xref: /netbsd-src/sys/kern/init_main.c (revision 88fcb00c0357f2d7c1774f86a352637bfda96184)
1 /*	$NetBSD: init_main.c,v 1.433 2011/07/02 17:53:50 bouyer 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 1.433 2011/07/02 17:53:50 bouyer 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 	/* Initialize sockets thread(s) */
499 	soinit1();
500 
501 
502 	/* Configure the system hardware.  This will enable interrupts. */
503 	configure();
504 
505 	ssp_init();
506 
507 	ubc_init();		/* must be after autoconfig */
508 
509 	mm_init();
510 
511 	configure2();
512 	/* Now timer is working.  Enable preemption. */
513 	kpreempt_enable();
514 
515 #ifdef SYSVSHM
516 	/* Initialize System V style shared memory. */
517 	shminit();
518 #endif
519 
520 	vmem_rehash_start();	/* must be before exec_init */
521 
522 	/* Initialize exec structures */
523 	exec_init(1);		/* seminit calls exithook_establish() */
524 
525 #ifdef SYSVSEM
526 	/* Initialize System V style semaphores. */
527 	seminit();
528 #endif
529 
530 #ifdef SYSVMSG
531 	/* Initialize System V style message queues. */
532 	msginit();
533 #endif
534 
535 #if NVERIEXEC > 0
536 	/*
537 	 * Initialise the Veriexec subsystem.
538 	 */
539 	veriexec_init();
540 #endif /* NVERIEXEC > 0 */
541 
542 #if defined(PAX_MPROTECT) || defined(PAX_SEGVGUARD) || defined(PAX_ASLR)
543 	pax_init();
544 #endif /* PAX_MPROTECT || PAX_SEGVGUARD || PAX_ASLR */
545 
546 #ifdef	FAST_IPSEC
547 	/* Attach network crypto subsystem */
548 	ipsec_attach();
549 #endif
550 
551 	/*
552 	 * Initialize protocols.  Block reception of incoming packets
553 	 * until everything is ready.
554 	 */
555 	s = splnet();
556 	ifinit();
557 	domaininit(true);
558 	if_attachdomain();
559 	splx(s);
560 
561 #ifdef GPROF
562 	/* Initialize kernel profiling. */
563 	kmstartup();
564 #endif
565 
566 	/* Initialize system accounting. */
567 	acct_init();
568 
569 #ifndef PIPE_SOCKETPAIR
570 	/* Initialize pipes. */
571 	pipe_init();
572 #endif
573 
574 #ifdef KTRACE
575 	/* Initialize ktrace. */
576 	ktrinit();
577 #endif
578 
579 #ifdef PTRACE
580 	/* Initialize ptrace. */
581 	ptrace_init();
582 #endif /* PTRACE */
583 
584 	/* Initialize the UUID system calls. */
585 	uuid_init();
586 
587 	machdep_init();
588 
589 	procinit_sysctl();
590 
591 	/*
592 	 * Create process 1 (init(8)).  We do this now, as Unix has
593 	 * historically had init be process 1, and changing this would
594 	 * probably upset a lot of people.
595 	 *
596 	 * Note that process 1 won't immediately exec init(8), but will
597 	 * wait for us to inform it that the root file system has been
598 	 * mounted.
599 	 */
600 	if (fork1(l, 0, SIGCHLD, NULL, 0, start_init, NULL, NULL, &initproc))
601 		panic("fork init");
602 
603 	/*
604 	 * Load any remaining builtin modules, and hand back temporary
605 	 * storage to the VM system.  Then require force when loading any
606 	 * remaining un-init'ed built-in modules to avoid later surprises.
607 	 */
608 	module_init_class(MODULE_CLASS_ANY);
609 	module_builtin_require_force();
610 
611 	/*
612 	 * Finalize configuration now that all real devices have been
613 	 * found.  This needs to be done before the root device is
614 	 * selected, since finalization may create the root device.
615 	 */
616 	config_finalize();
617 
618 	sysctl_finalize();
619 
620 	/*
621 	 * Now that autoconfiguration has completed, we can determine
622 	 * the root and dump devices.
