xref: /freebsd/sys/kern/init_main.c (revision d0b2dbfa0ecf2bbc9709efc5e20baf8e4b44bbbf)
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/sched.h>
77 #include <sys/signalvar.h>
78 #include <sys/sx.h>
79 #include <sys/syscallsubr.h>
80 #include <sys/sysctl.h>
81 #include <sys/sysent.h>
82 #include <sys/sysproto.h>
83 #include <sys/unistd.h>
84 #include <sys/vmmeter.h>
85 #include <sys/vnode.h>
86 
87 #include <machine/cpu.h>
88 
89 #include <security/audit/audit.h>
90 #include <security/mac/mac_framework.h>
91 
92 #include <vm/vm.h>
93 #include <vm/vm_param.h>
94 #include <vm/vm_extern.h>
95 #include <vm/pmap.h>
96 #include <vm/vm_map.h>
97 #include <sys/copyright.h>
98 
99 #include <ddb/ddb.h>
100 #include <ddb/db_sym.h>
101 
102 void mi_startup(void);				/* Should be elsewhere */
103 
104 /* Components of the first process -- never freed. */
105 static struct session session0;
106 static struct pgrp pgrp0;
107 struct	proc proc0;
108 struct thread0_storage thread0_st __aligned(32);
109 struct	vmspace vmspace0;
110 struct	proc *initproc;
111 
112 int
113 linux_alloc_current_noop(struct thread *td __unused, int flags __unused)
114 {
115 	return (0);
116 }
117 int (*lkpi_alloc_current)(struct thread *, int) = linux_alloc_current_noop;
118 
119 #ifndef BOOTHOWTO
120 #define	BOOTHOWTO	0
121 #endif
122 int	boothowto = BOOTHOWTO;	/* initialized so that it can be patched */
123 SYSCTL_INT(_debug, OID_AUTO, boothowto, CTLFLAG_RD, &boothowto, 0,
124 	"Boot control flags, passed from loader");
125 
126 #ifndef BOOTVERBOSE
127 #define	BOOTVERBOSE	0
128 #endif
129 int	bootverbose = BOOTVERBOSE;
130 SYSCTL_INT(_debug, OID_AUTO, bootverbose, CTLFLAG_RW, &bootverbose, 0,
131 	"Control the output of verbose kernel messages");
132 
133 #ifdef VERBOSE_SYSINIT
134 /*
135  * We'll use the defined value of VERBOSE_SYSINIT from the kernel config to
136  * dictate the default VERBOSE_SYSINIT behavior.  Significant values for this
137  * option and associated tunable are:
138  * - 0, 'compiled in but silent by default'
139  * - 1, 'compiled in but verbose by default' (default)
140  */
141 int	verbose_sysinit = VERBOSE_SYSINIT;
142 TUNABLE_INT("debug.verbose_sysinit", &verbose_sysinit);
143 #endif
144 
145 #ifdef INVARIANTS
146 FEATURE(invariants, "Kernel compiled with INVARIANTS, may affect performance");
147 #endif
148 
149 /*
150  * This ensures that there is at least one entry so that the sysinit_set
151  * symbol is not undefined.  A sybsystem ID of SI_SUB_DUMMY is never
152  * executed.
153  */
154 SYSINIT(placeholder, SI_SUB_DUMMY, SI_ORDER_ANY, NULL, NULL);
155 
156 /*
157  * The sysinit table itself.  Items are checked off as the are run.
158  * If we want to register new sysinit types, add them to newsysinit.
159  */
160 SET_DECLARE(sysinit_set, struct sysinit);
161 struct sysinit **sysinit, **sysinit_end;
162 struct sysinit **newsysinit, **newsysinit_end;
163 
164 /*
165  * Merge a new sysinit set into the current set, reallocating it if
166  * necessary.  This can only be called after malloc is running.
