xref: /freebsd/sys/kern/init_main.c (revision 5861f9665471e98e544f6fa3ce73c4912229ff82)
1 /*-
2  * Copyright (c) 1995 Terrence R. Lambert
3  * All rights reserved.
4  *
5  * Copyright (c) 1982, 1986, 1989, 1991, 1992, 1993
6  *	The Regents of the University of California.  All rights reserved.
7  * (c) UNIX System Laboratories, Inc.
8  * All or some portions of this file are derived from material licensed
9  * to the University of California by American Telephone and Telegraph
10  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
11  * the permission of UNIX System Laboratories, Inc.
12  *
13  * Redistribution and use in source and binary forms, with or without
14  * modification, are permitted provided that the following conditions
15  * are met:
16  * 1. Redistributions of source code must retain the above copyright
17  *    notice, this list of conditions and the following disclaimer.
18  * 2. Redistributions in binary form must reproduce the above copyright
19  *    notice, this list of conditions and the following disclaimer in the
20  *    documentation and/or other materials provided with the distribution.
21  * 3. All advertising materials mentioning features or use of this software
22  *    must display the following acknowledgement:
23  *	This product includes software developed by the University of
24  *	California, Berkeley and its contributors.
25  * 4. Neither the name of the University nor the names of its contributors
26  *    may be used to endorse or promote products derived from this software
27  *    without specific prior written permission.
28  *
29  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
30  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
31  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
32  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
33  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
34  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
35  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
36  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
37  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
38  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
39  * SUCH DAMAGE.
40  *
41  *	@(#)init_main.c	8.9 (Berkeley) 1/21/94
42  */
43 
44 #include <sys/cdefs.h>
45 __FBSDID("$FreeBSD$");
46 
47 #include "opt_ddb.h"
48 #include "opt_init_path.h"
49 
50 #include <sys/param.h>
51 #include <sys/kernel.h>
52 #include <sys/exec.h>
53 #include <sys/file.h>
54 #include <sys/filedesc.h>
55 #include <sys/jail.h>
56 #include <sys/ktr.h>
57 #include <sys/lock.h>
58 #include <sys/mount.h>
59 #include <sys/mutex.h>
60 #include <sys/syscallsubr.h>
61 #include <sys/sysctl.h>
62 #include <sys/proc.h>
63 #include <sys/resourcevar.h>
64 #include <sys/systm.h>
65 #include <sys/signalvar.h>
66 #include <sys/vnode.h>
67 #include <sys/sysent.h>
68 #include <sys/reboot.h>
69 #include <sys/sched.h>
70 #include <sys/sx.h>
71 #include <sys/sysproto.h>
72 #include <sys/vmmeter.h>
73 #include <sys/unistd.h>
74 #include <sys/malloc.h>
75 #include <sys/conf.h>
76 #include <sys/cpuset.h>
77 #include <sys/vimage.h>
78 
79 #include <machine/cpu.h>
80 
81 #include <security/audit/audit.h>
82 #include <security/mac/mac_framework.h>
83 
84 #include <vm/vm.h>
85 #include <vm/vm_param.h>
86 #include <vm/pmap.h>
87 #include <vm/vm_map.h>
88 #include <sys/copyright.h>
89 
90 #include <ddb/ddb.h>
91 #include <ddb/db_sym.h>
92 
93 void mi_startup(void);				/* Should be elsewhere */
94 
95 /* Components of the first process -- never freed. */
96 static struct session session0;
97 static struct pgrp pgrp0;
98 struct	proc proc0;
99 struct	thread thread0 __aligned(16);
100 struct	vmspace vmspace0;
101 struct	proc *initproc;
102 
103 int	boothowto = 0;		/* initialized so that it can be patched */
104 SYSCTL_INT(_debug, OID_AUTO, boothowto, CTLFLAG_RD, &boothowto, 0, "");
105 int	bootverbose;
106 SYSCTL_INT(_debug, OID_AUTO, bootverbose, CTLFLAG_RW, &bootverbose, 0, "");
107 
108 /*
109  * This ensures that there is at least one entry so that the sysinit_set
110  * symbol is not undefined.  A sybsystem ID of SI_SUB_DUMMY is never
111  * executed.
112  */
113 SYSINIT(placeholder, SI_SUB_DUMMY, SI_ORDER_ANY, NULL, NULL);
114 
115 /*
116  * The sysinit table itself.  Items are checked off as the are run.
