xref: /linux-6.15/include/linux/wait.h (revision 04303f8e)
1 #ifndef _LINUX_WAIT_H
2 #define _LINUX_WAIT_H
3 /*
4  * Linux wait queue related types and methods
5  */
6 #include <linux/list.h>
7 #include <linux/stddef.h>
8 #include <linux/spinlock.h>
9 #include <asm/current.h>
10 #include <uapi/linux/wait.h>
11 
12 typedef struct __wait_queue wait_queue_t;
13 typedef int (*wait_queue_func_t)(wait_queue_t *wait, unsigned mode, int flags, void *key);
14 int default_wake_function(wait_queue_t *wait, unsigned mode, int flags, void *key);
15 
16 /* __wait_queue::flags */
17 #define WQ_FLAG_EXCLUSIVE	0x01
18 #define WQ_FLAG_WOKEN		0x02
19 
20 struct __wait_queue {
21 	unsigned int		flags;
22 	void			*private;
23 	wait_queue_func_t	func;
24 	struct list_head	task_list;
25 };
26 
27 struct wait_bit_key {
28 	void			*flags;
29 	int			bit_nr;
30 #define WAIT_ATOMIC_T_BIT_NR	-1
31 	unsigned long		timeout;
32 };
33 
34 struct wait_bit_queue {
35 	struct wait_bit_key	key;
36 	wait_queue_t		wait;
37 };
38 
39 struct __wait_queue_head {
40 	spinlock_t		lock;
41 	struct list_head	task_list;
42 };
43 typedef struct __wait_queue_head wait_queue_head_t;
44 
45 struct task_struct;
46 
47 /*
48  * Macros for declaration and initialisaton of the datatypes
49  */
50 
51 #define __WAITQUEUE_INITIALIZER(name, tsk) {				\
52 	.private	= tsk,						\
53 	.func		= default_wake_function,			\
54 	.task_list	= { NULL, NULL } }
55 
56 #define DECLARE_WAITQUEUE(name, tsk)					\
57 	wait_queue_t name = __WAITQUEUE_INITIALIZER(name, tsk)
58 
59 #define __WAIT_QUEUE_HEAD_INITIALIZER(name) {				\
60 	.lock		= __SPIN_LOCK_UNLOCKED(name.lock),		\
61 	.task_list	= { &(name).task_list, &(name).task_list } }
62 
63 #define DECLARE_WAIT_QUEUE_HEAD(name) \
64 	wait_queue_head_t name = __WAIT_QUEUE_HEAD_INITIALIZER(name)
65 
66 #define __WAIT_BIT_KEY_INITIALIZER(word, bit)				\
67 	{ .flags = word, .bit_nr = bit, }
68 
69 #define __WAIT_ATOMIC_T_KEY_INITIALIZER(p)				\
70 	{ .flags = p, .bit_nr = WAIT_ATOMIC_T_BIT_NR, }
71 
72 extern void __init_waitqueue_head(wait_queue_head_t *q, const char *name, struct lock_class_key *);
73 
74 #define init_waitqueue_head(q)				\
75 	do {						\
76 		static struct lock_class_key __key;	\
77 							\
78 		__init_waitqueue_head((q), #q, &__key);	\
79 	} while (0)
80 
81 #ifdef CONFIG_LOCKDEP
82 # define __WAIT_QUEUE_HEAD_INIT_ONSTACK(name) \
83 	({ init_waitqueue_head(&name); name; })
84 # define DECLARE_WAIT_QUEUE_HEAD_ONSTACK(name) \
85 	wait_queue_head_t name = __WAIT_QUEUE_HEAD_INIT_ONSTACK(name)
86 #else
87 # define DECLARE_WAIT_QUEUE_HEAD_ONSTACK(name) DECLARE_WAIT_QUEUE_HEAD(name)
88 #endif
89 
90 static inline void init_waitqueue_entry(wait_queue_t *q, struct task_struct *p)
91 {
92 	q->flags	= 0;
93 	q->private	= p;
94 	q->func		= default_wake_function;
95 }
96 
97 static inline void
98 init_waitqueue_func_entry(wait_queue_t *q, wait_queue_func_t func)
99 {
100 	q->flags	= 0;
101 	q->private	= NULL;
102 	q->func		= func;
103 }
104 
105 static inline int waitqueue_active(wait_queue_head_t *q)
106 {
107 	return !list_empty(&q->task_list);
108 }
109 
110 extern void add_wait_queue(wait_queue_head_t *q, wait_queue_t *wait);
111 extern void add_wait_queue_exclusive(wait_queue_head_t *q, wait_queue_t *wait);
112 extern void remove_wait_queue(wait_queue_head_t *q, wait_queue_t *wait);
113 
114 static inline void __add_wait_queue(wait_queue_head_t *head, wait_queue_t *new)
115 {
116 	list_add(&new->task_list, &head->task_list);
117 }
118 
119 /*
120  * Used for wake-one threads:
121  */
122 static inline void
123 __add_wait_queue_exclusive(wait_queue_head_t *q, wait_queue_t *wait)
124 {
125 	wait->flags |= WQ_FLAG_EXCLUSIVE;
126 	__add_wait_queue(q, wait);
127 }
128 
129 static inline void __add_wait_queue_tail(wait_queue_head_t *head,
130 					 wait_queue_t *new)
131 {
132 	list_add_tail(&new->task_list, &head->task_list);
133 }
134 
135 static inline void
136 __add_wait_queue_tail_exclusive(wait_queue_head_t *q, wait_queue_t *wait)
137 {
138 	wait->flags |= WQ_FLAG_EXCLUSIVE;
139 	__add_wait_queue_tail(q, wait);
140 }
141 
142 static inline void
143 __remove_wait_queue(wait_queue_head_t *head, wait_queue_t *old)
144 {
145 	list_del(&old->task_list);
146 }
147 
148 typedef int wait_bit_action_f(struct wait_bit_key *);
149 void __wake_up(wait_queue_head_t *q, unsigned int mode, int nr, void *key);
150 void __wake_up_locked_key(wait_queue_head_t *q, unsigned int mode, void *key);
151 void __wake_up_sync_key(wait_queue_head_t *q, unsigned int mode, int nr, void *key);
152 void __wake_up_locked(wait_queue_head_t *q, unsigned int mode, int nr);
153 void __wake_up_sync(wait_queue_head_t *q, unsigned int mode, int nr);
154 void __wake_up_bit(wait_queue_head_t *, void *, int);
155 int __wait_on_bit(wait_queue_head_t *, struct wait_bit_queue *, wait_bit_action_f *, unsigned);
