]> git.saurik.com Git - apple/xnu.git/blob - osfmk/kern/startup.c
e67c01a8b493e48e08a70ebb61ced5f02fef83af
[apple/xnu.git] / osfmk / kern / startup.c
1 /*
2 * Copyright (c) 2000-2010 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 Software.Distribution@CS.CMU.EDU
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 <xpr_debug.h>
71 #include <mach_kdp.h>
72
73 #include <mach/boolean.h>
74 #include <mach/machine.h>
75 #include <mach/thread_act.h>
76 #include <mach/task_special_ports.h>
77 #include <mach/vm_param.h>
78 #include <ipc/ipc_init.h>
79 #include <kern/assert.h>
80 #include <kern/mach_param.h>
81 #include <kern/misc_protos.h>
82 #include <kern/clock.h>
83 #include <kern/coalition.h>
84 #include <kern/cpu_number.h>
85 #include <kern/ledger.h>
86 #include <kern/machine.h>
87 #include <kern/processor.h>
88 #include <kern/sched_prim.h>
89 #include <kern/sfi.h>
90 #include <kern/startup.h>
91 #include <kern/task.h>
92 #include <kern/thread.h>
93 #include <kern/timer.h>
94 #if CONFIG_TELEMETRY
95 #include <kern/telemetry.h>
96 #endif
97 #include <kern/wait_queue.h>
98 #include <kern/xpr.h>
99 #include <kern/zalloc.h>
100 #include <kern/locks.h>
101 #include <prng/random.h>
102 #include <console/serial_protos.h>
103 #include <vm/vm_kern.h>
104 #include <vm/vm_init.h>
105 #include <vm/vm_map.h>
106 #include <vm/vm_object.h>
107 #include <vm/vm_page.h>
108 #include <vm/vm_pageout.h>
109 #include <vm/vm_shared_region.h>
110 #include <machine/pmap.h>
111 #include <machine/commpage.h>
112 #include <libkern/version.h>
113 #include <sys/codesign.h>
114 #include <sys/kdebug.h>
115 #include <sys/random.h>
116
117 #if CONFIG_ATM
118 #include <atm/atm_internal.h>
119 #endif
120
121 #if CONFIG_CSR
122 #include <sys/csr.h>
123 #endif
124
125 #if CONFIG_BANK
126 #include <bank/bank_internal.h>
127 #endif
128
129 #if ALTERNATE_DEBUGGER
130 #include <arm64/alternate_debugger.h>
131 #endif
132
133 #if MACH_KDP
134 #include <kdp/kdp.h>
135 #endif
136
137 #if CONFIG_MACF
138 #include <security/mac_mach_internal.h>
139 #endif
140
141 #if CONFIG_COUNTERS
142 #include <pmc/pmc.h>
143 #endif
144
145 #if KPC
146 #include <kern/kpc.h>
147 #endif
148
149 #if KPERF
150 #include <kperf/kperf.h>
151 #endif
152
153 #if HYPERVISOR
154 #include <kern/hv_support.h>
155 #endif
156
157
158 #include <i386/pmCPU.h>
159 static void kernel_bootstrap_thread(void);
160
161 static void load_context(
162 thread_t thread);
163 #if (defined(__i386__) || defined(__x86_64__)) && NCOPY_WINDOWS > 0
164 extern void cpu_userwindow_init(int);
165 extern void cpu_physwindow_init(int);
166 #endif
167
168 #if CONFIG_ECC_LOGGING
169 #include <kern/ecc.h>
170 #endif
171
172 #if (defined(__i386__) || defined(__x86_64__)) && CONFIG_VMX
173 #include <i386/vmx/vmx_cpu.h>
174 #endif
175
176 // libkern/OSKextLib.cpp
177 extern void OSKextRemoveKextBootstrap(void);
178
179 void scale_setup(void);
180 extern void bsd_scale_setup(int);
181 extern unsigned int semaphore_max;
182 extern void stackshot_lock_init(void);
183
184 /*
185 * Running in virtual memory, on the interrupt stack.
