xref: /freebsd/sys/kern/init_main.c (revision dadef94c7a762d05890e2891bc4a7d1dfe0cf758)
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 
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 
92 void mi_startup(void);				/* Should be elsewhere */
93 
94 /* Components of the first process -- never freed. */
95 static struct session session0;
96 static struct pgrp pgrp0;
97 struct	proc proc0;
98 struct	thread thread0 __aligned(16);
99 struct	vmspace vmspace0;
100 struct	proc *initproc;
101 
102 int	boothowto = 0;		/* initialized so that it can be patched */
103 SYSCTL_INT(_debug, OID_AUTO, boothowto, CTLFLAG_RD, &boothowto, 0,
104 	"Boot control flags, passed from loader");
105 int	bootverbose;
106 SYSCTL_INT(_debug, OID_AUTO, bootverbose, CTLFLAG_RW, &bootverbose, 0,
107 	"Control the output of verbose kernel messages");
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)
290 {
291 	printf("%s", (char *)data);
292 }
293 
294 static void
295 print_version(void *data __unused)
296 {
297 	int len;
298 
299 	/* Strip a trailing newline from version. */
300 	len = strlen(version);
301 	while (len > 0 && version[len - 1] == '\n')
302 		len--;
303 	printf("%.*s %s\n", len, version, machine);
304 }
305 
306 SYSINIT(announce, SI_SUB_COPYRIGHT, SI_ORDER_FIRST, print_caddr_t,
307     copyright);
308 SYSINIT(trademark, SI_SUB_COPYRIGHT, SI_ORDER_SECOND, print_caddr_t,
309     trademark);
310 SYSINIT(version, SI_SUB_COPYRIGHT, SI_ORDER_THIRD, print_version, NULL);
311 
312 #ifdef WITNESS
313 static char wit_warn[] =
314      "WARNING: WITNESS option enabled, expect reduced performance.\n";
315 SYSINIT(witwarn, SI_SUB_COPYRIGHT, SI_ORDER_THIRD + 1,
316    print_caddr_t, wit_warn);
317 SYSINIT(witwarn2, SI_SUB_RUN_SCHEDULER, SI_ORDER_THIRD + 1,
318    print_caddr_t, wit_warn);
319 #endif
320 
321 #ifdef DIAGNOSTIC
322 static char diag_warn[] =
323      "WARNING: DIAGNOSTIC option enabled, expect reduced performance.\n";
324 SYSINIT(diagwarn, SI_SUB_COPYRIGHT, SI_ORDER_THIRD + 2,
325     print_caddr_t, diag_warn);
326 SYSINIT(diagwarn2, SI_SUB_RUN_SCHEDULER, SI_ORDER_THIRD + 2,
327     print_caddr_t, diag_warn);
328 #endif
329 
330 static void
331 set_boot_verbose(void *data __unused)
332 {
333 
334 	if (boothowto & RB_VERBOSE)
335 		bootverbose++;
336 }
337 SYSINIT(boot_verbose, SI_SUB_TUNABLES, SI_ORDER_ANY, set_boot_verbose, NULL);
338 
339 static int
340 null_fetch_syscall_args(struct thread *td __unused,
341     struct syscall_args *sa __unused)
342 {
343 
344 	panic("null_fetch_syscall_args");
345 }
346 
347 static void
348 null_set_syscall_retval(struct thread *td __unused, int error __unused)
349 {
350 
351 	panic("null_set_syscall_retval");
352 }
353 
354 struct sysentvec null_sysvec = {
355 	.sv_size	= 0,
356 	.sv_table	= NULL,
357 	.sv_mask	= 0,
358 	.sv_sigsize	= 0,
359 	.sv_sigtbl	= NULL,
360 	.sv_errsize	= 0,
361 	.sv_errtbl	= NULL,
362 	.sv_transtrap	= NULL,
363 	.sv_fixup	= NULL,
364 	.sv_sendsig	= NULL,
365 	.sv_sigcode	= NULL,
366 	.sv_szsigcode	= NULL,
367 	.sv_prepsyscall	= NULL,
368 	.sv_name	= "null",
369 	.sv_coredump	= NULL,
370 	.sv_imgact_try	= NULL,
371 	.sv_minsigstksz	= 0,
372 	.sv_pagesize	= PAGE_SIZE,
373 	.sv_minuser	= VM_MIN_ADDRESS,
374 	.sv_maxuser	= VM_MAXUSER_ADDRESS,
375 	.sv_usrstack	= USRSTACK,
376 	.sv_psstrings	= PS_STRINGS,
377 	.sv_stackprot	= VM_PROT_ALL,
378 	.sv_copyout_strings	= NULL,
379 	.sv_setregs	= NULL,
380 	.sv_fixlimit	= NULL,
381 	.sv_maxssiz	= NULL,
382 	.sv_flags	= 0,
383 	.sv_set_syscall_retval = null_set_syscall_retval,
384 	.sv_fetch_syscall_args = null_fetch_syscall_args,
385 	.sv_syscallnames = NULL,
386 };
387 
388 /*
389  ***************************************************************************
390  ****
391  **** The two following SYSINIT's are proc0 specific glue code.  I am not
392  **** convinced that they can not be safely combined, but their order of
393  **** operation has been maintained as the same as the original init_main.c
394  **** for right now.
