osfmk/kern/startup.c

   1 /*
   2  * Copyright (c) 2000-2018 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 #include <corpses/task_corpse.h>
 105 #include <prng/random.h>
 106 #include <console/serial_protos.h>
 107 #include <vm/vm_kern.h>
 108 #include <vm/vm_init.h>
 109 #include <vm/vm_map.h>
 110 #include <vm/vm_object.h>
 111 #include <vm/vm_page.h>
 112 #include <vm/vm_pageout.h>
 113 #include <vm/vm_shared_region.h>
 114 #include <machine/pmap.h>
 115 #include <machine/commpage.h>
 116 #include <libkern/version.h>
 117 #include <sys/codesign.h>
 118 #include <sys/kdebug.h>
 119 #include <sys/random.h>
 120 #include <sys/ktrace.h>
 121 #include <libkern/section_keywords.h>
 122
 123 #include <kern/ltable.h>
 124 #include <kern/waitq.h>
 125 #include <ipc/ipc_voucher.h>
 126 #include <voucher/ipc_pthread_priority_internal.h>
 127 #include <mach/host_info.h>
 128 #include <pthread/workqueue_internal.h>
 129
 130 #if CONFIG_XNUPOST
 131 #include <tests/ktest.h>
 132 #include <tests/xnupost.h>
 133 #endif
 134
 135 #if CONFIG_ATM
 136 #include <atm/atm_internal.h>
 137 #endif
 138
 139 #if CONFIG_CSR
 140 #include <sys/csr.h>
 141 #endif
 142
 143 #include <bank/bank_internal.h>
 144
 145 #if ALTERNATE_DEBUGGER
 146 #include <arm64/alternate_debugger.h>
 147 #endif
 148
 149 #if MACH_KDP
 150 #include <kdp/kdp.h>
 151 #endif
 152
 153 #if CONFIG_MACF
 154 #include <security/mac_mach_internal.h>
 155 #if CONFIG_VNGUARD
 156 extern void vnguard_policy_init(void);
 157 #endif
 158 #endif
 159
 160 #if KPC
 161 #include <kern/kpc.h>
 162 #endif
 163
 164 #if HYPERVISOR
 165 #include <kern/hv_support.h>
 166 #endif
 167
 168 #include <san/kasan.h>
 169
 170 #if defined(__arm__) || defined(__arm64__)
 171 #include <arm/misc_protos.h> // for arm_vm_prot_finalize
 172 #endif
 173
 174 #include <i386/pmCPU.h>
 175 static void             kernel_bootstrap_thread(void);
 176
 177 static void             load_context(
 178         thread_t        thread);
 179 #if (defined(__i386__) || defined(__x86_64__)) && NCOPY_WINDOWS > 0
 180 extern void cpu_userwindow_init(int);
 181 extern void cpu_physwindow_init(int);
 182 #endif
 183
 184 #if CONFIG_ECC_LOGGING
 185 #include <kern/ecc.h>
 186 #endif
 187
 188 #if (defined(__i386__) || defined(__x86_64__)) && CONFIG_VMX
 189 #include <i386/vmx/vmx_cpu.h>
 190 #endif
 191
 192 #if CONFIG_DTRACE
 193 extern void dtrace_early_init(void);
 194 extern void sdt_early_init(void);
 195 #endif
 196
 197 // libkern/OSKextLib.cpp
 198 extern void OSKextRemoveKextBootstrap(void);
 199
 200 void scale_setup(void);
 201 extern void bsd_scale_setup(int);
 202 extern unsigned int semaphore_max;
 203 extern void stackshot_init(void);
 204 extern void ktrace_init(void);
 205 extern void oslog_init(void);
 206
 207 /*
 208  *      Running in virtual memory, on the interrupt stack.
