xref: /freebsd/sys/kern/init_main.c (revision 7be9a3b45356747f9fcb6d69a722c1c95f8060bf)
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/dtrace_bsd.h>
69 #include <sys/syscallsubr.h>
70 #include <sys/sysctl.h>
71 #include <sys/proc.h>
72 #include <sys/racct.h>
73 #include <sys/resourcevar.h>
74 #include <sys/systm.h>
75 #include <sys/signalvar.h>
76 #include <sys/vnode.h>
77 #include <sys/sysent.h>
78 #include <sys/reboot.h>
79 #include <sys/sched.h>
80 #include <sys/sx.h>
81 #include <sys/sysproto.h>
82 #include <sys/vmmeter.h>
83 #include <sys/unistd.h>
84 #include <sys/malloc.h>
85 #include <sys/conf.h>
86 #include <sys/cpuset.h>
87 
88 #include <machine/cpu.h>
89 
90 #include <security/audit/audit.h>
91 #include <security/mac/mac_framework.h>
92 
93 #include <vm/vm.h>
94 #include <vm/vm_param.h>
95 #include <vm/vm_extern.h>
96 #include <vm/pmap.h>
97 #include <vm/vm_map.h>
98 #include <sys/copyright.h>
99 
100 #include <ddb/ddb.h>
101 #include <ddb/db_sym.h>
102 
103 void mi_startup(void);				/* Should be elsewhere */
104 
105 /* Components of the first process -- never freed. */
106 static struct session session0;
107 static struct pgrp pgrp0;
108 struct	proc proc0;
109 struct thread0_storage thread0_st __aligned(32);
110 struct	vmspace vmspace0;
111 struct	proc *initproc;
112 
113 int
114 linux_alloc_current_noop(struct thread *td __unused, int flags __unused)
115 {
116 	return (0);
117 }
118 int (*lkpi_alloc_current)(struct thread *, int) = linux_alloc_current_noop;
119 
120 #ifndef BOOTHOWTO
121 #define	BOOTHOWTO	0
122 #endif
123 int	boothowto = BOOTHOWTO;	/* initialized so that it can be patched */
124 SYSCTL_INT(_debug, OID_AUTO, boothowto, CTLFLAG_RD, &boothowto, 0,
125 	"Boot control flags, passed from loader");
126 
127 #ifndef BOOTVERBOSE
128 #define	BOOTVERBOSE	0
129 #endif
130 int	bootverbose = BOOTVERBOSE;
131 SYSCTL_INT(_debug, OID_AUTO, bootverbose, CTLFLAG_RW, &bootverbose, 0,
132 	"Control the output of verbose kernel messages");
133 
134 #ifdef VERBOSE_SYSINIT
135 /*
136  * We'll use the defined value of VERBOSE_SYSINIT from the kernel config to
137  * dictate the default VERBOSE_SYSINIT behavior.  Significant values for this
138  * option and associated tunable are:
139  * - 0, 'compiled in but silent by default'
140  * - 1, 'compiled in but verbose by default' (default)
141  */
142 int	verbose_sysinit = VERBOSE_SYSINIT;
143 TUNABLE_INT("debug.verbose_sysinit", &verbose_sysinit);
144 #endif
145 
146 #ifdef INVARIANTS
147 FEATURE(invariants, "Kernel compiled with INVARIANTS, may affect performance");
148 #endif
149 
150 /*
151  * This ensures that there is at least one entry so that the sysinit_set
152  * symbol is not undefined.  A sybsystem ID of SI_SUB_DUMMY is never
153  * executed.
154  */
155 SYSINIT(placeholder, SI_SUB_DUMMY, SI_ORDER_ANY, NULL, NULL);
156 
157 /*
158  * The sysinit table itself.  Items are checked off as the are run.
159  * If we want to register new sysinit types, add them to newsysinit.
160  */
161 SET_DECLARE(sysinit_set, struct sysinit);
162 struct sysinit **sysinit, **sysinit_end;
163 struct sysinit **newsysinit, **newsysinit_end;
164 
165 /*
166  * Merge a new sysinit set into the current set, reallocating it if
167  * necessary.  This can only be called after malloc is running.
