xref: /netbsd-src/sys/kern/init_main.c (revision 4724848cf0da353df257f730694b7882798e5daf)
1 /*	$NetBSD: init_main.c,v 1.541 2022/10/26 23:20:47 riastradh Exp $	*/
2 
3 /*-
4  * Copyright (c) 2008, 2009, 2019 The NetBSD Foundation, Inc.
5  * All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  *
16  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
17  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
18  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
19  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
20  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
21  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
22  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
23  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
24  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
25  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
26  * POSSIBILITY OF SUCH DAMAGE.
27  */
28 
29 /*
30  * Copyright (c) 1982, 1986, 1989, 1991, 1992, 1993
31  *	The Regents of the University of California.  All rights reserved.
32  * (c) UNIX System Laboratories, Inc.
33  * All or some portions of this file are derived from material licensed
34  * to the University of California by American Telephone and Telegraph
35  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
36  * the permission of UNIX System Laboratories, Inc.
37  *
38  * Redistribution and use in source and binary forms, with or without
39  * modification, are permitted provided that the following conditions
40  * are met:
41  * 1. Redistributions of source code must retain the above copyright
42  *    notice, this list of conditions and the following disclaimer.
43  * 2. Redistributions in binary form must reproduce the above copyright
44  *    notice, this list of conditions and the following disclaimer in the
45  *    documentation and/or other materials provided with the distribution.
46  * 3. Neither the name of the University nor the names of its contributors
47  *    may be used to endorse or promote products derived from this software
48  *    without specific prior written permission.
49  *
50  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
51  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
52  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
53  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
54  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
55  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
56  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
57  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
58  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
59  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
60  * SUCH DAMAGE.
61  *
62  *	@(#)init_main.c	8.16 (Berkeley) 5/14/95
63  */
64 
65 /*
66  * Copyright (c) 1995 Christopher G. Demetriou.  All rights reserved.
67  *
68  * Redistribution and use in source and binary forms, with or without
69  * modification, are permitted provided that the following conditions
70  * are met:
71  * 1. Redistributions of source code must retain the above copyright
72  *    notice, this list of conditions and the following disclaimer.
73  * 2. Redistributions in binary form must reproduce the above copyright
74  *    notice, this list of conditions and the following disclaimer in the
75  *    documentation and/or other materials provided with the distribution.
76  * 3. All advertising materials mentioning features or use of this software
77  *    must display the following acknowledgement:
78  *	This product includes software developed by the University of
79  *	California, Berkeley and its contributors.
80  * 4. Neither the name of the University nor the names of its contributors
81  *    may be used to endorse or promote products derived from this software
82  *    without specific prior written permission.
83  *
84  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
85  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
86  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
87  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
88  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
89  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
90  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
91  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
92  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
93  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
94  * SUCH DAMAGE.
95  *
96  *	@(#)init_main.c	8.16 (Berkeley) 5/14/95
97  */
98 
99 #include <sys/cdefs.h>
100 __KERNEL_RCSID(0, "$NetBSD: init_main.c,v 1.541 2022/10/26 23:20:47 riastradh Exp $");
101 
102 #include "opt_cnmagic.h"
103 #include "opt_ddb.h"
104 #include "opt_inet.h"
105 #include "opt_ipsec.h"
106 #include "opt_modular.h"
107 #include "opt_ntp.h"
108 #include "opt_pipe.h"
