xref: /freebsd/sys/kern/init_main.c (revision ae41709ab46305df80f7f35bb478a3c8ebf22ebb)
1 /*-
2  * SPDX-License-Identifier: BSD-4-Clause
3  *
4  * Copyright (c) 1995 Terrence R. Lambert
5  * All rights reserved.
6  *
7  * Copyright (c) 1982, 1986, 1989, 1991, 1992, 1993
8  *	The Regents of the University of California.  All rights reserved.
9  * (c) UNIX System Laboratories, Inc.
10  * All or some portions of this file are derived from material licensed
11  * to the University of California by American Telephone and Telegraph
12  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
13  * the permission of UNIX System Laboratories, Inc.
14  *
15  * Redistribution and use in source and binary forms, with or without
16  * modification, are permitted provided that the following conditions
17  * are met:
18  * 1. Redistributions of source code must retain the above copyright
19  *    notice, this list of conditions and the following disclaimer.
20  * 2. Redistributions in binary form must reproduce the above copyright
21  *    notice, this list of conditions and the following disclaimer in the
22  *    documentation and/or other materials provided with the distribution.
23  * 3. All advertising materials mentioning features or use of this software
24  *    must display the following acknowledgement:
25  *	This product includes software developed by the University of
26  *	California, Berkeley and its contributors.
27  * 4. Neither the name of the University nor the names of its contributors
28  *    may be used to endorse or promote products derived from this software
29  *    without specific prior written permission.
30  *
31  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
32  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
33  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
34  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
35  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
36  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
37  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
38  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
39  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
40  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
41  * SUCH DAMAGE.
42  *
43  *	@(#)init_main.c	8.9 (Berkeley) 1/21/94
44  */
45 
46 #include <sys/cdefs.h>
47 __FBSDID("$FreeBSD$");
48 
49 #include "opt_ddb.h"
50 #include "opt_kdb.h"
51 #include "opt_init_path.h"
52 #include "opt_verbose_sysinit.h"
53 
54 #include <sys/param.h>
55 #include <sys/kernel.h>
56 #include <sys/epoch.h>
57 #include <sys/eventhandler.h>
58 #include <sys/exec.h>
59 #include <sys/file.h>
60 #include <sys/filedesc.h>
61 #include <sys/imgact.h>
62 #include <sys/jail.h>
63 #include <sys/ktr.h>
64 #include <sys/lock.h>
65 #include <sys/loginclass.h>
66 #include <sys/mount.h>
67 #include <sys/mutex.h>
68 #include <sys/syscallsubr.h>
69 #include <sys/sysctl.h>
70 #include <sys/proc.h>
71 #include <sys/racct.h>
72 #include <sys/resourcevar.h>
73 #include <sys/systm.h>
74 #include <sys/signalvar.h>
75 #include <sys/vnode.h>
76 #include <sys/sysent.h>
77 #include <sys/reboot.h>
78 #include <sys/sched.h>
79 #include <sys/sx.h>
80 #include <sys/sysproto.h>
81 #include <sys/vmmeter.h>
82 #include <sys/unistd.h>
83 #include <sys/malloc.h>
84 #include <sys/conf.h>
85 #include <sys/cpuset.h>
86 
87 #include <machine/cpu.h>
88 
89 #include <security/audit/audit.h>
90 #include <security/mac/mac_framework.h>
91 
92 #include <vm/vm.h>
93 #include <vm/vm_param.h>
94 #include <vm/vm_extern.h>
95 #include <vm/pmap.h>
96 #include <vm/vm_map.h>
97 #include <sys/copyright.h>
98 
99 #include <ddb/ddb.h>
100 #include <ddb/db_sym.h>
101 
102 void mi_startup(void);				/* Should be elsewhere */
103 
104 /* Components of the first process -- never freed. */
105 static struct session session0;
106 static struct pgrp pgrp0;
107 struct	proc proc0;
108 struct thread0_storage thread0_st __aligned(32);
109 struct	vmspace vmspace0;
110 struct	proc *initproc;
111 
112 int
113 linux_alloc_current_noop(struct thread *td __unused, int flags __unused)
114 {
115 	return (0);
116 }
117 int (*lkpi_alloc_current)(struct thread *, int) = linux_alloc_current_noop;
118 
119 #ifndef BOOTHOWTO
120 #define	BOOTHOWTO	0
121 #endif
122 int	boothowto = BOOTHOWTO;	/* initialized so that it can be patched */
123 SYSCTL_INT(_debug, OID_AUTO, boothowto, CTLFLAG_RD, &boothowto, 0,
124 	"Boot control flags, passed from loader");
125 
126 #ifndef BOOTVERBOSE
127 #define	BOOTVERBOSE	0
128 #endif
129 int	bootverbose = BOOTVERBOSE;
130 SYSCTL_INT(_debug, OID_AUTO, bootverbose, CTLFLAG_RW, &bootverbose, 0,
131 	"Control the output of verbose kernel messages");
132 
133 #ifdef VERBOSE_SYSINIT
134 /*
135  * We'll use the defined value of VERBOSE_SYSINIT from the kernel config to
136  * dictate the default VERBOSE_SYSINIT behavior.  Significant values for this
137  * option and associated tunable are:
138  * - 0, 'compiled in but silent by default'
139  * - 1, 'compiled in but verbose by default' (default)
140  */
141 int	verbose_sysinit = VERBOSE_SYSINIT;
142 TUNABLE_INT("debug.verbose_sysinit", &verbose_sysinit);
143 #endif
144 
145 #ifdef INVARIANTS
146 FEATURE(invariants, "Kernel compiled with INVARIANTS, may affect performance");
147 #endif
148 
149 /*
150  * This ensures that there is at least one entry so that the sysinit_set
151  * symbol is not undefined.  A sybsystem ID of SI_SUB_DUMMY is never
152  * executed.
