osfmk/kern/startup.c

   1 /*
   2  * Copyright (c) 2000-2010 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 <xpr_debug.h>
  71 #include <mach_kdp.h>
  72
  73 #include <mach/boolean.h>
  74 #include <mach/machine.h>
  75 #include <mach/thread_act.h>
  76 #include <mach/task_special_ports.h>
  77 #include <mach/vm_param.h>
  78 #include <ipc/ipc_init.h>
  79 #include <kern/assert.h>
  80 #include <kern/mach_param.h>
  81 #include <kern/misc_protos.h>
  82 #include <kern/clock.h>
  83 #include <kern/coalition.h>
  84 #include <kern/cpu_number.h>
  85 #include <kern/cpu_quiesce.h>
  86 #include <kern/ledger.h>
  87 #include <kern/machine.h>
  88 #include <kern/processor.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/xpr.h>
 102 #include <kern/zalloc.h>
 103 #include <kern/locks.h>
 104 #include <kern/debug.h>
 105 #include <corpses/task_corpse.h>
 106 #include <prng/random.h>
 107 #include <console/serial_protos.h>
 108 #include <vm/vm_kern.h>
 109 #include <vm/vm_init.h>
 110 #include <vm/vm_map.h>
 111 #include <vm/vm_object.h>
 112 #include <vm/vm_page.h>
 113 #include <vm/vm_pageout.h>
 114 #include <vm/vm_shared_region.h>
 115 #include <machine/pmap.h>
 116 #include <machine/commpage.h>
 117 #include <libkern/version.h>
 118 #include <sys/codesign.h>
 119 #include <sys/kdebug.h>
 120 #include <sys/random.h>
 121 #include <sys/ktrace.h>
 122 #include <libkern/section_keywords.h>
 123
 124 #include <kern/ltable.h>
 125 #include <kern/waitq.h>
 126 #include <ipc/ipc_voucher.h>
 127 #include <voucher/ipc_pthread_priority_internal.h>
 128 #include <mach/host_info.h>
 129 #include <pthread/workqueue_internal.h>
 130
 131 #if CONFIG_XNUPOST
 132 #include <tests/ktest.h>
 133 #include <tests/xnupost.h>
 134 #endif
 135
 136 #if CONFIG_ATM
 137 #include <atm/atm_internal.h>
 138 #endif
 139
 140 #if CONFIG_CSR
 141 #include <sys/csr.h>
 142 #endif
 143
 144 #include <bank/bank_internal.h>
 145
 146 #if ALTERNATE_DEBUGGER
 147 #include <arm64/alternate_debugger.h>
 148 #endif
 149
 150 #if MACH_KDP
 151 #include <kdp/kdp.h>
 152 #endif
 153
 154 #if CONFIG_MACF
 155 #include <security/mac_mach_internal.h>
 156 #if CONFIG_VNGUARD
 157 extern void vnguard_policy_init(void);
 158 #endif
 159 #endif
 160
 161 #if KPC
 162 #include <kern/kpc.h>
 163 #endif
 164
 165 #if HYPERVISOR
 166 #include <kern/hv_support.h>
 167 #endif
 168
 169 #include <san/kasan.h>
 170
 171 #if defined(__arm__) || defined(__arm64__)
 172 #include <arm/misc_protos.h> // for arm_vm_prot_finalize
 173 #endif
 174
 175 #include <i386/pmCPU.h>
 176 static void             kernel_bootstrap_thread(void);
 177
 178 static void             load_context(
 179                                         thread_t        thread);
 180 #if (defined(__i386__) || defined(__x86_64__)) && NCOPY_WINDOWS > 0
 181 extern void cpu_userwindow_init(int);
 182 extern void cpu_physwindow_init(int);
 183 #endif
 184
 185 #if CONFIG_ECC_LOGGING
 186 #include <kern/ecc.h>
 187 #endif
 188
 189 #if (defined(__i386__) || defined(__x86_64__)) && CONFIG_VMX
 190 #include <i386/vmx/vmx_cpu.h>
 191 #endif
 192
 193 #if CONFIG_DTRACE
 194 extern void dtrace_early_init(void);
 195 extern void sdt_early_init(void);
 196 #endif
 197
 198 // libkern/OSKextLib.cpp
 199 extern void OSKextRemoveKextBootstrap(void);
 200
 201 void scale_setup(void);
 202 extern void bsd_scale_setup(int);
 203 extern unsigned int semaphore_max;
 204 extern void stackshot_init(void);
 205 extern void ktrace_init(void);
 206 extern void oslog_init(void);
 207
 208 /*
 209  *      Running in virtual memory, on the interrupt stack.
