xref: /freebsd/sys/kern/init_main.c (revision 730cecb05aaf016ac52ef7cfc691ccec3a0408cd)
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/loginclass.h>
59 #include <sys/mount.h>
60 #include <sys/mutex.h>
61 #include <sys/syscallsubr.h>
62 #include <sys/sysctl.h>
63 #include <sys/proc.h>
64 #include <sys/racct.h>
65 #include <sys/resourcevar.h>
66 #include <sys/systm.h>
67 #include <sys/signalvar.h>
68 #include <sys/vnode.h>
69 #include <sys/sysent.h>
70 #include <sys/reboot.h>
71 #include <sys/sched.h>
72 #include <sys/sx.h>
73 #include <sys/sysproto.h>
74 #include <sys/vmmeter.h>
75 #include <sys/unistd.h>
76 #include <sys/malloc.h>
77 #include <sys/conf.h>
78 #include <sys/cpuset.h>
79 
80 #include <machine/cpu.h>
81 
82 #include <security/audit/audit.h>
83 #include <security/mac/mac_framework.h>
84 
85 #include <vm/vm.h>
86 #include <vm/vm_param.h>
87 #include <vm/pmap.h>
88 #include <vm/vm_map.h>
89 #include <sys/copyright.h>
90 
91 #include <ddb/ddb.h>
92 #include <ddb/db_sym.h>
93 
94 void mi_startup(void);				/* Should be elsewhere */
95 
96 /* Components of the first process -- never freed. */
97 static struct session session0;
98 static struct pgrp pgrp0;
99 struct	proc proc0;
100 struct	thread thread0 __aligned(16);
101 struct	vmspace vmspace0;
102 struct	proc *initproc;
103 
104 int	boothowto = 0;		/* initialized so that it can be patched */
105 SYSCTL_INT(_debug, OID_AUTO, boothowto, CTLFLAG_RD, &boothowto, 0,
106 	"Boot control flags, passed from loader");
107 int	bootverbose;
108 SYSCTL_INT(_debug, OID_AUTO, bootverbose, CTLFLAG_RW, &bootverbose, 0,
109 	"Control the output of verbose kernel messages");
110 
111 /*
112  * This ensures that there is at least one entry so that the sysinit_set
113  * symbol is not undefined.  A sybsystem ID of SI_SUB_DUMMY is never
114  * executed.
115  */
116 SYSINIT(placeholder, SI_SUB_DUMMY, SI_ORDER_ANY, NULL, NULL);
117 
118 /*
119  * The sysinit table itself.  Items are checked off as the are run.
120  * If we want to register new sysinit types, add them to newsysinit.
121  */
122 SET_DECLARE(sysinit_set, struct sysinit);
123 struct sysinit **sysinit, **sysinit_end;
124 struct sysinit **newsysinit, **newsysinit_end;
125 
126 /*
127  * Merge a new sysinit set into the current set, reallocating it if
128  * necessary.  This can only be called after malloc is running.
