xref: /netbsd-src/sys/kern/init_main.c (revision da9817918ec7e88db2912a2882967c7570a83f47)
1 /*	$NetBSD: init_main.c,v 1.390 2009/05/27 23:44:36 pooka 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 1.390 2009/05/27 23:44:36 pooka 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 
115 #include "ksyms.h"
116 #include "rnd.h"
117 #include "sysmon_envsys.h"
118 #include "sysmon_power.h"
119 #include "sysmon_taskq.h"
120 #include "sysmon_wdog.h"
121 #include "veriexec.h"
122 
123 #include <sys/param.h>
124 #include <sys/acct.h>
125 #include <sys/filedesc.h>
126 #include <sys/file.h>
127 #include <sys/errno.h>
128 #include <sys/callout.h>
129 #include <sys/cpu.h>
130 #include <sys/kernel.h>
131 #include <sys/mount.h>
132 #include <sys/proc.h>
133 #include <sys/kthread.h>
134 #include <sys/resourcevar.h>
135 #include <sys/signalvar.h>
136 #include <sys/systm.h>
137 #include <sys/vnode.h>
138 #include <sys/fstrans.h>
139 #include <sys/tty.h>
140 #include <sys/conf.h>
141 #include <sys/disklabel.h>
142 #include <sys/buf.h>
143 #include <sys/device.h>
144 #include <sys/exec.h>
145 #include <sys/socketvar.h>
146 #include <sys/protosw.h>
147 #include <sys/percpu.h>
148 #include <sys/pset.h>
149 #include <sys/sysctl.h>
150 #include <sys/reboot.h>
151 #include <sys/user.h>
152 #include <sys/sysctl.h>
153 #include <sys/event.h>
154 #include <sys/mbuf.h>
155 #include <sys/sched.h>
156 #include <sys/sleepq.h>
157 #include <sys/iostat.h>
158 #include <sys/vmem.h>
159 #include <sys/uuid.h>
160 #include <sys/extent.h>
161 #include <sys/disk.h>
162 #include <sys/mqueue.h>
163 #include <sys/msgbuf.h>
164 #include <sys/module.h>
165 #include <sys/event.h>
166 #include <sys/lockf.h>
167 #include <sys/once.h>
168 #include <sys/ksyms.h>
169 #include <sys/uidinfo.h>
170 #include <sys/kprintf.h>
171 #ifdef FAST_IPSEC
172 #include <netipsec/ipsec.h>
173 #endif
174 #ifdef SYSVSHM
175 #include <sys/shm.h>
176 #endif
177 #ifdef SYSVSEM
178 #include <sys/sem.h>
179 #endif
180 #ifdef SYSVMSG
181 #include <sys/msg.h>
182 #endif
183 #include <sys/domain.h>
184 #include <sys/namei.h>
185 #if NRND > 0
186 #include <sys/rnd.h>
187 #endif
188 #include <sys/pipe.h>
189 #if NVERIEXEC > 0
190 #include <sys/verified_exec.h>
191 #endif /* NVERIEXEC > 0 */
192 #ifdef KTRACE
193 #include <sys/ktrace.h>
194 #endif
195 #include <sys/kauth.h>
196 #ifdef WAPBL
197 #include <sys/wapbl.h>
198 #endif
199 #include <net80211/ieee80211_netbsd.h>
200 
201 #include <sys/syscall.h>
202 #include <sys/syscallargs.h>
203 
204 #if defined(PAX_MPROTECT) || defined(PAX_SEGVGUARD) || defined(PAX_ASLR)
205 #include <sys/pax.h>
206 #endif /* PAX_MPROTECT || PAX_SEGVGUARD || PAX_ASLR */
207 
208 #include <ufs/ufs/quota.h>
209 
210 #include <miscfs/genfs/genfs.h>
211 #include <miscfs/syncfs/syncfs.h>
212 
213 #include <sys/cpu.h>
214 
215 #include <uvm/uvm.h>
216 
217 #if NSYSMON_TASKQ > 0
218 #include <dev/sysmon/sysmon_taskq.h>
219 #endif
220 
221 #include <dev/cons.h>
222 
223 #if NSYSMON_ENVSYS > 0 || NSYSMON_POWER > 0 || NSYSMON_WDOG > 0
224 #include <dev/sysmon/sysmonvar.h>
225 #endif
226 
227 #include <net/if.h>
228 #include <net/raw_cb.h>
229 
230 #include <secmodel/secmodel.h>
231 
232 #ifdef COMPAT_50
233 #include <compat/sys/time.h>
234 struct timeval50 boottime50;
235 #endif
236 
237 extern struct proc proc0;
238 extern struct lwp lwp0;
239 extern struct cwdinfo cwdi0;
240 extern time_t rootfstime;
241 
242 #ifndef curlwp
243 struct	lwp *curlwp = &lwp0;
244 #endif
245 struct	proc *initproc;
246 
247 struct	vnode *rootvp, *swapdev_vp;
248 int	boothowto;
