xref: /netbsd-src/sys/kern/init_main.c (revision 2de962bd804263c16657f586aa00f1704045df8e)
1 /*	$NetBSD: init_main.c,v 1.355 2008/05/01 14:44:48 ad Exp $	*/
2 
3 /*-
4  * Copyright (c) 2008 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 1.355 2008/05/01 14:44:48 ad Exp $");
101 
102 #include "opt_ipsec.h"
103 #include "opt_ntp.h"
104 #include "opt_pipe.h"
105 #include "opt_posix.h"
106 #include "opt_syscall_debug.h"
107 #include "opt_sysv.h"
108 #include "opt_fileassoc.h"
109 #include "opt_ktrace.h"
110 #include "opt_pax.h"
111 
112 #include "rnd.h"
113 #include "sysmon_envsys.h"
114 #include "sysmon_power.h"
115 #include "sysmon_taskq.h"
116 #include "sysmon_wdog.h"
117 #include "veriexec.h"
118 
119 #include <sys/param.h>
120 #include <sys/acct.h>
121 #include <sys/filedesc.h>
122 #include <sys/file.h>
123 #include <sys/errno.h>
124 #include <sys/callout.h>
125 #include <sys/cpu.h>
126 #include <sys/kernel.h>
127 #include <sys/kmem.h>
128 #include <sys/mount.h>
129 #include <sys/proc.h>
130 #include <sys/kthread.h>
131 #include <sys/resourcevar.h>
132 #include <sys/signalvar.h>
133 #include <sys/systm.h>
134 #include <sys/vnode.h>
135 #include <sys/fstrans.h>
136 #include <sys/tty.h>
137 #include <sys/conf.h>
138 #include <sys/disklabel.h>
139 #include <sys/buf.h>
140 #include <sys/device.h>
141 #include <sys/exec.h>
142 #include <sys/socketvar.h>
143 #include <sys/protosw.h>
144 #include <sys/percpu.h>
145 #include <sys/pset.h>
146 #include <sys/sysctl.h>
147 #include <sys/reboot.h>
148 #include <sys/user.h>
149 #include <sys/sysctl.h>
150 #include <sys/event.h>
151 #include <sys/mbuf.h>
152 #include <sys/sched.h>
153 #include <sys/sleepq.h>
154 #include <sys/iostat.h>
155 #include <sys/vmem.h>
156 #include <sys/uuid.h>
157 #include <sys/extent.h>
158 #include <sys/disk.h>
159 #include <sys/mqueue.h>
160 #include <sys/msgbuf.h>
161 #include <sys/module.h>
162 #include <sys/event.h>
163 #ifdef FAST_IPSEC
164 #include <netipsec/ipsec.h>
165 #endif
166 #ifdef SYSVSHM
167 #include <sys/shm.h>
168 #endif
169 #ifdef SYSVSEM
170 #include <sys/sem.h>
171 #endif
172 #ifdef SYSVMSG
173 #include <sys/msg.h>
174 #endif
175 #ifdef P1003_1B_SEMAPHORE
176 #include <sys/ksem.h>
177 #endif
178 #include <sys/domain.h>
179 #include <sys/namei.h>
180 #if NRND > 0
181 #include <sys/rnd.h>
182 #endif
183 #include <sys/pipe.h>
184 #ifdef LKM
185 #include <sys/lkm.h>
186 #endif
187 #if NVERIEXEC > 0
188 #include <sys/verified_exec.h>
189 #endif /* NVERIEXEC > 0 */
190 #ifdef KTRACE
191 #include <sys/ktrace.h>
192 #endif
193 #include <sys/kauth.h>
194 #include <net80211/ieee80211_netbsd.h>
195 
196 #include <sys/syscall.h>
197 #include <sys/syscallargs.h>
198 
199 #if defined(PAX_MPROTECT) || defined(PAX_SEGVGUARD) || defined(PAX_ASLR)
200 #include <sys/pax.h>
201 #endif /* PAX_MPROTECT || PAX_SEGVGUARD || PAX_ASLR */
202 
203 #include <ufs/ufs/quota.h>
204 
205 #include <miscfs/genfs/genfs.h>
