xref: /freebsd/sys/kern/init_main.c (revision 61ba55bcf70f2340f9c943c9571113b3fd8eda69)
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 lists.  Items are moved to sysinit_done_list when done.
167  */
168 static STAILQ_HEAD(sysinitlist, sysinit) sysinit_list;
169 static struct sysinitlist sysinit_done_list =
170     STAILQ_HEAD_INITIALIZER(sysinit_done_list);
171 
172 /*
173  * Compare two sysinits; return -1, 0, or 1 if a comes before, at the same time
174  * as, or after b.
175  */
176 static int
177 sysinit_compar(struct sysinit *a, struct sysinit *b, void *thunk __unused)
178 {
179 
180 	if (a->subsystem < b->subsystem)
181 		return (-1);
182 	if (a->subsystem > b->subsystem)
183 		return (1);
184 	if (a->order < b->order)
185 		return (-1);
186 	if (a->order > b->order)
187 		return (1);
188 	return (0);
189 }
190 
191 static void
192 sysinit_mklist(struct sysinitlist *list, struct sysinit **set,
193     struct sysinit **set_end)
194 {
195 	struct sysinit **sipp;
196 
197 	TSENTER();
198 	TSENTER2("listify");
199 	STAILQ_INIT(list);
200 	for (sipp = set; sipp < set_end; sipp++)
201 		STAILQ_INSERT_TAIL(list, *sipp, next);
202 	TSEXIT2("listify");
203 	TSENTER2("mergesort");
204 	STAILQ_MERGESORT(list, NULL, sysinit_compar, sysinit, next);
205 	TSEXIT2("mergesort");
206 	TSEXIT();
207 }
208 
209 /*
210  * Merge a new sysinit set into the sysinit list.
211  */
212 void
213 sysinit_add(struct sysinit **set, struct sysinit **set_end)
214 {
215 	struct sysinitlist new_list;
216 
217 	TSENTER();
218 
219 	/* Construct a sorted list from the new sysinits. */
220 	sysinit_mklist(&new_list, set, set_end);
221 
222 	/* Merge the new list into the existing one. */
223 	TSENTER2("STAILQ_MERGE");
224 	STAILQ_MERGE(&sysinit_list, &new_list, NULL, sysinit_compar, sysinit, next);
225 	TSEXIT2("STAILQ_MERGE");
226 
227 	TSEXIT();
228 }
229 
230 #if defined (DDB) && defined(VERBOSE_SYSINIT)
231 static const char *
232 symbol_name(vm_offset_t va, db_strategy_t strategy)
233 {
234 	const char *name;
235 	c_db_sym_t sym;
236 	db_expr_t  offset;
237 
238 	if (va == 0)
239 		return (NULL);
240 	sym = db_search_symbol(va, strategy, &offset);
241 	if (offset != 0)
242 		return (NULL);
243 	db_symbol_values(sym, &name, NULL);
244 	return (name);
245 }
246 #endif
247 
248 /*
249  * System startup; initialize the world, create process 0, mount root
250  * filesystem, and fork to create init and pagedaemon.  Most of the
251  * hard work is done in the lower-level initialization routines including
252  * startup(), which does memory initialization and autoconfiguration.
253  *
254  * This allows simple addition of new kernel subsystems that require
255  * boot time initialization.  It also allows substitution of subsystem
256  * (for instance, a scheduler, kernel profiler, or VM system) by object
257  * module.  Finally, it allows for optional "kernel threads".
