xref: /freebsd/sys/kern/init_main.c (revision 52ec752989b2e6d4e9a59a8ff25d8ff596d85e62)
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_init_path.h"
48 #include "opt_mac.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/ktr.h>
56 #include <sys/lock.h>
57 #include <sys/mac.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 
77 #include <machine/cpu.h>
78 
79 #include <vm/vm.h>
80 #include <vm/vm_param.h>
81 #include <vm/pmap.h>
82 #include <vm/vm_map.h>
83 #include <sys/user.h>
84 #include <sys/copyright.h>
85 
86 void mi_startup(void);				/* Should be elsewhere */
87 
88 /* Components of the first process -- never freed. */
89 static struct session session0;
90 static struct pgrp pgrp0;
91 struct	proc proc0;
92 struct	thread thread0;
93 struct	kse kse0;
94 struct	ksegrp ksegrp0;
95 static struct filedesc0 filedesc0;
96 static struct plimit limit0;
97 struct	vmspace vmspace0;
98 struct	proc *initproc;
99 
100 struct	vnode *rootvp;
101 int	boothowto = 0;		/* initialized so that it can be patched */
102 SYSCTL_INT(_debug, OID_AUTO, boothowto, CTLFLAG_RD, &boothowto, 0, "");
103 int	bootverbose;
104 SYSCTL_INT(_debug, OID_AUTO, bootverbose, CTLFLAG_RW, &bootverbose, 0, "");
105 
106 /*
107  * This ensures that there is at least one entry so that the sysinit_set
108  * symbol is not undefined.  A sybsystem ID of SI_SUB_DUMMY is never
109  * executed.
110  */
111 SYSINIT(placeholder, SI_SUB_DUMMY, SI_ORDER_ANY, NULL, NULL)
112 
113 /*
114  * The sysinit table itself.  Items are checked off as the are run.
115  * If we want to register new sysinit types, add them to newsysinit.
116  */
117 SET_DECLARE(sysinit_set, struct sysinit);
118 struct sysinit **sysinit, **sysinit_end;
119 struct sysinit **newsysinit, **newsysinit_end;
120 
121 /*
122  * Merge a new sysinit set into the current set, reallocating it if
123  * necessary.  This can only be called after malloc is running.
124  */
125 void
126 sysinit_add(struct sysinit **set, struct sysinit **set_end)
127 {
128 	struct sysinit **newset;
129 	struct sysinit **sipp;
130 	struct sysinit **xipp;
131 	int count;
132 
133 	count = set_end - set;
134 	if (newsysinit)
135 		count += newsysinit_end - newsysinit;
136 	else
137 		count += sysinit_end - sysinit;
138 	newset = malloc(count * sizeof(*sipp), M_TEMP, M_NOWAIT);
139 	if (newset == NULL)
140 		panic("cannot malloc for sysinit");
141 	xipp = newset;
142 	if (newsysinit)
143 		for (sipp = newsysinit; sipp < newsysinit_end; sipp++)
144 			*xipp++ = *sipp;
145 	else
146 		for (sipp = sysinit; sipp < sysinit_end; sipp++)
147 			*xipp++ = *sipp;
148 	for (sipp = set; sipp < set_end; sipp++)
149 		*xipp++ = *sipp;
150 	if (newsysinit)
151 		free(newsysinit, M_TEMP);
152 	newsysinit = newset;
153 	newsysinit_end = newset + count;
154 }
155 
156 /*
157  * System startup; initialize the world, create process 0, mount root
158  * filesystem, and fork to create init and pagedaemon.  Most of the
159  * hard work is done in the lower-level initialization routines including
160  * startup(), which does memory initialization and autoconfiguration.
161  *
162  * This allows simple addition of new kernel subsystems that require
163  * boot time initialization.  It also allows substitution of subsystem
164  * (for instance, a scheduler, kernel profiler, or VM system) by object
165  * module.  Finally, it allows for optional "kernel threads".
166  */
167 void
168 mi_startup(void)
169 {
170 
171 	register struct sysinit **sipp;		/* system initialization*/
172 	register struct sysinit **xipp;		/* interior loop of sort*/
173 	register struct sysinit *save;		/* bubble*/
174 
175 	if (sysinit == NULL) {
176 		sysinit = SET_BEGIN(sysinit_set);
177 		sysinit_end = SET_LIMIT(sysinit_set);
178 	}
179 
180 restart:
181 	/*
182 	 * Perform a bubble sort of the system initialization objects by
183 	 * their subsystem (primary key) and order (secondary key).
