xref: /xnu-8020.140.41/osfmk/kern/sched_prim.h (revision 27b03b360a988dfd3dfdf34262bb0042026747cc) !
1 /*
2  * Copyright (c) 2000-2012 Apple Inc. All rights reserved.
3  *
4  * @APPLE_OSREFERENCE_LICENSE_HEADER_START@
5  *
6  * This file contains Original Code and/or Modifications of Original Code
7  * as defined in and that are subject to the Apple Public Source License
8  * Version 2.0 (the 'License'). You may not use this file except in
9  * compliance with the License. The rights granted to you under the License
10  * may not be used to create, or enable the creation or redistribution of,
11  * unlawful or unlicensed copies of an Apple operating system, or to
12  * circumvent, violate, or enable the circumvention or violation of, any
13  * terms of an Apple operating system software license agreement.
14  *
15  * Please obtain a copy of the License at
16  * http://www.opensource.apple.com/apsl/ and read it before using this file.
17  *
18  * The Original Code and all software distributed under the License are
19  * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER
20  * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
21  * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
22  * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT.
23  * Please see the License for the specific language governing rights and
24  * limitations under the License.
25  *
26  * @APPLE_OSREFERENCE_LICENSE_HEADER_END@
27  */
28 /*
29  * @OSF_COPYRIGHT@
30  */
31 /*
32  * Mach Operating System
33  * Copyright (c) 1991,1990,1989,1988,1987 Carnegie Mellon University
34  * All Rights Reserved.
35  *
36  * Permission to use, copy, modify and distribute this software and its
37  * documentation is hereby granted, provided that both the copyright
38  * notice and this permission notice appear in all copies of the
39  * software, derivative works or modified versions, and any portions
40  * thereof, and that both notices appear in supporting documentation.
41  *
42  * CARNEGIE MELLON ALLOWS FREE USE OF THIS SOFTWARE IN ITS "AS IS"
43  * CONDITION.  CARNEGIE MELLON DISCLAIMS ANY LIABILITY OF ANY KIND FOR
44  * ANY DAMAGES WHATSOEVER RESULTING FROM THE USE OF THIS SOFTWARE.
45  *
46  * Carnegie Mellon requests users of this software to return to
47  *
48  *  Software Distribution Coordinator  or  [email protected]
49  *  School of Computer Science
50  *  Carnegie Mellon University
51  *  Pittsburgh PA 15213-3890
52  *
53  * any improvements or extensions that they make and grant Carnegie Mellon
54  * the rights to redistribute these changes.
55  */
56 /*
57  */
58 /*
59  *	File:	sched_prim.h
60  *	Author:	David Golub
61  *
62  *	Scheduling primitive definitions file
63  *
64  */
65 
66 #ifndef _KERN_SCHED_PRIM_H_
67 #define _KERN_SCHED_PRIM_H_
68 
69 #include <sys/cdefs.h>
70 #include <mach/boolean.h>
71 #include <mach/machine/vm_types.h>
72 #include <mach/kern_return.h>
73 #include <kern/clock.h>
74 #include <kern/kern_types.h>
75 #include <kern/percpu.h>
76 #include <kern/thread.h>
77 #include <kern/block_hint.h>
78 
79 extern int              thread_get_current_cpuid(void);
80 
81 #ifdef  MACH_KERNEL_PRIVATE
82 
83 #include <kern/sched_urgency.h>
84 #include <kern/thread_group.h>
85 #include <kern/waitq.h>
86 
87 /* Initialization */
88 extern void             sched_init(void);
89 
90 extern void             sched_startup(void);
91 
92 extern void             sched_timebase_init(void);
93 
94 extern void             pset_rt_init(processor_set_t pset);
95 
96 extern void             sched_rtlocal_init(processor_set_t pset);
97 
98 extern rt_queue_t       sched_rtlocal_runq(processor_set_t pset);
99 
100 extern void             sched_rtlocal_queue_shutdown(processor_t processor);
101 
102 extern int64_t          sched_rtlocal_runq_count_sum(void);
103 
104 extern thread_t         sched_rtlocal_steal_thread(processor_set_t stealing_pset, uint64_t earliest_deadline);
105 
106 extern thread_t         sched_rt_choose_thread(processor_set_t pset);
107 
108 extern void             sched_check_spill(processor_set_t pset, thread_t thread);
109 
110 extern bool             sched_thread_should_yield(processor_t processor, thread_t thread);
111 
112 extern bool             sched_steal_thread_DISABLED(processor_set_t pset);
113 extern bool             sched_steal_thread_enabled(processor_set_t pset);
