Home | History | Annotate | Line # | Download | only in kern
init_main.c revision 1.325
      1 /*	$NetBSD: init_main.c,v 1.325 2007/11/07 00:19:08 ad 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 1.325 2007/11/07 00:19:08 ad Exp $");
     75 
     76 #include "opt_ipsec.h"
     77 #include "opt_multiprocessor.h"
     78 #include "opt_ntp.h"
     79 #include "opt_pipe.h"
     80 #include "opt_posix.h"
     81 #include "opt_syscall_debug.h"
     82 #include "opt_sysv.h"
     83 #include "opt_systrace.h"
     84 #include "opt_fileassoc.h"
     85 #include "opt_ktrace.h"
     86 #include "opt_pax.h"
     87 
     88 #include "rnd.h"
     89 #include "sysmon_envsys.h"
     90 #include "sysmon_power.h"
     91 #include "sysmon_taskq.h"
     92 #include "sysmon_wdog.h"
     93 #include "veriexec.h"
     94 
     95 #include <sys/param.h>
     96 #include <sys/acct.h>
     97 #include <sys/filedesc.h>
     98 #include <sys/file.h>
     99 #include <sys/errno.h>
    100 #include <sys/callout.h>
    101 #include <sys/cpu.h>
    102 #include <sys/kernel.h>
    103 #include <sys/kmem.h>
    104 #include <sys/mount.h>
    105 #include <sys/proc.h>
    106 #include <sys/kthread.h>
    107 #include <sys/resourcevar.h>
    108 #include <sys/signalvar.h>
    109 #include <sys/systm.h>
    110 #include <sys/vnode.h>
    111 #include <sys/fstrans.h>
    112 #include <sys/tty.h>
    113 #include <sys/conf.h>
    114 #include <sys/disklabel.h>
    115 #include <sys/buf.h>
    116 #include <sys/device.h>
    117 #include <sys/exec.h>
    118 #include <sys/socketvar.h>
    119 #include <sys/protosw.h>
    120 #include <sys/reboot.h>
    121 #include <sys/user.h>
    122 #include <sys/sysctl.h>
    123 #include <sys/event.h>
    124 #include <sys/mbuf.h>
    125 #include <sys/sched.h>
    126 #include <sys/sleepq.h>
    127 #include <sys/iostat.h>
    128 #include <sys/vmem.h>
    129 #include <sys/uuid.h>
    130 #include <sys/extent.h>
    131 #include <sys/disk.h>
    132 #include <sys/mqueue.h>
    133 #include <sys/msgbuf.h>
    134 #ifdef FAST_IPSEC
    135 #include <netipsec/ipsec.h>
    136 #endif
    137 #ifdef SYSVSHM
    138 #include <sys/shm.h>
    139 #endif
    140 #ifdef SYSVSEM
    141 #include <sys/sem.h>
    142 #endif
    143 #ifdef SYSVMSG
    144 #include <sys/msg.h>
    145 #endif
    146 #ifdef SYSTRACE
    147 #include <sys/systrace.h>
    148 #endif
    149 #ifdef P1003_1B_SEMAPHORE
    150 #include <sys/ksem.h>
    151 #endif
    152 #include <sys/domain.h>
    153 #include <sys/namei.h>
    154 #if NRND > 0
    155 #include <sys/rnd.h>
    156 #endif
    157 #include <sys/pipe.h>
    158 #ifdef LKM
    159 #include <sys/lkm.h>
    160 #endif
    161 #if NVERIEXEC > 0
    162 #include <sys/verified_exec.h>
    163 #endif /* NVERIEXEC > 0 */
    164 #ifdef KTRACE
    165 #include <sys/ktrace.h>
    166 #endif
    167 #include <sys/debug.h>
    168 #include <sys/kauth.h>
    169 #include <net80211/ieee80211_netbsd.h>
    170 
    171 #include <sys/syscall.h>
    172 #include <sys/syscallargs.h>
    173 
    174 #if defined(PAX_MPROTECT) || defined(PAX_SEGVGUARD)
    175 #include <sys/pax.h>
    176 #endif /* PAX_MPROTECT || PAX_SEGVGUARD */
    177 #include <ufs/ufs/quota.h>
    178 
    179 #include <miscfs/genfs/genfs.h>
    180 #include <miscfs/syncfs/syncfs.h>
    181 
    182 #include <sys/cpu.h>
    183 
    184 #include <uvm/uvm.h>
    185 
    186 #if NSYSMON_TASKQ > 0
    187 #include <dev/sysmon/sysmon_taskq.h>
    188 #endif
    189 
    190 #include <dev/cons.h>
    191 
    192 #if NSYSMON_ENVSYS > 0 || NSYSMON_POWER > 0 || NSYSMON_WDOG > 0
