xref: /netbsd-src/sys/kern/init_main.c (revision bd1f9e680783606ba327ccadf6002a15ffbdf514)
1 /*	$NetBSD: init_main.c,v 1.407 2009/10/26 19:03:17 rmind Exp $	*/
2 
3 /*-
4  * Copyright (c) 2008, 2009 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 1.407 2009/10/26 19:03:17 rmind Exp $");
101 
102 #include "opt_ddb.h"
103 #include "opt_ipsec.h"
104 #include "opt_modular.h"
105 #include "opt_ntp.h"
106 #include "opt_pipe.h"
107 #include "opt_sa.h"
108 #include "opt_syscall_debug.h"
109 #include "opt_sysv.h"
110 #include "opt_fileassoc.h"
111 #include "opt_ktrace.h"
112 #include "opt_pax.h"
113 #include "opt_compat_netbsd.h"
114 #include "opt_wapbl.h"
115 #include "opt_ptrace.h"
116 
117 #include "drvctl.h"
118 #include "ksyms.h"
119 #include "rnd.h"
120 #include "sysmon_envsys.h"
121 #include "sysmon_power.h"
122 #include "sysmon_taskq.h"
123 #include "sysmon_wdog.h"
124 #include "veriexec.h"
125 
126 #include <sys/param.h>
127 #include <sys/acct.h>
128 #include <sys/filedesc.h>
129 #include <sys/file.h>
130 #include <sys/errno.h>
131 #include <sys/callout.h>
132 #include <sys/cpu.h>
133 #include <sys/kernel.h>
134 #include <sys/mount.h>
135 #include <sys/proc.h>
136 #include <sys/kthread.h>
137 #include <sys/resourcevar.h>
138 #include <sys/signalvar.h>
139 #include <sys/systm.h>
140 #include <sys/vnode.h>
141 #include <sys/fstrans.h>
142 #include <sys/tty.h>
143 #include <sys/conf.h>
144 #include <sys/disklabel.h>
145 #include <sys/buf.h>
146 #include <sys/device.h>
147 #include <sys/exec.h>
148 #include <sys/socketvar.h>
149 #include <sys/protosw.h>
150 #include <sys/percpu.h>
151 #include <sys/pset.h>
152 #include <sys/sysctl.h>
153 #include <sys/reboot.h>
154 #include <sys/user.h>
155 #include <sys/sysctl.h>
156 #include <sys/event.h>
157 #include <sys/mbuf.h>
158 #include <sys/sched.h>
159 #include <sys/sleepq.h>
160 #include <sys/iostat.h>
161 #include <sys/vmem.h>
162 #include <sys/uuid.h>
163 #include <sys/extent.h>
164 #include <sys/disk.h>
165 #include <sys/msgbuf.h>
166 #include <sys/module.h>
167 #include <sys/event.h>
168 #include <sys/lockf.h>
169 #include <sys/once.h>
170 #include <sys/ksyms.h>
171 #include <sys/uidinfo.h>
172 #include <sys/kprintf.h>
173 #ifdef FAST_IPSEC
174 #include <netipsec/ipsec.h>
175 #endif
176 #ifdef SYSVSHM
177 #include <sys/shm.h>
178 #endif
179 #ifdef SYSVSEM
180 #include <sys/sem.h>
181 #endif
182 #ifdef SYSVMSG
183 #include <sys/msg.h>
184 #endif
185 #include <sys/domain.h>
186 #include <sys/namei.h>
187 #if NRND > 0
188 #include <sys/rnd.h>
189 #endif
190 #include <sys/pipe.h>
191 #if NVERIEXEC > 0
192 #include <sys/verified_exec.h>
193 #endif /* NVERIEXEC > 0 */
194 #ifdef KTRACE
195 #include <sys/ktrace.h>
196 #endif
197 #include <sys/kauth.h>
198 #ifdef WAPBL
199 #include <sys/wapbl.h>
200 #endif
201 #ifdef KERN_SA
202 #include <sys/savar.h>
203 #endif
204 #include <net80211/ieee80211_netbsd.h>
205 #ifdef PTRACE
206 #include <sys/ptrace.h>
207 #endif /* PTRACE */
208 
209 #include <sys/syscall.h>
210 #include <sys/syscallargs.h>
211 
212 #if defined(PAX_MPROTECT) || defined(PAX_SEGVGUARD) || defined(PAX_ASLR)
213 #include <sys/pax.h>
214 #endif /* PAX_MPROTECT || PAX_SEGVGUARD || PAX_ASLR */
215 
216 #include <ufs/ufs/quota.h>
217 
218 #include <miscfs/genfs/genfs.h>
