xref: /netbsd-src/sys/kern/init_main.c (revision 4fee23f98c45552038ad6b5bd05124a41302fb01)
1 /*	$NetBSD: init_main.c,v 1.432 2011/06/12 03:35:56 rmind Exp $	*/
2 
3 /*-
4  * Copyright (c) 2008, 2009 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 1.432 2011/06/12 03:35:56 rmind Exp $");
101 
102 #include "opt_ddb.h"
103 #include "opt_ipsec.h"
104 #include "opt_modular.h"
105 #include "opt_ntp.h"
106 #include "opt_pipe.h"
107 #include "opt_sa.h"
108 #include "opt_syscall_debug.h"
109 #include "opt_sysv.h"
110 #include "opt_fileassoc.h"
111 #include "opt_ktrace.h"
112 #include "opt_pax.h"
113 #include "opt_compat_netbsd.h"
114 #include "opt_wapbl.h"
115 #include "opt_ptrace.h"
116 
117 #include "drvctl.h"
118 #include "ksyms.h"
119 #include "rnd.h"
120 #include "sysmon_envsys.h"
121 #include "sysmon_power.h"
122 #include "sysmon_taskq.h"
123 #include "sysmon_wdog.h"
124 #include "veriexec.h"
125 
126 #include <sys/param.h>
127 #include <sys/acct.h>
128 #include <sys/filedesc.h>
129 #include <sys/file.h>
130 #include <sys/errno.h>
131 #include <sys/callout.h>
132 #include <sys/cpu.h>
133 #include <sys/spldebug.h>
134 #include <sys/kernel.h>
135 #include <sys/mount.h>
136 #include <sys/proc.h>
137 #include <sys/kthread.h>
138 #include <sys/resourcevar.h>
139 #include <sys/signalvar.h>
140 #include <sys/systm.h>
141 #include <sys/vnode.h>
142 #include <sys/fstrans.h>
143 #include <sys/tty.h>
144 #include <sys/conf.h>
145 #include <sys/disklabel.h>
146 #include <sys/buf.h>
147 #include <sys/device.h>
148 #include <sys/exec.h>
149 #include <sys/socketvar.h>
150 #include <sys/protosw.h>
151 #include <sys/percpu.h>
152 #include <sys/pset.h>
153 #include <sys/sysctl.h>
154 #include <sys/reboot.h>
155 #include <sys/event.h>
156 #include <sys/mbuf.h>
157 #include <sys/sched.h>
158 #include <sys/sleepq.h>
159 #include <sys/iostat.h>
160 #include <sys/vmem.h>
161 #include <sys/uuid.h>
162 #include <sys/extent.h>
163 #include <sys/disk.h>
164 #include <sys/msgbuf.h>
165 #include <sys/module.h>
166 #include <sys/event.h>
167 #include <sys/lockf.h>
168 #include <sys/once.h>
169 #include <sys/ksyms.h>
170 #include <sys/uidinfo.h>
171 #include <sys/kprintf.h>
172 #ifdef FAST_IPSEC
173 #include <netipsec/ipsec.h>
174 #endif
175 #ifdef SYSVSHM
176 #include <sys/shm.h>
177 #endif
178 #ifdef SYSVSEM
179 #include <sys/sem.h>
180 #endif
181 #ifdef SYSVMSG
182 #include <sys/msg.h>
183 #endif
184 #include <sys/domain.h>
185 #include <sys/namei.h>
186 #if NRND > 0
187 #include <sys/rnd.h>
188 #endif
189 #include <sys/pipe.h>
190 #if NVERIEXEC > 0
191 #include <sys/verified_exec.h>
192 #endif /* NVERIEXEC > 0 */
193 #ifdef KTRACE
194 #include <sys/ktrace.h>
195 #endif
196 #include <sys/kauth.h>
197 #ifdef KERN_SA
198 #include <sys/savar.h>
199 #endif
200 #include <net80211/ieee80211_netbsd.h>
201 #ifdef PTRACE
202 #include <sys/ptrace.h>
203 #endif /* PTRACE */
204 
205 #include <sys/syscall.h>
206 #include <sys/syscallargs.h>
207 
208 #if defined(PAX_MPROTECT) || defined(PAX_SEGVGUARD) || defined(PAX_ASLR)
209 #include <sys/pax.h>
210 #endif /* PAX_MPROTECT || PAX_SEGVGUARD || PAX_ASLR */
211 
212 #include <ufs/ufs/quota.h>
213 
