xref: /freebsd-13-stable/sys/kern/init_main.c (revision 4078a7051ab146d146e88720fac85ca164af2d4a)
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/conf.h>
55 #include <sys/cpuset.h>
56 #include <sys/dtrace_bsd.h>
57 #include <sys/epoch.h>
58 #include <sys/eventhandler.h>
59 #include <sys/exec.h>
60 #include <sys/file.h>
61 #include <sys/filedesc.h>
62 #include <sys/imgact.h>
63 #include <sys/jail.h>
64 #include <sys/kernel.h>
65 #include <sys/ktr.h>
66 #include <sys/lock.h>
67 #include <sys/loginclass.h>
68 #include <sys/malloc.h>
69 #include <sys/mount.h>
70 #include <sys/mutex.h>
71 #include <sys/proc.h>
72 #include <sys/racct.h>
73 #include <sys/reboot.h>
74 #include <sys/resourcevar.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
linux_alloc_current_noop(struct thread * td __unused,int flags __unused)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 #endif
142 
143 #ifdef INVARIANTS
144 FEATURE(invariants, "Kernel compiled with INVARIANTS, may affect performance");
145 #endif
146 
147 /*
148  * This ensures that there is at least one entry so that the sysinit_set
149  * symbol is not undefined.  A sybsystem ID of SI_SUB_DUMMY is never
150  * executed.
151  */
152 SYSINIT(placeholder, SI_SUB_DUMMY, SI_ORDER_ANY, NULL, NULL);
153 
154 /*
155  * The sysinit table itself.  Items are checked off as the are run.
156  * If we want to register new sysinit types, add them to newsysinit.
157  */
158 SET_DECLARE(sysinit_set, struct sysinit);
159 struct sysinit **sysinit, **sysinit_end;
160 struct sysinit **newsysinit, **newsysinit_end;
161 
162 /*
163  * Merge a new sysinit set into the current set, reallocating it if
164  * necessary.  This can only be called after malloc is running.
165  */
166 void
sysinit_add(struct sysinit ** set,struct sysinit ** set_end)167 sysinit_add(struct sysinit **set, struct sysinit **set_end)
168 {
169 	struct sysinit **newset;
170 	struct sysinit **sipp;
171 	struct sysinit **xipp;
172 	int count;
173 
174 	count = set_end - set;
175 	if (newsysinit)
176 		count += newsysinit_end - newsysinit;
177 	else
178 		count += sysinit_end - sysinit;
179 	newset = malloc(count * sizeof(*sipp), M_TEMP, M_NOWAIT);
180 	if (newset == NULL)
181 		panic("cannot malloc for sysinit");
182 	xipp = newset;
183 	if (newsysinit)
184 		for (sipp = newsysinit; sipp < newsysinit_end; sipp++)
185 			*xipp++ = *sipp;
186 	else
187 		for (sipp = sysinit; sipp < sysinit_end; sipp++)
188 			*xipp++ = *sipp;
189 	for (sipp = set; sipp < set_end; sipp++)
190 		*xipp++ = *sipp;
191 	if (newsysinit)
192 		free(newsysinit, M_TEMP);
193 	newsysinit = newset;
194 	newsysinit_end = newset + count;
195 }
196 
197 #if defined (DDB) && defined(VERBOSE_SYSINIT)
198 static const char *
symbol_name(vm_offset_t va,db_strategy_t strategy)199 symbol_name(vm_offset_t va, db_strategy_t strategy)
200 {
201 	const char *name;
202 	c_db_sym_t sym;
203 	db_expr_t  offset;
204 
205 	if (va == 0)
206 		return (NULL);
207 	sym = db_search_symbol(va, strategy, &offset);
208 	if (offset != 0)
209 		return (NULL);
210 	db_symbol_values(sym, &name, NULL);
211 	return (name);
212 }
213 #endif
214 
215 /*
216  * System startup; initialize the world, create process 0, mount root
217  * filesystem, and fork to create init and pagedaemon.  Most of the
218  * hard work is done in the lower-level initialization routines including
219  * startup(), which does memory initialization and autoconfiguration.
220  *
221  * This allows simple addition of new kernel subsystems that require
222  * boot time initialization.  It also allows substitution of subsystem
223  * (for instance, a scheduler, kernel profiler, or VM system) by object
224  * module.  Finally, it allows for optional "kernel threads".
