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> |
5a0e3ad6 | 42 | #include <linux/slab.h> |
1da177e4 LT |
43 | |
44 | #include <asm/uaccess.h> | |
45 | #include <asm/unistd.h> | |
46 | #include <asm/div64.h> | |
47 | #include <asm/timex.h> | |
48 | #include <asm/io.h> | |
49 | ||
2b022e3d XG |
50 | #define CREATE_TRACE_POINTS |
51 | #include <trace/events/timer.h> | |
52 | ||
ecea8d19 TG |
53 | u64 jiffies_64 __cacheline_aligned_in_smp = INITIAL_JIFFIES; |
54 | ||
55 | EXPORT_SYMBOL(jiffies_64); | |
56 | ||
1da177e4 LT |
57 | /* |
58 | * per-CPU timer vector definitions: | |
59 | */ | |
1da177e4 LT |
60 | #define TVN_BITS (CONFIG_BASE_SMALL ? 4 : 6) |
61 | #define TVR_BITS (CONFIG_BASE_SMALL ? 6 : 8) | |
62 | #define TVN_SIZE (1 << TVN_BITS) | |
63 | #define TVR_SIZE (1 << TVR_BITS) | |
64 | #define TVN_MASK (TVN_SIZE - 1) | |
65 | #define TVR_MASK (TVR_SIZE - 1) | |
66 | ||
a6fa8e5a | 67 | struct tvec { |
1da177e4 | 68 | struct list_head vec[TVN_SIZE]; |
a6fa8e5a | 69 | }; |
1da177e4 | 70 | |
a6fa8e5a | 71 | struct tvec_root { |
1da177e4 | 72 | struct list_head vec[TVR_SIZE]; |
a6fa8e5a | 73 | }; |
1da177e4 | 74 | |
a6fa8e5a | 75 | struct tvec_base { |
3691c519 ON |
76 | spinlock_t lock; |
77 | struct timer_list *running_timer; | |
1da177e4 | 78 | unsigned long timer_jiffies; |
97fd9ed4 | 79 | unsigned long next_timer; |
a6fa8e5a PM |
80 | struct tvec_root tv1; |
81 | struct tvec tv2; | |
82 | struct tvec tv3; | |
83 | struct tvec tv4; | |
84 | struct tvec tv5; | |
6e453a67 | 85 | } ____cacheline_aligned; |
1da177e4 | 86 | |
a6fa8e5a | 87 | struct tvec_base boot_tvec_bases; |
3691c519 | 88 | EXPORT_SYMBOL(boot_tvec_bases); |
a6fa8e5a | 89 | static DEFINE_PER_CPU(struct tvec_base *, tvec_bases) = &boot_tvec_bases; |
1da177e4 | 90 | |
6e453a67 | 91 | /* |
a6fa8e5a | 92 | * Note that all tvec_bases are 2 byte aligned and lower bit of |
866e2611 BF |
93 | * base in timer_list is guaranteed to be zero. Use the LSB to |
94 | * indicate whether the timer is deferrable. | |
95 | * | |
96 | * A deferrable timer will work normally when the system is busy, but | |
97 | * will not cause a CPU to come out of idle just to service it; instead, | |
98 | * the timer will be serviced when the CPU eventually wakes up with a | |
99 | * subsequent non-deferrable timer. | |
6e453a67 VP |
100 | */ |
101 | #define TBASE_DEFERRABLE_FLAG (0x1) | |
102 | ||
103 | /* Functions below help us manage 'deferrable' flag */ | |
a6fa8e5a | 104 | static inline unsigned int tbase_get_deferrable(struct tvec_base *base) |
6e453a67 | 105 | { |
e9910846 | 106 | return ((unsigned int)(unsigned long)base & TBASE_DEFERRABLE_FLAG); |
6e453a67 VP |
107 | } |
108 | ||
a6fa8e5a | 109 | static inline struct tvec_base *tbase_get_base(struct tvec_base *base) |
6e453a67 | 110 | { |
a6fa8e5a | 111 | return ((struct tvec_base *)((unsigned long)base & ~TBASE_DEFERRABLE_FLAG)); |
6e453a67 VP |
112 | } |
113 | ||
114 | static inline void timer_set_deferrable(struct timer_list *timer) | |
115 | { | |
a6fa8e5a | 116 | timer->base = ((struct tvec_base *)((unsigned long)(timer->base) | |
6819457d | 117 | TBASE_DEFERRABLE_FLAG)); |
6e453a67 VP |
118 | } |
119 | ||
120 | static inline void | |
a6fa8e5a | 121 | timer_set_base(struct timer_list *timer, struct tvec_base *new_base) |
6e453a67 | 122 | { |
a6fa8e5a | 123 | timer->base = (struct tvec_base *)((unsigned long)(new_base) | |
6819457d | 124 | tbase_get_deferrable(timer->base)); |
6e453a67 VP |
125 | } |
126 | ||
9c133c46 AS |
127 | static unsigned long round_jiffies_common(unsigned long j, int cpu, |
128 | bool force_up) | |
4c36a5de AV |
129 | { |
130 | int rem; | |
131 | unsigned long original = j; | |
132 | ||
133 | /* | |
134 | * We don't want all cpus firing their timers at once hitting the | |
135 | * same lock or cachelines, so we skew each extra cpu with an extra | |
136 | * 3 jiffies. This 3 jiffies came originally from the mm/ code which | |
137 | * already did this. | |
138 | * The skew is done by adding 3*cpunr, then round, then subtract this | |
139 | * extra offset again. | |
140 | */ | |
141 | j += cpu * 3; | |
142 | ||
143 | rem = j % HZ; | |
144 | ||
145 | /* | |
146 | * If the target jiffie is just after a whole second (which can happen | |
147 | * due to delays of the timer irq, long irq off times etc etc) then | |
148 | * we should round down to the whole second, not up. Use 1/4th second | |
149 | * as cutoff for this rounding as an extreme upper bound for this. | |
9c133c46 | 150 | * But never round down if @force_up is set. |
4c36a5de | 151 | */ |
9c133c46 | 152 | if (rem < HZ/4 && !force_up) /* round down */ |
4c36a5de AV |
153 | j = j - rem; |
154 | else /* round up */ | |
155 | j = j - rem + HZ; | |
156 | ||
157 | /* now that we have rounded, subtract the extra skew again */ | |
158 | j -= cpu * 3; | |
159 | ||
160 | if (j <= jiffies) /* rounding ate our timeout entirely; */ | |
161 | return original; | |
162 | return j; | |
163 | } | |
9c133c46 AS |
164 | |
165 | /** | |
166 | * __round_jiffies - function to round jiffies to a full second | |
167 | * @j: the time in (absolute) jiffies that should be rounded | |
168 | * @cpu: the processor number on which the timeout will happen | |
169 | * | |
170 | * __round_jiffies() rounds an absolute time in the future (in jiffies) | |
171 | * up or down to (approximately) full seconds. This is useful for timers | |
172 | * for which the exact time they fire does not matter too much, as long as | |
173 | * they fire approximately every X seconds. | |
174 | * | |
175 | * By rounding these timers to whole seconds, all such timers will fire | |
176 | * at the same time, rather than at various times spread out. The goal | |
177 | * of this is to have the CPU wake up less, which saves power. | |
178 | * | |
179 | * The exact rounding is skewed for each processor to avoid all | |
180 | * processors firing at the exact same time, which could lead | |
181 | * to lock contention or spurious cache line bouncing. | |
182 | * | |
183 | * The return value is the rounded version of the @j parameter. | |
184 | */ | |
185 | unsigned long __round_jiffies(unsigned long j, int cpu) | |
186 | { | |
187 | return round_jiffies_common(j, cpu, false); | |
188 | } | |
4c36a5de AV |
189 | EXPORT_SYMBOL_GPL(__round_jiffies); |
190 | ||
191 | /** | |
192 | * __round_jiffies_relative - function to round jiffies to a full second | |
193 | * @j: the time in (relative) jiffies that should be rounded | |
194 | * @cpu: the processor number on which the timeout will happen | |
195 | * | |
72fd4a35 | 196 | * __round_jiffies_relative() rounds a time delta in the future (in jiffies) |
4c36a5de AV |
197 | * up or down to (approximately) full seconds. This is useful for timers |
198 | * for which the exact time they fire does not matter too much, as long as | |
199 | * they fire approximately every X seconds. | |
200 | * | |
201 | * By rounding these timers to whole seconds, all such timers will fire | |
202 | * at the same time, rather than at various times spread out. The goal | |
203 | * of this is to have the CPU wake up less, which saves power. | |
204 | * | |
205 | * The exact rounding is skewed for each processor to avoid all | |
206 | * processors firing at the exact same time, which could lead | |
207 | * to lock contention or spurious cache line bouncing. | |
208 | * | |
72fd4a35 | 209 | * The return value is the rounded version of the @j parameter. |
4c36a5de AV |
210 | */ |
211 | unsigned long __round_jiffies_relative(unsigned long j, int cpu) | |
212 | { | |
9c133c46 AS |
213 | unsigned long j0 = jiffies; |
214 | ||
215 | /* Use j0 because jiffies might change while we run */ | |
216 | return round_jiffies_common(j + j0, cpu, false) - j0; | |
4c36a5de AV |
217 | } |
218 | EXPORT_SYMBOL_GPL(__round_jiffies_relative); | |
219 | ||
220 | /** | |
221 | * round_jiffies - function to round jiffies to a full second | |
222 | * @j: the time in (absolute) jiffies that should be rounded | |
223 | * | |
72fd4a35 | 224 | * round_jiffies() rounds an absolute time in the future (in jiffies) |
4c36a5de AV |
225 | * up or down to (approximately) full seconds. This is useful for timers |
226 | * for which the exact time they fire does not matter too much, as long as | |
227 | * they fire approximately every X seconds. | |
228 | * | |
229 | * By rounding these timers to whole seconds, all such timers will fire | |
230 | * at the same time, rather than at various times spread out. The goal | |
231 | * of this is to have the CPU wake up less, which saves power. | |
232 | * | |
