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 1.70 2006/12/28 21:24:01 dillon 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/file2.h>
66 #include <sys/thread2.h>
68 #include <machine/cpu.h>
71 #include <vm/vm_param.h>
74 #include <vm/vm_map.h>
76 #include <sys/copyright.h>
78 /* Components of the first process -- never freed. */
79 static struct session session0;
80 static struct pgrp pgrp0;
81 static struct procsig procsig0;
82 static struct filedesc filedesc0;
83 static struct plimit limit0;
84 static struct vmspace vmspace0;
85 struct proc *initproc;
87 struct thread thread0;
90 extern struct user *proc0paddr;
91 extern int fallback_elf_brand;
93 int boothowto = 0; /* initialized so that it can be patched */
94 SYSCTL_INT(_debug, OID_AUTO, boothowto, CTLFLAG_RD, &boothowto, 0, "");
97 * This ensures that there is at least one entry so that the sysinit_set
98 * symbol is not undefined. A sybsystem ID of SI_SUB_DUMMY is never
101 SYSINIT(placeholder, SI_SUB_DUMMY, SI_ORDER_ANY, NULL, NULL)
104 * The sysinit table itself. Items are checked off as the are run.
105 * If we want to register new sysinit types, add them to newsysinit.
107 SET_DECLARE(sysinit_set, struct sysinit);
108 struct sysinit **sysinit, **sysinit_end;
109 struct sysinit **newsysinit, **newsysinit_end;
113 * Merge a new sysinit set into the current set, reallocating it if
114 * necessary. This can only be called after malloc is running.
117 sysinit_add(struct sysinit **set, struct sysinit **set_end)
119 struct sysinit **newset;
120 struct sysinit **sipp;
121 struct sysinit **xipp;
124 count = set_end - set;
126 count += newsysinit_end - newsysinit;
128 count += sysinit_end - sysinit;
129 newset = kmalloc(count * sizeof(*sipp), M_TEMP, M_WAITOK);
131 panic("cannot malloc for sysinit");
134 for (sipp = newsysinit; sipp < newsysinit_end; sipp++)
137 for (sipp = sysinit; sipp < sysinit_end; sipp++)
140 for (sipp = set; sipp < set_end; sipp++)
143 kfree(newsysinit, M_TEMP);
145 newsysinit_end = newset + count;
149 * Callbacks from machine-dependant startup code (e.g. init386) to set
150 * up low level entities related to cpu #0's globaldata.
152 * Called from very low level boot code.
155 mi_proc0init(struct globaldata *gd, struct user *proc0paddr)
157 lwkt_init_thread(&thread0, proc0paddr, LWKT_THREAD_STACK, 0, gd);
158 lwkt_set_comm(&thread0, "thread0");
159 proc0.p_addr = (void *)thread0.td_kstack;
160 LIST_INIT(&proc0.p_lwps);
161 LIST_INSERT_HEAD(&proc0.p_lwps, &proc0.p_lwp, lwp_list);
162 proc0.p_lwp.lwp_thread = &thread0;
163 proc0.p_lwp.lwp_proc = &proc0;
164 proc0.p_usched = usched_init();
165 proc0.p_lwp.lwp_cpumask = 0xFFFFFFFF;
166 varsymset_init(&proc0.p_varsymset, NULL);
167 thread0.td_flags |= TDF_RUNNING;
168 thread0.td_proc = &proc0;
169 thread0.td_lwp = &proc0.p_lwp;
170 thread0.td_switch = cpu_heavy_switch; /* YYY eventually LWKT */
174 * System startup; initialize the world, create process 0, mount root
175 * filesystem, and fork to create init and pagedaemon. Most of the
176 * hard work is done in the lower-level initialization routines including
177 * startup(), which does memory initialization and autoconfiguration.
179 * This allows simple addition of new kernel subsystems that require
180 * boot time initialization. It also allows substitution of subsystem
181 * (for instance, a scheduler, kernel profiler, or VM system) by object
182 * module. Finally, it allows for optional "kernel threads".
