xref: /netbsd-src/sys/kern/init_main.c (revision 23c8222edbfb0f0932d88a8351d3a0cf817dfb9e)
1 /*	$NetBSD: init_main.c,v 1.242 2004/10/15 07:22:02 thorpej Exp $	*/
2 
3 /*
4  * Copyright (c) 1982, 1986, 1989, 1991, 1992, 1993
5  *	The Regents of the University of California.  All rights reserved.
6  * (c) UNIX System Laboratories, Inc.
7  * All or some portions of this file are derived from material licensed
8  * to the University of California by American Telephone and Telegraph
9  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
10  * the permission of UNIX System Laboratories, Inc.
11  *
12  * Redistribution and use in source and binary forms, with or without
13  * modification, are permitted provided that the following conditions
14  * are met:
15  * 1. Redistributions of source code must retain the above copyright
16  *    notice, this list of conditions and the following disclaimer.
17  * 2. Redistributions in binary form must reproduce the above copyright
18  *    notice, this list of conditions and the following disclaimer in the
19  *    documentation and/or other materials provided with the distribution.
20  * 3. Neither the name of the University nor the names of its contributors
21  *    may be used to endorse or promote products derived from this software
22  *    without specific prior written permission.
23  *
24  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
25  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
28  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
29  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
30  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
31  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
32  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
33  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
34  * SUCH DAMAGE.
35  *
36  *	@(#)init_main.c	8.16 (Berkeley) 5/14/95
37  */
38 
39 /*
40  * Copyright (c) 1995 Christopher G. Demetriou.  All rights reserved.
41  *
42  * Redistribution and use in source and binary forms, with or without
43  * modification, are permitted provided that the following conditions
44  * are met:
45  * 1. Redistributions of source code must retain the above copyright
46  *    notice, this list of conditions and the following disclaimer.
47  * 2. Redistributions in binary form must reproduce the above copyright
48  *    notice, this list of conditions and the following disclaimer in the
49  *    documentation and/or other materials provided with the distribution.
50  * 3. All advertising materials mentioning features or use of this software
51  *    must display the following acknowledgement:
52  *	This product includes software developed by the University of
53  *	California, Berkeley and its contributors.
54  * 4. Neither the name of the University nor the names of its contributors
55  *    may be used to endorse or promote products derived from this software
56  *    without specific prior written permission.
57  *
58  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
59  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
60  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
61  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
62  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
63  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
64  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
65  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
66  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
67  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
68  * SUCH DAMAGE.
69  *
70  *	@(#)init_main.c	8.16 (Berkeley) 5/14/95
71  */
72 
73 #include <sys/cdefs.h>
74 __KERNEL_RCSID(0, "$NetBSD: init_main.c,v 1.242 2004/10/15 07:22:02 thorpej Exp $");
75 
76 #include "fs_nfs.h"
77 #include "opt_nfsserver.h"
78 #include "opt_ipsec.h"
79 #include "opt_sysv.h"
80 #include "opt_maxuprc.h"
81 #include "opt_multiprocessor.h"
82 #include "opt_pipe.h"
83 #include "opt_syscall_debug.h"