214 #include <miscfs/genfs/genfs.h>
215 #include <miscfs/syncfs/syncfs.h>
216 #include <miscfs/specfs/specdev.h>
217 
218 #include <sys/cpu.h>
219 
220 #include <uvm/uvm.h>	/* extern struct uvm uvm */
221 
222 #if NSYSMON_TASKQ > 0
223 #include <dev/sysmon/sysmon_taskq.h>
224 #endif
225 
226 #include <dev/cons.h>
227 
228 #if NSYSMON_ENVSYS > 0 || NSYSMON_POWER > 0 || NSYSMON_WDOG > 0
229 #include <dev/sysmon/sysmonvar.h>
230 #endif
231 
232 #include <net/bpf.h>
233 #include <net/if.h>
234 #include <net/raw_cb.h>
235 
236 #include <prop/proplib.h>
237 
238 #ifdef COMPAT_50
239 #include <compat/sys/time.h>
240 struct timeval50 boottime50;
241 #endif
242 
243 #include <sys/userconf.h>
244 
245 extern struct lwp lwp0;
246 extern time_t rootfstime;
247 
248 #ifndef curlwp
249 struct	lwp *curlwp = &lwp0;
250 #endif
251 struct	proc *initproc;
252 
253 struct	vnode *rootvp, *swapdev_vp;
254 int	boothowto;
255 int	cold = 1;			/* still working on startup */
256 struct timespec boottime;	        /* time at system startup - will only follow settime deltas */
257 
258 int	start_init_exec;		/* semaphore for start_init() */
259 
260 static void check_console(struct lwp *l);
261 static void start_init(void *);
262 static void configure(void);
263 static void configure2(void);
264 static void configure3(void);
265 void main(void);
266 
267 /*
268  * System startup; initialize the world, create process 0, mount root
269  * filesystem, and fork to create init and pagedaemon.  Most of the
270  * hard work is done in the lower-level initialization routines including
271  * startup(), which does memory initialization and autoconfiguration.
272  */
273 void
274 main(void)
275 {
276 	struct timespec time;
277 	struct lwp *l;
278 	struct proc *p;
279 	int s, error;
280 #ifdef NVNODE_IMPLICIT
281 	int usevnodes;
282 #endif
283 	CPU_INFO_ITERATOR cii;
284 	struct cpu_info *ci;
285 
286 	l = &lwp0;
287 #ifndef LWP0_CPU_INFO
288 	l->l_cpu = curcpu();
289 #endif
290 	l->l_pflag |= LP_RUNNING;
291 
292 	/*
293 	 * Attempt to find console and initialize
294 	 * in case of early panic or other messages.
295 	 */
296 	consinit();
297 
298 	kernel_lock_init();
299 	once_init();
300 	mutex_init(&cpu_lock, MUTEX_DEFAULT, IPL_NONE);
301 	kernconfig_lock_init();
302 	kthread_sysinit();
303 
304 	/* Initialize the device switch tables. */
305 	devsw_init();
306 
307 	/* Initialize event counters. */
308 	evcnt_init();
309 
310 	uvm_init();
311 
312 	prop_kern_init();
313 
314 #if ((NKSYMS > 0) || (NDDB > 0) || (NMODULAR > 0))
315 	ksyms_init();
316 #endif
317 	kprintf_init();
318 
319 	percpu_init();
320 
321 	/* Initialize lock caches. */
322 	mutex_obj_init();
323 	rw_obj_init();
324 
325 	/* Initialize the extent manager. */
326 	extent_init();
327 
328 	/* Do machine-dependent initialization. */
329 	cpu_startup();
330 
331 	/* Initialize the sysctl subsystem. */
332 	sysctl_init();
333 
334 	/* Initialize callouts, part 1. */
335 	callout_startup();
336 
337 	/* Initialize the kernel authorization subsystem. */
338 	kauth_init();
339 
340 	spec_init();
341 
342 	/*
343 	 * Set BPF op vector.  Can't do this in bpf attach, since
344 	 * network drivers attach before bpf.
