xref: /netbsd-src/sys/kern/init_main.c (revision a6f3f22f245acb8ee3bbf6871d7dce989204fa97)
1 /*	$NetBSD: init_main.c,v 1.471 2015/10/17 13:52:52 jmcneill 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 1.471 2015/10/17 13:52:52 jmcneill 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(makeoptions_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 "ether.h"
125 #include "ksyms.h"
126 
127 #include "veriexec.h"
128 
129 #include <sys/param.h>
130 #include <sys/acct.h>
131 #include <sys/filedesc.h>
132 #include <sys/file.h>
133 #include <sys/errno.h>
134 #include <sys/callout.h>
135 #include <sys/cpu.h>
136 #include <sys/cpufreq.h>
137 #include <sys/spldebug.h>
138 #include <sys/kernel.h>
139 #include <sys/mount.h>
140 #include <sys/proc.h>
141 #include <sys/kthread.h>
142 #include <sys/resourcevar.h>
143 #include <sys/signalvar.h>
144 #include <sys/systm.h>
145 #include <sys/vnode.h>
146 #include <sys/fstrans.h>
147 #include <sys/tty.h>
148 #include <sys/conf.h>
149 #include <sys/disklabel.h>
150 #include <sys/buf.h>
151 #include <sys/device.h>
152 #include <sys/exec.h>
153 #include <sys/socketvar.h>
154 #include <sys/protosw.h>
155 #include <sys/percpu.h>
156 #include <sys/pserialize.h>
157 #include <sys/pset.h>
158 #include <sys/sysctl.h>
159 #include <sys/reboot.h>
160 #include <sys/event.h>
161 #include <sys/mbuf.h>
162 #include <sys/sched.h>
163 #include <sys/sleepq.h>
164 #include <sys/ipi.h>
165 #include <sys/iostat.h>
166 #include <sys/vmem.h>
167 #include <sys/uuid.h>
168 #include <sys/extent.h>
169 #include <sys/disk.h>
170 #include <sys/msgbuf.h>
171 #include <sys/module.h>
172 #include <sys/event.h>
173 #include <sys/lockf.h>
174 #include <sys/once.h>
175 #include <sys/kcpuset.h>
176 #include <sys/ksyms.h>
177 #include <sys/uidinfo.h>
178 #include <sys/kprintf.h>
179 #ifdef IPSEC
180 #include <netipsec/ipsec.h>
181 #endif
182 #ifdef SYSVSHM
183 #include <sys/shm.h>
184 #endif
185 #ifdef SYSVSEM
186 #include <sys/sem.h>
187 #endif
188 #ifdef SYSVMSG
189 #include <sys/msg.h>
190 #endif
191 #include <sys/domain.h>
192 #include <sys/namei.h>
193 #include <sys/rnd.h>
194 #include <sys/pipe.h>
195 #if NVERIEXEC > 0
196 #include <sys/verified_exec.h>
197 #endif /* NVERIEXEC > 0 */
198 #ifdef KTRACE
199 #include <sys/ktrace.h>
200 #endif
201 #include <sys/kauth.h>
202 #include <net80211/ieee80211_netbsd.h>
203 #ifdef PTRACE
204 #include <sys/ptrace.h>
205 #endif /* PTRACE */
206 #include <sys/cprng.h>
207 
208 #include <sys/syscall.h>
209 #include <sys/syscallargs.h>
210 
211 #if defined(PAX_MPROTECT) || defined(PAX_SEGVGUARD) || defined(PAX_ASLR)
212 #include <sys/pax.h>
213 #endif /* PAX_MPROTECT || PAX_SEGVGUARD || PAX_ASLR */
214 
215 #include <secmodel/secmodel.h>
216 
217 #include <ufs/ufs/quota.h>
218 
219 #include <miscfs/genfs/genfs.h>
220 #include <miscfs/specfs/specdev.h>
221 
222 #include <sys/cpu.h>
223 
224 #include <uvm/uvm.h>	/* extern struct uvm uvm */
225 
226 #include <dev/cons.h>
227 #include <dev/splash/splash.h>
228 
229 #include <net/bpf.h>
230 #include <net/if.h>
231 #include <net/raw_cb.h>
232 #include <net/if_llatbl.h>
233 
234 #include <prop/proplib.h>
235 
236 #include <sys/userconf.h>
237 
238 extern struct lwp lwp0;
239 extern time_t rootfstime;
240 
241 #ifndef curlwp
242 struct	lwp *curlwp = &lwp0;
243 #endif
244 struct	proc *initproc;
245 
246 struct	vnode *rootvp, *swapdev_vp;
247 int	boothowto;
248 int	cold = 1;			/* still working on startup */
249 struct timespec boottime;	        /* time at system startup - will only follow settime deltas */
250 
251 int	start_init_exec;		/* semaphore for start_init() */
252 
253 cprng_strong_t	*kern_cprng;
254 
255 static void check_console(struct lwp *l);
256 static void start_init(void *);
257 static void configure(void);
258 static void configure2(void);
259 static void configure3(void);
260 void main(void);
261 
262 /*
263  * System startup; initialize the world, create process 0, mount root
264  * filesystem, and fork to create init and pagedaemon.  Most of the
265  * hard work is done in the lower-level initialization routines including
266  * startup(), which does memory initialization and autoconfiguration.
