osfmk/kern/sched.h

   1 /*
   2  * Copyright (c) 2000-2009 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.h
  60  *      Author: Avadis Tevanian, Jr.
  61  *      Date:   1985
  62  *
  63  *      Header file for scheduler.
  64  *
  65  */
  66
  67 #ifndef _KERN_SCHED_H_
  68 #define _KERN_SCHED_H_
  69
  70 #include <mach/policy.h>
  71 #include <kern/kern_types.h>
  72 #include <kern/queue.h>
  73 #include <kern/macro_help.h>
  74 #include <kern/timer_call.h>
  75 #include <kern/ast.h>
  76
  77 #define NRQS            128                             /* 128 levels per run queue */
  78 #define NRQBM           (NRQS / 32)             /* number of words per bit map */
  79
  80 #define MAXPRI          (NRQS-1)
  81 #define MINPRI          IDLEPRI                 /* lowest legal priority schedulable */
  82 #define IDLEPRI         0                               /* idle thread priority */
  83
  84 /*
  85  *      High-level priority assignments
  86  *
  87  *************************************************************************
  88  * 127          Reserved (real-time)
  89  *                              A
  90  *                              +
  91  *                      (32 levels)
  92  *                              +
  93  *                              V
  94  * 96           Reserved (real-time)
  95  * 95           Kernel mode only
  96  *                              A
  97  *                              +
  98  *                      (16 levels)
  99  *                              +
 100  *                              V
 101  * 80           Kernel mode only
 102  * 79           System high priority
 103  *                              A
 104  *                              +
 105  *                      (16 levels)
 106  *                              +
 107  *                              V
 108  * 64           System high priority
 109  * 63           Elevated priorities
 110  *                              A
 111  *                              +
 112  *                      (12 levels)
 113  *                              +
 114  *                              V
 115  * 52           Elevated priorities
 116  * 51           Elevated priorities (incl. BSD +nice)
 117  *                              A
 118  *                              +
 119  *                      (20 levels)
 120  *                              +
 121  *                              V
 122  * 32           Elevated priorities (incl. BSD +nice)
 123  * 31           Default (default base for threads)
 124  * 30           Lowered priorities (incl. BSD -nice)
 125  *                              A
 126  *                              +
 127  *                      (20 levels)
 128  *                              +
 129  *                              V
 130  * 11           Lowered priorities (incl. BSD -nice)
 131  * 10           Lowered priorities (aged pri's)
 132  *                              A
 133  *                              +
 134  *                      (11 levels)
 135  *                              +
 136  *                              V
 137  * 0            Lowered priorities (aged pri's / idle)
 138  *************************************************************************
 139  */
 140
 141 #define BASEPRI_RTQUEUES        (BASEPRI_REALTIME + 1)                          /* 97 */
 142 #define BASEPRI_REALTIME        (MAXPRI - (NRQS / 4) + 1)                       /* 96 */
 143
 144 #define MAXPRI_KERNEL           (BASEPRI_REALTIME - 1)                          /* 95 */
 145 #define BASEPRI_PREEMPT         (MAXPRI_KERNEL - 2)                                     /* 93 */
 146 #define BASEPRI_KERNEL          (MINPRI_KERNEL + 1)                                     /* 81 */
 147 #define MINPRI_KERNEL           (MAXPRI_KERNEL - (NRQS / 8) + 1)        /* 80 */
 148
 149 #define MAXPRI_RESERVED         (MINPRI_KERNEL - 1)                                     /* 79 */
 150 #define BASEPRI_GRAPHICS        (MAXPRI_RESERVED - 3)                           /* 76 */
 151 #define MINPRI_RESERVED         (MAXPRI_RESERVED - (NRQS / 8) + 1)      /* 64 */
 152
 153 #define MAXPRI_USER                     (MINPRI_RESERVED - 1)                           /* 63 */
 154 #define BASEPRI_CONTROL         (BASEPRI_DEFAULT + 17)                          /* 48 */
 155 #define BASEPRI_FOREGROUND      (BASEPRI_DEFAULT + 16)                          /* 47 */
 156 #define BASEPRI_BACKGROUND      (BASEPRI_DEFAULT + 15)                          /* 46 */
 157 #define BASEPRI_USER_INITIATED  (BASEPRI_DEFAULT +  6)                          /* 37 */
 158 #define BASEPRI_DEFAULT         (MAXPRI_USER - (NRQS / 4))                      /* 31 */
 159 #define MAXPRI_SUPPRESSED       (BASEPRI_DEFAULT - 3)                           /* 28 */
 160 #define BASEPRI_UTILITY         (BASEPRI_DEFAULT - 11)                          /* 20 */
 161 #define MAXPRI_THROTTLE         (MINPRI + 4)                                            /*  4 */
 162 #define MINPRI_USER                     MINPRI                                                          /*  0 */
 163
 164 #define DEPRESSPRI      MINPRI                  /* depress priority */
 165 #define MAXPRI_PROMOTE          (MAXPRI_KERNEL)         /* ceiling for mutex promotion */
 166
 167 /* Type used for thread->sched_mode and saved_mode */
 168 typedef enum {
 169         TH_MODE_NONE = 0,                                       /* unassigned, usually for saved_mode only */
 170         TH_MODE_REALTIME,                                       /* time constraints supplied */
 171         TH_MODE_FIXED,                                          /* use fixed priorities, no decay */
 172         TH_MODE_TIMESHARE,                                      /* use timesharing algorithm */
 173 } sched_mode_t;
 174
 175 /*
 176  *      Macro to check for invalid priorities.