168  */
169 void
170 sysinit_add(struct sysinit **set, struct sysinit **set_end)
171 {
172 	struct sysinit **newset;
173 	struct sysinit **sipp;
174 	struct sysinit **xipp;
175 	int count;
176 
177 	count = set_end - set;
178 	if (newsysinit)
179 		count += newsysinit_end - newsysinit;
180 	else
181 		count += sysinit_end - sysinit;
182 	newset = malloc(count * sizeof(*sipp), M_TEMP, M_NOWAIT);
183 	if (newset == NULL)
184 		panic("cannot malloc for sysinit");
185 	xipp = newset;
186 	if (newsysinit)
187 		for (sipp = newsysinit; sipp < newsysinit_end; sipp++)
188 			*xipp++ = *sipp;
189 	else
190 		for (sipp = sysinit; sipp < sysinit_end; sipp++)
191 			*xipp++ = *sipp;
192 	for (sipp = set; sipp < set_end; sipp++)
193 		*xipp++ = *sipp;
194 	if (newsysinit)
195 		free(newsysinit, M_TEMP);
196 	newsysinit = newset;
197 	newsysinit_end = newset + count;
198 }
199 
200 #if defined (DDB) && defined(VERBOSE_SYSINIT)
201 static const char *
202 symbol_name(vm_offset_t va, db_strategy_t strategy)
203 {
204 	const char *name;
205 	c_db_sym_t sym;
206 	db_expr_t  offset;
207 
208 	if (va == 0)
209 		return (NULL);
210 	sym = db_search_symbol(va, strategy, &offset);
211 	if (offset != 0)
212 		return (NULL);
213 	db_symbol_values(sym, &name, NULL);
214 	return (name);
215 }
216 #endif
217 
218 /*
219  * System startup; initialize the world, create process 0, mount root
220  * filesystem, and fork to create init and pagedaemon.  Most of the
221  * hard work is done in the lower-level initialization routines including
222  * startup(), which does memory initialization and autoconfiguration.
223  *
224  * This allows simple addition of new kernel subsystems that require
225  * boot time initialization.  It also allows substitution of subsystem
226  * (for instance, a scheduler, kernel profiler, or VM system) by object
227  * module.  Finally, it allows for optional "kernel threads".
228  */
229 void
230 mi_startup(void)
231 {
232 
233 	struct sysinit **sipp;	/* system initialization*/
234 	struct sysinit **xipp;	/* interior loop of sort*/
235 	struct sysinit *save;	/* bubble*/
236 
237 #if defined(VERBOSE_SYSINIT)
238 	int last;
239 	int verbose;
240 #endif
241 
242 	TSENTER();
243 
244 	if (boothowto & RB_VERBOSE)
245 		bootverbose++;
246 
247 	if (sysinit == NULL) {
248 		sysinit = SET_BEGIN(sysinit_set);
249 		sysinit_end = SET_LIMIT(sysinit_set);
250 	}
251 
252 restart:
253 	/*
254 	 * Perform a bubble sort of the system initialization objects by
255 	 * their subsystem (primary key) and order (secondary key).
256 	 */
257 	for (sipp = sysinit; sipp < sysinit_end; sipp++) {
258 		for (xipp = sipp + 1; xipp < sysinit_end; xipp++) {
259 			if ((*sipp)->subsystem < (*xipp)->subsystem ||
260 			     ((*sipp)->subsystem == (*xipp)->subsystem &&
261 			      (*sipp)->order <= (*xipp)->order))
262 				continue;	/* skip*/
263 			save = *sipp;
264 			*sipp = *xipp;
265 			*xipp = save;
266 		}
267 	}
268 
269 #if defined(VERBOSE_SYSINIT)
270 	last = SI_SUB_COPYRIGHT;
271 	verbose = 0;
272 #if !defined(DDB)
273 	printf("VERBOSE_SYSINIT: DDB not enabled, symbol lookups disabled.\n");
274 #endif
275 #endif
276 
277 	/*
278 	 * Traverse the (now) ordered list of system initialization tasks.
