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/loginclass.h>
59 #include <sys/mount.h>
60 #include <sys/mutex.h>
61 #include <sys/syscallsubr.h>
62 #include <sys/sysctl.h>
64 #include <sys/resourcevar.h>
65 #include <sys/systm.h>
66 #include <sys/signalvar.h>
67 #include <sys/vnode.h>
68 #include <sys/sysent.h>
69 #include <sys/reboot.h>
70 #include <sys/sched.h>
72 #include <sys/sysproto.h>
73 #include <sys/vmmeter.h>
74 #include <sys/unistd.h>
75 #include <sys/malloc.h>
77 #include <sys/cpuset.h>
79 #include <machine/cpu.h>
81 #include <security/audit/audit.h>
82 #include <security/mac/mac_framework.h>
85 #include <vm/vm_param.h>
87 #include <vm/vm_map.h>
88 #include <sys/copyright.h>
91 #include <ddb/db_sym.h>
93 void mi_startup(void); /* Should be elsewhere */
95 /* Components of the first process -- never freed. */
96 static struct session session0;
97 static struct pgrp pgrp0;
99 struct thread thread0 __aligned(16);
100 struct vmspace vmspace0;
101 struct proc *initproc;
103 int boothowto = 0; /* initialized so that it can be patched */
104 SYSCTL_INT(_debug, OID_AUTO, boothowto, CTLFLAG_RD, &boothowto, 0,
105 "Boot control flags, passed from loader");
107 SYSCTL_INT(_debug, OID_AUTO, bootverbose, CTLFLAG_RW, &bootverbose, 0,
108 "Control the output of verbose kernel messages");
111 * This ensures that there is at least one entry so that the sysinit_set
112 * symbol is not undefined. A sybsystem ID of SI_SUB_DUMMY is never
115 SYSINIT(placeholder, SI_SUB_DUMMY, SI_ORDER_ANY, NULL, NULL);
118 * The sysinit table itself. Items are checked off as the are run.
119 * If we want to register new sysinit types, add them to newsysinit.
121 SET_DECLARE(sysinit_set, struct sysinit);
122 struct sysinit **sysinit, **sysinit_end;
123 struct sysinit **newsysinit, **newsysinit_end;
126 * Merge a new sysinit set into the current set, reallocating it if
127 * necessary. This can only be called after malloc is running.
130 sysinit_add(struct sysinit **set, struct sysinit **set_end)
132 struct sysinit **newset;
133 struct sysinit **sipp;
134 struct sysinit **xipp;
137 count = set_end - set;
139 count += newsysinit_end - newsysinit;
141 count += sysinit_end - sysinit;
142 newset = malloc(count * sizeof(*sipp), M_TEMP, M_NOWAIT);
144 panic("cannot malloc for sysinit");
147 for (sipp = newsysinit; sipp < newsysinit_end; sipp++)
150 for (sipp = sysinit; sipp < sysinit_end; sipp++)
152 for (sipp = set; sipp < set_end; sipp++)
155 free(newsysinit, M_TEMP);
157 newsysinit_end = newset + count;
161 * System startup; initialize the world, create process 0, mount root
162 * filesystem, and fork to create init and pagedaemon. Most of the
163 * hard work is done in the lower-level initialization routines including
164 * startup(), which does memory initialization and autoconfiguration.
166 * This allows simple addition of new kernel subsystems that require
167 * boot time initialization. It also allows substitution of subsystem
168 * (for instance, a scheduler, kernel profiler, or VM system) by object
169 * module. Finally, it allows for optional "kernel threads".
175 register struct sysinit **sipp; /* system initialization*/
176 register struct sysinit **xipp; /* interior loop of sort*/
177 register struct sysinit *save; /* bubble*/
179 #if defined(VERBOSE_SYSINIT)
184 if (boothowto & RB_VERBOSE)
187 if (sysinit == NULL) {
188 sysinit = SET_BEGIN(sysinit_set);
189 sysinit_end = SET_LIMIT(sysinit_set);
194 * Perform a bubble sort of the system initialization objects by
195 * their subsystem (primary key) and order (secondary key).
197 for (sipp = sysinit; sipp < sysinit_end; sipp++) {
198 for (xipp = sipp + 1; xipp < sysinit_end; xipp++) {
199 if ((*sipp)->subsystem < (*xipp)->subsystem ||
200 ((*sipp)->subsystem == (*xipp)->subsystem &&
201 (*sipp)->order <= (*xipp)->order))
209 #if defined(VERBOSE_SYSINIT)
210 last = SI_SUB_COPYRIGHT;
213 printf("VERBOSE_SYSINIT: DDB not enabled, symbol lookups disabled.\n");
218 * Traverse the (now) ordered list of system initialization tasks.
