xref: /freebsd/sys/kern/init_main.c (revision 05c7a37afb48ddd5ee1bd921a5d46fe59cc70b15)
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 1.39 1996/03/02 18:24:05 peter Exp $
43  */
44 
45 #include "opt_rlimit.h"
46 
47 #include <sys/param.h>
48 #include <sys/filedesc.h>
49 #include <sys/errno.h>
50 #include <sys/exec.h>
51 #include <sys/kernel.h>
52 #include <sys/sysctl.h>
53 #ifdef GPROF
54 #include <sys/gmon.h>
55 #endif
56 #include <sys/mount.h>
57 #include <sys/proc.h>
58 #include <sys/resourcevar.h>
59 #include <sys/signalvar.h>
60 #include <sys/systm.h>
61 #include <sys/vnode.h>
62 #include <sys/sysent.h>
63 #include <sys/conf.h>
64 #include <sys/buf.h>
65 #include <sys/clist.h>
66 #include <sys/msg.h>
67 #include <sys/protosw.h>
68 #include <sys/reboot.h>
69 #include <sys/sem.h>
70 #include <sys/shm.h>
71 #include <sys/sysproto.h>
72 #include <sys/vmmeter.h>
73 
74 #include <ufs/ufs/quota.h>
75 
76 #include <machine/cpu.h>
77 
78 #include <vm/vm.h>
79 #include <vm/vm_param.h>
80 #include <vm/vm_prot.h>
81 #include <vm/lock.h>
82 #include <vm/pmap.h>
83 #include <vm/vm_map.h>
84 #include <vm/vm_extern.h>
85 #include <sys/user.h>
86 
87 extern struct linker_set	sysinit_set;	/* XXX */
88 
89 extern void __main __P((void));
90 extern void main __P((void *framep));
91 
92 /* Components of the first process -- never freed. */
93 static struct session session0;
94 static struct pgrp pgrp0;
95 struct	proc proc0;
96 static struct pcred cred0;
97 static struct filedesc0 filedesc0;
98 static struct plimit limit0;
99 static struct vmspace vmspace0;
100 struct	proc *curproc = &proc0;
101 struct	proc *initproc;
102 
103 int cmask = CMASK;
104 extern	struct user *proc0paddr;
105 
106 struct	vnode *rootvp;
107 int	boothowto;
108 
109 struct	timeval boottime;
110 SYSCTL_STRUCT(_kern, KERN_BOOTTIME, boottime,
111 	CTLFLAG_RW, &boottime, timeval, "");
112 
113 struct	timeval runtime;
114 
115 /*
116  * Promiscuous argument pass for start_init()
117  *
118  * This is a kludge because we use a return from main() rather than a call
119  * to a new reoutine in locore.s to kick the kernel alive from locore.s.
120  */
121 static void	*init_framep;
122 
123 
124 #if __GNUC__ >= 2
125 void __main() {}
126 #endif
127 
128 
129 /*
130  * This ensures that there is at least one entry so that the sysinit_set
131  * symbol is not undefined.  A sybsystem ID of SI_SUB_DUMMY is never
132  * executed.
133  */
134 SYSINIT(placeholder, SI_SUB_DUMMY,SI_ORDER_ANY, NULL, NULL)
135 
136 
137 /*
138  * System startup; initialize the world, create process 0, mount root
139  * filesystem, and fork to create init and pagedaemon.  Most of the
140  * hard work is done in the lower-level initialization routines including
141  * startup(), which does memory initialization and autoconfiguration.
142  *
143  * This allows simple addition of new kernel subsystems that require
144  * boot time initialization.  It also allows substitution of subsystem
145  * (for instance, a scheduler, kernel profiler, or VM system) by object
146  * module.  Finally, it allows for optional "kernel threads", like an LFS
147  * cleaner.
148  */
149 void
150 main(framep)
151 	void *framep;
152 {
153 
154 	register struct sysinit **sipp;		/* system initialization*/
155 	register struct sysinit **xipp;		/* interior loop of sort*/
156 	register struct sysinit *save;		/* bubble*/
157 	int			rval[2];	/* SI_TYPE_KTHREAD support*/
158 
159 	/*
160 	 * Save the locore.s frame pointer for start_init().
161 	 */
162 	init_framep = framep;
163 
164 	/*
165 	 * Perform a bubble sort of the system initialization objects by
166 	 * their subsystem (primary key) and order (secondary key).