117  * If we want to register new sysinit types, add them to newsysinit.
118  */
119 SET_DECLARE(sysinit_set, struct sysinit);
120 struct sysinit **sysinit, **sysinit_end;
121 struct sysinit **newsysinit, **newsysinit_end;
122 
123 /*
124  * Merge a new sysinit set into the current set, reallocating it if
125  * necessary.  This can only be called after malloc is running.
126  */
127 void
128 sysinit_add(struct sysinit **set, struct sysinit **set_end)
129 {
130 	struct sysinit **newset;
131 	struct sysinit **sipp;
132 	struct sysinit **xipp;
133 	int count;
134 
135 	count = set_end - set;
136 	if (newsysinit)
137 		count += newsysinit_end - newsysinit;
138 	else
139 		count += sysinit_end - sysinit;
140 	newset = malloc(count * sizeof(*sipp), M_TEMP, M_NOWAIT);
141 	if (newset == NULL)
142 		panic("cannot malloc for sysinit");
143 	xipp = newset;
144 	if (newsysinit)
145 		for (sipp = newsysinit; sipp < newsysinit_end; sipp++)
146 			*xipp++ = *sipp;
147 	else
148 		for (sipp = sysinit; sipp < sysinit_end; sipp++)
149 			*xipp++ = *sipp;
150 	for (sipp = set; sipp < set_end; sipp++)
151 		*xipp++ = *sipp;
152 	if (newsysinit)
153 		free(newsysinit, M_TEMP);
154 	newsysinit = newset;
155 	newsysinit_end = newset + count;
156 }
157 
158 /*
159  * System startup; initialize the world, create process 0, mount root
160  * filesystem, and fork to create init and pagedaemon.  Most of the
161  * hard work is done in the lower-level initialization routines including
162  * startup(), which does memory initialization and autoconfiguration.
163  *
164  * This allows simple addition of new kernel subsystems that require
165  * boot time initialization.  It also allows substitution of subsystem
166  * (for instance, a scheduler, kernel profiler, or VM system) by object
167  * module.  Finally, it allows for optional "kernel threads".
168  */
169 void
170 mi_startup(void)
171 {
172 
173 	register struct sysinit **sipp;		/* system initialization*/
174 	register struct sysinit **xipp;		/* interior loop of sort*/
175 	register struct sysinit *save;		/* bubble*/
176 
177 #if defined(VERBOSE_SYSINIT)
178 	int last;
179 	int verbose;
180 #endif
181 
182 	if (sysinit == NULL) {
183 		sysinit = SET_BEGIN(sysinit_set);
184 		sysinit_end = SET_LIMIT(sysinit_set);
185 	}
186 
187 restart:
188 	/*
189 	 * Perform a bubble sort of the system initialization objects by
190 	 * their subsystem (primary key) and order (secondary key).
191 	 */
192 	for (sipp = sysinit; sipp < sysinit_end; sipp++) {
193 		for (xipp = sipp + 1; xipp < sysinit_end; xipp++) {
194 			if ((*sipp)->subsystem < (*xipp)->subsystem ||
195 			     ((*sipp)->subsystem == (*xipp)->subsystem &&
196 			      (*sipp)->order <= (*xipp)->order))
197 				continue;	/* skip*/
198 			save = *sipp;
199 			*sipp = *xipp;
200 			*xipp = save;
201 		}
202 	}
203 
204 #if defined(VERBOSE_SYSINIT)
205 	last = SI_SUB_COPYRIGHT;
206 	verbose = 0;
207 #if !defined(DDB)
208 	printf("VERBOSE_SYSINIT: DDB not enabled, symbol lookups disabled.\n");
209 #endif
210 #endif
211 
212 	/*
213 	 * Traverse the (now) ordered list of system initialization tasks.
214 	 * Perform each task, and continue on to the next task.
215 	 *
216 	 * The last item on the list is expected to be the scheduler,
217 	 * which will not return.
