xref: /freebsd/sys/kern/init_main.c (revision fb2971ccd2bab42f5406f919812ceb8e014773ea)
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 (boothowto & RB_VERBOSE)
184 		bootverbose++;
185 
186 	if (sysinit == NULL) {
187 		sysinit = SET_BEGIN(sysinit_set);
188 		sysinit_end = SET_LIMIT(sysinit_set);
189 	}
190 
191 restart:
192 	/*
193 	 * Perform a bubble sort of the system initialization objects by
194 	 * their subsystem (primary key) and order (secondary key).
195 	 */
196 	for (sipp = sysinit; sipp < sysinit_end; sipp++) {
197 		for (xipp = sipp + 1; xipp < sysinit_end; xipp++) {
198 			if ((*sipp)->subsystem < (*xipp)->subsystem ||
199 			     ((*sipp)->subsystem == (*xipp)->subsystem &&
200 			      (*sipp)->order <= (*xipp)->order))
201 				continue;	/* skip*/
202 			save = *sipp;
203 			*sipp = *xipp;
204 			*xipp = save;
205 		}
206 	}
207 
208 #if defined(VERBOSE_SYSINIT)
209 	last = SI_SUB_COPYRIGHT;
210 	verbose = 0;
211 #if !defined(DDB)
212 	printf("VERBOSE_SYSINIT: DDB not enabled, symbol lookups disabled.\n");
213 #endif
214 #endif
215 
216 	/*
217 	 * Traverse the (now) ordered list of system initialization tasks.
218 	 * Perform each task, and continue on to the next task.
219 	 *
220 	 * The last item on the list is expected to be the scheduler,
221 	 * which will not return.
222 	 */
223 	for (sipp = sysinit; sipp < sysinit_end; sipp++) {
224 
225 		if ((*sipp)->subsystem == SI_SUB_DUMMY)
226 			continue;	/* skip dummy task(s)*/
227 
228 		if ((*sipp)->subsystem == SI_SUB_DONE)
229 			continue;
230 
231 #if defined(VERBOSE_SYSINIT)
232 		if ((*sipp)->subsystem > last) {
233 			verbose = 1;
234 			last = (*sipp)->subsystem;
235 			printf("subsystem %x\n", last);
236 		}
237 		if (verbose) {
238 #if defined(DDB)
239 			const char *name;
240 			c_db_sym_t sym;
241 			db_expr_t  offset;
242 
243 			sym = db_search_symbol((vm_offset_t)(*sipp)->func,
244 			    DB_STGY_PROC, &offset);
245 			db_symbol_values(sym, &name, NULL);
246 			if (name != NULL)
247 				printf("   %s(%p)... ", name, (*sipp)->udata);
248 			else
249 #endif
250 				printf("   %p(%p)... ", (*sipp)->func,
251 				    (*sipp)->udata);
252 		}
253 #endif
254 
255 		/* Call function */
256 		(*((*sipp)->func))((*sipp)->udata);
257 
258 #if defined(VERBOSE_SYSINIT)
259 		if (verbose)
260 			printf("done.\n");
261 #endif
262 
263 		/* Check off the one we're just done */
264 		(*sipp)->subsystem = SI_SUB_DONE;
265 
266 		/* Check if we've installed more sysinit items via KLD */
267 		if (newsysinit != NULL) {
268 			if (sysinit != SET_BEGIN(sysinit_set))
269 				free(sysinit, M_TEMP);
270 			sysinit = newsysinit;
271 			sysinit_end = newsysinit_end;
272 			newsysinit = NULL;
273 			newsysinit_end = NULL;
274 			goto restart;
275 		}
276 	}
277 
278 	panic("Shouldn't get here!");
279 	/* NOTREACHED*/
280 }
281 
282 
283 /*
284  ***************************************************************************
285  ****
286  **** The following SYSINIT's belong elsewhere, but have not yet
287  **** been moved.