258  */
259 void
260 mi_startup(void)
261 {
262 
263 	struct sysinit *sip;
264 	int last;
265 #if defined(VERBOSE_SYSINIT)
266 	int verbose;
267 #endif
268 
269 	TSENTER();
270 
271 	if (boothowto & RB_VERBOSE)
272 		bootverbose++;
273 
274 	/* Construct and sort sysinit list. */
275 	sysinit_mklist(&sysinit_list, SET_BEGIN(sysinit_set), SET_LIMIT(sysinit_set));
276 
277 	last = SI_SUB_COPYRIGHT;
278 #if defined(VERBOSE_SYSINIT)
279 	verbose = 0;
280 #if !defined(DDB)
281 	printf("VERBOSE_SYSINIT: DDB not enabled, symbol lookups disabled.\n");
282 #endif
283 #endif
284 
285 	/*
286 	 * Perform each system initialization task from the ordered list.  Note
287 	 * that if sysinit_list is modified (e.g. by a KLD) we will nonetheless
288 	 * always perform the earlist-sorted sysinit at each step; using the
289 	 * STAILQ_FOREACH macro would result in items being skipped if inserted
290 	 * earlier than the "current item".
291 	 */
292 	while ((sip = STAILQ_FIRST(&sysinit_list)) != NULL) {
293 		STAILQ_REMOVE_HEAD(&sysinit_list, next);
294 		STAILQ_INSERT_TAIL(&sysinit_done_list, sip, next);
295 
296 		if (sip->subsystem == SI_SUB_DUMMY)
297 			continue;	/* skip dummy task(s)*/
298 
299 		if (sip->subsystem > last)
300 			BOOTTRACE_INIT("sysinit 0x%7x", sip->subsystem);
301 
302 #if defined(VERBOSE_SYSINIT)
303 		if (sip->subsystem > last && verbose_sysinit != 0) {
304 			verbose = 1;
305 			printf("subsystem %x\n", last);
306 		}
307 		if (verbose) {
308 #if defined(DDB)
309 			const char *func, *data;
310 
311 			func = symbol_name((vm_offset_t)sip->func,
312 			    DB_STGY_PROC);
313 			data = symbol_name((vm_offset_t)sip->udata,
314 			    DB_STGY_ANY);
315 			if (func != NULL && data != NULL)
316 				printf("   %s(&%s)... ", func, data);
317 			else if (func != NULL)
318 				printf("   %s(%p)... ", func, sip->udata);
319 			else
320 #endif
321 				printf("   %p(%p)... ", sip->func,
322 				    sip->udata);
323 		}
324 #endif
325 
326 		/* Call function */
327 		(*(sip->func))(sip->udata);
328 
329 #if defined(VERBOSE_SYSINIT)
330 		if (verbose)
331 			printf("done.\n");
332 #endif
333 
334 		/* Check off the one we're just done */
335 		last = sip->subsystem;
336 	}
337 
338 	TSEXIT();	/* Here so we don't overlap with start_init. */
339 	BOOTTRACE("mi_startup done");
340 
341 	mtx_assert(&Giant, MA_OWNED | MA_NOTRECURSED);
342 	mtx_unlock(&Giant);
343 
344 	/*
345 	 * Now hand over this thread to swapper.
346 	 */
347 	swapper();
348 	/* NOTREACHED*/
349 }
350 
351 static void
352 print_caddr_t(void *data)
353 {
354 	printf("%s", (char *)data);
355 }
356 
357 static void
358 print_version(void *data __unused)
359 {
360 	int len;
361 
362 	/* Strip a trailing newline from version. */
363 	len = strlen(version);
364 	while (len > 0 && version[len - 1] == '\n')
365 		len--;
366 	printf("%.*s %s\n", len, version, machine);
367 	printf("%s\n", compiler_version);
368 }
369 
370 SYSINIT(announce, SI_SUB_COPYRIGHT, SI_ORDER_FIRST, print_caddr_t,
371     copyright);
372 SYSINIT(trademark, SI_SUB_COPYRIGHT, SI_ORDER_SECOND, print_caddr_t,