249 int	cold = 1;			/* still working on startup */
250 struct timespec boottime;	        /* time at system startup - will only follow settime deltas */
251 
252 int	start_init_exec;		/* semaphore for start_init() */
253 
254 static void check_console(struct lwp *l);
255 static void start_init(void *);
256 void main(void);
257 
258 void __secmodel_none(void);
259 __weak_alias(secmodel_start,__secmodel_none);
260 void
261 __secmodel_none(void)
262 {
263 	return;
264 }
265 
266 /*
267  * System startup; initialize the world, create process 0, mount root
268  * filesystem, and fork to create init and pagedaemon.  Most of the
269  * hard work is done in the lower-level initialization routines including
270  * startup(), which does memory initialization and autoconfiguration.
271  */
272 void
273 main(void)
274 {
275 	struct timespec time;
276 	struct lwp *l;
277 	struct proc *p;
278 	int s, error;
279 #ifdef NVNODE_IMPLICIT
280 	int usevnodes;
281 #endif
282 	CPU_INFO_ITERATOR cii;
283 	struct cpu_info *ci;
284 
285 	l = &lwp0;
286 #ifndef LWP0_CPU_INFO
287 	l->l_cpu = curcpu();
288 #endif
289 
290 	/*
291 	 * Attempt to find console and initialize
292 	 * in case of early panic or other messages.
293 	 */
294 	consinit();
295 
296 	kernel_lock_init();
297 	once_init();
298 	mutex_init(&cpu_lock, MUTEX_DEFAULT, IPL_NONE);
299 
300 	uvm_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 	/* Initialize the extent manager. */
314 	extent_init();
315 
316 	/* Do machine-dependent initialization. */
317 	cpu_startup();
318 
319 	/* Initialize the sysctl subsystem. */
320 	sysctl_init();
321 
322 	/* Initialize callouts, part 1. */
323 	callout_startup();
324 
325 	/*
326 	 * Initialize the kernel authorization subsystem and start the
327 	 * default security model, if any. We need to do this early
328 	 * enough so that subsystems relying on any of the aforementioned
329 	 * can work properly. Since the security model may dictate the
330 	 * credential inheritance policy, it is needed at least before
331 	 * any process is created, specifically proc0.
332 	 */
333 	kauth_init();
334 	secmodel_start();
335 
336 	/* Initialize the buffer cache */
337 	bufinit();
338 
339 	/* Initialize sockets. */
340 	soinit();
341 
342 	/*
343 	 * The following things must be done before autoconfiguration.
344 	 */
345 	evcnt_init();		/* initialize event counters */
346 #if NRND > 0
347 	rnd_init();		/* initialize random number generator */
348 #endif
349 
350 	/* Initialize process and pgrp structures. */
351 	procinit();
352 	lwpinit();
353 
354 	/* Initialize signal-related data structures. */
355 	signal_init();
356 
357 	/* Initialize resource management. */
358 	resource_init();
359 
360 	/* Create process 0 (the swapper). */
361 	proc0_init();
362 
363 	/* Disable preemption during boot. */
364 	kpreempt_disable();
365 
366 	/* Initialize the UID hash table. */
367 	uid_init();
368 
369 	/* Charge root for one process. */
370 	(void)chgproccnt(0, 1);
371 
372 	/* Initialize timekeeping. */
373 	time_init();
374 
375 	/* Initialize the run queues, turnstiles and sleep queues. */
376 	sched_rqinit();
377 	turnstile_init();
378 	sleeptab_init(&sleeptab);
379 
380 	/* Initialize processor-sets */
381 	psets_init();
382 
383 	/* MI initialization of the boot cpu */
384 	error = mi_cpu_attach(curcpu());
385 	KASSERT(error == 0);
386 
387 	/* Initialize timekeeping, part 2. */
388 	time_init2();
389 
390 	/*
391 	 * Initialize mbuf's.  Do this now because we might attempt to
392 	 * allocate mbufs or mbuf clusters during autoconfiguration.
