osfmk/kern/startup.c

   1 /*
   2  * Copyright (c) 2000-2020 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  * NOTICE: This file was modified by McAfee Research in 2004 to introduce
  58  * support for mandatory and extensible security protections.  This notice
  59  * is included in support of clause 2.2 (b) of the Apple Public License,
  60  * Version 2.0.
  61  */
  62 /*
  63  */
  64
  65 /*
  66  *      Mach kernel startup.
  67  */
  68
  69 #include <debug.h>
  70 #include <mach_kdp.h>
  71
  72 #include <mach/boolean.h>
  73 #include <mach/machine.h>
  74 #include <mach/thread_act.h>
  75 #include <mach/task_special_ports.h>
  76 #include <mach/vm_param.h>
  77 #include <ipc/ipc_init.h>
  78 #include <kern/assert.h>
  79 #include <kern/mach_param.h>
  80 #include <kern/misc_protos.h>
  81 #include <kern/clock.h>
  82 #include <kern/coalition.h>
  83 #include <kern/cpu_number.h>
  84 #include <kern/cpu_quiesce.h>
  85 #include <kern/ledger.h>
  86 #include <kern/machine.h>
  87 #include <kern/processor.h>
  88 #include <kern/restartable.h>
  89 #include <kern/sched_prim.h>
  90 #include <kern/turnstile.h>
  91 #if CONFIG_SCHED_SFI
  92 #include <kern/sfi.h>
  93 #endif
  94 #include <kern/startup.h>
  95 #include <kern/task.h>
  96 #include <kern/thread.h>
  97 #include <kern/timer.h>
  98 #if CONFIG_TELEMETRY
  99 #include <kern/telemetry.h>
 100 #endif
 101 #include <kern/zalloc.h>
 102 #include <kern/locks.h>
 103 #include <kern/debug.h>
 104 #if KPERF
 105 #include <kperf/kperf.h>
 106 #endif /* KPERF */
 107 #include <corpses/task_corpse.h>
 108 #include <prng/random.h>
 109 #include <console/serial_protos.h>
 110 #include <vm/vm_kern.h>
 111 #include <vm/vm_init.h>
 112 #include <vm/vm_map.h>
 113 #include <vm/vm_object.h>
 114 #include <vm/vm_page.h>
 115 #include <vm/vm_pageout.h>
 116 #include <vm/vm_shared_region.h>
 117 #include <machine/pmap.h>
 118 #include <machine/commpage.h>
 119 #include <machine/machine_routines.h>
 120 #include <libkern/version.h>
 121 #include <sys/codesign.h>
 122 #include <sys/kdebug.h>
 123 #include <sys/random.h>
 124 #include <sys/ktrace.h>
 125 #include <libkern/section_keywords.h>
 126
 127 #include <kern/ltable.h>
 128 #include <kern/waitq.h>
 129 #include <ipc/ipc_voucher.h>
 130 #include <voucher/ipc_pthread_priority_internal.h>
 131 #include <mach/host_info.h>
 132 #include <pthread/workqueue_internal.h>
 133
 134 #if CONFIG_XNUPOST
 135 #include <tests/ktest.h>
 136 #include <tests/xnupost.h>
 137 #endif
 138
 139 #if CONFIG_ATM
 140 #include <atm/atm_internal.h>
 141 #endif
 142
 143 #if CONFIG_CSR
 144 #include <sys/csr.h>
 145 #endif
 146
 147 #include <bank/bank_internal.h>
 148
 149 #if ALTERNATE_DEBUGGER
 150 #include <arm64/alternate_debugger.h>
 151 #endif
 152
 153 #if MACH_KDP
 154 #include <kdp/kdp.h>
 155 #endif
 156
 157 #if CONFIG_MACF
 158 #include <security/mac_mach_internal.h>
 159 #if CONFIG_VNGUARD
 160 extern void vnguard_policy_init(void);
 161 #endif
 162 #endif
 163
 164 #if KPC
 165 #include <kern/kpc.h>
 166 #endif
 167
 168 #if HYPERVISOR
 169 #include <kern/hv_support.h>
 170 #endif
 171
 172 #include <san/kasan.h>
 173
 174 #include <i386/pmCPU.h>
 175 static void             kernel_bootstrap_thread(void);
 176
 177 static void             load_context(
 178         thread_t        thread);
 179
 180 #if CONFIG_ECC_LOGGING
 181 #include <kern/ecc.h>
 182 #endif
 183
 184 #if (defined(__i386__) || defined(__x86_64__)) && CONFIG_VMX
 185 #include <i386/vmx/vmx_cpu.h>
 186 #endif
 187
 188 #if CONFIG_DTRACE
 189 extern void dtrace_early_init(void);
 190 extern void sdt_early_init(void);
 191 #endif
 192
 193 // libkern/OSKextLib.cpp
 194 extern void OSKextRemoveKextBootstrap(void);
 195
 196 void scale_setup(void);
 197 extern void bsd_scale_setup(int);
 198 extern unsigned int semaphore_max;
 199 extern void stackshot_init(void);
 200 extern void ktrace_init(void);
 201
 202 /*
 203  *      Running in virtual memory, on the interrupt stack.
