xref: /netbsd-src/sys/kern/init_main.c (revision 062cd249892f31a10349de47b78d800cd8ae2c25)
1 /*	$NetBSD: init_main.c,v 1.444 2012/06/14 20:18:16 martin 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 1.444 2012/06/14 20:18:16 martin 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 
116 #include "drvctl.h"
117 #include "ksyms.h"
118 
119 #include "sysmon_envsys.h"
120 #include "sysmon_power.h"
121 #include "sysmon_taskq.h"
122 #include "sysmon_wdog.h"
123 #include "veriexec.h"
124 
125 #include <sys/param.h>
126 #include <sys/acct.h>
127 #include <sys/filedesc.h>
128 #include <sys/file.h>
129 #include <sys/errno.h>
130 #include <sys/callout.h>
131 #include <sys/cpu.h>
132 #include <sys/cpufreq.h>
133 #include <sys/spldebug.h>
134 #include <sys/kernel.h>
135 #include <sys/mount.h>
136 #include <sys/proc.h>
137 #include <sys/kthread.h>
138 #include <sys/resourcevar.h>
139 #include <sys/signalvar.h>
140 #include <sys/systm.h>
141 #include <sys/vnode.h>
142 #include <sys/fstrans.h>
143 #include <sys/tty.h>
144 #include <sys/conf.h>
145 #include <sys/disklabel.h>
146 #include <sys/buf.h>
147 #include <sys/device.h>
148 #include <sys/exec.h>
149 #include <sys/socketvar.h>
150 #include <sys/protosw.h>
151 #include <sys/percpu.h>
152 #include <sys/pserialize.h>
153 #include <sys/pset.h>
154 #include <sys/sysctl.h>
155 #include <sys/reboot.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/kcpuset.h>
171 #include <sys/ksyms.h>
172 #include <sys/uidinfo.h>
173 #include <sys/kprintf.h>
174 #ifdef FAST_IPSEC
175 #include <netipsec/ipsec.h>
176 #endif
177 #ifdef SYSVSHM
178 #include <sys/shm.h>
179 #endif
180 #ifdef SYSVSEM
181 #include <sys/sem.h>
182 #endif
183 #ifdef SYSVMSG
184 #include <sys/msg.h>
185 #endif
186 #include <sys/domain.h>
187 #include <sys/namei.h>
188 #include <sys/rnd.h>
189 #include <sys/pipe.h>
190 #if NVERIEXEC > 0
191 #include <sys/verified_exec.h>
192 #endif /* NVERIEXEC > 0 */
193 #ifdef KTRACE
194 #include <sys/ktrace.h>
195 #endif
196 #include <sys/kauth.h>
197 #include <net80211/ieee80211_netbsd.h>
198 #ifdef PTRACE
199 #include <sys/ptrace.h>
200 #endif /* PTRACE */
201 #include <sys/cprng.h>
202 
203 #include <sys/syscall.h>
204 #include <sys/syscallargs.h>
205 
206 #if defined(PAX_MPROTECT) || defined(PAX_SEGVGUARD) || defined(PAX_ASLR)
207 #include <sys/pax.h>
208 #endif /* PAX_MPROTECT || PAX_SEGVGUARD || PAX_ASLR */
209 
210 #include <secmodel/secmodel.h>
211 
212 #include <ufs/ufs/quota.h>
213 
214 #include <miscfs/genfs/genfs.h>
215 #include <miscfs/syncfs/syncfs.h>
216 #include <miscfs/specfs/specdev.h>
217 
218 #include <sys/cpu.h>
219 
220 #include <uvm/uvm.h>	/* extern struct uvm uvm */
221 
222 #if NSYSMON_TASKQ > 0
223 #include <dev/sysmon/sysmon_taskq.h>
224 #endif
225 
226 #include <dev/cons.h>
227 
228 #if NSYSMON_ENVSYS > 0 || NSYSMON_POWER > 0 || NSYSMON_WDOG > 0
229 #include <dev/sysmon/sysmonvar.h>
230 #endif
231 
232 #include <net/bpf.h>
233 #include <net/if.h>
234 #include <net/raw_cb.h>
235 
236 #include <prop/proplib.h>
237 
238 #ifdef COMPAT_50
239 #include <compat/sys/time.h>
240 struct timeval50 boottime50;
241 #endif
242 
243 #include <sys/userconf.h>
244 
245 extern struct lwp lwp0;
246 extern time_t rootfstime;
247 
248 #ifndef curlwp
249 struct	lwp *curlwp = &lwp0;
250 #endif
251 struct	proc *initproc;
252 
253 struct	vnode *rootvp, *swapdev_vp;
254 int	boothowto;
255 int	cold = 1;			/* still working on startup */
256 struct timespec boottime;	        /* time at system startup - will only follow settime deltas */
257 
258 int	start_init_exec;		/* semaphore for start_init() */
259 
260 cprng_strong_t	*kern_cprng;
261 
262 static void rootconf(void);
263 static void check_console(struct lwp *l);
264 static void start_init(void *);
265 static void configure(void);
266 static void configure2(void);
267 static void configure3(void);
268 void main(void);
269 
270 /*
271  * System startup; initialize the world, create process 0, mount root
272  * filesystem, and fork to create init and pagedaemon.  Most of the
273  * hard work is done in the lower-level initialization routines including
274  * startup(), which does memory initialization and autoconfiguration.
