1/*
2 * Copyright (c) 1995 Terrence R. Lambert
3 * All rights reserved.
4 *
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.
12 *
13 * Redistribution and use in source and binary forms, with or without
14 * modification, are permitted provided that the following conditions
15 * are met:
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.
28 *
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
39 * SUCH DAMAGE.
40 *
41 * @(#)init_main.c 8.9 (Berkeley) 1/21/94
42 * $FreeBSD: head/sys/kern/init_main.c 108338 2002-12-28 01:23:07Z julian $
43 */
44
45#include "opt_init_path.h"
46#include "opt_mac.h"
47
48#include <sys/param.h>
49#include <sys/kernel.h>
50#include <sys/exec.h>
51#include <sys/file.h>
52#include <sys/filedesc.h>
53#include <sys/ktr.h>
54#include <sys/lock.h>
55#include <sys/mac.h>
56#include <sys/mount.h>
57#include <sys/mutex.h>
58#include <sys/syscallsubr.h>
59#include <sys/sysctl.h>
60#include <sys/proc.h>
61#include <sys/resourcevar.h>
62#include <sys/systm.h>
63#include <sys/signalvar.h>
64#include <sys/vnode.h>
65#include <sys/sysent.h>
66#include <sys/reboot.h>
67#include <sys/sched.h>
68#include <sys/sx.h>
69#include <sys/sysproto.h>
70#include <sys/vmmeter.h>
71#include <sys/unistd.h>
72#include <sys/malloc.h>
73#include <sys/conf.h>
74
75#include <machine/cpu.h>
76
77#include <vm/vm.h>
78#include <vm/vm_param.h>
79#include <vm/pmap.h>
80#include <vm/vm_map.h>
81#include <sys/user.h>
82#include <sys/copyright.h>
83
84void mi_startup(void); /* Should be elsewhere */
85
86/* Components of the first process -- never freed. */
87static struct session session0;
88static struct pgrp pgrp0;
89struct proc proc0;
90struct thread thread0;
91struct kse kse0;
92struct ksegrp ksegrp0;
93static struct procsig procsig0;
94static struct filedesc0 filedesc0;
95static struct plimit limit0;
96static struct vmspace vmspace0;
97struct proc *initproc;
98
99int cmask = CMASK;
100extern int fallback_elf_brand;
101
102struct vnode *rootvp;
103int boothowto = 0; /* initialized so that it can be patched */
104SYSCTL_INT(_debug, OID_AUTO, boothowto, CTLFLAG_RD, &boothowto, 0, "");
105int bootverbose;
106SYSCTL_INT(_debug, OID_AUTO, bootverbose, CTLFLAG_RW, &bootverbose, 0, "");
107
108/*
109 * This ensures that there is at least one entry so that the sysinit_set
110 * symbol is not undefined. A sybsystem ID of SI_SUB_DUMMY is never
111 * executed.
112 */
113SYSINIT(placeholder, SI_SUB_DUMMY, SI_ORDER_ANY, NULL, NULL)
114
115/*
116 * The sysinit table itself. Items are checked off as the are run.
117 * If we want to register new sysinit types, add them to newsysinit.
118 */
119SET_DECLARE(sysinit_set, struct sysinit);
120struct sysinit **sysinit, **sysinit_end;
121struct sysinit **newsysinit, **newsysinit_end;
122
123/*
124 * Merge a new sysinit set into the current set, reallocating it if
125 * necessary. This can only be called after malloc is running.
126 */
127void
128sysinit_add(struct sysinit **set, struct sysinit **set_end)
129{
130 struct sysinit **newset;
131 struct sysinit **sipp;
132 struct sysinit **xipp;
133 int count;
134
135 count = set_end - set;
136 if (newsysinit)
137 count += newsysinit_end - newsysinit;
138 else
139 count += sysinit_end - sysinit;
140 newset = malloc(count * sizeof(*sipp), M_TEMP, M_NOWAIT);
141 if (newset == NULL)
142 panic("cannot malloc for sysinit");
143 xipp = newset;
144 if (newsysinit)
145 for (sipp = newsysinit; sipp < newsysinit_end; sipp++)
146 *xipp++ = *sipp;
147 else
148 for (sipp = sysinit; sipp < sysinit_end; sipp++)
149 *xipp++ = *sipp;
150 for (sipp = set; sipp < set_end; sipp++)
151 *xipp++ = *sipp;
152 if (newsysinit)
153 free(newsysinit, M_TEMP);
154 newsysinit = newset;
155 newsysinit_end = newset + count;
156}
157
158/*
159 * System startup; initialize the world, create process 0, mount root
160 * filesystem, and fork to create init and pagedaemon. Most of the
161 * hard work is done in the lower-level initialization routines including
162 * startup(), which does memory initialization and autoconfiguration.
163 *
164 * This allows simple addition of new kernel subsystems that require
165 * boot time initialization. It also allows substitution of subsystem
166 * (for instance, a scheduler, kernel profiler, or VM system) by object
167 * module. Finally, it allows for optional "kernel threads".
168 */
169void
170mi_startup(void)
171{
172
173 register struct sysinit **sipp; /* system initialization*/
174 register struct sysinit **xipp; /* interior loop of sort*/
175 register struct sysinit *save; /* bubble*/
176
177 if (sysinit == NULL) {
178 sysinit = SET_BEGIN(sysinit_set);
179 sysinit_end = SET_LIMIT(sysinit_set);
180 }
181
182restart:
183 /*
184 * Perform a bubble sort of the system initialization objects by
185 * their subsystem (primary key) and order (secondary key).
186 */
187 for (sipp = sysinit; sipp < sysinit_end; sipp++) {
188 for (xipp = sipp + 1; xipp < sysinit_end; xipp++) {
189 if ((*sipp)->subsystem < (*xipp)->subsystem ||
190 ((*sipp)->subsystem == (*xipp)->subsystem &&
191 (*sipp)->order <= (*xipp)->order))
192 continue; /* skip*/
193 save = *sipp;
194 *sipp = *xipp;
195 *xipp = save;
196 }
197 }
198
199 /*
200 * Traverse the (now) ordered list of system initialization tasks.
201 * Perform each task, and continue on to the next task.
202 *
203 * The last item on the list is expected to be the scheduler,
204 * which will not return.