184 	 */
185 	for (sipp = sysinit; sipp < sysinit_end; sipp++) {
186 		for (xipp = sipp + 1; xipp < sysinit_end; xipp++) {
187 			if ((*sipp)->subsystem < (*xipp)->subsystem ||
188 			     ((*sipp)->subsystem == (*xipp)->subsystem &&
189 			      (*sipp)->order <= (*xipp)->order))
190 				continue;	/* skip*/
191 			save = *sipp;
192 			*sipp = *xipp;
193 			*xipp = save;
194 		}
195 	}
196 
197 	/*
198 	 * Traverse the (now) ordered list of system initialization tasks.
199 	 * Perform each task, and continue on to the next task.
200 	 *
201 	 * The last item on the list is expected to be the scheduler,
202 	 * which will not return.
203 	 */
204 	for (sipp = sysinit; sipp < sysinit_end; sipp++) {
205 
206 		if ((*sipp)->subsystem == SI_SUB_DUMMY)
207 			continue;	/* skip dummy task(s)*/
208 
209 		if ((*sipp)->subsystem == SI_SUB_DONE)
210 			continue;
211 
212 		/* Call function */
213 		(*((*sipp)->func))((*sipp)->udata);
214 
215 		/* Check off the one we're just done */
216 		(*sipp)->subsystem = SI_SUB_DONE;
217 
218 		/* Check if we've installed more sysinit items via KLD */
219 		if (newsysinit != NULL) {
220 			if (sysinit != SET_BEGIN(sysinit_set))
221 				free(sysinit, M_TEMP);
222 			sysinit = newsysinit;
223 			sysinit_end = newsysinit_end;
224 			newsysinit = NULL;
225 			newsysinit_end = NULL;
226 			goto restart;
227 		}
228 	}
229 
230 	panic("Shouldn't get here!");
231 	/* NOTREACHED*/
232 }
233 
234 
235 /*
236  ***************************************************************************
237  ****
238  **** The following SYSINIT's belong elsewhere, but have not yet
239  **** been moved.
240  ****
241  ***************************************************************************
242  */
243 static void
244 print_caddr_t(void *data __unused)
245 {
246 	printf("%s", (char *)data);
247 }
248 SYSINIT(announce, SI_SUB_COPYRIGHT, SI_ORDER_FIRST, print_caddr_t, copyright)
249 SYSINIT(version, SI_SUB_COPYRIGHT, SI_ORDER_SECOND, print_caddr_t, version)
250 
251 static void
252 set_boot_verbose(void *data __unused)
253 {
254 
255 	if (boothowto & RB_VERBOSE)
256 		bootverbose++;
257 }
258 SYSINIT(boot_verbose, SI_SUB_TUNABLES, SI_ORDER_ANY, set_boot_verbose, NULL)
259 
260 struct sysentvec null_sysvec = {
261 	0,
262 	NULL,
263 	0,
264 	0,
265 	NULL,
266 	0,
267 	NULL,
268 	NULL,
269 	NULL,
270 	NULL,
271 	NULL,
272 	NULL,
273 	NULL,
274 	"null",
275 	NULL,
276 	NULL,
277 	0,
278 	PAGE_SIZE,
279 	VM_MIN_ADDRESS,
280 	VM_MAXUSER_ADDRESS,
281 	USRSTACK,
282 	PS_STRINGS,
283 	VM_PROT_ALL,
284 	NULL,
285 	NULL,
286 	NULL
287 };
288 
289 /*
290  ***************************************************************************
291  ****
292  **** The two following SYSINIT's are proc0 specific glue code.  I am not
293  **** convinced that they can not be safely combined, but their order of
294  **** operation has been maintained as the same as the original init_main.c
295  **** for right now.
296  ****
297  **** These probably belong in init_proc.c or kern_proc.c, since they
298  **** deal with proc0 (the fork template process).
