xref: /netbsd-src/sys/kern/init_main.c (revision 56a34939419542e88b386b2229be7565f4f45461)
1 /*	$NetBSD: init_main.c,v 1.378 2008/12/07 20:58:46 pooka 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 1.378 2008/12/07 20:58:46 pooka Exp $");
101 
102 #include "opt_ddb.h"
103 #include "opt_ipsec.h"
104 #include "opt_ntp.h"
105 #include "opt_pipe.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 #include "opt_wapbl.h"
112 
113 #include "ksyms.h"
114 #include "rnd.h"
115 #include "sysmon_envsys.h"
116 #include "sysmon_power.h"
117 #include "sysmon_taskq.h"
118 #include "sysmon_wdog.h"
119 #include "veriexec.h"
120 
121 #include <sys/param.h>
122 #include <sys/acct.h>
123 #include <sys/filedesc.h>
124 #include <sys/file.h>
125 #include <sys/errno.h>
126 #include <sys/callout.h>
127 #include <sys/cpu.h>
128 #include <sys/kernel.h>
129 #include <sys/mount.h>
130 #include <sys/proc.h>
131 #include <sys/kthread.h>
132 #include <sys/resourcevar.h>
133 #include <sys/signalvar.h>
134 #include <sys/systm.h>
135 #include <sys/vnode.h>
136 #include <sys/fstrans.h>
137 #include <sys/tty.h>
138 #include <sys/conf.h>
139 #include <sys/disklabel.h>
140 #include <sys/buf.h>
141 #include <sys/device.h>
142 #include <sys/exec.h>
143 #include <sys/socketvar.h>
144 #include <sys/protosw.h>
145 #include <sys/percpu.h>
146 #include <sys/pset.h>
147 #include <sys/sysctl.h>
148 #include <sys/reboot.h>
149 #include <sys/user.h>
150 #include <sys/sysctl.h>
151 #include <sys/event.h>
152 #include <sys/mbuf.h>
153 #include <sys/sched.h>
154 #include <sys/sleepq.h>
155 #include <sys/iostat.h>
156 #include <sys/vmem.h>
157 #include <sys/uuid.h>
158 #include <sys/extent.h>
159 #include <sys/disk.h>
160 #include <sys/mqueue.h>
161 #include <sys/msgbuf.h>
162 #include <sys/module.h>
163 #include <sys/event.h>
164 #include <sys/lockf.h>
165 #include <sys/once.h>
166 #include <sys/ksyms.h>
167 #include <sys/uidinfo.h>
168 #ifdef FAST_IPSEC
169 #include <netipsec/ipsec.h>
170 #endif
171 #ifdef SYSVSHM
172 #include <sys/shm.h>
173 #endif
174 #ifdef SYSVSEM
175 #include <sys/sem.h>
176 #endif
177 #ifdef SYSVMSG
178 #include <sys/msg.h>
179 #endif
180 #include <sys/domain.h>
181 #include <sys/namei.h>
182 #if NRND > 0
183 #include <sys/rnd.h>
184 #endif
185 #include <sys/pipe.h>
186 #if NVERIEXEC > 0
187 #include <sys/verified_exec.h>
188 #endif /* NVERIEXEC > 0 */
189 #ifdef KTRACE
190 #include <sys/ktrace.h>
191 #endif
192 #include <sys/kauth.h>
193 #ifdef WAPBL
194 #include <sys/wapbl.h>
195 #endif
196 #include <net80211/ieee80211_netbsd.h>
197 
198 #include <sys/syscall.h>
199 #include <sys/syscallargs.h>
200 
201 #if defined(PAX_MPROTECT) || defined(PAX_SEGVGUARD) || defined(PAX_ASLR)
202 #include <sys/pax.h>
203 #endif /* PAX_MPROTECT || PAX_SEGVGUARD || PAX_ASLR */
204 
205 #include <ufs/ufs/quota.h>
206 
207 #include <miscfs/genfs/genfs.h>
