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_init_path.h"
51#include "opt_verbose_sysinit.h"
52
53#include <sys/param.h>
54#include <sys/kernel.h>
55#include <sys/exec.h>
56#include <sys/file.h>
57#include <sys/filedesc.h>
58#include <sys/jail.h>
59#include <sys/ktr.h>
60#include <sys/lock.h>
61#include <sys/loginclass.h>
62#include <sys/mount.h>
63#include <sys/mutex.h>
64#include <sys/syscallsubr.h>
65#include <sys/sysctl.h>
66#include <sys/proc.h>
67#include <sys/racct.h>
68#include <sys/resourcevar.h>
69#include <sys/systm.h>
70#include <sys/signalvar.h>
71#include <sys/vnode.h>
72#include <sys/sysent.h>
73#include <sys/reboot.h>
74#include <sys/sched.h>
75#include <sys/sx.h>
76#include <sys/sysproto.h>
77#include <sys/vmmeter.h>
78#include <sys/unistd.h>
79#include <sys/malloc.h>
80#include <sys/conf.h>
81#include <sys/cpuset.h>
82
83#include <machine/cpu.h>
84
85#include <security/audit/audit.h>
86#include <security/mac/mac_framework.h>
87
88#include <vm/vm.h>
89#include <vm/vm_param.h>
90#include <vm/vm_extern.h>
91#include <vm/pmap.h>
92#include <vm/vm_map.h>
93#include <sys/copyright.h>
94
95#include <ddb/ddb.h>
96#include <ddb/db_sym.h>
97
98void mi_startup(void);				/* Should be elsewhere */
99
100/* Components of the first process -- never freed. */
101static struct session session0;
102static struct pgrp pgrp0;
103struct	proc proc0;
104struct thread0_storage thread0_st __aligned(32);
105struct	vmspace vmspace0;
106struct	proc *initproc;
107
108#ifndef BOOTHOWTO
109#define	BOOTHOWTO	0
110#endif
111int	boothowto = BOOTHOWTO;	/* initialized so that it can be patched */
112SYSCTL_INT(_debug, OID_AUTO, boothowto, CTLFLAG_RD, &boothowto, 0,
113	"Boot control flags, passed from loader");
114
115#ifndef BOOTVERBOSE
116#define	BOOTVERBOSE	0
117#endif
118int	bootverbose = BOOTVERBOSE;
119SYSCTL_INT(_debug, OID_AUTO, bootverbose, CTLFLAG_RW, &bootverbose, 0,
120	"Control the output of verbose kernel messages");
121
122#ifdef VERBOSE_SYSINIT
123/*
124 * We'll use the defined value of VERBOSE_SYSINIT from the kernel config to
125 * dictate the default VERBOSE_SYSINIT behavior.  Significant values for this
126 * option and associated tunable are:
127 * - 0, 'compiled in but silent by default'
128 * - 1, 'compiled in but verbose by default' (default)
129 */
130int	verbose_sysinit = VERBOSE_SYSINIT;
131TUNABLE_INT("debug.verbose_sysinit", &verbose_sysinit);
132#endif
133
134#ifdef INVARIANTS
135FEATURE(invariants, "Kernel compiled with INVARIANTS, may affect performance");
136#endif
137
138/*
139 * This ensures that there is at least one entry so that the sysinit_set
140 * symbol is not undefined.  A sybsystem ID of SI_SUB_DUMMY is never
141 * executed.
142 */
143SYSINIT(placeholder, SI_SUB_DUMMY, SI_ORDER_ANY, NULL, NULL);
144
145/*
146 * The sysinit table itself.  Items are checked off as the are run.
147 * If we want to register new sysinit types, add them to newsysinit.
148 */
149SET_DECLARE(sysinit_set, struct sysinit);
150struct sysinit **sysinit, **sysinit_end;
151struct sysinit **newsysinit, **newsysinit_end;
152
153EVENTHANDLER_LIST_DECLARE(process_init);
154EVENTHANDLER_LIST_DECLARE(thread_init);
155EVENTHANDLER_LIST_DECLARE(process_ctor);
156EVENTHANDLER_LIST_DECLARE(thread_ctor);
157
158/*
159 * Merge a new sysinit set into the current set, reallocating it if
160 * necessary.  This can only be called after malloc is running.
