xref: /freebsd/sys/kern/init_main.c (revision 5129159789cc9d7bc514e4546b88e3427695002d)
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  * $FreeBSD$
43  */
44 
45 #include "opt_init_path.h"
46 
47 #include <sys/param.h>
48 #include <sys/file.h>
49 #include <sys/filedesc.h>
50 #include <sys/kernel.h>
51 #include <sys/mount.h>
52 #include <sys/sysctl.h>
53 #include <sys/proc.h>
54 #include <sys/resourcevar.h>
55 #include <sys/signalvar.h>
56 #include <sys/systm.h>
57 #include <sys/vnode.h>
58 #include <sys/sysent.h>
59 #include <sys/reboot.h>
60 #include <sys/sysproto.h>
61 #include <sys/vmmeter.h>
62 #include <sys/unistd.h>
63 #include <sys/malloc.h>
64 
65 #include <machine/cpu.h>
66 
67 #include <vm/vm.h>
68 #include <vm/vm_param.h>
69 #include <sys/lock.h>
70 #include <vm/pmap.h>
71 #include <vm/vm_map.h>
72 #include <sys/user.h>
73 #include <sys/copyright.h>
74 
75 extern struct linker_set	sysinit_set;	/* XXX */
76 
77 extern void mi_startup __P((void *framep));
78 
79 /* Components of the first process -- never freed. */
80 static struct session session0;
81 static struct pgrp pgrp0;
82 struct	proc proc0;
83 static struct pcred cred0;
84 static struct procsig procsig0;
85 static struct filedesc0 filedesc0;
86 static struct plimit limit0;
87 static struct vmspace vmspace0;
88 struct	proc *initproc;
89 
90 int cmask = CMASK;
91 extern	struct user *proc0paddr;
92 
93 struct	vnode *rootvp;
94 int	boothowto = 0;		/* initialized so that it can be patched */
95 
96 /*
97  * Promiscuous argument pass for start_init()
98  *
99  * This is a kludge because we use a return from mi_startup() rather than a call
100  * to a new routine in locore.s to kick the kernel alive from locore.s.
101  */
102 static void	*init_framep;
103 
104 /*
105  * This ensures that there is at least one entry so that the sysinit_set
106  * symbol is not undefined.  A sybsystem ID of SI_SUB_DUMMY is never
107  * executed.
108  */
109 SYSINIT(placeholder, SI_SUB_DUMMY,SI_ORDER_ANY, NULL, NULL)
110 
111 /*
112  * The sysinit table itself.  Items are checked off as the are run.
113  * If we want to register new sysinit types, add them to newsysinit.
114  */
115 struct sysinit **sysinit = (struct sysinit **)sysinit_set.ls_items;
116 struct sysinit **newsysinit;
117 
118 /*
119  * Merge a new sysinit set into the current set, reallocating it if
120  * necessary.  This can only be called after malloc is running.
121  */
122 void
123 sysinit_add(set)
124 	struct sysinit **set;
125 {
126 	struct sysinit **newset;
127 	struct sysinit **sipp;
128 	struct sysinit **xipp;
129 	int count = 0;
130 
131 	if (newsysinit)
132 		for (sipp = newsysinit; *sipp; sipp++)
133 			count++;
134 	else
135 		for (sipp = sysinit; *sipp; sipp++)
136 			count++;
137 	for (sipp = set; *sipp; sipp++)
138 		count++;
139 	count++;		/* Trailing NULL */
140 	newset = malloc(count * sizeof(*sipp), M_TEMP, M_NOWAIT);
141 	if (newset == NULL)
142 		panic("cannot malloc for sysinit");
143 	xipp = newset;
144 	if (newsysinit)
145 		for (sipp = newsysinit; *sipp; sipp++)
146 			*xipp++ = *sipp;
147 	else
148 		for (sipp = sysinit; *sipp; sipp++)
149 			*xipp++ = *sipp;
150 	for (sipp = set; *sipp; sipp++)
151 		*xipp++ = *sipp;
152 	*xipp = NULL;
153 	if (newsysinit)
154 		free(newsysinit, M_TEMP);
155 	newsysinit = newset;
156 }
157 
158 /*
159  * System startup; initialize the world, create process 0, mount root
160  * filesystem, and fork to create init and pagedaemon.  Most of the
161  * hard work is done in the lower-level initialization routines including
162  * startup(), which does memory initialization and autoconfiguration.
