init_main.c revision 198295 
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: head/sys/kern/init_main.c 198295 2009-10-20 16:36:51Z ru $");
46
47#include "opt_ddb.h"
48#include "opt_init_path.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/jail.h>
56#include <sys/ktr.h>
57#include <sys/lock.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#include <sys/cpuset.h>
77
78#include <machine/cpu.h>
79
80#include <security/audit/audit.h>
81#include <security/mac/mac_framework.h>
82
83#include <vm/vm.h>
84#include <vm/vm_param.h>
85#include <vm/pmap.h>
86#include <vm/vm_map.h>
87#include <sys/copyright.h>
88
89#include <ddb/ddb.h>
90#include <ddb/db_sym.h>
91
92void mi_startup(void);				/* Should be elsewhere */
93
94/* Components of the first process -- never freed. */
95static struct session session0;
96static struct pgrp pgrp0;
97struct	proc proc0;
98struct	thread thread0 __aligned(16);
99struct	vmspace vmspace0;
100struct	proc *initproc;
101
102int	boothowto = 0;		/* initialized so that it can be patched */
103SYSCTL_INT(_debug, OID_AUTO, boothowto, CTLFLAG_RD, &boothowto, 0, "");
104int	bootverbose;
105SYSCTL_INT(_debug, OID_AUTO, bootverbose, CTLFLAG_RW, &bootverbose, 0, "");
106
107/*
108 * This ensures that there is at least one entry so that the sysinit_set
109 * symbol is not undefined.  A sybsystem ID of SI_SUB_DUMMY is never
110 * executed.
111 */
112SYSINIT(placeholder, SI_SUB_DUMMY, SI_ORDER_ANY, NULL, NULL);
113
114/*
115 * The sysinit table itself.  Items are checked off as the are run.
116 * If we want to register new sysinit types, add them to newsysinit.
117 */
118SET_DECLARE(sysinit_set, struct sysinit);
119struct sysinit **sysinit, **sysinit_end;
120struct sysinit **newsysinit, **newsysinit_end;
121
122/*
123 * Merge a new sysinit set into the current set, reallocating it if
124 * necessary.  This can only be called after malloc is running.
125 */
126void
127sysinit_add(struct sysinit **set, struct sysinit **set_end)
128{
129	struct sysinit **newset;
130	struct sysinit **sipp;
131	struct sysinit **xipp;
132	int count;
133
134	count = set_end - set;
135	if (newsysinit)
136		count += newsysinit_end - newsysinit;
137	else
138		count += sysinit_end - sysinit;
139	newset = malloc(count * sizeof(*sipp), M_TEMP, M_NOWAIT);
140	if (newset == NULL)
141		panic("cannot malloc for sysinit");
142	xipp = newset;
143	if (newsysinit)
144		for (sipp = newsysinit; sipp < newsysinit_end; sipp++)
145			*xipp++ = *sipp;
146	else
147		for (sipp = sysinit; sipp < sysinit_end; sipp++)
148			*xipp++ = *sipp;
149	for (sipp = set; sipp < set_end; sipp++)
150		*xipp++ = *sipp;
151	if (newsysinit)
152		free(newsysinit, M_TEMP);
153	newsysinit = newset;
154	newsysinit_end = newset + count;
155}
156
157/*
158 * System startup; initialize the world, create process 0, mount root
159 * filesystem, and fork to create init and pagedaemon.  Most of the
160 * hard work is done in the lower-level initialization routines including
161 * startup(), which does memory initialization and autoconfiguration.
162 *
163 * This allows simple addition of new kernel subsystems that require
164 * boot time initialization.  It also allows substitution of subsystem
165 * (for instance, a scheduler, kernel profiler, or VM system) by object
166 * module.  Finally, it allows for optional "kernel threads".