    193 #include <dev/sysmon/sysmonvar.h>
    194 #endif
    195 
    196 #include <net/if.h>
    197 #include <net/raw_cb.h>
    198 
    199 #include <secmodel/secmodel.h>
    200 
    201 extern struct proc proc0;
    202 extern struct lwp lwp0;
    203 extern struct cwdinfo cwdi0;
    204 extern time_t rootfstime;
    205 
    206 #ifndef curlwp
    207 struct	lwp *curlwp = &lwp0;
    208 #endif
    209 struct	proc *initproc;
    210 
    211 struct	vnode *rootvp, *swapdev_vp;
    212 int	boothowto;
    213 int	cold = 1;			/* still working on startup */
    214 struct timeval boottime;	        /* time at system startup - will only follow settime deltas */
    215 
    216 volatile int start_init_exec;		/* semaphore for start_init() */
    217 
    218 static void check_console(struct lwp *l);
    219 static void start_init(void *);
    220 void main(void);
    221 
    222 #if defined(__SSP__) || defined(__SSP_ALL__)
    223 long __stack_chk_guard[8] = {0, 0, 0, 0, 0, 0, 0, 0};
    224 void __stack_chk_fail(void);
    225 
    226 void
    227 __stack_chk_fail(void)
    228 {
    229 	panic("stack overflow detected; terminated");
    230 }
    231 #endif
    232 
    233 void __secmodel_none(void);
    234 __weak_alias(secmodel_start,__secmodel_none);
    235 void
    236 __secmodel_none(void)
    237 {
    238 	return;
    239 }
    240 
    241 /*
    242  * System startup; initialize the world, create process 0, mount root
    243  * filesystem, and fork to create init and pagedaemon.  Most of the
    244  * hard work is done in the lower-level initialization routines including
    245  * startup(), which does memory initialization and autoconfiguration.
    246  */
    247 void
    248 main(void)
    249 {
    250 #ifdef __HAVE_TIMECOUNTER
    251 	struct timeval time;
    252 #endif
    253 	struct lwp *l;
    254 	struct proc *p;
    255 	struct pdevinit *pdev;
    256 	int s, error;
    257 	extern struct pdevinit pdevinit[];
    258 #ifdef NVNODE_IMPLICIT
    259 	int usevnodes;
    260 #endif
    261 	CPU_INFO_ITERATOR cii;
    262 	struct cpu_info *ci;
    263 
    264 	l = &lwp0;
    265 	l->l_cpu = curcpu();
    266 	l->l_proc = &proc0;
    267 	l->l_lid = 1;
    268 
    269 	/*
    270 	 * XXX This is a temporary check to be removed before
    271 	 * NetBSD 5.0 is released.
    272 	 */
    273 #if !defined(__i386__ ) && !defined(__x86_64__)
    274 	if (curlwp != l) {
    275 		printf("NOTICE: curlwp should be set before main()\n");
    276 		DELAY(250000);
    277 		curlwp = l;
    278 	}
    279 #endif
    280 
    281 	/*
    282 	 * Attempt to find console and initialize
    283 	 * in case of early panic or other messages.
    284 	 */
    285 	consinit();
    286 
    287 	kernel_lock_init();
    288 
    289 	uvm_init();
    290 
    291 #ifdef DEBUG
    292 	debug_init();
    293 #endif
    294 
    295 	kmem_init();
    296 
    297 	/* Initialize the extent manager. */
    298 	extent_init();
    299 
    300 	/* Do machine-dependent initialization. */
    301 	cpu_startup();
    302 
    303 	/* Initialize callouts, part 1. */
    304 	callout_startup();
    305 
    306 	/*
    307 	 * Initialize the kernel authorization subsystem and start the
    308 	 * default security model, if any. We need to do this early
    309 	 * enough so that subsystems relying on any of the aforementioned
    310 	 * can work properly. Since the security model may dictate the
    311 	 * credential inheritance policy, it is needed at least before
    312 	 * any process is created, specifically proc0.
