xref: /netbsd-src/sys/kern/init_main.c (revision e89934bbf778a6d6d6894877c4da59d0c7835b0f)
1 /*	$NetBSD: init_main.c,v 1.490 2017/01/16 09:28:40 ryo 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 1.490 2017/01/16 09:28:40 ryo 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 
120 #if defined(SPLASHSCREEN) && defined(makeoptions_SPLASHSCREEN_IMAGE)
121 extern void *_binary_splash_image_start;
122 extern void *_binary_splash_image_end;
123 #endif
124 
125 #include "ksyms.h"
126 
127 #include "veriexec.h"
128 
129 #include <sys/param.h>
130 #include <sys/acct.h>
131 #include <sys/filedesc.h>
132 #include <sys/file.h>
133 #include <sys/errno.h>
134 #include <sys/callout.h>
135 #include <sys/cpu.h>
136 #include <sys/cpufreq.h>
137 #include <sys/spldebug.h>
138 #include <sys/kernel.h>
139 #include <sys/mount.h>
140 #include <sys/proc.h>
141 #include <sys/kthread.h>
142 #include <sys/resourcevar.h>
143 #include <sys/signalvar.h>
144 #include <sys/systm.h>
145 #include <sys/vnode.h>
146 #include <sys/fstrans.h>
147 #include <sys/tty.h>
148 #include <sys/conf.h>
149 #include <sys/disklabel.h>
150 #include <sys/buf.h>
151 #include <sys/device.h>
152 #include <sys/exec.h>
153 #include <sys/socketvar.h>
154 #include <sys/protosw.h>
155 #include <sys/percpu.h>
156 #include <sys/pserialize.h>
157 #include <sys/pset.h>
158 #include <sys/sysctl.h>
159 #include <sys/reboot.h>
160 #include <sys/event.h>
161 #include <sys/mbuf.h>
162 #include <sys/sched.h>
163 #include <sys/sleepq.h>
164 #include <sys/ipi.h>
165 #include <sys/iostat.h>
166 #include <sys/vmem.h>
167 #include <sys/uuid.h>
168 #include <sys/extent.h>
169 #include <sys/disk.h>
170 #include <sys/msgbuf.h>
171 #include <sys/module.h>
172 #include <sys/event.h>
173 #include <sys/lockf.h>
174 #include <sys/once.h>
175 #include <sys/kcpuset.h>
176 #include <sys/ksyms.h>
177 #include <sys/uidinfo.h>
178 #include <sys/kprintf.h>
179 #include <sys/bufq.h>
180 #ifdef IPSEC
181 #include <netipsec/ipsec.h>
182 #endif
183 #include <sys/domain.h>
184 #include <sys/namei.h>
185 #include <sys/rnd.h>
186 #include <sys/pipe.h>
187 #if NVERIEXEC > 0
188 #include <sys/verified_exec.h>
189 #endif /* NVERIEXEC > 0 */
190 #ifdef KTRACE
191 #include <sys/ktrace.h>
192 #endif
193 #include <sys/kauth.h>
194 #include <net80211/ieee80211_netbsd.h>
195 #include <sys/cprng.h>
196 
197 #include <sys/syscall.h>
198 #include <sys/syscallargs.h>
199 
200 #include <sys/pax.h>
201 
202 #include <secmodel/secmodel.h>
203 
204 #include <ufs/ufs/quota.h>
205 
206 #include <miscfs/genfs/genfs.h>
207 #include <miscfs/specfs/specdev.h>
208 
209 #include <sys/cpu.h>
210 
211 #include <uvm/uvm.h>	/* extern struct uvm uvm */
212 
213 #include <dev/cons.h>
214 #include <dev/splash/splash.h>
215 
216 #include <net/bpf.h>
217 #include <net/if.h>
218 #include <net/pfil.h>
219 #include <net/raw_cb.h>
220 #include <net/if_llatbl.h>
221 
222 #include <prop/proplib.h>
223 
224 #include <sys/userconf.h>
225 
226 extern struct lwp lwp0;
227 extern time_t rootfstime;
228 
229 #ifndef curlwp
230 struct	lwp *curlwp = &lwp0;
231 #endif
232 struct	proc *initproc;
233 
234 struct	vnode *rootvp, *swapdev_vp;
235 int	boothowto;
236 int	cold = 1;			/* still working on startup */
237 struct timespec boottime;	        /* time at system startup - will only follow settime deltas */
238 
239 int	start_init_exec;		/* semaphore for start_init() */
240 
241 cprng_strong_t	*kern_cprng;
242 
243 static void check_console(struct lwp *l);
244 static void start_init(void *);
245 static void configure(void);
246 static void configure2(void);
247 static void configure3(void);
248 void main(void);
249 
250 /*
251  * System startup; initialize the world, create process 0, mount root
252  * filesystem, and fork to create init and pagedaemon.  Most of the
253  * hard work is done in the lower-level initialization routines including
254  * startup(), which does memory initialization and autoconfiguration.