205 */
206 for (sipp = sysinit; sipp < sysinit_end; sipp++) {
207
208 if ((*sipp)->subsystem == SI_SUB_DUMMY)
209 continue; /* skip dummy task(s)*/
210
211 if ((*sipp)->subsystem == SI_SUB_DONE)
212 continue;
213
214 /* Call function */
215 (*((*sipp)->func))((*sipp)->udata);
216
217 /* Check off the one we're just done */
218 (*sipp)->subsystem = SI_SUB_DONE;
219
220 /* Check if we've installed more sysinit items via KLD */
221 if (newsysinit != NULL) {
222 if (sysinit != SET_BEGIN(sysinit_set))
223 free(sysinit, M_TEMP);
224 sysinit = newsysinit;
225 sysinit_end = newsysinit_end;
226 newsysinit = NULL;
227 newsysinit_end = NULL;
228 goto restart;
229 }
230 }
231
232 panic("Shouldn't get here!");
233 /* NOTREACHED*/
234}
235
236
237/*
238 ***************************************************************************
239 ****
240 **** The following SYSINIT's belong elsewhere, but have not yet
241 **** been moved.
242 ****
243 ***************************************************************************
244 */
245static void
246print_caddr_t(void *data __unused)
247{
248 printf("%s", (char *)data);
249}
250SYSINIT(announce, SI_SUB_COPYRIGHT, SI_ORDER_FIRST, print_caddr_t, copyright)
251SYSINIT(version, SI_SUB_COPYRIGHT, SI_ORDER_SECOND, print_caddr_t, version)
252
253static void
254set_boot_verbose(void *data __unused)
255{
256
257 if (boothowto & RB_VERBOSE)
258 bootverbose++;
259}
260SYSINIT(boot_verbose, SI_SUB_TUNABLES, SI_ORDER_ANY, set_boot_verbose, NULL)
261
262struct sysentvec null_sysvec = {
263 0,
264 NULL,
265 0,
266 0,
267 NULL,
268 0,
269 NULL,
270 NULL,
271 NULL,
272 NULL,
273 NULL,
274 NULL,
275 NULL,
276 "null",
277 NULL,
278 NULL,
279 0,
280 PAGE_SIZE,
281 VM_MIN_ADDRESS,
282 VM_MAXUSER_ADDRESS,
283 USRSTACK,
284 PS_STRINGS,
285 VM_PROT_ALL,
286 NULL,
287 NULL
288};
289
290/*
291 ***************************************************************************
292 ****
293 **** The two following SYSINIT's are proc0 specific glue code. I am not
294 **** convinced that they can not be safely combined, but their order of
295 **** operation has been maintained as the same as the original init_main.c
296 **** for right now.
297 ****
298 **** These probably belong in init_proc.c or kern_proc.c, since they
299 **** deal with proc0 (the fork template process).
300 ****
301 ***************************************************************************
302 */
303/* ARGSUSED*/
304static void
305proc0_init(void *dummy __unused)
306{
307 register struct proc *p;
308 register struct filedesc0 *fdp;
309 register unsigned i;
310 struct thread *td;
311 struct ksegrp *kg;
312 struct kse *ke;
313
314 GIANT_REQUIRED;
315 p = &proc0;
316 td = &thread0;
317 ke = &kse0;
318 kg = &ksegrp0;
319
320 ke->ke_sched = kse0_sched;
321 kg->kg_sched = ksegrp0_sched;
322 p->p_sched = proc0_sched;
323 td->td_sched = thread0_sched;
324
325 /*
326 * Initialize magic number.
327 */
328 p->p_magic = P_MAGIC;
329
330 /*
331 * Initialize thread, process and pgrp structures.
332 */
333 procinit();
334 threadinit();
335
336 /*
337 * Initialize sleep queue hash table
338 */
339 sleepinit();
340
341 /*
342 * additional VM structures
343 */
344 vm_init2();
345
346 /*
347 * Create process 0 (the swapper).
348 */
349 LIST_INSERT_HEAD(&allproc, p, p_list);
350 LIST_INSERT_HEAD(PIDHASH(0), p, p_hash);
351 mtx_init(&pgrp0.pg_mtx, "process group", NULL, MTX_DEF | MTX_DUPOK);
352 p->p_pgrp = &pgrp0;
353 LIST_INSERT_HEAD(PGRPHASH(0), &pgrp0, pg_hash);
354 LIST_INIT(&pgrp0.pg_members);
355 LIST_INSERT_HEAD(&pgrp0.pg_members, p, p_pglist);
356
357 pgrp0.pg_session = &session0;
358 mtx_init(&session0.s_mtx, "session", NULL, MTX_DEF);
359 session0.s_count = 1;
360 session0.s_leader = p;
361
362 p->p_sysent = &null_sysvec;
363
364 /*
365 * proc_linkup was already done in init_i386() or alphainit() etc.
366 * because the earlier code needed to follow td->td_proc. Otherwise
367 * I would have done it here.. maybe this means this should be
368 * done earlier too.
