xref: /netbsd-src/sys/kern/init_main.c (revision 946379e7b37692fc43f68eb0d1c10daa0a7f3b6c)
1 /*	$NetBSD: init_main.c,v 1.477 2015/12/08 20:36:15 christos 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 1.477 2015/12/08 20:36:15 christos 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 
119 #if defined(SPLASHSCREEN) && defined(makeoptions_SPLASHSCREEN_IMAGE)
120 extern void *_binary_splash_image_start;
121 extern void *_binary_splash_image_end;
122 #endif
123 
124 #include "ksyms.h"
125 
126 #include "veriexec.h"
127 
128 #include <sys/param.h>
129 #include <sys/acct.h>
130 #include <sys/filedesc.h>
131 #include <sys/file.h>
132 #include <sys/errno.h>
133 #include <sys/callout.h>
134 #include <sys/cpu.h>
135 #include <sys/cpufreq.h>
136 #include <sys/spldebug.h>
137 #include <sys/kernel.h>
138 #include <sys/mount.h>
139 #include <sys/proc.h>
140 #include <sys/kthread.h>
141 #include <sys/resourcevar.h>
142 #include <sys/signalvar.h>
143 #include <sys/systm.h>
144 #include <sys/vnode.h>
145 #include <sys/fstrans.h>
146 #include <sys/tty.h>
147 #include <sys/conf.h>
148 #include <sys/disklabel.h>
149 #include <sys/buf.h>
150 #include <sys/device.h>
151 #include <sys/exec.h>
152 #include <sys/socketvar.h>
153 #include <sys/protosw.h>
154 #include <sys/percpu.h>
155 #include <sys/pserialize.h>
156 #include <sys/pset.h>
157 #include <sys/sysctl.h>
158 #include <sys/reboot.h>
159 #include <sys/event.h>
160 #include <sys/mbuf.h>
161 #include <sys/sched.h>
162 #include <sys/sleepq.h>
163 #include <sys/ipi.h>
164 #include <sys/iostat.h>
165 #include <sys/vmem.h>
166 #include <sys/uuid.h>
167 #include <sys/extent.h>
168 #include <sys/disk.h>
169 #include <sys/msgbuf.h>
170 #include <sys/module.h>
171 #include <sys/event.h>
172 #include <sys/lockf.h>
173 #include <sys/once.h>
174 #include <sys/kcpuset.h>
175 #include <sys/ksyms.h>
176 #include <sys/uidinfo.h>
177 #include <sys/kprintf.h>
178 #ifdef IPSEC
179 #include <netipsec/ipsec.h>
180 #endif
181 #include <sys/domain.h>
182 #include <sys/namei.h>
183 #include <sys/rnd.h>
184 #include <sys/pipe.h>
185 #if NVERIEXEC > 0
186 #include <sys/verified_exec.h>
187 #endif /* NVERIEXEC > 0 */
188 #ifdef KTRACE
189 #include <sys/ktrace.h>
190 #endif
191 #include <sys/kauth.h>
192 #include <net80211/ieee80211_netbsd.h>
193 #ifdef PTRACE
194 #include <sys/ptrace.h>
195 #endif /* PTRACE */
196 #include <sys/cprng.h>
197 
198 #include <sys/syscall.h>
199 #include <sys/syscallargs.h>
200 
201 #if defined(PAX_MPROTECT) || defined(PAX_SEGVGUARD) || defined(PAX_ASLR)
202 #include <sys/pax.h>
203 #endif /* PAX_MPROTECT || PAX_SEGVGUARD || PAX_ASLR */
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/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 = 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 	l = &lwp0;
272 #ifndef LWP0_CPU_INFO
273 	l->l_cpu = curcpu();
274 #endif
275 	l->l_pflag |= LP_RUNNING;
276 
277 	/*
278 	 * Attempt to find console and initialize
279 	 * in case of early panic or other messages.