167 	 *
168 	 * Since some things care about execution order, this is the
169 	 * operation which ensures continued function.
170 	 */
171 	for( sipp = (struct sysinit **)sysinit_set.ls_items; *sipp; sipp++) {
172 		for( xipp = sipp + 1; *xipp; xipp++) {
173 			if( (*sipp)->subsystem < (*xipp)->subsystem ||
174 			    ( (*sipp)->subsystem == (*xipp)->subsystem &&
175 			      (*sipp)->order < (*xipp)->order))
176 				continue;	/* skip*/
177 			save = *sipp;
178 			*sipp = *xipp;
179 			*xipp = save;
180 		}
181 	}
182 
183 	/*
184 	 * Traverse the (now) ordered list of system initialization tasks.
185 	 * Perform each task, and continue on to the next task.
186 	 *
187 	 * The last item on the list is expected to be the scheduler,
188 	 * which will not return.
189 	 */
190 	for( sipp = (struct sysinit **)sysinit_set.ls_items; *sipp; sipp++) {
191 		if( (*sipp)->subsystem == SI_SUB_DUMMY)
192 			continue;	/* skip dummy task(s)*/
193 
194 		switch( (*sipp)->type) {
195 		case SI_TYPE_DEFAULT:
196 			/* no special processing*/
197 			(*((*sipp)->func))( (*sipp)->udata);
198 			break;
199 
200 		case SI_TYPE_KTHREAD:
201 			/* kernel thread*/
202 			if (fork(&proc0, NULL, rval))
203 				panic("fork kernel process");
204 			if (rval[1]) {
205 				(*((*sipp)->func))( (*sipp)->udata);
206 				/*
207 				 * The call to start "init" returns
208 				 * here after the scheduler has been
209 				 * started, and returns to the caller
210 				 * in i386/i386/locore.s.  This is a
211 				 * necessary part of initialization
212 				 * and is rather non-obvious.
213 				 *
214 				 * No other "kernel threads" should
215 				 * return here.  Call panic() instead.
216 				 */
217 				return;
218 			}
219 			break;
220 
221 		default:
222 			panic( "init_main: unrecognized init type");
223 		}
224 	}
225 
226 	/* NOTREACHED*/
227 }
228 
229 
230 /*
231  * Start a kernel process.  This is called after a fork() call in
232  * main() in the file kern/init_main.c.
233  *
234  * This function is used to start "internal" daemons.
235  */
236 /* ARGSUSED*/
237 void
238 kproc_start(udata)
239 	void *udata;
240 {
241 	struct kproc_desc	*kp = udata;
242 	struct proc		*p = curproc;
243 
244 	/* save a global descriptor, if desired*/
245 	if( kp->global_procpp != NULL)
246 		*kp->global_procpp	= p;
247 
248 	/* this is a non-swapped system process*/
249 	p->p_flag |= P_INMEM | P_SYSTEM;
250 
251 	/* set up arg0 for 'ps', et al*/
252 	strcpy( p->p_comm, kp->arg0);
253 
254 	/* call the processes' main()...*/
255 	(*kp->func)();
256 
257 	/* NOTREACHED */
258 	panic("kproc_start: %s", kp->arg0);
259 }
260 
261 
262 /*
263  ***************************************************************************
264  ****
265  **** The following SYSINIT's belong elsewhere, but have not yet
266  **** been moved.
267  ****
268  ***************************************************************************
269  */
270 #ifdef OMIT
271 /*
272  * Handled by vfs_mountroot (bad idea) at this time... should be
273  * done the same as 4.4Lite2.
274  */
275 SYSINIT(swapinit, SI_SUB_SWAP, SI_ORDER_FIRST, swapinit, NULL)
276 #endif	/* OMIT*/
277 
278 /*
279  * Should get its own file...
280  */
281 #ifdef HPFPLIB
282 char	copyright[] =
283 "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";
284 #else
285 char	copyright[] =
286 "Copyright (c) 1982, 1986, 1989, 1991, 1993\n\tThe Regents of the University of California.  All rights reserved.\n\n";
287 #endif
288 static void print_caddr_t __P((void *data));
289 static void
290 print_caddr_t(data)
291 	void *data;
292 {
293 	printf("%s", (char *)data);
294 }
295 SYSINIT(announce, SI_SUB_COPYRIGHT, SI_ORDER_FIRST, print_caddr_t, copyright)
296 
297 
298 /*
299  ***************************************************************************
300  ****
301  **** The two following SYSINT's are proc0 specific glue code.  I am not
302  **** convinced that they can not be safely combined, but their order of
303  **** operation has been maintained as the same as the original init_main.c
304  **** for right now.
