xref: /freebsd/sys/kern/init_main.c (revision edca4938f74db18d091868237592abbf7e718669)
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 
119 #ifndef BOOTHOWTO
120 #define	BOOTHOWTO	0
121 #endif
122 int	boothowto = BOOTHOWTO;	/* initialized so that it can be patched */
123 SYSCTL_INT(_debug, OID_AUTO, boothowto, CTLFLAG_RD, &boothowto, 0,
124 	"Boot control flags, passed from loader");
125 
126 #ifndef BOOTVERBOSE
127 #define	BOOTVERBOSE	0
128 #endif
129 int	bootverbose = BOOTVERBOSE;
130 SYSCTL_INT(_debug, OID_AUTO, bootverbose, CTLFLAG_RW, &bootverbose, 0,
131 	"Control the output of verbose kernel messages");
132 
133 #ifdef VERBOSE_SYSINIT
134 /*
135  * We'll use the defined value of VERBOSE_SYSINIT from the kernel config to
136  * dictate the default VERBOSE_SYSINIT behavior.  Significant values for this
137  * option and associated tunable are:
138  * - 0, 'compiled in but silent by default'
139  * - 1, 'compiled in but verbose by default' (default)
140  */
141 int	verbose_sysinit = VERBOSE_SYSINIT;
142 TUNABLE_INT("debug.verbose_sysinit", &verbose_sysinit);
143 #endif
144 
145 #ifdef INVARIANTS
146 FEATURE(invariants, "Kernel compiled with INVARIANTS, may affect performance");
147 #endif
148 
149 /*
150  * This ensures that there is at least one entry so that the sysinit_set
151  * symbol is not undefined.  A sybsystem ID of SI_SUB_DUMMY is never
152  * executed.
153  */
154 SYSINIT(placeholder, SI_SUB_DUMMY, SI_ORDER_ANY, NULL, NULL);
155 
156 /*
157  * The sysinit table itself.  Items are checked off as the are run.
158  * If we want to register new sysinit types, add them to newsysinit.
159  */
160 SET_DECLARE(sysinit_set, struct sysinit);
161 struct sysinit **sysinit, **sysinit_end;
162 struct sysinit **newsysinit, **newsysinit_end;
163 
164 /*
165  * Merge a new sysinit set into the current set, reallocating it if
166  * necessary.  This can only be called after malloc is running.
167  */
168 void
169 sysinit_add(struct sysinit **set, struct sysinit **set_end)
170 {
171 	struct sysinit **newset;
172 	struct sysinit **sipp;
173 	struct sysinit **xipp;
174 	int count;
175 
176 	count = set_end - set;
177 	if (newsysinit)
178 		count += newsysinit_end - newsysinit;
179 	else
180 		count += sysinit_end - sysinit;
181 	newset = malloc(count * sizeof(*sipp), M_TEMP, M_NOWAIT);
182 	if (newset == NULL)
183 		panic("cannot malloc for sysinit");
184 	xipp = newset;
185 	if (newsysinit)
186 		for (sipp = newsysinit; sipp < newsysinit_end; sipp++)
187 			*xipp++ = *sipp;
188 	else
189 		for (sipp = sysinit; sipp < sysinit_end; sipp++)
190 			*xipp++ = *sipp;
191 	for (sipp = set; sipp < set_end; sipp++)
192 		*xipp++ = *sipp;
193 	if (newsysinit)
194 		free(newsysinit, M_TEMP);
195 	newsysinit = newset;
196 	newsysinit_end = newset + count;
197 }
198 
199 #if defined (DDB) && defined(VERBOSE_SYSINIT)
200 static const char *
201 symbol_name(vm_offset_t va, db_strategy_t strategy)
202 {
203 	const char *name;
204 	c_db_sym_t sym;
205 	db_expr_t  offset;
206 
207 	if (va == 0)
208 		return (NULL);
209 	sym = db_search_symbol(va, strategy, &offset);
210 	if (offset != 0)
211 		return (NULL);
212 	db_symbol_values(sym, &name, NULL);
213 	return (name);
214 }
215 #endif
216 
217 /*
218  * System startup; initialize the world, create process 0, mount root
219  * filesystem, and fork to create init and pagedaemon.  Most of the
220  * hard work is done in the lower-level initialization routines including
221  * startup(), which does memory initialization and autoconfiguration.