 209  */
 210
 211 extern int serverperfmode;
 212
 213 /* size of kernel trace buffer, disabled by default */
 214 unsigned int new_nkdbufs = 0;
 215 unsigned int wake_nkdbufs = 0;
 216 unsigned int write_trace_on_panic = 0;
 217 static char trace_typefilter[64] = { 0 };
 218 unsigned int trace_wrap = 0;
 219 boolean_t trace_serial = FALSE;
 220 boolean_t early_boot_complete = FALSE;
 221
 222 /* mach leak logging */
 223 int log_leaks = 0;
 224
 225 static inline void
 226 kernel_bootstrap_log(const char *message)
 227 {
 228 //      kprintf("kernel_bootstrap: %s\n", message);
 229         kernel_debug_string_early(message);
 230 }
 231
 232 static inline void
 233 kernel_bootstrap_thread_log(const char *message)
 234 {
 235 //      kprintf("kernel_bootstrap_thread: %s\n", message);
 236         kernel_debug_string_early(message);
 237 }
 238
 239 void
 240 kernel_early_bootstrap(void)
 241 {
 242         /* serverperfmode is needed by timer setup */
 243         if (PE_parse_boot_argn("serverperfmode", &serverperfmode, sizeof(serverperfmode))) {
 244                 serverperfmode = 1;
 245         }
 246
 247 #if CONFIG_SCHED_SFI
 248         /*
 249          * Configure SFI classes
 250          */
 251         sfi_early_init();
 252 #endif
 253 }
 254
 255
 256 void
 257 kernel_bootstrap(void)
 258 {
 259         kern_return_t   result;
 260         thread_t        thread;
 261         char            namep[16];
 262
 263         printf("%s\n", version); /* log kernel version */
 264
 265         if (PE_parse_boot_argn("-l", namep, sizeof(namep))) { /* leaks logging */
 266                 log_leaks = 1;
 267         }
 268
 269         PE_parse_boot_argn("trace", &new_nkdbufs, sizeof(new_nkdbufs));
 270         PE_parse_boot_argn("trace_wake", &wake_nkdbufs, sizeof(wake_nkdbufs));
 271         PE_parse_boot_argn("trace_panic", &write_trace_on_panic, sizeof(write_trace_on_panic));
 272         PE_parse_boot_arg_str("trace_typefilter", trace_typefilter, sizeof(trace_typefilter));
 273         PE_parse_boot_argn("trace_wrap", &trace_wrap, sizeof(trace_wrap));
 274
 275         scale_setup();
 276
 277         kernel_bootstrap_log("vm_mem_bootstrap");
 278         vm_mem_bootstrap();
 279
 280         kernel_bootstrap_log("cs_init");
 281         cs_init();
 282
 283         kernel_bootstrap_log("vm_mem_init");
 284         vm_mem_init();
 285
 286         machine_info.memory_size = (uint32_t)mem_size;
 287         machine_info.max_mem = max_mem;
 288         machine_info.major_version = version_major;
 289         machine_info.minor_version = version_minor;
 290
 291         oslog_init();
 292
 293 #if KASAN
 294         kernel_bootstrap_log("kasan_late_init");
 295         kasan_late_init();
 296 #endif
 297
 298 #if CONFIG_TELEMETRY
 299         kernel_bootstrap_log("telemetry_init");
 300         telemetry_init();
 301 #endif
 302
 303 #if CONFIG_CSR
 304         kernel_bootstrap_log("csr_init");
 305         csr_init();
 306 #endif
 307
 308         if (PE_i_can_has_debugger(NULL)) {
 309                 if (PE_parse_boot_argn("-show_pointers", &namep, sizeof(namep))) {
 310                         doprnt_hide_pointers = FALSE;
 311                 }
 312                 if (PE_parse_boot_argn("-no_slto_panic", &namep, sizeof(namep))) {
 313                         extern boolean_t spinlock_timeout_panic;
 314                         spinlock_timeout_panic = FALSE;
 315                 }
 316         }
 317
 318         kernel_bootstrap_log("console_init");
 319         console_init();
 320
 321         kernel_bootstrap_log("stackshot_init");
 322         stackshot_init();
 323
 324         kernel_bootstrap_log("sched_init");
 325         sched_init();
 326
 327         kernel_bootstrap_log("ltable_bootstrap");
 328         ltable_bootstrap();
 329
 330         kernel_bootstrap_log("waitq_bootstrap");
 331         waitq_bootstrap();
 332
 333         kernel_bootstrap_log("ipc_bootstrap");
 334         ipc_bootstrap();
 335
 336 #if CONFIG_MACF
 337         kernel_bootstrap_log("mac_policy_init");
 338         mac_policy_init();
 339 #endif
 340
 341         kernel_bootstrap_log("ipc_init");
 342         ipc_init();
 343
 344         /*
 345          * As soon as the virtual memory system is up, we record
 346          * that this CPU is using the kernel pmap.