267  */
268 void
269 main(void)
270 {
271 	struct timespec time;
272 	struct lwp *l;
273 	struct proc *p;
274 	int s, error;
275 #ifdef NVNODE_IMPLICIT
276 	int usevnodes;
277 #endif
278 	CPU_INFO_ITERATOR cii;
279 	struct cpu_info *ci;
280 
281 	l = &lwp0;
282 #ifndef LWP0_CPU_INFO
283 	l->l_cpu = curcpu();
284 #endif
285 	l->l_pflag |= LP_RUNNING;
286 
287 	/*
288 	 * Attempt to find console and initialize
289 	 * in case of early panic or other messages.
290 	 */
291 	consinit();
292 
293 	kernel_lock_init();
294 	once_init();
295 
296 	mi_cpu_init();
297 	kernconfig_lock_init();
298 	kthread_sysinit();
299 
300 	/* Initialize the device switch tables. */
301 	devsw_init();
302 
303 	/* Initialize event counters. */
304 	evcnt_init();
305 
306 	uvm_init();
307 	ubchist_init();
308 	kcpuset_sysinit();
309 
310 	prop_kern_init();
311 
312 #if ((NKSYMS > 0) || (NDDB > 0) || (NMODULAR > 0))
313 	ksyms_init();
314 #endif
315 	kprintf_init();
316 
317 	percpu_init();
318 
319 	/* Initialize lock caches. */
320 	mutex_obj_init();
321 	rw_obj_init();
322 
323 	/* Passive serialization. */
324 	pserialize_init();
325 
326 	/* Initialize the extent manager. */
327 	extent_init();
328 
329 	/* Do machine-dependent initialization. */
330 	cpu_startup();
331 
332 	/* Initialize the sysctl subsystem. */
333 	sysctl_init();
334 
335 	/* Initialize callouts, part 1. */
336 	callout_startup();
337 
338 	/* Initialize the kernel authorization subsystem. */
339 	kauth_init();
340 
341 	secmodel_init();
342 
343 	spec_init();
344 
345 	/*
346 	 * Set BPF op vector.  Can't do this in bpf attach, since
347 	 * network drivers attach before bpf.
348 	 */
349 	bpf_setops();
350 
351 	/* Start module system. */
352 	module_init();
353 
354 	/*
355 	 * Initialize the kernel authorization subsystem and start the
356 	 * default security model, if any. We need to do this early
357 	 * enough so that subsystems relying on any of the aforementioned
358 	 * can work properly. Since the security model may dictate the
359 	 * credential inheritance policy, it is needed at least before
360 	 * any process is created, specifically proc0.
361 	 */
362 	module_init_class(MODULE_CLASS_SECMODEL);
363 
364 	/* Initialize the buffer cache */
365 	bufinit();
366 
367 
368 #if defined(SPLASHSCREEN) && defined(SPLASHSCREEN_IMAGE)
369 	size_t splash_size = (&_binary_splash_image_end -
370 	    &_binary_splash_image_start) * sizeof(void *);
371 	splash_setimage(&_binary_splash_image_start, splash_size);
372 #endif
373 
374 	/* Initialize sockets. */
375 	soinit();
376 
377 	/*
378 	 * The following things must be done before autoconfiguration.
