xref: /freebsd/sys/kern/init_main.c (revision e627b39baccd1ec9129690167cf5e6d860509655)
1 /*
2  * Copyright (c) 1995 Terrence R. Lambert
3  * All rights reserved.
4  *
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  * $Id: init_main.c,v 1.48 1996/09/03 10:22:58 asami Exp $
43  */
44 
45 #include "opt_rlimit.h"
46 
47 #include <sys/param.h>
48 #include <sys/filedesc.h>
49 #include <sys/kernel.h>
50 #include <sys/sysctl.h>
51 #include <sys/proc.h>
52 #include <sys/resourcevar.h>
53 #include <sys/signalvar.h>
54 #include <sys/systm.h>
55 #include <sys/vnode.h>
56 #include <sys/sysent.h>
57 #include <sys/reboot.h>
58 #include <sys/sysproto.h>
59 #include <sys/vmmeter.h>
60 
61 #include <machine/cpu.h>
62 
63 #include <vm/vm.h>
64 #include <vm/vm_param.h>
65 #include <vm/vm_prot.h>
66 #include <vm/lock.h>
67 #include <vm/pmap.h>
68 #include <vm/vm_map.h>
69 #include <sys/user.h>
70 
71 extern struct linker_set	sysinit_set;	/* XXX */
72 
73 extern void __main __P((void));
74 extern void main __P((void *framep));
75 
76 /* Components of the first process -- never freed. */
77 static struct session session0;
78 static struct pgrp pgrp0;
79 struct	proc proc0;
80 static struct pcred cred0;
81 static struct filedesc0 filedesc0;
82 static struct plimit limit0;
83 static struct vmspace vmspace0;
84 struct	proc *curproc = &proc0;
85 struct	proc *initproc;
86 
87 int cmask = CMASK;
88 extern	struct user *proc0paddr;
89 
90 struct	vnode *rootvp;
91 int	boothowto;
92 
93 struct	timeval boottime;
94 SYSCTL_STRUCT(_kern, KERN_BOOTTIME, boottime,
95 	CTLFLAG_RW, &boottime, timeval, "");
96 
97 struct	timeval runtime;
98 
99 /*
100  * Promiscuous argument pass for start_init()
101  *
102  * This is a kludge because we use a return from main() rather than a call
103  * to a new reoutine in locore.s to kick the kernel alive from locore.s.
104  */
105 static void	*init_framep;
106 
107 
108 #if __GNUC__ >= 2
109 void __main() {}
110 #endif
111 
112 
113 /*
114  * This ensures that there is at least one entry so that the sysinit_set
115  * symbol is not undefined.  A sybsystem ID of SI_SUB_DUMMY is never
116  * executed.
117  */
118 SYSINIT(placeholder, SI_SUB_DUMMY,SI_ORDER_ANY, NULL, NULL)
119 
120 
121 /*
122  * System startup; initialize the world, create process 0, mount root
123  * filesystem, and fork to create init and pagedaemon.  Most of the
124  * hard work is done in the lower-level initialization routines including
125  * startup(), which does memory initialization and autoconfiguration.
126  *
127  * This allows simple addition of new kernel subsystems that require
128  * boot time initialization.  It also allows substitution of subsystem
129  * (for instance, a scheduler, kernel profiler, or VM system) by object
130  * module.  Finally, it allows for optional "kernel threads", like an LFS
131  * cleaner.
132  */
133 void
134 main(framep)
135 	void *framep;
136 {
137 
138 	register struct sysinit **sipp;		/* system initialization*/
139 	register struct sysinit **xipp;		/* interior loop of sort*/
140 	register struct sysinit *save;		/* bubble*/
141 	int			rval[2];	/* SI_TYPE_KTHREAD support*/
142 
143 	/*
144 	 * Save the locore.s frame pointer for start_init().
145 	 */
146 	init_framep = framep;
147 
148 	/*
149 	 * Perform a bubble sort of the system initialization objects by
150 	 * their subsystem (primary key) and order (secondary key).
151 	 *
152 	 * Since some things care about execution order, this is the
153 	 * operation which ensures continued function.
154 	 */
155 	for( sipp = (struct sysinit **)sysinit_set.ls_items; *sipp; sipp++) {
156 		for( xipp = sipp + 1; *xipp; xipp++) {
157 			if( (*sipp)->subsystem < (*xipp)->subsystem ||
158 			    ( (*sipp)->subsystem == (*xipp)->subsystem &&
159 			      (*sipp)->order < (*xipp)->order))
160 				continue;	/* skip*/
161 			save = *sipp;
162 			*sipp = *xipp;
163 			*xipp = save;
164 		}
165 	}
166 
167 	/*
168 	 * Traverse the (now) ordered list of system initialization tasks.