163  *
164  * This allows simple addition of new kernel subsystems that require
165  * boot time initialization.  It also allows substitution of subsystem
166  * (for instance, a scheduler, kernel profiler, or VM system) by object
167  * module.  Finally, it allows for optional "kernel threads".
168  */
169 void
170 mi_startup(framep)
171 	void *framep;
172 {
173 
174 	register struct sysinit **sipp;		/* system initialization*/
175 	register struct sysinit **xipp;		/* interior loop of sort*/
176 	register struct sysinit *save;		/* bubble*/
177 
178 	/*
179 	 * Copy the locore.s frame pointer for proc0, this is forked into
180 	 * all other processes.
181 	 */
182 	init_framep = framep;
183 
184 restart:
185 	/*
186 	 * Perform a bubble sort of the system initialization objects by
187 	 * their subsystem (primary key) and order (secondary key).
188 	 */
189 	for (sipp = sysinit; *sipp; sipp++) {
190 		for (xipp = sipp + 1; *xipp; xipp++) {
191 			if ((*sipp)->subsystem < (*xipp)->subsystem ||
192 			     ((*sipp)->subsystem == (*xipp)->subsystem &&
193 			      (*sipp)->order < (*xipp)->order))
194 				continue;	/* skip*/
195 			save = *sipp;
196 			*sipp = *xipp;
197 			*xipp = save;
198 		}
199 	}
200 
201 	/*
202 	 * Traverse the (now) ordered list of system initialization tasks.
203 	 * Perform each task, and continue on to the next task.
204 	 *
205 	 * The last item on the list is expected to be the scheduler,
206 	 * which will not return.
207 	 */
208 	for (sipp = sysinit; *sipp; sipp++) {
209 
210 		if ((*sipp)->subsystem == SI_SUB_DUMMY)
211 			continue;	/* skip dummy task(s)*/
212 
213 		if ((*sipp)->subsystem == SI_SUB_DONE)
214 			continue;
215 
216 		/* Call function */
217 		(*((*sipp)->func))((*sipp)->udata);
218 
219 		/* Check off the one we're just done */
220 		(*sipp)->subsystem = SI_SUB_DONE;
221 
222 		/* Check if we've installed more sysinit items via KLD */
223 		if (newsysinit != NULL) {
224 			if (sysinit != (struct sysinit **)sysinit_set.ls_items)
225 				free(sysinit, M_TEMP);
226 			sysinit = newsysinit;
227 			newsysinit = NULL;
228 			goto restart;
229 		}
230 	}
231 
232 	panic("Shouldn't get here!");
233 	/* NOTREACHED*/
234 }
235 
236 
237 /*
238  ***************************************************************************
239  ****
240  **** The following SYSINIT's belong elsewhere, but have not yet
241  **** been moved.
242  ****
243  ***************************************************************************
244  */
245 static void print_caddr_t __P((void *data));
246 static void
247 print_caddr_t(data)
248 	void *data;
249 {
250 	printf("%s", (char *)data);
251 }
252 SYSINIT(announce, SI_SUB_COPYRIGHT, SI_ORDER_FIRST, print_caddr_t, copyright)
253 
254 
255 /*
256  ***************************************************************************
257  ****
258  **** The two following SYSINT's are proc0 specific glue code.  I am not
259  **** convinced that they can not be safely combined, but their order of
260  **** operation has been maintained as the same as the original init_main.c
261  **** for right now.
262  ****
263  **** These probably belong in init_proc.c or kern_proc.c, since they
264  **** deal with proc0 (the fork template process).
265  ****
266  ***************************************************************************
267  */
268 /* ARGSUSED*/
269 static void proc0_init __P((void *dummy));
270 static void
271 proc0_init(dummy)
272 	void *dummy;
273 {
274 	register struct proc		*p;
275 	register struct filedesc0	*fdp;
276 	register unsigned i;
277 
278 	p = &proc0;
279 
280 	/*
281 	 * Initialize process and pgrp structures.
282 	 */
283 	procinit();
284 
285 	/*
286 	 * Initialize sleep queue hash table
287 	 */
288 	sleepinit();
289 
290 	/*
291 	 * additional VM structures
292 	 */
293 	vm_init2();
294 
295 	/*
296 	 * Create process 0 (the swapper).