129  */
130 void
131 sysinit_add(struct sysinit **set, struct sysinit **set_end)
132 {
133 	struct sysinit **newset;
134 	struct sysinit **sipp;
135 	struct sysinit **xipp;
136 	int count;
137 
138 	count = set_end - set;
139 	if (newsysinit)
140 		count += newsysinit_end - newsysinit;
141 	else
142 		count += sysinit_end - sysinit;
143 	newset = malloc(count * sizeof(*sipp), M_TEMP, M_NOWAIT);
144 	if (newset == NULL)
145 		panic("cannot malloc for sysinit");
146 	xipp = newset;
147 	if (newsysinit)
148 		for (sipp = newsysinit; sipp < newsysinit_end; sipp++)
149 			*xipp++ = *sipp;
150 	else
151 		for (sipp = sysinit; sipp < sysinit_end; sipp++)
152 			*xipp++ = *sipp;
153 	for (sipp = set; sipp < set_end; sipp++)
154 		*xipp++ = *sipp;
155 	if (newsysinit)
156 		free(newsysinit, M_TEMP);
157 	newsysinit = newset;
158 	newsysinit_end = newset + count;
159 }
160 
161 #if defined (DDB) && defined(VERBOSE_SYSINIT)
162 static const char *
163 symbol_name(vm_offset_t va, db_strategy_t strategy)
164 {
165 	const char *name;
166 	c_db_sym_t sym;
167 	db_expr_t  offset;
168 
169 	if (va == 0)
170 		return (NULL);
171 	sym = db_search_symbol(va, strategy, &offset);
172 	if (offset != 0)
173 		return (NULL);
174 	db_symbol_values(sym, &name, NULL);
175 	return (name);
176 }
177 #endif
178 
179 /*
180  * System startup; initialize the world, create process 0, mount root
181  * filesystem, and fork to create init and pagedaemon.  Most of the
182  * hard work is done in the lower-level initialization routines including
183  * startup(), which does memory initialization and autoconfiguration.
184  *
185  * This allows simple addition of new kernel subsystems that require
186  * boot time initialization.  It also allows substitution of subsystem
187  * (for instance, a scheduler, kernel profiler, or VM system) by object
188  * module.  Finally, it allows for optional "kernel threads".
189  */
190 void
191 mi_startup(void)
192 {
193 
194 	register struct sysinit **sipp;		/* system initialization*/
195 	register struct sysinit **xipp;		/* interior loop of sort*/
196 	register struct sysinit *save;		/* bubble*/
197 
198 #if defined(VERBOSE_SYSINIT)
199 	int last;
200 	int verbose;
201 #endif
202 
203 	if (boothowto & RB_VERBOSE)
204 		bootverbose++;
205 
206 	if (sysinit == NULL) {
207 		sysinit = SET_BEGIN(sysinit_set);
208 		sysinit_end = SET_LIMIT(sysinit_set);
209 	}
210 
211 restart:
212 	/*
213 	 * Perform a bubble sort of the system initialization objects by
214 	 * their subsystem (primary key) and order (secondary key).
215 	 */
216 	for (sipp = sysinit; sipp < sysinit_end; sipp++) {
217 		for (xipp = sipp + 1; xipp < sysinit_end; xipp++) {
218 			if ((*sipp)->subsystem < (*xipp)->subsystem ||
219 			     ((*sipp)->subsystem == (*xipp)->subsystem &&
220 			      (*sipp)->order <= (*xipp)->order))
221 				continue;	/* skip*/
222 			save = *sipp;
223 			*sipp = *xipp;
224 			*xipp = save;
225 		}
226 	}
227 
228 #if defined(VERBOSE_SYSINIT)
229 	last = SI_SUB_COPYRIGHT;
230 	verbose = 0;
231 #if !defined(DDB)
232 	printf("VERBOSE_SYSINIT: DDB not enabled, symbol lookups disabled.\n");
233 #endif
234 #endif
235 
236 	/*
237 	 * Traverse the (now) ordered list of system initialization tasks.
238 	 * Perform each task, and continue on to the next task.
239 	 *
240 	 * The last item on the list is expected to be the scheduler,
241 	 * which will not return.