275  */
276 void
277 main(void)
278 {
279 	struct timespec time;
280 	struct lwp *l;
281 	struct proc *p;
282 	int s, error;
283 #ifdef NVNODE_IMPLICIT
284 	int usevnodes;
285 #endif
286 	CPU_INFO_ITERATOR cii;
287 	struct cpu_info *ci;
288 
289 	l = &lwp0;
290 #ifndef LWP0_CPU_INFO
291 	l->l_cpu = curcpu();
292 #endif
293 	l->l_pflag |= LP_RUNNING;
294 
295 	/*
296 	 * Attempt to find console and initialize
297 	 * in case of early panic or other messages.
298 	 */
299 	consinit();
300 
301 	kernel_lock_init();
302 	once_init();
303 
304 	mi_cpu_init();
305 	kernconfig_lock_init();
306 	kthread_sysinit();
307 
308 	/* Initialize the device switch tables. */
309 	devsw_init();
310 
311 	/* Initialize event counters. */
312 	evcnt_init();
313 
314 	uvm_init();
315 	kcpuset_sysinit();
316 
317 	prop_kern_init();
318 
319 #if ((NKSYMS > 0) || (NDDB > 0) || (NMODULAR > 0))
320 	ksyms_init();
321 #endif
322 	kprintf_init();
323 
324 	percpu_init();
325 
326 	/* Initialize lock caches. */
327 	mutex_obj_init();
328 	rw_obj_init();
329 
330 	/* Passive serialization. */
331 	pserialize_init();
332 
333 	/* Initialize the extent manager. */
334 	extent_init();
335 
336 	/* Do machine-dependent initialization. */
337 	cpu_startup();
338 
339 	/* Initialize the sysctl subsystem. */
340 	sysctl_init();
341 
342 	/* Initialize callouts, part 1. */
343 	callout_startup();
344 
345 	/* Initialize the kernel authorization subsystem. */
346 	kauth_init();
347 
348 	secmodel_init();
349 
350 	spec_init();
351 
352 	/*
353 	 * Set BPF op vector.  Can't do this in bpf attach, since
354 	 * network drivers attach before bpf.
355 	 */
356 	bpf_setops();
357 
358 	/* Start module system. */
359 	module_init();
360 
361 	/*
362 	 * Initialize the kernel authorization subsystem and start the
363 	 * default security model, if any. We need to do this early
364 	 * enough so that subsystems relying on any of the aforementioned
365 	 * can work properly. Since the security model may dictate the
366 	 * credential inheritance policy, it is needed at least before
367 	 * any process is created, specifically proc0.
368 	 */
369 	module_init_class(MODULE_CLASS_SECMODEL);
370 
371 	/* Initialize the buffer cache */
372 	bufinit();
373 
374 	/* Initialize sockets. */
375 	soinit();
376 
377 	/*
378 	 * The following things must be done before autoconfiguration.