84 #include "opt_systrace.h"
85 #include "opt_posix.h"
86 #include "opt_kcont.h"
87 
88 #include "opencrypto.h"
89 #include "rnd.h"
90 
91 #include <sys/param.h>
92 #include <sys/acct.h>
93 #include <sys/filedesc.h>
94 #include <sys/file.h>
95 #include <sys/errno.h>
96 #include <sys/callout.h>
97 #include <sys/kernel.h>
98 #include <sys/kcont.h>
99 #include <sys/mount.h>
100 #include <sys/proc.h>
101 #include <sys/kthread.h>
102 #include <sys/resourcevar.h>
103 #include <sys/signalvar.h>
104 #include <sys/systm.h>
105 #include <sys/vnode.h>
106 #include <sys/tty.h>
107 #include <sys/conf.h>
108 #include <sys/disklabel.h>
109 #include <sys/buf.h>
110 #include <sys/device.h>
111 #include <sys/exec.h>
112 #include <sys/socketvar.h>
113 #include <sys/protosw.h>
114 #include <sys/reboot.h>
115 #include <sys/user.h>
116 #include <sys/sysctl.h>
117 #include <sys/event.h>
118 #include <sys/mbuf.h>
119 #ifdef FAST_IPSEC
120 #include <netipsec/ipsec.h>
121 #endif
122 #ifdef SYSVSHM
123 #include <sys/shm.h>
124 #endif
125 #ifdef SYSVSEM
126 #include <sys/sem.h>
127 #endif
128 #ifdef SYSVMSG
129 #include <sys/msg.h>
130 #endif
131 #ifdef P1003_1B_SEMAPHORE
132 #include <sys/ksem.h>
133 #endif
134 #ifdef SYSTRACE
135 #include <sys/systrace.h>
136 #endif
137 #include <sys/domain.h>
138 #include <sys/namei.h>
139 #if NOPENCRYPTO > 0
140 #include <opencrypto/cryptodev.h>	/* XXX really the framework */
141 #endif
142 #if NRND > 0
143 #include <sys/rnd.h>
144 #endif
145 #ifndef PIPE_SOCKETPAIR
146 #include <sys/pipe.h>
147 #endif
148 #ifdef LKM
149 #include <sys/lkm.h>
150 #endif
151 
152 #include <sys/syscall.h>
153 #include <sys/sa.h>
154 #include <sys/syscallargs.h>
155 
156 #include <ufs/ufs/quota.h>
157 
158 #include <miscfs/genfs/genfs.h>
159 #include <miscfs/syncfs/syncfs.h>
160 
161 #include <machine/cpu.h>
162 
163 #include <uvm/uvm.h>
164 
165 #include <dev/cons.h>
166 
167 #include <net/if.h>
168 #include <net/raw_cb.h>
169 
170 /* Components of the first process -- never freed. */
171 struct	session session0;
172 struct	pgrp pgrp0;
173 struct	proc proc0;
174 struct	lwp lwp0;
175 struct	pcred cred0;
176 struct	filedesc0 filedesc0;
177 struct	cwdinfo cwdi0;
178 struct	plimit limit0;
179 struct	pstats pstat0;
180 struct	vmspace vmspace0;
181 struct	sigacts sigacts0;
182 #ifndef curlwp
183 struct	lwp *curlwp = &lwp0;
184 #endif
185 struct	proc *initproc;
186 
187 int	nofile = NOFILE;
188 int	maxuprc = MAXUPRC;
189 int	cmask = CMASK;
190 extern	struct user *proc0paddr;
191 
192 struct	vnode *rootvp, *swapdev_vp;
193 int	boothowto;
194 int	cold = 1;			/* still working on startup */
195 struct	timeval boottime;
196 time_t	rootfstime;			/* recorded root fs time, if known */
197 
198 __volatile int start_init_exec;		/* semaphore for start_init() */
199 
200 static void check_console(struct proc *p);
201 static void start_init(void *);
202 void main(void);
203 
204 extern const struct emul emul_netbsd;	/* defined in kern_exec.c */
205 
206 /*
207  * System startup; initialize the world, create process 0, mount root
208  * filesystem, and fork to create init and pagedaemon.  Most of the
209  * hard work is done in the lower-level initialization routines including
210  * startup(), which does memory initialization and autoconfiguration.
211  */
212 void
213 main(void)
214 {
215 	struct lwp *l;
216 	struct proc *p;
217 	struct pdevinit *pdev;
218 	int s, error;
219 	u_int i;
220 	rlim_t lim;
221 	extern struct pdevinit pdevinit[];
222 	extern void schedcpu(void *);
223 #if defined(NFSSERVER) || defined(NFS)
224 	extern void nfs_init(void);
225 #endif
226 #ifdef NVNODE_IMPLICIT
227 	int usevnodes;
228 #endif
229 
230 	/*
231 	 * Initialize the current LWP pointer (curlwp) before
232 	 * any possible traps/probes to simplify trap processing.
233 	 */
234 	l = &lwp0;
235 	curlwp = l;
236 	l->l_cpu = curcpu();
237 	l->l_proc = &proc0;
238 	l->l_lid = 1;
239 
240 	/*
241 	 * Attempt to find console and initialize
242 	 * in case of early panic or other messages.
