xref: /freebsd/sys/kern/init_main.c (revision ca987d4641cdcd7f27e153db17c5bf064934faf5)
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_verbose_sysinit.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/loginclass.h>
60 #include <sys/mount.h>
61 #include <sys/mutex.h>
62 #include <sys/syscallsubr.h>
63 #include <sys/sysctl.h>
64 #include <sys/proc.h>
65 #include <sys/racct.h>
66 #include <sys/resourcevar.h>
67 #include <sys/systm.h>
68 #include <sys/signalvar.h>
69 #include <sys/vnode.h>
70 #include <sys/sysent.h>
71 #include <sys/reboot.h>
72 #include <sys/sched.h>
73 #include <sys/sx.h>
74 #include <sys/sysproto.h>
75 #include <sys/vmmeter.h>
76 #include <sys/unistd.h>
77 #include <sys/malloc.h>
78 #include <sys/conf.h>
79 #include <sys/cpuset.h>
80 
81 #include <machine/cpu.h>
82 
83 #include <security/audit/audit.h>
84 #include <security/mac/mac_framework.h>
85 
86 #include <vm/vm.h>
87 #include <vm/vm_param.h>
88 #include <vm/pmap.h>
89 #include <vm/vm_map.h>
90 #include <vm/vm_domain.h>
91 #include <sys/copyright.h>
92 
93 #include <ddb/ddb.h>
94 #include <ddb/db_sym.h>
95 
96 void mi_startup(void);				/* Should be elsewhere */
97 
98 /* Components of the first process -- never freed. */
99 static struct session session0;
100 static struct pgrp pgrp0;
101 struct	proc proc0;
102 struct thread0_storage thread0_st __aligned(32);
103 struct	vmspace vmspace0;
104 struct	proc *initproc;
105 
106 #ifndef BOOTHOWTO
107 #define	BOOTHOWTO	0
108 #endif
109 int	boothowto = BOOTHOWTO;	/* initialized so that it can be patched */
110 SYSCTL_INT(_debug, OID_AUTO, boothowto, CTLFLAG_RD, &boothowto, 0,
111 	"Boot control flags, passed from loader");
112 
113 #ifndef BOOTVERBOSE
114 #define	BOOTVERBOSE	0
115 #endif
116 int	bootverbose = BOOTVERBOSE;
117 SYSCTL_INT(_debug, OID_AUTO, bootverbose, CTLFLAG_RW, &bootverbose, 0,
118 	"Control the output of verbose kernel messages");
119 
120 #ifdef INVARIANTS
121 FEATURE(invariants, "Kernel compiled with INVARIANTS, may affect performance");
122 #endif
123 
124 /*
125  * This ensures that there is at least one entry so that the sysinit_set
126  * symbol is not undefined.  A sybsystem ID of SI_SUB_DUMMY is never
127  * executed.
128  */
129 SYSINIT(placeholder, SI_SUB_DUMMY, SI_ORDER_ANY, NULL, NULL);
130 
131 /*
132  * The sysinit table itself.  Items are checked off as the are run.
133  * If we want to register new sysinit types, add them to newsysinit.
134  */
135 SET_DECLARE(sysinit_set, struct sysinit);
136 struct sysinit **sysinit, **sysinit_end;
137 struct sysinit **newsysinit, **newsysinit_end;
138 
139 EVENTHANDLER_LIST_DECLARE(process_init);
140 EVENTHANDLER_LIST_DECLARE(thread_init);
141 EVENTHANDLER_LIST_DECLARE(process_ctor);
142 EVENTHANDLER_LIST_DECLARE(thread_ctor);
143 
144 /*
145  * Merge a new sysinit set into the current set, reallocating it if
146  * necessary.  This can only be called after malloc is running.
