xref: /netbsd-src/sys/kern/init_main.c (revision 89c5a767f8fc7a4633b2d409966e2becbb98ff92)
1 /*	$NetBSD: init_main.c,v 1.164 2000/03/10 01:13:18 enami 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_sysv.h"
47 #include "opt_maxuprc.h"
48 #include "opt_multiprocessor.h"
49 
50 #include "rnd.h"
51 
52 #include <sys/param.h>
53 #include <sys/acct.h>
54 #include <sys/filedesc.h>
55 #include <sys/file.h>
56 #include <sys/errno.h>
57 #include <sys/exec.h>
58 #include <sys/callout.h>
59 #include <sys/kernel.h>
60 #include <sys/mount.h>
61 #include <sys/map.h>
62 #include <sys/proc.h>
63 #include <sys/kthread.h>
64 #include <sys/resourcevar.h>
65 #include <sys/signalvar.h>
66 #include <sys/systm.h>
67 #include <sys/vnode.h>
68 #include <sys/tty.h>
69 #include <sys/conf.h>
70 #include <sys/disklabel.h>
71 #include <sys/buf.h>
72 #include <sys/device.h>
73 #include <sys/socketvar.h>
74 #include <sys/protosw.h>
75 #include <sys/reboot.h>
76 #include <sys/user.h>
77 #ifdef SYSVSHM
78 #include <sys/shm.h>
79 #endif
80 #ifdef SYSVSEM
81 #include <sys/sem.h>
82 #endif
83 #ifdef SYSVMSG
84 #include <sys/msg.h>
85 #endif
86 #include <sys/domain.h>
87 #include <sys/mbuf.h>
88 #include <sys/namei.h>
89 #if NRND > 0
90 #include <sys/rnd.h>
91 #endif
92 
93 #include <sys/syscall.h>
94 #include <sys/syscallargs.h>
95 
96 #include <ufs/ufs/quota.h>
97 
98 #include <miscfs/genfs/genfs.h>
99 #include <miscfs/syncfs/syncfs.h>
100 
101 #include <machine/cpu.h>
102 
103 #include <vm/vm.h>
104 #include <vm/vm_pageout.h>
105 
106 #include <uvm/uvm.h>
107 
108 #include <net/if.h>
109 #include <net/raw_cb.h>
110 
111 char	copyright[] = "\
112 Copyright (c) 1996, 1997, 1998, 1999, 2000
113     The NetBSD Foundation, Inc.  All rights reserved.
114 Copyright (c) 1982, 1986, 1989, 1991, 1993
115     The Regents of the University of California.  All rights reserved.
116 
117 ";
118 
119 /* Components of the first process -- never freed. */
120 struct	session session0;
121 struct	pgrp pgrp0;
122 struct	proc proc0;
123 struct	pcred cred0;
124 struct	filedesc0 filedesc0;
125 struct	cwdinfo cwdi0;
126 struct	plimit limit0;
127 struct	vmspace vmspace0;
128 struct	sigacts sigacts0;
129 #ifndef curproc
130 struct	proc *curproc = &proc0;
131 #endif
132 struct	proc *initproc;
133 
134 int	cmask = CMASK;
135 extern	struct user *proc0paddr;
136 
137 struct	vnode *rootvp, *swapdev_vp;
138 int	boothowto;
139 int	cold = 1;			/* still working on startup */
140 struct	timeval boottime;
141 struct	timeval runtime;
142 
143 __volatile int start_init_exec;		/* semaphore for start_init() */
144 
145 static void check_console __P((struct proc *p));
146 static void start_init __P((void *));
147 static void start_pagedaemon __P((void *));
148 static void start_reaper __P((void *));
149 void main __P((void));
150 
151 extern char sigcode[], esigcode[];
152 #ifdef SYSCALL_DEBUG
153 extern char *syscallnames[];
154 #endif
155 
156 struct emul emul_netbsd = {
157 	"netbsd",
158 	NULL,
159 	sendsig,
160 	SYS_syscall,
161 	SYS_MAXSYSCALL,
162 	sysent,
163 #ifdef SYSCALL_DEBUG
164 	syscallnames,
165 #else
166 	NULL,
167 #endif
168 	0,
169 	copyargs,
170 	setregs,
171 	sigcode,
172 	esigcode,
173 };
174 
175 /*
176  * System startup; initialize the world, create process 0, mount root
177  * filesystem, and fork to create init and pagedaemon.  Most of the
178  * hard work is done in the lower-level initialization routines including
179  * startup(), which does memory initialization and autoconfiguration.
