xref: /netbsd-src/sys/kern/init_main.c (revision 404fbe5fb94ca1e054339640cabb2801ce52dd30)
1 /*	$NetBSD: init_main.c,v 1.380 2009/01/11 02:45:51 christos 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 1.380 2009/01/11 02:45:51 christos 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_compat_netbsd.h"
112 #include "opt_wapbl.h"
113 
114 #include "ksyms.h"
115 #include "rnd.h"
116 #include "sysmon_envsys.h"
117 #include "sysmon_power.h"
118 #include "sysmon_taskq.h"
119 #include "sysmon_wdog.h"
120 #include "veriexec.h"
121 
122 #include <sys/param.h>
123 #include <sys/acct.h>
124 #include <sys/filedesc.h>
125 #include <sys/file.h>
126 #include <sys/errno.h>
127 #include <sys/callout.h>
128 #include <sys/cpu.h>
129 #include <sys/kernel.h>
130 #include <sys/mount.h>
131 #include <sys/proc.h>
132 #include <sys/kthread.h>
133 #include <sys/resourcevar.h>
134 #include <sys/signalvar.h>
135 #include <sys/systm.h>
136 #include <sys/vnode.h>
137 #include <sys/fstrans.h>
138 #include <sys/tty.h>
139 #include <sys/conf.h>
140 #include <sys/disklabel.h>
141 #include <sys/buf.h>
142 #include <sys/device.h>
143 #include <sys/exec.h>
144 #include <sys/socketvar.h>
145 #include <sys/protosw.h>
146 #include <sys/percpu.h>
147 #include <sys/pset.h>
148 #include <sys/sysctl.h>
149 #include <sys/reboot.h>
150 #include <sys/user.h>
151 #include <sys/sysctl.h>
152 #include <sys/event.h>
153 #include <sys/mbuf.h>
154 #include <sys/sched.h>
155 #include <sys/sleepq.h>
156 #include <sys/iostat.h>
157 #include <sys/vmem.h>
158 #include <sys/uuid.h>
159 #include <sys/extent.h>
160 #include <sys/disk.h>
161 #include <sys/mqueue.h>
162 #include <sys/msgbuf.h>
163 #include <sys/module.h>
164 #include <sys/event.h>
165 #include <sys/lockf.h>
166 #include <sys/once.h>
167 #include <sys/ksyms.h>
168 #include <sys/uidinfo.h>
169 #include <sys/kprintf.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 #include <sys/domain.h>
183 #include <sys/namei.h>
184 #if NRND > 0
185 #include <sys/rnd.h>
186 #endif
187 #include <sys/pipe.h>
188 #if NVERIEXEC > 0
189 #include <sys/verified_exec.h>
190 #endif /* NVERIEXEC > 0 */
191 #ifdef KTRACE
192 #include <sys/ktrace.h>
193 #endif
194 #include <sys/kauth.h>
195 #ifdef WAPBL
196 #include <sys/wapbl.h>
197 #endif
198 #include <net80211/ieee80211_netbsd.h>
199 
200 #include <sys/syscall.h>
201 #include <sys/syscallargs.h>
202 
203 #if defined(PAX_MPROTECT) || defined(PAX_SEGVGUARD) || defined(PAX_ASLR)
204 #include <sys/pax.h>
205 #endif /* PAX_MPROTECT || PAX_SEGVGUARD || PAX_ASLR */
206 
207 #include <ufs/ufs/quota.h>
208 
209 #include <miscfs/genfs/genfs.h>
210 #include <miscfs/syncfs/syncfs.h>
211 
212 #include <sys/cpu.h>
213 
214 #include <uvm/uvm.h>
215 
216 #if NSYSMON_TASKQ > 0
217 #include <dev/sysmon/sysmon_taskq.h>
218 #endif
219 
220 #include <dev/cons.h>
221 
222 #if NSYSMON_ENVSYS > 0 || NSYSMON_POWER > 0 || NSYSMON_WDOG > 0
223 #include <dev/sysmon/sysmonvar.h>
224 #endif
225 
226 #include <net/if.h>
227 #include <net/raw_cb.h>
228 
229 #include <secmodel/secmodel.h>
230 
231 #ifdef COMPAT_50
232 #include <compat/sys/time.h>
233 struct timeval50 boottime50;
234 #endif
235 
236 extern struct proc proc0;
237 extern struct lwp lwp0;
238 extern struct cwdinfo cwdi0;
239 extern time_t rootfstime;
240 