147  */
148 void
149 sysinit_add(struct sysinit **set, struct sysinit **set_end)
150 {
151 	struct sysinit **newset;
152 	struct sysinit **sipp;
153 	struct sysinit **xipp;
154 	int count;
155 
156 	count = set_end - set;
157 	if (newsysinit)
158 		count += newsysinit_end - newsysinit;
159 	else
160 		count += sysinit_end - sysinit;
161 	newset = malloc(count * sizeof(*sipp), M_TEMP, M_NOWAIT);
162 	if (newset == NULL)
163 		panic("cannot malloc for sysinit");
164 	xipp = newset;
165 	if (newsysinit)
166 		for (sipp = newsysinit; sipp < newsysinit_end; sipp++)
167 			*xipp++ = *sipp;
168 	else
169 		for (sipp = sysinit; sipp < sysinit_end; sipp++)
170 			*xipp++ = *sipp;
171 	for (sipp = set; sipp < set_end; sipp++)
172 		*xipp++ = *sipp;
173 	if (newsysinit)
174 		free(newsysinit, M_TEMP);
175 	newsysinit = newset;
176 	newsysinit_end = newset + count;
177 }
178 
179 #if defined (DDB) && defined(VERBOSE_SYSINIT)
180 static const char *
181 symbol_name(vm_offset_t va, db_strategy_t strategy)
182 {
183 	const char *name;
184 	c_db_sym_t sym;
185 	db_expr_t  offset;
186 
187 	if (va == 0)
188 		return (NULL);
189 	sym = db_search_symbol(va, strategy, &offset);
190 	if (offset != 0)
191 		return (NULL);
192 	db_symbol_values(sym, &name, NULL);
193 	return (name);
194 }
195 #endif
196 
197 /*
198  * System startup; initialize the world, create process 0, mount root
199  * filesystem, and fork to create init and pagedaemon.  Most of the
200  * hard work is done in the lower-level initialization routines including
201  * startup(), which does memory initialization and autoconfiguration.
202  *
203  * This allows simple addition of new kernel subsystems that require
204  * boot time initialization.  It also allows substitution of subsystem
205  * (for instance, a scheduler, kernel profiler, or VM system) by object
206  * module.  Finally, it allows for optional "kernel threads".
207  */
208 void
209 mi_startup(void)
210 {
211 
212 	struct sysinit **sipp;	/* system initialization*/
213 	struct sysinit **xipp;	/* interior loop of sort*/
214 	struct sysinit *save;	/* bubble*/
215 
216 #if defined(VERBOSE_SYSINIT)
217 	int last;
218 	int verbose;
219 #endif
220 
221 	if (boothowto & RB_VERBOSE)
222 		bootverbose++;
223 
224 	if (sysinit == NULL) {
225 		sysinit = SET_BEGIN(sysinit_set);
226 		sysinit_end = SET_LIMIT(sysinit_set);
227 	}
228 
229 restart:
230 	/*
231 	 * Perform a bubble sort of the system initialization objects by
232 	 * their subsystem (primary key) and order (secondary key).
233 	 */
234 	for (sipp = sysinit; sipp < sysinit_end; sipp++) {
235 		for (xipp = sipp + 1; xipp < sysinit_end; xipp++) {
236 			if ((*sipp)->subsystem < (*xipp)->subsystem ||
237 			     ((*sipp)->subsystem == (*xipp)->subsystem &&
238 			      (*sipp)->order <= (*xipp)->order))
239 				continue;	/* skip*/
240 			save = *sipp;
241 			*sipp = *xipp;
242 			*xipp = save;
243 		}
244 	}
245 
246 #if defined(VERBOSE_SYSINIT)
247 	last = SI_SUB_COPYRIGHT;
248 	verbose = 0;
249 #if !defined(DDB)
250 	printf("VERBOSE_SYSINIT: DDB not enabled, symbol lookups disabled.\n");
251 #endif
252 #endif
253 
254 	/*
255 	 * Traverse the (now) ordered list of system initialization tasks.
256 	 * Perform each task, and continue on to the next task.
