Home | History | Annotate | Line # | Download | only in kern
init_main.c revision 1.304
      1 /*	$NetBSD: init_main.c,v 1.304 2007/06/17 13:34:43 yamt 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 1.304 2007/06/17 13:34:43 yamt Exp $");
     75 
     76 #include "opt_ipsec.h"
     77 #include "opt_kcont.h"
     78 #include "opt_multiprocessor.h"
     79 #include "opt_ntp.h"
     80 #include "opt_pipe.h"
     81 #include "opt_posix.h"
     82 #include "opt_syscall_debug.h"
     83 #include "opt_sysv.h"
     84 #include "opt_systrace.h"
     85 #include "opt_fileassoc.h"
     86 #include "opt_ktrace.h"
     87 #include "opt_pax.h"
     88 
     89 #include "rnd.h"
     90 #include "veriexec.h"
     91 
     92 #include <sys/param.h>
     93 #include <sys/acct.h>
     94 #include <sys/filedesc.h>
     95 #include <sys/file.h>
     96 #include <sys/errno.h>
     97 #include <sys/callout.h>
     98 #include <sys/cpu.h>
     99 #include <sys/kernel.h>
    100 #include <sys/kcont.h>
    101 #include <sys/kmem.h>
    102 #include <sys/mount.h>
    103 #include <sys/proc.h>
    104 #include <sys/kthread.h>
    105 #include <sys/resourcevar.h>
    106 #include <sys/signalvar.h>
    107 #include <sys/systm.h>
    108 #include <sys/vnode.h>
    109 #include <sys/fstrans.h>
    110 #include <sys/tty.h>
    111 #include <sys/conf.h>
    112 #include <sys/disklabel.h>
    113 #include <sys/buf.h>
    114 #include <sys/device.h>
    115 #include <sys/exec.h>
    116 #include <sys/socketvar.h>
    117 #include <sys/protosw.h>
    118 #include <sys/reboot.h>
    119 #include <sys/user.h>
    120 #include <sys/sysctl.h>
    121 #include <sys/event.h>
    122 #include <sys/mbuf.h>
    123 #include <sys/sched.h>
    124 #include <sys/sleepq.h>
    125 #include <sys/iostat.h>
    126 #include <sys/vmem.h>
    127 #ifdef FAST_IPSEC
    128 #include <netipsec/ipsec.h>
    129 #endif
    130 #ifdef SYSVSHM
    131 #include <sys/shm.h>
    132 #endif
    133 #ifdef SYSVSEM
    134 #include <sys/sem.h>
    135 #endif
    136 #ifdef SYSVMSG
    137 #include <sys/msg.h>
    138 #endif
    139 #ifdef SYSTRACE
    140 #include <sys/systrace.h>
    141 #endif
    142 #ifdef P1003_1B_SEMAPHORE
    143 #include <sys/ksem.h>
    144 #endif
    145 #include <sys/domain.h>
    146 #include <sys/namei.h>
    147 #if NRND > 0
    148 #include <sys/rnd.h>
    149 #endif
    150 #include <sys/pipe.h>
    151 #ifdef LKM
    152 #include <sys/lkm.h>
    153 #endif
    154 #if NVERIEXEC > 0
    155 #include <sys/verified_exec.h>
    156 #endif /* NVERIEXEC > 0 */
    157 #ifdef KTRACE
    158 #include <sys/ktrace.h>
    159 #endif
    160 #include <sys/debug.h>
    161 #include <sys/kauth.h>
    162 #include <net80211/ieee80211_netbsd.h>
    163 
    164 #include <sys/syscall.h>
    165 #include <sys/syscallargs.h>
    166 
    167 #if defined(PAX_MPROTECT) || defined(PAX_SEGVGUARD)
    168 #include <sys/pax.h>
    169 #endif /* PAX_MPROTECT || PAX_SEGVGUARD */
    170 #include <ufs/ufs/quota.h>
    171 
    172 #include <miscfs/genfs/genfs.h>
    173 #include <miscfs/syncfs/syncfs.h>
    174 
    175 #include <machine/cpu.h>
    176 
    177 #include <uvm/uvm.h>
    178 
    179 #include <dev/cons.h>
    180 
    181 #include <net/if.h>