114 
115 /* Force a preemption point for a thread and wait for it to stop running */
116 extern boolean_t        thread_stop(
117 	thread_t        thread,
118 	boolean_t       until_not_runnable);
119 
120 /* Release a previous stop request */
121 extern void                     thread_unstop(
122 	thread_t        thread);
123 
124 /* Wait for a thread to stop running */
125 extern void                     thread_wait(
126 	thread_t        thread,
127 	boolean_t       until_not_runnable);
128 
129 /* Unblock thread on wake up */
130 extern boolean_t        thread_unblock(
131 	thread_t                thread,
132 	wait_result_t   wresult);
133 
134 /* Unblock and dispatch thread */
135 extern kern_return_t    thread_go(
136 	thread_t                thread,
137 	wait_result_t   wresult,
138 	waitq_options_t option);
139 
140 /* Check if direct handoff is allowed */
141 extern boolean_t
142 thread_allowed_for_handoff(
143 	thread_t         thread);
144 
145 /* Handle threads at context switch */
146 extern void                     thread_dispatch(
147 	thread_t                old_thread,
148 	thread_t                new_thread);
149 
150 /* Switch directly to a particular thread */
151 extern int                      thread_run(
152 	thread_t                        self,
153 	thread_continue_t       continuation,
154 	void                            *parameter,
155 	thread_t                        new_thread);
156 
157 /* Resume thread with new stack */
158 extern __dead2 void     thread_continue(thread_t old_thread);
159 
160 /* Invoke continuation */
161 extern __dead2 void     call_continuation(
162 	thread_continue_t       continuation,
163 	void                    *parameter,
164 	wait_result_t           wresult,
165 	boolean_t               enable_interrupts);
166 
167 /*
168  * Flags that can be passed to set_sched_pri
169  * to skip side effects
170  */
171 __options_decl(set_sched_pri_options_t, uint32_t, {
172 	SETPRI_DEFAULT  = 0x0,
173 	SETPRI_LAZY     = 0x1,  /* Avoid setting AST flags or sending IPIs */
174 });
175 
176 /* Set the current scheduled priority */
177 extern void set_sched_pri(
178 	thread_t      thread,
179 	int16_t       priority,
180 	set_sched_pri_options_t options);
181 
182 /* Set base priority of the specified thread */
183 extern void             sched_set_thread_base_priority(
184 	thread_t                thread,
185 	int                             priority);
186 
187 /* Set absolute base priority of the specified thread */
188 extern void             sched_set_kernel_thread_priority(
189 	thread_t                thread,
190 	int                             priority);
191 
192 
193 /* Set the thread's true scheduling mode */
194 extern void             sched_set_thread_mode(thread_t thread,
195     sched_mode_t mode);
196 /* Demote the true scheduler mode */
197 extern void             sched_thread_mode_demote(thread_t thread,
198     uint32_t reason);
199 /* Un-demote the true scheduler mode */
200 extern void             sched_thread_mode_undemote(thread_t thread,
201     uint32_t reason);
202 
203 extern void sched_thread_promote_reason(thread_t thread, uint32_t reason, uintptr_t trace_obj);
204 extern void sched_thread_unpromote_reason(thread_t thread, uint32_t reason, uintptr_t trace_obj);
205 
206 /* Re-evaluate base priority of thread (thread locked) */
207 void thread_recompute_priority(thread_t thread);
208 
209 /* Re-evaluate scheduled priority of thread (thread locked) */
210 extern void thread_recompute_sched_pri(
211 	thread_t thread,
212 	set_sched_pri_options_t options);
213 
214 /* Periodic scheduler activity */
215 extern void             sched_init_thread(void);
216 
217 /* Perform sched_tick housekeeping activities */
218 extern boolean_t                can_update_priority(
219 	thread_t                thread);
220 
221 extern void             update_priority(
222 	thread_t                thread);
223 
224 extern void             lightweight_update_priority(
225 	thread_t                thread);
226 
227 extern void             sched_default_quantum_expire(thread_t thread);
228 
229 /* Idle processor thread continuation */
230 extern void             idle_thread(
231 	void*           parameter,
232 	wait_result_t   result);
233 
234 extern kern_return_t    idle_thread_create(