280 	 */
281 	consinit();
282 
283 	kernel_lock_init();
284 	once_init();
285 
286 	mi_cpu_init();
287 	kernconfig_lock_init();
288 	kthread_sysinit();
289 
290 	/* Initialize the device switch tables. */
291 	devsw_init();
292 
293 	/* Initialize event counters. */
294 	evcnt_init();
295 
296 	uvm_init();
297 	ubchist_init();
298 	kcpuset_sysinit();
299 
300 	prop_kern_init();
301 
302 #if ((NKSYMS > 0) || (NDDB > 0) || (NMODULAR > 0))
303 	ksyms_init();
304 #endif
305 	kprintf_init();
306 
307 	percpu_init();
308 
309 	/* Initialize lock caches. */
310 	mutex_obj_init();
311 	rw_obj_init();
312 
313 	/* Passive serialization. */
314 	pserialize_init();
315 
316 	/* Initialize the extent manager. */
317 	extent_init();
318 
319 	/* Do machine-dependent initialization. */
320 	cpu_startup();
321 
322 	/* Initialize the sysctl subsystem. */
323 	sysctl_init();
324 
325 	/* Initialize callouts, part 1. */
326 	callout_startup();
327 
328 	/* Initialize the kernel authorization subsystem. */
329 	kauth_init();
330 
331 	secmodel_init();
332 
333 	spec_init();
334 
335 	/*
336 	 * Set BPF op vector.  Can't do this in bpf attach, since
337 	 * network drivers attach before bpf.
338 	 */
339 	bpf_setops();
340 
341 	/* Start module system. */
342 	module_init();
343 
344 	/*
345 	 * Initialize the kernel authorization subsystem and start the
346 	 * default security model, if any. We need to do this early
347 	 * enough so that subsystems relying on any of the aforementioned
348 	 * can work properly. Since the security model may dictate the
349 	 * credential inheritance policy, it is needed at least before
350 	 * any process is created, specifically proc0.
351 	 */
352 	module_init_class(MODULE_CLASS_SECMODEL);
353 
354 	/* Initialize the buffer cache */
355 	bufinit();
356 
357 
358 #if defined(SPLASHSCREEN) && defined(SPLASHSCREEN_IMAGE)
359 	size_t splash_size = (&_binary_splash_image_end -
360 	    &_binary_splash_image_start) * sizeof(void *);
361 	splash_setimage(&_binary_splash_image_start, splash_size);
362 #endif
363 
364 	/* Initialize sockets. */
365 	soinit();
366 
367 	/*
368 	 * The following things must be done before autoconfiguration.
369 	 */
370 	rnd_init();		/* initialize entropy pool */
371 
372 	cprng_init();		/* initialize cryptographic PRNG */
373 
374 	/* Initialize process and pgrp structures. */
375 	procinit();
376 	lwpinit();
377 
378 	/* Initialize signal-related data structures. */
379 	signal_init();
380 
381 	/* Initialize resource management. */
382 	resource_init();
383 
384 	/* Create process 0. */
385 	proc0_init();
386 	lwp0_init();
387 
388 	/* Disable preemption during boot. */
389 	kpreempt_disable();
390 
391 	/* Initialize the UID hash table. */
392 	uid_init();
393 
394 	/* Charge root for one process. */
395 	(void)chgproccnt(0, 1);
396 
397 	/* Initialize timekeeping. */
398 	time_init();
399 
400 	/* Initialize the run queues, turnstiles and sleep queues. */
401 	sched_rqinit();
402 	turnstile_init();
403 	sleeptab_init(&sleeptab);
404 
405 	sched_init();
406 
407 	/* Initialize processor-sets */
408 	psets_init();
409 
410 	/* Initialize cpufreq(9) */
411 	cpufreq_init();
412 
413 	/* MI initialization of the boot cpu */
414 	error = mi_cpu_attach(curcpu());
415 	KASSERT(error == 0);
416 
417 	/* Initialize timekeeping, part 2. */
418 	time_init2();
419 
420 	/*
421 	 * Initialize mbuf's.  Do this now because we might attempt to
422 	 * allocate mbufs or mbuf clusters during autoconfiguration.
