xref: /freebsd/sys/kern/init_main.c (revision d34048812292b714a0bf99967270d18fe3097c62)
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/exec.h>
57 #include <sys/file.h>
58 #include <sys/filedesc.h>
59 #include <sys/jail.h>
60 #include <sys/ktr.h>
61 #include <sys/lock.h>
62 #include <sys/loginclass.h>
63 #include <sys/mount.h>
64 #include <sys/mutex.h>
65 #include <sys/syscallsubr.h>
66 #include <sys/sysctl.h>
67 #include <sys/proc.h>
68 #include <sys/racct.h>
69 #include <sys/resourcevar.h>
70 #include <sys/systm.h>
71 #include <sys/signalvar.h>
72 #include <sys/vnode.h>
73 #include <sys/sysent.h>
74 #include <sys/reboot.h>
75 #include <sys/sched.h>
76 #include <sys/sx.h>
77 #include <sys/sysproto.h>
78 #include <sys/vmmeter.h>
79 #include <sys/unistd.h>
80 #include <sys/malloc.h>
81 #include <sys/conf.h>
82 #include <sys/cpuset.h>
83 
84 #include <machine/cpu.h>
85 
86 #include <security/audit/audit.h>
87 #include <security/mac/mac_framework.h>
88 
89 #include <vm/vm.h>
90 #include <vm/vm_param.h>
91 #include <vm/vm_extern.h>
92 #include <vm/pmap.h>
93 #include <vm/vm_map.h>
94 #include <sys/copyright.h>
95 
96 #include <ddb/ddb.h>
97 #include <ddb/db_sym.h>
98 
99 void mi_startup(void);				/* Should be elsewhere */
100 
101 /* Components of the first process -- never freed. */
102 static struct session session0;
103 static struct pgrp pgrp0;
104 struct	proc proc0;
105 struct thread0_storage thread0_st __aligned(32);
106 struct	vmspace vmspace0;
107 struct	proc *initproc;
108 
109 #ifndef BOOTHOWTO
110 #define	BOOTHOWTO	0
111 #endif
112 int	boothowto = BOOTHOWTO;	/* initialized so that it can be patched */
113 SYSCTL_INT(_debug, OID_AUTO, boothowto, CTLFLAG_RD, &boothowto, 0,
114 	"Boot control flags, passed from loader");
115 
116 #ifndef BOOTVERBOSE
117 #define	BOOTVERBOSE	0
118 #endif
119 int	bootverbose = BOOTVERBOSE;
120 SYSCTL_INT(_debug, OID_AUTO, bootverbose, CTLFLAG_RW, &bootverbose, 0,
121 	"Control the output of verbose kernel messages");
122 
123 #ifdef VERBOSE_SYSINIT
124 /*
125  * We'll use the defined value of VERBOSE_SYSINIT from the kernel config to
126  * dictate the default VERBOSE_SYSINIT behavior.  Significant values for this
127  * option and associated tunable are:
128  * - 0, 'compiled in but silent by default'
129  * - 1, 'compiled in but verbose by default' (default)
130  */
131 int	verbose_sysinit = VERBOSE_SYSINIT;
132 TUNABLE_INT("debug.verbose_sysinit", &verbose_sysinit);
133 #endif
134 
135 #ifdef INVARIANTS
136 FEATURE(invariants, "Kernel compiled with INVARIANTS, may affect performance");
137 #endif
138 
139 /*
140  * This ensures that there is at least one entry so that the sysinit_set
141  * symbol is not undefined.  A sybsystem ID of SI_SUB_DUMMY is never
142  * executed.
143  */
144 SYSINIT(placeholder, SI_SUB_DUMMY, SI_ORDER_ANY, NULL, NULL);
145 
146 /*
147  * The sysinit table itself.  Items are checked off as the are run.
148  * If we want to register new sysinit types, add them to newsysinit.
