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_xnu.h>
110 #include <vm/vm_init_xnu.h>
111 #include <vm/vm_map.h>
112 #include <vm/vm_object_xnu.h>
113 #include <vm/vm_page.h>
114 #include <vm/vm_pageout_xnu.h>
115 #include <vm/vm_shared_region_xnu.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 ALTERNATE_DEBUGGER
148 #include <arm64/alternate_debugger.h>
149 #endif
150
151 #if MACH_KDP
152 #include <kdp/kdp.h>
153 #endif
154
155 #if CONFIG_MACF
156 #include <security/mac_mach_internal.h>
157 #if CONFIG_VNGUARD
158 extern void vnguard_policy_init(void);
159 #endif
160 #endif
161
162 #if HYPERVISOR
163 #include <kern/hv_support.h>
164 #endif
165
166 #if CONFIG_UBSAN_MINIMAL
167 #include <san/ubsan_minimal.h>
168 #endif
169
170 #include <san/kasan.h>
171
172 #include <i386/pmCPU.h>
173 static void kernel_bootstrap_thread(void);
174
175 static void load_context(
176 thread_t thread);
177
178 #if CONFIG_ECC_LOGGING
179 #include <kern/ecc.h>
180 #endif
181
182 #if (defined(__i386__) || defined(__x86_64__)) && CONFIG_VMX
183 #include <i386/vmx/vmx_cpu.h>
184 #endif
185
186 #if CONFIG_DTRACE
187 extern void dtrace_early_init(void);
188 extern void sdt_early_init(void);
189 #endif
190
191 #ifdef CONFIG_BTI_TELEMETRY
192 #include <arm64/bti_telemetry.h>
193 #endif /* CONFIG_BTI_TELEMETRY */
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 TUNABLE(startup_debug_t, startup_debug, "startup_debug", 0);
218
219 /* Indicates a server boot when set */
220 TUNABLE(int, serverperfmode, "serverperfmode", 0);
221
222 static inline void
kernel_bootstrap_log(const char * message)223 kernel_bootstrap_log(const char *message)
224 {
225 if ((startup_debug & STARTUP_DEBUG_VERBOSE) &&
226 startup_phase >= STARTUP_SUB_KPRINTF) {
227 kprintf("kernel_bootstrap: %s\n", message);
228 }
229 kernel_debug_string_early(message);
230 }
231
232 static inline void
kernel_bootstrap_thread_log(const char * message)233 kernel_bootstrap_thread_log(const char *message)
234 {
235 if ((startup_debug & STARTUP_DEBUG_VERBOSE) &&
236 startup_phase >= STARTUP_SUB_KPRINTF) {
237 kprintf("kernel_bootstrap_thread: %s\n", message);
238 }
239 kernel_debug_string_early(message);
240 }
241
242 extern void
243 qsort(void *a, size_t n, size_t es, int (*cmp)(const void *, const void *));
244
245 __startup_func
246 static int
startup_entry_cmp(const void * e1,const void * e2)247 startup_entry_cmp(const void *e1, const void *e2)
248 {
249 const struct startup_entry *a = e1;
250 const struct startup_entry *b = e2;
251 if (a->subsystem == b->subsystem) {
252 if (a->rank == b->rank) {
253 return 0;
254 }
255 return a->rank > b->rank ? 1 : -1;
256 }
257 return a->subsystem > b->subsystem ? 1 : -1;
258 }
259
260 __startup_func
261 void
kernel_startup_bootstrap(void)262 kernel_startup_bootstrap(void)
263 {
264 /*
265 * Sort the various STARTUP() entries by subsystem/rank.