241 #ifndef curlwp
242 struct	lwp *curlwp = &lwp0;
243 #endif
244 struct	proc *initproc;
245 
246 struct	vnode *rootvp, *swapdev_vp;
247 int	boothowto;
248 int	cold = 1;			/* still working on startup */
249 struct timespec boottime;	        /* time at system startup - will only follow settime deltas */
250 
251 int	start_init_exec;		/* semaphore for start_init() */
252 
253 static void check_console(struct lwp *l);
254 static void start_init(void *);
255 void main(void);
256 void ssp_init(void);
257 
258 #if defined(__SSP__) || defined(__SSP_ALL__)
259 long __stack_chk_guard[8] = {0, 0, 0, 0, 0, 0, 0, 0};
260 void __stack_chk_fail(void);
261 
262 void
263 __stack_chk_fail(void)
264 {
265 	panic("stack overflow detected; terminated");
266 }
267 
268 void
269 ssp_init(void)
270 {
271 	int s;
272 
273 #ifdef DIAGNOSTIC
274 	printf("Initializing SSP:");
275 #endif
276 	/*
277 	 * We initialize ssp here carefully:
278 	 *	1. after we got some entropy
279 	 *	2. without calling a function
280 	 */
281 	size_t i;
282 	long guard[__arraycount(__stack_chk_guard)];
283 
284 	arc4randbytes(guard, sizeof(guard));
285 	s = splhigh();
286 	for (i = 0; i < __arraycount(guard); i++)
287 		__stack_chk_guard[i] = guard[i];
288 	splx(s);
289 #ifdef DIAGNOSTIC
290 	for (i = 0; i < __arraycount(guard); i++)
291 		printf("%lx ", guard[i]);
292 	printf("\n");
293 #endif
294 }
295 #else
296 void
297 ssp_init(void)
298 {
299 
300 }
301 #endif
302 
303 void __secmodel_none(void);
304 __weak_alias(secmodel_start,__secmodel_none);
305 void
306 __secmodel_none(void)
307 {
308 	return;
309 }
310 
311 /*
312  * System startup; initialize the world, create process 0, mount root
313  * filesystem, and fork to create init and pagedaemon.  Most of the
314  * hard work is done in the lower-level initialization routines including
315  * startup(), which does memory initialization and autoconfiguration.
316  */
317 void
318 main(void)
319 {
320 	struct timespec time;
321 	struct lwp *l;
322 	struct proc *p;
323 	int s, error;
324 #ifdef NVNODE_IMPLICIT
325 	int usevnodes;
326 #endif
327 	CPU_INFO_ITERATOR cii;
328 	struct cpu_info *ci;
329 
330 	l = &lwp0;
331 #ifndef LWP0_CPU_INFO
332 	l->l_cpu = curcpu();
333 #endif
334 
335 	/*
336 	 * Attempt to find console and initialize
337 	 * in case of early panic or other messages.
338 	 */
339 	consinit();
340 
341 	kernel_lock_init();
342 	once_init();
343 
344 	uvm_init();
345 
346 #if ((NKSYMS > 0) || (NDDB > 0) || (NMODULAR > 0))
347 	ksyms_init();
348 #endif
349 	kprintf_init();
350 
351 	percpu_init();
352 
353 	/* Initialize lock caches. */
354 	mutex_obj_init();
355 
356 	/* Initialize the extent manager. */
357 	extent_init();
358 
359 	/* Do machine-dependent initialization. */
360 	cpu_startup();
361 
362 	/* Initialize the sysctl subsystem. */
363 	sysctl_init();
364 
365 	/* Initialize callouts, part 1. */
366 	callout_startup();
367 
368 	/*
369 	 * Initialize the kernel authorization subsystem and start the
370 	 * default security model, if any. We need to do this early
371 	 * enough so that subsystems relying on any of the aforementioned
372 	 * can work properly. Since the security model may dictate the
373 	 * credential inheritance policy, it is needed at least before
374 	 * any process is created, specifically proc0.