423 	 */
424 	mbinit();
425 
426 	/* Initialize I/O statistics. */
427 	iostat_init();
428 
429 	/* Initialize the log device. */
430 	loginit();
431 
432 	/* Second part of module system initialization. */
433 	module_start_unload_thread();
434 
435 	/* Initialize the file systems. */
436 #ifdef NVNODE_IMPLICIT
437 	/*
438 	 * If maximum number of vnodes in namei vnode cache is not explicitly
439 	 * defined in kernel config, adjust the number such as we use roughly
440 	 * 10% of memory for vnodes and associated data structures in the
441 	 * assumed worst case.  Do not provide fewer than NVNODE vnodes.
442 	 */
443 	usevnodes = calc_cache_size(vmem_size(kmem_arena, VMEM_FREE|VMEM_ALLOC),
444 	    10, VNODE_KMEM_MAXPCT) / VNODE_COST;
445 	if (usevnodes > desiredvnodes)
446 		desiredvnodes = usevnodes;
447 #endif
448 	vfsinit();
449 	lf_init();
450 
451 	/* Initialize fstrans. */
452 	fstrans_init();
453 
454 	/* Initialize the file descriptor system. */
455 	fd_sys_init();
456 
457 	/* Initialize cwd structures */
458 	cwd_sys_init();
459 
460 	/* Initialize kqueue. */
461 	kqueue_init();
462 
463 	inittimecounter();
464 	ntp_init();
465 
466 	/* Initialize tty subsystem. */
467 	tty_init();
468 	ttyldisc_init();
469 
470 	/* Initialize the buffer cache, part 2. */
471 	bufinit2();
472 
473 	/* Initialize the disk wedge subsystem. */
474 	dkwedge_init();
475 
476 	/* Initialize the kernel strong PRNG. */
477 	kern_cprng = cprng_strong_create("kernel", IPL_VM,
478 					 CPRNG_INIT_ANY|CPRNG_REKEY_ANY);
479 
480 	/* Initialize interfaces. */
481 	ifinit1();
482 
483 	spldebug_start();
484 
485 	/* Initialize sockets thread(s) */
486 	soinit1();
487 
488 	/* Configure the system hardware.  This will enable interrupts. */
489 	configure();
490 
491 	/* Once all CPUs are detected, initialize the per-CPU cprng_fast.  */
492 	cprng_fast_init();
493 
494 	ssp_init();
495 
496 	ubc_init();		/* must be after autoconfig */
497 
498 	mm_init();
499 
500 	configure2();
501 
502 	ipi_sysinit();
503 
504 	/* Now timer is working.  Enable preemption. */
505 	kpreempt_enable();
506 
507 	/* Get the threads going and into any sleeps before continuing. */
508 	yield();
509 
510 	/* Enable deferred processing of RNG samples */
511 	rnd_init_softint();
512 
513 #ifdef RND_PRINTF
514 	/* Enable periodic injection of console output into entropy pool */
515 	kprintf_init_callout();
516 #endif
517 
518 	vmem_rehash_start();	/* must be before exec_init */
519 
520 	/* Initialize exec structures */
521 	exec_init(1);		/* seminit calls exithook_establish() */
522 
523 #if NVERIEXEC > 0
524 	/*
525 	 * Initialise the Veriexec subsystem.
526 	 */
527 	veriexec_init();
528 #endif /* NVERIEXEC > 0 */
529 
530 #if defined(PAX_MPROTECT) || defined(PAX_SEGVGUARD) || defined(PAX_ASLR)
531 	pax_init();
532 #endif /* PAX_MPROTECT || PAX_SEGVGUARD || PAX_ASLR */
533 
534 #ifdef	IPSEC
535 	/* Attach network crypto subsystem */
536 	ipsec_attach();
537 #endif
538 
539 	/*
540 	 * Initialize protocols.  Block reception of incoming packets
541 	 * until everything is ready.
