xref: /netbsd-src/sys/kern/init_main.c (revision bdc22b2e01993381dcefeff2bc9b56ca75a4235c)
1 /*	$NetBSD: init_main.c,v 1.498 2018/07/03 03:37:03 ozaki-r 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 1.498 2018/07/03 03:37:03 ozaki-r 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 #ifdef IPSEC
182 #include <netipsec/ipsec.h>
183 #endif
184 #include <sys/domain.h>
185 #include <sys/namei.h>
186 #include <sys/rnd.h>
187 #include <sys/pipe.h>
188 #if NVERIEXEC > 0
189 #include <sys/verified_exec.h>
190 #endif /* NVERIEXEC > 0 */
191 #ifdef KTRACE
192 #include <sys/ktrace.h>
193 #endif
194 #include <sys/kauth.h>
195 #include <net80211/ieee80211_netbsd.h>
196 #include <sys/cprng.h>
197 
198 #include <sys/syscall.h>
199 #include <sys/syscallargs.h>
200 
201 #include <sys/pax.h>
202 
203 #include <secmodel/secmodel.h>
204 
205 #include <ufs/ufs/quota.h>
206 
207 #include <miscfs/genfs/genfs.h>
208 #include <miscfs/specfs/specdev.h>
209 
210 #include <sys/cpu.h>
211 
212 #include <uvm/uvm.h>	/* extern struct uvm uvm */
213 
214 #include <dev/cons.h>
215 #include <dev/splash/splash.h>
216 
217 #include <net/bpf.h>
218 #include <net/if.h>
219 #include <net/pfil.h>
220 #include <net/raw_cb.h>
221 #include <net/if_llatbl.h>
222 
223 #include <prop/proplib.h>
224 
225 #include <sys/userconf.h>
226 
227 extern struct lwp lwp0;
228 extern time_t rootfstime;
229 
230 #ifndef curlwp
231 struct	lwp *curlwp = &lwp0;
232 #endif
233 struct	proc *initproc;
234 
235 struct	vnode *rootvp, *swapdev_vp;
236 int	boothowto;
237 int	cold __read_mostly = 1;		/* still working on startup */
238 struct timespec boottime;	        /* time at system startup - will only follow settime deltas */
239 
240 int	start_init_exec;		/* semaphore for start_init() */
241 
242 cprng_strong_t	*kern_cprng;
243 
244 static void check_console(struct lwp *l);
245 static void start_init(void *);
246 static void configure(void);
247 static void configure2(void);
248 static void configure3(void);
249 void main(void);
250 
251 /*
252  * System startup; initialize the world, create process 0, mount root
253  * filesystem, and fork to create init and pagedaemon.  Most of the
254  * hard work is done in the lower-level initialization routines including
255  * startup(), which does memory initialization and autoconfiguration.
256  */
257 void
258 main(void)
259 {
260 	struct timespec time;
261 	struct lwp *l;
262 	struct proc *p;
263 	int s, error;
264 #ifdef NVNODE_IMPLICIT
265 	int usevnodes;
266 #endif
267 	CPU_INFO_ITERATOR cii;
268 	struct cpu_info *ci;
269 
270 #ifdef DIAGNOSTIC
271 	/*
272 	 * Verify that CPU_INFO_FOREACH() knows about the boot CPU
273 	 * and only the boot CPU at this point.
274 	 */
275 	int cpucount = 0;
276 	for (CPU_INFO_FOREACH(cii, ci)) {
277 		KASSERT(ci == curcpu());
278 		cpucount++;
279 	}
280 	KASSERT(cpucount == 1);
281 #endif
282 
283 	l = &lwp0;
284 #ifndef LWP0_CPU_INFO
285 	l->l_cpu = curcpu();
286 #endif
287 	l->l_pflag |= LP_RUNNING;
288 
289 	/*
290 	 * Attempt to find console and initialize
291 	 * in case of early panic or other messages.