208 #include <miscfs/syncfs/syncfs.h>
209 
210 #include <sys/cpu.h>
211 
212 #include <uvm/uvm.h>
213 
214 #if NSYSMON_TASKQ > 0
215 #include <dev/sysmon/sysmon_taskq.h>
216 #endif
217 
218 #include <dev/cons.h>
219 
220 #if NSYSMON_ENVSYS > 0 || NSYSMON_POWER > 0 || NSYSMON_WDOG > 0
221 #include <dev/sysmon/sysmonvar.h>
222 #endif
223 
224 #include <net/if.h>
225 #include <net/raw_cb.h>
226 
227 #include <secmodel/secmodel.h>
228 
229 extern struct proc proc0;
230 extern struct lwp lwp0;
231 extern struct cwdinfo cwdi0;
232 extern time_t rootfstime;
233 
234 #ifndef curlwp
235 struct	lwp *curlwp = &lwp0;
236 #endif
237 struct	proc *initproc;
238 
239 struct	vnode *rootvp, *swapdev_vp;
240 int	boothowto;
241 int	cold = 1;			/* still working on startup */
242 struct timeval boottime;	        /* time at system startup - will only follow settime deltas */
243 
244 int	start_init_exec;		/* semaphore for start_init() */
245 
246 static void check_console(struct lwp *l);
247 static void start_init(void *);
248 void main(void);
249 void ssp_init(void);
250 
251 #if defined(__SSP__) || defined(__SSP_ALL__)
252 long __stack_chk_guard[8] = {0, 0, 0, 0, 0, 0, 0, 0};
253 void __stack_chk_fail(void);
254 
255 void
256 __stack_chk_fail(void)
257 {
258 	panic("stack overflow detected; terminated");
259 }
260 
261 void
262 ssp_init(void)
263 {
264 	int s;
265 
266 #ifdef DIAGNOSTIC
267 	printf("Initializing SSP:");
268 #endif
269 	/*
270 	 * We initialize ssp here carefully:
271 	 *	1. after we got some entropy
272 	 *	2. without calling a function
273 	 */
274 	size_t i;
275 	long guard[__arraycount(__stack_chk_guard)];
276 
277 	arc4randbytes(guard, sizeof(guard));
278 	s = splhigh();
279 	for (i = 0; i < __arraycount(guard); i++)
280 		__stack_chk_guard[i] = guard[i];
281 	splx(s);
282 #ifdef DIAGNOSTIC
283 	for (i = 0; i < __arraycount(guard); i++)
284 		printf("%lx ", guard[i]);
285 	printf("\n");
286 #endif
287 }
288 #else
289 void
290 ssp_init(void)
291 {
292 
293 }
294 #endif
295 
296 void __secmodel_none(void);
297 __weak_alias(secmodel_start,__secmodel_none);
298 void
299 __secmodel_none(void)
300 {
301 	return;
302 }
303 
304 /*
305  * System startup; initialize the world, create process 0, mount root
306  * filesystem, and fork to create init and pagedaemon.  Most of the
307  * hard work is done in the lower-level initialization routines including
308  * startup(), which does memory initialization and autoconfiguration.
309  */
310 void
311 main(void)
312 {
313 	struct timeval time;
314 	struct lwp *l;
315 	struct proc *p;
316 	int s, error;
317 #ifdef NVNODE_IMPLICIT
318 	int usevnodes;
319 #endif
320 	CPU_INFO_ITERATOR cii;
321 	struct cpu_info *ci;
322 
323 	l = &lwp0;
324 #ifndef LWP0_CPU_INFO
325 	l->l_cpu = curcpu();
326 #endif
327 
328 	/*
329 	 * Attempt to find console and initialize
330 	 * in case of early panic or other messages.