222  *
223  * This allows simple addition of new kernel subsystems that require
224  * boot time initialization.  It also allows substitution of subsystem
225  * (for instance, a scheduler, kernel profiler, or VM system) by object
226  * module.  Finally, it allows for optional "kernel threads".
227  */
228 void
229 mi_startup(void)
230 {
231 
232 	struct sysinit **sipp;	/* system initialization*/
233 	struct sysinit **xipp;	/* interior loop of sort*/
234 	struct sysinit *save;	/* bubble*/
235 
236 #if defined(VERBOSE_SYSINIT)
237 	int last;
238 	int verbose;
239 #endif
240 
241 	TSENTER();
242 
243 	if (boothowto & RB_VERBOSE)
244 		bootverbose++;
245 
246 	if (sysinit == NULL) {
247 		sysinit = SET_BEGIN(sysinit_set);
248 		sysinit_end = SET_LIMIT(sysinit_set);
249 	}
250 
251 restart:
252 	/*
253 	 * Perform a bubble sort of the system initialization objects by
254 	 * their subsystem (primary key) and order (secondary key).
255 	 */
256 	for (sipp = sysinit; sipp < sysinit_end; sipp++) {
257 		for (xipp = sipp + 1; xipp < sysinit_end; xipp++) {
258 			if ((*sipp)->subsystem < (*xipp)->subsystem ||
259 			     ((*sipp)->subsystem == (*xipp)->subsystem &&
260 			      (*sipp)->order <= (*xipp)->order))
261 				continue;	/* skip*/
262 			save = *sipp;
263 			*sipp = *xipp;
264 			*xipp = save;
265 		}
266 	}
267 
268 #if defined(VERBOSE_SYSINIT)
269 	last = SI_SUB_COPYRIGHT;
270 	verbose = 0;
271 #if !defined(DDB)
272 	printf("VERBOSE_SYSINIT: DDB not enabled, symbol lookups disabled.\n");
273 #endif
274 #endif
275 
276 	/*
277 	 * Traverse the (now) ordered list of system initialization tasks.
278 	 * Perform each task, and continue on to the next task.
279 	 */
280 	for (sipp = sysinit; sipp < sysinit_end; sipp++) {
281 
282 		if ((*sipp)->subsystem == SI_SUB_DUMMY)
283 			continue;	/* skip dummy task(s)*/
284 
285 		if ((*sipp)->subsystem == SI_SUB_DONE)
286 			continue;
287 
288 #if defined(VERBOSE_SYSINIT)
289 		if ((*sipp)->subsystem > last && verbose_sysinit != 0) {
290 			verbose = 1;
291 			last = (*sipp)->subsystem;
292 			printf("subsystem %x\n", last);
293 		}
294 		if (verbose) {
295 #if defined(DDB)
296 			const char *func, *data;
297 
298 			func = symbol_name((vm_offset_t)(*sipp)->func,
299 			    DB_STGY_PROC);
300 			data = symbol_name((vm_offset_t)(*sipp)->udata,
301 			    DB_STGY_ANY);
302 			if (func != NULL && data != NULL)
303 				printf("   %s(&%s)... ", func, data);
304 			else if (func != NULL)
305 				printf("   %s(%p)... ", func, (*sipp)->udata);
306 			else
307 #endif
308 				printf("   %p(%p)... ", (*sipp)->func,
309 				    (*sipp)->udata);
310 		}
311 #endif
312 
313 		/* Call function */
314 		(*((*sipp)->func))((*sipp)->udata);
315 
316 #if defined(VERBOSE_SYSINIT)
317 		if (verbose)
318 			printf("done.\n");
319 #endif
320 
321 		/* Check off the one we're just done */
322 		(*sipp)->subsystem = SI_SUB_DONE;
323 
324 		/* Check if we've installed more sysinit items via KLD */
325 		if (newsysinit != NULL) {
326 			if (sysinit != SET_BEGIN(sysinit_set))
327 				free(sysinit, M_TEMP);
328 			sysinit = newsysinit;
329 			sysinit_end = newsysinit_end;
330 			newsysinit = NULL;
331 			newsysinit_end = NULL;
332 			goto restart;
333 		}
334 	}
335 
336 	TSEXIT();	/* Here so we don't overlap with start_init. */
337 
338 	mtx_assert(&Giant, MA_OWNED | MA_NOTRECURSED);
339 	mtx_unlock(&Giant);
340 
341 	/*
342 	 * Now hand over this thread to swapper.
