xref: /freebsd/sys/kern/init_main.c (revision ab0b9f6b3073e6c4d1dfbf07444d7db67a189a96)
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 #ifndef BOOTHOWTO
105 #define	BOOTHOWTO	0
106 #endif
107 int	boothowto = BOOTHOWTO;	/* initialized so that it can be patched */
108 SYSCTL_INT(_debug, OID_AUTO, boothowto, CTLFLAG_RD, &boothowto, 0,
109 	"Boot control flags, passed from loader");
110 
111 #ifndef BOOTVERBOSE
112 #define	BOOTVERBOSE	0
113 #endif
114 int	bootverbose = BOOTVERBOSE;
115 SYSCTL_INT(_debug, OID_AUTO, bootverbose, CTLFLAG_RW, &bootverbose, 0,
116 	"Control the output of verbose kernel messages");
117 
118 /*
119  * This ensures that there is at least one entry so that the sysinit_set
120  * symbol is not undefined.  A sybsystem ID of SI_SUB_DUMMY is never
121  * executed.
122  */
123 SYSINIT(placeholder, SI_SUB_DUMMY, SI_ORDER_ANY, NULL, NULL);
124 
125 /*
126  * The sysinit table itself.  Items are checked off as the are run.
127  * If we want to register new sysinit types, add them to newsysinit.
128  */
129 SET_DECLARE(sysinit_set, struct sysinit);
130 struct sysinit **sysinit, **sysinit_end;
131 struct sysinit **newsysinit, **newsysinit_end;
132 
133 /*
134  * Merge a new sysinit set into the current set, reallocating it if
135  * necessary.  This can only be called after malloc is running.
136  */
137 void
138 sysinit_add(struct sysinit **set, struct sysinit **set_end)
139 {
140 	struct sysinit **newset;
141 	struct sysinit **sipp;
142 	struct sysinit **xipp;
143 	int count;
144 
145 	count = set_end - set;
146 	if (newsysinit)
147 		count += newsysinit_end - newsysinit;
148 	else
149 		count += sysinit_end - sysinit;
150 	newset = malloc(count * sizeof(*sipp), M_TEMP, M_NOWAIT);
151 	if (newset == NULL)
152 		panic("cannot malloc for sysinit");
153 	xipp = newset;
154 	if (newsysinit)
155 		for (sipp = newsysinit; sipp < newsysinit_end; sipp++)
156 			*xipp++ = *sipp;
157 	else
158 		for (sipp = sysinit; sipp < sysinit_end; sipp++)
159 			*xipp++ = *sipp;
160 	for (sipp = set; sipp < set_end; sipp++)
161 		*xipp++ = *sipp;
162 	if (newsysinit)
163 		free(newsysinit, M_TEMP);
164 	newsysinit = newset;
165 	newsysinit_end = newset + count;
166 }
167 
168 #if defined (DDB) && defined(VERBOSE_SYSINIT)
169 static const char *
170 symbol_name(vm_offset_t va, db_strategy_t strategy)
171 {
172 	const char *name;
173 	c_db_sym_t sym;
174 	db_expr_t  offset;
175 
176 	if (va == 0)
177 		return (NULL);
178 	sym = db_search_symbol(va, strategy, &offset);
179 	if (offset != 0)
180 		return (NULL);
181 	db_symbol_values(sym, &name, NULL);
182 	return (name);
183 }
184 #endif
185 
186 /*
187  * System startup; initialize the world, create process 0, mount root
188  * filesystem, and fork to create init and pagedaemon.  Most of the
189  * hard work is done in the lower-level initialization routines including
190  * startup(), which does memory initialization and autoconfiguration.
191  *
192  * This allows simple addition of new kernel subsystems that require
193  * boot time initialization.  It also allows substitution of subsystem
194  * (for instance, a scheduler, kernel profiler, or VM system) by object
195  * module.  Finally, it allows for optional "kernel threads".
196  */
197 void
198 mi_startup(void)
199 {
200 
201 	register struct sysinit **sipp;		/* system initialization*/
202 	register struct sysinit **xipp;		/* interior loop of sort*/
203 	register struct sysinit *save;		/* bubble*/
204 
205 #if defined(VERBOSE_SYSINIT)
206 	int last;
207 	int verbose;
208 #endif
209 
210 	if (boothowto & RB_VERBOSE)
211 		bootverbose++;
212 
213 	if (sysinit == NULL) {
214 		sysinit = SET_BEGIN(sysinit_set);
215 		sysinit_end = SET_LIMIT(sysinit_set);
216 	}
217 
218 restart:
219 	/*
220 	 * Perform a bubble sort of the system initialization objects by
221 	 * their subsystem (primary key) and order (secondary key).
