xref: /netbsd-src/sys/kern/init_main.c (revision f983e71d70cfccf7b3de601eb4d998b2d886ede4)
1 /*	$NetBSD: init_main.c,v 1.350 2008/04/21 00:13:46 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  * 3. All advertising materials mentioning features or use of this software
16  *    must display the following acknowledgement:
17  *	This product includes software developed by the NetBSD
18  *	Foundation, Inc. and its contributors.
19  * 4. Neither the name of The NetBSD Foundation nor the names of its
20  *    contributors may be used to endorse or promote products derived
21  *    from this software without specific prior written permission.
22  *
23  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
24  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
25  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
26  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
27  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
28  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
29  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
30  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
31  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
32  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
33  * POSSIBILITY OF SUCH DAMAGE.
34  */
35 
36 /*
37  * Copyright (c) 1982, 1986, 1989, 1991, 1992, 1993
38  *	The Regents of the University of California.  All rights reserved.
39  * (c) UNIX System Laboratories, Inc.
40  * All or some portions of this file are derived from material licensed
41  * to the University of California by American Telephone and Telegraph
42  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
43  * the permission of UNIX System Laboratories, Inc.
44  *
45  * Redistribution and use in source and binary forms, with or without
46  * modification, are permitted provided that the following conditions
47  * are met:
48  * 1. Redistributions of source code must retain the above copyright
49  *    notice, this list of conditions and the following disclaimer.
50  * 2. Redistributions in binary form must reproduce the above copyright
51  *    notice, this list of conditions and the following disclaimer in the
52  *    documentation and/or other materials provided with the distribution.
53  * 3. Neither the name of the University nor the names of its contributors
54  *    may be used to endorse or promote products derived from this software
55  *    without specific prior written permission.
56  *
57  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
58  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
59  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
60  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
61  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
62  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
63  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
64  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
65  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
66  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
67  * SUCH DAMAGE.
68  *
69  *	@(#)init_main.c	8.16 (Berkeley) 5/14/95
70  */
71 
72 /*
73  * Copyright (c) 1995 Christopher G. Demetriou.  All rights reserved.
74  *
75  * Redistribution and use in source and binary forms, with or without
76  * modification, are permitted provided that the following conditions
77  * are met:
78  * 1. Redistributions of source code must retain the above copyright
79  *    notice, this list of conditions and the following disclaimer.
80  * 2. Redistributions in binary form must reproduce the above copyright
81  *    notice, this list of conditions and the following disclaimer in the
82  *    documentation and/or other materials provided with the distribution.
83  * 3. All advertising materials mentioning features or use of this software
84  *    must display the following acknowledgement:
85  *	This product includes software developed by the University of
86  *	California, Berkeley and its contributors.
87  * 4. Neither the name of the University nor the names of its contributors
88  *    may be used to endorse or promote products derived from this software
89  *    without specific prior written permission.
90  *
91  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
92  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
93  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
94  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
95  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
96  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
97  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
98  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
99  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
100  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
101  * SUCH DAMAGE.