180  */
181 void
182 main()
183 {
184 	struct proc *p;
185 	struct pdevinit *pdev;
186 	int i, s, error;
187 	extern struct pdevinit pdevinit[];
188 	extern void roundrobin __P((void *));
189 	extern void schedcpu __P((void *));
190 	extern void disk_init __P((void));
191 #if defined(NFSSERVER) || defined(NFS)
192 	extern void nfs_init __P((void));
193 #endif
194 
195 	/*
196 	 * Initialize the current process pointer (curproc) before
197 	 * any possible traps/probes to simplify trap processing.
198 	 */
199 	p = &proc0;
200 	curproc = p;
201 	/*
202 	 * Attempt to find console and initialize
203 	 * in case of early panic or other messages.
204 	 */
205 	consinit();
206 	printf("%s", copyright);
207 
208 	uvm_init();
209 
210 	/* Do machine-dependent initialization. */
211 	cpu_startup();
212 
213 	/* Initialize callouts. */
214 	callout_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 	/* Initialize sockets. */
223 	soinit();
224 
225 	/*
226 	 * The following 3 things must be done before autoconfiguration.
227 	 */
228 	disk_init();		/* initialize disk list */
229 	tty_init();		/* initialize tty list */
230 #if NRND > 0
231 	rnd_init();		/* initialize RNG */
232 #endif
233 
234 	/*
235 	 * Initialize process and pgrp structures.
236 	 */
237 	procinit();
238 
239 	/*
240 	 * Create process 0 (the swapper).
241 	 */
242 	s = proclist_lock_write();
243 	LIST_INSERT_HEAD(&allproc, p, p_list);
244 	proclist_unlock_write(s);
245 
246 	p->p_pgrp = &pgrp0;
247 	LIST_INSERT_HEAD(PGRPHASH(0), &pgrp0, pg_hash);
248 	LIST_INIT(&pgrp0.pg_members);
249 	LIST_INSERT_HEAD(&pgrp0.pg_members, p, p_pglist);
250 
251 	pgrp0.pg_session = &session0;
252 	session0.s_count = 1;
253 	session0.s_sid = p->p_pid;
254 	session0.s_leader = p;
255 
256 	/*
257 	 * Set P_NOCLDWAIT so that kernel threads are reparented to
258 	 * init(8) when they exit.  init(8) can easily wait them out
259 	 * for us.
260 	 */
261 	p->p_flag = P_INMEM | P_SYSTEM | P_NOCLDWAIT;
262 	p->p_stat = SRUN;
263 	p->p_nice = NZERO;
264 	p->p_emul = &emul_netbsd;
265 	strncpy(p->p_comm, "swapper", MAXCOMLEN);
266 
267 	/* Create credentials. */
268 	cred0.p_refcnt = 1;
269 	p->p_cred = &cred0;
270 	p->p_ucred = crget();
271 	p->p_ucred->cr_ngroups = 1;	/* group 0 */
272 
273 	/* Create the file descriptor table. */
274 	finit();
275 	p->p_fd = &filedesc0.fd_fd;
276 	fdinit1(&filedesc0);
277 
278 	/* Create the CWD info. */
279 	p->p_cwdi = &cwdi0;
280 	cwdi0.cwdi_cmask = cmask;
281 	cwdi0.cwdi_refcnt = 1;
282 
283 	/* Create the limits structures. */
284 	p->p_limit = &limit0;
285 	for (i = 0; i < sizeof(p->p_rlimit)/sizeof(p->p_rlimit[0]); i++)
286 		limit0.pl_rlimit[i].rlim_cur =
287 		    limit0.pl_rlimit[i].rlim_max = RLIM_INFINITY;
288 
289 	limit0.pl_rlimit[RLIMIT_NOFILE].rlim_max = maxfiles;
290 	limit0.pl_rlimit[RLIMIT_NOFILE].rlim_cur =
291 	    maxfiles < NOFILE ? maxfiles : NOFILE;
292 
293 	limit0.pl_rlimit[RLIMIT_NPROC].rlim_max = maxproc;
294 	limit0.pl_rlimit[RLIMIT_NPROC].rlim_cur =
295 	    maxproc < MAXUPRC ? maxproc : MAXUPRC;
296 
297 	i = ptoa(uvmexp.free);
298 	limit0.pl_rlimit[RLIMIT_RSS].rlim_max = i;
299 	limit0.pl_rlimit[RLIMIT_MEMLOCK].rlim_max = i;
300 	limit0.pl_rlimit[RLIMIT_MEMLOCK].rlim_cur = i / 3;
301 	limit0.pl_corename = defcorename;
302 	limit0.p_refcnt = 1;
303 
304 	/*
305 	 * Initialize proc0's vmspace, which uses the kernel pmap.
