Home | History | Annotate | Line # | Download | only in kern
init_main.c revision 1.232
      1 /*	$NetBSD: init_main.c,v 1.232 2004/01/09 00:04:53 tls 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 1.232 2004/01/09 00:04:53 tls Exp $");
     75 
     76 #include "fs_nfs.h"
     77 #include "opt_nfsserver.h"
     78 #include "opt_ipsec.h"
     79 #include "opt_sysv.h"
     80 #include "opt_maxuprc.h"
     81 #include "opt_multiprocessor.h"
     82 #include "opt_pipe.h"
     83 #include "opt_syscall_debug.h"
     84 #include "opt_systrace.h"
     85 #include "opt_posix.h"
     86 
     87 #include "opencrypto.h"
     88 #include "rnd.h"
     89 
     90 #include <sys/param.h>
     91 #include <sys/acct.h>
     92 #include <sys/filedesc.h>
     93 #include <sys/file.h>
     94 #include <sys/errno.h>
     95 #include <sys/callout.h>
     96 #include <sys/kernel.h>
     97 #include <sys/mount.h>
     98 #include <sys/proc.h>
     99 #include <sys/kthread.h>
    100 #include <sys/resourcevar.h>
    101 #include <sys/signalvar.h>
    102 #include <sys/systm.h>
    103 #include <sys/vnode.h>
    104 #include <sys/tty.h>
    105 #include <sys/conf.h>
    106 #include <sys/disklabel.h>
    107 #include <sys/buf.h>
    108 #include <sys/device.h>
    109 #include <sys/disk.h>
    110 #include <sys/exec.h>
    111 #include <sys/socketvar.h>
    112 #include <sys/protosw.h>
    113 #include <sys/reboot.h>
    114 #include <sys/user.h>
    115 #include <sys/sysctl.h>
    116 #include <sys/event.h>
    117 #include <sys/mbuf.h>
    118 #ifdef	FAST_IPSEC
    119 #include <netipsec/ipsec.h>
    120 #endif
    121 #ifdef SYSVSHM
    122 #include <sys/shm.h>
    123 #endif
    124 #ifdef SYSVSEM
    125 #include <sys/sem.h>
    126 #endif
    127 #ifdef SYSVMSG
    128 #include <sys/msg.h>
    129 #endif
    130 #ifdef P1003_1B_SEMAPHORE
    131 #include <sys/ksem.h>
    132 #endif
    133 #ifdef SYSTRACE
    134 #include <sys/systrace.h>
    135 #endif
    136 #include <sys/domain.h>
    137 #include <sys/namei.h>
    138 #if NOPENCRYPTO > 0
    139 #include <opencrypto/cryptodev.h>	/* XXX really  the framework */
    140 #endif
    141 #if NRND > 0
    142 #include <sys/rnd.h>
    143 #endif
    144 #ifndef PIPE_SOCKETPAIR
    145 #include <sys/pipe.h>
    146 #endif
    147 #ifdef LKM
    148 #include <sys/lkm.h>
    149 #endif
    150 
    151 #include <sys/syscall.h>
    152 #include <sys/sa.h>
    153 #include <sys/syscallargs.h>
    154 
    155 #include <ufs/ufs/quota.h>
    156 
    157 #include <miscfs/genfs/genfs.h>
    158 #include <miscfs/syncfs/syncfs.h>
    159 
    160 #include <machine/cpu.h>
    161 
    162 #include <uvm/uvm.h>
    163 
    164 #include <dev/cons.h>
    165 
    166 #include <net/if.h>
    167 #include <net/raw_cb.h>
    168 
    169 /* Components of the first process -- never freed. */
    170 struct	session session0;
    171 struct	pgrp pgrp0;
    172 struct	proc proc0;
    173 struct	lwp lwp0;
    174 struct	pcred cred0;
    175 struct	filedesc0 filedesc0;
    176 struct	cwdinfo cwdi0;
    177 struct	plimit limit0;
    178 struct	pstats pstat0;
    179 struct	vmspace vmspace0;
    180 struct	sigacts sigacts0;