149  */
150 SET_DECLARE(sysinit_set, struct sysinit);
151 struct sysinit **sysinit, **sysinit_end;
152 struct sysinit **newsysinit, **newsysinit_end;
153 
154 EVENTHANDLER_LIST_DECLARE(process_init);
155 EVENTHANDLER_LIST_DECLARE(thread_init);
156 EVENTHANDLER_LIST_DECLARE(process_ctor);
157 EVENTHANDLER_LIST_DECLARE(thread_ctor);
158 
159 /*
160  * Merge a new sysinit set into the current set, reallocating it if
161  * necessary.  This can only be called after malloc is running.
162  */
163 void
164 sysinit_add(struct sysinit **set, struct sysinit **set_end)
165 {
166 	struct sysinit **newset;
167 	struct sysinit **sipp;
168 	struct sysinit **xipp;
169 	int count;
170 
171 	count = set_end - set;
172 	if (newsysinit)
173 		count += newsysinit_end - newsysinit;
174 	else
175 		count += sysinit_end - sysinit;
176 	newset = malloc(count * sizeof(*sipp), M_TEMP, M_NOWAIT);
177 	if (newset == NULL)
178 		panic("cannot malloc for sysinit");
179 	xipp = newset;
180 	if (newsysinit)
181 		for (sipp = newsysinit; sipp < newsysinit_end; sipp++)
182 			*xipp++ = *sipp;
183 	else
184 		for (sipp = sysinit; sipp < sysinit_end; sipp++)
185 			*xipp++ = *sipp;
186 	for (sipp = set; sipp < set_end; sipp++)
187 		*xipp++ = *sipp;
188 	if (newsysinit)
189 		free(newsysinit, M_TEMP);
190 	newsysinit = newset;
191 	newsysinit_end = newset + count;
192 }
193 
194 #if defined (DDB) && defined(VERBOSE_SYSINIT)
195 static const char *
196 symbol_name(vm_offset_t va, db_strategy_t strategy)
197 {
198 	const char *name;
199 	c_db_sym_t sym;
200 	db_expr_t  offset;
201 
202 	if (va == 0)
203 		return (NULL);
204 	sym = db_search_symbol(va, strategy, &offset);
205 	if (offset != 0)
206 		return (NULL);
207 	db_symbol_values(sym, &name, NULL);
208 	return (name);
209 }
210 #endif
211 
212 /*
213  * System startup; initialize the world, create process 0, mount root
214  * filesystem, and fork to create init and pagedaemon.  Most of the
215  * hard work is done in the lower-level initialization routines including
216  * startup(), which does memory initialization and autoconfiguration.
217  *
218  * This allows simple addition of new kernel subsystems that require
219  * boot time initialization.  It also allows substitution of subsystem
220  * (for instance, a scheduler, kernel profiler, or VM system) by object
221  * module.  Finally, it allows for optional "kernel threads".
222  */
223 void
224 mi_startup(void)
225 {
226 
227 	struct sysinit **sipp;	/* system initialization*/
228 	struct sysinit **xipp;	/* interior loop of sort*/
229 	struct sysinit *save;	/* bubble*/
230 
231 #if defined(VERBOSE_SYSINIT)
232 	int last;
233 	int verbose;
234 #endif
235 
236 	TSENTER();
237 
238 	if (boothowto & RB_VERBOSE)
239 		bootverbose++;
240 
241 	if (sysinit == NULL) {
242 		sysinit = SET_BEGIN(sysinit_set);
243 		sysinit_end = SET_LIMIT(sysinit_set);
244 	}
245 
246 restart:
247 	/*
248 	 * Perform a bubble sort of the system initialization objects by
249 	 * their subsystem (primary key) and order (secondary key).
250 	 */
251 	for (sipp = sysinit; sipp < sysinit_end; sipp++) {
252 		for (xipp = sipp + 1; xipp < sysinit_end; xipp++) {
253 			if ((*sipp)->subsystem < (*xipp)->subsystem ||
254 			     ((*sipp)->subsystem == (*xipp)->subsystem &&
255 			      (*sipp)->order <= (*xipp)->order))
256 				continue;	/* skip*/
257 			save = *sipp;
258 			*sipp = *xipp;
259 			*xipp = save;
260 		}
261 	}
262 
263 #if defined(VERBOSE_SYSINIT)
264 	last = SI_SUB_COPYRIGHT;
265 	verbose = 0;
266 #if !defined(DDB)
267 	printf("VERBOSE_SYSINIT: DDB not enabled, symbol lookups disabled.\n");
268 #endif
269 #endif
270 
271 	/*
272 	 * Traverse the (now) ordered list of system initialization tasks.
