xref: /openbsd-src/sys/kern/init_main.c (revision 06ccd5dad87daa7ede3ae33928857a729a8043ea)
1 /*	$OpenBSD: init_main.c,v 1.119 2004/07/28 17:15:12 tholo 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 #if !defined(__HAVE_CPUINFO) && !defined(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 #ifndef __HAVE_CPUINFO
126 struct	timeval runtime;
127 #endif
128 int	ncpus =  1;
129 
130 #if !defined(NO_PROPOLICE)
131 long	__guard[8];
132 #endif
133 
134 /* XXX return int so gcc -Werror won't complain */
135 int	main(void *);
136 void	check_console(struct proc *);
137 void	start_init(void *);
138 void	start_cleaner(void *);
139 void	start_update(void *);
140 void	start_reaper(void *);
141 void	start_crypto(void *);
142 void	init_exec(void);
143 void	kqueue_init(void);
144 
145 extern char sigcode[], esigcode[];
146 #ifdef SYSCALL_DEBUG
147 extern char *syscallnames[];
148 #endif
149 
150 struct emul emul_native = {
151 	"native",
152 	NULL,
153 	sendsig,
154 	SYS_syscall,
155 	SYS_MAXSYSCALL,
156 	sysent,
157 #ifdef SYSCALL_DEBUG
158 	syscallnames,
159 #else
160 	NULL,
161 #endif
162 	0,
163 	copyargs,
164 	setregs,
165 	NULL,
166 	sigcode,
167 	esigcode,
168 	EMUL_ENABLED | EMUL_NATIVE,
169 };
170 
171 
172 /*
173  * System startup; initialize the world, create process 0, mount root
174  * filesystem, and fork to create init and pagedaemon.  Most of the
175  * hard work is done in the lower-level initialization routines including
176  * startup(), which does memory initialization and autoconfiguration.
177  */
178 /* XXX return int, so gcc -Werror won't complain */
179 int
180 main(framep)
181 	void *framep;				/* XXX should go away */
182 {
183 	struct proc *p;
184 	struct pdevinit *pdev;
185 	struct timeval rtv;
186 	quad_t lim;
187 	int s, i;
188 	register_t rval[2];
189 	extern struct pdevinit pdevinit[];
190 	extern void scheduler_start(void);
191 	extern void disk_init(void);
192 	extern void endtsleep(void *);
193 	extern void realitexpire(void *);
194 
195 	/*
196 	 * Initialize the current process pointer (curproc) before
197 	 * any possible traps/probes to simplify trap processing.
198 	 */
199 	curproc = p = &proc0;
200 #ifdef __HAVE_CPUINFO
201 	p->p_cpu = curcpu();
202 #endif
203 
204 	/*
205 	 * Initialize timeouts.
206 	 */
207 	timeout_startup();
208 
209 	/*
210 	 * Attempt to find console and initialize
211 	 * in case of early panic or other messages.
212 	 */
213 	config_init();		/* init autoconfiguration data structures */
214 	consinit();
215 
216 	printf("%s\n", copyright);
217 
218 	KERNEL_LOCK_INIT();
219 
220 	uvm_init();
221 	disk_init();		/* must come before autoconfiguration */
222 	tty_init();		/* initialise tty's */
223 	cpu_startup();
224 
225 	/*
226 	 * Initialize mbuf's.  Do this now because we might attempt to
227 	 * allocate mbufs or mbuf clusters during autoconfiguration.
228 	 */
229 	mbinit();
230 
231 	/* Initalize sockets. */
232 	soinit();
233 
234 	/* Initialize sysctls (must be done before any processes run) */
235 	sysctl_init();
236 
237 	/*
238 	 * Initialize process and pgrp structures.
239 	 */
240 	procinit();
241 
242 	/* Initialize file locking. */
243 	lf_init();
244 
245 	/*
246 	 * Initialize filedescriptors.
247 	 */
248 	filedesc_init();
249 
250 	/*
251 	 * Initialize pipes.