    181 #ifndef curlwp
    182 struct	lwp *curlwp = &lwp0;
    183 #endif
    184 struct	proc *initproc;
    185 
    186 int	nofile = NOFILE;
    187 int	maxuprc = MAXUPRC;
    188 int	cmask = CMASK;
    189 extern	struct user *proc0paddr;
    190 
    191 struct	vnode *rootvp, *swapdev_vp;
    192 int	boothowto;
    193 int	cold = 1;			/* still working on startup */
    194 struct	timeval boottime;
    195 
    196 __volatile int start_init_exec;		/* semaphore for start_init() */
    197 
    198 static void check_console(struct proc *p);
    199 static void start_init(void *);
    200 void main(void);
    201 
    202 extern const struct emul emul_netbsd;	/* defined in kern_exec.c */
    203 
    204 /*
    205  * System startup; initialize the world, create process 0, mount root
    206  * filesystem, and fork to create init and pagedaemon.  Most of the
    207  * hard work is done in the lower-level initialization routines including
    208  * startup(), which does memory initialization and autoconfiguration.
    209  */
    210 void
    211 main(void)
    212 {
    213 	struct lwp *l;
    214 	struct proc *p;
    215 	struct pdevinit *pdev;
    216 	int s, error;
    217 	u_int i;
    218 	rlim_t lim;
    219 	extern struct pdevinit pdevinit[];
    220 	extern void schedcpu(void *);
    221 #if defined(NFSSERVER) || defined(NFS)
    222 	extern void nfs_init(void);
    223 #endif
    224 #ifdef NVNODE_IMPLICIT
    225 	int usevnodes;
    226 #endif
    227 
    228 	/*
    229 	 * Initialize the current LWP pointer (curlwp) before
    230 	 * any possible traps/probes to simplify trap processing.
    231 	 */
    232 	l = &lwp0;
    233 	curlwp = l;
    234 	l->l_cpu = curcpu();
    235 	l->l_proc = &proc0;
    236 	l->l_lid = 1;
    237 	/*
    238 	 * Attempt to find console and initialize
    239 	 * in case of early panic or other messages.
    240 	 */
    241 	consinit();
    242 	printf("%s", copyright);
    243 
    244 	KERNEL_LOCK_INIT();
    245 
    246 	uvm_init();
    247 
    248 	/* Do machine-dependent initialization. */
    249 	cpu_startup();
    250 
    251 	/* Initialize callouts. */
    252 	callout_startup();
    253 
    254 	/* Initialize the buffer cache */
    255 	bufinit();
    256 
    257 	/*
    258 	 * Initialize mbuf's.  Do this now because we might attempt to
    259 	 * allocate mbufs or mbuf clusters during autoconfiguration.
    260 	 */
    261 	mbinit();
    262 
    263 	/* Initialize kqueues. */
    264 	kqueue_init();
    265 
    266 	/* Initialize sockets. */
    267 	soinit();
    268 
    269 	/*
    270 	 * The following things must be done before autoconfiguration.
    271 	 */
    272 	evcnt_init();		/* initialize event counters */
    273 	disk_init();		/* initialize disk list */
    274 	tty_init();		/* initialize tty list */
    275 #if NRND > 0
    276 	rnd_init();		/* initialize RNG */
    277 #endif
    278 #if NOPENCRYPTO > 0
    279 	/* Initialize crypto subsystem before configuring crypto hardware. */
    280 	(void)crypto_init();
    281 #endif
    282 	/* Initialize the sysctl subsystem. */
    283 	sysctl_init();
    284 
    285 	/*
    286 	 * Initialize process and pgrp structures.