305  ****
306  **** These probably belong in init_proc.c or kern_proc.c, since they
307  **** deal with proc0 (the fork template process).
308  ****
309  ***************************************************************************
310  */
311 /* ARGSUSED*/
312 static void proc0_init __P((void *dummy));
313 static void
314 proc0_init(dummy)
315 	void *dummy;
316 {
317 	register struct proc		*p;
318 	register struct filedesc0	*fdp;
319 	register int i;
320 
321 	/*
322 	 * Initialize the current process pointer (curproc) before
323 	 * any possible traps/probes to simplify trap processing.
324 	 */
325 	p = &proc0;
326 	curproc = p;			/* XXX redundant*/
327 
328 	/*
329 	 * Initialize process and pgrp structures.
330 	 */
331 	procinit();
332 
333 	/*
334 	 * Create process 0 (the swapper).
335 	 */
336 	LIST_INSERT_HEAD(&allproc, p, p_list);
337 	p->p_pgrp = &pgrp0;
338 	LIST_INSERT_HEAD(PGRPHASH(0), &pgrp0, pg_hash);
339 	LIST_INIT(&pgrp0.pg_members);
340 	LIST_INSERT_HEAD(&pgrp0.pg_members, p, p_pglist);
341 
342 	pgrp0.pg_session = &session0;
343 	session0.s_count = 1;
344 	session0.s_leader = p;
345 
346 	p->p_sysent = &aout_sysvec;
347 
348 	p->p_flag = P_INMEM | P_SYSTEM;
349 	p->p_stat = SRUN;
350 	p->p_nice = NZERO;
351 	p->p_rtprio.type = RTP_PRIO_NORMAL;
352 	p->p_rtprio.prio = 0;
353 
354 	bcopy("swapper", p->p_comm, sizeof ("swapper"));
355 
356 	/* Create credentials. */
357 	cred0.p_refcnt = 1;
358 	p->p_cred = &cred0;
359 	p->p_ucred = crget();
360 	p->p_ucred->cr_ngroups = 1;	/* group 0 */
361 
362 	/* Create the file descriptor table. */
363 	fdp = &filedesc0;
364 	p->p_fd = &fdp->fd_fd;
365 	fdp->fd_fd.fd_refcnt = 1;
366 	fdp->fd_fd.fd_cmask = cmask;
367 	fdp->fd_fd.fd_ofiles = fdp->fd_dfiles;
368 	fdp->fd_fd.fd_ofileflags = fdp->fd_dfileflags;
369 	fdp->fd_fd.fd_nfiles = NDFILE;
370 
371 	/* Create the limits structures. */
372 	p->p_limit = &limit0;
373 	for (i = 0; i < sizeof(p->p_rlimit)/sizeof(p->p_rlimit[0]); i++)
374 		limit0.pl_rlimit[i].rlim_cur =
375 		    limit0.pl_rlimit[i].rlim_max = RLIM_INFINITY;
376 	limit0.pl_rlimit[RLIMIT_NOFILE].rlim_cur = NOFILE;
377 	limit0.pl_rlimit[RLIMIT_NPROC].rlim_cur = MAXUPRC;
378 	i = ptoa(cnt.v_free_count);
379 	limit0.pl_rlimit[RLIMIT_RSS].rlim_max = i;
380 	limit0.pl_rlimit[RLIMIT_MEMLOCK].rlim_max = i;
381 	limit0.pl_rlimit[RLIMIT_MEMLOCK].rlim_cur = i / 3;
382 	limit0.p_refcnt = 1;
383 
384 	/* Allocate a prototype map so we have something to fork. */
385 	p->p_vmspace = &vmspace0;
386 	vmspace0.vm_refcnt = 1;
387 	pmap_pinit(&vmspace0.vm_pmap);
388 	vm_map_init(&vmspace0.vm_map, round_page(VM_MIN_ADDRESS),
389 	    trunc_page(VM_MAX_ADDRESS), TRUE);
390 	vmspace0.vm_map.pmap = &vmspace0.vm_pmap;
391 	p->p_addr = proc0paddr;				/* XXX */
392 
393 #define INCOMPAT_LITES2
394 #ifdef INCOMPAT_LITES2
395 	/*
396 	 * proc0 needs to have a coherent frame base, too.
