xref: /netbsd-src/sys/kern/init_main.c (revision aad9773e38ed2370a628a6416e098f9008fc10a7)
1 /*	$NetBSD: init_main.c,v 1.461 2014/11/27 14:38:09 uebayasi 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 1.461 2014/11/27 14:38:09 uebayasi 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_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_compat_netbsd.h"
113 #include "opt_wapbl.h"
114 #include "opt_ptrace.h"
115 #include "opt_rnd_printf.h"
116 
117 #include "drvctl.h"
118 #include "ksyms.h"
119 
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/cpufreq.h>
134 #include <sys/spldebug.h>
135 #include <sys/kernel.h>
136 #include <sys/mount.h>
137 #include <sys/proc.h>
138 #include <sys/kthread.h>
139 #include <sys/resourcevar.h>
140 #include <sys/signalvar.h>
141 #include <sys/systm.h>
142 #include <sys/vnode.h>
143 #include <sys/fstrans.h>
144 #include <sys/tty.h>
145 #include <sys/conf.h>
146 #include <sys/disklabel.h>
147 #include <sys/buf.h>
148 #include <sys/device.h>
149 #include <sys/exec.h>
150 #include <sys/socketvar.h>
151 #include <sys/protosw.h>
152 #include <sys/percpu.h>
153 #include <sys/pserialize.h>
154 #include <sys/pset.h>
155 #include <sys/sysctl.h>
156 #include <sys/reboot.h>
157 #include <sys/event.h>
158 #include <sys/mbuf.h>
159 #include <sys/sched.h>
160 #include <sys/sleepq.h>
161 #include <sys/ipi.h>
162 #include <sys/iostat.h>
163 #include <sys/vmem.h>
164 #include <sys/uuid.h>
165 #include <sys/extent.h>
166 #include <sys/disk.h>
167 #include <sys/msgbuf.h>
168 #include <sys/module.h>
169 #include <sys/event.h>
170 #include <sys/lockf.h>
171 #include <sys/once.h>
172 #include <sys/kcpuset.h>
173 #include <sys/ksyms.h>
174 #include <sys/uidinfo.h>
175 #include <sys/kprintf.h>
176 #ifdef IPSEC
177 #include <netipsec/ipsec.h>
178 #endif
179 #ifdef SYSVSHM
180 #include <sys/shm.h>
181 #endif
182 #ifdef SYSVSEM
183 #include <sys/sem.h>
184 #endif
185 #ifdef SYSVMSG
186 #include <sys/msg.h>
187 #endif
188 #include <sys/domain.h>
189 #include <sys/namei.h>
190 #include <sys/rnd.h>
191 #include <sys/pipe.h>
192 #if NVERIEXEC > 0
193 #include <sys/verified_exec.h>
194 #endif /* NVERIEXEC > 0 */
195 #ifdef KTRACE
196 #include <sys/ktrace.h>
197 #endif
198 #include <sys/kauth.h>
199 #include <net80211/ieee80211_netbsd.h>
200 #ifdef PTRACE
201 #include <sys/ptrace.h>
202 #endif /* PTRACE */
203 #include <sys/cprng.h>
204 
205 #include <sys/syscall.h>
206 #include <sys/syscallargs.h>
207 
208 #if defined(PAX_MPROTECT) || defined(PAX_SEGVGUARD) || defined(PAX_ASLR)
209 #include <sys/pax.h>
210 #endif /* PAX_MPROTECT || PAX_SEGVGUARD || PAX_ASLR */
211 
212 #include <secmodel/secmodel.h>
213 
214 #include <ufs/ufs/quota.h>
215 
216 #include <miscfs/genfs/genfs.h>
217 #include <miscfs/syncfs/syncfs.h>
218 #include <miscfs/specfs/specdev.h>
219 
220 #include <sys/cpu.h>
221 
222 #include <uvm/uvm.h>	/* extern struct uvm uvm */
223 
224 #if NSYSMON_TASKQ > 0
225 #include <dev/sysmon/sysmon_taskq.h>
226 #endif
227 
228 #include <dev/cons.h>
229 
230 #if NSYSMON_ENVSYS > 0 || NSYSMON_POWER > 0 || NSYSMON_WDOG > 0
231 #include <dev/sysmon/sysmonvar.h>
232 #endif
233 
234 #include <net/bpf.h>
235 #include <net/if.h>
236 #include <net/raw_cb.h>
237 
238 #include <prop/proplib.h>
239 
240 #include <sys/userconf.h>
241 
242 extern struct lwp lwp0;
243 extern time_t rootfstime;
244 
245 #ifndef curlwp
246 struct	lwp *curlwp = &lwp0;
247 #endif
248 struct	proc *initproc;
249 
250 struct	vnode *rootvp, *swapdev_vp;
251 int	boothowto;
252 int	cold = 1;			/* still working on startup */
253 struct timespec boottime;	        /* time at system startup - will only follow settime deltas */
254 
255 int	start_init_exec;		/* semaphore for start_init() */
256 
257 cprng_strong_t	*kern_cprng;
258 
259 static void check_console(struct lwp *l);
260 static void start_init(void *);
261 static void configure(void);
262 static void configure2(void);
263 static void configure3(void);
264 void main(void);
265 
266 /*
267  * System startup; initialize the world, create process 0, mount root
268  * filesystem, and fork to create init and pagedaemon.  Most of the
269  * hard work is done in the lower-level initialization routines including
270  * startup(), which does memory initialization and autoconfiguration.