109 #include "opt_syscall_debug.h"
110 #include "opt_sysv.h"
111 #include "opt_fileassoc.h"
112 #include "opt_ktrace.h"
113 #include "opt_pax.h"
114 #include "opt_compat_netbsd.h"
115 #include "opt_ptrace.h"
116 #include "opt_splash.h"
117 #include "opt_kernhist.h"
118 #include "opt_gprof.h"
119 
120 #if defined(SPLASHSCREEN) && defined(makeoptions_SPLASHSCREEN_IMAGE)
121 extern void *_binary_splash_image_start;
122 extern void *_binary_splash_image_end;
123 #endif
124 
125 #include "ksyms.h"
126 
127 #include "veriexec.h"
128 
129 #include <sys/param.h>
130 #include <sys/acct.h>
131 #include <sys/filedesc.h>
132 #include <sys/file.h>
133 #include <sys/errno.h>
134 #include <sys/callout.h>
135 #include <sys/cpu.h>
136 #include <sys/cpufreq.h>
137 #include <sys/spldebug.h>
138 #include <sys/kernel.h>
139 #include <sys/mount.h>
140 #include <sys/proc.h>
141 #include <sys/lwp.h>
142 #include <sys/kthread.h>
143 #include <sys/resourcevar.h>
144 #include <sys/signalvar.h>
145 #include <sys/systm.h>
146 #include <sys/vnode.h>
147 #include <sys/fstrans.h>
148 #include <sys/tty.h>
149 #include <sys/conf.h>
150 #include <sys/disklabel.h>
151 #include <sys/buf.h>
152 #include <sys/device.h>
153 #include <sys/exec.h>
154 #include <sys/socketvar.h>
155 #include <sys/protosw.h>
156 #include <sys/percpu.h>
157 #include <sys/pserialize.h>
158 #include <sys/pset.h>
159 #include <sys/sysctl.h>
160 #include <sys/reboot.h>
161 #include <sys/event.h>
162 #include <sys/mbuf.h>
163 #include <sys/sched.h>
164 #include <sys/sleepq.h>
165 #include <sys/ipi.h>
166 #include <sys/iostat.h>
167 #include <sys/vmem.h>
168 #include <sys/uuid.h>
169 #include <sys/extent.h>
170 #include <sys/disk.h>
171 #include <sys/msgbuf.h>
172 #include <sys/module.h>
173 #include <sys/module_hook.h>
174 #include <sys/event.h>
175 #include <sys/lockf.h>
176 #include <sys/once.h>
177 #include <sys/kcpuset.h>
178 #include <sys/ksyms.h>
179 #include <sys/uidinfo.h>
180 #include <sys/kprintf.h>
181 #include <sys/bufq.h>
182 #include <sys/threadpool.h>
183 #include <sys/futex.h>
184 #ifdef IPSEC
185 #include <netipsec/ipsec.h>
186 #endif
187 #include <sys/domain.h>
188 #include <sys/namei.h>
189 #include <sys/rnd.h>
190 #include <sys/pipe.h>
191 #if NVERIEXEC > 0
192 #include <sys/verified_exec.h>
193 #endif /* NVERIEXEC > 0 */
194 #ifdef KTRACE
195 #include <sys/ktrace.h>
196 #endif
197 #include <sys/kauth.h>
198 #include <net80211/ieee80211_netbsd.h>
199 #include <sys/cprng.h>
200 #include <sys/psref.h>
201 #include <sys/radixtree.h>
202 
203 #include <sys/syscall.h>
204 #include <sys/syscallargs.h>
205 
206 #include <sys/pax.h>
207 
208 #include <dev/clock_subr.h>
209 
210 #include <secmodel/secmodel.h>
211 
212 #include <ufs/ufs/quota.h>
213 
214 #include <miscfs/genfs/genfs.h>
215 #include <miscfs/specfs/specdev.h>
216 
217 #include <sys/cpu.h>
218 
219 #include <uvm/uvm.h>	/* extern struct uvm uvm */
220 
221 #include <dev/cons.h>
222 #include <dev/splash/splash.h>
223 
224 #include <net/bpf.h>
225 #include <net/if.h>
226 #include <net/pfil.h>
227 #include <net/raw_cb.h>
228 #include <net/if_llatbl.h>
229 
230 #include <prop/proplib.h>
231 
232 #include <sys/userconf.h>
233 
234 extern time_t rootfstime;
235 
236 #ifndef curlwp
237 struct	lwp *curlwp = &lwp0;
238 #endif
239 struct	proc *initproc;
240 
241 struct	vnode *rootvp, *swapdev_vp;
242 int	boothowto;
243 int	cold __read_mostly = 1;		/* still working on startup */
244 int	shutting_down __read_mostly;	/* system is shutting down */
245 
246 int	start_init_exec;		/* semaphore for start_init() */
247 
248 static void check_console(struct lwp *l);
249 static void start_init(void *);
250 static void configure(void);
251 static void configure2(void);
252 static void configure3(void);
253 void main(void);
254 
255 /*
256  * System startup; initialize the world, create process 0, mount root
257  * filesystem, and fork to create init and pagedaemon.  Most of the
258  * hard work is done in the lower-level initialization routines including
259  * startup(), which does memory initialization and autoconfiguration.
260  */
261 void
262 main(void)
263 {
264 	struct timespec time;
265 	struct lwp *l;
266 	struct proc *p;
267 	int s, error;
268 #ifdef NVNODE_IMPLICIT
269 	int usevnodes;
270 #endif
271 	CPU_INFO_ITERATOR cii;
272 	struct cpu_info *ci;
273 
274 #ifdef DIAGNOSTIC
275 	/*
276 	 * Verify that CPU_INFO_FOREACH() knows about the boot CPU
277 	 * and only the boot CPU at this point.
278 	 */
279 	int cpucount = 0;
280 	for (CPU_INFO_FOREACH(cii, ci)) {
281 		KASSERT(ci == curcpu());
282 		cpucount++;
283 	}
284 	KASSERT(cpucount == 1);
285 #endif
286 
287 	l = &lwp0;
288 #ifndef LWP0_CPU_INFO
289 	l->l_cpu = curcpu();
290 #endif
291 	l->l_pflag |= LP_RUNNING;
292 
293 	/*
294 	 * Attempt to find console and initialize
295 	 * in case of early panic or other messages.