266 */
267 size_t n = startup_entries_end - startup_entries;
268
269 if (n == 0) {
270 panic("Section %s,%s missing",
271 STARTUP_HOOK_SEGMENT, STARTUP_HOOK_SECTION);
272 }
273 if (((uintptr_t)startup_entries_end - (uintptr_t)startup_entries) %
274 sizeof(struct startup_entry)) {
275 panic("Section %s,%s has invalid size",
276 STARTUP_HOOK_SEGMENT, STARTUP_HOOK_SECTION);
277 }
278
279 qsort(startup_entries, n, sizeof(struct startup_entry), startup_entry_cmp);
280
281 /*
282 * Then initialize all tunables, timeouts, and locks
283 */
284 kernel_startup_initialize_upto(STARTUP_SUB_LOCKS);
285 }
286
287 __startup_func
288 void
kernel_startup_tunable_init(const struct startup_tunable_spec * spec)289 kernel_startup_tunable_init(const struct startup_tunable_spec *spec)
290 {
291 if (spec->var_is_str) {
292 PE_parse_boot_arg_str(spec->name, spec->var_addr, spec->var_len);
293 } else if (PE_parse_boot_argn(spec->name, spec->var_addr, spec->var_len)) {
294 if (spec->var_is_bool) {
295 /* make sure bool's are valued in {0, 1} */
296 *(bool *)spec->var_addr = *(uint8_t *)spec->var_addr;
297 }
298 }
299 }
300
301 __startup_func
302 void
kernel_startup_tunable_dt_init(const struct startup_tunable_dt_spec * spec)303 kernel_startup_tunable_dt_init(const struct startup_tunable_dt_spec *spec)
304 {
305 DTEntry base;
306
307 if (SecureDTLookupEntry(NULL, spec->dt_base, &base) != kSuccess) {
308 base = NULL;
309 }
310
311 bool found_in_chosen = false;
312
313 if (spec->dt_chosen_override) {
314 DTEntry chosen, chosen_base;
315
316 if (SecureDTLookupEntry(NULL, "chosen", &chosen) != kSuccess) {
317 chosen = NULL;
318 }
319
320 if (chosen != NULL && SecureDTLookupEntry(chosen, spec->dt_base, &chosen_base) == kSuccess) {
321 base = chosen_base;
322 found_in_chosen = true;
323 }
324 }
325
326 uint64_t const *data;
327 unsigned int data_size = spec->var_len;
328
329 if (base != NULL && SecureDTGetProperty(base, spec->dt_name, (const void **)&data, &data_size) == kSuccess) {
330 if (data_size != spec->var_len) {
331 panic("unexpected tunable size %u in DT entry %s/%s/%s",
332 data_size, found_in_chosen ? "/chosen" : "", spec->dt_base, spec->dt_name);
333 }
334
335 /* No need to handle bools specially, they are 1 byte integers in the DT. */
336 memcpy(spec->var_addr, data, spec->var_len);
337 }
338
339 /* boot-arg overrides. */
340
341 if (PE_parse_boot_argn(spec->boot_arg_name, spec->var_addr, spec->var_len)) {
342 if (spec->var_is_bool) {
343 *(bool *)spec->var_addr = *(uint8_t *)spec->var_addr;
344 }
345 }
346 }
347
348 static void
kernel_startup_log(startup_subsystem_id_t subsystem)349 kernel_startup_log(startup_subsystem_id_t subsystem)
350 {
351 static const char *names[] = {
352 [STARTUP_SUB_TUNABLES] = "tunables",
353 [STARTUP_SUB_TIMEOUTS] = "timeouts",
354 [STARTUP_SUB_LOCKS] = "locks",
355 [STARTUP_SUB_KPRINTF] = "kprintf",
356
357 [STARTUP_SUB_PMAP_STEAL] = "pmap_steal",
358 [STARTUP_SUB_KMEM] = "kmem",
359 [STARTUP_SUB_ZALLOC] = "zalloc",
360 [STARTUP_SUB_PERCPU] = "percpu",
361 [STARTUP_SUB_EVENT] = "event",
362
363 [STARTUP_SUB_CODESIGNING] = "codesigning",
364 [STARTUP_SUB_KTRACE] = "ktrace",
365 [STARTUP_SUB_OSLOG] = "oslog",
366 [STARTUP_SUB_MACH_IPC] = "mach_ipc",
367 [STARTUP_SUB_THREAD_CALL] = "thread_call",
368 [STARTUP_SUB_SYSCTL] = "sysctl",
369 [STARTUP_SUB_EARLY_BOOT] = "early_boot",
370
371 /* LOCKDOWN is special and its value won't fit here. */
372 };
373 static startup_subsystem_id_t logged = STARTUP_SUB_NONE;
