Home | History | Annotate | Line # | Download | only in kern
init_main.c revision 1.134
      1 /*	$NetBSD: init_main.c,v 1.134 1998/10/19 22:19:26 tron 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_uvm.h"
     47 #include "opt_sysv.h"
     48 
     49 #include "rnd.h"
     50 
     51 #include <sys/param.h>
     52 #include <sys/filedesc.h>
     53 #include <sys/file.h>
     54 #include <sys/errno.h>
     55 #include <sys/exec.h>
     56 #include <sys/kernel.h>
     57 #include <sys/mount.h>
     58 #include <sys/map.h>
     59 #include <sys/proc.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 #ifdef REAL_CLISTS
     69 #include <sys/clist.h>
     70 #endif
     71 #include <sys/device.h>
     72 #include <sys/socketvar.h>
     73 #include <sys/protosw.h>
     74 #include <sys/reboot.h>
     75 #include <sys/user.h>
     76 #ifdef SYSVSHM
     77 #include <sys/shm.h>
     78 #endif
     79 #ifdef SYSVSEM
     80 #include <sys/sem.h>
     81 #endif
     82 #ifdef SYSVMSG
     83 #include <sys/msg.h>
     84 #endif
     85 #include <sys/domain.h>
     86 #include <sys/mbuf.h>
     87 #include <sys/namei.h>
     88 #if NRND > 0
     89 #include <sys/rnd.h>
     90 #endif
     91 
     92 #include <sys/syscall.h>
     93 #include <sys/syscallargs.h>
     94 
     95 #include <ufs/ufs/quota.h>
     96 
     97 #include <machine/cpu.h>
     98 
     99 #include <vm/vm.h>
    100 #include <vm/vm_pageout.h>
    101 
    102 #if defined(UVM)
    103 #include <uvm/uvm.h>
    104 #endif
    105 
    106 #include <net/if.h>
    107 #include <net/raw_cb.h>
    108 
    109 char	copyright[] = "\
    110 Copyright (c) 1996, 1997, 1998
    111     The NetBSD Foundation, Inc.  All rights reserved.
    112 Copyright (c) 1982, 1986, 1989, 1991, 1993
    113     The Regents of the University of California.  All rights reserved.
    114 
    115 ";
    116 
    117 /* Components of the first process -- never freed. */
    118 struct	session session0;
    119 struct	pgrp pgrp0;
    120 struct	proc proc0;
    121 struct	pcred cred0;
    122 struct	filedesc0 filedesc0;
    123 struct	plimit limit0;
    124 struct	vmspace vmspace0;
    125 #ifndef curproc
    126 struct	proc *curproc = &proc0;
    127 #endif
    128 struct	proc *initproc;
    129 
    130 int	cmask = CMASK;
    131 extern	struct user *proc0paddr;
    132 
    133 struct	vnode *rootvp, *swapdev_vp;
    134 int	boothowto;
    135 struct	timeval boottime;
    136 struct	timeval runtime;
    137 
    138 static void check_console __P((struct proc *p));
    139 static void start_init __P((struct proc *));
    140 static void start_pagedaemon __P((struct proc *));
    141 static void start_reaper __P((struct proc *));
    142 void main __P((void));
    143 
    144 extern char sigcode[], esigcode[];
    145 #ifdef SYSCALL_DEBUG
    146 extern char *syscallnames[];
    147 #endif
    148 
    149 struct emul emul_netbsd = {
    150 	"netbsd",
    151 	NULL,
    152 	sendsig,
    153 	SYS_syscall,
    154 	SYS_MAXSYSCALL,
    155 	sysent,
    156 #ifdef SYSCALL_DEBUG
    157 	syscallnames,
    158 #else
    159 	NULL,
    160 #endif
    161 	0,
    162 	copyargs,
    163 	setregs,
    164 	sigcode,
    165 	esigcode,
    166 };
    167 
    168 /*
    169  * System startup; initialize the world, create process 0, mount root
    170  * filesystem, and fork to create init and pagedaemon.  Most of the
    171  * hard work is done in the lower-level initialization routines including
    172  * startup(), which does memory initialization and autoconfiguration.