393 	 */
394 	mbinit();
395 
396 	/* Initialize I/O statistics. */
397 	iostat_init();
398 
399 	/* Initialize the log device. */
400 	loginit();
401 
402 	/* Start module system. */
403 	module_init();
404 
405 	/* Initialize the file systems. */
406 #ifdef NVNODE_IMPLICIT
407 	/*
408 	 * If maximum number of vnodes in namei vnode cache is not explicitly
409 	 * defined in kernel config, adjust the number such as we use roughly
410 	 * 10% of memory for vnodes and associated data structures in the
411 	 * assumed worst case.  Do not provide fewer than NVNODE vnodes.
412 	 */
413 	usevnodes =
414 	    calc_cache_size(kernel_map, 10, VNODE_VA_MAXPCT) / VNODE_COST;
415 	if (usevnodes > desiredvnodes)
416 		desiredvnodes = usevnodes;
417 #endif
418 	vfsinit();
419 	lf_init();
420 
421 	/* Initialize fstrans. */
422 	fstrans_init();
423 
424 	/* Initialize the file descriptor system. */
425 	fd_sys_init();
426 
427 	/* Initialize cwd structures */
428 	cwd_sys_init();
429 
430 	/* Initialize kqueue. */
431 	kqueue_init();
432 
433 	/* Initialize message queues. */
434 	mqueue_sysinit();
435 
436 	/* Initialize the system monitor subsystems. */
437 #if NSYSMON_TASKQ > 0
438 	sysmon_task_queue_preinit();
439 #endif
440 
441 #if NSYSMON_ENVSYS > 0
442 	sysmon_envsys_init();
443 #endif
444 
445 #if NSYSMON_POWER > 0
446 	sysmon_power_init();
447 #endif
448 
449 #if NSYSMON_WDOG > 0
450 	sysmon_wdog_init();
451 #endif
452 
453 	inittimecounter();
454 	ntp_init();
455 
456 	/* Initialize the device switch tables. */
457 	devsw_init();
458 
459 	/* Initialize tty subsystem. */
460 	tty_init();
461 	ttyldisc_init();
462 
463 	/* Initialize the buffer cache, part 2. */
464 	bufinit2();
465 
466 	/* Initialize the disk wedge subsystem. */
467 	dkwedge_init();
468 
469 	/* Initialize interfaces. */
470 	ifinit1();
471 
472 	/* Configure the system hardware.  This will enable interrupts. */
473 	configure();
474 
475 	ssp_init();
476 
477 	configure2();
478 
479 	/* Now timer is working.  Enable preemption. */
480 	kpreempt_enable();
481 
482 	ubc_init();		/* must be after autoconfig */
483 
484 #ifdef SYSVSHM
485 	/* Initialize System V style shared memory. */
486 	shminit();
487 #endif
488 
489 #ifdef SYSVSEM
490 	/* Initialize System V style semaphores. */
491 	seminit();
492 #endif
493 
494 #ifdef SYSVMSG
495 	/* Initialize System V style message queues. */
496 	msginit();
497 #endif
498 
499 #if NVERIEXEC > 0
500 	/*
501 	 * Initialise the Veriexec subsystem.
