xref: /freebsd/sys/kern/init_main.c (revision 195ebc7e9e4b129de810833791a19dfb4349d6a9)
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_ddb.h"
48 #include "opt_init_path.h"
49 #include "opt_mac.h"
50 
51 #include <sys/param.h>
52 #include <sys/kernel.h>
53 #include <sys/exec.h>
54 #include <sys/file.h>
55 #include <sys/filedesc.h>
56 #include <sys/jail.h>
57 #include <sys/ktr.h>
58 #include <sys/lock.h>
59 #include <sys/mount.h>
60 #include <sys/mutex.h>
61 #include <sys/syscallsubr.h>
62 #include <sys/sysctl.h>
63 #include <sys/proc.h>
64 #include <sys/resourcevar.h>
65 #include <sys/systm.h>
66 #include <sys/signalvar.h>
67 #include <sys/vnode.h>
68 #include <sys/sysent.h>
69 #include <sys/reboot.h>
70 #include <sys/sched.h>
71 #include <sys/sx.h>
72 #include <sys/sysproto.h>
73 #include <sys/vmmeter.h>
74 #include <sys/unistd.h>
75 #include <sys/malloc.h>
76 #include <sys/conf.h>
77 #include <sys/cpuset.h>
78 #include <sys/vimage.h>
79 
80 #include <machine/cpu.h>
81 
82 #include <security/audit/audit.h>
83 #include <security/mac/mac_framework.h>
84 
85 #include <vm/vm.h>
86 #include <vm/vm_param.h>
87 #include <vm/pmap.h>
88 #include <vm/vm_map.h>
89 #include <sys/copyright.h>
90 
91 #include <ddb/ddb.h>
92 #include <ddb/db_sym.h>
93 
94 void mi_startup(void);				/* Should be elsewhere */
95 
96 /* Components of the first process -- never freed. */
97 static struct session session0;
98 static struct pgrp pgrp0;
99 struct	proc proc0;
100 struct	thread thread0 __aligned(16);
101 struct	vmspace vmspace0;
102 struct	proc *initproc;
103 
104 int	boothowto = 0;		/* initialized so that it can be patched */
105 SYSCTL_INT(_debug, OID_AUTO, boothowto, CTLFLAG_RD, &boothowto, 0, "");
106 int	bootverbose;
107 SYSCTL_INT(_debug, OID_AUTO, bootverbose, CTLFLAG_RW, &bootverbose, 0, "");
108 
109 /*
110  * This ensures that there is at least one entry so that the sysinit_set
111  * symbol is not undefined.  A sybsystem ID of SI_SUB_DUMMY is never
112  * executed.
113  */
114 SYSINIT(placeholder, SI_SUB_DUMMY, SI_ORDER_ANY, NULL, NULL);
115 
116 /*
117  * The sysinit table itself.  Items are checked off as the are run.
118  * If we want to register new sysinit types, add them to newsysinit.
119  */
120 SET_DECLARE(sysinit_set, struct sysinit);
121 struct sysinit **sysinit, **sysinit_end;
122 struct sysinit **newsysinit, **newsysinit_end;
123 
124 /*
125  * Merge a new sysinit set into the current set, reallocating it if
126  * necessary.  This can only be called after malloc is running.
127  */
128 void
129 sysinit_add(struct sysinit **set, struct sysinit **set_end)
130 {
131 	struct sysinit **newset;
132 	struct sysinit **sipp;
133 	struct sysinit **xipp;
134 	int count;
135 
136 	count = set_end - set;
137 	if (newsysinit)
138 		count += newsysinit_end - newsysinit;
139 	else
140 		count += sysinit_end - sysinit;
141 	newset = malloc(count * sizeof(*sipp), M_TEMP, M_NOWAIT);
142 	if (newset == NULL)
143 		panic("cannot malloc for sysinit");
144 	xipp = newset;
145 	if (newsysinit)
146 		for (sipp = newsysinit; sipp < newsysinit_end; sipp++)
147 			*xipp++ = *sipp;
148 	else
149 		for (sipp = sysinit; sipp < sysinit_end; sipp++)
150 			*xipp++ = *sipp;
151 	for (sipp = set; sipp < set_end; sipp++)
152 		*xipp++ = *sipp;
153 	if (newsysinit)
154 		free(newsysinit, M_TEMP);
155 	newsysinit = newset;
156 	newsysinit_end = newset + count;
157 }
158 
159 /*
160  * System startup; initialize the world, create process 0, mount root
161  * filesystem, and fork to create init and pagedaemon.  Most of the
162  * hard work is done in the lower-level initialization routines including
163  * startup(), which does memory initialization and autoconfiguration.
