xref: /netbsd-src/sys/kern/init_main.c (revision c505c4429840c353a86d4eb53b5e2bfc0092264e)
1 /*	$NetBSD: init_main.c,v 1.422 2010/06/26 07:23:57 pgoyette 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 1.422 2010/06/26 07:23:57 pgoyette 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>
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 #ifdef _KERNEL_OPT
244 #include "opt_userconf.h"
245 #endif
246 #ifdef USERCONF
247 #include <sys/userconf.h>
248 #endif
249 
250 extern struct lwp lwp0;
251 extern time_t rootfstime;
252 
253 #ifndef curlwp
254 struct	lwp *curlwp = &lwp0;
255 #endif
256 struct	proc *initproc;
257 
258 struct	vnode *rootvp, *swapdev_vp;
259 int	boothowto;
260 int	cold = 1;			/* still working on startup */
261 struct timespec boottime;	        /* time at system startup - will only follow settime deltas */
262 
263 int	start_init_exec;		/* semaphore for start_init() */
264 
265 static void check_console(struct lwp *l);
266 static void start_init(void *);
267 static void configure(void);
268 static void configure2(void);
269 static void configure3(void);
270 void main(void);
271 
272 /*
273  * System startup; initialize the world, create process 0, mount root
274  * filesystem, and fork to create init and pagedaemon.  Most of the
275  * hard work is done in the lower-level initialization routines including
276  * startup(), which does memory initialization and autoconfiguration.
277  */
278 void
279 main(void)
280 {
281 	struct timespec time;
282 	struct lwp *l;
283 	struct proc *p;
284 	int s, error;
285 #ifdef NVNODE_IMPLICIT
286 	int usevnodes;
287 #endif
288 	CPU_INFO_ITERATOR cii;
289 	struct cpu_info *ci;
290 
291 	l = &lwp0;
292 #ifndef LWP0_CPU_INFO
293 	l->l_cpu = curcpu();
294 #endif
295 	l->l_pflag |= LP_RUNNING;
296 
297 	/*
298 	 * Attempt to find console and initialize
299 	 * in case of early panic or other messages.
300 	 */
301 	consinit();
302 
303 	kernel_lock_init();
304 	once_init();
305 	mutex_init(&cpu_lock, MUTEX_DEFAULT, IPL_NONE);
306 
307 	/* Initialize the device switch tables. */
308 	devsw_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 	evcnt_init();		/* initialize event counters */
371 #if NRND > 0
372 	rnd_init();		/* initialize random number generator */
373 #endif
374 
375 	/* Initialize process and pgrp structures. */
376 #ifdef KERN_SA
377 	sa_init();
378 #endif
379 	procinit();
380 	lwpinit();
381 
382 	/* Initialize signal-related data structures. */
383 	signal_init();
384 
385 	/* Initialize resource management. */
386 	resource_init();
387 
388 	/* Create process 0. */
389 	proc0_init();
390 	lwp0_init();
391 
392 	/* Disable preemption during boot. */
393 	kpreempt_disable();
394 
395 	/* Initialize the UID hash table. */
396 	uid_init();
397 
398 	/* Charge root for one process. */
399 	(void)chgproccnt(0, 1);
400 
401 	/* Initialize timekeeping. */
402 	time_init();
403 
404 	/* Initialize the run queues, turnstiles and sleep queues. */
405 	sched_rqinit();
406 	turnstile_init();
407 	sleeptab_init(&sleeptab);
408 
409 	sched_init();
410 
411 	/* Initialize processor-sets */
412 	psets_init();
413 
414 	/* MI initialization of the boot cpu */
415 	error = mi_cpu_attach(curcpu());
416 	KASSERT(error == 0);
417 
418 	/* Initialize timekeeping, part 2. */
419 	time_init2();
420 
421 	/*
422 	 * Initialize mbuf's.  Do this now because we might attempt to
423 	 * allocate mbufs or mbuf clusters during autoconfiguration.
424 	 */
425 	mbinit();
426 
427 	/* Initialize I/O statistics. */
428 	iostat_init();
429 
430 	/* Initialize the log device. */
431 	loginit();
432 
433 	/* Second part of module system initialization. */
434 	module_start_unload_thread();
435 
436 	/* Initialize the file systems. */
437 #ifdef NVNODE_IMPLICIT
438 	/*
439 	 * If maximum number of vnodes in namei vnode cache is not explicitly
440 	 * defined in kernel config, adjust the number such as we use roughly
441 	 * 10% of memory for vnodes and associated data structures in the
442 	 * assumed worst case.  Do not provide fewer than NVNODE vnodes.
