xref: /freebsd/sys/kern/init_main.c (revision 3c4ba5f55438f7afd4f4b0b56f88f2bb505fd6a6)
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/systm.h>
56 #include <sys/boottrace.h>
57 #include <sys/conf.h>
58 #include <sys/cpuset.h>
59 #include <sys/dtrace_bsd.h>
60 #include <sys/epoch.h>
61 #include <sys/eventhandler.h>
62 #include <sys/exec.h>
63 #include <sys/file.h>
64 #include <sys/filedesc.h>
65 #include <sys/imgact.h>
66 #include <sys/jail.h>
67 #include <sys/kernel.h>
68 #include <sys/ktr.h>
69 #include <sys/lock.h>
70 #include <sys/loginclass.h>
71 #include <sys/malloc.h>
72 #include <sys/mount.h>
73 #include <sys/mutex.h>
74 #include <sys/proc.h>
75 #include <sys/racct.h>
76 #include <sys/reboot.h>
77 #include <sys/resourcevar.h>
78 #include <sys/sched.h>
79 #include <sys/signalvar.h>
80 #include <sys/sx.h>
81 #include <sys/syscallsubr.h>
82 #include <sys/sysctl.h>
83 #include <sys/sysent.h>
84 #include <sys/sysproto.h>
85 #include <sys/unistd.h>
86 #include <sys/vmmeter.h>
87 #include <sys/vnode.h>
88 
89 #include <machine/cpu.h>
90 
91 #include <security/audit/audit.h>
92 #include <security/mac/mac_framework.h>
93 
94 #include <vm/vm.h>
95 #include <vm/vm_param.h>
96 #include <vm/vm_extern.h>
97 #include <vm/pmap.h>
98 #include <vm/vm_map.h>
99 #include <sys/copyright.h>
100 
101 #include <ddb/ddb.h>
102 #include <ddb/db_sym.h>
103 
104 void mi_startup(void);				/* Should be elsewhere */
105 
106 /* Components of the first process -- never freed. */
107 static struct session session0;
108 static struct pgrp pgrp0;
109 struct	proc proc0;
110 struct thread0_storage thread0_st __aligned(32);
111 struct	vmspace vmspace0;
112 struct	proc *initproc;
113 
114 int
115 linux_alloc_current_noop(struct thread *td __unused, int flags __unused)
116 {
117 	return (0);
118 }
119 int (*lkpi_alloc_current)(struct thread *, int) = linux_alloc_current_noop;
120 
121 #ifndef BOOTHOWTO
122 #define	BOOTHOWTO	0
123 #endif
124 int	boothowto = BOOTHOWTO;	/* initialized so that it can be patched */
125 SYSCTL_INT(_debug, OID_AUTO, boothowto, CTLFLAG_RD, &boothowto, 0,
126 	"Boot control flags, passed from loader");
127 
128 #ifndef BOOTVERBOSE
129 #define	BOOTVERBOSE	0
130 #endif
131 int	bootverbose = BOOTVERBOSE;
132 SYSCTL_INT(_debug, OID_AUTO, bootverbose, CTLFLAG_RW, &bootverbose, 0,
133 	"Control the output of verbose kernel messages");
134 
135 #ifdef VERBOSE_SYSINIT
136 /*
137  * We'll use the defined value of VERBOSE_SYSINIT from the kernel config to
138  * dictate the default VERBOSE_SYSINIT behavior.  Significant values for this
139  * option and associated tunable are:
140  * - 0, 'compiled in but silent by default'
141  * - 1, 'compiled in but verbose by default' (default)
142  */
143 int	verbose_sysinit = VERBOSE_SYSINIT;
144 TUNABLE_INT("debug.verbose_sysinit", &verbose_sysinit);
145 #endif
146 
147 #ifdef INVARIANTS
148 FEATURE(invariants, "Kernel compiled with INVARIANTS, may affect performance");
149 #endif
150 
151 /*
152  * This ensures that there is at least one entry so that the sysinit_set
153  * symbol is not undefined.  A sybsystem ID of SI_SUB_DUMMY is never
154  * executed.
155  */
156 SYSINIT(placeholder, SI_SUB_DUMMY, SI_ORDER_ANY, NULL, NULL);
157 
158 /*
159  * The sysinit table itself.  Items are checked off as the are run.
160  * If we want to register new sysinit types, add them to newsysinit.