379 	 */
380 	rnd_init();		/* initialize entropy pool */
381 
382 	cprng_init();		/* initialize cryptographic PRNG */
383 
384 	/* Initialize process and pgrp structures. */
385 	procinit();
386 	lwpinit();
387 
388 	/* Initialize signal-related data structures. */
389 	signal_init();
390 
391 	/* Initialize resource management. */
392 	resource_init();
393 
394 	/* Create process 0. */
395 	proc0_init();
396 	lwp0_init();
397 
398 	/* Disable preemption during boot. */
399 	kpreempt_disable();
400 
401 	/* Initialize the UID hash table. */
402 	uid_init();
403 
404 	/* Charge root for one process. */
405 	(void)chgproccnt(0, 1);
406 
407 	/* Initialize timekeeping. */
408 	time_init();
409 
410 	/* Initialize the run queues, turnstiles and sleep queues. */
411 	sched_rqinit();
412 	turnstile_init();
413 	sleeptab_init(&sleeptab);
414 
415 	sched_init();
416 
417 	/* Initialize processor-sets */
418 	psets_init();
419 
420 	/* Initialize cpufreq(9) */
421 	cpufreq_init();
422 
423 	/* MI initialization of the boot cpu */
424 	error = mi_cpu_attach(curcpu());
425 	KASSERT(error == 0);
426 
427 	/* Initialize timekeeping, part 2. */
428 	time_init2();
429 
430 	/*
431 	 * Initialize mbuf's.  Do this now because we might attempt to
432 	 * allocate mbufs or mbuf clusters during autoconfiguration.
433 	 */
434 	mbinit();
435 
436 	/* Initialize I/O statistics. */
437 	iostat_init();
438 
439 	/* Initialize the log device. */
440 	loginit();
441 
442 	/* Second part of module system initialization. */
443 	module_start_unload_thread();
444 
445 	/* Initialize the file systems. */
446 #ifdef NVNODE_IMPLICIT
447 	/*
448 	 * If maximum number of vnodes in namei vnode cache is not explicitly
449 	 * defined in kernel config, adjust the number such as we use roughly
450 	 * 10% of memory for vnodes and associated data structures in the
451 	 * assumed worst case.  Do not provide fewer than NVNODE vnodes.
452 	 */
453 	usevnodes =
454 	    calc_cache_size(kernel_map, 10, VNODE_VA_MAXPCT) / VNODE_COST;
455 	if (usevnodes > desiredvnodes)
456 		desiredvnodes = usevnodes;
457 #endif
458 	vfsinit();
459 	lf_init();
460 
461 	/* Initialize fstrans. */
462 	fstrans_init();
463 
464 	/* Initialize the file descriptor system. */
465 	fd_sys_init();
466 
467 	/* Initialize cwd structures */
468 	cwd_sys_init();
469 
470 	/* Initialize kqueue. */
471 	kqueue_init();
472 
473 	/* Initialize the system monitor subsystems. */
474 #if NSYSMON_TASKQ > 0
475 	sysmon_task_queue_preinit();
476 #endif
477 
478 #if NSYSMON_ENVSYS > 0
479 	sysmon_envsys_init();
480 #endif
481 
482 #if NSYSMON_POWER > 0
483 	sysmon_power_init();
484 #endif
485 
486 #if NSYSMON_WDOG > 0
487 	sysmon_wdog_init();
488 #endif
489 
490 	inittimecounter();
491 	ntp_init();
492 
493 	/* Initialize tty subsystem. */
494 	tty_init();
495 	ttyldisc_init();
496 
497 	/* Initialize the buffer cache, part 2. */
498 	bufinit2();
499 
500 	/* Initialize the disk wedge subsystem. */
501 	dkwedge_init();
502 
503 	/* Initialize the kernel strong PRNG. */
504 	kern_cprng = cprng_strong_create("kernel", IPL_VM,
505 					 CPRNG_INIT_ANY|CPRNG_REKEY_ANY);
506 
507 	/* Initialize interfaces. */
508 	ifinit1();
509 
510 	spldebug_start();
511 
512 	/* Initialize sockets thread(s) */
513 	soinit1();
514 
515 	/* Configure the system hardware.  This will enable interrupts. */
516 	configure();
517 
518 	ssp_init();
519 
520 	ubc_init();		/* must be after autoconfig */
521 
522 	mm_init();
523 
524 	configure2();
525 	/* Now timer is working.  Enable preemption. */
526 	kpreempt_enable();
527 
528 #ifdef SYSVSHM
529 	/* Initialize System V style shared memory. */
530 	shminit();
531 #endif
532 
533 	vmem_rehash_start();	/* must be before exec_init */
534 
535 	/* Initialize exec structures */
536 	exec_init(1);		/* seminit calls exithook_establish() */
537 
538 #ifdef SYSVSEM
539 	/* Initialize System V style semaphores. */
540 	seminit();
541 #endif
542 
543 #ifdef SYSVMSG
544 	/* Initialize System V style message queues. */
545 	msginit();
546 #endif
547 
548 #if NVERIEXEC > 0
549 	/*
550 	 * Initialise the Veriexec subsystem.
