xref: /freebsd/sys/kern/init_main.c (revision 8af5d536ffbd13a97c1332dc307ebf3a8a64c6f4)
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 1.28 1995/08/29 23:59:22 bde Exp $
43  */
44 
45 #include <sys/param.h>
46 #include <sys/filedesc.h>
47 #include <sys/errno.h>
48 #include <sys/exec.h>
49 #include <sys/kernel.h>
50 #ifdef GPROF
51 #include <sys/gmon.h>
52 #endif
53 #include <sys/mount.h>
54 #include <sys/proc.h>
55 #include <sys/resourcevar.h>
56 #include <sys/signalvar.h>
57 #include <sys/systm.h>
58 #include <sys/vnode.h>
59 #include <sys/sysent.h>
60 #include <sys/conf.h>
61 #include <sys/buf.h>
62 #include <sys/clist.h>
63 #include <sys/device.h>
64 #include <sys/msg.h>
65 #include <sys/protosw.h>
66 #include <sys/reboot.h>
67 #include <sys/sem.h>
68 #include <sys/shm.h>
69 #include <sys/user.h>
70 
71 #include <ufs/ufs/quota.h>
72 
73 #include <machine/cpu.h>
74 
75 #include <vm/vm.h>
76 #include <vm/vm_pageout.h>
77 
78 extern struct linker_set	sysinit_set;	/* XXX */
79 
80 extern void __main __P((void));
81 extern void main __P((void *framep));
82 
83 /* Components of the first process -- never freed. */
84 struct	session session0;
85 struct	pgrp pgrp0;
86 struct	proc proc0;
87 struct	pcred cred0;
88 struct	filedesc0 filedesc0;
89 struct	plimit limit0;
90 struct	vmspace vmspace0;
91 struct	proc *curproc = &proc0;
92 struct	proc *initproc;
93 
94 int	cmask = CMASK;
95 extern	struct user *proc0paddr;
96 
97 struct	vnode *rootvp;
98 int	boothowto;
99 struct	timeval boottime;
100 struct	timeval runtime;
101 
102 /*
103  * Promiscuous argument pass for start_init()
104  *
105  * This is a kludge because we use a return from main() rather than a call
106  * to a new reoutine in locore.s to kick the kernel alive from locore.s.
107  */
108 static void	*init_framep;
109 
110 
111 #if __GNUC__ >= 2
112 void __main() {}
113 #endif
114 
115 
116 /*
117  * This ensures that there is at least one entry so that the sysinit_set
118  * symbol is not undefined.  A sybsystem ID of SI_SUB_DUMMY is never
119  * executed.
120  */
121 SYSINIT(placeholder, SI_SUB_DUMMY,SI_ORDER_ANY, NULL, NULL)
122 
123 
124 /*
125  * System startup; initialize the world, create process 0, mount root
126  * filesystem, and fork to create init and pagedaemon.  Most of the
127  * hard work is done in the lower-level initialization routines including
128  * startup(), which does memory initialization and autoconfiguration.
129  *
130  * This allows simple addition of new kernel subsystems that require
131  * boot time initialization.  It also allows substitution of subsystem
132  * (for instance, a scheduler, kernel profiler, or VM system) by object
133  * module.  Finally, it allows for optional "kernel threads", like an LFS
134  * cleaner.
135  */
136 void
137 main(framep)
138 	void *framep;
139 {
140 
141 	register struct sysinit **sipp;		/* system initialization*/
142 	register struct sysinit **xipp;		/* interior loop of sort*/
143 	register struct sysinit *save;		/* bubble*/
144 	int			rval[2];	/* SI_TYPE_KTHREAD support*/
145 
146 	/*
147 	 * Save the locore.s frame pointer for start_init().
148 	 */
149 	init_framep = framep;
150 
151 	/*
152 	 * Perform a bubble sort of the system initialization objects by
153 	 * their subsystem (primary key) and order (secondary key).
154 	 *
155 	 * Since some things care about execution order, this is the
156 	 * operation which ensures continued function.
