xref: /freebsd/sys/kern/init_main.c (revision ce834215a70ff69e7e222827437116eee2f9ac6f)
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 1.65 1997/06/22 16:04:09 peter Exp $
43  */
44 
45 #include "opt_rlimit.h"
46 #include "opt_devfs.h"
47 
48 #include <sys/param.h>
49 #include <sys/file.h>
50 #include <sys/filedesc.h>
51 #include <sys/kernel.h>
52 #include <sys/mount.h>
53 #include <sys/sysctl.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/reboot.h>
61 #include <sys/sysproto.h>
62 #include <sys/vmmeter.h>
63 
64 #include <machine/cpu.h>
65 #if defined(SMP)
66 #include <machine/smp.h>
67 #endif /* SMP */
68 
69 #include <vm/vm.h>
70 #include <vm/vm_param.h>
71 #include <vm/vm_prot.h>
72 #include <sys/lock.h>
73 #include <vm/pmap.h>
74 #include <vm/vm_map.h>
75 #include <sys/user.h>
76 #include <sys/copyright.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 extern void secondary_main __P((void));
83 
84 /* Components of the first process -- never freed. */
85 static struct session session0;
86 static struct pgrp pgrp0;
87 struct	proc proc0;
88 static struct pcred cred0;
89 static struct filedesc0 filedesc0;
90 static struct plimit limit0;
91 static struct vmspace vmspace0;
92 #ifndef SMP	/* per-cpu on smp */
93 struct	proc *curproc = &proc0;
94 #endif
95 struct	proc *initproc;
96 
97 int cmask = CMASK;
98 extern	struct user *proc0paddr;
99 
100 struct	vnode *rootvp;
101 int	boothowto;
102 
103 struct	timeval boottime;
104 SYSCTL_STRUCT(_kern, KERN_BOOTTIME, boottime,
105 	CTLFLAG_RW, &boottime, timeval, "");
106 
107 static int shutdowntimeout = 120;
108 SYSCTL_INT(_kern, OID_AUTO, shutdown_timeout,
109 	CTLFLAG_RW, &shutdowntimeout, 0, "");
110 
111 #ifndef SMP	/* per-cpu on smp */
112 struct	timeval runtime;
113 #endif
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 routine 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 	 * Copy the locore.s frame pointer for proc0, this is forked into
161 	 * all other processes.
162 	 */
163 	init_framep = framep;
164 
165 	/*
166 	 * Perform a bubble sort of the system initialization objects by
167 	 * their subsystem (primary key) and order (secondary key).
168 	 *
169 	 * Since some things care about execution order, this is the
170 	 * operation which ensures continued function.
171 	 */
172 	for( sipp = (struct sysinit **)sysinit_set.ls_items; *sipp; sipp++) {
173 		for( xipp = sipp + 1; *xipp; xipp++) {
174 			if( (*sipp)->subsystem < (*xipp)->subsystem ||
175 			    ( (*sipp)->subsystem == (*xipp)->subsystem &&
176 			      (*sipp)->order < (*xipp)->order))
177 				continue;	/* skip*/
178 			save = *sipp;
179 			*sipp = *xipp;
180 			*xipp = save;
181 		}
182 	}
183 
184 	/*
185 	 * Traverse the (now) ordered list of system initialization tasks.
186 	 * Perform each task, and continue on to the next task.
187 	 *
188 	 * The last item on the list is expected to be the scheduler,
189 	 * which will not return.
190 	 */
191 	for( sipp = (struct sysinit **)sysinit_set.ls_items; *sipp; sipp++) {
192 		if( (*sipp)->subsystem == SI_SUB_DUMMY)
193 			continue;	/* skip dummy task(s)*/
194 
195 		switch( (*sipp)->type) {
196 		case SI_TYPE_DEFAULT:
197 			/* no special processing*/
198 			(*((*sipp)->func))( (*sipp)->udata);
199 			break;
200 
201 		case SI_TYPE_KTHREAD:
202 			/* kernel thread*/
203 			if (fork(&proc0, NULL, rval))
204 				panic("fork kernel process");
205 			cpu_set_fork_handler(pfind(rval[0]), (*sipp)->func, (*sipp)->udata);
206 			break;
207 
208 		default:
209 			panic( "init_main: unrecognized init type");
210 		}
211 	}
212 
213 	panic("Shouldn't get here!");
214 	/* NOTREACHED*/
215 }
216 
217 
218 /*
219  * Start a kernel process.  This is called after a fork() call in
220  * main() in the file kern/init_main.c.