164  *
165  * This allows simple addition of new kernel subsystems that require
166  * boot time initialization.  It also allows substitution of subsystem
167  * (for instance, a scheduler, kernel profiler, or VM system) by object
168  * module.  Finally, it allows for optional "kernel threads".
169  */
170 void
171 mi_startup(void)
172 {
173 
174 	register struct sysinit **sipp;		/* system initialization*/
175 	register struct sysinit **xipp;		/* interior loop of sort*/
176 	register struct sysinit *save;		/* bubble*/
177 
178 #if defined(VERBOSE_SYSINIT)
179 	int last;
180 	int verbose;
181 #endif
182 
183 	if (sysinit == NULL) {
184 		sysinit = SET_BEGIN(sysinit_set);
185 		sysinit_end = SET_LIMIT(sysinit_set);
186 	}
187 
188 restart:
189 	/*
190 	 * Perform a bubble sort of the system initialization objects by
191 	 * their subsystem (primary key) and order (secondary key).
192 	 */
193 	for (sipp = sysinit; sipp < sysinit_end; sipp++) {
194 		for (xipp = sipp + 1; xipp < sysinit_end; xipp++) {
195 			if ((*sipp)->subsystem < (*xipp)->subsystem ||
196 			     ((*sipp)->subsystem == (*xipp)->subsystem &&
197 			      (*sipp)->order <= (*xipp)->order))
198 				continue;	/* skip*/
199 			save = *sipp;
200 			*sipp = *xipp;
201 			*xipp = save;
202 		}
203 	}
204 
205 #if defined(VERBOSE_SYSINIT)
206 	last = SI_SUB_COPYRIGHT;
207 	verbose = 0;
208 #if !defined(DDB)
209 	printf("VERBOSE_SYSINIT: DDB not enabled, symbol lookups disabled.\n");
210 #endif
211 #endif
212 
213 	/*
214 	 * Traverse the (now) ordered list of system initialization tasks.
215 	 * Perform each task, and continue on to the next task.
216 	 *
217 	 * The last item on the list is expected to be the scheduler,
218 	 * which will not return.
219 	 */
220 	for (sipp = sysinit; sipp < sysinit_end; sipp++) {
221 
222 		if ((*sipp)->subsystem == SI_SUB_DUMMY)
223 			continue;	/* skip dummy task(s)*/
224 
225 		if ((*sipp)->subsystem == SI_SUB_DONE)
226 			continue;
227 
228 #if defined(VERBOSE_SYSINIT)
229 		if ((*sipp)->subsystem > last) {
230 			verbose = 1;
231 			last = (*sipp)->subsystem;
232 			printf("subsystem %x\n", last);
233 		}
234 		if (verbose) {
235 #if defined(DDB)
236 			const char *name;
237 			c_db_sym_t sym;
238 			db_expr_t  offset;
239 
240 			sym = db_search_symbol((vm_offset_t)(*sipp)->func,
241 			    DB_STGY_PROC, &offset);
242 			db_symbol_values(sym, &name, NULL);
243 			if (name != NULL)
244 				printf("   %s(%p)... ", name, (*sipp)->udata);
245 			else
246 #endif
247 				printf("   %p(%p)... ", (*sipp)->func,
248 				    (*sipp)->udata);
249 		}
250 #endif
251 
252 		/* Call function */
253 		(*((*sipp)->func))((*sipp)->udata);
254 
255 #if defined(VERBOSE_SYSINIT)
256 		if (verbose)
257 			printf("done.\n");
258 #endif
259 
260 		/* Check off the one we're just done */
261 		(*sipp)->subsystem = SI_SUB_DONE;
262 
263 		/* Check if we've installed more sysinit items via KLD */
264 		if (newsysinit != NULL) {
265 			if (sysinit != SET_BEGIN(sysinit_set))
266 				free(sysinit, M_TEMP);
267 			sysinit = newsysinit;
268 			sysinit_end = newsysinit_end;
269 			newsysinit = NULL;
270 			newsysinit_end = NULL;
271 			goto restart;
272 		}
273 	}
274 
275 	panic("Shouldn't get here!");
276 	/* NOTREACHED*/
277 }
278 
279 
280 /*
281  ***************************************************************************
282  ****
283  **** The following SYSINIT's belong elsewhere, but have not yet
284  **** been moved.
