xref: /netbsd-src/sys/kern/init_main.c (revision 4b71a66d0f279143147d63ebfcfd8a59499a3684)
1 /*	$NetBSD: init_main.c,v 1.357 2008/05/28 13:35:32 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 1.357 2008/05/28 13:35:32 ad Exp $");
101 
102 #include "opt_ipsec.h"
103 #include "opt_ntp.h"
104 #include "opt_pipe.h"
105 #include "opt_posix.h"
106 #include "opt_syscall_debug.h"
107 #include "opt_sysv.h"
108 #include "opt_fileassoc.h"
109 #include "opt_ktrace.h"
110 #include "opt_pax.h"
111 
112 #include "rnd.h"
113 #include "sysmon_envsys.h"
114 #include "sysmon_power.h"
115 #include "sysmon_taskq.h"
116 #include "sysmon_wdog.h"
117 #include "veriexec.h"
118 
119 #include <sys/param.h>
120 #include <sys/acct.h>
121 #include <sys/filedesc.h>
122 #include <sys/file.h>
123 #include <sys/errno.h>
124 #include <sys/callout.h>
125 #include <sys/cpu.h>
126 #include <sys/kernel.h>
127 #include <sys/kmem.h>
128 #include <sys/mount.h>
129 #include <sys/proc.h>
130 #include <sys/kthread.h>
131 #include <sys/resourcevar.h>
132 #include <sys/signalvar.h>
133 #include <sys/systm.h>
134 #include <sys/vnode.h>
135 #include <sys/fstrans.h>
136 #include <sys/tty.h>
137 #include <sys/conf.h>
138 #include <sys/disklabel.h>
139 #include <sys/buf.h>
140 #include <sys/device.h>
141 #include <sys/exec.h>
142 #include <sys/socketvar.h>
143 #include <sys/protosw.h>
144 #include <sys/percpu.h>
145 #include <sys/pset.h>
146 #include <sys/sysctl.h>
147 #include <sys/reboot.h>
148 #include <sys/user.h>
149 #include <sys/sysctl.h>
150 #include <sys/event.h>
151 #include <sys/mbuf.h>
152 #include <sys/sched.h>
153 #include <sys/sleepq.h>
154 #include <sys/iostat.h>
155 #include <sys/vmem.h>
156 #include <sys/uuid.h>
157 #include <sys/extent.h>
158 #include <sys/disk.h>
159 #include <sys/mqueue.h>
160 #include <sys/msgbuf.h>
161 #include <sys/module.h>
162 #include <sys/event.h>
163 #include <sys/lockf.h>
164 #ifdef FAST_IPSEC
165 #include <netipsec/ipsec.h>
166 #endif
167 #ifdef SYSVSHM
168 #include <sys/shm.h>
169 #endif
170 #ifdef SYSVSEM
171 #include <sys/sem.h>
172 #endif
173 #ifdef SYSVMSG
174 #include <sys/msg.h>
175 #endif
176 #ifdef P1003_1B_SEMAPHORE
177 #include <sys/ksem.h>
178 #endif
179 #include <sys/domain.h>
180 #include <sys/namei.h>
181 #if NRND > 0
182 #include <sys/rnd.h>
183 #endif
184 #include <sys/pipe.h>
185 #ifdef LKM
186 #include <sys/lkm.h>
187 #endif
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 #include <net80211/ieee80211_netbsd.h>
196 
197 #include <sys/syscall.h>
198 #include <sys/syscallargs.h>
199 
200 #if defined(PAX_MPROTECT) || defined(PAX_SEGVGUARD) || defined(PAX_ASLR)
201 #include <sys/pax.h>
202 #endif /* PAX_MPROTECT || PAX_SEGVGUARD || PAX_ASLR */
203 
204 #include <ufs/ufs/quota.h>
205 
206 #include <miscfs/genfs/genfs.h>
207 #include <miscfs/syncfs/syncfs.h>
208 
