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/zalloc.h>
100 #include <kern/locks.h>
101 #include <kern/debug.h>
102 #if KPERF
103 #include <kperf/kperf.h>
104 #endif /* KPERF */
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 <machine/machine_routines.h>
118 #include <libkern/version.h>
119 #include <pexpert/device_tree.h>
120 #include <sys/codesign.h>
121 #include <sys/kdebug.h>
122 #include <sys/random.h>
123 #include <sys/ktrace.h>
124 #include <sys/trust_caches.h>
125 #include <sys/code_signing.h>
126 #include <libkern/section_keywords.h>
127
128 #include <kern/waitq.h>
129 #include <ipc/ipc_voucher.h>
130 #include <mach/host_info.h>
131 #include <pthread/workqueue_internal.h>
132
133 #if SOCKETS
134 extern void mbuf_tag_init(void);
135 #endif
136
137 #if CONFIG_XNUPOST
138 #include <tests/ktest.h>
139 #include <tests/xnupost.h>
140 #endif
141
142 #if CONFIG_ATM
143 #include <atm/atm_internal.h>
144 #endif
145
146 #if CONFIG_CSR
147 #include <sys/csr.h>
148 #endif
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 [STARTUP_SUB_EVENT] = "event",
364
365 [STARTUP_SUB_CODESIGNING] = "codesigning",
366 [STARTUP_SUB_KTRACE] = "ktrace",
367 [STARTUP_SUB_OSLOG] = "oslog",
368 [STARTUP_SUB_MACH_IPC] = "mach_ipc",
369 [STARTUP_SUB_THREAD_CALL] = "thread_call",
370 [STARTUP_SUB_SYSCTL] = "sysctl",
371 [STARTUP_SUB_EARLY_BOOT] = "early_boot",
372
373 /* LOCKDOWN is special and its value won't fit here. */
374 };
375 static startup_subsystem_id_t logged = STARTUP_SUB_NONE;
376
377 if (subsystem <= logged) {
378 return;
379 }
380
381 if (subsystem < sizeof(names) / sizeof(names[0]) && names[subsystem]) {
382 kernel_bootstrap_log(names[subsystem]);
383 }
384 logged = subsystem;
385 }
386
387 __startup_func
388 void
event_register_handler(struct event_hdr * hdr)389 event_register_handler(struct event_hdr *hdr)
390 {
391 struct event_hdr *head = hdr->next;
392
393 hdr->next = head->next;
394 head->next = hdr;
395 }
396
397 __startup_func
398 void
kernel_startup_initialize_upto(startup_subsystem_id_t upto)399 kernel_startup_initialize_upto(startup_subsystem_id_t upto)
400 {
401 struct startup_entry *cur = startup_entry_cur;
402
403 assert(startup_phase < upto);
404
405 while (cur < startup_entries_end && cur->subsystem <= upto) {
406 if ((startup_debug & STARTUP_DEBUG_VERBOSE) &&
407 startup_phase >= STARTUP_SUB_KPRINTF) {
408 kprintf("%s[%d, rank %d]: %p(%p)\n", __func__,
409 cur->subsystem, cur->rank, cur->func, cur->arg);
410 }
411 startup_phase = cur->subsystem - 1;
412 kernel_startup_log(cur->subsystem);
413 cur->func(cur->arg);
414 startup_entry_cur = ++cur;
415 }
416 kernel_startup_log(upto);
417
418 if ((startup_debug & STARTUP_DEBUG_VERBOSE) &&
419 upto >= STARTUP_SUB_KPRINTF) {
420 kprintf("%s: reached phase %d\n", __func__, upto);
421 }
422 startup_phase = upto;
423 }
424
425 void
kernel_bootstrap(void)426 kernel_bootstrap(void)
427 {
428 kern_return_t result;
429 thread_t thread;
430 char namep[16];
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 #if KASAN
461 kernel_bootstrap_log("kasan_late_init");
462 kasan_late_init();
463 #endif
464
465 #if CONFIG_TELEMETRY
466 kernel_bootstrap_log("telemetry_init");
467 telemetry_init();
468 #endif
469
470 if (PE_i_can_has_debugger(NULL)) {
471 if (PE_parse_boot_argn("-show_pointers", &namep, sizeof(namep))) {
472 doprnt_hide_pointers = FALSE;
473 }
474 if (PE_parse_boot_argn("-no_slto_panic", &namep, sizeof(namep))) {
475 extern boolean_t spinlock_timeout_panic;
476 spinlock_timeout_panic = FALSE;
477 }
478 }
479
480 kernel_bootstrap_log("console_init");
481 console_init();
482
483 kernel_bootstrap_log("stackshot_init");
484 stackshot_init();
485
486 kernel_bootstrap_log("sched_init");
487 sched_init();
488
489 #if CONFIG_MACF
490 kernel_bootstrap_log("mac_policy_init");
491 mac_policy_init();
492 #endif
493
494 kernel_startup_initialize_upto(STARTUP_SUB_MACH_IPC);
495
496 /*
497 * As soon as the virtual memory system is up, we record
498 * that this CPU is using the kernel pmap.