219 #include <miscfs/syncfs/syncfs.h>
220 
221 #include <sys/cpu.h>
222 
223 #include <uvm/uvm.h>
224 
225 #if NSYSMON_TASKQ > 0
226 #include <dev/sysmon/sysmon_taskq.h>
227 #endif
228 
229 #include <dev/cons.h>
230 
231 #if NSYSMON_ENVSYS > 0 || NSYSMON_POWER > 0 || NSYSMON_WDOG > 0
232 #include <dev/sysmon/sysmonvar.h>
233 #endif
234 
235 #include <net/if.h>
236 #include <net/raw_cb.h>
237 
238 #include <prop/proplib.h>
239 
240 #ifdef COMPAT_50
241 #include <compat/sys/time.h>
242 struct timeval50 boottime50;
243 #endif
244 
245 #ifdef _KERNEL_OPT
246 #include "opt_userconf.h"
247 #endif
248 #ifdef USERCONF
249 #include <sys/userconf.h>
250 #endif
251 
252 extern struct proc proc0;
253 extern struct lwp lwp0;
254 extern struct cwdinfo cwdi0;
255 extern time_t rootfstime;
256 
257 #ifndef curlwp
258 struct	lwp *curlwp = &lwp0;
259 #endif
260 struct	proc *initproc;
261 
262 struct	vnode *rootvp, *swapdev_vp;
263 int	boothowto;
264 int	cold = 1;			/* still working on startup */
265 struct timespec boottime;	        /* time at system startup - will only follow settime deltas */
266 
267 int	start_init_exec;		/* semaphore for start_init() */
268 
269 static void check_console(struct lwp *l);
270 static void start_init(void *);
271 static void configure(void);
272 static void configure2(void);
273 void main(void);
274 
275 /*
276  * System startup; initialize the world, create process 0, mount root
277  * filesystem, and fork to create init and pagedaemon.  Most of the
278  * hard work is done in the lower-level initialization routines including
279  * startup(), which does memory initialization and autoconfiguration.
280  */
281 void
282 main(void)
283 {
284 	struct timespec time;
285 	struct lwp *l;
286 	struct proc *p;
287 	int s, error;
288 #ifdef NVNODE_IMPLICIT
289 	int usevnodes;
290 #endif
291 	CPU_INFO_ITERATOR cii;
292 	struct cpu_info *ci;
293 
294 	l = &lwp0;
295 #ifndef LWP0_CPU_INFO
296 	l->l_cpu = curcpu();
297 #endif
298 	l->l_pflag |= LP_RUNNING;
299 
300 	/*
301 	 * Attempt to find console and initialize
302 	 * in case of early panic or other messages.
303 	 */
304 	consinit();
305 
306 	kernel_lock_init();
307 	once_init();
308 	mutex_init(&cpu_lock, MUTEX_DEFAULT, IPL_NONE);
309 
310 	/* Initialize the device switch tables. */
311 	devsw_init();
312 
313 	uvm_init();
314 
315 	prop_kern_init();
316 
317 #if ((NKSYMS > 0) || (NDDB > 0) || (NMODULAR > 0))
318 	ksyms_init();
319 #endif
320 	kprintf_init();
321 
322 	percpu_init();
323 
324 	/* Initialize lock caches. */
325 	mutex_obj_init();
326 	rw_obj_init();
327 
328 	/* Initialize the extent manager. */
329 	extent_init();
330 
331 	/* Do machine-dependent initialization. */
332 	cpu_startup();
333 
334 	/* Initialize the sysctl subsystem. */
335 	sysctl_init();
336 
337 	/* Initialize callouts, part 1. */
338 	callout_startup();
339 
340 	/* Start module system. */
341 	module_init();
342 
343 	/*
344 	 * Initialize the kernel authorization subsystem and start the
345 	 * default security model, if any. We need to do this early
346 	 * enough so that subsystems relying on any of the aforementioned
347 	 * can work properly. Since the security model may dictate the
348 	 * credential inheritance policy, it is needed at least before
349 	 * any process is created, specifically proc0.
