xref: /freebsd/sys/kern/init_main.c (revision 7937bfbc0ca53fe7cdd0d54414f9296e273a518e)
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 
44 #include "opt_ddb.h"
45 #include "opt_kdb.h"
46 #include "opt_init_path.h"
47 #include "opt_verbose_sysinit.h"
48 
49 #include <sys/param.h>
50 #include <sys/systm.h>
51 #include <sys/boottrace.h>
52 #include <sys/conf.h>
53 #include <sys/cpuset.h>
54 #include <sys/dtrace_bsd.h>
55 #include <sys/epoch.h>
56 #include <sys/eventhandler.h>
57 #include <sys/exec.h>
58 #include <sys/file.h>
59 #include <sys/filedesc.h>
60 #include <sys/imgact.h>
61 #include <sys/jail.h>
62 #include <sys/kernel.h>
63 #include <sys/ktr.h>
64 #include <sys/lock.h>
65 #include <sys/loginclass.h>
66 #include <sys/malloc.h>
67 #include <sys/mount.h>
68 #include <sys/mutex.h>
69 #include <sys/proc.h>
70 #include <sys/racct.h>
71 #include <sys/reboot.h>
72 #include <sys/resourcevar.h>
73 #include <sys/queue.h>
74 #include <sys/queue_mergesort.h>
75 #include <sys/sched.h>
76 #include <sys/signalvar.h>
77 #include <sys/sx.h>
78 #include <sys/syscallsubr.h>
79 #include <sys/sysctl.h>
80 #include <sys/sysent.h>
81 #include <sys/sysproto.h>
82 #include <sys/unistd.h>
83 #include <sys/vmmeter.h>
84 #include <sys/vnode.h>
85 
86 #include <machine/cpu.h>
87 
88 #include <security/audit/audit.h>
89 #include <security/mac/mac_framework.h>
90 
91 #include <vm/vm.h>
92 #include <vm/vm_param.h>
93 #include <vm/vm_extern.h>
94 #include <vm/pmap.h>
95 #include <vm/vm_map.h>
96 #include <sys/copyright.h>
97 
98 #include <ddb/ddb.h>
99 #include <ddb/db_sym.h>
100 
101 void mi_startup(void);				/* Should be elsewhere */
102 
103 /* Components of the first process -- never freed. */
104 static struct session session0;
105 static struct pgrp pgrp0;
106 struct	proc proc0;
107 struct thread0_storage thread0_st __aligned(32);
108 struct	vmspace vmspace0;
109 struct	proc *initproc;
110 
111 int
112 linux_alloc_current_noop(struct thread *td __unused, int flags __unused)
113 {
114 	return (0);
115 }
116 int (*lkpi_alloc_current)(struct thread *, int) = linux_alloc_current_noop;
117 
118 #ifndef BOOTHOWTO
119 #define	BOOTHOWTO	0
120 #endif
121 int	boothowto = BOOTHOWTO;	/* initialized so that it can be patched */
122 SYSCTL_INT(_debug, OID_AUTO, boothowto, CTLFLAG_RD, &boothowto, 0,
123 	"Boot control flags, passed from loader");
124 
125 #ifndef BOOTVERBOSE
126 #define	BOOTVERBOSE	0
127 #endif
128 int	bootverbose = BOOTVERBOSE;
129 SYSCTL_INT(_debug, OID_AUTO, bootverbose, CTLFLAG_RW, &bootverbose, 0,
130 	"Control the output of verbose kernel messages");
131 
132 #ifdef VERBOSE_SYSINIT
133 /*
134  * We'll use the defined value of VERBOSE_SYSINIT from the kernel config to
135  * dictate the default VERBOSE_SYSINIT behavior.  Significant values for this
136  * option and associated tunable are:
137  * - 0, 'compiled in but silent by default'
138  * - 1, 'compiled in but verbose by default' (default)
139  */
140 int	verbose_sysinit = VERBOSE_SYSINIT;
141 TUNABLE_INT("debug.verbose_sysinit", &verbose_sysinit);
142 #endif
143 
144 #ifdef INVARIANTS
145 FEATURE(invariants, "Kernel compiled with INVARIANTS, may affect performance");
146 #endif
147 
148 /*
149  * This ensures that there is at least one entry so that the sysinit_set
150  * symbol is not undefined.  A sybsystem ID of SI_SUB_DUMMY is never
151  * executed.
152  */
153 SYSINIT(placeholder, SI_SUB_DUMMY, SI_ORDER_ANY, NULL, NULL);
154 
155 /*
156  * The sysinit linker set compiled into the kernel.  These are placed onto the
157  * sysinit list by mi_startup; sysinit_add can add (e.g., from klds) additional
158  * sysinits to the linked list but the linker set here does not change.
