]>
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 |
6e453a67 VP |
93 | * base in timer_list is guaranteed to be zero. Use the LSB for |
94 | * the new flag to indicate whether the timer is deferrable | |
95 | */ | |
96 | #define TBASE_DEFERRABLE_FLAG (0x1) | |
97 | ||
98 | /* Functions below help us manage 'deferrable' flag */ | |
a6fa8e5a | 99 | static inline unsigned int tbase_get_deferrable(struct tvec_base *base) |
6e453a67 | 100 | { |
e9910846 | 101 | return ((unsigned int)(unsigned long)base & TBASE_DEFERRABLE_FLAG); |
6e453a67 VP |
102 | } |
103 | ||
a6fa8e5a | 104 | static inline struct tvec_base *tbase_get_base(struct tvec_base *base) |
6e453a67 | 105 | { |
a6fa8e5a | 106 | return ((struct tvec_base *)((unsigned long)base & ~TBASE_DEFERRABLE_FLAG)); |
6e453a67 VP |
107 | } |
108 | ||
109 | static inline void timer_set_deferrable(struct timer_list *timer) | |
110 | { | |
a6fa8e5a | 111 | timer->base = ((struct tvec_base *)((unsigned long)(timer->base) | |
6819457d | 112 | TBASE_DEFERRABLE_FLAG)); |
6e453a67 VP |
113 | } |
114 | ||
115 | static inline void | |
a6fa8e5a | 116 | timer_set_base(struct timer_list *timer, struct tvec_base *new_base) |
6e453a67 | 117 | { |
a6fa8e5a | 118 | timer->base = (struct tvec_base *)((unsigned long)(new_base) | |
6819457d | 119 | tbase_get_deferrable(timer->base)); |
6e453a67 VP |
120 | } |
121 | ||
9c133c46 AS |
122 | static unsigned long round_jiffies_common(unsigned long j, int cpu, |
123 | bool force_up) | |
4c36a5de AV |
124 | { |
125 | int rem; | |
126 | unsigned long original = j; | |
127 | ||
128 | /* | |
129 | * We don't want all cpus firing their timers at once hitting the | |
130 | * same lock or cachelines, so we skew each extra cpu with an extra | |
131 | * 3 jiffies. This 3 jiffies came originally from the mm/ code which | |
132 | * already did this. | |
133 | * The skew is done by adding 3*cpunr, then round, then subtract this | |
134 | * extra offset again. | |
135 | */ | |
136 | j += cpu * 3; | |
137 | ||
138 | rem = j % HZ; | |
139 | ||
140 | /* | |
141 | * If the target jiffie is just after a whole second (which can happen | |
142 | * due to delays of the timer irq, long irq off times etc etc) then | |
143 | * we should round down to the whole second, not up. Use 1/4th second | |
144 | * as cutoff for this rounding as an extreme upper bound for this. | |
9c133c46 | 145 | * But never round down if @force_up is set. |
4c36a5de | 146 | */ |
9c133c46 | 147 | if (rem < HZ/4 && !force_up) /* round down */ |
4c36a5de AV |
148 | j = j - rem; |
149 | else /* round up */ | |
150 | j = j - rem + HZ; | |
151 | ||
152 | /* now that we have rounded, subtract the extra skew again */ | |
153 | j -= cpu * 3; | |
154 | ||
155 | if (j <= jiffies) /* rounding ate our timeout entirely; */ | |
156 | return original; | |
157 | return j; | |
158 | } | |
9c133c46 AS |
159 | |
160 | /** | |
161 | * __round_jiffies - function to round jiffies to a full second | |
162 | * @j: the time in (absolute) jiffies that should be rounded | |
163 | * @cpu: the processor number on which the timeout will happen | |
164 | * | |
165 | * __round_jiffies() rounds an absolute time in the future (in jiffies) | |
166 | * up or down to (approximately) full seconds. This is useful for timers | |
167 | * for which the exact time they fire does not matter too much, as long as | |
168 | * they fire approximately every X seconds. | |
169 | * | |
170 | * By rounding these timers to whole seconds, all such timers will fire | |
171 | * at the same time, rather than at various times spread out. The goal | |
172 | * of this is to have the CPU wake up less, which saves power. | |
173 | * | |
174 | * The exact rounding is skewed for each processor to avoid all | |
175 | * processors firing at the exact same time, which could lead | |
176 | * to lock contention or spurious cache line bouncing. | |
177 | * | |
178 | * The return value is the rounded version of the @j parameter. | |
179 | */ | |
180 | unsigned long __round_jiffies(unsigned long j, int cpu) | |
181 | { | |
182 | return round_jiffies_common(j, cpu, false); | |
183 | } | |
4c36a5de AV |
184 | EXPORT_SYMBOL_GPL(__round_jiffies); |
185 | ||
186 | /** | |
187 | * __round_jiffies_relative - function to round jiffies to a full second | |
188 | * @j: the time in (relative) jiffies that should be rounded | |
189 | * @cpu: the processor number on which the timeout will happen | |
190 | * | |
72fd4a35 | 191 | * __round_jiffies_relative() rounds a time delta in the future (in jiffies) |
4c36a5de AV |
192 | * up or down to (approximately) full seconds. This is useful for timers |
193 | * for which the exact time they fire does not matter too much, as long as | |
194 | * they fire approximately every X seconds. | |
195 | * | |
196 | * By rounding these timers to whole seconds, all such timers will fire | |
197 | * at the same time, rather than at various times spread out. The goal | |
198 | * of this is to have the CPU wake up less, which saves power. | |
199 | * | |
200 | * The exact rounding is skewed for each processor to avoid all | |
201 | * processors firing at the exact same time, which could lead | |
202 | * to lock contention or spurious cache line bouncing. | |
203 | * | |
72fd4a35 | 204 | * The return value is the rounded version of the @j parameter. |
4c36a5de AV |
205 | */ |
206 | unsigned long __round_jiffies_relative(unsigned long j, int cpu) | |
207 | { | |
9c133c46 AS |
208 | unsigned long j0 = jiffies; |
209 | ||
210 | /* Use j0 because jiffies might change while we run */ | |
211 | return round_jiffies_common(j + j0, cpu, false) - j0; | |
4c36a5de AV |
212 | } |
213 | EXPORT_SYMBOL_GPL(__round_jiffies_relative); | |
214 | ||
215 | /** | |
216 | * round_jiffies - function to round jiffies to a full second | |
217 | * @j: the time in (absolute) jiffies that should be rounded | |
218 | * | |
72fd4a35 | 219 | * round_jiffies() rounds an absolute time in the future (in jiffies) |
4c36a5de AV |
220 | * up or down to (approximately) full seconds. This is useful for timers |
