xref: /netbsd-src/sys/kern/init_main.c (revision 7330f729ccf0bd976a06f95fad452fe774fc7fd1)
1 /*	$NetBSD: init_main.c,v 1.506 2019/10/03 22:29:17 kamil 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 1.506 2019/10/03 22:29:17 kamil 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 #include <sys/psref.h>
199 
200 #include <sys/syscall.h>
201 #include <sys/syscallargs.h>
202 
203 #include <sys/pax.h>
204 
205 #include <secmodel/secmodel.h>
206 
207 #include <ufs/ufs/quota.h>
208 
209 #include <miscfs/genfs/genfs.h>
210 #include <miscfs/specfs/specdev.h>
211 
212 #include <sys/cpu.h>
213 
214 #include <uvm/uvm.h>	/* extern struct uvm uvm */
215 
216 #include <dev/cons.h>
217 #include <dev/splash/splash.h>
218 
219 #include <net/bpf.h>
220 #include <net/if.h>
221 #include <net/pfil.h>
222 #include <net/raw_cb.h>
223 #include <net/if_llatbl.h>
224 
225 #include <prop/proplib.h>
226 
227 #include <sys/userconf.h>
228 
229 extern struct lwp lwp0;
230 extern time_t rootfstime;
231 
232 #ifndef curlwp
233 struct	lwp *curlwp = &lwp0;
234 #endif
235 struct	proc *initproc;
236 
237 struct	vnode *rootvp, *swapdev_vp;
238 int	boothowto;
239 int	cold __read_mostly = 1;		/* still working on startup */
240 struct timespec boottime;	        /* time at system startup - will only follow settime deltas */
241 
242 int	start_init_exec;		/* semaphore for start_init() */
243 
244 cprng_strong_t	*kern_cprng;
245 
246 static void check_console(struct lwp *l);
247 static void start_init(void *);
248 static void configure(void);
249 static void configure2(void);
250 static void configure3(void);
251 void main(void);
252 
253 /*
254  * System startup; initialize the world, create process 0, mount root
255  * filesystem, and fork to create init and pagedaemon.  Most of the
256  * hard work is done in the lower-level initialization routines including
257  * startup(), which does memory initialization and autoconfiguration.
258  */
259 void
260 main(void)
261 {
262 	struct timespec time;
263 	struct lwp *l;
264 	struct proc *p;
265 	int s, error;
266 #ifdef NVNODE_IMPLICIT
267 	int usevnodes;
268 #endif
269 	CPU_INFO_ITERATOR cii;
270 	struct cpu_info *ci;
271 
272 #ifdef DIAGNOSTIC
273 	/*
274 	 * Verify that CPU_INFO_FOREACH() knows about the boot CPU
275 	 * and only the boot CPU at this point.
276 	 */
277 	int cpucount = 0;
278 	for (CPU_INFO_FOREACH(cii, ci)) {
279 		KASSERT(ci == curcpu());
280 		cpucount++;
281 	}
282 	KASSERT(cpucount == 1);
283 #endif
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 	biohist_init();
371 
372 #ifdef KERNHIST
373 	sysctl_kernhist_init();
374 #endif
375 
376 
377 #if defined(SPLASHSCREEN) && defined(SPLASHSCREEN_IMAGE)
378 	size_t splash_size = (&_binary_splash_image_end -
379 	    &_binary_splash_image_start) * sizeof(void *);
380 	splash_setimage(&_binary_splash_image_start, splash_size);
381 #endif
382 
383 	/* Initialize sockets. */
384 	soinit();
385 
386 	/*
387 	 * The following things must be done before autoconfiguration.
