xref: /netbsd-src/sys/kern/init_main.c (revision f17b710f3d406bee67aa39c65053114ab78297c5)
1 /*	$NetBSD: init_main.c,v 1.467 2015/05/06 15:57:08 hannken Exp $	*/
2 
3 /*-
4  * Copyright (c) 2008, 2009 The NetBSD Foundation, Inc.
5  * All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  *
16  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
17  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
18  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
19  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
20  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
21  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
22  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
23  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
24  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
25  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
26  * POSSIBILITY OF SUCH DAMAGE.
27  */
28 
29 /*
30  * Copyright (c) 1982, 1986, 1989, 1991, 1992, 1993
31  *	The Regents of the University of California.  All rights reserved.
32  * (c) UNIX System Laboratories, Inc.
33  * All or some portions of this file are derived from material licensed
34  * to the University of California by American Telephone and Telegraph
35  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
36  * the permission of UNIX System Laboratories, Inc.
37  *
38  * Redistribution and use in source and binary forms, with or without
39  * modification, are permitted provided that the following conditions
40  * are met:
41  * 1. Redistributions of source code must retain the above copyright
42  *    notice, this list of conditions and the following disclaimer.
43  * 2. Redistributions in binary form must reproduce the above copyright
44  *    notice, this list of conditions and the following disclaimer in the
45  *    documentation and/or other materials provided with the distribution.
46  * 3. Neither the name of the University nor the names of its contributors
47  *    may be used to endorse or promote products derived from this software
48  *    without specific prior written permission.
49  *
50  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
51  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
52  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
53  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
54  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
55  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
56  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
57  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
58  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
59  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
60  * SUCH DAMAGE.
61  *
62  *	@(#)init_main.c	8.16 (Berkeley) 5/14/95
63  */
64 
65 /*
66  * Copyright (c) 1995 Christopher G. Demetriou.  All rights reserved.
67  *
68  * Redistribution and use in source and binary forms, with or without
69  * modification, are permitted provided that the following conditions
70  * are met:
71  * 1. Redistributions of source code must retain the above copyright
72  *    notice, this list of conditions and the following disclaimer.
73  * 2. Redistributions in binary form must reproduce the above copyright
74  *    notice, this list of conditions and the following disclaimer in the
75  *    documentation and/or other materials provided with the distribution.
76  * 3. All advertising materials mentioning features or use of this software
77  *    must display the following acknowledgement:
78  *	This product includes software developed by the University of
79  *	California, Berkeley and its contributors.
80  * 4. Neither the name of the University nor the names of its contributors
81  *    may be used to endorse or promote products derived from this software
82  *    without specific prior written permission.
83  *
84  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
85  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
86  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
87  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
88  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
89  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
90  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
91  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
92  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
93  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
94  * SUCH DAMAGE.
95  *
96  *	@(#)init_main.c	8.16 (Berkeley) 5/14/95
97  */
98 
99 #include <sys/cdefs.h>
100 __KERNEL_RCSID(0, "$NetBSD: init_main.c,v 1.467 2015/05/06 15:57:08 hannken Exp $");
