xref: /freebsd/sys/kern/init_main.c (revision 83d98828867591b5d842573ed6edcec7392f82df)
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 	.t0st_thread = {
109 		/*
110 		 * thread0.td_pflags is set with TDP_NOFAULTING to
111 		 * short-cut the vm page fault handler until it is
112 		 * ready.  It is cleared in vm_init() after VM
113 		 * initialization.
114 		 */
115 		.td_pflags = TDP_NOFAULTING,
116 	},
117 };
118 struct	vmspace vmspace0;
119 struct	proc *initproc;
120 
121 int
linux_alloc_current_noop(struct thread * td __unused,int flags __unused)122 linux_alloc_current_noop(struct thread *td __unused, int flags __unused)
123 {
124 	return (0);
125 }
126 int (*lkpi_alloc_current)(struct thread *, int) = linux_alloc_current_noop;
127 
128 #ifndef BOOTHOWTO
129 #define	BOOTHOWTO	0
130 #endif
131 int	boothowto = BOOTHOWTO;	/* initialized so that it can be patched */
132 SYSCTL_INT(_debug, OID_AUTO, boothowto, CTLFLAG_RD, &boothowto, 0,
133 	"Boot control flags, passed from loader");
134 
135 #ifndef BOOTVERBOSE
136 #define	BOOTVERBOSE	0
137 #endif
138 int	bootverbose = BOOTVERBOSE;
139 SYSCTL_INT(_debug, OID_AUTO, bootverbose, CTLFLAG_RW, &bootverbose, 0,
140 	"Control the output of verbose kernel messages");
141 
142 #ifdef VERBOSE_SYSINIT
143 /*
144  * We'll use the defined value of VERBOSE_SYSINIT from the kernel config to
145  * dictate the default VERBOSE_SYSINIT behavior.  Significant values for this
146  * option and associated tunable are:
147  * - 0, 'compiled in but silent by default'
148  * - 1, 'compiled in but verbose by default' (default)
149  */
150 int	verbose_sysinit = VERBOSE_SYSINIT;
151 #endif
152 
153 #ifdef INVARIANTS
154 FEATURE(invariants, "Kernel compiled with INVARIANTS, may affect performance");
155 #endif
156 
157 /*
158  * The sysinit linker set compiled into the kernel.  These are placed onto the
159  * sysinit list by mi_startup; sysinit_add can add (e.g., from klds) additional
160  * sysinits to the linked list but the linker set here does not change.
161  */
162 SET_DECLARE(sysinit_set, struct sysinit);
163 
164 /*
165  * The sysinit lists.  Items are moved to sysinit_done_list when done.
166  */
167 static STAILQ_HEAD(sysinitlist, sysinit) sysinit_list;
168 static struct sysinitlist sysinit_done_list =
169     STAILQ_HEAD_INITIALIZER(sysinit_done_list);
170 
171 /*
172  * Compare two sysinits; return -1, 0, or 1 if a comes before, at the same time
173  * as, or after b.
174  */
175 static int
sysinit_compar(struct sysinit * a,struct sysinit * b,void * thunk __unused)176 sysinit_compar(struct sysinit *a, struct sysinit *b, void *thunk __unused)
177 {
178 
179 	if (a->subsystem < b->subsystem)
180 		return (-1);
181 	if (a->subsystem > b->subsystem)
182 		return (1);
183 	if (a->order < b->order)
184 		return (-1);
185 	if (a->order > b->order)
186 		return (1);
187 	return (0);
188 }
189 
190 static void
sysinit_mklist(struct sysinitlist * list,struct sysinit ** set,struct sysinit ** set_end)191 sysinit_mklist(struct sysinitlist *list, struct sysinit **set,
192     struct sysinit **set_end)
193 {
194 	struct sysinit **sipp;
195 
196 	TSENTER();
197 	TSENTER2("listify");
198 	STAILQ_INIT(list);
199 	for (sipp = set; sipp < set_end; sipp++)
200 		STAILQ_INSERT_TAIL(list, *sipp, next);
201 	TSEXIT2("listify");
202 	TSENTER2("mergesort");
203 	STAILQ_MERGESORT(list, NULL, sysinit_compar, sysinit, next);
204 	TSEXIT2("mergesort");
205 	TSEXIT();
206 }
207 
208 /*
209  * Merge a new sysinit set into the sysinit list.