499 */
500 kernel_bootstrap_log("PMAP_ACTIVATE_KERNEL");
501 PMAP_ACTIVATE_KERNEL(master_cpu);
502
503 kernel_bootstrap_log("mapping_free_prime");
504 mapping_free_prime(); /* Load up with temporary mapping blocks */
505
506 kernel_bootstrap_log("machine_init");
507 machine_init();
508
509 kernel_bootstrap_log("thread_machine_init_template");
510 thread_machine_init_template();
511
512 kernel_bootstrap_log("clock_init");
513 clock_init();
514
515 /*
516 * Initialize the IPC, task, and thread subsystems.
517 */
518 #if CONFIG_THREAD_GROUPS
519 kernel_bootstrap_log("thread_group_init");
520 thread_group_init();
521 #endif
522
523 #if CONFIG_COALITIONS
524 kernel_bootstrap_log("coalitions_init");
525 coalitions_init();
526 #endif
527
528 kernel_bootstrap_log("code_signing_init");
529 code_signing_init();
530 code_signing_configuration(NULL, NULL);
531
532 kernel_bootstrap_log("task_init");
533 task_init();
534
535 kernel_bootstrap_log("thread_init");
536 thread_init();
537
538 kernel_bootstrap_log("restartable_init");
539 restartable_init();
540
541 kernel_bootstrap_log("workq_init");
542 workq_init();
543
544 kernel_bootstrap_log("turnstiles_init");
545 turnstiles_init();
546
547 kernel_bootstrap_log("mach_init_activity_id");
548 mach_init_activity_id();
549
550 /* initialize host_statistics */
551 host_statistics_init();
552
553 /* initialize exceptions */
554 kernel_bootstrap_log("exception_init");
555 exception_init();
556
557 #if CONFIG_SCHED_SFI
558 kernel_bootstrap_log("sfi_init");
559 sfi_init();
560 #endif
561
562 /*
563 * Create a kernel thread to execute the kernel bootstrap.
564 */
565
566 kernel_bootstrap_log("kernel_thread_create");
567 result = kernel_thread_create((thread_continue_t)kernel_bootstrap_thread, NULL, MAXPRI_KERNEL, &thread);
568
569 if (result != KERN_SUCCESS) {
570 panic("kernel_bootstrap: result = %08X", result);
571 }
572
573 /* TODO: do a proper thread_start() (without the thread_setrun()) */
574 thread->state = TH_RUN;
575 thread->last_made_runnable_time = mach_absolute_time();
576 thread_set_thread_name(thread, "kernel_bootstrap_thread");
577
578 thread_deallocate(thread);
579
580 kernel_bootstrap_log("load_context - done");
581 load_context(thread);
582 /*NOTREACHED*/
583 }
584
585 SECURITY_READ_ONLY_LATE(vm_offset_t) vm_kernel_addrperm;
586 SECURITY_READ_ONLY_LATE(vm_offset_t) buf_kernel_addrperm;
587 SECURITY_READ_ONLY_LATE(vm_offset_t) vm_kernel_addrperm_ext;
588 SECURITY_READ_ONLY_LATE(uint64_t) vm_kernel_addrhash_salt;
589 SECURITY_READ_ONLY_LATE(uint64_t) vm_kernel_addrhash_salt_ext;
590
591 /*
592 * Now running in a thread. Kick off other services,
593 * invoke user bootstrap, enter pageout loop.
594 */
595 static void
kernel_bootstrap_thread(void)596 kernel_bootstrap_thread(void)
597 {
598 processor_t processor = current_processor();
599
600 #if (DEVELOPMENT || DEBUG)
601 platform_stall_panic_or_spin(PLATFORM_STALL_XNU_LOCATION_KERNEL_BOOTSTRAP);
602 #endif
603
604 kernel_bootstrap_thread_log("idle_thread_create");
605 /*
606 * Create the idle processor thread.