379 	 */
380 	rnd_init();		/* initialize entropy pool */
381 
382 	cprng_init();		/* initialize cryptographic PRNG */
383 
384 	/* Initialize process and pgrp structures. */
385 	procinit();
386 	lwpinit();
387 
388 	/* Initialize signal-related data structures. */
389 	signal_init();
390 
391 	/* Initialize resource management. */
392 	resource_init();
393 
394 	/* Create process 0. */
395 	proc0_init();
396 	lwp0_init();
397 
398 	/* Disable preemption during boot. */
399 	kpreempt_disable();
400 
401 	/* Initialize the UID hash table. */
402 	uid_init();
403 
404 	/* Charge root for one process. */
405 	(void)chgproccnt(0, 1);
406 
407 	/* Initialize timekeeping. */
408 	time_init();
409 
410 	/* Initialize the run queues, turnstiles and sleep queues. */
411 	sched_rqinit();
412 	turnstile_init();
413 	sleeptab_init(&sleeptab);
414 
415 	sched_init();
416 
417 	/* Initialize processor-sets */
418 	psets_init();
419 
420 	/* Initialize cpufreq(9) */
421 	cpufreq_init();
422 
423 	/* MI initialization of the boot cpu */
424 	error = mi_cpu_attach(curcpu());
425 	KASSERT(error == 0);
426 
427 	/* Initialize timekeeping, part 2. */
428 	time_init2();
429 
430 	/*
431 	 * Initialize mbuf's.  Do this now because we might attempt to
432 	 * allocate mbufs or mbuf clusters during autoconfiguration.
433 	 */
434 	mbinit();
435 
436 	/* Initialize I/O statistics. */
437 	iostat_init();
438 
439 	/* Initialize the log device. */
440 	loginit();
441 
442 	/* Second part of module system initialization. */
443 	module_start_unload_thread();
444 
445 	/* Initialize the file systems. */
446 #ifdef NVNODE_IMPLICIT
447 	/*
448 	 * If maximum number of vnodes in namei vnode cache is not explicitly
449 	 * defined in kernel config, adjust the number such as we use roughly
450 	 * 10% of memory for vnodes and associated data structures in the
451 	 * assumed worst case.  Do not provide fewer than NVNODE vnodes.
452 	 */
453 	usevnodes = calc_cache_size(vmem_size(kmem_arena, VMEM_FREE|VMEM_ALLOC),
454 	    10, VNODE_KMEM_MAXPCT) / VNODE_COST;
455 	if (usevnodes > desiredvnodes)
456 		desiredvnodes = usevnodes;
457 #endif
458 	vfsinit();
459 	lf_init();
460 
461 	/* Initialize fstrans. */
462 	fstrans_init();
463 
464 	/* Initialize the file descriptor system. */
465 	fd_sys_init();
466 
467 	/* Initialize cwd structures */
468 	cwd_sys_init();
469 
470 	/* Initialize kqueue. */
471 	kqueue_init();
472 
473 	inittimecounter();
474 	ntp_init();
475 
476 	/* Initialize tty subsystem. */
477 	tty_init();
478 	ttyldisc_init();
479 
480 	/* Initialize the buffer cache, part 2. */
481 	bufinit2();
482 
483 	/* Initialize the disk wedge subsystem. */
484 	dkwedge_init();
485 
486 	/* Initialize the kernel strong PRNG. */
487 	kern_cprng = cprng_strong_create("kernel", IPL_VM,
488 					 CPRNG_INIT_ANY|CPRNG_REKEY_ANY);
489 
490 	/* Initialize interfaces. */
491 	ifinit1();
492 
493 	spldebug_start();
494 
495 	/* Initialize sockets thread(s) */
496 	soinit1();
497 
498 	/* Configure the system hardware.  This will enable interrupts. */
499 	configure();
500 
501 	/* Once all CPUs are detected, initialize the per-CPU cprng_fast.  */
502 	cprng_fast_init();
503 
504 	ssp_init();
505 
506 	ubc_init();		/* must be after autoconfig */
507 
508 	mm_init();
509 
510 	configure2();
511 
512 	ipi_sysinit();
513 
514 	/* Now timer is working.  Enable preemption. */
515 	kpreempt_enable();
516 
517 	/* Get the threads going and into any sleeps before continuing. */
518 	yield();
519 
520 	/* Enable deferred processing of RNG samples */
521 	rnd_init_softint();
522 
523 #ifdef RND_PRINTF
524 	/* Enable periodic injection of console output into entropy pool */
525 	kprintf_init_callout();
526 #endif
527 
528 #ifdef SYSVSHM
529 	/* Initialize System V style shared memory. */
530 	shminit();
531 #endif
532 
533 	vmem_rehash_start();	/* must be before exec_init */
534 
535 	/* Initialize exec structures */
536 	exec_init(1);		/* seminit calls exithook_establish() */
537 
538 #ifdef SYSVSEM
539 	/* Initialize System V style semaphores. */
540 	seminit();
541 #endif
542 
543 #ifdef SYSVMSG
544 	/* Initialize System V style message queues. */
545 	msginit();
546 #endif
547 
548 #if NVERIEXEC > 0
549 	/*
550 	 * Initialise the Veriexec subsystem.
