osfmk/kern/processor.c

   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 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 /*
  60  *      processor.c: processor and processor_set manipulation routines.
  61  */
  62
  63 #include <mach/boolean.h>
  64 #include <mach/policy.h>
  65 #include <mach/processor.h>
  66 #include <mach/processor_info.h>
  67 #include <mach/vm_param.h>
  68 #include <kern/cpu_number.h>
  69 #include <kern/host.h>
  70 #include <kern/machine.h>
  71 #include <kern/misc_protos.h>
  72 #include <kern/processor.h>
  73 #include <kern/sched.h>
  74 #include <kern/task.h>
  75 #include <kern/thread.h>
  76 #include <kern/ipc_host.h>
  77 #include <kern/ipc_tt.h>
  78 #include <ipc/ipc_port.h>
  79 #include <kern/kalloc.h>
  80
  81 /*
  82  * Exported interface
  83  */
  84 #include <mach/mach_host_server.h>
  85 #include <mach/processor_set_server.h>
  86
  87 struct processor_set    pset0;
  88 struct pset_node                pset_node0;
  89 decl_simple_lock_data(static,pset_node_lock)
  90
  91 queue_head_t                    tasks;
  92 queue_head_t                    terminated_tasks;       /* To be used ONLY for stackshot. */
  93 int                                             tasks_count;
  94 int                                             terminated_tasks_count;
  95 queue_head_t                    threads;
  96 int                                             threads_count;
  97 decl_lck_mtx_data(,tasks_threads_lock)
  98
  99 processor_t                             processor_list;
 100 unsigned int                    processor_count;
 101 static processor_t              processor_list_tail;
 102 decl_simple_lock_data(,processor_list_lock)
 103
 104 uint32_t                                processor_avail_count;
 105
 106 processor_t             master_processor;
 107 int                     master_cpu = 0;
 108 boolean_t               sched_stats_active = FALSE;
 109
 110 /* Forwards */
 111 kern_return_t   processor_set_things(
 112                 processor_set_t         pset,
 113                 mach_port_t             **thing_list,
 114                 mach_msg_type_number_t  *count,
 115                 int                     type);
 116
 117 void
 118 processor_bootstrap(void)
 119 {
 120         pset_init(&pset0, &pset_node0);
 121         pset_node0.psets = &pset0;
 122
 123         simple_lock_init(&pset_node_lock, 0);
 124
 125         queue_init(&tasks);
 126         queue_init(&terminated_tasks);
 127         queue_init(&threads);
 128
 129         simple_lock_init(&processor_list_lock, 0);
 130
 131         master_processor = cpu_to_processor(master_cpu);
 132
 133         processor_init(master_processor, master_cpu, &pset0);
 134 }
 135
 136 /*
 137  *      Initialize the given processor for the cpu
 138  *      indicated by cpu_id, and assign to the
 139  *      specified processor set.
 140  */
 141 void
 142 processor_init(
 143         processor_t                     processor,
 144         int                                     cpu_id,
 145         processor_set_t         pset)
 146 {
 147         spl_t           s;
 148
 149         if (processor != master_processor) {
 150                 /* Scheduler state deferred until sched_init() */
 151                 SCHED(processor_init)(processor);
 152         }
 153
 154         processor->state = PROCESSOR_OFF_LINE;
 155         processor->active_thread = processor->next_thread = processor->idle_thread = THREAD_NULL;
 156         processor->processor_set = pset;
 157         processor->current_pri = MINPRI;
 158         processor->current_thmode = TH_MODE_NONE;
 159         processor->cpu_id = cpu_id;
 160         timer_call_setup(&processor->quantum_timer, thread_quantum_expire, processor);
 161         processor->quantum_end = UINT64_MAX;
 162         processor->deadline = UINT64_MAX;
 163         processor->timeslice = 0;
 164         processor->processor_primary = processor; /* no SMT relationship known at this point */
 165         processor->processor_secondary = NULL;
 166         processor->is_SMT = FALSE;
 167         processor->processor_self = IP_NULL;
 168         processor_data_init(processor);
 169         processor->processor_list = NULL;
 170
 171         s = splsched();
 172         pset_lock(pset);
 173         if (pset->cpu_set_count++ == 0)
 174                 pset->cpu_set_low = pset->cpu_set_hi = cpu_id;
 175         else {
 176                 pset->cpu_set_low = (cpu_id < pset->cpu_set_low)? cpu_id: pset->cpu_set_low;
 177                 pset->cpu_set_hi = (cpu_id > pset->cpu_set_hi)? cpu_id: pset->cpu_set_hi;
 178         }
 179         pset_unlock(pset);
 180         splx(s);
 181
 182         simple_lock(&processor_list_lock);
 183         if (processor_list == NULL)
 184                 processor_list = processor;
 185         else
 186                 processor_list_tail->processor_list = processor;
 187         processor_list_tail = processor;
 188         processor_count++;
 189         simple_unlock(&processor_list_lock);
 190 }
 191
 192 void
 193 processor_set_primary(
 194         processor_t             processor,
 195         processor_t             primary)
 196 {
 197         assert(processor->processor_primary == primary || processor->processor_primary == processor);
 198         /* Re-adjust primary point for this (possibly) secondary processor */
 199         processor->processor_primary = primary;
 200
 201         assert(primary->processor_secondary == NULL || primary->processor_secondary == processor);
 202         if (primary != processor) {
 203                 /* Link primary to secondary, assumes a 2-way SMT model
 204                  * We'll need to move to a queue if any future architecture
 205                  * requires otherwise.
