Home | History | Annotate | Line # | Download | only in kern
init_main.c revision 1.236
      1 /*	$NetBSD: init_main.c,v 1.236 2004/04/25 16:42:41 simonb 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 1.236 2004/04/25 16:42:41 simonb 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 
    198 __volatile int start_init_exec;		/* semaphore for start_init() */
    199 
    200 static void check_console(struct proc *p);
    201 static void start_init(void *);
    202 void main(void);
    203 
    204 extern const struct emul emul_netbsd;	/* defined in kern_exec.c */
    205 
    206 /*
    207  * System startup; initialize the world, create process 0, mount root
    208  * filesystem, and fork to create init and pagedaemon.  Most of the
    209  * hard work is done in the lower-level initialization routines including
    210  * startup(), which does memory initialization and autoconfiguration.
    211  */
    212 void
    213 main(void)
    214 {
    215 	struct lwp *l;
    216 	struct proc *p;
    217 	struct pdevinit *pdev;
    218 	int s, error;
    219 	u_int i;
    220 	rlim_t lim;
    221 	extern struct pdevinit pdevinit[];
    222 	extern void schedcpu(void *);
    223 #if defined(NFSSERVER) || defined(NFS)
    224 	extern void nfs_init(void);
    225 #endif
    226 #ifdef NVNODE_IMPLICIT
    227 	int usevnodes;
    228 #endif
    229 
    230 	/*
    231 	 * Initialize the current LWP pointer (curlwp) before
    232 	 * any possible traps/probes to simplify trap processing.
    233 	 */
    234 	l = &lwp0;
    235 	curlwp = l;
    236 	l->l_cpu = curcpu();
    237 	l->l_proc = &proc0;
    238 	l->l_lid = 1;
    239 
    240 	/*
    241 	 * Attempt to find console and initialize
    242 	 * in case of early panic or other messages.
    243 	 */
    244 	consinit();
    245 	printf("%s", copyright);
    246 
    247 	KERNEL_LOCK_INIT();
    248 
    249 	uvm_init();
    250 
    251 	/* Do machine-dependent initialization. */
    252 	cpu_startup();
    253 
    254 	/* Initialise pools. */
    255 	link_pool_init();
    256 
    257 	/* Initialize callouts. */
    258 	callout_startup();
    259 
    260 	/* Initialize the buffer cache */
    261 	bufinit();
    262 
    263 	/*
    264 	 * Initialize mbuf's.  Do this now because we might attempt to
    265 	 * allocate mbufs or mbuf clusters during autoconfiguration.
    266 	 */
    267 	mbinit();
    268 
    269 	/* Initialize sockets. */
    270 	soinit();
    271 
    272 #ifdef KCONT
    273 	/* Initialize kcont. */
    274         kcont_init();
    275 #endif
    276 
    277 	/*
    278 	 * The following things must be done before autoconfiguration.
    279 	 */
    280 	evcnt_init();		/* initialize event counters */
    281 	disk_init();		/* initialize disk list */
    282 	tty_init();		/* initialize tty list */
    283 #if NRND > 0
    284 	rnd_init();		/* initialize RNG */
    285 #endif
    286 #if NOPENCRYPTO > 0
    287 	/* Initialize crypto subsystem before configuring crypto hardware. */
    288 	(void)crypto_init();
    289 #endif
    290 	/* Initialize the sysctl subsystem. */
    291 	sysctl_init();
    292 
    293 	/* Initialize process and pgrp structures. */
    294 	procinit();
    295 
    296 #ifdef LKM
    297 	/* Initialize the LKM system. */
    298 	lkm_init();
    299 #endif
    300 
    301 	/*
    302 	 * Create process 0 (the swapper).
    303 	 */
    304 	p = &proc0;
    305 	proc0_insert(p, l, &pgrp0, &session0);
    306 
    307 	/*
    308 	 * Set P_NOCLDWAIT so that kernel threads are reparented to
    309 	 * init(8) when they exit.  init(8) can easily wait them out
    310 	 * for us.
    311 	 */
    312 	p->p_flag = P_SYSTEM | P_NOCLDWAIT;
    313 	p->p_stat = SACTIVE;
    314 	p->p_nice = NZERO;
    315 	p->p_emul = &emul_netbsd;
    316 #ifdef __HAVE_SYSCALL_INTERN
    317 	(*p->p_emul->e_syscall_intern)(p);
    318 #endif
    319 	strncpy(p->p_comm, "swapper", MAXCOMLEN);
    320 
    321 	l->l_flag = L_INMEM;
    322 	l->l_stat = LSONPROC;
    323 	p->p_nrlwps = 1;
    324 
    325 	callout_init(&l->l_tsleep_ch);
    326 
    327 	/* Create credentials. */
    328 	cred0.p_refcnt = 1;
    329 	p->p_cred = &cred0;
    330 	p->p_ucred = crget();
    331 	p->p_ucred->cr_ngroups = 1;	/* group 0 */
    332 
    333 	/* Create the file descriptor table. */
    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 	 * 1.0% of memory for vnode cache (but not less than NVNODE vnodes).
