xref: /openbsd-src/sys/kern/init_main.c (revision b725ae7711052a2233e31a66fefb8a752c388d7a)
1 /*	$OpenBSD: init_main.c,v 1.113 2004/04/01 00:27:51 tedu Exp $	*/
2 /*	$NetBSD: init_main.c,v 1.84.4.1 1996/06/02 09:08:06 mrg Exp $	*/
3 
4 /*
5  * Copyright (c) 1995 Christopher G. Demetriou.  All rights reserved.
6  * Copyright (c) 1982, 1986, 1989, 1991, 1992, 1993
7  *	The Regents of the University of California.  All rights reserved.
8  * (c) UNIX System Laboratories, Inc.
9  * All or some portions of this file are derived from material licensed
10  * to the University of California by American Telephone and Telegraph
11  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
12  * the permission of UNIX System Laboratories, Inc.
13  *
14  * Redistribution and use in source and binary forms, with or without
15  * modification, are permitted provided that the following conditions
16  * are met:
17  * 1. Redistributions of source code must retain the above copyright
18  *    notice, this list of conditions and the following disclaimer.
19  * 2. Redistributions in binary form must reproduce the above copyright
20  *    notice, this list of conditions and the following disclaimer in the
21  *    documentation and/or other materials provided with the distribution.
22  * 3. Neither the name of the University nor the names of its contributors
23  *    may be used to endorse or promote products derived from this software
24  *    without specific prior written permission.
25  *
26  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
27  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
28  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
29  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
30  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
31  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
32  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
33  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
34  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
35  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
36  * SUCH DAMAGE.
37  *
38  *	@(#)init_main.c	8.9 (Berkeley) 1/21/94
39  */
40 
41 #include <sys/param.h>
42 #include <sys/filedesc.h>
43 #include <sys/file.h>
44 #include <sys/errno.h>
45 #include <sys/exec.h>
46 #include <sys/kernel.h>
47 #include <sys/kthread.h>
48 #include <sys/mount.h>
49 #include <sys/proc.h>
50 #include <sys/resourcevar.h>
51 #include <sys/signalvar.h>
52 #include <sys/systm.h>
53 #include <sys/namei.h>
54 #include <sys/vnode.h>
55 #include <sys/tty.h>
56 #include <sys/conf.h>
57 #include <sys/buf.h>
58 #include <sys/device.h>
59 #include <sys/socketvar.h>
60 #include <sys/lockf.h>
61 #include <sys/protosw.h>
62 #include <sys/reboot.h>
63 #include <sys/user.h>
64 #ifdef SYSVSHM
65 #include <sys/shm.h>
66 #endif
67 #ifdef SYSVSEM
68 #include <sys/sem.h>
69 #endif
70 #ifdef SYSVMSG
71 #include <sys/msg.h>
72 #endif
73 #include <sys/domain.h>
74 #include <sys/mbuf.h>
75 #include <sys/pipe.h>
76 
77 #include <sys/syscall.h>
78 #include <sys/syscallargs.h>
79 
80 #include <dev/rndvar.h>
81 
82 #include <ufs/ufs/quota.h>
83 
84 #include <machine/cpu.h>
85 
86 #include <uvm/uvm.h>
87 
88 #include <net/if.h>
89 #include <net/raw_cb.h>
90 
91 #if defined(CRYPTO)
92 #include <crypto/cryptodev.h>
93 #include <crypto/cryptosoft.h>
94 #endif
95 
96 #if defined(NFSSERVER) || defined(NFSCLIENT)
97 extern void nfs_init(void);
98 #endif
99 
100 const char	copyright[] =
101 "Copyright (c) 1982, 1986, 1989, 1991, 1993\n"
102 "\tThe Regents of the University of California.  All rights reserved.\n"
103 "Copyright (c) 1995-2004 OpenBSD. All rights reserved.  http://www.OpenBSD.org\n";
104 
105 /* Components of the first process -- never freed. */
106 struct	session session0;
107 struct	pgrp pgrp0;
108 struct	proc proc0;
109 struct	pcred cred0;
110 struct	plimit limit0;
111 struct	vmspace vmspace0;
112 struct	sigacts sigacts0;
113 #ifndef curproc
114 struct	proc *curproc;
115 #endif
116 struct	proc *initproc;
117 
118 int	cmask = CMASK;
119 extern	struct user *proc0paddr;
120 
121 void	(*md_diskconf)(void) = NULL;
122 struct	vnode *rootvp, *swapdev_vp;
123 int	boothowto;
124 struct	timeval boottime;
125 struct	timeval runtime;
126 
127 #if !defined(NO_PROPOLICE)
128 long	__guard[8];
129 #endif
130 
131 /* XXX return int so gcc -Werror won't complain */
132 int	main(void *);
133 void	check_console(struct proc *);
134 void	start_init(void *);
135 void	start_cleaner(void *);
136 void	start_update(void *);
137 void	start_reaper(void *);
138 void	start_crypto(void *);
139 void	init_exec(void);
140 void	kqueue_init(void);
141 
142 extern char sigcode[], esigcode[];
143 #ifdef SYSCALL_DEBUG
144 extern char *syscallnames[];
145 #endif
146 
147 struct emul emul_native = {
148 	"native",
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 	NULL,
163 	sigcode,
164 	esigcode,
165 	EMUL_ENABLED | EMUL_NATIVE,
166 };
167 
168 
169 /*
170  * System startup; initialize the world, create process 0, mount root
171  * filesystem, and fork to create init and pagedaemon.  Most of the
172  * hard work is done in the lower-level initialization routines including
173  * startup(), which does memory initialization and autoconfiguration.
