timers: allow deferrable timers for intervals tv2-tv5 to be deferred
[GitHub/MotorolaMobilityLLC/kernel-slsi.git] / kernel / timer.c
CommitLineData
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
47u64 jiffies_64 __cacheline_aligned_in_smp = INITIAL_JIFFIES;
48
49EXPORT_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 61struct tvec {
1da177e4 62 struct list_head vec[TVN_SIZE];
a6fa8e5a 63};
1da177e4 64
a6fa8e5a 65struct tvec_root {
1da177e4 66 struct list_head vec[TVR_SIZE];
a6fa8e5a 67};
1da177e4 68
a6fa8e5a 69struct 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 80struct tvec_base boot_tvec_bases;
3691c519 81EXPORT_SYMBOL(boot_tvec_bases);
a6fa8e5a 82static 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 92static 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 97static 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
102static 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
108static inline void
a6fa8e5a 109timer_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
9c133c46
AS
115static unsigned long round_jiffies_common(unsigned long j, int cpu,
116 bool force_up)
4c36a5de
AV
117{
118 int rem;
119 unsigned long original = j;
120
121 /*
122 * We don't want all cpus firing their timers at once hitting the
123 * same lock or cachelines, so we skew each extra cpu with an extra
124 * 3 jiffies. This 3 jiffies came originally from the mm/ code which
125 * already did this.
126 * The skew is done by adding 3*cpunr, then round, then subtract this
127 * extra offset again.
128 */
129 j += cpu * 3;
130
131 rem = j % HZ;
132
133 /*
134 * If the target jiffie is just after a whole second (which can happen
135 * due to delays of the timer irq, long irq off times etc etc) then
136 * we should round down to the whole second, not up. Use 1/4th second
137 * as cutoff for this rounding as an extreme upper bound for this.
9c133c46 138 * But never round down if @force_up is set.
4c36a5de 139 */
9c133c46 140 if (rem < HZ/4 && !force_up) /* round down */
4c36a5de
AV
141 j = j - rem;
142 else /* round up */
143 j = j - rem + HZ;
144
145 /* now that we have rounded, subtract the extra skew again */
146 j -= cpu * 3;
147
148 if (j <= jiffies) /* rounding ate our timeout entirely; */
149 return original;
150 return j;
151}
9c133c46
AS
152
153/**
154 * __round_jiffies - function to round jiffies to a full second
155 * @j: the time in (absolute) jiffies that should be rounded
156 * @cpu: the processor number on which the timeout will happen
157 *
158 * __round_jiffies() rounds an absolute time in the future (in jiffies)
159 * up or down to (approximately) full seconds. This is useful for timers
160 * for which the exact time they fire does not matter too much, as long as
161 * they fire approximately every X seconds.
162 *
163 * By rounding these timers to whole seconds, all such timers will fire
164 * at the same time, rather than at various times spread out. The goal
165 * of this is to have the CPU wake up less, which saves power.
166 *
167 * The exact rounding is skewed for each processor to avoid all
168 * processors firing at the exact same time, which could lead
169 * to lock contention or spurious cache line bouncing.
170 *
171 * The return value is the rounded version of the @j parameter.
172 */
173unsigned long __round_jiffies(unsigned long j, int cpu)
174{
175 return round_jiffies_common(j, cpu, false);
176}
4c36a5de
AV
177EXPORT_SYMBOL_GPL(__round_jiffies);
178
179/**
180 * __round_jiffies_relative - function to round jiffies to a full second
181 * @j: the time in (relative) jiffies that should be rounded
182 * @cpu: the processor number on which the timeout will happen
183 *
72fd4a35 184 * __round_jiffies_relative() rounds a time delta in the future (in jiffies)
4c36a5de
AV
185 * up or down to (approximately) full seconds. This is useful for timers
186 * for which the exact time they fire does not matter too much, as long as
187 * they fire approximately every X seconds.
188 *
189 * By rounding these timers to whole seconds, all such timers will fire
190 * at the same time, rather than at various times spread out. The goal
191 * of this is to have the CPU wake up less, which saves power.
192 *
193 * The exact rounding is skewed for each processor to avoid all
194 * processors firing at the exact same time, which could lead
195 * to lock contention or spurious cache line bouncing.
196 *
72fd4a35 197 * The return value is the rounded version of the @j parameter.
4c36a5de
AV
198 */
199unsigned long __round_jiffies_relative(unsigned long j, int cpu)
200{
9c133c46
AS
201 unsigned long j0 = jiffies;
202
203 /* Use j0 because jiffies might change while we run */
204 return round_jiffies_common(j + j0, cpu, false) - j0;
4c36a5de
AV
205}
206EXPORT_SYMBOL_GPL(__round_jiffies_relative);
207
208/**
209 * round_jiffies - function to round jiffies to a full second
210 * @j: the time in (absolute) jiffies that should be rounded
211 *
72fd4a35 212 * round_jiffies() rounds an absolute time in the future (in jiffies)
4c36a5de
AV
213 * up or down to (approximately) full seconds. This is useful for timers
214 * for which the exact time they fire does not matter too much, as long as
215 * they fire approximately every X seconds.
216 *
217 * By rounding these timers to whole seconds, all such timers will fire
218 * at the same time, rather than at various times spread out. The goal
219 * of this is to have the CPU wake up less, which saves power.
220 *
72fd4a35 221 * The return value is the rounded version of the @j parameter.
4c36a5de
AV
222 */
223unsigned long round_jiffies(unsigned long j)
224{
9c133c46 225 return round_jiffies_common(j, raw_smp_processor_id(), false);
4c36a5de
AV
226}
227EXPORT_SYMBOL_GPL(round_jiffies);
228
229/**
230 * round_jiffies_relative - function to round jiffies to a full second
231 * @j: the time in (relative) jiffies that should be rounded
232 *
72fd4a35 233 * round_jiffies_relative() rounds a time delta in the future (in jiffies)
4c36a5de
AV
234 * up or down to (approximately) full seconds. This is useful for timers
235 * for which the exact time they fire does not matter too much, as long as
236 * they fire approximately every X seconds.
237 *
238 * By rounding these timers to whole seconds, all such timers will fire
239 * at the same time, rather than at various times spread out. The goal
240 * of this is to have the CPU wake up less, which saves power.
241 *
72fd4a35 242 * The return value is the rounded version of the @j parameter.
4c36a5de
AV
243 */
244unsigned long round_jiffies_relative(unsigned long j)
245{
246 return __round_jiffies_relative(j, raw_smp_processor_id());
247}
248EXPORT_SYMBOL_GPL(round_jiffies_relative);
249
9c133c46
AS
250/**
251 * __round_jiffies_up - function to round jiffies up to a full second
252 * @j: the time in (absolute) jiffies that should be rounded
253 * @cpu: the processor number on which the timeout will happen
254 *
255 * This is the same as __round_jiffies() except that it will never
256 * round down. This is useful for timeouts for which the exact time
257 * of firing does not matter too much, as long as they don't fire too
258 * early.
259 */
260unsigned long __round_jiffies_up(unsigned long j, int cpu)
261{
262 return round_jiffies_common(j, cpu, true);
263}
264EXPORT_SYMBOL_GPL(__round_jiffies_up);
265
266/**
267 * __round_jiffies_up_relative - function to round jiffies up to a full second
268 * @j: the time in (relative) jiffies that should be rounded
269 * @cpu: the processor number on which the timeout will happen
270 *
271 * This is the same as __round_jiffies_relative() except that it will never
272 * round down. This is useful for timeouts for which the exact time
273 * of firing does not matter too much, as long as they don't fire too
274 * early.