    287 	 */
    288 	procinit();
    289 
    290 #ifdef LKM
    291 	/* Initialize the LKM system. */
    292 	lkm_init();
    293 #endif
    294 
    295 	/*
    296 	 * Create process 0 (the swapper).
    297 	 */
    298 	p = &proc0;
    299 	proc0_insert(p, l, &pgrp0, &session0);
    300 
    301 	/*
    302 	 * Set P_NOCLDWAIT so that kernel threads are reparented to
    303 	 * init(8) when they exit.  init(8) can easily wait them out
    304 	 * for us.
    305 	 */
    306 	p->p_flag = P_SYSTEM | P_NOCLDWAIT;
    307 	p->p_stat = SACTIVE;
    308 	p->p_nice = NZERO;
    309 	p->p_emul = &emul_netbsd;
    310 #ifdef __HAVE_SYSCALL_INTERN
    311 	(*p->p_emul->e_syscall_intern)(p);
    312 #endif
    313 	strncpy(p->p_comm, "swapper", MAXCOMLEN);
    314 
    315 	l->l_flag = L_INMEM;
    316 	l->l_stat = LSONPROC;
    317 	p->p_nrlwps = 1;
    318 
    319 	callout_init(&l->l_tsleep_ch);
    320 
    321 	/* Create credentials. */
    322 	cred0.p_refcnt = 1;
    323 	p->p_cred = &cred0;
    324 	p->p_ucred = crget();
    325 	p->p_ucred->cr_ngroups = 1;	/* group 0 */
    326 
    327 	/* Create the file descriptor table. */
    328 	finit();
    329 	p->p_fd = &filedesc0.fd_fd;
    330 	fdinit1(&filedesc0);
    331 
    332 	/* Create the CWD info. */
    333 	p->p_cwdi = &cwdi0;
    334 	cwdi0.cwdi_cmask = cmask;
    335 	cwdi0.cwdi_refcnt = 1;
    336 
    337 	/* Create the limits structures. */
    338 	p->p_limit = &limit0;
    339 	for (i = 0; i < sizeof(p->p_rlimit)/sizeof(p->p_rlimit[0]); i++)
    340 		limit0.pl_rlimit[i].rlim_cur =
    341 		    limit0.pl_rlimit[i].rlim_max = RLIM_INFINITY;
    342 
    343 	limit0.pl_rlimit[RLIMIT_NOFILE].rlim_max = maxfiles;
    344 	limit0.pl_rlimit[RLIMIT_NOFILE].rlim_cur =
    345 	    maxfiles < nofile ? maxfiles : nofile;
    346 
    347 	limit0.pl_rlimit[RLIMIT_NPROC].rlim_max = maxproc;
    348 	limit0.pl_rlimit[RLIMIT_NPROC].rlim_cur =
    349 	    maxproc < maxuprc ? maxproc : maxuprc;
    350 
    351 	lim = ptoa(uvmexp.free);
    352 	limit0.pl_rlimit[RLIMIT_RSS].rlim_max = lim;
    353 	limit0.pl_rlimit[RLIMIT_MEMLOCK].rlim_max = lim;
    354 	limit0.pl_rlimit[RLIMIT_MEMLOCK].rlim_cur = lim / 3;
    355 	limit0.pl_corename = defcorename;
    356 	limit0.p_refcnt = 1;
    357 
    358 	/*
    359 	 * Initialize proc0's vmspace, which uses the kernel pmap.
    360 	 * All kernel processes (which never have user space mappings)
    361 	 * share proc0's vmspace, and thus, the kernel pmap.
    362 	 */
    363 	uvmspace_init(&vmspace0, pmap_kernel(), round_page(VM_MIN_ADDRESS),
    364 	    trunc_page(VM_MAX_ADDRESS));
    365 	p->p_vmspace = &vmspace0;
    366 
    367 	l->l_addr = proc0paddr;				/* XXX */
    368 
    369 	p->p_stats = &pstat0;
    370 
    371 	/*
    372 	 * Charge root for one process.