186 */
187
188 extern int serverperfmode;
189
190 /* size of kernel trace buffer, disabled by default */
191 unsigned int new_nkdbufs = 0;
192 unsigned int wake_nkdbufs = 0;
193 unsigned int write_trace_on_panic = 0;
194 unsigned int trace_typefilter = 0;
195
196 /* mach leak logging */
197 int log_leaks = 0;
198 int turn_on_log_leaks = 0;
199
200 static inline void
201 kernel_bootstrap_log(const char *message)
202 {
203 // kprintf("kernel_bootstrap: %s\n", message);
204 kernel_debug_string(message);
205 }
206
207 static inline void
208 kernel_bootstrap_thread_log(const char *message)
209 {
210 // kprintf("kernel_bootstrap_thread: %s\n", message);
211 kernel_debug_string(message);
212 }
213
214 void
215 kernel_early_bootstrap(void)
216 {
217 /* serverperfmode is needed by timer setup */
218 if (PE_parse_boot_argn("serverperfmode", &serverperfmode, sizeof (serverperfmode))) {
219 serverperfmode = 1;
220 }
221
222 lck_mod_init();
223
224 /*
225 * Initialize the timer callout world
226 */
227 timer_call_init();
228
229 /*
230 * Configure SFI classes
231 */
232 sfi_early_init();
233 }
234
235
236 void
237 kernel_bootstrap(void)
238 {
239 kern_return_t result;
240 thread_t thread;
241 char namep[16];
242
243 printf("%s\n", version); /* log kernel version */
244
245 if (PE_parse_boot_argn("-l", namep, sizeof (namep))) /* leaks logging */
246 turn_on_log_leaks = 1;
247
248 PE_parse_boot_argn("trace", &new_nkdbufs, sizeof (new_nkdbufs));
249 PE_parse_boot_argn("trace_wake", &wake_nkdbufs, sizeof (wake_nkdbufs));
250 PE_parse_boot_argn("trace_panic", &write_trace_on_panic, sizeof(write_trace_on_panic));
251 PE_parse_boot_argn("trace_typefilter", &trace_typefilter, sizeof(trace_typefilter));
252
253 scale_setup();
254
255 kernel_bootstrap_log("vm_mem_bootstrap");
256 vm_mem_bootstrap();
257
258 kernel_bootstrap_log("cs_init");
259 cs_init();
260
261 kernel_bootstrap_log("vm_mem_init");
262 vm_mem_init();
263
264 machine_info.memory_size = (uint32_t)mem_size;
265 machine_info.max_mem = max_mem;
266 machine_info.major_version = version_major;
267 machine_info.minor_version = version_minor;
268
269 #if CONFIG_TELEMETRY
270 kernel_bootstrap_log("telemetry_init");
271 telemetry_init();
272 #endif
273
274 kernel_bootstrap_log("stackshot_lock_init");
275 stackshot_lock_init();
276
277 kernel_bootstrap_log("sched_init");
278 sched_init();
279
280 kernel_bootstrap_log("wait_queue_bootstrap");
281 wait_queue_bootstrap();
282
283 kernel_bootstrap_log("ipc_bootstrap");
284 ipc_bootstrap();
285
286 #if CONFIG_MACF
287 mac_policy_init();
288 #endif
289 kernel_bootstrap_log("ipc_init");
290 ipc_init();
291
292 /*
293 * As soon as the virtual memory system is up, we record
294 * that this CPU is using the kernel pmap.
295 */
296 kernel_bootstrap_log("PMAP_ACTIVATE_KERNEL");
297 PMAP_ACTIVATE_KERNEL(master_cpu);
298
299 kernel_bootstrap_log("mapping_free_prime");
300 mapping_free_prime(); /* Load up with temporary mapping blocks */
301
302 kernel_bootstrap_log("machine_init");
303 machine_init();
304
305 kernel_bootstrap_log("clock_init");
306 clock_init();
307
308 ledger_init();
309
310 /*
311 * Initialize the IPC, task, and thread subsystems.