607 */
608 idle_thread_create(processor);
609
610 /*
611 * N.B. Do not stick anything else
612 * before this point.
613 *
614 * Start up the scheduler services.
615 */
616 kernel_bootstrap_thread_log("sched_startup");
617 sched_startup();
618
619 /*
620 * Thread lifecycle maintenance (teardown, stack allocation)
621 */
622 kernel_bootstrap_thread_log("thread_daemon_init");
623 thread_daemon_init();
624
625 /*
626 * Thread callout service.
627 */
628 kernel_startup_initialize_upto(STARTUP_SUB_THREAD_CALL);
629
630 /*
631 * Remain on current processor as
632 * additional processors come online.
633 */
634 kernel_bootstrap_thread_log("thread_bind");
635 thread_bind(processor);
636
637 /*
638 * Kick off memory mapping adjustments.
639 */
640 kernel_bootstrap_thread_log("mapping_adjust");
641 mapping_adjust();
642
643 /*
644 * Create the clock service.
645 */
646 kernel_bootstrap_thread_log("clock_service_create");
647 clock_service_create();
648
649 /*
650 * Create the device service.
651 */
652 device_service_create();
653
654 phys_carveout_init();
655
656 #if MACH_KDP
657 kernel_bootstrap_log("kdp_init");
658 kdp_init();
659 #endif
660
661 #if ALTERNATE_DEBUGGER
662 alternate_debugger_init();
663 #endif
664
665 #if HYPERVISOR
666 kernel_bootstrap_thread_log("hv_support_init");
667 hv_support_init();
668 #endif
669
670 #if CONFIG_TELEMETRY
671 kernel_bootstrap_log("bootprofile_init");
672 bootprofile_init();
673 #endif
674
675 kernel_startup_initialize_upto(STARTUP_SUB_SYSCTL);
676
677 /*
678 * Initialize the globals used for permuting kernel
679 * addresses that may be exported to userland as tokens
680 * using VM_KERNEL_ADDRPERM()/VM_KERNEL_ADDRPERM_EXTERNAL().
681 * Force the random number to be odd to avoid mapping a non-zero
682 * word-aligned address to zero via addition.
683 */
684 vm_kernel_addrperm = (vm_offset_t)(early_random() | 1);
685 buf_kernel_addrperm = (vm_offset_t)(early_random() | 1);
686 vm_kernel_addrperm_ext = (vm_offset_t)(early_random() | 1);
687 vm_kernel_addrhash_salt = early_random();
688 vm_kernel_addrhash_salt_ext = early_random();
689
690 #ifdef IOKIT
691 kernel_bootstrap_log("PE_init_iokit");
692 PE_init_iokit();
693 #endif
694
695 assert(ml_get_interrupts_enabled() == FALSE);
696
697 /*
698 * Past this point, kernel subsystems that expect to operate with
699 * interrupts or preemption enabled may begin enforcement.
700 */
701 kernel_startup_initialize_upto(STARTUP_SUB_EARLY_BOOT);
702
703 #if SCHED_HYGIENE_DEBUG
704 // Reset interrupts masked timeout before we enable interrupts
705 ml_spin_debug_clear_self();
706 #endif
707 (void) spllo(); /* Allow interruptions */
708
709 /*
710 * This will start displaying progress to the user, start as early as possible
711 */
712 initialize_screen(NULL, kPEAcquireScreen);
713
714 /*
715 * Initialize the shared region module.
716 */
717 vm_commpage_init();
718 vm_commpage_text_init();
719
720 #if CONFIG_MACF
721 kernel_bootstrap_log("mac_policy_initmach");
722 mac_policy_initmach();
723 #if CONFIG_VNGUARD
724 kernel_bootstrap_log("vnguard_policy_init");
725 vnguard_policy_init();
726 #endif
727 #endif
728
729 #if CONFIG_DTRACE
730 kernel_bootstrap_log("dtrace_early_init");
731 dtrace_early_init();
732 sdt_early_init();
733 #endif
734
735 #if CODE_SIGNING_MONITOR
736 kernel_bootstrap_log("provisioning_profile_init");
737 csm_initialize_provisioning_profiles();
738 #endif
739
740 #ifndef BCM2837
741 kernel_bootstrap_log("trust_cache_init");
742
743 /* Initialize the runtime for the trust cache interface */
744 trust_cache_runtime_init();
745
746 /* Load the static and engineering trust caches */
747 load_static_trust_cache();
748 #endif
749
750 kernel_startup_initialize_upto(STARTUP_SUB_LOCKDOWN);
751
752 /*
753 * Get rid of segments used to bootstrap kext loading. This removes
754 * the KLD, PRELINK symtab, LINKEDIT, and symtab segments/load commands.