306 	 * All kernel processes (which never have user space mappings)
307 	 * share proc0's vmspace, and thus, the kernel pmap.
308 	 */
309 	uvmspace_init(&vmspace0, pmap_kernel(), round_page(VM_MIN_ADDRESS),
310 	    trunc_page(VM_MAX_ADDRESS), TRUE);
311 	p->p_vmspace = &vmspace0;
312 
313 	p->p_addr = proc0paddr;				/* XXX */
314 
315 	/*
316 	 * We continue to place resource usage info in the
317 	 * user struct so they're pageable.
318 	 */
319 	p->p_stats = &p->p_addr->u_stats;
320 
321 	/*
322 	 * Charge root for one process.
323 	 */
324 	(void)chgproccnt(0, 1);
325 
326 	rqinit();
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(NFS)
333 	nfs_init();			/* initialize server/shared data */
334 #endif
335 	vfsinit();
336 
337 	/* Configure the system hardware.  This will enable interrupts. */
338 	configure();
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 	for (pdev = pdevinit; pdev->pdev_attach != NULL; pdev++)
357 		(*pdev->pdev_attach)(pdev->pdev_count);
358 
359 	/*
360 	 * Initialize protocols.  Block reception of incoming packets
361 	 * until everything is ready.
362 	 */
363 	s = splimp();
364 	ifinit();
365 	domaininit();
366 	splx(s);
367 
368 #ifdef GPROF
369 	/* Initialize kernel profiling. */
370 	kmstartup();
371 #endif
372 
373 	/* Initialize system accouting. */
374 	acct_init();
375 
376 	/*
377 	 * Initialize signal-related data structures, and signal state
378 	 * for proc0.
379 	 */
380 	signal_init();
381 	p->p_sigacts = &sigacts0;
382 	siginit(p);
383 
384 	/* Kick off timeout driven events by calling first time. */
385 	roundrobin(NULL);
386 	schedcpu(NULL);
387 
388 	/*
389 	 * Create process 1 (init(8)).  We do this now, as Unix has
390 	 * historically had init be process 1, and changing this would
391 	 * probably upset a lot of people.
392 	 *
393 	 * Note that process 1 won't immediately exec init(8), but will
394 	 * wait for us to inform it that the root file system has been
395 	 * mounted.
396 	 */
397 	if (fork1(p, 0, SIGCHLD, NULL, 0, NULL, &initproc))
398 		panic("fork init");
399 	cpu_set_kpc(initproc, start_init, initproc);
400 
401 	/*
402 	 * Create any kernel threads who's creation was deferred because
403 	 * initproc had not yet been created.
404 	 */
405 	kthread_run_deferred_queue();
406 
407 	/*
408 	 * Now that device driver threads have been created, wait for
409 	 * them to finish any deferred autoconfiguration.  Note we don't
410 	 * need to lock this semaphore, since we haven't booted any
411 	 * secondary processors, yet.
412 	 */
413 	while (config_pending)
414 		(void) tsleep((void *)&config_pending, PWAIT, "cfpend", 0);
415 
416 	/*
417 	 * Now that autoconfiguration has completed, we can determine
418 	 * the root and dump devices.
419 	 */
420 	cpu_rootconf();
421 	cpu_dumpconf();
422 
423 	/* Mount the root file system. */
424 	do {
425 		domountroothook();
426 		if ((error = vfs_mountroot())) {
427 			printf("cannot mount root, error = %d\n", error);
428 			boothowto |= RB_ASKNAME;
429 			setroot(root_device,
430 			    (rootdev != NODEV) ? DISKPART(rootdev) : 0);
431 		}
432 	} while (error != 0);
433 	mountroothook_destroy();
434 
435 	mountlist.cqh_first->mnt_flag |= MNT_ROOTFS;
436 	mountlist.cqh_first->mnt_op->vfs_refcount++;
437 
438 	/*
439 	 * Get the vnode for '/'.  Set filedesc0.fd_fd.fd_cdir to
440 	 * reference it.