275 */
276unsigned long __round_jiffies_up_relative(unsigned long j, int cpu)
277{
278 unsigned long j0 = jiffies;
279
280 /* Use j0 because jiffies might change while we run */
281 return round_jiffies_common(j + j0, cpu, true) - j0;
282}
283EXPORT_SYMBOL_GPL(__round_jiffies_up_relative);
284
285/**
286 * round_jiffies_up - function to round jiffies up to a full second
287 * @j: the time in (absolute) jiffies that should be rounded
288 *
289 * This is the same as round_jiffies() except that it will never
290 * round down. This is useful for timeouts for which the exact time
291 * of firing does not matter too much, as long as they don't fire too
292 * early.
293 */
294unsigned long round_jiffies_up(unsigned long j)
295{
296 return round_jiffies_common(j, raw_smp_processor_id(), true);
297}
298EXPORT_SYMBOL_GPL(round_jiffies_up);
299
300/**
301 * round_jiffies_up_relative - function to round jiffies up to a full second
302 * @j: the time in (relative) jiffies that should be rounded
303 *
304 * This is the same as round_jiffies_relative() except that it will never
305 * round down. This is useful for timeouts for which the exact time
306 * of firing does not matter too much, as long as they don't fire too
307 * early.
308 */
309unsigned long round_jiffies_up_relative(unsigned long j)
310{
311 return __round_jiffies_up_relative(j, raw_smp_processor_id());
312}
313EXPORT_SYMBOL_GPL(round_jiffies_up_relative);
314
4c36a5de 315
a6fa8e5a 316static inline void set_running_timer(struct tvec_base *base,
1da177e4
LT
317 struct timer_list *timer)
318{
319#ifdef CONFIG_SMP
3691c519 320 base->running_timer = timer;
1da177e4
LT
321#endif
322}
323
a6fa8e5a 324static void internal_add_timer(struct tvec_base *base, struct timer_list *timer)
1da177e4
LT
325{
326 unsigned long expires = timer->expires;
327 unsigned long idx = expires - base->timer_jiffies;
328 struct list_head *vec;
329
330 if (idx < TVR_SIZE) {
331 int i = expires & TVR_MASK;
332 vec = base->tv1.vec + i;
333 } else if (idx < 1 << (TVR_BITS + TVN_BITS)) {
334 int i = (expires >> TVR_BITS) & TVN_MASK;
335 vec = base->tv2.vec + i;
336 } else if (idx < 1 << (TVR_BITS + 2 * TVN_BITS)) {
337 int i = (expires >> (TVR_BITS + TVN_BITS)) & TVN_MASK;
338 vec = base->tv3.vec + i;
339 } else if (idx < 1 << (TVR_BITS + 3 * TVN_BITS)) {
340 int i = (expires >> (TVR_BITS + 2 * TVN_BITS)) & TVN_MASK;
341 vec = base->tv4.vec + i;
342 } else if ((signed long) idx < 0) {
343 /*
344 * Can happen if you add a timer with expires == jiffies,
345 * or you set a timer to go off in the past
346 */
347 vec = base->tv1.vec + (base->timer_jiffies & TVR_MASK);
348 } else {
349 int i;
350 /* If the timeout is larger than 0xffffffff on 64-bit
351 * architectures then we use the maximum timeout:
352 */
353 if (idx > 0xffffffffUL) {
354 idx = 0xffffffffUL;
355 expires = idx + base->timer_jiffies;
356 }
357 i = (expires >> (TVR_BITS + 3 * TVN_BITS)) & TVN_MASK;
358 vec = base->tv5.vec + i;
359 }
360 /*
361 * Timers are FIFO:
362 */
363 list_add_tail(&timer->entry, vec);
364}
365
82f67cd9
IM
366#ifdef CONFIG_TIMER_STATS
367void __timer_stats_timer_set_start_info(struct timer_list *timer, void *addr)
368{
369 if (timer->start_site)
370 return;
371
372 timer->start_site = addr;
373 memcpy(timer->start_comm, current->comm, TASK_COMM_LEN);
374 timer->start_pid = current->pid;
375}
c5c061b8
VP
376
377static void timer_stats_account_timer(struct timer_list *timer)
378{
379 unsigned int flag = 0;
380
381 if (unlikely(tbase_get_deferrable(timer->base)))
382 flag |= TIMER_STATS_FLAG_DEFERRABLE;
383
384 timer_stats_update_stats(timer, timer->start_pid, timer->start_site,
385 timer->function, timer->start_comm, flag);
386}
387
388#else
389static void timer_stats_account_timer(struct timer_list *timer) {}
82f67cd9
IM
390#endif
391
c6f3a97f
TG
392#ifdef CONFIG_DEBUG_OBJECTS_TIMERS
393
394static struct debug_obj_descr timer_debug_descr;
395
396/*
397 * fixup_init is called when:
398 * - an active object is initialized
55c888d6 399 */
c6f3a97f
TG
400static int timer_fixup_init(void *addr, enum debug_obj_state state)
401{
402 struct timer_list *timer = addr;
403
404 switch (state) {
405 case ODEBUG_STATE_ACTIVE:
406 del_timer_sync(timer);
407 debug_object_init(timer, &timer_debug_descr);
408 return 1;
409 default:
410 return 0;
411 }
412}
413
414/*
415 * fixup_activate is called when:
416 * - an active object is activated
417 * - an unknown object is activated (might be a statically initialized object)
418 */
419static int timer_fixup_activate(void *addr, enum debug_obj_state state)
420{
421 struct timer_list *timer = addr;
422
423 switch (state) {
424
425 case ODEBUG_STATE_NOTAVAILABLE:
426 /*
427 * This is not really a fixup. The timer was
428 * statically initialized. We just make sure that it
429 * is tracked in the object tracker.
430 */
431 if (timer->entry.next == NULL &&
432 timer->entry.prev == TIMER_ENTRY_STATIC) {
433 debug_object_init(timer, &timer_debug_descr);
434 debug_object_activate(timer, &timer_debug_descr);
435 return 0;
436 } else {
437 WARN_ON_ONCE(1);
438 }
439 return 0;
440
441 case ODEBUG_STATE_ACTIVE:
442 WARN_ON(1);
443
444 default:
445 return 0;
446 }
447}
448
449/*
450 * fixup_free is called when:
451 * - an active object is freed
452 */
453static int timer_fixup_free(void *addr, enum debug_obj_state state)
454{
455 struct timer_list *timer = addr;
456
457 switch (state) {
458 case ODEBUG_STATE_ACTIVE:
459 del_timer_sync(timer);
460 debug_object_free(timer, &timer_debug_descr);
461 return 1;
462 default:
463 return 0;
464 }
465}
466
467static struct debug_obj_descr timer_debug_descr = {
468 .name = "timer_list",
469 .fixup_init = timer_fixup_init,
470 .fixup_activate = timer_fixup_activate,
471 .fixup_free = timer_fixup_free,
472};
473
474static inline void debug_timer_init(struct timer_list *timer)
475{
476 debug_object_init(timer, &timer_debug_descr);
477}
478
479static inline void debug_timer_activate(struct timer_list *timer)
480{
481 debug_object_activate(timer, &timer_debug_descr);
482}
483
484static inline void debug_timer_deactivate(struct timer_list *timer)
485{
486 debug_object_deactivate(timer, &timer_debug_descr);
487}
488
489static inline void debug_timer_free(struct timer_list *timer)
490{
491 debug_object_free(timer, &timer_debug_descr);
492}
493
6f2b9b9a
JB
494static void __init_timer(struct timer_list *timer,
495 const char *name,
496 struct lock_class_key *key);
c6f3a97f 497
6f2b9b9a
JB
498void init_timer_on_stack_key(struct timer_list *timer,
499 const char *name,
500 struct lock_class_key *key)
c6f3a97f
TG
501{
502 debug_object_init_on_stack(timer, &timer_debug_descr);
6f2b9b9a 503 __init_timer(timer, name, key);
c6f3a97f 504}
6f2b9b9a 505EXPORT_SYMBOL_GPL(init_timer_on_stack_key);
c6f3a97f
TG
506
507void destroy_timer_on_stack(struct timer_list *timer)
508{
509 debug_object_free(timer, &timer_debug_descr);
510}
511EXPORT_SYMBOL_GPL(destroy_timer_on_stack);
512
513#else
514static inline void debug_timer_init(struct timer_list *timer) { }
515static inline void debug_timer_activate(struct timer_list *timer) { }
516static inline void debug_timer_deactivate(struct timer_list *timer) { }
517#endif
518
6f2b9b9a
JB
519static void __init_timer(struct timer_list *timer,
520 const char *name,
521 struct lock_class_key *key)
55c888d6
ON
522{
523 timer->entry.next = NULL;
bfe5d834 524 timer->base = __raw_get_cpu_var(tvec_bases);
82f67cd9
IM
525#ifdef CONFIG_TIMER_STATS
526 timer->start_site = NULL;
527 timer->start_pid = -1;
528 memset(timer->start_comm, 0, TASK_COMM_LEN);
529#endif
6f2b9b9a 530 lockdep_init_map(&timer->lockdep_map, name, key, 0);
55c888d6 531}
c6f3a97f
TG
532
533/**
633fe795 534 * init_timer_key - initialize a timer
c6f3a97f 535 * @timer: the timer to be initialized
633fe795
RD
536 * @name: name of the timer
537 * @key: lockdep class key of the fake lock used for tracking timer
538 * sync lock dependencies
c6f3a97f 539 *
633fe795 540 * init_timer_key() must be done to a timer prior calling *any* of the
c6f3a97f
TG
541 * other timer functions.