373     trademark);
374 SYSINIT(version, SI_SUB_COPYRIGHT, SI_ORDER_THIRD, print_version, NULL);
375 
376 #ifdef WITNESS
377 static char wit_warn[] =
378      "WARNING: WITNESS option enabled, expect reduced performance.\n";
379 SYSINIT(witwarn, SI_SUB_COPYRIGHT, SI_ORDER_FOURTH,
380    print_caddr_t, wit_warn);
381 SYSINIT(witwarn2, SI_SUB_LAST, SI_ORDER_FOURTH,
382    print_caddr_t, wit_warn);
383 #endif
384 
385 #ifdef DIAGNOSTIC
386 static char diag_warn[] =
387      "WARNING: DIAGNOSTIC option enabled, expect reduced performance.\n";
388 SYSINIT(diagwarn, SI_SUB_COPYRIGHT, SI_ORDER_FIFTH,
389     print_caddr_t, diag_warn);
390 SYSINIT(diagwarn2, SI_SUB_LAST, SI_ORDER_FIFTH,
391     print_caddr_t, diag_warn);
392 #endif
393 
394 #if __SIZEOF_LONG__ == 4
395 static char ilp32_warn[] =
396     "WARNING: 32-bit kernels are deprecated and may be removed in FreeBSD 15.0.\n";
397 SYSINIT(ilp32warn, SI_SUB_COPYRIGHT, SI_ORDER_FIFTH,
398     print_caddr_t, ilp32_warn);
399 SYSINIT(ilp32warn2, SI_SUB_LAST, SI_ORDER_FIFTH,
400     print_caddr_t, ilp32_warn);
401 #endif
402 
403 static int
404 null_fetch_syscall_args(struct thread *td __unused)
405 {
406 
407 	panic("null_fetch_syscall_args");
408 }
409 
410 static void
411 null_set_syscall_retval(struct thread *td __unused, int error __unused)
412 {
413 
414 	panic("null_set_syscall_retval");
415 }
416 
417 static void
418 null_set_fork_retval(struct thread *td __unused)
419 {
420 
421 }
422 
423 struct sysentvec null_sysvec = {
424 	.sv_size	= 0,
425 	.sv_table	= NULL,
426 	.sv_fixup	= NULL,
427 	.sv_sendsig	= NULL,
428 	.sv_sigcode	= NULL,
429 	.sv_szsigcode	= NULL,
430 	.sv_name	= "null",
431 	.sv_coredump	= NULL,
432 	.sv_minsigstksz	= 0,
433 	.sv_minuser	= VM_MIN_ADDRESS,
434 	.sv_maxuser	= VM_MAXUSER_ADDRESS,
435 	.sv_usrstack	= USRSTACK,
436 	.sv_psstrings	= PS_STRINGS,
437 	.sv_psstringssz	= sizeof(struct ps_strings),
438 	.sv_stackprot	= VM_PROT_ALL,
439 	.sv_copyout_strings	= NULL,
440 	.sv_setregs	= NULL,
441 	.sv_fixlimit	= NULL,
442 	.sv_maxssiz	= NULL,
443 	.sv_flags	= 0,
444 	.sv_set_syscall_retval = null_set_syscall_retval,
445 	.sv_fetch_syscall_args = null_fetch_syscall_args,
446 	.sv_syscallnames = NULL,
447 	.sv_schedtail	= NULL,
448 	.sv_thread_detach = NULL,
449 	.sv_trap	= NULL,
450 	.sv_set_fork_retval = null_set_fork_retval,
451 	.sv_regset_begin = NULL,
452 	.sv_regset_end  = NULL,
453 };
454 
455 /*
456  * The two following SYSINIT's are proc0 specific glue code.  I am not
457  * convinced that they can not be safely combined, but their order of
458  * operation has been maintained as the same as the original init_main.c
459  * for right now.
460  */
461 /* ARGSUSED*/
462 static void
463 proc0_init(void *dummy __unused)
464 {
465 	struct proc *p;
466 	struct thread *td;
467 	struct ucred *newcred;
468 	struct uidinfo tmpuinfo;
469 	struct loginclass tmplc = {
470 		.lc_name = "",
471 	};
472 	vm_paddr_t pageablemem;
473 	int i;
474 
475 	GIANT_REQUIRED;
476 	p = &proc0;
477 	td = &thread0;
478 
479 	/*
480 	 * Initialize magic number and osrel.