167 */
168void
169mi_startup(void)
170{
171
172	register struct sysinit **sipp;		/* system initialization*/
173	register struct sysinit **xipp;		/* interior loop of sort*/
174	register struct sysinit *save;		/* bubble*/
175
176#if defined(VERBOSE_SYSINIT)
177	int last;
178	int verbose;
179#endif
180
181	if (sysinit == NULL) {
182		sysinit = SET_BEGIN(sysinit_set);
183		sysinit_end = SET_LIMIT(sysinit_set);
184	}
185
186restart:
187	/*
188	 * Perform a bubble sort of the system initialization objects by
189	 * their subsystem (primary key) and order (secondary key).
190	 */
191	for (sipp = sysinit; sipp < sysinit_end; sipp++) {
192		for (xipp = sipp + 1; xipp < sysinit_end; xipp++) {
193			if ((*sipp)->subsystem < (*xipp)->subsystem ||
194			     ((*sipp)->subsystem == (*xipp)->subsystem &&
195			      (*sipp)->order <= (*xipp)->order))
196				continue;	/* skip*/
197			save = *sipp;
198			*sipp = *xipp;
199			*xipp = save;
200		}
201	}
202
203#if defined(VERBOSE_SYSINIT)
204	last = SI_SUB_COPYRIGHT;
205	verbose = 0;
206#if !defined(DDB)
207	printf("VERBOSE_SYSINIT: DDB not enabled, symbol lookups disabled.\n");
208#endif
209#endif
210
211	/*
212	 * Traverse the (now) ordered list of system initialization tasks.
213	 * Perform each task, and continue on to the next task.
214	 *
215	 * The last item on the list is expected to be the scheduler,
216	 * which will not return.
217	 */
218	for (sipp = sysinit; sipp < sysinit_end; sipp++) {
219
220		if ((*sipp)->subsystem == SI_SUB_DUMMY)
221			continue;	/* skip dummy task(s)*/
222
223		if ((*sipp)->subsystem == SI_SUB_DONE)
224			continue;
225
226#if defined(VERBOSE_SYSINIT)
227		if ((*sipp)->subsystem > last) {
228			verbose = 1;
229			last = (*sipp)->subsystem;
230			printf("subsystem %x\n", last);
231		}
232		if (verbose) {
233#if defined(DDB)
234			const char *name;
235			c_db_sym_t sym;
236			db_expr_t  offset;
237
238			sym = db_search_symbol((vm_offset_t)(*sipp)->func,
239			    DB_STGY_PROC, &offset);
240			db_symbol_values(sym, &name, NULL);
241			if (name != NULL)
242				printf("   %s(%p)... ", name, (*sipp)->udata);
243			else
244#endif
245				printf("   %p(%p)... ", (*sipp)->func,
246				    (*sipp)->udata);
247		}
248#endif
249
250		/* Call function */
251		(*((*sipp)->func))((*sipp)->udata);
252
253#if defined(VERBOSE_SYSINIT)
254		if (verbose)
255			printf("done.\n");
256#endif
257
258		/* Check off the one we're just done */
259		(*sipp)->subsystem = SI_SUB_DONE;
260
261		/* Check if we've installed more sysinit items via KLD */
262		if (newsysinit != NULL) {
263			if (sysinit != SET_BEGIN(sysinit_set))
264				free(sysinit, M_TEMP);
265			sysinit = newsysinit;
266			sysinit_end = newsysinit_end;
267			newsysinit = NULL;
268			newsysinit_end = NULL;
269			goto restart;
270		}
271	}
272
273	panic("Shouldn't get here!");
274	/* NOTREACHED*/
275}
276
277
278/*
279 ***************************************************************************
280 ****
281 **** The following SYSINIT's belong elsewhere, but have not yet
282 **** been moved.
