xref: /freebsd-13.1/sys/kern/init_main.c (revision 947e8491)
1 /*-
2  * SPDX-License-Identifier: BSD-4-Clause
3  *
4  * Copyright (c) 1995 Terrence R. Lambert
5  * All rights reserved.
6  *
7  * Copyright (c) 1982, 1986, 1989, 1991, 1992, 1993
8  *	The Regents of the University of California.  All rights reserved.
9  * (c) UNIX System Laboratories, Inc.
10  * All or some portions of this file are derived from material licensed
11  * to the University of California by American Telephone and Telegraph
12  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
13  * the permission of UNIX System Laboratories, Inc.
14  *
15  * Redistribution and use in source and binary forms, with or without
16  * modification, are permitted provided that the following conditions
17  * are met:
18  * 1. Redistributions of source code must retain the above copyright
19  *    notice, this list of conditions and the following disclaimer.
20  * 2. Redistributions in binary form must reproduce the above copyright
21  *    notice, this list of conditions and the following disclaimer in the
22  *    documentation and/or other materials provided with the distribution.
23  * 3. All advertising materials mentioning features or use of this software
24  *    must display the following acknowledgement:
25  *	This product includes software developed by the University of
26  *	California, Berkeley and its contributors.
27  * 4. Neither the name of the University nor the names of its contributors
28  *    may be used to endorse or promote products derived from this software
29  *    without specific prior written permission.
30  *
31  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
32  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
33  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
34  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
35  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
36  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
37  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
38  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
39  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
40  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
41  * SUCH DAMAGE.
42  *
43  *	@(#)init_main.c	8.9 (Berkeley) 1/21/94
44  */
45 
46 #include <sys/cdefs.h>
47 __FBSDID("$FreeBSD$");
48 
49 #include "opt_ddb.h"
50 #include "opt_kdb.h"
51 #include "opt_init_path.h"
52 #include "opt_verbose_sysinit.h"
53 
54 #include <sys/param.h>
55 #include <sys/kernel.h>
56 #include <sys/epoch.h>
57 #include <sys/eventhandler.h>
58 #include <sys/exec.h>
59 #include <sys/file.h>
60 #include <sys/filedesc.h>
61 #include <sys/imgact.h>
62 #include <sys/jail.h>
63 #include <sys/ktr.h>
64 #include <sys/lock.h>
65 #include <sys/loginclass.h>
66 #include <sys/mount.h>
67 #include <sys/mutex.h>
68 #include <sys/dtrace_bsd.h>
69 #include <sys/syscallsubr.h>
70 #include <sys/sysctl.h>
71 #include <sys/proc.h>
72 #include <sys/racct.h>
73 #include <sys/resourcevar.h>
74 #include <sys/systm.h>
75 #include <sys/signalvar.h>
76 #include <sys/vnode.h>
77 #include <sys/sysent.h>
78 #include <sys/reboot.h>
79 #include <sys/sched.h>
80 #include <sys/sx.h>
81 #include <sys/sysproto.h>
82 #include <sys/vmmeter.h>
83 #include <sys/unistd.h>
84 #include <sys/malloc.h>
85 #include <sys/conf.h>
86 #include <sys/cpuset.h>
87 
88 #include <machine/cpu.h>
89 
90 #include <security/audit/audit.h>
91 #include <security/mac/mac_framework.h>
92 
93 #include <vm/vm.h>
94 #include <vm/vm_param.h>
95 #include <vm/vm_extern.h>
96 #include <vm/pmap.h>
97 #include <vm/vm_map.h>
98 #include <sys/copyright.h>
99 
100 #include <ddb/ddb.h>
101 #include <ddb/db_sym.h>
102 
103 void mi_startup(void);				/* Should be elsewhere */
104 
105 /* Components of the first process -- never freed. */
106 static struct session session0;
107 static struct pgrp pgrp0;
108 struct	proc proc0;
109 struct thread0_storage thread0_st __aligned(32);
110 struct	vmspace vmspace0;
111 struct	proc *initproc;
112 
113 int
linux_alloc_current_noop(struct thread * td __unused,int flags __unused)114 linux_alloc_current_noop(struct thread *td __unused, int flags __unused)
115 {
116 	return (0);
117 }
118 int (*lkpi_alloc_current)(struct thread *, int) = linux_alloc_current_noop;
119 
120 #ifndef BOOTHOWTO
121 #define	BOOTHOWTO	0
122 #endif
123 int	boothowto = BOOTHOWTO;	/* initialized so that it can be patched */
124 SYSCTL_INT(_debug, OID_AUTO, boothowto, CTLFLAG_RD, &boothowto, 0,
125 	"Boot control flags, passed from loader");
126 
127 #ifndef BOOTVERBOSE
128 #define	BOOTVERBOSE	0
129 #endif
130 int	bootverbose = BOOTVERBOSE;
131 SYSCTL_INT(_debug, OID_AUTO, bootverbose, CTLFLAG_RW, &bootverbose, 0,
132 	"Control the output of verbose kernel messages");
133 
134 #ifdef VERBOSE_SYSINIT
135 /*
136  * We'll use the defined value of VERBOSE_SYSINIT from the kernel config to
137  * dictate the default VERBOSE_SYSINIT behavior.  Significant values for this
138  * option and associated tunable are:
139  * - 0, 'compiled in but silent by default'
140  * - 1, 'compiled in but verbose by default' (default)
141  */
142 int	verbose_sysinit = VERBOSE_SYSINIT;
143 TUNABLE_INT("debug.verbose_sysinit", &verbose_sysinit);
144 #endif
145 
146 #ifdef INVARIANTS
147 FEATURE(invariants, "Kernel compiled with INVARIANTS, may affect performance");
148 #endif
149 
150 /*
151  * This ensures that there is at least one entry so that the sysinit_set
152  * symbol is not undefined.  A sybsystem ID of SI_SUB_DUMMY is never
153  * executed.