221  *
222  * This function is used to start "internal" daemons.
223  */
224 /* ARGSUSED*/
225 void
226 kproc_start(udata)
227 	void *udata;
228 {
229 	struct kproc_desc	*kp = udata;
230 	struct proc		*p = curproc;
231 
232 #ifdef DIAGNOSTIC
233 	printf("Start pid=%d <%s>\n",p->p_pid, kp->arg0);
234 #endif
235 
236 
237 	/* save a global descriptor, if desired*/
238 	if( kp->global_procpp != NULL)
239 		*kp->global_procpp	= p;
240 
241 	/* this is a non-swapped system process*/
242 	p->p_flag |= P_INMEM | P_SYSTEM;
243 
244 	/* set up arg0 for 'ps', et al*/
245 	strcpy( p->p_comm, kp->arg0);
246 
247 	/* call the processes' main()...*/
248 	(*kp->func)();
249 
250 	/* NOTREACHED */
251 	panic("kproc_start: %s", kp->arg0);
252 }
253 
254 
255 /*
256  ***************************************************************************
257  ****
258  **** The following SYSINIT's belong elsewhere, but have not yet
259  **** been moved.
260  ****
261  ***************************************************************************
262  */
263 #ifdef OMIT
264 /*
265  * Handled by vfs_mountroot (bad idea) at this time... should be
266  * done the same as 4.4Lite2.
267  */
268 SYSINIT(swapinit, SI_SUB_SWAP, SI_ORDER_FIRST, swapinit, NULL)
269 #endif	/* OMIT*/
270 
271 static void print_caddr_t __P((void *data));
272 static void
273 print_caddr_t(data)
274 	void *data;
275 {
276 	printf("%s", (char *)data);
277 }
278 SYSINIT(announce, SI_SUB_COPYRIGHT, SI_ORDER_FIRST, print_caddr_t, copyright)
279 
280 
281 /*
282  ***************************************************************************
283  ****
284  **** The two following SYSINT's are proc0 specific glue code.  I am not
285  **** convinced that they can not be safely combined, but their order of
286  **** operation has been maintained as the same as the original init_main.c
287  **** for right now.
288  ****
289  **** These probably belong in init_proc.c or kern_proc.c, since they
290  **** deal with proc0 (the fork template process).
291  ****
292  ***************************************************************************
293  */
294 /* ARGSUSED*/
295 static void proc0_init __P((void *dummy));
296 static void
297 proc0_init(dummy)
298 	void *dummy;
299 {
300 	register struct proc		*p;
301 	register struct filedesc0	*fdp;
302 	register unsigned i;
303 
304 	/*
305 	 * Initialize the current process pointer (curproc) before
306 	 * any possible traps/probes to simplify trap processing.
307 	 */
308 	p = &proc0;
309 	curproc = p;			/* XXX redundant*/
310 
311 	/*
312 	 * Initialize process and pgrp structures.
313 	 */
314 	procinit();
315 
316 	/*
317 	 * Initialize sleep queue hash table
318 	 */
319 	sleepinit();
320 
321 	/*
322 	 * Create process 0 (the swapper).