481 	 */
482 	p->p_magic = P_MAGIC;
483 	p->p_osrel = osreldate;
484 
485 	/*
486 	 * Initialize thread and process structures.
487 	 */
488 	procinit();	/* set up proc zone */
489 	threadinit();	/* set up UMA zones */
490 
491 	/*
492 	 * Initialise scheduler resources.
493 	 * Add scheduler specific parts to proc, thread as needed.
494 	 */
495 	schedinit();	/* scheduler gets its house in order */
496 
497 	/*
498 	 * Create process 0 (the swapper).
499 	 */
500 	LIST_INSERT_HEAD(&allproc, p, p_list);
501 	LIST_INSERT_HEAD(PIDHASH(0), p, p_hash);
502 	mtx_init(&pgrp0.pg_mtx, "process group", NULL, MTX_DEF | MTX_DUPOK);
503 	sx_init(&pgrp0.pg_killsx, "killpg racer");
504 	p->p_pgrp = &pgrp0;
505 	LIST_INSERT_HEAD(PGRPHASH(0), &pgrp0, pg_hash);
506 	LIST_INIT(&pgrp0.pg_members);
507 	LIST_INSERT_HEAD(&pgrp0.pg_members, p, p_pglist);
508 
509 	pgrp0.pg_session = &session0;
510 	mtx_init(&session0.s_mtx, "session", NULL, MTX_DEF);
511 	refcount_init(&session0.s_count, 1);
512 	session0.s_leader = p;
513 
514 	p->p_sysent = &null_sysvec;
515 	p->p_flag = P_SYSTEM | P_INMEM | P_KPROC;
516 	p->p_flag2 = 0;
517 	p->p_state = PRS_NORMAL;
518 	p->p_klist = knlist_alloc(&p->p_mtx);
519 	STAILQ_INIT(&p->p_ktr);
520 	p->p_nice = NZERO;
521 	td->td_tid = THREAD0_TID;
522 	tidhash_add(td);
523 	TD_SET_STATE(td, TDS_RUNNING);
524 	td->td_pri_class = PRI_TIMESHARE;
525 	td->td_user_pri = PUSER;
526 	td->td_base_user_pri = PUSER;
527 	td->td_lend_user_pri = PRI_MAX;
528 	td->td_priority = PVM;
529 	td->td_base_pri = PVM;
530 	td->td_oncpu = curcpu;
531 	td->td_flags = TDF_INMEM;
532 	td->td_pflags = TDP_KTHREAD;
533 	td->td_cpuset = cpuset_thread0();
534 	td->td_domain.dr_policy = td->td_cpuset->cs_domain;
535 	prison0_init();
536 	p->p_peers = 0;
537 	p->p_leader = p;
538 	p->p_reaper = p;
539 	p->p_treeflag |= P_TREE_REAPER;
540 	LIST_INIT(&p->p_reaplist);
541 
542 	strncpy(p->p_comm, "kernel", sizeof (p->p_comm));
543 	strncpy(td->td_name, "swapper", sizeof (td->td_name));
544 
545 	callout_init_mtx(&p->p_itcallout, &p->p_mtx, 0);
546 	callout_init_mtx(&p->p_limco, &p->p_mtx, 0);
547 	callout_init(&td->td_slpcallout, 1);
548 	TAILQ_INIT(&p->p_kqtim_stop);
549 
550 	/* Create credentials. */
551 	newcred = crget();
552 	newcred->cr_ngroups = 1;	/* group 0 */
553 	/* A hack to prevent uifind from tripping over NULL pointers. */
554 	curthread->td_ucred = newcred;
555 	tmpuinfo.ui_uid = 1;
556 	newcred->cr_uidinfo = newcred->cr_ruidinfo = &tmpuinfo;
557 	newcred->cr_uidinfo = uifind(0);
558 	newcred->cr_ruidinfo = uifind(0);
559 	newcred->cr_loginclass = &tmplc;
560 	newcred->cr_loginclass = loginclass_find("default");