102  *
103  *	@(#)init_main.c	8.16 (Berkeley) 5/14/95
104  */
105 
106 #include <sys/cdefs.h>
107 __KERNEL_RCSID(0, "$NetBSD: init_main.c,v 1.350 2008/04/21 00:13:46 ad Exp $");
108 
109 #include "opt_ipsec.h"
110 #include "opt_ntp.h"
111 #include "opt_pipe.h"
112 #include "opt_posix.h"
113 #include "opt_syscall_debug.h"
114 #include "opt_sysv.h"
115 #include "opt_fileassoc.h"
116 #include "opt_ktrace.h"
117 #include "opt_pax.h"
118 
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/kernel.h>
134 #include <sys/kmem.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/user.h>
156 #include <sys/sysctl.h>
157 #include <sys/event.h>
158 #include <sys/mbuf.h>
159 #include <sys/sched.h>
160 #include <sys/sleepq.h>
161 #include <sys/iostat.h>
162 #include <sys/vmem.h>
163 #include <sys/uuid.h>
164 #include <sys/extent.h>
165 #include <sys/disk.h>
166 #include <sys/mqueue.h>
167 #include <sys/msgbuf.h>
168 #include <sys/module.h>
169 #include <sys/event.h>
170 #ifdef FAST_IPSEC
171 #include <netipsec/ipsec.h>
172 #endif
173 #ifdef SYSVSHM
174 #include <sys/shm.h>
175 #endif
176 #ifdef SYSVSEM
177 #include <sys/sem.h>
178 #endif
179 #ifdef SYSVMSG
180 #include <sys/msg.h>
181 #endif
182 #ifdef P1003_1B_SEMAPHORE
183 #include <sys/ksem.h>
184 #endif
185 #include <sys/domain.h>
186 #include <sys/namei.h>
187 #if NRND > 0
188 #include <sys/rnd.h>
189 #endif
190 #include <sys/pipe.h>
191 #ifdef LKM
192 #include <sys/lkm.h>
193 #endif
194 #if NVERIEXEC > 0
195 #include <sys/verified_exec.h>
196 #endif /* NVERIEXEC > 0 */
197 #ifdef KTRACE
198 #include <sys/ktrace.h>
199 #endif
200 #include <sys/kauth.h>
201 #include <net80211/ieee80211_netbsd.h>
202 
203 #include <sys/syscall.h>
204 #include <sys/syscallargs.h>
205 
206 #if defined(PAX_MPROTECT) || defined(PAX_SEGVGUARD) || defined(PAX_ASLR)
207 #include <sys/pax.h>
208 #endif /* PAX_MPROTECT || PAX_SEGVGUARD || PAX_ASLR */
209 
210 #include <ufs/ufs/quota.h>
211 
212 #include <miscfs/genfs/genfs.h>
213 #include <miscfs/syncfs/syncfs.h>
214 
215 #include <sys/cpu.h>
216 
217 #include <uvm/uvm.h>
218 
219 #if NSYSMON_TASKQ > 0
220 #include <dev/sysmon/sysmon_taskq.h>
221 #endif
222 
223 #include <dev/cons.h>
224 
225 #if NSYSMON_ENVSYS > 0 || NSYSMON_POWER > 0 || NSYSMON_WDOG > 0
226 #include <dev/sysmon/sysmonvar.h>
227 #endif
228 
229 #include <net/if.h>
230 #include <net/raw_cb.h>
231 
232 #include <secmodel/secmodel.h>
233 
234 extern struct proc proc0;
235 extern struct lwp lwp0;
236 extern struct cwdinfo cwdi0;
237 extern time_t rootfstime;
238 
239 #ifndef curlwp
240 struct	lwp *curlwp = &lwp0;
241 #endif
242 struct	proc *initproc;
243 
244 struct	vnode *rootvp, *swapdev_vp;
245 int	boothowto;
246 int	cold = 1;			/* still working on startup */
247 struct timeval boottime;	        /* time at system startup - will only follow settime deltas */
248 
249 volatile int start_init_exec;		/* semaphore for start_init() */
250 
251 static void check_console(struct lwp *l);
252 static void start_init(void *);
253 void main(void);
254 void ssp_init(void);
255 
256 #if defined(__SSP__) || defined(__SSP_ALL__)
257 long __stack_chk_guard[8] = {0, 0, 0, 0, 0, 0, 0, 0};
258 void __stack_chk_fail(void);
259 
260 void
261 __stack_chk_fail(void)
262 {
263 	panic("stack overflow detected; terminated");
264 }
265 
266 void
267 ssp_init(void)
268 {
269 	int s;
270 
271 #ifdef DIAGNOSTIC
272 	printf("Initializing SSP:");
273 #endif
274 	/*
275 	 * We initialize ssp here carefully:
276 	 *	1. after we got some entropy
277 	 *	2. without calling a function
278 	 */
279 	size_t i;
280 	long guard[__arraycount(__stack_chk_guard)];
281 
282 	arc4randbytes(guard, sizeof(guard));
283 	s = splhigh();
284 	for (i = 0; i < __arraycount(guard); i++)
285 		__stack_chk_guard[i] = guard[i];
286 	splx(s);
287 #ifdef DIAGNOSTIC
288 	for (i = 0; i < __arraycount(guard); i++)
289 		printf("%lx ", guard[i]);
290 	printf("\n");
291 #endif
292 }
293 #else
294 void
295 ssp_init(void)
296 {
297 
298 }
299 #endif
300 
301 void __secmodel_none(void);
302 __weak_alias(secmodel_start,__secmodel_none);
303 void
304 __secmodel_none(void)
305 {
306 	return;
307 }
308 
309 /*
310  * System startup; initialize the world, create process 0, mount root
311  * filesystem, and fork to create init and pagedaemon.  Most of the
312  * hard work is done in the lower-level initialization routines including
313  * startup(), which does memory initialization and autoconfiguration.