 204  */
 205
 206 extern struct startup_entry startup_entries[]
 207 __SECTION_START_SYM(STARTUP_HOOK_SEGMENT, STARTUP_HOOK_SECTION);
 208
 209 extern struct startup_entry startup_entries_end[]
 210 __SECTION_END_SYM(STARTUP_HOOK_SEGMENT, STARTUP_HOOK_SECTION);
 211
 212 static struct startup_entry *__startup_data startup_entry_cur = startup_entries;
 213
 214 SECURITY_READ_ONLY_LATE(startup_subsystem_id_t) startup_phase = STARTUP_SUB_NONE;
 215
 216 extern int serverperfmode;
 217
 218 #if DEBUG || DEVELOPMENT
 219 TUNABLE(startup_debug_t, startup_debug, "startup_debug", 0);
 220 #endif
 221
 222 /* size of kernel trace buffer, disabled by default */
 223 TUNABLE(unsigned int, new_nkdbufs, "trace", 0);
 224 TUNABLE(unsigned int, wake_nkdbufs, "trace_wake", 0);
 225 TUNABLE(unsigned int, write_trace_on_panic, "trace_panic", 0);
 226 TUNABLE(unsigned int, trace_wrap, "trace_wrap", 0);
 227
 228 /* mach leak logging */
 229 TUNABLE(int, log_leaks, "-l", 0);
 230
 231 static inline void
 232 kernel_bootstrap_log(const char *message)
 233 {
 234         if ((startup_debug & STARTUP_DEBUG_VERBOSE) &&
 235             startup_phase >= STARTUP_SUB_KPRINTF) {
 236                 kprintf("kernel_bootstrap: %s\n", message);
 237         }
 238         kernel_debug_string_early(message);
 239 }
 240
 241 static inline void
 242 kernel_bootstrap_thread_log(const char *message)
 243 {
 244         if ((startup_debug & STARTUP_DEBUG_VERBOSE) &&
 245             startup_phase >= STARTUP_SUB_KPRINTF) {
 246                 kprintf("kernel_bootstrap_thread: %s\n", message);
 247         }
 248         kernel_debug_string_early(message);
 249 }
 250
 251 extern void
 252 qsort(void *a, size_t n, size_t es, int (*cmp)(const void *, const void *));
 253
 254 __startup_func
 255 static int
 256 startup_entry_cmp(const void *e1, const void *e2)
 257 {
 258         const struct startup_entry *a = e1;
 259         const struct startup_entry *b = e2;
 260         if (a->subsystem == b->subsystem) {
 261                 if (a->rank == b->rank) {
 262                         return 0;
 263                 }
 264                 return a->rank > b->rank ? 1 : -1;
 265         }
 266         return a->subsystem > b->subsystem ? 1 : -1;
 267 }
 268
 269 __startup_func
 270 void
 271 kernel_startup_bootstrap(void)
 272 {
 273         /*
 274          * Sort the various STARTUP() entries by subsystem/rank.