542 	 */
543 	s = splnet();
544 	ifinit();
545 #if defined(INET) || defined(INET6)
546 	lltableinit();
547 #endif
548 	domaininit(true);
549 	if_attachdomain();
550 	splx(s);
551 
552 #ifdef GPROF
553 	/* Initialize kernel profiling. */
554 	kmstartup();
555 #endif
556 
557 	/* Initialize system accounting. */
558 	acct_init();
559 
560 #ifndef PIPE_SOCKETPAIR
561 	/* Initialize pipes. */
562 	pipe_init();
563 #endif
564 
565 #ifdef KTRACE
566 	/* Initialize ktrace. */
567 	ktrinit();
568 #endif
569 
570 #ifdef PTRACE
571 	/* Initialize ptrace. */
572 	ptrace_init();
573 #endif /* PTRACE */
574 
575 	machdep_init();
576 
577 	procinit_sysctl();
578 
579 	scdebug_init();
580 
581 	/*
582 	 * Create process 1 (init(8)).  We do this now, as Unix has
583 	 * historically had init be process 1, and changing this would
584 	 * probably upset a lot of people.
585 	 *
586 	 * Note that process 1 won't immediately exec init(8), but will
587 	 * wait for us to inform it that the root file system has been
588 	 * mounted.
589 	 */
590 	if (fork1(l, 0, SIGCHLD, NULL, 0, start_init, NULL, NULL, &initproc))
591 		panic("fork init");
592 
593 	/*
594 	 * Load any remaining builtin modules, and hand back temporary
595 	 * storage to the VM system.  Then require force when loading any
596 	 * remaining un-init'ed built-in modules to avoid later surprises.
597 	 */
598 	module_init_class(MODULE_CLASS_ANY);
599 	module_builtin_require_force();
600 
601 	/*
602 	 * Finalize configuration now that all real devices have been
603 	 * found.  This needs to be done before the root device is
604 	 * selected, since finalization may create the root device.
605 	 */
606 	config_finalize();
607 
608 	sysctl_finalize();
609 
610 	/*
611 	 * Now that autoconfiguration has completed, we can determine
612 	 * the root and dump devices.
613 	 */
614 	cpu_rootconf();
615 	cpu_dumpconf();
616 
617 	/* Mount the root file system. */
618 	do {
619 		domountroothook(root_device);
620 		if ((error = vfs_mountroot())) {
621 			printf("cannot mount root, error = %d\n", error);
622 			boothowto |= RB_ASKNAME;
623 			setroot(root_device,
624 			    (rootdev != NODEV) ? DISKPART(rootdev) : 0);
625 		}
626 	} while (error != 0);
627 	mountroothook_destroy();
628 
629 	configure3();
630 
631 	/*
632 	 * Initialise the time-of-day clock, passing the time recorded
633 	 * in the root filesystem (if any) for use by systems that
634 	 * don't have a non-volatile time-of-day device.
635 	 */
636 	inittodr(rootfstime);
637 
638 	/*
639 	 * Now can look at time, having had a chance to verify the time
640 	 * from the file system.  Reset l->l_rtime as it may have been
641 	 * munched in mi_switch() after the time got set.
642 	 */
643 	getnanotime(&time);
644 	boottime = time;
645 
646 	mutex_enter(proc_lock);
647 	LIST_FOREACH(p, &allproc, p_list) {
648 		KASSERT((p->p_flag & PK_MARKER) == 0);
649 		mutex_enter(p->p_lock);
650 		TIMESPEC_TO_TIMEVAL(&p->p_stats->p_start, &time);
651 		LIST_FOREACH(l, &p->p_lwps, l_sibling) {
652 			lwp_lock(l);
653 			memset(&l->l_rtime, 0, sizeof(l->l_rtime));
654 			lwp_unlock(l);
655 		}
656 		mutex_exit(p->p_lock);
657 	}
658 	mutex_exit(proc_lock);
659 	binuptime(&curlwp->l_stime);
660 
661 	for (CPU_INFO_FOREACH(cii, ci)) {
662 		ci->ci_schedstate.spc_lastmod = time_second;
663 	}
664 
665 	/* Create the pageout daemon kernel thread. */
666 	uvm_swap_init();
667 	if (kthread_create(PRI_PGDAEMON, KTHREAD_MPSAFE, NULL, uvm_pageout,
668 	    NULL, NULL, "pgdaemon"))
669 		panic("fork pagedaemon");
670 
671 	/* Create the filesystem syncer kernel thread. */
672 	if (kthread_create(PRI_IOFLUSH, KTHREAD_MPSAFE, NULL, sched_sync,
673 	    NULL, NULL, "ioflush"))
674 		panic("fork syncer");
675 
676 	/* Create the aiodone daemon kernel thread. */
677 	if (workqueue_create(&uvm.aiodone_queue, "aiodoned",
678 	    uvm_aiodone_worker, NULL, PRI_VM, IPL_NONE, WQ_MPSAFE))
679 		panic("fork aiodoned");
680 
681 	/* Wait for final configure threads to complete. */
682 	config_finalize_mountroot();
683 
684 	/*
685 	 * Okay, now we can let init(8) exec!  It's off to userland!