345 	 */
346 	bpf_setops();
347 
348 	/* Start module system. */
349 	module_init();
350 
351 	/*
352 	 * Initialize the kernel authorization subsystem and start the
353 	 * default security model, if any. We need to do this early
354 	 * enough so that subsystems relying on any of the aforementioned
355 	 * can work properly. Since the security model may dictate the
356 	 * credential inheritance policy, it is needed at least before
357 	 * any process is created, specifically proc0.
358 	 */
359 	module_init_class(MODULE_CLASS_SECMODEL);
360 
361 	/* Initialize the buffer cache */
362 	bufinit();
363 
364 	/* Initialize sockets. */
365 	soinit();
366 
367 	/*
368 	 * The following things must be done before autoconfiguration.
369 	 */
370 #if NRND > 0
371 	rnd_init();		/* initialize random number generator */
372 #endif
373 
374 	/* Initialize process and pgrp structures. */
375 #ifdef KERN_SA
376 	sa_init();
377 #endif
378 	procinit();
379 	lwpinit();
380 
381 	/* Initialize signal-related data structures. */
382 	signal_init();
383 
384 	/* Initialize resource management. */
385 	resource_init();
386 
387 	/* Create process 0. */
388 	proc0_init();
389 	lwp0_init();
390 
391 	/* Disable preemption during boot. */
392 	kpreempt_disable();
393 
394 	/* Initialize the UID hash table. */
395 	uid_init();
396 
397 	/* Charge root for one process. */
398 	(void)chgproccnt(0, 1);
399 
400 	/* Initialize timekeeping. */
401 	time_init();
402 
403 	/* Initialize the run queues, turnstiles and sleep queues. */
404 	sched_rqinit();
405 	turnstile_init();
406 	sleeptab_init(&sleeptab);
407 
408 	sched_init();
409 
410 	/* Initialize processor-sets */
411 	psets_init();
412 
413 	/* MI initialization of the boot cpu */
414 	error = mi_cpu_attach(curcpu());
415 	KASSERT(error == 0);
416 
417 	/* Initialize timekeeping, part 2. */
418 	time_init2();
419 
420 	/*
421 	 * Initialize mbuf's.  Do this now because we might attempt to
422 	 * allocate mbufs or mbuf clusters during autoconfiguration.
423 	 */
424 	mbinit();
425 
426 	/* Initialize I/O statistics. */
427 	iostat_init();
428 
429 	/* Initialize the log device. */
430 	loginit();
431 
432 	/* Second part of module system initialization. */
433 	module_start_unload_thread();
434 
435 	/* Initialize the file systems. */
436 #ifdef NVNODE_IMPLICIT
437 	/*
438 	 * If maximum number of vnodes in namei vnode cache is not explicitly
439 	 * defined in kernel config, adjust the number such as we use roughly
440 	 * 10% of memory for vnodes and associated data structures in the
441 	 * assumed worst case.  Do not provide fewer than NVNODE vnodes.
