xref: /netbsd-src/sys/kern/init_main.c (revision a0698ed9d41653d7a2378819ad501a285ca0d401)
1 /*	$NetBSD: init_main.c,v 1.503 2019/02/20 10:07:27 hannken 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 1.503 2019/02/20 10:07:27 hannken Exp $");
101 
102 #include "opt_ddb.h"
103 #include "opt_inet.h"
104 #include "opt_ipsec.h"
105 #include "opt_modular.h"
106 #include "opt_ntp.h"
107 #include "opt_pipe.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 #include "opt_rnd_printf.h"
117 #include "opt_splash.h"
118 #include "opt_kernhist.h"
119 #include "opt_gprof.h"
120 
121 #if defined(SPLASHSCREEN) && defined(makeoptions_SPLASHSCREEN_IMAGE)
122 extern void *_binary_splash_image_start;
123 extern void *_binary_splash_image_end;
124 #endif
125 
126 #include "ksyms.h"
127 
128 #include "veriexec.h"
129 
130 #include <sys/param.h>
131 #include <sys/acct.h>
132 #include <sys/filedesc.h>
133 #include <sys/file.h>
134 #include <sys/errno.h>
135 #include <sys/callout.h>
136 #include <sys/cpu.h>
137 #include <sys/cpufreq.h>
138 #include <sys/spldebug.h>
139 #include <sys/kernel.h>
140 #include <sys/mount.h>
141 #include <sys/proc.h>
142 #include <sys/kthread.h>
143 #include <sys/resourcevar.h>
144 #include <sys/signalvar.h>
145 #include <sys/systm.h>
146 #include <sys/vnode.h>
147 #include <sys/fstrans.h>
148 #include <sys/tty.h>
149 #include <sys/conf.h>
150 #include <sys/disklabel.h>
151 #include <sys/buf.h>
152 #include <sys/device.h>
153 #include <sys/exec.h>
154 #include <sys/socketvar.h>
155 #include <sys/protosw.h>
156 #include <sys/percpu.h>
157 #include <sys/pserialize.h>
158 #include <sys/pset.h>
159 #include <sys/sysctl.h>
160 #include <sys/reboot.h>
161 #include <sys/event.h>
162 #include <sys/mbuf.h>
163 #include <sys/sched.h>
164 #include <sys/sleepq.h>
165 #include <sys/ipi.h>
166 #include <sys/iostat.h>
167 #include <sys/vmem.h>
168 #include <sys/uuid.h>
169 #include <sys/extent.h>
170 #include <sys/disk.h>
171 #include <sys/msgbuf.h>
172 #include <sys/module.h>
173 #include <sys/event.h>
174 #include <sys/lockf.h>
175 #include <sys/once.h>
176 #include <sys/kcpuset.h>
177 #include <sys/ksyms.h>
178 #include <sys/uidinfo.h>
179 #include <sys/kprintf.h>
180 #include <sys/bufq.h>
181 #include <sys/threadpool.h>
182 #ifdef IPSEC
183 #include <netipsec/ipsec.h>
184 #endif
185 #include <sys/domain.h>
186 #include <sys/namei.h>
187 #include <sys/rnd.h>
188 #include <sys/pipe.h>
189 #if NVERIEXEC > 0
190 #include <sys/verified_exec.h>
191 #endif /* NVERIEXEC > 0 */
192 #ifdef KTRACE
193 #include <sys/ktrace.h>
194 #endif
195 #include <sys/kauth.h>
196 #include <net80211/ieee80211_netbsd.h>
197 #include <sys/cprng.h>
198 
199 #include <sys/syscall.h>
200 #include <sys/syscallargs.h>
201 
202 #include <sys/pax.h>
203 
204 #include <secmodel/secmodel.h>
205 
206 #include <ufs/ufs/quota.h>
207 
208 #include <miscfs/genfs/genfs.h>
209 #include <miscfs/specfs/specdev.h>
210 
211 #include <sys/cpu.h>
212 
213 #include <uvm/uvm.h>	/* extern struct uvm uvm */
214 
215 #include <dev/cons.h>
216 #include <dev/splash/splash.h>
217 
218 #include <net/bpf.h>
219 #include <net/if.h>
220 #include <net/pfil.h>
221 #include <net/raw_cb.h>
222 #include <net/if_llatbl.h>
223 
224 #include <prop/proplib.h>
225 
226 #include <sys/userconf.h>
227 
228 extern struct lwp lwp0;
229 extern time_t rootfstime;
230 
231 #ifndef curlwp
232 struct	lwp *curlwp = &lwp0;
233 #endif
234 struct	proc *initproc;
235 
236 struct	vnode *rootvp, *swapdev_vp;
237 int	boothowto;
238 int	cold __read_mostly = 1;		/* still working on startup */
239 struct timespec boottime;	        /* time at system startup - will only follow settime deltas */
240 
241 int	start_init_exec;		/* semaphore for start_init() */
242 
243 cprng_strong_t	*kern_cprng;
244 
245 static void check_console(struct lwp *l);
246 static void start_init(void *);
247 static void configure(void);
248 static void configure2(void);
249 static void configure3(void);
250 void main(void);
251 
252 /*
253  * System startup; initialize the world, create process 0, mount root
254  * filesystem, and fork to create init and pagedaemon.  Most of the
255  * hard work is done in the lower-level initialization routines including
256  * startup(), which does memory initialization and autoconfiguration.
