xref: /freebsd/sys/kern/init_main.c (revision 88f578841fd49ff796b62a8d3531bd73afd1a88e)
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	thread thread0 __aligned(16);
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 /*
121  * This ensures that there is at least one entry so that the sysinit_set
122  * symbol is not undefined.  A sybsystem ID of SI_SUB_DUMMY is never
123  * executed.
124  */
125 SYSINIT(placeholder, SI_SUB_DUMMY, SI_ORDER_ANY, NULL, NULL);
126 
127 /*
128  * The sysinit table itself.  Items are checked off as the are run.
129  * If we want to register new sysinit types, add them to newsysinit.
130  */
131 SET_DECLARE(sysinit_set, struct sysinit);
132 struct sysinit **sysinit, **sysinit_end;
133 struct sysinit **newsysinit, **newsysinit_end;
134 
135 /*
136  * Merge a new sysinit set into the current set, reallocating it if
137  * necessary.  This can only be called after malloc is running.
138  */
139 void
140 sysinit_add(struct sysinit **set, struct sysinit **set_end)
141 {
142 	struct sysinit **newset;
143 	struct sysinit **sipp;
144 	struct sysinit **xipp;
145 	int count;
146 
147 	count = set_end - set;
148 	if (newsysinit)
149 		count += newsysinit_end - newsysinit;
150 	else
151 		count += sysinit_end - sysinit;
152 	newset = malloc(count * sizeof(*sipp), M_TEMP, M_NOWAIT);
153 	if (newset == NULL)
154 		panic("cannot malloc for sysinit");
155 	xipp = newset;
156 	if (newsysinit)
157 		for (sipp = newsysinit; sipp < newsysinit_end; sipp++)
158 			*xipp++ = *sipp;
159 	else
160 		for (sipp = sysinit; sipp < sysinit_end; sipp++)
161 			*xipp++ = *sipp;
162 	for (sipp = set; sipp < set_end; sipp++)
163 		*xipp++ = *sipp;
164 	if (newsysinit)
165 		free(newsysinit, M_TEMP);
166 	newsysinit = newset;
167 	newsysinit_end = newset + count;
168 }
169 
170 #if defined (DDB) && defined(VERBOSE_SYSINIT)
171 static const char *
172 symbol_name(vm_offset_t va, db_strategy_t strategy)
173 {
174 	const char *name;
175 	c_db_sym_t sym;
176 	db_expr_t  offset;
177 
178 	if (va == 0)
179 		return (NULL);
180 	sym = db_search_symbol(va, strategy, &offset);
181 	if (offset != 0)
182 		return (NULL);
183 	db_symbol_values(sym, &name, NULL);
184 	return (name);
185 }
186 #endif
187 
188 /*
189  * System startup; initialize the world, create process 0, mount root
190  * filesystem, and fork to create init and pagedaemon.  Most of the
191  * hard work is done in the lower-level initialization routines including
192  * startup(), which does memory initialization and autoconfiguration.
193  *
194  * This allows simple addition of new kernel subsystems that require
195  * boot time initialization.  It also allows substitution of subsystem
196  * (for instance, a scheduler, kernel profiler, or VM system) by object
197  * module.  Finally, it allows for optional "kernel threads".
198  */
199 void
200 mi_startup(void)
201 {
202 
203 	register struct sysinit **sipp;		/* system initialization*/
204 	register struct sysinit **xipp;		/* interior loop of sort*/
205 	register struct sysinit *save;		/* bubble*/
206 
207 #if defined(VERBOSE_SYSINIT)
208 	int last;
209 	int verbose;
210 #endif
211 
212 	if (boothowto & RB_VERBOSE)
213 		bootverbose++;
214 
215 	if (sysinit == NULL) {
216 		sysinit = SET_BEGIN(sysinit_set);
217 		sysinit_end = SET_LIMIT(sysinit_set);
218 	}
219 
220 restart:
221 	/*
222 	 * Perform a bubble sort of the system initialization objects by
223 	 * their subsystem (primary key) and order (secondary key).