292 	 */
293 	consinit();
294 
295 	kernel_lock_init();
296 	once_init();
297 
298 	mi_cpu_init();
299 	kernconfig_lock_init();
300 	kthread_sysinit();
301 
302 	/* Initialize the device switch tables. */
303 	devsw_init();
304 
305 	/* Initialize event counters. */
306 	evcnt_init();
307 
308 	uvm_init();
309 	ubchist_init();
310 	kcpuset_sysinit();
311 
312 	prop_kern_init();
313 
314 #if ((NKSYMS > 0) || (NDDB > 0) || (NMODULAR > 0))
315 	ksyms_init();
316 #endif
317 	kprintf_init();
318 
319 	percpu_init();
320 
321 	/* Initialize lock caches. */
322 	mutex_obj_init();
323 	rw_obj_init();
324 
325 	/* Passive serialization. */
326 	pserialize_init();
327 
328 	/* Initialize the extent manager. */
329 	extent_init();
330 
331 	/* Do machine-dependent initialization. */
332 	cpu_startup();
333 
334 	/* Initialize the sysctl subsystem. */
335 	sysctl_init();
336 
337 	/* Initialize callouts, part 1. */
338 	callout_startup();
339 
340 	/* Initialize the kernel authorization subsystem. */
341 	kauth_init();
342 
343 	secmodel_init();
344 
345 	spec_init();
346 
347 	/*
348 	 * Set BPF op vector.  Can't do this in bpf attach, since
349 	 * network drivers attach before bpf.
350 	 */
351 	bpf_setops();
352 
353 	/* Start module system. */
354 	module_init();
355 
356 	/*
357 	 * Initialize the kernel authorization subsystem and start the
358 	 * default security model, if any. We need to do this early
359 	 * enough so that subsystems relying on any of the aforementioned
360 	 * can work properly. Since the security model may dictate the
361 	 * credential inheritance policy, it is needed at least before
362 	 * any process is created, specifically proc0.
363 	 */
364 	module_init_class(MODULE_CLASS_SECMODEL);
365 
366 	/* Initialize the buffer cache */
367 	bufinit();
368 	biohist_init();
369 
370 #ifdef KERNHIST
371 	sysctl_kernhist_init();
372 #endif
373 
374 
375 #if defined(SPLASHSCREEN) && defined(SPLASHSCREEN_IMAGE)
376 	size_t splash_size = (&_binary_splash_image_end -
377 	    &_binary_splash_image_start) * sizeof(void *);
378 	splash_setimage(&_binary_splash_image_start, splash_size);
379 #endif
380 
381 	/* Initialize sockets. */
382 	soinit();
383 
384 	/*
385 	 * The following things must be done before autoconfiguration.
386 	 */
387 	rnd_init();		/* initialize entropy pool */
388 
389 	cprng_init();		/* initialize cryptographic PRNG */
390 
391 	/* Initialize process and pgrp structures. */
392 	procinit();
393 	lwpinit();
394 
395 	/* Initialize signal-related data structures. */
396 	signal_init();
397 
398 	/* Initialize resource management. */
399 	resource_init();
400 
401 	/* Create process 0. */
402 	proc0_init();
403 	lwp0_init();
404 
405 	/* Disable preemption during boot. */
406 	kpreempt_disable();
407 
408 	/* Initialize the UID hash table. */
409 	uid_init();
410 
411 	/* Charge root for one process. */
412 	(void)chgproccnt(0, 1);
413 
414 	/* Initialize timekeeping. */
415 	time_init();
416 
417 	/* Initialize the run queues, turnstiles and sleep queues. */
418 	sched_rqinit();
419 	turnstile_init();
420 	sleeptab_init(&sleeptab);
421 
422 	sched_init();
423 
424 	/* Initialize processor-sets */
425 	psets_init();
426 
427 	/* Initialize cpufreq(9) */
428 	cpufreq_init();
429 
430 	/* MI initialization of the boot cpu */
431 	error = mi_cpu_attach(curcpu());
432 	KASSERT(error == 0);
433 
434 	/* Initialize timekeeping, part 2. */
435 	time_init2();
436 
437 	/*
438 	 * Initialize mbuf's.  Do this now because we might attempt to
439 	 * allocate mbufs or mbuf clusters during autoconfiguration.