 210  */
 211
 212 extern int serverperfmode;
 213
 214 /* size of kernel trace buffer, disabled by default */
 215 unsigned int new_nkdbufs = 0;
 216 unsigned int wake_nkdbufs = 0;
 217 unsigned int write_trace_on_panic = 0;
 218 static char trace_typefilter[64] = { 0 };
 219 unsigned int trace_wrap = 0;
 220 boolean_t trace_serial = FALSE;
 221 boolean_t early_boot_complete = FALSE;
 222
 223 /* physically contiguous carveouts */
 224 SECURITY_READ_ONLY_LATE(uintptr_t) phys_carveout = 0;
 225 SECURITY_READ_ONLY_LATE(uintptr_t) phys_carveout_pa = 0;
 226 SECURITY_READ_ONLY_LATE(size_t) phys_carveout_size = 0;
 227
 228 /* mach leak logging */
 229 int log_leaks = 0;
 230
 231 static inline void
 232 kernel_bootstrap_log(const char *message)
 233 {
 234 //      kprintf("kernel_bootstrap: %s\n", message);
 235         kernel_debug_string_early(message);
 236 }
 237
 238 static inline void
 239 kernel_bootstrap_thread_log(const char *message)
 240 {
 241 //      kprintf("kernel_bootstrap_thread: %s\n", message);
 242         kernel_debug_string_early(message);
 243 }
 244
 245 void
 246 kernel_early_bootstrap(void)
 247 {
 248         /* serverperfmode is needed by timer setup */
 249         if (PE_parse_boot_argn("serverperfmode", &serverperfmode, sizeof (serverperfmode))) {
 250                 serverperfmode = 1;
 251         }
 252
 253         lck_mod_init();
 254
 255         /*
 256          * Initialize the timer callout world
 257          */
 258         timer_call_init();
 259
 260 #if CONFIG_SCHED_SFI
 261         /*
 262          * Configure SFI classes
 263          */
 264         sfi_early_init();
 265 #endif
 266 }
 267
 268 extern boolean_t IORamDiskBSDRoot(void);
 269 extern kern_return_t cpm_preallocate_early(void);
 270
 271 void
 272 kernel_bootstrap(void)
 273 {
 274         kern_return_t   result;
 275         thread_t        thread;
 276         char            namep[16];
 277
 278         printf("%s\n", version); /* log kernel version */
 279
 280         if (PE_parse_boot_argn("-l", namep, sizeof (namep))) /* leaks logging */
 281                 log_leaks = 1;
 282
 283         PE_parse_boot_argn("trace", &new_nkdbufs, sizeof (new_nkdbufs));
 284         PE_parse_boot_argn("trace_wake", &wake_nkdbufs, sizeof (wake_nkdbufs));
 285         PE_parse_boot_argn("trace_panic", &write_trace_on_panic, sizeof(write_trace_on_panic));
 286         PE_parse_boot_arg_str("trace_typefilter", trace_typefilter, sizeof(trace_typefilter));
 287         PE_parse_boot_argn("trace_wrap", &trace_wrap, sizeof(trace_wrap));
 288
 289         scale_setup();
 290
 291         kernel_bootstrap_log("vm_mem_bootstrap");
 292         vm_mem_bootstrap();
 293
 294         kernel_bootstrap_log("cs_init");
 295         cs_init();
 296
 297         kernel_bootstrap_log("vm_mem_init");
 298         vm_mem_init();
 299
 300         machine_info.memory_size = (uint32_t)mem_size;
 301         machine_info.max_mem = max_mem;
 302         machine_info.major_version = version_major;
 303         machine_info.minor_version = version_minor;
 304
 305         oslog_init();
 306
 307 #if KASAN
 308         kernel_bootstrap_log("kasan_late_init");
 309         kasan_late_init();
 310 #endif
 311
 312 #if CONFIG_TELEMETRY
 313         kernel_bootstrap_log("telemetry_init");
 314         telemetry_init();