235 	processor_t             processor);
236 
237 /* Continuation return from syscall */
238 extern void     thread_syscall_return(
239 	kern_return_t   ret);
240 
241 /* Context switch */
242 extern wait_result_t    thread_block_reason(
243 	thread_continue_t       continuation,
244 	void                            *parameter,
245 	ast_t                           reason);
246 
247 __options_decl(sched_options_t, uint32_t, {
248 	SCHED_NONE      = 0x0,
249 	SCHED_TAILQ     = 0x1,
250 	SCHED_HEADQ     = 0x2,
251 	SCHED_PREEMPT   = 0x4,
252 	SCHED_REBALANCE = 0x8,
253 });
254 
255 /* Reschedule thread for execution */
256 extern void             thread_setrun(
257 	thread_t        thread,
258 	sched_options_t options);
259 
260 extern processor_set_t  task_choose_pset(
261 	task_t                  task);
262 
263 /* Bind the current thread to a particular processor */
264 extern processor_t              thread_bind(
265 	processor_t             processor);
266 
267 extern bool pset_has_stealable_threads(
268 	processor_set_t         pset);
269 
270 extern bool pset_has_stealable_rt_threads(
271 	processor_set_t         pset);
272 
273 extern processor_set_t choose_starting_pset(
274 	pset_node_t  node,
275 	thread_t     thread,
276 	processor_t *processor_hint);
277 
278 extern int pset_available_cpu_count(
279 	processor_set_t pset);
280 
281 extern pset_node_t sched_choose_node(
282 	thread_t     thread);
283 
284 /* Choose the best processor to run a thread */
285 extern processor_t      choose_processor(
286 	processor_set_t                pset,
287 	processor_t                    processor,
288 	thread_t                       thread);
289 
290 extern void sched_SMT_balance(
291 	processor_t processor,
292 	processor_set_t pset);
293 
294 extern void thread_quantum_init(
295 	thread_t thread);
296 
297 extern void             run_queue_init(
298 	run_queue_t             runq);
299 
300 extern thread_t run_queue_dequeue(
301 	run_queue_t           runq,
302 	sched_options_t       options);
303 
304 extern boolean_t        run_queue_enqueue(
305 	run_queue_t           runq,
306 	thread_t              thread,
307 	sched_options_t       options);
308 
309 extern void     run_queue_remove(
310 	run_queue_t            runq,
311 	thread_t                       thread);
312 
313 extern thread_t run_queue_peek(
314 	run_queue_t            runq);
315 
316 struct sched_update_scan_context {
317 	uint64_t        earliest_bg_make_runnable_time;
318 	uint64_t        earliest_normal_make_runnable_time;
319 	uint64_t        earliest_rt_make_runnable_time;
320 	uint64_t        sched_tick_last_abstime;
321 };
322 typedef struct sched_update_scan_context *sched_update_scan_context_t;
323 
324 extern void             sched_rtlocal_runq_scan(sched_update_scan_context_t scan_context);
325 
326 extern void sched_pset_made_schedulable(
327 	processor_t processor,
328 	processor_set_t pset,
329 	boolean_t drop_lock);
330 
331 extern void sched_cpu_init_completed(void);
332 
333 /*
334  * Enum to define various events which need IPIs. The IPI policy
335  * engine decides what kind of IPI to use based on destination
336  * processor state, thread and one of the following scheduling events.
337  */
338 typedef enum {
339 	SCHED_IPI_EVENT_BOUND_THR   = 0x1,
340 	SCHED_IPI_EVENT_PREEMPT     = 0x2,
341 	SCHED_IPI_EVENT_SMT_REBAL   = 0x3,
342 	SCHED_IPI_EVENT_SPILL       = 0x4,
343 	SCHED_IPI_EVENT_REBALANCE   = 0x5,
344 	SCHED_IPI_EVENT_RT_PREEMPT  = 0x6,
345 } sched_ipi_event_t;
346 
347 
348 /* Enum to define various IPI types used by the scheduler */
349 typedef enum {
350 	SCHED_IPI_NONE              = 0x0,
351 	SCHED_IPI_IMMEDIATE         = 0x1,
352 	SCHED_IPI_IDLE              = 0x2,
353 	SCHED_IPI_DEFERRED          = 0x3,
354 } sched_ipi_type_t;
355 
356 /* The IPI policy engine behaves in the following manner:
357  * - All scheduler events which need an IPI invoke sched_ipi_action() with
358  *   the appropriate destination processor, thread and event.
359  * - sched_ipi_action() performs basic checks, invokes the scheduler specific
360  *   ipi_policy routine and sets pending_AST bits based on the result.
361  * - Once the pset lock is dropped, the scheduler invokes sched_ipi_perform()
362  *   routine which actually sends the appropriate IPI to the destination core.
