Home | History | Annotate | Line # | Download | only in kern
init_main.c revision 1.240
      1 /*	$NetBSD: init_main.c,v 1.240 2004/10/01 16:30:52 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 1.240 2004/10/01 16:30:52 yamt 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/disk.h>
    112 #include <sys/exec.h>
    113 #include <sys/socketvar.h>
    114 #include <sys/protosw.h>
    115 #include <sys/reboot.h>
    116 #include <sys/user.h>
    117 #include <sys/sysctl.h>
    118 #include <sys/event.h>
    119 #include <sys/mbuf.h>
    120 #ifdef FAST_IPSEC
    121 #include <netipsec/ipsec.h>
    122 #endif
    123 #ifdef SYSVSHM
    124 #include <sys/shm.h>
    125 #endif
    126 #ifdef SYSVSEM
    127 #include <sys/sem.h>
    128 #endif
    129 #ifdef SYSVMSG
    130 #include <sys/msg.h>
    131 #endif
    132 #ifdef P1003_1B_SEMAPHORE
    133 #include <sys/ksem.h>
    134 #endif
    135 #ifdef SYSTRACE
    136 #include <sys/systrace.h>
    137 #endif
    138 #include <sys/domain.h>
    139 #include <sys/namei.h>
    140 #if NOPENCRYPTO > 0
    141 #include <opencrypto/cryptodev.h>	/* XXX really the framework */
    142 #endif
    143 #if NRND > 0
    144 #include <sys/rnd.h>
    145 #endif
    146 #ifndef PIPE_SOCKETPAIR
    147 #include <sys/pipe.h>
    148 #endif
    149 #ifdef LKM
    150 #include <sys/lkm.h>
    151 #endif
    152 
    153 #include <sys/syscall.h>
    154 #include <sys/sa.h>
    155 #include <sys/syscallargs.h>
    156 
    157 #include <ufs/ufs/quota.h>
    158 
    159 #include <miscfs/genfs/genfs.h>
    160 #include <miscfs/syncfs/syncfs.h>
    161 
    162 #include <machine/cpu.h>
    163 
    164 #include <uvm/uvm.h>
    165 
    166 #include <dev/cons.h>
    167 
    168 #include <net/if.h>
    169 #include <net/raw_cb.h>
    170 
    171 /* Components of the first process -- never freed. */
    172 struct	session session0;
    173 struct	pgrp pgrp0;
    174 struct	proc proc0;
    175 struct	lwp lwp0;
    176 struct	pcred cred0;
    177 struct	filedesc0 filedesc0;
    178 struct	cwdinfo cwdi0;
    179 struct	plimit limit0;
    180 struct	pstats pstat0;
    181 struct	vmspace vmspace0;
    182 struct	sigacts sigacts0;
    183 #ifndef curlwp
    184 struct	lwp *curlwp = &lwp0;
    185 #endif
    186 struct	proc *initproc;
    187 
    188 int	nofile = NOFILE;
    189 int	maxuprc = MAXUPRC;
    190 int	cmask = CMASK;
    191 extern	struct user *proc0paddr;
    192 
    193 struct	vnode *rootvp, *swapdev_vp;
    194 int	boothowto;
    195 int	cold = 1;			/* still working on startup */
    196 struct	timeval boottime;
    197 time_t	rootfstime;			/* recorded root fs time, if known */
    198 
    199 __volatile int start_init_exec;		/* semaphore for start_init() */
    200 
    201 static void check_console(struct proc *p);
    202 static void start_init(void *);
    203 void main(void);
    204 
    205 extern const struct emul emul_netbsd;	/* defined in kern_exec.c */
    206 
    207 /*
    208  * System startup; initialize the world, create process 0, mount root
    209  * filesystem, and fork to create init and pagedaemon.  Most of the
    210  * hard work is done in the lower-level initialization routines including
    211  * startup(), which does memory initialization and autoconfiguration.
    212  */
    213 void
    214 main(void)
    215 {
    216 	struct lwp *l;
    217 	struct proc *p;
    218 	struct pdevinit *pdev;
    219 	int s, error;
    220 	u_int i;
    221 	rlim_t lim;
    222 	extern struct pdevinit pdevinit[];
    223 	extern void schedcpu(void *);
    224 #if defined(NFSSERVER) || defined(NFS)
    225 	extern void nfs_init(void);
    226 #endif
    227 #ifdef NVNODE_IMPLICIT
    228 	int usevnodes;
    229 #endif
    230 
    231 	/*
    232 	 * Initialize the current LWP pointer (curlwp) before
    233 	 * any possible traps/probes to simplify trap processing.
