xref: /netbsd/sys/kern/init_main.c (revision de09df2f)
1 /*	$NetBSD: init_main.c,v 1.542 2023/07/07 12:34:50 riastradh Exp $	*/
2 
3 /*-
4  * Copyright (c) 2008, 2009, 2019 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 1.542 2023/07/07 12:34:50 riastradh Exp $");
101 
102 #include "opt_cnmagic.h"
103 #include "opt_ddb.h"
104 #include "opt_heartbeat.h"
105 #include "opt_inet.h"
106 #include "opt_ipsec.h"
107 #include "opt_modular.h"
108 #include "opt_ntp.h"
109 #include "opt_pipe.h"
110 #include "opt_syscall_debug.h"
111 #include "opt_sysv.h"
112 #include "opt_fileassoc.h"
113 #include "opt_ktrace.h"
114 #include "opt_pax.h"
115 #include "opt_compat_netbsd.h"
116 #include "opt_ptrace.h"
117 #include "opt_splash.h"
118 #include "opt_kernhist.h"
119 #include "opt_gprof.h"
120 
121 #if defined(SPLASHSCREEN) && defined(makeoptions_SPLASHSCREEN_IMAGE)
122 extern void *_binary_splash_image_start;
123 extern void *_binary_splash_image_end;
124 #endif
125 
126 #include "ksyms.h"
127 
128 #include "veriexec.h"
129 
130 #include <sys/param.h>
131 #include <sys/acct.h>
132 #include <sys/filedesc.h>
133 #include <sys/file.h>
134 #include <sys/errno.h>
135 #include <sys/callout.h>
136 #include <sys/cpu.h>
137 #include <sys/cpufreq.h>
138 #include <sys/spldebug.h>
139 #include <sys/kernel.h>
140 #include <sys/mount.h>
141 #include <sys/proc.h>
142 #include <sys/lwp.h>
143 #include <sys/kthread.h>
144 #include <sys/resourcevar.h>
145 #include <sys/signalvar.h>
146 #include <sys/systm.h>
147 #include <sys/vnode.h>
148 #include <sys/fstrans.h>
149 #include <sys/tty.h>
150 #include <sys/conf.h>
151 #include <sys/disklabel.h>
152 #include <sys/buf.h>
153 #include <sys/device.h>
154 #include <sys/exec.h>
155 #include <sys/socketvar.h>
156 #include <sys/protosw.h>
157 #include <sys/percpu.h>
158 #include <sys/pserialize.h>
159 #include <sys/pset.h>
160 #include <sys/sysctl.h>
161 #include <sys/reboot.h>
162 #include <sys/event.h>
163 #include <sys/mbuf.h>
164 #include <sys/sched.h>
165 #include <sys/sleepq.h>
166 #include <sys/ipi.h>
167 #include <sys/iostat.h>
168 #include <sys/vmem.h>
169 #include <sys/uuid.h>
170 #include <sys/extent.h>
171 #include <sys/disk.h>
172 #include <sys/msgbuf.h>
173 #include <sys/module.h>
174 #include <sys/module_hook.h>
175 #include <sys/event.h>
176 #include <sys/lockf.h>
177 #include <sys/once.h>
178 #include <sys/kcpuset.h>
179 #include <sys/ksyms.h>
180 #include <sys/uidinfo.h>
181 #include <sys/kprintf.h>
182 #include <sys/bufq.h>
183 #include <sys/threadpool.h>
184 #include <sys/futex.h>
185 #ifdef IPSEC
186 #include <netipsec/ipsec.h>
187 #endif
188 #include <sys/domain.h>
189 #include <sys/namei.h>
190 #include <sys/rnd.h>
191 #include <sys/pipe.h>
192 #if NVERIEXEC > 0
193 #include <sys/verified_exec.h>
194 #endif /* NVERIEXEC > 0 */
195 #ifdef KTRACE
196 #include <sys/ktrace.h>
197 #endif
198 #include <sys/kauth.h>
199 #include <net80211/ieee80211_netbsd.h>
200 #include <sys/cprng.h>
201 #include <sys/psref.h>
202 #include <sys/radixtree.h>
203 #include <sys/heartbeat.h>
204 
205 #include <sys/syscall.h>
206 #include <sys/syscallargs.h>
207 
208 #include <sys/pax.h>
209 
210 #include <dev/clock_subr.h>
211 
212 #include <secmodel/secmodel.h>
213 
214 #include <ufs/ufs/quota.h>
215 
216 #include <miscfs/genfs/genfs.h>
217 #include <miscfs/specfs/specdev.h>
218 
219 #include <sys/cpu.h>
220 
221 #include <uvm/uvm.h>	/* extern struct uvm uvm */
222 
223 #include <dev/cons.h>
224 #include <dev/splash/splash.h>
225 
226 #include <net/bpf.h>
227 #include <net/if.h>
228 #include <net/pfil.h>
229 #include <net/raw_cb.h>
230 #include <net/if_llatbl.h>
231 
232 #include <prop/proplib.h>
233 
234 #include <sys/userconf.h>
235 
236 extern time_t rootfstime;
237 
238 #ifndef curlwp
239 struct	lwp *curlwp = &lwp0;
240 #endif
241 struct	proc *initproc;
242 
243 struct	vnode *rootvp, *swapdev_vp;
244 int	boothowto;
245 int	cold __read_mostly = 1;		/* still working on startup */
246 int	shutting_down __read_mostly;	/* system is shutting down */
247 
248 int	start_init_exec;		/* semaphore for start_init() */
249 
250 static void check_console(struct lwp *l);
251 static void start_init(void *);
252 static void configure(void);
253 static void configure2(void);
254 static void configure3(void);
255 void main(void);
256 
257 /*
258  * System startup; initialize the world, create process 0, mount root
259  * filesystem, and fork to create init and pagedaemon.  Most of the
260  * hard work is done in the lower-level initialization routines including
261  * startup(), which does memory initialization and autoconfiguration.
