xref: /freebsd/sys/kern/init_main.c (revision 6de306ecee3831f48debaad1d0b22418faa48e10)
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/ktr.h>
52 #include <sys/mount.h>
53 #include <sys/mutex.h>
54 #include <sys/sysctl.h>
55 #include <sys/proc.h>
56 #include <sys/resourcevar.h>
57 #include <sys/systm.h>
58 #include <sys/signalvar.h>
59 #include <sys/vnode.h>
60 #include <sys/sysent.h>
61 #include <sys/reboot.h>
62 #include <sys/sysproto.h>
63 #include <sys/vmmeter.h>
64 #include <sys/unistd.h>
65 #include <sys/malloc.h>
66 #include <sys/conf.h>
67 
68 #include <machine/cpu.h>
69 #include <machine/globals.h>
70 
71 #include <vm/vm.h>
72 #include <vm/vm_param.h>
73 #include <sys/lock.h>
74 #include <vm/pmap.h>
75 #include <vm/vm_map.h>
76 #include <sys/user.h>
77 #include <sys/copyright.h>
78 
79 extern struct linker_set	sysinit_set;	/* XXX */
80 
81 void mi_startup(void);				/* Should be elsewhere */
82 
83 /* Components of the first process -- never freed. */
84 static struct session session0;
85 static struct pgrp pgrp0;
86 struct	proc proc0;
87 static struct pcred cred0;
88 static struct procsig procsig0;
89 static struct filedesc0 filedesc0;
90 static struct plimit limit0;
91 static struct vmspace vmspace0;
92 struct	proc *initproc;
93 
94 int cmask = CMASK;
95 extern	struct user *proc0paddr;
96 
97 struct	vnode *rootvp;
98 int	boothowto = 0;		/* initialized so that it can be patched */
99 SYSCTL_INT(_debug, OID_AUTO, boothowto, CTLFLAG_RD, &boothowto, 0, "");
100 
101 /*
102  * This ensures that there is at least one entry so that the sysinit_set
103  * symbol is not undefined.  A sybsystem ID of SI_SUB_DUMMY is never
104  * executed.
105  */
106 SYSINIT(placeholder, SI_SUB_DUMMY, SI_ORDER_ANY, NULL, NULL)
107 
108 /*
109  * The sysinit table itself.  Items are checked off as the are run.
110  * If we want to register new sysinit types, add them to newsysinit.
111  */
112 struct sysinit **sysinit = (struct sysinit **)sysinit_set.ls_items;
113 struct sysinit **newsysinit;
114 
115 /*
116  * Merge a new sysinit set into the current set, reallocating it if
117  * necessary.  This can only be called after malloc is running.
118  */
119 void
120 sysinit_add(struct sysinit **set)
121 {
122 	struct sysinit **newset;
123 	struct sysinit **sipp;
124 	struct sysinit **xipp;
125 	int count = 0;
126 
127 	if (newsysinit)
128 		for (sipp = newsysinit; *sipp; sipp++)
129 			count++;
130 	else
131 		for (sipp = sysinit; *sipp; sipp++)
132 			count++;
133 	for (sipp = set; *sipp; sipp++)
134 		count++;
135 	count++;		/* Trailing NULL */
136 	newset = malloc(count * sizeof(*sipp), M_TEMP, M_NOWAIT);
137 	if (newset == NULL)
138 		panic("cannot malloc for sysinit");
139 	xipp = newset;
140 	if (newsysinit)
141 		for (sipp = newsysinit; *sipp; sipp++)
142 			*xipp++ = *sipp;
143 	else
144 		for (sipp = sysinit; *sipp; sipp++)
145 			*xipp++ = *sipp;
146 	for (sipp = set; *sipp; sipp++)
147 		*xipp++ = *sipp;
148 	*xipp = NULL;
149 	if (newsysinit)
150 		free(newsysinit, M_TEMP);
151 	newsysinit = newset;
152 }
153 
154 /*
155  * System startup; initialize the world, create process 0, mount root
156  * filesystem, and fork to create init and pagedaemon.  Most of the
157  * hard work is done in the lower-level initialization routines including
158  * startup(), which does memory initialization and autoconfiguration.
