xref: /freebsd/sys/kern/init_main.c (revision f6a4109212fd8fbabc731f07b2dd5c7e07fbec33)
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 struct	vmspace vmspace0;
97 struct	proc *initproc;
98 
99 struct	vnode *rootvp;
100 int	boothowto = 0;		/* initialized so that it can be patched */
101 SYSCTL_INT(_debug, OID_AUTO, boothowto, CTLFLAG_RD, &boothowto, 0, "");
102 int	bootverbose;
103 SYSCTL_INT(_debug, OID_AUTO, bootverbose, CTLFLAG_RW, &bootverbose, 0, "");
104 
105 /*
106  * This ensures that there is at least one entry so that the sysinit_set
107  * symbol is not undefined.  A sybsystem ID of SI_SUB_DUMMY is never
108  * executed.
109  */
110 SYSINIT(placeholder, SI_SUB_DUMMY, SI_ORDER_ANY, NULL, NULL)
111 
112 /*
113  * The sysinit table itself.  Items are checked off as the are run.
114  * If we want to register new sysinit types, add them to newsysinit.
115  */
116 SET_DECLARE(sysinit_set, struct sysinit);
117 struct sysinit **sysinit, **sysinit_end;
118 struct sysinit **newsysinit, **newsysinit_end;
119 
120 /*
121  * Merge a new sysinit set into the current set, reallocating it if
122  * necessary.  This can only be called after malloc is running.
123  */
124 void
125 sysinit_add(struct sysinit **set, struct sysinit **set_end)
126 {
127 	struct sysinit **newset;
128 	struct sysinit **sipp;
129 	struct sysinit **xipp;
130 	int count;
131 
132 	count = set_end - set;
133 	if (newsysinit)
134 		count += newsysinit_end - newsysinit;
135 	else
136 		count += sysinit_end - sysinit;
137 	newset = malloc(count * sizeof(*sipp), M_TEMP, M_NOWAIT);
138 	if (newset == NULL)
139 		panic("cannot malloc for sysinit");
140 	xipp = newset;
141 	if (newsysinit)
142 		for (sipp = newsysinit; sipp < newsysinit_end; sipp++)
143 			*xipp++ = *sipp;
144 	else
145 		for (sipp = sysinit; sipp < sysinit_end; sipp++)
146 			*xipp++ = *sipp;
147 	for (sipp = set; sipp < set_end; sipp++)
148 		*xipp++ = *sipp;
149 	if (newsysinit)
150 		free(newsysinit, M_TEMP);
151 	newsysinit = newset;
152 	newsysinit_end = newset + count;
153 }
154 
155 /*
156  * System startup; initialize the world, create process 0, mount root
157  * filesystem, and fork to create init and pagedaemon.  Most of the
158  * hard work is done in the lower-level initialization routines including
159  * startup(), which does memory initialization and autoconfiguration.
160  *
161  * This allows simple addition of new kernel subsystems that require
162  * boot time initialization.  It also allows substitution of subsystem
163  * (for instance, a scheduler, kernel profiler, or VM system) by object
164  * module.  Finally, it allows for optional "kernel threads".
165  */
166 void
167 mi_startup(void)
168 {
169 
170 	register struct sysinit **sipp;		/* system initialization*/
171 	register struct sysinit **xipp;		/* interior loop of sort*/
172 	register struct sysinit *save;		/* bubble*/
173 
174 	if (sysinit == NULL) {
175 		sysinit = SET_BEGIN(sysinit_set);
176 		sysinit_end = SET_LIMIT(sysinit_set);
177 	}
178 
179 restart:
180 	/*
181 	 * Perform a bubble sort of the system initialization objects by
182 	 * their subsystem (primary key) and order (secondary key).
183 	 */
184 	for (sipp = sysinit; sipp < sysinit_end; sipp++) {
185 		for (xipp = sipp + 1; xipp < sysinit_end; xipp++) {
186 			if ((*sipp)->subsystem < (*xipp)->subsystem ||
187 			     ((*sipp)->subsystem == (*xipp)->subsystem &&
188 			      (*sipp)->order <= (*xipp)->order))
189 				continue;	/* skip*/
190 			save = *sipp;
191 			*sipp = *xipp;
192 			*xipp = save;
193 		}
194 	}
195 
196 	/*
197 	 * Traverse the (now) ordered list of system initialization tasks.