206 #include <miscfs/syncfs/syncfs.h>
207 
208 #include <sys/cpu.h>
209 
210 #include <uvm/uvm.h>
211 
212 #if NSYSMON_TASKQ > 0
213 #include <dev/sysmon/sysmon_taskq.h>
214 #endif
215 
216 #include <dev/cons.h>
217 
218 #if NSYSMON_ENVSYS > 0 || NSYSMON_POWER > 0 || NSYSMON_WDOG > 0
219 #include <dev/sysmon/sysmonvar.h>
220 #endif
221 
222 #include <net/if.h>
223 #include <net/raw_cb.h>
224 
225 #include <secmodel/secmodel.h>
226 
227 extern struct proc proc0;
228 extern struct lwp lwp0;
229 extern struct cwdinfo cwdi0;
230 extern time_t rootfstime;
231 
232 #ifndef curlwp
233 struct	lwp *curlwp = &lwp0;
234 #endif
235 struct	proc *initproc;
236 
237 struct	vnode *rootvp, *swapdev_vp;
238 int	boothowto;
239 int	cold = 1;			/* still working on startup */
240 struct timeval boottime;	        /* time at system startup - will only follow settime deltas */
241 
242 volatile int start_init_exec;		/* semaphore for start_init() */
243 
244 static void check_console(struct lwp *l);
245 static void start_init(void *);
246 void main(void);
247 void ssp_init(void);
248 
249 #if defined(__SSP__) || defined(__SSP_ALL__)
250 long __stack_chk_guard[8] = {0, 0, 0, 0, 0, 0, 0, 0};
251 void __stack_chk_fail(void);
252 
253 void
254 __stack_chk_fail(void)
255 {
256 	panic("stack overflow detected; terminated");
257 }
258 
259 void
260 ssp_init(void)
261 {
262 	int s;
263 
264 #ifdef DIAGNOSTIC
265 	printf("Initializing SSP:");
266 #endif
267 	/*
268 	 * We initialize ssp here carefully:
269 	 *	1. after we got some entropy
270 	 *	2. without calling a function
271 	 */
272 	size_t i;
273 	long guard[__arraycount(__stack_chk_guard)];
274 
275 	arc4randbytes(guard, sizeof(guard));
276 	s = splhigh();
277 	for (i = 0; i < __arraycount(guard); i++)
278 		__stack_chk_guard[i] = guard[i];
279 	splx(s);
280 #ifdef DIAGNOSTIC
281 	for (i = 0; i < __arraycount(guard); i++)
282 		printf("%lx ", guard[i]);
283 	printf("\n");
284 #endif
285 }
286 #else
287 void
288 ssp_init(void)
289 {
290 
291 }
292 #endif
293 
294 void __secmodel_none(void);
295 __weak_alias(secmodel_start,__secmodel_none);
296 void
297 __secmodel_none(void)
298 {
299 	return;
300 }
301 
302 /*
303  * System startup; initialize the world, create process 0, mount root
304  * filesystem, and fork to create init and pagedaemon.  Most of the
305  * hard work is done in the lower-level initialization routines including
306  * startup(), which does memory initialization and autoconfiguration.
307  */
308 void
309 main(void)
310 {
311 	struct timeval time;
312 	struct lwp *l;
313 	struct proc *p;
314 	int s, error;
315 #ifdef NVNODE_IMPLICIT
316 	int usevnodes;
317 #endif
318 	CPU_INFO_ITERATOR cii;
319 	struct cpu_info *ci;
320 
321 	l = &lwp0;
322 #ifndef LWP0_CPU_INFO
323 	l->l_cpu = curcpu();
324 #endif
325 
326 	/*
327 	 * Attempt to find console and initialize
328 	 * in case of early panic or other messages.
