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/refcount.h>
68 #include <sys/file2.h>
69 #include <sys/thread2.h>
70 #include <sys/sysref2.h>
71 #include <sys/spinlock2.h>
72 #include <sys/mplock2.h>
74 #include <machine/cpu.h>
77 #include <vm/vm_param.h>
80 #include <vm/vm_map.h>
81 #include <vm/vm_extern.h>
83 #include <sys/copyright.h>
85 int vfs_mountroot_devfs(void);
87 /* Components of the first process -- never freed. */
88 static struct session session0;
89 static struct pgrp pgrp0;
90 static struct sigacts sigacts0;
91 static struct filedesc filedesc0;
92 static struct plimit limit0;
93 static struct vmspace vmspace0;
94 struct proc *initproc;
97 struct thread thread0;
100 u_int cpu_mi_feature;
101 cpumask_t usched_global_cpumask;
102 extern struct user *proc0paddr;
103 extern int fallback_elf_brand;
105 int boothowto = 0; /* initialized so that it can be patched */
106 SYSCTL_INT(_debug, OID_AUTO, boothowto, CTLFLAG_RD, &boothowto, 0,
107 "Reboot flags, from console subsystem");
108 SYSCTL_ULONG(_kern, OID_AUTO, usched_global_cpumask, CTLFLAG_RW,
109 &usched_global_cpumask, 0, "global user scheduler cpumask");
112 * This ensures that there is at least one entry so that the sysinit_set
113 * symbol is not undefined. A subsystem ID of SI_SPECIAL_DUMMY is never
116 SYSINIT(placeholder, SI_SPECIAL_DUMMY, SI_ORDER_ANY, NULL, NULL)
119 * The sysinit table itself. Items are checked off as the are run.
120 * If we want to register new sysinit types, add them to newsysinit.
122 SET_DECLARE(sysinit_set, struct sysinit);
123 struct sysinit **sysinit, **sysinit_end;
124 struct sysinit **newsysinit, **newsysinit_end;
128 * Merge a new sysinit set into the current set, reallocating it if
129 * necessary. This can only be called after malloc is running.
132 sysinit_add(struct sysinit **set, struct sysinit **set_end)
134 struct sysinit **newset;
135 struct sysinit **sipp;
136 struct sysinit **xipp;
139 count = set_end - set;
141 count += newsysinit_end - newsysinit;
143 count += sysinit_end - sysinit;
144 newset = kmalloc(count * sizeof(*sipp), M_TEMP, M_WAITOK);
147 for (sipp = newsysinit; sipp < newsysinit_end; sipp++)
150 for (sipp = sysinit; sipp < sysinit_end; sipp++)
153 for (sipp = set; sipp < set_end; sipp++)
156 kfree(newsysinit, M_TEMP);
158 newsysinit_end = newset + count;
162 * Callbacks from machine-dependant startup code (e.g. init386) to set
163 * up low level entities related to cpu #0's globaldata.
165 * Called from very low level boot code.
168 mi_proc0init(struct globaldata *gd, struct user *proc0paddr)
170 lwkt_init_thread(&thread0, proc0paddr, LWKT_THREAD_STACK, 0, gd);
171 lwkt_set_comm(&thread0, "thread0");
172 RB_INIT(&proc0.p_lwp_tree);
173 spin_init(&proc0.p_spin);
174 lwkt_token_init(&proc0.p_token, "iproc");
175 proc0.p_lasttid = 0; /* +1 = next TID */
176 lwp_rb_tree_RB_INSERT(&proc0.p_lwp_tree, &lwp0);
177 lwp0.lwp_thread = &thread0;
178 lwp0.lwp_proc = &proc0;
179 proc0.p_usched = usched_init();
180 lwp0.lwp_cpumask = (cpumask_t)-1;
181 varsymset_init(&proc0.p_varsymset, NULL);
182 thread0.td_flags |= TDF_RUNNING;
183 thread0.td_proc = &proc0;
184 thread0.td_lwp = &lwp0;
185 thread0.td_switch = cpu_lwkt_switch;
186 lwkt_schedule_self(curthread);
190 * System startup; initialize the world, create process 0, mount root
191 * filesystem, and fork to create init and pagedaemon. Most of the
192 * hard work is done in the lower-level initialization routines including
193 * startup(), which does memory initialization and autoconfiguration.
195 * This allows simple addition of new kernel subsystems that require
196 * boot time initialization. It also allows substitution of subsystem
197 * (for instance, a scheduler, kernel profiler, or VM system) by object
198 * module. Finally, it allows for optional "kernel threads".
