xref: /freebsd/sys/kern/init_main.c (revision 7660b554bc59a07be0431c17e0e33815818baa69)
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 int cmask = CMASK;
101 
102 struct	vnode *rootvp;
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 (sysinit == NULL) {
178 		sysinit = SET_BEGIN(sysinit_set);
179 		sysinit_end = SET_LIMIT(sysinit_set);
180 	}
181 
182 restart:
183 	/*
184 	 * Perform a bubble sort of the system initialization objects by
185 	 * their subsystem (primary key) and order (secondary key).
186 	 */
187 	for (sipp = sysinit; sipp < sysinit_end; sipp++) {
188 		for (xipp = sipp + 1; xipp < sysinit_end; xipp++) {
189 			if ((*sipp)->subsystem < (*xipp)->subsystem ||
190 			     ((*sipp)->subsystem == (*xipp)->subsystem &&
191 			      (*sipp)->order <= (*xipp)->order))
192 				continue;	/* skip*/
193 			save = *sipp;
194 			*sipp = *xipp;
195 			*xipp = save;
196 		}
197 	}
198 
199 	/*
200 	 * Traverse the (now) ordered list of system initialization tasks.
201 	 * Perform each task, and continue on to the next task.
202 	 *
203 	 * The last item on the list is expected to be the scheduler,
204 	 * which will not return.
205 	 */
206 	for (sipp = sysinit; sipp < sysinit_end; sipp++) {
207 
208 		if ((*sipp)->subsystem == SI_SUB_DUMMY)
209 			continue;	/* skip dummy task(s)*/
210 
211 		if ((*sipp)->subsystem == SI_SUB_DONE)
212 			continue;
213 
214 		/* Call function */
215 		(*((*sipp)->func))((*sipp)->udata);
216 
217 		/* Check off the one we're just done */
218 		(*sipp)->subsystem = SI_SUB_DONE;
219 
220 		/* Check if we've installed more sysinit items via KLD */
221 		if (newsysinit != NULL) {
222 			if (sysinit != SET_BEGIN(sysinit_set))
223 				free(sysinit, M_TEMP);
224 			sysinit = newsysinit;
225 			sysinit_end = newsysinit_end;
226 			newsysinit = NULL;
227 			newsysinit_end = NULL;
228 			goto restart;
229 		}
230 	}
231 
232 	panic("Shouldn't get here!");
233 	/* NOTREACHED*/
234 }
235 
236 
237 /*
238  ***************************************************************************
239  ****
240  **** The following SYSINIT's belong elsewhere, but have not yet
241  **** been moved.
242  ****
243  ***************************************************************************
244  */
245 static void
246 print_caddr_t(void *data __unused)
247 {
248 	printf("%s", (char *)data);
249 }
250 SYSINIT(announce, SI_SUB_COPYRIGHT, SI_ORDER_FIRST, print_caddr_t, copyright)
251 SYSINIT(version, SI_SUB_COPYRIGHT, SI_ORDER_SECOND, print_caddr_t, version)
252 
253 static void
254 set_boot_verbose(void *data __unused)
255 {
256 
257 	if (boothowto & RB_VERBOSE)
258 		bootverbose++;
259 }
260 SYSINIT(boot_verbose, SI_SUB_TUNABLES, SI_ORDER_ANY, set_boot_verbose, NULL)
261 
262 struct sysentvec null_sysvec = {
263 	0,
264 	NULL,
265 	0,
266 	0,
267 	NULL,
268 	0,
269 	NULL,
270 	NULL,
271 	NULL,
272 	NULL,
273 	NULL,
274 	NULL,
275 	NULL,
276 	"null",
277 	NULL,
278 	NULL,
279 	0,
280 	PAGE_SIZE,
281 	VM_MIN_ADDRESS,
282 	VM_MAXUSER_ADDRESS,
283 	USRSTACK,
284 	PS_STRINGS,
285 	VM_PROT_ALL,
286 	NULL,
287 	NULL
288 };
289 
290 /*
291  ***************************************************************************
292  ****
293  **** The two following SYSINIT's are proc0 specific glue code.  I am not
294  **** convinced that they can not be safely combined, but their order of
295  **** operation has been maintained as the same as the original init_main.c
296  **** for right now.
297  ****
298  **** These probably belong in init_proc.c or kern_proc.c, since they
299  **** deal with proc0 (the fork template process).