369 */
370 p->p_flag = P_SYSTEM;
371 p->p_sflag = PS_INMEM;
372 p->p_state = PRS_NORMAL;
373 td->td_state = TDS_RUNNING;
374 kg->kg_nice = NZERO;
375 kg->kg_pri_class = PRI_TIMESHARE;
376 kg->kg_user_pri = PUSER;
377 td->td_priority = PVM;
378 td->td_base_pri = PUSER;
379 td->td_kse = ke; /* XXXKSE */
380 ke->ke_oncpu = 0;
381 ke->ke_state = KES_THREAD;
382 ke->ke_thread = td;
383 ke->ke_owner = td;
384 p->p_peers = 0;
385 p->p_leader = p;
386
387
388 bcopy("swapper", p->p_comm, sizeof ("swapper"));
389
390 callout_init(&p->p_itcallout, 0);
391 callout_init(&td->td_slpcallout, 1);
392
393 /* Create credentials. */
394 p->p_ucred = crget();
395 p->p_ucred->cr_ngroups = 1; /* group 0 */
396 p->p_ucred->cr_uidinfo = uifind(0);
397 p->p_ucred->cr_ruidinfo = uifind(0);
398 p->p_ucred->cr_prison = NULL; /* Don't jail it. */
399#ifdef MAC
400 mac_create_proc0(p->p_ucred);
401#endif
402 td->td_ucred = crhold(p->p_ucred);
403
404 /* Create procsig. */
405 p->p_procsig = &procsig0;
406 p->p_procsig->ps_refcnt = 1;
407
408 /* Initialize signal state for process 0. */
409 siginit(&proc0);
410
411 /* Create the file descriptor table. */
412 fdp = &filedesc0;
413 p->p_fd = &fdp->fd_fd;
414 mtx_init(&fdp->fd_fd.fd_mtx, FILEDESC_LOCK_DESC, NULL, MTX_DEF);
415 fdp->fd_fd.fd_refcnt = 1;
416 fdp->fd_fd.fd_cmask = cmask;
417 fdp->fd_fd.fd_ofiles = fdp->fd_dfiles;
418 fdp->fd_fd.fd_ofileflags = fdp->fd_dfileflags;
419 fdp->fd_fd.fd_nfiles = NDFILE;
420
421 /* Create the limits structures. */
422 p->p_limit = &limit0;
423 for (i = 0; i < sizeof(p->p_rlimit)/sizeof(p->p_rlimit[0]); i++)
424 limit0.pl_rlimit[i].rlim_cur =
425 limit0.pl_rlimit[i].rlim_max = RLIM_INFINITY;
426 limit0.pl_rlimit[RLIMIT_NOFILE].rlim_cur =
427 limit0.pl_rlimit[RLIMIT_NOFILE].rlim_max = maxfiles;
428 limit0.pl_rlimit[RLIMIT_NPROC].rlim_cur =
429 limit0.pl_rlimit[RLIMIT_NPROC].rlim_max = maxproc;
430 i = ptoa(cnt.v_free_count);
431 limit0.pl_rlimit[RLIMIT_RSS].rlim_max = i;
432 limit0.pl_rlimit[RLIMIT_MEMLOCK].rlim_max = i;
433 limit0.pl_rlimit[RLIMIT_MEMLOCK].rlim_cur = i / 3;
434 limit0.p_refcnt = 1;
435 p->p_cpulimit = RLIM_INFINITY;
436
437 /* Allocate a prototype map so we have something to fork. */
438 pmap_pinit0(vmspace_pmap(&vmspace0));
439 p->p_vmspace = &vmspace0;
440 vmspace0.vm_refcnt = 1;
441 vm_map_init(&vmspace0.vm_map, p->p_sysent->sv_minuser,
442 p->p_sysent->sv_maxuser);
443 vmspace0.vm_map.pmap = vmspace_pmap(&vmspace0);
444
445 /*
446 * We continue to place resource usage info and signal
447 * actions in the user struct so they're pageable.
448 */
449 p->p_stats = &p->p_uarea->u_stats;
450 p->p_sigacts = &p->p_uarea->u_sigacts;
451
452 /*
453 * Charge root for one process.
454 */
455 (void)chgproccnt(p->p_ucred->cr_ruidinfo, 1, 0);
456}
457SYSINIT(p0init, SI_SUB_INTRINSIC, SI_ORDER_FIRST, proc0_init, NULL)
458
459/* ARGSUSED*/
460static void
461proc0_post(void *dummy __unused)
462{
463 struct timespec ts;
464 struct proc *p;
465
466 /*
467 * Now we can look at the time, having had a chance to verify the
468 * time from the filesystem. Pretend that proc0 started now.
469 */
470 sx_slock(&allproc_lock);
471 LIST_FOREACH(p, &allproc, p_list) {
472 microtime(&p->p_stats->p_start);
473 p->p_runtime.sec = 0;
474 p->p_runtime.frac = 0;
475 }
476 sx_sunlock(&allproc_lock);
477 binuptime(PCPU_PTR(switchtime));
478 PCPU_SET(switchticks, ticks);
479
480 /*
481 * Give the ``random'' number generator a thump.
482 */
483 nanotime(&ts);
484 srandom(ts.tv_sec ^ ts.tv_nsec);
485}
486SYSINIT(p0post, SI_SUB_INTRINSIC_POST, SI_ORDER_FIRST, proc0_post, NULL)
487
488/*
489 ***************************************************************************
490 ****
491 **** The following SYSINIT's and glue code should be moved to the
492 **** respective files on a per subsystem basis.
493 ****
494 ***************************************************************************
495 */
496
497
498/*
499 ***************************************************************************
500 ****
501 **** The following code probably belongs in another file, like
502 **** kern/init_init.c.
503 ****
504 ***************************************************************************
505 */
506
507/*
508 * List of paths to try when searching for "init".
509 */
510static char init_path[MAXPATHLEN] =
511#ifdef INIT_PATH
512 __XSTRING(INIT_PATH);
513#else
514 "/sbin/init:/sbin/oinit:/sbin/init.bak:/stand/sysinstall";
515#endif
516SYSCTL_STRING(_kern, OID_AUTO, init_path, CTLFLAG_RD, init_path, 0,
517 "Path used to search the init process");
518
519/*
520 * Start the initial user process; try exec'ing each pathname in init_path.
521 * The program is invoked with one argument containing the boot flags.
522 */
523static void
524start_init(void *dummy)
525{
526 vm_offset_t addr;
527 struct execve_args args;
528 int options, error;
529 char *var, *path, *next, *s;
530 char *ucp, **uap, *arg0, *arg1;
531 struct thread *td;
532 struct proc *p;
533 int init_does_devfs = 0;
534
535 mtx_lock(&Giant);
536
537 GIANT_REQUIRED;
538
539 td = curthread;
540 p = td->td_proc;
541
542 vfs_mountroot();
543
544 /* Get the vnode for '/'. Set p->p_fd->fd_cdir to reference it. */
545 if (VFS_ROOT(TAILQ_FIRST(&mountlist), &rootvnode))
546 panic("cannot find root vnode");
547 FILEDESC_LOCK(p->p_fd);
548 p->p_fd->fd_cdir = rootvnode;
549 VREF(p->p_fd->fd_cdir);
550 p->p_fd->fd_rdir = rootvnode;
551 VREF(p->p_fd->fd_rdir);
552 FILEDESC_UNLOCK(p->p_fd);
553 VOP_UNLOCK(rootvnode, 0, td);
554#ifdef MAC
555 mac_create_root_mount(td->td_ucred, TAILQ_FIRST(&mountlist));
556#endif
557
558 if (devfs_present) {
559 /*
560 * For disk based systems, we probably cannot do this yet
561 * since the fs will be read-only. But a NFS root
562 * might be ok. It is worth a shot.