296 	 */
297 	consinit();
298 #ifdef CNMAGIC
299 	cn_set_magic(CNMAGIC);
300 #endif
301 
302 	kernel_lock_init();
303 	once_init();
304 	todr_init();
305 
306 	mi_cpu_init();
307 	kernconfig_lock_init();
308 	kthread_sysinit();
309 
310 	/* Initialize the device switch tables. */
311 	devsw_init();
312 
313 	/* Initialize event counters. */
314 	evcnt_init();
315 
316 	uvm_init();
317 	ubchist_init();
318 	kcpuset_sysinit();
319 
320 	prop_kern_init();
321 
322 #if ((NKSYMS > 0) || (NDDB > 0) || (NMODULAR > 0))
323 	ksyms_init();
324 #endif
325 	kprintf_init();
326 
327 	percpu_init();
328 
329 	/* Initialize lock caches. */
330 	mutex_obj_init();
331 
332 	/* Initialize radix trees (used by numerous subsystems). */
333 	radix_tree_init();
334 
335 	/* Passive serialization. */
336 	pserialize_init();
337 
338 	/* Initialize the extent manager. */
339 	extent_init();
340 
341 	/* Do machine-dependent initialization. */
342 	cpu_startup();
343 
344 	/* Initialize the sysctl subsystem. */
345 	sysctl_init();
346 
347 	/* Initialize callouts, part 1. */
348 	callout_startup();
349 
350 	/* Initialize the kernel authorization subsystem. */
351 	kauth_init();
352 
353 	secmodel_init();
354 
355 	spec_init();
356 
357 	/*
358 	 * Set BPF op vector.  Can't do this in bpf attach, since
359 	 * network drivers attach before bpf.
360 	 */
361 	bpf_setops();
362 
363 	/* Initialize what we can in ipi(9) before CPUs are detected. */
364 	ipi_sysinit();
365 
366 	/* Start module system. */
367 	module_init();
368 	module_hook_init();
369 
370 	/*
371 	 * Initialize the kernel authorization subsystem and start the
372 	 * default security model, if any. We need to do this early
373 	 * enough so that subsystems relying on any of the aforementioned
374 	 * can work properly. Since the security model may dictate the
375 	 * credential inheritance policy, it is needed at least before
376 	 * any process is created, specifically proc0.
377 	 */
378 	module_init_class(MODULE_CLASS_SECMODEL);
379 
380 	/* Initialize the buffer cache */
381 	bufinit();
382 	biohist_init();
383 
384 #ifdef KERNHIST
385 	sysctl_kernhist_init();
386 #endif
387 
388 
389 #if defined(SPLASHSCREEN) && defined(SPLASHSCREEN_IMAGE)
390 	size_t splash_size = (&_binary_splash_image_end -
391 	    &_binary_splash_image_start) * sizeof(void *);
392 	splash_setimage(&_binary_splash_image_start, splash_size);
393 #endif
394 
395 	/* Initialize sockets. */
396 	soinit();
397 
398 	/*
399 	 * The following things must be done before autoconfiguration.
400 	 */
401 	rnd_init();		/* initialize entropy pool */
402 
403 	cprng_init();		/* initialize cryptographic PRNG */
404 
405 	/* Initialize process and pgrp structures. */
406 	procinit();
407 	lwpinit();
408 
409 	/* Must be called after lwpinit (lwpinit_specificdata) */
410 	psref_init();
411 
412 	/* Initialize signal-related data structures. */
413 	signal_init();
414 
415 	/* Initialize resource management. */
416 	resource_init();
417 
418 	/* Create process 0. */
419 	proc0_init();
420 	lwp0_init();
421 
422 	/* Disable preemption during boot. */
423 	kpreempt_disable();
424 
425 	/* Initialize the threadpool system. */
426 	threadpools_init();
427 
428 	/* Initialize the UID hash table. */
429 	uid_init();
430 
431 	/* Charge root for one process. */
432 	(void)chgproccnt(0, 1);
433 
434 	/* Initialize the run queues, turnstiles and sleep queues. */
435 	sched_rqinit();
436 	turnstile_init();
437 	sleeptab_init(&sleeptab);
438 
439 	sched_init();
440 
441 	/* Initialize processor-sets */
442 	psets_init();
443 
444 	/* Initialize cpufreq(9) */
445 	cpufreq_init();
446 
447 	/* MI initialization of the boot cpu */
448 	error = mi_cpu_attach(curcpu());
449 	KASSERT(error == 0);
450 
451 	/* Initialize timekeeping. */
452 	time_init();
453 
454 	/*
455 	 * Initialize mbuf's.  Do this now because we might attempt to
456 	 * allocate mbufs or mbuf clusters during autoconfiguration.
457 	 */
458 	mbinit();
459 
460 	/* Initialize I/O statistics. */
461 	iostat_init();
462 
463 	/* Initialize the log device. */
464 	loginit();
465 
466 	/* Second part of module system initialization. */
467 	module_start_unload_thread();
468 
469 	/* Initialize autoconf data structures before any modules are loaded */
470 	config_init_mi();
471 
472 	/* Initialize the file systems. */
473 #ifdef NVNODE_IMPLICIT
474 	/*
475 	 * If maximum number of vnodes in namei vnode cache is not explicitly
476 	 * defined in kernel config, adjust the number such as we use roughly
477 	 * 10% of memory for vnodes and associated data structures in the
478 	 * assumed worst case.  Do not provide fewer than NVNODE vnodes.