101 
102 #include "opt_ddb.h"
103 #include "opt_ipsec.h"
104 #include "opt_modular.h"
105 #include "opt_ntp.h"
106 #include "opt_pipe.h"
107 #include "opt_syscall_debug.h"
108 #include "opt_sysv.h"
109 #include "opt_fileassoc.h"
110 #include "opt_ktrace.h"
111 #include "opt_pax.h"
112 #include "opt_compat_netbsd.h"
113 #include "opt_wapbl.h"
114 #include "opt_ptrace.h"
115 #include "opt_rnd_printf.h"
116 #include "opt_splash.h"
117 
118 #if defined(SPLASHSCREEN) && defined(SPLASHSCREEN_IMAGE)
119 extern void *_binary_splash_image_start;
120 extern void *_binary_splash_image_end;
121 #endif
122 
123 #include "drvctl.h"
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 #ifdef SYSVSHM
182 #include <sys/shm.h>
183 #endif
184 #ifdef SYSVSEM
185 #include <sys/sem.h>
186 #endif
187 #ifdef SYSVMSG
188 #include <sys/msg.h>
189 #endif
190 #include <sys/domain.h>
191 #include <sys/namei.h>
192 #include <sys/rnd.h>
193 #include <sys/pipe.h>
194 #if NVERIEXEC > 0
195 #include <sys/verified_exec.h>
196 #endif /* NVERIEXEC > 0 */
197 #ifdef KTRACE
198 #include <sys/ktrace.h>
199 #endif
200 #include <sys/kauth.h>
201 #include <net80211/ieee80211_netbsd.h>
202 #ifdef PTRACE
203 #include <sys/ptrace.h>
204 #endif /* PTRACE */
205 #include <sys/cprng.h>
206 
207 #include <sys/syscall.h>
208 #include <sys/syscallargs.h>
209 
210 #if defined(PAX_MPROTECT) || defined(PAX_SEGVGUARD) || defined(PAX_ASLR)
211 #include <sys/pax.h>
212 #endif /* PAX_MPROTECT || PAX_SEGVGUARD || PAX_ASLR */
213 
214 #include <secmodel/secmodel.h>
215 
216 #include <ufs/ufs/quota.h>
217 
218 #include <miscfs/genfs/genfs.h>
219 #include <miscfs/specfs/specdev.h>
220 
221 #include <sys/cpu.h>
222 
223 #include <uvm/uvm.h>	/* extern struct uvm uvm */
224 
225 #include <dev/cons.h>
226 #include <dev/splash/splash.h>
227 
228 #include <net/bpf.h>
229 #include <net/if.h>
230 #include <net/raw_cb.h>
231 
232 #include <prop/proplib.h>
233 
234 #include <sys/userconf.h>
235 
236 extern struct lwp lwp0;
237 extern time_t rootfstime;
238 
239 #ifndef curlwp
240 struct	lwp *curlwp = &lwp0;
241 #endif
242 struct	proc *initproc;
243 
244 struct	vnode *rootvp, *swapdev_vp;
245 int	boothowto;
246 int	cold = 1;			/* still working on startup */
247 struct timespec boottime;	        /* time at system startup - will only follow settime deltas */
248 
249 int	start_init_exec;		/* semaphore for start_init() */
250 
251 cprng_strong_t	*kern_cprng;
252 
253 static void check_console(struct lwp *l);
254 static void start_init(void *);
255 static void configure(void);
256 static void configure2(void);
257 static void configure3(void);
258 void main(void);
259 
260 /*
261  * System startup; initialize the world, create process 0, mount root
262  * filesystem, and fork to create init and pagedaemon.  Most of the
263  * hard work is done in the lower-level initialization routines including
264  * startup(), which does memory initialization and autoconfiguration.
265  */
266 void
267 main(void)
268 {
269 	struct timespec time;
270 	struct lwp *l;
271 	struct proc *p;
272 	int s, error;
273 #ifdef NVNODE_IMPLICIT
274 	int usevnodes;
275 #endif
276 	CPU_INFO_ITERATOR cii;
277 	struct cpu_info *ci;
278 
279 	l = &lwp0;
280 #ifndef LWP0_CPU_INFO
281 	l->l_cpu = curcpu();
282 #endif
283 	l->l_pflag |= LP_RUNNING;
284 
285 	/*
286 	 * Attempt to find console and initialize
287 	 * in case of early panic or other messages.
288 	 */
289 	consinit();
290 
291 	kernel_lock_init();
292 	once_init();
293 
294 	mi_cpu_init();
295 	kernconfig_lock_init();
296 	kthread_sysinit();
297 
298 	/* Initialize the device switch tables. */
299 	devsw_init();
300 
301 	/* Initialize event counters. */
302 	evcnt_init();
303 
304 	uvm_init();
305 	ubchist_init();
306 	kcpuset_sysinit();
307 
308 	prop_kern_init();
309 
310 #if ((NKSYMS > 0) || (NDDB > 0) || (NMODULAR > 0))
311 	ksyms_init();
312 #endif
313 	kprintf_init();
314 
315 	percpu_init();
316 
317 	/* Initialize lock caches. */
318 	mutex_obj_init();
319 	rw_obj_init();
320 
321 	/* Passive serialization. */
322 	pserialize_init();
323 
324 	/* Initialize the extent manager. */
325 	extent_init();
326 
327 	/* Do machine-dependent initialization. */
328 	cpu_startup();
329 
330 	/* Initialize the sysctl subsystem. */
331 	sysctl_init();
332 
333 	/* Initialize callouts, part 1. */
334 	callout_startup();
335 
336 	/* Initialize the kernel authorization subsystem. */
337 	kauth_init();
338 
339 	secmodel_init();
340 
341 	spec_init();
342 
343 	/*
344 	 * Set BPF op vector.  Can't do this in bpf attach, since
345 	 * network drivers attach before bpf.