375 	 */
376 	kauth_init();
377 	secmodel_start();
378 
379 	/* Initialize the buffer cache */
380 	bufinit();
381 
382 	/* Initialize sockets. */
383 	soinit();
384 
385 	/*
386 	 * The following things must be done before autoconfiguration.
387 	 */
388 	evcnt_init();		/* initialize event counters */
389 #if NRND > 0
390 	rnd_init();		/* initialize random number generator */
391 #endif
392 
393 	/* Initialize process and pgrp structures. */
394 	procinit();
395 	lwpinit();
396 
397 	/* Initialize signal-related data structures. */
398 	signal_init();
399 
400 	/* Initialize resource management. */
401 	resource_init();
402 
403 	/* Create process 0 (the swapper). */
404 	proc0_init();
405 
406 	/* Initialize the UID hash table. */
407 	uid_init();
408 
409 	/* Charge root for one process. */
410 	(void)chgproccnt(0, 1);
411 
412 	/* Initialize timekeeping. */
413 	time_init();
414 
415 	/* Initialize the run queues, turnstiles and sleep queues. */
416 	mutex_init(&cpu_lock, MUTEX_DEFAULT, IPL_NONE);
417 	sched_rqinit();
418 	turnstile_init();
419 	sleeptab_init(&sleeptab);
420 
421 	/* Initialize processor-sets */
422 	psets_init();
423 
424 	/* MI initialization of the boot cpu */
425 	error = mi_cpu_attach(curcpu());
426 	KASSERT(error == 0);
427 
428 	/* Initialize timekeeping, part 2. */
429 	time_init2();
430 
431 	/*
432 	 * Initialize mbuf's.  Do this now because we might attempt to
433 	 * allocate mbufs or mbuf clusters during autoconfiguration.
434 	 */
435 	mbinit();
436 
437 	/* Initialize I/O statistics. */
438 	iostat_init();
439 
440 	/* Initialize the log device. */
441 	loginit();
442 
443 	/* Start module system. */
444 	module_init();
445 
446 	/* Initialize the file systems. */
447 #ifdef NVNODE_IMPLICIT
448 	/*
449 	 * If maximum number of vnodes in namei vnode cache is not explicitly
450 	 * defined in kernel config, adjust the number such as we use roughly
451 	 * 10% of memory for vnodes and associated data structures in the
452 	 * assumed worst case.  Do not provide fewer than NVNODE vnodes.
453 	 */
454 	usevnodes =
455 	    calc_cache_size(kernel_map, 10, VNODE_VA_MAXPCT) / VNODE_COST;
456 	if (usevnodes > desiredvnodes)
457 		desiredvnodes = usevnodes;
458 #endif
459 	vfsinit();
460 	lf_init();
461 
462 	/* Initialize fstrans. */
463 	fstrans_init();
464 
465 	/* Initialize the file descriptor system. */
466 	fd_sys_init();
467 
468 	/* Initialize cwd structures */
469 	cwd_sys_init();
470 
471 	/* Initialize kqueue. */
472 	kqueue_init();
473 
474 	/* Initialize message queues. */
475 	mqueue_sysinit();
476 
477 	/* Initialize the system monitor subsystems. */
478 #if NSYSMON_TASKQ > 0
479 	sysmon_task_queue_preinit();
480 #endif
481 
482 #if NSYSMON_ENVSYS > 0
483 	sysmon_envsys_init();
484 #endif
485 
486 #if NSYSMON_POWER > 0
487 	sysmon_power_init();
488 #endif
489 
490 #if NSYSMON_WDOG > 0
491 	sysmon_wdog_init();
492 #endif
493 
494 	inittimecounter();
495 	ntp_init();
496 
497 	/* Initialize the device switch tables. */
498 	devsw_init();
499 
500 	/* Initialize tty subsystem. */
501 	tty_init();
502 	ttyldisc_init();
503 
504 	/* Initialize the buffer cache, part 2. */
505 	bufinit2();
506 
507 	/* Initialize the disk wedge subsystem. */
508 	dkwedge_init();
509 
510 	/* Initialize interfaces. */
511 	ifinit1();
512 
513 	/* Configure the system hardware.  This will enable interrupts. */
514 	configure();
515 
516 	ubc_init();		/* must be after autoconfig */
517 
518 #ifdef SYSVSHM
519 	/* Initialize System V style shared memory. */
520 	shminit();
521 #endif
522 
523 #ifdef SYSVSEM
524 	/* Initialize System V style semaphores. */
525 	seminit();
526 #endif
527 
528 #ifdef SYSVMSG
529 	/* Initialize System V style message queues. */
530 	msginit();
531 #endif
532 
533 #if NVERIEXEC > 0
534 	/*
535 	 * Initialise the Veriexec subsystem.