551 	 */
552 	veriexec_init();
553 #endif /* NVERIEXEC > 0 */
554 
555 #if defined(PAX_MPROTECT) || defined(PAX_SEGVGUARD) || defined(PAX_ASLR)
556 	pax_init();
557 #endif /* PAX_MPROTECT || PAX_SEGVGUARD || PAX_ASLR */
558 
559 #ifdef	FAST_IPSEC
560 	/* Attach network crypto subsystem */
561 	ipsec_attach();
562 #endif
563 
564 	/*
565 	 * Initialize protocols.  Block reception of incoming packets
566 	 * until everything is ready.
567 	 */
568 	s = splnet();
569 	ifinit();
570 	domaininit(true);
571 	if_attachdomain();
572 	splx(s);
573 
574 #ifdef GPROF
575 	/* Initialize kernel profiling. */
576 	kmstartup();
577 #endif
578 
579 	/* Initialize system accounting. */
580 	acct_init();
581 
582 #ifndef PIPE_SOCKETPAIR
583 	/* Initialize pipes. */
584 	pipe_init();
585 #endif
586 
587 #ifdef KTRACE
588 	/* Initialize ktrace. */
589 	ktrinit();
590 #endif
591 
592 #ifdef PTRACE
593 	/* Initialize ptrace. */
594 	ptrace_init();
595 #endif /* PTRACE */
596 
597 	/* Initialize the UUID system calls. */
598 	uuid_init();
599 
600 	machdep_init();
601 
602 	procinit_sysctl();
603 
604 	/*
605 	 * Create process 1 (init(8)).  We do this now, as Unix has
606 	 * historically had init be process 1, and changing this would
607 	 * probably upset a lot of people.
608 	 *
609 	 * Note that process 1 won't immediately exec init(8), but will
610 	 * wait for us to inform it that the root file system has been
611 	 * mounted.
612 	 */
613 	if (fork1(l, 0, SIGCHLD, NULL, 0, start_init, NULL, NULL, &initproc))
614 		panic("fork init");
615 
616 	/*
617 	 * Load any remaining builtin modules, and hand back temporary
618 	 * storage to the VM system.  Then require force when loading any
619 	 * remaining un-init'ed built-in modules to avoid later surprises.
620 	 */
621 	module_init_class(MODULE_CLASS_ANY);
622 	module_builtin_require_force();
623 
624 	/*
625 	 * Finalize configuration now that all real devices have been
626 	 * found.  This needs to be done before the root device is
627 	 * selected, since finalization may create the root device.
628 	 */
629 	config_finalize();
630 
631 	sysctl_finalize();
632 
633 	/*
634 	 * Now that autoconfiguration has completed, we can determine
635 	 * the root and dump devices.
636 	 */
637 	rootconf();
638 	cpu_dumpconf();
639 
640 	/* Mount the root file system. */
641 	do {
642 		domountroothook(root_device);
643 		if ((error = vfs_mountroot())) {
644 			printf("cannot mount root, error = %d\n", error);
645 			boothowto |= RB_ASKNAME;
646 			setroot(root_device,
647 			    (rootdev != NODEV) ? DISKPART(rootdev) : 0);
648 		}
649 	} while (error != 0);
650 	mountroothook_destroy();
651 
652 	configure3();
653 
654 	/*
655 	 * Initialise the time-of-day clock, passing the time recorded
656 	 * in the root filesystem (if any) for use by systems that
657 	 * don't have a non-volatile time-of-day device.
658 	 */
659 	inittodr(rootfstime);
660 
661 	/*
662 	 * Now can look at time, having had a chance to verify the time
663 	 * from the file system.  Reset l->l_rtime as it may have been
664 	 * munched in mi_switch() after the time got set.