502 	 */
503 	veriexec_init();
504 #endif /* NVERIEXEC > 0 */
505 
506 #if defined(PAX_MPROTECT) || defined(PAX_SEGVGUARD) || defined(PAX_ASLR)
507 	pax_init();
508 #endif /* PAX_MPROTECT || PAX_SEGVGUARD || PAX_ASLR */
509 
510 #ifdef	FAST_IPSEC
511 	/* Attach network crypto subsystem */
512 	ipsec_attach();
513 #endif
514 
515 	/*
516 	 * Initialize protocols.  Block reception of incoming packets
517 	 * until everything is ready.
518 	 */
519 	s = splnet();
520 	ifinit();
521 	domaininit(true);
522 	if_attachdomain();
523 	splx(s);
524 
525 #ifdef GPROF
526 	/* Initialize kernel profiling. */
527 	kmstartup();
528 #endif
529 
530 	/* Initialize system accounting. */
531 	acct_init();
532 
533 #ifndef PIPE_SOCKETPAIR
534 	/* Initialize pipes. */
535 	pipe_init();
536 #endif
537 
538 #ifdef KTRACE
539 	/* Initialize ktrace. */
540 	ktrinit();
541 #endif
542 
543 	/* Initialize the UUID system calls. */
544 	uuid_init();
545 
546 #ifdef WAPBL
547 	/* Initialize write-ahead physical block logging. */
548 	wapbl_init();
549 #endif
550 
551 	/*
552 	 * Create process 1 (init(8)).  We do this now, as Unix has
553 	 * historically had init be process 1, and changing this would
554 	 * probably upset a lot of people.
555 	 *
556 	 * Note that process 1 won't immediately exec init(8), but will
557 	 * wait for us to inform it that the root file system has been
558 	 * mounted.
559 	 */
560 	if (fork1(l, 0, SIGCHLD, NULL, 0, start_init, NULL, NULL, &initproc))
561 		panic("fork init");
562 
563 	/*
564 	 * Load any remaining builtin modules, and hand back temporary
565 	 * storage to the VM system.
566 	 */
567 	module_init_class(MODULE_CLASS_ANY);
568 
569 	/*
570 	 * Finalize configuration now that all real devices have been
571 	 * found.  This needs to be done before the root device is
572 	 * selected, since finalization may create the root device.
573 	 */
574 	config_finalize();
575 
576 	/*
577 	 * Now that autoconfiguration has completed, we can determine
578 	 * the root and dump devices.
579 	 */
580 	cpu_rootconf();
581 	cpu_dumpconf();
582 
583 	/* Mount the root file system. */
584 	do {
585 		domountroothook();
586 		if ((error = vfs_mountroot())) {
587 			printf("cannot mount root, error = %d\n", error);
588 			boothowto |= RB_ASKNAME;
589 			setroot(root_device,
590 			    (rootdev != NODEV) ? DISKPART(rootdev) : 0);
591 		}
592 	} while (error != 0);
593 	mountroothook_destroy();
594 
595 	/*
596 	 * Initialise the time-of-day clock, passing the time recorded
597 	 * in the root filesystem (if any) for use by systems that
598 	 * don't have a non-volatile time-of-day device.
599 	 */
600 	inittodr(rootfstime);
601 
602 	CIRCLEQ_FIRST(&mountlist)->mnt_flag |= MNT_ROOTFS;
603 	CIRCLEQ_FIRST(&mountlist)->mnt_op->vfs_refcount++;
604 
605 	/*
606 	 * Get the vnode for '/'.  Set filedesc0.fd_fd.fd_cdir to
607 	 * reference it.
608 	 */
609 	error = VFS_ROOT(CIRCLEQ_FIRST(&mountlist), &rootvnode);
610 	if (error)
611 		panic("cannot find root vnode, error=%d", error);
612 	cwdi0.cwdi_cdir = rootvnode;
613 	VREF(cwdi0.cwdi_cdir);
614 	VOP_UNLOCK(rootvnode, 0);
615 	cwdi0.cwdi_rdir = NULL;
616 
617 	/*
618 	 * Now that root is mounted, we can fixup initproc's CWD
619 	 * info.  All other processes are kthreads, which merely
620 	 * share proc0's CWD info.