 315 #endif
 316
 317 #if CONFIG_CSR
 318         kernel_bootstrap_log("csr_init");
 319         csr_init();
 320 #endif
 321
 322         if (PE_i_can_has_debugger(NULL)) {
 323                 if (PE_parse_boot_argn("-show_pointers", &namep, sizeof(namep))) {
 324                         doprnt_hide_pointers = FALSE;
 325                 }
 326                 if (PE_parse_boot_argn("-no_slto_panic", &namep, sizeof(namep))) {
 327                         extern boolean_t spinlock_timeout_panic;
 328                         spinlock_timeout_panic = FALSE;
 329                 }
 330         }
 331
 332         kernel_bootstrap_log("console_init");
 333         console_init();
 334
 335         kernel_bootstrap_log("stackshot_init");
 336         stackshot_init();
 337
 338         kernel_bootstrap_log("sched_init");
 339         sched_init();
 340
 341         kernel_bootstrap_log("ltable_bootstrap");
 342         ltable_bootstrap();
 343
 344         kernel_bootstrap_log("waitq_bootstrap");
 345         waitq_bootstrap();
 346
 347         kernel_bootstrap_log("ipc_bootstrap");
 348         ipc_bootstrap();
 349
 350 #if CONFIG_MACF
 351         kernel_bootstrap_log("mac_policy_init");
 352         mac_policy_init();
 353 #endif
 354
 355         kernel_bootstrap_log("ipc_init");
 356         ipc_init();
 357
 358         /*
 359          * As soon as the virtual memory system is up, we record
 360          * that this CPU is using the kernel pmap.
 361          */
 362         kernel_bootstrap_log("PMAP_ACTIVATE_KERNEL");
 363         PMAP_ACTIVATE_KERNEL(master_cpu);
 364
 365         kernel_bootstrap_log("mapping_free_prime");
 366         mapping_free_prime();                                           /* Load up with temporary mapping blocks */
 367
 368         kernel_bootstrap_log("machine_init");
 369         machine_init();
 370
 371         kernel_bootstrap_log("clock_init");
 372         clock_init();
 373
 374         ledger_init();
 375
 376         /*
 377          *      Initialize the IPC, task, and thread subsystems.
 378          */
 379
 380 #if CONFIG_COALITIONS
 381         kernel_bootstrap_log("coalitions_init");
 382         coalitions_init();
 383 #endif
 384
 385         kernel_bootstrap_log("task_init");
 386         task_init();
 387
 388         kernel_bootstrap_log("thread_init");
 389         thread_init();
 390
 391         kernel_bootstrap_log("workq_init");
 392         workq_init();
 393
 394         kernel_bootstrap_log("turnstiles_init");
 395         turnstiles_init();
 396
 397 #if CONFIG_ATM
 398         /* Initialize the Activity Trace Resource Manager. */
 399         kernel_bootstrap_log("atm_init");
 400         atm_init();
 401 #endif
 402         kernel_bootstrap_log("mach_init_activity_id");
 403         mach_init_activity_id();
 404
 405         /* Initialize the BANK Manager. */
 406         kernel_bootstrap_log("bank_init");
 407         bank_init();
 408
 409         kernel_bootstrap_log("ipc_pthread_priority_init");
 410         ipc_pthread_priority_init();
 411
 412         /* initialize the corpse config based on boot-args */
 413         corpses_init();
 414
 415         /* initialize host_statistics */
 416         host_statistics_init();
 417
 418         /*
 419          *      Create a kernel thread to execute the kernel bootstrap.
