Home | History | Annotate | Line # | Download | only in kern
init_main.c revision 1.250
      1 /*	$NetBSD: init_main.c,v 1.250 2005/07/16 22:47:18 christos 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 1.250 2005/07/16 22:47:18 christos 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 #include "opt_verified_exec.h"
     89 
     90 #include "opencrypto.h"
     91 #include "rnd.h"
     92 
     93 #include <sys/param.h>
     94 #include <sys/acct.h>
     95 #include <sys/filedesc.h>
     96 #include <sys/file.h>
     97 #include <sys/errno.h>
     98 #include <sys/callout.h>
     99 #include <sys/kernel.h>
    100 #include <sys/kcont.h>
    101 #include <sys/mount.h>
    102 #include <sys/proc.h>
    103 #include <sys/kthread.h>
    104 #include <sys/resourcevar.h>
    105 #include <sys/signalvar.h>
    106 #include <sys/systm.h>
    107 #include <sys/vnode.h>
    108 #include <sys/tty.h>
    109 #include <sys/conf.h>
    110 #include <sys/disklabel.h>
    111 #include <sys/buf.h>
    112 #include <sys/device.h>
    113 #include <sys/exec.h>
    114 #include <sys/socketvar.h>
    115 #include <sys/protosw.h>
    116 #include <sys/reboot.h>
    117 #include <sys/user.h>
    118 #include <sys/sysctl.h>
    119 #include <sys/event.h>
    120 #include <sys/mbuf.h>
    121 #ifdef FAST_IPSEC
    122 #include <netipsec/ipsec.h>
    123 #endif
    124 #ifdef SYSVSHM
    125 #include <sys/shm.h>
    126 #endif
    127 #ifdef SYSVSEM
    128 #include <sys/sem.h>
    129 #endif
    130 #ifdef SYSVMSG
    131 #include <sys/msg.h>
    132 #endif
    133 #ifdef P1003_1B_SEMAPHORE
    134 #include <sys/ksem.h>
    135 #endif
    136 #ifdef SYSTRACE
    137 #include <sys/systrace.h>
    138 #endif
    139 #include <sys/domain.h>
    140 #include <sys/namei.h>
    141 #if NOPENCRYPTO > 0
    142 #include <opencrypto/cryptodev.h>	/* XXX really the framework */
    143 #endif
    144 #if NRND > 0
    145 #include <sys/rnd.h>
    146 #endif
    147 #ifndef PIPE_SOCKETPAIR
    148 #include <sys/pipe.h>
    149 #endif
    150 #ifdef LKM
    151 #include <sys/lkm.h>
    152 #endif
    153 #ifdef VERIFIED_EXEC
    154 #include <sys/verified_exec.h>
    155 #endif
    156 
    157 #include <sys/syscall.h>
    158 #include <sys/sa.h>
    159 #include <sys/syscallargs.h>
    160 
    161 #include <ufs/ufs/quota.h>
    162 
    163 #include <miscfs/genfs/genfs.h>
    164 #include <miscfs/syncfs/syncfs.h>
    165 
    166 #include <machine/cpu.h>
    167 
    168 #include <uvm/uvm.h>
    169 
    170 #include <dev/cons.h>
    171 
    172 #include <net/if.h>
    173 #include <net/raw_cb.h>
    174 
    175 /* Components of the first process -- never freed. */
    176 struct	session session0;
    177 struct	pgrp pgrp0;
    178 struct	proc proc0;
    179 struct	lwp lwp0;
    180 struct	pcred cred0;
    181 struct	filedesc0 filedesc0;
    182 struct	cwdinfo cwdi0;
    183 struct	plimit limit0;
    184 struct	pstats pstat0;
    185 struct	vmspace vmspace0;
    186 struct	sigacts sigacts0;
    187 #ifndef curlwp
    188 struct	lwp *curlwp = &lwp0;
    189 #endif
    190 struct	proc *initproc;
    191 
    192 int	nofile = NOFILE;
    193 int	maxuprc = MAXUPRC;
    194 int	cmask = CMASK;
    195 extern	struct user *proc0paddr;
    196 
    197 struct	vnode *rootvp, *swapdev_vp;
    198 int	boothowto;
    199 int	cold = 1;			/* still working on startup */
    200 struct	timeval boottime;
    201 time_t	rootfstime;			/* recorded root fs time, if known */
    202 
    203 __volatile int start_init_exec;		/* semaphore for start_init() */
    204 
    205 static void check_console(struct proc *p);
    206 static void start_init(void *);
    207 void main(void);
    208 
    209 extern const struct emul emul_netbsd;	/* defined in kern_exec.c */
    210 
    211 /*
    212  * System startup; initialize the world, create process 0, mount root
    213  * filesystem, and fork to create init and pagedaemon.  Most of the
    214  * hard work is done in the lower-level initialization routines including
    215  * startup(), which does memory initialization and autoconfiguration.