242 	 */
243 	for (sipp = sysinit; sipp < sysinit_end; sipp++) {
244 
245 		if ((*sipp)->subsystem == SI_SUB_DUMMY)
246 			continue;	/* skip dummy task(s)*/
247 
248 		if ((*sipp)->subsystem == SI_SUB_DONE)
249 			continue;
250 
251 #if defined(VERBOSE_SYSINIT)
252 		if ((*sipp)->subsystem > last) {
253 			verbose = 1;
254 			last = (*sipp)->subsystem;
255 			printf("subsystem %x\n", last);
256 		}
257 		if (verbose) {
258 #if defined(DDB)
259 			const char *func, *data;
260 
261 			func = symbol_name((vm_offset_t)(*sipp)->func,
262 			    DB_STGY_PROC);
263 			data = symbol_name((vm_offset_t)(*sipp)->udata,
264 			    DB_STGY_ANY);
265 			if (func != NULL && data != NULL)
266 				printf("   %s(&%s)... ", func, data);
267 			else if (func != NULL)
268 				printf("   %s(%p)... ", func, (*sipp)->udata);
269 			else
270 #endif
271 				printf("   %p(%p)... ", (*sipp)->func,
272 				    (*sipp)->udata);
273 		}
274 #endif
275 
276 		/* Call function */
277 		(*((*sipp)->func))((*sipp)->udata);
278 
279 #if defined(VERBOSE_SYSINIT)
280 		if (verbose)
281 			printf("done.\n");
282 #endif
283 
284 		/* Check off the one we're just done */
285 		(*sipp)->subsystem = SI_SUB_DONE;
286 
287 		/* Check if we've installed more sysinit items via KLD */
288 		if (newsysinit != NULL) {
289 			if (sysinit != SET_BEGIN(sysinit_set))
290 				free(sysinit, M_TEMP);
291 			sysinit = newsysinit;
292 			sysinit_end = newsysinit_end;
293 			newsysinit = NULL;
294 			newsysinit_end = NULL;
295 			goto restart;
296 		}
297 	}
298 
299 	panic("Shouldn't get here!");
300 	/* NOTREACHED*/
301 }
302 
303 
304 /*
305  ***************************************************************************
306  ****
307  **** The following SYSINIT's belong elsewhere, but have not yet
308  **** been moved.
309  ****
310  ***************************************************************************
311  */
312 static void
313 print_caddr_t(void *data)
314 {
315 	printf("%s", (char *)data);
316 }
317 
318 static void
319 print_version(void *data __unused)
320 {
321 	int len;
322 
323 	/* Strip a trailing newline from version. */
324 	len = strlen(version);
325 	while (len > 0 && version[len - 1] == '\n')
326 		len--;
327 	printf("%.*s %s\n", len, version, machine);
328 	printf("%s\n", compiler_version);
329 }
330 
331 SYSINIT(announce, SI_SUB_COPYRIGHT, SI_ORDER_FIRST, print_caddr_t,
332     copyright);
333 SYSINIT(trademark, SI_SUB_COPYRIGHT, SI_ORDER_SECOND, print_caddr_t,
334     trademark);
335 SYSINIT(version, SI_SUB_COPYRIGHT, SI_ORDER_THIRD, print_version, NULL);
336 
337 #ifdef WITNESS
338 static char wit_warn[] =
339      "WARNING: WITNESS option enabled, expect reduced performance.\n";
340 SYSINIT(witwarn, SI_SUB_COPYRIGHT, SI_ORDER_THIRD + 1,
341    print_caddr_t, wit_warn);
342 SYSINIT(witwarn2, SI_SUB_RUN_SCHEDULER, SI_ORDER_THIRD + 1,
343    print_caddr_t, wit_warn);
344 #endif
345 
346 #ifdef DIAGNOSTIC
347 static char diag_warn[] =
348      "WARNING: DIAGNOSTIC option enabled, expect reduced performance.\n";
