xref: /freebsd/sys/kern/init_main.c (revision 2c48331d28f16c0efce5a72a81e7d71668c4a158)
1 /*-
2  * SPDX-License-Identifier: BSD-4-Clause
3  *
4  * Copyright (c) 1995 Terrence R. Lambert
5  * All rights reserved.
6  *
7  * Copyright (c) 1982, 1986, 1989, 1991, 1992, 1993
8  *	The Regents of the University of California.  All rights reserved.
9  * (c) UNIX System Laboratories, Inc.
10  * All or some portions of this file are derived from material licensed
11  * to the University of California by American Telephone and Telegraph
12  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
13  * the permission of UNIX System Laboratories, Inc.
14  *
15  * Redistribution and use in source and binary forms, with or without
16  * modification, are permitted provided that the following conditions
17  * are met:
18  * 1. Redistributions of source code must retain the above copyright
19  *    notice, this list of conditions and the following disclaimer.
20  * 2. Redistributions in binary form must reproduce the above copyright
21  *    notice, this list of conditions and the following disclaimer in the
22  *    documentation and/or other materials provided with the distribution.
23  * 3. All advertising materials mentioning features or use of this software
24  *    must display the following acknowledgement:
25  *	This product includes software developed by the University of
26  *	California, Berkeley and its contributors.
27  * 4. Neither the name of the University nor the names of its contributors
28  *    may be used to endorse or promote products derived from this software
29  *    without specific prior written permission.
30  *
31  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
32  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
33  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
34  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
35  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
36  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
37  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
38  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
39  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
40  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
41  * SUCH DAMAGE.
42  *
43  *	@(#)init_main.c	8.9 (Berkeley) 1/21/94
44  */
45 
46 #include <sys/cdefs.h>
47 __FBSDID("$FreeBSD$");
48 
49 #include "opt_ddb.h"
50 #include "opt_init_path.h"
51 #include "opt_verbose_sysinit.h"
52 
53 #include <sys/param.h>
54 #include <sys/kernel.h>
55 #include <sys/epoch.h>
56 #include <sys/eventhandler.h>
57 #include <sys/exec.h>
58 #include <sys/file.h>
59 #include <sys/filedesc.h>
60 #include <sys/imgact.h>
61 #include <sys/jail.h>
62 #include <sys/ktr.h>
63 #include <sys/lock.h>
64 #include <sys/loginclass.h>
65 #include <sys/mount.h>
66 #include <sys/mutex.h>
67 #include <sys/syscallsubr.h>
68 #include <sys/sysctl.h>
69 #include <sys/proc.h>
70 #include <sys/racct.h>
71 #include <sys/resourcevar.h>
72 #include <sys/systm.h>
73 #include <sys/signalvar.h>
74 #include <sys/vnode.h>
75 #include <sys/sysent.h>
76 #include <sys/reboot.h>
77 #include <sys/sched.h>
78 #include <sys/sx.h>
79 #include <sys/sysproto.h>
80 #include <sys/vmmeter.h>
81 #include <sys/unistd.h>
82 #include <sys/malloc.h>
83 #include <sys/conf.h>
84 #include <sys/cpuset.h>
85 
86 #include <machine/cpu.h>
87 
88 #include <security/audit/audit.h>
89 #include <security/mac/mac_framework.h>
90 
91 #include <vm/vm.h>
92 #include <vm/vm_param.h>
93 #include <vm/vm_extern.h>
94 #include <vm/pmap.h>
95 #include <vm/vm_map.h>
96 #include <sys/copyright.h>
97 
98 #include <ddb/ddb.h>
99 #include <ddb/db_sym.h>
100 
101 void mi_startup(void);				/* Should be elsewhere */
102 
103 /* Components of the first process -- never freed. */
104 static struct session session0;
105 static struct pgrp pgrp0;
106 struct	proc proc0;
107 struct thread0_storage thread0_st __aligned(32);