350 	 */
351 	kauth_init();
352 	module_init_class(MODULE_CLASS_SECMODEL);
353 
354 	/* Initialize the buffer cache */
355 	bufinit();
356 
357 	/* Initialize sockets. */
358 	soinit();
359 
360 	/*
361 	 * The following things must be done before autoconfiguration.
362 	 */
363 	evcnt_init();		/* initialize event counters */
364 #if NRND > 0
365 	rnd_init();		/* initialize random number generator */
366 #endif
367 
368 	/* Initialize process and pgrp structures. */
369 #ifdef KERN_SA
370 	sa_init();
371 #endif
372 	procinit();
373 	lwpinit();
374 
375 	/* Initialize signal-related data structures. */
376 	signal_init();
377 
378 	/* Initialize resource management. */
379 	resource_init();
380 
381 	/* Create process 0. */
382 	proc0_init();
383 
384 	/* Disable preemption during boot. */
385 	kpreempt_disable();
386 
387 	/* Initialize the UID hash table. */
388 	uid_init();
389 
390 	/* Charge root for one process. */
391 	(void)chgproccnt(0, 1);
392 
393 	/* Initialize timekeeping. */
394 	time_init();
395 
396 	/* Initialize the run queues, turnstiles and sleep queues. */
397 	sched_rqinit();
398 	turnstile_init();
399 	sleeptab_init(&sleeptab);
400 
401 	sched_init();
402 
403 	/* Initialize processor-sets */
404 	psets_init();
405 
406 	/* MI initialization of the boot cpu */
407 	error = mi_cpu_attach(curcpu());
408 	KASSERT(error == 0);
409 
410 	/* Initialize timekeeping, part 2. */
411 	time_init2();
412 
413 	/*
414 	 * Initialize mbuf's.  Do this now because we might attempt to
415 	 * allocate mbufs or mbuf clusters during autoconfiguration.
416 	 */
417 	mbinit();
418 
419 	/* Initialize I/O statistics. */
420 	iostat_init();
421 
422 	/* Initialize the log device. */
423 	loginit();
424 
425 	/* Second part of module system initialization. */
426 	module_init2();
427 
428 	/* Initialize the file systems. */
429 #ifdef NVNODE_IMPLICIT
430 	/*
431 	 * If maximum number of vnodes in namei vnode cache is not explicitly
432 	 * defined in kernel config, adjust the number such as we use roughly
433 	 * 10% of memory for vnodes and associated data structures in the
434 	 * assumed worst case.  Do not provide fewer than NVNODE vnodes.
435 	 */
436 	usevnodes =
437 	    calc_cache_size(kernel_map, 10, VNODE_VA_MAXPCT) / VNODE_COST;
438 	if (usevnodes > desiredvnodes)
439 		desiredvnodes = usevnodes;
440 #endif
441 	vfsinit();
442 	lf_init();
443 
444 	/* Initialize fstrans. */
445 	fstrans_init();
446 
447 	/* Initialize the file descriptor system. */
448 	fd_sys_init();
449 
450 	/* Initialize cwd structures */
451 	cwd_sys_init();
452 
453 	/* Initialize kqueue. */
454 	kqueue_init();
455 
456 	/* Initialize the system monitor subsystems. */
457 #if NSYSMON_TASKQ > 0
458 	sysmon_task_queue_preinit();
459 #endif
460 
461 #if NSYSMON_ENVSYS > 0
462 	sysmon_envsys_init();
463 #endif
464 
465 #if NSYSMON_POWER > 0
466 	sysmon_power_init();
467 #endif
468 
469 #if NSYSMON_WDOG > 0
470 	sysmon_wdog_init();
471 #endif
472 
473 	inittimecounter();
474 	ntp_init();
475 
476 	/* Initialize tty subsystem. */
477 	tty_init();
478 	ttyldisc_init();
479 
480 	/* Initialize the buffer cache, part 2. */
481 	bufinit2();
482 
483 	/* Initialize the disk wedge subsystem. */
484 	dkwedge_init();
485 
486 	/* Initialize interfaces. */
487 	ifinit1();
488 
489 	/* Configure the system hardware.  This will enable interrupts. */
490 	configure();
491 
492 	ssp_init();
493 
494 	configure2();
495 	/* Now timer is working.  Enable preemption. */
496 	kpreempt_enable();
497 
498 	ubc_init();		/* must be after autoconfig */
499 
500 #ifdef SYSVSHM
501 	/* Initialize System V style shared memory. */
502 	shminit();
503 #endif
504 
505 #ifdef SYSVSEM
506 	/* Initialize System V style semaphores. */
507 	seminit();
508 #endif
509 
510 #ifdef SYSVMSG
511 	/* Initialize System V style message queues. */
512 	msginit();
513 #endif
514 
515 #if NVERIEXEC > 0
516 	/*
517 	 * Initialise the Veriexec subsystem.