561 	/* End hack. creds get properly set later with thread_cow_get_proc */
562 	curthread->td_ucred = NULL;
563 	newcred->cr_prison = &prison0;
564 	newcred->cr_users++; /* avoid assertion failure */
565 	proc_set_cred_init(p, newcred);
566 	newcred->cr_users--;
567 	crfree(newcred);
568 #ifdef AUDIT
569 	audit_cred_kproc0(newcred);
570 #endif
571 #ifdef MAC
572 	mac_cred_create_swapper(newcred);
573 #endif
574 	/* Create sigacts. */
575 	p->p_sigacts = sigacts_alloc();
576 
577 	/* Initialize signal state for process 0. */
578 	siginit(&proc0);
579 
580 	/* Create the file descriptor table. */
581 	p->p_pd = pdinit(NULL, false);
582 	p->p_fd = fdinit();
583 	p->p_fdtol = NULL;
584 
585 	/* Create the limits structures. */
586 	p->p_limit = lim_alloc();
587 	for (i = 0; i < RLIM_NLIMITS; i++)
588 		p->p_limit->pl_rlimit[i].rlim_cur =
589 		    p->p_limit->pl_rlimit[i].rlim_max = RLIM_INFINITY;
590 	p->p_limit->pl_rlimit[RLIMIT_NOFILE].rlim_cur =
591 	    p->p_limit->pl_rlimit[RLIMIT_NOFILE].rlim_max = maxfiles;
592 	p->p_limit->pl_rlimit[RLIMIT_NPROC].rlim_cur =
593 	    p->p_limit->pl_rlimit[RLIMIT_NPROC].rlim_max = maxproc;
594 	p->p_limit->pl_rlimit[RLIMIT_DATA].rlim_cur = dfldsiz;
595 	p->p_limit->pl_rlimit[RLIMIT_DATA].rlim_max = maxdsiz;
596 	p->p_limit->pl_rlimit[RLIMIT_STACK].rlim_cur = dflssiz;
597 	p->p_limit->pl_rlimit[RLIMIT_STACK].rlim_max = maxssiz;
598 	/* Cast to avoid overflow on i386/PAE. */
599 	pageablemem = ptoa((vm_paddr_t)vm_free_count());
600 	p->p_limit->pl_rlimit[RLIMIT_RSS].rlim_cur =
601 	    p->p_limit->pl_rlimit[RLIMIT_RSS].rlim_max = pageablemem;
602 	p->p_limit->pl_rlimit[RLIMIT_MEMLOCK].rlim_cur = pageablemem / 3;
603 	p->p_limit->pl_rlimit[RLIMIT_MEMLOCK].rlim_max = pageablemem;
604 	p->p_cpulimit = RLIM_INFINITY;
605 
606 	PROC_LOCK(p);
607 	thread_cow_get_proc(td, p);
608 	PROC_UNLOCK(p);
609 
610 	/* Initialize resource accounting structures. */
611 	racct_create(&p->p_racct);
612 
613 	p->p_stats = pstats_alloc();
614 
615 	/* Allocate a prototype map so we have something to fork. */
616 	p->p_vmspace = &vmspace0;
617 	refcount_init(&vmspace0.vm_refcnt, 1);
618 	pmap_pinit0(vmspace_pmap(&vmspace0));
619 
620 	/*
621 	 * proc0 is not expected to enter usermode, so there is no special
622 	 * handling for sv_minuser here, like is done for exec_new_vmspace().
623 	 */
624 	vm_map_init(&vmspace0.vm_map, vmspace_pmap(&vmspace0),
625 	    p->p_sysent->sv_minuser, p->p_sysent->sv_maxuser);
626 
627 	/*
628 	 * Call the init and ctor for the new thread and proc.  We wait
629 	 * to do this until all other structures are fairly sane.