 347          */
 348         kernel_bootstrap_log("PMAP_ACTIVATE_KERNEL");
 349         PMAP_ACTIVATE_KERNEL(master_cpu);
 350
 351         kernel_bootstrap_log("mapping_free_prime");
 352         mapping_free_prime();                                           /* Load up with temporary mapping blocks */
 353
 354         kernel_bootstrap_log("machine_init");
 355         machine_init();
 356
 357         kernel_bootstrap_log("clock_init");
 358         clock_init();
 359
 360         ledger_init();
 361
 362         /*
 363          *      Initialize the IPC, task, and thread subsystems.
 364          */
 365
 366 #if CONFIG_COALITIONS
 367         kernel_bootstrap_log("coalitions_init");
 368         coalitions_init();
 369 #endif
 370
 371         kernel_bootstrap_log("task_init");
 372         task_init();
 373
 374         kernel_bootstrap_log("thread_init");
 375         thread_init();
 376
 377         kernel_bootstrap_log("restartable_init");
 378         restartable_init();
 379
 380         kernel_bootstrap_log("workq_init");
 381         workq_init();
 382
 383         kernel_bootstrap_log("turnstiles_init");
 384         turnstiles_init();
 385
 386 #if CONFIG_ATM
 387         /* Initialize the Activity Trace Resource Manager. */
 388         kernel_bootstrap_log("atm_init");
 389         atm_init();
 390 #endif
 391         kernel_bootstrap_log("mach_init_activity_id");
 392         mach_init_activity_id();
 393
 394         /* Initialize the BANK Manager. */
 395         kernel_bootstrap_log("bank_init");
 396         bank_init();
 397
 398         kernel_bootstrap_log("ipc_pthread_priority_init");
 399         ipc_pthread_priority_init();
 400
 401         /* initialize the corpse config based on boot-args */
 402         corpses_init();
 403
 404         /* initialize host_statistics */
 405         host_statistics_init();
 406
 407         /* initialize exceptions */
 408         exception_init();
 409
 410         /*
 411          *      Create a kernel thread to execute the kernel bootstrap.
 412          */
 413
 414         kernel_bootstrap_log("kernel_thread_create");
 415         result = kernel_thread_create((thread_continue_t)kernel_bootstrap_thread, NULL, MAXPRI_KERNEL, &thread);
 416
 417         if (result != KERN_SUCCESS) {
 418                 panic("kernel_bootstrap: result = %08X\n", result);
 419         }
 420
 421         /* The static init_thread is re-used as the bootstrap thread */
 422         assert(thread == current_thread());
 423
 424         /* TODO: do a proper thread_start() (without the thread_setrun()) */
 425         thread->state = TH_RUN;
 426         thread->last_made_runnable_time = mach_absolute_time();
 427         thread_set_thread_name(thread, "kernel_bootstrap_thread");
 428
 429         thread_deallocate(thread);
 430
 431         kernel_bootstrap_log("load_context - done");
 432         load_context(thread);
 433         /*NOTREACHED*/
 434 }
 435
 436 int kth_started = 0;
 437
 438 vm_offset_t vm_kernel_addrperm;
 439 vm_offset_t buf_kernel_addrperm;
 440 vm_offset_t vm_kernel_addrperm_ext;
 441 uint64_t vm_kernel_addrhash_salt;
 442 uint64_t vm_kernel_addrhash_salt_ext;
 443
 444 /*
 445  * Now running in a thread.  Kick off other services,
 446  * invoke user bootstrap, enter pageout loop.