161 */
162void
163sysinit_add(struct sysinit **set, struct sysinit **set_end)
164{
165	struct sysinit **newset;
166	struct sysinit **sipp;
167	struct sysinit **xipp;
168	int count;
169
170	count = set_end - set;
171	if (newsysinit)
172		count += newsysinit_end - newsysinit;
173	else
174		count += sysinit_end - sysinit;
175	newset = malloc(count * sizeof(*sipp), M_TEMP, M_NOWAIT);
176	if (newset == NULL)
177		panic("cannot malloc for sysinit");
178	xipp = newset;
179	if (newsysinit)
180		for (sipp = newsysinit; sipp < newsysinit_end; sipp++)
181			*xipp++ = *sipp;
182	else
183		for (sipp = sysinit; sipp < sysinit_end; sipp++)
184			*xipp++ = *sipp;
185	for (sipp = set; sipp < set_end; sipp++)
186		*xipp++ = *sipp;
187	if (newsysinit)
188		free(newsysinit, M_TEMP);
189	newsysinit = newset;
190	newsysinit_end = newset + count;
191}
192
193#if defined (DDB) && defined(VERBOSE_SYSINIT)
194static const char *
195symbol_name(vm_offset_t va, db_strategy_t strategy)
196{
197	const char *name;
198	c_db_sym_t sym;
199	db_expr_t  offset;
200
201	if (va == 0)
202		return (NULL);
203	sym = db_search_symbol(va, strategy, &offset);
204	if (offset != 0)
205		return (NULL);
206	db_symbol_values(sym, &name, NULL);
207	return (name);
208}
209#endif
210
211/*
212 * System startup; initialize the world, create process 0, mount root
213 * filesystem, and fork to create init and pagedaemon.  Most of the
214 * hard work is done in the lower-level initialization routines including
215 * startup(), which does memory initialization and autoconfiguration.
216 *
217 * This allows simple addition of new kernel subsystems that require
218 * boot time initialization.  It also allows substitution of subsystem
219 * (for instance, a scheduler, kernel profiler, or VM system) by object
220 * module.  Finally, it allows for optional "kernel threads".
221 */
222void
223mi_startup(void)
224{
225
226	struct sysinit **sipp;	/* system initialization*/
227	struct sysinit **xipp;	/* interior loop of sort*/
228	struct sysinit *save;	/* bubble*/
229
230#if defined(VERBOSE_SYSINIT)
231	int last;
232	int verbose;
233#endif
234
235	TSENTER();
236
237	if (boothowto & RB_VERBOSE)
238		bootverbose++;
239
240	if (sysinit == NULL) {
241		sysinit = SET_BEGIN(sysinit_set);
242		sysinit_end = SET_LIMIT(sysinit_set);
243	}
244
245restart:
246	/*
247	 * Perform a bubble sort of the system initialization objects by
248	 * their subsystem (primary key) and order (secondary key).
249	 */
250	for (sipp = sysinit; sipp < sysinit_end; sipp++) {
251		for (xipp = sipp + 1; xipp < sysinit_end; xipp++) {
252			if ((*sipp)->subsystem < (*xipp)->subsystem ||
253			     ((*sipp)->subsystem == (*xipp)->subsystem &&
254			      (*sipp)->order <= (*xipp)->order))
255				continue;	/* skip*/
256			save = *sipp;
257			*sipp = *xipp;
258			*xipp = save;
259		}
260	}
261
262#if defined(VERBOSE_SYSINIT)
263	last = SI_SUB_COPYRIGHT;
264	verbose = 0;
265#if !defined(DDB)
266	printf("VERBOSE_SYSINIT: DDB not enabled, symbol lookups disabled.\n");
267#endif
268#endif
269
270	/*
271	 * Traverse the (now) ordered list of system initialization tasks.
272	 * Perform each task, and continue on to the next task.