518 	 */
519 	veriexec_init();
520 #endif /* NVERIEXEC > 0 */
521 
522 #if defined(PAX_MPROTECT) || defined(PAX_SEGVGUARD) || defined(PAX_ASLR)
523 	pax_init();
524 #endif /* PAX_MPROTECT || PAX_SEGVGUARD || PAX_ASLR */
525 
526 #ifdef	FAST_IPSEC
527 	/* Attach network crypto subsystem */
528 	ipsec_attach();
529 #endif
530 
531 	/*
532 	 * Initialize protocols.  Block reception of incoming packets
533 	 * until everything is ready.
534 	 */
535 	s = splnet();
536 	ifinit();
537 	domaininit(true);
538 	if_attachdomain();
539 	splx(s);
540 
541 #ifdef GPROF
542 	/* Initialize kernel profiling. */
543 	kmstartup();
544 #endif
545 
546 	/* Initialize system accounting. */
547 	acct_init();
548 
549 #ifndef PIPE_SOCKETPAIR
550 	/* Initialize pipes. */
551 	pipe_init();
552 #endif
553 
554 #ifdef KTRACE
555 	/* Initialize ktrace. */
556 	ktrinit();
557 #endif
558 
559 #ifdef PTRACE
560 	/* Initialize ptrace. */
561 	ptrace_init();
562 #endif /* PTRACE */
563 
564 	/* Initialize the UUID system calls. */
565 	uuid_init();
566 
567 #ifdef WAPBL
568 	/* Initialize write-ahead physical block logging. */
569 	wapbl_init();
570 #endif
571 
572 	machdep_init();
573 
574 	/*
575 	 * Create process 1 (init(8)).  We do this now, as Unix has
576 	 * historically had init be process 1, and changing this would
577 	 * probably upset a lot of people.
578 	 *
579 	 * Note that process 1 won't immediately exec init(8), but will
580 	 * wait for us to inform it that the root file system has been
581 	 * mounted.
582 	 */
583 	if (fork1(l, 0, SIGCHLD, NULL, 0, start_init, NULL, NULL, &initproc))
584 		panic("fork init");
585 
586 	/*
587 	 * Load any remaining builtin modules, and hand back temporary
588 	 * storage to the VM system.
589 	 */
590 	module_init_class(MODULE_CLASS_ANY);
591 
592 	/*
593 	 * Finalize configuration now that all real devices have been
594 	 * found.  This needs to be done before the root device is
595 	 * selected, since finalization may create the root device.
596 	 */
597 	config_finalize();
598 
599 	sysctl_finalize();
600 
601 	/*
602 	 * Now that autoconfiguration has completed, we can determine
603 	 * the root and dump devices.
604 	 */
605 	cpu_rootconf();
606 	cpu_dumpconf();
607 
608 	/* Mount the root file system. */
609 	do {
610 		domountroothook();
611 		if ((error = vfs_mountroot())) {
612 			printf("cannot mount root, error = %d\n", error);
613 			boothowto |= RB_ASKNAME;
614 			setroot(root_device,
615 			    (rootdev != NODEV) ? DISKPART(rootdev) : 0);
616 		}
617 	} while (error != 0);
618 	mountroothook_destroy();
619 
620 	/*
621 	 * Initialise the time-of-day clock, passing the time recorded
622 	 * in the root filesystem (if any) for use by systems that
623 	 * don't have a non-volatile time-of-day device.