257  */
258 void
259 main(void)
260 {
261 	struct timespec time;
262 	struct lwp *l;
263 	struct proc *p;
264 	int s, error;
265 #ifdef NVNODE_IMPLICIT
266 	int usevnodes;
267 #endif
268 	CPU_INFO_ITERATOR cii;
269 	struct cpu_info *ci;
270 
271 #ifdef DIAGNOSTIC
272 	/*
273 	 * Verify that CPU_INFO_FOREACH() knows about the boot CPU
274 	 * and only the boot CPU at this point.
275 	 */
276 	int cpucount = 0;
277 	for (CPU_INFO_FOREACH(cii, ci)) {
278 		KASSERT(ci == curcpu());
279 		cpucount++;
280 	}
281 	KASSERT(cpucount == 1);
282 #endif
283 
284 	l = &lwp0;
285 #ifndef LWP0_CPU_INFO
286 	l->l_cpu = curcpu();
287 #endif
288 	l->l_pflag |= LP_RUNNING;
289 
290 	/*
291 	 * Attempt to find console and initialize
292 	 * in case of early panic or other messages.
293 	 */
294 	consinit();
295 
296 	kernel_lock_init();
297 	once_init();
298 
299 	mi_cpu_init();
300 	kernconfig_lock_init();
301 	kthread_sysinit();
302 
303 	/* Initialize the device switch tables. */
304 	devsw_init();
305 
306 	/* Initialize event counters. */
307 	evcnt_init();
308 
309 	uvm_init();
310 	ubchist_init();
311 	kcpuset_sysinit();
312 
313 	prop_kern_init();
314 
315 #if ((NKSYMS > 0) || (NDDB > 0) || (NMODULAR > 0))
316 	ksyms_init();
317 #endif
318 	kprintf_init();
319 
320 	percpu_init();
321 
322 	/* Initialize lock caches. */
323 	mutex_obj_init();
324 	rw_obj_init();
325 
326 	/* Passive serialization. */
327 	pserialize_init();
328 
329 	/* Initialize the extent manager. */
330 	extent_init();
331 
332 	/* Do machine-dependent initialization. */
333 	cpu_startup();
334 
335 	/* Initialize the sysctl subsystem. */
336 	sysctl_init();
337 
338 	/* Initialize callouts, part 1. */
339 	callout_startup();
340 
341 	/* Initialize the kernel authorization subsystem. */
342 	kauth_init();
343 
344 	secmodel_init();
345 
346 	spec_init();
347 
348 	/*
349 	 * Set BPF op vector.  Can't do this in bpf attach, since
350 	 * network drivers attach before bpf.
351 	 */
352 	bpf_setops();
353 
354 	/* Start module system. */
355 	module_init();
356 
357 	/*
358 	 * Initialize the kernel authorization subsystem and start the
359 	 * default security model, if any. We need to do this early
360 	 * enough so that subsystems relying on any of the aforementioned
361 	 * can work properly. Since the security model may dictate the
362 	 * credential inheritance policy, it is needed at least before
363 	 * any process is created, specifically proc0.
364 	 */
365 	module_init_class(MODULE_CLASS_SECMODEL);
366 
367 	/* Initialize the buffer cache */
368 	bufinit();
369 	biohist_init();
370 
371 #ifdef KERNHIST
372 	sysctl_kernhist_init();
373 #endif
374 
375 
376 #if defined(SPLASHSCREEN) && defined(SPLASHSCREEN_IMAGE)
377 	size_t splash_size = (&_binary_splash_image_end -
378 	    &_binary_splash_image_start) * sizeof(void *);
379 	splash_setimage(&_binary_splash_image_start, splash_size);
380 #endif
381 
382 	/* Initialize sockets. */
383 	soinit();
384 
385 	/*
386 	 * The following things must be done before autoconfiguration.