108 struct	vmspace vmspace0;
109 struct	proc *initproc;
110 
111 int
112 linux_alloc_current_noop(struct thread *td __unused, int flags __unused)
113 {
114 	return (0);
115 }
116 int (*lkpi_alloc_current)(struct thread *, int) = linux_alloc_current_noop;
117 
118 #ifndef BOOTHOWTO
119 #define	BOOTHOWTO	0
120 #endif
121 int	boothowto = BOOTHOWTO;	/* initialized so that it can be patched */
122 SYSCTL_INT(_debug, OID_AUTO, boothowto, CTLFLAG_RD, &boothowto, 0,
123 	"Boot control flags, passed from loader");
124 
125 #ifndef BOOTVERBOSE
126 #define	BOOTVERBOSE	0
127 #endif
128 int	bootverbose = BOOTVERBOSE;
129 SYSCTL_INT(_debug, OID_AUTO, bootverbose, CTLFLAG_RW, &bootverbose, 0,
130 	"Control the output of verbose kernel messages");
131 
132 #ifdef VERBOSE_SYSINIT
133 /*
134  * We'll use the defined value of VERBOSE_SYSINIT from the kernel config to
135  * dictate the default VERBOSE_SYSINIT behavior.  Significant values for this
136  * option and associated tunable are:
137  * - 0, 'compiled in but silent by default'
138  * - 1, 'compiled in but verbose by default' (default)
139  */
140 int	verbose_sysinit = VERBOSE_SYSINIT;
141 TUNABLE_INT("debug.verbose_sysinit", &verbose_sysinit);
142 #endif
143 
144 #ifdef INVARIANTS
145 FEATURE(invariants, "Kernel compiled with INVARIANTS, may affect performance");
146 #endif
147 
148 /*
149  * This ensures that there is at least one entry so that the sysinit_set
150  * symbol is not undefined.  A sybsystem ID of SI_SUB_DUMMY is never
151  * executed.
152  */
153 SYSINIT(placeholder, SI_SUB_DUMMY, SI_ORDER_ANY, NULL, NULL);
154 
155 /*
156  * The sysinit table itself.  Items are checked off as the are run.
157  * If we want to register new sysinit types, add them to newsysinit.
158  */
159 SET_DECLARE(sysinit_set, struct sysinit);
160 struct sysinit **sysinit, **sysinit_end;
161 struct sysinit **newsysinit, **newsysinit_end;
162 
163 /*
164  * Merge a new sysinit set into the current set, reallocating it if
165  * necessary.  This can only be called after malloc is running.
166  */
167 void
168 sysinit_add(struct sysinit **set, struct sysinit **set_end)
169 {
170 	struct sysinit **newset;
171 	struct sysinit **sipp;
172 	struct sysinit **xipp;
173 	int count;
174 
175 	count = set_end - set;
176 	if (newsysinit)
177 		count += newsysinit_end - newsysinit;
178 	else
179 		count += sysinit_end - sysinit;
180 	newset = malloc(count * sizeof(*sipp), M_TEMP, M_NOWAIT);
181 	if (newset == NULL)
182 		panic("cannot malloc for sysinit");
183 	xipp = newset;
184 	if (newsysinit)
185 		for (sipp = newsysinit; sipp < newsysinit_end; sipp++)
186 			*xipp++ = *sipp;
187 	else
188 		for (sipp = sysinit; sipp < sysinit_end; sipp++)
189 			*xipp++ = *sipp;
190 	for (sipp = set; sipp < set_end; sipp++)
191 		*xipp++ = *sipp;
192 	if (newsysinit)
193 		free(newsysinit, M_TEMP);
194 	newsysinit = newset;
195 	newsysinit_end = newset + count;
196 }
197 
198 #if defined (DDB) && defined(VERBOSE_SYSINIT)
199 static const char *
200 symbol_name(vm_offset_t va, db_strategy_t strategy)
201 {
202 	const char *name;
203 	c_db_sym_t sym;
204 	db_expr_t  offset;
205 
206 	if (va == 0)
207 		return (NULL);
208 	sym = db_search_symbol(va, strategy, &offset);
209 	if (offset != 0)
210 		return (NULL);
211 	db_symbol_values(sym, &name, NULL);
212 	return (name);
213 }
214 #endif
215 
216 /*
217  * System startup; initialize the world, create process 0, mount root
218  * filesystem, and fork to create init and pagedaemon.  Most of the
219  * hard work is done in the lower-level initialization routines including
220  * startup(), which does memory initialization and autoconfiguration.
221  *
222  * This allows simple addition of new kernel subsystems that require
223  * boot time initialization.  It also allows substitution of subsystem
224  * (for instance, a scheduler, kernel profiler, or VM system) by object
225  * module.  Finally, it allows for optional "kernel threads".