395  ****
396  **** These probably belong in init_proc.c or kern_proc.c, since they
397  **** deal with proc0 (the fork template process).
398  ****
399  ***************************************************************************
400  */
401 /* ARGSUSED*/
402 static void
403 proc0_init(void *dummy __unused)
404 {
405 	struct proc *p;
406 	struct thread *td;
407 	vm_paddr_t pageablemem;
408 	int i;
409 
410 	GIANT_REQUIRED;
411 	p = &proc0;
412 	td = &thread0;
413 
414 	/*
415 	 * Initialize magic number and osrel.
416 	 */
417 	p->p_magic = P_MAGIC;
418 	p->p_osrel = osreldate;
419 
420 	/*
421 	 * Initialize thread and process structures.
422 	 */
423 	procinit();	/* set up proc zone */
424 	threadinit();	/* set up UMA zones */
425 
426 	/*
427 	 * Initialise scheduler resources.
428 	 * Add scheduler specific parts to proc, thread as needed.
429 	 */
430 	schedinit();	/* scheduler gets its house in order */
431 	/*
432 	 * Initialize sleep queue hash table
433 	 */
434 	sleepinit();
435 
436 	/*
437 	 * additional VM structures
438 	 */
439 	vm_init2();
440 
441 	/*
442 	 * Create process 0 (the swapper).
443 	 */
444 	LIST_INSERT_HEAD(&allproc, p, p_list);
445 	LIST_INSERT_HEAD(PIDHASH(0), p, p_hash);
446 	mtx_init(&pgrp0.pg_mtx, "process group", NULL, MTX_DEF | MTX_DUPOK);
447 	p->p_pgrp = &pgrp0;
448 	LIST_INSERT_HEAD(PGRPHASH(0), &pgrp0, pg_hash);
449 	LIST_INIT(&pgrp0.pg_members);
450 	LIST_INSERT_HEAD(&pgrp0.pg_members, p, p_pglist);
451 
452 	pgrp0.pg_session = &session0;
453 	mtx_init(&session0.s_mtx, "session", NULL, MTX_DEF);
454 	refcount_init(&session0.s_count, 1);
455 	session0.s_leader = p;
456 
457 	p->p_sysent = &null_sysvec;
458 	p->p_flag = P_SYSTEM | P_INMEM;
459 	p->p_state = PRS_NORMAL;
460 	knlist_init_mtx(&p->p_klist, &p->p_mtx);
461 	STAILQ_INIT(&p->p_ktr);
462 	p->p_nice = NZERO;
463 	td->td_tid = PID_MAX + 1;
464 	LIST_INSERT_HEAD(TIDHASH(td->td_tid), td, td_hash);
465 	td->td_state = TDS_RUNNING;
466 	td->td_pri_class = PRI_TIMESHARE;
467 	td->td_user_pri = PUSER;
468 	td->td_base_user_pri = PUSER;
469 	td->td_priority = PVM;
470 	td->td_base_pri = PUSER;
471 	td->td_oncpu = 0;
472 	td->td_flags = TDF_INMEM|TDP_KTHREAD;
473 	td->td_cpuset = cpuset_thread0();
474 	prison0.pr_cpuset = cpuset_ref(td->td_cpuset);
475 	p->p_peers = 0;
476 	p->p_leader = p;
477 
478 
479 	strncpy(p->p_comm, "kernel", sizeof (p->p_comm));
480 	strncpy(td->td_name, "swapper", sizeof (td->td_name));
481 
482 	callout_init(&p->p_itcallout, CALLOUT_MPSAFE);
483 	callout_init_mtx(&p->p_limco, &p->p_mtx, 0);
484 	callout_init(&td->td_slpcallout, CALLOUT_MPSAFE);
485 
486 	/* Create credentials. */
487 	p->p_ucred = crget();
488 	p->p_ucred->cr_ngroups = 1;	/* group 0 */
489 	p->p_ucred->cr_uidinfo = uifind(0);
490 	p->p_ucred->cr_ruidinfo = uifind(0);
491 	p->p_ucred->cr_prison = &prison0;
492 #ifdef AUDIT