257 	 */
258 	for (sipp = sysinit; sipp < sysinit_end; sipp++) {
259 
260 		if ((*sipp)->subsystem == SI_SUB_DUMMY)
261 			continue;	/* skip dummy task(s)*/
262 
263 		if ((*sipp)->subsystem == SI_SUB_DONE)
264 			continue;
265 
266 #if defined(VERBOSE_SYSINIT)
267 		if ((*sipp)->subsystem > last) {
268 			verbose = 1;
269 			last = (*sipp)->subsystem;
270 			printf("subsystem %x\n", last);
271 		}
272 		if (verbose) {
273 #if defined(DDB)
274 			const char *func, *data;
275 
276 			func = symbol_name((vm_offset_t)(*sipp)->func,
277 			    DB_STGY_PROC);
278 			data = symbol_name((vm_offset_t)(*sipp)->udata,
279 			    DB_STGY_ANY);
280 			if (func != NULL && data != NULL)
281 				printf("   %s(&%s)... ", func, data);
282 			else if (func != NULL)
283 				printf("   %s(%p)... ", func, (*sipp)->udata);
284 			else
285 #endif
286 				printf("   %p(%p)... ", (*sipp)->func,
287 				    (*sipp)->udata);
288 		}
289 #endif
290 
291 		/* Call function */
292 		(*((*sipp)->func))((*sipp)->udata);
293 
294 #if defined(VERBOSE_SYSINIT)
295 		if (verbose)
296 			printf("done.\n");
297 #endif
298 
299 		/* Check off the one we're just done */
300 		(*sipp)->subsystem = SI_SUB_DONE;
301 
302 		/* Check if we've installed more sysinit items via KLD */
303 		if (newsysinit != NULL) {
304 			if (sysinit != SET_BEGIN(sysinit_set))
305 				free(sysinit, M_TEMP);
306 			sysinit = newsysinit;
307 			sysinit_end = newsysinit_end;
308 			newsysinit = NULL;
309 			newsysinit_end = NULL;
310 			goto restart;
311 		}
312 	}
313 
314 	mtx_assert(&Giant, MA_OWNED | MA_NOTRECURSED);
315 	mtx_unlock(&Giant);
316 
317 	/*
318 	 * Now hand over this thread to swapper.
319 	 */
320 	swapper();
321 	/* NOTREACHED*/
322 }
323 
324 static void
325 print_caddr_t(void *data)
326 {
327 	printf("%s", (char *)data);
328 }
329 
330 static void
331 print_version(void *data __unused)
332 {
333 	int len;
334 
335 	/* Strip a trailing newline from version. */
336 	len = strlen(version);
337 	while (len > 0 && version[len - 1] == '\n')
338 		len--;
339 	printf("%.*s %s\n", len, version, machine);
340 	printf("%s\n", compiler_version);
341 }
342 
343 SYSINIT(announce, SI_SUB_COPYRIGHT, SI_ORDER_FIRST, print_caddr_t,
344     copyright);
345 SYSINIT(trademark, SI_SUB_COPYRIGHT, SI_ORDER_SECOND, print_caddr_t,
346     trademark);
347 SYSINIT(version, SI_SUB_COPYRIGHT, SI_ORDER_THIRD, print_version, NULL);
348 
349 #ifdef WITNESS
350 static char wit_warn[] =
351      "WARNING: WITNESS option enabled, expect reduced performance.\n";
352 SYSINIT(witwarn, SI_SUB_COPYRIGHT, SI_ORDER_THIRD + 1,
353    print_caddr_t, wit_warn);
354 SYSINIT(witwarn2, SI_SUB_LAST, SI_ORDER_THIRD + 1,
355    print_caddr_t, wit_warn);
356 #endif
357 
358 #ifdef DIAGNOSTIC
359 static char diag_warn[] =
360      "WARNING: DIAGNOSTIC option enabled, expect reduced performance.\n";
361 SYSINIT(diagwarn, SI_SUB_COPYRIGHT, SI_ORDER_THIRD + 2,
362     print_caddr_t, diag_warn);
363 SYSINIT(diagwarn2, SI_SUB_LAST, SI_ORDER_THIRD + 2,
364     print_caddr_t, diag_warn);
365 #endif
366 
367 static int
368 null_fetch_syscall_args(struct thread *td __unused)
369 {
370 
371 	panic("null_fetch_syscall_args");
372 }
373 
374 static void
375 null_set_syscall_retval(struct thread *td __unused, int error __unused)
376 {
377 
378 	panic("null_set_syscall_retval");