255  */
256 void
257 main(void)
258 {
259 	struct timespec time;
260 	struct lwp *l;
261 	struct proc *p;
262 	int s, error;
263 #ifdef NVNODE_IMPLICIT
264 	int usevnodes;
265 #endif
266 	CPU_INFO_ITERATOR cii;
267 	struct cpu_info *ci;
268 
269 	l = &lwp0;
270 #ifndef LWP0_CPU_INFO
271 	l->l_cpu = curcpu();
272 #endif
273 	l->l_pflag |= LP_RUNNING;
274 
275 	/*
276 	 * Attempt to find console and initialize
277 	 * in case of early panic or other messages.
278 	 */
279 	consinit();
280 
281 	kernel_lock_init();
282 	once_init();
283 
284 	mi_cpu_init();
285 	kernconfig_lock_init();
286 	kthread_sysinit();
287 
288 	/* Initialize the device switch tables. */
289 	devsw_init();
290 
291 	/* Initialize event counters. */
292 	evcnt_init();
293 
294 	uvm_init();
295 	ubchist_init();
296 	kcpuset_sysinit();
297 
298 	prop_kern_init();
299 
300 #if ((NKSYMS > 0) || (NDDB > 0) || (NMODULAR > 0))
301 	ksyms_init();
302 #endif
303 	kprintf_init();
304 
305 	percpu_init();
306 
307 	/* Initialize lock caches. */
308 	mutex_obj_init();
309 	rw_obj_init();
310 
311 	/* Passive serialization. */
312 	pserialize_init();
313 
314 	/* Initialize the extent manager. */
315 	extent_init();
316 
317 	/* Do machine-dependent initialization. */
318 	cpu_startup();
319 
320 	/* Initialize the sysctl subsystem. */
321 	sysctl_init();
322 
323 	/* Initialize callouts, part 1. */
324 	callout_startup();
325 
326 	/* Initialize the kernel authorization subsystem. */
327 	kauth_init();
328 
329 	secmodel_init();
330 
331 	spec_init();
332 
333 	/*
334 	 * Set BPF op vector.  Can't do this in bpf attach, since
335 	 * network drivers attach before bpf.
336 	 */
337 	bpf_setops();
338 
339 	/* Start module system. */
340 	module_init();
341 
342 	/*
343 	 * Initialize the kernel authorization subsystem and start the
344 	 * default security model, if any. We need to do this early
345 	 * enough so that subsystems relying on any of the aforementioned
346 	 * can work properly. Since the security model may dictate the
347 	 * credential inheritance policy, it is needed at least before
348 	 * any process is created, specifically proc0.
349 	 */
350 	module_init_class(MODULE_CLASS_SECMODEL);
351 
352 	/* Initialize the buffer cache */
353 	bufinit();
354 	biohist_init();
355 
356 #ifdef KERNHIST
357 	sysctl_kernhist_init();
358 #endif
359 
360 
361 #if defined(SPLASHSCREEN) && defined(SPLASHSCREEN_IMAGE)
362 	size_t splash_size = (&_binary_splash_image_end -
363 	    &_binary_splash_image_start) * sizeof(void *);
364 	splash_setimage(&_binary_splash_image_start, splash_size);
365 #endif
366 
367 	/* Initialize sockets. */
368 	soinit();
369 
370 	/*
371 	 * The following things must be done before autoconfiguration.