279 	 * Perform each task, and continue on to the next task.
280 	 */
281 	for (sipp = sysinit; sipp < sysinit_end; sipp++) {
282 		if ((*sipp)->subsystem == SI_SUB_DUMMY)
283 			continue;	/* skip dummy task(s)*/
284 
285 		if ((*sipp)->subsystem == SI_SUB_DONE)
286 			continue;
287 
288 #if defined(VERBOSE_SYSINIT)
289 		if ((*sipp)->subsystem > last && verbose_sysinit != 0) {
290 			verbose = 1;
291 			last = (*sipp)->subsystem;
292 			printf("subsystem %x\n", last);
293 		}
294 		if (verbose) {
295 #if defined(DDB)
296 			const char *func, *data;
297 
298 			func = symbol_name((vm_offset_t)(*sipp)->func,
299 			    DB_STGY_PROC);
300 			data = symbol_name((vm_offset_t)(*sipp)->udata,
301 			    DB_STGY_ANY);
302 			if (func != NULL && data != NULL)
303 				printf("   %s(&%s)... ", func, data);
304 			else if (func != NULL)
305 				printf("   %s(%p)... ", func, (*sipp)->udata);
306 			else
307 #endif
308 				printf("   %p(%p)... ", (*sipp)->func,
309 				    (*sipp)->udata);
310 		}
311 #endif
312 
313 		/* Call function */
314 		(*((*sipp)->func))((*sipp)->udata);
315 
316 #if defined(VERBOSE_SYSINIT)
317 		if (verbose)
318 			printf("done.\n");
319 #endif
320 
321 		/* Check off the one we're just done */
322 		(*sipp)->subsystem = SI_SUB_DONE;
323 
324 		/* Check if we've installed more sysinit items via KLD */
325 		if (newsysinit != NULL) {
326 			if (sysinit != SET_BEGIN(sysinit_set))
327 				free(sysinit, M_TEMP);
328 			sysinit = newsysinit;
329 			sysinit_end = newsysinit_end;
330 			newsysinit = NULL;
331 			newsysinit_end = NULL;
332 			goto restart;
333 		}
334 	}
335 
336 	TSEXIT();	/* Here so we don't overlap with start_init. */
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 static int
392 null_fetch_syscall_args(struct thread *td __unused)
393 {
394 
395 	panic("null_fetch_syscall_args");
396 }
397 
398 static void
399 null_set_syscall_retval(struct thread *td __unused, int error __unused)
400 {
401 
402 	panic("null_set_syscall_retval");
403 }
404 
405 static void
406 null_set_fork_retval(struct thread *td __unused)
407 {
408 
409 }
410 
411 struct sysentvec null_sysvec = {
412 	.sv_size	= 0,
413 	.sv_table	= NULL,
414 	.sv_transtrap	= NULL,
415 	.sv_fixup	= NULL,
416 	.sv_sendsig	= NULL,
417 	.sv_sigcode	= NULL,
418 	.sv_szsigcode	= NULL,
419 	.sv_name	= "null",
420 	.sv_coredump	= NULL,
421 	.sv_imgact_try	= NULL,
422 	.sv_minsigstksz	= 0,
423 	.sv_minuser	= VM_MIN_ADDRESS,
424 	.sv_maxuser	= VM_MAXUSER_ADDRESS,
425 	.sv_usrstack	= USRSTACK,
426 	.sv_psstrings	= PS_STRINGS,
427 	.sv_psstringssz	= sizeof(struct ps_strings),
428 	.sv_stackprot	= VM_PROT_ALL,
429 	.sv_copyout_strings	= NULL,
430 	.sv_setregs	= NULL,
431 	.sv_fixlimit	= NULL,
432 	.sv_maxssiz	= NULL,
433 	.sv_flags	= 0,
434 	.sv_set_syscall_retval = null_set_syscall_retval,
435 	.sv_fetch_syscall_args = null_fetch_syscall_args,
436 	.sv_syscallnames = NULL,
437 	.sv_schedtail	= NULL,
438 	.sv_thread_detach = NULL,
439 	.sv_trap	= NULL,
440 	.sv_set_fork_retval = null_set_fork_retval,
441 	.sv_regset_begin = NULL,
442 	.sv_regset_end  = NULL,
443 };
444 
445 /*
446  * The two following SYSINIT's are proc0 specific glue code.  I am not
447  * convinced that they can not be safely combined, but their order of
448  * operation has been maintained as the same as the original init_main.c
449  * for right now.