262  */
263 void
main(void)264 main(void)
265 {
266 	struct timespec time;
267 	struct lwp *l;
268 	struct proc *p;
269 	int s, error;
270 #ifdef NVNODE_IMPLICIT
271 	int usevnodes;
272 #endif
273 	CPU_INFO_ITERATOR cii;
274 	struct cpu_info *ci;
275 
276 #ifdef DIAGNOSTIC
277 	/*
278 	 * Verify that CPU_INFO_FOREACH() knows about the boot CPU
279 	 * and only the boot CPU at this point.
280 	 */
281 	int cpucount = 0;
282 	for (CPU_INFO_FOREACH(cii, ci)) {
283 		KASSERT(ci == curcpu());
284 		cpucount++;
285 	}
286 	KASSERT(cpucount == 1);
287 #endif
288 
289 	l = &lwp0;
290 #ifndef LWP0_CPU_INFO
291 	l->l_cpu = curcpu();
292 #endif
293 	l->l_pflag |= LP_RUNNING;
294 
295 	/*
296 	 * Attempt to find console and initialize
297 	 * in case of early panic or other messages.
298 	 */
299 	consinit();
300 #ifdef CNMAGIC
301 	cn_set_magic(CNMAGIC);
302 #endif
303 
304 	kernel_lock_init();
305 	once_init();
306 	todr_init();
307 
308 	mi_cpu_init();
309 	kernconfig_lock_init();
310 	kthread_sysinit();
311 
312 	/* Initialize the device switch tables. */
313 	devsw_init();
314 
315 	/* Initialize event counters. */
316 	evcnt_init();
317 
318 	uvm_init();
319 	ubchist_init();
320 	kcpuset_sysinit();
321 
322 	prop_kern_init();
323 
324 #if ((NKSYMS > 0) || (NDDB > 0) || (NMODULAR > 0))
325 	ksyms_init();
326 #endif
327 	kprintf_init();
328 
329 	percpu_init();
330 
331 	/* Initialize lock caches. */
332 	mutex_obj_init();
333 
334 	/* Initialize radix trees (used by numerous subsystems). */
335 	radix_tree_init();
336 
337 	/* Passive serialization. */
338 	pserialize_init();
339 
340 	/* Initialize the extent manager. */
341 	extent_init();
342 
343 	/* Do machine-dependent initialization. */
344 	cpu_startup();
345 
346 	/* Initialize the sysctl subsystem. */
347 	sysctl_init();
348 
349 	/* Initialize callouts, part 1. */
350 	callout_startup();
351 
352 	/* Initialize the kernel authorization subsystem. */
353 	kauth_init();
354 
355 	secmodel_init();
356 
357 	spec_init();
358 
359 	/*
360 	 * Set BPF op vector.  Can't do this in bpf attach, since
361 	 * network drivers attach before bpf.
362 	 */
363 	bpf_setops();
364 
365 	/* Initialize what we can in ipi(9) before CPUs are detected. */
366 	ipi_sysinit();
367 
368 	/* Start module system. */
369 	module_init();
370 	module_hook_init();
371 
372 	/*
373 	 * Initialize the kernel authorization subsystem and start the
374 	 * default security model, if any. We need to do this early
375 	 * enough so that subsystems relying on any of the aforementioned
376 	 * can work properly. Since the security model may dictate the
377 	 * credential inheritance policy, it is needed at least before
378 	 * any process is created, specifically proc0.