167  */
168 void
169 sysinit_add(struct sysinit **set, struct sysinit **set_end)
170 {
171 	struct sysinit **newset;
172 	struct sysinit **sipp;
173 	struct sysinit **xipp;
174 	int count;
175 
176 	count = set_end - set;
177 	if (newsysinit)
178 		count += newsysinit_end - newsysinit;
179 	else
180 		count += sysinit_end - sysinit;
181 	newset = malloc(count * sizeof(*sipp), M_TEMP, M_NOWAIT);
182 	if (newset == NULL)
183 		panic("cannot malloc for sysinit");
184 	xipp = newset;
185 	if (newsysinit)
186 		for (sipp = newsysinit; sipp < newsysinit_end; sipp++)
187 			*xipp++ = *sipp;
188 	else
189 		for (sipp = sysinit; sipp < sysinit_end; sipp++)
190 			*xipp++ = *sipp;
191 	for (sipp = set; sipp < set_end; sipp++)
192 		*xipp++ = *sipp;
193 	if (newsysinit)
194 		free(newsysinit, M_TEMP);
195 	newsysinit = newset;
196 	newsysinit_end = newset + count;
197 }
198 
199 #if defined (DDB) && defined(VERBOSE_SYSINIT)
200 static const char *
201 symbol_name(vm_offset_t va, db_strategy_t strategy)
202 {
203 	const char *name;
204 	c_db_sym_t sym;
205 	db_expr_t  offset;
206 
207 	if (va == 0)
208 		return (NULL);
209 	sym = db_search_symbol(va, strategy, &offset);
210 	if (offset != 0)
211 		return (NULL);
212 	db_symbol_values(sym, &name, NULL);
213 	return (name);
214 }
215 #endif
216 
217 /*
218  * System startup; initialize the world, create process 0, mount root
219  * filesystem, and fork to create init and pagedaemon.  Most of the
220  * hard work is done in the lower-level initialization routines including
221  * startup(), which does memory initialization and autoconfiguration.
222  *
223  * This allows simple addition of new kernel subsystems that require
224  * boot time initialization.  It also allows substitution of subsystem
225  * (for instance, a scheduler, kernel profiler, or VM system) by object
226  * module.  Finally, it allows for optional "kernel threads".
227  */
228 void
229 mi_startup(void)
230 {
231 
232 	struct sysinit **sipp;	/* system initialization*/
233 	struct sysinit **xipp;	/* interior loop of sort*/
234 	struct sysinit *save;	/* bubble*/
235 
236 	int last;
237 #if defined(VERBOSE_SYSINIT)
238 	int verbose;
239 #endif
240 
241 	TSENTER();
242 
243 	if (boothowto & RB_VERBOSE)
244 		bootverbose++;
245 
246 	if (sysinit == NULL) {
247 		sysinit = SET_BEGIN(sysinit_set);
248 		sysinit_end = SET_LIMIT(sysinit_set);
249 	}
250 
251 restart:
252 	/*
253 	 * Perform a bubble sort of the system initialization objects by
254 	 * their subsystem (primary key) and order (secondary key).
255 	 */
256 	TSENTER2("bubblesort");
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 	TSEXIT2("bubblesort");
269 
270 	last = SI_SUB_COPYRIGHT;
271 #if defined(VERBOSE_SYSINIT)
272 	verbose = 0;
273 #if !defined(DDB)
274 	printf("VERBOSE_SYSINIT: DDB not enabled, symbol lookups disabled.\n");
275 #endif
276 #endif
277 
278 	/*
279 	 * Traverse the (now) ordered list of system initialization tasks.
280 	 * Perform each task, and continue on to the next task.