387 	 */
388 	rnd_init();		/* initialize entropy pool */
389 
390 	cprng_init();		/* initialize cryptographic PRNG */
391 
392 	/* Initialize process and pgrp structures. */
393 	procinit();
394 	lwpinit();
395 
396 	/* Initialize signal-related data structures. */
397 	signal_init();
398 
399 	/* Initialize resource management. */
400 	resource_init();
401 
402 	/* Create process 0. */
403 	proc0_init();
404 	lwp0_init();
405 
406 	/* Disable preemption during boot. */
407 	kpreempt_disable();
408 
409 	/* Initialize the threadpool system. */
410 	threadpools_init();
411 
412 	/* Initialize the UID hash table. */
413 	uid_init();
414 
415 	/* Charge root for one process. */
416 	(void)chgproccnt(0, 1);
417 
418 	/* Initialize timekeeping. */
419 	time_init();
420 
421 	/* Initialize the run queues, turnstiles and sleep queues. */
422 	sched_rqinit();
423 	turnstile_init();
424 	sleeptab_init(&sleeptab);
425 
426 	sched_init();
427 
428 	/* Initialize processor-sets */
429 	psets_init();
430 
431 	/* Initialize cpufreq(9) */
432 	cpufreq_init();
433 
434 	/* MI initialization of the boot cpu */
435 	error = mi_cpu_attach(curcpu());
436 	KASSERT(error == 0);
437 
438 	/* Initialize timekeeping, part 2. */
439 	time_init2();
440 
441 	/*
442 	 * Initialize mbuf's.  Do this now because we might attempt to
443 	 * allocate mbufs or mbuf clusters during autoconfiguration.
444 	 */
445 	mbinit();
446 
447 	/* Initialize I/O statistics. */
448 	iostat_init();
449 
450 	/* Initialize the log device. */
451 	loginit();
452 
453 	/* Second part of module system initialization. */
454 	module_start_unload_thread();
455 
456 	/* Initialize the file systems. */
457 #ifdef NVNODE_IMPLICIT
458 	/*
459 	 * If maximum number of vnodes in namei vnode cache is not explicitly
460 	 * defined in kernel config, adjust the number such as we use roughly
461 	 * 10% of memory for vnodes and associated data structures in the
462 	 * assumed worst case.  Do not provide fewer than NVNODE vnodes.
463 	 */
464 	usevnodes = calc_cache_size(vmem_size(kmem_arena, VMEM_FREE|VMEM_ALLOC),
465 	    10, VNODE_KMEM_MAXPCT) / VNODE_COST;
466 	if (usevnodes > desiredvnodes)
467 		desiredvnodes = usevnodes;
468 #endif
469 
470 	/* Initialize fstrans. */
471 	fstrans_init();
472 
473 	vfsinit();
474 	lf_init();
475 
476 	/* Initialize the file descriptor system. */
477 	fd_sys_init();
478 
479 	/* Initialize cwd structures */
480 	cwd_sys_init();
481 
482 	/* Initialize kqueue. */
483 	kqueue_init();
484 
485 	inittimecounter();
486 	ntp_init();
487 
488 	/* Initialize tty subsystem. */
489 	tty_init();
490 	ttyldisc_init();
491 
492 	/* Initialize the buffer cache, part 2. */
493 	bufinit2();
494 
495 	/* Initialize the disk wedge subsystem. */
496 	dkwedge_init();
497 
498 	/* Initialize the kernel strong PRNG. */
499 	kern_cprng = cprng_strong_create("kernel", IPL_VM,
500 					 CPRNG_INIT_ANY|CPRNG_REKEY_ANY);
501 
502 	/* Initialize pfil */
503 	pfil_init();
504 
505 	/* Initialize interfaces. */
506 	ifinit1();
507 
508 	spldebug_start();
509 
510 	/* Initialize sockets thread(s) */
511 	soinit1();
512 
513 	/*
514 	 * Initialize the bufq strategy sub-system and any built-in
515 	 * strategy modules - they may be needed by some devices during
516 	 * auto-configuration
517 	 */
518 	bufq_init();
519 	module_init_class(MODULE_CLASS_BUFQ);
520 
521 	/* Configure the system hardware.  This will enable interrupts. */
522 	configure();
523 
524 	/* Once all CPUs are detected, initialize the per-CPU cprng_fast.  */
525 	cprng_fast_init();
526 
527 	ssp_init();
528 
529 	ubc_init();		/* must be after autoconfig */
530 
531 	mm_init();
532 
533 	configure2();
534 
535 	ipi_sysinit();
536 
537 	/* Now timer is working.  Enable preemption. */
538 	kpreempt_enable();
539 
540 	/* Get the threads going and into any sleeps before continuing. */
541 	yield();
542 
543 	/* Enable deferred processing of RNG samples */
544 	rnd_init_softint();
545 
546 #ifdef RND_PRINTF
547 	/* Enable periodic injection of console output into entropy pool */
548 	kprintf_init_callout();
549 #endif
550 
551 	vmem_rehash_start();	/* must be before exec_init */
552 
553 	/* Initialize exec structures */
554 	exec_init(1);		/* seminit calls exithook_establish() */
555 
556 #if NVERIEXEC > 0
557 	/*
558 	 * Initialise the Veriexec subsystem.