343 	 */
344 	swapper();
345 	/* NOTREACHED*/
346 }
347 
348 static void
349 print_caddr_t(void *data)
350 {
351 	printf("%s", (char *)data);
352 }
353 
354 static void
355 print_version(void *data __unused)
356 {
357 	int len;
358 
359 	/* Strip a trailing newline from version. */
360 	len = strlen(version);
361 	while (len > 0 && version[len - 1] == '\n')
362 		len--;
363 	printf("%.*s %s\n", len, version, machine);
364 	printf("%s\n", compiler_version);
365 }
366 
367 SYSINIT(announce, SI_SUB_COPYRIGHT, SI_ORDER_FIRST, print_caddr_t,
368     copyright);
369 SYSINIT(trademark, SI_SUB_COPYRIGHT, SI_ORDER_SECOND, print_caddr_t,
370     trademark);
371 SYSINIT(version, SI_SUB_COPYRIGHT, SI_ORDER_THIRD, print_version, NULL);
372 
373 #ifdef WITNESS
374 static char wit_warn[] =
375      "WARNING: WITNESS option enabled, expect reduced performance.\n";
376 SYSINIT(witwarn, SI_SUB_COPYRIGHT, SI_ORDER_THIRD + 1,
377    print_caddr_t, wit_warn);
378 SYSINIT(witwarn2, SI_SUB_LAST, SI_ORDER_THIRD + 1,
379    print_caddr_t, wit_warn);
380 #endif
381 
382 #ifdef DIAGNOSTIC
383 static char diag_warn[] =
384      "WARNING: DIAGNOSTIC option enabled, expect reduced performance.\n";
385 SYSINIT(diagwarn, SI_SUB_COPYRIGHT, SI_ORDER_THIRD + 2,
386     print_caddr_t, diag_warn);
387 SYSINIT(diagwarn2, SI_SUB_LAST, SI_ORDER_THIRD + 2,
388     print_caddr_t, diag_warn);
389 #endif
390 
391 static int
392 null_fetch_syscall_args(struct thread *td __unused)
393 {
394 
395 	panic("null_fetch_syscall_args");
396 }
397 
398 static void
399 null_set_syscall_retval(struct thread *td __unused, int error __unused)
400 {
401 
402 	panic("null_set_syscall_retval");
403 }
404 
405 struct sysentvec null_sysvec = {
406 	.sv_size	= 0,
407 	.sv_table	= NULL,
408 	.sv_errsize	= 0,
409 	.sv_errtbl	= NULL,
410 	.sv_transtrap	= NULL,
411 	.sv_fixup	= NULL,
412 	.sv_sendsig	= NULL,
413 	.sv_sigcode	= NULL,
414 	.sv_szsigcode	= NULL,
415 	.sv_name	= "null",
416 	.sv_coredump	= NULL,
417 	.sv_imgact_try	= NULL,
418 	.sv_minsigstksz	= 0,
419 	.sv_minuser	= VM_MIN_ADDRESS,
420 	.sv_maxuser	= VM_MAXUSER_ADDRESS,
421 	.sv_usrstack	= USRSTACK,
422 	.sv_psstrings	= PS_STRINGS,
423 	.sv_stackprot	= VM_PROT_ALL,
424 	.sv_copyout_strings	= NULL,
425 	.sv_setregs	= NULL,
426 	.sv_fixlimit	= NULL,
427 	.sv_maxssiz	= NULL,
428 	.sv_flags	= 0,
429 	.sv_set_syscall_retval = null_set_syscall_retval,
430 	.sv_fetch_syscall_args = null_fetch_syscall_args,
431 	.sv_syscallnames = NULL,
432 	.sv_schedtail	= NULL,
433 	.sv_thread_detach = NULL,
434 	.sv_trap	= NULL,
435 };
436 
437 /*
438  * The two following SYSINIT's are proc0 specific glue code.  I am not
439  * convinced that they can not be safely combined, but their order of
440  * operation has been maintained as the same as the original init_main.c
441  * for right now.