363  */
364 extern sched_ipi_type_t sched_ipi_action(processor_t dst, thread_t thread, sched_ipi_event_t event);
365 extern void sched_ipi_perform(processor_t dst, sched_ipi_type_t ipi);
366 
367 /* sched_ipi_policy() is the global default IPI policy for all schedulers */
368 extern sched_ipi_type_t sched_ipi_policy(processor_t dst, thread_t thread,
369     boolean_t dst_idle, sched_ipi_event_t event);
370 
371 /* sched_ipi_deferred_policy() is the global default deferred IPI policy for all schedulers */
372 extern sched_ipi_type_t sched_ipi_deferred_policy(processor_set_t pset,
373     processor_t dst, thread_t thread, sched_ipi_event_t event);
374 
375 #if defined(CONFIG_SCHED_TIMESHARE_CORE)
376 
377 extern boolean_t        thread_update_add_thread(thread_t thread);
378 extern void             thread_update_process_threads(void);
379 extern boolean_t        runq_scan(run_queue_t runq, sched_update_scan_context_t scan_context);
380 
381 #if CONFIG_SCHED_CLUTCH
382 extern boolean_t        sched_clutch_timeshare_scan(queue_t thread_queue, uint16_t count, sched_update_scan_context_t scan_context);
383 #endif /* CONFIG_SCHED_CLUTCH */
384 
385 extern void sched_timeshare_init(void);
386 extern void sched_timeshare_timebase_init(void);
387 extern void sched_timeshare_maintenance_continue(void);
388 
389 extern boolean_t priority_is_urgent(int priority);
390 extern uint32_t sched_timeshare_initial_quantum_size(thread_t thread);
391 
392 extern int sched_compute_timeshare_priority(thread_t thread);
393 
394 #endif /* CONFIG_SCHED_TIMESHARE_CORE */
395 
396 /* Remove thread from its run queue */
397 extern boolean_t        thread_run_queue_remove(thread_t thread);
398 thread_t thread_run_queue_remove_for_handoff(thread_t thread);
399 
400 /* Put a thread back in the run queue after being yanked */
401 extern void thread_run_queue_reinsert(thread_t thread, sched_options_t options);
402 
403 extern void             thread_timer_expire(
404 	void                    *thread,
405 	void                    *p1);
406 
407 extern bool thread_is_eager_preempt(thread_t thread);
408 
409 extern boolean_t sched_generic_direct_dispatch_to_idle_processors;
410 
411 /* Set the maximum interrupt level for the thread */
412 __private_extern__ wait_interrupt_t thread_interrupt_level(
413 	wait_interrupt_t interruptible);
414 
415 __private_extern__ wait_result_t thread_mark_wait_locked(
416 	thread_t                 thread,
417 	wait_interrupt_t interruptible);
418 
419 /* Wake up locked thread directly, passing result */
420 __private_extern__ kern_return_t clear_wait_internal(
421 	thread_t                thread,
422 	wait_result_t   result);
423 
424 struct sched_statistics {
425 	uint32_t        csw_count;
426 	uint32_t        preempt_count;
427 	uint32_t        preempted_rt_count;
428 	uint32_t        preempted_by_rt_count;
429 	uint32_t        rt_sched_count;
430 	uint32_t        interrupt_count;
431 	uint32_t        ipi_count;
432 	uint32_t        timer_pop_count;
433 	uint32_t        idle_transitions;
434 	uint32_t        quantum_timer_expirations;
435 };
436 PERCPU_DECL(struct sched_statistics, sched_stats);
437 extern bool             sched_stats_active;
438 
439 extern void sched_stats_handle_csw(
440 	processor_t processor,
441 	int reasons,
442 	int selfpri,
443 	int otherpri);
444 
445 extern void sched_stats_handle_runq_change(
446 	struct runq_stats *stats,
447 	int old_count);
448 
449 #define SCHED_STATS_INC(field)                                                  \
450 MACRO_BEGIN                                                                     \
451 	if (__improbable(sched_stats_active)) {                                 \
452 	        PERCPU_GET(sched_stats)->field++;                               \
453 	}                                                                       \
454 MACRO_END
455 
456 #if DEBUG
457 
458 #define SCHED_STATS_CSW(processor, reasons, selfpri, otherpri)                  \
459 MACRO_BEGIN                                                                     \
460 	if (__improbable(sched_stats_active)) {                                 \
461 	        sched_stats_handle_csw((processor),                             \
462 	            (reasons), (selfpri), (otherpri));                          \
463 	}                                                                       \
464 MACRO_END
465 
466 
467 #define SCHED_STATS_RUNQ_CHANGE(stats, old_count)                               \