297 	 */
298 	LIST_INSERT_HEAD(&allproc, p, p_list);
299 	p->p_pgrp = &pgrp0;
300 	LIST_INSERT_HEAD(PGRPHASH(0), &pgrp0, pg_hash);
301 	LIST_INIT(&pgrp0.pg_members);
302 	LIST_INSERT_HEAD(&pgrp0.pg_members, p, p_pglist);
303 
304 	pgrp0.pg_session = &session0;
305 	session0.s_count = 1;
306 	session0.s_leader = p;
307 
308 	p->p_sysent = &aout_sysvec;
309 
310 	p->p_flag = P_INMEM | P_SYSTEM;
311 	p->p_stat = SRUN;
312 	p->p_nice = NZERO;
313 	p->p_rtprio.type = RTP_PRIO_NORMAL;
314 	p->p_rtprio.prio = 0;
315 
316 /*
317  * Link for kernel based threads
318  */
319 	p->p_peers = 0;
320 	p->p_leader = p;
321 
322 	bcopy("swapper", p->p_comm, sizeof ("swapper"));
323 
324 	/* Create credentials. */
325 	cred0.p_refcnt = 1;
326 	p->p_cred = &cred0;
327 	p->p_ucred = crget();
328 	p->p_ucred->cr_ngroups = 1;	/* group 0 */
329 
330 	/* Don't jail it */
331 	p->p_prison = 0;
332 
333 	/* Create procsig. */
334 	p->p_procsig = &procsig0;
335 	p->p_procsig->ps_refcnt = 1;
336 
337 	/* Create the file descriptor table. */
338 	fdp = &filedesc0;
339 	p->p_fd = &fdp->fd_fd;
340 	fdp->fd_fd.fd_refcnt = 1;
341 	fdp->fd_fd.fd_cmask = cmask;
342 	fdp->fd_fd.fd_ofiles = fdp->fd_dfiles;
343 	fdp->fd_fd.fd_ofileflags = fdp->fd_dfileflags;
344 	fdp->fd_fd.fd_nfiles = NDFILE;
345 
346 	/* Create the limits structures. */
347 	p->p_limit = &limit0;
348 	for (i = 0; i < sizeof(p->p_rlimit)/sizeof(p->p_rlimit[0]); i++)
349 		limit0.pl_rlimit[i].rlim_cur =
350 		    limit0.pl_rlimit[i].rlim_max = RLIM_INFINITY;
351 	limit0.pl_rlimit[RLIMIT_NOFILE].rlim_cur =
352 	    limit0.pl_rlimit[RLIMIT_NOFILE].rlim_max = maxfiles;
353 	limit0.pl_rlimit[RLIMIT_NPROC].rlim_cur =
354 	    limit0.pl_rlimit[RLIMIT_NPROC].rlim_max = maxproc;
355 	i = ptoa(cnt.v_free_count);
356 	limit0.pl_rlimit[RLIMIT_RSS].rlim_max = i;
357 	limit0.pl_rlimit[RLIMIT_MEMLOCK].rlim_max = i;
358 	limit0.pl_rlimit[RLIMIT_MEMLOCK].rlim_cur = i / 3;
359 	limit0.p_cpulimit = RLIM_INFINITY;
360 	limit0.p_refcnt = 1;
361 
362 
363 	/* Allocate a prototype map so we have something to fork. */
364 	pmap_pinit0(vmspace_pmap(&vmspace0));
365 	p->p_vmspace = &vmspace0;
366 	vmspace0.vm_refcnt = 1;
367 	vm_map_init(&vmspace0.vm_map, round_page(VM_MIN_ADDRESS),
368 	    trunc_page(VM_MAXUSER_ADDRESS));
369 	vmspace0.vm_map.pmap = vmspace_pmap(&vmspace0);
370 	p->p_addr = proc0paddr;				/* XXX */
371 
372 #ifdef cpu_set_init_frame
373 	/*
374 	 * proc0 needs to have a coherent frame base in its stack.
375 	 */
376 	cpu_set_init_frame(p, init_framep);			/* XXX! */
377 #endif
378 
379 	/*
380 	 * We continue to place resource usage info and signal
381 	 * actions in the user struct so they're pageable.
