xref: /freebsd/sys/kern/init_main.c (revision ce4946daa5ce852d28008dac492029500ab2ee95)
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/lock.h>
53 #include <sys/mount.h>
54 #include <sys/mutex.h>
55 #include <sys/sysctl.h>
56 #include <sys/proc.h>
57 #include <sys/resourcevar.h>
58 #include <sys/systm.h>
59 #include <sys/signalvar.h>
60 #include <sys/vnode.h>
61 #include <sys/sysent.h>
62 #include <sys/reboot.h>
63 #include <sys/sx.h>
64 #include <sys/sysproto.h>
65 #include <sys/vmmeter.h>
66 #include <sys/unistd.h>
67 #include <sys/malloc.h>
68 #include <sys/conf.h>
69 
70 #include <machine/cpu.h>
71 #include <machine/globals.h>
72 
73 #include <vm/vm.h>
74 #include <vm/vm_param.h>
75 #include <vm/pmap.h>
76 #include <vm/vm_map.h>
77 #include <sys/user.h>
78 #include <sys/copyright.h>
79 
80 extern struct linker_set	sysinit_set;	/* XXX */
81 
82 void mi_startup(void);				/* Should be elsewhere */
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 procsig procsig0;
90 static struct filedesc0 filedesc0;
91 static struct plimit limit0;
92 static struct vmspace vmspace0;
93 struct	proc *initproc;
94 
95 int cmask = CMASK;
96 extern	struct user *proc0paddr;
97 extern int fallback_elf_brand;
98 
99 struct	vnode *rootvp;
100 int	boothowto = 0;		/* initialized so that it can be patched */
101 SYSCTL_INT(_debug, OID_AUTO, boothowto, CTLFLAG_RD, &boothowto, 0, "");
102 
103 /*
104  * This ensures that there is at least one entry so that the sysinit_set
105  * symbol is not undefined.  A sybsystem ID of SI_SUB_DUMMY is never
106  * executed.
107  */
108 SYSINIT(placeholder, SI_SUB_DUMMY, SI_ORDER_ANY, NULL, NULL)
109 
110 /*
111  * The sysinit table itself.  Items are checked off as the are run.
112  * If we want to register new sysinit types, add them to newsysinit.
113  */
114 struct sysinit **sysinit = (struct sysinit **)sysinit_set.ls_items;
115 struct sysinit **newsysinit;
116 
117 /*
118  * Merge a new sysinit set into the current set, reallocating it if
119  * necessary.  This can only be called after malloc is running.
120  */
121 void
122 sysinit_add(struct sysinit **set)
123 {
124 	struct sysinit **newset;
125 	struct sysinit **sipp;
126 	struct sysinit **xipp;
127 	int count = 0;
128 
129 	if (newsysinit)
130 		for (sipp = newsysinit; *sipp; sipp++)
131 			count++;
132 	else
133 		for (sipp = sysinit; *sipp; sipp++)
134 			count++;
135 	for (sipp = set; *sipp; sipp++)
136 		count++;
137 	count++;		/* Trailing NULL */
138 	newset = malloc(count * sizeof(*sipp), M_TEMP, M_NOWAIT);
139 	if (newset == NULL)
140 		panic("cannot malloc for sysinit");
141 	xipp = newset;
142 	if (newsysinit)
143 		for (sipp = newsysinit; *sipp; sipp++)
144 			*xipp++ = *sipp;
145 	else
146 		for (sipp = sysinit; *sipp; sipp++)
147 			*xipp++ = *sipp;
148 	for (sipp = set; *sipp; sipp++)
149 		*xipp++ = *sipp;
150 	*xipp = NULL;
151 	if (newsysinit)
152 		free(newsysinit, M_TEMP);
153 	newsysinit = newset;
154 }
155 
156 /*
157  * System startup; initialize the world, create process 0, mount root
158  * filesystem, and fork to create init and pagedaemon.  Most of the
159  * hard work is done in the lower-level initialization routines including
160  * startup(), which does memory initialization and autoconfiguration.
161  *
162  * This allows simple addition of new kernel subsystems that require
163  * boot time initialization.  It also allows substitution of subsystem
164  * (for instance, a scheduler, kernel profiler, or VM system) by object
165  * module.  Finally, it allows for optional "kernel threads".