285  ****
286  ***************************************************************************
287  */
288 static void
289 print_caddr_t(void *data __unused)
290 {
291 	printf("%s", (char *)data);
292 }
293 SYSINIT(announce, SI_SUB_COPYRIGHT, SI_ORDER_FIRST, print_caddr_t,
294     copyright);
295 SYSINIT(trademark, SI_SUB_COPYRIGHT, SI_ORDER_SECOND, print_caddr_t,
296     trademark);
297 SYSINIT(version, SI_SUB_COPYRIGHT, SI_ORDER_THIRD, print_caddr_t, version);
298 
299 #ifdef WITNESS
300 static char wit_warn[] =
301      "WARNING: WITNESS option enabled, expect reduced performance.\n";
302 SYSINIT(witwarn, SI_SUB_COPYRIGHT, SI_ORDER_THIRD + 1,
303    print_caddr_t, wit_warn);
304 SYSINIT(witwarn2, SI_SUB_RUN_SCHEDULER, SI_ORDER_THIRD + 1,
305    print_caddr_t, wit_warn);
306 #endif
307 
308 #ifdef DIAGNOSTIC
309 static char diag_warn[] =
310      "WARNING: DIAGNOSTIC option enabled, expect reduced performance.\n";
311 SYSINIT(diagwarn, SI_SUB_COPYRIGHT, SI_ORDER_THIRD + 2,
312     print_caddr_t, diag_warn);
313 SYSINIT(diagwarn2, SI_SUB_RUN_SCHEDULER, SI_ORDER_THIRD + 2,
314     print_caddr_t, diag_warn);
315 #endif
316 
317 static void
318 set_boot_verbose(void *data __unused)
319 {
320 
321 	if (boothowto & RB_VERBOSE)
322 		bootverbose++;
323 }
324 SYSINIT(boot_verbose, SI_SUB_TUNABLES, SI_ORDER_ANY, set_boot_verbose, NULL);
325 
326 struct sysentvec null_sysvec = {
327 	.sv_size	= 0,
328 	.sv_table	= NULL,
329 	.sv_mask	= 0,
330 	.sv_sigsize	= 0,
331 	.sv_sigtbl	= NULL,
332 	.sv_errsize	= 0,
333 	.sv_errtbl	= NULL,
334 	.sv_transtrap	= NULL,
335 	.sv_fixup	= NULL,
336 	.sv_sendsig	= NULL,
337 	.sv_sigcode	= NULL,
338 	.sv_szsigcode	= NULL,
339 	.sv_prepsyscall	= NULL,
340 	.sv_name	= "null",
341 	.sv_coredump	= NULL,
342 	.sv_imgact_try	= NULL,
343 	.sv_minsigstksz	= 0,
344 	.sv_pagesize	= PAGE_SIZE,
345 	.sv_minuser	= VM_MIN_ADDRESS,
346 	.sv_maxuser	= VM_MAXUSER_ADDRESS,
347 	.sv_usrstack	= USRSTACK,
348 	.sv_psstrings	= PS_STRINGS,
349 	.sv_stackprot	= VM_PROT_ALL,
350 	.sv_copyout_strings	= NULL,
351 	.sv_setregs	= NULL,
352 	.sv_fixlimit	= NULL,
353 	.sv_maxssiz	= NULL
354 };
355 
356 /*
357  ***************************************************************************
358  ****
359  **** The two following SYSINIT's are proc0 specific glue code.  I am not
360  **** convinced that they can not be safely combined, but their order of
361  **** operation has been maintained as the same as the original init_main.c
362  **** for right now.