 275          */
 276         size_t n = startup_entries_end - startup_entries;
 277
 278         if (n == 0) {
 279                 panic("Section %s,%s missing",
 280                     STARTUP_HOOK_SEGMENT, STARTUP_HOOK_SECTION);
 281         }
 282         if (((uintptr_t)startup_entries_end - (uintptr_t)startup_entries) %
 283             sizeof(struct startup_entry)) {
 284                 panic("Section %s,%s has invalid size",
 285                     STARTUP_HOOK_SEGMENT, STARTUP_HOOK_SECTION);
 286         }
 287
 288         qsort(startup_entries, n, sizeof(struct startup_entry), startup_entry_cmp);
 289
 290         /*
 291          * Then initialize all tunables, and early locks
 292          */
 293         kernel_startup_initialize_upto(STARTUP_SUB_LOCKS_EARLY);
 294 }
 295
 296 __startup_func
 297 extern void
 298 kernel_startup_tunable_init(const struct startup_tunable_spec *spec)
 299 {
 300         if (PE_parse_boot_argn(spec->name, spec->var_addr, spec->var_len)) {
 301                 if (spec->var_is_bool) {
 302                         /* make sure bool's are valued in {0, 1} */
 303                         *(bool *)spec->var_addr = *(uint8_t *)spec->var_addr;
 304                 }
 305         }
 306 }
 307
 308 static void
 309 kernel_startup_log(startup_subsystem_id_t subsystem)
 310 {
 311         static const char *names[] = {
 312                 [STARTUP_SUB_TUNABLES] = "tunables",
 313                 [STARTUP_SUB_LOCKS_EARLY] = "locks_early",
 314                 [STARTUP_SUB_KPRINTF] = "kprintf",
 315
 316                 [STARTUP_SUB_PMAP_STEAL] = "pmap_steal",
 317                 [STARTUP_SUB_VM_KERNEL] = "vm_kernel",
 318                 [STARTUP_SUB_KMEM] = "kmem",
 319                 [STARTUP_SUB_KMEM_ALLOC] = "kmem_alloc",
 320                 [STARTUP_SUB_ZALLOC] = "zalloc",
 321                 [STARTUP_SUB_PERCPU] = "percpu",
 322                 [STARTUP_SUB_LOCKS] = "locks",
 323
 324                 [STARTUP_SUB_CODESIGNING] = "codesigning",
 325                 [STARTUP_SUB_OSLOG] = "oslog",
 326                 [STARTUP_SUB_MACH_IPC] = "mach_ipc",
 327                 [STARTUP_SUB_EARLY_BOOT] = "early_boot",
 328
 329                 /* LOCKDOWN is special and its value won't fit here. */
 330         };
 331         static startup_subsystem_id_t logged = STARTUP_SUB_NONE;
 332
 333         if (subsystem <= logged) {
 334                 return;
 335         }
 336
 337         if (subsystem < sizeof(names) / sizeof(names[0]) && names[subsystem]) {
 338                 kernel_bootstrap_log(names[subsystem]);
 339         }
 340         logged = subsystem;
 341 }
 342
 343 __startup_func
 344 void
 345 kernel_startup_initialize_upto(startup_subsystem_id_t upto)
 346 {
 347         struct startup_entry *cur = startup_entry_cur;
 348
 349         assert(startup_phase < upto);
 350
 351         while (cur < startup_entries_end && cur->subsystem <= upto) {
 352                 if ((startup_debug & STARTUP_DEBUG_VERBOSE) &&
 353                     startup_phase >= STARTUP_SUB_KPRINTF) {
 354                         kprintf("%s[%d, rank %d]: %p(%p)\n", __func__,
 355                             cur->subsystem, cur->rank, cur->func, cur->arg);
 356                 }
 357                 startup_phase = cur->subsystem - 1;
 358                 kernel_startup_log(cur->subsystem);
 359                 cur->func(cur->arg);
 360                 startup_entry_cur = ++cur;
 361         }
 362         kernel_startup_log(upto);
 363
 364         if ((startup_debug & STARTUP_DEBUG_VERBOSE) &&
 365             upto >= STARTUP_SUB_KPRINTF) {
 366                 kprintf("%s: reached phase %d\n", __func__, upto);
 367         }
 368         startup_phase = upto;
 369 }
 370
 371 void
 372 kernel_bootstrap(void)
 373 {
 374         kern_return_t   result;
 375         thread_t        thread;
 376         char            namep[16];
 377
 378         printf("%s\n", version); /* log kernel version */
 379
 380         scale_setup();
 381
 382         kernel_bootstrap_log("vm_mem_bootstrap");
 383         vm_mem_bootstrap();
 384
 385         machine_info.memory_size = (uint32_t)mem_size;
