xref: /netbsd-src/sys/kern/init_main.c (revision 73704c4ce4ee2a60eb617e693ce7e9f03902613e)
1 /*	$NetBSD: init_main.c,v 1.224 2003/08/07 16:31:41 agc 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 1.224 2003/08/07 16:31:41 agc 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 
87 #include "opencrypto.h"
88 #include "rnd.h"
89 
90 #include <sys/param.h>
91 #include <sys/acct.h>
92 #include <sys/filedesc.h>
93 #include <sys/file.h>
94 #include <sys/errno.h>
95 #include <sys/callout.h>
96 #include <sys/kernel.h>
97 #include <sys/mount.h>
98 #include <sys/proc.h>
99 #include <sys/kthread.h>
100 #include <sys/resourcevar.h>
101 #include <sys/signalvar.h>
102 #include <sys/systm.h>
103 #include <sys/vnode.h>
104 #include <sys/tty.h>
105 #include <sys/conf.h>
106 #include <sys/disklabel.h>
107 #include <sys/buf.h>
108 #include <sys/device.h>
109 #include <sys/disk.h>
110 #include <sys/exec.h>
111 #include <sys/socketvar.h>
112 #include <sys/protosw.h>
113 #include <sys/reboot.h>
114 #include <sys/user.h>
115 #include <sys/sysctl.h>
116 #include <sys/event.h>
117 #ifdef	FAST_IPSEC
118 #include <netipsec/ipsec.h>
119 #endif
120 #ifdef SYSVSHM
121 #include <sys/shm.h>
122 #endif
123 #ifdef SYSVSEM
124 #include <sys/sem.h>
125 #endif
126 #ifdef SYSVMSG
127 #include <sys/msg.h>
128 #endif
129 #ifdef P1003_1B_SEMAPHORE
130 #include <sys/ksem.h>
131 #endif
132 #ifdef SYSTRACE
133 #include <sys/systrace.h>
134 #endif
135 #include <sys/domain.h>
136 #include <sys/mbuf.h>
137 #include <sys/namei.h>
138 #if NOPENCRYPTO > 0
139 #include <opencrypto/cryptodev.h>	/* XXX really  the framework */
140 #endif
141 #if NRND > 0
142 #include <sys/rnd.h>
143 #endif
144 #ifndef PIPE_SOCKETPAIR
145 #include <sys/pipe.h>
146 #endif
147 #ifdef LKM
148 #include <sys/lkm.h>
149 #endif
150 
151 #include <sys/syscall.h>
152 #include <sys/sa.h>
153 #include <sys/syscallargs.h>
154 
155 #include <ufs/ufs/quota.h>
156 
157 #include <miscfs/genfs/genfs.h>
158 #include <miscfs/syncfs/syncfs.h>
159 
160 #include <machine/cpu.h>
161 
162 #include <uvm/uvm.h>
163 
164 #include <dev/cons.h>
165 
166 #include <net/if.h>
167 #include <net/raw_cb.h>
168 
169 const char copyright[] =
170 "Copyright (c) 1996, 1997, 1998, 1999, 2000, 2001, 2002, 2003\n"
171 "    The NetBSD Foundation, Inc.  All rights reserved.\n"
172 "Copyright (c) 1982, 1986, 1989, 1991, 1993\n"
173 "    The Regents of the University of California.  All rights reserved.\n"
174 "\n";
175 
176 /* Components of the first process -- never freed. */
177 struct	session session0;
178 struct	pgrp pgrp0;
179 struct	proc proc0;
180 struct	lwp lwp0;
181 struct	pcred cred0;
182 struct	filedesc0 filedesc0;
183 struct	cwdinfo cwdi0;
184 struct	plimit limit0;
185 struct	pstats pstat0;
186 struct	vmspace vmspace0;
187 struct	sigacts sigacts0;
188 #ifndef curlwp
189 struct	lwp *curlwp = &lwp0;
190 #endif
191 struct	proc *initproc;
192 
193 int	nofile = NOFILE;
194 int	maxuprc = MAXUPRC;
195 int	cmask = CMASK;
196 extern	struct user *proc0paddr;
197 
198 struct	vnode *rootvp, *swapdev_vp;
199 int	boothowto;
200 int	cold = 1;			/* still working on startup */
201 struct	timeval boottime;
202 
203 __volatile int start_init_exec;		/* semaphore for start_init() */
204 
205 static void check_console(struct proc *p);
206 static void start_init(void *);
207 void main(void);
208 
209 extern const struct emul emul_netbsd;	/* defined in kern_exec.c */
210 
211 /*
212  * System startup; initialize the world, create process 0, mount root
213  * filesystem, and fork to create init and pagedaemon.  Most of the
214  * hard work is done in the lower-level initialization routines including
215  * startup(), which does memory initialization and autoconfiguration.