288  ****
289  ***************************************************************************
290  */
291 static void
292 print_caddr_t(void *data)
293 {
294 	printf("%s", (char *)data);
295 }
296 
297 static void
298 print_version(void *data __unused)
299 {
300 	int len;
301 
302 	/* Strip a trailing newline from version. */
303 	len = strlen(version);
304 	while (len > 0 && version[len - 1] == '\n')
305 		len--;
306 	printf("%.*s %s\n", len, version, machine);
307 }
308 
309 SYSINIT(announce, SI_SUB_COPYRIGHT, SI_ORDER_FIRST, print_caddr_t,
310     copyright);
311 SYSINIT(trademark, SI_SUB_COPYRIGHT, SI_ORDER_SECOND, print_caddr_t,
312     trademark);
313 SYSINIT(version, SI_SUB_COPYRIGHT, SI_ORDER_THIRD, print_version, NULL);
314 
315 #ifdef WITNESS
316 static char wit_warn[] =
317      "WARNING: WITNESS option enabled, expect reduced performance.\n";
318 SYSINIT(witwarn, SI_SUB_COPYRIGHT, SI_ORDER_THIRD + 1,
319    print_caddr_t, wit_warn);
320 SYSINIT(witwarn2, SI_SUB_RUN_SCHEDULER, SI_ORDER_THIRD + 1,
321    print_caddr_t, wit_warn);
322 #endif
323 
324 #ifdef DIAGNOSTIC
325 static char diag_warn[] =
326      "WARNING: DIAGNOSTIC option enabled, expect reduced performance.\n";
327 SYSINIT(diagwarn, SI_SUB_COPYRIGHT, SI_ORDER_THIRD + 2,
328     print_caddr_t, diag_warn);
329 SYSINIT(diagwarn2, SI_SUB_RUN_SCHEDULER, SI_ORDER_THIRD + 2,
330     print_caddr_t, diag_warn);
331 #endif
332 
333 static int
334 null_fetch_syscall_args(struct thread *td __unused,
335     struct syscall_args *sa __unused)
336 {
337 
338 	panic("null_fetch_syscall_args");
339 }
340 
341 static void
342 null_set_syscall_retval(struct thread *td __unused, int error __unused)
343 {
344 
345 	panic("null_set_syscall_retval");
346 }
347 
348 struct sysentvec null_sysvec = {
349 	.sv_size	= 0,
350 	.sv_table	= NULL,
351 	.sv_mask	= 0,
352 	.sv_sigsize	= 0,
353 	.sv_sigtbl	= NULL,
354 	.sv_errsize	= 0,
355 	.sv_errtbl	= NULL,
356 	.sv_transtrap	= NULL,
357 	.sv_fixup	= NULL,
358 	.sv_sendsig	= NULL,
359 	.sv_sigcode	= NULL,
360 	.sv_szsigcode	= NULL,
361 	.sv_prepsyscall	= NULL,
362 	.sv_name	= "null",
363 	.sv_coredump	= NULL,
364 	.sv_imgact_try	= NULL,
365 	.sv_minsigstksz	= 0,
366 	.sv_pagesize	= PAGE_SIZE,
367 	.sv_minuser	= VM_MIN_ADDRESS,
368 	.sv_maxuser	= VM_MAXUSER_ADDRESS,
369 	.sv_usrstack	= USRSTACK,
370 	.sv_psstrings	= PS_STRINGS,
371 	.sv_stackprot	= VM_PROT_ALL,
372 	.sv_copyout_strings	= NULL,
373 	.sv_setregs	= NULL,
374 	.sv_fixlimit	= NULL,
375 	.sv_maxssiz	= NULL,
376 	.sv_flags	= 0,
377 	.sv_set_syscall_retval = null_set_syscall_retval,
378 	.sv_fetch_syscall_args = null_fetch_syscall_args,
379 	.sv_syscallnames = NULL,
380 };
381 
382 /*
383  ***************************************************************************
384  ****
385  **** The two following SYSINIT's are proc0 specific glue code.  I am not
386  **** convinced that they can not be safely combined, but their order of
387  **** operation has been maintained as the same as the original init_main.c
388  **** for right now.
389  ****
390  **** These probably belong in init_proc.c or kern_proc.c, since they
391  **** deal with proc0 (the fork template process).
392  ****
393  ***************************************************************************
394  */
395 /* ARGSUSED*/
396 static void
397 proc0_init(void *dummy __unused)
398 {
399 	struct proc *p;
400 	struct thread *td;
401 	vm_paddr_t pageablemem;
402 	int i;
403 
404 	GIANT_REQUIRED;
405 	p = &proc0;
406 	td = &thread0;
407 
408 	/*
409 	 * Initialize magic number and osrel.