224 	 */
225 	for (sipp = sysinit; sipp < sysinit_end; sipp++) {
226 		for (xipp = sipp + 1; xipp < sysinit_end; xipp++) {
227 			if ((*sipp)->subsystem < (*xipp)->subsystem ||
228 			     ((*sipp)->subsystem == (*xipp)->subsystem &&
229 			      (*sipp)->order <= (*xipp)->order))
230 				continue;	/* skip*/
231 			save = *sipp;
232 			*sipp = *xipp;
233 			*xipp = save;
234 		}
235 	}
236 
237 #if defined(VERBOSE_SYSINIT)
238 	last = SI_SUB_COPYRIGHT;
239 	verbose = 0;
240 #if !defined(DDB)
241 	printf("VERBOSE_SYSINIT: DDB not enabled, symbol lookups disabled.\n");
242 #endif
243 #endif
244 
245 	/*
246 	 * Traverse the (now) ordered list of system initialization tasks.
247 	 * Perform each task, and continue on to the next task.
248 	 */
249 	for (sipp = sysinit; sipp < sysinit_end; sipp++) {
250 
251 		if ((*sipp)->subsystem == SI_SUB_DUMMY)
252 			continue;	/* skip dummy task(s)*/
253 
254 		if ((*sipp)->subsystem == SI_SUB_DONE)
255 			continue;
256 
257 #if defined(VERBOSE_SYSINIT)
258 		if ((*sipp)->subsystem > last) {
259 			verbose = 1;
260 			last = (*sipp)->subsystem;
261 			printf("subsystem %x\n", last);
262 		}
263 		if (verbose) {
264 #if defined(DDB)
265 			const char *func, *data;
266 
267 			func = symbol_name((vm_offset_t)(*sipp)->func,
268 			    DB_STGY_PROC);
269 			data = symbol_name((vm_offset_t)(*sipp)->udata,
270 			    DB_STGY_ANY);
271 			if (func != NULL && data != NULL)
272 				printf("   %s(&%s)... ", func, data);
273 			else if (func != NULL)
274 				printf("   %s(%p)... ", func, (*sipp)->udata);
275 			else
276 #endif
277 				printf("   %p(%p)... ", (*sipp)->func,
278 				    (*sipp)->udata);
279 		}
280 #endif
281 
282 		/* Call function */
283 		(*((*sipp)->func))((*sipp)->udata);
284 
285 #if defined(VERBOSE_SYSINIT)
286 		if (verbose)
287 			printf("done.\n");
288 #endif
289 
290 		/* Check off the one we're just done */
291 		(*sipp)->subsystem = SI_SUB_DONE;
292 
293 		/* Check if we've installed more sysinit items via KLD */
294 		if (newsysinit != NULL) {
295 			if (sysinit != SET_BEGIN(sysinit_set))
296 				free(sysinit, M_TEMP);
297 			sysinit = newsysinit;
298 			sysinit_end = newsysinit_end;
299 			newsysinit = NULL;
300 			newsysinit_end = NULL;
301 			goto restart;
302 		}
303 	}
304 
305 	mtx_assert(&Giant, MA_OWNED | MA_NOTRECURSED);
306 	mtx_unlock(&Giant);
307 
308 	/*
309 	 * Now hand over this thread to swapper.
310 	 */
311 	swapper();
312 	/* NOTREACHED*/
313 }
314 
315 
316 /*
317  ***************************************************************************
318  ****
319  **** The following SYSINIT's belong elsewhere, but have not yet
320  **** been moved.
