xref: /netbsd-src/sys/kern/init_main.c (revision bada23909e740596d0a3785a73bd3583a9807fb8)
1 /*	$NetBSD: init_main.c,v 1.142 1999/03/05 07:26:21 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	8.16 (Berkeley) 5/14/95
42  */
43 
44 #include "fs_nfs.h"
45 #include "opt_nfsserver.h"
46 #include "opt_uvm.h"
47 #include "opt_sysv.h"
48 
49 #include "rnd.h"
50 
51 #include <sys/param.h>
52 #include <sys/filedesc.h>
53 #include <sys/file.h>
54 #include <sys/errno.h>
55 #include <sys/exec.h>
56 #include <sys/kernel.h>
57 #include <sys/mount.h>
58 #include <sys/map.h>
59 #include <sys/proc.h>
60 #include <sys/kthread.h>
61 #include <sys/resourcevar.h>
62 #include <sys/signalvar.h>
63 #include <sys/systm.h>
64 #include <sys/vnode.h>
65 #include <sys/tty.h>
66 #include <sys/conf.h>
67 #include <sys/disklabel.h>
68 #include <sys/buf.h>
69 #ifdef REAL_CLISTS
70 #include <sys/clist.h>
71 #endif
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 <machine/cpu.h>
99 
100 #include <vm/vm.h>
101 #include <vm/vm_pageout.h>
102 
103 #if defined(UVM)
104 #include <uvm/uvm.h>
105 #endif
106 
107 #include <net/if.h>
108 #include <net/raw_cb.h>
109 
110 char	copyright[] = "\
111 Copyright (c) 1996, 1997, 1998, 1999
112     The NetBSD Foundation, Inc.  All rights reserved.
113 Copyright (c) 1982, 1986, 1989, 1991, 1993
114     The Regents of the University of California.  All rights reserved.
115 
116 ";
117 
118 /* Components of the first process -- never freed. */
119 struct	session session0;
120 struct	pgrp pgrp0;
121 struct	proc proc0;
122 struct	pcred cred0;
123 struct	filedesc0 filedesc0;
124 struct	plimit limit0;
125 struct	vmspace vmspace0;
126 #ifndef curproc
127 struct	proc *curproc = &proc0;
128 #endif
129 struct	proc *initproc;
130 
131 int	cmask = CMASK;
132 extern	struct user *proc0paddr;
133 
134 struct	vnode *rootvp, *swapdev_vp;
135 int	boothowto;
136 struct	timeval boottime;
137 struct	timeval runtime;
138 
139 static void check_console __P((struct proc *p));
140 static void start_init __P((void *));
141 static void start_pagedaemon __P((void *));
142 static void start_reaper __P((void *));
143 void main __P((void));
144 
145 extern char sigcode[], esigcode[];
146 #ifdef SYSCALL_DEBUG
147 extern char *syscallnames[];
148 #endif
149 
150 struct emul emul_netbsd = {
151 	"netbsd",
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 	sigcode,
166 	esigcode,
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 void
176 main()
177 {
178 	struct proc *p, *p2;
179 	struct pdevinit *pdev;
180 	int i, s, error;
181 	extern struct pdevinit pdevinit[];
182 	extern void roundrobin __P((void *));
183 	extern void schedcpu __P((void *));
184 	extern void disk_init __P((void));
185 #if defined(NFSSERVER) || defined(NFS)
186 	extern void nfs_init __P((void));
187 #endif
188 
189 	/*
190 	 * Initialize the current process pointer (curproc) before
191 	 * any possible traps/probes to simplify trap processing.
192 	 */
193 	p = &proc0;
194 	curproc = p;
195 	/*
196 	 * Attempt to find console and initialize
197 	 * in case of early panic or other messages.
