xref: /netbsd-src/sys/kern/init_main.c (revision 0efea177e37295fed654875d875160c47500d6bd)
1 /*	$NetBSD: init_main.c,v 1.372 2008/11/12 12:36:16 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 1.372 2008/11/12 12:36:16 ad 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_posix.h"
107 #include "opt_syscall_debug.h"
108 #include "opt_sysv.h"
109 #include "opt_fileassoc.h"
110 #include "opt_ktrace.h"
111 #include "opt_pax.h"
112 #include "opt_wapbl.h"
113 
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 #ifdef P1003_1B_SEMAPHORE
181 #include <sys/ksem.h>
182 #endif
183 #include <sys/domain.h>
184 #include <sys/namei.h>
185 #if NRND > 0
186 #include <sys/rnd.h>
187 #endif
188 #include <sys/pipe.h>
189 #if NVERIEXEC > 0
190 #include <sys/verified_exec.h>
191 #endif /* NVERIEXEC > 0 */
192 #ifdef KTRACE
193 #include <sys/ktrace.h>
194 #endif
195 #include <sys/kauth.h>
196 #ifdef WAPBL
197 #include <sys/wapbl.h>
198 #endif
199 #include <net80211/ieee80211_netbsd.h>
200 
201 #include <sys/syscall.h>
202 #include <sys/syscallargs.h>
203 
204 #if defined(PAX_MPROTECT) || defined(PAX_SEGVGUARD) || defined(PAX_ASLR)
205 #include <sys/pax.h>
206 #endif /* PAX_MPROTECT || PAX_SEGVGUARD || PAX_ASLR */
207 
208 #include <ufs/ufs/quota.h>
209 
210 #include <miscfs/genfs/genfs.h>
211 #include <miscfs/syncfs/syncfs.h>
212 
213 #include <sys/cpu.h>
214 
215 #include <uvm/uvm.h>
216 
217 #if NSYSMON_TASKQ > 0
218 #include <dev/sysmon/sysmon_taskq.h>
219 #endif
220 
221 #include <dev/cons.h>
222 
223 #if NSYSMON_ENVSYS > 0 || NSYSMON_POWER > 0 || NSYSMON_WDOG > 0
224 #include <dev/sysmon/sysmonvar.h>
225 #endif
226 
227 #include <net/if.h>
228 #include <net/raw_cb.h>
229 
230 #include <secmodel/secmodel.h>
231 
232 extern struct proc proc0;
233 extern struct lwp lwp0;
234 extern struct cwdinfo cwdi0;
235 extern time_t rootfstime;
236 
237 #ifndef curlwp
238 struct	lwp *curlwp = &lwp0;
239 #endif
240 struct	proc *initproc;
241 
242 struct	vnode *rootvp, *swapdev_vp;
243 int	boothowto;
244 int	cold = 1;			/* still working on startup */
245 struct timeval boottime;	        /* time at system startup - will only follow settime deltas */
246 
247 int	start_init_exec;		/* semaphore for start_init() */
248 
249 static void check_console(struct lwp *l);
250 static void start_init(void *);
251 void main(void);
252 void ssp_init(void);
253 
254 #if defined(__SSP__) || defined(__SSP_ALL__)
255 long __stack_chk_guard[8] = {0, 0, 0, 0, 0, 0, 0, 0};
256 void __stack_chk_fail(void);
257 
258 void
259 __stack_chk_fail(void)
260 {
261 	panic("stack overflow detected; terminated");
262 }
263 
264 void
265 ssp_init(void)
266 {
267 	int s;
268 
269 #ifdef DIAGNOSTIC
270 	printf("Initializing SSP:");
271 #endif
272 	/*
273 	 * We initialize ssp here carefully:
274 	 *	1. after we got some entropy
275 	 *	2. without calling a function
276 	 */
277 	size_t i;
278 	long guard[__arraycount(__stack_chk_guard)];
279 
280 	arc4randbytes(guard, sizeof(guard));
281 	s = splhigh();
282 	for (i = 0; i < __arraycount(guard); i++)
283 		__stack_chk_guard[i] = guard[i];
284 	splx(s);
285 #ifdef DIAGNOSTIC
286 	for (i = 0; i < __arraycount(guard); i++)
287 		printf("%lx ", guard[i]);
288 	printf("\n");
289 #endif
290 }
291 #else
292 void
293 ssp_init(void)
294 {
295 
296 }
297 #endif
298 
299 void __secmodel_none(void);
300 __weak_alias(secmodel_start,__secmodel_none);
301 void
302 __secmodel_none(void)
303 {
304 	return;
305 }
306 
307 /*
308  * System startup; initialize the world, create process 0, mount root
309  * filesystem, and fork to create init and pagedaemon.  Most of the
310  * hard work is done in the lower-level initialization routines including
311  * startup(), which does memory initialization and autoconfiguration.
