
207 | unsigned long j0 = jiffies; |
208 | ||
209 | /* Use j0 because jiffies might change while we run */ | |
210 | return round_jiffies_common(j + j0, cpu, false) - j0; | |
4c36a5de AV |
211 | } |
212 | EXPORT_SYMBOL_GPL(__round_jiffies_relative); | |
213 | ||
214 | /** | |
215 | * round_jiffies - function to round jiffies to a full second | |
216 | * @j: the time in (absolute) jiffies that should be rounded | |
217 | * | |
72fd4a35 | 218 | * round_jiffies() rounds an absolute time in the future (in jiffies) |
4c36a5de AV |
219 | * up or down to (approximately) full seconds. This is useful for timers |
220 | * for which the exact time they fire does not matter too much, as long as | |
221 | * they fire approximately every X seconds. | |
222 | * | |
223 | * By rounding these timers to whole seconds, all such timers will fire | |
224 | * at the same time, rather than at various times spread out. The goal | |
225 | * of this is to have the CPU wake up less, which saves power. | |
226 | * | |
72fd4a35 | 227 | * The return value is the rounded version of the @j parameter. |
4c36a5de AV |
228 | */ |
229 | unsigned long round_jiffies(unsigned long j) | |
230 | { | |
9c133c46 | 231 | return round_jiffies_common(j, raw_smp_processor_id(), false); |
4c36a5de AV |
232 | } |
233 | EXPORT_SYMBOL_GPL(round_jiffies); | |
234 | ||
235 | /** | |
236 | * round_jiffies_relative - function to round jiffies to a full second | |
237 | * @j: the time in (relative) jiffies that should be rounded | |
238 | * | |
72fd4a35 | 239 | * round_jiffies_relative() rounds a time delta in the future (in jiffies) |
4c36a5de AV |
240 | * up or down to (approximately) full seconds. This is useful for timers |
241 | * for which the exact time they fire does not matter too much, as long as | |
242 | * they fire approximately every X seconds. | |
243 | * | |
244 | * By rounding these timers to whole seconds, all such timers will fire | |
245 | * at the same time, rather than at various times spread out. The goal | |
246 | * of this is to have the CPU wake up less, which saves power. | |
247 | * | |
72fd4a35 | 248 | * The return value is the rounded version of the @j parameter. |
4c36a5de AV |
249 | */ |
250 | unsigned long round_jiffies_relative(unsigned long j) | |
251 | { | |
252 | return __round_jiffies_relative(j, raw_smp_processor_id()); | |
253 | } | |
254 | EXPORT_SYMBOL_GPL(round_jiffies_relative); | |
255 | ||
9c133c46 AS |
256 | /** |
257 | * __round_jiffies_up - function to round jiffies up to a full second | |
258 | * @j: the time in (absolute) jiffies that should be rounded | |
259 | * @cpu: the processor number on which the timeout will happen | |
260 | * | |
261 | * This is the same as __round_jiffies() except that it will never | |
262 | * round down. This is useful for timeouts for which the exact time | |
263 | * of firing does not matter too much, as long as they don't fire too | |
264 | * early. | |
265 | */ | |
266 | unsigned long __round_jiffies_up(unsigned long j, int cpu) | |
267 | { | |
268 | return round_jiffies_common(j, cpu, true); | |
269 | } | |
270 | EXPORT_SYMBOL_GPL(__round_jiffies_up); | |
271 | ||
272 | /** | |
273 | * __round_jiffies_up_relative - function to round jiffies up to a full second | |
274 | * @j: the time in (relative) jiffies that should be rounded | |
275 | * @cpu: the processor number on which the timeout will happen | |
276 | * | |
277 | * This is the same as __round_jiffies_relative() except that it will never | |
278 | * round down. This is useful for timeouts for which the exact time | |
279 | * of firing does not matter too much, as long as they don't fire too | |
280 | * early. | |
281 | */ | |
282 | unsigned long __round_jiffies_up_relative(unsigned long j, int cpu) | |
283 | { | |
284 | unsigned long j0 = jiffies; | |
285 | ||
286 | /* Use j0 because jiffies might change while we run */ | |
287 | return round_jiffies_common(j + j0, cpu, true) - j0; | |
288 | } | |
289 | EXPORT_SYMBOL_GPL(__round_jiffies_up_relative); | |
290 | ||
291 | /** | |
292 | * round_jiffies_up - function to round jiffies up to a full second | |
293 | * @j: the time in (absolute) jiffies that should be rounded | |
294 | * | |
295 | * This is the same as round_jiffies() except that it will never | |
296 | * round down. This is useful for timeouts for which the exact time | |
297 | * of firing does not matter too much, as long as they don't fire too | |
298 | * early. | |
299 | */ | |
300 | unsigned long round_jiffies_up(unsigned long j) | |
301 | { | |
302 | return round_jiffies_common(j, raw_smp_processor_id(), true); | |
303 | } | |
304 | EXPORT_SYMBOL_GPL(round_jiffies_up); | |
305 | ||
306 | /** | |
307 | * round_jiffies_up_relative - function to round jiffies up to a full second | |
308 | * @j: the time in (relative) jiffies that should be rounded | |
309 | * | |
310 | * This is the same as round_jiffies_relative() except that it will never | |
311 | * round down. This is useful for timeouts for which the exact time | |
312 | * of firing does not matter too much, as long as they don't fire too | |
313 | * early. | |
314 | */ | |
315 | unsigned long round_jiffies_up_relative(unsigned long j) | |
316 | { | |
317 | return __round_jiffies_up_relative(j, raw_smp_processor_id()); | |
318 | } | |
319 | EXPORT_SYMBOL_GPL(round_jiffies_up_relative); | |
320 | ||
3bbb9ec9 AV |
321 | /** |
322 | * set_timer_slack - set the allowed slack for a timer | |
323 | * @slack_hz: the amount of time (in jiffies) allowed for rounding | |
324 | * | |
325 | * Set the amount of time, in jiffies, that a certain timer has | |
326 | * in terms of slack. By setting this value, the timer subsystem | |
327 | * will schedule the actual timer somewhere between | |
328 | * the time mod_timer() asks for, and that time plus the slack. | |
329 | * | |
330 | * By setting the slack to -1, a percentage of the delay is used | |
331 | * instead. | |
332 | */ | |
333 | void set_timer_slack(struct timer_list *timer, int slack_hz) | |
334 | { | |
335 | timer->slack = slack_hz; | |
336 | } | |
337 | EXPORT_SYMBOL_GPL(set_timer_slack); | |
338 | ||
4c36a5de | 339 | |
a6fa8e5a | 340 | static inline void set_running_timer(struct tvec_base *base, |
1da177e4 LT |
341 | struct timer_list *timer) |
342 | { | |
343 | #ifdef CONFIG_SMP | |
3691c519 | 344 | base->running_timer = timer; |
1da177e4 LT |
345 | #endif |
346 | } | |
347 | ||
a6fa8e5a | 348 | static void internal_add_timer(struct tvec_base *base, struct timer_list *timer) |
1da177e4 LT |
349 | { |
350 | unsigned long expires = timer->expires; | |
351 | unsigned long idx = expires - base->timer_jiffies; | |
352 | struct list_head *vec; | |
353 | ||
354 | if (idx < TVR_SIZE) { | |
355 | int i = expires & TVR_MASK; | |
356 | vec = base->tv1.vec + i; | |
357 | } else if (idx < 1 << (TVR_BITS + TVN_BITS)) { | |
358 | int i = (expires >> TVR_BITS) & TVN_MASK; | |
359 | vec = base->tv2.vec + i; | |
360 | } else if (idx < 1 << (TVR_BITS + 2 * TVN_BITS)) { | |
361 | int i = (expires >> (TVR_BITS + TVN_BITS)) & TVN_MASK; | |
362 | vec = base->tv3.vec + i; | |
363 | } else if (idx < 1 << (TVR_BITS + 3 * TVN_BITS)) { | |
364 | int i = (expires >> (TVR_BITS + 2 * TVN_BITS)) & TVN_MASK; | |
365 | vec = base->tv4.vec + i; | |
366 | } else if ((signed long) idx < 0) { | |
367 | /* | |
368 | * Can happen if you add a timer with expires == jiffies, | |
369 | * or you set a timer to go off in the past | |
370 | */ | |