210  */
211 void
sysinit_add(struct sysinit ** set,struct sysinit ** set_end)212 sysinit_add(struct sysinit **set, struct sysinit **set_end)
213 {
214 	struct sysinitlist new_list;
215 
216 	TSENTER();
217 
218 	/* Construct a sorted list from the new sysinits. */
219 	sysinit_mklist(&new_list, set, set_end);
220 
221 	/* Merge the new list into the existing one. */
222 	TSENTER2("STAILQ_MERGE");
223 	STAILQ_MERGE(&sysinit_list, &new_list, NULL, sysinit_compar, sysinit, next);
224 	TSEXIT2("STAILQ_MERGE");
225 
226 	TSEXIT();
227 }
228 
229 #if defined (DDB) && defined(VERBOSE_SYSINIT)
230 static const char *
symbol_name(vm_offset_t va,db_strategy_t strategy)231 symbol_name(vm_offset_t va, db_strategy_t strategy)
232 {
233 	const char *name;
234 	c_db_sym_t sym;
235 	db_expr_t  offset;
236 
237 	if (va == 0)
238 		return (NULL);
239 	sym = db_search_symbol(va, strategy, &offset);
240 	if (offset != 0)
241 		return (NULL);
242 	db_symbol_values(sym, &name, NULL);
243 	return (name);
244 }
245 #endif
246 
247 /*
248  * System startup; initialize the world, create process 0, mount root
249  * filesystem, and fork to create init and pagedaemon.  Most of the
250  * hard work is done in the lower-level initialization routines including
251  * startup(), which does memory initialization and autoconfiguration.
252  *
253  * This allows simple addition of new kernel subsystems that require
254  * boot time initialization.  It also allows substitution of subsystem
255  * (for instance, a scheduler, kernel profiler, or VM system) by object
256  * module.  Finally, it allows for optional "kernel threads".
257  */
258 void
mi_startup(void)259 mi_startup(void)
260 {
261 	struct sysinit *sip;
262 	int last;
263 #if defined(VERBOSE_SYSINIT)
264 	int verbose;
265 #endif
266 
267 	TSENTER();
268 
269 	if (boothowto & RB_VERBOSE)
270 		bootverbose++;
271 
272 	/* Construct and sort sysinit list. */
273 	sysinit_mklist(&sysinit_list, SET_BEGIN(sysinit_set), SET_LIMIT(sysinit_set));
274 
275 	last = SI_SUB_DUMMY;
276 #if defined(VERBOSE_SYSINIT)
277 	TUNABLE_INT_FETCH("debug.verbose_sysinit", &verbose_sysinit);
278 	verbose = 0;
279 #if !defined(DDB)
280 	printf("VERBOSE_SYSINIT: DDB not enabled, symbol lookups disabled.\n");
281 #endif
282 #endif
283 
284 	/*
285 	 * Perform each system initialization task from the ordered list.  Note
286 	 * that if sysinit_list is modified (e.g. by a KLD) we will nonetheless
287 	 * always perform the earlist-sorted sysinit at each step; using the
288 	 * STAILQ_FOREACH macro would result in items being skipped if inserted
289 	 * earlier than the "current item".
290 	 */
291 	while ((sip = STAILQ_FIRST(&sysinit_list)) != NULL) {
292 		STAILQ_REMOVE_HEAD(&sysinit_list, next);
293 		STAILQ_INSERT_TAIL(&sysinit_done_list, sip, next);
294 
295 		if (sip->subsystem == SI_SUB_DUMMY)
296 			continue;	/* skip dummy task(s)*/
297 
298 		if (sip->subsystem > last)
299 			BOOTTRACE_INIT("sysinit 0x%7x", sip->subsystem);
300 
301 #if defined(VERBOSE_SYSINIT)
302 		if (sip->subsystem != last && verbose_sysinit != 0) {
303 			verbose = 1;
304 			printf("subsystem %x\n", sip->subsystem);
305 		}
306 		if (verbose) {
307 #if defined(DDB)
308 			const char *func, *data;
309 
310 			func = symbol_name((vm_offset_t)sip->func,
311 			    DB_STGY_PROC);
312 			data = symbol_name((vm_offset_t)sip->udata,
313 			    DB_STGY_ANY);
314 			if (func != NULL && data != NULL)
315 				printf("   %s(&%s)... ", func, data);
316 			else if (func != NULL)
317 				printf("   %s(%p)... ", func, sip->udata);
318 			else
319 #endif
320 				printf("   %p(%p)... ", sip->func,
321 				    sip->udata);
322 		}
323 #endif
324 
325 		/* Call function */
326 		(*(sip->func))(sip->udata);
327 
328 #if defined(VERBOSE_SYSINIT)
329 		if (verbose)
330 			printf("done.\n");
331 #endif
332 
333 		/* Check off the one we're just done */
334 		last = sip->subsystem;
335 	}
336 
337 	TSEXIT();	/* Here so we don't overlap with start_init. */
338 	BOOTTRACE("mi_startup done");
339 
340 	mtx_assert(&Giant, MA_OWNED | MA_NOTRECURSED);
341 	mtx_unlock(&Giant);
342 
343 	/*
344 	 * We can't free our thread structure since it is statically allocated.