169 	 * Perform each task, and continue on to the next task.
170 	 *
171 	 * The last item on the list is expected to be the scheduler,
172 	 * which will not return.
173 	 */
174 	for( sipp = (struct sysinit **)sysinit_set.ls_items; *sipp; sipp++) {
175 		if( (*sipp)->subsystem == SI_SUB_DUMMY)
176 			continue;	/* skip dummy task(s)*/
177 
178 		switch( (*sipp)->type) {
179 		case SI_TYPE_DEFAULT:
180 			/* no special processing*/
181 			(*((*sipp)->func))( (*sipp)->udata);
182 			break;
183 
184 		case SI_TYPE_KTHREAD:
185 			/* kernel thread*/
186 			if (fork(&proc0, NULL, rval))
187 				panic("fork kernel process");
188 			if (rval[1]) {
189 				(*((*sipp)->func))( (*sipp)->udata);
190 				/*
191 				 * The call to start "init" returns
192 				 * here after the scheduler has been
193 				 * started, and returns to the caller
194 				 * in i386/i386/locore.s.  This is a
195 				 * necessary part of initialization
196 				 * and is rather non-obvious.
197 				 *
198 				 * No other "kernel threads" should
199 				 * return here.  Call panic() instead.
200 				 */
201 				return;
202 			}
203 			break;
204 
205 		default:
206 			panic( "init_main: unrecognized init type");
207 		}
208 	}
209 
210 	/* NOTREACHED*/
211 }
212 
213 
214 /*
215  * Start a kernel process.  This is called after a fork() call in
216  * main() in the file kern/init_main.c.
217  *
218  * This function is used to start "internal" daemons.
219  */
220 /* ARGSUSED*/
221 void
222 kproc_start(udata)
223 	void *udata;
224 {
225 	struct kproc_desc	*kp = udata;
226 	struct proc		*p = curproc;
227 
228 	/* save a global descriptor, if desired*/
229 	if( kp->global_procpp != NULL)
230 		*kp->global_procpp	= p;
231 
232 	/* this is a non-swapped system process*/
233 	p->p_flag |= P_INMEM | P_SYSTEM;
234 
235 	/* set up arg0 for 'ps', et al*/
236 	strcpy( p->p_comm, kp->arg0);
237 
238 	/* call the processes' main()...*/
239 	(*kp->func)();
240 
241 	/* NOTREACHED */
242 	panic("kproc_start: %s", kp->arg0);
243 }
244 
245 
246 /*
247  ***************************************************************************
248  ****
249  **** The following SYSINIT's belong elsewhere, but have not yet
250  **** been moved.
251  ****
252  ***************************************************************************
253  */
254 #ifdef OMIT
255 /*
256  * Handled by vfs_mountroot (bad idea) at this time... should be
257  * done the same as 4.4Lite2.
258  */
259 SYSINIT(swapinit, SI_SUB_SWAP, SI_ORDER_FIRST, swapinit, NULL)
260 #endif	/* OMIT*/
261 
262 /*
263  * Should get its own file...
264  */
265 #ifdef HPFPLIB
266 char	copyright[] =
267 "Copyright (c) 1982, 1986, 1989, 1991, 1993\n\tThe Regents of the University of California.\nCopyright (c) 1992 Hewlett-Packard Company\nCopyright (c) 1992 Motorola Inc.\nAll rights reserved.\n\n";
268 #else
269 char	copyright[] =
270 "Copyright (c) 1992-1996 FreeBSD Inc.\n"
271 #ifdef PC98
272 "Copyright (c) 1994-1996  FreeBSD(98) porting team.\n"
273 "Copyright (c) 1982, 1986, 1989, 1991, 1993\n\tThe Regents of the University of California.\n"
274 "Copyright (c) 1992  A.Kojima F.Ukai M.Ishii (KMC).\n"
275 "\tAll rights reserved.\n\n";
276 #else
277 "Copyright (c) 1982, 1986, 1989, 1991, 1993\n\tThe Regents of the University of California.  All rights reserved.\n\n";
278 #endif
279 #endif
280 static void print_caddr_t __P((void *data));
281 static void
282 print_caddr_t(data)
283 	void *data;
284 {
285 	printf("%s", (char *)data);
286 }
287 SYSINIT(announce, SI_SUB_COPYRIGHT, SI_ORDER_FIRST, print_caddr_t, copyright)
288 
289 
290 /*
291  ***************************************************************************
292  ****
293  **** The two following SYSINT's are proc0 specific glue code.  I am not
294  **** convinced that they can not be safely combined, but their order of
295  **** operation has been maintained as the same as the original init_main.c
296  **** for right now.