346 	 */
347 	bpf_setops();
348 
349 	/* Start module system. */
350 	module_init();
351 
352 	/*
353 	 * Initialize the kernel authorization subsystem and start the
354 	 * default security model, if any. We need to do this early
355 	 * enough so that subsystems relying on any of the aforementioned
356 	 * can work properly. Since the security model may dictate the
357 	 * credential inheritance policy, it is needed at least before
358 	 * any process is created, specifically proc0.
359 	 */
360 	module_init_class(MODULE_CLASS_SECMODEL);
361 
362 	/* Initialize the buffer cache */
363 	bufinit();
364 
365 
366 #if defined(SPLASHSCREEN) && defined(SPLASHSCREEN_IMAGE)
367 	size_t splash_size = (&_binary_splash_image_end -
368 	    &_binary_splash_image_start) * sizeof(void *);
369 	splash_setimage(&_binary_splash_image_start, splash_size);
370 #endif
371 
372 	/* Initialize sockets. */
373 	soinit();
374 
375 	/*
376 	 * The following things must be done before autoconfiguration.
377 	 */
378 	rnd_init();		/* initialize entropy pool */
379 
380 	cprng_init();		/* initialize cryptographic PRNG */
381 
382 	/* Initialize process and pgrp structures. */
383 	procinit();
384 	lwpinit();
385 
386 	/* Initialize signal-related data structures. */
387 	signal_init();
388 
389 	/* Initialize resource management. */
390 	resource_init();
391 
392 	/* Create process 0. */
393 	proc0_init();
394 	lwp0_init();
395 
396 	/* Disable preemption during boot. */
397 	kpreempt_disable();
398 
399 	/* Initialize the UID hash table. */
400 	uid_init();
401 
402 	/* Charge root for one process. */
403 	(void)chgproccnt(0, 1);
404 
405 	/* Initialize timekeeping. */
406 	time_init();
407 
408 	/* Initialize the run queues, turnstiles and sleep queues. */
409 	sched_rqinit();
410 	turnstile_init();
411 	sleeptab_init(&sleeptab);
412 
413 	sched_init();
414 
415 	/* Initialize processor-sets */
416 	psets_init();
417 
418 	/* Initialize cpufreq(9) */
419 	cpufreq_init();
420 
421 	/* MI initialization of the boot cpu */
422 	error = mi_cpu_attach(curcpu());
423 	KASSERT(error == 0);
424 
425 	/* Initialize timekeeping, part 2. */
426 	time_init2();
427 
428 	/*
429 	 * Initialize mbuf's.  Do this now because we might attempt to
430 	 * allocate mbufs or mbuf clusters during autoconfiguration.
431 	 */
432 	mbinit();
433 
434 	/* Initialize I/O statistics. */
435 	iostat_init();
436 
437 	/* Initialize the log device. */
438 	loginit();
439 
440 	/* Second part of module system initialization. */
441 	module_start_unload_thread();
442 
443 	/* Initialize the file systems. */
444 #ifdef NVNODE_IMPLICIT
445 	/*
446 	 * If maximum number of vnodes in namei vnode cache is not explicitly
447 	 * defined in kernel config, adjust the number such as we use roughly
448 	 * 10% of memory for vnodes and associated data structures in the
449 	 * assumed worst case.  Do not provide fewer than NVNODE vnodes.