159  */
160 SET_DECLARE(sysinit_set, struct sysinit);
161 
162 /*
163  * The sysinit lists.  Items are moved to sysinit_done_list when done.
164  */
165 static STAILQ_HEAD(sysinitlist, sysinit) sysinit_list;
166 static struct sysinitlist sysinit_done_list =
167     STAILQ_HEAD_INITIALIZER(sysinit_done_list);
168 
169 /*
170  * Compare two sysinits; return -1, 0, or 1 if a comes before, at the same time
171  * as, or after b.
172  */
173 static int
174 sysinit_compar(struct sysinit *a, struct sysinit *b, void *thunk __unused)
175 {
176 
177 	if (a->subsystem < b->subsystem)
178 		return (-1);
179 	if (a->subsystem > b->subsystem)
180 		return (1);
181 	if (a->order < b->order)
182 		return (-1);
183 	if (a->order > b->order)
184 		return (1);
185 	return (0);
186 }
187 
188 static void
189 sysinit_mklist(struct sysinitlist *list, struct sysinit **set,
190     struct sysinit **set_end)
191 {
192 	struct sysinit **sipp;
193 
194 	TSENTER();
195 	TSENTER2("listify");
196 	STAILQ_INIT(list);
197 	for (sipp = set; sipp < set_end; sipp++)
198 		STAILQ_INSERT_TAIL(list, *sipp, next);
199 	TSEXIT2("listify");
200 	TSENTER2("mergesort");
201 	STAILQ_MERGESORT(list, NULL, sysinit_compar, sysinit, next);
202 	TSEXIT2("mergesort");
203 	TSEXIT();
204 }
205 
206 /*
207  * Merge a new sysinit set into the sysinit list.
208  */
209 void
210 sysinit_add(struct sysinit **set, struct sysinit **set_end)
211 {
212 	struct sysinitlist new_list;
213 
214 	TSENTER();
215 
216 	/* Construct a sorted list from the new sysinits. */
217 	sysinit_mklist(&new_list, set, set_end);
218 
219 	/* Merge the new list into the existing one. */
220 	TSENTER2("STAILQ_MERGE");
221 	STAILQ_MERGE(&sysinit_list, &new_list, NULL, sysinit_compar, sysinit, next);
222 	TSEXIT2("STAILQ_MERGE");
223 
224 	TSEXIT();
225 }
226 
227 #if defined (DDB) && defined(VERBOSE_SYSINIT)
228 static const char *
229 symbol_name(vm_offset_t va, db_strategy_t strategy)
230 {
231 	const char *name;
232 	c_db_sym_t sym;
233 	db_expr_t  offset;
234 
235 	if (va == 0)
236 		return (NULL);
237 	sym = db_search_symbol(va, strategy, &offset);
238 	if (offset != 0)
239 		return (NULL);
240 	db_symbol_values(sym, &name, NULL);
241 	return (name);
242 }
243 #endif
244 
245 /*
246  * System startup; initialize the world, create process 0, mount root
247  * filesystem, and fork to create init and pagedaemon.  Most of the
248  * hard work is done in the lower-level initialization routines including
249  * startup(), which does memory initialization and autoconfiguration.
250  *
251  * This allows simple addition of new kernel subsystems that require
252  * boot time initialization.  It also allows substitution of subsystem
253  * (for instance, a scheduler, kernel profiler, or VM system) by object
254  * module.  Finally, it allows for optional "kernel threads".
255  */
256 void
257 mi_startup(void)
258 {
259 	struct sysinit *sip;
260 	int last;
261 #if defined(VERBOSE_SYSINIT)
262 	int verbose;
263 #endif
264 
265 	TSENTER();
266 
267 	if (boothowto & RB_VERBOSE)
268 		bootverbose++;
269 
270 	/* Construct and sort sysinit list. */
271 	sysinit_mklist(&sysinit_list, SET_BEGIN(sysinit_set), SET_LIMIT(sysinit_set));
272 
273 	last = SI_SUB_COPYRIGHT;
274 #if defined(VERBOSE_SYSINIT)
275 	verbose = 0;
276 #if !defined(DDB)
277 	printf("VERBOSE_SYSINIT: DDB not enabled, symbol lookups disabled.\n");
278 #endif
279 #endif
280 
281 	/*
282 	 * Perform each system initialization task from the ordered list.  Note
283 	 * that if sysinit_list is modified (e.g. by a KLD) we will nonetheless
284 	 * always perform the earlist-sorted sysinit at each step; using the
285 	 * STAILQ_FOREACH macro would result in items being skipped if inserted
286 	 * earlier than the "current item".