203 struct sysinit *sip; /* system initialization*/
204 struct sysinit **sipp; /* system initialization*/
205 struct sysinit **xipp; /* interior loop of sort*/
206 struct sysinit *save; /* bubble*/
208 if (sysinit == NULL) {
209 sysinit = SET_BEGIN(sysinit_set);
210 #if defined(__amd64__) && defined(_KERNEL_VIRTUAL)
212 * XXX For whatever reason, on 64-bit vkernels
213 * the value of sysinit obtained from the
214 * linker set is wrong.
216 if ((long)sysinit % 8 != 0) {
217 kprintf("Fixing sysinit value...\n");
218 sysinit = (long)sysinit + 4;
221 sysinit_end = SET_LIMIT(sysinit_set);
223 #if defined(__amd64__) && defined(_KERNEL_VIRTUAL)
224 KKASSERT((long)sysinit % 8 == 0);
229 * Perform a bubble sort of the system initialization objects by
230 * their subsystem (primary key) and order (secondary key).
232 for (sipp = sysinit; sipp < sysinit_end; sipp++) {
233 for (xipp = sipp + 1; xipp < sysinit_end; xipp++) {
234 if ((*sipp)->subsystem < (*xipp)->subsystem ||
235 ((*sipp)->subsystem == (*xipp)->subsystem &&
236 (*sipp)->order <= (*xipp)->order))
245 * Traverse the (now) ordered list of system initialization tasks.
246 * Perform each task, and continue on to the next task.
248 * The last item on the list is expected to be the scheduler,
249 * which will not return.
251 for (sipp = sysinit; sipp < sysinit_end; sipp++) {
253 if (sip->subsystem == SI_SPECIAL_DUMMY)
254 continue; /* skip dummy task(s)*/
256 if (sip->subsystem == SI_SPECIAL_DONE)
260 (*(sip->func))(sip->udata);
262 /* Check off the one we're just done */
263 sip->subsystem = SI_SPECIAL_DONE;
265 /* Check if we've installed more sysinit items via KLD */
266 if (newsysinit != NULL) {
267 if (sysinit != SET_BEGIN(sysinit_set))
268 kfree(sysinit, M_TEMP);
269 sysinit = newsysinit;
270 sysinit_end = newsysinit_end;
272 newsysinit_end = NULL;
277 panic("Shouldn't get here!");
283 ***************************************************************************
285 **** The following SYSINIT's belong elsewhere, but have not yet
288 ***************************************************************************
291 print_caddr_t(void *data __unused)
293 kprintf("%s", (char *)data);
295 SYSINIT(announce, SI_BOOT1_COPYRIGHT, SI_ORDER_FIRST, print_caddr_t, copyright)
298 * Leave the critical section that protected us from spurious interrupts
299 * so device probes work.
302 leavecrit(void *dummy __unused)
304 MachIntrABI.stabilize();
306 MachIntrABI.cleanup();
308 KKASSERT(!IN_CRITICAL_SECT(curthread));
311 kprintf("Leaving critical section, allowing interrupts\n");
313 SYSINIT(leavecrit, SI_BOOT2_LEAVE_CRIT, SI_ORDER_ANY, leavecrit, NULL)
316 * This is called after the threading system is up and running,
317 * including the softclock, clock interrupts, and SMP.
320 tsleepworks(void *dummy __unused)
322 tsleep_now_works = 1;
324 SYSINIT(tsleepworks, SI_BOOT2_FINISH_SMP, SI_ORDER_SECOND, tsleepworks, NULL)
327 * This is called after devices have configured. Tell the kernel we are
328 * no longer in cold boot.
331 endofcoldboot(void *dummy __unused)
335 SYSINIT(endofcoldboot, SI_SUB_ISWARM, SI_ORDER_ANY, endofcoldboot, NULL)
338 ***************************************************************************
340 **** The two following SYSINT's are proc0 specific glue code. I am not
341 **** convinced that they can not be safely combined, but their order of
342 **** operation has been maintained as the same as the original init_main.c
345 **** These probably belong in init_proc.c or kern_proc.c, since they
346 **** deal with proc0 (the fork template process).
348 ***************************************************************************
352 proc0_init(void *dummy __unused)
361 * Initialize process and pgrp structures.
366 * additional VM structures
371 * Create process 0 (the swapper).