443 	 */
444 	usevnodes =
445 	    calc_cache_size(kernel_map, 10, VNODE_VA_MAXPCT) / VNODE_COST;
446 	if (usevnodes > desiredvnodes)
447 		desiredvnodes = usevnodes;
448 #endif
449 	vfsinit();
450 	lf_init();
451 
452 	/* Initialize fstrans. */
453 	fstrans_init();
454 
455 	/* Initialize the file descriptor system. */
456 	fd_sys_init();
457 
458 	/* Initialize cwd structures */
459 	cwd_sys_init();
460 
461 	/* Initialize kqueue. */
462 	kqueue_init();
463 
464 	/* Initialize the system monitor subsystems. */
465 #if NSYSMON_TASKQ > 0
466 	sysmon_task_queue_preinit();
467 #endif
468 
469 #if NSYSMON_ENVSYS > 0
470 	sysmon_envsys_init();
471 #endif
472 
473 #if NSYSMON_POWER > 0
474 	sysmon_power_init();
475 #endif
476 
477 #if NSYSMON_WDOG > 0
478 	sysmon_wdog_init();
479 #endif
480 
481 	inittimecounter();
482 	ntp_init();
483 
484 	/* Initialize tty subsystem. */
485 	tty_init();
486 	ttyldisc_init();
487 
488 	/* Initialize the buffer cache, part 2. */
489 	bufinit2();
490 
491 	/* Initialize the disk wedge subsystem. */
492 	dkwedge_init();
493 
494 	/* Initialize interfaces. */
495 	ifinit1();
496 
497 	spldebug_start();
498 
499 	/* Configure the system hardware.  This will enable interrupts. */
500 	configure();
501 
502 	ssp_init();
503 
504 	configure2();
505 	/* Now timer is working.  Enable preemption. */
506 	kpreempt_enable();
507 
508 	ubc_init();		/* must be after autoconfig */
509 
510 #ifdef SYSVSHM
511 	/* Initialize System V style shared memory. */
512 	shminit();
513 #endif
514 
515 	vmem_rehash_start();	/* must be before exec_init */
516 
517 	/* Initialize exec structures */
518 	exec_init(1);		/* seminit calls exithook_establish() */
519 
520 #ifdef SYSVSEM
521 	/* Initialize System V style semaphores. */
522 	seminit();
523 #endif
524 
525 #ifdef SYSVMSG
526 	/* Initialize System V style message queues. */
527 	msginit();
528 #endif
529 
530 #if NVERIEXEC > 0
531 	/*
532 	 * Initialise the Veriexec subsystem.
533 	 */
534 	veriexec_init();
535 #endif /* NVERIEXEC > 0 */
536 
537 #if defined(PAX_MPROTECT) || defined(PAX_SEGVGUARD) || defined(PAX_ASLR)
538 	pax_init();
539 #endif /* PAX_MPROTECT || PAX_SEGVGUARD || PAX_ASLR */
540 
541 #ifdef	FAST_IPSEC
542 	/* Attach network crypto subsystem */
543 	ipsec_attach();
544 #endif
545 
546 	/*
547 	 * Initialize protocols.  Block reception of incoming packets
548 	 * until everything is ready.
549 	 */
550 	s = splnet();
551 	ifinit();
552 	domaininit(true);
553 	if_attachdomain();
554 	splx(s);
555 
556 #ifdef GPROF
557 	/* Initialize kernel profiling. */
558 	kmstartup();
559 #endif
560 
561 	/* Initialize system accounting. */
562 	acct_init();
563 
564 #ifndef PIPE_SOCKETPAIR
565 	/* Initialize pipes. */
566 	pipe_init();
567 #endif
568 
569 #ifdef KTRACE
570 	/* Initialize ktrace. */
571 	ktrinit();
572 #endif
573 
574 #ifdef PTRACE
575 	/* Initialize ptrace. */
576 	ptrace_init();
577 #endif /* PTRACE */
578 
579 	/* Initialize the UUID system calls. */
580 	uuid_init();
581 
582 	machdep_init();
583 
584 	/*
585 	 * Create process 1 (init(8)).  We do this now, as Unix has
586 	 * historically had init be process 1, and changing this would
587 	 * probably upset a lot of people.
