Home | History | Annotate | Line # | Download | only in kern
init_main.c revision 1.249
      1 /*	$NetBSD: init_main.c,v 1.249 2005/07/15 03:59:31 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 1.249 2005/07/15 03:59:31 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 #include "opt_rootfs_magiclinks.h"
     88 
     89 #include "opencrypto.h"
     90 #include "rnd.h"
     91 
     92 #include <sys/param.h>
     93 #include <sys/acct.h>
     94 #include <sys/filedesc.h>
     95 #include <sys/file.h>
     96 #include <sys/errno.h>
     97 #include <sys/callout.h>
     98 #include <sys/kernel.h>
     99 #include <sys/kcont.h>
    100 #include <sys/mount.h>
    101 #include <sys/proc.h>
    102 #include <sys/kthread.h>
    103 #include <sys/resourcevar.h>
    104 #include <sys/signalvar.h>
    105 #include <sys/systm.h>
    106 #include <sys/vnode.h>
    107 #include <sys/tty.h>
    108 #include <sys/conf.h>
    109 #include <sys/disklabel.h>
    110 #include <sys/buf.h>
    111 #include <sys/device.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 #ifdef VERIFIED_EXEC
    153 #include <sys/verified_exec.h>
    154 #endif
    155 
    156 #include <sys/syscall.h>
    157 #include <sys/sa.h>
    158 #include <sys/syscallargs.h>
    159 
    160 #include <ufs/ufs/quota.h>
    161 
    162 #include <miscfs/genfs/genfs.h>
    163 #include <miscfs/syncfs/syncfs.h>
    164 
    165 #include <machine/cpu.h>
    166 
    167 #include <uvm/uvm.h>
    168 
    169 #include <dev/cons.h>
    170 
    171 #include <net/if.h>
    172 #include <net/raw_cb.h>
    173 
    174 /* Components of the first process -- never freed. */
    175 struct	session session0;
    176 struct	pgrp pgrp0;
    177 struct	proc proc0;
    178 struct	lwp lwp0;
    179 struct	pcred cred0;
    180 struct	filedesc0 filedesc0;
    181 struct	cwdinfo cwdi0;
    182 struct	plimit limit0;
    183 struct	pstats pstat0;
    184 struct	vmspace vmspace0;
    185 struct	sigacts sigacts0;
    186 #ifndef curlwp
    187 struct	lwp *curlwp = &lwp0;
    188 #endif
    189 struct	proc *initproc;
    190 
    191 int	nofile = NOFILE;
    192 int	maxuprc = MAXUPRC;
    193 int	cmask = CMASK;
    194 extern	struct user *proc0paddr;
    195 
    196 struct	vnode *rootvp, *swapdev_vp;
    197 int	boothowto;
    198 int	cold = 1;			/* still working on startup */
    199 struct	timeval boottime;
    200 time_t	rootfstime;			/* recorded root fs time, if known */
    201 
    202 __volatile int start_init_exec;		/* semaphore for start_init() */
    203 
    204 static void check_console(struct proc *p);
    205 static void start_init(void *);
    206 void main(void);
    207 
    208 extern const struct emul emul_netbsd;	/* defined in kern_exec.c */
    209 
    210 /*
    211  * System startup; initialize the world, create process 0, mount root
    212  * filesystem, and fork to create init and pagedaemon.  Most of the
    213  * hard work is done in the lower-level initialization routines including
    214  * startup(), which does memory initialization and autoconfiguration.
    215  */
    216 void
    217 main(void)
    218 {
    219 	struct lwp *l;
    220 	struct proc *p;
    221 	struct pdevinit *pdev;
    222 	int s, error;
    223 	u_int i;
    224 	rlim_t lim;
    225 	extern struct pdevinit pdevinit[];
    226 	extern void schedcpu(void *);
    227 #if defined(NFSSERVER) || defined(NFS)
    228 	extern void nfs_init(void);
    229 #endif
    230 #ifdef NVNODE_IMPLICIT
    231 	int usevnodes;
    232 #endif
    233 
    234 	/*
    235 	 * Initialize the current LWP pointer (curlwp) before
    236 	 * any possible traps/probes to simplify trap processing.
    237 	 */
    238 	l = &lwp0;
    239 	curlwp = l;
    240 	l->l_cpu = curcpu();
    241 	l->l_proc = &proc0;
    242 	l->l_lid = 1;
    243 
    244 	/*
    245 	 * Attempt to find console and initialize
    246 	 * in case of early panic or other messages.
