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 int vfs_mountroot_devfs(void);
84 /* Components of the first process -- never freed. */
85 static struct session session0
;
86 static struct pgrp pgrp0
;
87 static struct sigacts sigacts0
;
88 static struct filedesc filedesc0
;
89 static struct plimit limit0
;
90 static struct vmspace vmspace0
;
91 struct proc
*initproc
;
94 struct thread thread0
;
97 extern struct user
*proc0paddr
;
98 extern int fallback_elf_brand
;
100 int boothowto
= 0; /* initialized so that it can be patched */
101 SYSCTL_INT(_debug
, OID_AUTO
, boothowto
, CTLFLAG_RD
, &boothowto
, 0, "");
104 * This ensures that there is at least one entry so that the sysinit_set
105 * symbol is not undefined. A subsystem ID of SI_SPECIAL_DUMMY is never
108 SYSINIT(placeholder
, SI_SPECIAL_DUMMY
, SI_ORDER_ANY
, NULL
, NULL
)
111 * The sysinit table itself. Items are checked off as the are run.
112 * If we want to register new sysinit types, add them to newsysinit.
114 SET_DECLARE(sysinit_set
, struct sysinit
);
115 struct sysinit
**sysinit
, **sysinit_end
;
116 struct sysinit
**newsysinit
, **newsysinit_end
;
120 * Merge a new sysinit set into the current set, reallocating it if
121 * necessary. This can only be called after malloc is running.
124 sysinit_add(struct sysinit
**set
, struct sysinit
**set_end
)
126 struct sysinit
**newset
;
127 struct sysinit
**sipp
;
128 struct sysinit
**xipp
;
131 count
= set_end
- set
;
133 count
+= newsysinit_end
- newsysinit
;
135 count
+= sysinit_end
- sysinit
;
136 newset
= kmalloc(count
* sizeof(*sipp
), M_TEMP
, M_WAITOK
);
139 for (sipp
= newsysinit
; sipp
< newsysinit_end
; sipp
++)
142 for (sipp
= sysinit
; sipp
< sysinit_end
; sipp
++)
145 for (sipp
= set
; sipp
< set_end
; sipp
++)
148 kfree(newsysinit
, M_TEMP
);
150 newsysinit_end
= newset
+ count
;
154 * Callbacks from machine-dependant startup code (e.g. init386) to set
155 * up low level entities related to cpu #0's globaldata.
157 * Called from very low level boot code.
160 mi_proc0init(struct globaldata
*gd
, struct user
*proc0paddr
)
162 lwkt_init_thread(&thread0
, proc0paddr
, LWKT_THREAD_STACK
, 0, gd
);
163 lwkt_set_comm(&thread0
, "thread0");
164 RB_INIT(&proc0
.p_lwp_tree
);
165 spin_init(&proc0
.p_spin
);
166 proc0
.p_lasttid
= 0; /* +1 = next TID */
167 lwp_rb_tree_RB_INSERT(&proc0
.p_lwp_tree
, &lwp0
);
168 lwp0
.lwp_thread
= &thread0
;
169 lwp0
.lwp_proc
= &proc0
;
170 proc0
.p_usched
= usched_init();
171 lwp0
.lwp_cpumask
= 0xFFFFFFFF;
172 varsymset_init(&proc0
.p_varsymset
, NULL
);
173 thread0
.td_flags
|= TDF_RUNNING
;
174 thread0
.td_proc
= &proc0
;
175 thread0
.td_lwp
= &lwp0
;
176 thread0
.td_switch
= cpu_lwkt_switch
;
177 lwkt_schedule_self(curthread
);
181 * System startup; initialize the world, create process 0, mount root
182 * filesystem, and fork to create init and pagedaemon. Most of the
183 * hard work is done in the lower-level initialization routines including
184 * startup(), which does memory initialization and autoconfiguration.
186 * This allows simple addition of new kernel subsystems that require
187 * boot time initialization. It also allows substitution of subsystem
188 * (for instance, a scheduler, kernel profiler, or VM system) by object
189 * module. Finally, it allows for optional "kernel threads".