 447  */
 448 static void
 449 kernel_bootstrap_thread(void)
 450 {
 451         processor_t             processor = current_processor();
 452
 453 #define kernel_bootstrap_thread_kprintf(x...) /* kprintf("kernel_bootstrap_thread: " x) */
 454         kernel_bootstrap_thread_log("idle_thread_create");
 455         /*
 456          * Create the idle processor thread.
 457          */
 458         idle_thread_create(processor);
 459
 460         /*
 461          * N.B. Do not stick anything else
 462          * before this point.
 463          *
 464          * Start up the scheduler services.
 465          */
 466         kernel_bootstrap_thread_log("sched_startup");
 467         sched_startup();
 468
 469         /*
 470          * Thread lifecycle maintenance (teardown, stack allocation)
 471          */
 472         kernel_bootstrap_thread_log("thread_daemon_init");
 473         thread_daemon_init();
 474
 475         /* Create kernel map entry reserve */
 476         vm_kernel_reserved_entry_init();
 477
 478         /*
 479          * Thread callout service.
 480          */
 481         kernel_bootstrap_thread_log("thread_call_initialize");
 482         thread_call_initialize();
 483
 484         /*
 485          * Remain on current processor as
 486          * additional processors come online.
 487          */
 488         kernel_bootstrap_thread_log("thread_bind");
 489         thread_bind(processor);
 490
 491         /*
 492          * Initialize ipc thread call support.
 493          */
 494         kernel_bootstrap_thread_log("ipc_thread_call_init");
 495         ipc_thread_call_init();
 496
 497         /*
 498          * Kick off memory mapping adjustments.
 499          */
 500         kernel_bootstrap_thread_log("mapping_adjust");
 501         mapping_adjust();
 502
 503         /*
 504          *      Create the clock service.
 505          */
 506         kernel_bootstrap_thread_log("clock_service_create");
 507         clock_service_create();
 508
 509         /*
 510          *      Create the device service.
 511          */
 512         device_service_create();
 513
 514         kth_started = 1;
 515
 516 #if (defined(__i386__) || defined(__x86_64__)) && NCOPY_WINDOWS > 0
 517         /*
 518          * Create and initialize the physical copy window for processor 0
 519          * This is required before starting kicking off IOKit.
 520          */
 521         cpu_physwindow_init(0);
 522 #endif
 523
 524         phys_carveout_init();
 525
 526 #if MACH_KDP
 527         kernel_bootstrap_log("kdp_init");
 528         kdp_init();
 529 #endif
 530
 531 #if ALTERNATE_DEBUGGER
 532         alternate_debugger_init();
 533 #endif
 534
 535 #if KPC
 536         kpc_init();
 537 #endif
 538
 539 #if CONFIG_ECC_LOGGING
 540         ecc_log_init();
 541 #endif
 542
 543 #if HYPERVISOR
 544         hv_support_init();
 545 #endif
 546
 547 #if CONFIG_TELEMETRY
 548         kernel_bootstrap_log("bootprofile_init");
 549         bootprofile_init();
 550 #endif
 551
 552 #if (defined(__i386__) || defined(__x86_64__)) && CONFIG_VMX
 553         vmx_init();
 554 #endif
 555
 556         kernel_bootstrap_thread_log("ktrace_init");
 557         ktrace_init();
 558
 559         kdebug_init(new_nkdbufs, trace_typefilter, trace_wrap);
 560
 561 #ifdef  MACH_BSD
 562         kernel_bootstrap_log("bsd_early_init");
 563         bsd_early_init();
 564 #endif
 565
 566 #if defined(__arm64__)
 567         ml_lockdown_init();
 568 #endif
 569
 570 #ifdef  IOKIT
 571         kernel_bootstrap_log("PE_init_iokit");
 572         PE_init_iokit();
 573 #endif
 574
 575         assert(ml_get_interrupts_enabled() == FALSE);
 576
 577         /*
 578          * Past this point, kernel subsystems that expect to operate with
 579          * interrupts or preemption enabled may begin enforcement.