    313 	 */
    314 	kauth_init();
    315 	secmodel_start();
    316 
    317 	/* Initialize the buffer cache */
    318 	bufinit();
    319 
    320 	/*
    321 	 * Initialize mbuf's.  Do this now because we might attempt to
    322 	 * allocate mbufs or mbuf clusters during autoconfiguration.
    323 	 */
    324 	mbinit();
    325 
    326 	/* Initialize sockets. */
    327 	soinit();
    328 
    329 	/*
    330 	 * The following things must be done before autoconfiguration.
    331 	 */
    332 	evcnt_init();		/* initialize event counters */
    333 #if NRND > 0
    334 	rnd_init();		/* initialize RNG */
    335 #endif
    336 
    337 	/* Initialize process and pgrp structures. */
    338 	procinit();
    339 	lwpinit();
    340 
    341 	/* Initialize signal-related data structures. */
    342 	signal_init();
    343 
    344 	/* Create process 0 (the swapper). */
    345 	proc0_init();
    346 
    347 	/* Initialize the UID hash table. */
    348 	uid_init();
    349 
    350 	/* Charge root for one process. */
    351 	(void)chgproccnt(0, 1);
    352 
    353 	/* Initialize the run queues, turnstiles and sleep queues. */
    354 	mutex_init(&cpu_lock, MUTEX_DEFAULT, IPL_NONE);
    355 	sched_rqinit();
    356 	turnstile_init();
    357 	sleeptab_init(&sleeptab);
    358 
    359 	/* MI initialization of the boot cpu */
    360 	error = mi_cpu_attach(curcpu());
    361 	KASSERT(error == 0);
    362 
    363 	/* Initialize the sysctl subsystem. */
    364 	sysctl_init();
    365 
    366 	/* Initialize I/O statistics. */
    367 	iostat_init();
    368 
    369 	/* Initialize the log device. */
    370 	loginit();
    371 
    372 	/* Initialize the file systems. */
    373 #ifdef NVNODE_IMPLICIT
    374 	/*
    375 	 * If maximum number of vnodes in namei vnode cache is not explicitly
    376 	 * defined in kernel config, adjust the number such as we use roughly
    377 	 * 1.0% of memory for vnode cache (but not less than NVNODE vnodes).
    378 	 */
    379 	usevnodes = (ptoa((unsigned)physmem) / 100) / sizeof(struct vnode);
    380 	if (usevnodes > desiredvnodes)
    381 		desiredvnodes = usevnodes;
    382 #endif
    383 	vfsinit();
    384 
    385 	/* Initialize fstrans. */
    386 	fstrans_init();
    387 
    388 	/* Initialize the file descriptor system. */
    389 	filedesc_init();
    390 
    391 	/* Initialize the select()/poll() system calls. */
    392 	selsysinit();
    393 
    394 	/* Initialize asynchronous I/O. */
    395 	aio_sysinit();
    396 
    397 	/* Initialize message queues. */
    398 	mqueue_sysinit();
    399 
    400 	/* Initialize the system monitor subsystems. */
    401 #if NSYSMON_TASKQ > 0
    402 	sysmon_task_queue_preinit();
    403 #endif
    404 
    405 #if NSYSMON_ENVSYS > 0
    406 	sysmon_envsys_init();
    407 #endif
    408 
    409 #if NSYSMON_POWER > 0
    410 	sysmon_power_init();
    411 #endif
    412 
    413 #if NSYSMON_WDOG > 0
    414 	sysmon_wdog_init();
    415 #endif
    416 
    417 #ifdef __HAVE_TIMECOUNTER
    418 	inittimecounter();
    419 	ntp_init();
    420 #endif /* __HAVE_TIMECOUNTER */
    421 
    422 	/* Initialize the device switch tables. */
    423 	devsw_init();
    424 
    425 	/* Initialize the disk wedge subsystem. */
    426 	dkwedge_init();
    427 
    428 	/* Configure the system hardware.  This will enable interrupts. */
    429 	configure();
    430 
    431 #if defined(__SSP__) || defined(__SSP_ALL__)
    432 	{
    433 #ifdef DIAGNOSTIC