 386 #if XNU_TARGET_OS_OSX
 387         machine_info.max_mem = max_mem_actual;
 388 #else
 389         machine_info.max_mem = max_mem;
 390 #endif /* XNU_TARGET_OS_OSX */
 391         machine_info.major_version = version_major;
 392         machine_info.minor_version = version_minor;
 393
 394         kernel_startup_initialize_upto(STARTUP_SUB_OSLOG);
 395
 396 #if KASAN
 397         kernel_bootstrap_log("kasan_late_init");
 398         kasan_late_init();
 399 #endif
 400
 401 #if CONFIG_TELEMETRY
 402         kernel_bootstrap_log("telemetry_init");
 403         telemetry_init();
 404 #endif
 405
 406         if (PE_i_can_has_debugger(NULL)) {
 407                 if (PE_parse_boot_argn("-show_pointers", &namep, sizeof(namep))) {
 408                         doprnt_hide_pointers = FALSE;
 409                 }
 410                 if (PE_parse_boot_argn("-no_slto_panic", &namep, sizeof(namep))) {
 411                         extern boolean_t spinlock_timeout_panic;
 412                         spinlock_timeout_panic = FALSE;
 413                 }
 414         }
 415
 416         kernel_bootstrap_log("console_init");
 417         console_init();
 418
 419         kernel_bootstrap_log("stackshot_init");
 420         stackshot_init();
 421
 422         kernel_bootstrap_log("sched_init");
 423         sched_init();
 424
 425         kernel_bootstrap_log("waitq_bootstrap");
 426         waitq_bootstrap();
 427
 428 #if CONFIG_MACF
 429         kernel_bootstrap_log("mac_policy_init");
 430         mac_policy_init();
 431 #endif
 432
 433         kernel_startup_initialize_upto(STARTUP_SUB_MACH_IPC);
 434
 435         /*
 436          * As soon as the virtual memory system is up, we record
 437          * that this CPU is using the kernel pmap.
 438          */
 439         kernel_bootstrap_log("PMAP_ACTIVATE_KERNEL");
 440         PMAP_ACTIVATE_KERNEL(master_cpu);
 441
 442         kernel_bootstrap_log("mapping_free_prime");
 443         mapping_free_prime();                                           /* Load up with temporary mapping blocks */
 444
 445         kernel_bootstrap_log("machine_init");
 446         machine_init();
 447
 448         kernel_bootstrap_log("thread_machine_init_template");
 449         thread_machine_init_template();
 450
 451         kernel_bootstrap_log("clock_init");
 452         clock_init();
 453
 454         /*
 455          *      Initialize the IPC, task, and thread subsystems.
 456          */
 457 #if CONFIG_THREAD_GROUPS
 458         kernel_bootstrap_log("thread_group_init");
 459         thread_group_init();
 460 #endif
 461
 462 #if CONFIG_COALITIONS
 463         kernel_bootstrap_log("coalitions_init");
 464         coalitions_init();
 465 #endif
 466
 467         kernel_bootstrap_log("task_init");
 468         task_init();
 469
 470         kernel_bootstrap_log("thread_init");
 471         thread_init();
 472
 473         kernel_bootstrap_log("restartable_init");
 474         restartable_init();
 475
 476         kernel_bootstrap_log("workq_init");
 477         workq_init();
 478
 479         kernel_bootstrap_log("turnstiles_init");
 480         turnstiles_init();
 481
 482 #if CONFIG_ATM
 483         /* Initialize the Activity Trace Resource Manager. */
 484         kernel_bootstrap_log("atm_init");
 485         atm_init();
 486 #endif
 487         kernel_bootstrap_log("mach_init_activity_id");
 488         mach_init_activity_id();
 489
 490         /* Initialize the BANK Manager. */
 491         kernel_bootstrap_log("bank_init");
 492         bank_init();
 493
 494         kernel_bootstrap_log("ipc_pthread_priority_init");
 495         ipc_pthread_priority_init();
 496
 497         /* initialize the corpse config based on boot-args */
 498         corpses_init();
 499
 500         /* initialize host_statistics */
 501         host_statistics_init();
 502
 503         /* initialize exceptions */
 504         kernel_bootstrap_log("exception_init");
 505         exception_init();
 506
 507 #if CONFIG_SCHED_SFI
 508         kernel_bootstrap_log("sfi_init");
 509         sfi_init();
 510 #endif
 511
 512         /*
 513          *      Create a kernel thread to execute the kernel bootstrap.