623 	 */
624 	cpu_rootconf();
625 	cpu_dumpconf();
626 
627 	/* Mount the root file system. */
628 	do {
629 		domountroothook(root_device);
630 		if ((error = vfs_mountroot())) {
631 			printf("cannot mount root, error = %d\n", error);
632 			boothowto |= RB_ASKNAME;
633 			setroot(root_device,
634 			    (rootdev != NODEV) ? DISKPART(rootdev) : 0);
635 		}
636 	} while (error != 0);
637 	mountroothook_destroy();
638 
639 	configure3();
640 
641 	/*
642 	 * Initialise the time-of-day clock, passing the time recorded
643 	 * in the root filesystem (if any) for use by systems that
644 	 * don't have a non-volatile time-of-day device.
645 	 */
646 	inittodr(rootfstime);
647 
648 	/*
649 	 * Now can look at time, having had a chance to verify the time
650 	 * from the file system.  Reset l->l_rtime as it may have been
651 	 * munched in mi_switch() after the time got set.
652 	 */
653 	getnanotime(&time);
654 	boottime = time;
655 #ifdef COMPAT_50
656 	{
657 		struct timeval tv;
658 		TIMESPEC_TO_TIMEVAL(&tv, &time);
659 		timeval_to_timeval50(&tv, &boottime50);
660 	}
661 #endif
662 	mutex_enter(proc_lock);
663 	LIST_FOREACH(p, &allproc, p_list) {
664 		KASSERT((p->p_flag & PK_MARKER) == 0);
665 		mutex_enter(p->p_lock);
666 		TIMESPEC_TO_TIMEVAL(&p->p_stats->p_start, &time);
667 		LIST_FOREACH(l, &p->p_lwps, l_sibling) {
668 			lwp_lock(l);
669 			memset(&l->l_rtime, 0, sizeof(l->l_rtime));
670 			lwp_unlock(l);
671 		}
672 		mutex_exit(p->p_lock);
673 	}
674 	mutex_exit(proc_lock);
675 	binuptime(&curlwp->l_stime);
676 
677 	for (CPU_INFO_FOREACH(cii, ci)) {
678 		ci->ci_schedstate.spc_lastmod = time_second;
679 	}
680 
681 	/* Create the pageout daemon kernel thread. */
682 	uvm_swap_init();
683 	if (kthread_create(PRI_PGDAEMON, KTHREAD_MPSAFE, NULL, uvm_pageout,
684 	    NULL, NULL, "pgdaemon"))
685 		panic("fork pagedaemon");
686 
687 	/* Create the filesystem syncer kernel thread. */
688 	if (kthread_create(PRI_IOFLUSH, KTHREAD_MPSAFE, NULL, sched_sync,
689 	    NULL, NULL, "ioflush"))
690 		panic("fork syncer");
691 
692 	/* Create the aiodone daemon kernel thread. */
693 	if (workqueue_create(&uvm.aiodone_queue, "aiodoned",
694 	    uvm_aiodone_worker, NULL, PRI_VM, IPL_NONE, WQ_MPSAFE))
695 		panic("fork aiodoned");
696 
697 	/*
698 	 * Okay, now we can let init(8) exec!  It's off to userland!
699 	 */
700 	mutex_enter(proc_lock);
701 	start_init_exec = 1;
702 	cv_broadcast(&lbolt);
703 	mutex_exit(proc_lock);
704 
705 	/* The scheduler is an infinite loop. */
706 	uvm_scheduler();
707 	/* NOTREACHED */
708 }
709 
710 /*
711  * Configure the system's hardware.
712  */
713 static void
714 configure(void)
715 {
716 
717 	/* Initialize autoconf data structures. */
718 	config_init_mi();
719 	/*
720 	 * XXX
721 	 * callout_setfunc() requires mutex(9) so it can't be in config_init()
722 	 * on amiga and atari which use config_init() and autoconf(9) fucntions
723 	 * to initialize console devices.