314  */
315 void
316 main(void)
317 {
318 	struct timeval time;
319 	struct lwp *l;
320 	struct proc *p;
321 	int s, error;
322 #ifdef NVNODE_IMPLICIT
323 	int usevnodes;
324 #endif
325 	CPU_INFO_ITERATOR cii;
326 	struct cpu_info *ci;
327 
328 	l = &lwp0;
329 #ifndef LWP0_CPU_INFO
330 	l->l_cpu = curcpu();
331 #endif
332 
333 	/*
334 	 * Attempt to find console and initialize
335 	 * in case of early panic or other messages.
336 	 */
337 	consinit();
338 
339 	kernel_lock_init();
340 
341 	uvm_init();
342 
343 	kmem_init();
344 
345 	percpu_init();
346 
347 	/* Initialize lock caches. */
348 	mutex_obj_init();
349 
350 	/* Initialize the extent manager. */
351 	extent_init();
352 
353 	/* Do machine-dependent initialization. */
354 	cpu_startup();
355 
356 	/* Start module system. */
357 	module_init();
358 
359 	/* Initialize callouts, part 1. */
360 	callout_startup();
361 
362 	/*
363 	 * Initialize the kernel authorization subsystem and start the
364 	 * default security model, if any. We need to do this early
365 	 * enough so that subsystems relying on any of the aforementioned
366 	 * can work properly. Since the security model may dictate the
367 	 * credential inheritance policy, it is needed at least before
368 	 * any process is created, specifically proc0.
369 	 */
370 	kauth_init();
371 	secmodel_start();
372 
373 	/* Initialize the buffer cache */
374 	bufinit();
375 
376 	/* Initialize sockets. */
377 	soinit();
378 
379 	/*
380 	 * The following things must be done before autoconfiguration.
381 	 */
382 	evcnt_init();		/* initialize event counters */
383 #if NRND > 0
384 	rnd_init();		/* initialize RNG */
385 #endif
386 
387 	/* Initialize process and pgrp structures. */
388 	procinit();
389 	lwpinit();
390 
391 	/* Initialize signal-related data structures. */
392 	signal_init();
393 
394 	/* Initialize resource management. */
395 	resource_init();
396 
397 	/* Create process 0 (the swapper). */
398 	proc0_init();
399 
400 	/* Initialize the UID hash table. */
401 	uid_init();
402 
403 	/* Charge root for one process. */
404 	(void)chgproccnt(0, 1);
405 
406 	/* Initialize timekeeping. */
407 	time_init();
408 
409 	/* Initialize the run queues, turnstiles and sleep queues. */
410 	mutex_init(&cpu_lock, MUTEX_DEFAULT, IPL_NONE);
411 	sched_rqinit();
412 	turnstile_init();
413 	sleeptab_init(&sleeptab);
414 
415 	/* Initialize processor-sets */
416 	psets_init();
417 
418 	/* MI initialization of the boot cpu */
419 	error = mi_cpu_attach(curcpu());
420 	KASSERT(error == 0);
421 
422 	/* Initialize timekeeping, part 2. */
423 	time_init2();
424 
425 	/*
426 	 * Initialize mbuf's.  Do this now because we might attempt to
427 	 * allocate mbufs or mbuf clusters during autoconfiguration.
428 	 */
429 	mbinit();
430 
431 	/* Initialize the sysctl subsystem. */
432 	sysctl_init();
433 
434 	/* Initialize I/O statistics. */
435 	iostat_init();
436 
437 	/* Initialize the log device. */
438 	loginit();
439 
440 	/* Initialize the file systems. */
441 #ifdef NVNODE_IMPLICIT
442 	/*
443 	 * If maximum number of vnodes in namei vnode cache is not explicitly
444 	 * defined in kernel config, adjust the number such as we use roughly
445 	 * 1.0% of memory for vnode cache (but not less than NVNODE vnodes).
