xref: /freebsd/sys/kern/init_main.c (revision cbb3ec25236ba72f91cbdf23f8b78b9d1af0cedf)
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 #include "opt_ddb.h"
48 #include "opt_kdb.h"
49 #include "opt_init_path.h"
50 #include "opt_verbose_sysinit.h"
51 
52 #include <sys/param.h>
53 #include <sys/systm.h>
54 #include <sys/boottrace.h>
55 #include <sys/conf.h>
56 #include <sys/cpuset.h>
57 #include <sys/dtrace_bsd.h>
58 #include <sys/epoch.h>
59 #include <sys/eventhandler.h>
60 #include <sys/exec.h>
61 #include <sys/file.h>
62 #include <sys/filedesc.h>
63 #include <sys/imgact.h>
64 #include <sys/jail.h>
65 #include <sys/kernel.h>
66 #include <sys/ktr.h>
67 #include <sys/lock.h>
68 #include <sys/loginclass.h>
69 #include <sys/malloc.h>
70 #include <sys/mount.h>
71 #include <sys/mutex.h>
72 #include <sys/proc.h>
73 #include <sys/racct.h>
74 #include <sys/reboot.h>
75 #include <sys/resourcevar.h>
76 #include <sys/queue.h>
77 #include <sys/queue_mergesort.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 linker set compiled into the kernel.  These are placed onto the
160  * sysinit list by mi_startup; sysinit_add can add (e.g., from klds) additional
161  * sysinits to the linked list but the linker set here does not change.
162  */
163 SET_DECLARE(sysinit_set, struct sysinit);
164 
165 /*
166  * The sysinit list itself.  Items are removed as they are run.
167  */
168 static SLIST_HEAD(sysinitlist, sysinit) sysinit_list;
169 
170 /*
171  * Compare two sysinits; return -1, 0, or 1 if a comes before, at the same time
172  * as, or after b.
173  */
174 static int
175 sysinit_compar(struct sysinit *a, struct sysinit *b, void *thunk __unused)
176 {
177 
178 	if (a->subsystem < b->subsystem)
179 		return (-1);
180 	if (a->subsystem > b->subsystem)
181 		return (1);
182 	if (a->order < b->order)
183 		return (-1);
184 	if (a->order > b->order)
185 		return (1);
186 	return (0);
187 }
188 
189 static void
190 sysinit_mklist(struct sysinitlist *list, struct sysinit **set,
191     struct sysinit **set_end)
192 {
193 	struct sysinit **sipp;
194 
195 	TSENTER();
196 	TSENTER2("listify");
197 	SLIST_INIT(list);
198 	for (sipp = set; sipp < set_end; sipp++)
199 		SLIST_INSERT_HEAD(list, *sipp, next);
200 	TSEXIT2("listify");
201 	TSENTER2("mergesort");
202 	SLIST_MERGESORT(list, NULL, sysinit_compar, sysinit, next);
203 	TSEXIT2("mergesort");
204 	TSEXIT();
205 }
206 
207 /*
208  * Merge a new sysinit set into the sysinit list.
209  */
210 void
211 sysinit_add(struct sysinit **set, struct sysinit **set_end)
212 {
213 	struct sysinitlist new_list;
214 
215 	TSENTER();
216 
217 	/* Construct a sorted list from the new sysinits. */
218 	sysinit_mklist(&new_list, set, set_end);
219 
220 	/* Merge the new list into the existing one. */
221 	TSENTER2("SLIST_MERGE");
222 	SLIST_MERGE(&sysinit_list, &new_list, NULL, sysinit_compar, sysinit, next);
223 	TSEXIT2("SLIST_MERGE");
224 
225 	TSEXIT();
226 }
227 
228 #if defined (DDB) && defined(VERBOSE_SYSINIT)
229 static const char *
230 symbol_name(vm_offset_t va, db_strategy_t strategy)
231 {
232 	const char *name;
233 	c_db_sym_t sym;
234 	db_expr_t  offset;
235 
236 	if (va == 0)
237 		return (NULL);
238 	sym = db_search_symbol(va, strategy, &offset);
239 	if (offset != 0)
240 		return (NULL);
241 	db_symbol_values(sym, &name, NULL);
242 	return (name);
243 }
244 #endif
245 
246 /*
247  * System startup; initialize the world, create process 0, mount root
248  * filesystem, and fork to create init and pagedaemon.  Most of the
249  * hard work is done in the lower-level initialization routines including
250  * startup(), which does memory initialization and autoconfiguration.