 177  */
 178 #define invalid_pri(pri) ((pri) < MINPRI || (pri) > MAXPRI)
 179
 180 struct runq_stats {
 181         uint64_t                                count_sum;
 182         uint64_t                                last_change_timestamp;
 183 };
 184
 185 #if defined(CONFIG_SCHED_TIMESHARE_CORE) || defined(CONFIG_SCHED_PROTO)
 186
 187 struct run_queue {
 188         int                                     highq;                          /* highest runnable queue */
 189         int                                     bitmap[NRQBM];          /* run queue bitmap array */
 190         int                                     count;                          /* # of threads total */
 191         int                                     urgency;                        /* level of preemption urgency */
 192         queue_head_t            queues[NRQS];           /* one for each priority */
 193
 194         struct runq_stats       runq_stats;
 195 };
 196
 197 #endif /* defined(CONFIG_SCHED_TIMESHARE_CORE) || defined(CONFIG_SCHED_PROTO) */
 198
 199 struct rt_queue {
 200         int                                     count;                          /* # of threads total */
 201         queue_head_t            queue;                          /* all runnable RT threads */
 202
 203         struct runq_stats       runq_stats;
 204 };
 205
 206 #if defined(CONFIG_SCHED_GRRR_CORE)
 207
 208 /*
 209  * We map standard Mach priorities to an abstract scale that more properly
 210  * indicates how we want processor time allocated under contention.
 211  */
 212 typedef uint8_t grrr_proportional_priority_t;
 213 typedef uint8_t grrr_group_index_t;
 214
 215 #define NUM_GRRR_PROPORTIONAL_PRIORITIES        256
 216 #define MAX_GRRR_PROPORTIONAL_PRIORITY ((grrr_proportional_priority_t)255)
 217
 218 #if 0
 219 #define NUM_GRRR_GROUPS 8                                       /* log(256) */
 220 #endif
 221
 222 #define NUM_GRRR_GROUPS 64                                      /* 256/4 */
 223
 224 struct grrr_group {
 225         queue_chain_t                   priority_order;                         /* next greatest weight group */
 226         grrr_proportional_priority_t            minpriority;
 227         grrr_group_index_t              index;
 228
 229         queue_head_t                    clients;
 230         int                                             count;
 231         uint32_t                                weight;
 232 #if 0
 233         uint32_t                                deferred_removal_weight;
 234 #endif
 235         uint32_t                                work;
 236         thread_t                                current_client;
 237 };
 238
 239 struct grrr_run_queue {
 240         int                                     count;
 241         uint32_t                        last_rescale_tick;
 242         struct grrr_group       groups[NUM_GRRR_GROUPS];
 243         queue_head_t            sorted_group_list;
 244         uint32_t                        weight;
 245         grrr_group_t            current_group;
 246
 247         struct runq_stats   runq_stats;
 248 };
 249
 250 #endif /* defined(CONFIG_SCHED_GRRR_CORE) */
 251
 252 extern struct rt_queue          rt_runq;
 253
 254 #if defined(CONFIG_SCHED_MULTIQ)
 255 sched_group_t   sched_group_create(void);
 256 void            sched_group_destroy(sched_group_t sched_group);
 257 #endif /* defined(CONFIG_SCHED_MULTIQ) */
 258
 259
 260
 261 /*
 262  *      Scheduler routines.
 263  */
 264
 265 /* Handle quantum expiration for an executing thread */
 266 extern void             thread_quantum_expire(
 267                                         timer_call_param_t      processor,
 268                                         timer_call_param_t      thread);
 269
 270 /* Context switch check for current processor */
 271 extern ast_t    csw_check(processor_t           processor,
 272                                                 ast_t                   check_reason);
 273
 274 #if defined(CONFIG_SCHED_TIMESHARE_CORE)
 275 extern uint32_t std_quantum, min_std_quantum;
 276 extern uint32_t std_quantum_us;
 277 #endif /* CONFIG_SCHED_TIMESHARE_CORE */
 278
 279 extern uint32_t thread_depress_time;
 280 extern uint32_t default_timeshare_computation;
 281 extern uint32_t default_timeshare_constraint;
 282
 283 extern uint32_t max_rt_quantum, min_rt_quantum;
 284
 285 extern int default_preemption_rate;
 286 extern int default_bg_preemption_rate;
 287
 288 #if defined(CONFIG_SCHED_TIMESHARE_CORE)
 289
 290 /*
 291  *      Age usage  at approximately (1 << SCHED_TICK_SHIFT) times per second
 292  *      Aging may be deferred during periods where all processors are idle
 293  *      and cumulatively applied during periods of activity.