446 	 */
447 	usevnodes =
448 	    calc_cache_size(kernel_map, 1, VNODE_VA_MAXPCT) / sizeof(vnode_t);
449 	if (usevnodes > desiredvnodes)
450 		desiredvnodes = usevnodes;
451 #endif
452 	vfsinit();
453 
454 	/* Initialize fstrans. */
455 	fstrans_init();
456 
457 	/* Initialize the file descriptor system. */
458 	fd_sys_init();
459 
460 	/* Initialize kqueue. */
461 	kqueue_init();
462 
463 	/* Initialize asynchronous I/O. */
464 	aio_sysinit();
465 
466 	/* Initialize message queues. */
467 	mqueue_sysinit();
468 
469 	/* Initialize the system monitor subsystems. */
470 #if NSYSMON_TASKQ > 0
471 	sysmon_task_queue_preinit();
472 #endif
473 
474 #if NSYSMON_ENVSYS > 0
475 	sysmon_envsys_init();
476 #endif
477 
478 #if NSYSMON_POWER > 0
479 	sysmon_power_init();
480 #endif
481 
482 #if NSYSMON_WDOG > 0
483 	sysmon_wdog_init();
484 #endif
485 
486 	inittimecounter();
487 	ntp_init();
488 
489 	/* Initialize the device switch tables. */
490 	devsw_init();
491 
492 	/* Initialize tty subsystem. */
493 	tty_init();
494 	ttyldisc_init();
495 
496 	/* Initialize the buffer cache, part 2. */
497 	bufinit2();
498 
499 	/* Initialize the disk wedge subsystem. */
500 	dkwedge_init();
501 
502 	/* Configure the system hardware.  This will enable interrupts. */
503 	configure();
504 
505 	ubc_init();		/* must be after autoconfig */
506 
507 #ifdef SYSVSHM
508 	/* Initialize System V style shared memory. */
509 	shminit();
510 #endif
511 
512 #ifdef SYSVSEM
513 	/* Initialize System V style semaphores. */
514 	seminit();
515 #endif
516 
517 #ifdef SYSVMSG
518 	/* Initialize System V style message queues. */
519 	msginit();
520 #endif
521 
522 #ifdef P1003_1B_SEMAPHORE
523 	/* Initialize posix semaphores */
524 	ksem_init();
525 #endif
526 
527 #if NVERIEXEC > 0
528 	/*
529 	 * Initialise the Veriexec subsystem.
530 	 */
531 	veriexec_init();
532 #endif /* NVERIEXEC > 0 */
533 
534 #if defined(PAX_MPROTECT) || defined(PAX_SEGVGUARD) || defined(PAX_ASLR)
535 	pax_init();
536 #endif /* PAX_MPROTECT || PAX_SEGVGUARD || PAX_ASLR */
537 
538 #ifdef	FAST_IPSEC
539 	/* Attach network crypto subsystem */
540 	ipsec_attach();
541 #endif
542 
543 	/*
544 	 * Initialize protocols.  Block reception of incoming packets
545 	 * until everything is ready.
546 	 */
547 	s = splnet();
548 	ifinit();
549 	domaininit();
550 	if_attachdomain();
551 	splx(s);
552 
553 #ifdef GPROF
554 	/* Initialize kernel profiling. */
555 	kmstartup();
556 #endif
557 
558 	/* Initialize system accouting. */
559 	acct_init();
560 
561 #ifndef PIPE_SOCKETPAIR
562 	/* Initialize pipes. */
563 	pipe_init();
564 #endif
565 
566 #ifdef KTRACE
567 	/* Initialize ktrace. */
568 	ktrinit();
569 #endif
570 
571 	/* Initialize the UUID system calls. */
572 	uuid_init();
573 
574 	/*
575 	 * Create process 1 (init(8)).  We do this now, as Unix has
576 	 * historically had init be process 1, and changing this would
577 	 * probably upset a lot of people.