    247 	 */
    248 	consinit();
    249 
    250 	KERNEL_LOCK_INIT();
    251 
    252 	uvm_init();
    253 
    254 	/* Do machine-dependent initialization. */
    255 	cpu_startup();
    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 	tty_init();		/* initialize tty list */
    282 #if NRND > 0
    283 	rnd_init();		/* initialize RNG */
    284 #endif
    285 #if NOPENCRYPTO > 0
    286 	/* Initialize crypto subsystem before configuring crypto hardware. */
    287 	(void)crypto_init();
    288 #endif
    289 	/* Initialize the sysctl subsystem. */
    290 	sysctl_init();
    291 
    292 	/* Initialize process and pgrp structures. */
    293 	procinit();
    294 
    295 #ifdef LKM
    296 	/* Initialize the LKM system. */
    297 	lkm_init();
    298 #endif
    299 
    300 	/*
    301 	 * Create process 0 (the swapper).
    302 	 */
    303 	p = &proc0;
    304 	proc0_insert(p, l, &pgrp0, &session0);
    305 
    306 	/*
    307 	 * Set P_NOCLDWAIT so that kernel threads are reparented to
    308 	 * init(8) when they exit.  init(8) can easily wait them out
    309 	 * for us.
    310 	 */
    311 	p->p_flag = P_SYSTEM | P_NOCLDWAIT;
    312 	p->p_stat = SACTIVE;
    313 	p->p_nice = NZERO;
    314 	p->p_emul = &emul_netbsd;
    315 #ifdef __HAVE_SYSCALL_INTERN
    316 	(*p->p_emul->e_syscall_intern)(p);
    317 #endif
    318 	strncpy(p->p_comm, "swapper", MAXCOMLEN);
    319 
    320 	l->l_flag = L_INMEM;
    321 	l->l_stat = LSONPROC;
    322 	p->p_nrlwps = 1;
    323 
    324 	callout_init(&l->l_tsleep_ch);
    325 
    326 	/* Create credentials. */
    327 	cred0.p_refcnt = 1;
    328 	p->p_cred = &cred0;
    329 	p->p_ucred = crget();
    330 	p->p_ucred->cr_ngroups = 1;	/* group 0 */
    331 
    332 	/* Create the file descriptor table. */
    333 	p->p_fd = &filedesc0.fd_fd;
    334 	fdinit1(&filedesc0);
    335 
    336 	/* Create the CWD info. */
    337 	p->p_cwdi = &cwdi0;
    338 	cwdi0.cwdi_cmask = cmask;
    339 	cwdi0.cwdi_refcnt = 1;
    340 	simple_lock_init(&cwdi0.cwdi_slock);
    341 
    342 	/* Create the limits structures. */
    343 	p->p_limit = &limit0;
    344 	simple_lock_init(&limit0.p_slock);
    345 	for (i = 0; i < sizeof(p->p_rlimit)/sizeof(p->p_rlimit[0]); i++)
    346 		limit0.pl_rlimit[i].rlim_cur =
    347 		    limit0.pl_rlimit[i].rlim_max = RLIM_INFINITY;
    348 
    349 	limit0.pl_rlimit[RLIMIT_NOFILE].rlim_max = maxfiles;
    350 	limit0.pl_rlimit[RLIMIT_NOFILE].rlim_cur =
    351 	    maxfiles < nofile ? maxfiles : nofile;
    352 
    353 	limit0.pl_rlimit[RLIMIT_NPROC].rlim_max = maxproc;
    354 	limit0.pl_rlimit[RLIMIT_NPROC].rlim_cur =
    355 	    maxproc < maxuprc ? maxproc : maxuprc;
    356 
    357 	lim = ptoa(uvmexp.free);
    358 	limit0.pl_rlimit[RLIMIT_RSS].rlim_max = lim;
    359 	limit0.pl_rlimit[RLIMIT_MEMLOCK].rlim_max = lim;
    360 	limit0.pl_rlimit[RLIMIT_MEMLOCK].rlim_cur = lim / 3;
    361 	limit0.pl_corename = defcorename;
    362 	limit0.p_refcnt = 1;
    363 
    364 	/*
    365 	 * Initialize proc0's vmspace, which uses the kernel pmap.