209 #include <sys/cpu.h>
210 
211 #include <uvm/uvm.h>
212 
213 #if NSYSMON_TASKQ > 0
214 #include <dev/sysmon/sysmon_taskq.h>
215 #endif
216 
217 #include <dev/cons.h>
218 
219 #if NSYSMON_ENVSYS > 0 || NSYSMON_POWER > 0 || NSYSMON_WDOG > 0
220 #include <dev/sysmon/sysmonvar.h>
221 #endif
222 
223 #include <net/if.h>
224 #include <net/raw_cb.h>
225 
226 #include <secmodel/secmodel.h>
227 
228 extern struct proc proc0;
229 extern struct lwp lwp0;
230 extern struct cwdinfo cwdi0;
231 extern time_t rootfstime;
232 
233 #ifndef curlwp
234 struct	lwp *curlwp = &lwp0;
235 #endif
236 struct	proc *initproc;
237 
238 struct	vnode *rootvp, *swapdev_vp;
239 int	boothowto;
240 int	cold = 1;			/* still working on startup */
241 struct timeval boottime;	        /* time at system startup - will only follow settime deltas */
242 
243 int	start_init_exec;		/* semaphore for start_init() */
244 
245 static void check_console(struct lwp *l);
246 static void start_init(void *);
247 void main(void);
248 void ssp_init(void);
249 
250 #if defined(__SSP__) || defined(__SSP_ALL__)
251 long __stack_chk_guard[8] = {0, 0, 0, 0, 0, 0, 0, 0};
252 void __stack_chk_fail(void);
253 
254 void
255 __stack_chk_fail(void)
256 {
257 	panic("stack overflow detected; terminated");
258 }
259 
260 void
261 ssp_init(void)
262 {
263 	int s;
264 
265 #ifdef DIAGNOSTIC
266 	printf("Initializing SSP:");
267 #endif
268 	/*
269 	 * We initialize ssp here carefully:
270 	 *	1. after we got some entropy
271 	 *	2. without calling a function
272 	 */
273 	size_t i;
274 	long guard[__arraycount(__stack_chk_guard)];
275 
276 	arc4randbytes(guard, sizeof(guard));
277 	s = splhigh();
278 	for (i = 0; i < __arraycount(guard); i++)
279 		__stack_chk_guard[i] = guard[i];
280 	splx(s);
281 #ifdef DIAGNOSTIC
282 	for (i = 0; i < __arraycount(guard); i++)
283 		printf("%lx ", guard[i]);
284 	printf("\n");
285 #endif
286 }
287 #else
288 void
289 ssp_init(void)
290 {
291 
292 }
293 #endif
294 
295 void __secmodel_none(void);
296 __weak_alias(secmodel_start,__secmodel_none);
297 void
298 __secmodel_none(void)
299 {
300 	return;
301 }
302 
303 /*
304  * System startup; initialize the world, create process 0, mount root
305  * filesystem, and fork to create init and pagedaemon.  Most of the
306  * hard work is done in the lower-level initialization routines including
307  * startup(), which does memory initialization and autoconfiguration.
308  */
309 void
310 main(void)
311 {
312 	struct timeval time;
313 	struct lwp *l;
314 	struct proc *p;
315 	int s, error;
316 #ifdef NVNODE_IMPLICIT
317 	int usevnodes;
318 #endif
319 	CPU_INFO_ITERATOR cii;
320 	struct cpu_info *ci;
321 
322 	l = &lwp0;
323 #ifndef LWP0_CPU_INFO
324 	l->l_cpu = curcpu();
325 #endif
326 
327 	/*
328 	 * Attempt to find console and initialize
329 	 * in case of early panic or other messages.
