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/lock.h>
  74 #include <kern/macro_help.h>
  75 #include <kern/timer_call.h>
  76 #include <kern/ast.h>
  77
  78 #define NRQS            128                             /* 128 levels per run queue */
  79 #define NRQBM           (NRQS / 32)             /* number of words per bit map */
  80
  81 #define MAXPRI          (NRQS-1)
  82 #define MINPRI          IDLEPRI                 /* lowest legal priority schedulable */
  83 #define IDLEPRI         0                               /* idle thread priority */
  84
  85 /*
  86  *      High-level priority assignments
  87  *
  88  *************************************************************************
  89  * 127          Reserved (real-time)
  90  *                              A
  91  *                              +
  92  *                      (32 levels)
  93  *                              +
  94  *                              V
  95  * 96           Reserved (real-time)
  96  * 95           Kernel mode only
  97  *                              A
  98  *                              +
  99  *                      (16 levels)
 100  *                              +
 101  *                              V
 102  * 80           Kernel mode only
 103  * 79           System high priority
 104  *                              A
 105  *                              +
 106  *                      (16 levels)
 107  *                              +
 108  *                              V
 109  * 64           System high priority
 110  * 63           Elevated priorities
 111  *                              A
 112  *                              +
 113  *                      (12 levels)
 114  *                              +
 115  *                              V
 116  * 52           Elevated priorities
 117  * 51           Elevated priorities (incl. BSD +nice)
 118  *                              A
 119  *                              +
 120  *                      (20 levels)
 121  *                              +
 122  *                              V
 123  * 32           Elevated priorities (incl. BSD +nice)
 124  * 31           Default (default base for threads)
 125  * 30           Lowered priorities (incl. BSD -nice)
 126  *                              A
 127  *                              +
 128  *                      (20 levels)
 129  *                              +
 130  *                              V
 131  * 11           Lowered priorities (incl. BSD -nice)
 132  * 10           Lowered priorities (aged pri's)
 133  *                              A
 134  *                              +
 135  *                      (11 levels)
 136  *                              +
 137  *                              V
 138  * 0            Lowered priorities (aged pri's / idle)
 139  *************************************************************************
 140  */
 141
 142 #define BASEPRI_RTQUEUES        (BASEPRI_REALTIME + 1)                          /* 97 */
 143 #define BASEPRI_REALTIME        (MAXPRI - (NRQS / 4) + 1)                       /* 96 */
 144
 145 #define MAXPRI_KERNEL           (BASEPRI_REALTIME - 1)                          /* 95 */
 146 #define BASEPRI_PREEMPT         (MAXPRI_KERNEL - 2)                                     /* 93 */
 147 #define BASEPRI_KERNEL          (MINPRI_KERNEL + 1)                                     /* 81 */
 148 #define MINPRI_KERNEL           (MAXPRI_KERNEL - (NRQS / 8) + 1)        /* 80 */
 149
 150 #define MAXPRI_RESERVED         (MINPRI_KERNEL - 1)                                     /* 79 */
 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_DEFAULT         (MAXPRI_USER - (NRQS / 4))                      /* 31 */
 158 #define MAXPRI_THROTTLE         (MINPRI + 4)                                            /*  4 */
 159 #define MINPRI_USER                     MINPRI                                                          /*  0 */
 160
 161 #ifdef CONFIG_EMBEDDED
 162 #define DEPRESSPRI      MAXPRI_THROTTLE
 163 #else
 164 #define DEPRESSPRI      MINPRI                  /* depress priority */
 165 #endif
 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         TH_MODE_FAIRSHARE                                       /* use fair-share scheduling */
 174 } sched_mode_t;
 175
 176 /*
 177  *      Macro to check for invalid priorities.