273 	 * Perform each task, and continue on to the next task.
274 	 */
275 	for (sipp = sysinit; sipp < sysinit_end; sipp++) {
276 
277 		if ((*sipp)->subsystem == SI_SUB_DUMMY)
278 			continue;	/* skip dummy task(s)*/
279 
280 		if ((*sipp)->subsystem == SI_SUB_DONE)
281 			continue;
282 
283 #if defined(VERBOSE_SYSINIT)
284 		if ((*sipp)->subsystem > last && verbose_sysinit != 0) {
285 			verbose = 1;
286 			last = (*sipp)->subsystem;
287 			printf("subsystem %x\n", last);
288 		}
289 		if (verbose) {
290 #if defined(DDB)
291 			const char *func, *data;
292 
293 			func = symbol_name((vm_offset_t)(*sipp)->func,
294 			    DB_STGY_PROC);
295 			data = symbol_name((vm_offset_t)(*sipp)->udata,
296 			    DB_STGY_ANY);
297 			if (func != NULL && data != NULL)
298 				printf("   %s(&%s)... ", func, data);
299 			else if (func != NULL)
300 				printf("   %s(%p)... ", func, (*sipp)->udata);
301 			else
302 #endif
303 				printf("   %p(%p)... ", (*sipp)->func,
304 				    (*sipp)->udata);
305 		}
306 #endif
307 
308 		/* Call function */
309 		(*((*sipp)->func))((*sipp)->udata);
310 
311 #if defined(VERBOSE_SYSINIT)
312 		if (verbose)
313 			printf("done.\n");
314 #endif
315 
316 		/* Check off the one we're just done */
317 		(*sipp)->subsystem = SI_SUB_DONE;
318 
319 		/* Check if we've installed more sysinit items via KLD */
320 		if (newsysinit != NULL) {
321 			if (sysinit != SET_BEGIN(sysinit_set))
322 				free(sysinit, M_TEMP);
323 			sysinit = newsysinit;
324 			sysinit_end = newsysinit_end;
325 			newsysinit = NULL;
326 			newsysinit_end = NULL;
327 			goto restart;
328 		}
329 	}
330 
331 	TSEXIT();	/* Here so we don't overlap with start_init. */
332 
333 	mtx_assert(&Giant, MA_OWNED | MA_NOTRECURSED);
334 	mtx_unlock(&Giant);
335 
336 	/*
337 	 * Now hand over this thread to swapper.
338 	 */
339 	swapper();
340 	/* NOTREACHED*/
341 }
342 
343 static void
344 print_caddr_t(void *data)
345 {
346 	printf("%s", (char *)data);
347 }
348 
349 static void
350 print_version(void *data __unused)
351 {
352 	int len;
353 
354 	/* Strip a trailing newline from version. */
355 	len = strlen(version);
356 	while (len > 0 && version[len - 1] == '\n')
357 		len--;
358 	printf("%.*s %s\n", len, version, machine);
359 	printf("%s\n", compiler_version);
360 }
361 
362 SYSINIT(announce, SI_SUB_COPYRIGHT, SI_ORDER_FIRST, print_caddr_t,
363     copyright);
364 SYSINIT(trademark, SI_SUB_COPYRIGHT, SI_ORDER_SECOND, print_caddr_t,
365     trademark);
366 SYSINIT(version, SI_SUB_COPYRIGHT, SI_ORDER_THIRD, print_version, NULL);
367 
368 #ifdef WITNESS
369 static char wit_warn[] =
370      "WARNING: WITNESS option enabled, expect reduced performance.\n";