450  */
451 /* ARGSUSED*/
452 static void
453 proc0_init(void *dummy __unused)
454 {
455 	struct proc *p;
456 	struct thread *td;
457 	struct ucred *newcred;
458 	struct uidinfo tmpuinfo;
459 	struct loginclass tmplc = {
460 		.lc_name = "",
461 	};
462 	vm_paddr_t pageablemem;
463 	int i;
464 
465 	GIANT_REQUIRED;
466 	p = &proc0;
467 	td = &thread0;
468 
469 	/*
470 	 * Initialize magic number and osrel.
471 	 */
472 	p->p_magic = P_MAGIC;
473 	p->p_osrel = osreldate;
474 
475 	/*
476 	 * Initialize thread and process structures.
477 	 */
478 	procinit();	/* set up proc zone */
479 	threadinit();	/* set up UMA zones */
480 
481 	/*
482 	 * Initialise scheduler resources.
483 	 * Add scheduler specific parts to proc, thread as needed.
484 	 */
485 	schedinit();	/* scheduler gets its house in order */
486 
487 	/*
488 	 * Create process 0 (the swapper).
489 	 */
490 	LIST_INSERT_HEAD(&allproc, p, p_list);
491 	LIST_INSERT_HEAD(PIDHASH(0), p, p_hash);
492 	mtx_init(&pgrp0.pg_mtx, "process group", NULL, MTX_DEF | MTX_DUPOK);
493 	p->p_pgrp = &pgrp0;
494 	LIST_INSERT_HEAD(PGRPHASH(0), &pgrp0, pg_hash);
495 	LIST_INIT(&pgrp0.pg_members);
496 	LIST_INSERT_HEAD(&pgrp0.pg_members, p, p_pglist);
497 
498 	pgrp0.pg_session = &session0;
499 	mtx_init(&session0.s_mtx, "session", NULL, MTX_DEF);
500 	refcount_init(&session0.s_count, 1);
501 	session0.s_leader = p;
502 
503 	p->p_sysent = &null_sysvec;
504 	p->p_flag = P_SYSTEM | P_INMEM | P_KPROC;
505 	p->p_flag2 = 0;
506 	p->p_state = PRS_NORMAL;
507 	p->p_klist = knlist_alloc(&p->p_mtx);
508 	STAILQ_INIT(&p->p_ktr);
509 	p->p_nice = NZERO;
510 	td->td_tid = THREAD0_TID;
511 	tidhash_add(td);
512 	TD_SET_STATE(td, TDS_RUNNING);
513 	td->td_pri_class = PRI_TIMESHARE;
514 	td->td_user_pri = PUSER;
515 	td->td_base_user_pri = PUSER;
516 	td->td_lend_user_pri = PRI_MAX;
517 	td->td_priority = PVM;
518 	td->td_base_pri = PVM;
519 	td->td_oncpu = curcpu;
520 	td->td_flags = TDF_INMEM;
521 	td->td_pflags = TDP_KTHREAD;
522 	td->td_cpuset = cpuset_thread0();
523 	td->td_domain.dr_policy = td->td_cpuset->cs_domain;
524 	prison0_init();
525 	p->p_peers = 0;
526 	p->p_leader = p;
527 	p->p_reaper = p;
528 	p->p_treeflag |= P_TREE_REAPER;
529 	LIST_INIT(&p->p_reaplist);
530 
531 	strncpy(p->p_comm, "kernel", sizeof (p->p_comm));
532 	strncpy(td->td_name, "swapper", sizeof (td->td_name));
533 
534 	callout_init_mtx(&p->p_itcallout, &p->p_mtx, 0);
535 	callout_init_mtx(&p->p_limco, &p->p_mtx, 0);
536 	callout_init(&td->td_slpcallout, 1);
537 	TAILQ_INIT(&p->p_kqtim_stop);
538 
539 	/* Create credentials. */
540 	newcred = crget();
541 	newcred->cr_ngroups = 1;	/* group 0 */
542 	/* A hack to prevent uifind from tripping over NULL pointers. */