283 ****
284 ***************************************************************************
285 */
286static void
287print_caddr_t(void *data)
288{
289	printf("%s", (char *)data);
290}
291
292static void
293print_version(void *data __unused)
294{
295	int len;
296
297	/* Strip a trailing newline from version. */
298	len = strlen(version);
299	while (len > 0 && version[len - 1] == '\n')
300		len--;
301	printf("%.*s %s\n", len, version, machine);
302}
303
304SYSINIT(announce, SI_SUB_COPYRIGHT, SI_ORDER_FIRST, print_caddr_t,
305    copyright);
306SYSINIT(trademark, SI_SUB_COPYRIGHT, SI_ORDER_SECOND, print_caddr_t,
307    trademark);
308SYSINIT(version, SI_SUB_COPYRIGHT, SI_ORDER_THIRD, print_version, NULL);
309
310#ifdef WITNESS
311static char wit_warn[] =
312     "WARNING: WITNESS option enabled, expect reduced performance.\n";
313SYSINIT(witwarn, SI_SUB_COPYRIGHT, SI_ORDER_THIRD + 1,
314   print_caddr_t, wit_warn);
315SYSINIT(witwarn2, SI_SUB_RUN_SCHEDULER, SI_ORDER_THIRD + 1,
316   print_caddr_t, wit_warn);
317#endif
318
319#ifdef DIAGNOSTIC
320static char diag_warn[] =
321     "WARNING: DIAGNOSTIC option enabled, expect reduced performance.\n";
322SYSINIT(diagwarn, SI_SUB_COPYRIGHT, SI_ORDER_THIRD + 2,
323    print_caddr_t, diag_warn);
324SYSINIT(diagwarn2, SI_SUB_RUN_SCHEDULER, SI_ORDER_THIRD + 2,
325    print_caddr_t, diag_warn);
326#endif
327
328static void
329set_boot_verbose(void *data __unused)
330{
331
332	if (boothowto & RB_VERBOSE)
333		bootverbose++;
334}
335SYSINIT(boot_verbose, SI_SUB_TUNABLES, SI_ORDER_ANY, set_boot_verbose, NULL);
336
337struct sysentvec null_sysvec = {
338	.sv_size	= 0,
339	.sv_table	= NULL,
340	.sv_mask	= 0,
341	.sv_sigsize	= 0,
342	.sv_sigtbl	= NULL,
343	.sv_errsize	= 0,
344	.sv_errtbl	= NULL,
345	.sv_transtrap	= NULL,
346	.sv_fixup	= NULL,
347	.sv_sendsig	= NULL,
348	.sv_sigcode	= NULL,
349	.sv_szsigcode	= NULL,
350	.sv_prepsyscall	= NULL,
351	.sv_name	= "null",
352	.sv_coredump	= NULL,
353	.sv_imgact_try	= NULL,
354	.sv_minsigstksz	= 0,
355	.sv_pagesize	= PAGE_SIZE,
356	.sv_minuser	= VM_MIN_ADDRESS,
357	.sv_maxuser	= VM_MAXUSER_ADDRESS,
358	.sv_usrstack	= USRSTACK,
359	.sv_psstrings	= PS_STRINGS,
360	.sv_stackprot	= VM_PROT_ALL,
361	.sv_copyout_strings	= NULL,
362	.sv_setregs	= NULL,
363	.sv_fixlimit	= NULL,
364	.sv_maxssiz	= NULL
365};
366
367/*
368 ***************************************************************************
369 ****
370 **** The two following SYSINIT's are proc0 specific glue code.  I am not
371 **** convinced that they can not be safely combined, but their order of
372 **** operation has been maintained as the same as the original init_main.c
373 **** for right now.
374 ****
375 **** These probably belong in init_proc.c or kern_proc.c, since they
376 **** deal with proc0 (the fork template process).
377 ****
378 ***************************************************************************
379 */
380/* ARGSUSED*/
381static void
382proc0_init(void *dummy __unused)
383{
384	struct proc *p;
385	unsigned i;
386	struct thread *td;
387
388	GIANT_REQUIRED;
389	p = &proc0;
390	td = &thread0;
391
392	/*
393	 * Initialize magic number and osrel.