159  *
160  * This allows simple addition of new kernel subsystems that require
161  * boot time initialization.  It also allows substitution of subsystem
162  * (for instance, a scheduler, kernel profiler, or VM system) by object
163  * module.  Finally, it allows for optional "kernel threads".
164  */
165 void
166 mi_startup(void)
167 {
168 
169 	register struct sysinit **sipp;		/* system initialization*/
170 	register struct sysinit **xipp;		/* interior loop of sort*/
171 	register struct sysinit *save;		/* bubble*/
172 
173 restart:
174 	/*
175 	 * Perform a bubble sort of the system initialization objects by
176 	 * their subsystem (primary key) and order (secondary key).
177 	 */
178 	for (sipp = sysinit; *sipp; sipp++) {
179 		for (xipp = sipp + 1; *xipp; xipp++) {
180 			if ((*sipp)->subsystem < (*xipp)->subsystem ||
181 			     ((*sipp)->subsystem == (*xipp)->subsystem &&
182 			      (*sipp)->order <= (*xipp)->order))
183 				continue;	/* skip*/
184 			save = *sipp;
185 			*sipp = *xipp;
186 			*xipp = save;
187 		}
188 	}
189 
190 	/*
191 	 * Traverse the (now) ordered list of system initialization tasks.
192 	 * Perform each task, and continue on to the next task.
193 	 *
194 	 * The last item on the list is expected to be the scheduler,
195 	 * which will not return.
196 	 */
197 	for (sipp = sysinit; *sipp; sipp++) {
198 
199 		if ((*sipp)->subsystem == SI_SUB_DUMMY)
200 			continue;	/* skip dummy task(s)*/
201 
202 		if ((*sipp)->subsystem == SI_SUB_DONE)
203 			continue;
204 
205 		/* Call function */
206 		(*((*sipp)->func))((*sipp)->udata);
207 
208 		/* Check off the one we're just done */
209 		(*sipp)->subsystem = SI_SUB_DONE;
210 
211 		/* Check if we've installed more sysinit items via KLD */
212 		if (newsysinit != NULL) {
213 			if (sysinit != (struct sysinit **)sysinit_set.ls_items)
214 				free(sysinit, M_TEMP);
215 			sysinit = newsysinit;
216 			newsysinit = NULL;
217 			goto restart;
218 		}
219 	}
220 
221 	panic("Shouldn't get here!");
222 	/* NOTREACHED*/
223 }
224 
225 
226 /*
227  ***************************************************************************
228  ****
229  **** The following SYSINIT's belong elsewhere, but have not yet
230  **** been moved.
231  ****
232  ***************************************************************************
233  */
234 static void
235 print_caddr_t(void *data __unused)
236 {
237 	printf("%s", (char *)data);
238 }
239 SYSINIT(announce, SI_SUB_COPYRIGHT, SI_ORDER_FIRST, print_caddr_t, copyright)
240 
241 /*
242  ***************************************************************************
243  ****
244  **** The two following SYSINT's are proc0 specific glue code.  I am not
245  **** convinced that they can not be safely combined, but their order of
246  **** operation has been maintained as the same as the original init_main.c
247  **** for right now.
248  ****
249  **** These probably belong in init_proc.c or kern_proc.c, since they
250  **** deal with proc0 (the fork template process).
251  ****
252  ***************************************************************************
253  */
254 /* ARGSUSED*/
255 static void
256 proc0_init(void *dummy __unused)
257 {
258 	register struct proc		*p;
259 	register struct filedesc0	*fdp;
260 	register unsigned i;
261 
262 	p = &proc0;
263 
264 	/*
265 	 * Initialize magic number.
266 	 */
267 	p->p_magic = P_MAGIC;
268 
269 	/*
270 	 * Initialize process and pgrp structures.
271 	 */
272 	procinit();
273 
274 	/*
275 	 * Initialize sleep queue hash table
276 	 */
277 	sleepinit();
278 
279 	/*
280 	 * additional VM structures
281 	 */
282 	vm_init2();
283 
284 	/*
285 	 * Create process 0 (the swapper).