251  *
252  * This allows simple addition of new kernel subsystems that require
253  * boot time initialization.  It also allows substitution of subsystem
254  * (for instance, a scheduler, kernel profiler, or VM system) by object
255  * module.  Finally, it allows for optional "kernel threads".
256  */
257 void
258 mi_startup(void)
259 {
260 
261 	struct sysinit *sip;
262 	int last;
263 #if defined(VERBOSE_SYSINIT)
264 	int verbose;
265 #endif
266 
267 	TSENTER();
268 
269 	if (boothowto & RB_VERBOSE)
270 		bootverbose++;
271 
272 	/* Construct and sort sysinit list. */
273 	sysinit_mklist(&sysinit_list, SET_BEGIN(sysinit_set), SET_LIMIT(sysinit_set));
274 
275 	last = SI_SUB_COPYRIGHT;
276 #if defined(VERBOSE_SYSINIT)
277 	verbose = 0;
278 #if !defined(DDB)
279 	printf("VERBOSE_SYSINIT: DDB not enabled, symbol lookups disabled.\n");
280 #endif
281 #endif
282 
283 	/*
284 	 * Perform each system initialization task from the ordered list.  Note
285 	 * that if sysinit_list is modified (e.g. by a KLD) we will nonetheless
286 	 * always perform the earlist-sorted sysinit at each step; using the
287 	 * SLIST_FOREACH macro would result in items being skipped if inserted
288 	 * earlier than the "current item".
289 	 */
290 	while ((sip = SLIST_FIRST(&sysinit_list)) != NULL) {
291 		SLIST_REMOVE_HEAD(&sysinit_list, next);
292 
293 		if (sip->subsystem == SI_SUB_DUMMY)
294 			continue;	/* skip dummy task(s)*/
295 
296 		if (sip->subsystem > last)
297 			BOOTTRACE_INIT("sysinit 0x%7x", sip->subsystem);
298 
299 #if defined(VERBOSE_SYSINIT)
300 		if (sip->subsystem > last && verbose_sysinit != 0) {
301 			verbose = 1;
302 			printf("subsystem %x\n", last);
303 		}
304 		if (verbose) {
305 #if defined(DDB)
306 			const char *func, *data;
307 
308 			func = symbol_name((vm_offset_t)sip->func,
309 			    DB_STGY_PROC);
310 			data = symbol_name((vm_offset_t)sip->udata,
311 			    DB_STGY_ANY);
312 			if (func != NULL && data != NULL)
313 				printf("   %s(&%s)... ", func, data);
314 			else if (func != NULL)
315 				printf("   %s(%p)... ", func, sip->udata);
316 			else
317 #endif
318 				printf("   %p(%p)... ", sip->func,
319 				    sip->udata);
320 		}
321 #endif
322 
323 		/* Call function */
324 		(*(sip->func))(sip->udata);
325 
326 #if defined(VERBOSE_SYSINIT)
327 		if (verbose)
328 			printf("done.\n");
329 #endif
330 
331 		/* Check off the one we're just done */
332 		last = sip->subsystem;
333 	}
334 
335 	TSEXIT();	/* Here so we don't overlap with start_init. */
336 	BOOTTRACE("mi_startup done");
337 
338 	mtx_assert(&Giant, MA_OWNED | MA_NOTRECURSED);
339 	mtx_unlock(&Giant);
340 
341 	/*
342 	 * Now hand over this thread to swapper.