72fd4a35 | 233 | * The return value is the rounded version of the @j parameter. |
4c36a5de AV |
234 | */ |
235 | unsigned long round_jiffies(unsigned long j) | |
236 | { | |
9c133c46 | 237 | return round_jiffies_common(j, raw_smp_processor_id(), false); |
4c36a5de AV |
238 | } |
239 | EXPORT_SYMBOL_GPL(round_jiffies); | |
240 | ||
241 | /** | |
242 | * round_jiffies_relative - function to round jiffies to a full second | |
243 | * @j: the time in (relative) jiffies that should be rounded | |
244 | * | |
72fd4a35 | 245 | * round_jiffies_relative() rounds a time delta in the future (in jiffies) |
4c36a5de AV |
246 | * up or down to (approximately) full seconds. This is useful for timers |
247 | * for which the exact time they fire does not matter too much, as long as | |
248 | * they fire approximately every X seconds. | |
249 | * | |
250 | * By rounding these timers to whole seconds, all such timers will fire | |
251 | * at the same time, rather than at various times spread out. The goal | |
252 | * of this is to have the CPU wake up less, which saves power. | |
253 | * | |
72fd4a35 | 254 | * The return value is the rounded version of the @j parameter. |
4c36a5de AV |
255 | */ |
256 | unsigned long round_jiffies_relative(unsigned long j) | |
257 | { | |
258 | return __round_jiffies_relative(j, raw_smp_processor_id()); | |
259 | } | |
260 | EXPORT_SYMBOL_GPL(round_jiffies_relative); | |
261 | ||
9c133c46 AS |
262 | /** |
263 | * __round_jiffies_up - function to round jiffies up to a full second | |
264 | * @j: the time in (absolute) jiffies that should be rounded | |
265 | * @cpu: the processor number on which the timeout will happen | |
266 | * | |
267 | * This is the same as __round_jiffies() except that it will never | |
268 | * round down. This is useful for timeouts for which the exact time | |
269 | * of firing does not matter too much, as long as they don't fire too | |
270 | * early. | |
271 | */ | |
272 | unsigned long __round_jiffies_up(unsigned long j, int cpu) | |
273 | { | |
274 | return round_jiffies_common(j, cpu, true); | |
275 | } | |
276 | EXPORT_SYMBOL_GPL(__round_jiffies_up); | |
277 | ||
278 | /** | |
279 | * __round_jiffies_up_relative - function to round jiffies up to a full second | |
280 | * @j: the time in (relative) jiffies that should be rounded | |
281 | * @cpu: the processor number on which the timeout will happen | |
282 | * | |
283 | * This is the same as __round_jiffies_relative() except that it will never | |
284 | * round down. This is useful for timeouts for which the exact time | |
285 | * of firing does not matter too much, as long as they don't fire too | |
286 | * early. | |
287 | */ | |
288 | unsigned long __round_jiffies_up_relative(unsigned long j, int cpu) | |
289 | { | |
290 | unsigned long j0 = jiffies; | |
291 | ||
292 | /* Use j0 because jiffies might change while we run */ | |
293 | return round_jiffies_common(j + j0, cpu, true) - j0; | |
294 | } | |
295 | EXPORT_SYMBOL_GPL(__round_jiffies_up_relative); | |
296 | ||
297 | /** | |
298 | * round_jiffies_up - function to round jiffies up to a full second | |
299 | * @j: the time in (absolute) jiffies that should be rounded | |
300 | * | |
301 | * This is the same as round_jiffies() except that it will never | |
302 | * round down. This is useful for timeouts for which the exact time | |
303 | * of firing does not matter too much, as long as they don't fire too | |
304 | * early. | |
305 | */ | |
306 | unsigned long round_jiffies_up(unsigned long j) | |
307 | { | |
308 | return round_jiffies_common(j, raw_smp_processor_id(), true); | |
309 | } | |
310 | EXPORT_SYMBOL_GPL(round_jiffies_up); | |
311 | ||
312 | /** | |
313 | * round_jiffies_up_relative - function to round jiffies up to a full second | |
314 | * @j: the time in (relative) jiffies that should be rounded | |
315 | * | |
316 | * This is the same as round_jiffies_relative() except that it will never | |
317 | * round down. This is useful for timeouts for which the exact time | |
318 | * of firing does not matter too much, as long as they don't fire too | |
319 | * early. | |
320 | */ | |
321 | unsigned long round_jiffies_up_relative(unsigned long j) | |
322 | { | |
323 | return __round_jiffies_up_relative(j, raw_smp_processor_id()); | |
324 | } | |
325 | EXPORT_SYMBOL_GPL(round_jiffies_up_relative); | |
326 | ||
3bbb9ec9 AV |
327 | /** |
328 | * set_timer_slack - set the allowed slack for a timer | |
329 | * @slack_hz: the amount of time (in jiffies) allowed for rounding | |
330 | * | |
331 | * Set the amount of time, in jiffies, that a certain timer has | |
332 | * in terms of slack. By setting this value, the timer subsystem | |
333 | * will schedule the actual timer somewhere between | |
334 | * the time mod_timer() asks for, and that time plus the slack. | |
335 | * | |
336 | * By setting the slack to -1, a percentage of the delay is used | |
337 | * instead. | |
338 | */ | |
339 | void set_timer_slack(struct timer_list *timer, int slack_hz) | |
340 | { | |
341 | timer->slack = slack_hz; | |
342 | } | |
343 | EXPORT_SYMBOL_GPL(set_timer_slack); | |
344 | ||
4c36a5de | 345 | |
a6fa8e5a | 346 | static inline void set_running_timer(struct tvec_base *base, |
1da177e4 LT |
347 | struct timer_list *timer) |
348 | { | |
349 | #ifdef CONFIG_SMP | |
3691c519 | 350 | base->running_timer = timer; |
1da177e4 LT |
351 | #endif |
352 | } | |
353 | ||
a6fa8e5a | 354 | static void internal_add_timer(struct tvec_base *base, struct timer_list *timer) |
1da177e4 LT |
355 | { |
356 | unsigned long expires = timer->expires; | |
357 | unsigned long idx = expires - base->timer_jiffies; | |
358 | struct list_head *vec; | |
359 | ||
360 | if (idx < TVR_SIZE) { | |
361 | int i = expires & TVR_MASK; | |
362 | vec = base->tv1.vec + i; | |
363 | } else if (idx < 1 << (TVR_BITS + TVN_BITS)) { | |
364 | int i = (expires >> TVR_BITS) & TVN_MASK; | |
365 | vec = base->tv2.vec + i; | |
366 | } else if (idx < 1 << (TVR_BITS + 2 * TVN_BITS)) { | |
367 | int i = (expires >> (TVR_BITS + TVN_BITS)) & TVN_MASK; | |
368 | vec = base->tv3.vec + i; | |
369 | } else if (idx < 1 << (TVR_BITS + 3 * TVN_BITS)) { | |
370 | int i = (expires >> (TVR_BITS + 2 * TVN_BITS)) & TVN_MASK; | |
371 | vec = base->tv4.vec + i; | |
372 | } else if ((signed long) idx < 0) { | |
373 | /* | |
374 | * Can happen if you add a timer with expires == jiffies, | |
375 | * or you set a timer to go off in the past | |
376 | */ | |
377 | vec = base->tv1.vec + (base->timer_jiffies & TVR_MASK); | |
378 | } else { | |
379 | int i; | |
380 | /* If the timeout is larger than 0xffffffff on 64-bit | |
381 | * architectures then we use the maximum timeout: | |
382 | */ | |
383 | if (idx > 0xffffffffUL) { | |
384 | idx = 0xffffffffUL; | |
385 | expires = idx + base->timer_jiffies; | |
386 | } | |
387 | i = (expires >> (TVR_BITS + 3 * TVN_BITS)) & TVN_MASK; | |
388 | vec = base->tv5.vec + i; | |
389 | } | |
390 | /* | |
391 | * Timers are FIFO: | |
392 | */ | |
393 | list_add_tail(&timer->entry, vec); | |
394 | } | |
395 | ||
82f67cd9 IM |
396 | #ifdef CONFIG_TIMER_STATS |
397 | void __timer_stats_timer_set_start_info(struct timer_list *timer, void *addr) | |
398 | { | |
399 | if (timer->start_site) | |
400 | return; | |
401 | ||
402 | timer->start_site = addr; | |
403 | memcpy(timer->start_comm, current->comm, TASK_COMM_LEN); | |
404 | timer->start_pid = current->pid; | |
405 | } | |
c5c061b8 VP |
406 | |
407 | static void timer_stats_account_timer(struct timer_list *timer) | |
408 | { | |
409 | unsigned int flag = 0; | |
410 | ||
507e1231 HC |
411 | if (likely(!timer->start_site)) |
412 | return; | |
c5c061b8 VP |
413 | if (unlikely(tbase_get_deferrable(timer->base))) |
414 | flag |= TIMER_STATS_FLAG_DEFERRABLE; | |
415 | ||
416 | timer_stats_update_stats(timer, timer->start_pid, timer->start_site, | |
417 | timer->function, timer->start_comm, flag); | |
418 | } | |
419 | ||
420 | #else | |
421 | static void timer_stats_account_timer(struct timer_list *timer) {} | |
82f67cd9 IM |
422 | #endif |
423 | ||
c6f3a97f TG |
424 | #ifdef CONFIG_DEBUG_OBJECTS_TIMERS |
425 | ||
426 | static struct debug_obj_descr timer_debug_descr; | |
427 | ||
428 | /* | |
429 | * fixup_init is called when: | |
430 | * - an active object is initialized | |
55c888d6 | 431 | */ |
c6f3a97f TG |
432 | static int timer_fixup_init(void *addr, enum debug_obj_state state) |
433 | { | |
434 | struct timer_list *timer = addr; | |
435 | ||
436 | switch (state) { | |
437 | case ODEBUG_STATE_ACTIVE: | |
438 | del_timer_sync(timer); | |
439 | debug_object_init(timer, &timer_debug_descr); | |
440 | return 1; | |
441 | default: | |