665 	 */
666 	getnanotime(&time);
667 	boottime = time;
668 #ifdef COMPAT_50
669 	{
670 		struct timeval tv;
671 		TIMESPEC_TO_TIMEVAL(&tv, &time);
672 		timeval_to_timeval50(&tv, &boottime50);
673 	}
674 #endif
675 	mutex_enter(proc_lock);
676 	LIST_FOREACH(p, &allproc, p_list) {
677 		KASSERT((p->p_flag & PK_MARKER) == 0);
678 		mutex_enter(p->p_lock);
679 		TIMESPEC_TO_TIMEVAL(&p->p_stats->p_start, &time);
680 		LIST_FOREACH(l, &p->p_lwps, l_sibling) {
681 			lwp_lock(l);
682 			memset(&l->l_rtime, 0, sizeof(l->l_rtime));
683 			lwp_unlock(l);
684 		}
685 		mutex_exit(p->p_lock);
686 	}
687 	mutex_exit(proc_lock);
688 	binuptime(&curlwp->l_stime);
689 
690 	for (CPU_INFO_FOREACH(cii, ci)) {
691 		ci->ci_schedstate.spc_lastmod = time_second;
692 	}
693 
694 	/* Create the pageout daemon kernel thread. */
695 	uvm_swap_init();
696 	if (kthread_create(PRI_PGDAEMON, KTHREAD_MPSAFE, NULL, uvm_pageout,
697 	    NULL, NULL, "pgdaemon"))
698 		panic("fork pagedaemon");
699 
700 	/* Create the filesystem syncer kernel thread. */
701 	if (kthread_create(PRI_IOFLUSH, KTHREAD_MPSAFE, NULL, sched_sync,
702 	    NULL, NULL, "ioflush"))
703 		panic("fork syncer");
704 
705 	/* Create the aiodone daemon kernel thread. */
706 	if (workqueue_create(&uvm.aiodone_queue, "aiodoned",
707 	    uvm_aiodone_worker, NULL, PRI_VM, IPL_NONE, WQ_MPSAFE))
708 		panic("fork aiodoned");
709 
710 	/*
711 	 * Okay, now we can let init(8) exec!  It's off to userland!
712 	 */
713 	mutex_enter(proc_lock);
714 	start_init_exec = 1;
715 	cv_broadcast(&lbolt);
716 	mutex_exit(proc_lock);
717 
718 	/* The scheduler is an infinite loop. */
719 	uvm_scheduler();
720 	/* NOTREACHED */
721 }
722 
723 /*
724  * Configure the system's hardware.
725  */
726 static void
727 configure(void)
728 {
729 
730 	/* Initialize autoconf data structures. */
731 	config_init_mi();
732 	/*
733 	 * XXX
734 	 * callout_setfunc() requires mutex(9) so it can't be in config_init()
735 	 * on amiga and atari which use config_init() and autoconf(9) fucntions
736 	 * to initialize console devices.
737 	 */
738 	config_twiddle_init();
739 
740 	pmf_init();
741 #if NDRVCTL > 0
742 	drvctl_init();
743 #endif
744 
745 	userconf_init();
746 	if (boothowto & RB_USERCONF)
747 		userconf_prompt();
748 
749 	if ((boothowto & (AB_SILENT|AB_VERBOSE)) == AB_SILENT) {
750 		printf_nolog("Detecting hardware...");
751 	}
752 
753 	/*
754 	 * Do the machine-dependent portion of autoconfiguration.  This
755 	 * sets the configuration machinery here in motion by "finding"
756 	 * the root bus.  When this function returns, we expect interrupts
757 	 * to be enabled.
758 	 */
759 	cpu_configure();
760 }
761 
762 static void
763 configure2(void)
764 {
765 	CPU_INFO_ITERATOR cii;
766 	struct cpu_info *ci;
767 	int s;
768 
769 	/*
770 	 * Now that we've found all the hardware, start the real time
771 	 * and statistics clocks.
772 	 */
773 	initclocks();
774 
775 	cold = 0;	/* clocks are running, we're warm now! */
776 	s = splsched();
777 	curcpu()->ci_schedstate.spc_flags |= SPCF_RUNNING;
778 	splx(s);
779 
780 	/* Boot the secondary processors. */
781 	for (CPU_INFO_FOREACH(cii, ci)) {
782 		uvm_cpu_attach(ci);
783 	}
784 	mp_online = true;
785 #if defined(MULTIPROCESSOR)
786 	cpu_boot_secondary_processors();
787 #endif
788 
789 	/* Setup the runqueues and scheduler. */
790 	runq_init();
791 	synch_init();
792 
793 	/*
794 	 * Bus scans can make it appear as if the system has paused, so
795 	 * twiddle constantly while config_interrupts() jobs are running.
