xref: /freebsd/sys/kern/init_main.c (revision 0b87f79976047c8f4332bbf7dc03146f6b0de79f)
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 
47 #include <sys/param.h>
48 #include <sys/kernel.h>
49 #include <sys/file.h>
50 #include <sys/filedesc.h>
51 #include <sys/ktr.h>
52 #include <sys/lock.h>
53 #include <sys/mount.h>
54 #include <sys/mutex.h>
55 #include <sys/sysctl.h>
56 #include <sys/proc.h>
57 #include <sys/resourcevar.h>
58 #include <sys/systm.h>
59 #include <sys/signalvar.h>
60 #include <sys/vnode.h>
61 #include <sys/sysent.h>
62 #include <sys/reboot.h>
63 #include <sys/sx.h>
64 #include <sys/sysproto.h>
65 #include <sys/vmmeter.h>
66 #include <sys/unistd.h>
67 #include <sys/malloc.h>
68 #include <sys/conf.h>
69 
70 #include <machine/cpu.h>
71 
72 #include <vm/vm.h>
73 #include <vm/vm_param.h>
74 #include <vm/pmap.h>
75 #include <vm/vm_map.h>
76 #include <sys/user.h>
77 #include <sys/copyright.h>
78 
79 void mi_startup(void);				/* Should be elsewhere */
80 
81 /* Components of the first process -- never freed. */
82 static struct session session0;
83 static struct pgrp pgrp0;
84 struct	proc proc0;
85 struct	thread thread0;
86 static struct procsig procsig0;
87 static struct filedesc0 filedesc0;
88 static struct plimit limit0;
89 static struct vmspace vmspace0;
90 struct	proc *initproc;
91 
92 int cmask = CMASK;
93 extern int fallback_elf_brand;
94 
95 struct	vnode *rootvp;
96 int	boothowto = 0;		/* initialized so that it can be patched */
97 SYSCTL_INT(_debug, OID_AUTO, boothowto, CTLFLAG_RD, &boothowto, 0, "");
98 int	bootverbose;
99 SYSCTL_INT(_debug, OID_AUTO, bootverbose, CTLFLAG_RW, &bootverbose, 0, "");
100 
101 /*
102  * This ensures that there is at least one entry so that the sysinit_set
103  * symbol is not undefined.  A sybsystem ID of SI_SUB_DUMMY is never
104  * executed.
105  */
106 SYSINIT(placeholder, SI_SUB_DUMMY, SI_ORDER_ANY, NULL, NULL)
107 
108 /*
109  * The sysinit table itself.  Items are checked off as the are run.
110  * If we want to register new sysinit types, add them to newsysinit.
111  */
112 SET_DECLARE(sysinit_set, struct sysinit);
113 struct sysinit **sysinit, **sysinit_end;
114 struct sysinit **newsysinit, **newsysinit_end;
115 
116 /*
117  * Merge a new sysinit set into the current set, reallocating it if
118  * necessary.  This can only be called after malloc is running.
119  */
120 void
121 sysinit_add(struct sysinit **set, struct sysinit **set_end)
122 {
123 	struct sysinit **newset;
124 	struct sysinit **sipp;
125 	struct sysinit **xipp;
126 	int count;
127 
128 	count = set_end - set;
129 	if (newsysinit)
130 		count += newsysinit_end - newsysinit;
131 	else
132 		count += sysinit_end - sysinit;
133 	newset = malloc(count * sizeof(*sipp), M_TEMP, M_NOWAIT);
134 	if (newset == NULL)
135 		panic("cannot malloc for sysinit");
136 	xipp = newset;
137 	if (newsysinit)
138 		for (sipp = newsysinit; sipp < newsysinit_end; sipp++)
139 			*xipp++ = *sipp;
140 	else
141 		for (sipp = sysinit; sipp < sysinit_end; sipp++)
142 			*xipp++ = *sipp;
143 	for (sipp = set; sipp < set_end; sipp++)
144 		*xipp++ = *sipp;
145 	if (newsysinit)
146 		free(newsysinit, M_TEMP);
147 	newsysinit = newset;
148 	newsysinit_end = newset + count;
149 }
150 
151 /*
152  * System startup; initialize the world, create process 0, mount root
153  * filesystem, and fork to create init and pagedaemon.  Most of the
154  * hard work is done in the lower-level initialization routines including
155  * startup(), which does memory initialization and autoconfiguration.
