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/spinlock2.h>
67 #include <machine/cpu.h>
70 #include <vm/vm_param.h>
73 #include <vm/vm_map.h>
74 #include <vm/vm_extern.h>
75 #include <sys/copyright.h>
77 int vfs_mountroot_devfs(void);
79 /* Components of the first process -- never freed. */
80 static struct session session0;
81 static struct pgrp pgrp0;
82 static struct sigacts sigacts0;
83 static struct filedesc filedesc0;
84 static struct plimit limit0;
85 static struct vmspace vmspace0;
86 struct proc *initproc;
89 struct thread thread0;
90 struct sys_kpmap *kpmap;
91 struct sysreaper initreaper;
95 cpumask_t usched_global_cpumask;
96 extern struct user *proc0paddr;
98 int boothowto = 0; /* initialized so that it can be patched */
99 SYSCTL_INT(_debug, OID_AUTO, boothowto, CTLFLAG_RD, &boothowto, 0,
100 "Reboot flags, from console subsystem");
101 SYSCTL_OPAQUE(_kern, OID_AUTO, usched_global_cpumask, CTLFLAG_RW,
102 &usched_global_cpumask, sizeof(usched_global_cpumask), "LU",
103 "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");
170 proc0.p_lasttid = lwp0.lwp_tid; /* +1 = next TID */
171 lwp_rb_tree_RB_INSERT(&proc0.p_lwp_tree, &lwp0);
172 lwp0.lwp_thread = &thread0;
173 lwp0.lwp_proc = &proc0;
174 proc0.p_usched = usched_init();
175 CPUMASK_ASSALLONES(lwp0.lwp_cpumask);
176 lwkt_token_init(&lwp0.lwp_token, "lwp_token");
177 spin_init(&lwp0.lwp_spin, "iproc_lwp0");
178 varsymset_init(&proc0.p_varsymset, NULL);
179 thread0.td_flags |= TDF_RUNNING;
180 thread0.td_proc = &proc0;
181 thread0.td_lwp = &lwp0;
182 thread0.td_switch = cpu_lwkt_switch;
183 lwkt_schedule_self(curthread);
187 * System startup; initialize the world, create process 0, mount root
188 * filesystem, and fork to create init and pagedaemon. Most of the
189 * hard work is done in the lower-level initialization routines including
190 * startup(), which does memory initialization and autoconfiguration.
192 * This allows simple addition of new kernel subsystems that require
193 * boot time initialization. It also allows substitution of subsystem
194 * (for instance, a scheduler, kernel profiler, or VM system) by object
195 * module. Finally, it allows for optional "kernel threads".
200 struct sysinit *sip; /* system initialization*/
201 struct sysinit **sipp; /* system initialization*/
202 struct sysinit **xipp; /* interior loop of sort*/
203 struct sysinit *save; /* bubble*/
205 if (sysinit == NULL) {
206 sysinit = SET_BEGIN(sysinit_set);
207 #if defined(__x86_64__) && defined(_KERNEL_VIRTUAL)
209 * XXX For whatever reason, on 64-bit vkernels
210 * the value of sysinit obtained from the
211 * linker set is wrong.
213 if ((long)sysinit % 8 != 0) {
214 kprintf("Fixing sysinit value...\n");
215 sysinit = (void *)((long)(intptr_t)sysinit + 4);
218 sysinit_end = SET_LIMIT(sysinit_set);
220 #if defined(__x86_64__) && defined(_KERNEL_VIRTUAL)
221 KKASSERT((long)sysinit % 8 == 0);
226 * Perform a bubble sort of the system initialization objects by
227 * their subsystem (primary key) and order (secondary key).
229 for (sipp = sysinit; sipp < sysinit_end; sipp++) {
230 for (xipp = sipp + 1; xipp < sysinit_end; xipp++) {
231 if ((*sipp)->subsystem < (*xipp)->subsystem ||
232 ((*sipp)->subsystem == (*xipp)->subsystem &&
233 (*sipp)->order <= (*xipp)->order))
242 * Traverse the (now) ordered list of system initialization tasks.