154  */
155 SYSINIT(placeholder, SI_SUB_DUMMY, SI_ORDER_ANY, NULL, NULL);
156 
157 /*
158  * The sysinit table itself.  Items are checked off as the are run.
159  * If we want to register new sysinit types, add them to newsysinit.
160  */
161 SET_DECLARE(sysinit_set, struct sysinit);
162 struct sysinit **sysinit, **sysinit_end;
163 struct sysinit **newsysinit, **newsysinit_end;
164 
165 /*
166  * Merge a new sysinit set into the current set, reallocating it if
167  * necessary.  This can only be called after malloc is running.
168  */
169 void
sysinit_add(struct sysinit ** set,struct sysinit ** set_end)170 sysinit_add(struct sysinit **set, struct sysinit **set_end)
171 {
172 	struct sysinit **newset;
173 	struct sysinit **sipp;
174 	struct sysinit **xipp;
175 	int count;
176 
177 	count = set_end - set;
178 	if (newsysinit)
179 		count += newsysinit_end - newsysinit;
180 	else
181 		count += sysinit_end - sysinit;
182 	newset = malloc(count * sizeof(*sipp), M_TEMP, M_NOWAIT);
183 	if (newset == NULL)
184 		panic("cannot malloc for sysinit");
185 	xipp = newset;
186 	if (newsysinit)
187 		for (sipp = newsysinit; sipp < newsysinit_end; sipp++)
188 			*xipp++ = *sipp;
189 	else
190 		for (sipp = sysinit; sipp < sysinit_end; sipp++)
191 			*xipp++ = *sipp;
192 	for (sipp = set; sipp < set_end; sipp++)
193 		*xipp++ = *sipp;
194 	if (newsysinit)
195 		free(newsysinit, M_TEMP);
196 	newsysinit = newset;
197 	newsysinit_end = newset + count;
198 }
199 
200 #if defined (DDB) && defined(VERBOSE_SYSINIT)
201 static const char *
symbol_name(vm_offset_t va,db_strategy_t strategy)202 symbol_name(vm_offset_t va, db_strategy_t strategy)
203 {
204 	const char *name;
205 	c_db_sym_t sym;
206 	db_expr_t  offset;
207 
208 	if (va == 0)
209 		return (NULL);
210 	sym = db_search_symbol(va, strategy, &offset);
211 	if (offset != 0)
212 		return (NULL);
213 	db_symbol_values(sym, &name, NULL);
214 	return (name);
215 }
216 #endif
217 
218 /*
219  * System startup; initialize the world, create process 0, mount root
220  * filesystem, and fork to create init and pagedaemon.  Most of the
221  * hard work is done in the lower-level initialization routines including
222  * startup(), which does memory initialization and autoconfiguration.
223  *
224  * This allows simple addition of new kernel subsystems that require
225  * boot time initialization.  It also allows substitution of subsystem
226  * (for instance, a scheduler, kernel profiler, or VM system) by object
227  * module.  Finally, it allows for optional "kernel threads".
228  */
229 void
mi_startup(void)230 mi_startup(void)
231 {
232 
233 	struct sysinit **sipp;	/* system initialization*/
234 	struct sysinit **xipp;	/* interior loop of sort*/
235 	struct sysinit *save;	/* bubble*/
236 
237 #if defined(VERBOSE_SYSINIT)
238 	int last;
239 	int verbose;
240 #endif
241 
242 	TSENTER();
243 
244 	if (boothowto & RB_VERBOSE)
245 		bootverbose++;
246 
247 	if (sysinit == NULL) {
248 		sysinit = SET_BEGIN(sysinit_set);
249 		sysinit_end = SET_LIMIT(sysinit_set);
250 	}
251 
252 restart:
253 	/*
254 	 * Perform a bubble sort of the system initialization objects by
255 	 * their subsystem (primary key) and order (secondary key).