312  */
313 void
314 main(void)
315 {
316 	struct timeval time;
317 	struct lwp *l;
318 	struct proc *p;
319 	int s, error;
320 #ifdef NVNODE_IMPLICIT
321 	int usevnodes;
322 #endif
323 	CPU_INFO_ITERATOR cii;
324 	struct cpu_info *ci;
325 
326 	l = &lwp0;
327 #ifndef LWP0_CPU_INFO
328 	l->l_cpu = curcpu();
329 #endif
330 
331 	/*
332 	 * Attempt to find console and initialize
333 	 * in case of early panic or other messages.
334 	 */
335 	consinit();
336 
337 	kernel_lock_init();
338 	once_init();
339 
340 	uvm_init();
341 
342 	percpu_init();
343 
344 	/* Initialize lock caches. */
345 	mutex_obj_init();
346 
347 	/* Initialize the extent manager. */
348 	extent_init();
349 
350 	/* Do machine-dependent initialization. */
351 	cpu_startup();
352 
353 	/* Initialize callouts, part 1. */
354 	callout_startup();
355 
356 	/*
357 	 * Initialize the kernel authorization subsystem and start the
358 	 * default security model, if any. We need to do this early
359 	 * enough so that subsystems relying on any of the aforementioned
360 	 * can work properly. Since the security model may dictate the
361 	 * credential inheritance policy, it is needed at least before
362 	 * any process is created, specifically proc0.
363 	 */
364 	kauth_init();
365 	secmodel_start();
366 
367 	/* Initialize the buffer cache */
368 	bufinit();
369 
370 	/* Initialize sockets. */
371 	soinit();
372 
373 	/*
374 	 * The following things must be done before autoconfiguration.
375 	 */
376 	evcnt_init();		/* initialize event counters */
377 #if NRND > 0
378 	rnd_init();		/* initialize random number generator */
379 #endif
380 
381 	/* Initialize process and pgrp structures. */
382 	procinit();
383 	lwpinit();
384 
385 	/* Initialize signal-related data structures. */
386 	signal_init();
387 
388 	/* Initialize resource management. */
389 	resource_init();
390 
391 	/* Create process 0 (the swapper). */
392 	proc0_init();
393 
394 	/* Initialize the UID hash table. */
395 	uid_init();
396 
397 	/* Charge root for one process. */
398 	(void)chgproccnt(0, 1);
399 
400 	/* Initialize timekeeping. */
401 	time_init();
402 
403 	/* Initialize the run queues, turnstiles and sleep queues. */
404 	mutex_init(&cpu_lock, MUTEX_DEFAULT, IPL_NONE);
405 	sched_rqinit();
406 	turnstile_init();
407 	sleeptab_init(&sleeptab);
408 
409 	/* Initialize processor-sets */
410 	psets_init();
411 
412 	/* MI initialization of the boot cpu */
413 	error = mi_cpu_attach(curcpu());
414 	KASSERT(error == 0);
415 
416 	/* Initialize timekeeping, part 2. */
417 	time_init2();
418 
419 	/*
420 	 * Initialize mbuf's.  Do this now because we might attempt to
421 	 * allocate mbufs or mbuf clusters during autoconfiguration.
422 	 */
423 	mbinit();
424 
425 	/* Initialize the sysctl subsystem. */
426 	sysctl_init();
427 
428 	/* Initialize I/O statistics. */
429 	iostat_init();
430 
431 	/* Initialize the log device. */
432 	loginit();
433 
434 	/* Start module system. */
435 	module_init();
436 
437 	/* Initialize the file systems. */
438 #ifdef NVNODE_IMPLICIT
439 	/*
440 	 * If maximum number of vnodes in namei vnode cache is not explicitly
441 	 * defined in kernel config, adjust the number such as we use roughly
442 	 * 10% of memory for vnodes and associated data structures in the
443 	 * assumed worst case.  Do not provide fewer than NVNODE vnodes.
