Home | History | Annotate | Line # | Download | only in kern
init_main.c revision 1.156
      1 /*	$NetBSD: init_main.c,v 1.156 1999/09/17 20:11:56 thorpej Exp $	*/
      2 
      3 /*
      4  * Copyright (c) 1995 Christopher G. Demetriou.  All rights reserved.
      5  * Copyright (c) 1982, 1986, 1989, 1991, 1992, 1993
      6  *	The Regents of the University of California.  All rights reserved.
      7  * (c) UNIX System Laboratories, Inc.
      8  * All or some portions of this file are derived from material licensed
      9  * to the University of California by American Telephone and Telegraph
     10  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
     11  * the permission of UNIX System Laboratories, Inc.
     12  *
     13  * Redistribution and use in source and binary forms, with or without
     14  * modification, are permitted provided that the following conditions
     15  * are met:
     16  * 1. Redistributions of source code must retain the above copyright
     17  *    notice, this list of conditions and the following disclaimer.
     18  * 2. Redistributions in binary form must reproduce the above copyright
     19  *    notice, this list of conditions and the following disclaimer in the
     20  *    documentation and/or other materials provided with the distribution.
     21  * 3. All advertising materials mentioning features or use of this software
     22  *    must display the following acknowledgement:
     23  *	This product includes software developed by the University of
     24  *	California, Berkeley and its contributors.
     25  * 4. Neither the name of the University nor the names of its contributors
     26  *    may be used to endorse or promote products derived from this software
     27  *    without specific prior written permission.
     28  *
     29  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
     30  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
     31  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
     32  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
     33  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
     34  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
     35  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
     36  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
     37  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
     38  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
     39  * SUCH DAMAGE.
     40  *
     41  *	@(#)init_main.c	8.16 (Berkeley) 5/14/95
     42  */
     43 
     44 #include "fs_nfs.h"
     45 #include "opt_nfsserver.h"
     46 #include "opt_sysv.h"
     47 
     48 #include "rnd.h"
