xref: /freebsd/sys/kern/init_main.c (revision 626ff2081fd054b971c71aedddcd45f42efb406a)
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	ksegrp ksegrp0;
94 struct	vmspace vmspace0;
95 struct	proc *initproc;
96 
97 int	boothowto = 0;		/* initialized so that it can be patched */
98 SYSCTL_INT(_debug, OID_AUTO, boothowto, CTLFLAG_RD, &boothowto, 0, "");
99 int	bootverbose;
100 SYSCTL_INT(_debug, OID_AUTO, bootverbose, CTLFLAG_RW, &bootverbose, 0, "");
101 
102 /*
103  * This ensures that there is at least one entry so that the sysinit_set
104  * symbol is not undefined.  A sybsystem ID of SI_SUB_DUMMY is never
105  * executed.
106  */
107 SYSINIT(placeholder, SI_SUB_DUMMY, SI_ORDER_ANY, NULL, NULL)
108 
109 /*
110  * The sysinit table itself.  Items are checked off as the are run.
111  * If we want to register new sysinit types, add them to newsysinit.
112  */
113 SET_DECLARE(sysinit_set, struct sysinit);
114 struct sysinit **sysinit, **sysinit_end;
115 struct sysinit **newsysinit, **newsysinit_end;
116 
117 /*
118  * Merge a new sysinit set into the current set, reallocating it if
119  * necessary.  This can only be called after malloc is running.
120  */
121 void
122 sysinit_add(struct sysinit **set, struct sysinit **set_end)
123 {
124 	struct sysinit **newset;
125 	struct sysinit **sipp;
126 	struct sysinit **xipp;
127 	int count;
128 
129 	count = set_end - set;
130 	if (newsysinit)
131 		count += newsysinit_end - newsysinit;
132 	else
133 		count += sysinit_end - sysinit;
134 	newset = malloc(count * sizeof(*sipp), M_TEMP, M_NOWAIT);
135 	if (newset == NULL)
136 		panic("cannot malloc for sysinit");
137 	xipp = newset;
138 	if (newsysinit)
139 		for (sipp = newsysinit; sipp < newsysinit_end; sipp++)
140 			*xipp++ = *sipp;
141 	else
142 		for (sipp = sysinit; sipp < sysinit_end; sipp++)
143 			*xipp++ = *sipp;
144 	for (sipp = set; sipp < set_end; sipp++)
145 		*xipp++ = *sipp;
146 	if (newsysinit)
147 		free(newsysinit, M_TEMP);
148 	newsysinit = newset;
149 	newsysinit_end = newset + count;
150 }
151 
152 /*
153  * System startup; initialize the world, create process 0, mount root
154  * filesystem, and fork to create init and pagedaemon.  Most of the
155  * hard work is done in the lower-level initialization routines including
156  * startup(), which does memory initialization and autoconfiguration.
157  *
158  * This allows simple addition of new kernel subsystems that require
159  * boot time initialization.  It also allows substitution of subsystem
160  * (for instance, a scheduler, kernel profiler, or VM system) by object
161  * module.  Finally, it allows for optional "kernel threads".
162  */
163 void
164 mi_startup(void)
165 {
166 
167 	register struct sysinit **sipp;		/* system initialization*/
168 	register struct sysinit **xipp;		/* interior loop of sort*/
169 	register struct sysinit *save;		/* bubble*/
170 
171 	if (sysinit == NULL) {
172 		sysinit = SET_BEGIN(sysinit_set);
173 		sysinit_end = SET_LIMIT(sysinit_set);
174 	}
175 
176 restart:
177 	/*
178 	 * Perform a bubble sort of the system initialization objects by
179 	 * their subsystem (primary key) and order (secondary key).