312 */
313 #if CONFIG_COALITIONS
314 kernel_bootstrap_log("coalition_init");
315 coalition_init();
316 #endif
317
318 kernel_bootstrap_log("task_init");
319 task_init();
320
321 kernel_bootstrap_log("thread_init");
322 thread_init();
323
324 #if CONFIG_ATM
325 /* Initialize the Activity Trace Resource Manager. */
326 kernel_bootstrap_log("atm_init");
327 atm_init();
328 #endif
329
330 #if CONFIG_CSR
331 kernel_bootstrap_log("csr_init");
332 csr_init();
333 #endif
334
335 #if CONFIG_BANK
336 /* Initialize the BANK Manager. */
337 kernel_bootstrap_log("bank_init");
338 bank_init();
339 #endif
340
341 /*
342 * Create a kernel thread to execute the kernel bootstrap.
343 */
344 kernel_bootstrap_log("kernel_thread_create");
345 result = kernel_thread_create((thread_continue_t)kernel_bootstrap_thread, NULL, MAXPRI_KERNEL, &thread);
346
347 if (result != KERN_SUCCESS) panic("kernel_bootstrap: result = %08X\n", result);
348
349 thread->state = TH_RUN;
350 thread_deallocate(thread);
351
352 kernel_bootstrap_log("load_context - done");
353 load_context(thread);
354 /*NOTREACHED*/
355 }
356
357 int kth_started = 0;
358
359 vm_offset_t vm_kernel_addrperm;
360 vm_offset_t buf_kernel_addrperm;
361
362 /*
363 * Now running in a thread. Kick off other services,
364 * invoke user bootstrap, enter pageout loop.
365 */
366 static void
367 kernel_bootstrap_thread(void)
368 {
369 processor_t processor = current_processor();
370
371 #define kernel_bootstrap_thread_kprintf(x...) /* kprintf("kernel_bootstrap_thread: " x) */
372 kernel_bootstrap_thread_log("idle_thread_create");
373 /*
374 * Create the idle processor thread.
375 */
376 idle_thread_create(processor);
377
378 /*
379 * N.B. Do not stick anything else
380 * before this point.
381 *
382 * Start up the scheduler services.
383 */
384 kernel_bootstrap_thread_log("sched_startup");
385 sched_startup();
386
387 /*
388 * Thread lifecycle maintenance (teardown, stack allocation)
389 */
390 kernel_bootstrap_thread_log("thread_daemon_init");
391 thread_daemon_init();
392
393 /* Create kernel map entry reserve */
394 vm_kernel_reserved_entry_init();
395
396 /*
397 * Thread callout service.
398 */
399 kernel_bootstrap_thread_log("thread_call_initialize");
400 thread_call_initialize();
401
402 /*
403 * Remain on current processor as
404 * additional processors come online.
405 */
406 kernel_bootstrap_thread_log("thread_bind");
407 thread_bind(processor);
408
409 /*
410 * Initialize ipc thread call support.
411 */
412 kernel_bootstrap_thread_log("ipc_thread_call_init");
413 ipc_thread_call_init();
414
415 /*
416 * Kick off memory mapping adjustments.
417 */
418 kernel_bootstrap_thread_log("mapping_adjust");
419 mapping_adjust();
420
421 /*
422 * Create the clock service.
423 */
424 kernel_bootstrap_thread_log("clock_service_create");
425 clock_service_create();
426
427 /*
428 * Create the device service.
429 */
430 device_service_create();
431
432 kth_started = 1;
433
434 #if (defined(__i386__) || defined(__x86_64__)) && NCOPY_WINDOWS > 0
435 /*
436 * Create and initialize the physical copy window for processor 0
437 * This is required before starting kicking off IOKit.