468 MACRO_BEGIN                                                                     \
469 	if (__improbable(sched_stats_active)) {                                 \
470 	        sched_stats_handle_runq_change((stats), (old_count));           \
471 	}                                                                       \
472 MACRO_END
473 
474 #else /* DEBUG */
475 
476 #define SCHED_STATS_CSW(processor, reasons, selfpri, otherpri) do { }while(0)
477 #define SCHED_STATS_RUNQ_CHANGE(stats, old_count) do { }while(0)
478 
479 #endif /* DEBUG */
480 
481 extern uint32_t sched_debug_flags;
482 #define SCHED_DEBUG_FLAG_PLATFORM_TRACEPOINTS           0x00000001
483 #define SCHED_DEBUG_FLAG_CHOOSE_PROCESSOR_TRACEPOINTS   0x00000002
484 
485 #define SCHED_DEBUG_PLATFORM_KERNEL_DEBUG_CONSTANT(...)                         \
486 MACRO_BEGIN                                                                     \
487 	if (__improbable(sched_debug_flags &                                    \
488 	    SCHED_DEBUG_FLAG_PLATFORM_TRACEPOINTS)) {                           \
489 	        KERNEL_DEBUG_CONSTANT(__VA_ARGS__);                             \
490 	}                                                                       \
491 MACRO_END
492 
493 #define SCHED_DEBUG_CHOOSE_PROCESSOR_KERNEL_DEBUG_CONSTANT(...)                 \
494 MACRO_BEGIN                                                                     \
495 	if (__improbable(sched_debug_flags &                                    \
496 	    SCHED_DEBUG_FLAG_CHOOSE_PROCESSOR_TRACEPOINTS)) {                   \
497 	        KERNEL_DEBUG_CONSTANT(__VA_ARGS__);                             \
498 	}                                                                       \
499 MACRO_END
500 
501 /* Tells if there are "active" RT threads in the system (provided by CPU PM) */
502 extern void     active_rt_threads(
503 	boolean_t       active);
504 
505 /* Returns the perfcontrol attribute for the thread */
506 extern perfcontrol_class_t thread_get_perfcontrol_class(
507 	thread_t        thread);
508 
509 /* Generic routine for Non-AMP schedulers to calculate parallelism */
510 extern uint32_t sched_qos_max_parallelism(int qos, uint64_t options);
511 
512 extern void check_monotonic_time(uint64_t ctime);
513 
514 #endif /* MACH_KERNEL_PRIVATE */
515 
516 __BEGIN_DECLS
517 
518 #ifdef  XNU_KERNEL_PRIVATE
519 
520 extern void thread_bind_cluster_type(thread_t, char cluster_type, bool soft_bind);
521 
522 __options_decl(thread_bind_option_t, uint64_t, {
523 	/* Unbind a previously cluster bound thread */
524 	THREAD_UNBIND                   = 0x1,
525 	/*
526 	 * Soft bind the thread to the cluster; soft binding means the thread will be
527 	 * moved to an available cluster if the bound cluster is de-recommended/offline.
528 	 */
529 	THREAD_BIND_SOFT                = 0x2,
530 	/*
531 	 * Bind thread to the cluster only if it is eligible to run on that cluster. If
532 	 * the thread is not eligible to run on the cluster, thread_bind_cluster_id()
533 	 * returns KERN_INVALID_POLICY.
534 	 */
535 	THREAD_BIND_ELIGIBLE_ONLY       = 0x4,
536 });
537 extern kern_return_t thread_bind_cluster_id(thread_t thread, uint32_t cluster_id, thread_bind_option_t options);
538 
539 extern int sched_get_rt_n_backup_processors(void);
540 extern void sched_set_rt_n_backup_processors(int n);
541 
542 extern int sched_get_rt_constraint_ll(void);
543 extern void sched_set_rt_constraint_ll(int new_constraint_us);
544 
545 extern int sched_get_rt_deadline_epsilon(void);
546 extern void sched_set_rt_deadline_epsilon(int new_epsilon_us);
547 
548 /* Toggles a global override to turn off CPU Throttling */
549 extern void     sys_override_cpu_throttle(boolean_t enable_override);
550 
551 /*
552  ****************** Only exported until BSD stops using ********************
553  */
554 
555 extern void                     thread_vm_bind_group_add(void);
556 
557 /* Wake up thread directly, passing result */
558 extern kern_return_t clear_wait(
559 	thread_t                thread,
560 	wait_result_t   result);
561 
562 /* Start thread running */
563 extern void             thread_bootstrap_return(void) __attribute__((noreturn));
564 
565 /* Return from exception (BSD-visible interface) */
566 extern void             thread_exception_return(void) __dead2;
567 
568 #define SCHED_STRING_MAX_LENGTH (48)
569 /* String declaring the name of the current scheduler */
570 extern char sched_string[SCHED_STRING_MAX_LENGTH];
571 
572 __options_decl(thread_handoff_option_t, uint32_t, {