252 	 */
253 	pipe_init();
254 
255 	/*
256 	 * Initialize kqueues.
257 	 */
258 	kqueue_init();
259 
260 	/*
261 	 * Create process 0 (the swapper).
262 	 */
263 	LIST_INSERT_HEAD(&allproc, p, p_list);
264 	p->p_pgrp = &pgrp0;
265 	LIST_INSERT_HEAD(PIDHASH(0), p, p_hash);
266 	LIST_INSERT_HEAD(PGRPHASH(0), &pgrp0, pg_hash);
267 	LIST_INIT(&pgrp0.pg_members);
268 	LIST_INSERT_HEAD(&pgrp0.pg_members, p, p_pglist);
269 
270 	pgrp0.pg_session = &session0;
271 	session0.s_count = 1;
272 	session0.s_leader = p;
273 
274 	p->p_flag = P_INMEM | P_SYSTEM | P_NOCLDWAIT;
275 	p->p_stat = SONPROC;
276 	p->p_nice = NZERO;
277 	p->p_emul = &emul_native;
278 	bcopy("swapper", p->p_comm, sizeof ("swapper"));
279 
280 	/* Init timeouts. */
281 	timeout_set(&p->p_sleep_to, endtsleep, p);
282 	timeout_set(&p->p_realit_to, realitexpire, p);
283 
284 	/* Create credentials. */
285 	cred0.p_refcnt = 1;
286 	p->p_cred = &cred0;
287 	p->p_ucred = crget();
288 	p->p_ucred->cr_ngroups = 1;	/* group 0 */
289 
290 	/* Initialize signal state for process 0. */
291 	signal_init();
292 	p->p_sigacts = &sigacts0;
293 	siginit(p);
294 
295 	/* Create the file descriptor table. */
296 	p->p_fd = fdinit(NULL);
297 
298 	/* Create the limits structures. */
299 	p->p_limit = &limit0;
300 	for (i = 0; i < sizeof(p->p_rlimit)/sizeof(p->p_rlimit[0]); i++)
301 		limit0.pl_rlimit[i].rlim_cur =
302 		    limit0.pl_rlimit[i].rlim_max = RLIM_INFINITY;
303 	limit0.pl_rlimit[RLIMIT_NOFILE].rlim_cur = NOFILE;
304 	limit0.pl_rlimit[RLIMIT_NOFILE].rlim_max = MIN(NOFILE_MAX,
305 	    (maxfiles - NOFILE > NOFILE) ?  maxfiles - NOFILE : NOFILE);
306 	limit0.pl_rlimit[RLIMIT_NPROC].rlim_cur = MAXUPRC;
307 	lim = ptoa(uvmexp.free);
308 	limit0.pl_rlimit[RLIMIT_RSS].rlim_max = lim;
309 	limit0.pl_rlimit[RLIMIT_MEMLOCK].rlim_max = lim;
310 	limit0.pl_rlimit[RLIMIT_MEMLOCK].rlim_cur = lim / 3;
311 	limit0.p_refcnt = 1;
312 
313 	/* Allocate a prototype map so we have something to fork. */
314 	uvmspace_init(&vmspace0, pmap_kernel(), round_page(VM_MIN_ADDRESS),
315 	    trunc_page(VM_MAX_ADDRESS), TRUE);
316 	p->p_vmspace = &vmspace0;
317 
318 	p->p_addr = proc0paddr;				/* XXX */
319 
320 	/*
321 	 * We continue to place resource usage info in the
322 	 * user struct so they're pageable.
323 	 */
324 	p->p_stats = &p->p_addr->u_stats;
325 
326 	/*
327 	 * Charge root for one process.