281 	 */
282 	for (sipp = sysinit; sipp < sysinit_end; sipp++) {
283 		if ((*sipp)->subsystem == SI_SUB_DUMMY)
284 			continue;	/* skip dummy task(s)*/
285 
286 		if ((*sipp)->subsystem == SI_SUB_DONE)
287 			continue;
288 
289 		if ((*sipp)->subsystem > last)
290 			BOOTTRACE_INIT("sysinit 0x%7x", (*sipp)->subsystem);
291 
292 #if defined(VERBOSE_SYSINIT)
293 		if ((*sipp)->subsystem > last && verbose_sysinit != 0) {
294 			verbose = 1;
295 			printf("subsystem %x\n", last);
296 		}
297 		if (verbose) {
298 #if defined(DDB)
299 			const char *func, *data;
300 
301 			func = symbol_name((vm_offset_t)(*sipp)->func,
302 			    DB_STGY_PROC);
303 			data = symbol_name((vm_offset_t)(*sipp)->udata,
304 			    DB_STGY_ANY);
305 			if (func != NULL && data != NULL)
306 				printf("   %s(&%s)... ", func, data);
307 			else if (func != NULL)
308 				printf("   %s(%p)... ", func, (*sipp)->udata);
309 			else
310 #endif
311 				printf("   %p(%p)... ", (*sipp)->func,
312 				    (*sipp)->udata);
313 		}
314 #endif
315 
316 		/* Call function */
317 		(*((*sipp)->func))((*sipp)->udata);
318 
319 #if defined(VERBOSE_SYSINIT)
320 		if (verbose)
321 			printf("done.\n");
322 #endif
323 
324 		/* Check off the one we're just done */
325 		last = (*sipp)->subsystem;
326 		(*sipp)->subsystem = SI_SUB_DONE;
327 
328 		/* Check if we've installed more sysinit items via KLD */
329 		if (newsysinit != NULL) {
330 			if (sysinit != SET_BEGIN(sysinit_set))
331 				free(sysinit, M_TEMP);
332 			sysinit = newsysinit;
333 			sysinit_end = newsysinit_end;
334 			newsysinit = NULL;
335 			newsysinit_end = NULL;
336 			goto restart;
337 		}
338 	}
339 
340 	TSEXIT();	/* Here so we don't overlap with start_init. */
341 	BOOTTRACE("mi_startup done");
342 
343 	mtx_assert(&Giant, MA_OWNED | MA_NOTRECURSED);
344 	mtx_unlock(&Giant);
345 
346 	/*
347 	 * Now hand over this thread to swapper.
348 	 */
349 	swapper();
350 	/* NOTREACHED*/
351 }
352 
353 static void
354 print_caddr_t(void *data)
355 {
356 	printf("%s", (char *)data);
357 }
358 
359 static void
360 print_version(void *data __unused)
361 {
362 	int len;
363 
364 	/* Strip a trailing newline from version. */
365 	len = strlen(version);
366 	while (len > 0 && version[len - 1] == '\n')
367 		len--;
368 	printf("%.*s %s\n", len, version, machine);
369 	printf("%s\n", compiler_version);
370 }
371 
372 SYSINIT(announce, SI_SUB_COPYRIGHT, SI_ORDER_FIRST, print_caddr_t,
373     copyright);
374 SYSINIT(trademark, SI_SUB_COPYRIGHT, SI_ORDER_SECOND, print_caddr_t,
375     trademark);
376 SYSINIT(version, SI_SUB_COPYRIGHT, SI_ORDER_THIRD, print_version, NULL);
377 
378 #ifdef WITNESS
379 static char wit_warn[] =
380      "WARNING: WITNESS option enabled, expect reduced performance.\n";
381 SYSINIT(witwarn, SI_SUB_COPYRIGHT, SI_ORDER_FOURTH,
382    print_caddr_t, wit_warn);
383 SYSINIT(witwarn2, SI_SUB_LAST, SI_ORDER_FOURTH,
384    print_caddr_t, wit_warn);
385 #endif
386 
387 #ifdef DIAGNOSTIC
388 static char diag_warn[] =
389      "WARNING: DIAGNOSTIC option enabled, expect reduced performance.\n";
390 SYSINIT(diagwarn, SI_SUB_COPYRIGHT, SI_ORDER_FIFTH,
391     print_caddr_t, diag_warn);
392 SYSINIT(diagwarn2, SI_SUB_LAST, SI_ORDER_FIFTH,
393     print_caddr_t, diag_warn);
394 #endif
395 
396 #if __SIZEOF_LONG__ == 4
397 static char ilp32_warn[] =
398     "WARNING: 32-bit kernels are deprecated and may be removed in FreeBSD 15.0.\n";
399 SYSINIT(ilp32warn, SI_SUB_COPYRIGHT, SI_ORDER_FIFTH,
400     print_caddr_t, ilp32_warn);