330 	 */
331 	consinit();
332 
333 	kernel_lock_init();
334 
335 	uvm_init();
336 
337 	kmem_init();
338 
339 	percpu_init();
340 
341 	/* Initialize lock caches. */
342 	mutex_obj_init();
343 
344 	/* Initialize the extent manager. */
345 	extent_init();
346 
347 	/* Do machine-dependent initialization. */
348 	cpu_startup();
349 
350 	/* Initialize callouts, part 1. */
351 	callout_startup();
352 
353 	/*
354 	 * Initialize the kernel authorization subsystem and start the
355 	 * default security model, if any. We need to do this early
356 	 * enough so that subsystems relying on any of the aforementioned
357 	 * can work properly. Since the security model may dictate the
358 	 * credential inheritance policy, it is needed at least before
359 	 * any process is created, specifically proc0.
360 	 */
361 	kauth_init();
362 	secmodel_start();
363 
364 	/* Initialize the buffer cache */
365 	bufinit();
366 
367 	/* Initialize sockets. */
368 	soinit();
369 
370 	/*
371 	 * The following things must be done before autoconfiguration.
372 	 */
373 	evcnt_init();		/* initialize event counters */
374 #if NRND > 0
375 	rnd_init();		/* initialize RNG */
376 #endif
377 
378 	/* Initialize process and pgrp structures. */
379 	procinit();
380 	lwpinit();
381 
382 	/* Initialize signal-related data structures. */
383 	signal_init();
384 
385 	/* Initialize resource management. */
386 	resource_init();
387 
388 	/* Create process 0 (the swapper). */
389 	proc0_init();
390 
391 	/* Initialize the UID hash table. */
392 	uid_init();
393 
394 	/* Charge root for one process. */
395 	(void)chgproccnt(0, 1);
396 
397 	/* Initialize timekeeping. */
398 	time_init();
399 
400 	/* Initialize the run queues, turnstiles and sleep queues. */
401 	mutex_init(&cpu_lock, MUTEX_DEFAULT, IPL_NONE);
402 	sched_rqinit();
403 	turnstile_init();
404 	sleeptab_init(&sleeptab);
405 
406 	/* Initialize processor-sets */
407 	psets_init();
408 
409 	/* MI initialization of the boot cpu */
410 	error = mi_cpu_attach(curcpu());
411 	KASSERT(error == 0);
412 
413 	/* Initialize timekeeping, part 2. */
414 	time_init2();
415 
416 	/*
417 	 * Initialize mbuf's.  Do this now because we might attempt to
418 	 * allocate mbufs or mbuf clusters during autoconfiguration.
419 	 */
420 	mbinit();
421 
422 	/* Initialize the sysctl subsystem. */
423 	sysctl_init();
424 
425 	/* Initialize I/O statistics. */
426 	iostat_init();
427 
428 	/* Initialize the log device. */
429 	loginit();
430 
431 	/* Start module system. */
432 	module_init();
433 
434 	/* Initialize the file systems. */
435 #ifdef NVNODE_IMPLICIT
436 	/*
437 	 * If maximum number of vnodes in namei vnode cache is not explicitly
438 	 * defined in kernel config, adjust the number such as we use roughly
439 	 * 1.0% of memory for vnode cache (but not less than NVNODE vnodes).