328 	 */
329 	(void)chgproccnt(0, 1);
330 
331 	/* Initialize run queues */
332 	rqinit();
333 
334 	/* Configure the devices */
335 	cpu_configure();
336 
337 	/* Configure virtual memory system, set vm rlimits. */
338 	uvm_init_limits(p);
339 
340 	/* Initialize the file systems. */
341 #if defined(NFSSERVER) || defined(NFSCLIENT)
342 	nfs_init();			/* initialize server/shared data */
343 #endif
344 	vfsinit();
345 
346 	/* Start real time and statistics clocks. */
347 	initclocks();
348 
349 	/* Lock the kernel on behalf of proc0. */
350 	KERNEL_PROC_LOCK(p);
351 
352 #ifdef SYSVSHM
353 	/* Initialize System V style shared memory. */
354 	shminit();
355 #endif
356 
357 #ifdef SYSVSEM
358 	/* Initialize System V style semaphores. */
359 	seminit();
360 #endif
361 
362 #ifdef SYSVMSG
363 	/* Initialize System V style message queues. */
364 	msginit();
365 #endif
366 
367 	/* Attach pseudo-devices. */
368 	randomattach();
369 	for (pdev = pdevinit; pdev->pdev_attach != NULL; pdev++)
370 		if (pdev->pdev_count > 0)
371 			(*pdev->pdev_attach)(pdev->pdev_count);
372 
373 #ifdef CRYPTO
374 	swcr_init();
375 #endif /* CRYPTO */
376 
377 	/*
378 	 * Initialize protocols.  Block reception of incoming packets
379 	 * until everything is ready.
380 	 */
381 	s = splimp();
382 	ifinit();
383 	domaininit();
384 	if_attachdomain();
385 	splx(s);
386 
387 #ifdef GPROF
388 	/* Initialize kernel profiling. */
389 	kmstartup();
390 #endif
391 
392 #if !defined(NO_PROPOLICE)
393 	arc4random_bytes(__guard, sizeof(__guard));
394 #endif
395 
396 	/* init exec and emul */
397 	init_exec();
398 
399 	/* Start the scheduler */
400 	scheduler_start();
401 
402 	dostartuphooks();
403 
404 	/* Configure root/swap devices */
405 	if (md_diskconf)
406 		(*md_diskconf)();
407 
408 	/* Mount the root file system. */
409 	if (vfs_mountroot())
410 		panic("cannot mount root");
411 	CIRCLEQ_FIRST(&mountlist)->mnt_flag |= MNT_ROOTFS;
412 
413 	/* Get the vnode for '/'.  Set p->p_fd->fd_cdir to reference it. */
414 	if (VFS_ROOT(mountlist.cqh_first, &rootvnode))
415 		panic("cannot find root vnode");
416 	p->p_fd->fd_cdir = rootvnode;
417 	VREF(p->p_fd->fd_cdir);
418 	VOP_UNLOCK(rootvnode, 0, p);
419 	p->p_fd->fd_rdir = NULL;
420 
421 	/*
422 	 * Now can look at time, having had a chance to verify the time
423 	 * from the file system.  Reset p->p_rtime as it may have been
424 	 * munched in mi_switch() after the time got set.