243 * Perform each task, and continue on to the next task.
245 * The last item on the list is expected to be the scheduler,
246 * which will not return.
248 for (sipp = sysinit; sipp < sysinit_end; sipp++) {
250 if (sip->subsystem == SI_SPECIAL_DUMMY)
251 continue; /* skip dummy task(s)*/
253 if (sip->subsystem == SI_SPECIAL_DONE)
258 kprintf("(%08x-%p)\n", sip->subsystem, sip->func);
262 (*(sip->func))(sip->udata);
264 /* Check off the one we're just done */
265 sip->subsystem = SI_SPECIAL_DONE;
267 /* Check if we've installed more sysinit items via KLD */
268 if (newsysinit != NULL) {
269 if (sysinit != SET_BEGIN(sysinit_set))
270 kfree(sysinit, M_TEMP);
271 sysinit = newsysinit;
272 sysinit_end = newsysinit_end;
274 newsysinit_end = NULL;
279 panic("Shouldn't get here!");
285 ***************************************************************************
287 **** The following SYSINIT's belong elsewhere, but have not yet
290 ***************************************************************************
293 print_caddr_t(void *data)
295 kprintf("%s", (char *)data);
297 SYSINIT(announce, SI_BOOT1_COPYRIGHT, SI_ORDER_FIRST, print_caddr_t, copyright);
300 * Leave the critical section that protected us from spurious interrupts
301 * so device probes work.
304 leavecrit(void *dummy __unused)
306 MachIntrABI.stabilize();
308 MachIntrABI.cleanup();
310 KKASSERT(!IN_CRITICAL_SECT(curthread));
313 kprintf("Leaving critical section, allowing interrupts\n");
315 SYSINIT(leavecrit, SI_BOOT2_LEAVE_CRIT, SI_ORDER_ANY, leavecrit, NULL);
318 * This is called after the threading system is up and running,
319 * including the softclock, clock interrupts, and SMP.
322 tsleepworks(void *dummy __unused)
324 tsleep_now_works = 1;
326 SYSINIT(tsleepworks, SI_BOOT2_FINISH_SMP, SI_ORDER_SECOND, tsleepworks, NULL);
329 * This is called after devices have configured. Tell the kernel we are
330 * no longer in cold boot.
333 endofcoldboot(void *dummy __unused)
337 SYSINIT(endofcoldboot, SI_SUB_ISWARM, SI_ORDER_ANY, endofcoldboot, NULL);
340 ***************************************************************************
342 **** The two following SYSINT's are proc0 specific glue code. I am not
343 **** convinced that they can not be safely combined, but their order of
344 **** operation has been maintained as the same as the original init_main.c
347 **** These probably belong in init_proc.c or kern_proc.c, since they
348 **** deal with proc0 (the fork template process).
350 ***************************************************************************
354 proc0_init(void *dummy __unused)
366 p->p_osrel = osreldate;
369 * Initialize process and pgrp structures.
374 * additional VM structures
379 * Create process 0 (the swapper).