543 	curthread->td_ucred = newcred;
544 	tmpuinfo.ui_uid = 1;
545 	newcred->cr_uidinfo = newcred->cr_ruidinfo = &tmpuinfo;
546 	newcred->cr_uidinfo = uifind(0);
547 	newcred->cr_ruidinfo = uifind(0);
548 	newcred->cr_loginclass = &tmplc;
549 	newcred->cr_loginclass = loginclass_find("default");
550 	/* End hack. creds get properly set later with thread_cow_get_proc */
551 	curthread->td_ucred = NULL;
552 	newcred->cr_prison = &prison0;
553 	newcred->cr_users++; /* avoid assertion failure */
554 	proc_set_cred_init(p, newcred);
555 	newcred->cr_users--;
556 	crfree(newcred);
557 #ifdef AUDIT
558 	audit_cred_kproc0(newcred);
559 #endif
560 #ifdef MAC
561 	mac_cred_create_swapper(newcred);
562 #endif
563 	/* Create sigacts. */
564 	p->p_sigacts = sigacts_alloc();
565 
566 	/* Initialize signal state for process 0. */
567 	siginit(&proc0);
568 
569 	/* Create the file descriptor table. */
570 	p->p_pd = pdinit(NULL, false);
571 	p->p_fd = fdinit(NULL, false, NULL);
572 	p->p_fdtol = NULL;
573 
574 	/* Create the limits structures. */
575 	p->p_limit = lim_alloc();
576 	for (i = 0; i < RLIM_NLIMITS; i++)
577 		p->p_limit->pl_rlimit[i].rlim_cur =
578 		    p->p_limit->pl_rlimit[i].rlim_max = RLIM_INFINITY;
579 	p->p_limit->pl_rlimit[RLIMIT_NOFILE].rlim_cur =
580 	    p->p_limit->pl_rlimit[RLIMIT_NOFILE].rlim_max = maxfiles;
581 	p->p_limit->pl_rlimit[RLIMIT_NPROC].rlim_cur =
582 	    p->p_limit->pl_rlimit[RLIMIT_NPROC].rlim_max = maxproc;
583 	p->p_limit->pl_rlimit[RLIMIT_DATA].rlim_cur = dfldsiz;
584 	p->p_limit->pl_rlimit[RLIMIT_DATA].rlim_max = maxdsiz;
585 	p->p_limit->pl_rlimit[RLIMIT_STACK].rlim_cur = dflssiz;
586 	p->p_limit->pl_rlimit[RLIMIT_STACK].rlim_max = maxssiz;
587 	/* Cast to avoid overflow on i386/PAE. */
588 	pageablemem = ptoa((vm_paddr_t)vm_free_count());
589 	p->p_limit->pl_rlimit[RLIMIT_RSS].rlim_cur =
590 	    p->p_limit->pl_rlimit[RLIMIT_RSS].rlim_max = pageablemem;
591 	p->p_limit->pl_rlimit[RLIMIT_MEMLOCK].rlim_cur = pageablemem / 3;
592 	p->p_limit->pl_rlimit[RLIMIT_MEMLOCK].rlim_max = pageablemem;
593 	p->p_cpulimit = RLIM_INFINITY;
594 
595 	PROC_LOCK(p);
596 	thread_cow_get_proc(td, p);
597 	PROC_UNLOCK(p);
598 
599 	/* Initialize resource accounting structures. */
600 	racct_create(&p->p_racct);
601 
602 	p->p_stats = pstats_alloc();
603 
604 	/* Allocate a prototype map so we have something to fork. */
605 	p->p_vmspace = &vmspace0;
606 	refcount_init(&vmspace0.vm_refcnt, 1);
607 	pmap_pinit0(vmspace_pmap(&vmspace0));
608 
609 	/*
610 	 * proc0 is not expected to enter usermode, so there is no special
611 	 * handling for sv_minuser here, like is done for exec_new_vmspace().