225  */
226 void
mi_startup(void)227 mi_startup(void)
228 {
229 
230 	struct sysinit **sipp;	/* system initialization*/
231 	struct sysinit **xipp;	/* interior loop of sort*/
232 	struct sysinit *save;	/* bubble*/
233 
234 #if defined(VERBOSE_SYSINIT)
235 	int last;
236 	int verbose;
237 #endif
238 
239 	TSENTER();
240 
241 	if (boothowto & RB_VERBOSE)
242 		bootverbose++;
243 
244 	if (sysinit == NULL) {
245 		sysinit = SET_BEGIN(sysinit_set);
246 		sysinit_end = SET_LIMIT(sysinit_set);
247 	}
248 
249 restart:
250 	/*
251 	 * Perform a bubble sort of the system initialization objects by
252 	 * their subsystem (primary key) and order (secondary key).
253 	 */
254 	for (sipp = sysinit; sipp < sysinit_end; sipp++) {
255 		for (xipp = sipp + 1; xipp < sysinit_end; xipp++) {
256 			if ((*sipp)->subsystem < (*xipp)->subsystem ||
257 			     ((*sipp)->subsystem == (*xipp)->subsystem &&
258 			      (*sipp)->order <= (*xipp)->order))
259 				continue;	/* skip*/
260 			save = *sipp;
261 			*sipp = *xipp;
262 			*xipp = save;
263 		}
264 	}
265 
266 #if defined(VERBOSE_SYSINIT)
267 	last = SI_SUB_DUMMY;
268 	TUNABLE_INT_FETCH("debug.verbose_sysinit", &verbose_sysinit);
269 	verbose = 0;
270 #if !defined(DDB)
271 	printf("VERBOSE_SYSINIT: DDB not enabled, symbol lookups disabled.\n");
272 #endif
273 #endif
274 
275 	/*
276 	 * Traverse the (now) ordered list of system initialization tasks.
277 	 * Perform each task, and continue on to the next task.
278 	 */
279 	for (sipp = sysinit; sipp < sysinit_end; sipp++) {
280 		if ((*sipp)->subsystem == SI_SUB_DUMMY)
281 			continue;	/* skip dummy task(s)*/
282 
283 		if ((*sipp)->subsystem == SI_SUB_DONE)
284 			continue;
285 
286 #if defined(VERBOSE_SYSINIT)
287 		if ((*sipp)->subsystem > last && verbose_sysinit != 0) {
288 			verbose = 1;
289 			last = (*sipp)->subsystem;
290 			printf("subsystem %x\n", last);
291 		}
292 		if (verbose) {
293 #if defined(DDB)
294 			const char *func, *data;
295 
296 			func = symbol_name((vm_offset_t)(*sipp)->func,
297 			    DB_STGY_PROC);
298 			data = symbol_name((vm_offset_t)(*sipp)->udata,
299 			    DB_STGY_ANY);
300 			if (func != NULL && data != NULL)
301 				printf("   %s(&%s)... ", func, data);
302 			else if (func != NULL)
303 				printf("   %s(%p)... ", func, (*sipp)->udata);
304 			else
305 #endif
306 				printf("   %p(%p)... ", (*sipp)->func,
307 				    (*sipp)->udata);
308 		}
309 #endif
310 
311 		/* Call function */
312 		(*((*sipp)->func))((*sipp)->udata);
313 
314 #if defined(VERBOSE_SYSINIT)
315 		if (verbose)
316 			printf("done.\n");
317 #endif
318 
319 		/* Check off the one we're just done */
320 		(*sipp)->subsystem = SI_SUB_DONE;
321 
322 		/* Check if we've installed more sysinit items via KLD */
323 		if (newsysinit != NULL) {
324 			if (sysinit != SET_BEGIN(sysinit_set))
325 				free(sysinit, M_TEMP);
326 			sysinit = newsysinit;
327 			sysinit_end = newsysinit_end;
328 			newsysinit = NULL;
329 			newsysinit_end = NULL;
330 			goto restart;
331 		}
332 	}
333 
334 	TSEXIT();	/* Here so we don't overlap with start_init. */
335 
336 	mtx_assert(&Giant, MA_OWNED | MA_NOTRECURSED);
337 	mtx_unlock(&Giant);
338 
339 	/*
340 	 * Now hand over this thread to swapper.
