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 <libkern/section_keywords.h>
126
127 #include <kern/waitq.h>
128 #include <ipc/ipc_voucher.h>
129 #include <voucher/ipc_pthread_priority_internal.h>
130 #include <mach/host_info.h>
131 #include <pthread/workqueue_internal.h>
132
133 #if CONFIG_XNUPOST
134 #include <tests/ktest.h>
135 #include <tests/xnupost.h>
136 #endif
137
138 #if CONFIG_ATM
139 #include <atm/atm_internal.h>
140 #endif
141
142 #if CONFIG_CSR
143 #include <sys/csr.h>
144 #endif
145
146 #include <bank/bank_internal.h>
147
148 #if ALTERNATE_DEBUGGER
149 #include <arm64/alternate_debugger.h>
150 #endif
151
152 #if MACH_KDP
153 #include <kdp/kdp.h>
154 #endif
155
156 #if CONFIG_MACF
157 #include <security/mac_mach_internal.h>
158 #if CONFIG_VNGUARD
159 extern void vnguard_policy_init(void);
160 #endif
161 #endif
162
163 #if KPC
164 #include <kern/kpc.h>
165 #endif
166
167 #if HYPERVISOR
168 #include <kern/hv_support.h>
169 #endif
170
171 #if CONFIG_UBSAN_MINIMAL
172 #include <san/ubsan_minimal.h>
173 #endif
174
175 #include <san/kasan.h>
176
177 #include <i386/pmCPU.h>
178 static void kernel_bootstrap_thread(void);
179
180 static void load_context(
181 thread_t thread);
182
183 #if CONFIG_ECC_LOGGING
184 #include <kern/ecc.h>
185 #endif
186
187 #if (defined(__i386__) || defined(__x86_64__)) && CONFIG_VMX
188 #include <i386/vmx/vmx_cpu.h>
189 #endif
190
191 #if CONFIG_DTRACE
192 extern void dtrace_early_init(void);
193 extern void sdt_early_init(void);
194 #endif
195
196 // libkern/OSKextLib.cpp
197 extern void OSKextRemoveKextBootstrap(void);
198
199 void scale_setup(void);
200 extern void bsd_scale_setup(int);
201 extern unsigned int semaphore_max;
202 extern void stackshot_init(void);
203
204 /*
205 * Running in virtual memory, on the interrupt stack.
206 */
207
208 extern struct startup_entry startup_entries[]
209 __SECTION_START_SYM(STARTUP_HOOK_SEGMENT, STARTUP_HOOK_SECTION);
210
211 extern struct startup_entry startup_entries_end[]
212 __SECTION_END_SYM(STARTUP_HOOK_SEGMENT, STARTUP_HOOK_SECTION);
213
214 static struct startup_entry *__startup_data startup_entry_cur = startup_entries;
215
216 SECURITY_READ_ONLY_LATE(startup_subsystem_id_t) startup_phase = STARTUP_SUB_NONE;
217
218 extern int serverperfmode;
219
220 TUNABLE(startup_debug_t, startup_debug, "startup_debug", 0);
221
222 /* size of kernel trace buffer, disabled by default */
223 TUNABLE(unsigned int, new_nkdbufs, "trace", 0);
224 TUNABLE(unsigned int, wake_nkdbufs, "trace_wake", 0);
225 TUNABLE(unsigned int, write_trace_on_panic, "trace_panic", 0);
226 TUNABLE(unsigned int, trace_wrap, "trace_wrap", 0);
227
228 /* mach leak logging */
229 TUNABLE(int, log_leaks, "-l", 0);
230
231 static inline void
kernel_bootstrap_log(const char * message)232 kernel_bootstrap_log(const char *message)
233 {
234 if ((startup_debug & STARTUP_DEBUG_VERBOSE) &&
235 startup_phase >= STARTUP_SUB_KPRINTF) {
236 kprintf("kernel_bootstrap: %s\n", message);
237 }
238 kernel_debug_string_early(message);
239 }
240
241 static inline void
kernel_bootstrap_thread_log(const char * message)242 kernel_bootstrap_thread_log(const char *message)
243 {
244 if ((startup_debug & STARTUP_DEBUG_VERBOSE) &&
245 startup_phase >= STARTUP_SUB_KPRINTF) {
246 kprintf("kernel_bootstrap_thread: %s\n", message);
247 }
248 kernel_debug_string_early(message);
249 }
250
251 extern void
252 qsort(void *a, size_t n, size_t es, int (*cmp)(const void *, const void *));
253
254 __startup_func
255 static int
startup_entry_cmp(const void * e1,const void * e2)256 startup_entry_cmp(const void *e1, const void *e2)
257 {
258 const struct startup_entry *a = e1;
259 const struct startup_entry *b = e2;
260 if (a->subsystem == b->subsystem) {
261 if (a->rank == b->rank) {
262 return 0;
263 }
264 return a->rank > b->rank ? 1 : -1;
265 }
266 return a->subsystem > b->subsystem ? 1 : -1;
267 }
268
269 __startup_func
270 void
kernel_startup_bootstrap(void)271 kernel_startup_bootstrap(void)
272 {
273 /*
274 * Sort the various STARTUP() entries by subsystem/rank.