153  */
154 SYSINIT(placeholder, SI_SUB_DUMMY, SI_ORDER_ANY, NULL, NULL);
155 
156 /*
157  * The sysinit table itself.  Items are checked off as the are run.
158  * If we want to register new sysinit types, add them to newsysinit.
159  */
160 SET_DECLARE(sysinit_set, struct sysinit);
161 struct sysinit **sysinit, **sysinit_end;
162 struct sysinit **newsysinit, **newsysinit_end;
163 
164 /*
165  * Merge a new sysinit set into the current set, reallocating it if
166  * necessary.  This can only be called after malloc is running.
167  */
168 void
169 sysinit_add(struct sysinit **set, struct sysinit **set_end)
170 {
171 	struct sysinit **newset;
172 	struct sysinit **sipp;
173 	struct sysinit **xipp;
174 	int count;
175 
176 	count = set_end - set;
177 	if (newsysinit)
178 		count += newsysinit_end - newsysinit;
179 	else
180 		count += sysinit_end - sysinit;
181 	newset = malloc(count * sizeof(*sipp), M_TEMP, M_NOWAIT);
182 	if (newset == NULL)
183 		panic("cannot malloc for sysinit");
184 	xipp = newset;
185 	if (newsysinit)
186 		for (sipp = newsysinit; sipp < newsysinit_end; sipp++)
187 			*xipp++ = *sipp;
188 	else
189 		for (sipp = sysinit; sipp < sysinit_end; sipp++)
190 			*xipp++ = *sipp;
191 	for (sipp = set; sipp < set_end; sipp++)
192 		*xipp++ = *sipp;
193 	if (newsysinit)
194 		free(newsysinit, M_TEMP);
195 	newsysinit = newset;
196 	newsysinit_end = newset + count;
197 }
198 
199 #if defined (DDB) && defined(VERBOSE_SYSINIT)
200 static const char *
201 symbol_name(vm_offset_t va, db_strategy_t strategy)
202 {
203 	const char *name;
204 	c_db_sym_t sym;
205 	db_expr_t  offset;
206 
207 	if (va == 0)
208 		return (NULL);
209 	sym = db_search_symbol(va, strategy, &offset);
210 	if (offset != 0)
211 		return (NULL);
212 	db_symbol_values(sym, &name, NULL);
213 	return (name);
214 }
215 #endif
216 
217 /*
218  * System startup; initialize the world, create process 0, mount root
219  * filesystem, and fork to create init and pagedaemon.  Most of the
220  * hard work is done in the lower-level initialization routines including
221  * startup(), which does memory initialization and autoconfiguration.
222  *
223  * This allows simple addition of new kernel subsystems that require
224  * boot time initialization.  It also allows substitution of subsystem
225  * (for instance, a scheduler, kernel profiler, or VM system) by object
226  * module.  Finally, it allows for optional "kernel threads".
227  */
228 void
229 mi_startup(void)
230 {
231 
232 	struct sysinit **sipp;	/* system initialization*/
233 	struct sysinit **xipp;	/* interior loop of sort*/
234 	struct sysinit *save;	/* bubble*/
235 
236 #if defined(VERBOSE_SYSINIT)
237 	int last;
238 	int verbose;
239 #endif
240 
241 	TSENTER();
242 
243 	if (boothowto & RB_VERBOSE)
244 		bootverbose++;
245 
246 	if (sysinit == NULL) {
247 		sysinit = SET_BEGIN(sysinit_set);
248 		sysinit_end = SET_LIMIT(sysinit_set);
249 	}
250 
251 restart:
252 	/*
253 	 * Perform a bubble sort of the system initialization objects by
254 	 * their subsystem (primary key) and order (secondary key).
