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