425 	 */
426 #ifdef __HAVE_TIMECOUNTER
427 	microtime(&boottime);
428 #else
429 	boottime = mono_time = time;
430 #endif
431 	p->p_stats->p_start = boottime;
432 #ifdef __HAVE_CPUINFO
433 	microuptime(&p->p_cpu->ci_schedstate.spc_runtime);
434 #else
435 	microuptime(&runtime);
436 #endif
437 	p->p_rtime.tv_sec = p->p_rtime.tv_usec = 0;
438 
439 	uvm_swap_init();
440 
441 	/* Create process 1 (init(8)). */
442 	if (fork1(p, SIGCHLD, FORK_FORK, NULL, 0, start_init, NULL, rval))
443 		panic("fork init");
444 
445 	/* Create process 2, the pageout daemon kernel thread. */
446 	if (kthread_create(uvm_pageout, NULL, NULL, "pagedaemon"))
447 		panic("fork pagedaemon");
448 
449 	/* Create process 3, the reaper daemon kernel thread. */
450 	if (kthread_create(start_reaper, NULL, NULL, "reaper"))
451 		panic("fork reaper");
452 
453 	/* Create process 4, the cleaner daemon kernel thread. */
454 	if (kthread_create(start_cleaner, NULL, NULL, "cleaner"))
455 		panic("fork cleaner");
456 
457 	/* Create process 5, the update daemon kernel thread. */
458 	if (kthread_create(start_update, NULL, NULL, "update"))
459 		panic("fork update");
460 
461 	/* Create process 6, the aiodone daemon kernel thread. */
462 	if (kthread_create(uvm_aiodone_daemon, NULL, NULL, "aiodoned"))
463 		panic("fork aiodoned");
464 
465 #ifdef CRYPTO
466 	/* Create process 7, the crypto kernel thread. */
467 	if (kthread_create(start_crypto, NULL, NULL, "crypto"))
468 		panic("crypto thread");
469 #endif /* CRYPTO */
470 
471 	/* Create any other deferred kernel threads. */
472 	kthread_run_deferred_queue();
473 
474 	microtime(&rtv);
475 	srandom((u_long)(rtv.tv_sec ^ rtv.tv_usec));
476 
477 	randompid = 1;
478 
479 #if defined(MULTIPROCESSOR)
480 	/* Boot the secondary processors. */
481 	cpu_boot_secondary_processors();
482 #endif
483 
484 	/* The scheduler is an infinite loop. */
485 	uvm_scheduler();
486 	/* NOTREACHED */
487 }
488 
489 /*
490  * List of paths to try when searching for "init".
491  */
492 static char *initpaths[] = {
493 	"/sbin/init",
494 	"/sbin/oinit",
495 	"/sbin/init.bak",
496 	NULL,
497 };
498 
499 void
500 check_console(p)
501 	struct proc *p;
502 {
503 	struct nameidata nd;
504 	int error;
505 
506 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_SYSSPACE, "/dev/console", p);
507 	error = namei(&nd);
508 	if (error) {
509 		if (error == ENOENT)
510 			printf("warning: /dev/console does not exist\n");
511 		else
512 			printf("warning: /dev/console error %d\n", error);
513 	} else
514 		vrele(nd.ni_vp);
515 }
516 
517 /*
518  * Start the initial user process; try exec'ing each pathname in "initpaths".
519  * The program is invoked with one argument containing the boot flags.
520  */
521 void
522 start_init(arg)
523 	void *arg;
524 {
525 	struct proc *p = arg;
526 	vaddr_t addr;
527 	struct sys_execve_args /* {
528 		syscallarg(const char *) path;
529 		syscallarg(char *const *) argp;
530 		syscallarg(char *const *) envp;
531 	} */ args;
532 	int options, error;
533 	long i;
534 	register_t retval[2];
535 	char flags[4], *flagsp;
536 	char **pathp, *path, *ucp, **uap, *arg0, *arg1 = NULL;
537 
538 	initproc = p;
539 
540 	/*
541 	 * Now in process 1.
542 	 */
543 	check_console(p);
544 
545 	/*
546 	 * Need just enough stack to hold the faked-up "execve()" arguments.
547 	 */
548 #ifdef MACHINE_STACK_GROWS_UP
549 	addr = USRSTACK;
550 #else
551 	addr = USRSTACK - PAGE_SIZE;
552 #endif
553 	if (uvm_map(&p->p_vmspace->vm_map, &addr, PAGE_SIZE,
554 	    NULL, UVM_UNKNOWN_OFFSET, 0,
555 	    UVM_MAPFLAG(UVM_PROT_RW, UVM_PROT_ALL, UVM_INH_COPY,
556 	    UVM_ADV_NORMAL, UVM_FLAG_FIXED|UVM_FLAG_OVERLAY|UVM_FLAG_COPYONW)))
557 		panic("init: couldn't allocate argument space");
558 	p->p_vmspace->vm_maxsaddr = (caddr_t)addr;
559 
560 	for (pathp = &initpaths[0]; (path = *pathp) != NULL; pathp++) {
561 #ifdef MACHINE_STACK_GROWS_UP
562 		ucp = (char *)addr;
563 #else
564 		ucp = (char *)(addr + PAGE_SIZE);
565 #endif
566 		/*
567 		 * Construct the boot flag argument.