588 	 *
589 	 * Note that process 1 won't immediately exec init(8), but will
590 	 * wait for us to inform it that the root file system has been
591 	 * mounted.
592 	 */
593 	if (fork1(l, 0, SIGCHLD, NULL, 0, start_init, NULL, NULL, &initproc))
594 		panic("fork init");
595 
596 	/*
597 	 * Load any remaining builtin modules, and hand back temporary
598 	 * storage to the VM system.  Then require force when loading any
599 	 * remaining un-init'ed built-in modules to avoid later surprises.
600 	 */
601 	module_init_class(MODULE_CLASS_ANY);
602 	module_builtin_require_force();
603 
604 	/*
605 	 * Finalize configuration now that all real devices have been
606 	 * found.  This needs to be done before the root device is
607 	 * selected, since finalization may create the root device.
608 	 */
609 	config_finalize();
610 
611 	sysctl_finalize();
612 
613 	/*
614 	 * Now that autoconfiguration has completed, we can determine
615 	 * the root and dump devices.
616 	 */
617 	cpu_rootconf();
618 	cpu_dumpconf();
619 
620 	/* Mount the root file system. */
621 	do {
622 		domountroothook(root_device);
623 		if ((error = vfs_mountroot())) {
624 			printf("cannot mount root, error = %d\n", error);
625 			boothowto |= RB_ASKNAME;
626 			setroot(root_device,
627 			    (rootdev != NODEV) ? DISKPART(rootdev) : 0);
628 		}
629 	} while (error != 0);
630 	mountroothook_destroy();
631 
632 	configure3();
633 
634 	/*
635 	 * Initialise the time-of-day clock, passing the time recorded
636 	 * in the root filesystem (if any) for use by systems that
637 	 * don't have a non-volatile time-of-day device.
638 	 */
639 	inittodr(rootfstime);
640 
641 	/*
642 	 * Now can look at time, having had a chance to verify the time
643 	 * from the file system.  Reset l->l_rtime as it may have been
644 	 * munched in mi_switch() after the time got set.
645 	 */
646 	getnanotime(&time);
647 	boottime = time;
648 #ifdef COMPAT_50
649 	{
650 		struct timeval tv;
651 		TIMESPEC_TO_TIMEVAL(&tv, &time);
652 		timeval_to_timeval50(&tv, &boottime50);
653 	}
654 #endif
655 	mutex_enter(proc_lock);
656 	LIST_FOREACH(p, &allproc, p_list) {
657 		KASSERT((p->p_flag & PK_MARKER) == 0);
658 		mutex_enter(p->p_lock);
659 		TIMESPEC_TO_TIMEVAL(&p->p_stats->p_start, &time);
660 		LIST_FOREACH(l, &p->p_lwps, l_sibling) {
661 			lwp_lock(l);
662 			memset(&l->l_rtime, 0, sizeof(l->l_rtime));
663 			lwp_unlock(l);
664 		}
665 		mutex_exit(p->p_lock);
666 	}
667 	mutex_exit(proc_lock);
668 	binuptime(&curlwp->l_stime);
669 
670 	for (CPU_INFO_FOREACH(cii, ci)) {
671 		ci->ci_schedstate.spc_lastmod = time_second;
672 	}
673 
674 	/* Create the pageout daemon kernel thread. */
675 	uvm_swap_init();
676 	if (kthread_create(PRI_PGDAEMON, KTHREAD_MPSAFE, NULL, uvm_pageout,
677 	    NULL, NULL, "pgdaemon"))
678 		panic("fork pagedaemon");
679 
680 	/* Create the filesystem syncer kernel thread. */
681 	if (kthread_create(PRI_IOFLUSH, KTHREAD_MPSAFE, NULL, sched_sync,
682 	    NULL, NULL, "ioflush"))
683 		panic("fork syncer");
684 
685 	/* Create the aiodone daemon kernel thread. */
686 	if (workqueue_create(&uvm.aiodone_queue, "aiodoned",
687 	    uvm_aiodone_worker, NULL, PRI_VM, IPL_NONE, WQ_MPSAFE))
688 		panic("fork aiodoned");
689 
690 	/*
691 	 * Okay, now we can let init(8) exec!  It's off to userland!