271  */
272 void
273 main(void)
274 {
275 	struct timespec time;
276 	struct lwp *l;
277 	struct proc *p;
278 	int s, error;
279 #ifdef NVNODE_IMPLICIT
280 	int usevnodes;
281 #endif
282 	CPU_INFO_ITERATOR cii;
283 	struct cpu_info *ci;
284 
285 	l = &lwp0;
286 #ifndef LWP0_CPU_INFO
287 	l->l_cpu = curcpu();
288 #endif
289 	l->l_pflag |= LP_RUNNING;
290 
291 	/*
292 	 * Attempt to find console and initialize
293 	 * in case of early panic or other messages.
294 	 */
295 	consinit();
296 
297 	kernel_lock_init();
298 	once_init();
299 
300 	mi_cpu_init();
301 	kernconfig_lock_init();
302 	kthread_sysinit();
303 
304 	/* Initialize the device switch tables. */
305 	devsw_init();
306 
307 	/* Initialize event counters. */
308 	evcnt_init();
309 
310 	uvm_init();
311 	ubchist_init();
312 	kcpuset_sysinit();
313 
314 	prop_kern_init();
315 
316 #if ((NKSYMS > 0) || (NDDB > 0) || (NMODULAR > 0))
317 	ksyms_init();
318 #endif
319 	kprintf_init();
320 
321 	percpu_init();
322 
323 	/* Initialize lock caches. */
324 	mutex_obj_init();
325 	rw_obj_init();
326 
327 	/* Passive serialization. */
328 	pserialize_init();
329 
330 	/* Initialize the extent manager. */
331 	extent_init();
332 
333 	/* Do machine-dependent initialization. */
334 	cpu_startup();
335 
336 	/* Initialize the sysctl subsystem. */
337 	sysctl_init();
338 
339 	/* Initialize callouts, part 1. */
340 	callout_startup();
341 
342 	/* Initialize the kernel authorization subsystem. */
343 	kauth_init();
344 
345 	secmodel_init();
346 
347 	spec_init();
348 
349 	/*
350 	 * Set BPF op vector.  Can't do this in bpf attach, since
351 	 * network drivers attach before bpf.
352 	 */
353 	bpf_setops();
354 
355 	/* Start module system. */
356 	module_init();
357 
358 	/*
359 	 * Initialize the kernel authorization subsystem and start the
360 	 * default security model, if any. We need to do this early
361 	 * enough so that subsystems relying on any of the aforementioned
362 	 * can work properly. Since the security model may dictate the
363 	 * credential inheritance policy, it is needed at least before
364 	 * any process is created, specifically proc0.
365 	 */
366 	module_init_class(MODULE_CLASS_SECMODEL);
367 
368 	/* Initialize the buffer cache */
369 	bufinit();
370 
371 	/* Initialize sockets. */
372 	soinit();
373 
374 	/*
375 	 * The following things must be done before autoconfiguration.