374
375 if (subsystem <= logged) {
376 return;
377 }
378
379 if (subsystem < sizeof(names) / sizeof(names[0]) && names[subsystem]) {
380 kernel_bootstrap_log(names[subsystem]);
381 }
382 logged = subsystem;
383 }
384
385 __startup_func
386 void
event_register_handler(struct event_hdr * hdr)387 event_register_handler(struct event_hdr *hdr)
388 {
389 struct event_hdr *head = hdr->next;
390
391 hdr->next = head->next;
392 head->next = hdr;
393 }
394
395 __startup_func
396 void
kernel_startup_initialize_upto(startup_subsystem_id_t upto)397 kernel_startup_initialize_upto(startup_subsystem_id_t upto)
398 {
399 struct startup_entry *cur = startup_entry_cur;
400
401 assert(startup_phase < upto);
402
403 while (cur < startup_entries_end && cur->subsystem <= upto) {
404 if ((startup_debug & STARTUP_DEBUG_VERBOSE) &&
405 startup_phase >= STARTUP_SUB_KPRINTF) {
406 kprintf("%s[%d, rank %d]: %p(%p)\n", __func__,
407 cur->subsystem, cur->rank, cur->func, cur->arg);
408 }
409 startup_phase = cur->subsystem - 1;
410 kernel_startup_log(cur->subsystem);
411 cur->func(cur->arg);
412 startup_entry_cur = ++cur;
413 }
414 kernel_startup_log(upto);
415
416 if ((startup_debug & STARTUP_DEBUG_VERBOSE) &&
417 upto >= STARTUP_SUB_KPRINTF) {
418 kprintf("%s: reached phase %d\n", __func__, upto);
419 }
420 startup_phase = upto;
421 }
422
423 void
kernel_bootstrap(void)424 kernel_bootstrap(void)
425 {
426 kern_return_t result;
427 thread_t thread;
428 char namep[16];
429
430 code_signing_config_t cs_config;
431
432 printf("%s\n", version); /* log kernel version */
433
434 scale_setup();
435
436 kernel_bootstrap_log("vm_mem_bootstrap");
437 vm_mem_bootstrap();
438
439 machine_info.memory_size = (uint32_t)mem_size;
440 #if XNU_TARGET_OS_OSX
441 machine_info.max_mem = max_mem_actual;
442 #else
443 machine_info.max_mem = max_mem;
444 #endif /* XNU_TARGET_OS_OSX */
445 machine_info.major_version = version_major;
446 machine_info.minor_version = version_minor;
447
448 #if CONFIG_ATM
449 /* Initialize the Activity Trace Resource Manager. */
450 kernel_bootstrap_log("atm_init");
451 atm_init();
452 #endif
453 kernel_startup_initialize_upto(STARTUP_SUB_OSLOG);
454
455 #if CONFIG_UBSAN_MINIMAL
456 kernel_bootstrap_log("UBSan minimal runtime init");
457 ubsan_minimal_init();
458 #endif
459
460 #ifdef CONFIG_BTI_TELEMETRY
461 kernel_bootstrap_log("BTI exception telemetry runtime init");
462 bti_telemetry_init();
463 #endif /* CONFIG_BTI_TELEMETRY */
464
465 #if KASAN
466 kernel_bootstrap_log("kasan_late_init");
467 kasan_late_init();
468 #endif
469
470 #if CONFIG_TELEMETRY
471 kernel_bootstrap_log("telemetry_init");
472 telemetry_init();
473 #endif
474
475 if (PE_i_can_has_debugger(NULL)) {
476 if (PE_parse_boot_argn("-show_pointers", &namep, sizeof(namep))) {
477 doprnt_hide_pointers = FALSE;
478 }
479 if (PE_parse_boot_argn("-no_slto_panic", &namep, sizeof(namep))) {
480 extern boolean_t spinlock_timeout_panic;
481 spinlock_timeout_panic = FALSE;
482 }
483 }
484
485 kernel_bootstrap_log("console_init");
486 console_init();
487
488 kernel_bootstrap_log("stackshot_init");
489 stackshot_init();
490
491 kernel_bootstrap_log("sched_init");
492 sched_init();
493
494 #if CONFIG_MACF
495 kernel_bootstrap_log("mac_policy_init");
496 mac_policy_init();
497 #endif
498
499 kernel_startup_initialize_upto(STARTUP_SUB_MACH_IPC);
500
501 /*
502 * As soon as the virtual memory system is up, we record
503 * that this CPU is using the kernel pmap.