    182 #include <net/raw_cb.h>
    183 
    184 #include <secmodel/secmodel.h>
    185 
    186 extern struct proc proc0;
    187 extern struct lwp lwp0;
    188 extern struct cwdinfo cwdi0;
    189 
    190 #ifndef curlwp
    191 struct	lwp *curlwp = &lwp0;
    192 #endif
    193 struct	proc *initproc;
    194 
    195 struct	vnode *rootvp, *swapdev_vp;
    196 int	boothowto;
    197 int	cold = 1;			/* still working on startup */
    198 struct timeval boottime;	        /* time at system startup - will only follow settime deltas */
    199 time_t	rootfstime;			/* recorded root fs time, if known */
    200 int	ncpu = 0;			/* number of CPUs configured, assume 1 */
    201 
    202 volatile int start_init_exec;		/* semaphore for start_init() */
    203 
    204 static void check_console(struct lwp *l);
    205 static void start_init(void *);
    206 void main(void);
    207 
    208 #if defined(__SSP__) || defined(__SSP_ALL__)
    209 long __stack_chk_guard[8] = {0, 0, 0, 0, 0, 0, 0, 0};
    210 void __stack_chk_fail(void);
    211 
    212 void
    213 __stack_chk_fail(void)
    214 {
    215 	panic("stack overflow detected; terminated");
    216 }
    217 #endif
    218 
    219 void __secmodel_none(void);
    220 __weak_alias(secmodel_start,__secmodel_none);
    221 void
    222 __secmodel_none(void)
    223 {
    224 	return;
    225 }
    226 
    227 /*
    228  * System startup; initialize the world, create process 0, mount root
    229  * filesystem, and fork to create init and pagedaemon.  Most of the
    230  * hard work is done in the lower-level initialization routines including
    231  * startup(), which does memory initialization and autoconfiguration.
    232  */
    233 void
    234 main(void)
    235 {
    236 #ifdef __HAVE_TIMECOUNTER
    237 	struct timeval time;
    238 #endif
    239 	struct lwp *l;
    240 	struct proc *p;
    241 	struct pdevinit *pdev;
    242 	int s, error;
    243 	extern struct pdevinit pdevinit[];
    244 #ifdef NVNODE_IMPLICIT
    245 	int usevnodes;
    246 #endif
    247 
    248 	/*
    249 	 * Initialize the current LWP pointer (curlwp) before
    250 	 * any possible traps/probes to simplify trap processing.
    251 	 */
    252 	l = &lwp0;
    253 	curlwp = l;
    254 	l->l_cpu = curcpu();
    255 	l->l_proc = &proc0;
    256 	l->l_lid = 1;
    257 
    258 	/*
    259 	 * Attempt to find console and initialize
    260 	 * in case of early panic or other messages.
    261 	 */
    262 	consinit();
    263 
    264 	KERNEL_LOCK_INIT();
    265 
    266 	uvm_init();
    267 
    268 #ifdef DEBUG
    269 	debug_init();
    270 #endif
    271 
    272 	kmem_init();
    273 
    274 	/* Do machine-dependent initialization. */
    275 	cpu_startup();
    276 
    277 	/* Initialize callouts. */
    278 	callout_startup();
    279 
    280 	/*
    281 	 * Initialize the kernel authorization subsystem and start the
    282 	 * default security model, if any. We need to do this early
    283 	 * enough so that subsystems relying on any of the aforementioned
    284 	 * can work properly. Since the security model may dictate the
    285 	 * credential inheritance policy, it is needed at least before
    286 	 * any process is created, specifically proc0.