442 | return 0; | |
443 | } | |
444 | } | |
445 | ||
446 | /* | |
447 | * fixup_activate is called when: | |
448 | * - an active object is activated | |
449 | * - an unknown object is activated (might be a statically initialized object) | |
450 | */ | |
451 | static int timer_fixup_activate(void *addr, enum debug_obj_state state) | |
452 | { | |
453 | struct timer_list *timer = addr; | |
454 | ||
455 | switch (state) { | |
456 | ||
457 | case ODEBUG_STATE_NOTAVAILABLE: | |
458 | /* | |
459 | * This is not really a fixup. The timer was | |
460 | * statically initialized. We just make sure that it | |
461 | * is tracked in the object tracker. | |
462 | */ | |
463 | if (timer->entry.next == NULL && | |
464 | timer->entry.prev == TIMER_ENTRY_STATIC) { | |
465 | debug_object_init(timer, &timer_debug_descr); | |
466 | debug_object_activate(timer, &timer_debug_descr); | |
467 | return 0; | |
468 | } else { | |
469 | WARN_ON_ONCE(1); | |
470 | } | |
471 | return 0; | |
472 | ||
473 | case ODEBUG_STATE_ACTIVE: | |
474 | WARN_ON(1); | |
475 | ||
476 | default: | |
477 | return 0; | |
478 | } | |
479 | } | |
480 | ||
481 | /* | |
482 | * fixup_free is called when: | |
483 | * - an active object is freed | |
484 | */ | |
485 | static int timer_fixup_free(void *addr, enum debug_obj_state state) | |
486 | { | |
487 | struct timer_list *timer = addr; | |
488 | ||
489 | switch (state) { | |
490 | case ODEBUG_STATE_ACTIVE: | |
491 | del_timer_sync(timer); | |
492 | debug_object_free(timer, &timer_debug_descr); | |
493 | return 1; | |
494 | default: | |
495 | return 0; | |
496 | } | |
497 | } | |
498 | ||
499 | static struct debug_obj_descr timer_debug_descr = { | |
500 | .name = "timer_list", | |
501 | .fixup_init = timer_fixup_init, | |
502 | .fixup_activate = timer_fixup_activate, | |
503 | .fixup_free = timer_fixup_free, | |
504 | }; | |
505 | ||
506 | static inline void debug_timer_init(struct timer_list *timer) | |
507 | { | |
508 | debug_object_init(timer, &timer_debug_descr); | |
509 | } | |
510 | ||
511 | static inline void debug_timer_activate(struct timer_list *timer) | |
512 | { | |
513 | debug_object_activate(timer, &timer_debug_descr); | |
514 | } | |
515 | ||
516 | static inline void debug_timer_deactivate(struct timer_list *timer) | |
517 | { | |
518 | debug_object_deactivate(timer, &timer_debug_descr); | |
519 | } | |
520 | ||
521 | static inline void debug_timer_free(struct timer_list *timer) | |
522 | { | |
523 | debug_object_free(timer, &timer_debug_descr); | |
524 | } | |
525 | ||
6f2b9b9a JB |
526 | static void __init_timer(struct timer_list *timer, |
527 | const char *name, | |
528 | struct lock_class_key *key); | |
c6f3a97f | 529 | |
6f2b9b9a JB |
530 | void init_timer_on_stack_key(struct timer_list *timer, |
531 | const char *name, | |
532 | struct lock_class_key *key) | |
c6f3a97f TG |
533 | { |
534 | debug_object_init_on_stack(timer, &timer_debug_descr); | |
6f2b9b9a | 535 | __init_timer(timer, name, key); |
c6f3a97f | 536 | } |
6f2b9b9a | 537 | EXPORT_SYMBOL_GPL(init_timer_on_stack_key); |
c6f3a97f TG |
538 | |
539 | void destroy_timer_on_stack(struct timer_list *timer) | |
540 | { | |
541 | debug_object_free(timer, &timer_debug_descr); | |
542 | } | |
543 | EXPORT_SYMBOL_GPL(destroy_timer_on_stack); | |
544 | ||
545 | #else | |
546 | static inline void debug_timer_init(struct timer_list *timer) { } | |
547 | static inline void debug_timer_activate(struct timer_list *timer) { } | |
548 | static inline void debug_timer_deactivate(struct timer_list *timer) { } | |
549 | #endif | |
550 | ||
2b022e3d XG |
551 | static inline void debug_init(struct timer_list *timer) |
552 | { | |
553 | debug_timer_init(timer); | |
554 | trace_timer_init(timer); | |
555 | } | |
556 | ||
557 | static inline void | |
558 | debug_activate(struct timer_list *timer, unsigned long expires) | |
559 | { | |
560 | debug_timer_activate(timer); | |
561 | trace_timer_start(timer, expires); | |
562 | } | |
563 | ||
564 | static inline void debug_deactivate(struct timer_list *timer) | |
565 | { | |
566 | debug_timer_deactivate(timer); | |
567 | trace_timer_cancel(timer); | |
568 | } | |
569 | ||
6f2b9b9a JB |
570 | static void __init_timer(struct timer_list *timer, |
571 | const char *name, | |
572 | struct lock_class_key *key) | |
55c888d6 ON |
573 | { |
574 | timer->entry.next = NULL; | |
bfe5d834 | 575 | timer->base = __raw_get_cpu_var(tvec_bases); |
3bbb9ec9 | 576 | timer->slack = -1; |
82f67cd9 IM |
577 | #ifdef CONFIG_TIMER_STATS |
578 | timer->start_site = NULL; | |
579 | timer->start_pid = -1; | |
580 | memset(timer->start_comm, 0, TASK_COMM_LEN); | |
581 | #endif | |
6f2b9b9a | 582 | lockdep_init_map(&timer->lockdep_map, name, key, 0); |
55c888d6 | 583 | } |
c6f3a97f | 584 | |
8cadd283 JB |
585 | void setup_deferrable_timer_on_stack_key(struct timer_list *timer, |
586 | const char *name, | |
587 | struct lock_class_key *key, | |
588 | void (*function)(unsigned long), | |
589 | unsigned long data) | |
590 | { | |
591 | timer->function = function; | |
592 | timer->data = data; | |
593 | init_timer_on_stack_key(timer, name, key); | |
594 | timer_set_deferrable(timer); | |
595 | } | |
596 | EXPORT_SYMBOL_GPL(setup_deferrable_timer_on_stack_key); | |
597 | ||
c6f3a97f | 598 | /** |
633fe795 | 599 | * init_timer_key - initialize a timer |
c6f3a97f | 600 | * @timer: the timer to be initialized |
633fe795 RD |
601 | * @name: name of the timer |
602 | * @key: lockdep class key of the fake lock used for tracking timer | |
603 | * sync lock dependencies | |
c6f3a97f | 604 | * |
633fe795 | 605 | * init_timer_key() must be done to a timer prior calling *any* of the |
c6f3a97f TG |
606 | * other timer functions. |
607 | */ | |
6f2b9b9a JB |
608 | void init_timer_key(struct timer_list *timer, |
609 | const char *name, | |
610 | struct lock_class_key *key) | |
c6f3a97f | 611 | { |
2b022e3d | 612 | debug_init(timer); |
6f2b9b9a | 613 | __init_timer(timer, name, key); |
c6f3a97f | 614 | } |
6f2b9b9a | 615 | EXPORT_SYMBOL(init_timer_key); |
55c888d6 | 616 | |
6f2b9b9a JB |
617 | void init_timer_deferrable_key(struct timer_list *timer, |
618 | const char *name, | |
619 | struct lock_class_key *key) | |
6e453a67 | 620 | { |
6f2b9b9a | 621 | init_timer_key(timer, name, key); |
6e453a67 VP |
622 | timer_set_deferrable(timer); |
623 | } | |
6f2b9b9a | 624 | EXPORT_SYMBOL(init_timer_deferrable_key); |
6e453a67 | 625 | |
55c888d6 | 626 | static inline void detach_timer(struct timer_list *timer, |
82f67cd9 | 627 | int clear_pending) |
55c888d6 ON |
628 | { |
629 | struct list_head *entry = &timer->entry; | |
630 | ||
2b022e3d | 631 | debug_deactivate(timer); |
c6f3a97f | 632 | |
55c888d6 ON |
633 | __list_del(entry->prev, entry->next); |
634 | if (clear_pending) | |
635 | entry->next = NULL; | |
636 | entry->prev = LIST_POISON2; | |
637 | } | |
638 | ||
639 | /* | |
3691c519 | 640 | * We are using hashed locking: holding per_cpu(tvec_bases).lock |
55c888d6 ON |
641 | * means that all timers which are tied to this base via timer->base are |
642 | * locked, and the base itself is locked too. | |
643 | * | |
644 | * So __run_timers/migrate_timers can safely modify all timers which could | |
645 | * be found on ->tvX lists. | |
646 | * | |
647 | * When the timer's base is locked, and the timer removed from list, it is | |
648 | * possible to set timer->base = NULL and drop the lock: the timer remains | |
649 | * locked. | |
650 | */ | |
a6fa8e5a | 651 | static struct tvec_base *lock_timer_base(struct timer_list *timer, |
55c888d6 | 652 | unsigned long *flags) |
89e7e374 | 653 | __acquires(timer->base->lock) |
55c888d6 | 654 | { |
a6fa8e5a | 655 | struct tvec_base *base; |
55c888d6 ON |
656 | |
657 | for (;;) { | |
a6fa8e5a | 658 | struct tvec_base *prelock_base = timer->base; |
6e453a67 | 659 | base = tbase_get_base(prelock_base); |
55c888d6 ON |
660 | if (likely(base != NULL)) { |
661 | spin_lock_irqsave(&base->lock, *flags); | |
6e453a67 | 662 | if (likely(prelock_base == timer->base)) |
55c888d6 ON |
663 | return base; |
664 | /* The timer has migrated to another CPU */ | |
665 | spin_unlock_irqrestore(&base->lock, *flags); | |
666 | } | |
667 | cpu_relax(); | |
668 | } | |
669 | } | |
670 | ||
74019224 | 671 | static inline int |
597d0275 AB |
672 | __mod_timer(struct timer_list *timer, unsigned long expires, |
673 | bool pending_only, int pinned) | |
1da177e4 | 674 | { |
a6fa8e5a | 675 | struct tvec_base *base, *new_base; |
1da177e4 | 676 | unsigned long flags; |
eea08f32 | 677 | int ret = 0 , cpu; |
1da177e4 | 678 | |
82f67cd9 | 679 | timer_stats_timer_set_start_info(timer); |