394	 */
395	p->p_magic = P_MAGIC;
396	p->p_osrel = osreldate;
397
398	/*
399	 * Initialize thread and process structures.
400	 */
401	procinit();	/* set up proc zone */
402	threadinit();	/* set up UMA zones */
403
404	/*
405	 * Initialise scheduler resources.
406	 * Add scheduler specific parts to proc, thread as needed.
407	 */
408	schedinit();	/* scheduler gets its house in order */
409	/*
410	 * Initialize sleep queue hash table
411	 */
412	sleepinit();
413
414	/*
415	 * additional VM structures
416	 */
417	vm_init2();
418
419	/*
420	 * Create process 0 (the swapper).
421	 */
422	LIST_INSERT_HEAD(&allproc, p, p_list);
423	LIST_INSERT_HEAD(PIDHASH(0), p, p_hash);
424	mtx_init(&pgrp0.pg_mtx, "process group", NULL, MTX_DEF | MTX_DUPOK);
425	p->p_pgrp = &pgrp0;
426	LIST_INSERT_HEAD(PGRPHASH(0), &pgrp0, pg_hash);
427	LIST_INIT(&pgrp0.pg_members);
428	LIST_INSERT_HEAD(&pgrp0.pg_members, p, p_pglist);
429
430	pgrp0.pg_session = &session0;
431	mtx_init(&session0.s_mtx, "session", NULL, MTX_DEF);
432	refcount_init(&session0.s_count, 1);
433	session0.s_leader = p;
434
435	p->p_sysent = &null_sysvec;
436	p->p_flag = P_SYSTEM | P_INMEM;
437	p->p_state = PRS_NORMAL;
438	knlist_init_mtx(&p->p_klist, &p->p_mtx);
439	STAILQ_INIT(&p->p_ktr);
440	p->p_nice = NZERO;
441	td->td_tid = PID_MAX + 1;
442	td->td_state = TDS_RUNNING;
443	td->td_pri_class = PRI_TIMESHARE;
444	td->td_user_pri = PUSER;
445	td->td_base_user_pri = PUSER;
446	td->td_priority = PVM;
447	td->td_base_pri = PUSER;
448	td->td_oncpu = 0;
449	td->td_flags = TDF_INMEM|TDP_KTHREAD;
450	td->td_cpuset = cpuset_thread0();
451	prison0.pr_cpuset = cpuset_ref(td->td_cpuset);
452	p->p_peers = 0;
453	p->p_leader = p;
454
455
456	strncpy(p->p_comm, "kernel", sizeof (p->p_comm));
457	strncpy(td->td_name, "swapper", sizeof (td->td_name));
458
459	callout_init(&p->p_itcallout, CALLOUT_MPSAFE);
460	callout_init_mtx(&p->p_limco, &p->p_mtx, 0);
461	callout_init(&td->td_slpcallout, CALLOUT_MPSAFE);
462
463	/* Create credentials. */
464	p->p_ucred = crget();
465	p->p_ucred->cr_ngroups = 1;	/* group 0 */
466	p->p_ucred->cr_uidinfo = uifind(0);
467	p->p_ucred->cr_ruidinfo = uifind(0);
468	p->p_ucred->cr_prison = &prison0;
469#ifdef AUDIT
470	audit_cred_kproc0(p->p_ucred);
471#endif
472#ifdef MAC
473	mac_cred_create_swapper(p->p_ucred);
474#endif
475	td->td_ucred = crhold(p->p_ucred);
476
477	/* Create sigacts. */
478	p->p_sigacts = sigacts_alloc();
479
480	/* Initialize signal state for process 0. */
481	siginit(&proc0);
482
483	/* Create the file descriptor table. */
484	p->p_fd = fdinit(NULL);
485	p->p_fdtol = NULL;