218 	 */
219 	for (sipp = sysinit; sipp < sysinit_end; sipp++) {
220 
221 		if ((*sipp)->subsystem == SI_SUB_DUMMY)
222 			continue;	/* skip dummy task(s)*/
223 
224 		if ((*sipp)->subsystem == SI_SUB_DONE)
225 			continue;
226 
227 #if defined(VERBOSE_SYSINIT)
228 		if ((*sipp)->subsystem > last) {
229 			verbose = 1;
230 			last = (*sipp)->subsystem;
231 			printf("subsystem %x\n", last);
232 		}
233 		if (verbose) {
234 #if defined(DDB)
235 			const char *name;
236 			c_db_sym_t sym;
237 			db_expr_t  offset;
238 
239 			sym = db_search_symbol((vm_offset_t)(*sipp)->func,
240 			    DB_STGY_PROC, &offset);
241 			db_symbol_values(sym, &name, NULL);
242 			if (name != NULL)
243 				printf("   %s(%p)... ", name, (*sipp)->udata);
244 			else
245 #endif
246 				printf("   %p(%p)... ", (*sipp)->func,
247 				    (*sipp)->udata);
248 		}
249 #endif
250 
251 		/* Call function */
252 		(*((*sipp)->func))((*sipp)->udata);
253 
254 #if defined(VERBOSE_SYSINIT)
255 		if (verbose)
256 			printf("done.\n");
257 #endif
258 
259 		/* Check off the one we're just done */
260 		(*sipp)->subsystem = SI_SUB_DONE;
261 
262 		/* Check if we've installed more sysinit items via KLD */
263 		if (newsysinit != NULL) {
264 			if (sysinit != SET_BEGIN(sysinit_set))
265 				free(sysinit, M_TEMP);
266 			sysinit = newsysinit;
267 			sysinit_end = newsysinit_end;
268 			newsysinit = NULL;
269 			newsysinit_end = NULL;
270 			goto restart;
271 		}
272 	}
273 
274 	panic("Shouldn't get here!");
275 	/* NOTREACHED*/
276 }
277 
278 
279 /*
280  ***************************************************************************
281  ****
282  **** The following SYSINIT's belong elsewhere, but have not yet
283  **** been moved.
284  ****
285  ***************************************************************************
286  */
287 static void
288 print_caddr_t(void *data __unused)
289 {
290 	printf("%s", (char *)data);
291 }
292 SYSINIT(announce, SI_SUB_COPYRIGHT, SI_ORDER_FIRST, print_caddr_t,
293     copyright);
294 SYSINIT(trademark, SI_SUB_COPYRIGHT, SI_ORDER_SECOND, print_caddr_t,
295     trademark);
296 SYSINIT(version, SI_SUB_COPYRIGHT, SI_ORDER_THIRD, print_caddr_t, version);
297 
298 #ifdef WITNESS
299 static char wit_warn[] =
300      "WARNING: WITNESS option enabled, expect reduced performance.\n";
301 SYSINIT(witwarn, SI_SUB_COPYRIGHT, SI_ORDER_THIRD + 1,
302    print_caddr_t, wit_warn);
303 SYSINIT(witwarn2, SI_SUB_RUN_SCHEDULER, SI_ORDER_THIRD + 1,
304    print_caddr_t, wit_warn);
305 #endif
306 
307 #ifdef DIAGNOSTIC
308 static char diag_warn[] =
309      "WARNING: DIAGNOSTIC option enabled, expect reduced performance.\n";
310 SYSINIT(diagwarn, SI_SUB_COPYRIGHT, SI_ORDER_THIRD + 2,
311     print_caddr_t, diag_warn);
312 SYSINIT(diagwarn2, SI_SUB_RUN_SCHEDULER, SI_ORDER_THIRD + 2,
313     print_caddr_t, diag_warn);
314 #endif
315 
316 static void
317 set_boot_verbose(void *data __unused)
318 {
319 
320 	if (boothowto & RB_VERBOSE)
321 		bootverbose++;
322 }
323 SYSINIT(boot_verbose, SI_SUB_TUNABLES, SI_ORDER_ANY, set_boot_verbose, NULL);
324 
325 struct sysentvec null_sysvec = {
326 	.sv_size	= 0,
327 	.sv_table	= NULL,
328 	.sv_mask	= 0,
329 	.sv_sigsize	= 0,
330 	.sv_sigtbl	= NULL,
331 	.sv_errsize	= 0,
332 	.sv_errtbl	= NULL,
333 	.sv_transtrap	= NULL,
334 	.sv_fixup	= NULL,
335 	.sv_sendsig	= NULL,
336 	.sv_sigcode	= NULL,
337 	.sv_szsigcode	= NULL,
338 	.sv_prepsyscall	= NULL,
339 	.sv_name	= "null",
340 	.sv_coredump	= NULL,
341 	.sv_imgact_try	= NULL,
342 	.sv_minsigstksz	= 0,
343 	.sv_pagesize	= PAGE_SIZE,
344 	.sv_minuser	= VM_MIN_ADDRESS,
345 	.sv_maxuser	= VM_MAXUSER_ADDRESS,
346 	.sv_usrstack	= USRSTACK,
347 	.sv_psstrings	= PS_STRINGS,
348 	.sv_stackprot	= VM_PROT_ALL,
349 	.sv_copyout_strings	= NULL,
350 	.sv_setregs	= NULL,
351 	.sv_fixlimit	= NULL,
352 	.sv_maxssiz	= NULL
353 };
354 
355 /*
356  ***************************************************************************
357  ****
358  **** The two following SYSINIT's are proc0 specific glue code.  I am not
359  **** convinced that they can not be safely combined, but their order of
360  **** operation has been maintained as the same as the original init_main.c
361  **** for right now.