450 	 */
451 	usevnodes = calc_cache_size(vmem_size(kmem_arena, VMEM_FREE|VMEM_ALLOC),
452 	    10, VNODE_KMEM_MAXPCT) / VNODE_COST;
453 	if (usevnodes > desiredvnodes)
454 		desiredvnodes = usevnodes;
455 #endif
456 	vfsinit();
457 	lf_init();
458 
459 	/* Initialize fstrans. */
460 	fstrans_init();
461 
462 	/* Initialize the file descriptor system. */
463 	fd_sys_init();
464 
465 	/* Initialize cwd structures */
466 	cwd_sys_init();
467 
468 	/* Initialize kqueue. */
469 	kqueue_init();
470 
471 	inittimecounter();
472 	ntp_init();
473 
474 	/* Initialize tty subsystem. */
475 	tty_init();
476 	ttyldisc_init();
477 
478 	/* Initialize the buffer cache, part 2. */
479 	bufinit2();
480 
481 	/* Initialize the disk wedge subsystem. */
482 	dkwedge_init();
483 
484 	/* Initialize the kernel strong PRNG. */
485 	kern_cprng = cprng_strong_create("kernel", IPL_VM,
486 					 CPRNG_INIT_ANY|CPRNG_REKEY_ANY);
487 
488 	/* Initialize interfaces. */
489 	ifinit1();
490 
491 	spldebug_start();
492 
493 	/* Initialize sockets thread(s) */
494 	soinit1();
495 
496 	/* Configure the system hardware.  This will enable interrupts. */
497 	configure();
498 
499 	/* Once all CPUs are detected, initialize the per-CPU cprng_fast.  */
500 	cprng_fast_init();
501 
502 	ssp_init();
503 
504 	ubc_init();		/* must be after autoconfig */
505 
506 	mm_init();
507 
508 	configure2();
509 
510 	ipi_sysinit();
511 
512 	/* Now timer is working.  Enable preemption. */
513 	kpreempt_enable();
514 
515 	/* Get the threads going and into any sleeps before continuing. */
516 	yield();
517 
518 	/* Enable deferred processing of RNG samples */
519 	rnd_init_softint();
520 
521 #ifdef RND_PRINTF
522 	/* Enable periodic injection of console output into entropy pool */
523 	kprintf_init_callout();
524 #endif
525 
526 #ifdef SYSVSHM
527 	/* Initialize System V style shared memory. */
528 	shminit();
529 #endif
530 
531 	vmem_rehash_start();	/* must be before exec_init */
532 
533 	/* Initialize exec structures */
534 	exec_init(1);		/* seminit calls exithook_establish() */
535 
536 #ifdef SYSVSEM
537 	/* Initialize System V style semaphores. */
538 	seminit();
539 #endif
540 
541 #ifdef SYSVMSG
542 	/* Initialize System V style message queues. */
543 	msginit();
544 #endif
545 
546 #if NVERIEXEC > 0
547 	/*
548 	 * Initialise the Veriexec subsystem.
549 	 */
550 	veriexec_init();
551 #endif /* NVERIEXEC > 0 */
552 
553 #if defined(PAX_MPROTECT) || defined(PAX_SEGVGUARD) || defined(PAX_ASLR)
554 	pax_init();
555 #endif /* PAX_MPROTECT || PAX_SEGVGUARD || PAX_ASLR */
556 
557 #ifdef	IPSEC
558 	/* Attach network crypto subsystem */
559 	ipsec_attach();
560 #endif
561 
562 	/*
563 	 * Initialize protocols.  Block reception of incoming packets
564 	 * until everything is ready.
565 	 */
566 	s = splnet();
567 	ifinit();
568 	domaininit(true);
569 	if_attachdomain();
570 	splx(s);
571 
572 #ifdef GPROF
573 	/* Initialize kernel profiling. */
574 	kmstartup();
575 #endif
576 
577 	/* Initialize system accounting. */
578 	acct_init();
579 
580 #ifndef PIPE_SOCKETPAIR
581 	/* Initialize pipes. */
582 	pipe_init();
583 #endif
584 
585 #ifdef KTRACE
586 	/* Initialize ktrace. */
587 	ktrinit();
588 #endif
589 
590 #ifdef PTRACE
591 	/* Initialize ptrace. */
592 	ptrace_init();
593 #endif /* PTRACE */
594 
595 	machdep_init();
596 
597 	procinit_sysctl();
598 
599 	/*
600 	 * Create process 1 (init(8)).  We do this now, as Unix has
601 	 * historically had init be process 1, and changing this would
602 	 * probably upset a lot of people.
603 	 *
604 	 * Note that process 1 won't immediately exec init(8), but will
605 	 * wait for us to inform it that the root file system has been
606 	 * mounted.
607 	 */
608 	if (fork1(l, 0, SIGCHLD, NULL, 0, start_init, NULL, NULL, &initproc))
609 		panic("fork init");
610 
611 	/*
612 	 * Load any remaining builtin modules, and hand back temporary
613 	 * storage to the VM system.  Then require force when loading any
614 	 * remaining un-init'ed built-in modules to avoid later surprises.
615 	 */
616 	module_init_class(MODULE_CLASS_ANY);
617 	module_builtin_require_force();
618 
619 	/*
620 	 * Finalize configuration now that all real devices have been
621 	 * found.  This needs to be done before the root device is
622 	 * selected, since finalization may create the root device.
623 	 */
624 	config_finalize();
625 
626 	sysctl_finalize();
627 
628 	/*
629 	 * Now that autoconfiguration has completed, we can determine
630 	 * the root and dump devices.