321  ****
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     struct syscall_args *sa __unused)
370 {
371 
372 	panic("null_fetch_syscall_args");
373 }
374 
375 static void
376 null_set_syscall_retval(struct thread *td __unused, int error __unused)
377 {
378 
379 	panic("null_set_syscall_retval");
380 }
381 
382 struct sysentvec null_sysvec = {
383 	.sv_size	= 0,
384 	.sv_table	= NULL,
385 	.sv_mask	= 0,
386 	.sv_sigsize	= 0,
387 	.sv_sigtbl	= NULL,
388 	.sv_errsize	= 0,
389 	.sv_errtbl	= NULL,
390 	.sv_transtrap	= NULL,
391 	.sv_fixup	= NULL,
392 	.sv_sendsig	= NULL,
393 	.sv_sigcode	= NULL,
394 	.sv_szsigcode	= NULL,
395 	.sv_prepsyscall	= NULL,
396 	.sv_name	= "null",
397 	.sv_coredump	= NULL,
398 	.sv_imgact_try	= NULL,
399 	.sv_minsigstksz	= 0,
400 	.sv_pagesize	= PAGE_SIZE,
401 	.sv_minuser	= VM_MIN_ADDRESS,
402 	.sv_maxuser	= VM_MAXUSER_ADDRESS,
403 	.sv_usrstack	= USRSTACK,
404 	.sv_psstrings	= PS_STRINGS,
405 	.sv_stackprot	= VM_PROT_ALL,
406 	.sv_copyout_strings	= NULL,
407 	.sv_setregs	= NULL,
408 	.sv_fixlimit	= NULL,
409 	.sv_maxssiz	= NULL,
410 	.sv_flags	= 0,
411 	.sv_set_syscall_retval = null_set_syscall_retval,
412 	.sv_fetch_syscall_args = null_fetch_syscall_args,
413 	.sv_syscallnames = NULL,
414 	.sv_schedtail	= NULL,
415 	.sv_thread_detach = NULL,
416 };
417 
418 /*
419  ***************************************************************************
420  ****
421  **** The two following SYSINIT's are proc0 specific glue code.  I am not
422  **** convinced that they can not be safely combined, but their order of
423  **** operation has been maintained as the same as the original init_main.c
424  **** for right now.
425  ****
426  **** These probably belong in init_proc.c or kern_proc.c, since they
427  **** deal with proc0 (the fork template process).
428  ****
429  ***************************************************************************
430  */
431 /* ARGSUSED*/
432 static void
433 proc0_init(void *dummy __unused)
434 {
435 	struct proc *p;
436 	struct thread *td;
437 	struct ucred *newcred;
438 	vm_paddr_t pageablemem;
439 	int i;
440 
441 	GIANT_REQUIRED;
442 	p = &proc0;
443 	td = &thread0;
444 
445 	/*
446 	 * Initialize magic number and osrel.
447 	 */
448 	p->p_magic = P_MAGIC;
449 	p->p_osrel = osreldate;
450 
451 	/*
452 	 * Initialize thread and process structures.
453 	 */
454 	procinit();	/* set up proc zone */
455 	threadinit();	/* set up UMA zones */
456 
457 	/*
458 	 * Initialise scheduler resources.
459 	 * Add scheduler specific parts to proc, thread as needed.
460 	 */
461 	schedinit();	/* scheduler gets its house in order */
462 
463 	/*
464 	 * Create process 0 (the swapper).
465 	 */
466 	LIST_INSERT_HEAD(&allproc, p, p_list);
467 	LIST_INSERT_HEAD(PIDHASH(0), p, p_hash);
468 	mtx_init(&pgrp0.pg_mtx, "process group", NULL, MTX_DEF | MTX_DUPOK);
469 	p->p_pgrp = &pgrp0;
470 	LIST_INSERT_HEAD(PGRPHASH(0), &pgrp0, pg_hash);
471 	LIST_INIT(&pgrp0.pg_members);
472 	LIST_INSERT_HEAD(&pgrp0.pg_members, p, p_pglist);
473 
474 	pgrp0.pg_session = &session0;
475 	mtx_init(&session0.s_mtx, "session", NULL, MTX_DEF);
476 	refcount_init(&session0.s_count, 1);
477 	session0.s_leader = p;
478 
479 	p->p_sysent = &null_sysvec;
480 	p->p_flag = P_SYSTEM | P_INMEM;
481 	p->p_flag2 = 0;
482 	p->p_state = PRS_NORMAL;
483 	knlist_init_mtx(&p->p_klist, &p->p_mtx);
484 	STAILQ_INIT(&p->p_ktr);
485 	p->p_nice = NZERO;