329 	 */
330 	consinit();
331 
332 	kernel_lock_init();
333 
334 	uvm_init();
335 
336 	kmem_init();
337 
338 	percpu_init();
339 
340 	/* Initialize lock caches. */
341 	mutex_obj_init();
342 
343 	/* Initialize the extent manager. */
344 	extent_init();
345 
346 	/* Do machine-dependent initialization. */
347 	cpu_startup();
348 
349 	/* Initialize callouts, part 1. */
350 	callout_startup();
351 
352 	/*
353 	 * Initialize the kernel authorization subsystem and start the
354 	 * default security model, if any. We need to do this early
355 	 * enough so that subsystems relying on any of the aforementioned
356 	 * can work properly. Since the security model may dictate the
357 	 * credential inheritance policy, it is needed at least before
358 	 * any process is created, specifically proc0.
359 	 */
360 	kauth_init();
361 	secmodel_start();
362 
363 	/* Initialize the buffer cache */
364 	bufinit();
365 
366 	/* Initialize sockets. */
367 	soinit();
368 
369 	/*
370 	 * The following things must be done before autoconfiguration.
371 	 */
372 	evcnt_init();		/* initialize event counters */
373 #if NRND > 0
374 	rnd_init();		/* initialize RNG */
375 #endif
376 
377 	/* Initialize process and pgrp structures. */
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 (the swapper). */
388 	proc0_init();
389 
390 	/* Initialize the UID hash table. */
391 	uid_init();
392 
393 	/* Charge root for one process. */
394 	(void)chgproccnt(0, 1);
395 
396 	/* Initialize timekeeping. */
397 	time_init();
398 
399 	/* Initialize the run queues, turnstiles and sleep queues. */
400 	mutex_init(&cpu_lock, MUTEX_DEFAULT, IPL_NONE);
401 	sched_rqinit();
402 	turnstile_init();
403 	sleeptab_init(&sleeptab);
404 
405 	/* Initialize processor-sets */
406 	psets_init();
407 
408 	/* MI initialization of the boot cpu */
409 	error = mi_cpu_attach(curcpu());
410 	KASSERT(error == 0);
411 
412 	/* Initialize timekeeping, part 2. */
413 	time_init2();
414 
415 	/*
416 	 * Initialize mbuf's.  Do this now because we might attempt to
417 	 * allocate mbufs or mbuf clusters during autoconfiguration.
418 	 */
419 	mbinit();
420 
421 	/* Initialize the sysctl subsystem. */
422 	sysctl_init();
423 
424 	/* Initialize I/O statistics. */
425 	iostat_init();
426 
427 	/* Initialize the log device. */
428 	loginit();
429 
430 	/* Start module system. */
431 	module_init();
432 
433 	/* Initialize the file systems. */
434 #ifdef NVNODE_IMPLICIT
435 	/*
436 	 * If maximum number of vnodes in namei vnode cache is not explicitly
437 	 * defined in kernel config, adjust the number such as we use roughly
438 	 * 1.0% of memory for vnode cache (but not less than NVNODE vnodes).