363  ****
364  **** These probably belong in init_proc.c or kern_proc.c, since they
365  **** deal with proc0 (the fork template process).
366  ****
367  ***************************************************************************
368  */
369 /* ARGSUSED*/
370 static void
371 proc0_init(void *dummy __unused)
372 {
373 	struct proc *p;
374 	unsigned i;
375 	struct thread *td;
376 
377 	GIANT_REQUIRED;
378 	p = &proc0;
379 	td = &thread0;
380 
381 	/*
382 	 * Initialize magic number and osrel.
383 	 */
384 	p->p_magic = P_MAGIC;
385 	p->p_osrel = osreldate;
386 
387 	/*
388 	 * Initialize thread and process structures.
389 	 */
390 	procinit();	/* set up proc zone */
391 	threadinit();	/* set up UMA zones */
392 
393 	/*
394 	 * Initialise scheduler resources.
395 	 * Add scheduler specific parts to proc, thread as needed.
396 	 */
397 	schedinit();	/* scheduler gets its house in order */
398 	/*
399 	 * Initialize sleep queue hash table
400 	 */
401 	sleepinit();
402 
403 	/*
404 	 * additional VM structures
405 	 */
406 	vm_init2();
407 
408 	/*
409 	 * Create process 0 (the swapper).
410 	 */
411 	LIST_INSERT_HEAD(&allproc, p, p_list);
412 	LIST_INSERT_HEAD(PIDHASH(0), p, p_hash);
413 	mtx_init(&pgrp0.pg_mtx, "process group", NULL, MTX_DEF | MTX_DUPOK);
414 	p->p_pgrp = &pgrp0;
415 	LIST_INSERT_HEAD(PGRPHASH(0), &pgrp0, pg_hash);
416 	LIST_INIT(&pgrp0.pg_members);
417 	LIST_INSERT_HEAD(&pgrp0.pg_members, p, p_pglist);
418 
419 	pgrp0.pg_session = &session0;
420 	mtx_init(&session0.s_mtx, "session", NULL, MTX_DEF);
421 	refcount_init(&session0.s_count, 1);
422 	session0.s_leader = p;
423 
424 	p->p_sysent = &null_sysvec;
425 	p->p_flag = P_SYSTEM | P_INMEM;
426 	p->p_state = PRS_NORMAL;
427 	knlist_init(&p->p_klist, &p->p_mtx, NULL, NULL, NULL);
428 	STAILQ_INIT(&p->p_ktr);
429 	p->p_nice = NZERO;
430 	td->td_tid = PID_MAX + 1;
431 	td->td_state = TDS_RUNNING;
432 	td->td_pri_class = PRI_TIMESHARE;
433 	td->td_user_pri = PUSER;
434 	td->td_base_user_pri = PUSER;
435 	td->td_priority = PVM;
436 	td->td_base_pri = PUSER;
437 	td->td_oncpu = 0;
438 	td->td_flags = TDF_INMEM|TDP_KTHREAD;
439 	td->td_cpuset = cpuset_thread0();
440 	prison0.pr_cpuset = cpuset_ref(td->td_cpuset);
441 	p->p_peers = 0;
442 	p->p_leader = p;
443 
444 
445 	strncpy(p->p_comm, "kernel", sizeof (p->p_comm));
446 	strncpy(td->td_name, "swapper", sizeof (td->td_name));
447 
448 	callout_init(&p->p_itcallout, CALLOUT_MPSAFE);
449 	callout_init_mtx(&p->p_limco, &p->p_mtx, 0);
450 	callout_init(&td->td_slpcallout, CALLOUT_MPSAFE);
451 
452 	/* Create credentials. */
453 	p->p_ucred = crget();
454 	p->p_ucred->cr_ngroups = 1;	/* group 0 */
455 	p->p_ucred->cr_uidinfo = uifind(0);
456 	p->p_ucred->cr_ruidinfo = uifind(0);
457 	p->p_ucred->cr_prison = &prison0;
458 #ifdef VIMAGE
459 	KASSERT(LIST_FIRST(&vimage_head) != NULL, ("vimage_head empty"));
460 	P_TO_VIMAGE(p) =  LIST_FIRST(&vimage_head); /* set ucred->cr_vimage */
461 	refcount_acquire(&P_TO_VIMAGE(p)->vi_ucredrefc);