275 */
276 size_t n = startup_entries_end - startup_entries;
277
278 if (n == 0) {
279 panic("Section %s,%s missing",
280 STARTUP_HOOK_SEGMENT, STARTUP_HOOK_SECTION);
281 }
282 if (((uintptr_t)startup_entries_end - (uintptr_t)startup_entries) %
283 sizeof(struct startup_entry)) {
284 panic("Section %s,%s has invalid size",
285 STARTUP_HOOK_SEGMENT, STARTUP_HOOK_SECTION);
286 }
287
288 qsort(startup_entries, n, sizeof(struct startup_entry), startup_entry_cmp);
289
290 /*
291 * Then initialize all tunables, timeouts, and early locks
292 */
293 kernel_startup_initialize_upto(STARTUP_SUB_LOCKS_EARLY);
294 }
295
296 __startup_func
297 void
kernel_startup_tunable_init(const struct startup_tunable_spec * spec)298 kernel_startup_tunable_init(const struct startup_tunable_spec *spec)
299 {
300 if (spec->var_is_str) {
301 PE_parse_boot_arg_str(spec->name, spec->var_addr, spec->var_len);
302 } else if (PE_parse_boot_argn(spec->name, spec->var_addr, spec->var_len)) {
303 if (spec->var_is_bool) {
304 /* make sure bool's are valued in {0, 1} */
305 *(bool *)spec->var_addr = *(uint8_t *)spec->var_addr;
306 }
307 }
308 }
309
310 __startup_func
311 void
kernel_startup_tunable_dt_init(const struct startup_tunable_dt_spec * spec)312 kernel_startup_tunable_dt_init(const struct startup_tunable_dt_spec *spec)
313 {
314 DTEntry base;
315
316 if (SecureDTLookupEntry(NULL, spec->dt_base, &base) != kSuccess) {
317 base = NULL;
318 }
319
320 bool found_in_chosen = false;
321
322 if (spec->dt_chosen_override) {
323 DTEntry chosen, chosen_base;
324
325 if (SecureDTLookupEntry(NULL, "chosen", &chosen) != kSuccess) {
326 chosen = NULL;
327 }
328
329 if (chosen != NULL && SecureDTLookupEntry(chosen, spec->dt_base, &chosen_base) == kSuccess) {
330 base = chosen_base;
331 found_in_chosen = true;
332 }
333 }
334
335 uint64_t const *data;
336 unsigned int data_size = spec->var_len;
337
338 if (base != NULL && SecureDTGetProperty(base, spec->dt_name, (const void **)&data, &data_size) == kSuccess) {
339 if (data_size != spec->var_len) {
340 panic("unexpected tunable size %u in DT entry %s/%s/%s",
341 data_size, found_in_chosen ? "/chosen" : "", spec->dt_base, spec->dt_name);
342 }
343
344 /* No need to handle bools specially, they are 1 byte integers in the DT. */
345 memcpy(spec->var_addr, data, spec->var_len);
346 }
347
348 /* boot-arg overrides. */
349
350 if (PE_parse_boot_argn(spec->boot_arg_name, spec->var_addr, spec->var_len)) {
351 if (spec->var_is_bool) {
352 *(bool *)spec->var_addr = *(uint8_t *)spec->var_addr;
353 }
354 }
355 }
356
357 static void
kernel_startup_log(startup_subsystem_id_t subsystem)358 kernel_startup_log(startup_subsystem_id_t subsystem)
359 {
360 static const char *names[] = {
361 [STARTUP_SUB_TUNABLES] = "tunables",