    173  */
    174 void
    175 main()
    176 {
    177 	struct proc *p, *p2;
    178 	struct pdevinit *pdev;
    179 	int i, s, error;
    180 	extern struct pdevinit pdevinit[];
    181 	extern void roundrobin __P((void *));
    182 	extern void schedcpu __P((void *));
    183 	extern void disk_init __P((void));
    184 #if defined(NFSSERVER) || defined(NFS)
    185 	extern void nfs_init __P((void));
    186 #endif
    187 
    188 	/*
    189 	 * Initialize the current process pointer (curproc) before
    190 	 * any possible traps/probes to simplify trap processing.
    191 	 */
    192 	p = &proc0;
    193 	curproc = p;
    194 	/*
    195 	 * Attempt to find console and initialize
    196 	 * in case of early panic or other messages.
    197 	 */
    198 	consinit();
    199 	printf(copyright);
    200 
    201 #if defined(UVM)
    202 	uvm_init();
    203 #else
    204 	vm_mem_init();
    205 	kmeminit();
    206 #if defined(MACHINE_NEW_NONCONTIG)
    207 	vm_page_physrehash();
    208 #endif
    209 #endif /* UVM */
    210 
    211 	/*
    212 	 * Initialize mbuf's.  Do this now because we might attempt to
    213 	 * allocate mbufs or mbuf clusters during autoconfiguration.
    214 	 */
    215 	mbinit();
    216 
    217 	/* Initialize sockets. */
    218 	soinit();
    219 
    220 	disk_init();		/* must come before autoconfiguration */
    221 	tty_init();		/* initialise tty list */
    222 #if NRND > 0
    223 	rnd_init();
    224 #endif
    225 	config_init();		/* init autoconfiguration data structures */
    226 	cpu_startup();
    227 
    228 	/*
    229 	 * Initialize process and pgrp structures.
    230 	 */
    231 	procinit();
    232 
    233 	/*
    234 	 * Create process 0 (the swapper).
    235 	 */
    236 	LIST_INSERT_HEAD(&allproc, p, p_list);
    237 	p->p_pgrp = &pgrp0;
    238 	LIST_INSERT_HEAD(PGRPHASH(0), &pgrp0, pg_hash);
    239 	LIST_INIT(&pgrp0.pg_members);
    240 	LIST_INSERT_HEAD(&pgrp0.pg_members, p, p_pglist);
    241 
    242 	pgrp0.pg_session = &session0;
    243 	session0.s_count = 1;
    244 	session0.s_sid = p->p_pid;
    245 	session0.s_leader = p;
    246 
    247 	p->p_flag = P_INMEM | P_SYSTEM;
    248 	p->p_stat = SRUN;
    249 	p->p_nice = NZERO;
    250 	p->p_emul = &emul_netbsd;
    251 	memcpy(p->p_comm, "swapper", sizeof("swapper"));
    252 
    253 	/* Create credentials. */
    254 	cred0.p_refcnt = 1;
    255 	p->p_cred = &cred0;
    256 	p->p_ucred = crget();
    257 	p->p_ucred->cr_ngroups = 1;	/* group 0 */
    258 
    259 	/* Create the file descriptor table. */
    260 	finit();
    261 	p->p_fd = &filedesc0.fd_fd;
    262 	fdinit1(&filedesc0);
    263 
    264 	/* Create the limits structures. */
    265 	p->p_limit = &limit0;
    266 	for (i = 0; i < sizeof(p->p_rlimit)/sizeof(p->p_rlimit[0]); i++)
    267 		limit0.pl_rlimit[i].rlim_cur =
    268 		    limit0.pl_rlimit[i].rlim_max = RLIM_INFINITY;
    269 	limit0.pl_rlimit[RLIMIT_NOFILE].rlim_cur = NOFILE;
    270 	limit0.pl_rlimit[RLIMIT_NPROC].rlim_cur = MAXUPRC;
    271 #if defined(UVM)
    272 	i = ptoa(uvmexp.free);
    273 #else
    274 	i = ptoa(cnt.v_free_count);
    275 #endif
    276 	limit0.pl_rlimit[RLIMIT_RSS].rlim_max = i;
    277 	limit0.pl_rlimit[RLIMIT_MEMLOCK].rlim_max = i;
    278 	limit0.pl_rlimit[RLIMIT_MEMLOCK].rlim_cur = i / 3;
    279 	limit0.p_refcnt = 1;
    280 
    281 	/*
    282 	 * Initialize proc0's vmspace, which uses the kernel pmap.