 514          */
 515
 516         kernel_bootstrap_log("kernel_thread_create");
 517         result = kernel_thread_create((thread_continue_t)kernel_bootstrap_thread, NULL, MAXPRI_KERNEL, &thread);
 518
 519         if (result != KERN_SUCCESS) {
 520                 panic("kernel_bootstrap: result = %08X\n", result);
 521         }
 522
 523         /* The static init_thread is re-used as the bootstrap thread */
 524         assert(thread == current_thread());
 525
 526         /* TODO: do a proper thread_start() (without the thread_setrun()) */
 527         thread->state = TH_RUN;
 528         thread->last_made_runnable_time = mach_absolute_time();
 529         thread_set_thread_name(thread, "kernel_bootstrap_thread");
 530
 531         thread_deallocate(thread);
 532
 533         kernel_bootstrap_log("load_context - done");
 534         load_context(thread);
 535         /*NOTREACHED*/
 536 }
 537
 538 SECURITY_READ_ONLY_LATE(vm_offset_t) vm_kernel_addrperm;
 539 SECURITY_READ_ONLY_LATE(vm_offset_t) buf_kernel_addrperm;
 540 SECURITY_READ_ONLY_LATE(vm_offset_t) vm_kernel_addrperm_ext;
 541 SECURITY_READ_ONLY_LATE(uint64_t) vm_kernel_addrhash_salt;
 542 SECURITY_READ_ONLY_LATE(uint64_t) vm_kernel_addrhash_salt_ext;
 543
 544 /*
 545  * Now running in a thread.  Kick off other services,
 546  * invoke user bootstrap, enter pageout loop.
 547  */
 548 static void
 549 kernel_bootstrap_thread(void)
 550 {
 551         processor_t             processor = current_processor();
 552
 553         kernel_bootstrap_thread_log("idle_thread_create");
 554         /*
 555          * Create the idle processor thread.
 556          */
 557         idle_thread_create(processor);
 558
 559         /*
 560          * N.B. Do not stick anything else
 561          * before this point.
 562          *
 563          * Start up the scheduler services.
 564          */
 565         kernel_bootstrap_thread_log("sched_startup");
 566         sched_startup();
 567
 568         /*
 569          * Thread lifecycle maintenance (teardown, stack allocation)
 570          */
 571         kernel_bootstrap_thread_log("thread_daemon_init");
 572         thread_daemon_init();
 573
 574         /* Create kernel map entry reserve */
 575         vm_kernel_reserved_entry_init();
 576
 577         /*
 578          * Thread callout service.
 579          */
 580         kernel_bootstrap_thread_log("thread_call_initialize");
 581         thread_call_initialize();
 582
 583         /*
 584          * Work interval subsystem initialization.
 585          * Needs to be done once thread calls have been initialized.
 586          */
 587         kernel_bootstrap_thread_log("work_interval_initialize");
 588         work_interval_subsystem_init();
 589
 590         /*
 591          * Remain on current processor as
 592          * additional processors come online.
 593          */
 594         kernel_bootstrap_thread_log("thread_bind");
 595         thread_bind(processor);
 596
 597         /*
 598          * Initialize ipc thread call support.
 599          */
 600         kernel_bootstrap_thread_log("ipc_thread_call_init");
 601         ipc_thread_call_init();
 602
 603         /*
 604          * Kick off memory mapping adjustments.
 605          */
 606         kernel_bootstrap_thread_log("mapping_adjust");
 607         mapping_adjust();
 608
 609         /*
 610          *      Create the clock service.
 611          */
 612         kernel_bootstrap_thread_log("clock_service_create");
 613         clock_service_create();
 614
 615         /*
 616          *      Create the device service.