486 	/* pid_max cannot be greater than PID_MAX */
487 	td->td_tid = PID_MAX + 1;
488 	LIST_INSERT_HEAD(TIDHASH(td->td_tid), td, td_hash);
489 	td->td_state = TDS_RUNNING;
490 	td->td_pri_class = PRI_TIMESHARE;
491 	td->td_user_pri = PUSER;
492 	td->td_base_user_pri = PUSER;
493 	td->td_lend_user_pri = PRI_MAX;
494 	td->td_priority = PVM;
495 	td->td_base_pri = PVM;
496 	td->td_oncpu = 0;
497 	td->td_flags = TDF_INMEM;
498 	td->td_pflags = TDP_KTHREAD;
499 	td->td_cpuset = cpuset_thread0();
500 	vm_domain_policy_init(&td->td_vm_dom_policy);
501 	vm_domain_policy_set(&td->td_vm_dom_policy, VM_POLICY_NONE, -1);
502 	vm_domain_policy_init(&p->p_vm_dom_policy);
503 	vm_domain_policy_set(&p->p_vm_dom_policy, VM_POLICY_NONE, -1);
504 	prison0_init();
505 	p->p_peers = 0;
506 	p->p_leader = p;
507 	p->p_reaper = p;
508 	LIST_INIT(&p->p_reaplist);
509 
510 	strncpy(p->p_comm, "kernel", sizeof (p->p_comm));
511 	strncpy(td->td_name, "swapper", sizeof (td->td_name));
512 
513 	callout_init_mtx(&p->p_itcallout, &p->p_mtx, 0);
514 	callout_init_mtx(&p->p_limco, &p->p_mtx, 0);
515 	callout_init(&td->td_slpcallout, 1);
516 
517 	/* Create credentials. */
518 	newcred = crget();
519 	newcred->cr_ngroups = 1;	/* group 0 */
520 	newcred->cr_uidinfo = uifind(0);
521 	newcred->cr_ruidinfo = uifind(0);
522 	newcred->cr_prison = &prison0;
523 	newcred->cr_loginclass = loginclass_find("default");
524 	proc_set_cred_init(p, newcred);
525 #ifdef AUDIT
526 	audit_cred_kproc0(newcred);
527 #endif
528 #ifdef MAC
529 	mac_cred_create_swapper(newcred);
530 #endif
531 	/* Create sigacts. */
532 	p->p_sigacts = sigacts_alloc();
533 
534 	/* Initialize signal state for process 0. */
535 	siginit(&proc0);
536 
537 	/* Create the file descriptor table. */
538 	p->p_fd = fdinit(NULL, false);
539 	p->p_fdtol = NULL;
540 
541 	/* Create the limits structures. */
542 	p->p_limit = lim_alloc();
543 	for (i = 0; i < RLIM_NLIMITS; i++)
544 		p->p_limit->pl_rlimit[i].rlim_cur =
545 		    p->p_limit->pl_rlimit[i].rlim_max = RLIM_INFINITY;
546 	p->p_limit->pl_rlimit[RLIMIT_NOFILE].rlim_cur =
547 	    p->p_limit->pl_rlimit[RLIMIT_NOFILE].rlim_max = maxfiles;
548 	p->p_limit->pl_rlimit[RLIMIT_NPROC].rlim_cur =
549 	    p->p_limit->pl_rlimit[RLIMIT_NPROC].rlim_max = maxproc;
550 	p->p_limit->pl_rlimit[RLIMIT_DATA].rlim_cur = dfldsiz;
551 	p->p_limit->pl_rlimit[RLIMIT_DATA].rlim_max = maxdsiz;
552 	p->p_limit->pl_rlimit[RLIMIT_STACK].rlim_cur = dflssiz;
553 	p->p_limit->pl_rlimit[RLIMIT_STACK].rlim_max = maxssiz;
554 	/* Cast to avoid overflow on i386/PAE. */
555 	pageablemem = ptoa((vm_paddr_t)vm_cnt.v_free_count);
556 	p->p_limit->pl_rlimit[RLIMIT_RSS].rlim_cur =
557 	    p->p_limit->pl_rlimit[RLIMIT_RSS].rlim_max = pageablemem;
558 	p->p_limit->pl_rlimit[RLIMIT_MEMLOCK].rlim_cur = pageablemem / 3;
559 	p->p_limit->pl_rlimit[RLIMIT_MEMLOCK].rlim_max = pageablemem;
560 	p->p_cpulimit = RLIM_INFINITY;
561 
562 	PROC_LOCK(p);
563 	thread_cow_get_proc(td, p);
564 	PROC_UNLOCK(p);
565 
566 	/* Initialize resource accounting structures. */
567 	racct_create(&p->p_racct);
568 
569 	p->p_stats = pstats_alloc();
570 
571 	/* Allocate a prototype map so we have something to fork. */
572 	p->p_vmspace = &vmspace0;
573 	vmspace0.vm_refcnt = 1;
574 	pmap_pinit0(vmspace_pmap(&vmspace0));
575 
576 	/*
577 	 * proc0 is not expected to enter usermode, so there is no special
578 	 * handling for sv_minuser here, like is done for exec_new_vmspace().
