Home | History | Annotate | Line # | Download | only in kern
init_main.c revision 1.95
      1 /*	$NetBSD: init_main.c,v 1.95 1997/01/31 02:25:47 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	8.9 (Berkeley) 1/21/94
     42  */
     43 
     44 #include <sys/param.h>
     45 #include <sys/filedesc.h>
     46 #include <sys/errno.h>
     47 #include <sys/exec.h>
     48 #include <sys/kernel.h>
     49 #include <sys/mount.h>
     50 #include <sys/map.h>
     51 #include <sys/proc.h>
     52 #include <sys/resourcevar.h>
     53 #include <sys/signalvar.h>
     54 #include <sys/systm.h>
     55 #include <sys/vnode.h>
     56 #include <sys/tty.h>
     57 #include <sys/conf.h>
     58 #include <sys/disklabel.h>
     59 #include <sys/buf.h>
     60 #ifdef REAL_CLISTS
     61 #include <sys/clist.h>
     62 #endif
     63 #include <sys/device.h>
     64 #include <sys/protosw.h>
     65 #include <sys/reboot.h>
     66 #include <sys/user.h>
     67 #ifdef SYSVSHM
     68 #include <sys/shm.h>
     69 #endif
     70 #ifdef SYSVSEM
     71 #include <sys/sem.h>
     72 #endif
     73 #ifdef SYSVMSG
     74 #include <sys/msg.h>
     75 #endif
     76 #include <sys/domain.h>
     77 #include <sys/mbuf.h>
     78 #include <sys/namei.h>
     79 
     80 #include <sys/syscall.h>
     81 #include <sys/syscallargs.h>
     82 
     83 #include <ufs/ufs/quota.h>
     84 
     85 #include <machine/cpu.h>
     86 
     87 #include <vm/vm.h>
     88 #include <vm/vm_pageout.h>
     89 
     90 #include <net/if.h>
     91 #include <net/raw_cb.h>
     92 
     93 char	copyright[] =
     94 "Copyright (c) 1982, 1986, 1989, 1991, 1993\n    The Regents of the University of California.  All rights reserved.\n\n";
     95 
     96 /* Components of the first process -- never freed. */
     97 struct	session session0;
     98 struct	pgrp pgrp0;
     99 struct	proc proc0;
    100 struct	pcred cred0;
    101 struct	filedesc0 filedesc0;
    102 struct	plimit limit0;
    103 struct	vmspace vmspace0;
    104 struct	proc *curproc = &proc0;
    105 struct	proc *initproc, *pageproc;
    106 
    107 int	cmask = CMASK;
    108 extern	struct user *proc0paddr;
    109 
    110 struct	vnode *rootvp, *swapdev_vp;
    111 int	boothowto;
    112 struct	timeval boottime;
    113 struct	timeval runtime;
    114 
    115 static void start_init __P((struct proc *));
    116 static void start_pagedaemon __P((struct proc *));
    117 void main __P((void *));
    118 
    119 #ifdef cpu_set_init_frame
    120 void *initframep;				/* XXX should go away */
    121 #endif
    122 
    123 extern char sigcode[], esigcode[];
    124 #ifdef SYSCALL_DEBUG
    125 extern char *syscallnames[];
    126 #endif
    127 
    128 struct emul emul_netbsd = {
    129 	"netbsd",
    130 	NULL,
    131 	sendsig,
    132 	SYS_syscall,
    133 	SYS_MAXSYSCALL,
    134 	sysent,
    135 #ifdef SYSCALL_DEBUG
    136 	syscallnames,
    137 #else
    138 	NULL,
    139 #endif
    140 	0,
    141 	copyargs,
    142 	setregs,
    143 	sigcode,
    144 	esigcode,
    145 };
    146 
    147 /*
    148  * System startup; initialize the world, create process 0, mount root
    149  * filesystem, and fork to create init and pagedaemon.  Most of the
    150  * hard work is done in the lower-level initialization routines including
    151  * startup(), which does memory initialization and autoconfiguration.