379 }
380 
381 struct sysentvec null_sysvec = {
382 	.sv_size	= 0,
383 	.sv_table	= NULL,
384 	.sv_mask	= 0,
385 	.sv_errsize	= 0,
386 	.sv_errtbl	= NULL,
387 	.sv_transtrap	= NULL,
388 	.sv_fixup	= NULL,
389 	.sv_sendsig	= NULL,
390 	.sv_sigcode	= NULL,
391 	.sv_szsigcode	= NULL,
392 	.sv_name	= "null",
393 	.sv_coredump	= NULL,
394 	.sv_imgact_try	= NULL,
395 	.sv_minsigstksz	= 0,
396 	.sv_pagesize	= PAGE_SIZE,
397 	.sv_minuser	= VM_MIN_ADDRESS,
398 	.sv_maxuser	= VM_MAXUSER_ADDRESS,
399 	.sv_usrstack	= USRSTACK,
400 	.sv_psstrings	= PS_STRINGS,
401 	.sv_stackprot	= VM_PROT_ALL,
402 	.sv_copyout_strings	= NULL,
403 	.sv_setregs	= NULL,
404 	.sv_fixlimit	= NULL,
405 	.sv_maxssiz	= NULL,
406 	.sv_flags	= 0,
407 	.sv_set_syscall_retval = null_set_syscall_retval,
408 	.sv_fetch_syscall_args = null_fetch_syscall_args,
409 	.sv_syscallnames = NULL,
410 	.sv_schedtail	= NULL,
411 	.sv_thread_detach = NULL,
412 	.sv_trap	= NULL,
413 };
414 
415 /*
416  * The two following SYSINIT's are proc0 specific glue code.  I am not
417  * convinced that they can not be safely combined, but their order of
418  * operation has been maintained as the same as the original init_main.c
419  * for right now.
420  */
421 /* ARGSUSED*/
422 static void
423 proc0_init(void *dummy __unused)
424 {
425 	struct proc *p;
426 	struct thread *td;
427 	struct ucred *newcred;
428 	struct uidinfo tmpuinfo;
429 	struct loginclass tmplc = {
430 		.lc_name = "",
431 	};
432 	vm_paddr_t pageablemem;
433 	int i;
434 
435 	GIANT_REQUIRED;
436 	p = &proc0;
437 	td = &thread0;
438 
439 	/*
440 	 * Initialize magic number and osrel.
441 	 */
442 	p->p_magic = P_MAGIC;
443 	p->p_osrel = osreldate;
444 
445 	/*
446 	 * Initialize thread and process structures.
447 	 */
448 	procinit();	/* set up proc zone */
449 	threadinit();	/* set up UMA zones */
450 
451 	/*
452 	 * Initialise scheduler resources.
453 	 * Add scheduler specific parts to proc, thread as needed.
454 	 */
455 	schedinit();	/* scheduler gets its house in order */
456 
457 	/*
458 	 * Create process 0 (the swapper).
459 	 */
460 	LIST_INSERT_HEAD(&allproc, p, p_list);
461 	LIST_INSERT_HEAD(PIDHASH(0), p, p_hash);
462 	mtx_init(&pgrp0.pg_mtx, "process group", NULL, MTX_DEF | MTX_DUPOK);
463 	p->p_pgrp = &pgrp0;
464 	LIST_INSERT_HEAD(PGRPHASH(0), &pgrp0, pg_hash);
465 	LIST_INIT(&pgrp0.pg_members);
466 	LIST_INSERT_HEAD(&pgrp0.pg_members, p, p_pglist);
467 
468 	pgrp0.pg_session = &session0;
469 	mtx_init(&session0.s_mtx, "session", NULL, MTX_DEF);
470 	refcount_init(&session0.s_count, 1);
471 	session0.s_leader = p;
472 
473 	p->p_sysent = &null_sysvec;
474 	p->p_flag = P_SYSTEM | P_INMEM | P_KPROC;
475 	p->p_flag2 = 0;
476 	p->p_state = PRS_NORMAL;
477 	p->p_klist = knlist_alloc(&p->p_mtx);
478 	STAILQ_INIT(&p->p_ktr);
479 	p->p_nice = NZERO;
480 	/* pid_max cannot be greater than PID_MAX */
481 	td->td_tid = PID_MAX + 1;
482 	LIST_INSERT_HEAD(TIDHASH(td->td_tid), td, td_hash);
483 	td->td_state = TDS_RUNNING;
484 	td->td_pri_class = PRI_TIMESHARE;
485 	td->td_user_pri = PUSER;
486 	td->td_base_user_pri = PUSER;
487 	td->td_lend_user_pri = PRI_MAX;
488 	td->td_priority = PVM;
489 	td->td_base_pri = PVM;
490 	td->td_oncpu = curcpu;
491 	td->td_flags = TDF_INMEM;
492 	td->td_pflags = TDP_KTHREAD;
493 	td->td_cpuset = cpuset_thread0();