    287 	 */
    288 	kauth_init();
    289 	secmodel_start();
    290 
    291 	/* Initialize the buffer cache */
    292 	bufinit();
    293 
    294 	/*
    295 	 * Initialize mbuf's.  Do this now because we might attempt to
    296 	 * allocate mbufs or mbuf clusters during autoconfiguration.
    297 	 */
    298 	mbinit();
    299 
    300 	/* Initialize sockets. */
    301 	soinit();
    302 
    303 #ifdef KCONT
    304 	/* Initialize kcont. */
    305         kcont_init();
    306 #endif
    307 
    308 	/*
    309 	 * The following things must be done before autoconfiguration.
    310 	 */
    311 	evcnt_init();		/* initialize event counters */
    312 #if NRND > 0
    313 	rnd_init();		/* initialize RNG */
    314 #endif
    315 
    316 	/* Initialize process and pgrp structures. */
    317 	procinit();
    318 	lwpinit();
    319 
    320 	/* Initialize signal-related data structures. */
    321 	signal_init();
    322 
    323 	/* Create process 0 (the swapper). */
    324 	proc0_init();
    325 
    326 	/*
    327 	 * Charge root for one process.
    328 	 */
    329 	(void)chgproccnt(0, 1);
    330 
    331 	/* Initialize the run queues, turnstiles and sleep queues. */
    332 	sched_rqinit();
    333 	turnstile_init();
    334 	sleeptab_init(&sleeptab);
    335 
    336 	/* MI initialization of the boot cpu */
    337 	error = mi_cpu_attach(curcpu());
    338 	KASSERT(error == 0);
    339 
    340 	/* Initialize the sysctl subsystem. */
    341 	sysctl_init();
    342 
    343 	/* Initialize I/O statistics. */
    344 	iostat_init();
    345 
    346 	/* Initialize the file systems. */
    347 #ifdef NVNODE_IMPLICIT
    348 	/*
    349 	 * If maximum number of vnodes in namei vnode cache is not explicitly
    350 	 * defined in kernel config, adjust the number such as we use roughly
    351 	 * 1.0% of memory for vnode cache (but not less than NVNODE vnodes).
    352 	 */
    353 	usevnodes = (ptoa((unsigned)physmem) / 100) / sizeof(struct vnode);
    354 	if (usevnodes > desiredvnodes)
    355 		desiredvnodes = usevnodes;
    356 #endif
    357 	vfsinit();
    358 
    359 	/* Initialize fstrans. */
    360 	fstrans_init();
    361 
    362 	/* Initialize asynchronous I/O. */
    363 	aio_sysinit();
    364 
    365 #ifdef __HAVE_TIMECOUNTER
    366 	inittimecounter();
    367 	ntp_init();
    368 #endif /* __HAVE_TIMECOUNTER */
    369 
    370 	/* Configure the system hardware.  This will enable interrupts. */
    371 	configure();
    372 
    373 #if defined(__SSP__) || defined(__SSP_ALL__)
    374 	{
    375 #ifdef DIAGNOSTIC
    376 		printf("Initializing SSP:");
    377 #endif
    378 		/*
    379 		 * We initialize ssp here carefully:
    380 		 *	1. after we got some entropy
    381 		 *	2. without calling a function
    382 		 */
    383 		size_t i;
    384 		long guard[__arraycount(__stack_chk_guard)];
    385 
    386 		arc4randbytes(guard, sizeof(guard));
    387 		for (i = 0; i < __arraycount(guard); i++)
    388 			__stack_chk_guard[i] = guard[i];
    389 #ifdef DIAGNOSTIC
    390 		for (i = 0; i < __arraycount(guard); i++)
    391 			printf("%lx ", guard[i]);
    392 		printf("\n");
    393 #endif
    394 	}
    395 #endif
    396 	ubc_init();		/* must be after autoconfig */
    397 
    398 	/* Lock the kernel on behalf of proc0. */
    399 	KERNEL_LOCK(1, l);
    400 
    401 #ifdef SYSTRACE
    402 	systrace_init();
    403 #endif
    404 
    405 #ifdef SYSVSHM
    406 	/* Initialize System V style shared memory. */
    407 	shminit();
    408 #endif
    409 
    410 #ifdef SYSVSEM
    411 	/* Initialize System V style semaphores. */
    412 	seminit();
    413 #endif
    414 
    415 #ifdef SYSVMSG
    416 	/* Initialize System V style message queues. */
    417 	msginit();
    418 #endif
    419 
    420 #ifdef P1003_1B_SEMAPHORE
    421 	/* Initialize posix semaphores */
    422 	ksem_init();
    423 #endif
    424 
    425 #if NVERIEXEC > 0
    426 	/*
    427 	 * Initialise the Veriexec subsystem.
