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 $
45 #include "opt_init_path.h"
47 #include <sys/param.h>
49 #include <sys/filedesc.h>
50 #include <sys/kernel.h>
51 #include <sys/mount.h>
52 #include <sys/sysctl.h>
54 #include <sys/resourcevar.h>
55 #include <sys/signalvar.h>
56 #include <sys/systm.h>
57 #include <sys/vnode.h>
58 #include <sys/sysent.h>
59 #include <sys/reboot.h>
60 #include <sys/sysproto.h>
61 #include <sys/vmmeter.h>
62 #include <sys/unistd.h>
63 #include <sys/malloc.h>
64 #include <sys/machintr.h>
66 #include <sys/refcount.h>
67 #include <sys/file2.h>
68 #include <sys/thread2.h>
69 #include <sys/sysref2.h>
70 #include <sys/spinlock2.h>
71 #include <sys/mplock2.h>
73 #include <machine/cpu.h>
76 #include <vm/vm_param.h>
79 #include <vm/vm_map.h>
80 #include <vm/vm_extern.h>
82 #include <sys/copyright.h>
84 int vfs_mountroot_devfs(void);
86 /* Components of the first process -- never freed. */
87 static struct session session0;
88 static struct pgrp pgrp0;
89 static struct sigacts sigacts0;
90 static struct filedesc filedesc0;
91 static struct plimit limit0;
92 static struct vmspace vmspace0;
93 struct proc *initproc;
96 struct thread thread0;
100 cpumask_t usched_global_cpumask;
101 extern struct user *proc0paddr;
103 int boothowto = 0; /* initialized so that it can be patched */
104 SYSCTL_INT(_debug, OID_AUTO, boothowto, CTLFLAG_RD, &boothowto, 0,
105 "Reboot flags, from console subsystem");
106 SYSCTL_ULONG(_kern, OID_AUTO, usched_global_cpumask, CTLFLAG_RW,
107 &usched_global_cpumask, 0, "global user scheduler cpumask");
110 * This ensures that there is at least one entry so that the sysinit_set
111 * symbol is not undefined. A subsystem ID of SI_SPECIAL_DUMMY is never
114 SYSINIT(placeholder, SI_SPECIAL_DUMMY, SI_ORDER_ANY, NULL, NULL)
117 * The sysinit table itself. Items are checked off as the are run.
118 * If we want to register new sysinit types, add them to newsysinit.
120 SET_DECLARE(sysinit_set, struct sysinit);
121 struct sysinit **sysinit, **sysinit_end;
122 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 = kmalloc(count * sizeof(*sipp), M_TEMP, M_WAITOK);
145 for (sipp = newsysinit; sipp < newsysinit_end; sipp++)
148 for (sipp = sysinit; sipp < sysinit_end; sipp++)
151 for (sipp = set; sipp < set_end; sipp++)
154 kfree(newsysinit, M_TEMP);
156 newsysinit_end = newset + count;
160 * Callbacks from machine-dependant startup code (e.g. init386) to set
161 * up low level entities related to cpu #0's globaldata.
163 * Called from very low level boot code.
166 mi_proc0init(struct globaldata *gd, struct user *proc0paddr)
168 lwkt_init_thread(&thread0, proc0paddr, LWKT_THREAD_STACK, 0, gd);
169 lwkt_set_comm(&thread0, "thread0");
170 RB_INIT(&proc0.p_lwp_tree);
171 spin_init(&proc0.p_spin);
172 lwkt_token_init(&proc0.p_token, "iproc");
173 proc0.p_lasttid = 0; /* +1 = next TID */
174 lwp_rb_tree_RB_INSERT(&proc0.p_lwp_tree, &lwp0);
175 lwp0.lwp_thread = &thread0;
176 lwp0.lwp_proc = &proc0;
177 proc0.p_usched = usched_init();
178 lwp0.lwp_cpumask = (cpumask_t)-1;
179 varsymset_init(&proc0.p_varsymset, NULL);
180 thread0.td_flags |= TDF_RUNNING;
181 thread0.td_proc = &proc0;
182 thread0.td_lwp = &lwp0;
183 thread0.td_switch = cpu_lwkt_switch;
184 lwkt_schedule_self(curthread);
188 * System startup; initialize the world, create process 0, mount root
189 * filesystem, and fork to create init and pagedaemon. Most of the
190 * hard work is done in the lower-level initialization routines including
191 * startup(), which does memory initialization and autoconfiguration.
193 * This allows simple addition of new kernel subsystems that require
194 * boot time initialization. It also allows substitution of subsystem
195 * (for instance, a scheduler, kernel profiler, or VM system) by object
196 * module. Finally, it allows for optional "kernel threads".