161  */
162 SET_DECLARE(sysinit_set, struct sysinit);
163 struct sysinit **sysinit, **sysinit_end;
164 struct sysinit **newsysinit, **newsysinit_end;
165 
166 /*
167  * Merge a new sysinit set into the current set, reallocating it if
168  * necessary.  This can only be called after malloc is running.
169  */
170 void
171 sysinit_add(struct sysinit **set, struct sysinit **set_end)
172 {
173 	struct sysinit **newset;
174 	struct sysinit **sipp;
175 	struct sysinit **xipp;
176 	int count;
177 
178 	count = set_end - set;
179 	if (newsysinit)
180 		count += newsysinit_end - newsysinit;
181 	else
182 		count += sysinit_end - sysinit;
183 	newset = malloc(count * sizeof(*sipp), M_TEMP, M_NOWAIT);
184 	if (newset == NULL)
185 		panic("cannot malloc for sysinit");
186 	xipp = newset;
187 	if (newsysinit)
188 		for (sipp = newsysinit; sipp < newsysinit_end; sipp++)
189 			*xipp++ = *sipp;
190 	else
191 		for (sipp = sysinit; sipp < sysinit_end; sipp++)
192 			*xipp++ = *sipp;
193 	for (sipp = set; sipp < set_end; sipp++)
194 		*xipp++ = *sipp;
195 	if (newsysinit)
196 		free(newsysinit, M_TEMP);
197 	newsysinit = newset;
198 	newsysinit_end = newset + count;
199 }
200 
201 #if defined (DDB) && defined(VERBOSE_SYSINIT)
202 static const char *
203 symbol_name(vm_offset_t va, db_strategy_t strategy)
204 {
205 	const char *name;
206 	c_db_sym_t sym;
207 	db_expr_t  offset;
208 
209 	if (va == 0)
210 		return (NULL);
211 	sym = db_search_symbol(va, strategy, &offset);
212 	if (offset != 0)
213 		return (NULL);
214 	db_symbol_values(sym, &name, NULL);
215 	return (name);
216 }
217 #endif
218 
219 /*
220  * System startup; initialize the world, create process 0, mount root
221  * filesystem, and fork to create init and pagedaemon.  Most of the
222  * hard work is done in the lower-level initialization routines including
223  * startup(), which does memory initialization and autoconfiguration.
224  *
225  * This allows simple addition of new kernel subsystems that require
226  * boot time initialization.  It also allows substitution of subsystem
227  * (for instance, a scheduler, kernel profiler, or VM system) by object
228  * module.  Finally, it allows for optional "kernel threads".
229  */
230 void
231 mi_startup(void)
232 {
233 
234 	struct sysinit **sipp;	/* system initialization*/
235 	struct sysinit **xipp;	/* interior loop of sort*/
236 	struct sysinit *save;	/* bubble*/
237 
238 	int last;
239 #if defined(VERBOSE_SYSINIT)
240 	int verbose;
241 #endif
242 
243 	TSENTER();
244 
245 	if (boothowto & RB_VERBOSE)
246 		bootverbose++;
247 
248 	if (sysinit == NULL) {
249 		sysinit = SET_BEGIN(sysinit_set);
250 		sysinit_end = SET_LIMIT(sysinit_set);
251 	}
252 
253 restart:
254 	/*
255 	 * Perform a bubble sort of the system initialization objects by
256 	 * their subsystem (primary key) and order (secondary key).
257 	 */
258 	for (sipp = sysinit; sipp < sysinit_end; sipp++) {
259 		for (xipp = sipp + 1; xipp < sysinit_end; xipp++) {
260 			if ((*sipp)->subsystem < (*xipp)->subsystem ||
261 			     ((*sipp)->subsystem == (*xipp)->subsystem &&
262 			      (*sipp)->order <= (*xipp)->order))
263 				continue;	/* skip*/
264 			save = *sipp;
265 			*sipp = *xipp;
266 			*xipp = save;
267 		}
268 	}
269 
270 	last = SI_SUB_COPYRIGHT;
271 #if defined(VERBOSE_SYSINIT)
272 	verbose = 0;
273 #if !defined(DDB)
274 	printf("VERBOSE_SYSINIT: DDB not enabled, symbol lookups disabled.\n");
275 #endif
276 #endif
277 
278 	/*
279 	 * Traverse the (now) ordered list of system initialization tasks.
280 	 * Perform each task, and continue on to the next task.