630 	 */
631 	EVENTHANDLER_DIRECT_INVOKE(process_init, p);
632 	EVENTHANDLER_DIRECT_INVOKE(thread_init, td);
633 #ifdef KDTRACE_HOOKS
634 	kdtrace_proc_ctor(p);
635 	kdtrace_thread_ctor(td);
636 #endif
637 	EVENTHANDLER_DIRECT_INVOKE(process_ctor, p);
638 	EVENTHANDLER_DIRECT_INVOKE(thread_ctor, td);
639 
640 	/*
641 	 * Charge root for one process.
642 	 */
643 	(void)chgproccnt(p->p_ucred->cr_ruidinfo, 1, 0);
644 	PROC_LOCK(p);
645 	racct_add_force(p, RACCT_NPROC, 1);
646 	PROC_UNLOCK(p);
647 }
648 SYSINIT(p0init, SI_SUB_INTRINSIC, SI_ORDER_FIRST, proc0_init, NULL);
649 
650 /* ARGSUSED*/
651 static void
652 proc0_post(void *dummy __unused)
653 {
654 	struct proc *p;
655 	struct rusage ru;
656 	struct thread *td;
657 
658 	/*
659 	 * Now we can look at the time, having had a chance to verify the
660 	 * time from the filesystem.  Pretend that proc0 started now.
661 	 */
662 	sx_slock(&allproc_lock);
663 	FOREACH_PROC_IN_SYSTEM(p) {
664 		PROC_LOCK(p);
665 		if (p->p_state == PRS_NEW) {
666 			PROC_UNLOCK(p);
667 			continue;
668 		}
669 		microuptime(&p->p_stats->p_start);
670 		PROC_STATLOCK(p);
671 		rufetch(p, &ru);	/* Clears thread stats */
672 		p->p_rux.rux_runtime = 0;
673 		p->p_rux.rux_uticks = 0;
674 		p->p_rux.rux_sticks = 0;
675 		p->p_rux.rux_iticks = 0;
676 		PROC_STATUNLOCK(p);
677 		FOREACH_THREAD_IN_PROC(p, td) {
678 			td->td_runtime = 0;
679 		}
680 		PROC_UNLOCK(p);
681 	}
682 	sx_sunlock(&allproc_lock);
683 	PCPU_SET(switchtime, cpu_ticks());
684 	PCPU_SET(switchticks, ticks);
685 }
686 SYSINIT(p0post, SI_SUB_INTRINSIC_POST, SI_ORDER_FIRST, proc0_post, NULL);
687 
688 /*
689  ***************************************************************************
690  ****
691  **** The following SYSINIT's and glue code should be moved to the
692  **** respective files on a per subsystem basis.
693  ****
694  ***************************************************************************
695  */
696 
697 /*
698  * List of paths to try when searching for "init".
699  */
700 static char init_path[MAXPATHLEN] =
701 #ifdef	INIT_PATH
702     __XSTRING(INIT_PATH);
703 #else
704     "/sbin/init:/sbin/oinit:/sbin/init.bak:/rescue/init";
705 #endif
706 SYSCTL_STRING(_kern, OID_AUTO, init_path, CTLFLAG_RD, init_path, 0,
707 	"Path used to search the init process");
708 
709 /*
710  * Shutdown timeout of init(8).
711  * Unused within kernel, but used to control init(8), hence do not remove.
712  */
713 #ifndef INIT_SHUTDOWN_TIMEOUT
714 #define INIT_SHUTDOWN_TIMEOUT 120
715 #endif
716 static int init_shutdown_timeout = INIT_SHUTDOWN_TIMEOUT;
717 SYSCTL_INT(_kern, OID_AUTO, init_shutdown_timeout,
718 	CTLFLAG_RW, &init_shutdown_timeout, 0, "Shutdown timeout of init(8). "
719 	"Unused within kernel, but used to control init(8)");
720 
721 /*
722  * Start the initial user process; try exec'ing each pathname in init_path.
723  * The program is invoked with one argument containing the boot flags.