624 	 */
625 	inittodr(rootfstime);
626 
627 	CIRCLEQ_FIRST(&mountlist)->mnt_flag |= MNT_ROOTFS;
628 	CIRCLEQ_FIRST(&mountlist)->mnt_op->vfs_refcount++;
629 
630 	/*
631 	 * Get the vnode for '/'.  Set filedesc0.fd_fd.fd_cdir to
632 	 * reference it.
633 	 */
634 	error = VFS_ROOT(CIRCLEQ_FIRST(&mountlist), &rootvnode);
635 	if (error)
636 		panic("cannot find root vnode, error=%d", error);
637 	cwdi0.cwdi_cdir = rootvnode;
638 	VREF(cwdi0.cwdi_cdir);
639 	VOP_UNLOCK(rootvnode, 0);
640 	cwdi0.cwdi_rdir = NULL;
641 
642 	/*
643 	 * Now that root is mounted, we can fixup initproc's CWD
644 	 * info.  All other processes are kthreads, which merely
645 	 * share proc0's CWD info.
646 	 */
647 	initproc->p_cwdi->cwdi_cdir = rootvnode;
648 	VREF(initproc->p_cwdi->cwdi_cdir);
649 	initproc->p_cwdi->cwdi_rdir = NULL;
650 
651 	/*
652 	 * Now can look at time, having had a chance to verify the time
653 	 * from the file system.  Reset l->l_rtime as it may have been
654 	 * munched in mi_switch() after the time got set.
655 	 */
656 	getnanotime(&time);
657 	boottime = time;
658 #ifdef COMPAT_50
659 	{
660 		struct timeval tv;
661 		TIMESPEC_TO_TIMEVAL(&tv, &time);
662 		timeval_to_timeval50(&tv, &boottime50);
663 	}
664 #endif
665 	mutex_enter(proc_lock);
666 	LIST_FOREACH(p, &allproc, p_list) {
667 		KASSERT((p->p_flag & PK_MARKER) == 0);
668 		mutex_enter(p->p_lock);
669 		TIMESPEC_TO_TIMEVAL(&p->p_stats->p_start, &time);
670 		LIST_FOREACH(l, &p->p_lwps, l_sibling) {
671 			lwp_lock(l);
672 			memset(&l->l_rtime, 0, sizeof(l->l_rtime));
673 			lwp_unlock(l);
674 		}
675 		mutex_exit(p->p_lock);
676 	}
677 	mutex_exit(proc_lock);
678 	binuptime(&curlwp->l_stime);
679 
680 	for (CPU_INFO_FOREACH(cii, ci)) {
681 		ci->ci_schedstate.spc_lastmod = time_second;
682 	}
683 
684 	/* Create the pageout daemon kernel thread. */
685 	uvm_swap_init();
686 	if (kthread_create(PRI_PGDAEMON, KTHREAD_MPSAFE, NULL, uvm_pageout,
687 	    NULL, NULL, "pgdaemon"))
688 		panic("fork pagedaemon");
689 
690 	/* Create the filesystem syncer kernel thread. */
691 	if (kthread_create(PRI_IOFLUSH, KTHREAD_MPSAFE, NULL, sched_sync,
692 	    NULL, NULL, "ioflush"))
693 		panic("fork syncer");
694 
695 	/* Create the aiodone daemon kernel thread. */
696 	if (workqueue_create(&uvm.aiodone_queue, "aiodoned",
697 	    uvm_aiodone_worker, NULL, PRI_VM, IPL_NONE, WQ_MPSAFE))
698 		panic("fork aiodoned");
699 
700 	vmem_rehash_start();
701 
702 	/* Initialize exec structures */
703 	exec_init(1);
704 
705 	/*
706 	 * Okay, now we can let init(8) exec!  It's off to userland!