438 */
439 cpu_physwindow_init(0);
440 #endif
441
442
443
444 #if MACH_KDP
445 kernel_bootstrap_log("kdp_init");
446 kdp_init();
447 #endif
448
449 #if ALTERNATE_DEBUGGER
450 alternate_debugger_init();
451 #endif
452
453 #if CONFIG_COUNTERS
454 pmc_bootstrap();
455 #endif
456
457 #if KPC
458 kpc_init();
459 #endif
460
461 #if CONFIG_ECC_LOGGING
462 ecc_log_init();
463 #endif
464
465 #if KPERF
466 kperf_bootstrap();
467 #endif
468
469 #if HYPERVISOR
470 hv_support_init();
471 #endif
472
473 #if CONFIG_TELEMETRY
474 kernel_bootstrap_log("bootprofile_init");
475 bootprofile_init();
476 #endif
477
478 #if (defined(__i386__) || defined(__x86_64__)) && CONFIG_VMX
479 vmx_init();
480 #endif
481
482 #if (defined(__i386__) || defined(__x86_64__))
483 if (turn_on_log_leaks && !new_nkdbufs)
484 new_nkdbufs = 200000;
485 if (trace_typefilter)
486 start_kern_tracing_with_typefilter(new_nkdbufs,
487 FALSE,
488 trace_typefilter);
489 else
490 start_kern_tracing(new_nkdbufs, FALSE);
491 if (turn_on_log_leaks)
492 log_leaks = 1;
493
494 #endif
495
496 kernel_bootstrap_log("prng_init");
497 prng_cpu_init(master_cpu);
498
499 #ifdef IOKIT
500 PE_init_iokit();
501 #endif
502
503 assert(ml_get_interrupts_enabled() == FALSE);
504 (void) spllo(); /* Allow interruptions */
505
506 #if (defined(__i386__) || defined(__x86_64__)) && NCOPY_WINDOWS > 0
507 /*
508 * Create and initialize the copy window for processor 0
509 * This also allocates window space for all other processors.
510 * However, this is dependent on the number of processors - so this call
511 * must be after IOKit has been started because IOKit performs processor
512 * discovery.
513 */
514 cpu_userwindow_init(0);
515 #endif
516
517 #if (!defined(__i386__) && !defined(__x86_64__))
518 if (turn_on_log_leaks && !new_nkdbufs)
519 new_nkdbufs = 200000;
520 if (trace_typefilter)
521 start_kern_tracing_with_typefilter(new_nkdbufs, FALSE, trace_typefilter);
522 else
523 start_kern_tracing(new_nkdbufs, FALSE);
524 if (turn_on_log_leaks)
525 log_leaks = 1;
526 #endif
527
528 /*
529 * Initialize the shared region module.
530 */
531 vm_shared_region_init();
532 vm_commpage_init();
533 vm_commpage_text_init();
534
535
536 #if CONFIG_MACF
537 kernel_bootstrap_log("mac_policy_initmach");
538 mac_policy_initmach();
539 #endif
540
541 kernel_bootstrap_log("sfi_init");
542 sfi_init();
543
544 /*
545 * Initialize the global used for permuting kernel
546 * addresses that may be exported to userland as tokens
547 * using VM_KERNEL_ADDRPERM(). Force the random number
548 * to be odd to avoid mapping a non-zero
549 * word-aligned address to zero via addition.
550 * Note: at this stage we can use the cryptographically secure PRNG
551 * rather than early_random().
552 */
553 read_random(&vm_kernel_addrperm, sizeof(vm_kernel_addrperm));
554 vm_kernel_addrperm |= 1;
555 read_random(&buf_kernel_addrperm, sizeof(buf_kernel_addrperm));
556 buf_kernel_addrperm |= 1;
557
558 /*
559 * Start the user bootstrap.
560 */
561 #ifdef MACH_BSD
562 bsd_init();
563 #endif
564
565 /*
566 * Get rid of segments used to bootstrap kext loading. This removes
567 * the KLD, PRELINK symtab, LINKEDIT, and symtab segments/load commands.
568 */
569 OSKextRemoveKextBootstrap();
570
571 serial_keyboard_init(); /* Start serial keyboard if wanted */
572
573 vm_page_init_local_q();
574
575 thread_bind(PROCESSOR_NULL);
576
577 /*
578 * Become the pageout daemon.
579 */
580 vm_pageout();
581 /*NOTREACHED*/
582 }
583
584 /*
585 * slave_main:
586 *
587 * Load the first thread to start a processor.
588 */
589 void
590 slave_main(void *machine_param)
591 {
592 processor_t processor = current_processor();
593 thread_t thread;
594
595 /*
596 * Use the idle processor thread if there
597 * is no dedicated start up thread.