724  */
725 static void
726 start_init(void *dummy)
727 {
728 	struct image_args args;
729 	int error;
730 	char *var, *path;
731 	char *free_init_path, *tmp_init_path;
732 	struct thread *td;
733 	struct proc *p;
734 	struct vmspace *oldvmspace;
735 
736 	TSENTER();	/* Here so we don't overlap with mi_startup. */
737 
738 	td = curthread;
739 	p = td->td_proc;
740 
741 	vfs_mountroot();
742 
743 	/* Wipe GELI passphrase from the environment. */
744 	kern_unsetenv("kern.geom.eli.passphrase");
745 
746 	/* For Multicons, report which console is primary to both */
747 	if (boothowto & RB_MULTIPLE) {
748 		if (boothowto & RB_SERIAL)
749 			printf("Dual Console: Serial Primary, Video Secondary\n");
750 		else
751 			printf("Dual Console: Video Primary, Serial Secondary\n");
752 	}
753 
754 	if ((var = kern_getenv("init_path")) != NULL) {
755 		strlcpy(init_path, var, sizeof(init_path));
756 		freeenv(var);
757 	}
758 	free_init_path = tmp_init_path = strdup(init_path, M_TEMP);
759 
760 	while ((path = strsep(&tmp_init_path, ":")) != NULL) {
761 		if (bootverbose)
762 			printf("start_init: trying %s\n", path);
763 
764 		memset(&args, 0, sizeof(args));
765 		error = exec_alloc_args(&args);
766 		if (error != 0)
767 			panic("%s: Can't allocate space for init arguments %d",
768 			    __func__, error);
769 
770 		error = exec_args_add_fname(&args, path, UIO_SYSSPACE);
771 		if (error != 0)
772 			panic("%s: Can't add fname %d", __func__, error);
773 		error = exec_args_add_arg(&args, path, UIO_SYSSPACE);
774 		if (error != 0)
775 			panic("%s: Can't add argv[0] %d", __func__, error);
776 		if (boothowto & RB_SINGLE)
777 			error = exec_args_add_arg(&args, "-s", UIO_SYSSPACE);
778 		if (error != 0)
779 			panic("%s: Can't add argv[0] %d", __func__, error);
780 
781 		/*
782 		 * Now try to exec the program.  If can't for any reason
783 		 * other than it doesn't exist, complain.
784 		 *
785 		 * Otherwise, return via fork_trampoline() all the way
786 		 * to user mode as init!
787 		 */
788 		KASSERT((td->td_pflags & TDP_EXECVMSPC) == 0,
789 		    ("nested execve"));
790 		oldvmspace = p->p_vmspace;
791 		error = kern_execve(td, &args, NULL, oldvmspace);
792 		KASSERT(error != 0,
793 		    ("kern_execve returned success, not EJUSTRETURN"));
794 		if (error == EJUSTRETURN) {
795 			exec_cleanup(td, oldvmspace);
796 			free(free_init_path, M_TEMP);
797 			TSEXIT();
798 			return;
799 		}
800 		if (error != ENOENT)
801 			printf("exec %s: error %d\n", path, error);
802 	}
803 	free(free_init_path, M_TEMP);
804 	printf("init: not found in path %s\n", init_path);
805 	panic("no init");
806 }
807 
808 /*
809  * Like kproc_create(), but runs in its own address space.  We do this
810  * early to reserve pid 1.  Note special case - do not make it
811  * runnable yet, init execution is started when userspace can be served.