255 	 */
256 	for (sipp = sysinit; sipp < sysinit_end; sipp++) {
257 		for (xipp = sipp + 1; xipp < sysinit_end; xipp++) {
258 			if ((*sipp)->subsystem < (*xipp)->subsystem ||
259 			     ((*sipp)->subsystem == (*xipp)->subsystem &&
260 			      (*sipp)->order <= (*xipp)->order))
261 				continue;	/* skip*/
262 			save = *sipp;
263 			*sipp = *xipp;
264 			*xipp = save;
265 		}
266 	}
267 
268 #if defined(VERBOSE_SYSINIT)
269 	last = SI_SUB_COPYRIGHT;
270 	verbose = 0;
271 #if !defined(DDB)
272 	printf("VERBOSE_SYSINIT: DDB not enabled, symbol lookups disabled.\n");
273 #endif
274 #endif
275 
276 	/*
277 	 * Traverse the (now) ordered list of system initialization tasks.
278 	 * Perform each task, and continue on to the next task.
279 	 */
280 	for (sipp = sysinit; sipp < sysinit_end; sipp++) {
281 		if ((*sipp)->subsystem == SI_SUB_DUMMY)
282 			continue;	/* skip dummy task(s)*/
283 
284 		if ((*sipp)->subsystem == SI_SUB_DONE)
285 			continue;
286 
287 #if defined(VERBOSE_SYSINIT)
288 		if ((*sipp)->subsystem > last && verbose_sysinit != 0) {
289 			verbose = 1;
290 			last = (*sipp)->subsystem;
291 			printf("subsystem %x\n", last);
292 		}
293 		if (verbose) {
294 #if defined(DDB)
295 			const char *func, *data;
296 
297 			func = symbol_name((vm_offset_t)(*sipp)->func,
298 			    DB_STGY_PROC);
299 			data = symbol_name((vm_offset_t)(*sipp)->udata,
300 			    DB_STGY_ANY);
301 			if (func != NULL && data != NULL)
302 				printf("   %s(&%s)... ", func, data);
303 			else if (func != NULL)
304 				printf("   %s(%p)... ", func, (*sipp)->udata);
305 			else
306 #endif
307 				printf("   %p(%p)... ", (*sipp)->func,
308 				    (*sipp)->udata);
309 		}
310 #endif
311 
312 		/* Call function */
313 		(*((*sipp)->func))((*sipp)->udata);
314 
315 #if defined(VERBOSE_SYSINIT)
316 		if (verbose)
317 			printf("done.\n");
318 #endif
319 
320 		/* Check off the one we're just done */
321 		(*sipp)->subsystem = SI_SUB_DONE;
322 
323 		/* Check if we've installed more sysinit items via KLD */
324 		if (newsysinit != NULL) {
325 			if (sysinit != SET_BEGIN(sysinit_set))
326 				free(sysinit, M_TEMP);
327 			sysinit = newsysinit;
328 			sysinit_end = newsysinit_end;
329 			newsysinit = NULL;
330 			newsysinit_end = NULL;
331 			goto restart;
332 		}
333 	}
334 
335 	TSEXIT();	/* Here so we don't overlap with start_init. */
336 
337 	mtx_assert(&Giant, MA_OWNED | MA_NOTRECURSED);
338 	mtx_unlock(&Giant);
339 
340 	/*
341 	 * Now hand over this thread to swapper.
