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 [email protected]
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 <kern/assert.h>
78 #include <kern/mach_param.h>
79 #include <kern/misc_protos.h>
80 #include <kern/clock.h>
81 #include <kern/coalition.h>
82 #include <kern/cpu_number.h>
83 #include <kern/ledger.h>
84 #include <kern/machine.h>
85 #include <kern/processor.h>
86 #include <kern/restartable.h>
87 #include <kern/sched_prim.h>
88 #include <kern/turnstile.h>
89 #if CONFIG_SCHED_SFI
90 #include <kern/sfi.h>
91 #endif
92 #include <kern/startup.h>
93 #include <kern/task.h>
94 #include <kern/thread.h>
95 #include <kern/timer.h>
96 #if CONFIG_TELEMETRY
97 #include <kern/telemetry.h>
98 #endif
99 #include <kern/kpc.h>
100 #include <kern/zalloc.h>
101 #include <kern/locks.h>
102 #include <kern/debug.h>
103 #if KPERF
104 #include <kperf/kperf.h>
105 #endif /* KPERF */
106 #include <corpses/task_corpse.h>
107 #include <prng/random.h>
108 #include <console/serial_protos.h>
109 #include <vm/vm_kern.h>
110 #include <vm/vm_init.h>
111 #include <vm/vm_map.h>
112 #include <vm/vm_object.h>
113 #include <vm/vm_page.h>
114 #include <vm/vm_pageout.h>
115 #include <vm/vm_shared_region.h>
116 #include <machine/pmap.h>
117 #include <machine/commpage.h>
118 #include <machine/machine_routines.h>
119 #include <libkern/version.h>
120 #include <pexpert/device_tree.h>
121 #include <sys/codesign.h>
122 #include <sys/kdebug.h>
123 #include <sys/random.h>
124 #include <sys/ktrace.h>
125 #include <sys/trust_caches.h>
126 #include <sys/code_signing.h>
127 #include <libkern/section_keywords.h>
128
129 #include <kern/waitq.h>
130 #include <ipc/ipc_voucher.h>
131 #include <mach/host_info.h>
132 #include <pthread/workqueue_internal.h>
133
134 #if SOCKETS
135 extern void mbuf_tag_init(void);
136 #endif
137
138 #if CONFIG_XNUPOST
139 #include <tests/ktest.h>
140 #include <tests/xnupost.h>
141 #endif
142
143 #if CONFIG_ATM
144 #include <atm/atm_internal.h>
145 #endif
146
147 #if CONFIG_CSR
148 #include <sys/csr.h>
149 #endif
150
151 #if ALTERNATE_DEBUGGER
152 #include <arm64/alternate_debugger.h>
153 #endif
154
155 #if MACH_KDP
156 #include <kdp/kdp.h>
157 #endif
158
159 #if CONFIG_MACF
160 #include <security/mac_mach_internal.h>
161 #if CONFIG_VNGUARD
162 extern void vnguard_policy_init(void);
163 #endif
164 #endif
165
166 #if HYPERVISOR
167 #include <kern/hv_support.h>
168 #endif
169
170 #if CONFIG_UBSAN_MINIMAL
171 #include <san/ubsan_minimal.h>
172 #endif
173
174 #include <san/kasan.h>
175
176 #include <i386/pmCPU.h>
177 static void kernel_bootstrap_thread(void);
178
179 static void load_context(
180 thread_t thread);
181
182 #if CONFIG_ECC_LOGGING
183 #include <kern/ecc.h>
184 #endif
185
186 #if (defined(__i386__) || defined(__x86_64__)) && CONFIG_VMX
187 #include <i386/vmx/vmx_cpu.h>
188 #endif
189
190 #if CONFIG_DTRACE
191 extern void dtrace_early_init(void);
192 extern void sdt_early_init(void);
193 #endif
194
195 // libkern/OSKextLib.cpp
196 extern void OSKextRemoveKextBootstrap(void);
197
198 void scale_setup(void);
199 extern void bsd_scale_setup(int);
200 extern unsigned int semaphore_max;
201 extern void stackshot_init(void);
202
203 /*
204 * Running in virtual memory, on the interrupt stack.