157 	 */
158 	for( sipp = (struct sysinit **)sysinit_set.ls_items; *sipp; sipp++) {
159 		for( xipp = sipp + 1; *xipp; xipp++) {
160 			if( (*sipp)->subsystem < (*xipp)->subsystem ||
161 			    ( (*sipp)->subsystem == (*xipp)->subsystem &&
162 			      (*sipp)->order < (*xipp)->order))
163 				continue;	/* skip*/
164 			save = *sipp;
165 			*sipp = *xipp;
166 			*xipp = save;
167 		}
168 	}
169 
170 	/*
171 	 * Traverse the (now) ordered list of system initialization tasks.
172 	 * Perform each task, and continue on to the next task.
173 	 *
174 	 * The last item on the list is expected to be the scheduler,
175 	 * which will not return.
176 	 */
177 	for( sipp = (struct sysinit **)sysinit_set.ls_items; *sipp; sipp++) {
178 		if( (*sipp)->subsystem == SI_SUB_DUMMY)
179 			continue;	/* skip dummy task(s)*/
180 
181 		switch( (*sipp)->type) {
182 		case SI_TYPE_DEFAULT:
183 			/* no special processing*/
184 			(*((*sipp)->func))( (*sipp)->udata);
185 			break;
186 
187 		case SI_TYPE_KTHREAD:
188 			/* kernel thread*/
189 			if (fork(&proc0, NULL, rval))
190 				panic("fork kernel process");
191 			if (rval[1]) {
192 				(*((*sipp)->func))( (*sipp)->udata);
193 				/*
194 				 * The call to start "init" returns
195 				 * here after the scheduler has been
196 				 * started, and returns to the caller
197 				 * in i386/i386/locore.s.  This is a
198 				 * necessary part of initialization
199 				 * and is rather non-obvious.
200 				 *
201 				 * No other "kernel threads" should
202 				 * return here.  Call panic() instead.
203 				 */
204 				return;
205 			}
206 			break;
207 
208 		default:
209 			panic( "init_main: unrecognized init type");
210 		}
211 	}
212 
213 	/* NOTREACHED*/
214 }
215 
216 
217 /*
218  * Start a kernel process.  This is called after a fork() call in
219  * main() in the file kern/init_main.c.
220  *
221  * This function is used to start "internal" daemons.
222  */
223 /* ARGSUSED*/
224 void
225 kproc_start( udata)
226 caddr_t		udata;	/* pointer to a 'kproc_desc' ? */
227 {
228 	struct kproc_desc	*kp = (struct kproc_desc *)udata;
229 	struct proc		*p = curproc;
230 
231 	/* save a global descriptor, if desired*/
232 	if( kp->global_procpp != NULL)
233 		*kp->global_procpp	= p;
234 
235 	/* this is a non-swapped system process*/
236 	p->p_flag |= P_INMEM | P_SYSTEM;
237 
238 	/* set up arg0 for 'ps', et al*/
239 	strcpy( p->p_comm, kp->arg0);
240 
241 	/* call the processes' main()...*/
242 	(*kp->func)();
243 
244 	/* NOTREACHED */
245 	panic( "kproc_start: %s", kp->arg0);
246 }
247 
248 
249 /*
250  ***************************************************************************
251  ****
252  **** The following SYSINIT's belong elsewhere, but have not yet
253  **** been moved.
254  ****
255  ***************************************************************************
256  */
257 #ifdef OMIT
258 /*
259  * Handled by vfs_mountroot (bad idea) at this time... should be
260  * done the same as 4.4Lite2.
261  */
262 SYSINIT(swapinit, SI_SUB_SWAP, SI_ORDER_FIRST, swapinit, NULL)
263 #endif	/* OMIT*/
264 
265 /*
266  * Should get its own file...
267  */
268 #ifdef HPFPLIB
269 char	copyright[] =
270 "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";
271 #else
272 char	copyright[] =
273 "Copyright (c) 1982, 1986, 1989, 1991, 1993\n\tThe Regents of the University of California.  All rights reserved.\n\n";
274 #endif
275 static void print_caddr_t __P((caddr_t data));
276 static void
277 print_caddr_t(data)
278 	caddr_t data;
279 {
280 	printf("%s", (char *)data);
281 }
282 SYSINIT(announce, SI_SUB_COPYRIGHT, SI_ORDER_FIRST, print_caddr_t,
283 	(caddr_t)copyright)
284 
285 
286 /*
287  ***************************************************************************
288  ****
289  **** The two following SYSINT's are proc0 specific glue code.  I am not
290  **** convinced that they can not be safely combined, but their order of
291  **** operation has been maintained as the same as the original init_main.c
292  **** for right now.