362  ****
363  **** These probably belong in init_proc.c or kern_proc.c, since they
364  **** deal with proc0 (the fork template process).
365  ****
366  ***************************************************************************
367  */
368 /* ARGSUSED*/
369 static void
370 proc0_init(void *dummy __unused)
371 {
372 	struct proc *p;
373 	unsigned i;
374 	struct thread *td;
375 
376 	GIANT_REQUIRED;
377 	p = &proc0;
378 	td = &thread0;
379 
380 	/*
381 	 * Initialize magic number and osrel.
382 	 */
383 	p->p_magic = P_MAGIC;
384 	p->p_osrel = osreldate;
385 
386 	/*
387 	 * Initialize thread and process structures.
388 	 */
389 	procinit();	/* set up proc zone */
390 	threadinit();	/* set up UMA zones */
391 
392 	/*
393 	 * Initialise scheduler resources.
394 	 * Add scheduler specific parts to proc, thread as needed.
395 	 */
396 	schedinit();	/* scheduler gets its house in order */
397 	/*
398 	 * Initialize sleep queue hash table
399 	 */
400 	sleepinit();
401 
402 	/*
403 	 * additional VM structures
404 	 */
405 	vm_init2();
406 
407 	/*
408 	 * Create process 0 (the swapper).
409 	 */
410 	LIST_INSERT_HEAD(&allproc, p, p_list);
411 	LIST_INSERT_HEAD(PIDHASH(0), p, p_hash);
412 	mtx_init(&pgrp0.pg_mtx, "process group", NULL, MTX_DEF | MTX_DUPOK);
413 	p->p_pgrp = &pgrp0;
414 	LIST_INSERT_HEAD(PGRPHASH(0), &pgrp0, pg_hash);
415 	LIST_INIT(&pgrp0.pg_members);
416 	LIST_INSERT_HEAD(&pgrp0.pg_members, p, p_pglist);
417 
418 	pgrp0.pg_session = &session0;
419 	mtx_init(&session0.s_mtx, "session", NULL, MTX_DEF);
420 	refcount_init(&session0.s_count, 1);
421 	session0.s_leader = p;
422 
423 	p->p_sysent = &null_sysvec;
424 	p->p_flag = P_SYSTEM | P_INMEM;
425 	p->p_state = PRS_NORMAL;
426 	knlist_init_mtx(&p->p_klist, &p->p_mtx);
427 	STAILQ_INIT(&p->p_ktr);
428 	p->p_nice = NZERO;
429 	td->td_tid = PID_MAX + 1;
430 	td->td_state = TDS_RUNNING;
431 	td->td_pri_class = PRI_TIMESHARE;
432 	td->td_user_pri = PUSER;
433 	td->td_base_user_pri = PUSER;
434 	td->td_priority = PVM;
435 	td->td_base_pri = PUSER;
436 	td->td_oncpu = 0;
437 	td->td_flags = TDF_INMEM|TDP_KTHREAD;
438 	td->td_cpuset = cpuset_thread0();
439 	prison0.pr_cpuset = cpuset_ref(td->td_cpuset);
440 	p->p_peers = 0;
441 	p->p_leader = p;
442 
443 
444 	strncpy(p->p_comm, "kernel", sizeof (p->p_comm));
445 	strncpy(td->td_name, "swapper", sizeof (td->td_name));
446 
447 	callout_init(&p->p_itcallout, CALLOUT_MPSAFE);
448 	callout_init_mtx(&p->p_limco, &p->p_mtx, 0);
449 	callout_init(&td->td_slpcallout, CALLOUT_MPSAFE);
450 
451 	/* Create credentials. */
452 	p->p_ucred = crget();
453 	p->p_ucred->cr_ngroups = 1;	/* group 0 */
454 	p->p_ucred->cr_uidinfo = uifind(0);