194 struct sysinit
*sip
; /* system initialization*/
195 struct sysinit
**sipp
; /* system initialization*/
196 struct sysinit
**xipp
; /* interior loop of sort*/
197 struct sysinit
*save
; /* bubble*/
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
)
237 (*(sip
->func
))(sip
->udata
);
239 /* Check off the one we're just done */
240 sip
->subsystem
= SI_SPECIAL_DONE
;
242 /* Check if we've installed more sysinit items via KLD */
243 if (newsysinit
!= NULL
) {
244 if (sysinit
!= SET_BEGIN(sysinit_set
))
245 kfree(sysinit
, M_TEMP
);
246 sysinit
= newsysinit
;
247 sysinit_end
= newsysinit_end
;
249 newsysinit_end
= NULL
;
254 panic("Shouldn't get here!");
260 ***************************************************************************
262 **** The following SYSINIT's belong elsewhere, but have not yet
265 ***************************************************************************
268 print_caddr_t(void *data __unused
)
270 kprintf("%s", (char *)data
);
272 SYSINIT(announce
, SI_BOOT1_COPYRIGHT
, SI_ORDER_FIRST
, print_caddr_t
, copyright
)
275 * Leave the critical section that protected us from spurious interrupts
276 * so device probes work.
279 leavecrit(void *dummy __unused
)
281 MachIntrABI
.finalize();
283 MachIntrABI
.cleanup();
285 KKASSERT(!IN_CRITICAL_SECT(curthread
));
288 kprintf("Leaving critical section, allowing interrupts\n");
290 SYSINIT(leavecrit
, SI_BOOT2_LEAVE_CRIT
, SI_ORDER_ANY
, leavecrit
, NULL
)
293 * This is called after the threading system is up and running,
294 * including the softclock, clock interrupts, and SMP.
297 tsleepworks(void *dummy __unused
)
299 tsleep_now_works
= 1;
301 SYSINIT(tsleepworks
, SI_BOOT2_FINISH_SMP
, SI_ORDER_SECOND
, tsleepworks
, NULL
)
304 * This is called after devices have configured. Tell the kernel we are
305 * no longer in cold boot.
308 endofcoldboot(void *dummy __unused
)
312 SYSINIT(endofcoldboot
, SI_SUB_ISWARM
, SI_ORDER_ANY
, endofcoldboot
, NULL
)
315 ***************************************************************************
317 **** The two following SYSINT's are proc0 specific glue code. I am not
318 **** convinced that they can not be safely combined, but their order of
319 **** operation has been maintained as the same as the original init_main.c
322 **** These probably belong in init_proc.c or kern_proc.c, since they
323 **** deal with proc0 (the fork template process).
325 ***************************************************************************
329 proc0_init(void *dummy __unused
)
338 * Initialize process and pgrp structures.
343 * additional VM structures
348 * Create process 0 (the swapper).
350 LIST_INSERT_HEAD(&allproc
, p
, p_list
);
352 LIST_INSERT_HEAD(PGRPHASH(0), &pgrp0
, pg_hash
);
353 LIST_INIT(&pgrp0
.pg_members
);
354 LIST_INSERT_HEAD(&pgrp0
.pg_members
, p
, p_pglist
);
356 pgrp0
.pg_session
= &session0
;
357 session0
.s_count
= 1;
358 session0
.s_leader
= p
;
360 p
->p_sysent
= &aout_sysvec
;
362 p
->p_flag
= P_SYSTEM
;
364 lp
->lwp_stat
= LSRUN
;
366 p
->p_rtprio
.type
= RTP_PRIO_NORMAL
;
367 p
->p_rtprio
.prio
= 0;
368 lp
->lwp_rtprio
= p
->p_rtprio
;
373 bcopy("swapper", p
->p_comm
, sizeof ("swapper"));
374 bcopy("swapper", thread0
.td_comm
, sizeof ("swapper"));
376 /* Create credentials. */
377 p
->p_ucred
= crget();
378 p
->p_ucred
->cr_ruidinfo
= uifind(0);
379 p
->p_ucred
->cr_ngroups
= 1; /* group 0 */
380 p
->p_ucred
->cr_uidinfo
= uifind(0);
383 p
->p_ucred
->cr_prison
= NULL
;
385 /* Create sigacts. */
386 p
->p_sigacts
= &sigacts0
;
387 p
->p_sigacts
->ps_refcnt
= 1;
389 /* Initialize signal state for process 0. */
392 /* Create the file descriptor table. */
393 fdinit_bootstrap(p
, &filedesc0
, cmask
);
395 /* Create the limits structures. */
396 plimit_init0(&limit0
);
397 p
->p_limit
= &limit0
;
399 /* Allocate a prototype map so we have something to fork. */
400 pmap_pinit0(vmspace_pmap(&vmspace0
));
401 p
->p_vmspace
= &vmspace0
;
402 lp
->lwp_vmspace
= p
->p_vmspace
;
403 sysref_init(&vmspace0
.vm_sysref
, &vmspace_sysref_class
);
404 vm_map_init(&vmspace0
.vm_map
,
405 round_page(VM_MIN_USER_ADDRESS
),
406 trunc_page(VM_MAX_USER_ADDRESS
),
407 vmspace_pmap(&vmspace0
));
408 sysref_activate(&vmspace0
.vm_sysref
);
411 * Charge root for one process.