341 	 */
342 	swapper();
343 	/* NOTREACHED*/
344 }
345 
346 static void
print_caddr_t(const void * data)347 print_caddr_t(const void *data)
348 {
349 	printf("%s", (const char *)data);
350 }
351 
352 static void
print_version(const void * data __unused)353 print_version(const void *data __unused)
354 {
355 	int len;
356 
357 	/* Strip a trailing newline from version. */
358 	len = strlen(version);
359 	while (len > 0 && version[len - 1] == '\n')
360 		len--;
361 	printf("%.*s %s\n", len, version, machine);
362 	printf("%s\n", compiler_version);
363 }
364 
365 C_SYSINIT(announce, SI_SUB_COPYRIGHT, SI_ORDER_FIRST, print_caddr_t,
366     copyright);
367 C_SYSINIT(trademark, SI_SUB_COPYRIGHT, SI_ORDER_SECOND, print_caddr_t,
368     trademark);
369 C_SYSINIT(version, SI_SUB_COPYRIGHT, SI_ORDER_THIRD, print_version, NULL);
370 
371 #ifdef WITNESS
372 static const char wit_warn[] =
373      "WARNING: WITNESS option enabled, expect reduced performance.\n";
374 C_SYSINIT(witwarn, SI_SUB_COPYRIGHT, SI_ORDER_FOURTH,
375    print_caddr_t, wit_warn);
376 C_SYSINIT(witwarn2, SI_SUB_LAST, SI_ORDER_FOURTH,
377    print_caddr_t, wit_warn);
378 #endif
379 
380 #ifdef DIAGNOSTIC
381 static const char diag_warn[] =
382      "WARNING: DIAGNOSTIC option enabled, expect reduced performance.\n";
383 C_SYSINIT(diagwarn, SI_SUB_COPYRIGHT, SI_ORDER_FIFTH,
384     print_caddr_t, diag_warn);
385 C_SYSINIT(diagwarn2, SI_SUB_LAST, SI_ORDER_FIFTH,
386     print_caddr_t, diag_warn);
387 #endif
388 
389 #if __SIZEOF_LONG__ == 4
390 static const char ilp32_warn[] =
391     "WARNING: 32-bit kernels are deprecated and may be removed in FreeBSD 15.0.\n";
392 C_SYSINIT(ilp32warn, SI_SUB_COPYRIGHT, SI_ORDER_FIFTH,
393     print_caddr_t, ilp32_warn);
394 C_SYSINIT(ilp32warn2, SI_SUB_LAST, SI_ORDER_FIFTH,
395     print_caddr_t, ilp32_warn);
396 #endif
397 
398 static int
null_fetch_syscall_args(struct thread * td __unused)399 null_fetch_syscall_args(struct thread *td __unused)
400 {
401 
402 	panic("null_fetch_syscall_args");
403 }
404 
405 static void
null_set_syscall_retval(struct thread * td __unused,int error __unused)406 null_set_syscall_retval(struct thread *td __unused, int error __unused)
407 {
408 
409 	panic("null_set_syscall_retval");
410 }
411 
412 static void
null_set_fork_retval(struct thread * td __unused)413 null_set_fork_retval(struct thread *td __unused)
414 {
415 
416 }
417 
418 struct sysentvec null_sysvec = {
419 	.sv_size	= 0,
420 	.sv_table	= NULL,
421 	.sv_fixup	= NULL,
422 	.sv_sendsig	= NULL,
423 	.sv_sigcode	= NULL,
424 	.sv_szsigcode	= NULL,
425 	.sv_name	= "null",
426 	.sv_coredump	= NULL,
427 	.sv_imgact_try	= NULL,
428 	.sv_minsigstksz	= 0,
429 	.sv_minuser	= VM_MIN_ADDRESS,
430 	.sv_maxuser	= VM_MAXUSER_ADDRESS,
431 	.sv_usrstack	= USRSTACK,
432 	.sv_psstrings	= PS_STRINGS,
433 	.sv_psstringssz	= sizeof(struct ps_strings),
434 	.sv_stackprot	= VM_PROT_ALL,
435 	.sv_copyout_strings	= NULL,
436 	.sv_setregs	= NULL,
437 	.sv_fixlimit	= NULL,
438 	.sv_maxssiz	= NULL,
439 	.sv_flags	= 0,
440 	.sv_set_syscall_retval = null_set_syscall_retval,
441 	.sv_fetch_syscall_args = null_fetch_syscall_args,
442 	.sv_syscallnames = NULL,
443 	.sv_schedtail	= NULL,
444 	.sv_thread_detach = NULL,
445 	.sv_trap	= NULL,
446 	.sv_regset_begin = NULL,
447 	.sv_regset_end  = NULL,
448 	.sv_set_fork_retval = null_set_fork_retval,
449 };
450 
451 /*
452  * The two following SYSINIT's are proc0 specific glue code.  I am not
453  * convinced that they can not be safely combined, but their order of
454  * operation has been maintained as the same as the original init_main.c
455  * for right now.