440 	 */
441 	mbinit();
442 
443 	/* Initialize I/O statistics. */
444 	iostat_init();
445 
446 	/* Initialize the log device. */
447 	loginit();
448 
449 	/* Second part of module system initialization. */
450 	module_start_unload_thread();
451 
452 	/* Initialize the file systems. */
453 #ifdef NVNODE_IMPLICIT
454 	/*
455 	 * If maximum number of vnodes in namei vnode cache is not explicitly
456 	 * defined in kernel config, adjust the number such as we use roughly
457 	 * 10% of memory for vnodes and associated data structures in the
458 	 * assumed worst case.  Do not provide fewer than NVNODE vnodes.
459 	 */
460 	usevnodes = calc_cache_size(vmem_size(kmem_arena, VMEM_FREE|VMEM_ALLOC),
461 	    10, VNODE_KMEM_MAXPCT) / VNODE_COST;
462 	if (usevnodes > desiredvnodes)
463 		desiredvnodes = usevnodes;
464 #endif
465 	vfsinit();
466 	lf_init();
467 
468 	/* Initialize fstrans. */
469 	fstrans_init();
470 
471 	/* Initialize the file descriptor system. */
472 	fd_sys_init();
473 
474 	/* Initialize cwd structures */
475 	cwd_sys_init();
476 
477 	/* Initialize kqueue. */
478 	kqueue_init();
479 
480 	inittimecounter();
481 	ntp_init();
482 
483 	/* Initialize tty subsystem. */
484 	tty_init();
485 	ttyldisc_init();
486 
487 	/* Initialize the buffer cache, part 2. */
488 	bufinit2();
489 
490 	/* Initialize the disk wedge subsystem. */
491 	dkwedge_init();
492 
493 	/* Initialize the kernel strong PRNG. */
494 	kern_cprng = cprng_strong_create("kernel", IPL_VM,
495 					 CPRNG_INIT_ANY|CPRNG_REKEY_ANY);
496 
497 	/* Initialize pfil */
498 	pfil_init();
499 
500 	/* Initialize interfaces. */
501 	ifinit1();
502 
503 	spldebug_start();
504 
505 	/* Initialize sockets thread(s) */
506 	soinit1();
507 
508 	/*
509 	 * Initialize the bufq strategy sub-system and any built-in
510 	 * strategy modules - they may be needed by some devices during
511 	 * auto-configuration
512 	 */
513 	bufq_init();
514 	module_init_class(MODULE_CLASS_BUFQ);
515 
516 	/* Configure the system hardware.  This will enable interrupts. */
517 	configure();
518 
519 	/* Once all CPUs are detected, initialize the per-CPU cprng_fast.  */
520 	cprng_fast_init();
521 
522 	ssp_init();
523 
524 	ubc_init();		/* must be after autoconfig */
525 
526 	mm_init();
527 
528 	configure2();
529 
530 	ipi_sysinit();
531 
532 	/* Now timer is working.  Enable preemption. */
533 	kpreempt_enable();
534 
535 	/* Get the threads going and into any sleeps before continuing. */
536 	yield();
537 
538 	/* Enable deferred processing of RNG samples */
539 	rnd_init_softint();
540 
541 #ifdef RND_PRINTF
542 	/* Enable periodic injection of console output into entropy pool */
543 	kprintf_init_callout();
544 #endif
545 
546 	vmem_rehash_start();	/* must be before exec_init */
547 
548 	/* Initialize exec structures */
549 	exec_init(1);		/* seminit calls exithook_establish() */
550 
551 #if NVERIEXEC > 0
552 	/*
553 	 * Initialise the Veriexec subsystem.
554 	 */
555 	veriexec_init();
556 #endif /* NVERIEXEC > 0 */
557 
558 	pax_init();
559 
560 #ifdef	IPSEC
561 	/* Attach network crypto subsystem */
562 	ipsec_attach();
563 #endif
564 
565 	/*
566 	 * Initialize protocols.  Block reception of incoming packets
567 	 * until everything is ready.