    434 		printf("Initializing SSP:");
    435 #endif
    436 		/*
    437 		 * We initialize ssp here carefully:
    438 		 *	1. after we got some entropy
    439 		 *	2. without calling a function
    440 		 */
    441 		size_t i;
    442 		long guard[__arraycount(__stack_chk_guard)];
    443 
    444 		arc4randbytes(guard, sizeof(guard));
    445 		for (i = 0; i < __arraycount(guard); i++)
    446 			__stack_chk_guard[i] = guard[i];
    447 #ifdef DIAGNOSTIC
    448 		for (i = 0; i < __arraycount(guard); i++)
    449 			printf("%lx ", guard[i]);
    450 		printf("\n");
    451 #endif
    452 	}
    453 #endif
    454 	ubc_init();		/* must be after autoconfig */
    455 
    456 	/* Lock the kernel on behalf of proc0. */
    457 	KERNEL_LOCK(1, l);
    458 
    459 #ifdef SYSTRACE
    460 	systrace_init();
    461 #endif
    462 
    463 #ifdef SYSVSHM
    464 	/* Initialize System V style shared memory. */
    465 	shminit();
    466 #endif
    467 
    468 #ifdef SYSVSEM
    469 	/* Initialize System V style semaphores. */
    470 	seminit();
    471 #endif
    472 
    473 #ifdef SYSVMSG
    474 	/* Initialize System V style message queues. */
    475 	msginit();
    476 #endif
    477 
    478 #ifdef P1003_1B_SEMAPHORE
    479 	/* Initialize posix semaphores */
    480 	ksem_init();
    481 #endif
    482 
    483 #if NVERIEXEC > 0
    484 	/*
    485 	 * Initialise the Veriexec subsystem.
    486 	 */
    487 	veriexec_init();
    488 #endif /* NVERIEXEC > 0 */
    489 
    490 #if defined(PAX_MPROTECT) || defined(PAX_SEGVGUARD)
    491 	pax_init();
    492 #endif /* PAX_MPROTECT || PAX_SEGVGUARD */
    493 
    494 	/* Attach pseudo-devices. */
    495 	for (pdev = pdevinit; pdev->pdev_attach != NULL; pdev++)
    496 		(*pdev->pdev_attach)(pdev->pdev_count);
    497 
    498 #ifdef	FAST_IPSEC
    499 	/* Attach network crypto subsystem */
    500 	ipsec_attach();
    501 #endif
    502 
    503 	/*
    504 	 * Initialize protocols.  Block reception of incoming packets
    505 	 * until everything is ready.
    506 	 */
    507 	s = splnet();
    508 	ifinit();
    509 	domaininit();
    510 	if_attachdomain();
    511 	splx(s);
    512 
    513 #ifdef GPROF
    514 	/* Initialize kernel profiling. */
    515 	kmstartup();
    516 #endif
    517 
    518 	/* Initialize system accouting. */
    519 	acct_init();
    520 
    521 #ifndef PIPE_SOCKETPAIR
    522 	/* Initialize pipes. */
    523 	pipe_init();
    524 #endif
    525 
    526 	/* Setup the scheduler */
    527 	sched_init();
    528 
    529 #ifdef KTRACE
    530 	/* Initialize ktrace. */
    531 	ktrinit();
    532 #endif
    533 
    534 	/* Initialize the UUID system calls. */
    535 	uuid_init();
    536 
    537 	/*
    538 	 * Create process 1 (init(8)).  We do this now, as Unix has
    539 	 * historically had init be process 1, and changing this would
    540 	 * probably upset a lot of people.
    541 	 *
    542 	 * Note that process 1 won't immediately exec init(8), but will
    543 	 * wait for us to inform it that the root file system has been
    544 	 * mounted.
    545 	 */
    546 	if (fork1(l, 0, SIGCHLD, NULL, 0, start_init, NULL, NULL, &initproc))
    547 		panic("fork init");
    548 
    549 	/*
    550 	 * Now that device driver threads have been created, wait for
    551 	 * them to finish any deferred autoconfiguration.  Note we don't
    552 	 * need to lock this semaphore, since we haven't booted any
    553 	 * secondary processors, yet.