    373 	 */
    374 	(void)chgproccnt(0, 1);
    375 
    376 	rqinit();
    377 
    378 	/* Configure virtual memory system, set vm rlimits. */
    379 	uvm_init_limits(p);
    380 
    381 	/* Initialize the file systems. */
    382 #if defined(NFSSERVER) || defined(NFS)
    383 	nfs_init();			/* initialize server/shared data */
    384 #endif
    385 #ifdef NVNODE_IMPLICIT
    386 	/*
    387 	 * If maximum number of vnodes in namei vnode cache is not explicitly
    388 	 * defined in kernel config, adjust the number such as we use roughly
    389 	 * 1.0% of memory for vnode cache (but not less than NVNODE vnodes).
    390 	 */
    391 	usevnodes = (ptoa((unsigned)physmem) / 100) / sizeof(struct vnode);
    392 	if (usevnodes > desiredvnodes)
    393 		desiredvnodes = usevnodes;
    394 #endif
    395 	vfsinit();
    396 
    397 	/* Configure the system hardware.  This will enable interrupts. */
    398 	configure();
    399 
    400 	ubc_init();		/* must be after autoconfig */
    401 
    402 	/* Lock the kernel on behalf of proc0. */
    403 	KERNEL_PROC_LOCK(l);
    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 
    421 #ifdef P1003_1B_SEMAPHORE
    422 	/* Initialize posix semaphores */
    423 	ksem_init();
    424 #endif
    425 	/* Attach pseudo-devices. */
    426 	for (pdev = pdevinit; pdev->pdev_attach != NULL; pdev++)
    427 		(*pdev->pdev_attach)(pdev->pdev_count);
    428 
    429 #ifdef	FAST_IPSEC
    430 	/* Attach network crypto subsystem */
    431 	ipsec_attach();
    432 #endif
    433 
    434 	/*
    435 	 * Initialize protocols.  Block reception of incoming packets
    436 	 * until everything is ready.
    437 	 */
    438 	s = splnet();
    439 	ifinit();
    440 	domaininit();
    441 	if_attachdomain();
    442 	splx(s);
    443 
    444 #ifdef GPROF
    445 	/* Initialize kernel profiling. */
    446 	kmstartup();
    447 #endif
    448 
    449 	/* Initialize system accouting. */
    450 	acct_init();
    451 
    452 #ifdef SYSTRACE
    453 	systrace_init();
    454 #endif
    455 	/*
    456 	 * Initialize signal-related data structures, and signal state
    457 	 * for proc0.
    458 	 */
    459 	signal_init();
    460 	p->p_sigacts = &sigacts0;
    461 	siginit(p);
    462 
    463 	/* Kick off timeout driven events by calling first time. */
    464 	schedcpu(NULL);
    465 
    466 	/*
    467 	 * Create process 1 (init(8)).  We do this now, as Unix has
    468 	 * historically had init be process 1, and changing this would
    469 	 * probably upset a lot of people.
    470 	 *
    471 	 * Note that process 1 won't immediately exec init(8), but will
    472 	 * wait for us to inform it that the root file system has been
    473 	 * mounted.
    474 	 */
    475 	if (fork1(l, 0, SIGCHLD, NULL, 0, start_init, NULL, NULL, &initproc))
    476 		panic("fork init");
    477 
    478 	/*
    479 	 * Create any kernel threads who's creation was deferred because
    480 	 * initproc had not yet been created.
    481 	 */
    482 	kthread_run_deferred_queue();
    483 
    484 	/*
    485 	 * Now that device driver threads have been created, wait for
    486 	 * them to finish any deferred autoconfiguration.  Note we don't
    487 	 * need to lock this semaphore, since we haven't booted any
    488 	 * secondary processors, yet.