456  */
457 /* ARGSUSED*/
458 static void
proc0_init(void * dummy __unused)459 proc0_init(void *dummy __unused)
460 {
461 	struct proc *p;
462 	struct thread *td;
463 	struct ucred *newcred;
464 	struct uidinfo tmpuinfo;
465 	struct loginclass tmplc = {
466 		.lc_name = "",
467 	};
468 	vm_paddr_t pageablemem;
469 	int i;
470 
471 	GIANT_REQUIRED;
472 	p = &proc0;
473 	td = &thread0;
474 
475 	/*
476 	 * Initialize magic number and osrel.
477 	 */
478 	p->p_magic = P_MAGIC;
479 	p->p_osrel = osreldate;
480 
481 	/*
482 	 * Initialize thread and process structures.
483 	 */
484 	procinit();	/* set up proc zone */
485 	threadinit();	/* set up UMA zones */
486 
487 	/*
488 	 * Initialise scheduler resources.
489 	 * Add scheduler specific parts to proc, thread as needed.
490 	 */
491 	schedinit();	/* scheduler gets its house in order */
492 
493 	/*
494 	 * Create process 0 (the swapper).
495 	 */
496 	LIST_INSERT_HEAD(&allproc, p, p_list);
497 	LIST_INSERT_HEAD(PIDHASH(0), p, p_hash);
498 	mtx_init(&pgrp0.pg_mtx, "process group", NULL, MTX_DEF | MTX_DUPOK);
499 	sx_init(&pgrp0.pg_killsx, "killpg racer");
500 	p->p_pgrp = &pgrp0;
501 	LIST_INSERT_HEAD(PGRPHASH(0), &pgrp0, pg_hash);
502 	LIST_INIT(&pgrp0.pg_members);
503 	LIST_INSERT_HEAD(&pgrp0.pg_members, p, p_pglist);
504 
505 	pgrp0.pg_session = &session0;
506 	mtx_init(&session0.s_mtx, "session", NULL, MTX_DEF);
507 	refcount_init(&session0.s_count, 1);
508 	session0.s_leader = p;
509 
510 	p->p_sysent = &null_sysvec;
511 	p->p_flag = P_SYSTEM | P_INMEM | P_KPROC;
512 	p->p_flag2 = 0;
513 	p->p_state = PRS_NORMAL;
514 	p->p_klist = knlist_alloc(&p->p_mtx);
515 	STAILQ_INIT(&p->p_ktr);
516 	p->p_nice = NZERO;
517 	td->td_tid = THREAD0_TID;
518 	tidhash_add(td);
519 	td->td_state = TDS_RUNNING;
520 	td->td_pri_class = PRI_TIMESHARE;
521 	td->td_user_pri = PUSER;
522 	td->td_base_user_pri = PUSER;
523 	td->td_lend_user_pri = PRI_MAX;
524 	td->td_priority = PVM;
525 	td->td_base_pri = PVM;
526 	td->td_oncpu = curcpu;
527 	td->td_flags = TDF_INMEM;
528 	td->td_pflags = TDP_KTHREAD;
529 	td->td_cpuset = cpuset_thread0();
530 	td->td_domain.dr_policy = td->td_cpuset->cs_domain;
531 	prison0_init();
532 	p->p_peers = 0;
533 	p->p_leader = p;
534 	p->p_reaper = p;
535 	p->p_treeflag |= P_TREE_REAPER;
536 	LIST_INIT(&p->p_reaplist);
537 
538 	strncpy(p->p_comm, "kernel", sizeof (p->p_comm));
539 	strncpy(td->td_name, "swapper", sizeof (td->td_name));
540 
541 	callout_init_mtx(&p->p_itcallout, &p->p_mtx, 0);
542 	callout_init_mtx(&p->p_limco, &p->p_mtx, 0);
543 	callout_init(&td->td_slpcallout, 1);
544 	TAILQ_INIT(&p->p_kqtim_stop);
545 
546 	/* Create credentials. */
547 	newcred = crget();
548 	newcred->cr_ngroups = 1;	/* group 0 */
