osfmk/kern/sched_prim.h

   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  Software.Distribution@CS.CMU.EDU
  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 <mach/boolean.h>
  70 #include <mach/machine/vm_types.h>
  71 #include <mach/kern_return.h>
  72 #include <kern/clock.h>
  73 #include <kern/kern_types.h>
  74 #include <kern/thread.h>
  75 #include <sys/cdefs.h>
  76 #include <kern/block_hint.h>
  77
  78 #ifdef  MACH_KERNEL_PRIVATE
  79
  80 #include <kern/sched_urgency.h>
  81
  82 /* Initialization */
  83 extern void             sched_init(void);
  84
  85 extern void             sched_startup(void);
  86
  87 extern void             sched_timebase_init(void);
  88
  89 extern void             pset_rt_init(processor_set_t pset);
  90
  91 extern void             sched_rtglobal_init(processor_set_t pset);
  92
  93 extern rt_queue_t       sched_rtglobal_runq(processor_set_t pset);
  94
  95 extern void             sched_rtglobal_queue_shutdown(processor_t processor);
  96
  97 extern int64_t          sched_rtglobal_runq_count_sum(void);
  98
  99 extern void             sched_check_spill(processor_set_t pset, thread_t thread);
 100
 101 extern bool             sched_thread_should_yield(processor_t processor, thread_t thread);
 102
 103 extern bool             sched_steal_thread_DISABLED(processor_set_t pset);
 104 extern bool             sched_steal_thread_enabled(processor_set_t pset);
 105
 106 /* Force a preemption point for a thread and wait for it to stop running */
 107 extern boolean_t        thread_stop(
 108         thread_t        thread,
 109         boolean_t       until_not_runnable);
 110
 111 /* Release a previous stop request */
 112 extern void                     thread_unstop(
 113         thread_t        thread);
 114
 115 /* Wait for a thread to stop running */
 116 extern void                     thread_wait(
 117         thread_t        thread,
 118         boolean_t       until_not_runnable);
 119
 120 /* Unblock thread on wake up */
 121 extern boolean_t        thread_unblock(
 122         thread_t                thread,
 123         wait_result_t   wresult);
 124
 125 /* Unblock and dispatch thread */
 126 extern kern_return_t    thread_go(
 127         thread_t                thread,
 128         wait_result_t   wresult);
 129
 130 /* Handle threads at context switch */
 131 extern void                     thread_dispatch(
 132         thread_t                old_thread,
 133         thread_t                new_thread);
 134
 135 /* Switch directly to a particular thread */
 136 extern int                      thread_run(
 137         thread_t                        self,
 138         thread_continue_t       continuation,
 139         void                            *parameter,
 140         thread_t                        new_thread);
 141
 142 /* Resume thread with new stack */
 143 extern void                     thread_continue(
 144         thread_t                old_thread);
 145
 146 /* Invoke continuation */
 147 extern void             call_continuation(
 148         thread_continue_t       continuation,
 149         void                            *parameter,
 150         wait_result_t           wresult,
 151         boolean_t           enable_interrupts);
 152
 153 /*
 154  * Flags that can be passed to set_sched_pri
 155  * to skip side effects
 156  */
 157 typedef enum {
 158         SETPRI_DEFAULT  = 0x0,
 159         SETPRI_LAZY     = 0x1,  /* Avoid setting AST flags or sending IPIs */
 160 } set_sched_pri_options_t;
 161
 162 /* Set the current scheduled priority */
 163 extern void set_sched_pri(
 164         thread_t      thread,
 165         int           priority,
 166         set_sched_pri_options_t options);
 167
 168 /* Set base priority of the specified thread */
 169 extern void             sched_set_thread_base_priority(
 170         thread_t                thread,
 171         int                             priority);
 172
 173 /* Set the thread's true scheduling mode */
 174 extern void             sched_set_thread_mode(thread_t thread,
 175     sched_mode_t mode);
 176 /* Demote the true scheduler mode */
 177 extern void             sched_thread_mode_demote(thread_t thread,
 178     uint32_t reason);
 179 /* Un-demote the true scheduler mode */
 180 extern void             sched_thread_mode_undemote(thread_t thread,