 580          */
 581         early_boot_complete = TRUE;
 582
 583 #if INTERRUPT_MASKED_DEBUG
 584         // Reset interrupts masked timeout before we enable interrupts
 585         ml_spin_debug_clear_self();
 586 #endif
 587         (void) spllo();         /* Allow interruptions */
 588
 589 #if (defined(__i386__) || defined(__x86_64__)) && NCOPY_WINDOWS > 0
 590         /*
 591          * Create and initialize the copy window for processor 0
 592          * This also allocates window space for all other processors.
 593          * However, this is dependent on the number of processors - so this call
 594          * must be after IOKit has been started because IOKit performs processor
 595          * discovery.
 596          */
 597         cpu_userwindow_init(0);
 598 #endif
 599
 600         /*
 601          *      Initialize the shared region module.
 602          */
 603         vm_shared_region_init();
 604         vm_commpage_init();
 605         vm_commpage_text_init();
 606
 607 #if CONFIG_MACF
 608         kernel_bootstrap_log("mac_policy_initmach");
 609         mac_policy_initmach();
 610 #if CONFIG_VNGUARD
 611         vnguard_policy_init();
 612 #endif
 613 #endif
 614
 615 #if CONFIG_DTRACE
 616         dtrace_early_init();
 617         sdt_early_init();
 618 #endif
 619
 620
 621         /*
 622          * Get rid of segments used to bootstrap kext loading. This removes
 623          * the KLD, PRELINK symtab, LINKEDIT, and symtab segments/load commands.
 624          * Must be done prior to lockdown so that we can free (and possibly relocate)
 625          * the static KVA mappings used for the jettisoned bootstrap segments.
 626          */
 627         OSKextRemoveKextBootstrap();
 628 #if defined(__arm__) || defined(__arm64__)
 629 #if CONFIG_KERNEL_INTEGRITY
 630         machine_lockdown_preflight();
 631 #endif
 632         /*
 633          *  Finalize protections on statically mapped pages now that comm page mapping is established.
 634          */
 635         arm_vm_prot_finalize(PE_state.bootArgs);
 636 #endif
 637
 638 #if CONFIG_SCHED_SFI
 639         kernel_bootstrap_log("sfi_init");
 640         sfi_init();
 641 #endif
 642
 643         /*
 644          * Initialize the globals used for permuting kernel
 645          * addresses that may be exported to userland as tokens
 646          * using VM_KERNEL_ADDRPERM()/VM_KERNEL_ADDRPERM_EXTERNAL().
 647          * Force the random number to be odd to avoid mapping a non-zero
 648          * word-aligned address to zero via addition.
 649          * Note: at this stage we can use the cryptographically secure PRNG
 650          * rather than early_random().
 651          */
 652         read_random(&vm_kernel_addrperm, sizeof(vm_kernel_addrperm));
 653         vm_kernel_addrperm |= 1;
 654         read_random(&buf_kernel_addrperm, sizeof(buf_kernel_addrperm));
 655         buf_kernel_addrperm |= 1;
 656         read_random(&vm_kernel_addrperm_ext, sizeof(vm_kernel_addrperm_ext));
 657         vm_kernel_addrperm_ext |= 1;
 658         read_random(&vm_kernel_addrhash_salt, sizeof(vm_kernel_addrhash_salt));
 659         read_random(&vm_kernel_addrhash_salt_ext, sizeof(vm_kernel_addrhash_salt_ext));
 660
 661         vm_set_restrictions();
 662
 663
 664 #ifdef CONFIG_XNUPOST
 665         kern_return_t result = kernel_list_tests();
 666         result = kernel_do_post();
 667         if (result != KERN_SUCCESS) {
 668                 panic("kernel_do_post: Tests failed with result = 0x%08x\n", result);
 669         }
 670         kernel_bootstrap_log("kernel_do_post - done");
 671 #endif /* CONFIG_XNUPOST */
 672
 673
 674         /*
 675          *      Start the user bootstrap.