549 	/* A hack to prevent uifind from tripping over NULL pointers. */
550 	curthread->td_ucred = newcred;
551 	tmpuinfo.ui_uid = 1;
552 	newcred->cr_uidinfo = newcred->cr_ruidinfo = &tmpuinfo;
553 	newcred->cr_uidinfo = uifind(0);
554 	newcred->cr_ruidinfo = uifind(0);
555 	newcred->cr_loginclass = &tmplc;
556 	newcred->cr_loginclass = loginclass_find("default");
557 	/* End hack. creds get properly set later with thread_cow_get_proc */
558 	curthread->td_ucred = NULL;
559 	newcred->cr_prison = &prison0;
560 	newcred->cr_users++; /* avoid assertion failure */
561 	p->p_ucred = crcowget(newcred);
562 	newcred->cr_users--;
563 	crfree(newcred);
564 #ifdef AUDIT
565 	audit_cred_kproc0(newcred);
566 #endif
567 #ifdef MAC
568 	mac_cred_create_swapper(newcred);
569 #endif
570 	/* Create sigacts. */
571 	p->p_sigacts = sigacts_alloc();
572 
573 	/* Initialize signal state for process 0. */
574 	siginit(&proc0);
575 
576 	/* Create the file descriptor table. */
577 	p->p_pd = pdinit(NULL, false);
578 	p->p_fd = fdinit(NULL, false, NULL);
579 	p->p_fdtol = NULL;
580 
581 	/* Create the limits structures. */
582 	p->p_limit = lim_alloc();
583 	for (i = 0; i < RLIM_NLIMITS; i++)
584 		p->p_limit->pl_rlimit[i].rlim_cur =
585 		    p->p_limit->pl_rlimit[i].rlim_max = RLIM_INFINITY;
586 	p->p_limit->pl_rlimit[RLIMIT_NOFILE].rlim_cur =
587 	    p->p_limit->pl_rlimit[RLIMIT_NOFILE].rlim_max = maxfiles;
588 	p->p_limit->pl_rlimit[RLIMIT_NPROC].rlim_cur =
589 	    p->p_limit->pl_rlimit[RLIMIT_NPROC].rlim_max = maxproc;
590 	p->p_limit->pl_rlimit[RLIMIT_DATA].rlim_cur = dfldsiz;
591 	p->p_limit->pl_rlimit[RLIMIT_DATA].rlim_max = maxdsiz;
592 	p->p_limit->pl_rlimit[RLIMIT_STACK].rlim_cur = dflssiz;
593 	p->p_limit->pl_rlimit[RLIMIT_STACK].rlim_max = maxssiz;
594 	/* Cast to avoid overflow on i386/PAE. */
595 	pageablemem = ptoa((vm_paddr_t)vm_free_count());
596 	p->p_limit->pl_rlimit[RLIMIT_RSS].rlim_cur =
597 	    p->p_limit->pl_rlimit[RLIMIT_RSS].rlim_max = pageablemem;
598 	p->p_limit->pl_rlimit[RLIMIT_MEMLOCK].rlim_cur = pageablemem / 3;
599 	p->p_limit->pl_rlimit[RLIMIT_MEMLOCK].rlim_max = pageablemem;
600 	p->p_cpulimit = RLIM_INFINITY;
601 
602 	PROC_LOCK(p);
603 	thread_cow_get_proc(td, p);
604 	PROC_UNLOCK(p);
605 
606 	/* Initialize resource accounting structures. */
607 	racct_create(&p->p_racct);
608 
609 	p->p_stats = pstats_alloc();
610 
611 	/* Allocate a prototype map so we have something to fork. */
612 	p->p_vmspace = &vmspace0;
613 	refcount_init(&vmspace0.vm_refcnt, 1);
614 	pmap_pinit0(vmspace_pmap(&vmspace0));
615 
616 	/*
617 	 * proc0 is not expected to enter usermode, so there is no special
618 	 * handling for sv_minuser here, like is done for exec_new_vmspace().