442  */
443 /* ARGSUSED*/
444 static void
445 proc0_init(void *dummy __unused)
446 {
447 	struct proc *p;
448 	struct thread *td;
449 	struct ucred *newcred;
450 	struct uidinfo tmpuinfo;
451 	struct loginclass tmplc = {
452 		.lc_name = "",
453 	};
454 	vm_paddr_t pageablemem;
455 	int i;
456 
457 	GIANT_REQUIRED;
458 	p = &proc0;
459 	td = &thread0;
460 
461 	/*
462 	 * Initialize magic number and osrel.
463 	 */
464 	p->p_magic = P_MAGIC;
465 	p->p_osrel = osreldate;
466 
467 	/*
468 	 * Initialize thread and process structures.
469 	 */
470 	procinit();	/* set up proc zone */
471 	threadinit();	/* set up UMA zones */
472 
473 	/*
474 	 * Initialise scheduler resources.
475 	 * Add scheduler specific parts to proc, thread as needed.
476 	 */
477 	schedinit();	/* scheduler gets its house in order */
478 
479 	/*
480 	 * Create process 0 (the swapper).
481 	 */
482 	LIST_INSERT_HEAD(&allproc, p, p_list);
483 	LIST_INSERT_HEAD(PIDHASH(0), p, p_hash);
484 	mtx_init(&pgrp0.pg_mtx, "process group", NULL, MTX_DEF | MTX_DUPOK);
485 	p->p_pgrp = &pgrp0;
486 	LIST_INSERT_HEAD(PGRPHASH(0), &pgrp0, pg_hash);
487 	LIST_INIT(&pgrp0.pg_members);
488 	LIST_INSERT_HEAD(&pgrp0.pg_members, p, p_pglist);
489 
490 	pgrp0.pg_session = &session0;
491 	mtx_init(&session0.s_mtx, "session", NULL, MTX_DEF);
492 	refcount_init(&session0.s_count, 1);
493 	session0.s_leader = p;
494 
495 	p->p_sysent = &null_sysvec;
496 	p->p_flag = P_SYSTEM | P_INMEM | P_KPROC;
497 	p->p_flag2 = 0;
498 	p->p_state = PRS_NORMAL;
499 	p->p_klist = knlist_alloc(&p->p_mtx);
500 	STAILQ_INIT(&p->p_ktr);
501 	p->p_nice = NZERO;
502 	/* pid_max cannot be greater than PID_MAX */
503 	td->td_tid = PID_MAX + 1;
504 	LIST_INSERT_HEAD(TIDHASH(td->td_tid), td, td_hash);
505 	td->td_state = TDS_RUNNING;
506 	td->td_pri_class = PRI_TIMESHARE;
507 	td->td_user_pri = PUSER;
508 	td->td_base_user_pri = PUSER;
509 	td->td_lend_user_pri = PRI_MAX;
510 	td->td_priority = PVM;
511 	td->td_base_pri = PVM;
512 	td->td_oncpu = curcpu;
513 	td->td_flags = TDF_INMEM;
514 	td->td_pflags = TDP_KTHREAD;
515 	td->td_cpuset = cpuset_thread0();
516 	td->td_domain.dr_policy = td->td_cpuset->cs_domain;
517 	epoch_thread_init(td);
518 	prison0_init();
519 	p->p_peers = 0;
520 	p->p_leader = p;
521 	p->p_reaper = p;
522 	p->p_treeflag |= P_TREE_REAPER;
523 	LIST_INIT(&p->p_reaplist);
524 
525 	strncpy(p->p_comm, "kernel", sizeof (p->p_comm));
526 	strncpy(td->td_name, "swapper", sizeof (td->td_name));
527 
528 	callout_init_mtx(&p->p_itcallout, &p->p_mtx, 0);
529 	callout_init_mtx(&p->p_limco, &p->p_mtx, 0);
530 	callout_init(&td->td_slpcallout, 1);
531 
532 	/* Create credentials. */