371 | vec = base->tv1.vec + (base->timer_jiffies & TVR_MASK); | |
372 | } else { | |
373 | int i; | |
374 | /* If the timeout is larger than 0xffffffff on 64-bit | |
375 | * architectures then we use the maximum timeout: | |
376 | */ | |
377 | if (idx > 0xffffffffUL) { | |
378 | idx = 0xffffffffUL; | |
379 | expires = idx + base->timer_jiffies; | |
380 | } | |
381 | i = (expires >> (TVR_BITS + 3 * TVN_BITS)) & TVN_MASK; | |
382 | vec = base->tv5.vec + i; | |
383 | } | |
384 | /* | |
385 | * Timers are FIFO: | |
386 | */ | |
387 | list_add_tail(&timer->entry, vec); | |
388 | } | |
389 | ||
82f67cd9 IM |
390 | #ifdef CONFIG_TIMER_STATS |
391 | void __timer_stats_timer_set_start_info(struct timer_list *timer, void *addr) | |
392 | { | |
393 | if (timer->start_site) | |
394 | return; | |
395 | ||
396 | timer->start_site = addr; | |
397 | memcpy(timer->start_comm, current->comm, TASK_COMM_LEN); | |
398 | timer->start_pid = current->pid; | |
399 | } | |
c5c061b8 VP |
400 | |
401 | static void timer_stats_account_timer(struct timer_list *timer) | |
402 | { | |
403 | unsigned int flag = 0; | |
404 | ||
507e1231 HC |
405 | if (likely(!timer->start_site)) |
406 | return; | |
c5c061b8 VP |
407 | if (unlikely(tbase_get_deferrable(timer->base))) |
408 | flag |= TIMER_STATS_FLAG_DEFERRABLE; | |
409 | ||
410 | timer_stats_update_stats(timer, timer->start_pid, timer->start_site, | |
411 | timer->function, timer->start_comm, flag); | |
412 | } | |
413 | ||
414 | #else | |
415 | static void timer_stats_account_timer(struct timer_list *timer) {} | |
82f67cd9 IM |
416 | #endif |
417 | ||
c6f3a97f TG |
418 | #ifdef CONFIG_DEBUG_OBJECTS_TIMERS |
419 | ||
420 | static struct debug_obj_descr timer_debug_descr; | |
421 | ||
422 | /* | |
423 | * fixup_init is called when: | |
424 | * - an active object is initialized | |
55c888d6 | 425 | */ |
c6f3a97f TG |
426 | static int timer_fixup_init(void *addr, enum debug_obj_state state) |
427 | { | |
428 | struct timer_list *timer = addr; | |
429 | ||
430 | switch (state) { | |
431 | case ODEBUG_STATE_ACTIVE: | |
432 | del_timer_sync(timer); | |
433 | debug_object_init(timer, &timer_debug_descr); | |
434 | return 1; | |
435 | default: | |
436 | return 0; | |
437 | } | |
438 | } | |
439 | ||
440 | /* | |
441 | * fixup_activate is called when: | |
442 | * - an active object is activated | |
443 | * - an unknown object is activated (might be a statically initialized object) | |
444 | */ | |
445 | static int timer_fixup_activate(void *addr, enum debug_obj_state state) | |
446 | { | |
447 | struct timer_list *timer = addr; | |
448 | ||
449 | switch (state) { | |
450 | ||
451 | case ODEBUG_STATE_NOTAVAILABLE: | |
452 | /* | |
453 | * This is not really a fixup. The timer was | |
454 | * statically initialized. We just make sure that it | |
455 | * is tracked in the object tracker. | |
456 | */ | |
457 | if (timer->entry.next == NULL && | |
458 | timer->entry.prev == TIMER_ENTRY_STATIC) { | |
459 | debug_object_init(timer, &timer_debug_descr); | |
460 | debug_object_activate(timer, &timer_debug_descr); | |
461 | return 0; | |
462 | } else { | |
463 | WARN_ON_ONCE(1); | |
464 | } | |
465 | return 0; | |
466 | ||
467 | case ODEBUG_STATE_ACTIVE: | |
468 | WARN_ON(1); | |
469 | ||
470 | default: | |
471 | return 0; | |
472 | } | |
473 | } | |
474 | ||
475 | /* | |
476 | * fixup_free is called when: | |
477 | * - an active object is freed | |
478 | */ | |
479 | static int timer_fixup_free(void *addr, enum debug_obj_state state) | |
480 | { | |
481 | struct timer_list *timer = addr; | |
482 | ||
483 | switch (state) { | |
484 | case ODEBUG_STATE_ACTIVE: | |
485 | del_timer_sync(timer); | |
486 | debug_object_free(timer, &timer_debug_descr); | |
487 | return 1; | |
488 | default: | |
489 | return 0; | |
490 | } | |
491 | } | |
492 | ||
493 | static struct debug_obj_descr timer_debug_descr = { | |
494 | .name = "timer_list", | |
495 | .fixup_init = timer_fixup_init, | |
496 | .fixup_activate = timer_fixup_activate, | |
497 | .fixup_free = timer_fixup_free, | |
498 | }; | |
499 | ||
500 | static inline void debug_timer_init(struct timer_list *timer) | |
501 | { | |
502 | debug_object_init(timer, &timer_debug_descr); | |
503 | } | |
504 | ||
505 | static inline void debug_timer_activate(struct timer_list *timer) | |
506 | { | |
507 | debug_object_activate(timer, &timer_debug_descr); | |
508 | } | |
509 | ||
510 | static inline void debug_timer_deactivate(struct timer_list *timer) | |
511 | { | |
512 | debug_object_deactivate(timer, &timer_debug_descr); | |
513 | } | |
514 | ||
515 | static inline void debug_timer_free(struct timer_list *timer) | |
516 | { | |
517 | debug_object_free(timer, &timer_debug_descr); | |
518 | } | |
519 | ||
6f2b9b9a JB |
520 | static void __init_timer(struct timer_list *timer, |
521 | const char *name, | |
522 | struct lock_class_key *key); | |
c6f3a97f | 523 | |
6f2b9b9a JB |
524 | void init_timer_on_stack_key(struct timer_list *timer, |
525 | const char *name, | |
526 | struct lock_class_key *key) | |
c6f3a97f TG |
527 | { |
528 | debug_object_init_on_stack(timer, &timer_debug_descr); | |
6f2b9b9a | 529 | __init_timer(timer, name, key); |
c6f3a97f | 530 | } |
6f2b9b9a | 531 | EXPORT_SYMBOL_GPL(init_timer_on_stack_key); |
c6f3a97f TG |
532 | |
533 | void destroy_timer_on_stack(struct timer_list *timer) | |
534 | { | |
535 | debug_object_free(timer, &timer_debug_descr); | |
536 | } | |
537 | EXPORT_SYMBOL_GPL(destroy_timer_on_stack); | |
538 | ||
539 | #else | |
540 | static inline void debug_timer_init(struct timer_list *timer) { } | |
541 | static inline void debug_timer_activate(struct timer_list *timer) { } | |
542 | static inline void debug_timer_deactivate(struct timer_list *timer) { } | |
543 | #endif | |
544 | ||
2b022e3d XG |
545 | static inline void debug_init(struct timer_list *timer) |
546 | { | |
547 | debug_timer_init(timer); | |
548 | trace_timer_init(timer); | |
549 | } | |
550 | ||
551 | static inline void | |
552 | debug_activate(struct timer_list *timer, unsigned long expires) | |
553 | { | |
554 | debug_timer_activate(timer); | |
555 | trace_timer_start(timer, expires); | |
556 | } | |
557 | ||
558 | static inline void debug_deactivate(struct timer_list *timer) | |
559 | { | |
560 | debug_timer_deactivate(timer); | |
561 | trace_timer_cancel(timer); | |
562 | } | |
563 | ||
6f2b9b9a JB |
564 | static void __init_timer(struct timer_list *timer, |
565 | const char *name, | |
566 | struct lock_class_key *key) | |
55c888d6 ON |
567 | { |
568 | timer->entry.next = NULL; | |
bfe5d834 | 569 | timer->base = __raw_get_cpu_var(tvec_bases); |
3bbb9ec9 | 570 | timer->slack = -1; |
82f67cd9 IM |
571 | #ifdef CONFIG_TIMER_STATS |
572 | timer->start_site = NULL; | |
573 | timer->start_pid = -1; | |
574 | memset(timer->start_comm, 0, TASK_COMM_LEN); | |
575 | #endif | |
6f2b9b9a | 576 | lockdep_init_map(&timer->lockdep_map, name, key, 0); |
55c888d6 | 577 | } |
c6f3a97f TG |
578 | |
579 | /** | |
633fe795 | 580 | * init_timer_key - initialize a timer |
c6f3a97f | 581 | * @timer: the timer to be initialized |
633fe795 RD |
582 | * @name: name of the timer |
583 | * @key: lockdep class key of the fake lock used for tracking timer | |
584 | * sync lock dependencies | |
c6f3a97f | 585 | * |
633fe795 | 586 | * init_timer_key() must be done to a timer prior calling *any* of the |
c6f3a97f TG |
587 | * other timer functions. |
588 | */ | |
6f2b9b9a JB |
589 | void init_timer_key(struct timer_list *timer, |