256 	 */
257 	for (sipp = sysinit; sipp < sysinit_end; sipp++) {
258 		for (xipp = sipp + 1; xipp < sysinit_end; xipp++) {
259 			if ((*sipp)->subsystem < (*xipp)->subsystem ||
260 			     ((*sipp)->subsystem == (*xipp)->subsystem &&
261 			      (*sipp)->order <= (*xipp)->order))
262 				continue;	/* skip*/
263 			save = *sipp;
264 			*sipp = *xipp;
265 			*xipp = save;
266 		}
267 	}
268 
269 #if defined(VERBOSE_SYSINIT)
270 	last = SI_SUB_COPYRIGHT;
271 	verbose = 0;
272 #if !defined(DDB)
273 	printf("VERBOSE_SYSINIT: DDB not enabled, symbol lookups disabled.\n");
274 #endif
275 #endif
276 
277 	/*
278 	 * Traverse the (now) ordered list of system initialization tasks.
279 	 * Perform each task, and continue on to the next task.
280 	 */
281 	for (sipp = sysinit; sipp < sysinit_end; sipp++) {
282 		if ((*sipp)->subsystem == SI_SUB_DUMMY)
283 			continue;	/* skip dummy task(s)*/
284 
285 		if ((*sipp)->subsystem == SI_SUB_DONE)
286 			continue;
287 
288 #if defined(VERBOSE_SYSINIT)
289 		if ((*sipp)->subsystem > last && verbose_sysinit != 0) {
290 			verbose = 1;
291 			last = (*sipp)->subsystem;
292 			printf("subsystem %x\n", last);
293 		}
294 		if (verbose) {
295 #if defined(DDB)
296 			const char *func, *data;
297 
298 			func = symbol_name((vm_offset_t)(*sipp)->func,
299 			    DB_STGY_PROC);
300 			data = symbol_name((vm_offset_t)(*sipp)->udata,
301 			    DB_STGY_ANY);
302 			if (func != NULL && data != NULL)
303 				printf("   %s(&%s)... ", func, data);
304 			else if (func != NULL)
305 				printf("   %s(%p)... ", func, (*sipp)->udata);
306 			else
307 #endif
308 				printf("   %p(%p)... ", (*sipp)->func,
309 				    (*sipp)->udata);
310 		}
311 #endif
312 
313 		/* Call function */
314 		(*((*sipp)->func))((*sipp)->udata);
315 
316 #if defined(VERBOSE_SYSINIT)
317 		if (verbose)
318 			printf("done.\n");
319 #endif
320 
321 		/* Check off the one we're just done */
322 		(*sipp)->subsystem = SI_SUB_DONE;
323 
324 		/* Check if we've installed more sysinit items via KLD */
325 		if (newsysinit != NULL) {
326 			if (sysinit != SET_BEGIN(sysinit_set))
327 				free(sysinit, M_TEMP);
328 			sysinit = newsysinit;
329 			sysinit_end = newsysinit_end;
330 			newsysinit = NULL;
331 			newsysinit_end = NULL;
332 			goto restart;
333 		}
334 	}
335 
336 	TSEXIT();	/* Here so we don't overlap with start_init. */
337 
338 	mtx_assert(&Giant, MA_OWNED | MA_NOTRECURSED);
339 	mtx_unlock(&Giant);
340 
341 	/*
342 	 * Now hand over this thread to swapper.
