xref: /freebsd/sys/kern/init_main.c (revision e6bfd18d21b225af6a0ed67ceeaf1293b7b9eba5)
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 	TSENTER2("bubblesort");
259 	for (sipp = sysinit; sipp < sysinit_end; sipp++) {
260 		for (xipp = sipp + 1; xipp < sysinit_end; xipp++) {
261 			if ((*sipp)->subsystem < (*xipp)->subsystem ||
262 			     ((*sipp)->subsystem == (*xipp)->subsystem &&
263 			      (*sipp)->order <= (*xipp)->order))
264 				continue;	/* skip*/
265 			save = *sipp;
266 			*sipp = *xipp;
267 			*xipp = save;
268 		}
269 	}
270 	TSEXIT2("bubblesort");
271 
272 	last = SI_SUB_COPYRIGHT;
273 #if defined(VERBOSE_SYSINIT)
274 	verbose = 0;
275 #if !defined(DDB)
276 	printf("VERBOSE_SYSINIT: DDB not enabled, symbol lookups disabled.\n");
277 #endif
278 #endif
279 
280 	/*
281 	 * Traverse the (now) ordered list of system initialization tasks.
282 	 * Perform each task, and continue on to the next task.
283 	 */
284 	for (sipp = sysinit; sipp < sysinit_end; sipp++) {
285 		if ((*sipp)->subsystem == SI_SUB_DUMMY)
286 			continue;	/* skip dummy task(s)*/
287 
288 		if ((*sipp)->subsystem == SI_SUB_DONE)
289 			continue;
290 
291 		if ((*sipp)->subsystem > last)
292 			BOOTTRACE_INIT("sysinit 0x%7x", (*sipp)->subsystem);
293 
294 #if defined(VERBOSE_SYSINIT)
295 		if ((*sipp)->subsystem > last && verbose_sysinit != 0) {
296 			verbose = 1;
297 			printf("subsystem %x\n", last);
298 		}
299 		if (verbose) {
300 #if defined(DDB)
301 			const char *func, *data;
302 
303 			func = symbol_name((vm_offset_t)(*sipp)->func,
304 			    DB_STGY_PROC);
305 			data = symbol_name((vm_offset_t)(*sipp)->udata,
306 			    DB_STGY_ANY);
307 			if (func != NULL && data != NULL)
308 				printf("   %s(&%s)... ", func, data);
309 			else if (func != NULL)
310 				printf("   %s(%p)... ", func, (*sipp)->udata);
311 			else
312 #endif
313 				printf("   %p(%p)... ", (*sipp)->func,
314 				    (*sipp)->udata);
315 		}
316 #endif
317 
318 		/* Call function */
319 		(*((*sipp)->func))((*sipp)->udata);
320 
321 #if defined(VERBOSE_SYSINIT)
322 		if (verbose)
323 			printf("done.\n");
324 #endif
325 
326 		/* Check off the one we're just done */
327 		last = (*sipp)->subsystem;
328 		(*sipp)->subsystem = SI_SUB_DONE;
329 
330 		/* Check if we've installed more sysinit items via KLD */
331 		if (newsysinit != NULL) {
332 			if (sysinit != SET_BEGIN(sysinit_set))
333 				free(sysinit, M_TEMP);
334 			sysinit = newsysinit;
335 			sysinit_end = newsysinit_end;
336 			newsysinit = NULL;
337 			newsysinit_end = NULL;
338 			goto restart;
339 		}
340 	}
341 
342 	TSEXIT();	/* Here so we don't overlap with start_init. */
343 	BOOTTRACE("mi_startup done");
344 
345 	mtx_assert(&Giant, MA_OWNED | MA_NOTRECURSED);
346 	mtx_unlock(&Giant);
347 
348 	/*
349 	 * Now hand over this thread to swapper.