493 	audit_cred_kproc0(p->p_ucred);
494 #endif
495 #ifdef MAC
496 	mac_cred_create_swapper(p->p_ucred);
497 #endif
498 	td->td_ucred = crhold(p->p_ucred);
499 
500 	/* Create sigacts. */
501 	p->p_sigacts = sigacts_alloc();
502 
503 	/* Initialize signal state for process 0. */
504 	siginit(&proc0);
505 
506 	/* Create the file descriptor table. */
507 	p->p_fd = fdinit(NULL);
508 	p->p_fdtol = NULL;
509 
510 	/* Create the limits structures. */
511 	p->p_limit = lim_alloc();
512 	for (i = 0; i < RLIM_NLIMITS; i++)
513 		p->p_limit->pl_rlimit[i].rlim_cur =
514 		    p->p_limit->pl_rlimit[i].rlim_max = RLIM_INFINITY;
515 	p->p_limit->pl_rlimit[RLIMIT_NOFILE].rlim_cur =
516 	    p->p_limit->pl_rlimit[RLIMIT_NOFILE].rlim_max = maxfiles;
517 	p->p_limit->pl_rlimit[RLIMIT_NPROC].rlim_cur =
518 	    p->p_limit->pl_rlimit[RLIMIT_NPROC].rlim_max = maxproc;
519 	p->p_limit->pl_rlimit[RLIMIT_DATA].rlim_cur = dfldsiz;
520 	p->p_limit->pl_rlimit[RLIMIT_DATA].rlim_max = maxdsiz;
521 	p->p_limit->pl_rlimit[RLIMIT_STACK].rlim_cur = dflssiz;
522 	p->p_limit->pl_rlimit[RLIMIT_STACK].rlim_max = maxssiz;
523 	/* Cast to avoid overflow on i386/PAE. */
524 	pageablemem = ptoa((vm_paddr_t)cnt.v_free_count);
525 	p->p_limit->pl_rlimit[RLIMIT_RSS].rlim_cur =
526 	    p->p_limit->pl_rlimit[RLIMIT_RSS].rlim_max = pageablemem;
527 	p->p_limit->pl_rlimit[RLIMIT_MEMLOCK].rlim_cur = pageablemem / 3;
528 	p->p_limit->pl_rlimit[RLIMIT_MEMLOCK].rlim_max = pageablemem;
529 	p->p_cpulimit = RLIM_INFINITY;
530 
531 	p->p_stats = pstats_alloc();
532 
533 	/* Allocate a prototype map so we have something to fork. */
534 	pmap_pinit0(vmspace_pmap(&vmspace0));
535 	p->p_vmspace = &vmspace0;
536 	vmspace0.vm_refcnt = 1;
537 
538 	/*
539 	 * proc0 is not expected to enter usermode, so there is no special
540 	 * handling for sv_minuser here, like is done for exec_new_vmspace().
541 	 */
542 	vm_map_init(&vmspace0.vm_map, vmspace_pmap(&vmspace0),
543 	    p->p_sysent->sv_minuser, p->p_sysent->sv_maxuser);
544 
545 	/*
546 	 * Call the init and ctor for the new thread and proc.  We wait
547 	 * to do this until all other structures are fairly sane.
548 	 */
549 	EVENTHANDLER_INVOKE(process_init, p);
550 	EVENTHANDLER_INVOKE(thread_init, td);
551 	EVENTHANDLER_INVOKE(process_ctor, p);
552 	EVENTHANDLER_INVOKE(thread_ctor, td);
553 
554 	/*
555 	 * Charge root for one process.
556 	 */
557 	(void)chgproccnt(p->p_ucred->cr_ruidinfo, 1, 0);
558 }
559 SYSINIT(p0init, SI_SUB_INTRINSIC, SI_ORDER_FIRST, proc0_init, NULL);
560 
561 /* ARGSUSED*/
562 static void
563 proc0_post(void *dummy __unused)
564 {
565 	struct timespec ts;
566 	struct proc *p;
567 	struct rusage ru;
568 	struct thread *td;
569 
570 	/*
571 	 * Now we can look at the time, having had a chance to verify the
572 	 * time from the filesystem.  Pretend that proc0 started now.
