xref: /freebsd/sys/kern/init_main.c (revision 77a0943ded95b9e6438f7db70c4a28e4d93946d4)
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  ****
245  **** The two following SYSINT's are proc0 specific glue code.  I am not
246  **** convinced that they can not be safely combined, but their order of
247  **** operation has been maintained as the same as the original init_main.c
248  **** for right now.
249  ****
250  **** These probably belong in init_proc.c or kern_proc.c, since they
251  **** deal with proc0 (the fork template process).
252  ****
253  ***************************************************************************
254  */
255 /* ARGSUSED*/
256 static void
257 proc0_init(void *dummy __unused)
258 {
259 	register struct proc		*p;
260 	register struct filedesc0	*fdp;
261 	register unsigned i;
262 
263 	p = &proc0;
264 
265 	/*
266 	 * Initialize magic number.
267 	 */
268 	p->p_magic = P_MAGIC;
269 
270 	/*
271 	 * Initialize process and pgrp structures.
272 	 */
273 	procinit();
274 
275 	/*
276 	 * Initialize sleep queue hash table
277 	 */
278 	sleepinit();
279 
280 	/*
281 	 * additional VM structures
282 	 */
283 	vm_init2();
284 
285 	/*
286 	 * Create process 0 (the swapper).
287 	 */
288 	LIST_INSERT_HEAD(&allproc, p, p_list);
289 	p->p_pgrp = &pgrp0;
290 	LIST_INSERT_HEAD(PGRPHASH(0), &pgrp0, pg_hash);
291 	LIST_INIT(&pgrp0.pg_members);
292 	LIST_INSERT_HEAD(&pgrp0.pg_members, p, p_pglist);
293 
294 	pgrp0.pg_session = &session0;
295 	session0.s_count = 1;
296 	session0.s_leader = p;
297 
298 #ifdef __ELF__
299 	p->p_sysent = &elf_freebsd_sysvec;
300 #else
301 	p->p_sysent = &aout_sysvec;
302 #endif
303 
304 	p->p_flag = P_INMEM | P_SYSTEM;
305 	p->p_stat = SRUN;
306 	p->p_nice = NZERO;
307 	p->p_rtprio.type = RTP_PRIO_NORMAL;
308 	p->p_rtprio.prio = 0;
309 
310 	p->p_peers = 0;
311 	p->p_leader = p;
312 
313 	bcopy("swapper", p->p_comm, sizeof ("swapper"));
314 
315 	/* Create credentials. */
316 	cred0.p_refcnt = 1;
317 	cred0.p_uidinfo = uifind(0);
318 	p->p_cred = &cred0;
319 	p->p_ucred = crget();
320 	p->p_ucred->cr_ngroups = 1;	/* group 0 */
321 	p->p_ucred->cr_uidinfo = uifind(0);
322 
323 	/* Don't jail it */
324 	p->p_prison = 0;
325 
326 	/* Create procsig. */
327 	p->p_procsig = &procsig0;
328 	p->p_procsig->ps_refcnt = 1;
329 
330 	/* Initialize signal state for process 0. */
331 	siginit(&proc0);
332 
333 	/* Create the file descriptor table. */
334 	fdp = &filedesc0;
335 	p->p_fd = &fdp->fd_fd;
336 	fdp->fd_fd.fd_refcnt = 1;
337 	fdp->fd_fd.fd_cmask = cmask;
338 	fdp->fd_fd.fd_ofiles = fdp->fd_dfiles;
339 	fdp->fd_fd.fd_ofileflags = fdp->fd_dfileflags;
340 	fdp->fd_fd.fd_nfiles = NDFILE;
341 
342 	/* Create the limits structures. */
343 	p->p_limit = &limit0;
344 	for (i = 0; i < sizeof(p->p_rlimit)/sizeof(p->p_rlimit[0]); i++)
345 		limit0.pl_rlimit[i].rlim_cur =
346 		    limit0.pl_rlimit[i].rlim_max = RLIM_INFINITY;
347 	limit0.pl_rlimit[RLIMIT_NOFILE].rlim_cur =
348 	    limit0.pl_rlimit[RLIMIT_NOFILE].rlim_max = maxfiles;
349 	limit0.pl_rlimit[RLIMIT_NPROC].rlim_cur =
350 	    limit0.pl_rlimit[RLIMIT_NPROC].rlim_max = maxproc;
351 	i = ptoa(cnt.v_free_count);
352 	limit0.pl_rlimit[RLIMIT_RSS].rlim_max = i;
353 	limit0.pl_rlimit[RLIMIT_MEMLOCK].rlim_max = i;
354 	limit0.pl_rlimit[RLIMIT_MEMLOCK].rlim_cur = i / 3;
355 	limit0.p_cpulimit = RLIM_INFINITY;
356 	limit0.p_refcnt = 1;
357 
358 	/* Allocate a prototype map so we have something to fork. */
359 	pmap_pinit0(vmspace_pmap(&vmspace0));
360 	p->p_vmspace = &vmspace0;
361 	vmspace0.vm_refcnt = 1;
362 	vm_map_init(&vmspace0.vm_map, round_page(VM_MIN_ADDRESS),
363 	    trunc_page(VM_MAXUSER_ADDRESS));
364 	vmspace0.vm_map.pmap = vmspace_pmap(&vmspace0);
365 	p->p_addr = proc0paddr;				/* XXX */
366 
367 	/*
368 	 * We continue to place resource usage info and signal
369 	 * actions in the user struct so they're pageable.