533 	newcred = crget();
534 	newcred->cr_ngroups = 1;	/* group 0 */
535 	/* A hack to prevent uifind from tripping over NULL pointers. */
536 	curthread->td_ucred = newcred;
537 	tmpuinfo.ui_uid = 1;
538 	newcred->cr_uidinfo = newcred->cr_ruidinfo = &tmpuinfo;
539 	newcred->cr_uidinfo = uifind(0);
540 	newcred->cr_ruidinfo = uifind(0);
541 	newcred->cr_loginclass = &tmplc;
542 	newcred->cr_loginclass = loginclass_find("default");
543 	/* End hack. creds get properly set later with thread_cow_get_proc */
544 	curthread->td_ucred = NULL;
545 	newcred->cr_prison = &prison0;
546 	proc_set_cred_init(p, newcred);
547 #ifdef AUDIT
548 	audit_cred_kproc0(newcred);
549 #endif
550 #ifdef MAC
551 	mac_cred_create_swapper(newcred);
552 #endif
553 	/* Create sigacts. */
554 	p->p_sigacts = sigacts_alloc();
555 
556 	/* Initialize signal state for process 0. */
557 	siginit(&proc0);
558 
559 	/* Create the file descriptor table. */
560 	p->p_fd = fdinit(NULL, false);
561 	p->p_fdtol = NULL;
562 
563 	/* Create the limits structures. */
564 	p->p_limit = lim_alloc();
565 	for (i = 0; i < RLIM_NLIMITS; i++)
566 		p->p_limit->pl_rlimit[i].rlim_cur =
567 		    p->p_limit->pl_rlimit[i].rlim_max = RLIM_INFINITY;
568 	p->p_limit->pl_rlimit[RLIMIT_NOFILE].rlim_cur =
569 	    p->p_limit->pl_rlimit[RLIMIT_NOFILE].rlim_max = maxfiles;
570 	p->p_limit->pl_rlimit[RLIMIT_NPROC].rlim_cur =
571 	    p->p_limit->pl_rlimit[RLIMIT_NPROC].rlim_max = maxproc;
572 	p->p_limit->pl_rlimit[RLIMIT_DATA].rlim_cur = dfldsiz;
573 	p->p_limit->pl_rlimit[RLIMIT_DATA].rlim_max = maxdsiz;
574 	p->p_limit->pl_rlimit[RLIMIT_STACK].rlim_cur = dflssiz;
575 	p->p_limit->pl_rlimit[RLIMIT_STACK].rlim_max = maxssiz;
576 	/* Cast to avoid overflow on i386/PAE. */
577 	pageablemem = ptoa((vm_paddr_t)vm_free_count());
578 	p->p_limit->pl_rlimit[RLIMIT_RSS].rlim_cur =
579 	    p->p_limit->pl_rlimit[RLIMIT_RSS].rlim_max = pageablemem;
580 	p->p_limit->pl_rlimit[RLIMIT_MEMLOCK].rlim_cur = pageablemem / 3;
581 	p->p_limit->pl_rlimit[RLIMIT_MEMLOCK].rlim_max = pageablemem;
582 	p->p_cpulimit = RLIM_INFINITY;
583 
584 	PROC_LOCK(p);
585 	thread_cow_get_proc(td, p);
586 	PROC_UNLOCK(p);
587 
588 	/* Initialize resource accounting structures. */
589 	racct_create(&p->p_racct);
590 
591 	p->p_stats = pstats_alloc();
592 
593 	/* Allocate a prototype map so we have something to fork. */
594 	p->p_vmspace = &vmspace0;
595 	vmspace0.vm_refcnt = 1;
596 	pmap_pinit0(vmspace_pmap(&vmspace0));
597 
598 	/*
599 	 * proc0 is not expected to enter usermode, so there is no special
600 	 * handling for sv_minuser here, like is done for exec_new_vmspace().