1da177e4 | 680 | BUG_ON(!timer->function); |
1da177e4 | 681 | |
55c888d6 ON |
682 | base = lock_timer_base(timer, &flags); |
683 | ||
684 | if (timer_pending(timer)) { | |
685 | detach_timer(timer, 0); | |
97fd9ed4 MS |
686 | if (timer->expires == base->next_timer && |
687 | !tbase_get_deferrable(timer->base)) | |
688 | base->next_timer = base->timer_jiffies; | |
55c888d6 | 689 | ret = 1; |
74019224 IM |
690 | } else { |
691 | if (pending_only) | |
692 | goto out_unlock; | |
55c888d6 ON |
693 | } |
694 | ||
2b022e3d | 695 | debug_activate(timer, expires); |
c6f3a97f | 696 | |
eea08f32 AB |
697 | cpu = smp_processor_id(); |
698 | ||
699 | #if defined(CONFIG_NO_HZ) && defined(CONFIG_SMP) | |
83cd4fe2 VP |
700 | if (!pinned && get_sysctl_timer_migration() && idle_cpu(cpu)) |
701 | cpu = get_nohz_timer_target(); | |
eea08f32 AB |
702 | #endif |
703 | new_base = per_cpu(tvec_bases, cpu); | |
704 | ||
3691c519 | 705 | if (base != new_base) { |
1da177e4 | 706 | /* |
55c888d6 ON |
707 | * We are trying to schedule the timer on the local CPU. |
708 | * However we can't change timer's base while it is running, | |
709 | * otherwise del_timer_sync() can't detect that the timer's | |
710 | * handler yet has not finished. This also guarantees that | |
711 | * the timer is serialized wrt itself. | |
1da177e4 | 712 | */ |
a2c348fe | 713 | if (likely(base->running_timer != timer)) { |
55c888d6 | 714 | /* See the comment in lock_timer_base() */ |
6e453a67 | 715 | timer_set_base(timer, NULL); |
55c888d6 | 716 | spin_unlock(&base->lock); |
a2c348fe ON |
717 | base = new_base; |
718 | spin_lock(&base->lock); | |
6e453a67 | 719 | timer_set_base(timer, base); |
1da177e4 LT |
720 | } |
721 | } | |
722 | ||
1da177e4 | 723 | timer->expires = expires; |
97fd9ed4 MS |
724 | if (time_before(timer->expires, base->next_timer) && |
725 | !tbase_get_deferrable(timer->base)) | |
726 | base->next_timer = timer->expires; | |
a2c348fe | 727 | internal_add_timer(base, timer); |
74019224 IM |
728 | |
729 | out_unlock: | |
a2c348fe | 730 | spin_unlock_irqrestore(&base->lock, flags); |
1da177e4 LT |
731 | |
732 | return ret; | |
733 | } | |
734 | ||
2aae4a10 | 735 | /** |
74019224 IM |
736 | * mod_timer_pending - modify a pending timer's timeout |
737 | * @timer: the pending timer to be modified | |
738 | * @expires: new timeout in jiffies | |
1da177e4 | 739 | * |
74019224 IM |
740 | * mod_timer_pending() is the same for pending timers as mod_timer(), |
741 | * but will not re-activate and modify already deleted timers. | |
742 | * | |
743 | * It is useful for unserialized use of timers. | |
1da177e4 | 744 | */ |
74019224 | 745 | int mod_timer_pending(struct timer_list *timer, unsigned long expires) |
1da177e4 | 746 | { |
597d0275 | 747 | return __mod_timer(timer, expires, true, TIMER_NOT_PINNED); |
1da177e4 | 748 | } |
74019224 | 749 | EXPORT_SYMBOL(mod_timer_pending); |
1da177e4 | 750 | |
3bbb9ec9 AV |
751 | /* |
752 | * Decide where to put the timer while taking the slack into account | |
753 | * | |
754 | * Algorithm: | |
755 | * 1) calculate the maximum (absolute) time | |
756 | * 2) calculate the highest bit where the expires and new max are different | |
757 | * 3) use this bit to make a mask | |
758 | * 4) use the bitmask to round down the maximum time, so that all last | |
759 | * bits are zeros | |
760 | */ | |
761 | static inline | |
762 | unsigned long apply_slack(struct timer_list *timer, unsigned long expires) | |
763 | { | |
764 | unsigned long expires_limit, mask; | |
765 | int bit; | |
766 | ||
f00e047e | 767 | expires_limit = expires; |
3bbb9ec9 | 768 | |
8e63d779 | 769 | if (timer->slack >= 0) { |
f00e047e | 770 | expires_limit = expires + timer->slack; |
8e63d779 | 771 | } else { |
2abfb9e1 | 772 | unsigned long now = jiffies; |
3bbb9ec9 | 773 | |
8e63d779 TG |
774 | /* No slack, if already expired else auto slack 0.4% */ |
775 | if (time_after(expires, now)) | |
776 | expires_limit = expires + (expires - now)/256; | |
777 | } | |
3bbb9ec9 | 778 | mask = expires ^ expires_limit; |
3bbb9ec9 AV |
779 | if (mask == 0) |
780 | return expires; | |
781 | ||
782 | bit = find_last_bit(&mask, BITS_PER_LONG); | |
783 | ||
784 | mask = (1 << bit) - 1; | |
785 | ||
786 | expires_limit = expires_limit & ~(mask); | |
787 | ||
788 | return expires_limit; | |
789 | } | |
790 | ||
2aae4a10 | 791 | /** |
1da177e4 LT |
792 | * mod_timer - modify a timer's timeout |
793 | * @timer: the timer to be modified | |
2aae4a10 | 794 | * @expires: new timeout in jiffies |
1da177e4 | 795 | * |
72fd4a35 | 796 | * mod_timer() is a more efficient way to update the expire field of an |
1da177e4 LT |
797 | * active timer (if the timer is inactive it will be activated) |
798 | * | |
799 | * mod_timer(timer, expires) is equivalent to: | |
800 | * | |
801 | * del_timer(timer); timer->expires = expires; add_timer(timer); | |
802 | * | |
803 | * Note that if there are multiple unserialized concurrent users of the | |
804 | * same timer, then mod_timer() is the only safe way to modify the timeout, | |
805 | * since add_timer() cannot modify an already running timer. | |
806 | * | |
807 | * The function returns whether it has modified a pending timer or not. | |
808 | * (ie. mod_timer() of an inactive timer returns 0, mod_timer() of an | |
809 | * active timer returns 1.) | |
810 | */ | |
811 | int mod_timer(struct timer_list *timer, unsigned long expires) | |
812 | { | |
1da177e4 LT |
813 | /* |
814 | * This is a common optimization triggered by the | |
815 | * networking code - if the timer is re-modified | |
816 | * to be the same thing then just return: | |
817 | */ | |
4841158b | 818 | if (timer_pending(timer) && timer->expires == expires) |
1da177e4 LT |
819 | return 1; |
820 | ||
3bbb9ec9 AV |
821 | expires = apply_slack(timer, expires); |
822 | ||
597d0275 | 823 | return __mod_timer(timer, expires, false, TIMER_NOT_PINNED); |
1da177e4 | 824 | } |
1da177e4 LT |
825 | EXPORT_SYMBOL(mod_timer); |
826 | ||
597d0275 AB |
827 | /** |
828 | * mod_timer_pinned - modify a timer's timeout | |
829 | * @timer: the timer to be modified | |
830 | * @expires: new timeout in jiffies | |
831 | * | |
832 | * mod_timer_pinned() is a way to update the expire field of an | |
833 | * active timer (if the timer is inactive it will be activated) | |
834 | * and not allow the timer to be migrated to a different CPU. | |
835 | * | |
836 | * mod_timer_pinned(timer, expires) is equivalent to: | |
837 | * | |
838 | * del_timer(timer); timer->expires = expires; add_timer(timer); | |
839 | */ | |
840 | int mod_timer_pinned(struct timer_list *timer, unsigned long expires) | |
841 | { | |
842 | if (timer->expires == expires && timer_pending(timer)) | |
843 | return 1; | |
844 | ||
845 | return __mod_timer(timer, expires, false, TIMER_PINNED); | |
846 | } | |
847 | EXPORT_SYMBOL(mod_timer_pinned); | |
848 | ||
74019224 IM |
849 | /** |
850 | * add_timer - start a timer | |
851 | * @timer: the timer to be added | |
852 | * | |
853 | * The kernel will do a ->function(->data) callback from the | |
854 | * timer interrupt at the ->expires point in the future. The | |
855 | * current time is 'jiffies'. | |
856 | * | |
857 | * The timer's ->expires, ->function (and if the handler uses it, ->data) | |
858 | * fields must be set prior calling this function. | |
859 | * | |
860 | * Timers with an ->expires field in the past will be executed in the next | |
861 | * timer tick. | |
862 | */ | |
863 | void add_timer(struct timer_list *timer) | |
864 | { | |
865 | BUG_ON(timer_pending(timer)); | |
866 | mod_timer(timer, timer->expires); | |
867 | } | |
868 | EXPORT_SYMBOL(add_timer); | |
869 | ||
870 | /** | |
871 | * add_timer_on - start a timer on a particular CPU | |
872 | * @timer: the timer to be added | |
873 | * @cpu: the CPU to start it on | |
874 | * | |
875 | * This is not very scalable on SMP. Double adds are not possible. | |
876 | */ | |
877 | void add_timer_on(struct timer_list *timer, int cpu) | |
878 | { | |
879 | struct tvec_base *base = per_cpu(tvec_bases, cpu); | |
880 | unsigned long flags; | |
881 | ||
882 | timer_stats_timer_set_start_info(timer); | |
883 | BUG_ON(timer_pending(timer) || !timer->function); | |
884 | spin_lock_irqsave(&base->lock, flags); | |
885 | timer_set_base(timer, base); | |
2b022e3d | 886 | debug_activate(timer, timer->expires); |
97fd9ed4 MS |
887 | if (time_before(timer->expires, base->next_timer) && |
888 | !tbase_get_deferrable(timer->base)) | |
889 | base->next_timer = timer->expires; | |
74019224 IM |
890 | internal_add_timer(base, timer); |
891 | /* | |