     49 
     50 #include <sys/param.h>
     51 #include <sys/filedesc.h>
     52 #include <sys/file.h>
     53 #include <sys/errno.h>
     54 #include <sys/exec.h>
     55 #include <sys/kernel.h>
     56 #include <sys/mount.h>
     57 #include <sys/map.h>
     58 #include <sys/proc.h>
     59 #include <sys/kthread.h>
     60 #include <sys/resourcevar.h>
     61 #include <sys/signalvar.h>
     62 #include <sys/systm.h>
     63 #include <sys/vnode.h>
     64 #include <sys/tty.h>
     65 #include <sys/conf.h>
     66 #include <sys/disklabel.h>
     67 #include <sys/buf.h>
     68 #include <sys/device.h>
     69 #include <sys/socketvar.h>
     70 #include <sys/protosw.h>
     71 #include <sys/reboot.h>
     72 #include <sys/user.h>
     73 #ifdef SYSVSHM
     74 #include <sys/shm.h>
     75 #endif
     76 #ifdef SYSVSEM
     77 #include <sys/sem.h>
     78 #endif
     79 #ifdef SYSVMSG
     80 #include <sys/msg.h>
     81 #endif
     82 #include <sys/domain.h>
     83 #include <sys/mbuf.h>
     84 #include <sys/namei.h>
     85 #if NRND > 0
     86 #include <sys/rnd.h>
     87 #endif
     88 
     89 #include <sys/syscall.h>
     90 #include <sys/syscallargs.h>
     91 
     92 #include <ufs/ufs/quota.h>
     93 
     94 #include <machine/cpu.h>
     95 
     96 #include <vm/vm.h>
     97 #include <vm/vm_pageout.h>
     98 
     99 #include <uvm/uvm.h>
    100 
    101 #include <net/if.h>
    102 #include <net/raw_cb.h>
    103 
    104 char	copyright[] = "\
    105 Copyright (c) 1996, 1997, 1998, 1999
    106     The NetBSD Foundation, Inc.  All rights reserved.
    107 Copyright (c) 1982, 1986, 1989, 1991, 1993
    108     The Regents of the University of California.  All rights reserved.
    109 
    110 ";
    111 
    112 /* Components of the first process -- never freed. */
    113 struct	session session0;
    114 struct	pgrp pgrp0;
    115 struct	proc proc0;
    116 struct	pcred cred0;
    117 struct	filedesc0 filedesc0;
    118 struct	cwdinfo cwdi0;
    119 struct	plimit limit0;
    120 struct	vmspace vmspace0;
    121 struct	sigacts sigacts0;
    122 #ifndef curproc
    123 struct	proc *curproc = &proc0;
    124 #endif
    125 struct	proc *initproc;
    126 
    127 int	cmask = CMASK;
    128 extern	struct user *proc0paddr;
    129 
    130 struct	vnode *rootvp, *swapdev_vp;
    131 int	boothowto;
    132 int	cold = 1;			/* still working on startup */
    133 struct	timeval boottime;
    134 struct	timeval runtime;
    135 
    136 static void check_console __P((struct proc *p));
    137 static void start_init __P((void *));
    138 static void start_pagedaemon __P((void *));
    139 static void start_reaper __P((void *));
    140 void main __P((void));
    141 
    142 extern char sigcode[], esigcode[];
    143 #ifdef SYSCALL_DEBUG
    144 extern char *syscallnames[];
    145 #endif
    146 
    147 struct emul emul_netbsd = {
    148 	"netbsd",
    149 	NULL,
    150 	sendsig,
    151 	SYS_syscall,
    152 	SYS_MAXSYSCALL,
    153 	sysent,
    154 #ifdef SYSCALL_DEBUG
    155 	syscallnames,
    156 #else
    157 	NULL,
    158 #endif
    159 	0,
    160 	copyargs,
    161 	setregs,
    162 	sigcode,
    163 	esigcode,
    164 };
    165 
    166 /*
    167  * System startup; initialize the world, create process 0, mount root
    168  * filesystem, and fork to create init and pagedaemon.  Most of the
    169  * hard work is done in the lower-level initialization routines including
    170  * startup(), which does memory initialization and autoconfiguration.
    171  */
    172 void
    173 main()
    174 {
    175 	struct proc *p;
    176 	struct pdevinit *pdev;
    177 	int i, s, error;
    178 	extern struct pdevinit pdevinit[];
    179 	extern void roundrobin __P((void *));
    180 	extern void schedcpu __P((void *));
    181 	extern void disk_init __P((void));
    182 #if defined(NFSSERVER) || defined(NFS)
    183 	extern void nfs_init __P((void));
    184 #endif
    185 
    186 	/*
    187 	 * Initialize the current process pointer (curproc) before
    188 	 * any possible traps/probes to simplify trap processing.
    189 	 */
    190 	p = &proc0;
    191 	curproc = p;
    192 	/*
    193 	 * Attempt to find console and initialize
    194 	 * in case of early panic or other messages.
    195 	 */
    196 	consinit();
    197 	printf("%s", copyright);
    198 
    199 	uvm_init();
    200 
    201 	/* Do machine-dependent initialization. */
    202 	cpu_startup();
    203 
    204 	/*
    205 	 * Initialize mbuf's.  Do this now because we might attempt to
    206 	 * allocate mbufs or mbuf clusters during autoconfiguration.
    207 	 */
    208 	mbinit();
    209 
    210 	/* Initialize sockets. */
    211 	soinit();
    212 
    213 	/*
    214 	 * The following 3 things must be done before autoconfiguration.
    215 	 */
    216 	disk_init();		/* initialize disk list */
    217 	tty_init();		/* initialize tty list */
    218 #if NRND > 0
    219 	rnd_init();		/* initialize RNG */
    220 #endif
    221 
    222 	/*
    223 	 * Initialize process and pgrp structures.
    224 	 */
    225 	procinit();
    226 
    227 	/*
    228 	 * Create process 0 (the swapper).
    229 	 */
    230 	s = proclist_lock_write();
    231 	LIST_INSERT_HEAD(&allproc, p, p_list);
    232 	proclist_unlock_write(s);
    233 
    234 	p->p_pgrp = &pgrp0;
    235 	LIST_INSERT_HEAD(PGRPHASH(0), &pgrp0, pg_hash);
    236 	LIST_INIT(&pgrp0.pg_members);
    237 	LIST_INSERT_HEAD(&pgrp0.pg_members, p, p_pglist);
    238 
    239 	pgrp0.pg_session = &session0;
    240 	session0.s_count = 1;
    241 	session0.s_sid = p->p_pid;
    242 	session0.s_leader = p;
    243 
    244 	/*
    245 	 * Set P_NOCLDWAIT so that kernel threads are reparented to
    246 	 * init(8) when they exit.  init(8) can easily wait them out
    247 	 * for us.