462 	LIST_FIRST(&vprocg_head)->nprocs++;
463 #endif
464 #ifdef AUDIT
465 	audit_cred_kproc0(p->p_ucred);
466 #endif
467 #ifdef MAC
468 	mac_cred_create_swapper(p->p_ucred);
469 #endif
470 	td->td_ucred = crhold(p->p_ucred);
471 
472 	/* Create sigacts. */
473 	p->p_sigacts = sigacts_alloc();
474 
475 	/* Initialize signal state for process 0. */
476 	siginit(&proc0);
477 
478 	/* Create the file descriptor table. */
479 	p->p_fd = fdinit(NULL);
480 	p->p_fdtol = NULL;
481 
482 	/* Create the limits structures. */
483 	p->p_limit = lim_alloc();
484 	for (i = 0; i < RLIM_NLIMITS; i++)
485 		p->p_limit->pl_rlimit[i].rlim_cur =
486 		    p->p_limit->pl_rlimit[i].rlim_max = RLIM_INFINITY;
487 	p->p_limit->pl_rlimit[RLIMIT_NOFILE].rlim_cur =
488 	    p->p_limit->pl_rlimit[RLIMIT_NOFILE].rlim_max = maxfiles;
489 	p->p_limit->pl_rlimit[RLIMIT_NPROC].rlim_cur =
490 	    p->p_limit->pl_rlimit[RLIMIT_NPROC].rlim_max = maxproc;
491 	i = ptoa(cnt.v_free_count);
492 	p->p_limit->pl_rlimit[RLIMIT_RSS].rlim_max = i;
493 	p->p_limit->pl_rlimit[RLIMIT_MEMLOCK].rlim_max = i;
494 	p->p_limit->pl_rlimit[RLIMIT_MEMLOCK].rlim_cur = i / 3;
495 	p->p_cpulimit = RLIM_INFINITY;
496 
497 	p->p_stats = pstats_alloc();
498 
499 	/* Allocate a prototype map so we have something to fork. */
500 	pmap_pinit0(vmspace_pmap(&vmspace0));
501 	p->p_vmspace = &vmspace0;
502 	vmspace0.vm_refcnt = 1;
503 	vm_map_init(&vmspace0.vm_map, p->p_sysent->sv_minuser,
504 	    p->p_sysent->sv_maxuser);
505 	vmspace0.vm_map.pmap = vmspace_pmap(&vmspace0);
506 
507 	/*-
508 	 * call the init and ctor for the new thread and proc
509 	 * we wait to do this until all other structures
510 	 * are fairly sane.
511 	 */
512 	EVENTHANDLER_INVOKE(process_init, p);
513 	EVENTHANDLER_INVOKE(thread_init, td);
514 	EVENTHANDLER_INVOKE(process_ctor, p);
515 	EVENTHANDLER_INVOKE(thread_ctor, td);
516 
517 	/*
518 	 * Charge root for one process.
519 	 */
520 	(void)chgproccnt(p->p_ucred->cr_ruidinfo, 1, 0);
521 }
522 SYSINIT(p0init, SI_SUB_INTRINSIC, SI_ORDER_FIRST, proc0_init, NULL);
523 
524 /* ARGSUSED*/
525 static void
526 proc0_post(void *dummy __unused)
527 {
528 	struct timespec ts;
529 	struct proc *p;
530 	struct rusage ru;
531 	struct thread *td;
532 
533 	/*
534 	 * Now we can look at the time, having had a chance to verify the
535 	 * time from the filesystem.  Pretend that proc0 started now.
536 	 */
537 	sx_slock(&allproc_lock);
538 	FOREACH_PROC_IN_SYSTEM(p) {
539 		microuptime(&p->p_stats->p_start);
540 		PROC_SLOCK(p);
541 		rufetch(p, &ru);	/* Clears thread stats */
542 		PROC_SUNLOCK(p);
543 		p->p_rux.rux_runtime = 0;
544 		p->p_rux.rux_uticks = 0;
545 		p->p_rux.rux_sticks = 0;
546 		p->p_rux.rux_iticks = 0;
547 		FOREACH_THREAD_IN_PROC(p, td) {
548 			td->td_runtime = 0;
549 		}
550 	}
551 	sx_sunlock(&allproc_lock);
552 	PCPU_SET(switchtime, cpu_ticks());
553 	PCPU_SET(switchticks, ticks);
554 
555 	/*
556 	 * Give the ``random'' number generator a thump.