621 	 */
622 	initproc->p_cwdi->cwdi_cdir = rootvnode;
623 	VREF(initproc->p_cwdi->cwdi_cdir);
624 	initproc->p_cwdi->cwdi_rdir = NULL;
625 
626 	/*
627 	 * Now can look at time, having had a chance to verify the time
628 	 * from the file system.  Reset l->l_rtime as it may have been
629 	 * munched in mi_switch() after the time got set.
630 	 */
631 	getnanotime(&time);
632 	boottime = time;
633 #ifdef COMPAT_50
634 	{
635 		struct timeval tv;
636 		TIMESPEC_TO_TIMEVAL(&tv, &time);
637 		timeval_to_timeval50(&tv, &boottime50);
638 	}
639 #endif
640 	mutex_enter(proc_lock);
641 	LIST_FOREACH(p, &allproc, p_list) {
642 		KASSERT((p->p_flag & PK_MARKER) == 0);
643 		mutex_enter(p->p_lock);
644 		TIMESPEC_TO_TIMEVAL(&p->p_stats->p_start, &time);
645 		LIST_FOREACH(l, &p->p_lwps, l_sibling) {
646 			lwp_lock(l);
647 			memset(&l->l_rtime, 0, sizeof(l->l_rtime));
648 			lwp_unlock(l);
649 		}
650 		mutex_exit(p->p_lock);
651 	}
652 	mutex_exit(proc_lock);
653 	binuptime(&curlwp->l_stime);
654 
655 	for (CPU_INFO_FOREACH(cii, ci)) {
656 		ci->ci_schedstate.spc_lastmod = time_second;
657 	}
658 
659 	/* Create the pageout daemon kernel thread. */
660 	uvm_swap_init();
661 	if (kthread_create(PRI_PGDAEMON, KTHREAD_MPSAFE, NULL, uvm_pageout,
662 	    NULL, NULL, "pgdaemon"))
663 		panic("fork pagedaemon");
664 
665 	/* Create the filesystem syncer kernel thread. */
666 	if (kthread_create(PRI_IOFLUSH, KTHREAD_MPSAFE, NULL, sched_sync,
667 	    NULL, NULL, "ioflush"))
668 		panic("fork syncer");
669 
670 	/* Create the aiodone daemon kernel thread. */
671 	if (workqueue_create(&uvm.aiodone_queue, "aiodoned",
672 	    uvm_aiodone_worker, NULL, PRI_VM, IPL_NONE, WQ_MPSAFE))
673 		panic("fork aiodoned");
674 
675 	vmem_rehash_start();
676 
677 	/* Initialize exec structures */
678 	exec_init(1);
679 
680 	/*
681 	 * Okay, now we can let init(8) exec!  It's off to userland!
682 	 */
683 	mutex_enter(proc_lock);
684 	start_init_exec = 1;
685 	cv_broadcast(&lbolt);
686 	mutex_exit(proc_lock);
687 
688 	/* The scheduler is an infinite loop. */
689 	uvm_scheduler();
690 	/* NOTREACHED */
691 }
692 
693 static void
694 check_console(struct lwp *l)
695 {
696 	struct nameidata nd;
697 	int error;
698 
699 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_SYSSPACE, "/dev/console");
700 	error = namei(&nd);
701 	if (error == 0)
702 		vrele(nd.ni_vp);
703 	else if (error == ENOENT)
704 		printf("warning: no /dev/console\n");
705 	else
706 		printf("warning: lookup /dev/console: error %d\n", error);
707 }
708 
709 /*
710  * List of paths to try when searching for "init".
711  */
712 static const char * const initpaths[] = {
713 	"/sbin/init",
714 	"/sbin/oinit",
715 	"/sbin/init.bak",
716 	NULL,
717 };
718 
719 /*
720  * Start the initial user process; try exec'ing each pathname in "initpaths".
721  * The program is invoked with one argument containing the boot flags.
722  */
723 static void
724 start_init(void *arg)
725 {
726 	struct lwp *l = arg;
727 	struct proc *p = l->l_proc;
728 	vaddr_t addr;
729 	struct sys_execve_args /* {
730 		syscallarg(const char *) path;
731 		syscallarg(char * const *) argp;
732 		syscallarg(char * const *) envp;
733 	} */ args;
734 	int options, i, error;
735 	register_t retval[2];
736 	char flags[4], *flagsp;
737 	const char *path, *slash;
738 	char *ucp, **uap, *arg0, *arg1 = NULL;
739 	char ipath[129];
740 	int ipx, len;
741 
742 	/*
743 	 * Now in process 1.