166  */
167 void
168 mi_startup(void)
169 {
170 
171 	register struct sysinit **sipp;		/* system initialization*/
172 	register struct sysinit **xipp;		/* interior loop of sort*/
173 	register struct sysinit *save;		/* bubble*/
174 
175 restart:
176 	/*
177 	 * Perform a bubble sort of the system initialization objects by
178 	 * their subsystem (primary key) and order (secondary key).
179 	 */
180 	for (sipp = sysinit; *sipp; sipp++) {
181 		for (xipp = sipp + 1; *xipp; xipp++) {
182 			if ((*sipp)->subsystem < (*xipp)->subsystem ||
183 			     ((*sipp)->subsystem == (*xipp)->subsystem &&
184 			      (*sipp)->order <= (*xipp)->order))
185 				continue;	/* skip*/
186 			save = *sipp;
187 			*sipp = *xipp;
188 			*xipp = save;
189 		}
190 	}
191 
192 	/*
193 	 * Traverse the (now) ordered list of system initialization tasks.
194 	 * Perform each task, and continue on to the next task.
195 	 *
196 	 * The last item on the list is expected to be the scheduler,
197 	 * which will not return.
198 	 */
199 	for (sipp = sysinit; *sipp; sipp++) {
200 
201 		if ((*sipp)->subsystem == SI_SUB_DUMMY)
202 			continue;	/* skip dummy task(s)*/
203 
204 		if ((*sipp)->subsystem == SI_SUB_DONE)
205 			continue;
206 
207 		/* Call function */
208 		(*((*sipp)->func))((*sipp)->udata);
209 
210 		/* Check off the one we're just done */
211 		(*sipp)->subsystem = SI_SUB_DONE;
212 
213 		/* Check if we've installed more sysinit items via KLD */
214 		if (newsysinit != NULL) {
215 			if (sysinit != (struct sysinit **)sysinit_set.ls_items)
216 				free(sysinit, M_TEMP);
217 			sysinit = newsysinit;
218 			newsysinit = NULL;
219 			goto restart;
220 		}
221 	}
222 
223 	panic("Shouldn't get here!");
224 	/* NOTREACHED*/
225 }
226 
227 
228 /*
229  ***************************************************************************
230  ****
231  **** The following SYSINIT's belong elsewhere, but have not yet
232  **** been moved.
233  ****
234  ***************************************************************************
235  */
236 static void
237 print_caddr_t(void *data __unused)
238 {
239 	printf("%s", (char *)data);
240 }
241 SYSINIT(announce, SI_SUB_COPYRIGHT, SI_ORDER_FIRST, print_caddr_t, copyright)
242 
243 /*
244  ***************************************************************************
245  ****
246  **** The two following SYSINT's are proc0 specific glue code.  I am not
247  **** convinced that they can not be safely combined, but their order of
248  **** operation has been maintained as the same as the original init_main.c
249  **** for right now.
250  ****
251  **** These probably belong in init_proc.c or kern_proc.c, since they
252  **** deal with proc0 (the fork template process).
253  ****
254  ***************************************************************************
255  */
256 /* ARGSUSED*/
257 static void
258 proc0_init(void *dummy __unused)
259 {
260 	register struct proc		*p;
261 	register struct filedesc0	*fdp;
262 	register unsigned i;
263 
264 	p = &proc0;
265 
266 	/*
267 	 * Initialize magic number.
268 	 */
269 	p->p_magic = P_MAGIC;
270 
271 	/*
272 	 * Initialize process and pgrp structures.
273 	 */
274 	procinit();
275 
276 	/*
277 	 * Initialize sleep queue hash table
278 	 */
279 	sleepinit();
280 
281 	/*
282 	 * additional VM structures
283 	 */
284 	vm_init2();
285 
286 	/*
287 	 * Create process 0 (the swapper).