 181     uint32_t reason);
 182
 183 extern void sched_thread_promote_to_pri(thread_t thread, int priority, uintptr_t trace_obj);
 184 extern void sched_thread_update_promotion_to_pri(thread_t thread, int priority, uintptr_t trace_obj);
 185 extern void sched_thread_unpromote(thread_t thread, uintptr_t trace_obj);
 186
 187 extern void assert_promotions_invariant(thread_t thread);
 188
 189 extern void sched_thread_promote_reason(thread_t thread, uint32_t reason, uintptr_t trace_obj);
 190 extern void sched_thread_unpromote_reason(thread_t thread, uint32_t reason, uintptr_t trace_obj);
 191
 192 /* Re-evaluate base priority of thread (thread locked) */
 193 void thread_recompute_priority(thread_t thread);
 194
 195 /* Re-evaluate scheduled priority of thread (thread locked) */
 196 extern void thread_recompute_sched_pri(
 197         thread_t thread,
 198         set_sched_pri_options_t options);
 199
 200 /* Periodic scheduler activity */
 201 extern void             sched_init_thread(void (*)(void));
 202
 203 /* Perform sched_tick housekeeping activities */
 204 extern boolean_t                can_update_priority(
 205         thread_t                thread);
 206
 207 extern void             update_priority(
 208         thread_t                thread);
 209
 210 extern void             lightweight_update_priority(
 211         thread_t                thread);
 212
 213 extern void             sched_default_quantum_expire(thread_t thread);
 214
 215 /* Idle processor thread */
 216 extern void             idle_thread(void);
 217
 218 extern kern_return_t    idle_thread_create(
 219         processor_t             processor);
 220
 221 /* Continuation return from syscall */
 222 extern void     thread_syscall_return(
 223         kern_return_t   ret);
 224
 225 /* Context switch */
 226 extern wait_result_t    thread_block_reason(
 227         thread_continue_t       continuation,
 228         void                            *parameter,
 229         ast_t                           reason);
 230
 231 /* Reschedule thread for execution */
 232 extern void             thread_setrun(
 233         thread_t        thread,
 234         integer_t       options);
 235
 236 typedef enum {
 237         SCHED_NONE      = 0x0,
 238         SCHED_TAILQ     = 0x1,
 239         SCHED_HEADQ     = 0x2,
 240         SCHED_PREEMPT   = 0x4,
 241         SCHED_REBALANCE = 0x8,
 242         SCHED_PEEK      = 0x10,
 243 } sched_options_t;
 244
 245 extern processor_set_t  task_choose_pset(
 246         task_t                  task);
 247
 248 /* Bind the current thread to a particular processor */
 249 extern processor_t              thread_bind(
 250         processor_t             processor);
 251
 252 /* Choose the best processor to run a thread */
 253 extern processor_t      choose_processor(
 254         processor_set_t                pset,
 255         processor_t                    processor,
 256         thread_t                       thread);
 257
 258 extern void sched_SMT_balance(
 259         processor_t processor,
 260         processor_set_t pset);
 261
 262 extern void thread_quantum_init(
 263         thread_t thread);
 264
 265 extern void             run_queue_init(
 266         run_queue_t             runq);
 267
 268 extern thread_t run_queue_dequeue(
 269         run_queue_t           runq,
 270         integer_t             options);
 271
 272 extern boolean_t        run_queue_enqueue(
 273         run_queue_t           runq,
 274         thread_t                      thread,
 275         integer_t             options);
 276
 277 extern void     run_queue_remove(
 278         run_queue_t            runq,
 279         thread_t                       thread);
 280
 281 struct sched_update_scan_context {
 282         uint64_t        earliest_bg_make_runnable_time;
 283         uint64_t        earliest_normal_make_runnable_time;
 284         uint64_t        earliest_rt_make_runnable_time;
 285 };
 286 typedef struct sched_update_scan_context *sched_update_scan_context_t;
 287
 288 extern void             sched_rtglobal_runq_scan(sched_update_scan_context_t scan_context);
 289
 290 /*
 291  * Enum to define various events which need IPIs. The IPI policy
 292  * engine decides what kind of IPI to use based on destination
 293  * processor state, thread and one of the following scheduling events.