180 	 */
181 	for (sipp = sysinit; sipp < sysinit_end; sipp++) {
182 		for (xipp = sipp + 1; xipp < sysinit_end; xipp++) {
183 			if ((*sipp)->subsystem < (*xipp)->subsystem ||
184 			     ((*sipp)->subsystem == (*xipp)->subsystem &&
185 			      (*sipp)->order <= (*xipp)->order))
186 				continue;	/* skip*/
187 			save = *sipp;
188 			*sipp = *xipp;
189 			*xipp = save;
190 		}
191 	}
192 
193 	/*
194 	 * Traverse the (now) ordered list of system initialization tasks.
195 	 * Perform each task, and continue on to the next task.
196 	 *
197 	 * The last item on the list is expected to be the scheduler,
198 	 * which will not return.
199 	 */
200 	for (sipp = sysinit; sipp < sysinit_end; sipp++) {
201 
202 		if ((*sipp)->subsystem == SI_SUB_DUMMY)
203 			continue;	/* skip dummy task(s)*/
204 
205 		if ((*sipp)->subsystem == SI_SUB_DONE)
206 			continue;
207 
208 		/* Call function */
209 		(*((*sipp)->func))((*sipp)->udata);
210 
211 		/* Check off the one we're just done */
212 		(*sipp)->subsystem = SI_SUB_DONE;
213 
214 		/* Check if we've installed more sysinit items via KLD */
215 		if (newsysinit != NULL) {
216 			if (sysinit != SET_BEGIN(sysinit_set))
217 				free(sysinit, M_TEMP);
218 			sysinit = newsysinit;
219 			sysinit_end = newsysinit_end;
220 			newsysinit = NULL;
221 			newsysinit_end = NULL;
222 			goto restart;
223 		}
224 	}
225 
226 	panic("Shouldn't get here!");
227 	/* NOTREACHED*/
228 }
229 
230 
231 /*
232  ***************************************************************************
233  ****
234  **** The following SYSINIT's belong elsewhere, but have not yet
235  **** been moved.
236  ****
237  ***************************************************************************
238  */
239 static void
240 print_caddr_t(void *data __unused)
241 {
242 	printf("%s", (char *)data);
243 }
244 SYSINIT(announce, SI_SUB_COPYRIGHT, SI_ORDER_FIRST, print_caddr_t, copyright)
245 SYSINIT(version, SI_SUB_COPYRIGHT, SI_ORDER_SECOND, print_caddr_t, version)
246 
247 #ifdef WITNESS
248 static char wit_warn[] =
249      "WARNING: WITNESS option enabled, expect reduced performance.\n";
250 SYSINIT(witwarn, SI_SUB_COPYRIGHT, SI_ORDER_SECOND + 1,
251    print_caddr_t, wit_warn)
252 #endif
253 
254 #ifdef DIAGNOSTIC
255 static char diag_warn[] =
256      "WARNING: DIAGNOSTIC option enabled, expect reduced performance.\n";
257 SYSINIT(diagwarn, SI_SUB_COPYRIGHT, SI_ORDER_SECOND + 2,
258     print_caddr_t, diag_warn)
259 #endif
260 
261 static void
262 set_boot_verbose(void *data __unused)
263 {
264 
265 	if (boothowto & RB_VERBOSE)
266 		bootverbose++;
267 }
268 SYSINIT(boot_verbose, SI_SUB_TUNABLES, SI_ORDER_ANY, set_boot_verbose, NULL)
269 
270 struct sysentvec null_sysvec = {
271 	0,
272 	NULL,
273 	0,
274 	0,
275 	NULL,
276 	0,
277 	NULL,
278 	NULL,
279 	NULL,
280 	NULL,
281 	NULL,
282 	NULL,
283 	NULL,
284 	"null",
285 	NULL,
286 	NULL,
287 	0,
288 	PAGE_SIZE,
289 	VM_MIN_ADDRESS,
290 	VM_MAXUSER_ADDRESS,
291 	USRSTACK,
292 	PS_STRINGS,
293 	VM_PROT_ALL,
294 	NULL,
295 	NULL,
296 	NULL
297 };
298 
299 /*
300  ***************************************************************************
301  ****
302  **** The two following SYSINIT's are proc0 specific glue code.  I am not
303  **** convinced that they can not be safely combined, but their order of
304  **** operation has been maintained as the same as the original init_main.c
305  **** for right now.