612 	 */
613 	vm_map_init(&vmspace0.vm_map, vmspace_pmap(&vmspace0),
614 	    p->p_sysent->sv_minuser, p->p_sysent->sv_maxuser);
615 
616 	/*
617 	 * Call the init and ctor for the new thread and proc.  We wait
618 	 * to do this until all other structures are fairly sane.
619 	 */
620 	EVENTHANDLER_DIRECT_INVOKE(process_init, p);
621 	EVENTHANDLER_DIRECT_INVOKE(thread_init, td);
622 #ifdef KDTRACE_HOOKS
623 	kdtrace_proc_ctor(p);
624 	kdtrace_thread_ctor(td);
625 #endif
626 	EVENTHANDLER_DIRECT_INVOKE(process_ctor, p);
627 	EVENTHANDLER_DIRECT_INVOKE(thread_ctor, td);
628 
629 	/*
630 	 * Charge root for one process.
631 	 */
632 	(void)chgproccnt(p->p_ucred->cr_ruidinfo, 1, 0);
633 	PROC_LOCK(p);
634 	racct_add_force(p, RACCT_NPROC, 1);
635 	PROC_UNLOCK(p);
636 }
637 SYSINIT(p0init, SI_SUB_INTRINSIC, SI_ORDER_FIRST, proc0_init, NULL);
638 
639 /* ARGSUSED*/
640 static void
641 proc0_post(void *dummy __unused)
642 {
643 	struct proc *p;
644 	struct rusage ru;
645 	struct thread *td;
646 
647 	/*
648 	 * Now we can look at the time, having had a chance to verify the
649 	 * time from the filesystem.  Pretend that proc0 started now.
650 	 */
651 	sx_slock(&allproc_lock);
652 	FOREACH_PROC_IN_SYSTEM(p) {
653 		PROC_LOCK(p);
654 		if (p->p_state == PRS_NEW) {
655 			PROC_UNLOCK(p);
656 			continue;
657 		}
658 		microuptime(&p->p_stats->p_start);
659 		PROC_STATLOCK(p);
660 		rufetch(p, &ru);	/* Clears thread stats */
661 		p->p_rux.rux_runtime = 0;
662 		p->p_rux.rux_uticks = 0;
663 		p->p_rux.rux_sticks = 0;
664 		p->p_rux.rux_iticks = 0;
665 		PROC_STATUNLOCK(p);
666 		FOREACH_THREAD_IN_PROC(p, td) {
667 			td->td_runtime = 0;
668 		}
669 		PROC_UNLOCK(p);
670 	}
671 	sx_sunlock(&allproc_lock);
672 	PCPU_SET(switchtime, cpu_ticks());
673 	PCPU_SET(switchticks, ticks);
674 }
675 SYSINIT(p0post, SI_SUB_INTRINSIC_POST, SI_ORDER_FIRST, proc0_post, NULL);
676 
677 /*
678  ***************************************************************************
679  ****
680  **** The following SYSINIT's and glue code should be moved to the
681  **** respective files on a per subsystem basis.
682  ****
683  ***************************************************************************
684  */
685 
686 /*
687  * List of paths to try when searching for "init".
688  */
689 static char init_path[MAXPATHLEN] =
690 #ifdef	INIT_PATH
691     __XSTRING(INIT_PATH);
692 #else
693     "/sbin/init:/sbin/oinit:/sbin/init.bak:/rescue/init";
694 #endif
695 SYSCTL_STRING(_kern, OID_AUTO, init_path, CTLFLAG_RD, init_path, 0,
696 	"Path used to search the init process");
697 
698 /*
699  * Shutdown timeout of init(8).
700  * Unused within kernel, but used to control init(8), hence do not remove.