345 	 * Just sleep forever.  This thread could be repurposed for something if
346 	 * the need arises.
347 	 */
348 	for (;;)
349 		tsleep(__builtin_frame_address(0), PNOLOCK, "-", 0);
350 }
351 
352 static void
print_caddr_t(const void * data)353 print_caddr_t(const void *data)
354 {
355 	printf("%s", (const char *)data);
356 }
357 
358 static void
print_version(const void * data __unused)359 print_version(const void *data __unused)
360 {
361 	int len;
362 
363 	/* Strip a trailing newline from version. */
364 	len = strlen(version);
365 	while (len > 0 && version[len - 1] == '\n')
366 		len--;
367 	printf("%.*s %s\n", len, version, machine);
368 	printf("%s\n", compiler_version);
369 }
370 
371 C_SYSINIT(announce, SI_SUB_COPYRIGHT, SI_ORDER_FIRST, print_caddr_t,
372     copyright);
373 C_SYSINIT(trademark, SI_SUB_COPYRIGHT, SI_ORDER_SECOND, print_caddr_t,
374     trademark);
375 C_SYSINIT(version, SI_SUB_COPYRIGHT, SI_ORDER_THIRD, print_version, NULL);
376 
377 #ifdef WITNESS
378 static const char wit_warn[] =
379      "WARNING: WITNESS option enabled, expect reduced performance.\n";
380 C_SYSINIT(witwarn, SI_SUB_COPYRIGHT, SI_ORDER_FOURTH,
381    print_caddr_t, wit_warn);
382 C_SYSINIT(witwarn2, SI_SUB_LAST, SI_ORDER_FOURTH,
383    print_caddr_t, wit_warn);
384 #endif
385 
386 #ifdef DIAGNOSTIC
387 static const char diag_warn[] =
388      "WARNING: DIAGNOSTIC option enabled, expect reduced performance.\n";
389 C_SYSINIT(diagwarn, SI_SUB_COPYRIGHT, SI_ORDER_FIFTH,
390     print_caddr_t, diag_warn);
391 C_SYSINIT(diagwarn2, SI_SUB_LAST, SI_ORDER_FIFTH,
392     print_caddr_t, diag_warn);
393 #endif
394 
395 #if __SIZEOF_LONG__ == 4
396 static const char ilp32_warn[] =
397     "WARNING: 32-bit kernels are deprecated and may be removed in FreeBSD 16.0.\n";
398 C_SYSINIT(ilp32warn, SI_SUB_COPYRIGHT, SI_ORDER_FIFTH,
399     print_caddr_t, ilp32_warn);
400 C_SYSINIT(ilp32warn2, SI_SUB_LAST, SI_ORDER_FIFTH,
401     print_caddr_t, ilp32_warn);
402 #endif
403 
404 static int
null_fetch_syscall_args(struct thread * td __unused)405 null_fetch_syscall_args(struct thread *td __unused)
406 {
407 
408 	panic("null_fetch_syscall_args");
409 }
410 
411 static void
null_set_syscall_retval(struct thread * td __unused,int error __unused)412 null_set_syscall_retval(struct thread *td __unused, int error __unused)
413 {
414 
415 	panic("null_set_syscall_retval");
416 }
417 
418 static void
null_set_fork_retval(struct thread * td __unused)419 null_set_fork_retval(struct thread *td __unused)
420 {
421 
422 }
423 
424 struct sysentvec null_sysvec = {
425 	.sv_size	= 0,
426 	.sv_table	= NULL,
427 	.sv_fixup	= NULL,
428 	.sv_sendsig	= NULL,
429 	.sv_sigcode	= NULL,
430 	.sv_szsigcode	= NULL,
431 	.sv_name	= "null",
432 	.sv_coredump	= NULL,
433 	.sv_minsigstksz	= 0,
434 	.sv_minuser	= VM_MIN_ADDRESS,
435 	.sv_maxuser	= VM_MAXUSER_ADDRESS,
436 	.sv_usrstack	= USRSTACK,
437 	.sv_psstrings	= PS_STRINGS,
438 	.sv_psstringssz	= sizeof(struct ps_strings),
439 	.sv_stackprot	= VM_PROT_ALL,
440 	.sv_copyout_strings	= NULL,