631 	 */
632 	cpu_rootconf();
633 	cpu_dumpconf();
634 
635 	/* Mount the root file system. */
636 	do {
637 		domountroothook(root_device);
638 		if ((error = vfs_mountroot())) {
639 			printf("cannot mount root, error = %d\n", error);
640 			boothowto |= RB_ASKNAME;
641 			setroot(root_device,
642 			    (rootdev != NODEV) ? DISKPART(rootdev) : 0);
643 		}
644 	} while (error != 0);
645 	mountroothook_destroy();
646 
647 	configure3();
648 
649 	/*
650 	 * Initialise the time-of-day clock, passing the time recorded
651 	 * in the root filesystem (if any) for use by systems that
652 	 * don't have a non-volatile time-of-day device.
653 	 */
654 	inittodr(rootfstime);
655 
656 	/*
657 	 * Now can look at time, having had a chance to verify the time
658 	 * from the file system.  Reset l->l_rtime as it may have been
659 	 * munched in mi_switch() after the time got set.
660 	 */
661 	getnanotime(&time);
662 	boottime = time;
663 
664 	mutex_enter(proc_lock);
665 	LIST_FOREACH(p, &allproc, p_list) {
666 		KASSERT((p->p_flag & PK_MARKER) == 0);
667 		mutex_enter(p->p_lock);
668 		TIMESPEC_TO_TIMEVAL(&p->p_stats->p_start, &time);
669 		LIST_FOREACH(l, &p->p_lwps, l_sibling) {
670 			lwp_lock(l);
671 			memset(&l->l_rtime, 0, sizeof(l->l_rtime));
672 			lwp_unlock(l);
673 		}
674 		mutex_exit(p->p_lock);
675 	}
676 	mutex_exit(proc_lock);
677 	binuptime(&curlwp->l_stime);
678 
679 	for (CPU_INFO_FOREACH(cii, ci)) {
680 		ci->ci_schedstate.spc_lastmod = time_second;
681 	}
682 
683 	/* Create the pageout daemon kernel thread. */
684 	uvm_swap_init();
685 	if (kthread_create(PRI_PGDAEMON, KTHREAD_MPSAFE, NULL, uvm_pageout,
686 	    NULL, NULL, "pgdaemon"))
687 		panic("fork pagedaemon");
688 
689 	/* Create the filesystem syncer kernel thread. */
690 	if (kthread_create(PRI_IOFLUSH, KTHREAD_MPSAFE, NULL, sched_sync,
691 	    NULL, NULL, "ioflush"))
692 		panic("fork syncer");
693 
694 	/* Create the aiodone daemon kernel thread. */
695 	if (workqueue_create(&uvm.aiodone_queue, "aiodoned",
696 	    uvm_aiodone_worker, NULL, PRI_VM, IPL_NONE, WQ_MPSAFE))
697 		panic("fork aiodoned");
698 
699 	/* Wait for final configure threads to complete. */
700 	config_finalize_mountroot();
701 
702 	/*
703 	 * Okay, now we can let init(8) exec!  It's off to userland!
704 	 */
705 	mutex_enter(proc_lock);
706 	start_init_exec = 1;
707 	cv_broadcast(&lbolt);
708 	mutex_exit(proc_lock);
709 
710 	/* The scheduler is an infinite loop. */
711 	uvm_scheduler();
712 	/* NOTREACHED */
713 }
714 
715 /*
716  * Configure the system's hardware.
717  */
718 static void
719 configure(void)
720 {
721 
722 	/* Initialize autoconf data structures. */
723 	config_init_mi();
724 	/*
725 	 * XXX
726 	 * callout_setfunc() requires mutex(9) so it can't be in config_init()
727 	 * on amiga and atari which use config_init() and autoconf(9) fucntions
728 	 * to initialize console devices.
729 	 */
730 	config_twiddle_init();
731 
732 	pmf_init();
733 #if NDRVCTL > 0
734 	drvctl_init();
735 #endif
736 
737 	userconf_init();
738 	if (boothowto & RB_USERCONF)
739 		userconf_prompt();
740 
741 	if ((boothowto & (AB_SILENT|AB_VERBOSE)) == AB_SILENT) {
742 		printf_nolog("Detecting hardware...");
743 	}
744 
745 	/*
746 	 * Do the machine-dependent portion of autoconfiguration.  This
747 	 * sets the configuration machinery here in motion by "finding"
748 	 * the root bus.  When this function returns, we expect interrupts
749 	 * to be enabled.