619 	 */
620 	vm_map_init(&vmspace0.vm_map, vmspace_pmap(&vmspace0),
621 	    p->p_sysent->sv_minuser, p->p_sysent->sv_maxuser);
622 
623 	/*
624 	 * Call the init and ctor for the new thread and proc.  We wait
625 	 * to do this until all other structures are fairly sane.
626 	 */
627 	EVENTHANDLER_DIRECT_INVOKE(process_init, p);
628 	EVENTHANDLER_DIRECT_INVOKE(thread_init, td);
629 #ifdef KDTRACE_HOOKS
630 	kdtrace_proc_ctor(p);
631 	kdtrace_thread_ctor(td);
632 #endif
633 	EVENTHANDLER_DIRECT_INVOKE(process_ctor, p);
634 	EVENTHANDLER_DIRECT_INVOKE(thread_ctor, td);
635 
636 	/*
637 	 * Charge root for one process.
638 	 */
639 	(void)chgproccnt(p->p_ucred->cr_ruidinfo, 1, 0);
640 	PROC_LOCK(p);
641 	racct_add_force(p, RACCT_NPROC, 1);
642 	PROC_UNLOCK(p);
643 }
644 SYSINIT(p0init, SI_SUB_INTRINSIC, SI_ORDER_FIRST, proc0_init, NULL);
645 
646 /* ARGSUSED*/
647 static void
proc0_post(void * dummy __unused)648 proc0_post(void *dummy __unused)
649 {
650 	struct proc *p;
651 	struct rusage ru;
652 	struct thread *td;
653 
654 	/*
655 	 * Now we can look at the time, having had a chance to verify the
656 	 * time from the filesystem.  Pretend that proc0 started now.
657 	 */
658 	sx_slock(&allproc_lock);
659 	FOREACH_PROC_IN_SYSTEM(p) {
660 		PROC_LOCK(p);
661 		if (p->p_state == PRS_NEW) {
662 			PROC_UNLOCK(p);
663 			continue;
664 		}
665 		microuptime(&p->p_stats->p_start);
666 		PROC_STATLOCK(p);
667 		rufetch(p, &ru);	/* Clears thread stats */
668 		p->p_rux.rux_runtime = 0;
669 		p->p_rux.rux_uticks = 0;
670 		p->p_rux.rux_sticks = 0;
671 		p->p_rux.rux_iticks = 0;
672 		PROC_STATUNLOCK(p);
673 		FOREACH_THREAD_IN_PROC(p, td) {
674 			td->td_runtime = 0;
675 		}
676 		PROC_UNLOCK(p);
677 	}
678 	sx_sunlock(&allproc_lock);
679 	PCPU_SET(switchtime, cpu_ticks());
680 	PCPU_SET(switchticks, ticks);
681 }
682 SYSINIT(p0post, SI_SUB_INTRINSIC_POST, SI_ORDER_FIRST, proc0_post, NULL);
683 
684 /*
685  ***************************************************************************
686  ****
687  **** The following SYSINIT's and glue code should be moved to the
688  **** respective files on a per subsystem basis.
689  ****
690  ***************************************************************************
691  */
692 
693 /*
694  * List of paths to try when searching for "init".
695  */
696 static char init_path[MAXPATHLEN] =
697 #ifdef	INIT_PATH
698     __XSTRING(INIT_PATH);
699 #else
700     "/sbin/init:/sbin/oinit:/sbin/init.bak:/rescue/init";
701 #endif
702 SYSCTL_STRING(_kern, OID_AUTO, init_path, CTLFLAG_RD, init_path, 0,
703 	"Path used to search the init process");
704 
705 /*
706  * Shutdown timeout of init(8).
707  * Unused within kernel, but used to control init(8), hence do not remove.