568 	 */
569 	s = splnet();
570 	ifinit();
571 #if defined(INET) || defined(INET6)
572 	lltableinit();
573 #endif
574 	domaininit(true);
575 	ifinit_post();
576 	if_attachdomain();
577 	splx(s);
578 
579 #ifdef GPROF
580 	/* Initialize kernel profiling. */
581 	kmstartup();
582 #endif
583 
584 	/* Initialize system accounting. */
585 	acct_init();
586 
587 #ifndef PIPE_SOCKETPAIR
588 	/* Initialize pipes. */
589 	pipe_init();
590 #endif
591 
592 #ifdef KTRACE
593 	/* Initialize ktrace. */
594 	ktrinit();
595 #endif
596 
597 	machdep_init();
598 
599 	procinit_sysctl();
600 
601 	scdebug_init();
602 
603 	/*
604 	 * Create process 1 (init(8)).  We do this now, as Unix has
605 	 * historically had init be process 1, and changing this would
606 	 * probably upset a lot of people.
607 	 *
608 	 * Note that process 1 won't immediately exec init(8), but will
609 	 * wait for us to inform it that the root file system has been
610 	 * mounted.
611 	 */
612 	if (fork1(l, 0, SIGCHLD, NULL, 0, start_init, NULL, NULL))
613 		panic("fork init");
614 
615 	/*
616 	 * Load any remaining builtin modules, and hand back temporary
617 	 * storage to the VM system.  Then require force when loading any
618 	 * remaining un-init'ed built-in modules to avoid later surprises.
619 	 */
620 	module_init_class(MODULE_CLASS_ANY);
621 	module_builtin_require_force();
622 
623 	/*
624 	 * Finalize configuration now that all real devices have been
625 	 * found.  This needs to be done before the root device is
626 	 * selected, since finalization may create the root device.
627 	 */
628 	config_finalize();
629 
630 	sysctl_finalize();
631 
632 	/*
633 	 * Now that autoconfiguration has completed, we can determine
634 	 * the root and dump devices.
635 	 */
636 	cpu_rootconf();
637 	cpu_dumpconf();
638 
639 	/* Mount the root file system. */
640 	do {
641 		domountroothook(root_device);
642 		if ((error = vfs_mountroot())) {
643 			printf("cannot mount root, error = %d\n", error);
644 			boothowto |= RB_ASKNAME;
645 			setroot(root_device,
646 			    (rootdev != NODEV) ? DISKPART(rootdev) : 0);
647 		}
648 	} while (error != 0);
649 	mountroothook_destroy();
650 
651 	configure3();
652 
653 	/*
654 	 * Initialise the time-of-day clock, passing the time recorded
655 	 * in the root filesystem (if any) for use by systems that
656 	 * don't have a non-volatile time-of-day device.
657 	 */
658 	inittodr(rootfstime);
659 
660 	/*
661 	 * Now can look at time, having had a chance to verify the time
662 	 * from the file system.  Reset l->l_rtime as it may have been
663 	 * munched in mi_switch() after the time got set.
664 	 */
665 	getnanotime(&time);
666 	boottime = time;
667 
668 	mutex_enter(proc_lock);
669 	LIST_FOREACH(p, &allproc, p_list) {
670 		KASSERT((p->p_flag & PK_MARKER) == 0);
671 		mutex_enter(p->p_lock);
672 		TIMESPEC_TO_TIMEVAL(&p->p_stats->p_start, &time);
673 		LIST_FOREACH(l, &p->p_lwps, l_sibling) {
674 			lwp_lock(l);
675 			memset(&l->l_rtime, 0, sizeof(l->l_rtime));
676 			lwp_unlock(l);
677 		}
678 		mutex_exit(p->p_lock);
679 	}
680 	mutex_exit(proc_lock);
681 	binuptime(&curlwp->l_stime);
682 
683 	for (CPU_INFO_FOREACH(cii, ci)) {
684 		ci->ci_schedstate.spc_lastmod = time_second;
685 	}
686 
687 	/* Create the pageout daemon kernel thread. */
688 	uvm_swap_init();
689 	if (kthread_create(PRI_PGDAEMON, KTHREAD_MPSAFE, NULL, uvm_pageout,
690 	    NULL, NULL, "pgdaemon"))
691 		panic("fork pagedaemon");
692 
693 	/* Create the filesystem syncer kernel thread. */
694 	if (kthread_create(PRI_IOFLUSH, KTHREAD_MPSAFE, NULL, sched_sync,
695 	    NULL, NULL, "ioflush"))
696 		panic("fork syncer");
697 
698 	/* Create the aiodone daemon kernel thread. */
699 	if (workqueue_create(&uvm.aiodone_queue, "aiodoned",
700 	    uvm_aiodone_worker, NULL, PRI_VM, IPL_NONE, WQ_MPSAFE))
701 		panic("fork aiodoned");
702 
703 	/* Wait for final configure threads to complete. */
704 	config_finalize_mountroot();
705 
706 	/*
707 	 * Okay, now we can let init(8) exec!  It's off to userland!