    283 	 * All kernel processes (which never have user space mappings)
    284 	 * share proc0's vmspace, and thus, the kernel pmap.
    285 	 */
    286 #if defined(UVM)
    287 	uvmspace_init(&vmspace0, pmap_kernel(), round_page(VM_MIN_ADDRESS),
    288 	    trunc_page(VM_MAX_ADDRESS), TRUE);
    289 #else
    290 	vmspace_init(&vmspace0, pmap_kernel(), round_page(VM_MIN_ADDRESS),
    291 	    trunc_page(VM_MAX_ADDRESS), TRUE);
    292 #endif
    293 	p->p_vmspace = &vmspace0;
    294 
    295 	p->p_addr = proc0paddr;				/* XXX */
    296 
    297 	/*
    298 	 * We continue to place resource usage info and signal
    299 	 * actions in the user struct so they're pageable.
    300 	 */
    301 	p->p_stats = &p->p_addr->u_stats;
    302 	p->p_sigacts = &p->p_addr->u_sigacts;
    303 
    304 	/*
    305 	 * Charge root for one process.
    306 	 */
    307 	(void)chgproccnt(0, 1);
    308 
    309 	rqinit();
    310 
    311 	/* Configure virtual memory system, set vm rlimits. */
    312 #if defined(UVM)
    313 	uvm_init_limits(p);
    314 #else
    315 	vm_init_limits(p);
    316 #endif
    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 	/* Start real time and statistics clocks. */
    325 	initclocks();
    326 
    327 #ifdef REAL_CLISTS
    328 	/* Initialize clists. */
    329 	clist_init();
    330 #endif
    331 
    332 #ifdef SYSVSHM
    333 	/* Initialize System V style shared memory. */
    334 	shminit();
    335 #endif
    336 
    337 #ifdef SYSVSEM
    338 	/* Initialize System V style semaphores. */
    339 	seminit();
    340 #endif
    341 
    342 #ifdef SYSVMSG
    343 	/* Initialize System V style message queues. */
    344 	msginit();
    345 #endif
    346 
    347 	/* Attach pseudo-devices. */
    348 	for (pdev = pdevinit; pdev->pdev_attach != NULL; pdev++)
    349 		(*pdev->pdev_attach)(pdev->pdev_count);
    350 
    351 	/*
    352 	 * Initialize protocols.  Block reception of incoming packets
    353 	 * until everything is ready.
    354 	 */
    355 	s = splimp();
    356 	ifinit();
    357 	domaininit();
    358 	splx(s);
    359 
    360 #ifdef GPROF
    361 	/* Initialize kernel profiling. */
    362 	kmstartup();
    363 #endif
    364 
    365 	/* Kick off timeout driven events by calling first time. */
    366 	roundrobin(NULL);
    367 	schedcpu(NULL);
    368 
    369 	/* Determine the root and dump devices. */
    370 	cpu_rootconf();
    371 	cpu_dumpconf();
    372 
    373 	/* Mount the root file system. */
    374 	do {
    375 		domountroothook();
    376 		if ((error = vfs_mountroot())) {
    377 			printf("cannot mount root, error = %d\n", error);
    378 			boothowto |= RB_ASKNAME;
    379 			setroot(root_device,
    380 			    (rootdev != NODEV) ? DISKPART(rootdev) : 0, NULL);
    381 		}
    382 	} while (error != 0);
    383 	mountroothook_destroy();
    384 
    385 	mountlist.cqh_first->mnt_flag |= MNT_ROOTFS;
    386 	mountlist.cqh_first->mnt_op->vfs_refcount++;
    387 
    388 	/*
    389 	 * Get the vnode for '/'.  Set filedesc0.fd_fd.fd_cdir to
    390 	 * reference it.