    152  */
    153 void
    154 main(framep)
    155 	void *framep;				/* XXX should go away */
    156 {
    157 	register struct proc *p;
    158 	register struct pdevinit *pdev;
    159 	register int i;
    160 	int s, error;
    161 	register_t rval[2];
    162 	extern struct pdevinit pdevinit[];
    163 	extern void roundrobin __P((void *));
    164 	extern void schedcpu __P((void *));
    165 	extern void disk_init __P((void));
    166 #if defined(NFSSERVER) || defined(NFS)
    167 	extern void nfs_init __P((void));
    168 #endif
    169 
    170 	/*
    171 	 * Initialize the current process pointer (curproc) before
    172 	 * any possible traps/probes to simplify trap processing.
    173 	 */
    174 	p = &proc0;
    175 	curproc = p;
    176 	/*
    177 	 * Attempt to find console and initialize
    178 	 * in case of early panic or other messages.
    179 	 */
    180 	consinit();
    181 	printf(copyright);
    182 
    183 	vm_mem_init();
    184 	kmeminit();
    185 	disk_init();		/* must come before autoconfiguration */
    186 	tty_init();		/* initialise tty list */
    187 	config_init();		/* init autoconfiguration data structures */
    188 	cpu_startup();
    189 
    190 	/*
    191 	 * Initialize process and pgrp structures.
    192 	 */
    193 	procinit();
    194 
    195 	/*
    196 	 * Create process 0 (the swapper).
    197 	 */
    198 	LIST_INSERT_HEAD(&allproc, p, p_list);
    199 	p->p_pgrp = &pgrp0;
    200 	LIST_INSERT_HEAD(PGRPHASH(0), &pgrp0, pg_hash);
    201 	LIST_INIT(&pgrp0.pg_members);
    202 	LIST_INSERT_HEAD(&pgrp0.pg_members, p, p_pglist);
    203 
    204 	pgrp0.pg_session = &session0;
    205 	session0.s_count = 1;
    206 	session0.s_leader = p;
    207 
    208 	p->p_flag = P_INMEM | P_SYSTEM;
    209 	p->p_stat = SRUN;
    210 	p->p_nice = NZERO;
    211 	p->p_emul = &emul_netbsd;
    212 	bcopy("swapper", p->p_comm, sizeof ("swapper"));
    213 
    214 	/* Create credentials. */
    215 	cred0.p_refcnt = 1;
    216 	p->p_cred = &cred0;
    217 	p->p_ucred = crget();
    218 	p->p_ucred->cr_ngroups = 1;	/* group 0 */
    219 
    220 	/* Create the file descriptor table. */
    221 	p->p_fd = &filedesc0.fd_fd;
    222 	filedesc0.fd_fd.fd_refcnt = 1;
    223 	filedesc0.fd_fd.fd_cmask = cmask;
    224 	filedesc0.fd_fd.fd_ofiles = filedesc0.fd_dfiles;
    225 	filedesc0.fd_fd.fd_ofileflags = filedesc0.fd_dfileflags;
    226 	filedesc0.fd_fd.fd_nfiles = NDFILE;
    227 
    228 	/* Create the limits structures. */
    229 	p->p_limit = &limit0;
    230 	for (i = 0; i < sizeof(p->p_rlimit)/sizeof(p->p_rlimit[0]); i++)
    231 		limit0.pl_rlimit[i].rlim_cur =
    232 		    limit0.pl_rlimit[i].rlim_max = RLIM_INFINITY;
    233 	limit0.pl_rlimit[RLIMIT_NOFILE].rlim_cur = NOFILE;
    234 	limit0.pl_rlimit[RLIMIT_NPROC].rlim_cur = MAXUPRC;
    235 	i = ptoa(cnt.v_free_count);
    236 	limit0.pl_rlimit[RLIMIT_RSS].rlim_max = i;
    237 	limit0.pl_rlimit[RLIMIT_MEMLOCK].rlim_max = i;
    238 	limit0.pl_rlimit[RLIMIT_MEMLOCK].rlim_cur = i / 3;
    239 	limit0.p_refcnt = 1;
    240 
    241 	/* Allocate a prototype map so we have something to fork. */
    242 	p->p_vmspace = &vmspace0;
    243 	vmspace0.vm_refcnt = 1;
    244 	pmap_pinit(&vmspace0.vm_pmap);
    245 	vm_map_init(&p->p_vmspace->vm_map, round_page(VM_MIN_ADDRESS),
    246 	    trunc_page(VM_MAX_ADDRESS), TRUE);
    247 	vmspace0.vm_map.pmap = &vmspace0.vm_pmap;
    248 	p->p_addr = proc0paddr;				/* XXX */
    249 
    250 	/*
    251 	 * We continue to place resource usage info and signal
    252 	 * actions in the user struct so they're pageable.