156  *
157  * This allows simple addition of new kernel subsystems that require
158  * boot time initialization.  It also allows substitution of subsystem
159  * (for instance, a scheduler, kernel profiler, or VM system) by object
160  * module.  Finally, it allows for optional "kernel threads".
161  */
162 void
163 mi_startup(void)
164 {
165 
166 	register struct sysinit **sipp;		/* system initialization*/
167 	register struct sysinit **xipp;		/* interior loop of sort*/
168 	register struct sysinit *save;		/* bubble*/
169 
170 	if (sysinit == NULL) {
171 		sysinit = SET_BEGIN(sysinit_set);
172 		sysinit_end = SET_LIMIT(sysinit_set);
173 	}
174 
175 restart:
176 	/*
177 	 * Perform a bubble sort of the system initialization objects by
178 	 * their subsystem (primary key) and order (secondary key).
179 	 */
180 	for (sipp = sysinit; sipp < sysinit_end; sipp++) {
181 		for (xipp = sipp + 1; xipp < sysinit_end; xipp++) {
182 			if ((*sipp)->subsystem < (*xipp)->subsystem ||
183 			     ((*sipp)->subsystem == (*xipp)->subsystem &&
184 			      (*sipp)->order <= (*xipp)->order))
185 				continue;	/* skip*/
186 			save = *sipp;
187 			*sipp = *xipp;
188 			*xipp = save;
189 		}
190 	}
191 
192 	/*
193 	 * Traverse the (now) ordered list of system initialization tasks.
194 	 * Perform each task, and continue on to the next task.
195 	 *
196 	 * The last item on the list is expected to be the scheduler,
197 	 * which will not return.
198 	 */
199 	for (sipp = sysinit; sipp < sysinit_end; sipp++) {
200 
201 		if ((*sipp)->subsystem == SI_SUB_DUMMY)
202 			continue;	/* skip dummy task(s)*/
203 
204 		if ((*sipp)->subsystem == SI_SUB_DONE)
205 			continue;
206 
207 		/* Call function */
208 		(*((*sipp)->func))((*sipp)->udata);
209 
210 		/* Check off the one we're just done */
211 		(*sipp)->subsystem = SI_SUB_DONE;
212 
213 		/* Check if we've installed more sysinit items via KLD */
214 		if (newsysinit != NULL) {
215 			if (sysinit != SET_BEGIN(sysinit_set))
216 				free(sysinit, M_TEMP);
217 			sysinit = newsysinit;
218 			sysinit_end = newsysinit_end;
219 			newsysinit = NULL;
220 			newsysinit_end = NULL;
221 			goto restart;
222 		}
223 	}
224 
225 	panic("Shouldn't get here!");
226 	/* NOTREACHED*/
227 }
228 
229 
230 /*
231  ***************************************************************************
232  ****
233  **** The following SYSINIT's belong elsewhere, but have not yet
234  **** been moved.
235  ****
236  ***************************************************************************
237  */
238 static void
239 print_caddr_t(void *data __unused)
240 {
241 	printf("%s", (char *)data);
242 }
243 SYSINIT(announce, SI_SUB_COPYRIGHT, SI_ORDER_FIRST, print_caddr_t, copyright)
244 SYSINIT(version, SI_SUB_COPYRIGHT, SI_ORDER_SECOND, print_caddr_t, version)
245 
246 static void
247 set_boot_verbose(void *data __unused)
248 {
249 
250 	if (boothowto & RB_VERBOSE)
251 		bootverbose++;
252 }
253 SYSINIT(boot_verbose, SI_SUB_TUNABLES, SI_ORDER_ANY, set_boot_verbose, NULL)
254 
255 /*
256  ***************************************************************************
257  ****
258  **** The two following SYSINT's are proc0 specific glue code.  I am not
259  **** convinced that they can not be safely combined, but their order of
260  **** operation has been maintained as the same as the original init_main.c
261  **** for right now.
262  ****
263  **** These probably belong in init_proc.c or kern_proc.c, since they
264  **** deal with proc0 (the fork template process).
