Home | History | Annotate | Line # | Download | only in kern
init_main.c revision 1.120
      1 /*	$NetBSD: init_main.c,v 1.120 1998/03/27 01:52:01 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_uvm.h"
     46 
     47 #include "rnd.h"
     48 
     49 #include <sys/param.h>
     50 #include <sys/filedesc.h>
     51 #include <sys/errno.h>
     52 #include <sys/exec.h>
     53 #include <sys/kernel.h>
     54 #include <sys/mount.h>
     55 #include <sys/map.h>
     56 #include <sys/proc.h>
     57 #include <sys/resourcevar.h>
     58 #include <sys/signalvar.h>
     59 #include <sys/systm.h>
     60 #include <sys/vnode.h>
     61 #include <sys/tty.h>
     62 #include <sys/conf.h>
     63 #include <sys/disklabel.h>
     64 #include <sys/buf.h>
     65 #ifdef REAL_CLISTS
     66 #include <sys/clist.h>
     67 #endif
     68 #include <sys/device.h>
     69 #include <sys/protosw.h>
     70 #include <sys/reboot.h>
     71 #include <sys/user.h>
     72 #ifdef SYSVSHM
     73 #include <sys/shm.h>
     74 #endif
     75 #ifdef SYSVSEM
     76 #include <sys/sem.h>
     77 #endif
     78 #ifdef SYSVMSG
     79 #include <sys/msg.h>
     80 #endif
     81 #include <sys/domain.h>
     82 #include <sys/mbuf.h>
     83 #include <sys/namei.h>
     84 #if NRND > 0
     85 #include <sys/rnd.h>
     86 #endif
     87 
     88 #include <sys/syscall.h>
     89 #include <sys/syscallargs.h>
     90 
     91 #include <ufs/ufs/quota.h>
     92 
     93 #include <machine/cpu.h>
     94 
     95 #include <vm/vm.h>
     96 #include <vm/vm_pageout.h>
     97 
     98 #if defined(UVM)
     99 #include <uvm/uvm.h>
    100 #endif
    101 
    102 #include <net/if.h>
    103 #include <net/raw_cb.h>
    104 
    105 char	copyright[] = "\
    106 Copyright (c) 1996, 1997, 1998
    107     The NetBSD Foundation, Inc.  All rights reserved.
    108 Copyright (c) 1982, 1986, 1989, 1991, 1993
    109     The Regents of the University of California.  All rights reserved.
    110 
    111 ";
    112 
    113 /* Components of the first process -- never freed. */
    114 struct	session session0;
    115 struct	pgrp pgrp0;
    116 struct	proc proc0;
    117 struct	pcred cred0;
    118 struct	filedesc0 filedesc0;
    119 struct	plimit limit0;
    120 struct	vmspace vmspace0;
    121 struct	proc *curproc = &proc0;
    122 struct	proc *initproc;
    123 
    124 int	cmask = CMASK;
    125 extern	struct user *proc0paddr;
    126 
    127 struct	vnode *rootvp, *swapdev_vp;
    128 int	boothowto;
    129 struct	timeval boottime;
    130 struct	timeval runtime;
    131 
    132 static void check_console __P((struct proc *p));
    133 static void start_init __P((struct proc *));
    134 static void start_pagedaemon __P((struct proc *));
    135 void main __P((void *));
    136 
    137 extern char sigcode[], esigcode[];
    138 #ifdef SYSCALL_DEBUG
    139 extern char *syscallnames[];
    140 #endif
    141 
    142 struct emul emul_netbsd = {
    143 	"netbsd",
    144 	NULL,
    145 	sendsig,
    146 	SYS_syscall,
    147 	SYS_MAXSYSCALL,
    148 	sysent,
    149 #ifdef SYSCALL_DEBUG
    150 	syscallnames,
    151 #else
    152 	NULL,
    153 #endif
    154 	0,
    155 	copyargs,
    156 	setregs,
    157 	sigcode,
    158 	esigcode,
    159 };
    160 
    161 /*
    162  * System startup; initialize the world, create process 0, mount root
    163  * filesystem, and fork to create init and pagedaemon.  Most of the
    164  * hard work is done in the lower-level initialization routines including
    165  * startup(), which does memory initialization and autoconfiguration.