343 	 */
344 	swapper();
345 	/* NOTREACHED*/
346 }
347 
348 static void
print_caddr_t(void * data)349 print_caddr_t(void *data)
350 {
351 	printf("%s", (char *)data);
352 }
353 
354 static void
print_version(void * data __unused)355 print_version(void *data __unused)
356 {
357 	int len;
358 
359 	/* Strip a trailing newline from version. */
360 	len = strlen(version);
361 	while (len > 0 && version[len - 1] == '\n')
362 		len--;
363 	printf("%.*s %s\n", len, version, machine);
364 	printf("%s\n", compiler_version);
365 }
366 
367 SYSINIT(announce, SI_SUB_COPYRIGHT, SI_ORDER_FIRST, print_caddr_t,
368     copyright);
369 SYSINIT(trademark, SI_SUB_COPYRIGHT, SI_ORDER_SECOND, print_caddr_t,
370     trademark);
371 SYSINIT(version, SI_SUB_COPYRIGHT, SI_ORDER_THIRD, print_version, NULL);
372 
373 #ifdef WITNESS
374 static char wit_warn[] =
375      "WARNING: WITNESS option enabled, expect reduced performance.\n";
376 SYSINIT(witwarn, SI_SUB_COPYRIGHT, SI_ORDER_FOURTH,
377    print_caddr_t, wit_warn);
378 SYSINIT(witwarn2, SI_SUB_LAST, SI_ORDER_FOURTH,
379    print_caddr_t, wit_warn);
380 #endif
381 
382 #ifdef DIAGNOSTIC
383 static char diag_warn[] =
384      "WARNING: DIAGNOSTIC option enabled, expect reduced performance.\n";
385 SYSINIT(diagwarn, SI_SUB_COPYRIGHT, SI_ORDER_FIFTH,
386     print_caddr_t, diag_warn);
387 SYSINIT(diagwarn2, SI_SUB_LAST, SI_ORDER_FIFTH,
388     print_caddr_t, diag_warn);
389 #endif
390 
391 static int
null_fetch_syscall_args(struct thread * td __unused)392 null_fetch_syscall_args(struct thread *td __unused)
393 {
394 
395 	panic("null_fetch_syscall_args");
396 }
397 
398 static void
null_set_syscall_retval(struct thread * td __unused,int error __unused)399 null_set_syscall_retval(struct thread *td __unused, int error __unused)
400 {
401 
402 	panic("null_set_syscall_retval");
403 }
404 
405 struct sysentvec null_sysvec = {
406 	.sv_size	= 0,
407 	.sv_table	= NULL,
408 	.sv_transtrap	= NULL,
409 	.sv_fixup	= NULL,
410 	.sv_sendsig	= NULL,
411 	.sv_sigcode	= NULL,
412 	.sv_szsigcode	= NULL,
413 	.sv_name	= "null",
414 	.sv_coredump	= NULL,
415 	.sv_imgact_try	= NULL,
416 	.sv_minsigstksz	= 0,
417 	.sv_minuser	= VM_MIN_ADDRESS,
418 	.sv_maxuser	= VM_MAXUSER_ADDRESS,
419 	.sv_usrstack	= USRSTACK,
420 	.sv_psstrings	= PS_STRINGS,
421 	.sv_psstringssz	= sizeof(struct ps_strings),
422 	.sv_stackprot	= VM_PROT_ALL,
423 	.sv_copyout_strings	= NULL,
424 	.sv_setregs	= NULL,
425 	.sv_fixlimit	= NULL,
426 	.sv_maxssiz	= NULL,
427 	.sv_flags	= 0,
428 	.sv_set_syscall_retval = null_set_syscall_retval,
429 	.sv_fetch_syscall_args = null_fetch_syscall_args,
430 	.sv_syscallnames = NULL,
431 	.sv_schedtail	= NULL,
432 	.sv_thread_detach = NULL,
433 	.sv_trap	= NULL,
434 };
435 
436 /*
437  * The two following SYSINIT's are proc0 specific glue code.  I am not
438  * convinced that they can not be safely combined, but their order of
439  * operation has been maintained as the same as the original init_main.c
440  * for right now.
441  */
442 /* ARGSUSED*/
443 static void
proc0_init(void * dummy __unused)444 proc0_init(void *dummy __unused)
445 {
446 	struct proc *p;
447 	struct thread *td;
448 	struct ucred *newcred;
449 	struct uidinfo tmpuinfo;
450 	struct loginclass tmplc = {
451 		.lc_name = "",
452 	};
453 	vm_paddr_t pageablemem;
454 	int i;
455 
456 	GIANT_REQUIRED;
457 	p = &proc0;
458 	td = &thread0;
459 
460 	/*
461 	 * Initialize magic number and osrel.
462 	 */
463 	p->p_magic = P_MAGIC;
464 	p->p_osrel = osreldate;
465 
466 	/*
467 	 * Initialize thread and process structures.
468 	 */
469 	procinit();	/* set up proc zone */
470 	threadinit();	/* set up UMA zones */
471 
472 	/*
473 	 * Initialise scheduler resources.
474 	 * Add scheduler specific parts to proc, thread as needed.
475 	 */
476 	schedinit();	/* scheduler gets its house in order */
477 
478 	/*
479 	 * Create process 0 (the swapper).