    391 	 */
    392 	if (VFS_ROOT(mountlist.cqh_first, &rootvnode))
    393 		panic("cannot find root vnode");
    394 	filedesc0.fd_fd.fd_cdir = rootvnode;
    395 	VREF(filedesc0.fd_fd.fd_cdir);
    396 	VOP_UNLOCK(rootvnode, 0);
    397 	filedesc0.fd_fd.fd_rdir = NULL;
    398 #if defined(UVM)
    399 	uvm_swap_init();
    400 #else
    401 	swapinit();
    402 #endif
    403 
    404 	/*
    405 	 * Now can look at time, having had a chance to verify the time
    406 	 * from the file system.  Reset p->p_rtime as it may have been
    407 	 * munched in mi_switch() after the time got set.
    408 	 */
    409 	p->p_stats->p_start = runtime = mono_time = boottime = time;
    410 	p->p_rtime.tv_sec = p->p_rtime.tv_usec = 0;
    411 
    412 	/* Initialize signal state for process 0. */
    413 	siginit(p);
    414 
    415 	/* Create process 1 (init(8)). */
    416 	if (fork1(p, 0, NULL, &p2))
    417 		panic("fork init");
    418 	cpu_set_kpc(p2, start_init);
    419 
    420 	/* Create process 2 (the pageout daemon). */
    421 	if (fork1(p, FORK_SHAREVM, NULL, &p2))
    422 		panic("fork pager");
    423 	cpu_set_kpc(p2, start_pagedaemon);
    424 
    425 	/* Create process 3 (the process reaper). */
    426 	if (fork1(p, FORK_SHAREVM, NULL, &p2))
    427 		panic("fork reaper");
    428 	cpu_set_kpc(p2, start_reaper);
    429 
    430 	/* The scheduler is an infinite loop. */
    431 #if defined(UVM)
    432 	uvm_scheduler();
    433 #else
    434 	scheduler();
    435 #endif
    436 	/* NOTREACHED */
    437 }
    438 
    439 static void
    440 check_console(p)
    441 	struct proc *p;
    442 {
    443 	struct nameidata nd;
    444 	int error;
    445 
    446 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_SYSSPACE, "/dev/console", p);
    447 	error = namei(&nd);
    448 	if (error == 0)
    449 		vrele(nd.ni_vp);
    450 	else if (error == ENOENT)
    451 		printf("warning: no /dev/console\n");
    452 	else
    453 		printf("warning: lookup /dev/console: error %d\n", error);
    454 }
    455 
    456 /*
    457  * List of paths to try when searching for "init".
    458  */
    459 static char *initpaths[] = {
    460 	"/sbin/init",
    461 	"/sbin/oinit",
    462 	"/sbin/init.bak",
    463 	NULL,
    464 };
    465 
    466 /*
    467  * Start the initial user process; try exec'ing each pathname in "initpaths".
    468  * The program is invoked with one argument containing the boot flags.
    469  */
    470 static void
    471 start_init(p)
    472 	struct proc *p;
    473 {
    474 	vaddr_t addr;
    475 	struct sys_execve_args /* {
    476 		syscallarg(const char *) path;
    477 		syscallarg(char * const *) argp;
    478 		syscallarg(char * const *) envp;
    479 	} */ args;
    480 	int options, i, error;
    481 	register_t retval[2];
    482 	char flags[4], *flagsp;
    483 	char **pathp, *path, *ucp, **uap, *arg0, *arg1 = NULL;
    484 
    485 	/*
    486 	 * Now in process 1.
    487 	 */
    488 	initproc = p;
    489 
    490 	/*
    491 	 * This is not the right way to do this.  We really should
    492 	 * hand-craft a descriptor onto /dev/console to hand to init,
    493 	 * but that's a _lot_ more work, and the benefit from this easy
    494 	 * hack makes up for the "good is the enemy of the best" effect.
    495 	 */
    496 	check_console(p);
    497 
    498 	/*
    499 	 * Need just enough stack to hold the faked-up "execve()" arguments.