    428 	 */
    429 	veriexec_init();
    430 #endif /* NVERIEXEC > 0 */
    431 
    432 #if defined(PAX_MPROTECT) || defined(PAX_SEGVGUARD)
    433 	pax_init();
    434 #endif /* PAX_MPROTECT || PAX_SEGVGUARD */
    435 
    436 	/* Attach pseudo-devices. */
    437 	for (pdev = pdevinit; pdev->pdev_attach != NULL; pdev++)
    438 		(*pdev->pdev_attach)(pdev->pdev_count);
    439 
    440 #ifdef	FAST_IPSEC
    441 	/* Attach network crypto subsystem */
    442 	ipsec_attach();
    443 #endif
    444 
    445 	/*
    446 	 * Initialize protocols.  Block reception of incoming packets
    447 	 * until everything is ready.
    448 	 */
    449 	s = splnet();
    450 	ifinit();
    451 	domaininit();
    452 	if_attachdomain();
    453 	splx(s);
    454 
    455 #ifdef GPROF
    456 	/* Initialize kernel profiling. */
    457 	kmstartup();
    458 #endif
    459 
    460 	/* Initialize system accouting. */
    461 	acct_init();
    462 
    463 #ifndef PIPE_SOCKETPAIR
    464 	/* Initialize pipes. */
    465 	pipe_init();
    466 #endif
    467 
    468 	/* Setup the scheduler */
    469 	sched_setup();
    470 
    471 #ifdef KTRACE
    472 	/* Initialize ktrace. */
    473 	ktrinit();
    474 #endif
    475 
    476 	/*
    477 	 * Create process 1 (init(8)).  We do this now, as Unix has
    478 	 * historically had init be process 1, and changing this would
    479 	 * probably upset a lot of people.
    480 	 *
    481 	 * Note that process 1 won't immediately exec init(8), but will
    482 	 * wait for us to inform it that the root file system has been
    483 	 * mounted.
    484 	 */
    485 	if (fork1(l, 0, SIGCHLD, NULL, 0, start_init, NULL, NULL, &initproc))
    486 		panic("fork init");
    487 
    488 	/*
    489 	 * Create any kernel threads who's creation was deferred because
    490 	 * initproc had not yet been created.
    491 	 */
    492 	kthread_run_deferred_queue();
    493 
    494 	/*
    495 	 * Now that device driver threads have been created, wait for
    496 	 * them to finish any deferred autoconfiguration.  Note we don't
    497 	 * need to lock this semaphore, since we haven't booted any
    498 	 * secondary processors, yet.
    499 	 */
    500 	while (config_pending)
    501 		(void) tsleep(&config_pending, PWAIT, "cfpend", 0);
    502 
    503 	/*
    504 	 * Finalize configuration now that all real devices have been
    505 	 * found.  This needs to be done before the root device is
    506 	 * selected, since finalization may create the root device.
    507 	 */
    508 	config_finalize();
    509 
    510 	/*
    511 	 * Now that autoconfiguration has completed, we can determine
    512 	 * the root and dump devices.