750 	 */
751 	cpu_configure();
752 }
753 
754 static void
755 configure2(void)
756 {
757 	CPU_INFO_ITERATOR cii;
758 	struct cpu_info *ci;
759 	int s;
760 
761 	/*
762 	 * Now that we've found all the hardware, start the real time
763 	 * and statistics clocks.
764 	 */
765 	initclocks();
766 
767 	cold = 0;	/* clocks are running, we're warm now! */
768 	s = splsched();
769 	curcpu()->ci_schedstate.spc_flags |= SPCF_RUNNING;
770 	splx(s);
771 
772 	/* Boot the secondary processors. */
773 	for (CPU_INFO_FOREACH(cii, ci)) {
774 		uvm_cpu_attach(ci);
775 	}
776 	mp_online = true;
777 #if defined(MULTIPROCESSOR)
778 	cpu_boot_secondary_processors();
779 #endif
780 
781 	/* Setup the runqueues and scheduler. */
782 	runq_init();
783 	synch_init();
784 
785 	/*
786 	 * Bus scans can make it appear as if the system has paused, so
787 	 * twiddle constantly while config_interrupts() jobs are running.
788 	 */
789 	config_twiddle_fn(NULL);
790 
791 	/*
792 	 * Create threads to call back and finish configuration for
793 	 * devices that want interrupts enabled.
794 	 */
795 	config_create_interruptthreads();
796 }
797 
798 static void
799 configure3(void)
800 {
801 
802 	/*
803 	 * Create threads to call back and finish configuration for
804 	 * devices that want the mounted root file system.
805 	 */
806 	config_create_mountrootthreads();
807 
808 	/* Get the threads going and into any sleeps before continuing. */
809 	yield();
810 }
811 
812 static void
813 rootconf_handle_wedges(void)
814 {
815 	struct partinfo dpart;
816 	struct partition *p;
817 	struct vnode *vp;
818 	daddr_t startblk;
819 	uint64_t nblks;
820 	device_t dev;
821 	int error;
822 
823 	if (booted_nblks) {
824 		/*
825 		 * bootloader passed geometry
826 		 */
827 		dev      = booted_device;
828 		startblk = booted_startblk;
829 		nblks    = booted_nblks;
830 
831 		/*
832 		 * keep booted_device and booted_partition
833 		 * in case the kernel doesn't identify a wedge
834 		 */
835 	} else {
836 		/*
837 		 * bootloader passed partition number
838 		 *
839 		 * We cannot ask the partition device directly when it is
840 		 * covered by a wedge. Instead we look up the geometry in
841 		 * the disklabel.
842 		 */
843 		vp = opendisk(booted_device);
844 
845 		if (vp == NULL)
846 			return;
847 
848 		error = VOP_IOCTL(vp, DIOCGPART, &dpart, FREAD, NOCRED);
849 		VOP_CLOSE(vp, FREAD, NOCRED);
850 		vput(vp);
851 		if (error)
852 			return;
853 
854 		KASSERT(booted_partition >= 0
855 			&& booted_partition < MAXPARTITIONS);
856 
857 		p = &dpart.disklab->d_partitions[booted_partition];
858 
859 		dev      = booted_device;
860 		startblk = p->p_offset;
861 		nblks    = p->p_size;
862 	}
863 
864 	dev = dkwedge_find_partition(dev, startblk, nblks);
865 	if (dev != NULL) {
866 		booted_device = dev;
867 		booted_partition = 0;
868 	}
869 }
870 
871 void
872 rootconf(void)
873 {
874 	if (booted_device != NULL)
875 		rootconf_handle_wedges();
876 
877 	setroot(booted_device, booted_partition);
878 }
879 
880 static void
881 check_console(struct lwp *l)
882 {
883 	struct vnode *vp;
884 	int error;
885 
886 	error = namei_simple_kernel("/dev/console",
887 				NSM_FOLLOW_NOEMULROOT, &vp);
888 	if (error == 0)
889 		vrele(vp);
890 	else if (error == ENOENT)
891 		printf("warning: no /dev/console\n");
892 	else
893 		printf("warning: lookup /dev/console: error %d\n", error);
894 }
895 
896 /*
897  * List of paths to try when searching for "init".
898  */
899 static const char * const initpaths[] = {
900 	"/sbin/init",
901 	"/sbin/oinit",
902 	"/sbin/init.bak",
903 	"/rescue/init",
904 	NULL,
905 };
906 
907 /*
908  * Start the initial user process; try exec'ing each pathname in "initpaths".
909  * The program is invoked with one argument containing the boot flags.