371 SYSINIT(witwarn, SI_SUB_COPYRIGHT, SI_ORDER_THIRD + 1,
372    print_caddr_t, wit_warn);
373 SYSINIT(witwarn2, SI_SUB_LAST, SI_ORDER_THIRD + 1,
374    print_caddr_t, wit_warn);
375 #endif
376 
377 #ifdef DIAGNOSTIC
378 static char diag_warn[] =
379      "WARNING: DIAGNOSTIC option enabled, expect reduced performance.\n";
380 SYSINIT(diagwarn, SI_SUB_COPYRIGHT, SI_ORDER_THIRD + 2,
381     print_caddr_t, diag_warn);
382 SYSINIT(diagwarn2, SI_SUB_LAST, SI_ORDER_THIRD + 2,
383     print_caddr_t, diag_warn);
384 #endif
385 
386 static int
387 null_fetch_syscall_args(struct thread *td __unused)
388 {
389 
390 	panic("null_fetch_syscall_args");
391 }
392 
393 static void
394 null_set_syscall_retval(struct thread *td __unused, int error __unused)
395 {
396 
397 	panic("null_set_syscall_retval");
398 }
399 
400 struct sysentvec null_sysvec = {
401 	.sv_size	= 0,
402 	.sv_table	= NULL,
403 	.sv_mask	= 0,
404 	.sv_errsize	= 0,
405 	.sv_errtbl	= 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_pagesize	= PAGE_SIZE,
416 	.sv_minuser	= VM_MIN_ADDRESS,
417 	.sv_maxuser	= VM_MAXUSER_ADDRESS,
418 	.sv_usrstack	= USRSTACK,
419 	.sv_psstrings	= PS_STRINGS,
420 	.sv_stackprot	= VM_PROT_ALL,
421 	.sv_copyout_strings	= NULL,
422 	.sv_setregs	= NULL,
423 	.sv_fixlimit	= NULL,
424 	.sv_maxssiz	= NULL,
425 	.sv_flags	= 0,
426 	.sv_set_syscall_retval = null_set_syscall_retval,
427 	.sv_fetch_syscall_args = null_fetch_syscall_args,
428 	.sv_syscallnames = NULL,
429 	.sv_schedtail	= NULL,
430 	.sv_thread_detach = NULL,
431 	.sv_trap	= NULL,
432 };
433 
434 /*
435  * The two following SYSINIT's are proc0 specific glue code.  I am not
436  * convinced that they can not be safely combined, but their order of
437  * operation has been maintained as the same as the original init_main.c
438  * for right now.
439  */
440 /* ARGSUSED*/
441 static void
442 proc0_init(void *dummy __unused)
443 {
444 	struct proc *p;
445 	struct thread *td;
446 	struct ucred *newcred;
447 	struct uidinfo tmpuinfo;
448 	struct loginclass tmplc = {
449 		.lc_name = "",
450 	};
451 	vm_paddr_t pageablemem;
452 	int i;
453 
454 	GIANT_REQUIRED;
455 	p = &proc0;
456 	td = &thread0;
457 
458 	/*
459 	 * Initialize magic number and osrel.
460 	 */
461 	p->p_magic = P_MAGIC;
462 	p->p_osrel = osreldate;
463 
464 	/*
465 	 * Initialize thread and process structures.
466 	 */
467 	procinit();	/* set up proc zone */
468 	threadinit();	/* set up UMA zones */
469 
470 	/*
471 	 * Initialise scheduler resources.
472 	 * Add scheduler specific parts to proc, thread as needed.
473 	 */
474 	schedinit();	/* scheduler gets its house in order */
475 
476 	/*
477 	 * Create process 0 (the swapper).