480 	 */
481 	LIST_INSERT_HEAD(&allproc, p, p_list);
482 	LIST_INSERT_HEAD(PIDHASH(0), p, p_hash);
483 	mtx_init(&pgrp0.pg_mtx, "process group", NULL, MTX_DEF | MTX_DUPOK);
484 	p->p_pgrp = &pgrp0;
485 	LIST_INSERT_HEAD(PGRPHASH(0), &pgrp0, pg_hash);
486 	LIST_INIT(&pgrp0.pg_members);
487 	LIST_INSERT_HEAD(&pgrp0.pg_members, p, p_pglist);
488 
489 	pgrp0.pg_session = &session0;
490 	mtx_init(&session0.s_mtx, "session", NULL, MTX_DEF);
491 	refcount_init(&session0.s_count, 1);
492 	session0.s_leader = p;
493 
494 	p->p_sysent = &null_sysvec;
495 	p->p_flag = P_SYSTEM | P_INMEM | P_KPROC;
496 	p->p_flag2 = 0;
497 	p->p_state = PRS_NORMAL;
498 	p->p_klist = knlist_alloc(&p->p_mtx);
499 	STAILQ_INIT(&p->p_ktr);
500 	p->p_nice = NZERO;
501 	td->td_tid = THREAD0_TID;
502 	tidhash_add(td);
503 	td->td_state = TDS_RUNNING;
504 	td->td_pri_class = PRI_TIMESHARE;
505 	td->td_user_pri = PUSER;
506 	td->td_base_user_pri = PUSER;
507 	td->td_lend_user_pri = PRI_MAX;
508 	td->td_priority = PVM;
509 	td->td_base_pri = PVM;
510 	td->td_oncpu = curcpu;
511 	td->td_flags = TDF_INMEM;
512 	td->td_pflags = TDP_KTHREAD;
513 	td->td_cpuset = cpuset_thread0();
514 	td->td_domain.dr_policy = td->td_cpuset->cs_domain;
515 	prison0_init();
516 	p->p_peers = 0;
517 	p->p_leader = p;
518 	p->p_reaper = p;
519 	p->p_treeflag |= P_TREE_REAPER;
520 	LIST_INIT(&p->p_reaplist);
521 
522 	strncpy(p->p_comm, "kernel", sizeof (p->p_comm));
523 	strncpy(td->td_name, "swapper", sizeof (td->td_name));
524 
525 	callout_init_mtx(&p->p_itcallout, &p->p_mtx, 0);
526 	callout_init_mtx(&p->p_limco, &p->p_mtx, 0);
527 	callout_init(&td->td_slpcallout, 1);
528 	TAILQ_INIT(&p->p_kqtim_stop);
529 
530 	/* Create credentials. */
531 	newcred = crget();
532 	newcred->cr_ngroups = 1;	/* group 0 */
533 	/* A hack to prevent uifind from tripping over NULL pointers. */
534 	curthread->td_ucred = newcred;
535 	tmpuinfo.ui_uid = 1;
536 	newcred->cr_uidinfo = newcred->cr_ruidinfo = &tmpuinfo;
537 	newcred->cr_uidinfo = uifind(0);
538 	newcred->cr_ruidinfo = uifind(0);
539 	newcred->cr_loginclass = &tmplc;
540 	newcred->cr_loginclass = loginclass_find("default");
541 	/* End hack. creds get properly set later with thread_cow_get_proc */
542 	curthread->td_ucred = NULL;
543 	newcred->cr_prison = &prison0;
544 	newcred->cr_users++; /* avoid assertion failure */
545 	proc_set_cred_init(p, newcred);
546 	newcred->cr_users--;
547 	crfree(newcred);
548 #ifdef AUDIT
549 	audit_cred_kproc0(newcred);
550 #endif
551 #ifdef MAC
552 	mac_cred_create_swapper(newcred);
553 #endif
554 	/* Create sigacts. */
555 	p->p_sigacts = sigacts_alloc();
556 
557 	/* Initialize signal state for process 0. */
558 	siginit(&proc0);
559 
560 	/* Create the file descriptor table. */
561 	p->p_pd = pdinit(NULL, false);
562 	p->p_fd = fdinit(NULL, false, NULL);
563 	p->p_fdtol = NULL;
564 
565 	/* Create the limits structures. */
566 	p->p_limit = lim_alloc();
567 	for (i = 0; i < RLIM_NLIMITS; i++)
568 		p->p_limit->pl_rlimit[i].rlim_cur =
569 		    p->p_limit->pl_rlimit[i].rlim_max = RLIM_INFINITY;
570 	p->p_limit->pl_rlimit[RLIMIT_NOFILE].rlim_cur =
571 	    p->p_limit->pl_rlimit[RLIMIT_NOFILE].rlim_max = maxfiles;
572 	p->p_limit->pl_rlimit[RLIMIT_NPROC].rlim_cur =
573 	    p->p_limit->pl_rlimit[RLIMIT_NPROC].rlim_max = maxproc;