 420          */
 421         kernel_bootstrap_log("kernel_thread_create");
 422         result = kernel_thread_create((thread_continue_t)kernel_bootstrap_thread, NULL, MAXPRI_KERNEL, &thread);
 423
 424         if (result != KERN_SUCCESS) panic("kernel_bootstrap: result = %08X\n", result);
 425
 426         thread->state = TH_RUN;
 427         thread->last_made_runnable_time = mach_absolute_time();
 428         thread_deallocate(thread);
 429
 430         kernel_bootstrap_log("load_context - done");
 431         load_context(thread);
 432         /*NOTREACHED*/
 433 }
 434
 435 int kth_started = 0;
 436
 437 vm_offset_t vm_kernel_addrperm;
 438 vm_offset_t buf_kernel_addrperm;
 439 vm_offset_t vm_kernel_addrperm_ext;
 440 uint64_t vm_kernel_addrhash_salt;
 441 uint64_t vm_kernel_addrhash_salt_ext;
 442
 443 /*
 444  * Now running in a thread.  Kick off other services,
 445  * invoke user bootstrap, enter pageout loop.
 446  */
 447 static void
 448 kernel_bootstrap_thread(void)
 449 {
 450         processor_t             processor = current_processor();
 451
 452 #define kernel_bootstrap_thread_kprintf(x...) /* kprintf("kernel_bootstrap_thread: " x) */
 453         kernel_bootstrap_thread_log("idle_thread_create");
 454         /*
 455          * Create the idle processor thread.
 456          */
 457         idle_thread_create(processor);
 458
 459         /*
 460          * N.B. Do not stick anything else
 461          * before this point.
 462          *
 463          * Start up the scheduler services.
 464          */
 465         kernel_bootstrap_thread_log("sched_startup");
 466         sched_startup();
 467
 468         /*
 469          * Thread lifecycle maintenance (teardown, stack allocation)
 470          */
 471         kernel_bootstrap_thread_log("thread_daemon_init");
 472         thread_daemon_init();
 473
 474         /* Create kernel map entry reserve */
 475         vm_kernel_reserved_entry_init();
 476
 477         /*
 478          * Thread callout service.
 479          */
 480         kernel_bootstrap_thread_log("thread_call_initialize");
 481         thread_call_initialize();
 482
 483         /*
 484          * Remain on current processor as
 485          * additional processors come online.
 486          */
 487         kernel_bootstrap_thread_log("thread_bind");
 488         thread_bind(processor);
 489
 490 #if __arm64__
 491         if (IORamDiskBSDRoot()) {
 492                 cpm_preallocate_early();
 493         }
 494 #endif /* __arm64__ */
 495
 496         /*
 497          * Initialize ipc thread call support.
 498          */
 499         kernel_bootstrap_thread_log("ipc_thread_call_init");
 500         ipc_thread_call_init();
 501
 502         /*
 503          * Kick off memory mapping adjustments.
 504          */
 505         kernel_bootstrap_thread_log("mapping_adjust");
 506         mapping_adjust();
 507
 508         /*
 509          *      Create the clock service.
 510          */
 511         kernel_bootstrap_thread_log("clock_service_create");
 512         clock_service_create();
 513
 514         /*
 515          *      Create the device service.
 516          */
 517         device_service_create();
 518
 519         kth_started = 1;
 520
 521 #if (defined(__i386__) || defined(__x86_64__)) && NCOPY_WINDOWS > 0
 522         /*
 523          * Create and initialize the physical copy window for processor 0
 524          * This is required before starting kicking off  IOKit.