273	 */
274	for (sipp = sysinit; sipp < sysinit_end; sipp++) {
275
276		if ((*sipp)->subsystem == SI_SUB_DUMMY)
277			continue;	/* skip dummy task(s)*/
278
279		if ((*sipp)->subsystem == SI_SUB_DONE)
280			continue;
281
282#if defined(VERBOSE_SYSINIT)
283		if ((*sipp)->subsystem > last && verbose_sysinit != 0) {
284			verbose = 1;
285			last = (*sipp)->subsystem;
286			printf("subsystem %x\n", last);
287		}
288		if (verbose) {
289#if defined(DDB)
290			const char *func, *data;
291
292			func = symbol_name((vm_offset_t)(*sipp)->func,
293			    DB_STGY_PROC);
294			data = symbol_name((vm_offset_t)(*sipp)->udata,
295			    DB_STGY_ANY);
296			if (func != NULL && data != NULL)
297				printf("   %s(&%s)... ", func, data);
298			else if (func != NULL)
299				printf("   %s(%p)... ", func, (*sipp)->udata);
300			else
301#endif
302				printf("   %p(%p)... ", (*sipp)->func,
303				    (*sipp)->udata);
304		}
305#endif
306
307		/* Call function */
308		(*((*sipp)->func))((*sipp)->udata);
309
310#if defined(VERBOSE_SYSINIT)
311		if (verbose)
312			printf("done.\n");
313#endif
314
315		/* Check off the one we're just done */
316		(*sipp)->subsystem = SI_SUB_DONE;
317
318		/* Check if we've installed more sysinit items via KLD */
319		if (newsysinit != NULL) {
320			if (sysinit != SET_BEGIN(sysinit_set))
321				free(sysinit, M_TEMP);
322			sysinit = newsysinit;
323			sysinit_end = newsysinit_end;
324			newsysinit = NULL;
325			newsysinit_end = NULL;
326			goto restart;
327		}
328	}
329
330	TSEXIT();	/* Here so we don't overlap with start_init. */
331
332	mtx_assert(&Giant, MA_OWNED | MA_NOTRECURSED);
333	mtx_unlock(&Giant);
334
335	/*
336	 * Now hand over this thread to swapper.
337	 */
338	swapper();
339	/* NOTREACHED*/
340}
341
342static void
343print_caddr_t(void *data)
344{
345	printf("%s", (char *)data);
346}
347
348static void
349print_version(void *data __unused)
350{
351	int len;
352
353	/* Strip a trailing newline from version. */
354	len = strlen(version);
355	while (len > 0 && version[len - 1] == '\n')
356		len--;
357	printf("%.*s %s\n", len, version, machine);
358	printf("%s\n", compiler_version);
359}
360
361SYSINIT(announce, SI_SUB_COPYRIGHT, SI_ORDER_FIRST, print_caddr_t,
362    copyright);
363SYSINIT(trademark, SI_SUB_COPYRIGHT, SI_ORDER_SECOND, print_caddr_t,
364    trademark);
365SYSINIT(version, SI_SUB_COPYRIGHT, SI_ORDER_THIRD, print_version, NULL);
366
367#ifdef WITNESS
368static char wit_warn[] =
369     "WARNING: WITNESS option enabled, expect reduced performance.\n";
370SYSINIT(witwarn, SI_SUB_COPYRIGHT, SI_ORDER_FOURTH,
371   print_caddr_t, wit_warn);
372SYSINIT(witwarn2, SI_SUB_LAST, SI_ORDER_FOURTH,
373   print_caddr_t, wit_warn);
374#endif
375
376#ifdef DIAGNOSTIC
377static char diag_warn[] =
378     "WARNING: DIAGNOSTIC option enabled, expect reduced performance.\n";
379SYSINIT(diagwarn, SI_SUB_COPYRIGHT, SI_ORDER_FIFTH,
380    print_caddr_t, diag_warn);
381SYSINIT(diagwarn2, SI_SUB_LAST, SI_ORDER_FIFTH,
382    print_caddr_t, diag_warn);
383#endif
384
385static int
386null_fetch_syscall_args(struct thread *td __unused)
387{
388
389	panic("null_fetch_syscall_args");
390}
391
392static void