478 	 */
479 	LIST_INSERT_HEAD(&allproc, p, p_list);
480 	LIST_INSERT_HEAD(PIDHASH(0), p, p_hash);
481 	mtx_init(&pgrp0.pg_mtx, "process group", NULL, MTX_DEF | MTX_DUPOK);
482 	p->p_pgrp = &pgrp0;
483 	LIST_INSERT_HEAD(PGRPHASH(0), &pgrp0, pg_hash);
484 	LIST_INIT(&pgrp0.pg_members);
485 	LIST_INSERT_HEAD(&pgrp0.pg_members, p, p_pglist);
486 
487 	pgrp0.pg_session = &session0;
488 	mtx_init(&session0.s_mtx, "session", NULL, MTX_DEF);
489 	refcount_init(&session0.s_count, 1);
490 	session0.s_leader = p;
491 
492 	p->p_sysent = &null_sysvec;
493 	p->p_flag = P_SYSTEM | P_INMEM | P_KPROC;
494 	p->p_flag2 = 0;
495 	p->p_state = PRS_NORMAL;
496 	p->p_klist = knlist_alloc(&p->p_mtx);
497 	STAILQ_INIT(&p->p_ktr);
498 	p->p_nice = NZERO;
499 	/* pid_max cannot be greater than PID_MAX */
500 	td->td_tid = PID_MAX + 1;
501 	LIST_INSERT_HEAD(TIDHASH(td->td_tid), td, td_hash);
502 	td->td_state = TDS_RUNNING;
503 	td->td_pri_class = PRI_TIMESHARE;
504 	td->td_user_pri = PUSER;
505 	td->td_base_user_pri = PUSER;
506 	td->td_lend_user_pri = PRI_MAX;
507 	td->td_priority = PVM;
508 	td->td_base_pri = PVM;
509 	td->td_oncpu = curcpu;
510 	td->td_flags = TDF_INMEM;
511 	td->td_pflags = TDP_KTHREAD;
512 	td->td_cpuset = cpuset_thread0();
513 	td->td_domain.dr_policy = td->td_cpuset->cs_domain;
514 	epoch_thread_init(td);
515 	prison0_init();
516 	p->p_peers = 0;
517 	p->p_leader = p;
518 	p->p_reaper = p;
519 	p->p_treeflag |= P_TREE_REAPER;
520 	LIST_INIT(&p->p_reaplist);
521 
522 	strncpy(p->p_comm, "kernel", sizeof (p->p_comm));
523 	strncpy(td->td_name, "swapper", sizeof (td->td_name));
524 
525 	callout_init_mtx(&p->p_itcallout, &p->p_mtx, 0);
526 	callout_init_mtx(&p->p_limco, &p->p_mtx, 0);
527 	callout_init(&td->td_slpcallout, 1);
528 
529 	/* Create credentials. */
530 	newcred = crget();
531 	newcred->cr_ngroups = 1;	/* group 0 */
532 	/* A hack to prevent uifind from tripping over NULL pointers. */
533 	curthread->td_ucred = newcred;
534 	tmpuinfo.ui_uid = 1;
535 	newcred->cr_uidinfo = newcred->cr_ruidinfo = &tmpuinfo;
536 	newcred->cr_uidinfo = uifind(0);
537 	newcred->cr_ruidinfo = uifind(0);
538 	newcred->cr_loginclass = &tmplc;
539 	newcred->cr_loginclass = loginclass_find("default");
540 	/* End hack. creds get properly set later with thread_cow_get_proc */
541 	curthread->td_ucred = NULL;
542 	newcred->cr_prison = &prison0;
543 	proc_set_cred_init(p, newcred);
544 #ifdef AUDIT
545 	audit_cred_kproc0(newcred);
546 #endif
547 #ifdef MAC
548 	mac_cred_create_swapper(newcred);
549 #endif
550 	/* Create sigacts. */
551 	p->p_sigacts = sigacts_alloc();
552 
553 	/* Initialize signal state for process 0. */
554 	siginit(&proc0);
555 
556 	/* Create the file descriptor table. */
557 	p->p_fd = fdinit(NULL, false);
558 	p->p_fdtol = NULL;
559 
560 	/* Create the limits structures. */
561 	p->p_limit = lim_alloc();
562 	for (i = 0; i < RLIM_NLIMITS; i++)
563 		p->p_limit->pl_rlimit[i].rlim_cur =
564 		    p->p_limit->pl_rlimit[i].rlim_max = RLIM_INFINITY;
565 	p->p_limit->pl_rlimit[RLIMIT_NOFILE].rlim_cur =