    216  */
    217 void
    218 main(void)
    219 {
    220 	struct lwp *l;
    221 	struct proc *p;
    222 	struct pdevinit *pdev;
    223 	int s, error;
    224 	u_int i;
    225 	rlim_t lim;
    226 	extern struct pdevinit pdevinit[];
    227 	extern void schedcpu(void *);
    228 #if defined(NFSSERVER) || defined(NFS)
    229 	extern void nfs_init(void);
    230 #endif
    231 #ifdef NVNODE_IMPLICIT
    232 	int usevnodes;
    233 #endif
    234 
    235 	/*
    236 	 * Initialize the current LWP pointer (curlwp) before
    237 	 * any possible traps/probes to simplify trap processing.
    238 	 */
    239 	l = &lwp0;
    240 	curlwp = l;
    241 	l->l_cpu = curcpu();
    242 	l->l_proc = &proc0;
    243 	l->l_lid = 1;
    244 
    245 	/*
    246 	 * Attempt to find console and initialize
    247 	 * in case of early panic or other messages.
    248 	 */
    249 	consinit();
    250 
    251 	KERNEL_LOCK_INIT();
    252 
    253 	uvm_init();
    254 
    255 	/* Do machine-dependent initialization. */
    256 	cpu_startup();
    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 	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 	simple_lock_init(&cwdi0.cwdi_slock);
    342 
    343 	/* Create the limits structures. */
    344 	p->p_limit = &limit0;
    345 	simple_lock_init(&limit0.p_slock);
    346 	for (i = 0; i < sizeof(p->p_rlimit)/sizeof(p->p_rlimit[0]); i++)
    347 		limit0.pl_rlimit[i].rlim_cur =
    348 		    limit0.pl_rlimit[i].rlim_max = RLIM_INFINITY;
    349 
    350 	limit0.pl_rlimit[RLIMIT_NOFILE].rlim_max = maxfiles;
    351 	limit0.pl_rlimit[RLIMIT_NOFILE].rlim_cur =
    352 	    maxfiles < nofile ? maxfiles : nofile;
    353 
    354 	limit0.pl_rlimit[RLIMIT_NPROC].rlim_max = maxproc;
    355 	limit0.pl_rlimit[RLIMIT_NPROC].rlim_cur =
    356 	    maxproc < maxuprc ? maxproc : maxuprc;
    357 
    358 	lim = ptoa(uvmexp.free);
    359 	limit0.pl_rlimit[RLIMIT_RSS].rlim_max = lim;
    360 	limit0.pl_rlimit[RLIMIT_MEMLOCK].rlim_max = lim;
    361 	limit0.pl_rlimit[RLIMIT_MEMLOCK].rlim_cur = lim / 3;
    362 	limit0.pl_corename = defcorename;
    363 	limit0.p_refcnt = 1;
    364 
    365 	/*
    366 	 * Initialize proc0's vmspace, which uses the kernel pmap.
    367 	 * All kernel processes (which never have user space mappings)
    368 	 * share proc0's vmspace, and thus, the kernel pmap.
    369 	 */
    370 	uvmspace_init(&vmspace0, pmap_kernel(), round_page(VM_MIN_ADDRESS),
    371 	    trunc_page(VM_MAX_ADDRESS));
    372 	p->p_vmspace = &vmspace0;
    373 
    374 	l->l_addr = proc0paddr;				/* XXX */
    375 
    376 	p->p_stats = &pstat0;
    377 
    378 	/*
    379 	 * Charge root for one process.