287 	 */
288 	while ((sip = STAILQ_FIRST(&sysinit_list)) != NULL) {
289 		STAILQ_REMOVE_HEAD(&sysinit_list, next);
290 		STAILQ_INSERT_TAIL(&sysinit_done_list, sip, next);
291 
292 		if (sip->subsystem == SI_SUB_DUMMY)
293 			continue;	/* skip dummy task(s)*/
294 
295 		if (sip->subsystem > last)
296 			BOOTTRACE_INIT("sysinit 0x%7x", sip->subsystem);
297 
298 #if defined(VERBOSE_SYSINIT)
299 		if (sip->subsystem > last && verbose_sysinit != 0) {
300 			verbose = 1;
301 			printf("subsystem %x\n", last);
302 		}
303 		if (verbose) {
304 #if defined(DDB)
305 			const char *func, *data;
306 
307 			func = symbol_name((vm_offset_t)sip->func,
308 			    DB_STGY_PROC);
309 			data = symbol_name((vm_offset_t)sip->udata,
310 			    DB_STGY_ANY);
311 			if (func != NULL && data != NULL)
312 				printf("   %s(&%s)... ", func, data);
313 			else if (func != NULL)
314 				printf("   %s(%p)... ", func, sip->udata);
315 			else
316 #endif
317 				printf("   %p(%p)... ", sip->func,
318 				    sip->udata);
319 		}
320 #endif
321 
322 		/* Call function */
323 		(*(sip->func))(sip->udata);
324 
325 #if defined(VERBOSE_SYSINIT)
326 		if (verbose)
327 			printf("done.\n");
328 #endif
329 
330 		/* Check off the one we're just done */
331 		last = sip->subsystem;
332 	}
333 
334 	TSEXIT();	/* Here so we don't overlap with start_init. */
335 	BOOTTRACE("mi_startup done");
336 
337 	mtx_assert(&Giant, MA_OWNED | MA_NOTRECURSED);
338 	mtx_unlock(&Giant);
339 
340 	/*
341 	 * We can't free our thread structure since it is statically allocated.
342 	 * Just sleep forever.  This thread could be repurposed for something if
343 	 * the need arises.
344 	 */
345 	for (;;)
346 		tsleep(__builtin_frame_address(0), PNOLOCK, "parked", 0);
347 }
348 
349 static void
350 print_caddr_t(const void *data)
351 {
352 	printf("%s", (const char *)data);
353 }
354 
355 static void
356 print_version(const void *data __unused)
357 {
358 	int len;
359 
360 	/* Strip a trailing newline from version. */
361 	len = strlen(version);
362 	while (len > 0 && version[len - 1] == '\n')
363 		len--;
364 	printf("%.*s %s\n", len, version, machine);
365 	printf("%s\n", compiler_version);
366 }
367 
368 C_SYSINIT(announce, SI_SUB_COPYRIGHT, SI_ORDER_FIRST, print_caddr_t,
369     copyright);
370 C_SYSINIT(trademark, SI_SUB_COPYRIGHT, SI_ORDER_SECOND, print_caddr_t,
371     trademark);
372 C_SYSINIT(version, SI_SUB_COPYRIGHT, SI_ORDER_THIRD, print_version, NULL);
373 
374 #ifdef WITNESS
375 static const char wit_warn[] =
376      "WARNING: WITNESS option enabled, expect reduced performance.\n";
377 C_SYSINIT(witwarn, SI_SUB_COPYRIGHT, SI_ORDER_FOURTH,
378    print_caddr_t, wit_warn);
379 C_SYSINIT(witwarn2, SI_SUB_LAST, SI_ORDER_FOURTH,
380    print_caddr_t, wit_warn);
381 #endif
382 
383 #ifdef DIAGNOSTIC
384 static const char diag_warn[] =
385      "WARNING: DIAGNOSTIC option enabled, expect reduced performance.\n";
386 C_SYSINIT(diagwarn, SI_SUB_COPYRIGHT, SI_ORDER_FIFTH,