397 	 * This probably makes the identical call for the init proc
398 	 * that happens later unnecessary since it should inherit
399 	 * it during the fork.
400 	 */
401 	cpu_set_init_frame(p, init_framep);			/* XXX! */
402 #endif	/* INCOMPAT_LITES2*/
403 
404 	/*
405 	 * We continue to place resource usage info and signal
406 	 * actions in the user struct so they're pageable.
407 	 */
408 	p->p_stats = &p->p_addr->u_stats;
409 	p->p_sigacts = &p->p_addr->u_sigacts;
410 
411 	/*
412 	 * Charge root for one process.
413 	 */
414 	(void)chgproccnt(0, 1);
415 }
416 SYSINIT(p0init, SI_SUB_INTRINSIC, SI_ORDER_FIRST, proc0_init, NULL)
417 
418 /* ARGSUSED*/
419 static void proc0_post __P((void *dummy));
420 static void
421 proc0_post(dummy)
422 	void *dummy;
423 {
424 	/*
425 	 * Now can look at time, having had a chance to verify the time
426 	 * from the file system.  Reset p->p_rtime as it may have been
427 	 * munched in mi_switch() after the time got set.
428 	 */
429 	proc0.p_stats->p_start = runtime = mono_time = boottime = time;
430 	proc0.p_rtime.tv_sec = proc0.p_rtime.tv_usec = 0;
431 
432 	/* Initialize signal state for process 0. */
433 	siginit(&proc0);
434 }
435 SYSINIT(p0post, SI_SUB_INTRINSIC_POST, SI_ORDER_FIRST, proc0_post, NULL)
436 
437 
438 
439 
440 /*
441  ***************************************************************************
442  ****
443  **** The following SYSINIT's and glue code should be moved to the
444  **** respective files on a per subsystem basis.
445  ****
446  ***************************************************************************
447  */
448 /* ARGSUSED*/
449 static void sched_setup __P((void *dummy));
450 static void
451 sched_setup(dummy)
452 	void *dummy;
453 {
454 	/* Kick off timeout driven events by calling first time. */
455 	roundrobin(NULL);
456 	schedcpu(NULL);
457 }
458 SYSINIT(sched_setup, SI_SUB_KICK_SCHEDULER, SI_ORDER_FIRST, sched_setup, NULL)
459 
460 /* ARGSUSED*/
461 static void xxx_vfs_mountroot __P((void *dummy));
462 static void
463 xxx_vfs_mountroot(dummy)
464 	void *dummy;
465 {
466 	/* Mount the root file system. */
467 	if ((*mountroot)(mountrootvfsops))
468 		panic("cannot mount root");
469 }
470 SYSINIT(mountroot, SI_SUB_ROOT, SI_ORDER_FIRST, xxx_vfs_mountroot, NULL)
471 
472 /* ARGSUSED*/
473 static void xxx_vfs_root_fdtab __P((void *dummy));
474 static void
475 xxx_vfs_root_fdtab(dummy)
476 	void *dummy;
477 {
478 	register struct filedesc0	*fdp = &filedesc0;
479 
480 	/* Get the vnode for '/'.  Set fdp->fd_fd.fd_cdir to reference it. */
481 	if (VFS_ROOT(mountlist.cqh_first, &rootvnode))
482 		panic("cannot find root vnode");
483 	fdp->fd_fd.fd_cdir = rootvnode;
484 	VREF(fdp->fd_fd.fd_cdir);
485 	VOP_UNLOCK(rootvnode);
486 	fdp->fd_fd.fd_rdir = NULL;
487 }
488 SYSINIT(retrofit, SI_SUB_ROOT_FDTAB, SI_ORDER_FIRST, xxx_vfs_root_fdtab, NULL)
489 
490 
491 /*
492  ***************************************************************************
493  ****
494  **** The following code probably belongs in another file, like
495  **** kern/init_init.c.  It is here for two reasons only:
496  ****
497  ****	1)	This code returns to startup the system; this is
498  ****		abnormal for a kernel thread.