 206                  */
 207                 assert(processor->processor_secondary == NULL);
 208                 primary->processor_secondary = processor;
 209                 /* Mark both processors as SMT siblings */
 210                 primary->is_SMT = TRUE;
 211                 processor->is_SMT = TRUE;
 212         }
 213 }
 214
 215 processor_set_t
 216 processor_pset(
 217         processor_t     processor)
 218 {
 219         return (processor->processor_set);
 220 }
 221
 222 pset_node_t
 223 pset_node_root(void)
 224 {
 225         return &pset_node0;
 226 }
 227
 228 processor_set_t
 229 pset_create(
 230         pset_node_t                     node)
 231 {
 232 #if defined(CONFIG_SCHED_MULTIQ)
 233         /* multiq scheduler is not currently compatible with multiple psets */
 234         if (sched_groups_enabled)
 235                 return processor_pset(master_processor);
 236 #endif /* defined(CONFIG_SCHED_MULTIQ) */
 237
 238         processor_set_t         *prev, pset = kalloc(sizeof (*pset));
 239
 240         if (pset != PROCESSOR_SET_NULL) {
 241                 pset_init(pset, node);
 242
 243                 simple_lock(&pset_node_lock);
 244
 245                 prev = &node->psets;
 246                 while (*prev != PROCESSOR_SET_NULL)
 247                         prev = &(*prev)->pset_list;
 248
 249                 *prev = pset;
 250
 251                 simple_unlock(&pset_node_lock);
 252         }
 253
 254         return (pset);
 255 }
 256
 257 /*
 258  *      Initialize the given processor_set structure.
 259  */
 260 void
 261 pset_init(
 262         processor_set_t         pset,
 263         pset_node_t                     node)
 264 {
 265         if (pset != &pset0) {
 266                 /* Scheduler state deferred until sched_init() */
 267                 SCHED(pset_init)(pset);
 268         }
 269
 270         queue_init(&pset->active_queue);
 271         queue_init(&pset->idle_queue);
 272         queue_init(&pset->idle_secondary_queue);
 273         pset->online_processor_count = 0;
 274         pset->cpu_set_low = pset->cpu_set_hi = 0;
 275         pset->cpu_set_count = 0;
 276         pset->pending_AST_cpu_mask = 0;
 277         pset_lock_init(pset);
 278         pset->pset_self = IP_NULL;
 279         pset->pset_name_self = IP_NULL;
 280         pset->pset_list = PROCESSOR_SET_NULL;
 281         pset->node = node;
 282 }
 283
 284 kern_return_t
 285 processor_info_count(
 286         processor_flavor_t              flavor,
 287         mach_msg_type_number_t  *count)
 288 {
 289         switch (flavor) {
 290
 291         case PROCESSOR_BASIC_INFO:
 292                 *count = PROCESSOR_BASIC_INFO_COUNT;
 293                 break;
 294
 295         case PROCESSOR_CPU_LOAD_INFO:
 296                 *count = PROCESSOR_CPU_LOAD_INFO_COUNT;
 297                 break;
 298
 299         default:
 300                 return (cpu_info_count(flavor, count));
 301         }
 302
 303         return (KERN_SUCCESS);
 304 }
 305
 306
 307 kern_return_t
 308 processor_info(
 309         register processor_t    processor,
 310         processor_flavor_t              flavor,
 311         host_t                                  *host,
 312         processor_info_t                info,
 313         mach_msg_type_number_t  *count)
 314 {
 315         register int    cpu_id, state;
 316         kern_return_t   result;
 317
 318         if (processor == PROCESSOR_NULL)
 319                 return (KERN_INVALID_ARGUMENT);
 320
 321         cpu_id = processor->cpu_id;
 322
 323         switch (flavor) {
 324
 325         case PROCESSOR_BASIC_INFO:
 326         {
 327                 register processor_basic_info_t         basic_info;
 328