444 	 */
445 	usevnodes =
446 	    calc_cache_size(kernel_map, 10, VNODE_VA_MAXPCT) / VNODE_COST;
447 	if (usevnodes > desiredvnodes)
448 		desiredvnodes = usevnodes;
449 #endif
450 	vfsinit();
451 	lf_init();
452 
453 	/* Initialize fstrans. */
454 	fstrans_init();
455 
456 	/* Initialize the file descriptor system. */
457 	fd_sys_init();
458 
459 	/* Initialize kqueue. */
460 	kqueue_init();
461 
462 	/* Initialize asynchronous I/O. */
463 	aio_sysinit();
464 
465 	/* Initialize message queues. */
466 	mqueue_sysinit();
467 
468 	/* Initialize the system monitor subsystems. */
469 #if NSYSMON_TASKQ > 0
470 	sysmon_task_queue_preinit();
471 #endif
472 
473 #if NSYSMON_ENVSYS > 0
474 	sysmon_envsys_init();
475 #endif
476 
477 #if NSYSMON_POWER > 0
478 	sysmon_power_init();
479 #endif
480 
481 #if NSYSMON_WDOG > 0
482 	sysmon_wdog_init();
483 #endif
484 
485 	inittimecounter();
486 	ntp_init();
487 
488 	/* Initialize the device switch tables. */
489 	devsw_init();
490 
491 	/* Initialize tty subsystem. */
492 	tty_init();
493 	ttyldisc_init();
494 
495 	/* Initialize the buffer cache, part 2. */
496 	bufinit2();
497 
498 	/* Initialize the disk wedge subsystem. */
499 	dkwedge_init();
500 
501 	/* Initialize interfaces. */
502 	ifinit1();
503 
504 	/* Configure the system hardware.  This will enable interrupts. */
505 	configure();
506 
507 	ubc_init();		/* must be after autoconfig */
508 
509 #ifdef SYSVSHM
510 	/* Initialize System V style shared memory. */
511 	shminit();
512 #endif
513 
514 #ifdef SYSVSEM
515 	/* Initialize System V style semaphores. */
516 	seminit();
517 #endif
518 
519 #ifdef SYSVMSG
520 	/* Initialize System V style message queues. */
521 	msginit();
522 #endif
523 
524 #ifdef P1003_1B_SEMAPHORE
525 	/* Initialize posix semaphores */
526 	ksem_init();
527 #endif
528 
529 #if NVERIEXEC > 0
530 	/*
531 	 * Initialise the Veriexec subsystem.
532 	 */
533 	veriexec_init();
534 #endif /* NVERIEXEC > 0 */
535 
536 #if defined(PAX_MPROTECT) || defined(PAX_SEGVGUARD) || defined(PAX_ASLR)
537 	pax_init();
538 #endif /* PAX_MPROTECT || PAX_SEGVGUARD || PAX_ASLR */
539 
540 #ifdef	FAST_IPSEC
541 	/* Attach network crypto subsystem */
542 	ipsec_attach();
543 #endif
544 
545 	/*
546 	 * Initialize protocols.  Block reception of incoming packets
547 	 * until everything is ready.
548 	 */
549 	s = splnet();
550 	ifinit();
551 	domaininit();
552 	if_attachdomain();
553 	splx(s);
554 
555 #ifdef GPROF
556 	/* Initialize kernel profiling. */
557 	kmstartup();
558 #endif
559 
560 	/* Initialize system accounting. */
561 	acct_init();
562 
563 #ifndef PIPE_SOCKETPAIR
564 	/* Initialize pipes. */
565 	pipe_init();
566 #endif
567 
568 #ifdef KTRACE
569 	/* Initialize ktrace. */
570 	ktrinit();
571 #endif
572 
573 	/* Initialize the UUID system calls. */
574 	uuid_init();
575 
576 #ifdef WAPBL
577 	/* Initialize write-ahead physical block logging. */
578 	wapbl_init();
579 #endif
580 
581 	/*
582 	 * Create process 1 (init(8)).  We do this now, as Unix has
583 	 * historically had init be process 1, and changing this would
584 	 * probably upset a lot of people.
585 	 *
586 	 * Note that process 1 won't immediately exec init(8), but will
587 	 * wait for us to inform it that the root file system has been
588 	 * mounted.
589 	 */
590 	if (fork1(l, 0, SIGCHLD, NULL, 0, start_init, NULL, NULL, &initproc))
591 		panic("fork init");
592 
593 	/*
594 	 * Load any remaining builtin modules, and hand back temporary
595 	 * storage to the VM system.