362 [STARTUP_SUB_TIMEOUTS] = "timeouts",
363 [STARTUP_SUB_LOCKS_EARLY] = "locks_early",
364 [STARTUP_SUB_KPRINTF] = "kprintf",
365
366 [STARTUP_SUB_PMAP_STEAL] = "pmap_steal",
367 [STARTUP_SUB_KMEM] = "kmem",
368 [STARTUP_SUB_ZALLOC] = "zalloc",
369 [STARTUP_SUB_PERCPU] = "percpu",
370 [STARTUP_SUB_LOCKS] = "locks",
371
372 [STARTUP_SUB_CODESIGNING] = "codesigning",
373 [STARTUP_SUB_OSLOG] = "oslog",
374 [STARTUP_SUB_MACH_IPC] = "mach_ipc",
375 [STARTUP_SUB_THREAD_CALL] = "thread_call",
376 [STARTUP_SUB_SYSCTL] = "sysctl",
377 [STARTUP_SUB_EARLY_BOOT] = "early_boot",
378
379 /* LOCKDOWN is special and its value won't fit here. */
380 };
381 static startup_subsystem_id_t logged = STARTUP_SUB_NONE;
382
383 if (subsystem <= logged) {
384 return;
385 }
386
387 if (subsystem < sizeof(names) / sizeof(names[0]) && names[subsystem]) {
388 kernel_bootstrap_log(names[subsystem]);
389 }
390 logged = subsystem;
391 }
392
393 __startup_func
394 void
kernel_startup_initialize_upto(startup_subsystem_id_t upto)395 kernel_startup_initialize_upto(startup_subsystem_id_t upto)
396 {
397 struct startup_entry *cur = startup_entry_cur;
398
399 assert(startup_phase < upto);
400
401 while (cur < startup_entries_end && cur->subsystem <= upto) {
402 if ((startup_debug & STARTUP_DEBUG_VERBOSE) &&
403 startup_phase >= STARTUP_SUB_KPRINTF) {
404 kprintf("%s[%d, rank %d]: %p(%p)\n", __func__,
405 cur->subsystem, cur->rank, cur->func, cur->arg);
406 }
407 startup_phase = cur->subsystem - 1;
408 kernel_startup_log(cur->subsystem);
409 cur->func(cur->arg);
410 startup_entry_cur = ++cur;
411 }
412 kernel_startup_log(upto);
413
414 if ((startup_debug & STARTUP_DEBUG_VERBOSE) &&
415 upto >= STARTUP_SUB_KPRINTF) {
416 kprintf("%s: reached phase %d\n", __func__, upto);
417 }
418 startup_phase = upto;
419 }
420
421 void
kernel_bootstrap(void)422 kernel_bootstrap(void)
423 {
424 kern_return_t result;
425 thread_t thread;
426 char namep[16];
427
428 printf("%s\n", version); /* log kernel version */
429
430 scale_setup();
431
432 kernel_bootstrap_log("vm_mem_bootstrap");
433 vm_mem_bootstrap();
434
435 machine_info.memory_size = (uint32_t)mem_size;
436 #if XNU_TARGET_OS_OSX
437 machine_info.max_mem = max_mem_actual;
438 #else
439 machine_info.max_mem = max_mem;
440 #endif /* XNU_TARGET_OS_OSX */
441 machine_info.major_version = version_major;
442 machine_info.minor_version = version_minor;
443
444 #if CONFIG_ATM
445 /* Initialize the Activity Trace Resource Manager. */
446 kernel_bootstrap_log("atm_init");
447 atm_init();
448 #endif
449 kernel_startup_initialize_upto(STARTUP_SUB_OSLOG);
450
451 #if CONFIG_UBSAN_MINIMAL
452 kernel_bootstrap_log("UBSan minimal runtime init");
453 ubsan_minimal_init();
454 #endif
455
456 #if KASAN