376 	 */
377 	rnd_init();		/* initialize entropy pool */
378 
379 	cprng_init();		/* initialize cryptographic PRNG */
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. */
392 	proc0_init();
393 	lwp0_init();
394 
395 	/* Disable preemption during boot. */
396 	kpreempt_disable();
397 
398 	/* Initialize the UID hash table. */
399 	uid_init();
400 
401 	/* Charge root for one process. */
402 	(void)chgproccnt(0, 1);
403 
404 	/* Initialize timekeeping. */
405 	time_init();
406 
407 	/* Initialize the run queues, turnstiles and sleep queues. */
408 	sched_rqinit();
409 	turnstile_init();
410 	sleeptab_init(&sleeptab);
411 
412 	sched_init();
413 
414 	/* Initialize processor-sets */
415 	psets_init();
416 
417 	/* Initialize cpufreq(9) */
418 	cpufreq_init();
419 
420 	/* MI initialization of the boot cpu */
421 	error = mi_cpu_attach(curcpu());
422 	KASSERT(error == 0);
423 
424 	/* Initialize timekeeping, part 2. */
425 	time_init2();
426 
427 	/*
428 	 * Initialize mbuf's.  Do this now because we might attempt to
429 	 * allocate mbufs or mbuf clusters during autoconfiguration.
430 	 */
431 	mbinit();
432 
433 	/* Initialize I/O statistics. */
434 	iostat_init();
435 
436 	/* Initialize the log device. */
437 	loginit();
438 
439 	/* Second part of module system initialization. */
440 	module_start_unload_thread();
441 
442 	/* Initialize the file systems. */
443 #ifdef NVNODE_IMPLICIT
444 	/*
445 	 * If maximum number of vnodes in namei vnode cache is not explicitly
446 	 * defined in kernel config, adjust the number such as we use roughly
447 	 * 10% of memory for vnodes and associated data structures in the
448 	 * assumed worst case.  Do not provide fewer than NVNODE vnodes.
449 	 */
450 	usevnodes = calc_cache_size(vmem_size(kmem_arena, VMEM_FREE|VMEM_ALLOC),
451 	    10, VNODE_KMEM_MAXPCT) / VNODE_COST;
452 	if (usevnodes > desiredvnodes)
453 		desiredvnodes = usevnodes;
454 #endif
455 	vfsinit();
456 	lf_init();
457 
458 	/* Initialize fstrans. */
459 	fstrans_init();
460 
461 	/* Initialize the file descriptor system. */
462 	fd_sys_init();
463 
464 	/* Initialize cwd structures */
465 	cwd_sys_init();
466 
467 	/* Initialize kqueue. */
468 	kqueue_init();
469 
470 	/* Initialize the system monitor subsystems. */
471 #if NSYSMON_TASKQ > 0
472 	sysmon_task_queue_preinit();
473 #endif
474 
475 #if NSYSMON_ENVSYS > 0
476 	sysmon_envsys_init();
477 #endif
478 
479 #if NSYSMON_POWER > 0
480 	sysmon_power_init();
481 #endif
482 
483 #if NSYSMON_WDOG > 0
484 	sysmon_wdog_init();
485 #endif
486 
487 	inittimecounter();
488 	ntp_init();
489 
490 	/* Initialize tty subsystem. */
491 	tty_init();
492 	ttyldisc_init();
493 
494 	/* Initialize the buffer cache, part 2. */
495 	bufinit2();
496 
497 	/* Initialize the disk wedge subsystem. */
498 	dkwedge_init();
499 
500 	/* Initialize the kernel strong PRNG. */
501 	kern_cprng = cprng_strong_create("kernel", IPL_VM,
502 					 CPRNG_INIT_ANY|CPRNG_REKEY_ANY);
503 
504 	/* Initialize interfaces. */
505 	ifinit1();
506 
507 	spldebug_start();
508 
509 	/* Initialize sockets thread(s) */
510 	soinit1();
511 
512 	/* Configure the system hardware.  This will enable interrupts. */
513 	configure();
514 
515 	/* Once all CPUs are detected, initialize the per-CPU cprng_fast.  */
516 	cprng_fast_init();
517 
518 	ssp_init();
519 
520 	ubc_init();		/* must be after autoconfig */
521 
522 	mm_init();
523 
524 	configure2();
525 
526 	ipi_sysinit();
527 
528 	/* Now timer is working.  Enable preemption. */
529 	kpreempt_enable();
530 
531 	/* Get the threads going and into any sleeps before continuing. */
532 	yield();
533 
534 	/* Enable deferred processing of RNG samples */
535 	rnd_init_softint();
536 
537 #ifdef RND_PRINTF
538 	/* Enable periodic injection of console output into entropy pool */
539 	kprintf_init_callout();
540 #endif
541 
542 #ifdef SYSVSHM
543 	/* Initialize System V style shared memory. */
544 	shminit();
545 #endif
546 
547 	vmem_rehash_start();	/* must be before exec_init */
548 
549 	/* Initialize exec structures */
550 	exec_init(1);		/* seminit calls exithook_establish() */
551 
552 #ifdef SYSVSEM
553 	/* Initialize System V style semaphores. */
554 	seminit();
555 #endif
556 
557 #ifdef SYSVMSG
558 	/* Initialize System V style message queues. */
559 	msginit();
560 #endif
561 
562 #if NVERIEXEC > 0
563 	/*
564 	 * Initialise the Veriexec subsystem.