796 	 */
797 	config_twiddle_fn(NULL);
798 
799 	/*
800 	 * Create threads to call back and finish configuration for
801 	 * devices that want interrupts enabled.
802 	 */
803 	config_create_interruptthreads();
804 
805 	/* Get the threads going and into any sleeps before continuing. */
806 	yield();
807 }
808 
809 static void
810 configure3(void)
811 {
812 
813 	/*
814 	 * Create threads to call back and finish configuration for
815 	 * devices that want the mounted root file system.
816 	 */
817 	config_create_mountrootthreads();
818 
819 	/* Get the threads going and into any sleeps before continuing. */
820 	yield();
821 }
822 
823 static void
824 rootconf_handle_wedges(void)
825 {
826 	struct partinfo dpart;
827 	struct partition *p;
828 	struct vnode *vp;
829 	daddr_t startblk;
830 	uint64_t nblks;
831 	device_t dev;
832 	int error;
833 
834 	if (booted_nblks) {
835 		/*
836 		 * bootloader passed geometry
837 		 */
838 		dev      = booted_device;
839 		startblk = booted_startblk;
840 		nblks    = booted_nblks;
841 
842 		/*
843 		 * keep booted_device and booted_partition
844 		 * in case the kernel doesn't identify a wedge
845 		 */
846 	} else {
847 		/*
848 		 * bootloader passed partition number
849 		 *
850 		 * We cannot ask the partition device directly when it is
851 		 * covered by a wedge. Instead we look up the geometry in
852 		 * the disklabel.
853 		 */
854 		vp = opendisk(booted_device);
855 
856 		if (vp == NULL)
857 			return;
858 
859 		error = VOP_IOCTL(vp, DIOCGPART, &dpart, FREAD, NOCRED);
860 		VOP_CLOSE(vp, FREAD, NOCRED);
861 		vput(vp);
862 		if (error)
863 			return;
864 
865 		KASSERT(booted_partition >= 0
866 			&& booted_partition < MAXPARTITIONS);
867 
868 		p = &dpart.disklab->d_partitions[booted_partition];
869 
870 		dev      = booted_device;
871 		startblk = p->p_offset;
872 		nblks    = p->p_size;
873 	}
874 
875 	dev = dkwedge_find_partition(dev, startblk, nblks);
876 	if (dev != NULL) {
877 		booted_device = dev;
878 		booted_partition = 0;
879 	}
880 }
881 
882 static void
883 rootconf(void)
884 {
885 	cpu_rootconf();
886 
887 	if (booted_device == NULL)
888 		return;
889 
890 	rootconf_handle_wedges();
891 
892 	setroot(booted_device, booted_partition);
893 }
894 
895 static void
896 check_console(struct lwp *l)
897 {
898 	struct vnode *vp;
899 	int error;
900 
901 	error = namei_simple_kernel("/dev/console",
902 				NSM_FOLLOW_NOEMULROOT, &vp);
903 	if (error == 0)
904 		vrele(vp);
905 	else if (error == ENOENT)
906 		printf("warning: no /dev/console\n");
907 	else
908 		printf("warning: lookup /dev/console: error %d\n", error);
909 }
910 
911 /*
912  * List of paths to try when searching for "init".
913  */
914 static const char * const initpaths[] = {
915 	"/sbin/init",
916 	"/sbin/oinit",
917 	"/sbin/init.bak",
918 	NULL,
919 };
920 
921 /*
922  * Start the initial user process; try exec'ing each pathname in "initpaths".
923  * The program is invoked with one argument containing the boot flags.
924  */
925 static void
926 start_init(void *arg)
927 {
928 	struct lwp *l = arg;
929 	struct proc *p = l->l_proc;
930 	vaddr_t addr;
931 	struct sys_execve_args /* {
932 		syscallarg(const char *) path;
933 		syscallarg(char * const *) argp;
934 		syscallarg(char * const *) envp;
935 	} */ args;
936 	int options, i, error;
937 	register_t retval[2];
938 	char flags[4], *flagsp;
939 	const char *path, *slash;
940 	char *ucp, **uap, *arg0, *arg1 = NULL;
941 	char ipath[129];
942 	int ipx, len;
943 
944 	/*
945 	 * Now in process 1.