559 	 */
560 	veriexec_init();
561 #endif /* NVERIEXEC > 0 */
562 
563 	pax_init();
564 
565 #ifdef	IPSEC
566 	/* Attach network crypto subsystem */
567 	ipsec_attach();
568 #endif
569 
570 	/*
571 	 * Initialize protocols.  Block reception of incoming packets
572 	 * until everything is ready.
573 	 */
574 	s = splnet();
575 	ifinit();
576 #if defined(INET) || defined(INET6)
577 	lltableinit();
578 #endif
579 	domaininit(true);
580 	ifinit_post();
581 	if_attachdomain();
582 	splx(s);
583 
584 #ifdef GPROF
585 	/* Initialize kernel profiling. */
586 	kmstartup();
587 #endif
588 
589 	/* Initialize system accounting. */
590 	acct_init();
591 
592 #ifndef PIPE_SOCKETPAIR
593 	/* Initialize pipes. */
594 	pipe_init();
595 #endif
596 
597 #ifdef KTRACE
598 	/* Initialize ktrace. */
599 	ktrinit();
600 #endif
601 
602 	machdep_init();
603 
604 	procinit_sysctl();
605 
606 	scdebug_init();
607 
608 	/*
609 	 * Create process 1 (init(8)).  We do this now, as Unix has
610 	 * historically had init be process 1, and changing this would
611 	 * probably upset a lot of people.
612 	 *
613 	 * Note that process 1 won't immediately exec init(8), but will
614 	 * wait for us to inform it that the root file system has been
615 	 * mounted.
616 	 */
617 	if (fork1(l, 0, SIGCHLD, NULL, 0, start_init, NULL, NULL))
618 		panic("fork init");
619 
620 	/*
621 	 * The initproc variable cannot be initialized in start_init as there
622 	 * is a race between vfs_mountroot and start_init.
623 	 */
624 	mutex_enter(proc_lock);
625 	initproc = proc_find_raw(1);
626 	mutex_exit(proc_lock);
627 
628 	/*
629 	 * Load any remaining builtin modules, and hand back temporary
630 	 * storage to the VM system.  Then require force when loading any
631 	 * remaining un-init'ed built-in modules to avoid later surprises.
632 	 */
633 	module_init_class(MODULE_CLASS_ANY);
634 	module_builtin_require_force();
635 
636 	/*
637 	 * Finalize configuration now that all real devices have been
638 	 * found.  This needs to be done before the root device is
639 	 * selected, since finalization may create the root device.
640 	 */
641 	config_finalize();
642 
643 	sysctl_finalize();
644 
645 	/*
646 	 * Now that autoconfiguration has completed, we can determine
647 	 * the root and dump devices.
648 	 */
649 	cpu_rootconf();
650 	cpu_dumpconf();
651 
652 	/* Mount the root file system. */
653 	do {
654 		domountroothook(root_device);
655 		if ((error = vfs_mountroot())) {
656 			printf("cannot mount root, error = %d\n", error);
657 			boothowto |= RB_ASKNAME;
658 			setroot(root_device,
659 			    (rootdev != NODEV) ? DISKPART(rootdev) : 0);
660 		}
661 	} while (error != 0);
662 	mountroothook_destroy();
663 
664 	configure3();
665 
666 	/*
667 	 * Initialise the time-of-day clock, passing the time recorded
668 	 * in the root filesystem (if any) for use by systems that
669 	 * don't have a non-volatile time-of-day device.