342 	 */
343 	swapper();
344 	/* NOTREACHED*/
345 }
346 
347 static void
348 print_caddr_t(void *data)
349 {
350 	printf("%s", (char *)data);
351 }
352 
353 static void
354 print_version(void *data __unused)
355 {
356 	int len;
357 
358 	/* Strip a trailing newline from version. */
359 	len = strlen(version);
360 	while (len > 0 && version[len - 1] == '\n')
361 		len--;
362 	printf("%.*s %s\n", len, version, machine);
363 	printf("%s\n", compiler_version);
364 }
365 
366 SYSINIT(announce, SI_SUB_COPYRIGHT, SI_ORDER_FIRST, print_caddr_t,
367     copyright);
368 SYSINIT(trademark, SI_SUB_COPYRIGHT, SI_ORDER_SECOND, print_caddr_t,
369     trademark);
370 SYSINIT(version, SI_SUB_COPYRIGHT, SI_ORDER_THIRD, print_version, NULL);
371 
372 #ifdef WITNESS
373 static char wit_warn[] =
374      "WARNING: WITNESS option enabled, expect reduced performance.\n";
375 SYSINIT(witwarn, SI_SUB_COPYRIGHT, SI_ORDER_FOURTH,
376    print_caddr_t, wit_warn);
377 SYSINIT(witwarn2, SI_SUB_LAST, SI_ORDER_FOURTH,
378    print_caddr_t, wit_warn);
379 #endif
380 
381 #ifdef DIAGNOSTIC
382 static char diag_warn[] =
383      "WARNING: DIAGNOSTIC option enabled, expect reduced performance.\n";
384 SYSINIT(diagwarn, SI_SUB_COPYRIGHT, SI_ORDER_FIFTH,
385     print_caddr_t, diag_warn);
386 SYSINIT(diagwarn2, SI_SUB_LAST, SI_ORDER_FIFTH,
387     print_caddr_t, diag_warn);
388 #endif
389 
390 static int
391 null_fetch_syscall_args(struct thread *td __unused)
392 {
393 
394 	panic("null_fetch_syscall_args");
395 }
396 
397 static void
398 null_set_syscall_retval(struct thread *td __unused, int error __unused)
399 {
400 
401 	panic("null_set_syscall_retval");
402 }
403 
404 struct sysentvec null_sysvec = {
405 	.sv_size	= 0,
406 	.sv_table	= NULL,
407 	.sv_transtrap	= NULL,
408 	.sv_fixup	= NULL,
409 	.sv_sendsig	= NULL,
410 	.sv_sigcode	= NULL,
411 	.sv_szsigcode	= NULL,
412 	.sv_name	= "null",
413 	.sv_coredump	= NULL,
414 	.sv_imgact_try	= NULL,
415 	.sv_minsigstksz	= 0,
416 	.sv_minuser	= VM_MIN_ADDRESS,
417 	.sv_maxuser	= VM_MAXUSER_ADDRESS,
418 	.sv_usrstack	= USRSTACK,
419 	.sv_psstrings	= PS_STRINGS,
420 	.sv_stackprot	= VM_PROT_ALL,
421 	.sv_copyout_strings	= NULL,
422 	.sv_setregs	= NULL,
423 	.sv_fixlimit	= NULL,
424 	.sv_maxssiz	= NULL,
425 	.sv_flags	= 0,
426 	.sv_set_syscall_retval = null_set_syscall_retval,
427 	.sv_fetch_syscall_args = null_fetch_syscall_args,
428 	.sv_syscallnames = NULL,
429 	.sv_schedtail	= NULL,
430 	.sv_thread_detach = NULL,
431 	.sv_trap	= NULL,
432 };
433 
434 /*
435  * The two following SYSINIT's are proc0 specific glue code.  I am not
436  * convinced that they can not be safely combined, but their order of
437  * operation has been maintained as the same as the original init_main.c
438  * for right now.
439  */
440 /* ARGSUSED*/
441 static void
442 proc0_init(void *dummy __unused)
443 {
444 	struct proc *p;
445 	struct thread *td;
446 	struct ucred *newcred;
447 	struct uidinfo tmpuinfo;
448 	struct loginclass tmplc = {
449 		.lc_name = "",
450 	};
451 	vm_paddr_t pageablemem;
452 	int i;
453 
454 	GIANT_REQUIRED;
455 	p = &proc0;
456 	td = &thread0;
457 
458 	/*
459 	 * Initialize magic number and osrel.
460 	 */
461 	p->p_magic = P_MAGIC;
462 	p->p_osrel = osreldate;
463 
464 	/*
465 	 * Initialize thread and process structures.
466 	 */
467 	procinit();	/* set up proc zone */
468 	threadinit();	/* set up UMA zones */
469 
470 	/*
471 	 * Initialise scheduler resources.
472 	 * Add scheduler specific parts to proc, thread as needed.
473 	 */
474 	schedinit();	/* scheduler gets its house in order */
475 
476 	/*
477 	 * Create process 0 (the swapper).