504 */
505 kernel_bootstrap_log("PMAP_ACTIVATE_KERNEL");
506 PMAP_ACTIVATE_KERNEL(master_cpu);
507
508 kernel_bootstrap_log("mapping_free_prime");
509 mapping_free_prime(); /* Load up with temporary mapping blocks */
510
511 kernel_bootstrap_log("machine_init");
512 machine_init();
513
514 kernel_bootstrap_log("thread_machine_init_template");
515 thread_machine_init_template();
516
517 kernel_bootstrap_log("clock_init");
518 clock_init();
519
520 /*
521 * Initialize the IPC, task, and thread subsystems.
522 */
523 #if CONFIG_THREAD_GROUPS
524 kernel_bootstrap_log("thread_group_init");
525 thread_group_init();
526 #endif
527
528 #if CONFIG_COALITIONS
529 kernel_bootstrap_log("coalitions_init");
530 coalitions_init();
531 #endif
532
533 kernel_bootstrap_log("code_signing_init");
534 code_signing_init();
535 code_signing_configuration(NULL, &cs_config);
536 #if XNU_TARGET_OS_OSX && (DEVELOPMENT || DEBUG)
537 if (cs_config & CS_CONFIG_GET_OUT_OF_MY_WAY) {
538 AMFI_bootarg_disable_mach_hardening = true;
539 }
540 #endif /* XNU_TARGET_OS_OSX && (DEVELOPMENT || DEBUG) */
541
542 kernel_bootstrap_log("task_init");
543 task_init();
544
545 kernel_bootstrap_log("thread_init");
546 thread_init();
547
548 kernel_bootstrap_log("restartable_init");
549 restartable_init();
550
551 kernel_bootstrap_log("workq_init");
552 workq_init();
553
554 kernel_bootstrap_log("turnstiles_init");
555 turnstiles_init();
556
557 #if PAGE_SLEEP_WITH_INHERITOR
558 kernel_bootstrap_log("page_worker_init");
559 page_worker_init();
560 #endif /* PAGE_SLEEP_WITH_INHERITOR */
561
562 kernel_bootstrap_log("mach_init_activity_id");
563 mach_init_activity_id();
564
565 /* initialize host_statistics */
566 host_statistics_init();
567
568 /* initialize exceptions */
569 kernel_bootstrap_log("exception_init");
570 exception_init();
571
572 #if CONFIG_SCHED_SFI
573 kernel_bootstrap_log("sfi_init");
574 sfi_init();
575 #endif
576
577 /*
578 * Create a kernel thread to execute the kernel bootstrap.