590 | const char *name, | |
591 | struct lock_class_key *key) | |
c6f3a97f | 592 | { |
2b022e3d | 593 | debug_init(timer); |
6f2b9b9a | 594 | __init_timer(timer, name, key); |
c6f3a97f | 595 | } |
6f2b9b9a | 596 | EXPORT_SYMBOL(init_timer_key); |
55c888d6 | 597 | |
6f2b9b9a JB |
598 | void init_timer_deferrable_key(struct timer_list *timer, |
599 | const char *name, | |
600 | struct lock_class_key *key) | |
6e453a67 | 601 | { |
6f2b9b9a | 602 | init_timer_key(timer, name, key); |
6e453a67 VP |
603 | timer_set_deferrable(timer); |
604 | } | |
6f2b9b9a | 605 | EXPORT_SYMBOL(init_timer_deferrable_key); |
6e453a67 | 606 | |
55c888d6 | 607 | static inline void detach_timer(struct timer_list *timer, |
82f67cd9 | 608 | int clear_pending) |
55c888d6 ON |
609 | { |
610 | struct list_head *entry = &timer->entry; | |
611 | ||
2b022e3d | 612 | debug_deactivate(timer); |
c6f3a97f | 613 | |
55c888d6 ON |
614 | __list_del(entry->prev, entry->next); |
615 | if (clear_pending) | |
616 | entry->next = NULL; | |
617 | entry->prev = LIST_POISON2; | |
618 | } | |
619 | ||
620 | /* | |
3691c519 | 621 | * We are using hashed locking: holding per_cpu(tvec_bases).lock |
55c888d6 ON |
622 | * means that all timers which are tied to this base via timer->base are |
623 | * locked, and the base itself is locked too. | |
624 | * | |
625 | * So __run_timers/migrate_timers can safely modify all timers which could | |
626 | * be found on ->tvX lists. | |
627 | * | |
628 | * When the timer's base is locked, and the timer removed from list, it is | |
629 | * possible to set timer->base = NULL and drop the lock: the timer remains | |
630 | * locked. | |
631 | */ | |
a6fa8e5a | 632 | static struct tvec_base *lock_timer_base(struct timer_list *timer, |
55c888d6 | 633 | unsigned long *flags) |
89e7e374 | 634 | __acquires(timer->base->lock) |
55c888d6 | 635 | { |
a6fa8e5a | 636 | struct tvec_base *base; |
55c888d6 ON |
637 | |
638 | for (;;) { | |
a6fa8e5a | 639 | struct tvec_base *prelock_base = timer->base; |
6e453a67 | 640 | base = tbase_get_base(prelock_base); |
55c888d6 ON |
641 | if (likely(base != NULL)) { |
642 | spin_lock_irqsave(&base->lock, *flags); | |
6e453a67 | 643 | if (likely(prelock_base == timer->base)) |
55c888d6 ON |
644 | return base; |
645 | /* The timer has migrated to another CPU */ | |
646 | spin_unlock_irqrestore(&base->lock, *flags); | |
647 | } | |
648 | cpu_relax(); | |
649 | } | |
650 | } | |
651 | ||
74019224 | 652 | static inline int |
597d0275 AB |
653 | __mod_timer(struct timer_list *timer, unsigned long expires, |
654 | bool pending_only, int pinned) | |
1da177e4 | 655 | { |
a6fa8e5a | 656 | struct tvec_base *base, *new_base; |
1da177e4 | 657 | unsigned long flags; |
eea08f32 | 658 | int ret = 0 , cpu; |
1da177e4 | 659 | |
82f67cd9 | 660 | timer_stats_timer_set_start_info(timer); |
1da177e4 | 661 | BUG_ON(!timer->function); |
1da177e4 | 662 | |
55c888d6 ON |
663 | base = lock_timer_base(timer, &flags); |
664 | ||
665 | if (timer_pending(timer)) { | |
666 | detach_timer(timer, 0); | |
97fd9ed4 MS |
667 | if (timer->expires == base->next_timer && |
668 | !tbase_get_deferrable(timer->base)) | |
669 | base->next_timer = base->timer_jiffies; | |
55c888d6 | 670 | ret = 1; |
74019224 IM |
671 | } else { |
672 | if (pending_only) | |
673 | goto out_unlock; | |
55c888d6 ON |
674 | } |
675 | ||
2b022e3d | 676 | debug_activate(timer, expires); |
c6f3a97f | 677 | |
eea08f32 AB |
678 | cpu = smp_processor_id(); |
679 | ||
680 | #if defined(CONFIG_NO_HZ) && defined(CONFIG_SMP) | |
681 | if (!pinned && get_sysctl_timer_migration() && idle_cpu(cpu)) { | |
682 | int preferred_cpu = get_nohz_load_balancer(); | |
683 | ||
684 | if (preferred_cpu >= 0) | |
685 | cpu = preferred_cpu; | |
686 | } | |
687 | #endif | |
688 | new_base = per_cpu(tvec_bases, cpu); | |
689 | ||
3691c519 | 690 | if (base != new_base) { |
1da177e4 | 691 | /* |
55c888d6 ON |
692 | * We are trying to schedule the timer on the local CPU. |
693 | * However we can't change timer's base while it is running, | |
694 | * otherwise del_timer_sync() can't detect that the timer's | |
695 | * handler yet has not finished. This also guarantees that | |
696 | * the timer is serialized wrt itself. | |
1da177e4 | 697 | */ |
a2c348fe | 698 | if (likely(base->running_timer != timer)) { |
55c888d6 | 699 | /* See the comment in lock_timer_base() */ |
6e453a67 | 700 | timer_set_base(timer, NULL); |
55c888d6 | 701 | spin_unlock(&base->lock); |
a2c348fe ON |
702 | base = new_base; |
703 | spin_lock(&base->lock); | |
6e453a67 | 704 | timer_set_base(timer, base); |
1da177e4 LT |
705 | } |
706 | } | |
707 | ||
1da177e4 | 708 | timer->expires = expires; |
97fd9ed4 MS |
709 | if (time_before(timer->expires, base->next_timer) && |
710 | !tbase_get_deferrable(timer->base)) | |
711 | base->next_timer = timer->expires; | |
a2c348fe | 712 | internal_add_timer(base, timer); |
74019224 IM |
713 | |
714 | out_unlock: | |
a2c348fe | 715 | spin_unlock_irqrestore(&base->lock, flags); |
1da177e4 LT |
716 | |
717 | return ret; | |
718 | } | |
719 | ||
2aae4a10 | 720 | /** |
74019224 IM |
721 | * mod_timer_pending - modify a pending timer's timeout |
722 | * @timer: the pending timer to be modified | |
723 | * @expires: new timeout in jiffies | |
1da177e4 | 724 | * |
74019224 IM |
725 | * mod_timer_pending() is the same for pending timers as mod_timer(), |
726 | * but will not re-activate and modify already deleted timers. | |
727 | * | |
728 | * It is useful for unserialized use of timers. | |
1da177e4 | 729 | */ |
74019224 | 730 | int mod_timer_pending(struct timer_list *timer, unsigned long expires) |
1da177e4 | 731 | { |
597d0275 | 732 | return __mod_timer(timer, expires, true, TIMER_NOT_PINNED); |
1da177e4 | 733 | } |
74019224 | 734 | EXPORT_SYMBOL(mod_timer_pending); |
1da177e4 | 735 | |
3bbb9ec9 AV |
736 | /* |
737 | * Decide where to put the timer while taking the slack into account | |
738 | * | |
739 | * Algorithm: | |
740 | * 1) calculate the maximum (absolute) time | |
741 | * 2) calculate the highest bit where the expires and new max are different | |
742 | * 3) use this bit to make a mask | |
743 | * 4) use the bitmask to round down the maximum time, so that all last | |
744 | * bits are zeros | |
745 | */ | |
746 | static inline | |
747 | unsigned long apply_slack(struct timer_list *timer, unsigned long expires) | |
748 | { | |
749 | unsigned long expires_limit, mask; | |
750 | int bit; | |
751 | ||
752 | expires_limit = expires + timer->slack; | |
753 | ||
754 | if (timer->slack < 0) /* auto slack: use 0.4% */ | |
755 | expires_limit = expires + (expires - jiffies)/256; | |
756 | ||
757 | mask = expires ^ expires_limit; | |
758 | ||
759 | if (mask == 0) | |
760 | return expires; | |
761 | ||
762 | bit = find_last_bit(&mask, BITS_PER_LONG); | |
763 | ||
764 | mask = (1 << bit) - 1; | |
765 | ||
766 | expires_limit = expires_limit & ~(mask); | |
767 | ||
768 | return expires_limit; | |
769 | } | |
770 | ||
2aae4a10 | 771 | /** |
1da177e4 LT |
772 | * mod_timer - modify a timer's timeout |
773 | * @timer: the timer to be modified | |
2aae4a10 | 774 | * @expires: new timeout in jiffies |
1da177e4 | 775 | * |
72fd4a35 | 776 | * mod_timer() is a more efficient way to update the expire field of an |
1da177e4 LT |
777 | * active timer (if the timer is inactive it will be activated) |
778 | * | |
779 | * mod_timer(timer, expires) is equivalent to: | |
780 | * | |