198 	 */
199 	consinit();
200 	printf("%s", copyright);
201 
202 #if defined(UVM)
203 	uvm_init();
204 #else
205 	vm_mem_init();
206 	kmeminit();
207 	vm_page_physrehash();
208 #endif /* UVM */
209 
210 	/*
211 	 * Initialize mbuf's.  Do this now because we might attempt to
212 	 * allocate mbufs or mbuf clusters during autoconfiguration.
213 	 */
214 	mbinit();
215 
216 	/* Initialize sockets. */
217 	soinit();
218 
219 	disk_init();		/* must come before autoconfiguration */
220 	tty_init();		/* initialise tty list */
221 #if NRND > 0
222 	rnd_init();
223 #endif
224 	config_init();		/* init autoconfiguration data structures */
225 	cpu_startup();
226 
227 	/*
228 	 * Initialize process and pgrp structures.
229 	 */
230 	procinit();
231 
232 	/*
233 	 * Create process 0 (the swapper).
234 	 */
235 	LIST_INSERT_HEAD(&allproc, p, p_list);
236 	p->p_pgrp = &pgrp0;
237 	LIST_INSERT_HEAD(PGRPHASH(0), &pgrp0, pg_hash);
238 	LIST_INIT(&pgrp0.pg_members);
239 	LIST_INSERT_HEAD(&pgrp0.pg_members, p, p_pglist);
240 
241 	pgrp0.pg_session = &session0;
242 	session0.s_count = 1;
243 	session0.s_sid = p->p_pid;
244 	session0.s_leader = p;
245 
246 	/*
247 	 * Set P_NOCLDWAIT so that kernel threads are reparented to
248 	 * init(8) when they exit.  init(8) can easily wait them out
249 	 * for us.
250 	 */
251 	p->p_flag = P_INMEM | P_SYSTEM | P_NOCLDWAIT;
252 	p->p_stat = SRUN;
253 	p->p_nice = NZERO;
254 	p->p_emul = &emul_netbsd;
255 	memcpy(p->p_comm, "swapper", sizeof("swapper"));
256 
257 	/* Create credentials. */
258 	cred0.p_refcnt = 1;
259 	p->p_cred = &cred0;
260 	p->p_ucred = crget();
261 	p->p_ucred->cr_ngroups = 1;	/* group 0 */
262 
263 	/* Create the file descriptor table. */
264 	finit();
265 	p->p_fd = &filedesc0.fd_fd;
266 	fdinit1(&filedesc0);
267 
268 	/* Create the limits structures. */
269 	p->p_limit = &limit0;
270 	for (i = 0; i < sizeof(p->p_rlimit)/sizeof(p->p_rlimit[0]); i++)
271 		limit0.pl_rlimit[i].rlim_cur =
272 		    limit0.pl_rlimit[i].rlim_max = RLIM_INFINITY;
273 
274 	limit0.pl_rlimit[RLIMIT_NOFILE].rlim_max = maxfiles;
275 	limit0.pl_rlimit[RLIMIT_NOFILE].rlim_cur =
276 	    maxfiles < NOFILE ? maxfiles : NOFILE;
277 
278 	limit0.pl_rlimit[RLIMIT_NPROC].rlim_max = maxproc;
279 	limit0.pl_rlimit[RLIMIT_NPROC].rlim_cur =
280 	    maxproc < MAXUPRC ? maxproc : MAXUPRC;
281 
282 #if defined(UVM)
283 	i = ptoa(uvmexp.free);
284 #else
285 	i = ptoa(cnt.v_free_count);
286 #endif
287 	limit0.pl_rlimit[RLIMIT_RSS].rlim_max = i;
288 	limit0.pl_rlimit[RLIMIT_MEMLOCK].rlim_max = i;
289 	limit0.pl_rlimit[RLIMIT_MEMLOCK].rlim_cur = i / 3;
290 	limit0.p_refcnt = 1;
291 
292 	/*
293 	 * Initialize proc0's vmspace, which uses the kernel pmap.