    366 	 * All kernel processes (which never have user space mappings)
    367 	 * share proc0's vmspace, and thus, the kernel pmap.
    368 	 */
    369 	uvmspace_init(&vmspace0, pmap_kernel(), round_page(VM_MIN_ADDRESS),
    370 	    trunc_page(VM_MAX_ADDRESS));
    371 	p->p_vmspace = &vmspace0;
    372 
    373 	l->l_addr = proc0paddr;				/* XXX */
    374 
    375 	p->p_stats = &pstat0;
    376 
    377 	/*
    378 	 * Charge root for one process.
    379 	 */
    380 	(void)chgproccnt(0, 1);
    381 
    382 	rqinit();
    383 
    384 	/* Configure virtual memory system, set vm rlimits. */
    385 	uvm_init_limits(p);
    386 
    387 	/* Initialize the file systems. */
    388 #if defined(NFSSERVER) || defined(NFS)
    389 	nfs_init();			/* initialize server/shared data */
    390 #endif
    391 #ifdef NVNODE_IMPLICIT
    392 	/*
    393 	 * If maximum number of vnodes in namei vnode cache is not explicitly
    394 	 * defined in kernel config, adjust the number such as we use roughly
    395 	 * 1.0% of memory for vnode cache (but not less than NVNODE vnodes).
    396 	 */
    397 	usevnodes = (ptoa((unsigned)physmem) / 100) / sizeof(struct vnode);
    398 	if (usevnodes > desiredvnodes)
    399 		desiredvnodes = usevnodes;
    400 #endif
    401 	vfsinit();
    402 
    403 	/* Configure the system hardware.  This will enable interrupts. */
    404 	configure();
    405 
    406 	ubc_init();		/* must be after autoconfig */
    407 
    408 	/* Lock the kernel on behalf of proc0. */
    409 	KERNEL_PROC_LOCK(l);
    410 
    411 #ifdef SYSVSHM
    412 	/* Initialize System V style shared memory. */
    413 	shminit();
    414 #endif
    415 
    416 #ifdef SYSVSEM
    417 	/* Initialize System V style semaphores. */
    418 	seminit();
    419 #endif
    420 
    421 #ifdef SYSVMSG
    422 	/* Initialize System V style message queues. */
    423 	msginit();
    424 #endif
    425 
    426 #ifdef P1003_1B_SEMAPHORE
    427 	/* Initialize posix semaphores */
    428 	ksem_init();
    429 #endif
    430 
    431 #ifdef VERIFIED_EXEC
    432 	  /*
    433 	   * Initialise the fingerprint operations vectors before
    434 	   * fingerprints can be loaded.
    435 	   */
    436 	veriexec_init_fp_ops();
    437 #endif
    438 
    439 	/* Attach pseudo-devices. */
    440 	for (pdev = pdevinit; pdev->pdev_attach != NULL; pdev++)
    441 		(*pdev->pdev_attach)(pdev->pdev_count);
    442 
    443 #ifdef	FAST_IPSEC
    444 	/* Attach network crypto subsystem */
    445 	ipsec_attach();
    446 #endif
    447 
    448 	/*
    449 	 * Initialize protocols.  Block reception of incoming packets
    450 	 * until everything is ready.
    451 	 */
    452 	s = splnet();
    453 	ifinit();
    454 	domaininit();
    455 	if_attachdomain();
    456 	splx(s);
    457 
    458 #ifdef GPROF
    459 	/* Initialize kernel profiling. */
    460 	kmstartup();
    461 #endif
    462 
    463 	/* Initialize system accouting. */
    464 	acct_init();
    465 
    466 #ifdef SYSTRACE
    467 	systrace_init();
    468 #endif
    469 	/*
    470 	 * Initialize signal-related data structures, and signal state
    471 	 * for proc0.
    472 	 */
    473 	signal_init();
    474 	p->p_sigacts = &sigacts0;
    475 	siginit(p);
    476 
    477 	/* Kick off timeout driven events by calling first time. */
    478 	schedcpu(NULL);
    479 
    480 	/*
    481 	 * Create process 1 (init(8)).  We do this now, as Unix has
    482 	 * historically had init be process 1, and changing this would
    483 	 * probably upset a lot of people.