479 	 */
480 	usevnodes = calc_cache_size(vmem_size(kmem_arena, VMEM_FREE|VMEM_ALLOC),
481 	    10, VNODE_KMEM_MAXPCT) / VNODE_COST;
482 	if (usevnodes > desiredvnodes)
483 		desiredvnodes = usevnodes;
484 #endif /* NVNODE_IMPLICIT */
485 #ifdef MAXFILES_IMPLICIT
486 	/*
487 	 * If maximum number of files is not explicitly defined in
488 	 * kernel config, adjust the number so that it is somewhat
489 	 * more reasonable on machines with larger memory sizes.
490 	 * Arbitrary numbers are 20,000 files for 16GB RAM or more
491 	 * and 10,000 files for 1GB RAM or more.
492 	 *
493 	 * XXXtodo: adjust this and other values totally dynamically
494 	 */
495 	if (ctob((uint64_t)physmem) >= 16ULL * 1024 * 1024 * 1024)
496 		maxfiles = MAX(maxfiles, 20000);
497 	if (ctob((uint64_t)physmem) >= 1024 * 1024 * 1024)
498 		maxfiles = MAX(maxfiles, 10000);
499 #endif /* MAXFILES_IMPLICIT */
500 
501 	/* Initialize fstrans. */
502 	fstrans_init();
503 
504 	vfsinit();
505 	lf_init();
506 
507 	/* Initialize the file descriptor system. */
508 	fd_sys_init();
509 
510 	/* Initialize cwd structures */
511 	cwd_sys_init();
512 
513 	/* Initialize kqueue. */
514 	kqueue_init();
515 
516 	inittimecounter();
517 	ntp_init();
518 
519 	/* Initialize tty subsystem. */
520 	tty_init();
521 	ttyldisc_init();
522 
523 	/* Initialize the buffer cache, part 2. */
524 	bufinit2();
525 
526 	/* Initialize the disk wedge subsystem. */
527 	dkwedge_init();
528 
529 	/* Initialize pfil */
530 	pfil_init();
531 
532 	/* Initialize interfaces. */
533 	ifinit1();
534 
535 	spldebug_start();
536 
537 	/* Initialize sockets thread(s) */
538 	soinit1();
539 
540 	/*
541 	 * Initialize the bufq strategy sub-system and any built-in
542 	 * strategy modules - they may be needed by some devices during
543 	 * auto-configuration
544 	 */
545 	bufq_init();
546 	module_init_class(MODULE_CLASS_BUFQ);
547 
548 	/* Configure the system hardware.  This will enable interrupts. */
549 	configure();
550 #ifdef __HAVE_LEGACY_INTRCNT
551 	evcnt_attach_legacy_intrcnt();
552 #endif
553 
554 	/* Enable deferred processing of RNG samples */
555 	rnd_init_softint();
556 
557 	/* Once all CPUs are detected, initialize the per-CPU cprng_fast.  */
558 	cprng_fast_init();
559 
560 	ssp_init();
561 
562 	ubc_init();		/* must be after autoconfig */
563 
564 	mm_init();
565 
566 	configure2();
567 
568 	/* Initialize the rest of ipi(9) after CPUs have been detected. */
569 	ipi_percpu_init();
570 
571 	futex_sys_init();
572 
573 	/* Now timer is working.  Enable preemption. */
574 	kpreempt_enable();
575 
576 	/* Get the threads going and into any sleeps before continuing. */
577 	yield();
578 
579 	vmem_rehash_start();	/* must be before exec_init */
580 
581 	/* Initialize exec structures */
582 	exec_init(1);		/* seminit calls exithook_establish() */
583 
584 #if NVERIEXEC > 0
585 	/*
586 	 * Initialise the Veriexec subsystem.
587 	 */
588 	veriexec_init();
589 #endif /* NVERIEXEC > 0 */
590 
591 	pax_init();
592 
593 #ifdef	IPSEC
594 	/* Attach network crypto subsystem */
595 	ipsec_attach();
596 #endif
597 
598 	/*
599 	 * Initialize protocols.  Block reception of incoming packets
600 	 * until everything is ready.
601 	 */
602 	s = splnet();
603 	ifinit();
604 #if defined(INET) || defined(INET6)
605 	lltableinit();
606 #endif
607 	domaininit(true);
608 	ifinit_post();
609 	if_attachdomain();
610 	splx(s);
611 
612 #ifdef GPROF
613 	/* Initialize kernel profiling. */
614 	kmstartup();
615 #endif
616 
617 	/* Initialize system accounting. */
618 	acct_init();
619 
620 #ifndef PIPE_SOCKETPAIR
621 	/* Initialize pipes. */
622 	pipe_init();
623 #endif
624 
625 #ifdef KTRACE
626 	/* Initialize ktrace. */
627 	ktrinit();
628 #endif
629 
630 	machdep_init();
631 
632 	procinit_sysctl();
633 
634 	scdebug_init();
635 
636 	/*
637 	 * Create process 1 (init(8)).  We do this now, as Unix has
638 	 * historically had init be process 1, and changing this would
639 	 * probably upset a lot of people.