379 	 */
380 	module_init_class(MODULE_CLASS_SECMODEL);
381 
382 	/* Initialize the buffer cache */
383 	bufinit();
384 	biohist_init();
385 
386 #ifdef KERNHIST
387 	sysctl_kernhist_init();
388 #endif
389 
390 
391 #if defined(SPLASHSCREEN) && defined(SPLASHSCREEN_IMAGE)
392 	size_t splash_size = (&_binary_splash_image_end -
393 	    &_binary_splash_image_start) * sizeof(void *);
394 	splash_setimage(&_binary_splash_image_start, splash_size);
395 #endif
396 
397 	/* Initialize sockets. */
398 	soinit();
399 
400 	/*
401 	 * The following things must be done before autoconfiguration.
402 	 */
403 	rnd_init();		/* initialize entropy pool */
404 
405 	cprng_init();		/* initialize cryptographic PRNG */
406 
407 	/* Initialize process and pgrp structures. */
408 	procinit();
409 	lwpinit();
410 
411 	/* Must be called after lwpinit (lwpinit_specificdata) */
412 	psref_init();
413 
414 	/* Initialize signal-related data structures. */
415 	signal_init();
416 
417 	/* Initialize resource management. */
418 	resource_init();
419 
420 	/* Create process 0. */
421 	proc0_init();
422 	lwp0_init();
423 
424 	/* Disable preemption during boot. */
425 	kpreempt_disable();
426 
427 	/* Initialize the threadpool system. */
428 	threadpools_init();
429 
430 	/* Initialize the UID hash table. */
431 	uid_init();
432 
433 	/* Charge root for one process. */
434 	(void)chgproccnt(0, 1);
435 
436 	/* Initialize the run queues, turnstiles and sleep queues. */
437 	sched_rqinit();
438 	turnstile_init();
439 	sleeptab_init(&sleeptab);
440 
441 	sched_init();
442 
443 	/* Initialize processor-sets */
444 	psets_init();
445 
446 	/* Initialize cpufreq(9) */
447 	cpufreq_init();
448 
449 	/* MI initialization of the boot cpu */
450 	error = mi_cpu_attach(curcpu());
451 	KASSERT(error == 0);
452 
453 	/* Initialize timekeeping. */
454 	time_init();
455 
456 	/*
457 	 * Initialize mbuf's.  Do this now because we might attempt to
458 	 * allocate mbufs or mbuf clusters during autoconfiguration.
459 	 */
460 	mbinit();
461 
462 	/* Initialize I/O statistics. */
463 	iostat_init();
464 
465 	/* Initialize the log device. */
466 	loginit();
467 
468 	/* Second part of module system initialization. */
469 	module_start_unload_thread();
470 
471 	/* Initialize autoconf data structures before any modules are loaded */
472 	config_init_mi();
473 
474 	/* Initialize the file systems. */
475 #ifdef NVNODE_IMPLICIT
476 	/*
477 	 * If maximum number of vnodes in namei vnode cache is not explicitly
478 	 * defined in kernel config, adjust the number such as we use roughly
479 	 * 10% of memory for vnodes and associated data structures in the
480 	 * assumed worst case.  Do not provide fewer than NVNODE vnodes.
481 	 */
482 	usevnodes = calc_cache_size(vmem_size(kmem_arena, VMEM_FREE|VMEM_ALLOC),
483 	    10, VNODE_KMEM_MAXPCT) / VNODE_COST;
484 	if (usevnodes > desiredvnodes)
485 		desiredvnodes = usevnodes;
486 #endif /* NVNODE_IMPLICIT */
487 #ifdef MAXFILES_IMPLICIT
488 	/*
489 	 * If maximum number of files is not explicitly defined in
490 	 * kernel config, adjust the number so that it is somewhat
491 	 * more reasonable on machines with larger memory sizes.
492 	 * Arbitrary numbers are 20,000 files for 16GB RAM or more
493 	 * and 10,000 files for 1GB RAM or more.