 329                 if (*count < PROCESSOR_BASIC_INFO_COUNT)
 330                         return (KERN_FAILURE);
 331
 332                 basic_info = (processor_basic_info_t) info;
 333                 basic_info->cpu_type = slot_type(cpu_id);
 334                 basic_info->cpu_subtype = slot_subtype(cpu_id);
 335                 state = processor->state;
 336                 if (state == PROCESSOR_OFF_LINE)
 337                         basic_info->running = FALSE;
 338                 else
 339                         basic_info->running = TRUE;
 340                 basic_info->slot_num = cpu_id;
 341                 if (processor == master_processor)
 342                         basic_info->is_master = TRUE;
 343                 else
 344                         basic_info->is_master = FALSE;
 345
 346                 *count = PROCESSOR_BASIC_INFO_COUNT;
 347                 *host = &realhost;
 348
 349             return (KERN_SUCCESS);
 350         }
 351
 352         case PROCESSOR_CPU_LOAD_INFO:
 353         {
 354                 processor_cpu_load_info_t       cpu_load_info;
 355                 timer_t         idle_state;
 356                 uint64_t        idle_time_snapshot1, idle_time_snapshot2;
 357                 uint64_t        idle_time_tstamp1, idle_time_tstamp2;
 358
 359                 /*
 360                  * We capture the accumulated idle time twice over
 361                  * the course of this function, as well as the timestamps
 362                  * when each were last updated. Since these are
 363                  * all done using non-atomic racy mechanisms, the
 364                  * most we can infer is whether values are stable.
 365                  * timer_grab() is the only function that can be
 366                  * used reliably on another processor's per-processor
 367                  * data.
 368                  */
 369
 370                 if (*count < PROCESSOR_CPU_LOAD_INFO_COUNT)
 371                         return (KERN_FAILURE);
 372
 373                 cpu_load_info = (processor_cpu_load_info_t) info;
 374                 if (precise_user_kernel_time) {
 375                         cpu_load_info->cpu_ticks[CPU_STATE_USER] =
 376                                                         (uint32_t)(timer_grab(&PROCESSOR_DATA(processor, user_state)) / hz_tick_interval);
 377                         cpu_load_info->cpu_ticks[CPU_STATE_SYSTEM] =
 378                                                         (uint32_t)(timer_grab(&PROCESSOR_DATA(processor, system_state)) / hz_tick_interval);
 379                 } else {
 380                         uint64_t tval = timer_grab(&PROCESSOR_DATA(processor, user_state)) +
 381                                 timer_grab(&PROCESSOR_DATA(processor, system_state));
 382
 383                         cpu_load_info->cpu_ticks[CPU_STATE_USER] = (uint32_t)(tval / hz_tick_interval);
 384                         cpu_load_info->cpu_ticks[CPU_STATE_SYSTEM] = 0;
 385                 }
 386
 387                 idle_state = &PROCESSOR_DATA(processor, idle_state);
 388                 idle_time_snapshot1 = timer_grab(idle_state);
 389                 idle_time_tstamp1 = idle_state->tstamp;
 390
 391                 /*
 392                  * Idle processors are not continually updating their
 393                  * per-processor idle timer, so it may be extremely
 394                  * out of date, resulting in an over-representation
 395                  * of non-idle time between two measurement
 396                  * intervals by e.g. top(1). If we are non-idle, or
 397                  * have evidence that the timer is being updated
 398                  * concurrently, we consider its value up-to-date.