156 int __wait_on_bit_lock(wait_queue_head_t *, struct wait_bit_queue *, wait_bit_action_f *, unsigned);
157 void wake_up_bit(void *, int);
158 void wake_up_atomic_t(atomic_t *);
159 int out_of_line_wait_on_bit(void *, int, wait_bit_action_f *, unsigned);
160 int out_of_line_wait_on_bit_timeout(void *, int, wait_bit_action_f *, unsigned, unsigned long);
161 int out_of_line_wait_on_bit_lock(void *, int, wait_bit_action_f *, unsigned);
162 int out_of_line_wait_on_atomic_t(atomic_t *, int (*)(atomic_t *), unsigned);
163 wait_queue_head_t *bit_waitqueue(void *, int);
164 
165 #define wake_up(x)			__wake_up(x, TASK_NORMAL, 1, NULL)
166 #define wake_up_nr(x, nr)		__wake_up(x, TASK_NORMAL, nr, NULL)
167 #define wake_up_all(x)			__wake_up(x, TASK_NORMAL, 0, NULL)
168 #define wake_up_locked(x)		__wake_up_locked((x), TASK_NORMAL, 1)
169 #define wake_up_all_locked(x)		__wake_up_locked((x), TASK_NORMAL, 0)
170 
171 #define wake_up_interruptible(x)	__wake_up(x, TASK_INTERRUPTIBLE, 1, NULL)
172 #define wake_up_interruptible_nr(x, nr)	__wake_up(x, TASK_INTERRUPTIBLE, nr, NULL)
173 #define wake_up_interruptible_all(x)	__wake_up(x, TASK_INTERRUPTIBLE, 0, NULL)
174 #define wake_up_interruptible_sync(x)	__wake_up_sync((x), TASK_INTERRUPTIBLE, 1)
175 
176 /*
177  * Wakeup macros to be used to report events to the targets.
178  */
179 #define wake_up_poll(x, m)						\
180 	__wake_up(x, TASK_NORMAL, 1, (void *) (m))
181 #define wake_up_locked_poll(x, m)					\
182 	__wake_up_locked_key((x), TASK_NORMAL, (void *) (m))
183 #define wake_up_interruptible_poll(x, m)				\
184 	__wake_up(x, TASK_INTERRUPTIBLE, 1, (void *) (m))
185 #define wake_up_interruptible_sync_poll(x, m)				\
186 	__wake_up_sync_key((x), TASK_INTERRUPTIBLE, 1, (void *) (m))
187 
188 #define ___wait_cond_timeout(condition)					\
189 ({									\
190 	bool __cond = (condition);					\
191 	if (__cond && !__ret)						\
192 		__ret = 1;						\
193 	__cond || !__ret;						\
194 })
195 
196 #define ___wait_is_interruptible(state)					\
197 	(!__builtin_constant_p(state) ||				\
198 		state == TASK_INTERRUPTIBLE || state == TASK_KILLABLE)	\
199 
200 /*
201  * The below macro ___wait_event() has an explicit shadow of the __ret
202  * variable when used from the wait_event_*() macros.
203  *
204  * This is so that both can use the ___wait_cond_timeout() construct
205  * to wrap the condition.
206  *
207  * The type inconsistency of the wait_event_*() __ret variable is also
208  * on purpose; we use long where we can return timeout values and int
209  * otherwise.
210  */
211 
212 #define ___wait_event(wq, condition, state, exclusive, ret, cmd)	\
213 ({									\
214 	__label__ __out;						\
215 	wait_queue_t __wait;						\
216 	long __ret = ret;	/* explicit shadow */			\
217 									\
218 	INIT_LIST_HEAD(&__wait.task_list);				\
219 	if (exclusive)							\
220 		__wait.flags = WQ_FLAG_EXCLUSIVE;			\
221 	else								\
222 		__wait.flags = 0;					\
223 									\
224 	for (;;) {							\
225 		long __int = prepare_to_wait_event(&wq, &__wait, state);\
226 									\
227 		if (condition)						\
228 			break;						\
229 									\
230 		if (___wait_is_interruptible(state) && __int) {		\
231 			__ret = __int;					\
232 			if (exclusive) {				\
233 				abort_exclusive_wait(&wq, &__wait,	\
234 						     state, NULL);	\
235 				goto __out;				\
236 			}						\
237 			break;						\
238 		}							\
239 									\
240 		cmd;							\
241 	}								\
242 	finish_wait(&wq, &__wait);					\
243 __out:	__ret;								\
244 })
245 
246 #define __wait_event(wq, condition)					\
247 	(void)___wait_event(wq, condition, TASK_UNINTERRUPTIBLE, 0, 0,	\
248 			    schedule())
249 
250 /**
251  * wait_event - sleep until a condition gets true
252  * @wq: the waitqueue to wait on
253  * @condition: a C expression for the event to wait for
254  *
255  * The process is put to sleep (TASK_UNINTERRUPTIBLE) until the
256  * @condition evaluates to true. The @condition is checked each time
257  * the waitqueue @wq is woken up.
258  *
259  * wake_up() has to be called after changing any variable that could
260  * change the result of the wait condition.
261  */
262 #define wait_event(wq, condition)					\
263 do {									\
264 	might_sleep();							\
265 	if (condition)							\
266 		break;							\
267 	__wait_event(wq, condition);					\
268 } while (0)
269 
270 #define __wait_event_freezable(wq, condition)				\
271 	___wait_event(wq, condition, TASK_INTERRUPTIBLE, 0, 0,		\
272 			    schedule(); try_to_freeze())
273 
274 /**
275  * wait_event - sleep (or freeze) until a condition gets true
276  * @wq: the waitqueue to wait on
277  * @condition: a C expression for the event to wait for
278  *
279  * The process is put to sleep (TASK_INTERRUPTIBLE -- so as not to contribute
280  * to system load) until the @condition evaluates to true. The
281  * @condition is checked each time the waitqueue @wq is woken up.
