2 * Copyright (c) 1995 Terrence R. Lambert
5 * Copyright (c) 1982, 1986, 1989, 1991, 1992, 1993
6 * The Regents of the University of California. All rights reserved.
7 * (c) UNIX System Laboratories, Inc.
8 * All or some portions of this file are derived from material licensed
9 * to the University of California by American Telephone and Telegraph
10 * Co. or Unix System Laboratories, Inc. and are reproduced herein with
11 * the permission of UNIX System Laboratories, Inc.
13 * Redistribution and use in source and binary forms, with or without
14 * modification, are permitted provided that the following conditions
16 * 1. Redistributions of source code must retain the above copyright
17 * notice, this list of conditions and the following disclaimer.
18 * 2. Redistributions in binary form must reproduce the above copyright
19 * notice, this list of conditions and the following disclaimer in the
20 * documentation and/or other materials provided with the distribution.
21 * 3. All advertising materials mentioning features or use of this software
22 * must display the following acknowledgement:
23 * This product includes software developed by the University of
24 * California, Berkeley and its contributors.
25 * 4. Neither the name of the University nor the names of its contributors
26 * may be used to endorse or promote products derived from this software
27 * without specific prior written permission.
29 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
30 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
31 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
32 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
33 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
34 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
35 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
36 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
37 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
38 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
41 * @(#)init_main.c 8.9 (Berkeley) 1/21/94
44 #include <sys/cdefs.h>
45 __FBSDID("$FreeBSD$");
48 #include "opt_init_path.h"
50 #include <sys/param.h>
51 #include <sys/kernel.h>
54 #include <sys/filedesc.h>
58 #include <sys/mount.h>
59 #include <sys/mutex.h>
60 #include <sys/syscallsubr.h>
61 #include <sys/sysctl.h>
63 #include <sys/resourcevar.h>
64 #include <sys/systm.h>
65 #include <sys/signalvar.h>
66 #include <sys/vnode.h>
67 #include <sys/sysent.h>
68 #include <sys/reboot.h>
69 #include <sys/sched.h>
71 #include <sys/sysproto.h>
72 #include <sys/vmmeter.h>
73 #include <sys/unistd.h>
74 #include <sys/malloc.h>
76 #include <sys/cpuset.h>
78 #include <machine/cpu.h>
80 #include <security/audit/audit.h>
81 #include <security/mac/mac_framework.h>
84 #include <vm/vm_param.h>
86 #include <vm/vm_map.h>
87 #include <sys/copyright.h>
90 #include <ddb/db_sym.h>
92 void mi_startup(void); /* Should be elsewhere */
94 /* Components of the first process -- never freed. */
95 static struct session session0;
96 static struct pgrp pgrp0;
98 struct thread thread0 __aligned(16);
99 struct vmspace vmspace0;
100 struct proc *initproc;
102 int boothowto = 0; /* initialized so that it can be patched */
103 SYSCTL_INT(_debug, OID_AUTO, boothowto, CTLFLAG_RD, &boothowto, 0, "");
105 SYSCTL_INT(_debug, OID_AUTO, bootverbose, CTLFLAG_RW, &bootverbose, 0, "");
108 * This ensures that there is at least one entry so that the sysinit_set
109 * symbol is not undefined. A sybsystem ID of SI_SUB_DUMMY is never
112 SYSINIT(placeholder, SI_SUB_DUMMY, SI_ORDER_ANY, NULL, NULL);
115 * The sysinit table itself. Items are checked off as the are run.
116 * If we want to register new sysinit types, add them to newsysinit.
118 SET_DECLARE(sysinit_set, struct sysinit);
119 struct sysinit **sysinit, **sysinit_end;
120 struct sysinit **newsysinit, **newsysinit_end;
123 * Merge a new sysinit set into the current set, reallocating it if
124 * necessary. This can only be called after malloc is running.