    234 	 */
    235 	l = &lwp0;
    236 	curlwp = l;
    237 	l->l_cpu = curcpu();
    238 	l->l_proc = &proc0;
    239 	l->l_lid = 1;
    240 
    241 	/*
    242 	 * Attempt to find console and initialize
    243 	 * in case of early panic or other messages.
    244 	 */
    245 	consinit();
    246 	printf("%s", copyright);
    247 
    248 	KERNEL_LOCK_INIT();
    249 
    250 	uvm_init();
    251 
    252 	/* Do machine-dependent initialization. */
    253 	cpu_startup();
    254 
    255 	/* Initialise pools. */
    256 	link_pool_init();
    257 
    258 	/* Initialize callouts. */
    259 	callout_startup();
    260 
    261 	/* Initialize the buffer cache */
    262 	bufinit();
    263 
    264 	/*
    265 	 * Initialize mbuf's.  Do this now because we might attempt to
    266 	 * allocate mbufs or mbuf clusters during autoconfiguration.
    267 	 */
    268 	mbinit();
    269 
    270 	/* Initialize sockets. */
    271 	soinit();
    272 
    273 #ifdef KCONT
    274 	/* Initialize kcont. */
    275         kcont_init();
    276 #endif
    277 
    278 	/*
    279 	 * The following things must be done before autoconfiguration.
    280 	 */
    281 	evcnt_init();		/* initialize event counters */
    282 	disk_init();		/* initialize disk list */
    283 	tty_init();		/* initialize tty list */
    284 #if NRND > 0
    285 	rnd_init();		/* initialize RNG */
    286 #endif
    287 #if NOPENCRYPTO > 0
    288 	/* Initialize crypto subsystem before configuring crypto hardware. */
    289 	(void)crypto_init();
    290 #endif
    291 	/* Initialize the sysctl subsystem. */
    292 	sysctl_init();
    293 
    294 	/* Initialize process and pgrp structures. */
    295 	procinit();
    296 
    297 #ifdef LKM
    298 	/* Initialize the LKM system. */
    299 	lkm_init();
    300 #endif
    301 
    302 	/*
    303 	 * Create process 0 (the swapper).
    304 	 */
    305 	p = &proc0;
    306 	proc0_insert(p, l, &pgrp0, &session0);
    307 
    308 	/*
    309 	 * Set P_NOCLDWAIT so that kernel threads are reparented to
    310 	 * init(8) when they exit.  init(8) can easily wait them out
    311 	 * for us.
    312 	 */
    313 	p->p_flag = P_SYSTEM | P_NOCLDWAIT;
    314 	p->p_stat = SACTIVE;
    315 	p->p_nice = NZERO;
    316 	p->p_emul = &emul_netbsd;
    317 #ifdef __HAVE_SYSCALL_INTERN
    318 	(*p->p_emul->e_syscall_intern)(p);
    319 #endif
    320 	strncpy(p->p_comm, "swapper", MAXCOMLEN);
    321 
    322 	l->l_flag = L_INMEM;
    323 	l->l_stat = LSONPROC;
    324 	p->p_nrlwps = 1;
    325 
    326 	callout_init(&l->l_tsleep_ch);
    327 
    328 	/* Create credentials. */
    329 	cred0.p_refcnt = 1;
    330 	p->p_cred = &cred0;
    331 	p->p_ucred = crget();
    332 	p->p_ucred->cr_ngroups = 1;	/* group 0 */
    333 
    334 	/* Create the file descriptor table. */
    335 	p->p_fd = &filedesc0.fd_fd;
    336 	fdinit1(&filedesc0);
    337 
    338 	/* Create the CWD info. */
    339 	p->p_cwdi = &cwdi0;
    340 	cwdi0.cwdi_cmask = cmask;
    341 	cwdi0.cwdi_refcnt = 1;
    342 	simple_lock_init(&cwdi0.cwdi_slock);
    343 
    344 	/* Create the limits structures. */
    345 	p->p_limit = &limit0;
    346 	simple_lock_init(&limit0.p_slock);
    347 	for (i = 0; i < sizeof(p->p_rlimit)/sizeof(p->p_rlimit[0]); i++)
    348 		limit0.pl_rlimit[i].rlim_cur =
    349 		    limit0.pl_rlimit[i].rlim_max = RLIM_INFINITY;
    350 
    351 	limit0.pl_rlimit[RLIMIT_NOFILE].rlim_max = maxfiles;
    352 	limit0.pl_rlimit[RLIMIT_NOFILE].rlim_cur =
    353 	    maxfiles < nofile ? maxfiles : nofile;
    354 
    355 	limit0.pl_rlimit[RLIMIT_NPROC].rlim_max = maxproc;
    356 	limit0.pl_rlimit[RLIMIT_NPROC].rlim_cur =
    357 	    maxproc < maxuprc ? maxproc : maxuprc;
    358 
    359 	lim = ptoa(uvmexp.free);
    360 	limit0.pl_rlimit[RLIMIT_RSS].rlim_max = lim;
    361 	limit0.pl_rlimit[RLIMIT_MEMLOCK].rlim_max = lim;
    362 	limit0.pl_rlimit[RLIMIT_MEMLOCK].rlim_cur = lim / 3;
    363 	limit0.pl_corename = defcorename;
    364 	limit0.p_refcnt = 1;
    365 
    366 	/*
    367 	 * Initialize proc0's vmspace, which uses the kernel pmap.