578 	 *
579 	 * Note that process 1 won't immediately exec init(8), but will
580 	 * wait for us to inform it that the root file system has been
581 	 * mounted.
582 	 */
583 	if (fork1(l, 0, SIGCHLD, NULL, 0, start_init, NULL, NULL, &initproc))
584 		panic("fork init");
585 
586 	/*
587 	 * Load any remaining builtin modules, and hand back temporary
588 	 * storage to the VM system.
589 	 */
590 	module_init_class(MODULE_CLASS_ANY);
591 	module_jettison();
592 
593 	/*
594 	 * Finalize configuration now that all real devices have been
595 	 * found.  This needs to be done before the root device is
596 	 * selected, since finalization may create the root device.
597 	 */
598 	config_finalize();
599 
600 	/*
601 	 * Now that autoconfiguration has completed, we can determine
602 	 * the root and dump devices.
603 	 */
604 	cpu_rootconf();
605 	cpu_dumpconf();
606 
607 	/* Mount the root file system. */
608 	do {
609 		domountroothook();
610 		if ((error = vfs_mountroot())) {
611 			printf("cannot mount root, error = %d\n", error);
612 			boothowto |= RB_ASKNAME;
613 			setroot(root_device,
614 			    (rootdev != NODEV) ? DISKPART(rootdev) : 0);
615 		}
616 	} while (error != 0);
617 	mountroothook_destroy();
618 
619 	/*
620 	 * Initialise the time-of-day clock, passing the time recorded
621 	 * in the root filesystem (if any) for use by systems that
622 	 * don't have a non-volatile time-of-day device.
623 	 */
624 	inittodr(rootfstime);
625 
626 	CIRCLEQ_FIRST(&mountlist)->mnt_flag |= MNT_ROOTFS;
627 	CIRCLEQ_FIRST(&mountlist)->mnt_op->vfs_refcount++;
628 
629 	/*
630 	 * Get the vnode for '/'.  Set filedesc0.fd_fd.fd_cdir to
631 	 * reference it.
632 	 */
633 	error = VFS_ROOT(CIRCLEQ_FIRST(&mountlist), &rootvnode);
634 	if (error)
635 		panic("cannot find root vnode, error=%d", error);
636 	cwdi0.cwdi_cdir = rootvnode;
637 	VREF(cwdi0.cwdi_cdir);
638 	VOP_UNLOCK(rootvnode, 0);
639 	cwdi0.cwdi_rdir = NULL;
640 
641 	/*
642 	 * Now that root is mounted, we can fixup initproc's CWD
643 	 * info.  All other processes are kthreads, which merely
644 	 * share proc0's CWD info.
645 	 */
646 	initproc->p_cwdi->cwdi_cdir = rootvnode;
647 	VREF(initproc->p_cwdi->cwdi_cdir);
648 	initproc->p_cwdi->cwdi_rdir = NULL;
649 
650 	/*
651 	 * Now can look at time, having had a chance to verify the time
652 	 * from the file system.  Reset l->l_rtime as it may have been
653 	 * munched in mi_switch() after the time got set.
654 	 */
655 	getmicrotime(&time);
656 	boottime = time;
657 	mutex_enter(&proclist_lock);
658 	LIST_FOREACH(p, &allproc, p_list) {
659 		KASSERT((p->p_flag & PK_MARKER) == 0);
660 		mutex_enter(&p->p_smutex);
661 		p->p_stats->p_start = time;
662 		LIST_FOREACH(l, &p->p_lwps, l_sibling) {
663 			lwp_lock(l);
664 			memset(&l->l_rtime, 0, sizeof(l->l_rtime));
665 			lwp_unlock(l);
666 		}
667 		mutex_exit(&p->p_smutex);
668 	}
669 	mutex_exit(&proclist_lock);
670 	binuptime(&curlwp->l_stime);
671 
672 	for (CPU_INFO_FOREACH(cii, ci)) {
673 		ci->ci_schedstate.spc_lastmod = time_second;
674 	}
675 
676 	/* Create the pageout daemon kernel thread. */
677 	uvm_swap_init();
678 	if (kthread_create(PRI_PGDAEMON, KTHREAD_MPSAFE, NULL, uvm_pageout,
679 	    NULL, NULL, "pgdaemon"))
680 		panic("fork pagedaemon");
681 
682 	/* Create the filesystem syncer kernel thread. */
683 	if (kthread_create(PRI_IOFLUSH, KTHREAD_MPSAFE, NULL, sched_sync,
684 	    NULL, NULL, "ioflush"))
685 		panic("fork syncer");
686 
687 	/* Create the aiodone daemon kernel thread. */
688 	if (workqueue_create(&uvm.aiodone_queue, "aiodoned",
689 	    uvm_aiodone_worker, NULL, PRI_VM, IPL_NONE, WQ_MPSAFE))
690 		panic("fork aiodoned");
691 
692 	vmem_rehash_start();
693 
694 	/* Initialize exec structures */
695 	exec_init(1);
696 
697 	/*
698 	 * Okay, now we can let init(8) exec!  It's off to userland!