    513 	 */
    514 	cpu_rootconf();
    515 	cpu_dumpconf();
    516 
    517 	/* Mount the root file system. */
    518 	do {
    519 		domountroothook();
    520 		if ((error = vfs_mountroot())) {
    521 			printf("cannot mount root, error = %d\n", error);
    522 			boothowto |= RB_ASKNAME;
    523 			setroot(root_device,
    524 			    (rootdev != NODEV) ? DISKPART(rootdev) : 0);
    525 		}
    526 	} while (error != 0);
    527 	mountroothook_destroy();
    528 
    529 	/*
    530 	 * Initialise the time-of-day clock, passing the time recorded
    531 	 * in the root filesystem (if any) for use by systems that
    532 	 * don't have a non-volatile time-of-day device.
    533 	 */
    534 	inittodr(rootfstime);
    535 
    536 	CIRCLEQ_FIRST(&mountlist)->mnt_flag |= MNT_ROOTFS;
    537 	CIRCLEQ_FIRST(&mountlist)->mnt_op->vfs_refcount++;
    538 
    539 	/*
    540 	 * Get the vnode for '/'.  Set filedesc0.fd_fd.fd_cdir to
    541 	 * reference it.
    542 	 */
    543 	error = VFS_ROOT(CIRCLEQ_FIRST(&mountlist), &rootvnode);
    544 	if (error)
    545 		panic("cannot find root vnode, error=%d", error);
    546 	cwdi0.cwdi_cdir = rootvnode;
    547 	VREF(cwdi0.cwdi_cdir);
    548 	VOP_UNLOCK(rootvnode, 0);
    549 	cwdi0.cwdi_rdir = NULL;
    550 
    551 	/*
    552 	 * Now that root is mounted, we can fixup initproc's CWD
    553 	 * info.  All other processes are kthreads, which merely
    554 	 * share proc0's CWD info.
    555 	 */
    556 	initproc->p_cwdi->cwdi_cdir = rootvnode;
    557 	VREF(initproc->p_cwdi->cwdi_cdir);
    558 	initproc->p_cwdi->cwdi_rdir = NULL;
    559 
    560 	/*
    561 	 * Now can look at time, having had a chance to verify the time
    562 	 * from the file system.  Reset l->l_rtime as it may have been
    563 	 * munched in mi_switch() after the time got set.
    564 	 */
    565 #ifdef __HAVE_TIMECOUNTER
    566 	getmicrotime(&time);
    567 #else
    568 	mono_time = time;
    569 #endif
    570 	boottime = time;
    571 	mutex_enter(&proclist_lock);
    572 	LIST_FOREACH(p, &allproc, p_list) {
    573 		KASSERT((p->p_flag & PK_MARKER) == 0);
    574 		mutex_enter(&p->p_smutex);
    575 		p->p_stats->p_start = time;
    576 		LIST_FOREACH(l, &p->p_lwps, l_sibling) {
    577 			lwp_lock(l);
    578 			l->l_cpu->ci_schedstate.spc_runtime = time;
    579 			l->l_rtime.tv_sec = l->l_rtime.tv_usec = 0;
    580 			lwp_unlock(l);
    581 		}
    582 		mutex_exit(&p->p_smutex);
    583 	}
    584 	mutex_exit(&proclist_lock);
    585 
    586 	/* Create the pageout daemon kernel thread. */
    587 	uvm_swap_init();
    588 	if (kthread_create1(uvm_pageout, NULL, NULL, "pagedaemon"))
    589 		panic("fork pagedaemon");
    590 
    591 	/* Create the filesystem syncer kernel thread. */
    592 	if (kthread_create1(sched_sync, NULL, NULL, "ioflush"))
    593 		panic("fork syncer");
    594 
    595 	/* Create the aiodone daemon kernel thread. */
    596 	if (workqueue_create(&uvm.aiodone_queue, "aiodoned",
    597 	    uvm_aiodone_worker, NULL, PVM, IPL_BIO, 0))
    598 		panic("fork aiodoned");
    599 
    600 	vmem_rehash_start();
    601 
    602 #if defined(MULTIPROCESSOR)
    603 	/* Boot the secondary processors. */
    604 	cpu_boot_secondary_processors();
    605 #endif
    606 
    607 	/* Initialize exec structures */
    608 	exec_init(1);
    609 
    610 	/*
    611 	 * Okay, now we can let init(8) exec!  It's off to userland!