221 | * for which the exact time they fire does not matter too much, as long as | |
222 | * they fire approximately every X seconds. | |
223 | * | |
224 | * By rounding these timers to whole seconds, all such timers will fire | |
225 | * at the same time, rather than at various times spread out. The goal | |
226 | * of this is to have the CPU wake up less, which saves power. | |
227 | * | |
72fd4a35 | 228 | * The return value is the rounded version of the @j parameter. |
4c36a5de AV |
229 | */ |
230 | unsigned long round_jiffies(unsigned long j) | |
231 | { | |
9c133c46 | 232 | return round_jiffies_common(j, raw_smp_processor_id(), false); |
4c36a5de AV |
233 | } |
234 | EXPORT_SYMBOL_GPL(round_jiffies); | |
235 | ||
236 | /** | |
237 | * round_jiffies_relative - function to round jiffies to a full second | |
238 | * @j: the time in (relative) jiffies that should be rounded | |
239 | * | |
72fd4a35 | 240 | * round_jiffies_relative() rounds a time delta in the future (in jiffies) |
4c36a5de AV |
241 | * up or down to (approximately) full seconds. This is useful for timers |
242 | * for which the exact time they fire does not matter too much, as long as | |
243 | * they fire approximately every X seconds. | |
244 | * | |
245 | * By rounding these timers to whole seconds, all such timers will fire | |
246 | * at the same time, rather than at various times spread out. The goal | |
247 | * of this is to have the CPU wake up less, which saves power. | |
248 | * | |
72fd4a35 | 249 | * The return value is the rounded version of the @j parameter. |
4c36a5de AV |
250 | */ |
251 | unsigned long round_jiffies_relative(unsigned long j) | |
252 | { | |
253 | return __round_jiffies_relative(j, raw_smp_processor_id()); | |
254 | } | |
255 | EXPORT_SYMBOL_GPL(round_jiffies_relative); | |
256 | ||
9c133c46 AS |
257 | /** |
258 | * __round_jiffies_up - function to round jiffies up to a full second | |
259 | * @j: the time in (absolute) jiffies that should be rounded | |
260 | * @cpu: the processor number on which the timeout will happen | |
261 | * | |
262 | * This is the same as __round_jiffies() except that it will never | |
263 | * round down. This is useful for timeouts for which the exact time | |
264 | * of firing does not matter too much, as long as they don't fire too | |
265 | * early. | |
266 | */ | |
267 | unsigned long __round_jiffies_up(unsigned long j, int cpu) | |
268 | { | |
269 | return round_jiffies_common(j, cpu, true); | |
270 | } | |
271 | EXPORT_SYMBOL_GPL(__round_jiffies_up); | |
272 | ||
273 | /** | |
274 | * __round_jiffies_up_relative - function to round jiffies up to a full second | |
275 | * @j: the time in (relative) jiffies that should be rounded | |
276 | * @cpu: the processor number on which the timeout will happen | |
277 | * | |
278 | * This is the same as __round_jiffies_relative() except that it will never | |
279 | * round down. This is useful for timeouts for which the exact time | |
280 | * of firing does not matter too much, as long as they don't fire too | |
281 | * early. | |
282 | */ | |
283 | unsigned long __round_jiffies_up_relative(unsigned long j, int cpu) | |
284 | { | |
285 | unsigned long j0 = jiffies; | |
286 | ||
287 | /* Use j0 because jiffies might change while we run */ | |
288 | return round_jiffies_common(j + j0, cpu, true) - j0; | |
289 | } | |
290 | EXPORT_SYMBOL_GPL(__round_jiffies_up_relative); | |
291 | ||
292 | /** | |
293 | * round_jiffies_up - function to round jiffies up to a full second | |
294 | * @j: the time in (absolute) jiffies that should be rounded | |
295 | * | |
296 | * This is the same as round_jiffies() except that it will never | |
297 | * round down. This is useful for timeouts for which the exact time | |
298 | * of firing does not matter too much, as long as they don't fire too | |
299 | * early. | |
300 | */ | |
301 | unsigned long round_jiffies_up(unsigned long j) | |
302 | { | |
303 | return round_jiffies_common(j, raw_smp_processor_id(), true); | |
304 | } | |
305 | EXPORT_SYMBOL_GPL(round_jiffies_up); | |
306 | ||
307 | /** | |
308 | * round_jiffies_up_relative - function to round jiffies up to a full second | |
309 | * @j: the time in (relative) jiffies that should be rounded | |
310 | * | |
311 | * This is the same as round_jiffies_relative() except that it will never | |
312 | * round down. This is useful for timeouts for which the exact time | |
313 | * of firing does not matter too much, as long as they don't fire too | |
314 | * early. | |
315 | */ | |
316 | unsigned long round_jiffies_up_relative(unsigned long j) | |
317 | { | |
318 | return __round_jiffies_up_relative(j, raw_smp_processor_id()); | |
319 | } | |
320 | EXPORT_SYMBOL_GPL(round_jiffies_up_relative); | |
321 | ||
3bbb9ec9 AV |
322 | /** |
323 | * set_timer_slack - set the allowed slack for a timer | |
324 | * @slack_hz: the amount of time (in jiffies) allowed for rounding | |
325 | * | |
326 | * Set the amount of time, in jiffies, that a certain timer has | |
327 | * in terms of slack. By setting this value, the timer subsystem | |
328 | * will schedule the actual timer somewhere between | |
329 | * the time mod_timer() asks for, and that time plus the slack. | |
330 | * | |
331 | * By setting the slack to -1, a percentage of the delay is used | |
332 | * instead. | |
333 | */ | |
334 | void set_timer_slack(struct timer_list *timer, int slack_hz) | |
335 | { | |
336 | timer->slack = slack_hz; | |
337 | } | |
338 | EXPORT_SYMBOL_GPL(set_timer_slack); | |
339 | ||
4c36a5de | 340 | |
a6fa8e5a | 341 | static inline void set_running_timer(struct tvec_base *base, |
1da177e4 LT |
342 | struct timer_list *timer) |
343 | { | |
344 | #ifdef CONFIG_SMP | |
3691c519 | 345 | base->running_timer = timer; |
1da177e4 LT |
346 | #endif |
347 | } | |
348 | ||
a6fa8e5a | 349 | static void internal_add_timer(struct tvec_base *base, struct timer_list *timer) |
1da177e4 LT |
350 | { |
351 | unsigned long expires = timer->expires; | |
352 | unsigned long idx = expires - base->timer_jiffies; | |
353 | struct list_head *vec; | |
354 | ||
355 | if (idx < TVR_SIZE) { | |