640 	 *
641 	 * Note that process 1 won't immediately exec init(8), but will
642 	 * wait for us to inform it that the root file system has been
643 	 * mounted.
644 	 */
645 	if (fork1(l, 0, SIGCHLD, NULL, 0, start_init, NULL, NULL))
646 		panic("fork init");
647 
648 	/*
649 	 * The initproc variable cannot be initialized in start_init as there
650 	 * is a race between vfs_mountroot and start_init.
651 	 */
652 	mutex_enter(&proc_lock);
653 	initproc = proc_find_raw(1);
654 	mutex_exit(&proc_lock);
655 
656 	/*
657 	 * Load any remaining builtin modules, and hand back temporary
658 	 * storage to the VM system.  Then require force when loading any
659 	 * remaining un-init'ed built-in modules to avoid later surprises.
660 	 */
661 	module_init_class(MODULE_CLASS_ANY);
662 	module_builtin_require_force();
663 
664 	/*
665 	 * Finalize configuration now that all real devices have been
666 	 * found.  This needs to be done before the root device is
667 	 * selected, since finalization may create the root device.
668 	 */
669 	config_finalize();
670 
671 	sysctl_finalize();
672 
673 	/*
674 	 * Now that autoconfiguration has completed, we can determine
675 	 * the root and dump devices.
676 	 */
677 	cpu_rootconf();
678 	cpu_dumpconf();
679 
680 	/* Mount the root file system. */
681 	do {
682 		domountroothook(root_device);
683 		if ((error = vfs_mountroot())) {
684 			printf("cannot mount root, error = %d\n", error);
685 			boothowto |= RB_ASKNAME;
686 			setroot(root_device,
687 			    (rootdev != NODEV) ? DISKPART(rootdev) : 0);
688 		}
689 	} while (error != 0);
690 	mountroothook_destroy();
691 
692 	configure3();
693 
694 	/*
695 	 * Initialise the time-of-day clock, passing the time recorded
696 	 * in the root filesystem (if any) for use by systems that
697 	 * don't have a non-volatile time-of-day device.
698 	 */
699 	inittodr(rootfstime);
700 
701 	/*
702 	 * Now can look at time, having had a chance to verify the time
703 	 * from the file system.  Reset l->l_rtime as it may have been
704 	 * munched in mi_switch() after the time got set.
705 	 */
706 	getnanotime(&time);
707 
708 	mutex_enter(&proc_lock);
709 	LIST_FOREACH(p, &allproc, p_list) {
710 		KASSERT((p->p_flag & PK_MARKER) == 0);
711 		mutex_enter(p->p_lock);
712 		TIMESPEC_TO_TIMEVAL(&p->p_stats->p_start, &time);
713 		LIST_FOREACH(l, &p->p_lwps, l_sibling) {
714 			lwp_lock(l);
715 			memset(&l->l_rtime, 0, sizeof(l->l_rtime));
716 			lwp_unlock(l);
717 		}
718 		mutex_exit(p->p_lock);
719 	}
720 	mutex_exit(&proc_lock);
721 	binuptime(&curlwp->l_stime);
722 
723 	for (CPU_INFO_FOREACH(cii, ci)) {
724 		ci->ci_schedstate.spc_lastmod = time_second;
725 	}
726 
727 	/* Create the pageout daemon kernel thread. */
728 	uvm_swap_init();
729 	if (kthread_create(PRI_PGDAEMON, KTHREAD_MPSAFE, NULL, uvm_pageout,
730 	    NULL, NULL, "pgdaemon"))
731 		panic("fork pagedaemon");
732 
733 	/* Create the filesystem syncer kernel thread. */
734 	if (kthread_create(PRI_IOFLUSH, KTHREAD_MPSAFE, NULL, sched_sync,
735 	    NULL, NULL, "ioflush"))
736 		panic("fork syncer");
737 
738 	/* Wait for final configure threads to complete. */
739 	config_finalize_mountroot();
740 
741 	/*
742 	 * Okay, now we can let init(8) exec!  It's off to userland!
743 	 */
744 	mutex_enter(&proc_lock);
745 	start_init_exec = 1;
746 	cv_broadcast(&lbolt);
747 	mutex_exit(&proc_lock);
748 
749 	/* The scheduler is an infinite loop. */
750 	uvm_scheduler();
751 	/* NOTREACHED */
752 }
753 
754 /*
755  * Configure the system's hardware.
756  */
757 static void
758 configure(void)
759 {
760 
761 	/*
762 	 * XXX
763 	 * callout_setfunc() requires mutex(9) so it can't be in config_init()
764 	 * on amiga and atari which use config_init() and autoconf(9) functions
765 	 * to initialize console devices.