699 	 */
700 	start_init_exec = 1;
701 	wakeup(&start_init_exec);
702 
703 	/* The scheduler is an infinite loop. */
704 	uvm_scheduler();
705 	/* NOTREACHED */
706 }
707 
708 static void
709 check_console(struct lwp *l)
710 {
711 	struct nameidata nd;
712 	int error;
713 
714 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_SYSSPACE, "/dev/console");
715 	error = namei(&nd);
716 	if (error == 0)
717 		vrele(nd.ni_vp);
718 	else if (error == ENOENT)
719 		printf("warning: no /dev/console\n");
720 	else
721 		printf("warning: lookup /dev/console: error %d\n", error);
722 }
723 
724 /*
725  * List of paths to try when searching for "init".
726  */
727 static const char * const initpaths[] = {
728 	"/sbin/init",
729 	"/sbin/oinit",
730 	"/sbin/init.bak",
731 	NULL,
732 };
733 
734 /*
735  * Start the initial user process; try exec'ing each pathname in "initpaths".
736  * The program is invoked with one argument containing the boot flags.
737  */
738 static void
739 start_init(void *arg)
740 {
741 	struct lwp *l = arg;
742 	struct proc *p = l->l_proc;
743 	vaddr_t addr;
744 	struct sys_execve_args /* {
745 		syscallarg(const char *) path;
746 		syscallarg(char * const *) argp;
747 		syscallarg(char * const *) envp;
748 	} */ args;
749 	int options, i, error;
750 	register_t retval[2];
751 	char flags[4], *flagsp;
752 	const char *path, *slash;
753 	char *ucp, **uap, *arg0, *arg1 = NULL;
754 	char ipath[129];
755 	int ipx, len;
756 
757 	/*
758 	 * Now in process 1.
759 	 */
760 	strncpy(p->p_comm, "init", MAXCOMLEN);
761 
762 	/*
763 	 * Wait for main() to tell us that it's safe to exec.
764 	 */
765 	while (start_init_exec == 0)
766 		(void) tsleep(&start_init_exec, PWAIT, "initexec", 0);
767 
768 	/*
769 	 * This is not the right way to do this.  We really should
770 	 * hand-craft a descriptor onto /dev/console to hand to init,
771 	 * but that's a _lot_ more work, and the benefit from this easy
772 	 * hack makes up for the "good is the enemy of the best" effect.
773 	 */
774 	check_console(l);
775 
776 	/*
777 	 * Need just enough stack to hold the faked-up "execve()" arguments.
778 	 */
779 	addr = (vaddr_t)STACK_ALLOC(USRSTACK, PAGE_SIZE);
780 	if (uvm_map(&p->p_vmspace->vm_map, &addr, PAGE_SIZE,
781                     NULL, UVM_UNKNOWN_OFFSET, 0,
782                     UVM_MAPFLAG(UVM_PROT_ALL, UVM_PROT_ALL, UVM_INH_COPY,
783 		    UVM_ADV_NORMAL,
784                     UVM_FLAG_FIXED|UVM_FLAG_OVERLAY|UVM_FLAG_COPYONW)) != 0)
785 		panic("init: couldn't allocate argument space");
786 	p->p_vmspace->vm_maxsaddr = (void *)STACK_MAX(addr, PAGE_SIZE);
787 
788 	ipx = 0;
789 	while (1) {
790 		if (boothowto & RB_ASKNAME) {
791 			printf("init path");
792 			if (initpaths[ipx])
793 				printf(" (default %s)", initpaths[ipx]);
794 			printf(": ");
795 			len = cngetsn(ipath, sizeof(ipath)-1);
796 			if (len == 0) {
797 				if (initpaths[ipx])
798 					path = initpaths[ipx++];
799 				else
800 					continue;
801 			} else {
802 				ipath[len] = '\0';
803 				path = ipath;
804 			}
805 		} else {
806 			if ((path = initpaths[ipx++]) == NULL)
807 				break;
808 		}
809 
810 		ucp = (char *)USRSTACK;
811 
812 		/*
813 		 * Construct the boot flag argument.