708  */
709 #ifndef INIT_SHUTDOWN_TIMEOUT
710 #define INIT_SHUTDOWN_TIMEOUT 120
711 #endif
712 static int init_shutdown_timeout = INIT_SHUTDOWN_TIMEOUT;
713 SYSCTL_INT(_kern, OID_AUTO, init_shutdown_timeout,
714 	CTLFLAG_RW, &init_shutdown_timeout, 0, "Shutdown timeout of init(8). "
715 	"Unused within kernel, but used to control init(8)");
716 
717 /*
718  * Start the initial user process; try exec'ing each pathname in init_path.
719  * The program is invoked with one argument containing the boot flags.
720  */
721 static void
start_init(void * dummy)722 start_init(void *dummy)
723 {
724 	struct image_args args;
725 	int error;
726 	char *var, *path;
727 	char *free_init_path, *tmp_init_path;
728 	struct thread *td;
729 	struct proc *p;
730 	struct vmspace *oldvmspace;
731 
732 	TSENTER();	/* Here so we don't overlap with mi_startup. */
733 
734 	td = curthread;
735 	p = td->td_proc;
736 
737 	vfs_mountroot();
738 
739 	/* Wipe GELI passphrase from the environment. */
740 	kern_unsetenv("kern.geom.eli.passphrase");
741 
742 	/* For Multicons, report which console is primary to both */
743 	if (boothowto & RB_MULTIPLE) {
744 		if (boothowto & RB_SERIAL)
745 			printf("Dual Console: Serial Primary, Video Secondary\n");
746 		else
747 			printf("Dual Console: Video Primary, Serial Secondary\n");
748 	}
749 
750 	if ((var = kern_getenv("init_path")) != NULL) {
751 		strlcpy(init_path, var, sizeof(init_path));
752 		freeenv(var);
753 	}
754 	free_init_path = tmp_init_path = strdup(init_path, M_TEMP);
755 
756 	while ((path = strsep(&tmp_init_path, ":")) != NULL) {
757 		if (bootverbose)
758 			printf("start_init: trying %s\n", path);
759 
760 		memset(&args, 0, sizeof(args));
761 		error = exec_alloc_args(&args);
762 		if (error != 0)
763 			panic("%s: Can't allocate space for init arguments %d",
764 			    __func__, error);
765 
766 		error = exec_args_add_fname(&args, path, UIO_SYSSPACE);
767 		if (error != 0)
768 			panic("%s: Can't add fname %d", __func__, error);
769 		error = exec_args_add_arg(&args, path, UIO_SYSSPACE);
770 		if (error != 0)
771 			panic("%s: Can't add argv[0] %d", __func__, error);
772 		if (boothowto & RB_SINGLE)
773 			error = exec_args_add_arg(&args, "-s", UIO_SYSSPACE);
774 		if (error != 0)
775 			panic("%s: Can't add argv[0] %d", __func__, error);
776 
777 		/*
778 		 * Now try to exec the program.  If can't for any reason
779 		 * other than it doesn't exist, complain.
780 		 *
781 		 * Otherwise, return via fork_trampoline() all the way
782 		 * to user mode as init!
783 		 */
784 		KASSERT((td->td_pflags & TDP_EXECVMSPC) == 0,
785 		    ("nested execve"));
786 		memset(td->td_frame, 0, sizeof(*td->td_frame));
787 		oldvmspace = p->p_vmspace;
788 		error = kern_execve(td, &args, NULL, oldvmspace);
789 		KASSERT(error != 0,
790 		    ("kern_execve returned success, not EJUSTRETURN"));
791 		if (error == EJUSTRETURN) {
792 			exec_cleanup(td, oldvmspace);
793 			free(free_init_path, M_TEMP);
794 			TSEXIT();
795 			return;
796 		}
797 		if (error != ENOENT)
798 			printf("exec %s: error %d\n", path, error);
799 	}
800 	free(free_init_path, M_TEMP);
801 	printf("init: not found in path %s\n", init_path);
802 	panic("no init");
803 }
804 
805 /*
806  * Like kproc_create(), but runs in its own address space.  We do this
807  * early to reserve pid 1.  Note special case - do not make it
808  * runnable yet, init execution is started when userspace can be served.