198 	 * Perform each task, and continue on to the next task.
199 	 *
200 	 * The last item on the list is expected to be the scheduler,
201 	 * which will not return.
202 	 */
203 	for (sipp = sysinit; sipp < sysinit_end; sipp++) {
204 
205 		if ((*sipp)->subsystem == SI_SUB_DUMMY)
206 			continue;	/* skip dummy task(s)*/
207 
208 		if ((*sipp)->subsystem == SI_SUB_DONE)
209 			continue;
210 
211 		/* Call function */
212 		(*((*sipp)->func))((*sipp)->udata);
213 
214 		/* Check off the one we're just done */
215 		(*sipp)->subsystem = SI_SUB_DONE;
216 
217 		/* Check if we've installed more sysinit items via KLD */
218 		if (newsysinit != NULL) {
219 			if (sysinit != SET_BEGIN(sysinit_set))
220 				free(sysinit, M_TEMP);
221 			sysinit = newsysinit;
222 			sysinit_end = newsysinit_end;
223 			newsysinit = NULL;
224 			newsysinit_end = NULL;
225 			goto restart;
226 		}
227 	}
228 
229 	panic("Shouldn't get here!");
230 	/* NOTREACHED*/
231 }
232 
233 
234 /*
235  ***************************************************************************
236  ****
237  **** The following SYSINIT's belong elsewhere, but have not yet
238  **** been moved.
239  ****
240  ***************************************************************************
241  */
242 static void
243 print_caddr_t(void *data __unused)
244 {
245 	printf("%s", (char *)data);
246 }
247 SYSINIT(announce, SI_SUB_COPYRIGHT, SI_ORDER_FIRST, print_caddr_t, copyright)
248 SYSINIT(version, SI_SUB_COPYRIGHT, SI_ORDER_SECOND, print_caddr_t, version)
249 
250 static void
251 set_boot_verbose(void *data __unused)
252 {
253 
254 	if (boothowto & RB_VERBOSE)
255 		bootverbose++;
256 }
257 SYSINIT(boot_verbose, SI_SUB_TUNABLES, SI_ORDER_ANY, set_boot_verbose, NULL)
258 
259 struct sysentvec null_sysvec = {
260 	0,
261 	NULL,
262 	0,
263 	0,
264 	NULL,
265 	0,
266 	NULL,
267 	NULL,
268 	NULL,
269 	NULL,
270 	NULL,
271 	NULL,
272 	NULL,
273 	"null",
274 	NULL,
275 	NULL,
276 	0,
277 	PAGE_SIZE,
278 	VM_MIN_ADDRESS,
279 	VM_MAXUSER_ADDRESS,
280 	USRSTACK,
281 	PS_STRINGS,
282 	VM_PROT_ALL,
283 	NULL,
284 	NULL,
285 	NULL
286 };
287 
288 /*
289  ***************************************************************************
290  ****
291  **** The two following SYSINIT's are proc0 specific glue code.  I am not
292  **** convinced that they can not be safely combined, but their order of
293  **** operation has been maintained as the same as the original init_main.c
294  **** for right now.
295  ****
296  **** These probably belong in init_proc.c or kern_proc.c, since they
297  **** deal with proc0 (the fork template process).
298  ****
299  ***************************************************************************
300  */
301 /* ARGSUSED*/
302 static void
303 proc0_init(void *dummy __unused)
304 {
305 	register struct proc		*p;
306 	register struct filedesc0	*fdp;
307 	register unsigned i;
308 	struct thread *td;
309 	struct ksegrp *kg;
310 	struct kse *ke;
311 
312 	GIANT_REQUIRED;
313 	p = &proc0;
314 	td = &thread0;
315 	ke = &kse0;
316 	kg = &ksegrp0;
317 
318 	ke->ke_sched = kse0_sched;
319 	kg->kg_sched = ksegrp0_sched;
320 	p->p_sched = proc0_sched;
321 	td->td_sched = thread0_sched;
322 
323 	/*
324 	 * Initialize magic number.