299  ****
300  ***************************************************************************
301  */
302 /* ARGSUSED*/
303 static void
304 proc0_init(void *dummy __unused)
305 {
306 	register struct proc		*p;
307 	register struct filedesc0	*fdp;
308 	register unsigned i;
309 	struct thread *td;
310 	struct ksegrp *kg;
311 	struct kse *ke;
312 
313 	GIANT_REQUIRED;
314 	p = &proc0;
315 	td = &thread0;
316 	ke = &kse0;
317 	kg = &ksegrp0;
318 
319 	ke->ke_sched = kse0_sched;
320 	kg->kg_sched = ksegrp0_sched;
321 	p->p_sched = proc0_sched;
322 	td->td_sched = thread0_sched;
323 
324 	/*
325 	 * Initialize magic number.
326 	 */
327 	p->p_magic = P_MAGIC;
328 
329 	/*
330 	 * Initialize thread, process and pgrp structures.
331 	 */
332 	procinit();
333 	threadinit();
334 
335 	/*
336 	 * Initialize sleep queue hash table
337 	 */
338 	sleepinit();
339 
340 	/*
341 	 * additional VM structures
342 	 */
343 	vm_init2();
344 
345 	/*
346 	 * Create process 0 (the swapper).
347 	 */
348 	LIST_INSERT_HEAD(&allproc, p, p_list);
349 	LIST_INSERT_HEAD(PIDHASH(0), p, p_hash);
350 	mtx_init(&pgrp0.pg_mtx, "process group", NULL, MTX_DEF | MTX_DUPOK);
351 	p->p_pgrp = &pgrp0;
352 	LIST_INSERT_HEAD(PGRPHASH(0), &pgrp0, pg_hash);
353 	LIST_INIT(&pgrp0.pg_members);
354 	LIST_INSERT_HEAD(&pgrp0.pg_members, p, p_pglist);
355 
356 	pgrp0.pg_session = &session0;
357 	mtx_init(&session0.s_mtx, "session", NULL, MTX_DEF);
358 	session0.s_count = 1;
359 	session0.s_leader = p;
360 
361 	p->p_sysent = &null_sysvec;
362 
363 	/*
364 	 * proc_linkup was already done in init_i386() or alphainit() etc.
365 	 * because the earlier code needed to follow td->td_proc. Otherwise
366 	 * I would have done it here.. maybe this means this should be
367 	 * done earlier too.
368 	 */
369 	p->p_flag = P_SYSTEM;
370 	p->p_sflag = PS_INMEM;
371 	p->p_state = PRS_NORMAL;
372 	td->td_state = TDS_RUNNING;
373 	kg->kg_nice = NZERO;
374 	kg->kg_pri_class = PRI_TIMESHARE;
375 	kg->kg_user_pri = PUSER;
376 	td->td_priority = PVM;
377 	td->td_base_pri = PUSER;
378 	td->td_kse = ke; /* XXXKSE */
379 	td->td_oncpu = 0;
380 	ke->ke_state = KES_THREAD;
381 	ke->ke_thread = td;
382 	p->p_peers = 0;
383 	p->p_leader = p;
384 
385 
386 	bcopy("swapper", p->p_comm, sizeof ("swapper"));
387 
388 	callout_init(&p->p_itcallout, CALLOUT_MPSAFE);
389 	callout_init(&td->td_slpcallout, CALLOUT_MPSAFE);
390 
391 	/* Create credentials. */
392 	p->p_ucred = crget();
393 	p->p_ucred->cr_ngroups = 1;	/* group 0 */
394 	p->p_ucred->cr_uidinfo = uifind(0);
395 	p->p_ucred->cr_ruidinfo = uifind(0);
396 	p->p_ucred->cr_prison = NULL;	/* Don't jail it. */
397 #ifdef MAC
398 	mac_create_proc0(p->p_ucred);
399 #endif
400 	td->td_ucred = crhold(p->p_ucred);
401 
402 	/* Create sigacts. */
403 	p->p_sigacts = sigacts_alloc();
404 
405 	/* Initialize signal state for process 0. */
406 	siginit(&proc0);
407 
408 	/* Create the file descriptor table. */
409 	/* XXX this duplicates part of fdinit() */
410 	fdp = &filedesc0;
411 	p->p_fd = &fdp->fd_fd;