243 	 */
244 	consinit();
245 	printf("%s", copyright);
246 
247 	KERNEL_LOCK_INIT();
248 
249 	uvm_init();
250 
251 	/* Do machine-dependent initialization. */
252 	cpu_startup();
253 
254 	/* Initialise pools. */
255 	link_pool_init();
256 
257 	/* Initialize callouts. */
258 	callout_startup();
259 
260 	/* Initialize the buffer cache */
261 	bufinit();
262 
263 	/*
264 	 * Initialize mbuf's.  Do this now because we might attempt to
265 	 * allocate mbufs or mbuf clusters during autoconfiguration.
266 	 */
267 	mbinit();
268 
269 	/* Initialize sockets. */
270 	soinit();
271 
272 #ifdef KCONT
273 	/* Initialize kcont. */
274         kcont_init();
275 #endif
276 
277 	/*
278 	 * The following things must be done before autoconfiguration.
279 	 */
280 	evcnt_init();		/* initialize event counters */
281 	tty_init();		/* initialize tty list */
282 #if NRND > 0
283 	rnd_init();		/* initialize RNG */
284 #endif
285 #if NOPENCRYPTO > 0
286 	/* Initialize crypto subsystem before configuring crypto hardware. */
287 	(void)crypto_init();
288 #endif
289 	/* Initialize the sysctl subsystem. */
290 	sysctl_init();
291 
292 	/* Initialize process and pgrp structures. */
293 	procinit();
294 
295 #ifdef LKM
296 	/* Initialize the LKM system. */
297 	lkm_init();
298 #endif
299 
300 	/*
301 	 * Create process 0 (the swapper).
302 	 */
303 	p = &proc0;
304 	proc0_insert(p, l, &pgrp0, &session0);
305 
306 	/*
307 	 * Set P_NOCLDWAIT so that kernel threads are reparented to
308 	 * init(8) when they exit.  init(8) can easily wait them out
309 	 * for us.
310 	 */
311 	p->p_flag = P_SYSTEM | P_NOCLDWAIT;
312 	p->p_stat = SACTIVE;
313 	p->p_nice = NZERO;
314 	p->p_emul = &emul_netbsd;
315 #ifdef __HAVE_SYSCALL_INTERN
316 	(*p->p_emul->e_syscall_intern)(p);
317 #endif
318 	strncpy(p->p_comm, "swapper", MAXCOMLEN);
319 
320 	l->l_flag = L_INMEM;
321 	l->l_stat = LSONPROC;
322 	p->p_nrlwps = 1;
323 
324 	callout_init(&l->l_tsleep_ch);
325 
326 	/* Create credentials. */
327 	cred0.p_refcnt = 1;
328 	p->p_cred = &cred0;
329 	p->p_ucred = crget();
330 	p->p_ucred->cr_ngroups = 1;	/* group 0 */
331 
332 	/* Create the file descriptor table. */
333 	p->p_fd = &filedesc0.fd_fd;
334 	fdinit1(&filedesc0);
335 
336 	/* Create the CWD info. */
337 	p->p_cwdi = &cwdi0;
338 	cwdi0.cwdi_cmask = cmask;
339 	cwdi0.cwdi_refcnt = 1;
340 	simple_lock_init(&cwdi0.cwdi_slock);
341 
342 	/* Create the limits structures. */
343 	p->p_limit = &limit0;
344 	simple_lock_init(&limit0.p_slock);
345 	for (i = 0; i < sizeof(p->p_rlimit)/sizeof(p->p_rlimit[0]); i++)
346 		limit0.pl_rlimit[i].rlim_cur =
347 		    limit0.pl_rlimit[i].rlim_max = RLIM_INFINITY;
348 
349 	limit0.pl_rlimit[RLIMIT_NOFILE].rlim_max = maxfiles;
350 	limit0.pl_rlimit[RLIMIT_NOFILE].rlim_cur =
351 	    maxfiles < nofile ? maxfiles : nofile;
352 
353 	limit0.pl_rlimit[RLIMIT_NPROC].rlim_max = maxproc;
354 	limit0.pl_rlimit[RLIMIT_NPROC].rlim_cur =
355 	    maxproc < maxuprc ? maxproc : maxuprc;
356 
357 	lim = ptoa(uvmexp.free);
358 	limit0.pl_rlimit[RLIMIT_RSS].rlim_max = lim;
359 	limit0.pl_rlimit[RLIMIT_MEMLOCK].rlim_max = lim;
360 	limit0.pl_rlimit[RLIMIT_MEMLOCK].rlim_cur = lim / 3;
361 	limit0.pl_corename = defcorename;
362 	limit0.p_refcnt = 1;
363 
364 	/*
365 	 * Initialize proc0's vmspace, which uses the kernel pmap.
