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
42 * $FreeBSD: src/sys/kern/init_main.c,v 1.134.2.8 2003/06/06 20:21:32 tegge Exp $
43 * $DragonFly: src/sys/kern/init_main.c,v 1.87 2008/06/07 11:37:23 mneumann Exp $
46 #include "opt_init_path.h"
48 #include <sys/param.h>
50 #include <sys/filedesc.h>
51 #include <sys/kernel.h>
52 #include <sys/mount.h>
53 #include <sys/sysctl.h>
55 #include <sys/resourcevar.h>
56 #include <sys/signalvar.h>
57 #include <sys/systm.h>
58 #include <sys/vnode.h>
59 #include <sys/sysent.h>
60 #include <sys/reboot.h>
61 #include <sys/sysproto.h>
62 #include <sys/vmmeter.h>
63 #include <sys/unistd.h>
64 #include <sys/malloc.h>
65 #include <sys/machintr.h>
66 #include <sys/file2.h>
67 #include <sys/thread2.h>
68 #include <sys/sysref2.h>
69 #include <sys/spinlock2.h>
71 #include <machine/cpu.h>
74 #include <vm/vm_param.h>
77 #include <vm/vm_map.h>
78 #include <vm/vm_extern.h>
80 #include <sys/copyright.h>
82 /* Components of the first process -- never freed. */
83 static struct session session0
;
84 static struct pgrp pgrp0
;
85 static struct sigacts sigacts0
;
86 static struct filedesc filedesc0
;
87 static struct plimit limit0
;
88 static struct vmspace vmspace0
;
89 struct proc
*initproc
;
92 struct thread thread0
;
95 extern struct user
*proc0paddr
;
96 extern int fallback_elf_brand
;
98 int boothowto
= 0; /* initialized so that it can be patched */
99 SYSCTL_INT(_debug
, OID_AUTO
, boothowto
, CTLFLAG_RD
, &boothowto
, 0, "");
102 * This ensures that there is at least one entry so that the sysinit_set
103 * symbol is not undefined. A subsystem ID of SI_SPECIAL_DUMMY is never
106 SYSINIT(placeholder
, SI_SPECIAL_DUMMY
, SI_ORDER_ANY
, NULL
, NULL
)
109 * The sysinit table itself. Items are checked off as the are run.
110 * If we want to register new sysinit types, add them to newsysinit.
112 SET_DECLARE(sysinit_set
, struct sysinit
);
113 struct sysinit
**sysinit
, **sysinit_end
;
114 struct sysinit
**newsysinit
, **newsysinit_end
;
118 * Merge a new sysinit set into the current set, reallocating it if
119 * necessary. This can only be called after malloc is running.
122 sysinit_add(struct sysinit
**set
, struct sysinit
**set_end
)
124 struct sysinit
**newset
;
125 struct sysinit
**sipp
;
126 struct sysinit
**xipp
;
129 count
= set_end
- set
;
131 count
+= newsysinit_end
- newsysinit
;
133 count
+= sysinit_end
- sysinit
;
134 newset
= kmalloc(count
* sizeof(*sipp
), M_TEMP
, M_WAITOK
);
137 for (sipp
= newsysinit
; sipp
< newsysinit_end
; sipp
++)
140 for (sipp
= sysinit
; sipp
< sysinit_end
; sipp
++)
143 for (sipp
= set
; sipp
< set_end
; sipp
++)
146 kfree(newsysinit
, M_TEMP
);
148 newsysinit_end
= newset
+ count
;
152 * Callbacks from machine-dependant startup code (e.g. init386) to set
153 * up low level entities related to cpu #0's globaldata.
155 * Called from very low level boot code.