297  ****
298  **** These probably belong in init_proc.c or kern_proc.c, since they
299  **** deal with proc0 (the fork template process).
300  ****
301  ***************************************************************************
302  */
303 /* ARGSUSED*/
304 static void proc0_init __P((void *dummy));
305 static void
306 proc0_init(dummy)
307 	void *dummy;
308 {
309 	register struct proc		*p;
310 	register struct filedesc0	*fdp;
311 	register int i;
312 
313 	/*
314 	 * Initialize the current process pointer (curproc) before
315 	 * any possible traps/probes to simplify trap processing.
316 	 */
317 	p = &proc0;
318 	curproc = p;			/* XXX redundant*/
319 
320 	/*
321 	 * Initialize process and pgrp structures.
322 	 */
323 	procinit();
324 
325 	/*
326 	 * Initialize sleep queue hash table
327 	 */
328 	sleepinit();
329 
330 	/*
331 	 * Create process 0 (the swapper).
332 	 */
333 	LIST_INSERT_HEAD(&allproc, p, p_list);
334 	p->p_pgrp = &pgrp0;
335 	LIST_INSERT_HEAD(PGRPHASH(0), &pgrp0, pg_hash);
336 	LIST_INIT(&pgrp0.pg_members);
337 	LIST_INSERT_HEAD(&pgrp0.pg_members, p, p_pglist);
338 
339 	pgrp0.pg_session = &session0;
340 	session0.s_count = 1;
341 	session0.s_leader = p;
342 
343 	p->p_sysent = &aout_sysvec;
344 
345 	p->p_flag = P_INMEM | P_SYSTEM;
346 	p->p_stat = SRUN;
347 	p->p_nice = NZERO;
348 	p->p_rtprio.type = RTP_PRIO_NORMAL;
349 	p->p_rtprio.prio = 0;
350 
351 	bcopy("swapper", p->p_comm, sizeof ("swapper"));
352 
353 	/* Create credentials. */
354 	cred0.p_refcnt = 1;
355 	p->p_cred = &cred0;
356 	p->p_ucred = crget();
357 	p->p_ucred->cr_ngroups = 1;	/* group 0 */
358 
359 	/* Create the file descriptor table. */
360 	fdp = &filedesc0;
361 	p->p_fd = &fdp->fd_fd;
362 	fdp->fd_fd.fd_refcnt = 1;
363 	fdp->fd_fd.fd_cmask = cmask;
364 	fdp->fd_fd.fd_ofiles = fdp->fd_dfiles;
365 	fdp->fd_fd.fd_ofileflags = fdp->fd_dfileflags;
366 	fdp->fd_fd.fd_nfiles = NDFILE;
367 
368 	/* Create the limits structures. */
369 	p->p_limit = &limit0;
370 	for (i = 0; i < sizeof(p->p_rlimit)/sizeof(p->p_rlimit[0]); i++)
371 		limit0.pl_rlimit[i].rlim_cur =
372 		    limit0.pl_rlimit[i].rlim_max = RLIM_INFINITY;
373 	limit0.pl_rlimit[RLIMIT_NOFILE].rlim_cur = NOFILE;
374 	limit0.pl_rlimit[RLIMIT_NPROC].rlim_cur = MAXUPRC;
375 	i = ptoa(cnt.v_free_count);
376 	limit0.pl_rlimit[RLIMIT_RSS].rlim_max = i;
377 	limit0.pl_rlimit[RLIMIT_MEMLOCK].rlim_max = i;
378 	limit0.pl_rlimit[RLIMIT_MEMLOCK].rlim_cur = i / 3;
379 	limit0.p_refcnt = 1;
380 
381 	/* Allocate a prototype map so we have something to fork. */
382 	p->p_vmspace = &vmspace0;
383 	vmspace0.vm_refcnt = 1;
384 	pmap_pinit(&vmspace0.vm_pmap);
385 	vm_map_init(&vmspace0.vm_map, round_page(VM_MIN_ADDRESS),
386 	    trunc_page(VM_MAXUSER_ADDRESS), TRUE);
387 	vmspace0.vm_map.pmap = &vmspace0.vm_pmap;
388 	p->p_addr = proc0paddr;				/* XXX */
389 
390 #define INCOMPAT_LITES2
391 #ifdef INCOMPAT_LITES2
392 	/*
393 	 * proc0 needs to have a coherent frame base, too.
