xref: /netbsd-src/sys/kern/init_main.c (revision 84d0ab551791493d2630bbef27063a9d514b9108)
1 /*	$NetBSD: init_main.c,v 1.106 1997/10/13 00:48:07 explorer 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) 1982, 1986, 1989, 1991, 1993\n    The Regents of the University of California.  All rights reserved.\n\n";
100 
101 /* Components of the first process -- never freed. */
102 struct	session session0;
103 struct	pgrp pgrp0;
104 struct	proc proc0;
105 struct	pcred cred0;
106 struct	filedesc0 filedesc0;
107 struct	plimit limit0;
108 struct	vmspace vmspace0;
109 struct	proc *curproc = &proc0;
110 struct	proc *initproc;
111 
112 int	cmask = CMASK;
113 extern	struct user *proc0paddr;
114 
115 struct	vnode *rootvp, *swapdev_vp;
116 int	boothowto;
117 struct	timeval boottime;
118 struct	timeval runtime;
119 
120 static void check_console __P((struct proc *p));
121 static void start_init __P((struct proc *));
122 static void start_pagedaemon __P((struct proc *));
123 void main __P((void *));
124 
125 #ifdef cpu_set_init_frame
126 void *initframep;				/* XXX should go away */
127 #endif
128 
129 extern char sigcode[], esigcode[];
130 #ifdef SYSCALL_DEBUG
131 extern char *syscallnames[];
132 #endif
133 
134 struct emul emul_netbsd = {
135 	"netbsd",
136 	NULL,
137 	sendsig,
138 	SYS_syscall,
139 	SYS_MAXSYSCALL,
140 	sysent,
141 #ifdef SYSCALL_DEBUG
142 	syscallnames,
143 #else
144 	NULL,
145 #endif
146 	0,
147 	copyargs,
148 	setregs,
149 	sigcode,
150 	esigcode,
151 };
152 
153 /*
154  * System startup; initialize the world, create process 0, mount root
155  * filesystem, and fork to create init and pagedaemon.  Most of the
156  * hard work is done in the lower-level initialization routines including
157  * startup(), which does memory initialization and autoconfiguration.
158  */
159 void
160 main(framep)
161 	void *framep;				/* XXX should go away */
162 {
163 	register struct proc *p;
164 	register struct pdevinit *pdev;
165 	register int i;
166 	int s, error;
167 	register_t rval[2];
168 	extern struct pdevinit pdevinit[];
169 	extern void roundrobin __P((void *));
170 	extern void schedcpu __P((void *));
171 	extern void disk_init __P((void));
172 #if defined(NFSSERVER) || defined(NFS)
173 	extern void nfs_init __P((void));
174 #endif
175 
176 	/*
177 	 * Initialize the current process pointer (curproc) before
178 	 * any possible traps/probes to simplify trap processing.
179 	 */
180 	p = &proc0;
181 	curproc = p;
182 	/*
183 	 * Attempt to find console and initialize
184 	 * in case of early panic or other messages.
185 	 */
186 	consinit();
187 	printf(copyright);
188 
189 	vm_mem_init();
190 	kmeminit();
191 	disk_init();		/* must come before autoconfiguration */
192 	tty_init();		/* initialise tty list */
193 #if NRND > 0
194 	rnd_init();
195 #endif
196 	config_init();		/* init autoconfiguration data structures */
197 	cpu_startup();
198 
199 	/*
200 	 * Initialize process and pgrp structures.
201 	 */
202 	procinit();
203 
204 	/*
205 	 * Create process 0 (the swapper).