670 	 */
671 	inittodr(rootfstime);
672 
673 	/*
674 	 * Now can look at time, having had a chance to verify the time
675 	 * from the file system.  Reset l->l_rtime as it may have been
676 	 * munched in mi_switch() after the time got set.
677 	 */
678 	getnanotime(&time);
679 	{
680 		struct timespec ut;
681 		/*
682 		 * was:
683 		 *	boottime = time;
684 		 * but we can do better
685 		 */
686 		nanouptime(&ut);
687 		timespecsub(&time, &ut, &boottime);
688 	}
689 
690 	mutex_enter(proc_lock);
691 	LIST_FOREACH(p, &allproc, p_list) {
692 		KASSERT((p->p_flag & PK_MARKER) == 0);
693 		mutex_enter(p->p_lock);
694 		TIMESPEC_TO_TIMEVAL(&p->p_stats->p_start, &time);
695 		LIST_FOREACH(l, &p->p_lwps, l_sibling) {
696 			lwp_lock(l);
697 			memset(&l->l_rtime, 0, sizeof(l->l_rtime));
698 			lwp_unlock(l);
699 		}
700 		mutex_exit(p->p_lock);
701 	}
702 	mutex_exit(proc_lock);
703 	binuptime(&curlwp->l_stime);
704 
705 	for (CPU_INFO_FOREACH(cii, ci)) {
706 		ci->ci_schedstate.spc_lastmod = time_second;
707 	}
708 
709 	/* Create the pageout daemon kernel thread. */
710 	uvm_swap_init();
711 	if (kthread_create(PRI_PGDAEMON, KTHREAD_MPSAFE, NULL, uvm_pageout,
712 	    NULL, NULL, "pgdaemon"))
713 		panic("fork pagedaemon");
714 
715 	/* Create the filesystem syncer kernel thread. */
716 	if (kthread_create(PRI_IOFLUSH, KTHREAD_MPSAFE, NULL, sched_sync,
717 	    NULL, NULL, "ioflush"))
718 		panic("fork syncer");
719 
720 	/* Create the aiodone daemon kernel thread. */
721 	if (workqueue_create(&uvm.aiodone_queue, "aiodoned",
722 	    uvm_aiodone_worker, NULL, PRI_VM, IPL_NONE, WQ_MPSAFE))
723 		panic("fork aiodoned");
724 
725 	/* Wait for final configure threads to complete. */
726 	config_finalize_mountroot();
727 
728 	/*
729 	 * Okay, now we can let init(8) exec!  It's off to userland!
730 	 */
731 	mutex_enter(proc_lock);
732 	start_init_exec = 1;
733 	cv_broadcast(&lbolt);
734 	mutex_exit(proc_lock);
735 
736 	/* The scheduler is an infinite loop. */
737 	uvm_scheduler();
738 	/* NOTREACHED */
739 }
740 
741 /*
742  * Configure the system's hardware.
743  */
744 static void
745 configure(void)
746 {
747 
748 	/* Initialize autoconf data structures. */
749 	config_init_mi();
750 	/*
751 	 * XXX
752 	 * callout_setfunc() requires mutex(9) so it can't be in config_init()
753 	 * on amiga and atari which use config_init() and autoconf(9) fucntions
754 	 * to initialize console devices.
755 	 */
756 	config_twiddle_init();
757 
758 	pmf_init();
759 
760 	/* Initialize driver modules */
761 	module_init_class(MODULE_CLASS_DRIVER);
762 
763 	userconf_init();
764 	if (boothowto & RB_USERCONF)
765 		userconf_prompt();
766 
767 	if ((boothowto & (AB_SILENT|AB_VERBOSE)) == AB_SILENT) {
768 		printf_nolog("Detecting hardware...");
769 	}
770 
771 	/*
772 	 * Do the machine-dependent portion of autoconfiguration.  This
773 	 * sets the configuration machinery here in motion by "finding"
774 	 * the root bus.  When this function returns, we expect interrupts
775 	 * to be enabled.
776 	 */
777 	cpu_configure();
778 }
779 
780 static void
781 configure2(void)
782 {
783 	CPU_INFO_ITERATOR cii;
784 	struct cpu_info *ci;
785 	int s;
786 
787 	/*
788 	 * Now that we've found all the hardware, start the real time
789 	 * and statistics clocks.