222 	 */
223 	for (sipp = sysinit; sipp < sysinit_end; sipp++) {
224 		for (xipp = sipp + 1; xipp < sysinit_end; xipp++) {
225 			if ((*sipp)->subsystem < (*xipp)->subsystem ||
226 			     ((*sipp)->subsystem == (*xipp)->subsystem &&
227 			      (*sipp)->order <= (*xipp)->order))
228 				continue;	/* skip*/
229 			save = *sipp;
230 			*sipp = *xipp;
231 			*xipp = save;
232 		}
233 	}
234 
235 #if defined(VERBOSE_SYSINIT)
236 	last = SI_SUB_COPYRIGHT;
237 	verbose = 0;
238 #if !defined(DDB)
239 	printf("VERBOSE_SYSINIT: DDB not enabled, symbol lookups disabled.\n");
240 #endif
241 #endif
242 
243 	/*
244 	 * Traverse the (now) ordered list of system initialization tasks.
245 	 * Perform each task, and continue on to the next task.
246 	 */
247 	for (sipp = sysinit; sipp < sysinit_end; sipp++) {
248 
249 		if ((*sipp)->subsystem == SI_SUB_DUMMY)
250 			continue;	/* skip dummy task(s)*/
251 
252 		if ((*sipp)->subsystem == SI_SUB_DONE)
253 			continue;
254 
255 #if defined(VERBOSE_SYSINIT)
256 		if ((*sipp)->subsystem > last) {
257 			verbose = 1;
258 			last = (*sipp)->subsystem;
259 			printf("subsystem %x\n", last);
260 		}
261 		if (verbose) {
262 #if defined(DDB)
263 			const char *func, *data;
264 
265 			func = symbol_name((vm_offset_t)(*sipp)->func,
266 			    DB_STGY_PROC);
267 			data = symbol_name((vm_offset_t)(*sipp)->udata,
268 			    DB_STGY_ANY);
269 			if (func != NULL && data != NULL)
270 				printf("   %s(&%s)... ", func, data);
271 			else if (func != NULL)
272 				printf("   %s(%p)... ", func, (*sipp)->udata);
273 			else
274 #endif
275 				printf("   %p(%p)... ", (*sipp)->func,
276 				    (*sipp)->udata);
277 		}
278 #endif
279 
280 		/* Call function */
281 		(*((*sipp)->func))((*sipp)->udata);
282 
283 #if defined(VERBOSE_SYSINIT)
284 		if (verbose)
285 			printf("done.\n");
286 #endif
287 
288 		/* Check off the one we're just done */
289 		(*sipp)->subsystem = SI_SUB_DONE;
290 
291 		/* Check if we've installed more sysinit items via KLD */
292 		if (newsysinit != NULL) {
293 			if (sysinit != SET_BEGIN(sysinit_set))
294 				free(sysinit, M_TEMP);
295 			sysinit = newsysinit;
296 			sysinit_end = newsysinit_end;
297 			newsysinit = NULL;
298 			newsysinit_end = NULL;
299 			goto restart;
300 		}
301 	}
302 
303 	mtx_assert(&Giant, MA_OWNED | MA_NOTRECURSED);
304 	mtx_unlock(&Giant);
305 
306 	/*
307 	 * Now hand over this thread to swapper.
308 	 */
309 	swapper();
310 	/* NOTREACHED*/
311 }
312 
313 
314 /*
315  ***************************************************************************
316  ****
317  **** The following SYSINIT's belong elsewhere, but have not yet
318  **** been moved.