486
487	/* Create the limits structures. */
488	p->p_limit = lim_alloc();
489	for (i = 0; i < RLIM_NLIMITS; i++)
490		p->p_limit->pl_rlimit[i].rlim_cur =
491		    p->p_limit->pl_rlimit[i].rlim_max = RLIM_INFINITY;
492	p->p_limit->pl_rlimit[RLIMIT_NOFILE].rlim_cur =
493	    p->p_limit->pl_rlimit[RLIMIT_NOFILE].rlim_max = maxfiles;
494	p->p_limit->pl_rlimit[RLIMIT_NPROC].rlim_cur =
495	    p->p_limit->pl_rlimit[RLIMIT_NPROC].rlim_max = maxproc;
496	i = ptoa(cnt.v_free_count);
497	p->p_limit->pl_rlimit[RLIMIT_RSS].rlim_max = i;
498	p->p_limit->pl_rlimit[RLIMIT_MEMLOCK].rlim_max = i;
499	p->p_limit->pl_rlimit[RLIMIT_MEMLOCK].rlim_cur = i / 3;
500	p->p_cpulimit = RLIM_INFINITY;
501
502	p->p_stats = pstats_alloc();
503
504	/* Allocate a prototype map so we have something to fork. */
505	pmap_pinit0(vmspace_pmap(&vmspace0));
506	p->p_vmspace = &vmspace0;
507	vmspace0.vm_refcnt = 1;
508
509	/*
510	 * proc0 is not expected to enter usermode, so there is no special
511	 * handling for sv_minuser here, like is done for exec_new_vmspace().
512	 */
513	vm_map_init(&vmspace0.vm_map, p->p_sysent->sv_minuser,
514	    p->p_sysent->sv_maxuser);
515	vmspace0.vm_map.pmap = vmspace_pmap(&vmspace0);
516
517	/*-
518	 * call the init and ctor for the new thread and proc
519	 * we wait to do this until all other structures
520	 * are fairly sane.
521	 */
522	EVENTHANDLER_INVOKE(process_init, p);
523	EVENTHANDLER_INVOKE(thread_init, td);
524	EVENTHANDLER_INVOKE(process_ctor, p);
525	EVENTHANDLER_INVOKE(thread_ctor, td);
526
527	/*
528	 * Charge root for one process.
529	 */
530	(void)chgproccnt(p->p_ucred->cr_ruidinfo, 1, 0);
531}
532SYSINIT(p0init, SI_SUB_INTRINSIC, SI_ORDER_FIRST, proc0_init, NULL);
533
534/* ARGSUSED*/
535static void
536proc0_post(void *dummy __unused)
537{
538	struct timespec ts;
539	struct proc *p;
540	struct rusage ru;
541	struct thread *td;
542
543	/*
544	 * Now we can look at the time, having had a chance to verify the
545	 * time from the filesystem.  Pretend that proc0 started now.
546	 */
547	sx_slock(&allproc_lock);
548	FOREACH_PROC_IN_SYSTEM(p) {
549		microuptime(&p->p_stats->p_start);
550		PROC_SLOCK(p);
551		rufetch(p, &ru);	/* Clears thread stats */
552		PROC_SUNLOCK(p);
553		p->p_rux.rux_runtime = 0;
554		p->p_rux.rux_uticks = 0;
555		p->p_rux.rux_sticks = 0;
556		p->p_rux.rux_iticks = 0;
557		FOREACH_THREAD_IN_PROC(p, td) {
558			td->td_runtime = 0;
559		}
560	}
561	sx_sunlock(&allproc_lock);
562	PCPU_SET(switchtime, cpu_ticks());
563	PCPU_SET(switchticks, ticks);
564
565	/*
566	 * Give the ``random'' number generator a thump.