441 	.sv_setregs	= NULL,
442 	.sv_fixlimit	= NULL,
443 	.sv_maxssiz	= NULL,
444 	.sv_flags	= 0,
445 	.sv_set_syscall_retval = null_set_syscall_retval,
446 	.sv_fetch_syscall_args = null_fetch_syscall_args,
447 	.sv_syscallnames = NULL,
448 	.sv_schedtail	= NULL,
449 	.sv_thread_detach = NULL,
450 	.sv_trap	= NULL,
451 	.sv_set_fork_retval = null_set_fork_retval,
452 	.sv_regset_begin = NULL,
453 	.sv_regset_end  = NULL,
454 };
455 
456 /*
457  * The two following SYSINIT's are proc0 specific glue code.  I am not
458  * convinced that they can not be safely combined, but their order of
459  * operation has been maintained as the same as the original init_main.c
460  * for right now.
461  */
462 /* ARGSUSED*/
463 static void
proc0_init(void * dummy __unused)464 proc0_init(void *dummy __unused)
465 {
466 	struct proc *p;
467 	struct thread *td;
468 	struct ucred *newcred;
469 	struct uidinfo tmpuinfo;
470 	struct loginclass tmplc = {
471 		.lc_name = "",
472 	};
473 	vm_paddr_t pageablemem;
474 	int i;
475 
476 	GIANT_REQUIRED;
477 	p = &proc0;
478 	td = &thread0;
479 
480 	/*
481 	 * Initialize magic number and osrel.
482 	 */
483 	p->p_magic = P_MAGIC;
484 	p->p_osrel = osreldate;
485 
486 	/*
487 	 * Initialize thread and process structures.
488 	 */
489 	procinit();	/* set up proc zone */
490 	threadinit();	/* set up UMA zones */
491 
492 	/*
493 	 * Initialise scheduler resources.
494 	 * Add scheduler specific parts to proc, thread as needed.
495 	 */
496 	schedinit();	/* scheduler gets its house in order */
497 
498 	/*
499 	 * Create process 0.
500 	 */
501 	LIST_INSERT_HEAD(&allproc, p, p_list);
502 	LIST_INSERT_HEAD(PIDHASH(0), p, p_hash);
503 	mtx_init(&pgrp0.pg_mtx, "process group", NULL, MTX_DEF | MTX_DUPOK);
504 	sx_init(&pgrp0.pg_killsx, "killpg racer");
505 	p->p_pgrp = &pgrp0;
506 	LIST_INSERT_HEAD(PGRPHASH(0), &pgrp0, pg_hash);
507 	LIST_INIT(&pgrp0.pg_members);
508 	LIST_INSERT_HEAD(&pgrp0.pg_members, p, p_pglist);
509 
510 	pgrp0.pg_session = &session0;
511 	mtx_init(&session0.s_mtx, "session", NULL, MTX_DEF);
512 	refcount_init(&session0.s_count, 1);
513 	session0.s_leader = p;
514 
515 	p->p_sysent = &null_sysvec;
516 	p->p_flag = P_SYSTEM | P_INMEM | P_KPROC;
517 	p->p_flag2 = 0;
518 	p->p_state = PRS_NORMAL;
519 	p->p_klist = knlist_alloc(&p->p_mtx);
520 	STAILQ_INIT(&p->p_ktr);
521 	p->p_nice = NZERO;
522 	td->td_tid = THREAD0_TID;
523 	tidhash_add(td);
524 	TD_SET_STATE(td, TDS_RUNNING);
525 	td->td_pri_class = PRI_TIMESHARE;
526 	td->td_user_pri = PUSER;
527 	td->td_base_user_pri = PUSER;
528 	td->td_lend_user_pri = PRI_MAX;
529 	td->td_priority = PVM;
530 	td->td_base_pri = PVM;
531 	td->td_oncpu = curcpu;
532 	td->td_flags = TDF_INMEM;
533 	td->td_pflags = TDP_KTHREAD;
534 	td->td_cpuset = cpuset_thread0();
535 	td->td_domain.dr_policy = td->td_cpuset->cs_domain;
536 	prison0_init();
537 	p->p_peers = 0;
538 	p->p_leader = p;
539 	p->p_reaper = p;
540 	p->p_treeflag |= P_TREE_REAPER;
541 	LIST_INIT(&p->p_reaplist);
542 
543 	strncpy(p->p_comm, "kernel", sizeof (p->p_comm));