205 */
206
207 extern struct startup_entry startup_entries[]
208 __SECTION_START_SYM(STARTUP_HOOK_SEGMENT, STARTUP_HOOK_SECTION);
209
210 extern struct startup_entry startup_entries_end[]
211 __SECTION_END_SYM(STARTUP_HOOK_SEGMENT, STARTUP_HOOK_SECTION);
212
213 static struct startup_entry *__startup_data startup_entry_cur = startup_entries;
214
215 SECURITY_READ_ONLY_LATE(startup_subsystem_id_t) startup_phase = STARTUP_SUB_NONE;
216
217 extern int serverperfmode;
218
219 TUNABLE(startup_debug_t, startup_debug, "startup_debug", 0);
220
221 static inline void
kernel_bootstrap_log(const char * message)222 kernel_bootstrap_log(const char *message)
223 {
224 if ((startup_debug & STARTUP_DEBUG_VERBOSE) &&
225 startup_phase >= STARTUP_SUB_KPRINTF) {
226 kprintf("kernel_bootstrap: %s\n", message);
227 }
228 kernel_debug_string_early(message);
229 }
230
231 static inline void
kernel_bootstrap_thread_log(const char * message)232 kernel_bootstrap_thread_log(const char *message)
233 {
234 if ((startup_debug & STARTUP_DEBUG_VERBOSE) &&
235 startup_phase >= STARTUP_SUB_KPRINTF) {
236 kprintf("kernel_bootstrap_thread: %s\n", message);
237 }
238 kernel_debug_string_early(message);
239 }
240
241 extern void
242 qsort(void *a, size_t n, size_t es, int (*cmp)(const void *, const void *));
243
244 __startup_func
245 static int
startup_entry_cmp(const void * e1,const void * e2)246 startup_entry_cmp(const void *e1, const void *e2)
247 {
248 const struct startup_entry *a = e1;
249 const struct startup_entry *b = e2;
250 if (a->subsystem == b->subsystem) {
251 if (a->rank == b->rank) {
252 return 0;
253 }
254 return a->rank > b->rank ? 1 : -1;
255 }
256 return a->subsystem > b->subsystem ? 1 : -1;
257 }
258
259 __startup_func
260 void
kernel_startup_bootstrap(void)261 kernel_startup_bootstrap(void)
262 {
263 /*
264 * Sort the various STARTUP() entries by subsystem/rank.
265 */
266 size_t n = startup_entries_end - startup_entries;
267
268 if (n == 0) {
269 panic("Section %s,%s missing",
270 STARTUP_HOOK_SEGMENT, STARTUP_HOOK_SECTION);
271 }
272 if (((uintptr_t)startup_entries_end - (uintptr_t)startup_entries) %
273 sizeof(struct startup_entry)) {
274 panic("Section %s,%s has invalid size",
275 STARTUP_HOOK_SEGMENT, STARTUP_HOOK_SECTION);
276 }
277
278 qsort(startup_entries, n, sizeof(struct startup_entry), startup_entry_cmp);
279
280 /*
281 * Then initialize all tunables, timeouts, and locks
282 */
283 kernel_startup_initialize_upto(STARTUP_SUB_LOCKS);
284 }
285
286 __startup_func
287 void
kernel_startup_tunable_init(const struct startup_tunable_spec * spec)288 kernel_startup_tunable_init(const struct startup_tunable_spec *spec)
289 {
290 if (spec->var_is_str) {
291 PE_parse_boot_arg_str(spec->name, spec->var_addr, spec->var_len);
292 } else if (PE_parse_boot_argn(spec->name, spec->var_addr, spec->var_len)) {
293 if (spec->var_is_bool) {
294 /* make sure bool's are valued in {0, 1} */
295 *(bool *)spec->var_addr = *(uint8_t *)spec->var_addr;
296 }
297 }
298 }
299
300 __startup_func
301 void