306  ****
307  **** These probably belong in init_proc.c or kern_proc.c, since they
308  **** deal with proc0 (the fork template process).
309  ****
310  ***************************************************************************
311  */
312 /* ARGSUSED*/
313 static void
314 proc0_init(void *dummy __unused)
315 {
316 	struct proc *p;
317 	unsigned i;
318 	struct thread *td;
319 	struct ksegrp *kg;
320 
321 	GIANT_REQUIRED;
322 	p = &proc0;
323 	td = &thread0;
324 	kg = &ksegrp0;
325 
326 	/*
327 	 * Initialize magic number.
328 	 */
329 	p->p_magic = P_MAGIC;
330 
331 	/*
332 	 * Initialize thread, process and ksegrp structures.
333 	 */
334 	procinit();	/* set up proc zone */
335 	threadinit();	/* set up thead, upcall and KSEGRP zones */
336 
337 	/*
338 	 * Initialise scheduler resources.
339 	 * Add scheduler specific parts to proc, ksegrp, thread as needed.
340 	 */
341 	schedinit();	/* scheduler gets its house in order */
342 	/*
343 	 * Initialize sleep queue hash table
344 	 */
345 	sleepinit();
346 
347 	/*
348 	 * additional VM structures
349 	 */
350 	vm_init2();
351 
352 	/*
353 	 * Create process 0 (the swapper).
354 	 */
355 	LIST_INSERT_HEAD(&allproc, p, p_list);
356 	LIST_INSERT_HEAD(PIDHASH(0), p, p_hash);
357 	mtx_init(&pgrp0.pg_mtx, "process group", NULL, MTX_DEF | MTX_DUPOK);
358 	p->p_pgrp = &pgrp0;
359 	LIST_INSERT_HEAD(PGRPHASH(0), &pgrp0, pg_hash);
360 	LIST_INIT(&pgrp0.pg_members);
361 	LIST_INSERT_HEAD(&pgrp0.pg_members, p, p_pglist);
362 
363 	pgrp0.pg_session = &session0;
364 	mtx_init(&session0.s_mtx, "session", NULL, MTX_DEF);
365 	session0.s_count = 1;
366 	session0.s_leader = p;
367 
368 	p->p_sysent = &null_sysvec;
369 	p->p_flag = P_SYSTEM;
370 	p->p_sflag = PS_INMEM;
371 	p->p_state = PRS_NORMAL;
372 	knlist_init(&p->p_klist, &p->p_mtx);
373 	p->p_nice = NZERO;
374 	td->td_state = TDS_RUNNING;
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_oncpu = 0;
380 	p->p_peers = 0;
381 	p->p_leader = p;
382 
383 
384 	bcopy("swapper", p->p_comm, sizeof ("swapper"));
385 
386 	callout_init(&p->p_itcallout, CALLOUT_MPSAFE);
387 	callout_init(&td->td_slpcallout, CALLOUT_MPSAFE);
388 
389 	/* Create credentials. */
390 	p->p_ucred = crget();
391 	p->p_ucred->cr_ngroups = 1;	/* group 0 */
392 	p->p_ucred->cr_uidinfo = uifind(0);
393 	p->p_ucred->cr_ruidinfo = uifind(0);
394 	p->p_ucred->cr_prison = NULL;	/* Don't jail it. */
395 #ifdef MAC
396 	mac_create_proc0(p->p_ucred);
397 #endif
398 	td->td_ucred = crhold(p->p_ucred);
399 
400 	/* Create sigacts. */
401 	p->p_sigacts = sigacts_alloc();
402 
403 	/* Initialize signal state for process 0. */
404 	siginit(&proc0);
405 
406 	/* Create the file descriptor table. */
407 	p->p_fd = fdinit(NULL);
408 	p->p_fdtol = NULL;
409 
410 	/* Create the limits structures. */
411 	p->p_limit = lim_alloc();
412 	for (i = 0; i < RLIM_NLIMITS; i++)
413 		p->p_limit->pl_rlimit[i].rlim_cur =
414 		    p->p_limit->pl_rlimit[i].rlim_max = RLIM_INFINITY;