331 	 */
332 	consinit();
333 
334 	kernel_lock_init();
335 	once_init();
336 
337 	uvm_init();
338 
339 #if ((NKSYMS > 0) || (NDDB > 0) || (NMODULAR > 0))
340 	ksyms_init();
341 #endif
342 
343 	percpu_init();
344 
345 	/* Initialize lock caches. */
346 	mutex_obj_init();
347 
348 	/* Initialize the extent manager. */
349 	extent_init();
350 
351 	/* Do machine-dependent initialization. */
352 	cpu_startup();
353 
354 	/* Initialize the sysctl subsystem. */
355 	sysctl_init();
356 
357 	/* Initialize callouts, part 1. */
358 	callout_startup();
359 
360 	/*
361 	 * Initialize the kernel authorization subsystem and start the
362 	 * default security model, if any. We need to do this early
363 	 * enough so that subsystems relying on any of the aforementioned
364 	 * can work properly. Since the security model may dictate the
365 	 * credential inheritance policy, it is needed at least before
366 	 * any process is created, specifically proc0.
367 	 */
368 	kauth_init();
369 	secmodel_start();
370 
371 	/* Initialize the buffer cache */
372 	bufinit();
373 
374 	/* Initialize sockets. */
375 	soinit();
376 
377 	/*
378 	 * The following things must be done before autoconfiguration.
379 	 */
380 	evcnt_init();		/* initialize event counters */
381 #if NRND > 0
382 	rnd_init();		/* initialize random number generator */
383 #endif
384 
385 	/* Initialize process and pgrp structures. */
386 	procinit();
387 	lwpinit();
388 
389 	/* Initialize signal-related data structures. */
390 	signal_init();
391 
392 	/* Initialize resource management. */
393 	resource_init();
394 
395 	/* Create process 0 (the swapper). */
396 	proc0_init();
397 
398 	/* Initialize the UID hash table. */
399 	uid_init();
400 
401 	/* Charge root for one process. */
402 	(void)chgproccnt(0, 1);
403 
404 	/* Initialize timekeeping. */
405 	time_init();
406 
407 	/* Initialize the run queues, turnstiles and sleep queues. */
408 	mutex_init(&cpu_lock, MUTEX_DEFAULT, IPL_NONE);
409 	sched_rqinit();
410 	turnstile_init();
411 	sleeptab_init(&sleeptab);
412 
413 	/* Initialize processor-sets */
414 	psets_init();
415 
416 	/* MI initialization of the boot cpu */
417 	error = mi_cpu_attach(curcpu());
418 	KASSERT(error == 0);
419 
420 	/* Initialize timekeeping, part 2. */
421 	time_init2();
422 
423 	/*
424 	 * Initialize mbuf's.  Do this now because we might attempt to
425 	 * allocate mbufs or mbuf clusters during autoconfiguration.
426 	 */
427 	mbinit();
428 
429 	/* Initialize I/O statistics. */
430 	iostat_init();
431 
432 	/* Initialize the log device. */
433 	loginit();
434 
435 	/* Start module system. */
436 	module_init();
437 
438 	/* Initialize the file systems. */
439 #ifdef NVNODE_IMPLICIT
440 	/*
441 	 * If maximum number of vnodes in namei vnode cache is not explicitly
442 	 * defined in kernel config, adjust the number such as we use roughly
443 	 * 10% of memory for vnodes and associated data structures in the
444 	 * assumed worst case.  Do not provide fewer than NVNODE vnodes.
445 	 */
446 	usevnodes =
447 	    calc_cache_size(kernel_map, 10, VNODE_VA_MAXPCT) / VNODE_COST;
448 	if (usevnodes > desiredvnodes)
449 		desiredvnodes = usevnodes;
450 #endif
451 	vfsinit();
452 	lf_init();
453 
454 	/* Initialize fstrans. */
455 	fstrans_init();
456 
457 	/* Initialize the file descriptor system. */
458 	fd_sys_init();
459 
460 	/* Initialize cwd structures */
461 	cwd_sys_init();
462 
463 	/* Initialize kqueue. */
464 	kqueue_init();
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 	/* Initialize interfaces. */
503 	ifinit1();
504 
505 	/* Configure the system hardware.  This will enable interrupts. */
506 	configure();
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 #ifdef SYSVSEM
516 	/* Initialize System V style semaphores. */
517 	seminit();
518 #endif
519 
520 #ifdef SYSVMSG
521 	/* Initialize System V style message queues. */
522 	msginit();
523 #endif
524 
525 #if NVERIEXEC > 0
526 	/*
527 	 * Initialise the Veriexec subsystem.