781 | * del_timer(timer); timer->expires = expires; add_timer(timer); | |
782 | * | |
783 | * Note that if there are multiple unserialized concurrent users of the | |
784 | * same timer, then mod_timer() is the only safe way to modify the timeout, | |
785 | * since add_timer() cannot modify an already running timer. | |
786 | * | |
787 | * The function returns whether it has modified a pending timer or not. | |
788 | * (ie. mod_timer() of an inactive timer returns 0, mod_timer() of an | |
789 | * active timer returns 1.) | |
790 | */ | |
791 | int mod_timer(struct timer_list *timer, unsigned long expires) | |
792 | { | |
1da177e4 LT |
793 | /* |
794 | * This is a common optimization triggered by the | |
795 | * networking code - if the timer is re-modified | |
796 | * to be the same thing then just return: | |
797 | */ | |
4841158b | 798 | if (timer_pending(timer) && timer->expires == expires) |
1da177e4 LT |
799 | return 1; |
800 | ||
3bbb9ec9 AV |
801 | expires = apply_slack(timer, expires); |
802 | ||
597d0275 | 803 | return __mod_timer(timer, expires, false, TIMER_NOT_PINNED); |
1da177e4 | 804 | } |
1da177e4 LT |
805 | EXPORT_SYMBOL(mod_timer); |
806 | ||
597d0275 AB |
807 | /** |
808 | * mod_timer_pinned - modify a timer's timeout | |
809 | * @timer: the timer to be modified | |
810 | * @expires: new timeout in jiffies | |
811 | * | |
812 | * mod_timer_pinned() is a way to update the expire field of an | |
813 | * active timer (if the timer is inactive it will be activated) | |
814 | * and not allow the timer to be migrated to a different CPU. | |
815 | * | |
816 | * mod_timer_pinned(timer, expires) is equivalent to: | |
817 | * | |
818 | * del_timer(timer); timer->expires = expires; add_timer(timer); | |
819 | */ | |
820 | int mod_timer_pinned(struct timer_list *timer, unsigned long expires) | |
821 | { | |
822 | if (timer->expires == expires && timer_pending(timer)) | |
823 | return 1; | |
824 | ||
825 | return __mod_timer(timer, expires, false, TIMER_PINNED); | |
826 | } | |
827 | EXPORT_SYMBOL(mod_timer_pinned); | |
828 | ||
74019224 IM |
829 | /** |
830 | * add_timer - start a timer | |
831 | * @timer: the timer to be added | |
832 | * | |
833 | * The kernel will do a ->function(->data) callback from the | |
834 | * timer interrupt at the ->expires point in the future. The | |
835 | * current time is 'jiffies'. | |
836 | * | |
837 | * The timer's ->expires, ->function (and if the handler uses it, ->data) | |
838 | * fields must be set prior calling this function. | |
839 | * | |
840 | * Timers with an ->expires field in the past will be executed in the next | |
841 | * timer tick. | |
842 | */ | |
843 | void add_timer(struct timer_list *timer) | |
844 | { | |
845 | BUG_ON(timer_pending(timer)); | |
846 | mod_timer(timer, timer->expires); | |
847 | } | |
848 | EXPORT_SYMBOL(add_timer); | |
849 | ||
850 | /** | |
851 | * add_timer_on - start a timer on a particular CPU | |
852 | * @timer: the timer to be added | |
853 | * @cpu: the CPU to start it on | |
854 | * | |
855 | * This is not very scalable on SMP. Double adds are not possible. | |
856 | */ | |
857 | void add_timer_on(struct timer_list *timer, int cpu) | |
858 | { | |
859 | struct tvec_base *base = per_cpu(tvec_bases, cpu); | |
860 | unsigned long flags; | |
861 | ||
862 | timer_stats_timer_set_start_info(timer); | |
863 | BUG_ON(timer_pending(timer) || !timer->function); | |
864 | spin_lock_irqsave(&base->lock, flags); | |
865 | timer_set_base(timer, base); | |
2b022e3d | 866 | debug_activate(timer, timer->expires); |
97fd9ed4 MS |
867 | if (time_before(timer->expires, base->next_timer) && |
868 | !tbase_get_deferrable(timer->base)) | |
869 | base->next_timer = timer->expires; | |
74019224 IM |
870 | internal_add_timer(base, timer); |
871 | /* | |
872 | * Check whether the other CPU is idle and needs to be | |
873 | * triggered to reevaluate the timer wheel when nohz is | |
874 | * active. We are protected against the other CPU fiddling | |
875 | * with the timer by holding the timer base lock. This also | |
876 | * makes sure that a CPU on the way to idle can not evaluate | |
877 | * the timer wheel. | |
878 | */ | |
879 | wake_up_idle_cpu(cpu); | |
880 | spin_unlock_irqrestore(&base->lock, flags); | |
881 | } | |
a9862e05 | 882 | EXPORT_SYMBOL_GPL(add_timer_on); |
74019224 | 883 | |
2aae4a10 | 884 | /** |
1da177e4 LT |
885 | * del_timer - deactive a timer. |
886 | * @timer: the timer to be deactivated | |
887 | * | |
888 | * del_timer() deactivates a timer - this works on both active and inactive | |
889 | * timers. | |
890 | * | |
891 | * The function returns whether it has deactivated a pending timer or not. | |
892 | * (ie. del_timer() of an inactive timer returns 0, del_timer() of an | |
893 | * active timer returns 1.) | |
894 | */ | |
895 | int del_timer(struct timer_list *timer) | |
896 | { | |
a6fa8e5a | 897 | struct tvec_base *base; |
1da177e4 | 898 | unsigned long flags; |
55c888d6 | 899 | int ret = 0; |
1da177e4 | 900 | |
82f67cd9 | 901 | timer_stats_timer_clear_start_info(timer); |
55c888d6 ON |
902 | if (timer_pending(timer)) { |
903 | base = lock_timer_base(timer, &flags); | |
904 | if (timer_pending(timer)) { | |
905 | detach_timer(timer, 1); | |
97fd9ed4 MS |
906 | if (timer->expires == base->next_timer && |
907 | !tbase_get_deferrable(timer->base)) | |
908 | base->next_timer = base->timer_jiffies; | |
55c888d6 ON |
909 | ret = 1; |
910 | } | |
1da177e4 | 911 | spin_unlock_irqrestore(&base->lock, flags); |
1da177e4 | 912 | } |
1da177e4 | 913 | |
55c888d6 | 914 | return ret; |
1da177e4 | 915 | } |
1da177e4 LT |
916 | EXPORT_SYMBOL(del_timer); |
917 | ||
918 | #ifdef CONFIG_SMP | |
2aae4a10 REB |
919 | /** |
920 | * try_to_del_timer_sync - Try to deactivate a timer | |
921 | * @timer: timer do del | |
922 | * | |
fd450b73 ON |
923 | * This function tries to deactivate a timer. Upon successful (ret >= 0) |
924 | * exit the timer is not queued and the handler is not running on any CPU. | |
925 | * | |
926 | * It must not be called from interrupt contexts. | |
927 | */ | |
928 | int try_to_del_timer_sync(struct timer_list *timer) | |
929 | { | |
a6fa8e5a | 930 | struct tvec_base *base; |
fd450b73 ON |
931 | unsigned long flags; |
932 | int ret = -1; | |
933 | ||
934 | base = lock_timer_base(timer, &flags); | |
935 | ||
936 | if (base->running_timer == timer) | |
937 | goto out; | |
938 | ||
939 | ret = 0; | |
940 | if (timer_pending(timer)) { | |
941 | detach_timer(timer, 1); | |
97fd9ed4 MS |
942 | if (timer->expires == base->next_timer && |
943 | !tbase_get_deferrable(timer->base)) | |
944 | base->next_timer = base->timer_jiffies; | |
fd450b73 ON |
945 | ret = 1; |
946 | } | |
947 | out: | |
948 | spin_unlock_irqrestore(&base->lock, flags); | |
949 | ||
950 | return ret; | |
951 | } | |
e19dff1f DH |
952 | EXPORT_SYMBOL(try_to_del_timer_sync); |
953 | ||
2aae4a10 | 954 | /** |
1da177e4 LT |
955 | * del_timer_sync - deactivate a timer and wait for the handler to finish. |
956 | * @timer: the timer to be deactivated | |
957 | * | |
958 | * This function only differs from del_timer() on SMP: besides deactivating | |
959 | * the timer it also makes sure the handler has finished executing on other | |
960 | * CPUs. | |
961 | * | |
72fd4a35 | 962 | * Synchronization rules: Callers must prevent restarting of the timer, |
1da177e4 LT |
963 | * otherwise this function is meaningless. It must not be called from |