382 pgrpinsertinit(&pgrp0);
383 LIST_INIT(&pgrp0.pg_members);
384 lwkt_token_init(&pgrp0.pg_token, "pgrp0");
385 refcount_init(&pgrp0.pg_refs, 1);
386 lockinit(&pgrp0.pg_lock, "pgwt0", 0, 0);
387 LIST_INSERT_HEAD(&pgrp0.pg_members, p, p_pglist);
389 pgrp0.pg_session = &session0;
390 session0.s_count = 1;
391 session0.s_leader = p;
392 sessinsertinit(&session0);
397 p->p_sysent = &aout_sysvec;
399 p->p_flags = P_SYSTEM;
401 lp->lwp_stat = LSRUN;
403 p->p_rtprio.type = RTP_PRIO_NORMAL;
404 p->p_rtprio.prio = 0;
405 lp->lwp_rtprio = p->p_rtprio;
410 bcopy("swapper", p->p_comm, sizeof ("swapper"));
411 bcopy("swapper", thread0.td_comm, sizeof ("swapper"));
413 /* Create credentials. */
414 uip = uicreate(0); /* for cr_ruidinfo */
415 uihold(uip); /* for cr_uidinfo */
416 p->p_ucred = crget();
417 p->p_ucred->cr_ruidinfo = uip;
418 p->p_ucred->cr_uidinfo = uip;
419 p->p_ucred->cr_ngroups = 1; /* group 0 */
420 thread0.td_ucred = crhold(p->p_ucred); /* bootstrap fork1() */
423 p->p_ucred->cr_prison = NULL;
425 /* Create sigacts. */
426 p->p_sigacts = &sigacts0;
427 refcount_init(&p->p_sigacts->ps_refcnt, 1);
429 /* Initialize signal state for process 0. */
432 /* Create the file descriptor table. */
433 fdinit_bootstrap(p, &filedesc0, cmask);
435 /* Create the limits structures. */
436 plimit_init0(&limit0);
437 p->p_limit = &limit0;
439 /* Allocate a prototype map so we have something to fork. */
440 pmap_pinit0(vmspace_pmap(&vmspace0));
441 p->p_vmspace = &vmspace0;
442 lp->lwp_vmspace = p->p_vmspace;
443 vmspace_initrefs(&vmspace0);
444 vm_map_init(&vmspace0.vm_map,
445 round_page(VM_MIN_USER_ADDRESS),
446 trunc_page(VM_MAX_USER_ADDRESS),
447 vmspace_pmap(&vmspace0));
449 kqueue_init(&lwp0.lwp_kqueue, &filedesc0);
452 * Charge root for one process.
454 (void)chgproccnt(p->p_ucred->cr_uidinfo, 1, 0);
457 SYSINIT(p0init, SI_BOOT2_PROC0, SI_ORDER_FIRST, proc0_init, NULL);
459 static int proc0_post_callback(struct proc *p, void *data __unused);
463 proc0_post(void *dummy __unused)
468 * Now we can look at the time, having had a chance to verify the
469 * time from the file system. Pretend that proc0 started now.
471 allproc_scan(proc0_post_callback, NULL, 0);
474 * Give the ``random'' number generator a thump.
475 * XXX: Does read_random() contain enough bits to be used here ?
478 skrandom(ts.tv_sec ^ ts.tv_nsec);
482 proc0_post_callback(struct proc *p, void *data __unused)
484 microtime(&p->p_start);
488 SYSINIT(p0post, SI_SUB_PROC0_POST, SI_ORDER_FIRST, proc0_post, NULL);
491 ***************************************************************************
493 **** The following SYSINIT's and glue code should be moved to the
494 **** respective files on a per subsystem basis.
496 ***************************************************************************
501 ***************************************************************************
503 **** The following code probably belongs in another file, like
504 **** kern/init_init.c.
506 ***************************************************************************
510 * List of paths to try when searching for "init".
512 static char init_path[MAXPATHLEN] =
514 __XSTRING(INIT_PATH);
516 "/sbin/init:/sbin/oinit:/sbin/init.bak";
518 SYSCTL_STRING(_kern, OID_AUTO, init_path, CTLFLAG_RD, init_path, 0, "");
521 * Shutdown timeout of init(8).
522 * Unused within kernel, but used to control init(8), hence do not remove.
524 #ifndef INIT_SHUTDOWN_TIMEOUT
525 #define INIT_SHUTDOWN_TIMEOUT 120
527 static int init_shutdown_timeout = INIT_SHUTDOWN_TIMEOUT;
528 SYSCTL_INT(_kern, OID_AUTO, init_shutdown_timeout,
529 CTLFLAG_RW, &init_shutdown_timeout, 0, "Shutdown timeout of init(8). "
530 "Unused within kernel, but used to control init(8)");
533 * Start the initial user process; try exec'ing each pathname in init_path.