442 	 */
443 	usevnodes =
444 	    calc_cache_size(kernel_map, 10, VNODE_VA_MAXPCT) / VNODE_COST;
445 	if (usevnodes > desiredvnodes)
446 		desiredvnodes = usevnodes;
447 #endif
448 	vfsinit();
449 	lf_init();
450 
451 	/* Initialize fstrans. */
452 	fstrans_init();
453 
454 	/* Initialize the file descriptor system. */
455 	fd_sys_init();
456 
457 	/* Initialize cwd structures */
458 	cwd_sys_init();
459 
460 	/* Initialize kqueue. */
461 	kqueue_init();
462 
463 	/* Initialize the system monitor subsystems. */
464 #if NSYSMON_TASKQ > 0
465 	sysmon_task_queue_preinit();
466 #endif
467 
468 #if NSYSMON_ENVSYS > 0
469 	sysmon_envsys_init();
470 #endif
471 
472 #if NSYSMON_POWER > 0
473 	sysmon_power_init();
474 #endif
475 
476 #if NSYSMON_WDOG > 0
477 	sysmon_wdog_init();
478 #endif
479 
480 	inittimecounter();
481 	ntp_init();
482 
483 	/* Initialize tty subsystem. */
484 	tty_init();
485 	ttyldisc_init();
486 
487 	/* Initialize the buffer cache, part 2. */
488 	bufinit2();
489 
490 	/* Initialize the disk wedge subsystem. */
491 	dkwedge_init();
492 
493 	/* Initialize interfaces. */
494 	ifinit1();
495 
496 	spldebug_start();
497 
498 	/* Configure the system hardware.  This will enable interrupts. */
499 	configure();
500 
501 	ssp_init();
502 
503 	ubc_init();		/* must be after autoconfig */
504 
505 	mm_init();
506 
507 	configure2();
508 	/* Now timer is working.  Enable preemption. */
509 	kpreempt_enable();
510 
511 #ifdef SYSVSHM
512 	/* Initialize System V style shared memory. */
513 	shminit();
514 #endif
515 
516 	vmem_rehash_start();	/* must be before exec_init */
517 
518 	/* Initialize exec structures */
519 	exec_init(1);		/* seminit calls exithook_establish() */
520 
521 #ifdef SYSVSEM
522 	/* Initialize System V style semaphores. */
523 	seminit();
524 #endif
525 
526 #ifdef SYSVMSG
527 	/* Initialize System V style message queues. */
528 	msginit();
529 #endif
530 
531 #if NVERIEXEC > 0
532 	/*
533 	 * Initialise the Veriexec subsystem.
534 	 */
535 	veriexec_init();
536 #endif /* NVERIEXEC > 0 */
537 
538 #if defined(PAX_MPROTECT) || defined(PAX_SEGVGUARD) || defined(PAX_ASLR)
539 	pax_init();
540 #endif /* PAX_MPROTECT || PAX_SEGVGUARD || PAX_ASLR */
541 
542 #ifdef	FAST_IPSEC
543 	/* Attach network crypto subsystem */
544 	ipsec_attach();
545 #endif
546 
547 	/*
548 	 * Initialize protocols.  Block reception of incoming packets
549 	 * until everything is ready.
550 	 */
551 	s = splnet();
552 	ifinit();
553 	domaininit(true);
554 	if_attachdomain();
555 	splx(s);
556 
557 #ifdef GPROF
558 	/* Initialize kernel profiling. */
559 	kmstartup();
560 #endif
561 
562 	/* Initialize system accounting. */
563 	acct_init();
564 
565 #ifndef PIPE_SOCKETPAIR
566 	/* Initialize pipes. */
567 	pipe_init();
568 #endif
569 
570 #ifdef KTRACE
571 	/* Initialize ktrace. */
572 	ktrinit();
573 #endif
574 
575 #ifdef PTRACE
576 	/* Initialize ptrace. */
577 	ptrace_init();
578 #endif /* PTRACE */
579 
580 	/* Initialize the UUID system calls. */
581 	uuid_init();
582 
583 	machdep_init();
584 
585 	procinit_sysctl();
586 
587 	/*
588 	 * Create process 1 (init(8)).  We do this now, as Unix has
589 	 * historically had init be process 1, and changing this would
590 	 * probably upset a lot of people.
591 	 *
592 	 * Note that process 1 won't immediately exec init(8), but will
593 	 * wait for us to inform it that the root file system has been
594 	 * mounted.
595 	 */
596 	if (fork1(l, 0, SIGCHLD, NULL, 0, start_init, NULL, NULL, &initproc))
597 		panic("fork init");
598 
599 	/*
600 	 * Load any remaining builtin modules, and hand back temporary
601 	 * storage to the VM system.  Then require force when loading any
602 	 * remaining un-init'ed built-in modules to avoid later surprises.
603 	 */
604 	module_init_class(MODULE_CLASS_ANY);
605 	module_builtin_require_force();
606 
607 	/*
608 	 * Finalize configuration now that all real devices have been
609 	 * found.  This needs to be done before the root device is
610 	 * selected, since finalization may create the root device.