382 	 */
383 	p->p_stats = &p->p_addr->u_stats;
384 	p->p_sigacts = &p->p_addr->u_sigacts;
385 
386 	/*
387 	 * Charge root for one process.
388 	 */
389 	(void)chgproccnt(0, 1);
390 
391 	/*
392 	 * Initialize the current process pointer (curproc) before
393 	 * any possible traps/probes to simplify trap processing.
394 	 */
395 	SET_CURPROC(p);
396 
397 }
398 SYSINIT(p0init, SI_SUB_INTRINSIC, SI_ORDER_FIRST, proc0_init, NULL)
399 
400 /* ARGSUSED*/
401 static void proc0_post __P((void *dummy));
402 static void
403 proc0_post(dummy)
404 	void *dummy;
405 {
406 	struct timespec ts;
407 
408 	/*
409 	 * Now we can look at the time, having had a chance to verify the
410 	 * time from the file system.  Pretend that proc0 started now.
411 	 */
412 	microtime(&proc0.p_stats->p_start);
413 	proc0.p_runtime = 0;
414 	microuptime(&switchtime);
415 	switchticks = ticks;
416 
417 	/*
418 	 * Give the ``random'' number generator a thump.
419 	 * XXX: Does read_random() contain enough bits to be used here ?
420 	 */
421 	nanotime(&ts);
422 	srandom(ts.tv_sec ^ ts.tv_nsec);
423 
424 	/* Initialize signal state for process 0. */
425 	siginit(&proc0);
426 }
427 SYSINIT(p0post, SI_SUB_INTRINSIC_POST, SI_ORDER_FIRST, proc0_post, NULL)
428 
429 
430 
431 
432 /*
433  ***************************************************************************
434  ****
435  **** The following SYSINIT's and glue code should be moved to the
436  **** respective files on a per subsystem basis.
437  ****
438  ***************************************************************************
439  */
440 
441 /* ARGSUSED*/
442 static void xxx_vfs_root_fdtab __P((void *dummy));
443 static void
444 xxx_vfs_root_fdtab(dummy)
445 	void *dummy;
446 {
447 	register struct filedesc0	*fdp = &filedesc0;
448 
449 	/* Get the vnode for '/'.  Set fdp->fd_fd.fd_cdir to reference it. */
450 	if (VFS_ROOT(TAILQ_FIRST(&mountlist), &rootvnode))
451 		panic("cannot find root vnode");
452 	fdp->fd_fd.fd_cdir = rootvnode;
453 	VREF(fdp->fd_fd.fd_cdir);
454 	VOP_UNLOCK(rootvnode, 0, &proc0);
455 	fdp->fd_fd.fd_rdir = rootvnode;
456 }
457 SYSINIT(retrofit, SI_SUB_ROOT_FDTAB, SI_ORDER_FIRST, xxx_vfs_root_fdtab, NULL)
458 
459 
460 /*
461  ***************************************************************************
462  ****
463  **** The following code probably belongs in another file, like
464  **** kern/init_init.c.  It is here for two reasons only:
465  ****
466  ****	1)	This code returns to startup the system; this is
467  ****		abnormal for a kernel thread.
468  ****	2)	This code promiscuously uses init_frame
469  ****
470  ***************************************************************************
471  */
472 
473 extern void prepare_usermode __P((void));
474 static void create_init __P((const void *dummy));
475 static void start_init __P((void *dummy));
476 SYSINIT(init,SI_SUB_KTHREAD_INIT, SI_ORDER_FIRST, create_init, NULL)
477 
478 /*
479  * Like kthread_create(), but runs in it's own address space.
480  */
481 static void
482 create_init(udata)
483 	const void *udata;
484 {
485 	int error;
486 
487 	error = fork1(&proc0, RFFDG | RFPROC, &initproc);
488 	if (error)
489 		panic("cannot fork init: %d\n", error);
490 	initproc->p_flag |= P_INMEM | P_SYSTEM;
491 	cpu_set_fork_handler(initproc, start_init, NULL);
492 }
493 
494 /*
495  * List of paths to try when searching for "init".