566 	    p->p_limit->pl_rlimit[RLIMIT_NOFILE].rlim_max = maxfiles;
567 	p->p_limit->pl_rlimit[RLIMIT_NPROC].rlim_cur =
568 	    p->p_limit->pl_rlimit[RLIMIT_NPROC].rlim_max = maxproc;
569 	p->p_limit->pl_rlimit[RLIMIT_DATA].rlim_cur = dfldsiz;
570 	p->p_limit->pl_rlimit[RLIMIT_DATA].rlim_max = maxdsiz;
571 	p->p_limit->pl_rlimit[RLIMIT_STACK].rlim_cur = dflssiz;
572 	p->p_limit->pl_rlimit[RLIMIT_STACK].rlim_max = maxssiz;
573 	/* Cast to avoid overflow on i386/PAE. */
574 	pageablemem = ptoa((vm_paddr_t)vm_free_count());
575 	p->p_limit->pl_rlimit[RLIMIT_RSS].rlim_cur =
576 	    p->p_limit->pl_rlimit[RLIMIT_RSS].rlim_max = pageablemem;
577 	p->p_limit->pl_rlimit[RLIMIT_MEMLOCK].rlim_cur = pageablemem / 3;
578 	p->p_limit->pl_rlimit[RLIMIT_MEMLOCK].rlim_max = pageablemem;
579 	p->p_cpulimit = RLIM_INFINITY;
580 
581 	PROC_LOCK(p);
582 	thread_cow_get_proc(td, p);
583 	PROC_UNLOCK(p);
584 
585 	/* Initialize resource accounting structures. */
586 	racct_create(&p->p_racct);
587 
588 	p->p_stats = pstats_alloc();
589 
590 	/* Allocate a prototype map so we have something to fork. */
591 	p->p_vmspace = &vmspace0;
592 	vmspace0.vm_refcnt = 1;
593 	pmap_pinit0(vmspace_pmap(&vmspace0));
594 
595 	/*
596 	 * proc0 is not expected to enter usermode, so there is no special
597 	 * handling for sv_minuser here, like is done for exec_new_vmspace().
598 	 */
599 	vm_map_init(&vmspace0.vm_map, vmspace_pmap(&vmspace0),
600 	    p->p_sysent->sv_minuser, p->p_sysent->sv_maxuser);
601 
602 	/*
603 	 * Call the init and ctor for the new thread and proc.  We wait
604 	 * to do this until all other structures are fairly sane.
605 	 */
606 	EVENTHANDLER_DIRECT_INVOKE(process_init, p);
607 	EVENTHANDLER_DIRECT_INVOKE(thread_init, td);
608 	EVENTHANDLER_DIRECT_INVOKE(process_ctor, p);
609 	EVENTHANDLER_DIRECT_INVOKE(thread_ctor, td);
610 
611 	/*
612 	 * Charge root for one process.
613 	 */
614 	(void)chgproccnt(p->p_ucred->cr_ruidinfo, 1, 0);
615 	PROC_LOCK(p);
616 	racct_add_force(p, RACCT_NPROC, 1);
617 	PROC_UNLOCK(p);
618 }
619 SYSINIT(p0init, SI_SUB_INTRINSIC, SI_ORDER_FIRST, proc0_init, NULL);
620 
621 /* ARGSUSED*/
622 static void
623 proc0_post(void *dummy __unused)
624 {
625 	struct timespec ts;
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 	/*
659 	 * Give the ``random'' number generator a thump.
660 	 */
661 	nanotime(&ts);
662 	srandom(ts.tv_sec ^ ts.tv_nsec);
663 }
664 SYSINIT(p0post, SI_SUB_INTRINSIC_POST, SI_ORDER_FIRST, proc0_post, NULL);
665 
666 static void
667 random_init(void *dummy __unused)
668 {
669 
670 	/*
671 	 * After CPU has been started we have some randomness on most
672 	 * platforms via get_cyclecount().  For platforms that don't
673 	 * we will reseed random(9) in proc0_post() as well.
674 	 */
675 	srandom(get_cyclecount());
676 }
677 SYSINIT(random, SI_SUB_RANDOM, SI_ORDER_FIRST, random_init, NULL);
678 
679 /*
680  ***************************************************************************
681  ****
682  **** The following SYSINIT's and glue code should be moved to the
683  **** respective files on a per subsystem basis.