528 	 */
529 	veriexec_init();
530 #endif /* NVERIEXEC > 0 */
531 
532 #if defined(PAX_MPROTECT) || defined(PAX_SEGVGUARD) || defined(PAX_ASLR)
533 	pax_init();
534 #endif /* PAX_MPROTECT || PAX_SEGVGUARD || PAX_ASLR */
535 
536 #ifdef	FAST_IPSEC
537 	/* Attach network crypto subsystem */
538 	ipsec_attach();
539 #endif
540 
541 	/*
542 	 * Initialize protocols.  Block reception of incoming packets
543 	 * until everything is ready.
544 	 */
545 	s = splnet();
546 	ifinit();
547 	domaininit();
548 	if_attachdomain();
549 	splx(s);
550 
551 #ifdef GPROF
552 	/* Initialize kernel profiling. */
553 	kmstartup();
554 #endif
555 
556 	/* Initialize system accounting. */
557 	acct_init();
558 
559 #ifndef PIPE_SOCKETPAIR
560 	/* Initialize pipes. */
561 	pipe_init();
562 #endif
563 
564 #ifdef KTRACE
565 	/* Initialize ktrace. */
566 	ktrinit();
567 #endif
568 
569 	/* Initialize the UUID system calls. */
570 	uuid_init();
571 
572 #ifdef WAPBL
573 	/* Initialize write-ahead physical block logging. */
574 	wapbl_init();
575 #endif
576 
577 	/*
578 	 * Create process 1 (init(8)).  We do this now, as Unix has
579 	 * historically had init be process 1, and changing this would
580 	 * probably upset a lot of people.
581 	 *
582 	 * Note that process 1 won't immediately exec init(8), but will
583 	 * wait for us to inform it that the root file system has been
584 	 * mounted.
585 	 */
586 	if (fork1(l, 0, SIGCHLD, NULL, 0, start_init, NULL, NULL, &initproc))
587 		panic("fork init");
588 
589 	/*
590 	 * Load any remaining builtin modules, and hand back temporary
591 	 * storage to the VM system.
592 	 */
593 	module_init_class(MODULE_CLASS_ANY);
594 
595 	/*
596 	 * Finalize configuration now that all real devices have been
597 	 * found.  This needs to be done before the root device is
598 	 * selected, since finalization may create the root device.
599 	 */
600 	config_finalize();
601 
602 	/*
603 	 * Now that autoconfiguration has completed, we can determine
604 	 * the root and dump devices.
605 	 */
606 	cpu_rootconf();
607 	cpu_dumpconf();
608 
609 	/* Mount the root file system. */
610 	do {
611 		domountroothook();
612 		if ((error = vfs_mountroot())) {
613 			printf("cannot mount root, error = %d\n", error);
614 			boothowto |= RB_ASKNAME;
615 			setroot(root_device,
616 			    (rootdev != NODEV) ? DISKPART(rootdev) : 0);
617 		}
618 	} while (error != 0);
619 	mountroothook_destroy();
620 
621 	/*
622 	 * Initialise the time-of-day clock, passing the time recorded
623 	 * in the root filesystem (if any) for use by systems that
624 	 * don't have a non-volatile time-of-day device.