    554 	 */
    555 	while (config_pending)
    556 		(void) tsleep(&config_pending, PWAIT, "cfpend", 0);
    557 
    558 	/*
    559 	 * Finalize configuration now that all real devices have been
    560 	 * found.  This needs to be done before the root device is
    561 	 * selected, since finalization may create the root device.
    562 	 */
    563 	config_finalize();
    564 
    565 	/*
    566 	 * Now that autoconfiguration has completed, we can determine
    567 	 * the root and dump devices.
    568 	 */
    569 	cpu_rootconf();
    570 	cpu_dumpconf();
    571 
    572 	/* Mount the root file system. */
    573 	do {
    574 		domountroothook();
    575 		if ((error = vfs_mountroot())) {
    576 			printf("cannot mount root, error = %d\n", error);
    577 			boothowto |= RB_ASKNAME;
    578 			setroot(root_device,
    579 			    (rootdev != NODEV) ? DISKPART(rootdev) : 0);
    580 		}
    581 	} while (error != 0);
    582 	mountroothook_destroy();
    583 
    584 	/*
    585 	 * Initialise the time-of-day clock, passing the time recorded
    586 	 * in the root filesystem (if any) for use by systems that
    587 	 * don't have a non-volatile time-of-day device.
    588 	 */
    589 	inittodr(rootfstime);
    590 
    591 	CIRCLEQ_FIRST(&mountlist)->mnt_flag |= MNT_ROOTFS;
    592 	CIRCLEQ_FIRST(&mountlist)->mnt_op->vfs_refcount++;
    593 
    594 	/*
    595 	 * Get the vnode for '/'.  Set filedesc0.fd_fd.fd_cdir to
    596 	 * reference it.
    597 	 */
    598 	error = VFS_ROOT(CIRCLEQ_FIRST(&mountlist), &rootvnode);
    599 	if (error)
    600 		panic("cannot find root vnode, error=%d", error);
    601 	cwdi0.cwdi_cdir = rootvnode;
    602 	VREF(cwdi0.cwdi_cdir);
    603 	VOP_UNLOCK(rootvnode, 0);
    604 	cwdi0.cwdi_rdir = NULL;
    605 
    606 	/*
    607 	 * Now that root is mounted, we can fixup initproc's CWD
    608 	 * info.  All other processes are kthreads, which merely
    609 	 * share proc0's CWD info.
    610 	 */
    611 	initproc->p_cwdi->cwdi_cdir = rootvnode;
    612 	VREF(initproc->p_cwdi->cwdi_cdir);
    613 	initproc->p_cwdi->cwdi_rdir = NULL;
    614 
    615 	/*
    616 	 * Now can look at time, having had a chance to verify the time
    617 	 * from the file system.  Reset l->l_rtime as it may have been
    618 	 * munched in mi_switch() after the time got set.
    619 	 */
    620 #ifdef __HAVE_TIMECOUNTER
    621 	getmicrotime(&time);
    622 #else
    623 	mono_time = time;
    624 #endif
    625 	boottime = time;
    626 	mutex_enter(&proclist_lock);
    627 	LIST_FOREACH(p, &allproc, p_list) {
    628 		KASSERT((p->p_flag & PK_MARKER) == 0);
    629 		mutex_enter(&p->p_smutex);
    630 		p->p_stats->p_start = time;
    631 		LIST_FOREACH(l, &p->p_lwps, l_sibling) {
    632 			lwp_lock(l);
    633 			l->l_rtime.tv_sec = l->l_rtime.tv_usec = 0;
    634 			lwp_unlock(l);
    635 		}
    636 		mutex_exit(&p->p_smutex);
    637 	}
    638 	mutex_exit(&proclist_lock);
    639 	curlwp->l_stime = time;
    640 
    641 	for (CPU_INFO_FOREACH(cii, ci)) {
    642 		ci->ci_schedstate.spc_lastmod = time_second;
    643 	}
    644 
    645 	/* Create the pageout daemon kernel thread. */
    646 	uvm_swap_init();
    647 	if (kthread_create(PRI_PGDAEMON, 0, NULL, uvm_pageout,
    648 	    NULL, NULL, "pgdaemon"))
    649 		panic("fork pagedaemon");
    650 
    651 	/* Create the filesystem syncer kernel thread. */
    652 	if (kthread_create(PRI_IOFLUSH, 0, NULL, sched_sync, NULL, NULL, "ioflush"))
    653 		panic("fork syncer");
    654 
    655 	/* Create the aiodone daemon kernel thread. */
    656 	if (workqueue_create(&uvm.aiodone_queue, "aiodoned",
    657 	    uvm_aiodone_worker, NULL, PRI_VM, IPL_BIO, 0))
    658 		panic("fork aiodoned");
    659 
    660 	vmem_rehash_start();
    661 
    662 #if defined(MULTIPROCESSOR)
    663 	/* Boot the secondary processors. */
    664 	cpu_boot_secondary_processors();
    665 #endif
    666 
    667 	/* Initialize exec structures */
    668 	exec_init(1);
    669 
    670 	/*
    671 	 * Okay, now we can let init(8) exec!  It's off to userland!
