Home | History | Annotate | Line # | Download | only in kern
init_main.c revision 1.98
      1 /*	$NetBSD: init_main.c,v 1.98 1997/03/26 22:38:45 gwr 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 check_console __P((struct proc *p));
    116 static void start_init __P((struct proc *));
    117 static void start_pagedaemon __P((struct proc *));
    118 void main __P((void *));
    119 
    120 #ifdef cpu_set_init_frame
    121 void *initframep;				/* XXX should go away */
    122 #endif
    123 
    124 extern char sigcode[], esigcode[];
    125 #ifdef SYSCALL_DEBUG
    126 extern char *syscallnames[];
    127 #endif
    128 
    129 struct emul emul_netbsd = {
    130 	"netbsd",
    131 	NULL,
    132 	sendsig,
    133 	SYS_syscall,
    134 	SYS_MAXSYSCALL,
    135 	sysent,
    136 #ifdef SYSCALL_DEBUG
    137 	syscallnames,
    138 #else
    139 	NULL,
    140 #endif
    141 	0,
    142 	copyargs,
    143 	setregs,
    144 	sigcode,
    145 	esigcode,
    146 };
    147 
    148 /*
    149  * System startup; initialize the world, create process 0, mount root
    150  * filesystem, and fork to create init and pagedaemon.  Most of the
    151  * hard work is done in the lower-level initialization routines including
    152  * startup(), which does memory initialization and autoconfiguration.
    153  */
    154 void
    155 main(framep)
    156 	void *framep;				/* XXX should go away */
    157 {
    158 	register struct proc *p;
    159 	register struct pdevinit *pdev;
    160 	register int i;
    161 	int s, error;
    162 	register_t rval[2];
    163 	extern struct pdevinit pdevinit[];
    164 	extern void roundrobin __P((void *));
    165 	extern void schedcpu __P((void *));
    166 	extern void disk_init __P((void));
    167 #if defined(NFSSERVER) || defined(NFS)
    168 	extern void nfs_init __P((void));
    169 #endif
    170 
    171 	/*
    172 	 * Initialize the current process pointer (curproc) before
    173 	 * any possible traps/probes to simplify trap processing.
    174 	 */
    175 	p = &proc0;
    176 	curproc = p;
    177 	/*
    178 	 * Attempt to find console and initialize
    179 	 * in case of early panic or other messages.
    180 	 */
    181 	consinit();
    182 	printf(copyright);
    183 
    184 	vm_mem_init();
    185 	kmeminit();
    186 	disk_init();		/* must come before autoconfiguration */
    187 	tty_init();		/* initialise tty list */
    188 	config_init();		/* init autoconfiguration data structures */
    189 	cpu_startup();
    190 
    191 	/*
    192 	 * Initialize process and pgrp structures.
    193 	 */
    194 	procinit();
    195 
    196 	/*
    197 	 * Create process 0 (the swapper).
    198 	 */
    199 	LIST_INSERT_HEAD(&allproc, p, p_list);
    200 	p->p_pgrp = &pgrp0;
    201 	LIST_INSERT_HEAD(PGRPHASH(0), &pgrp0, pg_hash);
    202 	LIST_INIT(&pgrp0.pg_members);
    203 	LIST_INSERT_HEAD(&pgrp0.pg_members, p, p_pglist);
    204 
    205 	pgrp0.pg_session = &session0;
    206 	session0.s_count = 1;
    207 	session0.s_leader = p;
    208 
    209 	p->p_flag = P_INMEM | P_SYSTEM;
    210 	p->p_stat = SRUN;
    211 	p->p_nice = NZERO;
    212 	p->p_emul = &emul_netbsd;