288 	 */
289 	LIST_INSERT_HEAD(&allproc, p, p_list);
290 	LIST_INSERT_HEAD(PIDHASH(0), p, p_hash);
291 	p->p_pgrp = &pgrp0;
292 	LIST_INSERT_HEAD(PGRPHASH(0), &pgrp0, pg_hash);
293 	LIST_INIT(&pgrp0.pg_members);
294 	LIST_INSERT_HEAD(&pgrp0.pg_members, p, p_pglist);
295 
296 	pgrp0.pg_session = &session0;
297 	session0.s_count = 1;
298 	session0.s_leader = p;
299 
300 #ifdef __ELF__
301 	p->p_sysent = &elf_freebsd_sysvec;
302 #else
303 	p->p_sysent = &aout_sysvec;
304 #endif
305 
306 	p->p_flag = P_SYSTEM;
307 	p->p_sflag = PS_INMEM;
308 	p->p_stat = SRUN;
309 	p->p_nice = NZERO;
310 	p->p_pri.pri_class = PRI_TIMESHARE;
311 	p->p_pri.pri_level = PVM;
312 	p->p_pri.pri_native = PUSER;
313 	p->p_pri.pri_user = PUSER;
314 
315 	p->p_peers = 0;
316 	p->p_leader = p;
317 
318 	bcopy("swapper", p->p_comm, sizeof ("swapper"));
319 
320 	callout_init(&p->p_itcallout, 0);
321 	callout_init(&p->p_slpcallout, 1);
322 
323 	/* Create credentials. */
324 	cred0.p_refcnt = 1;
325 	cred0.p_uidinfo = uifind(0);
326 	p->p_cred = &cred0;
327 	p->p_ucred = crget();
328 	p->p_ucred->cr_ngroups = 1;	/* group 0 */
329 	p->p_ucred->cr_uidinfo = uifind(0);
330 	p->p_ucred->cr_prison = NULL;	/* Don't jail it. */
331 
332 	/* Create procsig. */
333 	p->p_procsig = &procsig0;
334 	p->p_procsig->ps_refcnt = 1;
335 
336 	/* Initialize signal state for process 0. */
337 	siginit(&proc0);
338 
339 	/* Create the file descriptor table. */
340 	fdp = &filedesc0;
341 	p->p_fd = &fdp->fd_fd;
342 	fdp->fd_fd.fd_refcnt = 1;
343 	fdp->fd_fd.fd_cmask = cmask;
344 	fdp->fd_fd.fd_ofiles = fdp->fd_dfiles;
345 	fdp->fd_fd.fd_ofileflags = fdp->fd_dfileflags;
346 	fdp->fd_fd.fd_nfiles = NDFILE;
347 
348 	/* Create the limits structures. */
349 	p->p_limit = &limit0;
350 	for (i = 0; i < sizeof(p->p_rlimit)/sizeof(p->p_rlimit[0]); i++)
351 		limit0.pl_rlimit[i].rlim_cur =
352 		    limit0.pl_rlimit[i].rlim_max = RLIM_INFINITY;
353 	limit0.pl_rlimit[RLIMIT_NOFILE].rlim_cur =
354 	    limit0.pl_rlimit[RLIMIT_NOFILE].rlim_max = maxfiles;
355 	limit0.pl_rlimit[RLIMIT_NPROC].rlim_cur =
356 	    limit0.pl_rlimit[RLIMIT_NPROC].rlim_max = maxproc;
357 	i = ptoa(cnt.v_free_count);
358 	limit0.pl_rlimit[RLIMIT_RSS].rlim_max = i;
359 	limit0.pl_rlimit[RLIMIT_MEMLOCK].rlim_max = i;
360 	limit0.pl_rlimit[RLIMIT_MEMLOCK].rlim_cur = i / 3;
361 	limit0.p_cpulimit = RLIM_INFINITY;
362 	limit0.p_refcnt = 1;
363 
364 	/* Allocate a prototype map so we have something to fork. */
365 	pmap_pinit0(vmspace_pmap(&vmspace0));
366 	p->p_vmspace = &vmspace0;
367 	vmspace0.vm_refcnt = 1;
368 	vm_map_init(&vmspace0.vm_map, round_page(VM_MIN_ADDRESS),
369 	    trunc_page(VM_MAXUSER_ADDRESS));
370 	vmspace0.vm_map.pmap = vmspace_pmap(&vmspace0);
371 	p->p_addr = proc0paddr;				/* XXX */
372 
373 	/*
374 	 * We continue to place resource usage info and signal
375 	 * actions in the user struct so they're pageable.
376 	 */
377 	p->p_stats = &p->p_addr->u_stats;
378 	p->p_sigacts = &p->p_addr->u_sigacts;
379 
380 	/*
381 	 * Charge root for one process.