294 	 * All kernel processes (which never have user space mappings)
295 	 * share proc0's vmspace, and thus, the kernel pmap.
296 	 */
297 #if defined(UVM)
298 	uvmspace_init(&vmspace0, pmap_kernel(), round_page(VM_MIN_ADDRESS),
299 	    trunc_page(VM_MAX_ADDRESS), TRUE);
300 #else
301 	vmspace_init(&vmspace0, pmap_kernel(), round_page(VM_MIN_ADDRESS),
302 	    trunc_page(VM_MAX_ADDRESS), TRUE);
303 #endif
304 	p->p_vmspace = &vmspace0;
305 
306 	p->p_addr = proc0paddr;				/* XXX */
307 
308 	/*
309 	 * We continue to place resource usage info and signal
310 	 * actions in the user struct so they're pageable.
311 	 */
312 	p->p_stats = &p->p_addr->u_stats;
313 	p->p_sigacts = &p->p_addr->u_sigacts;
314 
315 	/*
316 	 * Charge root for one process.
317 	 */
318 	(void)chgproccnt(0, 1);
319 
320 	rqinit();
321 
322 	/* Configure virtual memory system, set vm rlimits. */
323 #if defined(UVM)
324 	uvm_init_limits(p);
325 #else
326 	vm_init_limits(p);
327 #endif
328 
329 	/* Initialize the file systems. */
330 #if defined(NFSSERVER) || defined(NFS)
331 	nfs_init();			/* initialize server/shared data */
332 #endif
333 	vfsinit();
334 
335 	/* Start real time and statistics clocks. */
336 	initclocks();
337 
338 #ifdef REAL_CLISTS
339 	/* Initialize clists. */
340 	clist_init();
341 #endif
342 
343 #ifdef SYSVSHM
344 	/* Initialize System V style shared memory. */
345 	shminit();
346 #endif
347 
348 #ifdef SYSVSEM
349 	/* Initialize System V style semaphores. */
350 	seminit();
351 #endif
352 
353 #ifdef SYSVMSG
354 	/* Initialize System V style message queues. */
355 	msginit();
356 #endif
357 
358 	/* Attach pseudo-devices. */
359 	for (pdev = pdevinit; pdev->pdev_attach != NULL; pdev++)
360 		(*pdev->pdev_attach)(pdev->pdev_count);
361 
362 	/*
363 	 * Initialize protocols.  Block reception of incoming packets
364 	 * until everything is ready.
365 	 */
366 	s = splimp();
367 	ifinit();
368 	domaininit();
369 	splx(s);
370 
371 #ifdef GPROF
372 	/* Initialize kernel profiling. */
373 	kmstartup();
374 #endif
375 
376 	/* Kick off timeout driven events by calling first time. */
377 	roundrobin(NULL);
378 	schedcpu(NULL);
379 
380 	/* Determine the root and dump devices. */
381 	cpu_rootconf();
382 	cpu_dumpconf();
383 
384 	/* Mount the root file system. */
385 	do {
386 		domountroothook();
387 		if ((error = vfs_mountroot())) {
388 			printf("cannot mount root, error = %d\n", error);
389 			boothowto |= RB_ASKNAME;
390 			setroot(root_device,
391 			    (rootdev != NODEV) ? DISKPART(rootdev) : 0, NULL);
392 		}
393 	} while (error != 0);
394 	mountroothook_destroy();
395 
396 	mountlist.cqh_first->mnt_flag |= MNT_ROOTFS;
397 	mountlist.cqh_first->mnt_op->vfs_refcount++;
398 
399 	/*
400 	 * Get the vnode for '/'.  Set filedesc0.fd_fd.fd_cdir to
401 	 * reference it.