    166  */
    167 void
    168 main(framep)
    169 	void *framep;				/* XXX should go away */
    170 {
    171 	struct proc *p, *p2;
    172 	struct pdevinit *pdev;
    173 	int i, s, error;
    174 	extern struct pdevinit pdevinit[];
    175 	extern void roundrobin __P((void *));
    176 	extern void schedcpu __P((void *));
    177 	extern void disk_init __P((void));
    178 #if defined(NFSSERVER) || defined(NFS)
    179 	extern void nfs_init __P((void));
    180 #endif
    181 
    182 	/*
    183 	 * Initialize the current process pointer (curproc) before
    184 	 * any possible traps/probes to simplify trap processing.
    185 	 */
    186 	p = &proc0;
    187 	curproc = p;
    188 	/*
    189 	 * Attempt to find console and initialize
    190 	 * in case of early panic or other messages.
    191 	 */
    192 	consinit();
    193 	printf(copyright);
    194 
    195 #if defined(UVM)
    196 	uvm_init();
    197 #else
    198 	vm_mem_init();
    199 	kmeminit();
    200 #if defined(MACHINE_NEW_NONCONTIG)
    201 	vm_page_physrehash();
    202 #endif
    203 #endif /* UVM */
    204 	disk_init();		/* must come before autoconfiguration */
    205 	tty_init();		/* initialise tty list */
    206 #if NRND > 0
    207 	rnd_init();
    208 #endif
    209 	config_init();		/* init autoconfiguration data structures */
    210 	cpu_startup();
    211 
    212 	/*
    213 	 * Initialize process and pgrp structures.
    214 	 */
    215 	procinit();
    216 
    217 	/*
    218 	 * Create process 0 (the swapper).
    219 	 */
    220 	LIST_INSERT_HEAD(&allproc, p, p_list);
    221 	p->p_pgrp = &pgrp0;
    222 	LIST_INSERT_HEAD(PGRPHASH(0), &pgrp0, pg_hash);
    223 	LIST_INIT(&pgrp0.pg_members);
    224 	LIST_INSERT_HEAD(&pgrp0.pg_members, p, p_pglist);
    225 
    226 	pgrp0.pg_session = &session0;
    227 	session0.s_count = 1;
    228 	session0.s_sid = p->p_pid;
    229 	session0.s_leader = p;
    230 
    231 	p->p_flag = P_INMEM | P_SYSTEM;
    232 	p->p_stat = SRUN;
    233 	p->p_nice = NZERO;
    234 	p->p_emul = &emul_netbsd;
    235 	bcopy("swapper", p->p_comm, sizeof ("swapper"));
    236 
    237 	/* Create credentials. */
    238 	cred0.p_refcnt = 1;
    239 	p->p_cred = &cred0;
    240 	p->p_ucred = crget();
    241 	p->p_ucred->cr_ngroups = 1;	/* group 0 */
    242 
    243 	/* Create the file descriptor table. */
    244 	p->p_fd = &filedesc0.fd_fd;
    245 	fdinit1(&filedesc0);
    246 
    247 	/* Create the limits structures. */
    248 	p->p_limit = &limit0;
    249 	for (i = 0; i < sizeof(p->p_rlimit)/sizeof(p->p_rlimit[0]); i++)
    250 		limit0.pl_rlimit[i].rlim_cur =
    251 		    limit0.pl_rlimit[i].rlim_max = RLIM_INFINITY;
    252 	limit0.pl_rlimit[RLIMIT_NOFILE].rlim_cur = NOFILE;
    253 	limit0.pl_rlimit[RLIMIT_NPROC].rlim_cur = MAXUPRC;
    254 #if defined(UVM)
    255 	i = ptoa(uvmexp.free);
    256 #else
    257 	i = ptoa(cnt.v_free_count);
    258 #endif
    259 	limit0.pl_rlimit[RLIMIT_RSS].rlim_max = i;
    260 	limit0.pl_rlimit[RLIMIT_MEMLOCK].rlim_max = i;
    261 	limit0.pl_rlimit[RLIMIT_MEMLOCK].rlim_cur = i / 3;
    262 	limit0.p_refcnt = 1;
    263 
    264 	/*
    265 	 * Initialize proc0's vmspace, which uses the kernel pmap.