393null_set_syscall_retval(struct thread *td __unused, int error __unused)
394{
395
396	panic("null_set_syscall_retval");
397}
398
399struct sysentvec null_sysvec = {
400	.sv_size	= 0,
401	.sv_table	= NULL,
402	.sv_mask	= 0,
403	.sv_errsize	= 0,
404	.sv_errtbl	= NULL,
405	.sv_transtrap	= NULL,
406	.sv_fixup	= NULL,
407	.sv_sendsig	= NULL,
408	.sv_sigcode	= NULL,
409	.sv_szsigcode	= NULL,
410	.sv_name	= "null",
411	.sv_coredump	= NULL,
412	.sv_imgact_try	= NULL,
413	.sv_minsigstksz	= 0,
414	.sv_minuser	= VM_MIN_ADDRESS,
415	.sv_maxuser	= VM_MAXUSER_ADDRESS,
416	.sv_usrstack	= USRSTACK,
417	.sv_psstrings	= PS_STRINGS,
418	.sv_stackprot	= VM_PROT_ALL,
419	.sv_copyout_strings	= NULL,
420	.sv_setregs	= NULL,
421	.sv_fixlimit	= NULL,
422	.sv_maxssiz	= NULL,
423	.sv_flags	= 0,
424	.sv_set_syscall_retval = null_set_syscall_retval,
425	.sv_fetch_syscall_args = null_fetch_syscall_args,
426	.sv_syscallnames = NULL,
427	.sv_schedtail	= NULL,
428	.sv_thread_detach = NULL,
429	.sv_trap	= NULL,
430};
431
432/*
433 * The two following SYSINIT's are proc0 specific glue code.  I am not
434 * convinced that they can not be safely combined, but their order of
435 * operation has been maintained as the same as the original init_main.c
436 * for right now.
437 */
438/* ARGSUSED*/
439static void
440proc0_init(void *dummy __unused)
441{
442	struct proc *p;
443	struct thread *td;
444	struct ucred *newcred;
445	struct uidinfo tmpuinfo;
446	struct loginclass tmplc = {
447		.lc_name = "",
448	};
449	vm_paddr_t pageablemem;
450	int i;
451
452	GIANT_REQUIRED;
453	p = &proc0;
454	td = &thread0;
455
456	/*
457	 * Initialize magic number and osrel.
458	 */
459	p->p_magic = P_MAGIC;
460	p->p_osrel = osreldate;
461
462	/*
463	 * Initialize thread and process structures.
464	 */
465	procinit();	/* set up proc zone */
466	threadinit();	/* set up UMA zones */
467
468	/*
469	 * Initialise scheduler resources.
470	 * Add scheduler specific parts to proc, thread as needed.
471	 */
472	schedinit();	/* scheduler gets its house in order */
473
474	/*
475	 * Create process 0 (the swapper).
476	 */
477	LIST_INSERT_HEAD(&allproc, p, p_list);
478	LIST_INSERT_HEAD(PIDHASH(0), p, p_hash);
479	mtx_init(&pgrp0.pg_mtx, "process group", NULL, MTX_DEF | MTX_DUPOK);
480	p->p_pgrp = &pgrp0;
481	LIST_INSERT_HEAD(PGRPHASH(0), &pgrp0, pg_hash);
482	LIST_INIT(&pgrp0.pg_members);
483	LIST_INSERT_HEAD(&pgrp0.pg_members, p, p_pglist);
484
485	pgrp0.pg_session = &session0;
486	mtx_init(&session0.s_mtx, "session", NULL, MTX_DEF);
487	refcount_init(&session0.s_count, 1);
488	session0.s_leader = p;
489
490	p->p_sysent = &null_sysvec;
491	p->p_flag = P_SYSTEM | P_INMEM | P_KPROC;
492	p->p_flag2 = 0;
493	p->p_state = PRS_NORMAL;
494	p->p_klist = knlist_alloc(&p->p_mtx);
495	STAILQ_INIT(&p->p_ktr);
496	p->p_nice = NZERO;
497	/* pid_max cannot be greater than PID_MAX */
498	td->td_tid = PID_MAX + 1;
499	LIST_INSERT_HEAD(TIDHASH(td->td_tid), td, td_hash);
500	td->td_state = TDS_RUNNING;
501	td->td_pri_class = PRI_TIMESHARE;
502	td->td_user_pri = PUSER;
503	td->td_base_user_pri = PUSER;
504	td->td_lend_user_pri = PRI_MAX;
505	td->td_priority = PVM;
506	td->td_base_pri = PVM;
507	td->td_oncpu = curcpu;