 617          */
 618         device_service_create();
 619
 620         phys_carveout_init();
 621
 622 #if MACH_KDP
 623         kernel_bootstrap_log("kdp_init");
 624         kdp_init();
 625 #endif
 626
 627 #if ALTERNATE_DEBUGGER
 628         alternate_debugger_init();
 629 #endif
 630
 631 #if KPC
 632         kpc_init();
 633 #endif
 634
 635 #if HYPERVISOR
 636         kernel_bootstrap_thread_log("hv_support_init");
 637         hv_support_init();
 638 #endif
 639
 640 #if CONFIG_TELEMETRY
 641         kernel_bootstrap_log("bootprofile_init");
 642         bootprofile_init();
 643 #endif
 644
 645 #if (defined(__i386__) || defined(__x86_64__)) && CONFIG_VMX
 646         vmx_init();
 647 #endif
 648
 649         kernel_bootstrap_thread_log("ktrace_init");
 650         ktrace_init();
 651
 652         char trace_typefilter[256] = {};
 653         PE_parse_boot_arg_str("trace_typefilter", trace_typefilter,
 654             sizeof(trace_typefilter));
 655 #if KPERF
 656         kperf_init();
 657 #endif /* KPERF */
 658         kdebug_init(new_nkdbufs, trace_typefilter,
 659             (trace_wrap ? KDOPT_WRAPPING : 0) | KDOPT_ATBOOT);
 660
 661 #ifdef  MACH_BSD
 662         kernel_bootstrap_log("bsd_early_init");
 663         bsd_early_init();
 664 #endif
 665
 666 #ifdef  IOKIT
 667         kernel_bootstrap_log("PE_init_iokit");
 668         PE_init_iokit();
 669 #endif
 670
 671         assert(ml_get_interrupts_enabled() == FALSE);
 672
 673         /*
 674          * Past this point, kernel subsystems that expect to operate with
 675          * interrupts or preemption enabled may begin enforcement.
 676          */
 677         kernel_startup_initialize_upto(STARTUP_SUB_EARLY_BOOT);
 678
 679 #if INTERRUPT_MASKED_DEBUG
 680         // Reset interrupts masked timeout before we enable interrupts
 681         ml_spin_debug_clear_self();
 682 #endif
 683         (void) spllo();         /* Allow interruptions */
 684
 685         /*
 686          * This will start displaying progress to the user, start as early as possible
 687          */
 688         initialize_screen(NULL, kPEAcquireScreen);
 689
 690         /*
 691          *      Initialize the shared region module.
 692          */
 693         vm_commpage_init();
 694         vm_commpage_text_init();
 695
 696 #if CONFIG_MACF
 697         kernel_bootstrap_log("mac_policy_initmach");
 698         mac_policy_initmach();
 699 #if CONFIG_VNGUARD
 700         kernel_bootstrap_log("vnguard_policy_init");
 701         vnguard_policy_init();
 702 #endif
 703 #endif
 704
 705 #if CONFIG_DTRACE
 706         kernel_bootstrap_log("dtrace_early_init");
 707         dtrace_early_init();
 708         sdt_early_init();
 709 #endif
 710
 711
 712         kernel_startup_initialize_upto(STARTUP_SUB_LOCKDOWN);
 713
 714         /*
 715          * Get rid of segments used to bootstrap kext loading. This removes
 716          * the KLD, PRELINK symtab, LINKEDIT, and symtab segments/load commands.
 717          * Must be done prior to lockdown so that we can free (and possibly relocate)
 718          * the static KVA mappings used for the jettisoned bootstrap segments.
 719          */
 720         kernel_bootstrap_log("OSKextRemoveKextBootstrap");
 721         OSKextRemoveKextBootstrap();
 722
 723         /*
 724          * Initialize the globals used for permuting kernel
 725          * addresses that may be exported to userland as tokens
 726          * using VM_KERNEL_ADDRPERM()/VM_KERNEL_ADDRPERM_EXTERNAL().
 727          * Force the random number to be odd to avoid mapping a non-zero
 728          * word-aligned address to zero via addition.
 729          * Note: at this stage we can use the cryptographically secure PRNG
 730          * rather than early_random().