226  */
227 void
228 mi_startup(void)
229 {
230 
231 	struct sysinit **sipp;	/* system initialization*/
232 	struct sysinit **xipp;	/* interior loop of sort*/
233 	struct sysinit *save;	/* bubble*/
234 
235 #if defined(VERBOSE_SYSINIT)
236 	int last;
237 	int verbose;
238 #endif
239 
240 	TSENTER();
241 
242 	if (boothowto & RB_VERBOSE)
243 		bootverbose++;
244 
245 	if (sysinit == NULL) {
246 		sysinit = SET_BEGIN(sysinit_set);
247 		sysinit_end = SET_LIMIT(sysinit_set);
248 	}
249 
250 restart:
251 	/*
252 	 * Perform a bubble sort of the system initialization objects by
253 	 * their subsystem (primary key) and order (secondary key).
254 	 */
255 	for (sipp = sysinit; sipp < sysinit_end; sipp++) {
256 		for (xipp = sipp + 1; xipp < sysinit_end; xipp++) {
257 			if ((*sipp)->subsystem < (*xipp)->subsystem ||
258 			     ((*sipp)->subsystem == (*xipp)->subsystem &&
259 			      (*sipp)->order <= (*xipp)->order))
260 				continue;	/* skip*/
261 			save = *sipp;
262 			*sipp = *xipp;
263 			*xipp = save;
264 		}
265 	}
266 
267 #if defined(VERBOSE_SYSINIT)
268 	last = SI_SUB_COPYRIGHT;
269 	verbose = 0;
270 #if !defined(DDB)
271 	printf("VERBOSE_SYSINIT: DDB not enabled, symbol lookups disabled.\n");
272 #endif
273 #endif
274 
275 	/*
276 	 * Traverse the (now) ordered list of system initialization tasks.
277 	 * Perform each task, and continue on to the next task.
278 	 */
279 	for (sipp = sysinit; sipp < sysinit_end; sipp++) {
280 		if ((*sipp)->subsystem == SI_SUB_DUMMY)
281 			continue;	/* skip dummy task(s)*/
282 
283 		if ((*sipp)->subsystem == SI_SUB_DONE)
284 			continue;
285 
286 #if defined(VERBOSE_SYSINIT)
287 		if ((*sipp)->subsystem > last && verbose_sysinit != 0) {
288 			verbose = 1;
289 			last = (*sipp)->subsystem;
290 			printf("subsystem %x\n", last);
291 		}
292 		if (verbose) {
293 #if defined(DDB)
294 			const char *func, *data;
295 
296 			func = symbol_name((vm_offset_t)(*sipp)->func,
297 			    DB_STGY_PROC);
298 			data = symbol_name((vm_offset_t)(*sipp)->udata,
299 			    DB_STGY_ANY);
300 			if (func != NULL && data != NULL)
301 				printf("   %s(&%s)... ", func, data);
302 			else if (func != NULL)
303 				printf("   %s(%p)... ", func, (*sipp)->udata);
304 			else
305 #endif
306 				printf("   %p(%p)... ", (*sipp)->func,
307 				    (*sipp)->udata);
308 		}
309 #endif
310 
311 		/* Call function */
312 		(*((*sipp)->func))((*sipp)->udata);
313 
314 #if defined(VERBOSE_SYSINIT)
315 		if (verbose)
316 			printf("done.\n");
317 #endif
318 
319 		/* Check off the one we're just done */
320 		(*sipp)->subsystem = SI_SUB_DONE;
321 
322 		/* Check if we've installed more sysinit items via KLD */
323 		if (newsysinit != NULL) {
324 			if (sysinit != SET_BEGIN(sysinit_set))
325 				free(sysinit, M_TEMP);
326 			sysinit = newsysinit;
327 			sysinit_end = newsysinit_end;
328 			newsysinit = NULL;
329 			newsysinit_end = NULL;
330 			goto restart;
331 		}
332 	}
333 
334 	TSEXIT();	/* Here so we don't overlap with start_init. */
335 
336 	mtx_assert(&Giant, MA_OWNED | MA_NOTRECURSED);
337 	mtx_unlock(&Giant);
338 
339 	/*
340 	 * Now hand over this thread to swapper.