281 	 */
282 	for (sipp = sysinit; sipp < sysinit_end; sipp++) {
283 		if ((*sipp)->subsystem == SI_SUB_DUMMY)
284 			continue;	/* skip dummy task(s)*/
285 
286 		if ((*sipp)->subsystem == SI_SUB_DONE)
287 			continue;
288 
289 		if ((*sipp)->subsystem > last)
290 			BOOTTRACE_INIT("sysinit 0x%7x", (*sipp)->subsystem);
291 
292 #if defined(VERBOSE_SYSINIT)
293 		if ((*sipp)->subsystem > last && verbose_sysinit != 0) {
294 			verbose = 1;
295 			printf("subsystem %x\n", last);
296 		}
297 		if (verbose) {
298 #if defined(DDB)
299 			const char *func, *data;
300 
301 			func = symbol_name((vm_offset_t)(*sipp)->func,
302 			    DB_STGY_PROC);
303 			data = symbol_name((vm_offset_t)(*sipp)->udata,
304 			    DB_STGY_ANY);
305 			if (func != NULL && data != NULL)
306 				printf("   %s(&%s)... ", func, data);
307 			else if (func != NULL)
308 				printf("   %s(%p)... ", func, (*sipp)->udata);
309 			else
310 #endif
311 				printf("   %p(%p)... ", (*sipp)->func,
312 				    (*sipp)->udata);
313 		}
314 #endif
315 
316 		/* Call function */
317 		(*((*sipp)->func))((*sipp)->udata);
318 
319 #if defined(VERBOSE_SYSINIT)
320 		if (verbose)
321 			printf("done.\n");
322 #endif
323 
324 		/* Check off the one we're just done */
325 		last = (*sipp)->subsystem;
326 		(*sipp)->subsystem = SI_SUB_DONE;
327 
328 		/* Check if we've installed more sysinit items via KLD */
329 		if (newsysinit != NULL) {
330 			if (sysinit != SET_BEGIN(sysinit_set))
331 				free(sysinit, M_TEMP);
332 			sysinit = newsysinit;
333 			sysinit_end = newsysinit_end;
334 			newsysinit = NULL;
335 			newsysinit_end = NULL;
336 			goto restart;
337 		}
338 	}
339 
340 	TSEXIT();	/* Here so we don't overlap with start_init. */
341 	BOOTTRACE("mi_startup done");
342 
343 	mtx_assert(&Giant, MA_OWNED | MA_NOTRECURSED);
344 	mtx_unlock(&Giant);
345 
346 	/*
347 	 * Now hand over this thread to swapper.
348 	 */
349 	swapper();
350 	/* NOTREACHED*/
351 }
352 
353 static void
354 print_caddr_t(void *data)
355 {
356 	printf("%s", (char *)data);
357 }
358 
359 static void
360 print_version(void *data __unused)
361 {
362 	int len;
363 
364 	/* Strip a trailing newline from version. */
365 	len = strlen(version);
366 	while (len > 0 && version[len - 1] == '\n')
367 		len--;
368 	printf("%.*s %s\n", len, version, machine);
369 	printf("%s\n", compiler_version);
370 }
371 
372 SYSINIT(announce, SI_SUB_COPYRIGHT, SI_ORDER_FIRST, print_caddr_t,
373     copyright);
374 SYSINIT(trademark, SI_SUB_COPYRIGHT, SI_ORDER_SECOND, print_caddr_t,
375     trademark);
376 SYSINIT(version, SI_SUB_COPYRIGHT, SI_ORDER_THIRD, print_version, NULL);
377 
378 #ifdef WITNESS
379 static char wit_warn[] =
380      "WARNING: WITNESS option enabled, expect reduced performance.\n";
381 SYSINIT(witwarn, SI_SUB_COPYRIGHT, SI_ORDER_FOURTH,
382    print_caddr_t, wit_warn);
383 SYSINIT(witwarn2, SI_SUB_LAST, SI_ORDER_FOURTH,