    266 	 * All kernel processes (which never have user space mappings)
    267 	 * share proc0's vmspace, and thus, the kernel pmap.
    268 	 */
    269 #if defined(UVM)
    270 	uvmspace_init(&vmspace0, pmap_kernel(), round_page(VM_MIN_ADDRESS),
    271 	    trunc_page(VM_MAX_ADDRESS), TRUE);
    272 #else
    273 	vmspace_init(&vmspace0, pmap_kernel(), round_page(VM_MIN_ADDRESS),
    274 	    trunc_page(VM_MAX_ADDRESS), TRUE);
    275 #endif
    276 	p->p_vmspace = &vmspace0;
    277 
    278 	p->p_addr = proc0paddr;				/* XXX */
    279 
    280 	/*
    281 	 * We continue to place resource usage info and signal
    282 	 * actions in the user struct so they're pageable.
    283 	 */
    284 	p->p_stats = &p->p_addr->u_stats;
    285 	p->p_sigacts = &p->p_addr->u_sigacts;
    286 
    287 	/*
    288 	 * Charge root for one process.
    289 	 */
    290 	(void)chgproccnt(0, 1);
    291 
    292 	rqinit();
    293 
    294 	/* Configure virtual memory system, set vm rlimits. */
    295 #if defined(UVM)
    296 	uvm_init_limits(p);
    297 #else
    298 	vm_init_limits(p);
    299 #endif
    300 
    301 	/* Initialize the file systems. */
    302 #if defined(NFSSERVER) || defined(NFS)
    303 	nfs_init();			/* initialize server/shared data */
    304 #endif
    305 	vfsinit();
    306 
    307 	/* Start real time and statistics clocks. */
    308 	initclocks();
    309 
    310 	/* Initialize mbuf's. */
    311 	mbinit();
    312 
    313 #ifdef REAL_CLISTS
    314 	/* Initialize clists. */
    315 	clist_init();
    316 #endif
    317 
    318 #ifdef SYSVSHM
    319 	/* Initialize System V style shared memory. */
    320 	shminit();
    321 #endif
    322 
    323 #ifdef SYSVSEM
    324 	/* Initialize System V style semaphores. */
    325 	seminit();
    326 #endif
    327 
    328 #ifdef SYSVMSG
    329 	/* Initialize System V style message queues. */
    330 	msginit();
    331 #endif
    332 
    333 	/* Attach pseudo-devices. */
    334 	for (pdev = pdevinit; pdev->pdev_attach != NULL; pdev++)
    335 		(*pdev->pdev_attach)(pdev->pdev_count);
    336 
    337 	/*
    338 	 * Initialize protocols.  Block reception of incoming packets
    339 	 * until everything is ready.
    340 	 */
    341 	s = splimp();
    342 	ifinit();
    343 	domaininit();
    344 	splx(s);
    345 
    346 #ifdef GPROF
    347 	/* Initialize kernel profiling. */
    348 	kmstartup();
    349 #endif
    350 
    351 	/* Kick off timeout driven events by calling first time. */
    352 	roundrobin(NULL);
    353 	schedcpu(NULL);
    354 
    355 	/* Determine the root and dump devices. */
    356 	cpu_rootconf();
    357 	cpu_dumpconf();
    358 
    359 	/* Mount the root file system. */
    360 	do {
    361 		domountroothook();
    362 		if ((error = vfs_mountroot())) {
    363 			printf("cannot mount root, error = %d\n", error);
    364 			boothowto |= RB_ASKNAME;
    365 			setroot(root_device,
    366 			    (rootdev != NODEV) ? DISKPART(rootdev) : 0, NULL);
    367 		}
    368 	} while (error != 0);
    369 	mountroothook_destroy();
    370 
    371 	mountlist.cqh_first->mnt_flag |= MNT_ROOTFS;
    372 	mountlist.cqh_first->mnt_op->vfs_refcount++;
    373 
    374 	/*
    375 	 * Get the vnode for '/'.  Set filedesc0.fd_fd.fd_cdir to
    376 	 * reference it.