401 SYSINIT(ilp32warn2, SI_SUB_LAST, SI_ORDER_FIFTH,
402     print_caddr_t, ilp32_warn);
403 #endif
404 
405 static int
406 null_fetch_syscall_args(struct thread *td __unused)
407 {
408 
409 	panic("null_fetch_syscall_args");
410 }
411 
412 static void
413 null_set_syscall_retval(struct thread *td __unused, int error __unused)
414 {
415 
416 	panic("null_set_syscall_retval");
417 }
418 
419 static void
420 null_set_fork_retval(struct thread *td __unused)
421 {
422 
423 }
424 
425 struct sysentvec null_sysvec = {
426 	.sv_size	= 0,
427 	.sv_table	= NULL,
428 	.sv_fixup	= NULL,
429 	.sv_sendsig	= NULL,
430 	.sv_sigcode	= NULL,
431 	.sv_szsigcode	= NULL,
432 	.sv_name	= "null",
433 	.sv_coredump	= NULL,
434 	.sv_minsigstksz	= 0,
435 	.sv_minuser	= VM_MIN_ADDRESS,
436 	.sv_maxuser	= VM_MAXUSER_ADDRESS,
437 	.sv_usrstack	= USRSTACK,
438 	.sv_psstrings	= PS_STRINGS,
439 	.sv_psstringssz	= sizeof(struct ps_strings),
440 	.sv_stackprot	= VM_PROT_ALL,
441 	.sv_copyout_strings	= NULL,
442 	.sv_setregs	= NULL,
443 	.sv_fixlimit	= NULL,
444 	.sv_maxssiz	= NULL,
445 	.sv_flags	= 0,
446 	.sv_set_syscall_retval = null_set_syscall_retval,
447 	.sv_fetch_syscall_args = null_fetch_syscall_args,
448 	.sv_syscallnames = NULL,
449 	.sv_schedtail	= NULL,
450 	.sv_thread_detach = NULL,
451 	.sv_trap	= NULL,
452 	.sv_set_fork_retval = null_set_fork_retval,
453 	.sv_regset_begin = NULL,
454 	.sv_regset_end  = NULL,
455 };
456 
457 /*
458  * The two following SYSINIT's are proc0 specific glue code.  I am not
459  * convinced that they can not be safely combined, but their order of
460  * operation has been maintained as the same as the original init_main.c
461  * for right now.
462  */
463 /* ARGSUSED*/
464 static void
465 proc0_init(void *dummy __unused)
466 {
467 	struct proc *p;
468 	struct thread *td;
469 	struct ucred *newcred;
470 	struct uidinfo tmpuinfo;
471 	struct loginclass tmplc = {
472 		.lc_name = "",
473 	};
474 	vm_paddr_t pageablemem;
475 	int i;
476 
477 	GIANT_REQUIRED;
478 	p = &proc0;
479 	td = &thread0;
480 
481 	/*
482 	 * Initialize magic number and osrel.
483 	 */
484 	p->p_magic = P_MAGIC;
485 	p->p_osrel = osreldate;
486 
487 	/*
488 	 * Initialize thread and process structures.
489 	 */
490 	procinit();	/* set up proc zone */
491 	threadinit();	/* set up UMA zones */
492 
493 	/*
494 	 * Initialise scheduler resources.
495 	 * Add scheduler specific parts to proc, thread as needed.
496 	 */
497 	schedinit();	/* scheduler gets its house in order */
498 
499 	/*
500 	 * Create process 0 (the swapper).
501 	 */
502 	LIST_INSERT_HEAD(&allproc, p, p_list);
503 	LIST_INSERT_HEAD(PIDHASH(0), p, p_hash);
504 	mtx_init(&pgrp0.pg_mtx, "process group", NULL, MTX_DEF | MTX_DUPOK);
505 	sx_init(&pgrp0.pg_killsx, "killpg racer");
506 	p->p_pgrp = &pgrp0;
507 	LIST_INSERT_HEAD(PGRPHASH(0), &pgrp0, pg_hash);
508 	LIST_INIT(&pgrp0.pg_members);
509 	LIST_INSERT_HEAD(&pgrp0.pg_members, p, p_pglist);
510 
511 	pgrp0.pg_session = &session0;
512 	mtx_init(&session0.s_mtx, "session", NULL, MTX_DEF);
513 	refcount_init(&session0.s_count, 1);
514 	session0.s_leader = p;
515 
516 	p->p_sysent = &null_sysvec;
517 	p->p_flag = P_SYSTEM | P_INMEM | P_KPROC;
518 	p->p_flag2 = 0;
519 	p->p_state = PRS_NORMAL;
520 	p->p_klist = knlist_alloc(&p->p_mtx);
521 	STAILQ_INIT(&p->p_ktr);
522 	p->p_nice = NZERO;
523 	td->td_tid = THREAD0_TID;
524 	tidhash_add(td);
525 	TD_SET_STATE(td, TDS_RUNNING);
526 	td->td_pri_class = PRI_TIMESHARE;
527 	td->td_user_pri = PUSER;
528 	td->td_base_user_pri = PUSER;
529 	td->td_lend_user_pri = PRI_MAX;
