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>
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;
102 extern int fallback_elf_brand;
104 int boothowto = 0; /* initialized so that it can be patched */
105 SYSCTL_INT(_debug, OID_AUTO, boothowto, CTLFLAG_RD, &boothowto, 0,
106 "Reboot flags, from console subsystem");
107 SYSCTL_ULONG(_kern, OID_AUTO, usched_global_cpumask, CTLFLAG_RW,
108 &usched_global_cpumask, 0, "global user scheduler cpumask");
111 * This ensures that there is at least one entry so that the sysinit_set
112 * symbol is not undefined. A subsystem ID of SI_SPECIAL_DUMMY is never
115 SYSINIT(placeholder, SI_SPECIAL_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;
127 * Merge a new sysinit set into the current set, reallocating it if
128 * necessary. This can only be called after malloc is running.
131 sysinit_add(struct sysinit **set, struct sysinit **set_end)
133 struct sysinit **newset;
134 struct sysinit **sipp;
135 struct sysinit **xipp;
138 count = set_end - set;
140 count += newsysinit_end - newsysinit;
142 count += sysinit_end - sysinit;
143 newset = kmalloc(count * sizeof(*sipp), M_TEMP, M_WAITOK);
146 for (sipp = newsysinit; sipp < newsysinit_end; sipp++)
149 for (sipp = sysinit; sipp < sysinit_end; sipp++)
152 for (sipp = set; sipp < set_end; sipp++)
155 kfree(newsysinit, M_TEMP);
157 newsysinit_end = newset + count;
161 * Callbacks from machine-dependant startup code (e.g. init386) to set
162 * up low level entities related to cpu #0's globaldata.
164 * Called from very low level boot code.
167 mi_proc0init(struct globaldata *gd, struct user *proc0paddr)
169 lwkt_init_thread(&thread0, proc0paddr, LWKT_THREAD_STACK, 0, gd);
170 lwkt_set_comm(&thread0, "thread0");
171 RB_INIT(&proc0.p_lwp_tree);
172 spin_init(&proc0.p_spin);
173 lwkt_token_init(&proc0.p_token, "iproc");
174 proc0.p_lasttid = 0; /* +1 = next TID */
175 lwp_rb_tree_RB_INSERT(&proc0.p_lwp_tree, &lwp0);
176 lwp0.lwp_thread = &thread0;
177 lwp0.lwp_proc = &proc0;
178 proc0.p_usched = usched_init();
179 lwp0.lwp_cpumask = (cpumask_t)-1;
180 varsymset_init(&proc0.p_varsymset, NULL);
181 thread0.td_flags |= TDF_RUNNING;
182 thread0.td_proc = &proc0;
183 thread0.td_lwp = &lwp0;
184 thread0.td_switch = cpu_lwkt_switch;
185 lwkt_schedule_self(curthread);
189 * System startup; initialize the world, create process 0, mount root
190 * filesystem, and fork to create init and pagedaemon. Most of the
191 * hard work is done in the lower-level initialization routines including
192 * startup(), which does memory initialization and autoconfiguration.
194 * This allows simple addition of new kernel subsystems that require
195 * boot time initialization. It also allows substitution of subsystem
196 * (for instance, a scheduler, kernel profiler, or VM system) by object
197 * module. Finally, it allows for optional "kernel threads".
202 struct sysinit *sip; /* system initialization*/
203 struct sysinit **sipp; /* system initialization*/
204 struct sysinit **xipp; /* interior loop of sort*/
205 struct sysinit *save; /* bubble*/
207 if (sysinit == NULL) {
208 sysinit = SET_BEGIN(sysinit_set);
209 #if defined(__amd64__) && defined(_KERNEL_VIRTUAL)
211 * XXX For whatever reason, on 64-bit vkernels
212 * the value of sysinit obtained from the
213 * linker set is wrong.
215 if ((long)sysinit % 8 != 0) {
216 kprintf("Fixing sysinit value...\n");
217 sysinit = (long)sysinit + 4;
220 sysinit_end = SET_LIMIT(sysinit_set);
222 #if defined(__amd64__) && defined(_KERNEL_VIRTUAL)
223 KKASSERT((long)sysinit % 8 == 0);
228 * Perform a bubble sort of the system initialization objects by
229 * their subsystem (primary key) and order (secondary key).