372 	 */
373 	rnd_init();		/* initialize entropy pool */
374 
375 	cprng_init();		/* initialize cryptographic PRNG */
376 
377 	/* Initialize process and pgrp structures. */
378 	procinit();
379 	lwpinit();
380 
381 	/* Initialize signal-related data structures. */
382 	signal_init();
383 
384 	/* Initialize resource management. */
385 	resource_init();
386 
387 	/* Create process 0. */
388 	proc0_init();
389 	lwp0_init();
390 
391 	/* Disable preemption during boot. */
392 	kpreempt_disable();
393 
394 	/* Initialize the UID hash table. */
395 	uid_init();
396 
397 	/* Charge root for one process. */
398 	(void)chgproccnt(0, 1);
399 
400 	/* Initialize timekeeping. */
401 	time_init();
402 
403 	/* Initialize the run queues, turnstiles and sleep queues. */
404 	sched_rqinit();
405 	turnstile_init();
406 	sleeptab_init(&sleeptab);
407 
408 	sched_init();
409 
410 	/* Initialize processor-sets */
411 	psets_init();
412 
413 	/* Initialize cpufreq(9) */
414 	cpufreq_init();
415 
416 	/* MI initialization of the boot cpu */
417 	error = mi_cpu_attach(curcpu());
418 	KASSERT(error == 0);
419 
420 	/* Initialize timekeeping, part 2. */
421 	time_init2();
422 
423 	/*
424 	 * Initialize mbuf's.  Do this now because we might attempt to
425 	 * allocate mbufs or mbuf clusters during autoconfiguration.
426 	 */
427 	mbinit();
428 
429 	/* Initialize I/O statistics. */
430 	iostat_init();
431 
432 	/* Initialize the log device. */
433 	loginit();
434 
435 	/* Second part of module system initialization. */
436 	module_start_unload_thread();
437 
438 	/* Initialize the file systems. */
439 #ifdef NVNODE_IMPLICIT
440 	/*
441 	 * If maximum number of vnodes in namei vnode cache is not explicitly
442 	 * defined in kernel config, adjust the number such as we use roughly
443 	 * 10% of memory for vnodes and associated data structures in the
444 	 * assumed worst case.  Do not provide fewer than NVNODE vnodes.
445 	 */
446 	usevnodes = calc_cache_size(vmem_size(kmem_arena, VMEM_FREE|VMEM_ALLOC),
447 	    10, VNODE_KMEM_MAXPCT) / VNODE_COST;
448 	if (usevnodes > desiredvnodes)
449 		desiredvnodes = usevnodes;
450 #endif
451 	vfsinit();
452 	lf_init();
453 
454 	/* Initialize fstrans. */
455 	fstrans_init();
456 
457 	/* Initialize the file descriptor system. */
458 	fd_sys_init();
459 
460 	/* Initialize cwd structures */
461 	cwd_sys_init();
462 
463 	/* Initialize kqueue. */
464 	kqueue_init();
465 
466 	inittimecounter();
467 	ntp_init();
468 
469 	/* Initialize tty subsystem. */
470 	tty_init();
471 	ttyldisc_init();
472 
473 	/* Initialize the buffer cache, part 2. */
474 	bufinit2();
475 
476 	/* Initialize the disk wedge subsystem. */
477 	dkwedge_init();
478 
479 	/* Initialize the kernel strong PRNG. */
480 	kern_cprng = cprng_strong_create("kernel", IPL_VM,
481 					 CPRNG_INIT_ANY|CPRNG_REKEY_ANY);
482 
483 	/* Initialize pfil */
484 	pfil_init();
485 
486 	/* Initialize interfaces. */
487 	ifinit1();
488 
489 	spldebug_start();
490 
491 	/* Initialize sockets thread(s) */
492 	soinit1();
493 
494 	/*
495 	 * Initialize the bufq strategy sub-system and any built-in
496 	 * strategy modules - they may be needed by some devices during
497 	 * auto-configuration
498 	 */
499 	bufq_init();
500 	module_init_class(MODULE_CLASS_BUFQ);
501 
502 	/* Configure the system hardware.  This will enable interrupts. */
503 	configure();
504 
505 	/* Once all CPUs are detected, initialize the per-CPU cprng_fast.  */
506 	cprng_fast_init();
507 
508 	ssp_init();
509 
510 	ubc_init();		/* must be after autoconfig */
511 
512 	mm_init();
513 
514 	configure2();
515 
516 	ipi_sysinit();
517 
518 	/* Now timer is working.  Enable preemption. */
519 	kpreempt_enable();
520 
521 	/* Get the threads going and into any sleeps before continuing. */
522 	yield();
523 
524 	/* Enable deferred processing of RNG samples */
525 	rnd_init_softint();
526 
527 #ifdef RND_PRINTF
528 	/* Enable periodic injection of console output into entropy pool */
529 	kprintf_init_callout();
530 #endif
531 
532 	vmem_rehash_start();	/* must be before exec_init */
533 
534 	/* Initialize exec structures */
535 	exec_init(1);		/* seminit calls exithook_establish() */
536 
537 #if NVERIEXEC > 0
538 	/*
539 	 * Initialise the Veriexec subsystem.