219 * Perform each task, and continue on to the next task.
221 * The last item on the list is expected to be the scheduler,
222 * which will not return.
224 for (sipp = sysinit; sipp < sysinit_end; sipp++) {
226 if ((*sipp)->subsystem == SI_SUB_DUMMY)
227 continue; /* skip dummy task(s)*/
229 if ((*sipp)->subsystem == SI_SUB_DONE)
232 #if defined(VERBOSE_SYSINIT)
233 if ((*sipp)->subsystem > last) {
235 last = (*sipp)->subsystem;
236 printf("subsystem %x\n", last);
244 sym = db_search_symbol((vm_offset_t)(*sipp)->func,
245 DB_STGY_PROC, &offset);
246 db_symbol_values(sym, &name, NULL);
248 printf(" %s(%p)... ", name, (*sipp)->udata);
251 printf(" %p(%p)... ", (*sipp)->func,
257 (*((*sipp)->func))((*sipp)->udata);
259 #if defined(VERBOSE_SYSINIT)
264 /* Check off the one we're just done */
265 (*sipp)->subsystem = SI_SUB_DONE;
267 /* Check if we've installed more sysinit items via KLD */
268 if (newsysinit != NULL) {
269 if (sysinit != SET_BEGIN(sysinit_set))
270 free(sysinit, M_TEMP);
271 sysinit = newsysinit;
272 sysinit_end = newsysinit_end;
274 newsysinit_end = NULL;
279 panic("Shouldn't get here!");
285 ***************************************************************************
287 **** The following SYSINIT's belong elsewhere, but have not yet
290 ***************************************************************************
293 print_caddr_t(void *data)
295 printf("%s", (char *)data);
299 print_version(void *data __unused)
303 /* Strip a trailing newline from version. */
304 len = strlen(version);
305 while (len > 0 && version[len - 1] == '\n')
307 printf("%.*s %s\n", len, version, machine);
310 SYSINIT(announce, SI_SUB_COPYRIGHT, SI_ORDER_FIRST, print_caddr_t,
312 SYSINIT(trademark, SI_SUB_COPYRIGHT, SI_ORDER_SECOND, print_caddr_t,
314 SYSINIT(version, SI_SUB_COPYRIGHT, SI_ORDER_THIRD, print_version, NULL);
317 static char wit_warn[] =
318 "WARNING: WITNESS option enabled, expect reduced performance.\n";
319 SYSINIT(witwarn, SI_SUB_COPYRIGHT, SI_ORDER_THIRD + 1,
320 print_caddr_t, wit_warn);
321 SYSINIT(witwarn2, SI_SUB_RUN_SCHEDULER, SI_ORDER_THIRD + 1,
322 print_caddr_t, wit_warn);
326 static char diag_warn[] =
327 "WARNING: DIAGNOSTIC option enabled, expect reduced performance.\n";
328 SYSINIT(diagwarn, SI_SUB_COPYRIGHT, SI_ORDER_THIRD + 2,
329 print_caddr_t, diag_warn);
330 SYSINIT(diagwarn2, SI_SUB_RUN_SCHEDULER, SI_ORDER_THIRD + 2,
331 print_caddr_t, diag_warn);
335 null_fetch_syscall_args(struct thread *td __unused,
336 struct syscall_args *sa __unused)
339 panic("null_fetch_syscall_args");
343 null_set_syscall_retval(struct thread *td __unused, int error __unused)
346 panic("null_set_syscall_retval");
349 struct sysentvec null_sysvec = {
357 .sv_transtrap = NULL,
361 .sv_szsigcode = NULL,
362 .sv_prepsyscall = NULL,
365 .sv_imgact_try = NULL,
367 .sv_pagesize = PAGE_SIZE,
368 .sv_minuser = VM_MIN_ADDRESS,
369 .sv_maxuser = VM_MAXUSER_ADDRESS,
370 .sv_usrstack = USRSTACK,
371 .sv_psstrings = PS_STRINGS,
372 .sv_stackprot = VM_PROT_ALL,
373 .sv_copyout_strings = NULL,
378 .sv_set_syscall_retval = null_set_syscall_retval,
379 .sv_fetch_syscall_args = null_fetch_syscall_args,
380 .sv_syscallnames = NULL,
381 .sv_schedtail = NULL,
385 ***************************************************************************
387 **** The two following SYSINIT's are proc0 specific glue code. I am not
388 **** convinced that they can not be safely combined, but their order of
389 **** operation has been maintained as the same as the original init_main.c
392 **** These probably belong in init_proc.c or kern_proc.c, since they
393 **** deal with proc0 (the fork template process).