565 	 */
566 	veriexec_init();
567 #endif /* NVERIEXEC > 0 */
568 
569 #if defined(PAX_MPROTECT) || defined(PAX_SEGVGUARD) || defined(PAX_ASLR)
570 	pax_init();
571 #endif /* PAX_MPROTECT || PAX_SEGVGUARD || PAX_ASLR */
572 
573 #ifdef	IPSEC
574 	/* Attach network crypto subsystem */
575 	ipsec_attach();
576 #endif
577 
578 	/*
579 	 * Initialize protocols.  Block reception of incoming packets
580 	 * until everything is ready.
581 	 */
582 	s = splnet();
583 	ifinit();
584 	domaininit(true);
585 	if_attachdomain();
586 	splx(s);
587 
588 #ifdef GPROF
589 	/* Initialize kernel profiling. */
590 	kmstartup();
591 #endif
592 
593 	/* Initialize system accounting. */
594 	acct_init();
595 
596 #ifndef PIPE_SOCKETPAIR
597 	/* Initialize pipes. */
598 	pipe_init();
599 #endif
600 
601 #ifdef KTRACE
602 	/* Initialize ktrace. */
603 	ktrinit();
604 #endif
605 
606 #ifdef PTRACE
607 	/* Initialize ptrace. */
608 	ptrace_init();
609 #endif /* PTRACE */
610 
611 	machdep_init();
612 
613 	procinit_sysctl();
614 
615 	/*
616 	 * Create process 1 (init(8)).  We do this now, as Unix has
617 	 * historically had init be process 1, and changing this would
618 	 * probably upset a lot of people.
619 	 *
620 	 * Note that process 1 won't immediately exec init(8), but will
621 	 * wait for us to inform it that the root file system has been
622 	 * mounted.
623 	 */
624 	if (fork1(l, 0, SIGCHLD, NULL, 0, start_init, NULL, NULL, &initproc))
625 		panic("fork init");
626 
627 	/*
628 	 * Load any remaining builtin modules, and hand back temporary
629 	 * storage to the VM system.  Then require force when loading any
630 	 * remaining un-init'ed built-in modules to avoid later surprises.
631 	 */
632 	module_init_class(MODULE_CLASS_ANY);
633 	module_builtin_require_force();
634 
635 	/*
636 	 * Finalize configuration now that all real devices have been
637 	 * found.  This needs to be done before the root device is
638 	 * selected, since finalization may create the root device.
639 	 */
640 	config_finalize();
641 
642 	sysctl_finalize();
643 
644 	/*
645 	 * Now that autoconfiguration has completed, we can determine
646 	 * the root and dump devices.
647 	 */
648 	cpu_rootconf();
649 	cpu_dumpconf();
650 
651 	/* Mount the root file system. */
652 	do {
653 		domountroothook(root_device);
654 		if ((error = vfs_mountroot())) {
655 			printf("cannot mount root, error = %d\n", error);
656 			boothowto |= RB_ASKNAME;
657 			setroot(root_device,
658 			    (rootdev != NODEV) ? DISKPART(rootdev) : 0);
659 		}
660 	} while (error != 0);
661 	mountroothook_destroy();
662 
663 	configure3();
664 
665 	/*
666 	 * Initialise the time-of-day clock, passing the time recorded
667 	 * in the root filesystem (if any) for use by systems that
668 	 * don't have a non-volatile time-of-day device.