201 struct sysinit *sip; /* system initialization*/
202 struct sysinit **sipp; /* system initialization*/
203 struct sysinit **xipp; /* interior loop of sort*/
204 struct sysinit *save; /* bubble*/
206 if (sysinit == NULL) {
207 sysinit = SET_BEGIN(sysinit_set);
208 #if defined(__amd64__) && defined(_KERNEL_VIRTUAL)
210 * XXX For whatever reason, on 64-bit vkernels
211 * the value of sysinit obtained from the
212 * linker set is wrong.
214 if ((long)sysinit % 8 != 0) {
215 kprintf("Fixing sysinit value...\n");
216 sysinit = (long)sysinit + 4;
219 sysinit_end = SET_LIMIT(sysinit_set);
221 #if defined(__amd64__) && defined(_KERNEL_VIRTUAL)
222 KKASSERT((long)sysinit % 8 == 0);
227 * Perform a bubble sort of the system initialization objects by
228 * their subsystem (primary key) and order (secondary key).
230 for (sipp = sysinit; sipp < sysinit_end; sipp++) {
231 for (xipp = sipp + 1; xipp < sysinit_end; xipp++) {
232 if ((*sipp)->subsystem < (*xipp)->subsystem ||
233 ((*sipp)->subsystem == (*xipp)->subsystem &&
234 (*sipp)->order <= (*xipp)->order))
243 * Traverse the (now) ordered list of system initialization tasks.
244 * Perform each task, and continue on to the next task.
246 * The last item on the list is expected to be the scheduler,
247 * which will not return.
249 for (sipp = sysinit; sipp < sysinit_end; sipp++) {
251 if (sip->subsystem == SI_SPECIAL_DUMMY)
252 continue; /* skip dummy task(s)*/
254 if (sip->subsystem == SI_SPECIAL_DONE)
258 (*(sip->func))(sip->udata);
260 /* Check off the one we're just done */
261 sip->subsystem = SI_SPECIAL_DONE;
263 /* Check if we've installed more sysinit items via KLD */
264 if (newsysinit != NULL) {
265 if (sysinit != SET_BEGIN(sysinit_set))
266 kfree(sysinit, M_TEMP);
267 sysinit = newsysinit;
268 sysinit_end = newsysinit_end;
270 newsysinit_end = NULL;
275 panic("Shouldn't get here!");
281 ***************************************************************************
283 **** The following SYSINIT's belong elsewhere, but have not yet
286 ***************************************************************************
289 print_caddr_t(void *data __unused)
291 kprintf("%s", (char *)data);
293 SYSINIT(announce, SI_BOOT1_COPYRIGHT, SI_ORDER_FIRST, print_caddr_t, copyright)
296 * Leave the critical section that protected us from spurious interrupts
297 * so device probes work.
300 leavecrit(void *dummy __unused)
302 MachIntrABI.stabilize();
304 MachIntrABI.cleanup();
306 KKASSERT(!IN_CRITICAL_SECT(curthread));
309 kprintf("Leaving critical section, allowing interrupts\n");
311 SYSINIT(leavecrit, SI_BOOT2_LEAVE_CRIT, SI_ORDER_ANY, leavecrit, NULL)
314 * This is called after the threading system is up and running,
315 * including the softclock, clock interrupts, and SMP.
318 tsleepworks(void *dummy __unused)
320 tsleep_now_works = 1;
322 SYSINIT(tsleepworks, SI_BOOT2_FINISH_SMP, SI_ORDER_SECOND, tsleepworks, NULL)
325 * This is called after devices have configured. Tell the kernel we are
326 * no longer in cold boot.
329 endofcoldboot(void *dummy __unused)
333 SYSINIT(endofcoldboot, SI_SUB_ISWARM, SI_ORDER_ANY, endofcoldboot, NULL)
336 ***************************************************************************
338 **** The two following SYSINT's are proc0 specific glue code. I am not
339 **** convinced that they can not be safely combined, but their order of
340 **** operation has been maintained as the same as the original init_main.c
343 **** These probably belong in init_proc.c or kern_proc.c, since they
344 **** deal with proc0 (the fork template process).
346 ***************************************************************************
350 proc0_init(void *dummy __unused)
361 p->p_osrel = osreldate;
364 * Initialize process and pgrp structures.
369 * additional VM structures
374 * Create process 0 (the swapper).