158 mi_proc0init(struct globaldata
*gd
, struct user
*proc0paddr
)
160 lwkt_init_thread(&thread0
, proc0paddr
, LWKT_THREAD_STACK
, 0, gd
);
161 lwkt_set_comm(&thread0
, "thread0");
162 RB_INIT(&proc0
.p_lwp_tree
);
163 spin_init(&proc0
.p_spin
);
164 proc0
.p_lasttid
= 0; /* +1 = next TID */
165 lwp_rb_tree_RB_INSERT(&proc0
.p_lwp_tree
, &lwp0
);
166 lwp0
.lwp_thread
= &thread0
;
167 lwp0
.lwp_proc
= &proc0
;
168 proc0
.p_usched
= usched_init();
169 lwp0
.lwp_cpumask
= 0xFFFFFFFF;
170 varsymset_init(&proc0
.p_varsymset
, NULL
);
171 thread0
.td_flags
|= TDF_RUNNING
;
172 thread0
.td_proc
= &proc0
;
173 thread0
.td_lwp
= &lwp0
;
174 thread0
.td_switch
= cpu_heavy_switch
; /* YYY eventually LWKT */
175 lwkt_schedule_self(curthread
);
179 * System startup; initialize the world, create process 0, mount root
180 * filesystem, and fork to create init and pagedaemon. Most of the
181 * hard work is done in the lower-level initialization routines including
182 * startup(), which does memory initialization and autoconfiguration.
184 * This allows simple addition of new kernel subsystems that require
185 * boot time initialization. It also allows substitution of subsystem
186 * (for instance, a scheduler, kernel profiler, or VM system) by object
187 * module. Finally, it allows for optional "kernel threads".
192 struct sysinit
*sip
; /* system initialization*/
193 struct sysinit
**sipp
; /* system initialization*/
194 struct sysinit
**xipp
; /* interior loop of sort*/
195 struct sysinit
*save
; /* bubble*/
197 //kprintf0("mi_startup\n");
199 if (sysinit
== NULL
) {
200 sysinit
= SET_BEGIN(sysinit_set
);
201 sysinit_end
= SET_LIMIT(sysinit_set
);
206 * Perform a bubble sort of the system initialization objects by
207 * their subsystem (primary key) and order (secondary key).
209 for (sipp
= sysinit
; sipp
< sysinit_end
; sipp
++) {
210 for (xipp
= sipp
+ 1; xipp
< sysinit_end
; xipp
++) {
211 if ((*sipp
)->subsystem
< (*xipp
)->subsystem
||
212 ((*sipp
)->subsystem
== (*xipp
)->subsystem
&&
213 (*sipp
)->order
<= (*xipp
)->order
))
222 * Traverse the (now) ordered list of system initialization tasks.
223 * Perform each task, and continue on to the next task.
225 * The last item on the list is expected to be the scheduler,
226 * which will not return.
228 for (sipp
= sysinit
; sipp
< sysinit_end
; sipp
++) {
230 if (sip
->subsystem
== SI_SPECIAL_DUMMY
)
231 continue; /* skip dummy task(s)*/
233 if (sip
->subsystem
== SI_SPECIAL_DONE
)
236 //kprintf0("!%s!", sip->name);
239 (*(sip
->func
))(sip
->udata
);
241 /* Check off the one we're just done */
242 sip
->subsystem
= SI_SPECIAL_DONE
;
244 /* Check if we've installed more sysinit items via KLD */
245 if (newsysinit
!= NULL
) {
246 if (sysinit
!= SET_BEGIN(sysinit_set
))
247 kfree(sysinit
, M_TEMP
);
248 sysinit
= newsysinit
;
249 sysinit_end
= newsysinit_end
;
251 newsysinit_end
= NULL
;
256 panic("Shouldn't get here!");
262 ***************************************************************************
264 **** The following SYSINIT's belong elsewhere, but have not yet
267 ***************************************************************************
270 print_caddr_t(void *data __unused
)
272 kprintf("%s", (char *)data
);
274 SYSINIT(announce
, SI_BOOT1_COPYRIGHT
, SI_ORDER_FIRST
, print_caddr_t
, copyright
)
277 * Leave the critical section that protected us from spurious interrupts
278 * so device probes work.