625 	 */
626 	inittodr(rootfstime);
627 
628 	CIRCLEQ_FIRST(&mountlist)->mnt_flag |= MNT_ROOTFS;
629 	CIRCLEQ_FIRST(&mountlist)->mnt_op->vfs_refcount++;
630 
631 	/*
632 	 * Get the vnode for '/'.  Set filedesc0.fd_fd.fd_cdir to
633 	 * reference it.
634 	 */
635 	error = VFS_ROOT(CIRCLEQ_FIRST(&mountlist), &rootvnode);
636 	if (error)
637 		panic("cannot find root vnode, error=%d", error);
638 	cwdi0.cwdi_cdir = rootvnode;
639 	VREF(cwdi0.cwdi_cdir);
640 	VOP_UNLOCK(rootvnode, 0);
641 	cwdi0.cwdi_rdir = NULL;
642 
643 	/*
644 	 * Now that root is mounted, we can fixup initproc's CWD
645 	 * info.  All other processes are kthreads, which merely
646 	 * share proc0's CWD info.
647 	 */
648 	initproc->p_cwdi->cwdi_cdir = rootvnode;
649 	VREF(initproc->p_cwdi->cwdi_cdir);
650 	initproc->p_cwdi->cwdi_rdir = NULL;
651 
652 	/*
653 	 * Now can look at time, having had a chance to verify the time
654 	 * from the file system.  Reset l->l_rtime as it may have been
655 	 * munched in mi_switch() after the time got set.
656 	 */
657 	getmicrotime(&time);
658 	boottime = time;
659 	mutex_enter(proc_lock);
660 	LIST_FOREACH(p, &allproc, p_list) {
661 		KASSERT((p->p_flag & PK_MARKER) == 0);
662 		mutex_enter(p->p_lock);
663 		p->p_stats->p_start = time;
664 		LIST_FOREACH(l, &p->p_lwps, l_sibling) {
665 			lwp_lock(l);
666 			memset(&l->l_rtime, 0, sizeof(l->l_rtime));
667 			lwp_unlock(l);
668 		}
669 		mutex_exit(p->p_lock);
670 	}
671 	mutex_exit(proc_lock);
672 	binuptime(&curlwp->l_stime);
673 
674 	for (CPU_INFO_FOREACH(cii, ci)) {
675 		ci->ci_schedstate.spc_lastmod = time_second;
676 	}
677 
678 	/* Create the pageout daemon kernel thread. */
679 	uvm_swap_init();
680 	if (kthread_create(PRI_PGDAEMON, KTHREAD_MPSAFE, NULL, uvm_pageout,
681 	    NULL, NULL, "pgdaemon"))
682 		panic("fork pagedaemon");
683 
684 	/* Create the filesystem syncer kernel thread. */
685 	if (kthread_create(PRI_IOFLUSH, KTHREAD_MPSAFE, NULL, sched_sync,
686 	    NULL, NULL, "ioflush"))
687 		panic("fork syncer");
688 
689 	/* Create the aiodone daemon kernel thread. */
690 	if (workqueue_create(&uvm.aiodone_queue, "aiodoned",
691 	    uvm_aiodone_worker, NULL, PRI_VM, IPL_NONE, WQ_MPSAFE))
692 		panic("fork aiodoned");
693 
694 	vmem_rehash_start();
695 
696 	/* Initialize exec structures */
697 	exec_init(1);
698 
699 	/*
700 	 * Okay, now we can let init(8) exec!  It's off to userland!
701 	 */
702 	mutex_enter(proc_lock);
703 	start_init_exec = 1;
704 	cv_broadcast(&lbolt);
705 	mutex_exit(proc_lock);
706 
707 	/* The scheduler is an infinite loop. */
708 	uvm_scheduler();
709 	/* NOTREACHED */
710 }
711 
712 static void
713 check_console(struct lwp *l)
714 {
715 	struct nameidata nd;
716 	int error;
717 
718 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_SYSSPACE, "/dev/console");
719 	error = namei(&nd);
720 	if (error == 0)
721 		vrele(nd.ni_vp);
722 	else if (error == ENOENT)
723 		printf("warning: no /dev/console\n");
724 	else
725 		printf("warning: lookup /dev/console: error %d\n", error);
726 }
727 
728 /*
729  * List of paths to try when searching for "init".