395 ***************************************************************************
399 proc0_init(void *dummy __unused)
403 vm_paddr_t pageablemem;
411 * Initialize magic number and osrel.
413 p->p_magic = P_MAGIC;
414 p->p_osrel = osreldate;
417 * Initialize thread and process structures.
419 procinit(); /* set up proc zone */
420 threadinit(); /* set up UMA zones */
423 * Initialise scheduler resources.
424 * Add scheduler specific parts to proc, thread as needed.
426 schedinit(); /* scheduler gets its house in order */
428 * Initialize sleep queue hash table
433 * additional VM structures
438 * Create process 0 (the swapper).
440 LIST_INSERT_HEAD(&allproc, p, p_list);
441 LIST_INSERT_HEAD(PIDHASH(0), p, p_hash);
442 mtx_init(&pgrp0.pg_mtx, "process group", NULL, MTX_DEF | MTX_DUPOK);
444 LIST_INSERT_HEAD(PGRPHASH(0), &pgrp0, pg_hash);
445 LIST_INIT(&pgrp0.pg_members);
446 LIST_INSERT_HEAD(&pgrp0.pg_members, p, p_pglist);
448 pgrp0.pg_session = &session0;
449 mtx_init(&session0.s_mtx, "session", NULL, MTX_DEF);
450 refcount_init(&session0.s_count, 1);
451 session0.s_leader = p;
453 p->p_sysent = &null_sysvec;
454 p->p_flag = P_SYSTEM | P_INMEM;
455 p->p_state = PRS_NORMAL;
456 knlist_init_mtx(&p->p_klist, &p->p_mtx);
457 STAILQ_INIT(&p->p_ktr);
459 td->td_tid = PID_MAX + 1;
460 LIST_INSERT_HEAD(TIDHASH(td->td_tid), td, td_hash);
461 td->td_state = TDS_RUNNING;
462 td->td_pri_class = PRI_TIMESHARE;
463 td->td_user_pri = PUSER;
464 td->td_base_user_pri = PUSER;
465 td->td_lend_user_pri = PRI_MAX;
466 td->td_priority = PVM;
467 td->td_base_pri = PVM;
469 td->td_flags = TDF_INMEM|TDP_KTHREAD;
470 td->td_cpuset = cpuset_thread0();
471 prison0.pr_cpuset = cpuset_ref(td->td_cpuset);
476 strncpy(p->p_comm, "kernel", sizeof (p->p_comm));
477 strncpy(td->td_name, "swapper", sizeof (td->td_name));
479 callout_init(&p->p_itcallout, CALLOUT_MPSAFE);
480 callout_init_mtx(&p->p_limco, &p->p_mtx, 0);
481 callout_init(&td->td_slpcallout, CALLOUT_MPSAFE);
483 /* Create credentials. */
484 p->p_ucred = crget();
485 p->p_ucred->cr_ngroups = 1; /* group 0 */
486 p->p_ucred->cr_uidinfo = uifind(0);
487 p->p_ucred->cr_ruidinfo = uifind(0);
488 p->p_ucred->cr_prison = &prison0;
489 p->p_ucred->cr_loginclass = loginclass_find("default");
491 audit_cred_kproc0(p->p_ucred);
494 mac_cred_create_swapper(p->p_ucred);
496 td->td_ucred = crhold(p->p_ucred);
498 /* Create sigacts. */
499 p->p_sigacts = sigacts_alloc();
501 /* Initialize signal state for process 0. */
504 /* Create the file descriptor table. */
505 p->p_fd = fdinit(NULL);
508 /* Create the limits structures. */
509 p->p_limit = lim_alloc();
510 for (i = 0; i < RLIM_NLIMITS; i++)
511 p->p_limit->pl_rlimit[i].rlim_cur =
512 p->p_limit->pl_rlimit[i].rlim_max = RLIM_INFINITY;
513 p->p_limit->pl_rlimit[RLIMIT_NOFILE].rlim_cur =
514 p->p_limit->pl_rlimit[RLIMIT_NOFILE].rlim_max = maxfiles;
515 p->p_limit->pl_rlimit[RLIMIT_NPROC].rlim_cur =
516 p->p_limit->pl_rlimit[RLIMIT_NPROC].rlim_max = maxproc;
517 p->p_limit->pl_rlimit[RLIMIT_DATA].rlim_cur = dfldsiz;
518 p->p_limit->pl_rlimit[RLIMIT_DATA].rlim_max = maxdsiz;
519 p->p_limit->pl_rlimit[RLIMIT_STACK].rlim_cur = dflssiz;
520 p->p_limit->pl_rlimit[RLIMIT_STACK].rlim_max = maxssiz;
521 /* Cast to avoid overflow on i386/PAE. */