611 	 */
612 	config_finalize();
613 
614 	sysctl_finalize();
615 
616 	/*
617 	 * Now that autoconfiguration has completed, we can determine
618 	 * the root and dump devices.
619 	 */
620 	cpu_rootconf();
621 	cpu_dumpconf();
622 
623 	/* Mount the root file system. */
624 	do {
625 		domountroothook(root_device);
626 		if ((error = vfs_mountroot())) {
627 			printf("cannot mount root, error = %d\n", error);
628 			boothowto |= RB_ASKNAME;
629 			setroot(root_device,
630 			    (rootdev != NODEV) ? DISKPART(rootdev) : 0);
631 		}
632 	} while (error != 0);
633 	mountroothook_destroy();
634 
635 	configure3();
636 
637 	/*
638 	 * Initialise the time-of-day clock, passing the time recorded
639 	 * in the root filesystem (if any) for use by systems that
640 	 * don't have a non-volatile time-of-day device.
641 	 */
642 	inittodr(rootfstime);
643 
644 	/*
645 	 * Now can look at time, having had a chance to verify the time
646 	 * from the file system.  Reset l->l_rtime as it may have been
647 	 * munched in mi_switch() after the time got set.
648 	 */
649 	getnanotime(&time);
650 	boottime = time;
651 #ifdef COMPAT_50
652 	{
653 		struct timeval tv;
654 		TIMESPEC_TO_TIMEVAL(&tv, &time);
655 		timeval_to_timeval50(&tv, &boottime50);
656 	}
657 #endif
658 	mutex_enter(proc_lock);
659 	LIST_FOREACH(p, &allproc, p_list) {
660 		KASSERT((p->p_flag & PK_MARKER) == 0);
661 		mutex_enter(p->p_lock);
662 		TIMESPEC_TO_TIMEVAL(&p->p_stats->p_start, &time);
663 		LIST_FOREACH(l, &p->p_lwps, l_sibling) {
664 			lwp_lock(l);
665 			memset(&l->l_rtime, 0, sizeof(l->l_rtime));
666 			lwp_unlock(l);
667 		}
668 		mutex_exit(p->p_lock);
669 	}
670 	mutex_exit(proc_lock);
671 	binuptime(&curlwp->l_stime);
672 
673 	for (CPU_INFO_FOREACH(cii, ci)) {
674 		ci->ci_schedstate.spc_lastmod = time_second;
675 	}
676 
677 	/* Create the pageout daemon kernel thread. */
678 	uvm_swap_init();
679 	if (kthread_create(PRI_PGDAEMON, KTHREAD_MPSAFE, NULL, uvm_pageout,
680 	    NULL, NULL, "pgdaemon"))
681 		panic("fork pagedaemon");
682 
683 	/* Create the filesystem syncer kernel thread. */
684 	if (kthread_create(PRI_IOFLUSH, KTHREAD_MPSAFE, NULL, sched_sync,
685 	    NULL, NULL, "ioflush"))
686 		panic("fork syncer");
687 
688 	/* Create the aiodone daemon kernel thread. */
689 	if (workqueue_create(&uvm.aiodone_queue, "aiodoned",
690 	    uvm_aiodone_worker, NULL, PRI_VM, IPL_NONE, WQ_MPSAFE))
691 		panic("fork aiodoned");
692 
693 	/*
694 	 * Okay, now we can let init(8) exec!  It's off to userland!
695 	 */
696 	mutex_enter(proc_lock);
697 	start_init_exec = 1;
698 	cv_broadcast(&lbolt);
699 	mutex_exit(proc_lock);
700 
701 	/* The scheduler is an infinite loop. */
702 	uvm_scheduler();
703 	/* NOTREACHED */
704 }
705 
706 /*
707  * Configure the system's hardware.
708  */
709 static void
710 configure(void)
711 {
712 
713 	/* Initialize autoconf data structures. */
714 	config_init_mi();
715 	/*
716 	 * XXX
717 	 * callout_setfunc() requires mutex(9) so it can't be in config_init()
718 	 * on amiga and atari which use config_init() and autoconf(9) fucntions
719 	 * to initialize console devices.