536 	 */
537 	veriexec_init();
538 #endif /* NVERIEXEC > 0 */
539 
540 #if defined(PAX_MPROTECT) || defined(PAX_SEGVGUARD) || defined(PAX_ASLR)
541 	pax_init();
542 #endif /* PAX_MPROTECT || PAX_SEGVGUARD || PAX_ASLR */
543 
544 #ifdef	FAST_IPSEC
545 	/* Attach network crypto subsystem */
546 	ipsec_attach();
547 #endif
548 
549 	/*
550 	 * Initialize protocols.  Block reception of incoming packets
551 	 * until everything is ready.
552 	 */
553 	s = splnet();
554 	ifinit();
555 	domaininit();
556 	if_attachdomain();
557 	splx(s);
558 
559 #ifdef GPROF
560 	/* Initialize kernel profiling. */
561 	kmstartup();
562 #endif
563 
564 	/* Initialize system accounting. */
565 	acct_init();
566 
567 #ifndef PIPE_SOCKETPAIR
568 	/* Initialize pipes. */
569 	pipe_init();
570 #endif
571 
572 #ifdef KTRACE
573 	/* Initialize ktrace. */
574 	ktrinit();
575 #endif
576 
577 	/* Initialize the UUID system calls. */
578 	uuid_init();
579 
580 #ifdef WAPBL
581 	/* Initialize write-ahead physical block logging. */
582 	wapbl_init();
583 #endif
584 
585 	/*
586 	 * Create process 1 (init(8)).  We do this now, as Unix has
587 	 * historically had init be process 1, and changing this would
588 	 * probably upset a lot of people.
589 	 *
590 	 * Note that process 1 won't immediately exec init(8), but will
591 	 * wait for us to inform it that the root file system has been
592 	 * mounted.
593 	 */
594 	if (fork1(l, 0, SIGCHLD, NULL, 0, start_init, NULL, NULL, &initproc))
595 		panic("fork init");
596 
597 	/*
598 	 * Load any remaining builtin modules, and hand back temporary
599 	 * storage to the VM system.
600 	 */
601 	module_init_class(MODULE_CLASS_ANY);
602 
603 	/*
604 	 * Finalize configuration now that all real devices have been
605 	 * found.  This needs to be done before the root device is
606 	 * selected, since finalization may create the root device.
607 	 */
608 	config_finalize();
609 
610 	/*
611 	 * Now that autoconfiguration has completed, we can determine
612 	 * the root and dump devices.
613 	 */
614 	cpu_rootconf();
615 	cpu_dumpconf();
616 
617 	/* Mount the root file system. */
618 	do {
619 		domountroothook();
620 		if ((error = vfs_mountroot())) {
621 			printf("cannot mount root, error = %d\n", error);
622 			boothowto |= RB_ASKNAME;
623 			setroot(root_device,
624 			    (rootdev != NODEV) ? DISKPART(rootdev) : 0);
625 		}
626 	} while (error != 0);
627 	mountroothook_destroy();
628 
629 	/*
630 	 * Initialise the time-of-day clock, passing the time recorded
631 	 * in the root filesystem (if any) for use by systems that
632 	 * don't have a non-volatile time-of-day device.