382 	 */
383 	(void)chgproccnt(cred0.p_uidinfo, 1, 0);
384 }
385 SYSINIT(p0init, SI_SUB_INTRINSIC, SI_ORDER_FIRST, proc0_init, NULL)
386 
387 /* ARGSUSED*/
388 static void
389 proc0_post(void *dummy __unused)
390 {
391 	struct timespec ts;
392 	struct proc *p;
393 
394 	/*
395 	 * Now we can look at the time, having had a chance to verify the
396 	 * time from the file system.  Pretend that proc0 started now.
397 	 */
398 	sx_slock(&allproc_lock);
399 	LIST_FOREACH(p, &allproc, p_list) {
400 		microtime(&p->p_stats->p_start);
401 		p->p_runtime = 0;
402 	}
403 	sx_sunlock(&allproc_lock);
404 	microuptime(PCPU_PTR(switchtime));
405 	PCPU_SET(switchticks, ticks);
406 
407 	/*
408 	 * Give the ``random'' number generator a thump.
409 	 */
410 	nanotime(&ts);
411 	srandom(ts.tv_sec ^ ts.tv_nsec);
412 }
413 SYSINIT(p0post, SI_SUB_INTRINSIC_POST, SI_ORDER_FIRST, proc0_post, NULL)
414 
415 /*
416  ***************************************************************************
417  ****
418  **** The following SYSINIT's and glue code should be moved to the
419  **** respective files on a per subsystem basis.
420  ****
421  ***************************************************************************
422  */
423 
424 
425 /*
426  ***************************************************************************
427  ****
428  **** The following code probably belongs in another file, like
429  **** kern/init_init.c.
430  ****
431  ***************************************************************************
432  */
433 
434 /*
435  * List of paths to try when searching for "init".
436  */
437 static char init_path[MAXPATHLEN] =
438 #ifdef	INIT_PATH
439     __XSTRING(INIT_PATH);
440 #else
441     "/sbin/init:/sbin/oinit:/sbin/init.bak:/stand/sysinstall";
442 #endif
443 SYSCTL_STRING(_kern, OID_AUTO, init_path, CTLFLAG_RD, init_path, 0, "");
444 
445 /*
446  * Start the initial user process; try exec'ing each pathname in init_path.
447  * The program is invoked with one argument containing the boot flags.
448  */
449 static void
450 start_init(void *dummy)
451 {
452 	vm_offset_t addr;
453 	struct execve_args args;
454 	int options, error;
455 	char *var, *path, *next, *s;
456 	char *ucp, **uap, *arg0, *arg1;
457 	struct proc *p;
458 
459 	mtx_lock(&Giant);
460 
461 	p = curproc;
462 
463 	/* Get the vnode for '/'.  Set p->p_fd->fd_cdir to reference it. */
464 	if (VFS_ROOT(TAILQ_FIRST(&mountlist), &rootvnode))
465 		panic("cannot find root vnode");
466 	p->p_fd->fd_cdir = rootvnode;
467 	VREF(p->p_fd->fd_cdir);
468 	p->p_fd->fd_rdir = rootvnode;
469 	VREF(p->p_fd->fd_rdir);
470 	VOP_UNLOCK(rootvnode, 0, p);
471 
472 	/*
473 	 * Need just enough stack to hold the faked-up "execve()" arguments.
474 	 */
475 	addr = trunc_page(USRSTACK - PAGE_SIZE);
476 	if (vm_map_find(&p->p_vmspace->vm_map, NULL, 0, &addr, PAGE_SIZE,
477 			FALSE, VM_PROT_ALL, VM_PROT_ALL, 0) != 0)
478 		panic("init: couldn't allocate argument space");
479 	p->p_vmspace->vm_maxsaddr = (caddr_t)addr;
480 	p->p_vmspace->vm_ssize = 1;
481 
482 	if ((var = getenv("init_path")) != NULL) {
483 		strncpy(init_path, var, sizeof init_path);
484 		init_path[sizeof init_path - 1] = 0;
485 	}
486 	if ((var = getenv("kern.fallback_elf_brand")) != NULL)
487 		fallback_elf_brand = strtol(var, NULL, 0);
488 
489 	for (path = init_path; *path != '\0'; path = next) {
490 		while (*path == ':')
491 			path++;
492 		if (*path == '\0')
493 			break;
494 		for (next = path; *next != '\0' && *next != ':'; next++)
495 			/* nothing */ ;
496 		if (bootverbose)
497 			printf("start_init: trying %.*s\n", (int)(next - path),
498 			    path);
499 
500 		/*
501 		 * Move out the boot flag argument.