343 	 */
344 	swapper();
345 	/* NOTREACHED*/
346 }
347 
348 static void
349 print_caddr_t(void *data)
350 {
351 	printf("%s", (char *)data);
352 }
353 
354 static void
355 print_version(void *data __unused)
356 {
357 	int len;
358 
359 	/* Strip a trailing newline from version. */
360 	len = strlen(version);
361 	while (len > 0 && version[len - 1] == '\n')
362 		len--;
363 	printf("%.*s %s\n", len, version, machine);
364 	printf("%s\n", compiler_version);
365 }
366 
367 SYSINIT(announce, SI_SUB_COPYRIGHT, SI_ORDER_FIRST, print_caddr_t,
368     copyright);
369 SYSINIT(trademark, SI_SUB_COPYRIGHT, SI_ORDER_SECOND, print_caddr_t,
370     trademark);
371 SYSINIT(version, SI_SUB_COPYRIGHT, SI_ORDER_THIRD, print_version, NULL);
372 
373 #ifdef WITNESS
374 static char wit_warn[] =
375      "WARNING: WITNESS option enabled, expect reduced performance.\n";
376 SYSINIT(witwarn, SI_SUB_COPYRIGHT, SI_ORDER_FOURTH,
377    print_caddr_t, wit_warn);
378 SYSINIT(witwarn2, SI_SUB_LAST, SI_ORDER_FOURTH,
379    print_caddr_t, wit_warn);
380 #endif
381 
382 #ifdef DIAGNOSTIC
383 static char diag_warn[] =
384      "WARNING: DIAGNOSTIC option enabled, expect reduced performance.\n";
385 SYSINIT(diagwarn, SI_SUB_COPYRIGHT, SI_ORDER_FIFTH,
386     print_caddr_t, diag_warn);
387 SYSINIT(diagwarn2, SI_SUB_LAST, SI_ORDER_FIFTH,
388     print_caddr_t, diag_warn);
389 #endif
390 
391 #if __SIZEOF_LONG__ == 4
392 static char ilp32_warn[] =
393     "WARNING: 32-bit kernels are deprecated and may be removed in FreeBSD 15.0.\n";
394 SYSINIT(ilp32warn, SI_SUB_COPYRIGHT, SI_ORDER_FIFTH,
395     print_caddr_t, ilp32_warn);
396 SYSINIT(ilp32warn2, SI_SUB_LAST, SI_ORDER_FIFTH,
397     print_caddr_t, ilp32_warn);
398 #endif
399 
400 static int
401 null_fetch_syscall_args(struct thread *td __unused)
402 {
403 
404 	panic("null_fetch_syscall_args");
405 }
406 
407 static void
408 null_set_syscall_retval(struct thread *td __unused, int error __unused)
409 {
410 
411 	panic("null_set_syscall_retval");
412 }
413 
414 static void
415 null_set_fork_retval(struct thread *td __unused)
416 {
417 
418 }
419 
420 struct sysentvec null_sysvec = {
421 	.sv_size	= 0,
422 	.sv_table	= NULL,
423 	.sv_fixup	= NULL,
424 	.sv_sendsig	= NULL,
425 	.sv_sigcode	= NULL,
426 	.sv_szsigcode	= NULL,
427 	.sv_name	= "null",
428 	.sv_coredump	= NULL,
429 	.sv_minsigstksz	= 0,
430 	.sv_minuser	= VM_MIN_ADDRESS,
431 	.sv_maxuser	= VM_MAXUSER_ADDRESS,
432 	.sv_usrstack	= USRSTACK,
433 	.sv_psstrings	= PS_STRINGS,
434 	.sv_psstringssz	= sizeof(struct ps_strings),
435 	.sv_stackprot	= VM_PROT_ALL,
436 	.sv_copyout_strings	= NULL,
437 	.sv_setregs	= NULL,
438 	.sv_fixlimit	= NULL,
439 	.sv_maxssiz	= NULL,
440 	.sv_flags	= 0,
441 	.sv_set_syscall_retval = null_set_syscall_retval,
442 	.sv_fetch_syscall_args = null_fetch_syscall_args,
443 	.sv_syscallnames = NULL,
444 	.sv_schedtail	= NULL,
445 	.sv_thread_detach = NULL,
446 	.sv_trap	= NULL,
447 	.sv_set_fork_retval = null_set_fork_retval,
448 	.sv_regset_begin = NULL,
449 	.sv_regset_end  = NULL,
450 };
451 
452 /*
453  * The two following SYSINIT's are proc0 specific glue code.  I am not
454  * convinced that they can not be safely combined, but their order of
455  * operation has been maintained as the same as the original init_main.c
456  * for right now.