    213 	bcopy("swapper", p->p_comm, sizeof ("swapper"));
    214 
    215 	/* Create credentials. */
    216 	cred0.p_refcnt = 1;
    217 	p->p_cred = &cred0;
    218 	p->p_ucred = crget();
    219 	p->p_ucred->cr_ngroups = 1;	/* group 0 */
    220 
    221 	/* Create the file descriptor table. */
    222 	p->p_fd = &filedesc0.fd_fd;
    223 	filedesc0.fd_fd.fd_refcnt = 1;
    224 	filedesc0.fd_fd.fd_cmask = cmask;
    225 	filedesc0.fd_fd.fd_ofiles = filedesc0.fd_dfiles;
    226 	filedesc0.fd_fd.fd_ofileflags = filedesc0.fd_dfileflags;
    227 	filedesc0.fd_fd.fd_nfiles = NDFILE;
    228 
    229 	/* Create the limits structures. */
    230 	p->p_limit = &limit0;
    231 	for (i = 0; i < sizeof(p->p_rlimit)/sizeof(p->p_rlimit[0]); i++)
    232 		limit0.pl_rlimit[i].rlim_cur =
    233 		    limit0.pl_rlimit[i].rlim_max = RLIM_INFINITY;
    234 	limit0.pl_rlimit[RLIMIT_NOFILE].rlim_cur = NOFILE;
    235 	limit0.pl_rlimit[RLIMIT_NPROC].rlim_cur = MAXUPRC;
    236 	i = ptoa(cnt.v_free_count);
    237 	limit0.pl_rlimit[RLIMIT_RSS].rlim_max = i;
    238 	limit0.pl_rlimit[RLIMIT_MEMLOCK].rlim_max = i;
    239 	limit0.pl_rlimit[RLIMIT_MEMLOCK].rlim_cur = i / 3;
    240 	limit0.p_refcnt = 1;
    241 
    242 	/* Allocate a prototype map so we have something to fork. */
    243 	p->p_vmspace = &vmspace0;
    244 	vmspace0.vm_refcnt = 1;
    245 	pmap_pinit(&vmspace0.vm_pmap);
    246 	vm_map_init(&p->p_vmspace->vm_map, round_page(VM_MIN_ADDRESS),
    247 	    trunc_page(VM_MAX_ADDRESS), TRUE);
    248 	vmspace0.vm_map.pmap = &vmspace0.vm_pmap;
    249 	p->p_addr = proc0paddr;				/* XXX */
    250 
    251 	/*
    252 	 * We continue to place resource usage info and signal
    253 	 * actions in the user struct so they're pageable.
    254 	 */
    255 	p->p_stats = &p->p_addr->u_stats;
    256 	p->p_sigacts = &p->p_addr->u_sigacts;
    257 
    258 	/*
    259 	 * Charge root for one process.
    260 	 */
    261 	(void)chgproccnt(0, 1);
    262 
    263 	rqinit();
    264 
    265 	/* Configure virtual memory system, set vm rlimits. */
    266 	vm_init_limits(p);
    267 
    268 	/* Initialize the file systems. */
    269 #if defined(NFSSERVER) || defined(NFS)
    270 	nfs_init();			/* initialize server/shared data */
    271 #endif
    272 	vfsinit();
    273 
    274 	/* Start real time and statistics clocks. */
    275 	initclocks();
    276 
    277 	/* Initialize mbuf's. */
    278 	mbinit();
    279 
    280 #ifdef REAL_CLISTS
    281 	/* Initialize clists. */
    282 	clist_init();
    283 #endif
    284 
    285 #ifdef SYSVSHM
    286 	/* Initialize System V style shared memory. */
    287 	shminit();
    288 #endif
    289 
    290 #ifdef SYSVSEM
    291 	/* Initialize System V style semaphores. */
    292 	seminit();
    293 #endif
    294 
    295 #ifdef SYSVMSG
    296 	/* Initialize System V style message queues. */
    297 	msginit();
    298 #endif
    299 
    300 	/* Attach pseudo-devices. */
    301 	for (pdev = pdevinit; pdev->pdev_attach != NULL; pdev++)
    302 		(*pdev->pdev_attach)(pdev->pdev_count);
    303 
    304 	/*
    305 	 * Initialize protocols.  Block reception of incoming packets
    306 	 * until everything is ready.