440 	 */
441 	usevnodes =
442 	    calc_cache_size(kernel_map, 1, VNODE_VA_MAXPCT) / sizeof(vnode_t);
443 	if (usevnodes > desiredvnodes)
444 		desiredvnodes = usevnodes;
445 #endif
446 	vfsinit();
447 	lf_init();
448 
449 	/* Initialize fstrans. */
450 	fstrans_init();
451 
452 	/* Initialize the file descriptor system. */
453 	fd_sys_init();
454 
455 	/* Initialize kqueue. */
456 	kqueue_init();
457 
458 	/* Initialize asynchronous I/O. */
459 	aio_sysinit();
460 
461 	/* Initialize message queues. */
462 	mqueue_sysinit();
463 
464 	/* Initialize the system monitor subsystems. */
465 #if NSYSMON_TASKQ > 0
466 	sysmon_task_queue_preinit();
467 #endif
468 
469 #if NSYSMON_ENVSYS > 0
470 	sysmon_envsys_init();
471 #endif
472 
473 #if NSYSMON_POWER > 0
474 	sysmon_power_init();
475 #endif
476 
477 #if NSYSMON_WDOG > 0
478 	sysmon_wdog_init();
479 #endif
480 
481 	inittimecounter();
482 	ntp_init();
483 
484 	/* Initialize the device switch tables. */
485 	devsw_init();
486 
487 	/* Initialize tty subsystem. */
488 	tty_init();
489 	ttyldisc_init();
490 
491 	/* Initialize the buffer cache, part 2. */
492 	bufinit2();
493 
494 	/* Initialize the disk wedge subsystem. */
495 	dkwedge_init();
496 
497 	/* Configure the system hardware.  This will enable interrupts. */
498 	configure();
499 
500 	ubc_init();		/* must be after autoconfig */
501 
502 #ifdef SYSVSHM
503 	/* Initialize System V style shared memory. */
504 	shminit();
505 #endif
506 
507 #ifdef SYSVSEM
508 	/* Initialize System V style semaphores. */
509 	seminit();
510 #endif
511 
512 #ifdef SYSVMSG
513 	/* Initialize System V style message queues. */
514 	msginit();
515 #endif
516 
517 #ifdef P1003_1B_SEMAPHORE
518 	/* Initialize posix semaphores */
519 	ksem_init();
520 #endif
521 
522 #if NVERIEXEC > 0
523 	/*
524 	 * Initialise the Veriexec subsystem.
525 	 */
526 	veriexec_init();
527 #endif /* NVERIEXEC > 0 */
528 
529 #if defined(PAX_MPROTECT) || defined(PAX_SEGVGUARD) || defined(PAX_ASLR)
530 	pax_init();
531 #endif /* PAX_MPROTECT || PAX_SEGVGUARD || PAX_ASLR */
532 
533 #ifdef	FAST_IPSEC
534 	/* Attach network crypto subsystem */
535 	ipsec_attach();
536 #endif
537 
538 	/*
539 	 * Initialize protocols.  Block reception of incoming packets
540 	 * until everything is ready.
541 	 */
542 	s = splnet();
543 	ifinit();
544 	domaininit();
545 	if_attachdomain();
546 	splx(s);
547 
548 #ifdef GPROF
549 	/* Initialize kernel profiling. */
550 	kmstartup();
551 #endif
552 
553 	/* Initialize system accounting. */
554 	acct_init();
555 
556 #ifndef PIPE_SOCKETPAIR
557 	/* Initialize pipes. */
558 	pipe_init();
559 #endif
560 
561 #ifdef KTRACE
562 	/* Initialize ktrace. */
563 	ktrinit();
564 #endif
565 
566 	/* Initialize the UUID system calls. */
567 	uuid_init();
568 
569 	/*
570 	 * Create process 1 (init(8)).  We do this now, as Unix has
571 	 * historically had init be process 1, and changing this would
572 	 * probably upset a lot of people.
573 	 *
574 	 * Note that process 1 won't immediately exec init(8), but will
575 	 * wait for us to inform it that the root file system has been
576 	 * mounted.
577 	 */
578 	if (fork1(l, 0, SIGCHLD, NULL, 0, start_init, NULL, NULL, &initproc))
579 		panic("fork init");
580 
581 	/*
582 	 * Load any remaining builtin modules, and hand back temporary
583 	 * storage to the VM system.
584 	 */
585 	module_init_class(MODULE_CLASS_ANY);
586 	module_jettison();
587 
588 	/*
589 	 * Finalize configuration now that all real devices have been
590 	 * found.  This needs to be done before the root device is
591 	 * selected, since finalization may create the root device.
592 	 */
593 	config_finalize();
594 
595 	/*
596 	 * Now that autoconfiguration has completed, we can determine
597 	 * the root and dump devices.