707 	 */
708 	mutex_enter(proc_lock);
709 	start_init_exec = 1;
710 	cv_broadcast(&lbolt);
711 	mutex_exit(proc_lock);
712 
713 	/* The scheduler is an infinite loop. */
714 	uvm_scheduler();
715 	/* NOTREACHED */
716 }
717 
718 /*
719  * Configure the system's hardware.
720  */
721 static void
722 configure(void)
723 {
724 
725 	/* Initialize autoconf data structures. */
726 	config_init_mi();
727 	/*
728 	 * XXX
729 	 * callout_setfunc() requires mutex(9) so it can't be in config_init()
730 	 * on amiga and atari which use config_init() and autoconf(9) fucntions
731 	 * to initialize console devices.
732 	 */
733 	config_twiddle_init();
734 
735 	pmf_init();
736 #if NDRVCTL > 0
737 	drvctl_init();
738 #endif
739 
740 #ifdef USERCONF
741 	if (boothowto & RB_USERCONF)
742 		user_config();
743 #endif
744 
745 	if ((boothowto & (AB_SILENT|AB_VERBOSE)) == AB_SILENT) {
746 		printf_nolog("Detecting hardware...");
747 	}
748 
749 	/*
750 	 * Do the machine-dependent portion of autoconfiguration.  This
751 	 * sets the configuration machinery here in motion by "finding"
752 	 * the root bus.  When this function returns, we expect interrupts
753 	 * to be enabled.
754 	 */
755 	cpu_configure();
756 }
757 
758 static void
759 configure2(void)
760 {
761 	CPU_INFO_ITERATOR cii;
762 	struct cpu_info *ci;
763 	int s;
764 
765 	/*
766 	 * Now that we've found all the hardware, start the real time
767 	 * and statistics clocks.
768 	 */
769 	initclocks();
770 
771 	cold = 0;	/* clocks are running, we're warm now! */
772 	s = splsched();
773 	curcpu()->ci_schedstate.spc_flags |= SPCF_RUNNING;
774 	splx(s);
775 
776 	/* Boot the secondary processors. */
777 	for (CPU_INFO_FOREACH(cii, ci)) {
778 		uvm_cpu_attach(ci);
779 	}
780 	mp_online = true;
781 #if defined(MULTIPROCESSOR)
782 	cpu_boot_secondary_processors();
783 #endif
784 
785 	/* Setup the runqueues and scheduler. */
786 	runq_init();
787 	synch_init();
788 
789 	/*
790 	 * Bus scans can make it appear as if the system has paused, so
791 	 * twiddle constantly while config_interrupts() jobs are running.
792 	 */
793 	config_twiddle_fn(NULL);
794 
795 	/*
796 	 * Create threads to call back and finish configuration for
797 	 * devices that want interrupts enabled.
798 	 */
799 	config_create_interruptthreads();
800 
801 	/* Get the threads going and into any sleeps before continuing. */
802 	yield();
803 }
804 
805 static void
806 check_console(struct lwp *l)
807 {
808 	struct vnode *vp;
809 	int error;
810 
811 	error = namei_simple_kernel("/dev/console",
812 				NSM_FOLLOW_NOEMULROOT, &vp);
813 	if (error == 0)
814 		vrele(vp);
815 	else if (error == ENOENT)
816 		printf("warning: no /dev/console\n");
817 	else
818 		printf("warning: lookup /dev/console: error %d\n", error);
819 }
820 
821 /*
822  * List of paths to try when searching for "init".
823  */
824 static const char * const initpaths[] = {
825 	"/sbin/init",
826 	"/sbin/oinit",
827 	"/sbin/init.bak",
828 	NULL,
829 };
830 
831 /*
832  * Start the initial user process; try exec'ing each pathname in "initpaths".
833  * The program is invoked with one argument containing the boot flags.
834  */
835 static void
836 start_init(void *arg)
837 {
838 	struct lwp *l = arg;
839 	struct proc *p = l->l_proc;
840 	vaddr_t addr;
841 	struct sys_execve_args /* {
842 		syscallarg(const char *) path;
843 		syscallarg(char * const *) argp;
844 		syscallarg(char * const *) envp;
845 	} */ args;
846 	int options, i, error;
847 	register_t retval[2];
848 	char flags[4], *flagsp;
849 	const char *path, *slash;
850 	char *ucp, **uap, *arg0, *arg1 = NULL;
851 	char ipath[129];
852 	int ipx, len;
853 
854 	/*
855 	 * Now in process 1.