325 	 */
326 	p->p_magic = P_MAGIC;
327 
328 	/*
329 	 * Initialize thread, process and pgrp structures.
330 	 */
331 	procinit();
332 	threadinit();
333 
334 	/*
335 	 * Initialize sleep queue hash table
336 	 */
337 	sleepinit();
338 
339 	/*
340 	 * additional VM structures
341 	 */
342 	vm_init2();
343 
344 	/*
345 	 * Create process 0 (the swapper).
346 	 */
347 	LIST_INSERT_HEAD(&allproc, p, p_list);
348 	LIST_INSERT_HEAD(PIDHASH(0), p, p_hash);
349 	mtx_init(&pgrp0.pg_mtx, "process group", NULL, MTX_DEF | MTX_DUPOK);
350 	p->p_pgrp = &pgrp0;
351 	LIST_INSERT_HEAD(PGRPHASH(0), &pgrp0, pg_hash);
352 	LIST_INIT(&pgrp0.pg_members);
353 	LIST_INSERT_HEAD(&pgrp0.pg_members, p, p_pglist);
354 
355 	pgrp0.pg_session = &session0;
356 	mtx_init(&session0.s_mtx, "session", NULL, MTX_DEF);
357 	session0.s_count = 1;
358 	session0.s_leader = p;
359 
360 	p->p_sysent = &null_sysvec;
361 
362 	/*
363 	 * proc_linkup was already done in init_i386() or alphainit() etc.
364 	 * because the earlier code needed to follow td->td_proc. Otherwise
365 	 * I would have done it here.. maybe this means this should be
366 	 * done earlier too.
367 	 */
368 	p->p_flag = P_SYSTEM;
369 	p->p_sflag = PS_INMEM;
370 	p->p_state = PRS_NORMAL;
371 	td->td_state = TDS_RUNNING;
372 	kg->kg_nice = NZERO;
373 	kg->kg_pri_class = PRI_TIMESHARE;
374 	kg->kg_user_pri = PUSER;
375 	td->td_priority = PVM;
376 	td->td_base_pri = PUSER;
377 	td->td_kse = ke; /* XXXKSE */
378 	td->td_oncpu = 0;
379 	ke->ke_state = KES_THREAD;
380 	ke->ke_thread = td;
381 	p->p_peers = 0;
382 	p->p_leader = p;
383 
384 
385 	bcopy("swapper", p->p_comm, sizeof ("swapper"));
386 
387 	callout_init(&p->p_itcallout, CALLOUT_MPSAFE);
388 	callout_init(&td->td_slpcallout, CALLOUT_MPSAFE);
389 
390 	/* Create credentials. */
391 	p->p_ucred = crget();
392 	p->p_ucred->cr_ngroups = 1;	/* group 0 */
393 	p->p_ucred->cr_uidinfo = uifind(0);
394 	p->p_ucred->cr_ruidinfo = uifind(0);
395 	p->p_ucred->cr_prison = NULL;	/* Don't jail it. */
396 #ifdef MAC
397 	mac_create_proc0(p->p_ucred);
398 #endif
399 	td->td_ucred = crhold(p->p_ucred);
400 
401 	/* Create sigacts. */
402 	p->p_sigacts = sigacts_alloc();
403 
404 	/* Initialize signal state for process 0. */
405 	siginit(&proc0);
406 
407 	/* Create the file descriptor table. */
408 	/* XXX this duplicates part of fdinit() */
409 	fdp = &filedesc0;
410 	p->p_fd = &fdp->fd_fd;
411 	p->p_fdtol = NULL;
412 	mtx_init(&fdp->fd_fd.fd_mtx, FILEDESC_LOCK_DESC, NULL, MTX_DEF);
413 	fdp->fd_fd.fd_refcnt = 1;
414 	fdp->fd_fd.fd_cmask = CMASK;
415 	fdp->fd_fd.fd_ofiles = fdp->fd_dfiles;
416 	fdp->fd_fd.fd_ofileflags = fdp->fd_dfileflags;
417 	fdp->fd_fd.fd_nfiles = NDFILE;
418 	fdp->fd_fd.fd_map = fdp->fd_dmap;
419 
420 	/* Create the limits structures. */