 525          */
 526         cpu_physwindow_init(0);
 527 #endif
 528
 529         if (PE_i_can_has_debugger(NULL)) {
 530                 unsigned int phys_carveout_mb = 0;
 531                 if (PE_parse_boot_argn("phys_carveout_mb", &phys_carveout_mb,
 532                                 sizeof(phys_carveout_mb)) && phys_carveout_mb > 0) {
 533                         phys_carveout_size = phys_carveout_mb * 1024 * 1024;
 534                         kern_return_t kr = kmem_alloc_contig(kernel_map,
 535                                         (vm_offset_t *)&phys_carveout, phys_carveout_size,
 536                                         VM_MAP_PAGE_MASK(kernel_map), 0, 0, KMA_NOPAGEWAIT,
 537                                         VM_KERN_MEMORY_DIAG);
 538                         if (kr != KERN_SUCCESS) {
 539                                 kprintf("failed to allocate %uMB for phys_carveout_mb: %u\n",
 540                                                 phys_carveout_mb, (unsigned int)kr);
 541                         } else {
 542                                 phys_carveout_pa = kvtophys((vm_offset_t)phys_carveout);
 543                         }
 544                 }
 545         }
 546
 547 #if MACH_KDP
 548         kernel_bootstrap_log("kdp_init");
 549         kdp_init();
 550 #endif
 551
 552 #if ALTERNATE_DEBUGGER
 553         alternate_debugger_init();
 554 #endif
 555
 556 #if KPC
 557         kpc_init();
 558 #endif
 559
 560 #if CONFIG_ECC_LOGGING
 561         ecc_log_init();
 562 #endif
 563
 564 #if HYPERVISOR
 565         hv_support_init();
 566 #endif
 567
 568 #if CONFIG_TELEMETRY
 569         kernel_bootstrap_log("bootprofile_init");
 570         bootprofile_init();
 571 #endif
 572
 573 #if (defined(__i386__) || defined(__x86_64__)) && CONFIG_VMX
 574         vmx_init();
 575 #endif
 576
 577         kernel_bootstrap_thread_log("ktrace_init");
 578         ktrace_init();
 579
 580         kdebug_init(new_nkdbufs, trace_typefilter, trace_wrap);
 581
 582 #ifdef  MACH_BSD
 583         kernel_bootstrap_log("bsd_early_init");
 584         bsd_early_init();
 585 #endif
 586
 587 #if defined(__arm64__)
 588         ml_lockdown_init();
 589 #endif
 590
 591 #ifdef  IOKIT
 592         kernel_bootstrap_log("PE_init_iokit");
 593         PE_init_iokit();
 594 #endif
 595
 596         assert(ml_get_interrupts_enabled() == FALSE);
 597
 598         /*
 599          * Past this point, kernel subsystems that expect to operate with
 600          * interrupts or preemption enabled may begin enforcement.
 601          */
 602         early_boot_complete = TRUE;
 603
 604 #if INTERRUPT_MASKED_DEBUG
 605         // Reset interrupts masked timeout before we enable interrupts
 606         ml_spin_debug_clear_self();
 607 #endif
 608         (void) spllo();         /* Allow interruptions */
 609
 610 #if (defined(__i386__) || defined(__x86_64__)) && NCOPY_WINDOWS > 0
 611         /*
 612          * Create and initialize the copy window for processor 0
 613          * This also allocates window space for all other processors.
 614          * However, this is dependent on the number of processors - so this call
 615          * must be after IOKit has been started because IOKit performs processor
 616          * discovery.
 617          */
 618         cpu_userwindow_init(0);
 619 #endif
 620
 621         /*
 622          *      Initialize the shared region module.
 623          */
 624         vm_shared_region_init();
 625         vm_commpage_init();
 626         vm_commpage_text_init();
 627
 628 #if CONFIG_MACF
 629         kernel_bootstrap_log("mac_policy_initmach");
 630         mac_policy_initmach();
 631 #if CONFIG_VNGUARD
 632         vnguard_policy_init();
 633 #endif
 634 #endif
 635
 636 #if CONFIG_DTRACE
 637         dtrace_early_init();
 638         sdt_early_init();
 639 #endif
 640
 641
 642         /*
 643          * Get rid of segments used to bootstrap kext loading. This removes
 644          * the KLD, PRELINK symtab, LINKEDIT, and symtab segments/load commands.
 645          * Must be done prior to lockdown so that we can free (and possibly relocate)
 646          * the static KVA mappings used for the jettisoned bootstrap segments.
