xref: /netbsd-src/sys/kern/init_main.c (revision 4d7e773266e3c3f48566c86c0ad52d51c6454fd1)
1 /*	$NetBSD: init_main.c,v 1.108 1997/10/19 02:00:19 mycroft 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	8.9 (Berkeley) 1/21/94
42  */
43 
44 #include "rnd.h"
45 
46 #include <sys/param.h>
47 #include <sys/filedesc.h>
48 #include <sys/errno.h>
49 #include <sys/exec.h>
50 #include <sys/kernel.h>
51 #include <sys/mount.h>
52 #include <sys/map.h>
53 #include <sys/proc.h>
54 #include <sys/resourcevar.h>
55 #include <sys/signalvar.h>
56 #include <sys/systm.h>
57 #include <sys/vnode.h>
58 #include <sys/tty.h>
59 #include <sys/conf.h>
60 #include <sys/disklabel.h>
61 #include <sys/buf.h>
62 #ifdef REAL_CLISTS
63 #include <sys/clist.h>
64 #endif
65 #include <sys/device.h>
66 #include <sys/protosw.h>
67 #include <sys/reboot.h>
68 #include <sys/user.h>
69 #ifdef SYSVSHM
70 #include <sys/shm.h>
71 #endif
72 #ifdef SYSVSEM
73 #include <sys/sem.h>
74 #endif
75 #ifdef SYSVMSG
76 #include <sys/msg.h>
77 #endif
78 #include <sys/domain.h>
79 #include <sys/mbuf.h>
80 #include <sys/namei.h>
81 #if NRND > 0
82 #include <sys/rnd.h>
83 #endif
84 
85 #include <sys/syscall.h>
86 #include <sys/syscallargs.h>
87 
88 #include <ufs/ufs/quota.h>
89 
90 #include <machine/cpu.h>
91 
92 #include <vm/vm.h>
93 #include <vm/vm_pageout.h>
94 
95 #include <net/if.h>
96 #include <net/raw_cb.h>
97 
98 char	copyright[] = "\
99 Copyright (c) 1996, 1997 The NetBSD Foundation, Inc.  All rights reserved.
100 Copyright (c) 1982, 1986, 1989, 1991, 1993
101     The Regents of the University of California.  All rights reserved.
102 
103 ";
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	filedesc0 filedesc0;
111 struct	plimit limit0;
112 struct	vmspace vmspace0;
113 struct	proc *curproc = &proc0;
114 struct	proc *initproc;
115 
116 int	cmask = CMASK;
117 extern	struct user *proc0paddr;
118 
119 struct	vnode *rootvp, *swapdev_vp;
120 int	boothowto;
121 struct	timeval boottime;
122 struct	timeval runtime;
123 
124 static void check_console __P((struct proc *p));
125 static void start_init __P((struct proc *));
126 static void start_pagedaemon __P((struct proc *));
127 void main __P((void *));
128 
129 #ifdef cpu_set_init_frame
130 void *initframep;				/* XXX should go away */
131 #endif
132 
133 extern char sigcode[], esigcode[];
134 #ifdef SYSCALL_DEBUG
135 extern char *syscallnames[];
136 #endif
137 
138 struct emul emul_netbsd = {
139 	"netbsd",
140 	NULL,
141 	sendsig,
142 	SYS_syscall,
143 	SYS_MAXSYSCALL,
144 	sysent,
145 #ifdef SYSCALL_DEBUG
146 	syscallnames,
147 #else
148 	NULL,
149 #endif
150 	0,
151 	copyargs,
152 	setregs,
153 	sigcode,
154 	esigcode,
155 };
156 
157 /*
158  * System startup; initialize the world, create process 0, mount root
159  * filesystem, and fork to create init and pagedaemon.  Most of the
160  * hard work is done in the lower-level initialization routines including
161  * startup(), which does memory initialization and autoconfiguration.
162  */
163 void
164 main(framep)
165 	void *framep;				/* XXX should go away */
166 {
167 	register struct proc *p;
168 	register struct pdevinit *pdev;
169 	register int i;
170 	int s, error;
171 	register_t rval[2];
172 	extern struct pdevinit pdevinit[];
173 	extern void roundrobin __P((void *));
174 	extern void schedcpu __P((void *));
175 	extern void disk_init __P((void));
176 #if defined(NFSSERVER) || defined(NFS)
177 	extern void nfs_init __P((void));
178 #endif
179 
180 	/*
181 	 * Initialize the current process pointer (curproc) before
182 	 * any possible traps/probes to simplify trap processing.
183 	 */
184 	p = &proc0;
185 	curproc = p;
186 	/*
187 	 * Attempt to find console and initialize
188 	 * in case of early panic or other messages.