964 | * interrupt contexts. The caller must not hold locks which would prevent | |
55c888d6 ON |
965 | * completion of the timer's handler. The timer's handler must not call |
966 | * add_timer_on(). Upon exit the timer is not queued and the handler is | |
967 | * not running on any CPU. | |
1da177e4 LT |
968 | * |
969 | * The function returns whether it has deactivated a pending timer or not. | |
1da177e4 LT |
970 | */ |
971 | int del_timer_sync(struct timer_list *timer) | |
972 | { | |
6f2b9b9a JB |
973 | #ifdef CONFIG_LOCKDEP |
974 | unsigned long flags; | |
975 | ||
976 | local_irq_save(flags); | |
977 | lock_map_acquire(&timer->lockdep_map); | |
978 | lock_map_release(&timer->lockdep_map); | |
979 | local_irq_restore(flags); | |
980 | #endif | |
981 | ||
fd450b73 ON |
982 | for (;;) { |
983 | int ret = try_to_del_timer_sync(timer); | |
984 | if (ret >= 0) | |
985 | return ret; | |
a0009652 | 986 | cpu_relax(); |
fd450b73 | 987 | } |
1da177e4 | 988 | } |
55c888d6 | 989 | EXPORT_SYMBOL(del_timer_sync); |
1da177e4 LT |
990 | #endif |
991 | ||
a6fa8e5a | 992 | static int cascade(struct tvec_base *base, struct tvec *tv, int index) |
1da177e4 LT |
993 | { |
994 | /* cascade all the timers from tv up one level */ | |
3439dd86 P |
995 | struct timer_list *timer, *tmp; |
996 | struct list_head tv_list; | |
997 | ||
998 | list_replace_init(tv->vec + index, &tv_list); | |
1da177e4 | 999 | |
1da177e4 | 1000 | /* |
3439dd86 P |
1001 | * We are removing _all_ timers from the list, so we |
1002 | * don't have to detach them individually. | |
1da177e4 | 1003 | */ |
3439dd86 | 1004 | list_for_each_entry_safe(timer, tmp, &tv_list, entry) { |
6e453a67 | 1005 | BUG_ON(tbase_get_base(timer->base) != base); |
3439dd86 | 1006 | internal_add_timer(base, timer); |
1da177e4 | 1007 | } |
1da177e4 LT |
1008 | |
1009 | return index; | |
1010 | } | |
1011 | ||
576da126 TG |
1012 | static void call_timer_fn(struct timer_list *timer, void (*fn)(unsigned long), |
1013 | unsigned long data) | |
1014 | { | |
1015 | int preempt_count = preempt_count(); | |
1016 | ||
1017 | #ifdef CONFIG_LOCKDEP | |
1018 | /* | |
1019 | * It is permissible to free the timer from inside the | |
1020 | * function that is called from it, this we need to take into | |
1021 | * account for lockdep too. To avoid bogus "held lock freed" | |
1022 | * warnings as well as problems when looking into | |
1023 | * timer->lockdep_map, make a copy and use that here. | |
1024 | */ | |
1025 | struct lockdep_map lockdep_map = timer->lockdep_map; | |
1026 | #endif | |
1027 | /* | |
1028 | * Couple the lock chain with the lock chain at | |
1029 | * del_timer_sync() by acquiring the lock_map around the fn() | |
1030 | * call here and in del_timer_sync(). | |
1031 | */ | |
1032 | lock_map_acquire(&lockdep_map); | |
1033 | ||
1034 | trace_timer_expire_entry(timer); | |
1035 | fn(data); | |
1036 | trace_timer_expire_exit(timer); | |
1037 | ||
1038 | lock_map_release(&lockdep_map); | |
1039 | ||
1040 | if (preempt_count != preempt_count()) { | |
802702e0 TG |
1041 | WARN_ONCE(1, "timer: %pF preempt leak: %08x -> %08x\n", |
1042 | fn, preempt_count, preempt_count()); | |
1043 | /* | |
1044 | * Restore the preempt count. That gives us a decent | |
1045 | * chance to survive and extract information. If the | |
1046 | * callback kept a lock held, bad luck, but not worse | |
1047 | * than the BUG() we had. | |
1048 | */ | |
1049 | preempt_count() = preempt_count; | |
576da126 TG |
1050 | } |
1051 | } | |
1052 | ||
2aae4a10 REB |
1053 | #define INDEX(N) ((base->timer_jiffies >> (TVR_BITS + (N) * TVN_BITS)) & TVN_MASK) |
1054 | ||
1055 | /** | |
1da177e4 LT |
1056 | * __run_timers - run all expired timers (if any) on this CPU. |
1057 | * @base: the timer vector to be processed. | |
1058 | * | |
1059 | * This function cascades all vectors and executes all expired timer | |
1060 | * vectors. | |
1061 | */ | |
a6fa8e5a | 1062 | static inline void __run_timers(struct tvec_base *base) |
1da177e4 LT |
1063 | { |
1064 | struct timer_list *timer; | |
1065 | ||
3691c519 | 1066 | spin_lock_irq(&base->lock); |
1da177e4 | 1067 | while (time_after_eq(jiffies, base->timer_jiffies)) { |
626ab0e6 | 1068 | struct list_head work_list; |
1da177e4 | 1069 | struct list_head *head = &work_list; |
6819457d | 1070 | int index = base->timer_jiffies & TVR_MASK; |
626ab0e6 | 1071 | |
1da177e4 LT |
1072 | /* |
1073 | * Cascade timers: | |
1074 | */ | |
1075 | if (!index && | |
1076 | (!cascade(base, &base->tv2, INDEX(0))) && | |
1077 | (!cascade(base, &base->tv3, INDEX(1))) && | |
1078 | !cascade(base, &base->tv4, INDEX(2))) | |
1079 | cascade(base, &base->tv5, INDEX(3)); | |
626ab0e6 ON |
1080 | ++base->timer_jiffies; |
1081 | list_replace_init(base->tv1.vec + index, &work_list); | |
55c888d6 | 1082 | while (!list_empty(head)) { |
1da177e4 LT |
1083 | void (*fn)(unsigned long); |
1084 | unsigned long data; | |
1085 | ||
b5e61818 | 1086 | timer = list_first_entry(head, struct timer_list,entry); |
6819457d TG |
1087 | fn = timer->function; |
1088 | data = timer->data; | |
1da177e4 | 1089 | |
82f67cd9 IM |
1090 | timer_stats_account_timer(timer); |
1091 | ||
1da177e4 | 1092 | set_running_timer(base, timer); |
55c888d6 | 1093 | detach_timer(timer, 1); |
6f2b9b9a | 1094 | |
3691c519 | 1095 | spin_unlock_irq(&base->lock); |
576da126 | 1096 | call_timer_fn(timer, fn, data); |
3691c519 | 1097 | spin_lock_irq(&base->lock); |
1da177e4 LT |
1098 | } |
1099 | } | |
1100 | set_running_timer(base, NULL); | |
3691c519 | 1101 | spin_unlock_irq(&base->lock); |
1da177e4 LT |
1102 | } |
1103 | ||
ee9c5785 | 1104 | #ifdef CONFIG_NO_HZ |
1da177e4 LT |
1105 | /* |
1106 | * Find out when the next timer event is due to happen. This | |
90cba64a RD |
1107 | * is used on S/390 to stop all activity when a CPU is idle. |
1108 | * This function needs to be called with interrupts disabled. | |
1da177e4 | 1109 | */ |
a6fa8e5a | 1110 | static unsigned long __next_timer_interrupt(struct tvec_base *base) |
1da177e4 | 1111 | { |
1cfd6849 | 1112 | unsigned long timer_jiffies = base->timer_jiffies; |
eaad084b | 1113 | unsigned long expires = timer_jiffies + NEXT_TIMER_MAX_DELTA; |
1cfd6849 | 1114 | int index, slot, array, found = 0; |
1da177e4 | 1115 | struct timer_list *nte; |
a6fa8e5a | 1116 | struct tvec *varray[4]; |
1da177e4 LT |
1117 | |
1118 | /* Look for timer events in tv1. */ | |
1cfd6849 | 1119 | index = slot = timer_jiffies & TVR_MASK; |
1da177e4 | 1120 | do { |
1cfd6849 | 1121 | list_for_each_entry(nte, base->tv1.vec + slot, entry) { |
6819457d TG |
1122 | if (tbase_get_deferrable(nte->base)) |
1123 | continue; | |
6e453a67 | 1124 | |
1cfd6849 | 1125 | found = 1; |
1da177e4 | 1126 | expires = nte->expires; |
1cfd6849 TG |
1127 | /* Look at the cascade bucket(s)? */ |
1128 | if (!index || slot < index) | |
1129 | goto cascade; | |
1130 | return expires; | |
1da177e4 | 1131 | } |
1cfd6849 TG |
1132 | slot = (slot + 1) & TVR_MASK; |
1133 | } while (slot != index); | |
1134 | ||
1135 | cascade: | |
1136 | /* Calculate the next cascade event */ | |
1137 | if (index) | |
1138 | timer_jiffies += TVR_SIZE - index; | |
1139 | timer_jiffies >>= TVR_BITS; | |
1da177e4 LT |
1140 | |
1141 | /* Check tv2-tv5. */ | |
1142 | varray[0] = &base->tv2; | |
1143 | varray[1] = &base->tv3; | |
1144 | varray[2] = &base->tv4; | |
1145 | varray[3] = &base->tv5; | |
1cfd6849 TG |
1146 | |
1147 | for (array = 0; array < 4; array++) { | |
a6fa8e5a | 1148 | struct tvec *varp = varray[array]; |