724 	 */
725 	config_twiddle_init();
726 
727 	pmf_init();
728 #if NDRVCTL > 0
729 	drvctl_init();
730 #endif
731 
732 	userconf_init();
733 	if (boothowto & RB_USERCONF)
734 		userconf_prompt();
735 
736 	if ((boothowto & (AB_SILENT|AB_VERBOSE)) == AB_SILENT) {
737 		printf_nolog("Detecting hardware...");
738 	}
739 
740 	/*
741 	 * Do the machine-dependent portion of autoconfiguration.  This
742 	 * sets the configuration machinery here in motion by "finding"
743 	 * the root bus.  When this function returns, we expect interrupts
744 	 * to be enabled.
745 	 */
746 	cpu_configure();
747 }
748 
749 static void
750 configure2(void)
751 {
752 	CPU_INFO_ITERATOR cii;
753 	struct cpu_info *ci;
754 	int s;
755 
756 	/*
757 	 * Now that we've found all the hardware, start the real time
758 	 * and statistics clocks.
759 	 */
760 	initclocks();
761 
762 	cold = 0;	/* clocks are running, we're warm now! */
763 	s = splsched();
764 	curcpu()->ci_schedstate.spc_flags |= SPCF_RUNNING;
765 	splx(s);
766 
767 	/* Boot the secondary processors. */
768 	for (CPU_INFO_FOREACH(cii, ci)) {
769 		uvm_cpu_attach(ci);
770 	}
771 	mp_online = true;
772 #if defined(MULTIPROCESSOR)
773 	cpu_boot_secondary_processors();
774 #endif
775 
776 	/* Setup the runqueues and scheduler. */
777 	runq_init();
778 	synch_init();
779 
780 	/*
781 	 * Bus scans can make it appear as if the system has paused, so
782 	 * twiddle constantly while config_interrupts() jobs are running.
783 	 */
784 	config_twiddle_fn(NULL);
785 
786 	/*
787 	 * Create threads to call back and finish configuration for
788 	 * devices that want interrupts enabled.
789 	 */
790 	config_create_interruptthreads();
791 
792 	/* Get the threads going and into any sleeps before continuing. */
793 	yield();
794 }
795 
796 static void
797 configure3(void)
798 {
799 
800 	/*
801 	 * Create threads to call back and finish configuration for
802 	 * devices that want the mounted root file system.
803 	 */
804 	config_create_mountrootthreads();
805 
806 	/* Get the threads going and into any sleeps before continuing. */
807 	yield();
808 }
809 
810 static void
811 check_console(struct lwp *l)
812 {
813 	struct vnode *vp;
814 	int error;
815 
816 	error = namei_simple_kernel("/dev/console",
817 				NSM_FOLLOW_NOEMULROOT, &vp);
818 	if (error == 0)
819 		vrele(vp);
820 	else if (error == ENOENT)
821 		printf("warning: no /dev/console\n");
822 	else
823 		printf("warning: lookup /dev/console: error %d\n", error);
824 }
825 
826 /*
827  * List of paths to try when searching for "init".
828  */
829 static const char * const initpaths[] = {
830 	"/sbin/init",
831 	"/sbin/oinit",
832 	"/sbin/init.bak",
833 	NULL,
834 };
835 
836 /*
837  * Start the initial user process; try exec'ing each pathname in "initpaths".
838  * The program is invoked with one argument containing the boot flags.
839  */
840 static void
841 start_init(void *arg)
842 {
843 	struct lwp *l = arg;
844 	struct proc *p = l->l_proc;
845 	vaddr_t addr;
846 	struct sys_execve_args /* {
847 		syscallarg(const char *) path;
848 		syscallarg(char * const *) argp;
849 		syscallarg(char * const *) envp;
850 	} */ args;
851 	int options, i, error;
852 	register_t retval[2];
853 	char flags[4], *flagsp;
854 	const char *path, *slash;
855 	char *ucp, **uap, *arg0, *arg1 = NULL;
856 	char ipath[129];
857 	int ipx, len;
858 
859 	/*
860 	 * Now in process 1.
861 	 */
862 	strncpy(p->p_comm, "init", MAXCOMLEN);
863 
864 	/*
865 	 * Wait for main() to tell us that it's safe to exec.