 294  */
 295 typedef enum {
 296         SCHED_IPI_EVENT_BOUND_THR   = 0x1,
 297         SCHED_IPI_EVENT_PREEMPT     = 0x2,
 298         SCHED_IPI_EVENT_SMT_REBAL   = 0x3,
 299         SCHED_IPI_EVENT_SPILL       = 0x4,
 300         SCHED_IPI_EVENT_REBALANCE   = 0x5,
 301 } sched_ipi_event_t;
 302
 303
 304 /* Enum to define various IPI types used by the scheduler */
 305 typedef enum {
 306         SCHED_IPI_NONE              = 0x0,
 307         SCHED_IPI_IMMEDIATE         = 0x1,
 308         SCHED_IPI_IDLE              = 0x2,
 309         SCHED_IPI_DEFERRED          = 0x3,
 310 } sched_ipi_type_t;
 311
 312 /* The IPI policy engine behaves in the following manner:
 313  * - All scheduler events which need an IPI invoke sched_ipi_action() with
 314  *   the appropriate destination processor, thread and event.
 315  * - sched_ipi_action() performs basic checks, invokes the scheduler specific
 316  *   ipi_policy routine and sets pending_AST bits based on the result.
 317  * - Once the pset lock is dropped, the scheduler invokes sched_ipi_perform()
 318  *   routine which actually sends the appropriate IPI to the destination core.
 319  */
 320 extern sched_ipi_type_t sched_ipi_action(processor_t dst, thread_t thread,
 321     boolean_t dst_idle, sched_ipi_event_t event);
 322 extern void sched_ipi_perform(processor_t dst, sched_ipi_type_t ipi);
 323
 324 /* sched_ipi_policy() is the global default IPI policy for all schedulers */
 325 extern sched_ipi_type_t sched_ipi_policy(processor_t dst, thread_t thread,
 326     boolean_t dst_idle, sched_ipi_event_t event);
 327
 328 /* sched_ipi_deferred_policy() is the global default deferred IPI policy for all schedulers */
 329 extern sched_ipi_type_t sched_ipi_deferred_policy(processor_set_t pset,
 330     processor_t dst, sched_ipi_event_t event);
 331
 332 #if defined(CONFIG_SCHED_TIMESHARE_CORE)
 333
 334 extern boolean_t        thread_update_add_thread(thread_t thread);
 335 extern void             thread_update_process_threads(void);
 336 extern boolean_t        runq_scan(run_queue_t runq, sched_update_scan_context_t scan_context);
 337
 338 extern void sched_timeshare_init(void);
 339 extern void sched_timeshare_timebase_init(void);
 340 extern void sched_timeshare_maintenance_continue(void);
 341
 342 extern boolean_t priority_is_urgent(int priority);
 343 extern uint32_t sched_timeshare_initial_quantum_size(thread_t thread);
 344
 345 extern int sched_compute_timeshare_priority(thread_t thread);
 346
 347 #endif /* CONFIG_SCHED_TIMESHARE_CORE */
 348
 349 /* Remove thread from its run queue */
 350 extern boolean_t        thread_run_queue_remove(thread_t thread);
 351 thread_t thread_run_queue_remove_for_handoff(thread_t thread);
 352
 353 /* Put a thread back in the run queue after being yanked */
 354 extern void thread_run_queue_reinsert(thread_t thread, integer_t options);
 355
 356 extern void             thread_timer_expire(
 357         void                    *thread,
 358         void                    *p1);
 359
 360 extern boolean_t        thread_eager_preemption(
 361         thread_t thread);
 362
 363 extern boolean_t sched_generic_direct_dispatch_to_idle_processors;
 364
 365 /* Set the maximum interrupt level for the thread */
 366 __private_extern__ wait_interrupt_t thread_interrupt_level(
 367         wait_interrupt_t interruptible);
 368
 369 __private_extern__ wait_result_t thread_mark_wait_locked(
 370         thread_t                 thread,
 371         wait_interrupt_t interruptible);
 372
 373 /* Wake up locked thread directly, passing result */
 374 __private_extern__ kern_return_t clear_wait_internal(
 375         thread_t                thread,
 376         wait_result_t   result);
 377
 378 extern void sched_stats_handle_csw(
 379         processor_t processor,
 380         int reasons,
 381         int selfpri,
 382         int otherpri);
 383
 384 extern void sched_stats_handle_runq_change(
 385         struct runq_stats *stats,
 386         int old_count);
 387
 388
 389 #if DEBUG
 390
 391 #define SCHED_STATS_CSW(processor, reasons, selfpri, otherpri)          \