127 sysinit_add(struct sysinit **set, struct sysinit **set_end)
129 struct sysinit **newset;
130 struct sysinit **sipp;
131 struct sysinit **xipp;
134 count = set_end - set;
136 count += newsysinit_end - newsysinit;
138 count += sysinit_end - sysinit;
139 newset = malloc(count * sizeof(*sipp), M_TEMP, M_NOWAIT);
141 panic("cannot malloc for sysinit");
144 for (sipp = newsysinit; sipp < newsysinit_end; sipp++)
147 for (sipp = sysinit; sipp < sysinit_end; sipp++)
149 for (sipp = set; sipp < set_end; sipp++)
152 free(newsysinit, M_TEMP);
154 newsysinit_end = newset + count;
158 * System startup; initialize the world, create process 0, mount root
159 * filesystem, and fork to create init and pagedaemon. Most of the
160 * hard work is done in the lower-level initialization routines including
161 * startup(), which does memory initialization and autoconfiguration.
163 * This allows simple addition of new kernel subsystems that require
164 * boot time initialization. It also allows substitution of subsystem
165 * (for instance, a scheduler, kernel profiler, or VM system) by object
166 * module. Finally, it allows for optional "kernel threads".
172 register struct sysinit **sipp; /* system initialization*/
173 register struct sysinit **xipp; /* interior loop of sort*/
174 register struct sysinit *save; /* bubble*/
176 #if defined(VERBOSE_SYSINIT)
181 if (sysinit == NULL) {
182 sysinit = SET_BEGIN(sysinit_set);
183 sysinit_end = SET_LIMIT(sysinit_set);
188 * Perform a bubble sort of the system initialization objects by
189 * their subsystem (primary key) and order (secondary key).
191 for (sipp = sysinit; sipp < sysinit_end; sipp++) {
192 for (xipp = sipp + 1; xipp < sysinit_end; xipp++) {
193 if ((*sipp)->subsystem < (*xipp)->subsystem ||
194 ((*sipp)->subsystem == (*xipp)->subsystem &&
195 (*sipp)->order <= (*xipp)->order))
203 #if defined(VERBOSE_SYSINIT)
204 last = SI_SUB_COPYRIGHT;
207 printf("VERBOSE_SYSINIT: DDB not enabled, symbol lookups disabled.\n");
212 * Traverse the (now) ordered list of system initialization tasks.
213 * Perform each task, and continue on to the next task.
215 * The last item on the list is expected to be the scheduler,
216 * which will not return.
218 for (sipp = sysinit; sipp < sysinit_end; sipp++) {
220 if ((*sipp)->subsystem == SI_SUB_DUMMY)
221 continue; /* skip dummy task(s)*/
223 if ((*sipp)->subsystem == SI_SUB_DONE)
226 #if defined(VERBOSE_SYSINIT)
227 if ((*sipp)->subsystem > last) {
229 last = (*sipp)->subsystem;
230 printf("subsystem %x\n", last);
238 sym = db_search_symbol((vm_offset_t)(*sipp)->func,
239 DB_STGY_PROC, &offset);
240 db_symbol_values(sym, &name, NULL);
242 printf(" %s(%p)... ", name, (*sipp)->udata);
245 printf(" %p(%p)... ", (*sipp)->func,
251 (*((*sipp)->func))((*sipp)->udata);
253 #if defined(VERBOSE_SYSINIT)
258 /* Check off the one we're just done */
259 (*sipp)->subsystem = SI_SUB_DONE;
261 /* Check if we've installed more sysinit items via KLD */
262 if (newsysinit != NULL) {
263 if (sysinit != SET_BEGIN(sysinit_set))