    377 	 */
    378 	if (VFS_ROOT(mountlist.cqh_first, &rootvnode))
    379 		panic("cannot find root vnode");
    380 	filedesc0.fd_fd.fd_cdir = rootvnode;
    381 	VREF(filedesc0.fd_fd.fd_cdir);
    382 	VOP_UNLOCK(rootvnode, 0);
    383 	filedesc0.fd_fd.fd_rdir = NULL;
    384 #if defined(UVM)
    385 	uvm_swap_init();
    386 #else
    387 	swapinit();
    388 #endif
    389 
    390 	/*
    391 	 * Now can look at time, having had a chance to verify the time
    392 	 * from the file system.  Reset p->p_rtime as it may have been
    393 	 * munched in mi_switch() after the time got set.
    394 	 */
    395 	p->p_stats->p_start = runtime = mono_time = boottime = time;
    396 	p->p_rtime.tv_sec = p->p_rtime.tv_usec = 0;
    397 
    398 	/* Initialize signal state for process 0. */
    399 	siginit(p);
    400 
    401 	/* Create process 1 (init(8)). */
    402 	if (fork1(p, 0, NULL, &p2))
    403 		panic("fork init");
    404 	cpu_set_kpc(p2, start_init);
    405 
    406 	/* Create process 2 (the pageout daemon). */
    407 	if (fork1(p, FORK_SHAREVM, NULL, &p2))
    408 		panic("fork pager");
    409 	cpu_set_kpc(p2, start_pagedaemon);
    410 
    411 	/* The scheduler is an infinite loop. */
    412 #if defined(UVM)
    413 	uvm_scheduler();
    414 #else
    415 	scheduler();
    416 #endif
    417 	/* NOTREACHED */
    418 }
    419 
    420 static void
    421 check_console(p)
    422 	struct proc *p;
    423 {
    424 	struct nameidata nd;
    425 	int error;
    426 
    427 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_SYSSPACE, "/dev/console", p);
    428 	error = namei(&nd);
    429 	if (error == 0)
    430 		vrele(nd.ni_vp);
    431 	else if (error == ENOENT)
    432 		printf("warning: no /dev/console\n");
    433 	else
    434 		printf("warning: lookup /dev/console: error %d\n", error);
    435 }
    436 
    437 /*
    438  * List of paths to try when searching for "init".
    439  */
    440 static char *initpaths[] = {
    441 	"/sbin/init",
    442 	"/sbin/oinit",
    443 	"/sbin/init.bak",
    444 	NULL,
    445 };
    446 
    447 /*
    448  * Start the initial user process; try exec'ing each pathname in "initpaths".
    449  * The program is invoked with one argument containing the boot flags.
    450  */
    451 static void
    452 start_init(p)
    453 	struct proc *p;
    454 {
    455 	vm_offset_t addr;
    456 	struct sys_execve_args /* {
    457 		syscallarg(const char *) path;
    458 		syscallarg(char * const *) argp;
    459 		syscallarg(char * const *) envp;
    460 	} */ args;
    461 	int options, i, error;
    462 	register_t retval[2];
    463 	char flags[4], *flagsp;
    464 	char **pathp, *path, *ucp, **uap, *arg0, *arg1 = NULL;
    465 
    466 	/*
    467 	 * Now in process 1.
    468 	 */
    469 	initproc = p;
    470 
    471 	/*
    472 	 * This is not the right way to do this.  We really should
    473 	 * hand-craft a descriptor onto /dev/console to hand to init,
    474 	 * but that's a _lot_ more work, and the benefit from this easy
    475 	 * hack makes up for the "good is the enemy of the best" effect.
    476 	 */
    477 	check_console(p);
    478 
    479 	/*
    480 	 * Need just enough stack to hold the faked-up "execve()" arguments.