216  */
217 void
218 main(void)
219 {
220 	struct lwp *l;
221 	struct proc *p;
222 	struct pdevinit *pdev;
223 	int s, error;
224 	u_int i;
225 	rlim_t lim;
226 	extern struct pdevinit pdevinit[];
227 	extern void schedcpu(void *);
228 #if defined(NFSSERVER) || defined(NFS)
229 	extern void nfs_init(void);
230 #endif
231 #ifdef NVNODE_IMPLICIT
232 	int usevnodes;
233 #endif
234 
235 	/*
236 	 * Initialize the current LWP pointer (curlwp) before
237 	 * any possible traps/probes to simplify trap processing.
238 	 */
239 	simple_lock_init(&proc0.p_raslock);
240 	l = &lwp0;
241 	curlwp = l;
242 	l->l_cpu = curcpu();
243 	l->l_proc = &proc0;
244 	l->l_lid = 1;
245 	/*
246 	 * Attempt to find console and initialize
247 	 * in case of early panic or other messages.
248 	 */
249 	consinit();
250 	printf("%s", copyright);
251 
252 	KERNEL_LOCK_INIT();
253 
254 	uvm_init();
255 
256 	/* Do machine-dependent initialization. */
257 	cpu_startup();
258 
259 	/* Initialize callouts. */
260 	callout_startup();
261 
262 	/*
263 	 * Initialize mbuf's.  Do this now because we might attempt to
264 	 * allocate mbufs or mbuf clusters during autoconfiguration.
265 	 */
266 	mbinit();
267 
268 	/* Initialize kqueues. */
269 	kqueue_init();
270 
271 	/* Initialize sockets. */
272 	soinit();
273 
274 	/*
275 	 * The following things must be done before autoconfiguration.
276 	 */
277 	evcnt_init();		/* initialize event counters */
278 	disk_init();		/* initialize disk list */
279 	tty_init();		/* initialize tty list */
280 #if NRND > 0
281 	rnd_init();		/* initialize RNG */
282 #endif
283 #if NOPENCRYPTO > 0
284 	/* Initialize crypto subsystem before configuring crypto hardware. */
285 	(void)crypto_init();
286 #endif
287 	/* Initialize the sysctl subsystem. */
288 	sysctl_init();
289 
290 	/*
291 	 * Initialize process and pgrp structures.
292 	 */
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 	finit();
334 	p->p_fd = &filedesc0.fd_fd;
335 	fdinit1(&filedesc0);
336 
337 	/* Create the CWD info. */
338 	p->p_cwdi = &cwdi0;
339 	cwdi0.cwdi_cmask = cmask;
340 	cwdi0.cwdi_refcnt = 1;
341 
342 	/* Create the limits structures. */
343 	p->p_limit = &limit0;
344 	for (i = 0; i < sizeof(p->p_rlimit)/sizeof(p->p_rlimit[0]); i++)
345 		limit0.pl_rlimit[i].rlim_cur =
346 		    limit0.pl_rlimit[i].rlim_max = RLIM_INFINITY;
347 
348 	limit0.pl_rlimit[RLIMIT_NOFILE].rlim_max = maxfiles;
349 	limit0.pl_rlimit[RLIMIT_NOFILE].rlim_cur =
350 	    maxfiles < nofile ? maxfiles : nofile;
351 
352 	limit0.pl_rlimit[RLIMIT_NPROC].rlim_max = maxproc;
353 	limit0.pl_rlimit[RLIMIT_NPROC].rlim_cur =
354 	    maxproc < maxuprc ? maxproc : maxuprc;
355 
356 	lim = ptoa(uvmexp.free);
357 	limit0.pl_rlimit[RLIMIT_RSS].rlim_max = lim;
358 	limit0.pl_rlimit[RLIMIT_MEMLOCK].rlim_max = lim;
359 	limit0.pl_rlimit[RLIMIT_MEMLOCK].rlim_cur = lim / 3;
360 	limit0.pl_corename = defcorename;
361 	limit0.p_refcnt = 1;
362 
363 	/*
364 	 * Initialize proc0's vmspace, which uses the kernel pmap.