598 	 */
599 	cpu_rootconf();
600 	cpu_dumpconf();
601 
602 	/* Mount the root file system. */
603 	do {
604 		domountroothook();
605 		if ((error = vfs_mountroot())) {
606 			printf("cannot mount root, error = %d\n", error);
607 			boothowto |= RB_ASKNAME;
608 			setroot(root_device,
609 			    (rootdev != NODEV) ? DISKPART(rootdev) : 0);
610 		}
611 	} while (error != 0);
612 	mountroothook_destroy();
613 
614 	/*
615 	 * Initialise the time-of-day clock, passing the time recorded
616 	 * in the root filesystem (if any) for use by systems that
617 	 * don't have a non-volatile time-of-day device.
618 	 */
619 	inittodr(rootfstime);
620 
621 	CIRCLEQ_FIRST(&mountlist)->mnt_flag |= MNT_ROOTFS;
622 	CIRCLEQ_FIRST(&mountlist)->mnt_op->vfs_refcount++;
623 
624 	/*
625 	 * Get the vnode for '/'.  Set filedesc0.fd_fd.fd_cdir to
626 	 * reference it.
627 	 */
628 	error = VFS_ROOT(CIRCLEQ_FIRST(&mountlist), &rootvnode);
629 	if (error)
630 		panic("cannot find root vnode, error=%d", error);
631 	cwdi0.cwdi_cdir = rootvnode;
632 	VREF(cwdi0.cwdi_cdir);
633 	VOP_UNLOCK(rootvnode, 0);
634 	cwdi0.cwdi_rdir = NULL;
635 
636 	/*
637 	 * Now that root is mounted, we can fixup initproc's CWD
638 	 * info.  All other processes are kthreads, which merely
639 	 * share proc0's CWD info.
640 	 */
641 	initproc->p_cwdi->cwdi_cdir = rootvnode;
642 	VREF(initproc->p_cwdi->cwdi_cdir);
643 	initproc->p_cwdi->cwdi_rdir = NULL;
644 
645 	/*
646 	 * Now can look at time, having had a chance to verify the time
647 	 * from the file system.  Reset l->l_rtime as it may have been
648 	 * munched in mi_switch() after the time got set.
649 	 */
650 	getmicrotime(&time);
651 	boottime = time;
652 	mutex_enter(proc_lock);
653 	LIST_FOREACH(p, &allproc, p_list) {
654 		KASSERT((p->p_flag & PK_MARKER) == 0);
655 		mutex_enter(p->p_lock);
656 		p->p_stats->p_start = time;
657 		LIST_FOREACH(l, &p->p_lwps, l_sibling) {
658 			lwp_lock(l);
659 			memset(&l->l_rtime, 0, sizeof(l->l_rtime));
660 			lwp_unlock(l);
661 		}
662 		mutex_exit(p->p_lock);
663 	}
664 	mutex_exit(proc_lock);
665 	binuptime(&curlwp->l_stime);
666 
667 	for (CPU_INFO_FOREACH(cii, ci)) {
668 		ci->ci_schedstate.spc_lastmod = time_second;
669 	}
670 
671 	/* Create the pageout daemon kernel thread. */
672 	uvm_swap_init();
673 	if (kthread_create(PRI_PGDAEMON, KTHREAD_MPSAFE, NULL, uvm_pageout,
674 	    NULL, NULL, "pgdaemon"))
675 		panic("fork pagedaemon");
676 
677 	/* Create the filesystem syncer kernel thread. */
678 	if (kthread_create(PRI_IOFLUSH, KTHREAD_MPSAFE, NULL, sched_sync,
679 	    NULL, NULL, "ioflush"))
680 		panic("fork syncer");
681 
682 	/* Create the aiodone daemon kernel thread. */
683 	if (workqueue_create(&uvm.aiodone_queue, "aiodoned",
684 	    uvm_aiodone_worker, NULL, PRI_VM, IPL_NONE, WQ_MPSAFE))
685 		panic("fork aiodoned");
686 
687 	vmem_rehash_start();
688 
689 	/* Initialize exec structures */
690 	exec_init(1);
691 
692 	/*
693 	 * Okay, now we can let init(8) exec!  It's off to userland!