410 	 */
411 	p->p_magic = P_MAGIC;
412 	p->p_osrel = osreldate;
413 
414 	/*
415 	 * Initialize thread and process structures.
416 	 */
417 	procinit();	/* set up proc zone */
418 	threadinit();	/* set up UMA zones */
419 
420 	/*
421 	 * Initialise scheduler resources.
422 	 * Add scheduler specific parts to proc, thread as needed.
423 	 */
424 	schedinit();	/* scheduler gets its house in order */
425 	/*
426 	 * Initialize sleep queue hash table
427 	 */
428 	sleepinit();
429 
430 	/*
431 	 * additional VM structures
432 	 */
433 	vm_init2();
434 
435 	/*
436 	 * Create process 0 (the swapper).
437 	 */
438 	LIST_INSERT_HEAD(&allproc, p, p_list);
439 	LIST_INSERT_HEAD(PIDHASH(0), p, p_hash);
440 	mtx_init(&pgrp0.pg_mtx, "process group", NULL, MTX_DEF | MTX_DUPOK);
441 	p->p_pgrp = &pgrp0;
442 	LIST_INSERT_HEAD(PGRPHASH(0), &pgrp0, pg_hash);
443 	LIST_INIT(&pgrp0.pg_members);
444 	LIST_INSERT_HEAD(&pgrp0.pg_members, p, p_pglist);
445 
446 	pgrp0.pg_session = &session0;
447 	mtx_init(&session0.s_mtx, "session", NULL, MTX_DEF);
448 	refcount_init(&session0.s_count, 1);
449 	session0.s_leader = p;
450 
451 	p->p_sysent = &null_sysvec;
452 	p->p_flag = P_SYSTEM | P_INMEM;
453 	p->p_state = PRS_NORMAL;
454 	knlist_init_mtx(&p->p_klist, &p->p_mtx);
455 	STAILQ_INIT(&p->p_ktr);
456 	p->p_nice = NZERO;
457 	td->td_tid = PID_MAX + 1;
458 	LIST_INSERT_HEAD(TIDHASH(td->td_tid), td, td_hash);
459 	td->td_state = TDS_RUNNING;
460 	td->td_pri_class = PRI_TIMESHARE;
461 	td->td_user_pri = PUSER;
462 	td->td_base_user_pri = PUSER;
463 	td->td_priority = PVM;
464 	td->td_base_pri = PUSER;
465 	td->td_oncpu = 0;
466 	td->td_flags = TDF_INMEM|TDP_KTHREAD;
467 	td->td_cpuset = cpuset_thread0();
468 	prison0.pr_cpuset = cpuset_ref(td->td_cpuset);
469 	p->p_peers = 0;
470 	p->p_leader = p;
471 
472 
473 	strncpy(p->p_comm, "kernel", sizeof (p->p_comm));
474 	strncpy(td->td_name, "swapper", sizeof (td->td_name));
475 
476 	callout_init(&p->p_itcallout, CALLOUT_MPSAFE);
477 	callout_init_mtx(&p->p_limco, &p->p_mtx, 0);
478 	callout_init(&td->td_slpcallout, CALLOUT_MPSAFE);
479 
480 	/* Create credentials. */
481 	p->p_ucred = crget();
482 	p->p_ucred->cr_ngroups = 1;	/* group 0 */
483 	p->p_ucred->cr_uidinfo = uifind(0);
484 	p->p_ucred->cr_ruidinfo = uifind(0);
485 	p->p_ucred->cr_prison = &prison0;
486 #ifdef AUDIT
487 	audit_cred_kproc0(p->p_ucred);
488 #endif
489 #ifdef MAC
490 	mac_cred_create_swapper(p->p_ucred);
491 #endif
492 	td->td_ucred = crhold(p->p_ucred);
493 
494 	/* Create sigacts. */
495 	p->p_sigacts = sigacts_alloc();
496 
497 	/* Initialize signal state for process 0. */
498 	siginit(&proc0);
499 
500 	/* Create the file descriptor table. */
501 	p->p_fd = fdinit(NULL);
502 	p->p_fdtol = NULL;
503 
504 	/* Create the limits structures. */
505 	p->p_limit = lim_alloc();
506 	for (i = 0; i < RLIM_NLIMITS; i++)
507 		p->p_limit->pl_rlimit[i].rlim_cur =