174  */
175 /* XXX return int, so gcc -Werror won't complain */
176 int
177 main(framep)
178 	void *framep;				/* XXX should go away */
179 {
180 	struct proc *p;
181 	struct pdevinit *pdev;
182 	struct timeval rtv;
183 	quad_t lim;
184 	int s, i;
185 	register_t rval[2];
186 	extern struct pdevinit pdevinit[];
187 	extern void scheduler_start(void);
188 	extern void disk_init(void);
189 	extern void endtsleep(void *);
190 	extern void realitexpire(void *);
191 
192 	/*
193 	 * Initialize the current process pointer (curproc) before
194 	 * any possible traps/probes to simplify trap processing.
195 	 */
196 	curproc = p = &proc0;
197 
198 	/*
199 	 * Initialize timeouts.
200 	 */
201 	timeout_startup();
202 
203 	/*
204 	 * Attempt to find console and initialize
205 	 * in case of early panic or other messages.
206 	 */
207 	config_init();		/* init autoconfiguration data structures */
208 	consinit();
209 	printf("%s\n", copyright);
210 
211 	uvm_init();
212 	disk_init();		/* must come before autoconfiguration */
213 	tty_init();		/* initialise tty's */
214 	cpu_startup();
215 
216 	/*
217 	 * Initialize mbuf's.  Do this now because we might attempt to
218 	 * allocate mbufs or mbuf clusters during autoconfiguration.
219 	 */
220 	mbinit();
221 
222 	/* Initalize sockets. */
223 	soinit();
224 
225 	/* Initialize sysctls (must be done before any processes run) */
226 	sysctl_init();
227 
228 	/*
229 	 * Initialize process and pgrp structures.
230 	 */
231 	procinit();
232 
233 	/* Initialize file locking. */
234 	lf_init();
235 
236 	/*
237 	 * Initialize filedescriptors.
238 	 */
239 	filedesc_init();
240 
241 	/*
242 	 * Initialize pipes.
243 	 */
244 	pipe_init();
245 
246 	/*
247 	 * Initialize kqueues.
248 	 */
249 	kqueue_init();
250 
251 	/*
252 	 * Create process 0 (the swapper).