508	td->td_flags = TDF_INMEM;
509	td->td_pflags = TDP_KTHREAD;
510	td->td_cpuset = cpuset_thread0();
511	td->td_domain.dr_policy = td->td_cpuset->cs_domain;
512	prison0_init();
513	p->p_peers = 0;
514	p->p_leader = p;
515	p->p_reaper = p;
516	p->p_treeflag |= P_TREE_REAPER;
517	LIST_INIT(&p->p_reaplist);
518
519	strncpy(p->p_comm, "kernel", sizeof (p->p_comm));
520	strncpy(td->td_name, "swapper", sizeof (td->td_name));
521
522	callout_init_mtx(&p->p_itcallout, &p->p_mtx, 0);
523	callout_init_mtx(&p->p_limco, &p->p_mtx, 0);
524	callout_init(&td->td_slpcallout, 1);
525
526	/* Create credentials. */
527	newcred = crget();
528	newcred->cr_ngroups = 1;	/* group 0 */
529	/* A hack to prevent uifind from tripping over NULL pointers. */
530	curthread->td_ucred = newcred;
531	tmpuinfo.ui_uid = 1;
532	newcred->cr_uidinfo = newcred->cr_ruidinfo = &tmpuinfo;
533	newcred->cr_uidinfo = uifind(0);
534	newcred->cr_ruidinfo = uifind(0);
535	newcred->cr_loginclass = &tmplc;
536	newcred->cr_loginclass = loginclass_find("default");
537	/* End hack. creds get properly set later with thread_cow_get_proc */
538	curthread->td_ucred = NULL;
539	newcred->cr_prison = &prison0;
540	proc_set_cred_init(p, newcred);
541#ifdef AUDIT
542	audit_cred_kproc0(newcred);
543#endif
544#ifdef MAC
545	mac_cred_create_swapper(newcred);
546#endif
547	/* Create sigacts. */
548	p->p_sigacts = sigacts_alloc();
549
550	/* Initialize signal state for process 0. */
551	siginit(&proc0);
552
553	/* Create the file descriptor table. */
554	p->p_fd = fdinit(NULL, false);
555	p->p_fdtol = NULL;
556
557	/* Create the limits structures. */
558	p->p_limit = lim_alloc();
559	for (i = 0; i < RLIM_NLIMITS; i++)
560		p->p_limit->pl_rlimit[i].rlim_cur =
561		    p->p_limit->pl_rlimit[i].rlim_max = RLIM_INFINITY;
562	p->p_limit->pl_rlimit[RLIMIT_NOFILE].rlim_cur =
563	    p->p_limit->pl_rlimit[RLIMIT_NOFILE].rlim_max = maxfiles;
564	p->p_limit->pl_rlimit[RLIMIT_NPROC].rlim_cur =
565	    p->p_limit->pl_rlimit[RLIMIT_NPROC].rlim_max = maxproc;
566	p->p_limit->pl_rlimit[RLIMIT_DATA].rlim_cur = dfldsiz;
567	p->p_limit->pl_rlimit[RLIMIT_DATA].rlim_max = maxdsiz;
568	p->p_limit->pl_rlimit[RLIMIT_STACK].rlim_cur = dflssiz;
569	p->p_limit->pl_rlimit[RLIMIT_STACK].rlim_max = maxssiz;
570	/* Cast to avoid overflow on i386/PAE. */
571	pageablemem = ptoa((vm_paddr_t)vm_free_count());
572	p->p_limit->pl_rlimit[RLIMIT_RSS].rlim_cur =
573	    p->p_limit->pl_rlimit[RLIMIT_RSS].rlim_max = pageablemem;
574	p->p_limit->pl_rlimit[RLIMIT_MEMLOCK].rlim_cur = pageablemem / 3;
575	p->p_limit->pl_rlimit[RLIMIT_MEMLOCK].rlim_max = pageablemem;
576	p->p_cpulimit = RLIM_INFINITY;
577
578	PROC_LOCK(p);
579	thread_cow_get_proc(td, p);
580	PROC_UNLOCK(p);
581
582	/* Initialize resource accounting structures. */
583	racct_create(&p->p_racct);
584
585	p->p_stats = pstats_alloc();
586
587	/* Allocate a prototype map so we have something to fork. */
588	p->p_vmspace = &vmspace0;
589	vmspace0.vm_refcnt = 1;
590	pmap_pinit0(vmspace_pmap(&vmspace0));
591
592	/*
593	 * proc0 is not expected to enter usermode, so there is no special
594	 * handling for sv_minuser here, like is done for exec_new_vmspace().