720 	 */
721 	config_twiddle_init();
722 
723 	pmf_init();
724 #if NDRVCTL > 0
725 	drvctl_init();
726 #endif
727 
728 	userconf_init();
729 	if (boothowto & RB_USERCONF)
730 		userconf_prompt();
731 
732 	if ((boothowto & (AB_SILENT|AB_VERBOSE)) == AB_SILENT) {
733 		printf_nolog("Detecting hardware...");
734 	}
735 
736 	/*
737 	 * Do the machine-dependent portion of autoconfiguration.  This
738 	 * sets the configuration machinery here in motion by "finding"
739 	 * the root bus.  When this function returns, we expect interrupts
740 	 * to be enabled.
741 	 */
742 	cpu_configure();
743 }
744 
745 static void
746 configure2(void)
747 {
748 	CPU_INFO_ITERATOR cii;
749 	struct cpu_info *ci;
750 	int s;
751 
752 	/*
753 	 * Now that we've found all the hardware, start the real time
754 	 * and statistics clocks.
755 	 */
756 	initclocks();
757 
758 	cold = 0;	/* clocks are running, we're warm now! */
759 	s = splsched();
760 	curcpu()->ci_schedstate.spc_flags |= SPCF_RUNNING;
761 	splx(s);
762 
763 	/* Boot the secondary processors. */
764 	for (CPU_INFO_FOREACH(cii, ci)) {
765 		uvm_cpu_attach(ci);
766 	}
767 	mp_online = true;
768 #if defined(MULTIPROCESSOR)
769 	cpu_boot_secondary_processors();
770 #endif
771 
772 	/* Setup the runqueues and scheduler. */
773 	runq_init();
774 	synch_init();
775 
776 	/*
777 	 * Bus scans can make it appear as if the system has paused, so
778 	 * twiddle constantly while config_interrupts() jobs are running.
779 	 */
780 	config_twiddle_fn(NULL);
781 
782 	/*
783 	 * Create threads to call back and finish configuration for
784 	 * devices that want interrupts enabled.
785 	 */
786 	config_create_interruptthreads();
787 
788 	/* Get the threads going and into any sleeps before continuing. */
789 	yield();
790 }
791 
792 static void
793 configure3(void)
794 {
795 
796 	/*
797 	 * Create threads to call back and finish configuration for
798 	 * devices that want the mounted root file system.
799 	 */
800 	config_create_mountrootthreads();
801 
802 	/* Get the threads going and into any sleeps before continuing. */
803 	yield();
804 }
805 
806 static void
807 check_console(struct lwp *l)
808 {
809 	struct vnode *vp;
810 	int error;
811 
812 	error = namei_simple_kernel("/dev/console",
813 				NSM_FOLLOW_NOEMULROOT, &vp);
814 	if (error == 0)
815 		vrele(vp);
816 	else if (error == ENOENT)
817 		printf("warning: no /dev/console\n");
818 	else
819 		printf("warning: lookup /dev/console: error %d\n", error);
820 }
821 
822 /*
823  * List of paths to try when searching for "init".
824  */
825 static const char * const initpaths[] = {
826 	"/sbin/init",
827 	"/sbin/oinit",
828 	"/sbin/init.bak",
829 	NULL,
830 };
831 
832 /*
833  * Start the initial user process; try exec'ing each pathname in "initpaths".
834  * The program is invoked with one argument containing the boot flags.
835  */
836 static void
837 start_init(void *arg)
838 {
839 	struct lwp *l = arg;
840 	struct proc *p = l->l_proc;
841 	vaddr_t addr;
842 	struct sys_execve_args /* {
843 		syscallarg(const char *) path;
844 		syscallarg(char * const *) argp;
845 		syscallarg(char * const *) envp;
846 	} */ args;
847 	int options, i, error;
848 	register_t retval[2];
849 	char flags[4], *flagsp;
850 	const char *path, *slash;
851 	char *ucp, **uap, *arg0, *arg1 = NULL;
852 	char ipath[129];
853 	int ipx, len;
854 
855 	/*
856 	 * Now in process 1.
