Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * linux/kernel/timer.c | |
3 | * | |
8524070b | 4 | * Kernel internal timers, basic process system calls |
1da177e4 LT |
5 | * |
6 | * Copyright (C) 1991, 1992 Linus Torvalds | |
7 | * | |
8 | * 1997-01-28 Modified by Finn Arne Gangstad to make timers scale better. | |
9 | * | |
10 | * 1997-09-10 Updated NTP code according to technical memorandum Jan '96 | |
11 | * "A Kernel Model for Precision Timekeeping" by Dave Mills | |
12 | * 1998-12-24 Fixed a xtime SMP race (we need the xtime_lock rw spinlock to | |
13 | * serialize accesses to xtime/lost_ticks). | |
14 | * Copyright (C) 1998 Andrea Arcangeli | |
15 | * 1999-03-10 Improved NTP compatibility by Ulrich Windl | |
16 | * 2002-05-31 Move sys_sysinfo here and make its locking sane, Robert Love | |
17 | * 2000-10-05 Implemented scalable SMP per-CPU timer handling. | |
18 | * Copyright (C) 2000, 2001, 2002 Ingo Molnar | |
19 | * Designed by David S. Miller, Alexey Kuznetsov and Ingo Molnar | |
20 | */ | |
21 | ||
22 | #include <linux/kernel_stat.h> | |
23 | #include <linux/module.h> | |
24 | #include <linux/interrupt.h> | |
25 | #include <linux/percpu.h> | |
26 | #include <linux/init.h> | |
27 | #include <linux/mm.h> | |
28 | #include <linux/swap.h> | |
b488893a | 29 | #include <linux/pid_namespace.h> |
1da177e4 LT |
30 | #include <linux/notifier.h> |
31 | #include <linux/thread_info.h> | |
32 | #include <linux/time.h> | |
33 | #include <linux/jiffies.h> | |
34 | #include <linux/posix-timers.h> | |
35 | #include <linux/cpu.h> | |
36 | #include <linux/syscalls.h> | |
97a41e26 | 37 | #include <linux/delay.h> |
79bf2bb3 | 38 | #include <linux/tick.h> |
82f67cd9 | 39 | #include <linux/kallsyms.h> |
1da177e4 LT |
40 | |
41 | #include <asm/uaccess.h> | |
42 | #include <asm/unistd.h> | |
43 | #include <asm/div64.h> | |
44 | #include <asm/timex.h> | |
45 | #include <asm/io.h> | |
46 | ||
ecea8d19 TG |
47 | u64 jiffies_64 __cacheline_aligned_in_smp = INITIAL_JIFFIES; |
48 | ||
49 | EXPORT_SYMBOL(jiffies_64); | |
50 | ||
1da177e4 LT |
51 | /* |
52 | * per-CPU timer vector definitions: | |
53 | */ | |
1da177e4 LT |
54 | #define TVN_BITS (CONFIG_BASE_SMALL ? 4 : 6) |
55 | #define TVR_BITS (CONFIG_BASE_SMALL ? 6 : 8) | |
56 | #define TVN_SIZE (1 << TVN_BITS) | |
57 | #define TVR_SIZE (1 << TVR_BITS) | |
58 | #define TVN_MASK (TVN_SIZE - 1) | |
59 | #define TVR_MASK (TVR_SIZE - 1) | |
60 | ||
a6fa8e5a | 61 | struct tvec { |
1da177e4 | 62 | struct list_head vec[TVN_SIZE]; |
a6fa8e5a | 63 | }; |
1da177e4 | 64 | |
a6fa8e5a | 65 | struct tvec_root { |
1da177e4 | 66 | struct list_head vec[TVR_SIZE]; |
a6fa8e5a | 67 | }; |
1da177e4 | 68 | |
a6fa8e5a | 69 | struct tvec_base { |
3691c519 ON |
70 | spinlock_t lock; |
71 | struct timer_list *running_timer; | |
1da177e4 | 72 | unsigned long timer_jiffies; |
a6fa8e5a PM |
73 | struct tvec_root tv1; |
74 | struct tvec tv2; | |
75 | struct tvec tv3; | |
76 | struct tvec tv4; | |
77 | struct tvec tv5; | |
6e453a67 | 78 | } ____cacheline_aligned; |
1da177e4 | 79 | |
a6fa8e5a | 80 | struct tvec_base boot_tvec_bases; |
3691c519 | 81 | EXPORT_SYMBOL(boot_tvec_bases); |
a6fa8e5a | 82 | static DEFINE_PER_CPU(struct tvec_base *, tvec_bases) = &boot_tvec_bases; |
1da177e4 | 83 | |
6e453a67 | 84 | /* |
a6fa8e5a | 85 | * Note that all tvec_bases are 2 byte aligned and lower bit of |
6e453a67 VP |
86 | * base in timer_list is guaranteed to be zero. Use the LSB for |
87 | * the new flag to indicate whether the timer is deferrable | |
88 | */ | |
89 | #define TBASE_DEFERRABLE_FLAG (0x1) | |
90 | ||
91 | /* Functions below help us manage 'deferrable' flag */ | |
a6fa8e5a | 92 | static inline unsigned int tbase_get_deferrable(struct tvec_base *base) |
6e453a67 | 93 | { |
e9910846 | 94 | return ((unsigned int)(unsigned long)base & TBASE_DEFERRABLE_FLAG); |
6e453a67 VP |
95 | } |
96 | ||
a6fa8e5a | 97 | static inline struct tvec_base *tbase_get_base(struct tvec_base *base) |
6e453a67 | 98 | { |
a6fa8e5a | 99 | return ((struct tvec_base *)((unsigned long)base & ~TBASE_DEFERRABLE_FLAG)); |
6e453a67 VP |
100 | } |
101 | ||
102 | static inline void timer_set_deferrable(struct timer_list *timer) | |
103 | { | |
a6fa8e5a | 104 | timer->base = ((struct tvec_base *)((unsigned long)(timer->base) | |
6819457d | 105 | TBASE_DEFERRABLE_FLAG)); |
6e453a67 VP |
106 | } |
107 | ||
108 | static inline void | |
a6fa8e5a | 109 | timer_set_base(struct timer_list *timer, struct tvec_base *new_base) |
6e453a67 | 110 | { |
a6fa8e5a | 111 | timer->base = (struct tvec_base *)((unsigned long)(new_base) | |
6819457d | 112 | tbase_get_deferrable(timer->base)); |
6e453a67 VP |
113 | } |
114 | ||
4c36a5de AV |
115 | /** |
116 | * __round_jiffies - function to round jiffies to a full second | |
117 | * @j: the time in (absolute) jiffies that should be rounded | |
118 | * @cpu: the processor number on which the timeout will happen | |
119 | * | |
72fd4a35 | 120 | * __round_jiffies() rounds an absolute time in the future (in jiffies) |
4c36a5de AV |
121 | * up or down to (approximately) full seconds. This is useful for timers |
122 | * for which the exact time they fire does not matter too much, as long as | |
123 | * they fire approximately every X seconds. | |
124 | * | |
125 | * By rounding these timers to whole seconds, all such timers will fire | |
126 | * at the same time, rather than at various times spread out. The goal | |
127 | * of this is to have the CPU wake up less, which saves power. | |
128 | * | |
129 | * The exact rounding is skewed for each processor to avoid all | |
130 | * processors firing at the exact same time, which could lead | |
131 | * to lock contention or spurious cache line bouncing. | |
132 | * | |
72fd4a35 | 133 | * The return value is the rounded version of the @j parameter. |
4c36a5de AV |
134 | */ |
135 | unsigned long __round_jiffies(unsigned long j, int cpu) | |
136 | { | |
137 | int rem; | |
138 | unsigned long original = j; | |
139 | ||
140 | /* | |
141 | * We don't want all cpus firing their timers at once hitting the | |
142 | * same lock or cachelines, so we skew each extra cpu with an extra | |
143 | * 3 jiffies. This 3 jiffies came originally from the mm/ code which | |
144 | * already did this. | |
145 | * The skew is done by adding 3*cpunr, then round, then subtract this | |
146 | * extra offset again. | |
147 | */ | |
148 | j += cpu * 3; | |
149 | ||
150 | rem = j % HZ; | |
151 | ||
152 | /* | |
153 | * If the target jiffie is just after a whole second (which can happen | |
154 | * due to delays of the timer irq, long irq off times etc etc) then | |
155 | * we should round down to the whole second, not up. Use 1/4th second | |
156 | * as cutoff for this rounding as an extreme upper bound for this. | |
157 | */ | |