    489 	 */
    490 	while (config_pending)
    491 		(void) tsleep((void *)&config_pending, PWAIT, "cfpend", 0);
    492 
    493 	/*
    494 	 * Finalize configuration now that all real devices have been
    495 	 * found.  This needs to be done before the root device is
    496 	 * selected, since finalization may create the root device.
    497 	 */
    498 	config_finalize();
    499 
    500 	/*
    501 	 * Now that autoconfiguration has completed, we can determine
    502 	 * the root and dump devices.
    503 	 */
    504 	cpu_rootconf();
    505 	cpu_dumpconf();
    506 
    507 	/* Mount the root file system. */
    508 	do {
    509 		domountroothook();
    510 		if ((error = vfs_mountroot())) {
    511 			printf("cannot mount root, error = %d\n", error);
    512 			boothowto |= RB_ASKNAME;
    513 			setroot(root_device,
    514 			    (rootdev != NODEV) ? DISKPART(rootdev) : 0);
    515 		}
    516 	} while (error != 0);
    517 	mountroothook_destroy();
    518 
    519 	CIRCLEQ_FIRST(&mountlist)->mnt_flag |= MNT_ROOTFS;
    520 	CIRCLEQ_FIRST(&mountlist)->mnt_op->vfs_refcount++;
    521 
    522 	/*
    523 	 * Get the vnode for '/'.  Set filedesc0.fd_fd.fd_cdir to
    524 	 * reference it.
    525 	 */
    526 	if (VFS_ROOT(CIRCLEQ_FIRST(&mountlist), &rootvnode))
    527 		panic("cannot find root vnode");
    528 	cwdi0.cwdi_cdir = rootvnode;
    529 	VREF(cwdi0.cwdi_cdir);
    530 	VOP_UNLOCK(rootvnode, 0);
    531 	cwdi0.cwdi_rdir = NULL;
    532 
    533 	/*
    534 	 * Now that root is mounted, we can fixup initproc's CWD
    535 	 * info.  All other processes are kthreads, which merely
    536 	 * share proc0's CWD info.
    537 	 */
    538 	initproc->p_cwdi->cwdi_cdir = rootvnode;
    539 	VREF(initproc->p_cwdi->cwdi_cdir);
    540 	initproc->p_cwdi->cwdi_rdir = NULL;
    541 
    542 	/*
    543 	 * Now can look at time, having had a chance to verify the time
    544 	 * from the file system.  Reset p->p_rtime as it may have been
    545 	 * munched in mi_switch() after the time got set.
    546 	 */
    547 	proclist_lock_read();
    548 	s = splsched();
    549 	LIST_FOREACH(p, &allproc, p_list) {
    550 		p->p_stats->p_start = mono_time = boottime = time;
    551 		LIST_FOREACH(l, &p->p_lwps, l_sibling) {
    552 			if (l->l_cpu != NULL)
    553 				l->l_cpu->ci_schedstate.spc_runtime = time;
    554 		}
    555 		p->p_rtime.tv_sec = p->p_rtime.tv_usec = 0;
    556 	}
    557 	splx(s);
    558 	proclist_unlock_read();
    559 
    560 	/* Create the pageout daemon kernel thread. */
    561 	uvm_swap_init();
    562 	if (kthread_create1(uvm_pageout, NULL, NULL, "pagedaemon"))
    563 		panic("fork pagedaemon");
    564 
    565 	/* Create the filesystem syncer kernel thread. */
    566 	if (kthread_create1(sched_sync, NULL, NULL, "ioflush"))
    567 		panic("fork syncer");
    568 
    569 	/* Create the aiodone daemon kernel thread. */
    570 	if (kthread_create1(uvm_aiodone_daemon, NULL, &uvm.aiodoned_proc,
    571 	    "aiodoned"))
    572 		panic("fork aiodoned");
    573 
    574 #if defined(MULTIPROCESSOR)
    575 	/* Boot the secondary processors. */
    576 	cpu_boot_secondary_processors();
    577 #endif
    578 
    579 	/* Initialize exec structures */
    580 	exec_init(1);
    581 
    582 #ifndef PIPE_SOCKETPAIR
    583 	/* Initialize pipe structures */
    584 	pipe_init();
    585 #endif
    586 
    587 	/*
    588 	 * Okay, now we can let init(8) exec!  It's off to userland!