    380 	 */
    381 	(void)chgproccnt(0, 1);
    382 
    383 	rqinit();
    384 
    385 	/* Configure virtual memory system, set vm rlimits. */
    386 	uvm_init_limits(p);
    387 
    388 	/* Initialize the file systems. */
    389 #if defined(NFSSERVER) || defined(NFS)
    390 	nfs_init();			/* initialize server/shared data */
    391 #endif
    392 #ifdef NVNODE_IMPLICIT
    393 	/*
    394 	 * If maximum number of vnodes in namei vnode cache is not explicitly
    395 	 * defined in kernel config, adjust the number such as we use roughly
    396 	 * 1.0% of memory for vnode cache (but not less than NVNODE vnodes).
    397 	 */
    398 	usevnodes = (ptoa((unsigned)physmem) / 100) / sizeof(struct vnode);
    399 	if (usevnodes > desiredvnodes)
    400 		desiredvnodes = usevnodes;
    401 #endif
    402 	vfsinit();
    403 
    404 	/* Configure the system hardware.  This will enable interrupts. */
    405 	configure();
    406 
    407 	ubc_init();		/* must be after autoconfig */
    408 
    409 	/* Lock the kernel on behalf of proc0. */
    410 	KERNEL_PROC_LOCK(l);
    411 
    412 #ifdef SYSVSHM
    413 	/* Initialize System V style shared memory. */
    414 	shminit();
    415 #endif
    416 
    417 #ifdef SYSVSEM
    418 	/* Initialize System V style semaphores. */
    419 	seminit();
    420 #endif
    421 
    422 #ifdef SYSVMSG
    423 	/* Initialize System V style message queues. */
    424 	msginit();
    425 #endif
    426 
    427 #ifdef P1003_1B_SEMAPHORE
    428 	/* Initialize posix semaphores */
    429 	ksem_init();
    430 #endif
    431 
    432 #ifdef VERIFIED_EXEC
    433 	  /*
    434 	   * Initialise the fingerprint operations vectors before
    435 	   * fingerprints can be loaded.
    436 	   */
    437 	veriexec_init_fp_ops();
    438 #endif
    439 
    440 	/* Attach pseudo-devices. */
    441 	for (pdev = pdevinit; pdev->pdev_attach != NULL; pdev++)
    442 		(*pdev->pdev_attach)(pdev->pdev_count);
    443 
    444 #ifdef	FAST_IPSEC
    445 	/* Attach network crypto subsystem */
    446 	ipsec_attach();
    447 #endif
    448 
    449 	/*
    450 	 * Initialize protocols.  Block reception of incoming packets
    451 	 * until everything is ready.
    452 	 */
    453 	s = splnet();
    454 	ifinit();
    455 	domaininit();
    456 	if_attachdomain();
    457 	splx(s);
    458 
    459 #ifdef GPROF
    460 	/* Initialize kernel profiling. */
    461 	kmstartup();
    462 #endif
    463 
    464 	/* Initialize system accouting. */
    465 	acct_init();
    466 
    467 #ifdef SYSTRACE
    468 	systrace_init();
    469 #endif
    470 	/*
    471 	 * Initialize signal-related data structures, and signal state
    472 	 * for proc0.
    473 	 */
    474 	signal_init();
    475 	p->p_sigacts = &sigacts0;
    476 	siginit(p);
    477 
    478 	/* Kick off timeout driven events by calling first time. */
    479 	schedcpu(NULL);
    480 
    481 	/*
    482 	 * Create process 1 (init(8)).  We do this now, as Unix has
    483 	 * historically had init be process 1, and changing this would
    484 	 * probably upset a lot of people.
    485 	 *
    486 	 * Note that process 1 won't immediately exec init(8), but will
    487 	 * wait for us to inform it that the root file system has been
    488 	 * mounted.
    489 	 */
    490 	if (fork1(l, 0, SIGCHLD, NULL, 0, start_init, NULL, NULL, &initproc))
    491 		panic("fork init");
    492 
    493 	/*
    494 	 * Create any kernel threads who's creation was deferred because
    495 	 * initproc had not yet been created.