544 	strncpy(td->td_name, "kernel", sizeof (td->td_name));
545 
546 	callout_init_mtx(&p->p_itcallout, &p->p_mtx, 0);
547 	callout_init_mtx(&p->p_limco, &p->p_mtx, 0);
548 	callout_init(&td->td_slpcallout, 1);
549 	TAILQ_INIT(&p->p_kqtim_stop);
550 
551 	/* Create credentials. */
552 	newcred = crget();
553 	newcred->cr_ngroups = 1;	/* group 0 */
554 	/* A hack to prevent uifind from tripping over NULL pointers. */
555 	curthread->td_ucred = newcred;
556 	tmpuinfo.ui_uid = 1;
557 	newcred->cr_uidinfo = newcred->cr_ruidinfo = &tmpuinfo;
558 	newcred->cr_uidinfo = uifind(0);
559 	newcred->cr_ruidinfo = uifind(0);
560 	newcred->cr_loginclass = &tmplc;
561 	newcred->cr_loginclass = loginclass_find("default");
562 	/* End hack. creds get properly set later with thread_cow_get_proc */
563 	curthread->td_ucred = NULL;
564 	newcred->cr_prison = &prison0;
565 	newcred->cr_users++; /* avoid assertion failure */
566 	p->p_ucred = crcowget(newcred);
567 	newcred->cr_users--;
568 	crfree(newcred);
569 #ifdef AUDIT
570 	audit_cred_kproc0(newcred);
571 #endif
572 #ifdef MAC
573 	mac_cred_create_kproc0(newcred);
574 #endif
575 	/* Create sigacts. */
576 	p->p_sigacts = sigacts_alloc();
577 
578 	/* Initialize signal state for process 0. */
579 	siginit(&proc0);
580 
581 	/* Create the file descriptor table. */
582 	p->p_pd = pdinit(NULL, false);
583 	p->p_fd = fdinit();
584 	p->p_fdtol = NULL;
585 
586 	/* Create the limits structures. */
587 	p->p_limit = lim_alloc();
588 	for (i = 0; i < RLIM_NLIMITS; i++)
589 		p->p_limit->pl_rlimit[i].rlim_cur =
590 		    p->p_limit->pl_rlimit[i].rlim_max = RLIM_INFINITY;
591 	p->p_limit->pl_rlimit[RLIMIT_NOFILE].rlim_cur =
592 	    p->p_limit->pl_rlimit[RLIMIT_NOFILE].rlim_max = maxfiles;
593 	p->p_limit->pl_rlimit[RLIMIT_NPROC].rlim_cur =
594 	    p->p_limit->pl_rlimit[RLIMIT_NPROC].rlim_max = maxproc;
595 	p->p_limit->pl_rlimit[RLIMIT_DATA].rlim_cur = dfldsiz;
596 	p->p_limit->pl_rlimit[RLIMIT_DATA].rlim_max = maxdsiz;
597 	p->p_limit->pl_rlimit[RLIMIT_STACK].rlim_cur = dflssiz;
598 	p->p_limit->pl_rlimit[RLIMIT_STACK].rlim_max = maxssiz;
599 	/* Cast to avoid overflow on i386/PAE. */
600 	pageablemem = ptoa((vm_paddr_t)vm_free_count());
601 	p->p_limit->pl_rlimit[RLIMIT_RSS].rlim_cur =
602 	    p->p_limit->pl_rlimit[RLIMIT_RSS].rlim_max = pageablemem;
603 	p->p_limit->pl_rlimit[RLIMIT_MEMLOCK].rlim_cur = pageablemem / 3;
604 	p->p_limit->pl_rlimit[RLIMIT_MEMLOCK].rlim_max = pageablemem;
605 	p->p_cpulimit = RLIM_INFINITY;
606 
607 	PROC_LOCK(p);
608 	thread_cow_get_proc(td, p);
609 	PROC_UNLOCK(p);
610 
611 	/* Initialize resource accounting structures. */
612 	racct_create(&p->p_racct);
613 
614 	p->p_stats = pstats_alloc();
615 
616 	/* Allocate a prototype map so we have something to fork. */
617 	p->p_vmspace = &vmspace0;
618 	refcount_init(&vmspace0.vm_refcnt, 1);
619 	pmap_pinit0(vmspace_pmap(&vmspace0));
620 
621 	/*
622 	 * proc0 is not expected to enter usermode, so there is no special
623 	 * handling for sv_minuser here, like is done for exec_new_vmspace().