356 | int i = expires & TVR_MASK; | |
357 | vec = base->tv1.vec + i; | |
358 | } else if (idx < 1 << (TVR_BITS + TVN_BITS)) { | |
359 | int i = (expires >> TVR_BITS) & TVN_MASK; | |
360 | vec = base->tv2.vec + i; | |
361 | } else if (idx < 1 << (TVR_BITS + 2 * TVN_BITS)) { | |
362 | int i = (expires >> (TVR_BITS + TVN_BITS)) & TVN_MASK; | |
363 | vec = base->tv3.vec + i; | |
364 | } else if (idx < 1 << (TVR_BITS + 3 * TVN_BITS)) { | |
365 | int i = (expires >> (TVR_BITS + 2 * TVN_BITS)) & TVN_MASK; | |
366 | vec = base->tv4.vec + i; | |
367 | } else if ((signed long) idx < 0) { | |
368 | /* | |
369 | * Can happen if you add a timer with expires == jiffies, | |
370 | * or you set a timer to go off in the past | |
371 | */ | |
372 | vec = base->tv1.vec + (base->timer_jiffies & TVR_MASK); | |
373 | } else { | |
374 | int i; | |
375 | /* If the timeout is larger than 0xffffffff on 64-bit | |
376 | * architectures then we use the maximum timeout: | |
377 | */ | |
378 | if (idx > 0xffffffffUL) { | |
379 | idx = 0xffffffffUL; | |
380 | expires = idx + base->timer_jiffies; | |
381 | } | |
382 | i = (expires >> (TVR_BITS + 3 * TVN_BITS)) & TVN_MASK; | |
383 | vec = base->tv5.vec + i; | |
384 | } | |
385 | /* | |
386 | * Timers are FIFO: | |
387 | */ | |
388 | list_add_tail(&timer->entry, vec); | |
389 | } | |
390 | ||
82f67cd9 IM |
391 | #ifdef CONFIG_TIMER_STATS |
392 | void __timer_stats_timer_set_start_info(struct timer_list *timer, void *addr) | |
393 | { | |
394 | if (timer->start_site) | |
395 | return; | |
396 | ||
397 | timer->start_site = addr; | |
398 | memcpy(timer->start_comm, current->comm, TASK_COMM_LEN); | |
399 | timer->start_pid = current->pid; | |
400 | } | |
c5c061b8 VP |
401 | |
402 | static void timer_stats_account_timer(struct timer_list *timer) | |
403 | { | |
404 | unsigned int flag = 0; | |
405 | ||
507e1231 HC |
406 | if (likely(!timer->start_site)) |
407 | return; | |
c5c061b8 VP |
408 | if (unlikely(tbase_get_deferrable(timer->base))) |
409 | flag |= TIMER_STATS_FLAG_DEFERRABLE; | |
410 | ||
411 | timer_stats_update_stats(timer, timer->start_pid, timer->start_site, | |
412 | timer->function, timer->start_comm, flag); | |
413 | } | |
414 | ||
415 | #else | |
416 | static void timer_stats_account_timer(struct timer_list *timer) {} | |
82f67cd9 IM |
417 | #endif |
418 | ||
c6f3a97f TG |
419 | #ifdef CONFIG_DEBUG_OBJECTS_TIMERS |
420 | ||
421 | static struct debug_obj_descr timer_debug_descr; | |
422 | ||
423 | /* | |
424 | * fixup_init is called when: | |
425 | * - an active object is initialized | |
55c888d6 | 426 | */ |
c6f3a97f TG |
427 | static int timer_fixup_init(void *addr, enum debug_obj_state state) |
428 | { | |
429 | struct timer_list *timer = addr; | |
430 | ||
431 | switch (state) { | |
432 | case ODEBUG_STATE_ACTIVE: | |
433 | del_timer_sync(timer); | |
434 | debug_object_init(timer, &timer_debug_descr); | |
435 | return 1; | |
436 | default: | |
437 | return 0; | |
438 | } | |
439 | } | |
440 | ||
441 | /* | |
442 | * fixup_activate is called when: | |
443 | * - an active object is activated | |
444 | * - an unknown object is activated (might be a statically initialized object) | |
445 | */ | |
446 | static int timer_fixup_activate(void *addr, enum debug_obj_state state) | |
447 | { | |
448 | struct timer_list *timer = addr; | |
449 | ||
450 | switch (state) { | |
451 | ||
452 | case ODEBUG_STATE_NOTAVAILABLE: | |
453 | /* | |
454 | * This is not really a fixup. The timer was | |
455 | * statically initialized. We just make sure that it | |
456 | * is tracked in the object tracker. | |
457 | */ | |
458 | if (timer->entry.next == NULL && | |
459 | timer->entry.prev == TIMER_ENTRY_STATIC) { | |
460 | debug_object_init(timer, &timer_debug_descr); | |
461 | debug_object_activate(timer, &timer_debug_descr); | |
462 | return 0; | |
463 | } else { | |
464 | WARN_ON_ONCE(1); | |
465 | } | |
466 | return 0; | |
467 | ||
468 | case ODEBUG_STATE_ACTIVE: | |
469 | WARN_ON(1); | |
470 | ||
471 | default: | |
472 | return 0; | |
473 | } | |
474 | } | |
475 | ||
476 | /* | |
477 | * fixup_free is called when: | |
478 | * - an active object is freed | |
479 | */ | |
480 | static int timer_fixup_free(void *addr, enum debug_obj_state state) | |
481 | { | |
482 | struct timer_list *timer = addr; | |
483 | ||
484 | switch (state) { | |
485 | case ODEBUG_STATE_ACTIVE: | |
486 | del_timer_sync(timer); | |
487 | debug_object_free(timer, &timer_debug_descr); | |
488 | return 1; | |
489 | default: | |
490 | return 0; | |
491 | } | |
492 | } | |
493 | ||
494 | static struct debug_obj_descr timer_debug_descr = { | |
495 | .name = "timer_list", | |
496 | .fixup_init = timer_fixup_init, | |
497 | .fixup_activate = timer_fixup_activate, | |
498 | .fixup_free = timer_fixup_free, | |
499 | }; | |
500 | ||
501 | static inline void debug_timer_init(struct timer_list *timer) | |
502 | { | |
503 | debug_object_init(timer, &timer_debug_descr); | |
504 | } | |
505 | ||
506 | static inline void debug_timer_activate(struct timer_list *timer) | |
507 | { | |
508 | debug_object_activate(timer, &timer_debug_descr); | |
509 | } | |
510 | ||
511 | static inline void debug_timer_deactivate(struct timer_list *timer) | |
512 | { | |
513 | debug_object_deactivate(timer, &timer_debug_descr); | |
514 | } | |
515 | ||
516 | static inline void debug_timer_free(struct timer_list *timer) | |
517 | { | |
518 | debug_object_free(timer, &timer_debug_descr); | |
519 | } | |
520 | ||
6f2b9b9a JB |
521 | static void __init_timer(struct timer_list *timer, |
522 | const char *name, | |
523 | struct lock_class_key *key); | |
c6f3a97f | 524 | |
6f2b9b9a JB |
525 | void init_timer_on_stack_key(struct timer_list *timer, |
526 | const char *name, | |
527 | struct lock_class_key *key) | |
c6f3a97f TG |
528 | { |
529 | debug_object_init_on_stack(timer, &timer_debug_descr); | |
6f2b9b9a | 530 | __init_timer(timer, name, key); |
c6f3a97f | 531 | } |
6f2b9b9a | 532 | EXPORT_SYMBOL_GPL(init_timer_on_stack_key); |
c6f3a97f TG |
533 | |
534 | void destroy_timer_on_stack(struct timer_list *timer) | |
535 | { | |