    589 	 */
    590 	start_init_exec = 1;
    591 	wakeup((void *)&start_init_exec);
    592 
    593 	/* The scheduler is an infinite loop. */
    594 	uvm_scheduler();
    595 	/* NOTREACHED */
    596 }
    597 
    598 static void
    599 check_console(struct proc *p)
    600 {
    601 	struct nameidata nd;
    602 	int error;
    603 
    604 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_SYSSPACE, "/dev/console", p);
    605 	error = namei(&nd);
    606 	if (error == 0)
    607 		vrele(nd.ni_vp);
    608 	else if (error == ENOENT)
    609 		printf("warning: no /dev/console\n");
    610 	else
    611 		printf("warning: lookup /dev/console: error %d\n", error);
    612 }
    613 
    614 /*
    615  * List of paths to try when searching for "init".
    616  */
    617 static const char *initpaths[] = {
    618 	"/sbin/init",
    619 	"/sbin/oinit",
    620 	"/sbin/init.bak",
    621 	NULL,
    622 };
    623 
    624 /*
    625  * Start the initial user process; try exec'ing each pathname in "initpaths".
    626  * The program is invoked with one argument containing the boot flags.
    627  */
    628 static void
    629 start_init(void *arg)
    630 {
    631 	struct lwp *l = arg;
    632 	struct proc *p = l->l_proc;
    633 	vaddr_t addr;
    634 	struct sys_execve_args /* {
    635 		syscallarg(const char *) path;
    636 		syscallarg(char * const *) argp;
    637 		syscallarg(char * const *) envp;
    638 	} */ args;
    639 	int options, i, error;
    640 	register_t retval[2];
    641 	char flags[4], *flagsp;
    642 	const char *path, *slash;
    643 	char *ucp, **uap, *arg0, *arg1 = NULL;
    644 	char ipath[129];
    645 	int ipx, len;
    646 
    647 	/*
    648 	 * Now in process 1.
    649 	 */
    650 	strncpy(p->p_comm, "init", MAXCOMLEN);
    651 
    652 	/*
    653 	 * Wait for main() to tell us that it's safe to exec.
    654 	 */
    655 	while (start_init_exec == 0)
    656 		(void) tsleep((void *)&start_init_exec, PWAIT, "initexec", 0);
    657 
    658 	/*
    659 	 * This is not the right way to do this.  We really should
    660 	 * hand-craft a descriptor onto /dev/console to hand to init,
    661 	 * but that's a _lot_ more work, and the benefit from this easy
    662 	 * hack makes up for the "good is the enemy of the best" effect.
    663 	 */
    664 	check_console(p);
    665 
    666 	/*
    667 	 * Need just enough stack to hold the faked-up "execve()" arguments.