413 (void)chgproccnt(p
->p_ucred
->cr_uidinfo
, 1, 0);
416 SYSINIT(p0init
, SI_BOOT2_PROC0
, SI_ORDER_FIRST
, proc0_init
, NULL
)
418 static int proc0_post_callback(struct proc
*p
, void *data __unused
);
422 proc0_post(void *dummy __unused
)
427 * Now we can look at the time, having had a chance to verify the
428 * time from the file system. Pretend that proc0 started now.
430 allproc_scan(proc0_post_callback
, NULL
);
433 * Give the ``random'' number generator a thump.
434 * XXX: Does read_random() contain enough bits to be used here ?
437 skrandom(ts
.tv_sec
^ ts
.tv_nsec
);
441 proc0_post_callback(struct proc
*p
, void *data __unused
)
443 microtime(&p
->p_start
);
447 SYSINIT(p0post
, SI_SUB_PROC0_POST
, SI_ORDER_FIRST
, proc0_post
, NULL
)
450 ***************************************************************************
452 **** The following SYSINIT's and glue code should be moved to the
453 **** respective files on a per subsystem basis.
455 ***************************************************************************
460 ***************************************************************************
462 **** The following code probably belongs in another file, like
463 **** kern/init_init.c.
465 ***************************************************************************
469 * List of paths to try when searching for "init".
471 static char init_path
[MAXPATHLEN
] =
473 __XSTRING(INIT_PATH
);
475 "/sbin/init:/sbin/oinit:/sbin/init.bak";
477 SYSCTL_STRING(_kern
, OID_AUTO
, init_path
, CTLFLAG_RD
, init_path
, 0, "");
480 * Start the initial user process; try exec'ing each pathname in init_path.
481 * The program is invoked with one argument containing the boot flags.
483 * The MP lock is held on entry.
486 start_init(void *dummy
, struct trapframe
*frame
)
489 struct execve_args args
;
491 char *var
, *path
, *next
, *s
;
492 char *ucp
, **uap
, *arg0
, *arg1
;
500 lp
= ONLY_LWP_IN_PROC(p
);
502 /* Get the vnode for '/'. Set p->p_fd->fd_cdir to reference it. */
503 mp
= mountlist_boot_getfirst();
504 if (VFS_ROOT(mp
, &vp
))
505 panic("cannot find root vnode");
506 if (mp
->mnt_ncmountpt
.ncp
== NULL
) {
507 cache_allocroot(&mp
->mnt_ncmountpt
, mp
, vp
);
508 cache_unlock(&mp
->mnt_ncmountpt
); /* leave ref intact */
510 p
->p_fd
->fd_cdir
= vp
;
511 vref(p
->p_fd
->fd_cdir
);
512 p
->p_fd
->fd_rdir
= vp
;
513 vref(p
->p_fd
->fd_rdir
);
514 vfs_cache_setroot(vp
, cache_hold(&mp
->mnt_ncmountpt
));
515 vn_unlock(vp
); /* leave ref intact */
516 cache_copy(&mp
->mnt_ncmountpt
, &p
->p_fd
->fd_ncdir
);
517 cache_copy(&mp
->mnt_ncmountpt
, &p
->p_fd
->fd_nrdir
);
519 kprintf("Mounting devfs\n");
520 vfs_mountroot_devfs();
523 * Need just enough stack to hold the faked-up "execve()" arguments.