457  */
458 /* ARGSUSED*/
459 static void
460 proc0_init(void *dummy __unused)
461 {
462 	struct proc *p;
463 	struct thread *td;
464 	struct ucred *newcred;
465 	struct uidinfo tmpuinfo;
466 	struct loginclass tmplc = {
467 		.lc_name = "",
468 	};
469 	vm_paddr_t pageablemem;
470 	int i;
471 
472 	GIANT_REQUIRED;
473 	p = &proc0;
474 	td = &thread0;
475 
476 	/*
477 	 * Initialize magic number and osrel.
478 	 */
479 	p->p_magic = P_MAGIC;
480 	p->p_osrel = osreldate;
481 
482 	/*
483 	 * Initialize thread and process structures.
484 	 */
485 	procinit();	/* set up proc zone */
486 	threadinit();	/* set up UMA zones */
487 
488 	/*
489 	 * Initialise scheduler resources.
490 	 * Add scheduler specific parts to proc, thread as needed.
491 	 */
492 	schedinit();	/* scheduler gets its house in order */
493 
494 	/*
495 	 * Create process 0 (the swapper).
496 	 */
497 	LIST_INSERT_HEAD(&allproc, p, p_list);
498 	LIST_INSERT_HEAD(PIDHASH(0), p, p_hash);
499 	mtx_init(&pgrp0.pg_mtx, "process group", NULL, MTX_DEF | MTX_DUPOK);
500 	sx_init(&pgrp0.pg_killsx, "killpg racer");
501 	p->p_pgrp = &pgrp0;
502 	LIST_INSERT_HEAD(PGRPHASH(0), &pgrp0, pg_hash);
503 	LIST_INIT(&pgrp0.pg_members);
504 	LIST_INSERT_HEAD(&pgrp0.pg_members, p, p_pglist);
505 
506 	pgrp0.pg_session = &session0;
507 	mtx_init(&session0.s_mtx, "session", NULL, MTX_DEF);
508 	refcount_init(&session0.s_count, 1);
509 	session0.s_leader = p;
510 
511 	p->p_sysent = &null_sysvec;
512 	p->p_flag = P_SYSTEM | P_INMEM | P_KPROC;
513 	p->p_flag2 = 0;
514 	p->p_state = PRS_NORMAL;
515 	p->p_klist = knlist_alloc(&p->p_mtx);
516 	STAILQ_INIT(&p->p_ktr);
517 	p->p_nice = NZERO;
518 	td->td_tid = THREAD0_TID;
519 	tidhash_add(td);
520 	TD_SET_STATE(td, TDS_RUNNING);
521 	td->td_pri_class = PRI_TIMESHARE;
522 	td->td_user_pri = PUSER;
523 	td->td_base_user_pri = PUSER;
524 	td->td_lend_user_pri = PRI_MAX;
525 	td->td_priority = PVM;
526 	td->td_base_pri = PVM;
527 	td->td_oncpu = curcpu;
528 	td->td_flags = TDF_INMEM;
529 	td->td_pflags = TDP_KTHREAD;
530 	td->td_cpuset = cpuset_thread0();
531 	td->td_domain.dr_policy = td->td_cpuset->cs_domain;
532 	prison0_init();
533 	p->p_peers = 0;
534 	p->p_leader = p;
535 	p->p_reaper = p;
536 	p->p_treeflag |= P_TREE_REAPER;
537 	LIST_INIT(&p->p_reaplist);
538 
539 	strncpy(p->p_comm, "kernel", sizeof (p->p_comm));
540 	strncpy(td->td_name, "swapper", sizeof (td->td_name));
541 
542 	callout_init_mtx(&p->p_itcallout, &p->p_mtx, 0);
543 	callout_init_mtx(&p->p_limco, &p->p_mtx, 0);
544 	callout_init(&td->td_slpcallout, 1);
545 	TAILQ_INIT(&p->p_kqtim_stop);
546 
547 	/* Create credentials. */
548 	newcred = crget();
549 	newcred->cr_ngroups = 1;	/* group 0 */