542 */
6f2b9b9a
JB
543void init_timer_key(struct timer_list *timer,
544 const char *name,
545 struct lock_class_key *key)
c6f3a97f
TG
546{
547 debug_timer_init(timer);
6f2b9b9a 548 __init_timer(timer, name, key);
c6f3a97f 549}
6f2b9b9a 550EXPORT_SYMBOL(init_timer_key);
55c888d6 551
6f2b9b9a
JB
552void init_timer_deferrable_key(struct timer_list *timer,
553 const char *name,
554 struct lock_class_key *key)
6e453a67 555{
6f2b9b9a 556 init_timer_key(timer, name, key);
6e453a67
VP
557 timer_set_deferrable(timer);
558}
6f2b9b9a 559EXPORT_SYMBOL(init_timer_deferrable_key);
6e453a67 560
55c888d6 561static inline void detach_timer(struct timer_list *timer,
82f67cd9 562 int clear_pending)
55c888d6
ON
563{
564 struct list_head *entry = &timer->entry;
565
c6f3a97f
TG
566 debug_timer_deactivate(timer);
567
55c888d6
ON
568 __list_del(entry->prev, entry->next);
569 if (clear_pending)
570 entry->next = NULL;
571 entry->prev = LIST_POISON2;
572}
573
574/*
3691c519 575 * We are using hashed locking: holding per_cpu(tvec_bases).lock
55c888d6
ON
576 * means that all timers which are tied to this base via timer->base are
577 * locked, and the base itself is locked too.
578 *
579 * So __run_timers/migrate_timers can safely modify all timers which could
580 * be found on ->tvX lists.
581 *
582 * When the timer's base is locked, and the timer removed from list, it is
583 * possible to set timer->base = NULL and drop the lock: the timer remains
584 * locked.
585 */
a6fa8e5a 586static struct tvec_base *lock_timer_base(struct timer_list *timer,
55c888d6 587 unsigned long *flags)
89e7e374 588 __acquires(timer->base->lock)
55c888d6 589{
a6fa8e5a 590 struct tvec_base *base;
55c888d6
ON
591
592 for (;;) {
a6fa8e5a 593 struct tvec_base *prelock_base = timer->base;
6e453a67 594 base = tbase_get_base(prelock_base);
55c888d6
ON
595 if (likely(base != NULL)) {
596 spin_lock_irqsave(&base->lock, *flags);
6e453a67 597 if (likely(prelock_base == timer->base))
55c888d6
ON
598 return base;
599 /* The timer has migrated to another CPU */
600 spin_unlock_irqrestore(&base->lock, *flags);
601 }
602 cpu_relax();
603 }
604}
605
74019224
IM
606static inline int
607__mod_timer(struct timer_list *timer, unsigned long expires, bool pending_only)
1da177e4 608{
a6fa8e5a 609 struct tvec_base *base, *new_base;
1da177e4 610 unsigned long flags;
74019224
IM
611 int ret;
612
613 ret = 0;
1da177e4 614
82f67cd9 615 timer_stats_timer_set_start_info(timer);
1da177e4 616 BUG_ON(!timer->function);
1da177e4 617
55c888d6
ON
618 base = lock_timer_base(timer, &flags);
619
620 if (timer_pending(timer)) {
621 detach_timer(timer, 0);
622 ret = 1;
74019224
IM
623 } else {
624 if (pending_only)
625 goto out_unlock;
55c888d6
ON
626 }
627
c6f3a97f
TG
628 debug_timer_activate(timer);
629
a4a6198b 630 new_base = __get_cpu_var(tvec_bases);
1da177e4 631
3691c519 632 if (base != new_base) {
1da177e4 633 /*
55c888d6
ON
634 * We are trying to schedule the timer on the local CPU.
635 * However we can't change timer's base while it is running,
636 * otherwise del_timer_sync() can't detect that the timer's
637 * handler yet has not finished. This also guarantees that
638 * the timer is serialized wrt itself.
1da177e4 639 */
a2c348fe 640 if (likely(base->running_timer != timer)) {
55c888d6 641 /* See the comment in lock_timer_base() */
6e453a67 642 timer_set_base(timer, NULL);
55c888d6 643 spin_unlock(&base->lock);
a2c348fe
ON
644 base = new_base;
645 spin_lock(&base->lock);
6e453a67 646 timer_set_base(timer, base);
1da177e4
LT
647 }
648 }
649
1da177e4 650 timer->expires = expires;
a2c348fe 651 internal_add_timer(base, timer);
74019224
IM
652
653out_unlock:
a2c348fe 654 spin_unlock_irqrestore(&base->lock, flags);
1da177e4
LT
655
656 return ret;
657}
658
2aae4a10 659/**
74019224
IM
660 * mod_timer_pending - modify a pending timer's timeout
661 * @timer: the pending timer to be modified
662 * @expires: new timeout in jiffies
1da177e4 663 *
74019224
IM
664 * mod_timer_pending() is the same for pending timers as mod_timer(),
665 * but will not re-activate and modify already deleted timers.
666 *
667 * It is useful for unserialized use of timers.
1da177e4 668 */
74019224 669int mod_timer_pending(struct timer_list *timer, unsigned long expires)
1da177e4 670{
74019224 671 return __mod_timer(timer, expires, true);
1da177e4 672}
74019224 673EXPORT_SYMBOL(mod_timer_pending);
1da177e4 674
2aae4a10 675/**
1da177e4
LT
676 * mod_timer - modify a timer's timeout
677 * @timer: the timer to be modified
2aae4a10 678 * @expires: new timeout in jiffies
1da177e4 679 *
72fd4a35 680 * mod_timer() is a more efficient way to update the expire field of an
1da177e4
LT
681 * active timer (if the timer is inactive it will be activated)
682 *
683 * mod_timer(timer, expires) is equivalent to:
684 *
685 * del_timer(timer); timer->expires = expires; add_timer(timer);
686 *
687 * Note that if there are multiple unserialized concurrent users of the
688 * same timer, then mod_timer() is the only safe way to modify the timeout,
689 * since add_timer() cannot modify an already running timer.
690 *
691 * The function returns whether it has modified a pending timer or not.
692 * (ie. mod_timer() of an inactive timer returns 0, mod_timer() of an
693 * active timer returns 1.)