579 */
580
581 kernel_bootstrap_log("kernel_thread_create");
582 result = kernel_thread_create((thread_continue_t)kernel_bootstrap_thread, NULL, MAXPRI_KERNEL, &thread);
583
584 if (result != KERN_SUCCESS) {
585 panic("kernel_bootstrap: result = %08X", result);
586 }
587
588 /* TODO: do a proper thread_start() (without the thread_setrun()) */
589 thread->state = TH_RUN;
590 thread->last_made_runnable_time = mach_absolute_time();
591 thread_set_thread_name(thread, "kernel_bootstrap_thread");
592
593 thread_deallocate(thread);
594
595 kernel_bootstrap_log("load_context - done");
596 load_context(thread);
597 /*NOTREACHED*/
598 }
599
600 SECURITY_READ_ONLY_LATE(vm_offset_t) vm_kernel_addrperm;
601 SECURITY_READ_ONLY_LATE(vm_offset_t) buf_kernel_addrperm;
602 SECURITY_READ_ONLY_LATE(vm_offset_t) vm_kernel_addrperm_ext;
603 SECURITY_READ_ONLY_LATE(uint64_t) vm_kernel_addrhash_salt;
604 SECURITY_READ_ONLY_LATE(uint64_t) vm_kernel_addrhash_salt_ext;
605
606 /*
607 * Now running in a thread. Kick off other services,
608 * invoke user bootstrap, enter pageout loop.
609 */
610 static void
kernel_bootstrap_thread(void)611 kernel_bootstrap_thread(void)
612 {
613 processor_t processor = current_processor();
614
615 #if (DEVELOPMENT || DEBUG)
616 platform_stall_panic_or_spin(PLATFORM_STALL_XNU_LOCATION_KERNEL_BOOTSTRAP);
617 #endif
618
619 kernel_bootstrap_thread_log("idle_thread_create");
620 /*
621 * Create the idle processor thread for the boot processor.
622 */
623 idle_thread_create(processor, idle_thread);
624
625 /*
626 * N.B. Do not stick anything else
627 * before this point.
628 *
629 * Start up the scheduler services.
630 */
631 kernel_bootstrap_thread_log("sched_startup");
632 sched_startup();
633
634 /*
635 * Thread lifecycle maintenance (teardown, stack allocation)
636 */
637 kernel_bootstrap_thread_log("thread_daemon_init");
638 thread_daemon_init();
639
640 /*
641 * Thread callout service.
642 */
643 kernel_startup_initialize_upto(STARTUP_SUB_THREAD_CALL);
644
645 /*
646 * Remain on current processor as
647 * additional processors come online.
648 */
649 kernel_bootstrap_thread_log("thread_bind");
650 suspend_cluster_powerdown();
651 thread_bind(processor);
652
653 /*
654 * Kick off memory mapping adjustments.
655 */
656 kernel_bootstrap_thread_log("mapping_adjust");
657 mapping_adjust();
658
659 /*
660 * Create the clock service.
661 */
662 kernel_bootstrap_thread_log("clock_service_create");
663 clock_service_create();
664
665 /*
666 * Create the device service.
667 */
668 device_service_create();
669
670 phys_carveout_init();
671
672 #if MACH_KDP
673 kernel_bootstrap_log("kdp_init");
674 kdp_init();
675 #endif
676
677 #if ALTERNATE_DEBUGGER
678 alternate_debugger_init();
679 #endif
680
681 #if HYPERVISOR
682 kernel_bootstrap_thread_log("hv_support_init");
683 hv_support_init();
684 #endif
685
686 #if CONFIG_TELEMETRY
687 kernel_bootstrap_log("bootprofile_init");
688 bootprofile_init();
689 #endif
690
691 kernel_startup_initialize_upto(STARTUP_SUB_SYSCTL);
692
693 /*
694 * Initialize the globals used for permuting kernel
695 * addresses that may be exported to userland as tokens
696 * using VM_KERNEL_ADDRPERM()/VM_KERNEL_ADDRPERM_EXTERNAL().
697 * Force the random number to be odd to avoid mapping a non-zero
698 * word-aligned address to zero via addition.
699 */
700 vm_kernel_addrperm = (vm_offset_t)(early_random() | 1);
701 buf_kernel_addrperm = (vm_offset_t)(early_random() | 1);
702 vm_kernel_addrperm_ext = (vm_offset_t)(early_random() | 1);
703 vm_kernel_addrhash_salt = early_random();
704 vm_kernel_addrhash_salt_ext = early_random();
705
706 #ifdef IOKIT
707 kernel_bootstrap_log("PE_init_iokit");
708 PE_init_iokit();
709 #endif
710
711 assert(ml_get_interrupts_enabled() == FALSE);
712
713 /*
714 * Past this point, kernel subsystems that expect to operate with
715 * interrupts or preemption enabled may begin enforcement.