540 	 */
541 	veriexec_init();
542 #endif /* NVERIEXEC > 0 */
543 
544 	pax_init();
545 
546 #ifdef	IPSEC
547 	/* Attach network crypto subsystem */
548 	ipsec_attach();
549 #endif
550 
551 	/*
552 	 * Initialize protocols.  Block reception of incoming packets
553 	 * until everything is ready.
554 	 */
555 	s = splnet();
556 	ifinit();
557 #if defined(INET) || defined(INET6)
558 	lltableinit();
559 #endif
560 	domaininit(true);
561 	if_attachdomain();
562 	splx(s);
563 
564 #ifdef GPROF
565 	/* Initialize kernel profiling. */
566 	kmstartup();
567 #endif
568 
569 	/* Initialize system accounting. */
570 	acct_init();
571 
572 #ifndef PIPE_SOCKETPAIR
573 	/* Initialize pipes. */
574 	pipe_init();
575 #endif
576 
577 #ifdef KTRACE
578 	/* Initialize ktrace. */
579 	ktrinit();
580 #endif
581 
582 	machdep_init();
583 
584 	procinit_sysctl();
585 
586 	scdebug_init();
587 
588 	/*
589 	 * Create process 1 (init(8)).  We do this now, as Unix has
590 	 * historically had init be process 1, and changing this would
591 	 * probably upset a lot of people.
592 	 *
593 	 * Note that process 1 won't immediately exec init(8), but will
594 	 * wait for us to inform it that the root file system has been
595 	 * mounted.
596 	 */
597 	if (fork1(l, 0, SIGCHLD, NULL, 0, start_init, NULL, NULL, &initproc))
598 		panic("fork init");
599 
600 	/*
601 	 * Load any remaining builtin modules, and hand back temporary
602 	 * storage to the VM system.  Then require force when loading any
603 	 * remaining un-init'ed built-in modules to avoid later surprises.
604 	 */
605 	module_init_class(MODULE_CLASS_ANY);
606 	module_builtin_require_force();
607 
608 	/*
609 	 * Finalize configuration now that all real devices have been
610 	 * found.  This needs to be done before the root device is
611 	 * selected, since finalization may create the root device.
612 	 */
613 	config_finalize();
614 
615 	sysctl_finalize();
616 
617 	/*
618 	 * Now that autoconfiguration has completed, we can determine
619 	 * the root and dump devices.
620 	 */
621 	cpu_rootconf();
622 	cpu_dumpconf();
623 
624 	/* Mount the root file system. */
625 	do {
626 		domountroothook(root_device);
627 		if ((error = vfs_mountroot())) {
628 			printf("cannot mount root, error = %d\n", error);
629 			boothowto |= RB_ASKNAME;
630 			setroot(root_device,
631 			    (rootdev != NODEV) ? DISKPART(rootdev) : 0);
632 		}
633 	} while (error != 0);
634 	mountroothook_destroy();
635 
636 	configure3();
637 
638 	/*
639 	 * Initialise the time-of-day clock, passing the time recorded
640 	 * in the root filesystem (if any) for use by systems that
641 	 * don't have a non-volatile time-of-day device.
642 	 */
643 	inittodr(rootfstime);
644 
645 	/*
646 	 * Now can look at time, having had a chance to verify the time
647 	 * from the file system.  Reset l->l_rtime as it may have been
648 	 * munched in mi_switch() after the time got set.