708 	 */
709 	mutex_enter(proc_lock);
710 	start_init_exec = 1;
711 	cv_broadcast(&lbolt);
712 	mutex_exit(proc_lock);
713 
714 	/* The scheduler is an infinite loop. */
715 	uvm_scheduler();
716 	/* NOTREACHED */
717 }
718 
719 /*
720  * Configure the system's hardware.
721  */
722 static void
723 configure(void)
724 {
725 
726 	/* Initialize autoconf data structures. */
727 	config_init_mi();
728 	/*
729 	 * XXX
730 	 * callout_setfunc() requires mutex(9) so it can't be in config_init()
731 	 * on amiga and atari which use config_init() and autoconf(9) fucntions
732 	 * to initialize console devices.
733 	 */
734 	config_twiddle_init();
735 
736 	pmf_init();
737 
738 	/* Initialize driver modules */
739 	module_init_class(MODULE_CLASS_DRIVER);
740 
741 	userconf_init();
742 	if (boothowto & RB_USERCONF)
743 		userconf_prompt();
744 
745 	if ((boothowto & (AB_SILENT|AB_VERBOSE)) == AB_SILENT) {
746 		printf_nolog("Detecting hardware...");
747 	}
748 
749 	/*
750 	 * Do the machine-dependent portion of autoconfiguration.  This
751 	 * sets the configuration machinery here in motion by "finding"
752 	 * the root bus.  When this function returns, we expect interrupts
753 	 * to be enabled.
754 	 */
755 	cpu_configure();
756 }
757 
758 static void
759 configure2(void)
760 {
761 	CPU_INFO_ITERATOR cii;
762 	struct cpu_info *ci;
763 	int s;
764 
765 	/*
766 	 * Now that we've found all the hardware, start the real time
767 	 * and statistics clocks.
768 	 */
769 	initclocks();
770 
771 	cold = 0;	/* clocks are running, we're warm now! */
772 	s = splsched();
773 	curcpu()->ci_schedstate.spc_flags |= SPCF_RUNNING;
774 	splx(s);
775 
776 	/* Boot the secondary processors. */
777 	for (CPU_INFO_FOREACH(cii, ci)) {
778 		uvm_cpu_attach(ci);
779 	}
780 	mp_online = true;
781 #if defined(MULTIPROCESSOR)
782 	cpu_boot_secondary_processors();
783 #endif
784 
785 	/* Setup the runqueues and scheduler. */
786 	runq_init();
787 	synch_init();
788 
789 	/*
790 	 * Bus scans can make it appear as if the system has paused, so
791 	 * twiddle constantly while config_interrupts() jobs are running.
792 	 */
793 	config_twiddle_fn(NULL);
794 
795 	/*
796 	 * Create threads to call back and finish configuration for
797 	 * devices that want interrupts enabled.
798 	 */
799 	config_create_interruptthreads();
800 }
801 
802 static void
803 configure3(void)
804 {
805 
806 	/*
807 	 * Create threads to call back and finish configuration for
808 	 * devices that want the mounted root file system.