730  */
731 static const char * const initpaths[] = {
732 	"/sbin/init",
733 	"/sbin/oinit",
734 	"/sbin/init.bak",
735 	NULL,
736 };
737 
738 /*
739  * Start the initial user process; try exec'ing each pathname in "initpaths".
740  * The program is invoked with one argument containing the boot flags.
741  */
742 static void
743 start_init(void *arg)
744 {
745 	struct lwp *l = arg;
746 	struct proc *p = l->l_proc;
747 	vaddr_t addr;
748 	struct sys_execve_args /* {
749 		syscallarg(const char *) path;
750 		syscallarg(char * const *) argp;
751 		syscallarg(char * const *) envp;
752 	} */ args;
753 	int options, i, error;
754 	register_t retval[2];
755 	char flags[4], *flagsp;
756 	const char *path, *slash;
757 	char *ucp, **uap, *arg0, *arg1 = NULL;
758 	char ipath[129];
759 	int ipx, len;
760 
761 	/*
762 	 * Now in process 1.
763 	 */
764 	strncpy(p->p_comm, "init", MAXCOMLEN);
765 
766 	/*
767 	 * Wait for main() to tell us that it's safe to exec.
768 	 */
769 	mutex_enter(proc_lock);
770 	while (start_init_exec == 0)
771 		cv_wait(&lbolt, proc_lock);
772 	mutex_exit(proc_lock);
773 
774 	/*
775 	 * This is not the right way to do this.  We really should
776 	 * hand-craft a descriptor onto /dev/console to hand to init,
777 	 * but that's a _lot_ more work, and the benefit from this easy
778 	 * hack makes up for the "good is the enemy of the best" effect.
779 	 */
780 	check_console(l);
781 
782 	/*
783 	 * Need just enough stack to hold the faked-up "execve()" arguments.
784 	 */
785 	addr = (vaddr_t)STACK_ALLOC(USRSTACK, PAGE_SIZE);
786 	if (uvm_map(&p->p_vmspace->vm_map, &addr, PAGE_SIZE,
787                     NULL, UVM_UNKNOWN_OFFSET, 0,
788                     UVM_MAPFLAG(UVM_PROT_ALL, UVM_PROT_ALL, UVM_INH_COPY,
789 		    UVM_ADV_NORMAL,
790                     UVM_FLAG_FIXED|UVM_FLAG_OVERLAY|UVM_FLAG_COPYONW)) != 0)
791 		panic("init: couldn't allocate argument space");
792 	p->p_vmspace->vm_maxsaddr = (void *)STACK_MAX(addr, PAGE_SIZE);
793 
794 	ipx = 0;
795 	while (1) {
796 		if (boothowto & RB_ASKNAME) {
797 			printf("init path");
798 			if (initpaths[ipx])
799 				printf(" (default %s)", initpaths[ipx]);
800 			printf(": ");
801 			len = cngetsn(ipath, sizeof(ipath)-1);
802 			if (len == 4 && strcmp(ipath, "halt") == 0) {
803 				cpu_reboot(RB_HALT, NULL);
804 			} else if (len == 6 && strcmp(ipath, "reboot") == 0) {
805 				cpu_reboot(0, NULL);
806 #if defined(DDB)
807 			} else if (len == 3 && strcmp(ipath, "ddb") == 0) {
808 				console_debugger();
809 				continue;
810 #endif
811 			} else if (len > 0 && ipath[0] == '/') {
812 				ipath[len] = '\0';
813 				path = ipath;
814 			} else if (len == 0 && initpaths[ipx] != NULL) {
815 				path = initpaths[ipx++];
816 			} else {
817 				printf("use absolute path, ");
818 #if defined(DDB)
819 				printf("\"ddb\", ");
820 #endif
821 				printf("\"halt\", or \"reboot\"\n");
822 				continue;
823 			}
824 		} else {
825 			if ((path = initpaths[ipx++]) == NULL) {
826 				ipx = 0;
827 				boothowto |= RB_ASKNAME;
828 				continue;
829 			}
830 		}
831 
832 		ucp = (char *)USRSTACK;
833 
834 		/*
835 		 * Construct the boot flag argument.