478 	 */
479 	LIST_INSERT_HEAD(&allproc, p, p_list);
480 	LIST_INSERT_HEAD(PIDHASH(0), p, p_hash);
481 	mtx_init(&pgrp0.pg_mtx, "process group", NULL, MTX_DEF | MTX_DUPOK);
482 	p->p_pgrp = &pgrp0;
483 	LIST_INSERT_HEAD(PGRPHASH(0), &pgrp0, pg_hash);
484 	LIST_INIT(&pgrp0.pg_members);
485 	LIST_INSERT_HEAD(&pgrp0.pg_members, p, p_pglist);
486 
487 	pgrp0.pg_session = &session0;
488 	mtx_init(&session0.s_mtx, "session", NULL, MTX_DEF);
489 	refcount_init(&session0.s_count, 1);
490 	session0.s_leader = p;
491 
492 	p->p_sysent = &null_sysvec;
493 	p->p_flag = P_SYSTEM | P_INMEM | P_KPROC;
494 	p->p_flag2 = 0;
495 	p->p_state = PRS_NORMAL;
496 	p->p_klist = knlist_alloc(&p->p_mtx);
497 	STAILQ_INIT(&p->p_ktr);
498 	p->p_nice = NZERO;
499 	/* pid_max cannot be greater than PID_MAX */
500 	td->td_tid = PID_MAX + 1;
501 	LIST_INSERT_HEAD(TIDHASH(td->td_tid), td, td_hash);
502 	td->td_state = TDS_RUNNING;
503 	td->td_pri_class = PRI_TIMESHARE;
504 	td->td_user_pri = PUSER;
505 	td->td_base_user_pri = PUSER;
506 	td->td_lend_user_pri = PRI_MAX;
507 	td->td_priority = PVM;
508 	td->td_base_pri = PVM;
509 	td->td_oncpu = curcpu;
510 	td->td_flags = TDF_INMEM;
511 	td->td_pflags = TDP_KTHREAD;
512 	td->td_cpuset = cpuset_thread0();
513 	td->td_domain.dr_policy = td->td_cpuset->cs_domain;
514 	prison0_init();
515 	p->p_peers = 0;
516 	p->p_leader = p;
517 	p->p_reaper = p;
518 	p->p_treeflag |= P_TREE_REAPER;
519 	LIST_INIT(&p->p_reaplist);
520 
521 	strncpy(p->p_comm, "kernel", sizeof (p->p_comm));
522 	strncpy(td->td_name, "swapper", sizeof (td->td_name));
523 
524 	callout_init_mtx(&p->p_itcallout, &p->p_mtx, 0);
525 	callout_init_mtx(&p->p_limco, &p->p_mtx, 0);
526 	callout_init(&td->td_slpcallout, 1);
527 
528 	/* Create credentials. */
529 	newcred = crget();
530 	newcred->cr_ngroups = 1;	/* group 0 */
531 	/* A hack to prevent uifind from tripping over NULL pointers. */
532 	curthread->td_ucred = newcred;
533 	tmpuinfo.ui_uid = 1;
534 	newcred->cr_uidinfo = newcred->cr_ruidinfo = &tmpuinfo;
535 	newcred->cr_uidinfo = uifind(0);
536 	newcred->cr_ruidinfo = uifind(0);
537 	newcred->cr_loginclass = &tmplc;
538 	newcred->cr_loginclass = loginclass_find("default");
539 	/* End hack. creds get properly set later with thread_cow_get_proc */
540 	curthread->td_ucred = NULL;
541 	newcred->cr_prison = &prison0;
542 	newcred->cr_users++; /* avoid assertion failure */
543 	proc_set_cred_init(p, newcred);
544 	newcred->cr_users--;
545 	crfree(newcred);
546 #ifdef AUDIT
547 	audit_cred_kproc0(newcred);
548 #endif
549 #ifdef MAC
550 	mac_cred_create_swapper(newcred);
551 #endif
552 	/* Create sigacts. */
553 	p->p_sigacts = sigacts_alloc();
554 
555 	/* Initialize signal state for process 0. */
556 	siginit(&proc0);
557 
558 	/* Create the file descriptor table. */
559 	p->p_fd = fdinit(NULL, false, NULL);
560 	p->p_fdtol = NULL;
561 
562 	/* Create the limits structures. */
563 	p->p_limit = lim_alloc();
564 	for (i = 0; i < RLIM_NLIMITS; i++)
565 		p->p_limit->pl_rlimit[i].rlim_cur =
566 		    p->p_limit->pl_rlimit[i].rlim_max = RLIM_INFINITY;
567 	p->p_limit->pl_rlimit[RLIMIT_NOFILE].rlim_cur =
568 	    p->p_limit->pl_rlimit[RLIMIT_NOFILE].rlim_max = maxfiles;
569 	p->p_limit->pl_rlimit[RLIMIT_NPROC].rlim_cur =