370 	 */
371 	p->p_stats = &p->p_addr->u_stats;
372 	p->p_sigacts = &p->p_addr->u_sigacts;
373 
374 	/*
375 	 * Charge root for one process.
376 	 */
377 	(void)chgproccnt(cred0.p_uidinfo, 1, 0);
378 
379 	LIST_INIT(&p->p_heldmtx);
380 	LIST_INIT(&p->p_contested);
381 
382 	/*
383 	 * Initialize the current process pointer (curproc) before
384 	 * any possible traps/probes to simplify trap processing.
385 	 */
386 	PCPU_SET(curproc, p);
387 
388 	/*
389 	 * Enter the Giant mutex.
390 	 * XXX This should be done BEFORE cpu_startup().
391 	 */
392 	mtx_enter(&Giant, MTX_DEF);
393 
394 }
395 SYSINIT(p0init, SI_SUB_INTRINSIC, SI_ORDER_FIRST, proc0_init, NULL)
396 
397 /* ARGSUSED*/
398 static void
399 proc0_post(void *dummy __unused)
400 {
401 	struct timespec ts;
402 	struct proc *p;
403 
404 	/*
405 	 * Now we can look at the time, having had a chance to verify the
406 	 * time from the file system.  Pretend that proc0 started now.
407 	 */
408 	lockmgr(&allproc_lock, LK_SHARED, NULL, CURPROC);
409 	LIST_FOREACH(p, &allproc, p_list) {
410 		microtime(&p->p_stats->p_start);
411 		p->p_runtime = 0;
412 	}
413 	lockmgr(&allproc_lock, LK_RELEASE, NULL, CURPROC);
414 	microuptime(&switchtime);
415 	PCPU_SET(switchticks, ticks);
416 
417 	/*
418 	 * Give the ``random'' number generator a thump.
419 	 */
420 	nanotime(&ts);
421 	srandom(ts.tv_sec ^ ts.tv_nsec);
422 }
423 SYSINIT(p0post, SI_SUB_INTRINSIC_POST, SI_ORDER_FIRST, proc0_post, NULL)
424 
425 /*
426  ***************************************************************************
427  ****
428  **** The following SYSINIT's and glue code should be moved to the
429  **** respective files on a per subsystem basis.
430  ****
431  ***************************************************************************
432  */
433 
434 
435 /*
436  ***************************************************************************
437  ****
438  **** The following code probably belongs in another file, like
439  **** kern/init_init.c.
440  ****
441  ***************************************************************************
442  */
443 
444 /*
445  * List of paths to try when searching for "init".
446  */
447 static char init_path[MAXPATHLEN] =
448 #ifdef	INIT_PATH
449     __XSTRING(INIT_PATH);
450 #else
451     "/sbin/init:/sbin/oinit:/sbin/init.bak:/stand/sysinstall";
452 #endif
453 SYSCTL_STRING(_kern, OID_AUTO, init_path, CTLFLAG_RD, init_path, 0, "");
454 
455 /*
456  * Start the initial user process; try exec'ing each pathname in init_path.
457  * The program is invoked with one argument containing the boot flags.
458  */
459 static void
460 start_init(void *dummy)
461 {
462 	vm_offset_t addr;
463 	struct execve_args args;
464 	int options, error;
465 	char *var, *path, *next, *s;
466 	char *ucp, **uap, *arg0, *arg1;
467 	struct proc *p;
468 
469 	mtx_enter(&Giant, MTX_DEF);
470 
471 	p = curproc;
472 
473 	/* Get the vnode for '/'.  Set p->p_fd->fd_cdir to reference it. */
474 	if (VFS_ROOT(TAILQ_FIRST(&mountlist), &rootvnode))
475 		panic("cannot find root vnode");
476 	p->p_fd->fd_cdir = rootvnode;
477 	VREF(p->p_fd->fd_cdir);
478 	p->p_fd->fd_rdir = rootvnode;
479 	VOP_UNLOCK(rootvnode, 0, p);
480 
481 	/*
482 	 * Need just enough stack to hold the faked-up "execve()" arguments.