530 	td->td_priority = PVM;
531 	td->td_base_pri = PVM;
532 	td->td_oncpu = curcpu;
533 	td->td_flags = TDF_INMEM;
534 	td->td_pflags = TDP_KTHREAD;
535 	td->td_cpuset = cpuset_thread0();
536 	td->td_domain.dr_policy = td->td_cpuset->cs_domain;
537 	prison0_init();
538 	p->p_peers = 0;
539 	p->p_leader = p;
540 	p->p_reaper = p;
541 	p->p_treeflag |= P_TREE_REAPER;
542 	LIST_INIT(&p->p_reaplist);
543 
544 	strncpy(p->p_comm, "kernel", sizeof (p->p_comm));
545 	strncpy(td->td_name, "swapper", sizeof (td->td_name));
546 
547 	callout_init_mtx(&p->p_itcallout, &p->p_mtx, 0);
548 	callout_init_mtx(&p->p_limco, &p->p_mtx, 0);
549 	callout_init(&td->td_slpcallout, 1);
550 	TAILQ_INIT(&p->p_kqtim_stop);
551 
552 	/* Create credentials. */
553 	newcred = crget();
554 	newcred->cr_ngroups = 1;	/* group 0 */
555 	/* A hack to prevent uifind from tripping over NULL pointers. */
556 	curthread->td_ucred = newcred;
557 	tmpuinfo.ui_uid = 1;
558 	newcred->cr_uidinfo = newcred->cr_ruidinfo = &tmpuinfo;
559 	newcred->cr_uidinfo = uifind(0);
560 	newcred->cr_ruidinfo = uifind(0);
561 	newcred->cr_loginclass = &tmplc;
562 	newcred->cr_loginclass = loginclass_find("default");
563 	/* End hack. creds get properly set later with thread_cow_get_proc */
564 	curthread->td_ucred = NULL;
565 	newcred->cr_prison = &prison0;
566 	newcred->cr_users++; /* avoid assertion failure */
567 	proc_set_cred_init(p, newcred);
568 	newcred->cr_users--;
569 	crfree(newcred);
570 #ifdef AUDIT
571 	audit_cred_kproc0(newcred);
572 #endif
573 #ifdef MAC
574 	mac_cred_create_swapper(newcred);
575 #endif
576 	/* Create sigacts. */
577 	p->p_sigacts = sigacts_alloc();
578 
579 	/* Initialize signal state for process 0. */
580 	siginit(&proc0);
581 
582 	/* Create the file descriptor table. */
583 	p->p_pd = pdinit(NULL, false);
584 	p->p_fd = fdinit();
585 	p->p_fdtol = NULL;
586 
587 	/* Create the limits structures. */
588 	p->p_limit = lim_alloc();
589 	for (i = 0; i < RLIM_NLIMITS; i++)
590 		p->p_limit->pl_rlimit[i].rlim_cur =
591 		    p->p_limit->pl_rlimit[i].rlim_max = RLIM_INFINITY;
592 	p->p_limit->pl_rlimit[RLIMIT_NOFILE].rlim_cur =
593 	    p->p_limit->pl_rlimit[RLIMIT_NOFILE].rlim_max = maxfiles;
594 	p->p_limit->pl_rlimit[RLIMIT_NPROC].rlim_cur =
595 	    p->p_limit->pl_rlimit[RLIMIT_NPROC].rlim_max = maxproc;
596 	p->p_limit->pl_rlimit[RLIMIT_DATA].rlim_cur = dfldsiz;
597 	p->p_limit->pl_rlimit[RLIMIT_DATA].rlim_max = maxdsiz;
598 	p->p_limit->pl_rlimit[RLIMIT_STACK].rlim_cur = dflssiz;
599 	p->p_limit->pl_rlimit[RLIMIT_STACK].rlim_max = maxssiz;
600 	/* Cast to avoid overflow on i386/PAE. */
601 	pageablemem = ptoa((vm_paddr_t)vm_free_count());
602 	p->p_limit->pl_rlimit[RLIMIT_RSS].rlim_cur =
603 	    p->p_limit->pl_rlimit[RLIMIT_RSS].rlim_max = pageablemem;
604 	p->p_limit->pl_rlimit[RLIMIT_MEMLOCK].rlim_cur = pageablemem / 3;
605 	p->p_limit->pl_rlimit[RLIMIT_MEMLOCK].rlim_max = pageablemem;
606 	p->p_cpulimit = RLIM_INFINITY;
607 
608 	PROC_LOCK(p);
609 	thread_cow_get_proc(td, p);
610 	PROC_UNLOCK(p);
611 
612 	/* Initialize resource accounting structures. */
613 	racct_create(&p->p_racct);
614 
615 	p->p_stats = pstats_alloc();
616 
617 	/* Allocate a prototype map so we have something to fork. */
618 	p->p_vmspace = &vmspace0;
619 	refcount_init(&vmspace0.vm_refcnt, 1);
620 	pmap_pinit0(vmspace_pmap(&vmspace0));
621 
622 	/*
623 	 * proc0 is not expected to enter usermode, so there is no special
624 	 * handling for sv_minuser here, like is done for exec_new_vmspace().