    496 	 */
    497 	kthread_run_deferred_queue();
    498 
    499 	/*
    500 	 * Now that device driver threads have been created, wait for
    501 	 * them to finish any deferred autoconfiguration.  Note we don't
    502 	 * need to lock this semaphore, since we haven't booted any
    503 	 * secondary processors, yet.
    504 	 */
    505 	while (config_pending)
    506 		(void) tsleep(&config_pending, PWAIT, "cfpend", 0);
    507 
    508 	/*
    509 	 * Finalize configuration now that all real devices have been
    510 	 * found.  This needs to be done before the root device is
    511 	 * selected, since finalization may create the root device.
    512 	 */
    513 	config_finalize();
    514 
    515 	/*
    516 	 * Now that autoconfiguration has completed, we can determine
    517 	 * the root and dump devices.
    518 	 */
    519 	cpu_rootconf();
    520 	cpu_dumpconf();
    521 
    522 	/* Mount the root file system. */
    523 	do {
    524 		domountroothook();
    525 		if ((error = vfs_mountroot())) {
    526 			printf("cannot mount root, error = %d\n", error);
    527 			boothowto |= RB_ASKNAME;
    528 			setroot(root_device,
    529 			    (rootdev != NODEV) ? DISKPART(rootdev) : 0);
    530 		}
    531 	} while (error != 0);
    532 	mountroothook_destroy();
    533 
    534 	/*
    535 	 * Initialise the time-of-day clock, passing the time recorded
    536 	 * in the root filesystem (if any) for use by systems that
    537 	 * don't have a non-volatile time-of-day device.
    538 	 */
    539 	inittodr(rootfstime);
    540 
    541 	CIRCLEQ_FIRST(&mountlist)->mnt_flag |= MNT_ROOTFS;
    542 #ifdef ROOTFS_MAGICLINKS
    543 	CIRCLEQ_FIRST(&mountlist)->mnt_flag |= MNT_MAGICLINKS;
    544 #endif
    545 	CIRCLEQ_FIRST(&mountlist)->mnt_op->vfs_refcount++;
    546 
    547 	/*
    548 	 * Get the vnode for '/'.  Set filedesc0.fd_fd.fd_cdir to
    549 	 * reference it.
    550 	 */
    551 	error = VFS_ROOT(CIRCLEQ_FIRST(&mountlist), &rootvnode);
    552 	if (error)
    553 		panic("cannot find root vnode, error=%d", error);
    554 	cwdi0.cwdi_cdir = rootvnode;
    555 	VREF(cwdi0.cwdi_cdir);
    556 	VOP_UNLOCK(rootvnode, 0);
    557 	cwdi0.cwdi_rdir = NULL;
    558 
    559 	/*
    560 	 * Now that root is mounted, we can fixup initproc's CWD
    561 	 * info.  All other processes are kthreads, which merely
    562 	 * share proc0's CWD info.
    563 	 */
    564 	initproc->p_cwdi->cwdi_cdir = rootvnode;
    565 	VREF(initproc->p_cwdi->cwdi_cdir);
    566 	initproc->p_cwdi->cwdi_rdir = NULL;
    567 
    568 	/*
    569 	 * Now can look at time, having had a chance to verify the time
    570 	 * from the file system.  Reset p->p_rtime as it may have been
    571 	 * munched in mi_switch() after the time got set.
    572 	 */
    573 	proclist_lock_read();
    574 	s = splsched();
    575 	LIST_FOREACH(p, &allproc, p_list) {
    576 		KASSERT((p->p_flag & P_MARKER) == 0);
    577 		p->p_stats->p_start = mono_time = boottime = time;
    578 		LIST_FOREACH(l, &p->p_lwps, l_sibling) {
    579 			if (l->l_cpu != NULL)
    580 				l->l_cpu->ci_schedstate.spc_runtime = time;
    581 		}
    582 		p->p_rtime.tv_sec = p->p_rtime.tv_usec = 0;
    583 	}
    584 	splx(s);
    585 	proclist_unlock_read();
    586 
    587 	/* Create the pageout daemon kernel thread. */
    588 	uvm_swap_init();
    589 	if (kthread_create1(uvm_pageout, NULL, NULL, "pagedaemon"))
    590 		panic("fork pagedaemon");
    591 
    592 	/* Create the filesystem syncer kernel thread. */
    593 	if (kthread_create1(sched_sync, NULL, NULL, "ioflush"))
    594 		panic("fork syncer");
    595 
    596 	/* Create the aiodone daemon kernel thread. */
    597 	if (kthread_create1(uvm_aiodone_daemon, NULL, &uvm.aiodoned_proc,
    598 	    "aiodoned"))
    599 		panic("fork aiodoned");
    600 
    601 #if defined(MULTIPROCESSOR)
    602 	/* Boot the secondary processors. */
    603 	cpu_boot_secondary_processors();
    604 #endif
    605 
    606 	/* Initialize exec structures */
    607 	exec_init(1);
    608 
    609 	/*
    610 	 * Okay, now we can let init(8) exec!  It's off to userland!