536 | debug_object_free(timer, &timer_debug_descr); | |
537 | } | |
538 | EXPORT_SYMBOL_GPL(destroy_timer_on_stack); | |
539 | ||
540 | #else | |
541 | static inline void debug_timer_init(struct timer_list *timer) { } | |
542 | static inline void debug_timer_activate(struct timer_list *timer) { } | |
543 | static inline void debug_timer_deactivate(struct timer_list *timer) { } | |
544 | #endif | |
545 | ||
2b022e3d XG |
546 | static inline void debug_init(struct timer_list *timer) |
547 | { | |
548 | debug_timer_init(timer); | |
549 | trace_timer_init(timer); | |
550 | } | |
551 | ||
552 | static inline void | |
553 | debug_activate(struct timer_list *timer, unsigned long expires) | |
554 | { | |
555 | debug_timer_activate(timer); | |
556 | trace_timer_start(timer, expires); | |
557 | } | |
558 | ||
559 | static inline void debug_deactivate(struct timer_list *timer) | |
560 | { | |
561 | debug_timer_deactivate(timer); | |
562 | trace_timer_cancel(timer); | |
563 | } | |
564 | ||
6f2b9b9a JB |
565 | static void __init_timer(struct timer_list *timer, |
566 | const char *name, | |
567 | struct lock_class_key *key) | |
55c888d6 ON |
568 | { |
569 | timer->entry.next = NULL; | |
bfe5d834 | 570 | timer->base = __raw_get_cpu_var(tvec_bases); |
3bbb9ec9 | 571 | timer->slack = -1; |
82f67cd9 IM |
572 | #ifdef CONFIG_TIMER_STATS |
573 | timer->start_site = NULL; | |
574 | timer->start_pid = -1; | |
575 | memset(timer->start_comm, 0, TASK_COMM_LEN); | |
576 | #endif | |
6f2b9b9a | 577 | lockdep_init_map(&timer->lockdep_map, name, key, 0); |
55c888d6 | 578 | } |
c6f3a97f | 579 | |
8cadd283 JB |
580 | void setup_deferrable_timer_on_stack_key(struct timer_list *timer, |
581 | const char *name, | |
582 | struct lock_class_key *key, | |
583 | void (*function)(unsigned long), | |
584 | unsigned long data) | |
585 | { | |
586 | timer->function = function; | |
587 | timer->data = data; | |
588 | init_timer_on_stack_key(timer, name, key); | |
589 | timer_set_deferrable(timer); | |
590 | } | |
591 | EXPORT_SYMBOL_GPL(setup_deferrable_timer_on_stack_key); | |
592 | ||
c6f3a97f | 593 | /** |
633fe795 | 594 | * init_timer_key - initialize a timer |
c6f3a97f | 595 | * @timer: the timer to be initialized |
633fe795 RD |
596 | * @name: name of the timer |
597 | * @key: lockdep class key of the fake lock used for tracking timer | |
598 | * sync lock dependencies | |
c6f3a97f | 599 | * |
633fe795 | 600 | * init_timer_key() must be done to a timer prior calling *any* of the |
c6f3a97f TG |
601 | * other timer functions. |
602 | */ | |
6f2b9b9a JB |
603 | void init_timer_key(struct timer_list *timer, |
604 | const char *name, | |
605 | struct lock_class_key *key) | |
c6f3a97f | 606 | { |
2b022e3d | 607 | debug_init(timer); |
6f2b9b9a | 608 | __init_timer(timer, name, key); |
c6f3a97f | 609 | } |
6f2b9b9a | 610 | EXPORT_SYMBOL(init_timer_key); |
55c888d6 | 611 | |
6f2b9b9a JB |
612 | void init_timer_deferrable_key(struct timer_list *timer, |
613 | const char *name, | |
614 | struct lock_class_key *key) | |
6e453a67 | 615 | { |
6f2b9b9a | 616 | init_timer_key(timer, name, key); |
6e453a67 VP |
617 | timer_set_deferrable(timer); |
618 | } | |
6f2b9b9a | 619 | EXPORT_SYMBOL(init_timer_deferrable_key); |
6e453a67 | 620 | |
55c888d6 | 621 | static inline void detach_timer(struct timer_list *timer, |
82f67cd9 | 622 | int clear_pending) |
55c888d6 ON |
623 | { |
624 | struct list_head *entry = &timer->entry; | |
625 | ||
2b022e3d | 626 | debug_deactivate(timer); |
c6f3a97f | 627 | |
55c888d6 ON |
628 | __list_del(entry->prev, entry->next); |
629 | if (clear_pending) | |
630 | entry->next = NULL; | |
631 | entry->prev = LIST_POISON2; | |
632 | } | |
633 | ||
634 | /* | |
3691c519 | 635 | * We are using hashed locking: holding per_cpu(tvec_bases).lock |
55c888d6 ON |
636 | * means that all timers which are tied to this base via timer->base are |
637 | * locked, and the base itself is locked too. | |
638 | * | |
639 | * So __run_timers/migrate_timers can safely modify all timers which could | |
640 | * be found on ->tvX lists. | |
641 | * | |
642 | * When the timer's base is locked, and the timer removed from list, it is | |
643 | * possible to set timer->base = NULL and drop the lock: the timer remains | |
644 | * locked. | |
645 | */ | |
a6fa8e5a | 646 | static struct tvec_base *lock_timer_base(struct timer_list *timer, |
55c888d6 | 647 | unsigned long *flags) |
89e7e374 | 648 | __acquires(timer->base->lock) |
55c888d6 | 649 | { |
a6fa8e5a | 650 | struct tvec_base *base; |
55c888d6 ON |
651 | |
652 | for (;;) { | |
a6fa8e5a | 653 | struct tvec_base *prelock_base = timer->base; |
6e453a67 | 654 | base = tbase_get_base(prelock_base); |
55c888d6 ON |
655 | if (likely(base != NULL)) { |
656 | spin_lock_irqsave(&base->lock, *flags); | |
6e453a67 | 657 | if (likely(prelock_base == timer->base)) |
55c888d6 ON |
658 | return base; |
659 | /* The timer has migrated to another CPU */ | |
660 | spin_unlock_irqrestore(&base->lock, *flags); | |
661 | } | |
662 | cpu_relax(); | |
663 | } | |
664 | } | |
665 | ||
74019224 | 666 | static inline int |
597d0275 AB |
667 | __mod_timer(struct timer_list *timer, unsigned long expires, |
668 | bool pending_only, int pinned) | |
1da177e4 | 669 | { |
a6fa8e5a | 670 | struct tvec_base *base, *new_base; |
1da177e4 | 671 | unsigned long flags; |
eea08f32 | 672 | int ret = 0 , cpu; |
1da177e4 | 673 | |
82f67cd9 | 674 | timer_stats_timer_set_start_info(timer); |
1da177e4 | 675 | BUG_ON(!timer->function); |
1da177e4 | 676 | |
55c888d6 ON |
677 | base = lock_timer_base(timer, &flags); |
678 | ||
679 | if (timer_pending(timer)) { | |
680 | detach_timer(timer, 0); | |
97fd9ed4 MS |
681 | if (timer->expires == base->next_timer && |
682 | !tbase_get_deferrable(timer->base)) | |
683 | base->next_timer = base->timer_jiffies; | |
55c888d6 | 684 | ret = 1; |
74019224 IM |
685 | } else { |
686 | if (pending_only) | |
687 | goto out_unlock; | |
55c888d6 ON |
688 | } |
689 | ||
2b022e3d | 690 | debug_activate(timer, expires); |
c6f3a97f | 691 | |
eea08f32 AB |
692 | cpu = smp_processor_id(); |
693 | ||