625 	 */
626 	vm_map_init(&vmspace0.vm_map, vmspace_pmap(&vmspace0),
627 	    p->p_sysent->sv_minuser, p->p_sysent->sv_maxuser);
628 
629 	/*
630 	 * Call the init and ctor for the new thread and proc.  We wait
631 	 * to do this until all other structures are fairly sane.
632 	 */
633 	EVENTHANDLER_DIRECT_INVOKE(process_init, p);
634 	EVENTHANDLER_DIRECT_INVOKE(thread_init, td);
635 #ifdef KDTRACE_HOOKS
636 	kdtrace_proc_ctor(p);
637 	kdtrace_thread_ctor(td);
638 #endif
639 	EVENTHANDLER_DIRECT_INVOKE(process_ctor, p);
640 	EVENTHANDLER_DIRECT_INVOKE(thread_ctor, td);
641 
642 	/*
643 	 * Charge root for one process.
644 	 */
645 	(void)chgproccnt(p->p_ucred->cr_ruidinfo, 1, 0);
646 	PROC_LOCK(p);
647 	racct_add_force(p, RACCT_NPROC, 1);
648 	PROC_UNLOCK(p);
649 }
650 SYSINIT(p0init, SI_SUB_INTRINSIC, SI_ORDER_FIRST, proc0_init, NULL);
651 
652 /* ARGSUSED*/
653 static void
654 proc0_post(void *dummy __unused)
655 {
656 	struct proc *p;
657 	struct rusage ru;
658 	struct thread *td;
659 
660 	/*
661 	 * Now we can look at the time, having had a chance to verify the
662 	 * time from the filesystem.  Pretend that proc0 started now.
663 	 */
664 	sx_slock(&allproc_lock);
665 	FOREACH_PROC_IN_SYSTEM(p) {
666 		PROC_LOCK(p);
667 		if (p->p_state == PRS_NEW) {
668 			PROC_UNLOCK(p);
669 			continue;
670 		}
671 		microuptime(&p->p_stats->p_start);
672 		PROC_STATLOCK(p);
673 		rufetch(p, &ru);	/* Clears thread stats */
674 		p->p_rux.rux_runtime = 0;
675 		p->p_rux.rux_uticks = 0;
676 		p->p_rux.rux_sticks = 0;
677 		p->p_rux.rux_iticks = 0;
678 		PROC_STATUNLOCK(p);
679 		FOREACH_THREAD_IN_PROC(p, td) {
680 			td->td_runtime = 0;
681 		}
682 		PROC_UNLOCK(p);
683 	}
684 	sx_sunlock(&allproc_lock);
685 	PCPU_SET(switchtime, cpu_ticks());
686 	PCPU_SET(switchticks, ticks);
687 }
688 SYSINIT(p0post, SI_SUB_INTRINSIC_POST, SI_ORDER_FIRST, proc0_post, NULL);
689 
690 /*
691  ***************************************************************************
692  ****
693  **** The following SYSINIT's and glue code should be moved to the
694  **** respective files on a per subsystem basis.
695  ****
696  ***************************************************************************
697  */
698 
699 /*
700  * List of paths to try when searching for "init".
701  */
702 static char init_path[MAXPATHLEN] =
703 #ifdef	INIT_PATH
704     __XSTRING(INIT_PATH);
705 #else
706     "/sbin/init:/sbin/oinit:/sbin/init.bak:/rescue/init";
707 #endif
708 SYSCTL_STRING(_kern, OID_AUTO, init_path, CTLFLAG_RD, init_path, 0,
709 	"Path used to search the init process");
710 
711 /*
712  * Shutdown timeout of init(8).
