xref: /freebsd/sys/kern/init_main.c (revision 06a31d6a6779b74405b41c8ad4579b5d81db4d4c)
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  * $FreeBSD$
43  */
44 
45 #include "opt_init_path.h"
46 #include "opt_mac.h"
47 
48 #include <sys/param.h>
49 #include <sys/kernel.h>
50 #include <sys/exec.h>
51 #include <sys/file.h>
52 #include <sys/filedesc.h>
53 #include <sys/ktr.h>
54 #include <sys/lock.h>
55 #include <sys/mac.h>
56 #include <sys/mount.h>
57 #include <sys/mutex.h>
58 #include <sys/syscallsubr.h>
59 #include <sys/sysctl.h>
60 #include <sys/proc.h>
61 #include <sys/resourcevar.h>
62 #include <sys/systm.h>
63 #include <sys/signalvar.h>
64 #include <sys/vnode.h>
65 #include <sys/sysent.h>
66 #include <sys/reboot.h>
67 #include <sys/sched.h>
68 #include <sys/sx.h>
69 #include <sys/sysproto.h>
70 #include <sys/vmmeter.h>
71 #include <sys/unistd.h>
72 #include <sys/malloc.h>
73 #include <sys/conf.h>
74 
75 #include <machine/cpu.h>
76 
77 #include <vm/vm.h>
78 #include <vm/vm_param.h>
79 #include <vm/pmap.h>
80 #include <vm/vm_map.h>
81 #include <sys/user.h>
82 #include <sys/copyright.h>
83 
84 void mi_startup(void);				/* Should be elsewhere */
85 
86 /* Components of the first process -- never freed. */
87 static struct session session0;
88 static struct pgrp pgrp0;
89 struct	proc proc0;
90 struct	thread thread0;
91 struct	kse kse0;
92 struct	ksegrp ksegrp0;
93 static struct filedesc0 filedesc0;
94 static struct plimit limit0;
95 struct	vmspace vmspace0;
96 struct	proc *initproc;
97 
98 int cmask = CMASK;
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 };
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, 1);
388 	callout_init(&td->td_slpcallout, 1);
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 	fdp = &filedesc0;
409 	p->p_fd = &fdp->fd_fd;
410 	p->p_fdtol = NULL;
411 	mtx_init(&fdp->fd_fd.fd_mtx, FILEDESC_LOCK_DESC, NULL, MTX_DEF);
412 	fdp->fd_fd.fd_refcnt = 1;
413 	fdp->fd_fd.fd_cmask = cmask;
414 	fdp->fd_fd.fd_ofiles = fdp->fd_dfiles;
415 	fdp->fd_fd.fd_ofileflags = fdp->fd_dfileflags;
416 	fdp->fd_fd.fd_nfiles = NDFILE;
417 
418 	/* Create the limits structures. */
419 	p->p_limit = &limit0;
420 	for (i = 0; i < sizeof(p->p_rlimit)/sizeof(p->p_rlimit[0]); i++)
421 		limit0.pl_rlimit[i].rlim_cur =
422 		    limit0.pl_rlimit[i].rlim_max = RLIM_INFINITY;
423 	limit0.pl_rlimit[RLIMIT_NOFILE].rlim_cur =
424 	    limit0.pl_rlimit[RLIMIT_NOFILE].rlim_max = maxfiles;
425 	limit0.pl_rlimit[RLIMIT_NPROC].rlim_cur =
426 	    limit0.pl_rlimit[RLIMIT_NPROC].rlim_max = maxproc;
427 	i = ptoa(cnt.v_free_count);
428 	limit0.pl_rlimit[RLIMIT_RSS].rlim_max = i;
429 	limit0.pl_rlimit[RLIMIT_MEMLOCK].rlim_max = i;
430 	limit0.pl_rlimit[RLIMIT_MEMLOCK].rlim_cur = i / 3;
431 	limit0.p_refcnt = 1;
432 	p->p_cpulimit = RLIM_INFINITY;
433 
434 	/* Allocate a prototype map so we have something to fork. */
435 	pmap_pinit0(vmspace_pmap(&vmspace0));
436 	p->p_vmspace = &vmspace0;
437 	vmspace0.vm_refcnt = 1;
438 	vm_map_init(&vmspace0.vm_map, p->p_sysent->sv_minuser,
439 	    p->p_sysent->sv_maxuser);
440 	vmspace0.vm_map.pmap = vmspace_pmap(&vmspace0);
441 
442 	/*
443 	 * We continue to place resource usage info
444 	 * in the user struct so that it's pageable.
445 	 */
446 	p->p_stats = &p->p_uarea->u_stats;
447 
448 	/*
449 	 * Charge root for one process.