415 	p->p_limit->pl_rlimit[RLIMIT_NOFILE].rlim_cur =
416 	    p->p_limit->pl_rlimit[RLIMIT_NOFILE].rlim_max = maxfiles;
417 	p->p_limit->pl_rlimit[RLIMIT_NPROC].rlim_cur =
418 	    p->p_limit->pl_rlimit[RLIMIT_NPROC].rlim_max = maxproc;
419 	i = ptoa(cnt.v_free_count);
420 	p->p_limit->pl_rlimit[RLIMIT_RSS].rlim_max = i;
421 	p->p_limit->pl_rlimit[RLIMIT_MEMLOCK].rlim_max = i;
422 	p->p_limit->pl_rlimit[RLIMIT_MEMLOCK].rlim_cur = i / 3;
423 	p->p_cpulimit = RLIM_INFINITY;
424 
425 	p->p_stats = pstats_alloc();
426 
427 	/* Allocate a prototype map so we have something to fork. */
428 	pmap_pinit0(vmspace_pmap(&vmspace0));
429 	p->p_vmspace = &vmspace0;
430 	vmspace0.vm_refcnt = 1;
431 	vm_map_init(&vmspace0.vm_map, p->p_sysent->sv_minuser,
432 	    p->p_sysent->sv_maxuser);
433 	vmspace0.vm_map.pmap = vmspace_pmap(&vmspace0);
434 
435 	/*
436 	 * Charge root for one process.
437 	 */
438 	(void)chgproccnt(p->p_ucred->cr_ruidinfo, 1, 0);
439 }
440 SYSINIT(p0init, SI_SUB_INTRINSIC, SI_ORDER_FIRST, proc0_init, NULL)
441 
442 /* ARGSUSED*/
443 static void
444 proc0_post(void *dummy __unused)
445 {
446 	struct timespec ts;
447 	struct proc *p;
448 
449 	/*
450 	 * Now we can look at the time, having had a chance to verify the
451 	 * time from the filesystem.  Pretend that proc0 started now.
452 	 */
453 	sx_slock(&allproc_lock);
454 	LIST_FOREACH(p, &allproc, p_list) {
455 		microuptime(&p->p_stats->p_start);
456 		p->p_rux.rux_runtime.sec = 0;
457 		p->p_rux.rux_runtime.frac = 0;
458 	}
459 	sx_sunlock(&allproc_lock);
460 	binuptime(PCPU_PTR(switchtime));
461 	PCPU_SET(switchticks, ticks);
462 
463 	/*
464 	 * Give the ``random'' number generator a thump.
465 	 */
466 	nanotime(&ts);
467 	srandom(ts.tv_sec ^ ts.tv_nsec);
468 }
469 SYSINIT(p0post, SI_SUB_INTRINSIC_POST, SI_ORDER_FIRST, proc0_post, NULL)
470 
471 /*
472  ***************************************************************************
473  ****
474  **** The following SYSINIT's and glue code should be moved to the
475  **** respective files on a per subsystem basis.
476  ****
477  ***************************************************************************
478  */
479 
480 
481 /*
482  ***************************************************************************
483  ****
484  **** The following code probably belongs in another file, like
485  **** kern/init_init.c.
486  ****
487  ***************************************************************************
488  */
489 
490 /*
491  * List of paths to try when searching for "init".
492  */
493 static char init_path[MAXPATHLEN] =
494 #ifdef	INIT_PATH
495     __XSTRING(INIT_PATH);
496 #else
497     "/sbin/init:/sbin/oinit:/sbin/init.bak:/stand/sysinstall";
498 #endif
499 SYSCTL_STRING(_kern, OID_AUTO, init_path, CTLFLAG_RD, init_path, 0,
500 	"Path used to search the init process");
501 
502 /*
503  * Start the initial user process; try exec'ing each pathname in init_path.