633 	 */
634 	inittodr(rootfstime);
635 
636 	CIRCLEQ_FIRST(&mountlist)->mnt_flag |= MNT_ROOTFS;
637 	CIRCLEQ_FIRST(&mountlist)->mnt_op->vfs_refcount++;
638 
639 	/*
640 	 * Get the vnode for '/'.  Set filedesc0.fd_fd.fd_cdir to
641 	 * reference it.
642 	 */
643 	error = VFS_ROOT(CIRCLEQ_FIRST(&mountlist), &rootvnode);
644 	if (error)
645 		panic("cannot find root vnode, error=%d", error);
646 	cwdi0.cwdi_cdir = rootvnode;
647 	VREF(cwdi0.cwdi_cdir);
648 	VOP_UNLOCK(rootvnode, 0);
649 	cwdi0.cwdi_rdir = NULL;
650 
651 	/*
652 	 * Now that root is mounted, we can fixup initproc's CWD
653 	 * info.  All other processes are kthreads, which merely
654 	 * share proc0's CWD info.
655 	 */
656 	initproc->p_cwdi->cwdi_cdir = rootvnode;
657 	VREF(initproc->p_cwdi->cwdi_cdir);
658 	initproc->p_cwdi->cwdi_rdir = NULL;
659 
660 	/*
661 	 * Now can look at time, having had a chance to verify the time
662 	 * from the file system.  Reset l->l_rtime as it may have been
663 	 * munched in mi_switch() after the time got set.
664 	 */
665 	getnanotime(&time);
666 	boottime = time;
667 #ifdef COMPAT_50
668 	{
669 		struct timeval tv;
670 		TIMESPEC_TO_TIMEVAL(&tv, &time);
671 		timeval_to_timeval50(&tv, &boottime50);
672 	}
673 #endif
674 	mutex_enter(proc_lock);
675 	LIST_FOREACH(p, &allproc, p_list) {
676 		KASSERT((p->p_flag & PK_MARKER) == 0);
677 		mutex_enter(p->p_lock);
678 		TIMESPEC_TO_TIMEVAL(&p->p_stats->p_start, &time);
679 		LIST_FOREACH(l, &p->p_lwps, l_sibling) {
680 			lwp_lock(l);
681 			memset(&l->l_rtime, 0, sizeof(l->l_rtime));
682 			lwp_unlock(l);
683 		}
684 		mutex_exit(p->p_lock);
685 	}
686 	mutex_exit(proc_lock);
687 	binuptime(&curlwp->l_stime);
688 
689 	for (CPU_INFO_FOREACH(cii, ci)) {
690 		ci->ci_schedstate.spc_lastmod = time_second;
691 	}
692 
693 	/* Create the pageout daemon kernel thread. */
694 	uvm_swap_init();
695 	if (kthread_create(PRI_PGDAEMON, KTHREAD_MPSAFE, NULL, uvm_pageout,
696 	    NULL, NULL, "pgdaemon"))
697 		panic("fork pagedaemon");
698 
699 	/* Create the filesystem syncer kernel thread. */
700 	if (kthread_create(PRI_IOFLUSH, KTHREAD_MPSAFE, NULL, sched_sync,
701 	    NULL, NULL, "ioflush"))
702 		panic("fork syncer");
703 
704 	/* Create the aiodone daemon kernel thread. */
705 	if (workqueue_create(&uvm.aiodone_queue, "aiodoned",
706 	    uvm_aiodone_worker, NULL, PRI_VM, IPL_NONE, WQ_MPSAFE))
707 		panic("fork aiodoned");
708 
709 	vmem_rehash_start();
710 
711 	/* Initialize exec structures */
712 	exec_init(1);
713 
714 	/*
715 	 * Okay, now we can let init(8) exec!  It's off to userland!