550 	/* A hack to prevent uifind from tripping over NULL pointers. */
551 	curthread->td_ucred = newcred;
552 	tmpuinfo.ui_uid = 1;
553 	newcred->cr_uidinfo = newcred->cr_ruidinfo = &tmpuinfo;
554 	newcred->cr_uidinfo = uifind(0);
555 	newcred->cr_ruidinfo = uifind(0);
556 	newcred->cr_loginclass = &tmplc;
557 	newcred->cr_loginclass = loginclass_find("default");
558 	/* End hack. creds get properly set later with thread_cow_get_proc */
559 	curthread->td_ucred = NULL;
560 	newcred->cr_prison = &prison0;
561 	newcred->cr_users++; /* avoid assertion failure */
562 	proc_set_cred_init(p, newcred);
563 	newcred->cr_users--;
564 	crfree(newcred);
565 #ifdef AUDIT
566 	audit_cred_kproc0(newcred);
567 #endif
568 #ifdef MAC
569 	mac_cred_create_swapper(newcred);
570 #endif
571 	/* Create sigacts. */
572 	p->p_sigacts = sigacts_alloc();
573 
574 	/* Initialize signal state for process 0. */
575 	siginit(&proc0);
576 
577 	/* Create the file descriptor table. */
578 	p->p_pd = pdinit(NULL, false);
579 	p->p_fd = fdinit();
580 	p->p_fdtol = NULL;
581 
582 	/* Create the limits structures. */
583 	p->p_limit = lim_alloc();
584 	for (i = 0; i < RLIM_NLIMITS; i++)
585 		p->p_limit->pl_rlimit[i].rlim_cur =
586 		    p->p_limit->pl_rlimit[i].rlim_max = RLIM_INFINITY;
587 	p->p_limit->pl_rlimit[RLIMIT_NOFILE].rlim_cur =
588 	    p->p_limit->pl_rlimit[RLIMIT_NOFILE].rlim_max = maxfiles;
589 	p->p_limit->pl_rlimit[RLIMIT_NPROC].rlim_cur =
590 	    p->p_limit->pl_rlimit[RLIMIT_NPROC].rlim_max = maxproc;
591 	p->p_limit->pl_rlimit[RLIMIT_DATA].rlim_cur = dfldsiz;
592 	p->p_limit->pl_rlimit[RLIMIT_DATA].rlim_max = maxdsiz;
593 	p->p_limit->pl_rlimit[RLIMIT_STACK].rlim_cur = dflssiz;
594 	p->p_limit->pl_rlimit[RLIMIT_STACK].rlim_max = maxssiz;
595 	/* Cast to avoid overflow on i386/PAE. */
596 	pageablemem = ptoa((vm_paddr_t)vm_free_count());
597 	p->p_limit->pl_rlimit[RLIMIT_RSS].rlim_cur =
598 	    p->p_limit->pl_rlimit[RLIMIT_RSS].rlim_max = pageablemem;
599 	p->p_limit->pl_rlimit[RLIMIT_MEMLOCK].rlim_cur = pageablemem / 3;
600 	p->p_limit->pl_rlimit[RLIMIT_MEMLOCK].rlim_max = pageablemem;
601 	p->p_cpulimit = RLIM_INFINITY;
602 
603 	PROC_LOCK(p);
604 	thread_cow_get_proc(td, p);
605 	PROC_UNLOCK(p);
606 
607 	/* Initialize resource accounting structures. */
608 	racct_create(&p->p_racct);
609 
610 	p->p_stats = pstats_alloc();
611 
612 	/* Allocate a prototype map so we have something to fork. */
613 	p->p_vmspace = &vmspace0;
614 	refcount_init(&vmspace0.vm_refcnt, 1);
615 	pmap_pinit0(vmspace_pmap(&vmspace0));
616 
617 	/*
618 	 * proc0 is not expected to enter usermode, so there is no special
619 	 * handling for sv_minuser here, like is done for exec_new_vmspace().