    612 	 */
    613 	start_init_exec = 1;
    614 	wakeup(&start_init_exec);
    615 
    616 	/* The scheduler is an infinite loop. */
    617 	uvm_scheduler();
    618 	/* NOTREACHED */
    619 }
    620 
    621 void
    622 setrootfstime(time_t t)
    623 {
    624 	rootfstime = t;
    625 }
    626 
    627 static void
    628 check_console(struct lwp *l)
    629 {
    630 	struct nameidata nd;
    631 	int error;
    632 
    633 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_SYSSPACE, "/dev/console", l);
    634 	error = namei(&nd);
    635 	if (error == 0)
    636 		vrele(nd.ni_vp);
    637 	else if (error == ENOENT)
    638 		printf("warning: no /dev/console\n");
    639 	else
    640 		printf("warning: lookup /dev/console: error %d\n", error);
    641 }
    642 
    643 /*
    644  * List of paths to try when searching for "init".
    645  */
    646 static const char *initpaths[] = {
    647 	"/sbin/init",
    648 	"/sbin/oinit",
    649 	"/sbin/init.bak",
    650 	NULL,
    651 };
    652 
    653 /*
    654  * Start the initial user process; try exec'ing each pathname in "initpaths".
    655  * The program is invoked with one argument containing the boot flags.
    656  */
    657 static void
    658 start_init(void *arg)
    659 {
    660 	struct lwp *l = arg;
    661 	struct proc *p = l->l_proc;
    662 	vaddr_t addr;
    663 	struct sys_execve_args /* {
    664 		syscallarg(const char *) path;
    665 		syscallarg(char * const *) argp;
    666 		syscallarg(char * const *) envp;
    667 	} */ args;
    668 	int options, i, error;
    669 	register_t retval[2];
    670 	char flags[4], *flagsp;
    671 	const char *path, *slash;
    672 	char *ucp, **uap, *arg0, *arg1 = NULL;
    673 	char ipath[129];
    674 	int ipx, len;
    675 
    676 	/*
    677 	 * Now in process 1.
    678 	 */
    679 	strncpy(p->p_comm, "init", MAXCOMLEN);
    680 
    681 	/*
    682 	 * Wait for main() to tell us that it's safe to exec.
    683 	 */
    684 	while (start_init_exec == 0)
    685 		(void) tsleep(&start_init_exec, PWAIT, "initexec", 0);
    686 
    687 	/*
    688 	 * This is not the right way to do this.  We really should
    689 	 * hand-craft a descriptor onto /dev/console to hand to init,
    690 	 * but that's a _lot_ more work, and the benefit from this easy
    691 	 * hack makes up for the "good is the enemy of the best" effect.
    692 	 */
    693 	check_console(l);
    694 
    695 	/*
    696 	 * Need just enough stack to hold the faked-up "execve()" arguments.