1cfd6849 TG |
1149 | |
1150 | index = slot = timer_jiffies & TVN_MASK; | |
1da177e4 | 1151 | do { |
1cfd6849 | 1152 | list_for_each_entry(nte, varp->vec + slot, entry) { |
a0419888 JH |
1153 | if (tbase_get_deferrable(nte->base)) |
1154 | continue; | |
1155 | ||
1cfd6849 | 1156 | found = 1; |
1da177e4 LT |
1157 | if (time_before(nte->expires, expires)) |
1158 | expires = nte->expires; | |
1cfd6849 TG |
1159 | } |
1160 | /* | |
1161 | * Do we still search for the first timer or are | |
1162 | * we looking up the cascade buckets ? | |
1163 | */ | |
1164 | if (found) { | |
1165 | /* Look at the cascade bucket(s)? */ | |
1166 | if (!index || slot < index) | |
1167 | break; | |
1168 | return expires; | |
1169 | } | |
1170 | slot = (slot + 1) & TVN_MASK; | |
1171 | } while (slot != index); | |
1172 | ||
1173 | if (index) | |
1174 | timer_jiffies += TVN_SIZE - index; | |
1175 | timer_jiffies >>= TVN_BITS; | |
1da177e4 | 1176 | } |
1cfd6849 TG |
1177 | return expires; |
1178 | } | |
69239749 | 1179 | |
1cfd6849 TG |
1180 | /* |
1181 | * Check, if the next hrtimer event is before the next timer wheel | |
1182 | * event: | |
1183 | */ | |
1184 | static unsigned long cmp_next_hrtimer_event(unsigned long now, | |
1185 | unsigned long expires) | |
1186 | { | |
1187 | ktime_t hr_delta = hrtimer_get_next_event(); | |
1188 | struct timespec tsdelta; | |
9501b6cf | 1189 | unsigned long delta; |
1cfd6849 TG |
1190 | |
1191 | if (hr_delta.tv64 == KTIME_MAX) | |
1192 | return expires; | |
0662b713 | 1193 | |
9501b6cf TG |
1194 | /* |
1195 | * Expired timer available, let it expire in the next tick | |
1196 | */ | |
1197 | if (hr_delta.tv64 <= 0) | |
1198 | return now + 1; | |
69239749 | 1199 | |
1cfd6849 | 1200 | tsdelta = ktime_to_timespec(hr_delta); |
9501b6cf | 1201 | delta = timespec_to_jiffies(&tsdelta); |
eaad084b TG |
1202 | |
1203 | /* | |
1204 | * Limit the delta to the max value, which is checked in | |
1205 | * tick_nohz_stop_sched_tick(): | |
1206 | */ | |
1207 | if (delta > NEXT_TIMER_MAX_DELTA) | |
1208 | delta = NEXT_TIMER_MAX_DELTA; | |
1209 | ||
9501b6cf TG |
1210 | /* |
1211 | * Take rounding errors in to account and make sure, that it | |
1212 | * expires in the next tick. Otherwise we go into an endless | |
1213 | * ping pong due to tick_nohz_stop_sched_tick() retriggering | |
1214 | * the timer softirq | |
1215 | */ | |
1216 | if (delta < 1) | |
1217 | delta = 1; | |
1218 | now += delta; | |
1cfd6849 TG |
1219 | if (time_before(now, expires)) |
1220 | return now; | |
1da177e4 LT |
1221 | return expires; |
1222 | } | |
1cfd6849 TG |
1223 | |
1224 | /** | |
8dce39c2 | 1225 | * get_next_timer_interrupt - return the jiffy of the next pending timer |
05fb6bf0 | 1226 | * @now: current time (in jiffies) |
1cfd6849 | 1227 | */ |
fd064b9b | 1228 | unsigned long get_next_timer_interrupt(unsigned long now) |
1cfd6849 | 1229 | { |
a6fa8e5a | 1230 | struct tvec_base *base = __get_cpu_var(tvec_bases); |
fd064b9b | 1231 | unsigned long expires; |
1cfd6849 TG |
1232 | |
1233 | spin_lock(&base->lock); | |
97fd9ed4 MS |
1234 | if (time_before_eq(base->next_timer, base->timer_jiffies)) |
1235 | base->next_timer = __next_timer_interrupt(base); | |
1236 | expires = base->next_timer; | |
1cfd6849 TG |
1237 | spin_unlock(&base->lock); |
1238 | ||
1239 | if (time_before_eq(expires, now)) | |
1240 | return now; | |
1241 | ||
1242 | return cmp_next_hrtimer_event(now, expires); | |
1243 | } | |
1da177e4 LT |
1244 | #endif |
1245 | ||
1da177e4 | 1246 | /* |
5b4db0c2 | 1247 | * Called from the timer interrupt handler to charge one tick to the current |
1da177e4 LT |
1248 | * process. user_tick is 1 if the tick is user time, 0 for system. |
1249 | */ | |
1250 | void update_process_times(int user_tick) | |
1251 | { | |
1252 | struct task_struct *p = current; | |
1253 | int cpu = smp_processor_id(); | |
1254 | ||
1255 | /* Note: this timer irq context must be accounted for as well. */ | |
fa13a5a1 | 1256 | account_process_tick(p, user_tick); |
1da177e4 | 1257 | run_local_timers(); |
a157229c | 1258 | rcu_check_callbacks(cpu, user_tick); |
b845b517 | 1259 | printk_tick(); |
fe432200 | 1260 | perf_event_do_pending(); |
1da177e4 | 1261 | scheduler_tick(); |
6819457d | 1262 | run_posix_cpu_timers(p); |
1da177e4 LT |
1263 | } |
1264 | ||
1da177e4 LT |
1265 | /* |
1266 | * This function runs timers and the timer-tq in bottom half context. | |
1267 | */ | |
1268 | static void run_timer_softirq(struct softirq_action *h) | |
1269 | { | |
a6fa8e5a | 1270 | struct tvec_base *base = __get_cpu_var(tvec_bases); |
1da177e4 | 1271 | |
d3d74453 | 1272 | hrtimer_run_pending(); |
82f67cd9 | 1273 | |
1da177e4 LT |
1274 | if (time_after_eq(jiffies, base->timer_jiffies)) |
1275 | __run_timers(base); | |
1276 | } | |
1277 | ||
1278 | /* | |
1279 | * Called by the local, per-CPU timer interrupt on SMP. | |
1280 | */ | |
1281 | void run_local_timers(void) | |
1282 | { | |
d3d74453 | 1283 | hrtimer_run_queues(); |
1da177e4 | 1284 | raise_softirq(TIMER_SOFTIRQ); |
6687a97d | 1285 | softlockup_tick(); |
1da177e4 LT |
1286 | } |
1287 | ||
1da177e4 LT |
1288 | /* |
1289 | * The 64-bit jiffies value is not atomic - you MUST NOT read it | |
1290 | * without sampling the sequence number in xtime_lock. | |
1291 | * jiffies is defined in the linker script... | |
1292 | */ | |
1293 | ||
3171a030 | 1294 | void do_timer(unsigned long ticks) |
1da177e4 | 1295 | { |
3171a030 | 1296 | jiffies_64 += ticks; |
dce48a84 TG |
1297 | update_wall_time(); |
1298 | calc_global_load(); | |
1da177e4 LT |
1299 | } |
1300 | ||
1301 | #ifdef __ARCH_WANT_SYS_ALARM | |
1302 | ||
1303 | /* | |
1304 | * For backwards compatibility? This can be done in libc so Alpha | |
1305 | * and all newer ports shouldn't need it. | |
1306 | */ | |
58fd3aa2 | 1307 | SYSCALL_DEFINE1(alarm, unsigned int, seconds) |
1da177e4 | 1308 | { |
c08b8a49 | 1309 | return alarm_setitimer(seconds); |
1da177e4 LT |
1310 | } |
1311 | ||
1312 | #endif | |
1313 | ||
1314 | #ifndef __alpha__ | |
1315 | ||
1316 | /* | |
1317 | * The Alpha uses getxpid, getxuid, and getxgid instead. Maybe this | |
1318 | * should be moved into arch/i386 instead? | |
1319 | */ | |
1320 | ||
1321 | /** | |
1322 | * sys_getpid - return the thread group id of the current process | |
1323 | * | |
1324 | * Note, despite the name, this returns the tgid not the pid. The tgid and | |
1325 | * the pid are identical unless CLONE_THREAD was specified on clone() in | |
1326 | * which case the tgid is the same in all threads of the same group. | |
1327 | * | |
1328 | * This is SMP safe as current->tgid does not change. | |
1329 | */ | |
58fd3aa2 | 1330 | SYSCALL_DEFINE0(getpid) |
1da177e4 | 1331 | { |
b488893a | 1332 | return task_tgid_vnr(current); |
1da177e4 LT |
1333 | } |
1334 | ||
1335 | /* | |
6997a6fa KK |
1336 | * Accessing ->real_parent is not SMP-safe, it could |
1337 | * change from under us. However, we can use a stale | |
1338 | * value of ->real_parent under rcu_read_lock(), see | |
1339 | * release_task()->call_rcu(delayed_put_task_struct). | |
1da177e4 | 1340 | */ |
dbf040d9 | 1341 | SYSCALL_DEFINE0(getppid) |
1da177e4 LT |
1342 | { |
1343 | int pid; | |
1da177e4 | 1344 | |
6997a6fa | 1345 | rcu_read_lock(); |
6c5f3e7b | 1346 | pid = task_tgid_vnr(current->real_parent); |
6997a6fa | 1347 | rcu_read_unlock(); |
1da177e4 | 1348 | |
1da177e4 LT |
1349 | return pid; |
1350 | } | |
1351 | ||
dbf040d9 | 1352 | SYSCALL_DEFINE0(getuid) |
