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. Neither the name of the University nor the names of its contributors
22 * may be used to endorse or promote products derived from this software
23 * without specific prior written permission.
25 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
26 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
27 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
28 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
29 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
30 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
31 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
32 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
33 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
34 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
37 * @(#)init_main.c 8.9 (Berkeley) 1/21/94
38 * $FreeBSD: src/sys/kern/init_main.c,v 1.134.2.8 2003/06/06 20:21:32 tegge Exp $
41 #include "opt_init_path.h"
43 #include <sys/param.h>
45 #include <sys/filedesc.h>
46 #include <sys/kernel.h>
47 #include <sys/mount.h>
48 #include <sys/sysctl.h>
50 #include <sys/resourcevar.h>
51 #include <sys/signalvar.h>
52 #include <sys/systm.h>
53 #include <sys/vnode.h>
54 #include <sys/sysent.h>
55 #include <sys/reboot.h>
56 #include <sys/sysproto.h>
57 #include <sys/vmmeter.h>
58 #include <sys/unistd.h>
59 #include <sys/malloc.h>
60 #include <sys/machintr.h>
62 #include <sys/refcount.h>
63 #include <sys/file2.h>
64 #include <sys/thread2.h>
65 #include <sys/sysref2.h>
66 #include <sys/spinlock2.h>
67 #include <sys/mplock2.h>
69 #include <machine/cpu.h>
72 #include <vm/vm_param.h>
75 #include <vm/vm_map.h>
76 #include <vm/vm_extern.h>
78 #include <sys/copyright.h>
80 int vfs_mountroot_devfs(void);
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;
96 cpumask_t usched_global_cpumask;
97 extern struct user *proc0paddr;
99 int boothowto = 0; /* initialized so that it can be patched */
100 SYSCTL_INT(_debug, OID_AUTO, boothowto, CTLFLAG_RD, &boothowto, 0,
101 "Reboot flags, from console subsystem");
102 SYSCTL_ULONG(_kern, OID_AUTO, usched_global_cpumask, CTLFLAG_RW,
103 &usched_global_cpumask, 0, "global user scheduler cpumask");
106 * This ensures that there is at least one entry so that the sysinit_set
107 * symbol is not undefined. A subsystem ID of SI_SPECIAL_DUMMY is never
110 SYSINIT(placeholder, SI_SPECIAL_DUMMY, SI_ORDER_ANY, NULL, NULL)
113 * The sysinit table itself. Items are checked off as the are run.
114 * If we want to register new sysinit types, add them to newsysinit.
116 SET_DECLARE(sysinit_set, struct sysinit);
117 struct sysinit **sysinit, **sysinit_end;
118 struct sysinit **newsysinit, **newsysinit_end;
122 * Merge a new sysinit set into the current set, reallocating it if
123 * necessary. This can only be called after malloc is running.
126 sysinit_add(struct sysinit **set, struct sysinit **set_end)
128 struct sysinit **newset;
129 struct sysinit **sipp;
130 struct sysinit **xipp;
133 count = set_end - set;
135 count += newsysinit_end - newsysinit;
137 count += sysinit_end - sysinit;
138 newset = kmalloc(count * sizeof(*sipp), M_TEMP, M_WAITOK);
141 for (sipp = newsysinit; sipp < newsysinit_end; sipp++)
144 for (sipp = sysinit; sipp < sysinit_end; sipp++)
147 for (sipp = set; sipp < set_end; sipp++)
150 kfree(newsysinit, M_TEMP);
152 newsysinit_end = newset + count;
156 * Callbacks from machine-dependant startup code (e.g. init386) to set
157 * up low level entities related to cpu #0's globaldata.
159 * Called from very low level boot code.