573 	 */
574 	sx_slock(&allproc_lock);
575 	FOREACH_PROC_IN_SYSTEM(p) {
576 		microuptime(&p->p_stats->p_start);
577 		PROC_SLOCK(p);
578 		rufetch(p, &ru);	/* Clears thread stats */
579 		PROC_SUNLOCK(p);
580 		p->p_rux.rux_runtime = 0;
581 		p->p_rux.rux_uticks = 0;
582 		p->p_rux.rux_sticks = 0;
583 		p->p_rux.rux_iticks = 0;
584 		FOREACH_THREAD_IN_PROC(p, td) {
585 			td->td_runtime = 0;
586 		}
587 	}
588 	sx_sunlock(&allproc_lock);
589 	PCPU_SET(switchtime, cpu_ticks());
590 	PCPU_SET(switchticks, ticks);
591 
592 	/*
593 	 * Give the ``random'' number generator a thump.
594 	 */
595 	nanotime(&ts);
596 	srandom(ts.tv_sec ^ ts.tv_nsec);
597 }
598 SYSINIT(p0post, SI_SUB_INTRINSIC_POST, SI_ORDER_FIRST, proc0_post, NULL);
599 
600 static void
601 random_init(void *dummy __unused)
602 {
603 
604 	/*
605 	 * After CPU has been started we have some randomness on most
606 	 * platforms via get_cyclecount().  For platforms that don't
607 	 * we will reseed random(9) in proc0_post() as well.
608 	 */
609 	srandom(get_cyclecount());
610 }
611 SYSINIT(random, SI_SUB_RANDOM, SI_ORDER_FIRST, random_init, NULL);
612 
613 /*
614  ***************************************************************************
615  ****
616  **** The following SYSINIT's and glue code should be moved to the
617  **** respective files on a per subsystem basis.
618  ****
619  ***************************************************************************
620  */
621 
622 
623 /*
624  ***************************************************************************
625  ****
626  **** The following code probably belongs in another file, like
627  **** kern/init_init.c.
628  ****
629  ***************************************************************************
630  */
631 
632 /*
633  * List of paths to try when searching for "init".
634  */
635 static char init_path[MAXPATHLEN] =
636 #ifdef	INIT_PATH
637     __XSTRING(INIT_PATH);
638 #else
639     "/sbin/init:/sbin/oinit:/sbin/init.bak:/rescue/init:/stand/sysinstall";
640 #endif
641 SYSCTL_STRING(_kern, OID_AUTO, init_path, CTLFLAG_RD, init_path, 0,
642 	"Path used to search the init process");
643 
644 /*
645  * Shutdown timeout of init(8).
646  * Unused within kernel, but used to control init(8), hence do not remove.
647  */
648 #ifndef INIT_SHUTDOWN_TIMEOUT
649 #define INIT_SHUTDOWN_TIMEOUT 120
650 #endif
651 static int init_shutdown_timeout = INIT_SHUTDOWN_TIMEOUT;
652 SYSCTL_INT(_kern, OID_AUTO, init_shutdown_timeout,
653 	CTLFLAG_RW, &init_shutdown_timeout, 0, "Shutdown timeout of init(8). "
654 	"Unused within kernel, but used to control init(8)");
655 
656 /*
657  * Start the initial user process; try exec'ing each pathname in init_path.
658  * The program is invoked with one argument containing the boot flags.
659  */
660 static void
661 start_init(void *dummy)
662 {
663 	vm_offset_t addr;
664 	struct execve_args args;
665 	int options, error;
666 	char *var, *path, *next, *s;
667 	char *ucp, **uap, *arg0, *arg1;
668 	struct thread *td;
669 	struct proc *p;
670 
671 	mtx_lock(&Giant);
672 
673 	GIANT_REQUIRED;
674 
675 	td = curthread;
676 	p = td->td_proc;
677 
678 	vfs_mountroot();
679 
680 	/*
681 	 * Need just enough stack to hold the faked-up "execve()" arguments.
682 	 */
683 	addr = p->p_sysent->sv_usrstack - PAGE_SIZE;
684 	if (vm_map_find(&p->p_vmspace->vm_map, NULL, 0, &addr, PAGE_SIZE,
685 			FALSE, VM_PROT_ALL, VM_PROT_ALL, 0) != 0)
686 		panic("init: couldn't allocate argument space");
687 	p->p_vmspace->vm_maxsaddr = (caddr_t)addr;
688 	p->p_vmspace->vm_ssize = 1;
689 
690 	if ((var = getenv("init_path")) != NULL) {
691 		strlcpy(init_path, var, sizeof(init_path));
692 		freeenv(var);
693 	}
694 
695 	for (path = init_path; *path != '\0'; path = next) {
696 		while (*path == ':')
697 			path++;
698 		if (*path == '\0')
699 			break;
700 		for (next = path; *next != '\0' && *next != ':'; next++)
701 			/* nothing */ ;
702 		if (bootverbose)
703 			printf("start_init: trying %.*s\n", (int)(next - path),
704 			    path);
705 
706 		/*
707 		 * Move out the boot flag argument.