686 	 */
687 	mutex_enter(proc_lock);
688 	start_init_exec = 1;
689 	cv_broadcast(&lbolt);
690 	mutex_exit(proc_lock);
691 
692 	/* The scheduler is an infinite loop. */
693 	uvm_scheduler();
694 	/* NOTREACHED */
695 }
696 
697 /*
698  * Configure the system's hardware.
699  */
700 static void
701 configure(void)
702 {
703 
704 	/* Initialize autoconf data structures. */
705 	config_init_mi();
706 	/*
707 	 * XXX
708 	 * callout_setfunc() requires mutex(9) so it can't be in config_init()
709 	 * on amiga and atari which use config_init() and autoconf(9) fucntions
710 	 * to initialize console devices.
711 	 */
712 	config_twiddle_init();
713 
714 	pmf_init();
715 
716 	/* Initialize driver modules */
717 	module_init_class(MODULE_CLASS_DRIVER);
718 
719 	userconf_init();
720 	if (boothowto & RB_USERCONF)
721 		userconf_prompt();
722 
723 	if ((boothowto & (AB_SILENT|AB_VERBOSE)) == AB_SILENT) {
724 		printf_nolog("Detecting hardware...");
725 	}
726 
727 	/*
728 	 * Do the machine-dependent portion of autoconfiguration.  This
729 	 * sets the configuration machinery here in motion by "finding"
730 	 * the root bus.  When this function returns, we expect interrupts
731 	 * to be enabled.
732 	 */
733 	cpu_configure();
734 }
735 
736 static void
737 configure2(void)
738 {
739 	CPU_INFO_ITERATOR cii;
740 	struct cpu_info *ci;
741 	int s;
742 
743 	/*
744 	 * Now that we've found all the hardware, start the real time
745 	 * and statistics clocks.
746 	 */
747 	initclocks();
748 
749 	cold = 0;	/* clocks are running, we're warm now! */
750 	s = splsched();
751 	curcpu()->ci_schedstate.spc_flags |= SPCF_RUNNING;
752 	splx(s);
753 
754 	/* Boot the secondary processors. */
755 	for (CPU_INFO_FOREACH(cii, ci)) {
756 		uvm_cpu_attach(ci);
757 	}
758 	mp_online = true;
759 #if defined(MULTIPROCESSOR)
760 	cpu_boot_secondary_processors();
761 #endif
762 
763 	/* Setup the runqueues and scheduler. */
764 	runq_init();
765 	synch_init();
766 
767 	/*
768 	 * Bus scans can make it appear as if the system has paused, so
769 	 * twiddle constantly while config_interrupts() jobs are running.
770 	 */
771 	config_twiddle_fn(NULL);
772 
773 	/*
774 	 * Create threads to call back and finish configuration for
775 	 * devices that want interrupts enabled.
776 	 */
777 	config_create_interruptthreads();
778 }
779 
780 static void
781 configure3(void)
782 {
783 
784 	/*
785 	 * Create threads to call back and finish configuration for
786 	 * devices that want the mounted root file system.