836 		 */
837 		flagsp = flags;
838 		*flagsp++ = '-';
839 		options = 0;
840 
841 		if (boothowto & RB_SINGLE) {
842 			*flagsp++ = 's';
843 			options = 1;
844 		}
845 #ifdef notyet
846 		if (boothowto & RB_FASTBOOT) {
847 			*flagsp++ = 'f';
848 			options = 1;
849 		}
850 #endif
851 
852 		/*
853 		 * Move out the flags (arg 1), if necessary.
854 		 */
855 		if (options != 0) {
856 			*flagsp++ = '\0';
857 			i = flagsp - flags;
858 #ifdef DEBUG
859 			printf("init: copying out flags `%s' %d\n", flags, i);
860 #endif
861 			arg1 = STACK_ALLOC(ucp, i);
862 			ucp = STACK_MAX(arg1, i);
863 			(void)copyout((void *)flags, arg1, i);
864 		}
865 
866 		/*
867 		 * Move out the file name (also arg 0).
868 		 */
869 		i = strlen(path) + 1;
870 #ifdef DEBUG
871 		printf("init: copying out path `%s' %d\n", path, i);
872 #else
873 		if (boothowto & RB_ASKNAME || path != initpaths[0])
874 			printf("init: trying %s\n", path);
875 #endif
876 		arg0 = STACK_ALLOC(ucp, i);
877 		ucp = STACK_MAX(arg0, i);
878 		(void)copyout(path, arg0, i);
879 
880 		/*
881 		 * Move out the arg pointers.
882 		 */
883 		ucp = (void *)STACK_ALIGN(ucp, ALIGNBYTES);
884 		uap = (char **)STACK_ALLOC(ucp, sizeof(char *) * 3);
885 		SCARG(&args, path) = arg0;
886 		SCARG(&args, argp) = uap;
887 		SCARG(&args, envp) = NULL;
888 		slash = strrchr(path, '/');
889 		if (slash)
890 			(void)suword((void *)uap++,
891 			    (long)arg0 + (slash + 1 - path));
892 		else
893 			(void)suword((void *)uap++, (long)arg0);
894 		if (options != 0)
895 			(void)suword((void *)uap++, (long)arg1);
896 		(void)suword((void *)uap++, 0);	/* terminator */
897 
898 		/*
899 		 * Now try to exec the program.  If can't for any reason
900 		 * other than it doesn't exist, complain.
901 		 */
902 		error = sys_execve(l, &args, retval);
903 		if (error == 0 || error == EJUSTRETURN) {
904 			KERNEL_UNLOCK_LAST(l);
905 			return;
906 		}
907 		printf("exec %s: error %d\n", path, error);
908 	}
909 	printf("init: not found\n");
910 	panic("no init");
911 }
912 
913 /*
914  * calculate cache size from physmem and vm_map size.
915  */
916 vaddr_t
917 calc_cache_size(struct vm_map *map, int pct, int va_pct)
918 {
919 	paddr_t t;
920 
921 	/* XXX should consider competing cache if any */
922 	/* XXX should consider submaps */
923 	t = (uintmax_t)physmem * pct / 100 * PAGE_SIZE;
924 	if (map != NULL) {
925 		vsize_t vsize;
926 
927 		vsize = vm_map_max(map) - vm_map_min(map);
928 		vsize = (uintmax_t)vsize * va_pct / 100;
929 		if (t > vsize) {
930 			t = vsize;
931 		}
932 	}
933 	return t;
934 }
935