 647          */
 648         OSKextRemoveKextBootstrap();
 649 #if defined(__arm__) || defined(__arm64__)
 650 #if CONFIG_KERNEL_INTEGRITY
 651         machine_lockdown_preflight();
 652 #endif
 653         /*
 654          *  Finalize protections on statically mapped pages now that comm page mapping is established.
 655          */
 656         arm_vm_prot_finalize(PE_state.bootArgs);
 657 #endif
 658
 659 #if CONFIG_SCHED_SFI
 660         kernel_bootstrap_log("sfi_init");
 661         sfi_init();
 662 #endif
 663
 664         /*
 665          * Initialize the globals used for permuting kernel
 666          * addresses that may be exported to userland as tokens
 667          * using VM_KERNEL_ADDRPERM()/VM_KERNEL_ADDRPERM_EXTERNAL().
 668          * Force the random number to be odd to avoid mapping a non-zero
 669          * word-aligned address to zero via addition.
 670          * Note: at this stage we can use the cryptographically secure PRNG
 671          * rather than early_random().
 672          */
 673         read_random(&vm_kernel_addrperm, sizeof(vm_kernel_addrperm));
 674         vm_kernel_addrperm |= 1;
 675         read_random(&buf_kernel_addrperm, sizeof(buf_kernel_addrperm));
 676         buf_kernel_addrperm |= 1;
 677         read_random(&vm_kernel_addrperm_ext, sizeof(vm_kernel_addrperm_ext));
 678         vm_kernel_addrperm_ext |= 1;
 679         read_random(&vm_kernel_addrhash_salt, sizeof(vm_kernel_addrhash_salt));
 680         read_random(&vm_kernel_addrhash_salt_ext, sizeof(vm_kernel_addrhash_salt_ext));
 681
 682         vm_set_restrictions();
 683
 684
 685 #ifdef CONFIG_XNUPOST
 686         kern_return_t result = kernel_list_tests();
 687         result = kernel_do_post();
 688         if (result != KERN_SUCCESS) {
 689                 panic("kernel_do_post: Tests failed with result = 0x%08x\n", result);
 690         }
 691         kernel_bootstrap_log("kernel_do_post - done");
 692 #endif /* CONFIG_XNUPOST */
 693
 694
 695         /*
 696          *      Start the user bootstrap.
 697          */
 698 #ifdef  MACH_BSD
 699         bsd_init();
 700 #endif
 701
 702
 703         /*
 704          * Get rid of pages used for early boot tracing.
 705          */
 706         kdebug_free_early_buf();
 707
 708         serial_keyboard_init();         /* Start serial keyboard if wanted */
 709
 710         vm_page_init_local_q();
 711
 712         thread_bind(PROCESSOR_NULL);
 713
 714         /*
 715          *      Become the pageout daemon.
 716          */
 717         vm_pageout();
 718         /*NOTREACHED*/
 719 }
 720
 721 /*
 722  *      slave_main:
 723  *
 724  *      Load the first thread to start a processor.
 725  */
 726 void
 727 slave_main(void *machine_param)
 728 {
 729         processor_t             processor = current_processor();
 730         thread_t                thread;
 731
 732         /*
 733          *      Use the idle processor thread if there
 734          *      is no dedicated start up thread.
 735          */
 736         if (processor->next_thread == THREAD_NULL) {
 737                 thread = processor->idle_thread;
 738                 thread->continuation = (thread_continue_t)processor_start_thread;
 739                 thread->parameter = machine_param;
 740         }
 741         else {
 742                 thread = processor->next_thread;
 743                 processor->next_thread = THREAD_NULL;
 744         }
 745
 746         load_context(thread);
 747         /*NOTREACHED*/
 748 }
 749
 750 /*
 751  *      processor_start_thread:
 752  *
 753  *      First thread to execute on a started processor.
 754  *
 755  *      Called at splsched.
 756  */
 757 void
 758 processor_start_thread(void *machine_param)
 759 {
 760         processor_t             processor = current_processor();
 761         thread_t                self = current_thread();
 762
 763         slave_machine_init(machine_param);
 764
 765         /*
 766          *      If running the idle processor thread,
 767          *      reenter the idle loop, else terminate.