365 	 * All kernel processes (which never have user space mappings)
366 	 * share proc0's vmspace, and thus, the kernel pmap.
367 	 */
368 	uvmspace_init(&vmspace0, pmap_kernel(), round_page(VM_MIN_ADDRESS),
369 	    trunc_page(VM_MAX_ADDRESS));
370 	p->p_vmspace = &vmspace0;
371 
372 	l->l_addr = proc0paddr;				/* XXX */
373 
374 	p->p_stats = &pstat0;
375 
376 	/*
377 	 * Charge root for one process.
378 	 */
379 	(void)chgproccnt(0, 1);
380 
381 	rqinit();
382 
383 	/* Configure virtual memory system, set vm rlimits. */
384 	uvm_init_limits(p);
385 
386 	/* Initialize the file systems. */
387 #if defined(NFSSERVER) || defined(NFS)
388 	nfs_init();			/* initialize server/shared data */
389 #endif
390 #ifdef NVNODE_IMPLICIT
391 	/*
392 	 * If maximum number of vnodes in namei vnode cache is not explicitly
393 	 * defined in kernel config, adjust the number such as we use roughly
394 	 * 0.5% of memory for vnode cache (but not less than NVNODE vnodes).
395 	 */
396 	usevnodes = (ptoa((unsigned)physmem) / 200) / sizeof(struct vnode);
397 	if (usevnodes > desiredvnodes)
398 		desiredvnodes = usevnodes;
399 #endif
400 	vfsinit();
401 
402 	/* Configure the system hardware.  This will enable interrupts. */
403 	configure();
404 
405 	ubc_init();		/* must be after autoconfig */
406 
407 	/* Lock the kernel on behalf of proc0. */
408 	KERNEL_PROC_LOCK(l);
409 
410 #ifdef SYSVSHM
411 	/* Initialize System V style shared memory. */
412 	shminit();
413 #endif
414 
415 #ifdef SYSVSEM
416 	/* Initialize System V style semaphores. */
417 	seminit();
418 #endif
419 
420 #ifdef SYSVMSG
421 	/* Initialize System V style message queues. */
422 	msginit();
423 #endif
424 
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 	CIRCLEQ_FIRST(&mountlist)->mnt_flag |= MNT_ROOTFS;
525 	CIRCLEQ_FIRST(&mountlist)->mnt_op->vfs_refcount++;
526 
527 	/*
528 	 * Get the vnode for '/'.  Set filedesc0.fd_fd.fd_cdir to
529 	 * reference it.
530 	 */
531 	if (VFS_ROOT(CIRCLEQ_FIRST(&mountlist), &rootvnode))
532 		panic("cannot find root vnode");
533 	cwdi0.cwdi_cdir = rootvnode;
534 	VREF(cwdi0.cwdi_cdir);
535 	VOP_UNLOCK(rootvnode, 0);
536 	cwdi0.cwdi_rdir = NULL;
537 
538 	/*
539 	 * Now that root is mounted, we can fixup initproc's CWD
540 	 * info.  All other processes are kthreads, which merely
541 	 * share proc0's CWD info.
542 	 */
543 	initproc->p_cwdi->cwdi_cdir = rootvnode;
544 	VREF(initproc->p_cwdi->cwdi_cdir);
545 	initproc->p_cwdi->cwdi_rdir = NULL;
546 
547 	/*
548 	 * Now can look at time, having had a chance to verify the time
549 	 * from the file system.  Reset p->p_rtime as it may have been
550 	 * munched in mi_switch() after the time got set.