892 | * Check whether the other CPU is idle and needs to be | |
893 | * triggered to reevaluate the timer wheel when nohz is | |
894 | * active. We are protected against the other CPU fiddling | |
895 | * with the timer by holding the timer base lock. This also | |
896 | * makes sure that a CPU on the way to idle can not evaluate | |
897 | * the timer wheel. | |
898 | */ | |
899 | wake_up_idle_cpu(cpu); | |
900 | spin_unlock_irqrestore(&base->lock, flags); | |
901 | } | |
a9862e05 | 902 | EXPORT_SYMBOL_GPL(add_timer_on); |
74019224 | 903 | |
2aae4a10 | 904 | /** |
1da177e4 LT |
905 | * del_timer - deactive a timer. |
906 | * @timer: the timer to be deactivated | |
907 | * | |
908 | * del_timer() deactivates a timer - this works on both active and inactive | |
909 | * timers. | |
910 | * | |
911 | * The function returns whether it has deactivated a pending timer or not. | |
912 | * (ie. del_timer() of an inactive timer returns 0, del_timer() of an | |
913 | * active timer returns 1.) | |
914 | */ | |
915 | int del_timer(struct timer_list *timer) | |
916 | { | |
a6fa8e5a | 917 | struct tvec_base *base; |
1da177e4 | 918 | unsigned long flags; |
55c888d6 | 919 | int ret = 0; |
1da177e4 | 920 | |
82f67cd9 | 921 | timer_stats_timer_clear_start_info(timer); |
55c888d6 ON |
922 | if (timer_pending(timer)) { |
923 | base = lock_timer_base(timer, &flags); | |
924 | if (timer_pending(timer)) { | |
925 | detach_timer(timer, 1); | |
97fd9ed4 MS |
926 | if (timer->expires == base->next_timer && |
927 | !tbase_get_deferrable(timer->base)) | |
928 | base->next_timer = base->timer_jiffies; | |
55c888d6 ON |
929 | ret = 1; |
930 | } | |
1da177e4 | 931 | spin_unlock_irqrestore(&base->lock, flags); |
1da177e4 | 932 | } |
1da177e4 | 933 | |
55c888d6 | 934 | return ret; |
1da177e4 | 935 | } |
1da177e4 LT |
936 | EXPORT_SYMBOL(del_timer); |
937 | ||
938 | #ifdef CONFIG_SMP | |
2aae4a10 REB |
939 | /** |
940 | * try_to_del_timer_sync - Try to deactivate a timer | |
941 | * @timer: timer do del | |
942 | * | |
fd450b73 ON |
943 | * This function tries to deactivate a timer. Upon successful (ret >= 0) |
944 | * exit the timer is not queued and the handler is not running on any CPU. | |
945 | * | |
946 | * It must not be called from interrupt contexts. | |
947 | */ | |
948 | int try_to_del_timer_sync(struct timer_list *timer) | |
949 | { | |
a6fa8e5a | 950 | struct tvec_base *base; |
fd450b73 ON |
951 | unsigned long flags; |
952 | int ret = -1; | |
953 | ||
954 | base = lock_timer_base(timer, &flags); | |
955 | ||
956 | if (base->running_timer == timer) | |
957 | goto out; | |
958 | ||
829b6c1e | 959 | timer_stats_timer_clear_start_info(timer); |
fd450b73 ON |
960 | ret = 0; |
961 | if (timer_pending(timer)) { | |
962 | detach_timer(timer, 1); | |
97fd9ed4 MS |
963 | if (timer->expires == base->next_timer && |
964 | !tbase_get_deferrable(timer->base)) | |
965 | base->next_timer = base->timer_jiffies; | |
fd450b73 ON |
966 | ret = 1; |
967 | } | |
968 | out: | |
969 | spin_unlock_irqrestore(&base->lock, flags); | |
970 | ||
971 | return ret; | |
972 | } | |
e19dff1f DH |
973 | EXPORT_SYMBOL(try_to_del_timer_sync); |
974 | ||
2aae4a10 | 975 | /** |
1da177e4 LT |
976 | * del_timer_sync - deactivate a timer and wait for the handler to finish. |
977 | * @timer: the timer to be deactivated | |
978 | * | |
979 | * This function only differs from del_timer() on SMP: besides deactivating | |
980 | * the timer it also makes sure the handler has finished executing on other | |
981 | * CPUs. | |
982 | * | |
72fd4a35 | 983 | * Synchronization rules: Callers must prevent restarting of the timer, |
1da177e4 LT |
984 | * otherwise this function is meaningless. It must not be called from |
985 | * interrupt contexts. The caller must not hold locks which would prevent | |
55c888d6 ON |
986 | * completion of the timer's handler. The timer's handler must not call |
987 | * add_timer_on(). Upon exit the timer is not queued and the handler is | |
988 | * not running on any CPU. | |
1da177e4 LT |
989 | * |
990 | * The function returns whether it has deactivated a pending timer or not. | |
1da177e4 LT |
991 | */ |
992 | int del_timer_sync(struct timer_list *timer) | |
993 | { | |
6f2b9b9a JB |
994 | #ifdef CONFIG_LOCKDEP |
995 | unsigned long flags; | |
996 | ||
997 | local_irq_save(flags); | |
998 | lock_map_acquire(&timer->lockdep_map); | |
999 | lock_map_release(&timer->lockdep_map); | |
1000 | local_irq_restore(flags); | |
1001 | #endif | |
1002 | ||
fd450b73 ON |
1003 | for (;;) { |
1004 | int ret = try_to_del_timer_sync(timer); | |
1005 | if (ret >= 0) | |
1006 | return ret; | |
a0009652 | 1007 | cpu_relax(); |
fd450b73 | 1008 | } |
1da177e4 | 1009 | } |
55c888d6 | 1010 | EXPORT_SYMBOL(del_timer_sync); |
1da177e4 LT |
1011 | #endif |
1012 | ||
a6fa8e5a | 1013 | static int cascade(struct tvec_base *base, struct tvec *tv, int index) |
1da177e4 LT |
1014 | { |
1015 | /* cascade all the timers from tv up one level */ | |
3439dd86 P |
1016 | struct timer_list *timer, *tmp; |
1017 | struct list_head tv_list; | |
1018 | ||
1019 | list_replace_init(tv->vec + index, &tv_list); | |
1da177e4 | 1020 | |
1da177e4 | 1021 | /* |
3439dd86 P |
1022 | * We are removing _all_ timers from the list, so we |
1023 | * don't have to detach them individually. | |
1da177e4 | 1024 | */ |
3439dd86 | 1025 | list_for_each_entry_safe(timer, tmp, &tv_list, entry) { |
6e453a67 | 1026 | BUG_ON(tbase_get_base(timer->base) != base); |
3439dd86 | 1027 | internal_add_timer(base, timer); |
1da177e4 | 1028 | } |
1da177e4 LT |
1029 | |
1030 | return index; | |
1031 | } | |
1032 | ||
576da126 TG |
1033 | static void call_timer_fn(struct timer_list *timer, void (*fn)(unsigned long), |
1034 | unsigned long data) | |
1035 | { | |
1036 | int preempt_count = preempt_count(); | |
1037 | ||
1038 | #ifdef CONFIG_LOCKDEP | |
1039 | /* | |
1040 | * It is permissible to free the timer from inside the | |
1041 | * function that is called from it, this we need to take into | |
1042 | * account for lockdep too. To avoid bogus "held lock freed" | |
1043 | * warnings as well as problems when looking into | |
1044 | * timer->lockdep_map, make a copy and use that here. | |
1045 | */ | |
1046 | struct lockdep_map lockdep_map = timer->lockdep_map; | |
1047 | #endif | |
1048 | /* | |
1049 | * Couple the lock chain with the lock chain at | |
1050 | * del_timer_sync() by acquiring the lock_map around the fn() | |
1051 | * call here and in del_timer_sync(). | |
1052 | */ | |
1053 | lock_map_acquire(&lockdep_map); | |
1054 | ||
1055 | trace_timer_expire_entry(timer); | |
1056 | fn(data); | |
1057 | trace_timer_expire_exit(timer); | |
1058 | ||
1059 | lock_map_release(&lockdep_map); | |
1060 | ||
1061 | if (preempt_count != preempt_count()) { | |
802702e0 TG |
1062 | WARN_ONCE(1, "timer: %pF preempt leak: %08x -> %08x\n", |
1063 | fn, preempt_count, preempt_count()); | |
1064 | /* | |
1065 | * Restore the preempt count. That gives us a decent | |
1066 | * chance to survive and extract information. If the | |
1067 | * callback kept a lock held, bad luck, but not worse | |
1068 | * than the BUG() we had. | |
1069 | */ | |
1070 | preempt_count() = preempt_count; | |
576da126 TG |
1071 | } |
1072 | } | |
1073 | ||
2aae4a10 REB |
1074 | #define INDEX(N) ((base->timer_jiffies >> (TVR_BITS + (N) * TVN_BITS)) & TVN_MASK) |
1075 | ||
1076 | /** | |
1da177e4 LT |
1077 | * __run_timers - run all expired timers (if any) on this CPU. |
1078 | * @base: the timer vector to be processed. | |
1079 | * | |
1080 | * This function cascades all vectors and executes all expired timer | |
1081 | * vectors. | |
1082 | */ | |
a6fa8e5a | 1083 | static inline void __run_timers(struct tvec_base *base) |
1da177e4 LT |
1084 | { |
1085 | struct timer_list *timer; | |
1086 | ||
3691c519 | 1087 | spin_lock_irq(&base->lock); |
1da177e4 | 1088 | while (time_after_eq(jiffies, base->timer_jiffies)) { |
626ab0e6 | 1089 | struct list_head work_list; |
1da177e4 | 1090 | struct list_head *head = &work_list; |
6819457d | 1091 | int index = base->timer_jiffies & TVR_MASK; |
626ab0e6 | 1092 | |
1da177e4 LT |
1093 | /* |
1094 | * Cascade timers: | |
1095 | */ | |
1096 | if (!index && | |
1097 | (!cascade(base, &base->tv2, INDEX(0))) && | |
1098 | (!cascade(base, &base->tv3, INDEX(1))) && | |
1099 | !cascade(base, &base->tv4, INDEX(2))) | |
1100 | cascade(base, &base->tv5, INDEX(3)); | |
626ab0e6 ON |
1101 | ++base->timer_jiffies; |
1102 | list_replace_init(base->tv1.vec + index, &work_list); | |
55c888d6 | 1103 | while (!list_empty(head)) { |
1da177e4 LT |
1104 | void (*fn)(unsigned long); |