708 		 */
709 		options = 0;
710 		ucp = (char *)p->p_sysent->sv_usrstack;
711 		(void)subyte(--ucp, 0);		/* trailing zero */
712 		if (boothowto & RB_SINGLE) {
713 			(void)subyte(--ucp, 's');
714 			options = 1;
715 		}
716 #ifdef notyet
717                 if (boothowto & RB_FASTBOOT) {
718 			(void)subyte(--ucp, 'f');
719 			options = 1;
720 		}
721 #endif
722 
723 #ifdef BOOTCDROM
724 		(void)subyte(--ucp, 'C');
725 		options = 1;
726 #endif
727 
728 		if (options == 0)
729 			(void)subyte(--ucp, '-');
730 		(void)subyte(--ucp, '-');		/* leading hyphen */
731 		arg1 = ucp;
732 
733 		/*
734 		 * Move out the file name (also arg 0).
735 		 */
736 		(void)subyte(--ucp, 0);
737 		for (s = next - 1; s >= path; s--)
738 			(void)subyte(--ucp, *s);
739 		arg0 = ucp;
740 
741 		/*
742 		 * Move out the arg pointers.
743 		 */
744 		uap = (char **)((intptr_t)ucp & ~(sizeof(intptr_t)-1));
745 		(void)suword((caddr_t)--uap, (long)0);	/* terminator */
746 		(void)suword((caddr_t)--uap, (long)(intptr_t)arg1);
747 		(void)suword((caddr_t)--uap, (long)(intptr_t)arg0);
748 
749 		/*
750 		 * Point at the arguments.
751 		 */
752 		args.fname = arg0;
753 		args.argv = uap;
754 		args.envv = NULL;
755 
756 		/*
757 		 * Now try to exec the program.  If can't for any reason
758 		 * other than it doesn't exist, complain.
759 		 *
760 		 * Otherwise, return via fork_trampoline() all the way
761 		 * to user mode as init!
762 		 */
763 		if ((error = execve(td, &args)) == 0) {
764 			mtx_unlock(&Giant);
765 			return;
766 		}
767 		if (error != ENOENT)
768 			printf("exec %.*s: error %d\n", (int)(next - path),
769 			    path, error);
770 	}
771 	printf("init: not found in path %s\n", init_path);
772 	panic("no init");
773 }
774 
775 /*
776  * Like kproc_create(), but runs in it's own address space.
777  * We do this early to reserve pid 1.
778  *
779  * Note special case - do not make it runnable yet.  Other work
780  * in progress will change this more.
781  */
782 static void
783 create_init(const void *udata __unused)
784 {
785 	struct ucred *newcred, *oldcred;
786 	int error;
787 
788 	error = fork1(&thread0, RFFDG | RFPROC | RFSTOPPED, 0, &initproc);
789 	if (error)
790 		panic("cannot fork init: %d\n", error);
791 	KASSERT(initproc->p_pid == 1, ("create_init: initproc->p_pid != 1"));
792 	/* divorce init's credentials from the kernel's */
793 	newcred = crget();
794 	PROC_LOCK(initproc);
795 	initproc->p_flag |= P_SYSTEM | P_INMEM;
796 	oldcred = initproc->p_ucred;
797 	crcopy(newcred, oldcred);
798 #ifdef MAC
799 	mac_cred_create_init(newcred);
800 #endif
801 #ifdef AUDIT
802 	audit_cred_proc1(newcred);
803 #endif
804 	initproc->p_ucred = newcred;
805 	PROC_UNLOCK(initproc);
806 	crfree(oldcred);
807 	cred_update_thread(FIRST_THREAD_IN_PROC(initproc));
808 	cpu_set_fork_handler(FIRST_THREAD_IN_PROC(initproc), start_init, NULL);
809 }
810 SYSINIT(init, SI_SUB_CREATE_INIT, SI_ORDER_FIRST, create_init, NULL);
811 
812 /*
813  * Make it runnable now.
814  */
815 static void
816 kick_init(const void *udata __unused)
817 {
818 	struct thread *td;
819 
820 	td = FIRST_THREAD_IN_PROC(initproc);
821 	thread_lock(td);
822 	TD_SET_CAN_RUN(td);
823 	sched_add(td, SRQ_BORING);
824 	thread_unlock(td);
825 }
826 SYSINIT(kickinit, SI_SUB_KTHREAD_INIT, SI_ORDER_FIRST, kick_init, NULL);
827