624 	 */
625 	vm_map_init(&vmspace0.vm_map, vmspace_pmap(&vmspace0),
626 	    p->p_sysent->sv_minuser, p->p_sysent->sv_maxuser);
627 
628 	/*
629 	 * Call the init and ctor for the new thread and proc.  We wait
630 	 * to do this until all other structures are fairly sane.
631 	 */
632 	EVENTHANDLER_DIRECT_INVOKE(process_init, p);
633 	EVENTHANDLER_DIRECT_INVOKE(thread_init, td);
634 #ifdef KDTRACE_HOOKS
635 	kdtrace_proc_ctor(p);
636 	kdtrace_thread_ctor(td);
637 #endif
638 	EVENTHANDLER_DIRECT_INVOKE(process_ctor, p);
639 	EVENTHANDLER_DIRECT_INVOKE(thread_ctor, td);
640 
641 	/*
642 	 * Charge root for one process.
643 	 */
644 	(void)chgproccnt(p->p_ucred->cr_ruidinfo, 1, 0);
645 	PROC_LOCK(p);
646 	racct_add_force(p, RACCT_NPROC, 1);
647 	PROC_UNLOCK(p);
648 }
649 SYSINIT(p0init, SI_SUB_INTRINSIC, SI_ORDER_FIRST, proc0_init, NULL);
650 
651 /* ARGSUSED*/
652 static void
proc0_post(void * dummy __unused)653 proc0_post(void *dummy __unused)
654 {
655 	struct proc *p;
656 	struct thread *td;
657 
658 	/*
659 	 * Now we can look at the time, having had a chance to verify the
660 	 * time from the filesystem.  Pretend that all current threads
661 	 * started now.
662 	 */
663 	sx_slock(&allproc_lock);
664 	FOREACH_PROC_IN_SYSTEM(p) {
665 		PROC_LOCK(p);
666 		if (p->p_state == PRS_NEW) {
667 			PROC_UNLOCK(p);
668 			continue;
669 		}
670 		microuptime(&p->p_stats->p_start);
671 		PROC_STATLOCK(p);
672 		ruxreset(&p->p_rux);
673 		FOREACH_THREAD_IN_PROC(p, td) {
674 			thread_lock(td);
675 			td->td_incruntime = 0;
676 			td->td_runtime = 0;
677 			td->td_pticks = 0;
678 			td->td_sticks = 0;
679 			td->td_iticks = 0;
680 			td->td_uticks = 0;
681 			ruxreset(&td->td_rux);
682 			thread_unlock(td);
683 		}
684 		PROC_STATUNLOCK(p);
685 		PROC_UNLOCK(p);
686 	}
687 	sx_sunlock(&allproc_lock);
688 	PCPU_SET(switchtime, cpu_ticks());
689 	PCPU_SET(switchticks, ticks);
690 }
691 SYSINIT(p0post, SI_SUB_INTRINSIC_POST, SI_ORDER_FIRST, proc0_post, NULL);
692 
693 /*
694  ***************************************************************************
695  ****
696  **** The following SYSINIT's and glue code should be moved to the
697  **** respective files on a per subsystem basis.
698  ****
699  ***************************************************************************
700  */
701 
702 /*
703  * List of paths to try when searching for "init".
704  */
705 static char init_path[MAXPATHLEN] =
706 #ifdef	INIT_PATH
707     __XSTRING(INIT_PATH);
708 #else
709     "/sbin/init:/sbin/oinit:/sbin/init.bak:/rescue/init";
710 #endif
711 SYSCTL_STRING(_kern, OID_AUTO, init_path, CTLFLAG_RD, init_path, 0,
712 	"Path used to search the init process");
713 
714 /*
715  * Shutdown timeout of init(8).
716  * Unused within kernel, but used to control init(8), hence do not remove.