    611 	 */
    612 	start_init_exec = 1;
    613 	wakeup(&start_init_exec);
    614 
    615 	/* The scheduler is an infinite loop. */
    616 	uvm_scheduler();
    617 	/* NOTREACHED */
    618 }
    619 
    620 void
    621 setrootfstime(time_t t)
    622 {
    623 
    624 	rootfstime = t;
    625 }
    626 
    627 static void
    628 check_console(struct proc *p)
    629 {
    630 	struct nameidata nd;
    631 	int error;
    632 
    633 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_SYSSPACE, "/dev/console", p);
    634 	error = namei(&nd);
    635 	if (error == 0)
    636 		vrele(nd.ni_vp);
    637 	else if (error == ENOENT)
    638 		printf("warning: no /dev/console\n");
    639 	else
    640 		printf("warning: lookup /dev/console: error %d\n", error);
    641 }
    642 
    643 /*
    644  * List of paths to try when searching for "init".
    645  */
    646 static const char *initpaths[] = {
    647 	"/sbin/init",
    648 	"/sbin/oinit",
    649 	"/sbin/init.bak",
    650 	NULL,
    651 };
    652 
    653 /*
    654  * Start the initial user process; try exec'ing each pathname in "initpaths".
    655  * The program is invoked with one argument containing the boot flags.
    656  */
    657 static void
    658 start_init(void *arg)
    659 {
    660 	struct lwp *l = arg;
    661 	struct proc *p = l->l_proc;
    662 	vaddr_t addr;
    663 	struct sys_execve_args /* {
    664 		syscallarg(const char *) path;
    665 		syscallarg(char * const *) argp;
    666 		syscallarg(char * const *) envp;
    667 	} */ args;
    668 	int options, i, error;
    669 	register_t retval[2];
    670 	char flags[4], *flagsp;
    671 	const char *path, *slash;
    672 	char *ucp, **uap, *arg0, *arg1 = NULL;
    673 	char ipath[129];
    674 	int ipx, len;
    675 
    676 	/*
    677 	 * Now in process 1.
    678 	 */
    679 	strncpy(p->p_comm, "init", MAXCOMLEN);
    680 
    681 	/*
    682 	 * Wait for main() to tell us that it's safe to exec.
    683 	 */
    684 	while (start_init_exec == 0)
    685 		(void) tsleep(&start_init_exec, PWAIT, "initexec", 0);
    686 
    687 	/*
    688 	 * This is not the right way to do this.  We really should
    689 	 * hand-craft a descriptor onto /dev/console to hand to init,
    690 	 * but that's a _lot_ more work, and the benefit from this easy
    691 	 * hack makes up for the "good is the enemy of the best" effect.
    692 	 */
    693 	check_console(p);
    694 
    695 	/*
    696 	 * Need just enough stack to hold the faked-up "execve()" arguments.