366 	 * All kernel processes (which never have user space mappings)
367 	 * share proc0's vmspace, and thus, the kernel pmap.
368 	 */
369 	uvmspace_init(&vmspace0, pmap_kernel(), round_page(VM_MIN_ADDRESS),
370 	    trunc_page(VM_MAX_ADDRESS));
371 	p->p_vmspace = &vmspace0;
372 
373 	l->l_addr = proc0paddr;				/* XXX */
374 
375 	p->p_stats = &pstat0;
376 
377 	/*
378 	 * Charge root for one process.
379 	 */
380 	(void)chgproccnt(0, 1);
381 
382 	rqinit();
383 
384 	/* Configure virtual memory system, set vm rlimits. */
385 	uvm_init_limits(p);
386 
387 	/* Initialize the file systems. */
388 #if defined(NFSSERVER) || defined(NFS)
389 	nfs_init();			/* initialize server/shared data */
390 #endif
391 #ifdef NVNODE_IMPLICIT
392 	/*
393 	 * If maximum number of vnodes in namei vnode cache is not explicitly
394 	 * defined in kernel config, adjust the number such as we use roughly
395 	 * 1.0% of memory for vnode cache (but not less than NVNODE vnodes).
396 	 */
397 	usevnodes = (ptoa((unsigned)physmem) / 100) / sizeof(struct vnode);
398 	if (usevnodes > desiredvnodes)
399 		desiredvnodes = usevnodes;
400 #endif
401 	vfsinit();
402 
403 	/* Configure the system hardware.  This will enable interrupts. */
404 	configure();
405 
406 	ubc_init();		/* must be after autoconfig */
407 
408 	/* Lock the kernel on behalf of proc0. */
409 	KERNEL_PROC_LOCK(l);
410 
411 #ifdef SYSVSHM
412 	/* Initialize System V style shared memory. */
413 	shminit();
414 #endif
415 
416 #ifdef SYSVSEM
417 	/* Initialize System V style semaphores. */
418 	seminit();
419 #endif
420 
421 #ifdef SYSVMSG
422 	/* Initialize System V style message queues. */
423 	msginit();
424 #endif
425 
426 #ifdef P1003_1B_SEMAPHORE
427 	/* Initialize posix semaphores */
428 	ksem_init();
429 #endif
430 	/* Attach pseudo-devices. */
431 	for (pdev = pdevinit; pdev->pdev_attach != NULL; pdev++)
432 		(*pdev->pdev_attach)(pdev->pdev_count);
433 
434 #ifdef	FAST_IPSEC
435 	/* Attach network crypto subsystem */
436 	ipsec_attach();
437 #endif
438 
439 	/*
440 	 * Initialize protocols.  Block reception of incoming packets
441 	 * until everything is ready.
442 	 */
443 	s = splnet();
444 	ifinit();
445 	domaininit();
446 	if_attachdomain();
447 	splx(s);
448 
449 #ifdef GPROF
450 	/* Initialize kernel profiling. */
451 	kmstartup();
452 #endif
453 
454 	/* Initialize system accouting. */
455 	acct_init();
456 
457 #ifdef SYSTRACE
458 	systrace_init();
459 #endif
460 	/*
461 	 * Initialize signal-related data structures, and signal state
462 	 * for proc0.