620 	 */
621 	vm_map_init(&vmspace0.vm_map, vmspace_pmap(&vmspace0),
622 	    p->p_sysent->sv_minuser, p->p_sysent->sv_maxuser);
623 
624 	/*
625 	 * Call the init and ctor for the new thread and proc.  We wait
626 	 * to do this until all other structures are fairly sane.
627 	 */
628 	EVENTHANDLER_DIRECT_INVOKE(process_init, p);
629 	EVENTHANDLER_DIRECT_INVOKE(thread_init, td);
630 #ifdef KDTRACE_HOOKS
631 	kdtrace_proc_ctor(p);
632 	kdtrace_thread_ctor(td);
633 #endif
634 	EVENTHANDLER_DIRECT_INVOKE(process_ctor, p);
635 	EVENTHANDLER_DIRECT_INVOKE(thread_ctor, td);
636 
637 	/*
638 	 * Charge root for one process.
639 	 */
640 	(void)chgproccnt(p->p_ucred->cr_ruidinfo, 1, 0);
641 	PROC_LOCK(p);
642 	racct_add_force(p, RACCT_NPROC, 1);
643 	PROC_UNLOCK(p);
644 }
645 SYSINIT(p0init, SI_SUB_INTRINSIC, SI_ORDER_FIRST, proc0_init, NULL);
646 
647 /* ARGSUSED*/
648 static void
649 proc0_post(void *dummy __unused)
650 {
651 	struct proc *p;
652 	struct rusage ru;
653 	struct thread *td;
654 
655 	/*
656 	 * Now we can look at the time, having had a chance to verify the
657 	 * time from the filesystem.  Pretend that proc0 started now.
658 	 */
659 	sx_slock(&allproc_lock);
660 	FOREACH_PROC_IN_SYSTEM(p) {
661 		PROC_LOCK(p);
662 		if (p->p_state == PRS_NEW) {
663 			PROC_UNLOCK(p);
664 			continue;
665 		}
666 		microuptime(&p->p_stats->p_start);
667 		PROC_STATLOCK(p);
668 		rufetch(p, &ru);	/* Clears thread stats */
669 		p->p_rux.rux_runtime = 0;
670 		p->p_rux.rux_uticks = 0;
671 		p->p_rux.rux_sticks = 0;
672 		p->p_rux.rux_iticks = 0;
673 		PROC_STATUNLOCK(p);
674 		FOREACH_THREAD_IN_PROC(p, td) {
675 			td->td_runtime = 0;
676 		}
677 		PROC_UNLOCK(p);
678 	}
679 	sx_sunlock(&allproc_lock);
680 	PCPU_SET(switchtime, cpu_ticks());
681 	PCPU_SET(switchticks, ticks);
682 }
683 SYSINIT(p0post, SI_SUB_INTRINSIC_POST, SI_ORDER_FIRST, proc0_post, NULL);
684 
685 /*
686  ***************************************************************************
687  ****
688  **** The following SYSINIT's and glue code should be moved to the
689  **** respective files on a per subsystem basis.
690  ****
691  ***************************************************************************
692  */
693 
694 /*
695  * List of paths to try when searching for "init".
696  */
697 static char init_path[MAXPATHLEN] =
698 #ifdef	INIT_PATH
699     __XSTRING(INIT_PATH);
700 #else
701     "/sbin/init:/sbin/oinit:/sbin/init.bak:/rescue/init";
702 #endif
703 SYSCTL_STRING(_kern, OID_AUTO, init_path, CTLFLAG_RD, init_path, 0,
704 	"Path used to search the init process");
705 
706 /*
707  * Shutdown timeout of init(8).
708  * Unused within kernel, but used to control init(8), hence do not remove.