570 	    p->p_limit->pl_rlimit[RLIMIT_NPROC].rlim_max = maxproc;
571 	p->p_limit->pl_rlimit[RLIMIT_DATA].rlim_cur = dfldsiz;
572 	p->p_limit->pl_rlimit[RLIMIT_DATA].rlim_max = maxdsiz;
573 	p->p_limit->pl_rlimit[RLIMIT_STACK].rlim_cur = dflssiz;
574 	p->p_limit->pl_rlimit[RLIMIT_STACK].rlim_max = maxssiz;
575 	/* Cast to avoid overflow on i386/PAE. */
576 	pageablemem = ptoa((vm_paddr_t)vm_free_count());
577 	p->p_limit->pl_rlimit[RLIMIT_RSS].rlim_cur =
578 	    p->p_limit->pl_rlimit[RLIMIT_RSS].rlim_max = pageablemem;
579 	p->p_limit->pl_rlimit[RLIMIT_MEMLOCK].rlim_cur = pageablemem / 3;
580 	p->p_limit->pl_rlimit[RLIMIT_MEMLOCK].rlim_max = pageablemem;
581 	p->p_cpulimit = RLIM_INFINITY;
582 
583 	PROC_LOCK(p);
584 	thread_cow_get_proc(td, p);
585 	PROC_UNLOCK(p);
586 
587 	/* Initialize resource accounting structures. */
588 	racct_create(&p->p_racct);
589 
590 	p->p_stats = pstats_alloc();
591 
592 	/* Allocate a prototype map so we have something to fork. */
593 	p->p_vmspace = &vmspace0;
594 	refcount_init(&vmspace0.vm_refcnt, 1);
595 	pmap_pinit0(vmspace_pmap(&vmspace0));
596 
597 	/*
598 	 * proc0 is not expected to enter usermode, so there is no special
599 	 * handling for sv_minuser here, like is done for exec_new_vmspace().
600 	 */
601 	vm_map_init(&vmspace0.vm_map, vmspace_pmap(&vmspace0),
602 	    p->p_sysent->sv_minuser, p->p_sysent->sv_maxuser);
603 
604 	/*
605 	 * Call the init and ctor for the new thread and proc.  We wait
606 	 * to do this until all other structures are fairly sane.
607 	 */
608 	EVENTHANDLER_DIRECT_INVOKE(process_init, p);
609 	EVENTHANDLER_DIRECT_INVOKE(thread_init, td);
610 	EVENTHANDLER_DIRECT_INVOKE(process_ctor, p);
611 	EVENTHANDLER_DIRECT_INVOKE(thread_ctor, td);
612 
613 	/*
614 	 * Charge root for one process.
615 	 */
616 	(void)chgproccnt(p->p_ucred->cr_ruidinfo, 1, 0);
617 	PROC_LOCK(p);
618 	racct_add_force(p, RACCT_NPROC, 1);
619 	PROC_UNLOCK(p);
620 }
621 SYSINIT(p0init, SI_SUB_INTRINSIC, SI_ORDER_FIRST, proc0_init, NULL);
622 
623 /* ARGSUSED*/
624 static void
625 proc0_post(void *dummy __unused)
626 {
627 	struct proc *p;
628 	struct rusage ru;
629 	struct thread *td;
630 
631 	/*
632 	 * Now we can look at the time, having had a chance to verify the
633 	 * time from the filesystem.  Pretend that proc0 started now.
634 	 */
635 	sx_slock(&allproc_lock);
636 	FOREACH_PROC_IN_SYSTEM(p) {
637 		PROC_LOCK(p);
638 		if (p->p_state == PRS_NEW) {
639 			PROC_UNLOCK(p);
640 			continue;
641 		}
642 		microuptime(&p->p_stats->p_start);
643 		PROC_STATLOCK(p);
644 		rufetch(p, &ru);	/* Clears thread stats */
645 		p->p_rux.rux_runtime = 0;
646 		p->p_rux.rux_uticks = 0;
647 		p->p_rux.rux_sticks = 0;
648 		p->p_rux.rux_iticks = 0;
649 		PROC_STATUNLOCK(p);
650 		FOREACH_THREAD_IN_PROC(p, td) {
651 			td->td_runtime = 0;
652 		}
653 		PROC_UNLOCK(p);
654 	}
655 	sx_sunlock(&allproc_lock);
656 	PCPU_SET(switchtime, cpu_ticks());
657 	PCPU_SET(switchticks, ticks);
658 }
659 SYSINIT(p0post, SI_SUB_INTRINSIC_POST, SI_ORDER_FIRST, proc0_post, NULL);
660 
661 /*
662  ***************************************************************************
663  ****
664  **** The following SYSINIT's and glue code should be moved to the
665  **** respective files on a per subsystem basis.