286 	 */
287 	LIST_INSERT_HEAD(&allproc, p, p_list);
288 	p->p_pgrp = &pgrp0;
289 	LIST_INSERT_HEAD(PGRPHASH(0), &pgrp0, pg_hash);
290 	LIST_INIT(&pgrp0.pg_members);
291 	LIST_INSERT_HEAD(&pgrp0.pg_members, p, p_pglist);
292 
293 	pgrp0.pg_session = &session0;
294 	session0.s_count = 1;
295 	session0.s_leader = p;
296 
297 #ifdef __ELF__
298 	p->p_sysent = &elf_freebsd_sysvec;
299 #else
300 	p->p_sysent = &aout_sysvec;
301 #endif
302 
303 	p->p_flag = P_SYSTEM;
304 	p->p_sflag = PS_INMEM;
305 	p->p_stat = SRUN;
306 	p->p_nice = NZERO;
307 	p->p_pri.pri_class = PRI_TIMESHARE;
308 	p->p_pri.pri_level = PVM;
309 	p->p_pri.pri_user = PUSER;
310 
311 	p->p_peers = 0;
312 	p->p_leader = p;
313 
314 	bcopy("swapper", p->p_comm, sizeof ("swapper"));
315 
316 	callout_init(&p->p_itcallout, 0);
317 	callout_init(&p->p_slpcallout, 1);
318 
319 	/* Create credentials. */
320 	cred0.p_refcnt = 1;
321 	cred0.p_uidinfo = uifind(0);
322 	p->p_cred = &cred0;
323 	p->p_ucred = crget();
324 	p->p_ucred->cr_ngroups = 1;	/* group 0 */
325 	p->p_ucred->cr_uidinfo = uifind(0);
326 	p->p_ucred->cr_prison = NULL;	/* Don't jail it. */
327 
328 	/* Create procsig. */
329 	p->p_procsig = &procsig0;
330 	p->p_procsig->ps_refcnt = 1;
331 
332 	/* Initialize signal state for process 0. */
333 	siginit(&proc0);
334 
335 	/* Create the file descriptor table. */
336 	fdp = &filedesc0;
337 	p->p_fd = &fdp->fd_fd;
338 	fdp->fd_fd.fd_refcnt = 1;
339 	fdp->fd_fd.fd_cmask = cmask;
340 	fdp->fd_fd.fd_ofiles = fdp->fd_dfiles;
341 	fdp->fd_fd.fd_ofileflags = fdp->fd_dfileflags;
342 	fdp->fd_fd.fd_nfiles = NDFILE;
343 
344 	/* Create the limits structures. */
345 	p->p_limit = &limit0;
346 	for (i = 0; i < sizeof(p->p_rlimit)/sizeof(p->p_rlimit[0]); i++)
347 		limit0.pl_rlimit[i].rlim_cur =
348 		    limit0.pl_rlimit[i].rlim_max = RLIM_INFINITY;
349 	limit0.pl_rlimit[RLIMIT_NOFILE].rlim_cur =
350 	    limit0.pl_rlimit[RLIMIT_NOFILE].rlim_max = maxfiles;
351 	limit0.pl_rlimit[RLIMIT_NPROC].rlim_cur =
352 	    limit0.pl_rlimit[RLIMIT_NPROC].rlim_max = maxproc;
353 	i = ptoa(cnt.v_free_count);
354 	limit0.pl_rlimit[RLIMIT_RSS].rlim_max = i;
355 	limit0.pl_rlimit[RLIMIT_MEMLOCK].rlim_max = i;
356 	limit0.pl_rlimit[RLIMIT_MEMLOCK].rlim_cur = i / 3;
357 	limit0.p_cpulimit = RLIM_INFINITY;
358 	limit0.p_refcnt = 1;
359 
360 	/* Allocate a prototype map so we have something to fork. */
361 	pmap_pinit0(vmspace_pmap(&vmspace0));
362 	p->p_vmspace = &vmspace0;
363 	vmspace0.vm_refcnt = 1;
364 	vm_map_init(&vmspace0.vm_map, round_page(VM_MIN_ADDRESS),
365 	    trunc_page(VM_MAXUSER_ADDRESS));
366 	vmspace0.vm_map.pmap = vmspace_pmap(&vmspace0);
367 	p->p_addr = proc0paddr;				/* XXX */
368 
369 	/*
370 	 * We continue to place resource usage info and signal
371 	 * actions in the user struct so they're pageable.