669 	 */
670 	inittodr(rootfstime);
671 
672 	/*
673 	 * Now can look at time, having had a chance to verify the time
674 	 * from the file system.  Reset l->l_rtime as it may have been
675 	 * munched in mi_switch() after the time got set.
676 	 */
677 	getnanotime(&time);
678 	boottime = time;
679 
680 	mutex_enter(proc_lock);
681 	LIST_FOREACH(p, &allproc, p_list) {
682 		KASSERT((p->p_flag & PK_MARKER) == 0);
683 		mutex_enter(p->p_lock);
684 		TIMESPEC_TO_TIMEVAL(&p->p_stats->p_start, &time);
685 		LIST_FOREACH(l, &p->p_lwps, l_sibling) {
686 			lwp_lock(l);
687 			memset(&l->l_rtime, 0, sizeof(l->l_rtime));
688 			lwp_unlock(l);
689 		}
690 		mutex_exit(p->p_lock);
691 	}
692 	mutex_exit(proc_lock);
693 	binuptime(&curlwp->l_stime);
694 
695 	for (CPU_INFO_FOREACH(cii, ci)) {
696 		ci->ci_schedstate.spc_lastmod = time_second;
697 	}
698 
699 	/* Create the pageout daemon kernel thread. */
700 	uvm_swap_init();
701 	if (kthread_create(PRI_PGDAEMON, KTHREAD_MPSAFE, NULL, uvm_pageout,
702 	    NULL, NULL, "pgdaemon"))
703 		panic("fork pagedaemon");
704 
705 	/* Create the filesystem syncer kernel thread. */
706 	if (kthread_create(PRI_IOFLUSH, KTHREAD_MPSAFE, NULL, sched_sync,
707 	    NULL, NULL, "ioflush"))
708 		panic("fork syncer");
709 
710 	/* Create the aiodone daemon kernel thread. */
711 	if (workqueue_create(&uvm.aiodone_queue, "aiodoned",
712 	    uvm_aiodone_worker, NULL, PRI_VM, IPL_NONE, WQ_MPSAFE))
713 		panic("fork aiodoned");
714 
715 	/*
716 	 * Okay, now we can let init(8) exec!  It's off to userland!
717 	 */
718 	mutex_enter(proc_lock);
719 	start_init_exec = 1;
720 	cv_broadcast(&lbolt);
721 	mutex_exit(proc_lock);
722 
723 	/* The scheduler is an infinite loop. */
724 	uvm_scheduler();
725 	/* NOTREACHED */
726 }
727 
728 /*
729  * Configure the system's hardware.
730  */
731 static void
732 configure(void)
733 {
734 
735 	/* Initialize autoconf data structures. */
736 	config_init_mi();
737 	/*
738 	 * XXX
739 	 * callout_setfunc() requires mutex(9) so it can't be in config_init()
740 	 * on amiga and atari which use config_init() and autoconf(9) fucntions
741 	 * to initialize console devices.
742 	 */
743 	config_twiddle_init();
744 
745 	pmf_init();
746 #if NDRVCTL > 0
747 	drvctl_init();
748 #endif
749 
750 	userconf_init();
751 	if (boothowto & RB_USERCONF)
752 		userconf_prompt();
753 
754 	if ((boothowto & (AB_SILENT|AB_VERBOSE)) == AB_SILENT) {
755 		printf_nolog("Detecting hardware...");
756 	}
757 
758 	/*
759 	 * Do the machine-dependent portion of autoconfiguration.  This
760 	 * sets the configuration machinery here in motion by "finding"
761 	 * the root bus.  When this function returns, we expect interrupts
762 	 * to be enabled.
763 	 */
764 	cpu_configure();
765 }
766 
767 static void
768 configure2(void)
769 {
770 	CPU_INFO_ITERATOR cii;
771 	struct cpu_info *ci;
772 	int s;
773 
774 	/*
775 	 * Now that we've found all the hardware, start the real time
776 	 * and statistics clocks.