496  */
497 static char init_path[MAXPATHLEN] =
498 #ifdef	INIT_PATH
499     __XSTRING(INIT_PATH);
500 #else
501     "/sbin/init:/sbin/oinit:/sbin/init.bak:/stand/sysinstall";
502 #endif
503 SYSCTL_STRING(_kern, OID_AUTO, init_path, CTLFLAG_RD, init_path, 0, "");
504 
505 /*
506  * Start the initial user process; try exec'ing each pathname in init_path.
507  * The program is invoked with one argument containing the boot flags.
508  */
509 static void
510 start_init(dummy)
511 	void *dummy;
512 {
513 	vm_offset_t addr;
514 	struct execve_args args;
515 	int options, error;
516 	char *var, *path, *next, *s;
517 	char *ucp, **uap, *arg0, *arg1;
518 	struct proc *p;
519 
520 	p = curproc;
521 
522 	/*
523 	 * Need just enough stack to hold the faked-up "execve()" arguments.
524 	 */
525 	addr = trunc_page(USRSTACK - PAGE_SIZE);
526 	if (vm_map_find(&p->p_vmspace->vm_map, NULL, 0, &addr, PAGE_SIZE,
527 			FALSE, VM_PROT_ALL, VM_PROT_ALL, 0) != 0)
528 		panic("init: couldn't allocate argument space");
529 	p->p_vmspace->vm_maxsaddr = (caddr_t)addr;
530 	p->p_vmspace->vm_ssize = 1;
531 
532 	if ((var = getenv("init_path")) != NULL) {
533 		strncpy(init_path, var, sizeof init_path);
534 		init_path[sizeof init_path - 1] = 0;
535 	}
536 
537 	for (path = init_path; *path != '\0'; path = next) {
538 		while (*path == ':')
539 			path++;
540 		if (*path == '\0')
541 			break;
542 		for (next = path; *next != '\0' && *next != ':'; next++)
543 			/* nothing */ ;
544 		if (bootverbose)
545 			printf("start_init: trying %.*s\n", (int)(next - path),
546 			    path);
547 
548 		/*
549 		 * Move out the boot flag argument.
550 		 */
551 		options = 0;
552 		ucp = (char *)USRSTACK;
553 		(void)subyte(--ucp, 0);		/* trailing zero */
554 		if (boothowto & RB_SINGLE) {
555 			(void)subyte(--ucp, 's');
556 			options = 1;
557 		}
558 #ifdef notyet
559                 if (boothowto & RB_FASTBOOT) {
560 			(void)subyte(--ucp, 'f');
561 			options = 1;
562 		}
563 #endif
564 
565 #ifdef BOOTCDROM
566 		(void)subyte(--ucp, 'C');
567 		options = 1;
568 #endif
569 		if (options == 0)
570 			(void)subyte(--ucp, '-');
571 		(void)subyte(--ucp, '-');		/* leading hyphen */
572 		arg1 = ucp;
573 
574 		/*
575 		 * Move out the file name (also arg 0).
576 		 */
577 		(void)subyte(--ucp, 0);
578 		for (s = next - 1; s >= path; s--)
579 			(void)subyte(--ucp, *s);
580 		arg0 = ucp;
581 
582 		/*
583 		 * Move out the arg pointers.
584 		 */
585 		uap = (char **)((intptr_t)ucp & ~(sizeof(intptr_t)-1));
586 		(void)suword((caddr_t)--uap, (long)0);	/* terminator */
587 		(void)suword((caddr_t)--uap, (long)(intptr_t)arg1);
588 		(void)suword((caddr_t)--uap, (long)(intptr_t)arg0);
589 
590 		/*
591 		 * Point at the arguments.
592 		 */
593 		args.fname = arg0;
594 		args.argv = uap;
595 		args.envv = NULL;
596 
597 		/*
598 		 * Now try to exec the program.  If can't for any reason
599 		 * other than it doesn't exist, complain.
600 		 *
601 		 * Otherwise, return via the fork trampoline all the way
602 		 * to user mode as init!
603 		 */
604 		if ((error = execve(p, &args)) == 0) {
605 			prepare_usermode();
606 			return;
607 		}
608 		if (error != ENOENT)
609 			printf("exec %.*s: error %d\n", (int)(next - path),
610 			    path, error);
611 	}
612 	printf("init: not found in path %s\n", init_path);
613 	panic("no init");
614 }
615