    253 	 */
    254 	p->p_stats = &p->p_addr->u_stats;
    255 	p->p_sigacts = &p->p_addr->u_sigacts;
    256 
    257 	/*
    258 	 * Charge root for one process.
    259 	 */
    260 	(void)chgproccnt(0, 1);
    261 
    262 	rqinit();
    263 
    264 	/* Configure virtual memory system, set vm rlimits. */
    265 	vm_init_limits(p);
    266 
    267 	/* Initialize the file systems. */
    268 #if defined(NFSSERVER) || defined(NFS)
    269 	nfs_init();			/* initialize server/shared data */
    270 #endif
    271 	vfsinit();
    272 
    273 	/* Start real time and statistics clocks. */
    274 	initclocks();
    275 
    276 	/* Initialize mbuf's. */
    277 	mbinit();
    278 
    279 #ifdef REAL_CLISTS
    280 	/* Initialize clists. */
    281 	clist_init();
    282 #endif
    283 
    284 #ifdef SYSVSHM
    285 	/* Initialize System V style shared memory. */
    286 	shminit();
    287 #endif
    288 
    289 #ifdef SYSVSEM
    290 	/* Initialize System V style semaphores. */
    291 	seminit();
    292 #endif
    293 
    294 #ifdef SYSVMSG
    295 	/* Initialize System V style message queues. */
    296 	msginit();
    297 #endif
    298 
    299 	/* Attach pseudo-devices. */
    300 	for (pdev = pdevinit; pdev->pdev_attach != NULL; pdev++)
    301 		(*pdev->pdev_attach)(pdev->pdev_count);
    302 
    303 	/*
    304 	 * Initialize protocols.  Block reception of incoming packets
    305 	 * until everything is ready.
    306 	 */
    307 	s = splimp();
    308 	ifinit();
    309 	domaininit();
    310 	splx(s);
    311 
    312 #ifdef GPROF
    313 	/* Initialize kernel profiling. */
    314 	kmstartup();
    315 #endif
    316 
    317 	/* Kick off timeout driven events by calling first time. */
    318 	roundrobin(NULL);
    319 	schedcpu(NULL);
    320 
    321 	/* Mount the root file system. */
    322 	do {
    323 		domountroothook();
    324 		if ((error = vfs_mountroot())) {
    325 			printf("cannot mount root, error = %d", error);
    326 			boothowto |= RB_ASKNAME;
    327 			setroot(root_device,
    328 			    (rootdev != NODEV) ? DISKPART(rootdev) : 0, NULL);
    329 		}
    330 	} while (error != 0);
    331 	mountroothook_destroy();
    332 
    333 	mountlist.cqh_first->mnt_flag |= MNT_ROOTFS;
    334 	mountlist.cqh_first->mnt_op->vfs_refcount++;
    335 
    336 	/* Get the vnode for '/'.  Set filedesc0.fd_fd.fd_cdir to reference it. */
    337 	if (VFS_ROOT(mountlist.cqh_first, &rootvnode))
    338 		panic("cannot find root vnode");
    339 	filedesc0.fd_fd.fd_cdir = rootvnode;
    340 	VREF(filedesc0.fd_fd.fd_cdir);
    341 	VOP_UNLOCK(rootvnode);
    342 	filedesc0.fd_fd.fd_rdir = NULL;
    343 	swapinit();
    344 
    345 	/*
    346 	 * Now can look at time, having had a chance to verify the time
    347 	 * from the file system.  Reset p->p_rtime as it may have been
    348 	 * munched in mi_switch() after the time got set.