494 	 *
495 	 * XXXtodo: adjust this and other values totally dynamically
496 	 */
497 	if (ctob((uint64_t)physmem) >= 16ULL * 1024 * 1024 * 1024)
498 		maxfiles = MAX(maxfiles, 20000);
499 	if (ctob((uint64_t)physmem) >= 1024 * 1024 * 1024)
500 		maxfiles = MAX(maxfiles, 10000);
501 #endif /* MAXFILES_IMPLICIT */
502 
503 	/* Initialize fstrans. */
504 	fstrans_init();
505 
506 	vfsinit();
507 	lf_init();
508 
509 	/* Initialize the file descriptor system. */
510 	fd_sys_init();
511 
512 	/* Initialize cwd structures */
513 	cwd_sys_init();
514 
515 	/* Initialize kqueue. */
516 	kqueue_init();
517 
518 	inittimecounter();
519 	ntp_init();
520 
521 	/* Initialize tty subsystem. */
522 	tty_init();
523 	ttyldisc_init();
524 
525 	/* Initialize the buffer cache, part 2. */
526 	bufinit2();
527 
528 	/* Initialize the disk wedge subsystem. */
529 	dkwedge_init();
530 
531 	/* Initialize pfil */
532 	pfil_init();
533 
534 	/* Initialize interfaces. */
535 	ifinit1();
536 
537 	spldebug_start();
538 
539 	/* Initialize sockets thread(s) */
540 	soinit1();
541 
542 	/*
543 	 * Initialize the bufq strategy sub-system and any built-in
544 	 * strategy modules - they may be needed by some devices during
545 	 * auto-configuration
546 	 */
547 	bufq_init();
548 	module_init_class(MODULE_CLASS_BUFQ);
549 
550 	/* Configure the system hardware.  This will enable interrupts. */
551 	configure();
552 #ifdef __HAVE_LEGACY_INTRCNT
553 	evcnt_attach_legacy_intrcnt();
554 #endif
555 
556 	/* Enable deferred processing of RNG samples */
557 	rnd_init_softint();
558 
559 	/* Once all CPUs are detected, initialize the per-CPU cprng_fast.  */
560 	cprng_fast_init();
561 
562 #ifdef HEARTBEAT
563 	/*
564 	 * Now that softints can be established, start monitoring
565 	 * system heartbeat on all CPUs.
566 	 */
567 	heartbeat_start();
568 #endif
569 
570 	ssp_init();
571 
572 	ubc_init();		/* must be after autoconfig */
573 
574 	mm_init();
575 
576 	configure2();
577 
578 	/* Initialize the rest of ipi(9) after CPUs have been detected. */
579 	ipi_percpu_init();
580 
581 	futex_sys_init();
582 
583 	/* Now timer is working.  Enable preemption. */
584 	kpreempt_enable();
585 
586 	/* Get the threads going and into any sleeps before continuing. */
587 	yield();
588 
589 	vmem_rehash_start();	/* must be before exec_init */
590 
591 	/* Initialize exec structures */
592 	exec_init(1);		/* seminit calls exithook_establish() */
593 
594 #if NVERIEXEC > 0
595 	/*
596 	 * Initialise the Veriexec subsystem.
597 	 */
598 	veriexec_init();
599 #endif /* NVERIEXEC > 0 */
600 
601 	pax_init();
602 
603 #ifdef	IPSEC
604 	/* Attach network crypto subsystem */
605 	ipsec_attach();
606 #endif
607 
608 	/*
609 	 * Initialize protocols.  Block reception of incoming packets
610 	 * until everything is ready.
611 	 */
612 	s = splnet();
613 	ifinit();
614 #if defined(INET) || defined(INET6)
615 	lltableinit();
616 #endif
617 	domaininit(true);
618 	ifinit_post();
619 	if_attachdomain();
620 	splx(s);
621 
622 #ifdef GPROF
623 	/* Initialize kernel profiling. */
624 	kmstartup();
625 #endif
626 
627 	/* Initialize system accounting. */
628 	acct_init();
629 
630 #ifndef PIPE_SOCKETPAIR
631 	/* Initialize pipes. */
632 	pipe_init();
633 #endif
634 
635 #ifdef KTRACE
636 	/* Initialize ktrace. */
637 	ktrinit();
638 #endif
639 
640 	machdep_init();
641 
642 	procinit_sysctl();
643 
644 	scdebug_init();
645 
646 	/*
647 	 * Create process 1 (init(8)).  We do this now, as Unix has
648 	 * historically had init be process 1, and changing this would
649 	 * probably upset a lot of people.
650 	 *
651 	 * Note that process 1 won't immediately exec init(8), but will
652 	 * wait for us to inform it that the root file system has been
653 	 * mounted.
654 	 */
655 	if (fork1(l, 0, SIGCHLD, NULL, 0, start_init, NULL, NULL))
656 		panic("fork init");
657 
658 	/*
659 	 * The initproc variable cannot be initialized in start_init as there
660 	 * is a race between vfs_mountroot and start_init.
661 	 */
662 	mutex_enter(&proc_lock);
663 	initproc = proc_find_raw(1);
664 	mutex_exit(&proc_lock);
665 
666 	/*
667 	 * Load any remaining builtin modules, and hand back temporary
668 	 * storage to the VM system.  Then require force when loading any
669 	 * remaining un-init'ed built-in modules to avoid later surprises.
670 	 */
671 	module_init_class(MODULE_CLASS_ANY);
672 	module_builtin_require_force();
673 
674 	/*
675 	 * Finalize configuration now that all real devices have been
676 	 * found.  This needs to be done before the root device is
677 	 * selected, since finalization may create the root device.