551 	 */
552 	proclist_lock_read();
553 	s = splsched();
554 	for (p = LIST_FIRST(&allproc); p != NULL;
555 	     p = LIST_NEXT(p, p_list)) {
556 		p->p_stats->p_start = mono_time = boottime = time;
557 		for (l = LIST_FIRST(&p->p_lwps); l != NULL;
558 		     l = LIST_NEXT(l, l_sibling))
559 			if (l->l_cpu != NULL)
560 				l->l_cpu->ci_schedstate.spc_runtime = time;
561 		p->p_rtime.tv_sec = p->p_rtime.tv_usec = 0;
562 	}
563 	splx(s);
564 	proclist_unlock_read();
565 
566 	/* Create the pageout daemon kernel thread. */
567 	uvm_swap_init();
568 	if (kthread_create1(uvm_pageout, NULL, NULL, "pagedaemon"))
569 		panic("fork pagedaemon");
570 
571 	/* Create the process reaper kernel thread. */
572 	if (kthread_create1(reaper, NULL, NULL, "reaper"))
573 		panic("fork reaper");
574 
575 	/* Create the filesystem syncer kernel thread. */
576 	if (kthread_create1(sched_sync, NULL, NULL, "ioflush"))
577 		panic("fork syncer");
578 
579 	/* Create the aiodone daemon kernel thread. */
580 	if (kthread_create1(uvm_aiodone_daemon, NULL, &uvm.aiodoned_proc,
581 	    "aiodoned"))
582 		panic("fork aiodoned");
583 
584 #if defined(MULTIPROCESSOR)
585 	/* Boot the secondary processors. */
586 	cpu_boot_secondary_processors();
587 #endif
588 
589 	/* Initialize exec structures */
590 	exec_init(1);
591 
592 #ifndef PIPE_SOCKETPAIR
593 	/* Initialize pipe structures */
594 	pipe_init();
595 #endif
596 
597 	/*
598 	 * Okay, now we can let init(8) exec!  It's off to userland!
599 	 */
600 	start_init_exec = 1;
601 	wakeup((void *)&start_init_exec);
602 
603 	/* The scheduler is an infinite loop. */
604 	uvm_scheduler();
605 	/* NOTREACHED */
606 }
607 
608 static void
609 check_console(struct proc *p)
610 {
611 	struct nameidata nd;
612 	int error;
613 
614 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_SYSSPACE, "/dev/console", p);
615 	error = namei(&nd);
616 	if (error == 0)
617 		vrele(nd.ni_vp);
618 	else if (error == ENOENT)
619 		printf("warning: no /dev/console\n");
620 	else
621 		printf("warning: lookup /dev/console: error %d\n", error);
622 }
623 
624 /*
625  * List of paths to try when searching for "init".
626  */
627 static const char *initpaths[] = {
628 	"/sbin/init",
629 	"/sbin/oinit",
630 	"/sbin/init.bak",
631 	NULL,
632 };
633 
634 /*
635  * Start the initial user process; try exec'ing each pathname in "initpaths".
636  * The program is invoked with one argument containing the boot flags.
637  */
638 static void
639 start_init(void *arg)
640 {
641 	struct lwp *l = arg;
642 	struct proc *p = l->l_proc;
643 	vaddr_t addr;
644 	struct sys_execve_args /* {
645 		syscallarg(const char *) path;
646 		syscallarg(char * const *) argp;
647 		syscallarg(char * const *) envp;
648 	} */ args;
649 	int options, i, error;
650 	register_t retval[2];
651 	char flags[4], *flagsp;
652 	const char *path, *slash;
653 	char *ucp, **uap, *arg0, *arg1 = NULL;
654 	char ipath[129];
655 	int ipx, len;
656 
657 	/*
658 	 * Now in process 1.
659 	 */
660 	strncpy(p->p_comm, "init", MAXCOMLEN);
661 
662 	/*
663 	 * Wait for main() to tell us that it's safe to exec.
664 	 */
665 	while (start_init_exec == 0)
666 		(void) tsleep((void *)&start_init_exec, PWAIT, "initexec", 0);
667 
668 	/*
669 	 * This is not the right way to do this.  We really should
670 	 * hand-craft a descriptor onto /dev/console to hand to init,
671 	 * but that's a _lot_ more work, and the benefit from this easy
672 	 * hack makes up for the "good is the enemy of the best" effect.
673 	 */
674 	check_console(p);
675 
676 	/*
677 	 * Need just enough stack to hold the faked-up "execve()" arguments.