684  ****
685  ***************************************************************************
686  */
687 
688 /*
689  * List of paths to try when searching for "init".
690  */
691 static char init_path[MAXPATHLEN] =
692 #ifdef	INIT_PATH
693     __XSTRING(INIT_PATH);
694 #else
695     "/sbin/init:/sbin/oinit:/sbin/init.bak:/rescue/init";
696 #endif
697 SYSCTL_STRING(_kern, OID_AUTO, init_path, CTLFLAG_RD, init_path, 0,
698 	"Path used to search the init process");
699 
700 /*
701  * Shutdown timeout of init(8).
702  * Unused within kernel, but used to control init(8), hence do not remove.
703  */
704 #ifndef INIT_SHUTDOWN_TIMEOUT
705 #define INIT_SHUTDOWN_TIMEOUT 120
706 #endif
707 static int init_shutdown_timeout = INIT_SHUTDOWN_TIMEOUT;
708 SYSCTL_INT(_kern, OID_AUTO, init_shutdown_timeout,
709 	CTLFLAG_RW, &init_shutdown_timeout, 0, "Shutdown timeout of init(8). "
710 	"Unused within kernel, but used to control init(8)");
711 
712 /*
713  * Start the initial user process; try exec'ing each pathname in init_path.
714  * The program is invoked with one argument containing the boot flags.
715  */
716 static void
717 start_init(void *dummy)
718 {
719 	vm_offset_t addr;
720 	struct execve_args args;
721 	int options, error;
722 	size_t pathlen;
723 	char *var, *path;
724 	char *free_init_path, *tmp_init_path;
725 	char *ucp, **uap, *arg0, *arg1;
726 	struct thread *td;
727 	struct proc *p;
728 
729 	TSENTER();	/* Here so we don't overlap with mi_startup. */
730 
731 	td = curthread;
732 	p = td->td_proc;
733 
734 	vfs_mountroot();
735 
736 	/* Wipe GELI passphrase from the environment. */
737 	kern_unsetenv("kern.geom.eli.passphrase");
738 
739 	/*
740 	 * Need just enough stack to hold the faked-up "execve()" arguments.
741 	 */
742 	addr = p->p_sysent->sv_usrstack - PAGE_SIZE;
743 	if (vm_map_find(&p->p_vmspace->vm_map, NULL, 0, &addr, PAGE_SIZE, 0,
744 	    VMFS_NO_SPACE, VM_PROT_ALL, VM_PROT_ALL, 0) != 0)
745 		panic("init: couldn't allocate argument space");
746 	p->p_vmspace->vm_maxsaddr = (caddr_t)addr;
747 	p->p_vmspace->vm_ssize = 1;
748 
749 	if ((var = kern_getenv("init_path")) != NULL) {
750 		strlcpy(init_path, var, sizeof(init_path));
751 		freeenv(var);
752 	}
753 	free_init_path = tmp_init_path = strdup(init_path, M_TEMP);
754 
755 	while ((path = strsep(&tmp_init_path, ":")) != NULL) {
756 		pathlen = strlen(path) + 1;
757 		if (bootverbose)
758 			printf("start_init: trying %s\n", path);
759 
760 		/*
761 		 * Move out the boot flag argument.
762 		 */
763 		options = 0;
764 		ucp = (char *)p->p_sysent->sv_usrstack;
765 		(void)subyte(--ucp, 0);		/* trailing zero */
766 		if (boothowto & RB_SINGLE) {
767 			(void)subyte(--ucp, 's');
768 			options = 1;
769 		}
770 #ifdef notyet
771                 if (boothowto & RB_FASTBOOT) {
772 			(void)subyte(--ucp, 'f');
773 			options = 1;
774 		}
775 #endif
776 
777 #ifdef BOOTCDROM
778 		(void)subyte(--ucp, 'C');
779 		options = 1;
780 #endif
781 
782 		if (options == 0)
783 			(void)subyte(--ucp, '-');
784 		(void)subyte(--ucp, '-');		/* leading hyphen */
785 		arg1 = ucp;
786 
787 		/*
788 		 * Move out the file name (also arg 0).