678 	 */
679 	config_finalize();
680 
681 	sysctl_finalize();
682 
683 	/*
684 	 * Now that autoconfiguration has completed, we can determine
685 	 * the root and dump devices.
686 	 */
687 	cpu_rootconf();
688 	cpu_dumpconf();
689 
690 	/* Mount the root file system. */
691 	do {
692 		domountroothook(root_device);
693 		if ((error = vfs_mountroot())) {
694 			printf("cannot mount root, error = %d\n", error);
695 			boothowto |= RB_ASKNAME;
696 			setroot(root_device,
697 			    (rootdev != NODEV) ? DISKPART(rootdev) : 0);
698 		}
699 	} while (error != 0);
700 	mountroothook_destroy();
701 
702 	configure3();
703 
704 	/*
705 	 * Initialise the time-of-day clock, passing the time recorded
706 	 * in the root filesystem (if any) for use by systems that
707 	 * don't have a non-volatile time-of-day device.
708 	 */
709 	inittodr(rootfstime);
710 
711 	/*
712 	 * Now can look at time, having had a chance to verify the time
713 	 * from the file system.  Reset l->l_rtime as it may have been
714 	 * munched in mi_switch() after the time got set.
715 	 */
716 	getnanotime(&time);
717 
718 	mutex_enter(&proc_lock);
719 	LIST_FOREACH(p, &allproc, p_list) {
720 		KASSERT((p->p_flag & PK_MARKER) == 0);
721 		mutex_enter(p->p_lock);
722 		TIMESPEC_TO_TIMEVAL(&p->p_stats->p_start, &time);
723 		LIST_FOREACH(l, &p->p_lwps, l_sibling) {
724 			lwp_lock(l);
725 			memset(&l->l_rtime, 0, sizeof(l->l_rtime));
726 			lwp_unlock(l);
727 		}
728 		mutex_exit(p->p_lock);
729 	}
730 	mutex_exit(&proc_lock);
731 	binuptime(&curlwp->l_stime);
732 
733 	for (CPU_INFO_FOREACH(cii, ci)) {
734 		ci->ci_schedstate.spc_lastmod = time_second;
735 	}
736 
737 	/* Create the pageout daemon kernel thread. */
738 	uvm_swap_init();
739 	if (kthread_create(PRI_PGDAEMON, KTHREAD_MPSAFE, NULL, uvm_pageout,
740 	    NULL, NULL, "pgdaemon"))
741 		panic("fork pagedaemon");
742 
743 	/* Create the filesystem syncer kernel thread. */
744 	if (kthread_create(PRI_IOFLUSH, KTHREAD_MPSAFE, NULL, sched_sync,
745 	    NULL, NULL, "ioflush"))
746 		panic("fork syncer");
747 
748 	/* Wait for final configure threads to complete. */
749 	config_finalize_mountroot();
750 
751 	/*
752 	 * Okay, now we can let init(8) exec!  It's off to userland!
753 	 */
754 	mutex_enter(&proc_lock);
755 	start_init_exec = 1;
756 	cv_broadcast(&lbolt);
757 	mutex_exit(&proc_lock);
758 
759 	/* The scheduler is an infinite loop. */
760 	uvm_scheduler();
761 	/* NOTREACHED */
762 }
763 
764 /*
765  * Configure the system's hardware.
766  */
767 static void
configure(void)768 configure(void)
769 {
770 
771 	/*
772 	 * XXX
773 	 * callout_setfunc() requires mutex(9) so it can't be in config_init()
774 	 * on amiga and atari which use config_init() and autoconf(9) functions
775 	 * to initialize console devices.
776 	 */
777 	config_twiddle_init();
778 
779 	pmf_init();
780 
781 	/* Initialize driver modules */
782 	module_init_class(MODULE_CLASS_DRIVER);
783 
784 	userconf_init();
785 	if (boothowto & RB_USERCONF)
786 		userconf_prompt();
787 
788 	if ((boothowto & (AB_SILENT|AB_VERBOSE)) == AB_SILENT) {
789 		printf_nolog("Detecting hardware...");
790 	}
791 
792 	/*
793 	 * Do the machine-dependent portion of autoconfiguration.  This
794 	 * sets the configuration machinery here in motion by "finding"
795 	 * the root bus.  When this function returns, we expect interrupts
796 	 * to be enabled.
797 	 */
798 	cpu_configure();
799 }
800 
801 static void
configure2(void)802 configure2(void)
803 {
804 	CPU_INFO_ITERATOR cii;
805 	struct cpu_info *ci;
806 	int s;
807 
808 	/* Fix up CPU topology info, which has all been collected by now. */
809 	cpu_topology_init();
810 
811 	/*
812 	 * Now that we've found all the hardware, start the real time
813 	 * and statistics clocks.