563 */
564 error = kern_mkdir(td, "/dev", UIO_SYSSPACE, 0700);
565 if (error == EEXIST)
566 error = 0;
567 if (error == 0)
568 error = kernel_vmount(0, "fstype", "devfs",
569 "fspath", "/dev", NULL);
570 if (error != 0)
571 init_does_devfs = 1;
572 }
573
574 /*
575 * Need just enough stack to hold the faked-up "execve()" arguments.
576 */
577 addr = p->p_sysent->sv_usrstack - PAGE_SIZE;
578 if (vm_map_find(&p->p_vmspace->vm_map, NULL, 0, &addr, PAGE_SIZE,
579 FALSE, VM_PROT_ALL, VM_PROT_ALL, 0) != 0)
580 panic("init: couldn't allocate argument space");
581 p->p_vmspace->vm_maxsaddr = (caddr_t)addr;
582 p->p_vmspace->vm_ssize = 1;
583
584 if ((var = getenv("init_path")) != NULL) {
585 strlcpy(init_path, var, sizeof(init_path));
586 freeenv(var);
587 }
588 if ((var = getenv("kern.fallback_elf_brand")) != NULL) {
589 fallback_elf_brand = strtol(var, NULL, 0);
590 freeenv(var);
591 }
592
593 for (path = init_path; *path != '\0'; path = next) {
594 while (*path == ':')
595 path++;
596 if (*path == '\0')
597 break;
598 for (next = path; *next != '\0' && *next != ':'; next++)
599 /* nothing */ ;
600 if (bootverbose)
601 printf("start_init: trying %.*s\n", (int)(next - path),
602 path);
603
604 /*
605 * Move out the boot flag argument.
606 */
607 options = 0;
608 ucp = (char *)p->p_sysent->sv_usrstack;
609 (void)subyte(--ucp, 0); /* trailing zero */
610 if (boothowto & RB_SINGLE) {
611 (void)subyte(--ucp, 's');
612 options = 1;
613 }
614#ifdef notyet
615 if (boothowto & RB_FASTBOOT) {
616 (void)subyte(--ucp, 'f');
617 options = 1;
618 }
619#endif
620
621#ifdef BOOTCDROM
622 (void)subyte(--ucp, 'C');
623 options = 1;
624#endif
625 if (init_does_devfs) {
626 (void)subyte(--ucp, 'd');
627 options = 1;
628 }
629
630 if (options == 0)
631 (void)subyte(--ucp, '-');
632 (void)subyte(--ucp, '-'); /* leading hyphen */
633 arg1 = ucp;
634
635 /*
636 * Move out the file name (also arg 0).
637 */
638 (void)subyte(--ucp, 0);
639 for (s = next - 1; s >= path; s--)
640 (void)subyte(--ucp, *s);
641 arg0 = ucp;
642
643 /*
644 * Move out the arg pointers.
645 */
646 uap = (char **)((intptr_t)ucp & ~(sizeof(intptr_t)-1));
647 (void)suword((caddr_t)--uap, (long)0); /* terminator */
648 (void)suword((caddr_t)--uap, (long)(intptr_t)arg1);
649 (void)suword((caddr_t)--uap, (long)(intptr_t)arg0);
650
651 /*
652 * Point at the arguments.
653 */
654 args.fname = arg0;
655 args.argv = uap;
656 args.envv = NULL;
657
658 /*
659 * Now try to exec the program. If can't for any reason
660 * other than it doesn't exist, complain.
661 *
662 * Otherwise, return via fork_trampoline() all the way
663 * to user mode as init!
664 */
665 if ((error = execve(td, &args)) == 0) {
666 mtx_unlock(&Giant);
667 return;
668 }
669 if (error != ENOENT)
670 printf("exec %.*s: error %d\n", (int)(next - path),
671 path, error);
672 }
673 printf("init: not found in path %s\n", init_path);
674 panic("no init");
675}
676
677/*
678 * Like kthread_create(), but runs in it's own address space.
679 * We do this early to reserve pid 1.
680 *
681 * Note special case - do not make it runnable yet. Other work
682 * in progress will change this more.
683 */
684static void
685create_init(const void *udata __unused)
686{
687 struct ucred *newcred, *oldcred;
688 int error;
689
690 error = fork1(&thread0, RFFDG | RFPROC | RFSTOPPED, 0, &initproc);
691 if (error)
692 panic("cannot fork init: %d\n", error);
693 /* divorce init's credentials from the kernel's */
694 newcred = crget();
695 PROC_LOCK(initproc);
696 initproc->p_flag |= P_SYSTEM;
697 oldcred = initproc->p_ucred;
698 crcopy(newcred, oldcred);
699#ifdef MAC
700 mac_create_proc1(newcred);
701#endif
702 initproc->p_ucred = newcred;
703 PROC_UNLOCK(initproc);
704 crfree(oldcred);
705 cred_update_thread(FIRST_THREAD_IN_PROC(initproc));
706 mtx_lock_spin(&sched_lock);
707 initproc->p_sflag |= PS_INMEM;
708 mtx_unlock_spin(&sched_lock);
709 cpu_set_fork_handler(FIRST_THREAD_IN_PROC(initproc), start_init, NULL);
710}
711SYSINIT(init, SI_SUB_CREATE_INIT, SI_ORDER_FIRST, create_init, NULL)
712
713/*
714 * Make it runnable now.
715 */
716static void
717kick_init(const void *udata __unused)
718{
719 struct thread *td;
720
721 td = FIRST_THREAD_IN_PROC(initproc);
722 mtx_lock_spin(&sched_lock);
723 TD_SET_CAN_RUN(td);
724 setrunqueue(td); /* XXXKSE */
725 mtx_unlock_spin(&sched_lock);
726}
727SYSINIT(kickinit, SI_SUB_KTHREAD_INIT, SI_ORDER_FIRST, kick_init, NULL)
2 * Copyright (c) 1995 Terrence R. Lambert
3 * All rights reserved.