kernel_startup_tunable_dt_init(const struct startup_tunable_dt_spec * spec)302 kernel_startup_tunable_dt_init(const struct startup_tunable_dt_spec *spec)
303 {
304 DTEntry base;
305
306 if (SecureDTLookupEntry(NULL, spec->dt_base, &base) != kSuccess) {
307 base = NULL;
308 }
309
310 bool found_in_chosen = false;
311
312 if (spec->dt_chosen_override) {
313 DTEntry chosen, chosen_base;
314
315 if (SecureDTLookupEntry(NULL, "chosen", &chosen) != kSuccess) {
316 chosen = NULL;
317 }
318
319 if (chosen != NULL && SecureDTLookupEntry(chosen, spec->dt_base, &chosen_base) == kSuccess) {
320 base = chosen_base;
321 found_in_chosen = true;
322 }
323 }
324
325 uint64_t const *data;
326 unsigned int data_size = spec->var_len;
327
328 if (base != NULL && SecureDTGetProperty(base, spec->dt_name, (const void **)&data, &data_size) == kSuccess) {
329 if (data_size != spec->var_len) {
330 panic("unexpected tunable size %u in DT entry %s/%s/%s",
331 data_size, found_in_chosen ? "/chosen" : "", spec->dt_base, spec->dt_name);
332 }
333
334 /* No need to handle bools specially, they are 1 byte integers in the DT. */
335 memcpy(spec->var_addr, data, spec->var_len);
336 }
337
338 /* boot-arg overrides. */
339
340 if (PE_parse_boot_argn(spec->boot_arg_name, spec->var_addr, spec->var_len)) {
341 if (spec->var_is_bool) {
342 *(bool *)spec->var_addr = *(uint8_t *)spec->var_addr;
343 }
344 }
345 }
346
347 static void
kernel_startup_log(startup_subsystem_id_t subsystem)348 kernel_startup_log(startup_subsystem_id_t subsystem)
349 {
350 static const char *names[] = {
351 [STARTUP_SUB_TUNABLES] = "tunables",
352 [STARTUP_SUB_TIMEOUTS] = "timeouts",
353 [STARTUP_SUB_LOCKS] = "locks",
354 [STARTUP_SUB_KPRINTF] = "kprintf",
355
356 [STARTUP_SUB_PMAP_STEAL] = "pmap_steal",
357 [STARTUP_SUB_KMEM] = "kmem",
358 [STARTUP_SUB_ZALLOC] = "zalloc",
359 [STARTUP_SUB_PERCPU] = "percpu",
360 [STARTUP_SUB_EVENT] = "event",
361
362 [STARTUP_SUB_CODESIGNING] = "codesigning",
363 [STARTUP_SUB_KTRACE] = "ktrace",
364 [STARTUP_SUB_OSLOG] = "oslog",
365 [STARTUP_SUB_MACH_IPC] = "mach_ipc",
366 [STARTUP_SUB_THREAD_CALL] = "thread_call",
367 [STARTUP_SUB_SYSCTL] = "sysctl",
368 [STARTUP_SUB_EARLY_BOOT] = "early_boot",
369
370 /* LOCKDOWN is special and its value won't fit here. */
371 };
372 static startup_subsystem_id_t logged = STARTUP_SUB_NONE;
373
374 if (subsystem <= logged) {
375 return;
376 }
377
378 if (subsystem < sizeof(names) / sizeof(names[0]) && names[subsystem]) {
379 kernel_bootstrap_log(names[subsystem]);
380 }
381 logged = subsystem;
382 }
383
384 __startup_func
385 void
event_register_handler(struct event_hdr * hdr)386 event_register_handler(struct event_hdr *hdr)
387 {
388 struct event_hdr *head = hdr->next;
389
390 hdr->next = head->next;
391 head->next = hdr;
392 }
393
394 __startup_func
395 void
kernel_startup_initialize_upto(startup_subsystem_id_t upto)396 kernel_startup_initialize_upto(startup_subsystem_id_t upto)
397 {
398 struct startup_entry *cur = startup_entry_cur;
399
400 assert(startup_phase < upto);
401
402 while (cur < startup_entries_end && cur->subsystem <= upto) {