716 */
717 kernel_startup_initialize_upto(STARTUP_SUB_EARLY_BOOT);
718
719 #if SCHED_HYGIENE_DEBUG
720 // Reset interrupts masked timeout before we enable interrupts
721 ml_spin_debug_clear_self();
722 #endif
723 (void) spllo(); /* Allow interruptions */
724
725 /*
726 * This will start displaying progress to the user, start as early as possible
727 */
728 initialize_screen(NULL, kPEAcquireScreen);
729
730 /*
731 * Initialize the shared region module.
732 */
733 vm_commpage_init();
734 vm_commpage_text_init();
735
736 #if CONFIG_MACF
737 kernel_bootstrap_log("mac_policy_initmach");
738 mac_policy_initmach();
739 #if CONFIG_VNGUARD
740 kernel_bootstrap_log("vnguard_policy_init");
741 vnguard_policy_init();
742 #endif
743 #endif
744
745 #if CONFIG_DTRACE
746 kernel_bootstrap_log("dtrace_early_init");
747 dtrace_early_init();
748 sdt_early_init();
749 #endif
750
751 #if CODE_SIGNING_MONITOR
752 /*
753 * Lockdown mode is initialized as a startup function within the early boot
754 * category, which means it has been initialized by now. Query the state and
755 * pass it to the code-signing-monitor if required.
756 */
757 kernel_bootstrap_log("code-signing-monitor lockdown mode");
758 csm_check_lockdown_mode();
759 #endif
760
761 #if CODE_SIGNING_MONITOR
762 kernel_bootstrap_log("provisioning_profile_init");
763 csm_initialize_provisioning_profiles();
764 #endif
765
766 kernel_bootstrap_log("trust_cache_init");
767
768 /* Initialize the runtime for the trust cache interface */
769 trust_cache_runtime_init();
770
771 /* Load the static and engineering trust caches */
772 load_static_trust_cache();
773
774 kernel_startup_initialize_upto(STARTUP_SUB_LOCKDOWN);
775
776 /*
777 * Get rid of segments used to bootstrap kext loading. This removes
778 * the KLD, PRELINK symtab, LINKEDIT, and symtab segments/load commands.
779 * Must be done prior to lockdown so that we can free (and possibly relocate)
780 * the static KVA mappings used for the jettisoned bootstrap segments.
781 */
782 kernel_bootstrap_log("OSKextRemoveKextBootstrap");
783 OSKextRemoveKextBootstrap();
784
785 #if SOCKETS
786 /*
787 * Initialize callback table before machine lockdown
788 */
789 mbuf_tag_init();
790 #endif
791
792 /* No changes to kernel text and rodata beyond this point. */
793 kernel_bootstrap_log("machine_lockdown");
794 machine_lockdown();
795
796 #ifdef CONFIG_XNUPOST
797 kern_return_t result = kernel_list_tests();
798 result = kernel_do_post();
799 if (result != KERN_SUCCESS) {
800 panic("kernel_do_post: Tests failed with result = 0x%08x", result);
801 }
802 kernel_bootstrap_log("kernel_do_post - done");
803 #endif /* CONFIG_XNUPOST */
804
805 #ifdef IOKIT
806 kernel_bootstrap_log("PE_lockdown_iokit");
807 PE_lockdown_iokit();
808 #endif
809 /*
810 * max_cpus must be nailed down by the time PE_lockdown_iokit() finishes,
811 * at the latest
812 */
813 vm_set_restrictions(machine_info.max_cpus);
814
815
816 #if KPERF
817 kperf_init_early();
818 #endif
819
820 /*
821 * Start the user bootstrap.
822 */
823 #ifdef MACH_BSD
824 bsd_init();
825 #endif
826
827
828 /*
829 * Get rid of pages used for early boot tracing.