692 	 */
693 	mutex_enter(proc_lock);
694 	start_init_exec = 1;
695 	cv_broadcast(&lbolt);
696 	mutex_exit(proc_lock);
697 
698 	/* The scheduler is an infinite loop. */
699 	uvm_scheduler();
700 	/* NOTREACHED */
701 }
702 
703 /*
704  * Configure the system's hardware.
705  */
706 static void
707 configure(void)
708 {
709 
710 	/* Initialize autoconf data structures. */
711 	config_init_mi();
712 	/*
713 	 * XXX
714 	 * callout_setfunc() requires mutex(9) so it can't be in config_init()
715 	 * on amiga and atari which use config_init() and autoconf(9) fucntions
716 	 * to initialize console devices.
717 	 */
718 	config_twiddle_init();
719 
720 	pmf_init();
721 #if NDRVCTL > 0
722 	drvctl_init();
723 #endif
724 
725 #ifdef USERCONF
726 	if (boothowto & RB_USERCONF)
727 		user_config();
728 #endif
729 
730 	if ((boothowto & (AB_SILENT|AB_VERBOSE)) == AB_SILENT) {
731 		printf_nolog("Detecting hardware...");
732 	}
733 
734 	/*
735 	 * Do the machine-dependent portion of autoconfiguration.  This
736 	 * sets the configuration machinery here in motion by "finding"
737 	 * the root bus.  When this function returns, we expect interrupts
738 	 * to be enabled.
739 	 */
740 	cpu_configure();
741 }
742 
743 static void
744 configure2(void)
745 {
746 	CPU_INFO_ITERATOR cii;
747 	struct cpu_info *ci;
748 	int s;
749 
750 	/*
751 	 * Now that we've found all the hardware, start the real time
752 	 * and statistics clocks.
753 	 */
754 	initclocks();
755 
756 	cold = 0;	/* clocks are running, we're warm now! */
757 	s = splsched();
758 	curcpu()->ci_schedstate.spc_flags |= SPCF_RUNNING;
759 	splx(s);
760 
761 	/* Boot the secondary processors. */
762 	for (CPU_INFO_FOREACH(cii, ci)) {
763 		uvm_cpu_attach(ci);
764 	}
765 	mp_online = true;
766 #if defined(MULTIPROCESSOR)
767 	cpu_boot_secondary_processors();
768 #endif
769 
770 	/* Setup the runqueues and scheduler. */
771 	runq_init();
772 	synch_init();
773 
774 	/*
775 	 * Bus scans can make it appear as if the system has paused, so
776 	 * twiddle constantly while config_interrupts() jobs are running.
777 	 */
778 	config_twiddle_fn(NULL);
779 
780 	/*
781 	 * Create threads to call back and finish configuration for
782 	 * devices that want interrupts enabled.
783 	 */
784 	config_create_interruptthreads();
785 
786 	/* Get the threads going and into any sleeps before continuing. */
787 	yield();
788 }
789 
790 static void
791 configure3(void)
792 {
793 
794 	/*
795 	 * Create threads to call back and finish configuration for
796 	 * devices that want the mounted root file system.
797 	 */
798 	config_create_mountrootthreads();
799 
800 	/* Get the threads going and into any sleeps before continuing. */
801 	yield();
802 }
803 
804 static void
805 check_console(struct lwp *l)
806 {
807 	struct vnode *vp;
808 	int error;
809 
810 	error = namei_simple_kernel("/dev/console",
811 				NSM_FOLLOW_NOEMULROOT, &vp);
812 	if (error == 0)
813 		vrele(vp);
814 	else if (error == ENOENT)
815 		printf("warning: no /dev/console\n");
816 	else
817 		printf("warning: lookup /dev/console: error %d\n", error);
818 }
819 
820 /*
821  * List of paths to try when searching for "init".
822  */
823 static const char * const initpaths[] = {
824 	"/sbin/init",
825 	"/sbin/oinit",
826 	"/sbin/init.bak",
827 	NULL,
828 };
829 
830 /*
831  * Start the initial user process; try exec'ing each pathname in "initpaths".
832  * The program is invoked with one argument containing the boot flags.