557 	 */
558 	nanotime(&ts);
559 	srandom(ts.tv_sec ^ ts.tv_nsec);
560 }
561 SYSINIT(p0post, SI_SUB_INTRINSIC_POST, SI_ORDER_FIRST, proc0_post, NULL);
562 
563 /*
564  ***************************************************************************
565  ****
566  **** The following SYSINIT's and glue code should be moved to the
567  **** respective files on a per subsystem basis.
568  ****
569  ***************************************************************************
570  */
571 
572 
573 /*
574  ***************************************************************************
575  ****
576  **** The following code probably belongs in another file, like
577  **** kern/init_init.c.
578  ****
579  ***************************************************************************
580  */
581 
582 /*
583  * List of paths to try when searching for "init".
584  */
585 static char init_path[MAXPATHLEN] =
586 #ifdef	INIT_PATH
587     __XSTRING(INIT_PATH);
588 #else
589     "/sbin/init:/sbin/oinit:/sbin/init.bak:/rescue/init:/stand/sysinstall";
590 #endif
591 SYSCTL_STRING(_kern, OID_AUTO, init_path, CTLFLAG_RD, init_path, 0,
592 	"Path used to search the init process");
593 
594 /*
595  * Shutdown timeout of init(8).
596  * Unused within kernel, but used to control init(8), hence do not remove.
597  */
598 #ifndef INIT_SHUTDOWN_TIMEOUT
599 #define INIT_SHUTDOWN_TIMEOUT 120
600 #endif
601 static int init_shutdown_timeout = INIT_SHUTDOWN_TIMEOUT;
602 SYSCTL_INT(_kern, OID_AUTO, init_shutdown_timeout,
603 	CTLFLAG_RW, &init_shutdown_timeout, 0, "");
604 
605 /*
606  * Start the initial user process; try exec'ing each pathname in init_path.
607  * The program is invoked with one argument containing the boot flags.
608  */
609 static void
610 start_init(void *dummy)
611 {
612 	vm_offset_t addr;
613 	struct execve_args args;
614 	int options, error;
615 	char *var, *path, *next, *s;
616 	char *ucp, **uap, *arg0, *arg1;
617 	struct thread *td;
618 	struct proc *p;
619 
620 	mtx_lock(&Giant);
621 
622 	GIANT_REQUIRED;
623 
624 	td = curthread;
625 	p = td->td_proc;
626 
627 	vfs_mountroot();
628 
629 	/*
630 	 * Need just enough stack to hold the faked-up "execve()" arguments.
631 	 */
632 	addr = p->p_sysent->sv_usrstack - PAGE_SIZE;
633 	if (vm_map_find(&p->p_vmspace->vm_map, NULL, 0, &addr, PAGE_SIZE,
634 			FALSE, VM_PROT_ALL, VM_PROT_ALL, 0) != 0)
635 		panic("init: couldn't allocate argument space");
636 	p->p_vmspace->vm_maxsaddr = (caddr_t)addr;
637 	p->p_vmspace->vm_ssize = 1;
638 
639 	if ((var = getenv("init_path")) != NULL) {
640 		strlcpy(init_path, var, sizeof(init_path));
641 		freeenv(var);
642 	}
643 
644 	for (path = init_path; *path != '\0'; path = next) {
645 		while (*path == ':')
646 			path++;
647 		if (*path == '\0')
648 			break;
649 		for (next = path; *next != '\0' && *next != ':'; next++)
650 			/* nothing */ ;
651 		if (bootverbose)
652 			printf("start_init: trying %.*s\n", (int)(next - path),
653 			    path);
654 
655 		/*
656 		 * Move out the boot flag argument.