694 */
695int mod_timer(struct timer_list *timer, unsigned long expires)
696{
1da177e4
LT
697 /*
698 * This is a common optimization triggered by the
699 * networking code - if the timer is re-modified
700 * to be the same thing then just return:
701 */
702 if (timer->expires == expires && timer_pending(timer))
703 return 1;
704
74019224 705 return __mod_timer(timer, expires, false);
1da177e4 706}
1da177e4
LT
707EXPORT_SYMBOL(mod_timer);
708
74019224
IM
709/**
710 * add_timer - start a timer
711 * @timer: the timer to be added
712 *
713 * The kernel will do a ->function(->data) callback from the
714 * timer interrupt at the ->expires point in the future. The
715 * current time is 'jiffies'.
716 *
717 * The timer's ->expires, ->function (and if the handler uses it, ->data)
718 * fields must be set prior calling this function.
719 *
720 * Timers with an ->expires field in the past will be executed in the next
721 * timer tick.
722 */
723void add_timer(struct timer_list *timer)
724{
725 BUG_ON(timer_pending(timer));
726 mod_timer(timer, timer->expires);
727}
728EXPORT_SYMBOL(add_timer);
729
730/**
731 * add_timer_on - start a timer on a particular CPU
732 * @timer: the timer to be added
733 * @cpu: the CPU to start it on
734 *
735 * This is not very scalable on SMP. Double adds are not possible.
736 */
737void add_timer_on(struct timer_list *timer, int cpu)
738{
739 struct tvec_base *base = per_cpu(tvec_bases, cpu);
740 unsigned long flags;
741
742 timer_stats_timer_set_start_info(timer);
743 BUG_ON(timer_pending(timer) || !timer->function);
744 spin_lock_irqsave(&base->lock, flags);
745 timer_set_base(timer, base);
746 debug_timer_activate(timer);
747 internal_add_timer(base, timer);
748 /*
749 * Check whether the other CPU is idle and needs to be
750 * triggered to reevaluate the timer wheel when nohz is
751 * active. We are protected against the other CPU fiddling
752 * with the timer by holding the timer base lock. This also
753 * makes sure that a CPU on the way to idle can not evaluate
754 * the timer wheel.
755 */
756 wake_up_idle_cpu(cpu);
757 spin_unlock_irqrestore(&base->lock, flags);
758}
759
2aae4a10 760/**
1da177e4
LT
761 * del_timer - deactive a timer.
762 * @timer: the timer to be deactivated
763 *
764 * del_timer() deactivates a timer - this works on both active and inactive
765 * timers.
766 *
767 * The function returns whether it has deactivated a pending timer or not.
768 * (ie. del_timer() of an inactive timer returns 0, del_timer() of an
769 * active timer returns 1.)
770 */
771int del_timer(struct timer_list *timer)
772{
a6fa8e5a 773 struct tvec_base *base;
1da177e4 774 unsigned long flags;
55c888d6 775 int ret = 0;
1da177e4 776
82f67cd9 777 timer_stats_timer_clear_start_info(timer);
55c888d6
ON
778 if (timer_pending(timer)) {
779 base = lock_timer_base(timer, &flags);
780 if (timer_pending(timer)) {
781 detach_timer(timer, 1);
782 ret = 1;
783 }
1da177e4 784 spin_unlock_irqrestore(&base->lock, flags);
1da177e4 785 }
1da177e4 786
55c888d6 787 return ret;
1da177e4 788}
1da177e4
LT
789EXPORT_SYMBOL(del_timer);
790
791#ifdef CONFIG_SMP
2aae4a10
REB
792/**
793 * try_to_del_timer_sync - Try to deactivate a timer
794 * @timer: timer do del
795 *
fd450b73
ON
796 * This function tries to deactivate a timer. Upon successful (ret >= 0)
797 * exit the timer is not queued and the handler is not running on any CPU.
798 *
799 * It must not be called from interrupt contexts.
800 */
801int try_to_del_timer_sync(struct timer_list *timer)
802{
a6fa8e5a 803 struct tvec_base *base;
fd450b73
ON
804 unsigned long flags;
805 int ret = -1;
806
807 base = lock_timer_base(timer, &flags);
808
809 if (base->running_timer == timer)
810 goto out;
811
812 ret = 0;
813 if (timer_pending(timer)) {
814 detach_timer(timer, 1);
815 ret = 1;
816 }
817out:
818 spin_unlock_irqrestore(&base->lock, flags);
819
820 return ret;
821}
e19dff1f
DH
822EXPORT_SYMBOL(try_to_del_timer_sync);
823
2aae4a10 824/**
1da177e4
LT
825 * del_timer_sync - deactivate a timer and wait for the handler to finish.
826 * @timer: the timer to be deactivated
827 *
828 * This function only differs from del_timer() on SMP: besides deactivating
829 * the timer it also makes sure the handler has finished executing on other
830 * CPUs.
831 *
72fd4a35 832 * Synchronization rules: Callers must prevent restarting of the timer,
1da177e4
LT
833 * otherwise this function is meaningless. It must not be called from
834 * interrupt contexts. The caller must not hold locks which would prevent
55c888d6
ON
835 * completion of the timer's handler. The timer's handler must not call
836 * add_timer_on(). Upon exit the timer is not queued and the handler is
837 * not running on any CPU.
1da177e4
LT
838 *
839 * The function returns whether it has deactivated a pending timer or not.
1da177e4
LT
840 */
841int del_timer_sync(struct timer_list *timer)
842{
6f2b9b9a
JB
843#ifdef CONFIG_LOCKDEP
844 unsigned long flags;
845
846 local_irq_save(flags);
847 lock_map_acquire(&timer->lockdep_map);
848 lock_map_release(&timer->lockdep_map);
849 local_irq_restore(flags);
850#endif
851
fd450b73
ON
852 for (;;) {
853 int ret = try_to_del_timer_sync(timer);
854 if (ret >= 0)
855 return ret;
a0009652 856 cpu_relax();
fd450b73 857 }
1da177e4 858}
55c888d6 859EXPORT_SYMBOL(del_timer_sync);
1da177e4
LT
860#endif
861
a6fa8e5a 862static int cascade(struct tvec_base *base, struct tvec *tv, int index)
1da177e4
LT
863{
864 /* cascade all the timers from tv up one level */
3439dd86
P
865 struct timer_list *timer, *tmp;
866 struct list_head tv_list;
867
868 list_replace_init(tv->vec + index, &tv_list);
1da177e4 869
1da177e4 870 /*
3439dd86
P
871 * We are removing _all_ timers from the list, so we
872 * don't have to detach them individually.
1da177e4 873 */
3439dd86 874 list_for_each_entry_safe(timer, tmp, &tv_list, entry) {
6e453a67 875 BUG_ON(tbase_get_base(timer->base) != base);
3439dd86 876 internal_add_timer(base, timer);
1da177e4 877 }
1da177e4
LT
878
879 return index;
880}
881
2aae4a10
REB
882#define INDEX(N) ((base->timer_jiffies >> (TVR_BITS + (N) * TVN_BITS)) & TVN_MASK)
883
884/**
1da177e4
LT
885 * __run_timers - run all expired timers (if any) on this CPU.
886 * @base: the timer vector to be processed.
887 *
888 * This function cascades all vectors and executes all expired timer
889 * vectors.