350 	 */
351 	swapper();
352 	/* NOTREACHED*/
353 }
354 
355 static void
356 print_caddr_t(void *data)
357 {
358 	printf("%s", (char *)data);
359 }
360 
361 static void
362 print_version(void *data __unused)
363 {
364 	int len;
365 
366 	/* Strip a trailing newline from version. */
367 	len = strlen(version);
368 	while (len > 0 && version[len - 1] == '\n')
369 		len--;
370 	printf("%.*s %s\n", len, version, machine);
371 	printf("%s\n", compiler_version);
372 }
373 
374 SYSINIT(announce, SI_SUB_COPYRIGHT, SI_ORDER_FIRST, print_caddr_t,
375     copyright);
376 SYSINIT(trademark, SI_SUB_COPYRIGHT, SI_ORDER_SECOND, print_caddr_t,
377     trademark);
378 SYSINIT(version, SI_SUB_COPYRIGHT, SI_ORDER_THIRD, print_version, NULL);
379 
380 #ifdef WITNESS
381 static char wit_warn[] =
382      "WARNING: WITNESS option enabled, expect reduced performance.\n";
383 SYSINIT(witwarn, SI_SUB_COPYRIGHT, SI_ORDER_FOURTH,
384    print_caddr_t, wit_warn);
385 SYSINIT(witwarn2, SI_SUB_LAST, SI_ORDER_FOURTH,
386    print_caddr_t, wit_warn);
387 #endif
388 
389 #ifdef DIAGNOSTIC
390 static char diag_warn[] =
391      "WARNING: DIAGNOSTIC option enabled, expect reduced performance.\n";
392 SYSINIT(diagwarn, SI_SUB_COPYRIGHT, SI_ORDER_FIFTH,
393     print_caddr_t, diag_warn);
394 SYSINIT(diagwarn2, SI_SUB_LAST, SI_ORDER_FIFTH,
395     print_caddr_t, diag_warn);
396 #endif
397 
398 static int
399 null_fetch_syscall_args(struct thread *td __unused)
400 {
401 
402 	panic("null_fetch_syscall_args");
403 }
404 
405 static void
406 null_set_syscall_retval(struct thread *td __unused, int error __unused)
407 {
408 
409 	panic("null_set_syscall_retval");
410 }
411 
412 static void
413 null_set_fork_retval(struct thread *td __unused)
414 {
415 
416 }
417 
418 struct sysentvec null_sysvec = {
419 	.sv_size	= 0,
420 	.sv_table	= NULL,
421 	.sv_fixup	= NULL,
422 	.sv_sendsig	= NULL,
423 	.sv_sigcode	= NULL,
424 	.sv_szsigcode	= NULL,
425 	.sv_name	= "null",
426 	.sv_coredump	= NULL,
427 	.sv_minsigstksz	= 0,
428 	.sv_minuser	= VM_MIN_ADDRESS,
429 	.sv_maxuser	= VM_MAXUSER_ADDRESS,
430 	.sv_usrstack	= USRSTACK,
431 	.sv_psstrings	= PS_STRINGS,
432 	.sv_psstringssz	= sizeof(struct ps_strings),
433 	.sv_stackprot	= VM_PROT_ALL,
434 	.sv_copyout_strings	= NULL,
435 	.sv_setregs	= NULL,
436 	.sv_fixlimit	= NULL,
437 	.sv_maxssiz	= NULL,
438 	.sv_flags	= 0,
439 	.sv_set_syscall_retval = null_set_syscall_retval,
440 	.sv_fetch_syscall_args = null_fetch_syscall_args,
441 	.sv_syscallnames = NULL,
442 	.sv_schedtail	= NULL,
443 	.sv_thread_detach = NULL,
444 	.sv_trap	= NULL,
445 	.sv_set_fork_retval = null_set_fork_retval,
446 	.sv_regset_begin = NULL,
447 	.sv_regset_end  = NULL,
448 };
449 
450 /*
451  * The two following SYSINIT's are proc0 specific glue code.  I am not
452  * convinced that they can not be safely combined, but their order of
453  * operation has been maintained as the same as the original init_main.c
454  * for right now.