601 	 */
602 	vm_map_init(&vmspace0.vm_map, vmspace_pmap(&vmspace0),
603 	    p->p_sysent->sv_minuser, p->p_sysent->sv_maxuser);
604 
605 	/*
606 	 * Call the init and ctor for the new thread and proc.  We wait
607 	 * to do this until all other structures are fairly sane.
608 	 */
609 	EVENTHANDLER_DIRECT_INVOKE(process_init, p);
610 	EVENTHANDLER_DIRECT_INVOKE(thread_init, td);
611 	EVENTHANDLER_DIRECT_INVOKE(process_ctor, p);
612 	EVENTHANDLER_DIRECT_INVOKE(thread_ctor, td);
613 
614 	/*
615 	 * Charge root for one process.
616 	 */
617 	(void)chgproccnt(p->p_ucred->cr_ruidinfo, 1, 0);
618 	PROC_LOCK(p);
619 	racct_add_force(p, RACCT_NPROC, 1);
620 	PROC_UNLOCK(p);
621 }
622 SYSINIT(p0init, SI_SUB_INTRINSIC, SI_ORDER_FIRST, proc0_init, NULL);
623 
624 /* ARGSUSED*/
625 static void
626 proc0_post(void *dummy __unused)
627 {
628 	struct timespec ts;
629 	struct proc *p;
630 	struct rusage ru;
631 	struct thread *td;
632 
633 	/*
634 	 * Now we can look at the time, having had a chance to verify the
635 	 * time from the filesystem.  Pretend that proc0 started now.
636 	 */
637 	sx_slock(&allproc_lock);
638 	FOREACH_PROC_IN_SYSTEM(p) {
639 		PROC_LOCK(p);
640 		if (p->p_state == PRS_NEW) {
641 			PROC_UNLOCK(p);
642 			continue;
643 		}
644 		microuptime(&p->p_stats->p_start);
645 		PROC_STATLOCK(p);
646 		rufetch(p, &ru);	/* Clears thread stats */
647 		p->p_rux.rux_runtime = 0;
648 		p->p_rux.rux_uticks = 0;
649 		p->p_rux.rux_sticks = 0;
650 		p->p_rux.rux_iticks = 0;
651 		PROC_STATUNLOCK(p);
652 		FOREACH_THREAD_IN_PROC(p, td) {
653 			td->td_runtime = 0;
654 		}
655 		PROC_UNLOCK(p);
656 	}
657 	sx_sunlock(&allproc_lock);
658 	PCPU_SET(switchtime, cpu_ticks());
659 	PCPU_SET(switchticks, ticks);
660 
661 	/*
662 	 * Give the ``random'' number generator a thump.