596 	 */
597 	module_init_class(MODULE_CLASS_ANY);
598 
599 	/*
600 	 * Finalize configuration now that all real devices have been
601 	 * found.  This needs to be done before the root device is
602 	 * selected, since finalization may create the root device.
603 	 */
604 	config_finalize();
605 
606 	/*
607 	 * Now that autoconfiguration has completed, we can determine
608 	 * the root and dump devices.
609 	 */
610 	cpu_rootconf();
611 	cpu_dumpconf();
612 
613 	/* Mount the root file system. */
614 	do {
615 		domountroothook();
616 		if ((error = vfs_mountroot())) {
617 			printf("cannot mount root, error = %d\n", error);
618 			boothowto |= RB_ASKNAME;
619 			setroot(root_device,
620 			    (rootdev != NODEV) ? DISKPART(rootdev) : 0);
621 		}
622 	} while (error != 0);
623 	mountroothook_destroy();
624 
625 	/*
626 	 * Initialise the time-of-day clock, passing the time recorded
627 	 * in the root filesystem (if any) for use by systems that
628 	 * don't have a non-volatile time-of-day device.
629 	 */
630 	inittodr(rootfstime);
631 
632 	CIRCLEQ_FIRST(&mountlist)->mnt_flag |= MNT_ROOTFS;
633 	CIRCLEQ_FIRST(&mountlist)->mnt_op->vfs_refcount++;
634 
635 	/*
636 	 * Get the vnode for '/'.  Set filedesc0.fd_fd.fd_cdir to
637 	 * reference it.
638 	 */
639 	error = VFS_ROOT(CIRCLEQ_FIRST(&mountlist), &rootvnode);
640 	if (error)
641 		panic("cannot find root vnode, error=%d", error);
642 	cwdi0.cwdi_cdir = rootvnode;
643 	VREF(cwdi0.cwdi_cdir);
644 	VOP_UNLOCK(rootvnode, 0);
645 	cwdi0.cwdi_rdir = NULL;
646 
647 	/*
648 	 * Now that root is mounted, we can fixup initproc's CWD
649 	 * info.  All other processes are kthreads, which merely
650 	 * share proc0's CWD info.
651 	 */
652 	initproc->p_cwdi->cwdi_cdir = rootvnode;
653 	VREF(initproc->p_cwdi->cwdi_cdir);
654 	initproc->p_cwdi->cwdi_rdir = NULL;
655 
656 	/*
657 	 * Now can look at time, having had a chance to verify the time
658 	 * from the file system.  Reset l->l_rtime as it may have been
659 	 * munched in mi_switch() after the time got set.
660 	 */
661 	getmicrotime(&time);
662 	boottime = time;
663 	mutex_enter(proc_lock);
664 	LIST_FOREACH(p, &allproc, p_list) {
665 		KASSERT((p->p_flag & PK_MARKER) == 0);
666 		mutex_enter(p->p_lock);
667 		p->p_stats->p_start = time;
668 		LIST_FOREACH(l, &p->p_lwps, l_sibling) {
669 			lwp_lock(l);
670 			memset(&l->l_rtime, 0, sizeof(l->l_rtime));
671 			lwp_unlock(l);
672 		}
673 		mutex_exit(p->p_lock);
674 	}
675 	mutex_exit(proc_lock);
676 	binuptime(&curlwp->l_stime);
677 
678 	for (CPU_INFO_FOREACH(cii, ci)) {
679 		ci->ci_schedstate.spc_lastmod = time_second;
680 	}
681 
682 	/* Create the pageout daemon kernel thread. */
683 	uvm_swap_init();
684 	if (kthread_create(PRI_PGDAEMON, KTHREAD_MPSAFE, NULL, uvm_pageout,
685 	    NULL, NULL, "pgdaemon"))
686 		panic("fork pagedaemon");
687 
688 	/* Create the filesystem syncer kernel thread. */
689 	if (kthread_create(PRI_IOFLUSH, KTHREAD_MPSAFE, NULL, sched_sync,
690 	    NULL, NULL, "ioflush"))
691 		panic("fork syncer");
692 
693 	/* Create the aiodone daemon kernel thread. */
694 	if (workqueue_create(&uvm.aiodone_queue, "aiodoned",
695 	    uvm_aiodone_worker, NULL, PRI_VM, IPL_NONE, WQ_MPSAFE))
696 		panic("fork aiodoned");
697 
698 	vmem_rehash_start();
699 
700 	/* Initialize exec structures */
701 	exec_init(1);
702 
703 	/*
704 	 * Okay, now we can let init(8) exec!  It's off to userland!