 676          */
 677 #ifdef  MACH_BSD
 678         bsd_init();
 679 #endif
 680
 681 #if defined (__x86_64__)
 682         x86_64_protect_data_const();
 683 #endif
 684
 685
 686         /*
 687          * Get rid of pages used for early boot tracing.
 688          */
 689         kdebug_free_early_buf();
 690
 691         serial_keyboard_init();         /* Start serial keyboard if wanted */
 692
 693         vm_page_init_local_q();
 694
 695         thread_bind(PROCESSOR_NULL);
 696
 697         /*
 698          * Now that all CPUs are available to run threads, this is essentially
 699          * a background thread. Take this opportunity to initialize and free
 700          * any remaining vm_pages that were delayed earlier by pmap_startup().
 701          */
 702         vm_free_delayed_pages();
 703
 704         /*
 705          *      Become the pageout daemon.
 706          */
 707         vm_pageout();
 708         /*NOTREACHED*/
 709 }
 710
 711 /*
 712  *      slave_main:
 713  *
 714  *      Load the first thread to start a processor.
 715  *      This path will also be used by the master processor
 716  *      after being offlined.
 717  */
 718 void
 719 slave_main(void *machine_param)
 720 {
 721         processor_t             processor = current_processor();
 722         thread_t                thread;
 723
 724         /*
 725          *      Use the idle processor thread if there
 726          *      is no dedicated start up thread.
 727          */
 728         if (processor->processor_offlined == true) {
 729                 /* Return to the saved processor_offline context */
 730                 assert(processor->startup_thread == THREAD_NULL);
 731
 732                 thread = processor->idle_thread;
 733                 thread->parameter = machine_param;
 734         } else if (processor->startup_thread) {
 735                 thread = processor->startup_thread;
 736                 processor->startup_thread = THREAD_NULL;
 737         } else {
 738                 thread = processor->idle_thread;
 739                 thread->continuation = processor_start_thread;
 740                 thread->parameter = machine_param;
 741         }
 742
 743         load_context(thread);
 744         /*NOTREACHED*/
 745 }
 746
 747 /*
 748  *      processor_start_thread:
 749  *
 750  *      First thread to execute on a started processor.
 751  *
 752  *      Called at splsched.
 753  */
 754 void
 755 processor_start_thread(void *machine_param,
 756     __unused wait_result_t result)
 757 {
 758         processor_t             processor = current_processor();
 759         thread_t                self = current_thread();
 760
 761         slave_machine_init(machine_param);
 762
 763         /*
 764          *      If running the idle processor thread,
 765          *      reenter the idle loop, else terminate.
 766          */
 767         if (self == processor->idle_thread) {
 768                 thread_block(idle_thread);
 769         }
 770
 771         thread_terminate(self);
 772         /*NOTREACHED*/
 773 }
 774
 775 /*
 776  *      load_context:
 777  *
 778  *      Start the first thread on a processor.
 779  *      This may be the first thread ever run on a processor, or
 780  *      it could be a processor that was previously offlined.
 781  */
 782 static void __attribute__((noreturn))
 783 load_context(
 784         thread_t                thread)
 785 {
 786         processor_t             processor = current_processor();
 787
 788
 789 #define load_context_kprintf(x...) /* kprintf("load_context: " x) */
 790
 791         load_context_kprintf("machine_set_current_thread\n");
 792         machine_set_current_thread(thread);
 793
 794         load_context_kprintf("processor_up\n");
 795
 796         PMAP_ACTIVATE_KERNEL(processor->cpu_id);
 797
 798         /*
 799          * Acquire a stack if none attached.  The panic
 800          * should never occur since the thread is expected
 801          * to have reserved stack.