777 	 */
778 	initclocks();
779 
780 	cold = 0;	/* clocks are running, we're warm now! */
781 	s = splsched();
782 	curcpu()->ci_schedstate.spc_flags |= SPCF_RUNNING;
783 	splx(s);
784 
785 	/* Boot the secondary processors. */
786 	for (CPU_INFO_FOREACH(cii, ci)) {
787 		uvm_cpu_attach(ci);
788 	}
789 	mp_online = true;
790 #if defined(MULTIPROCESSOR)
791 	cpu_boot_secondary_processors();
792 #endif
793 
794 	/* Setup the runqueues and scheduler. */
795 	runq_init();
796 	synch_init();
797 
798 	/*
799 	 * Bus scans can make it appear as if the system has paused, so
800 	 * twiddle constantly while config_interrupts() jobs are running.
801 	 */
802 	config_twiddle_fn(NULL);
803 
804 	/*
805 	 * Create threads to call back and finish configuration for
806 	 * devices that want interrupts enabled.
807 	 */
808 	config_create_interruptthreads();
809 }
810 
811 static void
812 configure3(void)
813 {
814 
815 	/*
816 	 * Create threads to call back and finish configuration for
817 	 * devices that want the mounted root file system.
818 	 */
819 	config_create_mountrootthreads();
820 
821 	/* Get the threads going and into any sleeps before continuing. */
822 	yield();
823 }
824 
825 static void
826 rootconf_handle_wedges(void)
827 {
828 	struct partinfo dpart;
829 	struct partition *p;
830 	struct vnode *vp;
831 	daddr_t startblk;
832 	uint64_t nblks;
833 	device_t dev;
834 	int error;
835 
836 	if (booted_nblks) {
837 		/*
838 		 * bootloader passed geometry
839 		 */
840 		dev      = booted_device;
841 		startblk = booted_startblk;
842 		nblks    = booted_nblks;
843 
844 		/*
845 		 * keep booted_device and booted_partition
846 		 * in case the kernel doesn't identify a wedge
847 		 */
848 	} else {
849 		/*
850 		 * bootloader passed partition number
851 		 *
852 		 * We cannot ask the partition device directly when it is
853 		 * covered by a wedge. Instead we look up the geometry in
854 		 * the disklabel.
855 		 */
856 		vp = opendisk(booted_device);
857 
858 		if (vp == NULL)
859 			return;
860 
861 		error = VOP_IOCTL(vp, DIOCGPART, &dpart, FREAD, NOCRED);
862 		VOP_CLOSE(vp, FREAD, NOCRED);
863 		vput(vp);
864 		if (error)
865 			return;
866 
867 		KASSERT(booted_partition >= 0
868 			&& booted_partition < MAXPARTITIONS);
869 
870 		p = &dpart.disklab->d_partitions[booted_partition];
871 
872 		dev      = booted_device;
873 		startblk = p->p_offset;
874 		nblks    = p->p_size;
875 	}
876 
877 	dev = dkwedge_find_partition(dev, startblk, nblks);
878 	if (dev != NULL) {
879 		booted_device = dev;
880 		booted_partition = 0;
881 	}
882 }
883 
884 void
885 rootconf(void)
886 {
887 	if (booted_device != NULL)
888 		rootconf_handle_wedges();
889 
890 	setroot(booted_device, booted_partition);
891 }
892 
893 static void
894 check_console(struct lwp *l)
895 {
896 	struct vnode *vp;
897 	int error;
898 
899 	error = namei_simple_kernel("/dev/console",
900 				NSM_FOLLOW_NOEMULROOT, &vp);
901 	if (error == 0)
902 		vrele(vp);
903 	else if (error == ENOENT)
904 		printf("warning: no /dev/console\n");
905 	else
906 		printf("warning: lookup /dev/console: error %d\n", error);
907 }
908 
909 /*
910  * List of paths to try when searching for "init".
911  */
912 static const char * const initpaths[] = {
913 	"/sbin/init",
914 	"/sbin/oinit",
915 	"/sbin/init.bak",
916 	"/rescue/init",
917 	NULL,
918 };
919 
920 /*
921  * Start the initial user process; try exec'ing each pathname in "initpaths".
922  * The program is invoked with one argument containing the boot flags.
923  */
924 static void
925 start_init(void *arg)
926 {
927 	struct lwp *l = arg;
928 	struct proc *p = l->l_proc;
929 	vaddr_t addr;
930 	struct sys_execve_args /* {
931 		syscallarg(const char *) path;
932 		syscallarg(char * const *) argp;
933 		syscallarg(char * const *) envp;
934 	} */ args;
935 	int options, i, error;
936 	register_t retval[2];
937 	char flags[4], *flagsp;
938 	const char *path, *slash;
939 	char *ucp, **uap, *arg0, *arg1 = NULL;
940 	char ipath[129];
941 	int ipx, len;
942 
943 	/*
944 	 * Now in process 1.