349 SYSINIT(diagwarn, SI_SUB_COPYRIGHT, SI_ORDER_THIRD + 2,
350     print_caddr_t, diag_warn);
351 SYSINIT(diagwarn2, SI_SUB_RUN_SCHEDULER, SI_ORDER_THIRD + 2,
352     print_caddr_t, diag_warn);
353 #endif
354 
355 static int
356 null_fetch_syscall_args(struct thread *td __unused,
357     struct syscall_args *sa __unused)
358 {
359 
360 	panic("null_fetch_syscall_args");
361 }
362 
363 static void
364 null_set_syscall_retval(struct thread *td __unused, int error __unused)
365 {
366 
367 	panic("null_set_syscall_retval");
368 }
369 
370 struct sysentvec null_sysvec = {
371 	.sv_size	= 0,
372 	.sv_table	= NULL,
373 	.sv_mask	= 0,
374 	.sv_sigsize	= 0,
375 	.sv_sigtbl	= NULL,
376 	.sv_errsize	= 0,
377 	.sv_errtbl	= NULL,
378 	.sv_transtrap	= NULL,
379 	.sv_fixup	= NULL,
380 	.sv_sendsig	= NULL,
381 	.sv_sigcode	= NULL,
382 	.sv_szsigcode	= NULL,
383 	.sv_prepsyscall	= NULL,
384 	.sv_name	= "null",
385 	.sv_coredump	= NULL,
386 	.sv_imgact_try	= NULL,
387 	.sv_minsigstksz	= 0,
388 	.sv_pagesize	= PAGE_SIZE,
389 	.sv_minuser	= VM_MIN_ADDRESS,
390 	.sv_maxuser	= VM_MAXUSER_ADDRESS,
391 	.sv_usrstack	= USRSTACK,
392 	.sv_psstrings	= PS_STRINGS,
393 	.sv_stackprot	= VM_PROT_ALL,
394 	.sv_copyout_strings	= NULL,
395 	.sv_setregs	= NULL,
396 	.sv_fixlimit	= NULL,
397 	.sv_maxssiz	= NULL,
398 	.sv_flags	= 0,
399 	.sv_set_syscall_retval = null_set_syscall_retval,
400 	.sv_fetch_syscall_args = null_fetch_syscall_args,
401 	.sv_syscallnames = NULL,
402 	.sv_schedtail	= NULL,
403 };
404 
405 /*
406  ***************************************************************************
407  ****
408  **** The two following SYSINIT's are proc0 specific glue code.  I am not
409  **** convinced that they can not be safely combined, but their order of
410  **** operation has been maintained as the same as the original init_main.c
411  **** for right now.
412  ****
413  **** These probably belong in init_proc.c or kern_proc.c, since they
414  **** deal with proc0 (the fork template process).
415  ****
416  ***************************************************************************
417  */
418 /* ARGSUSED*/
419 static void
420 proc0_init(void *dummy __unused)
421 {
422 	struct proc *p;
423 	struct thread *td;
424 	vm_paddr_t pageablemem;
425 	int i;
426 
427 	GIANT_REQUIRED;
428 	p = &proc0;
429 	td = &thread0;
430 
431 	/*
432 	 * Initialize magic number and osrel.
433 	 */
434 	p->p_magic = P_MAGIC;
435 	p->p_osrel = osreldate;
436 
437 	/*
438 	 * Initialize thread and process structures.
439 	 */
440 	procinit();	/* set up proc zone */
441 	threadinit();	/* set up UMA zones */
442 
443 	/*
444 	 * Initialise scheduler resources.
445 	 * Add scheduler specific parts to proc, thread as needed.
446 	 */
447 	schedinit();	/* scheduler gets its house in order */
448 	/*
449 	 * Initialize sleep queue hash table
450 	 */
451 	sleepinit();
452 
453 	/*
454 	 * additional VM structures
455 	 */
456 	vm_init2();
457 
458 	/*
459 	 * Create process 0 (the swapper).