206 	 */
207 	LIST_INSERT_HEAD(&allproc, p, p_list);
208 	p->p_pgrp = &pgrp0;
209 	LIST_INSERT_HEAD(PGRPHASH(0), &pgrp0, pg_hash);
210 	LIST_INIT(&pgrp0.pg_members);
211 	LIST_INSERT_HEAD(&pgrp0.pg_members, p, p_pglist);
212 
213 	pgrp0.pg_session = &session0;
214 	session0.s_count = 1;
215 	session0.s_leader = p;
216 
217 	p->p_flag = P_INMEM | P_SYSTEM;
218 	p->p_stat = SRUN;
219 	p->p_nice = NZERO;
220 	p->p_emul = &emul_netbsd;
221 	bcopy("swapper", p->p_comm, sizeof ("swapper"));
222 
223 	/* Create credentials. */
224 	cred0.p_refcnt = 1;
225 	p->p_cred = &cred0;
226 	p->p_ucred = crget();
227 	p->p_ucred->cr_ngroups = 1;	/* group 0 */
228 
229 	/* Create the file descriptor table. */
230 	p->p_fd = &filedesc0.fd_fd;
231 	filedesc0.fd_fd.fd_refcnt = 1;
232 	filedesc0.fd_fd.fd_cmask = cmask;
233 	filedesc0.fd_fd.fd_ofiles = filedesc0.fd_dfiles;
234 	filedesc0.fd_fd.fd_ofileflags = filedesc0.fd_dfileflags;
235 	filedesc0.fd_fd.fd_nfiles = NDFILE;
236 
237 	/* Create the limits structures. */
238 	p->p_limit = &limit0;
239 	for (i = 0; i < sizeof(p->p_rlimit)/sizeof(p->p_rlimit[0]); i++)
240 		limit0.pl_rlimit[i].rlim_cur =
241 		    limit0.pl_rlimit[i].rlim_max = RLIM_INFINITY;
242 	limit0.pl_rlimit[RLIMIT_NOFILE].rlim_cur = NOFILE;
243 	limit0.pl_rlimit[RLIMIT_NPROC].rlim_cur = MAXUPRC;
244 	i = ptoa(cnt.v_free_count);
245 	limit0.pl_rlimit[RLIMIT_RSS].rlim_max = i;
246 	limit0.pl_rlimit[RLIMIT_MEMLOCK].rlim_max = i;
247 	limit0.pl_rlimit[RLIMIT_MEMLOCK].rlim_cur = i / 3;
248 	limit0.p_refcnt = 1;
249 
250 	/* Allocate a prototype map so we have something to fork. */
251 	p->p_vmspace = vmspace_alloc(round_page(VM_MIN_ADDRESS),
252 				     trunc_page(VM_MAX_ADDRESS), TRUE);
253 
254 	p->p_addr = proc0paddr;				/* XXX */
255 
256 	/*
257 	 * We continue to place resource usage info and signal
258 	 * actions in the user struct so they're pageable.
259 	 */
260 	p->p_stats = &p->p_addr->u_stats;
261 	p->p_sigacts = &p->p_addr->u_sigacts;
262 
263 	/*
264 	 * Charge root for one process.
265 	 */
266 	(void)chgproccnt(0, 1);
267 
268 	rqinit();
269 
270 	/* Configure virtual memory system, set vm rlimits. */
271 	vm_init_limits(p);
272 
273 	/* Initialize the file systems. */
274 #if defined(NFSSERVER) || defined(NFS)
275 	nfs_init();			/* initialize server/shared data */
276 #endif
277 	vfsinit();
278 
279 	/* Start real time and statistics clocks. */
280 	initclocks();
281 
282 	/* Initialize mbuf's. */
283 	mbinit();
284 
285 #ifdef REAL_CLISTS
286 	/* Initialize clists. */
287 	clist_init();
288 #endif
289 
290 #ifdef SYSVSHM
291 	/* Initialize System V style shared memory. */
292 	shminit();
293 #endif
294 
295 #ifdef SYSVSEM
296 	/* Initialize System V style semaphores. */
297 	seminit();
298 #endif
299 
300 #ifdef SYSVMSG
301 	/* Initialize System V style message queues. */
302 	msginit();
303 #endif
304 
305 	/* Attach pseudo-devices. */
306 	for (pdev = pdevinit; pdev->pdev_attach != NULL; pdev++)
307 		(*pdev->pdev_attach)(pdev->pdev_count);
308 
309 	/*
310 	 * Initialize protocols.  Block reception of incoming packets
311 	 * until everything is ready.