522 pageablemem = ptoa((vm_paddr_t)cnt.v_free_count);
523 p->p_limit->pl_rlimit[RLIMIT_RSS].rlim_cur =
524 p->p_limit->pl_rlimit[RLIMIT_RSS].rlim_max = pageablemem;
525 p->p_limit->pl_rlimit[RLIMIT_MEMLOCK].rlim_cur = pageablemem / 3;
526 p->p_limit->pl_rlimit[RLIMIT_MEMLOCK].rlim_max = pageablemem;
527 p->p_cpulimit = RLIM_INFINITY;
529 p->p_stats = pstats_alloc();
531 /* Allocate a prototype map so we have something to fork. */
532 pmap_pinit0(vmspace_pmap(&vmspace0));
533 p->p_vmspace = &vmspace0;
534 vmspace0.vm_refcnt = 1;
537 * proc0 is not expected to enter usermode, so there is no special
538 * handling for sv_minuser here, like is done for exec_new_vmspace().
540 vm_map_init(&vmspace0.vm_map, vmspace_pmap(&vmspace0),
541 p->p_sysent->sv_minuser, p->p_sysent->sv_maxuser);
544 * Call the init and ctor for the new thread and proc. We wait
545 * to do this until all other structures are fairly sane.
547 EVENTHANDLER_INVOKE(process_init, p);
548 EVENTHANDLER_INVOKE(thread_init, td);
549 EVENTHANDLER_INVOKE(process_ctor, p);
550 EVENTHANDLER_INVOKE(thread_ctor, td);
553 * Charge root for one process.
555 (void)chgproccnt(p->p_ucred->cr_ruidinfo, 1, 0);
557 SYSINIT(p0init, SI_SUB_INTRINSIC, SI_ORDER_FIRST, proc0_init, NULL);
561 proc0_post(void *dummy __unused)
569 * Now we can look at the time, having had a chance to verify the
570 * time from the filesystem. Pretend that proc0 started now.
572 sx_slock(&allproc_lock);
573 FOREACH_PROC_IN_SYSTEM(p) {
574 microuptime(&p->p_stats->p_start);
576 rufetch(p, &ru); /* Clears thread stats */
578 p->p_rux.rux_runtime = 0;
579 p->p_rux.rux_uticks = 0;
580 p->p_rux.rux_sticks = 0;
581 p->p_rux.rux_iticks = 0;
582 FOREACH_THREAD_IN_PROC(p, td) {
586 sx_sunlock(&allproc_lock);
587 PCPU_SET(switchtime, cpu_ticks());
588 PCPU_SET(switchticks, ticks);
591 * Give the ``random'' number generator a thump.
594 srandom(ts.tv_sec ^ ts.tv_nsec);
596 SYSINIT(p0post, SI_SUB_INTRINSIC_POST, SI_ORDER_FIRST, proc0_post, NULL);
599 random_init(void *dummy __unused)
603 * After CPU has been started we have some randomness on most
604 * platforms via get_cyclecount(). For platforms that don't
605 * we will reseed random(9) in proc0_post() as well.
607 srandom(get_cyclecount());
609 SYSINIT(random, SI_SUB_RANDOM, SI_ORDER_FIRST, random_init, NULL);
612 ***************************************************************************
614 **** The following SYSINIT's and glue code should be moved to the
615 **** respective files on a per subsystem basis.
617 ***************************************************************************
622 ***************************************************************************
624 **** The following code probably belongs in another file, like
625 **** kern/init_init.c.
627 ***************************************************************************
631 * List of paths to try when searching for "init".
633 static char init_path[MAXPATHLEN] =
635 __XSTRING(INIT_PATH);
637 "/sbin/init:/sbin/oinit:/sbin/init.bak:/rescue/init:/stand/sysinstall";
639 SYSCTL_STRING(_kern, OID_AUTO, init_path, CTLFLAG_RD, init_path, 0,
640 "Path used to search the init process");
643 * Shutdown timeout of init(8).
644 * Unused within kernel, but used to control init(8), hence do not remove.