694 | #if defined(CONFIG_NO_HZ) && defined(CONFIG_SMP) | |
695 | if (!pinned && get_sysctl_timer_migration() && idle_cpu(cpu)) { | |
696 | int preferred_cpu = get_nohz_load_balancer(); | |
697 | ||
698 | if (preferred_cpu >= 0) | |
699 | cpu = preferred_cpu; | |
700 | } | |
701 | #endif | |
702 | new_base = per_cpu(tvec_bases, cpu); | |
703 | ||
3691c519 | 704 | if (base != new_base) { |
1da177e4 | 705 | /* |
55c888d6 ON |
706 | * We are trying to schedule the timer on the local CPU. |
707 | * However we can't change timer's base while it is running, | |
708 | * otherwise del_timer_sync() can't detect that the timer's | |
709 | * handler yet has not finished. This also guarantees that | |
710 | * the timer is serialized wrt itself. | |
1da177e4 | 711 | */ |
a2c348fe | 712 | if (likely(base->running_timer != timer)) { |
55c888d6 | 713 | /* See the comment in lock_timer_base() */ |
6e453a67 | 714 | timer_set_base(timer, NULL); |
55c888d6 | 715 | spin_unlock(&base->lock); |
a2c348fe ON |
716 | base = new_base; |
717 | spin_lock(&base->lock); | |
6e453a67 | 718 | timer_set_base(timer, base); |
1da177e4 LT |
719 | } |
720 | } | |
721 | ||
1da177e4 | 722 | timer->expires = expires; |
97fd9ed4 MS |
723 | if (time_before(timer->expires, base->next_timer) && |
724 | !tbase_get_deferrable(timer->base)) | |
725 | base->next_timer = timer->expires; | |
a2c348fe | 726 | internal_add_timer(base, timer); |
74019224 IM |
727 | |
728 | out_unlock: | |
a2c348fe | 729 | spin_unlock_irqrestore(&base->lock, flags); |
1da177e4 LT |
730 | |
731 | return ret; | |
732 | } | |
733 | ||
2aae4a10 | 734 | /** |
74019224 IM |
735 | * mod_timer_pending - modify a pending timer's timeout |
736 | * @timer: the pending timer to be modified | |
737 | * @expires: new timeout in jiffies | |
1da177e4 | 738 | * |
74019224 IM |
739 | * mod_timer_pending() is the same for pending timers as mod_timer(), |
740 | * but will not re-activate and modify already deleted timers. | |
741 | * | |
742 | * It is useful for unserialized use of timers. | |
1da177e4 | 743 | */ |
74019224 | 744 | int mod_timer_pending(struct timer_list *timer, unsigned long expires) |
1da177e4 | 745 | { |
597d0275 | 746 | return __mod_timer(timer, expires, true, TIMER_NOT_PINNED); |
1da177e4 | 747 | } |
74019224 | 748 | EXPORT_SYMBOL(mod_timer_pending); |
1da177e4 | 749 | |
3bbb9ec9 AV |
750 | /* |
751 | * Decide where to put the timer while taking the slack into account | |
752 | * | |
753 | * Algorithm: | |
754 | * 1) calculate the maximum (absolute) time | |
755 | * 2) calculate the highest bit where the expires and new max are different | |
756 | * 3) use this bit to make a mask | |
757 | * 4) use the bitmask to round down the maximum time, so that all last | |
758 | * bits are zeros | |
759 | */ | |
760 | static inline | |
761 | unsigned long apply_slack(struct timer_list *timer, unsigned long expires) | |
762 | { | |
763 | unsigned long expires_limit, mask; | |
764 | int bit; | |
765 | ||
f00e047e | 766 | expires_limit = expires; |
3bbb9ec9 | 767 | |
8e63d779 | 768 | if (timer->slack >= 0) { |
f00e047e | 769 | expires_limit = expires + timer->slack; |
8e63d779 | 770 | } else { |
2abfb9e1 | 771 | unsigned long now = jiffies; |
3bbb9ec9 | 772 | |
8e63d779 TG |
773 | /* No slack, if already expired else auto slack 0.4% */ |
774 | if (time_after(expires, now)) | |
775 | expires_limit = expires + (expires - now)/256; | |
776 | } | |
3bbb9ec9 | 777 | mask = expires ^ expires_limit; |
3bbb9ec9 AV |
778 | if (mask == 0) |
779 | return expires; | |
780 | ||
781 | bit = find_last_bit(&mask, BITS_PER_LONG); | |
782 | ||
783 | mask = (1 << bit) - 1; | |
784 | ||
785 | expires_limit = expires_limit & ~(mask); | |
786 | ||
787 | return expires_limit; | |
788 | } | |
789 | ||
2aae4a10 | 790 | /** |
1da177e4 LT |
791 | * mod_timer - modify a timer's timeout |
792 | * @timer: the timer to be modified | |
2aae4a10 | 793 | * @expires: new timeout in jiffies |
1da177e4 | 794 | * |
72fd4a35 | 795 | * mod_timer() is a more efficient way to update the expire field of an |
1da177e4 LT |
796 | * active timer (if the timer is inactive it will be activated) |
797 | * | |
798 | * mod_timer(timer, expires) is equivalent to: | |
799 | * | |
800 | * del_timer(timer); timer->expires = expires; add_timer(timer); | |
801 | * | |
802 | * Note that if there are multiple unserialized concurrent users of the | |
803 | * same timer, then mod_timer() is the only safe way to modify the timeout, | |
804 | * since add_timer() cannot modify an already running timer. | |
805 | * | |
806 | * The function returns whether it has modified a pending timer or not. | |
807 | * (ie. mod_timer() of an inactive timer returns 0, mod_timer() of an | |
808 | * active timer returns 1.) | |
809 | */ | |
810 | int mod_timer(struct timer_list *timer, unsigned long expires) | |
811 | { | |
1da177e4 LT |
812 | /* |
813 | * This is a common optimization triggered by the | |
814 | * networking code - if the timer is re-modified | |
815 | * to be the same thing then just return: | |
816 | */ | |
4841158b | 817 | if (timer_pending(timer) && timer->expires == expires) |
1da177e4 LT |
818 | return 1; |
819 | ||
3bbb9ec9 AV |
820 | expires = apply_slack(timer, expires); |
821 | ||
597d0275 | 822 | return __mod_timer(timer, expires, false, TIMER_NOT_PINNED); |
1da177e4 | 823 | } |
1da177e4 LT |
824 | EXPORT_SYMBOL(mod_timer); |
825 | ||
597d0275 AB |
826 | /** |
827 | * mod_timer_pinned - modify a timer's timeout | |
828 | * @timer: the timer to be modified | |
829 | * @expires: new timeout in jiffies | |
830 | * | |
831 | * mod_timer_pinned() is a way to update the expire field of an | |
832 | * active timer (if the timer is inactive it will be activated) | |
833 | * and not allow the timer to be migrated to a different CPU. | |
834 | * | |
835 | * mod_timer_pinned(timer, expires) is equivalent to: | |
836 | * | |
837 | * del_timer(timer); timer->expires = expires; add_timer(timer); | |
838 | */ | |
839 | int mod_timer_pinned(struct timer_list *timer, unsigned long expires) | |
840 | { | |