341 	 */
342 	swapper();
343 	/* NOTREACHED*/
344 }
345 
346 static void
347 print_caddr_t(void *data)
348 {
349 	printf("%s", (char *)data);
350 }
351 
352 static void
353 print_version(void *data __unused)
354 {
355 	int len;
356 
357 	/* Strip a trailing newline from version. */
358 	len = strlen(version);
359 	while (len > 0 && version[len - 1] == '\n')
360 		len--;
361 	printf("%.*s %s\n", len, version, machine);
362 	printf("%s\n", compiler_version);
363 }
364 
365 SYSINIT(announce, SI_SUB_COPYRIGHT, SI_ORDER_FIRST, print_caddr_t,
366     copyright);
367 SYSINIT(trademark, SI_SUB_COPYRIGHT, SI_ORDER_SECOND, print_caddr_t,
368     trademark);
369 SYSINIT(version, SI_SUB_COPYRIGHT, SI_ORDER_THIRD, print_version, NULL);
370 
371 #ifdef WITNESS
372 static char wit_warn[] =
373      "WARNING: WITNESS option enabled, expect reduced performance.\n";
374 SYSINIT(witwarn, SI_SUB_COPYRIGHT, SI_ORDER_FOURTH,
375    print_caddr_t, wit_warn);
376 SYSINIT(witwarn2, SI_SUB_LAST, SI_ORDER_FOURTH,
377    print_caddr_t, wit_warn);
378 #endif
379 
380 #ifdef DIAGNOSTIC
381 static char diag_warn[] =
382      "WARNING: DIAGNOSTIC option enabled, expect reduced performance.\n";
383 SYSINIT(diagwarn, SI_SUB_COPYRIGHT, SI_ORDER_FIFTH,
384     print_caddr_t, diag_warn);
385 SYSINIT(diagwarn2, SI_SUB_LAST, SI_ORDER_FIFTH,
386     print_caddr_t, diag_warn);
387 #endif
388 
389 static int
390 null_fetch_syscall_args(struct thread *td __unused)
391 {
392 
393 	panic("null_fetch_syscall_args");
394 }
395 
396 static void
397 null_set_syscall_retval(struct thread *td __unused, int error __unused)
398 {
399 
400 	panic("null_set_syscall_retval");
401 }
402 
403 struct sysentvec null_sysvec = {
404 	.sv_size	= 0,
405 	.sv_table	= NULL,
406 	.sv_transtrap	= NULL,
407 	.sv_fixup	= NULL,
408 	.sv_sendsig	= NULL,
409 	.sv_sigcode	= NULL,
410 	.sv_szsigcode	= NULL,
411 	.sv_name	= "null",
412 	.sv_coredump	= NULL,
413 	.sv_imgact_try	= NULL,
414 	.sv_minsigstksz	= 0,
415 	.sv_minuser	= VM_MIN_ADDRESS,
416 	.sv_maxuser	= VM_MAXUSER_ADDRESS,
417 	.sv_usrstack	= USRSTACK,
418 	.sv_psstrings	= PS_STRINGS,
419 	.sv_stackprot	= VM_PROT_ALL,
420 	.sv_copyout_strings	= NULL,
421 	.sv_setregs	= NULL,
422 	.sv_fixlimit	= NULL,
423 	.sv_maxssiz	= NULL,
424 	.sv_flags	= 0,
425 	.sv_set_syscall_retval = null_set_syscall_retval,
426 	.sv_fetch_syscall_args = null_fetch_syscall_args,
427 	.sv_syscallnames = NULL,
428 	.sv_schedtail	= NULL,
429 	.sv_thread_detach = NULL,
430 	.sv_trap	= NULL,
431 };
432 
433 /*
434  * The two following SYSINIT's are proc0 specific glue code.  I am not
435  * convinced that they can not be safely combined, but their order of
436  * operation has been maintained as the same as the original init_main.c
437  * for right now.
438  */
439 /* ARGSUSED*/
440 static void
441 proc0_init(void *dummy __unused)
442 {
443 	struct proc *p;
444 	struct thread *td;
445 	struct ucred *newcred;
446 	struct uidinfo tmpuinfo;
447 	struct loginclass tmplc = {
448 		.lc_name = "",
449 	};
450 	vm_paddr_t pageablemem;
451 	int i;
452 
453 	GIANT_REQUIRED;
454 	p = &proc0;
455 	td = &thread0;
456 
457 	/*
458 	 * Initialize magic number and osrel.