373 LIST_INSERT_HEAD(&allproc, p, p_list);
374 LIST_INSERT_HEAD(PGRPHASH(0), &pgrp0, pg_hash);
375 LIST_INIT(&pgrp0.pg_members);
376 lwkt_token_init(&pgrp0.pg_token, "pgrp0");
377 refcount_init(&pgrp0.pg_refs, 1);
378 lockinit(&pgrp0.pg_lock, "pgwt0", 0, 0);
379 LIST_INSERT_HEAD(&pgrp0.pg_members, p, p_pglist);
381 pgrp0.pg_session = &session0;
382 session0.s_count = 1;
383 session0.s_leader = p;
388 p->p_sysent = &aout_sysvec;
390 p->p_flag = P_SYSTEM;
392 lp->lwp_stat = LSRUN;
394 p->p_rtprio.type = RTP_PRIO_NORMAL;
395 p->p_rtprio.prio = 0;
396 lp->lwp_rtprio = p->p_rtprio;
401 bcopy("swapper", p->p_comm, sizeof ("swapper"));
402 bcopy("swapper", thread0.td_comm, sizeof ("swapper"));
404 /* Create credentials. */
405 p->p_ucred = crget();
406 p->p_ucred->cr_ruidinfo = uifind(0);
407 p->p_ucred->cr_ngroups = 1; /* group 0 */
408 p->p_ucred->cr_uidinfo = uifind(0);
409 thread0.td_ucred = crhold(p->p_ucred); /* bootstrap fork1() */
412 p->p_ucred->cr_prison = NULL;
414 /* Create sigacts. */
415 p->p_sigacts = &sigacts0;
416 refcount_init(&p->p_sigacts->ps_refcnt, 1);
418 /* Initialize signal state for process 0. */
421 /* Create the file descriptor table. */
422 fdinit_bootstrap(p, &filedesc0, cmask);
424 /* Create the limits structures. */
425 plimit_init0(&limit0);
426 p->p_limit = &limit0;
428 /* Allocate a prototype map so we have something to fork. */
429 pmap_pinit0(vmspace_pmap(&vmspace0));
430 p->p_vmspace = &vmspace0;
431 lp->lwp_vmspace = p->p_vmspace;
432 sysref_init(&vmspace0.vm_sysref, &vmspace_sysref_class);
433 vm_map_init(&vmspace0.vm_map,
434 round_page(VM_MIN_USER_ADDRESS),
435 trunc_page(VM_MAX_USER_ADDRESS),
436 vmspace_pmap(&vmspace0));
437 sysref_activate(&vmspace0.vm_sysref);
439 kqueue_init(&lwp0.lwp_kqueue, &filedesc0);
442 * Charge root for one process.
444 (void)chgproccnt(p->p_ucred->cr_uidinfo, 1, 0);
447 SYSINIT(p0init, SI_BOOT2_PROC0, SI_ORDER_FIRST, proc0_init, NULL)
449 static int proc0_post_callback(struct proc *p, void *data __unused);
453 proc0_post(void *dummy __unused)
458 * Now we can look at the time, having had a chance to verify the
459 * time from the file system. Pretend that proc0 started now.
461 allproc_scan(proc0_post_callback, NULL);
464 * Give the ``random'' number generator a thump.
465 * XXX: Does read_random() contain enough bits to be used here ?
468 skrandom(ts.tv_sec ^ ts.tv_nsec);
472 proc0_post_callback(struct proc *p, void *data __unused)
474 microtime(&p->p_start);
478 SYSINIT(p0post, SI_SUB_PROC0_POST, SI_ORDER_FIRST, proc0_post, NULL)
481 ***************************************************************************
483 **** The following SYSINIT's and glue code should be moved to the
484 **** respective files on a per subsystem basis.
486 ***************************************************************************
491 ***************************************************************************
493 **** The following code probably belongs in another file, like
494 **** kern/init_init.c.
496 ***************************************************************************
500 * List of paths to try when searching for "init".
502 static char init_path[MAXPATHLEN] =
504 __XSTRING(INIT_PATH);
506 "/sbin/init:/sbin/oinit:/sbin/init.bak";
508 SYSCTL_STRING(_kern, OID_AUTO, init_path, CTLFLAG_RD, init_path, 0, "");
511 * Start the initial user process; try exec'ing each pathname in init_path.
512 * The program is invoked with one argument containing the boot flags.
515 start_init(void *dummy, struct trapframe *frame)
518 struct execve_args args;
520 char *var, *path, *next, *s;
521 char *ucp, **uap, *arg0, *arg1;
529 * This is passed in by the bootloader
531 env = kgetenv("kernelname");
533 strlcpy(kernelname, env, sizeof(kernelname));
536 * The MP lock is not held on entry. We release it before
537 * returning to userland.
542 lp = ONLY_LWP_IN_PROC(p);
544 /* Get the vnode for '/'. Set p->p_fd->fd_cdir to reference it. */
545 mp = mountlist_boot_getfirst();
546 if (VFS_ROOT(mp, &vp))
547 panic("cannot find root vnode");
548 if (mp->mnt_ncmountpt.ncp == NULL) {
549 cache_allocroot(&mp->mnt_ncmountpt, mp, vp);
550 cache_unlock(&mp->mnt_ncmountpt); /* leave ref intact */
552 p->p_fd->fd_cdir = vp;
553 vref(p->p_fd->fd_cdir);
554 p->p_fd->fd_rdir = vp;
555 vref(p->p_fd->fd_rdir);
556 vfs_cache_setroot(vp, cache_hold(&mp->mnt_ncmountpt));
557 vn_unlock(vp); /* leave ref intact */
558 cache_copy(&mp->mnt_ncmountpt, &p->p_fd->fd_ncdir);
559 cache_copy(&mp->mnt_ncmountpt, &p->p_fd->fd_nrdir);
561 kprintf("Mounting devfs\n");
562 vfs_mountroot_devfs();
565 * Need just enough stack to hold the faked-up "execve()" arguments.