312 	 */
313 	s = splimp();
314 	ifinit();
315 	domaininit();
316 	splx(s);
317 
318 #ifdef GPROF
319 	/* Initialize kernel profiling. */
320 	kmstartup();
321 #endif
322 
323 	/* Kick off timeout driven events by calling first time. */
324 	roundrobin(NULL);
325 	schedcpu(NULL);
326 
327 	/* Determine the root and dump devices. */
328 	cpu_rootconf();
329 	cpu_dumpconf();
330 
331 	/* Mount the root file system. */
332 	do {
333 		domountroothook();
334 		if ((error = vfs_mountroot())) {
335 			printf("cannot mount root, error = %d\n", error);
336 			boothowto |= RB_ASKNAME;
337 			setroot(root_device,
338 			    (rootdev != NODEV) ? DISKPART(rootdev) : 0, NULL);
339 		}
340 	} while (error != 0);
341 	mountroothook_destroy();
342 
343 	mountlist.cqh_first->mnt_flag |= MNT_ROOTFS;
344 	mountlist.cqh_first->mnt_op->vfs_refcount++;
345 
346 	/* Get the vnode for '/'.  Set filedesc0.fd_fd.fd_cdir to reference it. */
347 	if (VFS_ROOT(mountlist.cqh_first, &rootvnode))
348 		panic("cannot find root vnode");
349 	filedesc0.fd_fd.fd_cdir = rootvnode;
350 	VREF(filedesc0.fd_fd.fd_cdir);
351 	VOP_UNLOCK(rootvnode);
352 	filedesc0.fd_fd.fd_rdir = NULL;
353 	swapinit();
354 
355 	/*
356 	 * Now can look at time, having had a chance to verify the time
357 	 * from the file system.  Reset p->p_rtime as it may have been
358 	 * munched in mi_switch() after the time got set.
359 	 */
360 	p->p_stats->p_start = runtime = mono_time = boottime = time;
361 	p->p_rtime.tv_sec = p->p_rtime.tv_usec = 0;
362 
363 	/* Initialize signal state for process 0. */
364 	siginit(p);
365 
366 	/* Create process 1 (init(8)). */
367 	if (sys_fork(p, NULL, rval))
368 		panic("fork init");
369 #ifdef cpu_set_init_frame			/* XXX should go away */
370 	if (rval[1]) {
371 		/*
372 		 * Now in process 1.
373 		 */
374 		initframep = framep;
375 		start_init(curproc);
376 		return;
377 	}
378 #else
379 	cpu_set_kpc(pfind(1), start_init);
380 #endif
381 
382 	/* Create process 2 (the pageout daemon). */
383 	if (sys_fork(p, NULL, rval))
384 		panic("fork pager");
385 #ifdef cpu_set_init_frame			/* XXX should go away */
386 	if (rval[1]) {
387 		/*
388 		 * Now in process 2.
389 		 */
390 		start_pagedaemon(curproc);
391 	}
392 #else
393 	cpu_set_kpc(pfind(2), start_pagedaemon);
394 #endif
395 
396 	/* The scheduler is an infinite loop. */
397 	scheduler();
398 	/* NOTREACHED */
399 }
400 
401 static void
402 check_console(p)
403 	struct proc *p;
404 {
405 	struct nameidata nd;
406 	int error;
407 
408 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_SYSSPACE, "/dev/console", p);
409 	error = namei(&nd);
410 	if (error == 0)
411 		vrele(nd.ni_vp);
412 	else if (error == ENOENT)
413 		printf("warning: no /dev/console\n");
414 	else
415 		printf("warning: lookup /dev/console: error %d\n", error);
416 }
417 
418 /*
419  * List of paths to try when searching for "init".
420  */
421 static char *initpaths[] = {
422 	"/sbin/init",
423 	"/sbin/oinit",
424 	"/sbin/init.bak",
425 	NULL,
426 };
427 
428 /*
429  * Start the initial user process; try exec'ing each pathname in "initpaths".
430  * The program is invoked with one argument containing the boot flags.
431  */
432 static void
433 start_init(p)
434 	struct proc *p;
435 {
436 	vm_offset_t addr;
437 	struct sys_execve_args /* {
438 		syscallarg(char *) path;
439 		syscallarg(char * const *) argp;
440 		syscallarg(char * const *) envp;
441 	} */ args;
442 	int options, i, error;
443 	register_t retval[2];
444 	char flags[4], *flagsp;
445 	char **pathp, *path, *ucp, **uap, *arg0, *arg1 = NULL;
446 
447 	/*
448 	 * Now in process 1.