455  */
456 /* ARGSUSED*/
457 static void
458 proc0_init(void *dummy __unused)
459 {
460 	struct proc *p;
461 	struct thread *td;
462 	struct ucred *newcred;
463 	struct uidinfo tmpuinfo;
464 	struct loginclass tmplc = {
465 		.lc_name = "",
466 	};
467 	vm_paddr_t pageablemem;
468 	int i;
469 
470 	GIANT_REQUIRED;
471 	p = &proc0;
472 	td = &thread0;
473 
474 	/*
475 	 * Initialize magic number and osrel.
476 	 */
477 	p->p_magic = P_MAGIC;
478 	p->p_osrel = osreldate;
479 
480 	/*
481 	 * Initialize thread and process structures.
482 	 */
483 	procinit();	/* set up proc zone */
484 	threadinit();	/* set up UMA zones */
485 
486 	/*
487 	 * Initialise scheduler resources.
488 	 * Add scheduler specific parts to proc, thread as needed.
489 	 */
490 	schedinit();	/* scheduler gets its house in order */
491 
492 	/*
493 	 * Create process 0 (the swapper).
494 	 */
495 	LIST_INSERT_HEAD(&allproc, p, p_list);
496 	LIST_INSERT_HEAD(PIDHASH(0), p, p_hash);
497 	mtx_init(&pgrp0.pg_mtx, "process group", NULL, MTX_DEF | MTX_DUPOK);
498 	p->p_pgrp = &pgrp0;
499 	LIST_INSERT_HEAD(PGRPHASH(0), &pgrp0, pg_hash);
500 	LIST_INIT(&pgrp0.pg_members);
501 	LIST_INSERT_HEAD(&pgrp0.pg_members, p, p_pglist);
502 
503 	pgrp0.pg_session = &session0;
504 	mtx_init(&session0.s_mtx, "session", NULL, MTX_DEF);
505 	refcount_init(&session0.s_count, 1);
506 	session0.s_leader = p;
507 
508 	p->p_sysent = &null_sysvec;
509 	p->p_flag = P_SYSTEM | P_INMEM | P_KPROC;
510 	p->p_flag2 = 0;
511 	p->p_state = PRS_NORMAL;
512 	p->p_klist = knlist_alloc(&p->p_mtx);
513 	STAILQ_INIT(&p->p_ktr);
514 	p->p_nice = NZERO;
515 	td->td_tid = THREAD0_TID;
516 	tidhash_add(td);
517 	TD_SET_STATE(td, TDS_RUNNING);
518 	td->td_pri_class = PRI_TIMESHARE;
519 	td->td_user_pri = PUSER;
520 	td->td_base_user_pri = PUSER;
521 	td->td_lend_user_pri = PRI_MAX;
522 	td->td_priority = PVM;
523 	td->td_base_pri = PVM;
524 	td->td_oncpu = curcpu;
525 	td->td_flags = TDF_INMEM;
526 	td->td_pflags = TDP_KTHREAD;
527 	td->td_cpuset = cpuset_thread0();
528 	td->td_domain.dr_policy = td->td_cpuset->cs_domain;
529 	prison0_init();
530 	p->p_peers = 0;
531 	p->p_leader = p;
532 	p->p_reaper = p;
533 	p->p_treeflag |= P_TREE_REAPER;
534 	LIST_INIT(&p->p_reaplist);
535 
536 	strncpy(p->p_comm, "kernel", sizeof (p->p_comm));
537 	strncpy(td->td_name, "swapper", sizeof (td->td_name));
538 
539 	callout_init_mtx(&p->p_itcallout, &p->p_mtx, 0);
540 	callout_init_mtx(&p->p_limco, &p->p_mtx, 0);
541 	callout_init(&td->td_slpcallout, 1);
542 	TAILQ_INIT(&p->p_kqtim_stop);
543 
544 	/* Create credentials. */
545 	newcred = crget();
546 	newcred->cr_ngroups = 1;	/* group 0 */