282  *
283  * wake_up() has to be called after changing any variable that could
284  * change the result of the wait condition.
285  */
286 #define wait_event_freezable(wq, condition)				\
287 ({									\
288 	int __ret = 0;							\
289 	might_sleep();							\
290 	if (!(condition))						\
291 		__ret = __wait_event_freezable(wq, condition);		\
292 	__ret;								\
293 })
294 
295 #define __wait_event_timeout(wq, condition, timeout)			\
296 	___wait_event(wq, ___wait_cond_timeout(condition),		\
297 		      TASK_UNINTERRUPTIBLE, 0, timeout,			\
298 		      __ret = schedule_timeout(__ret))
299 
300 /**
301  * wait_event_timeout - sleep until a condition gets true or a timeout elapses
302  * @wq: the waitqueue to wait on
303  * @condition: a C expression for the event to wait for
304  * @timeout: timeout, in jiffies
305  *
306  * The process is put to sleep (TASK_UNINTERRUPTIBLE) until the
307  * @condition evaluates to true. The @condition is checked each time
308  * the waitqueue @wq is woken up.
309  *
310  * wake_up() has to be called after changing any variable that could
311  * change the result of the wait condition.
312  *
313  * Returns:
314  * 0 if the @condition evaluated to %false after the @timeout elapsed,
315  * 1 if the @condition evaluated to %true after the @timeout elapsed,
316  * or the remaining jiffies (at least 1) if the @condition evaluated
317  * to %true before the @timeout elapsed.
318  */
319 #define wait_event_timeout(wq, condition, timeout)			\
320 ({									\
321 	long __ret = timeout;						\
322 	might_sleep();							\
323 	if (!___wait_cond_timeout(condition))				\
324 		__ret = __wait_event_timeout(wq, condition, timeout);	\
325 	__ret;								\
326 })
327 
328 #define __wait_event_freezable_timeout(wq, condition, timeout)		\
329 	___wait_event(wq, ___wait_cond_timeout(condition),		\
330 		      TASK_INTERRUPTIBLE, 0, timeout,			\
331 		      __ret = schedule_timeout(__ret); try_to_freeze())
332 
333 /*
334  * like wait_event_timeout() -- except it uses TASK_INTERRUPTIBLE to avoid
335  * increasing load and is freezable.
336  */
337 #define wait_event_freezable_timeout(wq, condition, timeout)		\
338 ({									\
339 	long __ret = timeout;						\
340 	might_sleep();							\
341 	if (!___wait_cond_timeout(condition))				\
342 		__ret = __wait_event_freezable_timeout(wq, condition, timeout);	\
343 	__ret;								\
344 })
345 
346 #define __wait_event_cmd(wq, condition, cmd1, cmd2)			\
347 	(void)___wait_event(wq, condition, TASK_UNINTERRUPTIBLE, 0, 0,	\
348 			    cmd1; schedule(); cmd2)
349 
350 /**
351  * wait_event_cmd - sleep until a condition gets true
352  * @wq: the waitqueue to wait on
353  * @condition: a C expression for the event to wait for
354  * @cmd1: the command will be executed before sleep
355  * @cmd2: the command will be executed after sleep
356  *
357  * The process is put to sleep (TASK_UNINTERRUPTIBLE) until the
358  * @condition evaluates to true. The @condition is checked each time
359  * the waitqueue @wq is woken up.
360  *
361  * wake_up() has to be called after changing any variable that could
362  * change the result of the wait condition.
363  */
364 #define wait_event_cmd(wq, condition, cmd1, cmd2)			\
365 do {									\
366 	if (condition)							\
367 		break;							\
368 	__wait_event_cmd(wq, condition, cmd1, cmd2);			\
369 } while (0)
370 
371 #define __wait_event_interruptible(wq, condition)			\
372 	___wait_event(wq, condition, TASK_INTERRUPTIBLE, 0, 0,		\
373 		      schedule())
374 
375 /**
376  * wait_event_interruptible - sleep until a condition gets true
377  * @wq: the waitqueue to wait on
378  * @condition: a C expression for the event to wait for
379  *
380  * The process is put to sleep (TASK_INTERRUPTIBLE) until the
381  * @condition evaluates to true or a signal is received.
382  * The @condition is checked each time the waitqueue @wq is woken up.
383  *
384  * wake_up() has to be called after changing any variable that could
385  * change the result of the wait condition.
386  *
387  * The function will return -ERESTARTSYS if it was interrupted by a
388  * signal and 0 if @condition evaluated to true.
389  */
390 #define wait_event_interruptible(wq, condition)				\
391 ({									\
392 	int __ret = 0;							\
393 	might_sleep();							\
394 	if (!(condition))						\
395 		__ret = __wait_event_interruptible(wq, condition);	\
396 	__ret;								\
397 })
398 
399 #define __wait_event_interruptible_timeout(wq, condition, timeout)	\
400 	___wait_event(wq, ___wait_cond_timeout(condition),		\
401 		      TASK_INTERRUPTIBLE, 0, timeout,			\
402 		      __ret = schedule_timeout(__ret))
403 
404 /**
405  * wait_event_interruptible_timeout - sleep until a condition gets true or a timeout elapses
406  * @wq: the waitqueue to wait on
407  * @condition: a C expression for the event to wait for
408  * @timeout: timeout, in jiffies
409  *
410  * The process is put to sleep (TASK_INTERRUPTIBLE) until the
411  * @condition evaluates to true or a signal is received.
412  * The @condition is checked each time the waitqueue @wq is woken up.
413  *
414  * wake_up() has to be called after changing any variable that could
415  * change the result of the wait condition.
416  *
417  * Returns:
418  * 0 if the @condition evaluated to %false after the @timeout elapsed,
419  * 1 if the @condition evaluated to %true after the @timeout elapsed,
420  * the remaining jiffies (at least 1) if the @condition evaluated
421  * to %true before the @timeout elapsed, or -%ERESTARTSYS if it was
422  * interrupted by a signal.