649 	 */
650 	getnanotime(&time);
651 	boottime = time;
652 
653 	mutex_enter(proc_lock);
654 	LIST_FOREACH(p, &allproc, p_list) {
655 		KASSERT((p->p_flag & PK_MARKER) == 0);
656 		mutex_enter(p->p_lock);
657 		TIMESPEC_TO_TIMEVAL(&p->p_stats->p_start, &time);
658 		LIST_FOREACH(l, &p->p_lwps, l_sibling) {
659 			lwp_lock(l);
660 			memset(&l->l_rtime, 0, sizeof(l->l_rtime));
661 			lwp_unlock(l);
662 		}
663 		mutex_exit(p->p_lock);
664 	}
665 	mutex_exit(proc_lock);
666 	binuptime(&curlwp->l_stime);
667 
668 	for (CPU_INFO_FOREACH(cii, ci)) {
669 		ci->ci_schedstate.spc_lastmod = time_second;
670 	}
671 
672 	/* Create the pageout daemon kernel thread. */
673 	uvm_swap_init();
674 	if (kthread_create(PRI_PGDAEMON, KTHREAD_MPSAFE, NULL, uvm_pageout,
675 	    NULL, NULL, "pgdaemon"))
676 		panic("fork pagedaemon");
677 
678 	/* Create the filesystem syncer kernel thread. */
679 	if (kthread_create(PRI_IOFLUSH, KTHREAD_MPSAFE, NULL, sched_sync,
680 	    NULL, NULL, "ioflush"))
681 		panic("fork syncer");
682 
683 	/* Create the aiodone daemon kernel thread. */
684 	if (workqueue_create(&uvm.aiodone_queue, "aiodoned",
685 	    uvm_aiodone_worker, NULL, PRI_VM, IPL_NONE, WQ_MPSAFE))
686 		panic("fork aiodoned");
687 
688 	/* Wait for final configure threads to complete. */
689 	config_finalize_mountroot();
690 
691 	/*
692 	 * Okay, now we can let init(8) exec!  It's off to userland!
693 	 */
694 	mutex_enter(proc_lock);
695 	start_init_exec = 1;
696 	cv_broadcast(&lbolt);
697 	mutex_exit(proc_lock);
698 
699 	/* The scheduler is an infinite loop. */
700 	uvm_scheduler();
701 	/* NOTREACHED */
702 }
703 
704 /*
705  * Configure the system's hardware.
706  */
707 static void
708 configure(void)
709 {
710 
711 	/* Initialize autoconf data structures. */
712 	config_init_mi();
713 	/*
714 	 * XXX
715 	 * callout_setfunc() requires mutex(9) so it can't be in config_init()
716 	 * on amiga and atari which use config_init() and autoconf(9) fucntions
717 	 * to initialize console devices.
718 	 */
719 	config_twiddle_init();
720 
721 	pmf_init();
722 
723 	/* Initialize driver modules */
724 	module_init_class(MODULE_CLASS_DRIVER);
725 
726 	userconf_init();
727 	if (boothowto & RB_USERCONF)
728 		userconf_prompt();
729 
730 	if ((boothowto & (AB_SILENT|AB_VERBOSE)) == AB_SILENT) {
731 		printf_nolog("Detecting hardware...");
732 	}
733 
734 	/*
735 	 * Do the machine-dependent portion of autoconfiguration.  This
736 	 * sets the configuration machinery here in motion by "finding"
737 	 * the root bus.  When this function returns, we expect interrupts
738 	 * to be enabled.
739 	 */
740 	cpu_configure();
741 }
742 
743 static void
744 configure2(void)
745 {
746 	CPU_INFO_ITERATOR cii;
747 	struct cpu_info *ci;
748 	int s;
749 
750 	/*
751 	 * Now that we've found all the hardware, start the real time
752 	 * and statistics clocks.
753 	 */
754 	initclocks();
755 
756 	cold = 0;	/* clocks are running, we're warm now! */
757 	s = splsched();
758 	curcpu()->ci_schedstate.spc_flags |= SPCF_RUNNING;
759 	splx(s);
760 
761 	/* Boot the secondary processors. */
762 	for (CPU_INFO_FOREACH(cii, ci)) {
763 		uvm_cpu_attach(ci);
764 	}
765 	mp_online = true;
766 #if defined(MULTIPROCESSOR)
767 	cpu_boot_secondary_processors();
768 #endif
769 
770 	/* Setup the runqueues and scheduler. */
771 	runq_init();
772 	synch_init();
773 
774 	/*
775 	 * Bus scans can make it appear as if the system has paused, so
776 	 * twiddle constantly while config_interrupts() jobs are running.