    484 	 *
    485 	 * Note that process 1 won't immediately exec init(8), but will
    486 	 * wait for us to inform it that the root file system has been
    487 	 * mounted.
    488 	 */
    489 	if (fork1(l, 0, SIGCHLD, NULL, 0, start_init, NULL, NULL, &initproc))
    490 		panic("fork init");
    491 
    492 	/*
    493 	 * Create any kernel threads who's creation was deferred because
    494 	 * initproc had not yet been created.
    495 	 */
    496 	kthread_run_deferred_queue();
    497 
    498 	/*
    499 	 * Now that device driver threads have been created, wait for
    500 	 * them to finish any deferred autoconfiguration.  Note we don't
    501 	 * need to lock this semaphore, since we haven't booted any
    502 	 * secondary processors, yet.
    503 	 */
    504 	while (config_pending)
    505 		(void) tsleep(&config_pending, PWAIT, "cfpend", 0);
    506 
    507 	/*
    508 	 * Finalize configuration now that all real devices have been
    509 	 * found.  This needs to be done before the root device is
    510 	 * selected, since finalization may create the root device.
    511 	 */
    512 	config_finalize();
    513 
    514 	/*
    515 	 * Now that autoconfiguration has completed, we can determine
    516 	 * the root and dump devices.
    517 	 */
    518 	cpu_rootconf();
    519 	cpu_dumpconf();
    520 
    521 	/* Mount the root file system. */
    522 	do {
    523 		domountroothook();
    524 		if ((error = vfs_mountroot())) {
    525 			printf("cannot mount root, error = %d\n", error);
    526 			boothowto |= RB_ASKNAME;
    527 			setroot(root_device,
    528 			    (rootdev != NODEV) ? DISKPART(rootdev) : 0);
    529 		}
    530 	} while (error != 0);
    531 	mountroothook_destroy();
    532 
    533 	/*
    534 	 * Initialise the time-of-day clock, passing the time recorded
    535 	 * in the root filesystem (if any) for use by systems that
    536 	 * don't have a non-volatile time-of-day device.
    537 	 */
    538 	inittodr(rootfstime);
    539 
    540 	CIRCLEQ_FIRST(&mountlist)->mnt_flag |= MNT_ROOTFS;
    541 #ifdef ROOTFS_MAGICLINKS
    542 	CIRCLEQ_FIRST(&mountlist)->mnt_flag |= MNT_MAGICLINKS;
    543 #endif
    544 	CIRCLEQ_FIRST(&mountlist)->mnt_op->vfs_refcount++;
    545 
    546 	/*
    547 	 * Get the vnode for '/'.  Set filedesc0.fd_fd.fd_cdir to
    548 	 * reference it.
    549 	 */
    550 	error = VFS_ROOT(CIRCLEQ_FIRST(&mountlist), &rootvnode);
    551 	if (error)
    552 		panic("cannot find root vnode, error=%d", error);
    553 	cwdi0.cwdi_cdir = rootvnode;
    554 	VREF(cwdi0.cwdi_cdir);
    555 	VOP_UNLOCK(rootvnode, 0);
    556 	cwdi0.cwdi_rdir = NULL;
    557 
    558 	/*
    559 	 * Now that root is mounted, we can fixup initproc's CWD
    560 	 * info.  All other processes are kthreads, which merely
    561 	 * share proc0's CWD info.
    562 	 */
    563 	initproc->p_cwdi->cwdi_cdir = rootvnode;
    564 	VREF(initproc->p_cwdi->cwdi_cdir);
    565 	initproc->p_cwdi->cwdi_rdir = NULL;
    566 
    567 	/*
    568 	 * Now can look at time, having had a chance to verify the time
    569 	 * from the file system.  Reset p->p_rtime as it may have been
    570 	 * munched in mi_switch() after the time got set.