717  */
718 #ifndef INIT_SHUTDOWN_TIMEOUT
719 #define INIT_SHUTDOWN_TIMEOUT 120
720 #endif
721 static int init_shutdown_timeout = INIT_SHUTDOWN_TIMEOUT;
722 SYSCTL_INT(_kern, OID_AUTO, init_shutdown_timeout,
723 	CTLFLAG_RW, &init_shutdown_timeout, 0, "Shutdown timeout of init(8). "
724 	"Unused within kernel, but used to control init(8)");
725 
726 /*
727  * Start the initial user process; try exec'ing each pathname in init_path.
728  * The program is invoked with one argument containing the boot flags.
729  */
730 static void
start_init(void * dummy)731 start_init(void *dummy)
732 {
733 	struct image_args args;
734 	int error;
735 	char *var, *path;
736 	char *free_init_path, *tmp_init_path;
737 	struct thread *td;
738 	struct proc *p;
739 	struct vmspace *oldvmspace;
740 
741 	TSENTER();	/* Here so we don't overlap with mi_startup. */
742 
743 	td = curthread;
744 	p = td->td_proc;
745 
746 	vfs_mountroot();
747 
748 	/* Wipe GELI passphrase from the environment. */
749 	kern_unsetenv("kern.geom.eli.passphrase");
750 
751 	/* For Multicons, report which console is primary to both */
752 	if (boothowto & RB_MULTIPLE) {
753 		if (boothowto & RB_SERIAL)
754 			printf("Dual Console: Serial Primary, Video Secondary\n");
755 		else
756 			printf("Dual Console: Video Primary, Serial Secondary\n");
757 	}
758 
759 	if ((var = kern_getenv("init_path")) != NULL) {
760 		strlcpy(init_path, var, sizeof(init_path));
761 		freeenv(var);
762 	}
763 	free_init_path = tmp_init_path = strdup(init_path, M_TEMP);
764 
765 	while ((path = strsep(&tmp_init_path, ":")) != NULL) {
766 		if (bootverbose)
767 			printf("start_init: trying %s\n", path);
768 
769 		memset(&args, 0, sizeof(args));
770 		error = exec_alloc_args(&args);
771 		if (error != 0)
772 			panic("%s: Can't allocate space for init arguments %d",
773 			    __func__, error);
774 
775 		error = exec_args_add_fname(&args, path, UIO_SYSSPACE);
776 		if (error != 0)
777 			panic("%s: Can't add fname %d", __func__, error);
778 		error = exec_args_add_arg(&args, path, UIO_SYSSPACE);
779 		if (error != 0)
780 			panic("%s: Can't add argv[0] %d", __func__, error);
781 		if (boothowto & RB_SINGLE)
782 			error = exec_args_add_arg(&args, "-s", UIO_SYSSPACE);
783 		if (error != 0)
784 			panic("%s: Can't add argv[0] %d", __func__, error);
785 
786 		/*
787 		 * Now try to exec the program.  If can't for any reason
788 		 * other than it doesn't exist, complain.
789 		 *
790 		 * Otherwise, return via fork_trampoline() all the way
791 		 * to user mode as init!
792 		 */
793 		KASSERT((td->td_pflags & TDP_EXECVMSPC) == 0,
794 		    ("nested execve"));
795 		memset(td->td_frame, 0, sizeof(*td->td_frame));
796 		oldvmspace = p->p_vmspace;
797 		error = kern_execve(td, &args, NULL, oldvmspace);
798 		KASSERT(error != 0,
799 		    ("kern_execve returned success, not EJUSTRETURN"));
800 		if (error == EJUSTRETURN) {
801 			exec_cleanup(td, oldvmspace);
802 			free(free_init_path, M_TEMP);
803 			TSEXIT();
804 			return;
805 		}
806 		if (error != ENOENT)
807 			printf("exec %s: error %d\n", path, error);
808 	}
809 	free(free_init_path, M_TEMP);
810 	printf("init: not found in path %s\n", init_path);
811 	panic("no init");
812 }
813 
814 /*
815  * Like kproc_create(), but runs in its own address space.  We do this
816  * early to reserve pid 1.  Note special case - do not make it
817  * runnable yet, init execution is started when userspace can be served.