4 *
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.
12 *
13 * Redistribution and use in source and binary forms, with or without
14 * modification, are permitted provided that the following conditions
15 * are met:
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.
28 *
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
39 * SUCH DAMAGE.
40 *
41 * @(#)init_main.c 8.9 (Berkeley) 1/21/94
42 * $FreeBSD: head/sys/kern/init_main.c 108338 2002-12-28 01:23:07Z julian $
43 */
44
45#include "opt_init_path.h"
46#include "opt_mac.h"
47
48#include <sys/param.h>
49#include <sys/kernel.h>
50#include <sys/exec.h>
51#include <sys/file.h>
52#include <sys/filedesc.h>
53#include <sys/ktr.h>
54#include <sys/lock.h>
55#include <sys/mac.h>
56#include <sys/mount.h>
57#include <sys/mutex.h>
58#include <sys/syscallsubr.h>
59#include <sys/sysctl.h>
60#include <sys/proc.h>
61#include <sys/resourcevar.h>
62#include <sys/systm.h>
63#include <sys/signalvar.h>
64#include <sys/vnode.h>
65#include <sys/sysent.h>
66#include <sys/reboot.h>
67#include <sys/sched.h>
68#include <sys/sx.h>
69#include <sys/sysproto.h>
70#include <sys/vmmeter.h>
71#include <sys/unistd.h>
72#include <sys/malloc.h>
73#include <sys/conf.h>
74
75#include <machine/cpu.h>
76
77#include <vm/vm.h>
78#include <vm/vm_param.h>
79#include <vm/pmap.h>
80#include <vm/vm_map.h>
81#include <sys/user.h>
82#include <sys/copyright.h>
83
84void mi_startup(void); /* Should be elsewhere */
85
86/* Components of the first process -- never freed. */
87static struct session session0;
88static struct pgrp pgrp0;
89struct proc proc0;
90struct thread thread0;
91struct kse kse0;
92struct ksegrp ksegrp0;
93static struct procsig procsig0;
94static struct filedesc0 filedesc0;
95static struct plimit limit0;
96static struct vmspace vmspace0;
97struct proc *initproc;
98
99int cmask = CMASK;
100extern int fallback_elf_brand;
101
102struct vnode *rootvp;
103int boothowto = 0; /* initialized so that it can be patched */
104SYSCTL_INT(_debug, OID_AUTO, boothowto, CTLFLAG_RD, &boothowto, 0, "");
105int bootverbose;
106SYSCTL_INT(_debug, OID_AUTO, bootverbose, CTLFLAG_RW, &bootverbose, 0, "");
107
108/*
109 * This ensures that there is at least one entry so that the sysinit_set
110 * symbol is not undefined. A sybsystem ID of SI_SUB_DUMMY is never
111 * executed.
112 */
113SYSINIT(placeholder, SI_SUB_DUMMY, SI_ORDER_ANY, NULL, NULL)
114
115/*
116 * The sysinit table itself. Items are checked off as the are run.
117 * If we want to register new sysinit types, add them to newsysinit.
118 */
119SET_DECLARE(sysinit_set, struct sysinit);
120struct sysinit **sysinit, **sysinit_end;
121struct sysinit **newsysinit, **newsysinit_end;
122
123/*
124 * Merge a new sysinit set into the current set, reallocating it if
125 * necessary. This can only be called after malloc is running.
126 */
127void
128sysinit_add(struct sysinit **set, struct sysinit **set_end)
129{
130 struct sysinit **newset;
131 struct sysinit **sipp;
132 struct sysinit **xipp;
133 int count;
134
135 count = set_end - set;
136 if (newsysinit)
137 count += newsysinit_end - newsysinit;
138 else
139 count += sysinit_end - sysinit;
140 newset = malloc(count * sizeof(*sipp), M_TEMP, M_NOWAIT);
141 if (newset == NULL)
142 panic("cannot malloc for sysinit");
143 xipp = newset;
144 if (newsysinit)
145 for (sipp = newsysinit; sipp < newsysinit_end; sipp++)
146 *xipp++ = *sipp;
147 else
148 for (sipp = sysinit; sipp < sysinit_end; sipp++)
149 *xipp++ = *sipp;
150 for (sipp = set; sipp < set_end; sipp++)
151 *xipp++ = *sipp;
152 if (newsysinit)
153 free(newsysinit, M_TEMP);
154 newsysinit = newset;
155 newsysinit_end = newset + count;
156}
157
158/*
159 * System startup; initialize the world, create process 0, mount root
160 * filesystem, and fork to create init and pagedaemon. Most of the
161 * hard work is done in the lower-level initialization routines including
162 * startup(), which does memory initialization and autoconfiguration.
163 *
164 * This allows simple addition of new kernel subsystems that require
165 * boot time initialization. It also allows substitution of subsystem
166 * (for instance, a scheduler, kernel profiler, or VM system) by object
167 * module. Finally, it allows for optional "kernel threads".
168 */
169void
170mi_startup(void)
171{
172
173 register struct sysinit **sipp; /* system initialization*/
174 register struct sysinit **xipp; /* interior loop of sort*/
175 register struct sysinit *save; /* bubble*/
176
177 if (sysinit == NULL) {
178 sysinit = SET_BEGIN(sysinit_set);
179 sysinit_end = SET_LIMIT(sysinit_set);
180 }
181
182restart:
183 /*
184 * Perform a bubble sort of the system initialization objects by
185 * their subsystem (primary key) and order (secondary key).
186 */
187 for (sipp = sysinit; sipp < sysinit_end; sipp++) {
188 for (xipp = sipp + 1; xipp < sysinit_end; xipp++) {
189 if ((*sipp)->subsystem < (*xipp)->subsystem ||
190 ((*sipp)->subsystem == (*xipp)->subsystem &&
191 (*sipp)->order <= (*xipp)->order))
192 continue; /* skip*/
193 save = *sipp;
194 *sipp = *xipp;
195 *xipp = save;
196 }
197 }
198
199 /*
200 * Traverse the (now) ordered list of system initialization tasks.
201 * Perform each task, and continue on to the next task.
202 *
203 * The last item on the list is expected to be the scheduler,
204 * which will not return.