189 	 */
190 	consinit();
191 	printf(copyright);
192 
193 	vm_mem_init();
194 	kmeminit();
195 	disk_init();		/* must come before autoconfiguration */
196 	tty_init();		/* initialise tty list */
197 #if NRND > 0
198 	rnd_init();
199 #endif
200 	config_init();		/* init autoconfiguration data structures */
201 	cpu_startup();
202 
203 	/*
204 	 * Initialize process and pgrp structures.
205 	 */
206 	procinit();
207 
208 	/*
209 	 * Create process 0 (the swapper).
210 	 */
211 	LIST_INSERT_HEAD(&allproc, p, p_list);
212 	p->p_pgrp = &pgrp0;
213 	LIST_INSERT_HEAD(PGRPHASH(0), &pgrp0, pg_hash);
214 	LIST_INIT(&pgrp0.pg_members);
215 	LIST_INSERT_HEAD(&pgrp0.pg_members, p, p_pglist);
216 
217 	pgrp0.pg_session = &session0;
218 	session0.s_count = 1;
219 	session0.s_leader = p;
220 
221 	p->p_flag = P_INMEM | P_SYSTEM;
222 	p->p_stat = SRUN;
223 	p->p_nice = NZERO;
224 	p->p_emul = &emul_netbsd;
225 	bcopy("swapper", p->p_comm, sizeof ("swapper"));
226 
227 	/* Create credentials. */
228 	cred0.p_refcnt = 1;
229 	p->p_cred = &cred0;
230 	p->p_ucred = crget();
231 	p->p_ucred->cr_ngroups = 1;	/* group 0 */
232 
233 	/* Create the file descriptor table. */
234 	p->p_fd = &filedesc0.fd_fd;
235 	filedesc0.fd_fd.fd_refcnt = 1;
236 	filedesc0.fd_fd.fd_cmask = cmask;
237 	filedesc0.fd_fd.fd_ofiles = filedesc0.fd_dfiles;
238 	filedesc0.fd_fd.fd_ofileflags = filedesc0.fd_dfileflags;
239 	filedesc0.fd_fd.fd_nfiles = NDFILE;
240 
241 	/* Create the limits structures. */
242 	p->p_limit = &limit0;
243 	for (i = 0; i < sizeof(p->p_rlimit)/sizeof(p->p_rlimit[0]); i++)
244 		limit0.pl_rlimit[i].rlim_cur =
245 		    limit0.pl_rlimit[i].rlim_max = RLIM_INFINITY;
246 	limit0.pl_rlimit[RLIMIT_NOFILE].rlim_cur = NOFILE;
247 	limit0.pl_rlimit[RLIMIT_NPROC].rlim_cur = MAXUPRC;
248 	i = ptoa(cnt.v_free_count);
249 	limit0.pl_rlimit[RLIMIT_RSS].rlim_max = i;
250 	limit0.pl_rlimit[RLIMIT_MEMLOCK].rlim_max = i;
251 	limit0.pl_rlimit[RLIMIT_MEMLOCK].rlim_cur = i / 3;
252 	limit0.p_refcnt = 1;
253 
254 	/* Allocate a prototype map so we have something to fork. */
255 	p->p_vmspace = vmspace_alloc(round_page(VM_MIN_ADDRESS),
256 				     trunc_page(VM_MAX_ADDRESS), TRUE);
257 
258 	p->p_addr = proc0paddr;				/* XXX */
259 
260 	/*
261 	 * We continue to place resource usage info and signal
262 	 * actions in the user struct so they're pageable.
263 	 */
264 	p->p_stats = &p->p_addr->u_stats;
265 	p->p_sigacts = &p->p_addr->u_sigacts;
266 
267 	/*
268 	 * Charge root for one process.
269 	 */
270 	(void)chgproccnt(0, 1);
271 
272 	rqinit();
273 
274 	/* Configure virtual memory system, set vm rlimits. */
275 	vm_init_limits(p);
276 
277 	/* Initialize the file systems. */
278 #if defined(NFSSERVER) || defined(NFS)
279 	nfs_init();			/* initialize server/shared data */
280 #endif
281 	vfsinit();
282 
283 	/* Start real time and statistics clocks. */
284 	initclocks();
285 
286 	/* Initialize mbuf's. */
287 	mbinit();
288 
289 #ifdef REAL_CLISTS
290 	/* Initialize clists. */
291 	clist_init();
292 #endif
293 
294 #ifdef SYSVSHM
295 	/* Initialize System V style shared memory. */
296 	shminit();
297 #endif
298 
299 #ifdef SYSVSEM
300 	/* Initialize System V style semaphores. */
301 	seminit();
302 #endif
303 
304 #ifdef SYSVMSG
305 	/* Initialize System V style message queues. */
306 	msginit();
307 #endif
308 
309 	/* Attach pseudo-devices. */
310 	for (pdev = pdevinit; pdev->pdev_attach != NULL; pdev++)
311 		(*pdev->pdev_attach)(pdev->pdev_count);
312 
313 	/*
314 	 * Initialize protocols.  Block reception of incoming packets
315 	 * until everything is ready.