158 | if (rem < HZ/4) /* round down */ | |
159 | j = j - rem; | |
160 | else /* round up */ | |
161 | j = j - rem + HZ; | |
162 | ||
163 | /* now that we have rounded, subtract the extra skew again */ | |
164 | j -= cpu * 3; | |
165 | ||
166 | if (j <= jiffies) /* rounding ate our timeout entirely; */ | |
167 | return original; | |
168 | return j; | |
169 | } | |
170 | EXPORT_SYMBOL_GPL(__round_jiffies); | |
171 | ||
172 | /** | |
173 | * __round_jiffies_relative - function to round jiffies to a full second | |
174 | * @j: the time in (relative) jiffies that should be rounded | |
175 | * @cpu: the processor number on which the timeout will happen | |
176 | * | |
72fd4a35 | 177 | * __round_jiffies_relative() rounds a time delta in the future (in jiffies) |
4c36a5de AV |
178 | * up or down to (approximately) full seconds. This is useful for timers |
179 | * for which the exact time they fire does not matter too much, as long as | |
180 | * they fire approximately every X seconds. | |
181 | * | |
182 | * By rounding these timers to whole seconds, all such timers will fire | |
183 | * at the same time, rather than at various times spread out. The goal | |
184 | * of this is to have the CPU wake up less, which saves power. | |
185 | * | |
186 | * The exact rounding is skewed for each processor to avoid all | |
187 | * processors firing at the exact same time, which could lead | |
188 | * to lock contention or spurious cache line bouncing. | |
189 | * | |
72fd4a35 | 190 | * The return value is the rounded version of the @j parameter. |
4c36a5de AV |
191 | */ |
192 | unsigned long __round_jiffies_relative(unsigned long j, int cpu) | |
193 | { | |
194 | /* | |
195 | * In theory the following code can skip a jiffy in case jiffies | |
196 | * increments right between the addition and the later subtraction. | |
197 | * However since the entire point of this function is to use approximate | |
198 | * timeouts, it's entirely ok to not handle that. | |
199 | */ | |
200 | return __round_jiffies(j + jiffies, cpu) - jiffies; | |
201 | } | |
202 | EXPORT_SYMBOL_GPL(__round_jiffies_relative); | |
203 | ||
204 | /** | |
205 | * round_jiffies - function to round jiffies to a full second | |
206 | * @j: the time in (absolute) jiffies that should be rounded | |
207 | * | |
72fd4a35 | 208 | * round_jiffies() rounds an absolute time in the future (in jiffies) |
4c36a5de AV |
209 | * up or down to (approximately) full seconds. This is useful for timers |
210 | * for which the exact time they fire does not matter too much, as long as | |
211 | * they fire approximately every X seconds. | |
212 | * | |
213 | * By rounding these timers to whole seconds, all such timers will fire | |
214 | * at the same time, rather than at various times spread out. The goal | |
215 | * of this is to have the CPU wake up less, which saves power. | |
216 | * | |
72fd4a35 | 217 | * The return value is the rounded version of the @j parameter. |
4c36a5de AV |
218 | */ |
219 | unsigned long round_jiffies(unsigned long j) | |
220 | { | |
221 | return __round_jiffies(j, raw_smp_processor_id()); | |
222 | } | |
223 | EXPORT_SYMBOL_GPL(round_jiffies); | |
224 | ||
225 | /** | |
226 | * round_jiffies_relative - function to round jiffies to a full second | |
227 | * @j: the time in (relative) jiffies that should be rounded | |
228 | * | |
72fd4a35 | 229 | * round_jiffies_relative() rounds a time delta in the future (in jiffies) |
4c36a5de AV |
230 | * up or down to (approximately) full seconds. This is useful for timers |
231 | * for which the exact time they fire does not matter too much, as long as | |
232 | * they fire approximately every X seconds. | |
233 | * | |
234 | * By rounding these timers to whole seconds, all such timers will fire | |
235 | * at the same time, rather than at various times spread out. The goal | |
236 | * of this is to have the CPU wake up less, which saves power. | |
237 | * | |
72fd4a35 | 238 | * The return value is the rounded version of the @j parameter. |
4c36a5de AV |
239 | */ |
240 | unsigned long round_jiffies_relative(unsigned long j) | |
241 | { | |
242 | return __round_jiffies_relative(j, raw_smp_processor_id()); | |
243 | } | |
244 | EXPORT_SYMBOL_GPL(round_jiffies_relative); | |
245 | ||
246 | ||
a6fa8e5a | 247 | static inline void set_running_timer(struct tvec_base *base, |
1da177e4 LT |
248 | struct timer_list *timer) |
249 | { | |
250 | #ifdef CONFIG_SMP | |
3691c519 | 251 | base->running_timer = timer; |
1da177e4 LT |
252 | #endif |
253 | } | |
254 | ||
a6fa8e5a | 255 | static void internal_add_timer(struct tvec_base *base, struct timer_list *timer) |
1da177e4 LT |
256 | { |
257 | unsigned long expires = timer->expires; | |
258 | unsigned long idx = expires - base->timer_jiffies; | |
259 | struct list_head *vec; | |
260 | ||
261 | if (idx < TVR_SIZE) { | |
262 | int i = expires & TVR_MASK; | |
263 | vec = base->tv1.vec + i; | |
264 | } else if (idx < 1 << (TVR_BITS + TVN_BITS)) { | |
265 | int i = (expires >> TVR_BITS) & TVN_MASK; | |
266 | vec = base->tv2.vec + i; | |
267 | } else if (idx < 1 << (TVR_BITS + 2 * TVN_BITS)) { | |
268 | int i = (expires >> (TVR_BITS + TVN_BITS)) & TVN_MASK; | |
269 | vec = base->tv3.vec + i; | |
270 | } else if (idx < 1 << (TVR_BITS + 3 * TVN_BITS)) { | |
271 | int i = (expires >> (TVR_BITS + 2 * TVN_BITS)) & TVN_MASK; | |
272 | vec = base->tv4.vec + i; | |
273 | } else if ((signed long) idx < 0) { | |
274 | /* | |
275 | * Can happen if you add a timer with expires == jiffies, | |
276 | * or you set a timer to go off in the past | |
277 | */ | |
278 | vec = base->tv1.vec + (base->timer_jiffies & TVR_MASK); | |
279 | } else { | |
280 | int i; | |
281 | /* If the timeout is larger than 0xffffffff on 64-bit | |
282 | * architectures then we use the maximum timeout: | |
283 | */ | |
284 | if (idx > 0xffffffffUL) { | |
285 | idx = 0xffffffffUL; | |
286 | expires = idx + base->timer_jiffies; | |
287 | } | |
288 | i = (expires >> (TVR_BITS + 3 * TVN_BITS)) & TVN_MASK; | |
289 | vec = base->tv5.vec + i; | |
290 | } | |
291 | /* | |
292 | * Timers are FIFO: | |
293 | */ | |
294 | list_add_tail(&timer->entry, vec); | |
295 | } | |
296 | ||
82f67cd9 IM |
297 | #ifdef CONFIG_TIMER_STATS |
298 | void __timer_stats_timer_set_start_info(struct timer_list *timer, void *addr) | |
299 | { | |
300 | if (timer->start_site) | |
301 | return; | |
302 | ||
303 | timer->start_site = addr; | |
304 | memcpy(timer->start_comm, current->comm, TASK_COMM_LEN); | |
305 | timer->start_pid = current->pid; | |
306 | } | |
c5c061b8 VP |
307 | |
308 | static void timer_stats_account_timer(struct timer_list *timer) | |
309 | { | |
310 | unsigned int flag = 0; | |
311 | ||
312 | if (unlikely(tbase_get_deferrable(timer->base))) | |
313 | flag |= TIMER_STATS_FLAG_DEFERRABLE; | |