319  ****
320  ***************************************************************************
321  */
322 static void
323 print_caddr_t(void *data)
324 {
325 	printf("%s", (char *)data);
326 }
327 
328 static void
329 print_version(void *data __unused)
330 {
331 	int len;
332 
333 	/* Strip a trailing newline from version. */
334 	len = strlen(version);
335 	while (len > 0 && version[len - 1] == '\n')
336 		len--;
337 	printf("%.*s %s\n", len, version, machine);
338 	printf("%s\n", compiler_version);
339 }
340 
341 SYSINIT(announce, SI_SUB_COPYRIGHT, SI_ORDER_FIRST, print_caddr_t,
342     copyright);
343 SYSINIT(trademark, SI_SUB_COPYRIGHT, SI_ORDER_SECOND, print_caddr_t,
344     trademark);
345 SYSINIT(version, SI_SUB_COPYRIGHT, SI_ORDER_THIRD, print_version, NULL);
346 
347 #ifdef WITNESS
348 static char wit_warn[] =
349      "WARNING: WITNESS option enabled, expect reduced performance.\n";
350 SYSINIT(witwarn, SI_SUB_COPYRIGHT, SI_ORDER_THIRD + 1,
351    print_caddr_t, wit_warn);
352 SYSINIT(witwarn2, SI_SUB_LAST, SI_ORDER_THIRD + 1,
353    print_caddr_t, wit_warn);
354 #endif
355 
356 #ifdef DIAGNOSTIC
357 static char diag_warn[] =
358      "WARNING: DIAGNOSTIC option enabled, expect reduced performance.\n";
359 SYSINIT(diagwarn, SI_SUB_COPYRIGHT, SI_ORDER_THIRD + 2,
360     print_caddr_t, diag_warn);
361 SYSINIT(diagwarn2, SI_SUB_LAST, SI_ORDER_THIRD + 2,
362     print_caddr_t, diag_warn);
363 #endif
364 
365 static int
366 null_fetch_syscall_args(struct thread *td __unused,
367     struct syscall_args *sa __unused)
368 {
369 
370 	panic("null_fetch_syscall_args");
371 }
372 
373 static void
374 null_set_syscall_retval(struct thread *td __unused, int error __unused)
375 {
376 
377 	panic("null_set_syscall_retval");
378 }
379 
380 struct sysentvec null_sysvec = {
381 	.sv_size	= 0,
382 	.sv_table	= NULL,
383 	.sv_mask	= 0,
384 	.sv_sigsize	= 0,
385 	.sv_sigtbl	= NULL,
386 	.sv_errsize	= 0,
387 	.sv_errtbl	= NULL,
388 	.sv_transtrap	= NULL,
389 	.sv_fixup	= NULL,
390 	.sv_sendsig	= NULL,
391 	.sv_sigcode	= NULL,
392 	.sv_szsigcode	= NULL,
393 	.sv_prepsyscall	= NULL,
394 	.sv_name	= "null",
395 	.sv_coredump	= NULL,
396 	.sv_imgact_try	= NULL,
397 	.sv_minsigstksz	= 0,
398 	.sv_pagesize	= PAGE_SIZE,
399 	.sv_minuser	= VM_MIN_ADDRESS,
400 	.sv_maxuser	= VM_MAXUSER_ADDRESS,
401 	.sv_usrstack	= USRSTACK,
402 	.sv_psstrings	= PS_STRINGS,
403 	.sv_stackprot	= VM_PROT_ALL,
404 	.sv_copyout_strings	= NULL,
405 	.sv_setregs	= NULL,
406 	.sv_fixlimit	= NULL,
407 	.sv_maxssiz	= NULL,
408 	.sv_flags	= 0,
409 	.sv_set_syscall_retval = null_set_syscall_retval,
410 	.sv_fetch_syscall_args = null_fetch_syscall_args,
411 	.sv_syscallnames = NULL,
412 	.sv_schedtail	= NULL,
413 };
414 
415 /*
416  ***************************************************************************
417  ****
418  **** The two following SYSINIT's are proc0 specific glue code.  I am not
419  **** convinced that they can not be safely combined, but their order of
420  **** operation has been maintained as the same as the original init_main.c
421  **** for right now.
422  ****
423  **** These probably belong in init_proc.c or kern_proc.c, since they
424  **** deal with proc0 (the fork template process).
425  ****
426  ***************************************************************************
427  */
428 /* ARGSUSED*/
429 static void
430 proc0_init(void *dummy __unused)
431 {
432 	struct proc *p;
433 	struct thread *td;
434 	vm_paddr_t pageablemem;
435 	int i;
436 
437 	GIANT_REQUIRED;
438 	p = &proc0;
439 	td = &thread0;
440 
441 	/*
442 	 * Initialize magic number and osrel.
443 	 */
444 	p->p_magic = P_MAGIC;
445 	p->p_osrel = osreldate;
446 
447 	/*
448 	 * Initialize thread and process structures.