384    print_caddr_t, wit_warn);
385 #endif
386 
387 #ifdef DIAGNOSTIC
388 static char diag_warn[] =
389      "WARNING: DIAGNOSTIC option enabled, expect reduced performance.\n";
390 SYSINIT(diagwarn, SI_SUB_COPYRIGHT, SI_ORDER_FIFTH,
391     print_caddr_t, diag_warn);
392 SYSINIT(diagwarn2, SI_SUB_LAST, SI_ORDER_FIFTH,
393     print_caddr_t, diag_warn);
394 #endif
395 
396 static int
397 null_fetch_syscall_args(struct thread *td __unused)
398 {
399 
400 	panic("null_fetch_syscall_args");
401 }
402 
403 static void
404 null_set_syscall_retval(struct thread *td __unused, int error __unused)
405 {
406 
407 	panic("null_set_syscall_retval");
408 }
409 
410 static void
411 null_set_fork_retval(struct thread *td __unused)
412 {
413 
414 }
415 
416 struct sysentvec null_sysvec = {
417 	.sv_size	= 0,
418 	.sv_table	= NULL,
419 	.sv_fixup	= NULL,
420 	.sv_sendsig	= NULL,
421 	.sv_sigcode	= NULL,
422 	.sv_szsigcode	= NULL,
423 	.sv_name	= "null",
424 	.sv_coredump	= NULL,
425 	.sv_imgact_try	= NULL,
426 	.sv_minsigstksz	= 0,
427 	.sv_minuser	= VM_MIN_ADDRESS,
428 	.sv_maxuser	= VM_MAXUSER_ADDRESS,
429 	.sv_usrstack	= USRSTACK,
430 	.sv_psstrings	= PS_STRINGS,
431 	.sv_psstringssz	= sizeof(struct ps_strings),
432 	.sv_stackprot	= VM_PROT_ALL,
433 	.sv_copyout_strings	= NULL,
434 	.sv_setregs	= NULL,
435 	.sv_fixlimit	= NULL,
436 	.sv_maxssiz	= NULL,
437 	.sv_flags	= 0,
438 	.sv_set_syscall_retval = null_set_syscall_retval,
439 	.sv_fetch_syscall_args = null_fetch_syscall_args,
440 	.sv_syscallnames = NULL,
441 	.sv_schedtail	= NULL,
442 	.sv_thread_detach = NULL,
443 	.sv_trap	= NULL,
444 	.sv_set_fork_retval = null_set_fork_retval,
445 	.sv_regset_begin = NULL,
446 	.sv_regset_end  = NULL,
447 };
448 
449 /*
450  * The two following SYSINIT's are proc0 specific glue code.  I am not
451  * convinced that they can not be safely combined, but their order of
452  * operation has been maintained as the same as the original init_main.c
453  * for right now.
454  */
455 /* ARGSUSED*/
456 static void
457 proc0_init(void *dummy __unused)
458 {
459 	struct proc *p;
460 	struct thread *td;
461 	struct ucred *newcred;
462 	struct uidinfo tmpuinfo;
463 	struct loginclass tmplc = {
464 		.lc_name = "",
465 	};
466 	vm_paddr_t pageablemem;
467 	int i;
468 
469 	GIANT_REQUIRED;
470 	p = &proc0;
471 	td = &thread0;
472 
473 	/*
474 	 * Initialize magic number and osrel.
475 	 */
476 	p->p_magic = P_MAGIC;
477 	p->p_osrel = osreldate;
478 
479 	/*
480 	 * Initialize thread and process structures.
481 	 */
482 	procinit();	/* set up proc zone */
483 	threadinit();	/* set up UMA zones */
484 
485 	/*
486 	 * Initialise scheduler resources.
487 	 * Add scheduler specific parts to proc, thread as needed.
488 	 */
489 	schedinit();	/* scheduler gets its house in order */
490 
491 	/*
492 	 * Create process 0 (the swapper).