388 	 */
389 	rnd_init();		/* initialize entropy pool */
390 
391 	cprng_init();		/* initialize cryptographic PRNG */
392 
393 	/* Initialize process and pgrp structures. */
394 	procinit();
395 	lwpinit();
396 
397 	/* Must be called after lwpinit (lwpinit_specificdata) */
398 	psref_init();
399 
400 	/* Initialize signal-related data structures. */
401 	signal_init();
402 
403 	/* Initialize resource management. */
404 	resource_init();
405 
406 	/* Create process 0. */
407 	proc0_init();
408 	lwp0_init();
409 
410 	/* Disable preemption during boot. */
411 	kpreempt_disable();
412 
413 	/* Initialize the threadpool system. */
414 	threadpools_init();
415 
416 	/* Initialize the UID hash table. */
417 	uid_init();
418 
419 	/* Charge root for one process. */
420 	(void)chgproccnt(0, 1);
421 
422 	/* Initialize timekeeping. */
423 	time_init();
424 
425 	/* Initialize the run queues, turnstiles and sleep queues. */
426 	sched_rqinit();
427 	turnstile_init();
428 	sleeptab_init(&sleeptab);
429 
430 	sched_init();
431 
432 	/* Initialize processor-sets */
433 	psets_init();
434 
435 	/* Initialize cpufreq(9) */
436 	cpufreq_init();
437 
438 	/* MI initialization of the boot cpu */
439 	error = mi_cpu_attach(curcpu());
440 	KASSERT(error == 0);
441 
442 	/* Initialize timekeeping, part 2. */
443 	time_init2();
444 
445 	/*
446 	 * Initialize mbuf's.  Do this now because we might attempt to
447 	 * allocate mbufs or mbuf clusters during autoconfiguration.
448 	 */
449 	mbinit();
450 
451 	/* Initialize I/O statistics. */
452 	iostat_init();
453 
454 	/* Initialize the log device. */
455 	loginit();
456 
457 	/* Second part of module system initialization. */
458 	module_start_unload_thread();
459 
460 	/* Initialize the file systems. */
461 #ifdef NVNODE_IMPLICIT
462 	/*
463 	 * If maximum number of vnodes in namei vnode cache is not explicitly
464 	 * defined in kernel config, adjust the number such as we use roughly
465 	 * 10% of memory for vnodes and associated data structures in the
466 	 * assumed worst case.  Do not provide fewer than NVNODE vnodes.
467 	 */
468 	usevnodes = calc_cache_size(vmem_size(kmem_arena, VMEM_FREE|VMEM_ALLOC),
469 	    10, VNODE_KMEM_MAXPCT) / VNODE_COST;
470 	if (usevnodes > desiredvnodes)
471 		desiredvnodes = usevnodes;
472 #endif
473 
474 	/* Initialize fstrans. */
475 	fstrans_init();
476 
477 	vfsinit();
478 	lf_init();
479 
480 	/* Initialize the file descriptor system. */
481 	fd_sys_init();
482 
483 	/* Initialize cwd structures */
484 	cwd_sys_init();
485 
486 	/* Initialize kqueue. */
487 	kqueue_init();
488 
489 	inittimecounter();
490 	ntp_init();
491 
492 	/* Initialize tty subsystem. */
493 	tty_init();
494 	ttyldisc_init();
495 
496 	/* Initialize the buffer cache, part 2. */
497 	bufinit2();
498 
499 	/* Initialize the disk wedge subsystem. */
500 	dkwedge_init();
501 
502 	/* Initialize the kernel strong PRNG. */
503 	kern_cprng = cprng_strong_create("kernel", IPL_VM,
504 					 CPRNG_INIT_ANY|CPRNG_REKEY_ANY);
505 
506 	/* Initialize pfil */
507 	pfil_init();
508 
509 	/* Initialize interfaces. */
510 	ifinit1();
511 
512 	spldebug_start();
513 
514 	/* Initialize sockets thread(s) */
515 	soinit1();
516 
517 	/*
518 	 * Initialize the bufq strategy sub-system and any built-in
519 	 * strategy modules - they may be needed by some devices during
520 	 * auto-configuration
521 	 */
522 	bufq_init();
523 	module_init_class(MODULE_CLASS_BUFQ);
524 
525 	/* Configure the system hardware.  This will enable interrupts. */
526 	configure();
527 
528 	/* Once all CPUs are detected, initialize the per-CPU cprng_fast.  */
529 	cprng_fast_init();
530 
531 	ssp_init();
532 
533 	ubc_init();		/* must be after autoconfig */
534 
535 	mm_init();
536 
537 	configure2();
538 
539 	ipi_sysinit();
540 
541 	/* Now timer is working.  Enable preemption. */
542 	kpreempt_enable();
543 
544 	/* Get the threads going and into any sleeps before continuing. */
545 	yield();
546 
547 	/* Enable deferred processing of RNG samples */
548 	rnd_init_softint();
549 
550 #ifdef RND_PRINTF
551 	/* Enable periodic injection of console output into entropy pool */
552 	kprintf_init_callout();
553 #endif
554 
555 	vmem_rehash_start();	/* must be before exec_init */
556 
557 	/* Initialize exec structures */
558 	exec_init(1);		/* seminit calls exithook_establish() */
559 
560 #if NVERIEXEC > 0
561 	/*
562 	 * Initialise the Veriexec subsystem.