890 */
a6fa8e5a 891static inline void __run_timers(struct tvec_base *base)
1da177e4
LT
892{
893 struct timer_list *timer;
894
3691c519 895 spin_lock_irq(&base->lock);
1da177e4 896 while (time_after_eq(jiffies, base->timer_jiffies)) {
626ab0e6 897 struct list_head work_list;
1da177e4 898 struct list_head *head = &work_list;
6819457d 899 int index = base->timer_jiffies & TVR_MASK;
626ab0e6 900
1da177e4
LT
901 /*
902 * Cascade timers:
903 */
904 if (!index &&
905 (!cascade(base, &base->tv2, INDEX(0))) &&
906 (!cascade(base, &base->tv3, INDEX(1))) &&
907 !cascade(base, &base->tv4, INDEX(2)))
908 cascade(base, &base->tv5, INDEX(3));
626ab0e6
ON
909 ++base->timer_jiffies;
910 list_replace_init(base->tv1.vec + index, &work_list);
55c888d6 911 while (!list_empty(head)) {
1da177e4
LT
912 void (*fn)(unsigned long);
913 unsigned long data;
914
b5e61818 915 timer = list_first_entry(head, struct timer_list,entry);
6819457d
TG
916 fn = timer->function;
917 data = timer->data;
1da177e4 918
82f67cd9
IM
919 timer_stats_account_timer(timer);
920
1da177e4 921 set_running_timer(base, timer);
55c888d6 922 detach_timer(timer, 1);
6f2b9b9a 923
3691c519 924 spin_unlock_irq(&base->lock);
1da177e4 925 {
be5b4fbd 926 int preempt_count = preempt_count();
6f2b9b9a
JB
927
928#ifdef CONFIG_LOCKDEP
929 /*
930 * It is permissible to free the timer from
931 * inside the function that is called from
932 * it, this we need to take into account for
933 * lockdep too. To avoid bogus "held lock
934 * freed" warnings as well as problems when
935 * looking into timer->lockdep_map, make a
936 * copy and use that here.
937 */
938 struct lockdep_map lockdep_map =
939 timer->lockdep_map;
940#endif
941 /*
942 * Couple the lock chain with the lock chain at
943 * del_timer_sync() by acquiring the lock_map
944 * around the fn() call here and in
945 * del_timer_sync().
946 */
947 lock_map_acquire(&lockdep_map);
948
1da177e4 949 fn(data);
6f2b9b9a
JB
950
951 lock_map_release(&lockdep_map);
952
1da177e4 953 if (preempt_count != preempt_count()) {
4c9dc641 954 printk(KERN_ERR "huh, entered %p "
be5b4fbd
JJ
955 "with preempt_count %08x, exited"
956 " with %08x?\n",
957 fn, preempt_count,
958 preempt_count());
1da177e4
LT
959 BUG();
960 }
961 }
3691c519 962 spin_lock_irq(&base->lock);
1da177e4
LT
963 }
964 }
965 set_running_timer(base, NULL);
3691c519 966 spin_unlock_irq(&base->lock);
1da177e4
LT
967}
968
ee9c5785 969#ifdef CONFIG_NO_HZ
1da177e4
LT
970/*
971 * Find out when the next timer event is due to happen. This
972 * is used on S/390 to stop all activity when a cpus is idle.
973 * This functions needs to be called disabled.
974 */
a6fa8e5a 975static unsigned long __next_timer_interrupt(struct tvec_base *base)
1da177e4 976{
1cfd6849 977 unsigned long timer_jiffies = base->timer_jiffies;
eaad084b 978 unsigned long expires = timer_jiffies + NEXT_TIMER_MAX_DELTA;
1cfd6849 979 int index, slot, array, found = 0;
1da177e4 980 struct timer_list *nte;
a6fa8e5a 981 struct tvec *varray[4];
1da177e4
LT
982
983 /* Look for timer events in tv1. */
1cfd6849 984 index = slot = timer_jiffies & TVR_MASK;
1da177e4 985 do {
1cfd6849 986 list_for_each_entry(nte, base->tv1.vec + slot, entry) {
6819457d
TG
987 if (tbase_get_deferrable(nte->base))
988 continue;
6e453a67 989
1cfd6849 990 found = 1;
1da177e4 991 expires = nte->expires;
1cfd6849
TG
992 /* Look at the cascade bucket(s)? */
993 if (!index || slot < index)
994 goto cascade;
995 return expires;
1da177e4 996 }
1cfd6849
TG
997 slot = (slot + 1) & TVR_MASK;
998 } while (slot != index);
999
1000cascade:
1001 /* Calculate the next cascade event */
1002 if (index)
1003 timer_jiffies += TVR_SIZE - index;
1004 timer_jiffies >>= TVR_BITS;
1da177e4
LT
1005
1006 /* Check tv2-tv5. */
1007 varray[0] = &base->tv2;
1008 varray[1] = &base->tv3;
1009 varray[2] = &base->tv4;
1010 varray[3] = &base->tv5;
1cfd6849
TG
1011
1012 for (array = 0; array < 4; array++) {
a6fa8e5a 1013 struct tvec *varp = varray[array];
1cfd6849
TG
1014
1015 index = slot = timer_jiffies & TVN_MASK;
1da177e4 1016 do {
1cfd6849 1017 list_for_each_entry(nte, varp->vec + slot, entry) {
a0419888
JH
1018 if (tbase_get_deferrable(nte->base))
1019 continue;
1020
1cfd6849 1021 found = 1;
1da177e4
LT
1022 if (time_before(nte->expires, expires))
1023 expires = nte->expires;
1cfd6849
TG
1024 }
1025 /*
1026 * Do we still search for the first timer or are
1027 * we looking up the cascade buckets ?
1028 */
1029 if (found) {
1030 /* Look at the cascade bucket(s)? */
1031 if (!index || slot < index)
1032 break;
1033 return expires;
1034 }
1035 slot = (slot + 1) & TVN_MASK;
1036 } while (slot != index);
1037
1038 if (index)
1039 timer_jiffies += TVN_SIZE - index;
1040 timer_jiffies >>= TVN_BITS;
1da177e4 1041 }
1cfd6849
TG
1042 return expires;
1043}
69239749 1044
1cfd6849
TG
1045/*
1046 * Check, if the next hrtimer event is before the next timer wheel
1047 * event:
1048 */
1049static unsigned long cmp_next_hrtimer_event(unsigned long now,
1050 unsigned long expires)
1051{
1052 ktime_t hr_delta = hrtimer_get_next_event();
1053 struct timespec tsdelta;
9501b6cf 1054 unsigned long delta;
1cfd6849
TG
1055
1056 if (hr_delta.tv64 == KTIME_MAX)
1057 return expires;
0662b713 1058
9501b6cf
TG
1059 /*
1060 * Expired timer available, let it expire in the next tick
1061 */
1062 if (hr_delta.tv64 <= 0)
1063 return now + 1;
69239749 1064
1cfd6849 1065 tsdelta = ktime_to_timespec(hr_delta);
9501b6cf 1066 delta = timespec_to_jiffies(&tsdelta);
eaad084b
TG
1067
1068 /*
1069 * Limit the delta to the max value, which is checked in
1070 * tick_nohz_stop_sched_tick():
1071 */
1072 if (delta > NEXT_TIMER_MAX_DELTA)
1073 delta = NEXT_TIMER_MAX_DELTA;
1074
9501b6cf
TG
1075 /*
1076 * Take rounding errors in to account and make sure, that it
1077 * expires in the next tick. Otherwise we go into an endless
1078 * ping pong due to tick_nohz_stop_sched_tick() retriggering
1079 * the timer softirq
1080 */
1081 if (delta < 1)
1082 delta = 1;
1083 now += delta;
1cfd6849
TG
1084 if (time_before(now, expires))
1085 return now;
1da177e4
LT
1086 return expires;
1087}
1cfd6849
TG
1088
1089/**
8dce39c2 1090 * get_next_timer_interrupt - return the jiffy of the next pending timer
05fb6bf0 1091 * @now: current time (in jiffies)
1cfd6849 1092 */
fd064b9b 1093unsigned long get_next_timer_interrupt(unsigned long now)
1cfd6849 1094{
a6fa8e5a 1095 struct tvec_base *base = __get_cpu_var(tvec_bases);
fd064b9b 1096 unsigned long expires;
1cfd6849
TG
1097
1098 spin_lock(&base->lock);
1099 expires = __next_timer_interrupt(base);
1100 spin_unlock(&base->lock);
1101
1102 if (time_before_eq(expires, now))
1103 return now;
1104
1105 return cmp_next_hrtimer_event(now, expires);
1106}
1da177e4
LT
1107#endif
1108
1da177e4 1109/*
5b4db0c2 1110 * Called from the timer interrupt handler to charge one tick to the current
1da177e4
LT
1111 * process. user_tick is 1 if the tick is user time, 0 for system.