809 	 */
810 	config_create_mountrootthreads();
811 
812 	/* Get the threads going and into any sleeps before continuing. */
813 	yield();
814 }
815 
816 static void
817 rootconf_handle_wedges(void)
818 {
819 	struct disklabel label;
820 	struct partition *p;
821 	struct vnode *vp;
822 	daddr_t startblk;
823 	uint64_t nblks;
824 	device_t dev;
825 	int error;
826 
827 	if (booted_nblks) {
828 		/*
829 		 * bootloader passed geometry
830 		 */
831 		dev      = booted_device;
832 		startblk = booted_startblk;
833 		nblks    = booted_nblks;
834 
835 		/*
836 		 * keep booted_device and booted_partition
837 		 * in case the kernel doesn't identify a wedge
838 		 */
839 	} else {
840 		/*
841 		 * bootloader passed partition number
842 		 *
843 		 * We cannot ask the partition device directly when it is
844 		 * covered by a wedge. Instead we look up the geometry in
845 		 * the disklabel.
846 		 */
847 		vp = opendisk(booted_device);
848 
849 		if (vp == NULL)
850 			return;
851 
852 		error = VOP_IOCTL(vp, DIOCGDINFO, &label, FREAD, NOCRED);
853 		VOP_CLOSE(vp, FREAD, NOCRED);
854 		vput(vp);
855 		if (error)
856 			return;
857 
858 		KASSERT(booted_partition >= 0
859 			&& booted_partition < MAXPARTITIONS);
860 
861 		p = &label.d_partitions[booted_partition];
862 
863 		dev      = booted_device;
864 		startblk = p->p_offset;
865 		nblks    = p->p_size;
866 	}
867 
868 	dev = dkwedge_find_partition(dev, startblk, nblks);
869 	if (dev != NULL) {
870 		booted_device = dev;
871 		booted_partition = 0;
872 	}
873 }
874 
875 void
876 rootconf(void)
877 {
878 	if (booted_device != NULL)
879 		rootconf_handle_wedges();
880 
881 	setroot(booted_device, booted_partition);
882 }
883 
884 static void
885 check_console(struct lwp *l)
886 {
887 	struct vnode *vp;
888 	int error;
889 
890 	error = namei_simple_kernel("/dev/console",
891 				NSM_FOLLOW_NOEMULROOT, &vp);
892 	if (error == 0)
893 		vrele(vp);
894 	else if (error == ENOENT)
895 		printf("warning: no /dev/console\n");
896 	else
897 		printf("warning: lookup /dev/console: error %d\n", error);
898 }
899 
900 /*
901  * List of paths to try when searching for "init".
902  */
903 static const char * const initpaths[] = {
904 	"/sbin/init",
905 	"/sbin/oinit",
906 	"/sbin/init.bak",
907 	"/rescue/init",
908 	NULL,
909 };
910 
911 /*
912  * Start the initial user process; try exec'ing each pathname in "initpaths".
913  * The program is invoked with one argument containing the boot flags.
914  */
915 static void
916 start_init(void *arg)
917 {
918 	struct lwp *l = arg;
919 	struct proc *p = l->l_proc;
920 	vaddr_t addr;
921 	struct sys_execve_args /* {
922 		syscallarg(const char *) path;
923 		syscallarg(char * const *) argp;
924 		syscallarg(char * const *) envp;
925 	} */ args;
926 	int options, i, error;
927 	register_t retval[2];
928 	char flags[4], *flagsp;
929 	const char *path, *slash;
930 	char *ucp, **uap, *arg0, *arg1, *argv[3];
931 	char ipath[129];
932 	int ipx, len;
933 
934 	initproc = p;
935 
936 	/*
937 	 * Now in process 1.
938 	 */
939 	strncpy(p->p_comm, "init", MAXCOMLEN);
940 
941 	/*
942 	 * Wait for main() to tell us that it's safe to exec.