460 	 */
461 	LIST_INSERT_HEAD(&allproc, p, p_list);
462 	LIST_INSERT_HEAD(PIDHASH(0), p, p_hash);
463 	mtx_init(&pgrp0.pg_mtx, "process group", NULL, MTX_DEF | MTX_DUPOK);
464 	p->p_pgrp = &pgrp0;
465 	LIST_INSERT_HEAD(PGRPHASH(0), &pgrp0, pg_hash);
466 	LIST_INIT(&pgrp0.pg_members);
467 	LIST_INSERT_HEAD(&pgrp0.pg_members, p, p_pglist);
468 
469 	pgrp0.pg_session = &session0;
470 	mtx_init(&session0.s_mtx, "session", NULL, MTX_DEF);
471 	refcount_init(&session0.s_count, 1);
472 	session0.s_leader = p;
473 
474 	p->p_sysent = &null_sysvec;
475 	p->p_flag = P_SYSTEM | P_INMEM;
476 	p->p_state = PRS_NORMAL;
477 	knlist_init_mtx(&p->p_klist, &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 = 0;
491 	td->td_flags = TDF_INMEM;
492 	td->td_pflags = TDP_KTHREAD;
493 	td->td_cpuset = cpuset_thread0();
494 	prison0.pr_cpuset = cpuset_ref(td->td_cpuset);
495 	p->p_peers = 0;
496 	p->p_leader = p;
497 
498 
499 	strncpy(p->p_comm, "kernel", sizeof (p->p_comm));
500 	strncpy(td->td_name, "swapper", sizeof (td->td_name));
501 
502 	callout_init_mtx(&p->p_itcallout, &p->p_mtx, 0);
503 	callout_init_mtx(&p->p_limco, &p->p_mtx, 0);
504 	callout_init(&td->td_slpcallout, CALLOUT_MPSAFE);
505 
506 	/* Create credentials. */
507 	p->p_ucred = crget();
508 	p->p_ucred->cr_ngroups = 1;	/* group 0 */
509 	p->p_ucred->cr_uidinfo = uifind(0);
510 	p->p_ucred->cr_ruidinfo = uifind(0);
511 	p->p_ucred->cr_prison = &prison0;
512 	p->p_ucred->cr_loginclass = loginclass_find("default");
513 #ifdef AUDIT
514 	audit_cred_kproc0(p->p_ucred);
515 #endif
516 #ifdef MAC
517 	mac_cred_create_swapper(p->p_ucred);
518 #endif
519 	td->td_ucred = crhold(p->p_ucred);
520 
521 	/* Create sigacts. */
522 	p->p_sigacts = sigacts_alloc();
523 
524 	/* Initialize signal state for process 0. */
525 	siginit(&proc0);
526 
527 	/* Create the file descriptor table. */
528 	p->p_fd = fdinit(NULL);
529 	p->p_fdtol = NULL;
530 
531 	/* Create the limits structures. */
532 	p->p_limit = lim_alloc();
533 	for (i = 0; i < RLIM_NLIMITS; i++)
534 		p->p_limit->pl_rlimit[i].rlim_cur =
535 		    p->p_limit->pl_rlimit[i].rlim_max = RLIM_INFINITY;
536 	p->p_limit->pl_rlimit[RLIMIT_NOFILE].rlim_cur =
537 	    p->p_limit->pl_rlimit[RLIMIT_NOFILE].rlim_max = maxfiles;
538 	p->p_limit->pl_rlimit[RLIMIT_NPROC].rlim_cur =
539 	    p->p_limit->pl_rlimit[RLIMIT_NPROC].rlim_max = maxproc;
540 	p->p_limit->pl_rlimit[RLIMIT_DATA].rlim_cur = dfldsiz;
541 	p->p_limit->pl_rlimit[RLIMIT_DATA].rlim_max = maxdsiz;
542 	p->p_limit->pl_rlimit[RLIMIT_STACK].rlim_cur = dflssiz;
543 	p->p_limit->pl_rlimit[RLIMIT_STACK].rlim_max = maxssiz;
544 	/* Cast to avoid overflow on i386/PAE. */
545 	pageablemem = ptoa((vm_paddr_t)cnt.v_free_count);
546 	p->p_limit->pl_rlimit[RLIMIT_RSS].rlim_cur =
547 	    p->p_limit->pl_rlimit[RLIMIT_RSS].rlim_max = pageablemem;
548 	p->p_limit->pl_rlimit[RLIMIT_MEMLOCK].rlim_cur = pageablemem / 3;
549 	p->p_limit->pl_rlimit[RLIMIT_MEMLOCK].rlim_max = pageablemem;
550 	p->p_cpulimit = RLIM_INFINITY;
551 
552 	/* Initialize resource accounting structures. */
553 	racct_create(&p->p_racct);
554 
555 	p->p_stats = pstats_alloc();
556 
557 	/* Allocate a prototype map so we have something to fork. */
558 	pmap_pinit0(vmspace_pmap(&vmspace0));
559 	p->p_vmspace = &vmspace0;
560 	vmspace0.vm_refcnt = 1;
561 
562 	/*
563 	 * proc0 is not expected to enter usermode, so there is no special
564 	 * handling for sv_minuser here, like is done for exec_new_vmspace().