 294  */
 295 #define SCHED_TICK_SHIFT        3
 296 #define SCHED_TICK_MAX_DELTA    (8)
 297
 298 extern unsigned         sched_tick;
 299 extern uint32_t         sched_tick_interval;
 300
 301 #endif /* CONFIG_SCHED_TIMESHARE_CORE */
 302
 303 extern uint64_t         sched_one_second_interval;
 304
 305 /* Periodic computation of various averages */
 306 extern void             compute_averages(uint64_t);
 307
 308 extern void             compute_averunnable(
 309                                         void                    *nrun);
 310
 311 extern void             compute_stack_target(
 312                                         void                    *arg);
 313
 314 extern void             compute_memory_pressure(
 315                                         void                    *arg);
 316
 317 extern void             compute_zone_gc_throttle(
 318                                         void                    *arg);
 319
 320 extern void             compute_pageout_gc_throttle(
 321                                         void                    *arg);
 322
 323 extern void             compute_pmap_gc_throttle(
 324                                         void                    *arg);
 325
 326 /*
 327  *      Conversion factor from usage
 328  *      to priority.
 329  */
 330 #if defined(CONFIG_SCHED_TIMESHARE_CORE)
 331 extern uint32_t         sched_pri_shift;
 332 extern uint32_t         sched_background_pri_shift;
 333 extern uint32_t         sched_combined_fgbg_pri_shift;
 334 extern uint32_t         sched_fixed_shift;
 335 extern int8_t           sched_load_shifts[NRQS];
 336 extern uint32_t         sched_decay_usage_age_factor;
 337 extern uint32_t         sched_use_combined_fgbg_decay;
 338 void sched_timeshare_consider_maintenance(uint64_t);
 339 #endif /* CONFIG_SCHED_TIMESHARE_CORE */
 340
 341 extern int32_t          sched_poll_yield_shift;
 342 extern uint64_t         sched_safe_duration;
 343
 344 extern uint32_t         sched_run_count, sched_share_count, sched_background_count;
 345 extern uint32_t         sched_load_average, sched_mach_factor;
 346
 347 extern uint32_t         avenrun[3], mach_factor[3];
 348
 349 extern uint64_t         max_unsafe_computation;
 350 extern uint64_t         max_poll_computation;
 351
 352 /* TH_RUN & !TH_IDLE controls whether a thread has a run count */
 353 #define sched_run_incr(th)                                      \
 354         hw_atomic_add(&sched_run_count, 1)                      \
 355
 356 #define sched_run_decr(th)                                      \
 357         hw_atomic_sub(&sched_run_count, 1)                      \
 358
 359 #if MACH_ASSERT
 360 extern void sched_share_incr(thread_t thread);
 361 extern void sched_share_decr(thread_t thread);
 362 extern void sched_background_incr(thread_t thread);
 363 extern void sched_background_decr(thread_t thread);
 364
 365 extern void assert_thread_sched_count(thread_t thread);
 366
 367 #else /* MACH_ASSERT */
 368 /* sched_mode == TH_MODE_TIMESHARE controls whether a thread has a timeshare count when it has a run count */
 369 #define sched_share_incr(th)                            \
 370 MACRO_BEGIN                                             \
 371         (void)hw_atomic_add(&sched_share_count, 1);     \
 372 MACRO_END
 373
 374 #define sched_share_decr(th)                            \
 375 MACRO_BEGIN                                             \
 376         (void)hw_atomic_sub(&sched_share_count, 1);     \
 377 MACRO_END
 378
 379 /* TH_SFLAG_THROTTLED controls whether a thread has a background count when it has a run count and a share count */
 380 #define sched_background_incr(th)                       \
 381 MACRO_BEGIN                                             \
 382         hw_atomic_add(&sched_background_count, 1);      \
 383 MACRO_END
 384
 385 #define sched_background_decr(th)                       \
 386 MACRO_BEGIN                                             \
 387         hw_atomic_sub(&sched_background_count, 1);      \
 388 MACRO_END
 389
 390 #define assert_thread_sched_count(th)                   \
 391 MACRO_BEGIN                                             \
 392 MACRO_END
 393
 394 #endif /* !MACH_ASSERT */
 395
 396 /*
 397  *      thread_timer_delta macro takes care of both thread timers.
 398  */
 399 #define thread_timer_delta(thread, delta)                                       \
 400 MACRO_BEGIN                                                                                                     \
 401         (delta) = (typeof(delta))timer_delta(&(thread)->system_timer,                   \
 402                                                         &(thread)->system_timer_save);  \
 403         (delta) += (typeof(delta))timer_delta(&(thread)->user_timer,                    \
 404                                                         &(thread)->user_timer_save);    \
 405 MACRO_END
 406
 407 #endif  /* _KERN_SCHED_H_ */