573 	THREAD_HANDOFF_NONE          = 0,
574 	THREAD_HANDOFF_SETRUN_NEEDED = 0x1,
575 });
576 
577 /* Remove thread from its run queue */
578 thread_t thread_prepare_for_handoff(thread_t thread, thread_handoff_option_t option);
579 
580 /* Attempt to context switch to a specific runnable thread */
581 extern wait_result_t thread_handoff_deallocate(thread_t thread, thread_handoff_option_t option);
582 
583 __attribute__((nonnull(2)))
584 extern void thread_handoff_parameter(thread_t thread,
585     thread_continue_t continuation, void *parameter, thread_handoff_option_t) __dead2;
586 
587 extern struct waitq     *assert_wait_queue(event_t event);
588 
589 extern kern_return_t thread_wakeup_one_with_pri(event_t event, int priority);
590 
591 extern thread_t thread_wakeup_identify(event_t event, int priority);
592 
593 #endif  /* XNU_KERNEL_PRIVATE */
594 
595 #ifdef KERNEL_PRIVATE
596 /* Set pending block hint for a particular object before we go into a wait state */
597 extern void             thread_set_pending_block_hint(
598 	thread_t                        thread,
599 	block_hint_t                    block_hint);
600 
601 #define QOS_PARALLELISM_COUNT_LOGICAL   0x1
602 #define QOS_PARALLELISM_REALTIME        0x2
603 #define QOS_PARALLELISM_CLUSTER_SHARED_RESOURCE              0x4
604 
605 extern uint32_t qos_max_parallelism(int qos, uint64_t options);
606 #endif /* KERNEL_PRIVATE */
607 
608 #if XNU_KERNEL_PRIVATE
609 extern void             thread_yield_with_continuation(
610 	thread_continue_t       continuation,
611 	void                            *parameter) __dead2;
612 #endif
613 
614 /* Context switch */
615 extern wait_result_t    thread_block(
616 	thread_continue_t       continuation);
617 
618 extern wait_result_t    thread_block_parameter(
619 	thread_continue_t       continuation,
620 	void                            *parameter);
621 
622 /* Declare thread will wait on a particular event */
623 extern wait_result_t    assert_wait(
624 	event_t                         event,
625 	wait_interrupt_t        interruptible);
626 
627 /* Assert that the thread intends to wait with a timeout */
628 extern wait_result_t    assert_wait_timeout(
629 	event_t                         event,
630 	wait_interrupt_t        interruptible,
631 	uint32_t                        interval,
632 	uint32_t                        scale_factor);
633 
634 /* Assert that the thread intends to wait with an urgency, timeout and leeway */
635 extern wait_result_t    assert_wait_timeout_with_leeway(
636 	event_t                         event,
637 	wait_interrupt_t        interruptible,
638 	wait_timeout_urgency_t  urgency,
639 	uint32_t                        interval,
640 	uint32_t                        leeway,
641 	uint32_t                        scale_factor);
642 
643 extern wait_result_t    assert_wait_deadline(
644 	event_t                         event,
645 	wait_interrupt_t        interruptible,
646 	uint64_t                        deadline);
647 
648 /* Assert that the thread intends to wait with an urgency, deadline, and leeway */
649 extern wait_result_t    assert_wait_deadline_with_leeway(
650 	event_t                         event,
651 	wait_interrupt_t        interruptible,
652 	wait_timeout_urgency_t  urgency,
653 	uint64_t                        deadline,
654 	uint64_t                        leeway);
655 
656 /* Wake up thread (or threads) waiting on a particular event */
657 extern kern_return_t    thread_wakeup_prim(
658 	event_t                         event,
659 	boolean_t                       one_thread,
660 	wait_result_t                   result);
661 
662 #define thread_wakeup(x)                                        \
663 	                thread_wakeup_prim((x), FALSE, THREAD_AWAKENED)
664 #define thread_wakeup_with_result(x, z)         \
665 	                thread_wakeup_prim((x), FALSE, (z))
666 #define thread_wakeup_one(x)                            \
667 	                thread_wakeup_prim((x), TRUE, THREAD_AWAKENED)
668 
669 /* Wakeup the specified thread if it is waiting on this event */
670 extern kern_return_t thread_wakeup_thread(event_t event, thread_t thread);
671 
672 extern boolean_t preemption_enabled(void);
673 
674 #ifdef MACH_KERNEL_PRIVATE
675 
676 /*
677  * Scheduler algorithm indirection. If only one algorithm is
678  * enabled at compile-time, a direction function call is used.
679  * If more than one is enabled, calls are dispatched through
680  * a function pointer table.