457 kernel_bootstrap_log("kasan_late_init");
458 kasan_late_init();
459 #endif
460
461 #if CONFIG_TELEMETRY
462 kernel_bootstrap_log("telemetry_init");
463 telemetry_init();
464 #endif
465
466 if (PE_i_can_has_debugger(NULL)) {
467 if (PE_parse_boot_argn("-show_pointers", &namep, sizeof(namep))) {
468 doprnt_hide_pointers = FALSE;
469 }
470 if (PE_parse_boot_argn("-no_slto_panic", &namep, sizeof(namep))) {
471 extern boolean_t spinlock_timeout_panic;
472 spinlock_timeout_panic = FALSE;
473 }
474 }
475
476 kernel_bootstrap_log("console_init");
477 console_init();
478
479 kernel_bootstrap_log("stackshot_init");
480 stackshot_init();
481
482 kernel_bootstrap_log("sched_init");
483 sched_init();
484
485 #if CONFIG_MACF
486 kernel_bootstrap_log("mac_policy_init");
487 mac_policy_init();
488 #endif
489
490 kernel_startup_initialize_upto(STARTUP_SUB_MACH_IPC);
491
492 /*
493 * As soon as the virtual memory system is up, we record
494 * that this CPU is using the kernel pmap.
495 */
496 kernel_bootstrap_log("PMAP_ACTIVATE_KERNEL");
497 PMAP_ACTIVATE_KERNEL(master_cpu);
498
499 kernel_bootstrap_log("mapping_free_prime");
500 mapping_free_prime(); /* Load up with temporary mapping blocks */
501
502 kernel_bootstrap_log("machine_init");
503 machine_init();
504
505 kernel_bootstrap_log("thread_machine_init_template");
506 thread_machine_init_template();
507
508 kernel_bootstrap_log("clock_init");
509 clock_init();
510
511 /*
512 * Initialize the IPC, task, and thread subsystems.
513 */
514 #if CONFIG_THREAD_GROUPS
515 kernel_bootstrap_log("thread_group_init");
516 thread_group_init();
517 #endif
518
519 #if CONFIG_COALITIONS
520 kernel_bootstrap_log("coalitions_init");
521 coalitions_init();
522 #endif
523
524 kernel_bootstrap_log("task_init");
525 task_init();
526
527 kernel_bootstrap_log("thread_init");
528 thread_init();
529
530 kernel_bootstrap_log("restartable_init");
531 restartable_init();
532
533 kernel_bootstrap_log("workq_init");
534 workq_init();
535
536 kernel_bootstrap_log("turnstiles_init");
537 turnstiles_init();
538
539 kernel_bootstrap_log("mach_init_activity_id");
540 mach_init_activity_id();
541
542 /* Initialize the BANK Manager. */
543 kernel_bootstrap_log("bank_init");
544 bank_init();
545
546 #if CONFIG_VOUCHER_DEPRECATED
547 kernel_bootstrap_log("ipc_pthread_priority_init");
548 ipc_pthread_priority_init();
549 #endif /* CONFIG_VOUCHER_DEPRECATED */
550
551 /* initialize host_statistics */
552 host_statistics_init();
553
554 /* initialize exceptions */
555 kernel_bootstrap_log("exception_init");
556 exception_init();
557
558 #if CONFIG_SCHED_SFI
559 kernel_bootstrap_log("sfi_init");
560 sfi_init();
561 #endif
562
563 /*
564 * Create a kernel thread to execute the kernel bootstrap.