    368 	 * All kernel processes (which never have user space mappings)
    369 	 * share proc0's vmspace, and thus, the kernel pmap.
    370 	 */
    371 	uvmspace_init(&vmspace0, pmap_kernel(), round_page(VM_MIN_ADDRESS),
    372 	    trunc_page(VM_MAX_ADDRESS));
    373 	p->p_vmspace = &vmspace0;
    374 
    375 	l->l_addr = proc0paddr;				/* XXX */
    376 
    377 	p->p_stats = &pstat0;
    378 
    379 	/*
    380 	 * Charge root for one process.
    381 	 */
    382 	(void)chgproccnt(0, 1);
    383 
    384 	rqinit();
    385 
    386 	/* Configure virtual memory system, set vm rlimits. */
    387 	uvm_init_limits(p);
    388 
    389 	/* Initialize the file systems. */
    390 #if defined(NFSSERVER) || defined(NFS)
    391 	nfs_init();			/* initialize server/shared data */
    392 #endif
    393 #ifdef NVNODE_IMPLICIT
    394 	/*
    395 	 * If maximum number of vnodes in namei vnode cache is not explicitly
    396 	 * defined in kernel config, adjust the number such as we use roughly
    397 	 * 1.0% of memory for vnode cache (but not less than NVNODE vnodes).
    398 	 */
    399 	usevnodes = (ptoa((unsigned)physmem) / 100) / sizeof(struct vnode);
    400 	if (usevnodes > desiredvnodes)
    401 		desiredvnodes = usevnodes;
    402 #endif
    403 	vfsinit();
    404 
    405 	/* Configure the system hardware.  This will enable interrupts. */
    406 	configure();
    407 
    408 	ubc_init();		/* must be after autoconfig */
    409 
    410 	/* Lock the kernel on behalf of proc0. */
    411 	KERNEL_PROC_LOCK(l);
    412 
    413 #ifdef SYSVSHM
    414 	/* Initialize System V style shared memory. */
    415 	shminit();
    416 #endif
    417 
    418 #ifdef SYSVSEM
    419 	/* Initialize System V style semaphores. */
    420 	seminit();
    421 #endif
    422 
    423 #ifdef SYSVMSG
    424 	/* Initialize System V style message queues. */
    425 	msginit();
    426 #endif
    427 
    428 #ifdef P1003_1B_SEMAPHORE
    429 	/* Initialize posix semaphores */
    430 	ksem_init();
    431 #endif
    432 	/* Attach pseudo-devices. */
    433 	for (pdev = pdevinit; pdev->pdev_attach != NULL; pdev++)
    434 		(*pdev->pdev_attach)(pdev->pdev_count);
    435 
    436 #ifdef	FAST_IPSEC
    437 	/* Attach network crypto subsystem */
    438 	ipsec_attach();
    439 #endif
    440 
    441 	/*
    442 	 * Initialize protocols.  Block reception of incoming packets
    443 	 * until everything is ready.