187 struct sysinit *sip; /* system initialization*/
188 struct sysinit **sipp; /* system initialization*/
189 struct sysinit **xipp; /* interior loop of sort*/
190 struct sysinit *save; /* bubble*/
192 if (sysinit == NULL) {
193 sysinit = SET_BEGIN(sysinit_set);
194 sysinit_end = SET_LIMIT(sysinit_set);
199 * Perform a bubble sort of the system initialization objects by
200 * their subsystem (primary key) and order (secondary key).
202 for (sipp = sysinit; sipp < sysinit_end; sipp++) {
203 for (xipp = sipp + 1; xipp < sysinit_end; xipp++) {
204 if ((*sipp)->subsystem < (*xipp)->subsystem ||
205 ((*sipp)->subsystem == (*xipp)->subsystem &&
206 (*sipp)->order <= (*xipp)->order))
215 * Traverse the (now) ordered list of system initialization tasks.
216 * Perform each task, and continue on to the next task.
218 * The last item on the list is expected to be the scheduler,
219 * which will not return.
221 for (sipp = sysinit; sipp < sysinit_end; sipp++) {
223 if (sip->subsystem == SI_SUB_DUMMY)
224 continue; /* skip dummy task(s)*/
226 if (sip->subsystem == SI_SUB_DONE)
230 (*(sip->func))(sip->udata);
232 /* Check off the one we're just done */
233 sip->subsystem = SI_SUB_DONE;
235 /* Check if we've installed more sysinit items via KLD */
236 if (newsysinit != NULL) {
237 if (sysinit != SET_BEGIN(sysinit_set))
238 kfree(sysinit, M_TEMP);
239 sysinit = newsysinit;
240 sysinit_end = newsysinit_end;
242 newsysinit_end = NULL;
247 panic("Shouldn't get here!");
253 ***************************************************************************
255 **** The following SYSINIT's belong elsewhere, but have not yet
258 ***************************************************************************
261 print_caddr_t(void *data __unused)
263 kprintf("%s", (char *)data);
265 SYSINIT(announce, SI_SUB_COPYRIGHT, SI_ORDER_FIRST, print_caddr_t, copyright)
268 * Leave the critical section that protected us from spurious interrupts
269 * so device probes work.
272 leavecrit(void *dummy __unused)
275 KKASSERT(!IN_CRITICAL_SECT(curthread));
277 kprintf("Leaving critical section, allowing interrupts\n");
279 SYSINIT(leavecrit, SI_SUB_LEAVE_CRIT, SI_ORDER_ANY, leavecrit, NULL)
282 ***************************************************************************
284 **** The two following SYSINT's are proc0 specific glue code. I am not
285 **** convinced that they can not be safely combined, but their order of
286 **** operation has been maintained as the same as the original init_main.c
289 **** These probably belong in init_proc.c or kern_proc.c, since they
290 **** deal with proc0 (the fork template process).
292 ***************************************************************************
296 proc0_init(void *dummy __unused)
302 lp = &proc0.p_lwp; /* XXX lwp to be: lwp0 */
305 * Initialize process and pgrp structures.
310 * additional VM structures
315 * Create process 0 (the swapper).
317 LIST_INSERT_HEAD(&allproc, p, p_list);
319 LIST_INSERT_HEAD(PGRPHASH(0), &pgrp0, pg_hash);
320 LIST_INIT(&pgrp0.pg_members);
321 LIST_INSERT_HEAD(&pgrp0.pg_members, p, p_pglist);
323 pgrp0.pg_session = &session0;
324 session0.s_count = 1;
325 session0.s_leader = p;
327 p->p_sysent = &aout_sysvec;
329 p->p_flag = P_SYSTEM;
332 p->p_rtprio.type = RTP_PRIO_NORMAL;
333 p->p_rtprio.prio = 0;
334 p->p_lwp.lwp_rtprio = p->p_rtprio;
339 bcopy("swapper", p->p_comm, sizeof ("swapper"));
340 bcopy("swapper", thread0.td_comm, sizeof ("swapper"));
342 /* Create credentials. */
343 p->p_ucred = crget();
344 p->p_ucred->cr_ruidinfo = uifind(0);
345 p->p_ucred->cr_ngroups = 1; /* group 0 */
346 p->p_ucred->cr_uidinfo = uifind(0);
349 p->p_ucred->cr_prison = NULL;
351 /* Create procsig. */
352 p->p_procsig = &procsig0;
353 p->p_procsig->ps_refcnt = 1;
355 /* Initialize signal state for process 0. */
358 /* Create the file descriptor table. */
359 fdinit_bootstrap(p, &filedesc0, cmask);
361 /* Create the limits structures. */
362 plimit_init0(&limit0);
363 p->p_limit = &limit0;
365 /* Allocate a prototype map so we have something to fork. */
366 pmap_pinit0(vmspace_pmap(&vmspace0));
367 p->p_vmspace = &vmspace0;
368 vmspace0.vm_refcnt = 1;
369 vm_map_init(&vmspace0.vm_map,
370 round_page(VM_MIN_USER_ADDRESS),
371 trunc_page(VM_MAX_USER_ADDRESS),
372 vmspace_pmap(&vmspace0));
375 * We continue to place resource usage info and signal
376 * actions in the user struct so they're pageable.