421 	p->p_limit = lim_alloc();
422 	for (i = 0; i < RLIM_NLIMITS; i++)
423 		p->p_limit->pl_rlimit[i].rlim_cur =
424 		    p->p_limit->pl_rlimit[i].rlim_max = RLIM_INFINITY;
425 	p->p_limit->pl_rlimit[RLIMIT_NOFILE].rlim_cur =
426 	    p->p_limit->pl_rlimit[RLIMIT_NOFILE].rlim_max = maxfiles;
427 	p->p_limit->pl_rlimit[RLIMIT_NPROC].rlim_cur =
428 	    p->p_limit->pl_rlimit[RLIMIT_NPROC].rlim_max = maxproc;
429 	i = ptoa(cnt.v_free_count);
430 	p->p_limit->pl_rlimit[RLIMIT_RSS].rlim_max = i;
431 	p->p_limit->pl_rlimit[RLIMIT_MEMLOCK].rlim_max = i;
432 	p->p_limit->pl_rlimit[RLIMIT_MEMLOCK].rlim_cur = i / 3;
433 	p->p_cpulimit = RLIM_INFINITY;
434 
435 	/* Allocate a prototype map so we have something to fork. */
436 	pmap_pinit0(vmspace_pmap(&vmspace0));
437 	p->p_vmspace = &vmspace0;
438 	vmspace0.vm_refcnt = 1;
439 	vm_map_init(&vmspace0.vm_map, p->p_sysent->sv_minuser,
440 	    p->p_sysent->sv_maxuser);
441 	vmspace0.vm_map.pmap = vmspace_pmap(&vmspace0);
442 
443 	/*
444 	 * We continue to place resource usage info
445 	 * in the user struct so that it's pageable.
446 	 */
447 	p->p_stats = &p->p_uarea->u_stats;
448 
449 	/*
450 	 * Charge root for one process.
451 	 */
452 	(void)chgproccnt(p->p_ucred->cr_ruidinfo, 1, 0);
453 }
454 SYSINIT(p0init, SI_SUB_INTRINSIC, SI_ORDER_FIRST, proc0_init, NULL)
455 
456 /* ARGSUSED*/
457 static void
458 proc0_post(void *dummy __unused)
459 {
460 	struct timespec ts;
461 	struct proc *p;
462 
463 	/*
464 	 * Now we can look at the time, having had a chance to verify the
465 	 * time from the filesystem.  Pretend that proc0 started now.
466 	 */
467 	sx_slock(&allproc_lock);
468 	LIST_FOREACH(p, &allproc, p_list) {
469 		microuptime(&p->p_stats->p_start);
470 		p->p_runtime.sec = 0;
471 		p->p_runtime.frac = 0;
472 	}
473 	sx_sunlock(&allproc_lock);
474 	binuptime(PCPU_PTR(switchtime));
475 	PCPU_SET(switchticks, ticks);
476 
477 	/*
478 	 * Give the ``random'' number generator a thump.
479 	 */
480 	nanotime(&ts);
481 	srandom(ts.tv_sec ^ ts.tv_nsec);
482 }
483 SYSINIT(p0post, SI_SUB_INTRINSIC_POST, SI_ORDER_FIRST, proc0_post, NULL)
484 
485 /*
486  ***************************************************************************
487  ****
488  **** The following SYSINIT's and glue code should be moved to the
489  **** respective files on a per subsystem basis.
490  ****
491  ***************************************************************************
492  */
493 
494 
495 /*
496  ***************************************************************************
497  ****
498  **** The following code probably belongs in another file, like
499  **** kern/init_init.c.
500  ****
501  ***************************************************************************
502  */
503 
504 /*
505  * List of paths to try when searching for "init".