463 	 */
464 	signal_init();
465 	p->p_sigacts = &sigacts0;
466 	siginit(p);
467 
468 	/* Kick off timeout driven events by calling first time. */
469 	schedcpu(NULL);
470 
471 	/*
472 	 * Create process 1 (init(8)).  We do this now, as Unix has
473 	 * historically had init be process 1, and changing this would
474 	 * probably upset a lot of people.
475 	 *
476 	 * Note that process 1 won't immediately exec init(8), but will
477 	 * wait for us to inform it that the root file system has been
478 	 * mounted.
479 	 */
480 	if (fork1(l, 0, SIGCHLD, NULL, 0, start_init, NULL, NULL, &initproc))
481 		panic("fork init");
482 
483 	/*
484 	 * Create any kernel threads who's creation was deferred because
485 	 * initproc had not yet been created.
486 	 */
487 	kthread_run_deferred_queue();
488 
489 	/*
490 	 * Now that device driver threads have been created, wait for
491 	 * them to finish any deferred autoconfiguration.  Note we don't
492 	 * need to lock this semaphore, since we haven't booted any
493 	 * secondary processors, yet.
494 	 */
495 	while (config_pending)
496 		(void) tsleep((void *)&config_pending, PWAIT, "cfpend", 0);
497 
498 	/*
499 	 * Finalize configuration now that all real devices have been
500 	 * found.  This needs to be done before the root device is
501 	 * selected, since finalization may create the root device.
502 	 */
503 	config_finalize();
504 
505 	/*
506 	 * Now that autoconfiguration has completed, we can determine
507 	 * the root and dump devices.
508 	 */
509 	cpu_rootconf();
510 	cpu_dumpconf();
511 
512 	/* Mount the root file system. */
513 	do {
514 		domountroothook();
515 		if ((error = vfs_mountroot())) {
516 			printf("cannot mount root, error = %d\n", error);
517 			boothowto |= RB_ASKNAME;
518 			setroot(root_device,
519 			    (rootdev != NODEV) ? DISKPART(rootdev) : 0);
520 		}
521 	} while (error != 0);
522 	mountroothook_destroy();
523 
524 	/*
525 	 * Initialise the time-of-day clock, passing the time recorded
526 	 * in the root filesystem (if any) for use by systems that
527 	 * don't have a non-volatile time-of-day device.
528 	 */
529 	inittodr(rootfstime);
530 
531 	CIRCLEQ_FIRST(&mountlist)->mnt_flag |= MNT_ROOTFS;
532 	CIRCLEQ_FIRST(&mountlist)->mnt_op->vfs_refcount++;
533 
534 	/*
535 	 * Get the vnode for '/'.  Set filedesc0.fd_fd.fd_cdir to
536 	 * reference it.
537 	 */
538 	if (VFS_ROOT(CIRCLEQ_FIRST(&mountlist), &rootvnode))
539 		panic("cannot find root vnode");
540 	cwdi0.cwdi_cdir = rootvnode;
541 	VREF(cwdi0.cwdi_cdir);
542 	VOP_UNLOCK(rootvnode, 0);
543 	cwdi0.cwdi_rdir = NULL;
544 
545 	/*
546 	 * Now that root is mounted, we can fixup initproc's CWD
547 	 * info.  All other processes are kthreads, which merely
548 	 * share proc0's CWD info.
549 	 */
550 	initproc->p_cwdi->cwdi_cdir = rootvnode;
551 	VREF(initproc->p_cwdi->cwdi_cdir);
552 	initproc->p_cwdi->cwdi_rdir = NULL;
553 
554 	/*
555 	 * Now can look at time, having had a chance to verify the time
556 	 * from the file system.  Reset p->p_rtime as it may have been
557 	 * munched in mi_switch() after the time got set.