162 mi_proc0init(struct globaldata *gd, struct user *proc0paddr)
164 lwkt_init_thread(&thread0, proc0paddr, LWKT_THREAD_STACK, 0, gd);
165 lwkt_set_comm(&thread0, "thread0");
166 RB_INIT(&proc0.p_lwp_tree);
167 spin_init(&proc0.p_spin, "iproc_proc0");
168 lwkt_token_init(&proc0.p_token, "iproc");
169 proc0.p_lasttid = 0; /* +1 = next TID */
170 lwp_rb_tree_RB_INSERT(&proc0.p_lwp_tree, &lwp0);
171 lwp0.lwp_thread = &thread0;
172 lwp0.lwp_proc = &proc0;
173 proc0.p_usched = usched_init();
174 CPUMASK_ASSALLONES(lwp0.lwp_cpumask);
175 lwkt_token_init(&lwp0.lwp_token, "lwp_token");
176 spin_init(&lwp0.lwp_spin, "iproc_lwp0");
177 varsymset_init(&proc0.p_varsymset, NULL);
178 thread0.td_flags |= TDF_RUNNING;
179 thread0.td_proc = &proc0;
180 thread0.td_lwp = &lwp0;
181 thread0.td_switch = cpu_lwkt_switch;
182 lwkt_schedule_self(curthread);
186 * System startup; initialize the world, create process 0, mount root
187 * filesystem, and fork to create init and pagedaemon. Most of the
188 * hard work is done in the lower-level initialization routines including
189 * startup(), which does memory initialization and autoconfiguration.
191 * This allows simple addition of new kernel subsystems that require
192 * boot time initialization. It also allows substitution of subsystem
193 * (for instance, a scheduler, kernel profiler, or VM system) by object
194 * module. Finally, it allows for optional "kernel threads".
199 struct sysinit *sip; /* system initialization*/
200 struct sysinit **sipp; /* system initialization*/
201 struct sysinit **xipp; /* interior loop of sort*/
202 struct sysinit *save; /* bubble*/
204 if (sysinit == NULL) {
205 sysinit = SET_BEGIN(sysinit_set);
206 #if defined(__x86_64__) && defined(_KERNEL_VIRTUAL)
208 * XXX For whatever reason, on 64-bit vkernels
209 * the value of sysinit obtained from the
210 * linker set is wrong.
212 if ((long)sysinit % 8 != 0) {
213 kprintf("Fixing sysinit value...\n");
214 sysinit = (void *)((long)(intptr_t)sysinit + 4);
217 sysinit_end = SET_LIMIT(sysinit_set);
219 #if defined(__x86_64__) && defined(_KERNEL_VIRTUAL)
220 KKASSERT((long)sysinit % 8 == 0);
225 * Perform a bubble sort of the system initialization objects by
226 * their subsystem (primary key) and order (secondary key).
228 for (sipp = sysinit; sipp < sysinit_end; sipp++) {
229 for (xipp = sipp + 1; xipp < sysinit_end; xipp++) {
230 if ((*sipp)->subsystem < (*xipp)->subsystem ||
231 ((*sipp)->subsystem == (*xipp)->subsystem &&
232 (*sipp)->order <= (*xipp)->order))
241 * Traverse the (now) ordered list of system initialization tasks.
242 * Perform each task, and continue on to the next task.
244 * The last item on the list is expected to be the scheduler,
245 * which will not return.
247 for (sipp = sysinit; sipp < sysinit_end; sipp++) {
249 if (sip->subsystem == SI_SPECIAL_DUMMY)
250 continue; /* skip dummy task(s)*/
252 if (sip->subsystem == SI_SPECIAL_DONE)
257 kprintf("(%08x-%p)\n", sip->subsystem, sip->func);
261 (*(sip->func))(sip->udata);
263 /* Check off the one we're just done */
264 sip->subsystem = SI_SPECIAL_DONE;
266 /* Check if we've installed more sysinit items via KLD */
267 if (newsysinit != NULL) {
268 if (sysinit != SET_BEGIN(sysinit_set))
269 kfree(sysinit, M_TEMP);
270 sysinit = newsysinit;
271 sysinit_end = newsysinit_end;
273 newsysinit_end = NULL;
278 panic("Shouldn't get here!");
284 ***************************************************************************
286 **** The following SYSINIT's belong elsewhere, but have not yet
289 ***************************************************************************
292 print_caddr_t(void *data)
294 kprintf("%s", (char *)data);
296 SYSINIT(announce, SI_BOOT1_COPYRIGHT, SI_ORDER_FIRST, print_caddr_t, copyright)
299 * Leave the critical section that protected us from spurious interrupts
300 * so device probes work.