441 	 */
442 	if (VFS_ROOT(mountlist.cqh_first, &rootvnode))
443 		panic("cannot find root vnode");
444 	cwdi0.cwdi_cdir = rootvnode;
445 	VREF(cwdi0.cwdi_cdir);
446 	VOP_UNLOCK(rootvnode, 0);
447 	cwdi0.cwdi_rdir = NULL;
448 
449 	/*
450 	 * Now that root is mounted, we can fixup initproc's CWD
451 	 * info.  All other processes are kthreads, which merely
452 	 * share proc0's CWD info.
453 	 */
454 	initproc->p_cwdi->cwdi_cdir = rootvnode;
455 	VREF(initproc->p_cwdi->cwdi_cdir);
456 	initproc->p_cwdi->cwdi_rdir = NULL;
457 
458 	/*
459 	 * Now can look at time, having had a chance to verify the time
460 	 * from the file system.  Reset p->p_rtime as it may have been
461 	 * munched in mi_switch() after the time got set.
462 	 */
463 	proclist_lock_read();
464 	s = splclock();		/* so we can read time */
465 	for (p = LIST_FIRST(&allproc); p != NULL;
466 	     p = LIST_NEXT(p, p_list)) {
467 		p->p_stats->p_start = runtime = mono_time = boottime = time;
468 		p->p_rtime.tv_sec = p->p_rtime.tv_usec = 0;
469 	}
470 	splx(s);
471 	proclist_unlock_read();
472 
473 	/* Create the pageout daemon kernel thread. */
474 	uvm_swap_init();
475 	if (kthread_create1(start_pagedaemon, NULL, NULL, "pagedaemon"))
476 		panic("fork pagedaemon");
477 
478 	/* Create the process reaper kernel thread. */
479 	if (kthread_create1(start_reaper, NULL, NULL, "reaper"))
480 		panic("fork reaper");
481 
482 	/* Create the filesystem syncer kernel thread. */
483 	if (kthread_create1(sched_sync, NULL, NULL, "ioflush"))
484 		panic("fork syncer");
485 
486 #if defined(MULTIPROCESSOR)
487 	/* Boot the secondary processors. */
488 	cpu_boot_secondary_processors();
489 #endif
490 
491 	/*
492 	 * Okay, now we can let init(8) exec!  It's off to userland!
493 	 */
494 	start_init_exec = 1;
495 	wakeup((void *)&start_init_exec);
496 
497 	/* The scheduler is an infinite loop. */
498 	uvm_scheduler();
499 	/* NOTREACHED */
500 }
501 
502 static void
503 check_console(p)
504 	struct proc *p;
505 {
506 	struct nameidata nd;
507 	int error;
508 
509 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_SYSSPACE, "/dev/console", p);
510 	error = namei(&nd);
511 	if (error == 0)
512 		vrele(nd.ni_vp);
513 	else if (error == ENOENT)
514 		printf("warning: no /dev/console\n");
515 	else
516 		printf("warning: lookup /dev/console: error %d\n", error);
517 }
518 
519 /*
520  * List of paths to try when searching for "init".
521  */
522 static char *initpaths[] = {
523 	"/sbin/init",
524 	"/sbin/oinit",
525 	"/sbin/init.bak",
526 	NULL,
527 };
528 
529 /*
530  * Start the initial user process; try exec'ing each pathname in "initpaths".
531  * The program is invoked with one argument containing the boot flags.
532  */
533 static void
534 start_init(arg)
535 	void *arg;
536 {
537 	struct proc *p = arg;
538 	vaddr_t addr;
539 	struct sys_execve_args /* {
540 		syscallarg(const char *) path;
541 		syscallarg(char * const *) argp;
542 		syscallarg(char * const *) envp;
543 	} */ args;
544 	int options, i, error;
545 	register_t retval[2];
546 	char flags[4], *flagsp;
547 	char **pathp, *path, *slash, *ucp, **uap, *arg0, *arg1 = NULL;
548 
549 	/*
550 	 * Now in process 1.
551 	 */
552 	strncpy(p->p_comm, "init", MAXCOMLEN);
553 
554 	/*
555 	 * Wait for main() to tell us that it's safe to exec.