387     print_caddr_t, diag_warn);
388 C_SYSINIT(diagwarn2, SI_SUB_LAST, SI_ORDER_FIFTH,
389     print_caddr_t, diag_warn);
390 #endif
391 
392 #if __SIZEOF_LONG__ == 4
393 static const char ilp32_warn[] =
394     "WARNING: 32-bit kernels are deprecated and may be removed in FreeBSD 15.0.\n";
395 C_SYSINIT(ilp32warn, SI_SUB_COPYRIGHT, SI_ORDER_FIFTH,
396     print_caddr_t, ilp32_warn);
397 C_SYSINIT(ilp32warn2, SI_SUB_LAST, SI_ORDER_FIFTH,
398     print_caddr_t, ilp32_warn);
399 #endif
400 
401 static int
402 null_fetch_syscall_args(struct thread *td __unused)
403 {
404 
405 	panic("null_fetch_syscall_args");
406 }
407 
408 static void
409 null_set_syscall_retval(struct thread *td __unused, int error __unused)
410 {
411 
412 	panic("null_set_syscall_retval");
413 }
414 
415 static void
416 null_set_fork_retval(struct thread *td __unused)
417 {
418 
419 }
420 
421 struct sysentvec null_sysvec = {
422 	.sv_size	= 0,
423 	.sv_table	= NULL,
424 	.sv_fixup	= NULL,
425 	.sv_sendsig	= NULL,
426 	.sv_sigcode	= NULL,
427 	.sv_szsigcode	= NULL,
428 	.sv_name	= "null",
429 	.sv_coredump	= NULL,
430 	.sv_minsigstksz	= 0,
431 	.sv_minuser	= VM_MIN_ADDRESS,
432 	.sv_maxuser	= VM_MAXUSER_ADDRESS,
433 	.sv_usrstack	= USRSTACK,
434 	.sv_psstrings	= PS_STRINGS,
435 	.sv_psstringssz	= sizeof(struct ps_strings),
436 	.sv_stackprot	= VM_PROT_ALL,
437 	.sv_copyout_strings	= NULL,
438 	.sv_setregs	= NULL,
439 	.sv_fixlimit	= NULL,
440 	.sv_maxssiz	= NULL,
441 	.sv_flags	= 0,
442 	.sv_set_syscall_retval = null_set_syscall_retval,
443 	.sv_fetch_syscall_args = null_fetch_syscall_args,
444 	.sv_syscallnames = NULL,
445 	.sv_schedtail	= NULL,
446 	.sv_thread_detach = NULL,
447 	.sv_trap	= NULL,
448 	.sv_set_fork_retval = null_set_fork_retval,
449 	.sv_regset_begin = NULL,
450 	.sv_regset_end  = NULL,
451 };
452 
453 /*
454  * The two following SYSINIT's are proc0 specific glue code.  I am not
455  * convinced that they can not be safely combined, but their order of
456  * operation has been maintained as the same as the original init_main.c
457  * for right now.
458  */
459 /* ARGSUSED*/
460 static void
461 proc0_init(void *dummy __unused)
462 {
463 	struct proc *p;
464 	struct thread *td;
465 	struct ucred *newcred;
466 	struct uidinfo tmpuinfo;
467 	struct loginclass tmplc = {
468 		.lc_name = "",
469 	};
470 	vm_paddr_t pageablemem;
471 	int i;
472 
473 	GIANT_REQUIRED;
474 	p = &proc0;
475 	td = &thread0;
476 
477 	/*
478 	 * Initialize magic number and osrel.
479 	 */
480 	p->p_magic = P_MAGIC;
481 	p->p_osrel = osreldate;
482 
483 	/*
484 	 * Initialize thread and process structures.
485 	 */
486 	procinit();	/* set up proc zone */
487 	threadinit();	/* set up UMA zones */
488 
489 	/*
490 	 * Initialise scheduler resources.
491 	 * Add scheduler specific parts to proc, thread as needed.
492 	 */
493 	schedinit();	/* scheduler gets its house in order */
494 
495 	/*
496 	 * Create process 0 (the swapper).