830 */
831 kdebug_free_early_buf();
832
833 serial_keyboard_init(); /* Start serial keyboard if wanted */
834
835 vm_page_init_local_q(machine_info.max_cpus);
836
837 thread_bind(PROCESSOR_NULL);
838 resume_cluster_powerdown();
839
840 /*
841 * Now that all CPUs are available to run threads, this is essentially
842 * a background thread. Take this opportunity to initialize and free
843 * any remaining vm_pages that were delayed earlier by pmap_startup().
844 */
845 vm_free_delayed_pages();
846
847 /*
848 * Become the pageout daemon.
849 */
850 vm_pageout();
851 /*NOTREACHED*/
852 }
853
854 /*
855 * secondary_cpu_main:
856 *
857 * Load the first thread to start a processor, or
858 * load the previous thread context when restarting a processor
859 * from shutdown.
860 * This path will also be used by the master processor
861 * after being offlined.
862 */
863 void
secondary_cpu_main(void * machine_param)864 secondary_cpu_main(void *machine_param)
865 {
866 processor_t processor = current_processor();
867 thread_t thread = processor->idle_thread;
868
869 thread->parameter = machine_param;
870
871 load_context(thread);
872 /*NOTREACHED*/
873 }
874
875 /*
876 * processor_start_thread:
877 *
878 * First thread to execute on a started processor.
879 *
880 * Called at splsched.
881 */
882 void
processor_start_thread(void * machine_param,__unused wait_result_t result)883 processor_start_thread(void *machine_param,
884 __unused wait_result_t result)
885 {
886 assert(ml_get_interrupts_enabled() == FALSE);
887 assert(current_thread() == current_processor()->idle_thread);
888
889 #if CONFIG_KCOV
890 kcov_start_cpu(current_processor()->cpu_id);
891 #endif
892
893 #if USE_APPLEARMSMP
894 /*
895 * On AppleARMSMP platforms, the cpu_boot_thread registers the AIC and
896 * FastIPI interrupt handlers before the secondary CPU is booted, so we
897 * can expect the self-IPI to deliver immediately.
898 */
899 bool wait_for_cpu_signal = true;
900 #else /* USE_APPLEARMSMP */
901 /*
902 * On AppleARMCPU platforms, the AIC and AppleARMCPU threads must be
903 * scheduled after the secondary CPUs boot in order to register the IPI
904 * interrupt handlers, so we can not be guaranteed when the self-IPI
905 * will deliver. The threads may even need to run on this CPU, so we
906 * can't spin against the self-IPI being delivered.
907 * See rdar://125383535.
908 */
909 bool wait_for_cpu_signal = false;
910 #endif /* USE_APPLEARMSMP */
911
912 processor_cpu_reinit(machine_param, wait_for_cpu_signal, false);
913
914 thread_block(idle_thread);
915 /*NOTREACHED*/
916 }
917
918 /*
919 * load_context:
920 *
921 * Start the first thread on a processor.
922 * This may be the first thread ever run on a processor, or
923 * it could be a processor that was previously offlined.
924 */
925 static void __attribute__((noreturn))
load_context(thread_t thread)926 load_context(
927 thread_t thread)
928 {
929 processor_t processor = current_processor();
930
931
932 #define load_context_kprintf(x...) /* kprintf("load_context: " x) */
933
934 load_context_kprintf("machine_set_current_thread\n");
935 machine_set_current_thread(thread);
936
937 load_context_kprintf("processor_up\n");
938
939 PMAP_ACTIVATE_KERNEL(processor->cpu_id);
940
941 /*
942 * Acquire a stack if none attached. The panic
943 * should never occur since the thread is expected
944 * to have reserved stack.
945 */
946 load_context_kprintf("thread %p, stack %lx, stackptr %lx\n", thread,
947 thread->kernel_stack, thread->machine.kstackptr);
948 if (!thread->kernel_stack) {
949 load_context_kprintf("stack_alloc_try\n");
950 if (!stack_alloc_try(thread)) {
951 panic("load_context");
952 }
953 }
954
955 /*
956 * The idle processor threads are not counted as
957 * running for load calculations.