402 	 */
403 	if (VFS_ROOT(mountlist.cqh_first, &rootvnode))
404 		panic("cannot find root vnode");
405 	filedesc0.fd_fd.fd_cdir = rootvnode;
406 	VREF(filedesc0.fd_fd.fd_cdir);
407 	VOP_UNLOCK(rootvnode, 0);
408 	filedesc0.fd_fd.fd_rdir = NULL;
409 #if defined(UVM)
410 	uvm_swap_init();
411 #else
412 	swapinit();
413 #endif
414 
415 	/*
416 	 * Now can look at time, having had a chance to verify the time
417 	 * from the file system.  Reset p->p_rtime as it may have been
418 	 * munched in mi_switch() after the time got set.
419 	 */
420 	p->p_stats->p_start = runtime = mono_time = boottime = time;
421 	p->p_rtime.tv_sec = p->p_rtime.tv_usec = 0;
422 
423 	/* Initialize signal state for process 0. */
424 	siginit(p);
425 
426 	/* Create process 1 (init(8)). */
427 	if (fork1(p, 0, NULL, &p2))
428 		panic("fork init");
429 	cpu_set_kpc(p2, start_init, p2);
430 
431 	/* Create process 2, the pageout daemon kernel thread. */
432 	if (kthread_create(start_pagedaemon, NULL, NULL, "pagedaemon"))
433 		panic("fork pagedaemon");
434 
435 	/* Create process 3, the process reaper kernel thread. */
436 	if (kthread_create(start_reaper, NULL, NULL, "reaper"))
437 		panic("fork reaper");
438 
439 	/* Create any other deferred kernel threads. */
440 	kthread_run_deferred_queue();
441 
442 	/* The scheduler is an infinite loop. */
443 #if defined(UVM)
444 	uvm_scheduler();
445 #else
446 	scheduler();
447 #endif
448 	/* NOTREACHED */
449 }
450 
451 static void
452 check_console(p)
453 	struct proc *p;
454 {
455 	struct nameidata nd;
456 	int error;
457 
458 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_SYSSPACE, "/dev/console", p);
459 	error = namei(&nd);
460 	if (error == 0)
461 		vrele(nd.ni_vp);
462 	else if (error == ENOENT)
463 		printf("warning: no /dev/console\n");
464 	else
465 		printf("warning: lookup /dev/console: error %d\n", error);
466 }
467 
468 /*
469  * List of paths to try when searching for "init".
470  */
471 static char *initpaths[] = {
472 	"/sbin/init",
473 	"/sbin/oinit",
474 	"/sbin/init.bak",
475 	NULL,
476 };
477 
478 /*
479  * Start the initial user process; try exec'ing each pathname in "initpaths".
480  * The program is invoked with one argument containing the boot flags.
481  */
482 static void
483 start_init(arg)
484 	void *arg;
485 {
486 	struct proc *p = arg;
487 	vaddr_t addr;
488 	struct sys_execve_args /* {
489 		syscallarg(const char *) path;
490 		syscallarg(char * const *) argp;
491 		syscallarg(char * const *) envp;
492 	} */ args;
493 	int options, i, error;
494 	register_t retval[2];
495 	char flags[4], *flagsp;
496 	char **pathp, *path, *slash, *ucp, **uap, *arg0, *arg1 = NULL;
497 
498 	/*
499 	 * Now in process 1.
500 	 */
501 	initproc = p;
502 
503 	/*
504 	 * This is not the right way to do this.  We really should
505 	 * hand-craft a descriptor onto /dev/console to hand to init,
506 	 * but that's a _lot_ more work, and the benefit from this easy
507 	 * hack makes up for the "good is the enemy of the best" effect.
508 	 */
509 	check_console(p);
510 
511 	/*
512 	 * Need just enough stack to hold the faked-up "execve()" arguments.