565 */
566
567 kernel_bootstrap_log("kernel_thread_create");
568 result = kernel_thread_create((thread_continue_t)kernel_bootstrap_thread, NULL, MAXPRI_KERNEL, &thread);
569
570 if (result != KERN_SUCCESS) {
571 panic("kernel_bootstrap: result = %08X", result);
572 }
573
574 /* TODO: do a proper thread_start() (without the thread_setrun()) */
575 thread->state = TH_RUN;
576 thread->last_made_runnable_time = mach_absolute_time();
577 thread_set_thread_name(thread, "kernel_bootstrap_thread");
578
579 thread_deallocate(thread);
580
581 kernel_bootstrap_log("load_context - done");
582 load_context(thread);
583 /*NOTREACHED*/
584 }
585
586 SECURITY_READ_ONLY_LATE(vm_offset_t) vm_kernel_addrperm;
587 SECURITY_READ_ONLY_LATE(vm_offset_t) buf_kernel_addrperm;
588 SECURITY_READ_ONLY_LATE(vm_offset_t) vm_kernel_addrperm_ext;
589 SECURITY_READ_ONLY_LATE(uint64_t) vm_kernel_addrhash_salt;
590 SECURITY_READ_ONLY_LATE(uint64_t) vm_kernel_addrhash_salt_ext;
591
592 extern void kdebug_lck_init(void);
593 extern int create_buffers_triage(bool);
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 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 KPC
666 kpc_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 char trace_typefilter[256] = {};
680 PE_parse_boot_arg_str("trace_typefilter", trace_typefilter,
681 sizeof(trace_typefilter));
682 #if KPERF
683 kperf_init();
684 #endif /* KPERF */
685
686 /* Init the spinlocks contained in the kd_ctrl_page_XXX
687 * structures because the kdebug system can try to use
688 * it very early. Eg. tailspind->kdbg_disable_typefilter
689 * even before create_buffers_trace() is called ie init-ing
690 * there would be too late.
691 */
692 kdebug_lck_init();
693 kdebug_init(new_nkdbufs, trace_typefilter,
694 (trace_wrap ? KDOPT_WRAPPING : 0) | KDOPT_ATBOOT);
695
696 create_buffers_triage(TRUE /*early trace*/);
697
698 kernel_startup_initialize_upto(STARTUP_SUB_SYSCTL);
699
700 #ifdef IOKIT
701 kernel_bootstrap_log("PE_init_iokit");
702 PE_init_iokit();
703 #endif
704
705 assert(ml_get_interrupts_enabled() == FALSE);
706
707 /*
708 * Past this point, kernel subsystems that expect to operate with
709 * interrupts or preemption enabled may begin enforcement.
710 */
711 kernel_startup_initialize_upto(STARTUP_SUB_EARLY_BOOT);
712
713 #if INTERRUPT_MASKED_DEBUG
714 // Reset interrupts masked timeout before we enable interrupts
715 ml_spin_debug_clear_self();
716 #endif
717 (void) spllo(); /* Allow interruptions */
718
719 /*
720 * This will start displaying progress to the user, start as early as possible
721 */
722 initialize_screen(NULL, kPEAcquireScreen);
723
724 /*
725 * Initialize the shared region module.
726 */
727 vm_commpage_init();
728 vm_commpage_text_init();
729
730 #if CONFIG_MACF
731 kernel_bootstrap_log("mac_policy_initmach");
732 mac_policy_initmach();
733 #if CONFIG_VNGUARD
734 kernel_bootstrap_log("vnguard_policy_init");
735 vnguard_policy_init();
736 #endif
737 #endif
738
739 #if CONFIG_DTRACE
740 kernel_bootstrap_log("dtrace_early_init");
741 dtrace_early_init();
742 sdt_early_init();
743 #endif
744
745 #ifndef BCM2837
746 kernel_bootstrap_log("trust_cache_init");
747 trust_cache_init();
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 /*
762 * Initialize the globals used for permuting kernel
763 * addresses that may be exported to userland as tokens
764 * using VM_KERNEL_ADDRPERM()/VM_KERNEL_ADDRPERM_EXTERNAL().
765 * Force the random number to be odd to avoid mapping a non-zero
766 * word-aligned address to zero via addition.
767 * Note: at this stage we can use the cryptographically secure PRNG
768 * rather than early_random().