812  */
813 static void
814 create_init(const void *udata __unused)
815 {
816 	struct fork_req fr;
817 	struct ucred *newcred, *oldcred;
818 	struct thread *td;
819 	int error;
820 
821 	bzero(&fr, sizeof(fr));
822 	fr.fr_flags = RFFDG | RFPROC | RFSTOPPED;
823 	fr.fr_procp = &initproc;
824 	error = fork1(&thread0, &fr);
825 	if (error)
826 		panic("cannot fork init: %d\n", error);
827 	KASSERT(initproc->p_pid == 1, ("create_init: initproc->p_pid != 1"));
828 	/* divorce init's credentials from the kernel's */
829 	newcred = crget();
830 	sx_xlock(&proctree_lock);
831 	PROC_LOCK(initproc);
832 	initproc->p_flag |= P_SYSTEM | P_INMEM;
833 	initproc->p_treeflag |= P_TREE_REAPER;
834 	oldcred = initproc->p_ucred;
835 	crcopy(newcred, oldcred);
836 #ifdef MAC
837 	mac_cred_create_init(newcred);
838 #endif
839 #ifdef AUDIT
840 	audit_cred_proc1(newcred);
841 #endif
842 	proc_set_cred(initproc, newcred);
843 	td = FIRST_THREAD_IN_PROC(initproc);
844 	crcowfree(td);
845 	td->td_realucred = crcowget(initproc->p_ucred);
846 	td->td_ucred = td->td_realucred;
847 	PROC_UNLOCK(initproc);
848 	sx_xunlock(&proctree_lock);
849 	crfree(oldcred);
850 	cpu_fork_kthread_handler(FIRST_THREAD_IN_PROC(initproc),
851 	    start_init, NULL);
852 }
853 SYSINIT(init, SI_SUB_CREATE_INIT, SI_ORDER_FIRST, create_init, NULL);
854 
855 /*
856  * Make it runnable now.
857  */
858 static void
859 kick_init(const void *udata __unused)
860 {
861 	struct thread *td;
862 
863 	td = FIRST_THREAD_IN_PROC(initproc);
864 	thread_lock(td);
865 	TD_SET_CAN_RUN(td);
866 	sched_add(td, SRQ_BORING);
867 }
868 SYSINIT(kickinit, SI_SUB_KTHREAD_INIT, SI_ORDER_MIDDLE, kick_init, NULL);
869 
870 /*
871  * DDB(4).
872  */
873 #ifdef DDB
874 static void
875 db_show_print_syinit(struct sysinit *sip, bool ddb)
876 {
877 	const char *sname, *funcname;
878 	c_db_sym_t sym;
879 	db_expr_t  offset;
880 
881 #define xprint(...)							\
882 	if (ddb)							\
883 		db_printf(__VA_ARGS__);					\
884 	else								\
885 		printf(__VA_ARGS__)
886 
887 	if (sip == NULL) {
888 		xprint("%s: no sysinit * given\n", __func__);
889 		return;
890 	}
891 
892 	sym = db_search_symbol((vm_offset_t)sip, DB_STGY_ANY, &offset);
893 	db_symbol_values(sym, &sname, NULL);
894 	sym = db_search_symbol((vm_offset_t)sip->func, DB_STGY_PROC, &offset);
895 	db_symbol_values(sym, &funcname, NULL);
896 	xprint("%s(%p)\n", (sname != NULL) ? sname : "", sip);
897 	xprint("  %#08x %#08x\n", sip->subsystem, sip->order);
898 	xprint("  %p(%s)(%p)\n",
899 	    sip->func, (funcname != NULL) ? funcname : "", sip->udata);
900 #undef xprint
901 }
902 
903 DB_SHOW_COMMAND_FLAGS(sysinit, db_show_sysinit, DB_CMD_MEMSAFE)
904 {
905 	struct sysinit *sip;
906 
907 	db_printf("SYSINIT vs Name(Ptr)\n");
908 	db_printf("  Subsystem  Order\n");
909 	db_printf("  Function(Name)(Arg)\n");
910 	STAILQ_FOREACH(sip, &sysinit_done_list, next) {
911 		db_show_print_syinit(sip, true);
912 		if (db_pager_quit)
913 			return;
914 	}
915 	STAILQ_FOREACH(sip, &sysinit_list, next) {
916 		db_show_print_syinit(sip, true);
917 		if (db_pager_quit)
918 			break;
919 	}
920 }
921 #endif /* DDB */
922