439 	 */
440 	usevnodes =
441 	    calc_cache_size(kernel_map, 1, VNODE_VA_MAXPCT) / sizeof(vnode_t);
442 	if (usevnodes > desiredvnodes)
443 		desiredvnodes = usevnodes;
444 #endif
445 	vfsinit();
446 
447 	/* Initialize fstrans. */
448 	fstrans_init();
449 
450 	/* Initialize the file descriptor system. */
451 	fd_sys_init();
452 
453 	/* Initialize kqueue. */
454 	kqueue_init();
455 
456 	/* Initialize asynchronous I/O. */
457 	aio_sysinit();
458 
459 	/* Initialize message queues. */
460 	mqueue_sysinit();
461 
462 	/* Initialize the system monitor subsystems. */
463 #if NSYSMON_TASKQ > 0
464 	sysmon_task_queue_preinit();
465 #endif
466 
467 #if NSYSMON_ENVSYS > 0
468 	sysmon_envsys_init();
469 #endif
470 
471 #if NSYSMON_POWER > 0
472 	sysmon_power_init();
473 #endif
474 
475 #if NSYSMON_WDOG > 0
476 	sysmon_wdog_init();
477 #endif
478 
479 	inittimecounter();
480 	ntp_init();
481 
482 	/* Initialize the device switch tables. */
483 	devsw_init();
484 
485 	/* Initialize tty subsystem. */
486 	tty_init();
487 	ttyldisc_init();
488 
489 	/* Initialize the buffer cache, part 2. */
490 	bufinit2();
491 
492 	/* Initialize the disk wedge subsystem. */
493 	dkwedge_init();
494 
495 	/* Configure the system hardware.  This will enable interrupts. */
496 	configure();
497 
498 	ubc_init();		/* must be after autoconfig */
499 
500 #ifdef SYSVSHM
501 	/* Initialize System V style shared memory. */
502 	shminit();
503 #endif
504 
505 #ifdef SYSVSEM
506 	/* Initialize System V style semaphores. */
507 	seminit();
508 #endif
509 
510 #ifdef SYSVMSG
511 	/* Initialize System V style message queues. */
512 	msginit();
513 #endif
514 
515 #ifdef P1003_1B_SEMAPHORE
516 	/* Initialize posix semaphores */
517 	ksem_init();
518 #endif
519 
520 #if NVERIEXEC > 0
521 	/*
522 	 * Initialise the Veriexec subsystem.
523 	 */
524 	veriexec_init();
525 #endif /* NVERIEXEC > 0 */
526 
527 #if defined(PAX_MPROTECT) || defined(PAX_SEGVGUARD) || defined(PAX_ASLR)
528 	pax_init();
529 #endif /* PAX_MPROTECT || PAX_SEGVGUARD || PAX_ASLR */
530 
531 #ifdef	FAST_IPSEC
532 	/* Attach network crypto subsystem */
533 	ipsec_attach();
534 #endif
535 
536 	/*
537 	 * Initialize protocols.  Block reception of incoming packets
538 	 * until everything is ready.
539 	 */
540 	s = splnet();
541 	ifinit();
542 	domaininit();
543 	if_attachdomain();
544 	splx(s);
545 
546 #ifdef GPROF
547 	/* Initialize kernel profiling. */
548 	kmstartup();
549 #endif
550 
551 	/* Initialize system accounting. */
552 	acct_init();
553 
554 #ifndef PIPE_SOCKETPAIR
555 	/* Initialize pipes. */
556 	pipe_init();
557 #endif
558 
559 #ifdef KTRACE
560 	/* Initialize ktrace. */
561 	ktrinit();
562 #endif
563 
564 	/* Initialize the UUID system calls. */
565 	uuid_init();
566 
567 	/*
568 	 * Create process 1 (init(8)).  We do this now, as Unix has
569 	 * historically had init be process 1, and changing this would
570 	 * probably upset a lot of people.
571 	 *
572 	 * Note that process 1 won't immediately exec init(8), but will
573 	 * wait for us to inform it that the root file system has been
574 	 * mounted.
575 	 */
576 	if (fork1(l, 0, SIGCHLD, NULL, 0, start_init, NULL, NULL, &initproc))
577 		panic("fork init");
578 
579 	/*
580 	 * Load any remaining builtin modules, and hand back temporary
581 	 * storage to the VM system.
582 	 */
583 	module_init_class(MODULE_CLASS_ANY);
584 	module_jettison();
585 
586 	/*
587 	 * Finalize configuration now that all real devices have been
588 	 * found.  This needs to be done before the root device is
589 	 * selected, since finalization may create the root device.
590 	 */
591 	config_finalize();
592 
593 	/*
594 	 * Now that autoconfiguration has completed, we can determine
595 	 * the root and dump devices.