547 	/* A hack to prevent uifind from tripping over NULL pointers. */
548 	curthread->td_ucred = newcred;
549 	tmpuinfo.ui_uid = 1;
550 	newcred->cr_uidinfo = newcred->cr_ruidinfo = &tmpuinfo;
551 	newcred->cr_uidinfo = uifind(0);
552 	newcred->cr_ruidinfo = uifind(0);
553 	newcred->cr_loginclass = &tmplc;
554 	newcred->cr_loginclass = loginclass_find("default");
555 	/* End hack. creds get properly set later with thread_cow_get_proc */
556 	curthread->td_ucred = NULL;
557 	newcred->cr_prison = &prison0;
558 	newcred->cr_users++; /* avoid assertion failure */
559 	proc_set_cred_init(p, newcred);
560 	newcred->cr_users--;
561 	crfree(newcred);
562 #ifdef AUDIT
563 	audit_cred_kproc0(newcred);
564 #endif
565 #ifdef MAC
566 	mac_cred_create_swapper(newcred);
567 #endif
568 	/* Create sigacts. */
569 	p->p_sigacts = sigacts_alloc();
570 
571 	/* Initialize signal state for process 0. */
572 	siginit(&proc0);
573 
574 	/* Create the file descriptor table. */
575 	p->p_pd = pdinit(NULL, false);
576 	p->p_fd = fdinit();
577 	p->p_fdtol = NULL;
578 
579 	/* Create the limits structures. */
580 	p->p_limit = lim_alloc();
581 	for (i = 0; i < RLIM_NLIMITS; i++)
582 		p->p_limit->pl_rlimit[i].rlim_cur =
583 		    p->p_limit->pl_rlimit[i].rlim_max = RLIM_INFINITY;
584 	p->p_limit->pl_rlimit[RLIMIT_NOFILE].rlim_cur =
585 	    p->p_limit->pl_rlimit[RLIMIT_NOFILE].rlim_max = maxfiles;
586 	p->p_limit->pl_rlimit[RLIMIT_NPROC].rlim_cur =
587 	    p->p_limit->pl_rlimit[RLIMIT_NPROC].rlim_max = maxproc;
588 	p->p_limit->pl_rlimit[RLIMIT_DATA].rlim_cur = dfldsiz;
589 	p->p_limit->pl_rlimit[RLIMIT_DATA].rlim_max = maxdsiz;
590 	p->p_limit->pl_rlimit[RLIMIT_STACK].rlim_cur = dflssiz;
591 	p->p_limit->pl_rlimit[RLIMIT_STACK].rlim_max = maxssiz;
592 	/* Cast to avoid overflow on i386/PAE. */
593 	pageablemem = ptoa((vm_paddr_t)vm_free_count());
594 	p->p_limit->pl_rlimit[RLIMIT_RSS].rlim_cur =
595 	    p->p_limit->pl_rlimit[RLIMIT_RSS].rlim_max = pageablemem;
596 	p->p_limit->pl_rlimit[RLIMIT_MEMLOCK].rlim_cur = pageablemem / 3;
597 	p->p_limit->pl_rlimit[RLIMIT_MEMLOCK].rlim_max = pageablemem;
598 	p->p_cpulimit = RLIM_INFINITY;
599 
600 	PROC_LOCK(p);
601 	thread_cow_get_proc(td, p);
602 	PROC_UNLOCK(p);
603 
604 	/* Initialize resource accounting structures. */
605 	racct_create(&p->p_racct);
606 
607 	p->p_stats = pstats_alloc();
608 
609 	/* Allocate a prototype map so we have something to fork. */
610 	p->p_vmspace = &vmspace0;
611 	refcount_init(&vmspace0.vm_refcnt, 1);
612 	pmap_pinit0(vmspace_pmap(&vmspace0));
613 
614 	/*
615 	 * proc0 is not expected to enter usermode, so there is no special
616 	 * handling for sv_minuser here, like is done for exec_new_vmspace().