412 	p->p_fdtol = NULL;
413 	mtx_init(&fdp->fd_fd.fd_mtx, FILEDESC_LOCK_DESC, NULL, MTX_DEF);
414 	fdp->fd_fd.fd_refcnt = 1;
415 	fdp->fd_fd.fd_cmask = CMASK;
416 	fdp->fd_fd.fd_ofiles = fdp->fd_dfiles;
417 	fdp->fd_fd.fd_ofileflags = fdp->fd_dfileflags;
418 	fdp->fd_fd.fd_nfiles = NDFILE;
419 	fdp->fd_fd.fd_map = fdp->fd_dmap;
420 
421 	/* Create the limits structures. */
422 	p->p_limit = &limit0;
423 	for (i = 0; i < sizeof(p->p_rlimit)/sizeof(p->p_rlimit[0]); i++)
424 		limit0.pl_rlimit[i].rlim_cur =
425 		    limit0.pl_rlimit[i].rlim_max = RLIM_INFINITY;
426 	limit0.pl_rlimit[RLIMIT_NOFILE].rlim_cur =
427 	    limit0.pl_rlimit[RLIMIT_NOFILE].rlim_max = maxfiles;
428 	limit0.pl_rlimit[RLIMIT_NPROC].rlim_cur =
429 	    limit0.pl_rlimit[RLIMIT_NPROC].rlim_max = maxproc;
430 	i = ptoa(cnt.v_free_count);
431 	limit0.pl_rlimit[RLIMIT_RSS].rlim_max = i;
432 	limit0.pl_rlimit[RLIMIT_MEMLOCK].rlim_max = i;
433 	limit0.pl_rlimit[RLIMIT_MEMLOCK].rlim_cur = i / 3;
434 	limit0.p_refcnt = 1;
435 	p->p_cpulimit = RLIM_INFINITY;
436 
437 	/* Allocate a prototype map so we have something to fork. */
438 	pmap_pinit0(vmspace_pmap(&vmspace0));
439 	p->p_vmspace = &vmspace0;
440 	vmspace0.vm_refcnt = 1;
441 	vm_map_init(&vmspace0.vm_map, p->p_sysent->sv_minuser,
442 	    p->p_sysent->sv_maxuser);
443 	vmspace0.vm_map.pmap = vmspace_pmap(&vmspace0);
444 
445 	/*
446 	 * We continue to place resource usage info
447 	 * in the user struct so that it's pageable.
448 	 */
449 	p->p_stats = &p->p_uarea->u_stats;
450 
451 	/*
452 	 * Charge root for one process.
453 	 */
454 	(void)chgproccnt(p->p_ucred->cr_ruidinfo, 1, 0);
455 }
456 SYSINIT(p0init, SI_SUB_INTRINSIC, SI_ORDER_FIRST, proc0_init, NULL)
457 
458 /* ARGSUSED*/
459 static void
460 proc0_post(void *dummy __unused)
461 {
462 	struct timespec ts;
463 	struct proc *p;
464 
465 	/*
466 	 * Now we can look at the time, having had a chance to verify the
467 	 * time from the filesystem.  Pretend that proc0 started now.
468 	 */
469 	sx_slock(&allproc_lock);
470 	LIST_FOREACH(p, &allproc, p_list) {
471 		microuptime(&p->p_stats->p_start);
472 		p->p_runtime.sec = 0;
473 		p->p_runtime.frac = 0;
474 	}
475 	sx_sunlock(&allproc_lock);
476 	binuptime(PCPU_PTR(switchtime));
477 	PCPU_SET(switchticks, ticks);
478 
479 	/*
480 	 * Give the ``random'' number generator a thump.
481 	 */
482 	nanotime(&ts);
483 	srandom(ts.tv_sec ^ ts.tv_nsec);
484 }
485 SYSINIT(p0post, SI_SUB_INTRINSIC_POST, SI_ORDER_FIRST, proc0_post, NULL)
486 
487 /*
488  ***************************************************************************
489  ****
490  **** The following SYSINIT's and glue code should be moved to the
491  **** respective files on a per subsystem basis.
492  ****
493  ***************************************************************************
494  */
495 
496 
497 /*
498  ***************************************************************************
499  ****
500  **** The following code probably belongs in another file, like
501  **** kern/init_init.c.