857 	 */
858 	strncpy(p->p_comm, "init", MAXCOMLEN);
859 
860 	/*
861 	 * Wait for main() to tell us that it's safe to exec.
862 	 */
863 	mutex_enter(proc_lock);
864 	while (start_init_exec == 0)
865 		cv_wait(&lbolt, proc_lock);
866 	mutex_exit(proc_lock);
867 
868 	/*
869 	 * This is not the right way to do this.  We really should
870 	 * hand-craft a descriptor onto /dev/console to hand to init,
871 	 * but that's a _lot_ more work, and the benefit from this easy
872 	 * hack makes up for the "good is the enemy of the best" effect.
873 	 */
874 	check_console(l);
875 
876 	/*
877 	 * Need just enough stack to hold the faked-up "execve()" arguments.
878 	 */
879 	addr = (vaddr_t)STACK_ALLOC(USRSTACK, PAGE_SIZE);
880 	if (uvm_map(&p->p_vmspace->vm_map, &addr, PAGE_SIZE,
881                     NULL, UVM_UNKNOWN_OFFSET, 0,
882                     UVM_MAPFLAG(UVM_PROT_ALL, UVM_PROT_ALL, UVM_INH_COPY,
883 		    UVM_ADV_NORMAL,
884                     UVM_FLAG_FIXED|UVM_FLAG_OVERLAY|UVM_FLAG_COPYONW)) != 0)
885 		panic("init: couldn't allocate argument space");
886 	p->p_vmspace->vm_maxsaddr = (void *)STACK_MAX(addr, PAGE_SIZE);
887 
888 	ipx = 0;
889 	while (1) {
890 		if (boothowto & RB_ASKNAME) {
891 			printf("init path");
892 			if (initpaths[ipx])
893 				printf(" (default %s)", initpaths[ipx]);
894 			printf(": ");
895 			len = cngetsn(ipath, sizeof(ipath)-1);
896 			if (len == 4 && strcmp(ipath, "halt") == 0) {
897 				cpu_reboot(RB_HALT, NULL);
898 			} else if (len == 6 && strcmp(ipath, "reboot") == 0) {
899 				cpu_reboot(0, NULL);
900 #if defined(DDB)
901 			} else if (len == 3 && strcmp(ipath, "ddb") == 0) {
902 				console_debugger();
903 				continue;
904 #endif
905 			} else if (len > 0 && ipath[0] == '/') {
906 				ipath[len] = '\0';
907 				path = ipath;
908 			} else if (len == 0 && initpaths[ipx] != NULL) {
909 				path = initpaths[ipx++];
910 			} else {
911 				printf("use absolute path, ");
912 #if defined(DDB)
913 				printf("\"ddb\", ");
914 #endif
915 				printf("\"halt\", or \"reboot\"\n");
916 				continue;
917 			}
918 		} else {
919 			if ((path = initpaths[ipx++]) == NULL) {
920 				ipx = 0;
921 				boothowto |= RB_ASKNAME;
922 				continue;
923 			}
924 		}
925 
926 		ucp = (char *)USRSTACK;
927 
928 		/*
929 		 * Construct the boot flag argument.
930 		 */
931 		flagsp = flags;
932 		*flagsp++ = '-';
933 		options = 0;
934 
935 		if (boothowto & RB_SINGLE) {
936 			*flagsp++ = 's';
937 			options = 1;
938 		}
939 #ifdef notyet
940 		if (boothowto & RB_FASTBOOT) {
941 			*flagsp++ = 'f';
942 			options = 1;
943 		}
944 #endif
945 
946 		/*
947 		 * Move out the flags (arg 1), if necessary.
948 		 */
949 		if (options != 0) {
950 			*flagsp++ = '\0';
951 			i = flagsp - flags;
952 #ifdef DEBUG
953 			aprint_normal("init: copying out flags `%s' %d\n", flags, i);
954 #endif
955 			arg1 = STACK_ALLOC(ucp, i);
956 			ucp = STACK_MAX(arg1, i);
957 			(void)copyout((void *)flags, arg1, i);
958 		}
959 
960 		/*
961 		 * Move out the file name (also arg 0).