574 	p->p_limit->pl_rlimit[RLIMIT_DATA].rlim_cur = dfldsiz;
575 	p->p_limit->pl_rlimit[RLIMIT_DATA].rlim_max = maxdsiz;
576 	p->p_limit->pl_rlimit[RLIMIT_STACK].rlim_cur = dflssiz;
577 	p->p_limit->pl_rlimit[RLIMIT_STACK].rlim_max = maxssiz;
578 	/* Cast to avoid overflow on i386/PAE. */
579 	pageablemem = ptoa((vm_paddr_t)vm_free_count());
580 	p->p_limit->pl_rlimit[RLIMIT_RSS].rlim_cur =
581 	    p->p_limit->pl_rlimit[RLIMIT_RSS].rlim_max = pageablemem;
582 	p->p_limit->pl_rlimit[RLIMIT_MEMLOCK].rlim_cur = pageablemem / 3;
583 	p->p_limit->pl_rlimit[RLIMIT_MEMLOCK].rlim_max = pageablemem;
584 	p->p_cpulimit = RLIM_INFINITY;
585 
586 	PROC_LOCK(p);
587 	thread_cow_get_proc(td, p);
588 	PROC_UNLOCK(p);
589 
590 	/* Initialize resource accounting structures. */
591 	racct_create(&p->p_racct);
592 
593 	p->p_stats = pstats_alloc();
594 
595 	/* Allocate a prototype map so we have something to fork. */
596 	p->p_vmspace = &vmspace0;
597 	refcount_init(&vmspace0.vm_refcnt, 1);
598 	pmap_pinit0(vmspace_pmap(&vmspace0));
599 
600 	/*
601 	 * proc0 is not expected to enter usermode, so there is no special
602 	 * handling for sv_minuser here, like is done for exec_new_vmspace().
603 	 */
604 	vm_map_init(&vmspace0.vm_map, vmspace_pmap(&vmspace0),
605 	    p->p_sysent->sv_minuser, p->p_sysent->sv_maxuser);
606 
607 	/*
608 	 * Call the init and ctor for the new thread and proc.  We wait
609 	 * to do this until all other structures are fairly sane.
610 	 */
611 	EVENTHANDLER_DIRECT_INVOKE(process_init, p);
612 	EVENTHANDLER_DIRECT_INVOKE(thread_init, td);
613 #ifdef KDTRACE_HOOKS
614 	kdtrace_proc_ctor(p);
615 	kdtrace_thread_ctor(td);
616 #endif
617 	EVENTHANDLER_DIRECT_INVOKE(process_ctor, p);
618 	EVENTHANDLER_DIRECT_INVOKE(thread_ctor, td);
619 
620 	/*
621 	 * Charge root for one process.
622 	 */
623 	(void)chgproccnt(p->p_ucred->cr_ruidinfo, 1, 0);
624 	PROC_LOCK(p);
625 	racct_add_force(p, RACCT_NPROC, 1);
626 	PROC_UNLOCK(p);
627 }
628 SYSINIT(p0init, SI_SUB_INTRINSIC, SI_ORDER_FIRST, proc0_init, NULL);
629 
630 /* ARGSUSED*/
631 static void
proc0_post(void * dummy __unused)632 proc0_post(void *dummy __unused)
633 {
634 	struct proc *p;
635 	struct rusage ru;
636 	struct thread *td;
637 
638 	/*
639 	 * Now we can look at the time, having had a chance to verify the
640 	 * time from the filesystem.  Pretend that proc0 started now.
641 	 */
642 	sx_slock(&allproc_lock);
643 	FOREACH_PROC_IN_SYSTEM(p) {
644 		PROC_LOCK(p);
645 		if (p->p_state == PRS_NEW) {
646 			PROC_UNLOCK(p);
647 			continue;
648 		}
649 		microuptime(&p->p_stats->p_start);
650 		PROC_STATLOCK(p);
651 		rufetch(p, &ru);	/* Clears thread stats */
652 		p->p_rux.rux_runtime = 0;
653 		p->p_rux.rux_uticks = 0;
654 		p->p_rux.rux_sticks = 0;
655 		p->p_rux.rux_iticks = 0;
656 		PROC_STATUNLOCK(p);
657 		FOREACH_THREAD_IN_PROC(p, td) {
658 			td->td_runtime = 0;
659 		}
660 		PROC_UNLOCK(p);
661 	}
662 	sx_sunlock(&allproc_lock);
663 	PCPU_SET(switchtime, cpu_ticks());
664 	PCPU_SET(switchticks, ticks);
665 }
666 SYSINIT(p0post, SI_SUB_INTRINSIC_POST, SI_ORDER_FIRST, proc0_post, NULL);
667 
668 /*
669  ***************************************************************************
670  ****
671  **** The following SYSINIT's and glue code should be moved to the
672  **** respective files on a per subsystem basis.