 399                  */
 400                 if (PROCESSOR_DATA(processor, current_state) != idle_state) {
 401                         cpu_load_info->cpu_ticks[CPU_STATE_IDLE] =
 402                                                         (uint32_t)(idle_time_snapshot1 / hz_tick_interval);
 403                 } else if ((idle_time_snapshot1 != (idle_time_snapshot2 = timer_grab(idle_state))) ||
 404                                    (idle_time_tstamp1 != (idle_time_tstamp2 = idle_state->tstamp))){
 405                         /* Idle timer is being updated concurrently, second stamp is good enough */
 406                         cpu_load_info->cpu_ticks[CPU_STATE_IDLE] =
 407                                                         (uint32_t)(idle_time_snapshot2 / hz_tick_interval);
 408                 } else {
 409                         /*
 410                          * Idle timer may be very stale. Fortunately we have established
 411                          * that idle_time_snapshot1 and idle_time_tstamp1 are unchanging
 412                          */
 413                         idle_time_snapshot1 += mach_absolute_time() - idle_time_tstamp1;
 414
 415                         cpu_load_info->cpu_ticks[CPU_STATE_IDLE] =
 416                                 (uint32_t)(idle_time_snapshot1 / hz_tick_interval);
 417                 }
 418
 419                 cpu_load_info->cpu_ticks[CPU_STATE_NICE] = 0;
 420
 421             *count = PROCESSOR_CPU_LOAD_INFO_COUNT;
 422             *host = &realhost;
 423
 424             return (KERN_SUCCESS);
 425         }
 426
 427         default:
 428             result = cpu_info(flavor, cpu_id, info, count);
 429             if (result == KERN_SUCCESS)
 430                         *host = &realhost;
 431
 432             return (result);
 433         }
 434 }
 435
 436 kern_return_t
 437 processor_start(
 438         processor_t                     processor)
 439 {
 440         processor_set_t         pset;
 441         thread_t                        thread;
 442         kern_return_t           result;
 443         spl_t                           s;
 444
 445         if (processor == PROCESSOR_NULL || processor->processor_set == PROCESSOR_SET_NULL)
 446                 return (KERN_INVALID_ARGUMENT);
 447
 448         if (processor == master_processor) {
 449                 processor_t             prev;
 450
 451                 prev = thread_bind(processor);
 452                 thread_block(THREAD_CONTINUE_NULL);
 453
 454                 result = cpu_start(processor->cpu_id);
 455
 456                 thread_bind(prev);
 457
 458                 return (result);
 459         }
 460
 461         s = splsched();
 462         pset = processor->processor_set;
 463         pset_lock(pset);
 464         if (processor->state != PROCESSOR_OFF_LINE) {
 465                 pset_unlock(pset);
 466                 splx(s);
 467
 468                 return (KERN_FAILURE);
 469         }
 470
 471         processor->state = PROCESSOR_START;
 472         pset_unlock(pset);
 473         splx(s);
 474
 475         /*
 476          *      Create the idle processor thread.
 477          */
 478         if (processor->idle_thread == THREAD_NULL) {
 479                 result = idle_thread_create(processor);
 480                 if (result != KERN_SUCCESS) {
 481                         s = splsched();
 482                         pset_lock(pset);
 483                         processor->state = PROCESSOR_OFF_LINE;
 484                         pset_unlock(pset);
 485                         splx(s);
 486
 487                         return (result);
 488                 }
 489         }
 490
 491         /*
 492          *      If there is no active thread, the processor
 493          *      has never been started.  Create a dedicated
 494          *      start up thread.