818  */
819 static void
create_init(const void * udata __unused)820 create_init(const void *udata __unused)
821 {
822 	struct fork_req fr;
823 	struct ucred *newcred, *oldcred;
824 	struct thread *td;
825 	int error;
826 
827 	bzero(&fr, sizeof(fr));
828 	fr.fr_flags = RFFDG | RFPROC | RFSTOPPED;
829 	fr.fr_procp = &initproc;
830 	error = fork1(&thread0, &fr);
831 	if (error)
832 		panic("cannot fork init: %d\n", error);
833 	KASSERT(initproc->p_pid == 1, ("create_init: initproc->p_pid != 1"));
834 	/* divorce init's credentials from the kernel's */
835 	newcred = crget();
836 	sx_xlock(&proctree_lock);
837 	PROC_LOCK(initproc);
838 	initproc->p_flag |= P_SYSTEM | P_INMEM;
839 	initproc->p_treeflag |= P_TREE_REAPER;
840 	oldcred = initproc->p_ucred;
841 	crcopy(newcred, oldcred);
842 #ifdef MAC
843 	mac_cred_create_init(newcred);
844 #endif
845 #ifdef AUDIT
846 	audit_cred_proc1(newcred);
847 #endif
848 	proc_set_cred(initproc, newcred);
849 	td = FIRST_THREAD_IN_PROC(initproc);
850 	crcowfree(td);
851 	td->td_realucred = crcowget(initproc->p_ucred);
852 	td->td_ucred = td->td_realucred;
853 	PROC_UNLOCK(initproc);
854 	sx_xunlock(&proctree_lock);
855 	crfree(oldcred);
856 	cpu_fork_kthread_handler(FIRST_THREAD_IN_PROC(initproc),
857 	    start_init, NULL);
858 }
859 SYSINIT(init, SI_SUB_CREATE_INIT, SI_ORDER_FIRST, create_init, NULL);
860 
861 /*
862  * Make it runnable now.
863  */
864 static void
kick_init(const void * udata __unused)865 kick_init(const void *udata __unused)
866 {
867 	struct thread *td;
868 
869 	td = FIRST_THREAD_IN_PROC(initproc);
870 	thread_lock(td);
871 	TD_SET_CAN_RUN(td);
872 	sched_add(td, SRQ_BORING);
873 }
874 SYSINIT(kickinit, SI_SUB_KTHREAD_INIT, SI_ORDER_MIDDLE, kick_init, NULL);
875 
876 /*
877  * DDB(4).
878  */
879 #ifdef DDB
880 static void
db_show_print_syinit(struct sysinit * sip,bool ddb)881 db_show_print_syinit(struct sysinit *sip, bool ddb)
882 {
883 	const char *sname, *funcname;
884 	c_db_sym_t sym;
885 	db_expr_t  offset;
886 
887 #define xprint(...)							\
888 	if (ddb)							\
889 		db_printf(__VA_ARGS__);					\
890 	else								\
891 		printf(__VA_ARGS__)
892 
893 	if (sip == NULL) {
894 		xprint("%s: no sysinit * given\n", __func__);
895 		return;
896 	}
897 
898 	sym = db_search_symbol((vm_offset_t)sip, DB_STGY_ANY, &offset);
899 	db_symbol_values(sym, &sname, NULL);
900 	sym = db_search_symbol((vm_offset_t)sip->func, DB_STGY_PROC, &offset);
901 	db_symbol_values(sym, &funcname, NULL);
902 	xprint("%s(%p)\n", (sname != NULL) ? sname : "", sip);
903 	xprint("  %#08x %#08x\n", sip->subsystem, sip->order);
904 	xprint("  %p(%s)(%p)\n",
905 	    sip->func, (funcname != NULL) ? funcname : "", sip->udata);
906 #undef xprint
907 }
908 
DB_SHOW_COMMAND_FLAGS(sysinit,db_show_sysinit,DB_CMD_MEMSAFE)909 DB_SHOW_COMMAND_FLAGS(sysinit, db_show_sysinit, DB_CMD_MEMSAFE)
910 {
911 	struct sysinit *sip;
912 
913 	db_printf("SYSINIT vs Name(Ptr)\n");
914 	db_printf("  Subsystem  Order\n");
915 	db_printf("  Function(Name)(Arg)\n");
916 	STAILQ_FOREACH(sip, &sysinit_done_list, next) {
917 		db_show_print_syinit(sip, true);
918 		if (db_pager_quit)
919 			return;
920 	}
921 	STAILQ_FOREACH(sip, &sysinit_list, next) {
922 		db_show_print_syinit(sip, true);
923 		if (db_pager_quit)
924 			break;
925 	}
926 }
927 #endif /* DDB */
928