    697 	 */
    698 	addr = (vaddr_t)STACK_ALLOC(USRSTACK, PAGE_SIZE);
    699 	if (uvm_map(&p->p_vmspace->vm_map, &addr, PAGE_SIZE,
    700                     NULL, UVM_UNKNOWN_OFFSET, 0,
    701                     UVM_MAPFLAG(UVM_PROT_ALL, UVM_PROT_ALL, UVM_INH_COPY,
    702 		    UVM_ADV_NORMAL,
    703                     UVM_FLAG_FIXED|UVM_FLAG_OVERLAY|UVM_FLAG_COPYONW)) != 0)
    704 		panic("init: couldn't allocate argument space");
    705 	p->p_vmspace->vm_maxsaddr = (void *)STACK_MAX(addr, PAGE_SIZE);
    706 
    707 	ipx = 0;
    708 	while (1) {
    709 		if (boothowto & RB_ASKNAME) {
    710 			printf("init path");
    711 			if (initpaths[ipx])
    712 				printf(" (default %s)", initpaths[ipx]);
    713 			printf(": ");
    714 			len = cngetsn(ipath, sizeof(ipath)-1);
    715 			if (len == 0) {
    716 				if (initpaths[ipx])
    717 					path = initpaths[ipx++];
    718 				else
    719 					continue;
    720 			} else {
    721 				ipath[len] = '\0';
    722 				path = ipath;
    723 			}
    724 		} else {
    725 			if ((path = initpaths[ipx++]) == NULL)
    726 				break;
    727 		}
    728 
    729 		ucp = (char *)USRSTACK;
    730 
    731 		/*
    732 		 * Construct the boot flag argument.
    733 		 */
    734 		flagsp = flags;
    735 		*flagsp++ = '-';
    736 		options = 0;
    737 
    738 		if (boothowto & RB_SINGLE) {
    739 			*flagsp++ = 's';
    740 			options = 1;
    741 		}
    742 #ifdef notyet
    743 		if (boothowto & RB_FASTBOOT) {
    744 			*flagsp++ = 'f';
    745 			options = 1;
    746 		}
    747 #endif
    748 
    749 		/*
    750 		 * Move out the flags (arg 1), if necessary.
    751 		 */
    752 		if (options != 0) {
    753 			*flagsp++ = '\0';
    754 			i = flagsp - flags;
    755 #ifdef DEBUG
    756 			printf("init: copying out flags `%s' %d\n", flags, i);
    757 #endif
    758 			arg1 = STACK_ALLOC(ucp, i);
    759 			ucp = STACK_MAX(arg1, i);
    760 			(void)copyout((void *)flags, arg1, i);
    761 		}
    762 
    763 		/*
    764 		 * Move out the file name (also arg 0).
    765 		 */
    766 		i = strlen(path) + 1;
    767 #ifdef DEBUG
    768 		printf("init: copying out path `%s' %d\n", path, i);
    769 #else
    770 		if (boothowto & RB_ASKNAME || path != initpaths[0])
    771 			printf("init: trying %s\n", path);
    772 #endif
    773 		arg0 = STACK_ALLOC(ucp, i);
    774 		ucp = STACK_MAX(arg0, i);
    775 		(void)copyout(path, arg0, i);
    776 
    777 		/*
    778 		 * Move out the arg pointers.
    779 		 */
    780 		ucp = (void *)STACK_ALIGN(ucp, ALIGNBYTES);
    781 		uap = (char **)STACK_ALLOC(ucp, sizeof(char *) * 3);
    782 		SCARG(&args, path) = arg0;
    783 		SCARG(&args, argp) = uap;
    784 		SCARG(&args, envp) = NULL;
    785 		slash = strrchr(path, '/');
    786 		if (slash)
    787 			(void)suword((void *)uap++,
    788 			    (long)arg0 + (slash + 1 - path));
    789 		else
    790 			(void)suword((void *)uap++, (long)arg0);
    791 		if (options != 0)
    792 			(void)suword((void *)uap++, (long)arg1);
    793 		(void)suword((void *)uap++, 0);	/* terminator */
    794 
    795 		/*
    796 		 * Now try to exec the program.  If can't for any reason
    797 		 * other than it doesn't exist, complain.
    798 		 */
    799 		error = sys_execve(l, &args, retval);
    800 		if (error == 0 || error == EJUSTRETURN) {
    801 			KERNEL_UNLOCK_LAST(l);
    802 			return;
    803 		}
    804 		printf("exec %s: error %d\n", path, error);
    805 	}
    806 	printf("init: not found\n");
    807 	panic("no init");
    808 }
    809