    307 	 */
    308 	s = splimp();
    309 	ifinit();
    310 	domaininit();
    311 	splx(s);
    312 
    313 #ifdef GPROF
    314 	/* Initialize kernel profiling. */
    315 	kmstartup();
    316 #endif
    317 
    318 	/* Kick off timeout driven events by calling first time. */
    319 	roundrobin(NULL);
    320 	schedcpu(NULL);
    321 
    322 	/* Determine the root/swap/dump devices, etc. */
    323 	cpu_rootconf();
    324 	swapconf();
    325 	cpu_dumpconf();
    326 
    327 	/* Mount the root file system. */
    328 	do {
    329 		domountroothook();
    330 		if ((error = vfs_mountroot())) {
    331 			printf("cannot mount root, error = %d\n", error);
    332 			boothowto |= RB_ASKNAME;
    333 			setroot(root_device,
    334 			    (rootdev != NODEV) ? DISKPART(rootdev) : 0, NULL);
    335 		}
    336 	} while (error != 0);
    337 	mountroothook_destroy();
    338 
    339 	mountlist.cqh_first->mnt_flag |= MNT_ROOTFS;
    340 	mountlist.cqh_first->mnt_op->vfs_refcount++;
    341 
    342 	/* Get the vnode for '/'.  Set filedesc0.fd_fd.fd_cdir to reference it. */
    343 	if (VFS_ROOT(mountlist.cqh_first, &rootvnode))
    344 		panic("cannot find root vnode");
    345 	filedesc0.fd_fd.fd_cdir = rootvnode;
    346 	VREF(filedesc0.fd_fd.fd_cdir);
    347 	VOP_UNLOCK(rootvnode);
    348 	filedesc0.fd_fd.fd_rdir = NULL;
    349 	swapinit();
    350 
    351 	/*
    352 	 * Now can look at time, having had a chance to verify the time
    353 	 * from the file system.  Reset p->p_rtime as it may have been
    354 	 * munched in mi_switch() after the time got set.
    355 	 */
    356 	p->p_stats->p_start = runtime = mono_time = boottime = time;
    357 	p->p_rtime.tv_sec = p->p_rtime.tv_usec = 0;
    358 
    359 	/* Initialize signal state for process 0. */
    360 	siginit(p);
    361 
    362 	/* Create process 1 (init(8)). */
    363 	if (sys_fork(p, NULL, rval))
    364 		panic("fork init");
    365 #ifdef cpu_set_init_frame			/* XXX should go away */
    366 	if (rval[1]) {
    367 		/*
    368 		 * Now in process 1.
    369 		 */
    370 		initframep = framep;
    371 		start_init(curproc);
    372 		return;
    373 	}
    374 #else
    375 	cpu_set_kpc(pfind(1), start_init);
    376 #endif
    377 
    378 	/* Create process 2 (the pageout daemon). */
    379 	if (sys_fork(p, NULL, rval))
    380 		panic("fork pager");
    381 #ifdef cpu_set_init_frame			/* XXX should go away */
    382 	if (rval[1]) {
    383 		/*
    384 		 * Now in process 2.
    385 		 */
    386 		start_pagedaemon(curproc);
    387 	}
    388 #else
    389 	cpu_set_kpc(pfind(2), start_pagedaemon);
    390 #endif
    391 
    392 	/* The scheduler is an infinite loop. */
    393 	scheduler();
    394 	/* NOTREACHED */
    395 }
    396 
    397 static void
    398 check_console(p)
    399 	struct proc *p;
    400 {
    401 	struct nameidata nd;
    402 	int error;
    403 
    404 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_SYSSPACE, "/dev/console", p);
    405 	error = namei(&nd);
    406 	if (error == 0)
    407 		vrele(nd.ni_vp);
    408 	else if (error == ENOENT)
    409 		printf("warning: no /dev/console\n");
    410 	else
    411 		printf("warning: lookup /dev/console: error %d\n", error);
    412 }
    413 
    414 /*
    415  * List of paths to try when searching for "init".
    416  */
    417 static char *initpaths[] = {
    418 	"/sbin/init",
    419 	"/sbin/oinit",
    420 	"/sbin/init.bak",
    421 	NULL,
    422 };
    423 
    424 /*
    425  * Start the initial user process; try exec'ing each pathname in "initpaths".
    426  * The program is invoked with one argument containing the boot flags.
    427  */
    428 static void
    429 start_init(p)
    430 	struct proc *p;
    431 {
    432 	vm_offset_t addr;
    433 	struct sys_execve_args /* {
    434 		syscallarg(char *) path;
    435 		syscallarg(char * const *) argp;
    436 		syscallarg(char * const *) envp;
    437 	} */ args;
    438 	int options, i, error;
    439 	register_t retval[2];
    440 	char flags[4], *flagsp;
    441 	char **pathp, *path, *ucp, **uap, *arg0, *arg1 = NULL;
    442 
    443 	/*
    444 	 * Now in process 1.