694 	 */
695 	mutex_enter(proc_lock);
696 	start_init_exec = 1;
697 	cv_broadcast(&lbolt);
698 	mutex_exit(proc_lock);
699 
700 	/* The scheduler is an infinite loop. */
701 	uvm_scheduler();
702 	/* NOTREACHED */
703 }
704 
705 static void
706 check_console(struct lwp *l)
707 {
708 	struct nameidata nd;
709 	int error;
710 
711 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_SYSSPACE, "/dev/console");
712 	error = namei(&nd);
713 	if (error == 0)
714 		vrele(nd.ni_vp);
715 	else if (error == ENOENT)
716 		printf("warning: no /dev/console\n");
717 	else
718 		printf("warning: lookup /dev/console: error %d\n", error);
719 }
720 
721 /*
722  * List of paths to try when searching for "init".
723  */
724 static const char * const initpaths[] = {
725 	"/sbin/init",
726 	"/sbin/oinit",
727 	"/sbin/init.bak",
728 	NULL,
729 };
730 
731 /*
732  * Start the initial user process; try exec'ing each pathname in "initpaths".
733  * The program is invoked with one argument containing the boot flags.
734  */
735 static void
736 start_init(void *arg)
737 {
738 	struct lwp *l = arg;
739 	struct proc *p = l->l_proc;
740 	vaddr_t addr;
741 	struct sys_execve_args /* {
742 		syscallarg(const char *) path;
743 		syscallarg(char * const *) argp;
744 		syscallarg(char * const *) envp;
745 	} */ args;
746 	int options, i, error;
747 	register_t retval[2];
748 	char flags[4], *flagsp;
749 	const char *path, *slash;
750 	char *ucp, **uap, *arg0, *arg1 = NULL;
751 	char ipath[129];
752 	int ipx, len;
753 
754 	/*
755 	 * Now in process 1.
756 	 */
757 	strncpy(p->p_comm, "init", MAXCOMLEN);
758 
759 	/*
760 	 * Wait for main() to tell us that it's safe to exec.
761 	 */
762 	mutex_enter(proc_lock);
763 	while (start_init_exec == 0)
764 		cv_wait(&lbolt, proc_lock);
765 	mutex_exit(proc_lock);
766 
767 	/*
768 	 * This is not the right way to do this.  We really should
769 	 * hand-craft a descriptor onto /dev/console to hand to init,
770 	 * but that's a _lot_ more work, and the benefit from this easy
771 	 * hack makes up for the "good is the enemy of the best" effect.
772 	 */
773 	check_console(l);
774 
775 	/*
776 	 * Need just enough stack to hold the faked-up "execve()" arguments.
777 	 */
778 	addr = (vaddr_t)STACK_ALLOC(USRSTACK, PAGE_SIZE);
779 	if (uvm_map(&p->p_vmspace->vm_map, &addr, PAGE_SIZE,
780                     NULL, UVM_UNKNOWN_OFFSET, 0,
781                     UVM_MAPFLAG(UVM_PROT_ALL, UVM_PROT_ALL, UVM_INH_COPY,
782 		    UVM_ADV_NORMAL,
783                     UVM_FLAG_FIXED|UVM_FLAG_OVERLAY|UVM_FLAG_COPYONW)) != 0)
784 		panic("init: couldn't allocate argument space");
785 	p->p_vmspace->vm_maxsaddr = (void *)STACK_MAX(addr, PAGE_SIZE);
786 
787 	ipx = 0;
788 	while (1) {
789 		if (boothowto & RB_ASKNAME) {
790 			printf("init path");
791 			if (initpaths[ipx])
792 				printf(" (default %s)", initpaths[ipx]);
793 			printf(": ");
794 			len = cngetsn(ipath, sizeof(ipath)-1);
795 			if (len == 0) {
796 				if (initpaths[ipx])
797 					path = initpaths[ipx++];
798 				else
799 					continue;
800 			} else {
801 				ipath[len] = '\0';
802 				path = ipath;
803 			}
804 		} else {
805 			if ((path = initpaths[ipx++]) == NULL)
806 				break;
807 		}
808 
809 		ucp = (char *)USRSTACK;
810 
811 		/*
812 		 * Construct the boot flag argument.