579 	 */
580 	vm_map_init(&vmspace0.vm_map, vmspace_pmap(&vmspace0),
581 	    p->p_sysent->sv_minuser, p->p_sysent->sv_maxuser);
582 
583 	/*
584 	 * Call the init and ctor for the new thread and proc.  We wait
585 	 * to do this until all other structures are fairly sane.
586 	 */
587 	EVENTHANDLER_INVOKE(process_init, p);
588 	EVENTHANDLER_INVOKE(thread_init, td);
589 	EVENTHANDLER_INVOKE(process_ctor, p);
590 	EVENTHANDLER_INVOKE(thread_ctor, td);
591 
592 	/*
593 	 * Charge root for one process.
594 	 */
595 	(void)chgproccnt(p->p_ucred->cr_ruidinfo, 1, 0);
596 	PROC_LOCK(p);
597 	racct_add_force(p, RACCT_NPROC, 1);
598 	PROC_UNLOCK(p);
599 }
600 SYSINIT(p0init, SI_SUB_INTRINSIC, SI_ORDER_FIRST, proc0_init, NULL);
601 
602 /* ARGSUSED*/
603 static void
604 proc0_post(void *dummy __unused)
605 {
606 	struct timespec ts;
607 	struct proc *p;
608 	struct rusage ru;
609 	struct thread *td;
610 
611 	/*
612 	 * Now we can look at the time, having had a chance to verify the
613 	 * time from the filesystem.  Pretend that proc0 started now.
614 	 */
615 	sx_slock(&allproc_lock);
616 	FOREACH_PROC_IN_SYSTEM(p) {
617 		microuptime(&p->p_stats->p_start);
618 		PROC_STATLOCK(p);
619 		rufetch(p, &ru);	/* Clears thread stats */
620 		PROC_STATUNLOCK(p);
621 		p->p_rux.rux_runtime = 0;
622 		p->p_rux.rux_uticks = 0;
623 		p->p_rux.rux_sticks = 0;
624 		p->p_rux.rux_iticks = 0;
625 		FOREACH_THREAD_IN_PROC(p, td) {
626 			td->td_runtime = 0;
627 		}
628 	}
629 	sx_sunlock(&allproc_lock);
630 	PCPU_SET(switchtime, cpu_ticks());
631 	PCPU_SET(switchticks, ticks);
632 
633 	/*
634 	 * Give the ``random'' number generator a thump.
635 	 */
636 	nanotime(&ts);
637 	srandom(ts.tv_sec ^ ts.tv_nsec);
638 }
639 SYSINIT(p0post, SI_SUB_INTRINSIC_POST, SI_ORDER_FIRST, proc0_post, NULL);
640 
641 static void
642 random_init(void *dummy __unused)
643 {
644 
645 	/*
646 	 * After CPU has been started we have some randomness on most
647 	 * platforms via get_cyclecount().  For platforms that don't
648 	 * we will reseed random(9) in proc0_post() as well.
649 	 */
650 	srandom(get_cyclecount());
651 }
652 SYSINIT(random, SI_SUB_RANDOM, SI_ORDER_FIRST, random_init, NULL);
653 
654 /*
655  ***************************************************************************
656  ****
657  **** The following SYSINIT's and glue code should be moved to the
658  **** respective files on a per subsystem basis.