450 	 */
451 	(void)chgproccnt(p->p_ucred->cr_ruidinfo, 1, 0);
452 }
453 SYSINIT(p0init, SI_SUB_INTRINSIC, SI_ORDER_FIRST, proc0_init, NULL)
454 
455 /* ARGSUSED*/
456 static void
457 proc0_post(void *dummy __unused)
458 {
459 	struct timespec ts;
460 	struct proc *p;
461 
462 	/*
463 	 * Now we can look at the time, having had a chance to verify the
464 	 * time from the filesystem.  Pretend that proc0 started now.
465 	 */
466 	sx_slock(&allproc_lock);
467 	LIST_FOREACH(p, &allproc, p_list) {
468 		microuptime(&p->p_stats->p_start);
469 		p->p_runtime.sec = 0;
470 		p->p_runtime.frac = 0;
471 	}
472 	sx_sunlock(&allproc_lock);
473 	binuptime(PCPU_PTR(switchtime));
474 	PCPU_SET(switchticks, ticks);
475 
476 	/*
477 	 * Give the ``random'' number generator a thump.
478 	 */
479 	nanotime(&ts);
480 	srandom(ts.tv_sec ^ ts.tv_nsec);
481 }
482 SYSINIT(p0post, SI_SUB_INTRINSIC_POST, SI_ORDER_FIRST, proc0_post, NULL)
483 
484 /*
485  ***************************************************************************
486  ****
487  **** The following SYSINIT's and glue code should be moved to the
488  **** respective files on a per subsystem basis.
489  ****
490  ***************************************************************************
491  */
492 
493 
494 /*
495  ***************************************************************************
496  ****
497  **** The following code probably belongs in another file, like
498  **** kern/init_init.c.
499  ****
500  ***************************************************************************
501  */
502 
503 /*
504  * List of paths to try when searching for "init".
505  */
506 static char init_path[MAXPATHLEN] =
507 #ifdef	INIT_PATH
508     __XSTRING(INIT_PATH);
509 #else
510     "/sbin/init:/sbin/oinit:/sbin/init.bak:/stand/sysinstall";
511 #endif
512 SYSCTL_STRING(_kern, OID_AUTO, init_path, CTLFLAG_RD, init_path, 0,
513 	"Path used to search the init process");
514 
515 /*
516  * Start the initial user process; try exec'ing each pathname in init_path.
517  * The program is invoked with one argument containing the boot flags.
518  */
519 static void
520 start_init(void *dummy)
521 {
522 	vm_offset_t addr;
523 	struct execve_args args;
524 	int options, error;
525 	char *var, *path, *next, *s;
526 	char *ucp, **uap, *arg0, *arg1;
527 	struct thread *td;
528 	struct proc *p;
529 	int init_does_devfs = 0;
530 
531 	mtx_lock(&Giant);
532 
533 	GIANT_REQUIRED;
534 
535 	td = curthread;
536 	p = td->td_proc;
537 
538 	vfs_mountroot();
539 
540 	/* Get the vnode for '/'.  Set p->p_fd->fd_cdir to reference it. */
541 	if (VFS_ROOT(TAILQ_FIRST(&mountlist), &rootvnode))
542 		panic("cannot find root vnode");
543 	FILEDESC_LOCK(p->p_fd);
544 	p->p_fd->fd_cdir = rootvnode;
545 	VREF(p->p_fd->fd_cdir);
546 	p->p_fd->fd_rdir = rootvnode;
547 	VREF(p->p_fd->fd_rdir);
548 	FILEDESC_UNLOCK(p->p_fd);
549 	VOP_UNLOCK(rootvnode, 0, td);
550 #ifdef MAC
551 	mac_create_root_mount(td->td_ucred, TAILQ_FIRST(&mountlist));
552 #endif
553 
554 	/*
555 	 * For disk based systems, we probably cannot do this yet
556 	 * since the fs will be read-only.  But a NFS root
557 	 * might be ok.  It is worth a shot.
558 	 */
559 	error = kern_mkdir(td, "/dev", UIO_SYSSPACE, 0700);
560 	if (error == EEXIST)
561 		error = 0;
562 	if (error == 0)
563 		error = kernel_vmount(0, "fstype", "devfs",
564 		    "fspath", "/dev", NULL);
565 	if (error != 0)
566 		init_does_devfs = 1;
567 
568 	/*
569 	 * Need just enough stack to hold the faked-up "execve()" arguments.
570 	 */
571 	addr = p->p_sysent->sv_usrstack - PAGE_SIZE;
572 	if (vm_map_find(&p->p_vmspace->vm_map, NULL, 0, &addr, PAGE_SIZE,
573 			FALSE, VM_PROT_ALL, VM_PROT_ALL, 0) != 0)
574 		panic("init: couldn't allocate argument space");
575 	p->p_vmspace->vm_maxsaddr = (caddr_t)addr;
576 	p->p_vmspace->vm_ssize = 1;
577 
578 	if ((var = getenv("init_path")) != NULL) {
579 		strlcpy(init_path, var, sizeof(init_path));
580 		freeenv(var);
581 	}
582 
583 	for (path = init_path; *path != '\0'; path = next) {
584 		while (*path == ':')
585 			path++;
586 		if (*path == '\0')
587 			break;
588 		for (next = path; *next != '\0' && *next != ':'; next++)
589 			/* nothing */ ;
590 		if (bootverbose)
591 			printf("start_init: trying %.*s\n", (int)(next - path),
592 			    path);
593 
594 		/*
595 		 * Move out the boot flag argument.