617 	 */
618 	vm_map_init(&vmspace0.vm_map, vmspace_pmap(&vmspace0),
619 	    p->p_sysent->sv_minuser, p->p_sysent->sv_maxuser);
620 
621 	/*
622 	 * Call the init and ctor for the new thread and proc.  We wait
623 	 * to do this until all other structures are fairly sane.
624 	 */
625 	EVENTHANDLER_DIRECT_INVOKE(process_init, p);
626 	EVENTHANDLER_DIRECT_INVOKE(thread_init, td);
627 #ifdef KDTRACE_HOOKS
628 	kdtrace_proc_ctor(p);
629 	kdtrace_thread_ctor(td);
630 #endif
631 	EVENTHANDLER_DIRECT_INVOKE(process_ctor, p);
632 	EVENTHANDLER_DIRECT_INVOKE(thread_ctor, td);
633 
634 	/*
635 	 * Charge root for one process.
636 	 */
637 	(void)chgproccnt(p->p_ucred->cr_ruidinfo, 1, 0);
638 	PROC_LOCK(p);
639 	racct_add_force(p, RACCT_NPROC, 1);
640 	PROC_UNLOCK(p);
641 }
642 SYSINIT(p0init, SI_SUB_INTRINSIC, SI_ORDER_FIRST, proc0_init, NULL);
643 
644 /* ARGSUSED*/
645 static void
646 proc0_post(void *dummy __unused)
647 {
648 	struct proc *p;
649 	struct rusage ru;
650 	struct thread *td;
651 
652 	/*
653 	 * Now we can look at the time, having had a chance to verify the
654 	 * time from the filesystem.  Pretend that proc0 started now.
655 	 */
656 	sx_slock(&allproc_lock);
657 	FOREACH_PROC_IN_SYSTEM(p) {
658 		PROC_LOCK(p);
659 		if (p->p_state == PRS_NEW) {
660 			PROC_UNLOCK(p);
661 			continue;
662 		}
663 		microuptime(&p->p_stats->p_start);
664 		PROC_STATLOCK(p);
665 		rufetch(p, &ru);	/* Clears thread stats */
666 		p->p_rux.rux_runtime = 0;
667 		p->p_rux.rux_uticks = 0;
668 		p->p_rux.rux_sticks = 0;
669 		p->p_rux.rux_iticks = 0;
670 		PROC_STATUNLOCK(p);
671 		FOREACH_THREAD_IN_PROC(p, td) {
672 			td->td_runtime = 0;
673 		}
674 		PROC_UNLOCK(p);
675 	}
676 	sx_sunlock(&allproc_lock);
677 	PCPU_SET(switchtime, cpu_ticks());
678 	PCPU_SET(switchticks, ticks);
679 }
680 SYSINIT(p0post, SI_SUB_INTRINSIC_POST, SI_ORDER_FIRST, proc0_post, NULL);
681 
682 /*
683  ***************************************************************************
684  ****
685  **** The following SYSINIT's and glue code should be moved to the
686  **** respective files on a per subsystem basis.
687  ****
688  ***************************************************************************
689  */
690 
691 /*
692  * List of paths to try when searching for "init".
693  */
694 static char init_path[MAXPATHLEN] =
695 #ifdef	INIT_PATH
696     __XSTRING(INIT_PATH);
697 #else
698     "/sbin/init:/sbin/oinit:/sbin/init.bak:/rescue/init";
699 #endif
700 SYSCTL_STRING(_kern, OID_AUTO, init_path, CTLFLAG_RD, init_path, 0,
701 	"Path used to search the init process");
702 
703 /*
704  * Shutdown timeout of init(8).
705  * Unused within kernel, but used to control init(8), hence do not remove.