459 	 */
460 	p->p_magic = P_MAGIC;
461 	p->p_osrel = osreldate;
462 
463 	/*
464 	 * Initialize thread and process structures.
465 	 */
466 	procinit();	/* set up proc zone */
467 	threadinit();	/* set up UMA zones */
468 
469 	/*
470 	 * Initialise scheduler resources.
471 	 * Add scheduler specific parts to proc, thread as needed.
472 	 */
473 	schedinit();	/* scheduler gets its house in order */
474 
475 	/*
476 	 * Create process 0 (the swapper).
477 	 */
478 	LIST_INSERT_HEAD(&allproc, p, p_list);
479 	LIST_INSERT_HEAD(PIDHASH(0), p, p_hash);
480 	mtx_init(&pgrp0.pg_mtx, "process group", NULL, MTX_DEF | MTX_DUPOK);
481 	p->p_pgrp = &pgrp0;
482 	LIST_INSERT_HEAD(PGRPHASH(0), &pgrp0, pg_hash);
483 	LIST_INIT(&pgrp0.pg_members);
484 	LIST_INSERT_HEAD(&pgrp0.pg_members, p, p_pglist);
485 
486 	pgrp0.pg_session = &session0;
487 	mtx_init(&session0.s_mtx, "session", NULL, MTX_DEF);
488 	refcount_init(&session0.s_count, 1);
489 	session0.s_leader = p;
490 
491 	p->p_sysent = &null_sysvec;
492 	p->p_flag = P_SYSTEM | P_INMEM | P_KPROC;
493 	p->p_flag2 = 0;
494 	p->p_state = PRS_NORMAL;
495 	p->p_klist = knlist_alloc(&p->p_mtx);
496 	STAILQ_INIT(&p->p_ktr);
497 	p->p_nice = NZERO;
498 	/* pid_max cannot be greater than PID_MAX */
499 	td->td_tid = PID_MAX + 1;
500 	LIST_INSERT_HEAD(TIDHASH(td->td_tid), td, td_hash);
501 	td->td_state = TDS_RUNNING;
502 	td->td_pri_class = PRI_TIMESHARE;
503 	td->td_user_pri = PUSER;
504 	td->td_base_user_pri = PUSER;
505 	td->td_lend_user_pri = PRI_MAX;
506 	td->td_priority = PVM;
507 	td->td_base_pri = PVM;
508 	td->td_oncpu = curcpu;
509 	td->td_flags = TDF_INMEM;
510 	td->td_pflags = TDP_KTHREAD;
511 	td->td_cpuset = cpuset_thread0();
512 	td->td_domain.dr_policy = td->td_cpuset->cs_domain;
513 	prison0_init();
514 	p->p_peers = 0;
515 	p->p_leader = p;
516 	p->p_reaper = p;
517 	p->p_treeflag |= P_TREE_REAPER;
518 	LIST_INIT(&p->p_reaplist);
519 
520 	strncpy(p->p_comm, "kernel", sizeof (p->p_comm));
521 	strncpy(td->td_name, "swapper", sizeof (td->td_name));
522 
523 	callout_init_mtx(&p->p_itcallout, &p->p_mtx, 0);
524 	callout_init_mtx(&p->p_limco, &p->p_mtx, 0);
525 	callout_init(&td->td_slpcallout, 1);
526 
527 	/* Create credentials. */
528 	newcred = crget();
529 	newcred->cr_ngroups = 1;	/* group 0 */
530 	/* A hack to prevent uifind from tripping over NULL pointers. */
531 	curthread->td_ucred = newcred;
532 	tmpuinfo.ui_uid = 1;
533 	newcred->cr_uidinfo = newcred->cr_ruidinfo = &tmpuinfo;
534 	newcred->cr_uidinfo = uifind(0);
535 	newcred->cr_ruidinfo = uifind(0);
536 	newcred->cr_loginclass = &tmplc;
537 	newcred->cr_loginclass = loginclass_find("default");
538 	/* End hack. creds get properly set later with thread_cow_get_proc */
539 	curthread->td_ucred = NULL;
540 	newcred->cr_prison = &prison0;
541 	newcred->cr_users++; /* avoid assertion failure */
542 	proc_set_cred_init(p, newcred);
543 	newcred->cr_users--;
544 	crfree(newcred);
545 #ifdef AUDIT
546 	audit_cred_kproc0(newcred);
547 #endif
548 #ifdef MAC