769 */
770 read_random(&vm_kernel_addrperm, sizeof(vm_kernel_addrperm));
771 vm_kernel_addrperm |= 1;
772 read_random(&buf_kernel_addrperm, sizeof(buf_kernel_addrperm));
773 buf_kernel_addrperm |= 1;
774 read_random(&vm_kernel_addrperm_ext, sizeof(vm_kernel_addrperm_ext));
775 vm_kernel_addrperm_ext |= 1;
776 read_random(&vm_kernel_addrhash_salt, sizeof(vm_kernel_addrhash_salt));
777 read_random(&vm_kernel_addrhash_salt_ext, sizeof(vm_kernel_addrhash_salt_ext));
778
779 /* No changes to kernel text and rodata beyond this point. */
780 kernel_bootstrap_log("machine_lockdown");
781 machine_lockdown();
782
783 #ifdef CONFIG_XNUPOST
784 kern_return_t result = kernel_list_tests();
785 result = kernel_do_post();
786 if (result != KERN_SUCCESS) {
787 panic("kernel_do_post: Tests failed with result = 0x%08x", result);
788 }
789 kernel_bootstrap_log("kernel_do_post - done");
790 #endif /* CONFIG_XNUPOST */
791
792 #ifdef IOKIT
793 kernel_bootstrap_log("PE_lockdown_iokit");
794 PE_lockdown_iokit();
795 #endif
796 /*
797 * max_cpus must be nailed down by the time PE_lockdown_iokit() finishes,
798 * at the latest
799 */
800 vm_set_restrictions(machine_info.max_cpus);
801
802
803 #if KPERF
804 kperf_init_early();
805 #endif
806
807 /*
808 * Start the user bootstrap.
809 */
810 #ifdef MACH_BSD
811 bsd_init();
812 #endif
813
814
815 /*
816 * Get rid of pages used for early boot tracing.
817 */
818 kdebug_free_early_buf();
819
820 serial_keyboard_init(); /* Start serial keyboard if wanted */
821
822 vm_page_init_local_q(machine_info.max_cpus);
823
824 thread_bind(PROCESSOR_NULL);
825
826 /*
827 * Now that all CPUs are available to run threads, this is essentially
828 * a background thread. Take this opportunity to initialize and free
829 * any remaining vm_pages that were delayed earlier by pmap_startup().
830 */
831 vm_free_delayed_pages();
832
833 /*
834 * Become the pageout daemon.
835 */
836 vm_pageout();
837 /*NOTREACHED*/
838 }
839
840 /*
841 * slave_main:
842 *
843 * Load the first thread to start a processor.
844 * This path will also be used by the master processor
845 * after being offlined.
846 */
847 void
slave_main(void * machine_param)848 slave_main(void *machine_param)
849 {
850 processor_t processor = current_processor();
851 thread_t thread;
852
853 /*
854 * Use the idle processor thread if there
855 * is no dedicated start up thread.
856 */
857 if (processor->processor_offlined == true) {
858 /* Return to the saved processor_offline context */
859 assert(processor->startup_thread == THREAD_NULL);
860
861 thread = processor->idle_thread;
862 thread->parameter = machine_param;
863 } else if (processor->startup_thread) {
864 thread = processor->startup_thread;
865 processor->startup_thread = THREAD_NULL;
866 } else {
867 thread = processor->idle_thread;
868 thread->continuation = processor_start_thread;
869 thread->parameter = machine_param;
870 }
871
872 load_context(thread);
873 /*NOTREACHED*/
874 }
875
876 /*
877 * processor_start_thread:
878 *
879 * First thread to execute on a started processor.
880 *
881 * Called at splsched.
882 */
883 void
processor_start_thread(void * machine_param,__unused wait_result_t result)884 processor_start_thread(void *machine_param,
885 __unused wait_result_t result)
886 {
887 processor_t processor = current_processor();
888 thread_t self = current_thread();
889
890 slave_machine_init(machine_param);
891
892 /*
893 * If running the idle processor thread,
894 * reenter the idle loop, else terminate.
895 */
896 if (self == processor->idle_thread) {
897 thread_block(idle_thread);
898 }
899
900 thread_terminate(self);
901 /*NOTREACHED*/
902 }
903
904 /*
905 * load_context:
906 *
907 * Start the first thread on a processor.
908 * This may be the first thread ever run on a processor, or
909 * it could be a processor that was previously offlined.