372 	 */
373 	p->p_stats = &p->p_addr->u_stats;
374 	p->p_sigacts = &p->p_addr->u_sigacts;
375 
376 	/*
377 	 * Charge root for one process.
378 	 */
379 	(void)chgproccnt(cred0.p_uidinfo, 1, 0);
380 }
381 SYSINIT(p0init, SI_SUB_INTRINSIC, SI_ORDER_FIRST, proc0_init, NULL)
382 
383 /* ARGSUSED*/
384 static void
385 proc0_post(void *dummy __unused)
386 {
387 	struct timespec ts;
388 	struct proc *p;
389 
390 	/*
391 	 * Now we can look at the time, having had a chance to verify the
392 	 * time from the file system.  Pretend that proc0 started now.
393 	 */
394 	ALLPROC_LOCK(AP_SHARED);
395 	LIST_FOREACH(p, &allproc, p_list) {
396 		microtime(&p->p_stats->p_start);
397 		p->p_runtime = 0;
398 	}
399 	ALLPROC_LOCK(AP_RELEASE);
400 	microuptime(PCPU_PTR(switchtime));
401 	PCPU_SET(switchticks, ticks);
402 
403 	/*
404 	 * Give the ``random'' number generator a thump.
405 	 */
406 	nanotime(&ts);
407 	srandom(ts.tv_sec ^ ts.tv_nsec);
408 }
409 SYSINIT(p0post, SI_SUB_INTRINSIC_POST, SI_ORDER_FIRST, proc0_post, NULL)
410 
411 /*
412  ***************************************************************************
413  ****
414  **** The following SYSINIT's and glue code should be moved to the
415  **** respective files on a per subsystem basis.
416  ****
417  ***************************************************************************
418  */
419 
420 
421 /*
422  ***************************************************************************
423  ****
424  **** The following code probably belongs in another file, like
425  **** kern/init_init.c.
426  ****
427  ***************************************************************************
428  */
429 
430 /*
431  * List of paths to try when searching for "init".
432  */
433 static char init_path[MAXPATHLEN] =
434 #ifdef	INIT_PATH
435     __XSTRING(INIT_PATH);
436 #else
437     "/sbin/init:/sbin/oinit:/sbin/init.bak:/stand/sysinstall";
438 #endif
439 SYSCTL_STRING(_kern, OID_AUTO, init_path, CTLFLAG_RD, init_path, 0, "");
440 
441 /*
442  * Start the initial user process; try exec'ing each pathname in init_path.
443  * The program is invoked with one argument containing the boot flags.
444  */
445 static void
446 start_init(void *dummy)
447 {
448 	vm_offset_t addr;
449 	struct execve_args args;
450 	int options, error;
451 	char *var, *path, *next, *s;
452 	char *ucp, **uap, *arg0, *arg1;
453 	struct proc *p;
454 
455 	mtx_lock(&Giant);
456 
457 	p = curproc;
458 
459 	/* Get the vnode for '/'.  Set p->p_fd->fd_cdir to reference it. */
460 	if (VFS_ROOT(TAILQ_FIRST(&mountlist), &rootvnode))
461 		panic("cannot find root vnode");
462 	p->p_fd->fd_cdir = rootvnode;
463 	VREF(p->p_fd->fd_cdir);
464 	p->p_fd->fd_rdir = rootvnode;
465 	VOP_UNLOCK(rootvnode, 0, p);
466 
467 	/*
468 	 * Need just enough stack to hold the faked-up "execve()" arguments.
469 	 */
470 	addr = trunc_page(USRSTACK - PAGE_SIZE);
471 	if (vm_map_find(&p->p_vmspace->vm_map, NULL, 0, &addr, PAGE_SIZE,
472 			FALSE, VM_PROT_ALL, VM_PROT_ALL, 0) != 0)
473 		panic("init: couldn't allocate argument space");
474 	p->p_vmspace->vm_maxsaddr = (caddr_t)addr;
475 	p->p_vmspace->vm_ssize = 1;
476 
477 	if ((var = getenv("init_path")) != NULL) {
478 		strncpy(init_path, var, sizeof init_path);
479 		init_path[sizeof init_path - 1] = 0;
480 	}
481 
482 	for (path = init_path; *path != '\0'; path = next) {
483 		while (*path == ':')
484 			path++;
485 		if (*path == '\0')
486 			break;
487 		for (next = path; *next != '\0' && *next != ':'; next++)
488 			/* nothing */ ;
489 		if (bootverbose)
490 			printf("start_init: trying %.*s\n", (int)(next - path),
491 			    path);
492 
493 		/*
494 		 * Move out the boot flag argument.