493 	 */
494 	LIST_INSERT_HEAD(&allproc, p, p_list);
495 	LIST_INSERT_HEAD(PIDHASH(0), p, p_hash);
496 	mtx_init(&pgrp0.pg_mtx, "process group", NULL, MTX_DEF | MTX_DUPOK);
497 	p->p_pgrp = &pgrp0;
498 	LIST_INSERT_HEAD(PGRPHASH(0), &pgrp0, pg_hash);
499 	LIST_INIT(&pgrp0.pg_members);
500 	LIST_INSERT_HEAD(&pgrp0.pg_members, p, p_pglist);
501 
502 	pgrp0.pg_session = &session0;
503 	mtx_init(&session0.s_mtx, "session", NULL, MTX_DEF);
504 	refcount_init(&session0.s_count, 1);
505 	session0.s_leader = p;
506 
507 	p->p_sysent = &null_sysvec;
508 	p->p_flag = P_SYSTEM | P_INMEM | P_KPROC;
509 	p->p_flag2 = 0;
510 	p->p_state = PRS_NORMAL;
511 	p->p_klist = knlist_alloc(&p->p_mtx);
512 	STAILQ_INIT(&p->p_ktr);
513 	p->p_nice = NZERO;
514 	td->td_tid = THREAD0_TID;
515 	tidhash_add(td);
516 	TD_SET_STATE(td, TDS_RUNNING);
517 	td->td_pri_class = PRI_TIMESHARE;
518 	td->td_user_pri = PUSER;
519 	td->td_base_user_pri = PUSER;
520 	td->td_lend_user_pri = PRI_MAX;
521 	td->td_priority = PVM;
522 	td->td_base_pri = PVM;
523 	td->td_oncpu = curcpu;
524 	td->td_flags = TDF_INMEM;
525 	td->td_pflags = TDP_KTHREAD;
526 	td->td_cpuset = cpuset_thread0();
527 	td->td_domain.dr_policy = td->td_cpuset->cs_domain;
528 	prison0_init();
529 	p->p_peers = 0;
530 	p->p_leader = p;
531 	p->p_reaper = p;
532 	p->p_treeflag |= P_TREE_REAPER;
533 	LIST_INIT(&p->p_reaplist);
534 
535 	strncpy(p->p_comm, "kernel", sizeof (p->p_comm));
536 	strncpy(td->td_name, "swapper", sizeof (td->td_name));
537 
538 	callout_init_mtx(&p->p_itcallout, &p->p_mtx, 0);
539 	callout_init_mtx(&p->p_limco, &p->p_mtx, 0);
540 	callout_init(&td->td_slpcallout, 1);
541 	TAILQ_INIT(&p->p_kqtim_stop);
542 
543 	/* Create credentials. */
544 	newcred = crget();
545 	newcred->cr_ngroups = 1;	/* group 0 */
546 	/* A hack to prevent uifind from tripping over NULL pointers. */
547 	curthread->td_ucred = newcred;
548 	tmpuinfo.ui_uid = 1;
549 	newcred->cr_uidinfo = newcred->cr_ruidinfo = &tmpuinfo;
550 	newcred->cr_uidinfo = uifind(0);
551 	newcred->cr_ruidinfo = uifind(0);
552 	newcred->cr_loginclass = &tmplc;
553 	newcred->cr_loginclass = loginclass_find("default");
554 	/* End hack. creds get properly set later with thread_cow_get_proc */
555 	curthread->td_ucred = NULL;
556 	newcred->cr_prison = &prison0;
557 	newcred->cr_users++; /* avoid assertion failure */
558 	proc_set_cred_init(p, newcred);
559 	newcred->cr_users--;
560 	crfree(newcred);
561 #ifdef AUDIT
562 	audit_cred_kproc0(newcred);
563 #endif
564 #ifdef MAC
565 	mac_cred_create_swapper(newcred);
566 #endif
567 	/* Create sigacts. */
568 	p->p_sigacts = sigacts_alloc();
569 
570 	/* Initialize signal state for process 0. */
571 	siginit(&proc0);
572 
573 	/* Create the file descriptor table. */
574 	p->p_pd = pdinit(NULL, false);
575 	p->p_fd = fdinit();
576 	p->p_fdtol = NULL;
577 
578 	/* Create the limits structures. */
579 	p->p_limit = lim_alloc();
580 	for (i = 0; i < RLIM_NLIMITS; i++)
581 		p->p_limit->pl_rlimit[i].rlim_cur =
582 		    p->p_limit->pl_rlimit[i].rlim_max = RLIM_INFINITY;
583 	p->p_limit->pl_rlimit[RLIMIT_NOFILE].rlim_cur =
584 	    p->p_limit->pl_rlimit[RLIMIT_NOFILE].rlim_max = maxfiles;
585 	p->p_limit->pl_rlimit[RLIMIT_NPROC].rlim_cur =
586 	    p->p_limit->pl_rlimit[RLIMIT_NPROC].rlim_max = maxproc;
587 	p->p_limit->pl_rlimit[RLIMIT_DATA].rlim_cur = dfldsiz;
588 	p->p_limit->pl_rlimit[RLIMIT_DATA].rlim_max = maxdsiz;
589 	p->p_limit->pl_rlimit[RLIMIT_STACK].rlim_cur = dflssiz;
590 	p->p_limit->pl_rlimit[RLIMIT_STACK].rlim_max = maxssiz;
591 	/* Cast to avoid overflow on i386/PAE. */
592 	pageablemem = ptoa((vm_paddr_t)vm_free_count());
593 	p->p_limit->pl_rlimit[RLIMIT_RSS].rlim_cur =
594 	    p->p_limit->pl_rlimit[RLIMIT_RSS].rlim_max = pageablemem;
595 	p->p_limit->pl_rlimit[RLIMIT_MEMLOCK].rlim_cur = pageablemem / 3;
596 	p->p_limit->pl_rlimit[RLIMIT_MEMLOCK].rlim_max = pageablemem;
597 	p->p_cpulimit = RLIM_INFINITY;
598 
599 	PROC_LOCK(p);
600 	thread_cow_get_proc(td, p);
601 	PROC_UNLOCK(p);
602 
603 	/* Initialize resource accounting structures. */
604 	racct_create(&p->p_racct);
605 
606 	p->p_stats = pstats_alloc();
607 
608 	/* Allocate a prototype map so we have something to fork. */
609 	p->p_vmspace = &vmspace0;
610 	refcount_init(&vmspace0.vm_refcnt, 1);
611 	pmap_pinit0(vmspace_pmap(&vmspace0));
612 
613 	/*
614 	 * proc0 is not expected to enter usermode, so there is no special
615 	 * handling for sv_minuser here, like is done for exec_new_vmspace().