946 	 */
947 	strncpy(p->p_comm, "init", MAXCOMLEN);
948 
949 	/*
950 	 * Wait for main() to tell us that it's safe to exec.
951 	 */
952 	mutex_enter(proc_lock);
953 	while (start_init_exec == 0)
954 		cv_wait(&lbolt, proc_lock);
955 	mutex_exit(proc_lock);
956 
957 	/*
958 	 * This is not the right way to do this.  We really should
959 	 * hand-craft a descriptor onto /dev/console to hand to init,
960 	 * but that's a _lot_ more work, and the benefit from this easy
961 	 * hack makes up for the "good is the enemy of the best" effect.
962 	 */
963 	check_console(l);
964 
965 	/*
966 	 * Need just enough stack to hold the faked-up "execve()" arguments.
967 	 */
968 	addr = (vaddr_t)STACK_ALLOC(USRSTACK, PAGE_SIZE);
969 	if (uvm_map(&p->p_vmspace->vm_map, &addr, PAGE_SIZE,
970                     NULL, UVM_UNKNOWN_OFFSET, 0,
971                     UVM_MAPFLAG(UVM_PROT_ALL, UVM_PROT_ALL, UVM_INH_COPY,
972 		    UVM_ADV_NORMAL,
973                     UVM_FLAG_FIXED|UVM_FLAG_OVERLAY|UVM_FLAG_COPYONW)) != 0)
974 		panic("init: couldn't allocate argument space");
975 	p->p_vmspace->vm_maxsaddr = (void *)STACK_MAX(addr, PAGE_SIZE);
976 
977 	ipx = 0;
978 	while (1) {
979 		if (boothowto & RB_ASKNAME) {
980 			printf("init path");
981 			if (initpaths[ipx])
982 				printf(" (default %s)", initpaths[ipx]);
983 			printf(": ");
984 			len = cngetsn(ipath, sizeof(ipath)-1);
985 			if (len == 4 && strcmp(ipath, "halt") == 0) {
986 				cpu_reboot(RB_HALT, NULL);
987 			} else if (len == 6 && strcmp(ipath, "reboot") == 0) {
988 				cpu_reboot(0, NULL);
989 #if defined(DDB)
990 			} else if (len == 3 && strcmp(ipath, "ddb") == 0) {
991 				console_debugger();
992 				continue;
993 #endif
994 			} else if (len > 0 && ipath[0] == '/') {
995 				ipath[len] = '\0';
996 				path = ipath;
997 			} else if (len == 0 && initpaths[ipx] != NULL) {
998 				path = initpaths[ipx++];
999 			} else {
1000 				printf("use absolute path, ");
1001 #if defined(DDB)
1002 				printf("\"ddb\", ");
1003 #endif
1004 				printf("\"halt\", or \"reboot\"\n");
1005 				continue;
1006 			}
1007 		} else {
1008 			if ((path = initpaths[ipx++]) == NULL) {
1009 				ipx = 0;
1010 				boothowto |= RB_ASKNAME;
1011 				continue;
1012 			}
1013 		}
1014 
1015 		ucp = (char *)USRSTACK;
1016 
1017 		/*
1018 		 * Construct the boot flag argument.
1019 		 */
1020 		flagsp = flags;
1021 		*flagsp++ = '-';
1022 		options = 0;
1023 
1024 		if (boothowto & RB_SINGLE) {
1025 			*flagsp++ = 's';
1026 			options = 1;
1027 		}
1028 #ifdef notyet
1029 		if (boothowto & RB_FASTBOOT) {
1030 			*flagsp++ = 'f';
1031 			options = 1;
1032 		}
1033 #endif
1034 
1035 		/*
1036 		 * Move out the flags (arg 1), if necessary.
1037 		 */
1038 		if (options != 0) {
1039 			*flagsp++ = '\0';
1040 			i = flagsp - flags;
1041 #ifdef DEBUG
1042 			aprint_normal("init: copying out flags `%s' %d\n", flags, i);
1043 #endif
1044 			arg1 = STACK_ALLOC(ucp, i);
1045 			ucp = STACK_MAX(arg1, i);
1046 			(void)copyout((void *)flags, arg1, i);
1047 		}
1048 
1049 		/*
1050 		 * Move out the file name (also arg 0).