595	 */
596	vm_map_init(&vmspace0.vm_map, vmspace_pmap(&vmspace0),
597	    p->p_sysent->sv_minuser, p->p_sysent->sv_maxuser);
598
599	/*
600	 * Call the init and ctor for the new thread and proc.  We wait
601	 * to do this until all other structures are fairly sane.
602	 */
603	EVENTHANDLER_DIRECT_INVOKE(process_init, p);
604	EVENTHANDLER_DIRECT_INVOKE(thread_init, td);
605	EVENTHANDLER_DIRECT_INVOKE(process_ctor, p);
606	EVENTHANDLER_DIRECT_INVOKE(thread_ctor, td);
607
608	/*
609	 * Charge root for one process.
610	 */
611	(void)chgproccnt(p->p_ucred->cr_ruidinfo, 1, 0);
612	PROC_LOCK(p);
613	racct_add_force(p, RACCT_NPROC, 1);
614	PROC_UNLOCK(p);
615}
616SYSINIT(p0init, SI_SUB_INTRINSIC, SI_ORDER_FIRST, proc0_init, NULL);
617
618/* ARGSUSED*/
619static void
620proc0_post(void *dummy __unused)
621{
622	struct timespec ts;
623	struct proc *p;
624	struct rusage ru;
625	struct thread *td;
626
627	/*
628	 * Now we can look at the time, having had a chance to verify the
629	 * time from the filesystem.  Pretend that proc0 started now.
630	 */
631	sx_slock(&allproc_lock);
632	FOREACH_PROC_IN_SYSTEM(p) {
633		PROC_LOCK(p);
634		if (p->p_state == PRS_NEW) {
635			PROC_UNLOCK(p);
636			continue;
637		}
638		microuptime(&p->p_stats->p_start);
639		PROC_STATLOCK(p);
640		rufetch(p, &ru);	/* Clears thread stats */
641		p->p_rux.rux_runtime = 0;
642		p->p_rux.rux_uticks = 0;
643		p->p_rux.rux_sticks = 0;
644		p->p_rux.rux_iticks = 0;
645		PROC_STATUNLOCK(p);
646		FOREACH_THREAD_IN_PROC(p, td) {
647			td->td_runtime = 0;
648		}
649		PROC_UNLOCK(p);
650	}
651	sx_sunlock(&allproc_lock);
652	PCPU_SET(switchtime, cpu_ticks());
653	PCPU_SET(switchticks, ticks);
654
655	/*
656	 * Give the ``random'' number generator a thump.
657	 */
658	nanotime(&ts);
659	srandom(ts.tv_sec ^ ts.tv_nsec);
660}
661SYSINIT(p0post, SI_SUB_INTRINSIC_POST, SI_ORDER_FIRST, proc0_post, NULL);
662
663static void
664random_init(void *dummy __unused)
665{
666
667	/*
668	 * After CPU has been started we have some randomness on most
669	 * platforms via get_cyclecount().  For platforms that don't
670	 * we will reseed random(9) in proc0_post() as well.