659  ****
660  ***************************************************************************
661  */
662 
663 
664 /*
665  ***************************************************************************
666  ****
667  **** The following code probably belongs in another file, like
668  **** kern/init_init.c.
669  ****
670  ***************************************************************************
671  */
672 
673 /*
674  * List of paths to try when searching for "init".
675  */
676 static char init_path[MAXPATHLEN] =
677 #ifdef	INIT_PATH
678     __XSTRING(INIT_PATH);
679 #else
680     "/sbin/init:/sbin/oinit:/sbin/init.bak:/rescue/init";
681 #endif
682 SYSCTL_STRING(_kern, OID_AUTO, init_path, CTLFLAG_RD, init_path, 0,
683 	"Path used to search the init process");
684 
685 /*
686  * Shutdown timeout of init(8).
687  * Unused within kernel, but used to control init(8), hence do not remove.
688  */
689 #ifndef INIT_SHUTDOWN_TIMEOUT
690 #define INIT_SHUTDOWN_TIMEOUT 120
691 #endif
692 static int init_shutdown_timeout = INIT_SHUTDOWN_TIMEOUT;
693 SYSCTL_INT(_kern, OID_AUTO, init_shutdown_timeout,
694 	CTLFLAG_RW, &init_shutdown_timeout, 0, "Shutdown timeout of init(8). "
695 	"Unused within kernel, but used to control init(8)");
696 
697 /*
698  * Start the initial user process; try exec'ing each pathname in init_path.
699  * The program is invoked with one argument containing the boot flags.
700  */
701 static void
702 start_init(void *dummy)
703 {
704 	vm_offset_t addr;
705 	struct execve_args args;
706 	int options, error;
707 	char *var, *path, *next, *s;
708 	char *ucp, **uap, *arg0, *arg1;
709 	struct thread *td;
710 	struct proc *p;
711 
712 	mtx_lock(&Giant);
713 
714 	GIANT_REQUIRED;
715 
716 	td = curthread;
717 	p = td->td_proc;
718 
719 	vfs_mountroot();
720 
721 	/* Wipe GELI passphrase from the environment. */
722 	kern_unsetenv("kern.geom.eli.passphrase");
723 
724 	/*
725 	 * Need just enough stack to hold the faked-up "execve()" arguments.
726 	 */
727 	addr = p->p_sysent->sv_usrstack - PAGE_SIZE;
728 	if (vm_map_find(&p->p_vmspace->vm_map, NULL, 0, &addr, PAGE_SIZE, 0,
729 	    VMFS_NO_SPACE, VM_PROT_ALL, VM_PROT_ALL, 0) != 0)
730 		panic("init: couldn't allocate argument space");
731 	p->p_vmspace->vm_maxsaddr = (caddr_t)addr;
732 	p->p_vmspace->vm_ssize = 1;
733 
734 	if ((var = kern_getenv("init_path")) != NULL) {
735 		strlcpy(init_path, var, sizeof(init_path));
736 		freeenv(var);
737 	}
738 
739 	for (path = init_path; *path != '\0'; path = next) {
740 		while (*path == ':')
741 			path++;
742 		if (*path == '\0')
743 			break;
744 		for (next = path; *next != '\0' && *next != ':'; next++)
745 			/* nothing */ ;
746 		if (bootverbose)
747 			printf("start_init: trying %.*s\n", (int)(next - path),
748 			    path);
749 
750 		/*
751 		 * Move out the boot flag argument.
752 		 */
753 		options = 0;
754 		ucp = (char *)p->p_sysent->sv_usrstack;
755 		(void)subyte(--ucp, 0);		/* trailing zero */
756 		if (boothowto & RB_SINGLE) {
757 			(void)subyte(--ucp, 's');
758 			options = 1;
759 		}
760 #ifdef notyet
761                 if (boothowto & RB_FASTBOOT) {
762 			(void)subyte(--ucp, 'f');
763 			options = 1;
764 		}
765 #endif
766 
767 #ifdef BOOTCDROM
768 		(void)subyte(--ucp, 'C');
769 		options = 1;
770 #endif
771 
772 		if (options == 0)
773 			(void)subyte(--ucp, '-');
774 		(void)subyte(--ucp, '-');		/* leading hyphen */
775 		arg1 = ucp;
776 
777 		/*
778 		 * Move out the file name (also arg 0).