449 	 */
450 	procinit();	/* set up proc zone */
451 	threadinit();	/* set up UMA zones */
452 
453 	/*
454 	 * Initialise scheduler resources.
455 	 * Add scheduler specific parts to proc, thread as needed.
456 	 */
457 	schedinit();	/* scheduler gets its house in order */
458 
459 	/*
460 	 * Create process 0 (the swapper).
461 	 */
462 	LIST_INSERT_HEAD(&allproc, p, p_list);
463 	LIST_INSERT_HEAD(PIDHASH(0), p, p_hash);
464 	mtx_init(&pgrp0.pg_mtx, "process group", NULL, MTX_DEF | MTX_DUPOK);
465 	p->p_pgrp = &pgrp0;
466 	LIST_INSERT_HEAD(PGRPHASH(0), &pgrp0, pg_hash);
467 	LIST_INIT(&pgrp0.pg_members);
468 	LIST_INSERT_HEAD(&pgrp0.pg_members, p, p_pglist);
469 
470 	pgrp0.pg_session = &session0;
471 	mtx_init(&session0.s_mtx, "session", NULL, MTX_DEF);
472 	refcount_init(&session0.s_count, 1);
473 	session0.s_leader = p;
474 
475 	p->p_sysent = &null_sysvec;
476 	p->p_flag = P_SYSTEM | P_INMEM;
477 	p->p_flag2 = 0;
478 	p->p_state = PRS_NORMAL;
479 	knlist_init_mtx(&p->p_klist, &p->p_mtx);
480 	STAILQ_INIT(&p->p_ktr);
481 	p->p_nice = NZERO;
482 	/* pid_max cannot be greater than PID_MAX */
483 	td->td_tid = PID_MAX + 1;
484 	LIST_INSERT_HEAD(TIDHASH(td->td_tid), td, td_hash);
485 	td->td_state = TDS_RUNNING;
486 	td->td_pri_class = PRI_TIMESHARE;
487 	td->td_user_pri = PUSER;
488 	td->td_base_user_pri = PUSER;
489 	td->td_lend_user_pri = PRI_MAX;
490 	td->td_priority = PVM;
491 	td->td_base_pri = PVM;
492 	td->td_oncpu = 0;
493 	td->td_flags = TDF_INMEM;
494 	td->td_pflags = TDP_KTHREAD;
495 	td->td_cpuset = cpuset_thread0();
496 	prison0.pr_cpuset = cpuset_ref(td->td_cpuset);
497 	p->p_peers = 0;
498 	p->p_leader = p;
499 
500 
501 	strncpy(p->p_comm, "kernel", sizeof (p->p_comm));
502 	strncpy(td->td_name, "swapper", sizeof (td->td_name));
503 
504 	callout_init_mtx(&p->p_itcallout, &p->p_mtx, 0);
505 	callout_init_mtx(&p->p_limco, &p->p_mtx, 0);
506 	callout_init(&td->td_slpcallout, CALLOUT_MPSAFE);
507 
508 	/* Create credentials. */
509 	p->p_ucred = crget();
510 	p->p_ucred->cr_ngroups = 1;	/* group 0 */
511 	p->p_ucred->cr_uidinfo = uifind(0);
512 	p->p_ucred->cr_ruidinfo = uifind(0);
513 	p->p_ucred->cr_prison = &prison0;
514 	p->p_ucred->cr_loginclass = loginclass_find("default");
515 #ifdef AUDIT
516 	audit_cred_kproc0(p->p_ucred);
517 #endif
518 #ifdef MAC
519 	mac_cred_create_swapper(p->p_ucred);
520 #endif
521 	td->td_ucred = crhold(p->p_ucred);
522 
523 	/* Create sigacts. */
524 	p->p_sigacts = sigacts_alloc();
525 
526 	/* Initialize signal state for process 0. */
527 	siginit(&proc0);
528 
529 	/* Create the file descriptor table. */
530 	p->p_fd = fdinit(NULL);
531 	p->p_fdtol = NULL;
532 
533 	/* Create the limits structures. */
534 	p->p_limit = lim_alloc();
535 	for (i = 0; i < RLIM_NLIMITS; i++)
536 		p->p_limit->pl_rlimit[i].rlim_cur =
537 		    p->p_limit->pl_rlimit[i].rlim_max = RLIM_INFINITY;
538 	p->p_limit->pl_rlimit[RLIMIT_NOFILE].rlim_cur =