681  */
682 
683 #if   !defined(CONFIG_SCHED_TRADITIONAL) && !defined(CONFIG_SCHED_PROTO) && !defined(CONFIG_SCHED_GRRR) && !defined(CONFIG_SCHED_MULTIQ) && !defined(CONFIG_SCHED_CLUTCH) && !defined(CONFIG_SCHED_EDGE)
684 #error Enable at least one scheduler algorithm in osfmk/conf/MASTER.XXX
685 #endif
686 
687 #if __AMP__
688 
689 #if CONFIG_SCHED_EDGE
690 extern const struct sched_dispatch_table sched_edge_dispatch;
691 #define SCHED(f) (sched_edge_dispatch.f)
692 #else /* CONFIG_SCHED_EDGE */
693 extern const struct sched_dispatch_table sched_amp_dispatch;
694 #define SCHED(f) (sched_amp_dispatch.f)
695 #endif /* CONFIG_SCHED_EDGE */
696 
697 #else /* __AMP__ */
698 
699 #if CONFIG_SCHED_CLUTCH
700 extern const struct sched_dispatch_table sched_clutch_dispatch;
701 #define SCHED(f) (sched_clutch_dispatch.f)
702 #else /* CONFIG_SCHED_CLUTCH */
703 extern const struct sched_dispatch_table sched_dualq_dispatch;
704 #define SCHED(f) (sched_dualq_dispatch.f)
705 #endif /* CONFIG_SCHED_CLUTCH */
706 
707 #endif /* __AMP__ */
708 
709 struct sched_dispatch_table {
710 	const char *sched_name;
711 	void    (*init)(void);                          /* Init global state */
712 	void    (*timebase_init)(void);         /* Timebase-dependent initialization */
713 	void    (*processor_init)(processor_t processor);       /* Per-processor scheduler init */
714 	void    (*pset_init)(processor_set_t pset);     /* Per-processor set scheduler init */
715 
716 	void    (*maintenance_continuation)(void);      /* Function called regularly */
717 
718 	/*
719 	 * Choose a thread of greater or equal priority from the per-processor
720 	 * runqueue for timeshare/fixed threads
721 	 */
722 	thread_t        (*choose_thread)(
723 		processor_t           processor,
724 		int                           priority,
725 		ast_t reason);
726 
727 	/* True if scheduler supports stealing threads for this pset */
728 	bool    (*steal_thread_enabled)(processor_set_t pset);
729 
730 	/*
731 	 * Steal a thread from another processor in the pset so that it can run
732 	 * immediately
733 	 */
734 	thread_t        (*steal_thread)(
735 		processor_set_t         pset);
736 
737 	/*
738 	 * Compute priority for a timeshare thread based on base priority.
739 	 */
740 	int (*compute_timeshare_priority)(thread_t thread);
741 
742 	/*
743 	 * Pick the best node for a thread to run on.
744 	 */
745 	pset_node_t (*choose_node)(
746 		thread_t                      thread);
747 
748 	/*
749 	 * Pick the best processor for a thread (any kind of thread) to run on.
750 	 */
751 	processor_t     (*choose_processor)(
752 		processor_set_t                pset,
753 		processor_t                    processor,
754 		thread_t                       thread);
755 	/*
756 	 * Enqueue a timeshare or fixed priority thread onto the per-processor
757 	 * runqueue
758 	 */
759 	boolean_t (*processor_enqueue)(
760 		processor_t                    processor,
761 		thread_t                       thread,
762 		sched_options_t                options);
763 
764 	/* Migrate threads away in preparation for processor shutdown */
765 	void (*processor_queue_shutdown)(
766 		processor_t                    processor);
767 
768 	/* Remove the specific thread from the per-processor runqueue */
769 	boolean_t       (*processor_queue_remove)(
770 		processor_t             processor,
771 		thread_t                thread);
772 
773 	/*
774 	 * Does the per-processor runqueue have any timeshare or fixed priority
775 	 * threads on it? Called without pset lock held, so should
776 	 * not assume immutability while executing.
777 	 */
778 	boolean_t       (*processor_queue_empty)(processor_t            processor);
779 
780 	/*
781 	 * Would this priority trigger an urgent preemption if it's sitting
782 	 * on the per-processor runqueue?
783 	 */
784 	boolean_t       (*priority_is_urgent)(int priority);
785 
786 	/*
787 	 * Does the per-processor runqueue contain runnable threads that
788 	 * should cause the currently-running thread to be preempted?
789 	 */
790 	ast_t           (*processor_csw_check)(processor_t processor);
791 
792 	/*
793 	 * Does the per-processor runqueue contain a runnable thread
794 	 * of > or >= priority, as a preflight for choose_thread() or other
795 	 * thread selection
796 	 */
797 	boolean_t       (*processor_queue_has_priority)(processor_t             processor,
798 	    int                             priority,
799 	    boolean_t               gte);
800 
801 	/* Quantum size for the specified non-realtime thread. */
802 	uint32_t        (*initial_quantum_size)(thread_t thread);
803 
804 	/* Scheduler mode for a new thread */
805 	sched_mode_t    (*initial_thread_sched_mode)(task_t parent_task);
806 
807 	/*
808 	 * Is it safe to call update_priority, which may change a thread's
809 	 * runqueue or other state. This can be used to throttle changes
810 	 * to dynamic priority.
811 	 */
812 	boolean_t       (*can_update_priority)(thread_t thread);
813 
814 	/*
815 	 * Update both scheduled priority and other persistent state.
816 	 * Side effects may including migration to another processor's runqueue.