563 	 */
564 	veriexec_init();
565 #endif /* NVERIEXEC > 0 */
566 
567 	pax_init();
568 
569 #ifdef	IPSEC
570 	/* Attach network crypto subsystem */
571 	ipsec_attach();
572 #endif
573 
574 	/*
575 	 * Initialize protocols.  Block reception of incoming packets
576 	 * until everything is ready.
577 	 */
578 	s = splnet();
579 	ifinit();
580 #if defined(INET) || defined(INET6)
581 	lltableinit();
582 #endif
583 	domaininit(true);
584 	ifinit_post();
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 	machdep_init();
607 
608 	procinit_sysctl();
609 
610 	scdebug_init();
611 
612 	/*
613 	 * Create process 1 (init(8)).  We do this now, as Unix has
614 	 * historically had init be process 1, and changing this would
615 	 * probably upset a lot of people.
616 	 *
617 	 * Note that process 1 won't immediately exec init(8), but will
618 	 * wait for us to inform it that the root file system has been
619 	 * mounted.
620 	 */
621 	if (fork1(l, 0, SIGCHLD, NULL, 0, start_init, NULL, NULL))
622 		panic("fork init");
623 
624 	/*
625 	 * The initproc variable cannot be initialized in start_init as there
626 	 * is a race between vfs_mountroot and start_init.
627 	 */
628 	mutex_enter(proc_lock);
629 	initproc = proc_find_raw(1);
630 	mutex_exit(proc_lock);
631 
632 	/*
633 	 * Load any remaining builtin modules, and hand back temporary
634 	 * storage to the VM system.  Then require force when loading any
635 	 * remaining un-init'ed built-in modules to avoid later surprises.
636 	 */
637 	module_init_class(MODULE_CLASS_ANY);
638 	module_builtin_require_force();
639 
640 	/*
641 	 * Finalize configuration now that all real devices have been
642 	 * found.  This needs to be done before the root device is
643 	 * selected, since finalization may create the root device.
644 	 */
645 	config_finalize();
646 
647 	sysctl_finalize();
648 
649 	/*
650 	 * Now that autoconfiguration has completed, we can determine
651 	 * the root and dump devices.
652 	 */
653 	cpu_rootconf();
654 	cpu_dumpconf();
655 
656 	/* Mount the root file system. */
657 	do {
658 		domountroothook(root_device);
659 		if ((error = vfs_mountroot())) {
660 			printf("cannot mount root, error = %d\n", error);
661 			boothowto |= RB_ASKNAME;
662 			setroot(root_device,
663 			    (rootdev != NODEV) ? DISKPART(rootdev) : 0);
664 		}
665 	} while (error != 0);
666 	mountroothook_destroy();
667 
668 	configure3();
669 
670 	/*
671 	 * Initialise the time-of-day clock, passing the time recorded
672 	 * in the root filesystem (if any) for use by systems that
673 	 * don't have a non-volatile time-of-day device.
674 	 */
675 	inittodr(rootfstime);
676 
677 	/*
678 	 * Now can look at time, having had a chance to verify the time
679 	 * from the file system.  Reset l->l_rtime as it may have been
680 	 * munched in mi_switch() after the time got set.