539 	    p->p_limit->pl_rlimit[RLIMIT_NOFILE].rlim_max = maxfiles;
540 	p->p_limit->pl_rlimit[RLIMIT_NPROC].rlim_cur =
541 	    p->p_limit->pl_rlimit[RLIMIT_NPROC].rlim_max = maxproc;
542 	p->p_limit->pl_rlimit[RLIMIT_DATA].rlim_cur = dfldsiz;
543 	p->p_limit->pl_rlimit[RLIMIT_DATA].rlim_max = maxdsiz;
544 	p->p_limit->pl_rlimit[RLIMIT_STACK].rlim_cur = dflssiz;
545 	p->p_limit->pl_rlimit[RLIMIT_STACK].rlim_max = maxssiz;
546 	/* Cast to avoid overflow on i386/PAE. */
547 	pageablemem = ptoa((vm_paddr_t)cnt.v_free_count);
548 	p->p_limit->pl_rlimit[RLIMIT_RSS].rlim_cur =
549 	    p->p_limit->pl_rlimit[RLIMIT_RSS].rlim_max = pageablemem;
550 	p->p_limit->pl_rlimit[RLIMIT_MEMLOCK].rlim_cur = pageablemem / 3;
551 	p->p_limit->pl_rlimit[RLIMIT_MEMLOCK].rlim_max = pageablemem;
552 	p->p_cpulimit = RLIM_INFINITY;
553 
554 	/* Initialize resource accounting structures. */
555 	racct_create(&p->p_racct);
556 
557 	p->p_stats = pstats_alloc();
558 
559 	/* Allocate a prototype map so we have something to fork. */
560 	pmap_pinit0(vmspace_pmap(&vmspace0));
561 	p->p_vmspace = &vmspace0;
562 	vmspace0.vm_refcnt = 1;
563 
564 	/*
565 	 * proc0 is not expected to enter usermode, so there is no special
566 	 * handling for sv_minuser here, like is done for exec_new_vmspace().
567 	 */
568 	vm_map_init(&vmspace0.vm_map, vmspace_pmap(&vmspace0),
569 	    p->p_sysent->sv_minuser, p->p_sysent->sv_maxuser);
570 
571 	/*
572 	 * Call the init and ctor for the new thread and proc.  We wait
573 	 * to do this until all other structures are fairly sane.
574 	 */
575 	EVENTHANDLER_INVOKE(process_init, p);
576 	EVENTHANDLER_INVOKE(thread_init, td);
577 	EVENTHANDLER_INVOKE(process_ctor, p);
578 	EVENTHANDLER_INVOKE(thread_ctor, td);
579 
580 	/*
581 	 * Charge root for one process.
582 	 */
583 	(void)chgproccnt(p->p_ucred->cr_ruidinfo, 1, 0);
584 	PROC_LOCK(p);
585 	racct_add_force(p, RACCT_NPROC, 1);
586 	PROC_UNLOCK(p);
587 }
588 SYSINIT(p0init, SI_SUB_INTRINSIC, SI_ORDER_FIRST, proc0_init, NULL);
589 
590 /* ARGSUSED*/
591 static void
592 proc0_post(void *dummy __unused)
593 {
594 	struct timespec ts;
595 	struct proc *p;
596 	struct rusage ru;
597 	struct thread *td;
598 
599 	/*
600 	 * Now we can look at the time, having had a chance to verify the
601 	 * time from the filesystem.  Pretend that proc0 started now.
602 	 */
603 	sx_slock(&allproc_lock);
604 	FOREACH_PROC_IN_SYSTEM(p) {
605 		microuptime(&p->p_stats->p_start);
606 		PROC_SLOCK(p);
607 		rufetch(p, &ru);	/* Clears thread stats */
608 		PROC_SUNLOCK(p);
609 		p->p_rux.rux_runtime = 0;
610 		p->p_rux.rux_uticks = 0;
611 		p->p_rux.rux_sticks = 0;
612 		p->p_rux.rux_iticks = 0;
613 		FOREACH_THREAD_IN_PROC(p, td) {
614 			td->td_runtime = 0;
615 		}
616 	}
617 	sx_sunlock(&allproc_lock);
618 	PCPU_SET(switchtime, cpu_ticks());
619 	PCPU_SET(switchticks, ticks);
620 
621 	/*
622 	 * Give the ``random'' number generator a thump.