551 	 */
552 	veriexec_init();
553 #endif /* NVERIEXEC > 0 */
554 
555 #if defined(PAX_MPROTECT) || defined(PAX_SEGVGUARD) || defined(PAX_ASLR)
556 	pax_init();
557 #endif /* PAX_MPROTECT || PAX_SEGVGUARD || PAX_ASLR */
558 
559 #ifdef	IPSEC
560 	/* Attach network crypto subsystem */
561 	ipsec_attach();
562 #endif
563 
564 	/*
565 	 * Initialize protocols.  Block reception of incoming packets
566 	 * until everything is ready.
567 	 */
568 	s = splnet();
569 	ifinit();
570 #if NETHER > 0
571 	lltableinit();
572 #endif
573 	domaininit(true);
574 	if_attachdomain();
575 	splx(s);
576 
577 #ifdef GPROF
578 	/* Initialize kernel profiling. */
579 	kmstartup();
580 #endif
581 
582 	/* Initialize system accounting. */
583 	acct_init();
584 
585 #ifndef PIPE_SOCKETPAIR
586 	/* Initialize pipes. */
587 	pipe_init();
588 #endif
589 
590 #ifdef KTRACE
591 	/* Initialize ktrace. */
592 	ktrinit();
593 #endif
594 
595 #ifdef PTRACE
596 	/* Initialize ptrace. */
597 	ptrace_init();
598 #endif /* PTRACE */
599 
600 	machdep_init();
601 
602 	procinit_sysctl();
603 
604 	/*
605 	 * Create process 1 (init(8)).  We do this now, as Unix has
606 	 * historically had init be process 1, and changing this would
607 	 * probably upset a lot of people.
608 	 *
609 	 * Note that process 1 won't immediately exec init(8), but will
610 	 * wait for us to inform it that the root file system has been
611 	 * mounted.
612 	 */
613 	if (fork1(l, 0, SIGCHLD, NULL, 0, start_init, NULL, NULL, &initproc))
614 		panic("fork init");
615 
616 	/*
617 	 * Load any remaining builtin modules, and hand back temporary
618 	 * storage to the VM system.  Then require force when loading any
619 	 * remaining un-init'ed built-in modules to avoid later surprises.
620 	 */
621 	module_init_class(MODULE_CLASS_ANY);
622 	module_builtin_require_force();
623 
624 	/*
625 	 * Finalize configuration now that all real devices have been
626 	 * found.  This needs to be done before the root device is
627 	 * selected, since finalization may create the root device.
628 	 */
629 	config_finalize();
630 
631 	sysctl_finalize();
632 
633 	/*
634 	 * Now that autoconfiguration has completed, we can determine
635 	 * the root and dump devices.
636 	 */
637 	cpu_rootconf();
638 	cpu_dumpconf();
639 
640 	/* Mount the root file system. */
641 	do {
642 		domountroothook(root_device);
643 		if ((error = vfs_mountroot())) {
644 			printf("cannot mount root, error = %d\n", error);
645 			boothowto |= RB_ASKNAME;
646 			setroot(root_device,
647 			    (rootdev != NODEV) ? DISKPART(rootdev) : 0);
648 		}
649 	} while (error != 0);
650 	mountroothook_destroy();
651 
652 	configure3();
653 
654 	/*
655 	 * Initialise the time-of-day clock, passing the time recorded
656 	 * in the root filesystem (if any) for use by systems that
657 	 * don't have a non-volatile time-of-day device.