293  ****
294  **** These probably belong in init_proc.c or kern_proc.c, since they
295  **** deal with proc0 (the fork template process).
296  ****
297  ***************************************************************************
298  */
299 /* ARGSUSED*/
300 void proc0_init __P((caddr_t udata));
301 void
302 proc0_init( udata)
303 caddr_t		udata;		/* not used*/
304 {
305 	register struct proc		*p;
306 	register struct filedesc0	*fdp;
307 	register int i;
308 
309 	/*
310 	 * Initialize the current process pointer (curproc) before
311 	 * any possible traps/probes to simplify trap processing.
312 	 */
313 	p = &proc0;
314 	curproc = p;			/* XXX redundant*/
315 
316 	/*
317 	 * Create process 0 (the swapper).
318 	 */
319 	allproc = (volatile struct proc *)p;
320 	p->p_prev = (struct proc **)&allproc;
321 	p->p_pgrp = &pgrp0;
322 	pgrphash[0] = &pgrp0;
323 	pgrp0.pg_mem = p;
324 	pgrp0.pg_session = &session0;
325 	session0.s_count = 1;
326 	session0.s_leader = p;
327 
328 	p->p_sysent = &aout_sysvec;
329 
330 	p->p_flag = P_INMEM | P_SYSTEM;
331 	p->p_stat = SRUN;
332 	p->p_nice = NZERO;
333 	p->p_rtprio.type = RTP_PRIO_NORMAL;
334 	p->p_rtprio.prio = 0;
335 
336 	bcopy("swapper", p->p_comm, sizeof ("swapper"));
337 
338 	/* Create credentials. */
339 	cred0.p_refcnt = 1;
340 	p->p_cred = &cred0;
341 	p->p_ucred = crget();
342 	p->p_ucred->cr_ngroups = 1;	/* group 0 */
343 
344 	/* Create the file descriptor table. */
345 	fdp = &filedesc0;
346 	p->p_fd = &fdp->fd_fd;
347 	fdp->fd_fd.fd_refcnt = 1;
348 	fdp->fd_fd.fd_cmask = cmask;
349 	fdp->fd_fd.fd_ofiles = fdp->fd_dfiles;
350 	fdp->fd_fd.fd_ofileflags = fdp->fd_dfileflags;
351 	fdp->fd_fd.fd_nfiles = NDFILE;
352 
353 	/* Create the limits structures. */
354 	p->p_limit = &limit0;
355 	for (i = 0; i < sizeof(p->p_rlimit)/sizeof(p->p_rlimit[0]); i++)
356 		limit0.pl_rlimit[i].rlim_cur =
357 		    limit0.pl_rlimit[i].rlim_max = RLIM_INFINITY;
358 	limit0.pl_rlimit[RLIMIT_NOFILE].rlim_cur = NOFILE;
359 	limit0.pl_rlimit[RLIMIT_NPROC].rlim_cur = MAXUPRC;
360 	i = ptoa(cnt.v_free_count);
361 	limit0.pl_rlimit[RLIMIT_RSS].rlim_max = i;
362 	limit0.pl_rlimit[RLIMIT_MEMLOCK].rlim_max = i;
363 	limit0.pl_rlimit[RLIMIT_MEMLOCK].rlim_cur = i / 3;
364 	limit0.p_refcnt = 1;
365 
366 	/* Allocate a prototype map so we have something to fork. */
367 	p->p_vmspace = &vmspace0;
368 	vmspace0.vm_refcnt = 1;
369 	pmap_pinit(&vmspace0.vm_pmap);
370 	vm_map_init(&vmspace0.vm_map, round_page(VM_MIN_ADDRESS),
371 	    trunc_page(VM_MAX_ADDRESS), TRUE);
372 	vmspace0.vm_map.pmap = &vmspace0.vm_pmap;
373 	p->p_addr = proc0paddr;				/* XXX */
374 
375 #define INCOMPAT_LITES2
376 #ifdef INCOMPAT_LITES2
377 	/*
378 	 * proc0 needs to have a coherent frame base, too.
379 	 * This probably makes the identical call for the init proc
380 	 * that happens later unnecessary since it should inherit
381 	 * it during the fork.