709  */
710 #ifndef INIT_SHUTDOWN_TIMEOUT
711 #define INIT_SHUTDOWN_TIMEOUT 120
712 #endif
713 static int init_shutdown_timeout = INIT_SHUTDOWN_TIMEOUT;
714 SYSCTL_INT(_kern, OID_AUTO, init_shutdown_timeout,
715 	CTLFLAG_RW, &init_shutdown_timeout, 0, "Shutdown timeout of init(8). "
716 	"Unused within kernel, but used to control init(8)");
717 
718 /*
719  * Start the initial user process; try exec'ing each pathname in init_path.
720  * The program is invoked with one argument containing the boot flags.
721  */
722 static void
723 start_init(void *dummy)
724 {
725 	struct image_args args;
726 	int error;
727 	char *var, *path;
728 	char *free_init_path, *tmp_init_path;
729 	struct thread *td;
730 	struct proc *p;
731 	struct vmspace *oldvmspace;
732 
733 	TSENTER();	/* Here so we don't overlap with mi_startup. */
734 
735 	td = curthread;
736 	p = td->td_proc;
737 
738 	vfs_mountroot();
739 
740 	/* Wipe GELI passphrase from the environment. */
741 	kern_unsetenv("kern.geom.eli.passphrase");
742 
743 	/* For Multicons, report which console is primary to both */
744 	if (boothowto & RB_MULTIPLE) {
745 		if (boothowto & RB_SERIAL)
746 			printf("Dual Console: Serial Primary, Video Secondary\n");
747 		else
748 			printf("Dual Console: Video Primary, Serial Secondary\n");
749 	}
750 
751 	if ((var = kern_getenv("init_path")) != NULL) {
752 		strlcpy(init_path, var, sizeof(init_path));
753 		freeenv(var);
754 	}
755 	free_init_path = tmp_init_path = strdup(init_path, M_TEMP);
756 
757 	while ((path = strsep(&tmp_init_path, ":")) != NULL) {
758 		if (bootverbose)
759 			printf("start_init: trying %s\n", path);
760 
761 		memset(&args, 0, sizeof(args));
762 		error = exec_alloc_args(&args);
763 		if (error != 0)
764 			panic("%s: Can't allocate space for init arguments %d",
765 			    __func__, error);
766 
767 		error = exec_args_add_fname(&args, path, UIO_SYSSPACE);
768 		if (error != 0)
769 			panic("%s: Can't add fname %d", __func__, error);
770 		error = exec_args_add_arg(&args, path, UIO_SYSSPACE);
771 		if (error != 0)
772 			panic("%s: Can't add argv[0] %d", __func__, error);
773 		if (boothowto & RB_SINGLE)
774 			error = exec_args_add_arg(&args, "-s", UIO_SYSSPACE);
775 		if (error != 0)
776 			panic("%s: Can't add argv[0] %d", __func__, error);
777 
778 		/*
779 		 * Now try to exec the program.  If can't for any reason
780 		 * other than it doesn't exist, complain.
781 		 *
782 		 * Otherwise, return via fork_trampoline() all the way
783 		 * to user mode as init!
784 		 */
785 		KASSERT((td->td_pflags & TDP_EXECVMSPC) == 0,
786 		    ("nested execve"));
787 		oldvmspace = p->p_vmspace;
788 		error = kern_execve(td, &args, NULL, oldvmspace);
789 		KASSERT(error != 0,
790 		    ("kern_execve returned success, not EJUSTRETURN"));
791 		if (error == EJUSTRETURN) {
792 			exec_cleanup(td, oldvmspace);
793 			free(free_init_path, M_TEMP);
794 			TSEXIT();
795 			return;
796 		}
797 		if (error != ENOENT)
798 			printf("exec %s: error %d\n", path, error);
799 	}
800 	free(free_init_path, M_TEMP);
801 	printf("init: not found in path %s\n", init_path);
802 	panic("no init");
803 }
804 
805 /*
806  * Like kproc_create(), but runs in its own address space.  We do this
807  * early to reserve pid 1.  Note special case - do not make it
808  * runnable yet, init execution is started when userspace can be served.