814 	 */
815 	initclocks();
816 
817 	cold = 0;	/* clocks are running, we're warm now! */
818 	s = splsched();
819 	curcpu()->ci_schedstate.spc_flags |= SPCF_RUNNING;
820 	splx(s);
821 
822 	/* Setup the runqueues and scheduler. */
823 	runq_init();
824 	synch_init();
825 
826 	/* Boot the secondary processors. */
827 	for (CPU_INFO_FOREACH(cii, ci)) {
828 		uvm_cpu_attach(ci);
829 	}
830 
831 	/* Decide how to partition free memory. */
832 	uvm_page_rebucket();
833 
834 	mp_online = true;
835 #if defined(MULTIPROCESSOR)
836 	cpu_boot_secondary_processors();
837 #endif
838 
839 	/*
840 	 * Bus scans can make it appear as if the system has paused, so
841 	 * twiddle constantly while config_interrupts() jobs are running.
842 	 */
843 	config_twiddle_fn(NULL);
844 
845 	/*
846 	 * Create threads to call back and finish configuration for
847 	 * devices that want interrupts enabled.
848 	 */
849 	config_create_interruptthreads();
850 }
851 
852 static void
configure3(void)853 configure3(void)
854 {
855 
856 	/*
857 	 * Create threads to call back and finish configuration for
858 	 * devices that want the mounted root file system.
859 	 */
860 	config_create_mountrootthreads();
861 
862 	/* Get the threads going and into any sleeps before continuing. */
863 	yield();
864 }
865 
866 static void
rootconf_handle_wedges(void)867 rootconf_handle_wedges(void)
868 {
869 	struct disklabel label;
870 	struct partition *p;
871 	struct vnode *vp;
872 	daddr_t startblk;
873 	uint64_t nblks;
874 	device_t dev;
875 	int error;
876 
877 	if (booted_nblks) {
878 		/*
879 		 * bootloader passed geometry
880 		 */
881 		dev      = booted_device;
882 		startblk = booted_startblk;
883 		nblks    = booted_nblks;
884 
885 		/*
886 		 * keep booted_device and booted_partition
887 		 * in case the kernel doesn't identify a wedge
888 		 */
889 	} else {
890 		/*
891 		 * bootloader passed partition number
892 		 *
893 		 * We cannot ask the partition device directly when it is
894 		 * covered by a wedge. Instead we look up the geometry in
895 		 * the disklabel.
896 		 */
897 		vp = opendisk(booted_device);
898 
899 		if (vp == NULL)
900 			return;
901 
902 		VOP_UNLOCK(vp);
903 		error = VOP_IOCTL(vp, DIOCGDINFO, &label, FREAD, NOCRED);
904 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
905 		VOP_CLOSE(vp, FREAD, NOCRED);
906 		vput(vp);
907 		if (error)
908 			return;
909 
910 		KASSERT(booted_partition >= 0
911 			&& booted_partition < MAXPARTITIONS);
912 
913 		p = &label.d_partitions[booted_partition];
914 
915 		dev      = booted_device;
916 		startblk = p->p_offset;
917 		nblks    = p->p_size;
918 	}
919 
920 	dev = dkwedge_find_partition(dev, startblk, nblks);
921 	if (dev != NULL) {
922 		booted_device = dev;
923 		booted_partition = 0;
924 	}
925 }
926 
927 void
rootconf(void)928 rootconf(void)
929 {
930 	if (booted_device != NULL)
931 		rootconf_handle_wedges();
932 
933 	setroot(booted_device, booted_partition);
934 }
935 
936 static void
check_console(struct lwp * l)937 check_console(struct lwp *l)
938 {
939 	struct vnode *vp;
940 	int error;
941 
942 	error = namei_simple_kernel("/dev/console",
943 				NSM_FOLLOW_NOEMULROOT, &vp);
944 	if (error == 0) {
945 		vrele(vp);
946 	} else if (error == ENOENT) {
947 		if (boothowto & (AB_VERBOSE|AB_DEBUG))
948 			printf("warning: no /dev/console\n");
949 	} else {
950 		printf("warning: lookup /dev/console: error %d\n", error);
951 	}
952 }
953 
954 /*
955  * List of paths to try when searching for "init".
956  */
957 static const char * const initpaths[] = {
958 	"/sbin/init",
959 	"/sbin/oinit",
960 	"/sbin/init.bak",
961 	"/rescue/init",
962 	NULL,
963 };
964 
965 /*
966  * Start the initial user process; try exec'ing each pathname in "initpaths".