663 	 */
664 	nanotime(&ts);
665 	srandom(ts.tv_sec ^ ts.tv_nsec);
666 }
667 SYSINIT(p0post, SI_SUB_INTRINSIC_POST, SI_ORDER_FIRST, proc0_post, NULL);
668 
669 static void
670 random_init(void *dummy __unused)
671 {
672 
673 	/*
674 	 * After CPU has been started we have some randomness on most
675 	 * platforms via get_cyclecount().  For platforms that don't
676 	 * we will reseed random(9) in proc0_post() as well.
677 	 */
678 	srandom(get_cyclecount());
679 }
680 SYSINIT(random, SI_SUB_RANDOM, SI_ORDER_FIRST, random_init, NULL);
681 
682 /*
683  ***************************************************************************
684  ****
685  **** The following SYSINIT's and glue code should be moved to the
686  **** respective files on a per subsystem basis.
687  ****
688  ***************************************************************************
689  */
690 
691 /*
692  * List of paths to try when searching for "init".
693  */
694 static char init_path[MAXPATHLEN] =
695 #ifdef	INIT_PATH
696     __XSTRING(INIT_PATH);
697 #else
698     "/sbin/init:/sbin/oinit:/sbin/init.bak:/rescue/init";
699 #endif
700 SYSCTL_STRING(_kern, OID_AUTO, init_path, CTLFLAG_RD, init_path, 0,
701 	"Path used to search the init process");
702 
703 /*
704  * Shutdown timeout of init(8).
705  * Unused within kernel, but used to control init(8), hence do not remove.
706  */
707 #ifndef INIT_SHUTDOWN_TIMEOUT
708 #define INIT_SHUTDOWN_TIMEOUT 120
709 #endif
710 static int init_shutdown_timeout = INIT_SHUTDOWN_TIMEOUT;
711 SYSCTL_INT(_kern, OID_AUTO, init_shutdown_timeout,
712 	CTLFLAG_RW, &init_shutdown_timeout, 0, "Shutdown timeout of init(8). "
713 	"Unused within kernel, but used to control init(8)");
714 
715 /*
716  * Start the initial user process; try exec'ing each pathname in init_path.
717  * The program is invoked with one argument containing the boot flags.
718  */
719 static void
720 start_init(void *dummy)
721 {
722 	struct image_args args;
723 	int error;
724 	char *var, *path;
725 	char *free_init_path, *tmp_init_path;
726 	struct thread *td;
727 	struct proc *p;
728 	struct vmspace *oldvmspace;
729 
730 	TSENTER();	/* Here so we don't overlap with mi_startup. */
731 
732 	td = curthread;
733 	p = td->td_proc;
734 
735 	vfs_mountroot();
736 
737 	/* Wipe GELI passphrase from the environment. */
738 	kern_unsetenv("kern.geom.eli.passphrase");
739 
740 	if ((var = kern_getenv("init_path")) != NULL) {
741 		strlcpy(init_path, var, sizeof(init_path));
742 		freeenv(var);
743 	}
744 	free_init_path = tmp_init_path = strdup(init_path, M_TEMP);
745 
746 	while ((path = strsep(&tmp_init_path, ":")) != NULL) {
747 		if (bootverbose)
748 			printf("start_init: trying %s\n", path);
749 
750 		memset(&args, 0, sizeof(args));
751 		error = exec_alloc_args(&args);
752 		if (error != 0)
753 			panic("%s: Can't allocate space for init arguments %d",
754 			    __func__, error);
755 
756 		error = exec_args_add_fname(&args, path, UIO_SYSSPACE);
757 		if (error != 0)
758 			panic("%s: Can't add fname %d", __func__, error);
759 		error = exec_args_add_arg(&args, path, UIO_SYSSPACE);
760 		if (error != 0)
761 			panic("%s: Can't add argv[0] %d", __func__, error);
762 		if (boothowto & RB_SINGLE)
763 			error = exec_args_add_arg(&args, "-s", UIO_SYSSPACE);
764 		if (error != 0)
765 			panic("%s: Can't add argv[0] %d", __func__, error);
766 
767 		/*
768 		 * Now try to exec the program.  If can't for any reason
769 		 * other than it doesn't exist, complain.