378 p->p_stats = &p->p_addr->u_stats;
379 p->p_sigacts = &p->p_addr->u_sigacts;
382 * Charge root for one process.
384 (void)chgproccnt(p->p_ucred->cr_uidinfo, 1, 0);
387 SYSINIT(p0init, SI_SUB_INTRINSIC, SI_ORDER_FIRST, proc0_init, NULL)
389 static int proc0_post_callback(struct proc *p, void *data __unused);
393 proc0_post(void *dummy __unused)
398 * Now we can look at the time, having had a chance to verify the
399 * time from the file system. Pretend that proc0 started now.
401 allproc_scan(proc0_post_callback, NULL);
404 * Give the ``random'' number generator a thump.
405 * XXX: Does read_random() contain enough bits to be used here ?
408 skrandom(ts.tv_sec ^ ts.tv_nsec);
412 proc0_post_callback(struct proc *p, void *data __unused)
414 microtime(&p->p_start);
418 SYSINIT(p0post, SI_SUB_INTRINSIC_POST, SI_ORDER_FIRST, proc0_post, NULL)
421 ***************************************************************************
423 **** The following SYSINIT's and glue code should be moved to the
424 **** respective files on a per subsystem basis.
426 ***************************************************************************
431 ***************************************************************************
433 **** The following code probably belongs in another file, like
434 **** kern/init_init.c.
436 ***************************************************************************
440 * List of paths to try when searching for "init".
442 static char init_path[MAXPATHLEN] =
444 __XSTRING(INIT_PATH);
446 "/sbin/init:/sbin/oinit:/sbin/init.bak:/stand/sysinstall";
448 SYSCTL_STRING(_kern, OID_AUTO, init_path, CTLFLAG_RD, init_path, 0, "");
451 * Start the initial user process; try exec'ing each pathname in init_path.
452 * The program is invoked with one argument containing the boot flags.
454 * The MP lock is held on entry.
457 start_init(void *dummy)
460 struct execve_args args;
462 char *var, *path, *next, *s;
463 char *ucp, **uap, *arg0, *arg1;
471 KKASSERT(p->p_nthreads == 1);
473 lp = LIST_FIRST(&p->p_lwps);
475 /* Get the vnode for '/'. Set p->p_fd->fd_cdir to reference it. */
476 mp = mountlist_boot_getfirst();
477 if (VFS_ROOT(mp, &vp))
478 panic("cannot find root vnode");
479 if (mp->mnt_ncmountpt.ncp == NULL) {
480 cache_allocroot(&mp->mnt_ncmountpt, mp, vp);
481 cache_unlock(&mp->mnt_ncmountpt); /* leave ref intact */
483 p->p_fd->fd_cdir = vp;
484 vref(p->p_fd->fd_cdir);
485 p->p_fd->fd_rdir = vp;
486 vref(p->p_fd->fd_rdir);
487 vfs_cache_setroot(vp, cache_hold(&mp->mnt_ncmountpt));
488 vn_unlock(vp); /* leave ref intact */
489 cache_copy(&mp->mnt_ncmountpt, &p->p_fd->fd_ncdir);
490 cache_copy(&mp->mnt_ncmountpt, &p->p_fd->fd_nrdir);
493 * Need just enough stack to hold the faked-up "execve()" arguments.