    395 	 */
    396 	usevnodes = (ptoa((unsigned)physmem) / 100) / 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 #ifdef P1003_1B_SEMAPHORE
    426 	/* Initialize posix semaphores */
    427 	ksem_init();
    428 #endif
    429 	/* Attach pseudo-devices. */
    430 	for (pdev = pdevinit; pdev->pdev_attach != NULL; pdev++)
    431 		(*pdev->pdev_attach)(pdev->pdev_count);
    432 
    433 #ifdef	FAST_IPSEC
    434 	/* Attach network crypto subsystem */
    435 	ipsec_attach();
    436 #endif
    437 
    438 	/*
    439 	 * Initialize protocols.  Block reception of incoming packets
    440 	 * until everything is ready.
    441 	 */
    442 	s = splnet();
    443 	ifinit();
    444 	domaininit();
    445 	if_attachdomain();
    446 	splx(s);
    447 
    448 #ifdef GPROF
    449 	/* Initialize kernel profiling. */
    450 	kmstartup();
    451 #endif
    452 
    453 	/* Initialize system accouting. */
    454 	acct_init();
    455 
    456 #ifdef SYSTRACE
    457 	systrace_init();
    458 #endif
    459 	/*
    460 	 * Initialize signal-related data structures, and signal state
    461 	 * for proc0.
    462 	 */
    463 	signal_init();
    464 	p->p_sigacts = &sigacts0;
    465 	siginit(p);
    466 
    467 	/* Kick off timeout driven events by calling first time. */
    468 	schedcpu(NULL);
    469 
    470 	/*
    471 	 * Create process 1 (init(8)).  We do this now, as Unix has
    472 	 * historically had init be process 1, and changing this would
    473 	 * probably upset a lot of people.
    474 	 *
    475 	 * Note that process 1 won't immediately exec init(8), but will
    476 	 * wait for us to inform it that the root file system has been
    477 	 * mounted.
    478 	 */
    479 	if (fork1(l, 0, SIGCHLD, NULL, 0, start_init, NULL, NULL, &initproc))
    480 		panic("fork init");
    481 
    482 	/*
    483 	 * Create any kernel threads who's creation was deferred because
    484 	 * initproc had not yet been created.
    485 	 */
    486 	kthread_run_deferred_queue();
    487 
    488 	/*
    489 	 * Now that device driver threads have been created, wait for
    490 	 * them to finish any deferred autoconfiguration.  Note we don't
    491 	 * need to lock this semaphore, since we haven't booted any
    492 	 * secondary processors, yet.
    493 	 */
    494 	while (config_pending)
    495 		(void) tsleep((void *)&config_pending, PWAIT, "cfpend", 0);
    496 
    497 	/*
    498 	 * Finalize configuration now that all real devices have been
    499 	 * found.  This needs to be done before the root device is
    500 	 * selected, since finalization may create the root device.
    501 	 */
    502 	config_finalize();
    503 
    504 	/*
    505 	 * Now that autoconfiguration has completed, we can determine
    506 	 * the root and dump devices.
    507 	 */
    508 	cpu_rootconf();
    509 	cpu_dumpconf();
    510 
    511 	/* Mount the root file system. */
    512 	do {
    513 		domountroothook();
    514 		if ((error = vfs_mountroot())) {
    515 			printf("cannot mount root, error = %d\n", error);
    516 			boothowto |= RB_ASKNAME;
    517 			setroot(root_device,
    518 			    (rootdev != NODEV) ? DISKPART(rootdev) : 0);
    519 		}
    520 	} while (error != 0);
    521 	mountroothook_destroy();
    522 
    523 	CIRCLEQ_FIRST(&mountlist)->mnt_flag |= MNT_ROOTFS;
    524 	CIRCLEQ_FIRST(&mountlist)->mnt_op->vfs_refcount++;
    525 
    526 	/*
    527 	 * Get the vnode for '/'.  Set filedesc0.fd_fd.fd_cdir to
    528 	 * reference it.
    529 	 */
    530 	if (VFS_ROOT(CIRCLEQ_FIRST(&mountlist), &rootvnode))
    531 		panic("cannot find root vnode");
    532 	cwdi0.cwdi_cdir = rootvnode;
    533 	VREF(cwdi0.cwdi_cdir);
    534 	VOP_UNLOCK(rootvnode, 0);
    535 	cwdi0.cwdi_rdir = NULL;
    536 
    537 	/*
    538 	 * Now that root is mounted, we can fixup initproc's CWD
    539 	 * info.  All other processes are kthreads, which merely
    540 	 * share proc0's CWD info.