502  ****
503  ***************************************************************************
504  */
505 
506 /*
507  * List of paths to try when searching for "init".
508  */
509 static char init_path[MAXPATHLEN] =
510 #ifdef	INIT_PATH
511     __XSTRING(INIT_PATH);
512 #else
513     "/sbin/init:/sbin/oinit:/sbin/init.bak:/stand/sysinstall";
514 #endif
515 SYSCTL_STRING(_kern, OID_AUTO, init_path, CTLFLAG_RD, init_path, 0,
516 	"Path used to search the init process");
517 
518 /*
519  * Start the initial user process; try exec'ing each pathname in init_path.
520  * The program is invoked with one argument containing the boot flags.
521  */
522 static void
523 start_init(void *dummy)
524 {
525 	vm_offset_t addr;
526 	struct execve_args args;
527 	int options, error;
528 	char *var, *path, *next, *s;
529 	char *ucp, **uap, *arg0, *arg1;
530 	struct thread *td;
531 	struct proc *p;
532 	int init_does_devfs = 0;
533 
534 	mtx_lock(&Giant);
535 
536 	GIANT_REQUIRED;
537 
538 	td = curthread;
539 	p = td->td_proc;
540 
541 	vfs_mountroot();
542 
543 	/* Get the vnode for '/'.  Set p->p_fd->fd_cdir to reference it. */
544 	if (VFS_ROOT(TAILQ_FIRST(&mountlist), &rootvnode))
545 		panic("cannot find root vnode");
546 	FILEDESC_LOCK(p->p_fd);
547 	p->p_fd->fd_cdir = rootvnode;
548 	VREF(p->p_fd->fd_cdir);
549 	p->p_fd->fd_rdir = rootvnode;
550 	VREF(p->p_fd->fd_rdir);
551 	FILEDESC_UNLOCK(p->p_fd);
552 	VOP_UNLOCK(rootvnode, 0, td);
553 #ifdef MAC
554 	mac_create_root_mount(td->td_ucred, TAILQ_FIRST(&mountlist));
555 #endif
556 
557 	/*
558 	 * For disk based systems, we probably cannot do this yet
559 	 * since the fs will be read-only.  But a NFS root
560 	 * might be ok.  It is worth a shot.
561 	 */
562 	error = kern_mkdir(td, "/dev", UIO_SYSSPACE, 0700);
563 	if (error == EEXIST)
564 		error = 0;
565 	if (error == 0)
566 		error = kernel_vmount(0, "fstype", "devfs",
567 		    "fspath", "/dev", NULL);
568 	if (error != 0)
569 		init_does_devfs = 1;
570 
571 	/*
572 	 * Need just enough stack to hold the faked-up "execve()" arguments.
573 	 */
574 	addr = p->p_sysent->sv_usrstack - PAGE_SIZE;
575 	if (vm_map_find(&p->p_vmspace->vm_map, NULL, 0, &addr, PAGE_SIZE,
576 			FALSE, VM_PROT_ALL, VM_PROT_ALL, 0) != 0)
577 		panic("init: couldn't allocate argument space");
578 	p->p_vmspace->vm_maxsaddr = (caddr_t)addr;
579 	p->p_vmspace->vm_ssize = 1;
580 
581 	if ((var = getenv("init_path")) != NULL) {
582 		strlcpy(init_path, var, sizeof(init_path));
583 		freeenv(var);
584 	}
585 
586 	for (path = init_path; *path != '\0'; path = next) {
587 		while (*path == ':')
588 			path++;
589 		if (*path == '\0')
590 			break;
591 		for (next = path; *next != '\0' && *next != ':'; next++)
592 			/* nothing */ ;
593 		if (bootverbose)
594 			printf("start_init: trying %.*s\n", (int)(next - path),
595 			    path);
596 
597 		/*
598 		 * Move out the boot flag argument.