790 	 */
791 	initclocks();
792 
793 	cold = 0;	/* clocks are running, we're warm now! */
794 	s = splsched();
795 	curcpu()->ci_schedstate.spc_flags |= SPCF_RUNNING;
796 	splx(s);
797 
798 	/* Boot the secondary processors. */
799 	for (CPU_INFO_FOREACH(cii, ci)) {
800 		uvm_cpu_attach(ci);
801 	}
802 	mp_online = true;
803 #if defined(MULTIPROCESSOR)
804 	cpu_boot_secondary_processors();
805 #endif
806 
807 	/* Setup the runqueues and scheduler. */
808 	runq_init();
809 	synch_init();
810 
811 	/*
812 	 * Bus scans can make it appear as if the system has paused, so
813 	 * twiddle constantly while config_interrupts() jobs are running.
814 	 */
815 	config_twiddle_fn(NULL);
816 
817 	/*
818 	 * Create threads to call back and finish configuration for
819 	 * devices that want interrupts enabled.
820 	 */
821 	config_create_interruptthreads();
822 }
823 
824 static void
825 configure3(void)
826 {
827 
828 	/*
829 	 * Create threads to call back and finish configuration for
830 	 * devices that want the mounted root file system.
831 	 */
832 	config_create_mountrootthreads();
833 
834 	/* Get the threads going and into any sleeps before continuing. */
835 	yield();
836 }
837 
838 static void
839 rootconf_handle_wedges(void)
840 {
841 	struct disklabel label;
842 	struct partition *p;
843 	struct vnode *vp;
844 	daddr_t startblk;
845 	uint64_t nblks;
846 	device_t dev;
847 	int error;
848 
849 	if (booted_nblks) {
850 		/*
851 		 * bootloader passed geometry
852 		 */
853 		dev      = booted_device;
854 		startblk = booted_startblk;
855 		nblks    = booted_nblks;
856 
857 		/*
858 		 * keep booted_device and booted_partition
859 		 * in case the kernel doesn't identify a wedge
860 		 */
861 	} else {
862 		/*
863 		 * bootloader passed partition number
864 		 *
865 		 * We cannot ask the partition device directly when it is
866 		 * covered by a wedge. Instead we look up the geometry in
867 		 * the disklabel.
868 		 */
869 		vp = opendisk(booted_device);
870 
871 		if (vp == NULL)
872 			return;
873 
874 		error = VOP_IOCTL(vp, DIOCGDINFO, &label, FREAD, NOCRED);
875 		VOP_CLOSE(vp, FREAD, NOCRED);
876 		vput(vp);
877 		if (error)
878 			return;
879 
880 		KASSERT(booted_partition >= 0
881 			&& booted_partition < MAXPARTITIONS);
882 
883 		p = &label.d_partitions[booted_partition];
884 
885 		dev      = booted_device;
886 		startblk = p->p_offset;
887 		nblks    = p->p_size;
888 	}
889 
890 	dev = dkwedge_find_partition(dev, startblk, nblks);
891 	if (dev != NULL) {
892 		booted_device = dev;
893 		booted_partition = 0;
894 	}
895 }
896 
897 void
898 rootconf(void)
899 {
900 	if (booted_device != NULL)
901 		rootconf_handle_wedges();
902 
903 	setroot(booted_device, booted_partition);
904 }
905 
906 static void
907 check_console(struct lwp *l)
908 {
909 	struct vnode *vp;
910 	int error;
911 
912 	error = namei_simple_kernel("/dev/console",
913 				NSM_FOLLOW_NOEMULROOT, &vp);
914 	if (error == 0) {
915 		vrele(vp);
916 	} else if (error == ENOENT) {
917 		if (boothowto & (AB_VERBOSE|AB_DEBUG))
918 			printf("warning: no /dev/console\n");
919 	} else {
920 		printf("warning: lookup /dev/console: error %d\n", error);
921 	}
922 }
923 
924 /*
925  * List of paths to try when searching for "init".
926  */
927 static const char * const initpaths[] = {
928 	"/sbin/init",
929 	"/sbin/oinit",
930 	"/sbin/init.bak",
931 	"/rescue/init",
932 	NULL,
933 };
934 
935 /*
936  * Start the initial user process; try exec'ing each pathname in "initpaths".
937  * The program is invoked with one argument containing the boot flags.