596 	 */
597 	cpu_rootconf();
598 	cpu_dumpconf();
599 
600 	/* Mount the root file system. */
601 	do {
602 		domountroothook();
603 		if ((error = vfs_mountroot())) {
604 			printf("cannot mount root, error = %d\n", error);
605 			boothowto |= RB_ASKNAME;
606 			setroot(root_device,
607 			    (rootdev != NODEV) ? DISKPART(rootdev) : 0);
608 		}
609 	} while (error != 0);
610 	mountroothook_destroy();
611 
612 	/*
613 	 * Initialise the time-of-day clock, passing the time recorded
614 	 * in the root filesystem (if any) for use by systems that
615 	 * don't have a non-volatile time-of-day device.
616 	 */
617 	inittodr(rootfstime);
618 
619 	CIRCLEQ_FIRST(&mountlist)->mnt_flag |= MNT_ROOTFS;
620 	CIRCLEQ_FIRST(&mountlist)->mnt_op->vfs_refcount++;
621 
622 	/*
623 	 * Get the vnode for '/'.  Set filedesc0.fd_fd.fd_cdir to
624 	 * reference it.
625 	 */
626 	error = VFS_ROOT(CIRCLEQ_FIRST(&mountlist), &rootvnode);
627 	if (error)
628 		panic("cannot find root vnode, error=%d", error);
629 	cwdi0.cwdi_cdir = rootvnode;
630 	VREF(cwdi0.cwdi_cdir);
631 	VOP_UNLOCK(rootvnode, 0);
632 	cwdi0.cwdi_rdir = NULL;
633 
634 	/*
635 	 * Now that root is mounted, we can fixup initproc's CWD
636 	 * info.  All other processes are kthreads, which merely
637 	 * share proc0's CWD info.
638 	 */
639 	initproc->p_cwdi->cwdi_cdir = rootvnode;
640 	VREF(initproc->p_cwdi->cwdi_cdir);
641 	initproc->p_cwdi->cwdi_rdir = NULL;
642 
643 	/*
644 	 * Now can look at time, having had a chance to verify the time
645 	 * from the file system.  Reset l->l_rtime as it may have been
646 	 * munched in mi_switch() after the time got set.
647 	 */
648 	getmicrotime(&time);
649 	boottime = time;
650 	mutex_enter(proc_lock);
651 	LIST_FOREACH(p, &allproc, p_list) {
652 		KASSERT((p->p_flag & PK_MARKER) == 0);
653 		mutex_enter(p->p_lock);
654 		p->p_stats->p_start = time;
655 		LIST_FOREACH(l, &p->p_lwps, l_sibling) {
656 			lwp_lock(l);
657 			memset(&l->l_rtime, 0, sizeof(l->l_rtime));
658 			lwp_unlock(l);
659 		}
660 		mutex_exit(p->p_lock);
661 	}
662 	mutex_exit(proc_lock);
663 	binuptime(&curlwp->l_stime);
664 
665 	for (CPU_INFO_FOREACH(cii, ci)) {
666 		ci->ci_schedstate.spc_lastmod = time_second;
667 	}
668 
669 	/* Create the pageout daemon kernel thread. */
670 	uvm_swap_init();
671 	if (kthread_create(PRI_PGDAEMON, KTHREAD_MPSAFE, NULL, uvm_pageout,
672 	    NULL, NULL, "pgdaemon"))
673 		panic("fork pagedaemon");
674 
675 	/* Create the filesystem syncer kernel thread. */
676 	if (kthread_create(PRI_IOFLUSH, KTHREAD_MPSAFE, NULL, sched_sync,
677 	    NULL, NULL, "ioflush"))
678 		panic("fork syncer");
679 
680 	/* Create the aiodone daemon kernel thread. */
681 	if (workqueue_create(&uvm.aiodone_queue, "aiodoned",
682 	    uvm_aiodone_worker, NULL, PRI_VM, IPL_NONE, WQ_MPSAFE))
683 		panic("fork aiodoned");
684 
685 	vmem_rehash_start();
686 
687 	/* Initialize exec structures */
688 	exec_init(1);
689 
690 	/*
691 	 * Okay, now we can let init(8) exec!  It's off to userland!