623 	 */
624 	nanotime(&ts);
625 	srandom(ts.tv_sec ^ ts.tv_nsec);
626 }
627 SYSINIT(p0post, SI_SUB_INTRINSIC_POST, SI_ORDER_FIRST, proc0_post, NULL);
628 
629 static void
630 random_init(void *dummy __unused)
631 {
632 
633 	/*
634 	 * After CPU has been started we have some randomness on most
635 	 * platforms via get_cyclecount().  For platforms that don't
636 	 * we will reseed random(9) in proc0_post() as well.
637 	 */
638 	srandom(get_cyclecount());
639 }
640 SYSINIT(random, SI_SUB_RANDOM, SI_ORDER_FIRST, random_init, NULL);
641 
642 /*
643  ***************************************************************************
644  ****
645  **** The following SYSINIT's and glue code should be moved to the
646  **** respective files on a per subsystem basis.
647  ****
648  ***************************************************************************
649  */
650 
651 
652 /*
653  ***************************************************************************
654  ****
655  **** The following code probably belongs in another file, like
656  **** kern/init_init.c.
657  ****
658  ***************************************************************************
659  */
660 
661 /*
662  * List of paths to try when searching for "init".
663  */
664 static char init_path[MAXPATHLEN] =
665 #ifdef	INIT_PATH
666     __XSTRING(INIT_PATH);
667 #else
668     "/sbin/init:/sbin/oinit:/sbin/init.bak:/rescue/init";
669 #endif
670 SYSCTL_STRING(_kern, OID_AUTO, init_path, CTLFLAG_RD, init_path, 0,
671 	"Path used to search the init process");
672 
673 /*
674  * Shutdown timeout of init(8).
675  * Unused within kernel, but used to control init(8), hence do not remove.
676  */
677 #ifndef INIT_SHUTDOWN_TIMEOUT
678 #define INIT_SHUTDOWN_TIMEOUT 120
679 #endif
680 static int init_shutdown_timeout = INIT_SHUTDOWN_TIMEOUT;
681 SYSCTL_INT(_kern, OID_AUTO, init_shutdown_timeout,
682 	CTLFLAG_RW, &init_shutdown_timeout, 0, "Shutdown timeout of init(8). "
683 	"Unused within kernel, but used to control init(8)");
684 
685 /*
686  * Start the initial user process; try exec'ing each pathname in init_path.
687  * The program is invoked with one argument containing the boot flags.
688  */
689 static void
690 start_init(void *dummy)
691 {
692 	vm_offset_t addr;
693 	struct execve_args args;
694 	int options, error;
695 	char *var, *path, *next, *s;
696 	char *ucp, **uap, *arg0, *arg1;
697 	struct thread *td;
698 	struct proc *p;
699 
700 	mtx_lock(&Giant);
701 
702 	GIANT_REQUIRED;
703 
704 	td = curthread;
705 	p = td->td_proc;
706 
707 	vfs_mountroot();
708 
709 	/*
710 	 * Need just enough stack to hold the faked-up "execve()" arguments.
711 	 */
712 	addr = p->p_sysent->sv_usrstack - PAGE_SIZE;
713 	if (vm_map_find(&p->p_vmspace->vm_map, NULL, 0, &addr, PAGE_SIZE, 0,
714 	    VMFS_NO_SPACE, VM_PROT_ALL, VM_PROT_ALL, 0) != 0)
715 		panic("init: couldn't allocate argument space");
716 	p->p_vmspace->vm_maxsaddr = (caddr_t)addr;
717 	p->p_vmspace->vm_ssize = 1;
718 
719 	if ((var = getenv("init_path")) != NULL) {
720 		strlcpy(init_path, var, sizeof(init_path));
721 		freeenv(var);
722 	}
723 
724 	for (path = init_path; *path != '\0'; path = next) {
725 		while (*path == ':')
726 			path++;
727 		if (*path == '\0')
728 			break;
729 		for (next = path; *next != '\0' && *next != ':'; next++)
730 			/* nothing */ ;
731 		if (bootverbose)
732 			printf("start_init: trying %.*s\n", (int)(next - path),
733 			    path);
734 
735 		/*
736 		 * Move out the boot flag argument.