504  * The program is invoked with one argument containing the boot flags.
505  */
506 static void
507 start_init(void *dummy)
508 {
509 	vm_offset_t addr;
510 	struct execve_args args;
511 	int options, error;
512 	char *var, *path, *next, *s;
513 	char *ucp, **uap, *arg0, *arg1;
514 	struct thread *td;
515 	struct proc *p;
516 	int init_does_devfs = 0;
517 
518 	mtx_lock(&Giant);
519 
520 	GIANT_REQUIRED;
521 
522 	td = curthread;
523 	p = td->td_proc;
524 
525 	vfs_mountroot();
526 
527 	/* Get the vnode for '/'.  Set p->p_fd->fd_cdir to reference it. */
528 	if (VFS_ROOT(TAILQ_FIRST(&mountlist), &rootvnode, td))
529 		panic("cannot find root vnode");
530 	FILEDESC_LOCK(p->p_fd);
531 	p->p_fd->fd_cdir = rootvnode;
532 	VREF(p->p_fd->fd_cdir);
533 	p->p_fd->fd_rdir = rootvnode;
534 	VREF(p->p_fd->fd_rdir);
535 	FILEDESC_UNLOCK(p->p_fd);
536 	VOP_UNLOCK(rootvnode, 0, td);
537 #ifdef MAC
538 	mac_create_root_mount(td->td_ucred, TAILQ_FIRST(&mountlist));
539 #endif
540 
541 	/*
542 	 * For disk based systems, we probably cannot do this yet
543 	 * since the fs will be read-only.  But a NFS root
544 	 * might be ok.  It is worth a shot.
545 	 */
546 	error = kern_mkdir(td, "/dev", UIO_SYSSPACE, 0700);
547 	if (error == EEXIST)
548 		error = 0;
549 	if (error == 0)
550 		error = kernel_vmount(0, "fstype", "devfs",
551 		    "fspath", "/dev", NULL);
552 	if (error != 0)
553 		init_does_devfs = 1;
554 
555 	/*
556 	 * Need just enough stack to hold the faked-up "execve()" arguments.
557 	 */
558 	addr = p->p_sysent->sv_usrstack - PAGE_SIZE;
559 	if (vm_map_find(&p->p_vmspace->vm_map, NULL, 0, &addr, PAGE_SIZE,
560 			FALSE, VM_PROT_ALL, VM_PROT_ALL, 0) != 0)
561 		panic("init: couldn't allocate argument space");
562 	p->p_vmspace->vm_maxsaddr = (caddr_t)addr;
563 	p->p_vmspace->vm_ssize = 1;
564 
565 	if ((var = getenv("init_path")) != NULL) {
566 		strlcpy(init_path, var, sizeof(init_path));
567 		freeenv(var);
568 	}
569 
570 	for (path = init_path; *path != '\0'; path = next) {
571 		while (*path == ':')
572 			path++;
573 		if (*path == '\0')
574 			break;
575 		for (next = path; *next != '\0' && *next != ':'; next++)
576 			/* nothing */ ;
577 		if (bootverbose)
578 			printf("start_init: trying %.*s\n", (int)(next - path),
579 			    path);
580 
581 		/*
582 		 * Move out the boot flag argument.
583 		 */
584 		options = 0;
585 		ucp = (char *)p->p_sysent->sv_usrstack;
586 		(void)subyte(--ucp, 0);		/* trailing zero */
587 		if (boothowto & RB_SINGLE) {
588 			(void)subyte(--ucp, 's');
589 			options = 1;
590 		}
591 #ifdef notyet
592                 if (boothowto & RB_FASTBOOT) {
593 			(void)subyte(--ucp, 'f');
594 			options = 1;
595 		}
596 #endif
597 
598 #ifdef BOOTCDROM
599 		(void)subyte(--ucp, 'C');
600 		options = 1;
601 #endif
602 		if (init_does_devfs) {
603 			(void)subyte(--ucp, 'd');
604 			options = 1;
605 		}
606 
607 		if (options == 0)
608 			(void)subyte(--ucp, '-');
609 		(void)subyte(--ucp, '-');		/* leading hyphen */
610 		arg1 = ucp;
611 
612 		/*
613 		 * Move out the file name (also arg 0).