264 free(sysinit, M_TEMP);
265 sysinit = newsysinit;
266 sysinit_end = newsysinit_end;
268 newsysinit_end = NULL;
273 panic("Shouldn't get here!");
279 ***************************************************************************
281 **** The following SYSINIT's belong elsewhere, but have not yet
284 ***************************************************************************
287 print_caddr_t(void *data)
289 printf("%s", (char *)data);
293 print_version(void *data __unused)
297 /* Strip a trailing newline from version. */
298 len = strlen(version);
299 while (len > 0 && version[len - 1] == '\n')
301 printf("%.*s %s\n", len, version, machine);
304 SYSINIT(announce, SI_SUB_COPYRIGHT, SI_ORDER_FIRST, print_caddr_t,
306 SYSINIT(trademark, SI_SUB_COPYRIGHT, SI_ORDER_SECOND, print_caddr_t,
308 SYSINIT(version, SI_SUB_COPYRIGHT, SI_ORDER_THIRD, print_version, NULL);
311 static char wit_warn[] =
312 "WARNING: WITNESS option enabled, expect reduced performance.\n";
313 SYSINIT(witwarn, SI_SUB_COPYRIGHT, SI_ORDER_THIRD + 1,
314 print_caddr_t, wit_warn);
315 SYSINIT(witwarn2, SI_SUB_RUN_SCHEDULER, SI_ORDER_THIRD + 1,
316 print_caddr_t, wit_warn);
320 static char diag_warn[] =
321 "WARNING: DIAGNOSTIC option enabled, expect reduced performance.\n";
322 SYSINIT(diagwarn, SI_SUB_COPYRIGHT, SI_ORDER_THIRD + 2,
323 print_caddr_t, diag_warn);
324 SYSINIT(diagwarn2, SI_SUB_RUN_SCHEDULER, SI_ORDER_THIRD + 2,
325 print_caddr_t, diag_warn);
329 set_boot_verbose(void *data __unused)
332 if (boothowto & RB_VERBOSE)
335 SYSINIT(boot_verbose, SI_SUB_TUNABLES, SI_ORDER_ANY, set_boot_verbose, NULL);
337 struct sysentvec null_sysvec = {
345 .sv_transtrap = NULL,
349 .sv_szsigcode = NULL,
350 .sv_prepsyscall = NULL,
353 .sv_imgact_try = NULL,
355 .sv_pagesize = PAGE_SIZE,
356 .sv_minuser = VM_MIN_ADDRESS,
357 .sv_maxuser = VM_MAXUSER_ADDRESS,
358 .sv_usrstack = USRSTACK,
359 .sv_psstrings = PS_STRINGS,
360 .sv_stackprot = VM_PROT_ALL,
361 .sv_copyout_strings = NULL,
368 ***************************************************************************
370 **** The two following SYSINIT's are proc0 specific glue code. I am not
371 **** convinced that they can not be safely combined, but their order of
372 **** operation has been maintained as the same as the original init_main.c
375 **** These probably belong in init_proc.c or kern_proc.c, since they
376 **** deal with proc0 (the fork template process).
378 ***************************************************************************
382 proc0_init(void *dummy __unused)
386 vm_paddr_t pageablemem;
394 * Initialize magic number and osrel.
396 p->p_magic = P_MAGIC;
397 p->p_osrel = osreldate;
400 * Initialize thread and process structures.
402 procinit(); /* set up proc zone */
403 threadinit(); /* set up UMA zones */
406 * Initialise scheduler resources.
407 * Add scheduler specific parts to proc, thread as needed.
409 schedinit(); /* scheduler gets its house in order */
411 * Initialize sleep queue hash table
416 * additional VM structures
421 * Create process 0 (the swapper).