565 	 */
566 	vm_map_init(&vmspace0.vm_map, vmspace_pmap(&vmspace0),
567 	    p->p_sysent->sv_minuser, p->p_sysent->sv_maxuser);
568 
569 	/*
570 	 * Call the init and ctor for the new thread and proc.  We wait
571 	 * to do this until all other structures are fairly sane.
572 	 */
573 	EVENTHANDLER_INVOKE(process_init, p);
574 	EVENTHANDLER_INVOKE(thread_init, td);
575 	EVENTHANDLER_INVOKE(process_ctor, p);
576 	EVENTHANDLER_INVOKE(thread_ctor, td);
577 
578 	/*
579 	 * Charge root for one process.
580 	 */
581 	(void)chgproccnt(p->p_ucred->cr_ruidinfo, 1, 0);
582 	PROC_LOCK(p);
583 	racct_add_force(p, RACCT_NPROC, 1);
584 	PROC_UNLOCK(p);
585 }
586 SYSINIT(p0init, SI_SUB_INTRINSIC, SI_ORDER_FIRST, proc0_init, NULL);
587 
588 /* ARGSUSED*/
589 static void
590 proc0_post(void *dummy __unused)
591 {
592 	struct timespec ts;
593 	struct proc *p;
594 	struct rusage ru;
595 	struct thread *td;
596 
597 	/*
598 	 * Now we can look at the time, having had a chance to verify the
599 	 * time from the filesystem.  Pretend that proc0 started now.
600 	 */
601 	sx_slock(&allproc_lock);
602 	FOREACH_PROC_IN_SYSTEM(p) {
603 		microuptime(&p->p_stats->p_start);
604 		PROC_SLOCK(p);
605 		rufetch(p, &ru);	/* Clears thread stats */
606 		PROC_SUNLOCK(p);
607 		p->p_rux.rux_runtime = 0;
608 		p->p_rux.rux_uticks = 0;
609 		p->p_rux.rux_sticks = 0;
610 		p->p_rux.rux_iticks = 0;
611 		FOREACH_THREAD_IN_PROC(p, td) {
612 			td->td_runtime = 0;
613 		}
614 	}
615 	sx_sunlock(&allproc_lock);
616 	PCPU_SET(switchtime, cpu_ticks());
617 	PCPU_SET(switchticks, ticks);
618 
619 	/*
620 	 * Give the ``random'' number generator a thump.
621 	 */
622 	nanotime(&ts);
623 	srandom(ts.tv_sec ^ ts.tv_nsec);
624 }
625 SYSINIT(p0post, SI_SUB_INTRINSIC_POST, SI_ORDER_FIRST, proc0_post, NULL);
626 
627 static void
628 random_init(void *dummy __unused)
629 {
630 
631 	/*
632 	 * After CPU has been started we have some randomness on most
633 	 * platforms via get_cyclecount().  For platforms that don't
634 	 * we will reseed random(9) in proc0_post() as well.
635 	 */
636 	srandom(get_cyclecount());
637 }
638 SYSINIT(random, SI_SUB_RANDOM, SI_ORDER_FIRST, random_init, NULL);
639 
640 /*
641  ***************************************************************************
642  ****
643  **** The following SYSINIT's and glue code should be moved to the
644  **** respective files on a per subsystem basis.