403 if ((startup_debug & STARTUP_DEBUG_VERBOSE) &&
404 startup_phase >= STARTUP_SUB_KPRINTF) {
405 kprintf("%s[%d, rank %d]: %p(%p)\n", __func__,
406 cur->subsystem, cur->rank, cur->func, cur->arg);
407 }
408 startup_phase = cur->subsystem - 1;
409 kernel_startup_log(cur->subsystem);
410 cur->func(cur->arg);
411 startup_entry_cur = ++cur;
412 }
413 kernel_startup_log(upto);
414
415 if ((startup_debug & STARTUP_DEBUG_VERBOSE) &&
416 upto >= STARTUP_SUB_KPRINTF) {
417 kprintf("%s: reached phase %d\n", __func__, upto);
418 }
419 startup_phase = upto;
420 }
421
422 void
kernel_bootstrap(void)423 kernel_bootstrap(void)
424 {
425 kern_return_t result;
426 thread_t thread;
427 char namep[16];
428
429 code_signing_config_t cs_config;
430
431 printf("%s\n", version); /* log kernel version */
432
433 scale_setup();
434
435 kernel_bootstrap_log("vm_mem_bootstrap");
436 vm_mem_bootstrap();
437
438 machine_info.memory_size = (uint32_t)mem_size;
439 #if XNU_TARGET_OS_OSX
440 machine_info.max_mem = max_mem_actual;
441 #else
442 machine_info.max_mem = max_mem;
443 #endif /* XNU_TARGET_OS_OSX */
444 machine_info.major_version = version_major;
445 machine_info.minor_version = version_minor;
446
447 #if CONFIG_ATM
448 /* Initialize the Activity Trace Resource Manager. */
449 kernel_bootstrap_log("atm_init");
450 atm_init();
451 #endif
452 kernel_startup_initialize_upto(STARTUP_SUB_OSLOG);
453
454 #if CONFIG_UBSAN_MINIMAL
455 kernel_bootstrap_log("UBSan minimal runtime init");
456 ubsan_minimal_init();
457 #endif
458
459 #if KASAN
460 kernel_bootstrap_log("kasan_late_init");
461 kasan_late_init();
462 #endif
463
464 #if CONFIG_TELEMETRY
465 kernel_bootstrap_log("telemetry_init");
466 telemetry_init();
467 #endif
468
469 if (PE_i_can_has_debugger(NULL)) {
470 if (PE_parse_boot_argn("-show_pointers", &namep, sizeof(namep))) {
471 doprnt_hide_pointers = FALSE;
472 }
473 if (PE_parse_boot_argn("-no_slto_panic", &namep, sizeof(namep))) {
474 extern boolean_t spinlock_timeout_panic;
475 spinlock_timeout_panic = FALSE;
476 }
477 }
478
479 kernel_bootstrap_log("console_init");
480 console_init();
481
482 kernel_bootstrap_log("stackshot_init");
483 stackshot_init();
484
485 kernel_bootstrap_log("sched_init");
486 sched_init();
487
488 #if CONFIG_MACF
489 kernel_bootstrap_log("mac_policy_init");
490 mac_policy_init();
491 #endif
492
493 kernel_startup_initialize_upto(STARTUP_SUB_MACH_IPC);
494
495 /*
496 * As soon as the virtual memory system is up, we record
497 * that this CPU is using the kernel pmap.
498 */
499 kernel_bootstrap_log("PMAP_ACTIVATE_KERNEL");
500 PMAP_ACTIVATE_KERNEL(master_cpu);
501
502 kernel_bootstrap_log("mapping_free_prime");
503 mapping_free_prime(); /* Load up with temporary mapping blocks */
504
505 kernel_bootstrap_log("machine_init");
506 machine_init();
507
508 kernel_bootstrap_log("thread_machine_init_template");
509 thread_machine_init_template();
510
511 kernel_bootstrap_log("clock_init");
512 clock_init();
513
514 /*
515 * Initialize the IPC, task, and thread subsystems.