910  */
911 static void
912 start_init(void *arg)
913 {
914 	struct lwp *l = arg;
915 	struct proc *p = l->l_proc;
916 	vaddr_t addr;
917 	struct sys_execve_args /* {
918 		syscallarg(const char *) path;
919 		syscallarg(char * const *) argp;
920 		syscallarg(char * const *) envp;
921 	} */ args;
922 	int options, i, error;
923 	register_t retval[2];
924 	char flags[4], *flagsp;
925 	const char *path, *slash;
926 	char *ucp, **uap, *arg0, *arg1 = NULL;
927 	char ipath[129];
928 	int ipx, len;
929 
930 	/*
931 	 * Now in process 1.
932 	 */
933 	strncpy(p->p_comm, "init", MAXCOMLEN);
934 
935 	/*
936 	 * Wait for main() to tell us that it's safe to exec.
937 	 */
938 	mutex_enter(proc_lock);
939 	while (start_init_exec == 0)
940 		cv_wait(&lbolt, proc_lock);
941 	mutex_exit(proc_lock);
942 
943 	/*
944 	 * This is not the right way to do this.  We really should
945 	 * hand-craft a descriptor onto /dev/console to hand to init,
946 	 * but that's a _lot_ more work, and the benefit from this easy
947 	 * hack makes up for the "good is the enemy of the best" effect.
948 	 */
949 	check_console(l);
950 
951 	/*
952 	 * Need just enough stack to hold the faked-up "execve()" arguments.
953 	 */
954 	addr = (vaddr_t)STACK_ALLOC(USRSTACK, PAGE_SIZE);
955 	if (uvm_map(&p->p_vmspace->vm_map, &addr, PAGE_SIZE,
956                     NULL, UVM_UNKNOWN_OFFSET, 0,
957                     UVM_MAPFLAG(UVM_PROT_ALL, UVM_PROT_ALL, UVM_INH_COPY,
958 		    UVM_ADV_NORMAL,
959                     UVM_FLAG_FIXED|UVM_FLAG_OVERLAY|UVM_FLAG_COPYONW)) != 0)
960 		panic("init: couldn't allocate argument space");
961 	p->p_vmspace->vm_maxsaddr = (void *)STACK_MAX(addr, PAGE_SIZE);
962 
963 	ipx = 0;
964 	while (1) {
965 		if (boothowto & RB_ASKNAME) {
966 			printf("init path");
967 			if (initpaths[ipx])
968 				printf(" (default %s)", initpaths[ipx]);
969 			printf(": ");
970 			len = cngetsn(ipath, sizeof(ipath)-1);
971 			if (len == 4 && strcmp(ipath, "halt") == 0) {
972 				cpu_reboot(RB_HALT, NULL);
973 			} else if (len == 6 && strcmp(ipath, "reboot") == 0) {
974 				cpu_reboot(0, NULL);
975 #if defined(DDB)
976 			} else if (len == 3 && strcmp(ipath, "ddb") == 0) {
977 				console_debugger();
978 				continue;
979 #endif
980 			} else if (len > 0 && ipath[0] == '/') {
981 				ipath[len] = '\0';
982 				path = ipath;
983 			} else if (len == 0 && initpaths[ipx] != NULL) {
984 				path = initpaths[ipx++];
985 			} else {
986 				printf("use absolute path, ");
987 #if defined(DDB)
988 				printf("\"ddb\", ");
989 #endif
990 				printf("\"halt\", or \"reboot\"\n");
991 				continue;
992 			}
993 		} else {
994 			if ((path = initpaths[ipx++]) == NULL) {
995 				ipx = 0;
996 				boothowto |= RB_ASKNAME;
997 				continue;
998 			}
999 		}
1000 
1001 		ucp = (char *)USRSTACK;
1002 
1003 		/*
1004 		 * Construct the boot flag argument.
1005 		 */
1006 		flagsp = flags;
1007 		*flagsp++ = '-';
1008 		options = 0;
1009 
1010 		if (boothowto & RB_SINGLE) {
1011 			*flagsp++ = 's';
1012 			options = 1;
1013 		}
1014 #ifdef notyet
1015 		if (boothowto & RB_FASTBOOT) {
1016 			*flagsp++ = 'f';
1017 			options = 1;
1018 		}
1019 #endif
1020 
1021 		/*
1022 		 * Move out the flags (arg 1), if necessary.