314 | ||
315 | timer_stats_update_stats(timer, timer->start_pid, timer->start_site, | |
316 | timer->function, timer->start_comm, flag); | |
317 | } | |
318 | ||
319 | #else | |
320 | static void timer_stats_account_timer(struct timer_list *timer) {} | |
82f67cd9 IM |
321 | #endif |
322 | ||
2aae4a10 | 323 | /** |
55c888d6 ON |
324 | * init_timer - initialize a timer. |
325 | * @timer: the timer to be initialized | |
326 | * | |
327 | * init_timer() must be done to a timer prior calling *any* of the | |
328 | * other timer functions. | |
329 | */ | |
7ad5b3a5 | 330 | void init_timer(struct timer_list *timer) |
55c888d6 ON |
331 | { |
332 | timer->entry.next = NULL; | |
bfe5d834 | 333 | timer->base = __raw_get_cpu_var(tvec_bases); |
82f67cd9 IM |
334 | #ifdef CONFIG_TIMER_STATS |
335 | timer->start_site = NULL; | |
336 | timer->start_pid = -1; | |
337 | memset(timer->start_comm, 0, TASK_COMM_LEN); | |
338 | #endif | |
55c888d6 ON |
339 | } |
340 | EXPORT_SYMBOL(init_timer); | |
341 | ||
7ad5b3a5 | 342 | void init_timer_deferrable(struct timer_list *timer) |
6e453a67 VP |
343 | { |
344 | init_timer(timer); | |
345 | timer_set_deferrable(timer); | |
346 | } | |
347 | EXPORT_SYMBOL(init_timer_deferrable); | |
348 | ||
55c888d6 | 349 | static inline void detach_timer(struct timer_list *timer, |
82f67cd9 | 350 | int clear_pending) |
55c888d6 ON |
351 | { |
352 | struct list_head *entry = &timer->entry; | |
353 | ||
354 | __list_del(entry->prev, entry->next); | |
355 | if (clear_pending) | |
356 | entry->next = NULL; | |
357 | entry->prev = LIST_POISON2; | |
358 | } | |
359 | ||
360 | /* | |
3691c519 | 361 | * We are using hashed locking: holding per_cpu(tvec_bases).lock |
55c888d6 ON |
362 | * means that all timers which are tied to this base via timer->base are |
363 | * locked, and the base itself is locked too. | |
364 | * | |
365 | * So __run_timers/migrate_timers can safely modify all timers which could | |
366 | * be found on ->tvX lists. | |
367 | * | |
368 | * When the timer's base is locked, and the timer removed from list, it is | |
369 | * possible to set timer->base = NULL and drop the lock: the timer remains | |
370 | * locked. | |
371 | */ | |
a6fa8e5a | 372 | static struct tvec_base *lock_timer_base(struct timer_list *timer, |
55c888d6 | 373 | unsigned long *flags) |
89e7e374 | 374 | __acquires(timer->base->lock) |
55c888d6 | 375 | { |
a6fa8e5a | 376 | struct tvec_base *base; |
55c888d6 ON |
377 | |
378 | for (;;) { | |
a6fa8e5a | 379 | struct tvec_base *prelock_base = timer->base; |
6e453a67 | 380 | base = tbase_get_base(prelock_base); |
55c888d6 ON |
381 | if (likely(base != NULL)) { |
382 | spin_lock_irqsave(&base->lock, *flags); | |
6e453a67 | 383 | if (likely(prelock_base == timer->base)) |
55c888d6 ON |
384 | return base; |
385 | /* The timer has migrated to another CPU */ | |
386 | spin_unlock_irqrestore(&base->lock, *flags); | |
387 | } | |
388 | cpu_relax(); | |
389 | } | |
390 | } | |
391 | ||
1da177e4 LT |
392 | int __mod_timer(struct timer_list *timer, unsigned long expires) |
393 | { | |
a6fa8e5a | 394 | struct tvec_base *base, *new_base; |
1da177e4 LT |
395 | unsigned long flags; |
396 | int ret = 0; | |
397 | ||
82f67cd9 | 398 | timer_stats_timer_set_start_info(timer); |
1da177e4 | 399 | BUG_ON(!timer->function); |
1da177e4 | 400 | |
55c888d6 ON |
401 | base = lock_timer_base(timer, &flags); |
402 | ||
403 | if (timer_pending(timer)) { | |
404 | detach_timer(timer, 0); | |
405 | ret = 1; | |
406 | } | |
407 | ||
a4a6198b | 408 | new_base = __get_cpu_var(tvec_bases); |
1da177e4 | 409 | |
3691c519 | 410 | if (base != new_base) { |
1da177e4 | 411 | /* |
55c888d6 ON |
412 | * We are trying to schedule the timer on the local CPU. |
413 | * However we can't change timer's base while it is running, | |
414 | * otherwise del_timer_sync() can't detect that the timer's | |
415 | * handler yet has not finished. This also guarantees that | |
416 | * the timer is serialized wrt itself. | |
1da177e4 | 417 | */ |
a2c348fe | 418 | if (likely(base->running_timer != timer)) { |
55c888d6 | 419 | /* See the comment in lock_timer_base() */ |
6e453a67 | 420 | timer_set_base(timer, NULL); |
55c888d6 | 421 | spin_unlock(&base->lock); |
a2c348fe ON |
422 | base = new_base; |
423 | spin_lock(&base->lock); | |
6e453a67 | 424 | timer_set_base(timer, base); |
1da177e4 LT |
425 | } |
426 | } | |
427 | ||
1da177e4 | 428 | timer->expires = expires; |
a2c348fe ON |
429 | internal_add_timer(base, timer); |
430 | spin_unlock_irqrestore(&base->lock, flags); | |
1da177e4 LT |
431 | |
432 | return ret; | |
433 | } | |
434 | ||
435 | EXPORT_SYMBOL(__mod_timer); | |
436 | ||
2aae4a10 | 437 | /** |
1da177e4 LT |
438 | * add_timer_on - start a timer on a particular CPU |
439 | * @timer: the timer to be added | |
440 | * @cpu: the CPU to start it on | |
441 | * | |
442 | * This is not very scalable on SMP. Double adds are not possible. | |
443 | */ | |
444 | void add_timer_on(struct timer_list *timer, int cpu) | |
445 | { | |
a6fa8e5a | 446 | struct tvec_base *base = per_cpu(tvec_bases, cpu); |
6819457d | 447 | unsigned long flags; |
55c888d6 | 448 | |
82f67cd9 | 449 | timer_stats_timer_set_start_info(timer); |
6819457d | 450 | BUG_ON(timer_pending(timer) || !timer->function); |
3691c519 | 451 | spin_lock_irqsave(&base->lock, flags); |
6e453a67 | 452 | timer_set_base(timer, base); |
1da177e4 | 453 | internal_add_timer(base, timer); |
06d8308c TG |
454 | /* |
455 | * Check whether the other CPU is idle and needs to be | |
456 | * triggered to reevaluate the timer wheel when nohz is | |
457 | * active. We are protected against the other CPU fiddling | |
458 | * with the timer by holding the timer base lock. This also | |
459 | * makes sure that a CPU on the way to idle can not evaluate | |
460 | * the timer wheel. | |
461 | */ | |
462 | wake_up_idle_cpu(cpu); | |
3691c519 | 463 | spin_unlock_irqrestore(&base->lock, flags); |
1da177e4 LT |
464 | } |
465 | ||
2aae4a10 | 466 | /** |
1da177e4 LT |
467 | * mod_timer - modify a timer's timeout |
468 | * @timer: the timer to be modified | |
2aae4a10 | 469 | * @expires: new timeout in jiffies |
1da177e4 | 470 | * |
72fd4a35 | 471 | * mod_timer() is a more efficient way to update the expire field of an |
1da177e4 LT |
472 | * active timer (if the timer is inactive it will be activated) |
473 | * | |
474 | * mod_timer(timer, expires) is equivalent to: | |
475 | * | |
476 | * del_timer(timer); timer->expires = expires; add_timer(timer); | |
477 | * | |
478 | * Note that if there are multiple unserialized concurrent users of the | |
479 | * same timer, then mod_timer() is the only safe way to modify the timeout, | |
480 | * since add_timer() cannot modify an already running timer. | |
481 | * | |