705 	 */
706 	mutex_enter(proc_lock);
707 	start_init_exec = 1;
708 	cv_broadcast(&lbolt);
709 	mutex_exit(proc_lock);
710 
711 	/* The scheduler is an infinite loop. */
712 	uvm_scheduler();
713 	/* NOTREACHED */
714 }
715 
716 static void
717 check_console(struct lwp *l)
718 {
719 	struct nameidata nd;
720 	int error;
721 
722 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_SYSSPACE, "/dev/console");
723 	error = namei(&nd);
724 	if (error == 0)
725 		vrele(nd.ni_vp);
726 	else if (error == ENOENT)
727 		printf("warning: no /dev/console\n");
728 	else
729 		printf("warning: lookup /dev/console: error %d\n", error);
730 }
731 
732 /*
733  * List of paths to try when searching for "init".
734  */
735 static const char * const initpaths[] = {
736 	"/sbin/init",
737 	"/sbin/oinit",
738 	"/sbin/init.bak",
739 	NULL,
740 };
741 
742 /*
743  * Start the initial user process; try exec'ing each pathname in "initpaths".
744  * The program is invoked with one argument containing the boot flags.
745  */
746 static void
747 start_init(void *arg)
748 {
749 	struct lwp *l = arg;
750 	struct proc *p = l->l_proc;
751 	vaddr_t addr;
752 	struct sys_execve_args /* {
753 		syscallarg(const char *) path;
754 		syscallarg(char * const *) argp;
755 		syscallarg(char * const *) envp;
756 	} */ args;
757 	int options, i, error;
758 	register_t retval[2];
759 	char flags[4], *flagsp;
760 	const char *path, *slash;
761 	char *ucp, **uap, *arg0, *arg1 = NULL;
762 	char ipath[129];
763 	int ipx, len;
764 
765 	/*
766 	 * Now in process 1.
767 	 */
768 	strncpy(p->p_comm, "init", MAXCOMLEN);
769 
770 	/*
771 	 * Wait for main() to tell us that it's safe to exec.
772 	 */
773 	mutex_enter(proc_lock);
774 	while (start_init_exec == 0)
775 		cv_wait(&lbolt, proc_lock);
776 	mutex_exit(proc_lock);
777 
778 	/*
779 	 * This is not the right way to do this.  We really should
780 	 * hand-craft a descriptor onto /dev/console to hand to init,
781 	 * but that's a _lot_ more work, and the benefit from this easy
782 	 * hack makes up for the "good is the enemy of the best" effect.
783 	 */
784 	check_console(l);
785 
786 	/*
787 	 * Need just enough stack to hold the faked-up "execve()" arguments.
788 	 */
789 	addr = (vaddr_t)STACK_ALLOC(USRSTACK, PAGE_SIZE);
790 	if (uvm_map(&p->p_vmspace->vm_map, &addr, PAGE_SIZE,
791                     NULL, UVM_UNKNOWN_OFFSET, 0,
792                     UVM_MAPFLAG(UVM_PROT_ALL, UVM_PROT_ALL, UVM_INH_COPY,
793 		    UVM_ADV_NORMAL,
794                     UVM_FLAG_FIXED|UVM_FLAG_OVERLAY|UVM_FLAG_COPYONW)) != 0)
795 		panic("init: couldn't allocate argument space");
796 	p->p_vmspace->vm_maxsaddr = (void *)STACK_MAX(addr, PAGE_SIZE);
797 
798 	ipx = 0;
799 	while (1) {
800 		if (boothowto & RB_ASKNAME) {
801 			printf("init path");
802 			if (initpaths[ipx])
803 				printf(" (default %s)", initpaths[ipx]);
804 			printf(": ");
805 			len = cngetsn(ipath, sizeof(ipath)-1);
806 			if (len == 4 && strcmp(ipath, "halt") == 0) {
807 				cpu_reboot(RB_HALT, NULL);
808 			} else if (len == 6 && strcmp(ipath, "reboot") == 0) {
809 				cpu_reboot(0, NULL);
810 #if defined(DDB)
811 			} else if (len == 3 && strcmp(ipath, "ddb") == 0) {
812 				console_debugger();
813 				continue;
814 #endif
815 			} else if (len > 0 && ipath[0] == '/') {
816 				ipath[len] = '\0';
817 				path = ipath;
818 			} else if (len == 0 && initpaths[ipx] != NULL) {
819 				path = initpaths[ipx++];
820 			} else {
821 				printf("use absolute path, ");
822 #if defined(DDB)
823 				printf("\"ddb\", ");
824 #endif
825 				printf("\"halt\", or \"reboot\"\n");
826 				continue;
827 			}
828 		} else {
829 			if ((path = initpaths[ipx++]) == NULL) {
830 				ipx = 0;
831 				boothowto |= RB_ASKNAME;
832 				continue;
833 			}
834 		}
835 
836 		ucp = (char *)USRSTACK;
837 
838 		/*
839 		 * Construct the boot flag argument.