789 		 */
790 		ucp -= pathlen;
791 		copyout(path, ucp, pathlen);
792 		arg0 = ucp;
793 
794 		/*
795 		 * Move out the arg pointers.
796 		 */
797 		uap = (char **)rounddown2((intptr_t)ucp, sizeof(intptr_t));
798 		(void)suword((caddr_t)--uap, (long)0);	/* terminator */
799 		(void)suword((caddr_t)--uap, (long)(intptr_t)arg1);
800 		(void)suword((caddr_t)--uap, (long)(intptr_t)arg0);
801 
802 		/*
803 		 * Point at the arguments.
804 		 */
805 		args.fname = arg0;
806 		args.argv = uap;
807 		args.envv = NULL;
808 
809 		/*
810 		 * Now try to exec the program.  If can't for any reason
811 		 * other than it doesn't exist, complain.
812 		 *
813 		 * Otherwise, return via fork_trampoline() all the way
814 		 * to user mode as init!
815 		 */
816 		if ((error = sys_execve(td, &args)) == EJUSTRETURN) {
817 			free(free_init_path, M_TEMP);
818 			TSEXIT();
819 			return;
820 		}
821 		if (error != ENOENT)
822 			printf("exec %s: error %d\n", path, error);
823 	}
824 	free(free_init_path, M_TEMP);
825 	printf("init: not found in path %s\n", init_path);
826 	panic("no init");
827 }
828 
829 /*
830  * Like kproc_create(), but runs in its own address space.
831  * We do this early to reserve pid 1.
832  *
833  * Note special case - do not make it runnable yet.  Other work
834  * in progress will change this more.
835  */
836 static void
837 create_init(const void *udata __unused)
838 {
839 	struct fork_req fr;
840 	struct ucred *newcred, *oldcred;
841 	struct thread *td;
842 	int error;
843 
844 	bzero(&fr, sizeof(fr));
845 	fr.fr_flags = RFFDG | RFPROC | RFSTOPPED;
846 	fr.fr_procp = &initproc;
847 	error = fork1(&thread0, &fr);
848 	if (error)
849 		panic("cannot fork init: %d\n", error);
850 	KASSERT(initproc->p_pid == 1, ("create_init: initproc->p_pid != 1"));
851 	/* divorce init's credentials from the kernel's */
852 	newcred = crget();
853 	sx_xlock(&proctree_lock);
854 	PROC_LOCK(initproc);
855 	initproc->p_flag |= P_SYSTEM | P_INMEM;
856 	initproc->p_treeflag |= P_TREE_REAPER;
857 	oldcred = initproc->p_ucred;
858 	crcopy(newcred, oldcred);
859 #ifdef MAC
860 	mac_cred_create_init(newcred);
861 #endif
862 #ifdef AUDIT
863 	audit_cred_proc1(newcred);
864 #endif
865 	proc_set_cred(initproc, newcred);
866 	td = FIRST_THREAD_IN_PROC(initproc);
867 	crfree(td->td_ucred);
868 	td->td_ucred = crhold(initproc->p_ucred);
869 	PROC_UNLOCK(initproc);
870 	sx_xunlock(&proctree_lock);
871 	crfree(oldcred);
872 	cpu_fork_kthread_handler(FIRST_THREAD_IN_PROC(initproc),
873 	    start_init, NULL);
874 }
875 SYSINIT(init, SI_SUB_CREATE_INIT, SI_ORDER_FIRST, create_init, NULL);
876 
877 /*
878  * Make it runnable now.
879  */
880 static void
881 kick_init(const void *udata __unused)
882 {
883 	struct thread *td;
884 
885 	td = FIRST_THREAD_IN_PROC(initproc);
886 	thread_lock(td);
887 	TD_SET_CAN_RUN(td);
888 	sched_add(td, SRQ_BORING);
889 	thread_unlock(td);
890 }
891 SYSINIT(kickinit, SI_SUB_KTHREAD_INIT, SI_ORDER_MIDDLE, kick_init, NULL);
892