744 	 */
745 	strncpy(p->p_comm, "init", MAXCOMLEN);
746 
747 	/*
748 	 * Wait for main() to tell us that it's safe to exec.
749 	 */
750 	mutex_enter(proc_lock);
751 	while (start_init_exec == 0)
752 		cv_wait(&lbolt, proc_lock);
753 	mutex_exit(proc_lock);
754 
755 	/*
756 	 * This is not the right way to do this.  We really should
757 	 * hand-craft a descriptor onto /dev/console to hand to init,
758 	 * but that's a _lot_ more work, and the benefit from this easy
759 	 * hack makes up for the "good is the enemy of the best" effect.
760 	 */
761 	check_console(l);
762 
763 	/*
764 	 * Need just enough stack to hold the faked-up "execve()" arguments.
765 	 */
766 	addr = (vaddr_t)STACK_ALLOC(USRSTACK, PAGE_SIZE);
767 	if (uvm_map(&p->p_vmspace->vm_map, &addr, PAGE_SIZE,
768                     NULL, UVM_UNKNOWN_OFFSET, 0,
769                     UVM_MAPFLAG(UVM_PROT_ALL, UVM_PROT_ALL, UVM_INH_COPY,
770 		    UVM_ADV_NORMAL,
771                     UVM_FLAG_FIXED|UVM_FLAG_OVERLAY|UVM_FLAG_COPYONW)) != 0)
772 		panic("init: couldn't allocate argument space");
773 	p->p_vmspace->vm_maxsaddr = (void *)STACK_MAX(addr, PAGE_SIZE);
774 
775 	ipx = 0;
776 	while (1) {
777 		if (boothowto & RB_ASKNAME) {
778 			printf("init path");
779 			if (initpaths[ipx])
780 				printf(" (default %s)", initpaths[ipx]);
781 			printf(": ");
782 			len = cngetsn(ipath, sizeof(ipath)-1);
783 			if (len == 4 && strcmp(ipath, "halt") == 0) {
784 				cpu_reboot(RB_HALT, NULL);
785 			} else if (len == 6 && strcmp(ipath, "reboot") == 0) {
786 				cpu_reboot(0, NULL);
787 #if defined(DDB)
788 			} else if (len == 3 && strcmp(ipath, "ddb") == 0) {
789 				console_debugger();
790 				continue;
791 #endif
792 			} else if (len > 0 && ipath[0] == '/') {
793 				ipath[len] = '\0';
794 				path = ipath;
795 			} else if (len == 0 && initpaths[ipx] != NULL) {
796 				path = initpaths[ipx++];
797 			} else {
798 				printf("use absolute path, ");
799 #if defined(DDB)
800 				printf("\"ddb\", ");
801 #endif
802 				printf("\"halt\", or \"reboot\"\n");
803 				continue;
804 			}
805 		} else {
806 			if ((path = initpaths[ipx++]) == NULL) {
807 				ipx = 0;
808 				boothowto |= RB_ASKNAME;
809 				continue;
810 			}
811 		}
812 
813 		ucp = (char *)USRSTACK;
814 
815 		/*
816 		 * Construct the boot flag argument.
817 		 */
818 		flagsp = flags;
819 		*flagsp++ = '-';
820 		options = 0;
821 
822 		if (boothowto & RB_SINGLE) {
823 			*flagsp++ = 's';
824 			options = 1;
825 		}
826 #ifdef notyet
827 		if (boothowto & RB_FASTBOOT) {
828 			*flagsp++ = 'f';
829 			options = 1;
830 		}
831 #endif
832 
833 		/*
834 		 * Move out the flags (arg 1), if necessary.
835 		 */
836 		if (options != 0) {
837 			*flagsp++ = '\0';
838 			i = flagsp - flags;
839 #ifdef DEBUG
840 			printf("init: copying out flags `%s' %d\n", flags, i);
841 #endif
842 			arg1 = STACK_ALLOC(ucp, i);
843 			ucp = STACK_MAX(arg1, i);
844 			(void)copyout((void *)flags, arg1, i);
845 		}
846 
847 		/*
848 		 * Move out the file name (also arg 0).