482 | * The function returns whether it has modified a pending timer or not. | |
483 | * (ie. mod_timer() of an inactive timer returns 0, mod_timer() of an | |
484 | * active timer returns 1.) | |
485 | */ | |
486 | int mod_timer(struct timer_list *timer, unsigned long expires) | |
487 | { | |
488 | BUG_ON(!timer->function); | |
489 | ||
82f67cd9 | 490 | timer_stats_timer_set_start_info(timer); |
1da177e4 LT |
491 | /* |
492 | * This is a common optimization triggered by the | |
493 | * networking code - if the timer is re-modified | |
494 | * to be the same thing then just return: | |
495 | */ | |
496 | if (timer->expires == expires && timer_pending(timer)) | |
497 | return 1; | |
498 | ||
499 | return __mod_timer(timer, expires); | |
500 | } | |
501 | ||
502 | EXPORT_SYMBOL(mod_timer); | |
503 | ||
2aae4a10 | 504 | /** |
1da177e4 LT |
505 | * del_timer - deactive a timer. |
506 | * @timer: the timer to be deactivated | |
507 | * | |
508 | * del_timer() deactivates a timer - this works on both active and inactive | |
509 | * timers. | |
510 | * | |
511 | * The function returns whether it has deactivated a pending timer or not. | |
512 | * (ie. del_timer() of an inactive timer returns 0, del_timer() of an | |
513 | * active timer returns 1.) | |
514 | */ | |
515 | int del_timer(struct timer_list *timer) | |
516 | { | |
a6fa8e5a | 517 | struct tvec_base *base; |
1da177e4 | 518 | unsigned long flags; |
55c888d6 | 519 | int ret = 0; |
1da177e4 | 520 | |
82f67cd9 | 521 | timer_stats_timer_clear_start_info(timer); |
55c888d6 ON |
522 | if (timer_pending(timer)) { |
523 | base = lock_timer_base(timer, &flags); | |
524 | if (timer_pending(timer)) { | |
525 | detach_timer(timer, 1); | |
526 | ret = 1; | |
527 | } | |
1da177e4 | 528 | spin_unlock_irqrestore(&base->lock, flags); |
1da177e4 | 529 | } |
1da177e4 | 530 | |
55c888d6 | 531 | return ret; |
1da177e4 LT |
532 | } |
533 | ||
534 | EXPORT_SYMBOL(del_timer); | |
535 | ||
536 | #ifdef CONFIG_SMP | |
2aae4a10 REB |
537 | /** |
538 | * try_to_del_timer_sync - Try to deactivate a timer | |
539 | * @timer: timer do del | |
540 | * | |
fd450b73 ON |
541 | * This function tries to deactivate a timer. Upon successful (ret >= 0) |
542 | * exit the timer is not queued and the handler is not running on any CPU. | |
543 | * | |
544 | * It must not be called from interrupt contexts. | |
545 | */ | |
546 | int try_to_del_timer_sync(struct timer_list *timer) | |
547 | { | |
a6fa8e5a | 548 | struct tvec_base *base; |
fd450b73 ON |
549 | unsigned long flags; |
550 | int ret = -1; | |
551 | ||
552 | base = lock_timer_base(timer, &flags); | |
553 | ||
554 | if (base->running_timer == timer) | |
555 | goto out; | |
556 | ||
557 | ret = 0; | |
558 | if (timer_pending(timer)) { | |
559 | detach_timer(timer, 1); | |
560 | ret = 1; | |
561 | } | |
562 | out: | |
563 | spin_unlock_irqrestore(&base->lock, flags); | |
564 | ||
565 | return ret; | |
566 | } | |
567 | ||
e19dff1f DH |
568 | EXPORT_SYMBOL(try_to_del_timer_sync); |
569 | ||
2aae4a10 | 570 | /** |
1da177e4 LT |
571 | * del_timer_sync - deactivate a timer and wait for the handler to finish. |
572 | * @timer: the timer to be deactivated | |
573 | * | |
574 | * This function only differs from del_timer() on SMP: besides deactivating | |
575 | * the timer it also makes sure the handler has finished executing on other | |
576 | * CPUs. | |
577 | * | |
72fd4a35 | 578 | * Synchronization rules: Callers must prevent restarting of the timer, |
1da177e4 LT |
579 | * otherwise this function is meaningless. It must not be called from |
580 | * interrupt contexts. The caller must not hold locks which would prevent | |
55c888d6 ON |
581 | * completion of the timer's handler. The timer's handler must not call |
582 | * add_timer_on(). Upon exit the timer is not queued and the handler is | |
583 | * not running on any CPU. | |
1da177e4 LT |
584 | * |
585 | * The function returns whether it has deactivated a pending timer or not. | |
1da177e4 LT |
586 | */ |
587 | int del_timer_sync(struct timer_list *timer) | |
588 | { | |
fd450b73 ON |
589 | for (;;) { |
590 | int ret = try_to_del_timer_sync(timer); | |
591 | if (ret >= 0) | |
592 | return ret; | |
a0009652 | 593 | cpu_relax(); |
fd450b73 | 594 | } |
1da177e4 | 595 | } |
1da177e4 | 596 | |
55c888d6 | 597 | EXPORT_SYMBOL(del_timer_sync); |
1da177e4 LT |
598 | #endif |
599 | ||
a6fa8e5a | 600 | static int cascade(struct tvec_base *base, struct tvec *tv, int index) |
1da177e4 LT |
601 | { |
602 | /* cascade all the timers from tv up one level */ | |
3439dd86 P |
603 | struct timer_list *timer, *tmp; |
604 | struct list_head tv_list; | |
605 | ||
606 | list_replace_init(tv->vec + index, &tv_list); | |
1da177e4 | 607 | |
1da177e4 | 608 | /* |
3439dd86 P |
609 | * We are removing _all_ timers from the list, so we |
610 | * don't have to detach them individually. | |
1da177e4 | 611 | */ |
3439dd86 | 612 | list_for_each_entry_safe(timer, tmp, &tv_list, entry) { |
6e453a67 | 613 | BUG_ON(tbase_get_base(timer->base) != base); |
3439dd86 | 614 | internal_add_timer(base, timer); |
1da177e4 | 615 | } |
1da177e4 LT |
616 | |
617 | return index; | |
618 | } | |
619 | ||
2aae4a10 REB |
620 | #define INDEX(N) ((base->timer_jiffies >> (TVR_BITS + (N) * TVN_BITS)) & TVN_MASK) |
621 | ||
622 | /** | |
1da177e4 LT |
623 | * __run_timers - run all expired timers (if any) on this CPU. |
624 | * @base: the timer vector to be processed. | |
625 | * | |
626 | * This function cascades all vectors and executes all expired timer | |
627 | * vectors. | |
628 | */ | |
a6fa8e5a | 629 | static inline void __run_timers(struct tvec_base *base) |
1da177e4 LT |
630 | { |
631 | struct timer_list *timer; | |
632 | ||
3691c519 | 633 | spin_lock_irq(&base->lock); |
1da177e4 | 634 | while (time_after_eq(jiffies, base->timer_jiffies)) { |
626ab0e6 | 635 | struct list_head work_list; |
1da177e4 | 636 | struct list_head *head = &work_list; |
6819457d | 637 | int index = base->timer_jiffies & TVR_MASK; |
626ab0e6 | 638 | |
1da177e4 LT |
639 | /* |
640 | * Cascade timers: | |
641 | */ | |
642 | if (!index && | |
643 | (!cascade(base, &base->tv2, INDEX(0))) && | |
644 | (!cascade(base, &base->tv3, INDEX(1))) && | |
645 | !cascade(base, &base->tv4, INDEX(2))) | |
646 | cascade(base, &base->tv5, INDEX(3)); | |
626ab0e6 ON |
647 | ++base->timer_jiffies; |
648 | list_replace_init(base->tv1.vec + index, &work_list); | |
55c888d6 | 649 | while (!list_empty(head)) { |
1da177e4 LT |
650 | void (*fn)(unsigned long); |
651 | unsigned long data; | |
652 | ||
b5e61818 | 653 | timer = list_first_entry(head, struct timer_list,entry); |
6819457d TG |
654 | fn = timer->function; |
655 | data = timer->data; | |
1da177e4 | 656 | |
82f67cd9 IM |
657 | timer_stats_account_timer(timer); |
658 | ||