658 	 */
659 	inittodr(rootfstime);
660 
661 	/*
662 	 * Now can look at time, having had a chance to verify the time
663 	 * from the file system.  Reset l->l_rtime as it may have been
664 	 * munched in mi_switch() after the time got set.
665 	 */
666 	getnanotime(&time);
667 	boottime = time;
668 
669 	mutex_enter(proc_lock);
670 	LIST_FOREACH(p, &allproc, p_list) {
671 		KASSERT((p->p_flag & PK_MARKER) == 0);
672 		mutex_enter(p->p_lock);
673 		TIMESPEC_TO_TIMEVAL(&p->p_stats->p_start, &time);
674 		LIST_FOREACH(l, &p->p_lwps, l_sibling) {
675 			lwp_lock(l);
676 			memset(&l->l_rtime, 0, sizeof(l->l_rtime));
677 			lwp_unlock(l);
678 		}
679 		mutex_exit(p->p_lock);
680 	}
681 	mutex_exit(proc_lock);
682 	binuptime(&curlwp->l_stime);
683 
684 	for (CPU_INFO_FOREACH(cii, ci)) {
685 		ci->ci_schedstate.spc_lastmod = time_second;
686 	}
687 
688 	/* Create the pageout daemon kernel thread. */
689 	uvm_swap_init();
690 	if (kthread_create(PRI_PGDAEMON, KTHREAD_MPSAFE, NULL, uvm_pageout,
691 	    NULL, NULL, "pgdaemon"))
692 		panic("fork pagedaemon");
693 
694 	/* Create the filesystem syncer kernel thread. */
695 	if (kthread_create(PRI_IOFLUSH, KTHREAD_MPSAFE, NULL, sched_sync,
696 	    NULL, NULL, "ioflush"))
697 		panic("fork syncer");
698 
699 	/* Create the aiodone daemon kernel thread. */
700 	if (workqueue_create(&uvm.aiodone_queue, "aiodoned",
701 	    uvm_aiodone_worker, NULL, PRI_VM, IPL_NONE, WQ_MPSAFE))
702 		panic("fork aiodoned");
703 
704 	/* Wait for final configure threads to complete. */
705 	config_finalize_mountroot();
706 
707 	/*
708 	 * Okay, now we can let init(8) exec!  It's off to userland!
709 	 */
710 	mutex_enter(proc_lock);
711 	start_init_exec = 1;
712 	cv_broadcast(&lbolt);
713 	mutex_exit(proc_lock);
714 
715 	/* The scheduler is an infinite loop. */
716 	uvm_scheduler();
717 	/* NOTREACHED */
718 }
719 
720 /*
721  * Configure the system's hardware.
722  */
723 static void
724 configure(void)
725 {
726 
727 	/* Initialize autoconf data structures. */
728 	config_init_mi();
729 	/*
730 	 * XXX
731 	 * callout_setfunc() requires mutex(9) so it can't be in config_init()
732 	 * on amiga and atari which use config_init() and autoconf(9) fucntions
733 	 * to initialize console devices.
734 	 */
735 	config_twiddle_init();
736 
737 	pmf_init();
738 #if NDRVCTL > 0
739 	drvctl_init();
740 #endif
741 
742 	/* Initialize driver modules */
743 	module_init_class(MODULE_CLASS_DRIVER);
744 
745 	userconf_init();
746 	if (boothowto & RB_USERCONF)
747 		userconf_prompt();
748 
749 	if ((boothowto & (AB_SILENT|AB_VERBOSE)) == AB_SILENT) {
750 		printf_nolog("Detecting hardware...");
751 	}
752 
753 	/*
754 	 * Do the machine-dependent portion of autoconfiguration.  This
755 	 * sets the configuration machinery here in motion by "finding"
756 	 * the root bus.  When this function returns, we expect interrupts
757 	 * to be enabled.
758 	 */
759 	cpu_configure();
760 }
761 
762 static void
763 configure2(void)
764 {
765 	CPU_INFO_ITERATOR cii;
766 	struct cpu_info *ci;
767 	int s;
768 
769 	/*
770 	 * Now that we've found all the hardware, start the real time
771 	 * and statistics clocks.