205 */
206 for (sipp = sysinit; sipp < sysinit_end; sipp++) {
207
208 if ((*sipp)->subsystem == SI_SUB_DUMMY)
209 continue; /* skip dummy task(s)*/
210
211 if ((*sipp)->subsystem == SI_SUB_DONE)
212 continue;
213
214 /* Call function */
215 (*((*sipp)->func))((*sipp)->udata);
216
217 /* Check off the one we're just done */
218 (*sipp)->subsystem = SI_SUB_DONE;
219
220 /* Check if we've installed more sysinit items via KLD */
221 if (newsysinit != NULL) {
222 if (sysinit != SET_BEGIN(sysinit_set))
223 free(sysinit, M_TEMP);
224 sysinit = newsysinit;
225 sysinit_end = newsysinit_end;
226 newsysinit = NULL;
227 newsysinit_end = NULL;
228 goto restart;
229 }
230 }
231
232 panic("Shouldn't get here!");
233 /* NOTREACHED*/
234}
235
236
237/*
238 ***************************************************************************
239 ****
240 **** The following SYSINIT's belong elsewhere, but have not yet
241 **** been moved.
242 ****
243 ***************************************************************************
244 */
245static void
246print_caddr_t(void *data __unused)
247{
248 printf("%s", (char *)data);
249}
250SYSINIT(announce, SI_SUB_COPYRIGHT, SI_ORDER_FIRST, print_caddr_t, copyright)
251SYSINIT(version, SI_SUB_COPYRIGHT, SI_ORDER_SECOND, print_caddr_t, version)
252
253static void
254set_boot_verbose(void *data __unused)
255{
256
257 if (boothowto & RB_VERBOSE)
258 bootverbose++;
259}
260SYSINIT(boot_verbose, SI_SUB_TUNABLES, SI_ORDER_ANY, set_boot_verbose, NULL)
261
262struct sysentvec null_sysvec = {
263 0,
264 NULL,
265 0,
266 0,
267 NULL,
268 0,
269 NULL,
270 NULL,
271 NULL,
272 NULL,
273 NULL,
274 NULL,
275 NULL,
276 "null",
277 NULL,
278 NULL,
279 0,
280 PAGE_SIZE,
281 VM_MIN_ADDRESS,
282 VM_MAXUSER_ADDRESS,
283 USRSTACK,
284 PS_STRINGS,
285 VM_PROT_ALL,
286 NULL,
287 NULL
288};
289
290/*
291 ***************************************************************************
292 ****
293 **** The two following SYSINIT's are proc0 specific glue code. I am not
294 **** convinced that they can not be safely combined, but their order of
295 **** operation has been maintained as the same as the original init_main.c
296 **** for right now.
297 ****
298 **** These probably belong in init_proc.c or kern_proc.c, since they
299 **** deal with proc0 (the fork template process).
300 ****
301 ***************************************************************************
302 */
303/* ARGSUSED*/
304static void
305proc0_init(void *dummy __unused)
306{
307 register struct proc *p;
308 register struct filedesc0 *fdp;
309 register unsigned i;
310 struct thread *td;
311 struct ksegrp *kg;
312 struct kse *ke;
313
314 GIANT_REQUIRED;
315 p = &proc0;
316 td = &thread0;
317 ke = &kse0;
318 kg = &ksegrp0;
319
320 ke->ke_sched = kse0_sched;
321 kg->kg_sched = ksegrp0_sched;
322 p->p_sched = proc0_sched;
323 td->td_sched = thread0_sched;
324
325 /*
326 * Initialize magic number.
327 */
328 p->p_magic = P_MAGIC;
329
330 /*
331 * Initialize thread, process and pgrp structures.
332 */
333 procinit();
334 threadinit();
335
336 /*
337 * Initialize sleep queue hash table
338 */
339 sleepinit();
340
341 /*
342 * additional VM structures
343 */
344 vm_init2();
345
346 /*
347 * Create process 0 (the swapper).
348 */
349 LIST_INSERT_HEAD(&allproc, p, p_list);
350 LIST_INSERT_HEAD(PIDHASH(0), p, p_hash);
351 mtx_init(&pgrp0.pg_mtx, "process group", NULL, MTX_DEF | MTX_DUPOK);
352 p->p_pgrp = &pgrp0;
353 LIST_INSERT_HEAD(PGRPHASH(0), &pgrp0, pg_hash);
354 LIST_INIT(&pgrp0.pg_members);
355 LIST_INSERT_HEAD(&pgrp0.pg_members, p, p_pglist);
356
357 pgrp0.pg_session = &session0;
358 mtx_init(&session0.s_mtx, "session", NULL, MTX_DEF);
359 session0.s_count = 1;
360 session0.s_leader = p;
361
362 p->p_sysent = &null_sysvec;
363
364 /*
365 * proc_linkup was already done in init_i386() or alphainit() etc.
366 * because the earlier code needed to follow td->td_proc. Otherwise
367 * I would have done it here.. maybe this means this should be
368 * done earlier too.