1da177e4 | 659 | set_running_timer(base, timer); |
55c888d6 | 660 | detach_timer(timer, 1); |
3691c519 | 661 | spin_unlock_irq(&base->lock); |
1da177e4 | 662 | { |
be5b4fbd | 663 | int preempt_count = preempt_count(); |
1da177e4 LT |
664 | fn(data); |
665 | if (preempt_count != preempt_count()) { | |
4c9dc641 | 666 | printk(KERN_ERR "huh, entered %p " |
be5b4fbd JJ |
667 | "with preempt_count %08x, exited" |
668 | " with %08x?\n", | |
669 | fn, preempt_count, | |
670 | preempt_count()); | |
1da177e4 LT |
671 | BUG(); |
672 | } | |
673 | } | |
3691c519 | 674 | spin_lock_irq(&base->lock); |
1da177e4 LT |
675 | } |
676 | } | |
677 | set_running_timer(base, NULL); | |
3691c519 | 678 | spin_unlock_irq(&base->lock); |
1da177e4 LT |
679 | } |
680 | ||
fd064b9b | 681 | #if defined(CONFIG_NO_IDLE_HZ) || defined(CONFIG_NO_HZ) |
1da177e4 LT |
682 | /* |
683 | * Find out when the next timer event is due to happen. This | |
684 | * is used on S/390 to stop all activity when a cpus is idle. | |
685 | * This functions needs to be called disabled. | |
686 | */ | |
a6fa8e5a | 687 | static unsigned long __next_timer_interrupt(struct tvec_base *base) |
1da177e4 | 688 | { |
1cfd6849 | 689 | unsigned long timer_jiffies = base->timer_jiffies; |
eaad084b | 690 | unsigned long expires = timer_jiffies + NEXT_TIMER_MAX_DELTA; |
1cfd6849 | 691 | int index, slot, array, found = 0; |
1da177e4 | 692 | struct timer_list *nte; |
a6fa8e5a | 693 | struct tvec *varray[4]; |
1da177e4 LT |
694 | |
695 | /* Look for timer events in tv1. */ | |
1cfd6849 | 696 | index = slot = timer_jiffies & TVR_MASK; |
1da177e4 | 697 | do { |
1cfd6849 | 698 | list_for_each_entry(nte, base->tv1.vec + slot, entry) { |
6819457d TG |
699 | if (tbase_get_deferrable(nte->base)) |
700 | continue; | |
6e453a67 | 701 | |
1cfd6849 | 702 | found = 1; |
1da177e4 | 703 | expires = nte->expires; |
1cfd6849 TG |
704 | /* Look at the cascade bucket(s)? */ |
705 | if (!index || slot < index) | |
706 | goto cascade; | |
707 | return expires; | |
1da177e4 | 708 | } |
1cfd6849 TG |
709 | slot = (slot + 1) & TVR_MASK; |
710 | } while (slot != index); | |
711 | ||
712 | cascade: | |
713 | /* Calculate the next cascade event */ | |
714 | if (index) | |
715 | timer_jiffies += TVR_SIZE - index; | |
716 | timer_jiffies >>= TVR_BITS; | |
1da177e4 LT |
717 | |
718 | /* Check tv2-tv5. */ | |
719 | varray[0] = &base->tv2; | |
720 | varray[1] = &base->tv3; | |
721 | varray[2] = &base->tv4; | |
722 | varray[3] = &base->tv5; | |
1cfd6849 TG |
723 | |
724 | for (array = 0; array < 4; array++) { | |
a6fa8e5a | 725 | struct tvec *varp = varray[array]; |
1cfd6849 TG |
726 | |
727 | index = slot = timer_jiffies & TVN_MASK; | |
1da177e4 | 728 | do { |
1cfd6849 TG |
729 | list_for_each_entry(nte, varp->vec + slot, entry) { |
730 | found = 1; | |
1da177e4 LT |
731 | if (time_before(nte->expires, expires)) |
732 | expires = nte->expires; | |
1cfd6849 TG |
733 | } |
734 | /* | |
735 | * Do we still search for the first timer or are | |
736 | * we looking up the cascade buckets ? | |
737 | */ | |
738 | if (found) { | |
739 | /* Look at the cascade bucket(s)? */ | |
740 | if (!index || slot < index) | |
741 | break; | |
742 | return expires; | |
743 | } | |
744 | slot = (slot + 1) & TVN_MASK; | |
745 | } while (slot != index); | |
746 | ||
747 | if (index) | |
748 | timer_jiffies += TVN_SIZE - index; | |
749 | timer_jiffies >>= TVN_BITS; | |
1da177e4 | 750 | } |
1cfd6849 TG |
751 | return expires; |
752 | } | |
69239749 | 753 | |
1cfd6849 TG |
754 | /* |
755 | * Check, if the next hrtimer event is before the next timer wheel | |
756 | * event: | |
757 | */ | |
758 | static unsigned long cmp_next_hrtimer_event(unsigned long now, | |
759 | unsigned long expires) | |
760 | { | |
761 | ktime_t hr_delta = hrtimer_get_next_event(); | |
762 | struct timespec tsdelta; | |
9501b6cf | 763 | unsigned long delta; |
1cfd6849 TG |
764 | |
765 | if (hr_delta.tv64 == KTIME_MAX) | |
766 | return expires; | |
0662b713 | 767 | |
9501b6cf TG |
768 | /* |
769 | * Expired timer available, let it expire in the next tick | |
770 | */ | |
771 | if (hr_delta.tv64 <= 0) | |
772 | return now + 1; | |
69239749 | 773 | |
1cfd6849 | 774 | tsdelta = ktime_to_timespec(hr_delta); |
9501b6cf | 775 | delta = timespec_to_jiffies(&tsdelta); |
eaad084b TG |
776 | |
777 | /* | |
778 | * Limit the delta to the max value, which is checked in | |
779 | * tick_nohz_stop_sched_tick(): | |
780 | */ | |
781 | if (delta > NEXT_TIMER_MAX_DELTA) | |
782 | delta = NEXT_TIMER_MAX_DELTA; | |
783 | ||
9501b6cf TG |
784 | /* |
785 | * Take rounding errors in to account and make sure, that it | |
786 | * expires in the next tick. Otherwise we go into an endless | |
787 | * ping pong due to tick_nohz_stop_sched_tick() retriggering | |
788 | * the timer softirq | |
789 | */ | |
790 | if (delta < 1) | |
791 | delta = 1; | |
792 | now += delta; | |
1cfd6849 TG |
793 | if (time_before(now, expires)) |
794 | return now; | |
1da177e4 LT |
795 | return expires; |
796 | } | |
1cfd6849 TG |
797 | |
798 | /** | |
8dce39c2 | 799 | * get_next_timer_interrupt - return the jiffy of the next pending timer |
05fb6bf0 | 800 | * @now: current time (in jiffies) |
1cfd6849 | 801 | */ |
fd064b9b | 802 | unsigned long get_next_timer_interrupt(unsigned long now) |
1cfd6849 | 803 | { |
a6fa8e5a | 804 | struct tvec_base *base = __get_cpu_var(tvec_bases); |
fd064b9b | 805 | unsigned long expires; |
1cfd6849 TG |
806 | |
807 | spin_lock(&base->lock); | |
808 | expires = __next_timer_interrupt(base); | |
809 | spin_unlock(&base->lock); | |
810 | ||
811 | if (time_before_eq(expires, now)) | |
812 | return now; | |
813 | ||
814 | return cmp_next_hrtimer_event(now, expires); | |
815 | } | |
fd064b9b TG |
816 | |
817 | #ifdef CONFIG_NO_IDLE_HZ | |
818 | unsigned long next_timer_interrupt(void) | |
819 | { | |
820 | return get_next_timer_interrupt(jiffies); | |
821 | } | |
822 | #endif | |
823 | ||
1da177e4 LT |
824 | #endif |
825 | ||
fa13a5a1 PM |
826 | #ifndef CONFIG_VIRT_CPU_ACCOUNTING |
827 | void account_process_tick(struct task_struct *p, int user_tick) | |
828 | { | |
06b8e878 MN |
829 | cputime_t one_jiffy = jiffies_to_cputime(1); |
830 | ||
fa13a5a1 | 831 | if (user_tick) { |
06b8e878 MN |
832 | account_user_time(p, one_jiffy); |
833 | account_user_time_scaled(p, cputime_to_scaled(one_jiffy)); | |
fa13a5a1 | 834 | } else { |
06b8e878 MN |
835 | account_system_time(p, HARDIRQ_OFFSET, one_jiffy); |
836 | account_system_time_scaled(p, cputime_to_scaled(one_jiffy)); | |
fa13a5a1 PM |
837 | } |
838 | } | |
839 | #endif | |
840 | ||
1da177e4 | 841 | /* |
5b4db0c2 | 842 | * Called from the timer interrupt handler to charge one tick to the current |
1da177e4 LT |
843 | * process. user_tick is 1 if the tick is user time, 0 for system. |