866 	 */
867 	mutex_enter(proc_lock);
868 	while (start_init_exec == 0)
869 		cv_wait(&lbolt, proc_lock);
870 	mutex_exit(proc_lock);
871 
872 	/*
873 	 * This is not the right way to do this.  We really should
874 	 * hand-craft a descriptor onto /dev/console to hand to init,
875 	 * but that's a _lot_ more work, and the benefit from this easy
876 	 * hack makes up for the "good is the enemy of the best" effect.
877 	 */
878 	check_console(l);
879 
880 	/*
881 	 * Need just enough stack to hold the faked-up "execve()" arguments.
882 	 */
883 	addr = (vaddr_t)STACK_ALLOC(USRSTACK, PAGE_SIZE);
884 	if (uvm_map(&p->p_vmspace->vm_map, &addr, PAGE_SIZE,
885                     NULL, UVM_UNKNOWN_OFFSET, 0,
886                     UVM_MAPFLAG(UVM_PROT_ALL, UVM_PROT_ALL, UVM_INH_COPY,
887 		    UVM_ADV_NORMAL,
888                     UVM_FLAG_FIXED|UVM_FLAG_OVERLAY|UVM_FLAG_COPYONW)) != 0)
889 		panic("init: couldn't allocate argument space");
890 	p->p_vmspace->vm_maxsaddr = (void *)STACK_MAX(addr, PAGE_SIZE);
891 
892 	ipx = 0;
893 	while (1) {
894 		if (boothowto & RB_ASKNAME) {
895 			printf("init path");
896 			if (initpaths[ipx])
897 				printf(" (default %s)", initpaths[ipx]);
898 			printf(": ");
899 			len = cngetsn(ipath, sizeof(ipath)-1);
900 			if (len == 4 && strcmp(ipath, "halt") == 0) {
901 				cpu_reboot(RB_HALT, NULL);
902 			} else if (len == 6 && strcmp(ipath, "reboot") == 0) {
903 				cpu_reboot(0, NULL);
904 #if defined(DDB)
905 			} else if (len == 3 && strcmp(ipath, "ddb") == 0) {
906 				console_debugger();
907 				continue;
908 #endif
909 			} else if (len > 0 && ipath[0] == '/') {
910 				ipath[len] = '\0';
911 				path = ipath;
912 			} else if (len == 0 && initpaths[ipx] != NULL) {
913 				path = initpaths[ipx++];
914 			} else {
915 				printf("use absolute path, ");
916 #if defined(DDB)
917 				printf("\"ddb\", ");
918 #endif
919 				printf("\"halt\", or \"reboot\"\n");
920 				continue;
921 			}
922 		} else {
923 			if ((path = initpaths[ipx++]) == NULL) {
924 				ipx = 0;
925 				boothowto |= RB_ASKNAME;
926 				continue;
927 			}
928 		}
929 
930 		ucp = (char *)USRSTACK;
931 
932 		/*
933 		 * Construct the boot flag argument.
934 		 */
935 		flagsp = flags;
936 		*flagsp++ = '-';
937 		options = 0;
938 
939 		if (boothowto & RB_SINGLE) {
940 			*flagsp++ = 's';
941 			options = 1;
942 		}
943 #ifdef notyet
944 		if (boothowto & RB_FASTBOOT) {
945 			*flagsp++ = 'f';
946 			options = 1;
947 		}
948 #endif
949 
950 		/*
951 		 * Move out the flags (arg 1), if necessary.
952 		 */
953 		if (options != 0) {
954 			*flagsp++ = '\0';
955 			i = flagsp - flags;
956 #ifdef DEBUG
957 			aprint_normal("init: copying out flags `%s' %d\n", flags, i);
958 #endif
959 			arg1 = STACK_ALLOC(ucp, i);
960 			ucp = STACK_MAX(arg1, i);
961 			(void)copyout((void *)flags, arg1, i);
962 		}
963 
964 		/*
965 		 * Move out the file name (also arg 0).