549 	mac_cred_create_swapper(newcred);
550 #endif
551 	/* Create sigacts. */
552 	p->p_sigacts = sigacts_alloc();
553 
554 	/* Initialize signal state for process 0. */
555 	siginit(&proc0);
556 
557 	/* Create the file descriptor table. */
558 	p->p_fd = fdinit(NULL, false, NULL);
559 	p->p_fdtol = NULL;
560 
561 	/* Create the limits structures. */
562 	p->p_limit = lim_alloc();
563 	for (i = 0; i < RLIM_NLIMITS; i++)
564 		p->p_limit->pl_rlimit[i].rlim_cur =
565 		    p->p_limit->pl_rlimit[i].rlim_max = RLIM_INFINITY;
566 	p->p_limit->pl_rlimit[RLIMIT_NOFILE].rlim_cur =
567 	    p->p_limit->pl_rlimit[RLIMIT_NOFILE].rlim_max = maxfiles;
568 	p->p_limit->pl_rlimit[RLIMIT_NPROC].rlim_cur =
569 	    p->p_limit->pl_rlimit[RLIMIT_NPROC].rlim_max = maxproc;
570 	p->p_limit->pl_rlimit[RLIMIT_DATA].rlim_cur = dfldsiz;
571 	p->p_limit->pl_rlimit[RLIMIT_DATA].rlim_max = maxdsiz;
572 	p->p_limit->pl_rlimit[RLIMIT_STACK].rlim_cur = dflssiz;
573 	p->p_limit->pl_rlimit[RLIMIT_STACK].rlim_max = maxssiz;
574 	/* Cast to avoid overflow on i386/PAE. */
575 	pageablemem = ptoa((vm_paddr_t)vm_free_count());
576 	p->p_limit->pl_rlimit[RLIMIT_RSS].rlim_cur =
577 	    p->p_limit->pl_rlimit[RLIMIT_RSS].rlim_max = pageablemem;
578 	p->p_limit->pl_rlimit[RLIMIT_MEMLOCK].rlim_cur = pageablemem / 3;
579 	p->p_limit->pl_rlimit[RLIMIT_MEMLOCK].rlim_max = pageablemem;
580 	p->p_cpulimit = RLIM_INFINITY;
581 
582 	PROC_LOCK(p);
583 	thread_cow_get_proc(td, p);
584 	PROC_UNLOCK(p);
585 
586 	/* Initialize resource accounting structures. */
587 	racct_create(&p->p_racct);
588 
589 	p->p_stats = pstats_alloc();
590 
591 	/* Allocate a prototype map so we have something to fork. */
592 	p->p_vmspace = &vmspace0;
593 	vmspace0.vm_refcnt = 1;
594 	pmap_pinit0(vmspace_pmap(&vmspace0));
595 
596 	/*
597 	 * proc0 is not expected to enter usermode, so there is no special
598 	 * handling for sv_minuser here, like is done for exec_new_vmspace().
599 	 */
600 	vm_map_init(&vmspace0.vm_map, vmspace_pmap(&vmspace0),
601 	    p->p_sysent->sv_minuser, p->p_sysent->sv_maxuser);
602 
603 	/*
604 	 * Call the init and ctor for the new thread and proc.  We wait
605 	 * to do this until all other structures are fairly sane.
606 	 */
607 	EVENTHANDLER_DIRECT_INVOKE(process_init, p);
608 	EVENTHANDLER_DIRECT_INVOKE(thread_init, td);
609 	EVENTHANDLER_DIRECT_INVOKE(process_ctor, p);
610 	EVENTHANDLER_DIRECT_INVOKE(thread_ctor, td);
611 
612 	/*
613 	 * Charge root for one process.
614 	 */
615 	(void)chgproccnt(p->p_ucred->cr_ruidinfo, 1, 0);
616 	PROC_LOCK(p);
617 	racct_add_force(p, RACCT_NPROC, 1);
618 	PROC_UNLOCK(p);
619 }
620 SYSINIT(p0init, SI_SUB_INTRINSIC, SI_ORDER_FIRST, proc0_init, NULL);
621 
622 /* ARGSUSED*/
623 static void
624 proc0_post(void *dummy __unused)
625 {
626 	struct proc *p;
627 	struct rusage ru;
628 	struct thread *td;
629 
630 	/*
631 	 * Now we can look at the time, having had a chance to verify the
632 	 * time from the filesystem.  Pretend that proc0 started now.