945 	 */
946 	strncpy(p->p_comm, "init", MAXCOMLEN);
947 
948 	/*
949 	 * Wait for main() to tell us that it's safe to exec.
950 	 */
951 	mutex_enter(proc_lock);
952 	while (start_init_exec == 0)
953 		cv_wait(&lbolt, proc_lock);
954 	mutex_exit(proc_lock);
955 
956 	/*
957 	 * This is not the right way to do this.  We really should
958 	 * hand-craft a descriptor onto /dev/console to hand to init,
959 	 * but that's a _lot_ more work, and the benefit from this easy
960 	 * hack makes up for the "good is the enemy of the best" effect.
961 	 */
962 	check_console(l);
963 
964 	/*
965 	 * Need just enough stack to hold the faked-up "execve()" arguments.
966 	 */
967 	addr = (vaddr_t)STACK_ALLOC(USRSTACK, PAGE_SIZE);
968 	if (uvm_map(&p->p_vmspace->vm_map, &addr, PAGE_SIZE,
969                     NULL, UVM_UNKNOWN_OFFSET, 0,
970                     UVM_MAPFLAG(UVM_PROT_ALL, UVM_PROT_ALL, UVM_INH_COPY,
971 		    UVM_ADV_NORMAL,
972                     UVM_FLAG_FIXED|UVM_FLAG_OVERLAY|UVM_FLAG_COPYONW)) != 0)
973 		panic("init: couldn't allocate argument space");
974 	p->p_vmspace->vm_maxsaddr = (void *)STACK_MAX(addr, PAGE_SIZE);
975 
976 	ipx = 0;
977 	while (1) {
978 		if (boothowto & RB_ASKNAME) {
979 			printf("init path");
980 			if (initpaths[ipx])
981 				printf(" (default %s)", initpaths[ipx]);
982 			printf(": ");
983 			len = cngetsn(ipath, sizeof(ipath)-1);
984 			if (len == 4 && strcmp(ipath, "halt") == 0) {
985 				cpu_reboot(RB_HALT, NULL);
986 			} else if (len == 6 && strcmp(ipath, "reboot") == 0) {
987 				cpu_reboot(0, NULL);
988 #if defined(DDB)
989 			} else if (len == 3 && strcmp(ipath, "ddb") == 0) {
990 				console_debugger();
991 				continue;
992 #endif
993 			} else if (len > 0 && ipath[0] == '/') {
994 				ipath[len] = '\0';
995 				path = ipath;
996 			} else if (len == 0 && initpaths[ipx] != NULL) {
997 				path = initpaths[ipx++];
998 			} else {
999 				printf("use absolute path, ");
1000 #if defined(DDB)
1001 				printf("\"ddb\", ");
1002 #endif
1003 				printf("\"halt\", or \"reboot\"\n");
1004 				continue;
1005 			}
1006 		} else {
1007 			if ((path = initpaths[ipx++]) == NULL) {
1008 				ipx = 0;
1009 				boothowto |= RB_ASKNAME;
1010 				continue;
1011 			}
1012 		}
1013 
1014 		ucp = (char *)USRSTACK;
1015 
1016 		/*
1017 		 * Construct the boot flag argument.
1018 		 */
1019 		flagsp = flags;
1020 		*flagsp++ = '-';
1021 		options = 0;
1022 
1023 		if (boothowto & RB_SINGLE) {
1024 			*flagsp++ = 's';
1025 			options = 1;
1026 		}
1027 #ifdef notyet
1028 		if (boothowto & RB_FASTBOOT) {
1029 			*flagsp++ = 'f';
1030 			options = 1;
1031 		}
1032 #endif
1033 
1034 		/*
1035 		 * Move out the flags (arg 1), if necessary.
1036 		 */
1037 		if (options != 0) {
1038 			*flagsp++ = '\0';
1039 			i = flagsp - flags;
1040 #ifdef DEBUG
1041 			aprint_normal("init: copying out flags `%s' %d\n", flags, i);
1042 #endif
1043 			arg1 = STACK_ALLOC(ucp, i);
1044 			ucp = STACK_MAX(arg1, i);
1045 			(void)copyout((void *)flags, arg1, i);
1046 		}
1047 
1048 		/*
1049 		 * Move out the file name (also arg 0).