455 	p->p_ucred->cr_ruidinfo = uifind(0);
456 	p->p_ucred->cr_prison = &prison0;
457 #ifdef VIMAGE
458 	KASSERT(LIST_FIRST(&vimage_head) != NULL, ("vimage_head empty"));
459 	P_TO_VIMAGE(p) =  LIST_FIRST(&vimage_head); /* set ucred->cr_vimage */
460 	refcount_acquire(&P_TO_VIMAGE(p)->vi_ucredrefc);
461 	LIST_FIRST(&vprocg_head)->nprocs++;
462 #endif
463 #ifdef AUDIT
464 	audit_cred_kproc0(p->p_ucred);
465 #endif
466 #ifdef MAC
467 	mac_cred_create_swapper(p->p_ucred);
468 #endif
469 	td->td_ucred = crhold(p->p_ucred);
470 
471 	/* Create sigacts. */
472 	p->p_sigacts = sigacts_alloc();
473 
474 	/* Initialize signal state for process 0. */
475 	siginit(&proc0);
476 
477 	/* Create the file descriptor table. */
478 	p->p_fd = fdinit(NULL);
479 	p->p_fdtol = NULL;
480 
481 	/* Create the limits structures. */
482 	p->p_limit = lim_alloc();
483 	for (i = 0; i < RLIM_NLIMITS; i++)
484 		p->p_limit->pl_rlimit[i].rlim_cur =
485 		    p->p_limit->pl_rlimit[i].rlim_max = RLIM_INFINITY;
486 	p->p_limit->pl_rlimit[RLIMIT_NOFILE].rlim_cur =
487 	    p->p_limit->pl_rlimit[RLIMIT_NOFILE].rlim_max = maxfiles;
488 	p->p_limit->pl_rlimit[RLIMIT_NPROC].rlim_cur =
489 	    p->p_limit->pl_rlimit[RLIMIT_NPROC].rlim_max = maxproc;
490 	i = ptoa(cnt.v_free_count);
491 	p->p_limit->pl_rlimit[RLIMIT_RSS].rlim_max = i;
492 	p->p_limit->pl_rlimit[RLIMIT_MEMLOCK].rlim_max = i;
493 	p->p_limit->pl_rlimit[RLIMIT_MEMLOCK].rlim_cur = i / 3;
494 	p->p_cpulimit = RLIM_INFINITY;
495 
496 	p->p_stats = pstats_alloc();
497 
498 	/* Allocate a prototype map so we have something to fork. */
499 	pmap_pinit0(vmspace_pmap(&vmspace0));
500 	p->p_vmspace = &vmspace0;
501 	vmspace0.vm_refcnt = 1;
502 	vm_map_init(&vmspace0.vm_map, p->p_sysent->sv_minuser,
503 	    p->p_sysent->sv_maxuser);
504 	vmspace0.vm_map.pmap = vmspace_pmap(&vmspace0);
505 
506 	/*-
507 	 * call the init and ctor for the new thread and proc
508 	 * we wait to do this until all other structures
509 	 * are fairly sane.
510 	 */
511 	EVENTHANDLER_INVOKE(process_init, p);
512 	EVENTHANDLER_INVOKE(thread_init, td);
513 	EVENTHANDLER_INVOKE(process_ctor, p);
514 	EVENTHANDLER_INVOKE(thread_ctor, td);
515 
516 	/*
517 	 * Charge root for one process.
518 	 */
519 	(void)chgproccnt(p->p_ucred->cr_ruidinfo, 1, 0);
520 }
521 SYSINIT(p0init, SI_SUB_INTRINSIC, SI_ORDER_FIRST, proc0_init, NULL);
522 
523 /* ARGSUSED*/
524 static void
525 proc0_post(void *dummy __unused)
526 {
527 	struct timespec ts;
528 	struct proc *p;
529 	struct rusage ru;
530 	struct thread *td;
531 
532 	/*
533 	 * Now we can look at the time, having had a chance to verify the
534 	 * time from the filesystem.  Pretend that proc0 started now.