766 	 */
767 	config_twiddle_init();
768 
769 	pmf_init();
770 
771 	/* Initialize driver modules */
772 	module_init_class(MODULE_CLASS_DRIVER);
773 
774 	userconf_init();
775 	if (boothowto & RB_USERCONF)
776 		userconf_prompt();
777 
778 	if ((boothowto & (AB_SILENT|AB_VERBOSE)) == AB_SILENT) {
779 		printf_nolog("Detecting hardware...");
780 	}
781 
782 	/*
783 	 * Do the machine-dependent portion of autoconfiguration.  This
784 	 * sets the configuration machinery here in motion by "finding"
785 	 * the root bus.  When this function returns, we expect interrupts
786 	 * to be enabled.
787 	 */
788 	cpu_configure();
789 }
790 
791 static void
792 configure2(void)
793 {
794 	CPU_INFO_ITERATOR cii;
795 	struct cpu_info *ci;
796 	int s;
797 
798 	/* Fix up CPU topology info, which has all been collected by now. */
799 	cpu_topology_init();
800 
801 	/*
802 	 * Now that we've found all the hardware, start the real time
803 	 * and statistics clocks.
804 	 */
805 	initclocks();
806 
807 	cold = 0;	/* clocks are running, we're warm now! */
808 	s = splsched();
809 	curcpu()->ci_schedstate.spc_flags |= SPCF_RUNNING;
810 	splx(s);
811 
812 	/* Setup the runqueues and scheduler. */
813 	runq_init();
814 	synch_init();
815 
816 	/* Boot the secondary processors. */
817 	for (CPU_INFO_FOREACH(cii, ci)) {
818 		uvm_cpu_attach(ci);
819 	}
820 
821 	/* Decide how to partition free memory. */
822 	uvm_page_rebucket();
823 
824 	mp_online = true;
825 #if defined(MULTIPROCESSOR)
826 	cpu_boot_secondary_processors();
827 #endif
828 
829 	/*
830 	 * Bus scans can make it appear as if the system has paused, so
831 	 * twiddle constantly while config_interrupts() jobs are running.
832 	 */
833 	config_twiddle_fn(NULL);
834 
835 	/*
836 	 * Create threads to call back and finish configuration for
837 	 * devices that want interrupts enabled.
838 	 */
839 	config_create_interruptthreads();
840 }
841 
842 static void
843 configure3(void)
844 {
845 
846 	/*
847 	 * Create threads to call back and finish configuration for
848 	 * devices that want the mounted root file system.
849 	 */
850 	config_create_mountrootthreads();
851 
852 	/* Get the threads going and into any sleeps before continuing. */
853 	yield();
854 }
855 
856 static void
857 rootconf_handle_wedges(void)
858 {
859 	struct disklabel label;
860 	struct partition *p;
861 	struct vnode *vp;
862 	daddr_t startblk;
863 	uint64_t nblks;
864 	device_t dev;
865 	int error;
866 
867 	if (booted_nblks) {
868 		/*
869 		 * bootloader passed geometry
870 		 */
871 		dev      = booted_device;
872 		startblk = booted_startblk;
873 		nblks    = booted_nblks;
874 
875 		/*
876 		 * keep booted_device and booted_partition
877 		 * in case the kernel doesn't identify a wedge
878 		 */
879 	} else {
880 		/*
881 		 * bootloader passed partition number
882 		 *
883 		 * We cannot ask the partition device directly when it is
884 		 * covered by a wedge. Instead we look up the geometry in
885 		 * the disklabel.
886 		 */
887 		vp = opendisk(booted_device);
888 
889 		if (vp == NULL)
890 			return;
891 
892 		VOP_UNLOCK(vp);
893 		error = VOP_IOCTL(vp, DIOCGDINFO, &label, FREAD, NOCRED);
894 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
895 		VOP_CLOSE(vp, FREAD, NOCRED);
896 		vput(vp);
897 		if (error)
898 			return;
899 
900 		KASSERT(booted_partition >= 0
901 			&& booted_partition < MAXPARTITIONS);
902 
903 		p = &label.d_partitions[booted_partition];
904 
905 		dev      = booted_device;
906 		startblk = p->p_offset;
907 		nblks    = p->p_size;
908 	}
909 
910 	dev = dkwedge_find_partition(dev, startblk, nblks);
911 	if (dev != NULL) {
912 		booted_device = dev;
913 		booted_partition = 0;
914 	}
915 }
916 
917 void
918 rootconf(void)
919 {
920 	if (booted_device != NULL)
921 		rootconf_handle_wedges();
922 
923 	setroot(booted_device, booted_partition);
924 }
925 
926 static void
927 check_console(struct lwp *l)
928 {
929 	struct vnode *vp;
930 	int error;
931 
932 	error = namei_simple_kernel("/dev/console",
933 				NSM_FOLLOW_NOEMULROOT, &vp);
934 	if (error == 0) {
935 		vrele(vp);
936 	} else if (error == ENOENT) {
937 		if (boothowto & (AB_VERBOSE|AB_DEBUG))
938 			printf("warning: no /dev/console\n");
939 	} else {
940 		printf("warning: lookup /dev/console: error %d\n", error);
941 	}
942 }
943 
944 /*
945  * List of paths to try when searching for "init".