755 * Must be done prior to lockdown so that we can free (and possibly relocate)
756 * the static KVA mappings used for the jettisoned bootstrap segments.
757 */
758 kernel_bootstrap_log("OSKextRemoveKextBootstrap");
759 OSKextRemoveKextBootstrap();
760
761 #if SOCKETS
762 /*
763 * Initialize callback table before machine lockdown
764 */
765 mbuf_tag_init();
766 #endif
767
768 /* No changes to kernel text and rodata beyond this point. */
769 kernel_bootstrap_log("machine_lockdown");
770 machine_lockdown();
771
772 #ifdef CONFIG_XNUPOST
773 kern_return_t result = kernel_list_tests();
774 result = kernel_do_post();
775 if (result != KERN_SUCCESS) {
776 panic("kernel_do_post: Tests failed with result = 0x%08x", result);
777 }
778 kernel_bootstrap_log("kernel_do_post - done");
779 #endif /* CONFIG_XNUPOST */
780
781 #ifdef IOKIT
782 kernel_bootstrap_log("PE_lockdown_iokit");
783 PE_lockdown_iokit();
784 #endif
785 /*
786 * max_cpus must be nailed down by the time PE_lockdown_iokit() finishes,
787 * at the latest
788 */
789 vm_set_restrictions(machine_info.max_cpus);
790
791
792 #if KPERF
793 kperf_init_early();
794 #endif
795
796 /*
797 * Start the user bootstrap.
798 */
799 #ifdef MACH_BSD
800 bsd_init();
801 #endif
802
803
804 /*
805 * Get rid of pages used for early boot tracing.
806 */
807 kdebug_free_early_buf();
808
809 serial_keyboard_init(); /* Start serial keyboard if wanted */
810
811 vm_page_init_local_q(machine_info.max_cpus);
812
813 thread_bind(PROCESSOR_NULL);
814
815 /*
816 * Now that all CPUs are available to run threads, this is essentially
817 * a background thread. Take this opportunity to initialize and free
818 * any remaining vm_pages that were delayed earlier by pmap_startup().
819 */
820 vm_free_delayed_pages();
821
822 /*
823 * Become the pageout daemon.
824 */
825 vm_pageout();
826 /*NOTREACHED*/
827 }
828
829 /*
830 * slave_main:
831 *
832 * Load the first thread to start a processor.
833 * This path will also be used by the master processor
834 * after being offlined.
835 */
836 void
slave_main(void * machine_param)837 slave_main(void *machine_param)
838 {
839 processor_t processor = current_processor();
840 thread_t thread;
841
842 /*
843 * Use the idle processor thread if there
844 * is no dedicated start up thread.
845 */
846 if (processor->processor_offlined == true) {
847 /* Return to the saved processor_offline context */
848 assert(processor->startup_thread == THREAD_NULL);
849
850 thread = processor->idle_thread;
851 thread->parameter = machine_param;
852 } else if (processor->startup_thread) {
853 thread = processor->startup_thread;
854 processor->startup_thread = THREAD_NULL;
855 } else {
856 thread = processor->idle_thread;
857 thread->continuation = processor_start_thread;
858 thread->parameter = machine_param;
859 }
860
861 load_context(thread);
862 /*NOTREACHED*/
863 }
864
865 /*
866 * processor_start_thread:
867 *
868 * First thread to execute on a started processor.
869 *
870 * Called at splsched.
871 */
872 void
processor_start_thread(void * machine_param,__unused wait_result_t result)873 processor_start_thread(void *machine_param,
874 __unused wait_result_t result)
875 {
876 processor_t processor = current_processor();
877 thread_t self = current_thread();
878
879 slave_machine_init(machine_param);
880
881 /*
882 * If running the idle processor thread,
883 * reenter the idle loop, else terminate.
884 */
885 if (self == processor->idle_thread) {
886 thread_block(idle_thread);
887 }
888
889 thread_terminate(self);
890 /*NOTREACHED*/
891 }
892
893 /*
894 * load_context:
895 *
896 * Start the first thread on a processor.