    571 	 */
    572 	proclist_lock_read();
    573 	s = splsched();
    574 	LIST_FOREACH(p, &allproc, p_list) {
    575 		KASSERT((p->p_flag & P_MARKER) == 0);
    576 		p->p_stats->p_start = mono_time = boottime = time;
    577 		LIST_FOREACH(l, &p->p_lwps, l_sibling) {
    578 			if (l->l_cpu != NULL)
    579 				l->l_cpu->ci_schedstate.spc_runtime = time;
    580 		}
    581 		p->p_rtime.tv_sec = p->p_rtime.tv_usec = 0;
    582 	}
    583 	splx(s);
    584 	proclist_unlock_read();
    585 
    586 	/* Create the pageout daemon kernel thread. */
    587 	uvm_swap_init();
    588 	if (kthread_create1(uvm_pageout, NULL, NULL, "pagedaemon"))
    589 		panic("fork pagedaemon");
    590 
    591 	/* Create the filesystem syncer kernel thread. */
    592 	if (kthread_create1(sched_sync, NULL, NULL, "ioflush"))
    593 		panic("fork syncer");
    594 
    595 	/* Create the aiodone daemon kernel thread. */
    596 	if (kthread_create1(uvm_aiodone_daemon, NULL, &uvm.aiodoned_proc,
    597 	    "aiodoned"))
    598 		panic("fork aiodoned");
    599 
    600 #if defined(MULTIPROCESSOR)
    601 	/* Boot the secondary processors. */
    602 	cpu_boot_secondary_processors();
    603 #endif
    604 
    605 	/* Initialize exec structures */
    606 	exec_init(1);
    607 
    608 	/*
    609 	 * Okay, now we can let init(8) exec!  It's off to userland!
    610 	 */
    611 	start_init_exec = 1;
    612 	wakeup(&start_init_exec);
    613 
    614 	/* The scheduler is an infinite loop. */
    615 	uvm_scheduler();
    616 	/* NOTREACHED */
    617 }
    618 
    619 void
    620 setrootfstime(time_t t)
    621 {
    622 
    623 	rootfstime = t;
    624 }
    625 
    626 static void
    627 check_console(struct proc *p)
    628 {
    629 	struct nameidata nd;
    630 	int error;
    631 
    632 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_SYSSPACE, "/dev/console", p);
    633 	error = namei(&nd);
    634 	if (error == 0)
    635 		vrele(nd.ni_vp);
    636 	else if (error == ENOENT)
    637 		printf("warning: no /dev/console\n");
    638 	else
    639 		printf("warning: lookup /dev/console: error %d\n", error);
    640 }
    641 
    642 /*
    643  * List of paths to try when searching for "init".
    644  */
    645 static const char *initpaths[] = {
    646 	"/sbin/init",
    647 	"/sbin/oinit",
    648 	"/sbin/init.bak",
    649 	NULL,
    650 };
    651 
    652 /*
    653  * Start the initial user process; try exec'ing each pathname in "initpaths".
    654  * The program is invoked with one argument containing the boot flags.
    655  */
    656 static void
    657 start_init(void *arg)
    658 {
    659 	struct lwp *l = arg;
    660 	struct proc *p = l->l_proc;
    661 	vaddr_t addr;
    662 	struct sys_execve_args /* {
    663 		syscallarg(const char *) path;
    664 		syscallarg(char * const *) argp;
    665 		syscallarg(char * const *) envp;
    666 	} */ args;
    667 	int options, i, error;
    668 	register_t retval[2];
    669 	char flags[4], *flagsp;
    670 	const char *path, *slash;
    671 	char *ucp, **uap, *arg0, *arg1 = NULL;
    672 	char ipath[129];
    673 	int ipx, len;
    674 
    675 	/*
    676 	 * Now in process 1.
    677 	 */
    678 	strncpy(p->p_comm, "init", MAXCOMLEN);
    679 
    680 	/*
    681 	 * Wait for main() to tell us that it's safe to exec.
    682 	 */
    683 	while (start_init_exec == 0)
    684 		(void) tsleep(&start_init_exec, PWAIT, "initexec", 0);
    685 
    686 	/*
    687 	 * This is not the right way to do this.  We really should
    688 	 * hand-craft a descriptor onto /dev/console to hand to init,
    689 	 * but that's a _lot_ more work, and the benefit from this easy
    690 	 * hack makes up for the "good is the enemy of the best" effect.
    691 	 */
    692 	check_console(p);
    693 
    694 	/*
    695 	 * Need just enough stack to hold the faked-up "execve()" arguments.