265  ****
266  ***************************************************************************
267  */
268 /* ARGSUSED*/
269 static void
270 proc0_init(void *dummy __unused)
271 {
272 	register struct proc		*p;
273 	register struct filedesc0	*fdp;
274 	register unsigned i;
275 	struct thread *td;
276 	struct ksegrp *kg;
277 	struct kse *ke;
278 
279 	GIANT_REQUIRED;
280 	p = &proc0;
281 	td = &thread0;
282 
283 	/*
284 	 * Initialize magic number.
285 	 */
286 	p->p_magic = P_MAGIC;
287 
288 	/*
289 	 * Initialize thread, process and pgrp structures.
290 	 */
291 	procinit();
292 	threadinit();
293 
294 	/*
295 	 * Initialize sleep queue hash table
296 	 */
297 	sleepinit();
298 
299 	/*
300 	 * additional VM structures
301 	 */
302 	vm_init2();
303 
304 	/*
305 	 * Create process 0 (the swapper).
306 	 */
307 	LIST_INSERT_HEAD(&allproc, p, p_list);
308 	LIST_INSERT_HEAD(PIDHASH(0), p, p_hash);
309 	mtx_init(&pgrp0.pg_mtx, "process group", NULL, MTX_DEF | MTX_DUPOK);
310 	p->p_pgrp = &pgrp0;
311 	LIST_INSERT_HEAD(PGRPHASH(0), &pgrp0, pg_hash);
312 	LIST_INIT(&pgrp0.pg_members);
313 	LIST_INSERT_HEAD(&pgrp0.pg_members, p, p_pglist);
314 
315 	pgrp0.pg_session = &session0;
316 	mtx_init(&session0.s_mtx, "session", NULL, MTX_DEF);
317 	session0.s_count = 1;
318 	session0.s_leader = p;
319 
320 #ifdef __ELF__
321 	p->p_sysent = &elf_freebsd_sysvec;
322 #else
323 	p->p_sysent = &aout_sysvec;
324 #endif
325 
326 	/*
327 	 * proc_linkup was already done in init_i386() or alphainit() etc.
328 	 * because the earlier code needed to follow td->td_proc. Otherwise
329 	 * I would have done it here.. maybe this means this should be
330 	 * done earlier too.
331 	 */
332 	ke = &proc0.p_kse;	/* XXXKSE */
333 	kg = &proc0.p_ksegrp;	/* XXXKSE */
334 	p->p_flag = P_SYSTEM;
335 	p->p_sflag = PS_INMEM;
336 	p->p_state = PRS_NORMAL;
337 	td->td_state = TDS_RUNNING;
338 	kg->kg_nice = NZERO;
339 	kg->kg_pri_class = PRI_TIMESHARE;
340 	kg->kg_user_pri = PUSER;
341 	td->td_priority = PVM;
342 	td->td_base_pri = PUSER;
343 	td->td_kse = ke; /* XXXKSE */
344 	ke->ke_oncpu = 0;
345 	ke->ke_state = KES_THREAD;
346 	ke->ke_thread = td;
347 	/* proc_linkup puts it in the idle queue, that's not what we want. */
348 	TAILQ_REMOVE(&kg->kg_iq, ke, ke_kgrlist);
349 	kg->kg_idle_kses--;
350 	p->p_peers = 0;
351 	p->p_leader = p;
352 KASSERT((ke->ke_kgrlist.tqe_next != ke), ("linked to self!"));
353 
354 
355 	bcopy("swapper", p->p_comm, sizeof ("swapper"));
356 
357 	callout_init(&p->p_itcallout, 0);
358 	callout_init(&td->td_slpcallout, 1);
359 
360 	/* Create credentials. */
361 	p->p_ucred = crget();
362 	p->p_ucred->cr_ngroups = 1;	/* group 0 */
363 	p->p_ucred->cr_uidinfo = uifind(0);
364 	p->p_ucred->cr_ruidinfo = uifind(0);
365 	p->p_ucred->cr_prison = NULL;	/* Don't jail it. */
366 	td->td_ucred = crhold(p->p_ucred);
367 
368 	/* Create procsig. */
369 	p->p_procsig = &procsig0;
370 	p->p_procsig->ps_refcnt = 1;