497 	 */
498 	LIST_INSERT_HEAD(&allproc, p, p_list);
499 	LIST_INSERT_HEAD(PIDHASH(0), p, p_hash);
500 	mtx_init(&pgrp0.pg_mtx, "process group", NULL, MTX_DEF | MTX_DUPOK);
501 	sx_init(&pgrp0.pg_killsx, "killpg racer");
502 	p->p_pgrp = &pgrp0;
503 	LIST_INSERT_HEAD(PGRPHASH(0), &pgrp0, pg_hash);
504 	LIST_INIT(&pgrp0.pg_members);
505 	LIST_INSERT_HEAD(&pgrp0.pg_members, p, p_pglist);
506 
507 	pgrp0.pg_session = &session0;
508 	mtx_init(&session0.s_mtx, "session", NULL, MTX_DEF);
509 	refcount_init(&session0.s_count, 1);
510 	session0.s_leader = p;
511 
512 	p->p_sysent = &null_sysvec;
513 	p->p_flag = P_SYSTEM | P_INMEM | P_KPROC;
514 	p->p_flag2 = 0;
515 	p->p_state = PRS_NORMAL;
516 	p->p_klist = knlist_alloc(&p->p_mtx);
517 	STAILQ_INIT(&p->p_ktr);
518 	p->p_nice = NZERO;
519 	td->td_tid = THREAD0_TID;
520 	tidhash_add(td);
521 	TD_SET_STATE(td, TDS_RUNNING);
522 	td->td_pri_class = PRI_TIMESHARE;
523 	td->td_user_pri = PUSER;
524 	td->td_base_user_pri = PUSER;
525 	td->td_lend_user_pri = PRI_MAX;
526 	td->td_priority = PVM;
527 	td->td_base_pri = PVM;
528 	td->td_oncpu = curcpu;
529 	td->td_flags = TDF_INMEM;
530 	td->td_pflags = TDP_KTHREAD;
531 	td->td_cpuset = cpuset_thread0();
532 	td->td_domain.dr_policy = td->td_cpuset->cs_domain;
533 	prison0_init();
534 	p->p_peers = 0;
535 	p->p_leader = p;
536 	p->p_reaper = p;
537 	p->p_treeflag |= P_TREE_REAPER;
538 	LIST_INIT(&p->p_reaplist);
539 
540 	strncpy(p->p_comm, "kernel", sizeof (p->p_comm));
541 	strncpy(td->td_name, "swapper", sizeof (td->td_name));
542 
543 	callout_init_mtx(&p->p_itcallout, &p->p_mtx, 0);
544 	callout_init_mtx(&p->p_limco, &p->p_mtx, 0);
545 	callout_init(&td->td_slpcallout, 1);
546 	TAILQ_INIT(&p->p_kqtim_stop);
547 
548 	/* Create credentials. */
549 	newcred = crget();
550 	newcred->cr_ngroups = 1;	/* group 0 */
551 	/* A hack to prevent uifind from tripping over NULL pointers. */
552 	curthread->td_ucred = newcred;
553 	tmpuinfo.ui_uid = 1;
554 	newcred->cr_uidinfo = newcred->cr_ruidinfo = &tmpuinfo;
555 	newcred->cr_uidinfo = uifind(0);
556 	newcred->cr_ruidinfo = uifind(0);
557 	newcred->cr_loginclass = &tmplc;
558 	newcred->cr_loginclass = loginclass_find("default");
559 	/* End hack. creds get properly set later with thread_cow_get_proc */
560 	curthread->td_ucred = NULL;
561 	newcred->cr_prison = &prison0;
562 	newcred->cr_users++; /* avoid assertion failure */
563 	p->p_ucred = crcowget(newcred);
564 	newcred->cr_users--;
565 	crfree(newcred);
566 #ifdef AUDIT
567 	audit_cred_kproc0(newcred);
568 #endif
569 #ifdef MAC
570 	mac_cred_create_swapper(newcred);
571 #endif
572 	/* Create sigacts. */
573 	p->p_sigacts = sigacts_alloc();
574 
575 	/* Initialize signal state for process 0. */
576 	siginit(&proc0);
577 
578 	/* Create the file descriptor table. */
579 	p->p_pd = pdinit(NULL, false);
580 	p->p_fd = fdinit();
581 	p->p_fdtol = NULL;
582 
583 	/* Create the limits structures. */
584 	p->p_limit = lim_alloc();
585 	for (i = 0; i < RLIM_NLIMITS; i++)
586 		p->p_limit->pl_rlimit[i].rlim_cur =
587 		    p->p_limit->pl_rlimit[i].rlim_max = RLIM_INFINITY;
588 	p->p_limit->pl_rlimit[RLIMIT_NOFILE].rlim_cur =
589 	    p->p_limit->pl_rlimit[RLIMIT_NOFILE].rlim_max = maxfiles;
590 	p->p_limit->pl_rlimit[RLIMIT_NPROC].rlim_cur =
591 	    p->p_limit->pl_rlimit[RLIMIT_NPROC].rlim_max = maxproc;
592 	p->p_limit->pl_rlimit[RLIMIT_DATA].rlim_cur = dfldsiz;