833  */
834 static void
835 start_init(void *arg)
836 {
837 	struct lwp *l = arg;
838 	struct proc *p = l->l_proc;
839 	vaddr_t addr;
840 	struct sys_execve_args /* {
841 		syscallarg(const char *) path;
842 		syscallarg(char * const *) argp;
843 		syscallarg(char * const *) envp;
844 	} */ args;
845 	int options, i, error;
846 	register_t retval[2];
847 	char flags[4], *flagsp;
848 	const char *path, *slash;
849 	char *ucp, **uap, *arg0, *arg1 = NULL;
850 	char ipath[129];
851 	int ipx, len;
852 
853 	/*
854 	 * Now in process 1.
855 	 */
856 	strncpy(p->p_comm, "init", MAXCOMLEN);
857 
858 	/*
859 	 * Wait for main() to tell us that it's safe to exec.
860 	 */
861 	mutex_enter(proc_lock);
862 	while (start_init_exec == 0)
863 		cv_wait(&lbolt, proc_lock);
864 	mutex_exit(proc_lock);
865 
866 	/*
867 	 * This is not the right way to do this.  We really should
868 	 * hand-craft a descriptor onto /dev/console to hand to init,
869 	 * but that's a _lot_ more work, and the benefit from this easy
870 	 * hack makes up for the "good is the enemy of the best" effect.
871 	 */
872 	check_console(l);
873 
874 	/*
875 	 * Need just enough stack to hold the faked-up "execve()" arguments.
876 	 */
877 	addr = (vaddr_t)STACK_ALLOC(USRSTACK, PAGE_SIZE);
878 	if (uvm_map(&p->p_vmspace->vm_map, &addr, PAGE_SIZE,
879                     NULL, UVM_UNKNOWN_OFFSET, 0,
880                     UVM_MAPFLAG(UVM_PROT_ALL, UVM_PROT_ALL, UVM_INH_COPY,
881 		    UVM_ADV_NORMAL,
882                     UVM_FLAG_FIXED|UVM_FLAG_OVERLAY|UVM_FLAG_COPYONW)) != 0)
883 		panic("init: couldn't allocate argument space");
884 	p->p_vmspace->vm_maxsaddr = (void *)STACK_MAX(addr, PAGE_SIZE);
885 
886 	ipx = 0;
887 	while (1) {
888 		if (boothowto & RB_ASKNAME) {
889 			printf("init path");
890 			if (initpaths[ipx])
891 				printf(" (default %s)", initpaths[ipx]);
892 			printf(": ");
893 			len = cngetsn(ipath, sizeof(ipath)-1);
894 			if (len == 4 && strcmp(ipath, "halt") == 0) {
895 				cpu_reboot(RB_HALT, NULL);
896 			} else if (len == 6 && strcmp(ipath, "reboot") == 0) {
897 				cpu_reboot(0, NULL);
898 #if defined(DDB)
899 			} else if (len == 3 && strcmp(ipath, "ddb") == 0) {
900 				console_debugger();
901 				continue;
902 #endif
903 			} else if (len > 0 && ipath[0] == '/') {
904 				ipath[len] = '\0';
905 				path = ipath;
906 			} else if (len == 0 && initpaths[ipx] != NULL) {
907 				path = initpaths[ipx++];
908 			} else {
909 				printf("use absolute path, ");
910 #if defined(DDB)
911 				printf("\"ddb\", ");
912 #endif
913 				printf("\"halt\", or \"reboot\"\n");
914 				continue;
915 			}
916 		} else {
917 			if ((path = initpaths[ipx++]) == NULL) {
918 				ipx = 0;
919 				boothowto |= RB_ASKNAME;
920 				continue;
921 			}
922 		}
923 
924 		ucp = (char *)USRSTACK;
925 
926 		/*
927 		 * Construct the boot flag argument.
928 		 */
929 		flagsp = flags;
930 		*flagsp++ = '-';
931 		options = 0;
932 
933 		if (boothowto & RB_SINGLE) {
934 			*flagsp++ = 's';
935 			options = 1;
936 		}
937 #ifdef notyet
938 		if (boothowto & RB_FASTBOOT) {
939 			*flagsp++ = 'f';
940 			options = 1;
941 		}
942 #endif
943 
944 		/*
945 		 * Move out the flags (arg 1), if necessary.
946 		 */
947 		if (options != 0) {
948 			*flagsp++ = '\0';
949 			i = flagsp - flags;
950 #ifdef DEBUG
951 			aprint_normal("init: copying out flags `%s' %d\n", flags, i);
952 #endif
953 			arg1 = STACK_ALLOC(ucp, i);
954 			ucp = STACK_MAX(arg1, i);
955 			(void)copyout((void *)flags, arg1, i);
956 		}
957 
958 		/*
959 		 * Move out the file name (also arg 0).