716 	 */
717 	mutex_enter(proc_lock);
718 	start_init_exec = 1;
719 	cv_broadcast(&lbolt);
720 	mutex_exit(proc_lock);
721 
722 	/* The scheduler is an infinite loop. */
723 	uvm_scheduler();
724 	/* NOTREACHED */
725 }
726 
727 static void
728 check_console(struct lwp *l)
729 {
730 	struct nameidata nd;
731 	int error;
732 
733 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_SYSSPACE, "/dev/console");
734 	error = namei(&nd);
735 	if (error == 0)
736 		vrele(nd.ni_vp);
737 	else if (error == ENOENT)
738 		printf("warning: no /dev/console\n");
739 	else
740 		printf("warning: lookup /dev/console: error %d\n", error);
741 }
742 
743 /*
744  * List of paths to try when searching for "init".
745  */
746 static const char * const initpaths[] = {
747 	"/sbin/init",
748 	"/sbin/oinit",
749 	"/sbin/init.bak",
750 	NULL,
751 };
752 
753 /*
754  * Start the initial user process; try exec'ing each pathname in "initpaths".
755  * The program is invoked with one argument containing the boot flags.
756  */
757 static void
758 start_init(void *arg)
759 {
760 	struct lwp *l = arg;
761 	struct proc *p = l->l_proc;
762 	vaddr_t addr;
763 	struct sys_execve_args /* {
764 		syscallarg(const char *) path;
765 		syscallarg(char * const *) argp;
766 		syscallarg(char * const *) envp;
767 	} */ args;
768 	int options, i, error;
769 	register_t retval[2];
770 	char flags[4], *flagsp;
771 	const char *path, *slash;
772 	char *ucp, **uap, *arg0, *arg1 = NULL;
773 	char ipath[129];
774 	int ipx, len;
775 
776 	/*
777 	 * Now in process 1.
778 	 */
779 	strncpy(p->p_comm, "init", MAXCOMLEN);
780 
781 	/*
782 	 * Wait for main() to tell us that it's safe to exec.
783 	 */
784 	mutex_enter(proc_lock);
785 	while (start_init_exec == 0)
786 		cv_wait(&lbolt, proc_lock);
787 	mutex_exit(proc_lock);
788 
789 	/*
790 	 * This is not the right way to do this.  We really should
791 	 * hand-craft a descriptor onto /dev/console to hand to init,
792 	 * but that's a _lot_ more work, and the benefit from this easy
793 	 * hack makes up for the "good is the enemy of the best" effect.
794 	 */
795 	check_console(l);
796 
797 	/*
798 	 * Need just enough stack to hold the faked-up "execve()" arguments.
799 	 */
800 	addr = (vaddr_t)STACK_ALLOC(USRSTACK, PAGE_SIZE);
801 	if (uvm_map(&p->p_vmspace->vm_map, &addr, PAGE_SIZE,
802                     NULL, UVM_UNKNOWN_OFFSET, 0,
803                     UVM_MAPFLAG(UVM_PROT_ALL, UVM_PROT_ALL, UVM_INH_COPY,
804 		    UVM_ADV_NORMAL,
805                     UVM_FLAG_FIXED|UVM_FLAG_OVERLAY|UVM_FLAG_COPYONW)) != 0)
806 		panic("init: couldn't allocate argument space");
807 	p->p_vmspace->vm_maxsaddr = (void *)STACK_MAX(addr, PAGE_SIZE);
808 
809 	ipx = 0;
810 	while (1) {
811 		if (boothowto & RB_ASKNAME) {
812 			printf("init path");
813 			if (initpaths[ipx])
814 				printf(" (default %s)", initpaths[ipx]);
815 			printf(": ");
816 			len = cngetsn(ipath, sizeof(ipath)-1);
817 			if (len == 4 && strcmp(ipath, "halt") == 0) {
818 				cpu_reboot(RB_HALT, NULL);
819 			} else if (len == 6 && strcmp(ipath, "reboot") == 0) {
820 				cpu_reboot(0, NULL);
821 #if defined(DDB)
822 			} else if (len == 3 && strcmp(ipath, "ddb") == 0) {
823 				console_debugger();
824 				continue;
825 #endif
826 			} else if (len > 0 && ipath[0] == '/') {
827 				ipath[len] = '\0';
828 				path = ipath;
829 			} else if (len == 0 && initpaths[ipx] != NULL) {
830 				path = initpaths[ipx++];
831 			} else {
832 				printf("use absolute path, ");
833 #if defined(DDB)
834 				printf("\"ddb\", ");
835 #endif
836 				printf("\"halt\", or \"reboot\"\n");
837 				continue;
838 			}
839 		} else {
840 			if ((path = initpaths[ipx++]) == NULL) {
841 				ipx = 0;
842 				boothowto |= RB_ASKNAME;
843 				continue;
844 			}
845 		}
846 
847 		ucp = (char *)USRSTACK;
848 
849 		/*
850 		 * Construct the boot flag argument.