303 leavecrit(void *dummy __unused)
305 MachIntrABI.stabilize();
307 MachIntrABI.cleanup();
309 KKASSERT(!IN_CRITICAL_SECT(curthread));
312 kprintf("Leaving critical section, allowing interrupts\n");
314 SYSINIT(leavecrit, SI_BOOT2_LEAVE_CRIT, SI_ORDER_ANY, leavecrit, NULL)
317 * This is called after the threading system is up and running,
318 * including the softclock, clock interrupts, and SMP.
321 tsleepworks(void *dummy __unused)
323 tsleep_now_works = 1;
325 SYSINIT(tsleepworks, SI_BOOT2_FINISH_SMP, SI_ORDER_SECOND, tsleepworks, NULL)
328 * This is called after devices have configured. Tell the kernel we are
329 * no longer in cold boot.
332 endofcoldboot(void *dummy __unused)
336 SYSINIT(endofcoldboot, SI_SUB_ISWARM, SI_ORDER_ANY, endofcoldboot, NULL)
339 ***************************************************************************
341 **** The two following SYSINT's are proc0 specific glue code. I am not
342 **** convinced that they can not be safely combined, but their order of
343 **** operation has been maintained as the same as the original init_main.c
346 **** These probably belong in init_proc.c or kern_proc.c, since they
347 **** deal with proc0 (the fork template process).
349 ***************************************************************************
353 proc0_init(void *dummy __unused)
364 p->p_osrel = osreldate;
367 * Initialize process and pgrp structures.
372 * additional VM structures
377 * Create process 0 (the swapper).
380 pgrpinsertinit(&pgrp0);
381 LIST_INIT(&pgrp0.pg_members);
382 lwkt_token_init(&pgrp0.pg_token, "pgrp0");
383 refcount_init(&pgrp0.pg_refs, 1);
384 lockinit(&pgrp0.pg_lock, "pgwt0", 0, 0);
385 LIST_INSERT_HEAD(&pgrp0.pg_members, p, p_pglist);
387 pgrp0.pg_session = &session0;
388 session0.s_count = 1;
389 session0.s_leader = p;
390 sessinsertinit(&session0);
395 p->p_sysent = &aout_sysvec;
397 p->p_flags = P_SYSTEM;
399 lp->lwp_stat = LSRUN;
401 p->p_rtprio.type = RTP_PRIO_NORMAL;
402 p->p_rtprio.prio = 0;
403 lp->lwp_rtprio = p->p_rtprio;
408 bcopy("swapper", p->p_comm, sizeof ("swapper"));
409 bcopy("swapper", thread0.td_comm, sizeof ("swapper"));
411 /* Create credentials. */
412 p->p_ucred = crget();
413 p->p_ucred->cr_ruidinfo = uifind(0);
414 p->p_ucred->cr_ngroups = 1; /* group 0 */
415 p->p_ucred->cr_uidinfo = uifind(0);
416 thread0.td_ucred = crhold(p->p_ucred); /* bootstrap fork1() */
419 p->p_ucred->cr_prison = NULL;
421 /* Create sigacts. */
422 p->p_sigacts = &sigacts0;
423 refcount_init(&p->p_sigacts->ps_refcnt, 1);
425 /* Initialize signal state for process 0. */
428 /* Create the file descriptor table. */
429 fdinit_bootstrap(p, &filedesc0, cmask);
431 /* Create the limits structures. */
432 plimit_init0(&limit0);
433 p->p_limit = &limit0;
435 /* Allocate a prototype map so we have something to fork. */
436 pmap_pinit0(vmspace_pmap(&vmspace0));
437 p->p_vmspace = &vmspace0;
438 lp->lwp_vmspace = p->p_vmspace;
439 vmspace_initrefs(&vmspace0);
440 vm_map_init(&vmspace0.vm_map,
441 round_page(VM_MIN_USER_ADDRESS),
442 trunc_page(VM_MAX_USER_ADDRESS),
443 vmspace_pmap(&vmspace0));
445 kqueue_init(&lwp0.lwp_kqueue, &filedesc0);
448 * Charge root for one process.