516 */
517 #if CONFIG_THREAD_GROUPS
518 kernel_bootstrap_log("thread_group_init");
519 thread_group_init();
520 #endif
521
522 #if CONFIG_COALITIONS
523 kernel_bootstrap_log("coalitions_init");
524 coalitions_init();
525 #endif
526
527 kernel_bootstrap_log("code_signing_init");
528 code_signing_init();
529 code_signing_configuration(NULL, &cs_config);
530 #if XNU_TARGET_OS_OSX && (DEVELOPMENT || DEBUG)
531 if (cs_config & CS_CONFIG_GET_OUT_OF_MY_WAY) {
532 AMFI_bootarg_disable_mach_hardening = true;
533 }
534 #endif /* XNU_TARGET_OS_OSX && (DEVELOPMENT || DEBUG) */
535
536 kernel_bootstrap_log("task_init");
537 task_init();
538
539 kernel_bootstrap_log("thread_init");
540 thread_init();
541
542 kernel_bootstrap_log("restartable_init");
543 restartable_init();
544
545 kernel_bootstrap_log("workq_init");
546 workq_init();
547
548 kernel_bootstrap_log("turnstiles_init");
549 turnstiles_init();
550
551 kernel_bootstrap_log("mach_init_activity_id");
552 mach_init_activity_id();
553
554 /* initialize host_statistics */
555 host_statistics_init();
556
557 /* initialize exceptions */
558 kernel_bootstrap_log("exception_init");
559 exception_init();
560
561 #if CONFIG_SCHED_SFI
562 kernel_bootstrap_log("sfi_init");
563 sfi_init();
564 #endif
565
566 /*
567 * Create a kernel thread to execute the kernel bootstrap.
568 */
569
570 kernel_bootstrap_log("kernel_thread_create");
571 result = kernel_thread_create((thread_continue_t)kernel_bootstrap_thread, NULL, MAXPRI_KERNEL, &thread);
572
573 if (result != KERN_SUCCESS) {
574 panic("kernel_bootstrap: result = %08X", result);
575 }
576
577 /* TODO: do a proper thread_start() (without the thread_setrun()) */
578 thread->state = TH_RUN;
579 thread->last_made_runnable_time = mach_absolute_time();
580 thread_set_thread_name(thread, "kernel_bootstrap_thread");
581
582 thread_deallocate(thread);
583
584 kernel_bootstrap_log("load_context - done");
585 load_context(thread);
586 /*NOTREACHED*/
587 }
588
589 SECURITY_READ_ONLY_LATE(vm_offset_t) vm_kernel_addrperm;
590 SECURITY_READ_ONLY_LATE(vm_offset_t) buf_kernel_addrperm;
591 SECURITY_READ_ONLY_LATE(vm_offset_t) vm_kernel_addrperm_ext;
592 SECURITY_READ_ONLY_LATE(uint64_t) vm_kernel_addrhash_salt;
593 SECURITY_READ_ONLY_LATE(uint64_t) vm_kernel_addrhash_salt_ext;
594
595 /*
596 * Now running in a thread. Kick off other services,
597 * invoke user bootstrap, enter pageout loop.
598 */
599 static void
kernel_bootstrap_thread(void)600 kernel_bootstrap_thread(void)
601 {
602 processor_t processor = current_processor();
603
604 #if (DEVELOPMENT || DEBUG)
605 platform_stall_panic_or_spin(PLATFORM_STALL_XNU_LOCATION_KERNEL_BOOTSTRAP);
606 #endif
607
608 kernel_bootstrap_thread_log("idle_thread_create");
609 /*
610 * Create the idle processor thread.
611 */
612 idle_thread_create(processor);
613
614 /*
615 * N.B. Do not stick anything else
616 * before this point.
617 *
618 * Start up the scheduler services.
619 */
620 kernel_bootstrap_thread_log("sched_startup");
621 sched_startup();
622
623 /*
624 * Thread lifecycle maintenance (teardown, stack allocation)
625 */
626 kernel_bootstrap_thread_log("thread_daemon_init");
627 thread_daemon_init();
628
629 /*
630 * Thread callout service.
631 */
632 kernel_startup_initialize_upto(STARTUP_SUB_THREAD_CALL);
633
634 /*
635 * Remain on current processor as
636 * additional processors come online.
637 */
638 kernel_bootstrap_thread_log("thread_bind");
639 thread_bind(processor);
640
641 /*
642 * Kick off memory mapping adjustments.
643 */
644 kernel_bootstrap_thread_log("mapping_adjust");
645 mapping_adjust();
646
647 /*
648 * Create the clock service.