814 		 */
815 		flagsp = flags;
816 		*flagsp++ = '-';
817 		options = 0;
818 
819 		if (boothowto & RB_SINGLE) {
820 			*flagsp++ = 's';
821 			options = 1;
822 		}
823 #ifdef notyet
824 		if (boothowto & RB_FASTBOOT) {
825 			*flagsp++ = 'f';
826 			options = 1;
827 		}
828 #endif
829 
830 		/*
831 		 * Move out the flags (arg 1), if necessary.
832 		 */
833 		if (options != 0) {
834 			*flagsp++ = '\0';
835 			i = flagsp - flags;
836 #ifdef DEBUG
837 			printf("init: copying out flags `%s' %d\n", flags, i);
838 #endif
839 			arg1 = STACK_ALLOC(ucp, i);
840 			ucp = STACK_MAX(arg1, i);
841 			(void)copyout((void *)flags, arg1, i);
842 		}
843 
844 		/*
845 		 * Move out the file name (also arg 0).
846 		 */
847 		i = strlen(path) + 1;
848 #ifdef DEBUG
849 		printf("init: copying out path `%s' %d\n", path, i);
850 #else
851 		if (boothowto & RB_ASKNAME || path != initpaths[0])
852 			printf("init: trying %s\n", path);
853 #endif
854 		arg0 = STACK_ALLOC(ucp, i);
855 		ucp = STACK_MAX(arg0, i);
856 		(void)copyout(path, arg0, i);
857 
858 		/*
859 		 * Move out the arg pointers.
860 		 */
861 		ucp = (void *)STACK_ALIGN(ucp, ALIGNBYTES);
862 		uap = (char **)STACK_ALLOC(ucp, sizeof(char *) * 3);
863 		SCARG(&args, path) = arg0;
864 		SCARG(&args, argp) = uap;
865 		SCARG(&args, envp) = NULL;
866 		slash = strrchr(path, '/');
867 		if (slash)
868 			(void)suword((void *)uap++,
869 			    (long)arg0 + (slash + 1 - path));
870 		else
871 			(void)suword((void *)uap++, (long)arg0);
872 		if (options != 0)
873 			(void)suword((void *)uap++, (long)arg1);
874 		(void)suword((void *)uap++, 0);	/* terminator */
875 
876 		/*
877 		 * Now try to exec the program.  If can't for any reason
878 		 * other than it doesn't exist, complain.
879 		 */
880 		error = sys_execve(l, &args, retval);
881 		if (error == 0 || error == EJUSTRETURN) {
882 			KERNEL_UNLOCK_LAST(l);
883 			return;
884 		}
885 		printf("exec %s: error %d\n", path, error);
886 	}
887 	printf("init: not found\n");
888 	panic("no init");
889 }
890 
891 /*
892  * calculate cache size from physmem and vm_map size.
893  */
894 vaddr_t
895 calc_cache_size(struct vm_map *map, int pct, int va_pct)
896 {
897 	paddr_t t;
898 
899 	/* XXX should consider competing cache if any */
900 	/* XXX should consider submaps */
901 	t = (uintmax_t)physmem * pct / 100 * PAGE_SIZE;
902 	if (map != NULL) {
903 		vsize_t vsize;
904 
905 		vsize = vm_map_max(map) - vm_map_min(map);
906 		vsize = (uintmax_t)vsize * va_pct / 100;
907 		if (t > vsize) {
908 			t = vsize;
909 		}
910 	}
911 	return t;
912 }
913