897 * This may be the first thread ever run on a processor, or
898 * it could be a processor that was previously offlined.
899 */
900 static void __attribute__((noreturn))
load_context(thread_t thread)901 load_context(
902 thread_t thread)
903 {
904 processor_t processor = current_processor();
905
906
907 #define load_context_kprintf(x...) /* kprintf("load_context: " x) */
908
909 load_context_kprintf("machine_set_current_thread\n");
910 machine_set_current_thread(thread);
911
912 load_context_kprintf("processor_up\n");
913
914 PMAP_ACTIVATE_KERNEL(processor->cpu_id);
915
916 /*
917 * Acquire a stack if none attached. The panic
918 * should never occur since the thread is expected
919 * to have reserved stack.
920 */
921 load_context_kprintf("thread %p, stack %lx, stackptr %lx\n", thread,
922 thread->kernel_stack, thread->machine.kstackptr);
923 if (!thread->kernel_stack) {
924 load_context_kprintf("stack_alloc_try\n");
925 if (!stack_alloc_try(thread)) {
926 panic("load_context");
927 }
928 }
929
930 /*
931 * The idle processor threads are not counted as
932 * running for load calculations.
933 */
934 if (!(thread->state & TH_IDLE)) {
935 SCHED(run_count_incr)(thread);
936 }
937
938 processor->active_thread = thread;
939 processor_state_update_explicit(processor, thread->sched_pri,
940 SFI_CLASS_KERNEL, PSET_SMP, thread_get_perfcontrol_class(thread), THREAD_URGENCY_NONE,
941 ((thread->state & TH_IDLE) || (thread->bound_processor != PROCESSOR_NULL)) ? TH_BUCKET_SCHED_MAX : thread->th_sched_bucket);
942 processor->current_is_bound = thread->bound_processor != PROCESSOR_NULL;
943 processor->current_is_NO_SMT = false;
944 processor->current_is_eagerpreempt = false;
945 #if CONFIG_THREAD_GROUPS
946 processor->current_thread_group = thread_group_get(thread);
947 #endif
948 processor->starting_pri = thread->sched_pri;
949 processor->deadline = UINT64_MAX;
950 thread->last_processor = processor;
951 processor_up(processor);
952 struct recount_snap snap = { 0 };
953 recount_snapshot(&snap);
954 processor->last_dispatch = snap.rsn_time_mach;
955 recount_processor_online(processor, &snap);
956
957 smr_cpu_join(processor, processor->last_dispatch);
958
959 PMAP_ACTIVATE_USER(thread, processor->cpu_id);
960
961 load_context_kprintf("machine_load_context\n");
962
963 #if KASAN_TBI
964 __asan_handle_no_return();
965 #endif /* KASAN_TBI */
966
967 machine_load_context(thread);
968 /*NOTREACHED*/
969 }
970
971 void
scale_setup(void)972 scale_setup(void)
973 {
974 int scale = 0;
975 #if defined(__LP64__)
976 typeof(task_max) task_max_base = task_max;
977
978 /* Raise limits for servers with >= 16G */
979 if ((serverperfmode != 0) && ((uint64_t)max_mem_actual >= (uint64_t)(16 * 1024 * 1024 * 1024ULL))) {
980 scale = (int)((uint64_t)sane_size / (uint64_t)(8 * 1024 * 1024 * 1024ULL));
981 /* limit to 128 G */
982 if (scale > 16) {
983 scale = 16;
984 }
985 task_max_base = 2500;
986 /* Raise limits for machines with >= 3GB */
987 } else if ((uint64_t)max_mem_actual >= (uint64_t)(3 * 1024 * 1024 * 1024ULL)) {
988 if ((uint64_t)max_mem_actual < (uint64_t)(8 * 1024 * 1024 * 1024ULL)) {
989 scale = 2;
990 } else {
991 /* limit to 64GB */
992 scale = MIN(16, (int)((uint64_t)max_mem_actual / (uint64_t)(4 * 1024 * 1024 * 1024ULL)));
993 }
994 }
995
996 task_max = MAX(task_max, task_max_base * scale);
997
998 if (scale != 0) {
999 task_threadmax = task_max;
1000 thread_max = task_max * 5;
1001 }
1002
1003 #endif
1004
1005 bsd_scale_setup(scale);
1006 }
1007