    349 	 */
    350 	p->p_stats->p_start = runtime = mono_time = boottime = time;
    351 	p->p_rtime.tv_sec = p->p_rtime.tv_usec = 0;
    352 
    353 	/* Initialize signal state for process 0. */
    354 	siginit(p);
    355 
    356 	/* Create process 1 (init(8)). */
    357 	if (sys_fork(p, NULL, rval))
    358 		panic("fork init");
    359 #ifdef cpu_set_init_frame			/* XXX should go away */
    360 	if (rval[1]) {
    361 		/*
    362 		 * Now in process 1.
    363 		 */
    364 		initframep = framep;
    365 		start_init(curproc);
    366 		return;
    367 	}
    368 #else
    369 	cpu_set_kpc(pfind(1), start_init);
    370 #endif
    371 
    372 	/* Create process 2 (the pageout daemon). */
    373 	if (sys_fork(p, NULL, rval))
    374 		panic("fork pager");
    375 #ifdef cpu_set_init_frame			/* XXX should go away */
    376 	if (rval[1]) {
    377 		/*
    378 		 * Now in process 2.
    379 		 */
    380 		start_pagedaemon(curproc);
    381 	}
    382 #else
    383 	cpu_set_kpc(pfind(2), start_pagedaemon);
    384 #endif
    385 
    386 	/* The scheduler is an infinite loop. */
    387 	scheduler();
    388 	/* NOTREACHED */
    389 }
    390 
    391 static void
    392 check_console(p)
    393 	struct proc    *p
    394 {
    395 	struct nameidata nd;
    396 	int error;
    397 
    398 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_SYSSPACE, "/dev/console", p);
    399 	error = namei(&nd);
    400 	if (error)
    401 		if (error == ENOENT)
    402 			printf("warning: no /dev/console\n");
    403 		else
    404 			printf("warning: lookup /dev/console: error %d\n", error);
    405 	else
    406 		vrele(nd.ni_vp);
    407 }
    408 
    409 /*
    410  * List of paths to try when searching for "init".
    411  */
    412 static char *initpaths[] = {
    413 	"/sbin/init",
    414 	"/sbin/oinit",
    415 	"/sbin/init.bak",
    416 	NULL,
    417 };
    418 
    419 /*
    420  * Start the initial user process; try exec'ing each pathname in "initpaths".
    421  * The program is invoked with one argument containing the boot flags.
    422  */
    423 static void
    424 start_init(p)
    425 	struct proc *p;
    426 {
    427 	vm_offset_t addr;
    428 	struct sys_execve_args /* {
    429 		syscallarg(char *) path;
    430 		syscallarg(char * const *) argp;
    431 		syscallarg(char * const *) envp;
    432 	} */ args;
    433 	int options, i, error;
    434 	register_t retval[2];
    435 	char flags[4], *flagsp;
    436 	char **pathp, *path, *ucp, **uap, *arg0, *arg1 = NULL;
    437 
    438 	/*
    439 	 * Now in process 1.
    440 	 */
    441 	initproc = p;
    442 
    443 #ifdef cpu_set_init_frame			/* XXX should go away */
    444 	/*
    445 	 * We need to set the system call frame as if we were entered through
    446 	 * a syscall() so that when we call sys_execve() below, it will be able
    447 	 * to set the entry point (see setregs) when it tries to exec.  The
    448 	 * startup code in "locore.s" has allocated space for the frame and
    449 	 * passed a pointer to that space as main's argument.