817 	 */
818 	void            (*update_priority)(thread_t thread);
819 
820 	/* Lower overhead update to scheduled priority and state. */
821 	void            (*lightweight_update_priority)(thread_t thread);
822 
823 	/* Callback for non-realtime threads when the quantum timer fires */
824 	void            (*quantum_expire)(thread_t thread);
825 
826 	/*
827 	 * Runnable threads on per-processor runqueue. Should only
828 	 * be used for relative comparisons of load between processors.
829 	 */
830 	int                     (*processor_runq_count)(processor_t     processor);
831 
832 	/* Aggregate runcount statistics for per-processor runqueue */
833 	uint64_t    (*processor_runq_stats_count_sum)(processor_t   processor);
834 
835 	boolean_t       (*processor_bound_count)(processor_t processor);
836 
837 	void            (*thread_update_scan)(sched_update_scan_context_t scan_context);
838 
839 	/* Supports more than one pset */
840 	boolean_t   multiple_psets_enabled;
841 	/* Supports scheduler groups */
842 	boolean_t   sched_groups_enabled;
843 
844 	/* Supports avoid-processor */
845 	boolean_t   avoid_processor_enabled;
846 
847 	/* Returns true if this processor should avoid running this thread. */
848 	bool    (*thread_avoid_processor)(processor_t processor, thread_t thread);
849 
850 	/*
851 	 * Invoked when a processor is about to choose the idle thread
852 	 * Used to send IPIs to a processor which would be preferred to be idle instead.
853 	 * Called with pset lock held, returns pset lock unlocked.
854 	 */
855 	void    (*processor_balance)(processor_t processor, processor_set_t pset);
856 	rt_queue_t      (*rt_runq)(processor_set_t pset);
857 	void    (*rt_init)(processor_set_t pset);
858 	void    (*rt_queue_shutdown)(processor_t processor);
859 	void    (*rt_runq_scan)(sched_update_scan_context_t scan_context);
860 	int64_t (*rt_runq_count_sum)(void);
861 	thread_t (*rt_steal_thread)(processor_set_t pset, uint64_t earliest_deadline);
862 
863 	uint32_t (*qos_max_parallelism)(int qos, uint64_t options);
864 	void    (*check_spill)(processor_set_t pset, thread_t thread);
865 	sched_ipi_type_t (*ipi_policy)(processor_t dst, thread_t thread, boolean_t dst_idle, sched_ipi_event_t event);
866 	bool    (*thread_should_yield)(processor_t processor, thread_t thread);
867 
868 	/* Routine to update run counts */
869 	uint32_t (*run_count_incr)(thread_t thread);
870 	uint32_t (*run_count_decr)(thread_t thread);
871 
872 	/* Routine to update scheduling bucket for a thread */
873 	void (*update_thread_bucket)(thread_t thread);
874 
875 	/* Routine to inform the scheduler when a new pset becomes schedulable */
876 	void (*pset_made_schedulable)(processor_t processor, processor_set_t pset, boolean_t drop_lock);
877 #if CONFIG_THREAD_GROUPS
878 	/* Routine to inform the scheduler when CLPC changes a thread group recommendation */
879 	void (*thread_group_recommendation_change)(struct thread_group *tg, cluster_type_t new_recommendation);
880 #endif
881 	/* Routine to inform the scheduler when all CPUs have finished initializing */
882 	void (*cpu_init_completed)(void);
883 	/* Routine to check if a thread is eligible to execute on a specific pset */
884 	bool (*thread_eligible_for_pset)(thread_t thread, processor_set_t pset);
885 };
886 
887 #if defined(CONFIG_SCHED_TRADITIONAL)
888 extern const struct sched_dispatch_table sched_traditional_dispatch;
889 extern const struct sched_dispatch_table sched_traditional_with_pset_runqueue_dispatch;
890 #endif
891 
892 #if defined(CONFIG_SCHED_MULTIQ)
893 extern const struct sched_dispatch_table sched_multiq_dispatch;
894 extern const struct sched_dispatch_table sched_dualq_dispatch;
895 #if __AMP__
896 extern const struct sched_dispatch_table sched_amp_dispatch;
897 #endif
898 #endif
899 
900 #if defined(CONFIG_SCHED_PROTO)
901 extern const struct sched_dispatch_table sched_proto_dispatch;
902 #endif
903 
904 #if defined(CONFIG_SCHED_GRRR)
905 extern const struct sched_dispatch_table sched_grrr_dispatch;
906 #endif
907 
908 #if defined(CONFIG_SCHED_CLUTCH)
909 extern const struct sched_dispatch_table sched_clutch_dispatch;
910 #endif
911 
912 #if defined(CONFIG_SCHED_EDGE)
913 extern const struct sched_dispatch_table sched_edge_dispatch;
914 #endif
915 
916 
917 #endif  /* MACH_KERNEL_PRIVATE */
918 
919 __END_DECLS
920 
921 #endif  /* _KERN_SCHED_PRIM_H_ */
922