1051 		 */
1052 		i = strlen(path) + 1;
1053 #ifdef DEBUG
1054 		aprint_normal("init: copying out path `%s' %d\n", path, i);
1055 #else
1056 		if (boothowto & RB_ASKNAME || path != initpaths[0])
1057 			printf("init: trying %s\n", path);
1058 #endif
1059 		arg0 = STACK_ALLOC(ucp, i);
1060 		ucp = STACK_MAX(arg0, i);
1061 		(void)copyout(path, arg0, i);
1062 
1063 		/*
1064 		 * Move out the arg pointers.
1065 		 */
1066 		ucp = (void *)STACK_ALIGN(ucp, STACK_ALIGNBYTES);
1067 		uap = (char **)STACK_ALLOC(ucp, sizeof(char *) * 3);
1068 		SCARG(&args, path) = arg0;
1069 		SCARG(&args, argp) = uap;
1070 		SCARG(&args, envp) = NULL;
1071 		slash = strrchr(path, '/');
1072 		if (slash)
1073 			(void)suword((void *)uap++,
1074 			    (long)arg0 + (slash + 1 - path));
1075 		else
1076 			(void)suword((void *)uap++, (long)arg0);
1077 		if (options != 0)
1078 			(void)suword((void *)uap++, (long)arg1);
1079 		(void)suword((void *)uap++, 0);	/* terminator */
1080 
1081 		/*
1082 		 * Now try to exec the program.  If can't for any reason
1083 		 * other than it doesn't exist, complain.
1084 		 */
1085 		error = sys_execve(l, &args, retval);
1086 		if (error == 0 || error == EJUSTRETURN) {
1087 			KERNEL_UNLOCK_LAST(l);
1088 			return;
1089 		}
1090 		printf("exec %s: error %d\n", path, error);
1091 	}
1092 	printf("init: not found\n");
1093 	panic("no init");
1094 }
1095 
1096 /*
1097  * calculate cache size (in bytes) from physmem and vm_map size.
1098  */
1099 vaddr_t
1100 calc_cache_size(struct vm_map *map, int pct, int va_pct)
1101 {
1102 	paddr_t t;
1103 
1104 	/* XXX should consider competing cache if any */
1105 	/* XXX should consider submaps */
1106 	t = (uintmax_t)physmem * pct / 100 * PAGE_SIZE;
1107 	if (map != NULL) {
1108 		vsize_t vsize;
1109 
1110 		vsize = vm_map_max(map) - vm_map_min(map);
1111 		vsize = (uintmax_t)vsize * va_pct / 100;
1112 		if (t > vsize) {
1113 			t = vsize;
1114 		}
1115 	}
1116 	return t;
1117 }
1118 
1119 /*
1120  * Print the system start up banner.
1121  *
1122  * - Print a limited banner if AB_SILENT.
1123  * - Always send normal banner to the log.
1124  */
1125 #define MEM_PBUFSIZE	sizeof("99999 MB")
1126 
1127 void
1128 banner(void)
1129 {
1130 	static char notice[] = " Notice: this software is "
1131 	    "protected by copyright";
1132 	char pbuf[81];
1133 	void (*pr)(const char *, ...);
1134 	int i;
1135 
1136 	if ((boothowto & AB_SILENT) != 0) {
1137 		snprintf(pbuf, sizeof(pbuf), "%s %s (%s)",
1138 		    ostype, osrelease, kernel_ident);
1139 		printf_nolog("%s", pbuf);
1140 		for (i = 80 - strlen(pbuf) - sizeof(notice); i > 0; i--)
1141 			printf(" ");
1142 		printf_nolog("%s\n", notice);
1143 		pr = aprint_normal;
1144 	} else {
1145 		pr = printf;
1146 	}
1147 
1148 	memset(pbuf, 0, sizeof(pbuf));
1149 	(*pr)("%s%s", copyright, version);
1150 	format_bytes(pbuf, MEM_PBUFSIZE, ctob((uint64_t)physmem));
1151 	(*pr)("total memory = %s\n", pbuf);
1152 	format_bytes(pbuf, MEM_PBUFSIZE, ctob((uint64_t)uvmexp.free));
1153 	(*pr)("avail memory = %s\n", pbuf);
1154 }
1155