840 		 */
841 		flagsp = flags;
842 		*flagsp++ = '-';
843 		options = 0;
844 
845 		if (boothowto & RB_SINGLE) {
846 			*flagsp++ = 's';
847 			options = 1;
848 		}
849 #ifdef notyet
850 		if (boothowto & RB_FASTBOOT) {
851 			*flagsp++ = 'f';
852 			options = 1;
853 		}
854 #endif
855 
856 		/*
857 		 * Move out the flags (arg 1), if necessary.
858 		 */
859 		if (options != 0) {
860 			*flagsp++ = '\0';
861 			i = flagsp - flags;
862 #ifdef DEBUG
863 			printf("init: copying out flags `%s' %d\n", flags, i);
864 #endif
865 			arg1 = STACK_ALLOC(ucp, i);
866 			ucp = STACK_MAX(arg1, i);
867 			(void)copyout((void *)flags, arg1, i);
868 		}
869 
870 		/*
871 		 * Move out the file name (also arg 0).
872 		 */
873 		i = strlen(path) + 1;
874 #ifdef DEBUG
875 		printf("init: copying out path `%s' %d\n", path, i);
876 #else
877 		if (boothowto & RB_ASKNAME || path != initpaths[0])
878 			printf("init: trying %s\n", path);
879 #endif
880 		arg0 = STACK_ALLOC(ucp, i);
881 		ucp = STACK_MAX(arg0, i);
882 		(void)copyout(path, arg0, i);
883 
884 		/*
885 		 * Move out the arg pointers.
886 		 */
887 		ucp = (void *)STACK_ALIGN(ucp, ALIGNBYTES);
888 		uap = (char **)STACK_ALLOC(ucp, sizeof(char *) * 3);
889 		SCARG(&args, path) = arg0;
890 		SCARG(&args, argp) = uap;
891 		SCARG(&args, envp) = NULL;
892 		slash = strrchr(path, '/');
893 		if (slash)
894 			(void)suword((void *)uap++,
895 			    (long)arg0 + (slash + 1 - path));
896 		else
897 			(void)suword((void *)uap++, (long)arg0);
898 		if (options != 0)
899 			(void)suword((void *)uap++, (long)arg1);
900 		(void)suword((void *)uap++, 0);	/* terminator */
901 
902 		/*
903 		 * Now try to exec the program.  If can't for any reason
904 		 * other than it doesn't exist, complain.
905 		 */
906 		error = sys_execve(l, &args, retval);
907 		if (error == 0 || error == EJUSTRETURN) {
908 			KERNEL_UNLOCK_LAST(l);
909 			return;
910 		}
911 		printf("exec %s: error %d\n", path, error);
912 	}
913 	printf("init: not found\n");
914 	panic("no init");
915 }
916 
917 /*
918  * calculate cache size from physmem and vm_map size.
919  */
920 vaddr_t
921 calc_cache_size(struct vm_map *map, int pct, int va_pct)
922 {
923 	paddr_t t;
924 
925 	/* XXX should consider competing cache if any */
926 	/* XXX should consider submaps */
927 	t = (uintmax_t)physmem * pct / 100 * PAGE_SIZE;
928 	if (map != NULL) {
929 		vsize_t vsize;
930 
931 		vsize = vm_map_max(map) - vm_map_min(map);
932 		vsize = (uintmax_t)vsize * va_pct / 100;
933 		if (t > vsize) {
934 			t = vsize;
935 		}
936 	}
937 	return t;
938 }
939