    541 	 */
    542 	initproc->p_cwdi->cwdi_cdir = rootvnode;
    543 	VREF(initproc->p_cwdi->cwdi_cdir);
    544 	initproc->p_cwdi->cwdi_rdir = NULL;
    545 
    546 	/*
    547 	 * Now can look at time, having had a chance to verify the time
    548 	 * from the file system.  Reset p->p_rtime as it may have been
    549 	 * munched in mi_switch() after the time got set.
    550 	 */
    551 	proclist_lock_read();
    552 	s = splsched();
    553 	LIST_FOREACH(p, &allproc, p_list) {
    554 		p->p_stats->p_start = mono_time = boottime = time;
    555 		LIST_FOREACH(l, &p->p_lwps, l_sibling) {
    556 			if (l->l_cpu != NULL)
    557 				l->l_cpu->ci_schedstate.spc_runtime = time;
    558 		}
    559 		p->p_rtime.tv_sec = p->p_rtime.tv_usec = 0;
    560 	}
    561 	splx(s);
    562 	proclist_unlock_read();
    563 
    564 	/* Create the pageout daemon kernel thread. */
    565 	uvm_swap_init();
    566 	if (kthread_create1(uvm_pageout, NULL, NULL, "pagedaemon"))
    567 		panic("fork pagedaemon");
    568 
    569 	/* Create the filesystem syncer kernel thread. */
    570 	if (kthread_create1(sched_sync, NULL, NULL, "ioflush"))
    571 		panic("fork syncer");
    572 
    573 	/* Create the aiodone daemon kernel thread. */
    574 	if (kthread_create1(uvm_aiodone_daemon, NULL, &uvm.aiodoned_proc,
    575 	    "aiodoned"))
    576 		panic("fork aiodoned");
    577 
    578 #if defined(MULTIPROCESSOR)
    579 	/* Boot the secondary processors. */
    580 	cpu_boot_secondary_processors();
    581 #endif
    582 
    583 	/* Initialize exec structures */
    584 	exec_init(1);
    585 
    586 	/*
    587 	 * Okay, now we can let init(8) exec!  It's off to userland!
    588 	 */
    589 	start_init_exec = 1;
    590 	wakeup((void *)&start_init_exec);
    591 
    592 	/* The scheduler is an infinite loop. */
    593 	uvm_scheduler();
    594 	/* NOTREACHED */
    595 }
    596 
    597 static void
    598 check_console(struct proc *p)
    599 {
    600 	struct nameidata nd;
    601 	int error;
    602 
    603 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_SYSSPACE, "/dev/console", p);
    604 	error = namei(&nd);
    605 	if (error == 0)
    606 		vrele(nd.ni_vp);
    607 	else if (error == ENOENT)
    608 		printf("warning: no /dev/console\n");
    609 	else
    610 		printf("warning: lookup /dev/console: error %d\n", error);
    611 }
    612 
    613 /*
    614  * List of paths to try when searching for "init".
    615  */
    616 static const char *initpaths[] = {
    617 	"/sbin/init",
    618 	"/sbin/oinit",
    619 	"/sbin/init.bak",
    620 	NULL,
    621 };
    622 
    623 /*
    624  * Start the initial user process; try exec'ing each pathname in "initpaths".
    625  * The program is invoked with one argument containing the boot flags.
    626  */
    627 static void
    628 start_init(void *arg)
    629 {
    630 	struct lwp *l = arg;
    631 	struct proc *p = l->l_proc;
    632 	vaddr_t addr;
    633 	struct sys_execve_args /* {
    634 		syscallarg(const char *) path;
    635 		syscallarg(char * const *) argp;
    636 		syscallarg(char * const *) envp;
    637 	} */ args;
    638 	int options, i, error;
    639 	register_t retval[2];
    640 	char flags[4], *flagsp;
    641 	const char *path, *slash;
    642 	char *ucp, **uap, *arg0, *arg1 = NULL;
    643 	char ipath[129];
    644 	int ipx, len;
    645 
    646 	/*
    647 	 * Now in process 1.
    648 	 */
    649 	strncpy(p->p_comm, "init", MAXCOMLEN);
    650 
    651 	/*
    652 	 * Wait for main() to tell us that it's safe to exec.
    653 	 */
    654 	while (start_init_exec == 0)
    655 		(void) tsleep((void *)&start_init_exec, PWAIT, "initexec", 0);
    656 
    657 	/*
    658 	 * This is not the right way to do this.  We really should
    659 	 * hand-craft a descriptor onto /dev/console to hand to init,
    660 	 * but that's a _lot_ more work, and the benefit from this easy
    661 	 * hack makes up for the "good is the enemy of the best" effect.