423  */
424 #define wait_event_interruptible_timeout(wq, condition, timeout)	\
425 ({									\
426 	long __ret = timeout;						\
427 	might_sleep();							\
428 	if (!___wait_cond_timeout(condition))				\
429 		__ret = __wait_event_interruptible_timeout(wq,		\
430 						condition, timeout);	\
431 	__ret;								\
432 })
433 
434 #define __wait_event_hrtimeout(wq, condition, timeout, state)		\
435 ({									\
436 	int __ret = 0;							\
437 	struct hrtimer_sleeper __t;					\
438 									\
439 	hrtimer_init_on_stack(&__t.timer, CLOCK_MONOTONIC,		\
440 			      HRTIMER_MODE_REL);			\
441 	hrtimer_init_sleeper(&__t, current);				\
442 	if ((timeout).tv64 != KTIME_MAX)				\
443 		hrtimer_start_range_ns(&__t.timer, timeout,		\
444 				       current->timer_slack_ns,		\
445 				       HRTIMER_MODE_REL);		\
446 									\
447 	__ret = ___wait_event(wq, condition, state, 0, 0,		\
448 		if (!__t.task) {					\
449 			__ret = -ETIME;					\
450 			break;						\
451 		}							\
452 		schedule());						\
453 									\
454 	hrtimer_cancel(&__t.timer);					\
455 	destroy_hrtimer_on_stack(&__t.timer);				\
456 	__ret;								\
457 })
458 
459 /**
460  * wait_event_hrtimeout - sleep until a condition gets true or a timeout elapses
461  * @wq: the waitqueue to wait on
462  * @condition: a C expression for the event to wait for
463  * @timeout: timeout, as a ktime_t
464  *
465  * The process is put to sleep (TASK_UNINTERRUPTIBLE) until the
466  * @condition evaluates to true or a signal is received.
467  * The @condition is checked each time the waitqueue @wq is woken up.
468  *
469  * wake_up() has to be called after changing any variable that could
470  * change the result of the wait condition.
471  *
472  * The function returns 0 if @condition became true, or -ETIME if the timeout
473  * elapsed.
474  */
475 #define wait_event_hrtimeout(wq, condition, timeout)			\
476 ({									\
477 	int __ret = 0;							\
478 	might_sleep();							\
479 	if (!(condition))						\
480 		__ret = __wait_event_hrtimeout(wq, condition, timeout,	\
481 					       TASK_UNINTERRUPTIBLE);	\
482 	__ret;								\
483 })
484 
485 /**
486  * wait_event_interruptible_hrtimeout - sleep until a condition gets true or a timeout elapses
487  * @wq: the waitqueue to wait on
488  * @condition: a C expression for the event to wait for
489  * @timeout: timeout, as a ktime_t
490  *
491  * The process is put to sleep (TASK_INTERRUPTIBLE) until the
492  * @condition evaluates to true or a signal is received.
493  * The @condition is checked each time the waitqueue @wq is woken up.
494  *
495  * wake_up() has to be called after changing any variable that could
496  * change the result of the wait condition.
497  *
498  * The function returns 0 if @condition became true, -ERESTARTSYS if it was
499  * interrupted by a signal, or -ETIME if the timeout elapsed.
500  */
501 #define wait_event_interruptible_hrtimeout(wq, condition, timeout)	\
502 ({									\
503 	long __ret = 0;							\
504 	might_sleep();							\
505 	if (!(condition))						\
506 		__ret = __wait_event_hrtimeout(wq, condition, timeout,	\
507 					       TASK_INTERRUPTIBLE);	\
508 	__ret;								\
509 })
510 
511 #define __wait_event_interruptible_exclusive(wq, condition)		\
512 	___wait_event(wq, condition, TASK_INTERRUPTIBLE, 1, 0,		\
513 		      schedule())
514 
515 #define wait_event_interruptible_exclusive(wq, condition)		\
516 ({									\
517 	int __ret = 0;							\
518 	might_sleep();							\
519 	if (!(condition))						\
520 		__ret = __wait_event_interruptible_exclusive(wq, condition);\
521 	__ret;								\
522 })
523 
524 
525 #define __wait_event_freezable_exclusive(wq, condition)			\
526 	___wait_event(wq, condition, TASK_INTERRUPTIBLE, 1, 0,		\
527 			schedule(); try_to_freeze())
528 
529 #define wait_event_freezable_exclusive(wq, condition)			\
530 ({									\
531 	int __ret = 0;							\
532 	might_sleep();							\
533 	if (!(condition))						\
534 		__ret = __wait_event_freezable_exclusive(wq, condition);\
535 	__ret;								\
536 })
537 
538 
539 #define __wait_event_interruptible_locked(wq, condition, exclusive, irq) \
540 ({									\
541 	int __ret = 0;							\
542 	DEFINE_WAIT(__wait);						\
543 	if (exclusive)							\
544 		__wait.flags |= WQ_FLAG_EXCLUSIVE;			\
545 	do {								\
546 		if (likely(list_empty(&__wait.task_list)))		\
547 			__add_wait_queue_tail(&(wq), &__wait);		\
548 		set_current_state(TASK_INTERRUPTIBLE);			\
549 		if (signal_pending(current)) {				\
550 			__ret = -ERESTARTSYS;				\
551 			break;						\
552 		}							\
553 		if (irq)						\
554 			spin_unlock_irq(&(wq).lock);			\
555 		else							\
556 			spin_unlock(&(wq).lock);			\
557 		schedule();						\
558 		if (irq)						\
559 			spin_lock_irq(&(wq).lock);			\
560 		else							\
561 			spin_lock(&(wq).lock);				\
562 	} while (!(condition));						\
563 	__remove_wait_queue(&(wq), &__wait);				\
564 	__set_current_state(TASK_RUNNING);				\
565 	__ret;								\
566 })
567 
568 
569 /**
570  * wait_event_interruptible_locked - sleep until a condition gets true
571  * @wq: the waitqueue to wait on
572  * @condition: a C expression for the event to wait for
573  *
574  * The process is put to sleep (TASK_INTERRUPTIBLE) until the
575  * @condition evaluates to true or a signal is received.