841 | if (timer->expires == expires && timer_pending(timer)) | |
842 | return 1; | |
843 | ||
844 | return __mod_timer(timer, expires, false, TIMER_PINNED); | |
845 | } | |
846 | EXPORT_SYMBOL(mod_timer_pinned); | |
847 | ||
74019224 IM |
848 | /** |
849 | * add_timer - start a timer | |
850 | * @timer: the timer to be added | |
851 | * | |
852 | * The kernel will do a ->function(->data) callback from the | |
853 | * timer interrupt at the ->expires point in the future. The | |
854 | * current time is 'jiffies'. | |
855 | * | |
856 | * The timer's ->expires, ->function (and if the handler uses it, ->data) | |
857 | * fields must be set prior calling this function. | |
858 | * | |
859 | * Timers with an ->expires field in the past will be executed in the next | |
860 | * timer tick. | |
861 | */ | |
862 | void add_timer(struct timer_list *timer) | |
863 | { | |
864 | BUG_ON(timer_pending(timer)); | |
865 | mod_timer(timer, timer->expires); | |
866 | } | |
867 | EXPORT_SYMBOL(add_timer); | |
868 | ||
869 | /** | |
870 | * add_timer_on - start a timer on a particular CPU | |
871 | * @timer: the timer to be added | |
872 | * @cpu: the CPU to start it on | |
873 | * | |
874 | * This is not very scalable on SMP. Double adds are not possible. | |
875 | */ | |
876 | void add_timer_on(struct timer_list *timer, int cpu) | |
877 | { | |
878 | struct tvec_base *base = per_cpu(tvec_bases, cpu); | |
879 | unsigned long flags; | |
880 | ||
881 | timer_stats_timer_set_start_info(timer); | |
882 | BUG_ON(timer_pending(timer) || !timer->function); | |
883 | spin_lock_irqsave(&base->lock, flags); | |
884 | timer_set_base(timer, base); | |
2b022e3d | 885 | debug_activate(timer, timer->expires); |
97fd9ed4 MS |
886 | if (time_before(timer->expires, base->next_timer) && |
887 | !tbase_get_deferrable(timer->base)) | |
888 | base->next_timer = timer->expires; | |
74019224 IM |
889 | internal_add_timer(base, timer); |
890 | /* | |
891 | * Check whether the other CPU is idle and needs to be | |
892 | * triggered to reevaluate the timer wheel when nohz is | |
893 | * active. We are protected against the other CPU fiddling | |
894 | * with the timer by holding the timer base lock. This also | |
895 | * makes sure that a CPU on the way to idle can not evaluate | |
896 | * the timer wheel. | |
897 | */ | |
898 | wake_up_idle_cpu(cpu); | |
899 | spin_unlock_irqrestore(&base->lock, flags); | |
900 | } | |
a9862e05 | 901 | EXPORT_SYMBOL_GPL(add_timer_on); |
74019224 | 902 | |
2aae4a10 | 903 | /** |
1da177e4 LT |
904 | * del_timer - deactive a timer. |
905 | * @timer: the timer to be deactivated | |
906 | * | |
907 | * del_timer() deactivates a timer - this works on both active and inactive | |
908 | * timers. | |
909 | * | |
910 | * The function returns whether it has deactivated a pending timer or not. | |
911 | * (ie. del_timer() of an inactive timer returns 0, del_timer() of an | |
912 | * active timer returns 1.) | |
913 | */ | |
914 | int del_timer(struct timer_list *timer) | |
915 | { | |
a6fa8e5a | 916 | struct tvec_base *base; |
1da177e4 | 917 | unsigned long flags; |
55c888d6 | 918 | int ret = 0; |
1da177e4 | 919 | |
82f67cd9 | 920 | timer_stats_timer_clear_start_info(timer); |
55c888d6 ON |
921 | if (timer_pending(timer)) { |
922 | base = lock_timer_base(timer, &flags); | |
923 | if (timer_pending(timer)) { | |
924 | detach_timer(timer, 1); | |
97fd9ed4 MS |
925 | if (timer->expires == base->next_timer && |
926 | !tbase_get_deferrable(timer->base)) | |
927 | base->next_timer = base->timer_jiffies; | |
55c888d6 ON |
928 | ret = 1; |
929 | } | |
1da177e4 | 930 | spin_unlock_irqrestore(&base->lock, flags); |
1da177e4 | 931 | } |
1da177e4 | 932 | |
55c888d6 | 933 | return ret; |
1da177e4 | 934 | } |
1da177e4 LT |
935 | EXPORT_SYMBOL(del_timer); |
936 | ||
937 | #ifdef CONFIG_SMP | |
2aae4a10 REB |
938 | /** |
939 | * try_to_del_timer_sync - Try to deactivate a timer | |
940 | * @timer: timer do del | |
941 | * | |
fd450b73 ON |
942 | * This function tries to deactivate a timer. Upon successful (ret >= 0) |
943 | * exit the timer is not queued and the handler is not running on any CPU. | |
944 | * | |
945 | * It must not be called from interrupt contexts. | |
946 | */ | |
947 | int try_to_del_timer_sync(struct timer_list *timer) | |
948 | { | |
a6fa8e5a | 949 | struct tvec_base *base; |
fd450b73 ON |
950 | unsigned long flags; |
951 | int ret = -1; | |
952 | ||
953 | base = lock_timer_base(timer, &flags); | |
954 | ||
955 | if (base->running_timer == timer) | |
956 | goto out; | |
957 | ||
829b6c1e | 958 | timer_stats_timer_clear_start_info(timer); |
fd450b73 ON |
959 | ret = 0; |
960 | if (timer_pending(timer)) { | |
961 | detach_timer(timer, 1); | |
97fd9ed4 MS |
962 | if (timer->expires == base->next_timer && |
963 | !tbase_get_deferrable(timer->base)) | |
964 | base->next_timer = base->timer_jiffies; | |
fd450b73 ON |
965 | ret = 1; |
966 | } | |
967 | out: | |
968 | spin_unlock_irqrestore(&base->lock, flags); | |
969 | ||
970 | return ret; | |
971 | } | |
e19dff1f DH |
972 | EXPORT_SYMBOL(try_to_del_timer_sync); |
973 | ||
2aae4a10 | 974 | /** |
1da177e4 LT |
975 | * del_timer_sync - deactivate a timer and wait for the handler to finish. |
976 | * @timer: the timer to be deactivated | |
977 | * | |
978 | * This function only differs from del_timer() on SMP: besides deactivating | |
979 | * the timer it also makes sure the handler has finished executing on other | |
980 | * CPUs. | |
981 | * | |
72fd4a35 | 982 | * Synchronization rules: Callers must prevent restarting of the timer, |
1da177e4 LT |
983 | * otherwise this function is meaningless. It must not be called from |
984 | * interrupt contexts. The caller must not hold locks which would prevent | |
55c888d6 ON |
985 | * completion of the timer's handler. The timer's handler must not call |
986 | * add_timer_on(). Upon exit the timer is not queued and the handler is | |
987 | * not running on any CPU. | |
1da177e4 LT |
988 | * |
989 | * The function returns whether it has deactivated a pending timer or not. | |
1da177e4 LT |
990 | */ |