844 | */ | |
845 | void update_process_times(int user_tick) | |
846 | { | |
847 | struct task_struct *p = current; | |
848 | int cpu = smp_processor_id(); | |
849 | ||
850 | /* Note: this timer irq context must be accounted for as well. */ | |
fa13a5a1 | 851 | account_process_tick(p, user_tick); |
1da177e4 LT |
852 | run_local_timers(); |
853 | if (rcu_pending(cpu)) | |
854 | rcu_check_callbacks(cpu, user_tick); | |
855 | scheduler_tick(); | |
6819457d | 856 | run_posix_cpu_timers(p); |
1da177e4 LT |
857 | } |
858 | ||
859 | /* | |
860 | * Nr of active tasks - counted in fixed-point numbers | |
861 | */ | |
862 | static unsigned long count_active_tasks(void) | |
863 | { | |
db1b1fef | 864 | return nr_active() * FIXED_1; |
1da177e4 LT |
865 | } |
866 | ||
867 | /* | |
868 | * Hmm.. Changed this, as the GNU make sources (load.c) seems to | |
869 | * imply that avenrun[] is the standard name for this kind of thing. | |
870 | * Nothing else seems to be standardized: the fractional size etc | |
871 | * all seem to differ on different machines. | |
872 | * | |
873 | * Requires xtime_lock to access. | |
874 | */ | |
875 | unsigned long avenrun[3]; | |
876 | ||
877 | EXPORT_SYMBOL(avenrun); | |
878 | ||
879 | /* | |
880 | * calc_load - given tick count, update the avenrun load estimates. | |
881 | * This is called while holding a write_lock on xtime_lock. | |
882 | */ | |
883 | static inline void calc_load(unsigned long ticks) | |
884 | { | |
885 | unsigned long active_tasks; /* fixed-point */ | |
886 | static int count = LOAD_FREQ; | |
887 | ||
cd7175ed ED |
888 | count -= ticks; |
889 | if (unlikely(count < 0)) { | |
890 | active_tasks = count_active_tasks(); | |
891 | do { | |
892 | CALC_LOAD(avenrun[0], EXP_1, active_tasks); | |
893 | CALC_LOAD(avenrun[1], EXP_5, active_tasks); | |
894 | CALC_LOAD(avenrun[2], EXP_15, active_tasks); | |
895 | count += LOAD_FREQ; | |
896 | } while (count < 0); | |
1da177e4 LT |
897 | } |
898 | } | |
899 | ||
1da177e4 LT |
900 | /* |
901 | * This function runs timers and the timer-tq in bottom half context. | |
902 | */ | |
903 | static void run_timer_softirq(struct softirq_action *h) | |
904 | { | |
a6fa8e5a | 905 | struct tvec_base *base = __get_cpu_var(tvec_bases); |
1da177e4 | 906 | |
d3d74453 | 907 | hrtimer_run_pending(); |
82f67cd9 | 908 | |
1da177e4 LT |
909 | if (time_after_eq(jiffies, base->timer_jiffies)) |
910 | __run_timers(base); | |
911 | } | |
912 | ||
913 | /* | |
914 | * Called by the local, per-CPU timer interrupt on SMP. | |
915 | */ | |
916 | void run_local_timers(void) | |
917 | { | |
d3d74453 | 918 | hrtimer_run_queues(); |
1da177e4 | 919 | raise_softirq(TIMER_SOFTIRQ); |
6687a97d | 920 | softlockup_tick(); |
1da177e4 LT |
921 | } |
922 | ||
923 | /* | |
924 | * Called by the timer interrupt. xtime_lock must already be taken | |
925 | * by the timer IRQ! | |
926 | */ | |
3171a030 | 927 | static inline void update_times(unsigned long ticks) |
1da177e4 | 928 | { |
ad596171 | 929 | update_wall_time(); |
1da177e4 LT |
930 | calc_load(ticks); |
931 | } | |
6819457d | 932 | |
1da177e4 LT |
933 | /* |
934 | * The 64-bit jiffies value is not atomic - you MUST NOT read it | |
935 | * without sampling the sequence number in xtime_lock. | |
936 | * jiffies is defined in the linker script... | |
937 | */ | |
938 | ||
3171a030 | 939 | void do_timer(unsigned long ticks) |
1da177e4 | 940 | { |
3171a030 AN |
941 | jiffies_64 += ticks; |
942 | update_times(ticks); | |
1da177e4 LT |
943 | } |
944 | ||
945 | #ifdef __ARCH_WANT_SYS_ALARM | |
946 | ||
947 | /* | |
948 | * For backwards compatibility? This can be done in libc so Alpha | |
949 | * and all newer ports shouldn't need it. | |
950 | */ | |
951 | asmlinkage unsigned long sys_alarm(unsigned int seconds) | |
952 | { | |
c08b8a49 | 953 | return alarm_setitimer(seconds); |
1da177e4 LT |
954 | } |
955 | ||
956 | #endif | |
957 | ||
958 | #ifndef __alpha__ | |
959 | ||
960 | /* | |
961 | * The Alpha uses getxpid, getxuid, and getxgid instead. Maybe this | |
962 | * should be moved into arch/i386 instead? | |
963 | */ | |
964 | ||
965 | /** | |
966 | * sys_getpid - return the thread group id of the current process | |
967 | * | |
968 | * Note, despite the name, this returns the tgid not the pid. The tgid and | |
969 | * the pid are identical unless CLONE_THREAD was specified on clone() in | |
970 | * which case the tgid is the same in all threads of the same group. | |
971 | * | |
972 | * This is SMP safe as current->tgid does not change. | |
973 | */ | |
974 | asmlinkage long sys_getpid(void) | |
975 | { | |
b488893a | 976 | return task_tgid_vnr(current); |
1da177e4 LT |
977 | } |
978 | ||
979 | /* | |
6997a6fa KK |
980 | * Accessing ->real_parent is not SMP-safe, it could |
981 | * change from under us. However, we can use a stale | |
982 | * value of ->real_parent under rcu_read_lock(), see | |
983 | * release_task()->call_rcu(delayed_put_task_struct). | |
1da177e4 LT |
984 | */ |
985 | asmlinkage long sys_getppid(void) | |
986 | { | |
987 | int pid; | |
1da177e4 | 988 | |
6997a6fa | 989 | rcu_read_lock(); |
6c5f3e7b | 990 | pid = task_tgid_vnr(current->real_parent); |
6997a6fa | 991 | rcu_read_unlock(); |
1da177e4 | 992 | |
1da177e4 LT |
993 | return pid; |
994 | } | |
995 | ||
996 | asmlinkage long sys_getuid(void) | |
997 | { | |
998 | /* Only we change this so SMP safe */ | |
999 | return current->uid; | |
1000 | } | |
1001 | ||
1002 | asmlinkage long sys_geteuid(void) | |
1003 | { | |
1004 | /* Only we change this so SMP safe */ | |
1005 | return current->euid; | |
1006 | } | |
1007 | ||
1008 | asmlinkage long sys_getgid(void) | |
1009 | { | |
1010 | /* Only we change this so SMP safe */ | |
1011 | return current->gid; | |
1012 | } | |
1013 | ||
1014 | asmlinkage long sys_getegid(void) | |
1015 | { | |
1016 | /* Only we change this so SMP safe */ | |
1017 | return current->egid; | |
1018 | } | |
1019 | ||
1020 | #endif | |
1021 | ||
1022 | static void process_timeout(unsigned long __data) | |
1023 | { | |
36c8b586 | 1024 | wake_up_process((struct task_struct *)__data); |
1da177e4 LT |
1025 | } |
1026 | ||
1027 | /** | |
1028 | * schedule_timeout - sleep until timeout | |
1029 | * @timeout: timeout value in jiffies | |
1030 | * | |
1031 | * Make the current task sleep until @timeout jiffies have | |
1032 | * elapsed. The routine will return immediately unless | |
1033 | * the current task state has been set (see set_current_state()). | |
1034 | * | |
1035 | * You can set the task state as follows - | |
1036 | * | |
1037 | * %TASK_UNINTERRUPTIBLE - at least @timeout jiffies are guaranteed to | |
1038 | * pass before the routine returns. The routine will return 0 | |
1039 | * | |
1040 | * %TASK_INTERRUPTIBLE - the routine may return early if a signal is | |