 495          */
 496         if (    processor->active_thread == THREAD_NULL         &&
 497                         processor->next_thread == THREAD_NULL           ) {
 498                 result = kernel_thread_create((thread_continue_t)processor_start_thread, NULL, MAXPRI_KERNEL, &thread);
 499                 if (result != KERN_SUCCESS) {
 500                         s = splsched();
 501                         pset_lock(pset);
 502                         processor->state = PROCESSOR_OFF_LINE;
 503                         pset_unlock(pset);
 504                         splx(s);
 505
 506                         return (result);
 507                 }
 508
 509                 s = splsched();
 510                 thread_lock(thread);
 511                 thread->bound_processor = processor;
 512                 processor->next_thread = thread;
 513                 thread->state = TH_RUN;
 514                 thread_unlock(thread);
 515                 splx(s);
 516
 517                 thread_deallocate(thread);
 518         }
 519
 520         if (processor->processor_self == IP_NULL)
 521                 ipc_processor_init(processor);
 522
 523         result = cpu_start(processor->cpu_id);
 524         if (result != KERN_SUCCESS) {
 525                 s = splsched();
 526                 pset_lock(pset);
 527                 processor->state = PROCESSOR_OFF_LINE;
 528                 pset_unlock(pset);
 529                 splx(s);
 530
 531                 return (result);
 532         }
 533
 534         ipc_processor_enable(processor);
 535
 536         return (KERN_SUCCESS);
 537 }
 538
 539 kern_return_t
 540 processor_exit(
 541         processor_t     processor)
 542 {
 543         if (processor == PROCESSOR_NULL)
 544                 return(KERN_INVALID_ARGUMENT);
 545
 546         return(processor_shutdown(processor));
 547 }
 548
 549 kern_return_t
 550 processor_control(
 551         processor_t             processor,
 552         processor_info_t        info,
 553         mach_msg_type_number_t  count)
 554 {
 555         if (processor == PROCESSOR_NULL)
 556                 return(KERN_INVALID_ARGUMENT);
 557
 558         return(cpu_control(processor->cpu_id, info, count));
 559 }
 560
 561 kern_return_t
 562 processor_set_create(
 563         __unused host_t         host,
 564         __unused processor_set_t        *new_set,
 565         __unused processor_set_t        *new_name)
 566 {
 567         return(KERN_FAILURE);
 568 }
 569
 570 kern_return_t
 571 processor_set_destroy(
 572         __unused processor_set_t        pset)
 573 {
 574         return(KERN_FAILURE);
 575 }
 576
 577 kern_return_t
 578 processor_get_assignment(
 579         processor_t     processor,
 580         processor_set_t *pset)
 581 {
 582         int state;
 583
 584         if (processor == PROCESSOR_NULL)
 585                 return(KERN_INVALID_ARGUMENT);
 586
 587         state = processor->state;
 588         if (state == PROCESSOR_SHUTDOWN || state == PROCESSOR_OFF_LINE)
 589                 return(KERN_FAILURE);
 590
 591         *pset = &pset0;
 592
 593         return(KERN_SUCCESS);
 594 }
 595
 596 kern_return_t
 597 processor_set_info(
 598         processor_set_t         pset,
 599         int                     flavor,
 600         host_t                  *host,
 601         processor_set_info_t    info,
 602         mach_msg_type_number_t  *count)
 603 {
 604         if (pset == PROCESSOR_SET_NULL)
 605                 return(KERN_INVALID_ARGUMENT);
 606
 607         if (flavor == PROCESSOR_SET_BASIC_INFO) {
 608                 register processor_set_basic_info_t     basic_info;
 609
 610                 if (*count < PROCESSOR_SET_BASIC_INFO_COUNT)
 611                         return(KERN_FAILURE);
 612
 613                 basic_info = (processor_set_basic_info_t) info;
 614                 basic_info->processor_count = processor_avail_count;
 615                 basic_info->default_policy = POLICY_TIMESHARE;
 616
 617                 *count = PROCESSOR_SET_BASIC_INFO_COUNT;
 618                 *host = &realhost;
 619                 return(KERN_SUCCESS);
 620         }
 621         else if (flavor == PROCESSOR_SET_TIMESHARE_DEFAULT) {
 622                 register policy_timeshare_base_t        ts_base;
 623
 624                 if (*count < POLICY_TIMESHARE_BASE_COUNT)
 625                         return(KERN_FAILURE);
 626
 627                 ts_base = (policy_timeshare_base_t) info;
 628                 ts_base->base_priority = BASEPRI_DEFAULT;
 629
 630                 *count = POLICY_TIMESHARE_BASE_COUNT;
 631                 *host = &realhost;
 632                 return(KERN_SUCCESS);
 633         }
 634         else if (flavor == PROCESSOR_SET_FIFO_DEFAULT) {