534 * The program is invoked with one argument containing the boot flags.
537 start_init(void *dummy, struct trapframe *frame)
540 struct execve_args args;
542 char *var, *path, *next, *s;
543 char *ucp, **uap, *arg0, *arg1;
551 * This is passed in by the bootloader
553 env = kgetenv("kernelname");
555 strlcpy(kernelname, env, sizeof(kernelname));
559 lp = ONLY_LWP_IN_PROC(p);
561 /* Get the vnode for '/'. Set p->p_fd->fd_cdir to reference it. */
562 mp = mountlist_boot_getfirst();
563 if (VFS_ROOT(mp, &vp))
564 panic("cannot find root vnode");
565 if (mp->mnt_ncmountpt.ncp == NULL) {
566 cache_allocroot(&mp->mnt_ncmountpt, mp, vp);
567 cache_unlock(&mp->mnt_ncmountpt); /* leave ref intact */
569 p->p_fd->fd_cdir = vp;
570 vref(p->p_fd->fd_cdir);
571 p->p_fd->fd_rdir = vp;
572 vref(p->p_fd->fd_rdir);
573 vfs_cache_setroot(vp, cache_hold(&mp->mnt_ncmountpt));
574 vn_unlock(vp); /* leave ref intact */
575 cache_copy(&mp->mnt_ncmountpt, &p->p_fd->fd_ncdir);
576 cache_copy(&mp->mnt_ncmountpt, &p->p_fd->fd_nrdir);
578 kprintf("Mounting devfs\n");
579 vfs_mountroot_devfs();
582 * Need just enough stack to hold the faked-up "execve()" arguments.
584 addr = trunc_page(USRSTACK - PAGE_SIZE);
585 error = vm_map_find(&p->p_vmspace->vm_map, NULL, NULL,
588 VM_MAPTYPE_NORMAL, VM_SUBSYS_INIT,
589 VM_PROT_ALL, VM_PROT_ALL, 0);
591 panic("init: couldn't allocate argument space");
592 p->p_vmspace->vm_maxsaddr = (caddr_t)addr;
593 p->p_vmspace->vm_ssize = PAGE_SIZE;
595 if ((var = kgetenv("init_path")) != NULL) {
596 strncpy(init_path, var, sizeof init_path);
597 init_path[sizeof init_path - 1] = 0;
600 for (path = init_path; *path != '\0'; path = next) {
605 for (next = path; *next != '\0' && *next != ':'; next++)
608 kprintf("start_init: trying %.*s\n", (int)(next - path),
612 * Move out the boot flag argument.
615 ucp = (char *)USRSTACK;
616 (void)subyte(--ucp, 0); /* trailing zero */
617 if (boothowto & RB_SINGLE) {
618 (void)subyte(--ucp, 's');
622 if (boothowto & RB_FASTBOOT) {
623 (void)subyte(--ucp, 'f');
629 (void)subyte(--ucp, 'C');
633 (void)subyte(--ucp, '-');
634 (void)subyte(--ucp, '-'); /* leading hyphen */
638 * Move out the file name (also arg 0).
640 (void)subyte(--ucp, 0);
641 for (s = next - 1; s >= path; s--)
642 (void)subyte(--ucp, *s);
646 * Move out the arg pointers.
648 uap = (char **)((intptr_t)ucp & ~(sizeof(intptr_t)-1));
651 suword64((uint64_t *)(caddr_t)--uap, (uint64_t)0);
653 suword64((uint64_t *)(caddr_t)--uap, (uint64_t)(intptr_t)arg1);
654 suword64((uint64_t *)(caddr_t)--uap, (uint64_t)(intptr_t)arg0);
657 * Point at the arguments.
664 * Now try to exec the program. If can't for any reason
665 * other than it doesn't exist, complain.
667 * Otherwise, return via fork_trampoline() all the way
668 * to user mode as init!