323 	 */
324 	LIST_INSERT_HEAD(&allproc, p, p_list);
325 	p->p_pgrp = &pgrp0;
326 	LIST_INSERT_HEAD(PGRPHASH(0), &pgrp0, pg_hash);
327 	LIST_INIT(&pgrp0.pg_members);
328 	LIST_INSERT_HEAD(&pgrp0.pg_members, p, p_pglist);
329 
330 	pgrp0.pg_session = &session0;
331 	session0.s_count = 1;
332 	session0.s_leader = p;
333 
334 	p->p_sysent = &aout_sysvec;
335 
336 	p->p_flag = P_INMEM | P_SYSTEM;
337 	p->p_stat = SRUN;
338 	p->p_nice = NZERO;
339 	p->p_rtprio.type = RTP_PRIO_NORMAL;
340 	p->p_rtprio.prio = 0;
341 
342 /*
343  * Link for kernel based threads
344  */
345 	p->p_peers = 0;
346 	p->p_leader = p;
347 
348 	bcopy("swapper", p->p_comm, sizeof ("swapper"));
349 
350 	/* Create credentials. */
351 	cred0.p_refcnt = 1;
352 	p->p_cred = &cred0;
353 	p->p_ucred = crget();
354 	p->p_ucred->cr_ngroups = 1;	/* group 0 */
355 
356 	/* Create the file descriptor table. */
357 	fdp = &filedesc0;
358 	p->p_fd = &fdp->fd_fd;
359 	fdp->fd_fd.fd_refcnt = 1;
360 	fdp->fd_fd.fd_cmask = cmask;
361 	fdp->fd_fd.fd_ofiles = fdp->fd_dfiles;
362 	fdp->fd_fd.fd_ofileflags = fdp->fd_dfileflags;
363 	fdp->fd_fd.fd_nfiles = NDFILE;
364 
365 	/* Create the limits structures. */
366 	p->p_limit = &limit0;
367 	for (i = 0; i < sizeof(p->p_rlimit)/sizeof(p->p_rlimit[0]); i++)
368 		limit0.pl_rlimit[i].rlim_cur =
369 		    limit0.pl_rlimit[i].rlim_max = RLIM_INFINITY;
370 	limit0.pl_rlimit[RLIMIT_NOFILE].rlim_cur =
371 	    limit0.pl_rlimit[RLIMIT_NOFILE].rlim_max = maxfiles;
372 	limit0.pl_rlimit[RLIMIT_NPROC].rlim_cur =
373 	    limit0.pl_rlimit[RLIMIT_NPROC].rlim_max = maxproc;
374 	i = ptoa(cnt.v_free_count);
375 	limit0.pl_rlimit[RLIMIT_RSS].rlim_max = i;
376 	limit0.pl_rlimit[RLIMIT_MEMLOCK].rlim_max = i;
377 	limit0.pl_rlimit[RLIMIT_MEMLOCK].rlim_cur = i / 3;
378 	limit0.p_refcnt = 1;
379 
380 	/* Allocate a prototype map so we have something to fork. */
381 	p->p_vmspace = &vmspace0;
382 	vmspace0.vm_refcnt = 1;
383 	pmap_pinit(&vmspace0.vm_pmap);
384 	vm_map_init(&vmspace0.vm_map, round_page(VM_MIN_ADDRESS),
385 	    trunc_page(VM_MAXUSER_ADDRESS), TRUE);
386 	vmspace0.vm_map.pmap = &vmspace0.vm_pmap;
387 	p->p_addr = proc0paddr;				/* XXX */
388 
389 #define INCOMPAT_LITES2
390 #ifdef INCOMPAT_LITES2
391 	/*
392 	 * proc0 needs to have a coherent frame base in it's stack.
393 	 */
394 	cpu_set_init_frame(p, init_framep);			/* XXX! */
395 #endif	/* INCOMPAT_LITES2*/
396 
397 	/*
398 	 * We continue to place resource usage info and signal
399 	 * actions in the user struct so they're pageable.
400 	 */
401 	p->p_stats = &p->p_addr->u_stats;
402 	p->p_sigacts = &p->p_addr->u_sigacts;
403 
404 	/*
405 	 * Charge root for one process.
406 	 */
407 	(void)chgproccnt(0, 1);
408 }
409 SYSINIT(p0init, SI_SUB_INTRINSIC, SI_ORDER_FIRST, proc0_init, NULL)
410 
411 /* ARGSUSED*/
412 static void proc0_post __P((void *dummy));
413 static void
414 proc0_post(dummy)
415 	void *dummy;
416 {
417 	struct timeval tv;
418 
419 	/*
420 	 * Now can look at time, having had a chance to verify the time
421 	 * from the file system.  Reset p->p_rtime as it may have been
422 	 * munched in mi_switch() after the time got set.