499  ****	2)	This code promiscuously uses init_frame
500  ****
501  ***************************************************************************
502  */
503 
504 static void kthread_init __P((void *dummy));
505 SYSINIT_KT(init,SI_SUB_KTHREAD_INIT, SI_ORDER_FIRST, kthread_init, NULL)
506 
507 
508 static void start_init __P((struct proc *p, void *framep));
509 
510 /* ARGSUSED*/
511 static void
512 kthread_init(dummy)
513 	void *dummy;
514 {
515 
516 	/* Create process 1 (init(8)). */
517 	start_init(curproc, init_framep);
518 
519 	/*
520 	 * This is the only kernel thread allowed to return yo the
521 	 * caller!!!
522 	 */
523 	return;
524 }
525 
526 
527 /*
528  * List of paths to try when searching for "init".
529  */
530 static char *initpaths[] = {
531 	"/sbin/init",
532 	"/sbin/oinit",
533 	"/sbin/init.bak",
534 	"/stand/sysinstall",
535 	NULL,
536 };
537 
538 /*
539  * Start the initial user process; try exec'ing each pathname in "initpaths".
540  * The program is invoked with one argument containing the boot flags.
541  */
542 static void
543 start_init(p, framep)
544 	struct proc *p;
545 	void *framep;
546 {
547 	vm_offset_t addr;
548 	struct execve_args args;
549 	int options, i, retval[2], error;
550 	char **pathp, *path, *ucp, **uap, *arg0, *arg1;
551 
552 	initproc = p;
553 
554 	/*
555 	 * We need to set the system call frame as if we were entered through
556 	 * a syscall() so that when we call execve() below, it will be able
557 	 * to set the entry point (see setregs) when it tries to exec.  The
558 	 * startup code in "locore.s" has allocated space for the frame and
559 	 * passed a pointer to that space as main's argument.
560 	 */
561 	cpu_set_init_frame(p, framep);
562 
563 	/*
564 	 * Need just enough stack to hold the faked-up "execve()" arguments.
565 	 */
566 	addr = trunc_page(VM_MAXUSER_ADDRESS - PAGE_SIZE);
567 	if (vm_map_find(&p->p_vmspace->vm_map, NULL, 0, &addr, PAGE_SIZE, FALSE, VM_PROT_ALL, VM_PROT_ALL, 0) != 0)
568 		panic("init: couldn't allocate argument space");
569 	p->p_vmspace->vm_maxsaddr = (caddr_t)addr;
570 	p->p_vmspace->vm_ssize = 1;
571 
572 	for (pathp = &initpaths[0]; (path = *pathp) != NULL; pathp++) {
573 		/*
574 		 * Move out the boot flag argument.
575 		 */
576 		options = 0;
577 		ucp = (char *)USRSTACK;
578 		(void)subyte(--ucp, 0);		/* trailing zero */
579 		if (boothowto & RB_SINGLE) {
580 			(void)subyte(--ucp, 's');
581 			options = 1;
582 		}
583 #ifdef notyet
584                 if (boothowto & RB_FASTBOOT) {
585 			(void)subyte(--ucp, 'f');
586 			options = 1;
587 		}
588 #endif
589 
590 #ifdef BOOTCDROM
591 		(void)subyte(--ucp, 'C');
592 		options = 1;
593 #endif
594 		if (options == 0)
595 			(void)subyte(--ucp, '-');
596 		(void)subyte(--ucp, '-');		/* leading hyphen */
597 		arg1 = ucp;
598 
599 		/*
600 		 * Move out the file name (also arg 0).
601 		 */
602 		for (i = strlen(path) + 1; i >= 0; i--)
603 			(void)subyte(--ucp, path[i]);
604 		arg0 = ucp;
605 
606 		/*
607 		 * Move out the arg pointers.
608 		 */
609 		uap = (char **)((int)ucp & ~(NBPW-1));
610 		(void)suword((caddr_t)--uap, 0);	/* terminator */
611 		(void)suword((caddr_t)--uap, (int)arg1);
612 		(void)suword((caddr_t)--uap, (int)arg0);
613 
614 		/*
615 		 * Point at the arguments.
616 		 */
617 		args.fname = arg0;
618 		args.argv = uap;
619 		args.envv = NULL;
620 
621 		/*
622 		 * Now try to exec the program.  If can't for any reason
623 		 * other than it doesn't exist, complain.
624 		 *
625 		 * Otherwise return to main() which returns to btext
626 		 * which completes the system startup.
627 		 */
628 		if ((error = execve(p, &args, &retval[0])) == 0)
629 			return;
630 		if (error != ENOENT)
631 			printf("exec %s: error %d\n", path, error);
632 	}
633 	printf("init: not found\n");
634 	panic("no init");
635 }
636