787 	 */
788 	config_create_mountrootthreads();
789 
790 	/* Get the threads going and into any sleeps before continuing. */
791 	yield();
792 }
793 
794 static void
795 rootconf_handle_wedges(void)
796 {
797 	struct disklabel label;
798 	struct partition *p;
799 	struct vnode *vp;
800 	daddr_t startblk;
801 	uint64_t nblks;
802 	device_t dev;
803 	int error;
804 
805 	if (booted_nblks) {
806 		/*
807 		 * bootloader passed geometry
808 		 */
809 		dev      = booted_device;
810 		startblk = booted_startblk;
811 		nblks    = booted_nblks;
812 
813 		/*
814 		 * keep booted_device and booted_partition
815 		 * in case the kernel doesn't identify a wedge
816 		 */
817 	} else {
818 		/*
819 		 * bootloader passed partition number
820 		 *
821 		 * We cannot ask the partition device directly when it is
822 		 * covered by a wedge. Instead we look up the geometry in
823 		 * the disklabel.
824 		 */
825 		vp = opendisk(booted_device);
826 
827 		if (vp == NULL)
828 			return;
829 
830 		error = VOP_IOCTL(vp, DIOCGDINFO, &label, FREAD, NOCRED);
831 		VOP_CLOSE(vp, FREAD, NOCRED);
832 		vput(vp);
833 		if (error)
834 			return;
835 
836 		KASSERT(booted_partition >= 0
837 			&& booted_partition < MAXPARTITIONS);
838 
839 		p = &label.d_partitions[booted_partition];
840 
841 		dev      = booted_device;
842 		startblk = p->p_offset;
843 		nblks    = p->p_size;
844 	}
845 
846 	dev = dkwedge_find_partition(dev, startblk, nblks);
847 	if (dev != NULL) {
848 		booted_device = dev;
849 		booted_partition = 0;
850 	}
851 }
852 
853 void
854 rootconf(void)
855 {
856 	if (booted_device != NULL)
857 		rootconf_handle_wedges();
858 
859 	setroot(booted_device, booted_partition);
860 }
861 
862 static void
863 check_console(struct lwp *l)
864 {
865 	struct vnode *vp;
866 	int error;
867 
868 	error = namei_simple_kernel("/dev/console",
869 				NSM_FOLLOW_NOEMULROOT, &vp);
870 	if (error == 0)
871 		vrele(vp);
872 	else if (error == ENOENT)
873 		printf("warning: no /dev/console\n");
874 	else
875 		printf("warning: lookup /dev/console: error %d\n", error);
876 }
877 
878 /*
879  * List of paths to try when searching for "init".
880  */
881 static const char * const initpaths[] = {
882 	"/sbin/init",
883 	"/sbin/oinit",
884 	"/sbin/init.bak",
885 	"/rescue/init",
886 	NULL,
887 };
888 
889 /*
890  * Start the initial user process; try exec'ing each pathname in "initpaths".
891  * The program is invoked with one argument containing the boot flags.
892  */
893 static void
894 start_init(void *arg)
895 {
896 	struct lwp *l = arg;
897 	struct proc *p = l->l_proc;
898 	vaddr_t addr;
899 	struct sys_execve_args /* {
900 		syscallarg(const char *) path;
901 		syscallarg(char * const *) argp;
902 		syscallarg(char * const *) envp;
903 	} */ args;
904 	int options, i, error;
905 	register_t retval[2];
906 	char flags[4], *flagsp;
907 	const char *path, *slash;
908 	char *ucp, **uap, *arg0, *arg1, *argv[3];
909 	char ipath[129];
910 	int ipx, len;
911 
912 	/*
913 	 * Now in process 1.
914 	 */
915 	strncpy(p->p_comm, "init", MAXCOMLEN);
916 
917 	/*
918 	 * Wait for main() to tell us that it's safe to exec.