576  * The @condition is checked each time the waitqueue @wq is woken up.
577  *
578  * It must be called with wq.lock being held.  This spinlock is
579  * unlocked while sleeping but @condition testing is done while lock
580  * is held and when this macro exits the lock is held.
581  *
582  * The lock is locked/unlocked using spin_lock()/spin_unlock()
583  * functions which must match the way they are locked/unlocked outside
584  * of this macro.
585  *
586  * wake_up_locked() has to be called after changing any variable that could
587  * change the result of the wait condition.
588  *
589  * The function will return -ERESTARTSYS if it was interrupted by a
590  * signal and 0 if @condition evaluated to true.
591  */
592 #define wait_event_interruptible_locked(wq, condition)			\
593 	((condition)							\
594 	 ? 0 : __wait_event_interruptible_locked(wq, condition, 0, 0))
595 
596 /**
597  * wait_event_interruptible_locked_irq - sleep until a condition gets true
598  * @wq: the waitqueue to wait on
599  * @condition: a C expression for the event to wait for
600  *
601  * The process is put to sleep (TASK_INTERRUPTIBLE) until the
602  * @condition evaluates to true or a signal is received.
603  * The @condition is checked each time the waitqueue @wq is woken up.
604  *
605  * It must be called with wq.lock being held.  This spinlock is
606  * unlocked while sleeping but @condition testing is done while lock
607  * is held and when this macro exits the lock is held.
608  *
609  * The lock is locked/unlocked using spin_lock_irq()/spin_unlock_irq()
610  * functions which must match the way they are locked/unlocked outside
611  * of this macro.
612  *
613  * wake_up_locked() has to be called after changing any variable that could
614  * change the result of the wait condition.
615  *
616  * The function will return -ERESTARTSYS if it was interrupted by a
617  * signal and 0 if @condition evaluated to true.
618  */
619 #define wait_event_interruptible_locked_irq(wq, condition)		\
620 	((condition)							\
621 	 ? 0 : __wait_event_interruptible_locked(wq, condition, 0, 1))
622 
623 /**
624  * wait_event_interruptible_exclusive_locked - sleep exclusively until a condition gets true
625  * @wq: the waitqueue to wait on
626  * @condition: a C expression for the event to wait for
627  *
628  * The process is put to sleep (TASK_INTERRUPTIBLE) until the
629  * @condition evaluates to true or a signal is received.
630  * The @condition is checked each time the waitqueue @wq is woken up.
631  *
632  * It must be called with wq.lock being held.  This spinlock is
633  * unlocked while sleeping but @condition testing is done while lock
634  * is held and when this macro exits the lock is held.
635  *
636  * The lock is locked/unlocked using spin_lock()/spin_unlock()
637  * functions which must match the way they are locked/unlocked outside
638  * of this macro.
639  *
640  * The process is put on the wait queue with an WQ_FLAG_EXCLUSIVE flag
641  * set thus when other process waits process on the list if this
642  * process is awaken further processes are not considered.
643  *
644  * wake_up_locked() has to be called after changing any variable that could
645  * change the result of the wait condition.
646  *
647  * The function will return -ERESTARTSYS if it was interrupted by a
648  * signal and 0 if @condition evaluated to true.
649  */
650 #define wait_event_interruptible_exclusive_locked(wq, condition)	\
651 	((condition)							\
652 	 ? 0 : __wait_event_interruptible_locked(wq, condition, 1, 0))
653 
654 /**
655  * wait_event_interruptible_exclusive_locked_irq - sleep until a condition gets true
656  * @wq: the waitqueue to wait on
657  * @condition: a C expression for the event to wait for
658  *
659  * The process is put to sleep (TASK_INTERRUPTIBLE) until the
660  * @condition evaluates to true or a signal is received.
661  * The @condition is checked each time the waitqueue @wq is woken up.
662  *
663  * It must be called with wq.lock being held.  This spinlock is
664  * unlocked while sleeping but @condition testing is done while lock
665  * is held and when this macro exits the lock is held.
666  *
667  * The lock is locked/unlocked using spin_lock_irq()/spin_unlock_irq()
668  * functions which must match the way they are locked/unlocked outside
669  * of this macro.
670  *
671  * The process is put on the wait queue with an WQ_FLAG_EXCLUSIVE flag
672  * set thus when other process waits process on the list if this
673  * process is awaken further processes are not considered.
674  *
675  * wake_up_locked() has to be called after changing any variable that could
676  * change the result of the wait condition.
677  *
678  * The function will return -ERESTARTSYS if it was interrupted by a
679  * signal and 0 if @condition evaluated to true.
680  */
681 #define wait_event_interruptible_exclusive_locked_irq(wq, condition)	\
682 	((condition)							\
683 	 ? 0 : __wait_event_interruptible_locked(wq, condition, 1, 1))
684 
685 
686 #define __wait_event_killable(wq, condition)				\
687 	___wait_event(wq, condition, TASK_KILLABLE, 0, 0, schedule())
688 
689 /**
690  * wait_event_killable - sleep until a condition gets true
691  * @wq: the waitqueue to wait on
692  * @condition: a C expression for the event to wait for
693  *
694  * The process is put to sleep (TASK_KILLABLE) until the
695  * @condition evaluates to true or a signal is received.
696  * The @condition is checked each time the waitqueue @wq is woken up.
697  *
698  * wake_up() has to be called after changing any variable that could
699  * change the result of the wait condition.
700  *
701  * The function will return -ERESTARTSYS if it was interrupted by a
702  * signal and 0 if @condition evaluated to true.