649 */
650 kernel_bootstrap_thread_log("clock_service_create");
651 clock_service_create();
652
653 /*
654 * Create the device service.
655 */
656 device_service_create();
657
658 phys_carveout_init();
659
660 #if MACH_KDP
661 kernel_bootstrap_log("kdp_init");
662 kdp_init();
663 #endif
664
665 #if ALTERNATE_DEBUGGER
666 alternate_debugger_init();
667 #endif
668
669 #if HYPERVISOR
670 kernel_bootstrap_thread_log("hv_support_init");
671 hv_support_init();
672 #endif
673
674 #if CONFIG_TELEMETRY
675 kernel_bootstrap_log("bootprofile_init");
676 bootprofile_init();
677 #endif
678
679 kernel_startup_initialize_upto(STARTUP_SUB_SYSCTL);
680
681 /*
682 * Initialize the globals used for permuting kernel
683 * addresses that may be exported to userland as tokens
684 * using VM_KERNEL_ADDRPERM()/VM_KERNEL_ADDRPERM_EXTERNAL().
685 * Force the random number to be odd to avoid mapping a non-zero
686 * word-aligned address to zero via addition.
687 */
688 vm_kernel_addrperm = (vm_offset_t)(early_random() | 1);
689 buf_kernel_addrperm = (vm_offset_t)(early_random() | 1);
690 vm_kernel_addrperm_ext = (vm_offset_t)(early_random() | 1);
691 vm_kernel_addrhash_salt = early_random();
692 vm_kernel_addrhash_salt_ext = early_random();
693
694 #ifdef IOKIT
695 kernel_bootstrap_log("PE_init_iokit");
696 PE_init_iokit();
697 #endif
698
699 assert(ml_get_interrupts_enabled() == FALSE);
700
701 /*
702 * Past this point, kernel subsystems that expect to operate with
703 * interrupts or preemption enabled may begin enforcement.
704 */
705 kernel_startup_initialize_upto(STARTUP_SUB_EARLY_BOOT);
706
707 #if SCHED_HYGIENE_DEBUG
708 // Reset interrupts masked timeout before we enable interrupts
709 ml_spin_debug_clear_self();
710 #endif
711 (void) spllo(); /* Allow interruptions */
712
713 /*
714 * This will start displaying progress to the user, start as early as possible
715 */
716 initialize_screen(NULL, kPEAcquireScreen);
717
718 /*
719 * Initialize the shared region module.
720 */
721 vm_commpage_init();
722 vm_commpage_text_init();
723
724 #if CONFIG_MACF
725 kernel_bootstrap_log("mac_policy_initmach");
726 mac_policy_initmach();
727 #if CONFIG_VNGUARD
728 kernel_bootstrap_log("vnguard_policy_init");
729 vnguard_policy_init();
730 #endif
731 #endif
732
733 #if CONFIG_DTRACE
734 kernel_bootstrap_log("dtrace_early_init");
735 dtrace_early_init();
736 sdt_early_init();
737 #endif
738
739 #if CODE_SIGNING_MONITOR
740 /*
741 * Lockdown mode is initialized as a startup function within the early boot
742 * category, which means it has been initialized by now. Query the state and
743 * pass it to the code-signing-monitor if required.
744 */
745 kernel_bootstrap_log("code-signing-monitor lockdown mode");
746 csm_check_lockdown_mode();
747 #endif
748
749 #if CODE_SIGNING_MONITOR
750 kernel_bootstrap_log("provisioning_profile_init");
751 csm_initialize_provisioning_profiles();
752 #endif
753
754 #ifndef BCM2837
755 kernel_bootstrap_log("trust_cache_init");
756
757 /* Initialize the runtime for the trust cache interface */
758 trust_cache_runtime_init();
759
760 /* Load the static and engineering trust caches */
761 load_static_trust_cache();
762 #endif
763
764 kernel_startup_initialize_upto(STARTUP_SUB_LOCKDOWN);
765
766 /*
767 * Get rid of segments used to bootstrap kext loading. This removes
768 * the KLD, PRELINK symtab, LINKEDIT, and symtab segments/load commands.