701  */
702 #ifndef INIT_SHUTDOWN_TIMEOUT
703 #define INIT_SHUTDOWN_TIMEOUT 120
704 #endif
705 static int init_shutdown_timeout = INIT_SHUTDOWN_TIMEOUT;
706 SYSCTL_INT(_kern, OID_AUTO, init_shutdown_timeout,
707 	CTLFLAG_RW, &init_shutdown_timeout, 0, "Shutdown timeout of init(8). "
708 	"Unused within kernel, but used to control init(8)");
709 
710 /*
711  * Start the initial user process; try exec'ing each pathname in init_path.
712  * The program is invoked with one argument containing the boot flags.
713  */
714 static void
715 start_init(void *dummy)
716 {
717 	struct image_args args;
718 	int error;
719 	char *var, *path;
720 	char *free_init_path, *tmp_init_path;
721 	struct thread *td;
722 	struct proc *p;
723 	struct vmspace *oldvmspace;
724 
725 	TSENTER();	/* Here so we don't overlap with mi_startup. */
726 
727 	td = curthread;
728 	p = td->td_proc;
729 
730 	vfs_mountroot();
731 
732 	/* Wipe GELI passphrase from the environment. */
733 	kern_unsetenv("kern.geom.eli.passphrase");
734 
735 	/* For Multicons, report which console is primary to both */
736 	if (boothowto & RB_MULTIPLE) {
737 		if (boothowto & RB_SERIAL)
738 			printf("Dual Console: Serial Primary, Video Secondary\n");
739 		else
740 			printf("Dual Console: Video Primary, Serial Secondary\n");
741 	}
742 
743 	if ((var = kern_getenv("init_path")) != NULL) {
744 		strlcpy(init_path, var, sizeof(init_path));
745 		freeenv(var);
746 	}
747 	free_init_path = tmp_init_path = strdup(init_path, M_TEMP);
748 
749 	while ((path = strsep(&tmp_init_path, ":")) != NULL) {
750 		if (bootverbose)
751 			printf("start_init: trying %s\n", path);
752 
753 		memset(&args, 0, sizeof(args));
754 		error = exec_alloc_args(&args);
755 		if (error != 0)
756 			panic("%s: Can't allocate space for init arguments %d",
757 			    __func__, error);
758 
759 		error = exec_args_add_fname(&args, path, UIO_SYSSPACE);
760 		if (error != 0)
761 			panic("%s: Can't add fname %d", __func__, error);
762 		error = exec_args_add_arg(&args, path, UIO_SYSSPACE);
763 		if (error != 0)
764 			panic("%s: Can't add argv[0] %d", __func__, error);
765 		if (boothowto & RB_SINGLE)
766 			error = exec_args_add_arg(&args, "-s", UIO_SYSSPACE);
767 		if (error != 0)
768 			panic("%s: Can't add argv[0] %d", __func__, error);
769 
770 		/*
771 		 * Now try to exec the program.  If can't for any reason
772 		 * other than it doesn't exist, complain.
773 		 *
774 		 * Otherwise, return via fork_trampoline() all the way
775 		 * to user mode as init!
776 		 */
777 		KASSERT((td->td_pflags & TDP_EXECVMSPC) == 0,
778 		    ("nested execve"));
779 		oldvmspace = p->p_vmspace;
780 		error = kern_execve(td, &args, NULL, oldvmspace);
781 		KASSERT(error != 0,
782 		    ("kern_execve returned success, not EJUSTRETURN"));
783 		if (error == EJUSTRETURN) {
784 			exec_cleanup(td, oldvmspace);
785 			free(free_init_path, M_TEMP);
786 			TSEXIT();
787 			return;
788 		}
789 		if (error != ENOENT)
790 			printf("exec %s: error %d\n", path, error);
791 	}
792 	free(free_init_path, M_TEMP);
793 	printf("init: not found in path %s\n", init_path);
794 	panic("no init");
795 }
796 
797 /*
798  * Like kproc_create(), but runs in its own address space.  We do this
799  * early to reserve pid 1.  Note special case - do not make it
800  * runnable yet, init execution is started when userspace can be served.