369 */
370 p->p_flag = P_SYSTEM;
371 p->p_sflag = PS_INMEM;
372 p->p_state = PRS_NORMAL;
373 td->td_state = TDS_RUNNING;
374 kg->kg_nice = NZERO;
375 kg->kg_pri_class = PRI_TIMESHARE;
376 kg->kg_user_pri = PUSER;
377 td->td_priority = PVM;
378 td->td_base_pri = PUSER;
379 td->td_kse = ke; /* XXXKSE */
380 ke->ke_oncpu = 0;
381 ke->ke_state = KES_THREAD;
382 ke->ke_thread = td;
383 ke->ke_owner = td;
384 p->p_peers = 0;
385 p->p_leader = p;
386
387
388 bcopy("swapper", p->p_comm, sizeof ("swapper"));
389
390 callout_init(&p->p_itcallout, 0);
391 callout_init(&td->td_slpcallout, 1);
392
393 /* Create credentials. */
394 p->p_ucred = crget();
395 p->p_ucred->cr_ngroups = 1; /* group 0 */
396 p->p_ucred->cr_uidinfo = uifind(0);
397 p->p_ucred->cr_ruidinfo = uifind(0);
398 p->p_ucred->cr_prison = NULL; /* Don't jail it. */
399#ifdef MAC
400 mac_create_proc0(p->p_ucred);
401#endif
402 td->td_ucred = crhold(p->p_ucred);
403
404 /* Create procsig. */
405 p->p_procsig = &procsig0;
406 p->p_procsig->ps_refcnt = 1;
407
408 /* Initialize signal state for process 0. */
409 siginit(&proc0);
410
411 /* Create the file descriptor table. */
412 fdp = &filedesc0;
413 p->p_fd = &fdp->fd_fd;
414 mtx_init(&fdp->fd_fd.fd_mtx, FILEDESC_LOCK_DESC, NULL, MTX_DEF);
415 fdp->fd_fd.fd_refcnt = 1;
416 fdp->fd_fd.fd_cmask = cmask;
417 fdp->fd_fd.fd_ofiles = fdp->fd_dfiles;
418 fdp->fd_fd.fd_ofileflags = fdp->fd_dfileflags;
419 fdp->fd_fd.fd_nfiles = NDFILE;
420
421 /* Create the limits structures. */
422 p->p_limit = &limit0;
423 for (i = 0; i < sizeof(p->p_rlimit)/sizeof(p->p_rlimit[0]); i++)
424 limit0.pl_rlimit[i].rlim_cur =
425 limit0.pl_rlimit[i].rlim_max = RLIM_INFINITY;
426 limit0.pl_rlimit[RLIMIT_NOFILE].rlim_cur =
427 limit0.pl_rlimit[RLIMIT_NOFILE].rlim_max = maxfiles;
428 limit0.pl_rlimit[RLIMIT_NPROC].rlim_cur =
429 limit0.pl_rlimit[RLIMIT_NPROC].rlim_max = maxproc;
430 i = ptoa(cnt.v_free_count);
431 limit0.pl_rlimit[RLIMIT_RSS].rlim_max = i;
432 limit0.pl_rlimit[RLIMIT_MEMLOCK].rlim_max = i;
433 limit0.pl_rlimit[RLIMIT_MEMLOCK].rlim_cur = i / 3;
434 limit0.p_refcnt = 1;
435 p->p_cpulimit = RLIM_INFINITY;
436
437 /* Allocate a prototype map so we have something to fork. */
438 pmap_pinit0(vmspace_pmap(&vmspace0));
439 p->p_vmspace = &vmspace0;
440 vmspace0.vm_refcnt = 1;
441 vm_map_init(&vmspace0.vm_map, p->p_sysent->sv_minuser,
442 p->p_sysent->sv_maxuser);
443 vmspace0.vm_map.pmap = vmspace_pmap(&vmspace0);
444
445 /*
446 * We continue to place resource usage info and signal
447 * actions in the user struct so they're pageable.
448 */
449 p->p_stats = &p->p_uarea->u_stats;
450 p->p_sigacts = &p->p_uarea->u_sigacts;
451
452 /*
453 * Charge root for one process.
454 */
455 (void)chgproccnt(p->p_ucred->cr_ruidinfo, 1, 0);
456}
457SYSINIT(p0init, SI_SUB_INTRINSIC, SI_ORDER_FIRST, proc0_init, NULL)
458
459/* ARGSUSED*/
460static void
461proc0_post(void *dummy __unused)
462{
463 struct timespec ts;
464 struct proc *p;
465
466 /*
467 * Now we can look at the time, having had a chance to verify the
468 * time from the filesystem. Pretend that proc0 started now.
469 */
470 sx_slock(&allproc_lock);
471 LIST_FOREACH(p, &allproc, p_list) {
472 microtime(&p->p_stats->p_start);
473 p->p_runtime.sec = 0;
474 p->p_runtime.frac = 0;
475 }
476 sx_sunlock(&allproc_lock);
477 binuptime(PCPU_PTR(switchtime));
478 PCPU_SET(switchticks, ticks);
479
480 /*
481 * Give the ``random'' number generator a thump.
482 */
483 nanotime(&ts);
484 srandom(ts.tv_sec ^ ts.tv_nsec);
485}
486SYSINIT(p0post, SI_SUB_INTRINSIC_POST, SI_ORDER_FIRST, proc0_post, NULL)
487
488/*
489 ***************************************************************************
490 ****
491 **** The following SYSINIT's and glue code should be moved to the
492 **** respective files on a per subsystem basis.
493 ****
494 ***************************************************************************
495 */
496
497
498/*
499 ***************************************************************************
500 ****
501 **** The following code probably belongs in another file, like
502 **** kern/init_init.c.
503 ****
504 ***************************************************************************
505 */
506
507/*
508 * List of paths to try when searching for "init".
509 */
510static char init_path[MAXPATHLEN] =
511#ifdef INIT_PATH
512 __XSTRING(INIT_PATH);
513#else
514 "/sbin/init:/sbin/oinit:/sbin/init.bak:/stand/sysinstall";
515#endif
516SYSCTL_STRING(_kern, OID_AUTO, init_path, CTLFLAG_RD, init_path, 0,
517 "Path used to search the init process");
518
519/*
520 * Start the initial user process; try exec'ing each pathname in init_path.
521 * The program is invoked with one argument containing the boot flags.
522 */
523static void
524start_init(void *dummy)
525{
526 vm_offset_t addr;
527 struct execve_args args;
528 int options, error;
529 char *var, *path, *next, *s;
530 char *ucp, **uap, *arg0, *arg1;
531 struct thread *td;
532 struct proc *p;
533 int init_does_devfs = 0;
534
535 mtx_lock(&Giant);
536
537 GIANT_REQUIRED;
538
539 td = curthread;
540 p = td->td_proc;
541
542 vfs_mountroot();
543
544 /* Get the vnode for '/'. Set p->p_fd->fd_cdir to reference it. */
545 if (VFS_ROOT(TAILQ_FIRST(&mountlist), &rootvnode))
546 panic("cannot find root vnode");
547 FILEDESC_LOCK(p->p_fd);
548 p->p_fd->fd_cdir = rootvnode;
549 VREF(p->p_fd->fd_cdir);
550 p->p_fd->fd_rdir = rootvnode;
551 VREF(p->p_fd->fd_rdir);
552 FILEDESC_UNLOCK(p->p_fd);
553 VOP_UNLOCK(rootvnode, 0, td);
554#ifdef MAC
555 mac_create_root_mount(td->td_ucred, TAILQ_FIRST(&mountlist));
556#endif
557
558 if (devfs_present) {
559 /*
560 * For disk based systems, we probably cannot do this yet
561 * since the fs will be read-only. But a NFS root
562 * might be ok. It is worth a shot.