494 	vm_domain_policy_init(&td->td_vm_dom_policy);
495 	vm_domain_policy_set(&td->td_vm_dom_policy, VM_POLICY_NONE, -1);
496 	vm_domain_policy_init(&p->p_vm_dom_policy);
497 	vm_domain_policy_set(&p->p_vm_dom_policy, VM_POLICY_NONE, -1);
498 	prison0_init();
499 	p->p_peers = 0;
500 	p->p_leader = p;
501 	p->p_reaper = p;
502 	LIST_INIT(&p->p_reaplist);
503 
504 	strncpy(p->p_comm, "kernel", sizeof (p->p_comm));
505 	strncpy(td->td_name, "swapper", sizeof (td->td_name));
506 
507 	callout_init_mtx(&p->p_itcallout, &p->p_mtx, 0);
508 	callout_init_mtx(&p->p_limco, &p->p_mtx, 0);
509 	callout_init(&td->td_slpcallout, 1);
510 
511 	/* Create credentials. */
512 	newcred = crget();
513 	newcred->cr_ngroups = 1;	/* group 0 */
514 	/* A hack to prevent uifind from tripping over NULL pointers. */
515 	curthread->td_ucred = newcred;
516 	tmpuinfo.ui_uid = 1;
517 	newcred->cr_uidinfo = newcred->cr_ruidinfo = &tmpuinfo;
518 	newcred->cr_uidinfo = uifind(0);
519 	newcred->cr_ruidinfo = uifind(0);
520 	newcred->cr_loginclass = &tmplc;
521 	newcred->cr_loginclass = loginclass_find("default");
522 	/* End hack. creds get properly set later with thread_cow_get_proc */
523 	curthread->td_ucred = NULL;
524 	newcred->cr_prison = &prison0;
525 	proc_set_cred_init(p, newcred);
526 #ifdef AUDIT
527 	audit_cred_kproc0(newcred);
528 #endif
529 #ifdef MAC
530 	mac_cred_create_swapper(newcred);
531 #endif
532 	/* Create sigacts. */
533 	p->p_sigacts = sigacts_alloc();
534 
535 	/* Initialize signal state for process 0. */
536 	siginit(&proc0);
537 
538 	/* Create the file descriptor table. */
539 	p->p_fd = fdinit(NULL, false);
540 	p->p_fdtol = NULL;
541 
542 	/* Create the limits structures. */
543 	p->p_limit = lim_alloc();
544 	for (i = 0; i < RLIM_NLIMITS; i++)
545 		p->p_limit->pl_rlimit[i].rlim_cur =
546 		    p->p_limit->pl_rlimit[i].rlim_max = RLIM_INFINITY;
547 	p->p_limit->pl_rlimit[RLIMIT_NOFILE].rlim_cur =
548 	    p->p_limit->pl_rlimit[RLIMIT_NOFILE].rlim_max = maxfiles;
549 	p->p_limit->pl_rlimit[RLIMIT_NPROC].rlim_cur =
550 	    p->p_limit->pl_rlimit[RLIMIT_NPROC].rlim_max = maxproc;
551 	p->p_limit->pl_rlimit[RLIMIT_DATA].rlim_cur = dfldsiz;
552 	p->p_limit->pl_rlimit[RLIMIT_DATA].rlim_max = maxdsiz;
553 	p->p_limit->pl_rlimit[RLIMIT_STACK].rlim_cur = dflssiz;
554 	p->p_limit->pl_rlimit[RLIMIT_STACK].rlim_max = maxssiz;
555 	/* Cast to avoid overflow on i386/PAE. */
556 	pageablemem = ptoa((vm_paddr_t)vm_cnt.v_free_count);
557 	p->p_limit->pl_rlimit[RLIMIT_RSS].rlim_cur =
558 	    p->p_limit->pl_rlimit[RLIMIT_RSS].rlim_max = pageablemem;
559 	p->p_limit->pl_rlimit[RLIMIT_MEMLOCK].rlim_cur = pageablemem / 3;
560 	p->p_limit->pl_rlimit[RLIMIT_MEMLOCK].rlim_max = pageablemem;
561 	p->p_cpulimit = RLIM_INFINITY;
562 
563 	PROC_LOCK(p);
564 	thread_cow_get_proc(td, p);
565 	PROC_UNLOCK(p);
566 
567 	/* Initialize resource accounting structures. */
568 	racct_create(&p->p_racct);
569 
570 	p->p_stats = pstats_alloc();
571 
572 	/* Allocate a prototype map so we have something to fork. */
573 	p->p_vmspace = &vmspace0;
574 	vmspace0.vm_refcnt = 1;
575 	pmap_pinit0(vmspace_pmap(&vmspace0));
576 
577 	/*
578 	 * proc0 is not expected to enter usermode, so there is no special
579 	 * handling for sv_minuser here, like is done for exec_new_vmspace().