506  */
507 static char init_path[MAXPATHLEN] =
508 #ifdef	INIT_PATH
509     __XSTRING(INIT_PATH);
510 #else
511     "/sbin/init:/sbin/oinit:/sbin/init.bak:/stand/sysinstall";
512 #endif
513 SYSCTL_STRING(_kern, OID_AUTO, init_path, CTLFLAG_RD, init_path, 0,
514 	"Path used to search the init process");
515 
516 /*
517  * Start the initial user process; try exec'ing each pathname in init_path.
518  * The program is invoked with one argument containing the boot flags.
519  */
520 static void
521 start_init(void *dummy)
522 {
523 	vm_offset_t addr;
524 	struct execve_args args;
525 	int options, error;
526 	char *var, *path, *next, *s;
527 	char *ucp, **uap, *arg0, *arg1;
528 	struct thread *td;
529 	struct proc *p;
530 	int init_does_devfs = 0;
531 
532 	mtx_lock(&Giant);
533 
534 	GIANT_REQUIRED;
535 
536 	td = curthread;
537 	p = td->td_proc;
538 
539 	vfs_mountroot();
540 
541 	/* Get the vnode for '/'.  Set p->p_fd->fd_cdir to reference it. */
542 	if (VFS_ROOT(TAILQ_FIRST(&mountlist), &rootvnode))
543 		panic("cannot find root vnode");
544 	FILEDESC_LOCK(p->p_fd);
545 	p->p_fd->fd_cdir = rootvnode;
546 	VREF(p->p_fd->fd_cdir);
547 	p->p_fd->fd_rdir = rootvnode;
548 	VREF(p->p_fd->fd_rdir);
549 	FILEDESC_UNLOCK(p->p_fd);
550 	VOP_UNLOCK(rootvnode, 0, td);
551 #ifdef MAC
552 	mac_create_root_mount(td->td_ucred, TAILQ_FIRST(&mountlist));
553 #endif
554 
555 	/*
556 	 * For disk based systems, we probably cannot do this yet
557 	 * since the fs will be read-only.  But a NFS root
558 	 * might be ok.  It is worth a shot.
559 	 */
560 	error = kern_mkdir(td, "/dev", UIO_SYSSPACE, 0700);
561 	if (error == EEXIST)
562 		error = 0;
563 	if (error == 0)
564 		error = kernel_vmount(0, "fstype", "devfs",
565 		    "fspath", "/dev", NULL);
566 	if (error != 0)
567 		init_does_devfs = 1;
568 
569 	/*
570 	 * Need just enough stack to hold the faked-up "execve()" arguments.
571 	 */
572 	addr = p->p_sysent->sv_usrstack - PAGE_SIZE;
573 	if (vm_map_find(&p->p_vmspace->vm_map, NULL, 0, &addr, PAGE_SIZE,
574 			FALSE, VM_PROT_ALL, VM_PROT_ALL, 0) != 0)
575 		panic("init: couldn't allocate argument space");
576 	p->p_vmspace->vm_maxsaddr = (caddr_t)addr;
577 	p->p_vmspace->vm_ssize = 1;
578 
579 	if ((var = getenv("init_path")) != NULL) {
580 		strlcpy(init_path, var, sizeof(init_path));
581 		freeenv(var);
582 	}
583 
584 	for (path = init_path; *path != '\0'; path = next) {
585 		while (*path == ':')
586 			path++;
587 		if (*path == '\0')
588 			break;
589 		for (next = path; *next != '\0' && *next != ':'; next++)
590 			/* nothing */ ;
591 		if (bootverbose)
592 			printf("start_init: trying %.*s\n", (int)(next - path),
593 			    path);
594 
595 		/*
596 		 * Move out the boot flag argument.
597 		 */
598 		options = 0;
599 		ucp = (char *)p->p_sysent->sv_usrstack;
600 		(void)subyte(--ucp, 0);		/* trailing zero */
601 		if (boothowto & RB_SINGLE) {
602 			(void)subyte(--ucp, 's');
603 			options = 1;
604 		}
605 #ifdef notyet
606                 if (boothowto & RB_FASTBOOT) {
607 			(void)subyte(--ucp, 'f');
608 			options = 1;
609 		}
610 #endif
611 
612 #ifdef BOOTCDROM
613 		(void)subyte(--ucp, 'C');
614 		options = 1;
615 #endif
616 		if (init_does_devfs) {
617 			(void)subyte(--ucp, 'd');
618 			options = 1;
619 		}
620 
621 		if (options == 0)
622 			(void)subyte(--ucp, '-');
623 		(void)subyte(--ucp, '-');		/* leading hyphen */
624 		arg1 = ucp;
625 
626 		/*
627 		 * Move out the file name (also arg 0).