946  */
947 static const char * const initpaths[] = {
948 	"/sbin/init",
949 	"/sbin/oinit",
950 	"/sbin/init.bak",
951 	"/rescue/init",
952 	NULL,
953 };
954 
955 /*
956  * Start the initial user process; try exec'ing each pathname in "initpaths".
957  * The program is invoked with one argument containing the boot flags.
958  */
959 static void
960 start_init(void *arg)
961 {
962 	struct lwp *l = arg;
963 	struct proc *p = l->l_proc;
964 	vaddr_t addr;
965 	struct sys_execve_args /* {
966 		syscallarg(const char *) path;
967 		syscallarg(char * const *) argp;
968 		syscallarg(char * const *) envp;
969 	} */ args;
970 	int options, i, error;
971 	register_t retval[2];
972 	char flags[4], *flagsp;
973 	const char *path, *slash;
974 	char *ucp, **uap, *arg0, *arg1, *argv[3];
975 	char ipath[129];
976 	int ipx, len;
977 
978 	/*
979 	 * Now in process 1.
980 	 */
981 	strncpy(p->p_comm, "init", MAXCOMLEN);
982 
983 	/*
984 	 * Wait for main() to tell us that it's safe to exec.
985 	 */
986 	mutex_enter(&proc_lock);
987 	while (start_init_exec == 0)
988 		cv_wait(&lbolt, &proc_lock);
989 	mutex_exit(&proc_lock);
990 
991 	/*
992 	 * This is not the right way to do this.  We really should
993 	 * hand-craft a descriptor onto /dev/console to hand to init,
994 	 * but that's a _lot_ more work, and the benefit from this easy
995 	 * hack makes up for the "good is the enemy of the best" effect.
996 	 */
997 	check_console(l);
998 
999 	/*
1000 	 * Need just enough stack to hold the faked-up "execve()" arguments.
1001 	 */
1002 	addr = (vaddr_t)STACK_ALLOC(USRSTACK, PAGE_SIZE);
1003 	if (uvm_map(&p->p_vmspace->vm_map, &addr, PAGE_SIZE,
1004 	    NULL, UVM_UNKNOWN_OFFSET, 0,
1005 	    UVM_MAPFLAG(UVM_PROT_RW, UVM_PROT_RW, UVM_INH_COPY,
1006 	    UVM_ADV_NORMAL,
1007 	    UVM_FLAG_FIXED|UVM_FLAG_OVERLAY|UVM_FLAG_COPYONW)) != 0)
1008 		panic("init: couldn't allocate argument space");
1009 	p->p_vmspace->vm_maxsaddr = (void *)STACK_MAX(addr, PAGE_SIZE);
1010 
1011 	ipx = 0;
1012 	while (1) {
1013 		if (boothowto & RB_ASKNAME) {
1014 			printf("init path");
1015 			if (initpaths[ipx])
1016 				printf(" (default %s)", initpaths[ipx]);
1017 			printf(": ");
1018 			len = cngetsn(ipath, sizeof(ipath)-1);
1019 			if (len == 4 && strcmp(ipath, "halt") == 0) {
1020 				kern_reboot(RB_HALT, NULL);
1021 			} else if (len == 6 && strcmp(ipath, "reboot") == 0) {
1022 				kern_reboot(0, NULL);
1023 #if defined(DDB)
1024 			} else if (len == 3 && strcmp(ipath, "ddb") == 0) {
1025 				console_debugger();
1026 				continue;
1027 #endif
1028 			} else if (len > 0 && ipath[0] == '/') {
1029 				ipath[len] = '\0';
1030 				path = ipath;
1031 			} else if (len == 0 && initpaths[ipx] != NULL) {
1032 				path = initpaths[ipx++];
1033 			} else {
1034 				printf("use absolute path, ");
1035 #if defined(DDB)
1036 				printf("\"ddb\", ");
1037 #endif
1038 				printf("\"halt\", or \"reboot\"\n");
1039 				continue;
1040 			}
1041 		} else {
1042 			if ((path = initpaths[ipx++]) == NULL) {
1043 				ipx = 0;
1044 				boothowto |= RB_ASKNAME;
1045 				continue;
1046 			}
1047 		}
1048 
1049 		ucp = (char *)USRSTACK;
1050 
1051 		/*
1052 		 * Construct the boot flag argument.
1053 		 */
1054 		flagsp = flags;
1055 		*flagsp++ = '-';
1056 		options = 0;
1057 
1058 		if (boothowto & RB_SINGLE) {
1059 			*flagsp++ = 's';
1060 			options = 1;
1061 		}
1062 #ifdef notyet
1063 		if (boothowto & RB_FASTBOOT) {
1064 			*flagsp++ = 'f';
1065 			options = 1;
1066 		}
1067 #endif
1068 
1069 		/*
1070 		 * Move out the flags (arg 1), if necessary.