966 		 */
967 		i = strlen(path) + 1;
968 #ifdef DEBUG
969 		aprint_normal("init: copying out path `%s' %d\n", path, i);
970 #else
971 		if (boothowto & RB_ASKNAME || path != initpaths[0])
972 			printf("init: trying %s\n", path);
973 #endif
974 		arg0 = STACK_ALLOC(ucp, i);
975 		ucp = STACK_MAX(arg0, i);
976 		(void)copyout(path, arg0, i);
977 
978 		/*
979 		 * Move out the arg pointers.
980 		 */
981 		ucp = (void *)STACK_ALIGN(ucp, ALIGNBYTES);
982 		uap = (char **)STACK_ALLOC(ucp, sizeof(char *) * 3);
983 		SCARG(&args, path) = arg0;
984 		SCARG(&args, argp) = uap;
985 		SCARG(&args, envp) = NULL;
986 		slash = strrchr(path, '/');
987 		if (slash)
988 			(void)suword((void *)uap++,
989 			    (long)arg0 + (slash + 1 - path));
990 		else
991 			(void)suword((void *)uap++, (long)arg0);
992 		if (options != 0)
993 			(void)suword((void *)uap++, (long)arg1);
994 		(void)suword((void *)uap++, 0);	/* terminator */
995 
996 		/*
997 		 * Now try to exec the program.  If can't for any reason
998 		 * other than it doesn't exist, complain.
999 		 */
1000 		error = sys_execve(l, &args, retval);
1001 		if (error == 0 || error == EJUSTRETURN) {
1002 			KERNEL_UNLOCK_LAST(l);
1003 			return;
1004 		}
1005 		printf("exec %s: error %d\n", path, error);
1006 	}
1007 	printf("init: not found\n");
1008 	panic("no init");
1009 }
1010 
1011 /*
1012  * calculate cache size (in bytes) from physmem and vm_map size.
1013  */
1014 vaddr_t
1015 calc_cache_size(struct vm_map *map, int pct, int va_pct)
1016 {
1017 	paddr_t t;
1018 
1019 	/* XXX should consider competing cache if any */
1020 	/* XXX should consider submaps */
1021 	t = (uintmax_t)physmem * pct / 100 * PAGE_SIZE;
1022 	if (map != NULL) {
1023 		vsize_t vsize;
1024 
1025 		vsize = vm_map_max(map) - vm_map_min(map);
1026 		vsize = (uintmax_t)vsize * va_pct / 100;
1027 		if (t > vsize) {
1028 			t = vsize;
1029 		}
1030 	}
1031 	return t;
1032 }
1033 
1034 /*
1035  * Print the system start up banner.
1036  *
1037  * - Print a limited banner if AB_SILENT.
1038  * - Always send normal banner to the log.
1039  */
1040 #define MEM_PBUFSIZE	sizeof("99999 MB")
1041 
1042 void
1043 banner(void)
1044 {
1045 	static char notice[] = " Notice: this software is "
1046 	    "protected by copyright";
1047 	char pbuf[81];
1048 	void (*pr)(const char *, ...);
1049 	int i;
1050 
1051 	if ((boothowto & AB_SILENT) != 0) {
1052 		snprintf(pbuf, sizeof(pbuf), "%s %s (%s)",
1053 		    ostype, osrelease, kernel_ident);
1054 		printf_nolog("%s", pbuf);
1055 		for (i = 80 - strlen(pbuf) - sizeof(notice); i > 0; i--)
1056 			printf(" ");
1057 		printf_nolog("%s\n", notice);
1058 		pr = aprint_normal;
1059 	} else {
1060 		pr = printf;
1061 	}
1062 
1063 	memset(pbuf, 0, sizeof(pbuf));
1064 	(*pr)("%s%s", copyright, version);
1065 	format_bytes(pbuf, MEM_PBUFSIZE, ctob((uint64_t)physmem));
1066 	(*pr)("total memory = %s\n", pbuf);
1067 	format_bytes(pbuf, MEM_PBUFSIZE, ctob((uint64_t)uvmexp.free));
1068 	(*pr)("avail memory = %s\n", pbuf);
1069 }
1070