    444 	 */
    445 	s = splnet();
    446 	ifinit();
    447 	domaininit();
    448 	if_attachdomain();
    449 	splx(s);
    450 
    451 #ifdef GPROF
    452 	/* Initialize kernel profiling. */
    453 	kmstartup();
    454 #endif
    455 
    456 	/* Initialize system accouting. */
    457 	acct_init();
    458 
    459 #ifdef SYSTRACE
    460 	systrace_init();
    461 #endif
    462 	/*
    463 	 * Initialize signal-related data structures, and signal state
    464 	 * for proc0.
    465 	 */
    466 	signal_init();
    467 	p->p_sigacts = &sigacts0;
    468 	siginit(p);
    469 
    470 	/* Kick off timeout driven events by calling first time. */
    471 	schedcpu(NULL);
    472 
    473 	/*
    474 	 * Create process 1 (init(8)).  We do this now, as Unix has
    475 	 * historically had init be process 1, and changing this would
    476 	 * probably upset a lot of people.
    477 	 *
    478 	 * Note that process 1 won't immediately exec init(8), but will
    479 	 * wait for us to inform it that the root file system has been
    480 	 * mounted.
    481 	 */
    482 	if (fork1(l, 0, SIGCHLD, NULL, 0, start_init, NULL, NULL, &initproc))
    483 		panic("fork init");
    484 
    485 	/*
    486 	 * Create any kernel threads who's creation was deferred because
    487 	 * initproc had not yet been created.
    488 	 */
    489 	kthread_run_deferred_queue();
    490 
    491 	/*
    492 	 * Now that device driver threads have been created, wait for
    493 	 * them to finish any deferred autoconfiguration.  Note we don't
    494 	 * need to lock this semaphore, since we haven't booted any
    495 	 * secondary processors, yet.
    496 	 */
    497 	while (config_pending)
    498 		(void) tsleep((void *)&config_pending, PWAIT, "cfpend", 0);
    499 
    500 	/*
    501 	 * Finalize configuration now that all real devices have been
    502 	 * found.  This needs to be done before the root device is
    503 	 * selected, since finalization may create the root device.
    504 	 */
    505 	config_finalize();
    506 
    507 	/*
    508 	 * Now that autoconfiguration has completed, we can determine
    509 	 * the root and dump devices.
    510 	 */
    511 	cpu_rootconf();
    512 	cpu_dumpconf();
    513 
    514 	/* Mount the root file system. */
    515 	do {
    516 		domountroothook();
    517 		if ((error = vfs_mountroot())) {
    518 			printf("cannot mount root, error = %d\n", error);
    519 			boothowto |= RB_ASKNAME;
    520 			setroot(root_device,
    521 			    (rootdev != NODEV) ? DISKPART(rootdev) : 0);
    522 		}
    523 	} while (error != 0);
    524 	mountroothook_destroy();
    525 
    526 	/*
    527 	 * Initialise the time-of-day clock, passing the time recorded
    528 	 * in the root filesystem (if any) for use by systems that
    529 	 * don't have a non-volatile time-of-day device.
    530 	 */
    531 	inittodr(rootfstime);
    532 
    533 	CIRCLEQ_FIRST(&mountlist)->mnt_flag |= MNT_ROOTFS;
    534 	CIRCLEQ_FIRST(&mountlist)->mnt_op->vfs_refcount++;
    535 
    536 	/*
    537 	 * Get the vnode for '/'.  Set filedesc0.fd_fd.fd_cdir to
    538 	 * reference it.
    539 	 */
    540 	if (VFS_ROOT(CIRCLEQ_FIRST(&mountlist), &rootvnode))
    541 		panic("cannot find root vnode");
    542 	cwdi0.cwdi_cdir = rootvnode;
    543 	VREF(cwdi0.cwdi_cdir);
    544 	VOP_UNLOCK(rootvnode, 0);
    545 	cwdi0.cwdi_rdir = NULL;
    546 
    547 	/*
    548 	 * Now that root is mounted, we can fixup initproc's CWD
    549 	 * info.  All other processes are kthreads, which merely
    550 	 * share proc0's CWD info.
    551 	 */
    552 	initproc->p_cwdi->cwdi_cdir = rootvnode;
    553 	VREF(initproc->p_cwdi->cwdi_cdir);
    554 	initproc->p_cwdi->cwdi_rdir = NULL;
    555 
    556 	/*
    557 	 * Now can look at time, having had a chance to verify the time
    558 	 * from the file system.  Reset p->p_rtime as it may have been
    559 	 * munched in mi_switch() after the time got set.