502 		 */
503 		options = 0;
504 		ucp = (char *)USRSTACK;
505 		(void)subyte(--ucp, 0);		/* trailing zero */
506 		if (boothowto & RB_SINGLE) {
507 			(void)subyte(--ucp, 's');
508 			options = 1;
509 		}
510 #ifdef notyet
511                 if (boothowto & RB_FASTBOOT) {
512 			(void)subyte(--ucp, 'f');
513 			options = 1;
514 		}
515 #endif
516 
517 #ifdef BOOTCDROM
518 		(void)subyte(--ucp, 'C');
519 		options = 1;
520 #endif
521 		if (devfs_present) {
522 			(void)subyte(--ucp, 'd');
523 			options = 1;
524 		}
525 
526 		if (options == 0)
527 			(void)subyte(--ucp, '-');
528 		(void)subyte(--ucp, '-');		/* leading hyphen */
529 		arg1 = ucp;
530 
531 		/*
532 		 * Move out the file name (also arg 0).
533 		 */
534 		(void)subyte(--ucp, 0);
535 		for (s = next - 1; s >= path; s--)
536 			(void)subyte(--ucp, *s);
537 		arg0 = ucp;
538 
539 		/*
540 		 * Move out the arg pointers.
541 		 */
542 		uap = (char **)((intptr_t)ucp & ~(sizeof(intptr_t)-1));
543 		(void)suword((caddr_t)--uap, (long)0);	/* terminator */
544 		(void)suword((caddr_t)--uap, (long)(intptr_t)arg1);
545 		(void)suword((caddr_t)--uap, (long)(intptr_t)arg0);
546 
547 		/*
548 		 * Point at the arguments.
549 		 */
550 		args.fname = arg0;
551 		args.argv = uap;
552 		args.envv = NULL;
553 
554 		/*
555 		 * Now try to exec the program.  If can't for any reason
556 		 * other than it doesn't exist, complain.
557 		 *
558 		 * Otherwise, return via fork_trampoline() all the way
559 		 * to user mode as init!
560 		 */
561 		if ((error = execve(p, &args)) == 0) {
562 			mtx_unlock(&Giant);
563 			return;
564 		}
565 		if (error != ENOENT)
566 			printf("exec %.*s: error %d\n", (int)(next - path),
567 			    path, error);
568 	}
569 	printf("init: not found in path %s\n", init_path);
570 	panic("no init");
571 }
572 
573 /*
574  * Like kthread_create(), but runs in it's own address space.
575  * We do this early to reserve pid 1.
576  *
577  * Note special case - do not make it runnable yet.  Other work
578  * in progress will change this more.
579  */
580 static void
581 create_init(const void *udata __unused)
582 {
583 	int error;
584 
585 	error = fork1(&proc0, RFFDG | RFPROC | RFSTOPPED, &initproc);
586 	if (error)
587 		panic("cannot fork init: %d\n", error);
588 	PROC_LOCK(initproc);
589 	initproc->p_flag |= P_SYSTEM;
590 	PROC_UNLOCK(initproc);
591 	mtx_lock_spin(&sched_lock);
592 	initproc->p_sflag |= PS_INMEM;
593 	mtx_unlock_spin(&sched_lock);
594 	cpu_set_fork_handler(initproc, start_init, NULL);
595 }
596 SYSINIT(init, SI_SUB_CREATE_INIT, SI_ORDER_FIRST, create_init, NULL)
597 
598 /*
599  * Make it runnable now.
600  */
601 static void
602 kick_init(const void *udata __unused)
603 {
604 
605 	mtx_lock_spin(&sched_lock);
606 	initproc->p_stat = SRUN;
607 	setrunqueue(initproc);
608 	mtx_unlock_spin(&sched_lock);
609 }
610 SYSINIT(kickinit, SI_SUB_KTHREAD_INIT, SI_ORDER_FIRST, kick_init, NULL)
611