1050 		 */
1051 		i = strlen(path) + 1;
1052 #ifdef DEBUG
1053 		aprint_normal("init: copying out path `%s' %d\n", path, i);
1054 #else
1055 		if (boothowto & RB_ASKNAME || path != initpaths[0])
1056 			printf("init: trying %s\n", path);
1057 #endif
1058 		arg0 = STACK_ALLOC(ucp, i);
1059 		ucp = STACK_MAX(arg0, i);
1060 		(void)copyout(path, arg0, i);
1061 
1062 		/*
1063 		 * Move out the arg pointers.
1064 		 */
1065 		ucp = (void *)STACK_ALIGN(ucp, STACK_ALIGNBYTES);
1066 		uap = (char **)STACK_ALLOC(ucp, sizeof(char *) * 3);
1067 		SCARG(&args, path) = arg0;
1068 		SCARG(&args, argp) = uap;
1069 		SCARG(&args, envp) = NULL;
1070 		slash = strrchr(path, '/');
1071 		if (slash)
1072 			(void)suword((void *)uap++,
1073 			    (long)arg0 + (slash + 1 - path));
1074 		else
1075 			(void)suword((void *)uap++, (long)arg0);
1076 		if (options != 0)
1077 			(void)suword((void *)uap++, (long)arg1);
1078 		(void)suword((void *)uap++, 0);	/* terminator */
1079 
1080 		/*
1081 		 * Now try to exec the program.  If can't for any reason
1082 		 * other than it doesn't exist, complain.
1083 		 */
1084 		error = sys_execve(l, &args, retval);
1085 		if (error == 0 || error == EJUSTRETURN) {
1086 			KERNEL_UNLOCK_LAST(l);
1087 			return;
1088 		}
1089 		printf("exec %s: error %d\n", path, error);
1090 	}
1091 	printf("init: not found\n");
1092 	panic("no init");
1093 }
1094 
1095 /*
1096  * calculate cache size (in bytes) from physmem and vsize.
1097  */
1098 vaddr_t
1099 calc_cache_size(vsize_t vsize, int pct, int va_pct)
1100 {
1101 	paddr_t t;
1102 
1103 	/* XXX should consider competing cache if any */
1104 	/* XXX should consider submaps */
1105 	t = (uintmax_t)physmem * pct / 100 * PAGE_SIZE;
1106 	if (vsize != 0) {
1107 		vsize = (uintmax_t)vsize * va_pct / 100;
1108 		if (t > vsize) {
1109 			t = vsize;
1110 		}
1111 	}
1112 	return t;
1113 }
1114 
1115 /*
1116  * Print the system start up banner.
1117  *
1118  * - Print a limited banner if AB_SILENT.
1119  * - Always send normal banner to the log.
1120  */
1121 #define MEM_PBUFSIZE	sizeof("99999 MB")
1122 
1123 void
1124 banner(void)
1125 {
1126 	static char notice[] = " Notice: this software is "
1127 	    "protected by copyright";
1128 	char pbuf[81];
1129 	void (*pr)(const char *, ...) __printflike(1, 2);
1130 	int i;
1131 
1132 	if ((boothowto & AB_SILENT) != 0) {
1133 		snprintf(pbuf, sizeof(pbuf), "%s %s (%s)",
1134 		    ostype, osrelease, kernel_ident);
1135 		printf_nolog("%s", pbuf);
1136 		for (i = 80 - strlen(pbuf) - sizeof(notice); i > 0; i--)
1137 			printf(" ");
1138 		printf_nolog("%s\n", notice);
1139 		pr = aprint_normal;
1140 	} else {
1141 		pr = printf;
1142 	}
1143 
1144 	memset(pbuf, 0, sizeof(pbuf));
1145 	(*pr)("%s%s", copyright, version);
1146 	format_bytes(pbuf, MEM_PBUFSIZE, ctob((uint64_t)physmem));
1147 	(*pr)("total memory = %s\n", pbuf);
1148 	format_bytes(pbuf, MEM_PBUFSIZE, ctob((uint64_t)uvmexp.free));
1149 	(*pr)("avail memory = %s\n", pbuf);
1150 }
1151