508 		    p->p_limit->pl_rlimit[i].rlim_max = RLIM_INFINITY;
509 	p->p_limit->pl_rlimit[RLIMIT_NOFILE].rlim_cur =
510 	    p->p_limit->pl_rlimit[RLIMIT_NOFILE].rlim_max = maxfiles;
511 	p->p_limit->pl_rlimit[RLIMIT_NPROC].rlim_cur =
512 	    p->p_limit->pl_rlimit[RLIMIT_NPROC].rlim_max = maxproc;
513 	p->p_limit->pl_rlimit[RLIMIT_DATA].rlim_cur = dfldsiz;
514 	p->p_limit->pl_rlimit[RLIMIT_DATA].rlim_max = maxdsiz;
515 	p->p_limit->pl_rlimit[RLIMIT_STACK].rlim_cur = dflssiz;
516 	p->p_limit->pl_rlimit[RLIMIT_STACK].rlim_max = maxssiz;
517 	/* Cast to avoid overflow on i386/PAE. */
518 	pageablemem = ptoa((vm_paddr_t)cnt.v_free_count);
519 	p->p_limit->pl_rlimit[RLIMIT_RSS].rlim_cur =
520 	    p->p_limit->pl_rlimit[RLIMIT_RSS].rlim_max = pageablemem;
521 	p->p_limit->pl_rlimit[RLIMIT_MEMLOCK].rlim_cur = pageablemem / 3;
522 	p->p_limit->pl_rlimit[RLIMIT_MEMLOCK].rlim_max = pageablemem;
523 	p->p_cpulimit = RLIM_INFINITY;
524 
525 	p->p_stats = pstats_alloc();
526 
527 	/* Allocate a prototype map so we have something to fork. */
528 	pmap_pinit0(vmspace_pmap(&vmspace0));
529 	p->p_vmspace = &vmspace0;
530 	vmspace0.vm_refcnt = 1;
531 
532 	/*
533 	 * proc0 is not expected to enter usermode, so there is no special
534 	 * handling for sv_minuser here, like is done for exec_new_vmspace().
535 	 */
536 	vm_map_init(&vmspace0.vm_map, vmspace_pmap(&vmspace0),
537 	    p->p_sysent->sv_minuser, p->p_sysent->sv_maxuser);
538 
539 	/*
540 	 * Call the init and ctor for the new thread and proc.  We wait
541 	 * to do this until all other structures are fairly sane.
542 	 */
543 	EVENTHANDLER_INVOKE(process_init, p);
544 	EVENTHANDLER_INVOKE(thread_init, td);
545 	EVENTHANDLER_INVOKE(process_ctor, p);
546 	EVENTHANDLER_INVOKE(thread_ctor, td);
547 
548 	/*
549 	 * Charge root for one process.
550 	 */
551 	(void)chgproccnt(p->p_ucred->cr_ruidinfo, 1, 0);
552 }
553 SYSINIT(p0init, SI_SUB_INTRINSIC, SI_ORDER_FIRST, proc0_init, NULL);
554 
555 /* ARGSUSED*/
556 static void
557 proc0_post(void *dummy __unused)
558 {
559 	struct timespec ts;
560 	struct proc *p;
561 	struct rusage ru;
562 	struct thread *td;
563 
564 	/*
565 	 * Now we can look at the time, having had a chance to verify the
566 	 * time from the filesystem.  Pretend that proc0 started now.
567 	 */
568 	sx_slock(&allproc_lock);
569 	FOREACH_PROC_IN_SYSTEM(p) {
570 		microuptime(&p->p_stats->p_start);
571 		PROC_SLOCK(p);
572 		rufetch(p, &ru);	/* Clears thread stats */
573 		PROC_SUNLOCK(p);
574 		p->p_rux.rux_runtime = 0;
575 		p->p_rux.rux_uticks = 0;
576 		p->p_rux.rux_sticks = 0;
577 		p->p_rux.rux_iticks = 0;
578 		FOREACH_THREAD_IN_PROC(p, td) {
579 			td->td_runtime = 0;
580 		}
581 	}
582 	sx_sunlock(&allproc_lock);
583 	PCPU_SET(switchtime, cpu_ticks());
584 	PCPU_SET(switchticks, ticks);
585 
586 	/*
587 	 * Give the ``random'' number generator a thump.