253 	 */
254 	LIST_INSERT_HEAD(&allproc, p, p_list);
255 	p->p_pgrp = &pgrp0;
256 	LIST_INSERT_HEAD(PIDHASH(0), p, p_hash);
257 	LIST_INSERT_HEAD(PGRPHASH(0), &pgrp0, pg_hash);
258 	LIST_INIT(&pgrp0.pg_members);
259 	LIST_INSERT_HEAD(&pgrp0.pg_members, p, p_pglist);
260 
261 	pgrp0.pg_session = &session0;
262 	session0.s_count = 1;
263 	session0.s_leader = p;
264 
265 	p->p_flag = P_INMEM | P_SYSTEM | P_NOCLDWAIT;
266 	p->p_stat = SRUN;
267 	p->p_nice = NZERO;
268 	p->p_emul = &emul_native;
269 	bcopy("swapper", p->p_comm, sizeof ("swapper"));
270 
271 	/* Init timeouts. */
272 	timeout_set(&p->p_sleep_to, endtsleep, p);
273 	timeout_set(&p->p_realit_to, realitexpire, p);
274 
275 	/* Create credentials. */
276 	cred0.p_refcnt = 1;
277 	p->p_cred = &cred0;
278 	p->p_ucred = crget();
279 	p->p_ucred->cr_ngroups = 1;	/* group 0 */
280 
281 	/* Initialize signal state for process 0. */
282 	signal_init();
283 	p->p_sigacts = &sigacts0;
284 	siginit(p);
285 
286 	/* Create the file descriptor table. */
287 	p->p_fd = fdinit(NULL);
288 
289 	/* Create the limits structures. */
290 	p->p_limit = &limit0;
291 	for (i = 0; i < sizeof(p->p_rlimit)/sizeof(p->p_rlimit[0]); i++)
292 		limit0.pl_rlimit[i].rlim_cur =
293 		    limit0.pl_rlimit[i].rlim_max = RLIM_INFINITY;
294 	limit0.pl_rlimit[RLIMIT_NOFILE].rlim_cur = NOFILE;
295 	limit0.pl_rlimit[RLIMIT_NOFILE].rlim_max = MIN(NOFILE_MAX,
296 	    (maxfiles - NOFILE > NOFILE) ?  maxfiles - NOFILE : NOFILE);
297 	limit0.pl_rlimit[RLIMIT_NPROC].rlim_cur = MAXUPRC;
298 	lim = ptoa(uvmexp.free);
299 	limit0.pl_rlimit[RLIMIT_RSS].rlim_max = lim;
300 	limit0.pl_rlimit[RLIMIT_MEMLOCK].rlim_max = lim;
301 	limit0.pl_rlimit[RLIMIT_MEMLOCK].rlim_cur = lim / 3;
302 	limit0.p_refcnt = 1;
303 
304 	/* Allocate a prototype map so we have something to fork. */
305 	uvmspace_init(&vmspace0, pmap_kernel(), round_page(VM_MIN_ADDRESS),
306 	    trunc_page(VM_MAX_ADDRESS), TRUE);
307 	p->p_vmspace = &vmspace0;
308 
309 	p->p_addr = proc0paddr;				/* XXX */
310 
311 	/*
312 	 * We continue to place resource usage info in the
313 	 * user struct so they're pageable.
314 	 */
315 	p->p_stats = &p->p_addr->u_stats;
316 
317 	/*
318 	 * Charge root for one process.
319 	 */
320 	(void)chgproccnt(0, 1);
321 
322 	/* Initialize run queues */
323 	rqinit();
324 
325 	/* Configure the devices */
326 	cpu_configure();
327 
328 	/* Configure virtual memory system, set vm rlimits. */
329 	uvm_init_limits(p);
330 
331 	/* Initialize the file systems. */
332 #if defined(NFSSERVER) || defined(NFSCLIENT)
333 	nfs_init();			/* initialize server/shared data */
334 #endif
335 	vfsinit();
336 
337 	/* Start real time and statistics clocks. */
338 	initclocks();
339 
340 #ifdef SYSVSHM
341 	/* Initialize System V style shared memory. */
342 	shminit();
343 #endif
344 
345 #ifdef SYSVSEM
346 	/* Initialize System V style semaphores. */
347 	seminit();
348 #endif
349 
350 #ifdef SYSVMSG
351 	/* Initialize System V style message queues. */
352 	msginit();
353 #endif
354 
355 	/* Attach pseudo-devices. */
356 	randomattach();
357 	for (pdev = pdevinit; pdev->pdev_attach != NULL; pdev++)
358 		if (pdev->pdev_count > 0)
359 			(*pdev->pdev_attach)(pdev->pdev_count);
360 
361 #ifdef CRYPTO
362 	swcr_init();
363 #endif /* CRYPTO */
364 
365 	/*
366 	 * Initialize protocols.  Block reception of incoming packets
367 	 * until everything is ready.
368 	 */
369 	s = splimp();
370 	ifinit();
371 	domaininit();
372 	if_attachdomain();
373 	splx(s);
374 
375 #ifdef GPROF
376 	/* Initialize kernel profiling. */
377 	kmstartup();
378 #endif
379 
380 #if !defined(NO_PROPOLICE)
381 	arc4random_bytes(__guard, sizeof(__guard));
382 #endif
383 
384 	/* init exec and emul */
385 	init_exec();
386 
387 	/* Start the scheduler */
388 	scheduler_start();
389 
390 	dostartuphooks();
391 
392 	/* Configure root/swap devices */
393 	if (md_diskconf)
394 		(*md_diskconf)();
395 
396 	/* Mount the root file system. */
397 	if (vfs_mountroot())
398 		panic("cannot mount root");
399 	CIRCLEQ_FIRST(&mountlist)->mnt_flag |= MNT_ROOTFS;
400 
401 	/* Get the vnode for '/'.  Set p->p_fd->fd_cdir to reference it. */
402 	if (VFS_ROOT(mountlist.cqh_first, &rootvnode))
403 		panic("cannot find root vnode");
404 	p->p_fd->fd_cdir = rootvnode;
405 	VREF(p->p_fd->fd_cdir);
406 	VOP_UNLOCK(rootvnode, 0, p);
407 	p->p_fd->fd_rdir = NULL;
408 
409 	uvm_swap_init();
410 
411 	/*
412 	 * Now can look at time, having had a chance to verify the time
413 	 * from the file system.  Reset p->p_rtime as it may have been
414 	 * munched in mi_switch() after the time got set.