779 		 */
780 		(void)subyte(--ucp, 0);
781 		for (s = next - 1; s >= path; s--)
782 			(void)subyte(--ucp, *s);
783 		arg0 = ucp;
784 
785 		/*
786 		 * Move out the arg pointers.
787 		 */
788 		uap = (char **)((intptr_t)ucp & ~(sizeof(intptr_t)-1));
789 		(void)suword((caddr_t)--uap, (long)0);	/* terminator */
790 		(void)suword((caddr_t)--uap, (long)(intptr_t)arg1);
791 		(void)suword((caddr_t)--uap, (long)(intptr_t)arg0);
792 
793 		/*
794 		 * Point at the arguments.
795 		 */
796 		args.fname = arg0;
797 		args.argv = uap;
798 		args.envv = NULL;
799 
800 		/*
801 		 * Now try to exec the program.  If can't for any reason
802 		 * other than it doesn't exist, complain.
803 		 *
804 		 * Otherwise, return via fork_trampoline() all the way
805 		 * to user mode as init!
806 		 */
807 		if ((error = sys_execve(td, &args)) == 0) {
808 			mtx_unlock(&Giant);
809 			return;
810 		}
811 		if (error != ENOENT)
812 			printf("exec %.*s: error %d\n", (int)(next - path),
813 			    path, error);
814 	}
815 	printf("init: not found in path %s\n", init_path);
816 	panic("no init");
817 }
818 
819 /*
820  * Like kproc_create(), but runs in it's own address space.
821  * We do this early to reserve pid 1.
822  *
823  * Note special case - do not make it runnable yet.  Other work
824  * in progress will change this more.
825  */
826 static void
827 create_init(const void *udata __unused)
828 {
829 	struct ucred *newcred, *oldcred;
830 	struct thread *td;
831 	int error;
832 
833 	error = fork1(&thread0, RFFDG | RFPROC | RFSTOPPED, 0, &initproc,
834 	    NULL, 0, NULL);
835 	if (error)
836 		panic("cannot fork init: %d\n", error);
837 	KASSERT(initproc->p_pid == 1, ("create_init: initproc->p_pid != 1"));
838 	/* divorce init's credentials from the kernel's */
839 	newcred = crget();
840 	sx_xlock(&proctree_lock);
841 	PROC_LOCK(initproc);
842 	initproc->p_flag |= P_SYSTEM | P_INMEM;
843 	initproc->p_treeflag |= P_TREE_REAPER;
844 	LIST_INSERT_HEAD(&initproc->p_reaplist, &proc0, p_reapsibling);
845 	oldcred = initproc->p_ucred;
846 	crcopy(newcred, oldcred);
847 #ifdef MAC
848 	mac_cred_create_init(newcred);
849 #endif
850 #ifdef AUDIT
851 	audit_cred_proc1(newcred);
852 #endif
853 	proc_set_cred(initproc, newcred);
854 	td = FIRST_THREAD_IN_PROC(initproc);
855 	crfree(td->td_ucred);
856 	td->td_ucred = crhold(initproc->p_ucred);
857 	PROC_UNLOCK(initproc);
858 	sx_xunlock(&proctree_lock);
859 	crfree(oldcred);
860 	cpu_set_fork_handler(FIRST_THREAD_IN_PROC(initproc), start_init, NULL);
861 }
862 SYSINIT(init, SI_SUB_CREATE_INIT, SI_ORDER_FIRST, create_init, NULL);
863 
864 /*
865  * Make it runnable now.
866  */
867 static void
868 kick_init(const void *udata __unused)
869 {
870 	struct thread *td;
871 
872 	td = FIRST_THREAD_IN_PROC(initproc);
873 	thread_lock(td);
874 	TD_SET_CAN_RUN(td);
875 	sched_add(td, SRQ_BORING);
876 	thread_unlock(td);
877 }
878 SYSINIT(kickinit, SI_SUB_KTHREAD_INIT, SI_ORDER_MIDDLE, kick_init, NULL);
879