633 	 */
634 	sx_slock(&allproc_lock);
635 	FOREACH_PROC_IN_SYSTEM(p) {
636 		PROC_LOCK(p);
637 		if (p->p_state == PRS_NEW) {
638 			PROC_UNLOCK(p);
639 			continue;
640 		}
641 		microuptime(&p->p_stats->p_start);
642 		PROC_STATLOCK(p);
643 		rufetch(p, &ru);	/* Clears thread stats */
644 		p->p_rux.rux_runtime = 0;
645 		p->p_rux.rux_uticks = 0;
646 		p->p_rux.rux_sticks = 0;
647 		p->p_rux.rux_iticks = 0;
648 		PROC_STATUNLOCK(p);
649 		FOREACH_THREAD_IN_PROC(p, td) {
650 			td->td_runtime = 0;
651 		}
652 		PROC_UNLOCK(p);
653 	}
654 	sx_sunlock(&allproc_lock);
655 	PCPU_SET(switchtime, cpu_ticks());
656 	PCPU_SET(switchticks, ticks);
657 }
658 SYSINIT(p0post, SI_SUB_INTRINSIC_POST, SI_ORDER_FIRST, proc0_post, NULL);
659 
660 /*
661  ***************************************************************************
662  ****
663  **** The following SYSINIT's and glue code should be moved to the
664  **** respective files on a per subsystem basis.
665  ****
666  ***************************************************************************
667  */
668 
669 /*
670  * List of paths to try when searching for "init".
671  */
672 static char init_path[MAXPATHLEN] =
673 #ifdef	INIT_PATH
674     __XSTRING(INIT_PATH);
675 #else
676     "/sbin/init:/sbin/oinit:/sbin/init.bak:/rescue/init";
677 #endif
678 SYSCTL_STRING(_kern, OID_AUTO, init_path, CTLFLAG_RD, init_path, 0,
679 	"Path used to search the init process");
680 
681 /*
682  * Shutdown timeout of init(8).
683  * Unused within kernel, but used to control init(8), hence do not remove.
684  */
685 #ifndef INIT_SHUTDOWN_TIMEOUT
686 #define INIT_SHUTDOWN_TIMEOUT 120
687 #endif
688 static int init_shutdown_timeout = INIT_SHUTDOWN_TIMEOUT;
689 SYSCTL_INT(_kern, OID_AUTO, init_shutdown_timeout,
690 	CTLFLAG_RW, &init_shutdown_timeout, 0, "Shutdown timeout of init(8). "
691 	"Unused within kernel, but used to control init(8)");
692 
693 /*
694  * Start the initial user process; try exec'ing each pathname in init_path.
695  * The program is invoked with one argument containing the boot flags.
696  */
697 static void
698 start_init(void *dummy)
699 {
700 	struct image_args args;
701 	int error;
702 	char *var, *path;
703 	char *free_init_path, *tmp_init_path;
704 	struct thread *td;
705 	struct proc *p;
706 	struct vmspace *oldvmspace;
707 
708 	TSENTER();	/* Here so we don't overlap with mi_startup. */
709 
710 	td = curthread;
711 	p = td->td_proc;
712 
713 	vfs_mountroot();
714 
715 	/* Wipe GELI passphrase from the environment. */
716 	kern_unsetenv("kern.geom.eli.passphrase");
717 
718 	if ((var = kern_getenv("init_path")) != NULL) {
719 		strlcpy(init_path, var, sizeof(init_path));
720 		freeenv(var);
721 	}
722 	free_init_path = tmp_init_path = strdup(init_path, M_TEMP);
723 
724 	while ((path = strsep(&tmp_init_path, ":")) != NULL) {
725 		if (bootverbose)
726 			printf("start_init: trying %s\n", path);
727 
728 		memset(&args, 0, sizeof(args));
729 		error = exec_alloc_args(&args);
730 		if (error != 0)
731 			panic("%s: Can't allocate space for init arguments %d",
732 			    __func__, error);
733 
734 		error = exec_args_add_fname(&args, path, UIO_SYSSPACE);
735 		if (error != 0)
736 			panic("%s: Can't add fname %d", __func__, error);
737 		error = exec_args_add_arg(&args, path, UIO_SYSSPACE);
738 		if (error != 0)
739 			panic("%s: Can't add argv[0] %d", __func__, error);
740 		if (boothowto & RB_SINGLE)
741 			error = exec_args_add_arg(&args, "-s", UIO_SYSSPACE);
742 		if (error != 0)
743 			panic("%s: Can't add argv[0] %d", __func__, error);
744 
745 		/*
746 		 * Now try to exec the program.  If can't for any reason
747 		 * other than it doesn't exist, complain.