991 | int del_timer_sync(struct timer_list *timer) | |
992 | { | |
6f2b9b9a JB |
993 | #ifdef CONFIG_LOCKDEP |
994 | unsigned long flags; | |
995 | ||
996 | local_irq_save(flags); | |
997 | lock_map_acquire(&timer->lockdep_map); | |
998 | lock_map_release(&timer->lockdep_map); | |
999 | local_irq_restore(flags); | |
1000 | #endif | |
1001 | ||
fd450b73 ON |
1002 | for (;;) { |
1003 | int ret = try_to_del_timer_sync(timer); | |
1004 | if (ret >= 0) | |
1005 | return ret; | |
a0009652 | 1006 | cpu_relax(); |
fd450b73 | 1007 | } |
1da177e4 | 1008 | } |
55c888d6 | 1009 | EXPORT_SYMBOL(del_timer_sync); |
1da177e4 LT |
1010 | #endif |
1011 | ||
a6fa8e5a | 1012 | static int cascade(struct tvec_base *base, struct tvec *tv, int index) |
1da177e4 LT |
1013 | { |
1014 | /* cascade all the timers from tv up one level */ | |
3439dd86 P |
1015 | struct timer_list *timer, *tmp; |
1016 | struct list_head tv_list; | |
1017 | ||
1018 | list_replace_init(tv->vec + index, &tv_list); | |
1da177e4 | 1019 | |
1da177e4 | 1020 | /* |
3439dd86 P |
1021 | * We are removing _all_ timers from the list, so we |
1022 | * don't have to detach them individually. | |
1da177e4 | 1023 | */ |
3439dd86 | 1024 | list_for_each_entry_safe(timer, tmp, &tv_list, entry) { |
6e453a67 | 1025 | BUG_ON(tbase_get_base(timer->base) != base); |
3439dd86 | 1026 | internal_add_timer(base, timer); |
1da177e4 | 1027 | } |
1da177e4 LT |
1028 | |
1029 | return index; | |
1030 | } | |
1031 | ||
576da126 TG |
1032 | static void call_timer_fn(struct timer_list *timer, void (*fn)(unsigned long), |
1033 | unsigned long data) | |
1034 | { | |
1035 | int preempt_count = preempt_count(); | |
1036 | ||
1037 | #ifdef CONFIG_LOCKDEP | |
1038 | /* | |
1039 | * It is permissible to free the timer from inside the | |
1040 | * function that is called from it, this we need to take into | |
1041 | * account for lockdep too. To avoid bogus "held lock freed" | |
1042 | * warnings as well as problems when looking into | |
1043 | * timer->lockdep_map, make a copy and use that here. | |
1044 | */ | |
1045 | struct lockdep_map lockdep_map = timer->lockdep_map; | |
1046 | #endif | |
1047 | /* | |
1048 | * Couple the lock chain with the lock chain at | |
1049 | * del_timer_sync() by acquiring the lock_map around the fn() | |
1050 | * call here and in del_timer_sync(). | |
1051 | */ | |
1052 | lock_map_acquire(&lockdep_map); | |
1053 | ||
1054 | trace_timer_expire_entry(timer); | |
1055 | fn(data); | |
1056 | trace_timer_expire_exit(timer); | |
1057 | ||
1058 | lock_map_release(&lockdep_map); | |
1059 | ||
1060 | if (preempt_count != preempt_count()) { | |
802702e0 TG |
1061 | WARN_ONCE(1, "timer: %pF preempt leak: %08x -> %08x\n", |
1062 | fn, preempt_count, preempt_count()); | |
1063 | /* | |
1064 | * Restore the preempt count. That gives us a decent | |
1065 | * chance to survive and extract information. If the | |
1066 | * callback kept a lock held, bad luck, but not worse | |
1067 | * than the BUG() we had. | |
1068 | */ | |
1069 | preempt_count() = preempt_count; | |
576da126 TG |
1070 | } |
1071 | } | |
1072 | ||
2aae4a10 REB |
1073 | #define INDEX(N) ((base->timer_jiffies >> (TVR_BITS + (N) * TVN_BITS)) & TVN_MASK) |
1074 | ||
1075 | /** | |
1da177e4 LT |
1076 | * __run_timers - run all expired timers (if any) on this CPU. |
1077 | * @base: the timer vector to be processed. | |
1078 | * | |
1079 | * This function cascades all vectors and executes all expired timer | |
1080 | * vectors. | |
1081 | */ | |
a6fa8e5a | 1082 | static inline void __run_timers(struct tvec_base *base) |
1da177e4 LT |
1083 | { |
1084 | struct timer_list *timer; | |
1085 | ||
3691c519 | 1086 | spin_lock_irq(&base->lock); |
1da177e4 | 1087 | while (time_after_eq(jiffies, base->timer_jiffies)) { |
626ab0e6 | 1088 | struct list_head work_list; |
1da177e4 | 1089 | struct list_head *head = &work_list; |
6819457d | 1090 | int index = base->timer_jiffies & TVR_MASK; |
626ab0e6 | 1091 | |
1da177e4 LT |
1092 | /* |
1093 | * Cascade timers: | |
1094 | */ | |
1095 | if (!index && | |
1096 | (!cascade(base, &base->tv2, INDEX(0))) && | |
1097 | (!cascade(base, &base->tv3, INDEX(1))) && | |
1098 | !cascade(base, &base->tv4, INDEX(2))) | |
1099 | cascade(base, &base->tv5, INDEX(3)); | |
626ab0e6 ON |
1100 | ++base->timer_jiffies; |
1101 | list_replace_init(base->tv1.vec + index, &work_list); | |
55c888d6 | 1102 | while (!list_empty(head)) { |
1da177e4 LT |
1103 | void (*fn)(unsigned long); |
1104 | unsigned long data; | |
1105 | ||
b5e61818 | 1106 | timer = list_first_entry(head, struct timer_list,entry); |
6819457d TG |
1107 | fn = timer->function; |
1108 | data = timer->data; | |
1da177e4 | 1109 | |
82f67cd9 IM |
1110 | timer_stats_account_timer(timer); |
1111 | ||
1da177e4 | 1112 | set_running_timer(base, timer); |
55c888d6 | 1113 | detach_timer(timer, 1); |
6f2b9b9a | 1114 | |
3691c519 | 1115 | spin_unlock_irq(&base->lock); |
576da126 | 1116 | call_timer_fn(timer, fn, data); |
3691c519 | 1117 | spin_lock_irq(&base->lock); |
1da177e4 LT |
1118 | } |
1119 | } | |
1120 | set_running_timer(base, NULL); | |
3691c519 | 1121 | spin_unlock_irq(&base->lock); |
1da177e4 LT |
1122 | } |
1123 | ||
ee9c5785 | 1124 | #ifdef CONFIG_NO_HZ |
1da177e4 LT |
1125 | /* |
1126 | * Find out when the next timer event is due to happen. This | |
90cba64a RD |
1127 | * is used on S/390 to stop all activity when a CPU is idle. |
1128 | * This function needs to be called with interrupts disabled. | |
1da177e4 | 1129 | */ |
a6fa8e5a | 1130 | static unsigned long __next_timer_interrupt(struct tvec_base *base) |
1da177e4 | 1131 | { |
1cfd6849 | 1132 | unsigned long timer_jiffies = base->timer_jiffies; |
eaad084b | 1133 | unsigned long expires = timer_jiffies + NEXT_TIMER_MAX_DELTA; |
1cfd6849 | 1134 | int index, slot, array, found = 0; |
1da177e4 | 1135 | struct timer_list *nte; |
a6fa8e5a | 1136 | struct tvec *varray[4]; |
1da177e4 LT |
1137 | |
1138 | /* Look for timer events in tv1. */ | |
1cfd6849 | 1139 | index = slot = timer_jiffies & TVR_MASK; |
1da177e4 | 1140 | do { |
1cfd6849 | 1141 | list_for_each_entry(nte, base->tv1.vec + slot, entry) { |