616 	 */
617 	vm_map_init(&vmspace0.vm_map, vmspace_pmap(&vmspace0),
618 	    p->p_sysent->sv_minuser, p->p_sysent->sv_maxuser);
619 
620 	/*
621 	 * Call the init and ctor for the new thread and proc.  We wait
622 	 * to do this until all other structures are fairly sane.
623 	 */
624 	EVENTHANDLER_DIRECT_INVOKE(process_init, p);
625 	EVENTHANDLER_DIRECT_INVOKE(thread_init, td);
626 #ifdef KDTRACE_HOOKS
627 	kdtrace_proc_ctor(p);
628 	kdtrace_thread_ctor(td);
629 #endif
630 	EVENTHANDLER_DIRECT_INVOKE(process_ctor, p);
631 	EVENTHANDLER_DIRECT_INVOKE(thread_ctor, td);
632 
633 	/*
634 	 * Charge root for one process.
635 	 */
636 	(void)chgproccnt(p->p_ucred->cr_ruidinfo, 1, 0);
637 	PROC_LOCK(p);
638 	racct_add_force(p, RACCT_NPROC, 1);
639 	PROC_UNLOCK(p);
640 }
641 SYSINIT(p0init, SI_SUB_INTRINSIC, SI_ORDER_FIRST, proc0_init, NULL);
642 
643 /* ARGSUSED*/
644 static void
645 proc0_post(void *dummy __unused)
646 {
647 	struct proc *p;
648 	struct rusage ru;
649 	struct thread *td;
650 
651 	/*
652 	 * Now we can look at the time, having had a chance to verify the
653 	 * time from the filesystem.  Pretend that proc0 started now.
654 	 */
655 	sx_slock(&allproc_lock);
656 	FOREACH_PROC_IN_SYSTEM(p) {
657 		PROC_LOCK(p);
658 		if (p->p_state == PRS_NEW) {
659 			PROC_UNLOCK(p);
660 			continue;
661 		}
662 		microuptime(&p->p_stats->p_start);
663 		PROC_STATLOCK(p);
664 		rufetch(p, &ru);	/* Clears thread stats */
665 		p->p_rux.rux_runtime = 0;
666 		p->p_rux.rux_uticks = 0;
667 		p->p_rux.rux_sticks = 0;
668 		p->p_rux.rux_iticks = 0;
669 		PROC_STATUNLOCK(p);
670 		FOREACH_THREAD_IN_PROC(p, td) {
671 			td->td_runtime = 0;
672 		}
673 		PROC_UNLOCK(p);
674 	}
675 	sx_sunlock(&allproc_lock);
676 	PCPU_SET(switchtime, cpu_ticks());
677 	PCPU_SET(switchticks, ticks);
678 }
679 SYSINIT(p0post, SI_SUB_INTRINSIC_POST, SI_ORDER_FIRST, proc0_post, NULL);
680 
681 /*
682  ***************************************************************************
683  ****
684  **** The following SYSINIT's and glue code should be moved to the
685  **** respective files on a per subsystem basis.
686  ****
687  ***************************************************************************
688  */
689 
690 /*
691  * List of paths to try when searching for "init".