770 		 *
771 		 * Otherwise, return via fork_trampoline() all the way
772 		 * to user mode as init!
773 		 */
774 		KASSERT((td->td_pflags & TDP_EXECVMSPC) == 0,
775 		    ("nested execve"));
776 		oldvmspace = td->td_proc->p_vmspace;
777 		error = kern_execve(td, &args, NULL);
778 		KASSERT(error != 0,
779 		    ("kern_execve returned success, not EJUSTRETURN"));
780 		if (error == EJUSTRETURN) {
781 			if ((td->td_pflags & TDP_EXECVMSPC) != 0) {
782 				KASSERT(p->p_vmspace != oldvmspace,
783 				    ("oldvmspace still used"));
784 				vmspace_free(oldvmspace);
785 				td->td_pflags &= ~TDP_EXECVMSPC;
786 			}
787 			free(free_init_path, M_TEMP);
788 			TSEXIT();
789 			return;
790 		}
791 		if (error != ENOENT)
792 			printf("exec %s: error %d\n", path, error);
793 	}
794 	free(free_init_path, M_TEMP);
795 	printf("init: not found in path %s\n", init_path);
796 	panic("no init");
797 }
798 
799 /*
800  * Like kproc_create(), but runs in its own address space.  We do this
801  * early to reserve pid 1.  Note special case - do not make it
802  * runnable yet, init execution is started when userspace can be served.
803  */
804 static void
805 create_init(const void *udata __unused)
806 {
807 	struct fork_req fr;
808 	struct ucred *newcred, *oldcred;
809 	struct thread *td;
810 	int error;
811 
812 	bzero(&fr, sizeof(fr));
813 	fr.fr_flags = RFFDG | RFPROC | RFSTOPPED;
814 	fr.fr_procp = &initproc;
815 	error = fork1(&thread0, &fr);
816 	if (error)
817 		panic("cannot fork init: %d\n", error);
818 	KASSERT(initproc->p_pid == 1, ("create_init: initproc->p_pid != 1"));
819 	/* divorce init's credentials from the kernel's */
820 	newcred = crget();
821 	sx_xlock(&proctree_lock);
822 	PROC_LOCK(initproc);
823 	initproc->p_flag |= P_SYSTEM | P_INMEM;
824 	initproc->p_treeflag |= P_TREE_REAPER;
825 	oldcred = initproc->p_ucred;
826 	crcopy(newcred, oldcred);
827 #ifdef MAC
828 	mac_cred_create_init(newcred);
829 #endif
830 #ifdef AUDIT
831 	audit_cred_proc1(newcred);
832 #endif
833 	proc_set_cred(initproc, newcred);
834 	td = FIRST_THREAD_IN_PROC(initproc);
835 	crfree(td->td_ucred);
836 	td->td_ucred = crhold(initproc->p_ucred);
837 	PROC_UNLOCK(initproc);
838 	sx_xunlock(&proctree_lock);
839 	crfree(oldcred);
840 	cpu_fork_kthread_handler(FIRST_THREAD_IN_PROC(initproc),
841 	    start_init, NULL);
842 }
843 SYSINIT(init, SI_SUB_CREATE_INIT, SI_ORDER_FIRST, create_init, NULL);
844 
845 /*
846  * Make it runnable now.
847  */
848 static void
849 kick_init(const void *udata __unused)
850 {
851 	struct thread *td;
852 
853 	td = FIRST_THREAD_IN_PROC(initproc);
854 	thread_lock(td);
855 	TD_SET_CAN_RUN(td);
856 	sched_add(td, SRQ_BORING);
857 	thread_unlock(td);
858 }
859 SYSINIT(kickinit, SI_SUB_KTHREAD_INIT, SI_ORDER_MIDDLE, kick_init, NULL);
860