1105 | unsigned long data; | |
1106 | ||
b5e61818 | 1107 | timer = list_first_entry(head, struct timer_list,entry); |
6819457d TG |
1108 | fn = timer->function; |
1109 | data = timer->data; | |
1da177e4 | 1110 | |
82f67cd9 IM |
1111 | timer_stats_account_timer(timer); |
1112 | ||
1da177e4 | 1113 | set_running_timer(base, timer); |
55c888d6 | 1114 | detach_timer(timer, 1); |
6f2b9b9a | 1115 | |
3691c519 | 1116 | spin_unlock_irq(&base->lock); |
576da126 | 1117 | call_timer_fn(timer, fn, data); |
3691c519 | 1118 | spin_lock_irq(&base->lock); |
1da177e4 LT |
1119 | } |
1120 | } | |
1121 | set_running_timer(base, NULL); | |
3691c519 | 1122 | spin_unlock_irq(&base->lock); |
1da177e4 LT |
1123 | } |
1124 | ||
ee9c5785 | 1125 | #ifdef CONFIG_NO_HZ |
1da177e4 LT |
1126 | /* |
1127 | * Find out when the next timer event is due to happen. This | |
90cba64a RD |
1128 | * is used on S/390 to stop all activity when a CPU is idle. |
1129 | * This function needs to be called with interrupts disabled. | |
1da177e4 | 1130 | */ |
a6fa8e5a | 1131 | static unsigned long __next_timer_interrupt(struct tvec_base *base) |
1da177e4 | 1132 | { |
1cfd6849 | 1133 | unsigned long timer_jiffies = base->timer_jiffies; |
eaad084b | 1134 | unsigned long expires = timer_jiffies + NEXT_TIMER_MAX_DELTA; |
1cfd6849 | 1135 | int index, slot, array, found = 0; |
1da177e4 | 1136 | struct timer_list *nte; |
a6fa8e5a | 1137 | struct tvec *varray[4]; |
1da177e4 LT |
1138 | |
1139 | /* Look for timer events in tv1. */ | |
1cfd6849 | 1140 | index = slot = timer_jiffies & TVR_MASK; |
1da177e4 | 1141 | do { |
1cfd6849 | 1142 | list_for_each_entry(nte, base->tv1.vec + slot, entry) { |
6819457d TG |
1143 | if (tbase_get_deferrable(nte->base)) |
1144 | continue; | |
6e453a67 | 1145 | |
1cfd6849 | 1146 | found = 1; |
1da177e4 | 1147 | expires = nte->expires; |
1cfd6849 TG |
1148 | /* Look at the cascade bucket(s)? */ |
1149 | if (!index || slot < index) | |
1150 | goto cascade; | |
1151 | return expires; | |
1da177e4 | 1152 | } |
1cfd6849 TG |
1153 | slot = (slot + 1) & TVR_MASK; |
1154 | } while (slot != index); | |
1155 | ||
1156 | cascade: | |
1157 | /* Calculate the next cascade event */ | |
1158 | if (index) | |
1159 | timer_jiffies += TVR_SIZE - index; | |
1160 | timer_jiffies >>= TVR_BITS; | |
1da177e4 LT |
1161 | |
1162 | /* Check tv2-tv5. */ | |
1163 | varray[0] = &base->tv2; | |
1164 | varray[1] = &base->tv3; | |
1165 | varray[2] = &base->tv4; | |
1166 | varray[3] = &base->tv5; | |
1cfd6849 TG |
1167 | |
1168 | for (array = 0; array < 4; array++) { | |
a6fa8e5a | 1169 | struct tvec *varp = varray[array]; |
1cfd6849 TG |
1170 | |
1171 | index = slot = timer_jiffies & TVN_MASK; | |
1da177e4 | 1172 | do { |
1cfd6849 | 1173 | list_for_each_entry(nte, varp->vec + slot, entry) { |
a0419888 JH |
1174 | if (tbase_get_deferrable(nte->base)) |
1175 | continue; | |
1176 | ||
1cfd6849 | 1177 | found = 1; |
1da177e4 LT |
1178 | if (time_before(nte->expires, expires)) |
1179 | expires = nte->expires; | |
1cfd6849 TG |
1180 | } |
1181 | /* | |
1182 | * Do we still search for the first timer or are | |
1183 | * we looking up the cascade buckets ? | |
1184 | */ | |
1185 | if (found) { | |
1186 | /* Look at the cascade bucket(s)? */ | |
1187 | if (!index || slot < index) | |
1188 | break; | |
1189 | return expires; | |
1190 | } | |
1191 | slot = (slot + 1) & TVN_MASK; | |
1192 | } while (slot != index); | |
1193 | ||
1194 | if (index) | |
1195 | timer_jiffies += TVN_SIZE - index; | |
1196 | timer_jiffies >>= TVN_BITS; | |
1da177e4 | 1197 | } |
1cfd6849 TG |
1198 | return expires; |
1199 | } | |
69239749 | 1200 | |
1cfd6849 TG |
1201 | /* |
1202 | * Check, if the next hrtimer event is before the next timer wheel | |
1203 | * event: | |
1204 | */ | |
1205 | static unsigned long cmp_next_hrtimer_event(unsigned long now, | |
1206 | unsigned long expires) | |
1207 | { | |
1208 | ktime_t hr_delta = hrtimer_get_next_event(); | |
1209 | struct timespec tsdelta; | |
9501b6cf | 1210 | unsigned long delta; |
1cfd6849 TG |
1211 | |
1212 | if (hr_delta.tv64 == KTIME_MAX) | |
1213 | return expires; | |
0662b713 | 1214 | |
9501b6cf TG |
1215 | /* |
1216 | * Expired timer available, let it expire in the next tick | |
1217 | */ | |
1218 | if (hr_delta.tv64 <= 0) | |
1219 | return now + 1; | |
69239749 | 1220 | |
1cfd6849 | 1221 | tsdelta = ktime_to_timespec(hr_delta); |
9501b6cf | 1222 | delta = timespec_to_jiffies(&tsdelta); |
eaad084b TG |
1223 | |
1224 | /* | |
1225 | * Limit the delta to the max value, which is checked in | |
1226 | * tick_nohz_stop_sched_tick(): | |
1227 | */ | |
1228 | if (delta > NEXT_TIMER_MAX_DELTA) | |
1229 | delta = NEXT_TIMER_MAX_DELTA; | |
1230 | ||
9501b6cf TG |
1231 | /* |
1232 | * Take rounding errors in to account and make sure, that it | |
1233 | * expires in the next tick. Otherwise we go into an endless | |
1234 | * ping pong due to tick_nohz_stop_sched_tick() retriggering | |
1235 | * the timer softirq | |
1236 | */ | |
1237 | if (delta < 1) | |
1238 | delta = 1; | |
1239 | now += delta; | |
1cfd6849 TG |
1240 | if (time_before(now, expires)) |
1241 | return now; | |
1da177e4 LT |
1242 | return expires; |
1243 | } | |
1cfd6849 TG |
1244 | |
1245 | /** | |
8dce39c2 | 1246 | * get_next_timer_interrupt - return the jiffy of the next pending timer |
05fb6bf0 | 1247 | * @now: current time (in jiffies) |
1cfd6849 | 1248 | */ |
fd064b9b | 1249 | unsigned long get_next_timer_interrupt(unsigned long now) |
1cfd6849 | 1250 | { |
a6fa8e5a | 1251 | struct tvec_base *base = __get_cpu_var(tvec_bases); |
fd064b9b | 1252 | unsigned long expires; |
1cfd6849 TG |
1253 | |
1254 | spin_lock(&base->lock); | |
97fd9ed4 MS |
1255 | if (time_before_eq(base->next_timer, base->timer_jiffies)) |
1256 | base->next_timer = __next_timer_interrupt(base); | |
1257 | expires = base->next_timer; | |
1cfd6849 TG |
1258 | spin_unlock(&base->lock); |
1259 | ||
1260 | if (time_before_eq(expires, now)) | |
1261 | return now; | |
1262 | ||
1263 | return cmp_next_hrtimer_event(now, expires); | |
1264 | } | |
1da177e4 LT |
1265 | #endif |
1266 | ||
1da177e4 | 1267 | /* |
5b4db0c2 | 1268 | * Called from the timer interrupt handler to charge one tick to the current |
1da177e4 LT |
1269 | * process. user_tick is 1 if the tick is user time, 0 for system. |
1270 | */ | |
1271 | void update_process_times(int user_tick) | |
1272 | { | |
1273 | struct task_struct *p = current; | |
1274 | int cpu = smp_processor_id(); | |
1275 | ||
1276 | /* Note: this timer irq context must be accounted for as well. */ | |
fa13a5a1 | 1277 | account_process_tick(p, user_tick); |
1da177e4 | 1278 | run_local_timers(); |
a157229c | 1279 | rcu_check_callbacks(cpu, user_tick); |
b845b517 | 1280 | printk_tick(); |
fe432200 | 1281 | perf_event_do_pending(); |
1da177e4 | 1282 | scheduler_tick(); |
6819457d | 1283 | run_posix_cpu_timers(p); |
1da177e4 LT |
1284 | } |
1285 | ||
1da177e4 LT |
1286 | /* |
1287 | * This function runs timers and the timer-tq in bottom half context. | |
1288 | */ | |
1289 | static void run_timer_softirq(struct softirq_action *h) | |
1290 | { | |
a6fa8e5a | 1291 | struct tvec_base *base = __get_cpu_var(tvec_bases); |
1da177e4 | 1292 | |
d3d74453 | 1293 | hrtimer_run_pending(); |
82f67cd9 | 1294 | |
1da177e4 LT |
1295 | if (time_after_eq(jiffies, base->timer_jiffies)) |
1296 | __run_timers(base); | |
1297 | } | |
1298 | ||
1299 | /* | |
1300 | * Called by the local, per-CPU timer interrupt on SMP. | |
1301 | */ | |
1302 | void run_local_timers(void) | |
1303 | { | |
d3d74453 | 1304 | hrtimer_run_queues(); |
1da177e4 LT |
1305 | raise_softirq(TIMER_SOFTIRQ); |
1306 | } | |
1307 | ||
1da177e4 LT |
1308 | /* |
1309 | * The 64-bit jiffies value is not atomic - you MUST NOT read it | |
1310 | * without sampling the sequence number in xtime_lock. | |
1311 | * jiffies is defined in the linker script... | |
1312 | */ | |
1313 | ||
3171a030 | 1314 | void do_timer(unsigned long ticks) |
1da177e4 | 1315 | { |
3171a030 | 1316 | jiffies_64 += ticks; |
dce48a84 TG |
1317 | update_wall_time(); |
1318 | calc_global_load(); | |
1da177e4 LT |
1319 | } |
1320 | ||
1321 | #ifdef __ARCH_WANT_SYS_ALARM | |
1322 | ||
1323 | /* | |
1324 | * For backwards compatibility? This can be done in libc so Alpha | |
1325 | * and all newer ports shouldn't need it. | |
1326 | */ | |
58fd3aa2 | 1327 | SYSCALL_DEFINE1(alarm, unsigned int, seconds) |
1da177e4 | 1328 | { |
c08b8a49 | 1329 | return alarm_setitimer(seconds); |
1da177e4 LT |
1330 | } |
1331 | ||
1332 | #endif | |
1333 | ||
1334 | #ifndef __alpha__ | |
1335 | ||
1336 | /* | |