    672 	 */
    673 	start_init_exec = 1;
    674 	wakeup(&start_init_exec);
    675 
    676 	/* The scheduler is an infinite loop. */
    677 	uvm_scheduler();
    678 	/* NOTREACHED */
    679 }
    680 
    681 static void
    682 check_console(struct lwp *l)
    683 {
    684 	struct nameidata nd;
    685 	int error;
    686 
    687 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_SYSSPACE, "/dev/console", l);
    688 	error = namei(&nd);
    689 	if (error == 0)
    690 		vrele(nd.ni_vp);
    691 	else if (error == ENOENT)
    692 		printf("warning: no /dev/console\n");
    693 	else
    694 		printf("warning: lookup /dev/console: error %d\n", error);
    695 }
    696 
    697 /*
    698  * List of paths to try when searching for "init".
    699  */
    700 static const char *initpaths[] = {
    701 	"/sbin/init",
    702 	"/sbin/oinit",
    703 	"/sbin/init.bak",
    704 	NULL,
    705 };
    706 
    707 /*
    708  * Start the initial user process; try exec'ing each pathname in "initpaths".
    709  * The program is invoked with one argument containing the boot flags.
    710  */
    711 static void
    712 start_init(void *arg)
    713 {
    714 	struct lwp *l = arg;
    715 	struct proc *p = l->l_proc;
    716 	vaddr_t addr;
    717 	struct sys_execve_args /* {
    718 		syscallarg(const char *) path;
    719 		syscallarg(char * const *) argp;
    720 		syscallarg(char * const *) envp;
    721 	} */ args;
    722 	int options, i, error;
    723 	register_t retval[2];
    724 	char flags[4], *flagsp;
    725 	const char *path, *slash;
    726 	char *ucp, **uap, *arg0, *arg1 = NULL;
    727 	char ipath[129];
    728 	int ipx, len;
    729 
    730 	/*
    731 	 * Now in process 1.
    732 	 */
    733 	strncpy(p->p_comm, "init", MAXCOMLEN);
    734 
    735 	/*
    736 	 * Wait for main() to tell us that it's safe to exec.
    737 	 */
    738 	while (start_init_exec == 0)
    739 		(void) tsleep(&start_init_exec, PWAIT, "initexec", 0);
    740 
    741 	/*
    742 	 * This is not the right way to do this.  We really should
    743 	 * hand-craft a descriptor onto /dev/console to hand to init,
    744 	 * but that's a _lot_ more work, and the benefit from this easy
    745 	 * hack makes up for the "good is the enemy of the best" effect.
    746 	 */
    747 	check_console(l);
    748 
    749 	/*
    750 	 * Need just enough stack to hold the faked-up "execve()" arguments.