    662 	 */
    663 	check_console(p);
    664 
    665 	/*
    666 	 * Need just enough stack to hold the faked-up "execve()" arguments.
    667 	 */
    668 	addr = (vaddr_t)STACK_ALLOC(USRSTACK, PAGE_SIZE);
    669 	if (uvm_map(&p->p_vmspace->vm_map, &addr, PAGE_SIZE,
    670                     NULL, UVM_UNKNOWN_OFFSET, 0,
    671                     UVM_MAPFLAG(UVM_PROT_ALL, UVM_PROT_ALL, UVM_INH_COPY,
    672 		    UVM_ADV_NORMAL,
    673                     UVM_FLAG_FIXED|UVM_FLAG_OVERLAY|UVM_FLAG_COPYONW)) != 0)
    674 		panic("init: couldn't allocate argument space");
    675 	p->p_vmspace->vm_maxsaddr = (caddr_t)STACK_MAX(addr, PAGE_SIZE);
    676 
    677 	ipx = 0;
    678 	while (1) {
    679 		if (boothowto & RB_ASKNAME) {
    680 			printf("init path");
    681 			if (initpaths[ipx])
    682 				printf(" (default %s)", initpaths[ipx]);
    683 			printf(": ");
    684 			len = cngetsn(ipath, sizeof(ipath)-1);
    685 			if (len == 0) {
    686 				if (initpaths[ipx])
    687 					path = initpaths[ipx++];
    688 				else
    689 					continue;
    690 			} else {
    691 				ipath[len] = '\0';
    692 				path = ipath;
    693 			}
    694 		} else {
    695 			if ((path = initpaths[ipx++]) == NULL)
    696 				break;
    697 		}
    698 
    699 		ucp = (char *)USRSTACK;
    700 
    701 		/*
    702 		 * Construct the boot flag argument.
    703 		 */
    704 		flagsp = flags;
    705 		*flagsp++ = '-';
    706 		options = 0;
    707 
    708 		if (boothowto & RB_SINGLE) {
    709 			*flagsp++ = 's';
    710 			options = 1;
    711 		}
    712 #ifdef notyet
    713 		if (boothowto & RB_FASTBOOT) {
    714 			*flagsp++ = 'f';
    715 			options = 1;
    716 		}
    717 #endif
    718 
    719 		/*
    720 		 * Move out the flags (arg 1), if necessary.
    721 		 */
    722 		if (options != 0) {
    723 			*flagsp++ = '\0';
    724 			i = flagsp - flags;
    725 #ifdef DEBUG
    726 			printf("init: copying out flags `%s' %d\n", flags, i);
    727 #endif
    728 			arg1 = STACK_ALLOC(ucp, i);
    729 			ucp = STACK_MAX(arg1, i);
    730 			(void)copyout((caddr_t)flags, arg1, i);
    731 		}
    732 
    733 		/*
    734 		 * Move out the file name (also arg 0).
    735 		 */
    736 		i = strlen(path) + 1;
    737 #ifdef DEBUG
    738 		printf("init: copying out path `%s' %d\n", path, i);
    739 #else
    740 		if (boothowto & RB_ASKNAME || path != initpaths[0])
    741 			printf("init: trying %s\n", path);
    742 #endif
    743 		arg0 = STACK_ALLOC(ucp, i);
    744 		ucp = STACK_MAX(arg0, i);
    745 		(void)copyout((caddr_t)path, arg0, i);
    746 
    747 		/*
    748 		 * Move out the arg pointers.
    749 		 */
    750 		ucp = (caddr_t)STACK_ALIGN(ucp, ALIGNBYTES);
    751 		uap = (char **)STACK_ALLOC(ucp, sizeof(char *) * 3);
    752 		SCARG(&args, path) = arg0;
    753 		SCARG(&args, argp) = uap;
    754 		SCARG(&args, envp) = NULL;
    755 		slash = strrchr(path, '/');
    756 		if (slash)
    757 			(void)suword((caddr_t)uap++,
    758 			    (long)arg0 + (slash + 1 - path));
    759 		else
    760 			(void)suword((caddr_t)uap++, (long)arg0);
    761 		if (options != 0)
    762 			(void)suword((caddr_t)uap++, (long)arg1);
    763 		(void)suword((caddr_t)uap++, 0);	/* terminator */
    764 
    765 		/*
    766 		 * Now try to exec the program.  If can't for any reason
    767 		 * other than it doesn't exist, complain.
    768 		 */
    769 		error = sys_execve(LIST_FIRST(&p->p_lwps), &args, retval);
    770 		if (error == 0 || error == EJUSTRETURN) {
    771 			KERNEL_PROC_UNLOCK(l);
    772 			return;
    773 		}
    774 		printf("exec %s: error %d\n", path, error);
    775 	}
    776 	printf("init: not found\n");
    777 	panic("no init");
    778 }
    779