646 #ifndef INIT_SHUTDOWN_TIMEOUT
647 #define INIT_SHUTDOWN_TIMEOUT 120
649 static int init_shutdown_timeout = INIT_SHUTDOWN_TIMEOUT;
650 SYSCTL_INT(_kern, OID_AUTO, init_shutdown_timeout,
651 CTLFLAG_RW, &init_shutdown_timeout, 0, "Shutdown timeout of init(8). "
652 "Unused within kernel, but used to control init(8)");
655 * Start the initial user process; try exec'ing each pathname in init_path.
656 * The program is invoked with one argument containing the boot flags.
659 start_init(void *dummy)
662 struct execve_args args;
664 char *var, *path, *next, *s;
665 char *ucp, **uap, *arg0, *arg1;
679 * Need just enough stack to hold the faked-up "execve()" arguments.
681 addr = p->p_sysent->sv_usrstack - PAGE_SIZE;
682 if (vm_map_find(&p->p_vmspace->vm_map, NULL, 0, &addr, PAGE_SIZE,
683 FALSE, VM_PROT_ALL, VM_PROT_ALL, 0) != 0)
684 panic("init: couldn't allocate argument space");
685 p->p_vmspace->vm_maxsaddr = (caddr_t)addr;
686 p->p_vmspace->vm_ssize = 1;
688 if ((var = getenv("init_path")) != NULL) {
689 strlcpy(init_path, var, sizeof(init_path));
693 for (path = init_path; *path != '\0'; path = next) {
698 for (next = path; *next != '\0' && *next != ':'; next++)
701 printf("start_init: trying %.*s\n", (int)(next - path),
705 * Move out the boot flag argument.
708 ucp = (char *)p->p_sysent->sv_usrstack;
709 (void)subyte(--ucp, 0); /* trailing zero */
710 if (boothowto & RB_SINGLE) {
711 (void)subyte(--ucp, 's');
715 if (boothowto & RB_FASTBOOT) {
716 (void)subyte(--ucp, 'f');
722 (void)subyte(--ucp, 'C');
727 (void)subyte(--ucp, '-');
728 (void)subyte(--ucp, '-'); /* leading hyphen */
732 * Move out the file name (also arg 0).
734 (void)subyte(--ucp, 0);
735 for (s = next - 1; s >= path; s--)
736 (void)subyte(--ucp, *s);
740 * Move out the arg pointers.
742 uap = (char **)((intptr_t)ucp & ~(sizeof(intptr_t)-1));
743 (void)suword((caddr_t)--uap, (long)0); /* terminator */
744 (void)suword((caddr_t)--uap, (long)(intptr_t)arg1);
745 (void)suword((caddr_t)--uap, (long)(intptr_t)arg0);
748 * Point at the arguments.
755 * Now try to exec the program. If can't for any reason
756 * other than it doesn't exist, complain.
758 * Otherwise, return via fork_trampoline() all the way
759 * to user mode as init!
761 if ((error = execve(td, &args)) == 0) {
766 printf("exec %.*s: error %d\n", (int)(next - path),
769 printf("init: not found in path %s\n", init_path);
774 * Like kproc_create(), but runs in it's own address space.
775 * We do this early to reserve pid 1.
777 * Note special case - do not make it runnable yet. Other work
778 * in progress will change this more.
781 create_init(const void *udata __unused)
783 struct ucred *newcred, *oldcred;
786 error = fork1(&thread0, RFFDG | RFPROC | RFSTOPPED, 0, &initproc);
788 panic("cannot fork init: %d\n", error);
789 KASSERT(initproc->p_pid == 1, ("create_init: initproc->p_pid != 1"));
790 /* divorce init's credentials from the kernel's */
793 initproc->p_flag |= P_SYSTEM | P_INMEM;
794 oldcred = initproc->p_ucred;
795 crcopy(newcred, oldcred);
797 mac_cred_create_init(newcred);
800 audit_cred_proc1(newcred);
802 initproc->p_ucred = newcred;
803 PROC_UNLOCK(initproc);
805 cred_update_thread(FIRST_THREAD_IN_PROC(initproc));
806 cpu_set_fork_handler(FIRST_THREAD_IN_PROC(initproc), start_init, NULL);
808 SYSINIT(init, SI_SUB_CREATE_INIT, SI_ORDER_FIRST, create_init, NULL);
811 * Make it runnable now.
814 kick_init(const void *udata __unused)
818 td = FIRST_THREAD_IN_PROC(initproc);
821 sched_add(td, SRQ_BORING);
824 SYSINIT(kickinit, SI_SUB_KTHREAD_INIT, SI_ORDER_FIRST, kick_init, NULL);