1041 | * delivered to the current task. In this case the remaining time | |
1042 | * in jiffies will be returned, or 0 if the timer expired in time | |
1043 | * | |
1044 | * The current task state is guaranteed to be TASK_RUNNING when this | |
1045 | * routine returns. | |
1046 | * | |
1047 | * Specifying a @timeout value of %MAX_SCHEDULE_TIMEOUT will schedule | |
1048 | * the CPU away without a bound on the timeout. In this case the return | |
1049 | * value will be %MAX_SCHEDULE_TIMEOUT. | |
1050 | * | |
1051 | * In all cases the return value is guaranteed to be non-negative. | |
1052 | */ | |
7ad5b3a5 | 1053 | signed long __sched schedule_timeout(signed long timeout) |
1da177e4 LT |
1054 | { |
1055 | struct timer_list timer; | |
1056 | unsigned long expire; | |
1057 | ||
1058 | switch (timeout) | |
1059 | { | |
1060 | case MAX_SCHEDULE_TIMEOUT: | |
1061 | /* | |
1062 | * These two special cases are useful to be comfortable | |
1063 | * in the caller. Nothing more. We could take | |
1064 | * MAX_SCHEDULE_TIMEOUT from one of the negative value | |
1065 | * but I' d like to return a valid offset (>=0) to allow | |
1066 | * the caller to do everything it want with the retval. | |
1067 | */ | |
1068 | schedule(); | |
1069 | goto out; | |
1070 | default: | |
1071 | /* | |
1072 | * Another bit of PARANOID. Note that the retval will be | |
1073 | * 0 since no piece of kernel is supposed to do a check | |
1074 | * for a negative retval of schedule_timeout() (since it | |
1075 | * should never happens anyway). You just have the printk() | |
1076 | * that will tell you if something is gone wrong and where. | |
1077 | */ | |
5b149bcc | 1078 | if (timeout < 0) { |
1da177e4 | 1079 | printk(KERN_ERR "schedule_timeout: wrong timeout " |
5b149bcc AM |
1080 | "value %lx\n", timeout); |
1081 | dump_stack(); | |
1da177e4 LT |
1082 | current->state = TASK_RUNNING; |
1083 | goto out; | |
1084 | } | |
1085 | } | |
1086 | ||
1087 | expire = timeout + jiffies; | |
1088 | ||
a8db2db1 ON |
1089 | setup_timer(&timer, process_timeout, (unsigned long)current); |
1090 | __mod_timer(&timer, expire); | |
1da177e4 LT |
1091 | schedule(); |
1092 | del_singleshot_timer_sync(&timer); | |
1093 | ||
1094 | timeout = expire - jiffies; | |
1095 | ||
1096 | out: | |
1097 | return timeout < 0 ? 0 : timeout; | |
1098 | } | |
1da177e4 LT |
1099 | EXPORT_SYMBOL(schedule_timeout); |
1100 | ||
8a1c1757 AM |
1101 | /* |
1102 | * We can use __set_current_state() here because schedule_timeout() calls | |
1103 | * schedule() unconditionally. | |
1104 | */ | |
64ed93a2 NA |
1105 | signed long __sched schedule_timeout_interruptible(signed long timeout) |
1106 | { | |
a5a0d52c AM |
1107 | __set_current_state(TASK_INTERRUPTIBLE); |
1108 | return schedule_timeout(timeout); | |
64ed93a2 NA |
1109 | } |
1110 | EXPORT_SYMBOL(schedule_timeout_interruptible); | |
1111 | ||
294d5cc2 MW |
1112 | signed long __sched schedule_timeout_killable(signed long timeout) |
1113 | { | |
1114 | __set_current_state(TASK_KILLABLE); | |
1115 | return schedule_timeout(timeout); | |
1116 | } | |
1117 | EXPORT_SYMBOL(schedule_timeout_killable); | |
1118 | ||
64ed93a2 NA |
1119 | signed long __sched schedule_timeout_uninterruptible(signed long timeout) |
1120 | { | |
a5a0d52c AM |
1121 | __set_current_state(TASK_UNINTERRUPTIBLE); |
1122 | return schedule_timeout(timeout); | |
64ed93a2 NA |
1123 | } |
1124 | EXPORT_SYMBOL(schedule_timeout_uninterruptible); | |
1125 | ||
1da177e4 LT |
1126 | /* Thread ID - the internal kernel "pid" */ |
1127 | asmlinkage long sys_gettid(void) | |
1128 | { | |
b488893a | 1129 | return task_pid_vnr(current); |
1da177e4 LT |
1130 | } |
1131 | ||
2aae4a10 | 1132 | /** |
d4d23add | 1133 | * do_sysinfo - fill in sysinfo struct |
2aae4a10 | 1134 | * @info: pointer to buffer to fill |
6819457d | 1135 | */ |
d4d23add | 1136 | int do_sysinfo(struct sysinfo *info) |
1da177e4 | 1137 | { |
1da177e4 LT |
1138 | unsigned long mem_total, sav_total; |
1139 | unsigned int mem_unit, bitcount; | |
1140 | unsigned long seq; | |
1141 | ||
d4d23add | 1142 | memset(info, 0, sizeof(struct sysinfo)); |
1da177e4 LT |
1143 | |
1144 | do { | |
1145 | struct timespec tp; | |
1146 | seq = read_seqbegin(&xtime_lock); | |
1147 | ||
1148 | /* | |
1149 | * This is annoying. The below is the same thing | |
1150 | * posix_get_clock_monotonic() does, but it wants to | |
1151 | * take the lock which we want to cover the loads stuff | |
1152 | * too. | |
1153 | */ | |
1154 | ||
1155 | getnstimeofday(&tp); | |
1156 | tp.tv_sec += wall_to_monotonic.tv_sec; | |
1157 | tp.tv_nsec += wall_to_monotonic.tv_nsec; | |
d6214141 | 1158 | monotonic_to_bootbased(&tp); |
1da177e4 LT |
1159 | if (tp.tv_nsec - NSEC_PER_SEC >= 0) { |
1160 | tp.tv_nsec = tp.tv_nsec - NSEC_PER_SEC; | |
1161 | tp.tv_sec++; | |
1162 | } | |
d4d23add | 1163 | info->uptime = tp.tv_sec + (tp.tv_nsec ? 1 : 0); |
1da177e4 | 1164 | |
d4d23add KM |
1165 | info->loads[0] = avenrun[0] << (SI_LOAD_SHIFT - FSHIFT); |
1166 | info->loads[1] = avenrun[1] << (SI_LOAD_SHIFT - FSHIFT); | |
1167 | info->loads[2] = avenrun[2] << (SI_LOAD_SHIFT - FSHIFT); | |
1da177e4 | 1168 | |
d4d23add | 1169 | info->procs = nr_threads; |
1da177e4 LT |
1170 | } while (read_seqretry(&xtime_lock, seq)); |
1171 | ||
d4d23add KM |
1172 | si_meminfo(info); |
1173 | si_swapinfo(info); | |
1da177e4 LT |
1174 | |
1175 | /* | |
1176 | * If the sum of all the available memory (i.e. ram + swap) | |
1177 | * is less than can be stored in a 32 bit unsigned long then | |
1178 | * we can be binary compatible with 2.2.x kernels. If not, | |
1179 | * well, in that case 2.2.x was broken anyways... | |
1180 | * | |
1181 | * -Erik Andersen <andersee@debian.org> | |
1182 | */ | |
1183 | ||
d4d23add KM |
1184 | mem_total = info->totalram + info->totalswap; |
1185 | if (mem_total < info->totalram || mem_total < info->totalswap) | |
1da177e4 LT |
1186 | goto out; |
1187 | bitcount = 0; | |
d4d23add | 1188 | mem_unit = info->mem_unit; |
1da177e4 LT |
1189 | while (mem_unit > 1) { |
1190 | bitcount++; | |
1191 | mem_unit >>= 1; | |
1192 | sav_total = mem_total; | |
1193 | mem_total <<= 1; | |
1194 | if (mem_total < sav_total) | |
1195 | goto out; | |
1196 | } | |
1197 | ||
1198 | /* | |
1199 | * If mem_total did not overflow, multiply all memory values by | |
d4d23add | 1200 | * info->mem_unit and set it to 1. This leaves things compatible |
1da177e4 LT |
1201 | * with 2.2.x, and also retains compatibility with earlier 2.4.x |
1202 | * kernels... | |
1203 | */ | |
1204 | ||
d4d23add KM |
1205 | info->mem_unit = 1; |
1206 | info->totalram <<= bitcount; | |
1207 | info->freeram <<= bitcount; | |
1208 | info->sharedram <<= bitcount; | |
1209 | info->bufferram <<= bitcount; | |
1210 | info->totalswap <<= bitcount; | |