692 	 */
693 	start_init_exec = 1;
694 	wakeup(&start_init_exec);
695 
696 	/* The scheduler is an infinite loop. */
697 	uvm_scheduler();
698 	/* NOTREACHED */
699 }
700 
701 static void
702 check_console(struct lwp *l)
703 {
704 	struct nameidata nd;
705 	int error;
706 
707 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_SYSSPACE, "/dev/console");
708 	error = namei(&nd);
709 	if (error == 0)
710 		vrele(nd.ni_vp);
711 	else if (error == ENOENT)
712 		printf("warning: no /dev/console\n");
713 	else
714 		printf("warning: lookup /dev/console: error %d\n", error);
715 }
716 
717 /*
718  * List of paths to try when searching for "init".
719  */
720 static const char * const initpaths[] = {
721 	"/sbin/init",
722 	"/sbin/oinit",
723 	"/sbin/init.bak",
724 	NULL,
725 };
726 
727 /*
728  * Start the initial user process; try exec'ing each pathname in "initpaths".
729  * The program is invoked with one argument containing the boot flags.
730  */
731 static void
732 start_init(void *arg)
733 {
734 	struct lwp *l = arg;
735 	struct proc *p = l->l_proc;
736 	vaddr_t addr;
737 	struct sys_execve_args /* {
738 		syscallarg(const char *) path;
739 		syscallarg(char * const *) argp;
740 		syscallarg(char * const *) envp;
741 	} */ args;
742 	int options, i, error;
743 	register_t retval[2];
744 	char flags[4], *flagsp;
745 	const char *path, *slash;
746 	char *ucp, **uap, *arg0, *arg1 = NULL;
747 	char ipath[129];
748 	int ipx, len;
749 
750 	/*
751 	 * Now in process 1.
752 	 */
753 	strncpy(p->p_comm, "init", MAXCOMLEN);
754 
755 	/*
756 	 * Wait for main() to tell us that it's safe to exec.
757 	 */
758 	while (start_init_exec == 0)
759 		(void) tsleep(&start_init_exec, PWAIT, "initexec", 0);
760 
761 	/*
762 	 * This is not the right way to do this.  We really should
763 	 * hand-craft a descriptor onto /dev/console to hand to init,
764 	 * but that's a _lot_ more work, and the benefit from this easy
765 	 * hack makes up for the "good is the enemy of the best" effect.
766 	 */
767 	check_console(l);
768 
769 	/*
770 	 * Need just enough stack to hold the faked-up "execve()" arguments.
771 	 */
772 	addr = (vaddr_t)STACK_ALLOC(USRSTACK, PAGE_SIZE);
773 	if (uvm_map(&p->p_vmspace->vm_map, &addr, PAGE_SIZE,
774                     NULL, UVM_UNKNOWN_OFFSET, 0,
775                     UVM_MAPFLAG(UVM_PROT_ALL, UVM_PROT_ALL, UVM_INH_COPY,
776 		    UVM_ADV_NORMAL,
777                     UVM_FLAG_FIXED|UVM_FLAG_OVERLAY|UVM_FLAG_COPYONW)) != 0)
778 		panic("init: couldn't allocate argument space");
779 	p->p_vmspace->vm_maxsaddr = (void *)STACK_MAX(addr, PAGE_SIZE);
780 
781 	ipx = 0;
782 	while (1) {
783 		if (boothowto & RB_ASKNAME) {
784 			printf("init path");
785 			if (initpaths[ipx])
786 				printf(" (default %s)", initpaths[ipx]);
787 			printf(": ");
788 			len = cngetsn(ipath, sizeof(ipath)-1);
789 			if (len == 0) {
790 				if (initpaths[ipx])
791 					path = initpaths[ipx++];
792 				else
793 					continue;
794 			} else {
795 				ipath[len] = '\0';
796 				path = ipath;
797 			}
798 		} else {
799 			if ((path = initpaths[ipx++]) == NULL)
800 				break;
801 		}
802 
803 		ucp = (char *)USRSTACK;
804 
805 		/*
806 		 * Construct the boot flag argument.