938  */
939 static void
940 start_init(void *arg)
941 {
942 	struct lwp *l = arg;
943 	struct proc *p = l->l_proc;
944 	vaddr_t addr;
945 	struct sys_execve_args /* {
946 		syscallarg(const char *) path;
947 		syscallarg(char * const *) argp;
948 		syscallarg(char * const *) envp;
949 	} */ args;
950 	int options, i, error;
951 	register_t retval[2];
952 	char flags[4], *flagsp;
953 	const char *path, *slash;
954 	char *ucp, **uap, *arg0, *arg1, *argv[3];
955 	char ipath[129];
956 	int ipx, len;
957 
958 	/*
959 	 * Now in process 1.
960 	 */
961 	strncpy(p->p_comm, "init", MAXCOMLEN);
962 
963 	/*
964 	 * Wait for main() to tell us that it's safe to exec.
965 	 */
966 	mutex_enter(proc_lock);
967 	while (start_init_exec == 0)
968 		cv_wait(&lbolt, proc_lock);
969 	mutex_exit(proc_lock);
970 
971 	/*
972 	 * This is not the right way to do this.  We really should
973 	 * hand-craft a descriptor onto /dev/console to hand to init,
974 	 * but that's a _lot_ more work, and the benefit from this easy
975 	 * hack makes up for the "good is the enemy of the best" effect.
976 	 */
977 	check_console(l);
978 
979 	/*
980 	 * Need just enough stack to hold the faked-up "execve()" arguments.
981 	 */
982 	addr = (vaddr_t)STACK_ALLOC(USRSTACK, PAGE_SIZE);
983 	if (uvm_map(&p->p_vmspace->vm_map, &addr, PAGE_SIZE,
984 	    NULL, UVM_UNKNOWN_OFFSET, 0,
985 	    UVM_MAPFLAG(UVM_PROT_RW, UVM_PROT_RW, UVM_INH_COPY,
986 	    UVM_ADV_NORMAL,
987 	    UVM_FLAG_FIXED|UVM_FLAG_OVERLAY|UVM_FLAG_COPYONW)) != 0)
988 		panic("init: couldn't allocate argument space");
989 	p->p_vmspace->vm_maxsaddr = (void *)STACK_MAX(addr, PAGE_SIZE);
990 
991 	ipx = 0;
992 	while (1) {
993 		if (boothowto & RB_ASKNAME) {
994 			printf("init path");
995 			if (initpaths[ipx])
996 				printf(" (default %s)", initpaths[ipx]);
997 			printf(": ");
998 			len = cngetsn(ipath, sizeof(ipath)-1);
999 			if (len == 4 && strcmp(ipath, "halt") == 0) {
1000 				cpu_reboot(RB_HALT, NULL);
1001 			} else if (len == 6 && strcmp(ipath, "reboot") == 0) {
1002 				cpu_reboot(0, NULL);
1003 #if defined(DDB)
1004 			} else if (len == 3 && strcmp(ipath, "ddb") == 0) {
1005 				console_debugger();
1006 				continue;
1007 #endif
1008 			} else if (len > 0 && ipath[0] == '/') {
1009 				ipath[len] = '\0';
1010 				path = ipath;
1011 			} else if (len == 0 && initpaths[ipx] != NULL) {
1012 				path = initpaths[ipx++];
1013 			} else {
1014 				printf("use absolute path, ");
1015 #if defined(DDB)
1016 				printf("\"ddb\", ");
1017 #endif
1018 				printf("\"halt\", or \"reboot\"\n");
1019 				continue;
1020 			}
1021 		} else {
1022 			if ((path = initpaths[ipx++]) == NULL) {
1023 				ipx = 0;
1024 				boothowto |= RB_ASKNAME;
1025 				continue;
1026 			}
1027 		}
1028 
1029 		ucp = (char *)USRSTACK;
1030 
1031 		/*
1032 		 * Construct the boot flag argument.
1033 		 */
1034 		flagsp = flags;
1035 		*flagsp++ = '-';
1036 		options = 0;
1037 
1038 		if (boothowto & RB_SINGLE) {
1039 			*flagsp++ = 's';
1040 			options = 1;
1041 		}
1042 #ifdef notyet
1043 		if (boothowto & RB_FASTBOOT) {
1044 			*flagsp++ = 'f';
1045 			options = 1;
1046 		}
1047 #endif
1048 
1049 		/*
1050 		 * Move out the flags (arg 1), if necessary.