 392 do {                                                            \
 393         if (__builtin_expect(sched_stats_active, 0)) {  \
 394                 sched_stats_handle_csw((processor),             \
 395                                 (reasons), (selfpri), (otherpri));      \
 396         }                                                       \
 397 } while (0)
 398
 399
 400 #define SCHED_STATS_RUNQ_CHANGE(stats, old_count)               \
 401 do {                                                            \
 402         if (__builtin_expect(sched_stats_active, 0)) {  \
 403                 sched_stats_handle_runq_change((stats),         \
 404                                                                 (old_count));           \
 405         }                                                       \
 406 } while (0)
 407
 408 #else /* DEBUG */
 409
 410 #define SCHED_STATS_CSW(processor, reasons, selfpri, otherpri) do { }while(0)
 411 #define SCHED_STATS_RUNQ_CHANGE(stats, old_count) do { }while(0)
 412
 413 #endif /* DEBUG */
 414
 415 extern uint32_t sched_debug_flags;
 416 #define SCHED_DEBUG_FLAG_PLATFORM_TRACEPOINTS   0x00000001
 417 #define SCHED_DEBUG_FLAG_CHOOSE_PROCESSOR_TRACEPOINTS   0x00000002
 418
 419 #define SCHED_DEBUG_PLATFORM_KERNEL_DEBUG_CONSTANT(...) do {                                            \
 420                 if (__improbable(sched_debug_flags & SCHED_DEBUG_FLAG_PLATFORM_TRACEPOINTS)) { \
 421                         KERNEL_DEBUG_CONSTANT(__VA_ARGS__);                                                     \
 422                 }                                                                                                                               \
 423         } while(0)
 424
 425 #define SCHED_DEBUG_CHOOSE_PROCESSOR_KERNEL_DEBUG_CONSTANT(...) do {                                            \
 426                 if (__improbable(sched_debug_flags & SCHED_DEBUG_FLAG_CHOOSE_PROCESSOR_TRACEPOINTS)) { \
 427                         KERNEL_DEBUG_CONSTANT(__VA_ARGS__);                                                     \
 428                 }                                                                                                                               \
 429         } while(0)
 430
 431 /* Tells if there are "active" RT threads in the system (provided by CPU PM) */
 432 extern void     active_rt_threads(
 433         boolean_t       active);
 434
 435 /* Returns the perfcontrol attribute for the thread */
 436 extern perfcontrol_class_t thread_get_perfcontrol_class(
 437         thread_t        thread);
 438
 439 /* Generic routine for Non-AMP schedulers to calculate parallelism */
 440 extern uint32_t sched_qos_max_parallelism(int qos, uint64_t options);
 441
 442 #endif /* MACH_KERNEL_PRIVATE */
 443
 444 __BEGIN_DECLS
 445
 446 #ifdef  XNU_KERNEL_PRIVATE
 447
 448 /* Toggles a global override to turn off CPU Throttling */
 449 extern void     sys_override_cpu_throttle(boolean_t enable_override);
 450
 451 /*
 452  ****************** Only exported until BSD stops using ********************
 453  */
 454
 455 extern void                     thread_vm_bind_group_add(void);
 456
 457 /* Wake up thread directly, passing result */
 458 extern kern_return_t clear_wait(
 459         thread_t                thread,
 460         wait_result_t   result);
 461
 462 /* Start thread running */
 463 extern void             thread_bootstrap_return(void) __attribute__((noreturn));
 464
 465 /* Return from exception (BSD-visible interface) */
 466 extern void             thread_exception_return(void) __dead2;
 467
 468 #define SCHED_STRING_MAX_LENGTH (48)
 469 /* String declaring the name of the current scheduler */
 470 extern char sched_string[SCHED_STRING_MAX_LENGTH];
 471
 472 extern thread_t port_name_to_thread_for_ulock(mach_port_name_t  thread_name);
 473
 474 /* Attempt to context switch to a specific runnable thread */
 475 extern wait_result_t thread_handoff_deallocate(thread_t thread);
 476
 477 __attribute__((nonnull(1, 2)))
 478 extern void thread_handoff_parameter(thread_t thread,
 479     thread_continue_t continuation, void *parameter) __dead2;