281 leavecrit(void *dummy __unused
)
283 MachIntrABI
.finalize();
285 MachIntrABI
.cleanup();
287 KKASSERT(!IN_CRITICAL_SECT(curthread
));
290 kprintf("Leaving critical section, allowing interrupts\n");
292 SYSINIT(leavecrit
, SI_BOOT2_LEAVE_CRIT
, SI_ORDER_ANY
, leavecrit
, NULL
)
295 * This is called after the threading system is up and running,
296 * including the softclock, clock interrupts, and SMP.
299 tsleepworks(void *dummy __unused
)
301 tsleep_now_works
= 1;
303 SYSINIT(tsleepworks
, SI_BOOT2_FINISH_SMP
, SI_ORDER_SECOND
, tsleepworks
, NULL
)
306 * This is called after devices have configured. Tell the kernel we are
307 * no longer in cold boot.
310 endofcoldboot(void *dummy __unused
)
314 SYSINIT(endofcoldboot
, SI_SUB_ISWARM
, SI_ORDER_ANY
, endofcoldboot
, NULL
)
317 ***************************************************************************
319 **** The two following SYSINT's are proc0 specific glue code. I am not
320 **** convinced that they can not be safely combined, but their order of
321 **** operation has been maintained as the same as the original init_main.c
324 **** These probably belong in init_proc.c or kern_proc.c, since they
325 **** deal with proc0 (the fork template process).
327 ***************************************************************************
331 proc0_init(void *dummy __unused
)
340 * Initialize process and pgrp structures.
345 * additional VM structures
350 * Create process 0 (the swapper).
352 LIST_INSERT_HEAD(&allproc
, p
, p_list
);
354 LIST_INSERT_HEAD(PGRPHASH(0), &pgrp0
, pg_hash
);
355 LIST_INIT(&pgrp0
.pg_members
);
356 LIST_INSERT_HEAD(&pgrp0
.pg_members
, p
, p_pglist
);
358 pgrp0
.pg_session
= &session0
;
359 session0
.s_count
= 1;
360 session0
.s_leader
= p
;
362 p
->p_sysent
= &aout_sysvec
;
364 p
->p_flag
= P_SYSTEM
;
366 lp
->lwp_stat
= LSRUN
;
368 p
->p_rtprio
.type
= RTP_PRIO_NORMAL
;
369 p
->p_rtprio
.prio
= 0;
370 lp
->lwp_rtprio
= p
->p_rtprio
;
375 bcopy("swapper", p
->p_comm
, sizeof ("swapper"));
376 bcopy("swapper", thread0
.td_comm
, sizeof ("swapper"));
378 /* Create credentials. */
379 p
->p_ucred
= crget();
380 p
->p_ucred
->cr_ruidinfo
= uifind(0);
381 p
->p_ucred
->cr_ngroups
= 1; /* group 0 */
382 p
->p_ucred
->cr_uidinfo
= uifind(0);
385 p
->p_ucred
->cr_prison
= NULL
;
387 /* Create sigacts. */
388 p
->p_sigacts
= &sigacts0
;
389 p
->p_sigacts
->ps_refcnt
= 1;
391 /* Initialize signal state for process 0. */
394 /* Create the file descriptor table. */
395 fdinit_bootstrap(p
, &filedesc0
, cmask
);
397 /* Create the limits structures. */
398 plimit_init0(&limit0
);
399 p
->p_limit
= &limit0
;
401 /* Allocate a prototype map so we have something to fork. */
402 pmap_pinit0(vmspace_pmap(&vmspace0
));
403 p
->p_vmspace
= &vmspace0
;
404 lp
->lwp_vmspace
= p
->p_vmspace
;
405 sysref_init(&vmspace0
.vm_sysref
, &vmspace_sysref_class
);
406 vm_map_init(&vmspace0
.vm_map
,
407 round_page(VM_MIN_USER_ADDRESS
),
408 trunc_page(VM_MAX_USER_ADDRESS
),
409 vmspace_pmap(&vmspace0
));
410 sysref_activate(&vmspace0
.vm_sysref
);
413 * Charge root for one process.