 635                 register policy_fifo_base_t             fifo_base;
 636
 637                 if (*count < POLICY_FIFO_BASE_COUNT)
 638                         return(KERN_FAILURE);
 639
 640                 fifo_base = (policy_fifo_base_t) info;
 641                 fifo_base->base_priority = BASEPRI_DEFAULT;
 642
 643                 *count = POLICY_FIFO_BASE_COUNT;
 644                 *host = &realhost;
 645                 return(KERN_SUCCESS);
 646         }
 647         else if (flavor == PROCESSOR_SET_RR_DEFAULT) {
 648                 register policy_rr_base_t               rr_base;
 649
 650                 if (*count < POLICY_RR_BASE_COUNT)
 651                         return(KERN_FAILURE);
 652
 653                 rr_base = (policy_rr_base_t) info;
 654                 rr_base->base_priority = BASEPRI_DEFAULT;
 655                 rr_base->quantum = 1;
 656
 657                 *count = POLICY_RR_BASE_COUNT;
 658                 *host = &realhost;
 659                 return(KERN_SUCCESS);
 660         }
 661         else if (flavor == PROCESSOR_SET_TIMESHARE_LIMITS) {
 662                 register policy_timeshare_limit_t       ts_limit;
 663
 664                 if (*count < POLICY_TIMESHARE_LIMIT_COUNT)
 665                         return(KERN_FAILURE);
 666
 667                 ts_limit = (policy_timeshare_limit_t) info;
 668                 ts_limit->max_priority = MAXPRI_KERNEL;
 669
 670                 *count = POLICY_TIMESHARE_LIMIT_COUNT;
 671                 *host = &realhost;
 672                 return(KERN_SUCCESS);
 673         }
 674         else if (flavor == PROCESSOR_SET_FIFO_LIMITS) {
 675                 register policy_fifo_limit_t            fifo_limit;
 676
 677                 if (*count < POLICY_FIFO_LIMIT_COUNT)
 678                         return(KERN_FAILURE);
 679
 680                 fifo_limit = (policy_fifo_limit_t) info;
 681                 fifo_limit->max_priority = MAXPRI_KERNEL;
 682
 683                 *count = POLICY_FIFO_LIMIT_COUNT;
 684                 *host = &realhost;
 685                 return(KERN_SUCCESS);
 686         }
 687         else if (flavor == PROCESSOR_SET_RR_LIMITS) {
 688                 register policy_rr_limit_t              rr_limit;
 689
 690                 if (*count < POLICY_RR_LIMIT_COUNT)
 691                         return(KERN_FAILURE);
 692
 693                 rr_limit = (policy_rr_limit_t) info;
 694                 rr_limit->max_priority = MAXPRI_KERNEL;
 695
 696                 *count = POLICY_RR_LIMIT_COUNT;
 697                 *host = &realhost;
 698                 return(KERN_SUCCESS);
 699         }
 700         else if (flavor == PROCESSOR_SET_ENABLED_POLICIES) {
 701                 register int                            *enabled;
 702
 703                 if (*count < (sizeof(*enabled)/sizeof(int)))
 704                         return(KERN_FAILURE);
 705
 706                 enabled = (int *) info;
 707                 *enabled = POLICY_TIMESHARE | POLICY_RR | POLICY_FIFO;
 708
 709                 *count = sizeof(*enabled)/sizeof(int);
 710                 *host = &realhost;
 711                 return(KERN_SUCCESS);
 712         }
 713
 714
 715         *host = HOST_NULL;
 716         return(KERN_INVALID_ARGUMENT);
 717 }
 718
 719 /*
 720  *      processor_set_statistics
 721  *
 722  *      Returns scheduling statistics for a processor set.
 723  */
 724 kern_return_t
 725 processor_set_statistics(
 726         processor_set_t         pset,
 727         int                     flavor,
 728         processor_set_info_t    info,
 729         mach_msg_type_number_t  *count)
 730 {
 731         if (pset == PROCESSOR_SET_NULL || pset != &pset0)
 732                 return (KERN_INVALID_PROCESSOR_SET);
 733
 734         if (flavor == PROCESSOR_SET_LOAD_INFO) {
 735                 register processor_set_load_info_t     load_info;
 736
 737                 if (*count < PROCESSOR_SET_LOAD_INFO_COUNT)
 738                         return(KERN_FAILURE);
 739
 740                 load_info = (processor_set_load_info_t) info;
 741
 742                 load_info->mach_factor = sched_mach_factor;
 743                 load_info->load_average = sched_load_average;
 744
 745                 load_info->task_count = tasks_count;
 746                 load_info->thread_count = threads_count;
 747
 748                 *count = PROCESSOR_SET_LOAD_INFO_COUNT;
 749                 return(KERN_SUCCESS);
 750         }
 751
 752         return(KERN_INVALID_ARGUMENT);
 753 }
 754
 755 /*
 756  *      processor_set_max_priority:
 757  *
 758  *      Specify max priority permitted on processor set.  This affects
 759  *      newly created and assigned threads.  Optionally change existing
 760  *      ones.