713  * Unused within kernel, but used to control init(8), hence do not remove.
714  */
715 #ifndef INIT_SHUTDOWN_TIMEOUT
716 #define INIT_SHUTDOWN_TIMEOUT 120
717 #endif
718 static int init_shutdown_timeout = INIT_SHUTDOWN_TIMEOUT;
719 SYSCTL_INT(_kern, OID_AUTO, init_shutdown_timeout,
720 	CTLFLAG_RW, &init_shutdown_timeout, 0, "Shutdown timeout of init(8). "
721 	"Unused within kernel, but used to control init(8)");
722 
723 /*
724  * Start the initial user process; try exec'ing each pathname in init_path.
725  * The program is invoked with one argument containing the boot flags.
726  */
727 static void
728 start_init(void *dummy)
729 {
730 	struct image_args args;
731 	int error;
732 	char *var, *path;
733 	char *free_init_path, *tmp_init_path;
734 	struct thread *td;
735 	struct proc *p;
736 	struct vmspace *oldvmspace;
737 
738 	TSENTER();	/* Here so we don't overlap with mi_startup. */
739 
740 	td = curthread;
741 	p = td->td_proc;
742 
743 	vfs_mountroot();
744 
745 	/* Wipe GELI passphrase from the environment. */
746 	kern_unsetenv("kern.geom.eli.passphrase");
747 
748 	/* For Multicons, report which console is primary to both */
749 	if (boothowto & RB_MULTIPLE) {
750 		if (boothowto & RB_SERIAL)
751 			printf("Dual Console: Serial Primary, Video Secondary\n");
752 		else
753 			printf("Dual Console: Video Primary, Serial Secondary\n");
754 	}
755 
756 	if ((var = kern_getenv("init_path")) != NULL) {
757 		strlcpy(init_path, var, sizeof(init_path));
758 		freeenv(var);
759 	}
760 	free_init_path = tmp_init_path = strdup(init_path, M_TEMP);
761 
762 	while ((path = strsep(&tmp_init_path, ":")) != NULL) {
763 		if (bootverbose)
764 			printf("start_init: trying %s\n", path);
765 
766 		memset(&args, 0, sizeof(args));
767 		error = exec_alloc_args(&args);
768 		if (error != 0)
769 			panic("%s: Can't allocate space for init arguments %d",
770 			    __func__, error);
771 
772 		error = exec_args_add_fname(&args, path, UIO_SYSSPACE);
773 		if (error != 0)
774 			panic("%s: Can't add fname %d", __func__, error);
775 		error = exec_args_add_arg(&args, path, UIO_SYSSPACE);
776 		if (error != 0)
777 			panic("%s: Can't add argv[0] %d", __func__, error);
778 		if (boothowto & RB_SINGLE)
779 			error = exec_args_add_arg(&args, "-s", UIO_SYSSPACE);
780 		if (error != 0)
781 			panic("%s: Can't add argv[0] %d", __func__, error);
782 
783 		/*
784 		 * Now try to exec the program.  If can't for any reason
785 		 * other than it doesn't exist, complain.
786 		 *
787 		 * Otherwise, return via fork_trampoline() all the way
788 		 * to user mode as init!
789 		 */
790 		KASSERT((td->td_pflags & TDP_EXECVMSPC) == 0,
791 		    ("nested execve"));
792 		oldvmspace = p->p_vmspace;
793 		error = kern_execve(td, &args, NULL, oldvmspace);
794 		KASSERT(error != 0,
795 		    ("kern_execve returned success, not EJUSTRETURN"));
796 		if (error == EJUSTRETURN) {
797 			exec_cleanup(td, oldvmspace);
798 			free(free_init_path, M_TEMP);
799 			TSEXIT();
800 			return;
801 		}
802 		if (error != ENOENT)
803 			printf("exec %s: error %d\n", path, error);
804 	}
805 	free(free_init_path, M_TEMP);
806 	printf("init: not found in path %s\n", init_path);
807 	panic("no init");
808 }
809 
810 /*
811  * Like kproc_create(), but runs in its own address space.  We do this
812  * early to reserve pid 1.  Note special case - do not make it
813  * runnable yet, init execution is started when userspace can be served.