673  ****
674  ***************************************************************************
675  */
676 
677 /*
678  * List of paths to try when searching for "init".
679  */
680 static char init_path[MAXPATHLEN] =
681 #ifdef	INIT_PATH
682     __XSTRING(INIT_PATH);
683 #else
684     "/sbin/init:/sbin/oinit:/sbin/init.bak:/rescue/init";
685 #endif
686 SYSCTL_STRING(_kern, OID_AUTO, init_path, CTLFLAG_RD, init_path, 0,
687 	"Path used to search the init process");
688 
689 /*
690  * Shutdown timeout of init(8).
691  * Unused within kernel, but used to control init(8), hence do not remove.
692  */
693 #ifndef INIT_SHUTDOWN_TIMEOUT
694 #define INIT_SHUTDOWN_TIMEOUT 120
695 #endif
696 static int init_shutdown_timeout = INIT_SHUTDOWN_TIMEOUT;
697 SYSCTL_INT(_kern, OID_AUTO, init_shutdown_timeout,
698 	CTLFLAG_RW, &init_shutdown_timeout, 0, "Shutdown timeout of init(8). "
699 	"Unused within kernel, but used to control init(8)");
700 
701 /*
702  * Start the initial user process; try exec'ing each pathname in init_path.
703  * The program is invoked with one argument containing the boot flags.
704  */
705 static void
start_init(void * dummy)706 start_init(void *dummy)
707 {
708 	struct image_args args;
709 	int error;
710 	char *var, *path;
711 	char *free_init_path, *tmp_init_path;
712 	struct thread *td;
713 	struct proc *p;
714 	struct vmspace *oldvmspace;
715 
716 	TSENTER();	/* Here so we don't overlap with mi_startup. */
717 
718 	td = curthread;
719 	p = td->td_proc;
720 
721 	vfs_mountroot();
722 
723 	/* Wipe GELI passphrase from the environment. */
724 	kern_unsetenv("kern.geom.eli.passphrase");
725 
726 	/* For Multicons, report which console is primary to both */
727 	if (boothowto & RB_MULTIPLE) {
728 		if (boothowto & RB_SERIAL)
729 			printf("Dual Console: Serial Primary, Video Secondary\n");
730 		else
731 			printf("Dual Console: Video Primary, Serial Secondary\n");
732 	}
733 
734 	if ((var = kern_getenv("init_path")) != NULL) {
735 		strlcpy(init_path, var, sizeof(init_path));
736 		freeenv(var);
737 	}
738 	free_init_path = tmp_init_path = strdup(init_path, M_TEMP);
739 
740 	while ((path = strsep(&tmp_init_path, ":")) != NULL) {
741 		if (bootverbose)
742 			printf("start_init: trying %s\n", path);
743 
744 		memset(&args, 0, sizeof(args));
745 		error = exec_alloc_args(&args);
746 		if (error != 0)
747 			panic("%s: Can't allocate space for init arguments %d",
748 			    __func__, error);
749 
750 		error = exec_args_add_fname(&args, path, UIO_SYSSPACE);
751 		if (error != 0)
752 			panic("%s: Can't add fname %d", __func__, error);
753 		error = exec_args_add_arg(&args, path, UIO_SYSSPACE);
754 		if (error != 0)
755 			panic("%s: Can't add argv[0] %d", __func__, error);
756 		if (boothowto & RB_SINGLE)
757 			error = exec_args_add_arg(&args, "-s", UIO_SYSSPACE);
758 		if (error != 0)
759 			panic("%s: Can't add argv[0] %d", __func__, error);
760 
761 		/*
762 		 * Now try to exec the program.  If can't for any reason
763 		 * other than it doesn't exist, complain.
764 		 *
765 		 * Otherwise, return via fork_trampoline() all the way
766 		 * to user mode as init!
767 		 */
768 		KASSERT((td->td_pflags & TDP_EXECVMSPC) == 0,
769 		    ("nested execve"));
770 		oldvmspace = p->p_vmspace;
771 		error = kern_execve(td, &args, NULL, oldvmspace);
772 		KASSERT(error != 0,
773 		    ("kern_execve returned success, not EJUSTRETURN"));
774 		if (error == EJUSTRETURN) {
775 			exec_cleanup(td, oldvmspace);
776 			free(free_init_path, M_TEMP);
777 			TSEXIT();
778 			return;
779 		}
780 		if (error != ENOENT)
781 			printf("exec %s: error %d\n", path, error);
782 	}
783 	free(free_init_path, M_TEMP);
784 	printf("init: not found in path %s\n", init_path);
785 	panic("no init");
786 }
787 
788 /*
789  * Like kproc_create(), but runs in its own address space.  We do this
790  * early to reserve pid 1.  Note special case - do not make it
791  * runnable yet, init execution is started when userspace can be served.