1112 */
1113void update_process_times(int user_tick)
1114{
1115 struct task_struct *p = current;
1116 int cpu = smp_processor_id();
1117
1118 /* Note: this timer irq context must be accounted for as well. */
fa13a5a1 1119 account_process_tick(p, user_tick);
1da177e4
LT
1120 run_local_timers();
1121 if (rcu_pending(cpu))
1122 rcu_check_callbacks(cpu, user_tick);
b845b517 1123 printk_tick();
1da177e4 1124 scheduler_tick();
6819457d 1125 run_posix_cpu_timers(p);
1da177e4
LT
1126}
1127
1128/*
1129 * Nr of active tasks - counted in fixed-point numbers
1130 */
1131static unsigned long count_active_tasks(void)
1132{
db1b1fef 1133 return nr_active() * FIXED_1;
1da177e4
LT
1134}
1135
1136/*
1137 * Hmm.. Changed this, as the GNU make sources (load.c) seems to
1138 * imply that avenrun[] is the standard name for this kind of thing.
1139 * Nothing else seems to be standardized: the fractional size etc
1140 * all seem to differ on different machines.
1141 *
1142 * Requires xtime_lock to access.
1143 */
1144unsigned long avenrun[3];
1145
1146EXPORT_SYMBOL(avenrun);
1147
1148/*
1149 * calc_load - given tick count, update the avenrun load estimates.
1150 * This is called while holding a write_lock on xtime_lock.
1151 */
1152static inline void calc_load(unsigned long ticks)
1153{
1154 unsigned long active_tasks; /* fixed-point */
1155 static int count = LOAD_FREQ;
1156
cd7175ed
ED
1157 count -= ticks;
1158 if (unlikely(count < 0)) {
1159 active_tasks = count_active_tasks();
1160 do {
1161 CALC_LOAD(avenrun[0], EXP_1, active_tasks);
1162 CALC_LOAD(avenrun[1], EXP_5, active_tasks);
1163 CALC_LOAD(avenrun[2], EXP_15, active_tasks);
1164 count += LOAD_FREQ;
1165 } while (count < 0);
1da177e4
LT
1166 }
1167}
1168
1da177e4
LT
1169/*
1170 * This function runs timers and the timer-tq in bottom half context.
1171 */
1172static void run_timer_softirq(struct softirq_action *h)
1173{
a6fa8e5a 1174 struct tvec_base *base = __get_cpu_var(tvec_bases);
1da177e4 1175
d3d74453 1176 hrtimer_run_pending();
82f67cd9 1177
1da177e4
LT
1178 if (time_after_eq(jiffies, base->timer_jiffies))
1179 __run_timers(base);
1180}
1181
1182/*
1183 * Called by the local, per-CPU timer interrupt on SMP.
1184 */
1185void run_local_timers(void)
1186{
d3d74453 1187 hrtimer_run_queues();
1da177e4 1188 raise_softirq(TIMER_SOFTIRQ);
6687a97d 1189 softlockup_tick();
1da177e4
LT
1190}
1191
1192/*
1193 * Called by the timer interrupt. xtime_lock must already be taken
1194 * by the timer IRQ!
1195 */
3171a030 1196static inline void update_times(unsigned long ticks)
1da177e4 1197{
ad596171 1198 update_wall_time();
1da177e4
LT
1199 calc_load(ticks);
1200}
6819457d 1201
1da177e4
LT
1202/*
1203 * The 64-bit jiffies value is not atomic - you MUST NOT read it
1204 * without sampling the sequence number in xtime_lock.
1205 * jiffies is defined in the linker script...
1206 */
1207
3171a030 1208void do_timer(unsigned long ticks)
1da177e4 1209{
3171a030
AN
1210 jiffies_64 += ticks;
1211 update_times(ticks);
1da177e4
LT
1212}
1213
1214#ifdef __ARCH_WANT_SYS_ALARM
1215
1216/*
1217 * For backwards compatibility? This can be done in libc so Alpha
1218 * and all newer ports shouldn't need it.
1219 */
58fd3aa2 1220SYSCALL_DEFINE1(alarm, unsigned int, seconds)
1da177e4 1221{
c08b8a49 1222 return alarm_setitimer(seconds);
1da177e4
LT
1223}
1224
1225#endif
1226
1227#ifndef __alpha__
1228
1229/*
1230 * The Alpha uses getxpid, getxuid, and getxgid instead. Maybe this
1231 * should be moved into arch/i386 instead?
1232 */
1233
1234/**
1235 * sys_getpid - return the thread group id of the current process
1236 *
1237 * Note, despite the name, this returns the tgid not the pid. The tgid and
1238 * the pid are identical unless CLONE_THREAD was specified on clone() in
1239 * which case the tgid is the same in all threads of the same group.
1240 *
1241 * This is SMP safe as current->tgid does not change.
1242 */
58fd3aa2 1243SYSCALL_DEFINE0(getpid)
1da177e4 1244{
b488893a 1245 return task_tgid_vnr(current);
1da177e4
LT
1246}
1247
1248/*
6997a6fa
KK
1249 * Accessing ->real_parent is not SMP-safe, it could
1250 * change from under us. However, we can use a stale
1251 * value of ->real_parent under rcu_read_lock(), see
1252 * release_task()->call_rcu(delayed_put_task_struct).
1da177e4 1253 */
dbf040d9 1254SYSCALL_DEFINE0(getppid)
1da177e4
LT
1255{
1256 int pid;
1da177e4 1257
6997a6fa 1258 rcu_read_lock();
6c5f3e7b 1259 pid = task_tgid_vnr(current->real_parent);
6997a6fa 1260 rcu_read_unlock();
1da177e4 1261
1da177e4
LT
1262 return pid;
1263}
1264
dbf040d9 1265SYSCALL_DEFINE0(getuid)
1da177e4
LT
1266{
1267 /* Only we change this so SMP safe */
76aac0e9 1268 return current_uid();
1da177e4
LT
1269}
1270
dbf040d9 1271SYSCALL_DEFINE0(geteuid)
1da177e4
LT
1272{
1273 /* Only we change this so SMP safe */
76aac0e9 1274 return current_euid();
1da177e4
LT
1275}
1276
dbf040d9 1277SYSCALL_DEFINE0(getgid)
1da177e4
LT
1278{
1279 /* Only we change this so SMP safe */
76aac0e9 1280 return current_gid();
1da177e4
LT
1281}
1282
dbf040d9 1283SYSCALL_DEFINE0(getegid)
1da177e4
LT
1284{
1285 /* Only we change this so SMP safe */
76aac0e9 1286 return current_egid();
1da177e4
LT
1287}
1288
1289#endif
1290
1291static void process_timeout(unsigned long __data)
1292{
36c8b586 1293 wake_up_process((struct task_struct *)__data);
1da177e4
LT
1294}
1295
1296/**
1297 * schedule_timeout - sleep until timeout
1298 * @timeout: timeout value in jiffies
1299 *
1300 * Make the current task sleep until @timeout jiffies have
1301 * elapsed. The routine will return immediately unless
1302 * the current task state has been set (see set_current_state()).
1303 *
1304 * You can set the task state as follows -
1305 *
1306 * %TASK_UNINTERRUPTIBLE - at least @timeout jiffies are guaranteed to
1307 * pass before the routine returns. The routine will return 0
1308 *
1309 * %TASK_INTERRUPTIBLE - the routine may return early if a signal is
1310 * delivered to the current task. In this case the remaining time
1311 * in jiffies will be returned, or 0 if the timer expired in time
1312 *
1313 * The current task state is guaranteed to be TASK_RUNNING when this
1314 * routine returns.