563 */
564 error = kern_mkdir(td, "/dev", UIO_SYSSPACE, 0700);
565 if (error == EEXIST)
566 error = 0;
567 if (error == 0)
568 error = kernel_vmount(0, "fstype", "devfs",
569 "fspath", "/dev", NULL);
570 if (error != 0)
571 init_does_devfs = 1;
572 }
573
574 /*
575 * Need just enough stack to hold the faked-up "execve()" arguments.
576 */
577 addr = p->p_sysent->sv_usrstack - PAGE_SIZE;
578 if (vm_map_find(&p->p_vmspace->vm_map, NULL, 0, &addr, PAGE_SIZE,
579 FALSE, VM_PROT_ALL, VM_PROT_ALL, 0) != 0)
580 panic("init: couldn't allocate argument space");
581 p->p_vmspace->vm_maxsaddr = (caddr_t)addr;
582 p->p_vmspace->vm_ssize = 1;
583
584 if ((var = getenv("init_path")) != NULL) {
585 strlcpy(init_path, var, sizeof(init_path));
586 freeenv(var);
587 }
588 if ((var = getenv("kern.fallback_elf_brand")) != NULL) {
589 fallback_elf_brand = strtol(var, NULL, 0);
590 freeenv(var);
591 }
592
593 for (path = init_path; *path != '\0'; path = next) {
594 while (*path == ':')
595 path++;
596 if (*path == '\0')
597 break;
598 for (next = path; *next != '\0' && *next != ':'; next++)
599 /* nothing */ ;
600 if (bootverbose)
601 printf("start_init: trying %.*s\n", (int)(next - path),
602 path);
603
604 /*
605 * Move out the boot flag argument.
606 */
607 options = 0;
608 ucp = (char *)p->p_sysent->sv_usrstack;
609 (void)subyte(--ucp, 0); /* trailing zero */
610 if (boothowto & RB_SINGLE) {
611 (void)subyte(--ucp, 's');
612 options = 1;
613 }
614#ifdef notyet
615 if (boothowto & RB_FASTBOOT) {
616 (void)subyte(--ucp, 'f');
617 options = 1;
618 }
619#endif
620
621#ifdef BOOTCDROM
622 (void)subyte(--ucp, 'C');
623 options = 1;
624#endif
625 if (init_does_devfs) {
626 (void)subyte(--ucp, 'd');
627 options = 1;
628 }
629
630 if (options == 0)
631 (void)subyte(--ucp, '-');
632 (void)subyte(--ucp, '-'); /* leading hyphen */
633 arg1 = ucp;
634
635 /*
636 * Move out the file name (also arg 0).
637 */
638 (void)subyte(--ucp, 0);
639 for (s = next - 1; s >= path; s--)
640 (void)subyte(--ucp, *s);
641 arg0 = ucp;
642
643 /*
644 * Move out the arg pointers.
645 */
646 uap = (char **)((intptr_t)ucp & ~(sizeof(intptr_t)-1));
647 (void)suword((caddr_t)--uap, (long)0); /* terminator */
648 (void)suword((caddr_t)--uap, (long)(intptr_t)arg1);
649 (void)suword((caddr_t)--uap, (long)(intptr_t)arg0);
650
651 /*
652 * Point at the arguments.
653 */
654 args.fname = arg0;
655 args.argv = uap;
656 args.envv = NULL;
657
658 /*
659 * Now try to exec the program. If can't for any reason
660 * other than it doesn't exist, complain.
661 *
662 * Otherwise, return via fork_trampoline() all the way
663 * to user mode as init!
664 */
665 if ((error = execve(td, &args)) == 0) {
666 mtx_unlock(&Giant);
667 return;
668 }
669 if (error != ENOENT)
670 printf("exec %.*s: error %d\n", (int)(next - path),
671 path, error);
672 }
673 printf("init: not found in path %s\n", init_path);
674 panic("no init");
675}
676
677/*
678 * Like kthread_create(), but runs in it's own address space.
679 * We do this early to reserve pid 1.
680 *
681 * Note special case - do not make it runnable yet. Other work
682 * in progress will change this more.
683 */
684static void
685create_init(const void *udata __unused)
686{
687 struct ucred *newcred, *oldcred;
688 int error;
689
690 error = fork1(&thread0, RFFDG | RFPROC | RFSTOPPED, 0, &initproc);
691 if (error)
692 panic("cannot fork init: %d\n", error);
693 /* divorce init's credentials from the kernel's */
694 newcred = crget();
695 PROC_LOCK(initproc);
696 initproc->p_flag |= P_SYSTEM;
697 oldcred = initproc->p_ucred;
698 crcopy(newcred, oldcred);
699#ifdef MAC
700 mac_create_proc1(newcred);
701#endif
702 initproc->p_ucred = newcred;
703 PROC_UNLOCK(initproc);
704 crfree(oldcred);
705 cred_update_thread(FIRST_THREAD_IN_PROC(initproc));
706 mtx_lock_spin(&sched_lock);
707 initproc->p_sflag |= PS_INMEM;
708 mtx_unlock_spin(&sched_lock);
709 cpu_set_fork_handler(FIRST_THREAD_IN_PROC(initproc), start_init, NULL);
710}
711SYSINIT(init, SI_SUB_CREATE_INIT, SI_ORDER_FIRST, create_init, NULL)
712
713/*
714 * Make it runnable now.
715 */
716static void
717kick_init(const void *udata __unused)
718{
719 struct thread *td;
720
721 td = FIRST_THREAD_IN_PROC(initproc);
722 mtx_lock_spin(&sched_lock);
723 TD_SET_CAN_RUN(td);
724 setrunqueue(td); /* XXXKSE */
725 mtx_unlock_spin(&sched_lock);
726}
727SYSINIT(kickinit, SI_SUB_KTHREAD_INIT, SI_ORDER_FIRST, kick_init, NULL)