423 	 */
424 	gettime(&boottime);
425 	proc0.p_stats->p_start = runtime = mono_time = boottime;
426 	proc0.p_rtime.tv_sec = proc0.p_rtime.tv_usec = 0;
427 
428 	/*
429 	 * Give the ``random'' number generator a thump.
430 	 */
431 	microtime(&tv);
432 	srandom(tv.tv_sec ^ tv.tv_usec);
433 
434 	/* Initialize signal state for process 0. */
435 	siginit(&proc0);
436 }
437 SYSINIT(p0post, SI_SUB_INTRINSIC_POST, SI_ORDER_FIRST, proc0_post, NULL)
438 
439 
440 
441 
442 /*
443  ***************************************************************************
444  ****
445  **** The following SYSINIT's and glue code should be moved to the
446  **** respective files on a per subsystem basis.
447  ****
448  ***************************************************************************
449  */
450 /* ARGSUSED*/
451 static void sched_setup __P((void *dummy));
452 static void
453 sched_setup(dummy)
454 	void *dummy;
455 {
456 	/* Kick off timeout driven events by calling first time. */
457 	roundrobin(NULL);
458 	schedcpu(NULL);
459 }
460 SYSINIT(sched_setup, SI_SUB_KICK_SCHEDULER, SI_ORDER_FIRST, sched_setup, NULL)
461 
462 /* ARGSUSED*/
463 static void xxx_vfs_mountroot __P((void *fsnamep));
464 #ifdef BOOTP
465 extern void bootpc_init __P((void));
466 #endif
467 static void
468 xxx_vfs_mountroot(fsnamep)
469 	void *fsnamep;
470 {
471   /* XXX Add a separate SYSINIT entry */
472 #ifdef BOOTP
473 	bootpc_init();
474 #endif
475 	/* Mount the root file system. */
476 	if (vfs_mountrootfs(*((char **) fsnamep)))
477 		panic("cannot mount root");
478 }
479 SYSINIT(mountroot, SI_SUB_ROOT, SI_ORDER_FIRST, xxx_vfs_mountroot, &mountrootfsname)
480 
481 /* ARGSUSED*/
482 static void xxx_vfs_root_fdtab __P((void *dummy));
483 static void
484 xxx_vfs_root_fdtab(dummy)
485 	void *dummy;
486 {
487 	register struct filedesc0	*fdp = &filedesc0;
488 
489 	/* Get the vnode for '/'.  Set fdp->fd_fd.fd_cdir to reference it. */
490 	if (VFS_ROOT(mountlist.cqh_first, &rootvnode))
491 		panic("cannot find root vnode");
492 	fdp->fd_fd.fd_cdir = rootvnode;
493 	VREF(fdp->fd_fd.fd_cdir);
494 	VOP_UNLOCK(rootvnode, 0, &proc0);
495 	fdp->fd_fd.fd_rdir = NULL;
496 }
497 SYSINIT(retrofit, SI_SUB_ROOT_FDTAB, SI_ORDER_FIRST, xxx_vfs_root_fdtab, NULL)
498 
499 
500 /*
501  ***************************************************************************
502  ****
503  **** The following code probably belongs in another file, like
504  **** kern/init_init.c.  It is here for two reasons only:
505  ****
506  ****	1)	This code returns to startup the system; this is
507  ****		abnormal for a kernel thread.
508  ****	2)	This code promiscuously uses init_frame
509  ****
510  ***************************************************************************
511  */
512 
513 static void kthread_init __P((void *dummy));
514 SYSINIT_KT(init,SI_SUB_KTHREAD_INIT, SI_ORDER_FIRST, kthread_init, NULL)
515 
516 
517 extern void prepare_usermode __P((void));
518 static void start_init __P((struct proc *p));
519 
520 /* ARGSUSED*/
521 static void
522 kthread_init(dummy)
523 	void *dummy;
524 {
525 
526 	/* Create process 1 (init(8)). */
527 	start_init(curproc);
528 
529 	prepare_usermode();
530 
531 	/*
532 	 * This returns to the fork trampoline, then to user mode.