 761  */
 762 kern_return_t
 763 processor_set_max_priority(
 764         __unused processor_set_t        pset,
 765         __unused int                    max_priority,
 766         __unused boolean_t              change_threads)
 767 {
 768         return (KERN_INVALID_ARGUMENT);
 769 }
 770
 771 /*
 772  *      processor_set_policy_enable:
 773  *
 774  *      Allow indicated policy on processor set.
 775  */
 776
 777 kern_return_t
 778 processor_set_policy_enable(
 779         __unused processor_set_t        pset,
 780         __unused int                    policy)
 781 {
 782         return (KERN_INVALID_ARGUMENT);
 783 }
 784
 785 /*
 786  *      processor_set_policy_disable:
 787  *
 788  *      Forbid indicated policy on processor set.  Time sharing cannot
 789  *      be forbidden.
 790  */
 791 kern_return_t
 792 processor_set_policy_disable(
 793         __unused processor_set_t        pset,
 794         __unused int                    policy,
 795         __unused boolean_t              change_threads)
 796 {
 797         return (KERN_INVALID_ARGUMENT);
 798 }
 799
 800 #define THING_TASK      0
 801 #define THING_THREAD    1
 802
 803 /*
 804  *      processor_set_things:
 805  *
 806  *      Common internals for processor_set_{threads,tasks}
 807  */
 808 kern_return_t
 809 processor_set_things(
 810         processor_set_t                 pset,
 811         mach_port_t                             **thing_list,
 812         mach_msg_type_number_t  *count,
 813         int                                             type)
 814 {
 815         unsigned int actual;    /* this many things */
 816         unsigned int maxthings;
 817         unsigned int i;
 818
 819         vm_size_t size, size_needed;
 820         void  *addr;
 821
 822         if (pset == PROCESSOR_SET_NULL || pset != &pset0)
 823                 return (KERN_INVALID_ARGUMENT);
 824
 825         size = 0;
 826         addr = NULL;
 827
 828         for (;;) {
 829                 lck_mtx_lock(&tasks_threads_lock);
 830
 831                 if (type == THING_TASK)
 832                         maxthings = tasks_count;
 833                 else
 834                         maxthings = threads_count;
 835
 836                 /* do we have the memory we need? */
 837
 838                 size_needed = maxthings * sizeof (mach_port_t);
 839                 if (size_needed <= size)
 840                         break;
 841
 842                 /* unlock and allocate more memory */
 843                 lck_mtx_unlock(&tasks_threads_lock);
 844
 845                 if (size != 0)
 846                         kfree(addr, size);
 847
 848                 assert(size_needed > 0);
 849                 size = size_needed;
 850
 851                 addr = kalloc(size);
 852                 if (addr == 0)
 853                         return (KERN_RESOURCE_SHORTAGE);
 854         }
 855
 856         /* OK, have memory and the list locked */
 857
 858         actual = 0;
 859         switch (type) {
 860
 861         case THING_TASK: {
 862                 task_t          task, *task_list = (task_t *)addr;
 863
 864                 for (task = (task_t)queue_first(&tasks);
 865                                                 !queue_end(&tasks, (queue_entry_t)task);
 866                                                                 task = (task_t)queue_next(&task->tasks)) {
 867 #if defined(SECURE_KERNEL)
 868                         if (task != kernel_task) {
 869 #endif
 870                                 task_reference_internal(task);
 871                                 task_list[actual++] = task;
 872 #if defined(SECURE_KERNEL)
 873                         }
 874 #endif
 875                 }
 876
 877                 break;
 878         }
 879
 880         case THING_THREAD: {
 881                 thread_t        thread, *thread_list = (thread_t *)addr;
 882
 883                 for (thread = (thread_t)queue_first(&threads);
 884                                                 !queue_end(&threads, (queue_entry_t)thread);
 885                                                                 thread = (thread_t)queue_next(&thread->threads)) {
 886                         thread_reference_internal(thread);
 887                         thread_list[actual++] = thread;
 888                 }
 889
 890                 break;
 891         }
 892
 893         }
 894
 895         lck_mtx_unlock(&tasks_threads_lock);
 896
 897         if (actual < maxthings)
 898                 size_needed = actual * sizeof (mach_port_t);
 899
 900         if (actual == 0) {
 901                 /* no things, so return null pointer and deallocate memory */
 902                 *thing_list = NULL;
 903                 *count = 0;
 904
 905                 if (size != 0)