300  ****
301  ***************************************************************************
302  */
303 /* ARGSUSED*/
304 static void
305 proc0_init(void *dummy __unused)
306 {
307 	register struct proc		*p;
308 	register struct filedesc0	*fdp;
309 	register unsigned i;
310 	struct thread *td;
311 	struct ksegrp *kg;
312 	struct kse *ke;
313 
314 	GIANT_REQUIRED;
315 	p = &proc0;
316 	td = &thread0;
317 	ke = &kse0;
318 	kg = &ksegrp0;
319 
320 	ke->ke_sched = kse0_sched;
321 	kg->kg_sched = ksegrp0_sched;
322 	p->p_sched = proc0_sched;
323 	td->td_sched = thread0_sched;
324 
325 	/*
326 	 * Initialize magic number.
327 	 */
328 	p->p_magic = P_MAGIC;
329 
330 	/*
331 	 * Initialize thread, process and pgrp structures.
332 	 */
333 	procinit();
334 	threadinit();
335 
336 	/*
337 	 * Initialize sleep queue hash table
338 	 */
339 	sleepinit();
340 
341 	/*
342 	 * additional VM structures
343 	 */
344 	vm_init2();
345 
346 	/*
347 	 * Create process 0 (the swapper).
348 	 */
349 	LIST_INSERT_HEAD(&allproc, p, p_list);
350 	LIST_INSERT_HEAD(PIDHASH(0), p, p_hash);
351 	mtx_init(&pgrp0.pg_mtx, "process group", NULL, MTX_DEF | MTX_DUPOK);
352 	p->p_pgrp = &pgrp0;
353 	LIST_INSERT_HEAD(PGRPHASH(0), &pgrp0, pg_hash);
354 	LIST_INIT(&pgrp0.pg_members);
355 	LIST_INSERT_HEAD(&pgrp0.pg_members, p, p_pglist);
356 
357 	pgrp0.pg_session = &session0;
358 	mtx_init(&session0.s_mtx, "session", NULL, MTX_DEF);
359 	session0.s_count = 1;
360 	session0.s_leader = p;
361 
362 	p->p_sysent = &null_sysvec;
363 
364 	/*
365 	 * proc_linkup was already done in init_i386() or alphainit() etc.
366 	 * because the earlier code needed to follow td->td_proc. Otherwise
367 	 * I would have done it here.. maybe this means this should be
368 	 * done earlier too.
369 	 */
370 	p->p_flag = P_SYSTEM;
371 	p->p_sflag = PS_INMEM;
372 	p->p_state = PRS_NORMAL;
373 	td->td_state = TDS_RUNNING;
374 	kg->kg_nice = NZERO;
375 	kg->kg_pri_class = PRI_TIMESHARE;
376 	kg->kg_user_pri = PUSER;
377 	td->td_priority = PVM;
378 	td->td_base_pri = PUSER;
379 	td->td_kse = ke; /* XXXKSE */
380 	td->td_oncpu = 0;
381 	ke->ke_state = KES_THREAD;
382 	ke->ke_thread = td;
383 	p->p_peers = 0;
384 	p->p_leader = p;
385 
386 
387 	bcopy("swapper", p->p_comm, sizeof ("swapper"));
388 
389 	callout_init(&p->p_itcallout, CALLOUT_MPSAFE);
390 	callout_init(&td->td_slpcallout, CALLOUT_MPSAFE);
391 
392 	/* Create credentials. */
393 	p->p_ucred = crget();
394 	p->p_ucred->cr_ngroups = 1;	/* group 0 */
395 	p->p_ucred->cr_uidinfo = uifind(0);
396 	p->p_ucred->cr_ruidinfo = uifind(0);
397 	p->p_ucred->cr_prison = NULL;	/* Don't jail it. */
398 #ifdef MAC
399 	mac_create_proc0(p->p_ucred);
400 #endif
401 	td->td_ucred = crhold(p->p_ucred);
402 
403 	/* Create sigacts. */
404 	p->p_sigacts = sigacts_alloc();
405 
406 	/* Initialize signal state for process 0. */
407 	siginit(&proc0);
408 
409 	/* Create the file descriptor table. */
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 
420 	/* Create the limits structures. */
421 	p->p_limit = &limit0;
422 	for (i = 0; i < sizeof(p->p_rlimit)/sizeof(p->p_rlimit[0]); i++)
423 		limit0.pl_rlimit[i].rlim_cur =
424 		    limit0.pl_rlimit[i].rlim_max = RLIM_INFINITY;
425 	limit0.pl_rlimit[RLIMIT_NOFILE].rlim_cur =
426 	    limit0.pl_rlimit[RLIMIT_NOFILE].rlim_max = maxfiles;
427 	limit0.pl_rlimit[RLIMIT_NPROC].rlim_cur =
428 	    limit0.pl_rlimit[RLIMIT_NPROC].rlim_max = maxproc;
429 	i = ptoa(cnt.v_free_count);
430 	limit0.pl_rlimit[RLIMIT_RSS].rlim_max = i;
431 	limit0.pl_rlimit[RLIMIT_MEMLOCK].rlim_max = i;
432 	limit0.pl_rlimit[RLIMIT_MEMLOCK].rlim_cur = i / 3;
433 	limit0.p_refcnt = 1;
434 	p->p_cpulimit = RLIM_INFINITY;
435 
436 	/* Allocate a prototype map so we have something to fork. */
437 	pmap_pinit0(vmspace_pmap(&vmspace0));
438 	p->p_vmspace = &vmspace0;
439 	vmspace0.vm_refcnt = 1;
440 	vm_map_init(&vmspace0.vm_map, p->p_sysent->sv_minuser,
441 	    p->p_sysent->sv_maxuser);
442 	vmspace0.vm_map.pmap = vmspace_pmap(&vmspace0);
443 
444 	/*
445 	 * We continue to place resource usage info
446 	 * in the user struct so that it's pageable.