809  */
810 static void
create_init(const void * udata __unused)811 create_init(const void *udata __unused)
812 {
813 	struct fork_req fr;
814 	struct ucred *newcred, *oldcred;
815 	struct thread *td;
816 	int error;
817 
818 	bzero(&fr, sizeof(fr));
819 	fr.fr_flags = RFFDG | RFPROC | RFSTOPPED;
820 	fr.fr_procp = &initproc;
821 	error = fork1(&thread0, &fr);
822 	if (error)
823 		panic("cannot fork init: %d\n", error);
824 	KASSERT(initproc->p_pid == 1, ("create_init: initproc->p_pid != 1"));
825 	/* divorce init's credentials from the kernel's */
826 	newcred = crget();
827 	sx_xlock(&proctree_lock);
828 	PROC_LOCK(initproc);
829 	initproc->p_flag |= P_SYSTEM | P_INMEM;
830 	initproc->p_treeflag |= P_TREE_REAPER;
831 	oldcred = initproc->p_ucred;
832 	crcopy(newcred, oldcred);
833 #ifdef MAC
834 	mac_cred_create_init(newcred);
835 #endif
836 #ifdef AUDIT
837 	audit_cred_proc1(newcred);
838 #endif
839 	proc_set_cred(initproc, newcred);
840 	td = FIRST_THREAD_IN_PROC(initproc);
841 	crcowfree(td);
842 	td->td_realucred = crcowget(initproc->p_ucred);
843 	td->td_ucred = td->td_realucred;
844 	PROC_UNLOCK(initproc);
845 	sx_xunlock(&proctree_lock);
846 	crfree(oldcred);
847 	cpu_fork_kthread_handler(FIRST_THREAD_IN_PROC(initproc),
848 	    start_init, NULL);
849 }
850 SYSINIT(init, SI_SUB_CREATE_INIT, SI_ORDER_FIRST, create_init, NULL);
851 
852 /*
853  * Make it runnable now.
854  */
855 static void
kick_init(const void * udata __unused)856 kick_init(const void *udata __unused)
857 {
858 	struct thread *td;
859 
860 	td = FIRST_THREAD_IN_PROC(initproc);
861 	thread_lock(td);
862 	TD_SET_CAN_RUN(td);
863 	sched_add(td, SRQ_BORING);
864 }
865 SYSINIT(kickinit, SI_SUB_KTHREAD_INIT, SI_ORDER_MIDDLE, kick_init, NULL);
866 
867 /*
868  * DDB(4).
869  */
870 #ifdef DDB
871 static void
db_show_print_syinit(struct sysinit * sip,bool ddb)872 db_show_print_syinit(struct sysinit *sip, bool ddb)
873 {
874 	const char *sname, *funcname;
875 	c_db_sym_t sym;
876 	db_expr_t  offset;
877 
878 #define xprint(...)							\
879 	if (ddb)							\
880 		db_printf(__VA_ARGS__);					\
881 	else								\
882 		printf(__VA_ARGS__)
883 
884 	if (sip == NULL) {
885 		xprint("%s: no sysinit * given\n", __func__);
886 		return;
887 	}
888 
889 	sym = db_search_symbol((vm_offset_t)sip, DB_STGY_ANY, &offset);
890 	db_symbol_values(sym, &sname, NULL);
891 	sym = db_search_symbol((vm_offset_t)sip->func, DB_STGY_PROC, &offset);
892 	db_symbol_values(sym, &funcname, NULL);
893 	xprint("%s(%p)\n", (sname != NULL) ? sname : "", sip);
894 	xprint("  %#08x %#08x\n", sip->subsystem, sip->order);
895 	xprint("  %p(%s)(%p)\n",
896 	    sip->func, (funcname != NULL) ? funcname : "", sip->udata);
897 #undef xprint
898 }
899 
DB_SHOW_COMMAND(sysinit,db_show_sysinit)900 DB_SHOW_COMMAND(sysinit, db_show_sysinit)
901 {
902 	struct sysinit **sipp;
903 
904 	db_printf("SYSINIT vs Name(Ptr)\n");
905 	db_printf("  Subsystem  Order\n");
906 	db_printf("  Function(Name)(Arg)\n");
907 	for (sipp = sysinit; sipp < sysinit_end; sipp++) {
908 		db_show_print_syinit(*sipp, true);
909 		if (db_pager_quit)
910 			break;
911 	}
912 }
913 #endif /* DDB */
914