415 	 */
416 	p->p_stats->p_start = runtime = mono_time = boottime = time;
417 	p->p_rtime.tv_sec = p->p_rtime.tv_usec = 0;
418 
419 	/* Create process 1 (init(8)). */
420 	if (fork1(p, SIGCHLD, FORK_FORK, NULL, 0, start_init, NULL, rval))
421 		panic("fork init");
422 
423 	/* Create process 2, the pageout daemon kernel thread. */
424 	if (kthread_create(uvm_pageout, NULL, NULL, "pagedaemon"))
425 		panic("fork pagedaemon");
426 
427 	/* Create process 3, the reaper daemon kernel thread. */
428 	if (kthread_create(start_reaper, NULL, NULL, "reaper"))
429 		panic("fork reaper");
430 
431 	/* Create process 4, the cleaner daemon kernel thread. */
432 	if (kthread_create(start_cleaner, NULL, NULL, "cleaner"))
433 		panic("fork cleaner");
434 
435 	/* Create process 5, the update daemon kernel thread. */
436 	if (kthread_create(start_update, NULL, NULL, "update"))
437 		panic("fork update");
438 
439 	/* Create process 6, the aiodone daemon kernel thread. */
440 	if (kthread_create(uvm_aiodone_daemon, NULL, NULL, "aiodoned"))
441 		panic("fork aiodoned");
442 
443 #ifdef CRYPTO
444 	/* Create process 7, the crypto kernel thread. */
445 	if (kthread_create(start_crypto, NULL, NULL, "crypto"))
446 		panic("crypto thread");
447 #endif /* CRYPTO */
448 
449 	/* Create any other deferred kernel threads. */
450 	kthread_run_deferred_queue();
451 
452 	microtime(&rtv);
453 	srandom((u_long)(rtv.tv_sec ^ rtv.tv_usec));
454 
455 	randompid = 1;
456 	/* The scheduler is an infinite loop. */
457 	uvm_scheduler();
458 	/* NOTREACHED */
459 }
460 
461 /*
462  * List of paths to try when searching for "init".
463  */
464 static char *initpaths[] = {
465 	"/sbin/init",
466 	"/sbin/oinit",
467 	"/sbin/init.bak",
468 	NULL,
469 };
470 
471 void
472 check_console(p)
473 	struct proc *p;
474 {
475 	struct nameidata nd;
476 	int error;
477 
478 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_SYSSPACE, "/dev/console", p);
479 	error = namei(&nd);
480 	if (error) {
481 		if (error == ENOENT)
482 			printf("warning: /dev/console does not exist\n");
483 		else
484 			printf("warning: /dev/console error %d\n", error);
485 	} else
486 		vrele(nd.ni_vp);
487 }
488 
489 /*
490  * Start the initial user process; try exec'ing each pathname in "initpaths".
491  * The program is invoked with one argument containing the boot flags.
492  */
493 void
494 start_init(arg)
495 	void *arg;
496 {
497 	struct proc *p = arg;
498 	vaddr_t addr;
499 	struct sys_execve_args /* {
500 		syscallarg(const char *) path;
501 		syscallarg(char *const *) argp;
502 		syscallarg(char *const *) envp;
503 	} */ args;
504 	int options, error;
505 	long i;
506 	register_t retval[2];
507 	char flags[4], *flagsp;
508 	char **pathp, *path, *ucp, **uap, *arg0, *arg1 = NULL;
509 
510 	initproc = p;
511 
512 	/*
513 	 * Now in process 1.
514 	 */
515 	check_console(p);
516 
517 	/*
518 	 * Need just enough stack to hold the faked-up "execve()" arguments.