 480
 481 extern struct waitq     *assert_wait_queue(event_t event);
 482
 483 extern kern_return_t thread_wakeup_one_with_pri(event_t event, int priority);
 484
 485 extern thread_t thread_wakeup_identify(event_t event, int priority);
 486
 487 #endif  /* XNU_KERNEL_PRIVATE */
 488
 489 #ifdef KERNEL_PRIVATE
 490 /* Set pending block hint for a particular object before we go into a wait state */
 491 extern void             thread_set_pending_block_hint(
 492         thread_t                        thread,
 493         block_hint_t                    block_hint);
 494
 495 #define QOS_PARALLELISM_COUNT_LOGICAL   0x1
 496 #define QOS_PARALLELISM_REALTIME        0x2
 497 extern uint32_t qos_max_parallelism(int qos, uint64_t options);
 498 #endif /* KERNEL_PRIVATE */
 499
 500 #if XNU_KERNEL_PRIVATE
 501 extern void             thread_yield_with_continuation(
 502         thread_continue_t       continuation,
 503         void                            *parameter) __dead2;
 504 #endif
 505
 506 /* Context switch */
 507 extern wait_result_t    thread_block(
 508         thread_continue_t       continuation);
 509
 510 extern wait_result_t    thread_block_parameter(
 511         thread_continue_t       continuation,
 512         void                            *parameter);
 513
 514 /* Declare thread will wait on a particular event */
 515 extern wait_result_t    assert_wait(
 516         event_t                         event,
 517         wait_interrupt_t        interruptible);
 518
 519 /* Assert that the thread intends to wait with a timeout */
 520 extern wait_result_t    assert_wait_timeout(
 521         event_t                         event,
 522         wait_interrupt_t        interruptible,
 523         uint32_t                        interval,
 524         uint32_t                        scale_factor);
 525
 526 /* Assert that the thread intends to wait with an urgency, timeout and leeway */
 527 extern wait_result_t    assert_wait_timeout_with_leeway(
 528         event_t                         event,
 529         wait_interrupt_t        interruptible,
 530         wait_timeout_urgency_t  urgency,
 531         uint32_t                        interval,
 532         uint32_t                        leeway,
 533         uint32_t                        scale_factor);
 534
 535 extern wait_result_t    assert_wait_deadline(
 536         event_t                         event,
 537         wait_interrupt_t        interruptible,
 538         uint64_t                        deadline);
 539
 540 /* Assert that the thread intends to wait with an urgency, deadline, and leeway */
 541 extern wait_result_t    assert_wait_deadline_with_leeway(
 542         event_t                         event,
 543         wait_interrupt_t        interruptible,
 544         wait_timeout_urgency_t  urgency,
 545         uint64_t                        deadline,
 546         uint64_t                        leeway);
 547
 548 /* Wake up thread (or threads) waiting on a particular event */
 549 extern kern_return_t    thread_wakeup_prim(
 550         event_t                         event,
 551         boolean_t                       one_thread,
 552         wait_result_t                   result);
 553
 554 #define thread_wakeup(x)                                        \
 555                         thread_wakeup_prim((x), FALSE, THREAD_AWAKENED)
 556 #define thread_wakeup_with_result(x, z)         \
 557                         thread_wakeup_prim((x), FALSE, (z))
 558 #define thread_wakeup_one(x)                            \
 559                         thread_wakeup_prim((x), TRUE, THREAD_AWAKENED)
 560
 561 /* Wakeup the specified thread if it is waiting on this event */
 562 extern kern_return_t thread_wakeup_thread(event_t event, thread_t thread);
 563
 564 extern boolean_t preemption_enabled(void);
 565
 566 #ifdef MACH_KERNEL_PRIVATE
 567
 568 /*
 569  * Scheduler algorithm indirection. If only one algorithm is
 570  * enabled at compile-time, a direction function call is used.
 571  * If more than one is enabled, calls are dispatched through
 572  * a function pointer table.