495 		 */
496 		options = 0;
497 		ucp = (char *)USRSTACK;
498 		(void)subyte(--ucp, 0);		/* trailing zero */
499 		if (boothowto & RB_SINGLE) {
500 			(void)subyte(--ucp, 's');
501 			options = 1;
502 		}
503 #ifdef notyet
504                 if (boothowto & RB_FASTBOOT) {
505 			(void)subyte(--ucp, 'f');
506 			options = 1;
507 		}
508 #endif
509 
510 #ifdef BOOTCDROM
511 		(void)subyte(--ucp, 'C');
512 		options = 1;
513 #endif
514 		if (devfs_present) {
515 			(void)subyte(--ucp, 'd');
516 			options = 1;
517 		}
518 
519 		if (options == 0)
520 			(void)subyte(--ucp, '-');
521 		(void)subyte(--ucp, '-');		/* leading hyphen */
522 		arg1 = ucp;
523 
524 		/*
525 		 * Move out the file name (also arg 0).
526 		 */
527 		(void)subyte(--ucp, 0);
528 		for (s = next - 1; s >= path; s--)
529 			(void)subyte(--ucp, *s);
530 		arg0 = ucp;
531 
532 		/*
533 		 * Move out the arg pointers.
534 		 */
535 		uap = (char **)((intptr_t)ucp & ~(sizeof(intptr_t)-1));
536 		(void)suword((caddr_t)--uap, (long)0);	/* terminator */
537 		(void)suword((caddr_t)--uap, (long)(intptr_t)arg1);
538 		(void)suword((caddr_t)--uap, (long)(intptr_t)arg0);
539 
540 		/*
541 		 * Point at the arguments.
542 		 */
543 		args.fname = arg0;
544 		args.argv = uap;
545 		args.envv = NULL;
546 
547 		/*
548 		 * Now try to exec the program.  If can't for any reason
549 		 * other than it doesn't exist, complain.
550 		 *
551 		 * Otherwise, return via fork_trampoline() all the way
552 		 * to user mode as init!
553 		 */
554 		if ((error = execve(p, &args)) == 0) {
555 			mtx_unlock(&Giant);
556 			return;
557 		}
558 		if (error != ENOENT)
559 			printf("exec %.*s: error %d\n", (int)(next - path),
560 			    path, error);
561 	}
562 	printf("init: not found in path %s\n", init_path);
563 	panic("no init");
564 }
565 
566 /*
567  * Like kthread_create(), but runs in it's own address space.
568  * We do this early to reserve pid 1.
569  *
570  * Note special case - do not make it runnable yet.  Other work
571  * in progress will change this more.
572  */
573 static void
574 create_init(const void *udata __unused)
575 {
576 	int error;
577 
578 	error = fork1(&proc0, RFFDG | RFPROC | RFSTOPPED, &initproc);
579 	if (error)
580 		panic("cannot fork init: %d\n", error);
581 	PROC_LOCK(initproc);
582 	initproc->p_flag |= P_SYSTEM;
583 	PROC_UNLOCK(initproc);
584 	mtx_lock_spin(&sched_lock);
585 	initproc->p_sflag |= PS_INMEM;
586 	mtx_unlock_spin(&sched_lock);
587 	cpu_set_fork_handler(initproc, start_init, NULL);
588 }
589 SYSINIT(init, SI_SUB_CREATE_INIT, SI_ORDER_FIRST, create_init, NULL)
590 
591 /*
592  * Make it runnable now.
593  */
594 static void
595 kick_init(const void *udata __unused)
596 {
597 
598 	mtx_lock_spin(&sched_lock);
599 	initproc->p_stat = SRUN;
600 	setrunqueue(initproc);
601 	mtx_unlock_spin(&sched_lock);
602 }
603 SYSINIT(kickinit, SI_SUB_KTHREAD_INIT, SI_ORDER_FIRST, kick_init, NULL)
604