593 	p->p_limit->pl_rlimit[RLIMIT_DATA].rlim_max = maxdsiz;
594 	p->p_limit->pl_rlimit[RLIMIT_STACK].rlim_cur = dflssiz;
595 	p->p_limit->pl_rlimit[RLIMIT_STACK].rlim_max = maxssiz;
596 	/* Cast to avoid overflow on i386/PAE. */
597 	pageablemem = ptoa((vm_paddr_t)vm_free_count());
598 	p->p_limit->pl_rlimit[RLIMIT_RSS].rlim_cur =
599 	    p->p_limit->pl_rlimit[RLIMIT_RSS].rlim_max = pageablemem;
600 	p->p_limit->pl_rlimit[RLIMIT_MEMLOCK].rlim_cur = pageablemem / 3;
601 	p->p_limit->pl_rlimit[RLIMIT_MEMLOCK].rlim_max = pageablemem;
602 	p->p_cpulimit = RLIM_INFINITY;
603 
604 	PROC_LOCK(p);
605 	thread_cow_get_proc(td, p);
606 	PROC_UNLOCK(p);
607 
608 	/* Initialize resource accounting structures. */
609 	racct_create(&p->p_racct);
610 
611 	p->p_stats = pstats_alloc();
612 
613 	/* Allocate a prototype map so we have something to fork. */
614 	p->p_vmspace = &vmspace0;
615 	refcount_init(&vmspace0.vm_refcnt, 1);
616 	pmap_pinit0(vmspace_pmap(&vmspace0));
617 
618 	/*
619 	 * proc0 is not expected to enter usermode, so there is no special
620 	 * handling for sv_minuser here, like is done for exec_new_vmspace().
621 	 */
622 	vm_map_init(&vmspace0.vm_map, vmspace_pmap(&vmspace0),
623 	    p->p_sysent->sv_minuser, p->p_sysent->sv_maxuser);
624 
625 	/*
626 	 * Call the init and ctor for the new thread and proc.  We wait
627 	 * to do this until all other structures are fairly sane.
628 	 */
629 	EVENTHANDLER_DIRECT_INVOKE(process_init, p);
630 	EVENTHANDLER_DIRECT_INVOKE(thread_init, td);
631 #ifdef KDTRACE_HOOKS
632 	kdtrace_proc_ctor(p);
633 	kdtrace_thread_ctor(td);
634 #endif
635 	EVENTHANDLER_DIRECT_INVOKE(process_ctor, p);
636 	EVENTHANDLER_DIRECT_INVOKE(thread_ctor, td);
637 
638 	/*
639 	 * Charge root for one process.
640 	 */
641 	(void)chgproccnt(p->p_ucred->cr_ruidinfo, 1, 0);
642 	PROC_LOCK(p);
643 	racct_add_force(p, RACCT_NPROC, 1);
644 	PROC_UNLOCK(p);
645 }
646 SYSINIT(p0init, SI_SUB_INTRINSIC, SI_ORDER_FIRST, proc0_init, NULL);
647 
648 /* ARGSUSED*/
649 static void
650 proc0_post(void *dummy __unused)
651 {
652 	struct proc *p;
653 	struct rusage ru;
654 	struct thread *td;
655 
656 	/*
657 	 * Now we can look at the time, having had a chance to verify the
658 	 * time from the filesystem.  Pretend that proc0 started now.
659 	 */
660 	sx_slock(&allproc_lock);
661 	FOREACH_PROC_IN_SYSTEM(p) {
662 		PROC_LOCK(p);
663 		if (p->p_state == PRS_NEW) {
664 			PROC_UNLOCK(p);
665 			continue;
666 		}
667 		microuptime(&p->p_stats->p_start);
668 		PROC_STATLOCK(p);
669 		rufetch(p, &ru);	/* Clears thread stats */
670 		p->p_rux.rux_runtime = 0;
671 		p->p_rux.rux_uticks = 0;
672 		p->p_rux.rux_sticks = 0;
673 		p->p_rux.rux_iticks = 0;
674 		PROC_STATUNLOCK(p);
675 		FOREACH_THREAD_IN_PROC(p, td) {
676 			td->td_runtime = 0;
677 		}
678 		PROC_UNLOCK(p);
679 	}
680 	sx_sunlock(&allproc_lock);
681 	PCPU_SET(switchtime, cpu_ticks());
682 	PCPU_SET(switchticks, ticks);
683 }
684 SYSINIT(p0post, SI_SUB_INTRINSIC_POST, SI_ORDER_FIRST, proc0_post, NULL);
685 
686 /*
687  ***************************************************************************
688  ****
689  **** The following SYSINIT's and glue code should be moved to the
690  **** respective files on a per subsystem basis.
691  ****
692  ***************************************************************************
693  */
694 
695 /*
696  * List of paths to try when searching for "init".