    697 	 */
    698 	addr = (vaddr_t)STACK_ALLOC(USRSTACK, PAGE_SIZE);
    699 	if (uvm_map(&p->p_vmspace->vm_map, &addr, PAGE_SIZE,
    700                     NULL, UVM_UNKNOWN_OFFSET, 0,
    701                     UVM_MAPFLAG(UVM_PROT_ALL, UVM_PROT_ALL, UVM_INH_COPY,
    702 		    UVM_ADV_NORMAL,
    703                     UVM_FLAG_FIXED|UVM_FLAG_OVERLAY|UVM_FLAG_COPYONW)) != 0)
    704 		panic("init: couldn't allocate argument space");
    705 	p->p_vmspace->vm_maxsaddr = (caddr_t)STACK_MAX(addr, PAGE_SIZE);
    706 
    707 	ipx = 0;
    708 	while (1) {
    709 		if (boothowto & RB_ASKNAME) {
    710 			printf("init path");
    711 			if (initpaths[ipx])
    712 				printf(" (default %s)", initpaths[ipx]);
    713 			printf(": ");
    714 			len = cngetsn(ipath, sizeof(ipath)-1);
    715 			if (len == 0) {
    716 				if (initpaths[ipx])
    717 					path = initpaths[ipx++];
    718 				else
    719 					continue;
    720 			} else {
    721 				ipath[len] = '\0';
    722 				path = ipath;
    723 			}
    724 		} else {
    725 			if ((path = initpaths[ipx++]) == NULL)
    726 				break;
    727 		}
    728 
    729 		ucp = (char *)USRSTACK;
    730 
    731 		/*
    732 		 * Construct the boot flag argument.
    733 		 */
    734 		flagsp = flags;
    735 		*flagsp++ = '-';
    736 		options = 0;
    737 
    738 		if (boothowto & RB_SINGLE) {
    739 			*flagsp++ = 's';
    740 			options = 1;
    741 		}
    742 #ifdef notyet
    743 		if (boothowto & RB_FASTBOOT) {
    744 			*flagsp++ = 'f';
    745 			options = 1;
    746 		}
    747 #endif
    748 
    749 		/*
    750 		 * Move out the flags (arg 1), if necessary.
    751 		 */
    752 		if (options != 0) {
    753 			*flagsp++ = '\0';
    754 			i = flagsp - flags;
    755 #ifdef DEBUG
    756 			printf("init: copying out flags `%s' %d\n", flags, i);
    757 #endif
    758 			arg1 = STACK_ALLOC(ucp, i);
    759 			ucp = STACK_MAX(arg1, i);
    760 			(void)copyout((caddr_t)flags, arg1, i);
    761 		}
    762 
    763 		/*
    764 		 * Move out the file name (also arg 0).
    765 		 */
    766 		i = strlen(path) + 1;
    767 #ifdef DEBUG
    768 		printf("init: copying out path `%s' %d\n", path, i);
    769 #else
    770 		if (boothowto & RB_ASKNAME || path != initpaths[0])
    771 			printf("init: trying %s\n", path);
    772 #endif
    773 		arg0 = STACK_ALLOC(ucp, i);
    774 		ucp = STACK_MAX(arg0, i);
    775 		(void)copyout(path, arg0, i);
    776 
    777 		/*
    778 		 * Move out the arg pointers.
    779 		 */
    780 		ucp = (caddr_t)STACK_ALIGN(ucp, ALIGNBYTES);
    781 		uap = (char **)STACK_ALLOC(ucp, sizeof(char *) * 3);
    782 		SCARG(&args, path) = arg0;
    783 		SCARG(&args, argp) = uap;
    784 		SCARG(&args, envp) = NULL;
    785 		slash = strrchr(path, '/');
    786 		if (slash)
    787 			(void)suword((caddr_t)uap++,
    788 			    (long)arg0 + (slash + 1 - path));
    789 		else
    790 			(void)suword((caddr_t)uap++, (long)arg0);
    791 		if (options != 0)
    792 			(void)suword((caddr_t)uap++, (long)arg1);
    793 		(void)suword((caddr_t)uap++, 0);	/* terminator */
    794 
    795 		/*
    796 		 * Now try to exec the program.  If can't for any reason
    797 		 * other than it doesn't exist, complain.
    798 		 */
    799 		error = sys_execve(LIST_FIRST(&p->p_lwps), &args, retval);
    800 		if (error == 0 || error == EJUSTRETURN) {
    801 			KERNEL_PROC_UNLOCK(l);
    802 			return;
    803 		}
    804 		printf("exec %s: error %d\n", path, error);
    805 	}
    806 	printf("init: not found\n");
    807 	panic("no init");
    808 }
    809