814  */
815 static void
816 create_init(const void *udata __unused)
817 {
818 	struct fork_req fr;
819 	struct ucred *newcred, *oldcred;
820 	struct thread *td;
821 	int error;
822 
823 	bzero(&fr, sizeof(fr));
824 	fr.fr_flags = RFFDG | RFPROC | RFSTOPPED;
825 	fr.fr_procp = &initproc;
826 	error = fork1(&thread0, &fr);
827 	if (error)
828 		panic("cannot fork init: %d\n", error);
829 	KASSERT(initproc->p_pid == 1, ("create_init: initproc->p_pid != 1"));
830 	/* divorce init's credentials from the kernel's */
831 	newcred = crget();
832 	sx_xlock(&proctree_lock);
833 	PROC_LOCK(initproc);
834 	initproc->p_flag |= P_SYSTEM | P_INMEM;
835 	initproc->p_treeflag |= P_TREE_REAPER;
836 	oldcred = initproc->p_ucred;
837 	crcopy(newcred, oldcred);
838 #ifdef MAC
839 	mac_cred_create_init(newcred);
840 #endif
841 #ifdef AUDIT
842 	audit_cred_proc1(newcred);
843 #endif
844 	proc_set_cred(initproc, newcred);
845 	td = FIRST_THREAD_IN_PROC(initproc);
846 	crcowfree(td);
847 	td->td_realucred = crcowget(initproc->p_ucred);
848 	td->td_ucred = td->td_realucred;
849 	PROC_UNLOCK(initproc);
850 	sx_xunlock(&proctree_lock);
851 	crfree(oldcred);
852 	cpu_fork_kthread_handler(FIRST_THREAD_IN_PROC(initproc),
853 	    start_init, NULL);
854 }
855 SYSINIT(init, SI_SUB_CREATE_INIT, SI_ORDER_FIRST, create_init, NULL);
856 
857 /*
858  * Make it runnable now.
859  */
860 static void
861 kick_init(const void *udata __unused)
862 {
863 	struct thread *td;
864 
865 	td = FIRST_THREAD_IN_PROC(initproc);
866 	thread_lock(td);
867 	TD_SET_CAN_RUN(td);
868 	sched_add(td, SRQ_BORING);
869 }
870 SYSINIT(kickinit, SI_SUB_KTHREAD_INIT, SI_ORDER_MIDDLE, kick_init, NULL);
871 
872 /*
873  * DDB(4).
874  */
875 #ifdef DDB
876 static void
877 db_show_print_syinit(struct sysinit *sip, bool ddb)
878 {
879 	const char *sname, *funcname;
880 	c_db_sym_t sym;
881 	db_expr_t  offset;
882 
883 #define xprint(...)							\
884 	if (ddb)							\
885 		db_printf(__VA_ARGS__);					\
886 	else								\
887 		printf(__VA_ARGS__)
888 
889 	if (sip == NULL) {
890 		xprint("%s: no sysinit * given\n", __func__);
891 		return;
892 	}
893 
894 	sym = db_search_symbol((vm_offset_t)sip, DB_STGY_ANY, &offset);
895 	db_symbol_values(sym, &sname, NULL);
896 	sym = db_search_symbol((vm_offset_t)sip->func, DB_STGY_PROC, &offset);
897 	db_symbol_values(sym, &funcname, NULL);
898 	xprint("%s(%p)\n", (sname != NULL) ? sname : "", sip);
899 	xprint("  %#08x %#08x\n", sip->subsystem, sip->order);
900 	xprint("  %p(%s)(%p)\n",
901 	    sip->func, (funcname != NULL) ? funcname : "", sip->udata);
902 #undef xprint
903 }
904 
905 DB_SHOW_COMMAND_FLAGS(sysinit, db_show_sysinit, DB_CMD_MEMSAFE)
906 {
907 	struct sysinit **sipp;
908 
909 	db_printf("SYSINIT vs Name(Ptr)\n");
910 	db_printf("  Subsystem  Order\n");
911 	db_printf("  Function(Name)(Arg)\n");
912 	for (sipp = sysinit; sipp < sysinit_end; sipp++) {
913 		db_show_print_syinit(*sipp, true);
914 		if (db_pager_quit)
915 			break;
916 	}
917 }
918 #endif /* DDB */
919