1023 		 */
1024 		if (options != 0) {
1025 			*flagsp++ = '\0';
1026 			i = flagsp - flags;
1027 #ifdef DEBUG
1028 			aprint_normal("init: copying out flags `%s' %d\n", flags, i);
1029 #endif
1030 			arg1 = STACK_ALLOC(ucp, i);
1031 			ucp = STACK_MAX(arg1, i);
1032 			(void)copyout((void *)flags, arg1, i);
1033 		}
1034 
1035 		/*
1036 		 * Move out the file name (also arg 0).
1037 		 */
1038 		i = strlen(path) + 1;
1039 #ifdef DEBUG
1040 		aprint_normal("init: copying out path `%s' %d\n", path, i);
1041 #else
1042 		if (boothowto & RB_ASKNAME || path != initpaths[0])
1043 			printf("init: trying %s\n", path);
1044 #endif
1045 		arg0 = STACK_ALLOC(ucp, i);
1046 		ucp = STACK_MAX(arg0, i);
1047 		(void)copyout(path, arg0, i);
1048 
1049 		/*
1050 		 * Move out the arg pointers.
1051 		 */
1052 		ucp = (void *)STACK_ALIGN(ucp, STACK_ALIGNBYTES);
1053 		uap = (char **)STACK_ALLOC(ucp, sizeof(char *) * 3);
1054 		SCARG(&args, path) = arg0;
1055 		SCARG(&args, argp) = uap;
1056 		SCARG(&args, envp) = NULL;
1057 		slash = strrchr(path, '/');
1058 		if (slash)
1059 			(void)suword((void *)uap++,
1060 			    (long)arg0 + (slash + 1 - path));
1061 		else
1062 			(void)suword((void *)uap++, (long)arg0);
1063 		if (options != 0)
1064 			(void)suword((void *)uap++, (long)arg1);
1065 		(void)suword((void *)uap++, 0);	/* terminator */
1066 
1067 		/*
1068 		 * Now try to exec the program.  If can't for any reason
1069 		 * other than it doesn't exist, complain.
1070 		 */
1071 		error = sys_execve(l, &args, retval);
1072 		if (error == 0 || error == EJUSTRETURN) {
1073 			KERNEL_UNLOCK_LAST(l);
1074 			return;
1075 		}
1076 		printf("exec %s: error %d\n", path, error);
1077 	}
1078 	printf("init: not found\n");
1079 	panic("no init");
1080 }
1081 
1082 /*
1083  * calculate cache size (in bytes) from physmem and vsize.
1084  */
1085 vaddr_t
1086 calc_cache_size(vsize_t vsize, int pct, int va_pct)
1087 {
1088 	paddr_t t;
1089 
1090 	/* XXX should consider competing cache if any */
1091 	/* XXX should consider submaps */
1092 	t = (uintmax_t)physmem * pct / 100 * PAGE_SIZE;
1093 	if (vsize != 0) {
1094 		vsize = (uintmax_t)vsize * va_pct / 100;
1095 		if (t > vsize) {
1096 			t = vsize;
1097 		}
1098 	}
1099 	return t;
1100 }
1101 
1102 /*
1103  * Print the system start up banner.
1104  *
1105  * - Print a limited banner if AB_SILENT.
1106  * - Always send normal banner to the log.
1107  */
1108 #define MEM_PBUFSIZE	sizeof("99999 MB")
1109 
1110 void
1111 banner(void)
1112 {
1113 	static char notice[] = " Notice: this software is "
1114 	    "protected by copyright";
1115 	char pbuf[81];
1116 	void (*pr)(const char *, ...) __printflike(1, 2);
1117 	int i;
1118 
1119 	if ((boothowto & AB_SILENT) != 0) {
1120 		snprintf(pbuf, sizeof(pbuf), "%s %s (%s)",
1121 		    ostype, osrelease, kernel_ident);
1122 		printf_nolog("%s", pbuf);
1123 		for (i = 80 - strlen(pbuf) - sizeof(notice); i > 0; i--)
1124 			printf(" ");
1125 		printf_nolog("%s\n", notice);
1126 		pr = aprint_normal;
1127 	} else {
1128 		pr = printf;
1129 	}
1130 
1131 	memset(pbuf, 0, sizeof(pbuf));
1132 	(*pr)("%s%s", copyright, version);
1133 	format_bytes(pbuf, MEM_PBUFSIZE, ctob((uint64_t)physmem));
1134 	(*pr)("total memory = %s\n", pbuf);
1135 	format_bytes(pbuf, MEM_PBUFSIZE, ctob((uint64_t)uvmexp.free));
1136 	(*pr)("avail memory = %s\n", pbuf);
1137 }
1138