596 		 */
597 		options = 0;
598 		ucp = (char *)p->p_sysent->sv_usrstack;
599 		(void)subyte(--ucp, 0);		/* trailing zero */
600 		if (boothowto & RB_SINGLE) {
601 			(void)subyte(--ucp, 's');
602 			options = 1;
603 		}
604 #ifdef notyet
605                 if (boothowto & RB_FASTBOOT) {
606 			(void)subyte(--ucp, 'f');
607 			options = 1;
608 		}
609 #endif
610 
611 #ifdef BOOTCDROM
612 		(void)subyte(--ucp, 'C');
613 		options = 1;
614 #endif
615 		if (init_does_devfs) {
616 			(void)subyte(--ucp, 'd');
617 			options = 1;
618 		}
619 
620 		if (options == 0)
621 			(void)subyte(--ucp, '-');
622 		(void)subyte(--ucp, '-');		/* leading hyphen */
623 		arg1 = ucp;
624 
625 		/*
626 		 * Move out the file name (also arg 0).
627 		 */
628 		(void)subyte(--ucp, 0);
629 		for (s = next - 1; s >= path; s--)
630 			(void)subyte(--ucp, *s);
631 		arg0 = ucp;
632 
633 		/*
634 		 * Move out the arg pointers.
635 		 */
636 		uap = (char **)((intptr_t)ucp & ~(sizeof(intptr_t)-1));
637 		(void)suword((caddr_t)--uap, (long)0);	/* terminator */
638 		(void)suword((caddr_t)--uap, (long)(intptr_t)arg1);
639 		(void)suword((caddr_t)--uap, (long)(intptr_t)arg0);
640 
641 		/*
642 		 * Point at the arguments.
643 		 */
644 		args.fname = arg0;
645 		args.argv = uap;
646 		args.envv = NULL;
647 
648 		/*
649 		 * Now try to exec the program.  If can't for any reason
650 		 * other than it doesn't exist, complain.
651 		 *
652 		 * Otherwise, return via fork_trampoline() all the way
653 		 * to user mode as init!
654 		 */
655 		if ((error = execve(td, &args)) == 0) {
656 			mtx_unlock(&Giant);
657 			return;
658 		}
659 		if (error != ENOENT)
660 			printf("exec %.*s: error %d\n", (int)(next - path),
661 			    path, error);
662 	}
663 	printf("init: not found in path %s\n", init_path);
664 	panic("no init");
665 }
666 
667 /*
668  * Like kthread_create(), but runs in it's own address space.
669  * We do this early to reserve pid 1.
670  *
671  * Note special case - do not make it runnable yet.  Other work
672  * in progress will change this more.
673  */
674 static void
675 create_init(const void *udata __unused)
676 {
677 	struct ucred *newcred, *oldcred;
678 	int error;
679 
680 	error = fork1(&thread0, RFFDG | RFPROC | RFSTOPPED, 0, &initproc);
681 	if (error)
682 		panic("cannot fork init: %d\n", error);
683 	/* divorce init's credentials from the kernel's */
684 	newcred = crget();
685 	PROC_LOCK(initproc);
686 	initproc->p_flag |= P_SYSTEM;
687 	oldcred = initproc->p_ucred;
688 	crcopy(newcred, oldcred);
689 #ifdef MAC
690 	mac_create_proc1(newcred);
691 #endif
692 	initproc->p_ucred = newcred;
693 	PROC_UNLOCK(initproc);
694 	crfree(oldcred);
695 	cred_update_thread(FIRST_THREAD_IN_PROC(initproc));
696 	mtx_lock_spin(&sched_lock);
697 	initproc->p_sflag |= PS_INMEM;
698 	mtx_unlock_spin(&sched_lock);
699 	cpu_set_fork_handler(FIRST_THREAD_IN_PROC(initproc), start_init, NULL);
700 }
701 SYSINIT(init, SI_SUB_CREATE_INIT, SI_ORDER_FIRST, create_init, NULL)
702 
703 /*
704  * Make it runnable now.
705  */
706 static void
707 kick_init(const void *udata __unused)
708 {
709 	struct thread *td;
710 
711 	td = FIRST_THREAD_IN_PROC(initproc);
712 	mtx_lock_spin(&sched_lock);
713 	TD_SET_CAN_RUN(td);
714 	setrunqueue(td);	/* XXXKSE */
715 	mtx_unlock_spin(&sched_lock);
716 }
717 SYSINIT(kickinit, SI_SUB_KTHREAD_INIT, SI_ORDER_FIRST, kick_init, NULL)
718