919 	 */
920 	mutex_enter(proc_lock);
921 	while (start_init_exec == 0)
922 		cv_wait(&lbolt, proc_lock);
923 	mutex_exit(proc_lock);
924 
925 	/*
926 	 * This is not the right way to do this.  We really should
927 	 * hand-craft a descriptor onto /dev/console to hand to init,
928 	 * but that's a _lot_ more work, and the benefit from this easy
929 	 * hack makes up for the "good is the enemy of the best" effect.
930 	 */
931 	check_console(l);
932 
933 	/*
934 	 * Need just enough stack to hold the faked-up "execve()" arguments.
935 	 */
936 	addr = (vaddr_t)STACK_ALLOC(USRSTACK, PAGE_SIZE);
937 	if (uvm_map(&p->p_vmspace->vm_map, &addr, PAGE_SIZE,
938                     NULL, UVM_UNKNOWN_OFFSET, 0,
939                     UVM_MAPFLAG(UVM_PROT_ALL, UVM_PROT_ALL, UVM_INH_COPY,
940 		    UVM_ADV_NORMAL,
941                     UVM_FLAG_FIXED|UVM_FLAG_OVERLAY|UVM_FLAG_COPYONW)) != 0)
942 		panic("init: couldn't allocate argument space");
943 	p->p_vmspace->vm_maxsaddr = (void *)STACK_MAX(addr, PAGE_SIZE);
944 
945 	ipx = 0;
946 	while (1) {
947 		if (boothowto & RB_ASKNAME) {
948 			printf("init path");
949 			if (initpaths[ipx])
950 				printf(" (default %s)", initpaths[ipx]);
951 			printf(": ");
952 			len = cngetsn(ipath, sizeof(ipath)-1);
953 			if (len == 4 && strcmp(ipath, "halt") == 0) {
954 				cpu_reboot(RB_HALT, NULL);
955 			} else if (len == 6 && strcmp(ipath, "reboot") == 0) {
956 				cpu_reboot(0, NULL);
957 #if defined(DDB)
958 			} else if (len == 3 && strcmp(ipath, "ddb") == 0) {
959 				console_debugger();
960 				continue;
961 #endif
962 			} else if (len > 0 && ipath[0] == '/') {
963 				ipath[len] = '\0';
964 				path = ipath;
965 			} else if (len == 0 && initpaths[ipx] != NULL) {
966 				path = initpaths[ipx++];
967 			} else {
968 				printf("use absolute path, ");
969 #if defined(DDB)
970 				printf("\"ddb\", ");
971 #endif
972 				printf("\"halt\", or \"reboot\"\n");
973 				continue;
974 			}
975 		} else {
976 			if ((path = initpaths[ipx++]) == NULL) {
977 				ipx = 0;
978 				boothowto |= RB_ASKNAME;
979 				continue;
980 			}
981 		}
982 
983 		ucp = (char *)USRSTACK;
984 
985 		/*
986 		 * Construct the boot flag argument.
987 		 */
988 		flagsp = flags;
989 		*flagsp++ = '-';
990 		options = 0;
991 
992 		if (boothowto & RB_SINGLE) {
993 			*flagsp++ = 's';
994 			options = 1;
995 		}
996 #ifdef notyet
997 		if (boothowto & RB_FASTBOOT) {
998 			*flagsp++ = 'f';
999 			options = 1;
1000 		}
1001 #endif
1002 
1003 		/*
1004 		 * Move out the flags (arg 1), if necessary.
1005 		 */
1006 		if (options != 0) {
1007 			*flagsp++ = '\0';
1008 			i = flagsp - flags;
1009 #ifdef DEBUG
1010 			aprint_normal("init: copying out flags `%s' %d\n", flags, i);
1011 #endif
1012 			arg1 = STACK_ALLOC(ucp, i);
1013 			ucp = STACK_MAX(arg1, i);
1014 			if ((error = copyout((void *)flags, arg1, i)) != 0)
1015 				goto copyerr;
1016 		} else
1017 			arg1 = NULL;
1018 
1019 		/*
1020 		 * Move out the file name (also arg 0).