 178  */
 179 #define invalid_pri(pri) ((pri) < MINPRI || (pri) > MAXPRI)
 180
 181 struct runq_stats {
 182         uint64_t                                count_sum;
 183         uint64_t                                last_change_timestamp;
 184 };
 185
 186 #if defined(CONFIG_SCHED_TRADITIONAL) || defined(CONFIG_SCHED_PROTO) || defined(CONFIG_SCHED_FIXEDPRIORITY)
 187
 188 struct run_queue {
 189         int                                     highq;                          /* highest runnable queue */
 190         int                                     bitmap[NRQBM];          /* run queue bitmap array */
 191         int                                     count;                          /* # of threads total */
 192         int                                     urgency;                        /* level of preemption urgency */
 193         queue_head_t            queues[NRQS];           /* one for each priority */
 194
 195         struct runq_stats       runq_stats;
 196 };
 197
 198 #endif /* defined(CONFIG_SCHED_TRADITIONAL) || defined(CONFIG_SCHED_PROTO) || defined(CONFIG_SCHED_FIXEDPRIORITY) */
 199
 200 struct rt_queue {
 201         int                                     count;                          /* # of threads total */
 202         queue_head_t            queue;                          /* all runnable RT threads */
 203
 204         struct runq_stats       runq_stats;
 205 };
 206
 207 #if defined(CONFIG_SCHED_TRADITIONAL) || defined(CONFIG_SCHED_PROTO) || defined(CONFIG_SCHED_FIXEDPRIORITY)
 208 struct fairshare_queue {
 209         int                                     count;                          /* # of threads total */
 210         queue_head_t            queue;                          /* all runnable threads demoted to fairshare scheduling */
 211
 212         struct runq_stats       runq_stats;
 213 };
 214 #endif
 215
 216 #if defined(CONFIG_SCHED_GRRR_CORE)
 217
 218 /*
 219  * We map standard Mach priorities to an abstract scale that more properly
 220  * indicates how we want processor time allocated under contention.
 221  */
 222 typedef uint8_t grrr_proportional_priority_t;
 223 typedef uint8_t grrr_group_index_t;
 224
 225 #define NUM_GRRR_PROPORTIONAL_PRIORITIES        256
 226 #define MAX_GRRR_PROPORTIONAL_PRIORITY ((grrr_proportional_priority_t)255)
 227
 228 #if 0
 229 #define NUM_GRRR_GROUPS 8                                       /* log(256) */
 230 #endif
 231
 232 #define NUM_GRRR_GROUPS 64                                      /* 256/4 */
 233
 234 struct grrr_group {
 235         queue_chain_t                   priority_order;                         /* next greatest weight group */
 236         grrr_proportional_priority_t            minpriority;
 237         grrr_group_index_t              index;
 238
 239         queue_head_t                    clients;
 240         int                                             count;
 241         uint32_t                                weight;
 242 #if 0
 243         uint32_t                                deferred_removal_weight;
 244 #endif
 245         uint32_t                                work;
 246         thread_t                                current_client;
 247 };
 248
 249 struct grrr_run_queue {
 250         int                                     count;
 251         uint32_t                        last_rescale_tick;
 252         struct grrr_group       groups[NUM_GRRR_GROUPS];
 253         queue_head_t            sorted_group_list;
 254         uint32_t                        weight;
 255         grrr_group_t            current_group;
 256
 257         struct runq_stats   runq_stats;
 258 };
 259
 260 #endif /* defined(CONFIG_SCHED_GRRR_CORE) */
 261
 262 #define first_timeslice(processor)              ((processor)->timeslice > 0)
 263
 264 extern struct rt_queue          rt_runq;
 265
 266 /*
 267  *      Scheduler routines.