    248 	 */
    249 	p->p_flag = P_INMEM | P_SYSTEM | P_NOCLDWAIT;
    250 	p->p_stat = SRUN;
    251 	p->p_nice = NZERO;
    252 	p->p_emul = &emul_netbsd;
    253 	strncpy(p->p_comm, "swapper", MAXCOMLEN);
    254 
    255 	/* Create credentials. */
    256 	cred0.p_refcnt = 1;
    257 	p->p_cred = &cred0;
    258 	p->p_ucred = crget();
    259 	p->p_ucred->cr_ngroups = 1;	/* group 0 */
    260 
    261 	/* Create the file descriptor table. */
    262 	finit();
    263 	p->p_fd = &filedesc0.fd_fd;
    264 	fdinit1(&filedesc0);
    265 
    266 	/* Create the CWD info. */
    267 	p->p_cwdi = &cwdi0;
    268 	cwdi0.cwdi_cmask = cmask;
    269 	cwdi0.cwdi_refcnt = 1;
    270 
    271 	/* Create the limits structures. */
    272 	p->p_limit = &limit0;
    273 	for (i = 0; i < sizeof(p->p_rlimit)/sizeof(p->p_rlimit[0]); i++)
    274 		limit0.pl_rlimit[i].rlim_cur =
    275 		    limit0.pl_rlimit[i].rlim_max = RLIM_INFINITY;
    276 
    277 	limit0.pl_rlimit[RLIMIT_NOFILE].rlim_max = maxfiles;
    278 	limit0.pl_rlimit[RLIMIT_NOFILE].rlim_cur =
    279 	    maxfiles < NOFILE ? maxfiles : NOFILE;
    280 
    281 	limit0.pl_rlimit[RLIMIT_NPROC].rlim_max = maxproc;
    282 	limit0.pl_rlimit[RLIMIT_NPROC].rlim_cur =
    283 	    maxproc < MAXUPRC ? maxproc : MAXUPRC;
    284 
    285 	i = ptoa(uvmexp.free);
    286 	limit0.pl_rlimit[RLIMIT_RSS].rlim_max = i;
    287 	limit0.pl_rlimit[RLIMIT_MEMLOCK].rlim_max = i;
    288 	limit0.pl_rlimit[RLIMIT_MEMLOCK].rlim_cur = i / 3;
    289 	limit0.p_refcnt = 1;
    290 
    291 	/*
    292 	 * Initialize proc0's vmspace, which uses the kernel pmap.
    293 	 * All kernel processes (which never have user space mappings)
    294 	 * share proc0's vmspace, and thus, the kernel pmap.
    295 	 */
    296 	uvmspace_init(&vmspace0, pmap_kernel(), round_page(VM_MIN_ADDRESS),
    297 	    trunc_page(VM_MAX_ADDRESS), TRUE);
    298 	p->p_vmspace = &vmspace0;
    299 
    300 	p->p_addr = proc0paddr;				/* XXX */
    301 
    302 	/*
    303 	 * We continue to place resource usage info in the
    304 	 * user struct so they're pageable.
    305 	 */
    306 	p->p_stats = &p->p_addr->u_stats;
    307 
    308 	/*
    309 	 * Charge root for one process.
    310 	 */
    311 	(void)chgproccnt(0, 1);
    312 
    313 	rqinit();
    314 
    315 	/* Configure virtual memory system, set vm rlimits. */
    316 	uvm_init_limits(p);
    317 
    318 	/* Initialize the file systems. */
    319 #if defined(NFSSERVER) || defined(NFS)
    320 	nfs_init();			/* initialize server/shared data */
    321 #endif
    322 	vfsinit();
    323 
    324 	/* Configure the system hardware.  This will enable interrupts. */
    325 	configure();
    326 
    327 #ifdef SYSVSHM
    328 	/* Initialize System V style shared memory. */
    329 	shminit();
    330 #endif
    331 
    332 #ifdef SYSVSEM
    333 	/* Initialize System V style semaphores. */
    334 	seminit();
    335 #endif
    336 
    337 #ifdef SYSVMSG
    338 	/* Initialize System V style message queues. */
    339 	msginit();
    340 #endif
    341 
    342 	/* Attach pseudo-devices. */
    343 	for (pdev = pdevinit; pdev->pdev_attach != NULL; pdev++)
    344 		(*pdev->pdev_attach)(pdev->pdev_count);
    345 
    346 	/*
    347 	 * Initialize protocols.  Block reception of incoming packets
    348 	 * until everything is ready.