645  ****
646  ***************************************************************************
647  */
648 
649 
650 /*
651  ***************************************************************************
652  ****
653  **** The following code probably belongs in another file, like
654  **** kern/init_init.c.
655  ****
656  ***************************************************************************
657  */
658 
659 /*
660  * List of paths to try when searching for "init".
661  */
662 static char init_path[MAXPATHLEN] =
663 #ifdef	INIT_PATH
664     __XSTRING(INIT_PATH);
665 #else
666     "/sbin/init:/sbin/oinit:/sbin/init.bak:/rescue/init";
667 #endif
668 SYSCTL_STRING(_kern, OID_AUTO, init_path, CTLFLAG_RD, init_path, 0,
669 	"Path used to search the init process");
670 
671 /*
672  * Shutdown timeout of init(8).
673  * Unused within kernel, but used to control init(8), hence do not remove.
674  */
675 #ifndef INIT_SHUTDOWN_TIMEOUT
676 #define INIT_SHUTDOWN_TIMEOUT 120
677 #endif
678 static int init_shutdown_timeout = INIT_SHUTDOWN_TIMEOUT;
679 SYSCTL_INT(_kern, OID_AUTO, init_shutdown_timeout,
680 	CTLFLAG_RW, &init_shutdown_timeout, 0, "Shutdown timeout of init(8). "
681 	"Unused within kernel, but used to control init(8)");
682 
683 /*
684  * Start the initial user process; try exec'ing each pathname in init_path.
685  * The program is invoked with one argument containing the boot flags.
686  */
687 static void
688 start_init(void *dummy)
689 {
690 	vm_offset_t addr;
691 	struct execve_args args;
692 	int options, error;
693 	char *var, *path, *next, *s;
694 	char *ucp, **uap, *arg0, *arg1;
695 	struct thread *td;
696 	struct proc *p;
697 
698 	mtx_lock(&Giant);
699 
700 	GIANT_REQUIRED;
701 
702 	td = curthread;
703 	p = td->td_proc;
704 
705 	vfs_mountroot();
706 
707 	/*
708 	 * Need just enough stack to hold the faked-up "execve()" arguments.
709 	 */
710 	addr = p->p_sysent->sv_usrstack - PAGE_SIZE;
711 	if (vm_map_find(&p->p_vmspace->vm_map, NULL, 0, &addr, PAGE_SIZE,
712 			FALSE, VM_PROT_ALL, VM_PROT_ALL, 0) != 0)
713 		panic("init: couldn't allocate argument space");
714 	p->p_vmspace->vm_maxsaddr = (caddr_t)addr;
715 	p->p_vmspace->vm_ssize = 1;
716 
717 	if ((var = getenv("init_path")) != NULL) {
718 		strlcpy(init_path, var, sizeof(init_path));
719 		freeenv(var);
720 	}
721 
722 	for (path = init_path; *path != '\0'; path = next) {
723 		while (*path == ':')
724 			path++;
725 		if (*path == '\0')
726 			break;
727 		for (next = path; *next != '\0' && *next != ':'; next++)
728 			/* nothing */ ;
729 		if (bootverbose)
730 			printf("start_init: trying %.*s\n", (int)(next - path),
731 			    path);
732 
733 		/*
734 		 * Move out the boot flag argument.
735 		 */
736 		options = 0;
737 		ucp = (char *)p->p_sysent->sv_usrstack;
738 		(void)subyte(--ucp, 0);		/* trailing zero */
739 		if (boothowto & RB_SINGLE) {
740 			(void)subyte(--ucp, 's');
741 			options = 1;
742 		}
743 #ifdef notyet
744                 if (boothowto & RB_FASTBOOT) {
745 			(void)subyte(--ucp, 'f');
746 			options = 1;
747 		}
748 #endif
749 
750 #ifdef BOOTCDROM
751 		(void)subyte(--ucp, 'C');
752 		options = 1;
753 #endif
754 
755 		if (options == 0)
756 			(void)subyte(--ucp, '-');
757 		(void)subyte(--ucp, '-');		/* leading hyphen */
758 		arg1 = ucp;
759 
760 		/*
761 		 * Move out the file name (also arg 0).