809  */
810 static void
811 create_init(const void *udata __unused)
812 {
813 	struct fork_req fr;
814 	struct ucred *newcred, *oldcred;
815 	struct thread *td;
816 	int error;
817 
818 	bzero(&fr, sizeof(fr));
819 	fr.fr_flags = RFFDG | RFPROC | RFSTOPPED;
820 	fr.fr_procp = &initproc;
821 	error = fork1(&thread0, &fr);
822 	if (error)
823 		panic("cannot fork init: %d\n", error);
824 	KASSERT(initproc->p_pid == 1, ("create_init: initproc->p_pid != 1"));
825 	/* divorce init's credentials from the kernel's */
826 	newcred = crget();
827 	sx_xlock(&proctree_lock);
828 	PROC_LOCK(initproc);
829 	initproc->p_flag |= P_SYSTEM | P_INMEM;
830 	initproc->p_treeflag |= P_TREE_REAPER;
831 	oldcred = initproc->p_ucred;
832 	crcopy(newcred, oldcred);
833 #ifdef MAC
834 	mac_cred_create_init(newcred);
835 #endif
836 #ifdef AUDIT
837 	audit_cred_proc1(newcred);
838 #endif
839 	proc_set_cred(initproc, newcred);
840 	td = FIRST_THREAD_IN_PROC(initproc);
841 	crcowfree(td);
842 	td->td_realucred = crcowget(initproc->p_ucred);
843 	td->td_ucred = td->td_realucred;
844 	PROC_UNLOCK(initproc);
845 	sx_xunlock(&proctree_lock);
846 	crfree(oldcred);
847 	cpu_fork_kthread_handler(FIRST_THREAD_IN_PROC(initproc),
848 	    start_init, NULL);
849 }
850 SYSINIT(init, SI_SUB_CREATE_INIT, SI_ORDER_FIRST, create_init, NULL);
851 
852 /*
853  * Make it runnable now.
854  */
855 static void
856 kick_init(const void *udata __unused)
857 {
858 	struct thread *td;
859 
860 	td = FIRST_THREAD_IN_PROC(initproc);
861 	thread_lock(td);
862 	TD_SET_CAN_RUN(td);
863 	sched_add(td, SRQ_BORING);
864 }
865 SYSINIT(kickinit, SI_SUB_KTHREAD_INIT, SI_ORDER_MIDDLE, kick_init, NULL);
866 
867 /*
868  * DDB(4).
869  */
870 #ifdef DDB
871 static void
872 db_show_print_syinit(struct sysinit *sip, bool ddb)
873 {
874 	const char *sname, *funcname;
875 	c_db_sym_t sym;
876 	db_expr_t  offset;
877 
878 #define xprint(...)							\
879 	if (ddb)							\
880 		db_printf(__VA_ARGS__);					\
881 	else								\
882 		printf(__VA_ARGS__)
883 
884 	if (sip == NULL) {
885 		xprint("%s: no sysinit * given\n", __func__);
886 		return;
887 	}
888 
889 	sym = db_search_symbol((vm_offset_t)sip, DB_STGY_ANY, &offset);
890 	db_symbol_values(sym, &sname, NULL);
891 	sym = db_search_symbol((vm_offset_t)sip->func, DB_STGY_PROC, &offset);
892 	db_symbol_values(sym, &funcname, NULL);
893 	xprint("%s(%p)\n", (sname != NULL) ? sname : "", sip);
894 	xprint("  %#08x %#08x\n", sip->subsystem, sip->order);
895 	xprint("  %p(%s)(%p)\n",
896 	    sip->func, (funcname != NULL) ? funcname : "", sip->udata);
897 #undef xprint
898 }
899 
900 DB_SHOW_COMMAND_FLAGS(sysinit, db_show_sysinit, DB_CMD_MEMSAFE)
901 {
902 	struct sysinit *sip;
903 
904 	db_printf("SYSINIT vs Name(Ptr)\n");
905 	db_printf("  Subsystem  Order\n");
906 	db_printf("  Function(Name)(Arg)\n");
907 	SLIST_FOREACH(sip, &sysinit_list, next) {
908 		db_show_print_syinit(sip, true);
909 		if (db_pager_quit)
910 			break;
911 	}
912 }
913 #endif /* DDB */
914