xref: /netbsd-src/sys/kern/init_main.c (revision dc306354b0b29af51801a7632f1e95265a68cd81)
1 /*	$NetBSD: init_main.c,v 1.138 1999/01/06 13:51:09 lukem 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(copyright);
201 
202 #if defined(UVM)
203 	uvm_init();
204 #else
205 	vm_mem_init();
206 	kmeminit();
207 #if defined(MACHINE_NEW_NONCONTIG)
208 	vm_page_physrehash();
209 #endif
210 #endif /* UVM */
211 
212 	/*
213 	 * Initialize mbuf's.  Do this now because we might attempt to
214 	 * allocate mbufs or mbuf clusters during autoconfiguration.
215 	 */
216 	mbinit();
217 
218 	/* Initialize sockets. */
219 	soinit();
220 
221 	disk_init();		/* must come before autoconfiguration */
222 	tty_init();		/* initialise tty list */
223 #if NRND > 0
224 	rnd_init();
225 #endif
226 	config_init();		/* init autoconfiguration data structures */
227 	cpu_startup();
228 
229 	/*
230 	 * Initialize process and pgrp structures.
231 	 */
232 	procinit();
233 
234 	/*
235 	 * Create process 0 (the swapper).
236 	 */
237 	LIST_INSERT_HEAD(&allproc, p, p_list);
238 	p->p_pgrp = &pgrp0;
239 	LIST_INSERT_HEAD(PGRPHASH(0), &pgrp0, pg_hash);
240 	LIST_INIT(&pgrp0.pg_members);
241 	LIST_INSERT_HEAD(&pgrp0.pg_members, p, p_pglist);
242 
243 	pgrp0.pg_session = &session0;
244 	session0.s_count = 1;
245 	session0.s_sid = p->p_pid;
246 	session0.s_leader = p;
247 
248 	/*
249 	 * Set P_NOCLDWAIT so that kernel threads are reparented to
250 	 * init(8) when they exit.  init(8) can easily wait them out
251 	 * for us.
252 	 */
253 	p->p_flag = P_INMEM | P_SYSTEM | P_NOCLDWAIT;
254 	p->p_stat = SRUN;
255 	p->p_nice = NZERO;
256 	p->p_emul = &emul_netbsd;
257 	memcpy(p->p_comm, "swapper", sizeof("swapper"));
258 
259 	/* Create credentials. */
260 	cred0.p_refcnt = 1;
261 	p->p_cred = &cred0;
262 	p->p_ucred = crget();
263 	p->p_ucred->cr_ngroups = 1;	/* group 0 */
264 
265 	/* Create the file descriptor table. */
266 	finit();
267 	p->p_fd = &filedesc0.fd_fd;
268 	fdinit1(&filedesc0);
269 
270 	/* Create the limits structures. */
271 	p->p_limit = &limit0;
272 	for (i = 0; i < sizeof(p->p_rlimit)/sizeof(p->p_rlimit[0]); i++)
273 		limit0.pl_rlimit[i].rlim_cur =
274 		    limit0.pl_rlimit[i].rlim_max = RLIM_INFINITY;
275 	limit0.pl_rlimit[RLIMIT_NOFILE].rlim_cur = NOFILE;
276 	limit0.pl_rlimit[RLIMIT_NPROC].rlim_cur = MAXUPRC;
277 #if defined(UVM)
278 	i = ptoa(uvmexp.free);
279 #else
280 	i = ptoa(cnt.v_free_count);
281 #endif
282 	limit0.pl_rlimit[RLIMIT_RSS].rlim_max = i;
283 	limit0.pl_rlimit[RLIMIT_MEMLOCK].rlim_max = i;
284 	limit0.pl_rlimit[RLIMIT_MEMLOCK].rlim_cur = i / 3;
285 	limit0.p_refcnt = 1;
286 
287 	/*
288 	 * Initialize proc0's vmspace, which uses the kernel pmap.
289 	 * All kernel processes (which never have user space mappings)
290 	 * share proc0's vmspace, and thus, the kernel pmap.
291 	 */
292 #if defined(UVM)
293 	uvmspace_init(&vmspace0, pmap_kernel(), round_page(VM_MIN_ADDRESS),
294 	    trunc_page(VM_MAX_ADDRESS), TRUE);
295 #else
296 	vmspace_init(&vmspace0, pmap_kernel(), round_page(VM_MIN_ADDRESS),
297 	    trunc_page(VM_MAX_ADDRESS), TRUE);
298 #endif
299 	p->p_vmspace = &vmspace0;
300 
301 	p->p_addr = proc0paddr;				/* XXX */
302 
303 	/*
304 	 * We continue to place resource usage info and signal
305 	 * actions in the user struct so they're pageable.
306 	 */
307 	p->p_stats = &p->p_addr->u_stats;
308 	p->p_sigacts = &p->p_addr->u_sigacts;
309 
310 	/*
311 	 * Charge root for one process.