706  */
707 #ifndef INIT_SHUTDOWN_TIMEOUT
708 #define INIT_SHUTDOWN_TIMEOUT 120
709 #endif
710 static int init_shutdown_timeout = INIT_SHUTDOWN_TIMEOUT;
711 SYSCTL_INT(_kern, OID_AUTO, init_shutdown_timeout,
712 	CTLFLAG_RW, &init_shutdown_timeout, 0, "Shutdown timeout of init(8). "
713 	"Unused within kernel, but used to control init(8)");
714 
715 /*
716  * Start the initial user process; try exec'ing each pathname in init_path.
717  * The program is invoked with one argument containing the boot flags.
718  */
719 static void
720 start_init(void *dummy)
721 {
722 	struct image_args args;
723 	int error;
724 	char *var, *path;
725 	char *free_init_path, *tmp_init_path;
726 	struct thread *td;
727 	struct proc *p;
728 	struct vmspace *oldvmspace;
729 
730 	TSENTER();	/* Here so we don't overlap with mi_startup. */
731 
732 	td = curthread;
733 	p = td->td_proc;
734 
735 	vfs_mountroot();
736 
737 	/* Wipe GELI passphrase from the environment. */
738 	kern_unsetenv("kern.geom.eli.passphrase");
739 
740 	/* For Multicons, report which console is primary to both */
741 	if (boothowto & RB_MULTIPLE) {
742 		if (boothowto & RB_SERIAL)
743 			printf("Dual Console: Serial Primary, Video Secondary\n");
744 		else
745 			printf("Dual Console: Video Primary, Serial Secondary\n");
746 	}
747 
748 	if ((var = kern_getenv("init_path")) != NULL) {
749 		strlcpy(init_path, var, sizeof(init_path));
750 		freeenv(var);
751 	}
752 	free_init_path = tmp_init_path = strdup(init_path, M_TEMP);
753 
754 	while ((path = strsep(&tmp_init_path, ":")) != NULL) {
755 		if (bootverbose)
756 			printf("start_init: trying %s\n", path);
757 
758 		memset(&args, 0, sizeof(args));
759 		error = exec_alloc_args(&args);
760 		if (error != 0)
761 			panic("%s: Can't allocate space for init arguments %d",
762 			    __func__, error);
763 
764 		error = exec_args_add_fname(&args, path, UIO_SYSSPACE);
765 		if (error != 0)
766 			panic("%s: Can't add fname %d", __func__, error);
767 		error = exec_args_add_arg(&args, path, UIO_SYSSPACE);
768 		if (error != 0)
769 			panic("%s: Can't add argv[0] %d", __func__, error);
770 		if (boothowto & RB_SINGLE)
771 			error = exec_args_add_arg(&args, "-s", UIO_SYSSPACE);
772 		if (error != 0)
773 			panic("%s: Can't add argv[0] %d", __func__, error);
774 
775 		/*
776 		 * Now try to exec the program.  If can't for any reason
777 		 * other than it doesn't exist, complain.
778 		 *
779 		 * Otherwise, return via fork_trampoline() all the way
780 		 * to user mode as init!
781 		 */
782 		KASSERT((td->td_pflags & TDP_EXECVMSPC) == 0,
783 		    ("nested execve"));
784 		oldvmspace = p->p_vmspace;
785 		error = kern_execve(td, &args, NULL, oldvmspace);
786 		KASSERT(error != 0,
787 		    ("kern_execve returned success, not EJUSTRETURN"));
788 		if (error == EJUSTRETURN) {
789 			exec_cleanup(td, oldvmspace);
790 			free(free_init_path, M_TEMP);
791 			TSEXIT();
792 			return;
793 		}
794 		if (error != ENOENT)
795 			printf("exec %s: error %d\n", path, error);
796 	}
797 	free(free_init_path, M_TEMP);
798 	printf("init: not found in path %s\n", init_path);
799 	panic("no init");
800 }
801 
802 /*
803  * Like kproc_create(), but runs in its own address space.  We do this
804  * early to reserve pid 1.  Note special case - do not make it
805  * runnable yet, init execution is started when userspace can be served.