801  */
802 static void
803 create_init(const void *udata __unused)
804 {
805 	struct fork_req fr;
806 	struct ucred *newcred, *oldcred;
807 	struct thread *td;
808 	int error;
809 
810 	bzero(&fr, sizeof(fr));
811 	fr.fr_flags = RFFDG | RFPROC | RFSTOPPED;
812 	fr.fr_procp = &initproc;
813 	error = fork1(&thread0, &fr);
814 	if (error)
815 		panic("cannot fork init: %d\n", error);
816 	KASSERT(initproc->p_pid == 1, ("create_init: initproc->p_pid != 1"));
817 	/* divorce init's credentials from the kernel's */
818 	newcred = crget();
819 	sx_xlock(&proctree_lock);
820 	PROC_LOCK(initproc);
821 	initproc->p_flag |= P_SYSTEM | P_INMEM;
822 	initproc->p_treeflag |= P_TREE_REAPER;
823 	oldcred = initproc->p_ucred;
824 	crcopy(newcred, oldcred);
825 #ifdef MAC
826 	mac_cred_create_init(newcred);
827 #endif
828 #ifdef AUDIT
829 	audit_cred_proc1(newcred);
830 #endif
831 	proc_set_cred(initproc, newcred);
832 	td = FIRST_THREAD_IN_PROC(initproc);
833 	crcowfree(td);
834 	td->td_realucred = crcowget(initproc->p_ucred);
835 	td->td_ucred = td->td_realucred;
836 	PROC_UNLOCK(initproc);
837 	sx_xunlock(&proctree_lock);
838 	crfree(oldcred);
839 	cpu_fork_kthread_handler(FIRST_THREAD_IN_PROC(initproc),
840 	    start_init, NULL);
841 }
842 SYSINIT(init, SI_SUB_CREATE_INIT, SI_ORDER_FIRST, create_init, NULL);
843 
844 /*
845  * Make it runnable now.
846  */
847 static void
848 kick_init(const void *udata __unused)
849 {
850 	struct thread *td;
851 
852 	td = FIRST_THREAD_IN_PROC(initproc);
853 	thread_lock(td);
854 	TD_SET_CAN_RUN(td);
855 	sched_add(td, SRQ_BORING);
856 }
857 SYSINIT(kickinit, SI_SUB_KTHREAD_INIT, SI_ORDER_MIDDLE, kick_init, NULL);
858 
859 /*
860  * DDB(4).
861  */
862 #ifdef DDB
863 static void
864 db_show_print_syinit(struct sysinit *sip, bool ddb)
865 {
866 	const char *sname, *funcname;
867 	c_db_sym_t sym;
868 	db_expr_t  offset;
869 
870 #define xprint(...)							\
871 	if (ddb)							\
872 		db_printf(__VA_ARGS__);					\
873 	else								\
874 		printf(__VA_ARGS__)
875 
876 	if (sip == NULL) {
877 		xprint("%s: no sysinit * given\n", __func__);
878 		return;
879 	}
880 
881 	sym = db_search_symbol((vm_offset_t)sip, DB_STGY_ANY, &offset);
882 	db_symbol_values(sym, &sname, NULL);
883 	sym = db_search_symbol((vm_offset_t)sip->func, DB_STGY_PROC, &offset);
884 	db_symbol_values(sym, &funcname, NULL);
885 	xprint("%s(%p)\n", (sname != NULL) ? sname : "", sip);
886 	xprint("  %#08x %#08x\n", sip->subsystem, sip->order);
887 	xprint("  %p(%s)(%p)\n",
888 	    sip->func, (funcname != NULL) ? funcname : "", sip->udata);
889 #undef xprint
890 }
891 
892 DB_SHOW_COMMAND(sysinit, db_show_sysinit)
893 {
894 	struct sysinit **sipp;
895 
896 	db_printf("SYSINIT vs Name(Ptr)\n");
897 	db_printf("  Subsystem  Order\n");
898 	db_printf("  Function(Name)(Arg)\n");
899 	for (sipp = sysinit; sipp < sysinit_end; sipp++) {
900 		db_show_print_syinit(*sipp, true);
901 		if (db_pager_quit)
902 			break;
903 	}
904 }
905 #endif /* DDB */
906