312 	 */
313 	(void)chgproccnt(0, 1);
314 
315 	rqinit();
316 
317 	/* Configure virtual memory system, set vm rlimits. */
318 #if defined(UVM)
319 	uvm_init_limits(p);
320 #else
321 	vm_init_limits(p);
322 #endif
323 
324 	/* Initialize the file systems. */
325 #if defined(NFSSERVER) || defined(NFS)
326 	nfs_init();			/* initialize server/shared data */
327 #endif
328 	vfsinit();
329 
330 	/* Start real time and statistics clocks. */
331 	initclocks();
332 
333 #ifdef REAL_CLISTS
334 	/* Initialize clists. */
335 	clist_init();
336 #endif
337 
338 #ifdef SYSVSHM
339 	/* Initialize System V style shared memory. */
340 	shminit();
341 #endif
342 
343 #ifdef SYSVSEM
344 	/* Initialize System V style semaphores. */
345 	seminit();
346 #endif
347 
348 #ifdef SYSVMSG
349 	/* Initialize System V style message queues. */
350 	msginit();
351 #endif
352 
353 	/* Attach pseudo-devices. */
354 	for (pdev = pdevinit; pdev->pdev_attach != NULL; pdev++)
355 		(*pdev->pdev_attach)(pdev->pdev_count);
356 
357 	/*
358 	 * Initialize protocols.  Block reception of incoming packets
359 	 * until everything is ready.
360 	 */
361 	s = splimp();
362 	ifinit();
363 	domaininit();
364 	splx(s);
365 
366 #ifdef GPROF
367 	/* Initialize kernel profiling. */
368 	kmstartup();
369 #endif
370 
371 	/* Kick off timeout driven events by calling first time. */
372 	roundrobin(NULL);
373 	schedcpu(NULL);
374 
375 	/* Determine the root and dump devices. */
376 	cpu_rootconf();
377 	cpu_dumpconf();
378 
379 	/* Mount the root file system. */
380 	do {
381 		domountroothook();
382 		if ((error = vfs_mountroot())) {
383 			printf("cannot mount root, error = %d\n", error);
384 			boothowto |= RB_ASKNAME;
385 			setroot(root_device,
386 			    (rootdev != NODEV) ? DISKPART(rootdev) : 0, NULL);
387 		}
388 	} while (error != 0);
389 	mountroothook_destroy();
390 
391 	mountlist.cqh_first->mnt_flag |= MNT_ROOTFS;
392 	mountlist.cqh_first->mnt_op->vfs_refcount++;
393 
394 	/*
395 	 * Get the vnode for '/'.  Set filedesc0.fd_fd.fd_cdir to
396 	 * reference it.
397 	 */
398 	if (VFS_ROOT(mountlist.cqh_first, &rootvnode))
399 		panic("cannot find root vnode");
400 	filedesc0.fd_fd.fd_cdir = rootvnode;
401 	VREF(filedesc0.fd_fd.fd_cdir);
402 	VOP_UNLOCK(rootvnode, 0);
403 	filedesc0.fd_fd.fd_rdir = NULL;
404 #if defined(UVM)
405 	uvm_swap_init();
406 #else
407 	swapinit();
408 #endif
409 
410 	/*
411 	 * Now can look at time, having had a chance to verify the time
412 	 * from the file system.  Reset p->p_rtime as it may have been
413 	 * munched in mi_switch() after the time got set.
414 	 */
415 	p->p_stats->p_start = runtime = mono_time = boottime = time;
416 	p->p_rtime.tv_sec = p->p_rtime.tv_usec = 0;
417 
418 	/* Initialize signal state for process 0. */
419 	siginit(p);
420 
421 	/* Create process 1 (init(8)). */
422 	if (fork1(p, 0, NULL, &p2))
423 		panic("fork init");
424 	cpu_set_kpc(p2, start_init, p2);
425 
426 	/* Create process 2, the pageout daemon kernel thread. */
427 	if (kthread_create(start_pagedaemon, NULL, NULL, "pagedaemon"))
428 		panic("fork pagedaemon");
429 
430 	/* Create process 3, the process reaper kernel thread. */
431 	if (kthread_create(start_reaper, NULL, NULL, "reaper"))
432 		panic("fork reaper");
433 
434 	/* Create any other deferred kernel threads. */
435 	kthread_run_deferred_queue();
436 
437 	/* The scheduler is an infinite loop. */
438 #if defined(UVM)
439 	uvm_scheduler();
440 #else
441 	scheduler();
442 #endif
443 	/* NOTREACHED */
444 }
445 
446 static void
447 check_console(p)
448 	struct proc *p;
449 {
450 	struct nameidata nd;
451 	int error;
452 
453 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_SYSSPACE, "/dev/console", p);
454 	error = namei(&nd);
455 	if (error == 0)
456 		vrele(nd.ni_vp);
457 	else if (error == ENOENT)
458 		printf("warning: no /dev/console\n");
459 	else
460 		printf("warning: lookup /dev/console: error %d\n", error);
461 }
462 
463 /*
464  * List of paths to try when searching for "init".
465  */
466 static char *initpaths[] = {
467 	"/sbin/init",
468 	"/sbin/oinit",
469 	"/sbin/init.bak",
470 	NULL,
471 };
472 
473 /*
474  * Start the initial user process; try exec'ing each pathname in "initpaths".