580 	 */
581 	vm_map_init(&vmspace0.vm_map, vmspace_pmap(&vmspace0),
582 	    p->p_sysent->sv_minuser, p->p_sysent->sv_maxuser);
583 
584 	/*
585 	 * Call the init and ctor for the new thread and proc.  We wait
586 	 * to do this until all other structures are fairly sane.
587 	 */
588 	EVENTHANDLER_DIRECT_INVOKE(process_init, p);
589 	EVENTHANDLER_DIRECT_INVOKE(thread_init, td);
590 	EVENTHANDLER_DIRECT_INVOKE(process_ctor, p);
591 	EVENTHANDLER_DIRECT_INVOKE(thread_ctor, td);
592 
593 	/*
594 	 * Charge root for one process.
595 	 */
596 	(void)chgproccnt(p->p_ucred->cr_ruidinfo, 1, 0);
597 	PROC_LOCK(p);
598 	racct_add_force(p, RACCT_NPROC, 1);
599 	PROC_UNLOCK(p);
600 }
601 SYSINIT(p0init, SI_SUB_INTRINSIC, SI_ORDER_FIRST, proc0_init, NULL);
602 
603 /* ARGSUSED*/
604 static void
605 proc0_post(void *dummy __unused)
606 {
607 	struct timespec ts;
608 	struct proc *p;
609 	struct rusage ru;
610 	struct thread *td;
611 
612 	/*
613 	 * Now we can look at the time, having had a chance to verify the
614 	 * time from the filesystem.  Pretend that proc0 started now.
615 	 */
616 	sx_slock(&allproc_lock);
617 	FOREACH_PROC_IN_SYSTEM(p) {
618 		microuptime(&p->p_stats->p_start);
619 		PROC_STATLOCK(p);
620 		rufetch(p, &ru);	/* Clears thread stats */
621 		PROC_STATUNLOCK(p);
622 		p->p_rux.rux_runtime = 0;
623 		p->p_rux.rux_uticks = 0;
624 		p->p_rux.rux_sticks = 0;
625 		p->p_rux.rux_iticks = 0;
626 		FOREACH_THREAD_IN_PROC(p, td) {
627 			td->td_runtime = 0;
628 		}
629 	}
630 	sx_sunlock(&allproc_lock);
631 	PCPU_SET(switchtime, cpu_ticks());
632 	PCPU_SET(switchticks, ticks);
633 
634 	/*
635 	 * Give the ``random'' number generator a thump.
636 	 */
637 	nanotime(&ts);
638 	srandom(ts.tv_sec ^ ts.tv_nsec);
639 }
640 SYSINIT(p0post, SI_SUB_INTRINSIC_POST, SI_ORDER_FIRST, proc0_post, NULL);
641 
642 static void
643 random_init(void *dummy __unused)
644 {
645 
646 	/*
647 	 * After CPU has been started we have some randomness on most
648 	 * platforms via get_cyclecount().  For platforms that don't
649 	 * we will reseed random(9) in proc0_post() as well.