1315 *
1316 * Specifying a @timeout value of %MAX_SCHEDULE_TIMEOUT will schedule
1317 * the CPU away without a bound on the timeout. In this case the return
1318 * value will be %MAX_SCHEDULE_TIMEOUT.
1319 *
1320 * In all cases the return value is guaranteed to be non-negative.
1321 */
7ad5b3a5 1322signed long __sched schedule_timeout(signed long timeout)
1da177e4
LT
1323{
1324 struct timer_list timer;
1325 unsigned long expire;
1326
1327 switch (timeout)
1328 {
1329 case MAX_SCHEDULE_TIMEOUT:
1330 /*
1331 * These two special cases are useful to be comfortable
1332 * in the caller. Nothing more. We could take
1333 * MAX_SCHEDULE_TIMEOUT from one of the negative value
1334 * but I' d like to return a valid offset (>=0) to allow
1335 * the caller to do everything it want with the retval.
1336 */
1337 schedule();
1338 goto out;
1339 default:
1340 /*
1341 * Another bit of PARANOID. Note that the retval will be
1342 * 0 since no piece of kernel is supposed to do a check
1343 * for a negative retval of schedule_timeout() (since it
1344 * should never happens anyway). You just have the printk()
1345 * that will tell you if something is gone wrong and where.
1346 */
5b149bcc 1347 if (timeout < 0) {
1da177e4 1348 printk(KERN_ERR "schedule_timeout: wrong timeout "
5b149bcc
AM
1349 "value %lx\n", timeout);
1350 dump_stack();
1da177e4
LT
1351 current->state = TASK_RUNNING;
1352 goto out;
1353 }
1354 }
1355
1356 expire = timeout + jiffies;
1357
c6f3a97f 1358 setup_timer_on_stack(&timer, process_timeout, (unsigned long)current);
74019224 1359 __mod_timer(&timer, expire, false);
1da177e4
LT
1360 schedule();
1361 del_singleshot_timer_sync(&timer);
1362
c6f3a97f
TG
1363 /* Remove the timer from the object tracker */
1364 destroy_timer_on_stack(&timer);
1365
1da177e4
LT
1366 timeout = expire - jiffies;
1367
1368 out:
1369 return timeout < 0 ? 0 : timeout;
1370}
1da177e4
LT
1371EXPORT_SYMBOL(schedule_timeout);
1372
8a1c1757
AM
1373/*
1374 * We can use __set_current_state() here because schedule_timeout() calls
1375 * schedule() unconditionally.
1376 */
64ed93a2
NA
1377signed long __sched schedule_timeout_interruptible(signed long timeout)
1378{
a5a0d52c
AM
1379 __set_current_state(TASK_INTERRUPTIBLE);
1380 return schedule_timeout(timeout);
64ed93a2
NA
1381}
1382EXPORT_SYMBOL(schedule_timeout_interruptible);
1383
294d5cc2
MW
1384signed long __sched schedule_timeout_killable(signed long timeout)
1385{
1386 __set_current_state(TASK_KILLABLE);
1387 return schedule_timeout(timeout);
1388}
1389EXPORT_SYMBOL(schedule_timeout_killable);
1390
64ed93a2
NA
1391signed long __sched schedule_timeout_uninterruptible(signed long timeout)
1392{
a5a0d52c
AM
1393 __set_current_state(TASK_UNINTERRUPTIBLE);
1394 return schedule_timeout(timeout);
64ed93a2
NA
1395}
1396EXPORT_SYMBOL(schedule_timeout_uninterruptible);
1397
1da177e4 1398/* Thread ID - the internal kernel "pid" */
58fd3aa2 1399SYSCALL_DEFINE0(gettid)
1da177e4 1400{
b488893a 1401 return task_pid_vnr(current);
1da177e4
LT
1402}
1403
2aae4a10 1404/**
d4d23add 1405 * do_sysinfo - fill in sysinfo struct
2aae4a10 1406 * @info: pointer to buffer to fill
6819457d 1407 */
d4d23add 1408int do_sysinfo(struct sysinfo *info)
1da177e4 1409{
1da177e4
LT
1410 unsigned long mem_total, sav_total;
1411 unsigned int mem_unit, bitcount;
1412 unsigned long seq;
1413
d4d23add 1414 memset(info, 0, sizeof(struct sysinfo));
1da177e4
LT
1415
1416 do {
1417 struct timespec tp;
1418 seq = read_seqbegin(&xtime_lock);
1419
1420 /*
1421 * This is annoying. The below is the same thing
1422 * posix_get_clock_monotonic() does, but it wants to
1423 * take the lock which we want to cover the loads stuff
1424 * too.
1425 */
1426
1427 getnstimeofday(&tp);
1428 tp.tv_sec += wall_to_monotonic.tv_sec;
1429 tp.tv_nsec += wall_to_monotonic.tv_nsec;
d6214141 1430 monotonic_to_bootbased(&tp);
1da177e4
LT
1431 if (tp.tv_nsec - NSEC_PER_SEC >= 0) {
1432 tp.tv_nsec = tp.tv_nsec - NSEC_PER_SEC;
1433 tp.tv_sec++;
1434 }
d4d23add 1435 info->uptime = tp.tv_sec + (tp.tv_nsec ? 1 : 0);
1da177e4 1436
d4d23add
KM
1437 info->loads[0] = avenrun[0] << (SI_LOAD_SHIFT - FSHIFT);
1438 info->loads[1] = avenrun[1] << (SI_LOAD_SHIFT - FSHIFT);
1439 info->loads[2] = avenrun[2] << (SI_LOAD_SHIFT - FSHIFT);
1da177e4 1440
d4d23add 1441 info->procs = nr_threads;
1da177e4
LT
1442 } while (read_seqretry(&xtime_lock, seq));
1443
d4d23add
KM
1444 si_meminfo(info);
1445 si_swapinfo(info);
1da177e4
LT
1446
1447 /*
1448 * If the sum of all the available memory (i.e. ram + swap)
1449 * is less than can be stored in a 32 bit unsigned long then
1450 * we can be binary compatible with 2.2.x kernels. If not,
1451 * well, in that case 2.2.x was broken anyways...
1452 *
1453 * -Erik Andersen <andersee@debian.org>
1454 */
1455
d4d23add
KM
1456 mem_total = info->totalram + info->totalswap;
1457 if (mem_total < info->totalram || mem_total < info->totalswap)
1da177e4
LT
1458 goto out;
1459 bitcount = 0;
d4d23add 1460 mem_unit = info->mem_unit;
1da177e4
LT
1461 while (mem_unit > 1) {
1462 bitcount++;
1463 mem_unit >>= 1;
1464 sav_total = mem_total;
1465 mem_total <<= 1;
1466 if (mem_total < sav_total)
1467 goto out;
1468 }
1469
1470 /*
1471 * If mem_total did not overflow, multiply all memory values by
d4d23add 1472 * info->mem_unit and set it to 1. This leaves things compatible
1da177e4
LT
1473 * with 2.2.x, and also retains compatibility with earlier 2.4.x
1474 * kernels...