371 
372 	/* Initialize signal state for process 0. */
373 	siginit(&proc0);
374 
375 	/* Create the file descriptor table. */
376 	fdp = &filedesc0;
377 	p->p_fd = &fdp->fd_fd;
378 	mtx_init(&fdp->fd_fd.fd_mtx, FILEDESC_LOCK_DESC, NULL, MTX_DEF);
379 	fdp->fd_fd.fd_refcnt = 1;
380 	fdp->fd_fd.fd_cmask = cmask;
381 	fdp->fd_fd.fd_ofiles = fdp->fd_dfiles;
382 	fdp->fd_fd.fd_ofileflags = fdp->fd_dfileflags;
383 	fdp->fd_fd.fd_nfiles = NDFILE;
384 
385 	/* Create the limits structures. */
386 	p->p_limit = &limit0;
387 	for (i = 0; i < sizeof(p->p_rlimit)/sizeof(p->p_rlimit[0]); i++)
388 		limit0.pl_rlimit[i].rlim_cur =
389 		    limit0.pl_rlimit[i].rlim_max = RLIM_INFINITY;
390 	limit0.pl_rlimit[RLIMIT_NOFILE].rlim_cur =
391 	    limit0.pl_rlimit[RLIMIT_NOFILE].rlim_max = maxfiles;
392 	limit0.pl_rlimit[RLIMIT_NPROC].rlim_cur =
393 	    limit0.pl_rlimit[RLIMIT_NPROC].rlim_max = maxproc;
394 	i = ptoa(cnt.v_free_count);
395 	limit0.pl_rlimit[RLIMIT_RSS].rlim_max = i;
396 	limit0.pl_rlimit[RLIMIT_MEMLOCK].rlim_max = i;
397 	limit0.pl_rlimit[RLIMIT_MEMLOCK].rlim_cur = i / 3;
398 	limit0.p_cpulimit = RLIM_INFINITY;
399 	limit0.p_refcnt = 1;
400 
401 	/* Allocate a prototype map so we have something to fork. */
402 	pmap_pinit0(vmspace_pmap(&vmspace0));
403 	p->p_vmspace = &vmspace0;
404 	vmspace0.vm_refcnt = 1;
405 	vm_map_init(&vmspace0.vm_map, round_page(VM_MIN_ADDRESS),
406 	    trunc_page(VM_MAXUSER_ADDRESS));
407 	vmspace0.vm_map.pmap = vmspace_pmap(&vmspace0);
408 
409 	/*
410 	 * We continue to place resource usage info and signal
411 	 * actions in the user struct so they're pageable.
412 	 */
413 	p->p_stats = &p->p_uarea->u_stats;
414 	p->p_sigacts = &p->p_uarea->u_sigacts;
415 
416 	/*
417 	 * Charge root for one process.
418 	 */
419 	(void)chgproccnt(p->p_ucred->cr_ruidinfo, 1, 0);
420 }
421 SYSINIT(p0init, SI_SUB_INTRINSIC, SI_ORDER_FIRST, proc0_init, NULL)
422 
423 /* ARGSUSED*/
424 static void
425 proc0_post(void *dummy __unused)
426 {
427 	struct timespec ts;
428 	struct proc *p;
429 
430 	/*
431 	 * Now we can look at the time, having had a chance to verify the
432 	 * time from the filesystem.  Pretend that proc0 started now.
433 	 */
434 	sx_slock(&allproc_lock);
435 	LIST_FOREACH(p, &allproc, p_list) {
436 		microtime(&p->p_stats->p_start);
437 		p->p_runtime.sec = 0;
438 		p->p_runtime.frac = 0;
439 	}
440 	sx_sunlock(&allproc_lock);
441 	binuptime(PCPU_PTR(switchtime));
442 	PCPU_SET(switchticks, ticks);
443 
444 	/*
445 	 * Give the ``random'' number generator a thump.
446 	 */
447 	nanotime(&ts);
448 	srandom(ts.tv_sec ^ ts.tv_nsec);
449 }
450 SYSINIT(p0post, SI_SUB_INTRINSIC_POST, SI_ORDER_FIRST, proc0_post, NULL)
451 
452 /*
453  ***************************************************************************
454  ****
455  **** The following SYSINIT's and glue code should be moved to the
456  **** respective files on a per subsystem basis.