681 	 */
682 	getnanotime(&time);
683 	{
684 		struct timespec ut;
685 		/*
686 		 * was:
687 		 *	boottime = time;
688 		 * but we can do better
689 		 */
690 		nanouptime(&ut);
691 		timespecsub(&time, &ut, &boottime);
692 	}
693 
694 	mutex_enter(proc_lock);
695 	LIST_FOREACH(p, &allproc, p_list) {
696 		KASSERT((p->p_flag & PK_MARKER) == 0);
697 		mutex_enter(p->p_lock);
698 		TIMESPEC_TO_TIMEVAL(&p->p_stats->p_start, &time);
699 		LIST_FOREACH(l, &p->p_lwps, l_sibling) {
700 			lwp_lock(l);
701 			memset(&l->l_rtime, 0, sizeof(l->l_rtime));
702 			lwp_unlock(l);
703 		}
704 		mutex_exit(p->p_lock);
705 	}
706 	mutex_exit(proc_lock);
707 	binuptime(&curlwp->l_stime);
708 
709 	for (CPU_INFO_FOREACH(cii, ci)) {
710 		ci->ci_schedstate.spc_lastmod = time_second;
711 	}
712 
713 	/* Create the pageout daemon kernel thread. */
714 	uvm_swap_init();
715 	if (kthread_create(PRI_PGDAEMON, KTHREAD_MPSAFE, NULL, uvm_pageout,
716 	    NULL, NULL, "pgdaemon"))
717 		panic("fork pagedaemon");
718 
719 	/* Create the filesystem syncer kernel thread. */
720 	if (kthread_create(PRI_IOFLUSH, KTHREAD_MPSAFE, NULL, sched_sync,
721 	    NULL, NULL, "ioflush"))
722 		panic("fork syncer");
723 
724 	/* Create the aiodone daemon kernel thread. */
725 	if (workqueue_create(&uvm.aiodone_queue, "aiodoned",
726 	    uvm_aiodone_worker, NULL, PRI_VM, IPL_NONE, WQ_MPSAFE))
727 		panic("fork aiodoned");
728 
729 	/* Wait for final configure threads to complete. */
730 	config_finalize_mountroot();
731 
732 	/*
733 	 * Okay, now we can let init(8) exec!  It's off to userland!
734 	 */
735 	mutex_enter(proc_lock);
736 	start_init_exec = 1;
737 	cv_broadcast(&lbolt);
738 	mutex_exit(proc_lock);
739 
740 	/* The scheduler is an infinite loop. */
741 	uvm_scheduler();
742 	/* NOTREACHED */
743 }
744 
745 /*
746  * Configure the system's hardware.
747  */
748 static void
749 configure(void)
750 {
751 
752 	/* Initialize autoconf data structures. */
753 	config_init_mi();
754 	/*
755 	 * XXX
756 	 * callout_setfunc() requires mutex(9) so it can't be in config_init()
757 	 * on amiga and atari which use config_init() and autoconf(9) fucntions
758 	 * to initialize console devices.
759 	 */
760 	config_twiddle_init();
761 
762 	pmf_init();
763 
764 	/* Initialize driver modules */
765 	module_init_class(MODULE_CLASS_DRIVER);
766 
767 	userconf_init();
768 	if (boothowto & RB_USERCONF)
769 		userconf_prompt();
770 
771 	if ((boothowto & (AB_SILENT|AB_VERBOSE)) == AB_SILENT) {
772 		printf_nolog("Detecting hardware...");
773 	}
774 
775 	/*
776 	 * Do the machine-dependent portion of autoconfiguration.  This
777 	 * sets the configuration machinery here in motion by "finding"
778 	 * the root bus.  When this function returns, we expect interrupts
779 	 * to be enabled.
780 	 */
781 	cpu_configure();
782 }
783 
784 static void
785 configure2(void)
786 {
787 	CPU_INFO_ITERATOR cii;
788 	struct cpu_info *ci;
789 	int s;
790 
791 	/*
792 	 * Now that we've found all the hardware, start the real time
793 	 * and statistics clocks.
794 	 */
795 	initclocks();
796 
797 	cold = 0;	/* clocks are running, we're warm now! */
798 	s = splsched();
799 	curcpu()->ci_schedstate.spc_flags |= SPCF_RUNNING;
800 	splx(s);
801 
802 	/* Boot the secondary processors. */
803 	for (CPU_INFO_FOREACH(cii, ci)) {
804 		uvm_cpu_attach(ci);
805 	}
806 	mp_online = true;
807 #if defined(MULTIPROCESSOR)
808 	cpu_boot_secondary_processors();
809 #endif
810 
811 	/* Setup the runqueues and scheduler. */
812 	runq_init();
813 	synch_init();
814 
815 	/*
816 	 * Bus scans can make it appear as if the system has paused, so
817 	 * twiddle constantly while config_interrupts() jobs are running.