671	 */
672	srandom(get_cyclecount());
673}
674SYSINIT(random, SI_SUB_RANDOM, SI_ORDER_FIRST, random_init, NULL);
675
676/*
677 ***************************************************************************
678 ****
679 **** The following SYSINIT's and glue code should be moved to the
680 **** respective files on a per subsystem basis.
681 ****
682 ***************************************************************************
683 */
684
685/*
686 * List of paths to try when searching for "init".
687 */
688static char init_path[MAXPATHLEN] =
689#ifdef	INIT_PATH
690    __XSTRING(INIT_PATH);
691#else
692    "/sbin/init:/sbin/oinit:/sbin/init.bak:/rescue/init";
693#endif
694SYSCTL_STRING(_kern, OID_AUTO, init_path, CTLFLAG_RD, init_path, 0,
695	"Path used to search the init process");
696
697/*
698 * Shutdown timeout of init(8).
699 * Unused within kernel, but used to control init(8), hence do not remove.
700 */
701#ifndef INIT_SHUTDOWN_TIMEOUT
702#define INIT_SHUTDOWN_TIMEOUT 120
703#endif
704static int init_shutdown_timeout = INIT_SHUTDOWN_TIMEOUT;
705SYSCTL_INT(_kern, OID_AUTO, init_shutdown_timeout,
706	CTLFLAG_RW, &init_shutdown_timeout, 0, "Shutdown timeout of init(8). "
707	"Unused within kernel, but used to control init(8)");
708
709/*
710 * Start the initial user process; try exec'ing each pathname in init_path.
711 * The program is invoked with one argument containing the boot flags.
712 */
713static void
714start_init(void *dummy)
715{
716	vm_offset_t addr;
717	struct execve_args args;
718	int options, error;
719	size_t pathlen;
720	char *var, *path;
721	char *free_init_path, *tmp_init_path;
722	char *ucp, **uap, *arg0, *arg1;
723	struct thread *td;
724	struct proc *p;
725
726	TSENTER();	/* Here so we don't overlap with mi_startup. */
727
728	td = curthread;
729	p = td->td_proc;
730
731	vfs_mountroot();
732
733	/* Wipe GELI passphrase from the environment. */
734	kern_unsetenv("kern.geom.eli.passphrase");
735
736	/*
737	 * Need just enough stack to hold the faked-up "execve()" arguments.
738	 */
739	addr = p->p_sysent->sv_usrstack - PAGE_SIZE;
740	if (vm_map_find(&p->p_vmspace->vm_map, NULL, 0, &addr, PAGE_SIZE, 0,
741	    VMFS_NO_SPACE, VM_PROT_ALL, VM_PROT_ALL, 0) != 0)
742		panic("init: couldn't allocate argument space");
743	p->p_vmspace->vm_maxsaddr = (caddr_t)addr;
744	p->p_vmspace->vm_ssize = 1;
745
746	/* For Multicons, report which console is primary to both */
747	if (boothowto & RB_MULTIPLE) {
748		if (boothowto & RB_SERIAL)
749			printf("Dual Console: Serial Primary, Video Secondary\n");
750		else
751			printf("Dual Console: Video Primary, Serial Secondary\n");
752	}
753
754	if ((var = kern_getenv("init_path")) != NULL) {
755		strlcpy(init_path, var, sizeof(init_path));
756		freeenv(var);
757	}
758	free_init_path = tmp_init_path = strdup(init_path, M_TEMP);
759
760	while ((path = strsep(&tmp_init_path, ":")) != NULL) {
761		pathlen = strlen(path) + 1;
762		if (bootverbose)
763			printf("start_init: trying %s\n", path);
764
765		/*
766		 * Move out the boot flag argument.
767		 */
768		options = 0;
769		ucp = (char *)p->p_sysent->sv_usrstack;
770		(void)subyte(--ucp, 0);		/* trailing zero */
771		if (boothowto & RB_SINGLE) {
772			(void)subyte(--ucp, 's');
773			options = 1;
774		}
775#ifdef notyet
776                if (boothowto & RB_FASTBOOT) {
777			(void)subyte(--ucp, 'f');
778			options = 1;
779		}
780#endif
781
782#ifdef BOOTCDROM
783		(void)subyte(--ucp, 'C');
784		options = 1;
785#endif
786
787		if (options == 0)
788			(void)subyte(--ucp, '-');
789		(void)subyte(--ucp, '-');		/* leading hyphen */
790		arg1 = ucp;
791
792		/*
793		 * Move out the file name (also arg 0).