 573  */
 574
 575 #if   !defined(CONFIG_SCHED_TRADITIONAL) && !defined(CONFIG_SCHED_PROTO) && !defined(CONFIG_SCHED_GRRR) && !defined(CONFIG_SCHED_MULTIQ)
 576 #error Enable at least one scheduler algorithm in osfmk/conf/MASTER.XXX
 577 #endif
 578
 579 #if DEBUG
 580 #define SCHED(f) (sched_current_dispatch->f)
 581 #else /* DEBUG */
 582
 583 /*
 584  * For DEV & REL kernels, use a static dispatch table instead of
 585  * using the indirect function table.
 586  */
 587 extern const struct sched_dispatch_table sched_dualq_dispatch;
 588 #define SCHED(f) (sched_dualq_dispatch.f)
 589
 590 #endif /* DEBUG */
 591
 592 struct sched_dispatch_table {
 593         const char *sched_name;
 594         void    (*init)(void);                          /* Init global state */
 595         void    (*timebase_init)(void);         /* Timebase-dependent initialization */
 596         void    (*processor_init)(processor_t processor);       /* Per-processor scheduler init */
 597         void    (*pset_init)(processor_set_t pset);     /* Per-processor set scheduler init */
 598
 599         void    (*maintenance_continuation)(void);      /* Function called regularly */
 600
 601         /*
 602          * Choose a thread of greater or equal priority from the per-processor
 603          * runqueue for timeshare/fixed threads
 604          */
 605         thread_t        (*choose_thread)(
 606                 processor_t           processor,
 607                 int                           priority,
 608                 ast_t reason);
 609
 610         /* True if scheduler supports stealing threads for this pset */
 611         bool    (*steal_thread_enabled)(processor_set_t pset);
 612
 613         /*
 614          * Steal a thread from another processor in the pset so that it can run
 615          * immediately
 616          */
 617         thread_t        (*steal_thread)(
 618                 processor_set_t         pset);
 619
 620         /*
 621          * Compute priority for a timeshare thread based on base priority.
 622          */
 623         int (*compute_timeshare_priority)(thread_t thread);
 624
 625         /*
 626          * Pick the best processor for a thread (any kind of thread) to run on.
 627          */
 628         processor_t     (*choose_processor)(
 629                 processor_set_t                pset,
 630                 processor_t                    processor,
 631                 thread_t                       thread);
 632         /*
 633          * Enqueue a timeshare or fixed priority thread onto the per-processor
 634          * runqueue
 635          */
 636         boolean_t (*processor_enqueue)(
 637                 processor_t                    processor,
 638                 thread_t                       thread,
 639                 integer_t                      options);
 640
 641         /* Migrate threads away in preparation for processor shutdown */
 642         void (*processor_queue_shutdown)(
 643                 processor_t                    processor);
 644
 645         /* Remove the specific thread from the per-processor runqueue */
 646         boolean_t       (*processor_queue_remove)(
 647                 processor_t             processor,
 648                 thread_t                thread);
 649
 650         /*
 651          * Does the per-processor runqueue have any timeshare or fixed priority
 652          * threads on it? Called without pset lock held, so should
 653          * not assume immutability while executing.
 654          */
 655         boolean_t       (*processor_queue_empty)(processor_t            processor);
 656
 657         /*
 658          * Would this priority trigger an urgent preemption if it's sitting
 659          * on the per-processor runqueue?
 660          */
 661         boolean_t       (*priority_is_urgent)(int priority);
 662
 663         /*
 664          * Does the per-processor runqueue contain runnable threads that
 665          * should cause the currently-running thread to be preempted?
 666          */
 667         ast_t           (*processor_csw_check)(processor_t processor);
 668
 669         /*
 670          * Does the per-processor runqueue contain a runnable thread
 671          * of > or >= priority, as a preflight for choose_thread() or other
 672          * thread selection
 673          */
 674         boolean_t       (*processor_queue_has_priority)(processor_t             processor,
 675             int                             priority,
 676             boolean_t               gte);
 677
 678         /* Quantum size for the specified non-realtime thread. */
 679         uint32_t        (*initial_quantum_size)(thread_t thread);
 680
 681         /* Scheduler mode for a new thread */
 682         sched_mode_t    (*initial_thread_sched_mode)(task_t parent_task);
 683
 684         /*
 685          * Is it safe to call update_priority, which may change a thread's
 686          * runqueue or other state. This can be used to throttle changes
 687          * to dynamic priority.