588 	 */
589 	nanotime(&ts);
590 	srandom(ts.tv_sec ^ ts.tv_nsec);
591 }
592 SYSINIT(p0post, SI_SUB_INTRINSIC_POST, SI_ORDER_FIRST, proc0_post, NULL);
593 
594 static void
595 random_init(void *dummy __unused)
596 {
597 
598 	/*
599 	 * After CPU has been started we have some randomness on most
600 	 * platforms via get_cyclecount().  For platforms that don't
601 	 * we will reseed random(9) in proc0_post() as well.
602 	 */
603 	srandom(get_cyclecount());
604 }
605 SYSINIT(random, SI_SUB_RANDOM, SI_ORDER_FIRST, random_init, NULL);
606 
607 /*
608  ***************************************************************************
609  ****
610  **** The following SYSINIT's and glue code should be moved to the
611  **** respective files on a per subsystem basis.
612  ****
613  ***************************************************************************
614  */
615 
616 
617 /*
618  ***************************************************************************
619  ****
620  **** The following code probably belongs in another file, like
621  **** kern/init_init.c.
622  ****
623  ***************************************************************************
624  */
625 
626 /*
627  * List of paths to try when searching for "init".
628  */
629 static char init_path[MAXPATHLEN] =
630 #ifdef	INIT_PATH
631     __XSTRING(INIT_PATH);
632 #else
633     "/sbin/init:/sbin/oinit:/sbin/init.bak:/rescue/init:/stand/sysinstall";
634 #endif
635 SYSCTL_STRING(_kern, OID_AUTO, init_path, CTLFLAG_RD, init_path, 0,
636 	"Path used to search the init process");
637 
638 /*
639  * Shutdown timeout of init(8).
640  * Unused within kernel, but used to control init(8), hence do not remove.
641  */
642 #ifndef INIT_SHUTDOWN_TIMEOUT
643 #define INIT_SHUTDOWN_TIMEOUT 120
644 #endif
645 static int init_shutdown_timeout = INIT_SHUTDOWN_TIMEOUT;
646 SYSCTL_INT(_kern, OID_AUTO, init_shutdown_timeout,
647 	CTLFLAG_RW, &init_shutdown_timeout, 0, "Shutdown timeout of init(8). "
648 	"Unused within kernel, but used to control init(8)");
649 
650 /*
651  * Start the initial user process; try exec'ing each pathname in init_path.
652  * The program is invoked with one argument containing the boot flags.
653  */
654 static void
655 start_init(void *dummy)
656 {
657 	vm_offset_t addr;
658 	struct execve_args args;
659 	int options, error;
660 	char *var, *path, *next, *s;
661 	char *ucp, **uap, *arg0, *arg1;
662 	struct thread *td;
663 	struct proc *p;
664 
665 	mtx_lock(&Giant);
666 
667 	GIANT_REQUIRED;
668 
669 	td = curthread;
670 	p = td->td_proc;
671 
672 	vfs_mountroot();
673 
674 	/*
675 	 * Need just enough stack to hold the faked-up "execve()" arguments.
676 	 */
677 	addr = p->p_sysent->sv_usrstack - PAGE_SIZE;
678 	if (vm_map_find(&p->p_vmspace->vm_map, NULL, 0, &addr, PAGE_SIZE,
679 			FALSE, VM_PROT_ALL, VM_PROT_ALL, 0) != 0)
680 		panic("init: couldn't allocate argument space");
681 	p->p_vmspace->vm_maxsaddr = (caddr_t)addr;
682 	p->p_vmspace->vm_ssize = 1;
683 
684 	if ((var = getenv("init_path")) != NULL) {
685 		strlcpy(init_path, var, sizeof(init_path));
686 		freeenv(var);
687 	}
688 
689 	for (path = init_path; *path != '\0'; path = next) {
690 		while (*path == ':')
691 			path++;
692 		if (*path == '\0')
693 			break;
694 		for (next = path; *next != '\0' && *next != ':'; next++)
695 			/* nothing */ ;
696 		if (bootverbose)
697 			printf("start_init: trying %.*s\n", (int)(next - path),
698 			    path);
699 
700 		/*
701 		 * Move out the boot flag argument.