772 	 */
773 	initclocks();
774 
775 	cold = 0;	/* clocks are running, we're warm now! */
776 	s = splsched();
777 	curcpu()->ci_schedstate.spc_flags |= SPCF_RUNNING;
778 	splx(s);
779 
780 	/* Boot the secondary processors. */
781 	for (CPU_INFO_FOREACH(cii, ci)) {
782 		uvm_cpu_attach(ci);
783 	}
784 	mp_online = true;
785 #if defined(MULTIPROCESSOR)
786 	cpu_boot_secondary_processors();
787 #endif
788 
789 	/* Setup the runqueues and scheduler. */
790 	runq_init();
791 	synch_init();
792 
793 	/*
794 	 * Bus scans can make it appear as if the system has paused, so
795 	 * twiddle constantly while config_interrupts() jobs are running.
796 	 */
797 	config_twiddle_fn(NULL);
798 
799 	/*
800 	 * Create threads to call back and finish configuration for
801 	 * devices that want interrupts enabled.
802 	 */
803 	config_create_interruptthreads();
804 }
805 
806 static void
807 configure3(void)
808 {
809 
810 	/*
811 	 * Create threads to call back and finish configuration for
812 	 * devices that want the mounted root file system.
813 	 */
814 	config_create_mountrootthreads();
815 
816 	/* Get the threads going and into any sleeps before continuing. */
817 	yield();
818 }
819 
820 static void
821 rootconf_handle_wedges(void)
822 {
823 	struct partinfo dpart;
824 	struct partition *p;
825 	struct vnode *vp;
826 	daddr_t startblk;
827 	uint64_t nblks;
828 	device_t dev;
829 	int error;
830 
831 	if (booted_nblks) {
832 		/*
833 		 * bootloader passed geometry
834 		 */
835 		dev      = booted_device;
836 		startblk = booted_startblk;
837 		nblks    = booted_nblks;
838 
839 		/*
840 		 * keep booted_device and booted_partition
841 		 * in case the kernel doesn't identify a wedge
842 		 */
843 	} else {
844 		/*
845 		 * bootloader passed partition number
846 		 *
847 		 * We cannot ask the partition device directly when it is
848 		 * covered by a wedge. Instead we look up the geometry in
849 		 * the disklabel.
850 		 */
851 		vp = opendisk(booted_device);
852 
853 		if (vp == NULL)
854 			return;
855 
856 		error = VOP_IOCTL(vp, DIOCGPART, &dpart, FREAD, NOCRED);
857 		VOP_CLOSE(vp, FREAD, NOCRED);
858 		vput(vp);
859 		if (error)
860 			return;
861 
862 		KASSERT(booted_partition >= 0
863 			&& booted_partition < MAXPARTITIONS);
864 
865 		p = &dpart.disklab->d_partitions[booted_partition];
866 
867 		dev      = booted_device;
868 		startblk = p->p_offset;
869 		nblks    = p->p_size;
870 	}
871 
872 	dev = dkwedge_find_partition(dev, startblk, nblks);
873 	if (dev != NULL) {
874 		booted_device = dev;
875 		booted_partition = 0;
876 	}
877 }
878 
879 void
880 rootconf(void)
881 {
882 	if (booted_device != NULL)
883 		rootconf_handle_wedges();
884 
885 	setroot(booted_device, booted_partition);
886 }
887 
888 static void
889 check_console(struct lwp *l)
890 {
891 	struct vnode *vp;
892 	int error;
893 
894 	error = namei_simple_kernel("/dev/console",
895 				NSM_FOLLOW_NOEMULROOT, &vp);
896 	if (error == 0)
897 		vrele(vp);
898 	else if (error == ENOENT)
899 		printf("warning: no /dev/console\n");
900 	else
901 		printf("warning: lookup /dev/console: error %d\n", error);
902 }
903 
904 /*
905  * List of paths to try when searching for "init".
906  */
907 static const char * const initpaths[] = {
908 	"/sbin/init",
909 	"/sbin/oinit",
910 	"/sbin/init.bak",
911 	"/rescue/init",
912 	NULL,
913 };
914 
915 /*
916  * Start the initial user process; try exec'ing each pathname in "initpaths".
917  * The program is invoked with one argument containing the boot flags.