813 		 */
814 		flagsp = flags;
815 		*flagsp++ = '-';
816 		options = 0;
817 
818 		if (boothowto & RB_SINGLE) {
819 			*flagsp++ = 's';
820 			options = 1;
821 		}
822 #ifdef notyet
823 		if (boothowto & RB_FASTBOOT) {
824 			*flagsp++ = 'f';
825 			options = 1;
826 		}
827 #endif
828 
829 		/*
830 		 * Move out the flags (arg 1), if necessary.
831 		 */
832 		if (options != 0) {
833 			*flagsp++ = '\0';
834 			i = flagsp - flags;
835 #ifdef DEBUG
836 			printf("init: copying out flags `%s' %d\n", flags, i);
837 #endif
838 			arg1 = STACK_ALLOC(ucp, i);
839 			ucp = STACK_MAX(arg1, i);
840 			(void)copyout((void *)flags, arg1, i);
841 		}
842 
843 		/*
844 		 * Move out the file name (also arg 0).
845 		 */
846 		i = strlen(path) + 1;
847 #ifdef DEBUG
848 		printf("init: copying out path `%s' %d\n", path, i);
849 #else
850 		if (boothowto & RB_ASKNAME || path != initpaths[0])
851 			printf("init: trying %s\n", path);
852 #endif
853 		arg0 = STACK_ALLOC(ucp, i);
854 		ucp = STACK_MAX(arg0, i);
855 		(void)copyout(path, arg0, i);
856 
857 		/*
858 		 * Move out the arg pointers.
859 		 */
860 		ucp = (void *)STACK_ALIGN(ucp, ALIGNBYTES);
861 		uap = (char **)STACK_ALLOC(ucp, sizeof(char *) * 3);
862 		SCARG(&args, path) = arg0;
863 		SCARG(&args, argp) = uap;
864 		SCARG(&args, envp) = NULL;
865 		slash = strrchr(path, '/');
866 		if (slash)
867 			(void)suword((void *)uap++,
868 			    (long)arg0 + (slash + 1 - path));
869 		else
870 			(void)suword((void *)uap++, (long)arg0);
871 		if (options != 0)
872 			(void)suword((void *)uap++, (long)arg1);
873 		(void)suword((void *)uap++, 0);	/* terminator */
874 
875 		/*
876 		 * Now try to exec the program.  If can't for any reason
877 		 * other than it doesn't exist, complain.
878 		 */
879 		error = sys_execve(l, &args, retval);
880 		if (error == 0 || error == EJUSTRETURN) {
881 			KERNEL_UNLOCK_LAST(l);
882 			return;
883 		}
884 		printf("exec %s: error %d\n", path, error);
885 	}
886 	printf("init: not found\n");
887 	panic("no init");
888 }
889 
890 /*
891  * calculate cache size from physmem and vm_map size.
892  */
893 vaddr_t
894 calc_cache_size(struct vm_map *map, int pct, int va_pct)
895 {
896 	paddr_t t;
897 
898 	/* XXX should consider competing cache if any */
899 	/* XXX should consider submaps */
900 	t = (uintmax_t)physmem * pct / 100 * PAGE_SIZE;
901 	if (map != NULL) {
902 		vsize_t vsize;
903 
904 		vsize = vm_map_max(map) - vm_map_min(map);
905 		vsize = (uintmax_t)vsize * va_pct / 100;
906 		if (t > vsize) {
907 			t = vsize;
908 		}
909 	}
910 	return t;
911 }
912