    500 	 */
    501 	addr = USRSTACK - PAGE_SIZE;
    502 #if defined(UVM)
    503 	if (uvm_map(&p->p_vmspace->vm_map, &addr, PAGE_SIZE,
    504                     NULL, UVM_UNKNOWN_OFFSET,
    505                     UVM_MAPFLAG(UVM_PROT_ALL, UVM_PROT_ALL, UVM_INH_COPY,
    506 		    UVM_ADV_NORMAL,
    507                     UVM_FLAG_FIXED|UVM_FLAG_OVERLAY|UVM_FLAG_COPYONW))
    508 		!= KERN_SUCCESS)
    509 		panic("init: couldn't allocate argument space");
    510 #else
    511 	if (vm_allocate(&p->p_vmspace->vm_map, &addr, (vsize_t)PAGE_SIZE,
    512 	    FALSE) != 0)
    513 		panic("init: couldn't allocate argument space");
    514 #endif
    515 	p->p_vmspace->vm_maxsaddr = (caddr_t)addr;
    516 
    517 	for (pathp = &initpaths[0]; (path = *pathp) != NULL; pathp++) {
    518 		ucp = (char *)(addr + PAGE_SIZE);
    519 
    520 		/*
    521 		 * Construct the boot flag argument.
    522 		 */
    523 		flagsp = flags;
    524 		*flagsp++ = '-';
    525 		options = 0;
    526 
    527 		if (boothowto & RB_SINGLE) {
    528 			*flagsp++ = 's';
    529 			options = 1;
    530 		}
    531 #ifdef notyet
    532 		if (boothowto & RB_FASTBOOT) {
    533 			*flagsp++ = 'f';
    534 			options = 1;
    535 		}
    536 #endif
    537 
    538 		/*
    539 		 * Move out the flags (arg 1), if necessary.
    540 		 */
    541 		if (options != 0) {
    542 			*flagsp++ = '\0';
    543 			i = flagsp - flags;
    544 #ifdef DEBUG
    545 			printf("init: copying out flags `%s' %d\n", flags, i);
    546 #endif
    547 			(void)copyout((caddr_t)flags, (caddr_t)(ucp -= i), i);
    548 			arg1 = ucp;
    549 		}
    550 
    551 		/*
    552 		 * Move out the file name (also arg 0).
    553 		 */
    554 		i = strlen(path) + 1;
    555 #ifdef DEBUG
    556 		printf("init: copying out path `%s' %d\n", path, i);
    557 #endif
    558 		(void)copyout((caddr_t)path, (caddr_t)(ucp -= i), i);
    559 		arg0 = ucp;
    560 
    561 		/*
    562 		 * Move out the arg pointers.
    563 		 */
    564 		uap = (char **)((long)ucp & ~ALIGNBYTES);
    565 		(void)suword((caddr_t)--uap, 0);	/* terminator */
    566 		if (options != 0)
    567 			(void)suword((caddr_t)--uap, (long)arg1);
    568 		(void)suword((caddr_t)--uap, (long)arg0);
    569 
    570 		/*
    571 		 * Point at the arguments.
    572 		 */
    573 		SCARG(&args, path) = arg0;
    574 		SCARG(&args, argp) = uap;
    575 		SCARG(&args, envp) = NULL;
    576 
    577 		/*
    578 		 * Now try to exec the program.  If can't for any reason
    579 		 * other than it doesn't exist, complain.
    580 		 */
    581 		error = sys_execve(p, &args, retval);
    582 		if (error == 0 || error == EJUSTRETURN)
    583 			return;
    584 		if (error != ENOENT)
    585 			printf("exec %s: error %d\n", path, error);
    586 	}
    587 	printf("init: not found\n");
    588 	panic("no init");
    589 }
    590 
    591 static void
    592 start_pagedaemon(p)
    593 	struct proc *p;
    594 {
    595 
    596 	/*
    597 	 * Now in process 2.
    598 	 */
    599 	p->p_flag |= P_INMEM | P_SYSTEM;	/* XXX */
    600 	memcpy(curproc->p_comm, "pagedaemon", sizeof("pagedaemon"));
    601 #if defined(UVM)
    602 	uvm_pageout();
    603 #else
    604 	vm_pageout();
    605 #endif
    606 	/* NOTREACHED */
    607 }
    608 
    609 static void
    610 start_reaper(p)
    611 	struct proc *p;
    612 {
    613 
    614 	/*
    615 	 * Now in process 3.
    616 	 */
    617 	p->p_flag |= P_INMEM | P_SYSTEM;	/* XXX */
    618 	memcpy(curproc->p_comm, "reaper", sizeof("reaper"));
    619 	reaper();
    620 	/* NOTREACHED */
    621 }
    622