657 		 */
658 		options = 0;
659 		ucp = (char *)p->p_sysent->sv_usrstack;
660 		(void)subyte(--ucp, 0);		/* trailing zero */
661 		if (boothowto & RB_SINGLE) {
662 			(void)subyte(--ucp, 's');
663 			options = 1;
664 		}
665 #ifdef notyet
666                 if (boothowto & RB_FASTBOOT) {
667 			(void)subyte(--ucp, 'f');
668 			options = 1;
669 		}
670 #endif
671 
672 #ifdef BOOTCDROM
673 		(void)subyte(--ucp, 'C');
674 		options = 1;
675 #endif
676 
677 		if (options == 0)
678 			(void)subyte(--ucp, '-');
679 		(void)subyte(--ucp, '-');		/* leading hyphen */
680 		arg1 = ucp;
681 
682 		/*
683 		 * Move out the file name (also arg 0).
684 		 */
685 		(void)subyte(--ucp, 0);
686 		for (s = next - 1; s >= path; s--)
687 			(void)subyte(--ucp, *s);
688 		arg0 = ucp;
689 
690 		/*
691 		 * Move out the arg pointers.
692 		 */
693 		uap = (char **)((intptr_t)ucp & ~(sizeof(intptr_t)-1));
694 		(void)suword((caddr_t)--uap, (long)0);	/* terminator */
695 		(void)suword((caddr_t)--uap, (long)(intptr_t)arg1);
696 		(void)suword((caddr_t)--uap, (long)(intptr_t)arg0);
697 
698 		/*
699 		 * Point at the arguments.
700 		 */
701 		args.fname = arg0;
702 		args.argv = uap;
703 		args.envv = NULL;
704 
705 		/*
706 		 * Now try to exec the program.  If can't for any reason
707 		 * other than it doesn't exist, complain.
708 		 *
709 		 * Otherwise, return via fork_trampoline() all the way
710 		 * to user mode as init!
711 		 */
712 		if ((error = execve(td, &args)) == 0) {
713 			mtx_unlock(&Giant);
714 			return;
715 		}
716 		if (error != ENOENT)
717 			printf("exec %.*s: error %d\n", (int)(next - path),
718 			    path, error);
719 	}
720 	printf("init: not found in path %s\n", init_path);
721 	panic("no init");
722 }
723 
724 /*
725  * Like kproc_create(), but runs in it's own address space.
726  * We do this early to reserve pid 1.
727  *
728  * Note special case - do not make it runnable yet.  Other work
729  * in progress will change this more.
730  */
731 static void
732 create_init(const void *udata __unused)
733 {
734 	struct ucred *newcred, *oldcred;
735 	int error;
736 
737 	error = fork1(&thread0, RFFDG | RFPROC | RFSTOPPED, 0, &initproc);
738 	if (error)
739 		panic("cannot fork init: %d\n", error);
740 	KASSERT(initproc->p_pid == 1, ("create_init: initproc->p_pid != 1"));
741 	/* divorce init's credentials from the kernel's */
742 	newcred = crget();
743 	PROC_LOCK(initproc);
744 	initproc->p_flag |= P_SYSTEM | P_INMEM;
745 	oldcred = initproc->p_ucred;
746 	crcopy(newcred, oldcred);
747 #ifdef MAC
748 	mac_cred_create_init(newcred);
749 #endif
750 #ifdef AUDIT
751 	audit_cred_proc1(newcred);
752 #endif
753 	initproc->p_ucred = newcred;
754 	PROC_UNLOCK(initproc);
755 	crfree(oldcred);
756 	cred_update_thread(FIRST_THREAD_IN_PROC(initproc));
757 	cpu_set_fork_handler(FIRST_THREAD_IN_PROC(initproc), start_init, NULL);
758 }
759 SYSINIT(init, SI_SUB_CREATE_INIT, SI_ORDER_FIRST, create_init, NULL);
760 
761 /*
762  * Make it runnable now.
763  */
764 static void
765 kick_init(const void *udata __unused)
766 {
767 	struct thread *td;
768 
769 	td = FIRST_THREAD_IN_PROC(initproc);
770 	thread_lock(td);
771 	TD_SET_CAN_RUN(td);
772 	sched_add(td, SRQ_BORING);
773 	thread_unlock(td);
774 }
775 SYSINIT(kickinit, SI_SUB_KTHREAD_INIT, SI_ORDER_FIRST, kick_init, NULL);
776