697  */
698 static char init_path[MAXPATHLEN] =
699 #ifdef	INIT_PATH
700     __XSTRING(INIT_PATH);
701 #else
702     "/sbin/init:/sbin/oinit:/sbin/init.bak:/rescue/init";
703 #endif
704 SYSCTL_STRING(_kern, OID_AUTO, init_path, CTLFLAG_RD, init_path, 0,
705 	"Path used to search the init process");
706 
707 /*
708  * Shutdown timeout of init(8).
709  * Unused within kernel, but used to control init(8), hence do not remove.
710  */
711 #ifndef INIT_SHUTDOWN_TIMEOUT
712 #define INIT_SHUTDOWN_TIMEOUT 120
713 #endif
714 static int init_shutdown_timeout = INIT_SHUTDOWN_TIMEOUT;
715 SYSCTL_INT(_kern, OID_AUTO, init_shutdown_timeout,
716 	CTLFLAG_RW, &init_shutdown_timeout, 0, "Shutdown timeout of init(8). "
717 	"Unused within kernel, but used to control init(8)");
718 
719 /*
720  * Start the initial user process; try exec'ing each pathname in init_path.
721  * The program is invoked with one argument containing the boot flags.
722  */
723 static void
724 start_init(void *dummy)
725 {
726 	struct image_args args;
727 	int error;
728 	char *var, *path;
729 	char *free_init_path, *tmp_init_path;
730 	struct thread *td;
731 	struct proc *p;
732 	struct vmspace *oldvmspace;
733 
734 	TSENTER();	/* Here so we don't overlap with mi_startup. */
735 
736 	td = curthread;
737 	p = td->td_proc;
738 
739 	vfs_mountroot();
740 
741 	/* Wipe GELI passphrase from the environment. */
742 	kern_unsetenv("kern.geom.eli.passphrase");
743 
744 	/* For Multicons, report which console is primary to both */
745 	if (boothowto & RB_MULTIPLE) {
746 		if (boothowto & RB_SERIAL)
747 			printf("Dual Console: Serial Primary, Video Secondary\n");
748 		else
749 			printf("Dual Console: Video Primary, Serial Secondary\n");
750 	}
751 
752 	if ((var = kern_getenv("init_path")) != NULL) {
753 		strlcpy(init_path, var, sizeof(init_path));
754 		freeenv(var);
755 	}
756 	free_init_path = tmp_init_path = strdup(init_path, M_TEMP);
757 
758 	while ((path = strsep(&tmp_init_path, ":")) != NULL) {
759 		if (bootverbose)
760 			printf("start_init: trying %s\n", path);
761 
762 		memset(&args, 0, sizeof(args));
763 		error = exec_alloc_args(&args);
764 		if (error != 0)
765 			panic("%s: Can't allocate space for init arguments %d",
766 			    __func__, error);
767 
768 		error = exec_args_add_fname(&args, path, UIO_SYSSPACE);
769 		if (error != 0)
770 			panic("%s: Can't add fname %d", __func__, error);
771 		error = exec_args_add_arg(&args, path, UIO_SYSSPACE);
772 		if (error != 0)
773 			panic("%s: Can't add argv[0] %d", __func__, error);
774 		if (boothowto & RB_SINGLE)
775 			error = exec_args_add_arg(&args, "-s", UIO_SYSSPACE);
776 		if (error != 0)
777 			panic("%s: Can't add argv[0] %d", __func__, error);
778 
779 		/*
780 		 * Now try to exec the program.  If can't for any reason
781 		 * other than it doesn't exist, complain.
782 		 *
783 		 * Otherwise, return via fork_trampoline() all the way
784 		 * to user mode as init!
785 		 */
786 		KASSERT((td->td_pflags & TDP_EXECVMSPC) == 0,
787 		    ("nested execve"));
788 		memset(td->td_frame, 0, sizeof(*td->td_frame));
789 		oldvmspace = p->p_vmspace;
790 		error = kern_execve(td, &args, NULL, oldvmspace);
791 		KASSERT(error != 0,
792 		    ("kern_execve returned success, not EJUSTRETURN"));
793 		if (error == EJUSTRETURN) {
794 			exec_cleanup(td, oldvmspace);
795 			free(free_init_path, M_TEMP);
796 			TSEXIT();
797 			return;
798 		}
799 		if (error != ENOENT)
800 			printf("exec %s: error %d\n", path, error);
801 	}
802 	free(free_init_path, M_TEMP);
803 	printf("init: not found in path %s\n", init_path);
804 	panic("no init");
805 }
806 
807 /*
808  * Like kproc_create(), but runs in its own address space.  We do this
809  * early to reserve pid 1.  Note special case - do not make it
810  * runnable yet, init execution is started when userspace can be served.