 906                         kfree(addr, size);
 907         }
 908         else {
 909                 /* if we allocated too much, must copy */
 910
 911                 if (size_needed < size) {
 912                         void *newaddr;
 913
 914                         newaddr = kalloc(size_needed);
 915                         if (newaddr == 0) {
 916                                 switch (type) {
 917
 918                                 case THING_TASK: {
 919                                         task_t          *task_list = (task_t *)addr;
 920
 921                                         for (i = 0; i < actual; i++)
 922                                                 task_deallocate(task_list[i]);
 923                                         break;
 924                                 }
 925
 926                                 case THING_THREAD: {
 927                                         thread_t        *thread_list = (thread_t *)addr;
 928
 929                                         for (i = 0; i < actual; i++)
 930                                                 thread_deallocate(thread_list[i]);
 931                                         break;
 932                                 }
 933
 934                                 }
 935
 936                                 kfree(addr, size);
 937                                 return (KERN_RESOURCE_SHORTAGE);
 938                         }
 939
 940                         bcopy((void *) addr, (void *) newaddr, size_needed);
 941                         kfree(addr, size);
 942                         addr = newaddr;
 943                 }
 944
 945                 *thing_list = (mach_port_t *)addr;
 946                 *count = actual;
 947
 948                 /* do the conversion that Mig should handle */
 949
 950                 switch (type) {
 951
 952                 case THING_TASK: {
 953                         task_t          *task_list = (task_t *)addr;
 954
 955                         for (i = 0; i < actual; i++)
 956                                 (*thing_list)[i] = convert_task_to_port(task_list[i]);
 957                         break;
 958                 }
 959
 960                 case THING_THREAD: {
 961                         thread_t        *thread_list = (thread_t *)addr;
 962
 963                         for (i = 0; i < actual; i++)
 964                                 (*thing_list)[i] = convert_thread_to_port(thread_list[i]);
 965                         break;
 966                 }
 967
 968                 }
 969         }
 970
 971         return (KERN_SUCCESS);
 972 }
 973
 974
 975 /*
 976  *      processor_set_tasks:
 977  *
 978  *      List all tasks in the processor set.
 979  */
 980 kern_return_t
 981 processor_set_tasks(
 982         processor_set_t         pset,
 983         task_array_t            *task_list,
 984         mach_msg_type_number_t  *count)
 985 {
 986     return(processor_set_things(pset, (mach_port_t **)task_list, count, THING_TASK));
 987 }
 988
 989 /*
 990  *      processor_set_threads:
 991  *
 992  *      List all threads in the processor set.
 993  */
 994 #if defined(SECURE_KERNEL)
 995 kern_return_t
 996 processor_set_threads(
 997         __unused processor_set_t                pset,
 998         __unused thread_array_t         *thread_list,
 999         __unused mach_msg_type_number_t *count)
1000 {
1001     return KERN_FAILURE;
1002 }
1003 #else
1004 kern_return_t
1005 processor_set_threads(
1006         processor_set_t         pset,
1007         thread_array_t          *thread_list,
1008         mach_msg_type_number_t  *count)
1009 {
1010     return(processor_set_things(pset, (mach_port_t **)thread_list, count, THING_THREAD));
1011 }
1012 #endif
1013
1014 /*
1015  *      processor_set_policy_control
1016  *
1017  *      Controls the scheduling attributes governing the processor set.
1018  *      Allows control of enabled policies, and per-policy base and limit
1019  *      priorities.
1020  */
1021 kern_return_t
1022 processor_set_policy_control(
1023         __unused processor_set_t                pset,
1024         __unused int                            flavor,
1025         __unused processor_set_info_t   policy_info,
1026         __unused mach_msg_type_number_t count,
1027         __unused boolean_t                      change)
1028 {
1029         return (KERN_INVALID_ARGUMENT);
1030 }
1031
1032 #undef pset_deallocate
1033 void pset_deallocate(processor_set_t pset);
1034 void
1035 pset_deallocate(
1036 __unused processor_set_t        pset)
1037 {
1038         return;
1039 }
1040
1041 #undef pset_reference
1042 void pset_reference(processor_set_t pset);
1043 void
1044 pset_reference(
1045 __unused processor_set_t        pset)
1046 {
1047         return;
1048 }