382 	 */
383 	cpu_set_init_frame(p, init_framep);			/* XXX! */
384 #endif	/* INCOMPAT_LITES2*/
385 
386 	/*
387 	 * We continue to place resource usage info and signal
388 	 * actions in the user struct so they're pageable.
389 	 */
390 	p->p_stats = &p->p_addr->u_stats;
391 	p->p_sigacts = &p->p_addr->u_sigacts;
392 
393 	/*
394 	 * Initialize per uid information structure and charge
395 	 * root for one process.
396 	 */
397 	usrinfoinit();
398 	(void)chgproccnt(0, 1);
399 }
400 SYSINIT(p0init, SI_SUB_INTRINSIC, SI_ORDER_FIRST, proc0_init, NULL)
401 
402 /* ARGSUSED*/
403 void proc0_post __P((caddr_t udata));
404 void
405 proc0_post( udata)
406 caddr_t		udata;		/* not used*/
407 {
408 	/*
409 	 * Now can look at time, having had a chance to verify the time
410 	 * from the file system.  Reset p->p_rtime as it may have been
411 	 * munched in mi_switch() after the time got set.
412 	 */
413 	proc0.p_stats->p_start = runtime = mono_time = boottime = time;
414 	proc0.p_rtime.tv_sec = proc0.p_rtime.tv_usec = 0;
415 
416 	/* Initialize signal state for process 0. */
417 	siginit(&proc0);
418 }
419 SYSINIT(p0post, SI_SUB_INTRINSIC_POST, SI_ORDER_FIRST, proc0_post, NULL)
420 
421 
422 
423 
424 /*
425  ***************************************************************************
426  ****
427  **** The following SYSINIT's and glue code should be moved to the
428  **** respective files on a per subsystem basis.
429  ****
430  ***************************************************************************
431  */
432 /* ARGSUSED*/
433 void sched_setup __P((caddr_t udata));
434 void
435 sched_setup( udata)
436 caddr_t		udata;		/* not used*/
437 {
438 	/* Kick off timeout driven events by calling first time. */
439 	roundrobin(NULL);
440 	schedcpu(NULL);
441 }
442 SYSINIT(sched_setup, SI_SUB_KICK_SCHEDULER, SI_ORDER_FIRST, sched_setup, NULL)
443 
444 /* ARGSUSED*/
445 void xxx_vfs_mountroot __P((caddr_t udata));
446 void
447 xxx_vfs_mountroot( udata)
448 caddr_t		udata;		/* not used*/
449 {
450 	/* Mount the root file system. */
451 	if ((*mountroot)( (caddr_t)mountrootvfsops))
452 		panic("cannot mount root");
453 }
454 SYSINIT(mountroot, SI_SUB_ROOT, SI_ORDER_FIRST, xxx_vfs_mountroot, NULL)
455 
456 /* ARGSUSED*/
457 void xxx_vfs_root_fdtab __P((caddr_t udata));
458 void
459 xxx_vfs_root_fdtab( udata)
460 caddr_t		udata;		/* not used*/
461 {
462 	register struct filedesc0	*fdp = &filedesc0;
463 
464 	/* Get the vnode for '/'.  Set fdp->fd_fd.fd_cdir to reference it. */
465 	if (VFS_ROOT(mountlist.cqh_first, &rootvnode))
466 		panic("cannot find root vnode");
467 	fdp->fd_fd.fd_cdir = rootvnode;
468 	VREF(fdp->fd_fd.fd_cdir);
469 	VOP_UNLOCK(rootvnode);
470 	fdp->fd_fd.fd_rdir = NULL;
471 }
472 SYSINIT(retrofit, SI_SUB_ROOT_FDTAB, SI_ORDER_FIRST, xxx_vfs_root_fdtab, NULL)
473 
474 
475 /*
476  ***************************************************************************
477  ****
478  **** The following code probably belongs in another file, like
479  **** kern/init_init.c.  It is here for two reasons only:
480  ****
481  ****	1)	This code returns to startup the system; this is
482  ****		abnormal for a kernel thread.
483  ****	2)	This code promiscuously uses init_frame
484  ****
485  ***************************************************************************
486  */
487 
488 static void kthread_init __P(( caddr_t udata));
489 SYSINIT_KT(init,SI_SUB_KTHREAD_INIT, SI_ORDER_FIRST, kthread_init, NULL)
490 
491 
492 static void start_init __P((struct proc *p, void *framep));
493 
494 /* ARGSUSED*/
495 static void
496 kthread_init( udata)
497 caddr_t		udata;		/* not used*/
498 {
499 
500 	/* Create process 1 (init(8)). */
501 	start_init(curproc, init_framep);
502 
503 	/*
504 	 * This is the only kernel thread allowed to return yo the
505 	 * caller!!!