666  ****
667  ***************************************************************************
668  */
669 
670 /*
671  * List of paths to try when searching for "init".
672  */
673 static char init_path[MAXPATHLEN] =
674 #ifdef	INIT_PATH
675     __XSTRING(INIT_PATH);
676 #else
677     "/sbin/init:/sbin/oinit:/sbin/init.bak:/rescue/init";
678 #endif
679 SYSCTL_STRING(_kern, OID_AUTO, init_path, CTLFLAG_RD, init_path, 0,
680 	"Path used to search the init process");
681 
682 /*
683  * Shutdown timeout of init(8).
684  * Unused within kernel, but used to control init(8), hence do not remove.
685  */
686 #ifndef INIT_SHUTDOWN_TIMEOUT
687 #define INIT_SHUTDOWN_TIMEOUT 120
688 #endif
689 static int init_shutdown_timeout = INIT_SHUTDOWN_TIMEOUT;
690 SYSCTL_INT(_kern, OID_AUTO, init_shutdown_timeout,
691 	CTLFLAG_RW, &init_shutdown_timeout, 0, "Shutdown timeout of init(8). "
692 	"Unused within kernel, but used to control init(8)");
693 
694 /*
695  * Start the initial user process; try exec'ing each pathname in init_path.
696  * The program is invoked with one argument containing the boot flags.
697  */
698 static void
699 start_init(void *dummy)
700 {
701 	struct image_args args;
702 	int error;
703 	char *var, *path;
704 	char *free_init_path, *tmp_init_path;
705 	struct thread *td;
706 	struct proc *p;
707 	struct vmspace *oldvmspace;
708 
709 	TSENTER();	/* Here so we don't overlap with mi_startup. */
710 
711 	td = curthread;
712 	p = td->td_proc;
713 
714 	vfs_mountroot();
715 
716 	/* Wipe GELI passphrase from the environment. */
717 	kern_unsetenv("kern.geom.eli.passphrase");
718 
719 	/* For Multicons, report which console is primary to both */
720 	if (boothowto & RB_MULTIPLE) {
721 		if (boothowto & RB_SERIAL)
722 			printf("Dual Console: Serial Primary, Video Secondary\n");
723 		else
724 			printf("Dual Console: Video Primary, Serial Secondary\n");
725 	}
726 
727 	if ((var = kern_getenv("init_path")) != NULL) {
728 		strlcpy(init_path, var, sizeof(init_path));
729 		freeenv(var);
730 	}
731 	free_init_path = tmp_init_path = strdup(init_path, M_TEMP);
732 
733 	while ((path = strsep(&tmp_init_path, ":")) != NULL) {
734 		if (bootverbose)
735 			printf("start_init: trying %s\n", path);
736 
737 		memset(&args, 0, sizeof(args));
738 		error = exec_alloc_args(&args);
739 		if (error != 0)
740 			panic("%s: Can't allocate space for init arguments %d",
741 			    __func__, error);
742 
743 		error = exec_args_add_fname(&args, path, UIO_SYSSPACE);
744 		if (error != 0)
745 			panic("%s: Can't add fname %d", __func__, error);
746 		error = exec_args_add_arg(&args, path, UIO_SYSSPACE);
747 		if (error != 0)
748 			panic("%s: Can't add argv[0] %d", __func__, error);
749 		if (boothowto & RB_SINGLE)
750 			error = exec_args_add_arg(&args, "-s", UIO_SYSSPACE);
751 		if (error != 0)
752 			panic("%s: Can't add argv[0] %d", __func__, error);
753 
754 		/*
755 		 * Now try to exec the program.  If can't for any reason
756 		 * other than it doesn't exist, complain.
757 		 *
758 		 * Otherwise, return via fork_trampoline() all the way
759 		 * to user mode as init!
760 		 */
761 		KASSERT((td->td_pflags & TDP_EXECVMSPC) == 0,
762 		    ("nested execve"));
763 		oldvmspace = td->td_proc->p_vmspace;
764 		error = kern_execve(td, &args, NULL, oldvmspace);
765 		KASSERT(error != 0,
766 		    ("kern_execve returned success, not EJUSTRETURN"));
767 		if (error == EJUSTRETURN) {
768 			exec_cleanup(td, oldvmspace);
769 			free(free_init_path, M_TEMP);
770 			TSEXIT();
771 			return;
772 		}
773 		if (error != ENOENT)
774 			printf("exec %s: error %d\n", path, error);
775 	}
776 	free(free_init_path, M_TEMP);
777 	printf("init: not found in path %s\n", init_path);
778 	panic("no init");
779 }
780 
781 /*
782  * Like kproc_create(), but runs in its own address space.  We do this
783  * early to reserve pid 1.  Note special case - do not make it
784  * runnable yet, init execution is started when userspace can be served.