918  */
919 static void
920 start_init(void *arg)
921 {
922 	struct lwp *l = arg;
923 	struct proc *p = l->l_proc;
924 	vaddr_t addr;
925 	struct sys_execve_args /* {
926 		syscallarg(const char *) path;
927 		syscallarg(char * const *) argp;
928 		syscallarg(char * const *) envp;
929 	} */ args;
930 	int options, i, error;
931 	register_t retval[2];
932 	char flags[4], *flagsp;
933 	const char *path, *slash;
934 	char *ucp, **uap, *arg0, *arg1 = NULL;
935 	char ipath[129];
936 	int ipx, len;
937 
938 	/*
939 	 * Now in process 1.
940 	 */
941 	strncpy(p->p_comm, "init", MAXCOMLEN);
942 
943 	/*
944 	 * Wait for main() to tell us that it's safe to exec.
945 	 */
946 	mutex_enter(proc_lock);
947 	while (start_init_exec == 0)
948 		cv_wait(&lbolt, proc_lock);
949 	mutex_exit(proc_lock);
950 
951 	/*
952 	 * This is not the right way to do this.  We really should
953 	 * hand-craft a descriptor onto /dev/console to hand to init,
954 	 * but that's a _lot_ more work, and the benefit from this easy
955 	 * hack makes up for the "good is the enemy of the best" effect.
956 	 */
957 	check_console(l);
958 
959 	/*
960 	 * Need just enough stack to hold the faked-up "execve()" arguments.
961 	 */
962 	addr = (vaddr_t)STACK_ALLOC(USRSTACK, PAGE_SIZE);
963 	if (uvm_map(&p->p_vmspace->vm_map, &addr, PAGE_SIZE,
964                     NULL, UVM_UNKNOWN_OFFSET, 0,
965                     UVM_MAPFLAG(UVM_PROT_ALL, UVM_PROT_ALL, UVM_INH_COPY,
966 		    UVM_ADV_NORMAL,
967                     UVM_FLAG_FIXED|UVM_FLAG_OVERLAY|UVM_FLAG_COPYONW)) != 0)
968 		panic("init: couldn't allocate argument space");
969 	p->p_vmspace->vm_maxsaddr = (void *)STACK_MAX(addr, PAGE_SIZE);
970 
971 	ipx = 0;
972 	while (1) {
973 		if (boothowto & RB_ASKNAME) {
974 			printf("init path");
975 			if (initpaths[ipx])
976 				printf(" (default %s)", initpaths[ipx]);
977 			printf(": ");
978 			len = cngetsn(ipath, sizeof(ipath)-1);
979 			if (len == 4 && strcmp(ipath, "halt") == 0) {
980 				cpu_reboot(RB_HALT, NULL);
981 			} else if (len == 6 && strcmp(ipath, "reboot") == 0) {
982 				cpu_reboot(0, NULL);
983 #if defined(DDB)
984 			} else if (len == 3 && strcmp(ipath, "ddb") == 0) {
985 				console_debugger();
986 				continue;
987 #endif
988 			} else if (len > 0 && ipath[0] == '/') {
989 				ipath[len] = '\0';
990 				path = ipath;
991 			} else if (len == 0 && initpaths[ipx] != NULL) {
992 				path = initpaths[ipx++];
993 			} else {
994 				printf("use absolute path, ");
995 #if defined(DDB)
996 				printf("\"ddb\", ");
997 #endif
998 				printf("\"halt\", or \"reboot\"\n");
999 				continue;
1000 			}
1001 		} else {
1002 			if ((path = initpaths[ipx++]) == NULL) {
1003 				ipx = 0;
1004 				boothowto |= RB_ASKNAME;
1005 				continue;
1006 			}
1007 		}
1008 
1009 		ucp = (char *)USRSTACK;
1010 
1011 		/*
1012 		 * Construct the boot flag argument.
1013 		 */
1014 		flagsp = flags;
1015 		*flagsp++ = '-';
1016 		options = 0;
1017 
1018 		if (boothowto & RB_SINGLE) {
1019 			*flagsp++ = 's';
1020 			options = 1;
1021 		}
1022 #ifdef notyet
1023 		if (boothowto & RB_FASTBOOT) {
1024 			*flagsp++ = 'f';
1025 			options = 1;
1026 		}
1027 #endif
1028 
1029 		/*
1030 		 * Move out the flags (arg 1), if necessary.