960 		 */
961 		i = strlen(path) + 1;
962 #ifdef DEBUG
963 		aprint_normal("init: copying out path `%s' %d\n", path, i);
964 #else
965 		if (boothowto & RB_ASKNAME || path != initpaths[0])
966 			printf("init: trying %s\n", path);
967 #endif
968 		arg0 = STACK_ALLOC(ucp, i);
969 		ucp = STACK_MAX(arg0, i);
970 		(void)copyout(path, arg0, i);
971 
972 		/*
973 		 * Move out the arg pointers.
974 		 */
975 		ucp = (void *)STACK_ALIGN(ucp, ALIGNBYTES);
976 		uap = (char **)STACK_ALLOC(ucp, sizeof(char *) * 3);
977 		SCARG(&args, path) = arg0;
978 		SCARG(&args, argp) = uap;
979 		SCARG(&args, envp) = NULL;
980 		slash = strrchr(path, '/');
981 		if (slash)
982 			(void)suword((void *)uap++,
983 			    (long)arg0 + (slash + 1 - path));
984 		else
985 			(void)suword((void *)uap++, (long)arg0);
986 		if (options != 0)
987 			(void)suword((void *)uap++, (long)arg1);
988 		(void)suword((void *)uap++, 0);	/* terminator */
989 
990 		/*
991 		 * Now try to exec the program.  If can't for any reason
992 		 * other than it doesn't exist, complain.
993 		 */
994 		error = sys_execve(l, &args, retval);
995 		if (error == 0 || error == EJUSTRETURN) {
996 			KERNEL_UNLOCK_LAST(l);
997 			return;
998 		}
999 		printf("exec %s: error %d\n", path, error);
1000 	}
1001 	printf("init: not found\n");
1002 	panic("no init");
1003 }
1004 
1005 /*
1006  * calculate cache size from physmem and vm_map size.
1007  */
1008 vaddr_t
1009 calc_cache_size(struct vm_map *map, int pct, int va_pct)
1010 {
1011 	paddr_t t;
1012 
1013 	/* XXX should consider competing cache if any */
1014 	/* XXX should consider submaps */
1015 	t = (uintmax_t)physmem * pct / 100 * PAGE_SIZE;
1016 	if (map != NULL) {
1017 		vsize_t vsize;
1018 
1019 		vsize = vm_map_max(map) - vm_map_min(map);
1020 		vsize = (uintmax_t)vsize * va_pct / 100;
1021 		if (t > vsize) {
1022 			t = vsize;
1023 		}
1024 	}
1025 	return t;
1026 }
1027 
1028 /*
1029  * Print the system start up banner.
1030  *
1031  * - Print a limited banner if AB_SILENT.
1032  * - Always send normal banner to the log.
1033  */
1034 #define MEM_PBUFSIZE	sizeof("99999 MB")
1035 
1036 void
1037 banner(void)
1038 {
1039 	static char notice[] = " Notice: this software is "
1040 	    "protected by copyright";
1041 	char pbuf[81];
1042 	void (*pr)(const char *, ...);
1043 	int i;
1044 
1045 	if ((boothowto & AB_SILENT) != 0) {
1046 		snprintf(pbuf, sizeof(pbuf), "%s %s (%s)",
1047 		    ostype, osrelease, kernel_ident);
1048 		printf_nolog("%s", pbuf);
1049 		for (i = 80 - strlen(pbuf) - sizeof(notice); i > 0; i--)
1050 			printf(" ");
1051 		printf_nolog("%s\n", notice);
1052 		pr = aprint_normal;
1053 	} else {
1054 		pr = printf;
1055 	}
1056 
1057 	memset(pbuf, 0, sizeof(pbuf));
1058 	(*pr)("%s%s", copyright, version);
1059 	format_bytes(pbuf, MEM_PBUFSIZE, ctob((uint64_t)physmem));
1060 	(*pr)("total memory = %s\n", pbuf);
1061 	format_bytes(pbuf, MEM_PBUFSIZE, ctob((uint64_t)uvmexp.free));
1062 	(*pr)("avail memory = %s\n", pbuf);
1063 }
1064