818 	 */
819 	config_twiddle_fn(NULL);
820 
821 	/*
822 	 * Create threads to call back and finish configuration for
823 	 * devices that want interrupts enabled.
824 	 */
825 	config_create_interruptthreads();
826 }
827 
828 static void
829 configure3(void)
830 {
831 
832 	/*
833 	 * Create threads to call back and finish configuration for
834 	 * devices that want the mounted root file system.
835 	 */
836 	config_create_mountrootthreads();
837 
838 	/* Get the threads going and into any sleeps before continuing. */
839 	yield();
840 }
841 
842 static void
843 rootconf_handle_wedges(void)
844 {
845 	struct disklabel label;
846 	struct partition *p;
847 	struct vnode *vp;
848 	daddr_t startblk;
849 	uint64_t nblks;
850 	device_t dev;
851 	int error;
852 
853 	if (booted_nblks) {
854 		/*
855 		 * bootloader passed geometry
856 		 */
857 		dev      = booted_device;
858 		startblk = booted_startblk;
859 		nblks    = booted_nblks;
860 
861 		/*
862 		 * keep booted_device and booted_partition
863 		 * in case the kernel doesn't identify a wedge
864 		 */
865 	} else {
866 		/*
867 		 * bootloader passed partition number
868 		 *
869 		 * We cannot ask the partition device directly when it is
870 		 * covered by a wedge. Instead we look up the geometry in
871 		 * the disklabel.
872 		 */
873 		vp = opendisk(booted_device);
874 
875 		if (vp == NULL)
876 			return;
877 
878 		error = VOP_IOCTL(vp, DIOCGDINFO, &label, FREAD, NOCRED);
879 		VOP_CLOSE(vp, FREAD, NOCRED);
880 		vput(vp);
881 		if (error)
882 			return;
883 
884 		KASSERT(booted_partition >= 0
885 			&& booted_partition < MAXPARTITIONS);
886 
887 		p = &label.d_partitions[booted_partition];
888 
889 		dev      = booted_device;
890 		startblk = p->p_offset;
891 		nblks    = p->p_size;
892 	}
893 
894 	dev = dkwedge_find_partition(dev, startblk, nblks);
895 	if (dev != NULL) {
896 		booted_device = dev;
897 		booted_partition = 0;
898 	}
899 }
900 
901 void
902 rootconf(void)
903 {
904 	if (booted_device != NULL)
905 		rootconf_handle_wedges();
906 
907 	setroot(booted_device, booted_partition);
908 }
909 
910 static void
911 check_console(struct lwp *l)
912 {
913 	struct vnode *vp;
914 	int error;
915 
916 	error = namei_simple_kernel("/dev/console",
917 				NSM_FOLLOW_NOEMULROOT, &vp);
918 	if (error == 0) {
919 		vrele(vp);
920 	} else if (error == ENOENT) {
921 		if (boothowto & (AB_VERBOSE|AB_DEBUG))
922 			printf("warning: no /dev/console\n");
923 	} else {
924 		printf("warning: lookup /dev/console: error %d\n", error);
925 	}
926 }
927 
928 /*
929  * List of paths to try when searching for "init".
930  */
931 static const char * const initpaths[] = {
932 	"/sbin/init",
933 	"/sbin/oinit",
934 	"/sbin/init.bak",
935 	"/rescue/init",
936 	NULL,
937 };
938 
939 /*
940  * Start the initial user process; try exec'ing each pathname in "initpaths".
941  * The program is invoked with one argument containing the boot flags.
942  */
943 static void
944 start_init(void *arg)
945 {
946 	struct lwp *l = arg;
947 	struct proc *p = l->l_proc;
948 	vaddr_t addr;
949 	struct sys_execve_args /* {
950 		syscallarg(const char *) path;
951 		syscallarg(char * const *) argp;
952 		syscallarg(char * const *) envp;
953 	} */ args;
954 	int options, i, error;
955 	register_t retval[2];
956 	char flags[4], *flagsp;
957 	const char *path, *slash;
958 	char *ucp, **uap, *arg0, *arg1, *argv[3];
959 	char ipath[129];
960 	int ipx, len;
961 
962 	/*
963 	 * Now in process 1.