769 * Must be done prior to lockdown so that we can free (and possibly relocate)
770 * the static KVA mappings used for the jettisoned bootstrap segments.
771 */
772 kernel_bootstrap_log("OSKextRemoveKextBootstrap");
773 OSKextRemoveKextBootstrap();
774
775 #if SOCKETS
776 /*
777 * Initialize callback table before machine lockdown
778 */
779 mbuf_tag_init();
780 #endif
781
782 /* No changes to kernel text and rodata beyond this point. */
783 kernel_bootstrap_log("machine_lockdown");
784 machine_lockdown();
785
786 #ifdef CONFIG_XNUPOST
787 kern_return_t result = kernel_list_tests();
788 result = kernel_do_post();
789 if (result != KERN_SUCCESS) {
790 panic("kernel_do_post: Tests failed with result = 0x%08x", result);
791 }
792 kernel_bootstrap_log("kernel_do_post - done");
793 #endif /* CONFIG_XNUPOST */
794
795 #ifdef IOKIT
796 kernel_bootstrap_log("PE_lockdown_iokit");
797 PE_lockdown_iokit();
798 #endif
799 /*
800 * max_cpus must be nailed down by the time PE_lockdown_iokit() finishes,
801 * at the latest
802 */
803 vm_set_restrictions(machine_info.max_cpus);
804
805
806 #if KPERF
807 kperf_init_early();
808 #endif
809
810 /*
811 * Start the user bootstrap.
812 */
813 #ifdef MACH_BSD
814 bsd_init();
815 #endif
816
817
818 /*
819 * Get rid of pages used for early boot tracing.
820 */
821 kdebug_free_early_buf();
822
823 serial_keyboard_init(); /* Start serial keyboard if wanted */
824
825 vm_page_init_local_q(machine_info.max_cpus);
826
827 thread_bind(PROCESSOR_NULL);
828
829 /*
830 * Now that all CPUs are available to run threads, this is essentially
831 * a background thread. Take this opportunity to initialize and free
832 * any remaining vm_pages that were delayed earlier by pmap_startup().
833 */
834 vm_free_delayed_pages();
835
836 /*
837 * Become the pageout daemon.
838 */
839 vm_pageout();
840 /*NOTREACHED*/
841 }
842
843 /*
844 * slave_main:
845 *
846 * Load the first thread to start a processor.
847 * This path will also be used by the master processor
848 * after being offlined.
849 */
850 void
slave_main(void * machine_param)851 slave_main(void *machine_param)
852 {
853 processor_t processor = current_processor();
854 thread_t thread;
855
856 /*
857 * Use the idle processor thread if there
858 * is no dedicated start up thread.
859 */
860 if (processor->processor_offlined == true) {
861 /* Return to the saved processor_offline context */
862 assert(processor->startup_thread == THREAD_NULL);
863
864 thread = processor->idle_thread;
865 thread->parameter = machine_param;
866 } else if (processor->startup_thread) {
867 thread = processor->startup_thread;
868 processor->startup_thread = THREAD_NULL;
869 } else {
870 thread = processor->idle_thread;
871 thread->continuation = processor_start_thread;
872 thread->parameter = machine_param;
873 }
874
875 load_context(thread);
876 /*NOTREACHED*/
877 }
878
879 /*
880 * processor_start_thread:
881 *
882 * First thread to execute on a started processor.
883 *
884 * Called at splsched.
885 */
886 void
processor_start_thread(void * machine_param,__unused wait_result_t result)887 processor_start_thread(void *machine_param,
888 __unused wait_result_t result)
889 {
890 processor_t processor = current_processor();
891 thread_t self = current_thread();
892
893 slave_machine_init(machine_param);
894
895 /*
896 * If running the idle processor thread,
897 * reenter the idle loop, else terminate.
898 */
899 if (self == processor->idle_thread) {
900 thread_block(idle_thread);
901 }
902
903 thread_terminate(self);
904 /*NOTREACHED*/
905 }
906
907 /*
908 * load_context:
909 *
910 * Start the first thread on a processor.
911 * This may be the first thread ever run on a processor, or
912 * it could be a processor that was previously offlined.