450 (void)chgproccnt(p->p_ucred->cr_uidinfo, 1, 0);
453 SYSINIT(p0init, SI_BOOT2_PROC0, SI_ORDER_FIRST, proc0_init, NULL)
455 static int proc0_post_callback(struct proc *p, void *data __unused);
459 proc0_post(void *dummy __unused)
464 * Now we can look at the time, having had a chance to verify the
465 * time from the file system. Pretend that proc0 started now.
467 allproc_scan(proc0_post_callback, NULL);
470 * Give the ``random'' number generator a thump.
471 * XXX: Does read_random() contain enough bits to be used here ?
474 skrandom(ts.tv_sec ^ ts.tv_nsec);
478 proc0_post_callback(struct proc *p, void *data __unused)
480 microtime(&p->p_start);
484 SYSINIT(p0post, SI_SUB_PROC0_POST, SI_ORDER_FIRST, proc0_post, NULL)
487 ***************************************************************************
489 **** The following SYSINIT's and glue code should be moved to the
490 **** respective files on a per subsystem basis.
492 ***************************************************************************
497 ***************************************************************************
499 **** The following code probably belongs in another file, like
500 **** kern/init_init.c.
502 ***************************************************************************
506 * List of paths to try when searching for "init".
508 static char init_path[MAXPATHLEN] =
510 __XSTRING(INIT_PATH);
512 "/sbin/init:/sbin/oinit:/sbin/init.bak";
514 SYSCTL_STRING(_kern, OID_AUTO, init_path, CTLFLAG_RD, init_path, 0, "");
517 * Shutdown timeout of init(8).
518 * Unused within kernel, but used to control init(8), hence do not remove.
520 #ifndef INIT_SHUTDOWN_TIMEOUT
521 #define INIT_SHUTDOWN_TIMEOUT 120
523 static int init_shutdown_timeout = INIT_SHUTDOWN_TIMEOUT;
524 SYSCTL_INT(_kern, OID_AUTO, init_shutdown_timeout,
525 CTLFLAG_RW, &init_shutdown_timeout, 0, "Shutdown timeout of init(8). "
526 "Unused within kernel, but used to control init(8)");
529 * Start the initial user process; try exec'ing each pathname in init_path.
530 * The program is invoked with one argument containing the boot flags.
533 start_init(void *dummy, struct trapframe *frame)
536 struct execve_args args;
538 char *var, *path, *next, *s;
539 char *ucp, **uap, *arg0, *arg1;
547 * This is passed in by the bootloader
549 env = kgetenv("kernelname");
551 strlcpy(kernelname, env, sizeof(kernelname));
554 * The MP lock is not held on entry. We release it before
555 * returning to userland.
560 lp = ONLY_LWP_IN_PROC(p);
562 /* Get the vnode for '/'. Set p->p_fd->fd_cdir to reference it. */
563 mp = mountlist_boot_getfirst();
564 if (VFS_ROOT(mp, &vp))
565 panic("cannot find root vnode");
566 if (mp->mnt_ncmountpt.ncp == NULL) {
567 cache_allocroot(&mp->mnt_ncmountpt, mp, vp);
568 cache_unlock(&mp->mnt_ncmountpt); /* leave ref intact */
570 p->p_fd->fd_cdir = vp;
571 vref(p->p_fd->fd_cdir);
572 p->p_fd->fd_rdir = vp;
573 vref(p->p_fd->fd_rdir);
574 vfs_cache_setroot(vp, cache_hold(&mp->mnt_ncmountpt));
575 vn_unlock(vp); /* leave ref intact */
576 cache_copy(&mp->mnt_ncmountpt, &p->p_fd->fd_ncdir);
577 cache_copy(&mp->mnt_ncmountpt, &p->p_fd->fd_nrdir);
579 kprintf("Mounting devfs\n");
580 vfs_mountroot_devfs();
583 * Need just enough stack to hold the faked-up "execve()" arguments.