777 	 */
778 	config_twiddle_fn(NULL);
779 
780 	/*
781 	 * Create threads to call back and finish configuration for
782 	 * devices that want interrupts enabled.
783 	 */
784 	config_create_interruptthreads();
785 }
786 
787 static void
788 configure3(void)
789 {
790 
791 	/*
792 	 * Create threads to call back and finish configuration for
793 	 * devices that want the mounted root file system.
794 	 */
795 	config_create_mountrootthreads();
796 
797 	/* Get the threads going and into any sleeps before continuing. */
798 	yield();
799 }
800 
801 static void
802 rootconf_handle_wedges(void)
803 {
804 	struct disklabel label;
805 	struct partition *p;
806 	struct vnode *vp;
807 	daddr_t startblk;
808 	uint64_t nblks;
809 	device_t dev;
810 	int error;
811 
812 	if (booted_nblks) {
813 		/*
814 		 * bootloader passed geometry
815 		 */
816 		dev      = booted_device;
817 		startblk = booted_startblk;
818 		nblks    = booted_nblks;
819 
820 		/*
821 		 * keep booted_device and booted_partition
822 		 * in case the kernel doesn't identify a wedge
823 		 */
824 	} else {
825 		/*
826 		 * bootloader passed partition number
827 		 *
828 		 * We cannot ask the partition device directly when it is
829 		 * covered by a wedge. Instead we look up the geometry in
830 		 * the disklabel.
831 		 */
832 		vp = opendisk(booted_device);
833 
834 		if (vp == NULL)
835 			return;
836 
837 		error = VOP_IOCTL(vp, DIOCGDINFO, &label, FREAD, NOCRED);
838 		VOP_CLOSE(vp, FREAD, NOCRED);
839 		vput(vp);
840 		if (error)
841 			return;
842 
843 		KASSERT(booted_partition >= 0
844 			&& booted_partition < MAXPARTITIONS);
845 
846 		p = &label.d_partitions[booted_partition];
847 
848 		dev      = booted_device;
849 		startblk = p->p_offset;
850 		nblks    = p->p_size;
851 	}
852 
853 	dev = dkwedge_find_partition(dev, startblk, nblks);
854 	if (dev != NULL) {
855 		booted_device = dev;
856 		booted_partition = 0;
857 	}
858 }
859 
860 void
861 rootconf(void)
862 {
863 	if (booted_device != NULL)
864 		rootconf_handle_wedges();
865 
866 	setroot(booted_device, booted_partition);
867 }
868 
869 static void
870 check_console(struct lwp *l)
871 {
872 	struct vnode *vp;
873 	int error;
874 
875 	error = namei_simple_kernel("/dev/console",
876 				NSM_FOLLOW_NOEMULROOT, &vp);
877 	if (error == 0)
878 		vrele(vp);
879 	else if (error == ENOENT)
880 		printf("warning: no /dev/console\n");
881 	else
882 		printf("warning: lookup /dev/console: error %d\n", error);
883 }
884 
885 /*
886  * List of paths to try when searching for "init".
887  */
888 static const char * const initpaths[] = {
889 	"/sbin/init",
890 	"/sbin/oinit",
891 	"/sbin/init.bak",
892 	"/rescue/init",
893 	NULL,
894 };
895 
896 /*
897  * Start the initial user process; try exec'ing each pathname in "initpaths".
898  * The program is invoked with one argument containing the boot flags.
899  */
900 static void
901 start_init(void *arg)
902 {
903 	struct lwp *l = arg;
904 	struct proc *p = l->l_proc;
905 	vaddr_t addr;
906 	struct sys_execve_args /* {
907 		syscallarg(const char *) path;
908 		syscallarg(char * const *) argp;
909 		syscallarg(char * const *) envp;
910 	} */ args;
911 	int options, i, error;
912 	register_t retval[2];
913 	char flags[4], *flagsp;
914 	const char *path, *slash;
915 	char *ucp, **uap, *arg0, *arg1, *argv[3];
916 	char ipath[129];
917 	int ipx, len;
918 
919 	/*
920 	 * Now in process 1.