    445 	 */
    446 	initproc = p;
    447 
    448 #ifdef cpu_set_init_frame			/* XXX should go away */
    449 	/*
    450 	 * We need to set the system call frame as if we were entered through
    451 	 * a syscall() so that when we call sys_execve() below, it will be able
    452 	 * to set the entry point (see setregs) when it tries to exec.  The
    453 	 * startup code in "locore.s" has allocated space for the frame and
    454 	 * passed a pointer to that space as main's argument.
    455 	 */
    456 	cpu_set_init_frame(p, initframep);
    457 #endif
    458 
    459 	/*
    460 	 * This is not the right way to do this.  We really should
    461 	 * hand-craft a descriptor onto /dev/console to hand to init,
    462 	 * but that's a _lot_ more work, and the benefit from this easy
    463 	 * hack makes up for the "good is the enemy of the best" effect.
    464 	 */
    465 	check_console(p);
    466 
    467 	/*
    468 	 * Need just enough stack to hold the faked-up "execve()" arguments.
    469 	 */
    470 	addr = USRSTACK - PAGE_SIZE;
    471 	if (vm_allocate(&p->p_vmspace->vm_map, &addr, (vm_size_t)PAGE_SIZE,
    472 	    FALSE) != 0)
    473 		panic("init: couldn't allocate argument space");
    474 	p->p_vmspace->vm_maxsaddr = (caddr_t)addr;
    475 
    476 	for (pathp = &initpaths[0]; (path = *pathp) != NULL; pathp++) {
    477 		ucp = (char *)(addr + PAGE_SIZE);
    478 
    479 		/*
    480 		 * Construct the boot flag argument.
    481 		 */
    482 		flagsp = flags;
    483 		*flagsp++ = '-';
    484 		options = 0;
    485 
    486 		if (boothowto & RB_SINGLE) {
    487 			*flagsp++ = 's';
    488 			options = 1;
    489 		}
    490 #ifdef notyet
    491 		if (boothowto & RB_FASTBOOT) {
    492 			*flagsp++ = 'f';
    493 			options = 1;
    494 		}
    495 #endif
    496 
    497 		/*
    498 		 * Move out the flags (arg 1), if necessary.
    499 		 */
    500 		if (options != 0) {
    501 			*flagsp++ = '\0';
    502 			i = flagsp - flags;
    503 #ifdef DEBUG
    504 			printf("init: copying out flags `%s' %d\n", flags, i);
    505 #endif
    506 			(void)copyout((caddr_t)flags, (caddr_t)(ucp -= i), i);
    507 			arg1 = ucp;
    508 		}
    509 
    510 		/*
    511 		 * Move out the file name (also arg 0).
    512 		 */
    513 		i = strlen(path) + 1;
    514 #ifdef DEBUG
    515 		printf("init: copying out path `%s' %d\n", path, i);
    516 #endif
    517 		(void)copyout((caddr_t)path, (caddr_t)(ucp -= i), i);
    518 		arg0 = ucp;
    519 
    520 		/*
    521 		 * Move out the arg pointers.
    522 		 */
    523 		uap = (char **)((long)ucp & ~ALIGNBYTES);
    524 		(void)suword((caddr_t)--uap, 0);	/* terminator */
    525 		if (options != 0)
    526 			(void)suword((caddr_t)--uap, (long)arg1);
    527 		(void)suword((caddr_t)--uap, (long)arg0);
    528 
    529 		/*
    530 		 * Point at the arguments.
    531 		 */
    532 		SCARG(&args, path) = arg0;
    533 		SCARG(&args, argp) = uap;
    534 		SCARG(&args, envp) = NULL;
    535 
    536 		/*
    537 		 * Now try to exec the program.  If can't for any reason
    538 		 * other than it doesn't exist, complain.
    539 		 */
    540 		if ((error = sys_execve(p, &args, retval)) == 0)
    541 			return;
    542 		if (error != ENOENT)
    543 			printf("exec %s: error %d\n", path, error);
    544 	}
    545 	printf("init: not found\n");
    546 	panic("no init");
    547 }
    548 
    549 static void
    550 start_pagedaemon(p)
    551 	struct proc *p;
    552 {
    553 
    554 	/*
    555 	 * Now in process 2.
    556 	 */
    557 	pageproc = p;
    558 	p->p_flag |= P_INMEM | P_SYSTEM;	/* XXX */
    559 	bcopy("pagedaemon", curproc->p_comm, sizeof ("pagedaemon"));
    560 	vm_pageout();
    561 	/* NOTREACHED */
    562 }
    563