692  */
693 static char init_path[MAXPATHLEN] =
694 #ifdef	INIT_PATH
695     __XSTRING(INIT_PATH);
696 #else
697     "/sbin/init:/sbin/oinit:/sbin/init.bak:/rescue/init";
698 #endif
699 SYSCTL_STRING(_kern, OID_AUTO, init_path, CTLFLAG_RD, init_path, 0,
700 	"Path used to search the init process");
701 
702 /*
703  * Shutdown timeout of init(8).
704  * Unused within kernel, but used to control init(8), hence do not remove.
705  */
706 #ifndef INIT_SHUTDOWN_TIMEOUT
707 #define INIT_SHUTDOWN_TIMEOUT 120
708 #endif
709 static int init_shutdown_timeout = INIT_SHUTDOWN_TIMEOUT;
710 SYSCTL_INT(_kern, OID_AUTO, init_shutdown_timeout,
711 	CTLFLAG_RW, &init_shutdown_timeout, 0, "Shutdown timeout of init(8). "
712 	"Unused within kernel, but used to control init(8)");
713 
714 /*
715  * Start the initial user process; try exec'ing each pathname in init_path.
716  * The program is invoked with one argument containing the boot flags.
717  */
718 static void
719 start_init(void *dummy)
720 {
721 	struct image_args args;
722 	int error;
723 	char *var, *path;
724 	char *free_init_path, *tmp_init_path;
725 	struct thread *td;
726 	struct proc *p;
727 	struct vmspace *oldvmspace;
728 
729 	TSENTER();	/* Here so we don't overlap with mi_startup. */
730 
731 	td = curthread;
732 	p = td->td_proc;
733 
734 	vfs_mountroot();
735 
736 	/* Wipe GELI passphrase from the environment. */
737 	kern_unsetenv("kern.geom.eli.passphrase");
738 
739 	/* For Multicons, report which console is primary to both */
740 	if (boothowto & RB_MULTIPLE) {
741 		if (boothowto & RB_SERIAL)
742 			printf("Dual Console: Serial Primary, Video Secondary\n");
743 		else
744 			printf("Dual Console: Video Primary, Serial Secondary\n");
745 	}
746 
747 	if ((var = kern_getenv("init_path")) != NULL) {
748 		strlcpy(init_path, var, sizeof(init_path));
749 		freeenv(var);
750 	}
751 	free_init_path = tmp_init_path = strdup(init_path, M_TEMP);
752 
753 	while ((path = strsep(&tmp_init_path, ":")) != NULL) {
754 		if (bootverbose)
755 			printf("start_init: trying %s\n", path);
756 
757 		memset(&args, 0, sizeof(args));
758 		error = exec_alloc_args(&args);
759 		if (error != 0)
760 			panic("%s: Can't allocate space for init arguments %d",
761 			    __func__, error);
762 
763 		error = exec_args_add_fname(&args, path, UIO_SYSSPACE);
764 		if (error != 0)
765 			panic("%s: Can't add fname %d", __func__, error);
766 		error = exec_args_add_arg(&args, path, UIO_SYSSPACE);
767 		if (error != 0)
768 			panic("%s: Can't add argv[0] %d", __func__, error);
769 		if (boothowto & RB_SINGLE)
770 			error = exec_args_add_arg(&args, "-s", UIO_SYSSPACE);
771 		if (error != 0)
772 			panic("%s: Can't add argv[0] %d", __func__, error);
773 
774 		/*
775 		 * Now try to exec the program.  If can't for any reason
776 		 * other than it doesn't exist, complain.
777 		 *
778 		 * Otherwise, return via fork_trampoline() all the way
779 		 * to user mode as init!
780 		 */
781 		KASSERT((td->td_pflags & TDP_EXECVMSPC) == 0,
782 		    ("nested execve"));
783 		oldvmspace = p->p_vmspace;
784 		error = kern_execve(td, &args, NULL, oldvmspace);
785 		KASSERT(error != 0,
786 		    ("kern_execve returned success, not EJUSTRETURN"));
787 		if (error == EJUSTRETURN) {
788 			exec_cleanup(td, oldvmspace);
789 			free(free_init_path, M_TEMP);
790 			TSEXIT();
791 			return;
792 		}
793 		if (error != ENOENT)
794 			printf("exec %s: error %d\n", path, error);
795 	}
796 	free(free_init_path, M_TEMP);
797 	printf("init: not found in path %s\n", init_path);
798 	panic("no init");
799 }
800 
801 /*
802  * Like kproc_create(), but runs in its own address space.  We do this
803  * early to reserve pid 1.  Note special case - do not make it
804  * runnable yet, init execution is started when userspace can be served.