811  */
812 static void
813 create_init(const void *udata __unused)
814 {
815 	struct fork_req fr;
816 	struct ucred *newcred, *oldcred;
817 	struct thread *td;
818 	int error;
819 
820 	bzero(&fr, sizeof(fr));
821 	fr.fr_flags = RFFDG | RFPROC | RFSTOPPED;
822 	fr.fr_procp = &initproc;
823 	error = fork1(&thread0, &fr);
824 	if (error)
825 		panic("cannot fork init: %d\n", error);
826 	KASSERT(initproc->p_pid == 1, ("create_init: initproc->p_pid != 1"));
827 	/* divorce init's credentials from the kernel's */
828 	newcred = crget();
829 	sx_xlock(&proctree_lock);
830 	PROC_LOCK(initproc);
831 	initproc->p_flag |= P_SYSTEM | P_INMEM;
832 	initproc->p_treeflag |= P_TREE_REAPER;
833 	oldcred = initproc->p_ucred;
834 	crcopy(newcred, oldcred);
835 #ifdef MAC
836 	mac_cred_create_init(newcred);
837 #endif
838 #ifdef AUDIT
839 	audit_cred_proc1(newcred);
840 #endif
841 	proc_set_cred(initproc, newcred);
842 	td = FIRST_THREAD_IN_PROC(initproc);
843 	crcowfree(td);
844 	td->td_realucred = crcowget(initproc->p_ucred);
845 	td->td_ucred = td->td_realucred;
846 	PROC_UNLOCK(initproc);
847 	sx_xunlock(&proctree_lock);
848 	crfree(oldcred);
849 	cpu_fork_kthread_handler(FIRST_THREAD_IN_PROC(initproc),
850 	    start_init, NULL);
851 }
852 SYSINIT(init, SI_SUB_CREATE_INIT, SI_ORDER_FIRST, create_init, NULL);
853 
854 /*
855  * Make it runnable now.
856  */
857 static void
858 kick_init(const void *udata __unused)
859 {
860 	struct thread *td;
861 
862 	td = FIRST_THREAD_IN_PROC(initproc);
863 	thread_lock(td);
864 	TD_SET_CAN_RUN(td);
865 	sched_add(td, SRQ_BORING);
866 }
867 SYSINIT(kickinit, SI_SUB_KTHREAD_INIT, SI_ORDER_MIDDLE, kick_init, NULL);
868 
869 /*
870  * DDB(4).
871  */
872 #ifdef DDB
873 static void
874 db_show_print_syinit(struct sysinit *sip, bool ddb)
875 {
876 	const char *sname, *funcname;
877 	c_db_sym_t sym;
878 	db_expr_t  offset;
879 
880 #define xprint(...)							\
881 	if (ddb)							\
882 		db_printf(__VA_ARGS__);					\
883 	else								\
884 		printf(__VA_ARGS__)
885 
886 	if (sip == NULL) {
887 		xprint("%s: no sysinit * given\n", __func__);
888 		return;
889 	}
890 
891 	sym = db_search_symbol((vm_offset_t)sip, DB_STGY_ANY, &offset);
892 	db_symbol_values(sym, &sname, NULL);
893 	sym = db_search_symbol((vm_offset_t)sip->func, DB_STGY_PROC, &offset);
894 	db_symbol_values(sym, &funcname, NULL);
895 	xprint("%s(%p)\n", (sname != NULL) ? sname : "", sip);
896 	xprint("  %#08x %#08x\n", sip->subsystem, sip->order);
897 	xprint("  %p(%s)(%p)\n",
898 	    sip->func, (funcname != NULL) ? funcname : "", sip->udata);
899 #undef xprint
900 }
901 
902 DB_SHOW_COMMAND_FLAGS(sysinit, db_show_sysinit, DB_CMD_MEMSAFE)
903 {
904 	struct sysinit *sip;
905 
906 	db_printf("SYSINIT vs Name(Ptr)\n");
907 	db_printf("  Subsystem  Order\n");
908 	db_printf("  Function(Name)(Arg)\n");
909 	STAILQ_FOREACH(sip, &sysinit_done_list, next) {
910 		db_show_print_syinit(sip, true);
911 		if (db_pager_quit)
912 			return;
913 	}
914 	STAILQ_FOREACH(sip, &sysinit_list, next) {
915 		db_show_print_syinit(sip, true);
916 		if (db_pager_quit)
917 			break;
918 	}
919 }
920 #endif /* DDB */
921