 802          */
 803         load_context_kprintf("thread %p, stack %lx, stackptr %lx\n", thread,
 804             thread->kernel_stack, thread->machine.kstackptr);
 805         if (!thread->kernel_stack) {
 806                 load_context_kprintf("stack_alloc_try\n");
 807                 if (!stack_alloc_try(thread)) {
 808                         panic("load_context");
 809                 }
 810         }
 811
 812         /*
 813          * The idle processor threads are not counted as
 814          * running for load calculations.
 815          */
 816         if (!(thread->state & TH_IDLE)) {
 817                 SCHED(run_count_incr)(thread);
 818         }
 819
 820         processor->active_thread = thread;
 821         processor_state_update_explicit(processor, thread->sched_pri,
 822             SFI_CLASS_KERNEL, PSET_SMP, thread_get_perfcontrol_class(thread), THREAD_URGENCY_NONE);
 823         processor->current_is_bound = thread->bound_processor != PROCESSOR_NULL;
 824         processor->current_is_NO_SMT = false;
 825         processor->starting_pri = thread->sched_pri;
 826         processor->deadline = UINT64_MAX;
 827         thread->last_processor = processor;
 828
 829         processor_up(processor);
 830
 831         processor->last_dispatch = mach_absolute_time();
 832         timer_start(&thread->system_timer, processor->last_dispatch);
 833         PROCESSOR_DATA(processor, thread_timer) = PROCESSOR_DATA(processor, kernel_timer) = &thread->system_timer;
 834
 835         timer_start(&PROCESSOR_DATA(processor, system_state), processor->last_dispatch);
 836         PROCESSOR_DATA(processor, current_state) = &PROCESSOR_DATA(processor, system_state);
 837
 838
 839         cpu_quiescent_counter_join(processor->last_dispatch);
 840
 841         PMAP_ACTIVATE_USER(thread, processor->cpu_id);
 842
 843         load_context_kprintf("machine_load_context\n");
 844
 845 #if __arm__ || __arm64__
 846 #if __SMP__
 847         /* TODO: Should this be ordered? */
 848         thread->machine.machine_thread_flags |= MACHINE_THREAD_FLAGS_ON_CPU;
 849 #endif /* __SMP__ */
 850 #endif /* __arm__ || __arm64__ */
 851
 852         machine_load_context(thread);
 853         /*NOTREACHED*/
 854 }
 855
 856 void
 857 scale_setup()
 858 {
 859         int scale = 0;
 860 #if defined(__LP64__)
 861         typeof(task_max) task_max_base = task_max;
 862
 863         /* Raise limits for servers with >= 16G */
 864         if ((serverperfmode != 0) && ((uint64_t)sane_size >= (uint64_t)(16 * 1024 * 1024 * 1024ULL))) {
 865                 scale = (int)((uint64_t)sane_size / (uint64_t)(8 * 1024 * 1024 * 1024ULL));
 866                 /* limit to 128 G */
 867                 if (scale > 16) {
 868                         scale = 16;
 869                 }
 870                 task_max_base = 2500;
 871                 /* Raise limits for machines with >= 3GB */
 872         } else if ((uint64_t)sane_size >= (uint64_t)(3 * 1024 * 1024 * 1024ULL)) {
 873                 if ((uint64_t)sane_size < (uint64_t)(8 * 1024 * 1024 * 1024ULL)) {
 874                         scale = 2;
 875                 } else {
 876                         /* limit to 64GB */
 877                         scale = MIN(16, (int)((uint64_t)sane_size / (uint64_t)(4 * 1024 * 1024 * 1024ULL)));
 878                 }
 879         }
 880
 881         task_max = MAX(task_max, task_max_base * scale);
 882
 883         if (scale != 0) {
 884                 task_threadmax = task_max;
 885                 thread_max = task_max * 5;
 886         }
 887
 888 #endif
 889
 890         bsd_scale_setup(scale);
 891
 892         ipc_space_max = SPACE_MAX;
 893         ipc_port_max = PORT_MAX;
 894         ipc_pset_max = SET_MAX;
 895         semaphore_max = SEMAPHORE_MAX;
 896 }