    751 	 */
    752 	addr = (vaddr_t)STACK_ALLOC(USRSTACK, PAGE_SIZE);
    753 	if (uvm_map(&p->p_vmspace->vm_map, &addr, PAGE_SIZE,
    754                     NULL, UVM_UNKNOWN_OFFSET, 0,
    755                     UVM_MAPFLAG(UVM_PROT_ALL, UVM_PROT_ALL, UVM_INH_COPY,
    756 		    UVM_ADV_NORMAL,
    757                     UVM_FLAG_FIXED|UVM_FLAG_OVERLAY|UVM_FLAG_COPYONW)) != 0)
    758 		panic("init: couldn't allocate argument space");
    759 	p->p_vmspace->vm_maxsaddr = (void *)STACK_MAX(addr, PAGE_SIZE);
    760 
    761 	ipx = 0;
    762 	while (1) {
    763 		if (boothowto & RB_ASKNAME) {
    764 			printf("init path");
    765 			if (initpaths[ipx])
    766 				printf(" (default %s)", initpaths[ipx]);
    767 			printf(": ");
    768 			len = cngetsn(ipath, sizeof(ipath)-1);
    769 			if (len == 0) {
    770 				if (initpaths[ipx])
    771 					path = initpaths[ipx++];
    772 				else
    773 					continue;
    774 			} else {
    775 				ipath[len] = '\0';
    776 				path = ipath;
    777 			}
    778 		} else {
    779 			if ((path = initpaths[ipx++]) == NULL)
    780 				break;
    781 		}
    782 
    783 		ucp = (char *)USRSTACK;
    784 
    785 		/*
    786 		 * Construct the boot flag argument.
    787 		 */
    788 		flagsp = flags;
    789 		*flagsp++ = '-';
    790 		options = 0;
    791 
    792 		if (boothowto & RB_SINGLE) {
    793 			*flagsp++ = 's';
    794 			options = 1;
    795 		}
    796 #ifdef notyet
    797 		if (boothowto & RB_FASTBOOT) {
    798 			*flagsp++ = 'f';
    799 			options = 1;
    800 		}
    801 #endif
    802 
    803 		/*
    804 		 * Move out the flags (arg 1), if necessary.
    805 		 */
    806 		if (options != 0) {
    807 			*flagsp++ = '\0';
    808 			i = flagsp - flags;
    809 #ifdef DEBUG
    810 			printf("init: copying out flags `%s' %d\n", flags, i);
    811 #endif
    812 			arg1 = STACK_ALLOC(ucp, i);
    813 			ucp = STACK_MAX(arg1, i);
    814 			(void)copyout((void *)flags, arg1, i);
    815 		}
    816 
    817 		/*
    818 		 * Move out the file name (also arg 0).
    819 		 */
    820 		i = strlen(path) + 1;
    821 #ifdef DEBUG
    822 		printf("init: copying out path `%s' %d\n", path, i);
    823 #else
    824 		if (boothowto & RB_ASKNAME || path != initpaths[0])
    825 			printf("init: trying %s\n", path);
    826 #endif
    827 		arg0 = STACK_ALLOC(ucp, i);
    828 		ucp = STACK_MAX(arg0, i);
    829 		(void)copyout(path, arg0, i);
    830 
    831 		/*
    832 		 * Move out the arg pointers.
    833 		 */
    834 		ucp = (void *)STACK_ALIGN(ucp, ALIGNBYTES);
    835 		uap = (char **)STACK_ALLOC(ucp, sizeof(char *) * 3);
    836 		SCARG(&args, path) = arg0;
    837 		SCARG(&args, argp) = uap;
    838 		SCARG(&args, envp) = NULL;
    839 		slash = strrchr(path, '/');
    840 		if (slash)
    841 			(void)suword((void *)uap++,
    842 			    (long)arg0 + (slash + 1 - path));
    843 		else
    844 			(void)suword((void *)uap++, (long)arg0);
    845 		if (options != 0)
    846 			(void)suword((void *)uap++, (long)arg1);
    847 		(void)suword((void *)uap++, 0);	/* terminator */
    848 
    849 		/*
    850 		 * Now try to exec the program.  If can't for any reason
    851 		 * other than it doesn't exist, complain.
    852 		 */
    853 		error = sys_execve(l, &args, retval);
    854 		if (error == 0 || error == EJUSTRETURN) {
    855 			KERNEL_UNLOCK_LAST(l);
    856 			return;
    857 		}
    858 		printf("exec %s: error %d\n", path, error);
    859 	}
    860 	printf("init: not found\n");
    861 	panic("no init");
    862 }
    863