475  * The program is invoked with one argument containing the boot flags.
476  */
477 static void
478 start_init(arg)
479 	void *arg;
480 {
481 	struct proc *p = arg;
482 	vaddr_t addr;
483 	struct sys_execve_args /* {
484 		syscallarg(const char *) path;
485 		syscallarg(char * const *) argp;
486 		syscallarg(char * const *) envp;
487 	} */ args;
488 	int options, i, error;
489 	register_t retval[2];
490 	char flags[4], *flagsp;
491 	char **pathp, *path, *ucp, **uap, *arg0, *arg1 = NULL;
492 
493 	/*
494 	 * Now in process 1.
495 	 */
496 	initproc = p;
497 
498 	/*
499 	 * This is not the right way to do this.  We really should
500 	 * hand-craft a descriptor onto /dev/console to hand to init,
501 	 * but that's a _lot_ more work, and the benefit from this easy
502 	 * hack makes up for the "good is the enemy of the best" effect.
503 	 */
504 	check_console(p);
505 
506 	/*
507 	 * Need just enough stack to hold the faked-up "execve()" arguments.
508 	 */
509 	addr = USRSTACK - PAGE_SIZE;
510 #if defined(UVM)
511 	if (uvm_map(&p->p_vmspace->vm_map, &addr, PAGE_SIZE,
512                     NULL, UVM_UNKNOWN_OFFSET,
513                     UVM_MAPFLAG(UVM_PROT_ALL, UVM_PROT_ALL, UVM_INH_COPY,
514 		    UVM_ADV_NORMAL,
515                     UVM_FLAG_FIXED|UVM_FLAG_OVERLAY|UVM_FLAG_COPYONW))
516 		!= KERN_SUCCESS)
517 		panic("init: couldn't allocate argument space");
518 #else
519 	if (vm_allocate(&p->p_vmspace->vm_map, &addr, (vsize_t)PAGE_SIZE,
520 	    FALSE) != 0)
521 		panic("init: couldn't allocate argument space");
522 #endif
523 	p->p_vmspace->vm_maxsaddr = (caddr_t)addr;
524 
525 	for (pathp = &initpaths[0]; (path = *pathp) != NULL; pathp++) {
526 		ucp = (char *)(addr + PAGE_SIZE);
527 
528 		/*
529 		 * Construct the boot flag argument.
530 		 */
531 		flagsp = flags;
532 		*flagsp++ = '-';
533 		options = 0;
534 
535 		if (boothowto & RB_SINGLE) {
536 			*flagsp++ = 's';
537 			options = 1;
538 		}
539 #ifdef notyet
540 		if (boothowto & RB_FASTBOOT) {
541 			*flagsp++ = 'f';
542 			options = 1;
543 		}
544 #endif
545 
546 		/*
547 		 * Move out the flags (arg 1), if necessary.
548 		 */
549 		if (options != 0) {
550 			*flagsp++ = '\0';
551 			i = flagsp - flags;
552 #ifdef DEBUG
553 			printf("init: copying out flags `%s' %d\n", flags, i);
554 #endif
555 			(void)copyout((caddr_t)flags, (caddr_t)(ucp -= i), i);
556 			arg1 = ucp;
557 		}
558 
559 		/*
560 		 * Move out the file name (also arg 0).
561 		 */
562 		i = strlen(path) + 1;
563 #ifdef DEBUG
564 		printf("init: copying out path `%s' %d\n", path, i);
565 #endif
566 		(void)copyout((caddr_t)path, (caddr_t)(ucp -= i), i);
567 		arg0 = ucp;
568 
569 		/*
570 		 * Move out the arg pointers.
571 		 */
572 		uap = (char **)((long)ucp & ~ALIGNBYTES);
573 		(void)suword((caddr_t)--uap, 0);	/* terminator */
574 		if (options != 0)
575 			(void)suword((caddr_t)--uap, (long)arg1);
576 		(void)suword((caddr_t)--uap, (long)arg0);
577 
578 		/*
579 		 * Point at the arguments.
580 		 */
581 		SCARG(&args, path) = arg0;
582 		SCARG(&args, argp) = uap;
583 		SCARG(&args, envp) = NULL;
584 
585 		/*
586 		 * Now try to exec the program.  If can't for any reason
587 		 * other than it doesn't exist, complain.
588 		 */
589 		error = sys_execve(p, &args, retval);
590 		if (error == 0 || error == EJUSTRETURN)
591 			return;
592 		if (error != ENOENT)
593 			printf("exec %s: error %d\n", path, error);
594 	}
595 	printf("init: not found\n");
596 	panic("no init");
597 }
598 
599 /* ARGSUSED */
600 static void
601 start_pagedaemon(arg)
602 	void *arg;
603 {
604 
605 #if defined(UVM)
606 	uvm_pageout();
607 #else
608 	vm_pageout();
609 #endif
610 	/* NOTREACHED */
611 }
612 
613 /* ARGSUSED */
614 static void
615 start_reaper(arg)
616 	void *arg;
617 {
618 
619 	reaper();
620 	/* NOTREACHED */
621 }
622