 731          */
 732         read_random(&vm_kernel_addrperm, sizeof(vm_kernel_addrperm));
 733         vm_kernel_addrperm |= 1;
 734         read_random(&buf_kernel_addrperm, sizeof(buf_kernel_addrperm));
 735         buf_kernel_addrperm |= 1;
 736         read_random(&vm_kernel_addrperm_ext, sizeof(vm_kernel_addrperm_ext));
 737         vm_kernel_addrperm_ext |= 1;
 738         read_random(&vm_kernel_addrhash_salt, sizeof(vm_kernel_addrhash_salt));
 739         read_random(&vm_kernel_addrhash_salt_ext, sizeof(vm_kernel_addrhash_salt_ext));
 740
 741         /* No changes to kernel text and rodata beyond this point. */
 742         kernel_bootstrap_log("machine_lockdown");
 743         machine_lockdown();
 744
 745 #ifdef  IOKIT
 746         kernel_bootstrap_log("PE_lockdown_iokit");
 747         PE_lockdown_iokit();
 748 #endif
 749         /*
 750          * max_cpus must be nailed down by the time PE_lockdown_iokit() finishes,
 751          * at the latest
 752          */
 753         vm_set_restrictions(machine_info.max_cpus);
 754
 755 #ifdef CONFIG_XNUPOST
 756         kern_return_t result = kernel_list_tests();
 757         result = kernel_do_post();
 758         if (result != KERN_SUCCESS) {
 759                 panic("kernel_do_post: Tests failed with result = 0x%08x\n", result);
 760         }
 761         kernel_bootstrap_log("kernel_do_post - done");
 762 #endif /* CONFIG_XNUPOST */
 763
 764
 765 #if KPERF
 766         kperf_init_early();
 767 #endif
 768
 769         /*
 770          *      Start the user bootstrap.
 771          */
 772 #ifdef  MACH_BSD
 773         bsd_init();
 774 #endif
 775
 776
 777         /*
 778          * Get rid of pages used for early boot tracing.
 779          */
 780         kdebug_free_early_buf();
 781
 782         serial_keyboard_init();         /* Start serial keyboard if wanted */
 783
 784         vm_page_init_local_q(machine_info.max_cpus);
 785
 786         thread_bind(PROCESSOR_NULL);
 787
 788         /*
 789          * Now that all CPUs are available to run threads, this is essentially
 790          * a background thread. Take this opportunity to initialize and free
 791          * any remaining vm_pages that were delayed earlier by pmap_startup().
 792          */
 793         vm_free_delayed_pages();
 794
 795         /*
 796          *      Become the pageout daemon.
 797          */
 798         vm_pageout();
 799         /*NOTREACHED*/
 800 }
 801
 802 /*
 803  *      slave_main:
 804  *
 805  *      Load the first thread to start a processor.
 806  *      This path will also be used by the master processor
 807  *      after being offlined.
 808  */
 809 void
 810 slave_main(void *machine_param)
 811 {
 812         processor_t             processor = current_processor();
 813         thread_t                thread;
 814
 815         /*
 816          *      Use the idle processor thread if there
 817          *      is no dedicated start up thread.
 818          */
 819         if (processor->processor_offlined == true) {
 820                 /* Return to the saved processor_offline context */
 821                 assert(processor->startup_thread == THREAD_NULL);
 822
 823                 thread = processor->idle_thread;
 824                 thread->parameter = machine_param;
 825         } else if (processor->startup_thread) {
 826                 thread = processor->startup_thread;
 827                 processor->startup_thread = THREAD_NULL;
 828         } else {
 829                 thread = processor->idle_thread;
 830                 thread->continuation = processor_start_thread;
 831                 thread->parameter = machine_param;
 832         }
 833
 834         load_context(thread);
 835         /*NOTREACHED*/
 836 }
 837
 838 /*
 839  *      processor_start_thread:
 840  *
 841  *      First thread to execute on a started processor.
 842  *
 843  *      Called at splsched.
 844  */
 845 void
 846 processor_start_thread(void *machine_param,
 847     __unused wait_result_t result)
 848 {
 849         processor_t             processor = current_processor();
 850         thread_t                self = current_thread();
 851
 852         slave_machine_init(machine_param);
 853
 854         /*
 855          *      If running the idle processor thread,
 856          *      reenter the idle loop, else terminate.
 857          */
 858         if (self == processor->idle_thread) {
 859                 thread_block(idle_thread);
 860         }
 861
 862         thread_terminate(self);
 863         /*NOTREACHED*/
 864 }
 865
 866 /*
 867  *      load_context:
 868  *
 869  *      Start the first thread on a processor.
 870  *      This may be the first thread ever run on a processor, or
 871  *      it could be a processor that was previously offlined.