    668 	 */
    669 	addr = (vaddr_t)STACK_ALLOC(USRSTACK, PAGE_SIZE);
    670 	if (uvm_map(&p->p_vmspace->vm_map, &addr, PAGE_SIZE,
    671                     NULL, UVM_UNKNOWN_OFFSET, 0,
    672                     UVM_MAPFLAG(UVM_PROT_ALL, UVM_PROT_ALL, UVM_INH_COPY,
    673 		    UVM_ADV_NORMAL,
    674                     UVM_FLAG_FIXED|UVM_FLAG_OVERLAY|UVM_FLAG_COPYONW)) != 0)
    675 		panic("init: couldn't allocate argument space");
    676 	p->p_vmspace->vm_maxsaddr = (caddr_t)STACK_MAX(addr, PAGE_SIZE);
    677 
    678 	ipx = 0;
    679 	while (1) {
    680 		if (boothowto & RB_ASKNAME) {
    681 			printf("init path");
    682 			if (initpaths[ipx])
    683 				printf(" (default %s)", initpaths[ipx]);
    684 			printf(": ");
    685 			len = cngetsn(ipath, sizeof(ipath)-1);
    686 			if (len == 0) {
    687 				if (initpaths[ipx])
    688 					path = initpaths[ipx++];
    689 				else
    690 					continue;
    691 			} else {
    692 				ipath[len] = '\0';
    693 				path = ipath;
    694 			}
    695 		} else {
    696 			if ((path = initpaths[ipx++]) == NULL)
    697 				break;
    698 		}
    699 
    700 		ucp = (char *)USRSTACK;
    701 
    702 		/*
    703 		 * Construct the boot flag argument.
    704 		 */
    705 		flagsp = flags;
    706 		*flagsp++ = '-';
    707 		options = 0;
    708 
    709 		if (boothowto & RB_SINGLE) {
    710 			*flagsp++ = 's';
    711 			options = 1;
    712 		}
    713 #ifdef notyet
    714 		if (boothowto & RB_FASTBOOT) {
    715 			*flagsp++ = 'f';
    716 			options = 1;
    717 		}
    718 #endif
    719 
    720 		/*
    721 		 * Move out the flags (arg 1), if necessary.
    722 		 */
    723 		if (options != 0) {
    724 			*flagsp++ = '\0';
    725 			i = flagsp - flags;
    726 #ifdef DEBUG
    727 			printf("init: copying out flags `%s' %d\n", flags, i);
    728 #endif
    729 			arg1 = STACK_ALLOC(ucp, i);
    730 			ucp = STACK_MAX(arg1, i);
    731 			(void)copyout((caddr_t)flags, arg1, i);
    732 		}
    733 
    734 		/*
    735 		 * Move out the file name (also arg 0).
    736 		 */
    737 		i = strlen(path) + 1;
    738 #ifdef DEBUG
    739 		printf("init: copying out path `%s' %d\n", path, i);
    740 #else
    741 		if (boothowto & RB_ASKNAME || path != initpaths[0])
    742 			printf("init: trying %s\n", path);
    743 #endif
    744 		arg0 = STACK_ALLOC(ucp, i);
    745 		ucp = STACK_MAX(arg0, i);
    746 		(void)copyout((caddr_t)path, arg0, i);
    747 
    748 		/*
    749 		 * Move out the arg pointers.
    750 		 */
    751 		ucp = (caddr_t)STACK_ALIGN(ucp, ALIGNBYTES);
    752 		uap = (char **)STACK_ALLOC(ucp, sizeof(char *) * 3);
    753 		SCARG(&args, path) = arg0;
    754 		SCARG(&args, argp) = uap;
    755 		SCARG(&args, envp) = NULL;
    756 		slash = strrchr(path, '/');
    757 		if (slash)
    758 			(void)suword((caddr_t)uap++,
    759 			    (long)arg0 + (slash + 1 - path));
    760 		else
    761 			(void)suword((caddr_t)uap++, (long)arg0);
    762 		if (options != 0)
    763 			(void)suword((caddr_t)uap++, (long)arg1);
    764 		(void)suword((caddr_t)uap++, 0);	/* terminator */
    765 
    766 		/*
    767 		 * Now try to exec the program.  If can't for any reason
    768 		 * other than it doesn't exist, complain.
    769 		 */
    770 		error = sys_execve(LIST_FIRST(&p->p_lwps), &args, retval);
    771 		if (error == 0 || error == EJUSTRETURN) {
    772 			KERNEL_PROC_UNLOCK(l);
    773 			return;
    774 		}
    775 		printf("exec %s: error %d\n", path, error);
    776 	}
    777 	printf("init: not found\n");
    778 	panic("no init");
    779 }
    780