457  ****
458  ***************************************************************************
459  */
460 
461 
462 /*
463  ***************************************************************************
464  ****
465  **** The following code probably belongs in another file, like
466  **** kern/init_init.c.
467  ****
468  ***************************************************************************
469  */
470 
471 /*
472  * List of paths to try when searching for "init".
473  */
474 static char init_path[MAXPATHLEN] =
475 #ifdef	INIT_PATH
476     __XSTRING(INIT_PATH);
477 #else
478     "/sbin/init:/sbin/oinit:/sbin/init.bak:/stand/sysinstall";
479 #endif
480 SYSCTL_STRING(_kern, OID_AUTO, init_path, CTLFLAG_RD, init_path, 0,
481 	"Path used to search the init process");
482 
483 /*
484  * Start the initial user process; try exec'ing each pathname in init_path.
485  * The program is invoked with one argument containing the boot flags.
486  */
487 static void
488 start_init(void *dummy)
489 {
490 	vm_offset_t addr;
491 	struct execve_args args;
492 	int options, error;
493 	char *var, *path, *next, *s;
494 	char *ucp, **uap, *arg0, *arg1;
495 	struct thread *td;
496 	struct proc *p;
497 	int init_does_devfs = 0;
498 
499 	mtx_lock(&Giant);
500 
501 	GIANT_REQUIRED;
502 
503 	td = curthread;
504 	p = td->td_proc;
505 
506 	vfs_mountroot();
507 
508 	/* Get the vnode for '/'.  Set p->p_fd->fd_cdir to reference it. */
509 	if (VFS_ROOT(TAILQ_FIRST(&mountlist), &rootvnode))
510 		panic("cannot find root vnode");
511 	FILEDESC_LOCK(p->p_fd);
512 	p->p_fd->fd_cdir = rootvnode;
513 	VREF(p->p_fd->fd_cdir);
514 	p->p_fd->fd_rdir = rootvnode;
515 	VREF(p->p_fd->fd_rdir);
516 	FILEDESC_UNLOCK(p->p_fd);
517 	VOP_UNLOCK(rootvnode, 0, td);
518 
519 	if (devfs_present) {
520 		/*
521 		 * For disk based systems, we probably cannot do this yet
522 		 * since the fs will be read-only.  But a NFS root
523 		 * might be ok.  It is worth a shot.
524 		 */
525 		error = vn_mkdir("/dev", 0700, UIO_SYSSPACE, td);
526 		if (error == EEXIST)
527 			error = 0;
528 		if (error == 0)
529 			error = kernel_vmount(0, "fstype", "devfs",
530 			    "fspath", "/dev", NULL);
531 		if (error != 0)
532 			init_does_devfs = 1;
533 	}
534 
535 	/*
536 	 * Need just enough stack to hold the faked-up "execve()" arguments.
537 	 */
538 	addr = trunc_page(USRSTACK - PAGE_SIZE);
539 	if (vm_map_find(&p->p_vmspace->vm_map, NULL, 0, &addr, PAGE_SIZE,
540 			FALSE, VM_PROT_ALL, VM_PROT_ALL, 0) != 0)
541 		panic("init: couldn't allocate argument space");
542 	p->p_vmspace->vm_maxsaddr = (caddr_t)addr;
543 	p->p_vmspace->vm_ssize = 1;
544 
545 	if ((var = getenv("init_path")) != NULL) {
546 		strncpy(init_path, var, sizeof init_path);
547 		init_path[sizeof init_path - 1] = 0;
548 		freeenv(var);
549 	}
550 	if ((var = getenv("kern.fallback_elf_brand")) != NULL) {
551 		fallback_elf_brand = strtol(var, NULL, 0);
552 		freeenv(var);
553 	}
554 
555 	for (path = init_path; *path != '\0'; path = next) {
556 		while (*path == ':')
557 			path++;
558 		if (*path == '\0')
559 			break;
560 		for (next = path; *next != '\0' && *next != ':'; next++)
561 			/* nothing */ ;
562 		if (bootverbose)
563 			printf("start_init: trying %.*s\n", (int)(next - path),
564 			    path);
565 
566 		/*
567 		 * Move out the boot flag argument.