231 for (sipp = sysinit; sipp < sysinit_end; sipp++) {
232 for (xipp = sipp + 1; xipp < sysinit_end; xipp++) {
233 if ((*sipp)->subsystem < (*xipp)->subsystem ||
234 ((*sipp)->subsystem == (*xipp)->subsystem &&
235 (*sipp)->order <= (*xipp)->order))
244 * Traverse the (now) ordered list of system initialization tasks.
245 * Perform each task, and continue on to the next task.
247 * The last item on the list is expected to be the scheduler,
248 * which will not return.
250 for (sipp = sysinit; sipp < sysinit_end; sipp++) {
252 if (sip->subsystem == SI_SPECIAL_DUMMY)
253 continue; /* skip dummy task(s)*/
255 if (sip->subsystem == SI_SPECIAL_DONE)
259 (*(sip->func))(sip->udata);
261 /* Check off the one we're just done */
262 sip->subsystem = SI_SPECIAL_DONE;
264 /* Check if we've installed more sysinit items via KLD */
265 if (newsysinit != NULL) {
266 if (sysinit != SET_BEGIN(sysinit_set))
267 kfree(sysinit, M_TEMP);
268 sysinit = newsysinit;
269 sysinit_end = newsysinit_end;
271 newsysinit_end = NULL;
276 panic("Shouldn't get here!");
282 ***************************************************************************
284 **** The following SYSINIT's belong elsewhere, but have not yet
287 ***************************************************************************
290 print_caddr_t(void *data __unused)
292 kprintf("%s", (char *)data);
294 SYSINIT(announce, SI_BOOT1_COPYRIGHT, SI_ORDER_FIRST, print_caddr_t, copyright)
297 * Leave the critical section that protected us from spurious interrupts
298 * so device probes work.
301 leavecrit(void *dummy __unused)
303 MachIntrABI.stabilize();
305 MachIntrABI.cleanup();
307 KKASSERT(!IN_CRITICAL_SECT(curthread));
310 kprintf("Leaving critical section, allowing interrupts\n");
312 SYSINIT(leavecrit, SI_BOOT2_LEAVE_CRIT, SI_ORDER_ANY, leavecrit, NULL)
315 * This is called after the threading system is up and running,
316 * including the softclock, clock interrupts, and SMP.
319 tsleepworks(void *dummy __unused)
321 tsleep_now_works = 1;
323 SYSINIT(tsleepworks, SI_BOOT2_FINISH_SMP, SI_ORDER_SECOND, tsleepworks, NULL)
326 * This is called after devices have configured. Tell the kernel we are
327 * no longer in cold boot.
330 endofcoldboot(void *dummy __unused)
334 SYSINIT(endofcoldboot, SI_SUB_ISWARM, SI_ORDER_ANY, endofcoldboot, NULL)
337 ***************************************************************************
339 **** The two following SYSINT's are proc0 specific glue code. I am not
340 **** convinced that they can not be safely combined, but their order of
341 **** operation has been maintained as the same as the original init_main.c
344 **** These probably belong in init_proc.c or kern_proc.c, since they
345 **** deal with proc0 (the fork template process).
347 ***************************************************************************
351 proc0_init(void *dummy __unused)
360 * Initialize process and pgrp structures.
365 * additional VM structures
370 * Create process 0 (the swapper).