567	 */
568	nanotime(&ts);
569	srandom(ts.tv_sec ^ ts.tv_nsec);
570}
571SYSINIT(p0post, SI_SUB_INTRINSIC_POST, SI_ORDER_FIRST, proc0_post, NULL);
572
573static void
574random_init(void *dummy __unused)
575{
576
577	/*
578	 * After CPU has been started we have some randomness on most
579	 * platforms via get_cyclecount().  For platforms that don't
580	 * we will reseed random(9) in proc0_post() as well.
581	 */
582	srandom(get_cyclecount());
583}
584SYSINIT(random, SI_SUB_RANDOM, SI_ORDER_FIRST, random_init, NULL);
585
586/*
587 ***************************************************************************
588 ****
589 **** The following SYSINIT's and glue code should be moved to the
590 **** respective files on a per subsystem basis.
591 ****
592 ***************************************************************************
593 */
594
595
596/*
597 ***************************************************************************
598 ****
599 **** The following code probably belongs in another file, like
600 **** kern/init_init.c.
601 ****
602 ***************************************************************************
603 */
604
605/*
606 * List of paths to try when searching for "init".
607 */
608static char init_path[MAXPATHLEN] =
609#ifdef	INIT_PATH
610    __XSTRING(INIT_PATH);
611#else
612    "/sbin/init:/sbin/oinit:/sbin/init.bak:/rescue/init:/stand/sysinstall";
613#endif
614SYSCTL_STRING(_kern, OID_AUTO, init_path, CTLFLAG_RD, init_path, 0,
615	"Path used to search the init process");
616
617/*
618 * Shutdown timeout of init(8).
619 * Unused within kernel, but used to control init(8), hence do not remove.
620 */
621#ifndef INIT_SHUTDOWN_TIMEOUT
622#define INIT_SHUTDOWN_TIMEOUT 120
623#endif
624static int init_shutdown_timeout = INIT_SHUTDOWN_TIMEOUT;
625SYSCTL_INT(_kern, OID_AUTO, init_shutdown_timeout,
626	CTLFLAG_RW, &init_shutdown_timeout, 0, "");
627
628/*
629 * Start the initial user process; try exec'ing each pathname in init_path.
630 * The program is invoked with one argument containing the boot flags.
631 */
632static void
633start_init(void *dummy)
634{
635	vm_offset_t addr;
636	struct execve_args args;
637	int options, error;
638	char *var, *path, *next, *s;
639	char *ucp, **uap, *arg0, *arg1;
640	struct thread *td;
641	struct proc *p;
642
643	mtx_lock(&Giant);
644
645	GIANT_REQUIRED;
646
647	td = curthread;
648	p = td->td_proc;
649
650	vfs_mountroot();
651
652	/*
653	 * Need just enough stack to hold the faked-up "execve()" arguments.
654	 */
655	addr = p->p_sysent->sv_usrstack - PAGE_SIZE;
656	if (vm_map_find(&p->p_vmspace->vm_map, NULL, 0, &addr, PAGE_SIZE,
657			FALSE, VM_PROT_ALL, VM_PROT_ALL, 0) != 0)
658		panic("init: couldn't allocate argument space");
659	p->p_vmspace->vm_maxsaddr = (caddr_t)addr;
660	p->p_vmspace->vm_ssize = 1;
661
662	if ((var = getenv("init_path")) != NULL) {
663		strlcpy(init_path, var, sizeof(init_path));
664		freeenv(var);
665	}
666
667	for (path = init_path; *path != '\0'; path = next) {
668		while (*path == ':')
669			path++;
670		if (*path == '\0')
671			break;
672		for (next = path; *next != '\0' && *next != ':'; next++)
673			/* nothing */ ;
674		if (bootverbose)
675			printf("start_init: trying %.*s\n", (int)(next - path),
676			    path);
677
678		/*
679		 * Move out the boot flag argument.