967  * The program is invoked with one argument containing the boot flags.
968  */
969 static void
start_init(void * arg)970 start_init(void *arg)
971 {
972 	struct lwp *l = arg;
973 	struct proc *p = l->l_proc;
974 	vaddr_t addr;
975 	struct sys_execve_args /* {
976 		syscallarg(const char *) path;
977 		syscallarg(char * const *) argp;
978 		syscallarg(char * const *) envp;
979 	} */ args;
980 	int options, i, error;
981 	register_t retval[2];
982 	char flags[4], *flagsp;
983 	const char *path, *slash;
984 	char *ucp, **uap, *arg0, *arg1, *argv[3];
985 	char ipath[129];
986 	int ipx, len;
987 
988 	/*
989 	 * Now in process 1.
990 	 */
991 	strncpy(p->p_comm, "init", MAXCOMLEN);
992 
993 	/*
994 	 * Wait for main() to tell us that it's safe to exec.
995 	 */
996 	mutex_enter(&proc_lock);
997 	while (start_init_exec == 0)
998 		cv_wait(&lbolt, &proc_lock);
999 	mutex_exit(&proc_lock);
1000 
1001 	/*
1002 	 * This is not the right way to do this.  We really should
1003 	 * hand-craft a descriptor onto /dev/console to hand to init,
1004 	 * but that's a _lot_ more work, and the benefit from this easy
1005 	 * hack makes up for the "good is the enemy of the best" effect.
1006 	 */
1007 	check_console(l);
1008 
1009 	/*
1010 	 * Need just enough stack to hold the faked-up "execve()" arguments.
1011 	 */
1012 	addr = (vaddr_t)STACK_ALLOC(USRSTACK, PAGE_SIZE);
1013 	if (uvm_map(&p->p_vmspace->vm_map, &addr, PAGE_SIZE,
1014 	    NULL, UVM_UNKNOWN_OFFSET, 0,
1015 	    UVM_MAPFLAG(UVM_PROT_RW, UVM_PROT_RW, UVM_INH_COPY,
1016 	    UVM_ADV_NORMAL,
1017 	    UVM_FLAG_FIXED|UVM_FLAG_OVERLAY|UVM_FLAG_COPYONW)) != 0)
1018 		panic("init: couldn't allocate argument space");
1019 	p->p_vmspace->vm_maxsaddr = (void *)STACK_MAX(addr, PAGE_SIZE);
1020 
1021 	ipx = 0;
1022 	while (1) {
1023 		if (boothowto & RB_ASKNAME) {
1024 			printf("init path");
1025 			if (initpaths[ipx])
1026 				printf(" (default %s)", initpaths[ipx]);
1027 			printf(": ");
1028 			len = cngetsn(ipath, sizeof(ipath)-1);
1029 			if (len == 4 && strcmp(ipath, "halt") == 0) {
1030 				kern_reboot(RB_HALT, NULL);
1031 			} else if (len == 6 && strcmp(ipath, "reboot") == 0) {
1032 				kern_reboot(0, NULL);
1033 #if defined(DDB)
1034 			} else if (len == 3 && strcmp(ipath, "ddb") == 0) {
1035 				console_debugger();
1036 				continue;
1037 #endif
1038 			} else if (len > 0 && ipath[0] == '/') {
1039 				ipath[len] = '\0';
1040 				path = ipath;
1041 			} else if (len == 0 && initpaths[ipx] != NULL) {
1042 				path = initpaths[ipx++];
1043 			} else {
1044 				printf("use absolute path, ");
1045 #if defined(DDB)
1046 				printf("\"ddb\", ");
1047 #endif
1048 				printf("\"halt\", or \"reboot\"\n");
1049 				continue;
1050 			}
1051 		} else {
1052 			if ((path = initpaths[ipx++]) == NULL) {
1053 				ipx = 0;
1054 				boothowto |= RB_ASKNAME;
1055 				continue;
1056 			}
1057 		}
1058 
1059 		ucp = (char *)USRSTACK;
1060 
1061 		/*
1062 		 * Construct the boot flag argument.
1063 		 */
1064 		flagsp = flags;
1065 		*flagsp++ = '-';
1066 		options = 0;
1067 
1068 		if (boothowto & RB_SINGLE) {
1069 			*flagsp++ = 's';
1070 			options = 1;
1071 		}
1072 #ifdef notyet
1073 		if (boothowto & RB_FASTBOOT) {
1074 			*flagsp++ = 'f';
1075 			options = 1;
1076 		}
1077 #endif
1078 
1079 		/*
1080 		 * Move out the flags (arg 1), if necessary.