449 	 */
450 	initproc = p;
451 
452 #ifdef cpu_set_init_frame			/* XXX should go away */
453 	/*
454 	 * We need to set the system call frame as if we were entered through
455 	 * a syscall() so that when we call sys_execve() below, it will be able
456 	 * to set the entry point (see setregs) when it tries to exec.  The
457 	 * startup code in "locore.s" has allocated space for the frame and
458 	 * passed a pointer to that space as main's argument.
459 	 */
460 	cpu_set_init_frame(p, initframep);
461 #endif
462 
463 	/*
464 	 * This is not the right way to do this.  We really should
465 	 * hand-craft a descriptor onto /dev/console to hand to init,
466 	 * but that's a _lot_ more work, and the benefit from this easy
467 	 * hack makes up for the "good is the enemy of the best" effect.
468 	 */
469 	check_console(p);
470 
471 	/*
472 	 * Need just enough stack to hold the faked-up "execve()" arguments.
473 	 */
474 	addr = USRSTACK - PAGE_SIZE;
475 	if (vm_allocate(&p->p_vmspace->vm_map, &addr, (vm_size_t)PAGE_SIZE,
476 	    FALSE) != 0)
477 		panic("init: couldn't allocate argument space");
478 	p->p_vmspace->vm_maxsaddr = (caddr_t)addr;
479 
480 	for (pathp = &initpaths[0]; (path = *pathp) != NULL; pathp++) {
481 		ucp = (char *)(addr + PAGE_SIZE);
482 
483 		/*
484 		 * Construct the boot flag argument.
485 		 */
486 		flagsp = flags;
487 		*flagsp++ = '-';
488 		options = 0;
489 
490 		if (boothowto & RB_SINGLE) {
491 			*flagsp++ = 's';
492 			options = 1;
493 		}
494 #ifdef notyet
495 		if (boothowto & RB_FASTBOOT) {
496 			*flagsp++ = 'f';
497 			options = 1;
498 		}
499 #endif
500 
501 		/*
502 		 * Move out the flags (arg 1), if necessary.
503 		 */
504 		if (options != 0) {
505 			*flagsp++ = '\0';
506 			i = flagsp - flags;
507 #ifdef DEBUG
508 			printf("init: copying out flags `%s' %d\n", flags, i);
509 #endif
510 			(void)copyout((caddr_t)flags, (caddr_t)(ucp -= i), i);
511 			arg1 = ucp;
512 		}
513 
514 		/*
515 		 * Move out the file name (also arg 0).
516 		 */
517 		i = strlen(path) + 1;
518 #ifdef DEBUG
519 		printf("init: copying out path `%s' %d\n", path, i);
520 #endif
521 		(void)copyout((caddr_t)path, (caddr_t)(ucp -= i), i);
522 		arg0 = ucp;
523 
524 		/*
525 		 * Move out the arg pointers.
526 		 */
527 		uap = (char **)((long)ucp & ~ALIGNBYTES);
528 		(void)suword((caddr_t)--uap, 0);	/* terminator */
529 		if (options != 0)
530 			(void)suword((caddr_t)--uap, (long)arg1);
531 		(void)suword((caddr_t)--uap, (long)arg0);
532 
533 		/*
534 		 * Point at the arguments.
535 		 */
536 		SCARG(&args, path) = arg0;
537 		SCARG(&args, argp) = uap;
538 		SCARG(&args, envp) = NULL;
539 
540 		/*
541 		 * Now try to exec the program.  If can't for any reason
542 		 * other than it doesn't exist, complain.
543 		 */
544 		error = sys_execve(p, &args, retval);
545 		if (error == 0 || error == EJUSTRETURN)
546 			return;
547 		if (error != ENOENT)
548 			printf("exec %s: error %d\n", path, error);
549 	}
550 	printf("init: not found\n");
551 	panic("no init");
552 }
553 
554 static void
555 start_pagedaemon(p)
556 	struct proc *p;
557 {
558 
559 	/*
560 	 * Now in process 2.
561 	 */
562 	p->p_flag |= P_INMEM | P_SYSTEM;	/* XXX */
563 	bcopy("pagedaemon", curproc->p_comm, sizeof ("pagedaemon"));
564 	vm_pageout();
565 	/* NOTREACHED */
566 }
567