585 addr = trunc_page(USRSTACK - PAGE_SIZE);
586 error = vm_map_find(&p->p_vmspace->vm_map, NULL, 0, &addr,
587 PAGE_SIZE, PAGE_SIZE,
588 FALSE, VM_MAPTYPE_NORMAL,
589 VM_PROT_ALL, VM_PROT_ALL,
592 panic("init: couldn't allocate argument space");
593 p->p_vmspace->vm_maxsaddr = (caddr_t)addr;
594 p->p_vmspace->vm_ssize = 1;
596 if ((var = kgetenv("init_path")) != NULL) {
597 strncpy(init_path, var, sizeof init_path);
598 init_path[sizeof init_path - 1] = 0;
601 for (path = init_path; *path != '\0'; path = next) {
606 for (next = path; *next != '\0' && *next != ':'; next++)
609 kprintf("start_init: trying %.*s\n", (int)(next - path),
613 * Move out the boot flag argument.
616 ucp = (char *)USRSTACK;
617 (void)subyte(--ucp, 0); /* trailing zero */
618 if (boothowto & RB_SINGLE) {
619 (void)subyte(--ucp, 's');
623 if (boothowto & RB_FASTBOOT) {
624 (void)subyte(--ucp, 'f');
630 (void)subyte(--ucp, 'C');
634 (void)subyte(--ucp, '-');
635 (void)subyte(--ucp, '-'); /* leading hyphen */
639 * Move out the file name (also arg 0).
641 (void)subyte(--ucp, 0);
642 for (s = next - 1; s >= path; s--)
643 (void)subyte(--ucp, *s);
647 * Move out the arg pointers.
649 uap = (char **)((intptr_t)ucp & ~(sizeof(intptr_t)-1));
650 (void)suword((caddr_t)--uap, (long)0); /* terminator */
651 (void)suword((caddr_t)--uap, (long)(intptr_t)arg1);
652 (void)suword((caddr_t)--uap, (long)(intptr_t)arg0);
655 * Point at the arguments.
662 * Now try to exec the program. If can't for any reason
663 * other than it doesn't exist, complain.
665 * Otherwise, return via fork_trampoline() all the way
666 * to user mode as init!
668 * WARNING! We may have been moved to another cpu after
669 * acquiring the current user process designation. The
670 * MP lock will migrate with us though so we still have to
673 if ((error = sys_execve(&args)) == 0) {
675 lp->lwp_proc->p_usched->acquire_curproc(lp);
679 kprintf("exec %.*s: error %d\n", (int)(next - path),
682 kprintf("init: not found in path %s\n", init_path);
687 * Like kthread_create(), but runs in it's own address space.
688 * We do this early to reserve pid 1.
690 * Note special case - do not make it runnable yet. Other work
691 * in progress will change this more.
694 create_init(const void *udata __unused)
700 error = fork1(&lwp0, RFFDG | RFPROC, &initproc);
702 panic("cannot fork init: %d", error);
703 initproc->p_flags |= P_SYSTEM;
704 lp = ONLY_LWP_IN_PROC(initproc);
705 cpu_set_fork_handler(lp, start_init, NULL);
708 SYSINIT(init, SI_SUB_CREATE_INIT, SI_ORDER_FIRST, create_init, NULL)
711 * Make it runnable now.
714 kick_init(const void *udata __unused)
716 start_forked_proc(&lwp0, initproc);
718 SYSINIT(kickinit, SI_SUB_KTHREAD_INIT, SI_ORDER_FIRST, kick_init, NULL)
721 * Machine independant globaldata initialization
723 * WARNING! Called from early boot, 'mycpu' may not work yet.
726 mi_gdinit(struct globaldata *gd, int cpuid)
728 TAILQ_INIT(&gd->gd_systimerq);
729 gd->gd_sysid_alloc = cpuid; /* prime low bits for cpu lookup */
730 gd->gd_cpuid = cpuid;
731 CPUMASK_ASSBIT(gd->gd_cpumask, cpuid);
733 vm_map_entry_reserve_cpu_init(gd);
735 ATOMIC_CPUMASK_ORBIT(usched_global_cpumask, cpuid);