447 	 */
448 	p->p_stats = &p->p_uarea->u_stats;
449 
450 	/*
451 	 * Charge root for one process.
452 	 */
453 	(void)chgproccnt(p->p_ucred->cr_ruidinfo, 1, 0);
454 }
455 SYSINIT(p0init, SI_SUB_INTRINSIC, SI_ORDER_FIRST, proc0_init, NULL)
456 
457 /* ARGSUSED*/
458 static void
459 proc0_post(void *dummy __unused)
460 {
461 	struct timespec ts;
462 	struct proc *p;
463 
464 	/*
465 	 * Now we can look at the time, having had a chance to verify the
466 	 * time from the filesystem.  Pretend that proc0 started now.
467 	 */
468 	sx_slock(&allproc_lock);
469 	LIST_FOREACH(p, &allproc, p_list) {
470 		microuptime(&p->p_stats->p_start);
471 		p->p_runtime.sec = 0;
472 		p->p_runtime.frac = 0;
473 	}
474 	sx_sunlock(&allproc_lock);
475 	binuptime(PCPU_PTR(switchtime));
476 	PCPU_SET(switchticks, ticks);
477 
478 	/*
479 	 * Give the ``random'' number generator a thump.
480 	 */
481 	nanotime(&ts);
482 	srandom(ts.tv_sec ^ ts.tv_nsec);
483 }
484 SYSINIT(p0post, SI_SUB_INTRINSIC_POST, SI_ORDER_FIRST, proc0_post, NULL)
485 
486 /*
487  ***************************************************************************
488  ****
489  **** The following SYSINIT's and glue code should be moved to the
490  **** respective files on a per subsystem basis.
491  ****
492  ***************************************************************************
493  */
494 
495 
496 /*
497  ***************************************************************************
498  ****
499  **** The following code probably belongs in another file, like
500  **** kern/init_init.c.
501  ****
502  ***************************************************************************
503  */
504 
505 /*
506  * List of paths to try when searching for "init".
507  */
508 static char init_path[MAXPATHLEN] =
509 #ifdef	INIT_PATH
510     __XSTRING(INIT_PATH);
511 #else
512     "/sbin/init:/sbin/oinit:/sbin/init.bak:/stand/sysinstall";
513 #endif
514 SYSCTL_STRING(_kern, OID_AUTO, init_path, CTLFLAG_RD, init_path, 0,
515 	"Path used to search the init process");
516 
517 /*
518  * Start the initial user process; try exec'ing each pathname in init_path.
519  * The program is invoked with one argument containing the boot flags.
520  */
521 static void
522 start_init(void *dummy)
523 {
524 	vm_offset_t addr;
525 	struct execve_args args;
526 	int options, error;
527 	char *var, *path, *next, *s;
528 	char *ucp, **uap, *arg0, *arg1;
529 	struct thread *td;
530 	struct proc *p;
531 	int init_does_devfs = 0;
532 
533 	mtx_lock(&Giant);
534 
535 	GIANT_REQUIRED;
536 
537 	td = curthread;
538 	p = td->td_proc;
539 
540 	vfs_mountroot();
541 
542 	/* Get the vnode for '/'.  Set p->p_fd->fd_cdir to reference it. */
543 	if (VFS_ROOT(TAILQ_FIRST(&mountlist), &rootvnode))
544 		panic("cannot find root vnode");
545 	FILEDESC_LOCK(p->p_fd);
546 	p->p_fd->fd_cdir = rootvnode;
547 	VREF(p->p_fd->fd_cdir);
548 	p->p_fd->fd_rdir = rootvnode;
549 	VREF(p->p_fd->fd_rdir);
550 	FILEDESC_UNLOCK(p->p_fd);
551 	VOP_UNLOCK(rootvnode, 0, td);
552 #ifdef MAC
553 	mac_create_root_mount(td->td_ucred, TAILQ_FIRST(&mountlist));
554 #endif
555 
556 	/*
557 	 * For disk based systems, we probably cannot do this yet
558 	 * since the fs will be read-only.  But a NFS root
559 	 * might be ok.  It is worth a shot.