599 		 */
600 		options = 0;
601 		ucp = (char *)p->p_sysent->sv_usrstack;
602 		(void)subyte(--ucp, 0);		/* trailing zero */
603 		if (boothowto & RB_SINGLE) {
604 			(void)subyte(--ucp, 's');
605 			options = 1;
606 		}
607 #ifdef notyet
608                 if (boothowto & RB_FASTBOOT) {
609 			(void)subyte(--ucp, 'f');
610 			options = 1;
611 		}
612 #endif
613 
614 #ifdef BOOTCDROM
615 		(void)subyte(--ucp, 'C');
616 		options = 1;
617 #endif
618 		if (init_does_devfs) {
619 			(void)subyte(--ucp, 'd');
620 			options = 1;
621 		}
622 
623 		if (options == 0)
624 			(void)subyte(--ucp, '-');
625 		(void)subyte(--ucp, '-');		/* leading hyphen */
626 		arg1 = ucp;
627 
628 		/*
629 		 * Move out the file name (also arg 0).
630 		 */
631 		(void)subyte(--ucp, 0);
632 		for (s = next - 1; s >= path; s--)
633 			(void)subyte(--ucp, *s);
634 		arg0 = ucp;
635 
636 		/*
637 		 * Move out the arg pointers.
638 		 */
639 		uap = (char **)((intptr_t)ucp & ~(sizeof(intptr_t)-1));
640 		(void)suword((caddr_t)--uap, (long)0);	/* terminator */
641 		(void)suword((caddr_t)--uap, (long)(intptr_t)arg1);
642 		(void)suword((caddr_t)--uap, (long)(intptr_t)arg0);
643 
644 		/*
645 		 * Point at the arguments.
646 		 */
647 		args.fname = arg0;
648 		args.argv = uap;
649 		args.envv = NULL;
650 
651 		/*
652 		 * Now try to exec the program.  If can't for any reason
653 		 * other than it doesn't exist, complain.
654 		 *
655 		 * Otherwise, return via fork_trampoline() all the way
656 		 * to user mode as init!
657 		 */
658 		if ((error = execve(td, &args)) == 0) {
659 			mtx_unlock(&Giant);
660 			return;
661 		}
662 		if (error != ENOENT)
663 			printf("exec %.*s: error %d\n", (int)(next - path),
664 			    path, error);
665 	}
666 	printf("init: not found in path %s\n", init_path);
667 	panic("no init");
668 }
669 
670 /*
671  * Like kthread_create(), but runs in it's own address space.
672  * We do this early to reserve pid 1.
673  *
674  * Note special case - do not make it runnable yet.  Other work
675  * in progress will change this more.
676  */
677 static void
678 create_init(const void *udata __unused)
679 {
680 	struct ucred *newcred, *oldcred;
681 	int error;
682 
683 	error = fork1(&thread0, RFFDG | RFPROC | RFSTOPPED, 0, &initproc);
684 	if (error)
685 		panic("cannot fork init: %d\n", error);
686 	KASSERT(initproc->p_pid == 1, ("create_init: initproc->p_pid != 1"));
687 	/* divorce init's credentials from the kernel's */
688 	newcred = crget();
689 	PROC_LOCK(initproc);
690 	initproc->p_flag |= P_SYSTEM;
691 	oldcred = initproc->p_ucred;
692 	crcopy(newcred, oldcred);
693 #ifdef MAC
694 	mac_create_proc1(newcred);
695 #endif
696 	initproc->p_ucred = newcred;
697 	PROC_UNLOCK(initproc);
698 	crfree(oldcred);
699 	cred_update_thread(FIRST_THREAD_IN_PROC(initproc));
700 	mtx_lock_spin(&sched_lock);
701 	initproc->p_sflag |= PS_INMEM;
702 	mtx_unlock_spin(&sched_lock);
703 	cpu_set_fork_handler(FIRST_THREAD_IN_PROC(initproc), start_init, NULL);
704 }
705 SYSINIT(init, SI_SUB_CREATE_INIT, SI_ORDER_FIRST, create_init, NULL)
706 
707 /*
708  * Make it runnable now.
709  */
710 static void
711 kick_init(const void *udata __unused)
712 {
713 	struct thread *td;
714 
715 	td = FIRST_THREAD_IN_PROC(initproc);
716 	mtx_lock_spin(&sched_lock);
717 	TD_SET_CAN_RUN(td);
718 	setrunqueue(td);	/* XXXKSE */
719 	mtx_unlock_spin(&sched_lock);
720 }
721 SYSINIT(kickinit, SI_SUB_KTHREAD_INIT, SI_ORDER_FIRST, kick_init, NULL)
722