513 	 */
514 	addr = USRSTACK - PAGE_SIZE;
515 #if defined(UVM)
516 	if (uvm_map(&p->p_vmspace->vm_map, &addr, PAGE_SIZE,
517                     NULL, UVM_UNKNOWN_OFFSET,
518                     UVM_MAPFLAG(UVM_PROT_ALL, UVM_PROT_ALL, UVM_INH_COPY,
519 		    UVM_ADV_NORMAL,
520                     UVM_FLAG_FIXED|UVM_FLAG_OVERLAY|UVM_FLAG_COPYONW))
521 		!= KERN_SUCCESS)
522 		panic("init: couldn't allocate argument space");
523 #else
524 	if (vm_allocate(&p->p_vmspace->vm_map, &addr, (vsize_t)PAGE_SIZE,
525 	    FALSE) != 0)
526 		panic("init: couldn't allocate argument space");
527 #endif
528 	p->p_vmspace->vm_maxsaddr = (caddr_t)addr;
529 
530 	for (pathp = &initpaths[0]; (path = *pathp) != NULL; pathp++) {
531 		ucp = (char *)(addr + PAGE_SIZE);
532 
533 		/*
534 		 * Construct the boot flag argument.
535 		 */
536 		flagsp = flags;
537 		*flagsp++ = '-';
538 		options = 0;
539 
540 		if (boothowto & RB_SINGLE) {
541 			*flagsp++ = 's';
542 			options = 1;
543 		}
544 #ifdef notyet
545 		if (boothowto & RB_FASTBOOT) {
546 			*flagsp++ = 'f';
547 			options = 1;
548 		}
549 #endif
550 
551 		/*
552 		 * Move out the flags (arg 1), if necessary.
553 		 */
554 		if (options != 0) {
555 			*flagsp++ = '\0';
556 			i = flagsp - flags;
557 #ifdef DEBUG
558 			printf("init: copying out flags `%s' %d\n", flags, i);
559 #endif
560 			(void)copyout((caddr_t)flags, (caddr_t)(ucp -= i), i);
561 			arg1 = ucp;
562 		}
563 
564 		/*
565 		 * Move out the file name (also arg 0).
566 		 */
567 		i = strlen(path) + 1;
568 #ifdef DEBUG
569 		printf("init: copying out path `%s' %d\n", path, i);
570 #endif
571 		(void)copyout((caddr_t)path, (caddr_t)(ucp -= i), i);
572 		arg0 = ucp;
573 
574 		/*
575 		 * Move out the arg pointers.
576 		 */
577 		uap = (char **)((long)ucp & ~ALIGNBYTES);
578 		(void)suword((caddr_t)--uap, 0);	/* terminator */
579 		if (options != 0)
580 			(void)suword((caddr_t)--uap, (long)arg1);
581 		slash = strrchr(path, '/');
582 		if (slash)
583 			(void)suword((caddr_t)--uap,
584 			    (long)arg0 + (slash + 1 - path));
585 		else
586 			(void)suword((caddr_t)--uap, (long)arg0);
587 
588 		/*
589 		 * Point at the arguments.
590 		 */
591 		SCARG(&args, path) = arg0;
592 		SCARG(&args, argp) = uap;
593 		SCARG(&args, envp) = NULL;
594 
595 		/*
596 		 * Now try to exec the program.  If can't for any reason
597 		 * other than it doesn't exist, complain.
598 		 */
599 		error = sys_execve(p, &args, retval);
600 		if (error == 0 || error == EJUSTRETURN)
601 			return;
602 		if (error != ENOENT)
603 			printf("exec %s: error %d\n", path, error);
604 	}
605 	printf("init: not found\n");
606 	panic("no init");
607 }
608 
609 /* ARGSUSED */
610 static void
611 start_pagedaemon(arg)
612 	void *arg;
613 {
614 
615 #if defined(UVM)
616 	uvm_pageout();
617 #else
618 	vm_pageout();
619 #endif
620 	/* NOTREACHED */
621 }
622 
623 /* ARGSUSED */
624 static void
625 start_reaper(arg)
626 	void *arg;
627 {
628 
629 	reaper();
630 	/* NOTREACHED */
631 }
632