1da177e4 LT |
1353 | { |
1354 | /* Only we change this so SMP safe */ | |
76aac0e9 | 1355 | return current_uid(); |
1da177e4 LT |
1356 | } |
1357 | ||
dbf040d9 | 1358 | SYSCALL_DEFINE0(geteuid) |
1da177e4 LT |
1359 | { |
1360 | /* Only we change this so SMP safe */ | |
76aac0e9 | 1361 | return current_euid(); |
1da177e4 LT |
1362 | } |
1363 | ||
dbf040d9 | 1364 | SYSCALL_DEFINE0(getgid) |
1da177e4 LT |
1365 | { |
1366 | /* Only we change this so SMP safe */ | |
76aac0e9 | 1367 | return current_gid(); |
1da177e4 LT |
1368 | } |
1369 | ||
dbf040d9 | 1370 | SYSCALL_DEFINE0(getegid) |
1da177e4 LT |
1371 | { |
1372 | /* Only we change this so SMP safe */ | |
76aac0e9 | 1373 | return current_egid(); |
1da177e4 LT |
1374 | } |
1375 | ||
1376 | #endif | |
1377 | ||
1378 | static void process_timeout(unsigned long __data) | |
1379 | { | |
36c8b586 | 1380 | wake_up_process((struct task_struct *)__data); |
1da177e4 LT |
1381 | } |
1382 | ||
1383 | /** | |
1384 | * schedule_timeout - sleep until timeout | |
1385 | * @timeout: timeout value in jiffies | |
1386 | * | |
1387 | * Make the current task sleep until @timeout jiffies have | |
1388 | * elapsed. The routine will return immediately unless | |
1389 | * the current task state has been set (see set_current_state()). | |
1390 | * | |
1391 | * You can set the task state as follows - | |
1392 | * | |
1393 | * %TASK_UNINTERRUPTIBLE - at least @timeout jiffies are guaranteed to | |
1394 | * pass before the routine returns. The routine will return 0 | |
1395 | * | |
1396 | * %TASK_INTERRUPTIBLE - the routine may return early if a signal is | |
1397 | * delivered to the current task. In this case the remaining time | |
1398 | * in jiffies will be returned, or 0 if the timer expired in time | |
1399 | * | |
1400 | * The current task state is guaranteed to be TASK_RUNNING when this | |
1401 | * routine returns. | |
1402 | * | |
1403 | * Specifying a @timeout value of %MAX_SCHEDULE_TIMEOUT will schedule | |
1404 | * the CPU away without a bound on the timeout. In this case the return | |
1405 | * value will be %MAX_SCHEDULE_TIMEOUT. | |
1406 | * | |
1407 | * In all cases the return value is guaranteed to be non-negative. | |
1408 | */ | |
7ad5b3a5 | 1409 | signed long __sched schedule_timeout(signed long timeout) |
1da177e4 LT |
1410 | { |
1411 | struct timer_list timer; | |
1412 | unsigned long expire; | |
1413 | ||
1414 | switch (timeout) | |
1415 | { | |
1416 | case MAX_SCHEDULE_TIMEOUT: | |
1417 | /* | |
1418 | * These two special cases are useful to be comfortable | |
1419 | * in the caller. Nothing more. We could take | |
1420 | * MAX_SCHEDULE_TIMEOUT from one of the negative value | |
1421 | * but I' d like to return a valid offset (>=0) to allow | |
1422 | * the caller to do everything it want with the retval. | |
1423 | */ | |
1424 | schedule(); | |
1425 | goto out; | |
1426 | default: | |
1427 | /* | |
1428 | * Another bit of PARANOID. Note that the retval will be | |
1429 | * 0 since no piece of kernel is supposed to do a check | |
1430 | * for a negative retval of schedule_timeout() (since it | |
1431 | * should never happens anyway). You just have the printk() | |
1432 | * that will tell you if something is gone wrong and where. | |
1433 | */ | |
5b149bcc | 1434 | if (timeout < 0) { |
1da177e4 | 1435 | printk(KERN_ERR "schedule_timeout: wrong timeout " |
5b149bcc AM |
1436 | "value %lx\n", timeout); |
1437 | dump_stack(); | |
1da177e4 LT |
1438 | current->state = TASK_RUNNING; |
1439 | goto out; | |
1440 | } | |
1441 | } | |
1442 | ||
1443 | expire = timeout + jiffies; | |
1444 | ||
c6f3a97f | 1445 | setup_timer_on_stack(&timer, process_timeout, (unsigned long)current); |
597d0275 | 1446 | __mod_timer(&timer, expire, false, TIMER_NOT_PINNED); |
1da177e4 LT |
1447 | schedule(); |
1448 | del_singleshot_timer_sync(&timer); | |
1449 | ||
c6f3a97f TG |
1450 | /* Remove the timer from the object tracker */ |
1451 | destroy_timer_on_stack(&timer); | |
1452 | ||
1da177e4 LT |
1453 | timeout = expire - jiffies; |
1454 | ||
1455 | out: | |
1456 | return timeout < 0 ? 0 : timeout; | |
1457 | } | |
1da177e4 LT |
1458 | EXPORT_SYMBOL(schedule_timeout); |
1459 | ||
8a1c1757 AM |
1460 | /* |
1461 | * We can use __set_current_state() here because schedule_timeout() calls | |
1462 | * schedule() unconditionally. | |
1463 | */ | |
64ed93a2 NA |
1464 | signed long __sched schedule_timeout_interruptible(signed long timeout) |
1465 | { | |
a5a0d52c AM |
1466 | __set_current_state(TASK_INTERRUPTIBLE); |
1467 | return schedule_timeout(timeout); | |
64ed93a2 NA |
1468 | } |
1469 | EXPORT_SYMBOL(schedule_timeout_interruptible); | |
1470 | ||
294d5cc2 MW |
1471 | signed long __sched schedule_timeout_killable(signed long timeout) |
1472 | { | |
1473 | __set_current_state(TASK_KILLABLE); | |
1474 | return schedule_timeout(timeout); | |
1475 | } | |
1476 | EXPORT_SYMBOL(schedule_timeout_killable); | |
1477 | ||
64ed93a2 NA |
1478 | signed long __sched schedule_timeout_uninterruptible(signed long timeout) |
1479 | { | |
a5a0d52c AM |
1480 | __set_current_state(TASK_UNINTERRUPTIBLE); |
1481 | return schedule_timeout(timeout); | |
64ed93a2 NA |
1482 | } |
1483 | EXPORT_SYMBOL(schedule_timeout_uninterruptible); | |
1484 | ||
1da177e4 | 1485 | /* Thread ID - the internal kernel "pid" */ |
58fd3aa2 | 1486 | SYSCALL_DEFINE0(gettid) |
1da177e4 | 1487 | { |
b488893a | 1488 | return task_pid_vnr(current); |
1da177e4 LT |
1489 | } |
1490 | ||
2aae4a10 | 1491 | /** |
d4d23add | 1492 | * do_sysinfo - fill in sysinfo struct |
2aae4a10 | 1493 | * @info: pointer to buffer to fill |
6819457d | 1494 | */ |
d4d23add | 1495 | int do_sysinfo(struct sysinfo *info) |
1da177e4 | 1496 | { |
1da177e4 LT |
1497 | unsigned long mem_total, sav_total; |
1498 | unsigned int mem_unit, bitcount; | |
2d02494f | 1499 | struct timespec tp; |
1da177e4 | 1500 | |
d4d23add | 1501 | memset(info, 0, sizeof(struct sysinfo)); |
1da177e4 | 1502 | |
2d02494f TG |
1503 | ktime_get_ts(&tp); |
1504 | monotonic_to_bootbased(&tp); | |
1505 | info->uptime = tp.tv_sec + (tp.tv_nsec ? 1 : 0); | |
1da177e4 | 1506 | |
2d02494f | 1507 | get_avenrun(info->loads, 0, SI_LOAD_SHIFT - FSHIFT); |
1da177e4 | 1508 | |
2d02494f | 1509 | info->procs = nr_threads; |
1da177e4 | 1510 | |
d4d23add KM |
1511 | si_meminfo(info); |
1512 | si_swapinfo(info); | |
1da177e4 LT |
1513 | |
1514 | /* | |
1515 | * If the sum of all the available memory (i.e. ram + swap) | |
1516 | * is less than can be stored in a 32 bit unsigned long then | |
1517 | * we can be binary compatible with 2.2.x kernels. If not, | |
1518 | * well, in that case 2.2.x was broken anyways... | |
1519 | * | |
1520 | * -Erik Andersen <andersee@debian.org> | |
1521 | */ | |
1522 | ||
d4d23add KM |
1523 | mem_total = info->totalram + info->totalswap; |
1524 | if (mem_total < info->totalram || mem_total < info->totalswap) | |
1da177e4 LT |
1525 | goto out; |
1526 | bitcount = 0; | |
d4d23add | 1527 | mem_unit = info->mem_unit; |
1da177e4 LT |
1528 | while (mem_unit > 1) { |
1529 | bitcount++; | |
1530 | mem_unit >>= 1; | |
1531 | sav_total = mem_total; | |
1532 | mem_total <<= 1; | |
1533 | if (mem_total < sav_total) | |
1534 | goto out; | |
1535 | } | |
1536 | ||
1537 | /* | |
1538 | * If mem_total did not overflow, multiply all memory values by | |
d4d23add | 1539 | * info->mem_unit and set it to 1. This leaves things compatible |
1da177e4 LT |
1540 | * with 2.2.x, and also retains compatibility with earlier 2.4.x |