423 LIST_INSERT_HEAD(&allproc, p, p_list);
424 LIST_INSERT_HEAD(PIDHASH(0), p, p_hash);
425 mtx_init(&pgrp0.pg_mtx, "process group", NULL, MTX_DEF | MTX_DUPOK);
427 LIST_INSERT_HEAD(PGRPHASH(0), &pgrp0, pg_hash);
428 LIST_INIT(&pgrp0.pg_members);
429 LIST_INSERT_HEAD(&pgrp0.pg_members, p, p_pglist);
431 pgrp0.pg_session = &session0;
432 mtx_init(&session0.s_mtx, "session", NULL, MTX_DEF);
433 refcount_init(&session0.s_count, 1);
434 session0.s_leader = p;
436 p->p_sysent = &null_sysvec;
437 p->p_flag = P_SYSTEM | P_INMEM;
438 p->p_state = PRS_NORMAL;
439 knlist_init_mtx(&p->p_klist, &p->p_mtx);
440 STAILQ_INIT(&p->p_ktr);
442 td->td_tid = PID_MAX + 1;
443 td->td_state = TDS_RUNNING;
444 td->td_pri_class = PRI_TIMESHARE;
445 td->td_user_pri = PUSER;
446 td->td_base_user_pri = PUSER;
447 td->td_priority = PVM;
448 td->td_base_pri = PUSER;
450 td->td_flags = TDF_INMEM|TDP_KTHREAD;
451 td->td_cpuset = cpuset_thread0();
452 prison0.pr_cpuset = cpuset_ref(td->td_cpuset);
457 strncpy(p->p_comm, "kernel", sizeof (p->p_comm));
458 strncpy(td->td_name, "swapper", sizeof (td->td_name));
460 callout_init(&p->p_itcallout, CALLOUT_MPSAFE);
461 callout_init_mtx(&p->p_limco, &p->p_mtx, 0);
462 callout_init(&td->td_slpcallout, CALLOUT_MPSAFE);
464 /* Create credentials. */
465 p->p_ucred = crget();
466 p->p_ucred->cr_ngroups = 1; /* group 0 */
467 p->p_ucred->cr_uidinfo = uifind(0);
468 p->p_ucred->cr_ruidinfo = uifind(0);
469 p->p_ucred->cr_prison = &prison0;
471 audit_cred_kproc0(p->p_ucred);
474 mac_cred_create_swapper(p->p_ucred);
476 td->td_ucred = crhold(p->p_ucred);
478 /* Create sigacts. */
479 p->p_sigacts = sigacts_alloc();
481 /* Initialize signal state for process 0. */
484 /* Create the file descriptor table. */
485 p->p_fd = fdinit(NULL);
488 /* Create the limits structures. */
489 p->p_limit = lim_alloc();
490 for (i = 0; i < RLIM_NLIMITS; i++)
491 p->p_limit->pl_rlimit[i].rlim_cur =
492 p->p_limit->pl_rlimit[i].rlim_max = RLIM_INFINITY;
493 p->p_limit->pl_rlimit[RLIMIT_NOFILE].rlim_cur =
494 p->p_limit->pl_rlimit[RLIMIT_NOFILE].rlim_max = maxfiles;
495 p->p_limit->pl_rlimit[RLIMIT_NPROC].rlim_cur =
496 p->p_limit->pl_rlimit[RLIMIT_NPROC].rlim_max = maxproc;
497 p->p_limit->pl_rlimit[RLIMIT_DATA].rlim_cur = dfldsiz;
498 p->p_limit->pl_rlimit[RLIMIT_DATA].rlim_max = maxdsiz;
499 p->p_limit->pl_rlimit[RLIMIT_STACK].rlim_cur = dflssiz;
500 p->p_limit->pl_rlimit[RLIMIT_STACK].rlim_max = maxssiz;
501 /* Cast to avoid overflow on i386/PAE. */
502 pageablemem = ptoa((vm_paddr_t)cnt.v_free_count);
503 p->p_limit->pl_rlimit[RLIMIT_RSS].rlim_cur =
504 p->p_limit->pl_rlimit[RLIMIT_RSS].rlim_max = pageablemem;
505 p->p_limit->pl_rlimit[RLIMIT_MEMLOCK].rlim_cur = pageablemem / 3;
506 p->p_limit->pl_rlimit[RLIMIT_MEMLOCK].rlim_max = pageablemem;
507 p->p_cpulimit = RLIM_INFINITY;
509 p->p_stats = pstats_alloc();
511 /* Allocate a prototype map so we have something to fork. */
512 pmap_pinit0(vmspace_pmap(&vmspace0));
513 p->p_vmspace = &vmspace0;
514 vmspace0.vm_refcnt = 1;
517 * proc0 is not expected to enter usermode, so there is no special
518 * handling for sv_minuser here, like is done for exec_new_vmspace().