    481 	 */
    482 	addr = USRSTACK - PAGE_SIZE;
    483 #if defined(UVM)
    484 	if (uvm_map(&p->p_vmspace->vm_map, &addr, PAGE_SIZE,
    485                     NULL, UVM_UNKNOWN_OFFSET,
    486                     UVM_MAPFLAG(UVM_PROT_ALL, UVM_PROT_ALL, UVM_INH_COPY,
    487 		    UVM_ADV_NORMAL,
    488                     UVM_FLAG_FIXED|UVM_FLAG_OVERLAY|UVM_FLAG_COPYONW))
    489 		!= KERN_SUCCESS)
    490 		panic("init: couldn't allocate argument space");
    491 #else
    492 	if (vm_allocate(&p->p_vmspace->vm_map, &addr, (vm_size_t)PAGE_SIZE,
    493 	    FALSE) != 0)
    494 		panic("init: couldn't allocate argument space");
    495 #endif
    496 	p->p_vmspace->vm_maxsaddr = (caddr_t)addr;
    497 
    498 	for (pathp = &initpaths[0]; (path = *pathp) != NULL; pathp++) {
    499 		ucp = (char *)(addr + PAGE_SIZE);
    500 
    501 		/*
    502 		 * Construct the boot flag argument.
    503 		 */
    504 		flagsp = flags;
    505 		*flagsp++ = '-';
    506 		options = 0;
    507 
    508 		if (boothowto & RB_SINGLE) {
    509 			*flagsp++ = 's';
    510 			options = 1;
    511 		}
    512 #ifdef notyet
    513 		if (boothowto & RB_FASTBOOT) {
    514 			*flagsp++ = 'f';
    515 			options = 1;
    516 		}
    517 #endif
    518 
    519 		/*
    520 		 * Move out the flags (arg 1), if necessary.
    521 		 */
    522 		if (options != 0) {
    523 			*flagsp++ = '\0';
    524 			i = flagsp - flags;
    525 #ifdef DEBUG
    526 			printf("init: copying out flags `%s' %d\n", flags, i);
    527 #endif
    528 			(void)copyout((caddr_t)flags, (caddr_t)(ucp -= i), i);
    529 			arg1 = ucp;
    530 		}
    531 
    532 		/*
    533 		 * Move out the file name (also arg 0).
    534 		 */
    535 		i = strlen(path) + 1;
    536 #ifdef DEBUG
    537 		printf("init: copying out path `%s' %d\n", path, i);
    538 #endif
    539 		(void)copyout((caddr_t)path, (caddr_t)(ucp -= i), i);
    540 		arg0 = ucp;
    541 
    542 		/*
    543 		 * Move out the arg pointers.
    544 		 */
    545 		uap = (char **)((long)ucp & ~ALIGNBYTES);
    546 		(void)suword((caddr_t)--uap, 0);	/* terminator */
    547 		if (options != 0)
    548 			(void)suword((caddr_t)--uap, (long)arg1);
    549 		(void)suword((caddr_t)--uap, (long)arg0);
    550 
    551 		/*
    552 		 * Point at the arguments.
    553 		 */
    554 		SCARG(&args, path) = arg0;
    555 		SCARG(&args, argp) = uap;
    556 		SCARG(&args, envp) = NULL;
    557 
    558 		/*
    559 		 * Now try to exec the program.  If can't for any reason
    560 		 * other than it doesn't exist, complain.
    561 		 */
    562 		error = sys_execve(p, &args, retval);
    563 		if (error == 0 || error == EJUSTRETURN)
    564 			return;
    565 		if (error != ENOENT)
    566 			printf("exec %s: error %d\n", path, error);
    567 	}
    568 	printf("init: not found\n");
    569 	panic("no init");
    570 }
    571 
    572 static void
    573 start_pagedaemon(p)
    574 	struct proc *p;
    575 {
    576 
    577 	/*
    578 	 * Now in process 2.
    579 	 */
    580 	p->p_flag |= P_INMEM | P_SYSTEM;	/* XXX */
    581 	bcopy("pagedaemon", curproc->p_comm, sizeof ("pagedaemon"));
    582 #if defined(UVM)
    583 	uvm_pageout();
    584 #else
    585 	vm_pageout();
    586 #endif
    587 	/* NOTREACHED */
    588 }
    589