558 	 */
559 	proclist_lock_read();
560 	s = splsched();
561 	LIST_FOREACH(p, &allproc, p_list) {
562 		KASSERT((p->p_flag & P_MARKER) == 0);
563 		p->p_stats->p_start = mono_time = boottime = time;
564 		LIST_FOREACH(l, &p->p_lwps, l_sibling) {
565 			if (l->l_cpu != NULL)
566 				l->l_cpu->ci_schedstate.spc_runtime = time;
567 		}
568 		p->p_rtime.tv_sec = p->p_rtime.tv_usec = 0;
569 	}
570 	splx(s);
571 	proclist_unlock_read();
572 
573 	/* Create the pageout daemon kernel thread. */
574 	uvm_swap_init();
575 	if (kthread_create1(uvm_pageout, NULL, NULL, "pagedaemon"))
576 		panic("fork pagedaemon");
577 
578 	/* Create the filesystem syncer kernel thread. */
579 	if (kthread_create1(sched_sync, NULL, NULL, "ioflush"))
580 		panic("fork syncer");
581 
582 	/* Create the aiodone daemon kernel thread. */
583 	if (kthread_create1(uvm_aiodone_daemon, NULL, &uvm.aiodoned_proc,
584 	    "aiodoned"))
585 		panic("fork aiodoned");
586 
587 #if defined(MULTIPROCESSOR)
588 	/* Boot the secondary processors. */
589 	cpu_boot_secondary_processors();
590 #endif
591 
592 	/* Initialize exec structures */
593 	exec_init(1);
594 
595 	/*
596 	 * Okay, now we can let init(8) exec!  It's off to userland!
597 	 */
598 	start_init_exec = 1;
599 	wakeup((void *)&start_init_exec);
600 
601 	/* The scheduler is an infinite loop. */
602 	uvm_scheduler();
603 	/* NOTREACHED */
604 }
605 
606 void
607 setrootfstime(time_t t)
608 {
609 	rootfstime = t;
610 }
611 
612 static void
613 check_console(struct proc *p)
614 {
615 	struct nameidata nd;
616 	int error;
617 
618 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_SYSSPACE, "/dev/console", p);
619 	error = namei(&nd);
620 	if (error == 0)
621 		vrele(nd.ni_vp);
622 	else if (error == ENOENT)
623 		printf("warning: no /dev/console\n");
624 	else
625 		printf("warning: lookup /dev/console: error %d\n", error);
626 }
627 
628 /*
629  * List of paths to try when searching for "init".
630  */
631 static const char *initpaths[] = {
632 	"/sbin/init",
633 	"/sbin/oinit",
634 	"/sbin/init.bak",
635 	NULL,
636 };
637 
638 /*
639  * Start the initial user process; try exec'ing each pathname in "initpaths".
640  * The program is invoked with one argument containing the boot flags.
641  */
642 static void
643 start_init(void *arg)
644 {
645 	struct lwp *l = arg;
646 	struct proc *p = l->l_proc;
647 	vaddr_t addr;
648 	struct sys_execve_args /* {
649 		syscallarg(const char *) path;
650 		syscallarg(char * const *) argp;
651 		syscallarg(char * const *) envp;
652 	} */ args;
653 	int options, i, error;
654 	register_t retval[2];
655 	char flags[4], *flagsp;
656 	const char *path, *slash;
657 	char *ucp, **uap, *arg0, *arg1 = NULL;
658 	char ipath[129];
659 	int ipx, len;
660 
661 	/*
662 	 * Now in process 1.
663 	 */
664 	strncpy(p->p_comm, "init", MAXCOMLEN);
665 
666 	/*
667 	 * Wait for main() to tell us that it's safe to exec.
668 	 */
669 	while (start_init_exec == 0)
670 		(void) tsleep((void *)&start_init_exec, PWAIT, "initexec", 0);
671 
672 	/*
673 	 * This is not the right way to do this.  We really should
674 	 * hand-craft a descriptor onto /dev/console to hand to init,
675 	 * but that's a _lot_ more work, and the benefit from this easy
676 	 * hack makes up for the "good is the enemy of the best" effect.
677 	 */
678 	check_console(p);
679 
680 	/*
681 	 * Need just enough stack to hold the faked-up "execve()" arguments.