702 		 */
703 		options = 0;
704 		ucp = (char *)p->p_sysent->sv_usrstack;
705 		(void)subyte(--ucp, 0);		/* trailing zero */
706 		if (boothowto & RB_SINGLE) {
707 			(void)subyte(--ucp, 's');
708 			options = 1;
709 		}
710 #ifdef notyet
711                 if (boothowto & RB_FASTBOOT) {
712 			(void)subyte(--ucp, 'f');
713 			options = 1;
714 		}
715 #endif
716 
717 #ifdef BOOTCDROM
718 		(void)subyte(--ucp, 'C');
719 		options = 1;
720 #endif
721 
722 		if (options == 0)
723 			(void)subyte(--ucp, '-');
724 		(void)subyte(--ucp, '-');		/* leading hyphen */
725 		arg1 = ucp;
726 
727 		/*
728 		 * Move out the file name (also arg 0).
729 		 */
730 		(void)subyte(--ucp, 0);
731 		for (s = next - 1; s >= path; s--)
732 			(void)subyte(--ucp, *s);
733 		arg0 = ucp;
734 
735 		/*
736 		 * Move out the arg pointers.
737 		 */
738 		uap = (char **)((intptr_t)ucp & ~(sizeof(intptr_t)-1));
739 		(void)suword((caddr_t)--uap, (long)0);	/* terminator */
740 		(void)suword((caddr_t)--uap, (long)(intptr_t)arg1);
741 		(void)suword((caddr_t)--uap, (long)(intptr_t)arg0);
742 
743 		/*
744 		 * Point at the arguments.
745 		 */
746 		args.fname = arg0;
747 		args.argv = uap;
748 		args.envv = NULL;
749 
750 		/*
751 		 * Now try to exec the program.  If can't for any reason
752 		 * other than it doesn't exist, complain.
753 		 *
754 		 * Otherwise, return via fork_trampoline() all the way
755 		 * to user mode as init!
756 		 */
757 		if ((error = execve(td, &args)) == 0) {
758 			mtx_unlock(&Giant);
759 			return;
760 		}
761 		if (error != ENOENT)
762 			printf("exec %.*s: error %d\n", (int)(next - path),
763 			    path, error);
764 	}
765 	printf("init: not found in path %s\n", init_path);
766 	panic("no init");
767 }
768 
769 /*
770  * Like kproc_create(), but runs in it's own address space.
771  * We do this early to reserve pid 1.
772  *
773  * Note special case - do not make it runnable yet.  Other work
774  * in progress will change this more.
775  */
776 static void
777 create_init(const void *udata __unused)
778 {
779 	struct ucred *newcred, *oldcred;
780 	int error;
781 
782 	error = fork1(&thread0, RFFDG | RFPROC | RFSTOPPED, 0, &initproc);
783 	if (error)
784 		panic("cannot fork init: %d\n", error);
785 	KASSERT(initproc->p_pid == 1, ("create_init: initproc->p_pid != 1"));
786 	/* divorce init's credentials from the kernel's */
787 	newcred = crget();
788 	PROC_LOCK(initproc);
789 	initproc->p_flag |= P_SYSTEM | P_INMEM;
790 	oldcred = initproc->p_ucred;
791 	crcopy(newcred, oldcred);
792 #ifdef MAC
793 	mac_cred_create_init(newcred);
794 #endif
795 #ifdef AUDIT
796 	audit_cred_proc1(newcred);
797 #endif
798 	initproc->p_ucred = newcred;
799 	PROC_UNLOCK(initproc);
800 	crfree(oldcred);
801 	cred_update_thread(FIRST_THREAD_IN_PROC(initproc));
802 	cpu_set_fork_handler(FIRST_THREAD_IN_PROC(initproc), start_init, NULL);
803 }
804 SYSINIT(init, SI_SUB_CREATE_INIT, SI_ORDER_FIRST, create_init, NULL);
805 
806 /*
807  * Make it runnable now.
808  */
809 static void
810 kick_init(const void *udata __unused)
811 {
812 	struct thread *td;
813 
814 	td = FIRST_THREAD_IN_PROC(initproc);
815 	thread_lock(td);
816 	TD_SET_CAN_RUN(td);
817 	sched_add(td, SRQ_BORING);
818 	thread_unlock(td);
819 }
820 SYSINIT(kickinit, SI_SUB_KTHREAD_INIT, SI_ORDER_FIRST, kick_init, NULL);
821