1051 		 */
1052 		if (options != 0) {
1053 			*flagsp++ = '\0';
1054 			i = flagsp - flags;
1055 #ifdef DEBUG
1056 			aprint_normal("init: copying out flags `%s' %d\n", flags, i);
1057 #endif
1058 			arg1 = STACK_ALLOC(ucp, i);
1059 			ucp = STACK_MAX(arg1, i);
1060 			if ((error = copyout((void *)flags, arg1, i)) != 0)
1061 				goto copyerr;
1062 		} else
1063 			arg1 = NULL;
1064 
1065 		/*
1066 		 * Move out the file name (also arg 0).
1067 		 */
1068 		i = strlen(path) + 1;
1069 #ifdef DEBUG
1070 		aprint_normal("init: copying out path `%s' %d\n", path, i);
1071 #else
1072 		if (boothowto & RB_ASKNAME || path != initpaths[0])
1073 			printf("init: trying %s\n", path);
1074 #endif
1075 		arg0 = STACK_ALLOC(ucp, i);
1076 		ucp = STACK_MAX(arg0, i);
1077 		if ((error = copyout(path, arg0, i)) != 0)
1078 			goto copyerr;
1079 
1080 		/*
1081 		 * Move out the arg pointers.
1082 		 */
1083 		ucp = (void *)STACK_ALIGN(ucp, STACK_ALIGNBYTES);
1084 		uap = (char **)STACK_ALLOC(ucp, sizeof(argv));
1085 		SCARG(&args, path) = arg0;
1086 		SCARG(&args, argp) = uap;
1087 		SCARG(&args, envp) = NULL;
1088 		slash = strrchr(path, '/');
1089 
1090 		argv[0] = slash ? arg0 + (slash + 1 - path) : arg0;
1091 		argv[1] = arg1;
1092 		argv[2] = NULL;
1093 		if ((error = copyout(argv, uap, sizeof(argv))) != 0)
1094 			goto copyerr;
1095 
1096 		/*
1097 		 * Now try to exec the program.  If it can't for any reason
1098 		 * other than it doesn't exist, complain.
1099 		 */
1100 		error = sys_execve(l, &args, retval);
1101 		if (error == 0 || error == EJUSTRETURN) {
1102 			KERNEL_UNLOCK_LAST(l);
1103 			return;
1104 		}
1105 		printf("exec %s: error %d\n", path, error);
1106 	}
1107 	printf("init: not found\n");
1108 	panic("no init");
1109 copyerr:
1110 	panic("copyout %d", error);
1111 }
1112 
1113 /*
1114  * calculate cache size (in bytes) from physmem and vsize.
1115  */
1116 vaddr_t
1117 calc_cache_size(vsize_t vsize, int pct, int va_pct)
1118 {
1119 	paddr_t t;
1120 
1121 	/* XXX should consider competing cache if any */
1122 	/* XXX should consider submaps */
1123 	t = (uintmax_t)physmem * pct / 100 * PAGE_SIZE;
1124 	if (vsize != 0) {
1125 		vsize = (uintmax_t)vsize * va_pct / 100;
1126 		if (t > vsize) {
1127 			t = vsize;
1128 		}
1129 	}
1130 	return t;
1131 }
1132 
1133 /*
1134  * Print the system start up banner.
1135  *
1136  * - Print a limited banner if AB_SILENT.
1137  * - Always send normal banner to the log.
1138  */
1139 #define MEM_PBUFSIZE	sizeof("99999 MB")
1140 
1141 void
1142 banner(void)
1143 {
1144 	static char notice[] = " Notice: this software is "
1145 	    "protected by copyright";
1146 	char pbuf[81];
1147 	void (*pr)(const char *, ...) __printflike(1, 2);
1148 	int i;
1149 
1150 	if ((boothowto & AB_SILENT) != 0) {
1151 		snprintf(pbuf, sizeof(pbuf), "%s %s (%s)",
1152 		    ostype, osrelease, kernel_ident);
1153 		printf_nolog("%s", pbuf);
1154 		for (i = 80 - strlen(pbuf) - sizeof(notice); i > 0; i--)
1155 			printf(" ");
1156 		printf_nolog("%s\n", notice);
1157 		pr = aprint_normal;
1158 	} else {
1159 		pr = printf;
1160 	}
1161 
1162 	memset(pbuf, 0, sizeof(pbuf));
1163 	(*pr)("%s%s", copyright, version);
1164 	format_bytes(pbuf, MEM_PBUFSIZE, ctob((uint64_t)physmem));
1165 	(*pr)("total memory = %s\n", pbuf);
1166 	format_bytes(pbuf, MEM_PBUFSIZE, ctob((uint64_t)uvmexp.free));
1167 	(*pr)("avail memory = %s\n", pbuf);
1168 }
1169