506 	 */
507 	return;
508 }
509 
510 
511 /*
512  * List of paths to try when searching for "init".
513  */
514 static char *initpaths[] = {
515 	"/sbin/init",
516 	"/sbin/oinit",
517 	"/sbin/init.bak",
518 	"/stand/sysinstall",
519 	NULL,
520 };
521 
522 /*
523  * Start the initial user process; try exec'ing each pathname in "initpaths".
524  * The program is invoked with one argument containing the boot flags.
525  */
526 static void
527 start_init(p, framep)
528 	struct proc *p;
529 	void *framep;
530 {
531 	vm_offset_t addr;
532 	struct execve_args args;
533 	int options, i, retval[2], error;
534 	char **pathp, *path, *ucp, **uap, *arg0, *arg1;
535 
536 	initproc = p;
537 
538 	/*
539 	 * We need to set the system call frame as if we were entered through
540 	 * a syscall() so that when we call execve() below, it will be able
541 	 * to set the entry point (see setregs) when it tries to exec.  The
542 	 * startup code in "locore.s" has allocated space for the frame and
543 	 * passed a pointer to that space as main's argument.
544 	 */
545 	cpu_set_init_frame(p, framep);
546 
547 	/*
548 	 * Need just enough stack to hold the faked-up "execve()" arguments.
549 	 */
550 	addr = trunc_page(VM_MAXUSER_ADDRESS - PAGE_SIZE);
551 	if (vm_map_find(&p->p_vmspace->vm_map, NULL, 0, &addr, PAGE_SIZE, FALSE) != 0)
552 		panic("init: couldn't allocate argument space");
553 	p->p_vmspace->vm_maxsaddr = (caddr_t)addr;
554 	p->p_vmspace->vm_ssize = 1;
555 
556 	for (pathp = &initpaths[0]; (path = *pathp) != NULL; pathp++) {
557 		/*
558 		 * Move out the boot flag argument.
559 		 */
560 		options = 0;
561 		ucp = (char *)USRSTACK;
562 		(void)subyte(--ucp, 0);		/* trailing zero */
563 		if (boothowto & RB_SINGLE) {
564 			(void)subyte(--ucp, 's');
565 			options = 1;
566 		}
567 #ifdef notyet
568                 if (boothowto & RB_FASTBOOT) {
569 			(void)subyte(--ucp, 'f');
570 			options = 1;
571 		}
572 #endif
573 
574 #ifdef BOOTCDROM
575 		(void)subyte(--ucp, 'C');
576 		options = 1;
577 #endif
578 		if (options == 0)
579 			(void)subyte(--ucp, '-');
580 		(void)subyte(--ucp, '-');		/* leading hyphen */
581 		arg1 = ucp;
582 
583 		/*
584 		 * Move out the file name (also arg 0).
585 		 */
586 		for (i = strlen(path) + 1; i >= 0; i--)
587 			(void)subyte(--ucp, path[i]);
588 		arg0 = ucp;
589 
590 		/*
591 		 * Move out the arg pointers.
592 		 */
593 		uap = (char **)((int)ucp & ~(NBPW-1));
594 		(void)suword((caddr_t)--uap, 0);	/* terminator */
595 		(void)suword((caddr_t)--uap, (int)arg1);
596 		(void)suword((caddr_t)--uap, (int)arg0);
597 
598 		/*
599 		 * Point at the arguments.
600 		 */
601 		args.fname = arg0;
602 		args.argv = uap;
603 		args.envv = NULL;
604 
605 		/*
606 		 * Now try to exec the program.  If can't for any reason
607 		 * other than it doesn't exist, complain.
608 		 *
609 		 * Otherwise return to main() which returns to btext
610 		 * which completes the system startup.
611 		 */
612 		if ((error = execve(p, &args, &retval[0])) == 0)
613 			return;
614 		if (error != ENOENT)
615 			printf("exec %s: error %d\n", path, error);
616 	}
617 	printf("init: not found\n");
618 	panic("no init");
619 }
620