1071 		 */
1072 		if (options != 0) {
1073 			*flagsp++ = '\0';
1074 			i = flagsp - flags;
1075 #ifdef DEBUG
1076 			aprint_normal("init: copying out flags `%s' %d\n", flags, i);
1077 #endif
1078 			arg1 = STACK_ALLOC(ucp, i);
1079 			ucp = STACK_MAX(arg1, i);
1080 			if ((error = copyout((void *)flags, arg1, i)) != 0)
1081 				goto copyerr;
1082 		} else
1083 			arg1 = NULL;
1084 
1085 		/*
1086 		 * Move out the file name (also arg 0).
1087 		 */
1088 		i = strlen(path) + 1;
1089 #ifdef DEBUG
1090 		aprint_normal("init: copying out path `%s' %d\n", path, i);
1091 #else
1092 		if (boothowto & RB_ASKNAME || path != initpaths[0])
1093 			printf("init: trying %s\n", path);
1094 #endif
1095 		arg0 = STACK_ALLOC(ucp, i);
1096 		ucp = STACK_MAX(arg0, i);
1097 		if ((error = copyout(path, arg0, i)) != 0)
1098 			goto copyerr;
1099 
1100 		/*
1101 		 * Move out the arg pointers.
1102 		 */
1103 		ucp = (void *)STACK_ALIGN(ucp, STACK_ALIGNBYTES);
1104 		uap = (char **)STACK_ALLOC(ucp, sizeof(argv));
1105 		SCARG(&args, path) = arg0;
1106 		SCARG(&args, argp) = uap;
1107 		SCARG(&args, envp) = NULL;
1108 		slash = strrchr(path, '/');
1109 
1110 		argv[0] = slash ? arg0 + (slash + 1 - path) : arg0;
1111 		argv[1] = arg1;
1112 		argv[2] = NULL;
1113 		if ((error = copyout(argv, uap, sizeof(argv))) != 0)
1114 			goto copyerr;
1115 
1116 		/*
1117 		 * Now try to exec the program.  If it can't for any reason
1118 		 * other than it doesn't exist, complain.
1119 		 */
1120 		error = sys_execve(l, &args, retval);
1121 		if (error == 0 || error == EJUSTRETURN) {
1122 			KERNEL_UNLOCK_LAST(l);
1123 			return;
1124 		}
1125 		printf("exec %s: error %d\n", path, error);
1126 	}
1127 	printf("init: not found\n");
1128 	panic("no init");
1129 copyerr:
1130 	panic("copyout %d", error);
1131 }
1132 
1133 /*
1134  * calculate cache size (in bytes) from physmem and vsize.
1135  */
1136 vaddr_t
1137 calc_cache_size(vsize_t vsize, int pct, int va_pct)
1138 {
1139 	paddr_t t;
1140 
1141 	/* XXX should consider competing cache if any */
1142 	/* XXX should consider submaps */
1143 	t = (uintmax_t)physmem * pct / 100 * PAGE_SIZE;
1144 	if (vsize != 0) {
1145 		vsize = (uintmax_t)vsize * va_pct / 100;
1146 		if (t > vsize) {
1147 			t = vsize;
1148 		}
1149 	}
1150 	return t;
1151 }
1152 
1153 /*
1154  * Print the system start up banner.
1155  *
1156  * - Print a limited banner if AB_SILENT.
1157  * - Always send normal banner to the log.
1158  */
1159 #define MEM_PBUFSIZE	sizeof("99999 MB")
1160 
1161 void
1162 banner(void)
1163 {
1164 	static char notice[] = " Notice: this software is "
1165 	    "protected by copyright";
1166 	char pbuf[81];
1167 	void (*pr)(const char *, ...) __printflike(1, 2);
1168 	int i;
1169 
1170 	if ((boothowto & AB_SILENT) != 0) {
1171 		snprintf(pbuf, sizeof(pbuf), "%s %s (%s)",
1172 		    ostype, osrelease, kernel_ident);
1173 		printf_nolog("%s", pbuf);
1174 		for (i = 80 - strlen(pbuf) - sizeof(notice); i > 0; i--)
1175 			printf(" ");
1176 		printf_nolog("%s\n", notice);
1177 		pr = aprint_normal;
1178 	} else {
1179 		pr = printf;
1180 	}
1181 
1182 	memset(pbuf, 0, sizeof(pbuf));
1183 	(*pr)("%s%s", copyright, version);
1184 	format_bytes(pbuf, MEM_PBUFSIZE, ctob((uint64_t)physmem));
1185 	(*pr)("total memory = %s\n", pbuf);
1186 	format_bytes(pbuf, MEM_PBUFSIZE, ctob((uint64_t)uvm_availmem(false)));
1187 	(*pr)("avail memory = %s\n", pbuf);
1188 }
1189