1475 */
1476
d4d23add
KM
1477 info->mem_unit = 1;
1478 info->totalram <<= bitcount;
1479 info->freeram <<= bitcount;
1480 info->sharedram <<= bitcount;
1481 info->bufferram <<= bitcount;
1482 info->totalswap <<= bitcount;
1483 info->freeswap <<= bitcount;
1484 info->totalhigh <<= bitcount;
1485 info->freehigh <<= bitcount;
1486
1487out:
1488 return 0;
1489}
1490
1e7bfb21 1491SYSCALL_DEFINE1(sysinfo, struct sysinfo __user *, info)
d4d23add
KM
1492{
1493 struct sysinfo val;
1494
1495 do_sysinfo(&val);
1da177e4 1496
1da177e4
LT
1497 if (copy_to_user(info, &val, sizeof(struct sysinfo)))
1498 return -EFAULT;
1499
1500 return 0;
1501}
1502
b4be6258 1503static int __cpuinit init_timers_cpu(int cpu)
1da177e4
LT
1504{
1505 int j;
a6fa8e5a 1506 struct tvec_base *base;
b4be6258 1507 static char __cpuinitdata tvec_base_done[NR_CPUS];
55c888d6 1508
ba6edfcd 1509 if (!tvec_base_done[cpu]) {
a4a6198b
JB
1510 static char boot_done;
1511
a4a6198b 1512 if (boot_done) {
ba6edfcd
AM
1513 /*
1514 * The APs use this path later in boot
1515 */
94f6030c
CL
1516 base = kmalloc_node(sizeof(*base),
1517 GFP_KERNEL | __GFP_ZERO,
a4a6198b
JB
1518 cpu_to_node(cpu));
1519 if (!base)
1520 return -ENOMEM;
6e453a67
VP
1521
1522 /* Make sure that tvec_base is 2 byte aligned */
1523 if (tbase_get_deferrable(base)) {
1524 WARN_ON(1);
1525 kfree(base);
1526 return -ENOMEM;
1527 }
ba6edfcd 1528 per_cpu(tvec_bases, cpu) = base;
a4a6198b 1529 } else {
ba6edfcd
AM
1530 /*
1531 * This is for the boot CPU - we use compile-time
1532 * static initialisation because per-cpu memory isn't
1533 * ready yet and because the memory allocators are not
1534 * initialised either.
1535 */
a4a6198b 1536 boot_done = 1;
ba6edfcd 1537 base = &boot_tvec_bases;
a4a6198b 1538 }
ba6edfcd
AM
1539 tvec_base_done[cpu] = 1;
1540 } else {
1541 base = per_cpu(tvec_bases, cpu);
a4a6198b 1542 }
ba6edfcd 1543
3691c519 1544 spin_lock_init(&base->lock);
d730e882 1545
1da177e4
LT
1546 for (j = 0; j < TVN_SIZE; j++) {
1547 INIT_LIST_HEAD(base->tv5.vec + j);
1548 INIT_LIST_HEAD(base->tv4.vec + j);
1549 INIT_LIST_HEAD(base->tv3.vec + j);
1550 INIT_LIST_HEAD(base->tv2.vec + j);
1551 }
1552 for (j = 0; j < TVR_SIZE; j++)
1553 INIT_LIST_HEAD(base->tv1.vec + j);
1554
1555 base->timer_jiffies = jiffies;
a4a6198b 1556 return 0;
1da177e4
LT
1557}
1558
1559#ifdef CONFIG_HOTPLUG_CPU
a6fa8e5a 1560static void migrate_timer_list(struct tvec_base *new_base, struct list_head *head)
1da177e4
LT
1561{
1562 struct timer_list *timer;
1563
1564 while (!list_empty(head)) {
b5e61818 1565 timer = list_first_entry(head, struct timer_list, entry);
55c888d6 1566 detach_timer(timer, 0);
6e453a67 1567 timer_set_base(timer, new_base);
1da177e4 1568 internal_add_timer(new_base, timer);
1da177e4 1569 }
1da177e4
LT
1570}
1571
48ccf3da 1572static void __cpuinit migrate_timers(int cpu)
1da177e4 1573{
a6fa8e5a
PM
1574 struct tvec_base *old_base;
1575 struct tvec_base *new_base;
1da177e4
LT
1576 int i;
1577
1578 BUG_ON(cpu_online(cpu));
a4a6198b
JB
1579 old_base = per_cpu(tvec_bases, cpu);
1580 new_base = get_cpu_var(tvec_bases);
d82f0b0f
ON
1581 /*
1582 * The caller is globally serialized and nobody else
1583 * takes two locks at once, deadlock is not possible.
1584 */
1585 spin_lock_irq(&new_base->lock);
0d180406 1586 spin_lock_nested(&old_base->lock, SINGLE_DEPTH_NESTING);
3691c519
ON
1587
1588 BUG_ON(old_base->running_timer);
1da177e4 1589
1da177e4 1590 for (i = 0; i < TVR_SIZE; i++)
55c888d6
ON
1591 migrate_timer_list(new_base, old_base->tv1.vec + i);
1592 for (i = 0; i < TVN_SIZE; i++) {
1593 migrate_timer_list(new_base, old_base->tv2.vec + i);
1594 migrate_timer_list(new_base, old_base->tv3.vec + i);
1595 migrate_timer_list(new_base, old_base->tv4.vec + i);
1596 migrate_timer_list(new_base, old_base->tv5.vec + i);
1597 }
1598
0d180406 1599 spin_unlock(&old_base->lock);
d82f0b0f 1600 spin_unlock_irq(&new_base->lock);
1da177e4 1601 put_cpu_var(tvec_bases);
1da177e4
LT
1602}
1603#endif /* CONFIG_HOTPLUG_CPU */
1604
8c78f307 1605static int __cpuinit timer_cpu_notify(struct notifier_block *self,
1da177e4
LT
1606 unsigned long action, void *hcpu)
1607{
1608 long cpu = (long)hcpu;
1609 switch(action) {
1610 case CPU_UP_PREPARE:
8bb78442 1611 case CPU_UP_PREPARE_FROZEN:
a4a6198b
JB
1612 if (init_timers_cpu(cpu) < 0)
1613 return NOTIFY_BAD;
1da177e4
LT
1614 break;
1615#ifdef CONFIG_HOTPLUG_CPU
1616 case CPU_DEAD:
8bb78442 1617 case CPU_DEAD_FROZEN:
1da177e4
LT
1618 migrate_timers(cpu);
1619 break;
1620#endif
1621 default:
1622 break;
1623 }
1624 return NOTIFY_OK;
1625}
1626
8c78f307 1627static struct notifier_block __cpuinitdata timers_nb = {
1da177e4
LT
1628 .notifier_call = timer_cpu_notify,
1629};
1630
1631
1632void __init init_timers(void)
1633{
07dccf33 1634 int err = timer_cpu_notify(&timers_nb, (unsigned long)CPU_UP_PREPARE,
1da177e4 1635 (void *)(long)smp_processor_id());
07dccf33 1636
82f67cd9
IM
1637 init_timer_stats();
1638
07dccf33 1639 BUG_ON(err == NOTIFY_BAD);
1da177e4 1640 register_cpu_notifier(&timers_nb);
962cf36c 1641 open_softirq(TIMER_SOFTIRQ, run_timer_softirq);
1da177e4
LT
1642}
1643
1da177e4
LT
1644/**
1645 * msleep - sleep safely even with waitqueue interruptions
1646 * @msecs: Time in milliseconds to sleep for
1647 */
1648void msleep(unsigned int msecs)
1649{
1650 unsigned long timeout = msecs_to_jiffies(msecs) + 1;
1651
75bcc8c5
NA
1652 while (timeout)
1653 timeout = schedule_timeout_uninterruptible(timeout);
1da177e4
LT
1654}
1655
1656EXPORT_SYMBOL(msleep);
1657
1658/**
96ec3efd 1659 * msleep_interruptible - sleep waiting for signals
1da177e4
LT
1660 * @msecs: Time in milliseconds to sleep for
1661 */
1662unsigned long msleep_interruptible(unsigned int msecs)
1663{
1664 unsigned long timeout = msecs_to_jiffies(msecs) + 1;
1665
75bcc8c5
NA
1666 while (timeout && !signal_pending(current))
1667 timeout = schedule_timeout_interruptible(timeout);
1da177e4
LT
1668 return jiffies_to_msecs(timeout);
1669}
1670
1671EXPORT_SYMBOL(msleep_interruptible);