921 	 */
922 	strncpy(p->p_comm, "init", MAXCOMLEN);
923 
924 	/*
925 	 * Wait for main() to tell us that it's safe to exec.
926 	 */
927 	mutex_enter(proc_lock);
928 	while (start_init_exec == 0)
929 		cv_wait(&lbolt, proc_lock);
930 	mutex_exit(proc_lock);
931 
932 	/*
933 	 * This is not the right way to do this.  We really should
934 	 * hand-craft a descriptor onto /dev/console to hand to init,
935 	 * but that's a _lot_ more work, and the benefit from this easy
936 	 * hack makes up for the "good is the enemy of the best" effect.
937 	 */
938 	check_console(l);
939 
940 	/*
941 	 * Need just enough stack to hold the faked-up "execve()" arguments.
942 	 */
943 	addr = (vaddr_t)STACK_ALLOC(USRSTACK, PAGE_SIZE);
944 	if (uvm_map(&p->p_vmspace->vm_map, &addr, PAGE_SIZE,
945 	    NULL, UVM_UNKNOWN_OFFSET, 0,
946 	    UVM_MAPFLAG(UVM_PROT_RW, UVM_PROT_RW, UVM_INH_COPY,
947 	    UVM_ADV_NORMAL,
948 	    UVM_FLAG_FIXED|UVM_FLAG_OVERLAY|UVM_FLAG_COPYONW)) != 0)
949 		panic("init: couldn't allocate argument space");
950 	p->p_vmspace->vm_maxsaddr = (void *)STACK_MAX(addr, PAGE_SIZE);
951 
952 	ipx = 0;
953 	while (1) {
954 		if (boothowto & RB_ASKNAME) {
955 			printf("init path");
956 			if (initpaths[ipx])
957 				printf(" (default %s)", initpaths[ipx]);
958 			printf(": ");
959 			len = cngetsn(ipath, sizeof(ipath)-1);
960 			if (len == 4 && strcmp(ipath, "halt") == 0) {
961 				cpu_reboot(RB_HALT, NULL);
962 			} else if (len == 6 && strcmp(ipath, "reboot") == 0) {
963 				cpu_reboot(0, NULL);
964 #if defined(DDB)
965 			} else if (len == 3 && strcmp(ipath, "ddb") == 0) {
966 				console_debugger();
967 				continue;
968 #endif
969 			} else if (len > 0 && ipath[0] == '/') {
970 				ipath[len] = '\0';
971 				path = ipath;
972 			} else if (len == 0 && initpaths[ipx] != NULL) {
973 				path = initpaths[ipx++];
974 			} else {
975 				printf("use absolute path, ");
976 #if defined(DDB)
977 				printf("\"ddb\", ");
978 #endif
979 				printf("\"halt\", or \"reboot\"\n");
980 				continue;
981 			}
982 		} else {
983 			if ((path = initpaths[ipx++]) == NULL) {
984 				ipx = 0;
985 				boothowto |= RB_ASKNAME;
986 				continue;
987 			}
988 		}
989 
990 		ucp = (char *)USRSTACK;
991 
992 		/*
993 		 * Construct the boot flag argument.
994 		 */
995 		flagsp = flags;
996 		*flagsp++ = '-';
997 		options = 0;
998 
999 		if (boothowto & RB_SINGLE) {
1000 			*flagsp++ = 's';
1001 			options = 1;
1002 		}
1003 #ifdef notyet
1004 		if (boothowto & RB_FASTBOOT) {
1005 			*flagsp++ = 'f';
1006 			options = 1;
1007 		}
1008 #endif
1009 
1010 		/*
1011 		 * Move out the flags (arg 1), if necessary.
1012 		 */
1013 		if (options != 0) {
1014 			*flagsp++ = '\0';
1015 			i = flagsp - flags;
1016 #ifdef DEBUG
1017 			aprint_normal("init: copying out flags `%s' %d\n", flags, i);
1018 #endif
1019 			arg1 = STACK_ALLOC(ucp, i);
1020 			ucp = STACK_MAX(arg1, i);
1021 			if ((error = copyout((void *)flags, arg1, i)) != 0)
1022 				goto copyerr;
1023 		} else
1024 			arg1 = NULL;
1025 
1026 		/*
1027 		 * Move out the file name (also arg 0).