    560 	 */
    561 	proclist_lock_read();
    562 	s = splsched();
    563 	LIST_FOREACH(p, &allproc, p_list) {
    564 		KASSERT((p->p_flag & P_MARKER) == 0);
    565 		p->p_stats->p_start = mono_time = boottime = time;
    566 		LIST_FOREACH(l, &p->p_lwps, l_sibling) {
    567 			if (l->l_cpu != NULL)
    568 				l->l_cpu->ci_schedstate.spc_runtime = time;
    569 		}
    570 		p->p_rtime.tv_sec = p->p_rtime.tv_usec = 0;
    571 	}
    572 	splx(s);
    573 	proclist_unlock_read();
    574 
    575 	/* Create the pageout daemon kernel thread. */
    576 	uvm_swap_init();
    577 	if (kthread_create1(uvm_pageout, NULL, NULL, "pagedaemon"))
    578 		panic("fork pagedaemon");
    579 
    580 	/* Create the filesystem syncer kernel thread. */
    581 	if (kthread_create1(sched_sync, NULL, NULL, "ioflush"))
    582 		panic("fork syncer");
    583 
    584 	/* Create the aiodone daemon kernel thread. */
    585 	if (kthread_create1(uvm_aiodone_daemon, NULL, &uvm.aiodoned_proc,
    586 	    "aiodoned"))
    587 		panic("fork aiodoned");
    588 
    589 #if defined(MULTIPROCESSOR)
    590 	/* Boot the secondary processors. */
    591 	cpu_boot_secondary_processors();
    592 #endif
    593 
    594 	/* Initialize exec structures */
    595 	exec_init(1);
    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 void
    609 setrootfstime(time_t t)
    610 {
    611 	rootfstime = t;
    612 }
    613 
    614 static void
    615 check_console(struct proc *p)
    616 {
    617 	struct nameidata nd;
    618 	int error;
    619 
    620 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_SYSSPACE, "/dev/console", p);
    621 	error = namei(&nd);
    622 	if (error == 0)
    623 		vrele(nd.ni_vp);
    624 	else if (error == ENOENT)
    625 		printf("warning: no /dev/console\n");
    626 	else
    627 		printf("warning: lookup /dev/console: error %d\n", error);
    628 }
    629 
    630 /*
    631  * List of paths to try when searching for "init".
    632  */
    633 static const char *initpaths[] = {
    634 	"/sbin/init",
    635 	"/sbin/oinit",
    636 	"/sbin/init.bak",
    637 	NULL,
    638 };
    639 
    640 /*
    641  * Start the initial user process; try exec'ing each pathname in "initpaths".
    642  * The program is invoked with one argument containing the boot flags.
    643  */
    644 static void
    645 start_init(void *arg)
    646 {
    647 	struct lwp *l = arg;
    648 	struct proc *p = l->l_proc;
    649 	vaddr_t addr;
    650 	struct sys_execve_args /* {
    651 		syscallarg(const char *) path;
    652 		syscallarg(char * const *) argp;
    653 		syscallarg(char * const *) envp;
    654 	} */ args;
    655 	int options, i, error;
    656 	register_t retval[2];
    657 	char flags[4], *flagsp;
    658 	const char *path, *slash;
    659 	char *ucp, **uap, *arg0, *arg1 = NULL;
    660 	char ipath[129];
    661 	int ipx, len;
    662 
    663 	/*
    664 	 * Now in process 1.
    665 	 */
    666 	strncpy(p->p_comm, "init", MAXCOMLEN);
    667 
    668 	/*
    669 	 * Wait for main() to tell us that it's safe to exec.
    670 	 */
    671 	while (start_init_exec == 0)
    672 		(void) tsleep((void *)&start_init_exec, PWAIT, "initexec", 0);
    673 
    674 	/*
    675 	 * This is not the right way to do this.  We really should
    676 	 * hand-craft a descriptor onto /dev/console to hand to init,
    677 	 * but that's a _lot_ more work, and the benefit from this easy
    678 	 * hack makes up for the "good is the enemy of the best" effect.
    679 	 */
    680 	check_console(p);
    681 
    682 	/*
    683 	 * Need just enough stack to hold the faked-up "execve()" arguments.