1337 | * The Alpha uses getxpid, getxuid, and getxgid instead. Maybe this | |
1338 | * should be moved into arch/i386 instead? | |
1339 | */ | |
1340 | ||
1341 | /** | |
1342 | * sys_getpid - return the thread group id of the current process | |
1343 | * | |
1344 | * Note, despite the name, this returns the tgid not the pid. The tgid and | |
1345 | * the pid are identical unless CLONE_THREAD was specified on clone() in | |
1346 | * which case the tgid is the same in all threads of the same group. | |
1347 | * | |
1348 | * This is SMP safe as current->tgid does not change. | |
1349 | */ | |
58fd3aa2 | 1350 | SYSCALL_DEFINE0(getpid) |
1da177e4 | 1351 | { |
b488893a | 1352 | return task_tgid_vnr(current); |
1da177e4 LT |
1353 | } |
1354 | ||
1355 | /* | |
6997a6fa KK |
1356 | * Accessing ->real_parent is not SMP-safe, it could |
1357 | * change from under us. However, we can use a stale | |
1358 | * value of ->real_parent under rcu_read_lock(), see | |
1359 | * release_task()->call_rcu(delayed_put_task_struct). | |
1da177e4 | 1360 | */ |
dbf040d9 | 1361 | SYSCALL_DEFINE0(getppid) |
1da177e4 LT |
1362 | { |
1363 | int pid; | |
1da177e4 | 1364 | |
6997a6fa | 1365 | rcu_read_lock(); |
6c5f3e7b | 1366 | pid = task_tgid_vnr(current->real_parent); |
6997a6fa | 1367 | rcu_read_unlock(); |
1da177e4 | 1368 | |
1da177e4 LT |
1369 | return pid; |
1370 | } | |
1371 | ||
dbf040d9 | 1372 | SYSCALL_DEFINE0(getuid) |
1da177e4 LT |
1373 | { |
1374 | /* Only we change this so SMP safe */ | |
76aac0e9 | 1375 | return current_uid(); |
1da177e4 LT |
1376 | } |
1377 | ||
dbf040d9 | 1378 | SYSCALL_DEFINE0(geteuid) |
1da177e4 LT |
1379 | { |
1380 | /* Only we change this so SMP safe */ | |
76aac0e9 | 1381 | return current_euid(); |
1da177e4 LT |
1382 | } |
1383 | ||
dbf040d9 | 1384 | SYSCALL_DEFINE0(getgid) |
1da177e4 LT |
1385 | { |
1386 | /* Only we change this so SMP safe */ | |
76aac0e9 | 1387 | return current_gid(); |
1da177e4 LT |
1388 | } |
1389 | ||
dbf040d9 | 1390 | SYSCALL_DEFINE0(getegid) |
1da177e4 LT |
1391 | { |
1392 | /* Only we change this so SMP safe */ | |
76aac0e9 | 1393 | return current_egid(); |
1da177e4 LT |
1394 | } |
1395 | ||
1396 | #endif | |
1397 | ||
1398 | static void process_timeout(unsigned long __data) | |
1399 | { | |
36c8b586 | 1400 | wake_up_process((struct task_struct *)__data); |
1da177e4 LT |
1401 | } |
1402 | ||
1403 | /** | |
1404 | * schedule_timeout - sleep until timeout | |
1405 | * @timeout: timeout value in jiffies | |
1406 | * | |
1407 | * Make the current task sleep until @timeout jiffies have | |
1408 | * elapsed. The routine will return immediately unless | |
1409 | * the current task state has been set (see set_current_state()). | |
1410 | * | |
1411 | * You can set the task state as follows - | |
1412 | * | |
1413 | * %TASK_UNINTERRUPTIBLE - at least @timeout jiffies are guaranteed to | |
1414 | * pass before the routine returns. The routine will return 0 | |
1415 | * | |
1416 | * %TASK_INTERRUPTIBLE - the routine may return early if a signal is | |
1417 | * delivered to the current task. In this case the remaining time | |
1418 | * in jiffies will be returned, or 0 if the timer expired in time | |
1419 | * | |
1420 | * The current task state is guaranteed to be TASK_RUNNING when this | |
1421 | * routine returns. | |
1422 | * | |
1423 | * Specifying a @timeout value of %MAX_SCHEDULE_TIMEOUT will schedule | |
1424 | * the CPU away without a bound on the timeout. In this case the return | |
1425 | * value will be %MAX_SCHEDULE_TIMEOUT. | |
1426 | * | |
1427 | * In all cases the return value is guaranteed to be non-negative. | |
1428 | */ | |
7ad5b3a5 | 1429 | signed long __sched schedule_timeout(signed long timeout) |
1da177e4 LT |
1430 | { |
1431 | struct timer_list timer; | |
1432 | unsigned long expire; | |
1433 | ||
1434 | switch (timeout) | |
1435 | { | |
1436 | case MAX_SCHEDULE_TIMEOUT: | |
1437 | /* | |
1438 | * These two special cases are useful to be comfortable | |
1439 | * in the caller. Nothing more. We could take | |
1440 | * MAX_SCHEDULE_TIMEOUT from one of the negative value | |
1441 | * but I' d like to return a valid offset (>=0) to allow | |
1442 | * the caller to do everything it want with the retval. | |
1443 | */ | |
1444 | schedule(); | |
1445 | goto out; | |
1446 | default: | |
1447 | /* | |
1448 | * Another bit of PARANOID. Note that the retval will be | |
1449 | * 0 since no piece of kernel is supposed to do a check | |
1450 | * for a negative retval of schedule_timeout() (since it | |
1451 | * should never happens anyway). You just have the printk() | |
1452 | * that will tell you if something is gone wrong and where. | |
1453 | */ | |
5b149bcc | 1454 | if (timeout < 0) { |
1da177e4 | 1455 | printk(KERN_ERR "schedule_timeout: wrong timeout " |
5b149bcc AM |
1456 | "value %lx\n", timeout); |
1457 | dump_stack(); | |
1da177e4 LT |
1458 | current->state = TASK_RUNNING; |
1459 | goto out; | |
1460 | } | |
1461 | } | |
1462 | ||
1463 | expire = timeout + jiffies; | |
1464 | ||
c6f3a97f | 1465 | setup_timer_on_stack(&timer, process_timeout, (unsigned long)current); |
597d0275 | 1466 | __mod_timer(&timer, expire, false, TIMER_NOT_PINNED); |
1da177e4 LT |
1467 | schedule(); |
1468 | del_singleshot_timer_sync(&timer); | |
1469 | ||
c6f3a97f TG |
1470 | /* Remove the timer from the object tracker */ |
1471 | destroy_timer_on_stack(&timer); | |
1472 | ||
1da177e4 LT |
1473 | timeout = expire - jiffies; |
1474 | ||
1475 | out: | |
1476 | return timeout < 0 ? 0 : timeout; | |
1477 | } | |
1da177e4 LT |
1478 | EXPORT_SYMBOL(schedule_timeout); |
1479 | ||
8a1c1757 AM |
1480 | /* |
1481 | * We can use __set_current_state() here because schedule_timeout() calls | |
1482 | * schedule() unconditionally. | |
1483 | */ | |
64ed93a2 NA |
1484 | signed long __sched schedule_timeout_interruptible(signed long timeout) |
1485 | { | |
a5a0d52c AM |
1486 | __set_current_state(TASK_INTERRUPTIBLE); |
1487 | return schedule_timeout(timeout); | |
64ed93a2 NA |
1488 | } |
1489 | EXPORT_SYMBOL(schedule_timeout_interruptible); | |
1490 | ||
294d5cc2 MW |
1491 | signed long __sched schedule_timeout_killable(signed long timeout) |
1492 | { | |
1493 | __set_current_state(TASK_KILLABLE); | |
1494 | return schedule_timeout(timeout); | |
1495 | } | |
1496 | EXPORT_SYMBOL(schedule_timeout_killable); | |
1497 | ||
64ed93a2 NA |
1498 | signed long __sched schedule_timeout_uninterruptible(signed long timeout) |
1499 | { | |
a5a0d52c AM |
1500 | __set_current_state(TASK_UNINTERRUPTIBLE); |
1501 | return schedule_timeout(timeout); | |
64ed93a2 NA |
1502 | } |
1503 | EXPORT_SYMBOL(schedule_timeout_uninterruptible); | |
1504 | ||
1da177e4 | 1505 | /* Thread ID - the internal kernel "pid" */ |
58fd3aa2 | 1506 | SYSCALL_DEFINE0(gettid) |
1da177e4 | 1507 | { |
b488893a | 1508 | return task_pid_vnr(current); |
1da177e4 LT |
1509 | } |
1510 | ||
2aae4a10 | 1511 | /** |
d4d23add | 1512 | * do_sysinfo - fill in sysinfo struct |
2aae4a10 | 1513 | * @info: pointer to buffer to fill |
6819457d | 1514 | */ |
d4d23add | 1515 | int do_sysinfo(struct sysinfo *info) |
1da177e4 | 1516 | { |
1da177e4 LT |
1517 | unsigned long mem_total, sav_total; |
1518 | unsigned int mem_unit, bitcount; | |
2d02494f | 1519 | struct timespec tp; |
1da177e4 | 1520 | |
d4d23add | 1521 | memset(info, 0, sizeof(struct sysinfo)); |
1da177e4 | 1522 | |
2d02494f TG |
1523 | ktime_get_ts(&tp); |
1524 | monotonic_to_bootbased(&tp); | |
1525 | info->uptime = tp.tv_sec + (tp.tv_nsec ? 1 : 0); | |
1da177e4 | 1526 | |
2d02494f | 1527 | get_avenrun(info->loads, 0, SI_LOAD_SHIFT - FSHIFT); |
1da177e4 | 1528 | |
2d02494f | 1529 | info->procs = nr_threads; |
1da177e4 | 1530 | |
d4d23add KM |
1531 | si_meminfo(info); |
1532 | si_swapinfo(info); | |
1da177e4 LT |
1533 | |
1534 | /* | |
1535 | * If the sum of all the available memory (i.e. ram + swap) | |
1536 | * is less than can be stored in a 32 bit unsigned long then | |
1537 | * we can be binary compatible with 2.2.x kernels. If not, | |
1538 | * well, in that case 2.2.x was broken anyways... | |
1539 | * | |
1540 | * -Erik Andersen <andersee@debian.org> | |
1541 | */ | |
1542 | ||
d4d23add KM |
1543 | mem_total = info->totalram + info->totalswap; |
1544 | if (mem_total < info->totalram || mem_total < info->totalswap) | |
1da177e4 LT |
1545 | goto out; |
1546 | bitcount = 0; | |
d4d23add | 1547 | mem_unit = info->mem_unit; |
1da177e4 LT |
1548 | while (mem_unit > 1) { |
1549 | bitcount++; | |
1550 | mem_unit >>= 1; | |
1551 | sav_total = mem_total; | |
1552 | mem_total <<= 1; | |
1553 | if (mem_total < sav_total) | |
1554 | goto out; | |
1555 | } | |
1556 | ||