628 		 */
629 		(void)subyte(--ucp, 0);
630 		for (s = next - 1; s >= path; s--)
631 			(void)subyte(--ucp, *s);
632 		arg0 = ucp;
633 
634 		/*
635 		 * Move out the arg pointers.
636 		 */
637 		uap = (char **)((intptr_t)ucp & ~(sizeof(intptr_t)-1));
638 		(void)suword((caddr_t)--uap, (long)0);	/* terminator */
639 		(void)suword((caddr_t)--uap, (long)(intptr_t)arg1);
640 		(void)suword((caddr_t)--uap, (long)(intptr_t)arg0);
641 
642 		/*
643 		 * Point at the arguments.
644 		 */
645 		args.fname = arg0;
646 		args.argv = uap;
647 		args.envv = NULL;
648 
649 		/*
650 		 * Now try to exec the program.  If can't for any reason
651 		 * other than it doesn't exist, complain.
652 		 *
653 		 * Otherwise, return via fork_trampoline() all the way
654 		 * to user mode as init!
655 		 */
656 		if ((error = execve(td, &args)) == 0) {
657 			mtx_unlock(&Giant);
658 			return;
659 		}
660 		if (error != ENOENT)
661 			printf("exec %.*s: error %d\n", (int)(next - path),
662 			    path, error);
663 	}
664 	printf("init: not found in path %s\n", init_path);
665 	panic("no init");
666 }
667 
668 /*
669  * Like kthread_create(), but runs in it's own address space.
670  * We do this early to reserve pid 1.
671  *
672  * Note special case - do not make it runnable yet.  Other work
673  * in progress will change this more.
674  */
675 static void
676 create_init(const void *udata __unused)
677 {
678 	struct ucred *newcred, *oldcred;
679 	int error;
680 
681 	error = fork1(&thread0, RFFDG | RFPROC | RFSTOPPED, 0, &initproc);
682 	if (error)
683 		panic("cannot fork init: %d\n", error);
684 	KASSERT(initproc->p_pid == 1, ("create_init: initproc->p_pid != 1"));
685 	/* divorce init's credentials from the kernel's */
686 	newcred = crget();
687 	PROC_LOCK(initproc);
688 	initproc->p_flag |= P_SYSTEM;
689 	oldcred = initproc->p_ucred;
690 	crcopy(newcred, oldcred);
691 #ifdef MAC
692 	mac_create_proc1(newcred);
693 #endif
694 	initproc->p_ucred = newcred;
695 	PROC_UNLOCK(initproc);
696 	crfree(oldcred);
697 	cred_update_thread(FIRST_THREAD_IN_PROC(initproc));
698 	mtx_lock_spin(&sched_lock);
699 	initproc->p_sflag |= PS_INMEM;
700 	mtx_unlock_spin(&sched_lock);
701 	cpu_set_fork_handler(FIRST_THREAD_IN_PROC(initproc), start_init, NULL);
702 }
703 SYSINIT(init, SI_SUB_CREATE_INIT, SI_ORDER_FIRST, create_init, NULL)
704 
705 /*
706  * Make it runnable now.
707  */
708 static void
709 kick_init(const void *udata __unused)
710 {
711 	struct thread *td;
712 
713 	td = FIRST_THREAD_IN_PROC(initproc);
714 	mtx_lock_spin(&sched_lock);
715 	TD_SET_CAN_RUN(td);
716 	setrunqueue(td);	/* XXXKSE */
717 	mtx_unlock_spin(&sched_lock);
718 }
719 SYSINIT(kickinit, SI_SUB_KTHREAD_INIT, SI_ORDER_FIRST, kick_init, NULL)
720