 872  */
 873 static void __attribute__((noreturn))
 874 load_context(
 875         thread_t                thread)
 876 {
 877         processor_t             processor = current_processor();
 878
 879
 880 #define load_context_kprintf(x...) /* kprintf("load_context: " x) */
 881
 882         load_context_kprintf("machine_set_current_thread\n");
 883         machine_set_current_thread(thread);
 884
 885         load_context_kprintf("processor_up\n");
 886
 887         PMAP_ACTIVATE_KERNEL(processor->cpu_id);
 888
 889         /*
 890          * Acquire a stack if none attached.  The panic
 891          * should never occur since the thread is expected
 892          * to have reserved stack.
 893          */
 894         load_context_kprintf("thread %p, stack %lx, stackptr %lx\n", thread,
 895             thread->kernel_stack, thread->machine.kstackptr);
 896         if (!thread->kernel_stack) {
 897                 load_context_kprintf("stack_alloc_try\n");
 898                 if (!stack_alloc_try(thread)) {
 899                         panic("load_context");
 900                 }
 901         }
 902
 903         /*
 904          * The idle processor threads are not counted as
 905          * running for load calculations.
 906          */
 907         if (!(thread->state & TH_IDLE)) {
 908                 SCHED(run_count_incr)(thread);
 909         }
 910
 911         processor->active_thread = thread;
 912         processor_state_update_explicit(processor, thread->sched_pri,
 913             SFI_CLASS_KERNEL, PSET_SMP, thread_get_perfcontrol_class(thread), THREAD_URGENCY_NONE,
 914             ((thread->state & TH_IDLE) || (thread->bound_processor != PROCESSOR_NULL)) ? TH_BUCKET_SCHED_MAX : thread->th_sched_bucket);
 915         processor->current_is_bound = thread->bound_processor != PROCESSOR_NULL;
 916         processor->current_is_NO_SMT = false;
 917 #if CONFIG_THREAD_GROUPS
 918         processor->current_thread_group = thread_group_get(thread);
 919 #endif
 920         processor->starting_pri = thread->sched_pri;
 921         processor->deadline = UINT64_MAX;
 922         thread->last_processor = processor;
 923         processor_up(processor);
 924         processor->last_dispatch = mach_absolute_time();
 925         timer_start(&thread->system_timer, processor->last_dispatch);
 926         processor->thread_timer = processor->kernel_timer = &thread->system_timer;
 927
 928         timer_start(&processor->system_state, processor->last_dispatch);
 929         processor->current_state = &processor->system_state;
 930
 931
 932         cpu_quiescent_counter_join(processor->last_dispatch);
 933
 934         PMAP_ACTIVATE_USER(thread, processor->cpu_id);
 935
 936         load_context_kprintf("machine_load_context\n");
 937
 938         machine_load_context(thread);
 939         /*NOTREACHED*/
 940 }
 941
 942 void
 943 scale_setup(void)
 944 {
 945         int scale = 0;
 946 #if defined(__LP64__)
 947         typeof(task_max) task_max_base = task_max;
 948
 949         /* Raise limits for servers with >= 16G */
 950         if ((serverperfmode != 0) && ((uint64_t)max_mem_actual >= (uint64_t)(16 * 1024 * 1024 * 1024ULL))) {
 951                 scale = (int)((uint64_t)sane_size / (uint64_t)(8 * 1024 * 1024 * 1024ULL));
 952                 /* limit to 128 G */
 953                 if (scale > 16) {
 954                         scale = 16;
 955                 }
 956                 task_max_base = 2500;
 957                 /* Raise limits for machines with >= 3GB */
 958         } else if ((uint64_t)max_mem_actual >= (uint64_t)(3 * 1024 * 1024 * 1024ULL)) {
 959                 if ((uint64_t)max_mem_actual < (uint64_t)(8 * 1024 * 1024 * 1024ULL)) {
 960                         scale = 2;
 961                 } else {
 962                         /* limit to 64GB */
 963                         scale = MIN(16, (int)((uint64_t)max_mem_actual / (uint64_t)(4 * 1024 * 1024 * 1024ULL)));
 964                 }
 965         }
 966
 967         task_max = MAX(task_max, task_max_base * scale);
 968
 969         if (scale != 0) {
 970                 task_threadmax = task_max;
 971                 thread_max = task_max * 5;
 972         }
 973
 974 #endif
 975
 976         bsd_scale_setup(scale);
 977 }