943 	 */
944 	mutex_enter(proc_lock);
945 	while (start_init_exec == 0)
946 		cv_wait(&lbolt, proc_lock);
947 	mutex_exit(proc_lock);
948 
949 	/*
950 	 * This is not the right way to do this.  We really should
951 	 * hand-craft a descriptor onto /dev/console to hand to init,
952 	 * but that's a _lot_ more work, and the benefit from this easy
953 	 * hack makes up for the "good is the enemy of the best" effect.
954 	 */
955 	check_console(l);
956 
957 	/*
958 	 * Need just enough stack to hold the faked-up "execve()" arguments.
959 	 */
960 	addr = (vaddr_t)STACK_ALLOC(USRSTACK, PAGE_SIZE);
961 	if (uvm_map(&p->p_vmspace->vm_map, &addr, PAGE_SIZE,
962 	    NULL, UVM_UNKNOWN_OFFSET, 0,
963 	    UVM_MAPFLAG(UVM_PROT_RW, UVM_PROT_RW, UVM_INH_COPY,
964 	    UVM_ADV_NORMAL,
965 	    UVM_FLAG_FIXED|UVM_FLAG_OVERLAY|UVM_FLAG_COPYONW)) != 0)
966 		panic("init: couldn't allocate argument space");
967 	p->p_vmspace->vm_maxsaddr = (void *)STACK_MAX(addr, PAGE_SIZE);
968 
969 	ipx = 0;
970 	while (1) {
971 		if (boothowto & RB_ASKNAME) {
972 			printf("init path");
973 			if (initpaths[ipx])
974 				printf(" (default %s)", initpaths[ipx]);
975 			printf(": ");
976 			len = cngetsn(ipath, sizeof(ipath)-1);
977 			if (len == 4 && strcmp(ipath, "halt") == 0) {
978 				cpu_reboot(RB_HALT, NULL);
979 			} else if (len == 6 && strcmp(ipath, "reboot") == 0) {
980 				cpu_reboot(0, NULL);
981 #if defined(DDB)
982 			} else if (len == 3 && strcmp(ipath, "ddb") == 0) {
983 				console_debugger();
984 				continue;
985 #endif
986 			} else if (len > 0 && ipath[0] == '/') {
987 				ipath[len] = '\0';
988 				path = ipath;
989 			} else if (len == 0 && initpaths[ipx] != NULL) {
990 				path = initpaths[ipx++];
991 			} else {
992 				printf("use absolute path, ");
993 #if defined(DDB)
994 				printf("\"ddb\", ");
995 #endif
996 				printf("\"halt\", or \"reboot\"\n");
997 				continue;
998 			}
999 		} else {
1000 			if ((path = initpaths[ipx++]) == NULL) {
1001 				ipx = 0;
1002 				boothowto |= RB_ASKNAME;
1003 				continue;
1004 			}
1005 		}
1006 
1007 		ucp = (char *)USRSTACK;
1008 
1009 		/*
1010 		 * Construct the boot flag argument.
1011 		 */
1012 		flagsp = flags;
1013 		*flagsp++ = '-';
1014 		options = 0;
1015 
1016 		if (boothowto & RB_SINGLE) {
1017 			*flagsp++ = 's';
1018 			options = 1;
1019 		}
1020 #ifdef notyet
1021 		if (boothowto & RB_FASTBOOT) {
1022 			*flagsp++ = 'f';
1023 			options = 1;
1024 		}
1025 #endif
1026 
1027 		/*
1028 		 * Move out the flags (arg 1), if necessary.
1029 		 */
1030 		if (options != 0) {
1031 			*flagsp++ = '\0';
1032 			i = flagsp - flags;
1033 #ifdef DEBUG
1034 			aprint_normal("init: copying out flags `%s' %d\n", flags, i);
1035 #endif
1036 			arg1 = STACK_ALLOC(ucp, i);
1037 			ucp = STACK_MAX(arg1, i);
1038 			if ((error = copyout((void *)flags, arg1, i)) != 0)
1039 				goto copyerr;
1040 		} else
1041 			arg1 = NULL;
1042 
1043 		/*
1044 		 * Move out the file name (also arg 0).