614 		 */
615 		(void)subyte(--ucp, 0);
616 		for (s = next - 1; s >= path; s--)
617 			(void)subyte(--ucp, *s);
618 		arg0 = ucp;
619 
620 		/*
621 		 * Move out the arg pointers.
622 		 */
623 		uap = (char **)((intptr_t)ucp & ~(sizeof(intptr_t)-1));
624 		(void)suword((caddr_t)--uap, (long)0);	/* terminator */
625 		(void)suword((caddr_t)--uap, (long)(intptr_t)arg1);
626 		(void)suword((caddr_t)--uap, (long)(intptr_t)arg0);
627 
628 		/*
629 		 * Point at the arguments.
630 		 */
631 		args.fname = arg0;
632 		args.argv = uap;
633 		args.envv = NULL;
634 
635 		/*
636 		 * Now try to exec the program.  If can't for any reason
637 		 * other than it doesn't exist, complain.
638 		 *
639 		 * Otherwise, return via fork_trampoline() all the way
640 		 * to user mode as init!
641 		 */
642 		if ((error = execve(td, &args)) == 0) {
643 			mtx_unlock(&Giant);
644 			return;
645 		}
646 		if (error != ENOENT)
647 			printf("exec %.*s: error %d\n", (int)(next - path),
648 			    path, error);
649 	}
650 	printf("init: not found in path %s\n", init_path);
651 	panic("no init");
652 }
653 
654 /*
655  * Like kthread_create(), but runs in it's own address space.
656  * We do this early to reserve pid 1.
657  *
658  * Note special case - do not make it runnable yet.  Other work
659  * in progress will change this more.
660  */
661 static void
662 create_init(const void *udata __unused)
663 {
664 	struct ucred *newcred, *oldcred;
665 	int error;
666 
667 	error = fork1(&thread0, RFFDG | RFPROC | RFSTOPPED, 0, &initproc);
668 	if (error)
669 		panic("cannot fork init: %d\n", error);
670 	KASSERT(initproc->p_pid == 1, ("create_init: initproc->p_pid != 1"));
671 	/* divorce init's credentials from the kernel's */
672 	newcred = crget();
673 	PROC_LOCK(initproc);
674 	initproc->p_flag |= P_SYSTEM;
675 	oldcred = initproc->p_ucred;
676 	crcopy(newcred, oldcred);
677 #ifdef MAC
678 	mac_create_proc1(newcred);
679 #endif
680 	initproc->p_ucred = newcred;
681 	PROC_UNLOCK(initproc);
682 	crfree(oldcred);
683 	cred_update_thread(FIRST_THREAD_IN_PROC(initproc));
684 	mtx_lock_spin(&sched_lock);
685 	initproc->p_sflag |= PS_INMEM;
686 	mtx_unlock_spin(&sched_lock);
687 	cpu_set_fork_handler(FIRST_THREAD_IN_PROC(initproc), start_init, NULL);
688 }
689 SYSINIT(init, SI_SUB_CREATE_INIT, SI_ORDER_FIRST, create_init, NULL)
690 
691 /*
692  * Make it runnable now.
693  */
694 static void
695 kick_init(const void *udata __unused)
696 {
697 	struct thread *td;
698 
699 	td = FIRST_THREAD_IN_PROC(initproc);
700 	mtx_lock_spin(&sched_lock);
701 	TD_SET_CAN_RUN(td);
702 	setrunqueue(td, SRQ_BORING);	/* XXXKSE */
703 	mtx_unlock_spin(&sched_lock);
704 }
705 SYSINIT(kickinit, SI_SUB_KTHREAD_INIT, SI_ORDER_FIRST, kick_init, NULL)
706