703  */
704 #define wait_event_killable(wq, condition)				\
705 ({									\
706 	int __ret = 0;							\
707 	might_sleep();							\
708 	if (!(condition))						\
709 		__ret = __wait_event_killable(wq, condition);		\
710 	__ret;								\
711 })
712 
713 
714 #define __wait_event_lock_irq(wq, condition, lock, cmd)			\
715 	(void)___wait_event(wq, condition, TASK_UNINTERRUPTIBLE, 0, 0,	\
716 			    spin_unlock_irq(&lock);			\
717 			    cmd;					\
718 			    schedule();					\
719 			    spin_lock_irq(&lock))
720 
721 /**
722  * wait_event_lock_irq_cmd - sleep until a condition gets true. The
723  *			     condition is checked under the lock. This
724  *			     is expected to be called with the lock
725  *			     taken.
726  * @wq: the waitqueue to wait on
727  * @condition: a C expression for the event to wait for
728  * @lock: a locked spinlock_t, which will be released before cmd
729  *	  and schedule() and reacquired afterwards.
730  * @cmd: a command which is invoked outside the critical section before
731  *	 sleep
732  *
733  * The process is put to sleep (TASK_UNINTERRUPTIBLE) until the
734  * @condition evaluates to true. The @condition is checked each time
735  * the waitqueue @wq is woken up.
736  *
737  * wake_up() has to be called after changing any variable that could
738  * change the result of the wait condition.
739  *
740  * This is supposed to be called while holding the lock. The lock is
741  * dropped before invoking the cmd and going to sleep and is reacquired
742  * afterwards.
743  */
744 #define wait_event_lock_irq_cmd(wq, condition, lock, cmd)		\
745 do {									\
746 	if (condition)							\
747 		break;							\
748 	__wait_event_lock_irq(wq, condition, lock, cmd);		\
749 } while (0)
750 
751 /**
752  * wait_event_lock_irq - sleep until a condition gets true. The
753  *			 condition is checked under the lock. This
754  *			 is expected to be called with the lock
755  *			 taken.
756  * @wq: the waitqueue to wait on
757  * @condition: a C expression for the event to wait for
758  * @lock: a locked spinlock_t, which will be released before schedule()
759  *	  and reacquired afterwards.
760  *
761  * The process is put to sleep (TASK_UNINTERRUPTIBLE) until the
762  * @condition evaluates to true. The @condition is checked each time
763  * the waitqueue @wq is woken up.
764  *
765  * wake_up() has to be called after changing any variable that could
766  * change the result of the wait condition.
767  *
768  * This is supposed to be called while holding the lock. The lock is
769  * dropped before going to sleep and is reacquired afterwards.
770  */
771 #define wait_event_lock_irq(wq, condition, lock)			\
772 do {									\
773 	if (condition)							\
774 		break;							\
775 	__wait_event_lock_irq(wq, condition, lock, );			\
776 } while (0)
777 
778 
779 #define __wait_event_interruptible_lock_irq(wq, condition, lock, cmd)	\
780 	___wait_event(wq, condition, TASK_INTERRUPTIBLE, 0, 0,		\
781 		      spin_unlock_irq(&lock);				\
782 		      cmd;						\
783 		      schedule();					\
784 		      spin_lock_irq(&lock))
785 
786 /**
787  * wait_event_interruptible_lock_irq_cmd - sleep until a condition gets true.
788  *		The condition is checked under the lock. This is expected to
789  *		be called with the lock taken.
790  * @wq: the waitqueue to wait on
791  * @condition: a C expression for the event to wait for
792  * @lock: a locked spinlock_t, which will be released before cmd and
793  *	  schedule() and reacquired afterwards.
794  * @cmd: a command which is invoked outside the critical section before
795  *	 sleep
796  *
797  * The process is put to sleep (TASK_INTERRUPTIBLE) until the
798  * @condition evaluates to true or a signal is received. The @condition is
799  * checked each time the waitqueue @wq is woken up.
800  *
801  * wake_up() has to be called after changing any variable that could
802  * change the result of the wait condition.
803  *
804  * This is supposed to be called while holding the lock. The lock is
805  * dropped before invoking the cmd and going to sleep and is reacquired
806  * afterwards.
807  *
808  * The macro will return -ERESTARTSYS if it was interrupted by a signal
809  * and 0 if @condition evaluated to true.
810  */
811 #define wait_event_interruptible_lock_irq_cmd(wq, condition, lock, cmd)	\
812 ({									\
813 	int __ret = 0;							\
814 	if (!(condition))						\
815 		__ret = __wait_event_interruptible_lock_irq(wq,		\
816 						condition, lock, cmd);	\
817 	__ret;								\
818 })
819 
820 /**
821  * wait_event_interruptible_lock_irq - sleep until a condition gets true.
822  *		The condition is checked under the lock. This is expected
823  *		to be called with the lock taken.
824  * @wq: the waitqueue to wait on
825  * @condition: a C expression for the event to wait for
826  * @lock: a locked spinlock_t, which will be released before schedule()
827  *	  and reacquired afterwards.
828  *
829  * The process is put to sleep (TASK_INTERRUPTIBLE) until the
830  * @condition evaluates to true or signal is received. The @condition is
831  * checked each time the waitqueue @wq is woken up.
832  *
833  * wake_up() has to be called after changing any variable that could
834  * change the result of the wait condition.
835  *
836  * This is supposed to be called while holding the lock. The lock is
837  * dropped before going to sleep and is reacquired afterwards.
838  *
839  * The macro will return -ERESTARTSYS if it was interrupted by a signal
840  * and 0 if @condition evaluated to true.
841  */
842 #define wait_event_interruptible_lock_irq(wq, condition, lock)		\
843 ({									\
844 	int __ret = 0;							\
845 	if (!(condition))						\
846 		__ret = __wait_event_interruptible_lock_irq(wq,		\
847 						condition, lock,);	\
848 	__ret;								\
849 })
850 
851 #define __wait_event_interruptible_lock_irq_timeout(wq, condition,	\
852 						    lock, timeout)	\
853 	___wait_event(wq, ___wait_cond_timeout(condition),		\
854 		      TASK_INTERRUPTIBLE, 0, timeout,			\
855 		      spin_unlock_irq(&lock);				\
856 		      __ret = schedule_timeout(__ret);			\
857 		      spin_lock_irq(&lock));
858 
859 /**
860  * wait_event_interruptible_lock_irq_timeout - sleep until a condition gets
861  *		true or a timeout elapses. The condition is checked under
862  *		the lock. This is expected to be called with the lock taken.