568 		 */
569 		flagsp = flags;
570 		*flagsp++ = '-';
571 		options = 0;
572 
573 		if (boothowto & RB_SINGLE) {
574 			*flagsp++ = 's';
575 			options = 1;
576 		}
577 #ifdef notyet
578 		if (boothowto & RB_FASTBOOT) {
579 			*flagsp++ = 'f';
580 			options = 1;
581 		}
582 #endif
583 
584 		/*
585 		 * Move out the flags (arg 1), if necessary.
586 		 */
587 		if (options != 0) {
588 			*flagsp++ = '\0';
589 			i = flagsp - flags;
590 #ifdef DEBUG
591 			printf("init: copying out flags `%s' %d\n", flags, i);
592 #endif
593 #ifdef MACHINE_STACK_GROWS_UP
594 			arg1 = ucp;
595 			(void)copyout((caddr_t)flags, (caddr_t)ucp, i);
596 			ucp += i;
597 #else
598 			(void)copyout((caddr_t)flags, (caddr_t)(ucp -= i), i);
599 			arg1 = ucp;
600 #endif
601 		}
602 
603 		/*
604 		 * Move out the file name (also arg 0).
605 		 */
606 		i = strlen(path) + 1;
607 #ifdef DEBUG
608 		printf("init: copying out path `%s' %d\n", path, i);
609 #endif
610 #ifdef MACHINE_STACK_GROWS_UP
611 		arg0 = ucp;
612 		(void)copyout((caddr_t)path, (caddr_t)ucp, i);
613 		ucp += i;
614 		ucp = (caddr_t)ALIGN((u_long)ucp);
615 		uap = (char **)ucp + 3;
616 #else
617 		(void)copyout((caddr_t)path, (caddr_t)(ucp -= i), i);
618 		arg0 = ucp;
619 		uap = (char **)((u_long)ucp & ~ALIGNBYTES);
620 #endif
621 
622 		/*
623 		 * Move out the arg pointers.
624 		 */
625 		i = 0;
626 		copyout(&i, (caddr_t)--uap, sizeof(register_t)); /* terminator */
627 		if (options != 0)
628 			copyout(&arg1, (caddr_t)--uap, sizeof(register_t));
629 		copyout(&arg0, (caddr_t)--uap, sizeof(register_t));
630 
631 		/*
632 		 * Point at the arguments.
633 		 */
634 		SCARG(&args, path) = arg0;
635 		SCARG(&args, argp) = uap;
636 		SCARG(&args, envp) = NULL;
637 
638 		/*
639 		 * Now try to exec the program.  If can't for any reason
640 		 * other than it doesn't exist, complain.
641 		 */
642 		if ((error = sys_execve(p, &args, retval)) == 0) {
643 			KERNEL_PROC_UNLOCK(p);
644 			return;
645 		}
646 		if (error != ENOENT)
647 			printf("exec %s: error %d\n", path, error);
648 	}
649 	printf("init: not found\n");
650 	panic("no init");
651 }
652 
653 void
654 start_update(arg)
655 	void *arg;
656 {
657 	sched_sync(curproc);
658 	/* NOTREACHED */
659 }
660 
661 void
662 start_cleaner(arg)
663 	void *arg;
664 {
665 	buf_daemon(curproc);
666 	/* NOTREACHED */
667 }
668 
669 void
670 start_reaper(arg)
671 	void *arg;
672 {
673 	reaper();
674 	/* NOTREACHED */
675 }
676 
677 #ifdef CRYPTO
678 void
679 start_crypto(arg)
680 	void *arg;
681 {
682 	crypto_thread();
683 	/* NOTREACHED */
684 }
685 #endif /* CRYPTO */
686