415 (void)chgproccnt(p
->p_ucred
->cr_uidinfo
, 1, 0);
418 SYSINIT(p0init
, SI_BOOT2_PROC0
, SI_ORDER_FIRST
, proc0_init
, NULL
)
420 static int proc0_post_callback(struct proc
*p
, void *data __unused
);
424 proc0_post(void *dummy __unused
)
429 * Now we can look at the time, having had a chance to verify the
430 * time from the file system. Pretend that proc0 started now.
432 allproc_scan(proc0_post_callback
, NULL
);
435 * Give the ``random'' number generator a thump.
436 * XXX: Does read_random() contain enough bits to be used here ?
439 skrandom(ts
.tv_sec
^ ts
.tv_nsec
);
443 proc0_post_callback(struct proc
*p
, void *data __unused
)
445 microtime(&p
->p_start
);
449 SYSINIT(p0post
, SI_SUB_PROC0_POST
, SI_ORDER_FIRST
, proc0_post
, NULL
)
452 ***************************************************************************
454 **** The following SYSINIT's and glue code should be moved to the
455 **** respective files on a per subsystem basis.
457 ***************************************************************************
462 ***************************************************************************
464 **** The following code probably belongs in another file, like
465 **** kern/init_init.c.
467 ***************************************************************************
471 * List of paths to try when searching for "init".
473 static char init_path
[MAXPATHLEN
] =
475 __XSTRING(INIT_PATH
);
477 "/sbin/init:/sbin/oinit:/sbin/init.bak";
479 SYSCTL_STRING(_kern
, OID_AUTO
, init_path
, CTLFLAG_RD
, init_path
, 0, "");
482 * Start the initial user process; try exec'ing each pathname in init_path.
483 * The program is invoked with one argument containing the boot flags.
485 * The MP lock is held on entry.
488 start_init(void *dummy
, struct trapframe
*frame
)
491 struct execve_args args
;
493 char *var
, *path
, *next
, *s
;
494 char *ucp
, **uap
, *arg0
, *arg1
;
502 lp
= ONLY_LWP_IN_PROC(p
);
504 /* Get the vnode for '/'. Set p->p_fd->fd_cdir to reference it. */
505 mp
= mountlist_boot_getfirst();
506 if (VFS_ROOT(mp
, &vp
))
507 panic("cannot find root vnode");
508 if (mp
->mnt_ncmountpt
.ncp
== NULL
) {
509 cache_allocroot(&mp
->mnt_ncmountpt
, mp
, vp
);
510 cache_unlock(&mp
->mnt_ncmountpt
); /* leave ref intact */
512 p
->p_fd
->fd_cdir
= vp
;
513 vref(p
->p_fd
->fd_cdir
);
514 p
->p_fd
->fd_rdir
= vp
;
515 vref(p
->p_fd
->fd_rdir
);
516 vfs_cache_setroot(vp
, cache_hold(&mp
->mnt_ncmountpt
));
517 vn_unlock(vp
); /* leave ref intact */
518 cache_copy(&mp
->mnt_ncmountpt
, &p
->p_fd
->fd_ncdir
);
519 cache_copy(&mp
->mnt_ncmountpt
, &p
->p_fd
->fd_nrdir
);
522 * Need just enough stack to hold the faked-up "execve()" arguments.