598 */
599 if (processor->next_thread == THREAD_NULL) {
600 thread = processor->idle_thread;
601 thread->continuation = (thread_continue_t)processor_start_thread;
602 thread->parameter = machine_param;
603 }
604 else {
605 thread = processor->next_thread;
606 processor->next_thread = THREAD_NULL;
607 }
608
609 load_context(thread);
610 /*NOTREACHED*/
611 }
612
613 /*
614 * processor_start_thread:
615 *
616 * First thread to execute on a started processor.
617 *
618 * Called at splsched.
619 */
620 void
621 processor_start_thread(void *machine_param)
622 {
623 processor_t processor = current_processor();
624 thread_t self = current_thread();
625
626 slave_machine_init(machine_param);
627
628 /*
629 * If running the idle processor thread,
630 * reenter the idle loop, else terminate.
631 */
632 if (self == processor->idle_thread)
633 thread_block((thread_continue_t)idle_thread);
634
635 thread_terminate(self);
636 /*NOTREACHED*/
637 }
638
639 /*
640 * load_context:
641 *
642 * Start the first thread on a processor.
643 */
644 static void
645 load_context(
646 thread_t thread)
647 {
648 processor_t processor = current_processor();
649
650
651 #define load_context_kprintf(x...) /* kprintf("load_context: " x) */
652
653 load_context_kprintf("machine_set_current_thread\n");
654 machine_set_current_thread(thread);
655
656 load_context_kprintf("processor_up\n");
657 processor_up(processor);
658
659 PMAP_ACTIVATE_KERNEL(processor->cpu_id);
660
661 /*
662 * Acquire a stack if none attached. The panic
663 * should never occur since the thread is expected
664 * to have reserved stack.
665 */
666 load_context_kprintf("thread %p, stack %lx, stackptr %lx\n", thread,
667 thread->kernel_stack, thread->machine.kstackptr);
668 if (!thread->kernel_stack) {
669 load_context_kprintf("stack_alloc_try\n");
670 if (!stack_alloc_try(thread))
671 panic("load_context");
672 }
673
674 /*
675 * The idle processor threads are not counted as
676 * running for load calculations.
677 */
678 if (!(thread->state & TH_IDLE))
679 sched_run_incr(thread);
680
681 processor->active_thread = thread;
682 processor->current_pri = thread->sched_pri;
683 processor->current_thmode = thread->sched_mode;
684 processor->deadline = UINT64_MAX;
685 thread->last_processor = processor;
686
687 processor->last_dispatch = mach_absolute_time();
688 timer_start(&thread->system_timer, processor->last_dispatch);
689 PROCESSOR_DATA(processor, thread_timer) = PROCESSOR_DATA(processor, kernel_timer) = &thread->system_timer;
690
691 timer_start(&PROCESSOR_DATA(processor, system_state), processor->last_dispatch);
692 PROCESSOR_DATA(processor, current_state) = &PROCESSOR_DATA(processor, system_state);
693
694 PMAP_ACTIVATE_USER(thread, processor->cpu_id);
695
696 load_context_kprintf("machine_load_context\n");
697 machine_load_context(thread);
698 /*NOTREACHED*/
699 }
700
701 void
702 scale_setup()
703 {
704 int scale = 0;
705 #if defined(__LP64__)
706 typeof(task_max) task_max_base = task_max;
707
708 /* Raise limits for servers with >= 16G */
709 if ((serverperfmode != 0) && ((uint64_t)sane_size >= (uint64_t)(16 * 1024 * 1024 *1024ULL))) {
710 scale = (int)((uint64_t)sane_size / (uint64_t)(8 * 1024 * 1024 *1024ULL));
711 /* limit to 128 G */
712 if (scale > 16)
713 scale = 16;
714 task_max_base = 2500;
715 } else if ((uint64_t)sane_size >= (uint64_t)(3 * 1024 * 1024 *1024ULL))
716 scale = 2;
717
718 task_max = MAX(task_max, task_max_base * scale);
719
720 if (scale != 0) {
721 task_threadmax = task_max;
722 thread_max = task_max * 5;
723 }
724
725 #endif
726
727 bsd_scale_setup(scale);
728
729 ipc_space_max = SPACE_MAX;
730 ipc_port_max = PORT_MAX;
731 ipc_pset_max = SET_MAX;
732 semaphore_max = SEMAPHORE_MAX;
733 }
734