806  */
807 static void
808 create_init(const void *udata __unused)
809 {
810 	struct fork_req fr;
811 	struct ucred *newcred, *oldcred;
812 	struct thread *td;
813 	int error;
814 
815 	bzero(&fr, sizeof(fr));
816 	fr.fr_flags = RFFDG | RFPROC | RFSTOPPED;
817 	fr.fr_procp = &initproc;
818 	error = fork1(&thread0, &fr);
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 	sx_xlock(&proctree_lock);
825 	PROC_LOCK(initproc);
826 	initproc->p_flag |= P_SYSTEM | P_INMEM;
827 	initproc->p_treeflag |= P_TREE_REAPER;
828 	oldcred = initproc->p_ucred;
829 	crcopy(newcred, oldcred);
830 #ifdef MAC
831 	mac_cred_create_init(newcred);
832 #endif
833 #ifdef AUDIT
834 	audit_cred_proc1(newcred);
835 #endif
836 	proc_set_cred(initproc, newcred);
837 	td = FIRST_THREAD_IN_PROC(initproc);
838 	crcowfree(td);
839 	td->td_realucred = crcowget(initproc->p_ucred);
840 	td->td_ucred = td->td_realucred;
841 	PROC_UNLOCK(initproc);
842 	sx_xunlock(&proctree_lock);
843 	crfree(oldcred);
844 	cpu_fork_kthread_handler(FIRST_THREAD_IN_PROC(initproc),
845 	    start_init, NULL);
846 }
847 SYSINIT(init, SI_SUB_CREATE_INIT, SI_ORDER_FIRST, create_init, NULL);
848 
849 /*
850  * Make it runnable now.
851  */
852 static void
853 kick_init(const void *udata __unused)
854 {
855 	struct thread *td;
856 
857 	td = FIRST_THREAD_IN_PROC(initproc);
858 	thread_lock(td);
859 	TD_SET_CAN_RUN(td);
860 	sched_add(td, SRQ_BORING);
861 }
862 SYSINIT(kickinit, SI_SUB_KTHREAD_INIT, SI_ORDER_MIDDLE, kick_init, NULL);
863 
864 /*
865  * DDB(4).
866  */
867 #ifdef DDB
868 static void
869 db_show_print_syinit(struct sysinit *sip, bool ddb)
870 {
871 	const char *sname, *funcname;
872 	c_db_sym_t sym;
873 	db_expr_t  offset;
874 
875 #define xprint(...)							\
876 	if (ddb)							\
877 		db_printf(__VA_ARGS__);					\
878 	else								\
879 		printf(__VA_ARGS__)
880 
881 	if (sip == NULL) {
882 		xprint("%s: no sysinit * given\n", __func__);
883 		return;
884 	}
885 
886 	sym = db_search_symbol((vm_offset_t)sip, DB_STGY_ANY, &offset);
887 	db_symbol_values(sym, &sname, NULL);
888 	sym = db_search_symbol((vm_offset_t)sip->func, DB_STGY_PROC, &offset);
889 	db_symbol_values(sym, &funcname, NULL);
890 	xprint("%s(%p)\n", (sname != NULL) ? sname : "", sip);
891 	xprint("  %#08x %#08x\n", sip->subsystem, sip->order);
892 	xprint("  %p(%s)(%p)\n",
893 	    sip->func, (funcname != NULL) ? funcname : "", sip->udata);
894 #undef xprint
895 }
896 
897 DB_SHOW_COMMAND_FLAGS(sysinit, db_show_sysinit, DB_CMD_MEMSAFE)
898 {
899 	struct sysinit **sipp;
900 
901 	db_printf("SYSINIT vs Name(Ptr)\n");
902 	db_printf("  Subsystem  Order\n");
903 	db_printf("  Function(Name)(Arg)\n");
904 	for (sipp = sysinit; sipp < sysinit_end; sipp++) {
905 		db_show_print_syinit(*sipp, true);
906 		if (db_pager_quit)
907 			break;
908 	}
909 }
910 #endif /* DDB */
911