376 LIST_INSERT_HEAD(&allproc, p, p_list);
377 LIST_INSERT_HEAD(PGRPHASH(0), &pgrp0, pg_hash);
378 LIST_INIT(&pgrp0.pg_members);
379 lwkt_token_init(&pgrp0.pg_token, "pgrp0");
380 refcount_init(&pgrp0.pg_refs, 1);
381 lockinit(&pgrp0.pg_lock, "pgwt0", 0, 0);
382 LIST_INSERT_HEAD(&pgrp0.pg_members, p, p_pglist);
384 pgrp0.pg_session = &session0;
385 session0.s_count = 1;
386 session0.s_leader = p;
391 p->p_sysent = &aout_sysvec;
393 p->p_flag = P_SYSTEM;
395 lp->lwp_stat = LSRUN;
397 p->p_rtprio.type = RTP_PRIO_NORMAL;
398 p->p_rtprio.prio = 0;
399 lp->lwp_rtprio = p->p_rtprio;
404 bcopy("swapper", p->p_comm, sizeof ("swapper"));
405 bcopy("swapper", thread0.td_comm, sizeof ("swapper"));
407 /* Create credentials. */
408 p->p_ucred = crget();
409 p->p_ucred->cr_ruidinfo = uifind(0);
410 p->p_ucred->cr_ngroups = 1; /* group 0 */
411 p->p_ucred->cr_uidinfo = uifind(0);
412 thread0.td_ucred = crhold(p->p_ucred); /* bootstrap fork1() */
415 p->p_ucred->cr_prison = NULL;
417 /* Create sigacts. */
418 p->p_sigacts = &sigacts0;
419 refcount_init(&p->p_sigacts->ps_refcnt, 1);
421 /* Initialize signal state for process 0. */
424 /* Create the file descriptor table. */
425 fdinit_bootstrap(p, &filedesc0, cmask);
427 /* Create the limits structures. */
428 plimit_init0(&limit0);
429 p->p_limit = &limit0;
431 /* Allocate a prototype map so we have something to fork. */
432 pmap_pinit0(vmspace_pmap(&vmspace0));
433 p->p_vmspace = &vmspace0;
434 lp->lwp_vmspace = p->p_vmspace;
435 sysref_init(&vmspace0.vm_sysref, &vmspace_sysref_class);
436 vm_map_init(&vmspace0.vm_map,
437 round_page(VM_MIN_USER_ADDRESS),
438 trunc_page(VM_MAX_USER_ADDRESS),
439 vmspace_pmap(&vmspace0));
440 sysref_activate(&vmspace0.vm_sysref);
442 kqueue_init(&lwp0.lwp_kqueue, &filedesc0);
445 * Charge root for one process.
447 (void)chgproccnt(p->p_ucred->cr_uidinfo, 1, 0);
450 SYSINIT(p0init, SI_BOOT2_PROC0, SI_ORDER_FIRST, proc0_init, NULL)
452 static int proc0_post_callback(struct proc *p, void *data __unused);
456 proc0_post(void *dummy __unused)
461 * Now we can look at the time, having had a chance to verify the
462 * time from the file system. Pretend that proc0 started now.
464 allproc_scan(proc0_post_callback, NULL);
467 * Give the ``random'' number generator a thump.
468 * XXX: Does read_random() contain enough bits to be used here ?
471 skrandom(ts.tv_sec ^ ts.tv_nsec);
475 proc0_post_callback(struct proc *p, void *data __unused)
477 microtime(&p->p_start);
481 SYSINIT(p0post, SI_SUB_PROC0_POST, SI_ORDER_FIRST, proc0_post, NULL)
484 ***************************************************************************
486 **** The following SYSINIT's and glue code should be moved to the
487 **** respective files on a per subsystem basis.
489 ***************************************************************************
494 ***************************************************************************
496 **** The following code probably belongs in another file, like
497 **** kern/init_init.c.
499 ***************************************************************************
503 * List of paths to try when searching for "init".
505 static char init_path[MAXPATHLEN] =
507 __XSTRING(INIT_PATH);
509 "/sbin/init:/sbin/oinit:/sbin/init.bak";
511 SYSCTL_STRING(_kern, OID_AUTO, init_path, CTLFLAG_RD, init_path, 0, "");
514 * Start the initial user process; try exec'ing each pathname in init_path.
515 * The program is invoked with one argument containing the boot flags.
518 start_init(void *dummy, struct trapframe *frame)
521 struct execve_args args;
523 char *var, *path, *next, *s;
524 char *ucp, **uap, *arg0, *arg1;
532 * This is passed in by the bootloader
534 env = kgetenv("kernelname");
536 strlcpy(kernelname, env, sizeof(kernelname));
539 * The MP lock is not held on entry. We release it before
540 * returning to userland.