316 	 */
317 	s = splimp();
318 	ifinit();
319 	domaininit();
320 	splx(s);
321 
322 #ifdef GPROF
323 	/* Initialize kernel profiling. */
324 	kmstartup();
325 #endif
326 
327 	/* Kick off timeout driven events by calling first time. */
328 	roundrobin(NULL);
329 	schedcpu(NULL);
330 
331 	/* Determine the root and dump devices. */
332 	cpu_rootconf();
333 	cpu_dumpconf();
334 
335 	/* Mount the root file system. */
336 	do {
337 		domountroothook();
338 		if ((error = vfs_mountroot())) {
339 			printf("cannot mount root, error = %d\n", error);
340 			boothowto |= RB_ASKNAME;
341 			setroot(root_device,
342 			    (rootdev != NODEV) ? DISKPART(rootdev) : 0, NULL);
343 		}
344 	} while (error != 0);
345 	mountroothook_destroy();
346 
347 	mountlist.cqh_first->mnt_flag |= MNT_ROOTFS;
348 	mountlist.cqh_first->mnt_op->vfs_refcount++;
349 
350 	/* Get the vnode for '/'.  Set filedesc0.fd_fd.fd_cdir to reference it. */
351 	if (VFS_ROOT(mountlist.cqh_first, &rootvnode))
352 		panic("cannot find root vnode");
353 	filedesc0.fd_fd.fd_cdir = rootvnode;
354 	VREF(filedesc0.fd_fd.fd_cdir);
355 	VOP_UNLOCK(rootvnode);
356 	filedesc0.fd_fd.fd_rdir = NULL;
357 	swapinit();
358 
359 	/*
360 	 * Now can look at time, having had a chance to verify the time
361 	 * from the file system.  Reset p->p_rtime as it may have been
362 	 * munched in mi_switch() after the time got set.
363 	 */
364 	p->p_stats->p_start = runtime = mono_time = boottime = time;
365 	p->p_rtime.tv_sec = p->p_rtime.tv_usec = 0;
366 
367 	/* Initialize signal state for process 0. */
368 	siginit(p);
369 
370 	/* Create process 1 (init(8)). */
371 	if (sys_fork(p, NULL, rval))
372 		panic("fork init");
373 #ifdef cpu_set_init_frame			/* XXX should go away */
374 	if (rval[1]) {
375 		/*
376 		 * Now in process 1.
377 		 */
378 		initframep = framep;
379 		start_init(curproc);
380 		return;
381 	}
382 #else
383 	cpu_set_kpc(pfind(1), start_init);
384 #endif
385 
386 	/* Create process 2 (the pageout daemon). */
387 	if (sys_fork(p, NULL, rval))
388 		panic("fork pager");
389 #ifdef cpu_set_init_frame			/* XXX should go away */
390 	if (rval[1]) {
391 		/*
392 		 * Now in process 2.
393 		 */
394 		start_pagedaemon(curproc);
395 	}
396 #else
397 	cpu_set_kpc(pfind(2), start_pagedaemon);
398 #endif
399 
400 	/* The scheduler is an infinite loop. */
401 	scheduler();
402 	/* NOTREACHED */
403 }
404 
405 static void
406 check_console(p)
407 	struct proc *p;
408 {
409 	struct nameidata nd;
410 	int error;
411 
412 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_SYSSPACE, "/dev/console", p);
413 	error = namei(&nd);
414 	if (error == 0)
415 		vrele(nd.ni_vp);
416 	else if (error == ENOENT)
417 		printf("warning: no /dev/console\n");
418 	else
419 		printf("warning: lookup /dev/console: error %d\n", error);
420 }
421 
422 /*
423  * List of paths to try when searching for "init".
424  */
425 static char *initpaths[] = {
426 	"/sbin/init",
427 	"/sbin/oinit",
428 	"/sbin/init.bak",
429 	NULL,
430 };
431 
432 /*
433  * Start the initial user process; try exec'ing each pathname in "initpaths".
434  * The program is invoked with one argument containing the boot flags.