520 vm_map_init(&vmspace0.vm_map, vmspace_pmap(&vmspace0),
521 p->p_sysent->sv_minuser, p->p_sysent->sv_maxuser);
524 * call the init and ctor for the new thread and proc
525 * we wait to do this until all other structures
528 EVENTHANDLER_INVOKE(process_init, p);
529 EVENTHANDLER_INVOKE(thread_init, td);
530 EVENTHANDLER_INVOKE(process_ctor, p);
531 EVENTHANDLER_INVOKE(thread_ctor, td);
534 * Charge root for one process.
536 (void)chgproccnt(p->p_ucred->cr_ruidinfo, 1, 0);
538 SYSINIT(p0init, SI_SUB_INTRINSIC, SI_ORDER_FIRST, proc0_init, NULL);
542 proc0_post(void *dummy __unused)
550 * Now we can look at the time, having had a chance to verify the
551 * time from the filesystem. Pretend that proc0 started now.
553 sx_slock(&allproc_lock);
554 FOREACH_PROC_IN_SYSTEM(p) {
555 microuptime(&p->p_stats->p_start);
557 rufetch(p, &ru); /* Clears thread stats */
559 p->p_rux.rux_runtime = 0;
560 p->p_rux.rux_uticks = 0;
561 p->p_rux.rux_sticks = 0;
562 p->p_rux.rux_iticks = 0;
563 FOREACH_THREAD_IN_PROC(p, td) {
567 sx_sunlock(&allproc_lock);
568 PCPU_SET(switchtime, cpu_ticks());
569 PCPU_SET(switchticks, ticks);
572 * Give the ``random'' number generator a thump.
575 srandom(ts.tv_sec ^ ts.tv_nsec);
577 SYSINIT(p0post, SI_SUB_INTRINSIC_POST, SI_ORDER_FIRST, proc0_post, NULL);
580 random_init(void *dummy __unused)
584 * After CPU has been started we have some randomness on most
585 * platforms via get_cyclecount(). For platforms that don't
586 * we will reseed random(9) in proc0_post() as well.
588 srandom(get_cyclecount());
590 SYSINIT(random, SI_SUB_RANDOM, SI_ORDER_FIRST, random_init, NULL);
593 ***************************************************************************
595 **** The following SYSINIT's and glue code should be moved to the
596 **** respective files on a per subsystem basis.
598 ***************************************************************************
603 ***************************************************************************
605 **** The following code probably belongs in another file, like
606 **** kern/init_init.c.
608 ***************************************************************************
612 * List of paths to try when searching for "init".
614 static char init_path[MAXPATHLEN] =
616 __XSTRING(INIT_PATH);
618 "/sbin/init:/sbin/oinit:/sbin/init.bak:/rescue/init:/stand/sysinstall";
620 SYSCTL_STRING(_kern, OID_AUTO, init_path, CTLFLAG_RD, init_path, 0,
621 "Path used to search the init process");
624 * Shutdown timeout of init(8).
625 * Unused within kernel, but used to control init(8), hence do not remove.
627 #ifndef INIT_SHUTDOWN_TIMEOUT
628 #define INIT_SHUTDOWN_TIMEOUT 120
630 static int init_shutdown_timeout = INIT_SHUTDOWN_TIMEOUT;
631 SYSCTL_INT(_kern, OID_AUTO, init_shutdown_timeout,
632 CTLFLAG_RW, &init_shutdown_timeout, 0, "");
635 * Start the initial user process; try exec'ing each pathname in init_path.
636 * The program is invoked with one argument containing the boot flags.