680		 */
681		options = 0;
682		ucp = (char *)p->p_sysent->sv_usrstack;
683		(void)subyte(--ucp, 0);		/* trailing zero */
684		if (boothowto & RB_SINGLE) {
685			(void)subyte(--ucp, 's');
686			options = 1;
687		}
688#ifdef notyet
689                if (boothowto & RB_FASTBOOT) {
690			(void)subyte(--ucp, 'f');
691			options = 1;
692		}
693#endif
694
695#ifdef BOOTCDROM
696		(void)subyte(--ucp, 'C');
697		options = 1;
698#endif
699
700		if (options == 0)
701			(void)subyte(--ucp, '-');
702		(void)subyte(--ucp, '-');		/* leading hyphen */
703		arg1 = ucp;
704
705		/*
706		 * Move out the file name (also arg 0).
707		 */
708		(void)subyte(--ucp, 0);
709		for (s = next - 1; s >= path; s--)
710			(void)subyte(--ucp, *s);
711		arg0 = ucp;
712
713		/*
714		 * Move out the arg pointers.
715		 */
716		uap = (char **)((intptr_t)ucp & ~(sizeof(intptr_t)-1));
717		(void)suword((caddr_t)--uap, (long)0);	/* terminator */
718		(void)suword((caddr_t)--uap, (long)(intptr_t)arg1);
719		(void)suword((caddr_t)--uap, (long)(intptr_t)arg0);
720
721		/*
722		 * Point at the arguments.
723		 */
724		args.fname = arg0;
725		args.argv = uap;
726		args.envv = NULL;
727
728		/*
729		 * Now try to exec the program.  If can't for any reason
730		 * other than it doesn't exist, complain.
731		 *
732		 * Otherwise, return via fork_trampoline() all the way
733		 * to user mode as init!
734		 */
735		if ((error = execve(td, &args)) == 0) {
736			mtx_unlock(&Giant);
737			return;
738		}
739		if (error != ENOENT)
740			printf("exec %.*s: error %d\n", (int)(next - path),
741			    path, error);
742	}
743	printf("init: not found in path %s\n", init_path);
744	panic("no init");
745}
746
747/*
748 * Like kproc_create(), but runs in it's own address space.
749 * We do this early to reserve pid 1.
750 *
751 * Note special case - do not make it runnable yet.  Other work
752 * in progress will change this more.
753 */
754static void
755create_init(const void *udata __unused)
756{
757	struct ucred *newcred, *oldcred;
758	int error;
759
760	error = fork1(&thread0, RFFDG | RFPROC | RFSTOPPED, 0, &initproc);
761	if (error)
762		panic("cannot fork init: %d\n", error);
763	KASSERT(initproc->p_pid == 1, ("create_init: initproc->p_pid != 1"));
764	/* divorce init's credentials from the kernel's */
765	newcred = crget();
766	PROC_LOCK(initproc);
767	initproc->p_flag |= P_SYSTEM | P_INMEM;
768	oldcred = initproc->p_ucred;
769	crcopy(newcred, oldcred);
770#ifdef MAC
771	mac_cred_create_init(newcred);
772#endif
773#ifdef AUDIT
774	audit_cred_proc1(newcred);
775#endif
776	initproc->p_ucred = newcred;
777	PROC_UNLOCK(initproc);
778	crfree(oldcred);
779	cred_update_thread(FIRST_THREAD_IN_PROC(initproc));
780	cpu_set_fork_handler(FIRST_THREAD_IN_PROC(initproc), start_init, NULL);
781}
782SYSINIT(init, SI_SUB_CREATE_INIT, SI_ORDER_FIRST, create_init, NULL);
783
784/*
785 * Make it runnable now.
786 */
787static void
788kick_init(const void *udata __unused)
789{
790	struct thread *td;
791
792	td = FIRST_THREAD_IN_PROC(initproc);
793	thread_lock(td);
794	TD_SET_CAN_RUN(td);
795	sched_add(td, SRQ_BORING);
796	thread_unlock(td);
797}
798SYSINIT(kickinit, SI_SUB_KTHREAD_INIT, SI_ORDER_FIRST, kick_init, NULL);
799