792  */
793 static void
create_init(const void * udata __unused)794 create_init(const void *udata __unused)
795 {
796 	struct fork_req fr;
797 	struct ucred *newcred, *oldcred;
798 	struct thread *td;
799 	int error;
800 
801 	bzero(&fr, sizeof(fr));
802 	fr.fr_flags = RFFDG | RFPROC | RFSTOPPED;
803 	fr.fr_procp = &initproc;
804 	error = fork1(&thread0, &fr);
805 	if (error)
806 		panic("cannot fork init: %d\n", error);
807 	KASSERT(initproc->p_pid == 1, ("create_init: initproc->p_pid != 1"));
808 	/* divorce init's credentials from the kernel's */
809 	newcred = crget();
810 	sx_xlock(&proctree_lock);
811 	PROC_LOCK(initproc);
812 	initproc->p_flag |= P_SYSTEM | P_INMEM;
813 	initproc->p_treeflag |= P_TREE_REAPER;
814 	oldcred = initproc->p_ucred;
815 	crcopy(newcred, oldcred);
816 #ifdef MAC
817 	mac_cred_create_init(newcred);
818 #endif
819 #ifdef AUDIT
820 	audit_cred_proc1(newcred);
821 #endif
822 	proc_set_cred(initproc, newcred);
823 	td = FIRST_THREAD_IN_PROC(initproc);
824 	crcowfree(td);
825 	td->td_realucred = crcowget(initproc->p_ucred);
826 	td->td_ucred = td->td_realucred;
827 	PROC_UNLOCK(initproc);
828 	sx_xunlock(&proctree_lock);
829 	crfree(oldcred);
830 	cpu_fork_kthread_handler(FIRST_THREAD_IN_PROC(initproc),
831 	    start_init, NULL);
832 }
833 SYSINIT(init, SI_SUB_CREATE_INIT, SI_ORDER_FIRST, create_init, NULL);
834 
835 /*
836  * Make it runnable now.
837  */
838 static void
kick_init(const void * udata __unused)839 kick_init(const void *udata __unused)
840 {
841 	struct thread *td;
842 
843 	td = FIRST_THREAD_IN_PROC(initproc);
844 	thread_lock(td);
845 	TD_SET_CAN_RUN(td);
846 	sched_add(td, SRQ_BORING);
847 }
848 SYSINIT(kickinit, SI_SUB_KTHREAD_INIT, SI_ORDER_MIDDLE, kick_init, NULL);
849 
850 /*
851  * DDB(4).
852  */
853 #ifdef DDB
854 static void
db_show_print_syinit(struct sysinit * sip,bool ddb)855 db_show_print_syinit(struct sysinit *sip, bool ddb)
856 {
857 	const char *sname, *funcname;
858 	c_db_sym_t sym;
859 	db_expr_t  offset;
860 
861 #define xprint(...)							\
862 	if (ddb)							\
863 		db_printf(__VA_ARGS__);					\
864 	else								\
865 		printf(__VA_ARGS__)
866 
867 	if (sip == NULL) {
868 		xprint("%s: no sysinit * given\n", __func__);
869 		return;
870 	}
871 
872 	sym = db_search_symbol((vm_offset_t)sip, DB_STGY_ANY, &offset);
873 	db_symbol_values(sym, &sname, NULL);
874 	sym = db_search_symbol((vm_offset_t)sip->func, DB_STGY_PROC, &offset);
875 	db_symbol_values(sym, &funcname, NULL);
876 	xprint("%s(%p)\n", (sname != NULL) ? sname : "", sip);
877 	xprint("  %#08x %#08x\n", sip->subsystem, sip->order);
878 	xprint("  %p(%s)(%p)\n",
879 	    sip->func, (funcname != NULL) ? funcname : "", sip->udata);
880 #undef xprint
881 }
882 
DB_SHOW_COMMAND(sysinit,db_show_sysinit)883 DB_SHOW_COMMAND(sysinit, db_show_sysinit)
884 {
885 	struct sysinit **sipp;
886 
887 	db_printf("SYSINIT vs Name(Ptr)\n");
888 	db_printf("  Subsystem  Order\n");
889 	db_printf("  Function(Name)(Arg)\n");
890 	for (sipp = sysinit; sipp < sysinit_end; sipp++) {
891 		db_show_print_syinit(*sipp, true);
892 		if (db_pager_quit)
893 			break;
894 	}
895 }
896 #endif /* DDB */
897