863  * @wq: the waitqueue to wait on
864  * @condition: a C expression for the event to wait for
865  * @lock: a locked spinlock_t, which will be released before schedule()
866  *	  and reacquired afterwards.
867  * @timeout: timeout, in jiffies
868  *
869  * The process is put to sleep (TASK_INTERRUPTIBLE) until the
870  * @condition evaluates to true or signal is received. The @condition is
871  * checked each time the waitqueue @wq is woken up.
872  *
873  * wake_up() has to be called after changing any variable that could
874  * change the result of the wait condition.
875  *
876  * This is supposed to be called while holding the lock. The lock is
877  * dropped before going to sleep and is reacquired afterwards.
878  *
879  * The function returns 0 if the @timeout elapsed, -ERESTARTSYS if it
880  * was interrupted by a signal, and the remaining jiffies otherwise
881  * if the condition evaluated to true before the timeout elapsed.
882  */
883 #define wait_event_interruptible_lock_irq_timeout(wq, condition, lock,	\
884 						  timeout)		\
885 ({									\
886 	long __ret = timeout;						\
887 	if (!___wait_cond_timeout(condition))				\
888 		__ret = __wait_event_interruptible_lock_irq_timeout(	\
889 					wq, condition, lock, timeout);	\
890 	__ret;								\
891 })
892 
893 /*
894  * Waitqueues which are removed from the waitqueue_head at wakeup time
895  */
896 void prepare_to_wait(wait_queue_head_t *q, wait_queue_t *wait, int state);
897 void prepare_to_wait_exclusive(wait_queue_head_t *q, wait_queue_t *wait, int state);
898 long prepare_to_wait_event(wait_queue_head_t *q, wait_queue_t *wait, int state);
899 void finish_wait(wait_queue_head_t *q, wait_queue_t *wait);
900 void abort_exclusive_wait(wait_queue_head_t *q, wait_queue_t *wait, unsigned int mode, void *key);
901 long wait_woken(wait_queue_t *wait, unsigned mode, long timeout);
902 int woken_wake_function(wait_queue_t *wait, unsigned mode, int sync, void *key);
903 int autoremove_wake_function(wait_queue_t *wait, unsigned mode, int sync, void *key);
904 int wake_bit_function(wait_queue_t *wait, unsigned mode, int sync, void *key);
905 
906 #define DEFINE_WAIT_FUNC(name, function)				\
907 	wait_queue_t name = {						\
908 		.private	= current,				\
909 		.func		= function,				\
910 		.task_list	= LIST_HEAD_INIT((name).task_list),	\
911 	}
912 
913 #define DEFINE_WAIT(name) DEFINE_WAIT_FUNC(name, autoremove_wake_function)
914 
915 #define DEFINE_WAIT_BIT(name, word, bit)				\
916 	struct wait_bit_queue name = {					\
917 		.key = __WAIT_BIT_KEY_INITIALIZER(word, bit),		\
918 		.wait	= {						\
919 			.private	= current,			\
920 			.func		= wake_bit_function,		\
921 			.task_list	=				\
922 				LIST_HEAD_INIT((name).wait.task_list),	\
923 		},							\
924 	}
925 
926 #define init_wait(wait)							\
927 	do {								\
928 		(wait)->private = current;				\
929 		(wait)->func = autoremove_wake_function;		\
930 		INIT_LIST_HEAD(&(wait)->task_list);			\
931 		(wait)->flags = 0;					\
932 	} while (0)
933 
934 
935 extern int bit_wait(struct wait_bit_key *);
936 extern int bit_wait_io(struct wait_bit_key *);
937 extern int bit_wait_timeout(struct wait_bit_key *);
938 extern int bit_wait_io_timeout(struct wait_bit_key *);
939 
940 /**
941  * wait_on_bit - wait for a bit to be cleared
942  * @word: the word being waited on, a kernel virtual address
943  * @bit: the bit of the word being waited on
944  * @mode: the task state to sleep in
945  *
946  * There is a standard hashed waitqueue table for generic use. This
947  * is the part of the hashtable's accessor API that waits on a bit.
948  * For instance, if one were to have waiters on a bitflag, one would
949  * call wait_on_bit() in threads waiting for the bit to clear.
950  * One uses wait_on_bit() where one is waiting for the bit to clear,
951  * but has no intention of setting it.
952  * Returned value will be zero if the bit was cleared, or non-zero
953  * if the process received a signal and the mode permitted wakeup
954  * on that signal.
955  */
956 static inline int
957 wait_on_bit(void *word, int bit, unsigned mode)
958 {
959 	might_sleep();
960 	if (!test_bit(bit, word))
961 		return 0;
962 	return out_of_line_wait_on_bit(word, bit,
963 				       bit_wait,
964 				       mode);
965 }
966 
967 /**
968  * wait_on_bit_io - wait for a bit to be cleared
969  * @word: the word being waited on, a kernel virtual address
970  * @bit: the bit of the word being waited on
971  * @mode: the task state to sleep in
972  *
973  * Use the standard hashed waitqueue table to wait for a bit
974  * to be cleared.  This is similar to wait_on_bit(), but calls
975  * io_schedule() instead of schedule() for the actual waiting.
976  *
977  * Returned value will be zero if the bit was cleared, or non-zero
978  * if the process received a signal and the mode permitted wakeup
979  * on that signal.
980  */
981 static inline int
982 wait_on_bit_io(void *word, int bit, unsigned mode)
983 {
984 	might_sleep();
985 	if (!test_bit(bit, word))
986 		return 0;
987 	return out_of_line_wait_on_bit(word, bit,
988 				       bit_wait_io,
989 				       mode);
990 }
991 
992 /**
993  * wait_on_bit_timeout - wait for a bit to be cleared or a timeout elapses
994  * @word: the word being waited on, a kernel virtual address
995  * @bit: the bit of the word being waited on
996  * @mode: the task state to sleep in
997  * @timeout: timeout, in jiffies
998  *
999  * Use the standard hashed waitqueue table to wait for a bit
1000  * to be cleared. This is similar to wait_on_bit(), except also takes a
1001  * timeout parameter.