762 		 */
763 		(void)subyte(--ucp, 0);
764 		for (s = next - 1; s >= path; s--)
765 			(void)subyte(--ucp, *s);
766 		arg0 = ucp;
767 
768 		/*
769 		 * Move out the arg pointers.
770 		 */
771 		uap = (char **)((intptr_t)ucp & ~(sizeof(intptr_t)-1));
772 		(void)suword((caddr_t)--uap, (long)0);	/* terminator */
773 		(void)suword((caddr_t)--uap, (long)(intptr_t)arg1);
774 		(void)suword((caddr_t)--uap, (long)(intptr_t)arg0);
775 
776 		/*
777 		 * Point at the arguments.
778 		 */
779 		args.fname = arg0;
780 		args.argv = uap;
781 		args.envv = NULL;
782 
783 		/*
784 		 * Now try to exec the program.  If can't for any reason
785 		 * other than it doesn't exist, complain.
786 		 *
787 		 * Otherwise, return via fork_trampoline() all the way
788 		 * to user mode as init!
789 		 */
790 		if ((error = sys_execve(td, &args)) == 0) {
791 			mtx_unlock(&Giant);
792 			return;
793 		}
794 		if (error != ENOENT)
795 			printf("exec %.*s: error %d\n", (int)(next - path),
796 			    path, error);
797 	}
798 	printf("init: not found in path %s\n", init_path);
799 	panic("no init");
800 }
801 
802 /*
803  * Like kproc_create(), but runs in it's own address space.
804  * We do this early to reserve pid 1.
805  *
806  * Note special case - do not make it runnable yet.  Other work
807  * in progress will change this more.
808  */
809 static void
810 create_init(const void *udata __unused)
811 {
812 	struct ucred *newcred, *oldcred;
813 	int error;
814 
815 	error = fork1(&thread0, RFFDG | RFPROC | RFSTOPPED, 0, &initproc,
816 	    NULL, 0);
817 	if (error)
818 		panic("cannot fork init: %d\n", error);
819 	KASSERT(initproc->p_pid == 1, ("create_init: initproc->p_pid != 1"));
820 	/* divorce init's credentials from the kernel's */
821 	newcred = crget();
822 	PROC_LOCK(initproc);
823 	initproc->p_flag |= P_SYSTEM | P_INMEM;
824 	oldcred = initproc->p_ucred;
825 	crcopy(newcred, oldcred);
826 #ifdef MAC
827 	mac_cred_create_init(newcred);
828 #endif
829 #ifdef AUDIT
830 	audit_cred_proc1(newcred);
831 #endif
832 	initproc->p_ucred = newcred;
833 	PROC_UNLOCK(initproc);
834 	crfree(oldcred);
835 	cred_update_thread(FIRST_THREAD_IN_PROC(initproc));
836 	cpu_set_fork_handler(FIRST_THREAD_IN_PROC(initproc), start_init, NULL);
837 }
838 SYSINIT(init, SI_SUB_CREATE_INIT, SI_ORDER_FIRST, create_init, NULL);
839 
840 /*
841  * Make it runnable now.
842  */
843 static void
844 kick_init(const void *udata __unused)
845 {
846 	struct thread *td;
847 
848 	td = FIRST_THREAD_IN_PROC(initproc);
849 	thread_lock(td);
850 	TD_SET_CAN_RUN(td);
851 	sched_add(td, SRQ_BORING);
852 	thread_unlock(td);
853 }
854 SYSINIT(kickinit, SI_SUB_KTHREAD_INIT, SI_ORDER_FIRST, kick_init, NULL);
855