807 		 */
808 		flagsp = flags;
809 		*flagsp++ = '-';
810 		options = 0;
811 
812 		if (boothowto & RB_SINGLE) {
813 			*flagsp++ = 's';
814 			options = 1;
815 		}
816 #ifdef notyet
817 		if (boothowto & RB_FASTBOOT) {
818 			*flagsp++ = 'f';
819 			options = 1;
820 		}
821 #endif
822 
823 		/*
824 		 * Move out the flags (arg 1), if necessary.
825 		 */
826 		if (options != 0) {
827 			*flagsp++ = '\0';
828 			i = flagsp - flags;
829 #ifdef DEBUG
830 			printf("init: copying out flags `%s' %d\n", flags, i);
831 #endif
832 			arg1 = STACK_ALLOC(ucp, i);
833 			ucp = STACK_MAX(arg1, i);
834 			(void)copyout((void *)flags, arg1, i);
835 		}
836 
837 		/*
838 		 * Move out the file name (also arg 0).
839 		 */
840 		i = strlen(path) + 1;
841 #ifdef DEBUG
842 		printf("init: copying out path `%s' %d\n", path, i);
843 #else
844 		if (boothowto & RB_ASKNAME || path != initpaths[0])
845 			printf("init: trying %s\n", path);
846 #endif
847 		arg0 = STACK_ALLOC(ucp, i);
848 		ucp = STACK_MAX(arg0, i);
849 		(void)copyout(path, arg0, i);
850 
851 		/*
852 		 * Move out the arg pointers.
853 		 */
854 		ucp = (void *)STACK_ALIGN(ucp, ALIGNBYTES);
855 		uap = (char **)STACK_ALLOC(ucp, sizeof(char *) * 3);
856 		SCARG(&args, path) = arg0;
857 		SCARG(&args, argp) = uap;
858 		SCARG(&args, envp) = NULL;
859 		slash = strrchr(path, '/');
860 		if (slash)
861 			(void)suword((void *)uap++,
862 			    (long)arg0 + (slash + 1 - path));
863 		else
864 			(void)suword((void *)uap++, (long)arg0);
865 		if (options != 0)
866 			(void)suword((void *)uap++, (long)arg1);
867 		(void)suword((void *)uap++, 0);	/* terminator */
868 
869 		/*
870 		 * Now try to exec the program.  If can't for any reason
871 		 * other than it doesn't exist, complain.
872 		 */
873 		error = sys_execve(l, &args, retval);
874 		if (error == 0 || error == EJUSTRETURN) {
875 			KERNEL_UNLOCK_LAST(l);
876 			return;
877 		}
878 		printf("exec %s: error %d\n", path, error);
879 	}
880 	printf("init: not found\n");
881 	panic("no init");
882 }
883 
884 /*
885  * calculate cache size from physmem and vm_map size.
886  */
887 vaddr_t
888 calc_cache_size(struct vm_map *map, int pct, int va_pct)
889 {
890 	paddr_t t;
891 
892 	/* XXX should consider competing cache if any */
893 	/* XXX should consider submaps */
894 	t = (uintmax_t)physmem * pct / 100 * PAGE_SIZE;
895 	if (map != NULL) {
896 		vsize_t vsize;
897 
898 		vsize = vm_map_max(map) - vm_map_min(map);
899 		vsize = (uintmax_t)vsize * va_pct / 100;
900 		if (t > vsize) {
901 			t = vsize;
902 		}
903 	}
904 	return t;
905 }
906