394 	 * This probably makes the identical call for the init proc
395 	 * that happens later unnecessary since it should inherit
396 	 * it during the fork.
397 	 */
398 	cpu_set_init_frame(p, init_framep);			/* XXX! */
399 #endif	/* INCOMPAT_LITES2*/
400 
401 	/*
402 	 * We continue to place resource usage info and signal
403 	 * actions in the user struct so they're pageable.
404 	 */
405 	p->p_stats = &p->p_addr->u_stats;
406 	p->p_sigacts = &p->p_addr->u_sigacts;
407 
408 	/*
409 	 * Charge root for one process.
410 	 */
411 	(void)chgproccnt(0, 1);
412 }
413 SYSINIT(p0init, SI_SUB_INTRINSIC, SI_ORDER_FIRST, proc0_init, NULL)
414 
415 /* ARGSUSED*/
416 static void proc0_post __P((void *dummy));
417 static void
418 proc0_post(dummy)
419 	void *dummy;
420 {
421 	struct timeval tv;
422 
423 	/*
424 	 * Now can look at time, having had a chance to verify the time
425 	 * from the file system.  Reset p->p_rtime as it may have been
426 	 * munched in mi_switch() after the time got set.
427 	 */
428 	proc0.p_stats->p_start = runtime = mono_time = boottime = time;
429 	proc0.p_rtime.tv_sec = proc0.p_rtime.tv_usec = 0;
430 
431 	/*
432 	 * Give the ``random'' number generator a thump.
433 	 */
434 	microtime(&tv);
435 	srandom(tv.tv_sec ^ tv.tv_usec);
436 
437 	/* Initialize signal state for process 0. */
438 	siginit(&proc0);
439 }
440 SYSINIT(p0post, SI_SUB_INTRINSIC_POST, SI_ORDER_FIRST, proc0_post, NULL)
441 
442 
443 
444 
445 /*
446  ***************************************************************************
447  ****
448  **** The following SYSINIT's and glue code should be moved to the
449  **** respective files on a per subsystem basis.
450  ****
451  ***************************************************************************
452  */
453 /* ARGSUSED*/
454 static void sched_setup __P((void *dummy));
455 static void
456 sched_setup(dummy)
457 	void *dummy;
458 {
459 	/* Kick off timeout driven events by calling first time. */
460 	roundrobin(NULL);
461 	schedcpu(NULL);
462 }
463 SYSINIT(sched_setup, SI_SUB_KICK_SCHEDULER, SI_ORDER_FIRST, sched_setup, NULL)
464 
465 /* ARGSUSED*/
466 static void xxx_vfs_mountroot __P((void *dummy));
467 static void
468 xxx_vfs_mountroot(dummy)
469 	void *dummy;
470 {
471 	/* Mount the root file system. */
472 	if ((*mountroot)(mountrootvfsops))
473 		panic("cannot mount root");
474 }
475 SYSINIT(mountroot, SI_SUB_ROOT, SI_ORDER_FIRST, xxx_vfs_mountroot, NULL)
476 
477 /* ARGSUSED*/
478 static void xxx_vfs_root_fdtab __P((void *dummy));
479 static void
480 xxx_vfs_root_fdtab(dummy)
481 	void *dummy;
482 {
483 	register struct filedesc0	*fdp = &filedesc0;
484 
485 	/* Get the vnode for '/'.  Set fdp->fd_fd.fd_cdir to reference it. */
486 	if (VFS_ROOT(mountlist.cqh_first, &rootvnode))
487 		panic("cannot find root vnode");
488 	fdp->fd_fd.fd_cdir = rootvnode;
489 	VREF(fdp->fd_fd.fd_cdir);
490 	VOP_UNLOCK(rootvnode);
491 	fdp->fd_fd.fd_rdir = NULL;
492 }
493 SYSINIT(retrofit, SI_SUB_ROOT_FDTAB, SI_ORDER_FIRST, xxx_vfs_root_fdtab, NULL)
494 
495 
496 /*
497  ***************************************************************************
498  ****
499  **** The following code probably belongs in another file, like
500  **** kern/init_init.c.  It is here for two reasons only:
501  ****
502  ****	1)	This code returns to startup the system; this is
503  ****		abnormal for a kernel thread.