525 addr
= trunc_page(USRSTACK
- PAGE_SIZE
);
526 error
= vm_map_find(&p
->p_vmspace
->vm_map
, NULL
, 0, &addr
, PAGE_SIZE
,
529 VM_PROT_ALL
, VM_PROT_ALL
,
532 panic("init: couldn't allocate argument space");
533 p
->p_vmspace
->vm_maxsaddr
= (caddr_t
)addr
;
534 p
->p_vmspace
->vm_ssize
= 1;
536 if ((var
= kgetenv("init_path")) != NULL
) {
537 strncpy(init_path
, var
, sizeof init_path
);
538 init_path
[sizeof init_path
- 1] = 0;
540 if ((var
= kgetenv("kern.fallback_elf_brand")) != NULL
)
541 fallback_elf_brand
= strtol(var
, NULL
, 0);
543 for (path
= init_path
; *path
!= '\0'; path
= next
) {
548 for (next
= path
; *next
!= '\0' && *next
!= ':'; next
++)
551 kprintf("start_init: trying %.*s\n", (int)(next
- path
),
555 * Move out the boot flag argument.
558 ucp
= (char *)USRSTACK
;
559 (void)subyte(--ucp
, 0); /* trailing zero */
560 if (boothowto
& RB_SINGLE
) {
561 (void)subyte(--ucp
, 's');
565 if (boothowto
& RB_FASTBOOT
) {
566 (void)subyte(--ucp
, 'f');
572 (void)subyte(--ucp
, 'C');
576 (void)subyte(--ucp
, '-');
577 (void)subyte(--ucp
, '-'); /* leading hyphen */
581 * Move out the file name (also arg 0).
583 (void)subyte(--ucp
, 0);
584 for (s
= next
- 1; s
>= path
; s
--)
585 (void)subyte(--ucp
, *s
);
589 * Move out the arg pointers.
591 uap
= (char **)((intptr_t)ucp
& ~(sizeof(intptr_t)-1));
592 (void)suword((caddr_t
)--uap
, (long)0); /* terminator */
593 (void)suword((caddr_t
)--uap
, (long)(intptr_t)arg1
);
594 (void)suword((caddr_t
)--uap
, (long)(intptr_t)arg0
);
597 * Point at the arguments.
604 * Now try to exec the program. If can't for any reason
605 * other than it doesn't exist, complain.
607 * Otherwise, return via fork_trampoline() all the way
608 * to user mode as init!
610 * WARNING! We may have been moved to another cpu after
611 * acquiring the current user process designation. The
612 * MP lock will migrate with us though so we still have to
615 if ((error
= sys_execve(&args
)) == 0) {
617 lp
->lwp_proc
->p_usched
->acquire_curproc(lp
);
621 kprintf("exec %.*s: error %d\n", (int)(next
- path
),
624 kprintf("init: not found in path %s\n", init_path
);
629 * Like kthread_create(), but runs in it's own address space.
630 * We do this early to reserve pid 1.
632 * Note special case - do not make it runnable yet. Other work
633 * in progress will change this more.
636 create_init(const void *udata __unused
)
642 error
= fork1(&lwp0
, RFFDG
| RFPROC
, &initproc
);
644 panic("cannot fork init: %d", error
);
645 initproc
->p_flag
|= P_SYSTEM
;
646 lp
= ONLY_LWP_IN_PROC(initproc
);
647 cpu_set_fork_handler(lp
, start_init
, NULL
);
650 SYSINIT(init
, SI_SUB_CREATE_INIT
, SI_ORDER_FIRST
, create_init
, NULL
)
653 * Make it runnable now.
656 kick_init(const void *udata __unused
)
658 start_forked_proc(&lwp0
, initproc
);
660 SYSINIT(kickinit
, SI_SUB_KTHREAD_INIT
, SI_ORDER_FIRST
, kick_init
, NULL
)
663 * Machine independant globaldata initialization
665 * WARNING! Called from early boot, 'mycpu' may not work yet.
668 mi_gdinit(struct globaldata
*gd
, int cpuid
)
670 TAILQ_INIT(&gd
->gd_systimerq
);
671 gd
->gd_sysid_alloc
= cpuid
; /* prime low bits for cpu lookup */
672 gd
->gd_cpuid
= cpuid
;
673 gd
->gd_cpumask
= (cpumask_t
)1 << cpuid
;
675 vm_map_entry_reserve_cpu_init(gd
);