1081 		 */
1082 		if (options != 0) {
1083 			*flagsp++ = '\0';
1084 			i = flagsp - flags;
1085 #ifdef DEBUG
1086 			aprint_normal("init: copying out flags `%s' %d\n", flags, i);
1087 #endif
1088 			arg1 = STACK_ALLOC(ucp, i);
1089 			ucp = STACK_MAX(arg1, i);
1090 			if ((error = copyout((void *)flags, arg1, i)) != 0)
1091 				goto copyerr;
1092 		} else
1093 			arg1 = NULL;
1094 
1095 		/*
1096 		 * Move out the file name (also arg 0).
1097 		 */
1098 		i = strlen(path) + 1;
1099 #ifdef DEBUG
1100 		aprint_normal("init: copying out path `%s' %d\n", path, i);
1101 #else
1102 		if (boothowto & RB_ASKNAME || path != initpaths[0])
1103 			printf("init: trying %s\n", path);
1104 #endif
1105 		arg0 = STACK_ALLOC(ucp, i);
1106 		ucp = STACK_MAX(arg0, i);
1107 		if ((error = copyout(path, arg0, i)) != 0)
1108 			goto copyerr;
1109 
1110 		/*
1111 		 * Move out the arg pointers.
1112 		 */
1113 		ucp = (void *)STACK_ALIGN(ucp, STACK_ALIGNBYTES);
1114 		uap = (char **)STACK_ALLOC(ucp, sizeof(argv));
1115 		SCARG(&args, path) = arg0;
1116 		SCARG(&args, argp) = uap;
1117 		SCARG(&args, envp) = NULL;
1118 		slash = strrchr(path, '/');
1119 
1120 		argv[0] = slash ? arg0 + (slash + 1 - path) : arg0;
1121 		argv[1] = arg1;
1122 		argv[2] = NULL;
1123 		if ((error = copyout(argv, uap, sizeof(argv))) != 0)
1124 			goto copyerr;
1125 
1126 		/*
1127 		 * Now try to exec the program.  If it can't for any reason
1128 		 * other than it doesn't exist, complain.
1129 		 */
1130 		error = sys_execve(l, &args, retval);
1131 		if (error == 0 || error == EJUSTRETURN) {
1132 			KERNEL_UNLOCK_LAST(l);
1133 			return;
1134 		}
1135 		printf("exec %s: error %d\n", path, error);
1136 	}
1137 	printf("init: not found\n");
1138 	panic("no init");
1139 copyerr:
1140 	panic("copyout %d", error);
1141 }
1142 
1143 /*
1144  * calculate cache size (in bytes) from physmem and vsize.
1145  */
1146 vaddr_t
calc_cache_size(vsize_t vsize,int pct,int va_pct)1147 calc_cache_size(vsize_t vsize, int pct, int va_pct)
1148 {
1149 	paddr_t t;
1150 
1151 	/* XXX should consider competing cache if any */
1152 	/* XXX should consider submaps */
1153 	t = (uintmax_t)physmem * pct / 100 * PAGE_SIZE;
1154 	if (vsize != 0) {
1155 		vsize = (uintmax_t)vsize * va_pct / 100;
1156 		if (t > vsize) {
1157 			t = vsize;
1158 		}
1159 	}
1160 	return t;
1161 }
1162 
1163 /*
1164  * Print the system start up banner.
1165  *
1166  * - Print a limited banner if AB_SILENT.
1167  * - Always send normal banner to the log.
1168  */
1169 #define MEM_PBUFSIZE	sizeof("99999 MB")
1170 
1171 void
banner(void)1172 banner(void)
1173 {
1174 	static char notice[] = " Notice: this software is "
1175 	    "protected by copyright";
1176 	char pbuf[81];
1177 	void (*pr)(const char *, ...) __printflike(1, 2);
1178 	int i;
1179 
1180 	if ((boothowto & AB_SILENT) != 0) {
1181 		snprintf(pbuf, sizeof(pbuf), "%s %s (%s)",
1182 		    ostype, osrelease, kernel_ident);
1183 		printf_nolog("%s", pbuf);
1184 		for (i = 80 - strlen(pbuf) - sizeof(notice); i > 0; i--)
1185 			printf(" ");
1186 		printf_nolog("%s\n", notice);
1187 		pr = aprint_normal;
1188 	} else {
1189 		pr = printf;
1190 	}
1191 
1192 	memset(pbuf, 0, sizeof(pbuf));
1193 	(*pr)("%s%s", copyright, version);
1194 	format_bytes(pbuf, MEM_PBUFSIZE, ctob((uint64_t)physmem));
1195 	(*pr)("total memory = %s\n", pbuf);
1196 	format_bytes(pbuf, MEM_PBUFSIZE, ctob((uint64_t)uvm_availmem(false)));
1197 	(*pr)("avail memory = %s\n", pbuf);
1198 }
1199