483 	 */
484 	addr = trunc_page(USRSTACK - PAGE_SIZE);
485 	if (vm_map_find(&p->p_vmspace->vm_map, NULL, 0, &addr, PAGE_SIZE,
486 			FALSE, VM_PROT_ALL, VM_PROT_ALL, 0) != 0)
487 		panic("init: couldn't allocate argument space");
488 	p->p_vmspace->vm_maxsaddr = (caddr_t)addr;
489 	p->p_vmspace->vm_ssize = 1;
490 
491 	if ((var = getenv("init_path")) != NULL) {
492 		strncpy(init_path, var, sizeof init_path);
493 		init_path[sizeof init_path - 1] = 0;
494 	}
495 
496 	for (path = init_path; *path != '\0'; path = next) {
497 		while (*path == ':')
498 			path++;
499 		if (*path == '\0')
500 			break;
501 		for (next = path; *next != '\0' && *next != ':'; next++)
502 			/* nothing */ ;
503 		if (bootverbose)
504 			printf("start_init: trying %.*s\n", (int)(next - path),
505 			    path);
506 
507 		/*
508 		 * Move out the boot flag argument.
509 		 */
510 		options = 0;
511 		ucp = (char *)USRSTACK;
512 		(void)subyte(--ucp, 0);		/* trailing zero */
513 		if (boothowto & RB_SINGLE) {
514 			(void)subyte(--ucp, 's');
515 			options = 1;
516 		}
517 #ifdef notyet
518                 if (boothowto & RB_FASTBOOT) {
519 			(void)subyte(--ucp, 'f');
520 			options = 1;
521 		}
522 #endif
523 
524 #ifdef BOOTCDROM
525 		(void)subyte(--ucp, 'C');
526 		options = 1;
527 #endif
528 		if (devfs_present) {
529 			(void)subyte(--ucp, 'd');
530 			options = 1;
531 		}
532 
533 		if (options == 0)
534 			(void)subyte(--ucp, '-');
535 		(void)subyte(--ucp, '-');		/* leading hyphen */
536 		arg1 = ucp;
537 
538 		/*
539 		 * Move out the file name (also arg 0).
540 		 */
541 		(void)subyte(--ucp, 0);
542 		for (s = next - 1; s >= path; s--)
543 			(void)subyte(--ucp, *s);
544 		arg0 = ucp;
545 
546 		/*
547 		 * Move out the arg pointers.
548 		 */
549 		uap = (char **)((intptr_t)ucp & ~(sizeof(intptr_t)-1));
550 		(void)suword((caddr_t)--uap, (long)0);	/* terminator */
551 		(void)suword((caddr_t)--uap, (long)(intptr_t)arg1);
552 		(void)suword((caddr_t)--uap, (long)(intptr_t)arg0);
553 
554 		/*
555 		 * Point at the arguments.
556 		 */
557 		args.fname = arg0;
558 		args.argv = uap;
559 		args.envv = NULL;
560 
561 		/*
562 		 * Now try to exec the program.  If can't for any reason
563 		 * other than it doesn't exist, complain.
564 		 *
565 		 * Otherwise, return via fork_trampoline() all the way
566 		 * to user mode as init!
567 		 */
568 		if ((error = execve(p, &args)) == 0) {
569 			mtx_exit(&Giant, MTX_DEF);
570 			return;
571 		}
572 		if (error != ENOENT)
573 			printf("exec %.*s: error %d\n", (int)(next - path),
574 			    path, error);
575 	}
576 	printf("init: not found in path %s\n", init_path);
577 	panic("no init");
578 }
579 
580 /*
581  * Like kthread_create(), but runs in it's own address space.
582  * We do this early to reserve pid 1.
583  *
584  * Note special case - do not make it runnable yet.  Other work
585  * in progress will change this more.
586  */
587 static void
588 create_init(const void *udata __unused)
589 {
590 	int error;
591 
592 	error = fork1(&proc0, RFFDG | RFPROC | RFSTOPPED, &initproc);
593 	if (error)
594 		panic("cannot fork init: %d\n", error);
595 	initproc->p_flag |= P_INMEM | P_SYSTEM;
596 	cpu_set_fork_handler(initproc, start_init, NULL);
597 }
598 SYSINIT(init,SI_SUB_CREATE_INIT, SI_ORDER_FIRST, create_init, NULL)
599 
600 /*
601  * Make it runnable now.
602  */
603 static void
604 kick_init(const void *udata __unused)
605 {
606 	mtx_enter(&sched_lock, MTX_SPIN);
607 	initproc->p_stat = SRUN;
608 	setrunqueue(initproc);
609 	mtx_exit(&sched_lock, MTX_SPIN);
610 }
611 SYSINIT(kickinit,SI_SUB_KTHREAD_INIT, SI_ORDER_FIRST, kick_init, NULL)
612