1045 		 */
1046 		i = strlen(path) + 1;
1047 #ifdef DEBUG
1048 		aprint_normal("init: copying out path `%s' %d\n", path, i);
1049 #else
1050 		if (boothowto & RB_ASKNAME || path != initpaths[0])
1051 			printf("init: trying %s\n", path);
1052 #endif
1053 		arg0 = STACK_ALLOC(ucp, i);
1054 		ucp = STACK_MAX(arg0, i);
1055 		if ((error = copyout(path, arg0, i)) != 0)
1056 			goto copyerr;
1057 
1058 		/*
1059 		 * Move out the arg pointers.
1060 		 */
1061 		ucp = (void *)STACK_ALIGN(ucp, STACK_ALIGNBYTES);
1062 		uap = (char **)STACK_ALLOC(ucp, sizeof(argv));
1063 		SCARG(&args, path) = arg0;
1064 		SCARG(&args, argp) = uap;
1065 		SCARG(&args, envp) = NULL;
1066 		slash = strrchr(path, '/');
1067 
1068 		argv[0] = slash ? arg0 + (slash + 1 - path) : arg0;
1069 		argv[1] = arg1;
1070 		argv[2] = NULL;
1071 		if ((error = copyout(argv, uap, sizeof(argv))) != 0)
1072 			goto copyerr;
1073 
1074 		/*
1075 		 * Now try to exec the program.  If it can't for any reason
1076 		 * other than it doesn't exist, complain.
1077 		 */
1078 		error = sys_execve(l, &args, retval);
1079 		if (error == 0 || error == EJUSTRETURN) {
1080 			KERNEL_UNLOCK_LAST(l);
1081 			return;
1082 		}
1083 		printf("exec %s: error %d\n", path, error);
1084 	}
1085 	printf("init: not found\n");
1086 	panic("no init");
1087 copyerr:
1088 	panic("copyout %d", error);
1089 }
1090 
1091 /*
1092  * calculate cache size (in bytes) from physmem and vsize.
1093  */
1094 vaddr_t
1095 calc_cache_size(vsize_t vsize, int pct, int va_pct)
1096 {
1097 	paddr_t t;
1098 
1099 	/* XXX should consider competing cache if any */
1100 	/* XXX should consider submaps */
1101 	t = (uintmax_t)physmem * pct / 100 * PAGE_SIZE;
1102 	if (vsize != 0) {
1103 		vsize = (uintmax_t)vsize * va_pct / 100;
1104 		if (t > vsize) {
1105 			t = vsize;
1106 		}
1107 	}
1108 	return t;
1109 }
1110 
1111 /*
1112  * Print the system start up banner.
1113  *
1114  * - Print a limited banner if AB_SILENT.
1115  * - Always send normal banner to the log.
1116  */
1117 #define MEM_PBUFSIZE	sizeof("99999 MB")
1118 
1119 void
1120 banner(void)
1121 {
1122 	static char notice[] = " Notice: this software is "
1123 	    "protected by copyright";
1124 	char pbuf[81];
1125 	void (*pr)(const char *, ...) __printflike(1, 2);
1126 	int i;
1127 
1128 	if ((boothowto & AB_SILENT) != 0) {
1129 		snprintf(pbuf, sizeof(pbuf), "%s %s (%s)",
1130 		    ostype, osrelease, kernel_ident);
1131 		printf_nolog("%s", pbuf);
1132 		for (i = 80 - strlen(pbuf) - sizeof(notice); i > 0; i--)
1133 			printf(" ");
1134 		printf_nolog("%s\n", notice);
1135 		pr = aprint_normal;
1136 	} else {
1137 		pr = printf;
1138 	}
1139 
1140 	memset(pbuf, 0, sizeof(pbuf));
1141 	(*pr)("%s%s", copyright, version);
1142 	format_bytes(pbuf, MEM_PBUFSIZE, ctob((uint64_t)physmem));
1143 	(*pr)("total memory = %s\n", pbuf);
1144 	format_bytes(pbuf, MEM_PBUFSIZE, ctob((uint64_t)uvmexp.free));
1145 	(*pr)("avail memory = %s\n", pbuf);
1146 }
1147