524 addr
= trunc_page(USRSTACK
- PAGE_SIZE
);
525 error
= vm_map_find(&p
->p_vmspace
->vm_map
, NULL
, 0, &addr
, PAGE_SIZE
,
528 VM_PROT_ALL
, VM_PROT_ALL
,
531 panic("init: couldn't allocate argument space");
532 p
->p_vmspace
->vm_maxsaddr
= (caddr_t
)addr
;
533 p
->p_vmspace
->vm_ssize
= 1;
535 if ((var
= kgetenv("init_path")) != NULL
) {
536 strncpy(init_path
, var
, sizeof init_path
);
537 init_path
[sizeof init_path
- 1] = 0;
539 if ((var
= kgetenv("kern.fallback_elf_brand")) != NULL
)
540 fallback_elf_brand
= strtol(var
, NULL
, 0);
542 for (path
= init_path
; *path
!= '\0'; path
= next
) {
547 for (next
= path
; *next
!= '\0' && *next
!= ':'; next
++)
550 kprintf("start_init: trying %.*s\n", (int)(next
- path
),
554 * Move out the boot flag argument.
557 ucp
= (char *)USRSTACK
;
558 (void)subyte(--ucp
, 0); /* trailing zero */
559 if (boothowto
& RB_SINGLE
) {
560 (void)subyte(--ucp
, 's');
564 if (boothowto
& RB_FASTBOOT
) {
565 (void)subyte(--ucp
, 'f');
571 (void)subyte(--ucp
, 'C');
575 (void)subyte(--ucp
, '-');
576 (void)subyte(--ucp
, '-'); /* leading hyphen */
580 * Move out the file name (also arg 0).
582 (void)subyte(--ucp
, 0);
583 for (s
= next
- 1; s
>= path
; s
--)
584 (void)subyte(--ucp
, *s
);
588 * Move out the arg pointers.
590 uap
= (char **)((intptr_t)ucp
& ~(sizeof(intptr_t)-1));
591 (void)suword((caddr_t
)--uap
, (long)0); /* terminator */
592 (void)suword((caddr_t
)--uap
, (long)(intptr_t)arg1
);
593 (void)suword((caddr_t
)--uap
, (long)(intptr_t)arg0
);
596 * Point at the arguments.
603 * Now try to exec the program. If can't for any reason
604 * other than it doesn't exist, complain.
606 * Otherwise, return via fork_trampoline() all the way
607 * to user mode as init!
609 * WARNING! We may have been moved to another cpu after
610 * acquiring the current user process designation. The
611 * MP lock will migrate with us though so we still have to
614 if ((error
= sys_execve(&args
)) == 0) {
616 lp
->lwp_proc
->p_usched
->acquire_curproc(lp
);
620 kprintf("exec %.*s: error %d\n", (int)(next
- path
),
623 kprintf("init: not found in path %s\n", init_path
);
628 * Like kthread_create(), but runs in it's own address space.
629 * We do this early to reserve pid 1.
631 * Note special case - do not make it runnable yet. Other work
632 * in progress will change this more.
635 create_init(const void *udata __unused
)
641 error
= fork1(&lwp0
, RFFDG
| RFPROC
, &initproc
);
643 panic("cannot fork init: %d", error
);
644 initproc
->p_flag
|= P_SYSTEM
;
645 lp
= ONLY_LWP_IN_PROC(initproc
);
646 cpu_set_fork_handler(lp
, start_init
, NULL
);
649 SYSINIT(init
, SI_SUB_CREATE_INIT
, SI_ORDER_FIRST
, create_init
, NULL
)
652 * Make it runnable now.
655 kick_init(const void *udata __unused
)
657 start_forked_proc(&lwp0
, initproc
);
659 SYSINIT(kickinit
, SI_SUB_KTHREAD_INIT
, SI_ORDER_FIRST
, kick_init
, NULL
)
662 * Machine independant globaldata initialization
664 * WARNING! Called from early boot, 'mycpu' may not work yet.
667 mi_gdinit(struct globaldata
*gd
, int cpuid
)
669 TAILQ_INIT(&gd
->gd_systimerq
);
670 gd
->gd_sysid_alloc
= cpuid
; /* prime low bits for cpu lookup */
671 gd
->gd_cpuid
= cpuid
;
672 gd
->gd_cpumask
= (cpumask_t
)1 << cpuid
;
674 vm_map_entry_reserve_cpu_init(gd
);