785  */
786 static void
787 create_init(const void *udata __unused)
788 {
789 	struct fork_req fr;
790 	struct ucred *newcred, *oldcred;
791 	struct thread *td;
792 	int error;
793 
794 	bzero(&fr, sizeof(fr));
795 	fr.fr_flags = RFFDG | RFPROC | RFSTOPPED;
796 	fr.fr_procp = &initproc;
797 	error = fork1(&thread0, &fr);
798 	if (error)
799 		panic("cannot fork init: %d\n", error);
800 	KASSERT(initproc->p_pid == 1, ("create_init: initproc->p_pid != 1"));
801 	/* divorce init's credentials from the kernel's */
802 	newcred = crget();
803 	sx_xlock(&proctree_lock);
804 	PROC_LOCK(initproc);
805 	initproc->p_flag |= P_SYSTEM | P_INMEM;
806 	initproc->p_treeflag |= P_TREE_REAPER;
807 	oldcred = initproc->p_ucred;
808 	crcopy(newcred, oldcred);
809 #ifdef MAC
810 	mac_cred_create_init(newcred);
811 #endif
812 #ifdef AUDIT
813 	audit_cred_proc1(newcred);
814 #endif
815 	proc_set_cred(initproc, newcred);
816 	td = FIRST_THREAD_IN_PROC(initproc);
817 	crcowfree(td);
818 	td->td_realucred = crcowget(initproc->p_ucred);
819 	td->td_ucred = td->td_realucred;
820 	PROC_UNLOCK(initproc);
821 	sx_xunlock(&proctree_lock);
822 	crfree(oldcred);
823 	cpu_fork_kthread_handler(FIRST_THREAD_IN_PROC(initproc),
824 	    start_init, NULL);
825 }
826 SYSINIT(init, SI_SUB_CREATE_INIT, SI_ORDER_FIRST, create_init, NULL);
827 
828 /*
829  * Make it runnable now.
830  */
831 static void
832 kick_init(const void *udata __unused)
833 {
834 	struct thread *td;
835 
836 	td = FIRST_THREAD_IN_PROC(initproc);
837 	thread_lock(td);
838 	TD_SET_CAN_RUN(td);
839 	sched_add(td, SRQ_BORING);
840 }
841 SYSINIT(kickinit, SI_SUB_KTHREAD_INIT, SI_ORDER_MIDDLE, kick_init, NULL);
842 
843 /*
844  * DDB(4).
845  */
846 #ifdef DDB
847 static void
848 db_show_print_syinit(struct sysinit *sip, bool ddb)
849 {
850 	const char *sname, *funcname;
851 	c_db_sym_t sym;
852 	db_expr_t  offset;
853 
854 #define xprint(...)							\
855 	if (ddb)							\
856 		db_printf(__VA_ARGS__);					\
857 	else								\
858 		printf(__VA_ARGS__)
859 
860 	if (sip == NULL) {
861 		xprint("%s: no sysinit * given\n", __func__);
862 		return;
863 	}
864 
865 	sym = db_search_symbol((vm_offset_t)sip, DB_STGY_ANY, &offset);
866 	db_symbol_values(sym, &sname, NULL);
867 	sym = db_search_symbol((vm_offset_t)sip->func, DB_STGY_PROC, &offset);
868 	db_symbol_values(sym, &funcname, NULL);
869 	xprint("%s(%p)\n", (sname != NULL) ? sname : "", sip);
870 	xprint("  %#08x %#08x\n", sip->subsystem, sip->order);
871 	xprint("  %p(%s)(%p)\n",
872 	    sip->func, (funcname != NULL) ? funcname : "", sip->udata);
873 #undef xprint
874 }
875 
876 DB_SHOW_COMMAND(sysinit, db_show_sysinit)
877 {
878 	struct sysinit **sipp;
879 
880 	db_printf("SYSINIT vs Name(Ptr)\n");
881 	db_printf("  Subsystem  Order\n");
882 	db_printf("  Function(Name)(Arg)\n");
883 	for (sipp = sysinit; sipp < sysinit_end; sipp++) {
884 		db_show_print_syinit(*sipp, true);
885 		if (db_pager_quit)
886 			break;
887 	}
888 }
889 #endif /* DDB */
890