682 	 */
683 	addr = (vaddr_t)STACK_ALLOC(USRSTACK, PAGE_SIZE);
684 	if (uvm_map(&p->p_vmspace->vm_map, &addr, PAGE_SIZE,
685                     NULL, UVM_UNKNOWN_OFFSET, 0,
686                     UVM_MAPFLAG(UVM_PROT_ALL, UVM_PROT_ALL, UVM_INH_COPY,
687 		    UVM_ADV_NORMAL,
688                     UVM_FLAG_FIXED|UVM_FLAG_OVERLAY|UVM_FLAG_COPYONW)) != 0)
689 		panic("init: couldn't allocate argument space");
690 	p->p_vmspace->vm_maxsaddr = (caddr_t)STACK_MAX(addr, PAGE_SIZE);
691 
692 	ipx = 0;
693 	while (1) {
694 		if (boothowto & RB_ASKNAME) {
695 			printf("init path");
696 			if (initpaths[ipx])
697 				printf(" (default %s)", initpaths[ipx]);
698 			printf(": ");
699 			len = cngetsn(ipath, sizeof(ipath)-1);
700 			if (len == 0) {
701 				if (initpaths[ipx])
702 					path = initpaths[ipx++];
703 				else
704 					continue;
705 			} else {
706 				ipath[len] = '\0';
707 				path = ipath;
708 			}
709 		} else {
710 			if ((path = initpaths[ipx++]) == NULL)
711 				break;
712 		}
713 
714 		ucp = (char *)USRSTACK;
715 
716 		/*
717 		 * Construct the boot flag argument.
718 		 */
719 		flagsp = flags;
720 		*flagsp++ = '-';
721 		options = 0;
722 
723 		if (boothowto & RB_SINGLE) {
724 			*flagsp++ = 's';
725 			options = 1;
726 		}
727 #ifdef notyet
728 		if (boothowto & RB_FASTBOOT) {
729 			*flagsp++ = 'f';
730 			options = 1;
731 		}
732 #endif
733 
734 		/*
735 		 * Move out the flags (arg 1), if necessary.
736 		 */
737 		if (options != 0) {
738 			*flagsp++ = '\0';
739 			i = flagsp - flags;
740 #ifdef DEBUG
741 			printf("init: copying out flags `%s' %d\n", flags, i);
742 #endif
743 			arg1 = STACK_ALLOC(ucp, i);
744 			ucp = STACK_MAX(arg1, i);
745 			(void)copyout((caddr_t)flags, arg1, i);
746 		}
747 
748 		/*
749 		 * Move out the file name (also arg 0).
750 		 */
751 		i = strlen(path) + 1;
752 #ifdef DEBUG
753 		printf("init: copying out path `%s' %d\n", path, i);
754 #else
755 		if (boothowto & RB_ASKNAME || path != initpaths[0])
756 			printf("init: trying %s\n", path);
757 #endif
758 		arg0 = STACK_ALLOC(ucp, i);
759 		ucp = STACK_MAX(arg0, i);
760 		(void)copyout((caddr_t)path, arg0, i);
761 
762 		/*
763 		 * Move out the arg pointers.
764 		 */
765 		ucp = (caddr_t)STACK_ALIGN(ucp, ALIGNBYTES);
766 		uap = (char **)STACK_ALLOC(ucp, sizeof(char *) * 3);
767 		SCARG(&args, path) = arg0;
768 		SCARG(&args, argp) = uap;
769 		SCARG(&args, envp) = NULL;
770 		slash = strrchr(path, '/');
771 		if (slash)
772 			(void)suword((caddr_t)uap++,
773 			    (long)arg0 + (slash + 1 - path));
774 		else
775 			(void)suword((caddr_t)uap++, (long)arg0);
776 		if (options != 0)
777 			(void)suword((caddr_t)uap++, (long)arg1);
778 		(void)suword((caddr_t)uap++, 0);	/* terminator */
779 
780 		/*
781 		 * Now try to exec the program.  If can't for any reason
782 		 * other than it doesn't exist, complain.
783 		 */
784 		error = sys_execve(LIST_FIRST(&p->p_lwps), &args, retval);
785 		if (error == 0 || error == EJUSTRETURN) {
786 			KERNEL_PROC_UNLOCK(l);
787 			return;
788 		}
789 		printf("exec %s: error %d\n", path, error);
790 	}
791 	printf("init: not found\n");
792 	panic("no init");
793 }
794