1021 		 */
1022 		i = strlen(path) + 1;
1023 #ifdef DEBUG
1024 		aprint_normal("init: copying out path `%s' %d\n", path, i);
1025 #else
1026 		if (boothowto & RB_ASKNAME || path != initpaths[0])
1027 			printf("init: trying %s\n", path);
1028 #endif
1029 		arg0 = STACK_ALLOC(ucp, i);
1030 		ucp = STACK_MAX(arg0, i);
1031 		if ((error = copyout(path, arg0, i)) != 0)
1032 			goto copyerr;
1033 
1034 		/*
1035 		 * Move out the arg pointers.
1036 		 */
1037 		ucp = (void *)STACK_ALIGN(ucp, STACK_ALIGNBYTES);
1038 		uap = (char **)STACK_ALLOC(ucp, sizeof(char *) * 3);
1039 		SCARG(&args, path) = arg0;
1040 		SCARG(&args, argp) = uap;
1041 		SCARG(&args, envp) = NULL;
1042 		slash = strrchr(path, '/');
1043 
1044 		argv[0] = slash ? arg0 + (slash + 1 - path) : arg0;
1045 		argv[1] = arg1;
1046 		argv[2] = NULL;
1047 		if ((error = copyout(argv, uap, sizeof(args))) != 0)
1048 			goto copyerr;
1049 
1050 		/*
1051 		 * Now try to exec the program.  If can't for any reason
1052 		 * other than it doesn't exist, complain.
1053 		 */
1054 		error = sys_execve(l, &args, retval);
1055 		if (error == 0 || error == EJUSTRETURN) {
1056 			KERNEL_UNLOCK_LAST(l);
1057 			return;
1058 		}
1059 		printf("exec %s: error %d\n", path, error);
1060 	}
1061 	printf("init: not found\n");
1062 	panic("no init");
1063 copyerr:
1064 	panic("copyout %d", error);
1065 }
1066 
1067 /*
1068  * calculate cache size (in bytes) from physmem and vsize.
1069  */
1070 vaddr_t
1071 calc_cache_size(vsize_t vsize, int pct, int va_pct)
1072 {
1073 	paddr_t t;
1074 
1075 	/* XXX should consider competing cache if any */
1076 	/* XXX should consider submaps */
1077 	t = (uintmax_t)physmem * pct / 100 * PAGE_SIZE;
1078 	if (vsize != 0) {
1079 		vsize = (uintmax_t)vsize * va_pct / 100;
1080 		if (t > vsize) {
1081 			t = vsize;
1082 		}
1083 	}
1084 	return t;
1085 }
1086 
1087 /*
1088  * Print the system start up banner.
1089  *
1090  * - Print a limited banner if AB_SILENT.
1091  * - Always send normal banner to the log.
1092  */
1093 #define MEM_PBUFSIZE	sizeof("99999 MB")
1094 
1095 void
1096 banner(void)
1097 {
1098 	static char notice[] = " Notice: this software is "
1099 	    "protected by copyright";
1100 	char pbuf[81];
1101 	void (*pr)(const char *, ...) __printflike(1, 2);
1102 	int i;
1103 
1104 	if ((boothowto & AB_SILENT) != 0) {
1105 		snprintf(pbuf, sizeof(pbuf), "%s %s (%s)",
1106 		    ostype, osrelease, kernel_ident);
1107 		printf_nolog("%s", pbuf);
1108 		for (i = 80 - strlen(pbuf) - sizeof(notice); i > 0; i--)
1109 			printf(" ");
1110 		printf_nolog("%s\n", notice);
1111 		pr = aprint_normal;
1112 	} else {
1113 		pr = printf;
1114 	}
1115 
1116 	memset(pbuf, 0, sizeof(pbuf));
1117 	(*pr)("%s%s", copyright, version);
1118 	format_bytes(pbuf, MEM_PBUFSIZE, ctob((uint64_t)physmem));
1119 	(*pr)("total memory = %s\n", pbuf);
1120 	format_bytes(pbuf, MEM_PBUFSIZE, ctob((uint64_t)uvmexp.free));
1121 	(*pr)("avail memory = %s\n", pbuf);
1122 }
1123