533 	 */
534 	return;
535 }
536 
537 
538 /*
539  * List of paths to try when searching for "init".
540  */
541 static char *initpaths[] = {
542 	"/sbin/init",
543 	"/sbin/oinit",
544 	"/sbin/init.bak",
545 	"/stand/sysinstall",
546 	NULL,
547 };
548 
549 /*
550  * Start the initial user process; try exec'ing each pathname in "initpaths".
551  * The program is invoked with one argument containing the boot flags.
552  */
553 static void
554 start_init(p)
555 	struct proc *p;
556 {
557 	vm_offset_t addr;
558 	struct execve_args args;
559 	int options, i, retval[2], error;
560 	char **pathp, *path, *ucp, **uap, *arg0, *arg1;
561 
562 	initproc = p;
563 
564 	/*
565 	 * Need just enough stack to hold the faked-up "execve()" arguments.
566 	 */
567 	addr = trunc_page(VM_MAXUSER_ADDRESS - PAGE_SIZE);
568 	if (vm_map_find(&p->p_vmspace->vm_map, NULL, 0, &addr, PAGE_SIZE, FALSE, VM_PROT_ALL, VM_PROT_ALL, 0) != 0)
569 		panic("init: couldn't allocate argument space");
570 	p->p_vmspace->vm_maxsaddr = (caddr_t)addr;
571 	p->p_vmspace->vm_ssize = 1;
572 
573 	for (pathp = &initpaths[0]; (path = *pathp) != NULL; pathp++) {
574 		/*
575 		 * Move out the boot flag argument.
576 		 */
577 		options = 0;
578 		ucp = (char *)USRSTACK;
579 		(void)subyte(--ucp, 0);		/* trailing zero */
580 		if (boothowto & RB_SINGLE) {
581 			(void)subyte(--ucp, 's');
582 			options = 1;
583 		}
584 #ifdef notyet
585                 if (boothowto & RB_FASTBOOT) {
586 			(void)subyte(--ucp, 'f');
587 			options = 1;
588 		}
589 #endif
590 
591 #ifdef BOOTCDROM
592 		(void)subyte(--ucp, 'C');
593 		options = 1;
594 #endif
595 
596 #if defined(DEVFS) && defined(DEVFS_ROOT)
597 		(void)subyte(--ucp, 'd');
598 		options = 1;
599 #endif
600 		if (options == 0)
601 			(void)subyte(--ucp, '-');
602 		(void)subyte(--ucp, '-');		/* leading hyphen */
603 		arg1 = ucp;
604 
605 		/*
606 		 * Move out the file name (also arg 0).
607 		 */
608 		for (i = strlen(path) + 1; i >= 0; i--)
609 			(void)subyte(--ucp, path[i]);
610 		arg0 = ucp;
611 
612 		/*
613 		 * Move out the arg pointers.
614 		 */
615 		uap = (char **)((int)ucp & ~(NBPW-1));
616 		(void)suword((caddr_t)--uap, 0);	/* terminator */
617 		(void)suword((caddr_t)--uap, (int)arg1);
618 		(void)suword((caddr_t)--uap, (int)arg0);
619 
620 		/*
621 		 * Point at the arguments.
622 		 */
623 		args.fname = arg0;
624 		args.argv = uap;
625 		args.envv = NULL;
626 
627 		/*
628 		 * Now try to exec the program.  If can't for any reason
629 		 * other than it doesn't exist, complain.
630 		 *
631 		 * Otherwise return to main() which returns to btext
632 		 * which completes the system startup.
633 		 */
634 		if ((error = execve(p, &args, &retval[0])) == 0)
635 			return;
636 		if (error != ENOENT)
637 			printf("exec %s: error %d\n", path, error);
638 	}
639 	printf("init: not found\n");
640 	panic("no init");
641 }
642