    684 	 */
    685 	addr = (vaddr_t)STACK_ALLOC(USRSTACK, PAGE_SIZE);
    686 	if (uvm_map(&p->p_vmspace->vm_map, &addr, PAGE_SIZE,
    687                     NULL, UVM_UNKNOWN_OFFSET, 0,
    688                     UVM_MAPFLAG(UVM_PROT_ALL, UVM_PROT_ALL, UVM_INH_COPY,
    689 		    UVM_ADV_NORMAL,
    690                     UVM_FLAG_FIXED|UVM_FLAG_OVERLAY|UVM_FLAG_COPYONW)) != 0)
    691 		panic("init: couldn't allocate argument space");
    692 	p->p_vmspace->vm_maxsaddr = (caddr_t)STACK_MAX(addr, PAGE_SIZE);
    693 
    694 	ipx = 0;
    695 	while (1) {
    696 		if (boothowto & RB_ASKNAME) {
    697 			printf("init path");
    698 			if (initpaths[ipx])
    699 				printf(" (default %s)", initpaths[ipx]);
    700 			printf(": ");
    701 			len = cngetsn(ipath, sizeof(ipath)-1);
    702 			if (len == 0) {
    703 				if (initpaths[ipx])
    704 					path = initpaths[ipx++];
    705 				else
    706 					continue;
    707 			} else {
    708 				ipath[len] = '\0';
    709 				path = ipath;
    710 			}
    711 		} else {
    712 			if ((path = initpaths[ipx++]) == NULL)
    713 				break;
    714 		}
    715 
    716 		ucp = (char *)USRSTACK;
    717 
    718 		/*
    719 		 * Construct the boot flag argument.
    720 		 */
    721 		flagsp = flags;
    722 		*flagsp++ = '-';
    723 		options = 0;
    724 
    725 		if (boothowto & RB_SINGLE) {
    726 			*flagsp++ = 's';
    727 			options = 1;
    728 		}
    729 #ifdef notyet
    730 		if (boothowto & RB_FASTBOOT) {
    731 			*flagsp++ = 'f';
    732 			options = 1;
    733 		}
    734 #endif
    735 
    736 		/*
    737 		 * Move out the flags (arg 1), if necessary.
    738 		 */
    739 		if (options != 0) {
    740 			*flagsp++ = '\0';
    741 			i = flagsp - flags;
    742 #ifdef DEBUG
    743 			printf("init: copying out flags `%s' %d\n", flags, i);
    744 #endif
    745 			arg1 = STACK_ALLOC(ucp, i);
    746 			ucp = STACK_MAX(arg1, i);
    747 			(void)copyout((caddr_t)flags, arg1, i);
    748 		}
    749 
    750 		/*
    751 		 * Move out the file name (also arg 0).
    752 		 */
    753 		i = strlen(path) + 1;
    754 #ifdef DEBUG
    755 		printf("init: copying out path `%s' %d\n", path, i);
    756 #else
    757 		if (boothowto & RB_ASKNAME || path != initpaths[0])
    758 			printf("init: trying %s\n", path);
    759 #endif
    760 		arg0 = STACK_ALLOC(ucp, i);
    761 		ucp = STACK_MAX(arg0, i);
    762 		(void)copyout((caddr_t)path, arg0, i);
    763 
    764 		/*
    765 		 * Move out the arg pointers.
    766 		 */
    767 		ucp = (caddr_t)STACK_ALIGN(ucp, ALIGNBYTES);
    768 		uap = (char **)STACK_ALLOC(ucp, sizeof(char *) * 3);
    769 		SCARG(&args, path) = arg0;
    770 		SCARG(&args, argp) = uap;
    771 		SCARG(&args, envp) = NULL;
    772 		slash = strrchr(path, '/');
    773 		if (slash)
    774 			(void)suword((caddr_t)uap++,
    775 			    (long)arg0 + (slash + 1 - path));
    776 		else
    777 			(void)suword((caddr_t)uap++, (long)arg0);
    778 		if (options != 0)
    779 			(void)suword((caddr_t)uap++, (long)arg1);
    780 		(void)suword((caddr_t)uap++, 0);	/* terminator */
    781 
    782 		/*
    783 		 * Now try to exec the program.  If can't for any reason
    784 		 * other than it doesn't exist, complain.
    785 		 */
    786 		error = sys_execve(LIST_FIRST(&p->p_lwps), &args, retval);
    787 		if (error == 0 || error == EJUSTRETURN) {
    788 			KERNEL_PROC_UNLOCK(l);
    789 			return;
    790 		}
    791 		printf("exec %s: error %d\n", path, error);
    792 	}
    793 	printf("init: not found\n");
    794 	panic("no init");
    795 }
    796