372 LIST_INSERT_HEAD(&allproc, p, p_list);
374 LIST_INSERT_HEAD(PGRPHASH(0), &pgrp0, pg_hash);
375 LIST_INIT(&pgrp0.pg_members);
376 LIST_INSERT_HEAD(&pgrp0.pg_members, p, p_pglist);
378 pgrp0.pg_session = &session0;
379 session0.s_count = 1;
380 session0.s_leader = p;
382 p->p_sysent = &aout_sysvec;
384 p->p_flag = P_SYSTEM;
386 lp->lwp_stat = LSRUN;
388 p->p_rtprio.type = RTP_PRIO_NORMAL;
389 p->p_rtprio.prio = 0;
390 lp->lwp_rtprio = p->p_rtprio;
395 bcopy("swapper", p->p_comm, sizeof ("swapper"));
396 bcopy("swapper", thread0.td_comm, sizeof ("swapper"));
398 /* Create credentials. */
399 p->p_ucred = crget();
400 p->p_ucred->cr_ruidinfo = uifind(0);
401 p->p_ucred->cr_ngroups = 1; /* group 0 */
402 p->p_ucred->cr_uidinfo = uifind(0);
403 thread0.td_ucred = crhold(p->p_ucred); /* bootstrap fork1() */
406 p->p_ucred->cr_prison = NULL;
408 /* Create sigacts. */
409 p->p_sigacts = &sigacts0;
410 p->p_sigacts->ps_refcnt = 1;
412 /* Initialize signal state for process 0. */
415 /* Create the file descriptor table. */
416 fdinit_bootstrap(p, &filedesc0, cmask);
418 /* Create the limits structures. */
419 plimit_init0(&limit0);
420 p->p_limit = &limit0;
422 /* Allocate a prototype map so we have something to fork. */
423 pmap_pinit0(vmspace_pmap(&vmspace0));
424 p->p_vmspace = &vmspace0;
425 lp->lwp_vmspace = p->p_vmspace;
426 sysref_init(&vmspace0.vm_sysref, &vmspace_sysref_class);
427 vm_map_init(&vmspace0.vm_map,
428 round_page(VM_MIN_USER_ADDRESS),
429 trunc_page(VM_MAX_USER_ADDRESS),
430 vmspace_pmap(&vmspace0));
431 sysref_activate(&vmspace0.vm_sysref);
433 kqueue_init(&lwp0.lwp_kqueue, &filedesc0);
436 * Charge root for one process.
438 (void)chgproccnt(p->p_ucred->cr_uidinfo, 1, 0);
441 SYSINIT(p0init, SI_BOOT2_PROC0, SI_ORDER_FIRST, proc0_init, NULL)
443 static int proc0_post_callback(struct proc *p, void *data __unused);
447 proc0_post(void *dummy __unused)
452 * Now we can look at the time, having had a chance to verify the
453 * time from the file system. Pretend that proc0 started now.
455 allproc_scan(proc0_post_callback, NULL);
458 * Give the ``random'' number generator a thump.
459 * XXX: Does read_random() contain enough bits to be used here ?
462 skrandom(ts.tv_sec ^ ts.tv_nsec);
466 proc0_post_callback(struct proc *p, void *data __unused)
468 microtime(&p->p_start);
472 SYSINIT(p0post, SI_SUB_PROC0_POST, SI_ORDER_FIRST, proc0_post, NULL)
475 ***************************************************************************
477 **** The following SYSINIT's and glue code should be moved to the
478 **** respective files on a per subsystem basis.
480 ***************************************************************************
485 ***************************************************************************
487 **** The following code probably belongs in another file, like
488 **** kern/init_init.c.
490 ***************************************************************************
494 * List of paths to try when searching for "init".
496 static char init_path[MAXPATHLEN] =
498 __XSTRING(INIT_PATH);
500 "/sbin/init:/sbin/oinit:/sbin/init.bak";
502 SYSCTL_STRING(_kern, OID_AUTO, init_path, CTLFLAG_RD, init_path, 0, "");
505 * Start the initial user process; try exec'ing each pathname in init_path.
506 * The program is invoked with one argument containing the boot flags.
509 start_init(void *dummy, struct trapframe *frame)
512 struct execve_args args;
514 char *var, *path, *next, *s;
515 char *ucp, **uap, *arg0, *arg1;
523 * This is passed in by the bootloader
525 env = kgetenv("kernelname");
527 strlcpy(kernelname, env, sizeof(kernelname));
530 * The MP lock is not held on entry. We release it before
531 * returning to userland.
536 lp = ONLY_LWP_IN_PROC(p);
538 /* Get the vnode for '/'. Set p->p_fd->fd_cdir to reference it. */
539 mp = mountlist_boot_getfirst();
540 if (VFS_ROOT(mp, &vp))
541 panic("cannot find root vnode");
542 if (mp->mnt_ncmountpt.ncp == NULL) {
543 cache_allocroot(&mp->mnt_ncmountpt, mp, vp);
544 cache_unlock(&mp->mnt_ncmountpt); /* leave ref intact */
546 p->p_fd->fd_cdir = vp;
547 vref(p->p_fd->fd_cdir);
548 p->p_fd->fd_rdir = vp;
549 vref(p->p_fd->fd_rdir);
550 vfs_cache_setroot(vp, cache_hold(&mp->mnt_ncmountpt));
551 vn_unlock(vp); /* leave ref intact */
552 cache_copy(&mp->mnt_ncmountpt, &p->p_fd->fd_ncdir);
553 cache_copy(&mp->mnt_ncmountpt, &p->p_fd->fd_nrdir);
555 kprintf("Mounting devfs\n");
556 vfs_mountroot_devfs();
559 * Need just enough stack to hold the faked-up "execve()" arguments.