495 addr = trunc_page(USRSTACK - PAGE_SIZE);
496 error = vm_map_find(&p->p_vmspace->vm_map, NULL, 0, &addr, PAGE_SIZE,
499 VM_PROT_ALL, VM_PROT_ALL,
502 panic("init: couldn't allocate argument space");
503 p->p_vmspace->vm_maxsaddr = (caddr_t)addr;
504 p->p_vmspace->vm_ssize = 1;
506 if ((var = kgetenv("init_path")) != NULL) {
507 strncpy(init_path, var, sizeof init_path);
508 init_path[sizeof init_path - 1] = 0;
510 if ((var = kgetenv("kern.fallback_elf_brand")) != NULL)
511 fallback_elf_brand = strtol(var, NULL, 0);
513 for (path = init_path; *path != '\0'; path = next) {
518 for (next = path; *next != '\0' && *next != ':'; next++)
521 kprintf("start_init: trying %.*s\n", (int)(next - path),
525 * Move out the boot flag argument.
528 ucp = (char *)USRSTACK;
529 (void)subyte(--ucp, 0); /* trailing zero */
530 if (boothowto & RB_SINGLE) {
531 (void)subyte(--ucp, 's');
535 if (boothowto & RB_FASTBOOT) {
536 (void)subyte(--ucp, 'f');
542 (void)subyte(--ucp, 'C');
546 (void)subyte(--ucp, '-');
547 (void)subyte(--ucp, '-'); /* leading hyphen */
551 * Move out the file name (also arg 0).
553 (void)subyte(--ucp, 0);
554 for (s = next - 1; s >= path; s--)
555 (void)subyte(--ucp, *s);
559 * Move out the arg pointers.
561 uap = (char **)((intptr_t)ucp & ~(sizeof(intptr_t)-1));
562 (void)suword((caddr_t)--uap, (long)0); /* terminator */
563 (void)suword((caddr_t)--uap, (long)(intptr_t)arg1);
564 (void)suword((caddr_t)--uap, (long)(intptr_t)arg0);
567 * Point at the arguments.
574 * Now try to exec the program. If can't for any reason
575 * other than it doesn't exist, complain.
577 * Otherwise, return via fork_trampoline() all the way
578 * to user mode as init!
580 * WARNING! We may have been moved to another cpu after
581 * acquiring the current user process designation. The
582 * MP lock will migrate with us though so we still have to
585 if ((error = sys_execve(&args)) == 0) {
587 lp->lwp_proc->p_usched->acquire_curproc(lp);
591 kprintf("exec %.*s: error %d\n", (int)(next - path),
594 kprintf("init: not found in path %s\n", init_path);
599 * Like kthread_create(), but runs in it's own address space.
600 * We do this early to reserve pid 1.
602 * Note special case - do not make it runnable yet. Other work
603 * in progress will change this more.
606 create_init(const void *udata __unused)
612 error = fork1(&proc0.p_lwp, RFFDG | RFPROC, &initproc);
614 panic("cannot fork init: %d", error);
615 initproc->p_flag |= P_SYSTEM;
616 lp = LIST_FIRST(&initproc->p_lwps);
617 cpu_set_fork_handler(lp, start_init, NULL);
620 SYSINIT(init,SI_SUB_CREATE_INIT, SI_ORDER_FIRST, create_init, NULL)
623 * Make it runnable now.
626 kick_init(const void *udata __unused)
628 start_forked_proc(&proc0.p_lwp, initproc);
630 SYSINIT(kickinit,SI_SUB_KTHREAD_INIT, SI_ORDER_FIRST, kick_init, NULL)
633 * Machine independant globaldata initialization
635 * WARNING! Called from early boot, 'mycpu' may not work yet.
638 mi_gdinit(struct globaldata *gd, int cpuid)
640 TAILQ_INIT(&gd->gd_tdfreeq); /* for pmap_{new,dispose}_thread() */
641 TAILQ_INIT(&gd->gd_systimerq);
642 gd->gd_cpuid = cpuid;
643 gd->gd_cpumask = (cpumask_t)1 << cpuid;
645 vm_map_entry_reserve_cpu_init(gd);