1557 | /* | |
1558 | * If mem_total did not overflow, multiply all memory values by | |
d4d23add | 1559 | * info->mem_unit and set it to 1. This leaves things compatible |
1da177e4 LT |
1560 | * with 2.2.x, and also retains compatibility with earlier 2.4.x |
1561 | * kernels... | |
1562 | */ | |
1563 | ||
d4d23add KM |
1564 | info->mem_unit = 1; |
1565 | info->totalram <<= bitcount; | |
1566 | info->freeram <<= bitcount; | |
1567 | info->sharedram <<= bitcount; | |
1568 | info->bufferram <<= bitcount; | |
1569 | info->totalswap <<= bitcount; | |
1570 | info->freeswap <<= bitcount; | |
1571 | info->totalhigh <<= bitcount; | |
1572 | info->freehigh <<= bitcount; | |
1573 | ||
1574 | out: | |
1575 | return 0; | |
1576 | } | |
1577 | ||
1e7bfb21 | 1578 | SYSCALL_DEFINE1(sysinfo, struct sysinfo __user *, info) |
d4d23add KM |
1579 | { |
1580 | struct sysinfo val; | |
1581 | ||
1582 | do_sysinfo(&val); | |
1da177e4 | 1583 | |
1da177e4 LT |
1584 | if (copy_to_user(info, &val, sizeof(struct sysinfo))) |
1585 | return -EFAULT; | |
1586 | ||
1587 | return 0; | |
1588 | } | |
1589 | ||
b4be6258 | 1590 | static int __cpuinit init_timers_cpu(int cpu) |
1da177e4 LT |
1591 | { |
1592 | int j; | |
a6fa8e5a | 1593 | struct tvec_base *base; |
b4be6258 | 1594 | static char __cpuinitdata tvec_base_done[NR_CPUS]; |
55c888d6 | 1595 | |
ba6edfcd | 1596 | if (!tvec_base_done[cpu]) { |
a4a6198b JB |
1597 | static char boot_done; |
1598 | ||
a4a6198b | 1599 | if (boot_done) { |
ba6edfcd AM |
1600 | /* |
1601 | * The APs use this path later in boot | |
1602 | */ | |
94f6030c CL |
1603 | base = kmalloc_node(sizeof(*base), |
1604 | GFP_KERNEL | __GFP_ZERO, | |
a4a6198b JB |
1605 | cpu_to_node(cpu)); |
1606 | if (!base) | |
1607 | return -ENOMEM; | |
6e453a67 VP |
1608 | |
1609 | /* Make sure that tvec_base is 2 byte aligned */ | |
1610 | if (tbase_get_deferrable(base)) { | |
1611 | WARN_ON(1); | |
1612 | kfree(base); | |
1613 | return -ENOMEM; | |
1614 | } | |
ba6edfcd | 1615 | per_cpu(tvec_bases, cpu) = base; |
a4a6198b | 1616 | } else { |
ba6edfcd AM |
1617 | /* |
1618 | * This is for the boot CPU - we use compile-time | |
1619 | * static initialisation because per-cpu memory isn't | |
1620 | * ready yet and because the memory allocators are not | |
1621 | * initialised either. | |
1622 | */ | |
a4a6198b | 1623 | boot_done = 1; |
ba6edfcd | 1624 | base = &boot_tvec_bases; |
a4a6198b | 1625 | } |
ba6edfcd AM |
1626 | tvec_base_done[cpu] = 1; |
1627 | } else { | |
1628 | base = per_cpu(tvec_bases, cpu); | |
a4a6198b | 1629 | } |
ba6edfcd | 1630 | |
3691c519 | 1631 | spin_lock_init(&base->lock); |
d730e882 | 1632 | |
1da177e4 LT |
1633 | for (j = 0; j < TVN_SIZE; j++) { |
1634 | INIT_LIST_HEAD(base->tv5.vec + j); | |
1635 | INIT_LIST_HEAD(base->tv4.vec + j); | |
1636 | INIT_LIST_HEAD(base->tv3.vec + j); | |
1637 | INIT_LIST_HEAD(base->tv2.vec + j); | |
1638 | } | |
1639 | for (j = 0; j < TVR_SIZE; j++) | |
1640 | INIT_LIST_HEAD(base->tv1.vec + j); | |
1641 | ||
1642 | base->timer_jiffies = jiffies; | |
97fd9ed4 | 1643 | base->next_timer = base->timer_jiffies; |
a4a6198b | 1644 | return 0; |
1da177e4 LT |
1645 | } |
1646 | ||
1647 | #ifdef CONFIG_HOTPLUG_CPU | |
a6fa8e5a | 1648 | static void migrate_timer_list(struct tvec_base *new_base, struct list_head *head) |
1da177e4 LT |
1649 | { |
1650 | struct timer_list *timer; | |
1651 | ||
1652 | while (!list_empty(head)) { | |
b5e61818 | 1653 | timer = list_first_entry(head, struct timer_list, entry); |
55c888d6 | 1654 | detach_timer(timer, 0); |
6e453a67 | 1655 | timer_set_base(timer, new_base); |
97fd9ed4 MS |
1656 | if (time_before(timer->expires, new_base->next_timer) && |
1657 | !tbase_get_deferrable(timer->base)) | |
1658 | new_base->next_timer = timer->expires; | |
1da177e4 | 1659 | internal_add_timer(new_base, timer); |
1da177e4 | 1660 | } |
1da177e4 LT |
1661 | } |
1662 | ||
48ccf3da | 1663 | static void __cpuinit migrate_timers(int cpu) |
1da177e4 | 1664 | { |
a6fa8e5a PM |
1665 | struct tvec_base *old_base; |
1666 | struct tvec_base *new_base; | |
1da177e4 LT |
1667 | int i; |
1668 | ||
1669 | BUG_ON(cpu_online(cpu)); | |
a4a6198b JB |
1670 | old_base = per_cpu(tvec_bases, cpu); |
1671 | new_base = get_cpu_var(tvec_bases); | |
d82f0b0f ON |
1672 | /* |
1673 | * The caller is globally serialized and nobody else | |
1674 | * takes two locks at once, deadlock is not possible. | |
1675 | */ | |
1676 | spin_lock_irq(&new_base->lock); | |
0d180406 | 1677 | spin_lock_nested(&old_base->lock, SINGLE_DEPTH_NESTING); |
3691c519 ON |
1678 | |
1679 | BUG_ON(old_base->running_timer); | |
1da177e4 | 1680 | |
1da177e4 | 1681 | for (i = 0; i < TVR_SIZE; i++) |
55c888d6 ON |
1682 | migrate_timer_list(new_base, old_base->tv1.vec + i); |
1683 | for (i = 0; i < TVN_SIZE; i++) { | |
1684 | migrate_timer_list(new_base, old_base->tv2.vec + i); | |
1685 | migrate_timer_list(new_base, old_base->tv3.vec + i); | |
1686 | migrate_timer_list(new_base, old_base->tv4.vec + i); | |
1687 | migrate_timer_list(new_base, old_base->tv5.vec + i); | |
1688 | } | |
1689 | ||
0d180406 | 1690 | spin_unlock(&old_base->lock); |
d82f0b0f | 1691 | spin_unlock_irq(&new_base->lock); |
1da177e4 | 1692 | put_cpu_var(tvec_bases); |
1da177e4 LT |
1693 | } |
1694 | #endif /* CONFIG_HOTPLUG_CPU */ | |
1695 | ||
8c78f307 | 1696 | static int __cpuinit timer_cpu_notify(struct notifier_block *self, |
1da177e4 LT |
1697 | unsigned long action, void *hcpu) |
1698 | { | |
1699 | long cpu = (long)hcpu; | |
80b5184c AM |
1700 | int err; |
1701 | ||
1da177e4 LT |
1702 | switch(action) { |
1703 | case CPU_UP_PREPARE: | |
8bb78442 | 1704 | case CPU_UP_PREPARE_FROZEN: |
80b5184c AM |
1705 | err = init_timers_cpu(cpu); |
1706 | if (err < 0) | |
1707 | return notifier_from_errno(err); | |
1da177e4 LT |
1708 | break; |
1709 | #ifdef CONFIG_HOTPLUG_CPU | |
1710 | case CPU_DEAD: | |
8bb78442 | 1711 | case CPU_DEAD_FROZEN: |
1da177e4 LT |
1712 | migrate_timers(cpu); |
1713 | break; | |
1714 | #endif | |
1715 | default: | |
1716 | break; | |
1717 | } | |
1718 | return NOTIFY_OK; | |
1719 | } | |
1720 | ||
8c78f307 | 1721 | static struct notifier_block __cpuinitdata timers_nb = { |
1da177e4 LT |
1722 | .notifier_call = timer_cpu_notify, |
1723 | }; | |
1724 | ||
1725 | ||
1726 | void __init init_timers(void) | |
1727 | { | |
07dccf33 | 1728 | int err = timer_cpu_notify(&timers_nb, (unsigned long)CPU_UP_PREPARE, |
1da177e4 | 1729 | (void *)(long)smp_processor_id()); |
07dccf33 | 1730 | |
82f67cd9 IM |
1731 | init_timer_stats(); |
1732 | ||
9e506f7a | 1733 | BUG_ON(err != NOTIFY_OK); |
1da177e4 | 1734 | register_cpu_notifier(&timers_nb); |
962cf36c | 1735 | open_softirq(TIMER_SOFTIRQ, run_timer_softirq); |
1da177e4 LT |
1736 | } |
1737 | ||
1da177e4 LT |
1738 | /** |
1739 | * msleep - sleep safely even with waitqueue interruptions | |
1740 | * @msecs: Time in milliseconds to sleep for | |
1741 | */ | |
1742 | void msleep(unsigned int msecs) | |
1743 | { | |
1744 | unsigned long timeout = msecs_to_jiffies(msecs) + 1; | |
1745 | ||
75bcc8c5 NA |
1746 | while (timeout) |
1747 | timeout = schedule_timeout_uninterruptible(timeout); | |
1da177e4 LT |
1748 | } |
1749 | ||
1750 | EXPORT_SYMBOL(msleep); | |
1751 | ||
1752 | /** | |
96ec3efd | 1753 | * msleep_interruptible - sleep waiting for signals |
1da177e4 LT |
1754 | * @msecs: Time in milliseconds to sleep for |
1755 | */ | |
1756 | unsigned long msleep_interruptible(unsigned int msecs) | |
1757 | { | |
1758 | unsigned long timeout = msecs_to_jiffies(msecs) + 1; | |
1759 | ||
75bcc8c5 NA |
1760 | while (timeout && !signal_pending(current)) |
1761 | timeout = schedule_timeout_interruptible(timeout); | |
1da177e4 LT |
1762 | return jiffies_to_msecs(timeout); |
1763 | } | |
1764 | ||
1765 | EXPORT_SYMBOL(msleep_interruptible); | |
5e7f5a17 PP |
1766 | |
1767 | static int __sched do_usleep_range(unsigned long min, unsigned long max) | |
1768 | { | |
1769 | ktime_t kmin; | |
1770 | unsigned long delta; | |
1771 | ||
1772 | kmin = ktime_set(0, min * NSEC_PER_USEC); | |
1773 | delta = (max - min) * NSEC_PER_USEC; | |
1774 | return schedule_hrtimeout_range(&kmin, delta, HRTIMER_MODE_REL); | |
1775 | } | |
1776 | ||
1777 | /** | |
1778 | * usleep_range - Drop in replacement for udelay where wakeup is flexible | |
1779 | * @min: Minimum time in usecs to sleep | |
1780 | * @max: Maximum time in usecs to sleep | |
1781 | */ | |
1782 | void usleep_range(unsigned long min, unsigned long max) | |
1783 | { | |
1784 | __set_current_state(TASK_UNINTERRUPTIBLE); | |
1785 | do_usleep_range(min, max); | |
1786 | } | |
1787 | EXPORT_SYMBOL(usleep_range); |