1211 | info->freeswap <<= bitcount; | |
1212 | info->totalhigh <<= bitcount; | |
1213 | info->freehigh <<= bitcount; | |
1214 | ||
1215 | out: | |
1216 | return 0; | |
1217 | } | |
1218 | ||
1219 | asmlinkage long sys_sysinfo(struct sysinfo __user *info) | |
1220 | { | |
1221 | struct sysinfo val; | |
1222 | ||
1223 | do_sysinfo(&val); | |
1da177e4 | 1224 | |
1da177e4 LT |
1225 | if (copy_to_user(info, &val, sizeof(struct sysinfo))) |
1226 | return -EFAULT; | |
1227 | ||
1228 | return 0; | |
1229 | } | |
1230 | ||
b4be6258 | 1231 | static int __cpuinit init_timers_cpu(int cpu) |
1da177e4 LT |
1232 | { |
1233 | int j; | |
a6fa8e5a | 1234 | struct tvec_base *base; |
b4be6258 | 1235 | static char __cpuinitdata tvec_base_done[NR_CPUS]; |
55c888d6 | 1236 | |
ba6edfcd | 1237 | if (!tvec_base_done[cpu]) { |
a4a6198b JB |
1238 | static char boot_done; |
1239 | ||
a4a6198b | 1240 | if (boot_done) { |
ba6edfcd AM |
1241 | /* |
1242 | * The APs use this path later in boot | |
1243 | */ | |
94f6030c CL |
1244 | base = kmalloc_node(sizeof(*base), |
1245 | GFP_KERNEL | __GFP_ZERO, | |
a4a6198b JB |
1246 | cpu_to_node(cpu)); |
1247 | if (!base) | |
1248 | return -ENOMEM; | |
6e453a67 VP |
1249 | |
1250 | /* Make sure that tvec_base is 2 byte aligned */ | |
1251 | if (tbase_get_deferrable(base)) { | |
1252 | WARN_ON(1); | |
1253 | kfree(base); | |
1254 | return -ENOMEM; | |
1255 | } | |
ba6edfcd | 1256 | per_cpu(tvec_bases, cpu) = base; |
a4a6198b | 1257 | } else { |
ba6edfcd AM |
1258 | /* |
1259 | * This is for the boot CPU - we use compile-time | |
1260 | * static initialisation because per-cpu memory isn't | |
1261 | * ready yet and because the memory allocators are not | |
1262 | * initialised either. | |
1263 | */ | |
a4a6198b | 1264 | boot_done = 1; |
ba6edfcd | 1265 | base = &boot_tvec_bases; |
a4a6198b | 1266 | } |
ba6edfcd AM |
1267 | tvec_base_done[cpu] = 1; |
1268 | } else { | |
1269 | base = per_cpu(tvec_bases, cpu); | |
a4a6198b | 1270 | } |
ba6edfcd | 1271 | |
3691c519 | 1272 | spin_lock_init(&base->lock); |
d730e882 | 1273 | |
1da177e4 LT |
1274 | for (j = 0; j < TVN_SIZE; j++) { |
1275 | INIT_LIST_HEAD(base->tv5.vec + j); | |
1276 | INIT_LIST_HEAD(base->tv4.vec + j); | |
1277 | INIT_LIST_HEAD(base->tv3.vec + j); | |
1278 | INIT_LIST_HEAD(base->tv2.vec + j); | |
1279 | } | |
1280 | for (j = 0; j < TVR_SIZE; j++) | |
1281 | INIT_LIST_HEAD(base->tv1.vec + j); | |
1282 | ||
1283 | base->timer_jiffies = jiffies; | |
a4a6198b | 1284 | return 0; |
1da177e4 LT |
1285 | } |
1286 | ||
1287 | #ifdef CONFIG_HOTPLUG_CPU | |
a6fa8e5a | 1288 | static void migrate_timer_list(struct tvec_base *new_base, struct list_head *head) |
1da177e4 LT |
1289 | { |
1290 | struct timer_list *timer; | |
1291 | ||
1292 | while (!list_empty(head)) { | |
b5e61818 | 1293 | timer = list_first_entry(head, struct timer_list, entry); |
55c888d6 | 1294 | detach_timer(timer, 0); |
6e453a67 | 1295 | timer_set_base(timer, new_base); |
1da177e4 | 1296 | internal_add_timer(new_base, timer); |
1da177e4 | 1297 | } |
1da177e4 LT |
1298 | } |
1299 | ||
48ccf3da | 1300 | static void __cpuinit migrate_timers(int cpu) |
1da177e4 | 1301 | { |
a6fa8e5a PM |
1302 | struct tvec_base *old_base; |
1303 | struct tvec_base *new_base; | |
1da177e4 LT |
1304 | int i; |
1305 | ||
1306 | BUG_ON(cpu_online(cpu)); | |
a4a6198b JB |
1307 | old_base = per_cpu(tvec_bases, cpu); |
1308 | new_base = get_cpu_var(tvec_bases); | |
1da177e4 LT |
1309 | |
1310 | local_irq_disable(); | |
0d180406 ON |
1311 | spin_lock(&new_base->lock); |
1312 | spin_lock_nested(&old_base->lock, SINGLE_DEPTH_NESTING); | |
3691c519 ON |
1313 | |
1314 | BUG_ON(old_base->running_timer); | |
1da177e4 | 1315 | |
1da177e4 | 1316 | for (i = 0; i < TVR_SIZE; i++) |
55c888d6 ON |
1317 | migrate_timer_list(new_base, old_base->tv1.vec + i); |
1318 | for (i = 0; i < TVN_SIZE; i++) { | |
1319 | migrate_timer_list(new_base, old_base->tv2.vec + i); | |
1320 | migrate_timer_list(new_base, old_base->tv3.vec + i); | |
1321 | migrate_timer_list(new_base, old_base->tv4.vec + i); | |
1322 | migrate_timer_list(new_base, old_base->tv5.vec + i); | |
1323 | } | |
1324 | ||
0d180406 ON |
1325 | spin_unlock(&old_base->lock); |
1326 | spin_unlock(&new_base->lock); | |
1da177e4 LT |
1327 | local_irq_enable(); |
1328 | put_cpu_var(tvec_bases); | |
1da177e4 LT |
1329 | } |
1330 | #endif /* CONFIG_HOTPLUG_CPU */ | |
1331 | ||
8c78f307 | 1332 | static int __cpuinit timer_cpu_notify(struct notifier_block *self, |
1da177e4 LT |
1333 | unsigned long action, void *hcpu) |
1334 | { | |
1335 | long cpu = (long)hcpu; | |
1336 | switch(action) { | |
1337 | case CPU_UP_PREPARE: | |
8bb78442 | 1338 | case CPU_UP_PREPARE_FROZEN: |
a4a6198b JB |
1339 | if (init_timers_cpu(cpu) < 0) |
1340 | return NOTIFY_BAD; | |
1da177e4 LT |
1341 | break; |
1342 | #ifdef CONFIG_HOTPLUG_CPU | |
1343 | case CPU_DEAD: | |
8bb78442 | 1344 | case CPU_DEAD_FROZEN: |
1da177e4 LT |
1345 | migrate_timers(cpu); |
1346 | break; | |
1347 | #endif | |
1348 | default: | |
1349 | break; | |
1350 | } | |
1351 | return NOTIFY_OK; | |
1352 | } | |
1353 | ||
8c78f307 | 1354 | static struct notifier_block __cpuinitdata timers_nb = { |
1da177e4 LT |
1355 | .notifier_call = timer_cpu_notify, |
1356 | }; | |
1357 | ||
1358 | ||
1359 | void __init init_timers(void) | |
1360 | { | |
07dccf33 | 1361 | int err = timer_cpu_notify(&timers_nb, (unsigned long)CPU_UP_PREPARE, |
1da177e4 | 1362 | (void *)(long)smp_processor_id()); |
07dccf33 | 1363 | |
82f67cd9 IM |
1364 | init_timer_stats(); |
1365 | ||
07dccf33 | 1366 | BUG_ON(err == NOTIFY_BAD); |
1da177e4 LT |
1367 | register_cpu_notifier(&timers_nb); |
1368 | open_softirq(TIMER_SOFTIRQ, run_timer_softirq, NULL); | |
1369 | } | |
1370 | ||
1da177e4 LT |
1371 | /** |
1372 | * msleep - sleep safely even with waitqueue interruptions | |
1373 | * @msecs: Time in milliseconds to sleep for | |
1374 | */ | |
1375 | void msleep(unsigned int msecs) | |
1376 | { | |
1377 | unsigned long timeout = msecs_to_jiffies(msecs) + 1; | |
1378 | ||
75bcc8c5 NA |
1379 | while (timeout) |
1380 | timeout = schedule_timeout_uninterruptible(timeout); | |
1da177e4 LT |
1381 | } |
1382 | ||
1383 | EXPORT_SYMBOL(msleep); | |
1384 | ||
1385 | /** | |
96ec3efd | 1386 | * msleep_interruptible - sleep waiting for signals |
1da177e4 LT |
1387 | * @msecs: Time in milliseconds to sleep for |
1388 | */ | |
1389 | unsigned long msleep_interruptible(unsigned int msecs) | |
1390 | { | |
1391 | unsigned long timeout = msecs_to_jiffies(msecs) + 1; | |
1392 | ||
75bcc8c5 NA |
1393 | while (timeout && !signal_pending(current)) |
1394 | timeout = schedule_timeout_interruptible(timeout); | |
1da177e4 LT |
1395 | return jiffies_to_msecs(timeout); |
1396 | } | |
1397 | ||
1398 | EXPORT_SYMBOL(msleep_interruptible); |