958 */
959 if (!(thread->state & TH_IDLE)) {
960 SCHED(run_count_incr)(thread);
961 }
962
963 processor->active_thread = thread;
964 processor_state_update_explicit(processor, thread->sched_pri,
965 SFI_CLASS_KERNEL, PSET_SMP, thread_get_perfcontrol_class(thread), THREAD_URGENCY_NONE,
966 ((thread->state & TH_IDLE) || (thread->bound_processor != PROCESSOR_NULL)) ? TH_BUCKET_SCHED_MAX : thread->th_sched_bucket);
967 processor->current_is_bound = thread->bound_processor != PROCESSOR_NULL;
968 processor->current_is_NO_SMT = false;
969 processor->current_is_eagerpreempt = false;
970 #if CONFIG_THREAD_GROUPS
971 processor->current_thread_group = thread_group_get(thread);
972 #endif
973 processor->starting_pri = thread->sched_pri;
974 processor->deadline = UINT64_MAX;
975 thread->last_processor = processor;
976 processor_up(processor);
977 struct recount_snap snap = { 0 };
978 recount_snapshot(&snap);
979 processor->last_dispatch = snap.rsn_time_mach;
980 recount_processor_online(processor, &snap);
981
982 smr_cpu_join(processor, processor->last_dispatch);
983
984 PMAP_ACTIVATE_USER(thread, processor->cpu_id);
985
986 load_context_kprintf("machine_load_context\n");
987
988 #if KASAN_TBI
989 __asan_handle_no_return();
990 #endif /* KASAN_TBI */
991
992 machine_load_context(thread);
993 /*NOTREACHED*/
994 }
995
996 #define SERVER_PERF_MODE_VALIDATION_DISABLES 0x5dee
997 extern unsigned int kern_feature_overrides;
998
999 void
scale_setup(void)1000 scale_setup(void)
1001 {
1002 boolean_t pe_serverperfmode = FALSE;
1003 int scale = 0;
1004
1005 /* First, check boot-arg only for the feature overrides. Only then we
1006 * take the device-tree setting into account.
1007 */
1008 if (serverperfmode) {
1009 /* If running in serverperfmode disable some internal only diagnostics. */
1010 kern_feature_overrides |= SERVER_PERF_MODE_VALIDATION_DISABLES;
1011 }
1012
1013 pe_serverperfmode = PE_get_default("kern.serverperfmode",
1014 &pe_serverperfmode, sizeof(pe_serverperfmode));
1015 if (pe_serverperfmode) {
1016 serverperfmode = pe_serverperfmode;
1017 }
1018 #if defined(__LP64__)
1019 typeof(task_max) task_max_base = task_max;
1020
1021
1022 /* Raise limits for servers with >= 16G */
1023 if ((serverperfmode != 0) && ((uint64_t)max_mem_actual >= (uint64_t)(16 * 1024 * 1024 * 1024ULL))) {
1024 scale = (int)((uint64_t)sane_size / (uint64_t)(8 * 1024 * 1024 * 1024ULL));
1025 /* limit to 128 G */
1026 if (scale > 16) {
1027 scale = 16;
1028 }
1029 task_max_base = 2500;
1030 /* Raise limits for machines with >= 3GB */
1031 } else if ((uint64_t)max_mem_actual >= (uint64_t)(3 * 1024 * 1024 * 1024ULL)) {
1032 if ((uint64_t)max_mem_actual < (uint64_t)(8 * 1024 * 1024 * 1024ULL)) {
1033 scale = 2;
1034 } else {
1035 /* limit to 64GB */
1036 scale = MIN(16, (int)((uint64_t)max_mem_actual / (uint64_t)(4 * 1024 * 1024 * 1024ULL)));
1037 }
1038 }
1039
1040 task_max = MAX(task_max, task_max_base * scale);
1041
1042 if (scale != 0) {
1043 task_threadmax = task_max;
1044 thread_max = task_max * 5;
1045 }
1046
1047 #endif
1048
1049 bsd_scale_setup(scale);
1050 }
1051