1541 | * kernels... | |
1542 | */ | |
1543 | ||
d4d23add KM |
1544 | info->mem_unit = 1; |
1545 | info->totalram <<= bitcount; | |
1546 | info->freeram <<= bitcount; | |
1547 | info->sharedram <<= bitcount; | |
1548 | info->bufferram <<= bitcount; | |
1549 | info->totalswap <<= bitcount; | |
1550 | info->freeswap <<= bitcount; | |
1551 | info->totalhigh <<= bitcount; | |
1552 | info->freehigh <<= bitcount; | |
1553 | ||
1554 | out: | |
1555 | return 0; | |
1556 | } | |
1557 | ||
1e7bfb21 | 1558 | SYSCALL_DEFINE1(sysinfo, struct sysinfo __user *, info) |
d4d23add KM |
1559 | { |
1560 | struct sysinfo val; | |
1561 | ||
1562 | do_sysinfo(&val); | |
1da177e4 | 1563 | |
1da177e4 LT |
1564 | if (copy_to_user(info, &val, sizeof(struct sysinfo))) |
1565 | return -EFAULT; | |
1566 | ||
1567 | return 0; | |
1568 | } | |
1569 | ||
b4be6258 | 1570 | static int __cpuinit init_timers_cpu(int cpu) |
1da177e4 LT |
1571 | { |
1572 | int j; | |
a6fa8e5a | 1573 | struct tvec_base *base; |
b4be6258 | 1574 | static char __cpuinitdata tvec_base_done[NR_CPUS]; |
55c888d6 | 1575 | |
ba6edfcd | 1576 | if (!tvec_base_done[cpu]) { |
a4a6198b JB |
1577 | static char boot_done; |
1578 | ||
a4a6198b | 1579 | if (boot_done) { |
ba6edfcd AM |
1580 | /* |
1581 | * The APs use this path later in boot | |
1582 | */ | |
94f6030c CL |
1583 | base = kmalloc_node(sizeof(*base), |
1584 | GFP_KERNEL | __GFP_ZERO, | |
a4a6198b JB |
1585 | cpu_to_node(cpu)); |
1586 | if (!base) | |
1587 | return -ENOMEM; | |
6e453a67 VP |
1588 | |
1589 | /* Make sure that tvec_base is 2 byte aligned */ | |
1590 | if (tbase_get_deferrable(base)) { | |
1591 | WARN_ON(1); | |
1592 | kfree(base); | |
1593 | return -ENOMEM; | |
1594 | } | |
ba6edfcd | 1595 | per_cpu(tvec_bases, cpu) = base; |
a4a6198b | 1596 | } else { |
ba6edfcd AM |
1597 | /* |
1598 | * This is for the boot CPU - we use compile-time | |
1599 | * static initialisation because per-cpu memory isn't | |
1600 | * ready yet and because the memory allocators are not | |
1601 | * initialised either. | |
1602 | */ | |
a4a6198b | 1603 | boot_done = 1; |
ba6edfcd | 1604 | base = &boot_tvec_bases; |
a4a6198b | 1605 | } |
ba6edfcd AM |
1606 | tvec_base_done[cpu] = 1; |
1607 | } else { | |
1608 | base = per_cpu(tvec_bases, cpu); | |
a4a6198b | 1609 | } |
ba6edfcd | 1610 | |
3691c519 | 1611 | spin_lock_init(&base->lock); |
d730e882 | 1612 | |
1da177e4 LT |
1613 | for (j = 0; j < TVN_SIZE; j++) { |
1614 | INIT_LIST_HEAD(base->tv5.vec + j); | |
1615 | INIT_LIST_HEAD(base->tv4.vec + j); | |
1616 | INIT_LIST_HEAD(base->tv3.vec + j); | |
1617 | INIT_LIST_HEAD(base->tv2.vec + j); | |
1618 | } | |
1619 | for (j = 0; j < TVR_SIZE; j++) | |
1620 | INIT_LIST_HEAD(base->tv1.vec + j); | |
1621 | ||
1622 | base->timer_jiffies = jiffies; | |
97fd9ed4 | 1623 | base->next_timer = base->timer_jiffies; |
a4a6198b | 1624 | return 0; |
1da177e4 LT |
1625 | } |
1626 | ||
1627 | #ifdef CONFIG_HOTPLUG_CPU | |
a6fa8e5a | 1628 | static void migrate_timer_list(struct tvec_base *new_base, struct list_head *head) |
1da177e4 LT |
1629 | { |
1630 | struct timer_list *timer; | |
1631 | ||
1632 | while (!list_empty(head)) { | |
b5e61818 | 1633 | timer = list_first_entry(head, struct timer_list, entry); |
55c888d6 | 1634 | detach_timer(timer, 0); |
6e453a67 | 1635 | timer_set_base(timer, new_base); |
97fd9ed4 MS |
1636 | if (time_before(timer->expires, new_base->next_timer) && |
1637 | !tbase_get_deferrable(timer->base)) | |
1638 | new_base->next_timer = timer->expires; | |
1da177e4 | 1639 | internal_add_timer(new_base, timer); |
1da177e4 | 1640 | } |
1da177e4 LT |
1641 | } |
1642 | ||
48ccf3da | 1643 | static void __cpuinit migrate_timers(int cpu) |
1da177e4 | 1644 | { |
a6fa8e5a PM |
1645 | struct tvec_base *old_base; |
1646 | struct tvec_base *new_base; | |
1da177e4 LT |
1647 | int i; |
1648 | ||
1649 | BUG_ON(cpu_online(cpu)); | |
a4a6198b JB |
1650 | old_base = per_cpu(tvec_bases, cpu); |
1651 | new_base = get_cpu_var(tvec_bases); | |
d82f0b0f ON |
1652 | /* |
1653 | * The caller is globally serialized and nobody else | |
1654 | * takes two locks at once, deadlock is not possible. | |
1655 | */ | |
1656 | spin_lock_irq(&new_base->lock); | |
0d180406 | 1657 | spin_lock_nested(&old_base->lock, SINGLE_DEPTH_NESTING); |
3691c519 ON |
1658 | |
1659 | BUG_ON(old_base->running_timer); | |
1da177e4 | 1660 | |
1da177e4 | 1661 | for (i = 0; i < TVR_SIZE; i++) |
55c888d6 ON |
1662 | migrate_timer_list(new_base, old_base->tv1.vec + i); |
1663 | for (i = 0; i < TVN_SIZE; i++) { | |
1664 | migrate_timer_list(new_base, old_base->tv2.vec + i); | |
1665 | migrate_timer_list(new_base, old_base->tv3.vec + i); | |
1666 | migrate_timer_list(new_base, old_base->tv4.vec + i); | |
1667 | migrate_timer_list(new_base, old_base->tv5.vec + i); | |
1668 | } | |
1669 | ||
0d180406 | 1670 | spin_unlock(&old_base->lock); |
d82f0b0f | 1671 | spin_unlock_irq(&new_base->lock); |
1da177e4 | 1672 | put_cpu_var(tvec_bases); |
1da177e4 LT |
1673 | } |
1674 | #endif /* CONFIG_HOTPLUG_CPU */ | |
1675 | ||
8c78f307 | 1676 | static int __cpuinit timer_cpu_notify(struct notifier_block *self, |
1da177e4 LT |
1677 | unsigned long action, void *hcpu) |
1678 | { | |
1679 | long cpu = (long)hcpu; | |
1680 | switch(action) { | |
1681 | case CPU_UP_PREPARE: | |
8bb78442 | 1682 | case CPU_UP_PREPARE_FROZEN: |
a4a6198b JB |
1683 | if (init_timers_cpu(cpu) < 0) |
1684 | return NOTIFY_BAD; | |
1da177e4 LT |
1685 | break; |
1686 | #ifdef CONFIG_HOTPLUG_CPU | |
1687 | case CPU_DEAD: | |
8bb78442 | 1688 | case CPU_DEAD_FROZEN: |
1da177e4 LT |
1689 | migrate_timers(cpu); |
1690 | break; | |
1691 | #endif | |
1692 | default: | |
1693 | break; | |
1694 | } | |
1695 | return NOTIFY_OK; | |
1696 | } | |
1697 | ||
8c78f307 | 1698 | static struct notifier_block __cpuinitdata timers_nb = { |
1da177e4 LT |
1699 | .notifier_call = timer_cpu_notify, |
1700 | }; | |
1701 | ||
1702 | ||
1703 | void __init init_timers(void) | |
1704 | { | |
07dccf33 | 1705 | int err = timer_cpu_notify(&timers_nb, (unsigned long)CPU_UP_PREPARE, |
1da177e4 | 1706 | (void *)(long)smp_processor_id()); |
07dccf33 | 1707 | |
82f67cd9 IM |
1708 | init_timer_stats(); |
1709 | ||
07dccf33 | 1710 | BUG_ON(err == NOTIFY_BAD); |
1da177e4 | 1711 | register_cpu_notifier(&timers_nb); |
962cf36c | 1712 | open_softirq(TIMER_SOFTIRQ, run_timer_softirq); |
1da177e4 LT |
1713 | } |
1714 | ||
1da177e4 LT |
1715 | /** |
1716 | * msleep - sleep safely even with waitqueue interruptions | |
1717 | * @msecs: Time in milliseconds to sleep for | |
1718 | */ | |
1719 | void msleep(unsigned int msecs) | |
1720 | { | |
1721 | unsigned long timeout = msecs_to_jiffies(msecs) + 1; | |
1722 | ||
75bcc8c5 NA |
1723 | while (timeout) |
1724 | timeout = schedule_timeout_uninterruptible(timeout); | |
1da177e4 LT |
1725 | } |
1726 | ||
1727 | EXPORT_SYMBOL(msleep); | |
1728 | ||
1729 | /** | |
96ec3efd | 1730 | * msleep_interruptible - sleep waiting for signals |
1da177e4 LT |
1731 | * @msecs: Time in milliseconds to sleep for |
1732 | */ | |
1733 | unsigned long msleep_interruptible(unsigned int msecs) | |
1734 | { | |
1735 | unsigned long timeout = msecs_to_jiffies(msecs) + 1; | |
1736 | ||
75bcc8c5 NA |
1737 | while (timeout && !signal_pending(current)) |
1738 | timeout = schedule_timeout_interruptible(timeout); | |
1da177e4 LT |
1739 | return jiffies_to_msecs(timeout); |
1740 | } | |
1741 | ||
1742 | EXPORT_SYMBOL(msleep_interruptible); |