504  ****	2)	This code promiscuously uses init_frame
505  ****
506  ***************************************************************************
507  */
508 
509 static void kthread_init __P((void *dummy));
510 SYSINIT_KT(init,SI_SUB_KTHREAD_INIT, SI_ORDER_FIRST, kthread_init, NULL)
511 
512 
513 static void start_init __P((struct proc *p, void *framep));
514 
515 /* ARGSUSED*/
516 static void
517 kthread_init(dummy)
518 	void *dummy;
519 {
520 
521 	/* Create process 1 (init(8)). */
522 	start_init(curproc, init_framep);
523 
524 	/*
525 	 * This is the only kernel thread allowed to return yo the
526 	 * caller!!!
527 	 */
528 	return;
529 }
530 
531 
532 /*
533  * List of paths to try when searching for "init".
534  */
535 static char *initpaths[] = {
536 	"/sbin/init",
537 	"/sbin/oinit",
538 	"/sbin/init.bak",
539 	"/stand/sysinstall",
540 	NULL,
541 };
542 
543 /*
544  * Start the initial user process; try exec'ing each pathname in "initpaths".
545  * The program is invoked with one argument containing the boot flags.
546  */
547 static void
548 start_init(p, framep)
549 	struct proc *p;
550 	void *framep;
551 {
552 	vm_offset_t addr;
553 	struct execve_args args;
554 	int options, i, retval[2], error;
555 	char **pathp, *path, *ucp, **uap, *arg0, *arg1;
556 
557 	initproc = p;
558 
559 	/*
560 	 * We need to set the system call frame as if we were entered through
561 	 * a syscall() so that when we call execve() below, it will be able
562 	 * to set the entry point (see setregs) when it tries to exec.  The
563 	 * startup code in "locore.s" has allocated space for the frame and
564 	 * passed a pointer to that space as main's argument.
565 	 */
566 	cpu_set_init_frame(p, framep);
567 
568 	/*
569 	 * Need just enough stack to hold the faked-up "execve()" arguments.
570 	 */
571 	addr = trunc_page(VM_MAXUSER_ADDRESS - PAGE_SIZE);
572 	if (vm_map_find(&p->p_vmspace->vm_map, NULL, 0, &addr, PAGE_SIZE, FALSE, VM_PROT_ALL, VM_PROT_ALL, 0) != 0)
573 		panic("init: couldn't allocate argument space");
574 	p->p_vmspace->vm_maxsaddr = (caddr_t)addr;
575 	p->p_vmspace->vm_ssize = 1;
576 
577 	for (pathp = &initpaths[0]; (path = *pathp) != NULL; pathp++) {
578 		/*
579 		 * Move out the boot flag argument.
580 		 */
581 		options = 0;
582 		ucp = (char *)USRSTACK;
583 		(void)subyte(--ucp, 0);		/* trailing zero */
584 		if (boothowto & RB_SINGLE) {
585 			(void)subyte(--ucp, 's');
586 			options = 1;
587 		}
588 #ifdef notyet
589                 if (boothowto & RB_FASTBOOT) {
590 			(void)subyte(--ucp, 'f');
591 			options = 1;
592 		}
593 #endif
594 
595 #ifdef BOOTCDROM
596 		(void)subyte(--ucp, 'C');
597 		options = 1;
598 #endif
599 		if (options == 0)
600 			(void)subyte(--ucp, '-');
601 		(void)subyte(--ucp, '-');		/* leading hyphen */
602 		arg1 = ucp;
603 
604 		/*
605 		 * Move out the file name (also arg 0).
606 		 */
607 		for (i = strlen(path) + 1; i >= 0; i--)
608 			(void)subyte(--ucp, path[i]);
609 		arg0 = ucp;
610 
611 		/*
612 		 * Move out the arg pointers.
613 		 */
614 		uap = (char **)((int)ucp & ~(NBPW-1));
615 		(void)suword((caddr_t)--uap, 0);	/* terminator */
616 		(void)suword((caddr_t)--uap, (int)arg1);
617 		(void)suword((caddr_t)--uap, (int)arg0);
618 
619 		/*
620 		 * Point at the arguments.
621 		 */
622 		args.fname = arg0;
623 		args.argv = uap;
624 		args.envv = NULL;
625 
626 		/*
627 		 * Now try to exec the program.  If can't for any reason
628 		 * other than it doesn't exist, complain.
629 		 *
630 		 * Otherwise return to main() which returns to btext
631 		 * which completes the system startup.
632 		 */
633 		if ((error = execve(p, &args, &retval[0])) == 0)
634 			return;
635 		if (error != ENOENT)
636 			printf("exec %s: error %d\n", path, error);
637 	}
638 	printf("init: not found\n");
639 	panic("no init");
640 }
641