535 	 */
536 	sx_slock(&allproc_lock);
537 	FOREACH_PROC_IN_SYSTEM(p) {
538 		microuptime(&p->p_stats->p_start);
539 		PROC_SLOCK(p);
540 		rufetch(p, &ru);	/* Clears thread stats */
541 		PROC_SUNLOCK(p);
542 		p->p_rux.rux_runtime = 0;
543 		p->p_rux.rux_uticks = 0;
544 		p->p_rux.rux_sticks = 0;
545 		p->p_rux.rux_iticks = 0;
546 		FOREACH_THREAD_IN_PROC(p, td) {
547 			td->td_runtime = 0;
548 		}
549 	}
550 	sx_sunlock(&allproc_lock);
551 	PCPU_SET(switchtime, cpu_ticks());
552 	PCPU_SET(switchticks, ticks);
553 
554 	/*
555 	 * Give the ``random'' number generator a thump.
556 	 */
557 	nanotime(&ts);
558 	srandom(ts.tv_sec ^ ts.tv_nsec);
559 }
560 SYSINIT(p0post, SI_SUB_INTRINSIC_POST, SI_ORDER_FIRST, proc0_post, NULL);
561 
562 /*
563  ***************************************************************************
564  ****
565  **** The following SYSINIT's and glue code should be moved to the
566  **** respective files on a per subsystem basis.
567  ****
568  ***************************************************************************
569  */
570 
571 
572 /*
573  ***************************************************************************
574  ****
575  **** The following code probably belongs in another file, like
576  **** kern/init_init.c.
577  ****
578  ***************************************************************************
579  */
580 
581 /*
582  * List of paths to try when searching for "init".
583  */
584 static char init_path[MAXPATHLEN] =
585 #ifdef	INIT_PATH
586     __XSTRING(INIT_PATH);
587 #else
588     "/sbin/init:/sbin/oinit:/sbin/init.bak:/rescue/init:/stand/sysinstall";
589 #endif
590 SYSCTL_STRING(_kern, OID_AUTO, init_path, CTLFLAG_RD, init_path, 0,
591 	"Path used to search the init process");
592 
593 /*
594  * Shutdown timeout of init(8).
595  * Unused within kernel, but used to control init(8), hence do not remove.
596  */
597 #ifndef INIT_SHUTDOWN_TIMEOUT
598 #define INIT_SHUTDOWN_TIMEOUT 120
599 #endif
600 static int init_shutdown_timeout = INIT_SHUTDOWN_TIMEOUT;
601 SYSCTL_INT(_kern, OID_AUTO, init_shutdown_timeout,
602 	CTLFLAG_RW, &init_shutdown_timeout, 0, "");
603 
604 /*
605  * Start the initial user process; try exec'ing each pathname in init_path.
606  * The program is invoked with one argument containing the boot flags.
607  */
608 static void
609 start_init(void *dummy)
610 {
611 	vm_offset_t addr;
612 	struct execve_args args;
613 	int options, error;
614 	char *var, *path, *next, *s;
615 	char *ucp, **uap, *arg0, *arg1;
616 	struct thread *td;
617 	struct proc *p;
618 
619 	mtx_lock(&Giant);
620 
621 	GIANT_REQUIRED;
622 
623 	td = curthread;
624 	p = td->td_proc;
625 
626 	vfs_mountroot();
627 
628 	/*
629 	 * Need just enough stack to hold the faked-up "execve()" arguments.
630 	 */
631 	addr = p->p_sysent->sv_usrstack - PAGE_SIZE;
632 	if (vm_map_find(&p->p_vmspace->vm_map, NULL, 0, &addr, PAGE_SIZE,
633 			FALSE, VM_PROT_ALL, VM_PROT_ALL, 0) != 0)
634 		panic("init: couldn't allocate argument space");
635 	p->p_vmspace->vm_maxsaddr = (caddr_t)addr;
636 	p->p_vmspace->vm_ssize = 1;
637 
638 	if ((var = getenv("init_path")) != NULL) {
639 		strlcpy(init_path, var, sizeof(init_path));
640 		freeenv(var);
641 	}
642 
643 	for (path = init_path; *path != '\0'; path = next) {
644 		while (*path == ':')
645 			path++;
646 		if (*path == '\0')
647 			break;
648 		for (next = path; *next != '\0' && *next != ':'; next++)
649 			/* nothing */ ;
650 		if (bootverbose)
651 			printf("start_init: trying %.*s\n", (int)(next - path),
652 			    path);
653 
654 		/*
655 		 * Move out the boot flag argument.