964 	 */
965 	strncpy(p->p_comm, "init", MAXCOMLEN);
966 
967 	/*
968 	 * Wait for main() to tell us that it's safe to exec.
969 	 */
970 	mutex_enter(proc_lock);
971 	while (start_init_exec == 0)
972 		cv_wait(&lbolt, proc_lock);
973 	mutex_exit(proc_lock);
974 
975 	/*
976 	 * This is not the right way to do this.  We really should
977 	 * hand-craft a descriptor onto /dev/console to hand to init,
978 	 * but that's a _lot_ more work, and the benefit from this easy
979 	 * hack makes up for the "good is the enemy of the best" effect.
980 	 */
981 	check_console(l);
982 
983 	/*
984 	 * Need just enough stack to hold the faked-up "execve()" arguments.
985 	 */
986 	addr = (vaddr_t)STACK_ALLOC(USRSTACK, PAGE_SIZE);
987 	if (uvm_map(&p->p_vmspace->vm_map, &addr, PAGE_SIZE,
988 	    NULL, UVM_UNKNOWN_OFFSET, 0,
989 	    UVM_MAPFLAG(UVM_PROT_RW, UVM_PROT_RW, UVM_INH_COPY,
990 	    UVM_ADV_NORMAL,
991 	    UVM_FLAG_FIXED|UVM_FLAG_OVERLAY|UVM_FLAG_COPYONW)) != 0)
992 		panic("init: couldn't allocate argument space");
993 	p->p_vmspace->vm_maxsaddr = (void *)STACK_MAX(addr, PAGE_SIZE);
994 
995 	ipx = 0;
996 	while (1) {
997 		if (boothowto & RB_ASKNAME) {
998 			printf("init path");
999 			if (initpaths[ipx])
1000 				printf(" (default %s)", initpaths[ipx]);
1001 			printf(": ");
1002 			len = cngetsn(ipath, sizeof(ipath)-1);
1003 			if (len == 4 && strcmp(ipath, "halt") == 0) {
1004 				cpu_reboot(RB_HALT, NULL);
1005 			} else if (len == 6 && strcmp(ipath, "reboot") == 0) {
1006 				cpu_reboot(0, NULL);
1007 #if defined(DDB)
1008 			} else if (len == 3 && strcmp(ipath, "ddb") == 0) {
1009 				console_debugger();
1010 				continue;
1011 #endif
1012 			} else if (len > 0 && ipath[0] == '/') {
1013 				ipath[len] = '\0';
1014 				path = ipath;
1015 			} else if (len == 0 && initpaths[ipx] != NULL) {
1016 				path = initpaths[ipx++];
1017 			} else {
1018 				printf("use absolute path, ");
1019 #if defined(DDB)
1020 				printf("\"ddb\", ");
1021 #endif
1022 				printf("\"halt\", or \"reboot\"\n");
1023 				continue;
1024 			}
1025 		} else {
1026 			if ((path = initpaths[ipx++]) == NULL) {
1027 				ipx = 0;
1028 				boothowto |= RB_ASKNAME;
1029 				continue;
1030 			}
1031 		}
1032 
1033 		ucp = (char *)USRSTACK;
1034 
1035 		/*
1036 		 * Construct the boot flag argument.
1037 		 */
1038 		flagsp = flags;
1039 		*flagsp++ = '-';
1040 		options = 0;
1041 
1042 		if (boothowto & RB_SINGLE) {
1043 			*flagsp++ = 's';
1044 			options = 1;
1045 		}
1046 #ifdef notyet
1047 		if (boothowto & RB_FASTBOOT) {
1048 			*flagsp++ = 'f';
1049 			options = 1;
1050 		}
1051 #endif
1052 
1053 		/*
1054 		 * Move out the flags (arg 1), if necessary.
1055 		 */
1056 		if (options != 0) {
1057 			*flagsp++ = '\0';
1058 			i = flagsp - flags;
1059 #ifdef DEBUG
1060 			aprint_normal("init: copying out flags `%s' %d\n", flags, i);
1061 #endif
1062 			arg1 = STACK_ALLOC(ucp, i);
1063 			ucp = STACK_MAX(arg1, i);
1064 			if ((error = copyout((void *)flags, arg1, i)) != 0)
1065 				goto copyerr;
1066 		} else
1067 			arg1 = NULL;
1068 
1069 		/*
1070 		 * Move out the file name (also arg 0).