856 	 */
857 	strncpy(p->p_comm, "init", MAXCOMLEN);
858 
859 	/*
860 	 * Wait for main() to tell us that it's safe to exec.
861 	 */
862 	mutex_enter(proc_lock);
863 	while (start_init_exec == 0)
864 		cv_wait(&lbolt, proc_lock);
865 	mutex_exit(proc_lock);
866 
867 	/*
868 	 * This is not the right way to do this.  We really should
869 	 * hand-craft a descriptor onto /dev/console to hand to init,
870 	 * but that's a _lot_ more work, and the benefit from this easy
871 	 * hack makes up for the "good is the enemy of the best" effect.
872 	 */
873 	check_console(l);
874 
875 	/*
876 	 * Need just enough stack to hold the faked-up "execve()" arguments.
877 	 */
878 	addr = (vaddr_t)STACK_ALLOC(USRSTACK, PAGE_SIZE);
879 	if (uvm_map(&p->p_vmspace->vm_map, &addr, PAGE_SIZE,
880                     NULL, UVM_UNKNOWN_OFFSET, 0,
881                     UVM_MAPFLAG(UVM_PROT_ALL, UVM_PROT_ALL, UVM_INH_COPY,
882 		    UVM_ADV_NORMAL,
883                     UVM_FLAG_FIXED|UVM_FLAG_OVERLAY|UVM_FLAG_COPYONW)) != 0)
884 		panic("init: couldn't allocate argument space");
885 	p->p_vmspace->vm_maxsaddr = (void *)STACK_MAX(addr, PAGE_SIZE);
886 
887 	ipx = 0;
888 	while (1) {
889 		if (boothowto & RB_ASKNAME) {
890 			printf("init path");
891 			if (initpaths[ipx])
892 				printf(" (default %s)", initpaths[ipx]);
893 			printf(": ");
894 			len = cngetsn(ipath, sizeof(ipath)-1);
895 			if (len == 4 && strcmp(ipath, "halt") == 0) {
896 				cpu_reboot(RB_HALT, NULL);
897 			} else if (len == 6 && strcmp(ipath, "reboot") == 0) {
898 				cpu_reboot(0, NULL);
899 #if defined(DDB)
900 			} else if (len == 3 && strcmp(ipath, "ddb") == 0) {
901 				console_debugger();
902 				continue;
903 #endif
904 			} else if (len > 0 && ipath[0] == '/') {
905 				ipath[len] = '\0';
906 				path = ipath;
907 			} else if (len == 0 && initpaths[ipx] != NULL) {
908 				path = initpaths[ipx++];
909 			} else {
910 				printf("use absolute path, ");
911 #if defined(DDB)
912 				printf("\"ddb\", ");
913 #endif
914 				printf("\"halt\", or \"reboot\"\n");
915 				continue;
916 			}
917 		} else {
918 			if ((path = initpaths[ipx++]) == NULL) {
919 				ipx = 0;
920 				boothowto |= RB_ASKNAME;
921 				continue;
922 			}
923 		}
924 
925 		ucp = (char *)USRSTACK;
926 
927 		/*
928 		 * Construct the boot flag argument.
929 		 */
930 		flagsp = flags;
931 		*flagsp++ = '-';
932 		options = 0;
933 
934 		if (boothowto & RB_SINGLE) {
935 			*flagsp++ = 's';
936 			options = 1;
937 		}
938 #ifdef notyet
939 		if (boothowto & RB_FASTBOOT) {
940 			*flagsp++ = 'f';
941 			options = 1;
942 		}
943 #endif
944 
945 		/*
946 		 * Move out the flags (arg 1), if necessary.
947 		 */
948 		if (options != 0) {
949 			*flagsp++ = '\0';
950 			i = flagsp - flags;
951 #ifdef DEBUG
952 			printf("init: copying out flags `%s' %d\n", flags, i);
953 #endif
954 			arg1 = STACK_ALLOC(ucp, i);
955 			ucp = STACK_MAX(arg1, i);
956 			(void)copyout((void *)flags, arg1, i);
957 		}
958 
959 		/*
960 		 * Move out the file name (also arg 0).