849 		 */
850 		i = strlen(path) + 1;
851 #ifdef DEBUG
852 		printf("init: copying out path `%s' %d\n", path, i);
853 #else
854 		if (boothowto & RB_ASKNAME || path != initpaths[0])
855 			printf("init: trying %s\n", path);
856 #endif
857 		arg0 = STACK_ALLOC(ucp, i);
858 		ucp = STACK_MAX(arg0, i);
859 		(void)copyout(path, arg0, i);
860 
861 		/*
862 		 * Move out the arg pointers.
863 		 */
864 		ucp = (void *)STACK_ALIGN(ucp, ALIGNBYTES);
865 		uap = (char **)STACK_ALLOC(ucp, sizeof(char *) * 3);
866 		SCARG(&args, path) = arg0;
867 		SCARG(&args, argp) = uap;
868 		SCARG(&args, envp) = NULL;
869 		slash = strrchr(path, '/');
870 		if (slash)
871 			(void)suword((void *)uap++,
872 			    (long)arg0 + (slash + 1 - path));
873 		else
874 			(void)suword((void *)uap++, (long)arg0);
875 		if (options != 0)
876 			(void)suword((void *)uap++, (long)arg1);
877 		(void)suword((void *)uap++, 0);	/* terminator */
878 
879 		/*
880 		 * Now try to exec the program.  If can't for any reason
881 		 * other than it doesn't exist, complain.
882 		 */
883 		error = sys_execve(l, &args, retval);
884 		if (error == 0 || error == EJUSTRETURN) {
885 			KERNEL_UNLOCK_LAST(l);
886 			return;
887 		}
888 		printf("exec %s: error %d\n", path, error);
889 	}
890 	printf("init: not found\n");
891 	panic("no init");
892 }
893 
894 /*
895  * calculate cache size from physmem and vm_map size.
896  */
897 vaddr_t
898 calc_cache_size(struct vm_map *map, int pct, int va_pct)
899 {
900 	paddr_t t;
901 
902 	/* XXX should consider competing cache if any */
903 	/* XXX should consider submaps */
904 	t = (uintmax_t)physmem * pct / 100 * PAGE_SIZE;
905 	if (map != NULL) {
906 		vsize_t vsize;
907 
908 		vsize = vm_map_max(map) - vm_map_min(map);
909 		vsize = (uintmax_t)vsize * va_pct / 100;
910 		if (t > vsize) {
911 			t = vsize;
912 		}
913 	}
914 	return t;
915 }
916 
917 /*
918  * Print the system start up banner.
919  *
920  * - Print a limited banner if AB_SILENT.
921  * - Always send normal banner to the log.
922  */
923 #define MEM_PBUFSIZE	sizeof("99999 MB")
924 
925 void
926 banner(void)
927 {
928 	static char notice[] = " Notice: this software is "
929 	    "protected by copyright";
930 	char pbuf[81];
931 	void (*pr)(const char *, ...);
932 	int i;
933 
934 	if ((boothowto & AB_SILENT) != 0) {
935 		snprintf(pbuf, sizeof(pbuf), "%s %s (%s)",
936 		    ostype, osrelease, kernel_ident);
937 		printf("%s", pbuf);
938 		for (i = 80 - strlen(pbuf) - sizeof(notice); i > 0; i--)
939 			printf(" ");
940 		printf("%s\n", notice);
941 		pr = aprint_normal;
942 	} else {
943 		pr = printf;
944 	}
945 
946 	memset(pbuf, 0, sizeof(pbuf));
947 	(*pr)("%s%s", copyright, version);
948 	format_bytes(pbuf, MEM_PBUFSIZE, ctob((uint64_t)physmem));
949 	(*pr)("total memory = %s\n", pbuf);
950 	format_bytes(pbuf, MEM_PBUFSIZE, ctob((uint64_t)uvmexp.free));
951 	(*pr)("avail memory = %s\n", pbuf);
952 }
953