    349 	 */
    350 	s = splimp();
    351 	ifinit();
    352 	domaininit();
    353 	splx(s);
    354 
    355 #ifdef GPROF
    356 	/* Initialize kernel profiling. */
    357 	kmstartup();
    358 #endif
    359 
    360 	/* Kick off timeout driven events by calling first time. */
    361 	roundrobin(NULL);
    362 	schedcpu(NULL);
    363 
    364 	/* Determine the root and dump devices. */
    365 	cpu_rootconf();
    366 	cpu_dumpconf();
    367 
    368 	/* Mount the root file system. */
    369 	do {
    370 		domountroothook();
    371 		if ((error = vfs_mountroot())) {
    372 			printf("cannot mount root, error = %d\n", error);
    373 			boothowto |= RB_ASKNAME;
    374 			setroot(root_device,
    375 			    (rootdev != NODEV) ? DISKPART(rootdev) : 0);
    376 		}
    377 	} while (error != 0);
    378 	mountroothook_destroy();
    379 
    380 	mountlist.cqh_first->mnt_flag |= MNT_ROOTFS;
    381 	mountlist.cqh_first->mnt_op->vfs_refcount++;
    382 
    383 	/*
    384 	 * Get the vnode for '/'.  Set filedesc0.fd_fd.fd_cdir to
    385 	 * reference it.
    386 	 */
    387 	if (VFS_ROOT(mountlist.cqh_first, &rootvnode))
    388 		panic("cannot find root vnode");
    389 	cwdi0.cwdi_cdir = rootvnode;
    390 	VREF(cwdi0.cwdi_cdir);
    391 	VOP_UNLOCK(rootvnode, 0);
    392 	cwdi0.cwdi_rdir = NULL;
    393 	uvm_swap_init();
    394 
    395 	/*
    396 	 * Now can look at time, having had a chance to verify the time
    397 	 * from the file system.  Reset p->p_rtime as it may have been
    398 	 * munched in mi_switch() after the time got set.
    399 	 */
    400 	p->p_stats->p_start = runtime = mono_time = boottime = time;
    401 	p->p_rtime.tv_sec = p->p_rtime.tv_usec = 0;
    402 
    403 	/*
    404 	 * Initialize signal-related data structures, and signal state
    405 	 * for proc0.
    406 	 */
    407 	signal_init();
    408 	p->p_sigacts = &sigacts0;
    409 	siginit(p);
    410 
    411 	/* Create process 1 (init(8)). */
    412 	if (fork1(p, 0, SIGCHLD, NULL, 0, NULL, &initproc))
    413 		panic("fork init");
    414 	cpu_set_kpc(initproc, start_init, initproc);
    415 
    416 	/* Create process 2, the pageout daemon kernel thread. */
    417 	if (kthread_create1(start_pagedaemon, NULL, NULL, "pagedaemon"))
    418 		panic("fork pagedaemon");
    419 
    420 	/* Create process 3, the process reaper kernel thread. */
    421 	if (kthread_create1(start_reaper, NULL, NULL, "reaper"))
    422 		panic("fork reaper");
    423 
    424 	/* Create any other deferred kernel threads. */
    425 	kthread_run_deferred_queue();
    426 
    427 	/* The scheduler is an infinite loop. */
    428 	uvm_scheduler();
    429 	/* NOTREACHED */
    430 }
    431 
    432 static void
    433 check_console(p)
    434 	struct proc *p;
    435 {
    436 	struct nameidata nd;
    437 	int error;
    438 
    439 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_SYSSPACE, "/dev/console", p);
    440 	error = namei(&nd);
    441 	if (error == 0)
    442 		vrele(nd.ni_vp);
    443 	else if (error == ENOENT)
    444 		printf("warning: no /dev/console\n");
    445 	else
    446 		printf("warning: lookup /dev/console: error %d\n", error);
    447 }
    448 
    449 /*
    450  * List of paths to try when searching for "init".
    451  */
    452 static char *initpaths[] = {
    453 	"/sbin/init",
    454 	"/sbin/oinit",
    455 	"/sbin/init.bak",
    456 	NULL,
    457 };
    458 
    459 /*
    460  * Start the initial user process; try exec'ing each pathname in "initpaths".
    461  * The program is invoked with one argument containing the boot flags.
    462  */
    463 static void
    464 start_init(arg)
    465 	void *arg;
    466 {
    467 	struct proc *p = arg;
    468 	vaddr_t addr;
    469 	struct sys_execve_args /* {
    470 		syscallarg(const char *) path;
    471 		syscallarg(char * const *) argp;
    472 		syscallarg(char * const *) envp;
    473 	} */ args;
    474 	int options, i, error;
    475 	register_t retval[2];
    476 	char flags[4], *flagsp;
    477 	char **pathp, *path, *slash, *ucp, **uap, *arg0, *arg1 = NULL;
    478 
    479 	/*
    480 	 * Now in process 1.