639 start_init(void *dummy)
642 struct execve_args args;
644 char *var, *path, *next, *s;
645 char *ucp, **uap, *arg0, *arg1;
659 * Need just enough stack to hold the faked-up "execve()" arguments.
661 addr = p->p_sysent->sv_usrstack - PAGE_SIZE;
662 if (vm_map_find(&p->p_vmspace->vm_map, NULL, 0, &addr, PAGE_SIZE,
663 FALSE, VM_PROT_ALL, VM_PROT_ALL, 0) != 0)
664 panic("init: couldn't allocate argument space");
665 p->p_vmspace->vm_maxsaddr = (caddr_t)addr;
666 p->p_vmspace->vm_ssize = 1;
668 if ((var = getenv("init_path")) != NULL) {
669 strlcpy(init_path, var, sizeof(init_path));
673 for (path = init_path; *path != '\0'; path = next) {
678 for (next = path; *next != '\0' && *next != ':'; next++)
681 printf("start_init: trying %.*s\n", (int)(next - path),
685 * Move out the boot flag argument.
688 ucp = (char *)p->p_sysent->sv_usrstack;
689 (void)subyte(--ucp, 0); /* trailing zero */
690 if (boothowto & RB_SINGLE) {
691 (void)subyte(--ucp, 's');
695 if (boothowto & RB_FASTBOOT) {
696 (void)subyte(--ucp, 'f');
702 (void)subyte(--ucp, 'C');
707 (void)subyte(--ucp, '-');
708 (void)subyte(--ucp, '-'); /* leading hyphen */
712 * Move out the file name (also arg 0).
714 (void)subyte(--ucp, 0);
715 for (s = next - 1; s >= path; s--)
716 (void)subyte(--ucp, *s);
720 * Move out the arg pointers.
722 uap = (char **)((intptr_t)ucp & ~(sizeof(intptr_t)-1));
723 (void)suword((caddr_t)--uap, (long)0); /* terminator */
724 (void)suword((caddr_t)--uap, (long)(intptr_t)arg1);
725 (void)suword((caddr_t)--uap, (long)(intptr_t)arg0);
728 * Point at the arguments.
735 * Now try to exec the program. If can't for any reason
736 * other than it doesn't exist, complain.
738 * Otherwise, return via fork_trampoline() all the way
739 * to user mode as init!
741 if ((error = execve(td, &args)) == 0) {
746 printf("exec %.*s: error %d\n", (int)(next - path),
749 printf("init: not found in path %s\n", init_path);
754 * Like kproc_create(), but runs in it's own address space.
755 * We do this early to reserve pid 1.
757 * Note special case - do not make it runnable yet. Other work
758 * in progress will change this more.
761 create_init(const void *udata __unused)
763 struct ucred *newcred, *oldcred;
766 error = fork1(&thread0, RFFDG | RFPROC | RFSTOPPED, 0, &initproc);
768 panic("cannot fork init: %d\n", error);
769 KASSERT(initproc->p_pid == 1, ("create_init: initproc->p_pid != 1"));
770 /* divorce init's credentials from the kernel's */
773 initproc->p_flag |= P_SYSTEM | P_INMEM;
774 oldcred = initproc->p_ucred;
775 crcopy(newcred, oldcred);
777 mac_cred_create_init(newcred);
780 audit_cred_proc1(newcred);
782 initproc->p_ucred = newcred;
783 PROC_UNLOCK(initproc);
785 cred_update_thread(FIRST_THREAD_IN_PROC(initproc));
786 cpu_set_fork_handler(FIRST_THREAD_IN_PROC(initproc), start_init, NULL);
788 SYSINIT(init, SI_SUB_CREATE_INIT, SI_ORDER_FIRST, create_init, NULL);
791 * Make it runnable now.
794 kick_init(const void *udata __unused)
798 td = FIRST_THREAD_IN_PROC(initproc);
801 sched_add(td, SRQ_BORING);
804 SYSINIT(kickinit, SI_SUB_KTHREAD_INIT, SI_ORDER_FIRST, kick_init, NULL);