567 addr = trunc_page(USRSTACK - PAGE_SIZE);
568 error = vm_map_find(&p->p_vmspace->vm_map, NULL, 0, &addr,
569 PAGE_SIZE, PAGE_SIZE,
570 FALSE, VM_MAPTYPE_NORMAL,
571 VM_PROT_ALL, VM_PROT_ALL,
574 panic("init: couldn't allocate argument space");
575 p->p_vmspace->vm_maxsaddr = (caddr_t)addr;
576 p->p_vmspace->vm_ssize = 1;
578 if ((var = kgetenv("init_path")) != NULL) {
579 strncpy(init_path, var, sizeof init_path);
580 init_path[sizeof init_path - 1] = 0;
582 if ((var = kgetenv("kern.fallback_elf_brand")) != NULL)
583 fallback_elf_brand = strtol(var, NULL, 0);
585 for (path = init_path; *path != '\0'; path = next) {
590 for (next = path; *next != '\0' && *next != ':'; next++)
593 kprintf("start_init: trying %.*s\n", (int)(next - path),
597 * Move out the boot flag argument.
600 ucp = (char *)USRSTACK;
601 (void)subyte(--ucp, 0); /* trailing zero */
602 if (boothowto & RB_SINGLE) {
603 (void)subyte(--ucp, 's');
607 if (boothowto & RB_FASTBOOT) {
608 (void)subyte(--ucp, 'f');
614 (void)subyte(--ucp, 'C');
618 (void)subyte(--ucp, '-');
619 (void)subyte(--ucp, '-'); /* leading hyphen */
623 * Move out the file name (also arg 0).
625 (void)subyte(--ucp, 0);
626 for (s = next - 1; s >= path; s--)
627 (void)subyte(--ucp, *s);
631 * Move out the arg pointers.
633 uap = (char **)((intptr_t)ucp & ~(sizeof(intptr_t)-1));
634 (void)suword((caddr_t)--uap, (long)0); /* terminator */
635 (void)suword((caddr_t)--uap, (long)(intptr_t)arg1);
636 (void)suword((caddr_t)--uap, (long)(intptr_t)arg0);
639 * Point at the arguments.
646 * Now try to exec the program. If can't for any reason
647 * other than it doesn't exist, complain.
649 * Otherwise, return via fork_trampoline() all the way
650 * to user mode as init!
652 * WARNING! We may have been moved to another cpu after
653 * acquiring the current user process designation. The
654 * MP lock will migrate with us though so we still have to
657 if ((error = sys_execve(&args)) == 0) {
659 lp->lwp_proc->p_usched->acquire_curproc(lp);
663 kprintf("exec %.*s: error %d\n", (int)(next - path),
666 kprintf("init: not found in path %s\n", init_path);
671 * Like kthread_create(), but runs in it's own address space.
672 * We do this early to reserve pid 1.
674 * Note special case - do not make it runnable yet. Other work
675 * in progress will change this more.
678 create_init(const void *udata __unused)
684 error = fork1(&lwp0, RFFDG | RFPROC, &initproc);
686 panic("cannot fork init: %d", error);
687 initproc->p_flag |= P_SYSTEM;
688 lp = ONLY_LWP_IN_PROC(initproc);
689 cpu_set_fork_handler(lp, start_init, NULL);
692 SYSINIT(init, SI_SUB_CREATE_INIT, SI_ORDER_FIRST, create_init, NULL)
695 * Make it runnable now.
698 kick_init(const void *udata __unused)
700 start_forked_proc(&lwp0, initproc);
702 SYSINIT(kickinit, SI_SUB_KTHREAD_INIT, SI_ORDER_FIRST, kick_init, NULL)
705 * Machine independant globaldata initialization
707 * WARNING! Called from early boot, 'mycpu' may not work yet.
710 mi_gdinit(struct globaldata *gd, int cpuid)
712 TAILQ_INIT(&gd->gd_systimerq);
713 gd->gd_sysid_alloc = cpuid; /* prime low bits for cpu lookup */
714 gd->gd_cpuid = cpuid;
715 gd->gd_cpumask = CPUMASK(cpuid);
717 vm_map_entry_reserve_cpu_init(gd);
719 usched_global_cpumask |= CPUMASK(cpuid);