1028 		 */
1029 		i = strlen(path) + 1;
1030 #ifdef DEBUG
1031 		aprint_normal("init: copying out path `%s' %d\n", path, i);
1032 #else
1033 		if (boothowto & RB_ASKNAME || path != initpaths[0])
1034 			printf("init: trying %s\n", path);
1035 #endif
1036 		arg0 = STACK_ALLOC(ucp, i);
1037 		ucp = STACK_MAX(arg0, i);
1038 		if ((error = copyout(path, arg0, i)) != 0)
1039 			goto copyerr;
1040 
1041 		/*
1042 		 * Move out the arg pointers.
1043 		 */
1044 		ucp = (void *)STACK_ALIGN(ucp, STACK_ALIGNBYTES);
1045 		uap = (char **)STACK_ALLOC(ucp, sizeof(argv));
1046 		SCARG(&args, path) = arg0;
1047 		SCARG(&args, argp) = uap;
1048 		SCARG(&args, envp) = NULL;
1049 		slash = strrchr(path, '/');
1050 
1051 		argv[0] = slash ? arg0 + (slash + 1 - path) : arg0;
1052 		argv[1] = arg1;
1053 		argv[2] = NULL;
1054 		if ((error = copyout(argv, uap, sizeof(argv))) != 0)
1055 			goto copyerr;
1056 
1057 		/*
1058 		 * Now try to exec the program.  If it can't for any reason
1059 		 * other than it doesn't exist, complain.
1060 		 */
1061 		error = sys_execve(l, &args, retval);
1062 		if (error == 0 || error == EJUSTRETURN) {
1063 			KERNEL_UNLOCK_LAST(l);
1064 			return;
1065 		}
1066 		printf("exec %s: error %d\n", path, error);
1067 	}
1068 	printf("init: not found\n");
1069 	panic("no init");
1070 copyerr:
1071 	panic("copyout %d", error);
1072 }
1073 
1074 /*
1075  * calculate cache size (in bytes) from physmem and vsize.
1076  */
1077 vaddr_t
1078 calc_cache_size(vsize_t vsize, int pct, int va_pct)
1079 {
1080 	paddr_t t;
1081 
1082 	/* XXX should consider competing cache if any */
1083 	/* XXX should consider submaps */
1084 	t = (uintmax_t)physmem * pct / 100 * PAGE_SIZE;
1085 	if (vsize != 0) {
1086 		vsize = (uintmax_t)vsize * va_pct / 100;
1087 		if (t > vsize) {
1088 			t = vsize;
1089 		}
1090 	}
1091 	return t;
1092 }
1093 
1094 /*
1095  * Print the system start up banner.
1096  *
1097  * - Print a limited banner if AB_SILENT.
1098  * - Always send normal banner to the log.
1099  */
1100 #define MEM_PBUFSIZE	sizeof("99999 MB")
1101 
1102 void
1103 banner(void)
1104 {
1105 	static char notice[] = " Notice: this software is "
1106 	    "protected by copyright";
1107 	char pbuf[81];
1108 	void (*pr)(const char *, ...) __printflike(1, 2);
1109 	int i;
1110 
1111 	if ((boothowto & AB_SILENT) != 0) {
1112 		snprintf(pbuf, sizeof(pbuf), "%s %s (%s)",
1113 		    ostype, osrelease, kernel_ident);
1114 		printf_nolog("%s", pbuf);
1115 		for (i = 80 - strlen(pbuf) - sizeof(notice); i > 0; i--)
1116 			printf(" ");
1117 		printf_nolog("%s\n", notice);
1118 		pr = aprint_normal;
1119 	} else {
1120 		pr = printf;
1121 	}
1122 
1123 	memset(pbuf, 0, sizeof(pbuf));
1124 	(*pr)("%s%s", copyright, version);
1125 	format_bytes(pbuf, MEM_PBUFSIZE, ctob((uint64_t)physmem));
1126 	(*pr)("total memory = %s\n", pbuf);
1127 	format_bytes(pbuf, MEM_PBUFSIZE, ctob((uint64_t)uvmexp.free));
1128 	(*pr)("avail memory = %s\n", pbuf);
1129 }
1130