650 	 */
651 	srandom(get_cyclecount());
652 }
653 SYSINIT(random, SI_SUB_RANDOM, SI_ORDER_FIRST, random_init, NULL);
654 
655 /*
656  ***************************************************************************
657  ****
658  **** The following SYSINIT's and glue code should be moved to the
659  **** respective files on a per subsystem basis.
660  ****
661  ***************************************************************************
662  */
663 
664 /*
665  * List of paths to try when searching for "init".
666  */
667 static char init_path[MAXPATHLEN] =
668 #ifdef	INIT_PATH
669     __XSTRING(INIT_PATH);
670 #else
671     "/sbin/init:/sbin/oinit:/sbin/init.bak:/rescue/init";
672 #endif
673 SYSCTL_STRING(_kern, OID_AUTO, init_path, CTLFLAG_RD, init_path, 0,
674 	"Path used to search the init process");
675 
676 /*
677  * Shutdown timeout of init(8).
678  * Unused within kernel, but used to control init(8), hence do not remove.
679  */
680 #ifndef INIT_SHUTDOWN_TIMEOUT
681 #define INIT_SHUTDOWN_TIMEOUT 120
682 #endif
683 static int init_shutdown_timeout = INIT_SHUTDOWN_TIMEOUT;
684 SYSCTL_INT(_kern, OID_AUTO, init_shutdown_timeout,
685 	CTLFLAG_RW, &init_shutdown_timeout, 0, "Shutdown timeout of init(8). "
686 	"Unused within kernel, but used to control init(8)");
687 
688 /*
689  * Start the initial user process; try exec'ing each pathname in init_path.
690  * The program is invoked with one argument containing the boot flags.
691  */
692 static void
693 start_init(void *dummy)
694 {
695 	vm_offset_t addr;
696 	struct execve_args args;
697 	int options, error;
698 	char *var, *path, *next, *s;
699 	char *ucp, **uap, *arg0, *arg1;
700 	struct thread *td;
701 	struct proc *p;
702 
703 	mtx_lock(&Giant);
704 
705 	GIANT_REQUIRED;
706 
707 	td = curthread;
708 	p = td->td_proc;
709 
710 	vfs_mountroot();
711 
712 	/* Wipe GELI passphrase from the environment. */
713 	kern_unsetenv("kern.geom.eli.passphrase");
714 
715 	/*
716 	 * Need just enough stack to hold the faked-up "execve()" arguments.
717 	 */
718 	addr = p->p_sysent->sv_usrstack - PAGE_SIZE;
719 	if (vm_map_find(&p->p_vmspace->vm_map, NULL, 0, &addr, PAGE_SIZE, 0,
720 	    VMFS_NO_SPACE, VM_PROT_ALL, VM_PROT_ALL, 0) != 0)
721 		panic("init: couldn't allocate argument space");
722 	p->p_vmspace->vm_maxsaddr = (caddr_t)addr;
723 	p->p_vmspace->vm_ssize = 1;
724 
725 	if ((var = kern_getenv("init_path")) != NULL) {
726 		strlcpy(init_path, var, sizeof(init_path));
727 		freeenv(var);
728 	}
729 
730 	for (path = init_path; *path != '\0'; path = next) {
731 		while (*path == ':')
732 			path++;
733 		if (*path == '\0')
734 			break;
735 		for (next = path; *next != '\0' && *next != ':'; next++)
736 			/* nothing */ ;
737 		if (bootverbose)
738 			printf("start_init: trying %.*s\n", (int)(next - path),
739 			    path);
740 
741 		/*
742 		 * Move out the boot flag argument.
743 		 */
744 		options = 0;
745 		ucp = (char *)p->p_sysent->sv_usrstack;
746 		(void)subyte(--ucp, 0);		/* trailing zero */
747 		if (boothowto & RB_SINGLE) {
748 			(void)subyte(--ucp, 's');
749 			options = 1;
750 		}
751 #ifdef notyet
752                 if (boothowto & RB_FASTBOOT) {
753 			(void)subyte(--ucp, 'f');
754 			options = 1;
755 		}
756 #endif
757 
758 #ifdef BOOTCDROM
759 		(void)subyte(--ucp, 'C');
760 		options = 1;
761 #endif
762 
763 		if (options == 0)
764 			(void)subyte(--ucp, '-');
765 		(void)subyte(--ucp, '-');		/* leading hyphen */
766 		arg1 = ucp;
767 
768 		/*
769 		 * Move out the file name (also arg 0).