910 */
911 static void __attribute__((noreturn))
load_context(thread_t thread)912 load_context(
913 thread_t thread)
914 {
915 processor_t processor = current_processor();
916
917
918 #define load_context_kprintf(x...) /* kprintf("load_context: " x) */
919
920 load_context_kprintf("machine_set_current_thread\n");
921 machine_set_current_thread(thread);
922
923 load_context_kprintf("processor_up\n");
924
925 PMAP_ACTIVATE_KERNEL(processor->cpu_id);
926
927 /*
928 * Acquire a stack if none attached. The panic
929 * should never occur since the thread is expected
930 * to have reserved stack.
931 */
932 load_context_kprintf("thread %p, stack %lx, stackptr %lx\n", thread,
933 thread->kernel_stack, thread->machine.kstackptr);
934 if (!thread->kernel_stack) {
935 load_context_kprintf("stack_alloc_try\n");
936 if (!stack_alloc_try(thread)) {
937 panic("load_context");
938 }
939 }
940
941 /*
942 * The idle processor threads are not counted as
943 * running for load calculations.
944 */
945 if (!(thread->state & TH_IDLE)) {
946 SCHED(run_count_incr)(thread);
947 }
948
949 processor->active_thread = thread;
950 processor_state_update_explicit(processor, thread->sched_pri,
951 SFI_CLASS_KERNEL, PSET_SMP, thread_get_perfcontrol_class(thread), THREAD_URGENCY_NONE,
952 ((thread->state & TH_IDLE) || (thread->bound_processor != PROCESSOR_NULL)) ? TH_BUCKET_SCHED_MAX : thread->th_sched_bucket);
953 processor->current_is_bound = thread->bound_processor != PROCESSOR_NULL;
954 processor->current_is_NO_SMT = false;
955 processor->current_is_eagerpreempt = false;
956 #if CONFIG_THREAD_GROUPS
957 processor->current_thread_group = thread_group_get(thread);
958 #endif
959 processor->starting_pri = thread->sched_pri;
960 processor->deadline = UINT64_MAX;
961 thread->last_processor = processor;
962 processor_up(processor);
963 processor->last_dispatch = mach_absolute_time();
964 timer_start(&thread->system_timer, processor->last_dispatch);
965 processor->thread_timer = processor->kernel_timer = &thread->system_timer;
966
967 timer_start(&processor->system_state, processor->last_dispatch);
968 processor->current_state = &processor->system_state;
969
970 #if __AMP__
971 if (processor->processor_set->pset_cluster_type == PSET_AMP_P) {
972 timer_start(&thread->ptime, processor->last_dispatch);
973 }
974 #endif
975
976 cpu_quiescent_counter_join(processor->last_dispatch);
977
978 PMAP_ACTIVATE_USER(thread, processor->cpu_id);
979
980 load_context_kprintf("machine_load_context\n");
981
982 machine_load_context(thread);
983 /*NOTREACHED*/
984 }
985
986 void
scale_setup(void)987 scale_setup(void)
988 {
989 int scale = 0;
990 #if defined(__LP64__)
991 typeof(task_max) task_max_base = task_max;
992
993 /* Raise limits for servers with >= 16G */
994 if ((serverperfmode != 0) && ((uint64_t)max_mem_actual >= (uint64_t)(16 * 1024 * 1024 * 1024ULL))) {
995 scale = (int)((uint64_t)sane_size / (uint64_t)(8 * 1024 * 1024 * 1024ULL));
996 /* limit to 128 G */
997 if (scale > 16) {
998 scale = 16;
999 }
1000 task_max_base = 2500;
1001 /* Raise limits for machines with >= 3GB */
1002 } else if ((uint64_t)max_mem_actual >= (uint64_t)(3 * 1024 * 1024 * 1024ULL)) {
1003 if ((uint64_t)max_mem_actual < (uint64_t)(8 * 1024 * 1024 * 1024ULL)) {
1004 scale = 2;
1005 } else {
1006 /* limit to 64GB */
1007 scale = MIN(16, (int)((uint64_t)max_mem_actual / (uint64_t)(4 * 1024 * 1024 * 1024ULL)));
1008 }
1009 }
1010
1011 task_max = MAX(task_max, task_max_base * scale);
1012
1013 if (scale != 0) {
1014 task_threadmax = task_max;
1015 thread_max = task_max * 5;
1016 }
1017
1018 #endif
1019
1020 bsd_scale_setup(scale);
1021 }
1022