737 		 */
738 		options = 0;
739 		ucp = (char *)p->p_sysent->sv_usrstack;
740 		(void)subyte(--ucp, 0);		/* trailing zero */
741 		if (boothowto & RB_SINGLE) {
742 			(void)subyte(--ucp, 's');
743 			options = 1;
744 		}
745 #ifdef notyet
746                 if (boothowto & RB_FASTBOOT) {
747 			(void)subyte(--ucp, 'f');
748 			options = 1;
749 		}
750 #endif
751 
752 #ifdef BOOTCDROM
753 		(void)subyte(--ucp, 'C');
754 		options = 1;
755 #endif
756 
757 		if (options == 0)
758 			(void)subyte(--ucp, '-');
759 		(void)subyte(--ucp, '-');		/* leading hyphen */
760 		arg1 = ucp;
761 
762 		/*
763 		 * Move out the file name (also arg 0).
764 		 */
765 		(void)subyte(--ucp, 0);
766 		for (s = next - 1; s >= path; s--)
767 			(void)subyte(--ucp, *s);
768 		arg0 = ucp;
769 
770 		/*
771 		 * Move out the arg pointers.
772 		 */
773 		uap = (char **)((intptr_t)ucp & ~(sizeof(intptr_t)-1));
774 		(void)suword((caddr_t)--uap, (long)0);	/* terminator */
775 		(void)suword((caddr_t)--uap, (long)(intptr_t)arg1);
776 		(void)suword((caddr_t)--uap, (long)(intptr_t)arg0);
777 
778 		/*
779 		 * Point at the arguments.
780 		 */
781 		args.fname = arg0;
782 		args.argv = uap;
783 		args.envv = NULL;
784 
785 		/*
786 		 * Now try to exec the program.  If can't for any reason
787 		 * other than it doesn't exist, complain.
788 		 *
789 		 * Otherwise, return via fork_trampoline() all the way
790 		 * to user mode as init!
791 		 */
792 		if ((error = sys_execve(td, &args)) == 0) {
793 			mtx_unlock(&Giant);
794 			return;
795 		}
796 		if (error != ENOENT)
797 			printf("exec %.*s: error %d\n", (int)(next - path),
798 			    path, error);
799 	}
800 	printf("init: not found in path %s\n", init_path);
801 	panic("no init");
802 }
803 
804 /*
805  * Like kproc_create(), but runs in it's own address space.
806  * We do this early to reserve pid 1.
807  *
808  * Note special case - do not make it runnable yet.  Other work
809  * in progress will change this more.
810  */
811 static void
812 create_init(const void *udata __unused)
813 {
814 	struct ucred *newcred, *oldcred;
815 	int error;
816 
817 	error = fork1(&thread0, RFFDG | RFPROC | RFSTOPPED, 0, &initproc,
818 	    NULL, 0);
819 	if (error)
820 		panic("cannot fork init: %d\n", error);
821 	KASSERT(initproc->p_pid == 1, ("create_init: initproc->p_pid != 1"));
822 	/* divorce init's credentials from the kernel's */
823 	newcred = crget();
824 	PROC_LOCK(initproc);
825 	initproc->p_flag |= P_SYSTEM | P_INMEM;
826 	oldcred = initproc->p_ucred;
827 	crcopy(newcred, oldcred);
828 #ifdef MAC
829 	mac_cred_create_init(newcred);
830 #endif
831 #ifdef AUDIT
832 	audit_cred_proc1(newcred);
833 #endif
834 	initproc->p_ucred = newcred;
835 	PROC_UNLOCK(initproc);
836 	crfree(oldcred);
837 	cred_update_thread(FIRST_THREAD_IN_PROC(initproc));
838 	cpu_set_fork_handler(FIRST_THREAD_IN_PROC(initproc), start_init, NULL);
839 }
840 SYSINIT(init, SI_SUB_CREATE_INIT, SI_ORDER_FIRST, create_init, NULL);
841 
842 /*
843  * Make it runnable now.
844  */
845 static void
846 kick_init(const void *udata __unused)
847 {
848 	struct thread *td;
849 
850 	td = FIRST_THREAD_IN_PROC(initproc);
851 	thread_lock(td);
852 	TD_SET_CAN_RUN(td);
853 	sched_add(td, SRQ_BORING);
854 	thread_unlock(td);
855 }
856 SYSINIT(kickinit, SI_SUB_KTHREAD_INIT, SI_ORDER_FIRST, kick_init, NULL);
857