670 * WARNING! We may have been moved to another cpu after
671 * acquiring the current user process designation. The
672 * MP lock will migrate with us though so we still have to
675 if ((error = sys_execve(&args)) == 0) {
676 lp->lwp_proc->p_usched->acquire_curproc(lp);
680 kprintf("exec %.*s: error %d\n", (int)(next - path),
683 kprintf("init: not found in path %s\n", init_path);
688 * Like kthread_create(), but runs in it's own address space.
689 * We do this early to reserve pid 1.
691 * Note special case - do not make it runnable yet. Other work
692 * in progress will change this more.
695 create_init(const void *udata __unused)
701 error = fork1(&lwp0, RFFDG | RFPROC, &initproc);
703 panic("cannot fork init: %d", error);
704 initproc->p_flags |= P_SYSTEM;
705 reaper_init(initproc, &initreaper);
706 lp = ONLY_LWP_IN_PROC(initproc);
707 cpu_set_fork_handler(lp, start_init, NULL);
710 SYSINIT(init, SI_SUB_CREATE_INIT, SI_ORDER_FIRST, create_init, NULL);
713 * Make it runnable now.
716 kick_init(const void *udata __unused)
718 start_forked_proc(&lwp0, initproc);
720 SYSINIT(kickinit, SI_SUB_KTHREAD_INIT, SI_ORDER_FIRST, kick_init, NULL);
723 kpmap_init(const void *udata __unused)
725 kpmap = kmalloc(roundup2(sizeof(*kpmap), PAGE_SIZE),
726 M_TEMP, M_ZERO | M_WAITOK);
728 kpmap->header[0].type = UKPTYPE_VERSION;
729 kpmap->header[0].offset = offsetof(struct sys_kpmap, version);
730 kpmap->header[1].type = KPTYPE_UPTICKS;
731 kpmap->header[1].offset = offsetof(struct sys_kpmap, upticks);
732 kpmap->header[2].type = KPTYPE_TS_UPTIME;
733 kpmap->header[2].offset = offsetof(struct sys_kpmap, ts_uptime);
734 kpmap->header[3].type = KPTYPE_TS_REALTIME;
735 kpmap->header[3].offset = offsetof(struct sys_kpmap, ts_realtime);
736 kpmap->header[4].type = KPTYPE_TSC_FREQ;
737 kpmap->header[4].offset = offsetof(struct sys_kpmap, tsc_freq);
738 kpmap->header[5].type = KPTYPE_TICK_FREQ;
739 kpmap->header[5].offset = offsetof(struct sys_kpmap, tick_freq);
740 kpmap->header[6].type = KPTYPE_FAST_GTOD;
741 kpmap->header[6].offset = offsetof(struct sys_kpmap, fast_gtod);
742 kpmap->version = KPMAP_VERSION;
744 SYSINIT(kpmapinit, SI_BOOT1_POST, SI_ORDER_FIRST, kpmap_init, NULL);
747 * Machine independant globaldata initialization
749 * WARNING! Called from early boot, 'mycpu' may not work yet.
752 mi_gdinit(struct globaldata *gd, int cpuid)
754 TAILQ_INIT(&gd->gd_systimerq);
755 gd->gd_sysid_alloc = cpuid; /* prime low bits for cpu lookup */
756 gd->gd_cpuid = cpuid;
757 CPUMASK_ASSBIT(gd->gd_cpumask, cpuid);
758 gd->gd_cpumask_simple = CPUMASK_SIMPLE(cpuid);
759 gd->gd_cpumask_offset = (uintptr_t)CPUMASK_ADDR(*(cpumask_t *)0, cpuid);
761 vm_map_entry_reserve_cpu_init(gd);
762 if (gd->gd_cpuid == 0)
763 sleep_early_gdinit(gd);
767 ATOMIC_CPUMASK_ORBIT(usched_global_cpumask, cpuid);
768 gd->gd_vmstats = vmstats;