545 lp = ONLY_LWP_IN_PROC(p);
547 /* Get the vnode for '/'. Set p->p_fd->fd_cdir to reference it. */
548 mp = mountlist_boot_getfirst();
549 if (VFS_ROOT(mp, &vp))
550 panic("cannot find root vnode");
551 if (mp->mnt_ncmountpt.ncp == NULL) {
552 cache_allocroot(&mp->mnt_ncmountpt, mp, vp);
553 cache_unlock(&mp->mnt_ncmountpt); /* leave ref intact */
555 p->p_fd->fd_cdir = vp;
556 vref(p->p_fd->fd_cdir);
557 p->p_fd->fd_rdir = vp;
558 vref(p->p_fd->fd_rdir);
559 vfs_cache_setroot(vp, cache_hold(&mp->mnt_ncmountpt));
560 vn_unlock(vp); /* leave ref intact */
561 cache_copy(&mp->mnt_ncmountpt, &p->p_fd->fd_ncdir);
562 cache_copy(&mp->mnt_ncmountpt, &p->p_fd->fd_nrdir);
564 kprintf("Mounting devfs\n");
565 vfs_mountroot_devfs();
568 * Need just enough stack to hold the faked-up "execve()" arguments.
570 addr = trunc_page(USRSTACK - PAGE_SIZE);
571 error = vm_map_find(&p->p_vmspace->vm_map, NULL, 0, &addr,
572 PAGE_SIZE, PAGE_SIZE,
573 FALSE, VM_MAPTYPE_NORMAL,
574 VM_PROT_ALL, VM_PROT_ALL,
577 panic("init: couldn't allocate argument space");
578 p->p_vmspace->vm_maxsaddr = (caddr_t)addr;
579 p->p_vmspace->vm_ssize = 1;
581 if ((var = kgetenv("init_path")) != NULL) {
582 strncpy(init_path, var, sizeof init_path);
583 init_path[sizeof init_path - 1] = 0;
586 for (path = init_path; *path != '\0'; path = next) {
591 for (next = path; *next != '\0' && *next != ':'; next++)
594 kprintf("start_init: trying %.*s\n", (int)(next - path),
598 * Move out the boot flag argument.
601 ucp = (char *)USRSTACK;
602 (void)subyte(--ucp, 0); /* trailing zero */
603 if (boothowto & RB_SINGLE) {
604 (void)subyte(--ucp, 's');
608 if (boothowto & RB_FASTBOOT) {
609 (void)subyte(--ucp, 'f');
615 (void)subyte(--ucp, 'C');
619 (void)subyte(--ucp, '-');
620 (void)subyte(--ucp, '-'); /* leading hyphen */
624 * Move out the file name (also arg 0).
626 (void)subyte(--ucp, 0);
627 for (s = next - 1; s >= path; s--)
628 (void)subyte(--ucp, *s);
632 * Move out the arg pointers.
634 uap = (char **)((intptr_t)ucp & ~(sizeof(intptr_t)-1));
635 (void)suword((caddr_t)--uap, (long)0); /* terminator */
636 (void)suword((caddr_t)--uap, (long)(intptr_t)arg1);
637 (void)suword((caddr_t)--uap, (long)(intptr_t)arg0);
640 * Point at the arguments.
647 * Now try to exec the program. If can't for any reason
648 * other than it doesn't exist, complain.
650 * Otherwise, return via fork_trampoline() all the way
651 * to user mode as init!
653 * WARNING! We may have been moved to another cpu after
654 * acquiring the current user process designation. The
655 * MP lock will migrate with us though so we still have to
658 if ((error = sys_execve(&args)) == 0) {
660 lp->lwp_proc->p_usched->acquire_curproc(lp);
664 kprintf("exec %.*s: error %d\n", (int)(next - path),
667 kprintf("init: not found in path %s\n", init_path);
672 * Like kthread_create(), but runs in it's own address space.
673 * We do this early to reserve pid 1.
675 * Note special case - do not make it runnable yet. Other work
676 * in progress will change this more.
679 create_init(const void *udata __unused)
685 error = fork1(&lwp0, RFFDG | RFPROC, &initproc);
687 panic("cannot fork init: %d", error);
688 initproc->p_flag |= P_SYSTEM;
689 lp = ONLY_LWP_IN_PROC(initproc);
690 cpu_set_fork_handler(lp, start_init, NULL);
693 SYSINIT(init, SI_SUB_CREATE_INIT, SI_ORDER_FIRST, create_init, NULL)
696 * Make it runnable now.
699 kick_init(const void *udata __unused)
701 start_forked_proc(&lwp0, initproc);
703 SYSINIT(kickinit, SI_SUB_KTHREAD_INIT, SI_ORDER_FIRST, kick_init, NULL)
706 * Machine independant globaldata initialization
708 * WARNING! Called from early boot, 'mycpu' may not work yet.
711 mi_gdinit(struct globaldata *gd, int cpuid)
713 TAILQ_INIT(&gd->gd_systimerq);
714 gd->gd_sysid_alloc = cpuid; /* prime low bits for cpu lookup */
715 gd->gd_cpuid = cpuid;
716 gd->gd_cpumask = CPUMASK(cpuid);
718 vm_map_entry_reserve_cpu_init(gd);
720 usched_global_cpumask |= CPUMASK(cpuid);