656 		 */
657 		options = 0;
658 		ucp = (char *)p->p_sysent->sv_usrstack;
659 		(void)subyte(--ucp, 0);		/* trailing zero */
660 		if (boothowto & RB_SINGLE) {
661 			(void)subyte(--ucp, 's');
662 			options = 1;
663 		}
664 #ifdef notyet
665                 if (boothowto & RB_FASTBOOT) {
666 			(void)subyte(--ucp, 'f');
667 			options = 1;
668 		}
669 #endif
670 
671 #ifdef BOOTCDROM
672 		(void)subyte(--ucp, 'C');
673 		options = 1;
674 #endif
675 
676 		if (options == 0)
677 			(void)subyte(--ucp, '-');
678 		(void)subyte(--ucp, '-');		/* leading hyphen */
679 		arg1 = ucp;
680 
681 		/*
682 		 * Move out the file name (also arg 0).
683 		 */
684 		(void)subyte(--ucp, 0);
685 		for (s = next - 1; s >= path; s--)
686 			(void)subyte(--ucp, *s);
687 		arg0 = ucp;
688 
689 		/*
690 		 * Move out the arg pointers.
691 		 */
692 		uap = (char **)((intptr_t)ucp & ~(sizeof(intptr_t)-1));
693 		(void)suword((caddr_t)--uap, (long)0);	/* terminator */
694 		(void)suword((caddr_t)--uap, (long)(intptr_t)arg1);
695 		(void)suword((caddr_t)--uap, (long)(intptr_t)arg0);
696 
697 		/*
698 		 * Point at the arguments.
699 		 */
700 		args.fname = arg0;
701 		args.argv = uap;
702 		args.envv = NULL;
703 
704 		/*
705 		 * Now try to exec the program.  If can't for any reason
706 		 * other than it doesn't exist, complain.
707 		 *
708 		 * Otherwise, return via fork_trampoline() all the way
709 		 * to user mode as init!
710 		 */
711 		if ((error = execve(td, &args)) == 0) {
712 			mtx_unlock(&Giant);
713 			return;
714 		}
715 		if (error != ENOENT)
716 			printf("exec %.*s: error %d\n", (int)(next - path),
717 			    path, error);
718 	}
719 	printf("init: not found in path %s\n", init_path);
720 	panic("no init");
721 }
722 
723 /*
724  * Like kproc_create(), but runs in it's own address space.
725  * We do this early to reserve pid 1.
726  *
727  * Note special case - do not make it runnable yet.  Other work
728  * in progress will change this more.
729  */
730 static void
731 create_init(const void *udata __unused)
732 {
733 	struct ucred *newcred, *oldcred;
734 	int error;
735 
736 	error = fork1(&thread0, RFFDG | RFPROC | RFSTOPPED, 0, &initproc);
737 	if (error)
738 		panic("cannot fork init: %d\n", error);
739 	KASSERT(initproc->p_pid == 1, ("create_init: initproc->p_pid != 1"));
740 	/* divorce init's credentials from the kernel's */
741 	newcred = crget();
742 	PROC_LOCK(initproc);
743 	initproc->p_flag |= P_SYSTEM | P_INMEM;
744 	oldcred = initproc->p_ucred;
745 	crcopy(newcred, oldcred);
746 #ifdef MAC
747 	mac_cred_create_init(newcred);
748 #endif
749 #ifdef AUDIT
750 	audit_cred_proc1(newcred);
751 #endif
752 	initproc->p_ucred = newcred;
753 	PROC_UNLOCK(initproc);
754 	crfree(oldcred);
755 	cred_update_thread(FIRST_THREAD_IN_PROC(initproc));
756 	cpu_set_fork_handler(FIRST_THREAD_IN_PROC(initproc), start_init, NULL);
757 }
758 SYSINIT(init, SI_SUB_CREATE_INIT, SI_ORDER_FIRST, create_init, NULL);
759 
760 /*
761  * Make it runnable now.
762  */
763 static void
764 kick_init(const void *udata __unused)
765 {
766 	struct thread *td;
767 
768 	td = FIRST_THREAD_IN_PROC(initproc);
769 	thread_lock(td);
770 	TD_SET_CAN_RUN(td);
771 	sched_add(td, SRQ_BORING);
772 	thread_unlock(td);
773 }
774 SYSINIT(kickinit, SI_SUB_KTHREAD_INIT, SI_ORDER_FIRST, kick_init, NULL);
775