    450 	 */
    451 	cpu_set_init_frame(p, initframep);
    452 #endif
    453 
    454 	/*
    455 	 * This is not the right way to do this.  We really should
    456 	 * hand-craft a descriptor onto /dev/console to hand to init,
    457 	 * but that's a _lot_ more work, and the benefit from this easy
    458 	 * hack makes up for the "good is the enemy of the best" effect.
    459 	 */
    460 	check_console(p);
    461 
    462 	/*
    463 	 * Need just enough stack to hold the faked-up "execve()" arguments.
    464 	 */
    465 	addr = USRSTACK - PAGE_SIZE;
    466 	if (vm_allocate(&p->p_vmspace->vm_map, &addr, (vm_size_t)PAGE_SIZE,
    467 	    FALSE) != 0)
    468 		panic("init: couldn't allocate argument space");
    469 	p->p_vmspace->vm_maxsaddr = (caddr_t)addr;
    470 
    471 	for (pathp = &initpaths[0]; (path = *pathp) != NULL; pathp++) {
    472 		ucp = (char *)(addr + PAGE_SIZE);
    473 
    474 		/*
    475 		 * Construct the boot flag argument.
    476 		 */
    477 		flagsp = flags;
    478 		*flagsp++ = '-';
    479 		options = 0;
    480 
    481 		if (boothowto & RB_SINGLE) {
    482 			*flagsp++ = 's';
    483 			options = 1;
    484 		}
    485 #ifdef notyet
    486 		if (boothowto & RB_FASTBOOT) {
    487 			*flagsp++ = 'f';
    488 			options = 1;
    489 		}
    490 #endif
    491 
    492 		/*
    493 		 * Move out the flags (arg 1), if necessary.
    494 		 */
    495 		if (options != 0) {
    496 			*flagsp++ = '\0';
    497 			i = flagsp - flags;
    498 #ifdef DEBUG
    499 			printf("init: copying out flags `%s' %d\n", flags, i);
    500 #endif
    501 			(void)copyout((caddr_t)flags, (caddr_t)(ucp -= i), i);
    502 			arg1 = ucp;
    503 		}
    504 
    505 		/*
    506 		 * Move out the file name (also arg 0).
    507 		 */
    508 		i = strlen(path) + 1;
    509 #ifdef DEBUG
    510 		printf("init: copying out path `%s' %d\n", path, i);
    511 #endif
    512 		(void)copyout((caddr_t)path, (caddr_t)(ucp -= i), i);
    513 		arg0 = ucp;
    514 
    515 		/*
    516 		 * Move out the arg pointers.
    517 		 */
    518 		uap = (char **)((long)ucp & ~ALIGNBYTES);
    519 		(void)suword((caddr_t)--uap, 0);	/* terminator */
    520 		if (options != 0)
    521 			(void)suword((caddr_t)--uap, (long)arg1);
    522 		(void)suword((caddr_t)--uap, (long)arg0);
    523 
    524 		/*
    525 		 * Point at the arguments.
    526 		 */
    527 		SCARG(&args, path) = arg0;
    528 		SCARG(&args, argp) = uap;
    529 		SCARG(&args, envp) = NULL;
    530 
    531 		/*
    532 		 * Now try to exec the program.  If can't for any reason
    533 		 * other than it doesn't exist, complain.
    534 		 */
    535 		if ((error = sys_execve(p, &args, retval)) == 0)
    536 			return;
    537 		if (error != ENOENT)
    538 			printf("exec %s: error %d\n", path, error);
    539 	}
    540 	printf("init: not found\n");
    541 	panic("no init");
    542 }
    543 
    544 static void
    545 start_pagedaemon(p)
    546 	struct proc *p;
    547 {
    548 
    549 	/*
    550 	 * Now in process 2.
    551 	 */
    552 	pageproc = p;
    553 	p->p_flag |= P_INMEM | P_SYSTEM;	/* XXX */
    554 	bcopy("pagedaemon", curproc->p_comm, sizeof ("pagedaemon"));
    555 	vm_pageout();
    556 	/* NOTREACHED */
    557 }
    558