851 		 */
852 		flagsp = flags;
853 		*flagsp++ = '-';
854 		options = 0;
855 
856 		if (boothowto & RB_SINGLE) {
857 			*flagsp++ = 's';
858 			options = 1;
859 		}
860 #ifdef notyet
861 		if (boothowto & RB_FASTBOOT) {
862 			*flagsp++ = 'f';
863 			options = 1;
864 		}
865 #endif
866 
867 		/*
868 		 * Move out the flags (arg 1), if necessary.
869 		 */
870 		if (options != 0) {
871 			*flagsp++ = '\0';
872 			i = flagsp - flags;
873 #ifdef DEBUG
874 			printf("init: copying out flags `%s' %d\n", flags, i);
875 #endif
876 			arg1 = STACK_ALLOC(ucp, i);
877 			ucp = STACK_MAX(arg1, i);
878 			(void)copyout((void *)flags, arg1, i);
879 		}
880 
881 		/*
882 		 * Move out the file name (also arg 0).
883 		 */
884 		i = strlen(path) + 1;
885 #ifdef DEBUG
886 		printf("init: copying out path `%s' %d\n", path, i);
887 #else
888 		if (boothowto & RB_ASKNAME || path != initpaths[0])
889 			printf("init: trying %s\n", path);
890 #endif
891 		arg0 = STACK_ALLOC(ucp, i);
892 		ucp = STACK_MAX(arg0, i);
893 		(void)copyout(path, arg0, i);
894 
895 		/*
896 		 * Move out the arg pointers.
897 		 */
898 		ucp = (void *)STACK_ALIGN(ucp, ALIGNBYTES);
899 		uap = (char **)STACK_ALLOC(ucp, sizeof(char *) * 3);
900 		SCARG(&args, path) = arg0;
901 		SCARG(&args, argp) = uap;
902 		SCARG(&args, envp) = NULL;
903 		slash = strrchr(path, '/');
904 		if (slash)
905 			(void)suword((void *)uap++,
906 			    (long)arg0 + (slash + 1 - path));
907 		else
908 			(void)suword((void *)uap++, (long)arg0);
909 		if (options != 0)
910 			(void)suword((void *)uap++, (long)arg1);
911 		(void)suword((void *)uap++, 0);	/* terminator */
912 
913 		/*
914 		 * Now try to exec the program.  If can't for any reason
915 		 * other than it doesn't exist, complain.
916 		 */
917 		error = sys_execve(l, &args, retval);
918 		if (error == 0 || error == EJUSTRETURN) {
919 			KERNEL_UNLOCK_LAST(l);
920 			return;
921 		}
922 		printf("exec %s: error %d\n", path, error);
923 	}
924 	printf("init: not found\n");
925 	panic("no init");
926 }
927 
928 /*
929  * calculate cache size from physmem and vm_map size.
930  */
931 vaddr_t
932 calc_cache_size(struct vm_map *map, int pct, int va_pct)
933 {
934 	paddr_t t;
935 
936 	/* XXX should consider competing cache if any */
937 	/* XXX should consider submaps */
938 	t = (uintmax_t)physmem * pct / 100 * PAGE_SIZE;
939 	if (map != NULL) {
940 		vsize_t vsize;
941 
942 		vsize = vm_map_max(map) - vm_map_min(map);
943 		vsize = (uintmax_t)vsize * va_pct / 100;
944 		if (t > vsize) {
945 			t = vsize;
946 		}
947 	}
948 	return t;
949 }
950