1002  *
1003  * Returned value will be zero if the bit was cleared before the
1004  * @timeout elapsed, or non-zero if the @timeout elapsed or process
1005  * received a signal and the mode permitted wakeup on that signal.
1006  */
1007 static inline int
1008 wait_on_bit_timeout(void *word, int bit, unsigned mode, unsigned long timeout)
1009 {
1010 	might_sleep();
1011 	if (!test_bit(bit, word))
1012 		return 0;
1013 	return out_of_line_wait_on_bit_timeout(word, bit,
1014 					       bit_wait_timeout,
1015 					       mode, timeout);
1016 }
1017 
1018 /**
1019  * wait_on_bit_action - wait for a bit to be cleared
1020  * @word: the word being waited on, a kernel virtual address
1021  * @bit: the bit of the word being waited on
1022  * @action: the function used to sleep, which may take special actions
1023  * @mode: the task state to sleep in
1024  *
1025  * Use the standard hashed waitqueue table to wait for a bit
1026  * to be cleared, and allow the waiting action to be specified.
1027  * This is like wait_on_bit() but allows fine control of how the waiting
1028  * is done.
1029  *
1030  * Returned value will be zero if the bit was cleared, or non-zero
1031  * if the process received a signal and the mode permitted wakeup
1032  * on that signal.
1033  */
1034 static inline int
1035 wait_on_bit_action(void *word, int bit, wait_bit_action_f *action, unsigned mode)
1036 {
1037 	might_sleep();
1038 	if (!test_bit(bit, word))
1039 		return 0;
1040 	return out_of_line_wait_on_bit(word, bit, action, mode);
1041 }
1042 
1043 /**
1044  * wait_on_bit_lock - wait for a bit to be cleared, when wanting to set it
1045  * @word: the word being waited on, a kernel virtual address
1046  * @bit: the bit of the word being waited on
1047  * @mode: the task state to sleep in
1048  *
1049  * There is a standard hashed waitqueue table for generic use. This
1050  * is the part of the hashtable's accessor API that waits on a bit
1051  * when one intends to set it, for instance, trying to lock bitflags.
1052  * For instance, if one were to have waiters trying to set bitflag
1053  * and waiting for it to clear before setting it, one would call
1054  * wait_on_bit() in threads waiting to be able to set the bit.
1055  * One uses wait_on_bit_lock() where one is waiting for the bit to
1056  * clear with the intention of setting it, and when done, clearing it.
1057  *
1058  * Returns zero if the bit was (eventually) found to be clear and was
1059  * set.  Returns non-zero if a signal was delivered to the process and
1060  * the @mode allows that signal to wake the process.
1061  */
1062 static inline int
1063 wait_on_bit_lock(void *word, int bit, unsigned mode)
1064 {
1065 	might_sleep();
1066 	if (!test_and_set_bit(bit, word))
1067 		return 0;
1068 	return out_of_line_wait_on_bit_lock(word, bit, bit_wait, mode);
1069 }
1070 
1071 /**
1072  * wait_on_bit_lock_io - wait for a bit to be cleared, when wanting to set it
1073  * @word: the word being waited on, a kernel virtual address
1074  * @bit: the bit of the word being waited on
1075  * @mode: the task state to sleep in
1076  *
1077  * Use the standard hashed waitqueue table to wait for a bit
1078  * to be cleared and then to atomically set it.  This is similar
1079  * to wait_on_bit(), but calls io_schedule() instead of schedule()
1080  * for the actual waiting.
1081  *
1082  * Returns zero if the bit was (eventually) found to be clear and was
1083  * set.  Returns non-zero if a signal was delivered to the process and
1084  * the @mode allows that signal to wake the process.
1085  */
1086 static inline int
1087 wait_on_bit_lock_io(void *word, int bit, unsigned mode)
1088 {
1089 	might_sleep();
1090 	if (!test_and_set_bit(bit, word))
1091 		return 0;
1092 	return out_of_line_wait_on_bit_lock(word, bit, bit_wait_io, mode);
1093 }
1094 
1095 /**
1096  * wait_on_bit_lock_action - wait for a bit to be cleared, when wanting to set it
1097  * @word: the word being waited on, a kernel virtual address
1098  * @bit: the bit of the word being waited on
1099  * @action: the function used to sleep, which may take special actions
1100  * @mode: the task state to sleep in
1101  *
1102  * Use the standard hashed waitqueue table to wait for a bit
1103  * to be cleared and then to set it, and allow the waiting action
1104  * to be specified.
1105  * This is like wait_on_bit() but allows fine control of how the waiting
1106  * is done.
1107  *
1108  * Returns zero if the bit was (eventually) found to be clear and was
1109  * set.  Returns non-zero if a signal was delivered to the process and
1110  * the @mode allows that signal to wake the process.
1111  */
1112 static inline int
1113 wait_on_bit_lock_action(void *word, int bit, wait_bit_action_f *action, unsigned mode)
1114 {
1115 	might_sleep();
1116 	if (!test_and_set_bit(bit, word))
1117 		return 0;
1118 	return out_of_line_wait_on_bit_lock(word, bit, action, mode);
1119 }
1120 
1121 /**
1122  * wait_on_atomic_t - Wait for an atomic_t to become 0
1123  * @val: The atomic value being waited on, a kernel virtual address
1124  * @action: the function used to sleep, which may take special actions
1125  * @mode: the task state to sleep in
1126  *
1127  * Wait for an atomic_t to become 0.  We abuse the bit-wait waitqueue table for
1128  * the purpose of getting a waitqueue, but we set the key to a bit number
1129  * outside of the target 'word'.
1130  */
1131 static inline
1132 int wait_on_atomic_t(atomic_t *val, int (*action)(atomic_t *), unsigned mode)
1133 {
1134 	might_sleep();
1135 	if (atomic_read(val) == 0)
1136 		return 0;
1137 	return out_of_line_wait_on_atomic_t(val, action, mode);
1138 }
1139 
1140 #endif /* _LINUX_WAIT_H */
1141