560 	 */
561 	error = kern_mkdir(td, "/dev", UIO_SYSSPACE, 0700);
562 	if (error == EEXIST)
563 		error = 0;
564 	if (error == 0)
565 		error = kernel_vmount(0, "fstype", "devfs",
566 		    "fspath", "/dev", NULL);
567 	if (error != 0)
568 		init_does_devfs = 1;
569 
570 	/*
571 	 * Need just enough stack to hold the faked-up "execve()" arguments.
572 	 */
573 	addr = p->p_sysent->sv_usrstack - PAGE_SIZE;
574 	if (vm_map_find(&p->p_vmspace->vm_map, NULL, 0, &addr, PAGE_SIZE,
575 			FALSE, VM_PROT_ALL, VM_PROT_ALL, 0) != 0)
576 		panic("init: couldn't allocate argument space");
577 	p->p_vmspace->vm_maxsaddr = (caddr_t)addr;
578 	p->p_vmspace->vm_ssize = 1;
579 
580 	if ((var = getenv("init_path")) != NULL) {
581 		strlcpy(init_path, var, sizeof(init_path));
582 		freeenv(var);
583 	}
584 
585 	for (path = init_path; *path != '\0'; path = next) {
586 		while (*path == ':')
587 			path++;
588 		if (*path == '\0')
589 			break;
590 		for (next = path; *next != '\0' && *next != ':'; next++)
591 			/* nothing */ ;
592 		if (bootverbose)
593 			printf("start_init: trying %.*s\n", (int)(next - path),
594 			    path);
595 
596 		/*
597 		 * Move out the boot flag argument.
598 		 */
599 		options = 0;
600 		ucp = (char *)p->p_sysent->sv_usrstack;
601 		(void)subyte(--ucp, 0);		/* trailing zero */
602 		if (boothowto & RB_SINGLE) {
603 			(void)subyte(--ucp, 's');
604 			options = 1;
605 		}
606 #ifdef notyet
607                 if (boothowto & RB_FASTBOOT) {
608 			(void)subyte(--ucp, 'f');
609 			options = 1;
610 		}
611 #endif
612 
613 #ifdef BOOTCDROM
614 		(void)subyte(--ucp, 'C');
615 		options = 1;
616 #endif
617 		if (init_does_devfs) {
618 			(void)subyte(--ucp, 'd');
619 			options = 1;
620 		}
621 
622 		if (options == 0)
623 			(void)subyte(--ucp, '-');
624 		(void)subyte(--ucp, '-');		/* leading hyphen */
625 		arg1 = ucp;
626 
627 		/*
628 		 * Move out the file name (also arg 0).
629 		 */
630 		(void)subyte(--ucp, 0);
631 		for (s = next - 1; s >= path; s--)
632 			(void)subyte(--ucp, *s);
633 		arg0 = ucp;
634 
635 		/*
636 		 * Move out the arg pointers.
637 		 */
638 		uap = (char **)((intptr_t)ucp & ~(sizeof(intptr_t)-1));
639 		(void)suword((caddr_t)--uap, (long)0);	/* terminator */
640 		(void)suword((caddr_t)--uap, (long)(intptr_t)arg1);
641 		(void)suword((caddr_t)--uap, (long)(intptr_t)arg0);
642 
643 		/*
644 		 * Point at the arguments.
645 		 */
646 		args.fname = arg0;
647 		args.argv = uap;
648 		args.envv = NULL;
649 
650 		/*
651 		 * Now try to exec the program.  If can't for any reason
652 		 * other than it doesn't exist, complain.
653 		 *
654 		 * Otherwise, return via fork_trampoline() all the way
655 		 * to user mode as init!
656 		 */
657 		if ((error = execve(td, &args)) == 0) {
658 			mtx_unlock(&Giant);
659 			return;
660 		}
661 		if (error != ENOENT)
662 			printf("exec %.*s: error %d\n", (int)(next - path),
663 			    path, error);
664 	}
665 	printf("init: not found in path %s\n", init_path);
666 	panic("no init");
667 }
668 
669 /*
670  * Like kthread_create(), but runs in it's own address space.
671  * We do this early to reserve pid 1.
672  *
673  * Note special case - do not make it runnable yet.  Other work
674  * in progress will change this more.
675  */
676 static void
677 create_init(const void *udata __unused)
678 {
679 	struct ucred *newcred, *oldcred;
680 	int error;
681 
682 	error = fork1(&thread0, RFFDG | RFPROC | RFSTOPPED, 0, &initproc);
683 	if (error)
684 		panic("cannot fork init: %d\n", error);
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