 268  */
 269
 270 /* Handle quantum expiration for an executing thread */
 271 extern void             thread_quantum_expire(
 272                                         timer_call_param_t      processor,
 273                                         timer_call_param_t      thread);
 274
 275 /* Context switch check for current processor */
 276 extern ast_t    csw_check(processor_t           processor);
 277
 278 #if defined(CONFIG_SCHED_TRADITIONAL)
 279 extern uint32_t std_quantum, min_std_quantum;
 280 extern uint32_t std_quantum_us;
 281 #endif
 282
 283 extern uint32_t thread_depress_time;
 284 extern uint32_t default_timeshare_computation;
 285 extern uint32_t default_timeshare_constraint;
 286
 287 extern uint32_t max_rt_quantum, min_rt_quantum;
 288
 289 extern int default_preemption_rate;
 290 extern int default_bg_preemption_rate;
 291
 292 #if defined(CONFIG_SCHED_TRADITIONAL)
 293
 294 /*
 295  *      Age usage (1 << SCHED_TICK_SHIFT) times per second.
 296  */
 297 #define SCHED_TICK_SHIFT        3
 298
 299 extern unsigned         sched_tick;
 300 extern uint32_t         sched_tick_interval;
 301
 302 #endif /* CONFIG_SCHED_TRADITIONAL */
 303
 304 extern uint64_t         sched_one_second_interval;
 305
 306 /* Periodic computation of various averages */
 307 extern void             compute_averages(void);
 308
 309 extern void             compute_averunnable(
 310                                         void                    *nrun);
 311
 312 extern void             compute_stack_target(
 313                                         void                    *arg);
 314
 315 extern void             compute_memory_pressure(
 316                                         void                    *arg);
 317
 318 extern void             compute_zone_gc_throttle(
 319                                         void                    *arg);
 320
 321 extern void             compute_pageout_gc_throttle(
 322                                         void                    *arg);
 323
 324 extern void             compute_pmap_gc_throttle(
 325                                         void                    *arg);
 326
 327 /*
 328  *      Conversion factor from usage
 329  *      to priority.
 330  */
 331 #if defined(CONFIG_SCHED_TRADITIONAL)
 332 extern uint32_t         sched_pri_shift;
 333 extern uint32_t         sched_fixed_shift;
 334 extern int8_t           sched_load_shifts[NRQS];
 335 #endif
 336
 337 extern int32_t          sched_poll_yield_shift;
 338 extern uint64_t         sched_safe_duration;
 339
 340 extern uint32_t         sched_run_count, sched_share_count;
 341 extern uint32_t         sched_load_average, sched_mach_factor;
 342
 343 extern uint32_t         avenrun[3], mach_factor[3];
 344
 345 extern uint64_t         max_unsafe_computation;
 346 extern uint64_t         max_poll_computation;
 347
 348 #define sched_run_incr()                        \
 349 MACRO_BEGIN                                     \
 350          hw_atomic_add(&sched_run_count, 1);    \
 351 MACRO_END
 352
 353 #define sched_run_decr()                        \
 354 MACRO_BEGIN                                     \
 355         hw_atomic_sub(&sched_run_count, 1);     \
 356 MACRO_END
 357
 358 #define sched_share_incr()                      \
 359 MACRO_BEGIN                                                                                     \
 360         (void)hw_atomic_add(&sched_share_count, 1);             \
 361 MACRO_END
 362
 363 #define sched_share_decr()                      \
 364 MACRO_BEGIN                                                                                     \
 365         (void)hw_atomic_sub(&sched_share_count, 1);             \
 366 MACRO_END
 367
 368 /*
 369  *      thread_timer_delta macro takes care of both thread timers.
 370  */
 371 #define thread_timer_delta(thread, delta)                                       \
 372 MACRO_BEGIN                                                                                                     \
 373         (delta) = (typeof(delta))timer_delta(&(thread)->system_timer,                   \
 374                                                         &(thread)->system_timer_save);  \
 375         (delta) += (typeof(delta))timer_delta(&(thread)->user_timer,                    \
 376                                                         &(thread)->user_timer_save);    \
 377 MACRO_END
 378
 379 #endif  /* _KERN_SCHED_H_ */