794		 */
795		ucp -= pathlen;
796		copyout(path, ucp, pathlen);
797		arg0 = ucp;
798
799		/*
800		 * Move out the arg pointers.
801		 */
802		uap = (char **)rounddown2((intptr_t)ucp, sizeof(intptr_t));
803		(void)suword((caddr_t)--uap, (long)0);	/* terminator */
804		(void)suword((caddr_t)--uap, (long)(intptr_t)arg1);
805		(void)suword((caddr_t)--uap, (long)(intptr_t)arg0);
806
807		/*
808		 * Point at the arguments.
809		 */
810		args.fname = arg0;
811		args.argv = uap;
812		args.envv = NULL;
813
814		/*
815		 * Now try to exec the program.  If can't for any reason
816		 * other than it doesn't exist, complain.
817		 *
818		 * Otherwise, return via fork_trampoline() all the way
819		 * to user mode as init!
820		 */
821		if ((error = sys_execve(td, &args)) == EJUSTRETURN) {
822			free(free_init_path, M_TEMP);
823			TSEXIT();
824			return;
825		}
826		if (error != ENOENT)
827			printf("exec %s: error %d\n", path, error);
828	}
829	free(free_init_path, M_TEMP);
830	printf("init: not found in path %s\n", init_path);
831	panic("no init");
832}
833
834/*
835 * Like kproc_create(), but runs in its own address space.  We do this
836 * early to reserve pid 1.  Note special case - do not make it
837 * runnable yet, init execution is started when userspace can be served.
838 */
839static void
840create_init(const void *udata __unused)
841{
842	struct fork_req fr;
843	struct ucred *newcred, *oldcred;
844	struct thread *td;
845	int error;
846
847	bzero(&fr, sizeof(fr));
848	fr.fr_flags = RFFDG | RFPROC | RFSTOPPED;
849	fr.fr_procp = &initproc;
850	error = fork1(&thread0, &fr);
851	if (error)
852		panic("cannot fork init: %d\n", error);
853	KASSERT(initproc->p_pid == 1, ("create_init: initproc->p_pid != 1"));
854	/* divorce init's credentials from the kernel's */
855	newcred = crget();
856	sx_xlock(&proctree_lock);
857	PROC_LOCK(initproc);
858	initproc->p_flag |= P_SYSTEM | P_INMEM;
859	initproc->p_treeflag |= P_TREE_REAPER;
860	oldcred = initproc->p_ucred;
861	crcopy(newcred, oldcred);
862#ifdef MAC
863	mac_cred_create_init(newcred);
864#endif
865#ifdef AUDIT
866	audit_cred_proc1(newcred);
867#endif
868	proc_set_cred(initproc, newcred);
869	td = FIRST_THREAD_IN_PROC(initproc);
870	crfree(td->td_ucred);
871	td->td_ucred = crhold(initproc->p_ucred);
872	PROC_UNLOCK(initproc);
873	sx_xunlock(&proctree_lock);
874	crfree(oldcred);
875	cpu_fork_kthread_handler(FIRST_THREAD_IN_PROC(initproc),
876	    start_init, NULL);
877}
878SYSINIT(init, SI_SUB_CREATE_INIT, SI_ORDER_FIRST, create_init, NULL);
879
880/*
881 * Make it runnable now.
882 */
883static void
884kick_init(const void *udata __unused)
885{
886	struct thread *td;
887
888	td = FIRST_THREAD_IN_PROC(initproc);
889	thread_lock(td);
890	TD_SET_CAN_RUN(td);
891	sched_add(td, SRQ_BORING);
892	thread_unlock(td);
893}
894SYSINIT(kickinit, SI_SUB_KTHREAD_INIT, SI_ORDER_MIDDLE, kick_init, NULL);
895