561 addr = trunc_page(USRSTACK - PAGE_SIZE);
562 error = vm_map_find(&p->p_vmspace->vm_map, NULL, 0, &addr,
563 PAGE_SIZE, PAGE_SIZE,
564 FALSE, VM_MAPTYPE_NORMAL,
565 VM_PROT_ALL, VM_PROT_ALL,
568 panic("init: couldn't allocate argument space");
569 p->p_vmspace->vm_maxsaddr = (caddr_t)addr;
570 p->p_vmspace->vm_ssize = 1;
572 if ((var = kgetenv("init_path")) != NULL) {
573 strncpy(init_path, var, sizeof init_path);
574 init_path[sizeof init_path - 1] = 0;
576 if ((var = kgetenv("kern.fallback_elf_brand")) != NULL)
577 fallback_elf_brand = strtol(var, NULL, 0);
579 for (path = init_path; *path != '\0'; path = next) {
584 for (next = path; *next != '\0' && *next != ':'; next++)
587 kprintf("start_init: trying %.*s\n", (int)(next - path),
591 * Move out the boot flag argument.
594 ucp = (char *)USRSTACK;
595 (void)subyte(--ucp, 0); /* trailing zero */
596 if (boothowto & RB_SINGLE) {
597 (void)subyte(--ucp, 's');
601 if (boothowto & RB_FASTBOOT) {
602 (void)subyte(--ucp, 'f');
608 (void)subyte(--ucp, 'C');
612 (void)subyte(--ucp, '-');
613 (void)subyte(--ucp, '-'); /* leading hyphen */
617 * Move out the file name (also arg 0).
619 (void)subyte(--ucp, 0);
620 for (s = next - 1; s >= path; s--)
621 (void)subyte(--ucp, *s);
625 * Move out the arg pointers.
627 uap = (char **)((intptr_t)ucp & ~(sizeof(intptr_t)-1));
628 (void)suword((caddr_t)--uap, (long)0); /* terminator */
629 (void)suword((caddr_t)--uap, (long)(intptr_t)arg1);
630 (void)suword((caddr_t)--uap, (long)(intptr_t)arg0);
633 * Point at the arguments.
640 * Now try to exec the program. If can't for any reason
641 * other than it doesn't exist, complain.
643 * Otherwise, return via fork_trampoline() all the way
644 * to user mode as init!
646 * WARNING! We may have been moved to another cpu after
647 * acquiring the current user process designation. The
648 * MP lock will migrate with us though so we still have to
651 if ((error = sys_execve(&args)) == 0) {
653 lp->lwp_proc->p_usched->acquire_curproc(lp);
657 kprintf("exec %.*s: error %d\n", (int)(next - path),
660 kprintf("init: not found in path %s\n", init_path);
665 * Like kthread_create(), but runs in it's own address space.
666 * We do this early to reserve pid 1.
668 * Note special case - do not make it runnable yet. Other work
669 * in progress will change this more.
672 create_init(const void *udata __unused)
678 error = fork1(&lwp0, RFFDG | RFPROC, &initproc);
680 panic("cannot fork init: %d", error);
681 initproc->p_flag |= P_SYSTEM;
682 lp = ONLY_LWP_IN_PROC(initproc);
683 cpu_set_fork_handler(lp, start_init, NULL);
686 SYSINIT(init, SI_SUB_CREATE_INIT, SI_ORDER_FIRST, create_init, NULL)
689 * Make it runnable now.
692 kick_init(const void *udata __unused)
694 start_forked_proc(&lwp0, initproc);
696 SYSINIT(kickinit, SI_SUB_KTHREAD_INIT, SI_ORDER_FIRST, kick_init, NULL)
699 * Machine independant globaldata initialization
701 * WARNING! Called from early boot, 'mycpu' may not work yet.
704 mi_gdinit(struct globaldata *gd, int cpuid)
706 TAILQ_INIT(&gd->gd_systimerq);
707 gd->gd_sysid_alloc = cpuid; /* prime low bits for cpu lookup */
708 gd->gd_cpuid = cpuid;
709 gd->gd_cpumask = CPUMASK(cpuid);
711 vm_map_entry_reserve_cpu_init(gd);
713 usched_global_cpumask |= CPUMASK(cpuid);