    481 	 */
    482 	strncpy(p->p_comm, "init", MAXCOMLEN);
    483 
    484 	/*
    485 	 * This is not the right way to do this.  We really should
    486 	 * hand-craft a descriptor onto /dev/console to hand to init,
    487 	 * but that's a _lot_ more work, and the benefit from this easy
    488 	 * hack makes up for the "good is the enemy of the best" effect.
    489 	 */
    490 	check_console(p);
    491 
    492 	/*
    493 	 * Need just enough stack to hold the faked-up "execve()" arguments.
    494 	 */
    495 	addr = USRSTACK - PAGE_SIZE;
    496 	if (uvm_map(&p->p_vmspace->vm_map, &addr, PAGE_SIZE,
    497                     NULL, UVM_UNKNOWN_OFFSET,
    498                     UVM_MAPFLAG(UVM_PROT_ALL, UVM_PROT_ALL, UVM_INH_COPY,
    499 		    UVM_ADV_NORMAL,
    500                     UVM_FLAG_FIXED|UVM_FLAG_OVERLAY|UVM_FLAG_COPYONW))
    501 		!= KERN_SUCCESS)
    502 		panic("init: couldn't allocate argument space");
    503 	p->p_vmspace->vm_maxsaddr = (caddr_t)addr;
    504 
    505 	for (pathp = &initpaths[0]; (path = *pathp) != NULL; pathp++) {
    506 		ucp = (char *)(addr + PAGE_SIZE);
    507 
    508 		/*
    509 		 * Construct the boot flag argument.
    510 		 */
    511 		flagsp = flags;
    512 		*flagsp++ = '-';
    513 		options = 0;
    514 
    515 		if (boothowto & RB_SINGLE) {
    516 			*flagsp++ = 's';
    517 			options = 1;
    518 		}
    519 #ifdef notyet
    520 		if (boothowto & RB_FASTBOOT) {
    521 			*flagsp++ = 'f';
    522 			options = 1;
    523 		}
    524 #endif
    525 
    526 		/*
    527 		 * Move out the flags (arg 1), if necessary.
    528 		 */
    529 		if (options != 0) {
    530 			*flagsp++ = '\0';
    531 			i = flagsp - flags;
    532 #ifdef DEBUG
    533 			printf("init: copying out flags `%s' %d\n", flags, i);
    534 #endif
    535 			(void)copyout((caddr_t)flags, (caddr_t)(ucp -= i), i);
    536 			arg1 = ucp;
    537 		}
    538 
    539 		/*
    540 		 * Move out the file name (also arg 0).
    541 		 */
    542 		i = strlen(path) + 1;
    543 #ifdef DEBUG
    544 		printf("init: copying out path `%s' %d\n", path, i);
    545 #endif
    546 		(void)copyout((caddr_t)path, (caddr_t)(ucp -= i), i);
    547 		arg0 = ucp;
    548 
    549 		/*
    550 		 * Move out the arg pointers.
    551 		 */
    552 		uap = (char **)((long)ucp & ~ALIGNBYTES);
    553 		(void)suword((caddr_t)--uap, 0);	/* terminator */
    554 		if (options != 0)
    555 			(void)suword((caddr_t)--uap, (long)arg1);
    556 		slash = strrchr(path, '/');
    557 		if (slash)
    558 			(void)suword((caddr_t)--uap,
    559 			    (long)arg0 + (slash + 1 - path));
    560 		else
    561 			(void)suword((caddr_t)--uap, (long)arg0);
    562 
    563 		/*
    564 		 * Point at the arguments.
    565 		 */
    566 		SCARG(&args, path) = arg0;
    567 		SCARG(&args, argp) = uap;
    568 		SCARG(&args, envp) = NULL;
    569 
    570 		/*
    571 		 * Now try to exec the program.  If can't for any reason
    572 		 * other than it doesn't exist, complain.
    573 		 */
    574 		error = sys_execve(p, &args, retval);
    575 		if (error == 0 || error == EJUSTRETURN)
    576 			return;
    577 		if (error != ENOENT)
    578 			printf("exec %s: error %d\n", path, error);
    579 	}
    580 	printf("init: not found\n");
    581 	panic("no init");
    582 }
    583 
    584 /* ARGSUSED */
    585 static void
    586 start_pagedaemon(arg)
    587 	void *arg;
    588 {
    589 
    590 	uvm_pageout();
    591 	/* NOTREACHED */
    592 }
    593 
    594 /* ARGSUSED */
    595 static void
    596 start_reaper(arg)
    597 	void *arg;
    598 {
    599 
    600 	reaper();
    601 	/* NOTREACHED */
    602 }
    603