 688          */
 689         boolean_t       (*can_update_priority)(thread_t thread);
 690
 691         /*
 692          * Update both scheduled priority and other persistent state.
 693          * Side effects may including migration to another processor's runqueue.
 694          */
 695         void            (*update_priority)(thread_t thread);
 696
 697         /* Lower overhead update to scheduled priority and state. */
 698         void            (*lightweight_update_priority)(thread_t thread);
 699
 700         /* Callback for non-realtime threads when the quantum timer fires */
 701         void            (*quantum_expire)(thread_t thread);
 702
 703         /*
 704          * Runnable threads on per-processor runqueue. Should only
 705          * be used for relative comparisons of load between processors.
 706          */
 707         int                     (*processor_runq_count)(processor_t     processor);
 708
 709         /* Aggregate runcount statistics for per-processor runqueue */
 710         uint64_t    (*processor_runq_stats_count_sum)(processor_t   processor);
 711
 712         boolean_t       (*processor_bound_count)(processor_t processor);
 713
 714         void            (*thread_update_scan)(sched_update_scan_context_t scan_context);
 715
 716         /*
 717          * Use processor->next_thread to pin a thread to an idle
 718          * processor. If FALSE, threads are enqueued and can
 719          * be stolen by other processors.
 720          */
 721         boolean_t   direct_dispatch_to_idle_processors;
 722
 723         /* Supports more than one pset */
 724         boolean_t   multiple_psets_enabled;
 725         /* Supports scheduler groups */
 726         boolean_t   sched_groups_enabled;
 727
 728         /* Supports avoid-processor */
 729         boolean_t   avoid_processor_enabled;
 730
 731         /* Returns true if this processor should avoid running this thread. */
 732         bool    (*thread_avoid_processor)(processor_t processor, thread_t thread);
 733
 734         /*
 735          * Invoked when a processor is about to choose the idle thread
 736          * Used to send IPIs to a processor which would be preferred to be idle instead.
 737          * Called with pset lock held, returns pset lock unlocked.
 738          */
 739         void    (*processor_balance)(processor_t processor, processor_set_t pset);
 740         rt_queue_t      (*rt_runq)(processor_set_t pset);
 741         void    (*rt_init)(processor_set_t pset);
 742         void    (*rt_queue_shutdown)(processor_t processor);
 743         void    (*rt_runq_scan)(sched_update_scan_context_t scan_context);
 744         int64_t (*rt_runq_count_sum)(void);
 745
 746         uint32_t (*qos_max_parallelism)(int qos, uint64_t options);
 747         void    (*check_spill)(processor_set_t pset, thread_t thread);
 748         sched_ipi_type_t (*ipi_policy)(processor_t dst, thread_t thread, boolean_t dst_idle, sched_ipi_event_t event);
 749         bool    (*thread_should_yield)(processor_t processor, thread_t thread);
 750 };
 751
 752 #if defined(CONFIG_SCHED_TRADITIONAL)
 753 extern const struct sched_dispatch_table sched_traditional_dispatch;
 754 extern const struct sched_dispatch_table sched_traditional_with_pset_runqueue_dispatch;
 755 #endif
 756
 757 #if defined(CONFIG_SCHED_MULTIQ)
 758 extern const struct sched_dispatch_table sched_multiq_dispatch;
 759 extern const struct sched_dispatch_table sched_dualq_dispatch;
 760 #endif
 761
 762 #if defined(CONFIG_SCHED_PROTO)
 763 extern const struct sched_dispatch_table sched_proto_dispatch;
 764 #endif
 765
 766 #if defined(CONFIG_SCHED_GRRR)
 767 extern const struct sched_dispatch_table sched_grrr_dispatch;
 768 #endif
 769
 770 /*
 771  * It is an error to invoke any scheduler-related code
 772  * before this is set up
 773  */
 774 extern const struct sched_dispatch_table *sched_current_dispatch;
 775
 776 #endif  /* MACH_KERNEL_PRIVATE */
 777
 778 __END_DECLS
 779
 780 #endif  /* _KERN_SCHED_PRIM_H_ */