519 	 */
520 #ifdef MACHINE_STACK_GROWS_UP
521 	addr = USRSTACK;
522 #else
523 	addr = USRSTACK - PAGE_SIZE;
524 #endif
525 	if (uvm_map(&p->p_vmspace->vm_map, &addr, PAGE_SIZE,
526 	    NULL, UVM_UNKNOWN_OFFSET, 0,
527 	    UVM_MAPFLAG(UVM_PROT_RW, UVM_PROT_ALL, UVM_INH_COPY,
528 	    UVM_ADV_NORMAL, UVM_FLAG_FIXED|UVM_FLAG_OVERLAY|UVM_FLAG_COPYONW)))
529 		panic("init: couldn't allocate argument space");
530 	p->p_vmspace->vm_maxsaddr = (caddr_t)addr;
531 
532 	for (pathp = &initpaths[0]; (path = *pathp) != NULL; pathp++) {
533 #ifdef MACHINE_STACK_GROWS_UP
534 		ucp = (char *)addr;
535 #else
536 		ucp = (char *)(addr + PAGE_SIZE);
537 #endif
538 		/*
539 		 * Construct the boot flag argument.
540 		 */
541 		flagsp = flags;
542 		*flagsp++ = '-';
543 		options = 0;
544 
545 		if (boothowto & RB_SINGLE) {
546 			*flagsp++ = 's';
547 			options = 1;
548 		}
549 #ifdef notyet
550 		if (boothowto & RB_FASTBOOT) {
551 			*flagsp++ = 'f';
552 			options = 1;
553 		}
554 #endif
555 
556 		/*
557 		 * Move out the flags (arg 1), if necessary.
558 		 */
559 		if (options != 0) {
560 			*flagsp++ = '\0';
561 			i = flagsp - flags;
562 #ifdef DEBUG
563 			printf("init: copying out flags `%s' %d\n", flags, i);
564 #endif
565 #ifdef MACHINE_STACK_GROWS_UP
566 			arg1 = ucp;
567 			(void)copyout((caddr_t)flags, (caddr_t)ucp, i);
568 			ucp += i;
569 #else
570 			(void)copyout((caddr_t)flags, (caddr_t)(ucp -= i), i);
571 			arg1 = ucp;
572 #endif
573 		}
574 
575 		/*
576 		 * Move out the file name (also arg 0).
577 		 */
578 		i = strlen(path) + 1;
579 #ifdef DEBUG
580 		printf("init: copying out path `%s' %d\n", path, i);
581 #endif
582 #ifdef MACHINE_STACK_GROWS_UP
583 		arg0 = ucp;
584 		(void)copyout((caddr_t)path, (caddr_t)ucp, i);
585 		ucp += i;
586 		ucp = (caddr_t)ALIGN((u_long)ucp);
587 		uap = (char **)ucp + 3;
588 #else
589 		(void)copyout((caddr_t)path, (caddr_t)(ucp -= i), i);
590 		arg0 = ucp;
591 		uap = (char **)((u_long)ucp & ~ALIGNBYTES);
592 #endif
593 
594 		/*
595 		 * Move out the arg pointers.
596 		 */
597 		i = 0;
598 		copyout(&i, (caddr_t)--uap, sizeof(register_t)); /* terminator */
599 		if (options != 0)
600 			copyout(&arg1, (caddr_t)--uap, sizeof(register_t));
601 		copyout(&arg0, (caddr_t)--uap, sizeof(register_t));
602 
603 		/*
604 		 * Point at the arguments.
605 		 */
606 		SCARG(&args, path) = arg0;
607 		SCARG(&args, argp) = uap;
608 		SCARG(&args, envp) = NULL;
609 
610 		/*
611 		 * Now try to exec the program.  If can't for any reason
612 		 * other than it doesn't exist, complain.
613 		 */
614 		if ((error = sys_execve(p, &args, retval)) == 0)
615 			return;
616 		if (error != ENOENT)
617 			printf("exec %s: error %d\n", path, error);
618 	}
619 	printf("init: not found\n");
620 	panic("no init");
621 }
622 
623 void
624 start_update(arg)
625 	void *arg;
626 {
627 	sched_sync(curproc);
628 	/* NOTREACHED */
629 }
630 
631 void
632 start_cleaner(arg)
633 	void *arg;
634 {
635 	buf_daemon(curproc);
636 	/* NOTREACHED */
637 }
638 
639 void
640 start_reaper(arg)
641 	void *arg;
642 {
643 	reaper();
644 	/* NOTREACHED */
645 }
646 
647 #ifdef CRYPTO
648 void
649 start_crypto(arg)
650 	void *arg;
651 {
652 	crypto_thread();
653 	/* NOTREACHED */
654 }
655 #endif /* CRYPTO */
656