1071 		 */
1072 		i = strlen(path) + 1;
1073 #ifdef DEBUG
1074 		aprint_normal("init: copying out path `%s' %d\n", path, i);
1075 #else
1076 		if (boothowto & RB_ASKNAME || path != initpaths[0])
1077 			printf("init: trying %s\n", path);
1078 #endif
1079 		arg0 = STACK_ALLOC(ucp, i);
1080 		ucp = STACK_MAX(arg0, i);
1081 		if ((error = copyout(path, arg0, i)) != 0)
1082 			goto copyerr;
1083 
1084 		/*
1085 		 * Move out the arg pointers.
1086 		 */
1087 		ucp = (void *)STACK_ALIGN(ucp, STACK_ALIGNBYTES);
1088 		uap = (char **)STACK_ALLOC(ucp, sizeof(argv));
1089 		SCARG(&args, path) = arg0;
1090 		SCARG(&args, argp) = uap;
1091 		SCARG(&args, envp) = NULL;
1092 		slash = strrchr(path, '/');
1093 
1094 		argv[0] = slash ? arg0 + (slash + 1 - path) : arg0;
1095 		argv[1] = arg1;
1096 		argv[2] = NULL;
1097 		if ((error = copyout(argv, uap, sizeof(argv))) != 0)
1098 			goto copyerr;
1099 
1100 		/*
1101 		 * Now try to exec the program.  If it can't for any reason
1102 		 * other than it doesn't exist, complain.
1103 		 */
1104 		error = sys_execve(l, &args, retval);
1105 		if (error == 0 || error == EJUSTRETURN) {
1106 			KERNEL_UNLOCK_LAST(l);
1107 			return;
1108 		}
1109 		printf("exec %s: error %d\n", path, error);
1110 	}
1111 	printf("init: not found\n");
1112 	panic("no init");
1113 copyerr:
1114 	panic("copyout %d", error);
1115 }
1116 
1117 /*
1118  * calculate cache size (in bytes) from physmem and vsize.
1119  */
1120 vaddr_t
1121 calc_cache_size(vsize_t vsize, int pct, int va_pct)
1122 {
1123 	paddr_t t;
1124 
1125 	/* XXX should consider competing cache if any */
1126 	/* XXX should consider submaps */
1127 	t = (uintmax_t)physmem * pct / 100 * PAGE_SIZE;
1128 	if (vsize != 0) {
1129 		vsize = (uintmax_t)vsize * va_pct / 100;
1130 		if (t > vsize) {
1131 			t = vsize;
1132 		}
1133 	}
1134 	return t;
1135 }
1136 
1137 /*
1138  * Print the system start up banner.
1139  *
1140  * - Print a limited banner if AB_SILENT.
1141  * - Always send normal banner to the log.
1142  */
1143 #define MEM_PBUFSIZE	sizeof("99999 MB")
1144 
1145 void
1146 banner(void)
1147 {
1148 	static char notice[] = " Notice: this software is "
1149 	    "protected by copyright";
1150 	char pbuf[81];
1151 	void (*pr)(const char *, ...) __printflike(1, 2);
1152 	int i;
1153 
1154 	if ((boothowto & AB_SILENT) != 0) {
1155 		snprintf(pbuf, sizeof(pbuf), "%s %s (%s)",
1156 		    ostype, osrelease, kernel_ident);
1157 		printf_nolog("%s", pbuf);
1158 		for (i = 80 - strlen(pbuf) - sizeof(notice); i > 0; i--)
1159 			printf(" ");
1160 		printf_nolog("%s\n", notice);
1161 		pr = aprint_normal;
1162 	} else {
1163 		pr = printf;
1164 	}
1165 
1166 	memset(pbuf, 0, sizeof(pbuf));
1167 	(*pr)("%s%s", copyright, version);
1168 	format_bytes(pbuf, MEM_PBUFSIZE, ctob((uint64_t)physmem));
1169 	(*pr)("total memory = %s\n", pbuf);
1170 	format_bytes(pbuf, MEM_PBUFSIZE, ctob((uint64_t)uvmexp.free));
1171 	(*pr)("avail memory = %s\n", pbuf);
1172 }
1173