962 		 */
963 		i = strlen(path) + 1;
964 #ifdef DEBUG
965 		aprint_normal("init: copying out path `%s' %d\n", path, i);
966 #else
967 		if (boothowto & RB_ASKNAME || path != initpaths[0])
968 			printf("init: trying %s\n", path);
969 #endif
970 		arg0 = STACK_ALLOC(ucp, i);
971 		ucp = STACK_MAX(arg0, i);
972 		(void)copyout(path, arg0, i);
973 
974 		/*
975 		 * Move out the arg pointers.
976 		 */
977 		ucp = (void *)STACK_ALIGN(ucp, ALIGNBYTES);
978 		uap = (char **)STACK_ALLOC(ucp, sizeof(char *) * 3);
979 		SCARG(&args, path) = arg0;
980 		SCARG(&args, argp) = uap;
981 		SCARG(&args, envp) = NULL;
982 		slash = strrchr(path, '/');
983 		if (slash)
984 			(void)suword((void *)uap++,
985 			    (long)arg0 + (slash + 1 - path));
986 		else
987 			(void)suword((void *)uap++, (long)arg0);
988 		if (options != 0)
989 			(void)suword((void *)uap++, (long)arg1);
990 		(void)suword((void *)uap++, 0);	/* terminator */
991 
992 		/*
993 		 * Now try to exec the program.  If can't for any reason
994 		 * other than it doesn't exist, complain.
995 		 */
996 		error = sys_execve(l, &args, retval);
997 		if (error == 0 || error == EJUSTRETURN) {
998 			KERNEL_UNLOCK_LAST(l);
999 			return;
1000 		}
1001 		printf("exec %s: error %d\n", path, error);
1002 	}
1003 	printf("init: not found\n");
1004 	panic("no init");
1005 }
1006 
1007 /*
1008  * calculate cache size (in bytes) from physmem and vm_map size.
1009  */
1010 vaddr_t
1011 calc_cache_size(struct vm_map *map, int pct, int va_pct)
1012 {
1013 	paddr_t t;
1014 
1015 	/* XXX should consider competing cache if any */
1016 	/* XXX should consider submaps */
1017 	t = (uintmax_t)physmem * pct / 100 * PAGE_SIZE;
1018 	if (map != NULL) {
1019 		vsize_t vsize;
1020 
1021 		vsize = vm_map_max(map) - vm_map_min(map);
1022 		vsize = (uintmax_t)vsize * va_pct / 100;
1023 		if (t > vsize) {
1024 			t = vsize;
1025 		}
1026 	}
1027 	return t;
1028 }
1029 
1030 /*
1031  * Print the system start up banner.
1032  *
1033  * - Print a limited banner if AB_SILENT.
1034  * - Always send normal banner to the log.
1035  */
1036 #define MEM_PBUFSIZE	sizeof("99999 MB")
1037 
1038 void
1039 banner(void)
1040 {
1041 	static char notice[] = " Notice: this software is "
1042 	    "protected by copyright";
1043 	char pbuf[81];
1044 	void (*pr)(const char *, ...);
1045 	int i;
1046 
1047 	if ((boothowto & AB_SILENT) != 0) {
1048 		snprintf(pbuf, sizeof(pbuf), "%s %s (%s)",
1049 		    ostype, osrelease, kernel_ident);
1050 		printf_nolog("%s", pbuf);
1051 		for (i = 80 - strlen(pbuf) - sizeof(notice); i > 0; i--)
1052 			printf(" ");
1053 		printf_nolog("%s\n", notice);
1054 		pr = aprint_normal;
1055 	} else {
1056 		pr = printf;
1057 	}
1058 
1059 	memset(pbuf, 0, sizeof(pbuf));
1060 	(*pr)("%s%s", copyright, version);
1061 	format_bytes(pbuf, MEM_PBUFSIZE, ctob((uint64_t)physmem));
1062 	(*pr)("total memory = %s\n", pbuf);
1063 	format_bytes(pbuf, MEM_PBUFSIZE, ctob((uint64_t)uvmexp.free));
1064 	(*pr)("avail memory = %s\n", pbuf);
1065 }
1066