556 	 */
557 	while (start_init_exec == 0)
558 		(void) tsleep((void *)&start_init_exec, PWAIT, "initexec", 0);
559 
560 	/*
561 	 * This is not the right way to do this.  We really should
562 	 * hand-craft a descriptor onto /dev/console to hand to init,
563 	 * but that's a _lot_ more work, and the benefit from this easy
564 	 * hack makes up for the "good is the enemy of the best" effect.
565 	 */
566 	check_console(p);
567 
568 	/*
569 	 * Need just enough stack to hold the faked-up "execve()" arguments.
570 	 */
571 	addr = USRSTACK - PAGE_SIZE;
572 	if (uvm_map(&p->p_vmspace->vm_map, &addr, PAGE_SIZE,
573                     NULL, UVM_UNKNOWN_OFFSET,
574                     UVM_MAPFLAG(UVM_PROT_ALL, UVM_PROT_ALL, UVM_INH_COPY,
575 		    UVM_ADV_NORMAL,
576                     UVM_FLAG_FIXED|UVM_FLAG_OVERLAY|UVM_FLAG_COPYONW))
577 		!= KERN_SUCCESS)
578 		panic("init: couldn't allocate argument space");
579 	p->p_vmspace->vm_maxsaddr = (caddr_t)addr;
580 
581 	for (pathp = &initpaths[0]; (path = *pathp) != NULL; pathp++) {
582 		ucp = (char *)(addr + PAGE_SIZE);
583 
584 		/*
585 		 * Construct the boot flag argument.
586 		 */
587 		flagsp = flags;
588 		*flagsp++ = '-';
589 		options = 0;
590 
591 		if (boothowto & RB_SINGLE) {
592 			*flagsp++ = 's';
593 			options = 1;
594 		}
595 #ifdef notyet
596 		if (boothowto & RB_FASTBOOT) {
597 			*flagsp++ = 'f';
598 			options = 1;
599 		}
600 #endif
601 
602 		/*
603 		 * Move out the flags (arg 1), if necessary.
604 		 */
605 		if (options != 0) {
606 			*flagsp++ = '\0';
607 			i = flagsp - flags;
608 #ifdef DEBUG
609 			printf("init: copying out flags `%s' %d\n", flags, i);
610 #endif
611 			(void)copyout((caddr_t)flags, (caddr_t)(ucp -= i), i);
612 			arg1 = ucp;
613 		}
614 
615 		/*
616 		 * Move out the file name (also arg 0).
617 		 */
618 		i = strlen(path) + 1;
619 #ifdef DEBUG
620 		printf("init: copying out path `%s' %d\n", path, i);
621 #endif
622 		(void)copyout((caddr_t)path, (caddr_t)(ucp -= i), i);
623 		arg0 = ucp;
624 
625 		/*
626 		 * Move out the arg pointers.
627 		 */
628 		uap = (char **)((long)ucp & ~ALIGNBYTES);
629 		(void)suword((caddr_t)--uap, 0);	/* terminator */
630 		if (options != 0)
631 			(void)suword((caddr_t)--uap, (long)arg1);
632 		slash = strrchr(path, '/');
633 		if (slash)
634 			(void)suword((caddr_t)--uap,
635 			    (long)arg0 + (slash + 1 - path));
636 		else
637 			(void)suword((caddr_t)--uap, (long)arg0);
638 
639 		/*
640 		 * Point at the arguments.
641 		 */
642 		SCARG(&args, path) = arg0;
643 		SCARG(&args, argp) = uap;
644 		SCARG(&args, envp) = NULL;
645 
646 		/*
647 		 * Now try to exec the program.  If can't for any reason
648 		 * other than it doesn't exist, complain.
649 		 */
650 		error = sys_execve(p, &args, retval);
651 		if (error == 0 || error == EJUSTRETURN)
652 			return;
653 		if (error != ENOENT)
654 			printf("exec %s: error %d\n", path, error);
655 	}
656 	printf("init: not found\n");
657 	panic("no init");
658 }
659 
660 /* ARGSUSED */
661 static void
662 start_pagedaemon(arg)
663 	void *arg;
664 {
665 
666 	uvm_pageout();
667 	/* NOTREACHED */
668 }
669 
670 /* ARGSUSED */
671 static void
672 start_reaper(arg)
673 	void *arg;
674 {
675 
676 	reaper();
677 	/* NOTREACHED */
678 }
679