678 	 */
679 	addr = (vaddr_t)STACK_ALLOC(USRSTACK, PAGE_SIZE);
680 	if (uvm_map(&p->p_vmspace->vm_map, &addr, PAGE_SIZE,
681                     NULL, UVM_UNKNOWN_OFFSET, 0,
682                     UVM_MAPFLAG(UVM_PROT_ALL, UVM_PROT_ALL, UVM_INH_COPY,
683 		    UVM_ADV_NORMAL,
684                     UVM_FLAG_FIXED|UVM_FLAG_OVERLAY|UVM_FLAG_COPYONW)) != 0)
685 		panic("init: couldn't allocate argument space");
686 	p->p_vmspace->vm_maxsaddr = (caddr_t)STACK_MAX(addr, PAGE_SIZE);
687 
688 	ipx = 0;
689 	while (1) {
690 		if (boothowto & RB_ASKNAME) {
691 			printf("init path");
692 			if (initpaths[ipx])
693 				printf(" (default %s)", initpaths[ipx]);
694 			printf(": ");
695 			len = cngetsn(ipath, sizeof(ipath)-1);
696 			if (len == 0) {
697 				if (initpaths[ipx])
698 					path = initpaths[ipx++];
699 				else
700 					continue;
701 			} else {
702 				ipath[len] = '\0';
703 				path = ipath;
704 			}
705 		} else {
706 			if ((path = initpaths[ipx++]) == NULL)
707 				break;
708 		}
709 
710 		ucp = (char *)USRSTACK;
711 
712 		/*
713 		 * Construct the boot flag argument.
714 		 */
715 		flagsp = flags;
716 		*flagsp++ = '-';
717 		options = 0;
718 
719 		if (boothowto & RB_SINGLE) {
720 			*flagsp++ = 's';
721 			options = 1;
722 		}
723 #ifdef notyet
724 		if (boothowto & RB_FASTBOOT) {
725 			*flagsp++ = 'f';
726 			options = 1;
727 		}
728 #endif
729 
730 		/*
731 		 * Move out the flags (arg 1), if necessary.
732 		 */
733 		if (options != 0) {
734 			*flagsp++ = '\0';
735 			i = flagsp - flags;
736 #ifdef DEBUG
737 			printf("init: copying out flags `%s' %d\n", flags, i);
738 #endif
739 			arg1 = STACK_ALLOC(ucp, i);
740 			ucp = STACK_MAX(arg1, i);
741 			(void)copyout((caddr_t)flags, arg1, i);
742 		}
743 
744 		/*
745 		 * Move out the file name (also arg 0).
746 		 */
747 		i = strlen(path) + 1;
748 #ifdef DEBUG
749 		printf("init: copying out path `%s' %d\n", path, i);
750 #else
751 		if (boothowto & RB_ASKNAME || path != initpaths[0])
752 			printf("init: trying %s\n", path);
753 #endif
754 		arg0 = STACK_ALLOC(ucp, i);
755 		ucp = STACK_MAX(arg0, i);
756 		(void)copyout((caddr_t)path, arg0, i);
757 
758 		/*
759 		 * Move out the arg pointers.
760 		 */
761 		ucp = (caddr_t)STACK_ALIGN(ucp, ALIGNBYTES);
762 		uap = (char **)STACK_ALLOC(ucp, sizeof(char *) * 3);
763 		SCARG(&args, path) = arg0;
764 		SCARG(&args, argp) = uap;
765 		SCARG(&args, envp) = NULL;
766 		slash = strrchr(path, '/');
767 		if (slash)
768 			(void)suword((caddr_t)uap++,
769 			    (long)arg0 + (slash + 1 - path));
770 		else
771 			(void)suword((caddr_t)uap++, (long)arg0);
772 		if (options != 0)
773 			(void)suword((caddr_t)uap++, (long)arg1);
774 		(void)suword((caddr_t)uap++, 0);	/* terminator */
775 
776 		/*
777 		 * Now try to exec the program.  If can't for any reason
778 		 * other than it doesn't exist, complain.
779 		 */
780 		error = sys_execve(LIST_FIRST(&p->p_lwps), &args, retval);
781 		if (error == 0 || error == EJUSTRETURN) {
782 			KERNEL_PROC_UNLOCK(l);
783 			return;
784 		}
785 		printf("exec %s: error %d\n", path, error);
786 	}
787 	printf("init: not found\n");
788 	panic("no init");
789 }
790