568 		 */
569 		options = 0;
570 		ucp = (char *)USRSTACK;
571 		(void)subyte(--ucp, 0);		/* trailing zero */
572 		if (boothowto & RB_SINGLE) {
573 			(void)subyte(--ucp, 's');
574 			options = 1;
575 		}
576 #ifdef notyet
577                 if (boothowto & RB_FASTBOOT) {
578 			(void)subyte(--ucp, 'f');
579 			options = 1;
580 		}
581 #endif
582 
583 #ifdef BOOTCDROM
584 		(void)subyte(--ucp, 'C');
585 		options = 1;
586 #endif
587 		if (init_does_devfs) {
588 			(void)subyte(--ucp, 'd');
589 			options = 1;
590 		}
591 
592 		if (options == 0)
593 			(void)subyte(--ucp, '-');
594 		(void)subyte(--ucp, '-');		/* leading hyphen */
595 		arg1 = ucp;
596 
597 		/*
598 		 * Move out the file name (also arg 0).
599 		 */
600 		(void)subyte(--ucp, 0);
601 		for (s = next - 1; s >= path; s--)
602 			(void)subyte(--ucp, *s);
603 		arg0 = ucp;
604 
605 		/*
606 		 * Move out the arg pointers.
607 		 */
608 		uap = (char **)((intptr_t)ucp & ~(sizeof(intptr_t)-1));
609 		(void)suword((caddr_t)--uap, (long)0);	/* terminator */
610 		(void)suword((caddr_t)--uap, (long)(intptr_t)arg1);
611 		(void)suword((caddr_t)--uap, (long)(intptr_t)arg0);
612 
613 		/*
614 		 * Point at the arguments.
615 		 */
616 		args.fname = arg0;
617 		args.argv = uap;
618 		args.envv = NULL;
619 
620 		/*
621 		 * Now try to exec the program.  If can't for any reason
622 		 * other than it doesn't exist, complain.
623 		 *
624 		 * Otherwise, return via fork_trampoline() all the way
625 		 * to user mode as init!
626 		 */
627 		if ((error = execve(td, &args)) == 0) {
628 			mtx_unlock(&Giant);
629 			return;
630 		}
631 		if (error != ENOENT)
632 			printf("exec %.*s: error %d\n", (int)(next - path),
633 			    path, error);
634 	}
635 	printf("init: not found in path %s\n", init_path);
636 	panic("no init");
637 }
638 
639 /*
640  * Like kthread_create(), but runs in it's own address space.
641  * We do this early to reserve pid 1.
642  *
643  * Note special case - do not make it runnable yet.  Other work
644  * in progress will change this more.
645  */
646 static void
647 create_init(const void *udata __unused)
648 {
649 	struct ucred *newcred, *oldcred;
650 	int error;
651 
652 	error = fork1(&thread0, RFFDG | RFPROC | RFSTOPPED, &initproc);
653 	if (error)
654 		panic("cannot fork init: %d\n", error);
655 	/* divorce init's credentials from the kernel's */
656 	newcred = crget();
657 	PROC_LOCK(initproc);
658 	initproc->p_flag |= P_SYSTEM;
659 	oldcred = initproc->p_ucred;
660 	crcopy(newcred, oldcred);
661 	initproc->p_ucred = newcred;
662 	PROC_UNLOCK(initproc);
663 	crfree(oldcred);
664 	mtx_lock_spin(&sched_lock);
665 	initproc->p_sflag |= PS_INMEM;
666 	mtx_unlock_spin(&sched_lock);
667 	cpu_set_fork_handler(FIRST_THREAD_IN_PROC(initproc), start_init, NULL);
668 }
669 SYSINIT(init, SI_SUB_CREATE_INIT, SI_ORDER_FIRST, create_init, NULL)
670 
671 /*
672  * Make it runnable now.
673  */
674 static void
675 kick_init(const void *udata __unused)
676 {
677 	struct thread *td;
678 
679 	td = FIRST_THREAD_IN_PROC(initproc);
680 	mtx_lock_spin(&sched_lock);
681 	setrunqueue(td);	/* XXXKSE */
682 	mtx_unlock_spin(&sched_lock);
683 }
684 SYSINIT(kickinit, SI_SUB_KTHREAD_INIT, SI_ORDER_FIRST, kick_init, NULL)
685