770 		 */
771 		(void)subyte(--ucp, 0);
772 		for (s = next - 1; s >= path; s--)
773 			(void)subyte(--ucp, *s);
774 		arg0 = ucp;
775 
776 		/*
777 		 * Move out the arg pointers.
778 		 */
779 		uap = (char **)rounddown2((intptr_t)ucp, sizeof(intptr_t));
780 		(void)suword((caddr_t)--uap, (long)0);	/* terminator */
781 		(void)suword((caddr_t)--uap, (long)(intptr_t)arg1);
782 		(void)suword((caddr_t)--uap, (long)(intptr_t)arg0);
783 
784 		/*
785 		 * Point at the arguments.
786 		 */
787 		args.fname = arg0;
788 		args.argv = uap;
789 		args.envv = NULL;
790 
791 		/*
792 		 * Now try to exec the program.  If can't for any reason
793 		 * other than it doesn't exist, complain.
794 		 *
795 		 * Otherwise, return via fork_trampoline() all the way
796 		 * to user mode as init!
797 		 */
798 		if ((error = sys_execve(td, &args)) == 0) {
799 			mtx_unlock(&Giant);
800 			return;
801 		}
802 		if (error != ENOENT)
803 			printf("exec %.*s: error %d\n", (int)(next - path),
804 			    path, error);
805 	}
806 	printf("init: not found in path %s\n", init_path);
807 	panic("no init");
808 }
809 
810 /*
811  * Like kproc_create(), but runs in its own address space.
812  * We do this early to reserve pid 1.
813  *
814  * Note special case - do not make it runnable yet.  Other work
815  * in progress will change this more.
816  */
817 static void
818 create_init(const void *udata __unused)
819 {
820 	struct fork_req fr;
821 	struct ucred *newcred, *oldcred;
822 	struct thread *td;
823 	int error;
824 
825 	bzero(&fr, sizeof(fr));
826 	fr.fr_flags = RFFDG | RFPROC | RFSTOPPED;
827 	fr.fr_procp = &initproc;
828 	error = fork1(&thread0, &fr);
829 	if (error)
830 		panic("cannot fork init: %d\n", error);
831 	KASSERT(initproc->p_pid == 1, ("create_init: initproc->p_pid != 1"));
832 	/* divorce init's credentials from the kernel's */
833 	newcred = crget();
834 	sx_xlock(&proctree_lock);
835 	PROC_LOCK(initproc);
836 	initproc->p_flag |= P_SYSTEM | P_INMEM;
837 	initproc->p_treeflag |= P_TREE_REAPER;
838 	LIST_INSERT_HEAD(&initproc->p_reaplist, &proc0, p_reapsibling);
839 	oldcred = initproc->p_ucred;
840 	crcopy(newcred, oldcred);
841 #ifdef MAC
842 	mac_cred_create_init(newcred);
843 #endif
844 #ifdef AUDIT
845 	audit_cred_proc1(newcred);
846 #endif
847 	proc_set_cred(initproc, newcred);
848 	td = FIRST_THREAD_IN_PROC(initproc);
849 	crfree(td->td_ucred);
850 	td->td_ucred = crhold(initproc->p_ucred);
851 	PROC_UNLOCK(initproc);
852 	sx_xunlock(&proctree_lock);
853 	crfree(oldcred);
854 	cpu_fork_kthread_handler(FIRST_THREAD_IN_PROC(initproc),
855 	    start_init, NULL);
856 }
857 SYSINIT(init, SI_SUB_CREATE_INIT, SI_ORDER_FIRST, create_init, NULL);
858 
859 /*
860  * Make it runnable now.
861  */
862 static void
863 kick_init(const void *udata __unused)
864 {
865 	struct thread *td;
866 
867 	td = FIRST_THREAD_IN_PROC(initproc);
868 	thread_lock(td);
869 	TD_SET_CAN_RUN(td);
870 	sched_add(td, SRQ_BORING);
871 	thread_unlock(td);
872 }
873 SYSINIT(kickinit, SI_SUB_KTHREAD_INIT, SI_ORDER_MIDDLE, kick_init, NULL);
874