805  */
806 static void
807 create_init(const void *udata __unused)
808 {
809 	struct fork_req fr;
810 	struct ucred *newcred, *oldcred;
811 	struct thread *td;
812 	int error;
813 
814 	bzero(&fr, sizeof(fr));
815 	fr.fr_flags = RFFDG | RFPROC | RFSTOPPED;
816 	fr.fr_procp = &initproc;
817 	error = fork1(&thread0, &fr);
818 	if (error)
819 		panic("cannot fork init: %d\n", error);
820 	KASSERT(initproc->p_pid == 1, ("create_init: initproc->p_pid != 1"));
821 	/* divorce init's credentials from the kernel's */
822 	newcred = crget();
823 	sx_xlock(&proctree_lock);
824 	PROC_LOCK(initproc);
825 	initproc->p_flag |= P_SYSTEM | P_INMEM;
826 	initproc->p_treeflag |= P_TREE_REAPER;
827 	oldcred = initproc->p_ucred;
828 	crcopy(newcred, oldcred);
829 #ifdef MAC
830 	mac_cred_create_init(newcred);
831 #endif
832 #ifdef AUDIT
833 	audit_cred_proc1(newcred);
834 #endif
835 	proc_set_cred(initproc, newcred);
836 	td = FIRST_THREAD_IN_PROC(initproc);
837 	crcowfree(td);
838 	td->td_realucred = crcowget(initproc->p_ucred);
839 	td->td_ucred = td->td_realucred;
840 	PROC_UNLOCK(initproc);
841 	sx_xunlock(&proctree_lock);
842 	crfree(oldcred);
843 	cpu_fork_kthread_handler(FIRST_THREAD_IN_PROC(initproc),
844 	    start_init, NULL);
845 }
846 SYSINIT(init, SI_SUB_CREATE_INIT, SI_ORDER_FIRST, create_init, NULL);
847 
848 /*
849  * Make it runnable now.
850  */
851 static void
852 kick_init(const void *udata __unused)
853 {
854 	struct thread *td;
855 
856 	td = FIRST_THREAD_IN_PROC(initproc);
857 	thread_lock(td);
858 	TD_SET_CAN_RUN(td);
859 	sched_add(td, SRQ_BORING);
860 }
861 SYSINIT(kickinit, SI_SUB_KTHREAD_INIT, SI_ORDER_MIDDLE, kick_init, NULL);
862 
863 /*
864  * DDB(4).
865  */
866 #ifdef DDB
867 static void
868 db_show_print_syinit(struct sysinit *sip, bool ddb)
869 {
870 	const char *sname, *funcname;
871 	c_db_sym_t sym;
872 	db_expr_t  offset;
873 
874 #define xprint(...)							\
875 	if (ddb)							\
876 		db_printf(__VA_ARGS__);					\
877 	else								\
878 		printf(__VA_ARGS__)
879 
880 	if (sip == NULL) {
881 		xprint("%s: no sysinit * given\n", __func__);
882 		return;
883 	}
884 
885 	sym = db_search_symbol((vm_offset_t)sip, DB_STGY_ANY, &offset);
886 	db_symbol_values(sym, &sname, NULL);
887 	sym = db_search_symbol((vm_offset_t)sip->func, DB_STGY_PROC, &offset);
888 	db_symbol_values(sym, &funcname, NULL);
889 	xprint("%s(%p)\n", (sname != NULL) ? sname : "", sip);
890 	xprint("  %#08x %#08x\n", sip->subsystem, sip->order);
891 	xprint("  %p(%s)(%p)\n",
892 	    sip->func, (funcname != NULL) ? funcname : "", sip->udata);
893 #undef xprint
894 }
895 
896 DB_SHOW_COMMAND_FLAGS(sysinit, db_show_sysinit, DB_CMD_MEMSAFE)
897 {
898 	struct sysinit **sipp;
899 
900 	db_printf("SYSINIT vs Name(Ptr)\n");
901 	db_printf("  Subsystem  Order\n");
902 	db_printf("  Function(Name)(Arg)\n");
903 	for (sipp = sysinit; sipp < sysinit_end; sipp++) {
904 		db_show_print_syinit(*sipp, true);
905 		if (db_pager_quit)
906 			break;
907 	}
908 }
909 #endif /* DDB */
910