1031 		 */
1032 		if (options != 0) {
1033 			*flagsp++ = '\0';
1034 			i = flagsp - flags;
1035 #ifdef DEBUG
1036 			aprint_normal("init: copying out flags `%s' %d\n", flags, i);
1037 #endif
1038 			arg1 = STACK_ALLOC(ucp, i);
1039 			ucp = STACK_MAX(arg1, i);
1040 			(void)copyout((void *)flags, arg1, i);
1041 		}
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 		(void)copyout(path, arg0, i);
1056 
1057 		/*
1058 		 * Move out the arg pointers.
1059 		 */
1060 		ucp = (void *)STACK_ALIGN(ucp, STACK_ALIGNBYTES);
1061 		uap = (char **)STACK_ALLOC(ucp, sizeof(char *) * 3);
1062 		SCARG(&args, path) = arg0;
1063 		SCARG(&args, argp) = uap;
1064 		SCARG(&args, envp) = NULL;
1065 		slash = strrchr(path, '/');
1066 		if (slash)
1067 			(void)suword((void *)uap++,
1068 			    (long)arg0 + (slash + 1 - path));
1069 		else
1070 			(void)suword((void *)uap++, (long)arg0);
1071 		if (options != 0)
1072 			(void)suword((void *)uap++, (long)arg1);
1073 		(void)suword((void *)uap++, 0);	/* terminator */
1074 
1075 		/*
1076 		 * Now try to exec the program.  If can't for any reason
1077 		 * other than it doesn't exist, complain.
1078 		 */
1079 		error = sys_execve(l, &args, retval);
1080 		if (error == 0 || error == EJUSTRETURN) {
1081 			KERNEL_UNLOCK_LAST(l);
1082 			return;
1083 		}
1084 		printf("exec %s: error %d\n", path, error);
1085 	}
1086 	printf("init: not found\n");
1087 	panic("no init");
1088 }
1089 
1090 /*
1091  * calculate cache size (in bytes) from physmem and vsize.
1092  */
1093 vaddr_t
1094 calc_cache_size(vsize_t vsize, int pct, int va_pct)
1095 {
1096 	paddr_t t;
1097 
1098 	/* XXX should consider competing cache if any */
1099 	/* XXX should consider submaps */
1100 	t = (uintmax_t)physmem * pct / 100 * PAGE_SIZE;
1101 	if (vsize != 0) {
1102 		vsize = (uintmax_t)vsize * va_pct / 100;
1103 		if (t > vsize) {
1104 			t = vsize;
1105 		}
1106 	}
1107 	return t;
1108 }
1109 
1110 /*
1111  * Print the system start up banner.
1112  *
1113  * - Print a limited banner if AB_SILENT.
1114  * - Always send normal banner to the log.
1115  */
1116 #define MEM_PBUFSIZE	sizeof("99999 MB")
1117 
1118 void
1119 banner(void)
1120 {
1121 	static char notice[] = " Notice: this software is "
1122 	    "protected by copyright";
1123 	char pbuf[81];
1124 	void (*pr)(const char *, ...) __printflike(1, 2);
1125 	int i;
1126 
1127 	if ((boothowto & AB_SILENT) != 0) {
1128 		snprintf(pbuf, sizeof(pbuf), "%s %s (%s)",
1129 		    ostype, osrelease, kernel_ident);
1130 		printf_nolog("%s", pbuf);
1131 		for (i = 80 - strlen(pbuf) - sizeof(notice); i > 0; i--)
1132 			printf(" ");
1133 		printf_nolog("%s\n", notice);
1134 		pr = aprint_normal;
1135 	} else {
1136 		pr = printf;
1137 	}
1138 
1139 	memset(pbuf, 0, sizeof(pbuf));
1140 	(*pr)("%s%s", copyright, version);
1141 	format_bytes(pbuf, MEM_PBUFSIZE, ctob((uint64_t)physmem));
1142 	(*pr)("total memory = %s\n", pbuf);
1143 	format_bytes(pbuf, MEM_PBUFSIZE, ctob((uint64_t)uvmexp.free));
1144 	(*pr)("avail memory = %s\n", pbuf);
1145 }
1146