748 		 *
749 		 * Otherwise, return via fork_trampoline() all the way
750 		 * to user mode as init!
751 		 */
752 		KASSERT((td->td_pflags & TDP_EXECVMSPC) == 0,
753 		    ("nested execve"));
754 		oldvmspace = td->td_proc->p_vmspace;
755 		error = kern_execve(td, &args, NULL);
756 		KASSERT(error != 0,
757 		    ("kern_execve returned success, not EJUSTRETURN"));
758 		if (error == EJUSTRETURN) {
759 			if ((td->td_pflags & TDP_EXECVMSPC) != 0) {
760 				KASSERT(p->p_vmspace != oldvmspace,
761 				    ("oldvmspace still used"));
762 				vmspace_free(oldvmspace);
763 				td->td_pflags &= ~TDP_EXECVMSPC;
764 			}
765 			free(free_init_path, M_TEMP);
766 			TSEXIT();
767 			return;
768 		}
769 		if (error != ENOENT)
770 			printf("exec %s: error %d\n", path, error);
771 	}
772 	free(free_init_path, M_TEMP);
773 	printf("init: not found in path %s\n", init_path);
774 	panic("no init");
775 }
776 
777 /*
778  * Like kproc_create(), but runs in its own address space.  We do this
779  * early to reserve pid 1.  Note special case - do not make it
780  * runnable yet, init execution is started when userspace can be served.
781  */
782 static void
783 create_init(const void *udata __unused)
784 {
785 	struct fork_req fr;
786 	struct ucred *newcred, *oldcred;
787 	struct thread *td;
788 	int error;
789 
790 	bzero(&fr, sizeof(fr));
791 	fr.fr_flags = RFFDG | RFPROC | RFSTOPPED;
792 	fr.fr_procp = &initproc;
793 	error = fork1(&thread0, &fr);
794 	if (error)
795 		panic("cannot fork init: %d\n", error);
796 	KASSERT(initproc->p_pid == 1, ("create_init: initproc->p_pid != 1"));
797 	/* divorce init's credentials from the kernel's */
798 	newcred = crget();
799 	sx_xlock(&proctree_lock);
800 	PROC_LOCK(initproc);
801 	initproc->p_flag |= P_SYSTEM | P_INMEM;
802 	initproc->p_treeflag |= P_TREE_REAPER;
803 	oldcred = initproc->p_ucred;
804 	crcopy(newcred, oldcred);
805 #ifdef MAC
806 	mac_cred_create_init(newcred);
807 #endif
808 #ifdef AUDIT
809 	audit_cred_proc1(newcred);
810 #endif
811 	proc_set_cred(initproc, newcred);
812 	td = FIRST_THREAD_IN_PROC(initproc);
813 	crcowfree(td);
814 	td->td_realucred = crcowget(initproc->p_ucred);
815 	td->td_ucred = td->td_realucred;
816 	PROC_UNLOCK(initproc);
817 	sx_xunlock(&proctree_lock);
818 	crfree(oldcred);
819 	cpu_fork_kthread_handler(FIRST_THREAD_IN_PROC(initproc),
820 	    start_init, NULL);
821 }
822 SYSINIT(init, SI_SUB_CREATE_INIT, SI_ORDER_FIRST, create_init, NULL);
823 
824 /*
825  * Make it runnable now.
826  */
827 static void
828 kick_init(const void *udata __unused)
829 {
830 	struct thread *td;
831 
832 	td = FIRST_THREAD_IN_PROC(initproc);
833 	thread_lock(td);
834 	TD_SET_CAN_RUN(td);
835 	sched_add(td, SRQ_BORING);
836 }
837 SYSINIT(kickinit, SI_SUB_KTHREAD_INIT, SI_ORDER_MIDDLE, kick_init, NULL);
838