 768          */
 769         if (self == processor->idle_thread)
 770                 thread_block((thread_continue_t)idle_thread);
 771
 772         thread_terminate(self);
 773         /*NOTREACHED*/
 774 }
 775
 776 /*
 777  *      load_context:
 778  *
 779  *      Start the first thread on a processor.
 780  */
 781 static void __attribute__((noreturn))
 782 load_context(
 783         thread_t                thread)
 784 {
 785         processor_t             processor = current_processor();
 786
 787
 788 #define load_context_kprintf(x...) /* kprintf("load_context: " x) */
 789
 790         load_context_kprintf("machine_set_current_thread\n");
 791         machine_set_current_thread(thread);
 792
 793         load_context_kprintf("processor_up\n");
 794         processor_up(processor);
 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          * The idle processor threads are not counted as
 813          * running for load calculations.
 814          */
 815         if (!(thread->state & TH_IDLE))
 816                 sched_run_incr(thread);
 817
 818         processor->active_thread = thread;
 819         processor_state_update_explicit(processor, thread->sched_pri,
 820                 SFI_CLASS_KERNEL, PSET_SMP, thread_get_perfcontrol_class(thread));
 821         processor->starting_pri = thread->sched_pri;
 822         processor->deadline = UINT64_MAX;
 823         thread->last_processor = processor;
 824
 825         processor->last_dispatch = mach_absolute_time();
 826         timer_start(&thread->system_timer, processor->last_dispatch);
 827         PROCESSOR_DATA(processor, thread_timer) = PROCESSOR_DATA(processor, kernel_timer) = &thread->system_timer;
 828
 829         timer_start(&PROCESSOR_DATA(processor, system_state), processor->last_dispatch);
 830         PROCESSOR_DATA(processor, current_state) = &PROCESSOR_DATA(processor, system_state);
 831
 832
 833         cpu_quiescent_counter_join(processor->last_dispatch);
 834
 835         PMAP_ACTIVATE_USER(thread, processor->cpu_id);
 836
 837         load_context_kprintf("machine_load_context\n");
 838         machine_load_context(thread);
 839         /*NOTREACHED*/
 840 }
 841
 842 void
 843 scale_setup()
 844 {
 845         int scale = 0;
 846 #if defined(__LP64__)
 847         typeof(task_max) task_max_base = task_max;
 848
 849         /* Raise limits for servers with >= 16G */
 850         if ((serverperfmode != 0) && ((uint64_t)sane_size >= (uint64_t)(16 * 1024 * 1024 *1024ULL))) {
 851                 scale = (int)((uint64_t)sane_size / (uint64_t)(8 * 1024 * 1024 *1024ULL));
 852                 /* limit to 128 G */
 853                 if (scale > 16)
 854                         scale = 16;
 855                 task_max_base = 2500;
 856         /* Raise limits for machines with >= 3GB */
 857         } else if ((uint64_t)sane_size >= (uint64_t)(3 * 1024 * 1024 *1024ULL)) {
 858                 if ((uint64_t)sane_size < (uint64_t)(8 * 1024 * 1024 *1024ULL)) {
 859                         scale = 2;
 860                 } else {
 861                         /* limit to 64GB */
 862                         scale = MIN(16, (int)((uint64_t)sane_size / (uint64_t)(4 * 1024 * 1024 *1024ULL)));
 863                 }
 864         }
 865
 866         task_max = MAX(task_max, task_max_base * scale);
 867
 868         if (scale != 0) {
 869                 task_threadmax = task_max;
 870                 thread_max = task_max * 5;
 871         }
 872
 873 #endif
 874
 875         bsd_scale_setup(scale);
 876
 877         ipc_space_max = SPACE_MAX;
 878         ipc_port_max = PORT_MAX;
 879         ipc_pset_max = SET_MAX;
 880         semaphore_max = SEMAPHORE_MAX;
 881 }
 882