435  */
436 static void
437 start_init(p)
438 	struct proc *p;
439 {
440 	vm_offset_t addr;
441 	struct sys_execve_args /* {
442 		syscallarg(const char *) path;
443 		syscallarg(char * const *) argp;
444 		syscallarg(char * const *) envp;
445 	} */ args;
446 	int options, i, error;
447 	register_t retval[2];
448 	char flags[4], *flagsp;
449 	char **pathp, *path, *ucp, **uap, *arg0, *arg1 = NULL;
450 
451 	/*
452 	 * Now in process 1.
453 	 */
454 	initproc = p;
455 
456 #ifdef cpu_set_init_frame			/* XXX should go away */
457 	/*
458 	 * We need to set the system call frame as if we were entered through
459 	 * a syscall() so that when we call sys_execve() below, it will be able
460 	 * to set the entry point (see setregs) when it tries to exec.  The
461 	 * startup code in "locore.s" has allocated space for the frame and
462 	 * passed a pointer to that space as main's argument.
463 	 */
464 	cpu_set_init_frame(p, initframep);
465 #endif
466 
467 	/*
468 	 * This is not the right way to do this.  We really should
469 	 * hand-craft a descriptor onto /dev/console to hand to init,
470 	 * but that's a _lot_ more work, and the benefit from this easy
471 	 * hack makes up for the "good is the enemy of the best" effect.
472 	 */
473 	check_console(p);
474 
475 	/*
476 	 * Need just enough stack to hold the faked-up "execve()" arguments.
477 	 */
478 	addr = USRSTACK - PAGE_SIZE;
479 	if (vm_allocate(&p->p_vmspace->vm_map, &addr, (vm_size_t)PAGE_SIZE,
480 	    FALSE) != 0)
481 		panic("init: couldn't allocate argument space");
482 	p->p_vmspace->vm_maxsaddr = (caddr_t)addr;
483 
484 	for (pathp = &initpaths[0]; (path = *pathp) != NULL; pathp++) {
485 		ucp = (char *)(addr + PAGE_SIZE);
486 
487 		/*
488 		 * Construct the boot flag argument.
489 		 */
490 		flagsp = flags;
491 		*flagsp++ = '-';
492 		options = 0;
493 
494 		if (boothowto & RB_SINGLE) {
495 			*flagsp++ = 's';
496 			options = 1;
497 		}
498 #ifdef notyet
499 		if (boothowto & RB_FASTBOOT) {
500 			*flagsp++ = 'f';
501 			options = 1;
502 		}
503 #endif
504 
505 		/*
506 		 * Move out the flags (arg 1), if necessary.
507 		 */
508 		if (options != 0) {
509 			*flagsp++ = '\0';
510 			i = flagsp - flags;
511 #ifdef DEBUG
512 			printf("init: copying out flags `%s' %d\n", flags, i);
513 #endif
514 			(void)copyout((caddr_t)flags, (caddr_t)(ucp -= i), i);
515 			arg1 = ucp;
516 		}
517 
518 		/*
519 		 * Move out the file name (also arg 0).
520 		 */
521 		i = strlen(path) + 1;
522 #ifdef DEBUG
523 		printf("init: copying out path `%s' %d\n", path, i);
524 #endif
525 		(void)copyout((caddr_t)path, (caddr_t)(ucp -= i), i);
526 		arg0 = ucp;
527 
528 		/*
529 		 * Move out the arg pointers.
530 		 */
531 		uap = (char **)((long)ucp & ~ALIGNBYTES);
532 		(void)suword((caddr_t)--uap, 0);	/* terminator */
533 		if (options != 0)
534 			(void)suword((caddr_t)--uap, (long)arg1);
535 		(void)suword((caddr_t)--uap, (long)arg0);
536 
537 		/*
538 		 * Point at the arguments.
539 		 */
540 		SCARG(&args, path) = arg0;
541 		SCARG(&args, argp) = uap;
542 		SCARG(&args, envp) = NULL;
543 
544 		/*
545 		 * Now try to exec the program.  If can't for any reason
546 		 * other than it doesn't exist, complain.
547 		 */
548 		error = sys_execve(p, &args, retval);
549 		if (error == 0 || error == EJUSTRETURN)
550 			return;
551 		if (error != ENOENT)
552 			printf("exec %s: error %d\n", path, error);
553 	}
554 	printf("init: not found\n");
555 	panic("no init");
556 }
557 
558 static void
559 start_pagedaemon(p)
560 	struct proc *p;
561 {
562 
563 	/*
564 	 * Now in process 2.
565 	 */
566 	p->p_flag |= P_INMEM | P_SYSTEM;	/* XXX */
567 	bcopy("pagedaemon", curproc->p_comm, sizeof ("pagedaemon"));
568 	vm_pageout();
569 	/* NOTREACHED */
570 }
571