    696 	 */
    697 	addr = (vaddr_t)STACK_ALLOC(USRSTACK, PAGE_SIZE);
    698 	if (uvm_map(&p->p_vmspace->vm_map, &addr, PAGE_SIZE,
    699                     NULL, UVM_UNKNOWN_OFFSET, 0,
    700                     UVM_MAPFLAG(UVM_PROT_ALL, UVM_PROT_ALL, UVM_INH_COPY,
    701 		    UVM_ADV_NORMAL,
    702                     UVM_FLAG_FIXED|UVM_FLAG_OVERLAY|UVM_FLAG_COPYONW)) != 0)
    703 		panic("init: couldn't allocate argument space");
    704 	p->p_vmspace->vm_maxsaddr = (caddr_t)STACK_MAX(addr, PAGE_SIZE);
    705 
    706 	ipx = 0;
    707 	while (1) {
    708 		if (boothowto & RB_ASKNAME) {
    709 			printf("init path");
    710 			if (initpaths[ipx])
    711 				printf(" (default %s)", initpaths[ipx]);
    712 			printf(": ");
    713 			len = cngetsn(ipath, sizeof(ipath)-1);
    714 			if (len == 0) {
    715 				if (initpaths[ipx])
    716 					path = initpaths[ipx++];
    717 				else
    718 					continue;
    719 			} else {
    720 				ipath[len] = '\0';
    721 				path = ipath;
    722 			}
    723 		} else {
    724 			if ((path = initpaths[ipx++]) == NULL)
    725 				break;
    726 		}
    727 
    728 		ucp = (char *)USRSTACK;
    729 
    730 		/*
    731 		 * Construct the boot flag argument.
    732 		 */
    733 		flagsp = flags;
    734 		*flagsp++ = '-';
    735 		options = 0;
    736 
    737 		if (boothowto & RB_SINGLE) {
    738 			*flagsp++ = 's';
    739 			options = 1;
    740 		}
    741 #ifdef notyet
    742 		if (boothowto & RB_FASTBOOT) {
    743 			*flagsp++ = 'f';
    744 			options = 1;
    745 		}
    746 #endif
    747 
    748 		/*
    749 		 * Move out the flags (arg 1), if necessary.
    750 		 */
    751 		if (options != 0) {
    752 			*flagsp++ = '\0';
    753 			i = flagsp - flags;
    754 #ifdef DEBUG
    755 			printf("init: copying out flags `%s' %d\n", flags, i);
    756 #endif
    757 			arg1 = STACK_ALLOC(ucp, i);
    758 			ucp = STACK_MAX(arg1, i);
    759 			(void)copyout((caddr_t)flags, arg1, i);
    760 		}
    761 
    762 		/*
    763 		 * Move out the file name (also arg 0).
    764 		 */
    765 		i = strlen(path) + 1;
    766 #ifdef DEBUG
    767 		printf("init: copying out path `%s' %d\n", path, i);
    768 #else
    769 		if (boothowto & RB_ASKNAME || path != initpaths[0])
    770 			printf("init: trying %s\n", path);
    771 #endif
    772 		arg0 = STACK_ALLOC(ucp, i);
    773 		ucp = STACK_MAX(arg0, i);
    774 		(void)copyout(path, arg0, i);
    775 
    776 		/*
    777 		 * Move out the arg pointers.
    778 		 */
    779 		ucp = (caddr_t)STACK_ALIGN(ucp, ALIGNBYTES);
    780 		uap = (char **)STACK_ALLOC(ucp, sizeof(char *) * 3);
    781 		SCARG(&args, path) = arg0;
    782 		SCARG(&args, argp) = uap;
    783 		SCARG(&args, envp) = NULL;
    784 		slash = strrchr(path, '/');
    785 		if (slash)
    786 			(void)suword((caddr_t)uap++,
    787 			    (long)arg0 + (slash + 1 - path));
    788 		else
    789 			(void)suword((caddr_t)uap++, (long)arg0);
    790 		if (options != 0)
    791 			(void)suword((caddr_t)uap++, (long)arg1);
    792 		(void)suword((caddr_t)uap++, 0);	/* terminator */
    793 
    794 		/*
    795 		 * Now try to exec the program.  If can't for any reason
    796 		 * other than it doesn't exist, complain.
    797 		 */
    798 		error = sys_execve(LIST_FIRST(&p->p_lwps), &args, retval);
    799 		if (error == 0 || error == EJUSTRETURN) {
    800 			KERNEL_PROC_UNLOCK(l);
    801 			return;
    802 		}
    803 		printf("exec %s: error %d\n", path, error);
    804 	}
    805 	printf("init: not found\n");
    806 	panic("no init");
    807 }
    808