913 */
914 static void __attribute__((noreturn))
load_context(thread_t thread)915 load_context(
916 thread_t thread)
917 {
918 processor_t processor = current_processor();
919
920
921 #define load_context_kprintf(x...) /* kprintf("load_context: " x) */
922
923 load_context_kprintf("machine_set_current_thread\n");
924 machine_set_current_thread(thread);
925
926 load_context_kprintf("processor_up\n");
927
928 PMAP_ACTIVATE_KERNEL(processor->cpu_id);
929
930 /*
931 * Acquire a stack if none attached. The panic
932 * should never occur since the thread is expected
933 * to have reserved stack.
934 */
935 load_context_kprintf("thread %p, stack %lx, stackptr %lx\n", thread,
936 thread->kernel_stack, thread->machine.kstackptr);
937 if (!thread->kernel_stack) {
938 load_context_kprintf("stack_alloc_try\n");
939 if (!stack_alloc_try(thread)) {
940 panic("load_context");
941 }
942 }
943
944 /*
945 * The idle processor threads are not counted as
946 * running for load calculations.
947 */
948 if (!(thread->state & TH_IDLE)) {
949 SCHED(run_count_incr)(thread);
950 }
951
952 processor->active_thread = thread;
953 processor_state_update_explicit(processor, thread->sched_pri,
954 SFI_CLASS_KERNEL, PSET_SMP, thread_get_perfcontrol_class(thread), THREAD_URGENCY_NONE,
955 ((thread->state & TH_IDLE) || (thread->bound_processor != PROCESSOR_NULL)) ? TH_BUCKET_SCHED_MAX : thread->th_sched_bucket);
956 processor->current_is_bound = thread->bound_processor != PROCESSOR_NULL;
957 processor->current_is_NO_SMT = false;
958 processor->current_is_eagerpreempt = false;
959 #if CONFIG_THREAD_GROUPS
960 processor->current_thread_group = thread_group_get(thread);
961 #endif
962 processor->starting_pri = thread->sched_pri;
963 processor->deadline = UINT64_MAX;
964 thread->last_processor = processor;
965 processor_up(processor);
966 struct recount_snap snap = { 0 };
967 recount_snapshot(&snap);
968 processor->last_dispatch = snap.rsn_time_mach;
969 recount_processor_online(processor, &snap);
970
971 smr_cpu_join(processor, processor->last_dispatch);
972
973 PMAP_ACTIVATE_USER(thread, processor->cpu_id);
974
975 load_context_kprintf("machine_load_context\n");
976
977 #if KASAN_TBI
978 __asan_handle_no_return();
979 #endif /* KASAN_TBI */
980
981 machine_load_context(thread);
982 /*NOTREACHED*/
983 }
984
985 void
scale_setup(void)986 scale_setup(void)
987 {
988 int scale = 0;
989 #if defined(__LP64__)
990 typeof(task_max) task_max_base = task_max;
991
992 /* Raise limits for servers with >= 16G */
993 if ((serverperfmode != 0) && ((uint64_t)max_mem_actual >= (uint64_t)(16 * 1024 * 1024 * 1024ULL))) {
994 scale = (int)((uint64_t)sane_size / (uint64_t)(8 * 1024 * 1024 * 1024ULL));
995 /* limit to 128 G */
996 if (scale > 16) {
997 scale = 16;
998 }
999 task_max_base = 2500;
1000 /* Raise limits for machines with >= 3GB */
1001 } else if ((uint64_t)max_mem_actual >= (uint64_t)(3 * 1024 * 1024 * 1024ULL)) {
1002 if ((uint64_t)max_mem_actual < (uint64_t)(8 * 1024 * 1024 * 1024ULL)) {
1003 scale = 2;
1004 } else {
1005 /* limit to 64GB */
1006 scale = MIN(16, (int)((uint64_t)max_mem_actual / (uint64_t)(4 * 1024 * 1024 * 1024ULL)));
1007 }
1008 }
1009
1010 task_max = MAX(task_max, task_max_base * scale);
1011
1012 if (scale != 0) {
1013 task_threadmax = task_max;
1014 thread_max = task_max * 5;
1015 }
1016
1017 #endif
1018
1019 bsd_scale_setup(scale);
1020 }
1021