6819457d TG |
1142 | if (tbase_get_deferrable(nte->base)) |
1143 | continue; | |
6e453a67 | 1144 | |
1cfd6849 | 1145 | found = 1; |
1da177e4 | 1146 | expires = nte->expires; |
1cfd6849 TG |
1147 | /* Look at the cascade bucket(s)? */ |
1148 | if (!index || slot < index) | |
1149 | goto cascade; | |
1150 | return expires; | |
1da177e4 | 1151 | } |
1cfd6849 TG |
1152 | slot = (slot + 1) & TVR_MASK; |
1153 | } while (slot != index); | |
1154 | ||
1155 | cascade: | |
1156 | /* Calculate the next cascade event */ | |
1157 | if (index) | |
1158 | timer_jiffies += TVR_SIZE - index; | |
1159 | timer_jiffies >>= TVR_BITS; | |
1da177e4 LT |
1160 | |
1161 | /* Check tv2-tv5. */ | |
1162 | varray[0] = &base->tv2; | |
1163 | varray[1] = &base->tv3; | |
1164 | varray[2] = &base->tv4; | |
1165 | varray[3] = &base->tv5; | |
1cfd6849 TG |
1166 | |
1167 | for (array = 0; array < 4; array++) { | |
a6fa8e5a | 1168 | struct tvec *varp = varray[array]; |
1cfd6849 TG |
1169 | |
1170 | index = slot = timer_jiffies & TVN_MASK; | |
1da177e4 | 1171 | do { |
1cfd6849 | 1172 | list_for_each_entry(nte, varp->vec + slot, entry) { |
a0419888 JH |
1173 | if (tbase_get_deferrable(nte->base)) |
1174 | continue; | |
1175 | ||
1cfd6849 | 1176 | found = 1; |
1da177e4 LT |
1177 | if (time_before(nte->expires, expires)) |
1178 | expires = nte->expires; | |
1cfd6849 TG |
1179 | } |
1180 | /* | |
1181 | * Do we still search for the first timer or are | |
1182 | * we looking up the cascade buckets ? | |
1183 | */ | |
1184 | if (found) { | |
1185 | /* Look at the cascade bucket(s)? */ | |
1186 | if (!index || slot < index) | |
1187 | break; | |
1188 | return expires; | |
1189 | } | |
1190 | slot = (slot + 1) & TVN_MASK; | |
1191 | } while (slot != index); | |
1192 | ||
1193 | if (index) | |
1194 | timer_jiffies += TVN_SIZE - index; | |
1195 | timer_jiffies >>= TVN_BITS; | |
1da177e4 | 1196 | } |
1cfd6849 TG |
1197 | return expires; |
1198 | } | |
69239749 | 1199 | |
1cfd6849 TG |
1200 | /* |
1201 | * Check, if the next hrtimer event is before the next timer wheel | |
1202 | * event: | |
1203 | */ | |
1204 | static unsigned long cmp_next_hrtimer_event(unsigned long now, | |
1205 | unsigned long expires) | |
1206 | { | |
1207 | ktime_t hr_delta = hrtimer_get_next_event(); | |
1208 | struct timespec tsdelta; | |
9501b6cf | 1209 | unsigned long delta; |
1cfd6849 TG |
1210 | |
1211 | if (hr_delta.tv64 == KTIME_MAX) | |
1212 | return expires; | |
0662b713 | 1213 | |
9501b6cf TG |
1214 | /* |
1215 | * Expired timer available, let it expire in the next tick | |
1216 | */ | |
1217 | if (hr_delta.tv64 <= 0) | |
1218 | return now + 1; | |
69239749 | 1219 | |
1cfd6849 | 1220 | tsdelta = ktime_to_timespec(hr_delta); |
9501b6cf | 1221 | delta = timespec_to_jiffies(&tsdelta); |
eaad084b TG |
1222 | |
1223 | /* | |
1224 | * Limit the delta to the max value, which is checked in | |
1225 | * tick_nohz_stop_sched_tick(): | |
1226 | */ | |
1227 | if (delta > NEXT_TIMER_MAX_DELTA) | |
1228 | delta = NEXT_TIMER_MAX_DELTA; | |
1229 | ||
9501b6cf TG |
1230 | /* |
1231 | * Take rounding errors in to account and make sure, that it | |
1232 | * expires in the next tick. Otherwise we go into an endless | |
1233 | * ping pong due to tick_nohz_stop_sched_tick() retriggering | |
1234 | * the timer softirq | |
1235 | */ | |
1236 | if (delta < 1) | |
1237 | delta = 1; | |
1238 | now += delta; | |
1cfd6849 TG |
1239 | if (time_before(now, expires)) |
1240 | return now; | |
1da177e4 LT |
1241 | return expires; |
1242 | } | |
1cfd6849 TG |
1243 | |
1244 | /** | |
8dce39c2 | 1245 | * get_next_timer_interrupt - return the jiffy of the next pending timer |
05fb6bf0 | 1246 | * @now: current time (in jiffies) |
1cfd6849 | 1247 | */ |
fd064b9b | 1248 | unsigned long get_next_timer_interrupt(unsigned long now) |
1cfd6849 | 1249 | { |
a6fa8e5a | 1250 | struct tvec_base *base = __get_cpu_var(tvec_bases); |
fd064b9b | 1251 | unsigned long expires; |
1cfd6849 TG |
1252 | |
1253 | spin_lock(&base->lock); | |
97fd9ed4 MS |
1254 | if (time_before_eq(base->next_timer, base->timer_jiffies)) |
1255 | base->next_timer = __next_timer_interrupt(base); | |
1256 | expires = base->next_timer; | |
1cfd6849 TG |
1257 | spin_unlock(&base->lock); |
1258 | ||
1259 | if (time_before_eq(expires, now)) | |
1260 | return now; | |
1261 | ||
1262 | return cmp_next_hrtimer_event(now, expires); | |
1263 | } | |
1da177e4 LT |
1264 | #endif |
1265 | ||
1da177e4 | 1266 | /* |
5b4db0c2 | 1267 | * Called from the timer interrupt handler to charge one tick to the current |
1da177e4 LT |
1268 | * process. user_tick is 1 if the tick is user time, 0 for system. |
1269 | */ | |
1270 | void update_process_times(int user_tick) | |
1271 | { | |
1272 | struct task_struct *p = current; | |
1273 | int cpu = smp_processor_id(); | |
1274 | ||
1275 | /* Note: this timer irq context must be accounted for as well. */ | |
fa13a5a1 | 1276 | account_process_tick(p, user_tick); |
1da177e4 | 1277 | run_local_timers(); |
a157229c | 1278 | rcu_check_callbacks(cpu, user_tick); |
b845b517 | 1279 | printk_tick(); |
fe432200 | 1280 | perf_event_do_pending(); |
1da177e4 | 1281 | scheduler_tick(); |
6819457d | 1282 | run_posix_cpu_timers(p); |
1da177e4 LT |
1283 | } |
1284 | ||
1da177e4 LT |
1285 | /* |
1286 | * This function runs timers and the timer-tq in bottom half context. | |
1287 | */ | |
1288 | static void run_timer_softirq(struct softirq_action *h) | |
1289 | { | |
a6fa8e5a | 1290 | struct tvec_base *base = __get_cpu_var(tvec_bases); |
1da177e4 | 1291 | |
d3d74453 | 1292 | hrtimer_run_pending(); |
82f67cd9 | 1293 | |
1da177e4 LT |
1294 | if (time_after_eq(jiffies, base->timer_jiffies)) |
1295 | __run_timers(base); | |
1296 | } | |
1297 | ||
1298 | /* | |
1299 | * Called by the local, per-CPU timer interrupt on SMP. | |
1300 | */ | |
1301 | void run_local_timers(void) | |
1302 | { | |
d3d74453 | 1303 | hrtimer_run_queues(); |
1da177e4 | 1304 | raise_softirq(TIMER_SOFTIRQ); |
6687a97d | 1305 | softlockup_tick(); |
1da177e4 LT |
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); |