961 		 */
962 		i = strlen(path) + 1;
963 #ifdef DEBUG
964 		printf("init: copying out path `%s' %d\n", path, i);
965 #else
966 		if (boothowto & RB_ASKNAME || path != initpaths[0])
967 			printf("init: trying %s\n", path);
968 #endif
969 		arg0 = STACK_ALLOC(ucp, i);
970 		ucp = STACK_MAX(arg0, i);
971 		(void)copyout(path, arg0, i);
972 
973 		/*
974 		 * Move out the arg pointers.
975 		 */
976 		ucp = (void *)STACK_ALIGN(ucp, ALIGNBYTES);
977 		uap = (char **)STACK_ALLOC(ucp, sizeof(char *) * 3);
978 		SCARG(&args, path) = arg0;
979 		SCARG(&args, argp) = uap;
980 		SCARG(&args, envp) = NULL;
981 		slash = strrchr(path, '/');
982 		if (slash)
983 			(void)suword((void *)uap++,
984 			    (long)arg0 + (slash + 1 - path));
985 		else
986 			(void)suword((void *)uap++, (long)arg0);
987 		if (options != 0)
988 			(void)suword((void *)uap++, (long)arg1);
989 		(void)suword((void *)uap++, 0);	/* terminator */
990 
991 		/*
992 		 * Now try to exec the program.  If can't for any reason
993 		 * other than it doesn't exist, complain.
994 		 */
995 		error = sys_execve(l, &args, retval);
996 		if (error == 0 || error == EJUSTRETURN) {
997 			KERNEL_UNLOCK_LAST(l);
998 			return;
999 		}
1000 		printf("exec %s: error %d\n", path, error);
1001 	}
1002 	printf("init: not found\n");
1003 	panic("no init");
1004 }
1005 
1006 /*
1007  * calculate cache size from physmem and vm_map size.
1008  */
1009 vaddr_t
1010 calc_cache_size(struct vm_map *map, int pct, int va_pct)
1011 {
1012 	paddr_t t;
1013 
1014 	/* XXX should consider competing cache if any */
1015 	/* XXX should consider submaps */
1016 	t = (uintmax_t)physmem * pct / 100 * PAGE_SIZE;
1017 	if (map != NULL) {
1018 		vsize_t vsize;
1019 
1020 		vsize = vm_map_max(map) - vm_map_min(map);
1021 		vsize = (uintmax_t)vsize * va_pct / 100;
1022 		if (t > vsize) {
1023 			t = vsize;
1024 		}
1025 	}
1026 	return t;
1027 }
1028 
1029 /*
1030  * Print the system start up banner.
1031  *
1032  * - Print a limited banner if AB_SILENT.
1033  * - Always send normal banner to the log.
1034  */
1035 #define MEM_PBUFSIZE	sizeof("99999 MB")
1036 
1037 void
1038 banner(void)
1039 {
1040 	static char notice[] = " Notice: this software is "
1041 	    "protected by copyright";
1042 	char pbuf[81];
1043 	void (*pr)(const char *, ...);
1044 	int i;
1045 
1046 	if ((boothowto & AB_SILENT) != 0) {
1047 		snprintf(pbuf, sizeof(pbuf), "%s %s (%s)",
1048 		    ostype, osrelease, kernel_ident);
1049 		printf_nolog("%s", pbuf);
1050 		for (i = 80 - strlen(pbuf) - sizeof(notice); i > 0; i--)
1051 			printf(" ");
1052 		printf_nolog("%s\n", notice);
1053 		pr = aprint_normal;
1054 	} else {
1055 		pr = printf;
1056 	}
1057 
1058 	memset(pbuf, 0, sizeof(pbuf));
1059 	(*pr)("%s%s", copyright, version);
1060 	format_bytes(pbuf, MEM_PBUFSIZE, ctob((uint64_t)physmem));
1061 	(*pr)("total memory = %s\n", pbuf);
1062 	format_bytes(pbuf, MEM_PBUFSIZE, ctob((uint64_t)uvmexp.free));
1063 	(*pr)("avail memory = %s\n", pbuf);
1064 }
1065